diff --git "a/checkpoint-70000/trainer_state.json" "b/checkpoint-70000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-70000/trainer_state.json"
@@ -0,0 +1,10463 @@
+{
+  "best_metric": 3.3429155349731445,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M__1208/checkpoint-70000",
+  "epoch": 7.547169811320755,
+  "eval_steps": 1000,
+  "global_step": 70000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005390835579514825,
+      "grad_norm": 1.3872060775756836,
+      "learning_rate": 0.0003,
+      "loss": 8.7938,
+      "step": 50
+    },
+    {
+      "epoch": 0.01078167115902965,
+      "grad_norm": 3.891347885131836,
+      "learning_rate": 0.0006,
+      "loss": 6.992,
+      "step": 100
+    },
+    {
+      "epoch": 0.016172506738544475,
+      "grad_norm": 1.848716378211975,
+      "learning_rate": 0.0005996762007555315,
+      "loss": 6.5045,
+      "step": 150
+    },
+    {
+      "epoch": 0.0215633423180593,
+      "grad_norm": 1.3053127527236938,
+      "learning_rate": 0.000599352401511063,
+      "loss": 6.2433,
+      "step": 200
+    },
+    {
+      "epoch": 0.026954177897574125,
+      "grad_norm": 1.2040534019470215,
+      "learning_rate": 0.0005990286022665946,
+      "loss": 6.0719,
+      "step": 250
+    },
+    {
+      "epoch": 0.03234501347708895,
+      "grad_norm": 1.105197548866272,
+      "learning_rate": 0.0005987048030221263,
+      "loss": 5.9606,
+      "step": 300
+    },
+    {
+      "epoch": 0.03773584905660377,
+      "grad_norm": 1.502025842666626,
+      "learning_rate": 0.0005983810037776578,
+      "loss": 5.8818,
+      "step": 350
+    },
+    {
+      "epoch": 0.0431266846361186,
+      "grad_norm": 2.493607759475708,
+      "learning_rate": 0.0005980572045331894,
+      "loss": 5.8101,
+      "step": 400
+    },
+    {
+      "epoch": 0.04851752021563342,
+      "grad_norm": 1.3217182159423828,
+      "learning_rate": 0.0005977334052887209,
+      "loss": 5.7175,
+      "step": 450
+    },
+    {
+      "epoch": 0.05390835579514825,
+      "grad_norm": 1.215240240097046,
+      "learning_rate": 0.0005974096060442526,
+      "loss": 5.6639,
+      "step": 500
+    },
+    {
+      "epoch": 0.05929919137466307,
+      "grad_norm": 1.3821635246276855,
+      "learning_rate": 0.0005970858067997841,
+      "loss": 5.5744,
+      "step": 550
+    },
+    {
+      "epoch": 0.0646900269541779,
+      "grad_norm": 1.378275752067566,
+      "learning_rate": 0.0005967620075553157,
+      "loss": 5.5287,
+      "step": 600
+    },
+    {
+      "epoch": 0.07008086253369272,
+      "grad_norm": 1.1444066762924194,
+      "learning_rate": 0.0005964382083108472,
+      "loss": 5.4332,
+      "step": 650
+    },
+    {
+      "epoch": 0.07547169811320754,
+      "grad_norm": 1.657812476158142,
+      "learning_rate": 0.0005961144090663788,
+      "loss": 5.4098,
+      "step": 700
+    },
+    {
+      "epoch": 0.08086253369272237,
+      "grad_norm": 1.3864067792892456,
+      "learning_rate": 0.0005957906098219104,
+      "loss": 5.3168,
+      "step": 750
+    },
+    {
+      "epoch": 0.0862533692722372,
+      "grad_norm": 1.0762931108474731,
+      "learning_rate": 0.0005954668105774419,
+      "loss": 5.2678,
+      "step": 800
+    },
+    {
+      "epoch": 0.09164420485175202,
+      "grad_norm": 1.2499072551727295,
+      "learning_rate": 0.0005951430113329735,
+      "loss": 5.2107,
+      "step": 850
+    },
+    {
+      "epoch": 0.09703504043126684,
+      "grad_norm": 0.9282752275466919,
+      "learning_rate": 0.0005948192120885051,
+      "loss": 5.1849,
+      "step": 900
+    },
+    {
+      "epoch": 0.10242587601078167,
+      "grad_norm": 1.071282982826233,
+      "learning_rate": 0.0005944954128440366,
+      "loss": 5.1275,
+      "step": 950
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "grad_norm": 1.3316810131072998,
+      "learning_rate": 0.0005941716135995682,
+      "loss": 5.1151,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "eval_accuracy": 0.22672739349001877,
+      "eval_loss": 5.027144432067871,
+      "eval_runtime": 184.654,
+      "eval_samples_per_second": 97.539,
+      "eval_steps_per_second": 6.098,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11320754716981132,
+      "grad_norm": 1.1736186742782593,
+      "learning_rate": 0.0005938478143550997,
+      "loss": 5.0742,
+      "step": 1050
+    },
+    {
+      "epoch": 0.11859838274932614,
+      "grad_norm": 1.217553973197937,
+      "learning_rate": 0.0005935240151106314,
+      "loss": 5.0002,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12398921832884097,
+      "grad_norm": 1.3116732835769653,
+      "learning_rate": 0.0005932002158661629,
+      "loss": 4.9959,
+      "step": 1150
+    },
+    {
+      "epoch": 0.1293800539083558,
+      "grad_norm": 1.10372793674469,
+      "learning_rate": 0.0005928764166216945,
+      "loss": 4.9613,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1347708894878706,
+      "grad_norm": 1.0251359939575195,
+      "learning_rate": 0.000592552617377226,
+      "loss": 4.9081,
+      "step": 1250
+    },
+    {
+      "epoch": 0.14016172506738545,
+      "grad_norm": 1.1212270259857178,
+      "learning_rate": 0.0005922288181327577,
+      "loss": 4.9098,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14555256064690028,
+      "grad_norm": 1.0358482599258423,
+      "learning_rate": 0.0005919050188882893,
+      "loss": 4.8835,
+      "step": 1350
+    },
+    {
+      "epoch": 0.1509433962264151,
+      "grad_norm": 1.127200722694397,
+      "learning_rate": 0.0005915812196438207,
+      "loss": 4.85,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15633423180592992,
+      "grad_norm": 0.9541453123092651,
+      "learning_rate": 0.0005912574203993524,
+      "loss": 4.8199,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16172506738544473,
+      "grad_norm": 0.9905341863632202,
+      "learning_rate": 0.0005909336211548839,
+      "loss": 4.812,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16711590296495957,
+      "grad_norm": 0.9883430600166321,
+      "learning_rate": 0.0005906098219104155,
+      "loss": 4.7776,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1725067385444744,
+      "grad_norm": 1.14104425907135,
+      "learning_rate": 0.000590286022665947,
+      "loss": 4.7877,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1778975741239892,
+      "grad_norm": 0.915554404258728,
+      "learning_rate": 0.0005899622234214787,
+      "loss": 4.7362,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18328840970350405,
+      "grad_norm": 0.8896968364715576,
+      "learning_rate": 0.0005896384241770102,
+      "loss": 4.6944,
+      "step": 1700
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 0.7295605540275574,
+      "learning_rate": 0.0005893146249325418,
+      "loss": 4.679,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1940700808625337,
+      "grad_norm": 0.8035542368888855,
+      "learning_rate": 0.0005889908256880733,
+      "loss": 4.669,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19946091644204852,
+      "grad_norm": 0.9095346927642822,
+      "learning_rate": 0.0005886670264436049,
+      "loss": 4.6768,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20485175202156333,
+      "grad_norm": 0.8672581315040588,
+      "learning_rate": 0.0005883432271991365,
+      "loss": 4.6219,
+      "step": 1900
+    },
+    {
+      "epoch": 0.21024258760107817,
+      "grad_norm": 0.7349112033843994,
+      "learning_rate": 0.0005880194279546681,
+      "loss": 4.6077,
+      "step": 1950
+    },
+    {
+      "epoch": 0.215633423180593,
+      "grad_norm": 0.9716205596923828,
+      "learning_rate": 0.0005876956287101996,
+      "loss": 4.5792,
+      "step": 2000
+    },
+    {
+      "epoch": 0.215633423180593,
+      "eval_accuracy": 0.27099464439050674,
+      "eval_loss": 4.504695415496826,
+      "eval_runtime": 184.1142,
+      "eval_samples_per_second": 97.825,
+      "eval_steps_per_second": 6.116,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2210242587601078,
+      "grad_norm": 1.203096628189087,
+      "learning_rate": 0.0005873718294657312,
+      "loss": 4.5695,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22641509433962265,
+      "grad_norm": 0.9952515959739685,
+      "learning_rate": 0.0005870480302212628,
+      "loss": 4.5412,
+      "step": 2100
+    },
+    {
+      "epoch": 0.23180592991913745,
+      "grad_norm": 0.8814882040023804,
+      "learning_rate": 0.0005867242309767943,
+      "loss": 4.5392,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2371967654986523,
+      "grad_norm": 1.0191090106964111,
+      "learning_rate": 0.0005864004317323259,
+      "loss": 4.5151,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24258760107816713,
+      "grad_norm": 0.8580870628356934,
+      "learning_rate": 0.0005860766324878575,
+      "loss": 4.5014,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24797843665768193,
+      "grad_norm": 0.8280946016311646,
+      "learning_rate": 0.000585752833243389,
+      "loss": 4.4755,
+      "step": 2300
+    },
+    {
+      "epoch": 0.25336927223719674,
+      "grad_norm": 0.9512032866477966,
+      "learning_rate": 0.0005854290339989206,
+      "loss": 4.4624,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2587601078167116,
+      "grad_norm": 1.000420093536377,
+      "learning_rate": 0.0005851052347544521,
+      "loss": 4.4523,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2641509433962264,
+      "grad_norm": 0.9130716919898987,
+      "learning_rate": 0.0005847814355099838,
+      "loss": 4.4414,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2695417789757412,
+      "grad_norm": 0.8131120800971985,
+      "learning_rate": 0.0005844576362655154,
+      "loss": 4.4556,
+      "step": 2500
+    },
+    {
+      "epoch": 0.2749326145552561,
+      "grad_norm": 0.9587961435317993,
+      "learning_rate": 0.0005841338370210469,
+      "loss": 4.4083,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2803234501347709,
+      "grad_norm": 0.846928060054779,
+      "learning_rate": 0.0005838100377765785,
+      "loss": 4.3825,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 0.9020141363143921,
+      "learning_rate": 0.0005834862385321101,
+      "loss": 4.3869,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29110512129380056,
+      "grad_norm": 1.0460323095321655,
+      "learning_rate": 0.0005831624392876417,
+      "loss": 4.381,
+      "step": 2700
+    },
+    {
+      "epoch": 0.29649595687331537,
+      "grad_norm": 0.8714868426322937,
+      "learning_rate": 0.0005828386400431731,
+      "loss": 4.3618,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3018867924528302,
+      "grad_norm": 0.7531731724739075,
+      "learning_rate": 0.0005825148407987048,
+      "loss": 4.361,
+      "step": 2800
+    },
+    {
+      "epoch": 0.30727762803234504,
+      "grad_norm": 0.7624960541725159,
+      "learning_rate": 0.0005821910415542363,
+      "loss": 4.3642,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31266846361185985,
+      "grad_norm": 0.8426908850669861,
+      "learning_rate": 0.0005818672423097679,
+      "loss": 4.3367,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31805929919137466,
+      "grad_norm": 0.9197669625282288,
+      "learning_rate": 0.0005815434430652994,
+      "loss": 4.3496,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "grad_norm": 0.8955463171005249,
+      "learning_rate": 0.0005812196438208311,
+      "loss": 4.2988,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "eval_accuracy": 0.2984499253065723,
+      "eval_loss": 4.238968372344971,
+      "eval_runtime": 185.178,
+      "eval_samples_per_second": 97.263,
+      "eval_steps_per_second": 6.081,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3288409703504043,
+      "grad_norm": 0.7810043096542358,
+      "learning_rate": 0.0005808958445763626,
+      "loss": 4.3082,
+      "step": 3050
+    },
+    {
+      "epoch": 0.33423180592991913,
+      "grad_norm": 0.7354643940925598,
+      "learning_rate": 0.0005805720453318942,
+      "loss": 4.2858,
+      "step": 3100
+    },
+    {
+      "epoch": 0.33962264150943394,
+      "grad_norm": 0.7374374270439148,
+      "learning_rate": 0.0005802482460874257,
+      "loss": 4.2772,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3450134770889488,
+      "grad_norm": 0.7399438619613647,
+      "learning_rate": 0.0005799244468429573,
+      "loss": 4.2993,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3504043126684636,
+      "grad_norm": 0.7688404321670532,
+      "learning_rate": 0.0005796006475984889,
+      "loss": 4.2841,
+      "step": 3250
+    },
+    {
+      "epoch": 0.3557951482479784,
+      "grad_norm": 0.7221906781196594,
+      "learning_rate": 0.0005792768483540205,
+      "loss": 4.2667,
+      "step": 3300
+    },
+    {
+      "epoch": 0.3611859838274933,
+      "grad_norm": 0.7419180274009705,
+      "learning_rate": 0.000578953049109552,
+      "loss": 4.2709,
+      "step": 3350
+    },
+    {
+      "epoch": 0.3665768194070081,
+      "grad_norm": 0.7906516790390015,
+      "learning_rate": 0.0005786292498650836,
+      "loss": 4.2394,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3719676549865229,
+      "grad_norm": 0.7784256339073181,
+      "learning_rate": 0.0005783054506206152,
+      "loss": 4.2519,
+      "step": 3450
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.7829201221466064,
+      "learning_rate": 0.0005779816513761467,
+      "loss": 4.2314,
+      "step": 3500
+    },
+    {
+      "epoch": 0.38274932614555257,
+      "grad_norm": 0.8310431241989136,
+      "learning_rate": 0.0005776578521316782,
+      "loss": 4.2097,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3881401617250674,
+      "grad_norm": 0.8430302739143372,
+      "learning_rate": 0.0005773340528872099,
+      "loss": 4.2306,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3935309973045822,
+      "grad_norm": 0.733945906162262,
+      "learning_rate": 0.0005770102536427414,
+      "loss": 4.2221,
+      "step": 3650
+    },
+    {
+      "epoch": 0.39892183288409705,
+      "grad_norm": 0.6383342742919922,
+      "learning_rate": 0.000576686454398273,
+      "loss": 4.2028,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40431266846361186,
+      "grad_norm": 0.6672252416610718,
+      "learning_rate": 0.0005763626551538045,
+      "loss": 4.1801,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40970350404312667,
+      "grad_norm": 0.7545523643493652,
+      "learning_rate": 0.0005760388559093362,
+      "loss": 4.1881,
+      "step": 3800
+    },
+    {
+      "epoch": 0.41509433962264153,
+      "grad_norm": 0.7340240478515625,
+      "learning_rate": 0.0005757150566648678,
+      "loss": 4.171,
+      "step": 3850
+    },
+    {
+      "epoch": 0.42048517520215634,
+      "grad_norm": 0.7903727889060974,
+      "learning_rate": 0.0005753912574203993,
+      "loss": 4.1784,
+      "step": 3900
+    },
+    {
+      "epoch": 0.42587601078167114,
+      "grad_norm": 0.8018081188201904,
+      "learning_rate": 0.0005750674581759309,
+      "loss": 4.1838,
+      "step": 3950
+    },
+    {
+      "epoch": 0.431266846361186,
+      "grad_norm": 0.722425103187561,
+      "learning_rate": 0.0005747436589314624,
+      "loss": 4.1617,
+      "step": 4000
+    },
+    {
+      "epoch": 0.431266846361186,
+      "eval_accuracy": 0.31149631215800827,
+      "eval_loss": 4.093442440032959,
+      "eval_runtime": 185.3481,
+      "eval_samples_per_second": 97.174,
+      "eval_steps_per_second": 6.075,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4366576819407008,
+      "grad_norm": 0.9414203763008118,
+      "learning_rate": 0.0005744198596869941,
+      "loss": 4.15,
+      "step": 4050
+    },
+    {
+      "epoch": 0.4420485175202156,
+      "grad_norm": 0.7414451241493225,
+      "learning_rate": 0.0005740960604425255,
+      "loss": 4.1367,
+      "step": 4100
+    },
+    {
+      "epoch": 0.4474393530997305,
+      "grad_norm": 0.693946361541748,
+      "learning_rate": 0.0005737722611980572,
+      "loss": 4.1503,
+      "step": 4150
+    },
+    {
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.8669963479042053,
+      "learning_rate": 0.0005734484619535887,
+      "loss": 4.1342,
+      "step": 4200
+    },
+    {
+      "epoch": 0.4582210242587601,
+      "grad_norm": 0.717522144317627,
+      "learning_rate": 0.0005731246627091203,
+      "loss": 4.1361,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4636118598382749,
+      "grad_norm": 0.5836500525474548,
+      "learning_rate": 0.0005728008634646518,
+      "loss": 4.1184,
+      "step": 4300
+    },
+    {
+      "epoch": 0.46900269541778977,
+      "grad_norm": 0.6450395584106445,
+      "learning_rate": 0.0005724770642201835,
+      "loss": 4.1149,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4743935309973046,
+      "grad_norm": 0.6113871335983276,
+      "learning_rate": 0.000572153264975715,
+      "loss": 4.1102,
+      "step": 4400
+    },
+    {
+      "epoch": 0.4797843665768194,
+      "grad_norm": 0.7142659425735474,
+      "learning_rate": 0.0005718294657312466,
+      "loss": 4.1382,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48517520215633425,
+      "grad_norm": 0.6929596066474915,
+      "learning_rate": 0.0005715056664867781,
+      "loss": 4.0957,
+      "step": 4500
+    },
+    {
+      "epoch": 0.49056603773584906,
+      "grad_norm": 0.5346873998641968,
+      "learning_rate": 0.0005711818672423097,
+      "loss": 4.1046,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49595687331536387,
+      "grad_norm": 0.702147901058197,
+      "learning_rate": 0.0005708580679978413,
+      "loss": 4.1,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5013477088948787,
+      "grad_norm": 0.5867359042167664,
+      "learning_rate": 0.0005705342687533729,
+      "loss": 4.0953,
+      "step": 4650
+    },
+    {
+      "epoch": 0.5067385444743935,
+      "grad_norm": 0.6331294178962708,
+      "learning_rate": 0.0005702104695089044,
+      "loss": 4.0979,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5121293800539084,
+      "grad_norm": 0.6199395656585693,
+      "learning_rate": 0.000569886670264436,
+      "loss": 4.0987,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5175202156334232,
+      "grad_norm": 0.6783130764961243,
+      "learning_rate": 0.0005695628710199675,
+      "loss": 4.0776,
+      "step": 4800
+    },
+    {
+      "epoch": 0.522911051212938,
+      "grad_norm": 0.6518664956092834,
+      "learning_rate": 0.0005692390717754991,
+      "loss": 4.1048,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5283018867924528,
+      "grad_norm": 0.6022597551345825,
+      "learning_rate": 0.0005689152725310306,
+      "loss": 4.0778,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5336927223719676,
+      "grad_norm": 0.6618348360061646,
+      "learning_rate": 0.0005685914732865623,
+      "loss": 4.0617,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "grad_norm": 0.5986215472221375,
+      "learning_rate": 0.0005682676740420939,
+      "loss": 4.0655,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "eval_accuracy": 0.32159864036122304,
+      "eval_loss": 3.9903128147125244,
+      "eval_runtime": 185.2705,
+      "eval_samples_per_second": 97.215,
+      "eval_steps_per_second": 6.078,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5444743935309974,
+      "grad_norm": 0.5965196490287781,
+      "learning_rate": 0.0005679438747976254,
+      "loss": 4.0629,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5498652291105122,
+      "grad_norm": 0.618198812007904,
+      "learning_rate": 0.000567620075553157,
+      "loss": 4.048,
+      "step": 5100
+    },
+    {
+      "epoch": 0.555256064690027,
+      "grad_norm": 0.6078091263771057,
+      "learning_rate": 0.0005672962763086886,
+      "loss": 4.0601,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5606469002695418,
+      "grad_norm": 0.5713509917259216,
+      "learning_rate": 0.0005669724770642202,
+      "loss": 4.054,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.7706556916236877,
+      "learning_rate": 0.0005666486778197517,
+      "loss": 4.0499,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 0.6719933748245239,
+      "learning_rate": 0.0005663248785752833,
+      "loss": 4.0337,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5768194070080862,
+      "grad_norm": 0.6685216426849365,
+      "learning_rate": 0.0005660010793308148,
+      "loss": 4.0071,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5822102425876011,
+      "grad_norm": 0.6304830312728882,
+      "learning_rate": 0.0005656772800863465,
+      "loss": 4.0476,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5876010781671159,
+      "grad_norm": 0.6301653385162354,
+      "learning_rate": 0.0005653534808418779,
+      "loss": 4.0282,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5929919137466307,
+      "grad_norm": 0.6767484545707703,
+      "learning_rate": 0.0005650296815974096,
+      "loss": 4.0349,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5983827493261455,
+      "grad_norm": 0.6710530519485474,
+      "learning_rate": 0.0005647058823529411,
+      "loss": 4.0179,
+      "step": 5550
+    },
+    {
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.5611121654510498,
+      "learning_rate": 0.0005643820831084727,
+      "loss": 4.0202,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6091644204851752,
+      "grad_norm": 0.6187557578086853,
+      "learning_rate": 0.0005640582838640042,
+      "loss": 3.9935,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6145552560646901,
+      "grad_norm": 0.6690448522567749,
+      "learning_rate": 0.0005637344846195358,
+      "loss": 4.0082,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6199460916442049,
+      "grad_norm": 0.6771255135536194,
+      "learning_rate": 0.0005634106853750674,
+      "loss": 4.0026,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6253369272237197,
+      "grad_norm": 0.5823601484298706,
+      "learning_rate": 0.000563086886130599,
+      "loss": 4.014,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6307277628032345,
+      "grad_norm": 0.6157869100570679,
+      "learning_rate": 0.0005627630868861305,
+      "loss": 3.9952,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6361185983827493,
+      "grad_norm": 0.6894795894622803,
+      "learning_rate": 0.0005624392876416621,
+      "loss": 3.9947,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6415094339622641,
+      "grad_norm": 0.6236327290534973,
+      "learning_rate": 0.0005621154883971937,
+      "loss": 3.996,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "grad_norm": 0.6344878077507019,
+      "learning_rate": 0.0005617916891527253,
+      "loss": 3.9934,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "eval_accuracy": 0.3280596843351111,
+      "eval_loss": 3.924257278442383,
+      "eval_runtime": 185.4361,
+      "eval_samples_per_second": 97.128,
+      "eval_steps_per_second": 6.072,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6522911051212938,
+      "grad_norm": 0.5802189111709595,
+      "learning_rate": 0.0005614743658931463,
+      "loss": 3.9855,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6576819407008087,
+      "grad_norm": 0.6164999604225159,
+      "learning_rate": 0.0005611505666486777,
+      "loss": 4.0042,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6630727762803235,
+      "grad_norm": 0.6946608424186707,
+      "learning_rate": 0.0005608267674042094,
+      "loss": 3.989,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6684636118598383,
+      "grad_norm": 0.6158466935157776,
+      "learning_rate": 0.0005605029681597409,
+      "loss": 3.9871,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6738544474393531,
+      "grad_norm": 0.6288596391677856,
+      "learning_rate": 0.0005601791689152725,
+      "loss": 3.9773,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6792452830188679,
+      "grad_norm": 0.7042039036750793,
+      "learning_rate": 0.000559855369670804,
+      "loss": 3.9791,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6846361185983828,
+      "grad_norm": 0.5582059025764465,
+      "learning_rate": 0.0005595315704263357,
+      "loss": 3.9639,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6900269541778976,
+      "grad_norm": 0.6060189604759216,
+      "learning_rate": 0.0005592077711818672,
+      "loss": 3.9775,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6954177897574124,
+      "grad_norm": 0.7476151585578918,
+      "learning_rate": 0.0005588839719373988,
+      "loss": 3.9536,
+      "step": 6450
+    },
+    {
+      "epoch": 0.7008086253369272,
+      "grad_norm": 0.6221930384635925,
+      "learning_rate": 0.0005585601726929303,
+      "loss": 3.9583,
+      "step": 6500
+    },
+    {
+      "epoch": 0.706199460916442,
+      "grad_norm": 0.6130439639091492,
+      "learning_rate": 0.0005582363734484619,
+      "loss": 3.9417,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7115902964959568,
+      "grad_norm": 0.5701190233230591,
+      "learning_rate": 0.0005579125742039935,
+      "loss": 3.9488,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7169811320754716,
+      "grad_norm": 0.5955173969268799,
+      "learning_rate": 0.0005575887749595251,
+      "loss": 3.9511,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7223719676549866,
+      "grad_norm": 0.6578053832054138,
+      "learning_rate": 0.0005572649757150566,
+      "loss": 3.9526,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7277628032345014,
+      "grad_norm": 0.7318617105484009,
+      "learning_rate": 0.0005569411764705882,
+      "loss": 3.9511,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7331536388140162,
+      "grad_norm": 0.6341660618782043,
+      "learning_rate": 0.0005566173772261198,
+      "loss": 3.918,
+      "step": 6800
+    },
+    {
+      "epoch": 0.738544474393531,
+      "grad_norm": 0.5913158059120178,
+      "learning_rate": 0.0005562935779816513,
+      "loss": 3.9336,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7439353099730458,
+      "grad_norm": 0.6217501759529114,
+      "learning_rate": 0.0005559697787371828,
+      "loss": 3.9485,
+      "step": 6900
+    },
+    {
+      "epoch": 0.7493261455525606,
+      "grad_norm": 0.629990816116333,
+      "learning_rate": 0.0005556459794927145,
+      "loss": 3.9304,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.598209798336029,
+      "learning_rate": 0.000555322180248246,
+      "loss": 3.9432,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "eval_accuracy": 0.33339541034990466,
+      "eval_loss": 3.867119312286377,
+      "eval_runtime": 185.4718,
+      "eval_samples_per_second": 97.109,
+      "eval_steps_per_second": 6.071,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7601078167115903,
+      "grad_norm": 0.7586105465888977,
+      "learning_rate": 0.0005549983810037776,
+      "loss": 3.911,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7654986522911051,
+      "grad_norm": 0.6012836694717407,
+      "learning_rate": 0.0005546745817593091,
+      "loss": 3.9126,
+      "step": 7100
+    },
+    {
+      "epoch": 0.77088948787062,
+      "grad_norm": 0.6788042187690735,
+      "learning_rate": 0.0005543507825148408,
+      "loss": 3.9149,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7762803234501348,
+      "grad_norm": 0.60019451379776,
+      "learning_rate": 0.0005540269832703723,
+      "loss": 3.8948,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7816711590296496,
+      "grad_norm": 0.5468128323554993,
+      "learning_rate": 0.0005537031840259039,
+      "loss": 3.9228,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7870619946091644,
+      "grad_norm": 0.560189425945282,
+      "learning_rate": 0.0005533793847814354,
+      "loss": 3.8974,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7924528301886793,
+      "grad_norm": 0.6238011717796326,
+      "learning_rate": 0.000553055585536967,
+      "loss": 3.9169,
+      "step": 7350
+    },
+    {
+      "epoch": 0.7978436657681941,
+      "grad_norm": 0.6309143304824829,
+      "learning_rate": 0.0005527317862924987,
+      "loss": 3.9167,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8032345013477089,
+      "grad_norm": 0.5579202771186829,
+      "learning_rate": 0.0005524079870480301,
+      "loss": 3.9102,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8086253369272237,
+      "grad_norm": 0.637144923210144,
+      "learning_rate": 0.0005520841878035618,
+      "loss": 3.9305,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8140161725067385,
+      "grad_norm": 0.560736358165741,
+      "learning_rate": 0.0005517603885590933,
+      "loss": 3.9012,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8194070080862533,
+      "grad_norm": 0.5624358057975769,
+      "learning_rate": 0.0005514365893146249,
+      "loss": 3.8973,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8247978436657682,
+      "grad_norm": 0.5344393253326416,
+      "learning_rate": 0.0005511127900701564,
+      "loss": 3.8894,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8301886792452831,
+      "grad_norm": 0.5487794280052185,
+      "learning_rate": 0.000550788990825688,
+      "loss": 3.9173,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8355795148247979,
+      "grad_norm": 0.5416852831840515,
+      "learning_rate": 0.0005504651915812196,
+      "loss": 3.9131,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8409703504043127,
+      "grad_norm": 0.6219443678855896,
+      "learning_rate": 0.0005501413923367512,
+      "loss": 3.8783,
+      "step": 7800
+    },
+    {
+      "epoch": 0.8463611859838275,
+      "grad_norm": 0.5484825968742371,
+      "learning_rate": 0.0005498175930922827,
+      "loss": 3.8889,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8517520215633423,
+      "grad_norm": 0.5506182909011841,
+      "learning_rate": 0.0005494937938478143,
+      "loss": 3.8945,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.5658506155014038,
+      "learning_rate": 0.0005491699946033459,
+      "loss": 3.883,
+      "step": 7950
+    },
+    {
+      "epoch": 0.862533692722372,
+      "grad_norm": 0.5695346593856812,
+      "learning_rate": 0.0005488461953588775,
+      "loss": 3.8869,
+      "step": 8000
+    },
+    {
+      "epoch": 0.862533692722372,
+      "eval_accuracy": 0.33750498852571176,
+      "eval_loss": 3.818067789077759,
+      "eval_runtime": 185.1439,
+      "eval_samples_per_second": 97.281,
+      "eval_steps_per_second": 6.082,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8679245283018868,
+      "grad_norm": 0.5362111926078796,
+      "learning_rate": 0.000548522396114409,
+      "loss": 3.8853,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8733153638814016,
+      "grad_norm": 0.6255955696105957,
+      "learning_rate": 0.0005482050728548299,
+      "loss": 3.8955,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8787061994609164,
+      "grad_norm": 0.4989166557788849,
+      "learning_rate": 0.0005478812736103615,
+      "loss": 3.8747,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8840970350404312,
+      "grad_norm": 0.5338106751441956,
+      "learning_rate": 0.0005475574743658931,
+      "loss": 3.8772,
+      "step": 8200
+    },
+    {
+      "epoch": 0.889487870619946,
+      "grad_norm": 0.6587180495262146,
+      "learning_rate": 0.0005472336751214246,
+      "loss": 3.8723,
+      "step": 8250
+    },
+    {
+      "epoch": 0.894878706199461,
+      "grad_norm": 0.6544439196586609,
+      "learning_rate": 0.0005469098758769562,
+      "loss": 3.856,
+      "step": 8300
+    },
+    {
+      "epoch": 0.9002695417789758,
+      "grad_norm": 0.5896086692810059,
+      "learning_rate": 0.0005465860766324878,
+      "loss": 3.8439,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.579795241355896,
+      "learning_rate": 0.0005462622773880194,
+      "loss": 3.892,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9110512129380054,
+      "grad_norm": 0.6916389465332031,
+      "learning_rate": 0.000545938478143551,
+      "loss": 3.8569,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9164420485175202,
+      "grad_norm": 0.5650646686553955,
+      "learning_rate": 0.0005456146788990825,
+      "loss": 3.8475,
+      "step": 8500
+    },
+    {
+      "epoch": 0.921832884097035,
+      "grad_norm": 0.5492226481437683,
+      "learning_rate": 0.000545290879654614,
+      "loss": 3.86,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9272237196765498,
+      "grad_norm": 0.6170901656150818,
+      "learning_rate": 0.0005449670804101457,
+      "loss": 3.86,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9326145552560647,
+      "grad_norm": 0.605499804019928,
+      "learning_rate": 0.0005446432811656773,
+      "loss": 3.8616,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9380053908355795,
+      "grad_norm": 0.5699788331985474,
+      "learning_rate": 0.0005443194819212088,
+      "loss": 3.8398,
+      "step": 8700
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.6459490656852722,
+      "learning_rate": 0.0005439956826767404,
+      "loss": 3.8561,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9487870619946092,
+      "grad_norm": 0.5878280401229858,
+      "learning_rate": 0.000543671883432272,
+      "loss": 3.8437,
+      "step": 8800
+    },
+    {
+      "epoch": 0.954177897574124,
+      "grad_norm": 0.5819820165634155,
+      "learning_rate": 0.0005433480841878035,
+      "loss": 3.8392,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9595687331536388,
+      "grad_norm": 0.6361739635467529,
+      "learning_rate": 0.000543024284943335,
+      "loss": 3.8312,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9649595687331537,
+      "grad_norm": 0.6311874389648438,
+      "learning_rate": 0.0005427004856988667,
+      "loss": 3.8481,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "grad_norm": 0.6704027056694031,
+      "learning_rate": 0.0005423766864543982,
+      "loss": 3.848,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "eval_accuracy": 0.3414659295509887,
+      "eval_loss": 3.779182195663452,
+      "eval_runtime": 184.1503,
+      "eval_samples_per_second": 97.806,
+      "eval_steps_per_second": 6.115,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9757412398921833,
+      "grad_norm": 0.5311286449432373,
+      "learning_rate": 0.0005420528872099298,
+      "loss": 3.8298,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9811320754716981,
+      "grad_norm": 0.5362455248832703,
+      "learning_rate": 0.0005417290879654613,
+      "loss": 3.8289,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9865229110512129,
+      "grad_norm": 0.6195308566093445,
+      "learning_rate": 0.000541405288720993,
+      "loss": 3.8465,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9919137466307277,
+      "grad_norm": 0.5365903973579407,
+      "learning_rate": 0.0005410814894765245,
+      "loss": 3.8248,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9973045822102425,
+      "grad_norm": 0.652845025062561,
+      "learning_rate": 0.0005407576902320561,
+      "loss": 3.8518,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0026954177897573,
+      "grad_norm": 0.60700523853302,
+      "learning_rate": 0.0005404338909875876,
+      "loss": 3.7881,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0080862533692723,
+      "grad_norm": 0.6487429738044739,
+      "learning_rate": 0.0005401100917431192,
+      "loss": 3.7647,
+      "step": 9350
+    },
+    {
+      "epoch": 1.013477088948787,
+      "grad_norm": 0.561632513999939,
+      "learning_rate": 0.0005397862924986508,
+      "loss": 3.7791,
+      "step": 9400
+    },
+    {
+      "epoch": 1.0188679245283019,
+      "grad_norm": 0.5459903478622437,
+      "learning_rate": 0.0005394624932541824,
+      "loss": 3.7855,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0242587601078168,
+      "grad_norm": 0.5852130055427551,
+      "learning_rate": 0.0005391386940097139,
+      "loss": 3.7552,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0296495956873315,
+      "grad_norm": 0.6032885909080505,
+      "learning_rate": 0.0005388148947652455,
+      "loss": 3.7671,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0350404312668464,
+      "grad_norm": 0.6321649551391602,
+      "learning_rate": 0.000538491095520777,
+      "loss": 3.795,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0404312668463611,
+      "grad_norm": 0.615997314453125,
+      "learning_rate": 0.0005381672962763086,
+      "loss": 3.7544,
+      "step": 9650
+    },
+    {
+      "epoch": 1.045822102425876,
+      "grad_norm": 0.6234314441680908,
+      "learning_rate": 0.0005378434970318403,
+      "loss": 3.7772,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0512129380053907,
+      "grad_norm": 0.638446569442749,
+      "learning_rate": 0.0005375196977873718,
+      "loss": 3.753,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0566037735849056,
+      "grad_norm": 0.6064075827598572,
+      "learning_rate": 0.0005371958985429034,
+      "loss": 3.765,
+      "step": 9800
+    },
+    {
+      "epoch": 1.0619946091644206,
+      "grad_norm": 0.572862982749939,
+      "learning_rate": 0.0005368720992984349,
+      "loss": 3.7706,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0673854447439353,
+      "grad_norm": 0.6439206600189209,
+      "learning_rate": 0.0005365483000539665,
+      "loss": 3.7649,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0727762803234502,
+      "grad_norm": 0.5527343153953552,
+      "learning_rate": 0.0005362245008094981,
+      "loss": 3.7804,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "grad_norm": 0.5652170777320862,
+      "learning_rate": 0.0005359007015650297,
+      "loss": 3.7701,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "eval_accuracy": 0.3446935722364057,
+      "eval_loss": 3.7489497661590576,
+      "eval_runtime": 184.1084,
+      "eval_samples_per_second": 97.828,
+      "eval_steps_per_second": 6.116,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0835579514824798,
+      "grad_norm": 0.5648168325424194,
+      "learning_rate": 0.0005355769023205612,
+      "loss": 3.7698,
+      "step": 10050
+    },
+    {
+      "epoch": 1.0889487870619945,
+      "grad_norm": 0.6025657057762146,
+      "learning_rate": 0.0005352531030760928,
+      "loss": 3.7595,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0943396226415094,
+      "grad_norm": 0.6286342144012451,
+      "learning_rate": 0.0005349357798165137,
+      "loss": 3.7772,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0997304582210243,
+      "grad_norm": 0.565532386302948,
+      "learning_rate": 0.0005346119805720453,
+      "loss": 3.7646,
+      "step": 10200
+    },
+    {
+      "epoch": 1.105121293800539,
+      "grad_norm": 0.5853555202484131,
+      "learning_rate": 0.0005342881813275768,
+      "loss": 3.7637,
+      "step": 10250
+    },
+    {
+      "epoch": 1.110512129380054,
+      "grad_norm": 0.6412824988365173,
+      "learning_rate": 0.0005339708580679978,
+      "loss": 3.7642,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1159029649595686,
+      "grad_norm": 0.5646089315414429,
+      "learning_rate": 0.0005336470588235293,
+      "loss": 3.7553,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1212938005390836,
+      "grad_norm": 0.6126633882522583,
+      "learning_rate": 0.000533323259579061,
+      "loss": 3.7789,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1266846361185983,
+      "grad_norm": 0.5688264966011047,
+      "learning_rate": 0.0005329994603345925,
+      "loss": 3.7638,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1320754716981132,
+      "grad_norm": 0.5909513235092163,
+      "learning_rate": 0.0005326756610901241,
+      "loss": 3.7545,
+      "step": 10500
+    },
+    {
+      "epoch": 1.137466307277628,
+      "grad_norm": 0.6165957450866699,
+      "learning_rate": 0.0005323518618456556,
+      "loss": 3.7571,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.6512620449066162,
+      "learning_rate": 0.0005320280626011872,
+      "loss": 3.7638,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1482479784366577,
+      "grad_norm": 0.5640995502471924,
+      "learning_rate": 0.0005317042633567188,
+      "loss": 3.7548,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1536388140161726,
+      "grad_norm": 0.5489684343338013,
+      "learning_rate": 0.0005313804641122504,
+      "loss": 3.745,
+      "step": 10700
+    },
+    {
+      "epoch": 1.1590296495956873,
+      "grad_norm": 0.5724853873252869,
+      "learning_rate": 0.0005310566648677819,
+      "loss": 3.7553,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1644204851752022,
+      "grad_norm": 0.6125072240829468,
+      "learning_rate": 0.0005307328656233135,
+      "loss": 3.7447,
+      "step": 10800
+    },
+    {
+      "epoch": 1.169811320754717,
+      "grad_norm": 0.551490843296051,
+      "learning_rate": 0.000530409066378845,
+      "loss": 3.7589,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1752021563342319,
+      "grad_norm": 0.5600878000259399,
+      "learning_rate": 0.0005300852671343766,
+      "loss": 3.7586,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1805929919137466,
+      "grad_norm": 0.5831260681152344,
+      "learning_rate": 0.0005297614678899081,
+      "loss": 3.7425,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "grad_norm": 0.6004246473312378,
+      "learning_rate": 0.0005294376686454398,
+      "loss": 3.7519,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "eval_accuracy": 0.3466813768189716,
+      "eval_loss": 3.723628520965576,
+      "eval_runtime": 183.7188,
+      "eval_samples_per_second": 98.036,
+      "eval_steps_per_second": 6.129,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1913746630727764,
+      "grad_norm": 0.5393942594528198,
+      "learning_rate": 0.0005291138694009714,
+      "loss": 3.753,
+      "step": 11050
+    },
+    {
+      "epoch": 1.196765498652291,
+      "grad_norm": 0.5937897562980652,
+      "learning_rate": 0.0005287900701565029,
+      "loss": 3.7424,
+      "step": 11100
+    },
+    {
+      "epoch": 1.202156334231806,
+      "grad_norm": 0.5454901456832886,
+      "learning_rate": 0.0005284662709120345,
+      "loss": 3.7416,
+      "step": 11150
+    },
+    {
+      "epoch": 1.2075471698113207,
+      "grad_norm": 0.5742169618606567,
+      "learning_rate": 0.0005281424716675661,
+      "loss": 3.7442,
+      "step": 11200
+    },
+    {
+      "epoch": 1.2129380053908356,
+      "grad_norm": 0.5434097051620483,
+      "learning_rate": 0.0005278186724230977,
+      "loss": 3.7337,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2183288409703503,
+      "grad_norm": 0.5956228971481323,
+      "learning_rate": 0.0005274948731786292,
+      "loss": 3.7535,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2237196765498652,
+      "grad_norm": 0.5586687922477722,
+      "learning_rate": 0.0005271710739341608,
+      "loss": 3.7113,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2291105121293802,
+      "grad_norm": 0.5536643266677856,
+      "learning_rate": 0.0005268472746896923,
+      "loss": 3.7524,
+      "step": 11400
+    },
+    {
+      "epoch": 1.2345013477088949,
+      "grad_norm": 0.5457449555397034,
+      "learning_rate": 0.000526523475445224,
+      "loss": 3.7376,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2398921832884098,
+      "grad_norm": 0.6064246892929077,
+      "learning_rate": 0.0005261996762007554,
+      "loss": 3.7229,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2452830188679245,
+      "grad_norm": 0.5753145217895508,
+      "learning_rate": 0.0005258758769562871,
+      "loss": 3.7512,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2506738544474394,
+      "grad_norm": 0.6125341057777405,
+      "learning_rate": 0.0005255520777118186,
+      "loss": 3.7342,
+      "step": 11600
+    },
+    {
+      "epoch": 1.256064690026954,
+      "grad_norm": 0.589798629283905,
+      "learning_rate": 0.0005252282784673502,
+      "loss": 3.7213,
+      "step": 11650
+    },
+    {
+      "epoch": 1.261455525606469,
+      "grad_norm": 0.6119717359542847,
+      "learning_rate": 0.0005249044792228817,
+      "loss": 3.7299,
+      "step": 11700
+    },
+    {
+      "epoch": 1.266846361185984,
+      "grad_norm": 0.6549661159515381,
+      "learning_rate": 0.0005245806799784133,
+      "loss": 3.7474,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2722371967654986,
+      "grad_norm": 0.5407906174659729,
+      "learning_rate": 0.0005242568807339449,
+      "loss": 3.7444,
+      "step": 11800
+    },
+    {
+      "epoch": 1.2776280323450135,
+      "grad_norm": 0.5647597312927246,
+      "learning_rate": 0.0005239330814894765,
+      "loss": 3.744,
+      "step": 11850
+    },
+    {
+      "epoch": 1.2830188679245282,
+      "grad_norm": 0.6143915057182312,
+      "learning_rate": 0.000523609282245008,
+      "loss": 3.7301,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2884097035040432,
+      "grad_norm": 0.5904483795166016,
+      "learning_rate": 0.0005232854830005396,
+      "loss": 3.7286,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "grad_norm": 0.5413955450057983,
+      "learning_rate": 0.0005229616837560712,
+      "loss": 3.7387,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "eval_accuracy": 0.3488454163855726,
+      "eval_loss": 3.6995937824249268,
+      "eval_runtime": 184.1584,
+      "eval_samples_per_second": 97.802,
+      "eval_steps_per_second": 6.114,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2991913746630728,
+      "grad_norm": 0.5844531059265137,
+      "learning_rate": 0.0005226378845116028,
+      "loss": 3.7271,
+      "step": 12050
+    },
+    {
+      "epoch": 1.3045822102425877,
+      "grad_norm": 0.692406415939331,
+      "learning_rate": 0.0005223140852671344,
+      "loss": 3.7326,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3099730458221024,
+      "grad_norm": 0.531926691532135,
+      "learning_rate": 0.0005219902860226659,
+      "loss": 3.7207,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3153638814016173,
+      "grad_norm": 0.5657473206520081,
+      "learning_rate": 0.0005216664867781975,
+      "loss": 3.7375,
+      "step": 12200
+    },
+    {
+      "epoch": 1.320754716981132,
+      "grad_norm": 0.5740625858306885,
+      "learning_rate": 0.000521342687533729,
+      "loss": 3.7009,
+      "step": 12250
+    },
+    {
+      "epoch": 1.326145552560647,
+      "grad_norm": 0.5976618528366089,
+      "learning_rate": 0.0005210188882892606,
+      "loss": 3.7069,
+      "step": 12300
+    },
+    {
+      "epoch": 1.3315363881401616,
+      "grad_norm": 0.5708423256874084,
+      "learning_rate": 0.0005206950890447922,
+      "loss": 3.7345,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3369272237196765,
+      "grad_norm": 0.6060131192207336,
+      "learning_rate": 0.0005203712898003238,
+      "loss": 3.7245,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3423180592991915,
+      "grad_norm": 0.5739725232124329,
+      "learning_rate": 0.0005200474905558553,
+      "loss": 3.7225,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3477088948787062,
+      "grad_norm": 0.5318633913993835,
+      "learning_rate": 0.0005197236913113869,
+      "loss": 3.6992,
+      "step": 12500
+    },
+    {
+      "epoch": 1.353099730458221,
+      "grad_norm": 0.5786360502243042,
+      "learning_rate": 0.0005193998920669184,
+      "loss": 3.7161,
+      "step": 12550
+    },
+    {
+      "epoch": 1.3584905660377358,
+      "grad_norm": 0.5843545198440552,
+      "learning_rate": 0.0005190760928224501,
+      "loss": 3.7138,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3638814016172507,
+      "grad_norm": 0.579613447189331,
+      "learning_rate": 0.0005187522935779816,
+      "loss": 3.7258,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3692722371967654,
+      "grad_norm": 0.6323828101158142,
+      "learning_rate": 0.0005184284943335132,
+      "loss": 3.7231,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3746630727762803,
+      "grad_norm": 0.5477507710456848,
+      "learning_rate": 0.0005181046950890447,
+      "loss": 3.7115,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3800539083557952,
+      "grad_norm": 0.5417460203170776,
+      "learning_rate": 0.0005177808958445764,
+      "loss": 3.7014,
+      "step": 12800
+    },
+    {
+      "epoch": 1.38544474393531,
+      "grad_norm": 0.5849714279174805,
+      "learning_rate": 0.0005174635725849973,
+      "loss": 3.7173,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3908355795148248,
+      "grad_norm": 0.5853115916252136,
+      "learning_rate": 0.0005171397733405288,
+      "loss": 3.7121,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3962264150943398,
+      "grad_norm": 0.5460019111633301,
+      "learning_rate": 0.0005168159740960603,
+      "loss": 3.738,
+      "step": 12950
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "grad_norm": 0.6247442960739136,
+      "learning_rate": 0.000516492174851592,
+      "loss": 3.7317,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "eval_accuracy": 0.3512436029254138,
+      "eval_loss": 3.6774609088897705,
+      "eval_runtime": 183.7689,
+      "eval_samples_per_second": 98.009,
+      "eval_steps_per_second": 6.127,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4070080862533692,
+      "grad_norm": 0.6057413220405579,
+      "learning_rate": 0.0005161683756071235,
+      "loss": 3.7133,
+      "step": 13050
+    },
+    {
+      "epoch": 1.412398921832884,
+      "grad_norm": 0.5644381642341614,
+      "learning_rate": 0.0005158445763626551,
+      "loss": 3.698,
+      "step": 13100
+    },
+    {
+      "epoch": 1.417789757412399,
+      "grad_norm": 0.5523656010627747,
+      "learning_rate": 0.0005155207771181866,
+      "loss": 3.6978,
+      "step": 13150
+    },
+    {
+      "epoch": 1.4231805929919137,
+      "grad_norm": 0.5388391613960266,
+      "learning_rate": 0.0005151969778737183,
+      "loss": 3.7136,
+      "step": 13200
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.6056439876556396,
+      "learning_rate": 0.0005148731786292498,
+      "loss": 3.7094,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4339622641509435,
+      "grad_norm": 0.6014870405197144,
+      "learning_rate": 0.0005145493793847814,
+      "loss": 3.7149,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4393530997304582,
+      "grad_norm": 0.6494507193565369,
+      "learning_rate": 0.000514225580140313,
+      "loss": 3.7201,
+      "step": 13350
+    },
+    {
+      "epoch": 1.444743935309973,
+      "grad_norm": 0.5282674431800842,
+      "learning_rate": 0.0005139017808958445,
+      "loss": 3.712,
+      "step": 13400
+    },
+    {
+      "epoch": 1.4501347708894878,
+      "grad_norm": 0.5607653856277466,
+      "learning_rate": 0.0005135779816513762,
+      "loss": 3.7172,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4555256064690028,
+      "grad_norm": 0.5750332474708557,
+      "learning_rate": 0.0005132541824069076,
+      "loss": 3.7082,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4609164420485174,
+      "grad_norm": 0.5525111556053162,
+      "learning_rate": 0.0005129303831624393,
+      "loss": 3.7189,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4663072776280324,
+      "grad_norm": 0.5573806762695312,
+      "learning_rate": 0.0005126065839179708,
+      "loss": 3.7074,
+      "step": 13600
+    },
+    {
+      "epoch": 1.4716981132075473,
+      "grad_norm": 0.5327258706092834,
+      "learning_rate": 0.0005122827846735024,
+      "loss": 3.7108,
+      "step": 13650
+    },
+    {
+      "epoch": 1.477088948787062,
+      "grad_norm": 0.6094257831573486,
+      "learning_rate": 0.0005119589854290339,
+      "loss": 3.7015,
+      "step": 13700
+    },
+    {
+      "epoch": 1.482479784366577,
+      "grad_norm": 0.5354958772659302,
+      "learning_rate": 0.0005116351861845655,
+      "loss": 3.7019,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4878706199460916,
+      "grad_norm": 0.5956083536148071,
+      "learning_rate": 0.0005113113869400971,
+      "loss": 3.696,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4932614555256065,
+      "grad_norm": 0.5589125156402588,
+      "learning_rate": 0.0005109875876956287,
+      "loss": 3.6968,
+      "step": 13850
+    },
+    {
+      "epoch": 1.4986522911051212,
+      "grad_norm": 0.5720492601394653,
+      "learning_rate": 0.0005106637884511602,
+      "loss": 3.6979,
+      "step": 13900
+    },
+    {
+      "epoch": 1.5040431266846361,
+      "grad_norm": 0.5775404572486877,
+      "learning_rate": 0.0005103399892066918,
+      "loss": 3.6881,
+      "step": 13950
+    },
+    {
+      "epoch": 1.509433962264151,
+      "grad_norm": 0.5249178409576416,
+      "learning_rate": 0.0005100161899622234,
+      "loss": 3.6971,
+      "step": 14000
+    },
+    {
+      "epoch": 1.509433962264151,
+      "eval_accuracy": 0.35375946054275814,
+      "eval_loss": 3.6579151153564453,
+      "eval_runtime": 184.1099,
+      "eval_samples_per_second": 97.827,
+      "eval_steps_per_second": 6.116,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5148247978436657,
+      "grad_norm": 0.5867551565170288,
+      "learning_rate": 0.000509692390717755,
+      "loss": 3.6801,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5202156334231804,
+      "grad_norm": 0.5887311100959778,
+      "learning_rate": 0.0005093685914732865,
+      "loss": 3.6936,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5256064690026954,
+      "grad_norm": 0.5415863394737244,
+      "learning_rate": 0.0005090447922288181,
+      "loss": 3.6752,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5309973045822103,
+      "grad_norm": 0.5622326731681824,
+      "learning_rate": 0.0005087209929843496,
+      "loss": 3.6959,
+      "step": 14200
+    },
+    {
+      "epoch": 1.536388140161725,
+      "grad_norm": 0.5287529230117798,
+      "learning_rate": 0.0005083971937398812,
+      "loss": 3.7028,
+      "step": 14250
+    },
+    {
+      "epoch": 1.54177897574124,
+      "grad_norm": 0.5576687455177307,
+      "learning_rate": 0.0005080733944954127,
+      "loss": 3.7062,
+      "step": 14300
+    },
+    {
+      "epoch": 1.5471698113207548,
+      "grad_norm": 0.5946956872940063,
+      "learning_rate": 0.0005077495952509444,
+      "loss": 3.7062,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5525606469002695,
+      "grad_norm": 0.5189834237098694,
+      "learning_rate": 0.0005074257960064759,
+      "loss": 3.6921,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5579514824797842,
+      "grad_norm": 0.5719146728515625,
+      "learning_rate": 0.0005071019967620075,
+      "loss": 3.6974,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5633423180592994,
+      "grad_norm": 0.5453420281410217,
+      "learning_rate": 0.000506778197517539,
+      "loss": 3.6744,
+      "step": 14500
+    },
+    {
+      "epoch": 1.568733153638814,
+      "grad_norm": 0.5827313661575317,
+      "learning_rate": 0.0005064543982730707,
+      "loss": 3.6753,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5741239892183287,
+      "grad_norm": 0.5802609920501709,
+      "learning_rate": 0.0005061305990286023,
+      "loss": 3.6732,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5795148247978437,
+      "grad_norm": 0.5037614703178406,
+      "learning_rate": 0.0005058067997841338,
+      "loss": 3.6892,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5849056603773586,
+      "grad_norm": 0.5056049823760986,
+      "learning_rate": 0.0005054830005396654,
+      "loss": 3.6858,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5902964959568733,
+      "grad_norm": 0.6599529385566711,
+      "learning_rate": 0.0005051592012951969,
+      "loss": 3.6799,
+      "step": 14750
+    },
+    {
+      "epoch": 1.595687331536388,
+      "grad_norm": 0.563842236995697,
+      "learning_rate": 0.0005048354020507286,
+      "loss": 3.6988,
+      "step": 14800
+    },
+    {
+      "epoch": 1.6010781671159031,
+      "grad_norm": 0.5434414744377136,
+      "learning_rate": 0.00050451160280626,
+      "loss": 3.6767,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6064690026954178,
+      "grad_norm": 0.61106938123703,
+      "learning_rate": 0.0005041878035617917,
+      "loss": 3.6979,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6118598382749325,
+      "grad_norm": 0.5468453764915466,
+      "learning_rate": 0.0005038640043173232,
+      "loss": 3.6902,
+      "step": 14950
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "grad_norm": 0.589815616607666,
+      "learning_rate": 0.0005035402050728548,
+      "loss": 3.6885,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "eval_accuracy": 0.3551121889960484,
+      "eval_loss": 3.6401309967041016,
+      "eval_runtime": 183.883,
+      "eval_samples_per_second": 97.948,
+      "eval_steps_per_second": 6.123,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6226415094339623,
+      "grad_norm": 0.5671663284301758,
+      "learning_rate": 0.0005032164058283863,
+      "loss": 3.6831,
+      "step": 15050
+    },
+    {
+      "epoch": 1.628032345013477,
+      "grad_norm": 0.6244860887527466,
+      "learning_rate": 0.0005028926065839179,
+      "loss": 3.6613,
+      "step": 15100
+    },
+    {
+      "epoch": 1.633423180592992,
+      "grad_norm": 0.5171533823013306,
+      "learning_rate": 0.0005025688073394495,
+      "loss": 3.6778,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6388140161725069,
+      "grad_norm": 0.5679621696472168,
+      "learning_rate": 0.0005022450080949811,
+      "loss": 3.6648,
+      "step": 15200
+    },
+    {
+      "epoch": 1.6442048517520216,
+      "grad_norm": 0.5312994718551636,
+      "learning_rate": 0.0005019212088505126,
+      "loss": 3.6837,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6495956873315363,
+      "grad_norm": 0.5288447737693787,
+      "learning_rate": 0.0005015974096060442,
+      "loss": 3.6684,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6549865229110512,
+      "grad_norm": 0.6825982928276062,
+      "learning_rate": 0.0005012736103615758,
+      "loss": 3.671,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6603773584905661,
+      "grad_norm": 0.5444086790084839,
+      "learning_rate": 0.0005009498111171074,
+      "loss": 3.7001,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6657681940700808,
+      "grad_norm": 0.5970629453659058,
+      "learning_rate": 0.0005006260118726389,
+      "loss": 3.6732,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6711590296495957,
+      "grad_norm": 0.6333000063896179,
+      "learning_rate": 0.0005003022126281705,
+      "loss": 3.6567,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6765498652291106,
+      "grad_norm": 0.5841792821884155,
+      "learning_rate": 0.000499978413383702,
+      "loss": 3.6866,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6819407008086253,
+      "grad_norm": 0.5716648697853088,
+      "learning_rate": 0.0004996546141392336,
+      "loss": 3.6464,
+      "step": 15600
+    },
+    {
+      "epoch": 1.68733153638814,
+      "grad_norm": 0.6867700219154358,
+      "learning_rate": 0.0004993308148947651,
+      "loss": 3.667,
+      "step": 15650
+    },
+    {
+      "epoch": 1.692722371967655,
+      "grad_norm": 0.5503226518630981,
+      "learning_rate": 0.0004990070156502968,
+      "loss": 3.6594,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6981132075471699,
+      "grad_norm": 0.5787608623504639,
+      "learning_rate": 0.0004986832164058284,
+      "loss": 3.6795,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7035040431266846,
+      "grad_norm": 0.5347179770469666,
+      "learning_rate": 0.0004983594171613599,
+      "loss": 3.6652,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7088948787061995,
+      "grad_norm": 0.5248307585716248,
+      "learning_rate": 0.0004980356179168915,
+      "loss": 3.6812,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.5549546480178833,
+      "learning_rate": 0.000497711818672423,
+      "loss": 3.6636,
+      "step": 15900
+    },
+    {
+      "epoch": 1.719676549865229,
+      "grad_norm": 0.6331003308296204,
+      "learning_rate": 0.0004973880194279547,
+      "loss": 3.6594,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "grad_norm": 0.6510477662086487,
+      "learning_rate": 0.0004970642201834862,
+      "loss": 3.6771,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "eval_accuracy": 0.3569796061789159,
+      "eval_loss": 3.619459867477417,
+      "eval_runtime": 183.9465,
+      "eval_samples_per_second": 97.914,
+      "eval_steps_per_second": 6.121,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7304582210242587,
+      "grad_norm": 0.6069450378417969,
+      "learning_rate": 0.0004967404209390178,
+      "loss": 3.654,
+      "step": 16050
+    },
+    {
+      "epoch": 1.7358490566037736,
+      "grad_norm": 0.5242086052894592,
+      "learning_rate": 0.0004964166216945493,
+      "loss": 3.6722,
+      "step": 16100
+    },
+    {
+      "epoch": 1.7412398921832883,
+      "grad_norm": 0.612231433391571,
+      "learning_rate": 0.000496092822450081,
+      "loss": 3.6597,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7466307277628033,
+      "grad_norm": 0.6022439002990723,
+      "learning_rate": 0.0004957690232056125,
+      "loss": 3.6801,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7520215633423182,
+      "grad_norm": 0.5812779664993286,
+      "learning_rate": 0.0004954452239611441,
+      "loss": 3.6582,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7574123989218329,
+      "grad_norm": 0.5933021306991577,
+      "learning_rate": 0.0004951214247166756,
+      "loss": 3.6652,
+      "step": 16300
+    },
+    {
+      "epoch": 1.7628032345013476,
+      "grad_norm": 0.5812273621559143,
+      "learning_rate": 0.0004947976254722072,
+      "loss": 3.6697,
+      "step": 16350
+    },
+    {
+      "epoch": 1.7681940700808625,
+      "grad_norm": 0.6487030386924744,
+      "learning_rate": 0.0004944738262277387,
+      "loss": 3.6677,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7735849056603774,
+      "grad_norm": 0.548700749874115,
+      "learning_rate": 0.0004941500269832703,
+      "loss": 3.658,
+      "step": 16450
+    },
+    {
+      "epoch": 1.778975741239892,
+      "grad_norm": 0.6098374724388123,
+      "learning_rate": 0.0004938262277388019,
+      "loss": 3.6542,
+      "step": 16500
+    },
+    {
+      "epoch": 1.784366576819407,
+      "grad_norm": 0.6051768660545349,
+      "learning_rate": 0.0004935024284943335,
+      "loss": 3.6566,
+      "step": 16550
+    },
+    {
+      "epoch": 1.789757412398922,
+      "grad_norm": 0.6026385426521301,
+      "learning_rate": 0.000493178629249865,
+      "loss": 3.653,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7951482479784366,
+      "grad_norm": 0.562525749206543,
+      "learning_rate": 0.0004928548300053966,
+      "loss": 3.6491,
+      "step": 16650
+    },
+    {
+      "epoch": 1.8005390835579513,
+      "grad_norm": 0.571699321269989,
+      "learning_rate": 0.0004925310307609282,
+      "loss": 3.6552,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8059299191374663,
+      "grad_norm": 0.5961069464683533,
+      "learning_rate": 0.0004922072315164598,
+      "loss": 3.6492,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8113207547169812,
+      "grad_norm": 0.5025362372398376,
+      "learning_rate": 0.0004918834322719913,
+      "loss": 3.6478,
+      "step": 16800
+    },
+    {
+      "epoch": 1.8167115902964959,
+      "grad_norm": 0.5551483631134033,
+      "learning_rate": 0.0004915661090124123,
+      "loss": 3.6596,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8221024258760108,
+      "grad_norm": 0.5412172079086304,
+      "learning_rate": 0.0004912423097679439,
+      "loss": 3.6498,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8274932614555257,
+      "grad_norm": 0.6038126945495605,
+      "learning_rate": 0.0004909185105234754,
+      "loss": 3.6542,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "grad_norm": 0.5766414403915405,
+      "learning_rate": 0.000490594711279007,
+      "loss": 3.6503,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "eval_accuracy": 0.35894415904348953,
+      "eval_loss": 3.6033706665039062,
+      "eval_runtime": 184.1497,
+      "eval_samples_per_second": 97.806,
+      "eval_steps_per_second": 6.115,
+      "step": 17000
+    },
+    {
+      "epoch": 1.838274932614555,
+      "grad_norm": 0.582373857498169,
+      "learning_rate": 0.0004902709120345385,
+      "loss": 3.641,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8436657681940702,
+      "grad_norm": 0.5415629148483276,
+      "learning_rate": 0.00048994711279007,
+      "loss": 3.6456,
+      "step": 17100
+    },
+    {
+      "epoch": 1.849056603773585,
+      "grad_norm": 0.5694625973701477,
+      "learning_rate": 0.0004896233135456017,
+      "loss": 3.6568,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8544474393530996,
+      "grad_norm": 0.547035276889801,
+      "learning_rate": 0.0004892995143011333,
+      "loss": 3.654,
+      "step": 17200
+    },
+    {
+      "epoch": 1.8598382749326146,
+      "grad_norm": 0.6010807752609253,
+      "learning_rate": 0.0004889757150566648,
+      "loss": 3.6348,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8652291105121295,
+      "grad_norm": 0.5640713572502136,
+      "learning_rate": 0.0004886519158121964,
+      "loss": 3.6555,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8706199460916442,
+      "grad_norm": 0.6125467419624329,
+      "learning_rate": 0.000488328116567728,
+      "loss": 3.656,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8760107816711589,
+      "grad_norm": 0.6842520833015442,
+      "learning_rate": 0.0004880043173232595,
+      "loss": 3.6642,
+      "step": 17400
+    },
+    {
+      "epoch": 1.881401617250674,
+      "grad_norm": 0.5659988522529602,
+      "learning_rate": 0.0004876805180787911,
+      "loss": 3.6325,
+      "step": 17450
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.6425742506980896,
+      "learning_rate": 0.0004873567188343227,
+      "loss": 3.6474,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8921832884097034,
+      "grad_norm": 0.5763974785804749,
+      "learning_rate": 0.0004870329195898542,
+      "loss": 3.6383,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8975741239892183,
+      "grad_norm": 0.5723247528076172,
+      "learning_rate": 0.00048670912034538583,
+      "loss": 3.6529,
+      "step": 17600
+    },
+    {
+      "epoch": 1.9029649595687332,
+      "grad_norm": 0.5686931610107422,
+      "learning_rate": 0.0004863853211009174,
+      "loss": 3.6496,
+      "step": 17650
+    },
+    {
+      "epoch": 1.908355795148248,
+      "grad_norm": 0.5761986970901489,
+      "learning_rate": 0.000486061521856449,
+      "loss": 3.6397,
+      "step": 17700
+    },
+    {
+      "epoch": 1.9137466307277629,
+      "grad_norm": 0.6086380481719971,
+      "learning_rate": 0.00048573772261198054,
+      "loss": 3.6421,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9191374663072778,
+      "grad_norm": 0.5803984999656677,
+      "learning_rate": 0.00048541392336751214,
+      "loss": 3.6658,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9245283018867925,
+      "grad_norm": 0.6012025475502014,
+      "learning_rate": 0.0004850901241230437,
+      "loss": 3.6396,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9299191374663072,
+      "grad_norm": 0.5470909476280212,
+      "learning_rate": 0.0004847663248785753,
+      "loss": 3.6366,
+      "step": 17900
+    },
+    {
+      "epoch": 1.935309973045822,
+      "grad_norm": 0.6374045014381409,
+      "learning_rate": 0.0004844425256341068,
+      "loss": 3.6164,
+      "step": 17950
+    },
+    {
+      "epoch": 1.940700808625337,
+      "grad_norm": 0.5931817889213562,
+      "learning_rate": 0.00048411872638963834,
+      "loss": 3.6523,
+      "step": 18000
+    },
+    {
+      "epoch": 1.940700808625337,
+      "eval_accuracy": 0.35995310975973477,
+      "eval_loss": 3.5907158851623535,
+      "eval_runtime": 183.7191,
+      "eval_samples_per_second": 98.036,
+      "eval_steps_per_second": 6.129,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9460916442048517,
+      "grad_norm": 0.5779642462730408,
+      "learning_rate": 0.00048379492714516995,
+      "loss": 3.6288,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9514824797843666,
+      "grad_norm": 0.5749161839485168,
+      "learning_rate": 0.0004834711279007015,
+      "loss": 3.6552,
+      "step": 18100
+    },
+    {
+      "epoch": 1.9568733153638815,
+      "grad_norm": 0.5829963684082031,
+      "learning_rate": 0.0004831473286562331,
+      "loss": 3.6656,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9622641509433962,
+      "grad_norm": 0.547026515007019,
+      "learning_rate": 0.00048282352941176465,
+      "loss": 3.6443,
+      "step": 18200
+    },
+    {
+      "epoch": 1.967654986522911,
+      "grad_norm": 0.5671616792678833,
+      "learning_rate": 0.00048249973016729626,
+      "loss": 3.6491,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9730458221024259,
+      "grad_norm": 0.5770370960235596,
+      "learning_rate": 0.0004821759309228278,
+      "loss": 3.6299,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9784366576819408,
+      "grad_norm": 0.5550295114517212,
+      "learning_rate": 0.00048185213167835936,
+      "loss": 3.6396,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9838274932614555,
+      "grad_norm": 0.5505232214927673,
+      "learning_rate": 0.00048152833243389096,
+      "loss": 3.6541,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9892183288409704,
+      "grad_norm": 0.5260711908340454,
+      "learning_rate": 0.0004812045331894225,
+      "loss": 3.6188,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9946091644204853,
+      "grad_norm": 0.5258703231811523,
+      "learning_rate": 0.0004808807339449541,
+      "loss": 3.6256,
+      "step": 18500
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.088941216468811,
+      "learning_rate": 0.0004805569347004856,
+      "loss": 3.6356,
+      "step": 18550
+    },
+    {
+      "epoch": 2.0053908355795147,
+      "grad_norm": 0.5310294032096863,
+      "learning_rate": 0.0004802331354560173,
+      "loss": 3.542,
+      "step": 18600
+    },
+    {
+      "epoch": 2.01078167115903,
+      "grad_norm": 0.5242578387260437,
+      "learning_rate": 0.00047990933621154877,
+      "loss": 3.5532,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0161725067385445,
+      "grad_norm": 0.5820230841636658,
+      "learning_rate": 0.0004795855369670804,
+      "loss": 3.5545,
+      "step": 18700
+    },
+    {
+      "epoch": 2.0215633423180592,
+      "grad_norm": 0.6573295593261719,
+      "learning_rate": 0.0004792617377226119,
+      "loss": 3.5472,
+      "step": 18750
+    },
+    {
+      "epoch": 2.026954177897574,
+      "grad_norm": 0.5828113555908203,
+      "learning_rate": 0.0004789379384781435,
+      "loss": 3.5532,
+      "step": 18800
+    },
+    {
+      "epoch": 2.032345013477089,
+      "grad_norm": 0.6123409867286682,
+      "learning_rate": 0.0004786141392336751,
+      "loss": 3.5318,
+      "step": 18850
+    },
+    {
+      "epoch": 2.0377358490566038,
+      "grad_norm": 0.5386969447135925,
+      "learning_rate": 0.00047829033998920663,
+      "loss": 3.5529,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0431266846361185,
+      "grad_norm": 0.5555720329284668,
+      "learning_rate": 0.00047796654074473824,
+      "loss": 3.5571,
+      "step": 18950
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "grad_norm": 0.5521024465560913,
+      "learning_rate": 0.0004776427415002698,
+      "loss": 3.5469,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "eval_accuracy": 0.3615377034756212,
+      "eval_loss": 3.5789687633514404,
+      "eval_runtime": 184.2139,
+      "eval_samples_per_second": 97.772,
+      "eval_steps_per_second": 6.112,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0539083557951483,
+      "grad_norm": 0.6329952478408813,
+      "learning_rate": 0.00047732541824069074,
+      "loss": 3.5373,
+      "step": 19050
+    },
+    {
+      "epoch": 2.059299191374663,
+      "grad_norm": 0.5793041586875916,
+      "learning_rate": 0.0004770016189962223,
+      "loss": 3.5498,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0646900269541777,
+      "grad_norm": 0.5503807067871094,
+      "learning_rate": 0.0004766778197517539,
+      "loss": 3.5437,
+      "step": 19150
+    },
+    {
+      "epoch": 2.070080862533693,
+      "grad_norm": 0.5668014883995056,
+      "learning_rate": 0.0004763540205072854,
+      "loss": 3.5563,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0754716981132075,
+      "grad_norm": 0.5484731197357178,
+      "learning_rate": 0.00047603022126281705,
+      "loss": 3.554,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0808625336927222,
+      "grad_norm": 0.6245366930961609,
+      "learning_rate": 0.00047570642201834855,
+      "loss": 3.5468,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0862533692722374,
+      "grad_norm": 0.6225091814994812,
+      "learning_rate": 0.00047538262277388015,
+      "loss": 3.5642,
+      "step": 19350
+    },
+    {
+      "epoch": 2.091644204851752,
+      "grad_norm": 0.5625278949737549,
+      "learning_rate": 0.0004750588235294117,
+      "loss": 3.5567,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0970350404312668,
+      "grad_norm": 0.632989764213562,
+      "learning_rate": 0.0004747350242849433,
+      "loss": 3.5484,
+      "step": 19450
+    },
+    {
+      "epoch": 2.1024258760107815,
+      "grad_norm": 0.5744699239730835,
+      "learning_rate": 0.00047441122504047486,
+      "loss": 3.5617,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1078167115902966,
+      "grad_norm": 0.5948736071586609,
+      "learning_rate": 0.0004740874257960064,
+      "loss": 3.542,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1132075471698113,
+      "grad_norm": 0.5679342746734619,
+      "learning_rate": 0.000473763626551538,
+      "loss": 3.5556,
+      "step": 19600
+    },
+    {
+      "epoch": 2.118598382749326,
+      "grad_norm": 0.5541220307350159,
+      "learning_rate": 0.00047343982730706956,
+      "loss": 3.5516,
+      "step": 19650
+    },
+    {
+      "epoch": 2.123989218328841,
+      "grad_norm": 0.544877290725708,
+      "learning_rate": 0.00047311602806260117,
+      "loss": 3.5489,
+      "step": 19700
+    },
+    {
+      "epoch": 2.129380053908356,
+      "grad_norm": 0.5709410309791565,
+      "learning_rate": 0.0004727922288181327,
+      "loss": 3.5652,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1347708894878705,
+      "grad_norm": 0.5732446908950806,
+      "learning_rate": 0.0004724684295736643,
+      "loss": 3.5745,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1401617250673857,
+      "grad_norm": 0.6136924028396606,
+      "learning_rate": 0.0004721446303291959,
+      "loss": 3.5503,
+      "step": 19850
+    },
+    {
+      "epoch": 2.1455525606469004,
+      "grad_norm": 0.5857067108154297,
+      "learning_rate": 0.0004718208310847275,
+      "loss": 3.5585,
+      "step": 19900
+    },
+    {
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.5760287046432495,
+      "learning_rate": 0.000471497031840259,
+      "loss": 3.5601,
+      "step": 19950
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "grad_norm": 0.5563052296638489,
+      "learning_rate": 0.0004711732325957905,
+      "loss": 3.553,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "eval_accuracy": 0.36297746289205085,
+      "eval_loss": 3.5687758922576904,
+      "eval_runtime": 183.8162,
+      "eval_samples_per_second": 97.984,
+      "eval_steps_per_second": 6.126,
+      "step": 20000
+    },
+    {
+      "epoch": 2.161725067385445,
+      "grad_norm": 0.6046416759490967,
+      "learning_rate": 0.00047084943335132213,
+      "loss": 3.5431,
+      "step": 20050
+    },
+    {
+      "epoch": 2.1671159029649596,
+      "grad_norm": 0.5706889033317566,
+      "learning_rate": 0.0004705256341068537,
+      "loss": 3.5522,
+      "step": 20100
+    },
+    {
+      "epoch": 2.1725067385444743,
+      "grad_norm": 0.6285442113876343,
+      "learning_rate": 0.0004702018348623853,
+      "loss": 3.5604,
+      "step": 20150
+    },
+    {
+      "epoch": 2.177897574123989,
+      "grad_norm": 0.567588746547699,
+      "learning_rate": 0.00046987803561791684,
+      "loss": 3.5649,
+      "step": 20200
+    },
+    {
+      "epoch": 2.183288409703504,
+      "grad_norm": 0.543428897857666,
+      "learning_rate": 0.00046955423637344844,
+      "loss": 3.5729,
+      "step": 20250
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 0.6000525951385498,
+      "learning_rate": 0.00046923043712898,
+      "loss": 3.5503,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1940700808625335,
+      "grad_norm": 0.6174874305725098,
+      "learning_rate": 0.0004689066378845116,
+      "loss": 3.5539,
+      "step": 20350
+    },
+    {
+      "epoch": 2.1994609164420487,
+      "grad_norm": 0.591032087802887,
+      "learning_rate": 0.00046858283864004315,
+      "loss": 3.5589,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2048517520215634,
+      "grad_norm": 0.5443971753120422,
+      "learning_rate": 0.0004682590393955747,
+      "loss": 3.551,
+      "step": 20450
+    },
+    {
+      "epoch": 2.210242587601078,
+      "grad_norm": 0.6590529084205627,
+      "learning_rate": 0.0004679352401511063,
+      "loss": 3.5497,
+      "step": 20500
+    },
+    {
+      "epoch": 2.215633423180593,
+      "grad_norm": 0.592398464679718,
+      "learning_rate": 0.0004676114409066378,
+      "loss": 3.5583,
+      "step": 20550
+    },
+    {
+      "epoch": 2.221024258760108,
+      "grad_norm": 0.5806289911270142,
+      "learning_rate": 0.00046728764166216946,
+      "loss": 3.5656,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2264150943396226,
+      "grad_norm": 0.5720568299293518,
+      "learning_rate": 0.00046696384241770095,
+      "loss": 3.5663,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2318059299191373,
+      "grad_norm": 0.6045317649841309,
+      "learning_rate": 0.00046664004317323256,
+      "loss": 3.5707,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2371967654986524,
+      "grad_norm": 0.587020218372345,
+      "learning_rate": 0.0004663162439287641,
+      "loss": 3.5501,
+      "step": 20750
+    },
+    {
+      "epoch": 2.242587601078167,
+      "grad_norm": 0.6044445037841797,
+      "learning_rate": 0.00046599244468429566,
+      "loss": 3.5433,
+      "step": 20800
+    },
+    {
+      "epoch": 2.247978436657682,
+      "grad_norm": 0.5762318968772888,
+      "learning_rate": 0.00046566864543982726,
+      "loss": 3.5415,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2533692722371965,
+      "grad_norm": 0.5729345083236694,
+      "learning_rate": 0.0004653448461953588,
+      "loss": 3.5473,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2587601078167117,
+      "grad_norm": 0.619462251663208,
+      "learning_rate": 0.0004650210469508904,
+      "loss": 3.5689,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "grad_norm": 0.5913404226303101,
+      "learning_rate": 0.00046469724770642197,
+      "loss": 3.5623,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "eval_accuracy": 0.3636772961424639,
+      "eval_loss": 3.5585999488830566,
+      "eval_runtime": 183.9266,
+      "eval_samples_per_second": 97.925,
+      "eval_steps_per_second": 6.122,
+      "step": 21000
+    },
+    {
+      "epoch": 2.269541778975741,
+      "grad_norm": 0.584413468837738,
+      "learning_rate": 0.0004643734484619536,
+      "loss": 3.5674,
+      "step": 21050
+    },
+    {
+      "epoch": 2.274932614555256,
+      "grad_norm": 0.5533130764961243,
+      "learning_rate": 0.0004640496492174851,
+      "loss": 3.5502,
+      "step": 21100
+    },
+    {
+      "epoch": 2.280323450134771,
+      "grad_norm": 0.5667862296104431,
+      "learning_rate": 0.0004637323259579061,
+      "loss": 3.5487,
+      "step": 21150
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.5526608824729919,
+      "learning_rate": 0.0004634085267134376,
+      "loss": 3.5586,
+      "step": 21200
+    },
+    {
+      "epoch": 2.2911051212938007,
+      "grad_norm": 0.6517942547798157,
+      "learning_rate": 0.00046308472746896923,
+      "loss": 3.5757,
+      "step": 21250
+    },
+    {
+      "epoch": 2.2964959568733154,
+      "grad_norm": 0.5420659184455872,
+      "learning_rate": 0.00046276092822450073,
+      "loss": 3.5505,
+      "step": 21300
+    },
+    {
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.5619015097618103,
+      "learning_rate": 0.00046243712898003233,
+      "loss": 3.5671,
+      "step": 21350
+    },
+    {
+      "epoch": 2.3072776280323453,
+      "grad_norm": 0.5901092886924744,
+      "learning_rate": 0.0004621133297355639,
+      "loss": 3.5691,
+      "step": 21400
+    },
+    {
+      "epoch": 2.31266846361186,
+      "grad_norm": 0.5045002102851868,
+      "learning_rate": 0.0004617895304910955,
+      "loss": 3.5569,
+      "step": 21450
+    },
+    {
+      "epoch": 2.3180592991913747,
+      "grad_norm": 0.6043163537979126,
+      "learning_rate": 0.00046146573124662704,
+      "loss": 3.5506,
+      "step": 21500
+    },
+    {
+      "epoch": 2.3234501347708894,
+      "grad_norm": 0.5670344829559326,
+      "learning_rate": 0.00046114193200215864,
+      "loss": 3.5562,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3288409703504045,
+      "grad_norm": 0.5510123372077942,
+      "learning_rate": 0.0004608181327576902,
+      "loss": 3.5494,
+      "step": 21600
+    },
+    {
+      "epoch": 2.334231805929919,
+      "grad_norm": 0.5462912917137146,
+      "learning_rate": 0.00046049433351322175,
+      "loss": 3.5591,
+      "step": 21650
+    },
+    {
+      "epoch": 2.339622641509434,
+      "grad_norm": 0.6254079341888428,
+      "learning_rate": 0.00046017053426875335,
+      "loss": 3.5383,
+      "step": 21700
+    },
+    {
+      "epoch": 2.3450134770889486,
+      "grad_norm": 0.5986773371696472,
+      "learning_rate": 0.0004598467350242849,
+      "loss": 3.5331,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3504043126684637,
+      "grad_norm": 0.5816046595573425,
+      "learning_rate": 0.0004595229357798165,
+      "loss": 3.5548,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3557951482479784,
+      "grad_norm": 0.5853009223937988,
+      "learning_rate": 0.00045919913653534806,
+      "loss": 3.552,
+      "step": 21850
+    },
+    {
+      "epoch": 2.361185983827493,
+      "grad_norm": 0.5687203407287598,
+      "learning_rate": 0.00045887533729087966,
+      "loss": 3.5502,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3665768194070083,
+      "grad_norm": 0.604446291923523,
+      "learning_rate": 0.00045855153804641116,
+      "loss": 3.5531,
+      "step": 21950
+    },
+    {
+      "epoch": 2.371967654986523,
+      "grad_norm": 0.570324182510376,
+      "learning_rate": 0.0004582277388019427,
+      "loss": 3.5464,
+      "step": 22000
+    },
+    {
+      "epoch": 2.371967654986523,
+      "eval_accuracy": 0.36487475961906735,
+      "eval_loss": 3.5480477809906006,
+      "eval_runtime": 183.8636,
+      "eval_samples_per_second": 97.958,
+      "eval_steps_per_second": 6.124,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 0.6100995540618896,
+      "learning_rate": 0.0004579039395574743,
+      "loss": 3.5415,
+      "step": 22050
+    },
+    {
+      "epoch": 2.382749326145553,
+      "grad_norm": 0.5746794939041138,
+      "learning_rate": 0.00045758014031300586,
+      "loss": 3.557,
+      "step": 22100
+    },
+    {
+      "epoch": 2.3881401617250675,
+      "grad_norm": 0.6097766757011414,
+      "learning_rate": 0.00045725634106853747,
+      "loss": 3.5718,
+      "step": 22150
+    },
+    {
+      "epoch": 2.393530997304582,
+      "grad_norm": 0.5994210243225098,
+      "learning_rate": 0.000456932541824069,
+      "loss": 3.5689,
+      "step": 22200
+    },
+    {
+      "epoch": 2.398921832884097,
+      "grad_norm": 0.5749094486236572,
+      "learning_rate": 0.0004566087425796006,
+      "loss": 3.538,
+      "step": 22250
+    },
+    {
+      "epoch": 2.404312668463612,
+      "grad_norm": 0.549342691898346,
+      "learning_rate": 0.0004562849433351322,
+      "loss": 3.5508,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4097035040431267,
+      "grad_norm": 0.5516054630279541,
+      "learning_rate": 0.0004559611440906638,
+      "loss": 3.5465,
+      "step": 22350
+    },
+    {
+      "epoch": 2.4150943396226414,
+      "grad_norm": 0.5950953364372253,
+      "learning_rate": 0.00045563734484619533,
+      "loss": 3.5643,
+      "step": 22400
+    },
+    {
+      "epoch": 2.420485175202156,
+      "grad_norm": 0.8571392297744751,
+      "learning_rate": 0.0004553135456017269,
+      "loss": 3.5725,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4258760107816713,
+      "grad_norm": 0.6473979949951172,
+      "learning_rate": 0.0004549897463572585,
+      "loss": 3.5584,
+      "step": 22500
+    },
+    {
+      "epoch": 2.431266846361186,
+      "grad_norm": 0.5623269081115723,
+      "learning_rate": 0.00045466594711279,
+      "loss": 3.5454,
+      "step": 22550
+    },
+    {
+      "epoch": 2.4366576819407006,
+      "grad_norm": 0.5944833159446716,
+      "learning_rate": 0.00045434214786832164,
+      "loss": 3.5497,
+      "step": 22600
+    },
+    {
+      "epoch": 2.442048517520216,
+      "grad_norm": 0.5693549513816833,
+      "learning_rate": 0.00045401834862385314,
+      "loss": 3.552,
+      "step": 22650
+    },
+    {
+      "epoch": 2.4474393530997305,
+      "grad_norm": 0.5360573530197144,
+      "learning_rate": 0.00045369454937938474,
+      "loss": 3.5426,
+      "step": 22700
+    },
+    {
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.5875699520111084,
+      "learning_rate": 0.0004533707501349163,
+      "loss": 3.5768,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4582210242587603,
+      "grad_norm": 0.5625613331794739,
+      "learning_rate": 0.0004530469508904479,
+      "loss": 3.543,
+      "step": 22800
+    },
+    {
+      "epoch": 2.463611859838275,
+      "grad_norm": 0.5869686603546143,
+      "learning_rate": 0.0004527296276308688,
+      "loss": 3.5705,
+      "step": 22850
+    },
+    {
+      "epoch": 2.4690026954177897,
+      "grad_norm": 0.5737190842628479,
+      "learning_rate": 0.0004524058283864004,
+      "loss": 3.5454,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4743935309973044,
+      "grad_norm": 0.5693169236183167,
+      "learning_rate": 0.00045208202914193195,
+      "loss": 3.5611,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "grad_norm": 0.6664920449256897,
+      "learning_rate": 0.00045175822989746355,
+      "loss": 3.543,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "eval_accuracy": 0.3658964227231869,
+      "eval_loss": 3.5396640300750732,
+      "eval_runtime": 184.0061,
+      "eval_samples_per_second": 97.883,
+      "eval_steps_per_second": 6.119,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4851752021563343,
+      "grad_norm": 0.5835683345794678,
+      "learning_rate": 0.0004514344306529951,
+      "loss": 3.5382,
+      "step": 23050
+    },
+    {
+      "epoch": 2.490566037735849,
+      "grad_norm": 0.5643031001091003,
+      "learning_rate": 0.0004511106314085267,
+      "loss": 3.537,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4959568733153636,
+      "grad_norm": 0.5767573118209839,
+      "learning_rate": 0.00045078683216405826,
+      "loss": 3.5492,
+      "step": 23150
+    },
+    {
+      "epoch": 2.501347708894879,
+      "grad_norm": 0.5756224989891052,
+      "learning_rate": 0.00045046303291958976,
+      "loss": 3.5463,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5067385444743935,
+      "grad_norm": 0.6499918103218079,
+      "learning_rate": 0.0004501392336751214,
+      "loss": 3.561,
+      "step": 23250
+    },
+    {
+      "epoch": 2.512129380053908,
+      "grad_norm": 0.6103466153144836,
+      "learning_rate": 0.0004498154344306529,
+      "loss": 3.5578,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5175202156334233,
+      "grad_norm": 0.6318963170051575,
+      "learning_rate": 0.0004494916351861845,
+      "loss": 3.5593,
+      "step": 23350
+    },
+    {
+      "epoch": 2.522911051212938,
+      "grad_norm": 0.6232575178146362,
+      "learning_rate": 0.00044916783594171607,
+      "loss": 3.5548,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5283018867924527,
+      "grad_norm": 0.5690689086914062,
+      "learning_rate": 0.00044884403669724767,
+      "loss": 3.5452,
+      "step": 23450
+    },
+    {
+      "epoch": 2.533692722371968,
+      "grad_norm": 0.6115185618400574,
+      "learning_rate": 0.0004485202374527792,
+      "loss": 3.5415,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5390835579514826,
+      "grad_norm": 0.5683715343475342,
+      "learning_rate": 0.00044819643820831083,
+      "loss": 3.5434,
+      "step": 23550
+    },
+    {
+      "epoch": 2.5444743935309972,
+      "grad_norm": 0.6168181896209717,
+      "learning_rate": 0.0004478726389638424,
+      "loss": 3.5498,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5498652291105124,
+      "grad_norm": 0.6147149205207825,
+      "learning_rate": 0.00044754883971937393,
+      "loss": 3.5289,
+      "step": 23650
+    },
+    {
+      "epoch": 2.555256064690027,
+      "grad_norm": 0.5647293925285339,
+      "learning_rate": 0.00044722504047490553,
+      "loss": 3.5448,
+      "step": 23700
+    },
+    {
+      "epoch": 2.560646900269542,
+      "grad_norm": 0.6555283665657043,
+      "learning_rate": 0.0004469012412304371,
+      "loss": 3.5521,
+      "step": 23750
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 0.5926530361175537,
+      "learning_rate": 0.0004465774419859687,
+      "loss": 3.5618,
+      "step": 23800
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.6275022029876709,
+      "learning_rate": 0.00044625364274150024,
+      "loss": 3.5548,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5768194070080863,
+      "grad_norm": 0.5570287704467773,
+      "learning_rate": 0.00044592984349703184,
+      "loss": 3.5318,
+      "step": 23900
+    },
+    {
+      "epoch": 2.582210242587601,
+      "grad_norm": 0.5836941599845886,
+      "learning_rate": 0.00044560604425256334,
+      "loss": 3.5265,
+      "step": 23950
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "grad_norm": 0.546286404132843,
+      "learning_rate": 0.000445282245008095,
+      "loss": 3.5598,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "eval_accuracy": 0.36674695752899866,
+      "eval_loss": 3.528761386871338,
+      "eval_runtime": 184.0106,
+      "eval_samples_per_second": 97.88,
+      "eval_steps_per_second": 6.119,
+      "step": 24000
+    },
+    {
+      "epoch": 2.592991913746631,
+      "grad_norm": 0.5817551612854004,
+      "learning_rate": 0.0004449584457636265,
+      "loss": 3.554,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5983827493261455,
+      "grad_norm": 0.6376777291297913,
+      "learning_rate": 0.00044463464651915805,
+      "loss": 3.5397,
+      "step": 24100
+    },
+    {
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.6055445671081543,
+      "learning_rate": 0.00044431084727468965,
+      "loss": 3.5507,
+      "step": 24150
+    },
+    {
+      "epoch": 2.6091644204851754,
+      "grad_norm": 0.5456017851829529,
+      "learning_rate": 0.0004439870480302212,
+      "loss": 3.526,
+      "step": 24200
+    },
+    {
+      "epoch": 2.61455525606469,
+      "grad_norm": 0.6268299221992493,
+      "learning_rate": 0.0004436632487857528,
+      "loss": 3.5493,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6199460916442048,
+      "grad_norm": 0.6223170757293701,
+      "learning_rate": 0.00044333944954128436,
+      "loss": 3.5491,
+      "step": 24300
+    },
+    {
+      "epoch": 2.62533692722372,
+      "grad_norm": 0.580644428730011,
+      "learning_rate": 0.00044301565029681596,
+      "loss": 3.5434,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6307277628032346,
+      "grad_norm": 0.5493316054344177,
+      "learning_rate": 0.0004426918510523475,
+      "loss": 3.5352,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6361185983827493,
+      "grad_norm": 0.5446970462799072,
+      "learning_rate": 0.0004423680518078791,
+      "loss": 3.5151,
+      "step": 24450
+    },
+    {
+      "epoch": 2.641509433962264,
+      "grad_norm": 0.5663889646530151,
+      "learning_rate": 0.00044204425256341067,
+      "loss": 3.5352,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6469002695417787,
+      "grad_norm": 0.5662615895271301,
+      "learning_rate": 0.0004417204533189422,
+      "loss": 3.5313,
+      "step": 24550
+    },
+    {
+      "epoch": 2.652291105121294,
+      "grad_norm": 0.5774913430213928,
+      "learning_rate": 0.0004413966540744738,
+      "loss": 3.5433,
+      "step": 24600
+    },
+    {
+      "epoch": 2.6576819407008085,
+      "grad_norm": 0.5947896838188171,
+      "learning_rate": 0.0004410728548300053,
+      "loss": 3.5317,
+      "step": 24650
+    },
+    {
+      "epoch": 2.6630727762803232,
+      "grad_norm": 0.5944833159446716,
+      "learning_rate": 0.0004407490555855369,
+      "loss": 3.519,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6684636118598384,
+      "grad_norm": 0.6091347932815552,
+      "learning_rate": 0.0004404252563410685,
+      "loss": 3.5494,
+      "step": 24750
+    },
+    {
+      "epoch": 2.673854447439353,
+      "grad_norm": 0.6208272576332092,
+      "learning_rate": 0.0004401014570966001,
+      "loss": 3.5352,
+      "step": 24800
+    },
+    {
+      "epoch": 2.6792452830188678,
+      "grad_norm": 0.5658981204032898,
+      "learning_rate": 0.00043977765785213163,
+      "loss": 3.5444,
+      "step": 24850
+    },
+    {
+      "epoch": 2.684636118598383,
+      "grad_norm": 0.632360577583313,
+      "learning_rate": 0.0004394538586076632,
+      "loss": 3.5415,
+      "step": 24900
+    },
+    {
+      "epoch": 2.6900269541778976,
+      "grad_norm": 0.6172274351119995,
+      "learning_rate": 0.0004391300593631948,
+      "loss": 3.5341,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "grad_norm": 0.5555505156517029,
+      "learning_rate": 0.00043880626011872634,
+      "loss": 3.533,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "eval_accuracy": 0.3680360153900296,
+      "eval_loss": 3.5185306072235107,
+      "eval_runtime": 183.9436,
+      "eval_samples_per_second": 97.916,
+      "eval_steps_per_second": 6.121,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7008086253369274,
+      "grad_norm": 0.6439934968948364,
+      "learning_rate": 0.00043848246087425794,
+      "loss": 3.5493,
+      "step": 25050
+    },
+    {
+      "epoch": 2.706199460916442,
+      "grad_norm": 0.6363284587860107,
+      "learning_rate": 0.0004381586616297895,
+      "loss": 3.5325,
+      "step": 25100
+    },
+    {
+      "epoch": 2.711590296495957,
+      "grad_norm": 0.6599604487419128,
+      "learning_rate": 0.0004378348623853211,
+      "loss": 3.5416,
+      "step": 25150
+    },
+    {
+      "epoch": 2.7169811320754715,
+      "grad_norm": 0.626559317111969,
+      "learning_rate": 0.00043751106314085265,
+      "loss": 3.5485,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7223719676549867,
+      "grad_norm": 0.6159871816635132,
+      "learning_rate": 0.00043718726389638425,
+      "loss": 3.5455,
+      "step": 25250
+    },
+    {
+      "epoch": 2.7277628032345014,
+      "grad_norm": 0.6169484853744507,
+      "learning_rate": 0.00043686346465191575,
+      "loss": 3.5336,
+      "step": 25300
+    },
+    {
+      "epoch": 2.733153638814016,
+      "grad_norm": 0.6067425012588501,
+      "learning_rate": 0.0004365396654074473,
+      "loss": 3.5217,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7385444743935308,
+      "grad_norm": 0.5563640594482422,
+      "learning_rate": 0.0004362158661629789,
+      "loss": 3.5439,
+      "step": 25400
+    },
+    {
+      "epoch": 2.743935309973046,
+      "grad_norm": 0.7104645371437073,
+      "learning_rate": 0.00043589206691851045,
+      "loss": 3.5473,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7493261455525606,
+      "grad_norm": 0.5590559244155884,
+      "learning_rate": 0.00043556826767404206,
+      "loss": 3.5257,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.5929838418960571,
+      "learning_rate": 0.0004352444684295736,
+      "loss": 3.5374,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7601078167115904,
+      "grad_norm": 0.6350831985473633,
+      "learning_rate": 0.0004349206691851052,
+      "loss": 3.5381,
+      "step": 25600
+    },
+    {
+      "epoch": 2.765498652291105,
+      "grad_norm": 0.7433158755302429,
+      "learning_rate": 0.00043459686994063676,
+      "loss": 3.5282,
+      "step": 25650
+    },
+    {
+      "epoch": 2.77088948787062,
+      "grad_norm": 0.6278929114341736,
+      "learning_rate": 0.00043427307069616837,
+      "loss": 3.5368,
+      "step": 25700
+    },
+    {
+      "epoch": 2.776280323450135,
+      "grad_norm": 0.6056564450263977,
+      "learning_rate": 0.0004339492714516999,
+      "loss": 3.5388,
+      "step": 25750
+    },
+    {
+      "epoch": 2.7816711590296497,
+      "grad_norm": 0.6050384640693665,
+      "learning_rate": 0.00043362547220723147,
+      "loss": 3.5446,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7870619946091644,
+      "grad_norm": 0.5863341689109802,
+      "learning_rate": 0.0004333016729627631,
+      "loss": 3.5392,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7924528301886795,
+      "grad_norm": 0.6051446795463562,
+      "learning_rate": 0.0004329778737182946,
+      "loss": 3.5314,
+      "step": 25900
+    },
+    {
+      "epoch": 2.797843665768194,
+      "grad_norm": 0.5601605772972107,
+      "learning_rate": 0.00043265407447382623,
+      "loss": 3.5512,
+      "step": 25950
+    },
+    {
+      "epoch": 2.803234501347709,
+      "grad_norm": 0.6006129384040833,
+      "learning_rate": 0.0004323302752293577,
+      "loss": 3.5272,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803234501347709,
+      "eval_accuracy": 0.3687935433238722,
+      "eval_loss": 3.511807441711426,
+      "eval_runtime": 183.9921,
+      "eval_samples_per_second": 97.89,
+      "eval_steps_per_second": 6.12,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8086253369272236,
+      "grad_norm": 0.5731135606765747,
+      "learning_rate": 0.00043200647598488933,
+      "loss": 3.5367,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8140161725067383,
+      "grad_norm": 0.5950818061828613,
+      "learning_rate": 0.0004316826767404209,
+      "loss": 3.5272,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8194070080862534,
+      "grad_norm": 0.6465256810188293,
+      "learning_rate": 0.00043136535348084183,
+      "loss": 3.5273,
+      "step": 26150
+    },
+    {
+      "epoch": 2.824797843665768,
+      "grad_norm": 0.6396045684814453,
+      "learning_rate": 0.0004310415542363734,
+      "loss": 3.5309,
+      "step": 26200
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.575483500957489,
+      "learning_rate": 0.000430717754991905,
+      "loss": 3.5388,
+      "step": 26250
+    },
+    {
+      "epoch": 2.835579514824798,
+      "grad_norm": 0.5966452360153198,
+      "learning_rate": 0.00043039395574743654,
+      "loss": 3.523,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8409703504043127,
+      "grad_norm": 0.5425668954849243,
+      "learning_rate": 0.00043007015650296814,
+      "loss": 3.5356,
+      "step": 26350
+    },
+    {
+      "epoch": 2.8463611859838274,
+      "grad_norm": 0.587317943572998,
+      "learning_rate": 0.0004297463572584997,
+      "loss": 3.5102,
+      "step": 26400
+    },
+    {
+      "epoch": 2.8517520215633425,
+      "grad_norm": 0.578007161617279,
+      "learning_rate": 0.0004294225580140313,
+      "loss": 3.542,
+      "step": 26450
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.5723671913146973,
+      "learning_rate": 0.00042909875876956285,
+      "loss": 3.5329,
+      "step": 26500
+    },
+    {
+      "epoch": 2.862533692722372,
+      "grad_norm": 0.5588173866271973,
+      "learning_rate": 0.0004287749595250944,
+      "loss": 3.5404,
+      "step": 26550
+    },
+    {
+      "epoch": 2.867924528301887,
+      "grad_norm": 0.5535792112350464,
+      "learning_rate": 0.000428451160280626,
+      "loss": 3.5266,
+      "step": 26600
+    },
+    {
+      "epoch": 2.8733153638814017,
+      "grad_norm": 0.573268711566925,
+      "learning_rate": 0.0004281273610361575,
+      "loss": 3.5303,
+      "step": 26650
+    },
+    {
+      "epoch": 2.8787061994609164,
+      "grad_norm": 0.5750014185905457,
+      "learning_rate": 0.0004278035617916891,
+      "loss": 3.5032,
+      "step": 26700
+    },
+    {
+      "epoch": 2.884097035040431,
+      "grad_norm": 0.6313278079032898,
+      "learning_rate": 0.00042747976254722066,
+      "loss": 3.5281,
+      "step": 26750
+    },
+    {
+      "epoch": 2.889487870619946,
+      "grad_norm": 0.5894135236740112,
+      "learning_rate": 0.00042715596330275226,
+      "loss": 3.5066,
+      "step": 26800
+    },
+    {
+      "epoch": 2.894878706199461,
+      "grad_norm": 0.5986575484275818,
+      "learning_rate": 0.0004268321640582838,
+      "loss": 3.5441,
+      "step": 26850
+    },
+    {
+      "epoch": 2.9002695417789757,
+      "grad_norm": 0.5755992531776428,
+      "learning_rate": 0.0004265083648138154,
+      "loss": 3.5257,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.6677845120429993,
+      "learning_rate": 0.00042618456556934697,
+      "loss": 3.544,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "grad_norm": 0.6155284643173218,
+      "learning_rate": 0.0004258607663248785,
+      "loss": 3.5211,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "eval_accuracy": 0.3694859881779139,
+      "eval_loss": 3.50223445892334,
+      "eval_runtime": 183.9135,
+      "eval_samples_per_second": 97.932,
+      "eval_steps_per_second": 6.122,
+      "step": 27000
+    },
+    {
+      "epoch": 2.91644204851752,
+      "grad_norm": 0.6333293318748474,
+      "learning_rate": 0.0004255369670804101,
+      "loss": 3.5293,
+      "step": 27050
+    },
+    {
+      "epoch": 2.921832884097035,
+      "grad_norm": 0.5835155248641968,
+      "learning_rate": 0.0004252131678359417,
+      "loss": 3.5198,
+      "step": 27100
+    },
+    {
+      "epoch": 2.92722371967655,
+      "grad_norm": 0.5927255749702454,
+      "learning_rate": 0.0004248893685914733,
+      "loss": 3.5244,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9326145552560647,
+      "grad_norm": 0.6329957842826843,
+      "learning_rate": 0.00042456556934700483,
+      "loss": 3.5264,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9380053908355794,
+      "grad_norm": 0.6049855351448059,
+      "learning_rate": 0.00042424177010253643,
+      "loss": 3.5223,
+      "step": 27250
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 0.5478748679161072,
+      "learning_rate": 0.00042391797085806793,
+      "loss": 3.52,
+      "step": 27300
+    },
+    {
+      "epoch": 2.9487870619946093,
+      "grad_norm": 0.599579930305481,
+      "learning_rate": 0.0004235941716135995,
+      "loss": 3.5254,
+      "step": 27350
+    },
+    {
+      "epoch": 2.954177897574124,
+      "grad_norm": 0.597079873085022,
+      "learning_rate": 0.0004232703723691311,
+      "loss": 3.5192,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9595687331536387,
+      "grad_norm": 0.5469141602516174,
+      "learning_rate": 0.00042294657312466264,
+      "loss": 3.5294,
+      "step": 27450
+    },
+    {
+      "epoch": 2.964959568733154,
+      "grad_norm": 0.6046788692474365,
+      "learning_rate": 0.00042262277388019424,
+      "loss": 3.5236,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9703504043126685,
+      "grad_norm": 0.5884106159210205,
+      "learning_rate": 0.0004222989746357258,
+      "loss": 3.524,
+      "step": 27550
+    },
+    {
+      "epoch": 2.975741239892183,
+      "grad_norm": 0.634127676486969,
+      "learning_rate": 0.0004219751753912574,
+      "loss": 3.5287,
+      "step": 27600
+    },
+    {
+      "epoch": 2.981132075471698,
+      "grad_norm": 0.5886887311935425,
+      "learning_rate": 0.00042165137614678895,
+      "loss": 3.5248,
+      "step": 27650
+    },
+    {
+      "epoch": 2.986522911051213,
+      "grad_norm": 0.5682351589202881,
+      "learning_rate": 0.00042132757690232055,
+      "loss": 3.5282,
+      "step": 27700
+    },
+    {
+      "epoch": 2.9919137466307277,
+      "grad_norm": 0.5956007242202759,
+      "learning_rate": 0.0004210037776578521,
+      "loss": 3.5301,
+      "step": 27750
+    },
+    {
+      "epoch": 2.9973045822102424,
+      "grad_norm": 0.6513469815254211,
+      "learning_rate": 0.00042067997841338365,
+      "loss": 3.499,
+      "step": 27800
+    },
+    {
+      "epoch": 3.0026954177897576,
+      "grad_norm": 0.5566192865371704,
+      "learning_rate": 0.00042035617916891526,
+      "loss": 3.4543,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0080862533692723,
+      "grad_norm": 0.6176077127456665,
+      "learning_rate": 0.0004200323799244468,
+      "loss": 3.423,
+      "step": 27900
+    },
+    {
+      "epoch": 3.013477088948787,
+      "grad_norm": 0.6107637286186218,
+      "learning_rate": 0.0004197085806799784,
+      "loss": 3.4323,
+      "step": 27950
+    },
+    {
+      "epoch": 3.018867924528302,
+      "grad_norm": 0.6046628355979919,
+      "learning_rate": 0.0004193847814355099,
+      "loss": 3.4424,
+      "step": 28000
+    },
+    {
+      "epoch": 3.018867924528302,
+      "eval_accuracy": 0.37057208244450746,
+      "eval_loss": 3.4983136653900146,
+      "eval_runtime": 183.6916,
+      "eval_samples_per_second": 98.05,
+      "eval_steps_per_second": 6.13,
+      "step": 28000
+    },
+    {
+      "epoch": 3.024258760107817,
+      "grad_norm": 0.6203792095184326,
+      "learning_rate": 0.0004190609821910415,
+      "loss": 3.4246,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0296495956873315,
+      "grad_norm": 0.5849449038505554,
+      "learning_rate": 0.00041873718294657306,
+      "loss": 3.413,
+      "step": 28100
+    },
+    {
+      "epoch": 3.035040431266846,
+      "grad_norm": 0.6401404142379761,
+      "learning_rate": 0.00041841338370210467,
+      "loss": 3.4293,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0404312668463613,
+      "grad_norm": 0.5675318241119385,
+      "learning_rate": 0.00041809606044252557,
+      "loss": 3.4324,
+      "step": 28200
+    },
+    {
+      "epoch": 3.045822102425876,
+      "grad_norm": 0.5723515152931213,
+      "learning_rate": 0.00041777226119805717,
+      "loss": 3.4211,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0512129380053907,
+      "grad_norm": 0.5614615678787231,
+      "learning_rate": 0.0004174484619535887,
+      "loss": 3.4325,
+      "step": 28300
+    },
+    {
+      "epoch": 3.056603773584906,
+      "grad_norm": 0.5907319784164429,
+      "learning_rate": 0.00041712466270912033,
+      "loss": 3.4343,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0619946091644206,
+      "grad_norm": 0.6139299869537354,
+      "learning_rate": 0.0004168008634646519,
+      "loss": 3.4437,
+      "step": 28400
+    },
+    {
+      "epoch": 3.0673854447439353,
+      "grad_norm": 0.6088677048683167,
+      "learning_rate": 0.0004164770642201835,
+      "loss": 3.4369,
+      "step": 28450
+    },
+    {
+      "epoch": 3.07277628032345,
+      "grad_norm": 0.6208965182304382,
+      "learning_rate": 0.00041615326497571503,
+      "loss": 3.4344,
+      "step": 28500
+    },
+    {
+      "epoch": 3.078167115902965,
+      "grad_norm": 0.6358165144920349,
+      "learning_rate": 0.0004158294657312466,
+      "loss": 3.4428,
+      "step": 28550
+    },
+    {
+      "epoch": 3.08355795148248,
+      "grad_norm": 0.6254170536994934,
+      "learning_rate": 0.0004155056664867782,
+      "loss": 3.4463,
+      "step": 28600
+    },
+    {
+      "epoch": 3.0889487870619945,
+      "grad_norm": 0.585701584815979,
+      "learning_rate": 0.0004151818672423097,
+      "loss": 3.4471,
+      "step": 28650
+    },
+    {
+      "epoch": 3.0943396226415096,
+      "grad_norm": 0.6081207394599915,
+      "learning_rate": 0.0004148580679978413,
+      "loss": 3.4386,
+      "step": 28700
+    },
+    {
+      "epoch": 3.0997304582210243,
+      "grad_norm": 0.6378210186958313,
+      "learning_rate": 0.00041453426875337284,
+      "loss": 3.4258,
+      "step": 28750
+    },
+    {
+      "epoch": 3.105121293800539,
+      "grad_norm": 0.6246366500854492,
+      "learning_rate": 0.00041421046950890445,
+      "loss": 3.4211,
+      "step": 28800
+    },
+    {
+      "epoch": 3.1105121293800537,
+      "grad_norm": 0.5723994374275208,
+      "learning_rate": 0.000413886670264436,
+      "loss": 3.4276,
+      "step": 28850
+    },
+    {
+      "epoch": 3.115902964959569,
+      "grad_norm": 0.6091293692588806,
+      "learning_rate": 0.0004135628710199676,
+      "loss": 3.4424,
+      "step": 28900
+    },
+    {
+      "epoch": 3.1212938005390836,
+      "grad_norm": 0.5650001168251038,
+      "learning_rate": 0.00041323907177549915,
+      "loss": 3.4413,
+      "step": 28950
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "grad_norm": 0.6242942214012146,
+      "learning_rate": 0.0004129152725310307,
+      "loss": 3.4427,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "eval_accuracy": 0.37073506177622917,
+      "eval_loss": 3.4975240230560303,
+      "eval_runtime": 183.9334,
+      "eval_samples_per_second": 97.921,
+      "eval_steps_per_second": 6.122,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1320754716981134,
+      "grad_norm": 0.5982543230056763,
+      "learning_rate": 0.0004125914732865623,
+      "loss": 3.4407,
+      "step": 29050
+    },
+    {
+      "epoch": 3.137466307277628,
+      "grad_norm": 0.584873616695404,
+      "learning_rate": 0.00041226767404209386,
+      "loss": 3.4539,
+      "step": 29100
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.6070249676704407,
+      "learning_rate": 0.00041194387479762546,
+      "loss": 3.4543,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1482479784366575,
+      "grad_norm": 0.6300134658813477,
+      "learning_rate": 0.000411620075553157,
+      "loss": 3.4457,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1536388140161726,
+      "grad_norm": 0.6087648868560791,
+      "learning_rate": 0.0004112962763086886,
+      "loss": 3.4461,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1590296495956873,
+      "grad_norm": 0.6131951808929443,
+      "learning_rate": 0.0004109724770642201,
+      "loss": 3.4512,
+      "step": 29300
+    },
+    {
+      "epoch": 3.164420485175202,
+      "grad_norm": 0.6422128677368164,
+      "learning_rate": 0.00041064867781975177,
+      "loss": 3.4648,
+      "step": 29350
+    },
+    {
+      "epoch": 3.169811320754717,
+      "grad_norm": 0.7360274791717529,
+      "learning_rate": 0.00041032487857528327,
+      "loss": 3.4428,
+      "step": 29400
+    },
+    {
+      "epoch": 3.175202156334232,
+      "grad_norm": 0.5676162838935852,
+      "learning_rate": 0.0004100010793308148,
+      "loss": 3.4475,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1805929919137466,
+      "grad_norm": 0.5553027987480164,
+      "learning_rate": 0.0004096772800863464,
+      "loss": 3.4486,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1859838274932613,
+      "grad_norm": 0.6182715892791748,
+      "learning_rate": 0.000409353480841878,
+      "loss": 3.4488,
+      "step": 29550
+    },
+    {
+      "epoch": 3.1913746630727764,
+      "grad_norm": 0.5933767557144165,
+      "learning_rate": 0.0004090296815974096,
+      "loss": 3.4744,
+      "step": 29600
+    },
+    {
+      "epoch": 3.196765498652291,
+      "grad_norm": 0.5704579949378967,
+      "learning_rate": 0.00040870588235294113,
+      "loss": 3.4456,
+      "step": 29650
+    },
+    {
+      "epoch": 3.202156334231806,
+      "grad_norm": 0.6044793128967285,
+      "learning_rate": 0.00040838208310847273,
+      "loss": 3.455,
+      "step": 29700
+    },
+    {
+      "epoch": 3.207547169811321,
+      "grad_norm": 0.5789558291435242,
+      "learning_rate": 0.0004080582838640043,
+      "loss": 3.4687,
+      "step": 29750
+    },
+    {
+      "epoch": 3.2129380053908356,
+      "grad_norm": 0.6031609177589417,
+      "learning_rate": 0.0004077344846195359,
+      "loss": 3.4433,
+      "step": 29800
+    },
+    {
+      "epoch": 3.2183288409703503,
+      "grad_norm": 0.6147695183753967,
+      "learning_rate": 0.00040741068537506744,
+      "loss": 3.4398,
+      "step": 29850
+    },
+    {
+      "epoch": 3.223719676549865,
+      "grad_norm": 0.6561545133590698,
+      "learning_rate": 0.000407086886130599,
+      "loss": 3.4364,
+      "step": 29900
+    },
+    {
+      "epoch": 3.22911051212938,
+      "grad_norm": 0.6324929594993591,
+      "learning_rate": 0.0004067630868861306,
+      "loss": 3.4486,
+      "step": 29950
+    },
+    {
+      "epoch": 3.234501347708895,
+      "grad_norm": 0.6108676195144653,
+      "learning_rate": 0.0004064392876416621,
+      "loss": 3.4427,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234501347708895,
+      "eval_accuracy": 0.3712915818676149,
+      "eval_loss": 3.490433931350708,
+      "eval_runtime": 183.7094,
+      "eval_samples_per_second": 98.041,
+      "eval_steps_per_second": 6.129,
+      "step": 30000
+    },
+    {
+      "epoch": 3.2398921832884096,
+      "grad_norm": 0.627756655216217,
+      "learning_rate": 0.0004061154883971937,
+      "loss": 3.4489,
+      "step": 30050
+    },
+    {
+      "epoch": 3.2452830188679247,
+      "grad_norm": 0.6552205681800842,
+      "learning_rate": 0.00040579168915272525,
+      "loss": 3.4595,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2506738544474394,
+      "grad_norm": 0.605854868888855,
+      "learning_rate": 0.00040546788990825685,
+      "loss": 3.4753,
+      "step": 30150
+    },
+    {
+      "epoch": 3.256064690026954,
+      "grad_norm": 0.6816122531890869,
+      "learning_rate": 0.0004051440906637884,
+      "loss": 3.4714,
+      "step": 30200
+    },
+    {
+      "epoch": 3.2614555256064692,
+      "grad_norm": 0.6047664880752563,
+      "learning_rate": 0.00040482029141931995,
+      "loss": 3.458,
+      "step": 30250
+    },
+    {
+      "epoch": 3.266846361185984,
+      "grad_norm": 0.6369194984436035,
+      "learning_rate": 0.00040449649217485156,
+      "loss": 3.4702,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2722371967654986,
+      "grad_norm": 0.5737183690071106,
+      "learning_rate": 0.0004041791689152725,
+      "loss": 3.4396,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2776280323450133,
+      "grad_norm": 0.6164860725402832,
+      "learning_rate": 0.00040385536967080406,
+      "loss": 3.443,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2830188679245285,
+      "grad_norm": 0.5741621255874634,
+      "learning_rate": 0.00040353157042633567,
+      "loss": 3.4596,
+      "step": 30450
+    },
+    {
+      "epoch": 3.288409703504043,
+      "grad_norm": 0.6739873886108398,
+      "learning_rate": 0.0004032077711818672,
+      "loss": 3.4586,
+      "step": 30500
+    },
+    {
+      "epoch": 3.293800539083558,
+      "grad_norm": 0.6138100624084473,
+      "learning_rate": 0.0004028839719373988,
+      "loss": 3.4622,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2991913746630726,
+      "grad_norm": 0.6170074939727783,
+      "learning_rate": 0.00040256017269293037,
+      "loss": 3.4439,
+      "step": 30600
+    },
+    {
+      "epoch": 3.3045822102425877,
+      "grad_norm": 0.6086063981056213,
+      "learning_rate": 0.00040223637344846187,
+      "loss": 3.4403,
+      "step": 30650
+    },
+    {
+      "epoch": 3.3099730458221024,
+      "grad_norm": 0.6176971793174744,
+      "learning_rate": 0.00040191257420399347,
+      "loss": 3.4583,
+      "step": 30700
+    },
+    {
+      "epoch": 3.315363881401617,
+      "grad_norm": 0.6030409336090088,
+      "learning_rate": 0.000401588774959525,
+      "loss": 3.4435,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3207547169811322,
+      "grad_norm": 0.5938195586204529,
+      "learning_rate": 0.00040126497571505663,
+      "loss": 3.457,
+      "step": 30800
+    },
+    {
+      "epoch": 3.326145552560647,
+      "grad_norm": 0.6070782542228699,
+      "learning_rate": 0.0004009411764705882,
+      "loss": 3.4617,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3315363881401616,
+      "grad_norm": 0.6172419190406799,
+      "learning_rate": 0.0004006173772261198,
+      "loss": 3.4423,
+      "step": 30900
+    },
+    {
+      "epoch": 3.3369272237196768,
+      "grad_norm": 0.5756070017814636,
+      "learning_rate": 0.00040029357798165133,
+      "loss": 3.4723,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "grad_norm": 0.7206381559371948,
+      "learning_rate": 0.00039996977873718294,
+      "loss": 3.445,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "eval_accuracy": 0.37233475824352175,
+      "eval_loss": 3.4826910495758057,
+      "eval_runtime": 184.0439,
+      "eval_samples_per_second": 97.862,
+      "eval_steps_per_second": 6.118,
+      "step": 31000
+    },
+    {
+      "epoch": 3.347708894878706,
+      "grad_norm": 0.6098358631134033,
+      "learning_rate": 0.0003996459794927145,
+      "loss": 3.4583,
+      "step": 31050
+    },
+    {
+      "epoch": 3.353099730458221,
+      "grad_norm": 0.625518798828125,
+      "learning_rate": 0.00039932218024824604,
+      "loss": 3.4487,
+      "step": 31100
+    },
+    {
+      "epoch": 3.358490566037736,
+      "grad_norm": 0.5937629342079163,
+      "learning_rate": 0.00039899838100377764,
+      "loss": 3.4645,
+      "step": 31150
+    },
+    {
+      "epoch": 3.3638814016172507,
+      "grad_norm": 0.5838909149169922,
+      "learning_rate": 0.0003986745817593092,
+      "loss": 3.4557,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3692722371967654,
+      "grad_norm": 0.6747406721115112,
+      "learning_rate": 0.0003983507825148408,
+      "loss": 3.4407,
+      "step": 31250
+    },
+    {
+      "epoch": 3.37466307277628,
+      "grad_norm": 0.6010386943817139,
+      "learning_rate": 0.00039802698327037235,
+      "loss": 3.4678,
+      "step": 31300
+    },
+    {
+      "epoch": 3.3800539083557952,
+      "grad_norm": 0.6182563900947571,
+      "learning_rate": 0.00039770318402590396,
+      "loss": 3.4691,
+      "step": 31350
+    },
+    {
+      "epoch": 3.38544474393531,
+      "grad_norm": 0.5988255143165588,
+      "learning_rate": 0.00039737938478143545,
+      "loss": 3.4774,
+      "step": 31400
+    },
+    {
+      "epoch": 3.3908355795148246,
+      "grad_norm": 0.5775631666183472,
+      "learning_rate": 0.000397055585536967,
+      "loss": 3.4581,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3962264150943398,
+      "grad_norm": 0.6162555813789368,
+      "learning_rate": 0.0003967317862924986,
+      "loss": 3.4486,
+      "step": 31500
+    },
+    {
+      "epoch": 3.4016172506738545,
+      "grad_norm": 0.6046192646026611,
+      "learning_rate": 0.00039640798704803016,
+      "loss": 3.4534,
+      "step": 31550
+    },
+    {
+      "epoch": 3.407008086253369,
+      "grad_norm": 0.6041988730430603,
+      "learning_rate": 0.00039608418780356176,
+      "loss": 3.4575,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4123989218328843,
+      "grad_norm": 0.623914897441864,
+      "learning_rate": 0.0003957603885590933,
+      "loss": 3.4494,
+      "step": 31650
+    },
+    {
+      "epoch": 3.417789757412399,
+      "grad_norm": 0.5987055897712708,
+      "learning_rate": 0.0003954365893146249,
+      "loss": 3.4583,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4231805929919137,
+      "grad_norm": 0.5991863012313843,
+      "learning_rate": 0.00039511279007015647,
+      "loss": 3.4641,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4285714285714284,
+      "grad_norm": 0.6010271310806274,
+      "learning_rate": 0.00039478899082568807,
+      "loss": 3.4433,
+      "step": 31800
+    },
+    {
+      "epoch": 3.4339622641509435,
+      "grad_norm": 0.5700473189353943,
+      "learning_rate": 0.0003944651915812196,
+      "loss": 3.4547,
+      "step": 31850
+    },
+    {
+      "epoch": 3.439353099730458,
+      "grad_norm": 0.5874118208885193,
+      "learning_rate": 0.0003941413923367512,
+      "loss": 3.47,
+      "step": 31900
+    },
+    {
+      "epoch": 3.444743935309973,
+      "grad_norm": 0.5711833834648132,
+      "learning_rate": 0.0003938175930922828,
+      "loss": 3.4649,
+      "step": 31950
+    },
+    {
+      "epoch": 3.450134770889488,
+      "grad_norm": 0.6957682967185974,
+      "learning_rate": 0.0003934937938478143,
+      "loss": 3.4717,
+      "step": 32000
+    },
+    {
+      "epoch": 3.450134770889488,
+      "eval_accuracy": 0.37300981863551314,
+      "eval_loss": 3.476459264755249,
+      "eval_runtime": 183.7361,
+      "eval_samples_per_second": 98.026,
+      "eval_steps_per_second": 6.128,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4555256064690028,
+      "grad_norm": 0.5690134167671204,
+      "learning_rate": 0.0003931699946033459,
+      "loss": 3.4463,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4609164420485174,
+      "grad_norm": 0.6350628733634949,
+      "learning_rate": 0.00039284619535887743,
+      "loss": 3.4583,
+      "step": 32100
+    },
+    {
+      "epoch": 3.466307277628032,
+      "grad_norm": 0.5968349575996399,
+      "learning_rate": 0.00039252239611440904,
+      "loss": 3.459,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4716981132075473,
+      "grad_norm": 0.6233981251716614,
+      "learning_rate": 0.0003921985968699406,
+      "loss": 3.4794,
+      "step": 32200
+    },
+    {
+      "epoch": 3.477088948787062,
+      "grad_norm": 0.6959568858146667,
+      "learning_rate": 0.0003918747976254722,
+      "loss": 3.4749,
+      "step": 32250
+    },
+    {
+      "epoch": 3.4824797843665767,
+      "grad_norm": 0.6333785653114319,
+      "learning_rate": 0.00039155099838100374,
+      "loss": 3.4438,
+      "step": 32300
+    },
+    {
+      "epoch": 3.487870619946092,
+      "grad_norm": 0.6264799237251282,
+      "learning_rate": 0.0003912336751214247,
+      "loss": 3.4413,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4932614555256065,
+      "grad_norm": 0.6589958071708679,
+      "learning_rate": 0.00039090987587695624,
+      "loss": 3.4783,
+      "step": 32400
+    },
+    {
+      "epoch": 3.498652291105121,
+      "grad_norm": 0.5954399704933167,
+      "learning_rate": 0.00039058607663248785,
+      "loss": 3.4627,
+      "step": 32450
+    },
+    {
+      "epoch": 3.5040431266846364,
+      "grad_norm": 0.6438902616500854,
+      "learning_rate": 0.0003902622773880194,
+      "loss": 3.4567,
+      "step": 32500
+    },
+    {
+      "epoch": 3.509433962264151,
+      "grad_norm": 0.7091028690338135,
+      "learning_rate": 0.000389938478143551,
+      "loss": 3.4599,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5148247978436657,
+      "grad_norm": 0.625196099281311,
+      "learning_rate": 0.00038961467889908255,
+      "loss": 3.4453,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5202156334231804,
+      "grad_norm": 0.6431267857551575,
+      "learning_rate": 0.00038929087965461405,
+      "loss": 3.453,
+      "step": 32650
+    },
+    {
+      "epoch": 3.525606469002695,
+      "grad_norm": 0.6067655682563782,
+      "learning_rate": 0.00038896708041014566,
+      "loss": 3.4535,
+      "step": 32700
+    },
+    {
+      "epoch": 3.5309973045822103,
+      "grad_norm": 0.5781969428062439,
+      "learning_rate": 0.0003886432811656772,
+      "loss": 3.4454,
+      "step": 32750
+    },
+    {
+      "epoch": 3.536388140161725,
+      "grad_norm": 0.6490238904953003,
+      "learning_rate": 0.0003883194819212088,
+      "loss": 3.4601,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5417789757412397,
+      "grad_norm": 0.6111918091773987,
+      "learning_rate": 0.00038799568267674036,
+      "loss": 3.4692,
+      "step": 32850
+    },
+    {
+      "epoch": 3.547169811320755,
+      "grad_norm": 0.6645634174346924,
+      "learning_rate": 0.00038767188343227197,
+      "loss": 3.4725,
+      "step": 32900
+    },
+    {
+      "epoch": 3.5525606469002695,
+      "grad_norm": 0.9656093716621399,
+      "learning_rate": 0.0003873480841878035,
+      "loss": 3.4587,
+      "step": 32950
+    },
+    {
+      "epoch": 3.557951482479784,
+      "grad_norm": 0.629120409488678,
+      "learning_rate": 0.0003870242849433351,
+      "loss": 3.4426,
+      "step": 33000
+    },
+    {
+      "epoch": 3.557951482479784,
+      "eval_accuracy": 0.37355123597549267,
+      "eval_loss": 3.4684126377105713,
+      "eval_runtime": 184.2402,
+      "eval_samples_per_second": 97.758,
+      "eval_steps_per_second": 6.112,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5633423180592994,
+      "grad_norm": 0.5992081165313721,
+      "learning_rate": 0.00038670048569886667,
+      "loss": 3.4547,
+      "step": 33050
+    },
+    {
+      "epoch": 3.568733153638814,
+      "grad_norm": 0.6156100630760193,
+      "learning_rate": 0.0003863766864543982,
+      "loss": 3.466,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5741239892183287,
+      "grad_norm": 0.6006748080253601,
+      "learning_rate": 0.00038605288720992983,
+      "loss": 3.4431,
+      "step": 33150
+    },
+    {
+      "epoch": 3.579514824797844,
+      "grad_norm": 0.6171281933784485,
+      "learning_rate": 0.0003857290879654614,
+      "loss": 3.4524,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5849056603773586,
+      "grad_norm": 0.781915545463562,
+      "learning_rate": 0.000385405288720993,
+      "loss": 3.4424,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5902964959568733,
+      "grad_norm": 0.6024929285049438,
+      "learning_rate": 0.00038508148947652453,
+      "loss": 3.4464,
+      "step": 33300
+    },
+    {
+      "epoch": 3.595687331536388,
+      "grad_norm": 0.6088691353797913,
+      "learning_rate": 0.00038475769023205614,
+      "loss": 3.4655,
+      "step": 33350
+    },
+    {
+      "epoch": 3.601078167115903,
+      "grad_norm": 0.6233068704605103,
+      "learning_rate": 0.00038443389098758763,
+      "loss": 3.4556,
+      "step": 33400
+    },
+    {
+      "epoch": 3.606469002695418,
+      "grad_norm": 0.6492874026298523,
+      "learning_rate": 0.00038411009174311924,
+      "loss": 3.4574,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6118598382749325,
+      "grad_norm": 0.588290810585022,
+      "learning_rate": 0.0003837862924986508,
+      "loss": 3.4696,
+      "step": 33500
+    },
+    {
+      "epoch": 3.617250673854447,
+      "grad_norm": 0.6026934385299683,
+      "learning_rate": 0.00038346249325418234,
+      "loss": 3.4575,
+      "step": 33550
+    },
+    {
+      "epoch": 3.6226415094339623,
+      "grad_norm": 0.635048508644104,
+      "learning_rate": 0.00038313869400971395,
+      "loss": 3.4538,
+      "step": 33600
+    },
+    {
+      "epoch": 3.628032345013477,
+      "grad_norm": 0.592736005783081,
+      "learning_rate": 0.0003828148947652455,
+      "loss": 3.4527,
+      "step": 33650
+    },
+    {
+      "epoch": 3.6334231805929917,
+      "grad_norm": 0.6240489482879639,
+      "learning_rate": 0.0003824910955207771,
+      "loss": 3.4532,
+      "step": 33700
+    },
+    {
+      "epoch": 3.638814016172507,
+      "grad_norm": 0.6334491968154907,
+      "learning_rate": 0.00038216729627630865,
+      "loss": 3.4497,
+      "step": 33750
+    },
+    {
+      "epoch": 3.6442048517520216,
+      "grad_norm": 0.6171436309814453,
+      "learning_rate": 0.00038184349703184026,
+      "loss": 3.4446,
+      "step": 33800
+    },
+    {
+      "epoch": 3.6495956873315363,
+      "grad_norm": 0.6213189363479614,
+      "learning_rate": 0.0003815196977873718,
+      "loss": 3.4504,
+      "step": 33850
+    },
+    {
+      "epoch": 3.6549865229110514,
+      "grad_norm": 0.5768414735794067,
+      "learning_rate": 0.0003811958985429034,
+      "loss": 3.4571,
+      "step": 33900
+    },
+    {
+      "epoch": 3.660377358490566,
+      "grad_norm": 0.5723716616630554,
+      "learning_rate": 0.00038087209929843496,
+      "loss": 3.4619,
+      "step": 33950
+    },
+    {
+      "epoch": 3.665768194070081,
+      "grad_norm": 0.5816717147827148,
+      "learning_rate": 0.00038054830005396646,
+      "loss": 3.4603,
+      "step": 34000
+    },
+    {
+      "epoch": 3.665768194070081,
+      "eval_accuracy": 0.37412351073561156,
+      "eval_loss": 3.4615986347198486,
+      "eval_runtime": 183.9553,
+      "eval_samples_per_second": 97.91,
+      "eval_steps_per_second": 6.121,
+      "step": 34000
+    },
+    {
+      "epoch": 3.671159029649596,
+      "grad_norm": 0.5690119862556458,
+      "learning_rate": 0.00038022450080949806,
+      "loss": 3.4572,
+      "step": 34050
+    },
+    {
+      "epoch": 3.6765498652291106,
+      "grad_norm": 0.6545220613479614,
+      "learning_rate": 0.0003799007015650296,
+      "loss": 3.4421,
+      "step": 34100
+    },
+    {
+      "epoch": 3.6819407008086253,
+      "grad_norm": 0.6318289041519165,
+      "learning_rate": 0.0003795769023205612,
+      "loss": 3.4508,
+      "step": 34150
+    },
+    {
+      "epoch": 3.68733153638814,
+      "grad_norm": 0.6190627217292786,
+      "learning_rate": 0.00037925310307609277,
+      "loss": 3.456,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6927223719676547,
+      "grad_norm": 0.5709021687507629,
+      "learning_rate": 0.0003789293038316244,
+      "loss": 3.4468,
+      "step": 34250
+    },
+    {
+      "epoch": 3.69811320754717,
+      "grad_norm": 0.6180399060249329,
+      "learning_rate": 0.0003786055045871559,
+      "loss": 3.4551,
+      "step": 34300
+    },
+    {
+      "epoch": 3.7035040431266846,
+      "grad_norm": 0.6174043416976929,
+      "learning_rate": 0.0003782817053426875,
+      "loss": 3.4556,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7088948787061993,
+      "grad_norm": 0.6184794902801514,
+      "learning_rate": 0.0003779579060982191,
+      "loss": 3.4495,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7142857142857144,
+      "grad_norm": 0.6114649176597595,
+      "learning_rate": 0.00037763410685375063,
+      "loss": 3.4795,
+      "step": 34450
+    },
+    {
+      "epoch": 3.719676549865229,
+      "grad_norm": 0.6328994035720825,
+      "learning_rate": 0.00037731030760928223,
+      "loss": 3.4406,
+      "step": 34500
+    },
+    {
+      "epoch": 3.725067385444744,
+      "grad_norm": 0.6460617184638977,
+      "learning_rate": 0.0003769929843497032,
+      "loss": 3.45,
+      "step": 34550
+    },
+    {
+      "epoch": 3.730458221024259,
+      "grad_norm": 0.7519969344139099,
+      "learning_rate": 0.00037666918510523474,
+      "loss": 3.4476,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7358490566037736,
+      "grad_norm": 0.591691792011261,
+      "learning_rate": 0.00037634538586076634,
+      "loss": 3.4637,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7412398921832883,
+      "grad_norm": 0.6448811292648315,
+      "learning_rate": 0.00037602158661629784,
+      "loss": 3.449,
+      "step": 34700
+    },
+    {
+      "epoch": 3.7466307277628035,
+      "grad_norm": 0.6556087136268616,
+      "learning_rate": 0.0003756977873718294,
+      "loss": 3.4546,
+      "step": 34750
+    },
+    {
+      "epoch": 3.752021563342318,
+      "grad_norm": 0.6165753602981567,
+      "learning_rate": 0.000375373988127361,
+      "loss": 3.4493,
+      "step": 34800
+    },
+    {
+      "epoch": 3.757412398921833,
+      "grad_norm": 0.6732285618782043,
+      "learning_rate": 0.00037505018888289254,
+      "loss": 3.4393,
+      "step": 34850
+    },
+    {
+      "epoch": 3.7628032345013476,
+      "grad_norm": 0.6393305063247681,
+      "learning_rate": 0.00037472638963842415,
+      "loss": 3.4489,
+      "step": 34900
+    },
+    {
+      "epoch": 3.7681940700808623,
+      "grad_norm": 0.6352855563163757,
+      "learning_rate": 0.0003744025903939557,
+      "loss": 3.4633,
+      "step": 34950
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "grad_norm": 0.6707683205604553,
+      "learning_rate": 0.0003740787911494873,
+      "loss": 3.4463,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "eval_accuracy": 0.3749070067096418,
+      "eval_loss": 3.4561455249786377,
+      "eval_runtime": 183.9041,
+      "eval_samples_per_second": 97.937,
+      "eval_steps_per_second": 6.123,
+      "step": 35000
+    },
+    {
+      "epoch": 3.778975741239892,
+      "grad_norm": 0.6262804269790649,
+      "learning_rate": 0.00037375499190501885,
+      "loss": 3.4567,
+      "step": 35050
+    },
+    {
+      "epoch": 3.784366576819407,
+      "grad_norm": 0.61308354139328,
+      "learning_rate": 0.0003734311926605504,
+      "loss": 3.4635,
+      "step": 35100
+    },
+    {
+      "epoch": 3.789757412398922,
+      "grad_norm": 0.6734222173690796,
+      "learning_rate": 0.000373107393416082,
+      "loss": 3.4569,
+      "step": 35150
+    },
+    {
+      "epoch": 3.7951482479784366,
+      "grad_norm": 0.5994874835014343,
+      "learning_rate": 0.00037278359417161356,
+      "loss": 3.4739,
+      "step": 35200
+    },
+    {
+      "epoch": 3.8005390835579513,
+      "grad_norm": 0.6053499579429626,
+      "learning_rate": 0.00037245979492714517,
+      "loss": 3.4575,
+      "step": 35250
+    },
+    {
+      "epoch": 3.8059299191374665,
+      "grad_norm": 0.6050474643707275,
+      "learning_rate": 0.0003721359956826767,
+      "loss": 3.4419,
+      "step": 35300
+    },
+    {
+      "epoch": 3.811320754716981,
+      "grad_norm": 0.7089765071868896,
+      "learning_rate": 0.0003718121964382083,
+      "loss": 3.471,
+      "step": 35350
+    },
+    {
+      "epoch": 3.816711590296496,
+      "grad_norm": 0.6321031451225281,
+      "learning_rate": 0.0003714883971937398,
+      "loss": 3.4344,
+      "step": 35400
+    },
+    {
+      "epoch": 3.822102425876011,
+      "grad_norm": 0.655995786190033,
+      "learning_rate": 0.0003711645979492714,
+      "loss": 3.457,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8274932614555257,
+      "grad_norm": 0.617946445941925,
+      "learning_rate": 0.00037084079870480297,
+      "loss": 3.4314,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8328840970350404,
+      "grad_norm": 0.6026831865310669,
+      "learning_rate": 0.0003705169994603345,
+      "loss": 3.464,
+      "step": 35550
+    },
+    {
+      "epoch": 3.838274932614555,
+      "grad_norm": 0.6153815388679504,
+      "learning_rate": 0.00037019320021586613,
+      "loss": 3.4441,
+      "step": 35600
+    },
+    {
+      "epoch": 3.8436657681940702,
+      "grad_norm": 0.6427329778671265,
+      "learning_rate": 0.0003698694009713977,
+      "loss": 3.4645,
+      "step": 35650
+    },
+    {
+      "epoch": 3.849056603773585,
+      "grad_norm": 0.6545435786247253,
+      "learning_rate": 0.0003695456017269293,
+      "loss": 3.4523,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8544474393530996,
+      "grad_norm": 0.6320939064025879,
+      "learning_rate": 0.00036922180248246083,
+      "loss": 3.4408,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8598382749326143,
+      "grad_norm": 0.6298137307167053,
+      "learning_rate": 0.00036889800323799244,
+      "loss": 3.46,
+      "step": 35800
+    },
+    {
+      "epoch": 3.8652291105121295,
+      "grad_norm": 0.6715940237045288,
+      "learning_rate": 0.000368574203993524,
+      "loss": 3.4544,
+      "step": 35850
+    },
+    {
+      "epoch": 3.870619946091644,
+      "grad_norm": 0.5963342189788818,
+      "learning_rate": 0.0003682504047490556,
+      "loss": 3.4677,
+      "step": 35900
+    },
+    {
+      "epoch": 3.876010781671159,
+      "grad_norm": 0.6674957275390625,
+      "learning_rate": 0.00036792660550458714,
+      "loss": 3.452,
+      "step": 35950
+    },
+    {
+      "epoch": 3.881401617250674,
+      "grad_norm": 0.626230001449585,
+      "learning_rate": 0.00036760280626011864,
+      "loss": 3.4731,
+      "step": 36000
+    },
+    {
+      "epoch": 3.881401617250674,
+      "eval_accuracy": 0.37561976965370475,
+      "eval_loss": 3.4497599601745605,
+      "eval_runtime": 184.0301,
+      "eval_samples_per_second": 97.87,
+      "eval_steps_per_second": 6.119,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8867924528301887,
+      "grad_norm": 0.6103031039237976,
+      "learning_rate": 0.0003672790070156503,
+      "loss": 3.4362,
+      "step": 36050
+    },
+    {
+      "epoch": 3.8921832884097034,
+      "grad_norm": 0.6194986701011658,
+      "learning_rate": 0.0003669552077711818,
+      "loss": 3.4354,
+      "step": 36100
+    },
+    {
+      "epoch": 3.8975741239892185,
+      "grad_norm": 0.6245982050895691,
+      "learning_rate": 0.0003666314085267134,
+      "loss": 3.4358,
+      "step": 36150
+    },
+    {
+      "epoch": 3.9029649595687332,
+      "grad_norm": 0.5864128470420837,
+      "learning_rate": 0.00036630760928224495,
+      "loss": 3.4409,
+      "step": 36200
+    },
+    {
+      "epoch": 3.908355795148248,
+      "grad_norm": 0.6105799674987793,
+      "learning_rate": 0.00036598381003777656,
+      "loss": 3.4513,
+      "step": 36250
+    },
+    {
+      "epoch": 3.913746630727763,
+      "grad_norm": 0.6365011930465698,
+      "learning_rate": 0.0003656600107933081,
+      "loss": 3.4527,
+      "step": 36300
+    },
+    {
+      "epoch": 3.9191374663072778,
+      "grad_norm": 0.6909796595573425,
+      "learning_rate": 0.0003653362115488397,
+      "loss": 3.4705,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9245283018867925,
+      "grad_norm": 0.795727014541626,
+      "learning_rate": 0.00036501241230437126,
+      "loss": 3.4567,
+      "step": 36400
+    },
+    {
+      "epoch": 3.929919137466307,
+      "grad_norm": 0.6463427543640137,
+      "learning_rate": 0.0003646886130599028,
+      "loss": 3.4496,
+      "step": 36450
+    },
+    {
+      "epoch": 3.935309973045822,
+      "grad_norm": 0.6198216676712036,
+      "learning_rate": 0.0003643648138154344,
+      "loss": 3.4657,
+      "step": 36500
+    },
+    {
+      "epoch": 3.940700808625337,
+      "grad_norm": 0.6134477853775024,
+      "learning_rate": 0.00036404101457096597,
+      "loss": 3.4736,
+      "step": 36550
+    },
+    {
+      "epoch": 3.9460916442048517,
+      "grad_norm": 0.6083407998085022,
+      "learning_rate": 0.00036371721532649757,
+      "loss": 3.4535,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9514824797843664,
+      "grad_norm": 0.617780864238739,
+      "learning_rate": 0.0003633934160820291,
+      "loss": 3.4614,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9568733153638815,
+      "grad_norm": 0.620871365070343,
+      "learning_rate": 0.00036306961683756073,
+      "loss": 3.4446,
+      "step": 36700
+    },
+    {
+      "epoch": 3.9622641509433962,
+      "grad_norm": 0.684502124786377,
+      "learning_rate": 0.0003627458175930922,
+      "loss": 3.452,
+      "step": 36750
+    },
+    {
+      "epoch": 3.967654986522911,
+      "grad_norm": 0.6206174492835999,
+      "learning_rate": 0.0003624220183486238,
+      "loss": 3.4475,
+      "step": 36800
+    },
+    {
+      "epoch": 3.973045822102426,
+      "grad_norm": 0.6756080389022827,
+      "learning_rate": 0.0003620982191041554,
+      "loss": 3.4326,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9784366576819408,
+      "grad_norm": 0.618306577205658,
+      "learning_rate": 0.00036177441985968693,
+      "loss": 3.4326,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9838274932614555,
+      "grad_norm": 0.6362138390541077,
+      "learning_rate": 0.00036145062061521854,
+      "loss": 3.4391,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "grad_norm": 0.6646049618721008,
+      "learning_rate": 0.0003611268213707501,
+      "loss": 3.4391,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "eval_accuracy": 0.3763430719278858,
+      "eval_loss": 3.442918539047241,
+      "eval_runtime": 183.962,
+      "eval_samples_per_second": 97.906,
+      "eval_steps_per_second": 6.121,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9946091644204853,
+      "grad_norm": 0.6206597089767456,
+      "learning_rate": 0.0003608030221262817,
+      "loss": 3.4462,
+      "step": 37050
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.2068943977355957,
+      "learning_rate": 0.00036047922288181324,
+      "loss": 3.4468,
+      "step": 37100
+    },
+    {
+      "epoch": 4.005390835579515,
+      "grad_norm": 0.6037599444389343,
+      "learning_rate": 0.00036015542363734485,
+      "loss": 3.3481,
+      "step": 37150
+    },
+    {
+      "epoch": 4.010781671159029,
+      "grad_norm": 0.6229018568992615,
+      "learning_rate": 0.0003598316243928764,
+      "loss": 3.352,
+      "step": 37200
+    },
+    {
+      "epoch": 4.0161725067385445,
+      "grad_norm": 0.6315988302230835,
+      "learning_rate": 0.00035950782514840795,
+      "loss": 3.345,
+      "step": 37250
+    },
+    {
+      "epoch": 4.02156334231806,
+      "grad_norm": 0.601076602935791,
+      "learning_rate": 0.00035918402590393955,
+      "loss": 3.3492,
+      "step": 37300
+    },
+    {
+      "epoch": 4.026954177897574,
+      "grad_norm": 0.6532601714134216,
+      "learning_rate": 0.00035886022665947105,
+      "loss": 3.3611,
+      "step": 37350
+    },
+    {
+      "epoch": 4.032345013477089,
+      "grad_norm": 0.6189490556716919,
+      "learning_rate": 0.0003585364274150027,
+      "loss": 3.3661,
+      "step": 37400
+    },
+    {
+      "epoch": 4.037735849056604,
+      "grad_norm": 0.6153668761253357,
+      "learning_rate": 0.0003582126281705342,
+      "loss": 3.3486,
+      "step": 37450
+    },
+    {
+      "epoch": 4.0431266846361185,
+      "grad_norm": 0.6020147204399109,
+      "learning_rate": 0.0003578888289260658,
+      "loss": 3.3675,
+      "step": 37500
+    },
+    {
+      "epoch": 4.048517520215634,
+      "grad_norm": 0.6158584356307983,
+      "learning_rate": 0.00035756502968159736,
+      "loss": 3.3505,
+      "step": 37550
+    },
+    {
+      "epoch": 4.053908355795148,
+      "grad_norm": 0.649570643901825,
+      "learning_rate": 0.00035724123043712896,
+      "loss": 3.3583,
+      "step": 37600
+    },
+    {
+      "epoch": 4.059299191374663,
+      "grad_norm": 0.6425063014030457,
+      "learning_rate": 0.0003569174311926605,
+      "loss": 3.3569,
+      "step": 37650
+    },
+    {
+      "epoch": 4.064690026954178,
+      "grad_norm": 0.6738588213920593,
+      "learning_rate": 0.00035659363194819206,
+      "loss": 3.3734,
+      "step": 37700
+    },
+    {
+      "epoch": 4.070080862533692,
+      "grad_norm": 0.6135583519935608,
+      "learning_rate": 0.00035626983270372367,
+      "loss": 3.3618,
+      "step": 37750
+    },
+    {
+      "epoch": 4.0754716981132075,
+      "grad_norm": 0.6319704055786133,
+      "learning_rate": 0.0003559460334592552,
+      "loss": 3.3697,
+      "step": 37800
+    },
+    {
+      "epoch": 4.080862533692723,
+      "grad_norm": 0.6221917271614075,
+      "learning_rate": 0.0003556222342147868,
+      "loss": 3.3572,
+      "step": 37850
+    },
+    {
+      "epoch": 4.086253369272237,
+      "grad_norm": 0.6539017558097839,
+      "learning_rate": 0.0003552984349703184,
+      "loss": 3.368,
+      "step": 37900
+    },
+    {
+      "epoch": 4.091644204851752,
+      "grad_norm": 0.6019672155380249,
+      "learning_rate": 0.00035497463572585,
+      "loss": 3.3568,
+      "step": 37950
+    },
+    {
+      "epoch": 4.097035040431267,
+      "grad_norm": 0.6425307989120483,
+      "learning_rate": 0.00035465083648138153,
+      "loss": 3.3773,
+      "step": 38000
+    },
+    {
+      "epoch": 4.097035040431267,
+      "eval_accuracy": 0.37657037376919367,
+      "eval_loss": 3.4480812549591064,
+      "eval_runtime": 183.7404,
+      "eval_samples_per_second": 98.024,
+      "eval_steps_per_second": 6.128,
+      "step": 38000
+    },
+    {
+      "epoch": 4.1024258760107815,
+      "grad_norm": 0.6474397778511047,
+      "learning_rate": 0.00035432703723691314,
+      "loss": 3.3589,
+      "step": 38050
+    },
+    {
+      "epoch": 4.107816711590297,
+      "grad_norm": 0.6612837314605713,
+      "learning_rate": 0.00035400323799244463,
+      "loss": 3.3826,
+      "step": 38100
+    },
+    {
+      "epoch": 4.113207547169812,
+      "grad_norm": 0.6642338037490845,
+      "learning_rate": 0.0003536794387479762,
+      "loss": 3.3693,
+      "step": 38150
+    },
+    {
+      "epoch": 4.118598382749326,
+      "grad_norm": 0.6186237931251526,
+      "learning_rate": 0.0003533556395035078,
+      "loss": 3.3729,
+      "step": 38200
+    },
+    {
+      "epoch": 4.123989218328841,
+      "grad_norm": 0.5964476466178894,
+      "learning_rate": 0.00035303184025903934,
+      "loss": 3.3645,
+      "step": 38250
+    },
+    {
+      "epoch": 4.129380053908355,
+      "grad_norm": 0.6504191160202026,
+      "learning_rate": 0.00035270804101457094,
+      "loss": 3.3925,
+      "step": 38300
+    },
+    {
+      "epoch": 4.1347708894878705,
+      "grad_norm": 0.6848699450492859,
+      "learning_rate": 0.0003523842417701025,
+      "loss": 3.3713,
+      "step": 38350
+    },
+    {
+      "epoch": 4.140161725067386,
+      "grad_norm": 0.6712897419929504,
+      "learning_rate": 0.00035206691851052345,
+      "loss": 3.3645,
+      "step": 38400
+    },
+    {
+      "epoch": 4.1455525606469,
+      "grad_norm": 0.6874973773956299,
+      "learning_rate": 0.000351743119266055,
+      "loss": 3.3804,
+      "step": 38450
+    },
+    {
+      "epoch": 4.150943396226415,
+      "grad_norm": 0.6249614357948303,
+      "learning_rate": 0.0003514193200215866,
+      "loss": 3.3727,
+      "step": 38500
+    },
+    {
+      "epoch": 4.15633423180593,
+      "grad_norm": 0.6266210675239563,
+      "learning_rate": 0.00035109552077711815,
+      "loss": 3.3783,
+      "step": 38550
+    },
+    {
+      "epoch": 4.1617250673854445,
+      "grad_norm": 0.6524613499641418,
+      "learning_rate": 0.00035077172153264976,
+      "loss": 3.3802,
+      "step": 38600
+    },
+    {
+      "epoch": 4.16711590296496,
+      "grad_norm": 0.6768172383308411,
+      "learning_rate": 0.0003504479222881813,
+      "loss": 3.3788,
+      "step": 38650
+    },
+    {
+      "epoch": 4.172506738544475,
+      "grad_norm": 0.7132668495178223,
+      "learning_rate": 0.0003501241230437129,
+      "loss": 3.3795,
+      "step": 38700
+    },
+    {
+      "epoch": 4.177897574123989,
+      "grad_norm": 0.6380689144134521,
+      "learning_rate": 0.0003498003237992444,
+      "loss": 3.379,
+      "step": 38750
+    },
+    {
+      "epoch": 4.183288409703504,
+      "grad_norm": 0.6313562393188477,
+      "learning_rate": 0.000349476524554776,
+      "loss": 3.3853,
+      "step": 38800
+    },
+    {
+      "epoch": 4.188679245283019,
+      "grad_norm": 0.6632422804832458,
+      "learning_rate": 0.00034915272531030756,
+      "loss": 3.3703,
+      "step": 38850
+    },
+    {
+      "epoch": 4.1940700808625335,
+      "grad_norm": 0.6951757073402405,
+      "learning_rate": 0.0003488289260658391,
+      "loss": 3.3957,
+      "step": 38900
+    },
+    {
+      "epoch": 4.199460916442049,
+      "grad_norm": 0.6744063496589661,
+      "learning_rate": 0.0003485051268213707,
+      "loss": 3.4029,
+      "step": 38950
+    },
+    {
+      "epoch": 4.204851752021563,
+      "grad_norm": 0.6867125034332275,
+      "learning_rate": 0.00034818132757690227,
+      "loss": 3.3765,
+      "step": 39000
+    },
+    {
+      "epoch": 4.204851752021563,
+      "eval_accuracy": 0.37701585060923304,
+      "eval_loss": 3.4399898052215576,
+      "eval_runtime": 183.7265,
+      "eval_samples_per_second": 98.032,
+      "eval_steps_per_second": 6.129,
+      "step": 39000
+    },
+    {
+      "epoch": 4.210242587601078,
+      "grad_norm": 0.7025331854820251,
+      "learning_rate": 0.0003478575283324339,
+      "loss": 3.3783,
+      "step": 39050
+    },
+    {
+      "epoch": 4.215633423180593,
+      "grad_norm": 0.620579719543457,
+      "learning_rate": 0.0003475337290879654,
+      "loss": 3.3775,
+      "step": 39100
+    },
+    {
+      "epoch": 4.2210242587601075,
+      "grad_norm": 0.6752045154571533,
+      "learning_rate": 0.00034720992984349703,
+      "loss": 3.3801,
+      "step": 39150
+    },
+    {
+      "epoch": 4.226415094339623,
+      "grad_norm": 0.6968411803245544,
+      "learning_rate": 0.0003468861305990286,
+      "loss": 3.3906,
+      "step": 39200
+    },
+    {
+      "epoch": 4.231805929919138,
+      "grad_norm": 0.6568784117698669,
+      "learning_rate": 0.0003465623313545602,
+      "loss": 3.3789,
+      "step": 39250
+    },
+    {
+      "epoch": 4.237196765498652,
+      "grad_norm": 0.6356040239334106,
+      "learning_rate": 0.00034623853211009173,
+      "loss": 3.3842,
+      "step": 39300
+    },
+    {
+      "epoch": 4.242587601078167,
+      "grad_norm": 0.658807635307312,
+      "learning_rate": 0.00034591473286562323,
+      "loss": 3.3939,
+      "step": 39350
+    },
+    {
+      "epoch": 4.247978436657682,
+      "grad_norm": 0.6783695816993713,
+      "learning_rate": 0.0003455909336211549,
+      "loss": 3.3761,
+      "step": 39400
+    },
+    {
+      "epoch": 4.2533692722371965,
+      "grad_norm": 0.692827045917511,
+      "learning_rate": 0.0003452671343766864,
+      "loss": 3.3755,
+      "step": 39450
+    },
+    {
+      "epoch": 4.258760107816712,
+      "grad_norm": 0.6296021938323975,
+      "learning_rate": 0.000344943335132218,
+      "loss": 3.3845,
+      "step": 39500
+    },
+    {
+      "epoch": 4.264150943396227,
+      "grad_norm": 0.6483592391014099,
+      "learning_rate": 0.00034461953588774954,
+      "loss": 3.3662,
+      "step": 39550
+    },
+    {
+      "epoch": 4.269541778975741,
+      "grad_norm": 0.6927255392074585,
+      "learning_rate": 0.00034429573664328115,
+      "loss": 3.3778,
+      "step": 39600
+    },
+    {
+      "epoch": 4.274932614555256,
+      "grad_norm": 0.6652512550354004,
+      "learning_rate": 0.0003439719373988127,
+      "loss": 3.3664,
+      "step": 39650
+    },
+    {
+      "epoch": 4.280323450134771,
+      "grad_norm": 0.6730897426605225,
+      "learning_rate": 0.00034364813815434425,
+      "loss": 3.3955,
+      "step": 39700
+    },
+    {
+      "epoch": 4.285714285714286,
+      "grad_norm": 0.6138418316841125,
+      "learning_rate": 0.00034332433890987585,
+      "loss": 3.3833,
+      "step": 39750
+    },
+    {
+      "epoch": 4.291105121293801,
+      "grad_norm": 0.646796464920044,
+      "learning_rate": 0.0003430005396654074,
+      "loss": 3.3601,
+      "step": 39800
+    },
+    {
+      "epoch": 4.296495956873315,
+      "grad_norm": 0.6964720487594604,
+      "learning_rate": 0.000342676740420939,
+      "loss": 3.3958,
+      "step": 39850
+    },
+    {
+      "epoch": 4.30188679245283,
+      "grad_norm": 0.6758812069892883,
+      "learning_rate": 0.00034235294117647056,
+      "loss": 3.3742,
+      "step": 39900
+    },
+    {
+      "epoch": 4.307277628032345,
+      "grad_norm": 0.6758851408958435,
+      "learning_rate": 0.00034202914193200216,
+      "loss": 3.3595,
+      "step": 39950
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "grad_norm": 0.5890964269638062,
+      "learning_rate": 0.0003417053426875337,
+      "loss": 3.3936,
+      "step": 40000
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "eval_accuracy": 0.37785552012626333,
+      "eval_loss": 3.4354093074798584,
+      "eval_runtime": 184.1833,
+      "eval_samples_per_second": 97.788,
+      "eval_steps_per_second": 6.113,
+      "step": 40000
+    },
+    {
+      "epoch": 4.318059299191375,
+      "grad_norm": 0.6625378131866455,
+      "learning_rate": 0.0003413815434430653,
+      "loss": 3.3808,
+      "step": 40050
+    },
+    {
+      "epoch": 4.32345013477089,
+      "grad_norm": 0.6554327011108398,
+      "learning_rate": 0.0003410577441985968,
+      "loss": 3.3846,
+      "step": 40100
+    },
+    {
+      "epoch": 4.328840970350404,
+      "grad_norm": 0.6583057045936584,
+      "learning_rate": 0.00034073394495412837,
+      "loss": 3.3761,
+      "step": 40150
+    },
+    {
+      "epoch": 4.334231805929919,
+      "grad_norm": 0.6082165837287903,
+      "learning_rate": 0.00034041014570965997,
+      "loss": 3.3949,
+      "step": 40200
+    },
+    {
+      "epoch": 4.339622641509434,
+      "grad_norm": 0.6874740719795227,
+      "learning_rate": 0.0003400863464651915,
+      "loss": 3.3726,
+      "step": 40250
+    },
+    {
+      "epoch": 4.345013477088949,
+      "grad_norm": 0.6384726762771606,
+      "learning_rate": 0.0003397625472207231,
+      "loss": 3.3946,
+      "step": 40300
+    },
+    {
+      "epoch": 4.350404312668464,
+      "grad_norm": 0.6966606378555298,
+      "learning_rate": 0.0003394387479762547,
+      "loss": 3.3995,
+      "step": 40350
+    },
+    {
+      "epoch": 4.355795148247978,
+      "grad_norm": 0.6895248293876648,
+      "learning_rate": 0.0003391149487317863,
+      "loss": 3.3819,
+      "step": 40400
+    },
+    {
+      "epoch": 4.361185983827493,
+      "grad_norm": 0.6699813604354858,
+      "learning_rate": 0.00033879114948731783,
+      "loss": 3.3903,
+      "step": 40450
+    },
+    {
+      "epoch": 4.366576819407008,
+      "grad_norm": 0.6132922768592834,
+      "learning_rate": 0.00033846735024284944,
+      "loss": 3.3853,
+      "step": 40500
+    },
+    {
+      "epoch": 4.3719676549865225,
+      "grad_norm": 0.6432852745056152,
+      "learning_rate": 0.000338143550998381,
+      "loss": 3.3811,
+      "step": 40550
+    },
+    {
+      "epoch": 4.377358490566038,
+      "grad_norm": 0.6538777351379395,
+      "learning_rate": 0.00033781975175391254,
+      "loss": 3.376,
+      "step": 40600
+    },
+    {
+      "epoch": 4.382749326145553,
+      "grad_norm": 0.7164791822433472,
+      "learning_rate": 0.00033749595250944414,
+      "loss": 3.3981,
+      "step": 40650
+    },
+    {
+      "epoch": 4.388140161725067,
+      "grad_norm": 0.6560450792312622,
+      "learning_rate": 0.00033717215326497564,
+      "loss": 3.3787,
+      "step": 40700
+    },
+    {
+      "epoch": 4.393530997304582,
+      "grad_norm": 0.687532901763916,
+      "learning_rate": 0.0003368483540205073,
+      "loss": 3.3732,
+      "step": 40750
+    },
+    {
+      "epoch": 4.398921832884097,
+      "grad_norm": 0.7162414789199829,
+      "learning_rate": 0.0003365245547760388,
+      "loss": 3.3951,
+      "step": 40800
+    },
+    {
+      "epoch": 4.404312668463612,
+      "grad_norm": 0.643511950969696,
+      "learning_rate": 0.0003362007555315704,
+      "loss": 3.4023,
+      "step": 40850
+    },
+    {
+      "epoch": 4.409703504043127,
+      "grad_norm": 0.6941072940826416,
+      "learning_rate": 0.00033587695628710195,
+      "loss": 3.3932,
+      "step": 40900
+    },
+    {
+      "epoch": 4.415094339622642,
+      "grad_norm": 0.5936976075172424,
+      "learning_rate": 0.00033555315704263355,
+      "loss": 3.3823,
+      "step": 40950
+    },
+    {
+      "epoch": 4.420485175202156,
+      "grad_norm": 0.6859274506568909,
+      "learning_rate": 0.0003352293577981651,
+      "loss": 3.3884,
+      "step": 41000
+    },
+    {
+      "epoch": 4.420485175202156,
+      "eval_accuracy": 0.3781056390740123,
+      "eval_loss": 3.4318952560424805,
+      "eval_runtime": 183.8135,
+      "eval_samples_per_second": 97.985,
+      "eval_steps_per_second": 6.126,
+      "step": 41000
+    },
+    {
+      "epoch": 4.425876010781671,
+      "grad_norm": 0.6612915992736816,
+      "learning_rate": 0.00033490555855369665,
+      "loss": 3.3839,
+      "step": 41050
+    },
+    {
+      "epoch": 4.431266846361186,
+      "grad_norm": 0.6748125553131104,
+      "learning_rate": 0.00033458175930922826,
+      "loss": 3.3767,
+      "step": 41100
+    },
+    {
+      "epoch": 4.436657681940701,
+      "grad_norm": 0.6383491158485413,
+      "learning_rate": 0.0003342579600647598,
+      "loss": 3.3884,
+      "step": 41150
+    },
+    {
+      "epoch": 4.442048517520216,
+      "grad_norm": 0.672412633895874,
+      "learning_rate": 0.0003339341608202914,
+      "loss": 3.3683,
+      "step": 41200
+    },
+    {
+      "epoch": 4.44743935309973,
+      "grad_norm": 0.6618825197219849,
+      "learning_rate": 0.00033361036157582297,
+      "loss": 3.3797,
+      "step": 41250
+    },
+    {
+      "epoch": 4.452830188679245,
+      "grad_norm": 0.6703552603721619,
+      "learning_rate": 0.00033328656233135457,
+      "loss": 3.3878,
+      "step": 41300
+    },
+    {
+      "epoch": 4.45822102425876,
+      "grad_norm": 0.677817165851593,
+      "learning_rate": 0.0003329627630868861,
+      "loss": 3.4008,
+      "step": 41350
+    },
+    {
+      "epoch": 4.463611859838275,
+      "grad_norm": 0.6925399303436279,
+      "learning_rate": 0.0003326389638424177,
+      "loss": 3.3772,
+      "step": 41400
+    },
+    {
+      "epoch": 4.46900269541779,
+      "grad_norm": 0.6213698387145996,
+      "learning_rate": 0.0003323151645979492,
+      "loss": 3.3777,
+      "step": 41450
+    },
+    {
+      "epoch": 4.474393530997305,
+      "grad_norm": 0.6628020405769348,
+      "learning_rate": 0.00033199136535348077,
+      "loss": 3.4046,
+      "step": 41500
+    },
+    {
+      "epoch": 4.479784366576819,
+      "grad_norm": 0.6591078639030457,
+      "learning_rate": 0.0003316675661090124,
+      "loss": 3.4125,
+      "step": 41550
+    },
+    {
+      "epoch": 4.485175202156334,
+      "grad_norm": 0.6607985496520996,
+      "learning_rate": 0.00033134376686454393,
+      "loss": 3.3717,
+      "step": 41600
+    },
+    {
+      "epoch": 4.490566037735849,
+      "grad_norm": 0.6439804434776306,
+      "learning_rate": 0.00033101996762007553,
+      "loss": 3.3877,
+      "step": 41650
+    },
+    {
+      "epoch": 4.495956873315364,
+      "grad_norm": 0.6608968377113342,
+      "learning_rate": 0.0003306961683756071,
+      "loss": 3.3859,
+      "step": 41700
+    },
+    {
+      "epoch": 4.501347708894879,
+      "grad_norm": 0.6322570443153381,
+      "learning_rate": 0.0003303723691311387,
+      "loss": 3.4031,
+      "step": 41750
+    },
+    {
+      "epoch": 4.506738544474393,
+      "grad_norm": 0.6550046801567078,
+      "learning_rate": 0.00033004856988667024,
+      "loss": 3.4043,
+      "step": 41800
+    },
+    {
+      "epoch": 4.512129380053908,
+      "grad_norm": 0.6358398795127869,
+      "learning_rate": 0.0003297247706422018,
+      "loss": 3.3698,
+      "step": 41850
+    },
+    {
+      "epoch": 4.517520215633423,
+      "grad_norm": 0.6461077332496643,
+      "learning_rate": 0.0003294009713977334,
+      "loss": 3.3978,
+      "step": 41900
+    },
+    {
+      "epoch": 4.5229110512129385,
+      "grad_norm": 0.6892066597938538,
+      "learning_rate": 0.00032907717215326494,
+      "loss": 3.362,
+      "step": 41950
+    },
+    {
+      "epoch": 4.528301886792453,
+      "grad_norm": 0.6454676389694214,
+      "learning_rate": 0.00032875337290879655,
+      "loss": 3.3762,
+      "step": 42000
+    },
+    {
+      "epoch": 4.528301886792453,
+      "eval_accuracy": 0.3790023513571452,
+      "eval_loss": 3.4271507263183594,
+      "eval_runtime": 184.2053,
+      "eval_samples_per_second": 97.777,
+      "eval_steps_per_second": 6.113,
+      "step": 42000
+    },
+    {
+      "epoch": 4.533692722371968,
+      "grad_norm": 0.6774193048477173,
+      "learning_rate": 0.00032842957366432805,
+      "loss": 3.3983,
+      "step": 42050
+    },
+    {
+      "epoch": 4.539083557951482,
+      "grad_norm": 0.6757726669311523,
+      "learning_rate": 0.0003281057744198597,
+      "loss": 3.3727,
+      "step": 42100
+    },
+    {
+      "epoch": 4.544474393530997,
+      "grad_norm": 0.6542174816131592,
+      "learning_rate": 0.0003277819751753912,
+      "loss": 3.3891,
+      "step": 42150
+    },
+    {
+      "epoch": 4.549865229110512,
+      "grad_norm": 0.6432123184204102,
+      "learning_rate": 0.0003274581759309228,
+      "loss": 3.385,
+      "step": 42200
+    },
+    {
+      "epoch": 4.555256064690027,
+      "grad_norm": 0.7210422158241272,
+      "learning_rate": 0.00032713437668645436,
+      "loss": 3.411,
+      "step": 42250
+    },
+    {
+      "epoch": 4.560646900269542,
+      "grad_norm": 0.773587167263031,
+      "learning_rate": 0.0003268105774419859,
+      "loss": 3.3781,
+      "step": 42300
+    },
+    {
+      "epoch": 4.566037735849057,
+      "grad_norm": 0.6960445642471313,
+      "learning_rate": 0.0003264867781975175,
+      "loss": 3.3946,
+      "step": 42350
+    },
+    {
+      "epoch": 4.571428571428571,
+      "grad_norm": 0.6441209316253662,
+      "learning_rate": 0.00032616297895304906,
+      "loss": 3.3652,
+      "step": 42400
+    },
+    {
+      "epoch": 4.576819407008086,
+      "grad_norm": 0.6879327893257141,
+      "learning_rate": 0.00032584565569347,
+      "loss": 3.3817,
+      "step": 42450
+    },
+    {
+      "epoch": 4.5822102425876015,
+      "grad_norm": 0.6669209003448486,
+      "learning_rate": 0.0003255218564490016,
+      "loss": 3.3879,
+      "step": 42500
+    },
+    {
+      "epoch": 4.587601078167116,
+      "grad_norm": 0.6996499300003052,
+      "learning_rate": 0.00032519805720453317,
+      "loss": 3.3882,
+      "step": 42550
+    },
+    {
+      "epoch": 4.592991913746631,
+      "grad_norm": 0.6694384217262268,
+      "learning_rate": 0.0003248742579600647,
+      "loss": 3.3783,
+      "step": 42600
+    },
+    {
+      "epoch": 4.598382749326145,
+      "grad_norm": 0.6417438387870789,
+      "learning_rate": 0.0003245504587155963,
+      "loss": 3.3931,
+      "step": 42650
+    },
+    {
+      "epoch": 4.60377358490566,
+      "grad_norm": 0.7074453234672546,
+      "learning_rate": 0.0003242266594711278,
+      "loss": 3.3937,
+      "step": 42700
+    },
+    {
+      "epoch": 4.609164420485175,
+      "grad_norm": 0.6900057196617126,
+      "learning_rate": 0.0003239028602266595,
+      "loss": 3.3728,
+      "step": 42750
+    },
+    {
+      "epoch": 4.6145552560646905,
+      "grad_norm": 0.6593276262283325,
+      "learning_rate": 0.000323579060982191,
+      "loss": 3.404,
+      "step": 42800
+    },
+    {
+      "epoch": 4.619946091644205,
+      "grad_norm": 0.6587552428245544,
+      "learning_rate": 0.0003232552617377226,
+      "loss": 3.3996,
+      "step": 42850
+    },
+    {
+      "epoch": 4.62533692722372,
+      "grad_norm": 0.6677352786064148,
+      "learning_rate": 0.00032293146249325413,
+      "loss": 3.3965,
+      "step": 42900
+    },
+    {
+      "epoch": 4.630727762803234,
+      "grad_norm": 0.6786492466926575,
+      "learning_rate": 0.00032260766324878574,
+      "loss": 3.3921,
+      "step": 42950
+    },
+    {
+      "epoch": 4.636118598382749,
+      "grad_norm": 0.6595161557197571,
+      "learning_rate": 0.0003222838640043173,
+      "loss": 3.375,
+      "step": 43000
+    },
+    {
+      "epoch": 4.636118598382749,
+      "eval_accuracy": 0.3788302451828471,
+      "eval_loss": 3.422513008117676,
+      "eval_runtime": 183.7825,
+      "eval_samples_per_second": 98.002,
+      "eval_steps_per_second": 6.127,
+      "step": 43000
+    },
+    {
+      "epoch": 4.6415094339622645,
+      "grad_norm": 0.6542896628379822,
+      "learning_rate": 0.00032196006475984884,
+      "loss": 3.397,
+      "step": 43050
+    },
+    {
+      "epoch": 4.646900269541779,
+      "grad_norm": 0.6302821040153503,
+      "learning_rate": 0.00032163626551538044,
+      "loss": 3.3931,
+      "step": 43100
+    },
+    {
+      "epoch": 4.652291105121294,
+      "grad_norm": 0.6723839640617371,
+      "learning_rate": 0.0003213189422558014,
+      "loss": 3.3764,
+      "step": 43150
+    },
+    {
+      "epoch": 4.657681940700809,
+      "grad_norm": 0.7077513933181763,
+      "learning_rate": 0.00032099514301133295,
+      "loss": 3.3969,
+      "step": 43200
+    },
+    {
+      "epoch": 4.663072776280323,
+      "grad_norm": 0.6480196118354797,
+      "learning_rate": 0.00032067134376686455,
+      "loss": 3.4047,
+      "step": 43250
+    },
+    {
+      "epoch": 4.668463611859838,
+      "grad_norm": 0.6761808395385742,
+      "learning_rate": 0.0003203475445223961,
+      "loss": 3.3932,
+      "step": 43300
+    },
+    {
+      "epoch": 4.6738544474393535,
+      "grad_norm": 0.6225863099098206,
+      "learning_rate": 0.0003200237452779277,
+      "loss": 3.3894,
+      "step": 43350
+    },
+    {
+      "epoch": 4.679245283018868,
+      "grad_norm": 0.6918447613716125,
+      "learning_rate": 0.00031969994603345926,
+      "loss": 3.3835,
+      "step": 43400
+    },
+    {
+      "epoch": 4.684636118598383,
+      "grad_norm": 0.6438918709754944,
+      "learning_rate": 0.00031937614678899075,
+      "loss": 3.3819,
+      "step": 43450
+    },
+    {
+      "epoch": 4.690026954177897,
+      "grad_norm": 0.6605272889137268,
+      "learning_rate": 0.00031905234754452236,
+      "loss": 3.3883,
+      "step": 43500
+    },
+    {
+      "epoch": 4.695417789757412,
+      "grad_norm": 0.6213443875312805,
+      "learning_rate": 0.0003187285483000539,
+      "loss": 3.3871,
+      "step": 43550
+    },
+    {
+      "epoch": 4.7008086253369274,
+      "grad_norm": 0.6892732977867126,
+      "learning_rate": 0.0003184047490555855,
+      "loss": 3.375,
+      "step": 43600
+    },
+    {
+      "epoch": 4.706199460916442,
+      "grad_norm": 0.6602584719657898,
+      "learning_rate": 0.00031808094981111706,
+      "loss": 3.3888,
+      "step": 43650
+    },
+    {
+      "epoch": 4.711590296495957,
+      "grad_norm": 0.6708425283432007,
+      "learning_rate": 0.00031775715056664867,
+      "loss": 3.3869,
+      "step": 43700
+    },
+    {
+      "epoch": 4.716981132075472,
+      "grad_norm": 0.6698873043060303,
+      "learning_rate": 0.0003174333513221802,
+      "loss": 3.373,
+      "step": 43750
+    },
+    {
+      "epoch": 4.722371967654986,
+      "grad_norm": 0.6338597536087036,
+      "learning_rate": 0.00031710955207771177,
+      "loss": 3.3879,
+      "step": 43800
+    },
+    {
+      "epoch": 4.727762803234501,
+      "grad_norm": 0.7395240664482117,
+      "learning_rate": 0.0003167857528332434,
+      "loss": 3.4032,
+      "step": 43850
+    },
+    {
+      "epoch": 4.7331536388140165,
+      "grad_norm": 0.6551410555839539,
+      "learning_rate": 0.0003164619535887749,
+      "loss": 3.394,
+      "step": 43900
+    },
+    {
+      "epoch": 4.738544474393531,
+      "grad_norm": 0.7224283814430237,
+      "learning_rate": 0.00031613815434430653,
+      "loss": 3.3859,
+      "step": 43950
+    },
+    {
+      "epoch": 4.743935309973046,
+      "grad_norm": 0.6312947273254395,
+      "learning_rate": 0.0003158143550998381,
+      "loss": 3.3747,
+      "step": 44000
+    },
+    {
+      "epoch": 4.743935309973046,
+      "eval_accuracy": 0.38003672684913903,
+      "eval_loss": 3.4164884090423584,
+      "eval_runtime": 184.2338,
+      "eval_samples_per_second": 97.762,
+      "eval_steps_per_second": 6.112,
+      "step": 44000
+    },
+    {
+      "epoch": 4.74932614555256,
+      "grad_norm": 0.6509761214256287,
+      "learning_rate": 0.0003154905558553697,
+      "loss": 3.3905,
+      "step": 44050
+    },
+    {
+      "epoch": 4.754716981132075,
+      "grad_norm": 0.6715309023857117,
+      "learning_rate": 0.0003151667566109012,
+      "loss": 3.3842,
+      "step": 44100
+    },
+    {
+      "epoch": 4.7601078167115904,
+      "grad_norm": 0.7263250350952148,
+      "learning_rate": 0.00031484295736643284,
+      "loss": 3.3871,
+      "step": 44150
+    },
+    {
+      "epoch": 4.765498652291106,
+      "grad_norm": 0.6736378073692322,
+      "learning_rate": 0.00031451915812196434,
+      "loss": 3.3561,
+      "step": 44200
+    },
+    {
+      "epoch": 4.77088948787062,
+      "grad_norm": 0.7403033971786499,
+      "learning_rate": 0.0003141953588774959,
+      "loss": 3.3938,
+      "step": 44250
+    },
+    {
+      "epoch": 4.776280323450135,
+      "grad_norm": 0.7216389179229736,
+      "learning_rate": 0.0003138715596330275,
+      "loss": 3.3794,
+      "step": 44300
+    },
+    {
+      "epoch": 4.781671159029649,
+      "grad_norm": 0.72726970911026,
+      "learning_rate": 0.00031354776038855904,
+      "loss": 3.3543,
+      "step": 44350
+    },
+    {
+      "epoch": 4.787061994609164,
+      "grad_norm": 0.6150906085968018,
+      "learning_rate": 0.00031322396114409065,
+      "loss": 3.3835,
+      "step": 44400
+    },
+    {
+      "epoch": 4.7924528301886795,
+      "grad_norm": 0.6653279066085815,
+      "learning_rate": 0.0003129066378845116,
+      "loss": 3.3903,
+      "step": 44450
+    },
+    {
+      "epoch": 4.797843665768194,
+      "grad_norm": 0.6354139447212219,
+      "learning_rate": 0.00031258283864004315,
+      "loss": 3.3956,
+      "step": 44500
+    },
+    {
+      "epoch": 4.803234501347709,
+      "grad_norm": 0.6651866436004639,
+      "learning_rate": 0.0003122590393955747,
+      "loss": 3.4122,
+      "step": 44550
+    },
+    {
+      "epoch": 4.808625336927224,
+      "grad_norm": 0.6413260698318481,
+      "learning_rate": 0.0003119352401511063,
+      "loss": 3.3788,
+      "step": 44600
+    },
+    {
+      "epoch": 4.814016172506738,
+      "grad_norm": 0.7094629406929016,
+      "learning_rate": 0.00031161144090663786,
+      "loss": 3.3893,
+      "step": 44650
+    },
+    {
+      "epoch": 4.819407008086253,
+      "grad_norm": 0.6571715474128723,
+      "learning_rate": 0.00031128764166216946,
+      "loss": 3.3826,
+      "step": 44700
+    },
+    {
+      "epoch": 4.824797843665769,
+      "grad_norm": 0.6378207206726074,
+      "learning_rate": 0.00031096384241770096,
+      "loss": 3.3812,
+      "step": 44750
+    },
+    {
+      "epoch": 4.830188679245283,
+      "grad_norm": 0.6692754030227661,
+      "learning_rate": 0.0003106400431732326,
+      "loss": 3.3762,
+      "step": 44800
+    },
+    {
+      "epoch": 4.835579514824798,
+      "grad_norm": 0.6558183431625366,
+      "learning_rate": 0.0003103162439287641,
+      "loss": 3.3784,
+      "step": 44850
+    },
+    {
+      "epoch": 4.840970350404312,
+      "grad_norm": 0.6741693615913391,
+      "learning_rate": 0.0003099924446842957,
+      "loss": 3.3902,
+      "step": 44900
+    },
+    {
+      "epoch": 4.846361185983827,
+      "grad_norm": 0.6589843034744263,
+      "learning_rate": 0.00030966864543982727,
+      "loss": 3.4018,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "grad_norm": 0.6733647584915161,
+      "learning_rate": 0.0003093448461953588,
+      "loss": 3.3985,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "eval_accuracy": 0.38011973765542934,
+      "eval_loss": 3.4101078510284424,
+      "eval_runtime": 183.7639,
+      "eval_samples_per_second": 98.012,
+      "eval_steps_per_second": 6.127,
+      "step": 45000
+    },
+    {
+      "epoch": 4.857142857142857,
+      "grad_norm": 0.679538905620575,
+      "learning_rate": 0.0003090210469508904,
+      "loss": 3.3971,
+      "step": 45050
+    },
+    {
+      "epoch": 4.862533692722372,
+      "grad_norm": 0.7209023237228394,
+      "learning_rate": 0.00030869724770642197,
+      "loss": 3.391,
+      "step": 45100
+    },
+    {
+      "epoch": 4.867924528301887,
+      "grad_norm": 0.6559917330741882,
+      "learning_rate": 0.0003083734484619536,
+      "loss": 3.4099,
+      "step": 45150
+    },
+    {
+      "epoch": 4.873315363881401,
+      "grad_norm": 0.6977805495262146,
+      "learning_rate": 0.00030804964921748513,
+      "loss": 3.3935,
+      "step": 45200
+    },
+    {
+      "epoch": 4.878706199460916,
+      "grad_norm": 0.7327739000320435,
+      "learning_rate": 0.00030772584997301673,
+      "loss": 3.3777,
+      "step": 45250
+    },
+    {
+      "epoch": 4.884097035040432,
+      "grad_norm": 0.6754328608512878,
+      "learning_rate": 0.0003074020507285483,
+      "loss": 3.3802,
+      "step": 45300
+    },
+    {
+      "epoch": 4.889487870619946,
+      "grad_norm": 0.6827671527862549,
+      "learning_rate": 0.0003070782514840799,
+      "loss": 3.3702,
+      "step": 45350
+    },
+    {
+      "epoch": 4.894878706199461,
+      "grad_norm": 0.6643690466880798,
+      "learning_rate": 0.00030675445223961144,
+      "loss": 3.3821,
+      "step": 45400
+    },
+    {
+      "epoch": 4.900269541778976,
+      "grad_norm": 0.707639217376709,
+      "learning_rate": 0.00030643065299514294,
+      "loss": 3.3884,
+      "step": 45450
+    },
+    {
+      "epoch": 4.90566037735849,
+      "grad_norm": 0.7274863123893738,
+      "learning_rate": 0.00030610685375067454,
+      "loss": 3.3776,
+      "step": 45500
+    },
+    {
+      "epoch": 4.9110512129380055,
+      "grad_norm": 0.6537758708000183,
+      "learning_rate": 0.0003057830545062061,
+      "loss": 3.3664,
+      "step": 45550
+    },
+    {
+      "epoch": 4.916442048517521,
+      "grad_norm": 0.6469917297363281,
+      "learning_rate": 0.0003054592552617377,
+      "loss": 3.3919,
+      "step": 45600
+    },
+    {
+      "epoch": 4.921832884097035,
+      "grad_norm": 0.6994689702987671,
+      "learning_rate": 0.00030513545601726925,
+      "loss": 3.3893,
+      "step": 45650
+    },
+    {
+      "epoch": 4.92722371967655,
+      "grad_norm": 0.667890727519989,
+      "learning_rate": 0.00030481165677280085,
+      "loss": 3.3883,
+      "step": 45700
+    },
+    {
+      "epoch": 4.932614555256064,
+      "grad_norm": 0.6717064380645752,
+      "learning_rate": 0.0003044878575283324,
+      "loss": 3.3864,
+      "step": 45750
+    },
+    {
+      "epoch": 4.938005390835579,
+      "grad_norm": 0.65635085105896,
+      "learning_rate": 0.000304164058283864,
+      "loss": 3.3935,
+      "step": 45800
+    },
+    {
+      "epoch": 4.943396226415095,
+      "grad_norm": 0.6819509863853455,
+      "learning_rate": 0.00030384025903939556,
+      "loss": 3.3761,
+      "step": 45850
+    },
+    {
+      "epoch": 4.948787061994609,
+      "grad_norm": 0.6925830245018005,
+      "learning_rate": 0.0003035164597949271,
+      "loss": 3.3798,
+      "step": 45900
+    },
+    {
+      "epoch": 4.954177897574124,
+      "grad_norm": 0.6927830576896667,
+      "learning_rate": 0.0003031926605504587,
+      "loss": 3.3879,
+      "step": 45950
+    },
+    {
+      "epoch": 4.959568733153639,
+      "grad_norm": 0.637687623500824,
+      "learning_rate": 0.00030286886130599026,
+      "loss": 3.3853,
+      "step": 46000
+    },
+    {
+      "epoch": 4.959568733153639,
+      "eval_accuracy": 0.38066202421851136,
+      "eval_loss": 3.406095504760742,
+      "eval_runtime": 184.1141,
+      "eval_samples_per_second": 97.825,
+      "eval_steps_per_second": 6.116,
+      "step": 46000
+    },
+    {
+      "epoch": 4.964959568733153,
+      "grad_norm": 0.6471781134605408,
+      "learning_rate": 0.00030254506206152187,
+      "loss": 3.375,
+      "step": 46050
+    },
+    {
+      "epoch": 4.9703504043126685,
+      "grad_norm": 0.6780849099159241,
+      "learning_rate": 0.00030222126281705336,
+      "loss": 3.3776,
+      "step": 46100
+    },
+    {
+      "epoch": 4.975741239892184,
+      "grad_norm": 1.019982099533081,
+      "learning_rate": 0.000301897463572585,
+      "loss": 3.3775,
+      "step": 46150
+    },
+    {
+      "epoch": 4.981132075471698,
+      "grad_norm": 0.6539048552513123,
+      "learning_rate": 0.0003015736643281165,
+      "loss": 3.3683,
+      "step": 46200
+    },
+    {
+      "epoch": 4.986522911051213,
+      "grad_norm": 0.6219794750213623,
+      "learning_rate": 0.00030124986508364807,
+      "loss": 3.3791,
+      "step": 46250
+    },
+    {
+      "epoch": 4.991913746630727,
+      "grad_norm": 0.6488443613052368,
+      "learning_rate": 0.0003009260658391797,
+      "loss": 3.3952,
+      "step": 46300
+    },
+    {
+      "epoch": 4.997304582210242,
+      "grad_norm": 0.6938602328300476,
+      "learning_rate": 0.0003006022665947112,
+      "loss": 3.3779,
+      "step": 46350
+    },
+    {
+      "epoch": 5.002695417789758,
+      "grad_norm": 0.6452742218971252,
+      "learning_rate": 0.00030027846735024283,
+      "loss": 3.3349,
+      "step": 46400
+    },
+    {
+      "epoch": 5.008086253369272,
+      "grad_norm": 0.6376250982284546,
+      "learning_rate": 0.00029995466810577443,
+      "loss": 3.2813,
+      "step": 46450
+    },
+    {
+      "epoch": 5.013477088948787,
+      "grad_norm": 0.7294584512710571,
+      "learning_rate": 0.00029963086886130593,
+      "loss": 3.2785,
+      "step": 46500
+    },
+    {
+      "epoch": 5.018867924528302,
+      "grad_norm": 0.661499559879303,
+      "learning_rate": 0.00029930706961683754,
+      "loss": 3.2773,
+      "step": 46550
+    },
+    {
+      "epoch": 5.024258760107816,
+      "grad_norm": 0.6674730181694031,
+      "learning_rate": 0.0002989832703723691,
+      "loss": 3.2923,
+      "step": 46600
+    },
+    {
+      "epoch": 5.0296495956873315,
+      "grad_norm": 0.678300142288208,
+      "learning_rate": 0.0002986594711279007,
+      "loss": 3.2873,
+      "step": 46650
+    },
+    {
+      "epoch": 5.035040431266847,
+      "grad_norm": 0.7316428422927856,
+      "learning_rate": 0.00029833567188343224,
+      "loss": 3.3123,
+      "step": 46700
+    },
+    {
+      "epoch": 5.040431266846361,
+      "grad_norm": 0.6575501561164856,
+      "learning_rate": 0.00029801187263896385,
+      "loss": 3.2837,
+      "step": 46750
+    },
+    {
+      "epoch": 5.045822102425876,
+      "grad_norm": 0.7176420092582703,
+      "learning_rate": 0.0002976880733944954,
+      "loss": 3.2904,
+      "step": 46800
+    },
+    {
+      "epoch": 5.051212938005391,
+      "grad_norm": 0.7104209661483765,
+      "learning_rate": 0.00029736427415002695,
+      "loss": 3.3006,
+      "step": 46850
+    },
+    {
+      "epoch": 5.056603773584905,
+      "grad_norm": 0.7049729228019714,
+      "learning_rate": 0.0002970404749055585,
+      "loss": 3.3261,
+      "step": 46900
+    },
+    {
+      "epoch": 5.061994609164421,
+      "grad_norm": 0.622776985168457,
+      "learning_rate": 0.0002967166756610901,
+      "loss": 3.2972,
+      "step": 46950
+    },
+    {
+      "epoch": 5.067385444743936,
+      "grad_norm": 0.697993814945221,
+      "learning_rate": 0.00029639287641662165,
+      "loss": 3.2829,
+      "step": 47000
+    },
+    {
+      "epoch": 5.067385444743936,
+      "eval_accuracy": 0.38110869624031674,
+      "eval_loss": 3.4087939262390137,
+      "eval_runtime": 183.8604,
+      "eval_samples_per_second": 97.96,
+      "eval_steps_per_second": 6.124,
+      "step": 47000
+    },
+    {
+      "epoch": 5.07277628032345,
+      "grad_norm": 0.6486660242080688,
+      "learning_rate": 0.00029606907717215326,
+      "loss": 3.3014,
+      "step": 47050
+    },
+    {
+      "epoch": 5.078167115902965,
+      "grad_norm": 0.69781494140625,
+      "learning_rate": 0.0002957452779276848,
+      "loss": 3.3001,
+      "step": 47100
+    },
+    {
+      "epoch": 5.083557951482479,
+      "grad_norm": 0.6906808614730835,
+      "learning_rate": 0.00029542147868321636,
+      "loss": 3.3136,
+      "step": 47150
+    },
+    {
+      "epoch": 5.0889487870619945,
+      "grad_norm": 0.6491060256958008,
+      "learning_rate": 0.00029509767943874796,
+      "loss": 3.3082,
+      "step": 47200
+    },
+    {
+      "epoch": 5.09433962264151,
+      "grad_norm": 0.6999182105064392,
+      "learning_rate": 0.0002947738801942795,
+      "loss": 3.3151,
+      "step": 47250
+    },
+    {
+      "epoch": 5.099730458221024,
+      "grad_norm": 0.6435338854789734,
+      "learning_rate": 0.0002944500809498111,
+      "loss": 3.3212,
+      "step": 47300
+    },
+    {
+      "epoch": 5.105121293800539,
+      "grad_norm": 0.6768415570259094,
+      "learning_rate": 0.00029412628170534267,
+      "loss": 3.2983,
+      "step": 47350
+    },
+    {
+      "epoch": 5.110512129380054,
+      "grad_norm": 0.6941831111907959,
+      "learning_rate": 0.0002938024824608742,
+      "loss": 3.3183,
+      "step": 47400
+    },
+    {
+      "epoch": 5.115902964959568,
+      "grad_norm": 0.6833462715148926,
+      "learning_rate": 0.00029347868321640577,
+      "loss": 3.3287,
+      "step": 47450
+    },
+    {
+      "epoch": 5.121293800539084,
+      "grad_norm": 0.6526753902435303,
+      "learning_rate": 0.0002931548839719374,
+      "loss": 3.3072,
+      "step": 47500
+    },
+    {
+      "epoch": 5.126684636118599,
+      "grad_norm": 0.6595203876495361,
+      "learning_rate": 0.0002928310847274689,
+      "loss": 3.3137,
+      "step": 47550
+    },
+    {
+      "epoch": 5.132075471698113,
+      "grad_norm": 0.689451277256012,
+      "learning_rate": 0.00029250728548300053,
+      "loss": 3.3211,
+      "step": 47600
+    },
+    {
+      "epoch": 5.137466307277628,
+      "grad_norm": 0.6616727709770203,
+      "learning_rate": 0.0002921834862385321,
+      "loss": 3.3066,
+      "step": 47650
+    },
+    {
+      "epoch": 5.142857142857143,
+      "grad_norm": 0.6385394930839539,
+      "learning_rate": 0.0002918596869940637,
+      "loss": 3.3176,
+      "step": 47700
+    },
+    {
+      "epoch": 5.1482479784366575,
+      "grad_norm": 0.6855962872505188,
+      "learning_rate": 0.0002915358877495952,
+      "loss": 3.3303,
+      "step": 47750
+    },
+    {
+      "epoch": 5.153638814016173,
+      "grad_norm": 0.6520223617553711,
+      "learning_rate": 0.0002912120885051268,
+      "loss": 3.319,
+      "step": 47800
+    },
+    {
+      "epoch": 5.159029649595688,
+      "grad_norm": 0.7033669948577881,
+      "learning_rate": 0.00029088828926065834,
+      "loss": 3.3262,
+      "step": 47850
+    },
+    {
+      "epoch": 5.164420485175202,
+      "grad_norm": 0.7349123358726501,
+      "learning_rate": 0.00029056449001618994,
+      "loss": 3.3109,
+      "step": 47900
+    },
+    {
+      "epoch": 5.169811320754717,
+      "grad_norm": 0.7183597683906555,
+      "learning_rate": 0.0002902406907717215,
+      "loss": 3.3172,
+      "step": 47950
+    },
+    {
+      "epoch": 5.175202156334231,
+      "grad_norm": 0.690875768661499,
+      "learning_rate": 0.0002899168915272531,
+      "loss": 3.3263,
+      "step": 48000
+    },
+    {
+      "epoch": 5.175202156334231,
+      "eval_accuracy": 0.38107121099402075,
+      "eval_loss": 3.410407781600952,
+      "eval_runtime": 184.0011,
+      "eval_samples_per_second": 97.885,
+      "eval_steps_per_second": 6.12,
+      "step": 48000
+    },
+    {
+      "epoch": 5.180592991913747,
+      "grad_norm": 0.6746882796287537,
+      "learning_rate": 0.00028959309228278465,
+      "loss": 3.3005,
+      "step": 48050
+    },
+    {
+      "epoch": 5.185983827493262,
+      "grad_norm": 0.6469348669052124,
+      "learning_rate": 0.00028926929303831625,
+      "loss": 3.3094,
+      "step": 48100
+    },
+    {
+      "epoch": 5.191374663072776,
+      "grad_norm": 0.683721125125885,
+      "learning_rate": 0.0002889454937938478,
+      "loss": 3.318,
+      "step": 48150
+    },
+    {
+      "epoch": 5.196765498652291,
+      "grad_norm": 0.782717227935791,
+      "learning_rate": 0.00028862169454937935,
+      "loss": 3.3255,
+      "step": 48200
+    },
+    {
+      "epoch": 5.202156334231806,
+      "grad_norm": 0.7064043283462524,
+      "learning_rate": 0.0002882978953049109,
+      "loss": 3.3129,
+      "step": 48250
+    },
+    {
+      "epoch": 5.2075471698113205,
+      "grad_norm": 0.7118088006973267,
+      "learning_rate": 0.0002879740960604425,
+      "loss": 3.3106,
+      "step": 48300
+    },
+    {
+      "epoch": 5.212938005390836,
+      "grad_norm": 0.6886478066444397,
+      "learning_rate": 0.00028765029681597406,
+      "loss": 3.322,
+      "step": 48350
+    },
+    {
+      "epoch": 5.218328840970351,
+      "grad_norm": 0.6945542097091675,
+      "learning_rate": 0.00028732649757150566,
+      "loss": 3.3282,
+      "step": 48400
+    },
+    {
+      "epoch": 5.223719676549865,
+      "grad_norm": 0.7024538516998291,
+      "learning_rate": 0.0002870026983270372,
+      "loss": 3.2974,
+      "step": 48450
+    },
+    {
+      "epoch": 5.22911051212938,
+      "grad_norm": 0.680157482624054,
+      "learning_rate": 0.00028667889908256877,
+      "loss": 3.3084,
+      "step": 48500
+    },
+    {
+      "epoch": 5.234501347708895,
+      "grad_norm": 0.7121500968933105,
+      "learning_rate": 0.00028635509983810037,
+      "loss": 3.3226,
+      "step": 48550
+    },
+    {
+      "epoch": 5.2398921832884096,
+      "grad_norm": 0.7295412421226501,
+      "learning_rate": 0.0002860313005936319,
+      "loss": 3.3359,
+      "step": 48600
+    },
+    {
+      "epoch": 5.245283018867925,
+      "grad_norm": 0.6911020874977112,
+      "learning_rate": 0.00028570750134916347,
+      "loss": 3.3272,
+      "step": 48650
+    },
+    {
+      "epoch": 5.250673854447439,
+      "grad_norm": 0.7227579355239868,
+      "learning_rate": 0.0002853837021046951,
+      "loss": 3.3175,
+      "step": 48700
+    },
+    {
+      "epoch": 5.256064690026954,
+      "grad_norm": 0.64582359790802,
+      "learning_rate": 0.00028505990286022663,
+      "loss": 3.317,
+      "step": 48750
+    },
+    {
+      "epoch": 5.261455525606469,
+      "grad_norm": 0.7002373337745667,
+      "learning_rate": 0.0002847361036157582,
+      "loss": 3.3165,
+      "step": 48800
+    },
+    {
+      "epoch": 5.2668463611859835,
+      "grad_norm": 0.7050011157989502,
+      "learning_rate": 0.0002844123043712898,
+      "loss": 3.3082,
+      "step": 48850
+    },
+    {
+      "epoch": 5.272237196765499,
+      "grad_norm": 0.6821650266647339,
+      "learning_rate": 0.00028408850512682133,
+      "loss": 3.3348,
+      "step": 48900
+    },
+    {
+      "epoch": 5.277628032345014,
+      "grad_norm": 0.6783719658851624,
+      "learning_rate": 0.00028376470588235294,
+      "loss": 3.3126,
+      "step": 48950
+    },
+    {
+      "epoch": 5.283018867924528,
+      "grad_norm": 0.729580283164978,
+      "learning_rate": 0.0002834409066378845,
+      "loss": 3.3098,
+      "step": 49000
+    },
+    {
+      "epoch": 5.283018867924528,
+      "eval_accuracy": 0.3817646337240527,
+      "eval_loss": 3.4049880504608154,
+      "eval_runtime": 183.9025,
+      "eval_samples_per_second": 97.938,
+      "eval_steps_per_second": 6.123,
+      "step": 49000
+    },
+    {
+      "epoch": 5.288409703504043,
+      "grad_norm": 0.7196259498596191,
+      "learning_rate": 0.00028311710739341604,
+      "loss": 3.3177,
+      "step": 49050
+    },
+    {
+      "epoch": 5.293800539083558,
+      "grad_norm": 0.7036062479019165,
+      "learning_rate": 0.00028279330814894764,
+      "loss": 3.3319,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2991913746630726,
+      "grad_norm": 0.7958604097366333,
+      "learning_rate": 0.0002824695089044792,
+      "loss": 3.3291,
+      "step": 49150
+    },
+    {
+      "epoch": 5.304582210242588,
+      "grad_norm": 0.6777556538581848,
+      "learning_rate": 0.00028214570966001075,
+      "loss": 3.3213,
+      "step": 49200
+    },
+    {
+      "epoch": 5.309973045822103,
+      "grad_norm": 0.7265875935554504,
+      "learning_rate": 0.00028182191041554235,
+      "loss": 3.3201,
+      "step": 49250
+    },
+    {
+      "epoch": 5.315363881401617,
+      "grad_norm": 0.6683766841888428,
+      "learning_rate": 0.0002814981111710739,
+      "loss": 3.3334,
+      "step": 49300
+    },
+    {
+      "epoch": 5.320754716981132,
+      "grad_norm": 0.6723276972770691,
+      "learning_rate": 0.0002811743119266055,
+      "loss": 3.3208,
+      "step": 49350
+    },
+    {
+      "epoch": 5.3261455525606465,
+      "grad_norm": 0.674947202205658,
+      "learning_rate": 0.00028085051268213706,
+      "loss": 3.3206,
+      "step": 49400
+    },
+    {
+      "epoch": 5.331536388140162,
+      "grad_norm": 0.6870039701461792,
+      "learning_rate": 0.0002805267134376686,
+      "loss": 3.3373,
+      "step": 49450
+    },
+    {
+      "epoch": 5.336927223719677,
+      "grad_norm": 0.6902183890342712,
+      "learning_rate": 0.00028020291419320016,
+      "loss": 3.3207,
+      "step": 49500
+    },
+    {
+      "epoch": 5.342318059299191,
+      "grad_norm": 0.7033862471580505,
+      "learning_rate": 0.00027987911494873176,
+      "loss": 3.3258,
+      "step": 49550
+    },
+    {
+      "epoch": 5.347708894878706,
+      "grad_norm": 0.673235297203064,
+      "learning_rate": 0.0002795553157042633,
+      "loss": 3.3198,
+      "step": 49600
+    },
+    {
+      "epoch": 5.353099730458221,
+      "grad_norm": 0.7145218849182129,
+      "learning_rate": 0.0002792315164597949,
+      "loss": 3.3315,
+      "step": 49650
+    },
+    {
+      "epoch": 5.3584905660377355,
+      "grad_norm": 0.6835147738456726,
+      "learning_rate": 0.00027891419320021587,
+      "loss": 3.3322,
+      "step": 49700
+    },
+    {
+      "epoch": 5.363881401617251,
+      "grad_norm": 0.6946256160736084,
+      "learning_rate": 0.0002785903939557474,
+      "loss": 3.3398,
+      "step": 49750
+    },
+    {
+      "epoch": 5.369272237196766,
+      "grad_norm": 0.7527667284011841,
+      "learning_rate": 0.00027826659471127897,
+      "loss": 3.3284,
+      "step": 49800
+    },
+    {
+      "epoch": 5.37466307277628,
+      "grad_norm": 0.7169890999794006,
+      "learning_rate": 0.0002779427954668105,
+      "loss": 3.3206,
+      "step": 49850
+    },
+    {
+      "epoch": 5.380053908355795,
+      "grad_norm": 0.6716258525848389,
+      "learning_rate": 0.0002776189962223421,
+      "loss": 3.3221,
+      "step": 49900
+    },
+    {
+      "epoch": 5.38544474393531,
+      "grad_norm": 0.7162343263626099,
+      "learning_rate": 0.0002772951969778737,
+      "loss": 3.3214,
+      "step": 49950
+    },
+    {
+      "epoch": 5.390835579514825,
+      "grad_norm": 0.7313789129257202,
+      "learning_rate": 0.0002769713977334053,
+      "loss": 3.348,
+      "step": 50000
+    },
+    {
+      "epoch": 5.390835579514825,
+      "eval_accuracy": 0.3818524252574068,
+      "eval_loss": 3.4028215408325195,
+      "eval_runtime": 183.8802,
+      "eval_samples_per_second": 97.95,
+      "eval_steps_per_second": 6.124,
+      "step": 50000
+    },
+    {
+      "epoch": 5.39622641509434,
+      "grad_norm": 0.6996383666992188,
+      "learning_rate": 0.00027664759848893683,
+      "loss": 3.3008,
+      "step": 50050
+    },
+    {
+      "epoch": 5.401617250673855,
+      "grad_norm": 0.6793367862701416,
+      "learning_rate": 0.00027632379924446844,
+      "loss": 3.3247,
+      "step": 50100
+    },
+    {
+      "epoch": 5.407008086253369,
+      "grad_norm": 0.6811390519142151,
+      "learning_rate": 0.000276,
+      "loss": 3.3223,
+      "step": 50150
+    },
+    {
+      "epoch": 5.412398921832884,
+      "grad_norm": 0.7058961391448975,
+      "learning_rate": 0.00027567620075553154,
+      "loss": 3.3379,
+      "step": 50200
+    },
+    {
+      "epoch": 5.4177897574123985,
+      "grad_norm": 0.7083661556243896,
+      "learning_rate": 0.0002753524015110631,
+      "loss": 3.3497,
+      "step": 50250
+    },
+    {
+      "epoch": 5.423180592991914,
+      "grad_norm": 0.6529145836830139,
+      "learning_rate": 0.0002750286022665947,
+      "loss": 3.3276,
+      "step": 50300
+    },
+    {
+      "epoch": 5.428571428571429,
+      "grad_norm": 0.6388300061225891,
+      "learning_rate": 0.00027470480302212624,
+      "loss": 3.3116,
+      "step": 50350
+    },
+    {
+      "epoch": 5.433962264150943,
+      "grad_norm": 0.6704314947128296,
+      "learning_rate": 0.00027438100377765785,
+      "loss": 3.3319,
+      "step": 50400
+    },
+    {
+      "epoch": 5.439353099730458,
+      "grad_norm": 0.7339666485786438,
+      "learning_rate": 0.0002740572045331894,
+      "loss": 3.3106,
+      "step": 50450
+    },
+    {
+      "epoch": 5.444743935309973,
+      "grad_norm": 0.695424497127533,
+      "learning_rate": 0.00027373340528872095,
+      "loss": 3.3386,
+      "step": 50500
+    },
+    {
+      "epoch": 5.450134770889488,
+      "grad_norm": 0.700162410736084,
+      "learning_rate": 0.00027340960604425255,
+      "loss": 3.3246,
+      "step": 50550
+    },
+    {
+      "epoch": 5.455525606469003,
+      "grad_norm": 0.7001470923423767,
+      "learning_rate": 0.0002730858067997841,
+      "loss": 3.3226,
+      "step": 50600
+    },
+    {
+      "epoch": 5.460916442048518,
+      "grad_norm": 0.6577569842338562,
+      "learning_rate": 0.00027276200755531565,
+      "loss": 3.3271,
+      "step": 50650
+    },
+    {
+      "epoch": 5.466307277628032,
+      "grad_norm": 0.6836386322975159,
+      "learning_rate": 0.00027243820831084726,
+      "loss": 3.3389,
+      "step": 50700
+    },
+    {
+      "epoch": 5.471698113207547,
+      "grad_norm": 0.7137777805328369,
+      "learning_rate": 0.0002721144090663788,
+      "loss": 3.327,
+      "step": 50750
+    },
+    {
+      "epoch": 5.4770889487870615,
+      "grad_norm": 0.6904305219650269,
+      "learning_rate": 0.00027179060982191036,
+      "loss": 3.3095,
+      "step": 50800
+    },
+    {
+      "epoch": 5.482479784366577,
+      "grad_norm": 0.7418445944786072,
+      "learning_rate": 0.00027146681057744197,
+      "loss": 3.3346,
+      "step": 50850
+    },
+    {
+      "epoch": 5.487870619946092,
+      "grad_norm": 0.6904473304748535,
+      "learning_rate": 0.0002711430113329735,
+      "loss": 3.3331,
+      "step": 50900
+    },
+    {
+      "epoch": 5.493261455525606,
+      "grad_norm": 0.7091851830482483,
+      "learning_rate": 0.0002708192120885051,
+      "loss": 3.3373,
+      "step": 50950
+    },
+    {
+      "epoch": 5.498652291105121,
+      "grad_norm": 0.7276546359062195,
+      "learning_rate": 0.00027049541284403667,
+      "loss": 3.331,
+      "step": 51000
+    },
+    {
+      "epoch": 5.498652291105121,
+      "eval_accuracy": 0.3823172423114772,
+      "eval_loss": 3.39615797996521,
+      "eval_runtime": 184.0329,
+      "eval_samples_per_second": 97.868,
+      "eval_steps_per_second": 6.118,
+      "step": 51000
+    },
+    {
+      "epoch": 5.504043126684636,
+      "grad_norm": 0.6925030946731567,
+      "learning_rate": 0.0002701716135995683,
+      "loss": 3.3433,
+      "step": 51050
+    },
+    {
+      "epoch": 5.509433962264151,
+      "grad_norm": 0.652752161026001,
+      "learning_rate": 0.0002698478143550998,
+      "loss": 3.3233,
+      "step": 51100
+    },
+    {
+      "epoch": 5.514824797843666,
+      "grad_norm": 0.7031452655792236,
+      "learning_rate": 0.0002695240151106314,
+      "loss": 3.3256,
+      "step": 51150
+    },
+    {
+      "epoch": 5.520215633423181,
+      "grad_norm": 0.7321648597717285,
+      "learning_rate": 0.00026920021586616293,
+      "loss": 3.307,
+      "step": 51200
+    },
+    {
+      "epoch": 5.525606469002695,
+      "grad_norm": 0.6757856011390686,
+      "learning_rate": 0.00026887641662169453,
+      "loss": 3.3193,
+      "step": 51250
+    },
+    {
+      "epoch": 5.53099730458221,
+      "grad_norm": 0.7370916604995728,
+      "learning_rate": 0.0002685526173772261,
+      "loss": 3.3289,
+      "step": 51300
+    },
+    {
+      "epoch": 5.536388140161725,
+      "grad_norm": 0.7094478011131287,
+      "learning_rate": 0.0002682288181327577,
+      "loss": 3.3272,
+      "step": 51350
+    },
+    {
+      "epoch": 5.54177897574124,
+      "grad_norm": 0.6774241924285889,
+      "learning_rate": 0.00026790501888828924,
+      "loss": 3.3227,
+      "step": 51400
+    },
+    {
+      "epoch": 5.547169811320755,
+      "grad_norm": 0.7134442925453186,
+      "learning_rate": 0.00026758121964382084,
+      "loss": 3.3302,
+      "step": 51450
+    },
+    {
+      "epoch": 5.55256064690027,
+      "grad_norm": 0.7070966958999634,
+      "learning_rate": 0.00026725742039935234,
+      "loss": 3.3249,
+      "step": 51500
+    },
+    {
+      "epoch": 5.557951482479784,
+      "grad_norm": 0.7182289361953735,
+      "learning_rate": 0.00026693362115488394,
+      "loss": 3.3378,
+      "step": 51550
+    },
+    {
+      "epoch": 5.563342318059299,
+      "grad_norm": 0.6908033490180969,
+      "learning_rate": 0.0002666098219104155,
+      "loss": 3.3393,
+      "step": 51600
+    },
+    {
+      "epoch": 5.568733153638814,
+      "grad_norm": 0.7248824238777161,
+      "learning_rate": 0.0002662860226659471,
+      "loss": 3.3423,
+      "step": 51650
+    },
+    {
+      "epoch": 5.574123989218329,
+      "grad_norm": 0.6812718510627747,
+      "learning_rate": 0.00026596222342147865,
+      "loss": 3.3402,
+      "step": 51700
+    },
+    {
+      "epoch": 5.579514824797844,
+      "grad_norm": 0.6986221671104431,
+      "learning_rate": 0.00026563842417701026,
+      "loss": 3.3352,
+      "step": 51750
+    },
+    {
+      "epoch": 5.584905660377358,
+      "grad_norm": 0.6742642521858215,
+      "learning_rate": 0.0002653146249325418,
+      "loss": 3.3482,
+      "step": 51800
+    },
+    {
+      "epoch": 5.590296495956873,
+      "grad_norm": 0.667993426322937,
+      "learning_rate": 0.00026499082568807336,
+      "loss": 3.3278,
+      "step": 51850
+    },
+    {
+      "epoch": 5.595687331536388,
+      "grad_norm": 0.739936888217926,
+      "learning_rate": 0.00026466702644360496,
+      "loss": 3.3371,
+      "step": 51900
+    },
+    {
+      "epoch": 5.601078167115903,
+      "grad_norm": 0.6672866344451904,
+      "learning_rate": 0.0002643432271991365,
+      "loss": 3.3375,
+      "step": 51950
+    },
+    {
+      "epoch": 5.606469002695418,
+      "grad_norm": 0.7498651742935181,
+      "learning_rate": 0.00026401942795466806,
+      "loss": 3.3412,
+      "step": 52000
+    },
+    {
+      "epoch": 5.606469002695418,
+      "eval_accuracy": 0.3828715893451067,
+      "eval_loss": 3.3931350708007812,
+      "eval_runtime": 183.9258,
+      "eval_samples_per_second": 97.925,
+      "eval_steps_per_second": 6.122,
+      "step": 52000
+    },
+    {
+      "epoch": 5.611859838274933,
+      "grad_norm": 0.7274742126464844,
+      "learning_rate": 0.00026369562871019967,
+      "loss": 3.3442,
+      "step": 52050
+    },
+    {
+      "epoch": 5.617250673854447,
+      "grad_norm": 0.7057594656944275,
+      "learning_rate": 0.0002633783054506206,
+      "loss": 3.3284,
+      "step": 52100
+    },
+    {
+      "epoch": 5.622641509433962,
+      "grad_norm": 0.671108067035675,
+      "learning_rate": 0.00026305450620615217,
+      "loss": 3.334,
+      "step": 52150
+    },
+    {
+      "epoch": 5.628032345013477,
+      "grad_norm": 0.7190834283828735,
+      "learning_rate": 0.0002627307069616837,
+      "loss": 3.3304,
+      "step": 52200
+    },
+    {
+      "epoch": 5.633423180592992,
+      "grad_norm": 0.6813502907752991,
+      "learning_rate": 0.0002624069077172153,
+      "loss": 3.3345,
+      "step": 52250
+    },
+    {
+      "epoch": 5.638814016172507,
+      "grad_norm": 0.6833723187446594,
+      "learning_rate": 0.0002620831084727469,
+      "loss": 3.3245,
+      "step": 52300
+    },
+    {
+      "epoch": 5.644204851752022,
+      "grad_norm": 0.7639450430870056,
+      "learning_rate": 0.0002617593092282784,
+      "loss": 3.3322,
+      "step": 52350
+    },
+    {
+      "epoch": 5.649595687331536,
+      "grad_norm": 0.6970094442367554,
+      "learning_rate": 0.00026143550998381003,
+      "loss": 3.3326,
+      "step": 52400
+    },
+    {
+      "epoch": 5.654986522911051,
+      "grad_norm": 0.6890552043914795,
+      "learning_rate": 0.0002611117107393416,
+      "loss": 3.3346,
+      "step": 52450
+    },
+    {
+      "epoch": 5.660377358490566,
+      "grad_norm": 0.6932628154754639,
+      "learning_rate": 0.00026078791149487313,
+      "loss": 3.3241,
+      "step": 52500
+    },
+    {
+      "epoch": 5.665768194070081,
+      "grad_norm": 0.683397650718689,
+      "learning_rate": 0.00026046411225040474,
+      "loss": 3.3241,
+      "step": 52550
+    },
+    {
+      "epoch": 5.671159029649596,
+      "grad_norm": 0.6974855661392212,
+      "learning_rate": 0.0002601403130059363,
+      "loss": 3.3463,
+      "step": 52600
+    },
+    {
+      "epoch": 5.67654986522911,
+      "grad_norm": 0.6833411455154419,
+      "learning_rate": 0.0002598165137614679,
+      "loss": 3.3375,
+      "step": 52650
+    },
+    {
+      "epoch": 5.681940700808625,
+      "grad_norm": 0.7359110713005066,
+      "learning_rate": 0.00025949271451699944,
+      "loss": 3.3277,
+      "step": 52700
+    },
+    {
+      "epoch": 5.6873315363881405,
+      "grad_norm": 0.7000057697296143,
+      "learning_rate": 0.000259168915272531,
+      "loss": 3.3307,
+      "step": 52750
+    },
+    {
+      "epoch": 5.692722371967655,
+      "grad_norm": 0.7026528120040894,
+      "learning_rate": 0.0002588451160280626,
+      "loss": 3.3244,
+      "step": 52800
+    },
+    {
+      "epoch": 5.69811320754717,
+      "grad_norm": 0.7201850414276123,
+      "learning_rate": 0.00025852131678359415,
+      "loss": 3.3252,
+      "step": 52850
+    },
+    {
+      "epoch": 5.703504043126685,
+      "grad_norm": 0.7258884310722351,
+      "learning_rate": 0.0002581975175391257,
+      "loss": 3.3325,
+      "step": 52900
+    },
+    {
+      "epoch": 5.708894878706199,
+      "grad_norm": 0.7302329540252686,
+      "learning_rate": 0.0002578737182946573,
+      "loss": 3.3446,
+      "step": 52950
+    },
+    {
+      "epoch": 5.714285714285714,
+      "grad_norm": 0.7130410075187683,
+      "learning_rate": 0.00025754991905018885,
+      "loss": 3.3414,
+      "step": 53000
+    },
+    {
+      "epoch": 5.714285714285714,
+      "eval_accuracy": 0.38328479627746515,
+      "eval_loss": 3.3882997035980225,
+      "eval_runtime": 183.7964,
+      "eval_samples_per_second": 97.994,
+      "eval_steps_per_second": 6.126,
+      "step": 53000
+    },
+    {
+      "epoch": 5.719676549865229,
+      "grad_norm": 0.7381608486175537,
+      "learning_rate": 0.00025722611980572046,
+      "loss": 3.3485,
+      "step": 53050
+    },
+    {
+      "epoch": 5.725067385444744,
+      "grad_norm": 0.6873294115066528,
+      "learning_rate": 0.000256902320561252,
+      "loss": 3.3458,
+      "step": 53100
+    },
+    {
+      "epoch": 5.730458221024259,
+      "grad_norm": 0.683340847492218,
+      "learning_rate": 0.00025657852131678356,
+      "loss": 3.3562,
+      "step": 53150
+    },
+    {
+      "epoch": 5.735849056603773,
+      "grad_norm": 0.7045843005180359,
+      "learning_rate": 0.0002562547220723151,
+      "loss": 3.3257,
+      "step": 53200
+    },
+    {
+      "epoch": 5.741239892183288,
+      "grad_norm": 0.7063679695129395,
+      "learning_rate": 0.0002559309228278467,
+      "loss": 3.3495,
+      "step": 53250
+    },
+    {
+      "epoch": 5.7466307277628035,
+      "grad_norm": 0.7272491455078125,
+      "learning_rate": 0.00025560712358337827,
+      "loss": 3.3404,
+      "step": 53300
+    },
+    {
+      "epoch": 5.752021563342318,
+      "grad_norm": 0.7092030644416809,
+      "learning_rate": 0.00025528332433890987,
+      "loss": 3.3261,
+      "step": 53350
+    },
+    {
+      "epoch": 5.757412398921833,
+      "grad_norm": 0.7118086218833923,
+      "learning_rate": 0.0002549595250944414,
+      "loss": 3.3423,
+      "step": 53400
+    },
+    {
+      "epoch": 5.762803234501348,
+      "grad_norm": 0.7487428188323975,
+      "learning_rate": 0.000254635725849973,
+      "loss": 3.3416,
+      "step": 53450
+    },
+    {
+      "epoch": 5.768194070080862,
+      "grad_norm": 0.665637731552124,
+      "learning_rate": 0.0002543119266055046,
+      "loss": 3.3426,
+      "step": 53500
+    },
+    {
+      "epoch": 5.773584905660377,
+      "grad_norm": 0.7153809666633606,
+      "learning_rate": 0.00025398812736103613,
+      "loss": 3.3474,
+      "step": 53550
+    },
+    {
+      "epoch": 5.7789757412398925,
+      "grad_norm": 0.7571694850921631,
+      "learning_rate": 0.0002536643281165677,
+      "loss": 3.3268,
+      "step": 53600
+    },
+    {
+      "epoch": 5.784366576819407,
+      "grad_norm": 0.7039183974266052,
+      "learning_rate": 0.0002533405288720993,
+      "loss": 3.3451,
+      "step": 53650
+    },
+    {
+      "epoch": 5.789757412398922,
+      "grad_norm": 0.7550934553146362,
+      "learning_rate": 0.00025301672962763083,
+      "loss": 3.3295,
+      "step": 53700
+    },
+    {
+      "epoch": 5.795148247978437,
+      "grad_norm": 0.7558539509773254,
+      "learning_rate": 0.00025269293038316244,
+      "loss": 3.3295,
+      "step": 53750
+    },
+    {
+      "epoch": 5.800539083557951,
+      "grad_norm": 0.6939939856529236,
+      "learning_rate": 0.000252369131138694,
+      "loss": 3.3275,
+      "step": 53800
+    },
+    {
+      "epoch": 5.8059299191374665,
+      "grad_norm": 0.7229671478271484,
+      "learning_rate": 0.00025204533189422554,
+      "loss": 3.3375,
+      "step": 53850
+    },
+    {
+      "epoch": 5.811320754716981,
+      "grad_norm": 0.7354257702827454,
+      "learning_rate": 0.00025172153264975714,
+      "loss": 3.3143,
+      "step": 53900
+    },
+    {
+      "epoch": 5.816711590296496,
+      "grad_norm": 0.7292751669883728,
+      "learning_rate": 0.0002513977334052887,
+      "loss": 3.3298,
+      "step": 53950
+    },
+    {
+      "epoch": 5.822102425876011,
+      "grad_norm": 0.7425629496574402,
+      "learning_rate": 0.00025107393416082025,
+      "loss": 3.3438,
+      "step": 54000
+    },
+    {
+      "epoch": 5.822102425876011,
+      "eval_accuracy": 0.3835859820824869,
+      "eval_loss": 3.382173776626587,
+      "eval_runtime": 184.3297,
+      "eval_samples_per_second": 97.711,
+      "eval_steps_per_second": 6.109,
+      "step": 54000
+    },
+    {
+      "epoch": 5.827493261455525,
+      "grad_norm": 0.7313019037246704,
+      "learning_rate": 0.00025075013491635185,
+      "loss": 3.3422,
+      "step": 54050
+    },
+    {
+      "epoch": 5.83288409703504,
+      "grad_norm": 0.8147013783454895,
+      "learning_rate": 0.0002504263356718834,
+      "loss": 3.3338,
+      "step": 54100
+    },
+    {
+      "epoch": 5.8382749326145555,
+      "grad_norm": 0.7204222083091736,
+      "learning_rate": 0.000250102536427415,
+      "loss": 3.3189,
+      "step": 54150
+    },
+    {
+      "epoch": 5.84366576819407,
+      "grad_norm": 0.6922837495803833,
+      "learning_rate": 0.00024977873718294656,
+      "loss": 3.317,
+      "step": 54200
+    },
+    {
+      "epoch": 5.849056603773585,
+      "grad_norm": 0.7175344228744507,
+      "learning_rate": 0.0002494549379384781,
+      "loss": 3.3275,
+      "step": 54250
+    },
+    {
+      "epoch": 5.8544474393531,
+      "grad_norm": 0.6663742065429688,
+      "learning_rate": 0.0002491311386940097,
+      "loss": 3.3034,
+      "step": 54300
+    },
+    {
+      "epoch": 5.859838274932614,
+      "grad_norm": 0.6701042056083679,
+      "learning_rate": 0.00024880733944954126,
+      "loss": 3.3596,
+      "step": 54350
+    },
+    {
+      "epoch": 5.8652291105121295,
+      "grad_norm": 0.7388389706611633,
+      "learning_rate": 0.0002484835402050728,
+      "loss": 3.3137,
+      "step": 54400
+    },
+    {
+      "epoch": 5.870619946091644,
+      "grad_norm": 0.7365710735321045,
+      "learning_rate": 0.0002481597409606044,
+      "loss": 3.3216,
+      "step": 54450
+    },
+    {
+      "epoch": 5.876010781671159,
+      "grad_norm": 0.710827112197876,
+      "learning_rate": 0.00024783594171613597,
+      "loss": 3.3269,
+      "step": 54500
+    },
+    {
+      "epoch": 5.881401617250674,
+      "grad_norm": 0.7242414951324463,
+      "learning_rate": 0.0002475121424716675,
+      "loss": 3.3325,
+      "step": 54550
+    },
+    {
+      "epoch": 5.886792452830189,
+      "grad_norm": 0.7195931077003479,
+      "learning_rate": 0.0002471883432271991,
+      "loss": 3.3223,
+      "step": 54600
+    },
+    {
+      "epoch": 5.892183288409703,
+      "grad_norm": 0.6709275841712952,
+      "learning_rate": 0.0002468645439827307,
+      "loss": 3.3036,
+      "step": 54650
+    },
+    {
+      "epoch": 5.8975741239892185,
+      "grad_norm": 0.6818590760231018,
+      "learning_rate": 0.0002465407447382623,
+      "loss": 3.331,
+      "step": 54700
+    },
+    {
+      "epoch": 5.902964959568733,
+      "grad_norm": 0.723003089427948,
+      "learning_rate": 0.00024621694549379383,
+      "loss": 3.3229,
+      "step": 54750
+    },
+    {
+      "epoch": 5.908355795148248,
+      "grad_norm": 0.6944183111190796,
+      "learning_rate": 0.00024589314624932543,
+      "loss": 3.351,
+      "step": 54800
+    },
+    {
+      "epoch": 5.913746630727763,
+      "grad_norm": 0.7226275205612183,
+      "learning_rate": 0.00024556934700485693,
+      "loss": 3.3272,
+      "step": 54850
+    },
+    {
+      "epoch": 5.919137466307277,
+      "grad_norm": 0.6976397633552551,
+      "learning_rate": 0.00024524554776038853,
+      "loss": 3.3186,
+      "step": 54900
+    },
+    {
+      "epoch": 5.9245283018867925,
+      "grad_norm": 0.7749685049057007,
+      "learning_rate": 0.0002449217485159201,
+      "loss": 3.3312,
+      "step": 54950
+    },
+    {
+      "epoch": 5.929919137466308,
+      "grad_norm": 0.7344666719436646,
+      "learning_rate": 0.0002445979492714517,
+      "loss": 3.3077,
+      "step": 55000
+    },
+    {
+      "epoch": 5.929919137466308,
+      "eval_accuracy": 0.3844160901453895,
+      "eval_loss": 3.3785133361816406,
+      "eval_runtime": 183.7531,
+      "eval_samples_per_second": 98.017,
+      "eval_steps_per_second": 6.128,
+      "step": 55000
+    },
+    {
+      "epoch": 5.935309973045822,
+      "grad_norm": 0.6995274424552917,
+      "learning_rate": 0.00024427415002698324,
+      "loss": 3.3377,
+      "step": 55050
+    },
+    {
+      "epoch": 5.940700808625337,
+      "grad_norm": 0.7057139277458191,
+      "learning_rate": 0.00024395035078251482,
+      "loss": 3.3364,
+      "step": 55100
+    },
+    {
+      "epoch": 5.946091644204852,
+      "grad_norm": 0.7679343819618225,
+      "learning_rate": 0.0002436265515380464,
+      "loss": 3.3456,
+      "step": 55150
+    },
+    {
+      "epoch": 5.951482479784366,
+      "grad_norm": 0.7506498098373413,
+      "learning_rate": 0.00024330275229357797,
+      "loss": 3.3296,
+      "step": 55200
+    },
+    {
+      "epoch": 5.9568733153638815,
+      "grad_norm": 0.7279877662658691,
+      "learning_rate": 0.00024297895304910952,
+      "loss": 3.3316,
+      "step": 55250
+    },
+    {
+      "epoch": 5.962264150943396,
+      "grad_norm": 0.686543881893158,
+      "learning_rate": 0.0002426551538046411,
+      "loss": 3.2993,
+      "step": 55300
+    },
+    {
+      "epoch": 5.967654986522911,
+      "grad_norm": 0.7167565226554871,
+      "learning_rate": 0.00024233135456017265,
+      "loss": 3.328,
+      "step": 55350
+    },
+    {
+      "epoch": 5.973045822102426,
+      "grad_norm": 0.7074522972106934,
+      "learning_rate": 0.00024200755531570423,
+      "loss": 3.3511,
+      "step": 55400
+    },
+    {
+      "epoch": 5.97843665768194,
+      "grad_norm": 0.6831624507904053,
+      "learning_rate": 0.0002416837560712358,
+      "loss": 3.3474,
+      "step": 55450
+    },
+    {
+      "epoch": 5.9838274932614555,
+      "grad_norm": 0.7267654538154602,
+      "learning_rate": 0.00024135995682676739,
+      "loss": 3.3315,
+      "step": 55500
+    },
+    {
+      "epoch": 5.989218328840971,
+      "grad_norm": 0.7426101565361023,
+      "learning_rate": 0.00024103615758229896,
+      "loss": 3.3361,
+      "step": 55550
+    },
+    {
+      "epoch": 5.994609164420485,
+      "grad_norm": 0.7580520510673523,
+      "learning_rate": 0.00024071235833783054,
+      "loss": 3.3454,
+      "step": 55600
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 1.4545377492904663,
+      "learning_rate": 0.00024038855909336212,
+      "loss": 3.3377,
+      "step": 55650
+    },
+    {
+      "epoch": 6.005390835579515,
+      "grad_norm": 0.6739314198493958,
+      "learning_rate": 0.00024006475984889364,
+      "loss": 3.2288,
+      "step": 55700
+    },
+    {
+      "epoch": 6.010781671159029,
+      "grad_norm": 0.6961607933044434,
+      "learning_rate": 0.00023974096060442522,
+      "loss": 3.2358,
+      "step": 55750
+    },
+    {
+      "epoch": 6.0161725067385445,
+      "grad_norm": 0.7214361429214478,
+      "learning_rate": 0.0002394171613599568,
+      "loss": 3.2439,
+      "step": 55800
+    },
+    {
+      "epoch": 6.02156334231806,
+      "grad_norm": 0.7168677449226379,
+      "learning_rate": 0.00023909336211548837,
+      "loss": 3.2532,
+      "step": 55850
+    },
+    {
+      "epoch": 6.026954177897574,
+      "grad_norm": 0.7242555618286133,
+      "learning_rate": 0.00023876956287101995,
+      "loss": 3.2447,
+      "step": 55900
+    },
+    {
+      "epoch": 6.032345013477089,
+      "grad_norm": 0.7259112596511841,
+      "learning_rate": 0.00023844576362655153,
+      "loss": 3.2433,
+      "step": 55950
+    },
+    {
+      "epoch": 6.037735849056604,
+      "grad_norm": 0.7300722002983093,
+      "learning_rate": 0.0002381219643820831,
+      "loss": 3.2654,
+      "step": 56000
+    },
+    {
+      "epoch": 6.037735849056604,
+      "eval_accuracy": 0.38417759705663673,
+      "eval_loss": 3.380847454071045,
+      "eval_runtime": 183.8731,
+      "eval_samples_per_second": 97.953,
+      "eval_steps_per_second": 6.124,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0431266846361185,
+      "grad_norm": 0.7531189322471619,
+      "learning_rate": 0.00023779816513761466,
+      "loss": 3.2651,
+      "step": 56050
+    },
+    {
+      "epoch": 6.048517520215634,
+      "grad_norm": 0.7250877022743225,
+      "learning_rate": 0.00023748084187803558,
+      "loss": 3.2445,
+      "step": 56100
+    },
+    {
+      "epoch": 6.053908355795148,
+      "grad_norm": 0.7453233599662781,
+      "learning_rate": 0.00023715704263356716,
+      "loss": 3.2522,
+      "step": 56150
+    },
+    {
+      "epoch": 6.059299191374663,
+      "grad_norm": 0.7315291166305542,
+      "learning_rate": 0.00023683324338909874,
+      "loss": 3.2403,
+      "step": 56200
+    },
+    {
+      "epoch": 6.064690026954178,
+      "grad_norm": 0.7161756157875061,
+      "learning_rate": 0.00023650944414463032,
+      "loss": 3.2553,
+      "step": 56250
+    },
+    {
+      "epoch": 6.070080862533692,
+      "grad_norm": 0.7513576745986938,
+      "learning_rate": 0.0002361856449001619,
+      "loss": 3.2478,
+      "step": 56300
+    },
+    {
+      "epoch": 6.0754716981132075,
+      "grad_norm": 0.7142834663391113,
+      "learning_rate": 0.00023586184565569347,
+      "loss": 3.2543,
+      "step": 56350
+    },
+    {
+      "epoch": 6.080862533692723,
+      "grad_norm": 0.6998046040534973,
+      "learning_rate": 0.00023553804641122502,
+      "loss": 3.2514,
+      "step": 56400
+    },
+    {
+      "epoch": 6.086253369272237,
+      "grad_norm": 0.7207847833633423,
+      "learning_rate": 0.00023521424716675657,
+      "loss": 3.2571,
+      "step": 56450
+    },
+    {
+      "epoch": 6.091644204851752,
+      "grad_norm": 0.7710952162742615,
+      "learning_rate": 0.00023489044792228815,
+      "loss": 3.2567,
+      "step": 56500
+    },
+    {
+      "epoch": 6.097035040431267,
+      "grad_norm": 0.7176562547683716,
+      "learning_rate": 0.00023456664867781973,
+      "loss": 3.2456,
+      "step": 56550
+    },
+    {
+      "epoch": 6.1024258760107815,
+      "grad_norm": 0.7144829630851746,
+      "learning_rate": 0.0002342428494333513,
+      "loss": 3.2536,
+      "step": 56600
+    },
+    {
+      "epoch": 6.107816711590297,
+      "grad_norm": 0.7680326104164124,
+      "learning_rate": 0.00023391905018888288,
+      "loss": 3.2511,
+      "step": 56650
+    },
+    {
+      "epoch": 6.113207547169812,
+      "grad_norm": 0.7204803824424744,
+      "learning_rate": 0.00023359525094441443,
+      "loss": 3.2494,
+      "step": 56700
+    },
+    {
+      "epoch": 6.118598382749326,
+      "grad_norm": 0.7161763310432434,
+      "learning_rate": 0.000233271451699946,
+      "loss": 3.2641,
+      "step": 56750
+    },
+    {
+      "epoch": 6.123989218328841,
+      "grad_norm": 0.7113860845565796,
+      "learning_rate": 0.0002329476524554776,
+      "loss": 3.2594,
+      "step": 56800
+    },
+    {
+      "epoch": 6.129380053908355,
+      "grad_norm": 0.710684061050415,
+      "learning_rate": 0.00023262385321100917,
+      "loss": 3.2647,
+      "step": 56850
+    },
+    {
+      "epoch": 6.1347708894878705,
+      "grad_norm": 0.7076157331466675,
+      "learning_rate": 0.00023230005396654072,
+      "loss": 3.2609,
+      "step": 56900
+    },
+    {
+      "epoch": 6.140161725067386,
+      "grad_norm": 0.7792293429374695,
+      "learning_rate": 0.0002319762547220723,
+      "loss": 3.2541,
+      "step": 56950
+    },
+    {
+      "epoch": 6.1455525606469,
+      "grad_norm": 0.7125359773635864,
+      "learning_rate": 0.00023165245547760387,
+      "loss": 3.2674,
+      "step": 57000
+    },
+    {
+      "epoch": 6.1455525606469,
+      "eval_accuracy": 0.38459482414584434,
+      "eval_loss": 3.3812313079833984,
+      "eval_runtime": 184.1175,
+      "eval_samples_per_second": 97.823,
+      "eval_steps_per_second": 6.116,
+      "step": 57000
+    },
+    {
+      "epoch": 6.150943396226415,
+      "grad_norm": 0.758549153804779,
+      "learning_rate": 0.00023132865623313542,
+      "loss": 3.2666,
+      "step": 57050
+    },
+    {
+      "epoch": 6.15633423180593,
+      "grad_norm": 0.7191378474235535,
+      "learning_rate": 0.000231004856988667,
+      "loss": 3.2799,
+      "step": 57100
+    },
+    {
+      "epoch": 6.1617250673854445,
+      "grad_norm": 0.7202212810516357,
+      "learning_rate": 0.00023068105774419858,
+      "loss": 3.2571,
+      "step": 57150
+    },
+    {
+      "epoch": 6.16711590296496,
+      "grad_norm": 0.7385558485984802,
+      "learning_rate": 0.00023035725849973016,
+      "loss": 3.2776,
+      "step": 57200
+    },
+    {
+      "epoch": 6.172506738544475,
+      "grad_norm": 0.740178644657135,
+      "learning_rate": 0.00023003345925526173,
+      "loss": 3.2691,
+      "step": 57250
+    },
+    {
+      "epoch": 6.177897574123989,
+      "grad_norm": 0.741432785987854,
+      "learning_rate": 0.00022970966001079328,
+      "loss": 3.2609,
+      "step": 57300
+    },
+    {
+      "epoch": 6.183288409703504,
+      "grad_norm": 0.7559853792190552,
+      "learning_rate": 0.00022938586076632484,
+      "loss": 3.2646,
+      "step": 57350
+    },
+    {
+      "epoch": 6.188679245283019,
+      "grad_norm": 0.7264957427978516,
+      "learning_rate": 0.0002290620615218564,
+      "loss": 3.2543,
+      "step": 57400
+    },
+    {
+      "epoch": 6.1940700808625335,
+      "grad_norm": 0.728132426738739,
+      "learning_rate": 0.000228738262277388,
+      "loss": 3.2698,
+      "step": 57450
+    },
+    {
+      "epoch": 6.199460916442049,
+      "grad_norm": 0.7514779567718506,
+      "learning_rate": 0.00022841446303291957,
+      "loss": 3.2635,
+      "step": 57500
+    },
+    {
+      "epoch": 6.204851752021563,
+      "grad_norm": 0.7553060054779053,
+      "learning_rate": 0.00022809066378845115,
+      "loss": 3.2622,
+      "step": 57550
+    },
+    {
+      "epoch": 6.210242587601078,
+      "grad_norm": 0.7538262605667114,
+      "learning_rate": 0.0002277733405288721,
+      "loss": 3.2754,
+      "step": 57600
+    },
+    {
+      "epoch": 6.215633423180593,
+      "grad_norm": 0.7464041113853455,
+      "learning_rate": 0.00022744954128440365,
+      "loss": 3.2429,
+      "step": 57650
+    },
+    {
+      "epoch": 6.2210242587601075,
+      "grad_norm": 0.7480874061584473,
+      "learning_rate": 0.0002271257420399352,
+      "loss": 3.266,
+      "step": 57700
+    },
+    {
+      "epoch": 6.226415094339623,
+      "grad_norm": 0.7676622867584229,
+      "learning_rate": 0.00022680194279546678,
+      "loss": 3.2746,
+      "step": 57750
+    },
+    {
+      "epoch": 6.231805929919138,
+      "grad_norm": 0.7216105461120605,
+      "learning_rate": 0.00022647814355099835,
+      "loss": 3.2647,
+      "step": 57800
+    },
+    {
+      "epoch": 6.237196765498652,
+      "grad_norm": 0.7596701383590698,
+      "learning_rate": 0.00022615434430652993,
+      "loss": 3.2692,
+      "step": 57850
+    },
+    {
+      "epoch": 6.242587601078167,
+      "grad_norm": 0.7444921135902405,
+      "learning_rate": 0.0002258305450620615,
+      "loss": 3.2524,
+      "step": 57900
+    },
+    {
+      "epoch": 6.247978436657682,
+      "grad_norm": 0.757045567035675,
+      "learning_rate": 0.0002255067458175931,
+      "loss": 3.2445,
+      "step": 57950
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "grad_norm": 0.7225556373596191,
+      "learning_rate": 0.00022518294657312467,
+      "loss": 3.2691,
+      "step": 58000
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "eval_accuracy": 0.3846125345665581,
+      "eval_loss": 3.3798320293426514,
+      "eval_runtime": 183.8445,
+      "eval_samples_per_second": 97.969,
+      "eval_steps_per_second": 6.125,
+      "step": 58000
+    },
+    {
+      "epoch": 6.258760107816712,
+      "grad_norm": 0.7398164868354797,
+      "learning_rate": 0.0002248591473286562,
+      "loss": 3.268,
+      "step": 58050
+    },
+    {
+      "epoch": 6.264150943396227,
+      "grad_norm": 0.7138086557388306,
+      "learning_rate": 0.00022453534808418777,
+      "loss": 3.2739,
+      "step": 58100
+    },
+    {
+      "epoch": 6.269541778975741,
+      "grad_norm": 0.7250698804855347,
+      "learning_rate": 0.00022421154883971934,
+      "loss": 3.2892,
+      "step": 58150
+    },
+    {
+      "epoch": 6.274932614555256,
+      "grad_norm": 0.735781192779541,
+      "learning_rate": 0.00022388774959525092,
+      "loss": 3.2771,
+      "step": 58200
+    },
+    {
+      "epoch": 6.280323450134771,
+      "grad_norm": 0.6944800019264221,
+      "learning_rate": 0.0002235639503507825,
+      "loss": 3.2606,
+      "step": 58250
+    },
+    {
+      "epoch": 6.285714285714286,
+      "grad_norm": 0.7499034404754639,
+      "learning_rate": 0.00022324015110631408,
+      "loss": 3.2808,
+      "step": 58300
+    },
+    {
+      "epoch": 6.291105121293801,
+      "grad_norm": 0.7447680234909058,
+      "learning_rate": 0.00022291635186184565,
+      "loss": 3.2746,
+      "step": 58350
+    },
+    {
+      "epoch": 6.296495956873315,
+      "grad_norm": 0.7814144492149353,
+      "learning_rate": 0.0002225925526173772,
+      "loss": 3.2653,
+      "step": 58400
+    },
+    {
+      "epoch": 6.30188679245283,
+      "grad_norm": 0.7258203625679016,
+      "learning_rate": 0.00022226875337290878,
+      "loss": 3.2543,
+      "step": 58450
+    },
+    {
+      "epoch": 6.307277628032345,
+      "grad_norm": 0.732068657875061,
+      "learning_rate": 0.00022194495412844033,
+      "loss": 3.2562,
+      "step": 58500
+    },
+    {
+      "epoch": 6.3126684636118595,
+      "grad_norm": 0.7250475883483887,
+      "learning_rate": 0.0002216211548839719,
+      "loss": 3.274,
+      "step": 58550
+    },
+    {
+      "epoch": 6.318059299191375,
+      "grad_norm": 0.725308895111084,
+      "learning_rate": 0.0002212973556395035,
+      "loss": 3.272,
+      "step": 58600
+    },
+    {
+      "epoch": 6.32345013477089,
+      "grad_norm": 0.7906202673912048,
+      "learning_rate": 0.00022097355639503507,
+      "loss": 3.2792,
+      "step": 58650
+    },
+    {
+      "epoch": 6.328840970350404,
+      "grad_norm": 0.8644693493843079,
+      "learning_rate": 0.00022064975715056664,
+      "loss": 3.2385,
+      "step": 58700
+    },
+    {
+      "epoch": 6.334231805929919,
+      "grad_norm": 0.7321583032608032,
+      "learning_rate": 0.0002203259579060982,
+      "loss": 3.276,
+      "step": 58750
+    },
+    {
+      "epoch": 6.339622641509434,
+      "grad_norm": 0.8114139437675476,
+      "learning_rate": 0.00022000215866162977,
+      "loss": 3.2713,
+      "step": 58800
+    },
+    {
+      "epoch": 6.345013477088949,
+      "grad_norm": 0.7152981758117676,
+      "learning_rate": 0.00021967835941716135,
+      "loss": 3.257,
+      "step": 58850
+    },
+    {
+      "epoch": 6.350404312668464,
+      "grad_norm": 0.7789607048034668,
+      "learning_rate": 0.0002193545601726929,
+      "loss": 3.2629,
+      "step": 58900
+    },
+    {
+      "epoch": 6.355795148247978,
+      "grad_norm": 0.7857829928398132,
+      "learning_rate": 0.00021903076092822448,
+      "loss": 3.2739,
+      "step": 58950
+    },
+    {
+      "epoch": 6.361185983827493,
+      "grad_norm": 0.7703624367713928,
+      "learning_rate": 0.00021870696168375606,
+      "loss": 3.276,
+      "step": 59000
+    },
+    {
+      "epoch": 6.361185983827493,
+      "eval_accuracy": 0.38511038209852405,
+      "eval_loss": 3.375075101852417,
+      "eval_runtime": 184.271,
+      "eval_samples_per_second": 97.742,
+      "eval_steps_per_second": 6.111,
+      "step": 59000
+    },
+    {
+      "epoch": 6.366576819407008,
+      "grad_norm": 0.7588159441947937,
+      "learning_rate": 0.0002183831624392876,
+      "loss": 3.2779,
+      "step": 59050
+    },
+    {
+      "epoch": 6.3719676549865225,
+      "grad_norm": 0.8037658333778381,
+      "learning_rate": 0.00021805936319481918,
+      "loss": 3.2748,
+      "step": 59100
+    },
+    {
+      "epoch": 6.377358490566038,
+      "grad_norm": 0.7042297124862671,
+      "learning_rate": 0.00021773556395035076,
+      "loss": 3.2582,
+      "step": 59150
+    },
+    {
+      "epoch": 6.382749326145553,
+      "grad_norm": 0.7502013444900513,
+      "learning_rate": 0.00021741176470588234,
+      "loss": 3.2693,
+      "step": 59200
+    },
+    {
+      "epoch": 6.388140161725067,
+      "grad_norm": 0.7545161843299866,
+      "learning_rate": 0.00021708796546141392,
+      "loss": 3.2658,
+      "step": 59250
+    },
+    {
+      "epoch": 6.393530997304582,
+      "grad_norm": 0.739264190196991,
+      "learning_rate": 0.0002167641662169455,
+      "loss": 3.2854,
+      "step": 59300
+    },
+    {
+      "epoch": 6.398921832884097,
+      "grad_norm": 0.8167062997817993,
+      "learning_rate": 0.00021644036697247702,
+      "loss": 3.2723,
+      "step": 59350
+    },
+    {
+      "epoch": 6.404312668463612,
+      "grad_norm": 0.7420883178710938,
+      "learning_rate": 0.0002161165677280086,
+      "loss": 3.2852,
+      "step": 59400
+    },
+    {
+      "epoch": 6.409703504043127,
+      "grad_norm": 0.7469702363014221,
+      "learning_rate": 0.00021579276848354017,
+      "loss": 3.2801,
+      "step": 59450
+    },
+    {
+      "epoch": 6.415094339622642,
+      "grad_norm": 0.796428382396698,
+      "learning_rate": 0.00021546896923907175,
+      "loss": 3.271,
+      "step": 59500
+    },
+    {
+      "epoch": 6.420485175202156,
+      "grad_norm": 0.7591307163238525,
+      "learning_rate": 0.00021514516999460333,
+      "loss": 3.2745,
+      "step": 59550
+    },
+    {
+      "epoch": 6.425876010781671,
+      "grad_norm": 0.7786388397216797,
+      "learning_rate": 0.0002148213707501349,
+      "loss": 3.2793,
+      "step": 59600
+    },
+    {
+      "epoch": 6.431266846361186,
+      "grad_norm": 0.8503186702728271,
+      "learning_rate": 0.00021449757150566648,
+      "loss": 3.2832,
+      "step": 59650
+    },
+    {
+      "epoch": 6.436657681940701,
+      "grad_norm": 0.7847234010696411,
+      "learning_rate": 0.00021417377226119806,
+      "loss": 3.2923,
+      "step": 59700
+    },
+    {
+      "epoch": 6.442048517520216,
+      "grad_norm": 0.7348524928092957,
+      "learning_rate": 0.0002138499730167296,
+      "loss": 3.2674,
+      "step": 59750
+    },
+    {
+      "epoch": 6.44743935309973,
+      "grad_norm": 0.7561467289924622,
+      "learning_rate": 0.00021352617377226116,
+      "loss": 3.2832,
+      "step": 59800
+    },
+    {
+      "epoch": 6.452830188679245,
+      "grad_norm": 0.7439683675765991,
+      "learning_rate": 0.00021320237452779274,
+      "loss": 3.2768,
+      "step": 59850
+    },
+    {
+      "epoch": 6.45822102425876,
+      "grad_norm": 0.7063082456588745,
+      "learning_rate": 0.00021287857528332432,
+      "loss": 3.2697,
+      "step": 59900
+    },
+    {
+      "epoch": 6.463611859838275,
+      "grad_norm": 0.7481423020362854,
+      "learning_rate": 0.0002125547760388559,
+      "loss": 3.2742,
+      "step": 59950
+    },
+    {
+      "epoch": 6.46900269541779,
+      "grad_norm": 0.7237406373023987,
+      "learning_rate": 0.00021223097679438747,
+      "loss": 3.2903,
+      "step": 60000
+    },
+    {
+      "epoch": 6.46900269541779,
+      "eval_accuracy": 0.38577588103638777,
+      "eval_loss": 3.3692877292633057,
+      "eval_runtime": 183.7708,
+      "eval_samples_per_second": 98.008,
+      "eval_steps_per_second": 6.127,
+      "step": 60000
+    },
+    {
+      "epoch": 6.474393530997305,
+      "grad_norm": 1.16843843460083,
+      "learning_rate": 0.00021190717754991905,
+      "loss": 3.2967,
+      "step": 60050
+    },
+    {
+      "epoch": 6.479784366576819,
+      "grad_norm": 0.7618022561073303,
+      "learning_rate": 0.0002115833783054506,
+      "loss": 3.2958,
+      "step": 60100
+    },
+    {
+      "epoch": 6.485175202156334,
+      "grad_norm": 0.7672574520111084,
+      "learning_rate": 0.00021125957906098218,
+      "loss": 3.2812,
+      "step": 60150
+    },
+    {
+      "epoch": 6.490566037735849,
+      "grad_norm": 0.7672653794288635,
+      "learning_rate": 0.00021093577981651373,
+      "loss": 3.2856,
+      "step": 60200
+    },
+    {
+      "epoch": 6.495956873315364,
+      "grad_norm": 0.7688764929771423,
+      "learning_rate": 0.0002106119805720453,
+      "loss": 3.2644,
+      "step": 60250
+    },
+    {
+      "epoch": 6.501347708894879,
+      "grad_norm": 0.715165376663208,
+      "learning_rate": 0.00021028818132757689,
+      "loss": 3.2857,
+      "step": 60300
+    },
+    {
+      "epoch": 6.506738544474393,
+      "grad_norm": 0.7562441229820251,
+      "learning_rate": 0.00020996438208310846,
+      "loss": 3.2896,
+      "step": 60350
+    },
+    {
+      "epoch": 6.512129380053908,
+      "grad_norm": 0.6920077204704285,
+      "learning_rate": 0.00020964058283864001,
+      "loss": 3.2756,
+      "step": 60400
+    },
+    {
+      "epoch": 6.517520215633423,
+      "grad_norm": 0.6866161227226257,
+      "learning_rate": 0.0002093167835941716,
+      "loss": 3.2792,
+      "step": 60450
+    },
+    {
+      "epoch": 6.5229110512129385,
+      "grad_norm": 0.7109043598175049,
+      "learning_rate": 0.00020899298434970317,
+      "loss": 3.2687,
+      "step": 60500
+    },
+    {
+      "epoch": 6.528301886792453,
+      "grad_norm": 0.7530231475830078,
+      "learning_rate": 0.00020866918510523475,
+      "loss": 3.2742,
+      "step": 60550
+    },
+    {
+      "epoch": 6.533692722371968,
+      "grad_norm": 0.7493744492530823,
+      "learning_rate": 0.00020834538586076632,
+      "loss": 3.2848,
+      "step": 60600
+    },
+    {
+      "epoch": 6.539083557951482,
+      "grad_norm": 0.733198344707489,
+      "learning_rate": 0.00020802158661629787,
+      "loss": 3.2755,
+      "step": 60650
+    },
+    {
+      "epoch": 6.544474393530997,
+      "grad_norm": 0.9269688129425049,
+      "learning_rate": 0.00020769778737182943,
+      "loss": 3.2978,
+      "step": 60700
+    },
+    {
+      "epoch": 6.549865229110512,
+      "grad_norm": 0.7194287776947021,
+      "learning_rate": 0.000207373988127361,
+      "loss": 3.278,
+      "step": 60750
+    },
+    {
+      "epoch": 6.555256064690027,
+      "grad_norm": 0.7988015413284302,
+      "learning_rate": 0.00020705018888289258,
+      "loss": 3.2922,
+      "step": 60800
+    },
+    {
+      "epoch": 6.560646900269542,
+      "grad_norm": 0.7761523127555847,
+      "learning_rate": 0.00020672638963842416,
+      "loss": 3.2839,
+      "step": 60850
+    },
+    {
+      "epoch": 6.566037735849057,
+      "grad_norm": 0.8137022852897644,
+      "learning_rate": 0.00020640259039395574,
+      "loss": 3.2923,
+      "step": 60900
+    },
+    {
+      "epoch": 6.571428571428571,
+      "grad_norm": 0.7631632089614868,
+      "learning_rate": 0.00020607879114948731,
+      "loss": 3.2751,
+      "step": 60950
+    },
+    {
+      "epoch": 6.576819407008086,
+      "grad_norm": 0.7450467348098755,
+      "learning_rate": 0.0002057549919050189,
+      "loss": 3.2681,
+      "step": 61000
+    },
+    {
+      "epoch": 6.576819407008086,
+      "eval_accuracy": 0.3860572920158272,
+      "eval_loss": 3.3661625385284424,
+      "eval_runtime": 183.9813,
+      "eval_samples_per_second": 97.896,
+      "eval_steps_per_second": 6.12,
+      "step": 61000
+    },
+    {
+      "epoch": 6.5822102425876015,
+      "grad_norm": 0.7303608059883118,
+      "learning_rate": 0.00020543119266055041,
+      "loss": 3.2785,
+      "step": 61050
+    },
+    {
+      "epoch": 6.587601078167116,
+      "grad_norm": 0.7955816984176636,
+      "learning_rate": 0.000205107393416082,
+      "loss": 3.2928,
+      "step": 61100
+    },
+    {
+      "epoch": 6.592991913746631,
+      "grad_norm": 0.772139310836792,
+      "learning_rate": 0.00020478359417161357,
+      "loss": 3.2785,
+      "step": 61150
+    },
+    {
+      "epoch": 6.598382749326145,
+      "grad_norm": 0.7001410126686096,
+      "learning_rate": 0.00020445979492714515,
+      "loss": 3.2655,
+      "step": 61200
+    },
+    {
+      "epoch": 6.60377358490566,
+      "grad_norm": 0.7417465448379517,
+      "learning_rate": 0.00020413599568267673,
+      "loss": 3.2931,
+      "step": 61250
+    },
+    {
+      "epoch": 6.609164420485175,
+      "grad_norm": 0.7509415149688721,
+      "learning_rate": 0.0002038121964382083,
+      "loss": 3.2842,
+      "step": 61300
+    },
+    {
+      "epoch": 6.6145552560646905,
+      "grad_norm": 0.8277612924575806,
+      "learning_rate": 0.00020348839719373988,
+      "loss": 3.2657,
+      "step": 61350
+    },
+    {
+      "epoch": 6.619946091644205,
+      "grad_norm": 0.7039164304733276,
+      "learning_rate": 0.00020316459794927146,
+      "loss": 3.2932,
+      "step": 61400
+    },
+    {
+      "epoch": 6.62533692722372,
+      "grad_norm": 0.7653436064720154,
+      "learning_rate": 0.000202840798704803,
+      "loss": 3.2882,
+      "step": 61450
+    },
+    {
+      "epoch": 6.630727762803234,
+      "grad_norm": 0.753315806388855,
+      "learning_rate": 0.00020251699946033456,
+      "loss": 3.2836,
+      "step": 61500
+    },
+    {
+      "epoch": 6.636118598382749,
+      "grad_norm": 0.7230058312416077,
+      "learning_rate": 0.00020219320021586614,
+      "loss": 3.2888,
+      "step": 61550
+    },
+    {
+      "epoch": 6.6415094339622645,
+      "grad_norm": 0.7700191140174866,
+      "learning_rate": 0.00020186940097139771,
+      "loss": 3.293,
+      "step": 61600
+    },
+    {
+      "epoch": 6.646900269541779,
+      "grad_norm": 0.7520590424537659,
+      "learning_rate": 0.0002015456017269293,
+      "loss": 3.2894,
+      "step": 61650
+    },
+    {
+      "epoch": 6.652291105121294,
+      "grad_norm": 0.7307371497154236,
+      "learning_rate": 0.00020122180248246087,
+      "loss": 3.2735,
+      "step": 61700
+    },
+    {
+      "epoch": 6.657681940700809,
+      "grad_norm": 0.7702900171279907,
+      "learning_rate": 0.00020089800323799242,
+      "loss": 3.3107,
+      "step": 61750
+    },
+    {
+      "epoch": 6.663072776280323,
+      "grad_norm": 0.7241050601005554,
+      "learning_rate": 0.000200574203993524,
+      "loss": 3.2667,
+      "step": 61800
+    },
+    {
+      "epoch": 6.668463611859838,
+      "grad_norm": 0.7302685379981995,
+      "learning_rate": 0.00020025040474905558,
+      "loss": 3.2695,
+      "step": 61850
+    },
+    {
+      "epoch": 6.6738544474393535,
+      "grad_norm": 0.7514686584472656,
+      "learning_rate": 0.0001999330814894765,
+      "loss": 3.2696,
+      "step": 61900
+    },
+    {
+      "epoch": 6.679245283018868,
+      "grad_norm": 0.7941187024116516,
+      "learning_rate": 0.00019960928224500808,
+      "loss": 3.2697,
+      "step": 61950
+    },
+    {
+      "epoch": 6.684636118598383,
+      "grad_norm": 0.7222111821174622,
+      "learning_rate": 0.00019928548300053966,
+      "loss": 3.2793,
+      "step": 62000
+    },
+    {
+      "epoch": 6.684636118598383,
+      "eval_accuracy": 0.38628166022916416,
+      "eval_loss": 3.3625741004943848,
+      "eval_runtime": 184.0577,
+      "eval_samples_per_second": 97.855,
+      "eval_steps_per_second": 6.118,
+      "step": 62000
+    },
+    {
+      "epoch": 6.690026954177897,
+      "grad_norm": 0.7426719665527344,
+      "learning_rate": 0.00019896168375607123,
+      "loss": 3.2672,
+      "step": 62050
+    },
+    {
+      "epoch": 6.695417789757412,
+      "grad_norm": 0.7612356543540955,
+      "learning_rate": 0.00019863788451160278,
+      "loss": 3.2768,
+      "step": 62100
+    },
+    {
+      "epoch": 6.7008086253369274,
+      "grad_norm": 0.7817362546920776,
+      "learning_rate": 0.00019831408526713436,
+      "loss": 3.2943,
+      "step": 62150
+    },
+    {
+      "epoch": 6.706199460916442,
+      "grad_norm": 0.7813862562179565,
+      "learning_rate": 0.00019799028602266594,
+      "loss": 3.28,
+      "step": 62200
+    },
+    {
+      "epoch": 6.711590296495957,
+      "grad_norm": 0.8101115822792053,
+      "learning_rate": 0.0001976664867781975,
+      "loss": 3.2889,
+      "step": 62250
+    },
+    {
+      "epoch": 6.716981132075472,
+      "grad_norm": 0.7497881054878235,
+      "learning_rate": 0.00019734916351861844,
+      "loss": 3.2753,
+      "step": 62300
+    },
+    {
+      "epoch": 6.722371967654986,
+      "grad_norm": 0.793246865272522,
+      "learning_rate": 0.00019702536427415002,
+      "loss": 3.283,
+      "step": 62350
+    },
+    {
+      "epoch": 6.727762803234501,
+      "grad_norm": 0.7845967411994934,
+      "learning_rate": 0.0001967015650296816,
+      "loss": 3.287,
+      "step": 62400
+    },
+    {
+      "epoch": 6.7331536388140165,
+      "grad_norm": 0.7628470063209534,
+      "learning_rate": 0.00019637776578521315,
+      "loss": 3.2814,
+      "step": 62450
+    },
+    {
+      "epoch": 6.738544474393531,
+      "grad_norm": 0.7873274683952332,
+      "learning_rate": 0.00019605396654074473,
+      "loss": 3.2863,
+      "step": 62500
+    },
+    {
+      "epoch": 6.743935309973046,
+      "grad_norm": 0.7793614268302917,
+      "learning_rate": 0.0001957301672962763,
+      "loss": 3.2713,
+      "step": 62550
+    },
+    {
+      "epoch": 6.74932614555256,
+      "grad_norm": 0.8013772964477539,
+      "learning_rate": 0.00019540636805180785,
+      "loss": 3.28,
+      "step": 62600
+    },
+    {
+      "epoch": 6.754716981132075,
+      "grad_norm": 0.7564038634300232,
+      "learning_rate": 0.00019508256880733943,
+      "loss": 3.2837,
+      "step": 62650
+    },
+    {
+      "epoch": 6.7601078167115904,
+      "grad_norm": 0.7453675270080566,
+      "learning_rate": 0.000194758769562871,
+      "loss": 3.2719,
+      "step": 62700
+    },
+    {
+      "epoch": 6.765498652291106,
+      "grad_norm": 0.8338208198547363,
+      "learning_rate": 0.00019443497031840256,
+      "loss": 3.2872,
+      "step": 62750
+    },
+    {
+      "epoch": 6.77088948787062,
+      "grad_norm": 0.7567179203033447,
+      "learning_rate": 0.00019411117107393414,
+      "loss": 3.2853,
+      "step": 62800
+    },
+    {
+      "epoch": 6.776280323450135,
+      "grad_norm": 0.770332396030426,
+      "learning_rate": 0.00019378737182946572,
+      "loss": 3.2586,
+      "step": 62850
+    },
+    {
+      "epoch": 6.781671159029649,
+      "grad_norm": 0.7977242469787598,
+      "learning_rate": 0.0001934635725849973,
+      "loss": 3.2788,
+      "step": 62900
+    },
+    {
+      "epoch": 6.787061994609164,
+      "grad_norm": 0.7252545952796936,
+      "learning_rate": 0.00019313977334052887,
+      "loss": 3.2967,
+      "step": 62950
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "grad_norm": 0.7643560171127319,
+      "learning_rate": 0.00019281597409606042,
+      "loss": 3.2753,
+      "step": 63000
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "eval_accuracy": 0.3866451041389036,
+      "eval_loss": 3.3582923412323,
+      "eval_runtime": 183.964,
+      "eval_samples_per_second": 97.905,
+      "eval_steps_per_second": 6.121,
+      "step": 63000
+    },
+    {
+      "epoch": 6.797843665768194,
+      "grad_norm": 0.769708514213562,
+      "learning_rate": 0.00019249217485159197,
+      "loss": 3.2593,
+      "step": 63050
+    },
+    {
+      "epoch": 6.803234501347709,
+      "grad_norm": 0.7512614130973816,
+      "learning_rate": 0.00019216837560712355,
+      "loss": 3.279,
+      "step": 63100
+    },
+    {
+      "epoch": 6.808625336927224,
+      "grad_norm": 0.7702354192733765,
+      "learning_rate": 0.00019184457636265513,
+      "loss": 3.2774,
+      "step": 63150
+    },
+    {
+      "epoch": 6.814016172506738,
+      "grad_norm": 0.771395742893219,
+      "learning_rate": 0.0001915207771181867,
+      "loss": 3.2896,
+      "step": 63200
+    },
+    {
+      "epoch": 6.819407008086253,
+      "grad_norm": 0.7912437319755554,
+      "learning_rate": 0.00019119697787371828,
+      "loss": 3.272,
+      "step": 63250
+    },
+    {
+      "epoch": 6.824797843665769,
+      "grad_norm": 0.7434420585632324,
+      "learning_rate": 0.00019087317862924986,
+      "loss": 3.3104,
+      "step": 63300
+    },
+    {
+      "epoch": 6.830188679245283,
+      "grad_norm": 0.775771975517273,
+      "learning_rate": 0.00019054937938478144,
+      "loss": 3.2854,
+      "step": 63350
+    },
+    {
+      "epoch": 6.835579514824798,
+      "grad_norm": 0.768345057964325,
+      "learning_rate": 0.00019022558014031302,
+      "loss": 3.2817,
+      "step": 63400
+    },
+    {
+      "epoch": 6.840970350404312,
+      "grad_norm": 0.7371838092803955,
+      "learning_rate": 0.00018990178089584454,
+      "loss": 3.2816,
+      "step": 63450
+    },
+    {
+      "epoch": 6.846361185983827,
+      "grad_norm": 0.7820178270339966,
+      "learning_rate": 0.00018957798165137612,
+      "loss": 3.2847,
+      "step": 63500
+    },
+    {
+      "epoch": 6.8517520215633425,
+      "grad_norm": 0.7918660044670105,
+      "learning_rate": 0.0001892541824069077,
+      "loss": 3.277,
+      "step": 63550
+    },
+    {
+      "epoch": 6.857142857142857,
+      "grad_norm": 0.73056960105896,
+      "learning_rate": 0.00018893038316243927,
+      "loss": 3.2855,
+      "step": 63600
+    },
+    {
+      "epoch": 6.862533692722372,
+      "grad_norm": 0.7188368439674377,
+      "learning_rate": 0.00018860658391797085,
+      "loss": 3.2765,
+      "step": 63650
+    },
+    {
+      "epoch": 6.867924528301887,
+      "grad_norm": 0.7747359275817871,
+      "learning_rate": 0.00018828278467350243,
+      "loss": 3.2962,
+      "step": 63700
+    },
+    {
+      "epoch": 6.873315363881401,
+      "grad_norm": 0.7617150545120239,
+      "learning_rate": 0.000187958985429034,
+      "loss": 3.2799,
+      "step": 63750
+    },
+    {
+      "epoch": 6.878706199460916,
+      "grad_norm": 0.8125529289245605,
+      "learning_rate": 0.00018763518618456556,
+      "loss": 3.2672,
+      "step": 63800
+    },
+    {
+      "epoch": 6.884097035040432,
+      "grad_norm": 0.8084162473678589,
+      "learning_rate": 0.0001873113869400971,
+      "loss": 3.2708,
+      "step": 63850
+    },
+    {
+      "epoch": 6.889487870619946,
+      "grad_norm": 0.735432505607605,
+      "learning_rate": 0.00018698758769562868,
+      "loss": 3.2802,
+      "step": 63900
+    },
+    {
+      "epoch": 6.894878706199461,
+      "grad_norm": 0.7418496012687683,
+      "learning_rate": 0.00018666378845116026,
+      "loss": 3.2816,
+      "step": 63950
+    },
+    {
+      "epoch": 6.900269541778976,
+      "grad_norm": 0.7551790475845337,
+      "learning_rate": 0.00018633998920669184,
+      "loss": 3.2875,
+      "step": 64000
+    },
+    {
+      "epoch": 6.900269541778976,
+      "eval_accuracy": 0.3873492835048292,
+      "eval_loss": 3.3529202938079834,
+      "eval_runtime": 183.9905,
+      "eval_samples_per_second": 97.891,
+      "eval_steps_per_second": 6.12,
+      "step": 64000
+    },
+    {
+      "epoch": 6.90566037735849,
+      "grad_norm": 0.7530993223190308,
+      "learning_rate": 0.00018601618996222342,
+      "loss": 3.282,
+      "step": 64050
+    },
+    {
+      "epoch": 6.9110512129380055,
+      "grad_norm": 0.7252298593521118,
+      "learning_rate": 0.00018569239071775497,
+      "loss": 3.283,
+      "step": 64100
+    },
+    {
+      "epoch": 6.916442048517521,
+      "grad_norm": 0.7530275583267212,
+      "learning_rate": 0.00018536859147328655,
+      "loss": 3.2738,
+      "step": 64150
+    },
+    {
+      "epoch": 6.921832884097035,
+      "grad_norm": 0.7903400659561157,
+      "learning_rate": 0.00018504479222881812,
+      "loss": 3.2877,
+      "step": 64200
+    },
+    {
+      "epoch": 6.92722371967655,
+      "grad_norm": 0.786756157875061,
+      "learning_rate": 0.0001847209929843497,
+      "loss": 3.2815,
+      "step": 64250
+    },
+    {
+      "epoch": 6.932614555256064,
+      "grad_norm": 0.8159517645835876,
+      "learning_rate": 0.00018439719373988125,
+      "loss": 3.2693,
+      "step": 64300
+    },
+    {
+      "epoch": 6.938005390835579,
+      "grad_norm": 0.7632095217704773,
+      "learning_rate": 0.00018407339449541283,
+      "loss": 3.2753,
+      "step": 64350
+    },
+    {
+      "epoch": 6.943396226415095,
+      "grad_norm": 0.7830843329429626,
+      "learning_rate": 0.00018374959525094438,
+      "loss": 3.2897,
+      "step": 64400
+    },
+    {
+      "epoch": 6.948787061994609,
+      "grad_norm": 0.8015329241752625,
+      "learning_rate": 0.00018342579600647596,
+      "loss": 3.2668,
+      "step": 64450
+    },
+    {
+      "epoch": 6.954177897574124,
+      "grad_norm": 0.8383193612098694,
+      "learning_rate": 0.00018310199676200753,
+      "loss": 3.2993,
+      "step": 64500
+    },
+    {
+      "epoch": 6.959568733153639,
+      "grad_norm": 0.7615668177604675,
+      "learning_rate": 0.0001827781975175391,
+      "loss": 3.2865,
+      "step": 64550
+    },
+    {
+      "epoch": 6.964959568733153,
+      "grad_norm": 0.7987204194068909,
+      "learning_rate": 0.0001824543982730707,
+      "loss": 3.291,
+      "step": 64600
+    },
+    {
+      "epoch": 6.9703504043126685,
+      "grad_norm": 0.7923724055290222,
+      "learning_rate": 0.00018213059902860227,
+      "loss": 3.2708,
+      "step": 64650
+    },
+    {
+      "epoch": 6.975741239892184,
+      "grad_norm": 0.8186677098274231,
+      "learning_rate": 0.00018180679978413382,
+      "loss": 3.2715,
+      "step": 64700
+    },
+    {
+      "epoch": 6.981132075471698,
+      "grad_norm": 0.8212941288948059,
+      "learning_rate": 0.00018148300053966537,
+      "loss": 3.2677,
+      "step": 64750
+    },
+    {
+      "epoch": 6.986522911051213,
+      "grad_norm": 0.8504579067230225,
+      "learning_rate": 0.00018115920129519695,
+      "loss": 3.2813,
+      "step": 64800
+    },
+    {
+      "epoch": 6.991913746630727,
+      "grad_norm": 0.7678773403167725,
+      "learning_rate": 0.00018083540205072852,
+      "loss": 3.2945,
+      "step": 64850
+    },
+    {
+      "epoch": 6.997304582210242,
+      "grad_norm": 0.7796960473060608,
+      "learning_rate": 0.0001805116028062601,
+      "loss": 3.287,
+      "step": 64900
+    },
+    {
+      "epoch": 7.002695417789758,
+      "grad_norm": 0.8160527944564819,
+      "learning_rate": 0.00018018780356179168,
+      "loss": 3.2342,
+      "step": 64950
+    },
+    {
+      "epoch": 7.008086253369272,
+      "grad_norm": 0.7837775945663452,
+      "learning_rate": 0.00017986400431732326,
+      "loss": 3.2082,
+      "step": 65000
+    },
+    {
+      "epoch": 7.008086253369272,
+      "eval_accuracy": 0.38750389656418927,
+      "eval_loss": 3.3548026084899902,
+      "eval_runtime": 183.8139,
+      "eval_samples_per_second": 97.985,
+      "eval_steps_per_second": 6.126,
+      "step": 65000
+    },
+    {
+      "epoch": 7.013477088948787,
+      "grad_norm": 0.8339738845825195,
+      "learning_rate": 0.00017954020507285483,
+      "loss": 3.1856,
+      "step": 65050
+    },
+    {
+      "epoch": 7.018867924528302,
+      "grad_norm": 0.8263199329376221,
+      "learning_rate": 0.0001792164058283864,
+      "loss": 3.1704,
+      "step": 65100
+    },
+    {
+      "epoch": 7.024258760107816,
+      "grad_norm": 0.7783221006393433,
+      "learning_rate": 0.00017889260658391794,
+      "loss": 3.2089,
+      "step": 65150
+    },
+    {
+      "epoch": 7.0296495956873315,
+      "grad_norm": 0.762937068939209,
+      "learning_rate": 0.00017856880733944951,
+      "loss": 3.2022,
+      "step": 65200
+    },
+    {
+      "epoch": 7.035040431266847,
+      "grad_norm": 0.7570405602455139,
+      "learning_rate": 0.0001782450080949811,
+      "loss": 3.1999,
+      "step": 65250
+    },
+    {
+      "epoch": 7.040431266846361,
+      "grad_norm": 0.833223819732666,
+      "learning_rate": 0.00017792120885051267,
+      "loss": 3.1842,
+      "step": 65300
+    },
+    {
+      "epoch": 7.045822102425876,
+      "grad_norm": 0.746315598487854,
+      "learning_rate": 0.00017760388559093362,
+      "loss": 3.1927,
+      "step": 65350
+    },
+    {
+      "epoch": 7.051212938005391,
+      "grad_norm": 0.7781280279159546,
+      "learning_rate": 0.0001772800863464652,
+      "loss": 3.1935,
+      "step": 65400
+    },
+    {
+      "epoch": 7.056603773584905,
+      "grad_norm": 0.839777946472168,
+      "learning_rate": 0.00017695628710199672,
+      "loss": 3.2116,
+      "step": 65450
+    },
+    {
+      "epoch": 7.061994609164421,
+      "grad_norm": 0.8046913146972656,
+      "learning_rate": 0.0001766324878575283,
+      "loss": 3.2046,
+      "step": 65500
+    },
+    {
+      "epoch": 7.067385444743936,
+      "grad_norm": 0.7587380409240723,
+      "learning_rate": 0.00017630868861305988,
+      "loss": 3.208,
+      "step": 65550
+    },
+    {
+      "epoch": 7.07277628032345,
+      "grad_norm": 0.7880898118019104,
+      "learning_rate": 0.00017598488936859146,
+      "loss": 3.2019,
+      "step": 65600
+    },
+    {
+      "epoch": 7.078167115902965,
+      "grad_norm": 0.7449747323989868,
+      "learning_rate": 0.00017566109012412303,
+      "loss": 3.1968,
+      "step": 65650
+    },
+    {
+      "epoch": 7.083557951482479,
+      "grad_norm": 0.7823360562324524,
+      "learning_rate": 0.0001753372908796546,
+      "loss": 3.2062,
+      "step": 65700
+    },
+    {
+      "epoch": 7.0889487870619945,
+      "grad_norm": 0.78197181224823,
+      "learning_rate": 0.0001750134916351862,
+      "loss": 3.2057,
+      "step": 65750
+    },
+    {
+      "epoch": 7.09433962264151,
+      "grad_norm": 0.7566399574279785,
+      "learning_rate": 0.00017468969239071774,
+      "loss": 3.2,
+      "step": 65800
+    },
+    {
+      "epoch": 7.099730458221024,
+      "grad_norm": 0.8071035742759705,
+      "learning_rate": 0.00017436589314624932,
+      "loss": 3.1944,
+      "step": 65850
+    },
+    {
+      "epoch": 7.105121293800539,
+      "grad_norm": 0.8223758339881897,
+      "learning_rate": 0.00017404209390178087,
+      "loss": 3.1912,
+      "step": 65900
+    },
+    {
+      "epoch": 7.110512129380054,
+      "grad_norm": 0.7952433824539185,
+      "learning_rate": 0.00017371829465731244,
+      "loss": 3.222,
+      "step": 65950
+    },
+    {
+      "epoch": 7.115902964959568,
+      "grad_norm": 0.7765783667564392,
+      "learning_rate": 0.00017339449541284402,
+      "loss": 3.2291,
+      "step": 66000
+    },
+    {
+      "epoch": 7.115902964959568,
+      "eval_accuracy": 0.38722042117988126,
+      "eval_loss": 3.3594112396240234,
+      "eval_runtime": 183.9219,
+      "eval_samples_per_second": 97.927,
+      "eval_steps_per_second": 6.122,
+      "step": 66000
+    },
+    {
+      "epoch": 7.121293800539084,
+      "grad_norm": 0.7651994824409485,
+      "learning_rate": 0.0001730706961683756,
+      "loss": 3.1816,
+      "step": 66050
+    },
+    {
+      "epoch": 7.126684636118599,
+      "grad_norm": 0.7809786796569824,
+      "learning_rate": 0.00017274689692390715,
+      "loss": 3.2177,
+      "step": 66100
+    },
+    {
+      "epoch": 7.132075471698113,
+      "grad_norm": 0.772335410118103,
+      "learning_rate": 0.00017242309767943873,
+      "loss": 3.2218,
+      "step": 66150
+    },
+    {
+      "epoch": 7.137466307277628,
+      "grad_norm": 0.8068752884864807,
+      "learning_rate": 0.0001720992984349703,
+      "loss": 3.2222,
+      "step": 66200
+    },
+    {
+      "epoch": 7.142857142857143,
+      "grad_norm": 0.7680201530456543,
+      "learning_rate": 0.00017177549919050188,
+      "loss": 3.21,
+      "step": 66250
+    },
+    {
+      "epoch": 7.1482479784366575,
+      "grad_norm": 0.7757328748703003,
+      "learning_rate": 0.00017145169994603346,
+      "loss": 3.2183,
+      "step": 66300
+    },
+    {
+      "epoch": 7.153638814016173,
+      "grad_norm": 0.7772490978240967,
+      "learning_rate": 0.000171127900701565,
+      "loss": 3.2189,
+      "step": 66350
+    },
+    {
+      "epoch": 7.159029649595688,
+      "grad_norm": 0.8381431102752686,
+      "learning_rate": 0.0001708041014570966,
+      "loss": 3.196,
+      "step": 66400
+    },
+    {
+      "epoch": 7.164420485175202,
+      "grad_norm": 0.8712314367294312,
+      "learning_rate": 0.00017048030221262814,
+      "loss": 3.2298,
+      "step": 66450
+    },
+    {
+      "epoch": 7.169811320754717,
+      "grad_norm": 0.7782764434814453,
+      "learning_rate": 0.00017015650296815972,
+      "loss": 3.2195,
+      "step": 66500
+    },
+    {
+      "epoch": 7.175202156334231,
+      "grad_norm": 0.7687488794326782,
+      "learning_rate": 0.0001698327037236913,
+      "loss": 3.2183,
+      "step": 66550
+    },
+    {
+      "epoch": 7.180592991913747,
+      "grad_norm": 0.7948805689811707,
+      "learning_rate": 0.00016950890447922287,
+      "loss": 3.2202,
+      "step": 66600
+    },
+    {
+      "epoch": 7.185983827493262,
+      "grad_norm": 0.8118099570274353,
+      "learning_rate": 0.00016918510523475445,
+      "loss": 3.2115,
+      "step": 66650
+    },
+    {
+      "epoch": 7.191374663072776,
+      "grad_norm": 0.8137215971946716,
+      "learning_rate": 0.00016886130599028603,
+      "loss": 3.2174,
+      "step": 66700
+    },
+    {
+      "epoch": 7.196765498652291,
+      "grad_norm": 0.7838942408561707,
+      "learning_rate": 0.00016853750674581755,
+      "loss": 3.2021,
+      "step": 66750
+    },
+    {
+      "epoch": 7.202156334231806,
+      "grad_norm": 0.7940311431884766,
+      "learning_rate": 0.00016821370750134913,
+      "loss": 3.2059,
+      "step": 66800
+    },
+    {
+      "epoch": 7.2075471698113205,
+      "grad_norm": 0.7829987406730652,
+      "learning_rate": 0.0001678899082568807,
+      "loss": 3.2321,
+      "step": 66850
+    },
+    {
+      "epoch": 7.212938005390836,
+      "grad_norm": 0.7903277277946472,
+      "learning_rate": 0.00016756610901241228,
+      "loss": 3.2087,
+      "step": 66900
+    },
+    {
+      "epoch": 7.218328840970351,
+      "grad_norm": 0.8051270246505737,
+      "learning_rate": 0.00016724230976794386,
+      "loss": 3.2256,
+      "step": 66950
+    },
+    {
+      "epoch": 7.223719676549865,
+      "grad_norm": 0.7584980726242065,
+      "learning_rate": 0.00016691851052347544,
+      "loss": 3.2144,
+      "step": 67000
+    },
+    {
+      "epoch": 7.223719676549865,
+      "eval_accuracy": 0.38720227614761626,
+      "eval_loss": 3.356732130050659,
+      "eval_runtime": 184.0554,
+      "eval_samples_per_second": 97.856,
+      "eval_steps_per_second": 6.118,
+      "step": 67000
+    },
+    {
+      "epoch": 7.22911051212938,
+      "grad_norm": 0.818000316619873,
+      "learning_rate": 0.00016659471127900702,
+      "loss": 3.2173,
+      "step": 67050
+    },
+    {
+      "epoch": 7.234501347708895,
+      "grad_norm": 0.8095887899398804,
+      "learning_rate": 0.0001662709120345386,
+      "loss": 3.2123,
+      "step": 67100
+    },
+    {
+      "epoch": 7.2398921832884096,
+      "grad_norm": 0.8081578612327576,
+      "learning_rate": 0.00016594711279007015,
+      "loss": 3.2229,
+      "step": 67150
+    },
+    {
+      "epoch": 7.245283018867925,
+      "grad_norm": 0.7935521006584167,
+      "learning_rate": 0.0001656233135456017,
+      "loss": 3.2172,
+      "step": 67200
+    },
+    {
+      "epoch": 7.250673854447439,
+      "grad_norm": 0.8998975157737732,
+      "learning_rate": 0.00016529951430113327,
+      "loss": 3.2108,
+      "step": 67250
+    },
+    {
+      "epoch": 7.256064690026954,
+      "grad_norm": 0.8090941309928894,
+      "learning_rate": 0.00016497571505666485,
+      "loss": 3.2352,
+      "step": 67300
+    },
+    {
+      "epoch": 7.261455525606469,
+      "grad_norm": 0.7772446870803833,
+      "learning_rate": 0.00016465191581219643,
+      "loss": 3.2357,
+      "step": 67350
+    },
+    {
+      "epoch": 7.2668463611859835,
+      "grad_norm": 0.8178821206092834,
+      "learning_rate": 0.000164328116567728,
+      "loss": 3.2374,
+      "step": 67400
+    },
+    {
+      "epoch": 7.272237196765499,
+      "grad_norm": 0.7874232530593872,
+      "learning_rate": 0.00016400431732325956,
+      "loss": 3.2318,
+      "step": 67450
+    },
+    {
+      "epoch": 7.277628032345014,
+      "grad_norm": 0.8284779191017151,
+      "learning_rate": 0.00016368051807879114,
+      "loss": 3.2212,
+      "step": 67500
+    },
+    {
+      "epoch": 7.283018867924528,
+      "grad_norm": 0.8559075593948364,
+      "learning_rate": 0.0001633567188343227,
+      "loss": 3.1971,
+      "step": 67550
+    },
+    {
+      "epoch": 7.288409703504043,
+      "grad_norm": 0.7753306031227112,
+      "learning_rate": 0.00016303291958985426,
+      "loss": 3.2145,
+      "step": 67600
+    },
+    {
+      "epoch": 7.293800539083558,
+      "grad_norm": 0.7762905359268188,
+      "learning_rate": 0.00016270912034538584,
+      "loss": 3.2352,
+      "step": 67650
+    },
+    {
+      "epoch": 7.2991913746630726,
+      "grad_norm": 0.7790316343307495,
+      "learning_rate": 0.00016238532110091742,
+      "loss": 3.1933,
+      "step": 67700
+    },
+    {
+      "epoch": 7.304582210242588,
+      "grad_norm": 0.8162766695022583,
+      "learning_rate": 0.000162061521856449,
+      "loss": 3.2162,
+      "step": 67750
+    },
+    {
+      "epoch": 7.309973045822103,
+      "grad_norm": 0.7991843223571777,
+      "learning_rate": 0.00016173772261198055,
+      "loss": 3.2277,
+      "step": 67800
+    },
+    {
+      "epoch": 7.315363881401617,
+      "grad_norm": 0.777722954750061,
+      "learning_rate": 0.00016141392336751212,
+      "loss": 3.2262,
+      "step": 67850
+    },
+    {
+      "epoch": 7.320754716981132,
+      "grad_norm": 0.7924916744232178,
+      "learning_rate": 0.0001610901241230437,
+      "loss": 3.221,
+      "step": 67900
+    },
+    {
+      "epoch": 7.3261455525606465,
+      "grad_norm": 0.8169491291046143,
+      "learning_rate": 0.00016076632487857528,
+      "loss": 3.2437,
+      "step": 67950
+    },
+    {
+      "epoch": 7.331536388140162,
+      "grad_norm": 0.7886592149734497,
+      "learning_rate": 0.00016044252563410686,
+      "loss": 3.2273,
+      "step": 68000
+    },
+    {
+      "epoch": 7.331536388140162,
+      "eval_accuracy": 0.3883553005931035,
+      "eval_loss": 3.351078987121582,
+      "eval_runtime": 185.9442,
+      "eval_samples_per_second": 96.862,
+      "eval_steps_per_second": 6.056,
+      "step": 68000
+    },
+    {
+      "epoch": 7.336927223719677,
+      "grad_norm": 0.8167215585708618,
+      "learning_rate": 0.0001601187263896384,
+      "loss": 3.2205,
+      "step": 68050
+    },
+    {
+      "epoch": 7.342318059299191,
+      "grad_norm": 0.7827270030975342,
+      "learning_rate": 0.00015979492714516996,
+      "loss": 3.213,
+      "step": 68100
+    },
+    {
+      "epoch": 7.347708894878706,
+      "grad_norm": 0.8101873397827148,
+      "learning_rate": 0.00015947112790070154,
+      "loss": 3.221,
+      "step": 68150
+    },
+    {
+      "epoch": 7.353099730458221,
+      "grad_norm": 0.7950315475463867,
+      "learning_rate": 0.00015914732865623311,
+      "loss": 3.2326,
+      "step": 68200
+    },
+    {
+      "epoch": 7.3584905660377355,
+      "grad_norm": 0.7943675518035889,
+      "learning_rate": 0.0001588235294117647,
+      "loss": 3.2038,
+      "step": 68250
+    },
+    {
+      "epoch": 7.363881401617251,
+      "grad_norm": 0.7641896605491638,
+      "learning_rate": 0.00015849973016729627,
+      "loss": 3.2036,
+      "step": 68300
+    },
+    {
+      "epoch": 7.369272237196766,
+      "grad_norm": 0.8040649890899658,
+      "learning_rate": 0.00015817593092282785,
+      "loss": 3.2418,
+      "step": 68350
+    },
+    {
+      "epoch": 7.37466307277628,
+      "grad_norm": 0.8314372897148132,
+      "learning_rate": 0.00015785213167835942,
+      "loss": 3.2144,
+      "step": 68400
+    },
+    {
+      "epoch": 7.380053908355795,
+      "grad_norm": 0.8147042393684387,
+      "learning_rate": 0.00015752833243389095,
+      "loss": 3.219,
+      "step": 68450
+    },
+    {
+      "epoch": 7.38544474393531,
+      "grad_norm": 0.767492413520813,
+      "learning_rate": 0.00015720453318942253,
+      "loss": 3.2212,
+      "step": 68500
+    },
+    {
+      "epoch": 7.390835579514825,
+      "grad_norm": 0.7924861311912537,
+      "learning_rate": 0.0001568807339449541,
+      "loss": 3.2114,
+      "step": 68550
+    },
+    {
+      "epoch": 7.39622641509434,
+      "grad_norm": 0.8257951736450195,
+      "learning_rate": 0.00015655693470048568,
+      "loss": 3.228,
+      "step": 68600
+    },
+    {
+      "epoch": 7.401617250673855,
+      "grad_norm": 0.7666079998016357,
+      "learning_rate": 0.00015623313545601726,
+      "loss": 3.2288,
+      "step": 68650
+    },
+    {
+      "epoch": 7.407008086253369,
+      "grad_norm": 0.8601511716842651,
+      "learning_rate": 0.00015590933621154884,
+      "loss": 3.239,
+      "step": 68700
+    },
+    {
+      "epoch": 7.412398921832884,
+      "grad_norm": 0.7925523519515991,
+      "learning_rate": 0.00015558553696708041,
+      "loss": 3.2177,
+      "step": 68750
+    },
+    {
+      "epoch": 7.4177897574123985,
+      "grad_norm": 0.818068265914917,
+      "learning_rate": 0.00015526173772261196,
+      "loss": 3.2129,
+      "step": 68800
+    },
+    {
+      "epoch": 7.423180592991914,
+      "grad_norm": 0.8102258443832397,
+      "learning_rate": 0.00015493793847814354,
+      "loss": 3.2238,
+      "step": 68850
+    },
+    {
+      "epoch": 7.428571428571429,
+      "grad_norm": 0.8156552314758301,
+      "learning_rate": 0.0001546141392336751,
+      "loss": 3.2322,
+      "step": 68900
+    },
+    {
+      "epoch": 7.433962264150943,
+      "grad_norm": 0.8134991526603699,
+      "learning_rate": 0.00015429033998920667,
+      "loss": 3.2442,
+      "step": 68950
+    },
+    {
+      "epoch": 7.439353099730458,
+      "grad_norm": 0.8857681155204773,
+      "learning_rate": 0.00015397301672962762,
+      "loss": 3.2405,
+      "step": 69000
+    },
+    {
+      "epoch": 7.439353099730458,
+      "eval_accuracy": 0.3885801034179917,
+      "eval_loss": 3.3481028079986572,
+      "eval_runtime": 184.4596,
+      "eval_samples_per_second": 97.642,
+      "eval_steps_per_second": 6.104,
+      "step": 69000
+    },
+    {
+      "epoch": 7.444743935309973,
+      "grad_norm": 0.8334435224533081,
+      "learning_rate": 0.0001536492174851592,
+      "loss": 3.2515,
+      "step": 69050
+    },
+    {
+      "epoch": 7.450134770889488,
+      "grad_norm": 0.8292680978775024,
+      "learning_rate": 0.00015332541824069078,
+      "loss": 3.2223,
+      "step": 69100
+    },
+    {
+      "epoch": 7.455525606469003,
+      "grad_norm": 0.819562554359436,
+      "learning_rate": 0.00015300161899622233,
+      "loss": 3.2417,
+      "step": 69150
+    },
+    {
+      "epoch": 7.460916442048518,
+      "grad_norm": 0.788324236869812,
+      "learning_rate": 0.00015267781975175388,
+      "loss": 3.2305,
+      "step": 69200
+    },
+    {
+      "epoch": 7.466307277628032,
+      "grad_norm": 0.8466244339942932,
+      "learning_rate": 0.00015235402050728546,
+      "loss": 3.2109,
+      "step": 69250
+    },
+    {
+      "epoch": 7.471698113207547,
+      "grad_norm": 0.7949193716049194,
+      "learning_rate": 0.00015203022126281703,
+      "loss": 3.2201,
+      "step": 69300
+    },
+    {
+      "epoch": 7.4770889487870615,
+      "grad_norm": 0.7899576425552368,
+      "learning_rate": 0.0001517064220183486,
+      "loss": 3.2475,
+      "step": 69350
+    },
+    {
+      "epoch": 7.482479784366577,
+      "grad_norm": 0.7817439436912537,
+      "learning_rate": 0.0001513826227738802,
+      "loss": 3.2138,
+      "step": 69400
+    },
+    {
+      "epoch": 7.487870619946092,
+      "grad_norm": 0.9032320380210876,
+      "learning_rate": 0.00015105882352941177,
+      "loss": 3.2157,
+      "step": 69450
+    },
+    {
+      "epoch": 7.493261455525606,
+      "grad_norm": 0.7862300276756287,
+      "learning_rate": 0.00015073502428494332,
+      "loss": 3.2233,
+      "step": 69500
+    },
+    {
+      "epoch": 7.498652291105121,
+      "grad_norm": 0.8258659839630127,
+      "learning_rate": 0.0001504112250404749,
+      "loss": 3.2201,
+      "step": 69550
+    },
+    {
+      "epoch": 7.504043126684636,
+      "grad_norm": 0.7886598110198975,
+      "learning_rate": 0.00015008742579600647,
+      "loss": 3.2187,
+      "step": 69600
+    },
+    {
+      "epoch": 7.509433962264151,
+      "grad_norm": 0.8414490222930908,
+      "learning_rate": 0.00014976362655153802,
+      "loss": 3.2265,
+      "step": 69650
+    },
+    {
+      "epoch": 7.514824797843666,
+      "grad_norm": 0.7889809012413025,
+      "learning_rate": 0.0001494398273070696,
+      "loss": 3.23,
+      "step": 69700
+    },
+    {
+      "epoch": 7.520215633423181,
+      "grad_norm": 0.8091080188751221,
+      "learning_rate": 0.00014911602806260118,
+      "loss": 3.2209,
+      "step": 69750
+    },
+    {
+      "epoch": 7.525606469002695,
+      "grad_norm": 0.8536641001701355,
+      "learning_rate": 0.00014879222881813273,
+      "loss": 3.2336,
+      "step": 69800
+    },
+    {
+      "epoch": 7.53099730458221,
+      "grad_norm": 0.8249064683914185,
+      "learning_rate": 0.0001484684295736643,
+      "loss": 3.2296,
+      "step": 69850
+    },
+    {
+      "epoch": 7.536388140161725,
+      "grad_norm": 0.7949578762054443,
+      "learning_rate": 0.00014814463032919589,
+      "loss": 3.2286,
+      "step": 69900
+    },
+    {
+      "epoch": 7.54177897574124,
+      "grad_norm": 0.9103390574455261,
+      "learning_rate": 0.00014782083108472744,
+      "loss": 3.2317,
+      "step": 69950
+    },
+    {
+      "epoch": 7.547169811320755,
+      "grad_norm": 0.8113278150558472,
+      "learning_rate": 0.00014749703184025901,
+      "loss": 3.2308,
+      "step": 70000
+    },
+    {
+      "epoch": 7.547169811320755,
+      "eval_accuracy": 0.38873004440317566,
+      "eval_loss": 3.3429155349731445,
+      "eval_runtime": 184.234,
+      "eval_samples_per_second": 97.762,
+      "eval_steps_per_second": 6.112,
+      "step": 70000
+    }
+  ],
+  "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": 5.852484255744e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}