diff --git "a/checkpoint-92750/trainer_state.json" "b/checkpoint-92750/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-92750/trainer_state.json"
@@ -0,0 +1,13846 @@
+{
+  "best_metric": 3.303354024887085,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M_high_500_8397/checkpoint-90000",
+  "epoch": 10.0,
+  "eval_steps": 1000,
+  "global_step": 92750,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005390835579514825,
+      "grad_norm": 2.1266844272613525,
+      "learning_rate": 0.0003,
+      "loss": 8.6323,
+      "step": 50
+    },
+    {
+      "epoch": 0.01078167115902965,
+      "grad_norm": 2.236001968383789,
+      "learning_rate": 0.0006,
+      "loss": 6.8959,
+      "step": 100
+    },
+    {
+      "epoch": 0.016172506738544475,
+      "grad_norm": 2.4629225730895996,
+      "learning_rate": 0.0005996762007555315,
+      "loss": 6.4523,
+      "step": 150
+    },
+    {
+      "epoch": 0.0215633423180593,
+      "grad_norm": 3.235063076019287,
+      "learning_rate": 0.000599352401511063,
+      "loss": 6.233,
+      "step": 200
+    },
+    {
+      "epoch": 0.026954177897574125,
+      "grad_norm": 1.762332797050476,
+      "learning_rate": 0.0005990286022665946,
+      "loss": 6.0855,
+      "step": 250
+    },
+    {
+      "epoch": 0.03234501347708895,
+      "grad_norm": 1.7542847394943237,
+      "learning_rate": 0.0005987048030221263,
+      "loss": 5.9844,
+      "step": 300
+    },
+    {
+      "epoch": 0.03773584905660377,
+      "grad_norm": 1.4935657978057861,
+      "learning_rate": 0.0005983810037776578,
+      "loss": 5.8773,
+      "step": 350
+    },
+    {
+      "epoch": 0.0431266846361186,
+      "grad_norm": 1.0085582733154297,
+      "learning_rate": 0.0005980572045331894,
+      "loss": 5.8265,
+      "step": 400
+    },
+    {
+      "epoch": 0.04851752021563342,
+      "grad_norm": 1.127617359161377,
+      "learning_rate": 0.0005977334052887209,
+      "loss": 5.7578,
+      "step": 450
+    },
+    {
+      "epoch": 0.05390835579514825,
+      "grad_norm": 1.225400686264038,
+      "learning_rate": 0.0005974096060442526,
+      "loss": 5.6615,
+      "step": 500
+    },
+    {
+      "epoch": 0.05929919137466307,
+      "grad_norm": 2.2902519702911377,
+      "learning_rate": 0.0005970858067997841,
+      "loss": 5.5826,
+      "step": 550
+    },
+    {
+      "epoch": 0.0646900269541779,
+      "grad_norm": 1.6163238286972046,
+      "learning_rate": 0.0005967620075553157,
+      "loss": 5.5064,
+      "step": 600
+    },
+    {
+      "epoch": 0.07008086253369272,
+      "grad_norm": 1.3570210933685303,
+      "learning_rate": 0.0005964382083108472,
+      "loss": 5.4486,
+      "step": 650
+    },
+    {
+      "epoch": 0.07547169811320754,
+      "grad_norm": 1.2458889484405518,
+      "learning_rate": 0.0005961144090663788,
+      "loss": 5.3952,
+      "step": 700
+    },
+    {
+      "epoch": 0.08086253369272237,
+      "grad_norm": 0.9130052924156189,
+      "learning_rate": 0.0005957906098219104,
+      "loss": 5.3212,
+      "step": 750
+    },
+    {
+      "epoch": 0.0862533692722372,
+      "grad_norm": 1.87125825881958,
+      "learning_rate": 0.0005954668105774419,
+      "loss": 5.2706,
+      "step": 800
+    },
+    {
+      "epoch": 0.09164420485175202,
+      "grad_norm": 1.3916375637054443,
+      "learning_rate": 0.0005951430113329735,
+      "loss": 5.2346,
+      "step": 850
+    },
+    {
+      "epoch": 0.09703504043126684,
+      "grad_norm": 1.6579746007919312,
+      "learning_rate": 0.0005948192120885051,
+      "loss": 5.1752,
+      "step": 900
+    },
+    {
+      "epoch": 0.10242587601078167,
+      "grad_norm": 1.5823800563812256,
+      "learning_rate": 0.0005944954128440366,
+      "loss": 5.1437,
+      "step": 950
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "grad_norm": 1.0492331981658936,
+      "learning_rate": 0.0005941716135995682,
+      "loss": 5.1075,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "eval_accuracy": 0.22580884197643514,
+      "eval_loss": 5.039173126220703,
+      "eval_runtime": 185.661,
+      "eval_samples_per_second": 97.01,
+      "eval_steps_per_second": 6.065,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11320754716981132,
+      "grad_norm": 0.9801273345947266,
+      "learning_rate": 0.0005938478143550997,
+      "loss": 5.057,
+      "step": 1050
+    },
+    {
+      "epoch": 0.11859838274932614,
+      "grad_norm": 1.28288996219635,
+      "learning_rate": 0.0005935240151106314,
+      "loss": 5.0318,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12398921832884097,
+      "grad_norm": 1.150609016418457,
+      "learning_rate": 0.0005932002158661629,
+      "loss": 5.0054,
+      "step": 1150
+    },
+    {
+      "epoch": 0.1293800539083558,
+      "grad_norm": 1.4465186595916748,
+      "learning_rate": 0.0005928764166216945,
+      "loss": 4.9628,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1347708894878706,
+      "grad_norm": 1.2153831720352173,
+      "learning_rate": 0.000592552617377226,
+      "loss": 4.9205,
+      "step": 1250
+    },
+    {
+      "epoch": 0.14016172506738545,
+      "grad_norm": 1.18161141872406,
+      "learning_rate": 0.0005922288181327577,
+      "loss": 4.8891,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14555256064690028,
+      "grad_norm": 0.8075419664382935,
+      "learning_rate": 0.0005919050188882893,
+      "loss": 4.8706,
+      "step": 1350
+    },
+    {
+      "epoch": 0.1509433962264151,
+      "grad_norm": 1.0504150390625,
+      "learning_rate": 0.0005915812196438207,
+      "loss": 4.8579,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15633423180592992,
+      "grad_norm": 0.7853661775588989,
+      "learning_rate": 0.0005912574203993524,
+      "loss": 4.8093,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16172506738544473,
+      "grad_norm": 1.0137704610824585,
+      "learning_rate": 0.0005909336211548839,
+      "loss": 4.8223,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16711590296495957,
+      "grad_norm": 1.0812759399414062,
+      "learning_rate": 0.0005906098219104155,
+      "loss": 4.7728,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1725067385444744,
+      "grad_norm": 1.1242338418960571,
+      "learning_rate": 0.000590286022665947,
+      "loss": 4.7731,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1778975741239892,
+      "grad_norm": 0.8945317268371582,
+      "learning_rate": 0.0005899622234214787,
+      "loss": 4.7251,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18328840970350405,
+      "grad_norm": 0.9912477731704712,
+      "learning_rate": 0.0005896384241770102,
+      "loss": 4.7168,
+      "step": 1700
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 0.9227299690246582,
+      "learning_rate": 0.0005893146249325418,
+      "loss": 4.682,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1940700808625337,
+      "grad_norm": 0.7492371201515198,
+      "learning_rate": 0.0005889908256880733,
+      "loss": 4.6867,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19946091644204852,
+      "grad_norm": 0.8827073574066162,
+      "learning_rate": 0.0005886670264436049,
+      "loss": 4.6577,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20485175202156333,
+      "grad_norm": 0.9550777673721313,
+      "learning_rate": 0.0005883432271991365,
+      "loss": 4.6425,
+      "step": 1900
+    },
+    {
+      "epoch": 0.21024258760107817,
+      "grad_norm": 1.091475248336792,
+      "learning_rate": 0.0005880194279546681,
+      "loss": 4.6154,
+      "step": 1950
+    },
+    {
+      "epoch": 0.215633423180593,
+      "grad_norm": 1.0472371578216553,
+      "learning_rate": 0.0005876956287101996,
+      "loss": 4.5986,
+      "step": 2000
+    },
+    {
+      "epoch": 0.215633423180593,
+      "eval_accuracy": 0.2688498363850489,
+      "eval_loss": 4.523857593536377,
+      "eval_runtime": 185.6509,
+      "eval_samples_per_second": 97.015,
+      "eval_steps_per_second": 6.065,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2210242587601078,
+      "grad_norm": 0.7373393177986145,
+      "learning_rate": 0.0005873718294657312,
+      "loss": 4.5762,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22641509433962265,
+      "grad_norm": 0.7511190176010132,
+      "learning_rate": 0.0005870480302212628,
+      "loss": 4.5611,
+      "step": 2100
+    },
+    {
+      "epoch": 0.23180592991913745,
+      "grad_norm": 1.0402768850326538,
+      "learning_rate": 0.0005867242309767943,
+      "loss": 4.536,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2371967654986523,
+      "grad_norm": 0.8001542091369629,
+      "learning_rate": 0.0005864004317323259,
+      "loss": 4.5156,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24258760107816713,
+      "grad_norm": 0.7276251912117004,
+      "learning_rate": 0.0005860766324878575,
+      "loss": 4.5009,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24797843665768193,
+      "grad_norm": 0.8818603157997131,
+      "learning_rate": 0.000585752833243389,
+      "loss": 4.5074,
+      "step": 2300
+    },
+    {
+      "epoch": 0.25336927223719674,
+      "grad_norm": 0.7163493037223816,
+      "learning_rate": 0.0005854290339989206,
+      "loss": 4.4634,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2587601078167116,
+      "grad_norm": 1.0233269929885864,
+      "learning_rate": 0.0005851052347544521,
+      "loss": 4.467,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2641509433962264,
+      "grad_norm": 1.0603983402252197,
+      "learning_rate": 0.0005847814355099838,
+      "loss": 4.4493,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2695417789757412,
+      "grad_norm": 0.9133924245834351,
+      "learning_rate": 0.0005844576362655154,
+      "loss": 4.4303,
+      "step": 2500
+    },
+    {
+      "epoch": 0.2749326145552561,
+      "grad_norm": 0.8938034772872925,
+      "learning_rate": 0.0005841338370210469,
+      "loss": 4.4168,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2803234501347709,
+      "grad_norm": 0.8274545073509216,
+      "learning_rate": 0.0005838100377765785,
+      "loss": 4.3874,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 0.8141469955444336,
+      "learning_rate": 0.0005834862385321101,
+      "loss": 4.4009,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29110512129380056,
+      "grad_norm": 0.9526525735855103,
+      "learning_rate": 0.0005831624392876417,
+      "loss": 4.399,
+      "step": 2700
+    },
+    {
+      "epoch": 0.29649595687331537,
+      "grad_norm": 0.8524094223976135,
+      "learning_rate": 0.0005828386400431731,
+      "loss": 4.3639,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3018867924528302,
+      "grad_norm": 0.6328749656677246,
+      "learning_rate": 0.0005825148407987048,
+      "loss": 4.3386,
+      "step": 2800
+    },
+    {
+      "epoch": 0.30727762803234504,
+      "grad_norm": 0.8146114945411682,
+      "learning_rate": 0.0005821910415542363,
+      "loss": 4.3433,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31266846361185985,
+      "grad_norm": 0.7974101901054382,
+      "learning_rate": 0.0005818672423097679,
+      "loss": 4.3344,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31805929919137466,
+      "grad_norm": 0.8762542605400085,
+      "learning_rate": 0.0005815434430652994,
+      "loss": 4.3216,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "grad_norm": 0.7969277501106262,
+      "learning_rate": 0.0005812196438208311,
+      "loss": 4.3212,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "eval_accuracy": 0.2972257332195665,
+      "eval_loss": 4.256451606750488,
+      "eval_runtime": 185.891,
+      "eval_samples_per_second": 96.89,
+      "eval_steps_per_second": 6.057,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3288409703504043,
+      "grad_norm": 0.9131352305412292,
+      "learning_rate": 0.0005808958445763626,
+      "loss": 4.3095,
+      "step": 3050
+    },
+    {
+      "epoch": 0.33423180592991913,
+      "grad_norm": 0.8107157349586487,
+      "learning_rate": 0.0005805720453318942,
+      "loss": 4.3159,
+      "step": 3100
+    },
+    {
+      "epoch": 0.33962264150943394,
+      "grad_norm": 0.647481381893158,
+      "learning_rate": 0.0005802482460874257,
+      "loss": 4.272,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3450134770889488,
+      "grad_norm": 0.7248066067695618,
+      "learning_rate": 0.0005799244468429573,
+      "loss": 4.268,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3504043126684636,
+      "grad_norm": 0.8651499152183533,
+      "learning_rate": 0.0005796006475984889,
+      "loss": 4.2713,
+      "step": 3250
+    },
+    {
+      "epoch": 0.3557951482479784,
+      "grad_norm": 0.6976764798164368,
+      "learning_rate": 0.0005792768483540205,
+      "loss": 4.2541,
+      "step": 3300
+    },
+    {
+      "epoch": 0.3611859838274933,
+      "grad_norm": 0.698954164981842,
+      "learning_rate": 0.000578953049109552,
+      "loss": 4.267,
+      "step": 3350
+    },
+    {
+      "epoch": 0.3665768194070081,
+      "grad_norm": 0.6978999376296997,
+      "learning_rate": 0.0005786292498650836,
+      "loss": 4.2384,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3719676549865229,
+      "grad_norm": 0.6306410431861877,
+      "learning_rate": 0.0005783054506206152,
+      "loss": 4.2154,
+      "step": 3450
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.6356645822525024,
+      "learning_rate": 0.0005779816513761467,
+      "loss": 4.2289,
+      "step": 3500
+    },
+    {
+      "epoch": 0.38274932614555257,
+      "grad_norm": 0.7585115432739258,
+      "learning_rate": 0.0005776578521316782,
+      "loss": 4.2245,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3881401617250674,
+      "grad_norm": 0.5995826125144958,
+      "learning_rate": 0.0005773340528872099,
+      "loss": 4.2194,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3935309973045822,
+      "grad_norm": 0.7846726179122925,
+      "learning_rate": 0.0005770102536427414,
+      "loss": 4.198,
+      "step": 3650
+    },
+    {
+      "epoch": 0.39892183288409705,
+      "grad_norm": 0.7590677738189697,
+      "learning_rate": 0.000576686454398273,
+      "loss": 4.1988,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40431266846361186,
+      "grad_norm": 0.6893619298934937,
+      "learning_rate": 0.0005763626551538045,
+      "loss": 4.2032,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40970350404312667,
+      "grad_norm": 0.7359482049942017,
+      "learning_rate": 0.0005760388559093362,
+      "loss": 4.1966,
+      "step": 3800
+    },
+    {
+      "epoch": 0.41509433962264153,
+      "grad_norm": 0.626078724861145,
+      "learning_rate": 0.0005757150566648678,
+      "loss": 4.1772,
+      "step": 3850
+    },
+    {
+      "epoch": 0.42048517520215634,
+      "grad_norm": 0.7311425805091858,
+      "learning_rate": 0.0005753912574203993,
+      "loss": 4.1749,
+      "step": 3900
+    },
+    {
+      "epoch": 0.42587601078167114,
+      "grad_norm": 0.7368374466896057,
+      "learning_rate": 0.0005750674581759309,
+      "loss": 4.1665,
+      "step": 3950
+    },
+    {
+      "epoch": 0.431266846361186,
+      "grad_norm": 0.7938806414604187,
+      "learning_rate": 0.0005747436589314624,
+      "loss": 4.1636,
+      "step": 4000
+    },
+    {
+      "epoch": 0.431266846361186,
+      "eval_accuracy": 0.31286490393291944,
+      "eval_loss": 4.090136528015137,
+      "eval_runtime": 185.719,
+      "eval_samples_per_second": 96.98,
+      "eval_steps_per_second": 6.063,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4366576819407008,
+      "grad_norm": 0.6016411185264587,
+      "learning_rate": 0.0005744198596869941,
+      "loss": 4.154,
+      "step": 4050
+    },
+    {
+      "epoch": 0.4420485175202156,
+      "grad_norm": 0.7181369066238403,
+      "learning_rate": 0.0005740960604425255,
+      "loss": 4.1586,
+      "step": 4100
+    },
+    {
+      "epoch": 0.4474393530997305,
+      "grad_norm": 0.689071536064148,
+      "learning_rate": 0.0005737722611980572,
+      "loss": 4.1578,
+      "step": 4150
+    },
+    {
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.6437532305717468,
+      "learning_rate": 0.0005734484619535887,
+      "loss": 4.1432,
+      "step": 4200
+    },
+    {
+      "epoch": 0.4582210242587601,
+      "grad_norm": 0.6600464582443237,
+      "learning_rate": 0.0005731246627091203,
+      "loss": 4.1285,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4636118598382749,
+      "grad_norm": 0.8179474472999573,
+      "learning_rate": 0.0005728008634646518,
+      "loss": 4.1285,
+      "step": 4300
+    },
+    {
+      "epoch": 0.46900269541778977,
+      "grad_norm": 0.7125003337860107,
+      "learning_rate": 0.0005724770642201835,
+      "loss": 4.1222,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4743935309973046,
+      "grad_norm": 0.6154434084892273,
+      "learning_rate": 0.000572153264975715,
+      "loss": 4.1229,
+      "step": 4400
+    },
+    {
+      "epoch": 0.4797843665768194,
+      "grad_norm": 0.7257527112960815,
+      "learning_rate": 0.0005718294657312466,
+      "loss": 4.1081,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48517520215633425,
+      "grad_norm": 0.6786518692970276,
+      "learning_rate": 0.0005715056664867781,
+      "loss": 4.1152,
+      "step": 4500
+    },
+    {
+      "epoch": 0.49056603773584906,
+      "grad_norm": 0.6423896551132202,
+      "learning_rate": 0.0005711818672423097,
+      "loss": 4.1087,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49595687331536387,
+      "grad_norm": 0.7800512313842773,
+      "learning_rate": 0.0005708580679978413,
+      "loss": 4.0953,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5013477088948787,
+      "grad_norm": 0.6792780756950378,
+      "learning_rate": 0.0005705342687533729,
+      "loss": 4.0844,
+      "step": 4650
+    },
+    {
+      "epoch": 0.5067385444743935,
+      "grad_norm": 0.6257168054580688,
+      "learning_rate": 0.0005702104695089044,
+      "loss": 4.099,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5121293800539084,
+      "grad_norm": 0.6129534840583801,
+      "learning_rate": 0.000569886670264436,
+      "loss": 4.0868,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5175202156334232,
+      "grad_norm": 0.7634662985801697,
+      "learning_rate": 0.0005695628710199675,
+      "loss": 4.0839,
+      "step": 4800
+    },
+    {
+      "epoch": 0.522911051212938,
+      "grad_norm": 0.6432034373283386,
+      "learning_rate": 0.0005692390717754991,
+      "loss": 4.0711,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5283018867924528,
+      "grad_norm": 0.6302817463874817,
+      "learning_rate": 0.0005689152725310306,
+      "loss": 4.0586,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5336927223719676,
+      "grad_norm": 0.6604870557785034,
+      "learning_rate": 0.0005685914732865623,
+      "loss": 4.0787,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "grad_norm": 0.6179905533790588,
+      "learning_rate": 0.0005682676740420939,
+      "loss": 4.0541,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "eval_accuracy": 0.32092564437410015,
+      "eval_loss": 3.9973857402801514,
+      "eval_runtime": 185.7325,
+      "eval_samples_per_second": 96.973,
+      "eval_steps_per_second": 6.062,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5444743935309974,
+      "grad_norm": 0.6523880362510681,
+      "learning_rate": 0.0005679438747976254,
+      "loss": 4.0653,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5498652291105122,
+      "grad_norm": 0.632424533367157,
+      "learning_rate": 0.000567620075553157,
+      "loss": 4.0535,
+      "step": 5100
+    },
+    {
+      "epoch": 0.555256064690027,
+      "grad_norm": 0.6873350143432617,
+      "learning_rate": 0.0005672962763086886,
+      "loss": 4.0546,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5606469002695418,
+      "grad_norm": 0.6244584918022156,
+      "learning_rate": 0.0005669724770642202,
+      "loss": 4.0717,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.6625449657440186,
+      "learning_rate": 0.0005666486778197517,
+      "loss": 4.049,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 0.6838392019271851,
+      "learning_rate": 0.0005663248785752833,
+      "loss": 4.0368,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5768194070080862,
+      "grad_norm": 0.5923067927360535,
+      "learning_rate": 0.0005660010793308148,
+      "loss": 4.0328,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5822102425876011,
+      "grad_norm": 0.5898548364639282,
+      "learning_rate": 0.0005656772800863465,
+      "loss": 4.0444,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5876010781671159,
+      "grad_norm": 0.578851580619812,
+      "learning_rate": 0.0005653534808418779,
+      "loss": 4.044,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5929919137466307,
+      "grad_norm": 0.695020318031311,
+      "learning_rate": 0.0005650296815974096,
+      "loss": 4.0243,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5983827493261455,
+      "grad_norm": 0.6661456227302551,
+      "learning_rate": 0.0005647058823529411,
+      "loss": 4.0146,
+      "step": 5550
+    },
+    {
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.5961174368858337,
+      "learning_rate": 0.0005643820831084727,
+      "loss": 4.0297,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6091644204851752,
+      "grad_norm": 0.6652942299842834,
+      "learning_rate": 0.0005640582838640042,
+      "loss": 4.0088,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6145552560646901,
+      "grad_norm": 0.5791774392127991,
+      "learning_rate": 0.0005637344846195358,
+      "loss": 4.0071,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6199460916442049,
+      "grad_norm": 0.6387227773666382,
+      "learning_rate": 0.0005634106853750674,
+      "loss": 3.9825,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6253369272237197,
+      "grad_norm": 0.6057057976722717,
+      "learning_rate": 0.000563086886130599,
+      "loss": 3.9925,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6307277628032345,
+      "grad_norm": 0.6438758373260498,
+      "learning_rate": 0.0005627630868861305,
+      "loss": 3.9996,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6361185983827493,
+      "grad_norm": 0.6182964444160461,
+      "learning_rate": 0.0005624392876416621,
+      "loss": 4.0009,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6415094339622641,
+      "grad_norm": 0.7100219130516052,
+      "learning_rate": 0.0005621154883971937,
+      "loss": 3.9875,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "grad_norm": 0.5899308919906616,
+      "learning_rate": 0.0005617916891527253,
+      "loss": 3.9992,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "eval_accuracy": 0.3278621533850644,
+      "eval_loss": 3.922297954559326,
+      "eval_runtime": 185.7282,
+      "eval_samples_per_second": 96.975,
+      "eval_steps_per_second": 6.063,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6522911051212938,
+      "grad_norm": 0.648195207118988,
+      "learning_rate": 0.0005614678899082568,
+      "loss": 3.9705,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6576819407008087,
+      "grad_norm": 0.6435657739639282,
+      "learning_rate": 0.0005611440906637884,
+      "loss": 4.003,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6630727762803235,
+      "grad_norm": 0.6395747661590576,
+      "learning_rate": 0.00056082029141932,
+      "loss": 3.9875,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6684636118598383,
+      "grad_norm": 0.6319500803947449,
+      "learning_rate": 0.0005604964921748515,
+      "loss": 3.9561,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6738544474393531,
+      "grad_norm": 0.5592441558837891,
+      "learning_rate": 0.000560172692930383,
+      "loss": 3.9777,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6792452830188679,
+      "grad_norm": 0.6740835905075073,
+      "learning_rate": 0.0005598488936859147,
+      "loss": 3.9795,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6846361185983828,
+      "grad_norm": 0.6001284122467041,
+      "learning_rate": 0.0005595250944414463,
+      "loss": 3.9655,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6900269541778976,
+      "grad_norm": 0.7173512578010559,
+      "learning_rate": 0.0005592012951969778,
+      "loss": 3.9756,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6954177897574124,
+      "grad_norm": 0.6381141543388367,
+      "learning_rate": 0.0005588774959525094,
+      "loss": 3.9718,
+      "step": 6450
+    },
+    {
+      "epoch": 0.7008086253369272,
+      "grad_norm": 0.6382541060447693,
+      "learning_rate": 0.000558553696708041,
+      "loss": 3.9636,
+      "step": 6500
+    },
+    {
+      "epoch": 0.706199460916442,
+      "grad_norm": 0.6090039610862732,
+      "learning_rate": 0.0005582298974635726,
+      "loss": 3.9616,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7115902964959568,
+      "grad_norm": 0.5756356120109558,
+      "learning_rate": 0.0005579060982191041,
+      "loss": 3.965,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7169811320754716,
+      "grad_norm": 0.6311583518981934,
+      "learning_rate": 0.0005575887749595251,
+      "loss": 3.974,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7223719676549866,
+      "grad_norm": 0.6047680377960205,
+      "learning_rate": 0.0005572649757150566,
+      "loss": 3.9372,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7277628032345014,
+      "grad_norm": 0.5887506604194641,
+      "learning_rate": 0.0005569411764705882,
+      "loss": 3.9657,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7331536388140162,
+      "grad_norm": 0.6374669671058655,
+      "learning_rate": 0.0005566173772261198,
+      "loss": 3.9388,
+      "step": 6800
+    },
+    {
+      "epoch": 0.738544474393531,
+      "grad_norm": 0.6812782287597656,
+      "learning_rate": 0.0005562935779816513,
+      "loss": 3.9383,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7439353099730458,
+      "grad_norm": 0.5230938196182251,
+      "learning_rate": 0.0005559697787371828,
+      "loss": 3.9392,
+      "step": 6900
+    },
+    {
+      "epoch": 0.7493261455525606,
+      "grad_norm": 0.6506929397583008,
+      "learning_rate": 0.0005556459794927145,
+      "loss": 3.9444,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.6598263382911682,
+      "learning_rate": 0.000555322180248246,
+      "loss": 3.9262,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "eval_accuracy": 0.33297796595492146,
+      "eval_loss": 3.8664188385009766,
+      "eval_runtime": 185.5664,
+      "eval_samples_per_second": 97.06,
+      "eval_steps_per_second": 6.068,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7601078167115903,
+      "grad_norm": 0.7245527505874634,
+      "learning_rate": 0.0005549983810037776,
+      "loss": 3.9384,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7654986522911051,
+      "grad_norm": 0.8341423869132996,
+      "learning_rate": 0.0005546745817593091,
+      "loss": 3.9113,
+      "step": 7100
+    },
+    {
+      "epoch": 0.77088948787062,
+      "grad_norm": 0.679886519908905,
+      "learning_rate": 0.0005543507825148408,
+      "loss": 3.9386,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7762803234501348,
+      "grad_norm": 0.5818988084793091,
+      "learning_rate": 0.0005540269832703723,
+      "loss": 3.9105,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7816711590296496,
+      "grad_norm": 0.6063448190689087,
+      "learning_rate": 0.0005537031840259039,
+      "loss": 3.9036,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7870619946091644,
+      "grad_norm": 0.5180264711380005,
+      "learning_rate": 0.0005533793847814354,
+      "loss": 3.9169,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7924528301886793,
+      "grad_norm": 0.6858142018318176,
+      "learning_rate": 0.000553055585536967,
+      "loss": 3.9194,
+      "step": 7350
+    },
+    {
+      "epoch": 0.7978436657681941,
+      "grad_norm": 0.5290060043334961,
+      "learning_rate": 0.0005527317862924987,
+      "loss": 3.9125,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8032345013477089,
+      "grad_norm": 0.6154426336288452,
+      "learning_rate": 0.0005524079870480301,
+      "loss": 3.8904,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8086253369272237,
+      "grad_norm": 0.5800319314002991,
+      "learning_rate": 0.0005520841878035618,
+      "loss": 3.9076,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8140161725067385,
+      "grad_norm": 0.5865637063980103,
+      "learning_rate": 0.0005517603885590933,
+      "loss": 3.9016,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8194070080862533,
+      "grad_norm": 0.6085811853408813,
+      "learning_rate": 0.0005514365893146249,
+      "loss": 3.8946,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8247978436657682,
+      "grad_norm": 0.5241173505783081,
+      "learning_rate": 0.0005511127900701564,
+      "loss": 3.9015,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8301886792452831,
+      "grad_norm": 0.6474875211715698,
+      "learning_rate": 0.000550788990825688,
+      "loss": 3.8947,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8355795148247979,
+      "grad_norm": 0.6207894086837769,
+      "learning_rate": 0.0005504651915812196,
+      "loss": 3.8954,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8409703504043127,
+      "grad_norm": 0.6342117190361023,
+      "learning_rate": 0.0005501413923367512,
+      "loss": 3.8791,
+      "step": 7800
+    },
+    {
+      "epoch": 0.8463611859838275,
+      "grad_norm": 0.5856869220733643,
+      "learning_rate": 0.0005498175930922827,
+      "loss": 3.8858,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8517520215633423,
+      "grad_norm": 0.5404996871948242,
+      "learning_rate": 0.0005494937938478143,
+      "loss": 3.8815,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.5928559899330139,
+      "learning_rate": 0.0005491699946033459,
+      "loss": 3.9119,
+      "step": 7950
+    },
+    {
+      "epoch": 0.862533692722372,
+      "grad_norm": 0.6332123279571533,
+      "learning_rate": 0.0005488461953588775,
+      "loss": 3.8584,
+      "step": 8000
+    },
+    {
+      "epoch": 0.862533692722372,
+      "eval_accuracy": 0.33728094627103833,
+      "eval_loss": 3.820244312286377,
+      "eval_runtime": 185.4099,
+      "eval_samples_per_second": 97.142,
+      "eval_steps_per_second": 6.073,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8679245283018868,
+      "grad_norm": 0.5683950185775757,
+      "learning_rate": 0.000548522396114409,
+      "loss": 3.899,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8733153638814016,
+      "grad_norm": 0.6757128238677979,
+      "learning_rate": 0.0005481985968699406,
+      "loss": 3.8782,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8787061994609164,
+      "grad_norm": 0.5428512096405029,
+      "learning_rate": 0.0005478747976254721,
+      "loss": 3.8809,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8840970350404312,
+      "grad_norm": 0.5965107679367065,
+      "learning_rate": 0.0005475509983810037,
+      "loss": 3.8741,
+      "step": 8200
+    },
+    {
+      "epoch": 0.889487870619946,
+      "grad_norm": 0.6327036023139954,
+      "learning_rate": 0.0005472271991365352,
+      "loss": 3.8721,
+      "step": 8250
+    },
+    {
+      "epoch": 0.894878706199461,
+      "grad_norm": 0.6294053196907043,
+      "learning_rate": 0.0005469033998920669,
+      "loss": 3.8905,
+      "step": 8300
+    },
+    {
+      "epoch": 0.9002695417789758,
+      "grad_norm": 0.557381808757782,
+      "learning_rate": 0.0005465796006475984,
+      "loss": 3.8606,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.5703429579734802,
+      "learning_rate": 0.00054625580140313,
+      "loss": 3.8674,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9110512129380054,
+      "grad_norm": 0.4978976249694824,
+      "learning_rate": 0.0005459320021586615,
+      "loss": 3.8631,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9164420485175202,
+      "grad_norm": 0.6423895955085754,
+      "learning_rate": 0.0005456082029141932,
+      "loss": 3.8677,
+      "step": 8500
+    },
+    {
+      "epoch": 0.921832884097035,
+      "grad_norm": 0.5929063558578491,
+      "learning_rate": 0.0005452844036697248,
+      "loss": 3.8608,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9272237196765498,
+      "grad_norm": 0.6158359050750732,
+      "learning_rate": 0.0005449606044252563,
+      "loss": 3.8686,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9326145552560647,
+      "grad_norm": 0.6147903800010681,
+      "learning_rate": 0.0005446368051807879,
+      "loss": 3.8496,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9380053908355795,
+      "grad_norm": 0.5660675764083862,
+      "learning_rate": 0.0005443194819212088,
+      "loss": 3.8557,
+      "step": 8700
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.6215540766716003,
+      "learning_rate": 0.0005439956826767404,
+      "loss": 3.8513,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9487870619946092,
+      "grad_norm": 0.5756900906562805,
+      "learning_rate": 0.000543671883432272,
+      "loss": 3.8535,
+      "step": 8800
+    },
+    {
+      "epoch": 0.954177897574124,
+      "grad_norm": 0.617630124092102,
+      "learning_rate": 0.0005433480841878035,
+      "loss": 3.8531,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9595687331536388,
+      "grad_norm": 0.5321480631828308,
+      "learning_rate": 0.000543024284943335,
+      "loss": 3.8434,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9649595687331537,
+      "grad_norm": 0.5432807803153992,
+      "learning_rate": 0.0005427004856988667,
+      "loss": 3.8327,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "grad_norm": 0.6044248938560486,
+      "learning_rate": 0.0005423766864543982,
+      "loss": 3.8593,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "eval_accuracy": 0.3409367899873322,
+      "eval_loss": 3.782257080078125,
+      "eval_runtime": 185.7982,
+      "eval_samples_per_second": 96.939,
+      "eval_steps_per_second": 6.06,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9757412398921833,
+      "grad_norm": 0.6063427329063416,
+      "learning_rate": 0.0005420528872099298,
+      "loss": 3.8722,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9811320754716981,
+      "grad_norm": 0.5438401103019714,
+      "learning_rate": 0.0005417290879654613,
+      "loss": 3.8278,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9865229110512129,
+      "grad_norm": 0.6635690927505493,
+      "learning_rate": 0.000541405288720993,
+      "loss": 3.8442,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9919137466307277,
+      "grad_norm": 0.5909547209739685,
+      "learning_rate": 0.0005410814894765245,
+      "loss": 3.8334,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9973045822102425,
+      "grad_norm": 0.56688392162323,
+      "learning_rate": 0.0005407576902320561,
+      "loss": 3.849,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0026954177897573,
+      "grad_norm": 0.584598183631897,
+      "learning_rate": 0.0005404338909875876,
+      "loss": 3.7986,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0080862533692723,
+      "grad_norm": 0.6157205104827881,
+      "learning_rate": 0.0005401100917431192,
+      "loss": 3.7658,
+      "step": 9350
+    },
+    {
+      "epoch": 1.013477088948787,
+      "grad_norm": 0.6105006337165833,
+      "learning_rate": 0.0005397862924986508,
+      "loss": 3.7606,
+      "step": 9400
+    },
+    {
+      "epoch": 1.0188679245283019,
+      "grad_norm": 0.5399526357650757,
+      "learning_rate": 0.0005394624932541824,
+      "loss": 3.7589,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0242587601078168,
+      "grad_norm": 0.5277557969093323,
+      "learning_rate": 0.0005391386940097139,
+      "loss": 3.7623,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0296495956873315,
+      "grad_norm": 0.6086890697479248,
+      "learning_rate": 0.0005388148947652455,
+      "loss": 3.7727,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0350404312668464,
+      "grad_norm": 0.6136114001274109,
+      "learning_rate": 0.000538491095520777,
+      "loss": 3.7704,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0404312668463611,
+      "grad_norm": 0.6060198545455933,
+      "learning_rate": 0.0005381672962763086,
+      "loss": 3.7771,
+      "step": 9650
+    },
+    {
+      "epoch": 1.045822102425876,
+      "grad_norm": 0.582412838935852,
+      "learning_rate": 0.0005378434970318403,
+      "loss": 3.7608,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0512129380053907,
+      "grad_norm": 0.5505673289299011,
+      "learning_rate": 0.0005375196977873718,
+      "loss": 3.7567,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0566037735849056,
+      "grad_norm": 0.5291816592216492,
+      "learning_rate": 0.0005371958985429034,
+      "loss": 3.7866,
+      "step": 9800
+    },
+    {
+      "epoch": 1.0619946091644206,
+      "grad_norm": 0.5005487203598022,
+      "learning_rate": 0.0005368720992984349,
+      "loss": 3.7863,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0673854447439353,
+      "grad_norm": 0.5914438962936401,
+      "learning_rate": 0.0005365483000539665,
+      "loss": 3.7772,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0727762803234502,
+      "grad_norm": 0.5673604607582092,
+      "learning_rate": 0.0005362245008094981,
+      "loss": 3.7877,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "grad_norm": 0.5731451511383057,
+      "learning_rate": 0.0005359007015650297,
+      "loss": 3.7652,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "eval_accuracy": 0.3447277978960672,
+      "eval_loss": 3.7502317428588867,
+      "eval_runtime": 185.8269,
+      "eval_samples_per_second": 96.924,
+      "eval_steps_per_second": 6.059,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0835579514824798,
+      "grad_norm": 0.5782760977745056,
+      "learning_rate": 0.0005355769023205612,
+      "loss": 3.7612,
+      "step": 10050
+    },
+    {
+      "epoch": 1.0889487870619945,
+      "grad_norm": 0.6518768668174744,
+      "learning_rate": 0.0005352531030760928,
+      "loss": 3.7643,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0943396226415094,
+      "grad_norm": 0.574468195438385,
+      "learning_rate": 0.0005349293038316244,
+      "loss": 3.7578,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0997304582210243,
+      "grad_norm": 0.6225487589836121,
+      "learning_rate": 0.0005346055045871559,
+      "loss": 3.7697,
+      "step": 10200
+    },
+    {
+      "epoch": 1.105121293800539,
+      "grad_norm": 0.6141579151153564,
+      "learning_rate": 0.0005342817053426874,
+      "loss": 3.756,
+      "step": 10250
+    },
+    {
+      "epoch": 1.110512129380054,
+      "grad_norm": 0.6093278527259827,
+      "learning_rate": 0.0005339579060982191,
+      "loss": 3.7653,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1159029649595686,
+      "grad_norm": 0.5892840623855591,
+      "learning_rate": 0.0005336341068537506,
+      "loss": 3.7616,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1212938005390836,
+      "grad_norm": 0.6228449940681458,
+      "learning_rate": 0.0005333103076092822,
+      "loss": 3.7674,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1266846361185983,
+      "grad_norm": 0.5601798295974731,
+      "learning_rate": 0.0005329865083648137,
+      "loss": 3.7639,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1320754716981132,
+      "grad_norm": 0.6022484302520752,
+      "learning_rate": 0.0005326627091203454,
+      "loss": 3.766,
+      "step": 10500
+    },
+    {
+      "epoch": 1.137466307277628,
+      "grad_norm": 0.6059293150901794,
+      "learning_rate": 0.0005323389098758769,
+      "loss": 3.7699,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.5978447794914246,
+      "learning_rate": 0.0005320151106314085,
+      "loss": 3.7596,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1482479784366577,
+      "grad_norm": 0.5976098775863647,
+      "learning_rate": 0.00053169131138694,
+      "loss": 3.7444,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1536388140161726,
+      "grad_norm": 0.5541641116142273,
+      "learning_rate": 0.000531373988127361,
+      "loss": 3.7588,
+      "step": 10700
+    },
+    {
+      "epoch": 1.1590296495956873,
+      "grad_norm": 0.579910159111023,
+      "learning_rate": 0.0005310501888828925,
+      "loss": 3.7386,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1644204851752022,
+      "grad_norm": 0.5855928063392639,
+      "learning_rate": 0.0005307263896384242,
+      "loss": 3.7429,
+      "step": 10800
+    },
+    {
+      "epoch": 1.169811320754717,
+      "grad_norm": 0.5664414167404175,
+      "learning_rate": 0.0005304025903939556,
+      "loss": 3.7341,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1752021563342319,
+      "grad_norm": 0.5997658967971802,
+      "learning_rate": 0.0005300787911494873,
+      "loss": 3.7482,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1805929919137466,
+      "grad_norm": 0.532941997051239,
+      "learning_rate": 0.0005297549919050189,
+      "loss": 3.7631,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "grad_norm": 0.597166895866394,
+      "learning_rate": 0.0005294311926605504,
+      "loss": 3.7744,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "eval_accuracy": 0.3471044711641212,
+      "eval_loss": 3.7261691093444824,
+      "eval_runtime": 185.6183,
+      "eval_samples_per_second": 97.032,
+      "eval_steps_per_second": 6.066,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1913746630727764,
+      "grad_norm": 0.5834847688674927,
+      "learning_rate": 0.000529107393416082,
+      "loss": 3.7594,
+      "step": 11050
+    },
+    {
+      "epoch": 1.196765498652291,
+      "grad_norm": 0.5739508867263794,
+      "learning_rate": 0.0005287835941716135,
+      "loss": 3.7444,
+      "step": 11100
+    },
+    {
+      "epoch": 1.202156334231806,
+      "grad_norm": 0.5817961692810059,
+      "learning_rate": 0.0005284597949271452,
+      "loss": 3.7436,
+      "step": 11150
+    },
+    {
+      "epoch": 1.2075471698113207,
+      "grad_norm": 0.5332579016685486,
+      "learning_rate": 0.0005281359956826767,
+      "loss": 3.7644,
+      "step": 11200
+    },
+    {
+      "epoch": 1.2129380053908356,
+      "grad_norm": 0.6148272752761841,
+      "learning_rate": 0.0005278121964382083,
+      "loss": 3.7533,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2183288409703503,
+      "grad_norm": 0.526567280292511,
+      "learning_rate": 0.0005274883971937398,
+      "loss": 3.7374,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2237196765498652,
+      "grad_norm": 0.5606340169906616,
+      "learning_rate": 0.0005271645979492714,
+      "loss": 3.7345,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2291105121293802,
+      "grad_norm": 0.6641140580177307,
+      "learning_rate": 0.000526840798704803,
+      "loss": 3.7478,
+      "step": 11400
+    },
+    {
+      "epoch": 1.2345013477088949,
+      "grad_norm": 0.5626732707023621,
+      "learning_rate": 0.0005265169994603346,
+      "loss": 3.7573,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2398921832884098,
+      "grad_norm": 0.5549135804176331,
+      "learning_rate": 0.0005261932002158661,
+      "loss": 3.7479,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2452830188679245,
+      "grad_norm": 0.5892812013626099,
+      "learning_rate": 0.0005258694009713977,
+      "loss": 3.7413,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2506738544474394,
+      "grad_norm": 0.5910210609436035,
+      "learning_rate": 0.0005255456017269292,
+      "loss": 3.7553,
+      "step": 11600
+    },
+    {
+      "epoch": 1.256064690026954,
+      "grad_norm": 0.5582822561264038,
+      "learning_rate": 0.0005252218024824608,
+      "loss": 3.7349,
+      "step": 11650
+    },
+    {
+      "epoch": 1.261455525606469,
+      "grad_norm": 0.5085506439208984,
+      "learning_rate": 0.0005248980032379924,
+      "loss": 3.7464,
+      "step": 11700
+    },
+    {
+      "epoch": 1.266846361185984,
+      "grad_norm": 0.5303859710693359,
+      "learning_rate": 0.000524574203993524,
+      "loss": 3.7452,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2722371967654986,
+      "grad_norm": 0.5914645791053772,
+      "learning_rate": 0.0005242504047490555,
+      "loss": 3.7586,
+      "step": 11800
+    },
+    {
+      "epoch": 1.2776280323450135,
+      "grad_norm": 0.5623352527618408,
+      "learning_rate": 0.0005239266055045871,
+      "loss": 3.7425,
+      "step": 11850
+    },
+    {
+      "epoch": 1.2830188679245282,
+      "grad_norm": 0.6508404612541199,
+      "learning_rate": 0.0005236028062601186,
+      "loss": 3.7365,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2884097035040432,
+      "grad_norm": 0.6292461156845093,
+      "learning_rate": 0.0005232790070156503,
+      "loss": 3.7201,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "grad_norm": 0.577590823173523,
+      "learning_rate": 0.0005229552077711818,
+      "loss": 3.7343,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "eval_accuracy": 0.3496051173771714,
+      "eval_loss": 3.6999623775482178,
+      "eval_runtime": 185.9657,
+      "eval_samples_per_second": 96.851,
+      "eval_steps_per_second": 6.055,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2991913746630728,
+      "grad_norm": 0.5652825832366943,
+      "learning_rate": 0.0005226314085267134,
+      "loss": 3.7482,
+      "step": 12050
+    },
+    {
+      "epoch": 1.3045822102425877,
+      "grad_norm": 0.6255879402160645,
+      "learning_rate": 0.000522307609282245,
+      "loss": 3.7434,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3099730458221024,
+      "grad_norm": 0.5456310510635376,
+      "learning_rate": 0.0005219838100377766,
+      "loss": 3.7317,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3153638814016173,
+      "grad_norm": 0.548086404800415,
+      "learning_rate": 0.000521660010793308,
+      "loss": 3.7471,
+      "step": 12200
+    },
+    {
+      "epoch": 1.320754716981132,
+      "grad_norm": 0.5980852842330933,
+      "learning_rate": 0.0005213362115488396,
+      "loss": 3.7216,
+      "step": 12250
+    },
+    {
+      "epoch": 1.326145552560647,
+      "grad_norm": 0.6635577082633972,
+      "learning_rate": 0.0005210124123043713,
+      "loss": 3.729,
+      "step": 12300
+    },
+    {
+      "epoch": 1.3315363881401616,
+      "grad_norm": 0.5238260626792908,
+      "learning_rate": 0.0005206886130599028,
+      "loss": 3.7393,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3369272237196765,
+      "grad_norm": 0.5679298043251038,
+      "learning_rate": 0.0005203648138154344,
+      "loss": 3.7351,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3423180592991915,
+      "grad_norm": 0.5958516597747803,
+      "learning_rate": 0.0005200410145709659,
+      "loss": 3.7232,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3477088948787062,
+      "grad_norm": 0.5935837626457214,
+      "learning_rate": 0.0005197172153264976,
+      "loss": 3.7423,
+      "step": 12500
+    },
+    {
+      "epoch": 1.353099730458221,
+      "grad_norm": 0.5568834543228149,
+      "learning_rate": 0.0005193934160820291,
+      "loss": 3.7151,
+      "step": 12550
+    },
+    {
+      "epoch": 1.3584905660377358,
+      "grad_norm": 0.5727585554122925,
+      "learning_rate": 0.0005190696168375607,
+      "loss": 3.7149,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3638814016172507,
+      "grad_norm": 0.5959082841873169,
+      "learning_rate": 0.0005187458175930922,
+      "loss": 3.7073,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3692722371967654,
+      "grad_norm": 0.5620747804641724,
+      "learning_rate": 0.0005184284943335132,
+      "loss": 3.7209,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3746630727762803,
+      "grad_norm": 0.6341102123260498,
+      "learning_rate": 0.0005181046950890447,
+      "loss": 3.712,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3800539083557952,
+      "grad_norm": 0.5566592812538147,
+      "learning_rate": 0.0005177808958445764,
+      "loss": 3.7116,
+      "step": 12800
+    },
+    {
+      "epoch": 1.38544474393531,
+      "grad_norm": 0.6411048173904419,
+      "learning_rate": 0.0005174570966001078,
+      "loss": 3.7213,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3908355795148248,
+      "grad_norm": 0.637323260307312,
+      "learning_rate": 0.0005171332973556395,
+      "loss": 3.7198,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3962264150943398,
+      "grad_norm": 0.5226753950119019,
+      "learning_rate": 0.000516809498111171,
+      "loss": 3.7298,
+      "step": 12950
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "grad_norm": 0.5814768671989441,
+      "learning_rate": 0.0005164856988667026,
+      "loss": 3.7098,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "eval_accuracy": 0.35190584227664307,
+      "eval_loss": 3.6751368045806885,
+      "eval_runtime": 185.5194,
+      "eval_samples_per_second": 97.084,
+      "eval_steps_per_second": 6.069,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4070080862533692,
+      "grad_norm": 0.5799607038497925,
+      "learning_rate": 0.0005161618996222341,
+      "loss": 3.7092,
+      "step": 13050
+    },
+    {
+      "epoch": 1.412398921832884,
+      "grad_norm": 0.5906285643577576,
+      "learning_rate": 0.0005158381003777657,
+      "loss": 3.702,
+      "step": 13100
+    },
+    {
+      "epoch": 1.417789757412399,
+      "grad_norm": 0.5892825126647949,
+      "learning_rate": 0.0005155143011332973,
+      "loss": 3.7085,
+      "step": 13150
+    },
+    {
+      "epoch": 1.4231805929919137,
+      "grad_norm": 0.5715658068656921,
+      "learning_rate": 0.0005151905018888289,
+      "loss": 3.7085,
+      "step": 13200
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.5430290699005127,
+      "learning_rate": 0.0005148667026443604,
+      "loss": 3.7192,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4339622641509435,
+      "grad_norm": 0.542829692363739,
+      "learning_rate": 0.000514542903399892,
+      "loss": 3.6924,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4393530997304582,
+      "grad_norm": 0.6185770034790039,
+      "learning_rate": 0.0005142191041554237,
+      "loss": 3.7115,
+      "step": 13350
+    },
+    {
+      "epoch": 1.444743935309973,
+      "grad_norm": 0.5668213963508606,
+      "learning_rate": 0.0005138953049109552,
+      "loss": 3.7281,
+      "step": 13400
+    },
+    {
+      "epoch": 1.4501347708894878,
+      "grad_norm": 0.5747724175453186,
+      "learning_rate": 0.0005135715056664868,
+      "loss": 3.7309,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4555256064690028,
+      "grad_norm": 0.6788982152938843,
+      "learning_rate": 0.0005132477064220183,
+      "loss": 3.7209,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4609164420485174,
+      "grad_norm": 0.5687934160232544,
+      "learning_rate": 0.0005129239071775499,
+      "loss": 3.7097,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4663072776280324,
+      "grad_norm": 0.559430718421936,
+      "learning_rate": 0.0005126001079330814,
+      "loss": 3.7277,
+      "step": 13600
+    },
+    {
+      "epoch": 1.4716981132075473,
+      "grad_norm": 0.5699163675308228,
+      "learning_rate": 0.000512276308688613,
+      "loss": 3.7151,
+      "step": 13650
+    },
+    {
+      "epoch": 1.477088948787062,
+      "grad_norm": 0.6224086880683899,
+      "learning_rate": 0.0005119525094441446,
+      "loss": 3.7028,
+      "step": 13700
+    },
+    {
+      "epoch": 1.482479784366577,
+      "grad_norm": 0.5891327261924744,
+      "learning_rate": 0.0005116287101996762,
+      "loss": 3.7059,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4878706199460916,
+      "grad_norm": 0.5879706740379333,
+      "learning_rate": 0.0005113049109552077,
+      "loss": 3.7059,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4932614555256065,
+      "grad_norm": 0.5521015524864197,
+      "learning_rate": 0.0005109811117107393,
+      "loss": 3.7116,
+      "step": 13850
+    },
+    {
+      "epoch": 1.4986522911051212,
+      "grad_norm": 0.6306660175323486,
+      "learning_rate": 0.0005106573124662708,
+      "loss": 3.6918,
+      "step": 13900
+    },
+    {
+      "epoch": 1.5040431266846361,
+      "grad_norm": 0.5480997562408447,
+      "learning_rate": 0.0005103335132218025,
+      "loss": 3.6945,
+      "step": 13950
+    },
+    {
+      "epoch": 1.509433962264151,
+      "grad_norm": 0.6052276492118835,
+      "learning_rate": 0.000510009713977334,
+      "loss": 3.706,
+      "step": 14000
+    },
+    {
+      "epoch": 1.509433962264151,
+      "eval_accuracy": 0.3533992762196531,
+      "eval_loss": 3.659663200378418,
+      "eval_runtime": 185.6467,
+      "eval_samples_per_second": 97.018,
+      "eval_steps_per_second": 6.065,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5148247978436657,
+      "grad_norm": 0.5557168126106262,
+      "learning_rate": 0.0005096859147328656,
+      "loss": 3.7037,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5202156334231804,
+      "grad_norm": 0.604371964931488,
+      "learning_rate": 0.0005093621154883971,
+      "loss": 3.7041,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5256064690026954,
+      "grad_norm": 0.5364857316017151,
+      "learning_rate": 0.0005090383162439288,
+      "loss": 3.6986,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5309973045822103,
+      "grad_norm": 0.5772760510444641,
+      "learning_rate": 0.0005087145169994602,
+      "loss": 3.7034,
+      "step": 14200
+    },
+    {
+      "epoch": 1.536388140161725,
+      "grad_norm": 0.6090425848960876,
+      "learning_rate": 0.0005083907177549918,
+      "loss": 3.7027,
+      "step": 14250
+    },
+    {
+      "epoch": 1.54177897574124,
+      "grad_norm": 0.5599976181983948,
+      "learning_rate": 0.0005080669185105234,
+      "loss": 3.7,
+      "step": 14300
+    },
+    {
+      "epoch": 1.5471698113207548,
+      "grad_norm": 0.520064651966095,
+      "learning_rate": 0.000507743119266055,
+      "loss": 3.6903,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5525606469002695,
+      "grad_norm": 0.5653887391090393,
+      "learning_rate": 0.0005074193200215865,
+      "loss": 3.6936,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5579514824797842,
+      "grad_norm": 0.5902999639511108,
+      "learning_rate": 0.0005070955207771181,
+      "loss": 3.6947,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5633423180592994,
+      "grad_norm": 0.5585593581199646,
+      "learning_rate": 0.0005067717215326498,
+      "loss": 3.6943,
+      "step": 14500
+    },
+    {
+      "epoch": 1.568733153638814,
+      "grad_norm": 0.561992347240448,
+      "learning_rate": 0.0005064479222881813,
+      "loss": 3.6928,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5741239892183287,
+      "grad_norm": 0.614081621170044,
+      "learning_rate": 0.0005061241230437129,
+      "loss": 3.6771,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5795148247978437,
+      "grad_norm": 0.5700899362564087,
+      "learning_rate": 0.0005058003237992444,
+      "loss": 3.6801,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5849056603773586,
+      "grad_norm": 0.589597225189209,
+      "learning_rate": 0.000505476524554776,
+      "loss": 3.6883,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5902964959568733,
+      "grad_norm": 0.5588701367378235,
+      "learning_rate": 0.0005051527253103076,
+      "loss": 3.6682,
+      "step": 14750
+    },
+    {
+      "epoch": 1.595687331536388,
+      "grad_norm": 0.571918785572052,
+      "learning_rate": 0.0005048289260658392,
+      "loss": 3.6607,
+      "step": 14800
+    },
+    {
+      "epoch": 1.6010781671159031,
+      "grad_norm": 0.5928280353546143,
+      "learning_rate": 0.00050451160280626,
+      "loss": 3.6858,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6064690026954178,
+      "grad_norm": 0.6332111954689026,
+      "learning_rate": 0.0005041878035617917,
+      "loss": 3.6894,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6118598382749325,
+      "grad_norm": 0.5789216160774231,
+      "learning_rate": 0.0005038640043173232,
+      "loss": 3.6746,
+      "step": 14950
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "grad_norm": 0.48511213064193726,
+      "learning_rate": 0.0005035402050728548,
+      "loss": 3.6668,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "eval_accuracy": 0.3555571225716487,
+      "eval_loss": 3.637941360473633,
+      "eval_runtime": 185.5425,
+      "eval_samples_per_second": 97.072,
+      "eval_steps_per_second": 6.069,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6226415094339623,
+      "grad_norm": 0.6082944273948669,
+      "learning_rate": 0.0005032164058283863,
+      "loss": 3.6663,
+      "step": 15050
+    },
+    {
+      "epoch": 1.628032345013477,
+      "grad_norm": 0.5705110430717468,
+      "learning_rate": 0.0005028926065839179,
+      "loss": 3.6771,
+      "step": 15100
+    },
+    {
+      "epoch": 1.633423180592992,
+      "grad_norm": 0.5626236200332642,
+      "learning_rate": 0.0005025688073394495,
+      "loss": 3.6959,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6388140161725069,
+      "grad_norm": 0.5518443584442139,
+      "learning_rate": 0.0005022450080949811,
+      "loss": 3.6863,
+      "step": 15200
+    },
+    {
+      "epoch": 1.6442048517520216,
+      "grad_norm": 0.5364320278167725,
+      "learning_rate": 0.0005019212088505126,
+      "loss": 3.6781,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6495956873315363,
+      "grad_norm": 0.5463019609451294,
+      "learning_rate": 0.0005015974096060442,
+      "loss": 3.6738,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6549865229110512,
+      "grad_norm": 0.5183331370353699,
+      "learning_rate": 0.0005012736103615758,
+      "loss": 3.6614,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6603773584905661,
+      "grad_norm": 0.5746192336082458,
+      "learning_rate": 0.0005009498111171074,
+      "loss": 3.6803,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6657681940700808,
+      "grad_norm": 0.6013946533203125,
+      "learning_rate": 0.0005006260118726389,
+      "loss": 3.669,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6711590296495957,
+      "grad_norm": 0.5722841620445251,
+      "learning_rate": 0.0005003022126281705,
+      "loss": 3.6688,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6765498652291106,
+      "grad_norm": 0.5859444737434387,
+      "learning_rate": 0.000499978413383702,
+      "loss": 3.6543,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6819407008086253,
+      "grad_norm": 0.6131587028503418,
+      "learning_rate": 0.0004996546141392336,
+      "loss": 3.6729,
+      "step": 15600
+    },
+    {
+      "epoch": 1.68733153638814,
+      "grad_norm": 0.5786921381950378,
+      "learning_rate": 0.0004993308148947651,
+      "loss": 3.6788,
+      "step": 15650
+    },
+    {
+      "epoch": 1.692722371967655,
+      "grad_norm": 0.6234875321388245,
+      "learning_rate": 0.0004990070156502968,
+      "loss": 3.6754,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6981132075471699,
+      "grad_norm": 0.5718642473220825,
+      "learning_rate": 0.0004986832164058284,
+      "loss": 3.6577,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7035040431266846,
+      "grad_norm": 0.6089641451835632,
+      "learning_rate": 0.0004983594171613599,
+      "loss": 3.6671,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7088948787061995,
+      "grad_norm": 0.5759063959121704,
+      "learning_rate": 0.0004980356179168915,
+      "loss": 3.6859,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.5659204125404358,
+      "learning_rate": 0.000497711818672423,
+      "loss": 3.6828,
+      "step": 15900
+    },
+    {
+      "epoch": 1.719676549865229,
+      "grad_norm": 0.538532018661499,
+      "learning_rate": 0.0004973880194279547,
+      "loss": 3.6715,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "grad_norm": 0.6466843485832214,
+      "learning_rate": 0.0004970642201834862,
+      "loss": 3.6663,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "eval_accuracy": 0.3573782536243072,
+      "eval_loss": 3.622516393661499,
+      "eval_runtime": 185.5467,
+      "eval_samples_per_second": 97.07,
+      "eval_steps_per_second": 6.069,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7304582210242587,
+      "grad_norm": 0.5568268299102783,
+      "learning_rate": 0.0004967404209390178,
+      "loss": 3.6554,
+      "step": 16050
+    },
+    {
+      "epoch": 1.7358490566037736,
+      "grad_norm": 0.5877635478973389,
+      "learning_rate": 0.0004964166216945493,
+      "loss": 3.6559,
+      "step": 16100
+    },
+    {
+      "epoch": 1.7412398921832883,
+      "grad_norm": 0.6343328952789307,
+      "learning_rate": 0.000496092822450081,
+      "loss": 3.6665,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7466307277628033,
+      "grad_norm": 0.6591050624847412,
+      "learning_rate": 0.0004957690232056125,
+      "loss": 3.6714,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7520215633423182,
+      "grad_norm": 0.6299484372138977,
+      "learning_rate": 0.0004954452239611441,
+      "loss": 3.6553,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7574123989218329,
+      "grad_norm": 0.5382270216941833,
+      "learning_rate": 0.0004951214247166756,
+      "loss": 3.6475,
+      "step": 16300
+    },
+    {
+      "epoch": 1.7628032345013476,
+      "grad_norm": 0.5545465350151062,
+      "learning_rate": 0.0004947976254722072,
+      "loss": 3.6716,
+      "step": 16350
+    },
+    {
+      "epoch": 1.7681940700808625,
+      "grad_norm": 0.5523202419281006,
+      "learning_rate": 0.0004944738262277387,
+      "loss": 3.6494,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7735849056603774,
+      "grad_norm": 0.5391619801521301,
+      "learning_rate": 0.0004941500269832703,
+      "loss": 3.6589,
+      "step": 16450
+    },
+    {
+      "epoch": 1.778975741239892,
+      "grad_norm": 0.5421316623687744,
+      "learning_rate": 0.0004938262277388019,
+      "loss": 3.646,
+      "step": 16500
+    },
+    {
+      "epoch": 1.784366576819407,
+      "grad_norm": 0.5299726724624634,
+      "learning_rate": 0.0004935024284943335,
+      "loss": 3.6525,
+      "step": 16550
+    },
+    {
+      "epoch": 1.789757412398922,
+      "grad_norm": 0.5056886076927185,
+      "learning_rate": 0.000493178629249865,
+      "loss": 3.6657,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7951482479784366,
+      "grad_norm": 0.5270845293998718,
+      "learning_rate": 0.0004928548300053966,
+      "loss": 3.6557,
+      "step": 16650
+    },
+    {
+      "epoch": 1.8005390835579513,
+      "grad_norm": 0.6346863508224487,
+      "learning_rate": 0.0004925310307609282,
+      "loss": 3.6629,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8059299191374663,
+      "grad_norm": 0.5599194169044495,
+      "learning_rate": 0.0004922072315164598,
+      "loss": 3.6619,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8113207547169812,
+      "grad_norm": 0.6051496267318726,
+      "learning_rate": 0.0004918834322719913,
+      "loss": 3.6696,
+      "step": 16800
+    },
+    {
+      "epoch": 1.8167115902964959,
+      "grad_norm": 0.5915637612342834,
+      "learning_rate": 0.0004915596330275229,
+      "loss": 3.6765,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8221024258760108,
+      "grad_norm": 0.5501888990402222,
+      "learning_rate": 0.0004912423097679439,
+      "loss": 3.6496,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8274932614555257,
+      "grad_norm": 0.5859174132347107,
+      "learning_rate": 0.0004909185105234754,
+      "loss": 3.6598,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "grad_norm": 0.5620100498199463,
+      "learning_rate": 0.000490594711279007,
+      "loss": 3.6412,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "eval_accuracy": 0.35890319690478345,
+      "eval_loss": 3.6049413681030273,
+      "eval_runtime": 185.9741,
+      "eval_samples_per_second": 96.847,
+      "eval_steps_per_second": 6.055,
+      "step": 17000
+    },
+    {
+      "epoch": 1.838274932614555,
+      "grad_norm": 0.5494559407234192,
+      "learning_rate": 0.0004902709120345385,
+      "loss": 3.6457,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8436657681940702,
+      "grad_norm": 0.5877504944801331,
+      "learning_rate": 0.00048994711279007,
+      "loss": 3.6385,
+      "step": 17100
+    },
+    {
+      "epoch": 1.849056603773585,
+      "grad_norm": 0.5871086120605469,
+      "learning_rate": 0.0004896233135456017,
+      "loss": 3.6429,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8544474393530996,
+      "grad_norm": 0.555230975151062,
+      "learning_rate": 0.0004892995143011333,
+      "loss": 3.649,
+      "step": 17200
+    },
+    {
+      "epoch": 1.8598382749326146,
+      "grad_norm": 0.5545013546943665,
+      "learning_rate": 0.0004889757150566648,
+      "loss": 3.6561,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8652291105121295,
+      "grad_norm": 0.6687287092208862,
+      "learning_rate": 0.0004886519158121964,
+      "loss": 3.6577,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8706199460916442,
+      "grad_norm": 0.5878035426139832,
+      "learning_rate": 0.000488328116567728,
+      "loss": 3.6603,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8760107816711589,
+      "grad_norm": 0.5427060127258301,
+      "learning_rate": 0.0004880043173232595,
+      "loss": 3.6529,
+      "step": 17400
+    },
+    {
+      "epoch": 1.881401617250674,
+      "grad_norm": 0.5963491201400757,
+      "learning_rate": 0.0004876805180787911,
+      "loss": 3.6421,
+      "step": 17450
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.5591420531272888,
+      "learning_rate": 0.0004873567188343227,
+      "loss": 3.6463,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8921832884097034,
+      "grad_norm": 0.6233649849891663,
+      "learning_rate": 0.0004870329195898542,
+      "loss": 3.6403,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8975741239892183,
+      "grad_norm": 0.5761533975601196,
+      "learning_rate": 0.00048670912034538583,
+      "loss": 3.655,
+      "step": 17600
+    },
+    {
+      "epoch": 1.9029649595687332,
+      "grad_norm": 0.6134516596794128,
+      "learning_rate": 0.0004863853211009174,
+      "loss": 3.647,
+      "step": 17650
+    },
+    {
+      "epoch": 1.908355795148248,
+      "grad_norm": 0.5410062074661255,
+      "learning_rate": 0.000486061521856449,
+      "loss": 3.642,
+      "step": 17700
+    },
+    {
+      "epoch": 1.9137466307277629,
+      "grad_norm": 0.6168045401573181,
+      "learning_rate": 0.00048573772261198054,
+      "loss": 3.6608,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9191374663072778,
+      "grad_norm": 0.566486656665802,
+      "learning_rate": 0.00048541392336751214,
+      "loss": 3.6455,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9245283018867925,
+      "grad_norm": 0.5486074090003967,
+      "learning_rate": 0.0004850901241230437,
+      "loss": 3.6385,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9299191374663072,
+      "grad_norm": 0.5504534244537354,
+      "learning_rate": 0.0004847663248785753,
+      "loss": 3.6489,
+      "step": 17900
+    },
+    {
+      "epoch": 1.935309973045822,
+      "grad_norm": 0.5466365814208984,
+      "learning_rate": 0.0004844425256341068,
+      "loss": 3.629,
+      "step": 17950
+    },
+    {
+      "epoch": 1.940700808625337,
+      "grad_norm": 0.5768356919288635,
+      "learning_rate": 0.00048411872638963834,
+      "loss": 3.634,
+      "step": 18000
+    },
+    {
+      "epoch": 1.940700808625337,
+      "eval_accuracy": 0.35977426710639215,
+      "eval_loss": 3.5927350521087646,
+      "eval_runtime": 185.3253,
+      "eval_samples_per_second": 97.186,
+      "eval_steps_per_second": 6.076,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9460916442048517,
+      "grad_norm": 0.6032684445381165,
+      "learning_rate": 0.00048379492714516995,
+      "loss": 3.6367,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9514824797843666,
+      "grad_norm": 0.5920586585998535,
+      "learning_rate": 0.0004834711279007015,
+      "loss": 3.6372,
+      "step": 18100
+    },
+    {
+      "epoch": 1.9568733153638815,
+      "grad_norm": 0.5113473534584045,
+      "learning_rate": 0.0004831473286562331,
+      "loss": 3.6331,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9622641509433962,
+      "grad_norm": 0.592510461807251,
+      "learning_rate": 0.00048282352941176465,
+      "loss": 3.6261,
+      "step": 18200
+    },
+    {
+      "epoch": 1.967654986522911,
+      "grad_norm": 0.6228882670402527,
+      "learning_rate": 0.00048249973016729626,
+      "loss": 3.6438,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9730458221024259,
+      "grad_norm": 0.6540427803993225,
+      "learning_rate": 0.0004821759309228278,
+      "loss": 3.6536,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9784366576819408,
+      "grad_norm": 0.563521683216095,
+      "learning_rate": 0.00048185213167835936,
+      "loss": 3.6395,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9838274932614555,
+      "grad_norm": 0.633597731590271,
+      "learning_rate": 0.00048152833243389096,
+      "loss": 3.6384,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9892183288409704,
+      "grad_norm": 0.5636687874794006,
+      "learning_rate": 0.0004812045331894225,
+      "loss": 3.6465,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9946091644204853,
+      "grad_norm": 0.6388487815856934,
+      "learning_rate": 0.0004808807339449541,
+      "loss": 3.6375,
+      "step": 18500
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.1785130500793457,
+      "learning_rate": 0.0004805569347004856,
+      "loss": 3.6438,
+      "step": 18550
+    },
+    {
+      "epoch": 2.0053908355795147,
+      "grad_norm": 0.5871391892433167,
+      "learning_rate": 0.0004802331354560173,
+      "loss": 3.5742,
+      "step": 18600
+    },
+    {
+      "epoch": 2.01078167115903,
+      "grad_norm": 0.6168270707130432,
+      "learning_rate": 0.00047990933621154877,
+      "loss": 3.543,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0161725067385445,
+      "grad_norm": 0.549731969833374,
+      "learning_rate": 0.0004795855369670804,
+      "loss": 3.5557,
+      "step": 18700
+    },
+    {
+      "epoch": 2.0215633423180592,
+      "grad_norm": 0.5464780330657959,
+      "learning_rate": 0.0004792617377226119,
+      "loss": 3.5353,
+      "step": 18750
+    },
+    {
+      "epoch": 2.026954177897574,
+      "grad_norm": 0.5794655680656433,
+      "learning_rate": 0.0004789379384781435,
+      "loss": 3.5544,
+      "step": 18800
+    },
+    {
+      "epoch": 2.032345013477089,
+      "grad_norm": 0.558387279510498,
+      "learning_rate": 0.0004786141392336751,
+      "loss": 3.5604,
+      "step": 18850
+    },
+    {
+      "epoch": 2.0377358490566038,
+      "grad_norm": 0.5876425504684448,
+      "learning_rate": 0.00047829033998920663,
+      "loss": 3.5486,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0431266846361185,
+      "grad_norm": 0.5909355282783508,
+      "learning_rate": 0.00047796654074473824,
+      "loss": 3.5405,
+      "step": 18950
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "grad_norm": 0.5822170376777649,
+      "learning_rate": 0.0004776492174851592,
+      "loss": 3.57,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "eval_accuracy": 0.3617789128865693,
+      "eval_loss": 3.581454038619995,
+      "eval_runtime": 185.6663,
+      "eval_samples_per_second": 97.007,
+      "eval_steps_per_second": 6.065,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0539083557951483,
+      "grad_norm": 0.5871404409408569,
+      "learning_rate": 0.00047732541824069074,
+      "loss": 3.5472,
+      "step": 19050
+    },
+    {
+      "epoch": 2.059299191374663,
+      "grad_norm": 0.6160888075828552,
+      "learning_rate": 0.0004770016189962223,
+      "loss": 3.569,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0646900269541777,
+      "grad_norm": 0.5916035771369934,
+      "learning_rate": 0.0004766778197517539,
+      "loss": 3.565,
+      "step": 19150
+    },
+    {
+      "epoch": 2.070080862533693,
+      "grad_norm": 0.6275314092636108,
+      "learning_rate": 0.0004763540205072854,
+      "loss": 3.5718,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0754716981132075,
+      "grad_norm": 0.554090142250061,
+      "learning_rate": 0.00047603022126281705,
+      "loss": 3.5529,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0808625336927222,
+      "grad_norm": 0.559099018573761,
+      "learning_rate": 0.00047570642201834855,
+      "loss": 3.5473,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0862533692722374,
+      "grad_norm": 0.5705540180206299,
+      "learning_rate": 0.00047538262277388015,
+      "loss": 3.5502,
+      "step": 19350
+    },
+    {
+      "epoch": 2.091644204851752,
+      "grad_norm": 0.5766929984092712,
+      "learning_rate": 0.0004750588235294117,
+      "loss": 3.5452,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0970350404312668,
+      "grad_norm": 0.6509124636650085,
+      "learning_rate": 0.0004747350242849433,
+      "loss": 3.5553,
+      "step": 19450
+    },
+    {
+      "epoch": 2.1024258760107815,
+      "grad_norm": 0.6113699674606323,
+      "learning_rate": 0.00047441122504047486,
+      "loss": 3.5484,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1078167115902966,
+      "grad_norm": 0.5691044926643372,
+      "learning_rate": 0.0004740874257960064,
+      "loss": 3.5544,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1132075471698113,
+      "grad_norm": 0.5695565342903137,
+      "learning_rate": 0.000473763626551538,
+      "loss": 3.5503,
+      "step": 19600
+    },
+    {
+      "epoch": 2.118598382749326,
+      "grad_norm": 0.5740593075752258,
+      "learning_rate": 0.00047343982730706956,
+      "loss": 3.5738,
+      "step": 19650
+    },
+    {
+      "epoch": 2.123989218328841,
+      "grad_norm": 0.5763875246047974,
+      "learning_rate": 0.00047311602806260117,
+      "loss": 3.5446,
+      "step": 19700
+    },
+    {
+      "epoch": 2.129380053908356,
+      "grad_norm": 0.6153304576873779,
+      "learning_rate": 0.0004727922288181327,
+      "loss": 3.5653,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1347708894878705,
+      "grad_norm": 0.5829182267189026,
+      "learning_rate": 0.0004724684295736643,
+      "loss": 3.5581,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1401617250673857,
+      "grad_norm": 0.5911663174629211,
+      "learning_rate": 0.0004721446303291959,
+      "loss": 3.5505,
+      "step": 19850
+    },
+    {
+      "epoch": 2.1455525606469004,
+      "grad_norm": 0.6006436944007874,
+      "learning_rate": 0.0004718208310847275,
+      "loss": 3.5604,
+      "step": 19900
+    },
+    {
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.5507996082305908,
+      "learning_rate": 0.000471497031840259,
+      "loss": 3.5543,
+      "step": 19950
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "grad_norm": 0.6772857904434204,
+      "learning_rate": 0.0004711732325957905,
+      "loss": 3.5772,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "eval_accuracy": 0.3622711104683689,
+      "eval_loss": 3.5757319927215576,
+      "eval_runtime": 185.6283,
+      "eval_samples_per_second": 97.027,
+      "eval_steps_per_second": 6.066,
+      "step": 20000
+    },
+    {
+      "epoch": 2.161725067385445,
+      "grad_norm": 0.580953061580658,
+      "learning_rate": 0.00047084943335132213,
+      "loss": 3.5802,
+      "step": 20050
+    },
+    {
+      "epoch": 2.1671159029649596,
+      "grad_norm": 0.5873518586158752,
+      "learning_rate": 0.0004705256341068537,
+      "loss": 3.5784,
+      "step": 20100
+    },
+    {
+      "epoch": 2.1725067385444743,
+      "grad_norm": 0.5810049176216125,
+      "learning_rate": 0.0004702018348623853,
+      "loss": 3.5728,
+      "step": 20150
+    },
+    {
+      "epoch": 2.177897574123989,
+      "grad_norm": 0.5831699371337891,
+      "learning_rate": 0.00046987803561791684,
+      "loss": 3.5693,
+      "step": 20200
+    },
+    {
+      "epoch": 2.183288409703504,
+      "grad_norm": 0.5833231806755066,
+      "learning_rate": 0.00046955423637344844,
+      "loss": 3.5708,
+      "step": 20250
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 0.5573728680610657,
+      "learning_rate": 0.00046923043712898,
+      "loss": 3.5622,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1940700808625335,
+      "grad_norm": 0.5548168420791626,
+      "learning_rate": 0.0004689066378845116,
+      "loss": 3.5761,
+      "step": 20350
+    },
+    {
+      "epoch": 2.1994609164420487,
+      "grad_norm": 0.5806989669799805,
+      "learning_rate": 0.00046858283864004315,
+      "loss": 3.5735,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2048517520215634,
+      "grad_norm": 0.5829126238822937,
+      "learning_rate": 0.0004682590393955747,
+      "loss": 3.5728,
+      "step": 20450
+    },
+    {
+      "epoch": 2.210242587601078,
+      "grad_norm": 0.6163429617881775,
+      "learning_rate": 0.0004679352401511063,
+      "loss": 3.5436,
+      "step": 20500
+    },
+    {
+      "epoch": 2.215633423180593,
+      "grad_norm": 0.5611751079559326,
+      "learning_rate": 0.0004676114409066378,
+      "loss": 3.5731,
+      "step": 20550
+    },
+    {
+      "epoch": 2.221024258760108,
+      "grad_norm": 0.68216872215271,
+      "learning_rate": 0.00046728764166216946,
+      "loss": 3.5495,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2264150943396226,
+      "grad_norm": 0.5706924200057983,
+      "learning_rate": 0.00046696384241770095,
+      "loss": 3.5558,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2318059299191373,
+      "grad_norm": 0.5555259585380554,
+      "learning_rate": 0.00046664004317323256,
+      "loss": 3.5539,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2371967654986524,
+      "grad_norm": 0.5713426470756531,
+      "learning_rate": 0.0004663162439287641,
+      "loss": 3.5596,
+      "step": 20750
+    },
+    {
+      "epoch": 2.242587601078167,
+      "grad_norm": 0.5891562700271606,
+      "learning_rate": 0.00046599244468429566,
+      "loss": 3.5542,
+      "step": 20800
+    },
+    {
+      "epoch": 2.247978436657682,
+      "grad_norm": 0.5680431723594666,
+      "learning_rate": 0.00046566864543982726,
+      "loss": 3.5711,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2533692722371965,
+      "grad_norm": 0.5731450915336609,
+      "learning_rate": 0.0004653448461953588,
+      "loss": 3.5651,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2587601078167117,
+      "grad_norm": 0.562228798866272,
+      "learning_rate": 0.0004650210469508904,
+      "loss": 3.5531,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "grad_norm": 0.5744984745979309,
+      "learning_rate": 0.00046469724770642197,
+      "loss": 3.5661,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "eval_accuracy": 0.36361503803774625,
+      "eval_loss": 3.562875270843506,
+      "eval_runtime": 185.65,
+      "eval_samples_per_second": 97.016,
+      "eval_steps_per_second": 6.065,
+      "step": 21000
+    },
+    {
+      "epoch": 2.269541778975741,
+      "grad_norm": 0.5437514185905457,
+      "learning_rate": 0.0004643799244468429,
+      "loss": 3.5609,
+      "step": 21050
+    },
+    {
+      "epoch": 2.274932614555256,
+      "grad_norm": 0.6143810153007507,
+      "learning_rate": 0.0004640561252023745,
+      "loss": 3.556,
+      "step": 21100
+    },
+    {
+      "epoch": 2.280323450134771,
+      "grad_norm": 0.6042497754096985,
+      "learning_rate": 0.0004637323259579061,
+      "loss": 3.5793,
+      "step": 21150
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.6476418375968933,
+      "learning_rate": 0.0004634085267134376,
+      "loss": 3.5645,
+      "step": 21200
+    },
+    {
+      "epoch": 2.2911051212938007,
+      "grad_norm": 0.6019425988197327,
+      "learning_rate": 0.00046308472746896923,
+      "loss": 3.5748,
+      "step": 21250
+    },
+    {
+      "epoch": 2.2964959568733154,
+      "grad_norm": 0.6152219772338867,
+      "learning_rate": 0.00046276092822450073,
+      "loss": 3.5401,
+      "step": 21300
+    },
+    {
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.6449258923530579,
+      "learning_rate": 0.00046243712898003233,
+      "loss": 3.5596,
+      "step": 21350
+    },
+    {
+      "epoch": 2.3072776280323453,
+      "grad_norm": 0.5655533075332642,
+      "learning_rate": 0.0004621133297355639,
+      "loss": 3.5532,
+      "step": 21400
+    },
+    {
+      "epoch": 2.31266846361186,
+      "grad_norm": 0.5761871933937073,
+      "learning_rate": 0.0004617895304910955,
+      "loss": 3.5781,
+      "step": 21450
+    },
+    {
+      "epoch": 2.3180592991913747,
+      "grad_norm": 0.5710723996162415,
+      "learning_rate": 0.00046146573124662704,
+      "loss": 3.549,
+      "step": 21500
+    },
+    {
+      "epoch": 2.3234501347708894,
+      "grad_norm": 0.6261053681373596,
+      "learning_rate": 0.00046114193200215864,
+      "loss": 3.5644,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3288409703504045,
+      "grad_norm": 0.6094996929168701,
+      "learning_rate": 0.0004608181327576902,
+      "loss": 3.556,
+      "step": 21600
+    },
+    {
+      "epoch": 2.334231805929919,
+      "grad_norm": 0.6269069910049438,
+      "learning_rate": 0.00046049433351322175,
+      "loss": 3.557,
+      "step": 21650
+    },
+    {
+      "epoch": 2.339622641509434,
+      "grad_norm": 0.6441256999969482,
+      "learning_rate": 0.00046017053426875335,
+      "loss": 3.5743,
+      "step": 21700
+    },
+    {
+      "epoch": 2.3450134770889486,
+      "grad_norm": 0.5836607217788696,
+      "learning_rate": 0.0004598467350242849,
+      "loss": 3.565,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3504043126684637,
+      "grad_norm": 0.6284753084182739,
+      "learning_rate": 0.0004595229357798165,
+      "loss": 3.5467,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3557951482479784,
+      "grad_norm": 0.5670480728149414,
+      "learning_rate": 0.00045919913653534806,
+      "loss": 3.5584,
+      "step": 21850
+    },
+    {
+      "epoch": 2.361185983827493,
+      "grad_norm": 0.6184543371200562,
+      "learning_rate": 0.00045887533729087966,
+      "loss": 3.5516,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3665768194070083,
+      "grad_norm": 0.5675870776176453,
+      "learning_rate": 0.00045855153804641116,
+      "loss": 3.5536,
+      "step": 21950
+    },
+    {
+      "epoch": 2.371967654986523,
+      "grad_norm": 0.5641400218009949,
+      "learning_rate": 0.0004582277388019427,
+      "loss": 3.5726,
+      "step": 22000
+    },
+    {
+      "epoch": 2.371967654986523,
+      "eval_accuracy": 0.3648121755556862,
+      "eval_loss": 3.5518908500671387,
+      "eval_runtime": 185.7462,
+      "eval_samples_per_second": 96.966,
+      "eval_steps_per_second": 6.062,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 0.5567072033882141,
+      "learning_rate": 0.0004579039395574743,
+      "loss": 3.5502,
+      "step": 22050
+    },
+    {
+      "epoch": 2.382749326145553,
+      "grad_norm": 0.6511866450309753,
+      "learning_rate": 0.00045758014031300586,
+      "loss": 3.5514,
+      "step": 22100
+    },
+    {
+      "epoch": 2.3881401617250675,
+      "grad_norm": 0.5884348750114441,
+      "learning_rate": 0.00045725634106853747,
+      "loss": 3.5552,
+      "step": 22150
+    },
+    {
+      "epoch": 2.393530997304582,
+      "grad_norm": 0.5839561223983765,
+      "learning_rate": 0.000456932541824069,
+      "loss": 3.5647,
+      "step": 22200
+    },
+    {
+      "epoch": 2.398921832884097,
+      "grad_norm": 0.5600352883338928,
+      "learning_rate": 0.0004566087425796006,
+      "loss": 3.5702,
+      "step": 22250
+    },
+    {
+      "epoch": 2.404312668463612,
+      "grad_norm": 0.5991252660751343,
+      "learning_rate": 0.0004562849433351322,
+      "loss": 3.5718,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4097035040431267,
+      "grad_norm": 0.60001140832901,
+      "learning_rate": 0.0004559611440906638,
+      "loss": 3.5526,
+      "step": 22350
+    },
+    {
+      "epoch": 2.4150943396226414,
+      "grad_norm": 0.5612289309501648,
+      "learning_rate": 0.00045563734484619533,
+      "loss": 3.5746,
+      "step": 22400
+    },
+    {
+      "epoch": 2.420485175202156,
+      "grad_norm": 0.5874634981155396,
+      "learning_rate": 0.0004553135456017269,
+      "loss": 3.5655,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4258760107816713,
+      "grad_norm": 0.547502338886261,
+      "learning_rate": 0.0004549897463572585,
+      "loss": 3.5597,
+      "step": 22500
+    },
+    {
+      "epoch": 2.431266846361186,
+      "grad_norm": 0.6048047542572021,
+      "learning_rate": 0.00045466594711279,
+      "loss": 3.554,
+      "step": 22550
+    },
+    {
+      "epoch": 2.4366576819407006,
+      "grad_norm": 0.5858154892921448,
+      "learning_rate": 0.00045434214786832164,
+      "loss": 3.5543,
+      "step": 22600
+    },
+    {
+      "epoch": 2.442048517520216,
+      "grad_norm": 0.5783501863479614,
+      "learning_rate": 0.00045401834862385314,
+      "loss": 3.5413,
+      "step": 22650
+    },
+    {
+      "epoch": 2.4474393530997305,
+      "grad_norm": 0.603816032409668,
+      "learning_rate": 0.00045369454937938474,
+      "loss": 3.5451,
+      "step": 22700
+    },
+    {
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.5959175825119019,
+      "learning_rate": 0.0004533707501349163,
+      "loss": 3.5488,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4582210242587603,
+      "grad_norm": 0.5687942504882812,
+      "learning_rate": 0.0004530469508904479,
+      "loss": 3.5459,
+      "step": 22800
+    },
+    {
+      "epoch": 2.463611859838275,
+      "grad_norm": 0.60494464635849,
+      "learning_rate": 0.00045272315164597945,
+      "loss": 3.5532,
+      "step": 22850
+    },
+    {
+      "epoch": 2.4690026954177897,
+      "grad_norm": 0.6027876138687134,
+      "learning_rate": 0.000452399352401511,
+      "loss": 3.5631,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4743935309973044,
+      "grad_norm": 0.6948409080505371,
+      "learning_rate": 0.0004520755531570426,
+      "loss": 3.565,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "grad_norm": 0.5531662106513977,
+      "learning_rate": 0.00045175175391257415,
+      "loss": 3.5424,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "eval_accuracy": 0.365721817532469,
+      "eval_loss": 3.540759325027466,
+      "eval_runtime": 185.6808,
+      "eval_samples_per_second": 97.0,
+      "eval_steps_per_second": 6.064,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4851752021563343,
+      "grad_norm": 0.5789402723312378,
+      "learning_rate": 0.0004514344306529951,
+      "loss": 3.5453,
+      "step": 23050
+    },
+    {
+      "epoch": 2.490566037735849,
+      "grad_norm": 0.6050681471824646,
+      "learning_rate": 0.0004511106314085267,
+      "loss": 3.5601,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4959568733153636,
+      "grad_norm": 0.574383020401001,
+      "learning_rate": 0.00045078683216405826,
+      "loss": 3.5615,
+      "step": 23150
+    },
+    {
+      "epoch": 2.501347708894879,
+      "grad_norm": 0.6034984588623047,
+      "learning_rate": 0.00045046303291958976,
+      "loss": 3.5536,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5067385444743935,
+      "grad_norm": 0.5547395944595337,
+      "learning_rate": 0.0004501392336751214,
+      "loss": 3.5473,
+      "step": 23250
+    },
+    {
+      "epoch": 2.512129380053908,
+      "grad_norm": 0.6038048267364502,
+      "learning_rate": 0.0004498154344306529,
+      "loss": 3.5472,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5175202156334233,
+      "grad_norm": 0.6408278942108154,
+      "learning_rate": 0.0004494916351861845,
+      "loss": 3.5442,
+      "step": 23350
+    },
+    {
+      "epoch": 2.522911051212938,
+      "grad_norm": 0.5872176885604858,
+      "learning_rate": 0.00044916783594171607,
+      "loss": 3.5355,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5283018867924527,
+      "grad_norm": 0.5819481015205383,
+      "learning_rate": 0.00044884403669724767,
+      "loss": 3.5484,
+      "step": 23450
+    },
+    {
+      "epoch": 2.533692722371968,
+      "grad_norm": 0.5719451904296875,
+      "learning_rate": 0.0004485202374527792,
+      "loss": 3.5431,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5390835579514826,
+      "grad_norm": 0.5324487090110779,
+      "learning_rate": 0.00044819643820831083,
+      "loss": 3.5532,
+      "step": 23550
+    },
+    {
+      "epoch": 2.5444743935309972,
+      "grad_norm": 0.5669713616371155,
+      "learning_rate": 0.0004478726389638424,
+      "loss": 3.5532,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5498652291105124,
+      "grad_norm": 0.5789257287979126,
+      "learning_rate": 0.00044754883971937393,
+      "loss": 3.5579,
+      "step": 23650
+    },
+    {
+      "epoch": 2.555256064690027,
+      "grad_norm": 0.5890941023826599,
+      "learning_rate": 0.00044722504047490553,
+      "loss": 3.5665,
+      "step": 23700
+    },
+    {
+      "epoch": 2.560646900269542,
+      "grad_norm": 0.5482243895530701,
+      "learning_rate": 0.0004469012412304371,
+      "loss": 3.5714,
+      "step": 23750
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 0.6566593050956726,
+      "learning_rate": 0.0004465774419859687,
+      "loss": 3.547,
+      "step": 23800
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.6310127973556519,
+      "learning_rate": 0.00044625364274150024,
+      "loss": 3.5627,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5768194070080863,
+      "grad_norm": 0.6267402768135071,
+      "learning_rate": 0.00044592984349703184,
+      "loss": 3.5446,
+      "step": 23900
+    },
+    {
+      "epoch": 2.582210242587601,
+      "grad_norm": 0.5680336952209473,
+      "learning_rate": 0.00044560604425256334,
+      "loss": 3.551,
+      "step": 23950
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "grad_norm": 0.595809280872345,
+      "learning_rate": 0.000445282245008095,
+      "loss": 3.5334,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "eval_accuracy": 0.3668765804241613,
+      "eval_loss": 3.5324785709381104,
+      "eval_runtime": 185.7652,
+      "eval_samples_per_second": 96.956,
+      "eval_steps_per_second": 6.061,
+      "step": 24000
+    },
+    {
+      "epoch": 2.592991913746631,
+      "grad_norm": 0.5670157074928284,
+      "learning_rate": 0.0004449584457636265,
+      "loss": 3.5466,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5983827493261455,
+      "grad_norm": 0.6169844269752502,
+      "learning_rate": 0.00044463464651915805,
+      "loss": 3.5402,
+      "step": 24100
+    },
+    {
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.556118905544281,
+      "learning_rate": 0.00044431084727468965,
+      "loss": 3.5667,
+      "step": 24150
+    },
+    {
+      "epoch": 2.6091644204851754,
+      "grad_norm": 0.612477719783783,
+      "learning_rate": 0.0004439870480302212,
+      "loss": 3.5321,
+      "step": 24200
+    },
+    {
+      "epoch": 2.61455525606469,
+      "grad_norm": 0.5984819531440735,
+      "learning_rate": 0.0004436632487857528,
+      "loss": 3.5418,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6199460916442048,
+      "grad_norm": 0.5921148061752319,
+      "learning_rate": 0.00044333944954128436,
+      "loss": 3.5518,
+      "step": 24300
+    },
+    {
+      "epoch": 2.62533692722372,
+      "grad_norm": 0.5681852102279663,
+      "learning_rate": 0.00044301565029681596,
+      "loss": 3.5658,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6307277628032346,
+      "grad_norm": 0.6255022883415222,
+      "learning_rate": 0.0004426918510523475,
+      "loss": 3.5286,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6361185983827493,
+      "grad_norm": 0.572178840637207,
+      "learning_rate": 0.0004423680518078791,
+      "loss": 3.5473,
+      "step": 24450
+    },
+    {
+      "epoch": 2.641509433962264,
+      "grad_norm": 0.5545793175697327,
+      "learning_rate": 0.00044204425256341067,
+      "loss": 3.5608,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6469002695417787,
+      "grad_norm": 0.6044329404830933,
+      "learning_rate": 0.0004417204533189422,
+      "loss": 3.53,
+      "step": 24550
+    },
+    {
+      "epoch": 2.652291105121294,
+      "grad_norm": 0.6175848841667175,
+      "learning_rate": 0.0004413966540744738,
+      "loss": 3.5345,
+      "step": 24600
+    },
+    {
+      "epoch": 2.6576819407008085,
+      "grad_norm": 0.5467705130577087,
+      "learning_rate": 0.0004410728548300053,
+      "loss": 3.5373,
+      "step": 24650
+    },
+    {
+      "epoch": 2.6630727762803232,
+      "grad_norm": 0.6676491498947144,
+      "learning_rate": 0.0004407490555855369,
+      "loss": 3.5415,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6684636118598384,
+      "grad_norm": 0.6156054139137268,
+      "learning_rate": 0.0004404252563410685,
+      "loss": 3.5401,
+      "step": 24750
+    },
+    {
+      "epoch": 2.673854447439353,
+      "grad_norm": 0.573746919631958,
+      "learning_rate": 0.0004401014570966001,
+      "loss": 3.5181,
+      "step": 24800
+    },
+    {
+      "epoch": 2.6792452830188678,
+      "grad_norm": 0.5712265372276306,
+      "learning_rate": 0.00043977765785213163,
+      "loss": 3.5158,
+      "step": 24850
+    },
+    {
+      "epoch": 2.684636118598383,
+      "grad_norm": 0.574403703212738,
+      "learning_rate": 0.0004394538586076632,
+      "loss": 3.5546,
+      "step": 24900
+    },
+    {
+      "epoch": 2.6900269541778976,
+      "grad_norm": 0.5693142414093018,
+      "learning_rate": 0.0004391300593631948,
+      "loss": 3.5382,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "grad_norm": 0.6114800572395325,
+      "learning_rate": 0.00043880626011872634,
+      "loss": 3.5294,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "eval_accuracy": 0.36749405478561104,
+      "eval_loss": 3.5217669010162354,
+      "eval_runtime": 185.6142,
+      "eval_samples_per_second": 97.035,
+      "eval_steps_per_second": 6.066,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7008086253369274,
+      "grad_norm": 0.552423894405365,
+      "learning_rate": 0.00043848246087425794,
+      "loss": 3.5387,
+      "step": 25050
+    },
+    {
+      "epoch": 2.706199460916442,
+      "grad_norm": 0.5499989986419678,
+      "learning_rate": 0.0004381586616297895,
+      "loss": 3.5253,
+      "step": 25100
+    },
+    {
+      "epoch": 2.711590296495957,
+      "grad_norm": 0.5870550274848938,
+      "learning_rate": 0.00043784133837021044,
+      "loss": 3.5156,
+      "step": 25150
+    },
+    {
+      "epoch": 2.7169811320754715,
+      "grad_norm": 0.5875838994979858,
+      "learning_rate": 0.00043751753912574205,
+      "loss": 3.5225,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7223719676549867,
+      "grad_norm": 0.6579532027244568,
+      "learning_rate": 0.0004371937398812736,
+      "loss": 3.5564,
+      "step": 25250
+    },
+    {
+      "epoch": 2.7277628032345014,
+      "grad_norm": 0.5777049660682678,
+      "learning_rate": 0.0004368699406368051,
+      "loss": 3.5132,
+      "step": 25300
+    },
+    {
+      "epoch": 2.733153638814016,
+      "grad_norm": 0.6111543774604797,
+      "learning_rate": 0.0004365461413923367,
+      "loss": 3.5361,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7385444743935308,
+      "grad_norm": 0.6190521121025085,
+      "learning_rate": 0.00043622234214786825,
+      "loss": 3.5407,
+      "step": 25400
+    },
+    {
+      "epoch": 2.743935309973046,
+      "grad_norm": 0.5720779895782471,
+      "learning_rate": 0.00043589854290339985,
+      "loss": 3.5111,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7493261455525606,
+      "grad_norm": 0.5570679306983948,
+      "learning_rate": 0.0004355747436589314,
+      "loss": 3.571,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.548724889755249,
+      "learning_rate": 0.000435250944414463,
+      "loss": 3.5611,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7601078167115904,
+      "grad_norm": 0.6523057222366333,
+      "learning_rate": 0.00043492714516999456,
+      "loss": 3.5365,
+      "step": 25600
+    },
+    {
+      "epoch": 2.765498652291105,
+      "grad_norm": 0.6245532035827637,
+      "learning_rate": 0.0004346033459255261,
+      "loss": 3.5264,
+      "step": 25650
+    },
+    {
+      "epoch": 2.77088948787062,
+      "grad_norm": 0.58823162317276,
+      "learning_rate": 0.0004342795466810577,
+      "loss": 3.5446,
+      "step": 25700
+    },
+    {
+      "epoch": 2.776280323450135,
+      "grad_norm": 0.6146056652069092,
+      "learning_rate": 0.00043395574743658927,
+      "loss": 3.534,
+      "step": 25750
+    },
+    {
+      "epoch": 2.7816711590296497,
+      "grad_norm": 0.5810890793800354,
+      "learning_rate": 0.00043363194819212087,
+      "loss": 3.5349,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7870619946091644,
+      "grad_norm": 0.6029720902442932,
+      "learning_rate": 0.0004333081489476524,
+      "loss": 3.5226,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7924528301886795,
+      "grad_norm": 0.6069687008857727,
+      "learning_rate": 0.000432984349703184,
+      "loss": 3.5309,
+      "step": 25900
+    },
+    {
+      "epoch": 2.797843665768194,
+      "grad_norm": 0.5441927313804626,
+      "learning_rate": 0.0004326605504587155,
+      "loss": 3.5211,
+      "step": 25950
+    },
+    {
+      "epoch": 2.803234501347709,
+      "grad_norm": 0.5820786952972412,
+      "learning_rate": 0.0004323367512142472,
+      "loss": 3.5349,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803234501347709,
+      "eval_accuracy": 0.3683050399402583,
+      "eval_loss": 3.512828826904297,
+      "eval_runtime": 185.704,
+      "eval_samples_per_second": 96.988,
+      "eval_steps_per_second": 6.063,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8086253369272236,
+      "grad_norm": 0.5784863829612732,
+      "learning_rate": 0.0004320129519697787,
+      "loss": 3.5331,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8140161725067383,
+      "grad_norm": 0.5669176578521729,
+      "learning_rate": 0.00043168915272531023,
+      "loss": 3.5053,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8194070080862534,
+      "grad_norm": 0.607105016708374,
+      "learning_rate": 0.00043136535348084183,
+      "loss": 3.5076,
+      "step": 26150
+    },
+    {
+      "epoch": 2.824797843665768,
+      "grad_norm": 0.593468189239502,
+      "learning_rate": 0.0004310415542363734,
+      "loss": 3.5354,
+      "step": 26200
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.5860703587532043,
+      "learning_rate": 0.000430717754991905,
+      "loss": 3.5463,
+      "step": 26250
+    },
+    {
+      "epoch": 2.835579514824798,
+      "grad_norm": 0.5652604699134827,
+      "learning_rate": 0.00043039395574743654,
+      "loss": 3.5609,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8409703504043127,
+      "grad_norm": 0.7070855498313904,
+      "learning_rate": 0.00043007015650296814,
+      "loss": 3.5316,
+      "step": 26350
+    },
+    {
+      "epoch": 2.8463611859838274,
+      "grad_norm": 0.5851431488990784,
+      "learning_rate": 0.0004297463572584997,
+      "loss": 3.5403,
+      "step": 26400
+    },
+    {
+      "epoch": 2.8517520215633425,
+      "grad_norm": 0.5997386574745178,
+      "learning_rate": 0.0004294225580140313,
+      "loss": 3.5462,
+      "step": 26450
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.5847904682159424,
+      "learning_rate": 0.00042909875876956285,
+      "loss": 3.5183,
+      "step": 26500
+    },
+    {
+      "epoch": 2.862533692722372,
+      "grad_norm": 0.5951882600784302,
+      "learning_rate": 0.0004287749595250944,
+      "loss": 3.5349,
+      "step": 26550
+    },
+    {
+      "epoch": 2.867924528301887,
+      "grad_norm": 0.5900058746337891,
+      "learning_rate": 0.000428451160280626,
+      "loss": 3.5408,
+      "step": 26600
+    },
+    {
+      "epoch": 2.8733153638814017,
+      "grad_norm": 0.6038098335266113,
+      "learning_rate": 0.0004281273610361575,
+      "loss": 3.5473,
+      "step": 26650
+    },
+    {
+      "epoch": 2.8787061994609164,
+      "grad_norm": 0.5806732177734375,
+      "learning_rate": 0.0004278035617916891,
+      "loss": 3.5336,
+      "step": 26700
+    },
+    {
+      "epoch": 2.884097035040431,
+      "grad_norm": 0.6176631450653076,
+      "learning_rate": 0.00042747976254722066,
+      "loss": 3.547,
+      "step": 26750
+    },
+    {
+      "epoch": 2.889487870619946,
+      "grad_norm": 0.6195639371871948,
+      "learning_rate": 0.00042715596330275226,
+      "loss": 3.5239,
+      "step": 26800
+    },
+    {
+      "epoch": 2.894878706199461,
+      "grad_norm": 0.6487550735473633,
+      "learning_rate": 0.0004268321640582838,
+      "loss": 3.5441,
+      "step": 26850
+    },
+    {
+      "epoch": 2.9002695417789757,
+      "grad_norm": 0.6173131465911865,
+      "learning_rate": 0.0004265083648138154,
+      "loss": 3.508,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.5861935019493103,
+      "learning_rate": 0.00042618456556934697,
+      "loss": 3.5308,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "grad_norm": 0.5515564680099487,
+      "learning_rate": 0.0004258607663248785,
+      "loss": 3.5472,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "eval_accuracy": 0.36972480722533013,
+      "eval_loss": 3.5035452842712402,
+      "eval_runtime": 185.6195,
+      "eval_samples_per_second": 97.032,
+      "eval_steps_per_second": 6.066,
+      "step": 27000
+    },
+    {
+      "epoch": 2.91644204851752,
+      "grad_norm": 0.5942745208740234,
+      "learning_rate": 0.0004255369670804101,
+      "loss": 3.5306,
+      "step": 27050
+    },
+    {
+      "epoch": 2.921832884097035,
+      "grad_norm": 0.5568719506263733,
+      "learning_rate": 0.0004252131678359417,
+      "loss": 3.5345,
+      "step": 27100
+    },
+    {
+      "epoch": 2.92722371967655,
+      "grad_norm": 0.671881914138794,
+      "learning_rate": 0.0004248893685914733,
+      "loss": 3.5308,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9326145552560647,
+      "grad_norm": 0.6649537682533264,
+      "learning_rate": 0.00042457204533189423,
+      "loss": 3.5158,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9380053908355794,
+      "grad_norm": 0.5932601094245911,
+      "learning_rate": 0.0004242482460874258,
+      "loss": 3.5333,
+      "step": 27250
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 0.5902482867240906,
+      "learning_rate": 0.0004239244468429573,
+      "loss": 3.5363,
+      "step": 27300
+    },
+    {
+      "epoch": 2.9487870619946093,
+      "grad_norm": 0.6011241674423218,
+      "learning_rate": 0.0004236006475984889,
+      "loss": 3.521,
+      "step": 27350
+    },
+    {
+      "epoch": 2.954177897574124,
+      "grad_norm": 0.6237984895706177,
+      "learning_rate": 0.00042327684835402043,
+      "loss": 3.5114,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9595687331536387,
+      "grad_norm": 0.5679391026496887,
+      "learning_rate": 0.00042295304910955204,
+      "loss": 3.5188,
+      "step": 27450
+    },
+    {
+      "epoch": 2.964959568733154,
+      "grad_norm": 0.6046316027641296,
+      "learning_rate": 0.0004226292498650836,
+      "loss": 3.5294,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9703504043126685,
+      "grad_norm": 0.5815837383270264,
+      "learning_rate": 0.0004223054506206152,
+      "loss": 3.5182,
+      "step": 27550
+    },
+    {
+      "epoch": 2.975741239892183,
+      "grad_norm": 0.579035222530365,
+      "learning_rate": 0.00042198165137614674,
+      "loss": 3.5268,
+      "step": 27600
+    },
+    {
+      "epoch": 2.981132075471698,
+      "grad_norm": 0.579261064529419,
+      "learning_rate": 0.00042165785213167835,
+      "loss": 3.5053,
+      "step": 27650
+    },
+    {
+      "epoch": 2.986522911051213,
+      "grad_norm": 0.5712079405784607,
+      "learning_rate": 0.0004213340528872099,
+      "loss": 3.519,
+      "step": 27700
+    },
+    {
+      "epoch": 2.9919137466307277,
+      "grad_norm": 0.6659417152404785,
+      "learning_rate": 0.00042101025364274145,
+      "loss": 3.5233,
+      "step": 27750
+    },
+    {
+      "epoch": 2.9973045822102424,
+      "grad_norm": 0.6029751300811768,
+      "learning_rate": 0.00042068645439827305,
+      "loss": 3.5248,
+      "step": 27800
+    },
+    {
+      "epoch": 3.0026954177897576,
+      "grad_norm": 0.5802356600761414,
+      "learning_rate": 0.0004203626551538046,
+      "loss": 3.4885,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0080862533692723,
+      "grad_norm": 0.658331573009491,
+      "learning_rate": 0.0004200388559093362,
+      "loss": 3.4289,
+      "step": 27900
+    },
+    {
+      "epoch": 3.013477088948787,
+      "grad_norm": 0.5875312089920044,
+      "learning_rate": 0.0004197150566648677,
+      "loss": 3.4339,
+      "step": 27950
+    },
+    {
+      "epoch": 3.018867924528302,
+      "grad_norm": 0.652990460395813,
+      "learning_rate": 0.00041939125742039936,
+      "loss": 3.4469,
+      "step": 28000
+    },
+    {
+      "epoch": 3.018867924528302,
+      "eval_accuracy": 0.3702155923195881,
+      "eval_loss": 3.5020062923431396,
+      "eval_runtime": 185.5986,
+      "eval_samples_per_second": 97.043,
+      "eval_steps_per_second": 6.067,
+      "step": 28000
+    },
+    {
+      "epoch": 3.024258760107817,
+      "grad_norm": 0.6217921376228333,
+      "learning_rate": 0.00041906745817593086,
+      "loss": 3.4335,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0296495956873315,
+      "grad_norm": 0.6388250589370728,
+      "learning_rate": 0.00041874365893146247,
+      "loss": 3.4265,
+      "step": 28100
+    },
+    {
+      "epoch": 3.035040431266846,
+      "grad_norm": 0.6219255924224854,
+      "learning_rate": 0.000418419859686994,
+      "loss": 3.4698,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0404312668463613,
+      "grad_norm": 0.6177701354026794,
+      "learning_rate": 0.00041809606044252557,
+      "loss": 3.4493,
+      "step": 28200
+    },
+    {
+      "epoch": 3.045822102425876,
+      "grad_norm": 0.5982428789138794,
+      "learning_rate": 0.00041777226119805717,
+      "loss": 3.4572,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0512129380053907,
+      "grad_norm": 0.6229225993156433,
+      "learning_rate": 0.0004174484619535887,
+      "loss": 3.4536,
+      "step": 28300
+    },
+    {
+      "epoch": 3.056603773584906,
+      "grad_norm": 0.6119500994682312,
+      "learning_rate": 0.00041712466270912033,
+      "loss": 3.4463,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0619946091644206,
+      "grad_norm": 0.5835257172584534,
+      "learning_rate": 0.0004168008634646519,
+      "loss": 3.4254,
+      "step": 28400
+    },
+    {
+      "epoch": 3.0673854447439353,
+      "grad_norm": 0.6458772420883179,
+      "learning_rate": 0.0004164770642201835,
+      "loss": 3.4668,
+      "step": 28450
+    },
+    {
+      "epoch": 3.07277628032345,
+      "grad_norm": 0.6388500332832336,
+      "learning_rate": 0.00041615326497571503,
+      "loss": 3.4587,
+      "step": 28500
+    },
+    {
+      "epoch": 3.078167115902965,
+      "grad_norm": 0.6281006932258606,
+      "learning_rate": 0.0004158294657312466,
+      "loss": 3.4468,
+      "step": 28550
+    },
+    {
+      "epoch": 3.08355795148248,
+      "grad_norm": 0.5893887877464294,
+      "learning_rate": 0.0004155056664867782,
+      "loss": 3.443,
+      "step": 28600
+    },
+    {
+      "epoch": 3.0889487870619945,
+      "grad_norm": 0.6501525640487671,
+      "learning_rate": 0.0004151818672423097,
+      "loss": 3.4405,
+      "step": 28650
+    },
+    {
+      "epoch": 3.0943396226415096,
+      "grad_norm": 0.5806996822357178,
+      "learning_rate": 0.0004148580679978413,
+      "loss": 3.4524,
+      "step": 28700
+    },
+    {
+      "epoch": 3.0997304582210243,
+      "grad_norm": 0.6186302304267883,
+      "learning_rate": 0.00041453426875337284,
+      "loss": 3.4418,
+      "step": 28750
+    },
+    {
+      "epoch": 3.105121293800539,
+      "grad_norm": 0.5666465759277344,
+      "learning_rate": 0.00041421046950890445,
+      "loss": 3.4418,
+      "step": 28800
+    },
+    {
+      "epoch": 3.1105121293800537,
+      "grad_norm": 0.5941797494888306,
+      "learning_rate": 0.000413886670264436,
+      "loss": 3.4283,
+      "step": 28850
+    },
+    {
+      "epoch": 3.115902964959569,
+      "grad_norm": 0.5796331167221069,
+      "learning_rate": 0.0004135628710199676,
+      "loss": 3.4461,
+      "step": 28900
+    },
+    {
+      "epoch": 3.1212938005390836,
+      "grad_norm": 0.6226239204406738,
+      "learning_rate": 0.00041323907177549915,
+      "loss": 3.4618,
+      "step": 28950
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "grad_norm": 0.6251356601715088,
+      "learning_rate": 0.0004129152725310307,
+      "loss": 3.4316,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "eval_accuracy": 0.37116760892261863,
+      "eval_loss": 3.4976043701171875,
+      "eval_runtime": 185.5867,
+      "eval_samples_per_second": 97.049,
+      "eval_steps_per_second": 6.067,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1320754716981134,
+      "grad_norm": 0.6396812200546265,
+      "learning_rate": 0.0004125914732865623,
+      "loss": 3.4517,
+      "step": 29050
+    },
+    {
+      "epoch": 3.137466307277628,
+      "grad_norm": 0.6136710047721863,
+      "learning_rate": 0.00041226767404209386,
+      "loss": 3.443,
+      "step": 29100
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.6114400625228882,
+      "learning_rate": 0.00041194387479762546,
+      "loss": 3.4614,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1482479784366575,
+      "grad_norm": 0.5910611748695374,
+      "learning_rate": 0.000411620075553157,
+      "loss": 3.4603,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1536388140161726,
+      "grad_norm": 0.603925347328186,
+      "learning_rate": 0.0004112962763086886,
+      "loss": 3.4462,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1590296495956873,
+      "grad_norm": 0.5798592567443848,
+      "learning_rate": 0.00041097895304910946,
+      "loss": 3.4539,
+      "step": 29300
+    },
+    {
+      "epoch": 3.164420485175202,
+      "grad_norm": 0.6233105063438416,
+      "learning_rate": 0.00041065515380464107,
+      "loss": 3.4674,
+      "step": 29350
+    },
+    {
+      "epoch": 3.169811320754717,
+      "grad_norm": 0.5787250399589539,
+      "learning_rate": 0.0004103313545601726,
+      "loss": 3.4604,
+      "step": 29400
+    },
+    {
+      "epoch": 3.175202156334232,
+      "grad_norm": 0.606056272983551,
+      "learning_rate": 0.0004100075553157042,
+      "loss": 3.4725,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1805929919137466,
+      "grad_norm": 0.6234708428382874,
+      "learning_rate": 0.00040968375607123577,
+      "loss": 3.4667,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1859838274932613,
+      "grad_norm": 0.6266785860061646,
+      "learning_rate": 0.0004093599568267674,
+      "loss": 3.4427,
+      "step": 29550
+    },
+    {
+      "epoch": 3.1913746630727764,
+      "grad_norm": 0.6004592180252075,
+      "learning_rate": 0.0004090361575822989,
+      "loss": 3.4649,
+      "step": 29600
+    },
+    {
+      "epoch": 3.196765498652291,
+      "grad_norm": 0.6037859320640564,
+      "learning_rate": 0.00040871235833783053,
+      "loss": 3.4508,
+      "step": 29650
+    },
+    {
+      "epoch": 3.202156334231806,
+      "grad_norm": 0.5727283358573914,
+      "learning_rate": 0.0004083885590933621,
+      "loss": 3.4559,
+      "step": 29700
+    },
+    {
+      "epoch": 3.207547169811321,
+      "grad_norm": 0.5819792747497559,
+      "learning_rate": 0.00040806475984889363,
+      "loss": 3.4652,
+      "step": 29750
+    },
+    {
+      "epoch": 3.2129380053908356,
+      "grad_norm": 0.599252462387085,
+      "learning_rate": 0.0004077474365893146,
+      "loss": 3.445,
+      "step": 29800
+    },
+    {
+      "epoch": 3.2183288409703503,
+      "grad_norm": 0.6133659482002258,
+      "learning_rate": 0.0004074236373448462,
+      "loss": 3.4693,
+      "step": 29850
+    },
+    {
+      "epoch": 3.223719676549865,
+      "grad_norm": 0.6273553967475891,
+      "learning_rate": 0.00040709983810037774,
+      "loss": 3.4755,
+      "step": 29900
+    },
+    {
+      "epoch": 3.22911051212938,
+      "grad_norm": 0.629896342754364,
+      "learning_rate": 0.00040677603885590934,
+      "loss": 3.466,
+      "step": 29950
+    },
+    {
+      "epoch": 3.234501347708895,
+      "grad_norm": 0.6052857041358948,
+      "learning_rate": 0.00040645223961144084,
+      "loss": 3.4541,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234501347708895,
+      "eval_accuracy": 0.37135927261672336,
+      "eval_loss": 3.4913108348846436,
+      "eval_runtime": 185.5243,
+      "eval_samples_per_second": 97.082,
+      "eval_steps_per_second": 6.069,
+      "step": 30000
+    },
+    {
+      "epoch": 3.2398921832884096,
+      "grad_norm": 0.5984300971031189,
+      "learning_rate": 0.0004061284403669725,
+      "loss": 3.4503,
+      "step": 30050
+    },
+    {
+      "epoch": 3.2452830188679247,
+      "grad_norm": 0.6453818082809448,
+      "learning_rate": 0.000405804641122504,
+      "loss": 3.4511,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2506738544474394,
+      "grad_norm": 0.5658475756645203,
+      "learning_rate": 0.00040548084187803555,
+      "loss": 3.452,
+      "step": 30150
+    },
+    {
+      "epoch": 3.256064690026954,
+      "grad_norm": 0.6257442235946655,
+      "learning_rate": 0.00040515704263356715,
+      "loss": 3.4722,
+      "step": 30200
+    },
+    {
+      "epoch": 3.2614555256064692,
+      "grad_norm": 0.6362524032592773,
+      "learning_rate": 0.0004048332433890987,
+      "loss": 3.4659,
+      "step": 30250
+    },
+    {
+      "epoch": 3.266846361185984,
+      "grad_norm": 0.6771724224090576,
+      "learning_rate": 0.0004045094441446303,
+      "loss": 3.4672,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2722371967654986,
+      "grad_norm": 0.6712861061096191,
+      "learning_rate": 0.00040418564490016186,
+      "loss": 3.4418,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2776280323450133,
+      "grad_norm": 0.7446663975715637,
+      "learning_rate": 0.00040386184565569346,
+      "loss": 3.4748,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2830188679245285,
+      "grad_norm": 0.6019281148910522,
+      "learning_rate": 0.000403538046411225,
+      "loss": 3.4587,
+      "step": 30450
+    },
+    {
+      "epoch": 3.288409703504043,
+      "grad_norm": 0.6105091571807861,
+      "learning_rate": 0.00040321424716675656,
+      "loss": 3.4432,
+      "step": 30500
+    },
+    {
+      "epoch": 3.293800539083558,
+      "grad_norm": 0.6371334195137024,
+      "learning_rate": 0.00040289044792228817,
+      "loss": 3.474,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2991913746630726,
+      "grad_norm": 0.6892150044441223,
+      "learning_rate": 0.0004025666486778197,
+      "loss": 3.4633,
+      "step": 30600
+    },
+    {
+      "epoch": 3.3045822102425877,
+      "grad_norm": 0.6133437156677246,
+      "learning_rate": 0.0004022428494333513,
+      "loss": 3.4439,
+      "step": 30650
+    },
+    {
+      "epoch": 3.3099730458221024,
+      "grad_norm": 0.6131065487861633,
+      "learning_rate": 0.0004019190501888828,
+      "loss": 3.462,
+      "step": 30700
+    },
+    {
+      "epoch": 3.315363881401617,
+      "grad_norm": 0.6125616431236267,
+      "learning_rate": 0.0004015952509444144,
+      "loss": 3.431,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3207547169811322,
+      "grad_norm": 0.6142444610595703,
+      "learning_rate": 0.000401271451699946,
+      "loss": 3.467,
+      "step": 30800
+    },
+    {
+      "epoch": 3.326145552560647,
+      "grad_norm": 0.6196108460426331,
+      "learning_rate": 0.0004009476524554776,
+      "loss": 3.4472,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3315363881401616,
+      "grad_norm": 0.6309120059013367,
+      "learning_rate": 0.00040062385321100913,
+      "loss": 3.4697,
+      "step": 30900
+    },
+    {
+      "epoch": 3.3369272237196768,
+      "grad_norm": 0.6041767001152039,
+      "learning_rate": 0.0004003000539665407,
+      "loss": 3.4474,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "grad_norm": 0.6576740741729736,
+      "learning_rate": 0.0003999762547220723,
+      "loss": 3.459,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "eval_accuracy": 0.37212234184784443,
+      "eval_loss": 3.4848906993865967,
+      "eval_runtime": 185.878,
+      "eval_samples_per_second": 96.897,
+      "eval_steps_per_second": 6.058,
+      "step": 31000
+    },
+    {
+      "epoch": 3.347708894878706,
+      "grad_norm": 0.6974944472312927,
+      "learning_rate": 0.00039965245547760384,
+      "loss": 3.4624,
+      "step": 31050
+    },
+    {
+      "epoch": 3.353099730458221,
+      "grad_norm": 0.6188839673995972,
+      "learning_rate": 0.00039932865623313544,
+      "loss": 3.4567,
+      "step": 31100
+    },
+    {
+      "epoch": 3.358490566037736,
+      "grad_norm": 0.6468076705932617,
+      "learning_rate": 0.000399004856988667,
+      "loss": 3.4677,
+      "step": 31150
+    },
+    {
+      "epoch": 3.3638814016172507,
+      "grad_norm": 0.6456601023674011,
+      "learning_rate": 0.0003986810577441986,
+      "loss": 3.451,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3692722371967654,
+      "grad_norm": 0.6178808808326721,
+      "learning_rate": 0.00039835725849973015,
+      "loss": 3.4563,
+      "step": 31250
+    },
+    {
+      "epoch": 3.37466307277628,
+      "grad_norm": 0.6076691150665283,
+      "learning_rate": 0.00039803345925526175,
+      "loss": 3.4816,
+      "step": 31300
+    },
+    {
+      "epoch": 3.3800539083557952,
+      "grad_norm": 0.6464076638221741,
+      "learning_rate": 0.00039770966001079325,
+      "loss": 3.4531,
+      "step": 31350
+    },
+    {
+      "epoch": 3.38544474393531,
+      "grad_norm": 0.6401315927505493,
+      "learning_rate": 0.0003973858607663248,
+      "loss": 3.4517,
+      "step": 31400
+    },
+    {
+      "epoch": 3.3908355795148246,
+      "grad_norm": 0.5781538486480713,
+      "learning_rate": 0.0003970620615218564,
+      "loss": 3.4566,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3962264150943398,
+      "grad_norm": 0.6438537836074829,
+      "learning_rate": 0.00039673826227738795,
+      "loss": 3.4531,
+      "step": 31500
+    },
+    {
+      "epoch": 3.4016172506738545,
+      "grad_norm": 0.6200994253158569,
+      "learning_rate": 0.00039641446303291956,
+      "loss": 3.4693,
+      "step": 31550
+    },
+    {
+      "epoch": 3.407008086253369,
+      "grad_norm": 0.6058845520019531,
+      "learning_rate": 0.0003960906637884511,
+      "loss": 3.447,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4123989218328843,
+      "grad_norm": 0.650837779045105,
+      "learning_rate": 0.0003957668645439827,
+      "loss": 3.4504,
+      "step": 31650
+    },
+    {
+      "epoch": 3.417789757412399,
+      "grad_norm": 0.6446054577827454,
+      "learning_rate": 0.00039544306529951426,
+      "loss": 3.4753,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4231805929919137,
+      "grad_norm": 0.6618728637695312,
+      "learning_rate": 0.00039511926605504587,
+      "loss": 3.4503,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4285714285714284,
+      "grad_norm": 0.5940080285072327,
+      "learning_rate": 0.0003947954668105774,
+      "loss": 3.4537,
+      "step": 31800
+    },
+    {
+      "epoch": 3.4339622641509435,
+      "grad_norm": 0.6101294159889221,
+      "learning_rate": 0.00039447166756610897,
+      "loss": 3.4812,
+      "step": 31850
+    },
+    {
+      "epoch": 3.439353099730458,
+      "grad_norm": 0.7297942638397217,
+      "learning_rate": 0.0003941478683216406,
+      "loss": 3.465,
+      "step": 31900
+    },
+    {
+      "epoch": 3.444743935309973,
+      "grad_norm": 0.6869504451751709,
+      "learning_rate": 0.0003938240690771721,
+      "loss": 3.4444,
+      "step": 31950
+    },
+    {
+      "epoch": 3.450134770889488,
+      "grad_norm": 0.6336163878440857,
+      "learning_rate": 0.00039350026983270373,
+      "loss": 3.46,
+      "step": 32000
+    },
+    {
+      "epoch": 3.450134770889488,
+      "eval_accuracy": 0.3732112610895212,
+      "eval_loss": 3.4762558937072754,
+      "eval_runtime": 185.4107,
+      "eval_samples_per_second": 97.141,
+      "eval_steps_per_second": 6.073,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4555256064690028,
+      "grad_norm": 0.610059916973114,
+      "learning_rate": 0.00039317647058823523,
+      "loss": 3.4635,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4609164420485174,
+      "grad_norm": 0.6818318367004395,
+      "learning_rate": 0.00039285267134376683,
+      "loss": 3.4566,
+      "step": 32100
+    },
+    {
+      "epoch": 3.466307277628032,
+      "grad_norm": 0.631006121635437,
+      "learning_rate": 0.0003925288720992984,
+      "loss": 3.4697,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4716981132075473,
+      "grad_norm": 0.6268739104270935,
+      "learning_rate": 0.00039220507285482993,
+      "loss": 3.4505,
+      "step": 32200
+    },
+    {
+      "epoch": 3.477088948787062,
+      "grad_norm": 0.6351046562194824,
+      "learning_rate": 0.00039188127361036154,
+      "loss": 3.4677,
+      "step": 32250
+    },
+    {
+      "epoch": 3.4824797843665767,
+      "grad_norm": 0.6072060465812683,
+      "learning_rate": 0.0003915574743658931,
+      "loss": 3.4813,
+      "step": 32300
+    },
+    {
+      "epoch": 3.487870619946092,
+      "grad_norm": 0.6056317090988159,
+      "learning_rate": 0.0003912336751214247,
+      "loss": 3.4421,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4932614555256065,
+      "grad_norm": 0.6372120976448059,
+      "learning_rate": 0.00039090987587695624,
+      "loss": 3.4727,
+      "step": 32400
+    },
+    {
+      "epoch": 3.498652291105121,
+      "grad_norm": 0.6165457963943481,
+      "learning_rate": 0.00039058607663248785,
+      "loss": 3.4454,
+      "step": 32450
+    },
+    {
+      "epoch": 3.5040431266846364,
+      "grad_norm": 0.5812312960624695,
+      "learning_rate": 0.0003902622773880194,
+      "loss": 3.475,
+      "step": 32500
+    },
+    {
+      "epoch": 3.509433962264151,
+      "grad_norm": 0.580613374710083,
+      "learning_rate": 0.000389938478143551,
+      "loss": 3.4623,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5148247978436657,
+      "grad_norm": 0.6294305324554443,
+      "learning_rate": 0.00038961467889908255,
+      "loss": 3.4584,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5202156334231804,
+      "grad_norm": 0.6396526098251343,
+      "learning_rate": 0.00038929087965461405,
+      "loss": 3.4683,
+      "step": 32650
+    },
+    {
+      "epoch": 3.525606469002695,
+      "grad_norm": 0.6367679238319397,
+      "learning_rate": 0.00038896708041014566,
+      "loss": 3.4441,
+      "step": 32700
+    },
+    {
+      "epoch": 3.5309973045822103,
+      "grad_norm": 0.6563025116920471,
+      "learning_rate": 0.0003886432811656772,
+      "loss": 3.4545,
+      "step": 32750
+    },
+    {
+      "epoch": 3.536388140161725,
+      "grad_norm": 0.5986255407333374,
+      "learning_rate": 0.0003883194819212088,
+      "loss": 3.4661,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5417789757412397,
+      "grad_norm": 0.617114782333374,
+      "learning_rate": 0.00038799568267674036,
+      "loss": 3.4273,
+      "step": 32850
+    },
+    {
+      "epoch": 3.547169811320755,
+      "grad_norm": 0.643477201461792,
+      "learning_rate": 0.00038767188343227197,
+      "loss": 3.4509,
+      "step": 32900
+    },
+    {
+      "epoch": 3.5525606469002695,
+      "grad_norm": 0.6376758217811584,
+      "learning_rate": 0.0003873480841878035,
+      "loss": 3.466,
+      "step": 32950
+    },
+    {
+      "epoch": 3.557951482479784,
+      "grad_norm": 0.6447632312774658,
+      "learning_rate": 0.0003870242849433351,
+      "loss": 3.4641,
+      "step": 33000
+    },
+    {
+      "epoch": 3.557951482479784,
+      "eval_accuracy": 0.37378331854386443,
+      "eval_loss": 3.4722912311553955,
+      "eval_runtime": 185.7305,
+      "eval_samples_per_second": 96.974,
+      "eval_steps_per_second": 6.063,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5633423180592994,
+      "grad_norm": 0.5970557332038879,
+      "learning_rate": 0.00038670048569886667,
+      "loss": 3.4653,
+      "step": 33050
+    },
+    {
+      "epoch": 3.568733153638814,
+      "grad_norm": 0.6381769180297852,
+      "learning_rate": 0.0003863766864543982,
+      "loss": 3.4537,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5741239892183287,
+      "grad_norm": 0.6807313561439514,
+      "learning_rate": 0.00038605288720992983,
+      "loss": 3.4621,
+      "step": 33150
+    },
+    {
+      "epoch": 3.579514824797844,
+      "grad_norm": 0.6421796083450317,
+      "learning_rate": 0.0003857290879654614,
+      "loss": 3.4613,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5849056603773586,
+      "grad_norm": 0.6369513273239136,
+      "learning_rate": 0.000385405288720993,
+      "loss": 3.4554,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5902964959568733,
+      "grad_norm": 0.6491246223449707,
+      "learning_rate": 0.00038508148947652453,
+      "loss": 3.4451,
+      "step": 33300
+    },
+    {
+      "epoch": 3.595687331536388,
+      "grad_norm": 0.626429557800293,
+      "learning_rate": 0.00038475769023205614,
+      "loss": 3.4556,
+      "step": 33350
+    },
+    {
+      "epoch": 3.601078167115903,
+      "grad_norm": 0.5887939929962158,
+      "learning_rate": 0.00038443389098758763,
+      "loss": 3.4556,
+      "step": 33400
+    },
+    {
+      "epoch": 3.606469002695418,
+      "grad_norm": 0.6233583092689514,
+      "learning_rate": 0.00038411009174311924,
+      "loss": 3.4661,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6118598382749325,
+      "grad_norm": 0.6247572302818298,
+      "learning_rate": 0.0003837862924986508,
+      "loss": 3.4714,
+      "step": 33500
+    },
+    {
+      "epoch": 3.617250673854447,
+      "grad_norm": 0.6160756349563599,
+      "learning_rate": 0.00038346249325418234,
+      "loss": 3.4511,
+      "step": 33550
+    },
+    {
+      "epoch": 3.6226415094339623,
+      "grad_norm": 0.6060353517532349,
+      "learning_rate": 0.00038313869400971395,
+      "loss": 3.4539,
+      "step": 33600
+    },
+    {
+      "epoch": 3.628032345013477,
+      "grad_norm": 0.7138940691947937,
+      "learning_rate": 0.0003828148947652455,
+      "loss": 3.4514,
+      "step": 33650
+    },
+    {
+      "epoch": 3.6334231805929917,
+      "grad_norm": 0.5955808162689209,
+      "learning_rate": 0.0003824910955207771,
+      "loss": 3.4388,
+      "step": 33700
+    },
+    {
+      "epoch": 3.638814016172507,
+      "grad_norm": 0.6242125034332275,
+      "learning_rate": 0.00038216729627630865,
+      "loss": 3.4609,
+      "step": 33750
+    },
+    {
+      "epoch": 3.6442048517520216,
+      "grad_norm": 0.6291961073875427,
+      "learning_rate": 0.00038184349703184026,
+      "loss": 3.4417,
+      "step": 33800
+    },
+    {
+      "epoch": 3.6495956873315363,
+      "grad_norm": 0.6095578074455261,
+      "learning_rate": 0.00038152617377226115,
+      "loss": 3.4738,
+      "step": 33850
+    },
+    {
+      "epoch": 3.6549865229110514,
+      "grad_norm": 0.6353351473808289,
+      "learning_rate": 0.00038120237452779276,
+      "loss": 3.4489,
+      "step": 33900
+    },
+    {
+      "epoch": 3.660377358490566,
+      "grad_norm": 0.6624910235404968,
+      "learning_rate": 0.0003808785752833243,
+      "loss": 3.4495,
+      "step": 33950
+    },
+    {
+      "epoch": 3.665768194070081,
+      "grad_norm": 0.6535131335258484,
+      "learning_rate": 0.0003805547760388559,
+      "loss": 3.4617,
+      "step": 34000
+    },
+    {
+      "epoch": 3.665768194070081,
+      "eval_accuracy": 0.37374366023981215,
+      "eval_loss": 3.4666104316711426,
+      "eval_runtime": 185.6691,
+      "eval_samples_per_second": 97.006,
+      "eval_steps_per_second": 6.065,
+      "step": 34000
+    },
+    {
+      "epoch": 3.671159029649596,
+      "grad_norm": 0.6361366510391235,
+      "learning_rate": 0.0003802309767943874,
+      "loss": 3.4543,
+      "step": 34050
+    },
+    {
+      "epoch": 3.6765498652291106,
+      "grad_norm": 0.6315745115280151,
+      "learning_rate": 0.000379907177549919,
+      "loss": 3.4721,
+      "step": 34100
+    },
+    {
+      "epoch": 3.6819407008086253,
+      "grad_norm": 0.6330356001853943,
+      "learning_rate": 0.0003795898542903399,
+      "loss": 3.4687,
+      "step": 34150
+    },
+    {
+      "epoch": 3.68733153638814,
+      "grad_norm": 0.6052745580673218,
+      "learning_rate": 0.0003792660550458715,
+      "loss": 3.4487,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6927223719676547,
+      "grad_norm": 0.621346652507782,
+      "learning_rate": 0.00037894225580140307,
+      "loss": 3.4428,
+      "step": 34250
+    },
+    {
+      "epoch": 3.69811320754717,
+      "grad_norm": 0.6476735472679138,
+      "learning_rate": 0.00037861845655693467,
+      "loss": 3.4435,
+      "step": 34300
+    },
+    {
+      "epoch": 3.7035040431266846,
+      "grad_norm": 0.6602584719657898,
+      "learning_rate": 0.0003782946573124662,
+      "loss": 3.4482,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7088948787061993,
+      "grad_norm": 0.5904750823974609,
+      "learning_rate": 0.00037797085806799783,
+      "loss": 3.4453,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7142857142857144,
+      "grad_norm": 0.629729151725769,
+      "learning_rate": 0.0003776470588235294,
+      "loss": 3.4641,
+      "step": 34450
+    },
+    {
+      "epoch": 3.719676549865229,
+      "grad_norm": 0.624595582485199,
+      "learning_rate": 0.000377323259579061,
+      "loss": 3.4481,
+      "step": 34500
+    },
+    {
+      "epoch": 3.725067385444744,
+      "grad_norm": 0.654861330986023,
+      "learning_rate": 0.00037699946033459253,
+      "loss": 3.4399,
+      "step": 34550
+    },
+    {
+      "epoch": 3.730458221024259,
+      "grad_norm": 0.7012645602226257,
+      "learning_rate": 0.0003766756610901241,
+      "loss": 3.4608,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7358490566037736,
+      "grad_norm": 0.6455208659172058,
+      "learning_rate": 0.0003763518618456557,
+      "loss": 3.458,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7412398921832883,
+      "grad_norm": 0.6590430736541748,
+      "learning_rate": 0.0003760280626011872,
+      "loss": 3.4696,
+      "step": 34700
+    },
+    {
+      "epoch": 3.7466307277628035,
+      "grad_norm": 0.6089630126953125,
+      "learning_rate": 0.0003757042633567188,
+      "loss": 3.4648,
+      "step": 34750
+    },
+    {
+      "epoch": 3.752021563342318,
+      "grad_norm": 0.639079213142395,
+      "learning_rate": 0.00037538046411225034,
+      "loss": 3.4445,
+      "step": 34800
+    },
+    {
+      "epoch": 3.757412398921833,
+      "grad_norm": 0.6706318855285645,
+      "learning_rate": 0.00037505666486778195,
+      "loss": 3.4467,
+      "step": 34850
+    },
+    {
+      "epoch": 3.7628032345013476,
+      "grad_norm": 0.6065903902053833,
+      "learning_rate": 0.0003747328656233135,
+      "loss": 3.4636,
+      "step": 34900
+    },
+    {
+      "epoch": 3.7681940700808623,
+      "grad_norm": 0.6488689184188843,
+      "learning_rate": 0.0003744090663788451,
+      "loss": 3.4549,
+      "step": 34950
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "grad_norm": 0.6796548366546631,
+      "learning_rate": 0.00037408526713437665,
+      "loss": 3.4621,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "eval_accuracy": 0.375207323291561,
+      "eval_loss": 3.4563443660736084,
+      "eval_runtime": 185.4472,
+      "eval_samples_per_second": 97.122,
+      "eval_steps_per_second": 6.072,
+      "step": 35000
+    },
+    {
+      "epoch": 3.778975741239892,
+      "grad_norm": 0.6050362586975098,
+      "learning_rate": 0.0003737614678899082,
+      "loss": 3.4729,
+      "step": 35050
+    },
+    {
+      "epoch": 3.784366576819407,
+      "grad_norm": 0.6607158184051514,
+      "learning_rate": 0.0003734376686454398,
+      "loss": 3.4396,
+      "step": 35100
+    },
+    {
+      "epoch": 3.789757412398922,
+      "grad_norm": 0.6147844791412354,
+      "learning_rate": 0.00037311386940097136,
+      "loss": 3.4619,
+      "step": 35150
+    },
+    {
+      "epoch": 3.7951482479784366,
+      "grad_norm": 0.6096859574317932,
+      "learning_rate": 0.00037279007015650296,
+      "loss": 3.4604,
+      "step": 35200
+    },
+    {
+      "epoch": 3.8005390835579513,
+      "grad_norm": 0.6354143023490906,
+      "learning_rate": 0.0003724662709120345,
+      "loss": 3.4424,
+      "step": 35250
+    },
+    {
+      "epoch": 3.8059299191374665,
+      "grad_norm": 0.6351900696754456,
+      "learning_rate": 0.0003721424716675661,
+      "loss": 3.4558,
+      "step": 35300
+    },
+    {
+      "epoch": 3.811320754716981,
+      "grad_norm": 0.6564791798591614,
+      "learning_rate": 0.0003718186724230976,
+      "loss": 3.4569,
+      "step": 35350
+    },
+    {
+      "epoch": 3.816711590296496,
+      "grad_norm": 0.6275109052658081,
+      "learning_rate": 0.0003714948731786293,
+      "loss": 3.4577,
+      "step": 35400
+    },
+    {
+      "epoch": 3.822102425876011,
+      "grad_norm": 0.6954398155212402,
+      "learning_rate": 0.00037117107393416077,
+      "loss": 3.4508,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8274932614555257,
+      "grad_norm": 0.6001200079917908,
+      "learning_rate": 0.0003708472746896923,
+      "loss": 3.4531,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8328840970350404,
+      "grad_norm": 0.6402223110198975,
+      "learning_rate": 0.0003705234754452239,
+      "loss": 3.4651,
+      "step": 35550
+    },
+    {
+      "epoch": 3.838274932614555,
+      "grad_norm": 0.6889019012451172,
+      "learning_rate": 0.0003701996762007555,
+      "loss": 3.4709,
+      "step": 35600
+    },
+    {
+      "epoch": 3.8436657681940702,
+      "grad_norm": 0.6361086964607239,
+      "learning_rate": 0.0003698758769562871,
+      "loss": 3.4676,
+      "step": 35650
+    },
+    {
+      "epoch": 3.849056603773585,
+      "grad_norm": 0.6023678183555603,
+      "learning_rate": 0.00036955207771181863,
+      "loss": 3.4698,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8544474393530996,
+      "grad_norm": 0.6561288237571716,
+      "learning_rate": 0.00036922827846735024,
+      "loss": 3.4508,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8598382749326143,
+      "grad_norm": 1.0986155271530151,
+      "learning_rate": 0.0003689044792228818,
+      "loss": 3.4579,
+      "step": 35800
+    },
+    {
+      "epoch": 3.8652291105121295,
+      "grad_norm": 0.6084561944007874,
+      "learning_rate": 0.0003685806799784134,
+      "loss": 3.4331,
+      "step": 35850
+    },
+    {
+      "epoch": 3.870619946091644,
+      "grad_norm": 0.6422281265258789,
+      "learning_rate": 0.00036825688073394494,
+      "loss": 3.4485,
+      "step": 35900
+    },
+    {
+      "epoch": 3.876010781671159,
+      "grad_norm": 0.6261025667190552,
+      "learning_rate": 0.0003679330814894765,
+      "loss": 3.4726,
+      "step": 35950
+    },
+    {
+      "epoch": 3.881401617250674,
+      "grad_norm": 0.6010216474533081,
+      "learning_rate": 0.0003676092822450081,
+      "loss": 3.4515,
+      "step": 36000
+    },
+    {
+      "epoch": 3.881401617250674,
+      "eval_accuracy": 0.37517222840879694,
+      "eval_loss": 3.452343702316284,
+      "eval_runtime": 185.5713,
+      "eval_samples_per_second": 97.057,
+      "eval_steps_per_second": 6.068,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8867924528301887,
+      "grad_norm": 0.7316910028457642,
+      "learning_rate": 0.0003672854830005396,
+      "loss": 3.4601,
+      "step": 36050
+    },
+    {
+      "epoch": 3.8921832884097034,
+      "grad_norm": 0.7084899544715881,
+      "learning_rate": 0.0003669616837560712,
+      "loss": 3.4555,
+      "step": 36100
+    },
+    {
+      "epoch": 3.8975741239892185,
+      "grad_norm": 0.6113714575767517,
+      "learning_rate": 0.00036663788451160275,
+      "loss": 3.459,
+      "step": 36150
+    },
+    {
+      "epoch": 3.9029649595687332,
+      "grad_norm": 0.6242182850837708,
+      "learning_rate": 0.00036631408526713435,
+      "loss": 3.4502,
+      "step": 36200
+    },
+    {
+      "epoch": 3.908355795148248,
+      "grad_norm": 0.6214504241943359,
+      "learning_rate": 0.0003659902860226659,
+      "loss": 3.4639,
+      "step": 36250
+    },
+    {
+      "epoch": 3.913746630727763,
+      "grad_norm": 0.6366416811943054,
+      "learning_rate": 0.00036566648677819745,
+      "loss": 3.4455,
+      "step": 36300
+    },
+    {
+      "epoch": 3.9191374663072778,
+      "grad_norm": 0.6250828504562378,
+      "learning_rate": 0.00036534268753372906,
+      "loss": 3.4678,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9245283018867925,
+      "grad_norm": 0.6313366293907166,
+      "learning_rate": 0.0003650188882892606,
+      "loss": 3.4671,
+      "step": 36400
+    },
+    {
+      "epoch": 3.929919137466307,
+      "grad_norm": 0.6488140225410461,
+      "learning_rate": 0.0003646950890447922,
+      "loss": 3.4442,
+      "step": 36450
+    },
+    {
+      "epoch": 3.935309973045822,
+      "grad_norm": 0.6659651398658752,
+      "learning_rate": 0.00036437128980032376,
+      "loss": 3.4529,
+      "step": 36500
+    },
+    {
+      "epoch": 3.940700808625337,
+      "grad_norm": 0.6575341820716858,
+      "learning_rate": 0.00036404749055585537,
+      "loss": 3.4537,
+      "step": 36550
+    },
+    {
+      "epoch": 3.9460916442048517,
+      "grad_norm": 0.635816216468811,
+      "learning_rate": 0.0003637236913113869,
+      "loss": 3.4504,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9514824797843664,
+      "grad_norm": 0.736781895160675,
+      "learning_rate": 0.0003633998920669185,
+      "loss": 3.4417,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9568733153638815,
+      "grad_norm": 0.6434977054595947,
+      "learning_rate": 0.0003630760928224501,
+      "loss": 3.4518,
+      "step": 36700
+    },
+    {
+      "epoch": 3.9622641509433962,
+      "grad_norm": 0.6709410548210144,
+      "learning_rate": 0.00036275229357798157,
+      "loss": 3.4436,
+      "step": 36750
+    },
+    {
+      "epoch": 3.967654986522911,
+      "grad_norm": 0.6328015327453613,
+      "learning_rate": 0.0003624284943335132,
+      "loss": 3.4378,
+      "step": 36800
+    },
+    {
+      "epoch": 3.973045822102426,
+      "grad_norm": 0.6492040157318115,
+      "learning_rate": 0.00036210469508904473,
+      "loss": 3.4524,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9784366576819408,
+      "grad_norm": 0.6740325689315796,
+      "learning_rate": 0.00036178089584457633,
+      "loss": 3.4438,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9838274932614555,
+      "grad_norm": 0.7030938863754272,
+      "learning_rate": 0.0003614570966001079,
+      "loss": 3.4547,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "grad_norm": 0.6631276607513428,
+      "learning_rate": 0.0003611332973556395,
+      "loss": 3.4491,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "eval_accuracy": 0.3760049441410071,
+      "eval_loss": 3.447164297103882,
+      "eval_runtime": 185.6169,
+      "eval_samples_per_second": 97.033,
+      "eval_steps_per_second": 6.066,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9946091644204853,
+      "grad_norm": 0.6529706120491028,
+      "learning_rate": 0.00036080949811117104,
+      "loss": 3.4619,
+      "step": 37050
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.2341878414154053,
+      "learning_rate": 0.00036048569886670264,
+      "loss": 3.4346,
+      "step": 37100
+    },
+    {
+      "epoch": 4.005390835579515,
+      "grad_norm": 0.6886982321739197,
+      "learning_rate": 0.0003601618996222342,
+      "loss": 3.3297,
+      "step": 37150
+    },
+    {
+      "epoch": 4.010781671159029,
+      "grad_norm": 0.617487370967865,
+      "learning_rate": 0.00035983810037776574,
+      "loss": 3.3581,
+      "step": 37200
+    },
+    {
+      "epoch": 4.0161725067385445,
+      "grad_norm": 0.6758488416671753,
+      "learning_rate": 0.00035951430113329735,
+      "loss": 3.3531,
+      "step": 37250
+    },
+    {
+      "epoch": 4.02156334231806,
+      "grad_norm": 0.63946133852005,
+      "learning_rate": 0.0003591905018888289,
+      "loss": 3.3568,
+      "step": 37300
+    },
+    {
+      "epoch": 4.026954177897574,
+      "grad_norm": 0.673195481300354,
+      "learning_rate": 0.0003588667026443605,
+      "loss": 3.3441,
+      "step": 37350
+    },
+    {
+      "epoch": 4.032345013477089,
+      "grad_norm": 0.651838481426239,
+      "learning_rate": 0.000358542903399892,
+      "loss": 3.3634,
+      "step": 37400
+    },
+    {
+      "epoch": 4.037735849056604,
+      "grad_norm": 0.677670955657959,
+      "learning_rate": 0.0003582191041554236,
+      "loss": 3.3613,
+      "step": 37450
+    },
+    {
+      "epoch": 4.0431266846361185,
+      "grad_norm": 0.6520940661430359,
+      "learning_rate": 0.00035789530491095516,
+      "loss": 3.3495,
+      "step": 37500
+    },
+    {
+      "epoch": 4.048517520215634,
+      "grad_norm": 0.6828324198722839,
+      "learning_rate": 0.00035757150566648676,
+      "loss": 3.3657,
+      "step": 37550
+    },
+    {
+      "epoch": 4.053908355795148,
+      "grad_norm": 0.6632188558578491,
+      "learning_rate": 0.0003572477064220183,
+      "loss": 3.3598,
+      "step": 37600
+    },
+    {
+      "epoch": 4.059299191374663,
+      "grad_norm": 0.6378858685493469,
+      "learning_rate": 0.00035692390717754986,
+      "loss": 3.3738,
+      "step": 37650
+    },
+    {
+      "epoch": 4.064690026954178,
+      "grad_norm": 0.6564689874649048,
+      "learning_rate": 0.00035660010793308147,
+      "loss": 3.3596,
+      "step": 37700
+    },
+    {
+      "epoch": 4.070080862533692,
+      "grad_norm": 0.6671766042709351,
+      "learning_rate": 0.000356276308688613,
+      "loss": 3.3797,
+      "step": 37750
+    },
+    {
+      "epoch": 4.0754716981132075,
+      "grad_norm": 0.7246608138084412,
+      "learning_rate": 0.0003559525094441446,
+      "loss": 3.3623,
+      "step": 37800
+    },
+    {
+      "epoch": 4.080862533692723,
+      "grad_norm": 0.6863735914230347,
+      "learning_rate": 0.00035562871019967617,
+      "loss": 3.3614,
+      "step": 37850
+    },
+    {
+      "epoch": 4.086253369272237,
+      "grad_norm": 0.634773313999176,
+      "learning_rate": 0.0003553049109552078,
+      "loss": 3.3608,
+      "step": 37900
+    },
+    {
+      "epoch": 4.091644204851752,
+      "grad_norm": 0.6256311535835266,
+      "learning_rate": 0.00035498111171073933,
+      "loss": 3.3732,
+      "step": 37950
+    },
+    {
+      "epoch": 4.097035040431267,
+      "grad_norm": 0.6742475628852844,
+      "learning_rate": 0.0003546573124662708,
+      "loss": 3.3645,
+      "step": 38000
+    },
+    {
+      "epoch": 4.097035040431267,
+      "eval_accuracy": 0.37620106260351227,
+      "eval_loss": 3.4507040977478027,
+      "eval_runtime": 185.6288,
+      "eval_samples_per_second": 97.027,
+      "eval_steps_per_second": 6.066,
+      "step": 38000
+    },
+    {
+      "epoch": 4.1024258760107815,
+      "grad_norm": 0.7692692279815674,
+      "learning_rate": 0.0003543335132218025,
+      "loss": 3.368,
+      "step": 38050
+    },
+    {
+      "epoch": 4.107816711590297,
+      "grad_norm": 0.6596481800079346,
+      "learning_rate": 0.000354009713977334,
+      "loss": 3.3734,
+      "step": 38100
+    },
+    {
+      "epoch": 4.113207547169812,
+      "grad_norm": 0.6713324785232544,
+      "learning_rate": 0.00035369239071775493,
+      "loss": 3.3814,
+      "step": 38150
+    },
+    {
+      "epoch": 4.118598382749326,
+      "grad_norm": 0.6394351124763489,
+      "learning_rate": 0.00035336859147328654,
+      "loss": 3.3613,
+      "step": 38200
+    },
+    {
+      "epoch": 4.123989218328841,
+      "grad_norm": 0.643153190612793,
+      "learning_rate": 0.0003530447922288181,
+      "loss": 3.3655,
+      "step": 38250
+    },
+    {
+      "epoch": 4.129380053908355,
+      "grad_norm": 0.6525295972824097,
+      "learning_rate": 0.0003527209929843497,
+      "loss": 3.38,
+      "step": 38300
+    },
+    {
+      "epoch": 4.1347708894878705,
+      "grad_norm": 0.6644986271858215,
+      "learning_rate": 0.00035239719373988124,
+      "loss": 3.356,
+      "step": 38350
+    },
+    {
+      "epoch": 4.140161725067386,
+      "grad_norm": 0.6527798771858215,
+      "learning_rate": 0.0003520733944954128,
+      "loss": 3.3795,
+      "step": 38400
+    },
+    {
+      "epoch": 4.1455525606469,
+      "grad_norm": 0.6304841041564941,
+      "learning_rate": 0.0003517495952509444,
+      "loss": 3.3634,
+      "step": 38450
+    },
+    {
+      "epoch": 4.150943396226415,
+      "grad_norm": 0.6663521528244019,
+      "learning_rate": 0.00035142579600647595,
+      "loss": 3.3738,
+      "step": 38500
+    },
+    {
+      "epoch": 4.15633423180593,
+      "grad_norm": 0.6594198346138,
+      "learning_rate": 0.00035110199676200755,
+      "loss": 3.4016,
+      "step": 38550
+    },
+    {
+      "epoch": 4.1617250673854445,
+      "grad_norm": 0.6243187785148621,
+      "learning_rate": 0.0003507781975175391,
+      "loss": 3.3745,
+      "step": 38600
+    },
+    {
+      "epoch": 4.16711590296496,
+      "grad_norm": 0.6264576315879822,
+      "learning_rate": 0.0003504543982730707,
+      "loss": 3.3872,
+      "step": 38650
+    },
+    {
+      "epoch": 4.172506738544475,
+      "grad_norm": 0.7496191263198853,
+      "learning_rate": 0.00035013059902860226,
+      "loss": 3.3533,
+      "step": 38700
+    },
+    {
+      "epoch": 4.177897574123989,
+      "grad_norm": 0.6559754610061646,
+      "learning_rate": 0.00034980679978413375,
+      "loss": 3.3792,
+      "step": 38750
+    },
+    {
+      "epoch": 4.183288409703504,
+      "grad_norm": 0.6516438722610474,
+      "learning_rate": 0.00034948300053966536,
+      "loss": 3.3822,
+      "step": 38800
+    },
+    {
+      "epoch": 4.188679245283019,
+      "grad_norm": 0.6474061608314514,
+      "learning_rate": 0.0003491592012951969,
+      "loss": 3.3848,
+      "step": 38850
+    },
+    {
+      "epoch": 4.1940700808625335,
+      "grad_norm": 0.6637696623802185,
+      "learning_rate": 0.0003488354020507285,
+      "loss": 3.3919,
+      "step": 38900
+    },
+    {
+      "epoch": 4.199460916442049,
+      "grad_norm": 0.6934957504272461,
+      "learning_rate": 0.00034851160280626007,
+      "loss": 3.3853,
+      "step": 38950
+    },
+    {
+      "epoch": 4.204851752021563,
+      "grad_norm": 0.6696598529815674,
+      "learning_rate": 0.00034818780356179167,
+      "loss": 3.3654,
+      "step": 39000
+    },
+    {
+      "epoch": 4.204851752021563,
+      "eval_accuracy": 0.3770448609302795,
+      "eval_loss": 3.4471023082733154,
+      "eval_runtime": 187.1555,
+      "eval_samples_per_second": 96.235,
+      "eval_steps_per_second": 6.016,
+      "step": 39000
+    },
+    {
+      "epoch": 4.210242587601078,
+      "grad_norm": 0.6707077026367188,
+      "learning_rate": 0.0003478640043173232,
+      "loss": 3.3941,
+      "step": 39050
+    },
+    {
+      "epoch": 4.215633423180593,
+      "grad_norm": 0.6716105937957764,
+      "learning_rate": 0.0003475402050728548,
+      "loss": 3.3861,
+      "step": 39100
+    },
+    {
+      "epoch": 4.2210242587601075,
+      "grad_norm": 0.6311998963356018,
+      "learning_rate": 0.0003472164058283864,
+      "loss": 3.3967,
+      "step": 39150
+    },
+    {
+      "epoch": 4.226415094339623,
+      "grad_norm": 0.6261934041976929,
+      "learning_rate": 0.0003468926065839179,
+      "loss": 3.3782,
+      "step": 39200
+    },
+    {
+      "epoch": 4.231805929919138,
+      "grad_norm": 0.6730643510818481,
+      "learning_rate": 0.00034656880733944953,
+      "loss": 3.3722,
+      "step": 39250
+    },
+    {
+      "epoch": 4.237196765498652,
+      "grad_norm": 0.6469711065292358,
+      "learning_rate": 0.0003462450080949811,
+      "loss": 3.3856,
+      "step": 39300
+    },
+    {
+      "epoch": 4.242587601078167,
+      "grad_norm": 0.6442502737045288,
+      "learning_rate": 0.0003459212088505127,
+      "loss": 3.3742,
+      "step": 39350
+    },
+    {
+      "epoch": 4.247978436657682,
+      "grad_norm": 0.6268736124038696,
+      "learning_rate": 0.00034560388559093364,
+      "loss": 3.3936,
+      "step": 39400
+    },
+    {
+      "epoch": 4.2533692722371965,
+      "grad_norm": 0.6241418719291687,
+      "learning_rate": 0.00034528008634646514,
+      "loss": 3.3806,
+      "step": 39450
+    },
+    {
+      "epoch": 4.258760107816712,
+      "grad_norm": 0.655057430267334,
+      "learning_rate": 0.00034495628710199674,
+      "loss": 3.3872,
+      "step": 39500
+    },
+    {
+      "epoch": 4.264150943396227,
+      "grad_norm": 0.7007227540016174,
+      "learning_rate": 0.0003446324878575283,
+      "loss": 3.3929,
+      "step": 39550
+    },
+    {
+      "epoch": 4.269541778975741,
+      "grad_norm": 0.6439916491508484,
+      "learning_rate": 0.00034430868861305984,
+      "loss": 3.3987,
+      "step": 39600
+    },
+    {
+      "epoch": 4.274932614555256,
+      "grad_norm": 0.6349661350250244,
+      "learning_rate": 0.00034398488936859145,
+      "loss": 3.395,
+      "step": 39650
+    },
+    {
+      "epoch": 4.280323450134771,
+      "grad_norm": 0.7105922698974609,
+      "learning_rate": 0.000343661090124123,
+      "loss": 3.3759,
+      "step": 39700
+    },
+    {
+      "epoch": 4.285714285714286,
+      "grad_norm": 0.7135196924209595,
+      "learning_rate": 0.0003433372908796546,
+      "loss": 3.3831,
+      "step": 39750
+    },
+    {
+      "epoch": 4.291105121293801,
+      "grad_norm": 0.6759126782417297,
+      "learning_rate": 0.00034301349163518615,
+      "loss": 3.3837,
+      "step": 39800
+    },
+    {
+      "epoch": 4.296495956873315,
+      "grad_norm": 0.6720559000968933,
+      "learning_rate": 0.00034268969239071776,
+      "loss": 3.3846,
+      "step": 39850
+    },
+    {
+      "epoch": 4.30188679245283,
+      "grad_norm": 0.7007958889007568,
+      "learning_rate": 0.0003423658931462493,
+      "loss": 3.3911,
+      "step": 39900
+    },
+    {
+      "epoch": 4.307277628032345,
+      "grad_norm": 0.7054687142372131,
+      "learning_rate": 0.00034204209390178086,
+      "loss": 3.3777,
+      "step": 39950
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "grad_norm": 0.6342300772666931,
+      "learning_rate": 0.00034171829465731246,
+      "loss": 3.3894,
+      "step": 40000
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "eval_accuracy": 0.3770942979942351,
+      "eval_loss": 3.4418702125549316,
+      "eval_runtime": 187.7591,
+      "eval_samples_per_second": 95.926,
+      "eval_steps_per_second": 5.997,
+      "step": 40000
+    },
+    {
+      "epoch": 4.318059299191375,
+      "grad_norm": 0.679790198802948,
+      "learning_rate": 0.00034139449541284396,
+      "loss": 3.3882,
+      "step": 40050
+    },
+    {
+      "epoch": 4.32345013477089,
+      "grad_norm": 0.6788046360015869,
+      "learning_rate": 0.00034107069616837556,
+      "loss": 3.386,
+      "step": 40100
+    },
+    {
+      "epoch": 4.328840970350404,
+      "grad_norm": 0.678306519985199,
+      "learning_rate": 0.0003407468969239071,
+      "loss": 3.3971,
+      "step": 40150
+    },
+    {
+      "epoch": 4.334231805929919,
+      "grad_norm": 0.6796455979347229,
+      "learning_rate": 0.0003404230976794387,
+      "loss": 3.3944,
+      "step": 40200
+    },
+    {
+      "epoch": 4.339622641509434,
+      "grad_norm": 0.6383653879165649,
+      "learning_rate": 0.00034009929843497027,
+      "loss": 3.3812,
+      "step": 40250
+    },
+    {
+      "epoch": 4.345013477088949,
+      "grad_norm": 0.6595935821533203,
+      "learning_rate": 0.0003397754991905019,
+      "loss": 3.3932,
+      "step": 40300
+    },
+    {
+      "epoch": 4.350404312668464,
+      "grad_norm": 0.6320056319236755,
+      "learning_rate": 0.0003394516999460334,
+      "loss": 3.3749,
+      "step": 40350
+    },
+    {
+      "epoch": 4.355795148247978,
+      "grad_norm": 0.6796987056732178,
+      "learning_rate": 0.000339127900701565,
+      "loss": 3.3897,
+      "step": 40400
+    },
+    {
+      "epoch": 4.361185983827493,
+      "grad_norm": 0.6579636931419373,
+      "learning_rate": 0.0003388041014570966,
+      "loss": 3.3836,
+      "step": 40450
+    },
+    {
+      "epoch": 4.366576819407008,
+      "grad_norm": 0.6425259709358215,
+      "learning_rate": 0.00033848030221262813,
+      "loss": 3.4008,
+      "step": 40500
+    },
+    {
+      "epoch": 4.3719676549865225,
+      "grad_norm": 0.6290802955627441,
+      "learning_rate": 0.00033815650296815974,
+      "loss": 3.3902,
+      "step": 40550
+    },
+    {
+      "epoch": 4.377358490566038,
+      "grad_norm": 0.7187379002571106,
+      "learning_rate": 0.0003378327037236913,
+      "loss": 3.3809,
+      "step": 40600
+    },
+    {
+      "epoch": 4.382749326145553,
+      "grad_norm": 0.9867428541183472,
+      "learning_rate": 0.0003375089044792229,
+      "loss": 3.3882,
+      "step": 40650
+    },
+    {
+      "epoch": 4.388140161725067,
+      "grad_norm": 0.6685853004455566,
+      "learning_rate": 0.00033718510523475444,
+      "loss": 3.3892,
+      "step": 40700
+    },
+    {
+      "epoch": 4.393530997304582,
+      "grad_norm": 0.6470058560371399,
+      "learning_rate": 0.00033686130599028605,
+      "loss": 3.3874,
+      "step": 40750
+    },
+    {
+      "epoch": 4.398921832884097,
+      "grad_norm": 0.7292867302894592,
+      "learning_rate": 0.00033653750674581754,
+      "loss": 3.3759,
+      "step": 40800
+    },
+    {
+      "epoch": 4.404312668463612,
+      "grad_norm": 0.6573560833930969,
+      "learning_rate": 0.0003362137075013491,
+      "loss": 3.3905,
+      "step": 40850
+    },
+    {
+      "epoch": 4.409703504043127,
+      "grad_norm": 0.6764604449272156,
+      "learning_rate": 0.0003358899082568807,
+      "loss": 3.4025,
+      "step": 40900
+    },
+    {
+      "epoch": 4.415094339622642,
+      "grad_norm": 0.6686357855796814,
+      "learning_rate": 0.00033556610901241225,
+      "loss": 3.3966,
+      "step": 40950
+    },
+    {
+      "epoch": 4.420485175202156,
+      "grad_norm": 0.6409183740615845,
+      "learning_rate": 0.00033524230976794385,
+      "loss": 3.3968,
+      "step": 41000
+    },
+    {
+      "epoch": 4.420485175202156,
+      "eval_accuracy": 0.3780720653316776,
+      "eval_loss": 3.434173107147217,
+      "eval_runtime": 187.7168,
+      "eval_samples_per_second": 95.948,
+      "eval_steps_per_second": 5.998,
+      "step": 41000
+    },
+    {
+      "epoch": 4.425876010781671,
+      "grad_norm": 0.6630507111549377,
+      "learning_rate": 0.0003349185105234754,
+      "loss": 3.4025,
+      "step": 41050
+    },
+    {
+      "epoch": 4.431266846361186,
+      "grad_norm": 0.629347562789917,
+      "learning_rate": 0.000334594711279007,
+      "loss": 3.3709,
+      "step": 41100
+    },
+    {
+      "epoch": 4.436657681940701,
+      "grad_norm": 0.6732147336006165,
+      "learning_rate": 0.00033427091203453856,
+      "loss": 3.3941,
+      "step": 41150
+    },
+    {
+      "epoch": 4.442048517520216,
+      "grad_norm": 0.6474014520645142,
+      "learning_rate": 0.00033394711279007016,
+      "loss": 3.4073,
+      "step": 41200
+    },
+    {
+      "epoch": 4.44743935309973,
+      "grad_norm": 0.6890389919281006,
+      "learning_rate": 0.0003336233135456017,
+      "loss": 3.4038,
+      "step": 41250
+    },
+    {
+      "epoch": 4.452830188679245,
+      "grad_norm": 0.6490774154663086,
+      "learning_rate": 0.00033329951430113326,
+      "loss": 3.3711,
+      "step": 41300
+    },
+    {
+      "epoch": 4.45822102425876,
+      "grad_norm": 0.7015817761421204,
+      "learning_rate": 0.00033297571505666487,
+      "loss": 3.3799,
+      "step": 41350
+    },
+    {
+      "epoch": 4.463611859838275,
+      "grad_norm": 0.6473625898361206,
+      "learning_rate": 0.00033265191581219637,
+      "loss": 3.3996,
+      "step": 41400
+    },
+    {
+      "epoch": 4.46900269541779,
+      "grad_norm": 0.6641436219215393,
+      "learning_rate": 0.00033232811656772797,
+      "loss": 3.4102,
+      "step": 41450
+    },
+    {
+      "epoch": 4.474393530997305,
+      "grad_norm": 0.7184997797012329,
+      "learning_rate": 0.0003320043173232595,
+      "loss": 3.4065,
+      "step": 41500
+    },
+    {
+      "epoch": 4.479784366576819,
+      "grad_norm": 0.6495400667190552,
+      "learning_rate": 0.0003316805180787911,
+      "loss": 3.3939,
+      "step": 41550
+    },
+    {
+      "epoch": 4.485175202156334,
+      "grad_norm": 0.6731253862380981,
+      "learning_rate": 0.0003313567188343227,
+      "loss": 3.3872,
+      "step": 41600
+    },
+    {
+      "epoch": 4.490566037735849,
+      "grad_norm": 0.6487905383110046,
+      "learning_rate": 0.00033103291958985423,
+      "loss": 3.3986,
+      "step": 41650
+    },
+    {
+      "epoch": 4.495956873315364,
+      "grad_norm": 0.6703952550888062,
+      "learning_rate": 0.00033070912034538583,
+      "loss": 3.3889,
+      "step": 41700
+    },
+    {
+      "epoch": 4.501347708894879,
+      "grad_norm": 0.6991889476776123,
+      "learning_rate": 0.0003303853211009174,
+      "loss": 3.3894,
+      "step": 41750
+    },
+    {
+      "epoch": 4.506738544474393,
+      "grad_norm": 0.7930850386619568,
+      "learning_rate": 0.000330061521856449,
+      "loss": 3.4065,
+      "step": 41800
+    },
+    {
+      "epoch": 4.512129380053908,
+      "grad_norm": 0.6711742877960205,
+      "learning_rate": 0.00032973772261198054,
+      "loss": 3.3829,
+      "step": 41850
+    },
+    {
+      "epoch": 4.517520215633423,
+      "grad_norm": 0.6374300718307495,
+      "learning_rate": 0.00032941392336751214,
+      "loss": 3.3968,
+      "step": 41900
+    },
+    {
+      "epoch": 4.5229110512129385,
+      "grad_norm": 0.6754950284957886,
+      "learning_rate": 0.0003290901241230437,
+      "loss": 3.4006,
+      "step": 41950
+    },
+    {
+      "epoch": 4.528301886792453,
+      "grad_norm": 0.6714204549789429,
+      "learning_rate": 0.0003287663248785753,
+      "loss": 3.3864,
+      "step": 42000
+    },
+    {
+      "epoch": 4.528301886792453,
+      "eval_accuracy": 0.3787578823595626,
+      "eval_loss": 3.429198741912842,
+      "eval_runtime": 186.0156,
+      "eval_samples_per_second": 96.825,
+      "eval_steps_per_second": 6.053,
+      "step": 42000
+    },
+    {
+      "epoch": 4.533692722371968,
+      "grad_norm": 0.6874658465385437,
+      "learning_rate": 0.00032844252563410685,
+      "loss": 3.3858,
+      "step": 42050
+    },
+    {
+      "epoch": 4.539083557951482,
+      "grad_norm": 0.695583164691925,
+      "learning_rate": 0.00032811872638963834,
+      "loss": 3.3934,
+      "step": 42100
+    },
+    {
+      "epoch": 4.544474393530997,
+      "grad_norm": 0.6853567361831665,
+      "learning_rate": 0.00032779492714516995,
+      "loss": 3.3937,
+      "step": 42150
+    },
+    {
+      "epoch": 4.549865229110512,
+      "grad_norm": 0.6582930088043213,
+      "learning_rate": 0.0003274711279007015,
+      "loss": 3.3921,
+      "step": 42200
+    },
+    {
+      "epoch": 4.555256064690027,
+      "grad_norm": 0.6429412961006165,
+      "learning_rate": 0.0003271473286562331,
+      "loss": 3.4041,
+      "step": 42250
+    },
+    {
+      "epoch": 4.560646900269542,
+      "grad_norm": 0.7131776213645935,
+      "learning_rate": 0.00032682352941176466,
+      "loss": 3.381,
+      "step": 42300
+    },
+    {
+      "epoch": 4.566037735849057,
+      "grad_norm": 0.8906499147415161,
+      "learning_rate": 0.00032649973016729626,
+      "loss": 3.4145,
+      "step": 42350
+    },
+    {
+      "epoch": 4.571428571428571,
+      "grad_norm": 0.6330589056015015,
+      "learning_rate": 0.0003261759309228278,
+      "loss": 3.3746,
+      "step": 42400
+    },
+    {
+      "epoch": 4.576819407008086,
+      "grad_norm": 0.6902533769607544,
+      "learning_rate": 0.0003258521316783594,
+      "loss": 3.3965,
+      "step": 42450
+    },
+    {
+      "epoch": 4.5822102425876015,
+      "grad_norm": 0.6204718947410583,
+      "learning_rate": 0.00032552833243389097,
+      "loss": 3.4021,
+      "step": 42500
+    },
+    {
+      "epoch": 4.587601078167116,
+      "grad_norm": 0.6694790124893188,
+      "learning_rate": 0.0003252045331894225,
+      "loss": 3.3811,
+      "step": 42550
+    },
+    {
+      "epoch": 4.592991913746631,
+      "grad_norm": 0.6045516133308411,
+      "learning_rate": 0.0003248807339449541,
+      "loss": 3.4073,
+      "step": 42600
+    },
+    {
+      "epoch": 4.598382749326145,
+      "grad_norm": 0.6847394108772278,
+      "learning_rate": 0.00032455693470048567,
+      "loss": 3.3929,
+      "step": 42650
+    },
+    {
+      "epoch": 4.60377358490566,
+      "grad_norm": 0.6484902501106262,
+      "learning_rate": 0.0003242331354560173,
+      "loss": 3.3908,
+      "step": 42700
+    },
+    {
+      "epoch": 4.609164420485175,
+      "grad_norm": 0.6607882976531982,
+      "learning_rate": 0.0003239093362115488,
+      "loss": 3.3996,
+      "step": 42750
+    },
+    {
+      "epoch": 4.6145552560646905,
+      "grad_norm": 0.6574293971061707,
+      "learning_rate": 0.00032358553696708043,
+      "loss": 3.3758,
+      "step": 42800
+    },
+    {
+      "epoch": 4.619946091644205,
+      "grad_norm": 0.6498639583587646,
+      "learning_rate": 0.00032326173772261193,
+      "loss": 3.3994,
+      "step": 42850
+    },
+    {
+      "epoch": 4.62533692722372,
+      "grad_norm": 0.6482625603675842,
+      "learning_rate": 0.00032293793847814353,
+      "loss": 3.3967,
+      "step": 42900
+    },
+    {
+      "epoch": 4.630727762803234,
+      "grad_norm": 0.7077571153640747,
+      "learning_rate": 0.0003226141392336751,
+      "loss": 3.3939,
+      "step": 42950
+    },
+    {
+      "epoch": 4.636118598382749,
+      "grad_norm": 0.7394276857376099,
+      "learning_rate": 0.00032229033998920663,
+      "loss": 3.3964,
+      "step": 43000
+    },
+    {
+      "epoch": 4.636118598382749,
+      "eval_accuracy": 0.378906410857205,
+      "eval_loss": 3.4243056774139404,
+      "eval_runtime": 185.5995,
+      "eval_samples_per_second": 97.042,
+      "eval_steps_per_second": 6.067,
+      "step": 43000
+    },
+    {
+      "epoch": 4.6415094339622645,
+      "grad_norm": 0.684695839881897,
+      "learning_rate": 0.00032196654074473824,
+      "loss": 3.3885,
+      "step": 43050
+    },
+    {
+      "epoch": 4.646900269541779,
+      "grad_norm": 0.6889744400978088,
+      "learning_rate": 0.0003216427415002698,
+      "loss": 3.3914,
+      "step": 43100
+    },
+    {
+      "epoch": 4.652291105121294,
+      "grad_norm": 0.7006178498268127,
+      "learning_rate": 0.0003213189422558014,
+      "loss": 3.4066,
+      "step": 43150
+    },
+    {
+      "epoch": 4.657681940700809,
+      "grad_norm": 0.662488579750061,
+      "learning_rate": 0.00032100161899622235,
+      "loss": 3.3784,
+      "step": 43200
+    },
+    {
+      "epoch": 4.663072776280323,
+      "grad_norm": 0.6467827558517456,
+      "learning_rate": 0.0003206778197517539,
+      "loss": 3.3972,
+      "step": 43250
+    },
+    {
+      "epoch": 4.668463611859838,
+      "grad_norm": 0.6481455564498901,
+      "learning_rate": 0.00032035402050728545,
+      "loss": 3.3843,
+      "step": 43300
+    },
+    {
+      "epoch": 4.6738544474393535,
+      "grad_norm": 0.7185537815093994,
+      "learning_rate": 0.00032003022126281705,
+      "loss": 3.3911,
+      "step": 43350
+    },
+    {
+      "epoch": 4.679245283018868,
+      "grad_norm": 0.6544119715690613,
+      "learning_rate": 0.00031970642201834855,
+      "loss": 3.3893,
+      "step": 43400
+    },
+    {
+      "epoch": 4.684636118598383,
+      "grad_norm": 0.6571384072303772,
+      "learning_rate": 0.0003193826227738802,
+      "loss": 3.3986,
+      "step": 43450
+    },
+    {
+      "epoch": 4.690026954177897,
+      "grad_norm": 0.6685782074928284,
+      "learning_rate": 0.0003190588235294117,
+      "loss": 3.3764,
+      "step": 43500
+    },
+    {
+      "epoch": 4.695417789757412,
+      "grad_norm": 0.6582465767860413,
+      "learning_rate": 0.0003187350242849433,
+      "loss": 3.393,
+      "step": 43550
+    },
+    {
+      "epoch": 4.7008086253369274,
+      "grad_norm": 0.6975694894790649,
+      "learning_rate": 0.00031841122504047486,
+      "loss": 3.399,
+      "step": 43600
+    },
+    {
+      "epoch": 4.706199460916442,
+      "grad_norm": 0.7016478180885315,
+      "learning_rate": 0.00031808742579600646,
+      "loss": 3.3951,
+      "step": 43650
+    },
+    {
+      "epoch": 4.711590296495957,
+      "grad_norm": 0.6719104051589966,
+      "learning_rate": 0.000317763626551538,
+      "loss": 3.393,
+      "step": 43700
+    },
+    {
+      "epoch": 4.716981132075472,
+      "grad_norm": 0.668164849281311,
+      "learning_rate": 0.00031743982730706957,
+      "loss": 3.4084,
+      "step": 43750
+    },
+    {
+      "epoch": 4.722371967654986,
+      "grad_norm": 0.7527414560317993,
+      "learning_rate": 0.00031711602806260117,
+      "loss": 3.3937,
+      "step": 43800
+    },
+    {
+      "epoch": 4.727762803234501,
+      "grad_norm": 0.7288227677345276,
+      "learning_rate": 0.0003167922288181327,
+      "loss": 3.4009,
+      "step": 43850
+    },
+    {
+      "epoch": 4.7331536388140165,
+      "grad_norm": 0.6558834314346313,
+      "learning_rate": 0.0003164684295736643,
+      "loss": 3.3885,
+      "step": 43900
+    },
+    {
+      "epoch": 4.738544474393531,
+      "grad_norm": 0.6408897638320923,
+      "learning_rate": 0.0003161446303291959,
+      "loss": 3.3892,
+      "step": 43950
+    },
+    {
+      "epoch": 4.743935309973046,
+      "grad_norm": 0.6316078305244446,
+      "learning_rate": 0.0003158208310847275,
+      "loss": 3.3894,
+      "step": 44000
+    },
+    {
+      "epoch": 4.743935309973046,
+      "eval_accuracy": 0.3792912593858439,
+      "eval_loss": 3.419440269470215,
+      "eval_runtime": 185.7744,
+      "eval_samples_per_second": 96.951,
+      "eval_steps_per_second": 6.061,
+      "step": 44000
+    },
+    {
+      "epoch": 4.74932614555256,
+      "grad_norm": 0.6755422949790955,
+      "learning_rate": 0.00031549703184025903,
+      "loss": 3.3876,
+      "step": 44050
+    },
+    {
+      "epoch": 4.754716981132075,
+      "grad_norm": 0.6657536029815674,
+      "learning_rate": 0.00031517323259579064,
+      "loss": 3.3955,
+      "step": 44100
+    },
+    {
+      "epoch": 4.7601078167115904,
+      "grad_norm": 0.7551212310791016,
+      "learning_rate": 0.00031484943335132213,
+      "loss": 3.3864,
+      "step": 44150
+    },
+    {
+      "epoch": 4.765498652291106,
+      "grad_norm": 0.7116191387176514,
+      "learning_rate": 0.0003145256341068537,
+      "loss": 3.3911,
+      "step": 44200
+    },
+    {
+      "epoch": 4.77088948787062,
+      "grad_norm": 0.7062914967536926,
+      "learning_rate": 0.0003142018348623853,
+      "loss": 3.3765,
+      "step": 44250
+    },
+    {
+      "epoch": 4.776280323450135,
+      "grad_norm": 0.7071515917778015,
+      "learning_rate": 0.00031387803561791684,
+      "loss": 3.3984,
+      "step": 44300
+    },
+    {
+      "epoch": 4.781671159029649,
+      "grad_norm": 0.6906786561012268,
+      "learning_rate": 0.00031355423637344844,
+      "loss": 3.378,
+      "step": 44350
+    },
+    {
+      "epoch": 4.787061994609164,
+      "grad_norm": 0.6172106266021729,
+      "learning_rate": 0.00031323043712898,
+      "loss": 3.42,
+      "step": 44400
+    },
+    {
+      "epoch": 4.7924528301886795,
+      "grad_norm": 0.6594890356063843,
+      "learning_rate": 0.0003129066378845116,
+      "loss": 3.3755,
+      "step": 44450
+    },
+    {
+      "epoch": 4.797843665768194,
+      "grad_norm": 0.6606116890907288,
+      "learning_rate": 0.00031258283864004315,
+      "loss": 3.3898,
+      "step": 44500
+    },
+    {
+      "epoch": 4.803234501347709,
+      "grad_norm": 0.6465703845024109,
+      "learning_rate": 0.0003122590393955747,
+      "loss": 3.3827,
+      "step": 44550
+    },
+    {
+      "epoch": 4.808625336927224,
+      "grad_norm": 0.6434739828109741,
+      "learning_rate": 0.0003119352401511063,
+      "loss": 3.4002,
+      "step": 44600
+    },
+    {
+      "epoch": 4.814016172506738,
+      "grad_norm": 0.7177310585975647,
+      "learning_rate": 0.00031161144090663786,
+      "loss": 3.3819,
+      "step": 44650
+    },
+    {
+      "epoch": 4.819407008086253,
+      "grad_norm": 0.651936411857605,
+      "learning_rate": 0.00031128764166216946,
+      "loss": 3.3966,
+      "step": 44700
+    },
+    {
+      "epoch": 4.824797843665769,
+      "grad_norm": 0.7027115225791931,
+      "learning_rate": 0.00031096384241770096,
+      "loss": 3.3892,
+      "step": 44750
+    },
+    {
+      "epoch": 4.830188679245283,
+      "grad_norm": 0.6689330339431763,
+      "learning_rate": 0.0003106400431732326,
+      "loss": 3.388,
+      "step": 44800
+    },
+    {
+      "epoch": 4.835579514824798,
+      "grad_norm": 0.7061042189598083,
+      "learning_rate": 0.0003103162439287641,
+      "loss": 3.3915,
+      "step": 44850
+    },
+    {
+      "epoch": 4.840970350404312,
+      "grad_norm": 0.6312277317047119,
+      "learning_rate": 0.0003099924446842957,
+      "loss": 3.3908,
+      "step": 44900
+    },
+    {
+      "epoch": 4.846361185983827,
+      "grad_norm": 0.651349663734436,
+      "learning_rate": 0.00030966864543982727,
+      "loss": 3.3831,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "grad_norm": 0.6578131914138794,
+      "learning_rate": 0.0003093448461953588,
+      "loss": 3.406,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "eval_accuracy": 0.38000423963568253,
+      "eval_loss": 3.4150288105010986,
+      "eval_runtime": 185.3702,
+      "eval_samples_per_second": 97.162,
+      "eval_steps_per_second": 6.074,
+      "step": 45000
+    },
+    {
+      "epoch": 4.857142857142857,
+      "grad_norm": 0.7515460848808289,
+      "learning_rate": 0.0003090210469508904,
+      "loss": 3.3778,
+      "step": 45050
+    },
+    {
+      "epoch": 4.862533692722372,
+      "grad_norm": 0.6967692375183105,
+      "learning_rate": 0.00030869724770642197,
+      "loss": 3.3817,
+      "step": 45100
+    },
+    {
+      "epoch": 4.867924528301887,
+      "grad_norm": 0.6564337015151978,
+      "learning_rate": 0.0003083734484619536,
+      "loss": 3.3934,
+      "step": 45150
+    },
+    {
+      "epoch": 4.873315363881401,
+      "grad_norm": 0.6927236914634705,
+      "learning_rate": 0.00030804964921748513,
+      "loss": 3.3789,
+      "step": 45200
+    },
+    {
+      "epoch": 4.878706199460916,
+      "grad_norm": 0.700776994228363,
+      "learning_rate": 0.00030772584997301673,
+      "loss": 3.3935,
+      "step": 45250
+    },
+    {
+      "epoch": 4.884097035040432,
+      "grad_norm": 0.662986695766449,
+      "learning_rate": 0.0003074020507285483,
+      "loss": 3.3799,
+      "step": 45300
+    },
+    {
+      "epoch": 4.889487870619946,
+      "grad_norm": 0.6615574955940247,
+      "learning_rate": 0.0003070782514840799,
+      "loss": 3.3895,
+      "step": 45350
+    },
+    {
+      "epoch": 4.894878706199461,
+      "grad_norm": 0.662390410900116,
+      "learning_rate": 0.00030675445223961144,
+      "loss": 3.3858,
+      "step": 45400
+    },
+    {
+      "epoch": 4.900269541778976,
+      "grad_norm": 0.6884881854057312,
+      "learning_rate": 0.00030643065299514294,
+      "loss": 3.3945,
+      "step": 45450
+    },
+    {
+      "epoch": 4.90566037735849,
+      "grad_norm": 0.6707163453102112,
+      "learning_rate": 0.00030610685375067454,
+      "loss": 3.3948,
+      "step": 45500
+    },
+    {
+      "epoch": 4.9110512129380055,
+      "grad_norm": 0.6818541288375854,
+      "learning_rate": 0.0003057830545062061,
+      "loss": 3.3949,
+      "step": 45550
+    },
+    {
+      "epoch": 4.916442048517521,
+      "grad_norm": 0.6906265616416931,
+      "learning_rate": 0.0003054592552617377,
+      "loss": 3.393,
+      "step": 45600
+    },
+    {
+      "epoch": 4.921832884097035,
+      "grad_norm": 0.6746832132339478,
+      "learning_rate": 0.00030513545601726925,
+      "loss": 3.4084,
+      "step": 45650
+    },
+    {
+      "epoch": 4.92722371967655,
+      "grad_norm": 0.6849566102027893,
+      "learning_rate": 0.00030481165677280085,
+      "loss": 3.4034,
+      "step": 45700
+    },
+    {
+      "epoch": 4.932614555256064,
+      "grad_norm": 0.7006532549858093,
+      "learning_rate": 0.0003044878575283324,
+      "loss": 3.3839,
+      "step": 45750
+    },
+    {
+      "epoch": 4.938005390835579,
+      "grad_norm": 0.6992759108543396,
+      "learning_rate": 0.000304164058283864,
+      "loss": 3.4078,
+      "step": 45800
+    },
+    {
+      "epoch": 4.943396226415095,
+      "grad_norm": 0.666161835193634,
+      "learning_rate": 0.00030384025903939556,
+      "loss": 3.3785,
+      "step": 45850
+    },
+    {
+      "epoch": 4.948787061994609,
+      "grad_norm": 0.6996541619300842,
+      "learning_rate": 0.0003035164597949271,
+      "loss": 3.3743,
+      "step": 45900
+    },
+    {
+      "epoch": 4.954177897574124,
+      "grad_norm": 0.69724041223526,
+      "learning_rate": 0.0003031926605504587,
+      "loss": 3.3755,
+      "step": 45950
+    },
+    {
+      "epoch": 4.959568733153639,
+      "grad_norm": 0.6941049098968506,
+      "learning_rate": 0.00030286886130599026,
+      "loss": 3.3737,
+      "step": 46000
+    },
+    {
+      "epoch": 4.959568733153639,
+      "eval_accuracy": 0.3803529067526792,
+      "eval_loss": 3.410797595977783,
+      "eval_runtime": 185.7942,
+      "eval_samples_per_second": 96.941,
+      "eval_steps_per_second": 6.06,
+      "step": 46000
+    },
+    {
+      "epoch": 4.964959568733153,
+      "grad_norm": 0.6991933584213257,
+      "learning_rate": 0.00030254506206152187,
+      "loss": 3.3807,
+      "step": 46050
+    },
+    {
+      "epoch": 4.9703504043126685,
+      "grad_norm": 0.6805256009101868,
+      "learning_rate": 0.00030222126281705336,
+      "loss": 3.3908,
+      "step": 46100
+    },
+    {
+      "epoch": 4.975741239892184,
+      "grad_norm": 0.6762968301773071,
+      "learning_rate": 0.000301897463572585,
+      "loss": 3.3899,
+      "step": 46150
+    },
+    {
+      "epoch": 4.981132075471698,
+      "grad_norm": 0.6691407561302185,
+      "learning_rate": 0.0003015736643281165,
+      "loss": 3.3947,
+      "step": 46200
+    },
+    {
+      "epoch": 4.986522911051213,
+      "grad_norm": 0.6724907755851746,
+      "learning_rate": 0.00030124986508364807,
+      "loss": 3.3932,
+      "step": 46250
+    },
+    {
+      "epoch": 4.991913746630727,
+      "grad_norm": 0.6928101181983948,
+      "learning_rate": 0.0003009260658391797,
+      "loss": 3.3862,
+      "step": 46300
+    },
+    {
+      "epoch": 4.997304582210242,
+      "grad_norm": 0.7020788192749023,
+      "learning_rate": 0.0003006022665947112,
+      "loss": 3.3797,
+      "step": 46350
+    },
+    {
+      "epoch": 5.002695417789758,
+      "grad_norm": 0.7493858337402344,
+      "learning_rate": 0.00030027846735024283,
+      "loss": 3.361,
+      "step": 46400
+    },
+    {
+      "epoch": 5.008086253369272,
+      "grad_norm": 0.6816010475158691,
+      "learning_rate": 0.00029995466810577443,
+      "loss": 3.2895,
+      "step": 46450
+    },
+    {
+      "epoch": 5.013477088948787,
+      "grad_norm": 0.6867018938064575,
+      "learning_rate": 0.00029963086886130593,
+      "loss": 3.2951,
+      "step": 46500
+    },
+    {
+      "epoch": 5.018867924528302,
+      "grad_norm": 0.6954084038734436,
+      "learning_rate": 0.00029930706961683754,
+      "loss": 3.308,
+      "step": 46550
+    },
+    {
+      "epoch": 5.024258760107816,
+      "grad_norm": 0.7072650790214539,
+      "learning_rate": 0.0002989832703723691,
+      "loss": 3.2996,
+      "step": 46600
+    },
+    {
+      "epoch": 5.0296495956873315,
+      "grad_norm": 0.7235918641090393,
+      "learning_rate": 0.0002986594711279007,
+      "loss": 3.2882,
+      "step": 46650
+    },
+    {
+      "epoch": 5.035040431266847,
+      "grad_norm": 0.6916717886924744,
+      "learning_rate": 0.00029833567188343224,
+      "loss": 3.2987,
+      "step": 46700
+    },
+    {
+      "epoch": 5.040431266846361,
+      "grad_norm": 0.6355169415473938,
+      "learning_rate": 0.00029801187263896385,
+      "loss": 3.2883,
+      "step": 46750
+    },
+    {
+      "epoch": 5.045822102425876,
+      "grad_norm": 0.7969847321510315,
+      "learning_rate": 0.0002976880733944954,
+      "loss": 3.2979,
+      "step": 46800
+    },
+    {
+      "epoch": 5.051212938005391,
+      "grad_norm": 0.695892333984375,
+      "learning_rate": 0.00029736427415002695,
+      "loss": 3.3124,
+      "step": 46850
+    },
+    {
+      "epoch": 5.056603773584905,
+      "grad_norm": 0.6695360541343689,
+      "learning_rate": 0.0002970404749055585,
+      "loss": 3.3185,
+      "step": 46900
+    },
+    {
+      "epoch": 5.061994609164421,
+      "grad_norm": 0.693292498588562,
+      "learning_rate": 0.0002967166756610901,
+      "loss": 3.3083,
+      "step": 46950
+    },
+    {
+      "epoch": 5.067385444743936,
+      "grad_norm": 0.7053436636924744,
+      "learning_rate": 0.00029639287641662165,
+      "loss": 3.317,
+      "step": 47000
+    },
+    {
+      "epoch": 5.067385444743936,
+      "eval_accuracy": 0.38064387918624637,
+      "eval_loss": 3.4121837615966797,
+      "eval_runtime": 185.6346,
+      "eval_samples_per_second": 97.024,
+      "eval_steps_per_second": 6.066,
+      "step": 47000
+    },
+    {
+      "epoch": 5.07277628032345,
+      "grad_norm": 0.6880952715873718,
+      "learning_rate": 0.00029606907717215326,
+      "loss": 3.2865,
+      "step": 47050
+    },
+    {
+      "epoch": 5.078167115902965,
+      "grad_norm": 0.6662627458572388,
+      "learning_rate": 0.0002957517539125742,
+      "loss": 3.3008,
+      "step": 47100
+    },
+    {
+      "epoch": 5.083557951482479,
+      "grad_norm": 0.7271195650100708,
+      "learning_rate": 0.00029542795466810576,
+      "loss": 3.325,
+      "step": 47150
+    },
+    {
+      "epoch": 5.0889487870619945,
+      "grad_norm": 0.7169229984283447,
+      "learning_rate": 0.0002951041554236373,
+      "loss": 3.3205,
+      "step": 47200
+    },
+    {
+      "epoch": 5.09433962264151,
+      "grad_norm": 0.7308028340339661,
+      "learning_rate": 0.00029478035617916886,
+      "loss": 3.3016,
+      "step": 47250
+    },
+    {
+      "epoch": 5.099730458221024,
+      "grad_norm": 0.6794512867927551,
+      "learning_rate": 0.00029445655693470047,
+      "loss": 3.3138,
+      "step": 47300
+    },
+    {
+      "epoch": 5.105121293800539,
+      "grad_norm": 0.724534809589386,
+      "learning_rate": 0.000294132757690232,
+      "loss": 3.3161,
+      "step": 47350
+    },
+    {
+      "epoch": 5.110512129380054,
+      "grad_norm": 0.6985588669776917,
+      "learning_rate": 0.0002938089584457636,
+      "loss": 3.3111,
+      "step": 47400
+    },
+    {
+      "epoch": 5.115902964959568,
+      "grad_norm": 0.7239158749580383,
+      "learning_rate": 0.00029348515920129517,
+      "loss": 3.3189,
+      "step": 47450
+    },
+    {
+      "epoch": 5.121293800539084,
+      "grad_norm": 0.6728734970092773,
+      "learning_rate": 0.0002931613599568267,
+      "loss": 3.3121,
+      "step": 47500
+    },
+    {
+      "epoch": 5.126684636118599,
+      "grad_norm": 0.6981966495513916,
+      "learning_rate": 0.00029283756071235833,
+      "loss": 3.3092,
+      "step": 47550
+    },
+    {
+      "epoch": 5.132075471698113,
+      "grad_norm": 0.7157015204429626,
+      "learning_rate": 0.0002925137614678899,
+      "loss": 3.3069,
+      "step": 47600
+    },
+    {
+      "epoch": 5.137466307277628,
+      "grad_norm": 0.6674967408180237,
+      "learning_rate": 0.00029218996222342143,
+      "loss": 3.3195,
+      "step": 47650
+    },
+    {
+      "epoch": 5.142857142857143,
+      "grad_norm": 0.6846231818199158,
+      "learning_rate": 0.00029186616297895303,
+      "loss": 3.322,
+      "step": 47700
+    },
+    {
+      "epoch": 5.1482479784366575,
+      "grad_norm": 0.6753125190734863,
+      "learning_rate": 0.0002915423637344846,
+      "loss": 3.3231,
+      "step": 47750
+    },
+    {
+      "epoch": 5.153638814016173,
+      "grad_norm": 0.7175103425979614,
+      "learning_rate": 0.00029121856449001613,
+      "loss": 3.333,
+      "step": 47800
+    },
+    {
+      "epoch": 5.159029649595688,
+      "grad_norm": 0.7339282035827637,
+      "learning_rate": 0.00029089476524554774,
+      "loss": 3.3232,
+      "step": 47850
+    },
+    {
+      "epoch": 5.164420485175202,
+      "grad_norm": 0.7188692688941956,
+      "learning_rate": 0.0002905709660010793,
+      "loss": 3.3076,
+      "step": 47900
+    },
+    {
+      "epoch": 5.169811320754717,
+      "grad_norm": 0.6953505873680115,
+      "learning_rate": 0.0002902471667566109,
+      "loss": 3.3244,
+      "step": 47950
+    },
+    {
+      "epoch": 5.175202156334231,
+      "grad_norm": 0.7171928882598877,
+      "learning_rate": 0.00028992336751214245,
+      "loss": 3.3154,
+      "step": 48000
+    },
+    {
+      "epoch": 5.175202156334231,
+      "eval_accuracy": 0.3810895733320614,
+      "eval_loss": 3.412214756011963,
+      "eval_runtime": 185.5686,
+      "eval_samples_per_second": 97.058,
+      "eval_steps_per_second": 6.068,
+      "step": 48000
+    },
+    {
+      "epoch": 5.180592991913747,
+      "grad_norm": 0.717573881149292,
+      "learning_rate": 0.00028959956826767405,
+      "loss": 3.3453,
+      "step": 48050
+    },
+    {
+      "epoch": 5.185983827493262,
+      "grad_norm": 0.7013468742370605,
+      "learning_rate": 0.00028927576902320555,
+      "loss": 3.3245,
+      "step": 48100
+    },
+    {
+      "epoch": 5.191374663072776,
+      "grad_norm": 0.6565865874290466,
+      "learning_rate": 0.00028895196977873715,
+      "loss": 3.3314,
+      "step": 48150
+    },
+    {
+      "epoch": 5.196765498652291,
+      "grad_norm": 0.7255452871322632,
+      "learning_rate": 0.0002886281705342687,
+      "loss": 3.3234,
+      "step": 48200
+    },
+    {
+      "epoch": 5.202156334231806,
+      "grad_norm": 0.7184955477714539,
+      "learning_rate": 0.0002883043712898003,
+      "loss": 3.3049,
+      "step": 48250
+    },
+    {
+      "epoch": 5.2075471698113205,
+      "grad_norm": 0.6771543025970459,
+      "learning_rate": 0.00028798057204533186,
+      "loss": 3.3322,
+      "step": 48300
+    },
+    {
+      "epoch": 5.212938005390836,
+      "grad_norm": 0.6928405165672302,
+      "learning_rate": 0.00028765677280086346,
+      "loss": 3.3166,
+      "step": 48350
+    },
+    {
+      "epoch": 5.218328840970351,
+      "grad_norm": 0.6911975145339966,
+      "learning_rate": 0.000287332973556395,
+      "loss": 3.3296,
+      "step": 48400
+    },
+    {
+      "epoch": 5.223719676549865,
+      "grad_norm": 0.7407052516937256,
+      "learning_rate": 0.0002870091743119266,
+      "loss": 3.3193,
+      "step": 48450
+    },
+    {
+      "epoch": 5.22911051212938,
+      "grad_norm": 0.7343896627426147,
+      "learning_rate": 0.00028668537506745817,
+      "loss": 3.3218,
+      "step": 48500
+    },
+    {
+      "epoch": 5.234501347708895,
+      "grad_norm": 0.7096381783485413,
+      "learning_rate": 0.0002863615758229897,
+      "loss": 3.3044,
+      "step": 48550
+    },
+    {
+      "epoch": 5.2398921832884096,
+      "grad_norm": 0.6762372255325317,
+      "learning_rate": 0.00028603777657852127,
+      "loss": 3.3304,
+      "step": 48600
+    },
+    {
+      "epoch": 5.245283018867925,
+      "grad_norm": 0.7038601636886597,
+      "learning_rate": 0.0002857139773340529,
+      "loss": 3.3241,
+      "step": 48650
+    },
+    {
+      "epoch": 5.250673854447439,
+      "grad_norm": 0.6746121048927307,
+      "learning_rate": 0.0002853901780895844,
+      "loss": 3.3214,
+      "step": 48700
+    },
+    {
+      "epoch": 5.256064690026954,
+      "grad_norm": 0.701580286026001,
+      "learning_rate": 0.00028506637884511603,
+      "loss": 3.3133,
+      "step": 48750
+    },
+    {
+      "epoch": 5.261455525606469,
+      "grad_norm": 0.6735265851020813,
+      "learning_rate": 0.0002847425796006476,
+      "loss": 3.3276,
+      "step": 48800
+    },
+    {
+      "epoch": 5.2668463611859835,
+      "grad_norm": 0.7302528619766235,
+      "learning_rate": 0.00028441878035617913,
+      "loss": 3.3404,
+      "step": 48850
+    },
+    {
+      "epoch": 5.272237196765499,
+      "grad_norm": 0.6821619868278503,
+      "learning_rate": 0.00028409498111171073,
+      "loss": 3.3191,
+      "step": 48900
+    },
+    {
+      "epoch": 5.277628032345014,
+      "grad_norm": 0.7001194953918457,
+      "learning_rate": 0.0002837711818672423,
+      "loss": 3.3354,
+      "step": 48950
+    },
+    {
+      "epoch": 5.283018867924528,
+      "grad_norm": 0.672749400138855,
+      "learning_rate": 0.00028344738262277384,
+      "loss": 3.346,
+      "step": 49000
+    },
+    {
+      "epoch": 5.283018867924528,
+      "eval_accuracy": 0.38151875223892856,
+      "eval_loss": 3.40685772895813,
+      "eval_runtime": 185.6212,
+      "eval_samples_per_second": 97.031,
+      "eval_steps_per_second": 6.066,
+      "step": 49000
+    },
+    {
+      "epoch": 5.288409703504043,
+      "grad_norm": 0.7032443284988403,
+      "learning_rate": 0.00028312358337830544,
+      "loss": 3.338,
+      "step": 49050
+    },
+    {
+      "epoch": 5.293800539083558,
+      "grad_norm": 0.7204844355583191,
+      "learning_rate": 0.000282799784133837,
+      "loss": 3.3394,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2991913746630726,
+      "grad_norm": 0.6861972808837891,
+      "learning_rate": 0.00028247598488936854,
+      "loss": 3.3154,
+      "step": 49150
+    },
+    {
+      "epoch": 5.304582210242588,
+      "grad_norm": 0.6550790071487427,
+      "learning_rate": 0.00028215218564490015,
+      "loss": 3.3412,
+      "step": 49200
+    },
+    {
+      "epoch": 5.309973045822103,
+      "grad_norm": 0.6890367865562439,
+      "learning_rate": 0.0002818283864004317,
+      "loss": 3.3276,
+      "step": 49250
+    },
+    {
+      "epoch": 5.315363881401617,
+      "grad_norm": 0.6811584830284119,
+      "learning_rate": 0.0002815045871559633,
+      "loss": 3.3252,
+      "step": 49300
+    },
+    {
+      "epoch": 5.320754716981132,
+      "grad_norm": 0.7011695504188538,
+      "learning_rate": 0.00028118078791149485,
+      "loss": 3.3412,
+      "step": 49350
+    },
+    {
+      "epoch": 5.3261455525606465,
+      "grad_norm": 0.6732707023620605,
+      "learning_rate": 0.0002808569886670264,
+      "loss": 3.3157,
+      "step": 49400
+    },
+    {
+      "epoch": 5.331536388140162,
+      "grad_norm": 0.6839603781700134,
+      "learning_rate": 0.00028053318942255795,
+      "loss": 3.3161,
+      "step": 49450
+    },
+    {
+      "epoch": 5.336927223719677,
+      "grad_norm": 0.7026116251945496,
+      "learning_rate": 0.00028020939017808956,
+      "loss": 3.3436,
+      "step": 49500
+    },
+    {
+      "epoch": 5.342318059299191,
+      "grad_norm": 0.7285355925559998,
+      "learning_rate": 0.0002798855909336211,
+      "loss": 3.3428,
+      "step": 49550
+    },
+    {
+      "epoch": 5.347708894878706,
+      "grad_norm": 0.7406883239746094,
+      "learning_rate": 0.0002795617916891527,
+      "loss": 3.3342,
+      "step": 49600
+    },
+    {
+      "epoch": 5.353099730458221,
+      "grad_norm": 0.7168139815330505,
+      "learning_rate": 0.00027923799244468426,
+      "loss": 3.3231,
+      "step": 49650
+    },
+    {
+      "epoch": 5.3584905660377355,
+      "grad_norm": 0.7335942387580872,
+      "learning_rate": 0.00027891419320021587,
+      "loss": 3.3497,
+      "step": 49700
+    },
+    {
+      "epoch": 5.363881401617251,
+      "grad_norm": 0.7040148377418518,
+      "learning_rate": 0.0002785903939557474,
+      "loss": 3.3466,
+      "step": 49750
+    },
+    {
+      "epoch": 5.369272237196766,
+      "grad_norm": 0.6648421287536621,
+      "learning_rate": 0.00027826659471127897,
+      "loss": 3.3205,
+      "step": 49800
+    },
+    {
+      "epoch": 5.37466307277628,
+      "grad_norm": 0.7032297849655151,
+      "learning_rate": 0.0002779427954668105,
+      "loss": 3.3508,
+      "step": 49850
+    },
+    {
+      "epoch": 5.380053908355795,
+      "grad_norm": 0.6933472156524658,
+      "learning_rate": 0.0002776189962223421,
+      "loss": 3.3301,
+      "step": 49900
+    },
+    {
+      "epoch": 5.38544474393531,
+      "grad_norm": 0.6965847611427307,
+      "learning_rate": 0.0002772951969778737,
+      "loss": 3.342,
+      "step": 49950
+    },
+    {
+      "epoch": 5.390835579514825,
+      "grad_norm": 0.708238422870636,
+      "learning_rate": 0.0002769713977334053,
+      "loss": 3.3443,
+      "step": 50000
+    },
+    {
+      "epoch": 5.390835579514825,
+      "eval_accuracy": 0.38157861998011433,
+      "eval_loss": 3.404581308364868,
+      "eval_runtime": 185.898,
+      "eval_samples_per_second": 96.886,
+      "eval_steps_per_second": 6.057,
+      "step": 50000
+    },
+    {
+      "epoch": 5.39622641509434,
+      "grad_norm": 0.7190195322036743,
+      "learning_rate": 0.00027664759848893683,
+      "loss": 3.3249,
+      "step": 50050
+    },
+    {
+      "epoch": 5.401617250673855,
+      "grad_norm": 0.7666674852371216,
+      "learning_rate": 0.00027632379924446844,
+      "loss": 3.3194,
+      "step": 50100
+    },
+    {
+      "epoch": 5.407008086253369,
+      "grad_norm": 0.7099547982215881,
+      "learning_rate": 0.000276,
+      "loss": 3.3352,
+      "step": 50150
+    },
+    {
+      "epoch": 5.412398921832884,
+      "grad_norm": 0.706015944480896,
+      "learning_rate": 0.00027567620075553154,
+      "loss": 3.3309,
+      "step": 50200
+    },
+    {
+      "epoch": 5.4177897574123985,
+      "grad_norm": 0.7028420567512512,
+      "learning_rate": 0.0002753524015110631,
+      "loss": 3.3395,
+      "step": 50250
+    },
+    {
+      "epoch": 5.423180592991914,
+      "grad_norm": 0.718315839767456,
+      "learning_rate": 0.0002750286022665947,
+      "loss": 3.3238,
+      "step": 50300
+    },
+    {
+      "epoch": 5.428571428571429,
+      "grad_norm": 0.6950214505195618,
+      "learning_rate": 0.00027470480302212624,
+      "loss": 3.3527,
+      "step": 50350
+    },
+    {
+      "epoch": 5.433962264150943,
+      "grad_norm": 0.7302674055099487,
+      "learning_rate": 0.00027438100377765785,
+      "loss": 3.3334,
+      "step": 50400
+    },
+    {
+      "epoch": 5.439353099730458,
+      "grad_norm": 0.7544096112251282,
+      "learning_rate": 0.0002740572045331894,
+      "loss": 3.3379,
+      "step": 50450
+    },
+    {
+      "epoch": 5.444743935309973,
+      "grad_norm": 0.7582064867019653,
+      "learning_rate": 0.00027373340528872095,
+      "loss": 3.3277,
+      "step": 50500
+    },
+    {
+      "epoch": 5.450134770889488,
+      "grad_norm": 0.7223314046859741,
+      "learning_rate": 0.00027340960604425255,
+      "loss": 3.3336,
+      "step": 50550
+    },
+    {
+      "epoch": 5.455525606469003,
+      "grad_norm": 0.7132956981658936,
+      "learning_rate": 0.0002730858067997841,
+      "loss": 3.3437,
+      "step": 50600
+    },
+    {
+      "epoch": 5.460916442048518,
+      "grad_norm": 0.6773133277893066,
+      "learning_rate": 0.00027276200755531565,
+      "loss": 3.3561,
+      "step": 50650
+    },
+    {
+      "epoch": 5.466307277628032,
+      "grad_norm": 0.7546108961105347,
+      "learning_rate": 0.00027243820831084726,
+      "loss": 3.3178,
+      "step": 50700
+    },
+    {
+      "epoch": 5.471698113207547,
+      "grad_norm": 0.7244606018066406,
+      "learning_rate": 0.0002721144090663788,
+      "loss": 3.3354,
+      "step": 50750
+    },
+    {
+      "epoch": 5.4770889487870615,
+      "grad_norm": 0.7497327923774719,
+      "learning_rate": 0.00027179060982191036,
+      "loss": 3.3365,
+      "step": 50800
+    },
+    {
+      "epoch": 5.482479784366577,
+      "grad_norm": 0.6621273159980774,
+      "learning_rate": 0.00027146681057744197,
+      "loss": 3.3294,
+      "step": 50850
+    },
+    {
+      "epoch": 5.487870619946092,
+      "grad_norm": 0.7253899574279785,
+      "learning_rate": 0.0002711430113329735,
+      "loss": 3.3279,
+      "step": 50900
+    },
+    {
+      "epoch": 5.493261455525606,
+      "grad_norm": 0.6984434723854065,
+      "learning_rate": 0.0002708192120885051,
+      "loss": 3.3432,
+      "step": 50950
+    },
+    {
+      "epoch": 5.498652291105121,
+      "grad_norm": 0.6961939930915833,
+      "learning_rate": 0.00027049541284403667,
+      "loss": 3.345,
+      "step": 51000
+    },
+    {
+      "epoch": 5.498652291105121,
+      "eval_accuracy": 0.381837105200225,
+      "eval_loss": 3.400423049926758,
+      "eval_runtime": 185.6407,
+      "eval_samples_per_second": 97.021,
+      "eval_steps_per_second": 6.065,
+      "step": 51000
+    },
+    {
+      "epoch": 5.504043126684636,
+      "grad_norm": 0.7146604657173157,
+      "learning_rate": 0.0002701716135995683,
+      "loss": 3.3515,
+      "step": 51050
+    },
+    {
+      "epoch": 5.509433962264151,
+      "grad_norm": 0.6723055839538574,
+      "learning_rate": 0.0002698478143550998,
+      "loss": 3.3397,
+      "step": 51100
+    },
+    {
+      "epoch": 5.514824797843666,
+      "grad_norm": 0.7232591509819031,
+      "learning_rate": 0.0002695304910955207,
+      "loss": 3.3353,
+      "step": 51150
+    },
+    {
+      "epoch": 5.520215633423181,
+      "grad_norm": 0.7060632109642029,
+      "learning_rate": 0.00026920669185105233,
+      "loss": 3.3191,
+      "step": 51200
+    },
+    {
+      "epoch": 5.525606469002695,
+      "grad_norm": 0.7194018363952637,
+      "learning_rate": 0.0002688828926065839,
+      "loss": 3.3424,
+      "step": 51250
+    },
+    {
+      "epoch": 5.53099730458221,
+      "grad_norm": 0.710332989692688,
+      "learning_rate": 0.0002685590933621155,
+      "loss": 3.3324,
+      "step": 51300
+    },
+    {
+      "epoch": 5.536388140161725,
+      "grad_norm": 0.722364604473114,
+      "learning_rate": 0.00026823529411764704,
+      "loss": 3.3231,
+      "step": 51350
+    },
+    {
+      "epoch": 5.54177897574124,
+      "grad_norm": 0.763502299785614,
+      "learning_rate": 0.0002679114948731786,
+      "loss": 3.3131,
+      "step": 51400
+    },
+    {
+      "epoch": 5.547169811320755,
+      "grad_norm": 0.7600301504135132,
+      "learning_rate": 0.0002675876956287102,
+      "loss": 3.3135,
+      "step": 51450
+    },
+    {
+      "epoch": 5.55256064690027,
+      "grad_norm": 0.7283404469490051,
+      "learning_rate": 0.00026726389638424174,
+      "loss": 3.3435,
+      "step": 51500
+    },
+    {
+      "epoch": 5.557951482479784,
+      "grad_norm": 0.719776451587677,
+      "learning_rate": 0.0002669400971397733,
+      "loss": 3.346,
+      "step": 51550
+    },
+    {
+      "epoch": 5.563342318059299,
+      "grad_norm": 0.6799510717391968,
+      "learning_rate": 0.0002666162978953049,
+      "loss": 3.3218,
+      "step": 51600
+    },
+    {
+      "epoch": 5.568733153638814,
+      "grad_norm": 0.7170778512954712,
+      "learning_rate": 0.00026629249865083645,
+      "loss": 3.3379,
+      "step": 51650
+    },
+    {
+      "epoch": 5.574123989218329,
+      "grad_norm": 0.7793388962745667,
+      "learning_rate": 0.00026596869940636805,
+      "loss": 3.3409,
+      "step": 51700
+    },
+    {
+      "epoch": 5.579514824797844,
+      "grad_norm": 0.692438006401062,
+      "learning_rate": 0.0002656449001618996,
+      "loss": 3.3332,
+      "step": 51750
+    },
+    {
+      "epoch": 5.584905660377358,
+      "grad_norm": 0.6874307990074158,
+      "learning_rate": 0.0002653211009174312,
+      "loss": 3.3242,
+      "step": 51800
+    },
+    {
+      "epoch": 5.590296495956873,
+      "grad_norm": 0.7971224188804626,
+      "learning_rate": 0.0002649973016729627,
+      "loss": 3.3349,
+      "step": 51850
+    },
+    {
+      "epoch": 5.595687331536388,
+      "grad_norm": 0.7730355858802795,
+      "learning_rate": 0.0002646735024284943,
+      "loss": 3.3236,
+      "step": 51900
+    },
+    {
+      "epoch": 5.601078167115903,
+      "grad_norm": 0.7166358828544617,
+      "learning_rate": 0.00026435617916891526,
+      "loss": 3.3386,
+      "step": 51950
+    },
+    {
+      "epoch": 5.606469002695418,
+      "grad_norm": 0.6776190400123596,
+      "learning_rate": 0.0002640323799244468,
+      "loss": 3.3191,
+      "step": 52000
+    },
+    {
+      "epoch": 5.606469002695418,
+      "eval_accuracy": 0.382201418333067,
+      "eval_loss": 3.3955237865448,
+      "eval_runtime": 185.6837,
+      "eval_samples_per_second": 96.998,
+      "eval_steps_per_second": 6.064,
+      "step": 52000
+    },
+    {
+      "epoch": 5.611859838274933,
+      "grad_norm": 0.7449101805686951,
+      "learning_rate": 0.0002637085806799784,
+      "loss": 3.3357,
+      "step": 52050
+    },
+    {
+      "epoch": 5.617250673854447,
+      "grad_norm": 0.717818558216095,
+      "learning_rate": 0.00026338478143550997,
+      "loss": 3.3301,
+      "step": 52100
+    },
+    {
+      "epoch": 5.622641509433962,
+      "grad_norm": 0.7085523009300232,
+      "learning_rate": 0.00026306098219104157,
+      "loss": 3.3492,
+      "step": 52150
+    },
+    {
+      "epoch": 5.628032345013477,
+      "grad_norm": 0.7379426956176758,
+      "learning_rate": 0.00026273718294657307,
+      "loss": 3.3512,
+      "step": 52200
+    },
+    {
+      "epoch": 5.633423180592992,
+      "grad_norm": 0.6585873365402222,
+      "learning_rate": 0.00026241338370210467,
+      "loss": 3.3322,
+      "step": 52250
+    },
+    {
+      "epoch": 5.638814016172507,
+      "grad_norm": 0.7225645780563354,
+      "learning_rate": 0.0002620895844576362,
+      "loss": 3.3192,
+      "step": 52300
+    },
+    {
+      "epoch": 5.644204851752022,
+      "grad_norm": 0.6852019429206848,
+      "learning_rate": 0.00026176578521316783,
+      "loss": 3.3444,
+      "step": 52350
+    },
+    {
+      "epoch": 5.649595687331536,
+      "grad_norm": 0.7016271948814392,
+      "learning_rate": 0.0002614419859686994,
+      "loss": 3.3192,
+      "step": 52400
+    },
+    {
+      "epoch": 5.654986522911051,
+      "grad_norm": 0.7598816752433777,
+      "learning_rate": 0.000261118186724231,
+      "loss": 3.3403,
+      "step": 52450
+    },
+    {
+      "epoch": 5.660377358490566,
+      "grad_norm": 0.8143914937973022,
+      "learning_rate": 0.00026079438747976253,
+      "loss": 3.3446,
+      "step": 52500
+    },
+    {
+      "epoch": 5.665768194070081,
+      "grad_norm": 0.7350433468818665,
+      "learning_rate": 0.0002604705882352941,
+      "loss": 3.3345,
+      "step": 52550
+    },
+    {
+      "epoch": 5.671159029649596,
+      "grad_norm": 0.7242854833602905,
+      "learning_rate": 0.00026014678899082563,
+      "loss": 3.3468,
+      "step": 52600
+    },
+    {
+      "epoch": 5.67654986522911,
+      "grad_norm": 0.716185450553894,
+      "learning_rate": 0.00025982298974635724,
+      "loss": 3.3455,
+      "step": 52650
+    },
+    {
+      "epoch": 5.681940700808625,
+      "grad_norm": 0.6937525868415833,
+      "learning_rate": 0.0002594991905018888,
+      "loss": 3.334,
+      "step": 52700
+    },
+    {
+      "epoch": 5.6873315363881405,
+      "grad_norm": 0.7629469633102417,
+      "learning_rate": 0.0002591753912574204,
+      "loss": 3.3464,
+      "step": 52750
+    },
+    {
+      "epoch": 5.692722371967655,
+      "grad_norm": 0.7545972466468811,
+      "learning_rate": 0.00025885159201295195,
+      "loss": 3.327,
+      "step": 52800
+    },
+    {
+      "epoch": 5.69811320754717,
+      "grad_norm": 0.7334835529327393,
+      "learning_rate": 0.0002585277927684835,
+      "loss": 3.3274,
+      "step": 52850
+    },
+    {
+      "epoch": 5.703504043126685,
+      "grad_norm": 0.7234471440315247,
+      "learning_rate": 0.0002582039935240151,
+      "loss": 3.3274,
+      "step": 52900
+    },
+    {
+      "epoch": 5.708894878706199,
+      "grad_norm": 0.7582783699035645,
+      "learning_rate": 0.00025788019427954665,
+      "loss": 3.3459,
+      "step": 52950
+    },
+    {
+      "epoch": 5.714285714285714,
+      "grad_norm": 0.7253233790397644,
+      "learning_rate": 0.00025755639503507826,
+      "loss": 3.3349,
+      "step": 53000
+    },
+    {
+      "epoch": 5.714285714285714,
+      "eval_accuracy": 0.38310399787214183,
+      "eval_loss": 3.3891968727111816,
+      "eval_runtime": 185.7248,
+      "eval_samples_per_second": 96.977,
+      "eval_steps_per_second": 6.063,
+      "step": 53000
+    },
+    {
+      "epoch": 5.719676549865229,
+      "grad_norm": 0.7292880415916443,
+      "learning_rate": 0.0002572325957906098,
+      "loss": 3.3148,
+      "step": 53050
+    },
+    {
+      "epoch": 5.725067385444744,
+      "grad_norm": 0.7712312936782837,
+      "learning_rate": 0.00025690879654614136,
+      "loss": 3.3303,
+      "step": 53100
+    },
+    {
+      "epoch": 5.730458221024259,
+      "grad_norm": 0.7618584632873535,
+      "learning_rate": 0.0002565849973016729,
+      "loss": 3.334,
+      "step": 53150
+    },
+    {
+      "epoch": 5.735849056603773,
+      "grad_norm": 0.7276248335838318,
+      "learning_rate": 0.0002562611980572045,
+      "loss": 3.3356,
+      "step": 53200
+    },
+    {
+      "epoch": 5.741239892183288,
+      "grad_norm": 0.7285054922103882,
+      "learning_rate": 0.00025593739881273606,
+      "loss": 3.3251,
+      "step": 53250
+    },
+    {
+      "epoch": 5.7466307277628035,
+      "grad_norm": 0.7207826375961304,
+      "learning_rate": 0.00025561359956826767,
+      "loss": 3.3526,
+      "step": 53300
+    },
+    {
+      "epoch": 5.752021563342318,
+      "grad_norm": 0.7463442087173462,
+      "learning_rate": 0.0002552898003237992,
+      "loss": 3.3276,
+      "step": 53350
+    },
+    {
+      "epoch": 5.757412398921833,
+      "grad_norm": 0.7145280241966248,
+      "learning_rate": 0.0002549660010793308,
+      "loss": 3.3152,
+      "step": 53400
+    },
+    {
+      "epoch": 5.762803234501348,
+      "grad_norm": 0.7333078980445862,
+      "learning_rate": 0.0002546422018348624,
+      "loss": 3.3489,
+      "step": 53450
+    },
+    {
+      "epoch": 5.768194070080862,
+      "grad_norm": 0.723423182964325,
+      "learning_rate": 0.0002543184025903939,
+      "loss": 3.3157,
+      "step": 53500
+    },
+    {
+      "epoch": 5.773584905660377,
+      "grad_norm": 0.7532666921615601,
+      "learning_rate": 0.0002539946033459255,
+      "loss": 3.3445,
+      "step": 53550
+    },
+    {
+      "epoch": 5.7789757412398925,
+      "grad_norm": 0.7549908757209778,
+      "learning_rate": 0.0002536708041014571,
+      "loss": 3.3179,
+      "step": 53600
+    },
+    {
+      "epoch": 5.784366576819407,
+      "grad_norm": 0.7116851210594177,
+      "learning_rate": 0.00025334700485698863,
+      "loss": 3.3517,
+      "step": 53650
+    },
+    {
+      "epoch": 5.789757412398922,
+      "grad_norm": 0.6998304128646851,
+      "learning_rate": 0.00025302320561252023,
+      "loss": 3.3231,
+      "step": 53700
+    },
+    {
+      "epoch": 5.795148247978437,
+      "grad_norm": 0.7829564213752747,
+      "learning_rate": 0.0002526994063680518,
+      "loss": 3.3368,
+      "step": 53750
+    },
+    {
+      "epoch": 5.800539083557951,
+      "grad_norm": 0.756128191947937,
+      "learning_rate": 0.0002523756071235834,
+      "loss": 3.3297,
+      "step": 53800
+    },
+    {
+      "epoch": 5.8059299191374665,
+      "grad_norm": 0.8946164846420288,
+      "learning_rate": 0.00025205180787911494,
+      "loss": 3.3353,
+      "step": 53850
+    },
+    {
+      "epoch": 5.811320754716981,
+      "grad_norm": 0.7046872973442078,
+      "learning_rate": 0.0002517280086346465,
+      "loss": 3.3262,
+      "step": 53900
+    },
+    {
+      "epoch": 5.816711590296496,
+      "grad_norm": 0.7093944549560547,
+      "learning_rate": 0.00025140420939017804,
+      "loss": 3.3185,
+      "step": 53950
+    },
+    {
+      "epoch": 5.822102425876011,
+      "grad_norm": 0.7125768065452576,
+      "learning_rate": 0.00025108041014570965,
+      "loss": 3.332,
+      "step": 54000
+    },
+    {
+      "epoch": 5.822102425876011,
+      "eval_accuracy": 0.3833044624501596,
+      "eval_loss": 3.385930061340332,
+      "eval_runtime": 185.5949,
+      "eval_samples_per_second": 97.045,
+      "eval_steps_per_second": 6.067,
+      "step": 54000
+    },
+    {
+      "epoch": 5.827493261455525,
+      "grad_norm": 0.7239689826965332,
+      "learning_rate": 0.0002507566109012412,
+      "loss": 3.3453,
+      "step": 54050
+    },
+    {
+      "epoch": 5.83288409703504,
+      "grad_norm": 0.6991035342216492,
+      "learning_rate": 0.0002504328116567728,
+      "loss": 3.3514,
+      "step": 54100
+    },
+    {
+      "epoch": 5.8382749326145555,
+      "grad_norm": 0.7742658257484436,
+      "learning_rate": 0.00025010901241230435,
+      "loss": 3.3242,
+      "step": 54150
+    },
+    {
+      "epoch": 5.84366576819407,
+      "grad_norm": 0.7066153287887573,
+      "learning_rate": 0.0002497852131678359,
+      "loss": 3.3347,
+      "step": 54200
+    },
+    {
+      "epoch": 5.849056603773585,
+      "grad_norm": 0.7707701921463013,
+      "learning_rate": 0.0002494614139233675,
+      "loss": 3.3646,
+      "step": 54250
+    },
+    {
+      "epoch": 5.8544474393531,
+      "grad_norm": 0.7194285988807678,
+      "learning_rate": 0.00024913761467889906,
+      "loss": 3.341,
+      "step": 54300
+    },
+    {
+      "epoch": 5.859838274932614,
+      "grad_norm": 0.6992973685264587,
+      "learning_rate": 0.0002488138154344306,
+      "loss": 3.3445,
+      "step": 54350
+    },
+    {
+      "epoch": 5.8652291105121295,
+      "grad_norm": 0.776282012462616,
+      "learning_rate": 0.0002484900161899622,
+      "loss": 3.3458,
+      "step": 54400
+    },
+    {
+      "epoch": 5.870619946091644,
+      "grad_norm": 0.7476275563240051,
+      "learning_rate": 0.00024816621694549376,
+      "loss": 3.3187,
+      "step": 54450
+    },
+    {
+      "epoch": 5.876010781671159,
+      "grad_norm": 0.756615936756134,
+      "learning_rate": 0.00024784241770102537,
+      "loss": 3.3467,
+      "step": 54500
+    },
+    {
+      "epoch": 5.881401617250674,
+      "grad_norm": 0.7004367709159851,
+      "learning_rate": 0.0002475186184565569,
+      "loss": 3.3346,
+      "step": 54550
+    },
+    {
+      "epoch": 5.886792452830189,
+      "grad_norm": 0.6643033027648926,
+      "learning_rate": 0.00024719481921208847,
+      "loss": 3.3249,
+      "step": 54600
+    },
+    {
+      "epoch": 5.892183288409703,
+      "grad_norm": 0.7226757407188416,
+      "learning_rate": 0.0002468710199676201,
+      "loss": 3.3478,
+      "step": 54650
+    },
+    {
+      "epoch": 5.8975741239892185,
+      "grad_norm": 0.7756559252738953,
+      "learning_rate": 0.0002465472207231516,
+      "loss": 3.3446,
+      "step": 54700
+    },
+    {
+      "epoch": 5.902964959568733,
+      "grad_norm": 0.7599494457244873,
+      "learning_rate": 0.0002462234214786832,
+      "loss": 3.3284,
+      "step": 54750
+    },
+    {
+      "epoch": 5.908355795148248,
+      "grad_norm": 0.7394720315933228,
+      "learning_rate": 0.0002458996222342148,
+      "loss": 3.3595,
+      "step": 54800
+    },
+    {
+      "epoch": 5.913746630727763,
+      "grad_norm": 0.72690349817276,
+      "learning_rate": 0.00024557582298974633,
+      "loss": 3.3269,
+      "step": 54850
+    },
+    {
+      "epoch": 5.919137466307277,
+      "grad_norm": 0.7528433203697205,
+      "learning_rate": 0.0002452520237452779,
+      "loss": 3.3567,
+      "step": 54900
+    },
+    {
+      "epoch": 5.9245283018867925,
+      "grad_norm": 0.6863235831260681,
+      "learning_rate": 0.0002449282245008095,
+      "loss": 3.3356,
+      "step": 54950
+    },
+    {
+      "epoch": 5.929919137466308,
+      "grad_norm": 0.7730789184570312,
+      "learning_rate": 0.00024461090124123044,
+      "loss": 3.3376,
+      "step": 55000
+    },
+    {
+      "epoch": 5.929919137466308,
+      "eval_accuracy": 0.38378905432981214,
+      "eval_loss": 3.381817102432251,
+      "eval_runtime": 185.5997,
+      "eval_samples_per_second": 97.042,
+      "eval_steps_per_second": 6.067,
+      "step": 55000
+    },
+    {
+      "epoch": 5.935309973045822,
+      "grad_norm": 0.7587157487869263,
+      "learning_rate": 0.000244287101996762,
+      "loss": 3.3562,
+      "step": 55050
+    },
+    {
+      "epoch": 5.940700808625337,
+      "grad_norm": 0.7801807522773743,
+      "learning_rate": 0.00024396330275229354,
+      "loss": 3.3311,
+      "step": 55100
+    },
+    {
+      "epoch": 5.946091644204852,
+      "grad_norm": 0.8184198141098022,
+      "learning_rate": 0.00024363950350782512,
+      "loss": 3.3441,
+      "step": 55150
+    },
+    {
+      "epoch": 5.951482479784366,
+      "grad_norm": 0.7494048476219177,
+      "learning_rate": 0.0002433157042633567,
+      "loss": 3.3285,
+      "step": 55200
+    },
+    {
+      "epoch": 5.9568733153638815,
+      "grad_norm": 0.7266256809234619,
+      "learning_rate": 0.00024299190501888827,
+      "loss": 3.3151,
+      "step": 55250
+    },
+    {
+      "epoch": 5.962264150943396,
+      "grad_norm": 0.738677978515625,
+      "learning_rate": 0.00024266810577441985,
+      "loss": 3.3311,
+      "step": 55300
+    },
+    {
+      "epoch": 5.967654986522911,
+      "grad_norm": 0.7299897074699402,
+      "learning_rate": 0.00024234430652995143,
+      "loss": 3.3284,
+      "step": 55350
+    },
+    {
+      "epoch": 5.973045822102426,
+      "grad_norm": 0.6893080472946167,
+      "learning_rate": 0.00024202050728548298,
+      "loss": 3.3408,
+      "step": 55400
+    },
+    {
+      "epoch": 5.97843665768194,
+      "grad_norm": 0.7181869745254517,
+      "learning_rate": 0.00024169670804101456,
+      "loss": 3.3488,
+      "step": 55450
+    },
+    {
+      "epoch": 5.9838274932614555,
+      "grad_norm": 0.7386427521705627,
+      "learning_rate": 0.0002413729087965461,
+      "loss": 3.3172,
+      "step": 55500
+    },
+    {
+      "epoch": 5.989218328840971,
+      "grad_norm": 0.7251370549201965,
+      "learning_rate": 0.00024104910955207768,
+      "loss": 3.3374,
+      "step": 55550
+    },
+    {
+      "epoch": 5.994609164420485,
+      "grad_norm": 0.7081345319747925,
+      "learning_rate": 0.00024072531030760926,
+      "loss": 3.3517,
+      "step": 55600
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 1.683327555656433,
+      "learning_rate": 0.00024040151106314084,
+      "loss": 3.32,
+      "step": 55650
+    },
+    {
+      "epoch": 6.005390835579515,
+      "grad_norm": 0.7327098846435547,
+      "learning_rate": 0.0002400777118186724,
+      "loss": 3.2341,
+      "step": 55700
+    },
+    {
+      "epoch": 6.010781671159029,
+      "grad_norm": 0.7467867136001587,
+      "learning_rate": 0.00023975391257420397,
+      "loss": 3.2359,
+      "step": 55750
+    },
+    {
+      "epoch": 6.0161725067385445,
+      "grad_norm": 0.7838574647903442,
+      "learning_rate": 0.00023943011332973555,
+      "loss": 3.2685,
+      "step": 55800
+    },
+    {
+      "epoch": 6.02156334231806,
+      "grad_norm": 0.7408933043479919,
+      "learning_rate": 0.00023910631408526712,
+      "loss": 3.2343,
+      "step": 55850
+    },
+    {
+      "epoch": 6.026954177897574,
+      "grad_norm": 0.7470614314079285,
+      "learning_rate": 0.0002387825148407987,
+      "loss": 3.2364,
+      "step": 55900
+    },
+    {
+      "epoch": 6.032345013477089,
+      "grad_norm": 0.7341518998146057,
+      "learning_rate": 0.00023845871559633025,
+      "loss": 3.2636,
+      "step": 55950
+    },
+    {
+      "epoch": 6.037735849056604,
+      "grad_norm": 0.8219843506813049,
+      "learning_rate": 0.0002381349163518618,
+      "loss": 3.2377,
+      "step": 56000
+    },
+    {
+      "epoch": 6.037735849056604,
+      "eval_accuracy": 0.38377612463616223,
+      "eval_loss": 3.3850560188293457,
+      "eval_runtime": 185.4498,
+      "eval_samples_per_second": 97.121,
+      "eval_steps_per_second": 6.072,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0431266846361185,
+      "grad_norm": 0.7303977608680725,
+      "learning_rate": 0.00023781111710739338,
+      "loss": 3.2518,
+      "step": 56050
+    },
+    {
+      "epoch": 6.048517520215634,
+      "grad_norm": 0.7296903729438782,
+      "learning_rate": 0.00023748731786292496,
+      "loss": 3.2434,
+      "step": 56100
+    },
+    {
+      "epoch": 6.053908355795148,
+      "grad_norm": 0.7722464799880981,
+      "learning_rate": 0.00023716351861845654,
+      "loss": 3.2624,
+      "step": 56150
+    },
+    {
+      "epoch": 6.059299191374663,
+      "grad_norm": 0.743776261806488,
+      "learning_rate": 0.0002368397193739881,
+      "loss": 3.2607,
+      "step": 56200
+    },
+    {
+      "epoch": 6.064690026954178,
+      "grad_norm": 0.7222715616226196,
+      "learning_rate": 0.0002365159201295197,
+      "loss": 3.2549,
+      "step": 56250
+    },
+    {
+      "epoch": 6.070080862533692,
+      "grad_norm": 0.7410303950309753,
+      "learning_rate": 0.00023619212088505127,
+      "loss": 3.2594,
+      "step": 56300
+    },
+    {
+      "epoch": 6.0754716981132075,
+      "grad_norm": 0.7342054843902588,
+      "learning_rate": 0.0002358683216405828,
+      "loss": 3.2606,
+      "step": 56350
+    },
+    {
+      "epoch": 6.080862533692723,
+      "grad_norm": 0.7345823645591736,
+      "learning_rate": 0.00023554452239611437,
+      "loss": 3.2742,
+      "step": 56400
+    },
+    {
+      "epoch": 6.086253369272237,
+      "grad_norm": 0.8020313382148743,
+      "learning_rate": 0.00023522072315164595,
+      "loss": 3.2558,
+      "step": 56450
+    },
+    {
+      "epoch": 6.091644204851752,
+      "grad_norm": 0.8014670610427856,
+      "learning_rate": 0.00023489692390717752,
+      "loss": 3.247,
+      "step": 56500
+    },
+    {
+      "epoch": 6.097035040431267,
+      "grad_norm": 0.8149689435958862,
+      "learning_rate": 0.0002345731246627091,
+      "loss": 3.2515,
+      "step": 56550
+    },
+    {
+      "epoch": 6.1024258760107815,
+      "grad_norm": 0.7491607069969177,
+      "learning_rate": 0.00023424932541824068,
+      "loss": 3.2667,
+      "step": 56600
+    },
+    {
+      "epoch": 6.107816711590297,
+      "grad_norm": 0.7217200994491577,
+      "learning_rate": 0.00023392552617377226,
+      "loss": 3.2623,
+      "step": 56650
+    },
+    {
+      "epoch": 6.113207547169812,
+      "grad_norm": 0.7023745179176331,
+      "learning_rate": 0.00023360172692930384,
+      "loss": 3.2616,
+      "step": 56700
+    },
+    {
+      "epoch": 6.118598382749326,
+      "grad_norm": 0.7833262085914612,
+      "learning_rate": 0.00023327792768483539,
+      "loss": 3.2703,
+      "step": 56750
+    },
+    {
+      "epoch": 6.123989218328841,
+      "grad_norm": 0.721172571182251,
+      "learning_rate": 0.00023295412844036694,
+      "loss": 3.2675,
+      "step": 56800
+    },
+    {
+      "epoch": 6.129380053908355,
+      "grad_norm": 0.7600164413452148,
+      "learning_rate": 0.00023263032919589851,
+      "loss": 3.261,
+      "step": 56850
+    },
+    {
+      "epoch": 6.1347708894878705,
+      "grad_norm": 0.7170189619064331,
+      "learning_rate": 0.0002323065299514301,
+      "loss": 3.2708,
+      "step": 56900
+    },
+    {
+      "epoch": 6.140161725067386,
+      "grad_norm": 0.7651640772819519,
+      "learning_rate": 0.00023198273070696167,
+      "loss": 3.269,
+      "step": 56950
+    },
+    {
+      "epoch": 6.1455525606469,
+      "grad_norm": 0.752663254737854,
+      "learning_rate": 0.00023165893146249325,
+      "loss": 3.2589,
+      "step": 57000
+    },
+    {
+      "epoch": 6.1455525606469,
+      "eval_accuracy": 0.38401787731154946,
+      "eval_loss": 3.3852553367614746,
+      "eval_runtime": 185.5998,
+      "eval_samples_per_second": 97.042,
+      "eval_steps_per_second": 6.067,
+      "step": 57000
+    },
+    {
+      "epoch": 6.150943396226415,
+      "grad_norm": 0.7720143795013428,
+      "learning_rate": 0.0002313351322180248,
+      "loss": 3.276,
+      "step": 57050
+    },
+    {
+      "epoch": 6.15633423180593,
+      "grad_norm": 0.7548387050628662,
+      "learning_rate": 0.00023101133297355638,
+      "loss": 3.2787,
+      "step": 57100
+    },
+    {
+      "epoch": 6.1617250673854445,
+      "grad_norm": 0.7913086414337158,
+      "learning_rate": 0.00023068753372908795,
+      "loss": 3.2767,
+      "step": 57150
+    },
+    {
+      "epoch": 6.16711590296496,
+      "grad_norm": 1.4145164489746094,
+      "learning_rate": 0.0002303637344846195,
+      "loss": 3.2629,
+      "step": 57200
+    },
+    {
+      "epoch": 6.172506738544475,
+      "grad_norm": 0.8271373510360718,
+      "learning_rate": 0.00023003993524015108,
+      "loss": 3.2703,
+      "step": 57250
+    },
+    {
+      "epoch": 6.177897574123989,
+      "grad_norm": 0.7342875003814697,
+      "learning_rate": 0.00022971613599568266,
+      "loss": 3.2771,
+      "step": 57300
+    },
+    {
+      "epoch": 6.183288409703504,
+      "grad_norm": 0.7747467756271362,
+      "learning_rate": 0.0002293923367512142,
+      "loss": 3.2819,
+      "step": 57350
+    },
+    {
+      "epoch": 6.188679245283019,
+      "grad_norm": 0.7500191330909729,
+      "learning_rate": 0.0002290685375067458,
+      "loss": 3.2669,
+      "step": 57400
+    },
+    {
+      "epoch": 6.1940700808625335,
+      "grad_norm": 0.7284804582595825,
+      "learning_rate": 0.00022874473826227736,
+      "loss": 3.2628,
+      "step": 57450
+    },
+    {
+      "epoch": 6.199460916442049,
+      "grad_norm": 0.7782807350158691,
+      "learning_rate": 0.00022842093901780894,
+      "loss": 3.2768,
+      "step": 57500
+    },
+    {
+      "epoch": 6.204851752021563,
+      "grad_norm": 0.772388756275177,
+      "learning_rate": 0.00022809713977334052,
+      "loss": 3.3015,
+      "step": 57550
+    },
+    {
+      "epoch": 6.210242587601078,
+      "grad_norm": 0.726718544960022,
+      "learning_rate": 0.0002277733405288721,
+      "loss": 3.2712,
+      "step": 57600
+    },
+    {
+      "epoch": 6.215633423180593,
+      "grad_norm": 0.8056442737579346,
+      "learning_rate": 0.00022744954128440365,
+      "loss": 3.2764,
+      "step": 57650
+    },
+    {
+      "epoch": 6.2210242587601075,
+      "grad_norm": 0.7551636099815369,
+      "learning_rate": 0.0002271257420399352,
+      "loss": 3.2712,
+      "step": 57700
+    },
+    {
+      "epoch": 6.226415094339623,
+      "grad_norm": 0.7155594229698181,
+      "learning_rate": 0.00022680194279546678,
+      "loss": 3.2609,
+      "step": 57750
+    },
+    {
+      "epoch": 6.231805929919138,
+      "grad_norm": 0.7783682346343994,
+      "learning_rate": 0.00022647814355099835,
+      "loss": 3.2688,
+      "step": 57800
+    },
+    {
+      "epoch": 6.237196765498652,
+      "grad_norm": 0.7531312704086304,
+      "learning_rate": 0.00022615434430652993,
+      "loss": 3.2767,
+      "step": 57850
+    },
+    {
+      "epoch": 6.242587601078167,
+      "grad_norm": 0.7374745607376099,
+      "learning_rate": 0.0002258305450620615,
+      "loss": 3.2768,
+      "step": 57900
+    },
+    {
+      "epoch": 6.247978436657682,
+      "grad_norm": 0.7557754516601562,
+      "learning_rate": 0.0002255067458175931,
+      "loss": 3.2824,
+      "step": 57950
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "grad_norm": 0.7885853052139282,
+      "learning_rate": 0.00022518294657312467,
+      "loss": 3.2784,
+      "step": 58000
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "eval_accuracy": 0.38443586497097176,
+      "eval_loss": 3.381211757659912,
+      "eval_runtime": 185.7125,
+      "eval_samples_per_second": 96.983,
+      "eval_steps_per_second": 6.063,
+      "step": 58000
+    },
+    {
+      "epoch": 6.258760107816712,
+      "grad_norm": 0.773819088935852,
+      "learning_rate": 0.0002248591473286562,
+      "loss": 3.2696,
+      "step": 58050
+    },
+    {
+      "epoch": 6.264150943396227,
+      "grad_norm": 0.7929461598396301,
+      "learning_rate": 0.00022453534808418777,
+      "loss": 3.2632,
+      "step": 58100
+    },
+    {
+      "epoch": 6.269541778975741,
+      "grad_norm": 0.7103016376495361,
+      "learning_rate": 0.00022421154883971934,
+      "loss": 3.2629,
+      "step": 58150
+    },
+    {
+      "epoch": 6.274932614555256,
+      "grad_norm": 0.7493858337402344,
+      "learning_rate": 0.00022388774959525092,
+      "loss": 3.3003,
+      "step": 58200
+    },
+    {
+      "epoch": 6.280323450134771,
+      "grad_norm": 0.7662702798843384,
+      "learning_rate": 0.0002235639503507825,
+      "loss": 3.2773,
+      "step": 58250
+    },
+    {
+      "epoch": 6.285714285714286,
+      "grad_norm": 0.7544426321983337,
+      "learning_rate": 0.00022324015110631408,
+      "loss": 3.2616,
+      "step": 58300
+    },
+    {
+      "epoch": 6.291105121293801,
+      "grad_norm": 0.7621305584907532,
+      "learning_rate": 0.00022291635186184565,
+      "loss": 3.291,
+      "step": 58350
+    },
+    {
+      "epoch": 6.296495956873315,
+      "grad_norm": 0.7406484484672546,
+      "learning_rate": 0.0002225925526173772,
+      "loss": 3.272,
+      "step": 58400
+    },
+    {
+      "epoch": 6.30188679245283,
+      "grad_norm": 0.7376567125320435,
+      "learning_rate": 0.00022227522935779813,
+      "loss": 3.2799,
+      "step": 58450
+    },
+    {
+      "epoch": 6.307277628032345,
+      "grad_norm": 0.8024041652679443,
+      "learning_rate": 0.0002219514301133297,
+      "loss": 3.2946,
+      "step": 58500
+    },
+    {
+      "epoch": 6.3126684636118595,
+      "grad_norm": 0.7516837120056152,
+      "learning_rate": 0.00022162763086886129,
+      "loss": 3.2534,
+      "step": 58550
+    },
+    {
+      "epoch": 6.318059299191375,
+      "grad_norm": 0.7777389287948608,
+      "learning_rate": 0.00022130383162439286,
+      "loss": 3.2879,
+      "step": 58600
+    },
+    {
+      "epoch": 6.32345013477089,
+      "grad_norm": 0.7404323220252991,
+      "learning_rate": 0.00022098003237992444,
+      "loss": 3.2658,
+      "step": 58650
+    },
+    {
+      "epoch": 6.328840970350404,
+      "grad_norm": 0.75152587890625,
+      "learning_rate": 0.00022065623313545602,
+      "loss": 3.27,
+      "step": 58700
+    },
+    {
+      "epoch": 6.334231805929919,
+      "grad_norm": 0.761157751083374,
+      "learning_rate": 0.00022033243389098757,
+      "loss": 3.2909,
+      "step": 58750
+    },
+    {
+      "epoch": 6.339622641509434,
+      "grad_norm": 0.751017689704895,
+      "learning_rate": 0.00022000863464651915,
+      "loss": 3.2811,
+      "step": 58800
+    },
+    {
+      "epoch": 6.345013477088949,
+      "grad_norm": 0.7324016094207764,
+      "learning_rate": 0.0002196848354020507,
+      "loss": 3.2854,
+      "step": 58850
+    },
+    {
+      "epoch": 6.350404312668464,
+      "grad_norm": 0.7989504337310791,
+      "learning_rate": 0.00021936103615758227,
+      "loss": 3.2918,
+      "step": 58900
+    },
+    {
+      "epoch": 6.355795148247978,
+      "grad_norm": 0.7545925974845886,
+      "learning_rate": 0.00021903723691311385,
+      "loss": 3.3018,
+      "step": 58950
+    },
+    {
+      "epoch": 6.361185983827493,
+      "grad_norm": 0.7616492509841919,
+      "learning_rate": 0.00021871343766864543,
+      "loss": 3.2895,
+      "step": 59000
+    },
+    {
+      "epoch": 6.361185983827493,
+      "eval_accuracy": 0.3846159028060803,
+      "eval_loss": 3.3771748542785645,
+      "eval_runtime": 185.5837,
+      "eval_samples_per_second": 97.051,
+      "eval_steps_per_second": 6.067,
+      "step": 59000
+    },
+    {
+      "epoch": 6.366576819407008,
+      "grad_norm": 0.8057339191436768,
+      "learning_rate": 0.00021838963842417698,
+      "loss": 3.2869,
+      "step": 59050
+    },
+    {
+      "epoch": 6.3719676549865225,
+      "grad_norm": 0.7496756315231323,
+      "learning_rate": 0.00021806583917970856,
+      "loss": 3.2844,
+      "step": 59100
+    },
+    {
+      "epoch": 6.377358490566038,
+      "grad_norm": 0.784780740737915,
+      "learning_rate": 0.00021774203993524014,
+      "loss": 3.2852,
+      "step": 59150
+    },
+    {
+      "epoch": 6.382749326145553,
+      "grad_norm": 0.7106476426124573,
+      "learning_rate": 0.00021741824069077171,
+      "loss": 3.2769,
+      "step": 59200
+    },
+    {
+      "epoch": 6.388140161725067,
+      "grad_norm": 0.7998796701431274,
+      "learning_rate": 0.00021709444144630326,
+      "loss": 3.2809,
+      "step": 59250
+    },
+    {
+      "epoch": 6.393530997304582,
+      "grad_norm": 0.7501516342163086,
+      "learning_rate": 0.00021677064220183484,
+      "loss": 3.2741,
+      "step": 59300
+    },
+    {
+      "epoch": 6.398921832884097,
+      "grad_norm": 0.7542091608047485,
+      "learning_rate": 0.00021644684295736642,
+      "loss": 3.2648,
+      "step": 59350
+    },
+    {
+      "epoch": 6.404312668463612,
+      "grad_norm": 0.7649316787719727,
+      "learning_rate": 0.00021612304371289797,
+      "loss": 3.2538,
+      "step": 59400
+    },
+    {
+      "epoch": 6.409703504043127,
+      "grad_norm": 0.8222479224205017,
+      "learning_rate": 0.00021579924446842955,
+      "loss": 3.2849,
+      "step": 59450
+    },
+    {
+      "epoch": 6.415094339622642,
+      "grad_norm": 0.7676225304603577,
+      "learning_rate": 0.00021547544522396113,
+      "loss": 3.2756,
+      "step": 59500
+    },
+    {
+      "epoch": 6.420485175202156,
+      "grad_norm": 0.7920733690261841,
+      "learning_rate": 0.0002151516459794927,
+      "loss": 3.2916,
+      "step": 59550
+    },
+    {
+      "epoch": 6.425876010781671,
+      "grad_norm": 0.8051831722259521,
+      "learning_rate": 0.00021482784673502428,
+      "loss": 3.2866,
+      "step": 59600
+    },
+    {
+      "epoch": 6.431266846361186,
+      "grad_norm": 0.7384151816368103,
+      "learning_rate": 0.00021450404749055586,
+      "loss": 3.2661,
+      "step": 59650
+    },
+    {
+      "epoch": 6.436657681940701,
+      "grad_norm": 0.7636124491691589,
+      "learning_rate": 0.00021418024824608738,
+      "loss": 3.2646,
+      "step": 59700
+    },
+    {
+      "epoch": 6.442048517520216,
+      "grad_norm": 0.7643054723739624,
+      "learning_rate": 0.00021385644900161896,
+      "loss": 3.2784,
+      "step": 59750
+    },
+    {
+      "epoch": 6.44743935309973,
+      "grad_norm": 0.7794228792190552,
+      "learning_rate": 0.00021353264975715054,
+      "loss": 3.2873,
+      "step": 59800
+    },
+    {
+      "epoch": 6.452830188679245,
+      "grad_norm": 0.760810911655426,
+      "learning_rate": 0.00021320885051268211,
+      "loss": 3.2929,
+      "step": 59850
+    },
+    {
+      "epoch": 6.45822102425876,
+      "grad_norm": 0.7666246891021729,
+      "learning_rate": 0.0002128850512682137,
+      "loss": 3.2801,
+      "step": 59900
+    },
+    {
+      "epoch": 6.463611859838275,
+      "grad_norm": 0.7073615789413452,
+      "learning_rate": 0.00021256125202374527,
+      "loss": 3.2791,
+      "step": 59950
+    },
+    {
+      "epoch": 6.46900269541779,
+      "grad_norm": 0.8043730854988098,
+      "learning_rate": 0.00021223745277927685,
+      "loss": 3.2908,
+      "step": 60000
+    },
+    {
+      "epoch": 6.46900269541779,
+      "eval_accuracy": 0.38509006400850276,
+      "eval_loss": 3.374033212661743,
+      "eval_runtime": 185.731,
+      "eval_samples_per_second": 96.974,
+      "eval_steps_per_second": 6.063,
+      "step": 60000
+    },
+    {
+      "epoch": 6.474393530997305,
+      "grad_norm": 0.8008505702018738,
+      "learning_rate": 0.00021191365353480843,
+      "loss": 3.2924,
+      "step": 60050
+    },
+    {
+      "epoch": 6.479784366576819,
+      "grad_norm": 0.7732535004615784,
+      "learning_rate": 0.00021158985429033995,
+      "loss": 3.2806,
+      "step": 60100
+    },
+    {
+      "epoch": 6.485175202156334,
+      "grad_norm": 0.75225430727005,
+      "learning_rate": 0.00021126605504587153,
+      "loss": 3.2797,
+      "step": 60150
+    },
+    {
+      "epoch": 6.490566037735849,
+      "grad_norm": 0.7321357131004333,
+      "learning_rate": 0.0002109422558014031,
+      "loss": 3.2647,
+      "step": 60200
+    },
+    {
+      "epoch": 6.495956873315364,
+      "grad_norm": 0.7730959057807922,
+      "learning_rate": 0.00021061845655693468,
+      "loss": 3.2917,
+      "step": 60250
+    },
+    {
+      "epoch": 6.501347708894879,
+      "grad_norm": 0.7390173673629761,
+      "learning_rate": 0.00021029465731246626,
+      "loss": 3.281,
+      "step": 60300
+    },
+    {
+      "epoch": 6.506738544474393,
+      "grad_norm": 0.8006965517997742,
+      "learning_rate": 0.00020997085806799784,
+      "loss": 3.2775,
+      "step": 60350
+    },
+    {
+      "epoch": 6.512129380053908,
+      "grad_norm": 0.7447074055671692,
+      "learning_rate": 0.0002096470588235294,
+      "loss": 3.2704,
+      "step": 60400
+    },
+    {
+      "epoch": 6.517520215633423,
+      "grad_norm": 0.7788031101226807,
+      "learning_rate": 0.00020932325957906097,
+      "loss": 3.272,
+      "step": 60450
+    },
+    {
+      "epoch": 6.5229110512129385,
+      "grad_norm": 0.7655622959136963,
+      "learning_rate": 0.00020899946033459254,
+      "loss": 3.2806,
+      "step": 60500
+    },
+    {
+      "epoch": 6.528301886792453,
+      "grad_norm": 0.7953072786331177,
+      "learning_rate": 0.0002086756610901241,
+      "loss": 3.2728,
+      "step": 60550
+    },
+    {
+      "epoch": 6.533692722371968,
+      "grad_norm": 0.7440159916877747,
+      "learning_rate": 0.00020835186184565567,
+      "loss": 3.2804,
+      "step": 60600
+    },
+    {
+      "epoch": 6.539083557951482,
+      "grad_norm": 0.7470718026161194,
+      "learning_rate": 0.00020802806260118725,
+      "loss": 3.2849,
+      "step": 60650
+    },
+    {
+      "epoch": 6.544474393530997,
+      "grad_norm": 0.7817756533622742,
+      "learning_rate": 0.00020770426335671883,
+      "loss": 3.2621,
+      "step": 60700
+    },
+    {
+      "epoch": 6.549865229110512,
+      "grad_norm": 0.8029948472976685,
+      "learning_rate": 0.00020738046411225038,
+      "loss": 3.2866,
+      "step": 60750
+    },
+    {
+      "epoch": 6.555256064690027,
+      "grad_norm": 0.7905914783477783,
+      "learning_rate": 0.00020705666486778196,
+      "loss": 3.3021,
+      "step": 60800
+    },
+    {
+      "epoch": 6.560646900269542,
+      "grad_norm": 0.7610495686531067,
+      "learning_rate": 0.00020673286562331353,
+      "loss": 3.2814,
+      "step": 60850
+    },
+    {
+      "epoch": 6.566037735849057,
+      "grad_norm": 0.7549566030502319,
+      "learning_rate": 0.0002064090663788451,
+      "loss": 3.277,
+      "step": 60900
+    },
+    {
+      "epoch": 6.571428571428571,
+      "grad_norm": 0.803224503993988,
+      "learning_rate": 0.00020608526713437666,
+      "loss": 3.2964,
+      "step": 60950
+    },
+    {
+      "epoch": 6.576819407008086,
+      "grad_norm": 0.7267265319824219,
+      "learning_rate": 0.00020576146788990824,
+      "loss": 3.2828,
+      "step": 61000
+    },
+    {
+      "epoch": 6.576819407008086,
+      "eval_accuracy": 0.38581195379514216,
+      "eval_loss": 3.3680148124694824,
+      "eval_runtime": 185.7026,
+      "eval_samples_per_second": 96.988,
+      "eval_steps_per_second": 6.063,
+      "step": 61000
+    },
+    {
+      "epoch": 6.5822102425876015,
+      "grad_norm": 0.7836355566978455,
+      "learning_rate": 0.0002054376686454398,
+      "loss": 3.2959,
+      "step": 61050
+    },
+    {
+      "epoch": 6.587601078167116,
+      "grad_norm": 0.746856689453125,
+      "learning_rate": 0.00020511386940097137,
+      "loss": 3.2642,
+      "step": 61100
+    },
+    {
+      "epoch": 6.592991913746631,
+      "grad_norm": 0.7780184745788574,
+      "learning_rate": 0.00020479007015650294,
+      "loss": 3.3018,
+      "step": 61150
+    },
+    {
+      "epoch": 6.598382749326145,
+      "grad_norm": 0.7593811750411987,
+      "learning_rate": 0.00020446627091203452,
+      "loss": 3.2866,
+      "step": 61200
+    },
+    {
+      "epoch": 6.60377358490566,
+      "grad_norm": 0.7430720925331116,
+      "learning_rate": 0.0002041424716675661,
+      "loss": 3.2725,
+      "step": 61250
+    },
+    {
+      "epoch": 6.609164420485175,
+      "grad_norm": 0.7618003487586975,
+      "learning_rate": 0.00020381867242309768,
+      "loss": 3.3022,
+      "step": 61300
+    },
+    {
+      "epoch": 6.6145552560646905,
+      "grad_norm": 0.7492671608924866,
+      "learning_rate": 0.00020349487317862926,
+      "loss": 3.2967,
+      "step": 61350
+    },
+    {
+      "epoch": 6.619946091644205,
+      "grad_norm": 0.8068253397941589,
+      "learning_rate": 0.00020317107393416078,
+      "loss": 3.2826,
+      "step": 61400
+    },
+    {
+      "epoch": 6.62533692722372,
+      "grad_norm": 0.7889710664749146,
+      "learning_rate": 0.00020284727468969236,
+      "loss": 3.269,
+      "step": 61450
+    },
+    {
+      "epoch": 6.630727762803234,
+      "grad_norm": 0.7961161732673645,
+      "learning_rate": 0.00020252347544522393,
+      "loss": 3.2928,
+      "step": 61500
+    },
+    {
+      "epoch": 6.636118598382749,
+      "grad_norm": 0.782105028629303,
+      "learning_rate": 0.0002021996762007555,
+      "loss": 3.2776,
+      "step": 61550
+    },
+    {
+      "epoch": 6.6415094339622645,
+      "grad_norm": 0.8133910298347473,
+      "learning_rate": 0.00020188235294117646,
+      "loss": 3.2791,
+      "step": 61600
+    },
+    {
+      "epoch": 6.646900269541779,
+      "grad_norm": 0.765106737613678,
+      "learning_rate": 0.00020155855369670804,
+      "loss": 3.2886,
+      "step": 61650
+    },
+    {
+      "epoch": 6.652291105121294,
+      "grad_norm": 0.7589500546455383,
+      "learning_rate": 0.00020123475445223956,
+      "loss": 3.2984,
+      "step": 61700
+    },
+    {
+      "epoch": 6.657681940700809,
+      "grad_norm": 0.7715422511100769,
+      "learning_rate": 0.00020091095520777114,
+      "loss": 3.2876,
+      "step": 61750
+    },
+    {
+      "epoch": 6.663072776280323,
+      "grad_norm": 0.7510877251625061,
+      "learning_rate": 0.00020058715596330272,
+      "loss": 3.2817,
+      "step": 61800
+    },
+    {
+      "epoch": 6.668463611859838,
+      "grad_norm": 0.8139302730560303,
+      "learning_rate": 0.0002002633567188343,
+      "loss": 3.2787,
+      "step": 61850
+    },
+    {
+      "epoch": 6.6738544474393535,
+      "grad_norm": 0.7686782479286194,
+      "learning_rate": 0.00019993955747436588,
+      "loss": 3.2758,
+      "step": 61900
+    },
+    {
+      "epoch": 6.679245283018868,
+      "grad_norm": 0.7832411527633667,
+      "learning_rate": 0.00019961575822989745,
+      "loss": 3.2746,
+      "step": 61950
+    },
+    {
+      "epoch": 6.684636118598383,
+      "grad_norm": 0.767423689365387,
+      "learning_rate": 0.00019929195898542903,
+      "loss": 3.3009,
+      "step": 62000
+    },
+    {
+      "epoch": 6.684636118598383,
+      "eval_accuracy": 0.38623570005761865,
+      "eval_loss": 3.362583875656128,
+      "eval_runtime": 185.4691,
+      "eval_samples_per_second": 97.111,
+      "eval_steps_per_second": 6.071,
+      "step": 62000
+    },
+    {
+      "epoch": 6.690026954177897,
+      "grad_norm": 0.8642590045928955,
+      "learning_rate": 0.0001989681597409606,
+      "loss": 3.2834,
+      "step": 62050
+    },
+    {
+      "epoch": 6.695417789757412,
+      "grad_norm": 0.8146366477012634,
+      "learning_rate": 0.00019864436049649216,
+      "loss": 3.2635,
+      "step": 62100
+    },
+    {
+      "epoch": 6.7008086253369274,
+      "grad_norm": 0.8580556511878967,
+      "learning_rate": 0.0001983205612520237,
+      "loss": 3.3066,
+      "step": 62150
+    },
+    {
+      "epoch": 6.706199460916442,
+      "grad_norm": 0.7940399646759033,
+      "learning_rate": 0.0001979967620075553,
+      "loss": 3.2976,
+      "step": 62200
+    },
+    {
+      "epoch": 6.711590296495957,
+      "grad_norm": 0.7609732747077942,
+      "learning_rate": 0.00019767296276308686,
+      "loss": 3.283,
+      "step": 62250
+    },
+    {
+      "epoch": 6.716981132075472,
+      "grad_norm": 0.801042914390564,
+      "learning_rate": 0.00019734916351861844,
+      "loss": 3.2817,
+      "step": 62300
+    },
+    {
+      "epoch": 6.722371967654986,
+      "grad_norm": 0.8280794024467468,
+      "learning_rate": 0.00019702536427415002,
+      "loss": 3.303,
+      "step": 62350
+    },
+    {
+      "epoch": 6.727762803234501,
+      "grad_norm": 0.7966591715812683,
+      "learning_rate": 0.0001967015650296816,
+      "loss": 3.2872,
+      "step": 62400
+    },
+    {
+      "epoch": 6.7331536388140165,
+      "grad_norm": 0.7792521119117737,
+      "learning_rate": 0.00019637776578521315,
+      "loss": 3.2853,
+      "step": 62450
+    },
+    {
+      "epoch": 6.738544474393531,
+      "grad_norm": 0.7710815668106079,
+      "learning_rate": 0.00019605396654074473,
+      "loss": 3.2934,
+      "step": 62500
+    },
+    {
+      "epoch": 6.743935309973046,
+      "grad_norm": 0.8334966897964478,
+      "learning_rate": 0.0001957301672962763,
+      "loss": 3.29,
+      "step": 62550
+    },
+    {
+      "epoch": 6.74932614555256,
+      "grad_norm": 0.7747363448143005,
+      "learning_rate": 0.00019540636805180785,
+      "loss": 3.2811,
+      "step": 62600
+    },
+    {
+      "epoch": 6.754716981132075,
+      "grad_norm": 0.8317515850067139,
+      "learning_rate": 0.00019508256880733943,
+      "loss": 3.2848,
+      "step": 62650
+    },
+    {
+      "epoch": 6.7601078167115904,
+      "grad_norm": 0.7724961638450623,
+      "learning_rate": 0.000194758769562871,
+      "loss": 3.2937,
+      "step": 62700
+    },
+    {
+      "epoch": 6.765498652291106,
+      "grad_norm": 0.8079445362091064,
+      "learning_rate": 0.00019443497031840256,
+      "loss": 3.2725,
+      "step": 62750
+    },
+    {
+      "epoch": 6.77088948787062,
+      "grad_norm": 0.7925755977630615,
+      "learning_rate": 0.00019411117107393414,
+      "loss": 3.2735,
+      "step": 62800
+    },
+    {
+      "epoch": 6.776280323450135,
+      "grad_norm": 0.7883620858192444,
+      "learning_rate": 0.00019378737182946572,
+      "loss": 3.2922,
+      "step": 62850
+    },
+    {
+      "epoch": 6.781671159029649,
+      "grad_norm": 0.9127089381217957,
+      "learning_rate": 0.0001934635725849973,
+      "loss": 3.2878,
+      "step": 62900
+    },
+    {
+      "epoch": 6.787061994609164,
+      "grad_norm": 0.7578275203704834,
+      "learning_rate": 0.00019313977334052887,
+      "loss": 3.3037,
+      "step": 62950
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "grad_norm": 0.7769895792007446,
+      "learning_rate": 0.00019281597409606042,
+      "loss": 3.2808,
+      "step": 63000
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "eval_accuracy": 0.3865326484000156,
+      "eval_loss": 3.360027313232422,
+      "eval_runtime": 185.7201,
+      "eval_samples_per_second": 96.979,
+      "eval_steps_per_second": 6.063,
+      "step": 63000
+    },
+    {
+      "epoch": 6.797843665768194,
+      "grad_norm": 0.8027949929237366,
+      "learning_rate": 0.00019249217485159197,
+      "loss": 3.2927,
+      "step": 63050
+    },
+    {
+      "epoch": 6.803234501347709,
+      "grad_norm": 0.7600734829902649,
+      "learning_rate": 0.00019216837560712355,
+      "loss": 3.2988,
+      "step": 63100
+    },
+    {
+      "epoch": 6.808625336927224,
+      "grad_norm": 0.7969046831130981,
+      "learning_rate": 0.00019184457636265513,
+      "loss": 3.2763,
+      "step": 63150
+    },
+    {
+      "epoch": 6.814016172506738,
+      "grad_norm": 0.8030910491943359,
+      "learning_rate": 0.0001915207771181867,
+      "loss": 3.2912,
+      "step": 63200
+    },
+    {
+      "epoch": 6.819407008086253,
+      "grad_norm": 0.7622592449188232,
+      "learning_rate": 0.00019119697787371828,
+      "loss": 3.2911,
+      "step": 63250
+    },
+    {
+      "epoch": 6.824797843665769,
+      "grad_norm": 0.7882322669029236,
+      "learning_rate": 0.00019087317862924986,
+      "loss": 3.2824,
+      "step": 63300
+    },
+    {
+      "epoch": 6.830188679245283,
+      "grad_norm": 0.7555896639823914,
+      "learning_rate": 0.00019054937938478144,
+      "loss": 3.2661,
+      "step": 63350
+    },
+    {
+      "epoch": 6.835579514824798,
+      "grad_norm": 0.7905471920967102,
+      "learning_rate": 0.00019022558014031302,
+      "loss": 3.2966,
+      "step": 63400
+    },
+    {
+      "epoch": 6.840970350404312,
+      "grad_norm": 0.7742801904678345,
+      "learning_rate": 0.00018990178089584454,
+      "loss": 3.2994,
+      "step": 63450
+    },
+    {
+      "epoch": 6.846361185983827,
+      "grad_norm": 0.7538597583770752,
+      "learning_rate": 0.00018957798165137612,
+      "loss": 3.2933,
+      "step": 63500
+    },
+    {
+      "epoch": 6.8517520215633425,
+      "grad_norm": 0.7583889365196228,
+      "learning_rate": 0.0001892541824069077,
+      "loss": 3.2884,
+      "step": 63550
+    },
+    {
+      "epoch": 6.857142857142857,
+      "grad_norm": 0.7588804364204407,
+      "learning_rate": 0.00018893038316243927,
+      "loss": 3.2693,
+      "step": 63600
+    },
+    {
+      "epoch": 6.862533692722372,
+      "grad_norm": 0.7821441292762756,
+      "learning_rate": 0.00018860658391797085,
+      "loss": 3.2706,
+      "step": 63650
+    },
+    {
+      "epoch": 6.867924528301887,
+      "grad_norm": 0.7843536734580994,
+      "learning_rate": 0.00018828278467350243,
+      "loss": 3.2791,
+      "step": 63700
+    },
+    {
+      "epoch": 6.873315363881401,
+      "grad_norm": 0.7533359527587891,
+      "learning_rate": 0.000187958985429034,
+      "loss": 3.2791,
+      "step": 63750
+    },
+    {
+      "epoch": 6.878706199460916,
+      "grad_norm": 0.8090384006500244,
+      "learning_rate": 0.00018763518618456556,
+      "loss": 3.2695,
+      "step": 63800
+    },
+    {
+      "epoch": 6.884097035040432,
+      "grad_norm": 0.8591648936271667,
+      "learning_rate": 0.0001873113869400971,
+      "loss": 3.2881,
+      "step": 63850
+    },
+    {
+      "epoch": 6.889487870619946,
+      "grad_norm": 0.8102157711982727,
+      "learning_rate": 0.00018698758769562868,
+      "loss": 3.2864,
+      "step": 63900
+    },
+    {
+      "epoch": 6.894878706199461,
+      "grad_norm": 0.8191260099411011,
+      "learning_rate": 0.00018666378845116026,
+      "loss": 3.3036,
+      "step": 63950
+    },
+    {
+      "epoch": 6.900269541778976,
+      "grad_norm": 0.7749097943305969,
+      "learning_rate": 0.00018633998920669184,
+      "loss": 3.2885,
+      "step": 64000
+    },
+    {
+      "epoch": 6.900269541778976,
+      "eval_accuracy": 0.387105792383237,
+      "eval_loss": 3.3557851314544678,
+      "eval_runtime": 185.6126,
+      "eval_samples_per_second": 97.035,
+      "eval_steps_per_second": 6.066,
+      "step": 64000
+    },
+    {
+      "epoch": 6.90566037735849,
+      "grad_norm": 0.783939778804779,
+      "learning_rate": 0.00018601618996222342,
+      "loss": 3.2897,
+      "step": 64050
+    },
+    {
+      "epoch": 6.9110512129380055,
+      "grad_norm": 0.8147432208061218,
+      "learning_rate": 0.00018569239071775497,
+      "loss": 3.2957,
+      "step": 64100
+    },
+    {
+      "epoch": 6.916442048517521,
+      "grad_norm": 0.8287383913993835,
+      "learning_rate": 0.00018536859147328655,
+      "loss": 3.2897,
+      "step": 64150
+    },
+    {
+      "epoch": 6.921832884097035,
+      "grad_norm": 0.7737496495246887,
+      "learning_rate": 0.00018504479222881812,
+      "loss": 3.2824,
+      "step": 64200
+    },
+    {
+      "epoch": 6.92722371967655,
+      "grad_norm": 0.8038462996482849,
+      "learning_rate": 0.0001847209929843497,
+      "loss": 3.2657,
+      "step": 64250
+    },
+    {
+      "epoch": 6.932614555256064,
+      "grad_norm": 0.7752224802970886,
+      "learning_rate": 0.00018439719373988125,
+      "loss": 3.307,
+      "step": 64300
+    },
+    {
+      "epoch": 6.938005390835579,
+      "grad_norm": 0.7654508352279663,
+      "learning_rate": 0.00018407339449541283,
+      "loss": 3.2857,
+      "step": 64350
+    },
+    {
+      "epoch": 6.943396226415095,
+      "grad_norm": 0.7717485427856445,
+      "learning_rate": 0.00018374959525094438,
+      "loss": 3.286,
+      "step": 64400
+    },
+    {
+      "epoch": 6.948787061994609,
+      "grad_norm": 0.775194525718689,
+      "learning_rate": 0.00018342579600647596,
+      "loss": 3.2815,
+      "step": 64450
+    },
+    {
+      "epoch": 6.954177897574124,
+      "grad_norm": 0.7448485493659973,
+      "learning_rate": 0.00018310199676200753,
+      "loss": 3.2772,
+      "step": 64500
+    },
+    {
+      "epoch": 6.959568733153639,
+      "grad_norm": 0.7653682231903076,
+      "learning_rate": 0.0001827781975175391,
+      "loss": 3.2832,
+      "step": 64550
+    },
+    {
+      "epoch": 6.964959568733153,
+      "grad_norm": 0.7664286494255066,
+      "learning_rate": 0.0001824543982730707,
+      "loss": 3.2747,
+      "step": 64600
+    },
+    {
+      "epoch": 6.9703504043126685,
+      "grad_norm": 0.7716772556304932,
+      "learning_rate": 0.00018213059902860227,
+      "loss": 3.2868,
+      "step": 64650
+    },
+    {
+      "epoch": 6.975741239892184,
+      "grad_norm": 0.7692857980728149,
+      "learning_rate": 0.00018180679978413382,
+      "loss": 3.282,
+      "step": 64700
+    },
+    {
+      "epoch": 6.981132075471698,
+      "grad_norm": 0.8032350540161133,
+      "learning_rate": 0.00018148300053966537,
+      "loss": 3.2926,
+      "step": 64750
+    },
+    {
+      "epoch": 6.986522911051213,
+      "grad_norm": 0.7604122757911682,
+      "learning_rate": 0.00018115920129519695,
+      "loss": 3.279,
+      "step": 64800
+    },
+    {
+      "epoch": 6.991913746630727,
+      "grad_norm": 0.7604785561561584,
+      "learning_rate": 0.00018083540205072852,
+      "loss": 3.2744,
+      "step": 64850
+    },
+    {
+      "epoch": 6.997304582210242,
+      "grad_norm": 0.8146815299987793,
+      "learning_rate": 0.0001805116028062601,
+      "loss": 3.2846,
+      "step": 64900
+    },
+    {
+      "epoch": 7.002695417789758,
+      "grad_norm": 0.817147433757782,
+      "learning_rate": 0.00018019427954668105,
+      "loss": 3.2321,
+      "step": 64950
+    },
+    {
+      "epoch": 7.008086253369272,
+      "grad_norm": 0.7436534762382507,
+      "learning_rate": 0.00017987048030221263,
+      "loss": 3.1848,
+      "step": 65000
+    },
+    {
+      "epoch": 7.008086253369272,
+      "eval_accuracy": 0.3869140200362444,
+      "eval_loss": 3.358355760574341,
+      "eval_runtime": 185.9532,
+      "eval_samples_per_second": 96.858,
+      "eval_steps_per_second": 6.055,
+      "step": 65000
+    },
+    {
+      "epoch": 7.013477088948787,
+      "grad_norm": 0.8559082746505737,
+      "learning_rate": 0.00017954668105774416,
+      "loss": 3.2084,
+      "step": 65050
+    },
+    {
+      "epoch": 7.018867924528302,
+      "grad_norm": 0.8314744234085083,
+      "learning_rate": 0.00017922288181327573,
+      "loss": 3.2043,
+      "step": 65100
+    },
+    {
+      "epoch": 7.024258760107816,
+      "grad_norm": 0.7711519598960876,
+      "learning_rate": 0.0001788990825688073,
+      "loss": 3.2146,
+      "step": 65150
+    },
+    {
+      "epoch": 7.0296495956873315,
+      "grad_norm": 0.7674431204795837,
+      "learning_rate": 0.0001785752833243389,
+      "loss": 3.2077,
+      "step": 65200
+    },
+    {
+      "epoch": 7.035040431266847,
+      "grad_norm": 0.8487887978553772,
+      "learning_rate": 0.00017825148407987047,
+      "loss": 3.2071,
+      "step": 65250
+    },
+    {
+      "epoch": 7.040431266846361,
+      "grad_norm": 0.8080512881278992,
+      "learning_rate": 0.00017792768483540204,
+      "loss": 3.2003,
+      "step": 65300
+    },
+    {
+      "epoch": 7.045822102425876,
+      "grad_norm": 0.7953256368637085,
+      "learning_rate": 0.00017760388559093362,
+      "loss": 3.2153,
+      "step": 65350
+    },
+    {
+      "epoch": 7.051212938005391,
+      "grad_norm": 0.7839633226394653,
+      "learning_rate": 0.0001772800863464652,
+      "loss": 3.1939,
+      "step": 65400
+    },
+    {
+      "epoch": 7.056603773584905,
+      "grad_norm": 0.7950543165206909,
+      "learning_rate": 0.00017695628710199672,
+      "loss": 3.1827,
+      "step": 65450
+    },
+    {
+      "epoch": 7.061994609164421,
+      "grad_norm": 0.8554292321205139,
+      "learning_rate": 0.0001766324878575283,
+      "loss": 3.202,
+      "step": 65500
+    },
+    {
+      "epoch": 7.067385444743936,
+      "grad_norm": 0.7931976914405823,
+      "learning_rate": 0.00017630868861305988,
+      "loss": 3.192,
+      "step": 65550
+    },
+    {
+      "epoch": 7.07277628032345,
+      "grad_norm": 0.7638124823570251,
+      "learning_rate": 0.00017598488936859146,
+      "loss": 3.2201,
+      "step": 65600
+    },
+    {
+      "epoch": 7.078167115902965,
+      "grad_norm": 0.7745422124862671,
+      "learning_rate": 0.00017566109012412303,
+      "loss": 3.2158,
+      "step": 65650
+    },
+    {
+      "epoch": 7.083557951482479,
+      "grad_norm": 0.8866512179374695,
+      "learning_rate": 0.0001753372908796546,
+      "loss": 3.2082,
+      "step": 65700
+    },
+    {
+      "epoch": 7.0889487870619945,
+      "grad_norm": 0.8236171007156372,
+      "learning_rate": 0.0001750134916351862,
+      "loss": 3.1911,
+      "step": 65750
+    },
+    {
+      "epoch": 7.09433962264151,
+      "grad_norm": 0.7753577828407288,
+      "learning_rate": 0.00017468969239071774,
+      "loss": 3.2218,
+      "step": 65800
+    },
+    {
+      "epoch": 7.099730458221024,
+      "grad_norm": 0.8109624981880188,
+      "learning_rate": 0.00017436589314624932,
+      "loss": 3.2192,
+      "step": 65850
+    },
+    {
+      "epoch": 7.105121293800539,
+      "grad_norm": 0.7832379937171936,
+      "learning_rate": 0.00017404209390178087,
+      "loss": 3.2301,
+      "step": 65900
+    },
+    {
+      "epoch": 7.110512129380054,
+      "grad_norm": 0.8430277705192566,
+      "learning_rate": 0.00017371829465731244,
+      "loss": 3.2167,
+      "step": 65950
+    },
+    {
+      "epoch": 7.115902964959568,
+      "grad_norm": 0.8423453569412231,
+      "learning_rate": 0.00017339449541284402,
+      "loss": 3.2273,
+      "step": 66000
+    },
+    {
+      "epoch": 7.115902964959568,
+      "eval_accuracy": 0.387448374938516,
+      "eval_loss": 3.359693765640259,
+      "eval_runtime": 185.5479,
+      "eval_samples_per_second": 97.069,
+      "eval_steps_per_second": 6.069,
+      "step": 66000
+    },
+    {
+      "epoch": 7.121293800539084,
+      "grad_norm": 0.7843069434165955,
+      "learning_rate": 0.0001730706961683756,
+      "loss": 3.2199,
+      "step": 66050
+    },
+    {
+      "epoch": 7.126684636118599,
+      "grad_norm": 0.8226370215415955,
+      "learning_rate": 0.00017274689692390715,
+      "loss": 3.2214,
+      "step": 66100
+    },
+    {
+      "epoch": 7.132075471698113,
+      "grad_norm": 0.8398003578186035,
+      "learning_rate": 0.00017242309767943873,
+      "loss": 3.2192,
+      "step": 66150
+    },
+    {
+      "epoch": 7.137466307277628,
+      "grad_norm": 0.8124863505363464,
+      "learning_rate": 0.0001720992984349703,
+      "loss": 3.2173,
+      "step": 66200
+    },
+    {
+      "epoch": 7.142857142857143,
+      "grad_norm": 0.8067982196807861,
+      "learning_rate": 0.00017177549919050188,
+      "loss": 3.2257,
+      "step": 66250
+    },
+    {
+      "epoch": 7.1482479784366575,
+      "grad_norm": 0.8476393818855286,
+      "learning_rate": 0.00017145169994603346,
+      "loss": 3.2198,
+      "step": 66300
+    },
+    {
+      "epoch": 7.153638814016173,
+      "grad_norm": 0.7765963077545166,
+      "learning_rate": 0.000171127900701565,
+      "loss": 3.2257,
+      "step": 66350
+    },
+    {
+      "epoch": 7.159029649595688,
+      "grad_norm": 0.8154024481773376,
+      "learning_rate": 0.0001708041014570966,
+      "loss": 3.209,
+      "step": 66400
+    },
+    {
+      "epoch": 7.164420485175202,
+      "grad_norm": 0.7860375642776489,
+      "learning_rate": 0.00017048030221262814,
+      "loss": 3.2185,
+      "step": 66450
+    },
+    {
+      "epoch": 7.169811320754717,
+      "grad_norm": 0.795101523399353,
+      "learning_rate": 0.00017015650296815972,
+      "loss": 3.2204,
+      "step": 66500
+    },
+    {
+      "epoch": 7.175202156334231,
+      "grad_norm": 0.7983394861221313,
+      "learning_rate": 0.0001698327037236913,
+      "loss": 3.2229,
+      "step": 66550
+    },
+    {
+      "epoch": 7.180592991913747,
+      "grad_norm": 0.7524290680885315,
+      "learning_rate": 0.00016950890447922287,
+      "loss": 3.2088,
+      "step": 66600
+    },
+    {
+      "epoch": 7.185983827493262,
+      "grad_norm": 0.7941445112228394,
+      "learning_rate": 0.00016918510523475445,
+      "loss": 3.2178,
+      "step": 66650
+    },
+    {
+      "epoch": 7.191374663072776,
+      "grad_norm": 0.7590154409408569,
+      "learning_rate": 0.00016886130599028603,
+      "loss": 3.22,
+      "step": 66700
+    },
+    {
+      "epoch": 7.196765498652291,
+      "grad_norm": 0.795070469379425,
+      "learning_rate": 0.00016853750674581755,
+      "loss": 3.2244,
+      "step": 66750
+    },
+    {
+      "epoch": 7.202156334231806,
+      "grad_norm": 0.7601382732391357,
+      "learning_rate": 0.00016821370750134913,
+      "loss": 3.2127,
+      "step": 66800
+    },
+    {
+      "epoch": 7.2075471698113205,
+      "grad_norm": 0.7475704550743103,
+      "learning_rate": 0.0001678899082568807,
+      "loss": 3.2314,
+      "step": 66850
+    },
+    {
+      "epoch": 7.212938005390836,
+      "grad_norm": 0.8114707469940186,
+      "learning_rate": 0.00016756610901241228,
+      "loss": 3.24,
+      "step": 66900
+    },
+    {
+      "epoch": 7.218328840970351,
+      "grad_norm": 0.7982012033462524,
+      "learning_rate": 0.00016724230976794386,
+      "loss": 3.2292,
+      "step": 66950
+    },
+    {
+      "epoch": 7.223719676549865,
+      "grad_norm": 0.7660938501358032,
+      "learning_rate": 0.00016691851052347544,
+      "loss": 3.2312,
+      "step": 67000
+    },
+    {
+      "epoch": 7.223719676549865,
+      "eval_accuracy": 0.3875677844622242,
+      "eval_loss": 3.3567051887512207,
+      "eval_runtime": 185.7015,
+      "eval_samples_per_second": 96.989,
+      "eval_steps_per_second": 6.063,
+      "step": 67000
+    },
+    {
+      "epoch": 7.22911051212938,
+      "grad_norm": 0.8399280309677124,
+      "learning_rate": 0.00016659471127900702,
+      "loss": 3.203,
+      "step": 67050
+    },
+    {
+      "epoch": 7.234501347708895,
+      "grad_norm": 0.7751797437667847,
+      "learning_rate": 0.0001662709120345386,
+      "loss": 3.2442,
+      "step": 67100
+    },
+    {
+      "epoch": 7.2398921832884096,
+      "grad_norm": 0.7996171116828918,
+      "learning_rate": 0.00016594711279007015,
+      "loss": 3.207,
+      "step": 67150
+    },
+    {
+      "epoch": 7.245283018867925,
+      "grad_norm": 0.8425281047821045,
+      "learning_rate": 0.0001656233135456017,
+      "loss": 3.2259,
+      "step": 67200
+    },
+    {
+      "epoch": 7.250673854447439,
+      "grad_norm": 0.8230305910110474,
+      "learning_rate": 0.00016529951430113327,
+      "loss": 3.2384,
+      "step": 67250
+    },
+    {
+      "epoch": 7.256064690026954,
+      "grad_norm": 0.7822522521018982,
+      "learning_rate": 0.00016497571505666485,
+      "loss": 3.2411,
+      "step": 67300
+    },
+    {
+      "epoch": 7.261455525606469,
+      "grad_norm": 0.7796124219894409,
+      "learning_rate": 0.00016465191581219643,
+      "loss": 3.22,
+      "step": 67350
+    },
+    {
+      "epoch": 7.2668463611859835,
+      "grad_norm": 0.7988195419311523,
+      "learning_rate": 0.000164328116567728,
+      "loss": 3.2445,
+      "step": 67400
+    },
+    {
+      "epoch": 7.272237196765499,
+      "grad_norm": 0.8658559918403625,
+      "learning_rate": 0.00016400431732325956,
+      "loss": 3.2413,
+      "step": 67450
+    },
+    {
+      "epoch": 7.277628032345014,
+      "grad_norm": 0.7902882099151611,
+      "learning_rate": 0.00016368051807879114,
+      "loss": 3.2314,
+      "step": 67500
+    },
+    {
+      "epoch": 7.283018867924528,
+      "grad_norm": 0.854999840259552,
+      "learning_rate": 0.0001633567188343227,
+      "loss": 3.2402,
+      "step": 67550
+    },
+    {
+      "epoch": 7.288409703504043,
+      "grad_norm": 0.8116405010223389,
+      "learning_rate": 0.00016303291958985426,
+      "loss": 3.2278,
+      "step": 67600
+    },
+    {
+      "epoch": 7.293800539083558,
+      "grad_norm": 0.8255453705787659,
+      "learning_rate": 0.00016270912034538584,
+      "loss": 3.2305,
+      "step": 67650
+    },
+    {
+      "epoch": 7.2991913746630726,
+      "grad_norm": 0.780562162399292,
+      "learning_rate": 0.00016238532110091742,
+      "loss": 3.2094,
+      "step": 67700
+    },
+    {
+      "epoch": 7.304582210242588,
+      "grad_norm": 0.7962255477905273,
+      "learning_rate": 0.000162061521856449,
+      "loss": 3.2278,
+      "step": 67750
+    },
+    {
+      "epoch": 7.309973045822103,
+      "grad_norm": 0.8192128539085388,
+      "learning_rate": 0.00016173772261198055,
+      "loss": 3.2252,
+      "step": 67800
+    },
+    {
+      "epoch": 7.315363881401617,
+      "grad_norm": 0.9176713824272156,
+      "learning_rate": 0.00016141392336751212,
+      "loss": 3.2249,
+      "step": 67850
+    },
+    {
+      "epoch": 7.320754716981132,
+      "grad_norm": 0.7980507612228394,
+      "learning_rate": 0.0001610901241230437,
+      "loss": 3.2384,
+      "step": 67900
+    },
+    {
+      "epoch": 7.3261455525606465,
+      "grad_norm": 0.7474843859672546,
+      "learning_rate": 0.00016076632487857528,
+      "loss": 3.2137,
+      "step": 67950
+    },
+    {
+      "epoch": 7.331536388140162,
+      "grad_norm": 0.8784814476966858,
+      "learning_rate": 0.00016044252563410686,
+      "loss": 3.2312,
+      "step": 68000
+    },
+    {
+      "epoch": 7.331536388140162,
+      "eval_accuracy": 0.38756169990050654,
+      "eval_loss": 3.356365442276001,
+      "eval_runtime": 185.9577,
+      "eval_samples_per_second": 96.855,
+      "eval_steps_per_second": 6.055,
+      "step": 68000
+    },
+    {
+      "epoch": 7.336927223719677,
+      "grad_norm": 0.7711509466171265,
+      "learning_rate": 0.0001601187263896384,
+      "loss": 3.2345,
+      "step": 68050
+    },
+    {
+      "epoch": 7.342318059299191,
+      "grad_norm": 0.8484320640563965,
+      "learning_rate": 0.00015979492714516996,
+      "loss": 3.2397,
+      "step": 68100
+    },
+    {
+      "epoch": 7.347708894878706,
+      "grad_norm": 0.8413980603218079,
+      "learning_rate": 0.00015947112790070154,
+      "loss": 3.2292,
+      "step": 68150
+    },
+    {
+      "epoch": 7.353099730458221,
+      "grad_norm": 0.7935882210731506,
+      "learning_rate": 0.00015914732865623311,
+      "loss": 3.2478,
+      "step": 68200
+    },
+    {
+      "epoch": 7.3584905660377355,
+      "grad_norm": 0.7697828412055969,
+      "learning_rate": 0.0001588235294117647,
+      "loss": 3.2341,
+      "step": 68250
+    },
+    {
+      "epoch": 7.363881401617251,
+      "grad_norm": 0.7938187122344971,
+      "learning_rate": 0.00015849973016729627,
+      "loss": 3.2208,
+      "step": 68300
+    },
+    {
+      "epoch": 7.369272237196766,
+      "grad_norm": 0.7700347304344177,
+      "learning_rate": 0.00015817593092282785,
+      "loss": 3.2324,
+      "step": 68350
+    },
+    {
+      "epoch": 7.37466307277628,
+      "grad_norm": 0.8395536541938782,
+      "learning_rate": 0.00015785213167835942,
+      "loss": 3.2272,
+      "step": 68400
+    },
+    {
+      "epoch": 7.380053908355795,
+      "grad_norm": 0.8341064453125,
+      "learning_rate": 0.00015752833243389095,
+      "loss": 3.2111,
+      "step": 68450
+    },
+    {
+      "epoch": 7.38544474393531,
+      "grad_norm": 0.8286171555519104,
+      "learning_rate": 0.00015720453318942253,
+      "loss": 3.233,
+      "step": 68500
+    },
+    {
+      "epoch": 7.390835579514825,
+      "grad_norm": 0.7822830080986023,
+      "learning_rate": 0.0001568807339449541,
+      "loss": 3.232,
+      "step": 68550
+    },
+    {
+      "epoch": 7.39622641509434,
+      "grad_norm": 0.7763746976852417,
+      "learning_rate": 0.00015655693470048568,
+      "loss": 3.2286,
+      "step": 68600
+    },
+    {
+      "epoch": 7.401617250673855,
+      "grad_norm": 0.7900704145431519,
+      "learning_rate": 0.00015623313545601726,
+      "loss": 3.2275,
+      "step": 68650
+    },
+    {
+      "epoch": 7.407008086253369,
+      "grad_norm": 0.7560325264930725,
+      "learning_rate": 0.00015590933621154884,
+      "loss": 3.2388,
+      "step": 68700
+    },
+    {
+      "epoch": 7.412398921832884,
+      "grad_norm": 0.8804982304573059,
+      "learning_rate": 0.00015558553696708041,
+      "loss": 3.225,
+      "step": 68750
+    },
+    {
+      "epoch": 7.4177897574123985,
+      "grad_norm": 0.8142456412315369,
+      "learning_rate": 0.00015526173772261196,
+      "loss": 3.2347,
+      "step": 68800
+    },
+    {
+      "epoch": 7.423180592991914,
+      "grad_norm": 0.8332946300506592,
+      "learning_rate": 0.00015493793847814354,
+      "loss": 3.2391,
+      "step": 68850
+    },
+    {
+      "epoch": 7.428571428571429,
+      "grad_norm": 0.7892976403236389,
+      "learning_rate": 0.0001546141392336751,
+      "loss": 3.2327,
+      "step": 68900
+    },
+    {
+      "epoch": 7.433962264150943,
+      "grad_norm": 0.8157640099525452,
+      "learning_rate": 0.00015429681597409605,
+      "loss": 3.2347,
+      "step": 68950
+    },
+    {
+      "epoch": 7.439353099730458,
+      "grad_norm": 0.7988810539245605,
+      "learning_rate": 0.00015397301672962762,
+      "loss": 3.2206,
+      "step": 69000
+    },
+    {
+      "epoch": 7.439353099730458,
+      "eval_accuracy": 0.38823252282987314,
+      "eval_loss": 3.3501293659210205,
+      "eval_runtime": 185.6242,
+      "eval_samples_per_second": 97.029,
+      "eval_steps_per_second": 6.066,
+      "step": 69000
+    },
+    {
+      "epoch": 7.444743935309973,
+      "grad_norm": 0.8131216764450073,
+      "learning_rate": 0.0001536492174851592,
+      "loss": 3.2219,
+      "step": 69050
+    },
+    {
+      "epoch": 7.450134770889488,
+      "grad_norm": 0.8403545022010803,
+      "learning_rate": 0.00015332541824069078,
+      "loss": 3.2298,
+      "step": 69100
+    },
+    {
+      "epoch": 7.455525606469003,
+      "grad_norm": 0.815885603427887,
+      "learning_rate": 0.00015300161899622233,
+      "loss": 3.2323,
+      "step": 69150
+    },
+    {
+      "epoch": 7.460916442048518,
+      "grad_norm": 1.0801024436950684,
+      "learning_rate": 0.00015267781975175388,
+      "loss": 3.2211,
+      "step": 69200
+    },
+    {
+      "epoch": 7.466307277628032,
+      "grad_norm": 0.8832733035087585,
+      "learning_rate": 0.00015235402050728546,
+      "loss": 3.2386,
+      "step": 69250
+    },
+    {
+      "epoch": 7.471698113207547,
+      "grad_norm": 0.8115150928497314,
+      "learning_rate": 0.00015203022126281703,
+      "loss": 3.2263,
+      "step": 69300
+    },
+    {
+      "epoch": 7.4770889487870615,
+      "grad_norm": 0.7896444797515869,
+      "learning_rate": 0.0001517064220183486,
+      "loss": 3.2395,
+      "step": 69350
+    },
+    {
+      "epoch": 7.482479784366577,
+      "grad_norm": 0.7713582515716553,
+      "learning_rate": 0.0001513826227738802,
+      "loss": 3.2239,
+      "step": 69400
+    },
+    {
+      "epoch": 7.487870619946092,
+      "grad_norm": 0.8341700434684753,
+      "learning_rate": 0.00015105882352941177,
+      "loss": 3.2239,
+      "step": 69450
+    },
+    {
+      "epoch": 7.493261455525606,
+      "grad_norm": 0.8017354011535645,
+      "learning_rate": 0.00015073502428494332,
+      "loss": 3.2248,
+      "step": 69500
+    },
+    {
+      "epoch": 7.498652291105121,
+      "grad_norm": 0.8698682188987732,
+      "learning_rate": 0.0001504112250404749,
+      "loss": 3.238,
+      "step": 69550
+    },
+    {
+      "epoch": 7.504043126684636,
+      "grad_norm": 0.796074390411377,
+      "learning_rate": 0.00015008742579600647,
+      "loss": 3.2117,
+      "step": 69600
+    },
+    {
+      "epoch": 7.509433962264151,
+      "grad_norm": 0.8275701999664307,
+      "learning_rate": 0.00014976362655153802,
+      "loss": 3.2317,
+      "step": 69650
+    },
+    {
+      "epoch": 7.514824797843666,
+      "grad_norm": 0.7581632137298584,
+      "learning_rate": 0.0001494398273070696,
+      "loss": 3.2388,
+      "step": 69700
+    },
+    {
+      "epoch": 7.520215633423181,
+      "grad_norm": NaN,
+      "learning_rate": 0.00014912250404749055,
+      "loss": 3.2454,
+      "step": 69750
+    },
+    {
+      "epoch": 7.525606469002695,
+      "grad_norm": 0.8626130223274231,
+      "learning_rate": 0.0001487987048030221,
+      "loss": 3.2268,
+      "step": 69800
+    },
+    {
+      "epoch": 7.53099730458221,
+      "grad_norm": 0.7908021211624146,
+      "learning_rate": 0.00014847490555855368,
+      "loss": 3.2295,
+      "step": 69850
+    },
+    {
+      "epoch": 7.536388140161725,
+      "grad_norm": 0.839931309223175,
+      "learning_rate": 0.00014815110631408526,
+      "loss": 3.2329,
+      "step": 69900
+    },
+    {
+      "epoch": 7.54177897574124,
+      "grad_norm": 0.823737621307373,
+      "learning_rate": 0.0001478273070696168,
+      "loss": 3.2273,
+      "step": 69950
+    },
+    {
+      "epoch": 7.547169811320755,
+      "grad_norm": 0.8635311126708984,
+      "learning_rate": 0.0001475035078251484,
+      "loss": 3.2253,
+      "step": 70000
+    },
+    {
+      "epoch": 7.547169811320755,
+      "eval_accuracy": 0.38823176225965844,
+      "eval_loss": 3.347209930419922,
+      "eval_runtime": 185.658,
+      "eval_samples_per_second": 97.012,
+      "eval_steps_per_second": 6.065,
+      "step": 70000
+    },
+    {
+      "epoch": 7.55256064690027,
+      "grad_norm": 0.8203630447387695,
+      "learning_rate": 0.00014717970858067997,
+      "loss": 3.235,
+      "step": 70050
+    },
+    {
+      "epoch": 7.557951482479784,
+      "grad_norm": 1.1502406597137451,
+      "learning_rate": 0.00014685590933621154,
+      "loss": 3.2361,
+      "step": 70100
+    },
+    {
+      "epoch": 7.563342318059299,
+      "grad_norm": 0.8077203631401062,
+      "learning_rate": 0.0001465321100917431,
+      "loss": 3.2361,
+      "step": 70150
+    },
+    {
+      "epoch": 7.568733153638814,
+      "grad_norm": 0.7776874303817749,
+      "learning_rate": 0.00014620831084727467,
+      "loss": 3.2188,
+      "step": 70200
+    },
+    {
+      "epoch": 7.574123989218329,
+      "grad_norm": 0.8516008853912354,
+      "learning_rate": 0.00014588451160280625,
+      "loss": 3.2334,
+      "step": 70250
+    },
+    {
+      "epoch": 7.579514824797844,
+      "grad_norm": 0.758550763130188,
+      "learning_rate": 0.0001455607123583378,
+      "loss": 3.2315,
+      "step": 70300
+    },
+    {
+      "epoch": 7.584905660377358,
+      "grad_norm": 0.8415988087654114,
+      "learning_rate": 0.00014523691311386938,
+      "loss": 3.246,
+      "step": 70350
+    },
+    {
+      "epoch": 7.590296495956873,
+      "grad_norm": 0.8056899309158325,
+      "learning_rate": 0.00014491311386940096,
+      "loss": 3.2286,
+      "step": 70400
+    },
+    {
+      "epoch": 7.595687331536388,
+      "grad_norm": 0.783736526966095,
+      "learning_rate": 0.00014458931462493253,
+      "loss": 3.2383,
+      "step": 70450
+    },
+    {
+      "epoch": 7.601078167115903,
+      "grad_norm": 0.8655029535293579,
+      "learning_rate": 0.00014427199136535346,
+      "loss": 3.2426,
+      "step": 70500
+    },
+    {
+      "epoch": 7.606469002695418,
+      "grad_norm": 0.7884918451309204,
+      "learning_rate": 0.00014394819212088504,
+      "loss": 3.2309,
+      "step": 70550
+    },
+    {
+      "epoch": 7.611859838274933,
+      "grad_norm": 0.8152871131896973,
+      "learning_rate": 0.0001436243928764166,
+      "loss": 3.2311,
+      "step": 70600
+    },
+    {
+      "epoch": 7.617250673854447,
+      "grad_norm": 0.8208385705947876,
+      "learning_rate": 0.00014330059363194816,
+      "loss": 3.2304,
+      "step": 70650
+    },
+    {
+      "epoch": 7.622641509433962,
+      "grad_norm": 0.8047370910644531,
+      "learning_rate": 0.00014297679438747974,
+      "loss": 3.2333,
+      "step": 70700
+    },
+    {
+      "epoch": 7.628032345013477,
+      "grad_norm": 0.8149837851524353,
+      "learning_rate": 0.00014265299514301132,
+      "loss": 3.2333,
+      "step": 70750
+    },
+    {
+      "epoch": 7.633423180592992,
+      "grad_norm": 0.8267185688018799,
+      "learning_rate": 0.0001423291958985429,
+      "loss": 3.2466,
+      "step": 70800
+    },
+    {
+      "epoch": 7.638814016172507,
+      "grad_norm": 0.8208780884742737,
+      "learning_rate": 0.00014200539665407445,
+      "loss": 3.2227,
+      "step": 70850
+    },
+    {
+      "epoch": 7.644204851752022,
+      "grad_norm": 0.8307817578315735,
+      "learning_rate": 0.00014168159740960602,
+      "loss": 3.2405,
+      "step": 70900
+    },
+    {
+      "epoch": 7.649595687331536,
+      "grad_norm": 0.814270555973053,
+      "learning_rate": 0.0001413577981651376,
+      "loss": 3.2215,
+      "step": 70950
+    },
+    {
+      "epoch": 7.654986522911051,
+      "grad_norm": 0.8136581182479858,
+      "learning_rate": 0.00014103399892066918,
+      "loss": 3.2527,
+      "step": 71000
+    },
+    {
+      "epoch": 7.654986522911051,
+      "eval_accuracy": 0.38887846424793027,
+      "eval_loss": 3.3424172401428223,
+      "eval_runtime": 185.9073,
+      "eval_samples_per_second": 96.882,
+      "eval_steps_per_second": 6.057,
+      "step": 71000
+    },
+    {
+      "epoch": 7.660377358490566,
+      "grad_norm": 0.8112232089042664,
+      "learning_rate": 0.00014071019967620073,
+      "loss": 3.2448,
+      "step": 71050
+    },
+    {
+      "epoch": 7.665768194070081,
+      "grad_norm": 0.8638885021209717,
+      "learning_rate": 0.0001403864004317323,
+      "loss": 3.2538,
+      "step": 71100
+    },
+    {
+      "epoch": 7.671159029649596,
+      "grad_norm": 0.8865724205970764,
+      "learning_rate": 0.00014006260118726389,
+      "loss": 3.2333,
+      "step": 71150
+    },
+    {
+      "epoch": 7.67654986522911,
+      "grad_norm": 0.835533082485199,
+      "learning_rate": 0.00013973880194279546,
+      "loss": 3.2504,
+      "step": 71200
+    },
+    {
+      "epoch": 7.681940700808625,
+      "grad_norm": 0.8290038704872131,
+      "learning_rate": 0.00013941500269832704,
+      "loss": 3.238,
+      "step": 71250
+    },
+    {
+      "epoch": 7.6873315363881405,
+      "grad_norm": 0.8244901299476624,
+      "learning_rate": 0.0001390912034538586,
+      "loss": 3.2364,
+      "step": 71300
+    },
+    {
+      "epoch": 7.692722371967655,
+      "grad_norm": 0.8304265141487122,
+      "learning_rate": 0.00013876740420939017,
+      "loss": 3.2484,
+      "step": 71350
+    },
+    {
+      "epoch": 7.69811320754717,
+      "grad_norm": 0.8117026686668396,
+      "learning_rate": 0.00013844360496492175,
+      "loss": 3.2539,
+      "step": 71400
+    },
+    {
+      "epoch": 7.703504043126685,
+      "grad_norm": 0.7902008295059204,
+      "learning_rate": 0.00013811980572045333,
+      "loss": 3.2291,
+      "step": 71450
+    },
+    {
+      "epoch": 7.708894878706199,
+      "grad_norm": 0.8188857436180115,
+      "learning_rate": 0.00013779600647598488,
+      "loss": 3.2401,
+      "step": 71500
+    },
+    {
+      "epoch": 7.714285714285714,
+      "grad_norm": 0.7785261273384094,
+      "learning_rate": 0.00013747220723151645,
+      "loss": 3.2294,
+      "step": 71550
+    },
+    {
+      "epoch": 7.719676549865229,
+      "grad_norm": 0.8393741250038147,
+      "learning_rate": 0.00013714840798704803,
+      "loss": 3.2221,
+      "step": 71600
+    },
+    {
+      "epoch": 7.725067385444744,
+      "grad_norm": 0.8368397355079651,
+      "learning_rate": 0.00013682460874257958,
+      "loss": 3.2449,
+      "step": 71650
+    },
+    {
+      "epoch": 7.730458221024259,
+      "grad_norm": 0.8051853775978088,
+      "learning_rate": 0.00013650080949811116,
+      "loss": 3.2458,
+      "step": 71700
+    },
+    {
+      "epoch": 7.735849056603773,
+      "grad_norm": 0.7938108444213867,
+      "learning_rate": 0.00013617701025364274,
+      "loss": 3.2427,
+      "step": 71750
+    },
+    {
+      "epoch": 7.741239892183288,
+      "grad_norm": 0.7889504432678223,
+      "learning_rate": 0.0001358532110091743,
+      "loss": 3.229,
+      "step": 71800
+    },
+    {
+      "epoch": 7.7466307277628035,
+      "grad_norm": 0.8129399418830872,
+      "learning_rate": 0.00013552941176470587,
+      "loss": 3.247,
+      "step": 71850
+    },
+    {
+      "epoch": 7.752021563342318,
+      "grad_norm": 0.8431523442268372,
+      "learning_rate": 0.00013520561252023744,
+      "loss": 3.2453,
+      "step": 71900
+    },
+    {
+      "epoch": 7.757412398921833,
+      "grad_norm": 0.805289626121521,
+      "learning_rate": 0.00013488181327576902,
+      "loss": 3.2475,
+      "step": 71950
+    },
+    {
+      "epoch": 7.762803234501348,
+      "grad_norm": 0.8263291120529175,
+      "learning_rate": 0.00013455801403130057,
+      "loss": 3.2532,
+      "step": 72000
+    },
+    {
+      "epoch": 7.762803234501348,
+      "eval_accuracy": 0.38924777541361166,
+      "eval_loss": 3.3399181365966797,
+      "eval_runtime": 185.5782,
+      "eval_samples_per_second": 97.053,
+      "eval_steps_per_second": 6.068,
+      "step": 72000
+    },
+    {
+      "epoch": 7.768194070080862,
+      "grad_norm": 0.854796290397644,
+      "learning_rate": 0.00013423421478683215,
+      "loss": 3.2378,
+      "step": 72050
+    },
+    {
+      "epoch": 7.773584905660377,
+      "grad_norm": 0.7867196202278137,
+      "learning_rate": 0.00013391041554236373,
+      "loss": 3.2449,
+      "step": 72100
+    },
+    {
+      "epoch": 7.7789757412398925,
+      "grad_norm": 0.943561315536499,
+      "learning_rate": 0.00013358661629789528,
+      "loss": 3.2314,
+      "step": 72150
+    },
+    {
+      "epoch": 7.784366576819407,
+      "grad_norm": 0.7987220287322998,
+      "learning_rate": 0.00013326281705342685,
+      "loss": 3.2218,
+      "step": 72200
+    },
+    {
+      "epoch": 7.789757412398922,
+      "grad_norm": 0.8089998960494995,
+      "learning_rate": 0.00013293901780895843,
+      "loss": 3.2348,
+      "step": 72250
+    },
+    {
+      "epoch": 7.795148247978437,
+      "grad_norm": 0.8148220777511597,
+      "learning_rate": 0.00013261521856449,
+      "loss": 3.2246,
+      "step": 72300
+    },
+    {
+      "epoch": 7.800539083557951,
+      "grad_norm": 0.797408401966095,
+      "learning_rate": 0.00013229141932002156,
+      "loss": 3.2114,
+      "step": 72350
+    },
+    {
+      "epoch": 7.8059299191374665,
+      "grad_norm": 0.8176799416542053,
+      "learning_rate": 0.00013196762007555314,
+      "loss": 3.2338,
+      "step": 72400
+    },
+    {
+      "epoch": 7.811320754716981,
+      "grad_norm": 0.8059448599815369,
+      "learning_rate": 0.00013164382083108472,
+      "loss": 3.2398,
+      "step": 72450
+    },
+    {
+      "epoch": 7.816711590296496,
+      "grad_norm": 0.8232355117797852,
+      "learning_rate": 0.0001313200215866163,
+      "loss": 3.2354,
+      "step": 72500
+    },
+    {
+      "epoch": 7.822102425876011,
+      "grad_norm": 0.8593564629554749,
+      "learning_rate": 0.00013099622234214784,
+      "loss": 3.2213,
+      "step": 72550
+    },
+    {
+      "epoch": 7.827493261455525,
+      "grad_norm": 0.798647403717041,
+      "learning_rate": 0.00013067242309767942,
+      "loss": 3.252,
+      "step": 72600
+    },
+    {
+      "epoch": 7.83288409703504,
+      "grad_norm": 0.888236403465271,
+      "learning_rate": 0.000130348623853211,
+      "loss": 3.236,
+      "step": 72650
+    },
+    {
+      "epoch": 7.8382749326145555,
+      "grad_norm": 0.8630709648132324,
+      "learning_rate": 0.00013002482460874258,
+      "loss": 3.2415,
+      "step": 72700
+    },
+    {
+      "epoch": 7.84366576819407,
+      "grad_norm": 0.8336551785469055,
+      "learning_rate": 0.00012970102536427413,
+      "loss": 3.2107,
+      "step": 72750
+    },
+    {
+      "epoch": 7.849056603773585,
+      "grad_norm": 0.7977820634841919,
+      "learning_rate": 0.0001293772261198057,
+      "loss": 3.2412,
+      "step": 72800
+    },
+    {
+      "epoch": 7.8544474393531,
+      "grad_norm": 0.8409989476203918,
+      "learning_rate": 0.00012905342687533728,
+      "loss": 3.2389,
+      "step": 72850
+    },
+    {
+      "epoch": 7.859838274932614,
+      "grad_norm": 0.8394231200218201,
+      "learning_rate": 0.00012872962763086886,
+      "loss": 3.24,
+      "step": 72900
+    },
+    {
+      "epoch": 7.8652291105121295,
+      "grad_norm": 0.8246923089027405,
+      "learning_rate": 0.00012840582838640044,
+      "loss": 3.2424,
+      "step": 72950
+    },
+    {
+      "epoch": 7.870619946091644,
+      "grad_norm": 0.8409177660942078,
+      "learning_rate": 0.000128082029141932,
+      "loss": 3.2332,
+      "step": 73000
+    },
+    {
+      "epoch": 7.870619946091644,
+      "eval_accuracy": 0.38957547252326014,
+      "eval_loss": 3.3348822593688965,
+      "eval_runtime": 185.7721,
+      "eval_samples_per_second": 96.952,
+      "eval_steps_per_second": 6.061,
+      "step": 73000
+    },
+    {
+      "epoch": 7.876010781671159,
+      "grad_norm": 0.7878761291503906,
+      "learning_rate": 0.00012775822989746357,
+      "loss": 3.2406,
+      "step": 73050
+    },
+    {
+      "epoch": 7.881401617250674,
+      "grad_norm": 0.7846654057502747,
+      "learning_rate": 0.00012743443065299514,
+      "loss": 3.2387,
+      "step": 73100
+    },
+    {
+      "epoch": 7.886792452830189,
+      "grad_norm": 0.815717339515686,
+      "learning_rate": 0.00012711063140852672,
+      "loss": 3.2362,
+      "step": 73150
+    },
+    {
+      "epoch": 7.892183288409703,
+      "grad_norm": 0.8608244061470032,
+      "learning_rate": 0.00012678683216405827,
+      "loss": 3.2305,
+      "step": 73200
+    },
+    {
+      "epoch": 7.8975741239892185,
+      "grad_norm": 0.7981069087982178,
+      "learning_rate": 0.00012646303291958985,
+      "loss": 3.2381,
+      "step": 73250
+    },
+    {
+      "epoch": 7.902964959568733,
+      "grad_norm": 0.8341260552406311,
+      "learning_rate": 0.00012613923367512143,
+      "loss": 3.2418,
+      "step": 73300
+    },
+    {
+      "epoch": 7.908355795148248,
+      "grad_norm": 0.8806238770484924,
+      "learning_rate": 0.00012581543443065298,
+      "loss": 3.2187,
+      "step": 73350
+    },
+    {
+      "epoch": 7.913746630727763,
+      "grad_norm": 0.8482242822647095,
+      "learning_rate": 0.00012549163518618456,
+      "loss": 3.2321,
+      "step": 73400
+    },
+    {
+      "epoch": 7.919137466307277,
+      "grad_norm": 0.8092681169509888,
+      "learning_rate": 0.00012516783594171613,
+      "loss": 3.2147,
+      "step": 73450
+    },
+    {
+      "epoch": 7.9245283018867925,
+      "grad_norm": 0.8765472173690796,
+      "learning_rate": 0.00012484403669724768,
+      "loss": 3.2332,
+      "step": 73500
+    },
+    {
+      "epoch": 7.929919137466308,
+      "grad_norm": 0.8127001523971558,
+      "learning_rate": 0.00012452023745277926,
+      "loss": 3.2242,
+      "step": 73550
+    },
+    {
+      "epoch": 7.935309973045822,
+      "grad_norm": 0.8233795166015625,
+      "learning_rate": 0.00012419643820831084,
+      "loss": 3.2236,
+      "step": 73600
+    },
+    {
+      "epoch": 7.940700808625337,
+      "grad_norm": 0.846750795841217,
+      "learning_rate": 0.0001238726389638424,
+      "loss": 3.2413,
+      "step": 73650
+    },
+    {
+      "epoch": 7.946091644204852,
+      "grad_norm": 0.8156243562698364,
+      "learning_rate": 0.00012354883971937397,
+      "loss": 3.2173,
+      "step": 73700
+    },
+    {
+      "epoch": 7.951482479784366,
+      "grad_norm": 0.8058876991271973,
+      "learning_rate": 0.00012322504047490555,
+      "loss": 3.2356,
+      "step": 73750
+    },
+    {
+      "epoch": 7.9568733153638815,
+      "grad_norm": 0.829677402973175,
+      "learning_rate": 0.00012290124123043712,
+      "loss": 3.2339,
+      "step": 73800
+    },
+    {
+      "epoch": 7.962264150943396,
+      "grad_norm": 0.8429958820343018,
+      "learning_rate": 0.00012257744198596867,
+      "loss": 3.2205,
+      "step": 73850
+    },
+    {
+      "epoch": 7.967654986522911,
+      "grad_norm": 0.8639125823974609,
+      "learning_rate": 0.00012225364274150025,
+      "loss": 3.2252,
+      "step": 73900
+    },
+    {
+      "epoch": 7.973045822102426,
+      "grad_norm": 0.8114444017410278,
+      "learning_rate": 0.00012192984349703183,
+      "loss": 3.2273,
+      "step": 73950
+    },
+    {
+      "epoch": 7.97843665768194,
+      "grad_norm": 0.8097919821739197,
+      "learning_rate": 0.0001216060442525634,
+      "loss": 3.2535,
+      "step": 74000
+    },
+    {
+      "epoch": 7.97843665768194,
+      "eval_accuracy": 0.3901027649878238,
+      "eval_loss": 3.331040382385254,
+      "eval_runtime": 185.9774,
+      "eval_samples_per_second": 96.845,
+      "eval_steps_per_second": 6.054,
+      "step": 74000
+    },
+    {
+      "epoch": 7.9838274932614555,
+      "grad_norm": 0.9089035391807556,
+      "learning_rate": 0.00012128224500809496,
+      "loss": 3.2396,
+      "step": 74050
+    },
+    {
+      "epoch": 7.989218328840971,
+      "grad_norm": 0.8275425434112549,
+      "learning_rate": 0.00012095844576362653,
+      "loss": 3.2418,
+      "step": 74100
+    },
+    {
+      "epoch": 7.994609164420485,
+      "grad_norm": 0.8922067284584045,
+      "learning_rate": 0.00012063464651915811,
+      "loss": 3.2427,
+      "step": 74150
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 1.7210991382598877,
+      "learning_rate": 0.00012031084727468969,
+      "loss": 3.239,
+      "step": 74200
+    },
+    {
+      "epoch": 8.005390835579515,
+      "grad_norm": 0.8409031629562378,
+      "learning_rate": 0.00011998704803022124,
+      "loss": 3.1553,
+      "step": 74250
+    },
+    {
+      "epoch": 8.01078167115903,
+      "grad_norm": 0.7952747941017151,
+      "learning_rate": 0.00011966324878575282,
+      "loss": 3.1633,
+      "step": 74300
+    },
+    {
+      "epoch": 8.016172506738544,
+      "grad_norm": 0.8238943815231323,
+      "learning_rate": 0.0001193394495412844,
+      "loss": 3.1622,
+      "step": 74350
+    },
+    {
+      "epoch": 8.021563342318059,
+      "grad_norm": 0.8651562333106995,
+      "learning_rate": 0.00011901565029681597,
+      "loss": 3.1588,
+      "step": 74400
+    },
+    {
+      "epoch": 8.026954177897574,
+      "grad_norm": 0.8449139595031738,
+      "learning_rate": 0.00011869185105234754,
+      "loss": 3.162,
+      "step": 74450
+    },
+    {
+      "epoch": 8.032345013477089,
+      "grad_norm": 0.8473836183547974,
+      "learning_rate": 0.0001183680518078791,
+      "loss": 3.1806,
+      "step": 74500
+    },
+    {
+      "epoch": 8.037735849056604,
+      "grad_norm": 0.8596470952033997,
+      "learning_rate": 0.00011804425256341068,
+      "loss": 3.1587,
+      "step": 74550
+    },
+    {
+      "epoch": 8.04312668463612,
+      "grad_norm": 0.8506665825843811,
+      "learning_rate": 0.00011772045331894224,
+      "loss": 3.1801,
+      "step": 74600
+    },
+    {
+      "epoch": 8.048517520215633,
+      "grad_norm": 0.8084812760353088,
+      "learning_rate": 0.00011739665407447382,
+      "loss": 3.1751,
+      "step": 74650
+    },
+    {
+      "epoch": 8.053908355795148,
+      "grad_norm": 0.8186056017875671,
+      "learning_rate": 0.00011707285483000539,
+      "loss": 3.1641,
+      "step": 74700
+    },
+    {
+      "epoch": 8.059299191374663,
+      "grad_norm": 0.8883737921714783,
+      "learning_rate": 0.00011674905558553695,
+      "loss": 3.1516,
+      "step": 74750
+    },
+    {
+      "epoch": 8.064690026954178,
+      "grad_norm": 0.8599704504013062,
+      "learning_rate": 0.00011642525634106853,
+      "loss": 3.1718,
+      "step": 74800
+    },
+    {
+      "epoch": 8.070080862533693,
+      "grad_norm": 0.8047420382499695,
+      "learning_rate": 0.0001161014570966001,
+      "loss": 3.1504,
+      "step": 74850
+    },
+    {
+      "epoch": 8.075471698113208,
+      "grad_norm": 0.7963536381721497,
+      "learning_rate": 0.00011577765785213166,
+      "loss": 3.1519,
+      "step": 74900
+    },
+    {
+      "epoch": 8.080862533692722,
+      "grad_norm": 1.6519439220428467,
+      "learning_rate": 0.00011545385860766323,
+      "loss": 3.1683,
+      "step": 74950
+    },
+    {
+      "epoch": 8.086253369272237,
+      "grad_norm": 0.8095781803131104,
+      "learning_rate": 0.00011513005936319481,
+      "loss": 3.1589,
+      "step": 75000
+    },
+    {
+      "epoch": 8.086253369272237,
+      "eval_accuracy": 0.390003673554137,
+      "eval_loss": 3.337833881378174,
+      "eval_runtime": 185.4504,
+      "eval_samples_per_second": 97.12,
+      "eval_steps_per_second": 6.072,
+      "step": 75000
+    },
+    {
+      "epoch": 8.091644204851752,
+      "grad_norm": 0.870051383972168,
+      "learning_rate": 0.00011480626011872639,
+      "loss": 3.1641,
+      "step": 75050
+    },
+    {
+      "epoch": 8.097035040431267,
+      "grad_norm": 0.8616654276847839,
+      "learning_rate": 0.00011448246087425794,
+      "loss": 3.1502,
+      "step": 75100
+    },
+    {
+      "epoch": 8.102425876010782,
+      "grad_norm": 0.806140661239624,
+      "learning_rate": 0.00011415866162978952,
+      "loss": 3.1751,
+      "step": 75150
+    },
+    {
+      "epoch": 8.107816711590296,
+      "grad_norm": 0.8271510601043701,
+      "learning_rate": 0.0001138348623853211,
+      "loss": 3.1687,
+      "step": 75200
+    },
+    {
+      "epoch": 8.11320754716981,
+      "grad_norm": 0.8450615406036377,
+      "learning_rate": 0.00011351106314085266,
+      "loss": 3.1739,
+      "step": 75250
+    },
+    {
+      "epoch": 8.118598382749326,
+      "grad_norm": 0.8098863363265991,
+      "learning_rate": 0.00011318726389638424,
+      "loss": 3.1824,
+      "step": 75300
+    },
+    {
+      "epoch": 8.123989218328841,
+      "grad_norm": 0.817785918712616,
+      "learning_rate": 0.0001128634646519158,
+      "loss": 3.1848,
+      "step": 75350
+    },
+    {
+      "epoch": 8.129380053908356,
+      "grad_norm": 0.8651502728462219,
+      "learning_rate": 0.00011253966540744738,
+      "loss": 3.1858,
+      "step": 75400
+    },
+    {
+      "epoch": 8.134770889487871,
+      "grad_norm": 0.8325152397155762,
+      "learning_rate": 0.00011221586616297894,
+      "loss": 3.172,
+      "step": 75450
+    },
+    {
+      "epoch": 8.140161725067385,
+      "grad_norm": 0.8323520421981812,
+      "learning_rate": 0.00011189206691851052,
+      "loss": 3.1783,
+      "step": 75500
+    },
+    {
+      "epoch": 8.1455525606469,
+      "grad_norm": 0.7873876690864563,
+      "learning_rate": 0.00011156826767404208,
+      "loss": 3.1606,
+      "step": 75550
+    },
+    {
+      "epoch": 8.150943396226415,
+      "grad_norm": 0.8682588338851929,
+      "learning_rate": 0.00011124446842957365,
+      "loss": 3.1594,
+      "step": 75600
+    },
+    {
+      "epoch": 8.15633423180593,
+      "grad_norm": 0.8208882212638855,
+      "learning_rate": 0.00011092066918510523,
+      "loss": 3.1664,
+      "step": 75650
+    },
+    {
+      "epoch": 8.161725067385445,
+      "grad_norm": 0.8171040415763855,
+      "learning_rate": 0.0001105968699406368,
+      "loss": 3.1855,
+      "step": 75700
+    },
+    {
+      "epoch": 8.167115902964959,
+      "grad_norm": 0.804966390132904,
+      "learning_rate": 0.00011027307069616835,
+      "loss": 3.1654,
+      "step": 75750
+    },
+    {
+      "epoch": 8.172506738544474,
+      "grad_norm": 0.8100362420082092,
+      "learning_rate": 0.00010994927145169993,
+      "loss": 3.16,
+      "step": 75800
+    },
+    {
+      "epoch": 8.177897574123989,
+      "grad_norm": 0.838396430015564,
+      "learning_rate": 0.00010962547220723151,
+      "loss": 3.1669,
+      "step": 75850
+    },
+    {
+      "epoch": 8.183288409703504,
+      "grad_norm": 0.8108015656471252,
+      "learning_rate": 0.00010930167296276309,
+      "loss": 3.1689,
+      "step": 75900
+    },
+    {
+      "epoch": 8.18867924528302,
+      "grad_norm": 0.8487863540649414,
+      "learning_rate": 0.00010897787371829465,
+      "loss": 3.1801,
+      "step": 75950
+    },
+    {
+      "epoch": 8.194070080862534,
+      "grad_norm": 0.8276529312133789,
+      "learning_rate": 0.00010865407447382622,
+      "loss": 3.1588,
+      "step": 76000
+    },
+    {
+      "epoch": 8.194070080862534,
+      "eval_accuracy": 0.39020739771878915,
+      "eval_loss": 3.334951162338257,
+      "eval_runtime": 185.8453,
+      "eval_samples_per_second": 96.914,
+      "eval_steps_per_second": 6.059,
+      "step": 76000
+    },
+    {
+      "epoch": 8.199460916442048,
+      "grad_norm": 0.8735484480857849,
+      "learning_rate": 0.00010833027522935779,
+      "loss": 3.1817,
+      "step": 76050
+    },
+    {
+      "epoch": 8.204851752021563,
+      "grad_norm": 0.8510221242904663,
+      "learning_rate": 0.00010800647598488936,
+      "loss": 3.2001,
+      "step": 76100
+    },
+    {
+      "epoch": 8.210242587601078,
+      "grad_norm": 0.8201325535774231,
+      "learning_rate": 0.00010768267674042093,
+      "loss": 3.1683,
+      "step": 76150
+    },
+    {
+      "epoch": 8.215633423180593,
+      "grad_norm": 0.8542962670326233,
+      "learning_rate": 0.0001073588774959525,
+      "loss": 3.1843,
+      "step": 76200
+    },
+    {
+      "epoch": 8.221024258760108,
+      "grad_norm": 0.8557514548301697,
+      "learning_rate": 0.00010703507825148406,
+      "loss": 3.1785,
+      "step": 76250
+    },
+    {
+      "epoch": 8.226415094339623,
+      "grad_norm": 0.9004441499710083,
+      "learning_rate": 0.00010671127900701564,
+      "loss": 3.159,
+      "step": 76300
+    },
+    {
+      "epoch": 8.231805929919137,
+      "grad_norm": 0.8508898019790649,
+      "learning_rate": 0.00010638747976254722,
+      "loss": 3.1793,
+      "step": 76350
+    },
+    {
+      "epoch": 8.237196765498652,
+      "grad_norm": 0.8295674324035645,
+      "learning_rate": 0.00010606368051807877,
+      "loss": 3.1762,
+      "step": 76400
+    },
+    {
+      "epoch": 8.242587601078167,
+      "grad_norm": 0.8416191935539246,
+      "learning_rate": 0.00010573988127361035,
+      "loss": 3.1768,
+      "step": 76450
+    },
+    {
+      "epoch": 8.247978436657682,
+      "grad_norm": 0.8672610521316528,
+      "learning_rate": 0.00010541608202914192,
+      "loss": 3.1617,
+      "step": 76500
+    },
+    {
+      "epoch": 8.253369272237197,
+      "grad_norm": 0.8373281955718994,
+      "learning_rate": 0.00010509875876956286,
+      "loss": 3.1909,
+      "step": 76550
+    },
+    {
+      "epoch": 8.25876010781671,
+      "grad_norm": 0.8693946599960327,
+      "learning_rate": 0.00010477495952509443,
+      "loss": 3.1658,
+      "step": 76600
+    },
+    {
+      "epoch": 8.264150943396226,
+      "grad_norm": 0.7996649146080017,
+      "learning_rate": 0.000104451160280626,
+      "loss": 3.163,
+      "step": 76650
+    },
+    {
+      "epoch": 8.269541778975741,
+      "grad_norm": 0.8543344140052795,
+      "learning_rate": 0.00010412736103615758,
+      "loss": 3.1847,
+      "step": 76700
+    },
+    {
+      "epoch": 8.274932614555256,
+      "grad_norm": 0.8052183389663696,
+      "learning_rate": 0.00010380356179168913,
+      "loss": 3.1765,
+      "step": 76750
+    },
+    {
+      "epoch": 8.280323450134771,
+      "grad_norm": 0.8508358001708984,
+      "learning_rate": 0.00010347976254722071,
+      "loss": 3.1817,
+      "step": 76800
+    },
+    {
+      "epoch": 8.285714285714286,
+      "grad_norm": 0.7949924468994141,
+      "learning_rate": 0.00010315596330275229,
+      "loss": 3.1906,
+      "step": 76850
+    },
+    {
+      "epoch": 8.2911051212938,
+      "grad_norm": 0.863523542881012,
+      "learning_rate": 0.00010283216405828387,
+      "loss": 3.2043,
+      "step": 76900
+    },
+    {
+      "epoch": 8.296495956873315,
+      "grad_norm": 0.8439539074897766,
+      "learning_rate": 0.00010250836481381542,
+      "loss": 3.188,
+      "step": 76950
+    },
+    {
+      "epoch": 8.30188679245283,
+      "grad_norm": 0.8541856408119202,
+      "learning_rate": 0.000102184565569347,
+      "loss": 3.1776,
+      "step": 77000
+    },
+    {
+      "epoch": 8.30188679245283,
+      "eval_accuracy": 0.39050369414385927,
+      "eval_loss": 3.3322830200195312,
+      "eval_runtime": 185.6486,
+      "eval_samples_per_second": 97.017,
+      "eval_steps_per_second": 6.065,
+      "step": 77000
+    },
+    {
+      "epoch": 8.307277628032345,
+      "grad_norm": 0.8105238080024719,
+      "learning_rate": 0.00010186076632487857,
+      "loss": 3.1893,
+      "step": 77050
+    },
+    {
+      "epoch": 8.31266846361186,
+      "grad_norm": 0.8430174589157104,
+      "learning_rate": 0.00010153696708041014,
+      "loss": 3.1755,
+      "step": 77100
+    },
+    {
+      "epoch": 8.318059299191376,
+      "grad_norm": 0.8541438579559326,
+      "learning_rate": 0.0001012131678359417,
+      "loss": 3.1628,
+      "step": 77150
+    },
+    {
+      "epoch": 8.323450134770889,
+      "grad_norm": 0.8847959041595459,
+      "learning_rate": 0.00010088936859147328,
+      "loss": 3.1636,
+      "step": 77200
+    },
+    {
+      "epoch": 8.328840970350404,
+      "grad_norm": 0.8577356934547424,
+      "learning_rate": 0.00010056556934700485,
+      "loss": 3.1628,
+      "step": 77250
+    },
+    {
+      "epoch": 8.33423180592992,
+      "grad_norm": 0.830810546875,
+      "learning_rate": 0.00010024177010253642,
+      "loss": 3.1705,
+      "step": 77300
+    },
+    {
+      "epoch": 8.339622641509434,
+      "grad_norm": 0.8565813302993774,
+      "learning_rate": 9.9917970858068e-05,
+      "loss": 3.1872,
+      "step": 77350
+    },
+    {
+      "epoch": 8.34501347708895,
+      "grad_norm": 0.8739506602287292,
+      "learning_rate": 9.959417161359956e-05,
+      "loss": 3.1396,
+      "step": 77400
+    },
+    {
+      "epoch": 8.350404312668463,
+      "grad_norm": 0.8647984266281128,
+      "learning_rate": 9.927037236913112e-05,
+      "loss": 3.1851,
+      "step": 77450
+    },
+    {
+      "epoch": 8.355795148247978,
+      "grad_norm": 0.8516337275505066,
+      "learning_rate": 9.89465731246627e-05,
+      "loss": 3.1902,
+      "step": 77500
+    },
+    {
+      "epoch": 8.361185983827493,
+      "grad_norm": 0.8887714147567749,
+      "learning_rate": 9.862277388019428e-05,
+      "loss": 3.178,
+      "step": 77550
+    },
+    {
+      "epoch": 8.366576819407008,
+      "grad_norm": 0.8771308660507202,
+      "learning_rate": 9.829897463572583e-05,
+      "loss": 3.1929,
+      "step": 77600
+    },
+    {
+      "epoch": 8.371967654986523,
+      "grad_norm": 0.8397874236106873,
+      "learning_rate": 9.797517539125741e-05,
+      "loss": 3.1598,
+      "step": 77650
+    },
+    {
+      "epoch": 8.377358490566039,
+      "grad_norm": 0.8755930662155151,
+      "learning_rate": 9.765137614678899e-05,
+      "loss": 3.197,
+      "step": 77700
+    },
+    {
+      "epoch": 8.382749326145552,
+      "grad_norm": 0.8410030007362366,
+      "learning_rate": 9.732757690232056e-05,
+      "loss": 3.1773,
+      "step": 77750
+    },
+    {
+      "epoch": 8.388140161725067,
+      "grad_norm": 0.9130405187606812,
+      "learning_rate": 9.700377765785211e-05,
+      "loss": 3.1722,
+      "step": 77800
+    },
+    {
+      "epoch": 8.393530997304582,
+      "grad_norm": 0.8973836898803711,
+      "learning_rate": 9.667997841338369e-05,
+      "loss": 3.1806,
+      "step": 77850
+    },
+    {
+      "epoch": 8.398921832884097,
+      "grad_norm": 0.9638514518737793,
+      "learning_rate": 9.635617916891527e-05,
+      "loss": 3.18,
+      "step": 77900
+    },
+    {
+      "epoch": 8.404312668463612,
+      "grad_norm": 0.9173029661178589,
+      "learning_rate": 9.603237992444683e-05,
+      "loss": 3.1851,
+      "step": 77950
+    },
+    {
+      "epoch": 8.409703504043126,
+      "grad_norm": 0.8818854093551636,
+      "learning_rate": 9.57085806799784e-05,
+      "loss": 3.1701,
+      "step": 78000
+    },
+    {
+      "epoch": 8.409703504043126,
+      "eval_accuracy": 0.39072067396082477,
+      "eval_loss": 3.329658031463623,
+      "eval_runtime": 185.9378,
+      "eval_samples_per_second": 96.866,
+      "eval_steps_per_second": 6.056,
+      "step": 78000
+    },
+    {
+      "epoch": 8.415094339622641,
+      "grad_norm": 0.8253286480903625,
+      "learning_rate": 9.539125742039935e-05,
+      "loss": 3.1827,
+      "step": 78050
+    },
+    {
+      "epoch": 8.420485175202156,
+      "grad_norm": 0.8262404203414917,
+      "learning_rate": 9.506745817593093e-05,
+      "loss": 3.1714,
+      "step": 78100
+    },
+    {
+      "epoch": 8.425876010781671,
+      "grad_norm": 0.8860495686531067,
+      "learning_rate": 9.474365893146248e-05,
+      "loss": 3.1784,
+      "step": 78150
+    },
+    {
+      "epoch": 8.431266846361186,
+      "grad_norm": 0.9161273241043091,
+      "learning_rate": 9.441985968699406e-05,
+      "loss": 3.1912,
+      "step": 78200
+    },
+    {
+      "epoch": 8.436657681940702,
+      "grad_norm": 0.8695592880249023,
+      "learning_rate": 9.409606044252563e-05,
+      "loss": 3.2149,
+      "step": 78250
+    },
+    {
+      "epoch": 8.442048517520215,
+      "grad_norm": 0.9244877099990845,
+      "learning_rate": 9.37722611980572e-05,
+      "loss": 3.1809,
+      "step": 78300
+    },
+    {
+      "epoch": 8.44743935309973,
+      "grad_norm": 0.8668970465660095,
+      "learning_rate": 9.344846195358876e-05,
+      "loss": 3.1775,
+      "step": 78350
+    },
+    {
+      "epoch": 8.452830188679245,
+      "grad_norm": 0.8342470526695251,
+      "learning_rate": 9.312466270912034e-05,
+      "loss": 3.1836,
+      "step": 78400
+    },
+    {
+      "epoch": 8.45822102425876,
+      "grad_norm": 0.9028128981590271,
+      "learning_rate": 9.28008634646519e-05,
+      "loss": 3.1925,
+      "step": 78450
+    },
+    {
+      "epoch": 8.463611859838275,
+      "grad_norm": 0.9058071374893188,
+      "learning_rate": 9.247706422018348e-05,
+      "loss": 3.1881,
+      "step": 78500
+    },
+    {
+      "epoch": 8.46900269541779,
+      "grad_norm": 0.8586152195930481,
+      "learning_rate": 9.215326497571505e-05,
+      "loss": 3.1791,
+      "step": 78550
+    },
+    {
+      "epoch": 8.474393530997304,
+      "grad_norm": 0.8235473036766052,
+      "learning_rate": 9.182946573124661e-05,
+      "loss": 3.1874,
+      "step": 78600
+    },
+    {
+      "epoch": 8.479784366576819,
+      "grad_norm": 0.8261722326278687,
+      "learning_rate": 9.150566648677819e-05,
+      "loss": 3.177,
+      "step": 78650
+    },
+    {
+      "epoch": 8.485175202156334,
+      "grad_norm": 0.8255433440208435,
+      "learning_rate": 9.118186724230976e-05,
+      "loss": 3.1802,
+      "step": 78700
+    },
+    {
+      "epoch": 8.49056603773585,
+      "grad_norm": 0.9035568833351135,
+      "learning_rate": 9.085806799784134e-05,
+      "loss": 3.1888,
+      "step": 78750
+    },
+    {
+      "epoch": 8.495956873315365,
+      "grad_norm": 0.8632428050041199,
+      "learning_rate": 9.053426875337289e-05,
+      "loss": 3.1952,
+      "step": 78800
+    },
+    {
+      "epoch": 8.501347708894878,
+      "grad_norm": 0.830828070640564,
+      "learning_rate": 9.021046950890447e-05,
+      "loss": 3.1892,
+      "step": 78850
+    },
+    {
+      "epoch": 8.506738544474393,
+      "grad_norm": 0.8392572999000549,
+      "learning_rate": 8.988667026443605e-05,
+      "loss": 3.1811,
+      "step": 78900
+    },
+    {
+      "epoch": 8.512129380053908,
+      "grad_norm": 0.8769532442092896,
+      "learning_rate": 8.956287101996761e-05,
+      "loss": 3.179,
+      "step": 78950
+    },
+    {
+      "epoch": 8.517520215633423,
+      "grad_norm": 0.7989556789398193,
+      "learning_rate": 8.923907177549918e-05,
+      "loss": 3.1819,
+      "step": 79000
+    },
+    {
+      "epoch": 8.517520215633423,
+      "eval_accuracy": 0.3910877034158621,
+      "eval_loss": 3.324533700942993,
+      "eval_runtime": 185.3897,
+      "eval_samples_per_second": 97.152,
+      "eval_steps_per_second": 6.074,
+      "step": 79000
+    },
+    {
+      "epoch": 8.522911051212938,
+      "grad_norm": 0.842538058757782,
+      "learning_rate": 8.891527253103075e-05,
+      "loss": 3.1978,
+      "step": 79050
+    },
+    {
+      "epoch": 8.528301886792454,
+      "grad_norm": 0.8743948340415955,
+      "learning_rate": 8.859147328656233e-05,
+      "loss": 3.1816,
+      "step": 79100
+    },
+    {
+      "epoch": 8.533692722371967,
+      "grad_norm": 0.8316393494606018,
+      "learning_rate": 8.82676740420939e-05,
+      "loss": 3.1885,
+      "step": 79150
+    },
+    {
+      "epoch": 8.539083557951482,
+      "grad_norm": 0.8724850416183472,
+      "learning_rate": 8.794387479762546e-05,
+      "loss": 3.2101,
+      "step": 79200
+    },
+    {
+      "epoch": 8.544474393530997,
+      "grad_norm": 0.8789516687393188,
+      "learning_rate": 8.762007555315704e-05,
+      "loss": 3.188,
+      "step": 79250
+    },
+    {
+      "epoch": 8.549865229110512,
+      "grad_norm": 0.9154621958732605,
+      "learning_rate": 8.72962763086886e-05,
+      "loss": 3.1615,
+      "step": 79300
+    },
+    {
+      "epoch": 8.555256064690028,
+      "grad_norm": 0.8666130304336548,
+      "learning_rate": 8.697247706422018e-05,
+      "loss": 3.1842,
+      "step": 79350
+    },
+    {
+      "epoch": 8.560646900269543,
+      "grad_norm": 0.8889215588569641,
+      "learning_rate": 8.664867781975174e-05,
+      "loss": 3.1808,
+      "step": 79400
+    },
+    {
+      "epoch": 8.566037735849056,
+      "grad_norm": 0.8791170120239258,
+      "learning_rate": 8.632487857528331e-05,
+      "loss": 3.1674,
+      "step": 79450
+    },
+    {
+      "epoch": 8.571428571428571,
+      "grad_norm": 0.882530689239502,
+      "learning_rate": 8.600107933081489e-05,
+      "loss": 3.1768,
+      "step": 79500
+    },
+    {
+      "epoch": 8.576819407008086,
+      "grad_norm": 0.9383884072303772,
+      "learning_rate": 8.567728008634646e-05,
+      "loss": 3.1783,
+      "step": 79550
+    },
+    {
+      "epoch": 8.582210242587601,
+      "grad_norm": 0.8669118285179138,
+      "learning_rate": 8.535348084187804e-05,
+      "loss": 3.1623,
+      "step": 79600
+    },
+    {
+      "epoch": 8.587601078167117,
+      "grad_norm": 0.8568759560585022,
+      "learning_rate": 8.502968159740959e-05,
+      "loss": 3.1753,
+      "step": 79650
+    },
+    {
+      "epoch": 8.59299191374663,
+      "grad_norm": 0.8259025812149048,
+      "learning_rate": 8.470588235294117e-05,
+      "loss": 3.176,
+      "step": 79700
+    },
+    {
+      "epoch": 8.598382749326145,
+      "grad_norm": 0.8500630855560303,
+      "learning_rate": 8.438208310847275e-05,
+      "loss": 3.1826,
+      "step": 79750
+    },
+    {
+      "epoch": 8.60377358490566,
+      "grad_norm": 0.84529709815979,
+      "learning_rate": 8.405828386400431e-05,
+      "loss": 3.1866,
+      "step": 79800
+    },
+    {
+      "epoch": 8.609164420485175,
+      "grad_norm": 0.8760417103767395,
+      "learning_rate": 8.373448461953587e-05,
+      "loss": 3.1649,
+      "step": 79850
+    },
+    {
+      "epoch": 8.61455525606469,
+      "grad_norm": 0.8787358403205872,
+      "learning_rate": 8.341068537506745e-05,
+      "loss": 3.2081,
+      "step": 79900
+    },
+    {
+      "epoch": 8.619946091644206,
+      "grad_norm": 0.8166438341140747,
+      "learning_rate": 8.308688613059902e-05,
+      "loss": 3.1845,
+      "step": 79950
+    },
+    {
+      "epoch": 8.625336927223719,
+      "grad_norm": 0.8741464614868164,
+      "learning_rate": 8.27630868861306e-05,
+      "loss": 3.1988,
+      "step": 80000
+    },
+    {
+      "epoch": 8.625336927223719,
+      "eval_accuracy": 0.39168105683621696,
+      "eval_loss": 3.3219919204711914,
+      "eval_runtime": 185.6858,
+      "eval_samples_per_second": 96.997,
+      "eval_steps_per_second": 6.064,
+      "step": 80000
+    },
+    {
+      "epoch": 8.630727762803234,
+      "grad_norm": 0.8296083211898804,
+      "learning_rate": 8.243928764166216e-05,
+      "loss": 3.1903,
+      "step": 80050
+    },
+    {
+      "epoch": 8.63611859838275,
+      "grad_norm": 0.85300213098526,
+      "learning_rate": 8.211548839719374e-05,
+      "loss": 3.1728,
+      "step": 80100
+    },
+    {
+      "epoch": 8.641509433962264,
+      "grad_norm": 0.868763267993927,
+      "learning_rate": 8.179816513761467e-05,
+      "loss": 3.1832,
+      "step": 80150
+    },
+    {
+      "epoch": 8.64690026954178,
+      "grad_norm": 0.8488466739654541,
+      "learning_rate": 8.147436589314624e-05,
+      "loss": 3.1971,
+      "step": 80200
+    },
+    {
+      "epoch": 8.652291105121293,
+      "grad_norm": 0.8555065393447876,
+      "learning_rate": 8.115056664867782e-05,
+      "loss": 3.1842,
+      "step": 80250
+    },
+    {
+      "epoch": 8.657681940700808,
+      "grad_norm": 0.8884170055389404,
+      "learning_rate": 8.082676740420938e-05,
+      "loss": 3.1733,
+      "step": 80300
+    },
+    {
+      "epoch": 8.663072776280323,
+      "grad_norm": 0.8410635590553284,
+      "learning_rate": 8.050296815974096e-05,
+      "loss": 3.1858,
+      "step": 80350
+    },
+    {
+      "epoch": 8.668463611859838,
+      "grad_norm": 0.9378628134727478,
+      "learning_rate": 8.017916891527252e-05,
+      "loss": 3.182,
+      "step": 80400
+    },
+    {
+      "epoch": 8.673854447439354,
+      "grad_norm": 0.854771077632904,
+      "learning_rate": 7.985536967080409e-05,
+      "loss": 3.1823,
+      "step": 80450
+    },
+    {
+      "epoch": 8.679245283018869,
+      "grad_norm": 0.8428061008453369,
+      "learning_rate": 7.953157042633566e-05,
+      "loss": 3.1824,
+      "step": 80500
+    },
+    {
+      "epoch": 8.684636118598382,
+      "grad_norm": 0.9225944876670837,
+      "learning_rate": 7.920777118186724e-05,
+      "loss": 3.1869,
+      "step": 80550
+    },
+    {
+      "epoch": 8.690026954177897,
+      "grad_norm": 0.8428223729133606,
+      "learning_rate": 7.888397193739879e-05,
+      "loss": 3.1919,
+      "step": 80600
+    },
+    {
+      "epoch": 8.695417789757412,
+      "grad_norm": 0.8682796359062195,
+      "learning_rate": 7.856017269293037e-05,
+      "loss": 3.1844,
+      "step": 80650
+    },
+    {
+      "epoch": 8.700808625336927,
+      "grad_norm": 0.8594951629638672,
+      "learning_rate": 7.823637344846195e-05,
+      "loss": 3.1718,
+      "step": 80700
+    },
+    {
+      "epoch": 8.706199460916443,
+      "grad_norm": 0.860109806060791,
+      "learning_rate": 7.791257420399353e-05,
+      "loss": 3.1797,
+      "step": 80750
+    },
+    {
+      "epoch": 8.711590296495956,
+      "grad_norm": 0.8337358236312866,
+      "learning_rate": 7.758877495952508e-05,
+      "loss": 3.1929,
+      "step": 80800
+    },
+    {
+      "epoch": 8.716981132075471,
+      "grad_norm": 0.8567845821380615,
+      "learning_rate": 7.726497571505665e-05,
+      "loss": 3.1878,
+      "step": 80850
+    },
+    {
+      "epoch": 8.722371967654986,
+      "grad_norm": 0.8769427537918091,
+      "learning_rate": 7.694117647058823e-05,
+      "loss": 3.1755,
+      "step": 80900
+    },
+    {
+      "epoch": 8.727762803234501,
+      "grad_norm": 0.8659102916717529,
+      "learning_rate": 7.661737722611981e-05,
+      "loss": 3.1753,
+      "step": 80950
+    },
+    {
+      "epoch": 8.733153638814017,
+      "grad_norm": 0.859886109828949,
+      "learning_rate": 7.629357798165137e-05,
+      "loss": 3.1876,
+      "step": 81000
+    },
+    {
+      "epoch": 8.733153638814017,
+      "eval_accuracy": 0.39203667773803375,
+      "eval_loss": 3.31907057762146,
+      "eval_runtime": 185.7569,
+      "eval_samples_per_second": 96.96,
+      "eval_steps_per_second": 6.062,
+      "step": 81000
+    },
+    {
+      "epoch": 8.738544474393532,
+      "grad_norm": 0.8443482518196106,
+      "learning_rate": 7.596977873718294e-05,
+      "loss": 3.1846,
+      "step": 81050
+    },
+    {
+      "epoch": 8.743935309973045,
+      "grad_norm": 0.9283762574195862,
+      "learning_rate": 7.564597949271451e-05,
+      "loss": 3.1862,
+      "step": 81100
+    },
+    {
+      "epoch": 8.74932614555256,
+      "grad_norm": 0.8833479881286621,
+      "learning_rate": 7.532218024824608e-05,
+      "loss": 3.1833,
+      "step": 81150
+    },
+    {
+      "epoch": 8.754716981132075,
+      "grad_norm": 0.8351959586143494,
+      "learning_rate": 7.499838100377764e-05,
+      "loss": 3.1639,
+      "step": 81200
+    },
+    {
+      "epoch": 8.76010781671159,
+      "grad_norm": 0.9914209842681885,
+      "learning_rate": 7.467458175930922e-05,
+      "loss": 3.1805,
+      "step": 81250
+    },
+    {
+      "epoch": 8.765498652291106,
+      "grad_norm": 0.8483186364173889,
+      "learning_rate": 7.435078251484078e-05,
+      "loss": 3.1796,
+      "step": 81300
+    },
+    {
+      "epoch": 8.77088948787062,
+      "grad_norm": 0.8663406372070312,
+      "learning_rate": 7.402698327037236e-05,
+      "loss": 3.1886,
+      "step": 81350
+    },
+    {
+      "epoch": 8.776280323450134,
+      "grad_norm": 0.830612063407898,
+      "learning_rate": 7.370318402590393e-05,
+      "loss": 3.1853,
+      "step": 81400
+    },
+    {
+      "epoch": 8.78167115902965,
+      "grad_norm": 0.8934940099716187,
+      "learning_rate": 7.33793847814355e-05,
+      "loss": 3.1917,
+      "step": 81450
+    },
+    {
+      "epoch": 8.787061994609164,
+      "grad_norm": 0.8502947688102722,
+      "learning_rate": 7.305558553696707e-05,
+      "loss": 3.1983,
+      "step": 81500
+    },
+    {
+      "epoch": 8.79245283018868,
+      "grad_norm": 0.9056462645530701,
+      "learning_rate": 7.273178629249865e-05,
+      "loss": 3.1873,
+      "step": 81550
+    },
+    {
+      "epoch": 8.797843665768195,
+      "grad_norm": 0.8715682029724121,
+      "learning_rate": 7.240798704803021e-05,
+      "loss": 3.1887,
+      "step": 81600
+    },
+    {
+      "epoch": 8.80323450134771,
+      "grad_norm": 0.9065247774124146,
+      "learning_rate": 7.208418780356179e-05,
+      "loss": 3.1864,
+      "step": 81650
+    },
+    {
+      "epoch": 8.808625336927223,
+      "grad_norm": 0.9041933417320251,
+      "learning_rate": 7.176038855909337e-05,
+      "loss": 3.1906,
+      "step": 81700
+    },
+    {
+      "epoch": 8.814016172506738,
+      "grad_norm": 0.8660803437232971,
+      "learning_rate": 7.143658931462493e-05,
+      "loss": 3.1824,
+      "step": 81750
+    },
+    {
+      "epoch": 8.819407008086253,
+      "grad_norm": 0.9310882091522217,
+      "learning_rate": 7.11127900701565e-05,
+      "loss": 3.1804,
+      "step": 81800
+    },
+    {
+      "epoch": 8.824797843665769,
+      "grad_norm": 0.8923643231391907,
+      "learning_rate": 7.078899082568807e-05,
+      "loss": 3.1797,
+      "step": 81850
+    },
+    {
+      "epoch": 8.830188679245284,
+      "grad_norm": 0.8718100190162659,
+      "learning_rate": 7.046519158121964e-05,
+      "loss": 3.1745,
+      "step": 81900
+    },
+    {
+      "epoch": 8.835579514824797,
+      "grad_norm": 0.8527472019195557,
+      "learning_rate": 7.014139233675121e-05,
+      "loss": 3.2002,
+      "step": 81950
+    },
+    {
+      "epoch": 8.840970350404312,
+      "grad_norm": 0.8355522155761719,
+      "learning_rate": 6.981759309228278e-05,
+      "loss": 3.1844,
+      "step": 82000
+    },
+    {
+      "epoch": 8.840970350404312,
+      "eval_accuracy": 0.3921403325930088,
+      "eval_loss": 3.3155720233917236,
+      "eval_runtime": 185.6245,
+      "eval_samples_per_second": 97.029,
+      "eval_steps_per_second": 6.066,
+      "step": 82000
+    },
+    {
+      "epoch": 8.846361185983827,
+      "grad_norm": 0.81675785779953,
+      "learning_rate": 6.949379384781434e-05,
+      "loss": 3.1889,
+      "step": 82050
+    },
+    {
+      "epoch": 8.851752021563343,
+      "grad_norm": 0.8978166580200195,
+      "learning_rate": 6.916999460334592e-05,
+      "loss": 3.1782,
+      "step": 82100
+    },
+    {
+      "epoch": 8.857142857142858,
+      "grad_norm": 0.8902832269668579,
+      "learning_rate": 6.884619535887748e-05,
+      "loss": 3.1878,
+      "step": 82150
+    },
+    {
+      "epoch": 8.862533692722373,
+      "grad_norm": 0.8617354035377502,
+      "learning_rate": 6.852239611440906e-05,
+      "loss": 3.1842,
+      "step": 82200
+    },
+    {
+      "epoch": 8.867924528301886,
+      "grad_norm": 0.856952965259552,
+      "learning_rate": 6.819859686994062e-05,
+      "loss": 3.1873,
+      "step": 82250
+    },
+    {
+      "epoch": 8.873315363881401,
+      "grad_norm": 0.8991112112998962,
+      "learning_rate": 6.78747976254722e-05,
+      "loss": 3.1938,
+      "step": 82300
+    },
+    {
+      "epoch": 8.878706199460916,
+      "grad_norm": 0.853913426399231,
+      "learning_rate": 6.755747436589314e-05,
+      "loss": 3.1961,
+      "step": 82350
+    },
+    {
+      "epoch": 8.884097035040432,
+      "grad_norm": 0.8485080599784851,
+      "learning_rate": 6.72336751214247e-05,
+      "loss": 3.1733,
+      "step": 82400
+    },
+    {
+      "epoch": 8.889487870619947,
+      "grad_norm": 0.8443381786346436,
+      "learning_rate": 6.690987587695628e-05,
+      "loss": 3.193,
+      "step": 82450
+    },
+    {
+      "epoch": 8.89487870619946,
+      "grad_norm": 0.8695724010467529,
+      "learning_rate": 6.658607663248785e-05,
+      "loss": 3.1714,
+      "step": 82500
+    },
+    {
+      "epoch": 8.900269541778975,
+      "grad_norm": 0.8814206719398499,
+      "learning_rate": 6.626227738801942e-05,
+      "loss": 3.1947,
+      "step": 82550
+    },
+    {
+      "epoch": 8.90566037735849,
+      "grad_norm": 0.8670693039894104,
+      "learning_rate": 6.593847814355099e-05,
+      "loss": 3.162,
+      "step": 82600
+    },
+    {
+      "epoch": 8.911051212938006,
+      "grad_norm": 0.9399033784866333,
+      "learning_rate": 6.561467889908257e-05,
+      "loss": 3.1753,
+      "step": 82650
+    },
+    {
+      "epoch": 8.91644204851752,
+      "grad_norm": 0.8525042533874512,
+      "learning_rate": 6.529087965461413e-05,
+      "loss": 3.1797,
+      "step": 82700
+    },
+    {
+      "epoch": 8.921832884097036,
+      "grad_norm": 0.9035705327987671,
+      "learning_rate": 6.496708041014571e-05,
+      "loss": 3.1747,
+      "step": 82750
+    },
+    {
+      "epoch": 8.92722371967655,
+      "grad_norm": 0.8558206558227539,
+      "learning_rate": 6.464328116567727e-05,
+      "loss": 3.1907,
+      "step": 82800
+    },
+    {
+      "epoch": 8.932614555256064,
+      "grad_norm": 0.9054684638977051,
+      "learning_rate": 6.431948192120885e-05,
+      "loss": 3.1862,
+      "step": 82850
+    },
+    {
+      "epoch": 8.93800539083558,
+      "grad_norm": 0.8330327272415161,
+      "learning_rate": 6.399568267674041e-05,
+      "loss": 3.1887,
+      "step": 82900
+    },
+    {
+      "epoch": 8.943396226415095,
+      "grad_norm": 0.9181053638458252,
+      "learning_rate": 6.367188343227199e-05,
+      "loss": 3.1898,
+      "step": 82950
+    },
+    {
+      "epoch": 8.94878706199461,
+      "grad_norm": 0.8530136942863464,
+      "learning_rate": 6.334808418780356e-05,
+      "loss": 3.1661,
+      "step": 83000
+    },
+    {
+      "epoch": 8.94878706199461,
+      "eval_accuracy": 0.3925132293039881,
+      "eval_loss": 3.3129775524139404,
+      "eval_runtime": 185.6884,
+      "eval_samples_per_second": 96.996,
+      "eval_steps_per_second": 6.064,
+      "step": 83000
+    },
+    {
+      "epoch": 8.954177897574123,
+      "grad_norm": 0.8998593688011169,
+      "learning_rate": 6.302428494333512e-05,
+      "loss": 3.1957,
+      "step": 83050
+    },
+    {
+      "epoch": 8.959568733153638,
+      "grad_norm": 0.8955749273300171,
+      "learning_rate": 6.27004856988667e-05,
+      "loss": 3.1811,
+      "step": 83100
+    },
+    {
+      "epoch": 8.964959568733153,
+      "grad_norm": 0.9160959124565125,
+      "learning_rate": 6.237668645439826e-05,
+      "loss": 3.1838,
+      "step": 83150
+    },
+    {
+      "epoch": 8.970350404312669,
+      "grad_norm": 0.8576794862747192,
+      "learning_rate": 6.205288720992984e-05,
+      "loss": 3.1833,
+      "step": 83200
+    },
+    {
+      "epoch": 8.975741239892184,
+      "grad_norm": 0.8886733651161194,
+      "learning_rate": 6.17290879654614e-05,
+      "loss": 3.1904,
+      "step": 83250
+    },
+    {
+      "epoch": 8.981132075471699,
+      "grad_norm": 0.8392205834388733,
+      "learning_rate": 6.140528872099298e-05,
+      "loss": 3.1966,
+      "step": 83300
+    },
+    {
+      "epoch": 8.986522911051212,
+      "grad_norm": 0.8943293690681458,
+      "learning_rate": 6.108148947652455e-05,
+      "loss": 3.1926,
+      "step": 83350
+    },
+    {
+      "epoch": 8.991913746630727,
+      "grad_norm": 0.8496133685112,
+      "learning_rate": 6.075769023205612e-05,
+      "loss": 3.1777,
+      "step": 83400
+    },
+    {
+      "epoch": 8.997304582210242,
+      "grad_norm": 0.8753985166549683,
+      "learning_rate": 6.043389098758769e-05,
+      "loss": 3.1804,
+      "step": 83450
+    },
+    {
+      "epoch": 9.002695417789758,
+      "grad_norm": 0.902191162109375,
+      "learning_rate": 6.0110091743119265e-05,
+      "loss": 3.1564,
+      "step": 83500
+    },
+    {
+      "epoch": 9.008086253369273,
+      "grad_norm": 0.8964595198631287,
+      "learning_rate": 5.978629249865083e-05,
+      "loss": 3.1177,
+      "step": 83550
+    },
+    {
+      "epoch": 9.013477088948788,
+      "grad_norm": 0.8484180569648743,
+      "learning_rate": 5.94624932541824e-05,
+      "loss": 3.1421,
+      "step": 83600
+    },
+    {
+      "epoch": 9.018867924528301,
+      "grad_norm": 0.870861828327179,
+      "learning_rate": 5.913869400971397e-05,
+      "loss": 3.1533,
+      "step": 83650
+    },
+    {
+      "epoch": 9.024258760107816,
+      "grad_norm": 0.8558200001716614,
+      "learning_rate": 5.881489476524554e-05,
+      "loss": 3.1038,
+      "step": 83700
+    },
+    {
+      "epoch": 9.029649595687331,
+      "grad_norm": 0.8519108891487122,
+      "learning_rate": 5.849109552077711e-05,
+      "loss": 3.106,
+      "step": 83750
+    },
+    {
+      "epoch": 9.035040431266847,
+      "grad_norm": 0.8885248899459839,
+      "learning_rate": 5.8167296276308683e-05,
+      "loss": 3.1195,
+      "step": 83800
+    },
+    {
+      "epoch": 9.040431266846362,
+      "grad_norm": 0.9055321216583252,
+      "learning_rate": 5.784349703184026e-05,
+      "loss": 3.1324,
+      "step": 83850
+    },
+    {
+      "epoch": 9.045822102425875,
+      "grad_norm": 0.891910195350647,
+      "learning_rate": 5.7519697787371825e-05,
+      "loss": 3.1197,
+      "step": 83900
+    },
+    {
+      "epoch": 9.05121293800539,
+      "grad_norm": 0.8641490936279297,
+      "learning_rate": 5.7195898542903396e-05,
+      "loss": 3.1011,
+      "step": 83950
+    },
+    {
+      "epoch": 9.056603773584905,
+      "grad_norm": 0.8966986536979675,
+      "learning_rate": 5.687209929843497e-05,
+      "loss": 3.1287,
+      "step": 84000
+    },
+    {
+      "epoch": 9.056603773584905,
+      "eval_accuracy": 0.39250475437873855,
+      "eval_loss": 3.315013885498047,
+      "eval_runtime": 185.6065,
+      "eval_samples_per_second": 97.039,
+      "eval_steps_per_second": 6.067,
+      "step": 84000
+    },
+    {
+      "epoch": 9.06199460916442,
+      "grad_norm": 0.8515587449073792,
+      "learning_rate": 5.654830005396654e-05,
+      "loss": 3.1257,
+      "step": 84050
+    },
+    {
+      "epoch": 9.067385444743936,
+      "grad_norm": 0.9007041454315186,
+      "learning_rate": 5.62245008094981e-05,
+      "loss": 3.129,
+      "step": 84100
+    },
+    {
+      "epoch": 9.07277628032345,
+      "grad_norm": 0.8735681176185608,
+      "learning_rate": 5.590070156502968e-05,
+      "loss": 3.1409,
+      "step": 84150
+    },
+    {
+      "epoch": 9.078167115902964,
+      "grad_norm": 0.8952208161354065,
+      "learning_rate": 5.5576902320561244e-05,
+      "loss": 3.1158,
+      "step": 84200
+    },
+    {
+      "epoch": 9.08355795148248,
+      "grad_norm": 0.8467687964439392,
+      "learning_rate": 5.525310307609282e-05,
+      "loss": 3.1386,
+      "step": 84250
+    },
+    {
+      "epoch": 9.088948787061994,
+      "grad_norm": 0.8729225397109985,
+      "learning_rate": 5.4929303831624386e-05,
+      "loss": 3.1221,
+      "step": 84300
+    },
+    {
+      "epoch": 9.09433962264151,
+      "grad_norm": 0.8786865472793579,
+      "learning_rate": 5.460550458715596e-05,
+      "loss": 3.1296,
+      "step": 84350
+    },
+    {
+      "epoch": 9.099730458221025,
+      "grad_norm": 0.8809643387794495,
+      "learning_rate": 5.428170534268753e-05,
+      "loss": 3.0965,
+      "step": 84400
+    },
+    {
+      "epoch": 9.10512129380054,
+      "grad_norm": 0.8985245823860168,
+      "learning_rate": 5.39579060982191e-05,
+      "loss": 3.1118,
+      "step": 84450
+    },
+    {
+      "epoch": 9.110512129380053,
+      "grad_norm": 0.8574363589286804,
+      "learning_rate": 5.363410685375067e-05,
+      "loss": 3.1332,
+      "step": 84500
+    },
+    {
+      "epoch": 9.115902964959568,
+      "grad_norm": 0.8448746800422668,
+      "learning_rate": 5.331030760928224e-05,
+      "loss": 3.1101,
+      "step": 84550
+    },
+    {
+      "epoch": 9.121293800539084,
+      "grad_norm": 0.8848583698272705,
+      "learning_rate": 5.2986508364813804e-05,
+      "loss": 3.1249,
+      "step": 84600
+    },
+    {
+      "epoch": 9.126684636118599,
+      "grad_norm": 0.8773154020309448,
+      "learning_rate": 5.266270912034538e-05,
+      "loss": 3.1249,
+      "step": 84650
+    },
+    {
+      "epoch": 9.132075471698114,
+      "grad_norm": 0.9096025824546814,
+      "learning_rate": 5.233890987587695e-05,
+      "loss": 3.1322,
+      "step": 84700
+    },
+    {
+      "epoch": 9.137466307277627,
+      "grad_norm": 0.8677944540977478,
+      "learning_rate": 5.2015110631408524e-05,
+      "loss": 3.1361,
+      "step": 84750
+    },
+    {
+      "epoch": 9.142857142857142,
+      "grad_norm": 0.8547237515449524,
+      "learning_rate": 5.1691311386940094e-05,
+      "loss": 3.1314,
+      "step": 84800
+    },
+    {
+      "epoch": 9.148247978436657,
+      "grad_norm": 0.8557731509208679,
+      "learning_rate": 5.136751214247166e-05,
+      "loss": 3.1198,
+      "step": 84850
+    },
+    {
+      "epoch": 9.153638814016173,
+      "grad_norm": 0.9102908968925476,
+      "learning_rate": 5.1050188882892604e-05,
+      "loss": 3.1234,
+      "step": 84900
+    },
+    {
+      "epoch": 9.159029649595688,
+      "grad_norm": 0.877302348613739,
+      "learning_rate": 5.072638963842417e-05,
+      "loss": 3.1115,
+      "step": 84950
+    },
+    {
+      "epoch": 9.164420485175203,
+      "grad_norm": 0.8711842894554138,
+      "learning_rate": 5.0402590393955746e-05,
+      "loss": 3.1215,
+      "step": 85000
+    },
+    {
+      "epoch": 9.164420485175203,
+      "eval_accuracy": 0.3926480675377658,
+      "eval_loss": 3.3146796226501465,
+      "eval_runtime": 185.9457,
+      "eval_samples_per_second": 96.862,
+      "eval_steps_per_second": 6.056,
+      "step": 85000
+    },
+    {
+      "epoch": 9.169811320754716,
+      "grad_norm": 0.8675227165222168,
+      "learning_rate": 5.007879114948731e-05,
+      "loss": 3.1339,
+      "step": 85050
+    },
+    {
+      "epoch": 9.175202156334231,
+      "grad_norm": 0.8620361685752869,
+      "learning_rate": 4.975499190501889e-05,
+      "loss": 3.1304,
+      "step": 85100
+    },
+    {
+      "epoch": 9.180592991913747,
+      "grad_norm": 0.8773727416992188,
+      "learning_rate": 4.943119266055045e-05,
+      "loss": 3.1371,
+      "step": 85150
+    },
+    {
+      "epoch": 9.185983827493262,
+      "grad_norm": 0.8646271228790283,
+      "learning_rate": 4.910739341608202e-05,
+      "loss": 3.1343,
+      "step": 85200
+    },
+    {
+      "epoch": 9.191374663072777,
+      "grad_norm": 0.8819538950920105,
+      "learning_rate": 4.87835941716136e-05,
+      "loss": 3.1311,
+      "step": 85250
+    },
+    {
+      "epoch": 9.19676549865229,
+      "grad_norm": 0.8820063471794128,
+      "learning_rate": 4.8459794927145164e-05,
+      "loss": 3.1447,
+      "step": 85300
+    },
+    {
+      "epoch": 9.202156334231805,
+      "grad_norm": 0.8823925852775574,
+      "learning_rate": 4.813599568267674e-05,
+      "loss": 3.1393,
+      "step": 85350
+    },
+    {
+      "epoch": 9.20754716981132,
+      "grad_norm": 0.8114355802536011,
+      "learning_rate": 4.7812196438208306e-05,
+      "loss": 3.1239,
+      "step": 85400
+    },
+    {
+      "epoch": 9.212938005390836,
+      "grad_norm": 0.8829519152641296,
+      "learning_rate": 4.748839719373988e-05,
+      "loss": 3.1238,
+      "step": 85450
+    },
+    {
+      "epoch": 9.21832884097035,
+      "grad_norm": 0.8762984871864319,
+      "learning_rate": 4.716459794927145e-05,
+      "loss": 3.1437,
+      "step": 85500
+    },
+    {
+      "epoch": 9.223719676549866,
+      "grad_norm": 0.8304125070571899,
+      "learning_rate": 4.684079870480302e-05,
+      "loss": 3.1455,
+      "step": 85550
+    },
+    {
+      "epoch": 9.22911051212938,
+      "grad_norm": 0.8846282362937927,
+      "learning_rate": 4.651699946033459e-05,
+      "loss": 3.1355,
+      "step": 85600
+    },
+    {
+      "epoch": 9.234501347708894,
+      "grad_norm": 0.8445920348167419,
+      "learning_rate": 4.619320021586616e-05,
+      "loss": 3.1317,
+      "step": 85650
+    },
+    {
+      "epoch": 9.23989218328841,
+      "grad_norm": 0.8530385494232178,
+      "learning_rate": 4.5869400971397725e-05,
+      "loss": 3.1307,
+      "step": 85700
+    },
+    {
+      "epoch": 9.245283018867925,
+      "grad_norm": 0.8589995503425598,
+      "learning_rate": 4.55456017269293e-05,
+      "loss": 3.1322,
+      "step": 85750
+    },
+    {
+      "epoch": 9.25067385444744,
+      "grad_norm": 0.9057507514953613,
+      "learning_rate": 4.5221802482460866e-05,
+      "loss": 3.1221,
+      "step": 85800
+    },
+    {
+      "epoch": 9.256064690026955,
+      "grad_norm": 0.8584359884262085,
+      "learning_rate": 4.4898003237992444e-05,
+      "loss": 3.1154,
+      "step": 85850
+    },
+    {
+      "epoch": 9.261455525606468,
+      "grad_norm": 0.8353499174118042,
+      "learning_rate": 4.457420399352401e-05,
+      "loss": 3.1325,
+      "step": 85900
+    },
+    {
+      "epoch": 9.266846361185983,
+      "grad_norm": 0.8638600707054138,
+      "learning_rate": 4.425040474905558e-05,
+      "loss": 3.1238,
+      "step": 85950
+    },
+    {
+      "epoch": 9.272237196765499,
+      "grad_norm": 0.8481729030609131,
+      "learning_rate": 4.392660550458716e-05,
+      "loss": 3.1458,
+      "step": 86000
+    },
+    {
+      "epoch": 9.272237196765499,
+      "eval_accuracy": 0.393139939160902,
+      "eval_loss": 3.3122284412384033,
+      "eval_runtime": 185.5689,
+      "eval_samples_per_second": 97.058,
+      "eval_steps_per_second": 6.068,
+      "step": 86000
+    },
+    {
+      "epoch": 9.277628032345014,
+      "grad_norm": 0.8732629418373108,
+      "learning_rate": 4.360280626011872e-05,
+      "loss": 3.1347,
+      "step": 86050
+    },
+    {
+      "epoch": 9.283018867924529,
+      "grad_norm": 0.8761304616928101,
+      "learning_rate": 4.32790070156503e-05,
+      "loss": 3.1199,
+      "step": 86100
+    },
+    {
+      "epoch": 9.288409703504042,
+      "grad_norm": 0.8342899680137634,
+      "learning_rate": 4.295520777118186e-05,
+      "loss": 3.1214,
+      "step": 86150
+    },
+    {
+      "epoch": 9.293800539083557,
+      "grad_norm": 0.8341020941734314,
+      "learning_rate": 4.263140852671344e-05,
+      "loss": 3.1309,
+      "step": 86200
+    },
+    {
+      "epoch": 9.299191374663073,
+      "grad_norm": 0.8385040760040283,
+      "learning_rate": 4.2307609282245004e-05,
+      "loss": 3.1302,
+      "step": 86250
+    },
+    {
+      "epoch": 9.304582210242588,
+      "grad_norm": 0.8832799792289734,
+      "learning_rate": 4.1983810037776575e-05,
+      "loss": 3.1289,
+      "step": 86300
+    },
+    {
+      "epoch": 9.309973045822103,
+      "grad_norm": 0.8826196789741516,
+      "learning_rate": 4.1660010793308146e-05,
+      "loss": 3.1227,
+      "step": 86350
+    },
+    {
+      "epoch": 9.315363881401618,
+      "grad_norm": 0.9327345490455627,
+      "learning_rate": 4.133621154883972e-05,
+      "loss": 3.1507,
+      "step": 86400
+    },
+    {
+      "epoch": 9.320754716981131,
+      "grad_norm": 0.9211543798446655,
+      "learning_rate": 4.101241230437128e-05,
+      "loss": 3.1244,
+      "step": 86450
+    },
+    {
+      "epoch": 9.326145552560646,
+      "grad_norm": 0.9178559184074402,
+      "learning_rate": 4.068861305990286e-05,
+      "loss": 3.1399,
+      "step": 86500
+    },
+    {
+      "epoch": 9.331536388140162,
+      "grad_norm": 0.8896996378898621,
+      "learning_rate": 4.036481381543442e-05,
+      "loss": 3.145,
+      "step": 86550
+    },
+    {
+      "epoch": 9.336927223719677,
+      "grad_norm": 0.8767986297607422,
+      "learning_rate": 4.0041014570966e-05,
+      "loss": 3.1338,
+      "step": 86600
+    },
+    {
+      "epoch": 9.342318059299192,
+      "grad_norm": 0.8779925107955933,
+      "learning_rate": 3.9717215326497565e-05,
+      "loss": 3.1334,
+      "step": 86650
+    },
+    {
+      "epoch": 9.347708894878707,
+      "grad_norm": 0.9277268648147583,
+      "learning_rate": 3.9393416082029136e-05,
+      "loss": 3.1463,
+      "step": 86700
+    },
+    {
+      "epoch": 9.35309973045822,
+      "grad_norm": 0.8903272747993469,
+      "learning_rate": 3.9069616837560706e-05,
+      "loss": 3.1574,
+      "step": 86750
+    },
+    {
+      "epoch": 9.358490566037736,
+      "grad_norm": 0.853990375995636,
+      "learning_rate": 3.874581759309228e-05,
+      "loss": 3.1387,
+      "step": 86800
+    },
+    {
+      "epoch": 9.36388140161725,
+      "grad_norm": 0.9055298566818237,
+      "learning_rate": 3.8422018348623855e-05,
+      "loss": 3.1293,
+      "step": 86850
+    },
+    {
+      "epoch": 9.369272237196766,
+      "grad_norm": 0.9027107357978821,
+      "learning_rate": 3.809821910415542e-05,
+      "loss": 3.1414,
+      "step": 86900
+    },
+    {
+      "epoch": 9.374663072776281,
+      "grad_norm": 0.9010524749755859,
+      "learning_rate": 3.7774419859687e-05,
+      "loss": 3.1183,
+      "step": 86950
+    },
+    {
+      "epoch": 9.380053908355794,
+      "grad_norm": 0.8620128631591797,
+      "learning_rate": 3.745062061521856e-05,
+      "loss": 3.1262,
+      "step": 87000
+    },
+    {
+      "epoch": 9.380053908355794,
+      "eval_accuracy": 0.3933640900684633,
+      "eval_loss": 3.3104584217071533,
+      "eval_runtime": 185.6797,
+      "eval_samples_per_second": 97.0,
+      "eval_steps_per_second": 6.064,
+      "step": 87000
+    },
+    {
+      "epoch": 9.38544474393531,
+      "grad_norm": 0.8787023425102234,
+      "learning_rate": 3.712682137075013e-05,
+      "loss": 3.1342,
+      "step": 87050
+    },
+    {
+      "epoch": 9.390835579514825,
+      "grad_norm": 0.8926424384117126,
+      "learning_rate": 3.68030221262817e-05,
+      "loss": 3.1222,
+      "step": 87100
+    },
+    {
+      "epoch": 9.39622641509434,
+      "grad_norm": 0.9072372317314148,
+      "learning_rate": 3.6479222881813274e-05,
+      "loss": 3.1252,
+      "step": 87150
+    },
+    {
+      "epoch": 9.401617250673855,
+      "grad_norm": 0.8576307892799377,
+      "learning_rate": 3.6155423637344844e-05,
+      "loss": 3.1507,
+      "step": 87200
+    },
+    {
+      "epoch": 9.40700808625337,
+      "grad_norm": 0.8560314178466797,
+      "learning_rate": 3.5831624392876415e-05,
+      "loss": 3.1373,
+      "step": 87250
+    },
+    {
+      "epoch": 9.412398921832883,
+      "grad_norm": 0.920363187789917,
+      "learning_rate": 3.5507825148407986e-05,
+      "loss": 3.1412,
+      "step": 87300
+    },
+    {
+      "epoch": 9.417789757412399,
+      "grad_norm": 0.8651599287986755,
+      "learning_rate": 3.518402590393956e-05,
+      "loss": 3.1383,
+      "step": 87350
+    },
+    {
+      "epoch": 9.423180592991914,
+      "grad_norm": 0.853699803352356,
+      "learning_rate": 3.486022665947113e-05,
+      "loss": 3.1338,
+      "step": 87400
+    },
+    {
+      "epoch": 9.428571428571429,
+      "grad_norm": 0.8881181478500366,
+      "learning_rate": 3.45364274150027e-05,
+      "loss": 3.1422,
+      "step": 87450
+    },
+    {
+      "epoch": 9.433962264150944,
+      "grad_norm": 0.878387987613678,
+      "learning_rate": 3.421262817053426e-05,
+      "loss": 3.1428,
+      "step": 87500
+    },
+    {
+      "epoch": 9.439353099730457,
+      "grad_norm": 0.8532453775405884,
+      "learning_rate": 3.3888828926065834e-05,
+      "loss": 3.1373,
+      "step": 87550
+    },
+    {
+      "epoch": 9.444743935309972,
+      "grad_norm": 0.8711896538734436,
+      "learning_rate": 3.3565029681597405e-05,
+      "loss": 3.1107,
+      "step": 87600
+    },
+    {
+      "epoch": 9.450134770889488,
+      "grad_norm": 0.8701107501983643,
+      "learning_rate": 3.3241230437128976e-05,
+      "loss": 3.1399,
+      "step": 87650
+    },
+    {
+      "epoch": 9.455525606469003,
+      "grad_norm": 0.8386976718902588,
+      "learning_rate": 3.2917431192660547e-05,
+      "loss": 3.132,
+      "step": 87700
+    },
+    {
+      "epoch": 9.460916442048518,
+      "grad_norm": 0.9316996335983276,
+      "learning_rate": 3.259363194819212e-05,
+      "loss": 3.1294,
+      "step": 87750
+    },
+    {
+      "epoch": 9.466307277628033,
+      "grad_norm": 0.881655216217041,
+      "learning_rate": 3.226983270372369e-05,
+      "loss": 3.1184,
+      "step": 87800
+    },
+    {
+      "epoch": 9.471698113207546,
+      "grad_norm": 0.8883360624313354,
+      "learning_rate": 3.194603345925526e-05,
+      "loss": 3.1308,
+      "step": 87850
+    },
+    {
+      "epoch": 9.477088948787062,
+      "grad_norm": 0.9252742528915405,
+      "learning_rate": 3.162223421478683e-05,
+      "loss": 3.1192,
+      "step": 87900
+    },
+    {
+      "epoch": 9.482479784366577,
+      "grad_norm": 0.8270439505577087,
+      "learning_rate": 3.12984349703184e-05,
+      "loss": 3.1203,
+      "step": 87950
+    },
+    {
+      "epoch": 9.487870619946092,
+      "grad_norm": 0.8697919249534607,
+      "learning_rate": 3.097463572584997e-05,
+      "loss": 3.1322,
+      "step": 88000
+    },
+    {
+      "epoch": 9.487870619946092,
+      "eval_accuracy": 0.3935639027291541,
+      "eval_loss": 3.3076882362365723,
+      "eval_runtime": 185.7269,
+      "eval_samples_per_second": 96.976,
+      "eval_steps_per_second": 6.063,
+      "step": 88000
+    },
+    {
+      "epoch": 9.493261455525607,
+      "grad_norm": 0.8772650957107544,
+      "learning_rate": 3.065083648138154e-05,
+      "loss": 3.13,
+      "step": 88050
+    },
+    {
+      "epoch": 9.498652291105122,
+      "grad_norm": 0.8854417204856873,
+      "learning_rate": 3.0327037236913114e-05,
+      "loss": 3.127,
+      "step": 88100
+    },
+    {
+      "epoch": 9.504043126684635,
+      "grad_norm": 0.8294955492019653,
+      "learning_rate": 3.000323799244468e-05,
+      "loss": 3.1534,
+      "step": 88150
+    },
+    {
+      "epoch": 9.50943396226415,
+      "grad_norm": 0.9030188322067261,
+      "learning_rate": 2.9679438747976252e-05,
+      "loss": 3.1342,
+      "step": 88200
+    },
+    {
+      "epoch": 9.514824797843666,
+      "grad_norm": 0.8873876929283142,
+      "learning_rate": 2.9355639503507823e-05,
+      "loss": 3.1212,
+      "step": 88250
+    },
+    {
+      "epoch": 9.520215633423181,
+      "grad_norm": 0.8966760635375977,
+      "learning_rate": 2.9031840259039394e-05,
+      "loss": 3.1399,
+      "step": 88300
+    },
+    {
+      "epoch": 9.525606469002696,
+      "grad_norm": 0.8984833359718323,
+      "learning_rate": 2.8708041014570965e-05,
+      "loss": 3.1334,
+      "step": 88350
+    },
+    {
+      "epoch": 9.530997304582211,
+      "grad_norm": 0.9016095399856567,
+      "learning_rate": 2.8384241770102532e-05,
+      "loss": 3.1206,
+      "step": 88400
+    },
+    {
+      "epoch": 9.536388140161725,
+      "grad_norm": 0.9135856628417969,
+      "learning_rate": 2.8060442525634103e-05,
+      "loss": 3.129,
+      "step": 88450
+    },
+    {
+      "epoch": 9.54177897574124,
+      "grad_norm": 0.858954668045044,
+      "learning_rate": 2.7736643281165674e-05,
+      "loss": 3.1412,
+      "step": 88500
+    },
+    {
+      "epoch": 9.547169811320755,
+      "grad_norm": 0.8992297053337097,
+      "learning_rate": 2.7412844036697245e-05,
+      "loss": 3.1215,
+      "step": 88550
+    },
+    {
+      "epoch": 9.55256064690027,
+      "grad_norm": 0.9010116457939148,
+      "learning_rate": 2.7089044792228816e-05,
+      "loss": 3.1365,
+      "step": 88600
+    },
+    {
+      "epoch": 9.557951482479785,
+      "grad_norm": 0.9245710968971252,
+      "learning_rate": 2.6771721532649758e-05,
+      "loss": 3.131,
+      "step": 88650
+    },
+    {
+      "epoch": 9.563342318059298,
+      "grad_norm": 0.8326699137687683,
+      "learning_rate": 2.6447922288181325e-05,
+      "loss": 3.1305,
+      "step": 88700
+    },
+    {
+      "epoch": 9.568733153638814,
+      "grad_norm": 0.8716727495193481,
+      "learning_rate": 2.6124123043712896e-05,
+      "loss": 3.1398,
+      "step": 88750
+    },
+    {
+      "epoch": 9.574123989218329,
+      "grad_norm": 0.8772605657577515,
+      "learning_rate": 2.5800323799244467e-05,
+      "loss": 3.1255,
+      "step": 88800
+    },
+    {
+      "epoch": 9.579514824797844,
+      "grad_norm": 0.8987815976142883,
+      "learning_rate": 2.5476524554776038e-05,
+      "loss": 3.1325,
+      "step": 88850
+    },
+    {
+      "epoch": 9.584905660377359,
+      "grad_norm": 0.9095204472541809,
+      "learning_rate": 2.515272531030761e-05,
+      "loss": 3.1554,
+      "step": 88900
+    },
+    {
+      "epoch": 9.590296495956874,
+      "grad_norm": 0.9186028838157654,
+      "learning_rate": 2.4828926065839176e-05,
+      "loss": 3.1346,
+      "step": 88950
+    },
+    {
+      "epoch": 9.595687331536388,
+      "grad_norm": 0.8903992772102356,
+      "learning_rate": 2.4505126821370747e-05,
+      "loss": 3.1109,
+      "step": 89000
+    },
+    {
+      "epoch": 9.595687331536388,
+      "eval_accuracy": 0.393875519211406,
+      "eval_loss": 3.305893898010254,
+      "eval_runtime": 185.8393,
+      "eval_samples_per_second": 96.917,
+      "eval_steps_per_second": 6.059,
+      "step": 89000
+    },
+    {
+      "epoch": 9.601078167115903,
+      "grad_norm": 0.8526430130004883,
+      "learning_rate": 2.4181327576902318e-05,
+      "loss": 3.1424,
+      "step": 89050
+    },
+    {
+      "epoch": 9.606469002695418,
+      "grad_norm": 0.8774048089981079,
+      "learning_rate": 2.385752833243389e-05,
+      "loss": 3.137,
+      "step": 89100
+    },
+    {
+      "epoch": 9.611859838274933,
+      "grad_norm": 0.8635159134864807,
+      "learning_rate": 2.353372908796546e-05,
+      "loss": 3.1438,
+      "step": 89150
+    },
+    {
+      "epoch": 9.617250673854448,
+      "grad_norm": 0.9066774845123291,
+      "learning_rate": 2.3209929843497027e-05,
+      "loss": 3.1302,
+      "step": 89200
+    },
+    {
+      "epoch": 9.622641509433961,
+      "grad_norm": 0.9051966667175293,
+      "learning_rate": 2.2886130599028598e-05,
+      "loss": 3.1364,
+      "step": 89250
+    },
+    {
+      "epoch": 9.628032345013477,
+      "grad_norm": 0.9027606248855591,
+      "learning_rate": 2.256233135456017e-05,
+      "loss": 3.1139,
+      "step": 89300
+    },
+    {
+      "epoch": 9.633423180592992,
+      "grad_norm": 0.8751065135002136,
+      "learning_rate": 2.2238532110091743e-05,
+      "loss": 3.1466,
+      "step": 89350
+    },
+    {
+      "epoch": 9.638814016172507,
+      "grad_norm": 0.8614014983177185,
+      "learning_rate": 2.1914732865623314e-05,
+      "loss": 3.1456,
+      "step": 89400
+    },
+    {
+      "epoch": 9.644204851752022,
+      "grad_norm": 0.880818784236908,
+      "learning_rate": 2.1590933621154885e-05,
+      "loss": 3.1423,
+      "step": 89450
+    },
+    {
+      "epoch": 9.649595687331537,
+      "grad_norm": 0.8860384821891785,
+      "learning_rate": 2.1267134376686453e-05,
+      "loss": 3.1372,
+      "step": 89500
+    },
+    {
+      "epoch": 9.65498652291105,
+      "grad_norm": 0.8799102306365967,
+      "learning_rate": 2.0943335132218024e-05,
+      "loss": 3.1339,
+      "step": 89550
+    },
+    {
+      "epoch": 9.660377358490566,
+      "grad_norm": 0.8428423404693604,
+      "learning_rate": 2.0619535887749594e-05,
+      "loss": 3.1398,
+      "step": 89600
+    },
+    {
+      "epoch": 9.66576819407008,
+      "grad_norm": 0.8757205009460449,
+      "learning_rate": 2.0295736643281165e-05,
+      "loss": 3.1258,
+      "step": 89650
+    },
+    {
+      "epoch": 9.671159029649596,
+      "grad_norm": 0.8617460131645203,
+      "learning_rate": 1.9971937398812736e-05,
+      "loss": 3.1458,
+      "step": 89700
+    },
+    {
+      "epoch": 9.676549865229111,
+      "grad_norm": 0.873406708240509,
+      "learning_rate": 1.9648138154344304e-05,
+      "loss": 3.1199,
+      "step": 89750
+    },
+    {
+      "epoch": 9.681940700808624,
+      "grad_norm": 0.8965728878974915,
+      "learning_rate": 1.9324338909875875e-05,
+      "loss": 3.1237,
+      "step": 89800
+    },
+    {
+      "epoch": 9.68733153638814,
+      "grad_norm": 0.8846343159675598,
+      "learning_rate": 1.9000539665407446e-05,
+      "loss": 3.136,
+      "step": 89850
+    },
+    {
+      "epoch": 9.692722371967655,
+      "grad_norm": 0.8631774187088013,
+      "learning_rate": 1.8676740420939016e-05,
+      "loss": 3.1244,
+      "step": 89900
+    },
+    {
+      "epoch": 9.69811320754717,
+      "grad_norm": 0.9396966099739075,
+      "learning_rate": 1.8352941176470587e-05,
+      "loss": 3.1418,
+      "step": 89950
+    },
+    {
+      "epoch": 9.703504043126685,
+      "grad_norm": 0.8590573072433472,
+      "learning_rate": 1.8029141932002158e-05,
+      "loss": 3.1378,
+      "step": 90000
+    },
+    {
+      "epoch": 9.703504043126685,
+      "eval_accuracy": 0.394136177489273,
+      "eval_loss": 3.303354024887085,
+      "eval_runtime": 185.6906,
+      "eval_samples_per_second": 96.995,
+      "eval_steps_per_second": 6.064,
+      "step": 90000
+    },
+    {
+      "epoch": 9.7088948787062,
+      "grad_norm": 0.9063498973846436,
+      "learning_rate": 1.770534268753373e-05,
+      "loss": 3.123,
+      "step": 90050
+    },
+    {
+      "epoch": 9.714285714285714,
+      "grad_norm": 0.888960599899292,
+      "learning_rate": 1.7381543443065297e-05,
+      "loss": 3.1225,
+      "step": 90100
+    },
+    {
+      "epoch": 9.719676549865229,
+      "grad_norm": 0.9259812235832214,
+      "learning_rate": 1.7057744198596867e-05,
+      "loss": 3.1325,
+      "step": 90150
+    },
+    {
+      "epoch": 9.725067385444744,
+      "grad_norm": 0.9101106524467468,
+      "learning_rate": 1.673394495412844e-05,
+      "loss": 3.1345,
+      "step": 90200
+    },
+    {
+      "epoch": 9.730458221024259,
+      "grad_norm": 0.8655868172645569,
+      "learning_rate": 1.641014570966001e-05,
+      "loss": 3.1231,
+      "step": 90250
+    },
+    {
+      "epoch": 9.735849056603774,
+      "grad_norm": 0.9329912066459656,
+      "learning_rate": 1.608634646519158e-05,
+      "loss": 3.1279,
+      "step": 90300
+    },
+    {
+      "epoch": 9.74123989218329,
+      "grad_norm": 0.9029722809791565,
+      "learning_rate": 1.576254722072315e-05,
+      "loss": 3.1371,
+      "step": 90350
+    },
+    {
+      "epoch": 9.746630727762803,
+      "grad_norm": 0.86487877368927,
+      "learning_rate": 1.5438747976254722e-05,
+      "loss": 3.1243,
+      "step": 90400
+    },
+    {
+      "epoch": 9.752021563342318,
+      "grad_norm": 0.8681514859199524,
+      "learning_rate": 1.5114948731786291e-05,
+      "loss": 3.1388,
+      "step": 90450
+    },
+    {
+      "epoch": 9.757412398921833,
+      "grad_norm": 0.8514367341995239,
+      "learning_rate": 1.4791149487317862e-05,
+      "loss": 3.1367,
+      "step": 90500
+    },
+    {
+      "epoch": 9.762803234501348,
+      "grad_norm": 0.8764932751655579,
+      "learning_rate": 1.4467350242849431e-05,
+      "loss": 3.143,
+      "step": 90550
+    },
+    {
+      "epoch": 9.768194070080863,
+      "grad_norm": 0.8792263269424438,
+      "learning_rate": 1.4143550998381002e-05,
+      "loss": 3.1218,
+      "step": 90600
+    },
+    {
+      "epoch": 9.773584905660378,
+      "grad_norm": 0.8678767681121826,
+      "learning_rate": 1.3819751753912573e-05,
+      "loss": 3.1383,
+      "step": 90650
+    },
+    {
+      "epoch": 9.778975741239892,
+      "grad_norm": 0.8487959504127502,
+      "learning_rate": 1.3495952509444142e-05,
+      "loss": 3.1216,
+      "step": 90700
+    },
+    {
+      "epoch": 9.784366576819407,
+      "grad_norm": 0.8672841191291809,
+      "learning_rate": 1.3172153264975715e-05,
+      "loss": 3.1294,
+      "step": 90750
+    },
+    {
+      "epoch": 9.789757412398922,
+      "grad_norm": 0.863354504108429,
+      "learning_rate": 1.2848354020507286e-05,
+      "loss": 3.132,
+      "step": 90800
+    },
+    {
+      "epoch": 9.795148247978437,
+      "grad_norm": 0.844878077507019,
+      "learning_rate": 1.2524554776038855e-05,
+      "loss": 3.1326,
+      "step": 90850
+    },
+    {
+      "epoch": 9.800539083557952,
+      "grad_norm": 0.8593101501464844,
+      "learning_rate": 1.2200755531570426e-05,
+      "loss": 3.1452,
+      "step": 90900
+    },
+    {
+      "epoch": 9.805929919137466,
+      "grad_norm": 0.9137566089630127,
+      "learning_rate": 1.1876956287101995e-05,
+      "loss": 3.1366,
+      "step": 90950
+    },
+    {
+      "epoch": 9.81132075471698,
+      "grad_norm": 0.888411819934845,
+      "learning_rate": 1.1553157042633566e-05,
+      "loss": 3.1275,
+      "step": 91000
+    },
+    {
+      "epoch": 9.81132075471698,
+      "eval_accuracy": 0.3942931809121649,
+      "eval_loss": 3.3018739223480225,
+      "eval_runtime": 185.5913,
+      "eval_samples_per_second": 97.047,
+      "eval_steps_per_second": 6.067,
+      "step": 91000
+    },
+    {
+      "epoch": 9.816711590296496,
+      "grad_norm": 0.8553943634033203,
+      "learning_rate": 1.1229357798165137e-05,
+      "loss": 3.1367,
+      "step": 91050
+    },
+    {
+      "epoch": 9.822102425876011,
+      "grad_norm": 0.8534170985221863,
+      "learning_rate": 1.0905558553696706e-05,
+      "loss": 3.1189,
+      "step": 91100
+    },
+    {
+      "epoch": 9.827493261455526,
+      "grad_norm": 0.8530728816986084,
+      "learning_rate": 1.0581759309228278e-05,
+      "loss": 3.1507,
+      "step": 91150
+    },
+    {
+      "epoch": 9.832884097035041,
+      "grad_norm": 0.8968914747238159,
+      "learning_rate": 1.025796006475985e-05,
+      "loss": 3.1262,
+      "step": 91200
+    },
+    {
+      "epoch": 9.838274932614555,
+      "grad_norm": 0.8507218360900879,
+      "learning_rate": 9.940636805180786e-06,
+      "loss": 3.1347,
+      "step": 91250
+    },
+    {
+      "epoch": 9.84366576819407,
+      "grad_norm": 0.841758131980896,
+      "learning_rate": 9.616837560712357e-06,
+      "loss": 3.1349,
+      "step": 91300
+    },
+    {
+      "epoch": 9.849056603773585,
+      "grad_norm": 0.8487846255302429,
+      "learning_rate": 9.293038316243928e-06,
+      "loss": 3.1433,
+      "step": 91350
+    },
+    {
+      "epoch": 9.8544474393531,
+      "grad_norm": 0.8784653544425964,
+      "learning_rate": 8.969239071775497e-06,
+      "loss": 3.1328,
+      "step": 91400
+    },
+    {
+      "epoch": 9.859838274932615,
+      "grad_norm": 0.8941159248352051,
+      "learning_rate": 8.64543982730707e-06,
+      "loss": 3.125,
+      "step": 91450
+    },
+    {
+      "epoch": 9.865229110512129,
+      "grad_norm": 0.8612044453620911,
+      "learning_rate": 8.321640582838639e-06,
+      "loss": 3.1277,
+      "step": 91500
+    },
+    {
+      "epoch": 9.870619946091644,
+      "grad_norm": 0.8660699129104614,
+      "learning_rate": 7.99784133837021e-06,
+      "loss": 3.1257,
+      "step": 91550
+    },
+    {
+      "epoch": 9.876010781671159,
+      "grad_norm": 0.8946954607963562,
+      "learning_rate": 7.674042093901779e-06,
+      "loss": 3.121,
+      "step": 91600
+    },
+    {
+      "epoch": 9.881401617250674,
+      "grad_norm": 0.8860076665878296,
+      "learning_rate": 7.350242849433351e-06,
+      "loss": 3.1299,
+      "step": 91650
+    },
+    {
+      "epoch": 9.88679245283019,
+      "grad_norm": 0.9342142343521118,
+      "learning_rate": 7.026443604964921e-06,
+      "loss": 3.1522,
+      "step": 91700
+    },
+    {
+      "epoch": 9.892183288409704,
+      "grad_norm": 0.8851354718208313,
+      "learning_rate": 6.702644360496492e-06,
+      "loss": 3.1266,
+      "step": 91750
+    },
+    {
+      "epoch": 9.897574123989218,
+      "grad_norm": 0.8632917404174805,
+      "learning_rate": 6.378845116028062e-06,
+      "loss": 3.1441,
+      "step": 91800
+    },
+    {
+      "epoch": 9.902964959568733,
+      "grad_norm": 0.8434616327285767,
+      "learning_rate": 6.055045871559633e-06,
+      "loss": 3.1195,
+      "step": 91850
+    },
+    {
+      "epoch": 9.908355795148248,
+      "grad_norm": 0.8468968272209167,
+      "learning_rate": 5.731246627091203e-06,
+      "loss": 3.1175,
+      "step": 91900
+    },
+    {
+      "epoch": 9.913746630727763,
+      "grad_norm": 0.8902624249458313,
+      "learning_rate": 5.407447382622774e-06,
+      "loss": 3.1263,
+      "step": 91950
+    },
+    {
+      "epoch": 9.919137466307278,
+      "grad_norm": 0.8296983242034912,
+      "learning_rate": 5.083648138154344e-06,
+      "loss": 3.122,
+      "step": 92000
+    },
+    {
+      "epoch": 9.919137466307278,
+      "eval_accuracy": 0.39445637754966223,
+      "eval_loss": 3.2999660968780518,
+      "eval_runtime": 185.5577,
+      "eval_samples_per_second": 97.064,
+      "eval_steps_per_second": 6.068,
+      "step": 92000
+    },
+    {
+      "epoch": 9.924528301886792,
+      "grad_norm": 0.8883585333824158,
+      "learning_rate": 4.7598488936859145e-06,
+      "loss": 3.1098,
+      "step": 92050
+    },
+    {
+      "epoch": 9.929919137466307,
+      "grad_norm": 0.8840335607528687,
+      "learning_rate": 4.4360496492174846e-06,
+      "loss": 3.1318,
+      "step": 92100
+    },
+    {
+      "epoch": 9.935309973045822,
+      "grad_norm": 0.9213301539421082,
+      "learning_rate": 4.112250404749055e-06,
+      "loss": 3.1132,
+      "step": 92150
+    },
+    {
+      "epoch": 9.940700808625337,
+      "grad_norm": 0.9292727708816528,
+      "learning_rate": 3.788451160280626e-06,
+      "loss": 3.1438,
+      "step": 92200
+    },
+    {
+      "epoch": 9.946091644204852,
+      "grad_norm": 0.9014574289321899,
+      "learning_rate": 3.464651915812196e-06,
+      "loss": 3.1289,
+      "step": 92250
+    },
+    {
+      "epoch": 9.951482479784367,
+      "grad_norm": 0.880344569683075,
+      "learning_rate": 3.1408526713437664e-06,
+      "loss": 3.1372,
+      "step": 92300
+    },
+    {
+      "epoch": 9.95687331536388,
+      "grad_norm": 0.8632800579071045,
+      "learning_rate": 2.817053426875337e-06,
+      "loss": 3.129,
+      "step": 92350
+    },
+    {
+      "epoch": 9.962264150943396,
+      "grad_norm": 0.9179752469062805,
+      "learning_rate": 2.4932541824069074e-06,
+      "loss": 3.1285,
+      "step": 92400
+    },
+    {
+      "epoch": 9.967654986522911,
+      "grad_norm": 0.8777956962585449,
+      "learning_rate": 2.169454937938478e-06,
+      "loss": 3.1195,
+      "step": 92450
+    },
+    {
+      "epoch": 9.973045822102426,
+      "grad_norm": 0.8777942061424255,
+      "learning_rate": 1.8456556934700485e-06,
+      "loss": 3.1261,
+      "step": 92500
+    },
+    {
+      "epoch": 9.978436657681941,
+      "grad_norm": 0.919086754322052,
+      "learning_rate": 1.521856449001619e-06,
+      "loss": 3.1084,
+      "step": 92550
+    },
+    {
+      "epoch": 9.983827493261456,
+      "grad_norm": 0.884113073348999,
+      "learning_rate": 1.1980572045331894e-06,
+      "loss": 3.1219,
+      "step": 92600
+    },
+    {
+      "epoch": 9.98921832884097,
+      "grad_norm": 0.8591415882110596,
+      "learning_rate": 8.742579600647598e-07,
+      "loss": 3.1279,
+      "step": 92650
+    },
+    {
+      "epoch": 9.994609164420485,
+      "grad_norm": 0.8975785970687866,
+      "learning_rate": 5.504587155963303e-07,
+      "loss": 3.1181,
+      "step": 92700
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 1.847061276435852,
+      "learning_rate": 2.2665947112790068e-07,
+      "loss": 3.117,
+      "step": 92750
+    }
+  ],
+  "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": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.75449427968e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}