diff --git "a/checkpoint-80000/trainer_state.json" "b/checkpoint-80000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-80000/trainer_state.json"
@@ -0,0 +1,11953 @@
+{
+  "best_metric": 3.3219223022460938,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M__8397/checkpoint-80000",
+  "epoch": 8.625336927223719,
+  "eval_steps": 1000,
+  "global_step": 80000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005390835579514825,
+      "grad_norm": 1.9401150941848755,
+      "learning_rate": 0.0003,
+      "loss": 8.6351,
+      "step": 50
+    },
+    {
+      "epoch": 0.01078167115902965,
+      "grad_norm": 1.2623108625411987,
+      "learning_rate": 0.0006,
+      "loss": 6.8864,
+      "step": 100
+    },
+    {
+      "epoch": 0.016172506738544475,
+      "grad_norm": 3.276719808578491,
+      "learning_rate": 0.0005996762007555315,
+      "loss": 6.4636,
+      "step": 150
+    },
+    {
+      "epoch": 0.0215633423180593,
+      "grad_norm": 2.706618309020996,
+      "learning_rate": 0.000599352401511063,
+      "loss": 6.2398,
+      "step": 200
+    },
+    {
+      "epoch": 0.026954177897574125,
+      "grad_norm": 1.0366517305374146,
+      "learning_rate": 0.0005990286022665946,
+      "loss": 6.0879,
+      "step": 250
+    },
+    {
+      "epoch": 0.03234501347708895,
+      "grad_norm": 2.069812297821045,
+      "learning_rate": 0.0005987048030221263,
+      "loss": 5.9796,
+      "step": 300
+    },
+    {
+      "epoch": 0.03773584905660377,
+      "grad_norm": 1.1950629949569702,
+      "learning_rate": 0.0005983810037776578,
+      "loss": 5.8688,
+      "step": 350
+    },
+    {
+      "epoch": 0.0431266846361186,
+      "grad_norm": 1.0007357597351074,
+      "learning_rate": 0.0005980572045331894,
+      "loss": 5.8233,
+      "step": 400
+    },
+    {
+      "epoch": 0.04851752021563342,
+      "grad_norm": 2.0554916858673096,
+      "learning_rate": 0.0005977334052887209,
+      "loss": 5.7452,
+      "step": 450
+    },
+    {
+      "epoch": 0.05390835579514825,
+      "grad_norm": 1.996433973312378,
+      "learning_rate": 0.0005974096060442526,
+      "loss": 5.6484,
+      "step": 500
+    },
+    {
+      "epoch": 0.05929919137466307,
+      "grad_norm": 1.439592719078064,
+      "learning_rate": 0.0005970858067997841,
+      "loss": 5.5734,
+      "step": 550
+    },
+    {
+      "epoch": 0.0646900269541779,
+      "grad_norm": 1.0077263116836548,
+      "learning_rate": 0.0005967620075553157,
+      "loss": 5.4903,
+      "step": 600
+    },
+    {
+      "epoch": 0.07008086253369272,
+      "grad_norm": 1.460601568222046,
+      "learning_rate": 0.0005964382083108472,
+      "loss": 5.4273,
+      "step": 650
+    },
+    {
+      "epoch": 0.07547169811320754,
+      "grad_norm": 1.5479509830474854,
+      "learning_rate": 0.0005961144090663788,
+      "loss": 5.3803,
+      "step": 700
+    },
+    {
+      "epoch": 0.08086253369272237,
+      "grad_norm": 1.2044662237167358,
+      "learning_rate": 0.0005957906098219104,
+      "loss": 5.305,
+      "step": 750
+    },
+    {
+      "epoch": 0.0862533692722372,
+      "grad_norm": 1.4456267356872559,
+      "learning_rate": 0.0005954668105774419,
+      "loss": 5.2577,
+      "step": 800
+    },
+    {
+      "epoch": 0.09164420485175202,
+      "grad_norm": 1.4306107759475708,
+      "learning_rate": 0.0005951430113329735,
+      "loss": 5.2168,
+      "step": 850
+    },
+    {
+      "epoch": 0.09703504043126684,
+      "grad_norm": 1.5723680257797241,
+      "learning_rate": 0.0005948192120885051,
+      "loss": 5.1595,
+      "step": 900
+    },
+    {
+      "epoch": 0.10242587601078167,
+      "grad_norm": 1.4030689001083374,
+      "learning_rate": 0.0005944954128440366,
+      "loss": 5.1322,
+      "step": 950
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "grad_norm": 0.9765119552612305,
+      "learning_rate": 0.0005941716135995682,
+      "loss": 5.0934,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "eval_accuracy": 0.22754229014862737,
+      "eval_loss": 5.022522926330566,
+      "eval_runtime": 185.9242,
+      "eval_samples_per_second": 96.873,
+      "eval_steps_per_second": 6.056,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11320754716981132,
+      "grad_norm": 0.8278682827949524,
+      "learning_rate": 0.0005938478143550997,
+      "loss": 5.0442,
+      "step": 1050
+    },
+    {
+      "epoch": 0.11859838274932614,
+      "grad_norm": 1.0204647779464722,
+      "learning_rate": 0.0005935240151106314,
+      "loss": 5.0209,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12398921832884097,
+      "grad_norm": 1.004250407218933,
+      "learning_rate": 0.0005932002158661629,
+      "loss": 4.9915,
+      "step": 1150
+    },
+    {
+      "epoch": 0.1293800539083558,
+      "grad_norm": 1.2949087619781494,
+      "learning_rate": 0.0005928764166216945,
+      "loss": 4.9493,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1347708894878706,
+      "grad_norm": 0.8809250593185425,
+      "learning_rate": 0.000592552617377226,
+      "loss": 4.9039,
+      "step": 1250
+    },
+    {
+      "epoch": 0.14016172506738545,
+      "grad_norm": 1.0270975828170776,
+      "learning_rate": 0.0005922288181327577,
+      "loss": 4.8792,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14555256064690028,
+      "grad_norm": 0.9649767279624939,
+      "learning_rate": 0.0005919050188882893,
+      "loss": 4.8622,
+      "step": 1350
+    },
+    {
+      "epoch": 0.1509433962264151,
+      "grad_norm": 0.9902446866035461,
+      "learning_rate": 0.0005915812196438207,
+      "loss": 4.8481,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15633423180592992,
+      "grad_norm": 0.7882956862449646,
+      "learning_rate": 0.0005912574203993524,
+      "loss": 4.7968,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16172506738544473,
+      "grad_norm": 0.8941258788108826,
+      "learning_rate": 0.0005909336211548839,
+      "loss": 4.8149,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16711590296495957,
+      "grad_norm": 1.3297462463378906,
+      "learning_rate": 0.0005906098219104155,
+      "loss": 4.7656,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1725067385444744,
+      "grad_norm": 1.497783899307251,
+      "learning_rate": 0.000590286022665947,
+      "loss": 4.7642,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1778975741239892,
+      "grad_norm": 1.0149500370025635,
+      "learning_rate": 0.0005899622234214787,
+      "loss": 4.7241,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18328840970350405,
+      "grad_norm": 1.067275047302246,
+      "learning_rate": 0.0005896384241770102,
+      "loss": 4.7115,
+      "step": 1700
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 0.7510073781013489,
+      "learning_rate": 0.0005893146249325418,
+      "loss": 4.6734,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1940700808625337,
+      "grad_norm": 1.0857677459716797,
+      "learning_rate": 0.0005889908256880733,
+      "loss": 4.6802,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19946091644204852,
+      "grad_norm": 0.9117745161056519,
+      "learning_rate": 0.0005886670264436049,
+      "loss": 4.6481,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20485175202156333,
+      "grad_norm": 0.9455132484436035,
+      "learning_rate": 0.0005883432271991365,
+      "loss": 4.6385,
+      "step": 1900
+    },
+    {
+      "epoch": 0.21024258760107817,
+      "grad_norm": 0.8344852924346924,
+      "learning_rate": 0.0005880194279546681,
+      "loss": 4.6113,
+      "step": 1950
+    },
+    {
+      "epoch": 0.215633423180593,
+      "grad_norm": 1.0357882976531982,
+      "learning_rate": 0.0005876956287101996,
+      "loss": 4.5937,
+      "step": 2000
+    },
+    {
+      "epoch": 0.215633423180593,
+      "eval_accuracy": 0.27015801715433524,
+      "eval_loss": 4.516484260559082,
+      "eval_runtime": 185.6539,
+      "eval_samples_per_second": 97.014,
+      "eval_steps_per_second": 6.065,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2210242587601078,
+      "grad_norm": 0.6943921446800232,
+      "learning_rate": 0.0005873718294657312,
+      "loss": 4.5747,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22641509433962265,
+      "grad_norm": 0.8243718147277832,
+      "learning_rate": 0.0005870480302212628,
+      "loss": 4.5582,
+      "step": 2100
+    },
+    {
+      "epoch": 0.23180592991913745,
+      "grad_norm": 1.0496394634246826,
+      "learning_rate": 0.0005867242309767943,
+      "loss": 4.5315,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2371967654986523,
+      "grad_norm": 0.8335784077644348,
+      "learning_rate": 0.0005864004317323259,
+      "loss": 4.5086,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24258760107816713,
+      "grad_norm": 0.8698639273643494,
+      "learning_rate": 0.0005860766324878575,
+      "loss": 4.4958,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24797843665768193,
+      "grad_norm": 0.7829259634017944,
+      "learning_rate": 0.000585752833243389,
+      "loss": 4.5067,
+      "step": 2300
+    },
+    {
+      "epoch": 0.25336927223719674,
+      "grad_norm": 0.8259062767028809,
+      "learning_rate": 0.0005854290339989206,
+      "loss": 4.4569,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2587601078167116,
+      "grad_norm": 0.8445196151733398,
+      "learning_rate": 0.0005851052347544521,
+      "loss": 4.4611,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2641509433962264,
+      "grad_norm": 0.8523595929145813,
+      "learning_rate": 0.0005847814355099838,
+      "loss": 4.4402,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2695417789757412,
+      "grad_norm": 0.8830829858779907,
+      "learning_rate": 0.0005844576362655154,
+      "loss": 4.4235,
+      "step": 2500
+    },
+    {
+      "epoch": 0.2749326145552561,
+      "grad_norm": 0.8365247845649719,
+      "learning_rate": 0.0005841338370210469,
+      "loss": 4.4101,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2803234501347709,
+      "grad_norm": 0.853497326374054,
+      "learning_rate": 0.0005838100377765785,
+      "loss": 4.3826,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 0.9241225719451904,
+      "learning_rate": 0.0005834862385321101,
+      "loss": 4.3931,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29110512129380056,
+      "grad_norm": 0.8265724778175354,
+      "learning_rate": 0.0005831624392876417,
+      "loss": 4.3887,
+      "step": 2700
+    },
+    {
+      "epoch": 0.29649595687331537,
+      "grad_norm": 0.8722224831581116,
+      "learning_rate": 0.0005828386400431731,
+      "loss": 4.3563,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3018867924528302,
+      "grad_norm": 0.7775854468345642,
+      "learning_rate": 0.0005825148407987048,
+      "loss": 4.3271,
+      "step": 2800
+    },
+    {
+      "epoch": 0.30727762803234504,
+      "grad_norm": 0.7281326055526733,
+      "learning_rate": 0.0005821910415542363,
+      "loss": 4.3363,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31266846361185985,
+      "grad_norm": 0.8523993492126465,
+      "learning_rate": 0.0005818672423097679,
+      "loss": 4.3284,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31805929919137466,
+      "grad_norm": 0.8152425289154053,
+      "learning_rate": 0.0005815434430652994,
+      "loss": 4.313,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "grad_norm": 0.9258395433425903,
+      "learning_rate": 0.0005812196438208311,
+      "loss": 4.3092,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "eval_accuracy": 0.29783223363934697,
+      "eval_loss": 4.246051788330078,
+      "eval_runtime": 185.609,
+      "eval_samples_per_second": 97.037,
+      "eval_steps_per_second": 6.067,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3288409703504043,
+      "grad_norm": 0.78269362449646,
+      "learning_rate": 0.0005808958445763626,
+      "loss": 4.304,
+      "step": 3050
+    },
+    {
+      "epoch": 0.33423180592991913,
+      "grad_norm": 0.7207179069519043,
+      "learning_rate": 0.0005805720453318942,
+      "loss": 4.3041,
+      "step": 3100
+    },
+    {
+      "epoch": 0.33962264150943394,
+      "grad_norm": 0.791796088218689,
+      "learning_rate": 0.0005802482460874257,
+      "loss": 4.2633,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3450134770889488,
+      "grad_norm": 0.6603590250015259,
+      "learning_rate": 0.0005799244468429573,
+      "loss": 4.2613,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3504043126684636,
+      "grad_norm": 0.922822892665863,
+      "learning_rate": 0.0005796006475984889,
+      "loss": 4.2617,
+      "step": 3250
+    },
+    {
+      "epoch": 0.3557951482479784,
+      "grad_norm": 0.7605053782463074,
+      "learning_rate": 0.0005792768483540205,
+      "loss": 4.2446,
+      "step": 3300
+    },
+    {
+      "epoch": 0.3611859838274933,
+      "grad_norm": 0.7679175734519958,
+      "learning_rate": 0.000578953049109552,
+      "loss": 4.26,
+      "step": 3350
+    },
+    {
+      "epoch": 0.3665768194070081,
+      "grad_norm": 0.7325921654701233,
+      "learning_rate": 0.0005786292498650836,
+      "loss": 4.2337,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3719676549865229,
+      "grad_norm": 0.6152936816215515,
+      "learning_rate": 0.0005783054506206152,
+      "loss": 4.21,
+      "step": 3450
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.7148654460906982,
+      "learning_rate": 0.0005779816513761467,
+      "loss": 4.2203,
+      "step": 3500
+    },
+    {
+      "epoch": 0.38274932614555257,
+      "grad_norm": 0.8136675953865051,
+      "learning_rate": 0.0005776578521316782,
+      "loss": 4.2143,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3881401617250674,
+      "grad_norm": 0.6423671245574951,
+      "learning_rate": 0.0005773340528872099,
+      "loss": 4.2164,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3935309973045822,
+      "grad_norm": 0.8761276602745056,
+      "learning_rate": 0.0005770102536427414,
+      "loss": 4.1887,
+      "step": 3650
+    },
+    {
+      "epoch": 0.39892183288409705,
+      "grad_norm": 0.7816224694252014,
+      "learning_rate": 0.000576686454398273,
+      "loss": 4.1937,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40431266846361186,
+      "grad_norm": 0.7834933400154114,
+      "learning_rate": 0.0005763626551538045,
+      "loss": 4.1965,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40970350404312667,
+      "grad_norm": 0.6208221912384033,
+      "learning_rate": 0.0005760388559093362,
+      "loss": 4.1864,
+      "step": 3800
+    },
+    {
+      "epoch": 0.41509433962264153,
+      "grad_norm": 0.6872389316558838,
+      "learning_rate": 0.0005757150566648678,
+      "loss": 4.1732,
+      "step": 3850
+    },
+    {
+      "epoch": 0.42048517520215634,
+      "grad_norm": 0.7426034808158875,
+      "learning_rate": 0.0005753912574203993,
+      "loss": 4.1682,
+      "step": 3900
+    },
+    {
+      "epoch": 0.42587601078167114,
+      "grad_norm": 0.7201637625694275,
+      "learning_rate": 0.0005750674581759309,
+      "loss": 4.157,
+      "step": 3950
+    },
+    {
+      "epoch": 0.431266846361186,
+      "grad_norm": 0.7274760007858276,
+      "learning_rate": 0.0005747436589314624,
+      "loss": 4.1586,
+      "step": 4000
+    },
+    {
+      "epoch": 0.431266846361186,
+      "eval_accuracy": 0.3127459290207626,
+      "eval_loss": 4.088518142700195,
+      "eval_runtime": 185.6568,
+      "eval_samples_per_second": 97.012,
+      "eval_steps_per_second": 6.065,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4366576819407008,
+      "grad_norm": 0.5464039444923401,
+      "learning_rate": 0.0005744198596869941,
+      "loss": 4.149,
+      "step": 4050
+    },
+    {
+      "epoch": 0.4420485175202156,
+      "grad_norm": 0.7344082593917847,
+      "learning_rate": 0.0005740960604425255,
+      "loss": 4.1532,
+      "step": 4100
+    },
+    {
+      "epoch": 0.4474393530997305,
+      "grad_norm": 0.6581411957740784,
+      "learning_rate": 0.0005737722611980572,
+      "loss": 4.1547,
+      "step": 4150
+    },
+    {
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.6639128923416138,
+      "learning_rate": 0.0005734484619535887,
+      "loss": 4.1367,
+      "step": 4200
+    },
+    {
+      "epoch": 0.4582210242587601,
+      "grad_norm": 0.734761655330658,
+      "learning_rate": 0.0005731246627091203,
+      "loss": 4.1221,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4636118598382749,
+      "grad_norm": 0.8860589265823364,
+      "learning_rate": 0.0005728008634646518,
+      "loss": 4.1218,
+      "step": 4300
+    },
+    {
+      "epoch": 0.46900269541778977,
+      "grad_norm": 0.5995060205459595,
+      "learning_rate": 0.0005724770642201835,
+      "loss": 4.1167,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4743935309973046,
+      "grad_norm": 0.7229450345039368,
+      "learning_rate": 0.000572153264975715,
+      "loss": 4.1148,
+      "step": 4400
+    },
+    {
+      "epoch": 0.4797843665768194,
+      "grad_norm": 0.731573224067688,
+      "learning_rate": 0.0005718294657312466,
+      "loss": 4.1018,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48517520215633425,
+      "grad_norm": 0.7716799974441528,
+      "learning_rate": 0.0005715056664867781,
+      "loss": 4.109,
+      "step": 4500
+    },
+    {
+      "epoch": 0.49056603773584906,
+      "grad_norm": 0.542386531829834,
+      "learning_rate": 0.0005711818672423097,
+      "loss": 4.1017,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49595687331536387,
+      "grad_norm": 0.794573187828064,
+      "learning_rate": 0.0005708580679978413,
+      "loss": 4.086,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5013477088948787,
+      "grad_norm": 0.6180745959281921,
+      "learning_rate": 0.0005705342687533729,
+      "loss": 4.0787,
+      "step": 4650
+    },
+    {
+      "epoch": 0.5067385444743935,
+      "grad_norm": 0.6277428269386292,
+      "learning_rate": 0.0005702104695089044,
+      "loss": 4.0932,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5121293800539084,
+      "grad_norm": 0.6428430676460266,
+      "learning_rate": 0.000569886670264436,
+      "loss": 4.0812,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5175202156334232,
+      "grad_norm": 0.6423416137695312,
+      "learning_rate": 0.0005695628710199675,
+      "loss": 4.0791,
+      "step": 4800
+    },
+    {
+      "epoch": 0.522911051212938,
+      "grad_norm": 0.6095171570777893,
+      "learning_rate": 0.0005692390717754991,
+      "loss": 4.0655,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5283018867924528,
+      "grad_norm": 0.6891219615936279,
+      "learning_rate": 0.0005689152725310306,
+      "loss": 4.0527,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5336927223719676,
+      "grad_norm": 0.6701120734214783,
+      "learning_rate": 0.0005685914732865623,
+      "loss": 4.0714,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "grad_norm": 0.5252517461776733,
+      "learning_rate": 0.0005682676740420939,
+      "loss": 4.0492,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "eval_accuracy": 0.32119879773406573,
+      "eval_loss": 3.9914467334747314,
+      "eval_runtime": 185.5695,
+      "eval_samples_per_second": 97.058,
+      "eval_steps_per_second": 6.068,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5444743935309974,
+      "grad_norm": 0.6187731623649597,
+      "learning_rate": 0.0005679438747976254,
+      "loss": 4.0591,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5498652291105122,
+      "grad_norm": 0.6102878451347351,
+      "learning_rate": 0.000567620075553157,
+      "loss": 4.0513,
+      "step": 5100
+    },
+    {
+      "epoch": 0.555256064690027,
+      "grad_norm": 0.708706259727478,
+      "learning_rate": 0.0005672962763086886,
+      "loss": 4.0489,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5606469002695418,
+      "grad_norm": 0.6461382508277893,
+      "learning_rate": 0.0005669724770642202,
+      "loss": 4.0661,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.7211845517158508,
+      "learning_rate": 0.0005666486778197517,
+      "loss": 4.0418,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 0.7692673802375793,
+      "learning_rate": 0.0005663248785752833,
+      "loss": 4.0325,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5768194070080862,
+      "grad_norm": 0.5925557613372803,
+      "learning_rate": 0.0005660010793308148,
+      "loss": 4.0266,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5822102425876011,
+      "grad_norm": 0.5735204815864563,
+      "learning_rate": 0.0005656772800863465,
+      "loss": 4.0372,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5876010781671159,
+      "grad_norm": 0.5482949018478394,
+      "learning_rate": 0.0005653534808418779,
+      "loss": 4.036,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5929919137466307,
+      "grad_norm": 0.7778071165084839,
+      "learning_rate": 0.0005650296815974096,
+      "loss": 4.0203,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5983827493261455,
+      "grad_norm": 0.603145956993103,
+      "learning_rate": 0.0005647058823529411,
+      "loss": 4.009,
+      "step": 5550
+    },
+    {
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.5814369916915894,
+      "learning_rate": 0.0005643820831084727,
+      "loss": 4.0256,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6091644204851752,
+      "grad_norm": 0.6685661673545837,
+      "learning_rate": 0.0005640582838640042,
+      "loss": 4.0055,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6145552560646901,
+      "grad_norm": 0.5345515608787537,
+      "learning_rate": 0.0005637344846195358,
+      "loss": 4.0021,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6199460916442049,
+      "grad_norm": 0.6599013209342957,
+      "learning_rate": 0.0005634106853750674,
+      "loss": 3.9773,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6253369272237197,
+      "grad_norm": 0.6814241409301758,
+      "learning_rate": 0.000563086886130599,
+      "loss": 3.985,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6307277628032345,
+      "grad_norm": 0.6166688203811646,
+      "learning_rate": 0.0005627630868861305,
+      "loss": 3.9948,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6361185983827493,
+      "grad_norm": 0.6375031471252441,
+      "learning_rate": 0.0005624392876416621,
+      "loss": 3.995,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6415094339622641,
+      "grad_norm": 0.6021779179573059,
+      "learning_rate": 0.0005621154883971937,
+      "loss": 3.9838,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "grad_norm": 0.6350939273834229,
+      "learning_rate": 0.0005617916891527253,
+      "loss": 3.9962,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "eval_accuracy": 0.3278974655736041,
+      "eval_loss": 3.919321060180664,
+      "eval_runtime": 185.6272,
+      "eval_samples_per_second": 97.028,
+      "eval_steps_per_second": 6.066,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6522911051212938,
+      "grad_norm": 0.6063998937606812,
+      "learning_rate": 0.0005614678899082568,
+      "loss": 3.9623,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6576819407008087,
+      "grad_norm": 0.6236124634742737,
+      "learning_rate": 0.0005611440906637884,
+      "loss": 3.9998,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6630727762803235,
+      "grad_norm": 0.7479600310325623,
+      "learning_rate": 0.0005608267674042094,
+      "loss": 3.9792,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6684636118598383,
+      "grad_norm": 0.5813738703727722,
+      "learning_rate": 0.0005605029681597409,
+      "loss": 3.953,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6738544474393531,
+      "grad_norm": 0.5313798785209656,
+      "learning_rate": 0.0005601791689152725,
+      "loss": 3.972,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6792452830188679,
+      "grad_norm": 0.6238154768943787,
+      "learning_rate": 0.000559855369670804,
+      "loss": 3.9769,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6846361185983828,
+      "grad_norm": 0.5220109224319458,
+      "learning_rate": 0.0005595315704263357,
+      "loss": 3.9613,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6900269541778976,
+      "grad_norm": 0.673324704170227,
+      "learning_rate": 0.0005592077711818672,
+      "loss": 3.9706,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6954177897574124,
+      "grad_norm": 0.6563860177993774,
+      "learning_rate": 0.0005588839719373988,
+      "loss": 3.969,
+      "step": 6450
+    },
+    {
+      "epoch": 0.7008086253369272,
+      "grad_norm": 0.6665467023849487,
+      "learning_rate": 0.0005585601726929303,
+      "loss": 3.9606,
+      "step": 6500
+    },
+    {
+      "epoch": 0.706199460916442,
+      "grad_norm": 0.5975165367126465,
+      "learning_rate": 0.0005582363734484619,
+      "loss": 3.959,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7115902964959568,
+      "grad_norm": 0.6370920538902283,
+      "learning_rate": 0.0005579125742039935,
+      "loss": 3.9606,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7169811320754716,
+      "grad_norm": 0.6234825849533081,
+      "learning_rate": 0.0005575887749595251,
+      "loss": 3.97,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7223719676549866,
+      "grad_norm": 0.6267138123512268,
+      "learning_rate": 0.0005572649757150566,
+      "loss": 3.9344,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7277628032345014,
+      "grad_norm": 0.5611273050308228,
+      "learning_rate": 0.0005569411764705882,
+      "loss": 3.9631,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7331536388140162,
+      "grad_norm": 0.6250568628311157,
+      "learning_rate": 0.0005566173772261198,
+      "loss": 3.9344,
+      "step": 6800
+    },
+    {
+      "epoch": 0.738544474393531,
+      "grad_norm": 0.5902653932571411,
+      "learning_rate": 0.0005562935779816513,
+      "loss": 3.9345,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7439353099730458,
+      "grad_norm": 0.5684419274330139,
+      "learning_rate": 0.0005559697787371828,
+      "loss": 3.9367,
+      "step": 6900
+    },
+    {
+      "epoch": 0.7493261455525606,
+      "grad_norm": 0.5950194001197815,
+      "learning_rate": 0.0005556459794927145,
+      "loss": 3.9418,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.6283013820648193,
+      "learning_rate": 0.000555322180248246,
+      "loss": 3.9219,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "eval_accuracy": 0.33346299244612526,
+      "eval_loss": 3.8631539344787598,
+      "eval_runtime": 185.3774,
+      "eval_samples_per_second": 97.159,
+      "eval_steps_per_second": 6.074,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7601078167115903,
+      "grad_norm": 0.6419661641120911,
+      "learning_rate": 0.0005549983810037776,
+      "loss": 3.9341,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7654986522911051,
+      "grad_norm": 0.8758557438850403,
+      "learning_rate": 0.0005546745817593091,
+      "loss": 3.9093,
+      "step": 7100
+    },
+    {
+      "epoch": 0.77088948787062,
+      "grad_norm": 0.6024784445762634,
+      "learning_rate": 0.0005543507825148408,
+      "loss": 3.9347,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7762803234501348,
+      "grad_norm": 0.5663654208183289,
+      "learning_rate": 0.0005540269832703723,
+      "loss": 3.9072,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7816711590296496,
+      "grad_norm": 0.6945359110832214,
+      "learning_rate": 0.0005537031840259039,
+      "loss": 3.899,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7870619946091644,
+      "grad_norm": 0.5187634229660034,
+      "learning_rate": 0.0005533793847814354,
+      "loss": 3.9124,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7924528301886793,
+      "grad_norm": 0.6671504378318787,
+      "learning_rate": 0.000553055585536967,
+      "loss": 3.9133,
+      "step": 7350
+    },
+    {
+      "epoch": 0.7978436657681941,
+      "grad_norm": 0.5317527055740356,
+      "learning_rate": 0.0005527317862924987,
+      "loss": 3.9103,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8032345013477089,
+      "grad_norm": 0.567512571811676,
+      "learning_rate": 0.0005524079870480301,
+      "loss": 3.8879,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8086253369272237,
+      "grad_norm": 0.5598046183586121,
+      "learning_rate": 0.0005520841878035618,
+      "loss": 3.9029,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8140161725067385,
+      "grad_norm": 0.6032840013504028,
+      "learning_rate": 0.0005517603885590933,
+      "loss": 3.8964,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8194070080862533,
+      "grad_norm": 0.5689701437950134,
+      "learning_rate": 0.0005514365893146249,
+      "loss": 3.8911,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8247978436657682,
+      "grad_norm": 0.5858896374702454,
+      "learning_rate": 0.0005511127900701564,
+      "loss": 3.8974,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8301886792452831,
+      "grad_norm": 0.6379143595695496,
+      "learning_rate": 0.000550788990825688,
+      "loss": 3.8913,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8355795148247979,
+      "grad_norm": 0.6346825361251831,
+      "learning_rate": 0.0005504651915812196,
+      "loss": 3.8938,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8409703504043127,
+      "grad_norm": 0.651748776435852,
+      "learning_rate": 0.0005501413923367512,
+      "loss": 3.8768,
+      "step": 7800
+    },
+    {
+      "epoch": 0.8463611859838275,
+      "grad_norm": 0.5529057383537292,
+      "learning_rate": 0.0005498175930922827,
+      "loss": 3.8814,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8517520215633423,
+      "grad_norm": 0.5781732797622681,
+      "learning_rate": 0.0005494937938478143,
+      "loss": 3.8778,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.599770188331604,
+      "learning_rate": 0.0005491699946033459,
+      "loss": 3.9077,
+      "step": 7950
+    },
+    {
+      "epoch": 0.862533692722372,
+      "grad_norm": 0.6126531958580017,
+      "learning_rate": 0.0005488461953588775,
+      "loss": 3.8566,
+      "step": 8000
+    },
+    {
+      "epoch": 0.862533692722372,
+      "eval_accuracy": 0.3376529737589151,
+      "eval_loss": 3.815525531768799,
+      "eval_runtime": 186.958,
+      "eval_samples_per_second": 96.337,
+      "eval_steps_per_second": 6.023,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8679245283018868,
+      "grad_norm": 0.6914803981781006,
+      "learning_rate": 0.000548522396114409,
+      "loss": 3.8954,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8733153638814016,
+      "grad_norm": 0.6908184289932251,
+      "learning_rate": 0.0005481985968699406,
+      "loss": 3.8747,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8787061994609164,
+      "grad_norm": 0.5247132778167725,
+      "learning_rate": 0.0005478747976254721,
+      "loss": 3.8785,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8840970350404312,
+      "grad_norm": 0.585969090461731,
+      "learning_rate": 0.0005475509983810037,
+      "loss": 3.8734,
+      "step": 8200
+    },
+    {
+      "epoch": 0.889487870619946,
+      "grad_norm": 0.5993847250938416,
+      "learning_rate": 0.0005472271991365352,
+      "loss": 3.8685,
+      "step": 8250
+    },
+    {
+      "epoch": 0.894878706199461,
+      "grad_norm": 0.6115968227386475,
+      "learning_rate": 0.0005469033998920669,
+      "loss": 3.8867,
+      "step": 8300
+    },
+    {
+      "epoch": 0.9002695417789758,
+      "grad_norm": 0.5844521522521973,
+      "learning_rate": 0.0005465860766324878,
+      "loss": 3.8566,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.5750303864479065,
+      "learning_rate": 0.0005462622773880194,
+      "loss": 3.8639,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9110512129380054,
+      "grad_norm": 0.5799904465675354,
+      "learning_rate": 0.000545938478143551,
+      "loss": 3.859,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9164420485175202,
+      "grad_norm": 0.6240545511245728,
+      "learning_rate": 0.0005456146788990825,
+      "loss": 3.8648,
+      "step": 8500
+    },
+    {
+      "epoch": 0.921832884097035,
+      "grad_norm": 0.5578630566596985,
+      "learning_rate": 0.000545290879654614,
+      "loss": 3.8597,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9272237196765498,
+      "grad_norm": 0.5732514262199402,
+      "learning_rate": 0.0005449670804101457,
+      "loss": 3.8652,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9326145552560647,
+      "grad_norm": 0.614829957485199,
+      "learning_rate": 0.0005446432811656773,
+      "loss": 3.845,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9380053908355795,
+      "grad_norm": 0.6199792623519897,
+      "learning_rate": 0.0005443194819212088,
+      "loss": 3.8484,
+      "step": 8700
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.5770998597145081,
+      "learning_rate": 0.0005439956826767404,
+      "loss": 3.8456,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9487870619946092,
+      "grad_norm": 0.5756552815437317,
+      "learning_rate": 0.000543671883432272,
+      "loss": 3.8492,
+      "step": 8800
+    },
+    {
+      "epoch": 0.954177897574124,
+      "grad_norm": 0.5448219180107117,
+      "learning_rate": 0.0005433480841878035,
+      "loss": 3.8518,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9595687331536388,
+      "grad_norm": 0.5587062239646912,
+      "learning_rate": 0.000543024284943335,
+      "loss": 3.8389,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9649595687331537,
+      "grad_norm": 0.5609190464019775,
+      "learning_rate": 0.0005427004856988667,
+      "loss": 3.8303,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "grad_norm": 0.6230307817459106,
+      "learning_rate": 0.0005423766864543982,
+      "loss": 3.856,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "eval_accuracy": 0.34079347682830485,
+      "eval_loss": 3.7813315391540527,
+      "eval_runtime": 186.6989,
+      "eval_samples_per_second": 96.471,
+      "eval_steps_per_second": 6.031,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9757412398921833,
+      "grad_norm": 0.6406134963035583,
+      "learning_rate": 0.0005420528872099298,
+      "loss": 3.871,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9811320754716981,
+      "grad_norm": 0.49706608057022095,
+      "learning_rate": 0.0005417290879654613,
+      "loss": 3.8273,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9865229110512129,
+      "grad_norm": 0.5765336751937866,
+      "learning_rate": 0.000541405288720993,
+      "loss": 3.8418,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9919137466307277,
+      "grad_norm": 0.5545505881309509,
+      "learning_rate": 0.0005410814894765245,
+      "loss": 3.8285,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9973045822102425,
+      "grad_norm": 0.5445045232772827,
+      "learning_rate": 0.0005407576902320561,
+      "loss": 3.8485,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0026954177897573,
+      "grad_norm": 0.5650607347488403,
+      "learning_rate": 0.0005404338909875876,
+      "loss": 3.7963,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0080862533692723,
+      "grad_norm": 0.5752312541007996,
+      "learning_rate": 0.0005401100917431192,
+      "loss": 3.7605,
+      "step": 9350
+    },
+    {
+      "epoch": 1.013477088948787,
+      "grad_norm": 0.5649228096008301,
+      "learning_rate": 0.0005397862924986508,
+      "loss": 3.7564,
+      "step": 9400
+    },
+    {
+      "epoch": 1.0188679245283019,
+      "grad_norm": 0.5715786218643188,
+      "learning_rate": 0.0005394624932541824,
+      "loss": 3.7554,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0242587601078168,
+      "grad_norm": 0.5207133889198303,
+      "learning_rate": 0.0005391386940097139,
+      "loss": 3.7592,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0296495956873315,
+      "grad_norm": 0.5603737235069275,
+      "learning_rate": 0.0005388148947652455,
+      "loss": 3.7704,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0350404312668464,
+      "grad_norm": 0.6097581386566162,
+      "learning_rate": 0.000538491095520777,
+      "loss": 3.7686,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0404312668463611,
+      "grad_norm": 0.6030071377754211,
+      "learning_rate": 0.0005381672962763086,
+      "loss": 3.7726,
+      "step": 9650
+    },
+    {
+      "epoch": 1.045822102425876,
+      "grad_norm": 0.546688437461853,
+      "learning_rate": 0.0005378434970318403,
+      "loss": 3.7567,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0512129380053907,
+      "grad_norm": 0.5595793724060059,
+      "learning_rate": 0.0005375196977873718,
+      "loss": 3.755,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0566037735849056,
+      "grad_norm": 0.5992977619171143,
+      "learning_rate": 0.0005371958985429034,
+      "loss": 3.7821,
+      "step": 9800
+    },
+    {
+      "epoch": 1.0619946091644206,
+      "grad_norm": 0.5221476554870605,
+      "learning_rate": 0.0005368720992984349,
+      "loss": 3.7845,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0673854447439353,
+      "grad_norm": 0.5831781029701233,
+      "learning_rate": 0.0005365483000539665,
+      "loss": 3.772,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0727762803234502,
+      "grad_norm": 0.5980172753334045,
+      "learning_rate": 0.0005362245008094981,
+      "loss": 3.7854,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "grad_norm": 0.5590375661849976,
+      "learning_rate": 0.0005359007015650297,
+      "loss": 3.7649,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "eval_accuracy": 0.34435424926776104,
+      "eval_loss": 3.7518138885498047,
+      "eval_runtime": 186.0381,
+      "eval_samples_per_second": 96.814,
+      "eval_steps_per_second": 6.053,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0835579514824798,
+      "grad_norm": 0.5709946751594543,
+      "learning_rate": 0.0005355769023205612,
+      "loss": 3.7581,
+      "step": 10050
+    },
+    {
+      "epoch": 1.0889487870619945,
+      "grad_norm": 0.618277370929718,
+      "learning_rate": 0.0005352531030760928,
+      "loss": 3.7592,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0943396226415094,
+      "grad_norm": 0.6374980211257935,
+      "learning_rate": 0.0005349293038316244,
+      "loss": 3.7555,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0997304582210243,
+      "grad_norm": 0.5958013534545898,
+      "learning_rate": 0.0005346055045871559,
+      "loss": 3.7666,
+      "step": 10200
+    },
+    {
+      "epoch": 1.105121293800539,
+      "grad_norm": 0.6115665435791016,
+      "learning_rate": 0.0005342817053426874,
+      "loss": 3.7524,
+      "step": 10250
+    },
+    {
+      "epoch": 1.110512129380054,
+      "grad_norm": 0.585763692855835,
+      "learning_rate": 0.0005339579060982191,
+      "loss": 3.764,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1159029649595686,
+      "grad_norm": 0.653211772441864,
+      "learning_rate": 0.0005336341068537506,
+      "loss": 3.7568,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1212938005390836,
+      "grad_norm": 0.5737949013710022,
+      "learning_rate": 0.0005333103076092822,
+      "loss": 3.7638,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1266846361185983,
+      "grad_norm": 0.5587084293365479,
+      "learning_rate": 0.0005329865083648137,
+      "loss": 3.7609,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1320754716981132,
+      "grad_norm": 0.6286759972572327,
+      "learning_rate": 0.0005326627091203454,
+      "loss": 3.7635,
+      "step": 10500
+    },
+    {
+      "epoch": 1.137466307277628,
+      "grad_norm": 0.5914649367332458,
+      "learning_rate": 0.0005323453858607662,
+      "loss": 3.7659,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.574865460395813,
+      "learning_rate": 0.0005320215866162979,
+      "loss": 3.7552,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1482479784366577,
+      "grad_norm": 0.570719838142395,
+      "learning_rate": 0.0005316977873718294,
+      "loss": 3.7441,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1536388140161726,
+      "grad_norm": 0.5437813997268677,
+      "learning_rate": 0.000531373988127361,
+      "loss": 3.7548,
+      "step": 10700
+    },
+    {
+      "epoch": 1.1590296495956873,
+      "grad_norm": 0.575222909450531,
+      "learning_rate": 0.0005310501888828925,
+      "loss": 3.7359,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1644204851752022,
+      "grad_norm": 0.5911815762519836,
+      "learning_rate": 0.0005307263896384242,
+      "loss": 3.7398,
+      "step": 10800
+    },
+    {
+      "epoch": 1.169811320754717,
+      "grad_norm": 0.5724892020225525,
+      "learning_rate": 0.0005304025903939556,
+      "loss": 3.7319,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1752021563342319,
+      "grad_norm": 0.6011896729469299,
+      "learning_rate": 0.0005300787911494873,
+      "loss": 3.7465,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1805929919137466,
+      "grad_norm": 0.5386224985122681,
+      "learning_rate": 0.0005297549919050189,
+      "loss": 3.7593,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "grad_norm": 0.5424002408981323,
+      "learning_rate": 0.0005294311926605504,
+      "loss": 3.769,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "eval_accuracy": 0.3473514391781235,
+      "eval_loss": 3.723193883895874,
+      "eval_runtime": 186.4484,
+      "eval_samples_per_second": 96.6,
+      "eval_steps_per_second": 6.039,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1913746630727764,
+      "grad_norm": 0.5702701210975647,
+      "learning_rate": 0.000529107393416082,
+      "loss": 3.7567,
+      "step": 11050
+    },
+    {
+      "epoch": 1.196765498652291,
+      "grad_norm": 0.5534523129463196,
+      "learning_rate": 0.0005287835941716135,
+      "loss": 3.7412,
+      "step": 11100
+    },
+    {
+      "epoch": 1.202156334231806,
+      "grad_norm": 0.5500940084457397,
+      "learning_rate": 0.0005284597949271452,
+      "loss": 3.7407,
+      "step": 11150
+    },
+    {
+      "epoch": 1.2075471698113207,
+      "grad_norm": 0.5379878878593445,
+      "learning_rate": 0.0005281359956826767,
+      "loss": 3.7631,
+      "step": 11200
+    },
+    {
+      "epoch": 1.2129380053908356,
+      "grad_norm": 0.5732437372207642,
+      "learning_rate": 0.0005278121964382083,
+      "loss": 3.7498,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2183288409703503,
+      "grad_norm": 0.5423194766044617,
+      "learning_rate": 0.0005274883971937398,
+      "loss": 3.7371,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2237196765498652,
+      "grad_norm": 0.542545735836029,
+      "learning_rate": 0.0005271645979492714,
+      "loss": 3.7301,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2291105121293802,
+      "grad_norm": 0.8030261397361755,
+      "learning_rate": 0.0005268472746896923,
+      "loss": 3.7452,
+      "step": 11400
+    },
+    {
+      "epoch": 1.2345013477088949,
+      "grad_norm": 0.5232232809066772,
+      "learning_rate": 0.000526523475445224,
+      "loss": 3.755,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2398921832884098,
+      "grad_norm": 0.5808193683624268,
+      "learning_rate": 0.0005261996762007554,
+      "loss": 3.7452,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2452830188679245,
+      "grad_norm": 0.5781920552253723,
+      "learning_rate": 0.0005258758769562871,
+      "loss": 3.7392,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2506738544474394,
+      "grad_norm": 0.5637895464897156,
+      "learning_rate": 0.0005255520777118186,
+      "loss": 3.7512,
+      "step": 11600
+    },
+    {
+      "epoch": 1.256064690026954,
+      "grad_norm": 0.5855022668838501,
+      "learning_rate": 0.0005252282784673502,
+      "loss": 3.7318,
+      "step": 11650
+    },
+    {
+      "epoch": 1.261455525606469,
+      "grad_norm": 0.5060122013092041,
+      "learning_rate": 0.0005249044792228817,
+      "loss": 3.7424,
+      "step": 11700
+    },
+    {
+      "epoch": 1.266846361185984,
+      "grad_norm": 0.5578184723854065,
+      "learning_rate": 0.0005245806799784133,
+      "loss": 3.7421,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2722371967654986,
+      "grad_norm": 0.5692754983901978,
+      "learning_rate": 0.0005242568807339449,
+      "loss": 3.7541,
+      "step": 11800
+    },
+    {
+      "epoch": 1.2776280323450135,
+      "grad_norm": 0.5293059349060059,
+      "learning_rate": 0.0005239330814894765,
+      "loss": 3.7389,
+      "step": 11850
+    },
+    {
+      "epoch": 1.2830188679245282,
+      "grad_norm": 0.5823394060134888,
+      "learning_rate": 0.000523609282245008,
+      "loss": 3.7336,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2884097035040432,
+      "grad_norm": 0.5567950010299683,
+      "learning_rate": 0.0005232854830005396,
+      "loss": 3.7176,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "grad_norm": 0.5795886516571045,
+      "learning_rate": 0.0005229616837560712,
+      "loss": 3.7288,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "eval_accuracy": 0.34931164592718456,
+      "eval_loss": 3.697058916091919,
+      "eval_runtime": 186.8168,
+      "eval_samples_per_second": 96.41,
+      "eval_steps_per_second": 6.027,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2991913746630728,
+      "grad_norm": 0.6153177618980408,
+      "learning_rate": 0.0005226378845116028,
+      "loss": 3.7453,
+      "step": 12050
+    },
+    {
+      "epoch": 1.3045822102425877,
+      "grad_norm": 0.6005005836486816,
+      "learning_rate": 0.0005223140852671344,
+      "loss": 3.7398,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3099730458221024,
+      "grad_norm": 0.5839811563491821,
+      "learning_rate": 0.0005219902860226659,
+      "loss": 3.7295,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3153638814016173,
+      "grad_norm": 0.54124516248703,
+      "learning_rate": 0.0005216664867781975,
+      "loss": 3.7451,
+      "step": 12200
+    },
+    {
+      "epoch": 1.320754716981132,
+      "grad_norm": 0.5872597098350525,
+      "learning_rate": 0.000521342687533729,
+      "loss": 3.7195,
+      "step": 12250
+    },
+    {
+      "epoch": 1.326145552560647,
+      "grad_norm": 0.6764320731163025,
+      "learning_rate": 0.0005210188882892606,
+      "loss": 3.7273,
+      "step": 12300
+    },
+    {
+      "epoch": 1.3315363881401616,
+      "grad_norm": 0.5127867460250854,
+      "learning_rate": 0.0005206950890447922,
+      "loss": 3.7354,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3369272237196765,
+      "grad_norm": 0.5766441822052002,
+      "learning_rate": 0.0005203712898003238,
+      "loss": 3.7329,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3423180592991915,
+      "grad_norm": 0.5768564343452454,
+      "learning_rate": 0.0005200474905558553,
+      "loss": 3.7196,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3477088948787062,
+      "grad_norm": 0.5573644638061523,
+      "learning_rate": 0.0005197236913113869,
+      "loss": 3.7398,
+      "step": 12500
+    },
+    {
+      "epoch": 1.353099730458221,
+      "grad_norm": 0.519527018070221,
+      "learning_rate": 0.0005193998920669184,
+      "loss": 3.7141,
+      "step": 12550
+    },
+    {
+      "epoch": 1.3584905660377358,
+      "grad_norm": 0.6151216626167297,
+      "learning_rate": 0.0005190760928224501,
+      "loss": 3.7121,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3638814016172507,
+      "grad_norm": 0.6031010746955872,
+      "learning_rate": 0.0005187522935779816,
+      "loss": 3.7035,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3692722371967654,
+      "grad_norm": 0.534813642501831,
+      "learning_rate": 0.0005184284943335132,
+      "loss": 3.7193,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3746630727762803,
+      "grad_norm": 0.5405371785163879,
+      "learning_rate": 0.0005181046950890447,
+      "loss": 3.7074,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3800539083557952,
+      "grad_norm": 0.537695050239563,
+      "learning_rate": 0.0005177808958445764,
+      "loss": 3.7067,
+      "step": 12800
+    },
+    {
+      "epoch": 1.38544474393531,
+      "grad_norm": 0.6358603239059448,
+      "learning_rate": 0.0005174570966001078,
+      "loss": 3.7174,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3908355795148248,
+      "grad_norm": 0.6034652590751648,
+      "learning_rate": 0.0005171332973556395,
+      "loss": 3.7168,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3962264150943398,
+      "grad_norm": 0.553503692150116,
+      "learning_rate": 0.000516809498111171,
+      "loss": 3.7269,
+      "step": 12950
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "grad_norm": 0.5765125155448914,
+      "learning_rate": 0.0005164856988667026,
+      "loss": 3.7086,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "eval_accuracy": 0.35153598784652257,
+      "eval_loss": 3.674765110015869,
+      "eval_runtime": 186.3343,
+      "eval_samples_per_second": 96.66,
+      "eval_steps_per_second": 6.043,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4070080862533692,
+      "grad_norm": 0.6072783470153809,
+      "learning_rate": 0.0005161618996222341,
+      "loss": 3.7063,
+      "step": 13050
+    },
+    {
+      "epoch": 1.412398921832884,
+      "grad_norm": 0.5731549859046936,
+      "learning_rate": 0.0005158381003777657,
+      "loss": 3.6997,
+      "step": 13100
+    },
+    {
+      "epoch": 1.417789757412399,
+      "grad_norm": 0.5856156945228577,
+      "learning_rate": 0.0005155143011332973,
+      "loss": 3.7057,
+      "step": 13150
+    },
+    {
+      "epoch": 1.4231805929919137,
+      "grad_norm": 0.566673755645752,
+      "learning_rate": 0.0005151905018888289,
+      "loss": 3.7061,
+      "step": 13200
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.527152419090271,
+      "learning_rate": 0.0005148667026443604,
+      "loss": 3.7152,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4339622641509435,
+      "grad_norm": 0.5248644948005676,
+      "learning_rate": 0.000514542903399892,
+      "loss": 3.6882,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4393530997304582,
+      "grad_norm": 0.5552476048469543,
+      "learning_rate": 0.0005142191041554237,
+      "loss": 3.708,
+      "step": 13350
+    },
+    {
+      "epoch": 1.444743935309973,
+      "grad_norm": 0.5420221090316772,
+      "learning_rate": 0.0005138953049109552,
+      "loss": 3.723,
+      "step": 13400
+    },
+    {
+      "epoch": 1.4501347708894878,
+      "grad_norm": 0.546251654624939,
+      "learning_rate": 0.0005135715056664868,
+      "loss": 3.7274,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4555256064690028,
+      "grad_norm": 0.656278669834137,
+      "learning_rate": 0.0005132477064220183,
+      "loss": 3.719,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4609164420485174,
+      "grad_norm": 0.5882102847099304,
+      "learning_rate": 0.0005129239071775499,
+      "loss": 3.7077,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4663072776280324,
+      "grad_norm": 0.5904348492622375,
+      "learning_rate": 0.0005126001079330814,
+      "loss": 3.7233,
+      "step": 13600
+    },
+    {
+      "epoch": 1.4716981132075473,
+      "grad_norm": 0.5616297721862793,
+      "learning_rate": 0.000512276308688613,
+      "loss": 3.7126,
+      "step": 13650
+    },
+    {
+      "epoch": 1.477088948787062,
+      "grad_norm": 0.588398277759552,
+      "learning_rate": 0.0005119525094441446,
+      "loss": 3.6976,
+      "step": 13700
+    },
+    {
+      "epoch": 1.482479784366577,
+      "grad_norm": 0.6392883062362671,
+      "learning_rate": 0.0005116287101996762,
+      "loss": 3.701,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4878706199460916,
+      "grad_norm": 0.5546556711196899,
+      "learning_rate": 0.0005113049109552077,
+      "loss": 3.7046,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4932614555256065,
+      "grad_norm": 0.5846448540687561,
+      "learning_rate": 0.0005109811117107393,
+      "loss": 3.7087,
+      "step": 13850
+    },
+    {
+      "epoch": 1.4986522911051212,
+      "grad_norm": 0.6359493732452393,
+      "learning_rate": 0.0005106573124662708,
+      "loss": 3.6907,
+      "step": 13900
+    },
+    {
+      "epoch": 1.5040431266846361,
+      "grad_norm": 0.5580483078956604,
+      "learning_rate": 0.0005103335132218025,
+      "loss": 3.688,
+      "step": 13950
+    },
+    {
+      "epoch": 1.509433962264151,
+      "grad_norm": 0.5794188976287842,
+      "learning_rate": 0.000510009713977334,
+      "loss": 3.703,
+      "step": 14000
+    },
+    {
+      "epoch": 1.509433962264151,
+      "eval_accuracy": 0.35349782438890087,
+      "eval_loss": 3.6583781242370605,
+      "eval_runtime": 185.6312,
+      "eval_samples_per_second": 97.026,
+      "eval_steps_per_second": 6.066,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5148247978436657,
+      "grad_norm": 0.5856167078018188,
+      "learning_rate": 0.0005096859147328656,
+      "loss": 3.6999,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5202156334231804,
+      "grad_norm": 0.6054800152778625,
+      "learning_rate": 0.0005093621154883971,
+      "loss": 3.6984,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5256064690026954,
+      "grad_norm": 0.5391426086425781,
+      "learning_rate": 0.0005090383162439288,
+      "loss": 3.695,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5309973045822103,
+      "grad_norm": 0.5351890921592712,
+      "learning_rate": 0.0005087145169994602,
+      "loss": 3.701,
+      "step": 14200
+    },
+    {
+      "epoch": 1.536388140161725,
+      "grad_norm": 0.5853469371795654,
+      "learning_rate": 0.0005083907177549918,
+      "loss": 3.7001,
+      "step": 14250
+    },
+    {
+      "epoch": 1.54177897574124,
+      "grad_norm": 0.5472508668899536,
+      "learning_rate": 0.0005080669185105234,
+      "loss": 3.697,
+      "step": 14300
+    },
+    {
+      "epoch": 1.5471698113207548,
+      "grad_norm": 0.5463733673095703,
+      "learning_rate": 0.000507743119266055,
+      "loss": 3.6871,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5525606469002695,
+      "grad_norm": 0.549136757850647,
+      "learning_rate": 0.0005074193200215865,
+      "loss": 3.6895,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5579514824797842,
+      "grad_norm": 0.5161502957344055,
+      "learning_rate": 0.0005070955207771181,
+      "loss": 3.693,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5633423180592994,
+      "grad_norm": 0.5291047096252441,
+      "learning_rate": 0.0005067717215326498,
+      "loss": 3.6909,
+      "step": 14500
+    },
+    {
+      "epoch": 1.568733153638814,
+      "grad_norm": 0.5388016700744629,
+      "learning_rate": 0.0005064479222881813,
+      "loss": 3.6906,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5741239892183287,
+      "grad_norm": 0.5703091621398926,
+      "learning_rate": 0.0005061241230437129,
+      "loss": 3.6751,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5795148247978437,
+      "grad_norm": 0.5753449201583862,
+      "learning_rate": 0.0005058003237992444,
+      "loss": 3.6803,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5849056603773586,
+      "grad_norm": 0.5543888807296753,
+      "learning_rate": 0.000505476524554776,
+      "loss": 3.6864,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5902964959568733,
+      "grad_norm": 0.5601723790168762,
+      "learning_rate": 0.0005051527253103076,
+      "loss": 3.6654,
+      "step": 14750
+    },
+    {
+      "epoch": 1.595687331536388,
+      "grad_norm": 0.5522553324699402,
+      "learning_rate": 0.0005048289260658392,
+      "loss": 3.6576,
+      "step": 14800
+    },
+    {
+      "epoch": 1.6010781671159031,
+      "grad_norm": 0.5866943001747131,
+      "learning_rate": 0.0005045051268213707,
+      "loss": 3.6834,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6064690026954178,
+      "grad_norm": 0.5801655650138855,
+      "learning_rate": 0.0005041813275769023,
+      "loss": 3.6865,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6118598382749325,
+      "grad_norm": 0.5756445527076721,
+      "learning_rate": 0.0005038575283324338,
+      "loss": 3.6724,
+      "step": 14950
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "grad_norm": 0.4960193932056427,
+      "learning_rate": 0.0005035337290879654,
+      "loss": 3.664,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "eval_accuracy": 0.35552974204391946,
+      "eval_loss": 3.637425184249878,
+      "eval_runtime": 185.9925,
+      "eval_samples_per_second": 96.837,
+      "eval_steps_per_second": 6.054,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6226415094339623,
+      "grad_norm": 0.5649005770683289,
+      "learning_rate": 0.000503209929843497,
+      "loss": 3.6632,
+      "step": 15050
+    },
+    {
+      "epoch": 1.628032345013477,
+      "grad_norm": 0.5567517876625061,
+      "learning_rate": 0.0005028861305990286,
+      "loss": 3.6753,
+      "step": 15100
+    },
+    {
+      "epoch": 1.633423180592992,
+      "grad_norm": 0.5907676219940186,
+      "learning_rate": 0.0005025623313545601,
+      "loss": 3.6967,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6388140161725069,
+      "grad_norm": 0.5697629451751709,
+      "learning_rate": 0.0005022385321100917,
+      "loss": 3.6818,
+      "step": 15200
+    },
+    {
+      "epoch": 1.6442048517520216,
+      "grad_norm": 0.5234145522117615,
+      "learning_rate": 0.0005019147328656232,
+      "loss": 3.6737,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6495956873315363,
+      "grad_norm": 0.5655122995376587,
+      "learning_rate": 0.0005015909336211549,
+      "loss": 3.6731,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6549865229110512,
+      "grad_norm": 0.5556879639625549,
+      "learning_rate": 0.0005012671343766864,
+      "loss": 3.6583,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6603773584905661,
+      "grad_norm": 0.5384864211082458,
+      "learning_rate": 0.000500943335132218,
+      "loss": 3.6766,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6657681940700808,
+      "grad_norm": 0.6065835952758789,
+      "learning_rate": 0.0005006195358877495,
+      "loss": 3.6646,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6711590296495957,
+      "grad_norm": 0.5641660690307617,
+      "learning_rate": 0.0005002957366432812,
+      "loss": 3.666,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6765498652291106,
+      "grad_norm": 0.5256580710411072,
+      "learning_rate": 0.0004999719373988127,
+      "loss": 3.6515,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6819407008086253,
+      "grad_norm": 0.5791158676147461,
+      "learning_rate": 0.0004996481381543442,
+      "loss": 3.6708,
+      "step": 15600
+    },
+    {
+      "epoch": 1.68733153638814,
+      "grad_norm": 0.5627703666687012,
+      "learning_rate": 0.0004993308148947651,
+      "loss": 3.6779,
+      "step": 15650
+    },
+    {
+      "epoch": 1.692722371967655,
+      "grad_norm": 0.6009765863418579,
+      "learning_rate": 0.0004990070156502968,
+      "loss": 3.6717,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6981132075471699,
+      "grad_norm": 0.6165124177932739,
+      "learning_rate": 0.0004986832164058284,
+      "loss": 3.6558,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7035040431266846,
+      "grad_norm": 0.573556661605835,
+      "learning_rate": 0.0004983594171613599,
+      "loss": 3.6659,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7088948787061995,
+      "grad_norm": 0.5521446466445923,
+      "learning_rate": 0.0004980356179168915,
+      "loss": 3.6825,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.5886697769165039,
+      "learning_rate": 0.000497711818672423,
+      "loss": 3.6819,
+      "step": 15900
+    },
+    {
+      "epoch": 1.719676549865229,
+      "grad_norm": 0.6262629628181458,
+      "learning_rate": 0.0004973880194279547,
+      "loss": 3.6689,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "grad_norm": 0.6005555391311646,
+      "learning_rate": 0.0004970642201834862,
+      "loss": 3.6639,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "eval_accuracy": 0.3572108195241851,
+      "eval_loss": 3.6206066608428955,
+      "eval_runtime": 185.4999,
+      "eval_samples_per_second": 97.094,
+      "eval_steps_per_second": 6.07,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7304582210242587,
+      "grad_norm": 0.5764271020889282,
+      "learning_rate": 0.0004967404209390178,
+      "loss": 3.6523,
+      "step": 16050
+    },
+    {
+      "epoch": 1.7358490566037736,
+      "grad_norm": 0.5741286277770996,
+      "learning_rate": 0.0004964166216945493,
+      "loss": 3.653,
+      "step": 16100
+    },
+    {
+      "epoch": 1.7412398921832883,
+      "grad_norm": 0.6296107172966003,
+      "learning_rate": 0.000496092822450081,
+      "loss": 3.6606,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7466307277628033,
+      "grad_norm": 0.6621016263961792,
+      "learning_rate": 0.0004957690232056125,
+      "loss": 3.667,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7520215633423182,
+      "grad_norm": 0.6163436770439148,
+      "learning_rate": 0.0004954452239611441,
+      "loss": 3.6514,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7574123989218329,
+      "grad_norm": 0.5336316823959351,
+      "learning_rate": 0.0004951214247166756,
+      "loss": 3.645,
+      "step": 16300
+    },
+    {
+      "epoch": 1.7628032345013476,
+      "grad_norm": 0.5538309812545776,
+      "learning_rate": 0.0004947976254722072,
+      "loss": 3.6681,
+      "step": 16350
+    },
+    {
+      "epoch": 1.7681940700808625,
+      "grad_norm": 0.5325199365615845,
+      "learning_rate": 0.0004944738262277387,
+      "loss": 3.6466,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7735849056603774,
+      "grad_norm": 0.576173722743988,
+      "learning_rate": 0.0004941500269832703,
+      "loss": 3.6569,
+      "step": 16450
+    },
+    {
+      "epoch": 1.778975741239892,
+      "grad_norm": 0.5102316737174988,
+      "learning_rate": 0.0004938262277388019,
+      "loss": 3.6425,
+      "step": 16500
+    },
+    {
+      "epoch": 1.784366576819407,
+      "grad_norm": 0.5195789933204651,
+      "learning_rate": 0.0004935024284943335,
+      "loss": 3.6499,
+      "step": 16550
+    },
+    {
+      "epoch": 1.789757412398922,
+      "grad_norm": 0.5175392031669617,
+      "learning_rate": 0.000493178629249865,
+      "loss": 3.6622,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7951482479784366,
+      "grad_norm": 0.5237419605255127,
+      "learning_rate": 0.0004928548300053966,
+      "loss": 3.6531,
+      "step": 16650
+    },
+    {
+      "epoch": 1.8005390835579513,
+      "grad_norm": 0.6199319958686829,
+      "learning_rate": 0.0004925310307609282,
+      "loss": 3.66,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8059299191374663,
+      "grad_norm": 0.5926080346107483,
+      "learning_rate": 0.0004922072315164598,
+      "loss": 3.659,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8113207547169812,
+      "grad_norm": 0.5379118919372559,
+      "learning_rate": 0.0004918834322719913,
+      "loss": 3.6662,
+      "step": 16800
+    },
+    {
+      "epoch": 1.8167115902964959,
+      "grad_norm": 0.6137577295303345,
+      "learning_rate": 0.0004915596330275229,
+      "loss": 3.6746,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8221024258760108,
+      "grad_norm": 0.5653969645500183,
+      "learning_rate": 0.0004912358337830544,
+      "loss": 3.6474,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8274932614555257,
+      "grad_norm": 0.56174635887146,
+      "learning_rate": 0.000490912034538586,
+      "loss": 3.6592,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "grad_norm": 0.5398900508880615,
+      "learning_rate": 0.0004905882352941175,
+      "loss": 3.6383,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "eval_accuracy": 0.3587294609371681,
+      "eval_loss": 3.6051440238952637,
+      "eval_runtime": 185.6742,
+      "eval_samples_per_second": 97.003,
+      "eval_steps_per_second": 6.064,
+      "step": 17000
+    },
+    {
+      "epoch": 1.838274932614555,
+      "grad_norm": 0.5739971399307251,
+      "learning_rate": 0.0004902644360496492,
+      "loss": 3.6431,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8436657681940702,
+      "grad_norm": 0.5605867505073547,
+      "learning_rate": 0.0004899406368051808,
+      "loss": 3.6362,
+      "step": 17100
+    },
+    {
+      "epoch": 1.849056603773585,
+      "grad_norm": 0.5735852718353271,
+      "learning_rate": 0.0004896168375607123,
+      "loss": 3.6416,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8544474393530996,
+      "grad_norm": 0.5647099614143372,
+      "learning_rate": 0.0004892930383162439,
+      "loss": 3.6467,
+      "step": 17200
+    },
+    {
+      "epoch": 1.8598382749326146,
+      "grad_norm": 0.5226011276245117,
+      "learning_rate": 0.0004889692390717754,
+      "loss": 3.6534,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8652291105121295,
+      "grad_norm": 0.5790987610816956,
+      "learning_rate": 0.0004886454398273071,
+      "loss": 3.6564,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8706199460916442,
+      "grad_norm": 0.5586245059967041,
+      "learning_rate": 0.0004883216405828386,
+      "loss": 3.6568,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8760107816711589,
+      "grad_norm": 0.5110759139060974,
+      "learning_rate": 0.00048799784133837017,
+      "loss": 3.6508,
+      "step": 17400
+    },
+    {
+      "epoch": 1.881401617250674,
+      "grad_norm": 0.5973682999610901,
+      "learning_rate": 0.0004876740420939017,
+      "loss": 3.6405,
+      "step": 17450
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.5328883528709412,
+      "learning_rate": 0.0004873502428494333,
+      "loss": 3.6423,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8921832884097034,
+      "grad_norm": 0.6016663312911987,
+      "learning_rate": 0.0004870264436049649,
+      "loss": 3.6395,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8975741239892183,
+      "grad_norm": 0.5924084186553955,
+      "learning_rate": 0.00048670264436049643,
+      "loss": 3.6529,
+      "step": 17600
+    },
+    {
+      "epoch": 1.9029649595687332,
+      "grad_norm": 0.5941994190216064,
+      "learning_rate": 0.00048637884511602803,
+      "loss": 3.6441,
+      "step": 17650
+    },
+    {
+      "epoch": 1.908355795148248,
+      "grad_norm": 0.5356640219688416,
+      "learning_rate": 0.0004860550458715596,
+      "loss": 3.638,
+      "step": 17700
+    },
+    {
+      "epoch": 1.9137466307277629,
+      "grad_norm": 0.6006718277931213,
+      "learning_rate": 0.0004857312466270912,
+      "loss": 3.6598,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9191374663072778,
+      "grad_norm": 0.5845404863357544,
+      "learning_rate": 0.00048540744738262274,
+      "loss": 3.6449,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9245283018867925,
+      "grad_norm": 0.5526080131530762,
+      "learning_rate": 0.00048508364813815434,
+      "loss": 3.6354,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9299191374663072,
+      "grad_norm": 0.5421946048736572,
+      "learning_rate": 0.0004847663248785753,
+      "loss": 3.6465,
+      "step": 17900
+    },
+    {
+      "epoch": 1.935309973045822,
+      "grad_norm": 0.5944849848747253,
+      "learning_rate": 0.0004844425256341068,
+      "loss": 3.6269,
+      "step": 17950
+    },
+    {
+      "epoch": 1.940700808625337,
+      "grad_norm": 0.5479409694671631,
+      "learning_rate": 0.00048411872638963834,
+      "loss": 3.6321,
+      "step": 18000
+    },
+    {
+      "epoch": 1.940700808625337,
+      "eval_accuracy": 0.35996886442846787,
+      "eval_loss": 3.5911636352539062,
+      "eval_runtime": 185.6549,
+      "eval_samples_per_second": 97.013,
+      "eval_steps_per_second": 6.065,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9460916442048517,
+      "grad_norm": 0.5909312963485718,
+      "learning_rate": 0.00048379492714516995,
+      "loss": 3.6343,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9514824797843666,
+      "grad_norm": 0.5868159532546997,
+      "learning_rate": 0.0004834711279007015,
+      "loss": 3.6358,
+      "step": 18100
+    },
+    {
+      "epoch": 1.9568733153638815,
+      "grad_norm": 0.5208358764648438,
+      "learning_rate": 0.0004831473286562331,
+      "loss": 3.6319,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9622641509433962,
+      "grad_norm": 0.5705773234367371,
+      "learning_rate": 0.00048282352941176465,
+      "loss": 3.6214,
+      "step": 18200
+    },
+    {
+      "epoch": 1.967654986522911,
+      "grad_norm": 0.5833699703216553,
+      "learning_rate": 0.00048249973016729626,
+      "loss": 3.6428,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9730458221024259,
+      "grad_norm": 0.594480574131012,
+      "learning_rate": 0.0004821759309228278,
+      "loss": 3.6519,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9784366576819408,
+      "grad_norm": 0.5162302851676941,
+      "learning_rate": 0.00048185213167835936,
+      "loss": 3.6368,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9838274932614555,
+      "grad_norm": 0.6090728640556335,
+      "learning_rate": 0.00048152833243389096,
+      "loss": 3.6373,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9892183288409704,
+      "grad_norm": 0.5224543809890747,
+      "learning_rate": 0.0004812045331894225,
+      "loss": 3.6452,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9946091644204853,
+      "grad_norm": 0.6286558508872986,
+      "learning_rate": 0.0004808807339449541,
+      "loss": 3.6383,
+      "step": 18500
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.1893088817596436,
+      "learning_rate": 0.0004805569347004856,
+      "loss": 3.6417,
+      "step": 18550
+    },
+    {
+      "epoch": 2.0053908355795147,
+      "grad_norm": 0.5484069585800171,
+      "learning_rate": 0.0004802331354560173,
+      "loss": 3.5707,
+      "step": 18600
+    },
+    {
+      "epoch": 2.01078167115903,
+      "grad_norm": 0.5843707323074341,
+      "learning_rate": 0.00047990933621154877,
+      "loss": 3.5399,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0161725067385445,
+      "grad_norm": 0.5230047106742859,
+      "learning_rate": 0.0004795855369670804,
+      "loss": 3.5512,
+      "step": 18700
+    },
+    {
+      "epoch": 2.0215633423180592,
+      "grad_norm": 0.5233069658279419,
+      "learning_rate": 0.0004792617377226119,
+      "loss": 3.5319,
+      "step": 18750
+    },
+    {
+      "epoch": 2.026954177897574,
+      "grad_norm": 0.5597031712532043,
+      "learning_rate": 0.0004789379384781435,
+      "loss": 3.5518,
+      "step": 18800
+    },
+    {
+      "epoch": 2.032345013477089,
+      "grad_norm": 0.5443568825721741,
+      "learning_rate": 0.0004786141392336751,
+      "loss": 3.5586,
+      "step": 18850
+    },
+    {
+      "epoch": 2.0377358490566038,
+      "grad_norm": 0.5472878217697144,
+      "learning_rate": 0.00047829033998920663,
+      "loss": 3.5469,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0431266846361185,
+      "grad_norm": 0.5626387000083923,
+      "learning_rate": 0.00047796654074473824,
+      "loss": 3.5408,
+      "step": 18950
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "grad_norm": 0.5888155698776245,
+      "learning_rate": 0.0004776427415002698,
+      "loss": 3.5672,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "eval_accuracy": 0.361499131700447,
+      "eval_loss": 3.5821456909179688,
+      "eval_runtime": 185.3913,
+      "eval_samples_per_second": 97.151,
+      "eval_steps_per_second": 6.074,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0539083557951483,
+      "grad_norm": 0.5674854516983032,
+      "learning_rate": 0.0004773189422558014,
+      "loss": 3.5436,
+      "step": 19050
+    },
+    {
+      "epoch": 2.059299191374663,
+      "grad_norm": 0.5584875345230103,
+      "learning_rate": 0.00047699514301133294,
+      "loss": 3.567,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0646900269541777,
+      "grad_norm": 0.5911595821380615,
+      "learning_rate": 0.00047667134376686455,
+      "loss": 3.5623,
+      "step": 19150
+    },
+    {
+      "epoch": 2.070080862533693,
+      "grad_norm": 0.5857378840446472,
+      "learning_rate": 0.0004763475445223961,
+      "loss": 3.5702,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0754716981132075,
+      "grad_norm": 0.5929583311080933,
+      "learning_rate": 0.0004760237452779276,
+      "loss": 3.5495,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0808625336927222,
+      "grad_norm": 0.5647702813148499,
+      "learning_rate": 0.0004756999460334592,
+      "loss": 3.5433,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0862533692722374,
+      "grad_norm": 0.5857425332069397,
+      "learning_rate": 0.00047537614678899075,
+      "loss": 3.547,
+      "step": 19350
+    },
+    {
+      "epoch": 2.091644204851752,
+      "grad_norm": 0.5703374743461609,
+      "learning_rate": 0.00047505234754452235,
+      "loss": 3.5461,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0970350404312668,
+      "grad_norm": 0.8500975966453552,
+      "learning_rate": 0.0004747285483000539,
+      "loss": 3.5538,
+      "step": 19450
+    },
+    {
+      "epoch": 2.1024258760107815,
+      "grad_norm": 0.6051983833312988,
+      "learning_rate": 0.0004744047490555855,
+      "loss": 3.5454,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1078167115902966,
+      "grad_norm": 0.7593019008636475,
+      "learning_rate": 0.00047408094981111706,
+      "loss": 3.5555,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1132075471698113,
+      "grad_norm": 0.5925678014755249,
+      "learning_rate": 0.000473763626551538,
+      "loss": 3.5514,
+      "step": 19600
+    },
+    {
+      "epoch": 2.118598382749326,
+      "grad_norm": 0.5761629343032837,
+      "learning_rate": 0.00047343982730706956,
+      "loss": 3.5705,
+      "step": 19650
+    },
+    {
+      "epoch": 2.123989218328841,
+      "grad_norm": 0.5896238088607788,
+      "learning_rate": 0.00047311602806260117,
+      "loss": 3.5423,
+      "step": 19700
+    },
+    {
+      "epoch": 2.129380053908356,
+      "grad_norm": 0.5506173968315125,
+      "learning_rate": 0.0004727922288181327,
+      "loss": 3.5619,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1347708894878705,
+      "grad_norm": 0.5747953653335571,
+      "learning_rate": 0.0004724684295736643,
+      "loss": 3.5549,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1401617250673857,
+      "grad_norm": 0.6055583953857422,
+      "learning_rate": 0.0004721446303291959,
+      "loss": 3.5466,
+      "step": 19850
+    },
+    {
+      "epoch": 2.1455525606469004,
+      "grad_norm": 0.5854771733283997,
+      "learning_rate": 0.0004718208310847275,
+      "loss": 3.5586,
+      "step": 19900
+    },
+    {
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.5524225234985352,
+      "learning_rate": 0.000471497031840259,
+      "loss": 3.5506,
+      "step": 19950
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "grad_norm": 0.6449847221374512,
+      "learning_rate": 0.0004711732325957905,
+      "loss": 3.5754,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "eval_accuracy": 0.3623671596211969,
+      "eval_loss": 3.573341131210327,
+      "eval_runtime": 185.7691,
+      "eval_samples_per_second": 96.954,
+      "eval_steps_per_second": 6.061,
+      "step": 20000
+    },
+    {
+      "epoch": 2.161725067385445,
+      "grad_norm": 0.5967618823051453,
+      "learning_rate": 0.00047084943335132213,
+      "loss": 3.5752,
+      "step": 20050
+    },
+    {
+      "epoch": 2.1671159029649596,
+      "grad_norm": 0.5871074795722961,
+      "learning_rate": 0.0004705256341068537,
+      "loss": 3.5762,
+      "step": 20100
+    },
+    {
+      "epoch": 2.1725067385444743,
+      "grad_norm": 0.5428000688552856,
+      "learning_rate": 0.0004702018348623853,
+      "loss": 3.5692,
+      "step": 20150
+    },
+    {
+      "epoch": 2.177897574123989,
+      "grad_norm": 0.606965959072113,
+      "learning_rate": 0.00046987803561791684,
+      "loss": 3.5672,
+      "step": 20200
+    },
+    {
+      "epoch": 2.183288409703504,
+      "grad_norm": 0.5462236404418945,
+      "learning_rate": 0.00046955423637344844,
+      "loss": 3.5669,
+      "step": 20250
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 0.5761914253234863,
+      "learning_rate": 0.00046923043712898,
+      "loss": 3.5573,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1940700808625335,
+      "grad_norm": 0.5329501628875732,
+      "learning_rate": 0.0004689066378845116,
+      "loss": 3.5705,
+      "step": 20350
+    },
+    {
+      "epoch": 2.1994609164420487,
+      "grad_norm": 0.5812281966209412,
+      "learning_rate": 0.00046858283864004315,
+      "loss": 3.5734,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2048517520215634,
+      "grad_norm": 0.5665324926376343,
+      "learning_rate": 0.0004682590393955747,
+      "loss": 3.5702,
+      "step": 20450
+    },
+    {
+      "epoch": 2.210242587601078,
+      "grad_norm": 0.5955948233604431,
+      "learning_rate": 0.0004679352401511063,
+      "loss": 3.5426,
+      "step": 20500
+    },
+    {
+      "epoch": 2.215633423180593,
+      "grad_norm": 0.5881284475326538,
+      "learning_rate": 0.0004676114409066378,
+      "loss": 3.5686,
+      "step": 20550
+    },
+    {
+      "epoch": 2.221024258760108,
+      "grad_norm": 0.6288923025131226,
+      "learning_rate": 0.00046728764166216946,
+      "loss": 3.5468,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2264150943396226,
+      "grad_norm": 0.5576397776603699,
+      "learning_rate": 0.00046696384241770095,
+      "loss": 3.5532,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2318059299191373,
+      "grad_norm": 0.5199394226074219,
+      "learning_rate": 0.00046664004317323256,
+      "loss": 3.551,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2371967654986524,
+      "grad_norm": 0.5937060713768005,
+      "learning_rate": 0.0004663162439287641,
+      "loss": 3.5577,
+      "step": 20750
+    },
+    {
+      "epoch": 2.242587601078167,
+      "grad_norm": 0.6562036871910095,
+      "learning_rate": 0.00046599244468429566,
+      "loss": 3.5528,
+      "step": 20800
+    },
+    {
+      "epoch": 2.247978436657682,
+      "grad_norm": 0.565091609954834,
+      "learning_rate": 0.00046566864543982726,
+      "loss": 3.5683,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2533692722371965,
+      "grad_norm": 0.5420801043510437,
+      "learning_rate": 0.0004653448461953588,
+      "loss": 3.5628,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2587601078167117,
+      "grad_norm": 0.5261815786361694,
+      "learning_rate": 0.0004650210469508904,
+      "loss": 3.5508,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "grad_norm": 0.5820636749267578,
+      "learning_rate": 0.00046469724770642197,
+      "loss": 3.5637,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "eval_accuracy": 0.36372227843801913,
+      "eval_loss": 3.5613205432891846,
+      "eval_runtime": 185.2619,
+      "eval_samples_per_second": 97.219,
+      "eval_steps_per_second": 6.078,
+      "step": 21000
+    },
+    {
+      "epoch": 2.269541778975741,
+      "grad_norm": 0.5688766837120056,
+      "learning_rate": 0.0004643734484619536,
+      "loss": 3.5568,
+      "step": 21050
+    },
+    {
+      "epoch": 2.274932614555256,
+      "grad_norm": 0.5932535529136658,
+      "learning_rate": 0.0004640496492174851,
+      "loss": 3.5535,
+      "step": 21100
+    },
+    {
+      "epoch": 2.280323450134771,
+      "grad_norm": 0.5635634064674377,
+      "learning_rate": 0.00046372584997301673,
+      "loss": 3.5756,
+      "step": 21150
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.6333921551704407,
+      "learning_rate": 0.0004634020507285483,
+      "loss": 3.5627,
+      "step": 21200
+    },
+    {
+      "epoch": 2.2911051212938007,
+      "grad_norm": 0.5962139964103699,
+      "learning_rate": 0.0004630782514840798,
+      "loss": 3.571,
+      "step": 21250
+    },
+    {
+      "epoch": 2.2964959568733154,
+      "grad_norm": 0.601997435092926,
+      "learning_rate": 0.0004627544522396114,
+      "loss": 3.5379,
+      "step": 21300
+    },
+    {
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.83173668384552,
+      "learning_rate": 0.00046243065299514293,
+      "loss": 3.559,
+      "step": 21350
+    },
+    {
+      "epoch": 2.3072776280323453,
+      "grad_norm": 0.550325334072113,
+      "learning_rate": 0.00046210685375067454,
+      "loss": 3.5504,
+      "step": 21400
+    },
+    {
+      "epoch": 2.31266846361186,
+      "grad_norm": 0.5884050130844116,
+      "learning_rate": 0.0004617830545062061,
+      "loss": 3.5759,
+      "step": 21450
+    },
+    {
+      "epoch": 2.3180592991913747,
+      "grad_norm": 0.5583473443984985,
+      "learning_rate": 0.0004614592552617377,
+      "loss": 3.5461,
+      "step": 21500
+    },
+    {
+      "epoch": 2.3234501347708894,
+      "grad_norm": 0.5997599363327026,
+      "learning_rate": 0.00046113545601726924,
+      "loss": 3.5626,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3288409703504045,
+      "grad_norm": 0.5541661977767944,
+      "learning_rate": 0.00046081165677280085,
+      "loss": 3.5549,
+      "step": 21600
+    },
+    {
+      "epoch": 2.334231805929919,
+      "grad_norm": 0.6619417071342468,
+      "learning_rate": 0.0004604878575283324,
+      "loss": 3.5543,
+      "step": 21650
+    },
+    {
+      "epoch": 2.339622641509434,
+      "grad_norm": 0.5892545580863953,
+      "learning_rate": 0.00046016405828386395,
+      "loss": 3.5725,
+      "step": 21700
+    },
+    {
+      "epoch": 2.3450134770889486,
+      "grad_norm": 0.5789877772331238,
+      "learning_rate": 0.00045984025903939555,
+      "loss": 3.5618,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3504043126684637,
+      "grad_norm": 0.5967502593994141,
+      "learning_rate": 0.0004595164597949271,
+      "loss": 3.5435,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3557951482479784,
+      "grad_norm": 0.5737834572792053,
+      "learning_rate": 0.0004591926605504587,
+      "loss": 3.5549,
+      "step": 21850
+    },
+    {
+      "epoch": 2.361185983827493,
+      "grad_norm": 0.5817111134529114,
+      "learning_rate": 0.0004588688613059902,
+      "loss": 3.5489,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3665768194070083,
+      "grad_norm": 0.5602347254753113,
+      "learning_rate": 0.00045854506206152186,
+      "loss": 3.5488,
+      "step": 21950
+    },
+    {
+      "epoch": 2.371967654986523,
+      "grad_norm": 0.5591195821762085,
+      "learning_rate": 0.00045822126281705336,
+      "loss": 3.5706,
+      "step": 22000
+    },
+    {
+      "epoch": 2.371967654986523,
+      "eval_accuracy": 0.3645871554250224,
+      "eval_loss": 3.5503523349761963,
+      "eval_runtime": 185.7321,
+      "eval_samples_per_second": 96.973,
+      "eval_steps_per_second": 6.062,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 0.5747423768043518,
+      "learning_rate": 0.00045789746357258497,
+      "loss": 3.5462,
+      "step": 22050
+    },
+    {
+      "epoch": 2.382749326145553,
+      "grad_norm": 0.6654125452041626,
+      "learning_rate": 0.0004575736643281165,
+      "loss": 3.5505,
+      "step": 22100
+    },
+    {
+      "epoch": 2.3881401617250675,
+      "grad_norm": 0.593645453453064,
+      "learning_rate": 0.00045724986508364807,
+      "loss": 3.5528,
+      "step": 22150
+    },
+    {
+      "epoch": 2.393530997304582,
+      "grad_norm": 0.5858222842216492,
+      "learning_rate": 0.00045692606583917967,
+      "loss": 3.5645,
+      "step": 22200
+    },
+    {
+      "epoch": 2.398921832884097,
+      "grad_norm": 0.6029301285743713,
+      "learning_rate": 0.0004566022665947112,
+      "loss": 3.5695,
+      "step": 22250
+    },
+    {
+      "epoch": 2.404312668463612,
+      "grad_norm": 0.584065854549408,
+      "learning_rate": 0.0004562784673502428,
+      "loss": 3.5708,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4097035040431267,
+      "grad_norm": 0.5794286727905273,
+      "learning_rate": 0.0004559546681057744,
+      "loss": 3.5516,
+      "step": 22350
+    },
+    {
+      "epoch": 2.4150943396226414,
+      "grad_norm": 0.6180821657180786,
+      "learning_rate": 0.000455630868861306,
+      "loss": 3.5726,
+      "step": 22400
+    },
+    {
+      "epoch": 2.420485175202156,
+      "grad_norm": 0.5881420373916626,
+      "learning_rate": 0.00045530706961683753,
+      "loss": 3.5624,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4258760107816713,
+      "grad_norm": 0.5598190426826477,
+      "learning_rate": 0.00045498327037236914,
+      "loss": 3.556,
+      "step": 22500
+    },
+    {
+      "epoch": 2.431266846361186,
+      "grad_norm": 0.5753728151321411,
+      "learning_rate": 0.0004546594711279007,
+      "loss": 3.5505,
+      "step": 22550
+    },
+    {
+      "epoch": 2.4366576819407006,
+      "grad_norm": 0.5574933886528015,
+      "learning_rate": 0.0004543356718834322,
+      "loss": 3.551,
+      "step": 22600
+    },
+    {
+      "epoch": 2.442048517520216,
+      "grad_norm": 0.5815372467041016,
+      "learning_rate": 0.0004540118726389638,
+      "loss": 3.5393,
+      "step": 22650
+    },
+    {
+      "epoch": 2.4474393530997305,
+      "grad_norm": 0.5866022706031799,
+      "learning_rate": 0.00045368807339449534,
+      "loss": 3.5423,
+      "step": 22700
+    },
+    {
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.5673452615737915,
+      "learning_rate": 0.00045336427415002694,
+      "loss": 3.547,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4582210242587603,
+      "grad_norm": 0.587451159954071,
+      "learning_rate": 0.0004530404749055585,
+      "loss": 3.5455,
+      "step": 22800
+    },
+    {
+      "epoch": 2.463611859838275,
+      "grad_norm": 0.6395877599716187,
+      "learning_rate": 0.0004527166756610901,
+      "loss": 3.5518,
+      "step": 22850
+    },
+    {
+      "epoch": 2.4690026954177897,
+      "grad_norm": 0.6238613724708557,
+      "learning_rate": 0.00045239287641662165,
+      "loss": 3.5595,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4743935309973044,
+      "grad_norm": 0.6073823571205139,
+      "learning_rate": 0.0004520690771721532,
+      "loss": 3.5612,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "grad_norm": 0.582822322845459,
+      "learning_rate": 0.0004517452779276848,
+      "loss": 3.5408,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "eval_accuracy": 0.3659634615549684,
+      "eval_loss": 3.5402395725250244,
+      "eval_runtime": 185.4548,
+      "eval_samples_per_second": 97.118,
+      "eval_steps_per_second": 6.072,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4851752021563343,
+      "grad_norm": 0.5883828401565552,
+      "learning_rate": 0.00045142147868321636,
+      "loss": 3.543,
+      "step": 23050
+    },
+    {
+      "epoch": 2.490566037735849,
+      "grad_norm": 0.5678288340568542,
+      "learning_rate": 0.00045109767943874796,
+      "loss": 3.5585,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4959568733153636,
+      "grad_norm": 0.57308030128479,
+      "learning_rate": 0.0004507738801942795,
+      "loss": 3.5585,
+      "step": 23150
+    },
+    {
+      "epoch": 2.501347708894879,
+      "grad_norm": 0.6100690364837646,
+      "learning_rate": 0.0004504500809498111,
+      "loss": 3.5512,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5067385444743935,
+      "grad_norm": 0.5841269493103027,
+      "learning_rate": 0.0004501262817053426,
+      "loss": 3.5446,
+      "step": 23250
+    },
+    {
+      "epoch": 2.512129380053908,
+      "grad_norm": 0.6322735548019409,
+      "learning_rate": 0.00044980248246087427,
+      "loss": 3.5466,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5175202156334233,
+      "grad_norm": 0.6177630424499512,
+      "learning_rate": 0.00044947868321640577,
+      "loss": 3.5425,
+      "step": 23350
+    },
+    {
+      "epoch": 2.522911051212938,
+      "grad_norm": 0.6240397691726685,
+      "learning_rate": 0.0004491548839719373,
+      "loss": 3.5322,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5283018867924527,
+      "grad_norm": 0.587017834186554,
+      "learning_rate": 0.0004488310847274689,
+      "loss": 3.5453,
+      "step": 23450
+    },
+    {
+      "epoch": 2.533692722371968,
+      "grad_norm": 0.5576516389846802,
+      "learning_rate": 0.0004485072854830005,
+      "loss": 3.5428,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5390835579514826,
+      "grad_norm": 0.5701577663421631,
+      "learning_rate": 0.0004481834862385321,
+      "loss": 3.5514,
+      "step": 23550
+    },
+    {
+      "epoch": 2.5444743935309972,
+      "grad_norm": 0.5828137397766113,
+      "learning_rate": 0.00044785968699406363,
+      "loss": 3.5517,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5498652291105124,
+      "grad_norm": 0.6262269020080566,
+      "learning_rate": 0.0004475423637344846,
+      "loss": 3.556,
+      "step": 23650
+    },
+    {
+      "epoch": 2.555256064690027,
+      "grad_norm": 0.5848754644393921,
+      "learning_rate": 0.00044721856449001613,
+      "loss": 3.5631,
+      "step": 23700
+    },
+    {
+      "epoch": 2.560646900269542,
+      "grad_norm": 0.5634021162986755,
+      "learning_rate": 0.00044689476524554774,
+      "loss": 3.5699,
+      "step": 23750
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 0.6959907412528992,
+      "learning_rate": 0.0004465709660010793,
+      "loss": 3.5458,
+      "step": 23800
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.5664647817611694,
+      "learning_rate": 0.0004462471667566109,
+      "loss": 3.5597,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5768194070080863,
+      "grad_norm": 0.5963898301124573,
+      "learning_rate": 0.00044592336751214244,
+      "loss": 3.5427,
+      "step": 23900
+    },
+    {
+      "epoch": 2.582210242587601,
+      "grad_norm": 0.5585032105445862,
+      "learning_rate": 0.00044559956826767405,
+      "loss": 3.5489,
+      "step": 23950
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "grad_norm": 0.6078963279724121,
+      "learning_rate": 0.00044527576902320554,
+      "loss": 3.532,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "eval_accuracy": 0.36688788032449404,
+      "eval_loss": 3.531247854232788,
+      "eval_runtime": 185.6875,
+      "eval_samples_per_second": 96.996,
+      "eval_steps_per_second": 6.064,
+      "step": 24000
+    },
+    {
+      "epoch": 2.592991913746631,
+      "grad_norm": 0.5676623582839966,
+      "learning_rate": 0.00044495196977873715,
+      "loss": 3.5458,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5983827493261455,
+      "grad_norm": 0.6343262195587158,
+      "learning_rate": 0.0004446281705342687,
+      "loss": 3.5379,
+      "step": 24100
+    },
+    {
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.5848167538642883,
+      "learning_rate": 0.00044430437128980025,
+      "loss": 3.5677,
+      "step": 24150
+    },
+    {
+      "epoch": 2.6091644204851754,
+      "grad_norm": 0.6168308854103088,
+      "learning_rate": 0.00044398057204533185,
+      "loss": 3.5298,
+      "step": 24200
+    },
+    {
+      "epoch": 2.61455525606469,
+      "grad_norm": 0.5785287022590637,
+      "learning_rate": 0.0004436567728008634,
+      "loss": 3.5411,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6199460916442048,
+      "grad_norm": 0.5798066258430481,
+      "learning_rate": 0.000443332973556395,
+      "loss": 3.5491,
+      "step": 24300
+    },
+    {
+      "epoch": 2.62533692722372,
+      "grad_norm": 0.5612218379974365,
+      "learning_rate": 0.00044300917431192656,
+      "loss": 3.5649,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6307277628032346,
+      "grad_norm": 0.5887892842292786,
+      "learning_rate": 0.00044268537506745816,
+      "loss": 3.5273,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6361185983827493,
+      "grad_norm": 0.6022453308105469,
+      "learning_rate": 0.0004423615758229897,
+      "loss": 3.5447,
+      "step": 24450
+    },
+    {
+      "epoch": 2.641509433962264,
+      "grad_norm": 0.5853219032287598,
+      "learning_rate": 0.0004420377765785213,
+      "loss": 3.5591,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6469002695417787,
+      "grad_norm": 0.5891228318214417,
+      "learning_rate": 0.00044171397733405287,
+      "loss": 3.5263,
+      "step": 24550
+    },
+    {
+      "epoch": 2.652291105121294,
+      "grad_norm": 0.6075050830841064,
+      "learning_rate": 0.00044139017808958437,
+      "loss": 3.5344,
+      "step": 24600
+    },
+    {
+      "epoch": 2.6576819407008085,
+      "grad_norm": 0.5453810095787048,
+      "learning_rate": 0.00044106637884511597,
+      "loss": 3.5361,
+      "step": 24650
+    },
+    {
+      "epoch": 2.6630727762803232,
+      "grad_norm": 0.6543773412704468,
+      "learning_rate": 0.0004407425796006475,
+      "loss": 3.5407,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6684636118598384,
+      "grad_norm": 0.5871316194534302,
+      "learning_rate": 0.00044041878035617913,
+      "loss": 3.5404,
+      "step": 24750
+    },
+    {
+      "epoch": 2.673854447439353,
+      "grad_norm": 0.5868498086929321,
+      "learning_rate": 0.0004400949811117107,
+      "loss": 3.5162,
+      "step": 24800
+    },
+    {
+      "epoch": 2.6792452830188678,
+      "grad_norm": 0.5242236852645874,
+      "learning_rate": 0.0004397711818672423,
+      "loss": 3.5146,
+      "step": 24850
+    },
+    {
+      "epoch": 2.684636118598383,
+      "grad_norm": 0.5566399097442627,
+      "learning_rate": 0.00043944738262277383,
+      "loss": 3.5539,
+      "step": 24900
+    },
+    {
+      "epoch": 2.6900269541778976,
+      "grad_norm": 0.5947811603546143,
+      "learning_rate": 0.00043912358337830544,
+      "loss": 3.5379,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "grad_norm": 0.5996220111846924,
+      "learning_rate": 0.000438799784133837,
+      "loss": 3.527,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "eval_accuracy": 0.3672525194159994,
+      "eval_loss": 3.5217697620391846,
+      "eval_runtime": 185.7585,
+      "eval_samples_per_second": 96.959,
+      "eval_steps_per_second": 6.062,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7008086253369274,
+      "grad_norm": 0.5733533501625061,
+      "learning_rate": 0.00043847598488936854,
+      "loss": 3.5383,
+      "step": 25050
+    },
+    {
+      "epoch": 2.706199460916442,
+      "grad_norm": 0.5578649044036865,
+      "learning_rate": 0.00043815218564490014,
+      "loss": 3.5233,
+      "step": 25100
+    },
+    {
+      "epoch": 2.711590296495957,
+      "grad_norm": 0.5936254858970642,
+      "learning_rate": 0.0004378283864004317,
+      "loss": 3.5137,
+      "step": 25150
+    },
+    {
+      "epoch": 2.7169811320754715,
+      "grad_norm": 0.591632068157196,
+      "learning_rate": 0.0004375045871559633,
+      "loss": 3.5193,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7223719676549867,
+      "grad_norm": 0.6370254158973694,
+      "learning_rate": 0.00043718078791149485,
+      "loss": 3.5535,
+      "step": 25250
+    },
+    {
+      "epoch": 2.7277628032345014,
+      "grad_norm": 0.5756546258926392,
+      "learning_rate": 0.00043685698866702645,
+      "loss": 3.5101,
+      "step": 25300
+    },
+    {
+      "epoch": 2.733153638814016,
+      "grad_norm": 0.595561146736145,
+      "learning_rate": 0.00043653318942255795,
+      "loss": 3.5358,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7385444743935308,
+      "grad_norm": 0.6051717400550842,
+      "learning_rate": 0.00043620939017808956,
+      "loss": 3.5376,
+      "step": 25400
+    },
+    {
+      "epoch": 2.743935309973046,
+      "grad_norm": 0.5473976731300354,
+      "learning_rate": 0.0004358855909336211,
+      "loss": 3.5087,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7493261455525606,
+      "grad_norm": 0.6153759360313416,
+      "learning_rate": 0.00043556179168915266,
+      "loss": 3.5703,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.5839317440986633,
+      "learning_rate": 0.00043523799244468426,
+      "loss": 3.5587,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7601078167115904,
+      "grad_norm": 0.6133059859275818,
+      "learning_rate": 0.0004349141932002158,
+      "loss": 3.5361,
+      "step": 25600
+    },
+    {
+      "epoch": 2.765498652291105,
+      "grad_norm": 0.6802798509597778,
+      "learning_rate": 0.0004345903939557474,
+      "loss": 3.5245,
+      "step": 25650
+    },
+    {
+      "epoch": 2.77088948787062,
+      "grad_norm": 0.5973533987998962,
+      "learning_rate": 0.00043426659471127897,
+      "loss": 3.5436,
+      "step": 25700
+    },
+    {
+      "epoch": 2.776280323450135,
+      "grad_norm": 0.6152960658073425,
+      "learning_rate": 0.0004339492714516999,
+      "loss": 3.5327,
+      "step": 25750
+    },
+    {
+      "epoch": 2.7816711590296497,
+      "grad_norm": 0.5712870359420776,
+      "learning_rate": 0.00043362547220723147,
+      "loss": 3.5337,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7870619946091644,
+      "grad_norm": 0.5867605209350586,
+      "learning_rate": 0.0004333016729627631,
+      "loss": 3.5213,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7924528301886795,
+      "grad_norm": 0.6075189113616943,
+      "learning_rate": 0.0004329778737182946,
+      "loss": 3.5308,
+      "step": 25900
+    },
+    {
+      "epoch": 2.797843665768194,
+      "grad_norm": 0.5805151462554932,
+      "learning_rate": 0.00043265407447382623,
+      "loss": 3.5207,
+      "step": 25950
+    },
+    {
+      "epoch": 2.803234501347709,
+      "grad_norm": 0.5881994962692261,
+      "learning_rate": 0.0004323302752293577,
+      "loss": 3.5333,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803234501347709,
+      "eval_accuracy": 0.368511371774218,
+      "eval_loss": 3.51200795173645,
+      "eval_runtime": 185.141,
+      "eval_samples_per_second": 97.283,
+      "eval_steps_per_second": 6.082,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8086253369272236,
+      "grad_norm": 0.5690694451332092,
+      "learning_rate": 0.00043200647598488933,
+      "loss": 3.5296,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8140161725067383,
+      "grad_norm": 0.5980532169342041,
+      "learning_rate": 0.0004316826767404209,
+      "loss": 3.5044,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8194070080862534,
+      "grad_norm": 0.5740624070167542,
+      "learning_rate": 0.0004313588774959525,
+      "loss": 3.5082,
+      "step": 26150
+    },
+    {
+      "epoch": 2.824797843665768,
+      "grad_norm": 0.6146470904350281,
+      "learning_rate": 0.00043103507825148404,
+      "loss": 3.532,
+      "step": 26200
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.5921385288238525,
+      "learning_rate": 0.0004307112790070156,
+      "loss": 3.5445,
+      "step": 26250
+    },
+    {
+      "epoch": 2.835579514824798,
+      "grad_norm": 0.5627673268318176,
+      "learning_rate": 0.0004303874797625472,
+      "loss": 3.5604,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8409703504043127,
+      "grad_norm": 0.6661468744277954,
+      "learning_rate": 0.00043006368051807874,
+      "loss": 3.5291,
+      "step": 26350
+    },
+    {
+      "epoch": 2.8463611859838274,
+      "grad_norm": 0.6103497743606567,
+      "learning_rate": 0.00042973988127361035,
+      "loss": 3.5391,
+      "step": 26400
+    },
+    {
+      "epoch": 2.8517520215633425,
+      "grad_norm": 0.6032962799072266,
+      "learning_rate": 0.0004294160820291419,
+      "loss": 3.545,
+      "step": 26450
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.5415673851966858,
+      "learning_rate": 0.0004290922827846735,
+      "loss": 3.5161,
+      "step": 26500
+    },
+    {
+      "epoch": 2.862533692722372,
+      "grad_norm": 0.5705322027206421,
+      "learning_rate": 0.00042876848354020505,
+      "loss": 3.5323,
+      "step": 26550
+    },
+    {
+      "epoch": 2.867924528301887,
+      "grad_norm": 0.5669793486595154,
+      "learning_rate": 0.00042844468429573655,
+      "loss": 3.5407,
+      "step": 26600
+    },
+    {
+      "epoch": 2.8733153638814017,
+      "grad_norm": 0.605086624622345,
+      "learning_rate": 0.00042812088505126815,
+      "loss": 3.5466,
+      "step": 26650
+    },
+    {
+      "epoch": 2.8787061994609164,
+      "grad_norm": 0.5618116855621338,
+      "learning_rate": 0.0004277970858067997,
+      "loss": 3.5318,
+      "step": 26700
+    },
+    {
+      "epoch": 2.884097035040431,
+      "grad_norm": 0.6186133027076721,
+      "learning_rate": 0.0004274732865623313,
+      "loss": 3.5432,
+      "step": 26750
+    },
+    {
+      "epoch": 2.889487870619946,
+      "grad_norm": 0.5849158763885498,
+      "learning_rate": 0.00042714948731786286,
+      "loss": 3.5203,
+      "step": 26800
+    },
+    {
+      "epoch": 2.894878706199461,
+      "grad_norm": 0.6805973052978516,
+      "learning_rate": 0.00042682568807339447,
+      "loss": 3.5437,
+      "step": 26850
+    },
+    {
+      "epoch": 2.9002695417789757,
+      "grad_norm": 0.6036573648452759,
+      "learning_rate": 0.000426501888828926,
+      "loss": 3.5082,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.6047782301902771,
+      "learning_rate": 0.0004261780895844576,
+      "loss": 3.5299,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "grad_norm": 0.5614576935768127,
+      "learning_rate": 0.00042585429033998917,
+      "loss": 3.5442,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "eval_accuracy": 0.3696261504031946,
+      "eval_loss": 3.5053839683532715,
+      "eval_runtime": 185.4428,
+      "eval_samples_per_second": 97.124,
+      "eval_steps_per_second": 6.072,
+      "step": 27000
+    },
+    {
+      "epoch": 2.91644204851752,
+      "grad_norm": 0.6214551329612732,
+      "learning_rate": 0.0004255304910955207,
+      "loss": 3.5288,
+      "step": 27050
+    },
+    {
+      "epoch": 2.921832884097035,
+      "grad_norm": 0.5811451077461243,
+      "learning_rate": 0.0004252066918510523,
+      "loss": 3.5338,
+      "step": 27100
+    },
+    {
+      "epoch": 2.92722371967655,
+      "grad_norm": 0.6257526874542236,
+      "learning_rate": 0.0004248828926065839,
+      "loss": 3.5296,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9326145552560647,
+      "grad_norm": 0.6285461187362671,
+      "learning_rate": 0.0004245590933621155,
+      "loss": 3.5139,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9380053908355794,
+      "grad_norm": 0.549082338809967,
+      "learning_rate": 0.00042423529411764703,
+      "loss": 3.5302,
+      "step": 27250
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 0.5633655786514282,
+      "learning_rate": 0.00042391149487317864,
+      "loss": 3.5334,
+      "step": 27300
+    },
+    {
+      "epoch": 2.9487870619946093,
+      "grad_norm": 0.5984169244766235,
+      "learning_rate": 0.00042358769562871013,
+      "loss": 3.5178,
+      "step": 27350
+    },
+    {
+      "epoch": 2.954177897574124,
+      "grad_norm": 0.6494591236114502,
+      "learning_rate": 0.00042326389638424174,
+      "loss": 3.5112,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9595687331536387,
+      "grad_norm": 0.5794746279716492,
+      "learning_rate": 0.0004229400971397733,
+      "loss": 3.5172,
+      "step": 27450
+    },
+    {
+      "epoch": 2.964959568733154,
+      "grad_norm": 0.5832740068435669,
+      "learning_rate": 0.00042261629789530484,
+      "loss": 3.5293,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9703504043126685,
+      "grad_norm": 0.5642590522766113,
+      "learning_rate": 0.00042229249865083644,
+      "loss": 3.5181,
+      "step": 27550
+    },
+    {
+      "epoch": 2.975741239892183,
+      "grad_norm": 0.5790451765060425,
+      "learning_rate": 0.000421968699406368,
+      "loss": 3.5252,
+      "step": 27600
+    },
+    {
+      "epoch": 2.981132075471698,
+      "grad_norm": 0.6082571744918823,
+      "learning_rate": 0.0004216449001618996,
+      "loss": 3.5057,
+      "step": 27650
+    },
+    {
+      "epoch": 2.986522911051213,
+      "grad_norm": 0.5873731374740601,
+      "learning_rate": 0.00042132110091743115,
+      "loss": 3.5168,
+      "step": 27700
+    },
+    {
+      "epoch": 2.9919137466307277,
+      "grad_norm": 0.6429144144058228,
+      "learning_rate": 0.00042099730167296275,
+      "loss": 3.5223,
+      "step": 27750
+    },
+    {
+      "epoch": 2.9973045822102424,
+      "grad_norm": 0.642909586429596,
+      "learning_rate": 0.00042067997841338365,
+      "loss": 3.5247,
+      "step": 27800
+    },
+    {
+      "epoch": 3.0026954177897576,
+      "grad_norm": 0.6168606281280518,
+      "learning_rate": 0.00042035617916891526,
+      "loss": 3.4874,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0080862533692723,
+      "grad_norm": 0.6800719499588013,
+      "learning_rate": 0.0004200323799244468,
+      "loss": 3.4281,
+      "step": 27900
+    },
+    {
+      "epoch": 3.013477088948787,
+      "grad_norm": 0.57563316822052,
+      "learning_rate": 0.0004197085806799784,
+      "loss": 3.4331,
+      "step": 27950
+    },
+    {
+      "epoch": 3.018867924528302,
+      "grad_norm": 0.607589602470398,
+      "learning_rate": 0.0004193847814355099,
+      "loss": 3.4451,
+      "step": 28000
+    },
+    {
+      "epoch": 3.018867924528302,
+      "eval_accuracy": 0.3699707973633421,
+      "eval_loss": 3.5011649131774902,
+      "eval_runtime": 185.7113,
+      "eval_samples_per_second": 96.984,
+      "eval_steps_per_second": 6.063,
+      "step": 28000
+    },
+    {
+      "epoch": 3.024258760107817,
+      "grad_norm": 0.6603455543518066,
+      "learning_rate": 0.0004190609821910415,
+      "loss": 3.4317,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0296495956873315,
+      "grad_norm": 0.63117516040802,
+      "learning_rate": 0.00041873718294657306,
+      "loss": 3.4264,
+      "step": 28100
+    },
+    {
+      "epoch": 3.035040431266846,
+      "grad_norm": 0.6193031072616577,
+      "learning_rate": 0.00041841338370210467,
+      "loss": 3.466,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0404312668463613,
+      "grad_norm": 0.6265783905982971,
+      "learning_rate": 0.0004180895844576362,
+      "loss": 3.4467,
+      "step": 28200
+    },
+    {
+      "epoch": 3.045822102425876,
+      "grad_norm": 0.5731920003890991,
+      "learning_rate": 0.00041776578521316777,
+      "loss": 3.4543,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0512129380053907,
+      "grad_norm": 0.6172985434532166,
+      "learning_rate": 0.0004174419859686994,
+      "loss": 3.4541,
+      "step": 28300
+    },
+    {
+      "epoch": 3.056603773584906,
+      "grad_norm": 0.5971702337265015,
+      "learning_rate": 0.0004171181867242309,
+      "loss": 3.4452,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0619946091644206,
+      "grad_norm": 0.6083709001541138,
+      "learning_rate": 0.00041679438747976253,
+      "loss": 3.4237,
+      "step": 28400
+    },
+    {
+      "epoch": 3.0673854447439353,
+      "grad_norm": 0.595926821231842,
+      "learning_rate": 0.0004164705882352941,
+      "loss": 3.4655,
+      "step": 28450
+    },
+    {
+      "epoch": 3.07277628032345,
+      "grad_norm": 0.6328946352005005,
+      "learning_rate": 0.0004161467889908257,
+      "loss": 3.4581,
+      "step": 28500
+    },
+    {
+      "epoch": 3.078167115902965,
+      "grad_norm": 0.62444669008255,
+      "learning_rate": 0.00041582298974635724,
+      "loss": 3.4453,
+      "step": 28550
+    },
+    {
+      "epoch": 3.08355795148248,
+      "grad_norm": 0.5992231965065002,
+      "learning_rate": 0.00041549919050188884,
+      "loss": 3.442,
+      "step": 28600
+    },
+    {
+      "epoch": 3.0889487870619945,
+      "grad_norm": 0.6712383031845093,
+      "learning_rate": 0.00041517539125742034,
+      "loss": 3.4401,
+      "step": 28650
+    },
+    {
+      "epoch": 3.0943396226415096,
+      "grad_norm": 0.58417809009552,
+      "learning_rate": 0.0004148515920129519,
+      "loss": 3.4495,
+      "step": 28700
+    },
+    {
+      "epoch": 3.0997304582210243,
+      "grad_norm": 0.6253139972686768,
+      "learning_rate": 0.0004145277927684835,
+      "loss": 3.4385,
+      "step": 28750
+    },
+    {
+      "epoch": 3.105121293800539,
+      "grad_norm": 0.5860809683799744,
+      "learning_rate": 0.00041420399352401504,
+      "loss": 3.4427,
+      "step": 28800
+    },
+    {
+      "epoch": 3.1105121293800537,
+      "grad_norm": 0.6193458437919617,
+      "learning_rate": 0.00041388019427954665,
+      "loss": 3.4264,
+      "step": 28850
+    },
+    {
+      "epoch": 3.115902964959569,
+      "grad_norm": 0.5895789861679077,
+      "learning_rate": 0.0004135563950350782,
+      "loss": 3.4446,
+      "step": 28900
+    },
+    {
+      "epoch": 3.1212938005390836,
+      "grad_norm": 0.6106541156768799,
+      "learning_rate": 0.0004132325957906098,
+      "loss": 3.4618,
+      "step": 28950
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "grad_norm": 0.6361096501350403,
+      "learning_rate": 0.00041290879654614135,
+      "loss": 3.4295,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "eval_accuracy": 0.3709342225195931,
+      "eval_loss": 3.496896266937256,
+      "eval_runtime": 184.8429,
+      "eval_samples_per_second": 97.439,
+      "eval_steps_per_second": 6.092,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1320754716981134,
+      "grad_norm": 0.665164589881897,
+      "learning_rate": 0.00041258499730167296,
+      "loss": 3.4525,
+      "step": 29050
+    },
+    {
+      "epoch": 3.137466307277628,
+      "grad_norm": 0.6416609883308411,
+      "learning_rate": 0.0004122611980572045,
+      "loss": 3.4413,
+      "step": 29100
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.6141297817230225,
+      "learning_rate": 0.00041193739881273606,
+      "loss": 3.4591,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1482479784366575,
+      "grad_norm": 0.5525941848754883,
+      "learning_rate": 0.00041161359956826766,
+      "loss": 3.4582,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1536388140161726,
+      "grad_norm": 0.5938381552696228,
+      "learning_rate": 0.0004112898003237992,
+      "loss": 3.4456,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1590296495956873,
+      "grad_norm": 0.6277801990509033,
+      "learning_rate": 0.0004109660010793308,
+      "loss": 3.4546,
+      "step": 29300
+    },
+    {
+      "epoch": 3.164420485175202,
+      "grad_norm": 0.6602087616920471,
+      "learning_rate": 0.0004106422018348623,
+      "loss": 3.4666,
+      "step": 29350
+    },
+    {
+      "epoch": 3.169811320754717,
+      "grad_norm": 0.5656922459602356,
+      "learning_rate": 0.0004103184025903939,
+      "loss": 3.458,
+      "step": 29400
+    },
+    {
+      "epoch": 3.175202156334232,
+      "grad_norm": 0.605190634727478,
+      "learning_rate": 0.00040999460334592547,
+      "loss": 3.4687,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1805929919137466,
+      "grad_norm": 0.6249133944511414,
+      "learning_rate": 0.000409670804101457,
+      "loss": 3.465,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1859838274932613,
+      "grad_norm": 0.5883227586746216,
+      "learning_rate": 0.00040934700485698863,
+      "loss": 3.4413,
+      "step": 29550
+    },
+    {
+      "epoch": 3.1913746630727764,
+      "grad_norm": 0.5973005294799805,
+      "learning_rate": 0.0004090232056125202,
+      "loss": 3.4628,
+      "step": 29600
+    },
+    {
+      "epoch": 3.196765498652291,
+      "grad_norm": 0.5894708633422852,
+      "learning_rate": 0.0004086994063680518,
+      "loss": 3.4505,
+      "step": 29650
+    },
+    {
+      "epoch": 3.202156334231806,
+      "grad_norm": 0.5822822451591492,
+      "learning_rate": 0.00040837560712358333,
+      "loss": 3.4559,
+      "step": 29700
+    },
+    {
+      "epoch": 3.207547169811321,
+      "grad_norm": 0.6240995526313782,
+      "learning_rate": 0.00040805180787911494,
+      "loss": 3.4638,
+      "step": 29750
+    },
+    {
+      "epoch": 3.2129380053908356,
+      "grad_norm": 0.5901066064834595,
+      "learning_rate": 0.0004077344846195359,
+      "loss": 3.4456,
+      "step": 29800
+    },
+    {
+      "epoch": 3.2183288409703503,
+      "grad_norm": 0.6369420289993286,
+      "learning_rate": 0.00040741068537506744,
+      "loss": 3.467,
+      "step": 29850
+    },
+    {
+      "epoch": 3.223719676549865,
+      "grad_norm": 0.6294394135475159,
+      "learning_rate": 0.000407086886130599,
+      "loss": 3.4735,
+      "step": 29900
+    },
+    {
+      "epoch": 3.22911051212938,
+      "grad_norm": 0.6184892654418945,
+      "learning_rate": 0.0004067630868861306,
+      "loss": 3.4679,
+      "step": 29950
+    },
+    {
+      "epoch": 3.234501347708895,
+      "grad_norm": 0.6455713510513306,
+      "learning_rate": 0.0004064392876416621,
+      "loss": 3.4506,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234501347708895,
+      "eval_accuracy": 0.3713301536427891,
+      "eval_loss": 3.490234613418579,
+      "eval_runtime": 185.1295,
+      "eval_samples_per_second": 97.289,
+      "eval_steps_per_second": 6.082,
+      "step": 30000
+    },
+    {
+      "epoch": 3.2398921832884096,
+      "grad_norm": 0.642624020576477,
+      "learning_rate": 0.0004061154883971937,
+      "loss": 3.4476,
+      "step": 30050
+    },
+    {
+      "epoch": 3.2452830188679247,
+      "grad_norm": 0.626826822757721,
+      "learning_rate": 0.00040579168915272525,
+      "loss": 3.4482,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2506738544474394,
+      "grad_norm": 0.5784972310066223,
+      "learning_rate": 0.00040546788990825685,
+      "loss": 3.4517,
+      "step": 30150
+    },
+    {
+      "epoch": 3.256064690026954,
+      "grad_norm": 0.64665287733078,
+      "learning_rate": 0.0004051440906637884,
+      "loss": 3.4719,
+      "step": 30200
+    },
+    {
+      "epoch": 3.2614555256064692,
+      "grad_norm": 0.6225249171257019,
+      "learning_rate": 0.00040482029141931995,
+      "loss": 3.464,
+      "step": 30250
+    },
+    {
+      "epoch": 3.266846361185984,
+      "grad_norm": 0.7104143500328064,
+      "learning_rate": 0.00040449649217485156,
+      "loss": 3.465,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2722371967654986,
+      "grad_norm": 0.6647838950157166,
+      "learning_rate": 0.0004041726929303831,
+      "loss": 3.4433,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2776280323450133,
+      "grad_norm": 0.678033173084259,
+      "learning_rate": 0.0004038488936859147,
+      "loss": 3.4726,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2830188679245285,
+      "grad_norm": 0.6125953197479248,
+      "learning_rate": 0.00040352509444144626,
+      "loss": 3.4584,
+      "step": 30450
+    },
+    {
+      "epoch": 3.288409703504043,
+      "grad_norm": 0.5959815979003906,
+      "learning_rate": 0.00040320129519697787,
+      "loss": 3.4427,
+      "step": 30500
+    },
+    {
+      "epoch": 3.293800539083558,
+      "grad_norm": 0.6709897518157959,
+      "learning_rate": 0.0004028774959525094,
+      "loss": 3.4719,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2991913746630726,
+      "grad_norm": 0.7089282870292664,
+      "learning_rate": 0.000402553696708041,
+      "loss": 3.4638,
+      "step": 30600
+    },
+    {
+      "epoch": 3.3045822102425877,
+      "grad_norm": 0.6118188500404358,
+      "learning_rate": 0.0004022298974635726,
+      "loss": 3.4443,
+      "step": 30650
+    },
+    {
+      "epoch": 3.3099730458221024,
+      "grad_norm": 0.6362579464912415,
+      "learning_rate": 0.00040190609821910407,
+      "loss": 3.4604,
+      "step": 30700
+    },
+    {
+      "epoch": 3.315363881401617,
+      "grad_norm": 0.6267250776290894,
+      "learning_rate": 0.0004015822989746357,
+      "loss": 3.4298,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3207547169811322,
+      "grad_norm": 0.5901178121566772,
+      "learning_rate": 0.0004012584997301672,
+      "loss": 3.4663,
+      "step": 30800
+    },
+    {
+      "epoch": 3.326145552560647,
+      "grad_norm": 0.6500908732414246,
+      "learning_rate": 0.00040093470048569883,
+      "loss": 3.446,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3315363881401616,
+      "grad_norm": 0.6370844841003418,
+      "learning_rate": 0.0004006109012412304,
+      "loss": 3.4689,
+      "step": 30900
+    },
+    {
+      "epoch": 3.3369272237196768,
+      "grad_norm": 0.6281841993331909,
+      "learning_rate": 0.000400287101996762,
+      "loss": 3.4456,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "grad_norm": 0.6539614200592041,
+      "learning_rate": 0.00039996330275229354,
+      "loss": 3.4582,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "eval_accuracy": 0.37187863342047656,
+      "eval_loss": 3.48494291305542,
+      "eval_runtime": 184.9741,
+      "eval_samples_per_second": 97.37,
+      "eval_steps_per_second": 6.087,
+      "step": 31000
+    },
+    {
+      "epoch": 3.347708894878706,
+      "grad_norm": 0.8110284209251404,
+      "learning_rate": 0.00039963950350782514,
+      "loss": 3.4593,
+      "step": 31050
+    },
+    {
+      "epoch": 3.353099730458221,
+      "grad_norm": 0.6377867460250854,
+      "learning_rate": 0.0003993157042633567,
+      "loss": 3.4541,
+      "step": 31100
+    },
+    {
+      "epoch": 3.358490566037736,
+      "grad_norm": 0.6073828935623169,
+      "learning_rate": 0.00039899190501888824,
+      "loss": 3.4652,
+      "step": 31150
+    },
+    {
+      "epoch": 3.3638814016172507,
+      "grad_norm": 0.6192326545715332,
+      "learning_rate": 0.00039866810577441985,
+      "loss": 3.4487,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3692722371967654,
+      "grad_norm": 0.6012443900108337,
+      "learning_rate": 0.0003983443065299514,
+      "loss": 3.4543,
+      "step": 31250
+    },
+    {
+      "epoch": 3.37466307277628,
+      "grad_norm": 0.5912206172943115,
+      "learning_rate": 0.000398020507285483,
+      "loss": 3.4795,
+      "step": 31300
+    },
+    {
+      "epoch": 3.3800539083557952,
+      "grad_norm": 0.615149974822998,
+      "learning_rate": 0.0003976967080410145,
+      "loss": 3.4505,
+      "step": 31350
+    },
+    {
+      "epoch": 3.38544474393531,
+      "grad_norm": 0.6301441192626953,
+      "learning_rate": 0.0003973729087965461,
+      "loss": 3.4511,
+      "step": 31400
+    },
+    {
+      "epoch": 3.3908355795148246,
+      "grad_norm": 0.6270235180854797,
+      "learning_rate": 0.00039704910955207765,
+      "loss": 3.4568,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3962264150943398,
+      "grad_norm": 0.5964970588684082,
+      "learning_rate": 0.00039672531030760926,
+      "loss": 3.449,
+      "step": 31500
+    },
+    {
+      "epoch": 3.4016172506738545,
+      "grad_norm": 0.6153132319450378,
+      "learning_rate": 0.0003964015110631408,
+      "loss": 3.4681,
+      "step": 31550
+    },
+    {
+      "epoch": 3.407008086253369,
+      "grad_norm": 0.6561875343322754,
+      "learning_rate": 0.00039607771181867236,
+      "loss": 3.4452,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4123989218328843,
+      "grad_norm": 0.6051456332206726,
+      "learning_rate": 0.00039575391257420397,
+      "loss": 3.4482,
+      "step": 31650
+    },
+    {
+      "epoch": 3.417789757412399,
+      "grad_norm": 0.6496569514274597,
+      "learning_rate": 0.0003954301133297355,
+      "loss": 3.4746,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4231805929919137,
+      "grad_norm": 0.6450279355049133,
+      "learning_rate": 0.0003951063140852671,
+      "loss": 3.4468,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4285714285714284,
+      "grad_norm": 0.6020802855491638,
+      "learning_rate": 0.00039478251484079867,
+      "loss": 3.4509,
+      "step": 31800
+    },
+    {
+      "epoch": 3.4339622641509435,
+      "grad_norm": 0.6902250647544861,
+      "learning_rate": 0.0003944587155963303,
+      "loss": 3.4817,
+      "step": 31850
+    },
+    {
+      "epoch": 3.439353099730458,
+      "grad_norm": 0.6622137427330017,
+      "learning_rate": 0.0003941349163518618,
+      "loss": 3.4645,
+      "step": 31900
+    },
+    {
+      "epoch": 3.444743935309973,
+      "grad_norm": 0.6333610415458679,
+      "learning_rate": 0.0003938175930922828,
+      "loss": 3.4427,
+      "step": 31950
+    },
+    {
+      "epoch": 3.450134770889488,
+      "grad_norm": 0.6233834624290466,
+      "learning_rate": 0.0003934937938478143,
+      "loss": 3.4597,
+      "step": 32000
+    },
+    {
+      "epoch": 3.450134770889488,
+      "eval_accuracy": 0.3733303446545658,
+      "eval_loss": 3.4757399559020996,
+      "eval_runtime": 185.0469,
+      "eval_samples_per_second": 97.332,
+      "eval_steps_per_second": 6.085,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4555256064690028,
+      "grad_norm": 0.6415245532989502,
+      "learning_rate": 0.0003931699946033459,
+      "loss": 3.4613,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4609164420485174,
+      "grad_norm": 0.6769406199455261,
+      "learning_rate": 0.00039284619535887743,
+      "loss": 3.4545,
+      "step": 32100
+    },
+    {
+      "epoch": 3.466307277628032,
+      "grad_norm": 0.6122919321060181,
+      "learning_rate": 0.00039252239611440904,
+      "loss": 3.4669,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4716981132075473,
+      "grad_norm": 0.6024535894393921,
+      "learning_rate": 0.0003921985968699406,
+      "loss": 3.4484,
+      "step": 32200
+    },
+    {
+      "epoch": 3.477088948787062,
+      "grad_norm": 0.6409429311752319,
+      "learning_rate": 0.0003918747976254722,
+      "loss": 3.4661,
+      "step": 32250
+    },
+    {
+      "epoch": 3.4824797843665767,
+      "grad_norm": 0.6143214106559753,
+      "learning_rate": 0.00039155099838100374,
+      "loss": 3.4796,
+      "step": 32300
+    },
+    {
+      "epoch": 3.487870619946092,
+      "grad_norm": 0.5818310976028442,
+      "learning_rate": 0.0003912271991365353,
+      "loss": 3.4407,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4932614555256065,
+      "grad_norm": 0.6170235276222229,
+      "learning_rate": 0.0003909033998920669,
+      "loss": 3.4722,
+      "step": 32400
+    },
+    {
+      "epoch": 3.498652291105121,
+      "grad_norm": 0.6212165951728821,
+      "learning_rate": 0.00039057960064759845,
+      "loss": 3.446,
+      "step": 32450
+    },
+    {
+      "epoch": 3.5040431266846364,
+      "grad_norm": 0.6115076541900635,
+      "learning_rate": 0.00039025580140313005,
+      "loss": 3.4707,
+      "step": 32500
+    },
+    {
+      "epoch": 3.509433962264151,
+      "grad_norm": 0.5680184960365295,
+      "learning_rate": 0.0003899320021586616,
+      "loss": 3.4612,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5148247978436657,
+      "grad_norm": 0.6204341650009155,
+      "learning_rate": 0.0003896082029141932,
+      "loss": 3.4577,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5202156334231804,
+      "grad_norm": 0.6625291705131531,
+      "learning_rate": 0.00038928440366972476,
+      "loss": 3.4669,
+      "step": 32650
+    },
+    {
+      "epoch": 3.525606469002695,
+      "grad_norm": 0.6184507608413696,
+      "learning_rate": 0.00038896060442525636,
+      "loss": 3.4411,
+      "step": 32700
+    },
+    {
+      "epoch": 3.5309973045822103,
+      "grad_norm": 0.6391364932060242,
+      "learning_rate": 0.00038863680518078786,
+      "loss": 3.4536,
+      "step": 32750
+    },
+    {
+      "epoch": 3.536388140161725,
+      "grad_norm": 0.6026434898376465,
+      "learning_rate": 0.0003883130059363194,
+      "loss": 3.4641,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5417789757412397,
+      "grad_norm": 0.642951250076294,
+      "learning_rate": 0.000387989206691851,
+      "loss": 3.4257,
+      "step": 32850
+    },
+    {
+      "epoch": 3.547169811320755,
+      "grad_norm": 0.6193267107009888,
+      "learning_rate": 0.00038766540744738256,
+      "loss": 3.4491,
+      "step": 32900
+    },
+    {
+      "epoch": 3.5525606469002695,
+      "grad_norm": 0.6317867636680603,
+      "learning_rate": 0.00038734160820291417,
+      "loss": 3.4632,
+      "step": 32950
+    },
+    {
+      "epoch": 3.557951482479784,
+      "grad_norm": 0.654092013835907,
+      "learning_rate": 0.0003870178089584457,
+      "loss": 3.4639,
+      "step": 33000
+    },
+    {
+      "epoch": 3.557951482479784,
+      "eval_accuracy": 0.37359350194885255,
+      "eval_loss": 3.4721765518188477,
+      "eval_runtime": 185.0176,
+      "eval_samples_per_second": 97.347,
+      "eval_steps_per_second": 6.086,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5633423180592994,
+      "grad_norm": 0.5980980396270752,
+      "learning_rate": 0.0003866940097139773,
+      "loss": 3.4638,
+      "step": 33050
+    },
+    {
+      "epoch": 3.568733153638814,
+      "grad_norm": 0.671788215637207,
+      "learning_rate": 0.0003863702104695089,
+      "loss": 3.4525,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5741239892183287,
+      "grad_norm": 0.6702181100845337,
+      "learning_rate": 0.0003860464112250404,
+      "loss": 3.4601,
+      "step": 33150
+    },
+    {
+      "epoch": 3.579514824797844,
+      "grad_norm": 0.6230493187904358,
+      "learning_rate": 0.00038572261198057203,
+      "loss": 3.4616,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5849056603773586,
+      "grad_norm": 0.6334137916564941,
+      "learning_rate": 0.0003853988127361036,
+      "loss": 3.4536,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5902964959568733,
+      "grad_norm": 0.6252257227897644,
+      "learning_rate": 0.0003850750134916352,
+      "loss": 3.4431,
+      "step": 33300
+    },
+    {
+      "epoch": 3.595687331536388,
+      "grad_norm": 0.5797781348228455,
+      "learning_rate": 0.0003847512142471667,
+      "loss": 3.454,
+      "step": 33350
+    },
+    {
+      "epoch": 3.601078167115903,
+      "grad_norm": 0.6160080432891846,
+      "learning_rate": 0.0003844274150026983,
+      "loss": 3.4549,
+      "step": 33400
+    },
+    {
+      "epoch": 3.606469002695418,
+      "grad_norm": 0.6341015696525574,
+      "learning_rate": 0.00038410361575822984,
+      "loss": 3.4639,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6118598382749325,
+      "grad_norm": 0.6332781910896301,
+      "learning_rate": 0.00038377981651376144,
+      "loss": 3.4719,
+      "step": 33500
+    },
+    {
+      "epoch": 3.617250673854447,
+      "grad_norm": 0.6350259184837341,
+      "learning_rate": 0.000383456017269293,
+      "loss": 3.4501,
+      "step": 33550
+    },
+    {
+      "epoch": 3.6226415094339623,
+      "grad_norm": 0.6050942540168762,
+      "learning_rate": 0.00038313221802482454,
+      "loss": 3.4529,
+      "step": 33600
+    },
+    {
+      "epoch": 3.628032345013477,
+      "grad_norm": 0.679898202419281,
+      "learning_rate": 0.00038280841878035615,
+      "loss": 3.4496,
+      "step": 33650
+    },
+    {
+      "epoch": 3.6334231805929917,
+      "grad_norm": 0.6030776500701904,
+      "learning_rate": 0.0003824846195358877,
+      "loss": 3.438,
+      "step": 33700
+    },
+    {
+      "epoch": 3.638814016172507,
+      "grad_norm": 0.6465497016906738,
+      "learning_rate": 0.0003821608202914193,
+      "loss": 3.46,
+      "step": 33750
+    },
+    {
+      "epoch": 3.6442048517520216,
+      "grad_norm": 0.6347126960754395,
+      "learning_rate": 0.00038183702104695085,
+      "loss": 3.4418,
+      "step": 33800
+    },
+    {
+      "epoch": 3.6495956873315363,
+      "grad_norm": 0.6211069822311401,
+      "learning_rate": 0.00038151322180248246,
+      "loss": 3.4722,
+      "step": 33850
+    },
+    {
+      "epoch": 3.6549865229110514,
+      "grad_norm": 0.6023736000061035,
+      "learning_rate": 0.000381189422558014,
+      "loss": 3.4463,
+      "step": 33900
+    },
+    {
+      "epoch": 3.660377358490566,
+      "grad_norm": 0.6519888043403625,
+      "learning_rate": 0.00038087209929843496,
+      "loss": 3.4477,
+      "step": 33950
+    },
+    {
+      "epoch": 3.665768194070081,
+      "grad_norm": 0.6298272013664246,
+      "learning_rate": 0.00038054830005396646,
+      "loss": 3.4602,
+      "step": 34000
+    },
+    {
+      "epoch": 3.665768194070081,
+      "eval_accuracy": 0.3735198352909143,
+      "eval_loss": 3.4653146266937256,
+      "eval_runtime": 185.1109,
+      "eval_samples_per_second": 97.298,
+      "eval_steps_per_second": 6.083,
+      "step": 34000
+    },
+    {
+      "epoch": 3.671159029649596,
+      "grad_norm": 0.6116703152656555,
+      "learning_rate": 0.00038022450080949806,
+      "loss": 3.4527,
+      "step": 34050
+    },
+    {
+      "epoch": 3.6765498652291106,
+      "grad_norm": 0.6061373949050903,
+      "learning_rate": 0.0003799007015650296,
+      "loss": 3.4695,
+      "step": 34100
+    },
+    {
+      "epoch": 3.6819407008086253,
+      "grad_norm": 0.6899125576019287,
+      "learning_rate": 0.0003795769023205612,
+      "loss": 3.4656,
+      "step": 34150
+    },
+    {
+      "epoch": 3.68733153638814,
+      "grad_norm": 0.6152036190032959,
+      "learning_rate": 0.00037925310307609277,
+      "loss": 3.4485,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6927223719676547,
+      "grad_norm": 0.6439741253852844,
+      "learning_rate": 0.0003789293038316244,
+      "loss": 3.4411,
+      "step": 34250
+    },
+    {
+      "epoch": 3.69811320754717,
+      "grad_norm": 0.5817996263504028,
+      "learning_rate": 0.0003786055045871559,
+      "loss": 3.4418,
+      "step": 34300
+    },
+    {
+      "epoch": 3.7035040431266846,
+      "grad_norm": 0.633815348148346,
+      "learning_rate": 0.0003782817053426875,
+      "loss": 3.446,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7088948787061993,
+      "grad_norm": 0.593890905380249,
+      "learning_rate": 0.0003779579060982191,
+      "loss": 3.4419,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7142857142857144,
+      "grad_norm": 0.6376861929893494,
+      "learning_rate": 0.00037763410685375063,
+      "loss": 3.4648,
+      "step": 34450
+    },
+    {
+      "epoch": 3.719676549865229,
+      "grad_norm": 0.616452157497406,
+      "learning_rate": 0.00037731030760928223,
+      "loss": 3.4457,
+      "step": 34500
+    },
+    {
+      "epoch": 3.725067385444744,
+      "grad_norm": 0.6425932049751282,
+      "learning_rate": 0.0003769865083648138,
+      "loss": 3.4385,
+      "step": 34550
+    },
+    {
+      "epoch": 3.730458221024259,
+      "grad_norm": 0.6993752717971802,
+      "learning_rate": 0.0003766627091203454,
+      "loss": 3.4588,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7358490566037736,
+      "grad_norm": 0.6306095123291016,
+      "learning_rate": 0.00037633890987587694,
+      "loss": 3.4574,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7412398921832883,
+      "grad_norm": 0.6183214783668518,
+      "learning_rate": 0.00037602158661629784,
+      "loss": 3.4664,
+      "step": 34700
+    },
+    {
+      "epoch": 3.7466307277628035,
+      "grad_norm": 0.616973340511322,
+      "learning_rate": 0.0003756977873718294,
+      "loss": 3.4632,
+      "step": 34750
+    },
+    {
+      "epoch": 3.752021563342318,
+      "grad_norm": 0.64264976978302,
+      "learning_rate": 0.000375373988127361,
+      "loss": 3.4428,
+      "step": 34800
+    },
+    {
+      "epoch": 3.757412398921833,
+      "grad_norm": 0.6926430463790894,
+      "learning_rate": 0.00037505018888289254,
+      "loss": 3.4469,
+      "step": 34850
+    },
+    {
+      "epoch": 3.7628032345013476,
+      "grad_norm": 0.6002678871154785,
+      "learning_rate": 0.00037472638963842415,
+      "loss": 3.4612,
+      "step": 34900
+    },
+    {
+      "epoch": 3.7681940700808623,
+      "grad_norm": 0.7651305794715881,
+      "learning_rate": 0.0003744025903939557,
+      "loss": 3.454,
+      "step": 34950
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "grad_norm": 0.6672982573509216,
+      "learning_rate": 0.0003740787911494873,
+      "loss": 3.4613,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "eval_accuracy": 0.37492178350238453,
+      "eval_loss": 3.4591267108917236,
+      "eval_runtime": 185.1296,
+      "eval_samples_per_second": 97.289,
+      "eval_steps_per_second": 6.082,
+      "step": 35000
+    },
+    {
+      "epoch": 3.778975741239892,
+      "grad_norm": 0.6062513589859009,
+      "learning_rate": 0.00037375499190501885,
+      "loss": 3.47,
+      "step": 35050
+    },
+    {
+      "epoch": 3.784366576819407,
+      "grad_norm": 0.6883726716041565,
+      "learning_rate": 0.0003734311926605504,
+      "loss": 3.4365,
+      "step": 35100
+    },
+    {
+      "epoch": 3.789757412398922,
+      "grad_norm": 0.6614878177642822,
+      "learning_rate": 0.000373107393416082,
+      "loss": 3.4592,
+      "step": 35150
+    },
+    {
+      "epoch": 3.7951482479784366,
+      "grad_norm": 0.6300371289253235,
+      "learning_rate": 0.00037278359417161356,
+      "loss": 3.4572,
+      "step": 35200
+    },
+    {
+      "epoch": 3.8005390835579513,
+      "grad_norm": 0.6380262970924377,
+      "learning_rate": 0.00037245979492714517,
+      "loss": 3.4392,
+      "step": 35250
+    },
+    {
+      "epoch": 3.8059299191374665,
+      "grad_norm": 0.6063967943191528,
+      "learning_rate": 0.0003721359956826767,
+      "loss": 3.4528,
+      "step": 35300
+    },
+    {
+      "epoch": 3.811320754716981,
+      "grad_norm": 0.6829068064689636,
+      "learning_rate": 0.0003718121964382083,
+      "loss": 3.4549,
+      "step": 35350
+    },
+    {
+      "epoch": 3.816711590296496,
+      "grad_norm": 0.6208524703979492,
+      "learning_rate": 0.0003714883971937398,
+      "loss": 3.4547,
+      "step": 35400
+    },
+    {
+      "epoch": 3.822102425876011,
+      "grad_norm": 0.6724206209182739,
+      "learning_rate": 0.0003711645979492714,
+      "loss": 3.4494,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8274932614555257,
+      "grad_norm": 0.586846113204956,
+      "learning_rate": 0.00037084079870480297,
+      "loss": 3.4517,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8328840970350404,
+      "grad_norm": 0.7098872065544128,
+      "learning_rate": 0.0003705169994603345,
+      "loss": 3.4652,
+      "step": 35550
+    },
+    {
+      "epoch": 3.838274932614555,
+      "grad_norm": 0.6315362453460693,
+      "learning_rate": 0.00037019320021586613,
+      "loss": 3.4693,
+      "step": 35600
+    },
+    {
+      "epoch": 3.8436657681940702,
+      "grad_norm": 0.6404989957809448,
+      "learning_rate": 0.0003698694009713977,
+      "loss": 3.4676,
+      "step": 35650
+    },
+    {
+      "epoch": 3.849056603773585,
+      "grad_norm": 0.5798198580741882,
+      "learning_rate": 0.0003695456017269293,
+      "loss": 3.469,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8544474393530996,
+      "grad_norm": 0.6681696772575378,
+      "learning_rate": 0.00036922180248246083,
+      "loss": 3.4485,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8598382749326143,
+      "grad_norm": 0.8003295660018921,
+      "learning_rate": 0.00036889800323799244,
+      "loss": 3.4547,
+      "step": 35800
+    },
+    {
+      "epoch": 3.8652291105121295,
+      "grad_norm": 0.6079601645469666,
+      "learning_rate": 0.000368574203993524,
+      "loss": 3.4312,
+      "step": 35850
+    },
+    {
+      "epoch": 3.870619946091644,
+      "grad_norm": 0.6316880583763123,
+      "learning_rate": 0.0003682504047490556,
+      "loss": 3.4479,
+      "step": 35900
+    },
+    {
+      "epoch": 3.876010781671159,
+      "grad_norm": 0.6452414393424988,
+      "learning_rate": 0.00036792660550458714,
+      "loss": 3.4706,
+      "step": 35950
+    },
+    {
+      "epoch": 3.881401617250674,
+      "grad_norm": 0.6125510931015015,
+      "learning_rate": 0.00036760280626011864,
+      "loss": 3.4517,
+      "step": 36000
+    },
+    {
+      "epoch": 3.881401617250674,
+      "eval_accuracy": 0.3754607018259444,
+      "eval_loss": 3.4510653018951416,
+      "eval_runtime": 185.2329,
+      "eval_samples_per_second": 97.234,
+      "eval_steps_per_second": 6.079,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8867924528301887,
+      "grad_norm": 0.7082051038742065,
+      "learning_rate": 0.0003672790070156503,
+      "loss": 3.4576,
+      "step": 36050
+    },
+    {
+      "epoch": 3.8921832884097034,
+      "grad_norm": 0.6857859492301941,
+      "learning_rate": 0.0003669552077711818,
+      "loss": 3.4543,
+      "step": 36100
+    },
+    {
+      "epoch": 3.8975741239892185,
+      "grad_norm": 0.5955175161361694,
+      "learning_rate": 0.0003666314085267134,
+      "loss": 3.4599,
+      "step": 36150
+    },
+    {
+      "epoch": 3.9029649595687332,
+      "grad_norm": 0.6065194010734558,
+      "learning_rate": 0.00036630760928224495,
+      "loss": 3.4475,
+      "step": 36200
+    },
+    {
+      "epoch": 3.908355795148248,
+      "grad_norm": 0.6187525391578674,
+      "learning_rate": 0.00036598381003777656,
+      "loss": 3.4643,
+      "step": 36250
+    },
+    {
+      "epoch": 3.913746630727763,
+      "grad_norm": 0.637075662612915,
+      "learning_rate": 0.0003656600107933081,
+      "loss": 3.4443,
+      "step": 36300
+    },
+    {
+      "epoch": 3.9191374663072778,
+      "grad_norm": 0.613892138004303,
+      "learning_rate": 0.0003653362115488397,
+      "loss": 3.4656,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9245283018867925,
+      "grad_norm": 0.598813533782959,
+      "learning_rate": 0.00036501241230437126,
+      "loss": 3.4665,
+      "step": 36400
+    },
+    {
+      "epoch": 3.929919137466307,
+      "grad_norm": 0.6649259924888611,
+      "learning_rate": 0.0003646886130599028,
+      "loss": 3.4416,
+      "step": 36450
+    },
+    {
+      "epoch": 3.935309973045822,
+      "grad_norm": 0.6432799696922302,
+      "learning_rate": 0.0003643648138154344,
+      "loss": 3.4519,
+      "step": 36500
+    },
+    {
+      "epoch": 3.940700808625337,
+      "grad_norm": 0.6609790921211243,
+      "learning_rate": 0.00036404101457096597,
+      "loss": 3.4532,
+      "step": 36550
+    },
+    {
+      "epoch": 3.9460916442048517,
+      "grad_norm": 0.5976753234863281,
+      "learning_rate": 0.00036371721532649757,
+      "loss": 3.4501,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9514824797843664,
+      "grad_norm": 0.7079935669898987,
+      "learning_rate": 0.0003633934160820291,
+      "loss": 3.4404,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9568733153638815,
+      "grad_norm": 0.6590920686721802,
+      "learning_rate": 0.00036306961683756073,
+      "loss": 3.452,
+      "step": 36700
+    },
+    {
+      "epoch": 3.9622641509433962,
+      "grad_norm": 0.6470325589179993,
+      "learning_rate": 0.0003627458175930922,
+      "loss": 3.4424,
+      "step": 36750
+    },
+    {
+      "epoch": 3.967654986522911,
+      "grad_norm": 0.625947117805481,
+      "learning_rate": 0.0003624220183486238,
+      "loss": 3.4346,
+      "step": 36800
+    },
+    {
+      "epoch": 3.973045822102426,
+      "grad_norm": 0.6342935562133789,
+      "learning_rate": 0.0003620982191041554,
+      "loss": 3.4505,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9784366576819408,
+      "grad_norm": 0.6525506973266602,
+      "learning_rate": 0.00036177441985968693,
+      "loss": 3.4421,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9838274932614555,
+      "grad_norm": 0.6463566422462463,
+      "learning_rate": 0.00036145062061521854,
+      "loss": 3.4535,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "grad_norm": 0.6641883254051208,
+      "learning_rate": 0.0003611268213707501,
+      "loss": 3.4488,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "eval_accuracy": 0.3760902366579415,
+      "eval_loss": 3.445734977722168,
+      "eval_runtime": 185.0242,
+      "eval_samples_per_second": 97.344,
+      "eval_steps_per_second": 6.086,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9946091644204853,
+      "grad_norm": 0.6596526503562927,
+      "learning_rate": 0.0003608030221262817,
+      "loss": 3.4613,
+      "step": 37050
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.2152858972549438,
+      "learning_rate": 0.00036047922288181324,
+      "loss": 3.4334,
+      "step": 37100
+    },
+    {
+      "epoch": 4.005390835579515,
+      "grad_norm": 0.6752334833145142,
+      "learning_rate": 0.00036015542363734485,
+      "loss": 3.3286,
+      "step": 37150
+    },
+    {
+      "epoch": 4.010781671159029,
+      "grad_norm": 0.6589771509170532,
+      "learning_rate": 0.0003598316243928764,
+      "loss": 3.3567,
+      "step": 37200
+    },
+    {
+      "epoch": 4.0161725067385445,
+      "grad_norm": 0.6467882990837097,
+      "learning_rate": 0.00035950782514840795,
+      "loss": 3.3498,
+      "step": 37250
+    },
+    {
+      "epoch": 4.02156334231806,
+      "grad_norm": 0.6317258477210999,
+      "learning_rate": 0.00035918402590393955,
+      "loss": 3.3546,
+      "step": 37300
+    },
+    {
+      "epoch": 4.026954177897574,
+      "grad_norm": 0.6882931590080261,
+      "learning_rate": 0.00035886022665947105,
+      "loss": 3.3413,
+      "step": 37350
+    },
+    {
+      "epoch": 4.032345013477089,
+      "grad_norm": 0.6179378628730774,
+      "learning_rate": 0.0003585364274150027,
+      "loss": 3.361,
+      "step": 37400
+    },
+    {
+      "epoch": 4.037735849056604,
+      "grad_norm": 0.7111193537712097,
+      "learning_rate": 0.0003582126281705342,
+      "loss": 3.3608,
+      "step": 37450
+    },
+    {
+      "epoch": 4.0431266846361185,
+      "grad_norm": 0.6103177666664124,
+      "learning_rate": 0.0003578888289260658,
+      "loss": 3.3477,
+      "step": 37500
+    },
+    {
+      "epoch": 4.048517520215634,
+      "grad_norm": 0.67775958776474,
+      "learning_rate": 0.00035756502968159736,
+      "loss": 3.3644,
+      "step": 37550
+    },
+    {
+      "epoch": 4.053908355795148,
+      "grad_norm": 0.663544774055481,
+      "learning_rate": 0.00035724123043712896,
+      "loss": 3.3578,
+      "step": 37600
+    },
+    {
+      "epoch": 4.059299191374663,
+      "grad_norm": 0.6494781374931335,
+      "learning_rate": 0.0003569174311926605,
+      "loss": 3.3734,
+      "step": 37650
+    },
+    {
+      "epoch": 4.064690026954178,
+      "grad_norm": 0.6415791511535645,
+      "learning_rate": 0.00035659363194819206,
+      "loss": 3.3584,
+      "step": 37700
+    },
+    {
+      "epoch": 4.070080862533692,
+      "grad_norm": 0.6472448110580444,
+      "learning_rate": 0.00035626983270372367,
+      "loss": 3.3797,
+      "step": 37750
+    },
+    {
+      "epoch": 4.0754716981132075,
+      "grad_norm": 0.6979560256004333,
+      "learning_rate": 0.0003559460334592552,
+      "loss": 3.3631,
+      "step": 37800
+    },
+    {
+      "epoch": 4.080862533692723,
+      "grad_norm": 0.6640351414680481,
+      "learning_rate": 0.0003556222342147868,
+      "loss": 3.359,
+      "step": 37850
+    },
+    {
+      "epoch": 4.086253369272237,
+      "grad_norm": 0.6525617837905884,
+      "learning_rate": 0.0003552984349703184,
+      "loss": 3.359,
+      "step": 37900
+    },
+    {
+      "epoch": 4.091644204851752,
+      "grad_norm": 0.6497607231140137,
+      "learning_rate": 0.00035497463572585,
+      "loss": 3.3734,
+      "step": 37950
+    },
+    {
+      "epoch": 4.097035040431267,
+      "grad_norm": 0.6940391063690186,
+      "learning_rate": 0.00035465083648138153,
+      "loss": 3.3644,
+      "step": 38000
+    },
+    {
+      "epoch": 4.097035040431267,
+      "eval_accuracy": 0.3764235837177563,
+      "eval_loss": 3.4489781856536865,
+      "eval_runtime": 185.0419,
+      "eval_samples_per_second": 97.335,
+      "eval_steps_per_second": 6.085,
+      "step": 38000
+    },
+    {
+      "epoch": 4.1024258760107815,
+      "grad_norm": 0.7142944931983948,
+      "learning_rate": 0.00035432703723691314,
+      "loss": 3.3681,
+      "step": 38050
+    },
+    {
+      "epoch": 4.107816711590297,
+      "grad_norm": 0.6247251629829407,
+      "learning_rate": 0.00035400323799244463,
+      "loss": 3.3706,
+      "step": 38100
+    },
+    {
+      "epoch": 4.113207547169812,
+      "grad_norm": 0.6462594866752625,
+      "learning_rate": 0.0003536794387479762,
+      "loss": 3.3803,
+      "step": 38150
+    },
+    {
+      "epoch": 4.118598382749326,
+      "grad_norm": 0.659847617149353,
+      "learning_rate": 0.0003533556395035078,
+      "loss": 3.3622,
+      "step": 38200
+    },
+    {
+      "epoch": 4.123989218328841,
+      "grad_norm": 0.6913278102874756,
+      "learning_rate": 0.00035303184025903934,
+      "loss": 3.3643,
+      "step": 38250
+    },
+    {
+      "epoch": 4.129380053908355,
+      "grad_norm": 0.6582707166671753,
+      "learning_rate": 0.00035270804101457094,
+      "loss": 3.3768,
+      "step": 38300
+    },
+    {
+      "epoch": 4.1347708894878705,
+      "grad_norm": 0.6678318977355957,
+      "learning_rate": 0.0003523842417701025,
+      "loss": 3.3553,
+      "step": 38350
+    },
+    {
+      "epoch": 4.140161725067386,
+      "grad_norm": 0.6429058909416199,
+      "learning_rate": 0.0003520604425256341,
+      "loss": 3.379,
+      "step": 38400
+    },
+    {
+      "epoch": 4.1455525606469,
+      "grad_norm": 0.6227869391441345,
+      "learning_rate": 0.00035173664328116565,
+      "loss": 3.3626,
+      "step": 38450
+    },
+    {
+      "epoch": 4.150943396226415,
+      "grad_norm": 0.6584265828132629,
+      "learning_rate": 0.00035141284403669725,
+      "loss": 3.3735,
+      "step": 38500
+    },
+    {
+      "epoch": 4.15633423180593,
+      "grad_norm": 0.6800707578659058,
+      "learning_rate": 0.0003510890447922288,
+      "loss": 3.398,
+      "step": 38550
+    },
+    {
+      "epoch": 4.1617250673854445,
+      "grad_norm": 0.6075555086135864,
+      "learning_rate": 0.00035077172153264976,
+      "loss": 3.3731,
+      "step": 38600
+    },
+    {
+      "epoch": 4.16711590296496,
+      "grad_norm": 0.6489547491073608,
+      "learning_rate": 0.0003504479222881813,
+      "loss": 3.3858,
+      "step": 38650
+    },
+    {
+      "epoch": 4.172506738544475,
+      "grad_norm": 0.7513884902000427,
+      "learning_rate": 0.0003501241230437129,
+      "loss": 3.3517,
+      "step": 38700
+    },
+    {
+      "epoch": 4.177897574123989,
+      "grad_norm": 0.6607750058174133,
+      "learning_rate": 0.0003498003237992444,
+      "loss": 3.3763,
+      "step": 38750
+    },
+    {
+      "epoch": 4.183288409703504,
+      "grad_norm": 0.6463075280189514,
+      "learning_rate": 0.000349476524554776,
+      "loss": 3.3807,
+      "step": 38800
+    },
+    {
+      "epoch": 4.188679245283019,
+      "grad_norm": 0.7830615639686584,
+      "learning_rate": 0.00034915272531030756,
+      "loss": 3.3839,
+      "step": 38850
+    },
+    {
+      "epoch": 4.1940700808625335,
+      "grad_norm": 0.6614366769790649,
+      "learning_rate": 0.0003488289260658391,
+      "loss": 3.3912,
+      "step": 38900
+    },
+    {
+      "epoch": 4.199460916442049,
+      "grad_norm": 0.7006166577339172,
+      "learning_rate": 0.0003485051268213707,
+      "loss": 3.3844,
+      "step": 38950
+    },
+    {
+      "epoch": 4.204851752021563,
+      "grad_norm": 0.6897702813148499,
+      "learning_rate": 0.00034818132757690227,
+      "loss": 3.3636,
+      "step": 39000
+    },
+    {
+      "epoch": 4.204851752021563,
+      "eval_accuracy": 0.37732518538084087,
+      "eval_loss": 3.445204734802246,
+      "eval_runtime": 184.9796,
+      "eval_samples_per_second": 97.367,
+      "eval_steps_per_second": 6.087,
+      "step": 39000
+    },
+    {
+      "epoch": 4.210242587601078,
+      "grad_norm": 0.6401898264884949,
+      "learning_rate": 0.0003478575283324339,
+      "loss": 3.3922,
+      "step": 39050
+    },
+    {
+      "epoch": 4.215633423180593,
+      "grad_norm": 0.6454992294311523,
+      "learning_rate": 0.0003475337290879654,
+      "loss": 3.3863,
+      "step": 39100
+    },
+    {
+      "epoch": 4.2210242587601075,
+      "grad_norm": 0.6486014723777771,
+      "learning_rate": 0.00034720992984349703,
+      "loss": 3.395,
+      "step": 39150
+    },
+    {
+      "epoch": 4.226415094339623,
+      "grad_norm": 0.6316015720367432,
+      "learning_rate": 0.0003468861305990286,
+      "loss": 3.3762,
+      "step": 39200
+    },
+    {
+      "epoch": 4.231805929919138,
+      "grad_norm": 0.6304581761360168,
+      "learning_rate": 0.0003465623313545602,
+      "loss": 3.3712,
+      "step": 39250
+    },
+    {
+      "epoch": 4.237196765498652,
+      "grad_norm": 0.6496357917785645,
+      "learning_rate": 0.00034623853211009173,
+      "loss": 3.3843,
+      "step": 39300
+    },
+    {
+      "epoch": 4.242587601078167,
+      "grad_norm": 0.6726136207580566,
+      "learning_rate": 0.00034591473286562323,
+      "loss": 3.3732,
+      "step": 39350
+    },
+    {
+      "epoch": 4.247978436657682,
+      "grad_norm": 0.6343677043914795,
+      "learning_rate": 0.0003455909336211549,
+      "loss": 3.3922,
+      "step": 39400
+    },
+    {
+      "epoch": 4.2533692722371965,
+      "grad_norm": 0.6977464556694031,
+      "learning_rate": 0.0003452671343766864,
+      "loss": 3.3806,
+      "step": 39450
+    },
+    {
+      "epoch": 4.258760107816712,
+      "grad_norm": 0.6857298016548157,
+      "learning_rate": 0.000344943335132218,
+      "loss": 3.3852,
+      "step": 39500
+    },
+    {
+      "epoch": 4.264150943396227,
+      "grad_norm": 0.6643324494361877,
+      "learning_rate": 0.00034461953588774954,
+      "loss": 3.3899,
+      "step": 39550
+    },
+    {
+      "epoch": 4.269541778975741,
+      "grad_norm": 0.6354033946990967,
+      "learning_rate": 0.00034429573664328115,
+      "loss": 3.3981,
+      "step": 39600
+    },
+    {
+      "epoch": 4.274932614555256,
+      "grad_norm": 0.6460668444633484,
+      "learning_rate": 0.0003439719373988127,
+      "loss": 3.3938,
+      "step": 39650
+    },
+    {
+      "epoch": 4.280323450134771,
+      "grad_norm": 0.7137424945831299,
+      "learning_rate": 0.00034364813815434425,
+      "loss": 3.3745,
+      "step": 39700
+    },
+    {
+      "epoch": 4.285714285714286,
+      "grad_norm": 0.7198200821876526,
+      "learning_rate": 0.00034332433890987585,
+      "loss": 3.3825,
+      "step": 39750
+    },
+    {
+      "epoch": 4.291105121293801,
+      "grad_norm": 0.6877724528312683,
+      "learning_rate": 0.0003430005396654074,
+      "loss": 3.383,
+      "step": 39800
+    },
+    {
+      "epoch": 4.296495956873315,
+      "grad_norm": 0.6581820249557495,
+      "learning_rate": 0.000342676740420939,
+      "loss": 3.3829,
+      "step": 39850
+    },
+    {
+      "epoch": 4.30188679245283,
+      "grad_norm": 0.7357942461967468,
+      "learning_rate": 0.00034235294117647056,
+      "loss": 3.3896,
+      "step": 39900
+    },
+    {
+      "epoch": 4.307277628032345,
+      "grad_norm": 0.6755037903785706,
+      "learning_rate": 0.00034202914193200216,
+      "loss": 3.3776,
+      "step": 39950
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "grad_norm": 0.604346752166748,
+      "learning_rate": 0.0003417053426875337,
+      "loss": 3.3899,
+      "step": 40000
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "eval_accuracy": 0.37746034957328206,
+      "eval_loss": 3.44191312789917,
+      "eval_runtime": 185.1596,
+      "eval_samples_per_second": 97.273,
+      "eval_steps_per_second": 6.081,
+      "step": 40000
+    },
+    {
+      "epoch": 4.318059299191375,
+      "grad_norm": 0.7762167453765869,
+      "learning_rate": 0.0003413815434430653,
+      "loss": 3.3884,
+      "step": 40050
+    },
+    {
+      "epoch": 4.32345013477089,
+      "grad_norm": 0.7179322242736816,
+      "learning_rate": 0.0003410577441985968,
+      "loss": 3.3853,
+      "step": 40100
+    },
+    {
+      "epoch": 4.328840970350404,
+      "grad_norm": 0.6719879508018494,
+      "learning_rate": 0.00034073394495412837,
+      "loss": 3.3954,
+      "step": 40150
+    },
+    {
+      "epoch": 4.334231805929919,
+      "grad_norm": 0.705895185470581,
+      "learning_rate": 0.00034041014570965997,
+      "loss": 3.3915,
+      "step": 40200
+    },
+    {
+      "epoch": 4.339622641509434,
+      "grad_norm": 0.6576878428459167,
+      "learning_rate": 0.0003400863464651915,
+      "loss": 3.378,
+      "step": 40250
+    },
+    {
+      "epoch": 4.345013477088949,
+      "grad_norm": 0.6394710540771484,
+      "learning_rate": 0.0003397625472207231,
+      "loss": 3.3928,
+      "step": 40300
+    },
+    {
+      "epoch": 4.350404312668464,
+      "grad_norm": 0.6182326078414917,
+      "learning_rate": 0.0003394387479762547,
+      "loss": 3.3743,
+      "step": 40350
+    },
+    {
+      "epoch": 4.355795148247978,
+      "grad_norm": 0.6548103094100952,
+      "learning_rate": 0.0003391149487317863,
+      "loss": 3.3867,
+      "step": 40400
+    },
+    {
+      "epoch": 4.361185983827493,
+      "grad_norm": 0.6793950796127319,
+      "learning_rate": 0.00033879114948731783,
+      "loss": 3.3844,
+      "step": 40450
+    },
+    {
+      "epoch": 4.366576819407008,
+      "grad_norm": 0.6479836702346802,
+      "learning_rate": 0.00033846735024284944,
+      "loss": 3.3977,
+      "step": 40500
+    },
+    {
+      "epoch": 4.3719676549865225,
+      "grad_norm": 0.6476645469665527,
+      "learning_rate": 0.000338143550998381,
+      "loss": 3.3889,
+      "step": 40550
+    },
+    {
+      "epoch": 4.377358490566038,
+      "grad_norm": 0.6995546221733093,
+      "learning_rate": 0.00033781975175391254,
+      "loss": 3.3794,
+      "step": 40600
+    },
+    {
+      "epoch": 4.382749326145553,
+      "grad_norm": 0.7507542371749878,
+      "learning_rate": 0.00033749595250944414,
+      "loss": 3.3889,
+      "step": 40650
+    },
+    {
+      "epoch": 4.388140161725067,
+      "grad_norm": 0.7036803364753723,
+      "learning_rate": 0.00033717215326497564,
+      "loss": 3.3894,
+      "step": 40700
+    },
+    {
+      "epoch": 4.393530997304582,
+      "grad_norm": 0.6556727886199951,
+      "learning_rate": 0.0003368483540205073,
+      "loss": 3.3865,
+      "step": 40750
+    },
+    {
+      "epoch": 4.398921832884097,
+      "grad_norm": 0.6941169500350952,
+      "learning_rate": 0.0003365245547760388,
+      "loss": 3.3765,
+      "step": 40800
+    },
+    {
+      "epoch": 4.404312668463612,
+      "grad_norm": 0.6608136892318726,
+      "learning_rate": 0.0003362007555315704,
+      "loss": 3.3915,
+      "step": 40850
+    },
+    {
+      "epoch": 4.409703504043127,
+      "grad_norm": 0.8930730223655701,
+      "learning_rate": 0.00033587695628710195,
+      "loss": 3.3998,
+      "step": 40900
+    },
+    {
+      "epoch": 4.415094339622642,
+      "grad_norm": 0.6588183045387268,
+      "learning_rate": 0.00033555315704263355,
+      "loss": 3.3953,
+      "step": 40950
+    },
+    {
+      "epoch": 4.420485175202156,
+      "grad_norm": 0.6629270315170288,
+      "learning_rate": 0.0003352293577981651,
+      "loss": 3.3969,
+      "step": 41000
+    },
+    {
+      "epoch": 4.420485175202156,
+      "eval_accuracy": 0.3781523598157725,
+      "eval_loss": 3.4341278076171875,
+      "eval_runtime": 185.2042,
+      "eval_samples_per_second": 97.249,
+      "eval_steps_per_second": 6.08,
+      "step": 41000
+    },
+    {
+      "epoch": 4.425876010781671,
+      "grad_norm": 0.6934705972671509,
+      "learning_rate": 0.00033490555855369665,
+      "loss": 3.402,
+      "step": 41050
+    },
+    {
+      "epoch": 4.431266846361186,
+      "grad_norm": 0.608988881111145,
+      "learning_rate": 0.00033458175930922826,
+      "loss": 3.3712,
+      "step": 41100
+    },
+    {
+      "epoch": 4.436657681940701,
+      "grad_norm": 0.6706856489181519,
+      "learning_rate": 0.0003342579600647598,
+      "loss": 3.3936,
+      "step": 41150
+    },
+    {
+      "epoch": 4.442048517520216,
+      "grad_norm": 0.6479840874671936,
+      "learning_rate": 0.0003339341608202914,
+      "loss": 3.4055,
+      "step": 41200
+    },
+    {
+      "epoch": 4.44743935309973,
+      "grad_norm": 0.680960476398468,
+      "learning_rate": 0.00033361036157582297,
+      "loss": 3.4043,
+      "step": 41250
+    },
+    {
+      "epoch": 4.452830188679245,
+      "grad_norm": 0.6902644634246826,
+      "learning_rate": 0.00033328656233135457,
+      "loss": 3.3727,
+      "step": 41300
+    },
+    {
+      "epoch": 4.45822102425876,
+      "grad_norm": 0.6928199529647827,
+      "learning_rate": 0.0003329627630868861,
+      "loss": 3.379,
+      "step": 41350
+    },
+    {
+      "epoch": 4.463611859838275,
+      "grad_norm": 0.7097663879394531,
+      "learning_rate": 0.0003326389638424177,
+      "loss": 3.3974,
+      "step": 41400
+    },
+    {
+      "epoch": 4.46900269541779,
+      "grad_norm": 0.6607456803321838,
+      "learning_rate": 0.0003323151645979492,
+      "loss": 3.4084,
+      "step": 41450
+    },
+    {
+      "epoch": 4.474393530997305,
+      "grad_norm": 0.7242568731307983,
+      "learning_rate": 0.00033199136535348077,
+      "loss": 3.4067,
+      "step": 41500
+    },
+    {
+      "epoch": 4.479784366576819,
+      "grad_norm": 0.6340367197990417,
+      "learning_rate": 0.0003316675661090124,
+      "loss": 3.3938,
+      "step": 41550
+    },
+    {
+      "epoch": 4.485175202156334,
+      "grad_norm": 0.6760000586509705,
+      "learning_rate": 0.00033134376686454393,
+      "loss": 3.3863,
+      "step": 41600
+    },
+    {
+      "epoch": 4.490566037735849,
+      "grad_norm": 0.6576944589614868,
+      "learning_rate": 0.00033101996762007553,
+      "loss": 3.3976,
+      "step": 41650
+    },
+    {
+      "epoch": 4.495956873315364,
+      "grad_norm": 0.6730291247367859,
+      "learning_rate": 0.0003306961683756071,
+      "loss": 3.3886,
+      "step": 41700
+    },
+    {
+      "epoch": 4.501347708894879,
+      "grad_norm": 0.685353696346283,
+      "learning_rate": 0.0003303723691311387,
+      "loss": 3.387,
+      "step": 41750
+    },
+    {
+      "epoch": 4.506738544474393,
+      "grad_norm": 0.9739151000976562,
+      "learning_rate": 0.00033004856988667024,
+      "loss": 3.4041,
+      "step": 41800
+    },
+    {
+      "epoch": 4.512129380053908,
+      "grad_norm": 0.6757647395133972,
+      "learning_rate": 0.0003297247706422018,
+      "loss": 3.3808,
+      "step": 41850
+    },
+    {
+      "epoch": 4.517520215633423,
+      "grad_norm": 0.6179267168045044,
+      "learning_rate": 0.0003294009713977334,
+      "loss": 3.3965,
+      "step": 41900
+    },
+    {
+      "epoch": 4.5229110512129385,
+      "grad_norm": 0.677100658416748,
+      "learning_rate": 0.00032907717215326494,
+      "loss": 3.3982,
+      "step": 41950
+    },
+    {
+      "epoch": 4.528301886792453,
+      "grad_norm": 0.6717151999473572,
+      "learning_rate": 0.00032875337290879655,
+      "loss": 3.3848,
+      "step": 42000
+    },
+    {
+      "epoch": 4.528301886792453,
+      "eval_accuracy": 0.37859935779624126,
+      "eval_loss": 3.430816888809204,
+      "eval_runtime": 185.1614,
+      "eval_samples_per_second": 97.272,
+      "eval_steps_per_second": 6.081,
+      "step": 42000
+    },
+    {
+      "epoch": 4.533692722371968,
+      "grad_norm": 0.647710919380188,
+      "learning_rate": 0.00032842957366432805,
+      "loss": 3.3855,
+      "step": 42050
+    },
+    {
+      "epoch": 4.539083557951482,
+      "grad_norm": 0.7008846998214722,
+      "learning_rate": 0.000328112250404749,
+      "loss": 3.3938,
+      "step": 42100
+    },
+    {
+      "epoch": 4.544474393530997,
+      "grad_norm": 0.6753333806991577,
+      "learning_rate": 0.00032778845116028066,
+      "loss": 3.3952,
+      "step": 42150
+    },
+    {
+      "epoch": 4.549865229110512,
+      "grad_norm": 0.6874195337295532,
+      "learning_rate": 0.00032746465191581215,
+      "loss": 3.3911,
+      "step": 42200
+    },
+    {
+      "epoch": 4.555256064690027,
+      "grad_norm": 0.6420570611953735,
+      "learning_rate": 0.0003271408526713437,
+      "loss": 3.4027,
+      "step": 42250
+    },
+    {
+      "epoch": 4.560646900269542,
+      "grad_norm": 0.7065884470939636,
+      "learning_rate": 0.0003268170534268753,
+      "loss": 3.3818,
+      "step": 42300
+    },
+    {
+      "epoch": 4.566037735849057,
+      "grad_norm": 0.6392266750335693,
+      "learning_rate": 0.00032649325418240686,
+      "loss": 3.4136,
+      "step": 42350
+    },
+    {
+      "epoch": 4.571428571428571,
+      "grad_norm": 0.662189781665802,
+      "learning_rate": 0.00032616945493793846,
+      "loss": 3.374,
+      "step": 42400
+    },
+    {
+      "epoch": 4.576819407008086,
+      "grad_norm": 0.6861265301704407,
+      "learning_rate": 0.00032584565569347,
+      "loss": 3.3951,
+      "step": 42450
+    },
+    {
+      "epoch": 4.5822102425876015,
+      "grad_norm": 0.6285902857780457,
+      "learning_rate": 0.0003255218564490016,
+      "loss": 3.4016,
+      "step": 42500
+    },
+    {
+      "epoch": 4.587601078167116,
+      "grad_norm": 0.6859892010688782,
+      "learning_rate": 0.00032519805720453317,
+      "loss": 3.3806,
+      "step": 42550
+    },
+    {
+      "epoch": 4.592991913746631,
+      "grad_norm": 0.6380650401115417,
+      "learning_rate": 0.0003248742579600647,
+      "loss": 3.4063,
+      "step": 42600
+    },
+    {
+      "epoch": 4.598382749326145,
+      "grad_norm": 0.6524938344955444,
+      "learning_rate": 0.0003245504587155963,
+      "loss": 3.3911,
+      "step": 42650
+    },
+    {
+      "epoch": 4.60377358490566,
+      "grad_norm": 0.6308937072753906,
+      "learning_rate": 0.0003242266594711278,
+      "loss": 3.3892,
+      "step": 42700
+    },
+    {
+      "epoch": 4.609164420485175,
+      "grad_norm": 0.648926854133606,
+      "learning_rate": 0.0003239028602266595,
+      "loss": 3.3987,
+      "step": 42750
+    },
+    {
+      "epoch": 4.6145552560646905,
+      "grad_norm": 0.6324000954627991,
+      "learning_rate": 0.000323579060982191,
+      "loss": 3.3729,
+      "step": 42800
+    },
+    {
+      "epoch": 4.619946091644205,
+      "grad_norm": 0.6731909513473511,
+      "learning_rate": 0.0003232552617377226,
+      "loss": 3.4003,
+      "step": 42850
+    },
+    {
+      "epoch": 4.62533692722372,
+      "grad_norm": 0.6357996463775635,
+      "learning_rate": 0.00032293146249325413,
+      "loss": 3.3966,
+      "step": 42900
+    },
+    {
+      "epoch": 4.630727762803234,
+      "grad_norm": 0.7101162075996399,
+      "learning_rate": 0.00032260766324878574,
+      "loss": 3.3918,
+      "step": 42950
+    },
+    {
+      "epoch": 4.636118598382749,
+      "grad_norm": 0.7121390104293823,
+      "learning_rate": 0.0003222838640043173,
+      "loss": 3.3952,
+      "step": 43000
+    },
+    {
+      "epoch": 4.636118598382749,
+      "eval_accuracy": 0.3787855888459553,
+      "eval_loss": 3.424402952194214,
+      "eval_runtime": 185.282,
+      "eval_samples_per_second": 97.209,
+      "eval_steps_per_second": 6.077,
+      "step": 43000
+    },
+    {
+      "epoch": 4.6415094339622645,
+      "grad_norm": 0.6777333617210388,
+      "learning_rate": 0.00032196006475984884,
+      "loss": 3.3872,
+      "step": 43050
+    },
+    {
+      "epoch": 4.646900269541779,
+      "grad_norm": 0.6901153922080994,
+      "learning_rate": 0.00032163626551538044,
+      "loss": 3.3889,
+      "step": 43100
+    },
+    {
+      "epoch": 4.652291105121294,
+      "grad_norm": 0.6530250310897827,
+      "learning_rate": 0.000321312466270912,
+      "loss": 3.4072,
+      "step": 43150
+    },
+    {
+      "epoch": 4.657681940700809,
+      "grad_norm": 0.7110807299613953,
+      "learning_rate": 0.0003209886670264436,
+      "loss": 3.3743,
+      "step": 43200
+    },
+    {
+      "epoch": 4.663072776280323,
+      "grad_norm": 0.648343563079834,
+      "learning_rate": 0.00032066486778197515,
+      "loss": 3.396,
+      "step": 43250
+    },
+    {
+      "epoch": 4.668463611859838,
+      "grad_norm": 0.6666106581687927,
+      "learning_rate": 0.00032034106853750675,
+      "loss": 3.3839,
+      "step": 43300
+    },
+    {
+      "epoch": 4.6738544474393535,
+      "grad_norm": 0.7402759194374084,
+      "learning_rate": 0.0003200172692930383,
+      "loss": 3.3912,
+      "step": 43350
+    },
+    {
+      "epoch": 4.679245283018868,
+      "grad_norm": 0.6613388061523438,
+      "learning_rate": 0.0003196934700485699,
+      "loss": 3.3894,
+      "step": 43400
+    },
+    {
+      "epoch": 4.684636118598383,
+      "grad_norm": 0.6671817898750305,
+      "learning_rate": 0.0003193696708041014,
+      "loss": 3.3974,
+      "step": 43450
+    },
+    {
+      "epoch": 4.690026954177897,
+      "grad_norm": 0.6724131107330322,
+      "learning_rate": 0.00031904587155963296,
+      "loss": 3.3766,
+      "step": 43500
+    },
+    {
+      "epoch": 4.695417789757412,
+      "grad_norm": 0.6639029383659363,
+      "learning_rate": 0.00031872207231516456,
+      "loss": 3.3921,
+      "step": 43550
+    },
+    {
+      "epoch": 4.7008086253369274,
+      "grad_norm": 0.6928759217262268,
+      "learning_rate": 0.0003183982730706961,
+      "loss": 3.3984,
+      "step": 43600
+    },
+    {
+      "epoch": 4.706199460916442,
+      "grad_norm": 0.6988749504089355,
+      "learning_rate": 0.0003180744738262277,
+      "loss": 3.3954,
+      "step": 43650
+    },
+    {
+      "epoch": 4.711590296495957,
+      "grad_norm": 0.675839900970459,
+      "learning_rate": 0.00031775067458175927,
+      "loss": 3.3914,
+      "step": 43700
+    },
+    {
+      "epoch": 4.716981132075472,
+      "grad_norm": 0.6638389229774475,
+      "learning_rate": 0.00031742687533729087,
+      "loss": 3.4078,
+      "step": 43750
+    },
+    {
+      "epoch": 4.722371967654986,
+      "grad_norm": 0.6970890164375305,
+      "learning_rate": 0.0003171030760928224,
+      "loss": 3.3917,
+      "step": 43800
+    },
+    {
+      "epoch": 4.727762803234501,
+      "grad_norm": 0.6743645071983337,
+      "learning_rate": 0.000316779276848354,
+      "loss": 3.399,
+      "step": 43850
+    },
+    {
+      "epoch": 4.7331536388140165,
+      "grad_norm": 0.6509005427360535,
+      "learning_rate": 0.0003164554776038856,
+      "loss": 3.3884,
+      "step": 43900
+    },
+    {
+      "epoch": 4.738544474393531,
+      "grad_norm": 0.6444371938705444,
+      "learning_rate": 0.00031613167835941713,
+      "loss": 3.3879,
+      "step": 43950
+    },
+    {
+      "epoch": 4.743935309973046,
+      "grad_norm": 0.6728962063789368,
+      "learning_rate": 0.00031580787911494873,
+      "loss": 3.3875,
+      "step": 44000
+    },
+    {
+      "epoch": 4.743935309973046,
+      "eval_accuracy": 0.37932092162421727,
+      "eval_loss": 3.421056032180786,
+      "eval_runtime": 185.3808,
+      "eval_samples_per_second": 97.157,
+      "eval_steps_per_second": 6.074,
+      "step": 44000
+    },
+    {
+      "epoch": 4.74932614555256,
+      "grad_norm": 0.6364440321922302,
+      "learning_rate": 0.00031548407987048023,
+      "loss": 3.3854,
+      "step": 44050
+    },
+    {
+      "epoch": 4.754716981132075,
+      "grad_norm": 0.7157145738601685,
+      "learning_rate": 0.0003151602806260119,
+      "loss": 3.3942,
+      "step": 44100
+    },
+    {
+      "epoch": 4.7601078167115904,
+      "grad_norm": 0.7098649144172668,
+      "learning_rate": 0.0003148364813815434,
+      "loss": 3.3875,
+      "step": 44150
+    },
+    {
+      "epoch": 4.765498652291106,
+      "grad_norm": 0.691125214099884,
+      "learning_rate": 0.000314512682137075,
+      "loss": 3.3911,
+      "step": 44200
+    },
+    {
+      "epoch": 4.77088948787062,
+      "grad_norm": 0.6948944330215454,
+      "learning_rate": 0.00031418888289260654,
+      "loss": 3.3756,
+      "step": 44250
+    },
+    {
+      "epoch": 4.776280323450135,
+      "grad_norm": 0.6892076134681702,
+      "learning_rate": 0.0003138650836481381,
+      "loss": 3.3961,
+      "step": 44300
+    },
+    {
+      "epoch": 4.781671159029649,
+      "grad_norm": 0.7187123894691467,
+      "learning_rate": 0.0003135412844036697,
+      "loss": 3.3782,
+      "step": 44350
+    },
+    {
+      "epoch": 4.787061994609164,
+      "grad_norm": 0.6381692886352539,
+      "learning_rate": 0.00031321748515920124,
+      "loss": 3.4169,
+      "step": 44400
+    },
+    {
+      "epoch": 4.7924528301886795,
+      "grad_norm": 0.6556568145751953,
+      "learning_rate": 0.00031289368591473285,
+      "loss": 3.3752,
+      "step": 44450
+    },
+    {
+      "epoch": 4.797843665768194,
+      "grad_norm": 0.6677922010421753,
+      "learning_rate": 0.0003125698866702644,
+      "loss": 3.3898,
+      "step": 44500
+    },
+    {
+      "epoch": 4.803234501347709,
+      "grad_norm": 0.6631757616996765,
+      "learning_rate": 0.000312246087425796,
+      "loss": 3.3838,
+      "step": 44550
+    },
+    {
+      "epoch": 4.808625336927224,
+      "grad_norm": 0.6678897738456726,
+      "learning_rate": 0.00031192228818132756,
+      "loss": 3.3972,
+      "step": 44600
+    },
+    {
+      "epoch": 4.814016172506738,
+      "grad_norm": 0.7124047875404358,
+      "learning_rate": 0.00031159848893685916,
+      "loss": 3.3808,
+      "step": 44650
+    },
+    {
+      "epoch": 4.819407008086253,
+      "grad_norm": 0.6782823801040649,
+      "learning_rate": 0.0003112746896923907,
+      "loss": 3.3945,
+      "step": 44700
+    },
+    {
+      "epoch": 4.824797843665769,
+      "grad_norm": 0.687823474407196,
+      "learning_rate": 0.0003109508904479222,
+      "loss": 3.3879,
+      "step": 44750
+    },
+    {
+      "epoch": 4.830188679245283,
+      "grad_norm": 0.7118351459503174,
+      "learning_rate": 0.0003106270912034538,
+      "loss": 3.3857,
+      "step": 44800
+    },
+    {
+      "epoch": 4.835579514824798,
+      "grad_norm": 0.6740395426750183,
+      "learning_rate": 0.00031030329195898536,
+      "loss": 3.3894,
+      "step": 44850
+    },
+    {
+      "epoch": 4.840970350404312,
+      "grad_norm": 0.6710334420204163,
+      "learning_rate": 0.00030997949271451697,
+      "loss": 3.392,
+      "step": 44900
+    },
+    {
+      "epoch": 4.846361185983827,
+      "grad_norm": 0.6429509520530701,
+      "learning_rate": 0.0003096556934700485,
+      "loss": 3.3808,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "grad_norm": 0.6663042306900024,
+      "learning_rate": 0.00030933837021046947,
+      "loss": 3.4071,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "eval_accuracy": 0.3799037357144541,
+      "eval_loss": 3.415198802947998,
+      "eval_runtime": 185.3679,
+      "eval_samples_per_second": 97.164,
+      "eval_steps_per_second": 6.074,
+      "step": 45000
+    },
+    {
+      "epoch": 4.857142857142857,
+      "grad_norm": 0.7604882717132568,
+      "learning_rate": 0.0003090145709660011,
+      "loss": 3.377,
+      "step": 45050
+    },
+    {
+      "epoch": 4.862533692722372,
+      "grad_norm": 0.694174587726593,
+      "learning_rate": 0.0003086907717215326,
+      "loss": 3.3797,
+      "step": 45100
+    },
+    {
+      "epoch": 4.867924528301887,
+      "grad_norm": 0.690619945526123,
+      "learning_rate": 0.0003083669724770642,
+      "loss": 3.391,
+      "step": 45150
+    },
+    {
+      "epoch": 4.873315363881401,
+      "grad_norm": 0.7311848402023315,
+      "learning_rate": 0.0003080431732325958,
+      "loss": 3.3779,
+      "step": 45200
+    },
+    {
+      "epoch": 4.878706199460916,
+      "grad_norm": 0.7099789381027222,
+      "learning_rate": 0.00030771937398812733,
+      "loss": 3.3939,
+      "step": 45250
+    },
+    {
+      "epoch": 4.884097035040432,
+      "grad_norm": 0.7288787961006165,
+      "learning_rate": 0.00030739557474365894,
+      "loss": 3.3783,
+      "step": 45300
+    },
+    {
+      "epoch": 4.889487870619946,
+      "grad_norm": 0.7027914524078369,
+      "learning_rate": 0.0003070717754991905,
+      "loss": 3.3861,
+      "step": 45350
+    },
+    {
+      "epoch": 4.894878706199461,
+      "grad_norm": 0.6941841244697571,
+      "learning_rate": 0.0003067479762547221,
+      "loss": 3.3837,
+      "step": 45400
+    },
+    {
+      "epoch": 4.900269541778976,
+      "grad_norm": 0.756320059299469,
+      "learning_rate": 0.0003064241770102536,
+      "loss": 3.3929,
+      "step": 45450
+    },
+    {
+      "epoch": 4.90566037735849,
+      "grad_norm": 0.6624318361282349,
+      "learning_rate": 0.00030610037776578514,
+      "loss": 3.3948,
+      "step": 45500
+    },
+    {
+      "epoch": 4.9110512129380055,
+      "grad_norm": 0.6968894600868225,
+      "learning_rate": 0.00030577657852131674,
+      "loss": 3.3916,
+      "step": 45550
+    },
+    {
+      "epoch": 4.916442048517521,
+      "grad_norm": 0.6658734083175659,
+      "learning_rate": 0.0003054527792768483,
+      "loss": 3.3918,
+      "step": 45600
+    },
+    {
+      "epoch": 4.921832884097035,
+      "grad_norm": 0.6801126003265381,
+      "learning_rate": 0.0003051289800323799,
+      "loss": 3.4074,
+      "step": 45650
+    },
+    {
+      "epoch": 4.92722371967655,
+      "grad_norm": 0.679326057434082,
+      "learning_rate": 0.00030480518078791145,
+      "loss": 3.401,
+      "step": 45700
+    },
+    {
+      "epoch": 4.932614555256064,
+      "grad_norm": 0.6885223984718323,
+      "learning_rate": 0.00030448138154344305,
+      "loss": 3.3811,
+      "step": 45750
+    },
+    {
+      "epoch": 4.938005390835579,
+      "grad_norm": 0.7149697542190552,
+      "learning_rate": 0.0003041575822989746,
+      "loss": 3.4065,
+      "step": 45800
+    },
+    {
+      "epoch": 4.943396226415095,
+      "grad_norm": 0.6741778254508972,
+      "learning_rate": 0.0003038337830545062,
+      "loss": 3.3787,
+      "step": 45850
+    },
+    {
+      "epoch": 4.948787061994609,
+      "grad_norm": 0.6654149889945984,
+      "learning_rate": 0.00030350998381003776,
+      "loss": 3.3748,
+      "step": 45900
+    },
+    {
+      "epoch": 4.954177897574124,
+      "grad_norm": 0.682928740978241,
+      "learning_rate": 0.0003031861845655693,
+      "loss": 3.3726,
+      "step": 45950
+    },
+    {
+      "epoch": 4.959568733153639,
+      "grad_norm": 0.7227015495300293,
+      "learning_rate": 0.0003028623853211009,
+      "loss": 3.3742,
+      "step": 46000
+    },
+    {
+      "epoch": 4.959568733153639,
+      "eval_accuracy": 0.3803156388121588,
+      "eval_loss": 3.4095218181610107,
+      "eval_runtime": 185.3393,
+      "eval_samples_per_second": 97.179,
+      "eval_steps_per_second": 6.075,
+      "step": 46000
+    },
+    {
+      "epoch": 4.964959568733153,
+      "grad_norm": 0.6969720721244812,
+      "learning_rate": 0.00030253858607663247,
+      "loss": 3.3797,
+      "step": 46050
+    },
+    {
+      "epoch": 4.9703504043126685,
+      "grad_norm": 0.7063660621643066,
+      "learning_rate": 0.00030221478683216407,
+      "loss": 3.3897,
+      "step": 46100
+    },
+    {
+      "epoch": 4.975741239892184,
+      "grad_norm": 0.6650516986846924,
+      "learning_rate": 0.00030189098758769557,
+      "loss": 3.3888,
+      "step": 46150
+    },
+    {
+      "epoch": 4.981132075471698,
+      "grad_norm": 0.6772712469100952,
+      "learning_rate": 0.00030156718834322717,
+      "loss": 3.3944,
+      "step": 46200
+    },
+    {
+      "epoch": 4.986522911051213,
+      "grad_norm": 0.7059131860733032,
+      "learning_rate": 0.0003012433890987587,
+      "loss": 3.392,
+      "step": 46250
+    },
+    {
+      "epoch": 4.991913746630727,
+      "grad_norm": 0.7540570497512817,
+      "learning_rate": 0.0003009195898542903,
+      "loss": 3.3841,
+      "step": 46300
+    },
+    {
+      "epoch": 4.997304582210242,
+      "grad_norm": 0.7052427530288696,
+      "learning_rate": 0.0003005957906098219,
+      "loss": 3.3774,
+      "step": 46350
+    },
+    {
+      "epoch": 5.002695417789758,
+      "grad_norm": 0.7645816206932068,
+      "learning_rate": 0.00030027199136535343,
+      "loss": 3.3627,
+      "step": 46400
+    },
+    {
+      "epoch": 5.008086253369272,
+      "grad_norm": 0.6885704398155212,
+      "learning_rate": 0.00029994819212088503,
+      "loss": 3.2878,
+      "step": 46450
+    },
+    {
+      "epoch": 5.013477088948787,
+      "grad_norm": 1.085796594619751,
+      "learning_rate": 0.0002996243928764166,
+      "loss": 3.2952,
+      "step": 46500
+    },
+    {
+      "epoch": 5.018867924528302,
+      "grad_norm": 0.6787880659103394,
+      "learning_rate": 0.0002993005936319482,
+      "loss": 3.3063,
+      "step": 46550
+    },
+    {
+      "epoch": 5.024258760107816,
+      "grad_norm": 0.7390757203102112,
+      "learning_rate": 0.00029897679438747974,
+      "loss": 3.2968,
+      "step": 46600
+    },
+    {
+      "epoch": 5.0296495956873315,
+      "grad_norm": 0.7662341594696045,
+      "learning_rate": 0.0002986529951430113,
+      "loss": 3.2848,
+      "step": 46650
+    },
+    {
+      "epoch": 5.035040431266847,
+      "grad_norm": 0.6801024675369263,
+      "learning_rate": 0.0002983291958985429,
+      "loss": 3.2963,
+      "step": 46700
+    },
+    {
+      "epoch": 5.040431266846361,
+      "grad_norm": 0.6480717658996582,
+      "learning_rate": 0.00029800539665407444,
+      "loss": 3.2874,
+      "step": 46750
+    },
+    {
+      "epoch": 5.045822102425876,
+      "grad_norm": 0.7626944780349731,
+      "learning_rate": 0.000297681597409606,
+      "loss": 3.2959,
+      "step": 46800
+    },
+    {
+      "epoch": 5.051212938005391,
+      "grad_norm": 0.6716665625572205,
+      "learning_rate": 0.0002973577981651376,
+      "loss": 3.3083,
+      "step": 46850
+    },
+    {
+      "epoch": 5.056603773584905,
+      "grad_norm": 0.6842003464698792,
+      "learning_rate": 0.00029703399892066915,
+      "loss": 3.316,
+      "step": 46900
+    },
+    {
+      "epoch": 5.061994609164421,
+      "grad_norm": 0.6711278557777405,
+      "learning_rate": 0.00029671019967620076,
+      "loss": 3.3085,
+      "step": 46950
+    },
+    {
+      "epoch": 5.067385444743936,
+      "grad_norm": 0.6872760653495789,
+      "learning_rate": 0.0002963864004317323,
+      "loss": 3.3158,
+      "step": 47000
+    },
+    {
+      "epoch": 5.067385444743936,
+      "eval_accuracy": 0.380729606314732,
+      "eval_loss": 3.410980463027954,
+      "eval_runtime": 185.6161,
+      "eval_samples_per_second": 97.034,
+      "eval_steps_per_second": 6.066,
+      "step": 47000
+    },
+    {
+      "epoch": 5.07277628032345,
+      "grad_norm": 0.7116502523422241,
+      "learning_rate": 0.00029606260118726386,
+      "loss": 3.2866,
+      "step": 47050
+    },
+    {
+      "epoch": 5.078167115902965,
+      "grad_norm": 0.6567325592041016,
+      "learning_rate": 0.0002957388019427954,
+      "loss": 3.3009,
+      "step": 47100
+    },
+    {
+      "epoch": 5.083557951482479,
+      "grad_norm": 0.7238079309463501,
+      "learning_rate": 0.000295415002698327,
+      "loss": 3.3246,
+      "step": 47150
+    },
+    {
+      "epoch": 5.0889487870619945,
+      "grad_norm": 0.7564477324485779,
+      "learning_rate": 0.00029509120345385856,
+      "loss": 3.3181,
+      "step": 47200
+    },
+    {
+      "epoch": 5.09433962264151,
+      "grad_norm": 0.7536827921867371,
+      "learning_rate": 0.00029476740420939017,
+      "loss": 3.2986,
+      "step": 47250
+    },
+    {
+      "epoch": 5.099730458221024,
+      "grad_norm": 0.7181034684181213,
+      "learning_rate": 0.0002944436049649217,
+      "loss": 3.3121,
+      "step": 47300
+    },
+    {
+      "epoch": 5.105121293800539,
+      "grad_norm": 0.7820267081260681,
+      "learning_rate": 0.00029412628170534267,
+      "loss": 3.3151,
+      "step": 47350
+    },
+    {
+      "epoch": 5.110512129380054,
+      "grad_norm": 0.676846444606781,
+      "learning_rate": 0.0002938024824608742,
+      "loss": 3.3098,
+      "step": 47400
+    },
+    {
+      "epoch": 5.115902964959568,
+      "grad_norm": 0.7352178692817688,
+      "learning_rate": 0.00029347868321640577,
+      "loss": 3.3171,
+      "step": 47450
+    },
+    {
+      "epoch": 5.121293800539084,
+      "grad_norm": 0.7145463824272156,
+      "learning_rate": 0.0002931548839719374,
+      "loss": 3.3112,
+      "step": 47500
+    },
+    {
+      "epoch": 5.126684636118599,
+      "grad_norm": 0.6998820900917053,
+      "learning_rate": 0.0002928310847274689,
+      "loss": 3.3068,
+      "step": 47550
+    },
+    {
+      "epoch": 5.132075471698113,
+      "grad_norm": 0.6971082091331482,
+      "learning_rate": 0.00029250728548300053,
+      "loss": 3.3033,
+      "step": 47600
+    },
+    {
+      "epoch": 5.137466307277628,
+      "grad_norm": 0.6818746328353882,
+      "learning_rate": 0.0002921834862385321,
+      "loss": 3.3174,
+      "step": 47650
+    },
+    {
+      "epoch": 5.142857142857143,
+      "grad_norm": 0.6603133678436279,
+      "learning_rate": 0.0002918596869940637,
+      "loss": 3.32,
+      "step": 47700
+    },
+    {
+      "epoch": 5.1482479784366575,
+      "grad_norm": 0.720582902431488,
+      "learning_rate": 0.0002915358877495952,
+      "loss": 3.3214,
+      "step": 47750
+    },
+    {
+      "epoch": 5.153638814016173,
+      "grad_norm": 0.7110748291015625,
+      "learning_rate": 0.0002912120885051268,
+      "loss": 3.3299,
+      "step": 47800
+    },
+    {
+      "epoch": 5.159029649595688,
+      "grad_norm": 0.6903175115585327,
+      "learning_rate": 0.00029088828926065834,
+      "loss": 3.3209,
+      "step": 47850
+    },
+    {
+      "epoch": 5.164420485175202,
+      "grad_norm": 0.7338415384292603,
+      "learning_rate": 0.00029056449001618994,
+      "loss": 3.3033,
+      "step": 47900
+    },
+    {
+      "epoch": 5.169811320754717,
+      "grad_norm": 0.723596453666687,
+      "learning_rate": 0.0002902406907717215,
+      "loss": 3.3219,
+      "step": 47950
+    },
+    {
+      "epoch": 5.175202156334231,
+      "grad_norm": 0.7516893148422241,
+      "learning_rate": 0.0002899168915272531,
+      "loss": 3.3119,
+      "step": 48000
+    },
+    {
+      "epoch": 5.175202156334231,
+      "eval_accuracy": 0.3811521573954425,
+      "eval_loss": 3.411118268966675,
+      "eval_runtime": 185.4994,
+      "eval_samples_per_second": 97.095,
+      "eval_steps_per_second": 6.07,
+      "step": 48000
+    },
+    {
+      "epoch": 5.180592991913747,
+      "grad_norm": 0.7215257883071899,
+      "learning_rate": 0.00028959309228278465,
+      "loss": 3.3424,
+      "step": 48050
+    },
+    {
+      "epoch": 5.185983827493262,
+      "grad_norm": 0.7489564418792725,
+      "learning_rate": 0.00028926929303831625,
+      "loss": 3.3215,
+      "step": 48100
+    },
+    {
+      "epoch": 5.191374663072776,
+      "grad_norm": 0.6841164231300354,
+      "learning_rate": 0.0002889454937938478,
+      "loss": 3.3318,
+      "step": 48150
+    },
+    {
+      "epoch": 5.196765498652291,
+      "grad_norm": 0.6923016309738159,
+      "learning_rate": 0.00028862169454937935,
+      "loss": 3.3199,
+      "step": 48200
+    },
+    {
+      "epoch": 5.202156334231806,
+      "grad_norm": 0.7165120840072632,
+      "learning_rate": 0.0002882978953049109,
+      "loss": 3.3028,
+      "step": 48250
+    },
+    {
+      "epoch": 5.2075471698113205,
+      "grad_norm": 0.6918339729309082,
+      "learning_rate": 0.0002879740960604425,
+      "loss": 3.3329,
+      "step": 48300
+    },
+    {
+      "epoch": 5.212938005390836,
+      "grad_norm": 0.6800699830055237,
+      "learning_rate": 0.00028765029681597406,
+      "loss": 3.3145,
+      "step": 48350
+    },
+    {
+      "epoch": 5.218328840970351,
+      "grad_norm": 0.7082724571228027,
+      "learning_rate": 0.00028732649757150566,
+      "loss": 3.3271,
+      "step": 48400
+    },
+    {
+      "epoch": 5.223719676549865,
+      "grad_norm": 0.719948947429657,
+      "learning_rate": 0.0002870026983270372,
+      "loss": 3.3173,
+      "step": 48450
+    },
+    {
+      "epoch": 5.22911051212938,
+      "grad_norm": 0.6641173362731934,
+      "learning_rate": 0.00028667889908256877,
+      "loss": 3.317,
+      "step": 48500
+    },
+    {
+      "epoch": 5.234501347708895,
+      "grad_norm": 0.6939883828163147,
+      "learning_rate": 0.00028635509983810037,
+      "loss": 3.3035,
+      "step": 48550
+    },
+    {
+      "epoch": 5.2398921832884096,
+      "grad_norm": 0.695296049118042,
+      "learning_rate": 0.0002860313005936319,
+      "loss": 3.3286,
+      "step": 48600
+    },
+    {
+      "epoch": 5.245283018867925,
+      "grad_norm": 0.7294591069221497,
+      "learning_rate": 0.00028570750134916347,
+      "loss": 3.3226,
+      "step": 48650
+    },
+    {
+      "epoch": 5.250673854447439,
+      "grad_norm": 0.710701048374176,
+      "learning_rate": 0.0002853837021046951,
+      "loss": 3.3203,
+      "step": 48700
+    },
+    {
+      "epoch": 5.256064690026954,
+      "grad_norm": 0.7199147343635559,
+      "learning_rate": 0.00028505990286022663,
+      "loss": 3.311,
+      "step": 48750
+    },
+    {
+      "epoch": 5.261455525606469,
+      "grad_norm": 0.6639696359634399,
+      "learning_rate": 0.0002847361036157582,
+      "loss": 3.3255,
+      "step": 48800
+    },
+    {
+      "epoch": 5.2668463611859835,
+      "grad_norm": 0.6906806826591492,
+      "learning_rate": 0.0002844123043712898,
+      "loss": 3.3368,
+      "step": 48850
+    },
+    {
+      "epoch": 5.272237196765499,
+      "grad_norm": 0.6999244689941406,
+      "learning_rate": 0.00028408850512682133,
+      "loss": 3.3156,
+      "step": 48900
+    },
+    {
+      "epoch": 5.277628032345014,
+      "grad_norm": 0.7082576751708984,
+      "learning_rate": 0.00028376470588235294,
+      "loss": 3.3318,
+      "step": 48950
+    },
+    {
+      "epoch": 5.283018867924528,
+      "grad_norm": 0.7592863440513611,
+      "learning_rate": 0.0002834409066378845,
+      "loss": 3.3434,
+      "step": 49000
+    },
+    {
+      "epoch": 5.283018867924528,
+      "eval_accuracy": 0.38130264164506555,
+      "eval_loss": 3.4073922634124756,
+      "eval_runtime": 185.3808,
+      "eval_samples_per_second": 97.157,
+      "eval_steps_per_second": 6.074,
+      "step": 49000
+    },
+    {
+      "epoch": 5.288409703504043,
+      "grad_norm": 0.7438406944274902,
+      "learning_rate": 0.00028311710739341604,
+      "loss": 3.3356,
+      "step": 49050
+    },
+    {
+      "epoch": 5.293800539083558,
+      "grad_norm": 0.7723414301872253,
+      "learning_rate": 0.00028279330814894764,
+      "loss": 3.3378,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2991913746630726,
+      "grad_norm": 0.7213961482048035,
+      "learning_rate": 0.0002824695089044792,
+      "loss": 3.3135,
+      "step": 49150
+    },
+    {
+      "epoch": 5.304582210242588,
+      "grad_norm": 0.6973984241485596,
+      "learning_rate": 0.00028214570966001075,
+      "loss": 3.3393,
+      "step": 49200
+    },
+    {
+      "epoch": 5.309973045822103,
+      "grad_norm": 0.6829140186309814,
+      "learning_rate": 0.00028182191041554235,
+      "loss": 3.327,
+      "step": 49250
+    },
+    {
+      "epoch": 5.315363881401617,
+      "grad_norm": 0.7045508027076721,
+      "learning_rate": 0.0002814981111710739,
+      "loss": 3.324,
+      "step": 49300
+    },
+    {
+      "epoch": 5.320754716981132,
+      "grad_norm": 0.6908060908317566,
+      "learning_rate": 0.0002811743119266055,
+      "loss": 3.3408,
+      "step": 49350
+    },
+    {
+      "epoch": 5.3261455525606465,
+      "grad_norm": 0.6872555613517761,
+      "learning_rate": 0.00028085051268213706,
+      "loss": 3.3154,
+      "step": 49400
+    },
+    {
+      "epoch": 5.331536388140162,
+      "grad_norm": 0.7134802341461182,
+      "learning_rate": 0.0002805267134376686,
+      "loss": 3.3139,
+      "step": 49450
+    },
+    {
+      "epoch": 5.336927223719677,
+      "grad_norm": 0.6903613805770874,
+      "learning_rate": 0.00028020291419320016,
+      "loss": 3.3408,
+      "step": 49500
+    },
+    {
+      "epoch": 5.342318059299191,
+      "grad_norm": 0.7270494103431702,
+      "learning_rate": 0.00027987911494873176,
+      "loss": 3.3396,
+      "step": 49550
+    },
+    {
+      "epoch": 5.347708894878706,
+      "grad_norm": 0.7505331635475159,
+      "learning_rate": 0.0002795553157042633,
+      "loss": 3.3326,
+      "step": 49600
+    },
+    {
+      "epoch": 5.353099730458221,
+      "grad_norm": 0.71560138463974,
+      "learning_rate": 0.0002792315164597949,
+      "loss": 3.3208,
+      "step": 49650
+    },
+    {
+      "epoch": 5.3584905660377355,
+      "grad_norm": 0.7327362895011902,
+      "learning_rate": 0.00027890771721532647,
+      "loss": 3.3467,
+      "step": 49700
+    },
+    {
+      "epoch": 5.363881401617251,
+      "grad_norm": 0.707087516784668,
+      "learning_rate": 0.00027858391797085807,
+      "loss": 3.3443,
+      "step": 49750
+    },
+    {
+      "epoch": 5.369272237196766,
+      "grad_norm": 0.6669909358024597,
+      "learning_rate": 0.0002782601187263896,
+      "loss": 3.3169,
+      "step": 49800
+    },
+    {
+      "epoch": 5.37466307277628,
+      "grad_norm": 0.6940752863883972,
+      "learning_rate": 0.0002779363194819212,
+      "loss": 3.3494,
+      "step": 49850
+    },
+    {
+      "epoch": 5.380053908355795,
+      "grad_norm": 0.7098177671432495,
+      "learning_rate": 0.0002776125202374527,
+      "loss": 3.3276,
+      "step": 49900
+    },
+    {
+      "epoch": 5.38544474393531,
+      "grad_norm": 0.7212302088737488,
+      "learning_rate": 0.00027728872099298433,
+      "loss": 3.3396,
+      "step": 49950
+    },
+    {
+      "epoch": 5.390835579514825,
+      "grad_norm": 0.7480193972587585,
+      "learning_rate": 0.0002769713977334053,
+      "loss": 3.3422,
+      "step": 50000
+    },
+    {
+      "epoch": 5.390835579514825,
+      "eval_accuracy": 0.3820361572907011,
+      "eval_loss": 3.4000582695007324,
+      "eval_runtime": 185.5499,
+      "eval_samples_per_second": 97.068,
+      "eval_steps_per_second": 6.068,
+      "step": 50000
+    },
+    {
+      "epoch": 5.39622641509434,
+      "grad_norm": 0.6794703602790833,
+      "learning_rate": 0.00027664759848893683,
+      "loss": 3.3215,
+      "step": 50050
+    },
+    {
+      "epoch": 5.401617250673855,
+      "grad_norm": 0.7536985874176025,
+      "learning_rate": 0.00027632379924446844,
+      "loss": 3.3188,
+      "step": 50100
+    },
+    {
+      "epoch": 5.407008086253369,
+      "grad_norm": 0.693673312664032,
+      "learning_rate": 0.000276,
+      "loss": 3.3313,
+      "step": 50150
+    },
+    {
+      "epoch": 5.412398921832884,
+      "grad_norm": 0.750146746635437,
+      "learning_rate": 0.00027567620075553154,
+      "loss": 3.3299,
+      "step": 50200
+    },
+    {
+      "epoch": 5.4177897574123985,
+      "grad_norm": 0.6878756880760193,
+      "learning_rate": 0.0002753524015110631,
+      "loss": 3.3373,
+      "step": 50250
+    },
+    {
+      "epoch": 5.423180592991914,
+      "grad_norm": 0.6811180710792542,
+      "learning_rate": 0.0002750286022665947,
+      "loss": 3.3215,
+      "step": 50300
+    },
+    {
+      "epoch": 5.428571428571429,
+      "grad_norm": 0.722919762134552,
+      "learning_rate": 0.00027470480302212624,
+      "loss": 3.3506,
+      "step": 50350
+    },
+    {
+      "epoch": 5.433962264150943,
+      "grad_norm": 0.7224785089492798,
+      "learning_rate": 0.0002743874797625472,
+      "loss": 3.3321,
+      "step": 50400
+    },
+    {
+      "epoch": 5.439353099730458,
+      "grad_norm": 0.7404599785804749,
+      "learning_rate": 0.0002740636805180788,
+      "loss": 3.3371,
+      "step": 50450
+    },
+    {
+      "epoch": 5.444743935309973,
+      "grad_norm": 0.7081565260887146,
+      "learning_rate": 0.00027373988127361035,
+      "loss": 3.325,
+      "step": 50500
+    },
+    {
+      "epoch": 5.450134770889488,
+      "grad_norm": 0.7478460669517517,
+      "learning_rate": 0.0002734160820291419,
+      "loss": 3.3324,
+      "step": 50550
+    },
+    {
+      "epoch": 5.455525606469003,
+      "grad_norm": 0.715099036693573,
+      "learning_rate": 0.00027309228278467345,
+      "loss": 3.3411,
+      "step": 50600
+    },
+    {
+      "epoch": 5.460916442048518,
+      "grad_norm": 0.6973506808280945,
+      "learning_rate": 0.00027276848354020506,
+      "loss": 3.3541,
+      "step": 50650
+    },
+    {
+      "epoch": 5.466307277628032,
+      "grad_norm": 0.800144612789154,
+      "learning_rate": 0.0002724446842957366,
+      "loss": 3.3159,
+      "step": 50700
+    },
+    {
+      "epoch": 5.471698113207547,
+      "grad_norm": 0.721748948097229,
+      "learning_rate": 0.0002721208850512682,
+      "loss": 3.3341,
+      "step": 50750
+    },
+    {
+      "epoch": 5.4770889487870615,
+      "grad_norm": 0.7428343296051025,
+      "learning_rate": 0.00027179708580679976,
+      "loss": 3.3349,
+      "step": 50800
+    },
+    {
+      "epoch": 5.482479784366577,
+      "grad_norm": 0.6495864391326904,
+      "learning_rate": 0.0002714732865623313,
+      "loss": 3.3282,
+      "step": 50850
+    },
+    {
+      "epoch": 5.487870619946092,
+      "grad_norm": 0.7069242000579834,
+      "learning_rate": 0.0002711494873178629,
+      "loss": 3.325,
+      "step": 50900
+    },
+    {
+      "epoch": 5.493261455525606,
+      "grad_norm": 0.6868419647216797,
+      "learning_rate": 0.00027082568807339447,
+      "loss": 3.3423,
+      "step": 50950
+    },
+    {
+      "epoch": 5.498652291105121,
+      "grad_norm": 0.7288803458213806,
+      "learning_rate": 0.000270501888828926,
+      "loss": 3.3445,
+      "step": 51000
+    },
+    {
+      "epoch": 5.498652291105121,
+      "eval_accuracy": 0.3819953038048829,
+      "eval_loss": 3.3993842601776123,
+      "eval_runtime": 185.485,
+      "eval_samples_per_second": 97.102,
+      "eval_steps_per_second": 6.071,
+      "step": 51000
+    },
+    {
+      "epoch": 5.504043126684636,
+      "grad_norm": 0.7231615781784058,
+      "learning_rate": 0.0002701780895844576,
+      "loss": 3.3506,
+      "step": 51050
+    },
+    {
+      "epoch": 5.509433962264151,
+      "grad_norm": 0.7433888912200928,
+      "learning_rate": 0.0002698542903399892,
+      "loss": 3.3371,
+      "step": 51100
+    },
+    {
+      "epoch": 5.514824797843666,
+      "grad_norm": 0.6915830969810486,
+      "learning_rate": 0.0002695304910955207,
+      "loss": 3.3342,
+      "step": 51150
+    },
+    {
+      "epoch": 5.520215633423181,
+      "grad_norm": 0.6907735466957092,
+      "learning_rate": 0.00026920669185105233,
+      "loss": 3.3171,
+      "step": 51200
+    },
+    {
+      "epoch": 5.525606469002695,
+      "grad_norm": 0.7327287197113037,
+      "learning_rate": 0.0002688828926065839,
+      "loss": 3.3385,
+      "step": 51250
+    },
+    {
+      "epoch": 5.53099730458221,
+      "grad_norm": 0.6716380715370178,
+      "learning_rate": 0.0002685590933621155,
+      "loss": 3.3296,
+      "step": 51300
+    },
+    {
+      "epoch": 5.536388140161725,
+      "grad_norm": 0.7345424890518188,
+      "learning_rate": 0.00026823529411764704,
+      "loss": 3.3215,
+      "step": 51350
+    },
+    {
+      "epoch": 5.54177897574124,
+      "grad_norm": 0.7785588502883911,
+      "learning_rate": 0.0002679114948731786,
+      "loss": 3.3123,
+      "step": 51400
+    },
+    {
+      "epoch": 5.547169811320755,
+      "grad_norm": 0.7506522536277771,
+      "learning_rate": 0.0002675876956287102,
+      "loss": 3.3115,
+      "step": 51450
+    },
+    {
+      "epoch": 5.55256064690027,
+      "grad_norm": 0.7358080744743347,
+      "learning_rate": 0.00026726389638424174,
+      "loss": 3.342,
+      "step": 51500
+    },
+    {
+      "epoch": 5.557951482479784,
+      "grad_norm": 0.7009307146072388,
+      "learning_rate": 0.0002669400971397733,
+      "loss": 3.3456,
+      "step": 51550
+    },
+    {
+      "epoch": 5.563342318059299,
+      "grad_norm": 0.6749496459960938,
+      "learning_rate": 0.0002666162978953049,
+      "loss": 3.3201,
+      "step": 51600
+    },
+    {
+      "epoch": 5.568733153638814,
+      "grad_norm": 0.7447863221168518,
+      "learning_rate": 0.00026629249865083645,
+      "loss": 3.336,
+      "step": 51650
+    },
+    {
+      "epoch": 5.574123989218329,
+      "grad_norm": 0.7776146531105042,
+      "learning_rate": 0.00026596869940636805,
+      "loss": 3.3398,
+      "step": 51700
+    },
+    {
+      "epoch": 5.579514824797844,
+      "grad_norm": 0.6787426471710205,
+      "learning_rate": 0.0002656449001618996,
+      "loss": 3.3314,
+      "step": 51750
+    },
+    {
+      "epoch": 5.584905660377358,
+      "grad_norm": 0.6856850385665894,
+      "learning_rate": 0.0002653211009174312,
+      "loss": 3.3232,
+      "step": 51800
+    },
+    {
+      "epoch": 5.590296495956873,
+      "grad_norm": 0.7217938899993896,
+      "learning_rate": 0.0002649973016729627,
+      "loss": 3.332,
+      "step": 51850
+    },
+    {
+      "epoch": 5.595687331536388,
+      "grad_norm": 0.7498999238014221,
+      "learning_rate": 0.0002646735024284943,
+      "loss": 3.3218,
+      "step": 51900
+    },
+    {
+      "epoch": 5.601078167115903,
+      "grad_norm": 0.7334463596343994,
+      "learning_rate": 0.00026434970318402586,
+      "loss": 3.339,
+      "step": 51950
+    },
+    {
+      "epoch": 5.606469002695418,
+      "grad_norm": 0.6884651780128479,
+      "learning_rate": 0.00026402590393955746,
+      "loss": 3.3169,
+      "step": 52000
+    },
+    {
+      "epoch": 5.606469002695418,
+      "eval_accuracy": 0.3826099531912494,
+      "eval_loss": 3.392883777618408,
+      "eval_runtime": 185.4171,
+      "eval_samples_per_second": 97.138,
+      "eval_steps_per_second": 6.073,
+      "step": 52000
+    },
+    {
+      "epoch": 5.611859838274933,
+      "grad_norm": 0.782909095287323,
+      "learning_rate": 0.000263702104695089,
+      "loss": 3.3343,
+      "step": 52050
+    },
+    {
+      "epoch": 5.617250673854447,
+      "grad_norm": 0.7251365184783936,
+      "learning_rate": 0.0002633783054506206,
+      "loss": 3.3271,
+      "step": 52100
+    },
+    {
+      "epoch": 5.622641509433962,
+      "grad_norm": 0.72756427526474,
+      "learning_rate": 0.00026305450620615217,
+      "loss": 3.3473,
+      "step": 52150
+    },
+    {
+      "epoch": 5.628032345013477,
+      "grad_norm": 0.7529804706573486,
+      "learning_rate": 0.0002627307069616837,
+      "loss": 3.3497,
+      "step": 52200
+    },
+    {
+      "epoch": 5.633423180592992,
+      "grad_norm": 0.6761258244514465,
+      "learning_rate": 0.0002624069077172153,
+      "loss": 3.3294,
+      "step": 52250
+    },
+    {
+      "epoch": 5.638814016172507,
+      "grad_norm": 0.7436754107475281,
+      "learning_rate": 0.0002620831084727469,
+      "loss": 3.3182,
+      "step": 52300
+    },
+    {
+      "epoch": 5.644204851752022,
+      "grad_norm": 0.7272395491600037,
+      "learning_rate": 0.0002617593092282784,
+      "loss": 3.3417,
+      "step": 52350
+    },
+    {
+      "epoch": 5.649595687331536,
+      "grad_norm": 0.6948999762535095,
+      "learning_rate": 0.00026143550998381003,
+      "loss": 3.3186,
+      "step": 52400
+    },
+    {
+      "epoch": 5.654986522911051,
+      "grad_norm": 0.7938417792320251,
+      "learning_rate": 0.0002611117107393416,
+      "loss": 3.3372,
+      "step": 52450
+    },
+    {
+      "epoch": 5.660377358490566,
+      "grad_norm": 0.7938036918640137,
+      "learning_rate": 0.00026078791149487313,
+      "loss": 3.3445,
+      "step": 52500
+    },
+    {
+      "epoch": 5.665768194070081,
+      "grad_norm": 0.669804036617279,
+      "learning_rate": 0.00026046411225040474,
+      "loss": 3.3305,
+      "step": 52550
+    },
+    {
+      "epoch": 5.671159029649596,
+      "grad_norm": 0.7671605348587036,
+      "learning_rate": 0.0002601403130059363,
+      "loss": 3.3441,
+      "step": 52600
+    },
+    {
+      "epoch": 5.67654986522911,
+      "grad_norm": 0.7261316180229187,
+      "learning_rate": 0.0002598165137614679,
+      "loss": 3.3439,
+      "step": 52650
+    },
+    {
+      "epoch": 5.681940700808625,
+      "grad_norm": 0.6810511946678162,
+      "learning_rate": 0.00025949271451699944,
+      "loss": 3.3326,
+      "step": 52700
+    },
+    {
+      "epoch": 5.6873315363881405,
+      "grad_norm": 0.7681779265403748,
+      "learning_rate": 0.000259168915272531,
+      "loss": 3.3468,
+      "step": 52750
+    },
+    {
+      "epoch": 5.692722371967655,
+      "grad_norm": 0.7268847227096558,
+      "learning_rate": 0.0002588451160280626,
+      "loss": 3.3262,
+      "step": 52800
+    },
+    {
+      "epoch": 5.69811320754717,
+      "grad_norm": 0.7206116914749146,
+      "learning_rate": 0.00025852131678359415,
+      "loss": 3.3247,
+      "step": 52850
+    },
+    {
+      "epoch": 5.703504043126685,
+      "grad_norm": 0.7164592146873474,
+      "learning_rate": 0.0002581975175391257,
+      "loss": 3.3254,
+      "step": 52900
+    },
+    {
+      "epoch": 5.708894878706199,
+      "grad_norm": 0.7586601972579956,
+      "learning_rate": 0.0002578737182946573,
+      "loss": 3.3434,
+      "step": 52950
+    },
+    {
+      "epoch": 5.714285714285714,
+      "grad_norm": 0.786045491695404,
+      "learning_rate": 0.00025754991905018885,
+      "loss": 3.333,
+      "step": 53000
+    },
+    {
+      "epoch": 5.714285714285714,
+      "eval_accuracy": 0.38318516157933924,
+      "eval_loss": 3.3877861499786377,
+      "eval_runtime": 185.5745,
+      "eval_samples_per_second": 97.055,
+      "eval_steps_per_second": 6.068,
+      "step": 53000
+    },
+    {
+      "epoch": 5.719676549865229,
+      "grad_norm": 0.6914837956428528,
+      "learning_rate": 0.00025722611980572046,
+      "loss": 3.311,
+      "step": 53050
+    },
+    {
+      "epoch": 5.725067385444744,
+      "grad_norm": 0.767452597618103,
+      "learning_rate": 0.000256902320561252,
+      "loss": 3.3264,
+      "step": 53100
+    },
+    {
+      "epoch": 5.730458221024259,
+      "grad_norm": 0.7509741187095642,
+      "learning_rate": 0.00025657852131678356,
+      "loss": 3.333,
+      "step": 53150
+    },
+    {
+      "epoch": 5.735849056603773,
+      "grad_norm": 0.735542356967926,
+      "learning_rate": 0.0002562547220723151,
+      "loss": 3.3335,
+      "step": 53200
+    },
+    {
+      "epoch": 5.741239892183288,
+      "grad_norm": 0.709091067314148,
+      "learning_rate": 0.0002559309228278467,
+      "loss": 3.3245,
+      "step": 53250
+    },
+    {
+      "epoch": 5.7466307277628035,
+      "grad_norm": 0.741787850856781,
+      "learning_rate": 0.00025560712358337827,
+      "loss": 3.3504,
+      "step": 53300
+    },
+    {
+      "epoch": 5.752021563342318,
+      "grad_norm": 0.7668866515159607,
+      "learning_rate": 0.00025528332433890987,
+      "loss": 3.3277,
+      "step": 53350
+    },
+    {
+      "epoch": 5.757412398921833,
+      "grad_norm": 0.7090975046157837,
+      "learning_rate": 0.0002549595250944414,
+      "loss": 3.3134,
+      "step": 53400
+    },
+    {
+      "epoch": 5.762803234501348,
+      "grad_norm": 0.7142593860626221,
+      "learning_rate": 0.000254635725849973,
+      "loss": 3.3448,
+      "step": 53450
+    },
+    {
+      "epoch": 5.768194070080862,
+      "grad_norm": 0.7309287786483765,
+      "learning_rate": 0.0002543119266055046,
+      "loss": 3.3142,
+      "step": 53500
+    },
+    {
+      "epoch": 5.773584905660377,
+      "grad_norm": 0.7920538783073425,
+      "learning_rate": 0.00025398812736103613,
+      "loss": 3.3432,
+      "step": 53550
+    },
+    {
+      "epoch": 5.7789757412398925,
+      "grad_norm": 0.7171128392219543,
+      "learning_rate": 0.0002536643281165677,
+      "loss": 3.3163,
+      "step": 53600
+    },
+    {
+      "epoch": 5.784366576819407,
+      "grad_norm": 0.6814385056495667,
+      "learning_rate": 0.0002533405288720993,
+      "loss": 3.3494,
+      "step": 53650
+    },
+    {
+      "epoch": 5.789757412398922,
+      "grad_norm": 0.7790846228599548,
+      "learning_rate": 0.00025301672962763083,
+      "loss": 3.3204,
+      "step": 53700
+    },
+    {
+      "epoch": 5.795148247978437,
+      "grad_norm": 0.7661272287368774,
+      "learning_rate": 0.00025269293038316244,
+      "loss": 3.3356,
+      "step": 53750
+    },
+    {
+      "epoch": 5.800539083557951,
+      "grad_norm": 0.7544566988945007,
+      "learning_rate": 0.000252369131138694,
+      "loss": 3.3269,
+      "step": 53800
+    },
+    {
+      "epoch": 5.8059299191374665,
+      "grad_norm": 0.7239667177200317,
+      "learning_rate": 0.00025204533189422554,
+      "loss": 3.3316,
+      "step": 53850
+    },
+    {
+      "epoch": 5.811320754716981,
+      "grad_norm": 0.7095721364021301,
+      "learning_rate": 0.00025172153264975714,
+      "loss": 3.3212,
+      "step": 53900
+    },
+    {
+      "epoch": 5.816711590296496,
+      "grad_norm": 0.7116101384162903,
+      "learning_rate": 0.0002513977334052887,
+      "loss": 3.3163,
+      "step": 53950
+    },
+    {
+      "epoch": 5.822102425876011,
+      "grad_norm": 0.7057174444198608,
+      "learning_rate": 0.00025107393416082025,
+      "loss": 3.331,
+      "step": 54000
+    },
+    {
+      "epoch": 5.822102425876011,
+      "eval_accuracy": 0.3834898242767711,
+      "eval_loss": 3.384427070617676,
+      "eval_runtime": 185.8424,
+      "eval_samples_per_second": 96.915,
+      "eval_steps_per_second": 6.059,
+      "step": 54000
+    },
+    {
+      "epoch": 5.827493261455525,
+      "grad_norm": 0.7220545411109924,
+      "learning_rate": 0.00025075013491635185,
+      "loss": 3.3434,
+      "step": 54050
+    },
+    {
+      "epoch": 5.83288409703504,
+      "grad_norm": 0.6828851103782654,
+      "learning_rate": 0.0002504263356718834,
+      "loss": 3.3495,
+      "step": 54100
+    },
+    {
+      "epoch": 5.8382749326145555,
+      "grad_norm": 0.7387149333953857,
+      "learning_rate": 0.000250102536427415,
+      "loss": 3.323,
+      "step": 54150
+    },
+    {
+      "epoch": 5.84366576819407,
+      "grad_norm": 0.7202645540237427,
+      "learning_rate": 0.00024977873718294656,
+      "loss": 3.3336,
+      "step": 54200
+    },
+    {
+      "epoch": 5.849056603773585,
+      "grad_norm": 0.7070654630661011,
+      "learning_rate": 0.0002494549379384781,
+      "loss": 3.3618,
+      "step": 54250
+    },
+    {
+      "epoch": 5.8544474393531,
+      "grad_norm": 0.71637362241745,
+      "learning_rate": 0.0002491311386940097,
+      "loss": 3.3402,
+      "step": 54300
+    },
+    {
+      "epoch": 5.859838274932614,
+      "grad_norm": 0.6806088089942932,
+      "learning_rate": 0.00024880733944954126,
+      "loss": 3.3416,
+      "step": 54350
+    },
+    {
+      "epoch": 5.8652291105121295,
+      "grad_norm": 0.73921799659729,
+      "learning_rate": 0.0002484835402050728,
+      "loss": 3.3452,
+      "step": 54400
+    },
+    {
+      "epoch": 5.870619946091644,
+      "grad_norm": 0.7796598076820374,
+      "learning_rate": 0.0002481597409606044,
+      "loss": 3.3176,
+      "step": 54450
+    },
+    {
+      "epoch": 5.876010781671159,
+      "grad_norm": 0.7413299679756165,
+      "learning_rate": 0.00024783594171613597,
+      "loss": 3.3442,
+      "step": 54500
+    },
+    {
+      "epoch": 5.881401617250674,
+      "grad_norm": 0.722176194190979,
+      "learning_rate": 0.0002475121424716675,
+      "loss": 3.332,
+      "step": 54550
+    },
+    {
+      "epoch": 5.886792452830189,
+      "grad_norm": 0.6924837231636047,
+      "learning_rate": 0.0002471883432271991,
+      "loss": 3.3228,
+      "step": 54600
+    },
+    {
+      "epoch": 5.892183288409703,
+      "grad_norm": 0.7435639500617981,
+      "learning_rate": 0.0002468645439827307,
+      "loss": 3.3468,
+      "step": 54650
+    },
+    {
+      "epoch": 5.8975741239892185,
+      "grad_norm": 0.7494300007820129,
+      "learning_rate": 0.0002465407447382623,
+      "loss": 3.3424,
+      "step": 54700
+    },
+    {
+      "epoch": 5.902964959568733,
+      "grad_norm": 0.7710332274436951,
+      "learning_rate": 0.00024621694549379383,
+      "loss": 3.3272,
+      "step": 54750
+    },
+    {
+      "epoch": 5.908355795148248,
+      "grad_norm": 0.7281633615493774,
+      "learning_rate": 0.00024589314624932543,
+      "loss": 3.3583,
+      "step": 54800
+    },
+    {
+      "epoch": 5.913746630727763,
+      "grad_norm": 0.7332462072372437,
+      "learning_rate": 0.00024556934700485693,
+      "loss": 3.3256,
+      "step": 54850
+    },
+    {
+      "epoch": 5.919137466307277,
+      "grad_norm": 0.7332950234413147,
+      "learning_rate": 0.00024524554776038853,
+      "loss": 3.3547,
+      "step": 54900
+    },
+    {
+      "epoch": 5.9245283018867925,
+      "grad_norm": 0.7006910443305969,
+      "learning_rate": 0.0002449217485159201,
+      "loss": 3.3328,
+      "step": 54950
+    },
+    {
+      "epoch": 5.929919137466308,
+      "grad_norm": 0.8082262277603149,
+      "learning_rate": 0.0002445979492714517,
+      "loss": 3.3342,
+      "step": 55000
+    },
+    {
+      "epoch": 5.929919137466308,
+      "eval_accuracy": 0.38402233207994985,
+      "eval_loss": 3.380251884460449,
+      "eval_runtime": 185.4976,
+      "eval_samples_per_second": 97.096,
+      "eval_steps_per_second": 6.07,
+      "step": 55000
+    },
+    {
+      "epoch": 5.935309973045822,
+      "grad_norm": 0.7710868716239929,
+      "learning_rate": 0.00024427415002698324,
+      "loss": 3.3547,
+      "step": 55050
+    },
+    {
+      "epoch": 5.940700808625337,
+      "grad_norm": 0.7845757007598877,
+      "learning_rate": 0.00024395035078251482,
+      "loss": 3.3294,
+      "step": 55100
+    },
+    {
+      "epoch": 5.946091644204852,
+      "grad_norm": 0.7587760090827942,
+      "learning_rate": 0.0002436265515380464,
+      "loss": 3.3422,
+      "step": 55150
+    },
+    {
+      "epoch": 5.951482479784366,
+      "grad_norm": 0.7302002310752869,
+      "learning_rate": 0.00024330275229357797,
+      "loss": 3.3277,
+      "step": 55200
+    },
+    {
+      "epoch": 5.9568733153638815,
+      "grad_norm": 0.7525575757026672,
+      "learning_rate": 0.00024297895304910952,
+      "loss": 3.3126,
+      "step": 55250
+    },
+    {
+      "epoch": 5.962264150943396,
+      "grad_norm": 0.719638466835022,
+      "learning_rate": 0.00024266162978953048,
+      "loss": 3.3299,
+      "step": 55300
+    },
+    {
+      "epoch": 5.967654986522911,
+      "grad_norm": 0.7314834594726562,
+      "learning_rate": 0.00024233783054506203,
+      "loss": 3.3279,
+      "step": 55350
+    },
+    {
+      "epoch": 5.973045822102426,
+      "grad_norm": 0.6763395071029663,
+      "learning_rate": 0.0002420140313005936,
+      "loss": 3.3395,
+      "step": 55400
+    },
+    {
+      "epoch": 5.97843665768194,
+      "grad_norm": 0.7260298132896423,
+      "learning_rate": 0.00024169023205612518,
+      "loss": 3.3475,
+      "step": 55450
+    },
+    {
+      "epoch": 5.9838274932614555,
+      "grad_norm": 0.7525773644447327,
+      "learning_rate": 0.00024136643281165676,
+      "loss": 3.3146,
+      "step": 55500
+    },
+    {
+      "epoch": 5.989218328840971,
+      "grad_norm": 0.7207843065261841,
+      "learning_rate": 0.00024104263356718834,
+      "loss": 3.3359,
+      "step": 55550
+    },
+    {
+      "epoch": 5.994609164420485,
+      "grad_norm": 0.698942244052887,
+      "learning_rate": 0.0002407188343227199,
+      "loss": 3.349,
+      "step": 55600
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 1.4941537380218506,
+      "learning_rate": 0.00024039503507825147,
+      "loss": 3.3174,
+      "step": 55650
+    },
+    {
+      "epoch": 6.005390835579515,
+      "grad_norm": 0.7332544326782227,
+      "learning_rate": 0.00024007123583378302,
+      "loss": 3.2308,
+      "step": 55700
+    },
+    {
+      "epoch": 6.010781671159029,
+      "grad_norm": 0.7470399737358093,
+      "learning_rate": 0.0002397474365893146,
+      "loss": 3.2343,
+      "step": 55750
+    },
+    {
+      "epoch": 6.0161725067385445,
+      "grad_norm": 0.7069420218467712,
+      "learning_rate": 0.00023942363734484617,
+      "loss": 3.2661,
+      "step": 55800
+    },
+    {
+      "epoch": 6.02156334231806,
+      "grad_norm": 0.7124414443969727,
+      "learning_rate": 0.00023909983810037775,
+      "loss": 3.2322,
+      "step": 55850
+    },
+    {
+      "epoch": 6.026954177897574,
+      "grad_norm": 0.7617751955986023,
+      "learning_rate": 0.00023877603885590933,
+      "loss": 3.2335,
+      "step": 55900
+    },
+    {
+      "epoch": 6.032345013477089,
+      "grad_norm": 0.7374067902565002,
+      "learning_rate": 0.0002384522396114409,
+      "loss": 3.2608,
+      "step": 55950
+    },
+    {
+      "epoch": 6.037735849056604,
+      "grad_norm": 0.7492246031761169,
+      "learning_rate": 0.00023812844036697248,
+      "loss": 3.2353,
+      "step": 56000
+    },
+    {
+      "epoch": 6.037735849056604,
+      "eval_accuracy": 0.384033957938946,
+      "eval_loss": 3.384352684020996,
+      "eval_runtime": 185.3605,
+      "eval_samples_per_second": 97.167,
+      "eval_steps_per_second": 6.075,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0431266846361185,
+      "grad_norm": 0.7165188193321228,
+      "learning_rate": 0.000237804641122504,
+      "loss": 3.2517,
+      "step": 56050
+    },
+    {
+      "epoch": 6.048517520215634,
+      "grad_norm": 0.7443341612815857,
+      "learning_rate": 0.00023748084187803558,
+      "loss": 3.2432,
+      "step": 56100
+    },
+    {
+      "epoch": 6.053908355795148,
+      "grad_norm": 0.731981098651886,
+      "learning_rate": 0.00023715704263356716,
+      "loss": 3.2602,
+      "step": 56150
+    },
+    {
+      "epoch": 6.059299191374663,
+      "grad_norm": 0.7504057288169861,
+      "learning_rate": 0.00023683324338909874,
+      "loss": 3.2583,
+      "step": 56200
+    },
+    {
+      "epoch": 6.064690026954178,
+      "grad_norm": 0.7301583290100098,
+      "learning_rate": 0.00023650944414463032,
+      "loss": 3.2514,
+      "step": 56250
+    },
+    {
+      "epoch": 6.070080862533692,
+      "grad_norm": 0.7722079753875732,
+      "learning_rate": 0.0002361856449001619,
+      "loss": 3.2588,
+      "step": 56300
+    },
+    {
+      "epoch": 6.0754716981132075,
+      "grad_norm": 0.7549823522567749,
+      "learning_rate": 0.00023586184565569347,
+      "loss": 3.2598,
+      "step": 56350
+    },
+    {
+      "epoch": 6.080862533692723,
+      "grad_norm": 0.7237383723258972,
+      "learning_rate": 0.00023553804641122502,
+      "loss": 3.2725,
+      "step": 56400
+    },
+    {
+      "epoch": 6.086253369272237,
+      "grad_norm": 0.7340690493583679,
+      "learning_rate": 0.00023521424716675657,
+      "loss": 3.2511,
+      "step": 56450
+    },
+    {
+      "epoch": 6.091644204851752,
+      "grad_norm": 0.7521067261695862,
+      "learning_rate": 0.00023489044792228815,
+      "loss": 3.2467,
+      "step": 56500
+    },
+    {
+      "epoch": 6.097035040431267,
+      "grad_norm": 0.7757521271705627,
+      "learning_rate": 0.00023456664867781973,
+      "loss": 3.2494,
+      "step": 56550
+    },
+    {
+      "epoch": 6.1024258760107815,
+      "grad_norm": 0.7578301429748535,
+      "learning_rate": 0.0002342428494333513,
+      "loss": 3.2627,
+      "step": 56600
+    },
+    {
+      "epoch": 6.107816711590297,
+      "grad_norm": 0.7615374326705933,
+      "learning_rate": 0.00023391905018888288,
+      "loss": 3.2605,
+      "step": 56650
+    },
+    {
+      "epoch": 6.113207547169812,
+      "grad_norm": 0.7228667736053467,
+      "learning_rate": 0.00023359525094441443,
+      "loss": 3.2608,
+      "step": 56700
+    },
+    {
+      "epoch": 6.118598382749326,
+      "grad_norm": 0.7355426549911499,
+      "learning_rate": 0.000233271451699946,
+      "loss": 3.2673,
+      "step": 56750
+    },
+    {
+      "epoch": 6.123989218328841,
+      "grad_norm": 0.7165455222129822,
+      "learning_rate": 0.0002329476524554776,
+      "loss": 3.2662,
+      "step": 56800
+    },
+    {
+      "epoch": 6.129380053908355,
+      "grad_norm": 0.7375234961509705,
+      "learning_rate": 0.00023262385321100917,
+      "loss": 3.2589,
+      "step": 56850
+    },
+    {
+      "epoch": 6.1347708894878705,
+      "grad_norm": 0.74793541431427,
+      "learning_rate": 0.00023230005396654072,
+      "loss": 3.2684,
+      "step": 56900
+    },
+    {
+      "epoch": 6.140161725067386,
+      "grad_norm": 0.7644237279891968,
+      "learning_rate": 0.0002319762547220723,
+      "loss": 3.2657,
+      "step": 56950
+    },
+    {
+      "epoch": 6.1455525606469,
+      "grad_norm": 0.7800267934799194,
+      "learning_rate": 0.00023165245547760387,
+      "loss": 3.2567,
+      "step": 57000
+    },
+    {
+      "epoch": 6.1455525606469,
+      "eval_accuracy": 0.38427158180459625,
+      "eval_loss": 3.3837149143218994,
+      "eval_runtime": 185.459,
+      "eval_samples_per_second": 97.116,
+      "eval_steps_per_second": 6.071,
+      "step": 57000
+    },
+    {
+      "epoch": 6.150943396226415,
+      "grad_norm": 0.7396030426025391,
+      "learning_rate": 0.00023132865623313542,
+      "loss": 3.2724,
+      "step": 57050
+    },
+    {
+      "epoch": 6.15633423180593,
+      "grad_norm": 0.7594435811042786,
+      "learning_rate": 0.000231004856988667,
+      "loss": 3.2756,
+      "step": 57100
+    },
+    {
+      "epoch": 6.1617250673854445,
+      "grad_norm": 0.7593262195587158,
+      "learning_rate": 0.00023068105774419858,
+      "loss": 3.2731,
+      "step": 57150
+    },
+    {
+      "epoch": 6.16711590296496,
+      "grad_norm": 0.7153212428092957,
+      "learning_rate": 0.00023035725849973016,
+      "loss": 3.2591,
+      "step": 57200
+    },
+    {
+      "epoch": 6.172506738544475,
+      "grad_norm": 0.7407529354095459,
+      "learning_rate": 0.00023003345925526173,
+      "loss": 3.2671,
+      "step": 57250
+    },
+    {
+      "epoch": 6.177897574123989,
+      "grad_norm": 0.7409720420837402,
+      "learning_rate": 0.00022970966001079328,
+      "loss": 3.2746,
+      "step": 57300
+    },
+    {
+      "epoch": 6.183288409703504,
+      "grad_norm": 0.7745881676673889,
+      "learning_rate": 0.00022938586076632484,
+      "loss": 3.2796,
+      "step": 57350
+    },
+    {
+      "epoch": 6.188679245283019,
+      "grad_norm": 0.8050203919410706,
+      "learning_rate": 0.0002290620615218564,
+      "loss": 3.2638,
+      "step": 57400
+    },
+    {
+      "epoch": 6.1940700808625335,
+      "grad_norm": 0.7250150442123413,
+      "learning_rate": 0.000228738262277388,
+      "loss": 3.2607,
+      "step": 57450
+    },
+    {
+      "epoch": 6.199460916442049,
+      "grad_norm": 0.8042988777160645,
+      "learning_rate": 0.00022841446303291957,
+      "loss": 3.2757,
+      "step": 57500
+    },
+    {
+      "epoch": 6.204851752021563,
+      "grad_norm": 0.7667548060417175,
+      "learning_rate": 0.00022809066378845115,
+      "loss": 3.2992,
+      "step": 57550
+    },
+    {
+      "epoch": 6.210242587601078,
+      "grad_norm": 0.7277979850769043,
+      "learning_rate": 0.00022776686454398272,
+      "loss": 3.2697,
+      "step": 57600
+    },
+    {
+      "epoch": 6.215633423180593,
+      "grad_norm": 0.7964656949043274,
+      "learning_rate": 0.0002274430652995143,
+      "loss": 3.2737,
+      "step": 57650
+    },
+    {
+      "epoch": 6.2210242587601075,
+      "grad_norm": 0.7558398842811584,
+      "learning_rate": 0.00022711926605504588,
+      "loss": 3.2685,
+      "step": 57700
+    },
+    {
+      "epoch": 6.226415094339623,
+      "grad_norm": 0.7219595313072205,
+      "learning_rate": 0.0002267954668105774,
+      "loss": 3.2591,
+      "step": 57750
+    },
+    {
+      "epoch": 6.231805929919138,
+      "grad_norm": 0.7618371248245239,
+      "learning_rate": 0.00022647166756610898,
+      "loss": 3.2666,
+      "step": 57800
+    },
+    {
+      "epoch": 6.237196765498652,
+      "grad_norm": 0.7510839700698853,
+      "learning_rate": 0.00022614786832164056,
+      "loss": 3.2738,
+      "step": 57850
+    },
+    {
+      "epoch": 6.242587601078167,
+      "grad_norm": 0.736240565776825,
+      "learning_rate": 0.00022582406907717214,
+      "loss": 3.2727,
+      "step": 57900
+    },
+    {
+      "epoch": 6.247978436657682,
+      "grad_norm": 0.7617259621620178,
+      "learning_rate": 0.0002255002698327037,
+      "loss": 3.2813,
+      "step": 57950
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "grad_norm": 0.7635646462440491,
+      "learning_rate": 0.0002251764705882353,
+      "loss": 3.2751,
+      "step": 58000
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "eval_accuracy": 0.3847077144962836,
+      "eval_loss": 3.3810763359069824,
+      "eval_runtime": 185.3536,
+      "eval_samples_per_second": 97.171,
+      "eval_steps_per_second": 6.075,
+      "step": 58000
+    },
+    {
+      "epoch": 6.258760107816712,
+      "grad_norm": 0.7703520059585571,
+      "learning_rate": 0.00022485267134376687,
+      "loss": 3.2669,
+      "step": 58050
+    },
+    {
+      "epoch": 6.264150943396227,
+      "grad_norm": 0.7938615679740906,
+      "learning_rate": 0.00022452887209929842,
+      "loss": 3.2608,
+      "step": 58100
+    },
+    {
+      "epoch": 6.269541778975741,
+      "grad_norm": 0.7358196973800659,
+      "learning_rate": 0.00022420507285482997,
+      "loss": 3.2608,
+      "step": 58150
+    },
+    {
+      "epoch": 6.274932614555256,
+      "grad_norm": 0.7554171085357666,
+      "learning_rate": 0.00022388127361036155,
+      "loss": 3.2987,
+      "step": 58200
+    },
+    {
+      "epoch": 6.280323450134771,
+      "grad_norm": 0.7923507690429688,
+      "learning_rate": 0.00022355747436589312,
+      "loss": 3.2738,
+      "step": 58250
+    },
+    {
+      "epoch": 6.285714285714286,
+      "grad_norm": 0.7709658145904541,
+      "learning_rate": 0.0002232336751214247,
+      "loss": 3.2603,
+      "step": 58300
+    },
+    {
+      "epoch": 6.291105121293801,
+      "grad_norm": 0.7847428917884827,
+      "learning_rate": 0.00022290987587695628,
+      "loss": 3.2872,
+      "step": 58350
+    },
+    {
+      "epoch": 6.296495956873315,
+      "grad_norm": 0.7624788284301758,
+      "learning_rate": 0.00022258607663248783,
+      "loss": 3.2701,
+      "step": 58400
+    },
+    {
+      "epoch": 6.30188679245283,
+      "grad_norm": 0.7347487211227417,
+      "learning_rate": 0.0002222622773880194,
+      "loss": 3.2768,
+      "step": 58450
+    },
+    {
+      "epoch": 6.307277628032345,
+      "grad_norm": 0.8133580088615417,
+      "learning_rate": 0.00022193847814355099,
+      "loss": 3.2917,
+      "step": 58500
+    },
+    {
+      "epoch": 6.3126684636118595,
+      "grad_norm": 0.7164966464042664,
+      "learning_rate": 0.00022161467889908256,
+      "loss": 3.253,
+      "step": 58550
+    },
+    {
+      "epoch": 6.318059299191375,
+      "grad_norm": 0.7584646940231323,
+      "learning_rate": 0.00022129087965461411,
+      "loss": 3.2864,
+      "step": 58600
+    },
+    {
+      "epoch": 6.32345013477089,
+      "grad_norm": 0.7676268815994263,
+      "learning_rate": 0.0002209670804101457,
+      "loss": 3.2635,
+      "step": 58650
+    },
+    {
+      "epoch": 6.328840970350404,
+      "grad_norm": 0.7437969446182251,
+      "learning_rate": 0.00022064328116567724,
+      "loss": 3.2665,
+      "step": 58700
+    },
+    {
+      "epoch": 6.334231805929919,
+      "grad_norm": 0.7605310082435608,
+      "learning_rate": 0.00022031948192120882,
+      "loss": 3.2876,
+      "step": 58750
+    },
+    {
+      "epoch": 6.339622641509434,
+      "grad_norm": 0.7673757076263428,
+      "learning_rate": 0.0002199956826767404,
+      "loss": 3.2797,
+      "step": 58800
+    },
+    {
+      "epoch": 6.345013477088949,
+      "grad_norm": 0.7611026167869568,
+      "learning_rate": 0.00021967188343227198,
+      "loss": 3.2833,
+      "step": 58850
+    },
+    {
+      "epoch": 6.350404312668464,
+      "grad_norm": 0.7772629857063293,
+      "learning_rate": 0.00021934808418780355,
+      "loss": 3.2885,
+      "step": 58900
+    },
+    {
+      "epoch": 6.355795148247978,
+      "grad_norm": 0.7738526463508606,
+      "learning_rate": 0.00021902428494333513,
+      "loss": 3.2982,
+      "step": 58950
+    },
+    {
+      "epoch": 6.361185983827493,
+      "grad_norm": 0.766217052936554,
+      "learning_rate": 0.00021870048569886665,
+      "loss": 3.2877,
+      "step": 59000
+    },
+    {
+      "epoch": 6.361185983827493,
+      "eval_accuracy": 0.38495500846894937,
+      "eval_loss": 3.376774787902832,
+      "eval_runtime": 185.626,
+      "eval_samples_per_second": 97.028,
+      "eval_steps_per_second": 6.066,
+      "step": 59000
+    },
+    {
+      "epoch": 6.366576819407008,
+      "grad_norm": 0.7514162659645081,
+      "learning_rate": 0.00021837668645439823,
+      "loss": 3.285,
+      "step": 59050
+    },
+    {
+      "epoch": 6.3719676549865225,
+      "grad_norm": 0.7757066488265991,
+      "learning_rate": 0.0002180528872099298,
+      "loss": 3.2815,
+      "step": 59100
+    },
+    {
+      "epoch": 6.377358490566038,
+      "grad_norm": 0.7878845930099487,
+      "learning_rate": 0.0002177290879654614,
+      "loss": 3.2825,
+      "step": 59150
+    },
+    {
+      "epoch": 6.382749326145553,
+      "grad_norm": 0.7204424142837524,
+      "learning_rate": 0.00021740528872099296,
+      "loss": 3.2754,
+      "step": 59200
+    },
+    {
+      "epoch": 6.388140161725067,
+      "grad_norm": 0.776679515838623,
+      "learning_rate": 0.00021708148947652454,
+      "loss": 3.2768,
+      "step": 59250
+    },
+    {
+      "epoch": 6.393530997304582,
+      "grad_norm": 0.7729427218437195,
+      "learning_rate": 0.0002167641662169455,
+      "loss": 3.2716,
+      "step": 59300
+    },
+    {
+      "epoch": 6.398921832884097,
+      "grad_norm": 0.7658233642578125,
+      "learning_rate": 0.00021644036697247702,
+      "loss": 3.2639,
+      "step": 59350
+    },
+    {
+      "epoch": 6.404312668463612,
+      "grad_norm": 0.7380334138870239,
+      "learning_rate": 0.0002161165677280086,
+      "loss": 3.2524,
+      "step": 59400
+    },
+    {
+      "epoch": 6.409703504043127,
+      "grad_norm": 0.7924899458885193,
+      "learning_rate": 0.00021579276848354017,
+      "loss": 3.2818,
+      "step": 59450
+    },
+    {
+      "epoch": 6.415094339622642,
+      "grad_norm": 0.7978893518447876,
+      "learning_rate": 0.00021546896923907175,
+      "loss": 3.2748,
+      "step": 59500
+    },
+    {
+      "epoch": 6.420485175202156,
+      "grad_norm": 0.7716631889343262,
+      "learning_rate": 0.00021514516999460333,
+      "loss": 3.2888,
+      "step": 59550
+    },
+    {
+      "epoch": 6.425876010781671,
+      "grad_norm": 0.7399250864982605,
+      "learning_rate": 0.0002148213707501349,
+      "loss": 3.2837,
+      "step": 59600
+    },
+    {
+      "epoch": 6.431266846361186,
+      "grad_norm": 0.7455103397369385,
+      "learning_rate": 0.00021449757150566648,
+      "loss": 3.264,
+      "step": 59650
+    },
+    {
+      "epoch": 6.436657681940701,
+      "grad_norm": 0.7875425815582275,
+      "learning_rate": 0.00021417377226119806,
+      "loss": 3.2622,
+      "step": 59700
+    },
+    {
+      "epoch": 6.442048517520216,
+      "grad_norm": 0.7843931913375854,
+      "learning_rate": 0.0002138499730167296,
+      "loss": 3.2782,
+      "step": 59750
+    },
+    {
+      "epoch": 6.44743935309973,
+      "grad_norm": 0.7751355171203613,
+      "learning_rate": 0.00021352617377226116,
+      "loss": 3.2843,
+      "step": 59800
+    },
+    {
+      "epoch": 6.452830188679245,
+      "grad_norm": 0.7746478319168091,
+      "learning_rate": 0.00021320237452779274,
+      "loss": 3.2896,
+      "step": 59850
+    },
+    {
+      "epoch": 6.45822102425876,
+      "grad_norm": 0.8506944179534912,
+      "learning_rate": 0.00021287857528332432,
+      "loss": 3.2784,
+      "step": 59900
+    },
+    {
+      "epoch": 6.463611859838275,
+      "grad_norm": 0.7042997479438782,
+      "learning_rate": 0.0002125547760388559,
+      "loss": 3.2759,
+      "step": 59950
+    },
+    {
+      "epoch": 6.46900269541779,
+      "grad_norm": 0.8357415199279785,
+      "learning_rate": 0.00021223097679438747,
+      "loss": 3.2882,
+      "step": 60000
+    },
+    {
+      "epoch": 6.46900269541779,
+      "eval_accuracy": 0.38543384174554773,
+      "eval_loss": 3.373171806335449,
+      "eval_runtime": 185.3306,
+      "eval_samples_per_second": 97.183,
+      "eval_steps_per_second": 6.076,
+      "step": 60000
+    },
+    {
+      "epoch": 6.474393530997305,
+      "grad_norm": 0.7789021134376526,
+      "learning_rate": 0.00021190717754991905,
+      "loss": 3.2919,
+      "step": 60050
+    },
+    {
+      "epoch": 6.479784366576819,
+      "grad_norm": 0.7376289963722229,
+      "learning_rate": 0.0002115833783054506,
+      "loss": 3.2784,
+      "step": 60100
+    },
+    {
+      "epoch": 6.485175202156334,
+      "grad_norm": 0.7288296222686768,
+      "learning_rate": 0.00021125957906098218,
+      "loss": 3.2763,
+      "step": 60150
+    },
+    {
+      "epoch": 6.490566037735849,
+      "grad_norm": 0.7110522389411926,
+      "learning_rate": 0.00021093577981651373,
+      "loss": 3.2611,
+      "step": 60200
+    },
+    {
+      "epoch": 6.495956873315364,
+      "grad_norm": 0.7700715661048889,
+      "learning_rate": 0.0002106119805720453,
+      "loss": 3.2893,
+      "step": 60250
+    },
+    {
+      "epoch": 6.501347708894879,
+      "grad_norm": 0.7219817638397217,
+      "learning_rate": 0.00021028818132757689,
+      "loss": 3.2778,
+      "step": 60300
+    },
+    {
+      "epoch": 6.506738544474393,
+      "grad_norm": 0.7898648977279663,
+      "learning_rate": 0.00020996438208310846,
+      "loss": 3.2772,
+      "step": 60350
+    },
+    {
+      "epoch": 6.512129380053908,
+      "grad_norm": 0.7658487558364868,
+      "learning_rate": 0.00020964058283864001,
+      "loss": 3.2698,
+      "step": 60400
+    },
+    {
+      "epoch": 6.517520215633423,
+      "grad_norm": 0.8089845180511475,
+      "learning_rate": 0.0002093167835941716,
+      "loss": 3.2709,
+      "step": 60450
+    },
+    {
+      "epoch": 6.5229110512129385,
+      "grad_norm": 0.8010355830192566,
+      "learning_rate": 0.00020899298434970317,
+      "loss": 3.2784,
+      "step": 60500
+    },
+    {
+      "epoch": 6.528301886792453,
+      "grad_norm": 0.7700260877609253,
+      "learning_rate": 0.00020866918510523475,
+      "loss": 3.2692,
+      "step": 60550
+    },
+    {
+      "epoch": 6.533692722371968,
+      "grad_norm": 0.7774828672409058,
+      "learning_rate": 0.00020834538586076632,
+      "loss": 3.279,
+      "step": 60600
+    },
+    {
+      "epoch": 6.539083557951482,
+      "grad_norm": 0.7341523170471191,
+      "learning_rate": 0.00020802158661629787,
+      "loss": 3.2821,
+      "step": 60650
+    },
+    {
+      "epoch": 6.544474393530997,
+      "grad_norm": 0.7971969246864319,
+      "learning_rate": 0.00020769778737182943,
+      "loss": 3.2581,
+      "step": 60700
+    },
+    {
+      "epoch": 6.549865229110512,
+      "grad_norm": 0.8375870585441589,
+      "learning_rate": 0.000207373988127361,
+      "loss": 3.2855,
+      "step": 60750
+    },
+    {
+      "epoch": 6.555256064690027,
+      "grad_norm": 0.7822794318199158,
+      "learning_rate": 0.00020705018888289258,
+      "loss": 3.3007,
+      "step": 60800
+    },
+    {
+      "epoch": 6.560646900269542,
+      "grad_norm": 0.7843235731124878,
+      "learning_rate": 0.00020672638963842416,
+      "loss": 3.2775,
+      "step": 60850
+    },
+    {
+      "epoch": 6.566037735849057,
+      "grad_norm": 0.7439322471618652,
+      "learning_rate": 0.00020640259039395574,
+      "loss": 3.2741,
+      "step": 60900
+    },
+    {
+      "epoch": 6.571428571428571,
+      "grad_norm": 0.7677661180496216,
+      "learning_rate": 0.00020607879114948731,
+      "loss": 3.2948,
+      "step": 60950
+    },
+    {
+      "epoch": 6.576819407008086,
+      "grad_norm": 0.7387002110481262,
+      "learning_rate": 0.0002057549919050189,
+      "loss": 3.2801,
+      "step": 61000
+    },
+    {
+      "epoch": 6.576819407008086,
+      "eval_accuracy": 0.3860143741251405,
+      "eval_loss": 3.3682520389556885,
+      "eval_runtime": 185.5315,
+      "eval_samples_per_second": 97.078,
+      "eval_steps_per_second": 6.069,
+      "step": 61000
+    },
+    {
+      "epoch": 6.5822102425876015,
+      "grad_norm": 0.7914436459541321,
+      "learning_rate": 0.00020543119266055041,
+      "loss": 3.294,
+      "step": 61050
+    },
+    {
+      "epoch": 6.587601078167116,
+      "grad_norm": 0.7589644193649292,
+      "learning_rate": 0.000205107393416082,
+      "loss": 3.2632,
+      "step": 61100
+    },
+    {
+      "epoch": 6.592991913746631,
+      "grad_norm": 0.7871928811073303,
+      "learning_rate": 0.00020479007015650294,
+      "loss": 3.3004,
+      "step": 61150
+    },
+    {
+      "epoch": 6.598382749326145,
+      "grad_norm": 0.7957470417022705,
+      "learning_rate": 0.00020446627091203452,
+      "loss": 3.2818,
+      "step": 61200
+    },
+    {
+      "epoch": 6.60377358490566,
+      "grad_norm": 0.7232425212860107,
+      "learning_rate": 0.0002041424716675661,
+      "loss": 3.2687,
+      "step": 61250
+    },
+    {
+      "epoch": 6.609164420485175,
+      "grad_norm": 0.7499574422836304,
+      "learning_rate": 0.00020381867242309768,
+      "loss": 3.2972,
+      "step": 61300
+    },
+    {
+      "epoch": 6.6145552560646905,
+      "grad_norm": 0.7162529826164246,
+      "learning_rate": 0.00020349487317862926,
+      "loss": 3.2937,
+      "step": 61350
+    },
+    {
+      "epoch": 6.619946091644205,
+      "grad_norm": 0.8264064788818359,
+      "learning_rate": 0.00020317107393416078,
+      "loss": 3.2797,
+      "step": 61400
+    },
+    {
+      "epoch": 6.62533692722372,
+      "grad_norm": 0.7366076111793518,
+      "learning_rate": 0.00020284727468969236,
+      "loss": 3.2673,
+      "step": 61450
+    },
+    {
+      "epoch": 6.630727762803234,
+      "grad_norm": 0.8459667563438416,
+      "learning_rate": 0.00020252347544522393,
+      "loss": 3.2902,
+      "step": 61500
+    },
+    {
+      "epoch": 6.636118598382749,
+      "grad_norm": 0.7387397885322571,
+      "learning_rate": 0.0002021996762007555,
+      "loss": 3.2755,
+      "step": 61550
+    },
+    {
+      "epoch": 6.6415094339622645,
+      "grad_norm": 0.7981934547424316,
+      "learning_rate": 0.0002018758769562871,
+      "loss": 3.2769,
+      "step": 61600
+    },
+    {
+      "epoch": 6.646900269541779,
+      "grad_norm": 0.7583616971969604,
+      "learning_rate": 0.00020155207771181867,
+      "loss": 3.2867,
+      "step": 61650
+    },
+    {
+      "epoch": 6.652291105121294,
+      "grad_norm": 0.8009341359138489,
+      "learning_rate": 0.00020122827846735024,
+      "loss": 3.2969,
+      "step": 61700
+    },
+    {
+      "epoch": 6.657681940700809,
+      "grad_norm": 0.7538726329803467,
+      "learning_rate": 0.00020090447922288182,
+      "loss": 3.2853,
+      "step": 61750
+    },
+    {
+      "epoch": 6.663072776280323,
+      "grad_norm": 0.7500268816947937,
+      "learning_rate": 0.00020058067997841335,
+      "loss": 3.2796,
+      "step": 61800
+    },
+    {
+      "epoch": 6.668463611859838,
+      "grad_norm": 0.8226208686828613,
+      "learning_rate": 0.00020025688073394492,
+      "loss": 3.2772,
+      "step": 61850
+    },
+    {
+      "epoch": 6.6738544474393535,
+      "grad_norm": 0.7617080211639404,
+      "learning_rate": 0.0001999330814894765,
+      "loss": 3.272,
+      "step": 61900
+    },
+    {
+      "epoch": 6.679245283018868,
+      "grad_norm": 0.8388587236404419,
+      "learning_rate": 0.00019960928224500808,
+      "loss": 3.2727,
+      "step": 61950
+    },
+    {
+      "epoch": 6.684636118598383,
+      "grad_norm": 0.7575362920761108,
+      "learning_rate": 0.00019928548300053966,
+      "loss": 3.2991,
+      "step": 62000
+    },
+    {
+      "epoch": 6.684636118598383,
+      "eval_accuracy": 0.3863006744845317,
+      "eval_loss": 3.362423896789551,
+      "eval_runtime": 185.5003,
+      "eval_samples_per_second": 97.094,
+      "eval_steps_per_second": 6.07,
+      "step": 62000
+    },
+    {
+      "epoch": 6.690026954177897,
+      "grad_norm": 0.7498572468757629,
+      "learning_rate": 0.00019896168375607123,
+      "loss": 3.2829,
+      "step": 62050
+    },
+    {
+      "epoch": 6.695417789757412,
+      "grad_norm": 0.7894664406776428,
+      "learning_rate": 0.00019863788451160278,
+      "loss": 3.261,
+      "step": 62100
+    },
+    {
+      "epoch": 6.7008086253369274,
+      "grad_norm": 0.7989409565925598,
+      "learning_rate": 0.00019831408526713436,
+      "loss": 3.3027,
+      "step": 62150
+    },
+    {
+      "epoch": 6.706199460916442,
+      "grad_norm": 0.7582972645759583,
+      "learning_rate": 0.00019799028602266594,
+      "loss": 3.2954,
+      "step": 62200
+    },
+    {
+      "epoch": 6.711590296495957,
+      "grad_norm": 0.7609683871269226,
+      "learning_rate": 0.0001976664867781975,
+      "loss": 3.2792,
+      "step": 62250
+    },
+    {
+      "epoch": 6.716981132075472,
+      "grad_norm": 0.8025651574134827,
+      "learning_rate": 0.00019734268753372907,
+      "loss": 3.2795,
+      "step": 62300
+    },
+    {
+      "epoch": 6.722371967654986,
+      "grad_norm": 0.8627368807792664,
+      "learning_rate": 0.00019701888828926065,
+      "loss": 3.3019,
+      "step": 62350
+    },
+    {
+      "epoch": 6.727762803234501,
+      "grad_norm": 0.7837508916854858,
+      "learning_rate": 0.0001966950890447922,
+      "loss": 3.2843,
+      "step": 62400
+    },
+    {
+      "epoch": 6.7331536388140165,
+      "grad_norm": 0.7840617299079895,
+      "learning_rate": 0.00019637128980032377,
+      "loss": 3.284,
+      "step": 62450
+    },
+    {
+      "epoch": 6.738544474393531,
+      "grad_norm": 0.7620540857315063,
+      "learning_rate": 0.00019604749055585535,
+      "loss": 3.2904,
+      "step": 62500
+    },
+    {
+      "epoch": 6.743935309973046,
+      "grad_norm": 0.7956631183624268,
+      "learning_rate": 0.00019572369131138693,
+      "loss": 3.2867,
+      "step": 62550
+    },
+    {
+      "epoch": 6.74932614555256,
+      "grad_norm": 0.7693567276000977,
+      "learning_rate": 0.0001953998920669185,
+      "loss": 3.278,
+      "step": 62600
+    },
+    {
+      "epoch": 6.754716981132075,
+      "grad_norm": 0.7932519316673279,
+      "learning_rate": 0.00019507609282245006,
+      "loss": 3.2808,
+      "step": 62650
+    },
+    {
+      "epoch": 6.7601078167115904,
+      "grad_norm": 0.8055360913276672,
+      "learning_rate": 0.00019475229357798164,
+      "loss": 3.2907,
+      "step": 62700
+    },
+    {
+      "epoch": 6.765498652291106,
+      "grad_norm": 0.8191863298416138,
+      "learning_rate": 0.00019442849433351319,
+      "loss": 3.269,
+      "step": 62750
+    },
+    {
+      "epoch": 6.77088948787062,
+      "grad_norm": 0.812945544719696,
+      "learning_rate": 0.00019410469508904476,
+      "loss": 3.2704,
+      "step": 62800
+    },
+    {
+      "epoch": 6.776280323450135,
+      "grad_norm": 0.7781937718391418,
+      "learning_rate": 0.00019378089584457634,
+      "loss": 3.2891,
+      "step": 62850
+    },
+    {
+      "epoch": 6.781671159029649,
+      "grad_norm": 0.8601032495498657,
+      "learning_rate": 0.00019345709660010792,
+      "loss": 3.2863,
+      "step": 62900
+    },
+    {
+      "epoch": 6.787061994609164,
+      "grad_norm": 0.7767324447631836,
+      "learning_rate": 0.0001931332973556395,
+      "loss": 3.3008,
+      "step": 62950
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "grad_norm": 0.8062757849693298,
+      "learning_rate": 0.00019280949811117107,
+      "loss": 3.2791,
+      "step": 63000
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "eval_accuracy": 0.3865814335466443,
+      "eval_loss": 3.3601419925689697,
+      "eval_runtime": 185.3224,
+      "eval_samples_per_second": 97.187,
+      "eval_steps_per_second": 6.076,
+      "step": 63000
+    },
+    {
+      "epoch": 6.797843665768194,
+      "grad_norm": 0.7886756658554077,
+      "learning_rate": 0.00019248569886670265,
+      "loss": 3.2896,
+      "step": 63050
+    },
+    {
+      "epoch": 6.803234501347709,
+      "grad_norm": 0.7697591185569763,
+      "learning_rate": 0.00019216189962223418,
+      "loss": 3.2975,
+      "step": 63100
+    },
+    {
+      "epoch": 6.808625336927224,
+      "grad_norm": 0.7850983738899231,
+      "learning_rate": 0.00019183810037776575,
+      "loss": 3.2726,
+      "step": 63150
+    },
+    {
+      "epoch": 6.814016172506738,
+      "grad_norm": 0.7991440892219543,
+      "learning_rate": 0.00019151430113329733,
+      "loss": 3.2879,
+      "step": 63200
+    },
+    {
+      "epoch": 6.819407008086253,
+      "grad_norm": 0.7532088160514832,
+      "learning_rate": 0.0001911905018888289,
+      "loss": 3.2893,
+      "step": 63250
+    },
+    {
+      "epoch": 6.824797843665769,
+      "grad_norm": 0.8358055949211121,
+      "learning_rate": 0.00019086670264436049,
+      "loss": 3.2792,
+      "step": 63300
+    },
+    {
+      "epoch": 6.830188679245283,
+      "grad_norm": 0.7301905155181885,
+      "learning_rate": 0.00019054290339989206,
+      "loss": 3.2635,
+      "step": 63350
+    },
+    {
+      "epoch": 6.835579514824798,
+      "grad_norm": 0.7704817652702332,
+      "learning_rate": 0.00019021910415542364,
+      "loss": 3.2945,
+      "step": 63400
+    },
+    {
+      "epoch": 6.840970350404312,
+      "grad_norm": 0.8155561089515686,
+      "learning_rate": 0.0001898953049109552,
+      "loss": 3.2971,
+      "step": 63450
+    },
+    {
+      "epoch": 6.846361185983827,
+      "grad_norm": 0.7301574349403381,
+      "learning_rate": 0.00018957150566648677,
+      "loss": 3.2914,
+      "step": 63500
+    },
+    {
+      "epoch": 6.8517520215633425,
+      "grad_norm": 0.8053485155105591,
+      "learning_rate": 0.00018924770642201832,
+      "loss": 3.286,
+      "step": 63550
+    },
+    {
+      "epoch": 6.857142857142857,
+      "grad_norm": 0.7629830241203308,
+      "learning_rate": 0.0001889239071775499,
+      "loss": 3.2661,
+      "step": 63600
+    },
+    {
+      "epoch": 6.862533692722372,
+      "grad_norm": 0.7930283546447754,
+      "learning_rate": 0.00018860010793308148,
+      "loss": 3.268,
+      "step": 63650
+    },
+    {
+      "epoch": 6.867924528301887,
+      "grad_norm": 0.8023165464401245,
+      "learning_rate": 0.00018827630868861305,
+      "loss": 3.277,
+      "step": 63700
+    },
+    {
+      "epoch": 6.873315363881401,
+      "grad_norm": 0.7296940684318542,
+      "learning_rate": 0.0001879525094441446,
+      "loss": 3.2752,
+      "step": 63750
+    },
+    {
+      "epoch": 6.878706199460916,
+      "grad_norm": 0.7824319005012512,
+      "learning_rate": 0.00018762871019967618,
+      "loss": 3.267,
+      "step": 63800
+    },
+    {
+      "epoch": 6.884097035040432,
+      "grad_norm": 0.7985427379608154,
+      "learning_rate": 0.00018730491095520776,
+      "loss": 3.2845,
+      "step": 63850
+    },
+    {
+      "epoch": 6.889487870619946,
+      "grad_norm": 0.7951188087463379,
+      "learning_rate": 0.00018698111171073934,
+      "loss": 3.2824,
+      "step": 63900
+    },
+    {
+      "epoch": 6.894878706199461,
+      "grad_norm": 0.8034949898719788,
+      "learning_rate": 0.0001866573124662709,
+      "loss": 3.3008,
+      "step": 63950
+    },
+    {
+      "epoch": 6.900269541778976,
+      "grad_norm": 0.7574693560600281,
+      "learning_rate": 0.00018633351322180246,
+      "loss": 3.288,
+      "step": 64000
+    },
+    {
+      "epoch": 6.900269541778976,
+      "eval_accuracy": 0.38711002984586174,
+      "eval_loss": 3.3546271324157715,
+      "eval_runtime": 185.647,
+      "eval_samples_per_second": 97.017,
+      "eval_steps_per_second": 6.065,
+      "step": 64000
+    },
+    {
+      "epoch": 6.90566037735849,
+      "grad_norm": 0.7891746759414673,
+      "learning_rate": 0.00018600971397733404,
+      "loss": 3.2859,
+      "step": 64050
+    },
+    {
+      "epoch": 6.9110512129380055,
+      "grad_norm": 0.8314854502677917,
+      "learning_rate": 0.0001856859147328656,
+      "loss": 3.2927,
+      "step": 64100
+    },
+    {
+      "epoch": 6.916442048517521,
+      "grad_norm": 0.8023466467857361,
+      "learning_rate": 0.00018536211548839717,
+      "loss": 3.2871,
+      "step": 64150
+    },
+    {
+      "epoch": 6.921832884097035,
+      "grad_norm": 0.8005847334861755,
+      "learning_rate": 0.00018503831624392875,
+      "loss": 3.2775,
+      "step": 64200
+    },
+    {
+      "epoch": 6.92722371967655,
+      "grad_norm": 0.7765096426010132,
+      "learning_rate": 0.00018471451699946033,
+      "loss": 3.2639,
+      "step": 64250
+    },
+    {
+      "epoch": 6.932614555256064,
+      "grad_norm": 0.7770959138870239,
+      "learning_rate": 0.0001843907177549919,
+      "loss": 3.3046,
+      "step": 64300
+    },
+    {
+      "epoch": 6.938005390835579,
+      "grad_norm": 0.7714718580245972,
+      "learning_rate": 0.00018406691851052348,
+      "loss": 3.2832,
+      "step": 64350
+    },
+    {
+      "epoch": 6.943396226415095,
+      "grad_norm": 0.7529462575912476,
+      "learning_rate": 0.000183743119266055,
+      "loss": 3.2839,
+      "step": 64400
+    },
+    {
+      "epoch": 6.948787061994609,
+      "grad_norm": 0.7635588049888611,
+      "learning_rate": 0.00018342579600647596,
+      "loss": 3.2811,
+      "step": 64450
+    },
+    {
+      "epoch": 6.954177897574124,
+      "grad_norm": 0.7985997200012207,
+      "learning_rate": 0.00018310199676200753,
+      "loss": 3.2737,
+      "step": 64500
+    },
+    {
+      "epoch": 6.959568733153639,
+      "grad_norm": 0.7837223410606384,
+      "learning_rate": 0.0001827781975175391,
+      "loss": 3.2801,
+      "step": 64550
+    },
+    {
+      "epoch": 6.964959568733153,
+      "grad_norm": 0.8346897959709167,
+      "learning_rate": 0.0001824543982730707,
+      "loss": 3.273,
+      "step": 64600
+    },
+    {
+      "epoch": 6.9703504043126685,
+      "grad_norm": 0.788210391998291,
+      "learning_rate": 0.00018213059902860227,
+      "loss": 3.2838,
+      "step": 64650
+    },
+    {
+      "epoch": 6.975741239892184,
+      "grad_norm": 0.7630437612533569,
+      "learning_rate": 0.00018180679978413382,
+      "loss": 3.2784,
+      "step": 64700
+    },
+    {
+      "epoch": 6.981132075471698,
+      "grad_norm": 0.8454469442367554,
+      "learning_rate": 0.00018148300053966537,
+      "loss": 3.2904,
+      "step": 64750
+    },
+    {
+      "epoch": 6.986522911051213,
+      "grad_norm": 0.7735024094581604,
+      "learning_rate": 0.00018115920129519695,
+      "loss": 3.2774,
+      "step": 64800
+    },
+    {
+      "epoch": 6.991913746630727,
+      "grad_norm": 0.7661256790161133,
+      "learning_rate": 0.00018083540205072852,
+      "loss": 3.2721,
+      "step": 64850
+    },
+    {
+      "epoch": 6.997304582210242,
+      "grad_norm": 0.8172721862792969,
+      "learning_rate": 0.0001805116028062601,
+      "loss": 3.2823,
+      "step": 64900
+    },
+    {
+      "epoch": 7.002695417789758,
+      "grad_norm": 0.836434543132782,
+      "learning_rate": 0.00018018780356179168,
+      "loss": 3.2301,
+      "step": 64950
+    },
+    {
+      "epoch": 7.008086253369272,
+      "grad_norm": 0.7450481653213501,
+      "learning_rate": 0.00017986400431732326,
+      "loss": 3.1829,
+      "step": 65000
+    },
+    {
+      "epoch": 7.008086253369272,
+      "eval_accuracy": 0.3870916675078211,
+      "eval_loss": 3.3573193550109863,
+      "eval_runtime": 185.3036,
+      "eval_samples_per_second": 97.197,
+      "eval_steps_per_second": 6.077,
+      "step": 65000
+    },
+    {
+      "epoch": 7.013477088948787,
+      "grad_norm": 0.8692488074302673,
+      "learning_rate": 0.00017954020507285483,
+      "loss": 3.2036,
+      "step": 65050
+    },
+    {
+      "epoch": 7.018867924528302,
+      "grad_norm": 0.7994346022605896,
+      "learning_rate": 0.0001792164058283864,
+      "loss": 3.201,
+      "step": 65100
+    },
+    {
+      "epoch": 7.024258760107816,
+      "grad_norm": 0.8160102963447571,
+      "learning_rate": 0.00017889260658391794,
+      "loss": 3.2111,
+      "step": 65150
+    },
+    {
+      "epoch": 7.0296495956873315,
+      "grad_norm": 0.7847365736961365,
+      "learning_rate": 0.00017856880733944951,
+      "loss": 3.2034,
+      "step": 65200
+    },
+    {
+      "epoch": 7.035040431266847,
+      "grad_norm": 0.8384340405464172,
+      "learning_rate": 0.0001782450080949811,
+      "loss": 3.2032,
+      "step": 65250
+    },
+    {
+      "epoch": 7.040431266846361,
+      "grad_norm": 0.8204463124275208,
+      "learning_rate": 0.00017792120885051267,
+      "loss": 3.1974,
+      "step": 65300
+    },
+    {
+      "epoch": 7.045822102425876,
+      "grad_norm": 0.7891581654548645,
+      "learning_rate": 0.00017759740960604425,
+      "loss": 3.214,
+      "step": 65350
+    },
+    {
+      "epoch": 7.051212938005391,
+      "grad_norm": 0.7553804516792297,
+      "learning_rate": 0.00017727361036157582,
+      "loss": 3.1921,
+      "step": 65400
+    },
+    {
+      "epoch": 7.056603773584905,
+      "grad_norm": 0.7528859972953796,
+      "learning_rate": 0.00017694981111710737,
+      "loss": 3.1798,
+      "step": 65450
+    },
+    {
+      "epoch": 7.061994609164421,
+      "grad_norm": 0.8152026534080505,
+      "learning_rate": 0.00017662601187263895,
+      "loss": 3.1979,
+      "step": 65500
+    },
+    {
+      "epoch": 7.067385444743936,
+      "grad_norm": 0.8077825903892517,
+      "learning_rate": 0.0001763022126281705,
+      "loss": 3.1878,
+      "step": 65550
+    },
+    {
+      "epoch": 7.07277628032345,
+      "grad_norm": 0.8011118769645691,
+      "learning_rate": 0.00017597841338370208,
+      "loss": 3.2192,
+      "step": 65600
+    },
+    {
+      "epoch": 7.078167115902965,
+      "grad_norm": 0.7679276466369629,
+      "learning_rate": 0.00017565461413923366,
+      "loss": 3.213,
+      "step": 65650
+    },
+    {
+      "epoch": 7.083557951482479,
+      "grad_norm": 0.8511323928833008,
+      "learning_rate": 0.00017533081489476524,
+      "loss": 3.2056,
+      "step": 65700
+    },
+    {
+      "epoch": 7.0889487870619945,
+      "grad_norm": 0.8467673659324646,
+      "learning_rate": 0.0001750070156502968,
+      "loss": 3.1899,
+      "step": 65750
+    },
+    {
+      "epoch": 7.09433962264151,
+      "grad_norm": 0.7860401272773743,
+      "learning_rate": 0.00017468321640582836,
+      "loss": 3.2195,
+      "step": 65800
+    },
+    {
+      "epoch": 7.099730458221024,
+      "grad_norm": 0.8050251007080078,
+      "learning_rate": 0.00017435941716135994,
+      "loss": 3.2154,
+      "step": 65850
+    },
+    {
+      "epoch": 7.105121293800539,
+      "grad_norm": 0.835083544254303,
+      "learning_rate": 0.00017403561791689152,
+      "loss": 3.2303,
+      "step": 65900
+    },
+    {
+      "epoch": 7.110512129380054,
+      "grad_norm": 0.7602958083152771,
+      "learning_rate": 0.0001737118186724231,
+      "loss": 3.2124,
+      "step": 65950
+    },
+    {
+      "epoch": 7.115902964959568,
+      "grad_norm": 0.851472795009613,
+      "learning_rate": 0.00017338801942795465,
+      "loss": 3.2234,
+      "step": 66000
+    },
+    {
+      "epoch": 7.115902964959568,
+      "eval_accuracy": 0.387299629135098,
+      "eval_loss": 3.360718250274658,
+      "eval_runtime": 185.5923,
+      "eval_samples_per_second": 97.046,
+      "eval_steps_per_second": 6.067,
+      "step": 66000
+    },
+    {
+      "epoch": 7.121293800539084,
+      "grad_norm": 0.8110368251800537,
+      "learning_rate": 0.00017306422018348623,
+      "loss": 3.2182,
+      "step": 66050
+    },
+    {
+      "epoch": 7.126684636118599,
+      "grad_norm": 0.8128678798675537,
+      "learning_rate": 0.00017274042093901778,
+      "loss": 3.2171,
+      "step": 66100
+    },
+    {
+      "epoch": 7.132075471698113,
+      "grad_norm": 0.8244554400444031,
+      "learning_rate": 0.00017241662169454935,
+      "loss": 3.2187,
+      "step": 66150
+    },
+    {
+      "epoch": 7.137466307277628,
+      "grad_norm": 0.7652295827865601,
+      "learning_rate": 0.00017209282245008093,
+      "loss": 3.2137,
+      "step": 66200
+    },
+    {
+      "epoch": 7.142857142857143,
+      "grad_norm": 0.840542733669281,
+      "learning_rate": 0.0001717690232056125,
+      "loss": 3.2205,
+      "step": 66250
+    },
+    {
+      "epoch": 7.1482479784366575,
+      "grad_norm": 0.795521080493927,
+      "learning_rate": 0.0001714452239611441,
+      "loss": 3.2147,
+      "step": 66300
+    },
+    {
+      "epoch": 7.153638814016173,
+      "grad_norm": 0.8073201775550842,
+      "learning_rate": 0.00017112142471667566,
+      "loss": 3.2236,
+      "step": 66350
+    },
+    {
+      "epoch": 7.159029649595688,
+      "grad_norm": 0.7793418765068054,
+      "learning_rate": 0.0001707976254722072,
+      "loss": 3.2051,
+      "step": 66400
+    },
+    {
+      "epoch": 7.164420485175202,
+      "grad_norm": 0.8074833154678345,
+      "learning_rate": 0.00017047382622773877,
+      "loss": 3.217,
+      "step": 66450
+    },
+    {
+      "epoch": 7.169811320754717,
+      "grad_norm": 0.7875019907951355,
+      "learning_rate": 0.00017015002698327034,
+      "loss": 3.2174,
+      "step": 66500
+    },
+    {
+      "epoch": 7.175202156334231,
+      "grad_norm": 0.7965167760848999,
+      "learning_rate": 0.00016982622773880192,
+      "loss": 3.22,
+      "step": 66550
+    },
+    {
+      "epoch": 7.180592991913747,
+      "grad_norm": 0.7714762687683105,
+      "learning_rate": 0.0001695024284943335,
+      "loss": 3.2062,
+      "step": 66600
+    },
+    {
+      "epoch": 7.185983827493262,
+      "grad_norm": 0.8042778372764587,
+      "learning_rate": 0.00016917862924986508,
+      "loss": 3.2157,
+      "step": 66650
+    },
+    {
+      "epoch": 7.191374663072776,
+      "grad_norm": 0.7633846402168274,
+      "learning_rate": 0.00016885483000539665,
+      "loss": 3.2174,
+      "step": 66700
+    },
+    {
+      "epoch": 7.196765498652291,
+      "grad_norm": 0.818200409412384,
+      "learning_rate": 0.00016853103076092823,
+      "loss": 3.2211,
+      "step": 66750
+    },
+    {
+      "epoch": 7.202156334231806,
+      "grad_norm": 0.8160050511360168,
+      "learning_rate": 0.00016820723151645978,
+      "loss": 3.2099,
+      "step": 66800
+    },
+    {
+      "epoch": 7.2075471698113205,
+      "grad_norm": 0.779509425163269,
+      "learning_rate": 0.0001678899082568807,
+      "loss": 3.2305,
+      "step": 66850
+    },
+    {
+      "epoch": 7.212938005390836,
+      "grad_norm": 0.7970914244651794,
+      "learning_rate": 0.00016756610901241228,
+      "loss": 3.2384,
+      "step": 66900
+    },
+    {
+      "epoch": 7.218328840970351,
+      "grad_norm": 0.8047239780426025,
+      "learning_rate": 0.00016724230976794386,
+      "loss": 3.2257,
+      "step": 66950
+    },
+    {
+      "epoch": 7.223719676549865,
+      "grad_norm": 0.7717013359069824,
+      "learning_rate": 0.00016691851052347544,
+      "loss": 3.2282,
+      "step": 67000
+    },
+    {
+      "epoch": 7.223719676549865,
+      "eval_accuracy": 0.38757376037105395,
+      "eval_loss": 3.3571853637695312,
+      "eval_runtime": 185.3414,
+      "eval_samples_per_second": 97.177,
+      "eval_steps_per_second": 6.075,
+      "step": 67000
+    },
+    {
+      "epoch": 7.22911051212938,
+      "grad_norm": 0.7892364263534546,
+      "learning_rate": 0.00016659471127900702,
+      "loss": 3.1987,
+      "step": 67050
+    },
+    {
+      "epoch": 7.234501347708895,
+      "grad_norm": 0.7793697118759155,
+      "learning_rate": 0.0001662709120345386,
+      "loss": 3.2414,
+      "step": 67100
+    },
+    {
+      "epoch": 7.2398921832884096,
+      "grad_norm": 0.8037189841270447,
+      "learning_rate": 0.00016594711279007015,
+      "loss": 3.2046,
+      "step": 67150
+    },
+    {
+      "epoch": 7.245283018867925,
+      "grad_norm": 0.8540467619895935,
+      "learning_rate": 0.0001656233135456017,
+      "loss": 3.224,
+      "step": 67200
+    },
+    {
+      "epoch": 7.250673854447439,
+      "grad_norm": 0.8086419701576233,
+      "learning_rate": 0.00016529951430113327,
+      "loss": 3.234,
+      "step": 67250
+    },
+    {
+      "epoch": 7.256064690026954,
+      "grad_norm": 0.7857149243354797,
+      "learning_rate": 0.00016497571505666485,
+      "loss": 3.237,
+      "step": 67300
+    },
+    {
+      "epoch": 7.261455525606469,
+      "grad_norm": 0.7541642785072327,
+      "learning_rate": 0.00016465191581219643,
+      "loss": 3.2163,
+      "step": 67350
+    },
+    {
+      "epoch": 7.2668463611859835,
+      "grad_norm": 0.7896484136581421,
+      "learning_rate": 0.000164328116567728,
+      "loss": 3.2406,
+      "step": 67400
+    },
+    {
+      "epoch": 7.272237196765499,
+      "grad_norm": 0.8298619389533997,
+      "learning_rate": 0.00016400431732325956,
+      "loss": 3.2366,
+      "step": 67450
+    },
+    {
+      "epoch": 7.277628032345014,
+      "grad_norm": 0.779678463935852,
+      "learning_rate": 0.00016368051807879114,
+      "loss": 3.2287,
+      "step": 67500
+    },
+    {
+      "epoch": 7.283018867924528,
+      "grad_norm": 0.8095333576202393,
+      "learning_rate": 0.0001633567188343227,
+      "loss": 3.2385,
+      "step": 67550
+    },
+    {
+      "epoch": 7.288409703504043,
+      "grad_norm": 0.837469756603241,
+      "learning_rate": 0.00016303291958985426,
+      "loss": 3.2236,
+      "step": 67600
+    },
+    {
+      "epoch": 7.293800539083558,
+      "grad_norm": 0.7788031697273254,
+      "learning_rate": 0.00016270912034538584,
+      "loss": 3.228,
+      "step": 67650
+    },
+    {
+      "epoch": 7.2991913746630726,
+      "grad_norm": 0.8133876919746399,
+      "learning_rate": 0.00016238532110091742,
+      "loss": 3.2086,
+      "step": 67700
+    },
+    {
+      "epoch": 7.304582210242588,
+      "grad_norm": 0.8191686272621155,
+      "learning_rate": 0.000162061521856449,
+      "loss": 3.2226,
+      "step": 67750
+    },
+    {
+      "epoch": 7.309973045822103,
+      "grad_norm": 0.814452588558197,
+      "learning_rate": 0.00016173772261198055,
+      "loss": 3.2222,
+      "step": 67800
+    },
+    {
+      "epoch": 7.315363881401617,
+      "grad_norm": 0.8741965889930725,
+      "learning_rate": 0.00016141392336751212,
+      "loss": 3.2219,
+      "step": 67850
+    },
+    {
+      "epoch": 7.320754716981132,
+      "grad_norm": 0.7747724056243896,
+      "learning_rate": 0.0001610901241230437,
+      "loss": 3.2371,
+      "step": 67900
+    },
+    {
+      "epoch": 7.3261455525606465,
+      "grad_norm": 0.7778286337852478,
+      "learning_rate": 0.00016076632487857528,
+      "loss": 3.213,
+      "step": 67950
+    },
+    {
+      "epoch": 7.331536388140162,
+      "grad_norm": 0.8529244065284729,
+      "learning_rate": 0.00016044252563410686,
+      "loss": 3.2267,
+      "step": 68000
+    },
+    {
+      "epoch": 7.331536388140162,
+      "eval_accuracy": 0.3877894363533657,
+      "eval_loss": 3.3537304401397705,
+      "eval_runtime": 185.5322,
+      "eval_samples_per_second": 97.077,
+      "eval_steps_per_second": 6.069,
+      "step": 68000
+    },
+    {
+      "epoch": 7.336927223719677,
+      "grad_norm": 0.7788181304931641,
+      "learning_rate": 0.0001601187263896384,
+      "loss": 3.2322,
+      "step": 68050
+    },
+    {
+      "epoch": 7.342318059299191,
+      "grad_norm": 0.826663076877594,
+      "learning_rate": 0.00015979492714516996,
+      "loss": 3.2372,
+      "step": 68100
+    },
+    {
+      "epoch": 7.347708894878706,
+      "grad_norm": 0.7754775881767273,
+      "learning_rate": 0.00015947112790070154,
+      "loss": 3.226,
+      "step": 68150
+    },
+    {
+      "epoch": 7.353099730458221,
+      "grad_norm": 0.7751636505126953,
+      "learning_rate": 0.00015914732865623311,
+      "loss": 3.2435,
+      "step": 68200
+    },
+    {
+      "epoch": 7.3584905660377355,
+      "grad_norm": 0.8280046582221985,
+      "learning_rate": 0.0001588235294117647,
+      "loss": 3.2315,
+      "step": 68250
+    },
+    {
+      "epoch": 7.363881401617251,
+      "grad_norm": 0.7916621565818787,
+      "learning_rate": 0.00015849973016729627,
+      "loss": 3.2189,
+      "step": 68300
+    },
+    {
+      "epoch": 7.369272237196766,
+      "grad_norm": 0.7694364190101624,
+      "learning_rate": 0.00015817593092282785,
+      "loss": 3.2306,
+      "step": 68350
+    },
+    {
+      "epoch": 7.37466307277628,
+      "grad_norm": 0.8083913326263428,
+      "learning_rate": 0.00015785213167835942,
+      "loss": 3.2246,
+      "step": 68400
+    },
+    {
+      "epoch": 7.380053908355795,
+      "grad_norm": 0.8321467041969299,
+      "learning_rate": 0.00015752833243389095,
+      "loss": 3.2072,
+      "step": 68450
+    },
+    {
+      "epoch": 7.38544474393531,
+      "grad_norm": 0.8279005885124207,
+      "learning_rate": 0.00015720453318942253,
+      "loss": 3.2285,
+      "step": 68500
+    },
+    {
+      "epoch": 7.390835579514825,
+      "grad_norm": 0.7874852418899536,
+      "learning_rate": 0.0001568807339449541,
+      "loss": 3.2297,
+      "step": 68550
+    },
+    {
+      "epoch": 7.39622641509434,
+      "grad_norm": 0.7745757699012756,
+      "learning_rate": 0.00015655693470048568,
+      "loss": 3.2251,
+      "step": 68600
+    },
+    {
+      "epoch": 7.401617250673855,
+      "grad_norm": 0.7697532176971436,
+      "learning_rate": 0.00015623313545601726,
+      "loss": 3.2258,
+      "step": 68650
+    },
+    {
+      "epoch": 7.407008086253369,
+      "grad_norm": 0.7562422752380371,
+      "learning_rate": 0.00015590933621154884,
+      "loss": 3.2357,
+      "step": 68700
+    },
+    {
+      "epoch": 7.412398921832884,
+      "grad_norm": 0.862626314163208,
+      "learning_rate": 0.00015558553696708041,
+      "loss": 3.2234,
+      "step": 68750
+    },
+    {
+      "epoch": 7.4177897574123985,
+      "grad_norm": 0.8310208916664124,
+      "learning_rate": 0.00015526173772261196,
+      "loss": 3.2323,
+      "step": 68800
+    },
+    {
+      "epoch": 7.423180592991914,
+      "grad_norm": 0.8276763558387756,
+      "learning_rate": 0.00015493793847814354,
+      "loss": 3.2361,
+      "step": 68850
+    },
+    {
+      "epoch": 7.428571428571429,
+      "grad_norm": 0.7850115299224854,
+      "learning_rate": 0.0001546141392336751,
+      "loss": 3.2284,
+      "step": 68900
+    },
+    {
+      "epoch": 7.433962264150943,
+      "grad_norm": 0.8095736503601074,
+      "learning_rate": 0.00015429033998920667,
+      "loss": 3.2325,
+      "step": 68950
+    },
+    {
+      "epoch": 7.439353099730458,
+      "grad_norm": 0.8068041205406189,
+      "learning_rate": 0.00015396654074473825,
+      "loss": 3.218,
+      "step": 69000
+    },
+    {
+      "epoch": 7.439353099730458,
+      "eval_accuracy": 0.3883189018756857,
+      "eval_loss": 3.3485703468322754,
+      "eval_runtime": 185.8026,
+      "eval_samples_per_second": 96.936,
+      "eval_steps_per_second": 6.06,
+      "step": 69000
+    },
+    {
+      "epoch": 7.444743935309973,
+      "grad_norm": 0.7911592125892639,
+      "learning_rate": 0.00015364274150026983,
+      "loss": 3.2181,
+      "step": 69050
+    },
+    {
+      "epoch": 7.450134770889488,
+      "grad_norm": 0.8405386805534363,
+      "learning_rate": 0.0001533189422558014,
+      "loss": 3.2278,
+      "step": 69100
+    },
+    {
+      "epoch": 7.455525606469003,
+      "grad_norm": 0.8172047138214111,
+      "learning_rate": 0.00015300161899622233,
+      "loss": 3.2291,
+      "step": 69150
+    },
+    {
+      "epoch": 7.460916442048518,
+      "grad_norm": 0.8096583485603333,
+      "learning_rate": 0.00015267781975175388,
+      "loss": 3.2192,
+      "step": 69200
+    },
+    {
+      "epoch": 7.466307277628032,
+      "grad_norm": 0.8990162014961243,
+      "learning_rate": 0.00015235402050728546,
+      "loss": 3.2372,
+      "step": 69250
+    },
+    {
+      "epoch": 7.471698113207547,
+      "grad_norm": 0.8064659237861633,
+      "learning_rate": 0.00015203022126281703,
+      "loss": 3.2242,
+      "step": 69300
+    },
+    {
+      "epoch": 7.4770889487870615,
+      "grad_norm": 0.7970738410949707,
+      "learning_rate": 0.0001517064220183486,
+      "loss": 3.2357,
+      "step": 69350
+    },
+    {
+      "epoch": 7.482479784366577,
+      "grad_norm": 0.78382807970047,
+      "learning_rate": 0.0001513826227738802,
+      "loss": 3.2216,
+      "step": 69400
+    },
+    {
+      "epoch": 7.487870619946092,
+      "grad_norm": 0.8397746086120605,
+      "learning_rate": 0.00015105882352941177,
+      "loss": 3.222,
+      "step": 69450
+    },
+    {
+      "epoch": 7.493261455525606,
+      "grad_norm": 0.8016370534896851,
+      "learning_rate": 0.00015073502428494332,
+      "loss": 3.2229,
+      "step": 69500
+    },
+    {
+      "epoch": 7.498652291105121,
+      "grad_norm": 0.8667572736740112,
+      "learning_rate": 0.0001504112250404749,
+      "loss": 3.2365,
+      "step": 69550
+    },
+    {
+      "epoch": 7.504043126684636,
+      "grad_norm": 0.7989738583564758,
+      "learning_rate": 0.00015008742579600647,
+      "loss": 3.2103,
+      "step": 69600
+    },
+    {
+      "epoch": 7.509433962264151,
+      "grad_norm": 0.8215728998184204,
+      "learning_rate": 0.00014976362655153802,
+      "loss": 3.2272,
+      "step": 69650
+    },
+    {
+      "epoch": 7.514824797843666,
+      "grad_norm": 0.7850006222724915,
+      "learning_rate": 0.0001494398273070696,
+      "loss": 3.2367,
+      "step": 69700
+    },
+    {
+      "epoch": 7.520215633423181,
+      "grad_norm": 0.8493239879608154,
+      "learning_rate": 0.00014911602806260118,
+      "loss": 3.2425,
+      "step": 69750
+    },
+    {
+      "epoch": 7.525606469002695,
+      "grad_norm": 0.8434281349182129,
+      "learning_rate": 0.00014879222881813273,
+      "loss": 3.2238,
+      "step": 69800
+    },
+    {
+      "epoch": 7.53099730458221,
+      "grad_norm": 0.8048841953277588,
+      "learning_rate": 0.0001484684295736643,
+      "loss": 3.2276,
+      "step": 69850
+    },
+    {
+      "epoch": 7.536388140161725,
+      "grad_norm": 0.8230905532836914,
+      "learning_rate": 0.00014814463032919589,
+      "loss": 3.2295,
+      "step": 69900
+    },
+    {
+      "epoch": 7.54177897574124,
+      "grad_norm": 0.855772078037262,
+      "learning_rate": 0.00014782083108472744,
+      "loss": 3.226,
+      "step": 69950
+    },
+    {
+      "epoch": 7.547169811320755,
+      "grad_norm": 0.8324517011642456,
+      "learning_rate": 0.00014749703184025901,
+      "loss": 3.2235,
+      "step": 70000
+    },
+    {
+      "epoch": 7.547169811320755,
+      "eval_accuracy": 0.3884776437447826,
+      "eval_loss": 3.3457727432250977,
+      "eval_runtime": 185.4004,
+      "eval_samples_per_second": 97.146,
+      "eval_steps_per_second": 6.073,
+      "step": 70000
+    },
+    {
+      "epoch": 7.55256064690027,
+      "grad_norm": 0.8509570956230164,
+      "learning_rate": 0.0001471732325957906,
+      "loss": 3.2324,
+      "step": 70050
+    },
+    {
+      "epoch": 7.557951482479784,
+      "grad_norm": 0.8302329182624817,
+      "learning_rate": 0.00014684943335132217,
+      "loss": 3.2349,
+      "step": 70100
+    },
+    {
+      "epoch": 7.563342318059299,
+      "grad_norm": 0.8065623641014099,
+      "learning_rate": 0.00014652563410685375,
+      "loss": 3.2328,
+      "step": 70150
+    },
+    {
+      "epoch": 7.568733153638814,
+      "grad_norm": 0.8313807249069214,
+      "learning_rate": 0.0001462018348623853,
+      "loss": 3.217,
+      "step": 70200
+    },
+    {
+      "epoch": 7.574123989218329,
+      "grad_norm": 0.8627298474311829,
+      "learning_rate": 0.00014587803561791687,
+      "loss": 3.2306,
+      "step": 70250
+    },
+    {
+      "epoch": 7.579514824797844,
+      "grad_norm": 0.7965494394302368,
+      "learning_rate": 0.00014555423637344845,
+      "loss": 3.2271,
+      "step": 70300
+    },
+    {
+      "epoch": 7.584905660377358,
+      "grad_norm": 0.8298637866973877,
+      "learning_rate": 0.00014523043712898003,
+      "loss": 3.2427,
+      "step": 70350
+    },
+    {
+      "epoch": 7.590296495956873,
+      "grad_norm": 0.8248470425605774,
+      "learning_rate": 0.00014490663788451158,
+      "loss": 3.2247,
+      "step": 70400
+    },
+    {
+      "epoch": 7.595687331536388,
+      "grad_norm": 0.787762463092804,
+      "learning_rate": 0.00014458283864004316,
+      "loss": 3.2358,
+      "step": 70450
+    },
+    {
+      "epoch": 7.601078167115903,
+      "grad_norm": 0.8552139401435852,
+      "learning_rate": 0.00014425903939557474,
+      "loss": 3.2397,
+      "step": 70500
+    },
+    {
+      "epoch": 7.606469002695418,
+      "grad_norm": 0.828631579875946,
+      "learning_rate": 0.00014393524015110631,
+      "loss": 3.2287,
+      "step": 70550
+    },
+    {
+      "epoch": 7.611859838274933,
+      "grad_norm": 0.8409473299980164,
+      "learning_rate": 0.00014361144090663786,
+      "loss": 3.2292,
+      "step": 70600
+    },
+    {
+      "epoch": 7.617250673854447,
+      "grad_norm": 0.7935662865638733,
+      "learning_rate": 0.00014328764166216944,
+      "loss": 3.2284,
+      "step": 70650
+    },
+    {
+      "epoch": 7.622641509433962,
+      "grad_norm": 0.7940767407417297,
+      "learning_rate": 0.00014296384241770102,
+      "loss": 3.2311,
+      "step": 70700
+    },
+    {
+      "epoch": 7.628032345013477,
+      "grad_norm": 0.8206398487091064,
+      "learning_rate": 0.0001426400431732326,
+      "loss": 3.2316,
+      "step": 70750
+    },
+    {
+      "epoch": 7.633423180592992,
+      "grad_norm": 0.857508659362793,
+      "learning_rate": 0.00014231624392876417,
+      "loss": 3.2439,
+      "step": 70800
+    },
+    {
+      "epoch": 7.638814016172507,
+      "grad_norm": 0.795734167098999,
+      "learning_rate": 0.00014199244468429573,
+      "loss": 3.2205,
+      "step": 70850
+    },
+    {
+      "epoch": 7.644204851752022,
+      "grad_norm": 0.8424890637397766,
+      "learning_rate": 0.0001416686454398273,
+      "loss": 3.2385,
+      "step": 70900
+    },
+    {
+      "epoch": 7.649595687331536,
+      "grad_norm": 0.824721097946167,
+      "learning_rate": 0.00014134484619535888,
+      "loss": 3.2192,
+      "step": 70950
+    },
+    {
+      "epoch": 7.654986522911051,
+      "grad_norm": 0.849852442741394,
+      "learning_rate": 0.00014102104695089043,
+      "loss": 3.2512,
+      "step": 71000
+    },
+    {
+      "epoch": 7.654986522911051,
+      "eval_accuracy": 0.3889413742699748,
+      "eval_loss": 3.342104911804199,
+      "eval_runtime": 185.5438,
+      "eval_samples_per_second": 97.071,
+      "eval_steps_per_second": 6.069,
+      "step": 71000
+    },
+    {
+      "epoch": 7.660377358490566,
+      "grad_norm": 0.7847785353660583,
+      "learning_rate": 0.000140697247706422,
+      "loss": 3.242,
+      "step": 71050
+    },
+    {
+      "epoch": 7.665768194070081,
+      "grad_norm": 0.8247793912887573,
+      "learning_rate": 0.0001403734484619536,
+      "loss": 3.2495,
+      "step": 71100
+    },
+    {
+      "epoch": 7.671159029649596,
+      "grad_norm": 0.8466809988021851,
+      "learning_rate": 0.00014004964921748514,
+      "loss": 3.2309,
+      "step": 71150
+    },
+    {
+      "epoch": 7.67654986522911,
+      "grad_norm": 0.8371172547340393,
+      "learning_rate": 0.00013972584997301671,
+      "loss": 3.2462,
+      "step": 71200
+    },
+    {
+      "epoch": 7.681940700808625,
+      "grad_norm": 0.7972482442855835,
+      "learning_rate": 0.0001394020507285483,
+      "loss": 3.2354,
+      "step": 71250
+    },
+    {
+      "epoch": 7.6873315363881405,
+      "grad_norm": 0.8721069097518921,
+      "learning_rate": 0.00013907825148407984,
+      "loss": 3.2327,
+      "step": 71300
+    },
+    {
+      "epoch": 7.692722371967655,
+      "grad_norm": 0.8658064603805542,
+      "learning_rate": 0.00013875445223961142,
+      "loss": 3.246,
+      "step": 71350
+    },
+    {
+      "epoch": 7.69811320754717,
+      "grad_norm": 0.8553919196128845,
+      "learning_rate": 0.000138430652995143,
+      "loss": 3.2504,
+      "step": 71400
+    },
+    {
+      "epoch": 7.703504043126685,
+      "grad_norm": 0.8206518888473511,
+      "learning_rate": 0.00013810685375067455,
+      "loss": 3.2292,
+      "step": 71450
+    },
+    {
+      "epoch": 7.708894878706199,
+      "grad_norm": 0.8005398511886597,
+      "learning_rate": 0.00013778305450620613,
+      "loss": 3.2377,
+      "step": 71500
+    },
+    {
+      "epoch": 7.714285714285714,
+      "grad_norm": 0.8112930059432983,
+      "learning_rate": 0.0001374592552617377,
+      "loss": 3.226,
+      "step": 71550
+    },
+    {
+      "epoch": 7.719676549865229,
+      "grad_norm": 0.8869747519493103,
+      "learning_rate": 0.00013713545601726928,
+      "loss": 3.22,
+      "step": 71600
+    },
+    {
+      "epoch": 7.725067385444744,
+      "grad_norm": 0.8616730570793152,
+      "learning_rate": 0.00013681165677280086,
+      "loss": 3.243,
+      "step": 71650
+    },
+    {
+      "epoch": 7.730458221024259,
+      "grad_norm": 0.8031706213951111,
+      "learning_rate": 0.0001364878575283324,
+      "loss": 3.2445,
+      "step": 71700
+    },
+    {
+      "epoch": 7.735849056603773,
+      "grad_norm": 0.8454236388206482,
+      "learning_rate": 0.000136164058283864,
+      "loss": 3.2418,
+      "step": 71750
+    },
+    {
+      "epoch": 7.741239892183288,
+      "grad_norm": 0.812794029712677,
+      "learning_rate": 0.00013584025903939557,
+      "loss": 3.2266,
+      "step": 71800
+    },
+    {
+      "epoch": 7.7466307277628035,
+      "grad_norm": 0.8233222365379333,
+      "learning_rate": 0.00013551645979492714,
+      "loss": 3.245,
+      "step": 71850
+    },
+    {
+      "epoch": 7.752021563342318,
+      "grad_norm": 0.7795098423957825,
+      "learning_rate": 0.0001351926605504587,
+      "loss": 3.244,
+      "step": 71900
+    },
+    {
+      "epoch": 7.757412398921833,
+      "grad_norm": 0.8167760968208313,
+      "learning_rate": 0.00013486886130599027,
+      "loss": 3.2451,
+      "step": 71950
+    },
+    {
+      "epoch": 7.762803234501348,
+      "grad_norm": 0.8600962162017822,
+      "learning_rate": 0.00013454506206152185,
+      "loss": 3.2503,
+      "step": 72000
+    },
+    {
+      "epoch": 7.762803234501348,
+      "eval_accuracy": 0.3895109327078983,
+      "eval_loss": 3.338362693786621,
+      "eval_runtime": 185.6422,
+      "eval_samples_per_second": 97.02,
+      "eval_steps_per_second": 6.065,
+      "step": 72000
+    },
+    {
+      "epoch": 7.768194070080862,
+      "grad_norm": 0.8626272082328796,
+      "learning_rate": 0.00013422126281705343,
+      "loss": 3.237,
+      "step": 72050
+    },
+    {
+      "epoch": 7.773584905660377,
+      "grad_norm": 0.8045924305915833,
+      "learning_rate": 0.00013389746357258498,
+      "loss": 3.2439,
+      "step": 72100
+    },
+    {
+      "epoch": 7.7789757412398925,
+      "grad_norm": 0.8124786019325256,
+      "learning_rate": 0.00013357366432811656,
+      "loss": 3.2279,
+      "step": 72150
+    },
+    {
+      "epoch": 7.784366576819407,
+      "grad_norm": 0.821742832660675,
+      "learning_rate": 0.00013324986508364813,
+      "loss": 3.2195,
+      "step": 72200
+    },
+    {
+      "epoch": 7.789757412398922,
+      "grad_norm": 0.8228744268417358,
+      "learning_rate": 0.0001329260658391797,
+      "loss": 3.2336,
+      "step": 72250
+    },
+    {
+      "epoch": 7.795148247978437,
+      "grad_norm": 0.813159704208374,
+      "learning_rate": 0.00013260226659471126,
+      "loss": 3.2222,
+      "step": 72300
+    },
+    {
+      "epoch": 7.800539083557951,
+      "grad_norm": 0.8435391187667847,
+      "learning_rate": 0.00013227846735024284,
+      "loss": 3.2063,
+      "step": 72350
+    },
+    {
+      "epoch": 7.8059299191374665,
+      "grad_norm": 0.8387914299964905,
+      "learning_rate": 0.00013195466810577442,
+      "loss": 3.2326,
+      "step": 72400
+    },
+    {
+      "epoch": 7.811320754716981,
+      "grad_norm": 0.8151735067367554,
+      "learning_rate": 0.000131630868861306,
+      "loss": 3.2368,
+      "step": 72450
+    },
+    {
+      "epoch": 7.816711590296496,
+      "grad_norm": 0.8179688453674316,
+      "learning_rate": 0.00013131354560172692,
+      "loss": 3.2313,
+      "step": 72500
+    },
+    {
+      "epoch": 7.822102425876011,
+      "grad_norm": 0.8553862571716309,
+      "learning_rate": 0.0001309897463572585,
+      "loss": 3.2186,
+      "step": 72550
+    },
+    {
+      "epoch": 7.827493261455525,
+      "grad_norm": 0.818985104560852,
+      "learning_rate": 0.00013066594711279007,
+      "loss": 3.2491,
+      "step": 72600
+    },
+    {
+      "epoch": 7.83288409703504,
+      "grad_norm": 0.8475792407989502,
+      "learning_rate": 0.00013034214786832162,
+      "loss": 3.2344,
+      "step": 72650
+    },
+    {
+      "epoch": 7.8382749326145555,
+      "grad_norm": 0.8516533970832825,
+      "learning_rate": 0.0001300183486238532,
+      "loss": 3.2391,
+      "step": 72700
+    },
+    {
+      "epoch": 7.84366576819407,
+      "grad_norm": 0.8353322148323059,
+      "learning_rate": 0.00012969454937938478,
+      "loss": 3.2081,
+      "step": 72750
+    },
+    {
+      "epoch": 7.849056603773585,
+      "grad_norm": 0.8211535811424255,
+      "learning_rate": 0.00012937075013491636,
+      "loss": 3.2394,
+      "step": 72800
+    },
+    {
+      "epoch": 7.8544474393531,
+      "grad_norm": 0.8574275970458984,
+      "learning_rate": 0.0001290469508904479,
+      "loss": 3.2372,
+      "step": 72850
+    },
+    {
+      "epoch": 7.859838274932614,
+      "grad_norm": 0.8538966178894043,
+      "learning_rate": 0.00012872315164597949,
+      "loss": 3.237,
+      "step": 72900
+    },
+    {
+      "epoch": 7.8652291105121295,
+      "grad_norm": 0.8543484210968018,
+      "learning_rate": 0.00012839935240151106,
+      "loss": 3.2408,
+      "step": 72950
+    },
+    {
+      "epoch": 7.870619946091644,
+      "grad_norm": 0.8760818243026733,
+      "learning_rate": 0.00012807555315704261,
+      "loss": 3.2303,
+      "step": 73000
+    },
+    {
+      "epoch": 7.870619946091644,
+      "eval_accuracy": 0.39012243116051826,
+      "eval_loss": 3.334336042404175,
+      "eval_runtime": 185.7846,
+      "eval_samples_per_second": 96.946,
+      "eval_steps_per_second": 6.061,
+      "step": 73000
+    },
+    {
+      "epoch": 7.876010781671159,
+      "grad_norm": 0.9315475821495056,
+      "learning_rate": 0.0001277517539125742,
+      "loss": 3.2402,
+      "step": 73050
+    },
+    {
+      "epoch": 7.881401617250674,
+      "grad_norm": 0.8197574615478516,
+      "learning_rate": 0.00012742795466810577,
+      "loss": 3.2364,
+      "step": 73100
+    },
+    {
+      "epoch": 7.886792452830189,
+      "grad_norm": 0.8208304643630981,
+      "learning_rate": 0.00012710415542363732,
+      "loss": 3.2348,
+      "step": 73150
+    },
+    {
+      "epoch": 7.892183288409703,
+      "grad_norm": 0.8868663907051086,
+      "learning_rate": 0.0001267803561791689,
+      "loss": 3.228,
+      "step": 73200
+    },
+    {
+      "epoch": 7.8975741239892185,
+      "grad_norm": 0.8241314888000488,
+      "learning_rate": 0.00012645655693470048,
+      "loss": 3.2348,
+      "step": 73250
+    },
+    {
+      "epoch": 7.902964959568733,
+      "grad_norm": 0.8251668810844421,
+      "learning_rate": 0.00012613275769023203,
+      "loss": 3.2396,
+      "step": 73300
+    },
+    {
+      "epoch": 7.908355795148248,
+      "grad_norm": 0.870736300945282,
+      "learning_rate": 0.0001258089584457636,
+      "loss": 3.2153,
+      "step": 73350
+    },
+    {
+      "epoch": 7.913746630727763,
+      "grad_norm": 0.8884828090667725,
+      "learning_rate": 0.00012548515920129518,
+      "loss": 3.2293,
+      "step": 73400
+    },
+    {
+      "epoch": 7.919137466307277,
+      "grad_norm": 0.8159066438674927,
+      "learning_rate": 0.00012516135995682676,
+      "loss": 3.2142,
+      "step": 73450
+    },
+    {
+      "epoch": 7.9245283018867925,
+      "grad_norm": 0.8352020978927612,
+      "learning_rate": 0.0001248375607123583,
+      "loss": 3.2314,
+      "step": 73500
+    },
+    {
+      "epoch": 7.929919137466308,
+      "grad_norm": 0.7894663214683533,
+      "learning_rate": 0.0001245137614678899,
+      "loss": 3.2215,
+      "step": 73550
+    },
+    {
+      "epoch": 7.935309973045822,
+      "grad_norm": 0.8059535026550293,
+      "learning_rate": 0.00012418996222342147,
+      "loss": 3.221,
+      "step": 73600
+    },
+    {
+      "epoch": 7.940700808625337,
+      "grad_norm": 0.8399400115013123,
+      "learning_rate": 0.00012386616297895304,
+      "loss": 3.2375,
+      "step": 73650
+    },
+    {
+      "epoch": 7.946091644204852,
+      "grad_norm": 0.8263852596282959,
+      "learning_rate": 0.0001235423637344846,
+      "loss": 3.2149,
+      "step": 73700
+    },
+    {
+      "epoch": 7.951482479784366,
+      "grad_norm": 0.8288939595222473,
+      "learning_rate": 0.00012321856449001617,
+      "loss": 3.2328,
+      "step": 73750
+    },
+    {
+      "epoch": 7.9568733153638815,
+      "grad_norm": 0.8177635073661804,
+      "learning_rate": 0.00012289476524554775,
+      "loss": 3.2332,
+      "step": 73800
+    },
+    {
+      "epoch": 7.962264150943396,
+      "grad_norm": 0.8735475540161133,
+      "learning_rate": 0.00012257096600107933,
+      "loss": 3.2179,
+      "step": 73850
+    },
+    {
+      "epoch": 7.967654986522911,
+      "grad_norm": 0.8652746677398682,
+      "learning_rate": 0.0001222471667566109,
+      "loss": 3.2224,
+      "step": 73900
+    },
+    {
+      "epoch": 7.973045822102426,
+      "grad_norm": 0.8226109743118286,
+      "learning_rate": 0.00012192336751214245,
+      "loss": 3.2273,
+      "step": 73950
+    },
+    {
+      "epoch": 7.97843665768194,
+      "grad_norm": 0.8543111681938171,
+      "learning_rate": 0.00012159956826767403,
+      "loss": 3.2506,
+      "step": 74000
+    },
+    {
+      "epoch": 7.97843665768194,
+      "eval_accuracy": 0.39014720401893993,
+      "eval_loss": 3.330275535583496,
+      "eval_runtime": 185.4384,
+      "eval_samples_per_second": 97.127,
+      "eval_steps_per_second": 6.072,
+      "step": 74000
+    },
+    {
+      "epoch": 7.9838274932614555,
+      "grad_norm": 0.9201006293296814,
+      "learning_rate": 0.00012127576902320561,
+      "loss": 3.2366,
+      "step": 74050
+    },
+    {
+      "epoch": 7.989218328840971,
+      "grad_norm": 0.8811962604522705,
+      "learning_rate": 0.00012095196977873717,
+      "loss": 3.2393,
+      "step": 74100
+    },
+    {
+      "epoch": 7.994609164420485,
+      "grad_norm": 0.8862497806549072,
+      "learning_rate": 0.00012062817053426874,
+      "loss": 3.2423,
+      "step": 74150
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 1.6737746000289917,
+      "learning_rate": 0.00012030437128980032,
+      "loss": 3.2376,
+      "step": 74200
+    },
+    {
+      "epoch": 8.005390835579515,
+      "grad_norm": 0.7967222929000854,
+      "learning_rate": 0.00011998057204533188,
+      "loss": 3.1517,
+      "step": 74250
+    },
+    {
+      "epoch": 8.01078167115903,
+      "grad_norm": 0.8131229877471924,
+      "learning_rate": 0.00011965677280086346,
+      "loss": 3.162,
+      "step": 74300
+    },
+    {
+      "epoch": 8.016172506738544,
+      "grad_norm": 0.8384157419204712,
+      "learning_rate": 0.00011933297355639502,
+      "loss": 3.1598,
+      "step": 74350
+    },
+    {
+      "epoch": 8.021563342318059,
+      "grad_norm": 0.9127787947654724,
+      "learning_rate": 0.00011900917431192659,
+      "loss": 3.1583,
+      "step": 74400
+    },
+    {
+      "epoch": 8.026954177897574,
+      "grad_norm": 0.864522397518158,
+      "learning_rate": 0.00011868537506745816,
+      "loss": 3.1603,
+      "step": 74450
+    },
+    {
+      "epoch": 8.032345013477089,
+      "grad_norm": 0.8500173091888428,
+      "learning_rate": 0.00011836157582298974,
+      "loss": 3.1775,
+      "step": 74500
+    },
+    {
+      "epoch": 8.037735849056604,
+      "grad_norm": 0.8668968081474304,
+      "learning_rate": 0.00011803777657852132,
+      "loss": 3.1564,
+      "step": 74550
+    },
+    {
+      "epoch": 8.04312668463612,
+      "grad_norm": 0.8468078374862671,
+      "learning_rate": 0.00011771397733405287,
+      "loss": 3.1771,
+      "step": 74600
+    },
+    {
+      "epoch": 8.048517520215633,
+      "grad_norm": 0.8505526185035706,
+      "learning_rate": 0.00011739017808958445,
+      "loss": 3.1743,
+      "step": 74650
+    },
+    {
+      "epoch": 8.053908355795148,
+      "grad_norm": 0.8244675993919373,
+      "learning_rate": 0.00011706637884511602,
+      "loss": 3.1622,
+      "step": 74700
+    },
+    {
+      "epoch": 8.059299191374663,
+      "grad_norm": 0.8527902364730835,
+      "learning_rate": 0.0001167425796006476,
+      "loss": 3.1502,
+      "step": 74750
+    },
+    {
+      "epoch": 8.064690026954178,
+      "grad_norm": 0.8415079116821289,
+      "learning_rate": 0.00011641878035617915,
+      "loss": 3.1693,
+      "step": 74800
+    },
+    {
+      "epoch": 8.070080862533693,
+      "grad_norm": 0.8204153180122375,
+      "learning_rate": 0.00011609498111171073,
+      "loss": 3.15,
+      "step": 74850
+    },
+    {
+      "epoch": 8.075471698113208,
+      "grad_norm": 0.7880483269691467,
+      "learning_rate": 0.00011577118186724231,
+      "loss": 3.1494,
+      "step": 74900
+    },
+    {
+      "epoch": 8.080862533692722,
+      "grad_norm": 0.8183393478393555,
+      "learning_rate": 0.00011544738262277387,
+      "loss": 3.1641,
+      "step": 74950
+    },
+    {
+      "epoch": 8.086253369272237,
+      "grad_norm": 0.8524279594421387,
+      "learning_rate": 0.00011512358337830544,
+      "loss": 3.155,
+      "step": 75000
+    },
+    {
+      "epoch": 8.086253369272237,
+      "eval_accuracy": 0.39016502309254153,
+      "eval_loss": 3.3376200199127197,
+      "eval_runtime": 185.5538,
+      "eval_samples_per_second": 97.066,
+      "eval_steps_per_second": 6.068,
+      "step": 75000
+    },
+    {
+      "epoch": 8.091644204851752,
+      "grad_norm": 0.8844255805015564,
+      "learning_rate": 0.00011479978413383701,
+      "loss": 3.1615,
+      "step": 75050
+    },
+    {
+      "epoch": 8.097035040431267,
+      "grad_norm": 0.8732317090034485,
+      "learning_rate": 0.00011447598488936858,
+      "loss": 3.1484,
+      "step": 75100
+    },
+    {
+      "epoch": 8.102425876010782,
+      "grad_norm": 0.9299524426460266,
+      "learning_rate": 0.00011415218564490016,
+      "loss": 3.1726,
+      "step": 75150
+    },
+    {
+      "epoch": 8.107816711590296,
+      "grad_norm": 0.8411099910736084,
+      "learning_rate": 0.00011382838640043172,
+      "loss": 3.1661,
+      "step": 75200
+    },
+    {
+      "epoch": 8.11320754716981,
+      "grad_norm": 0.8532853126525879,
+      "learning_rate": 0.00011350458715596328,
+      "loss": 3.1714,
+      "step": 75250
+    },
+    {
+      "epoch": 8.118598382749326,
+      "grad_norm": 0.7946939468383789,
+      "learning_rate": 0.00011318078791149486,
+      "loss": 3.1793,
+      "step": 75300
+    },
+    {
+      "epoch": 8.123989218328841,
+      "grad_norm": 0.8610842823982239,
+      "learning_rate": 0.00011285698866702644,
+      "loss": 3.1826,
+      "step": 75350
+    },
+    {
+      "epoch": 8.129380053908356,
+      "grad_norm": 0.9045069217681885,
+      "learning_rate": 0.00011253318942255802,
+      "loss": 3.185,
+      "step": 75400
+    },
+    {
+      "epoch": 8.134770889487871,
+      "grad_norm": 0.8936116099357605,
+      "learning_rate": 0.00011220939017808957,
+      "loss": 3.1709,
+      "step": 75450
+    },
+    {
+      "epoch": 8.140161725067385,
+      "grad_norm": 0.8157985806465149,
+      "learning_rate": 0.00011188559093362115,
+      "loss": 3.1754,
+      "step": 75500
+    },
+    {
+      "epoch": 8.1455525606469,
+      "grad_norm": 0.8028213381767273,
+      "learning_rate": 0.00011156179168915272,
+      "loss": 3.1595,
+      "step": 75550
+    },
+    {
+      "epoch": 8.150943396226415,
+      "grad_norm": 0.8442425727844238,
+      "learning_rate": 0.00011123799244468429,
+      "loss": 3.158,
+      "step": 75600
+    },
+    {
+      "epoch": 8.15633423180593,
+      "grad_norm": 0.8477099537849426,
+      "learning_rate": 0.00011091419320021585,
+      "loss": 3.1646,
+      "step": 75650
+    },
+    {
+      "epoch": 8.161725067385445,
+      "grad_norm": 0.871978759765625,
+      "learning_rate": 0.00011059039395574743,
+      "loss": 3.1836,
+      "step": 75700
+    },
+    {
+      "epoch": 8.167115902964959,
+      "grad_norm": 0.8002774119377136,
+      "learning_rate": 0.00011026659471127899,
+      "loss": 3.1646,
+      "step": 75750
+    },
+    {
+      "epoch": 8.172506738544474,
+      "grad_norm": 0.8479560017585754,
+      "learning_rate": 0.00010994279546681057,
+      "loss": 3.1601,
+      "step": 75800
+    },
+    {
+      "epoch": 8.177897574123989,
+      "grad_norm": 0.8153694868087769,
+      "learning_rate": 0.00010961899622234213,
+      "loss": 3.1643,
+      "step": 75850
+    },
+    {
+      "epoch": 8.183288409703504,
+      "grad_norm": 0.844157874584198,
+      "learning_rate": 0.00010929519697787371,
+      "loss": 3.1668,
+      "step": 75900
+    },
+    {
+      "epoch": 8.18867924528302,
+      "grad_norm": 0.8492550253868103,
+      "learning_rate": 0.00010897139773340528,
+      "loss": 3.1761,
+      "step": 75950
+    },
+    {
+      "epoch": 8.194070080862534,
+      "grad_norm": 0.8232309222221375,
+      "learning_rate": 0.00010864759848893685,
+      "loss": 3.1552,
+      "step": 76000
+    },
+    {
+      "epoch": 8.194070080862534,
+      "eval_accuracy": 0.39022684658570794,
+      "eval_loss": 3.3350107669830322,
+      "eval_runtime": 185.73,
+      "eval_samples_per_second": 96.974,
+      "eval_steps_per_second": 6.063,
+      "step": 76000
+    },
+    {
+      "epoch": 8.199460916442048,
+      "grad_norm": 0.9264254570007324,
+      "learning_rate": 0.00010832379924446842,
+      "loss": 3.1785,
+      "step": 76050
+    },
+    {
+      "epoch": 8.204851752021563,
+      "grad_norm": 0.8779247403144836,
+      "learning_rate": 0.00010800647598488936,
+      "loss": 3.1992,
+      "step": 76100
+    },
+    {
+      "epoch": 8.210242587601078,
+      "grad_norm": 0.8053860068321228,
+      "learning_rate": 0.00010768267674042093,
+      "loss": 3.1669,
+      "step": 76150
+    },
+    {
+      "epoch": 8.215633423180593,
+      "grad_norm": 0.87197345495224,
+      "learning_rate": 0.0001073588774959525,
+      "loss": 3.1809,
+      "step": 76200
+    },
+    {
+      "epoch": 8.221024258760108,
+      "grad_norm": 0.8206743001937866,
+      "learning_rate": 0.00010703507825148406,
+      "loss": 3.1762,
+      "step": 76250
+    },
+    {
+      "epoch": 8.226415094339623,
+      "grad_norm": 0.8729113340377808,
+      "learning_rate": 0.00010671127900701564,
+      "loss": 3.159,
+      "step": 76300
+    },
+    {
+      "epoch": 8.231805929919137,
+      "grad_norm": 0.8449000716209412,
+      "learning_rate": 0.00010638747976254722,
+      "loss": 3.1766,
+      "step": 76350
+    },
+    {
+      "epoch": 8.237196765498652,
+      "grad_norm": 0.8317974209785461,
+      "learning_rate": 0.00010606368051807877,
+      "loss": 3.1728,
+      "step": 76400
+    },
+    {
+      "epoch": 8.242587601078167,
+      "grad_norm": 0.8439301252365112,
+      "learning_rate": 0.00010573988127361035,
+      "loss": 3.1729,
+      "step": 76450
+    },
+    {
+      "epoch": 8.247978436657682,
+      "grad_norm": 0.8755947351455688,
+      "learning_rate": 0.00010541608202914192,
+      "loss": 3.1605,
+      "step": 76500
+    },
+    {
+      "epoch": 8.253369272237197,
+      "grad_norm": 0.8506140112876892,
+      "learning_rate": 0.0001050922827846735,
+      "loss": 3.1887,
+      "step": 76550
+    },
+    {
+      "epoch": 8.25876010781671,
+      "grad_norm": 0.8570579290390015,
+      "learning_rate": 0.00010476848354020505,
+      "loss": 3.1638,
+      "step": 76600
+    },
+    {
+      "epoch": 8.264150943396226,
+      "grad_norm": 0.7894355654716492,
+      "learning_rate": 0.00010444468429573663,
+      "loss": 3.161,
+      "step": 76650
+    },
+    {
+      "epoch": 8.269541778975741,
+      "grad_norm": 0.8321079015731812,
+      "learning_rate": 0.00010412088505126821,
+      "loss": 3.1819,
+      "step": 76700
+    },
+    {
+      "epoch": 8.274932614555256,
+      "grad_norm": 0.8256025910377502,
+      "learning_rate": 0.00010379708580679979,
+      "loss": 3.173,
+      "step": 76750
+    },
+    {
+      "epoch": 8.280323450134771,
+      "grad_norm": 0.8556420207023621,
+      "learning_rate": 0.00010347328656233135,
+      "loss": 3.1784,
+      "step": 76800
+    },
+    {
+      "epoch": 8.285714285714286,
+      "grad_norm": 0.8610917925834656,
+      "learning_rate": 0.00010314948731786291,
+      "loss": 3.1877,
+      "step": 76850
+    },
+    {
+      "epoch": 8.2911051212938,
+      "grad_norm": 0.8405426740646362,
+      "learning_rate": 0.00010282568807339449,
+      "loss": 3.204,
+      "step": 76900
+    },
+    {
+      "epoch": 8.296495956873315,
+      "grad_norm": 0.9084919691085815,
+      "learning_rate": 0.00010250188882892606,
+      "loss": 3.1854,
+      "step": 76950
+    },
+    {
+      "epoch": 8.30188679245283,
+      "grad_norm": 0.8620514273643494,
+      "learning_rate": 0.00010217808958445763,
+      "loss": 3.1737,
+      "step": 77000
+    },
+    {
+      "epoch": 8.30188679245283,
+      "eval_accuracy": 0.3907469679596759,
+      "eval_loss": 3.330815076828003,
+      "eval_runtime": 185.4048,
+      "eval_samples_per_second": 97.144,
+      "eval_steps_per_second": 6.073,
+      "step": 77000
+    },
+    {
+      "epoch": 8.307277628032345,
+      "grad_norm": 0.8218554258346558,
+      "learning_rate": 0.0001018542903399892,
+      "loss": 3.1878,
+      "step": 77050
+    },
+    {
+      "epoch": 8.31266846361186,
+      "grad_norm": 0.8750684857368469,
+      "learning_rate": 0.00010153049109552076,
+      "loss": 3.1723,
+      "step": 77100
+    },
+    {
+      "epoch": 8.318059299191376,
+      "grad_norm": 0.8547766208648682,
+      "learning_rate": 0.00010120669185105234,
+      "loss": 3.1601,
+      "step": 77150
+    },
+    {
+      "epoch": 8.323450134770889,
+      "grad_norm": 0.9241974949836731,
+      "learning_rate": 0.00010088289260658392,
+      "loss": 3.1607,
+      "step": 77200
+    },
+    {
+      "epoch": 8.328840970350404,
+      "grad_norm": 0.8532211780548096,
+      "learning_rate": 0.00010055909336211547,
+      "loss": 3.1603,
+      "step": 77250
+    },
+    {
+      "epoch": 8.33423180592992,
+      "grad_norm": 0.8546267151832581,
+      "learning_rate": 0.00010023529411764704,
+      "loss": 3.1686,
+      "step": 77300
+    },
+    {
+      "epoch": 8.339622641509434,
+      "grad_norm": 0.8596141934394836,
+      "learning_rate": 9.991149487317862e-05,
+      "loss": 3.1831,
+      "step": 77350
+    },
+    {
+      "epoch": 8.34501347708895,
+      "grad_norm": 0.8430399298667908,
+      "learning_rate": 9.95876956287102e-05,
+      "loss": 3.1368,
+      "step": 77400
+    },
+    {
+      "epoch": 8.350404312668463,
+      "grad_norm": 0.8879104256629944,
+      "learning_rate": 9.926389638424175e-05,
+      "loss": 3.1817,
+      "step": 77450
+    },
+    {
+      "epoch": 8.355795148247978,
+      "grad_norm": 0.8396323919296265,
+      "learning_rate": 9.894009713977333e-05,
+      "loss": 3.1877,
+      "step": 77500
+    },
+    {
+      "epoch": 8.361185983827493,
+      "grad_norm": 0.9354804754257202,
+      "learning_rate": 9.86162978953049e-05,
+      "loss": 3.1739,
+      "step": 77550
+    },
+    {
+      "epoch": 8.366576819407008,
+      "grad_norm": 0.921036422252655,
+      "learning_rate": 9.829249865083647e-05,
+      "loss": 3.1888,
+      "step": 77600
+    },
+    {
+      "epoch": 8.371967654986523,
+      "grad_norm": 0.8595395088195801,
+      "learning_rate": 9.796869940636805e-05,
+      "loss": 3.1577,
+      "step": 77650
+    },
+    {
+      "epoch": 8.377358490566039,
+      "grad_norm": 0.8779202699661255,
+      "learning_rate": 9.764490016189961e-05,
+      "loss": 3.1973,
+      "step": 77700
+    },
+    {
+      "epoch": 8.382749326145552,
+      "grad_norm": 0.8880534768104553,
+      "learning_rate": 9.732110091743119e-05,
+      "loss": 3.1751,
+      "step": 77750
+    },
+    {
+      "epoch": 8.388140161725067,
+      "grad_norm": 0.9138519763946533,
+      "learning_rate": 9.699730167296275e-05,
+      "loss": 3.1703,
+      "step": 77800
+    },
+    {
+      "epoch": 8.393530997304582,
+      "grad_norm": 0.8722093105316162,
+      "learning_rate": 9.667350242849433e-05,
+      "loss": 3.1775,
+      "step": 77850
+    },
+    {
+      "epoch": 8.398921832884097,
+      "grad_norm": 0.8912443518638611,
+      "learning_rate": 9.63497031840259e-05,
+      "loss": 3.1734,
+      "step": 77900
+    },
+    {
+      "epoch": 8.404312668463612,
+      "grad_norm": 0.9036107063293457,
+      "learning_rate": 9.602590393955746e-05,
+      "loss": 3.1824,
+      "step": 77950
+    },
+    {
+      "epoch": 8.409703504043126,
+      "grad_norm": 0.8936976790428162,
+      "learning_rate": 9.570210469508904e-05,
+      "loss": 3.167,
+      "step": 78000
+    },
+    {
+      "epoch": 8.409703504043126,
+      "eval_accuracy": 0.390828892237088,
+      "eval_loss": 3.3293895721435547,
+      "eval_runtime": 185.6367,
+      "eval_samples_per_second": 97.023,
+      "eval_steps_per_second": 6.066,
+      "step": 78000
+    },
+    {
+      "epoch": 8.415094339622641,
+      "grad_norm": 0.8334117531776428,
+      "learning_rate": 9.537830545062061e-05,
+      "loss": 3.1807,
+      "step": 78050
+    },
+    {
+      "epoch": 8.420485175202156,
+      "grad_norm": 0.8583893179893494,
+      "learning_rate": 9.505450620615217e-05,
+      "loss": 3.1683,
+      "step": 78100
+    },
+    {
+      "epoch": 8.425876010781671,
+      "grad_norm": 0.8793879151344299,
+      "learning_rate": 9.473070696168374e-05,
+      "loss": 3.1779,
+      "step": 78150
+    },
+    {
+      "epoch": 8.431266846361186,
+      "grad_norm": 0.8737629055976868,
+      "learning_rate": 9.440690771721532e-05,
+      "loss": 3.1893,
+      "step": 78200
+    },
+    {
+      "epoch": 8.436657681940702,
+      "grad_norm": 0.8955116868019104,
+      "learning_rate": 9.40831084727469e-05,
+      "loss": 3.2123,
+      "step": 78250
+    },
+    {
+      "epoch": 8.442048517520215,
+      "grad_norm": 0.8622949123382568,
+      "learning_rate": 9.375930922827845e-05,
+      "loss": 3.1808,
+      "step": 78300
+    },
+    {
+      "epoch": 8.44743935309973,
+      "grad_norm": 0.8935747742652893,
+      "learning_rate": 9.343550998381003e-05,
+      "loss": 3.1764,
+      "step": 78350
+    },
+    {
+      "epoch": 8.452830188679245,
+      "grad_norm": 0.874409019947052,
+      "learning_rate": 9.31117107393416e-05,
+      "loss": 3.1797,
+      "step": 78400
+    },
+    {
+      "epoch": 8.45822102425876,
+      "grad_norm": 0.915778636932373,
+      "learning_rate": 9.278791149487317e-05,
+      "loss": 3.1915,
+      "step": 78450
+    },
+    {
+      "epoch": 8.463611859838275,
+      "grad_norm": 0.8673213124275208,
+      "learning_rate": 9.246411225040475e-05,
+      "loss": 3.1852,
+      "step": 78500
+    },
+    {
+      "epoch": 8.46900269541779,
+      "grad_norm": 0.8483467102050781,
+      "learning_rate": 9.214031300593631e-05,
+      "loss": 3.1768,
+      "step": 78550
+    },
+    {
+      "epoch": 8.474393530997304,
+      "grad_norm": 0.8365896344184875,
+      "learning_rate": 9.181651376146787e-05,
+      "loss": 3.1854,
+      "step": 78600
+    },
+    {
+      "epoch": 8.479784366576819,
+      "grad_norm": 0.8353489637374878,
+      "learning_rate": 9.149271451699945e-05,
+      "loss": 3.175,
+      "step": 78650
+    },
+    {
+      "epoch": 8.485175202156334,
+      "grad_norm": 0.8398521542549133,
+      "learning_rate": 9.116891527253103e-05,
+      "loss": 3.1768,
+      "step": 78700
+    },
+    {
+      "epoch": 8.49056603773585,
+      "grad_norm": 0.9163660407066345,
+      "learning_rate": 9.084511602806258e-05,
+      "loss": 3.187,
+      "step": 78750
+    },
+    {
+      "epoch": 8.495956873315365,
+      "grad_norm": 0.8956058621406555,
+      "learning_rate": 9.052131678359416e-05,
+      "loss": 3.1911,
+      "step": 78800
+    },
+    {
+      "epoch": 8.501347708894878,
+      "grad_norm": 0.8277040719985962,
+      "learning_rate": 9.019751753912574e-05,
+      "loss": 3.1885,
+      "step": 78850
+    },
+    {
+      "epoch": 8.506738544474393,
+      "grad_norm": 0.8544255495071411,
+      "learning_rate": 8.987371829465731e-05,
+      "loss": 3.1792,
+      "step": 78900
+    },
+    {
+      "epoch": 8.512129380053908,
+      "grad_norm": 0.8472200036048889,
+      "learning_rate": 8.954991905018886e-05,
+      "loss": 3.1771,
+      "step": 78950
+    },
+    {
+      "epoch": 8.517520215633423,
+      "grad_norm": 0.8413991928100586,
+      "learning_rate": 8.922611980572044e-05,
+      "loss": 3.1798,
+      "step": 79000
+    },
+    {
+      "epoch": 8.517520215633423,
+      "eval_accuracy": 0.39121558786481975,
+      "eval_loss": 3.324694871902466,
+      "eval_runtime": 185.6121,
+      "eval_samples_per_second": 97.036,
+      "eval_steps_per_second": 6.066,
+      "step": 79000
+    },
+    {
+      "epoch": 8.522911051212938,
+      "grad_norm": 0.855174720287323,
+      "learning_rate": 8.890232056125202e-05,
+      "loss": 3.1962,
+      "step": 79050
+    },
+    {
+      "epoch": 8.528301886792454,
+      "grad_norm": 0.9736406803131104,
+      "learning_rate": 8.85785213167836e-05,
+      "loss": 3.1796,
+      "step": 79100
+    },
+    {
+      "epoch": 8.533692722371967,
+      "grad_norm": 0.8685256242752075,
+      "learning_rate": 8.825472207231516e-05,
+      "loss": 3.1868,
+      "step": 79150
+    },
+    {
+      "epoch": 8.539083557951482,
+      "grad_norm": 0.9153226613998413,
+      "learning_rate": 8.793092282784672e-05,
+      "loss": 3.2056,
+      "step": 79200
+    },
+    {
+      "epoch": 8.544474393530997,
+      "grad_norm": 0.8617915511131287,
+      "learning_rate": 8.76071235833783e-05,
+      "loss": 3.1855,
+      "step": 79250
+    },
+    {
+      "epoch": 8.549865229110512,
+      "grad_norm": 0.8766077756881714,
+      "learning_rate": 8.728332433890987e-05,
+      "loss": 3.163,
+      "step": 79300
+    },
+    {
+      "epoch": 8.555256064690028,
+      "grad_norm": 0.8733969330787659,
+      "learning_rate": 8.695952509444144e-05,
+      "loss": 3.183,
+      "step": 79350
+    },
+    {
+      "epoch": 8.560646900269543,
+      "grad_norm": 0.9204221963882446,
+      "learning_rate": 8.663572584997301e-05,
+      "loss": 3.1797,
+      "step": 79400
+    },
+    {
+      "epoch": 8.566037735849056,
+      "grad_norm": 0.859405517578125,
+      "learning_rate": 8.631192660550457e-05,
+      "loss": 3.1643,
+      "step": 79450
+    },
+    {
+      "epoch": 8.571428571428571,
+      "grad_norm": 0.8825210928916931,
+      "learning_rate": 8.598812736103615e-05,
+      "loss": 3.174,
+      "step": 79500
+    },
+    {
+      "epoch": 8.576819407008086,
+      "grad_norm": 0.9521044492721558,
+      "learning_rate": 8.566432811656773e-05,
+      "loss": 3.1751,
+      "step": 79550
+    },
+    {
+      "epoch": 8.582210242587601,
+      "grad_norm": 0.8510432243347168,
+      "learning_rate": 8.534052887209928e-05,
+      "loss": 3.1599,
+      "step": 79600
+    },
+    {
+      "epoch": 8.587601078167117,
+      "grad_norm": 0.8742794394493103,
+      "learning_rate": 8.501672962763086e-05,
+      "loss": 3.1736,
+      "step": 79650
+    },
+    {
+      "epoch": 8.59299191374663,
+      "grad_norm": 0.8450675010681152,
+      "learning_rate": 8.469293038316243e-05,
+      "loss": 3.1715,
+      "step": 79700
+    },
+    {
+      "epoch": 8.598382749326145,
+      "grad_norm": 0.8777676224708557,
+      "learning_rate": 8.436913113869401e-05,
+      "loss": 3.1811,
+      "step": 79750
+    },
+    {
+      "epoch": 8.60377358490566,
+      "grad_norm": 0.8752297759056091,
+      "learning_rate": 8.404533189422556e-05,
+      "loss": 3.1833,
+      "step": 79800
+    },
+    {
+      "epoch": 8.609164420485175,
+      "grad_norm": 0.8393634557723999,
+      "learning_rate": 8.372153264975714e-05,
+      "loss": 3.1636,
+      "step": 79850
+    },
+    {
+      "epoch": 8.61455525606469,
+      "grad_norm": 0.8718554377555847,
+      "learning_rate": 8.339773340528872e-05,
+      "loss": 3.2053,
+      "step": 79900
+    },
+    {
+      "epoch": 8.619946091644206,
+      "grad_norm": 0.8281534910202026,
+      "learning_rate": 8.307393416082028e-05,
+      "loss": 3.1834,
+      "step": 79950
+    },
+    {
+      "epoch": 8.625336927223719,
+      "grad_norm": 0.8530157804489136,
+      "learning_rate": 8.275013491635186e-05,
+      "loss": 3.1957,
+      "step": 80000
+    },
+    {
+      "epoch": 8.625336927223719,
+      "eval_accuracy": 0.3918415371515189,
+      "eval_loss": 3.3219223022460938,
+      "eval_runtime": 185.7657,
+      "eval_samples_per_second": 96.955,
+      "eval_steps_per_second": 6.061,
+      "step": 80000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 92750,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.688553435136e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}