diff --git "a/checkpoint-30000/trainer_state.json" "b/checkpoint-30000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-30000/trainer_state.json"
@@ -0,0 +1,4503 @@
+{
+  "best_metric": 3.490234613418579,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M__8397/checkpoint-30000",
+  "epoch": 3.234501347708895,
+  "eval_steps": 1000,
+  "global_step": 30000,
+  "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
+    }
+  ],
+  "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": 2.508207538176e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}