diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,13876 @@
+{
+  "best_metric": 3.3076210021972656,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M_8397/checkpoint-90000",
+  "epoch": 10.0,
+  "eval_steps": 1000,
+  "global_step": 92910,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005381552039608223,
+      "grad_norm": 1.1363136768341064,
+      "learning_rate": 0.0003,
+      "loss": 8.4632,
+      "step": 50
+    },
+    {
+      "epoch": 0.010763104079216447,
+      "grad_norm": 3.25538969039917,
+      "learning_rate": 0.0006,
+      "loss": 6.8415,
+      "step": 100
+    },
+    {
+      "epoch": 0.01614465611882467,
+      "grad_norm": 3.1040453910827637,
+      "learning_rate": 0.0005996767589699385,
+      "loss": 6.4387,
+      "step": 150
+    },
+    {
+      "epoch": 0.021526208158432893,
+      "grad_norm": 3.3955957889556885,
+      "learning_rate": 0.0005993535179398771,
+      "loss": 6.1922,
+      "step": 200
+    },
+    {
+      "epoch": 0.026907760198041114,
+      "grad_norm": 2.7429745197296143,
+      "learning_rate": 0.0005990302769098158,
+      "loss": 6.0811,
+      "step": 250
+    },
+    {
+      "epoch": 0.03228931223764934,
+      "grad_norm": 1.5651514530181885,
+      "learning_rate": 0.0005987070358797543,
+      "loss": 5.9862,
+      "step": 300
+    },
+    {
+      "epoch": 0.03767086427725756,
+      "grad_norm": 1.5071296691894531,
+      "learning_rate": 0.0005983837948496929,
+      "loss": 5.8565,
+      "step": 350
+    },
+    {
+      "epoch": 0.04305241631686579,
+      "grad_norm": 1.3275337219238281,
+      "learning_rate": 0.0005980605538196314,
+      "loss": 5.8096,
+      "step": 400
+    },
+    {
+      "epoch": 0.048433968356474004,
+      "grad_norm": 1.523363709449768,
+      "learning_rate": 0.0005977373127895701,
+      "loss": 5.7397,
+      "step": 450
+    },
+    {
+      "epoch": 0.05381552039608223,
+      "grad_norm": 1.786197304725647,
+      "learning_rate": 0.0005974140717595086,
+      "loss": 5.6499,
+      "step": 500
+    },
+    {
+      "epoch": 0.05919707243569045,
+      "grad_norm": 1.4374873638153076,
+      "learning_rate": 0.0005970908307294472,
+      "loss": 5.5871,
+      "step": 550
+    },
+    {
+      "epoch": 0.06457862447529868,
+      "grad_norm": 1.401210069656372,
+      "learning_rate": 0.0005967675896993858,
+      "loss": 5.5076,
+      "step": 600
+    },
+    {
+      "epoch": 0.0699601765149069,
+      "grad_norm": 1.2165864706039429,
+      "learning_rate": 0.0005964443486693243,
+      "loss": 5.4396,
+      "step": 650
+    },
+    {
+      "epoch": 0.07534172855451512,
+      "grad_norm": 1.5090405941009521,
+      "learning_rate": 0.000596121107639263,
+      "loss": 5.3853,
+      "step": 700
+    },
+    {
+      "epoch": 0.08072328059412334,
+      "grad_norm": 1.6507272720336914,
+      "learning_rate": 0.0005957978666092015,
+      "loss": 5.3282,
+      "step": 750
+    },
+    {
+      "epoch": 0.08610483263373157,
+      "grad_norm": 1.5006401538848877,
+      "learning_rate": 0.0005954746255791401,
+      "loss": 5.2854,
+      "step": 800
+    },
+    {
+      "epoch": 0.09148638467333979,
+      "grad_norm": 1.1712360382080078,
+      "learning_rate": 0.0005951513845490787,
+      "loss": 5.265,
+      "step": 850
+    },
+    {
+      "epoch": 0.09686793671294801,
+      "grad_norm": 1.2654850482940674,
+      "learning_rate": 0.0005948281435190174,
+      "loss": 5.1858,
+      "step": 900
+    },
+    {
+      "epoch": 0.10224948875255624,
+      "grad_norm": 1.3258107900619507,
+      "learning_rate": 0.0005945049024889559,
+      "loss": 5.1631,
+      "step": 950
+    },
+    {
+      "epoch": 0.10763104079216446,
+      "grad_norm": 1.1906688213348389,
+      "learning_rate": 0.0005941816614588944,
+      "loss": 5.1052,
+      "step": 1000
+    },
+    {
+      "epoch": 0.10763104079216446,
+      "eval_accuracy": 0.22547332185886404,
+      "eval_loss": 5.034008502960205,
+      "eval_runtime": 183.4102,
+      "eval_samples_per_second": 98.201,
+      "eval_steps_per_second": 6.139,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11301259283177269,
+      "grad_norm": 0.88250333070755,
+      "learning_rate": 0.000593858420428833,
+      "loss": 5.0516,
+      "step": 1050
+    },
+    {
+      "epoch": 0.1183941448713809,
+      "grad_norm": 0.965934693813324,
+      "learning_rate": 0.0005935351793987716,
+      "loss": 5.0322,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12377569691098914,
+      "grad_norm": 1.1457676887512207,
+      "learning_rate": 0.0005932119383687103,
+      "loss": 5.0043,
+      "step": 1150
+    },
+    {
+      "epoch": 0.12915724895059735,
+      "grad_norm": 1.3504283428192139,
+      "learning_rate": 0.0005928886973386488,
+      "loss": 4.9543,
+      "step": 1200
+    },
+    {
+      "epoch": 0.13453880099020557,
+      "grad_norm": 1.02091383934021,
+      "learning_rate": 0.0005925654563085874,
+      "loss": 4.9343,
+      "step": 1250
+    },
+    {
+      "epoch": 0.1399203530298138,
+      "grad_norm": 0.9588813781738281,
+      "learning_rate": 0.000592242215278526,
+      "loss": 4.8984,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14530190506942203,
+      "grad_norm": 0.8982344269752502,
+      "learning_rate": 0.0005919189742484645,
+      "loss": 4.8657,
+      "step": 1350
+    },
+    {
+      "epoch": 0.15068345710903025,
+      "grad_norm": 1.053356647491455,
+      "learning_rate": 0.0005915957332184032,
+      "loss": 4.8735,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15606500914863847,
+      "grad_norm": 1.0207793712615967,
+      "learning_rate": 0.0005912724921883417,
+      "loss": 4.8455,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16144656118824668,
+      "grad_norm": 0.9741514325141907,
+      "learning_rate": 0.0005909492511582803,
+      "loss": 4.8106,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1668281132278549,
+      "grad_norm": 1.1351131200790405,
+      "learning_rate": 0.0005906260101282189,
+      "loss": 4.7854,
+      "step": 1550
+    },
+    {
+      "epoch": 0.17220966526746315,
+      "grad_norm": 0.8748419880867004,
+      "learning_rate": 0.0005903027690981575,
+      "loss": 4.7504,
+      "step": 1600
+    },
+    {
+      "epoch": 0.17759121730707136,
+      "grad_norm": 0.7746927738189697,
+      "learning_rate": 0.000589979528068096,
+      "loss": 4.7232,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18297276934667958,
+      "grad_norm": 1.0927244424819946,
+      "learning_rate": 0.0005896562870380347,
+      "loss": 4.7024,
+      "step": 1700
+    },
+    {
+      "epoch": 0.1883543213862878,
+      "grad_norm": 1.1120930910110474,
+      "learning_rate": 0.0005893330460079732,
+      "loss": 4.7158,
+      "step": 1750
+    },
+    {
+      "epoch": 0.19373587342589602,
+      "grad_norm": 0.7173694372177124,
+      "learning_rate": 0.0005890098049779118,
+      "loss": 4.684,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19911742546550426,
+      "grad_norm": 0.964622974395752,
+      "learning_rate": 0.0005886865639478504,
+      "loss": 4.668,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20449897750511248,
+      "grad_norm": 0.8760665059089661,
+      "learning_rate": 0.0005883633229177889,
+      "loss": 4.64,
+      "step": 1900
+    },
+    {
+      "epoch": 0.2098805295447207,
+      "grad_norm": 1.0075942277908325,
+      "learning_rate": 0.0005880400818877276,
+      "loss": 4.6187,
+      "step": 1950
+    },
+    {
+      "epoch": 0.2152620815843289,
+      "grad_norm": 1.170619249343872,
+      "learning_rate": 0.0005877168408576662,
+      "loss": 4.6124,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2152620815843289,
+      "eval_accuracy": 0.2689746785531477,
+      "eval_loss": 4.52186918258667,
+      "eval_runtime": 182.9507,
+      "eval_samples_per_second": 98.447,
+      "eval_steps_per_second": 6.155,
+      "step": 2000
+    },
+    {
+      "epoch": 0.22064363362393713,
+      "grad_norm": 1.1905301809310913,
+      "learning_rate": 0.0005873935998276048,
+      "loss": 4.5808,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22602518566354537,
+      "grad_norm": 0.7720122933387756,
+      "learning_rate": 0.0005870703587975433,
+      "loss": 4.5765,
+      "step": 2100
+    },
+    {
+      "epoch": 0.2314067377031536,
+      "grad_norm": 0.9038923978805542,
+      "learning_rate": 0.0005867471177674818,
+      "loss": 4.5495,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2367882897427618,
+      "grad_norm": 0.954218327999115,
+      "learning_rate": 0.0005864238767374205,
+      "loss": 4.5188,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24216984178237003,
+      "grad_norm": 0.890593409538269,
+      "learning_rate": 0.0005861006357073591,
+      "loss": 4.5113,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24755139382197827,
+      "grad_norm": 1.0826386213302612,
+      "learning_rate": 0.0005857773946772977,
+      "loss": 4.5139,
+      "step": 2300
+    },
+    {
+      "epoch": 0.2529329458615865,
+      "grad_norm": 1.0157872438430786,
+      "learning_rate": 0.0005854541536472362,
+      "loss": 4.4663,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2583144979011947,
+      "grad_norm": 1.0314841270446777,
+      "learning_rate": 0.0005851309126171749,
+      "loss": 4.4576,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2636960499408029,
+      "grad_norm": 1.0260419845581055,
+      "learning_rate": 0.0005848076715871134,
+      "loss": 4.4447,
+      "step": 2450
+    },
+    {
+      "epoch": 0.26907760198041114,
+      "grad_norm": 0.8484777212142944,
+      "learning_rate": 0.000584484430557052,
+      "loss": 4.4135,
+      "step": 2500
+    },
+    {
+      "epoch": 0.27445915402001936,
+      "grad_norm": 0.9077085256576538,
+      "learning_rate": 0.0005841611895269906,
+      "loss": 4.4272,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2798407060596276,
+      "grad_norm": 0.7669521570205688,
+      "learning_rate": 0.0005838379484969291,
+      "loss": 4.402,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2852222580992358,
+      "grad_norm": 0.7865301966667175,
+      "learning_rate": 0.0005835147074668678,
+      "loss": 4.406,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29060381013884407,
+      "grad_norm": 0.9062381386756897,
+      "learning_rate": 0.0005831914664368063,
+      "loss": 4.3869,
+      "step": 2700
+    },
+    {
+      "epoch": 0.2959853621784523,
+      "grad_norm": 0.8867931962013245,
+      "learning_rate": 0.0005828682254067449,
+      "loss": 4.367,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3013669142180605,
+      "grad_norm": 0.7799649834632874,
+      "learning_rate": 0.0005825449843766835,
+      "loss": 4.3764,
+      "step": 2800
+    },
+    {
+      "epoch": 0.3067484662576687,
+      "grad_norm": 0.998032808303833,
+      "learning_rate": 0.0005822217433466221,
+      "loss": 4.3453,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31213001829727693,
+      "grad_norm": 0.7040951251983643,
+      "learning_rate": 0.0005818985023165607,
+      "loss": 4.3335,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31751157033688515,
+      "grad_norm": 0.8604708313941956,
+      "learning_rate": 0.0005815752612864992,
+      "loss": 4.3286,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32289312237649337,
+      "grad_norm": 0.6292029619216919,
+      "learning_rate": 0.0005812520202564378,
+      "loss": 4.3286,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32289312237649337,
+      "eval_accuracy": 0.2978722179020627,
+      "eval_loss": 4.24153995513916,
+      "eval_runtime": 183.0117,
+      "eval_samples_per_second": 98.414,
+      "eval_steps_per_second": 6.153,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3282746744161016,
+      "grad_norm": 0.9364116787910461,
+      "learning_rate": 0.0005809287792263764,
+      "loss": 4.3113,
+      "step": 3050
+    },
+    {
+      "epoch": 0.3336562264557098,
+      "grad_norm": 0.6825273633003235,
+      "learning_rate": 0.0005806055381963151,
+      "loss": 4.3019,
+      "step": 3100
+    },
+    {
+      "epoch": 0.3390377784953181,
+      "grad_norm": 0.6769887208938599,
+      "learning_rate": 0.0005802822971662536,
+      "loss": 4.287,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3444193305349263,
+      "grad_norm": 0.7461885809898376,
+      "learning_rate": 0.0005799590561361922,
+      "loss": 4.2488,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3498008825745345,
+      "grad_norm": 0.8458229899406433,
+      "learning_rate": 0.0005796358151061307,
+      "loss": 4.2704,
+      "step": 3250
+    },
+    {
+      "epoch": 0.35518243461414273,
+      "grad_norm": 0.7842735052108765,
+      "learning_rate": 0.0005793125740760694,
+      "loss": 4.2467,
+      "step": 3300
+    },
+    {
+      "epoch": 0.36056398665375095,
+      "grad_norm": 0.6246998310089111,
+      "learning_rate": 0.0005789893330460079,
+      "loss": 4.2382,
+      "step": 3350
+    },
+    {
+      "epoch": 0.36594553869335916,
+      "grad_norm": 0.7485172748565674,
+      "learning_rate": 0.0005786660920159465,
+      "loss": 4.2417,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3713270907329674,
+      "grad_norm": 0.8092179894447327,
+      "learning_rate": 0.0005783428509858851,
+      "loss": 4.2616,
+      "step": 3450
+    },
+    {
+      "epoch": 0.3767086427725756,
+      "grad_norm": 1.0175601243972778,
+      "learning_rate": 0.0005780196099558237,
+      "loss": 4.2304,
+      "step": 3500
+    },
+    {
+      "epoch": 0.3820901948121838,
+      "grad_norm": 0.6826550960540771,
+      "learning_rate": 0.0005776963689257623,
+      "loss": 4.2206,
+      "step": 3550
+    },
+    {
+      "epoch": 0.38747174685179203,
+      "grad_norm": 0.7783133387565613,
+      "learning_rate": 0.0005773731278957008,
+      "loss": 4.2282,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3928532988914003,
+      "grad_norm": 0.6485751867294312,
+      "learning_rate": 0.0005770498868656394,
+      "loss": 4.2179,
+      "step": 3650
+    },
+    {
+      "epoch": 0.3982348509310085,
+      "grad_norm": 0.778927206993103,
+      "learning_rate": 0.000576726645835578,
+      "loss": 4.1895,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40361640297061674,
+      "grad_norm": 0.8899291753768921,
+      "learning_rate": 0.0005764034048055167,
+      "loss": 4.2006,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40899795501022496,
+      "grad_norm": 0.6624945998191833,
+      "learning_rate": 0.0005760801637754552,
+      "loss": 4.2022,
+      "step": 3800
+    },
+    {
+      "epoch": 0.4143795070498332,
+      "grad_norm": 0.7508805394172668,
+      "learning_rate": 0.0005757569227453937,
+      "loss": 4.1884,
+      "step": 3850
+    },
+    {
+      "epoch": 0.4197610590894414,
+      "grad_norm": 0.6273350119590759,
+      "learning_rate": 0.0005754336817153324,
+      "loss": 4.1653,
+      "step": 3900
+    },
+    {
+      "epoch": 0.4251426111290496,
+      "grad_norm": 0.7216036319732666,
+      "learning_rate": 0.0005751104406852709,
+      "loss": 4.168,
+      "step": 3950
+    },
+    {
+      "epoch": 0.4305241631686578,
+      "grad_norm": 0.6389856934547424,
+      "learning_rate": 0.0005747871996552096,
+      "loss": 4.1493,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4305241631686578,
+      "eval_accuracy": 0.31214225357606534,
+      "eval_loss": 4.097915172576904,
+      "eval_runtime": 182.793,
+      "eval_samples_per_second": 98.532,
+      "eval_steps_per_second": 6.16,
+      "step": 4000
+    },
+    {
+      "epoch": 0.43590571520826604,
+      "grad_norm": 0.8230230212211609,
+      "learning_rate": 0.0005744639586251481,
+      "loss": 4.1597,
+      "step": 4050
+    },
+    {
+      "epoch": 0.44128726724787426,
+      "grad_norm": 0.7011985182762146,
+      "learning_rate": 0.0005741407175950867,
+      "loss": 4.1713,
+      "step": 4100
+    },
+    {
+      "epoch": 0.44666881928748253,
+      "grad_norm": 0.6618315577507019,
+      "learning_rate": 0.0005738174765650253,
+      "loss": 4.156,
+      "step": 4150
+    },
+    {
+      "epoch": 0.45205037132709075,
+      "grad_norm": 0.652692973613739,
+      "learning_rate": 0.0005734942355349638,
+      "loss": 4.1499,
+      "step": 4200
+    },
+    {
+      "epoch": 0.45743192336669897,
+      "grad_norm": 0.6924708485603333,
+      "learning_rate": 0.0005731709945049025,
+      "loss": 4.1527,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4628134754063072,
+      "grad_norm": 0.7289921641349792,
+      "learning_rate": 0.000572847753474841,
+      "loss": 4.1386,
+      "step": 4300
+    },
+    {
+      "epoch": 0.4681950274459154,
+      "grad_norm": 0.6634368896484375,
+      "learning_rate": 0.0005725245124447796,
+      "loss": 4.1338,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4735765794855236,
+      "grad_norm": 0.6518203616142273,
+      "learning_rate": 0.0005722012714147182,
+      "loss": 4.1474,
+      "step": 4400
+    },
+    {
+      "epoch": 0.47895813152513184,
+      "grad_norm": 0.5596990585327148,
+      "learning_rate": 0.0005718780303846568,
+      "loss": 4.115,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48433968356474005,
+      "grad_norm": 0.5837390422821045,
+      "learning_rate": 0.0005715547893545953,
+      "loss": 4.0993,
+      "step": 4500
+    },
+    {
+      "epoch": 0.48972123560434827,
+      "grad_norm": 0.5811730027198792,
+      "learning_rate": 0.000571231548324534,
+      "loss": 4.099,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49510278764395654,
+      "grad_norm": 0.5709436535835266,
+      "learning_rate": 0.0005709083072944725,
+      "loss": 4.1202,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5004843396835648,
+      "grad_norm": 0.6790245175361633,
+      "learning_rate": 0.0005705850662644111,
+      "loss": 4.0862,
+      "step": 4650
+    },
+    {
+      "epoch": 0.505865891723173,
+      "grad_norm": 0.629574179649353,
+      "learning_rate": 0.0005702618252343497,
+      "loss": 4.1059,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5112474437627812,
+      "grad_norm": 0.5848677754402161,
+      "learning_rate": 0.0005699385842042882,
+      "loss": 4.0903,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5166289958023894,
+      "grad_norm": 0.6429191827774048,
+      "learning_rate": 0.0005696153431742269,
+      "loss": 4.0856,
+      "step": 4800
+    },
+    {
+      "epoch": 0.5220105478419976,
+      "grad_norm": 0.6784705519676208,
+      "learning_rate": 0.0005692921021441655,
+      "loss": 4.0779,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5273920998816058,
+      "grad_norm": 0.6737756133079529,
+      "learning_rate": 0.0005689688611141041,
+      "loss": 4.0737,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5327736519212141,
+      "grad_norm": 0.614524781703949,
+      "learning_rate": 0.0005686456200840426,
+      "loss": 4.055,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5381552039608223,
+      "grad_norm": 0.5657824277877808,
+      "learning_rate": 0.0005683223790539811,
+      "loss": 4.0522,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5381552039608223,
+      "eval_accuracy": 0.3199428790038182,
+      "eval_loss": 4.0021162033081055,
+      "eval_runtime": 183.0646,
+      "eval_samples_per_second": 98.386,
+      "eval_steps_per_second": 6.151,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5435367560004305,
+      "grad_norm": 0.7021715044975281,
+      "learning_rate": 0.0005679991380239198,
+      "loss": 4.0655,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5489183080400387,
+      "grad_norm": 0.796134352684021,
+      "learning_rate": 0.0005676758969938584,
+      "loss": 4.0578,
+      "step": 5100
+    },
+    {
+      "epoch": 0.5542998600796469,
+      "grad_norm": 0.6416186094284058,
+      "learning_rate": 0.000567352655963797,
+      "loss": 4.0523,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5596814121192552,
+      "grad_norm": 0.6666920185089111,
+      "learning_rate": 0.0005670294149337355,
+      "loss": 4.0536,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5650629641588634,
+      "grad_norm": 0.6438349485397339,
+      "learning_rate": 0.0005667061739036742,
+      "loss": 4.0504,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5704445161984716,
+      "grad_norm": 0.5683118104934692,
+      "learning_rate": 0.0005663829328736127,
+      "loss": 4.0581,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5758260682380799,
+      "grad_norm": 0.6448424458503723,
+      "learning_rate": 0.0005660596918435512,
+      "loss": 4.0476,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5812076202776881,
+      "grad_norm": 0.6892325282096863,
+      "learning_rate": 0.0005657364508134899,
+      "loss": 4.0357,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5865891723172963,
+      "grad_norm": 0.6151754260063171,
+      "learning_rate": 0.0005654132097834284,
+      "loss": 4.0408,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5919707243569046,
+      "grad_norm": 0.5521050095558167,
+      "learning_rate": 0.0005650899687533671,
+      "loss": 4.0438,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5973522763965128,
+      "grad_norm": 0.5622237324714661,
+      "learning_rate": 0.0005647667277233056,
+      "loss": 4.0191,
+      "step": 5550
+    },
+    {
+      "epoch": 0.602733828436121,
+      "grad_norm": 0.6457247734069824,
+      "learning_rate": 0.0005644434866932442,
+      "loss": 4.0255,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6081153804757292,
+      "grad_norm": 0.6344576478004456,
+      "learning_rate": 0.0005641202456631828,
+      "loss": 4.0171,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8287795186042786,
+      "learning_rate": 0.0005637970046331214,
+      "loss": 3.9951,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6188784845549457,
+      "grad_norm": 0.6326580047607422,
+      "learning_rate": 0.00056347376360306,
+      "loss": 4.0027,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6242600365945539,
+      "grad_norm": 0.7404974102973938,
+      "learning_rate": 0.0005631505225729985,
+      "loss": 4.0087,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6296415886341621,
+      "grad_norm": 0.6510661244392395,
+      "learning_rate": 0.0005628272815429371,
+      "loss": 3.9942,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6350231406737703,
+      "grad_norm": 0.6300317049026489,
+      "learning_rate": 0.0005625040405128757,
+      "loss": 4.0002,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6404046927133785,
+      "grad_norm": 0.5819694399833679,
+      "learning_rate": 0.0005621807994828143,
+      "loss": 3.9987,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6457862447529867,
+      "grad_norm": 0.7095755338668823,
+      "learning_rate": 0.0005618575584527529,
+      "loss": 3.9857,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6457862447529867,
+      "eval_accuracy": 0.32717677096873066,
+      "eval_loss": 3.925938606262207,
+      "eval_runtime": 182.691,
+      "eval_samples_per_second": 98.587,
+      "eval_steps_per_second": 6.163,
+      "step": 6000
+    },
+    {
+      "epoch": 0.651167796792595,
+      "grad_norm": 0.6090443134307861,
+      "learning_rate": 0.0005615343174226915,
+      "loss": 3.9941,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6565493488322032,
+      "grad_norm": 0.5509337186813354,
+      "learning_rate": 0.00056121107639263,
+      "loss": 3.988,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6619309008718114,
+      "grad_norm": 0.681161642074585,
+      "learning_rate": 0.0005608943001831699,
+      "loss": 3.989,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6673124529114196,
+      "grad_norm": 0.7362217903137207,
+      "learning_rate": 0.0005605710591531085,
+      "loss": 3.9711,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6726940049510278,
+      "grad_norm": 0.718980610370636,
+      "learning_rate": 0.000560247818123047,
+      "loss": 3.9902,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6780755569906362,
+      "grad_norm": 0.6289951205253601,
+      "learning_rate": 0.0005599245770929855,
+      "loss": 3.9774,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6834571090302444,
+      "grad_norm": 0.6267488598823547,
+      "learning_rate": 0.0005596013360629242,
+      "loss": 3.964,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6888386610698526,
+      "grad_norm": 0.681925892829895,
+      "learning_rate": 0.0005592780950328628,
+      "loss": 3.9841,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6942202131094608,
+      "grad_norm": 0.6765555143356323,
+      "learning_rate": 0.0005589548540028014,
+      "loss": 3.9582,
+      "step": 6450
+    },
+    {
+      "epoch": 0.699601765149069,
+      "grad_norm": 0.6029344201087952,
+      "learning_rate": 0.0005586316129727399,
+      "loss": 3.9729,
+      "step": 6500
+    },
+    {
+      "epoch": 0.7049833171886772,
+      "grad_norm": 0.5697076320648193,
+      "learning_rate": 0.0005583083719426786,
+      "loss": 3.9529,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7103648692282855,
+      "grad_norm": 0.6241387724876404,
+      "learning_rate": 0.0005579851309126171,
+      "loss": 3.941,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7157464212678937,
+      "grad_norm": 0.5610880851745605,
+      "learning_rate": 0.0005576618898825558,
+      "loss": 3.948,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7211279733075019,
+      "grad_norm": 0.5819376111030579,
+      "learning_rate": 0.0005573386488524943,
+      "loss": 3.9528,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7265095253471101,
+      "grad_norm": 0.5621766448020935,
+      "learning_rate": 0.0005570154078224328,
+      "loss": 3.9367,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7318910773867183,
+      "grad_norm": 0.6376521587371826,
+      "learning_rate": 0.0005566921667923715,
+      "loss": 3.9523,
+      "step": 6800
+    },
+    {
+      "epoch": 0.7372726294263265,
+      "grad_norm": 0.5738341212272644,
+      "learning_rate": 0.00055636892576231,
+      "loss": 3.9213,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7426541814659348,
+      "grad_norm": 0.5444178581237793,
+      "learning_rate": 0.0005560456847322487,
+      "loss": 3.9256,
+      "step": 6900
+    },
+    {
+      "epoch": 0.748035733505543,
+      "grad_norm": 0.6856753826141357,
+      "learning_rate": 0.0005557224437021872,
+      "loss": 3.9396,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7534172855451512,
+      "grad_norm": 0.6148762106895447,
+      "learning_rate": 0.0005553992026721258,
+      "loss": 3.932,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7534172855451512,
+      "eval_accuracy": 0.33273132389958254,
+      "eval_loss": 3.8695037364959717,
+      "eval_runtime": 183.1477,
+      "eval_samples_per_second": 98.341,
+      "eval_steps_per_second": 6.148,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7587988375847594,
+      "grad_norm": 0.6166629791259766,
+      "learning_rate": 0.0005550759616420644,
+      "loss": 3.9215,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7641803896243676,
+      "grad_norm": 0.5288543701171875,
+      "learning_rate": 0.000554752720612003,
+      "loss": 3.9322,
+      "step": 7100
+    },
+    {
+      "epoch": 0.7695619416639758,
+      "grad_norm": 0.650581955909729,
+      "learning_rate": 0.0005544294795819415,
+      "loss": 3.9267,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7749434937035841,
+      "grad_norm": 0.5915279984474182,
+      "learning_rate": 0.0005541062385518801,
+      "loss": 3.9138,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7803250457431924,
+      "grad_norm": 0.645673930644989,
+      "learning_rate": 0.0005537829975218188,
+      "loss": 3.9376,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7857065977828006,
+      "grad_norm": 0.710521936416626,
+      "learning_rate": 0.0005534597564917573,
+      "loss": 3.9161,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7910881498224088,
+      "grad_norm": 0.6567710041999817,
+      "learning_rate": 0.0005531365154616959,
+      "loss": 3.9225,
+      "step": 7350
+    },
+    {
+      "epoch": 0.796469701862017,
+      "grad_norm": 0.6403442621231079,
+      "learning_rate": 0.0005528132744316344,
+      "loss": 3.9261,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8018512539016253,
+      "grad_norm": 0.5535980463027954,
+      "learning_rate": 0.0005524900334015731,
+      "loss": 3.9177,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8072328059412335,
+      "grad_norm": 0.5908585786819458,
+      "learning_rate": 0.0005521667923715117,
+      "loss": 3.9195,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8126143579808417,
+      "grad_norm": 0.5708103775978088,
+      "learning_rate": 0.0005518435513414502,
+      "loss": 3.9055,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8179959100204499,
+      "grad_norm": 0.604663074016571,
+      "learning_rate": 0.0005515203103113888,
+      "loss": 3.9147,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8233774620600581,
+      "grad_norm": 0.5916433930397034,
+      "learning_rate": 0.0005511970692813274,
+      "loss": 3.9041,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8287590140996663,
+      "grad_norm": 0.74225914478302,
+      "learning_rate": 0.000550873828251266,
+      "loss": 3.9067,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8341405661392746,
+      "grad_norm": 0.5728737115859985,
+      "learning_rate": 0.0005505505872212045,
+      "loss": 3.9031,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8395221181788828,
+      "grad_norm": 0.6314652562141418,
+      "learning_rate": 0.0005502273461911432,
+      "loss": 3.9059,
+      "step": 7800
+    },
+    {
+      "epoch": 0.844903670218491,
+      "grad_norm": 0.6741981506347656,
+      "learning_rate": 0.0005499041051610817,
+      "loss": 3.8996,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8502852222580992,
+      "grad_norm": 0.5826407074928284,
+      "learning_rate": 0.0005495808641310204,
+      "loss": 3.8754,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8556667742977074,
+      "grad_norm": 0.5577662587165833,
+      "learning_rate": 0.0005492576231009589,
+      "loss": 3.8799,
+      "step": 7950
+    },
+    {
+      "epoch": 0.8610483263373157,
+      "grad_norm": 0.6278002858161926,
+      "learning_rate": 0.0005489343820708974,
+      "loss": 3.8959,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8610483263373157,
+      "eval_accuracy": 0.33684915969486356,
+      "eval_loss": 3.826174020767212,
+      "eval_runtime": 183.0754,
+      "eval_samples_per_second": 98.38,
+      "eval_steps_per_second": 6.15,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8664298783769239,
+      "grad_norm": 0.6210545301437378,
+      "learning_rate": 0.0005486111410408361,
+      "loss": 3.8848,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8718114304165321,
+      "grad_norm": 0.622451901435852,
+      "learning_rate": 0.0005482879000107746,
+      "loss": 3.8772,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8771929824561403,
+      "grad_norm": 0.604086697101593,
+      "learning_rate": 0.0005479646589807133,
+      "loss": 3.8865,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8825745344957485,
+      "grad_norm": 0.5613585710525513,
+      "learning_rate": 0.0005476414179506518,
+      "loss": 3.8691,
+      "step": 8200
+    },
+    {
+      "epoch": 0.8879560865353568,
+      "grad_norm": 0.6614847779273987,
+      "learning_rate": 0.0005473181769205904,
+      "loss": 3.861,
+      "step": 8250
+    },
+    {
+      "epoch": 0.8933376385749651,
+      "grad_norm": 0.551531970500946,
+      "learning_rate": 0.000546994935890529,
+      "loss": 3.8908,
+      "step": 8300
+    },
+    {
+      "epoch": 0.8987191906145733,
+      "grad_norm": 0.7161784768104553,
+      "learning_rate": 0.0005466716948604677,
+      "loss": 3.8724,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9041007426541815,
+      "grad_norm": 0.6812533140182495,
+      "learning_rate": 0.0005463484538304062,
+      "loss": 3.8675,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9094822946937897,
+      "grad_norm": 0.5462357401847839,
+      "learning_rate": 0.000546031677620946,
+      "loss": 3.8776,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9148638467333979,
+      "grad_norm": 0.5445178747177124,
+      "learning_rate": 0.0005457084365908845,
+      "loss": 3.8794,
+      "step": 8500
+    },
+    {
+      "epoch": 0.9202453987730062,
+      "grad_norm": 0.5628532767295837,
+      "learning_rate": 0.0005453851955608232,
+      "loss": 3.865,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9256269508126144,
+      "grad_norm": 0.590034544467926,
+      "learning_rate": 0.0005450619545307617,
+      "loss": 3.8665,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9310085028522226,
+      "grad_norm": 0.5842359662055969,
+      "learning_rate": 0.0005447387135007003,
+      "loss": 3.8516,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9363900548918308,
+      "grad_norm": 0.5638464093208313,
+      "learning_rate": 0.0005444154724706389,
+      "loss": 3.8689,
+      "step": 8700
+    },
+    {
+      "epoch": 0.941771606931439,
+      "grad_norm": 0.5643623471260071,
+      "learning_rate": 0.0005440922314405775,
+      "loss": 3.8532,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9471531589710472,
+      "grad_norm": 0.5182409882545471,
+      "learning_rate": 0.0005437689904105161,
+      "loss": 3.8537,
+      "step": 8800
+    },
+    {
+      "epoch": 0.9525347110106555,
+      "grad_norm": 0.6119173765182495,
+      "learning_rate": 0.0005434457493804546,
+      "loss": 3.8483,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9579162630502637,
+      "grad_norm": 0.609887957572937,
+      "learning_rate": 0.0005431225083503932,
+      "loss": 3.8477,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9632978150898719,
+      "grad_norm": 0.6084715127944946,
+      "learning_rate": 0.0005427992673203318,
+      "loss": 3.8383,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9686793671294801,
+      "grad_norm": 0.6143397092819214,
+      "learning_rate": 0.0005424760262902704,
+      "loss": 3.8478,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9686793671294801,
+      "eval_accuracy": 0.3404499163970355,
+      "eval_loss": 3.785356044769287,
+      "eval_runtime": 182.7059,
+      "eval_samples_per_second": 98.579,
+      "eval_steps_per_second": 6.163,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9740609191690883,
+      "grad_norm": 0.5996887683868408,
+      "learning_rate": 0.000542152785260209,
+      "loss": 3.8303,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9794424712086965,
+      "grad_norm": 0.5821929574012756,
+      "learning_rate": 0.0005418295442301476,
+      "loss": 3.8485,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9848240232483048,
+      "grad_norm": 0.5654467940330505,
+      "learning_rate": 0.0005415063032000861,
+      "loss": 3.8499,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9902055752879131,
+      "grad_norm": 0.6263448596000671,
+      "learning_rate": 0.0005411830621700248,
+      "loss": 3.8372,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9955871273275213,
+      "grad_norm": 0.5956529974937439,
+      "learning_rate": 0.0005408598211399633,
+      "loss": 3.8455,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0009686793671295,
+      "grad_norm": 0.5786877274513245,
+      "learning_rate": 0.0005405365801099019,
+      "loss": 3.8393,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0063502314067376,
+      "grad_norm": 0.6130419373512268,
+      "learning_rate": 0.0005402133390798405,
+      "loss": 3.7682,
+      "step": 9350
+    },
+    {
+      "epoch": 1.011731783446346,
+      "grad_norm": 0.6110179424285889,
+      "learning_rate": 0.000539890098049779,
+      "loss": 3.7743,
+      "step": 9400
+    },
+    {
+      "epoch": 1.017113335485954,
+      "grad_norm": 0.5443481206893921,
+      "learning_rate": 0.0005395668570197177,
+      "loss": 3.7773,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0224948875255624,
+      "grad_norm": 0.5167891383171082,
+      "learning_rate": 0.0005392436159896562,
+      "loss": 3.7604,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0278764395651705,
+      "grad_norm": 0.647746741771698,
+      "learning_rate": 0.0005389203749595948,
+      "loss": 3.7733,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0332579916047788,
+      "grad_norm": 0.5472586750984192,
+      "learning_rate": 0.0005385971339295334,
+      "loss": 3.7681,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0386395436443872,
+      "grad_norm": 0.5735807418823242,
+      "learning_rate": 0.000538273892899472,
+      "loss": 3.7828,
+      "step": 9650
+    },
+    {
+      "epoch": 1.0440210956839953,
+      "grad_norm": 0.6225168704986572,
+      "learning_rate": 0.0005379506518694106,
+      "loss": 3.7848,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0494026477236036,
+      "grad_norm": 0.5470787286758423,
+      "learning_rate": 0.0005376274108393491,
+      "loss": 3.7819,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0547841997632117,
+      "grad_norm": 0.5520187616348267,
+      "learning_rate": 0.0005373041698092877,
+      "loss": 3.7574,
+      "step": 9800
+    },
+    {
+      "epoch": 1.06016575180282,
+      "grad_norm": 0.5762578845024109,
+      "learning_rate": 0.0005369809287792263,
+      "loss": 3.7761,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0655473038424281,
+      "grad_norm": 0.5505574345588684,
+      "learning_rate": 0.000536657687749165,
+      "loss": 3.7765,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0709288558820365,
+      "grad_norm": 0.6215469837188721,
+      "learning_rate": 0.0005363344467191035,
+      "loss": 3.7648,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0763104079216446,
+      "grad_norm": 0.5857725739479065,
+      "learning_rate": 0.000536011205689042,
+      "loss": 3.7833,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0763104079216446,
+      "eval_accuracy": 0.3440022139112421,
+      "eval_loss": 3.7528064250946045,
+      "eval_runtime": 182.9389,
+      "eval_samples_per_second": 98.454,
+      "eval_steps_per_second": 6.155,
+      "step": 10000
+    },
+    {
+      "epoch": 1.081691959961253,
+      "grad_norm": 0.5563129782676697,
+      "learning_rate": 0.0005356879646589807,
+      "loss": 3.7952,
+      "step": 10050
+    },
+    {
+      "epoch": 1.087073512000861,
+      "grad_norm": 0.5878000259399414,
+      "learning_rate": 0.0005353647236289192,
+      "loss": 3.7564,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0924550640404693,
+      "grad_norm": 0.5479933619499207,
+      "learning_rate": 0.0005350414825988579,
+      "loss": 3.7828,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0978366160800774,
+      "grad_norm": 0.6343529224395752,
+      "learning_rate": 0.0005347182415687964,
+      "loss": 3.7892,
+      "step": 10200
+    },
+    {
+      "epoch": 1.1032181681196858,
+      "grad_norm": 0.6030763387680054,
+      "learning_rate": 0.000534395000538735,
+      "loss": 3.779,
+      "step": 10250
+    },
+    {
+      "epoch": 1.1085997201592939,
+      "grad_norm": 0.6497897505760193,
+      "learning_rate": 0.0005340717595086736,
+      "loss": 3.7651,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1139812721989022,
+      "grad_norm": 0.7304083704948425,
+      "learning_rate": 0.0005337549832992134,
+      "loss": 3.7863,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1193628242385103,
+      "grad_norm": 0.5949697494506836,
+      "learning_rate": 0.000533431742269152,
+      "loss": 3.7632,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1247443762781186,
+      "grad_norm": 0.6299553513526917,
+      "learning_rate": 0.0005331085012390905,
+      "loss": 3.7438,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1301259283177267,
+      "grad_norm": 0.6024655103683472,
+      "learning_rate": 0.0005327852602090292,
+      "loss": 3.7659,
+      "step": 10500
+    },
+    {
+      "epoch": 1.135507480357335,
+      "grad_norm": 0.6806287169456482,
+      "learning_rate": 0.0005324620191789678,
+      "loss": 3.7551,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1408890323969434,
+      "grad_norm": 0.5712878108024597,
+      "learning_rate": 0.0005321387781489063,
+      "loss": 3.7723,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1462705844365515,
+      "grad_norm": 0.5776671171188354,
+      "learning_rate": 0.0005318155371188449,
+      "loss": 3.7644,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1516521364761596,
+      "grad_norm": 0.5746056437492371,
+      "learning_rate": 0.0005314922960887834,
+      "loss": 3.7811,
+      "step": 10700
+    },
+    {
+      "epoch": 1.157033688515768,
+      "grad_norm": 0.5841644406318665,
+      "learning_rate": 0.0005311690550587221,
+      "loss": 3.767,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1624152405553763,
+      "grad_norm": 0.6027348041534424,
+      "learning_rate": 0.0005308458140286607,
+      "loss": 3.7558,
+      "step": 10800
+    },
+    {
+      "epoch": 1.1677967925949844,
+      "grad_norm": 0.6047888994216919,
+      "learning_rate": 0.0005305225729985993,
+      "loss": 3.772,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1731783446345927,
+      "grad_norm": 0.5457755327224731,
+      "learning_rate": 0.0005301993319685378,
+      "loss": 3.7738,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1785598966742008,
+      "grad_norm": 0.6240300536155701,
+      "learning_rate": 0.0005298760909384765,
+      "loss": 3.7453,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1839414487138091,
+      "grad_norm": 0.5848184823989868,
+      "learning_rate": 0.000529552849908415,
+      "loss": 3.7495,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1839414487138091,
+      "eval_accuracy": 0.3466185754498148,
+      "eval_loss": 3.728104591369629,
+      "eval_runtime": 182.5387,
+      "eval_samples_per_second": 98.67,
+      "eval_steps_per_second": 6.169,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1893230007534172,
+      "grad_norm": 0.6290469765663147,
+      "learning_rate": 0.0005292296088783535,
+      "loss": 3.7552,
+      "step": 11050
+    },
+    {
+      "epoch": 1.1947045527930256,
+      "grad_norm": 0.5949827432632446,
+      "learning_rate": 0.0005289063678482922,
+      "loss": 3.7403,
+      "step": 11100
+    },
+    {
+      "epoch": 1.2000861048326337,
+      "grad_norm": 0.5734910368919373,
+      "learning_rate": 0.0005285831268182307,
+      "loss": 3.7353,
+      "step": 11150
+    },
+    {
+      "epoch": 1.205467656872242,
+      "grad_norm": 0.5530065894126892,
+      "learning_rate": 0.0005282598857881694,
+      "loss": 3.7436,
+      "step": 11200
+    },
+    {
+      "epoch": 1.21084920891185,
+      "grad_norm": 0.6502180695533752,
+      "learning_rate": 0.0005279366447581079,
+      "loss": 3.7462,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2162307609514584,
+      "grad_norm": 0.546893298625946,
+      "learning_rate": 0.0005276134037280465,
+      "loss": 3.7476,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2216123129910665,
+      "grad_norm": 0.5351948738098145,
+      "learning_rate": 0.0005272901626979851,
+      "loss": 3.768,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2269938650306749,
+      "grad_norm": 0.6159573793411255,
+      "learning_rate": 0.0005269669216679236,
+      "loss": 3.7397,
+      "step": 11400
+    },
+    {
+      "epoch": 1.232375417070283,
+      "grad_norm": 0.5443019270896912,
+      "learning_rate": 0.0005266436806378623,
+      "loss": 3.7555,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2377569691098913,
+      "grad_norm": 0.547874927520752,
+      "learning_rate": 0.0005263204396078008,
+      "loss": 3.7532,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2431385211494996,
+      "grad_norm": 0.672290563583374,
+      "learning_rate": 0.0005259971985777394,
+      "loss": 3.7461,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2485200731891077,
+      "grad_norm": 0.6494690775871277,
+      "learning_rate": 0.000525673957547678,
+      "loss": 3.7317,
+      "step": 11600
+    },
+    {
+      "epoch": 1.2539016252287158,
+      "grad_norm": 0.5554096698760986,
+      "learning_rate": 0.0005253507165176167,
+      "loss": 3.7485,
+      "step": 11650
+    },
+    {
+      "epoch": 1.2592831772683242,
+      "grad_norm": 0.5914068222045898,
+      "learning_rate": 0.0005250274754875552,
+      "loss": 3.744,
+      "step": 11700
+    },
+    {
+      "epoch": 1.2646647293079325,
+      "grad_norm": 0.5564110279083252,
+      "learning_rate": 0.0005247042344574938,
+      "loss": 3.734,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2700462813475406,
+      "grad_norm": 0.5499497056007385,
+      "learning_rate": 0.0005243809934274323,
+      "loss": 3.7457,
+      "step": 11800
+    },
+    {
+      "epoch": 1.275427833387149,
+      "grad_norm": 0.5675379633903503,
+      "learning_rate": 0.0005240577523973709,
+      "loss": 3.7338,
+      "step": 11850
+    },
+    {
+      "epoch": 1.280809385426757,
+      "grad_norm": 0.5705437660217285,
+      "learning_rate": 0.0005237345113673095,
+      "loss": 3.759,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2861909374663654,
+      "grad_norm": 0.5672619342803955,
+      "learning_rate": 0.0005234112703372481,
+      "loss": 3.7245,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2915724895059735,
+      "grad_norm": 0.553011417388916,
+      "learning_rate": 0.0005230880293071867,
+      "loss": 3.7503,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2915724895059735,
+      "eval_accuracy": 0.34867276694683536,
+      "eval_loss": 3.701185703277588,
+      "eval_runtime": 183.0501,
+      "eval_samples_per_second": 98.394,
+      "eval_steps_per_second": 6.151,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2969540415455818,
+      "grad_norm": 0.5194526314735413,
+      "learning_rate": 0.0005227647882771253,
+      "loss": 3.7416,
+      "step": 12050
+    },
+    {
+      "epoch": 1.30233559358519,
+      "grad_norm": 0.5847457051277161,
+      "learning_rate": 0.0005224415472470639,
+      "loss": 3.7315,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3077171456247982,
+      "grad_norm": 0.6009708046913147,
+      "learning_rate": 0.0005221183062170024,
+      "loss": 3.7389,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3130986976644063,
+      "grad_norm": 0.5650305151939392,
+      "learning_rate": 0.0005217950651869409,
+      "loss": 3.7334,
+      "step": 12200
+    },
+    {
+      "epoch": 1.3184802497040147,
+      "grad_norm": 0.4992525279521942,
+      "learning_rate": 0.0005214718241568796,
+      "loss": 3.7386,
+      "step": 12250
+    },
+    {
+      "epoch": 1.3238618017436228,
+      "grad_norm": 0.5877363681793213,
+      "learning_rate": 0.0005211485831268182,
+      "loss": 3.7243,
+      "step": 12300
+    },
+    {
+      "epoch": 1.329243353783231,
+      "grad_norm": 0.6050459146499634,
+      "learning_rate": 0.0005208253420967568,
+      "loss": 3.7248,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3346249058228392,
+      "grad_norm": 0.5322745442390442,
+      "learning_rate": 0.0005205021010666953,
+      "loss": 3.7197,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3400064578624475,
+      "grad_norm": 0.5668430328369141,
+      "learning_rate": 0.0005201788600366339,
+      "loss": 3.7243,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3453880099020559,
+      "grad_norm": 0.6442492604255676,
+      "learning_rate": 0.0005198556190065725,
+      "loss": 3.7402,
+      "step": 12500
+    },
+    {
+      "epoch": 1.350769561941664,
+      "grad_norm": 0.593773365020752,
+      "learning_rate": 0.0005195323779765112,
+      "loss": 3.7295,
+      "step": 12550
+    },
+    {
+      "epoch": 1.356151113981272,
+      "grad_norm": 0.6152236461639404,
+      "learning_rate": 0.0005192091369464497,
+      "loss": 3.7121,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3615326660208804,
+      "grad_norm": 0.579685628414154,
+      "learning_rate": 0.0005188858959163882,
+      "loss": 3.7207,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3669142180604887,
+      "grad_norm": 0.5556007623672485,
+      "learning_rate": 0.0005185626548863269,
+      "loss": 3.7207,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3722957701000968,
+      "grad_norm": 0.6135968565940857,
+      "learning_rate": 0.0005182394138562654,
+      "loss": 3.7287,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3776773221397052,
+      "grad_norm": 0.5700423121452332,
+      "learning_rate": 0.0005179161728262041,
+      "loss": 3.7336,
+      "step": 12800
+    },
+    {
+      "epoch": 1.3830588741793133,
+      "grad_norm": 0.5285822749137878,
+      "learning_rate": 0.0005175929317961426,
+      "loss": 3.7093,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3884404262189216,
+      "grad_norm": 0.5742802023887634,
+      "learning_rate": 0.0005172696907660812,
+      "loss": 3.7164,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3938219782585297,
+      "grad_norm": 0.6038579940795898,
+      "learning_rate": 0.0005169464497360198,
+      "loss": 3.7429,
+      "step": 12950
+    },
+    {
+      "epoch": 1.399203530298138,
+      "grad_norm": 0.5283321738243103,
+      "learning_rate": 0.0005166232087059583,
+      "loss": 3.7205,
+      "step": 13000
+    },
+    {
+      "epoch": 1.399203530298138,
+      "eval_accuracy": 0.3511695016559243,
+      "eval_loss": 3.6827125549316406,
+      "eval_runtime": 183.2259,
+      "eval_samples_per_second": 98.299,
+      "eval_steps_per_second": 6.145,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4045850823377461,
+      "grad_norm": 0.5694897770881653,
+      "learning_rate": 0.0005162999676758969,
+      "loss": 3.7338,
+      "step": 13050
+    },
+    {
+      "epoch": 1.4099666343773545,
+      "grad_norm": 0.5415697693824768,
+      "learning_rate": 0.0005159767266458355,
+      "loss": 3.7092,
+      "step": 13100
+    },
+    {
+      "epoch": 1.4153481864169626,
+      "grad_norm": 0.5408352017402649,
+      "learning_rate": 0.0005156534856157741,
+      "loss": 3.7065,
+      "step": 13150
+    },
+    {
+      "epoch": 1.420729738456571,
+      "grad_norm": 0.5147499442100525,
+      "learning_rate": 0.0005153302445857127,
+      "loss": 3.7059,
+      "step": 13200
+    },
+    {
+      "epoch": 1.426111290496179,
+      "grad_norm": 0.553383469581604,
+      "learning_rate": 0.0005150070035556513,
+      "loss": 3.7238,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4314928425357873,
+      "grad_norm": 0.6054697632789612,
+      "learning_rate": 0.0005146837625255898,
+      "loss": 3.7054,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4368743945753955,
+      "grad_norm": 0.5648375749588013,
+      "learning_rate": 0.0005143605214955285,
+      "loss": 3.7171,
+      "step": 13350
+    },
+    {
+      "epoch": 1.4422559466150038,
+      "grad_norm": 0.6254073977470398,
+      "learning_rate": 0.0005140372804654671,
+      "loss": 3.705,
+      "step": 13400
+    },
+    {
+      "epoch": 1.447637498654612,
+      "grad_norm": 0.6315209865570068,
+      "learning_rate": 0.0005137140394354056,
+      "loss": 3.7054,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4530190506942202,
+      "grad_norm": 0.5474207997322083,
+      "learning_rate": 0.0005133907984053442,
+      "loss": 3.7168,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4584006027338283,
+      "grad_norm": 0.5616752505302429,
+      "learning_rate": 0.0005130675573752827,
+      "loss": 3.7246,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4637821547734367,
+      "grad_norm": 0.5460011959075928,
+      "learning_rate": 0.0005127443163452214,
+      "loss": 3.6915,
+      "step": 13600
+    },
+    {
+      "epoch": 1.469163706813045,
+      "grad_norm": 0.5980861186981201,
+      "learning_rate": 0.00051242107531516,
+      "loss": 3.7136,
+      "step": 13650
+    },
+    {
+      "epoch": 1.474545258852653,
+      "grad_norm": 0.6376727223396301,
+      "learning_rate": 0.0005120978342850986,
+      "loss": 3.6905,
+      "step": 13700
+    },
+    {
+      "epoch": 1.4799268108922612,
+      "grad_norm": 0.5472317337989807,
+      "learning_rate": 0.0005117745932550371,
+      "loss": 3.683,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4853083629318695,
+      "grad_norm": 0.5763005614280701,
+      "learning_rate": 0.0005114513522249758,
+      "loss": 3.6773,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4906899149714778,
+      "grad_norm": 0.5707553625106812,
+      "learning_rate": 0.0005111281111949143,
+      "loss": 3.6924,
+      "step": 13850
+    },
+    {
+      "epoch": 1.496071467011086,
+      "grad_norm": 0.5680550932884216,
+      "learning_rate": 0.0005108048701648528,
+      "loss": 3.7102,
+      "step": 13900
+    },
+    {
+      "epoch": 1.501453019050694,
+      "grad_norm": 0.6450196504592896,
+      "learning_rate": 0.0005104816291347915,
+      "loss": 3.6982,
+      "step": 13950
+    },
+    {
+      "epoch": 1.5068345710903024,
+      "grad_norm": 0.5898568630218506,
+      "learning_rate": 0.00051015838810473,
+      "loss": 3.6956,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5068345710903024,
+      "eval_accuracy": 0.3528064660637373,
+      "eval_loss": 3.6613011360168457,
+      "eval_runtime": 182.9869,
+      "eval_samples_per_second": 98.428,
+      "eval_steps_per_second": 6.153,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5122161231299107,
+      "grad_norm": 0.61032634973526,
+      "learning_rate": 0.0005098351470746687,
+      "loss": 3.7066,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5175976751695188,
+      "grad_norm": 0.5454217791557312,
+      "learning_rate": 0.0005095119060446072,
+      "loss": 3.7021,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5229792272091272,
+      "grad_norm": 0.5414226055145264,
+      "learning_rate": 0.0005091886650145458,
+      "loss": 3.7104,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5283607792487355,
+      "grad_norm": 0.5376770496368408,
+      "learning_rate": 0.0005088654239844844,
+      "loss": 3.6958,
+      "step": 14200
+    },
+    {
+      "epoch": 1.5337423312883436,
+      "grad_norm": 0.5370742082595825,
+      "learning_rate": 0.0005085421829544229,
+      "loss": 3.6932,
+      "step": 14250
+    },
+    {
+      "epoch": 1.5391238833279517,
+      "grad_norm": 0.6070656776428223,
+      "learning_rate": 0.0005082189419243616,
+      "loss": 3.6957,
+      "step": 14300
+    },
+    {
+      "epoch": 1.54450543536756,
+      "grad_norm": 0.5474653244018555,
+      "learning_rate": 0.0005079021657149014,
+      "loss": 3.6808,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5498869874071683,
+      "grad_norm": 0.5363737344741821,
+      "learning_rate": 0.0005075789246848399,
+      "loss": 3.6977,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5552685394467765,
+      "grad_norm": 0.5836102366447449,
+      "learning_rate": 0.0005072556836547785,
+      "loss": 3.693,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5606500914863846,
+      "grad_norm": 0.5627076625823975,
+      "learning_rate": 0.0005069324426247171,
+      "loss": 3.6993,
+      "step": 14500
+    },
+    {
+      "epoch": 1.566031643525993,
+      "grad_norm": 0.5487858057022095,
+      "learning_rate": 0.0005066092015946557,
+      "loss": 3.6824,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5714131955656012,
+      "grad_norm": 0.5511041879653931,
+      "learning_rate": 0.0005062859605645943,
+      "loss": 3.6905,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5767947476052093,
+      "grad_norm": 0.5165522694587708,
+      "learning_rate": 0.0005059627195345329,
+      "loss": 3.6881,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5821762996448174,
+      "grad_norm": 0.5816965103149414,
+      "learning_rate": 0.0005056394785044715,
+      "loss": 3.697,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5875578516844258,
+      "grad_norm": 0.55551677942276,
+      "learning_rate": 0.00050531623747441,
+      "loss": 3.6841,
+      "step": 14750
+    },
+    {
+      "epoch": 1.592939403724034,
+      "grad_norm": 0.5633440017700195,
+      "learning_rate": 0.0005049929964443486,
+      "loss": 3.7111,
+      "step": 14800
+    },
+    {
+      "epoch": 1.5983209557636422,
+      "grad_norm": 0.534885048866272,
+      "learning_rate": 0.0005046697554142871,
+      "loss": 3.6722,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6037025078032503,
+      "grad_norm": 0.5363733768463135,
+      "learning_rate": 0.0005043465143842258,
+      "loss": 3.6866,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6090840598428586,
+      "grad_norm": 0.7027775049209595,
+      "learning_rate": 0.0005040232733541644,
+      "loss": 3.6653,
+      "step": 14950
+    },
+    {
+      "epoch": 1.614465611882467,
+      "grad_norm": 0.5722295045852661,
+      "learning_rate": 0.000503700032324103,
+      "loss": 3.6897,
+      "step": 15000
+    },
+    {
+      "epoch": 1.614465611882467,
+      "eval_accuracy": 0.3547539604249241,
+      "eval_loss": 3.641711950302124,
+      "eval_runtime": 182.9321,
+      "eval_samples_per_second": 98.457,
+      "eval_steps_per_second": 6.155,
+      "step": 15000
+    },
+    {
+      "epoch": 1.619847163922075,
+      "grad_norm": 0.5616506934165955,
+      "learning_rate": 0.0005033767912940415,
+      "loss": 3.6878,
+      "step": 15050
+    },
+    {
+      "epoch": 1.6252287159616834,
+      "grad_norm": 0.5472654700279236,
+      "learning_rate": 0.0005030535502639802,
+      "loss": 3.7087,
+      "step": 15100
+    },
+    {
+      "epoch": 1.6306102680012917,
+      "grad_norm": 0.6255720853805542,
+      "learning_rate": 0.0005027303092339187,
+      "loss": 3.6824,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6359918200408998,
+      "grad_norm": 0.49615591764450073,
+      "learning_rate": 0.0005024070682038573,
+      "loss": 3.6515,
+      "step": 15200
+    },
+    {
+      "epoch": 1.641373372080508,
+      "grad_norm": 0.559258759021759,
+      "learning_rate": 0.0005020838271737959,
+      "loss": 3.6834,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6467549241201163,
+      "grad_norm": 0.572930097579956,
+      "learning_rate": 0.0005017605861437344,
+      "loss": 3.689,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6521364761597246,
+      "grad_norm": 0.5224685668945312,
+      "learning_rate": 0.0005014373451136731,
+      "loss": 3.6909,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6575180281993327,
+      "grad_norm": 0.591471791267395,
+      "learning_rate": 0.0005011141040836116,
+      "loss": 3.6753,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6628995802389408,
+      "grad_norm": 0.5919022560119629,
+      "learning_rate": 0.0005007908630535503,
+      "loss": 3.687,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6682811322785491,
+      "grad_norm": 0.5808023810386658,
+      "learning_rate": 0.0005004676220234888,
+      "loss": 3.6756,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6736626843181575,
+      "grad_norm": 0.5336684584617615,
+      "learning_rate": 0.0005001443809934273,
+      "loss": 3.6766,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6790442363577656,
+      "grad_norm": 0.622586190700531,
+      "learning_rate": 0.000499821139963366,
+      "loss": 3.689,
+      "step": 15600
+    },
+    {
+      "epoch": 1.6844257883973737,
+      "grad_norm": 0.5987013578414917,
+      "learning_rate": 0.0004994978989333045,
+      "loss": 3.6845,
+      "step": 15650
+    },
+    {
+      "epoch": 1.689807340436982,
+      "grad_norm": 0.6158627271652222,
+      "learning_rate": 0.0004991746579032431,
+      "loss": 3.6704,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6951888924765903,
+      "grad_norm": 0.6078811287879944,
+      "learning_rate": 0.0004988514168731817,
+      "loss": 3.6819,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7005704445161984,
+      "grad_norm": 0.5467185974121094,
+      "learning_rate": 0.0004985281758431204,
+      "loss": 3.6727,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7059519965558065,
+      "grad_norm": 0.585032045841217,
+      "learning_rate": 0.0004982049348130589,
+      "loss": 3.6765,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7113335485954149,
+      "grad_norm": 0.6013224720954895,
+      "learning_rate": 0.0004978816937829975,
+      "loss": 3.6628,
+      "step": 15900
+    },
+    {
+      "epoch": 1.7167151006350232,
+      "grad_norm": 0.5670937299728394,
+      "learning_rate": 0.000497558452752936,
+      "loss": 3.6655,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7220966526746313,
+      "grad_norm": 0.5367864966392517,
+      "learning_rate": 0.0004972352117228746,
+      "loss": 3.6712,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7220966526746313,
+      "eval_accuracy": 0.3568852954722929,
+      "eval_loss": 3.62510085105896,
+      "eval_runtime": 182.7792,
+      "eval_samples_per_second": 98.54,
+      "eval_steps_per_second": 6.16,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7274782047142396,
+      "grad_norm": 0.5834771394729614,
+      "learning_rate": 0.0004969119706928133,
+      "loss": 3.6616,
+      "step": 16050
+    },
+    {
+      "epoch": 1.732859756753848,
+      "grad_norm": 0.552348256111145,
+      "learning_rate": 0.0004965887296627518,
+      "loss": 3.6731,
+      "step": 16100
+    },
+    {
+      "epoch": 1.738241308793456,
+      "grad_norm": 0.5560649037361145,
+      "learning_rate": 0.0004962654886326904,
+      "loss": 3.6797,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7436228608330642,
+      "grad_norm": 0.5562814474105835,
+      "learning_rate": 0.000495942247602629,
+      "loss": 3.6658,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7490044128726725,
+      "grad_norm": 0.5309078097343445,
+      "learning_rate": 0.0004956190065725676,
+      "loss": 3.6545,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7543859649122808,
+      "grad_norm": 0.5403803586959839,
+      "learning_rate": 0.0004952957655425062,
+      "loss": 3.661,
+      "step": 16300
+    },
+    {
+      "epoch": 1.759767516951889,
+      "grad_norm": 0.5780831575393677,
+      "learning_rate": 0.0004949789893330459,
+      "loss": 3.6783,
+      "step": 16350
+    },
+    {
+      "epoch": 1.765149068991497,
+      "grad_norm": 0.5269186496734619,
+      "learning_rate": 0.0004946557483029846,
+      "loss": 3.6483,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7705306210311054,
+      "grad_norm": 0.598962128162384,
+      "learning_rate": 0.0004943325072729231,
+      "loss": 3.6521,
+      "step": 16450
+    },
+    {
+      "epoch": 1.7759121730707137,
+      "grad_norm": 0.5480334758758545,
+      "learning_rate": 0.0004940092662428617,
+      "loss": 3.6613,
+      "step": 16500
+    },
+    {
+      "epoch": 1.7812937251103218,
+      "grad_norm": 0.5782729983329773,
+      "learning_rate": 0.0004936860252128003,
+      "loss": 3.6506,
+      "step": 16550
+    },
+    {
+      "epoch": 1.78667527714993,
+      "grad_norm": 0.5823274850845337,
+      "learning_rate": 0.0004933627841827388,
+      "loss": 3.6526,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7920568291895382,
+      "grad_norm": 0.5701773762702942,
+      "learning_rate": 0.0004930395431526775,
+      "loss": 3.6625,
+      "step": 16650
+    },
+    {
+      "epoch": 1.7974383812291466,
+      "grad_norm": 0.5889811515808105,
+      "learning_rate": 0.0004927163021226161,
+      "loss": 3.6565,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8028199332687547,
+      "grad_norm": 0.5805728435516357,
+      "learning_rate": 0.0004923930610925547,
+      "loss": 3.6663,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8082014853083628,
+      "grad_norm": 0.5075141191482544,
+      "learning_rate": 0.0004920698200624932,
+      "loss": 3.6442,
+      "step": 16800
+    },
+    {
+      "epoch": 1.813583037347971,
+      "grad_norm": 0.5800615549087524,
+      "learning_rate": 0.0004917465790324317,
+      "loss": 3.6608,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8189645893875794,
+      "grad_norm": 0.5941094160079956,
+      "learning_rate": 0.0004914233380023704,
+      "loss": 3.66,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8243461414271875,
+      "grad_norm": 0.5344933867454529,
+      "learning_rate": 0.0004911000969723089,
+      "loss": 3.6616,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8297276934667959,
+      "grad_norm": 0.6212116479873657,
+      "learning_rate": 0.0004907768559422476,
+      "loss": 3.6428,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8297276934667959,
+      "eval_accuracy": 0.3578398110917431,
+      "eval_loss": 3.610550880432129,
+      "eval_runtime": 182.6501,
+      "eval_samples_per_second": 98.609,
+      "eval_steps_per_second": 6.165,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8351092455064042,
+      "grad_norm": 0.6466174125671387,
+      "learning_rate": 0.0004904536149121861,
+      "loss": 3.6452,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8404907975460123,
+      "grad_norm": 0.6004204750061035,
+      "learning_rate": 0.0004901303738821248,
+      "loss": 3.6527,
+      "step": 17100
+    },
+    {
+      "epoch": 1.8458723495856204,
+      "grad_norm": 0.5209426283836365,
+      "learning_rate": 0.0004898071328520633,
+      "loss": 3.6282,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8512539016252287,
+      "grad_norm": 0.5604954361915588,
+      "learning_rate": 0.0004894838918220019,
+      "loss": 3.6619,
+      "step": 17200
+    },
+    {
+      "epoch": 1.856635453664837,
+      "grad_norm": 0.5392171144485474,
+      "learning_rate": 0.0004891606507919405,
+      "loss": 3.6596,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8620170057044452,
+      "grad_norm": 0.5408145785331726,
+      "learning_rate": 0.000488837409761879,
+      "loss": 3.6552,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8673985577440533,
+      "grad_norm": 0.5370055437088013,
+      "learning_rate": 0.0004885141687318177,
+      "loss": 3.6433,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8727801097836616,
+      "grad_norm": 0.5374568700790405,
+      "learning_rate": 0.00048819092770175623,
+      "loss": 3.6328,
+      "step": 17400
+    },
+    {
+      "epoch": 1.87816166182327,
+      "grad_norm": 0.5178585648536682,
+      "learning_rate": 0.0004878676866716948,
+      "loss": 3.6565,
+      "step": 17450
+    },
+    {
+      "epoch": 1.883543213862878,
+      "grad_norm": 0.5513762831687927,
+      "learning_rate": 0.00048754444564163337,
+      "loss": 3.675,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8889247659024861,
+      "grad_norm": 0.5339059829711914,
+      "learning_rate": 0.000487221204611572,
+      "loss": 3.6502,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8943063179420945,
+      "grad_norm": 0.6063997745513916,
+      "learning_rate": 0.00048689796358151056,
+      "loss": 3.6437,
+      "step": 17600
+    },
+    {
+      "epoch": 1.8996878699817028,
+      "grad_norm": 0.5742690563201904,
+      "learning_rate": 0.00048657472255144915,
+      "loss": 3.646,
+      "step": 17650
+    },
+    {
+      "epoch": 1.905069422021311,
+      "grad_norm": 0.5425062775611877,
+      "learning_rate": 0.00048625148152138775,
+      "loss": 3.6322,
+      "step": 17700
+    },
+    {
+      "epoch": 1.910450974060919,
+      "grad_norm": 0.5563521981239319,
+      "learning_rate": 0.0004859282404913263,
+      "loss": 3.6437,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9158325261005273,
+      "grad_norm": 0.5626557469367981,
+      "learning_rate": 0.0004856049994612649,
+      "loss": 3.6403,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9212140781401357,
+      "grad_norm": 0.5019347071647644,
+      "learning_rate": 0.00048528175843120353,
+      "loss": 3.6476,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9265956301797438,
+      "grad_norm": 0.6100844144821167,
+      "learning_rate": 0.0004849585174011421,
+      "loss": 3.6504,
+      "step": 17900
+    },
+    {
+      "epoch": 1.931977182219352,
+      "grad_norm": 0.5230750441551208,
+      "learning_rate": 0.00048463527637108067,
+      "loss": 3.6293,
+      "step": 17950
+    },
+    {
+      "epoch": 1.9373587342589604,
+      "grad_norm": 0.5869380831718445,
+      "learning_rate": 0.0004843120353410192,
+      "loss": 3.6477,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9373587342589604,
+      "eval_accuracy": 0.35974362699202844,
+      "eval_loss": 3.5920450687408447,
+      "eval_runtime": 182.6734,
+      "eval_samples_per_second": 98.597,
+      "eval_steps_per_second": 6.164,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9427402862985685,
+      "grad_norm": 0.5855605006217957,
+      "learning_rate": 0.0004839887943109578,
+      "loss": 3.6606,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9481218383381766,
+      "grad_norm": 0.6179500222206116,
+      "learning_rate": 0.00048366555328089645,
+      "loss": 3.6605,
+      "step": 18100
+    },
+    {
+      "epoch": 1.953503390377785,
+      "grad_norm": 0.5851153135299683,
+      "learning_rate": 0.000483342312250835,
+      "loss": 3.6513,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9588849424173933,
+      "grad_norm": 0.610054612159729,
+      "learning_rate": 0.0004830190712207736,
+      "loss": 3.6214,
+      "step": 18200
+    },
+    {
+      "epoch": 1.9642664944570014,
+      "grad_norm": 0.6112696528434753,
+      "learning_rate": 0.0004826958301907122,
+      "loss": 3.6286,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9696480464966095,
+      "grad_norm": 0.5437807440757751,
+      "learning_rate": 0.0004823725891606507,
+      "loss": 3.6459,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9750295985362178,
+      "grad_norm": 0.5649225115776062,
+      "learning_rate": 0.0004820493481305893,
+      "loss": 3.6386,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9804111505758262,
+      "grad_norm": 0.5825486183166504,
+      "learning_rate": 0.00048172610710052797,
+      "loss": 3.6294,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9857927026154343,
+      "grad_norm": 0.5508699417114258,
+      "learning_rate": 0.0004814093308910677,
+      "loss": 3.6338,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9911742546550424,
+      "grad_norm": 0.5655311942100525,
+      "learning_rate": 0.00048108608986100637,
+      "loss": 3.6588,
+      "step": 18500
+    },
+    {
+      "epoch": 1.9965558066946507,
+      "grad_norm": 0.5678587555885315,
+      "learning_rate": 0.0004807628488309449,
+      "loss": 3.6348,
+      "step": 18550
+    },
+    {
+      "epoch": 2.001937358734259,
+      "grad_norm": 0.5703258514404297,
+      "learning_rate": 0.0004804396078008835,
+      "loss": 3.6138,
+      "step": 18600
+    },
+    {
+      "epoch": 2.007318910773867,
+      "grad_norm": 0.5701274871826172,
+      "learning_rate": 0.00048011636677082204,
+      "loss": 3.5586,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0127004628134753,
+      "grad_norm": 0.567638635635376,
+      "learning_rate": 0.00047979312574076064,
+      "loss": 3.5602,
+      "step": 18700
+    },
+    {
+      "epoch": 2.018082014853084,
+      "grad_norm": 0.5551114678382874,
+      "learning_rate": 0.0004794698847106992,
+      "loss": 3.5546,
+      "step": 18750
+    },
+    {
+      "epoch": 2.023463566892692,
+      "grad_norm": 0.6258234977722168,
+      "learning_rate": 0.0004791466436806378,
+      "loss": 3.5536,
+      "step": 18800
+    },
+    {
+      "epoch": 2.0288451189323,
+      "grad_norm": 0.7323607206344604,
+      "learning_rate": 0.0004788234026505764,
+      "loss": 3.5522,
+      "step": 18850
+    },
+    {
+      "epoch": 2.034226670971908,
+      "grad_norm": 0.5760945677757263,
+      "learning_rate": 0.00047850016162051496,
+      "loss": 3.5633,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0396082230115167,
+      "grad_norm": 0.561445951461792,
+      "learning_rate": 0.00047817692059045356,
+      "loss": 3.5486,
+      "step": 18950
+    },
+    {
+      "epoch": 2.044989775051125,
+      "grad_norm": 0.5980292558670044,
+      "learning_rate": 0.00047786014438099336,
+      "loss": 3.5527,
+      "step": 19000
+    },
+    {
+      "epoch": 2.044989775051125,
+      "eval_accuracy": 0.36137385492079693,
+      "eval_loss": 3.583522081375122,
+      "eval_runtime": 182.8509,
+      "eval_samples_per_second": 98.501,
+      "eval_steps_per_second": 6.158,
+      "step": 19000
+    },
+    {
+      "epoch": 2.050371327090733,
+      "grad_norm": 0.5768479108810425,
+      "learning_rate": 0.00047753690335093196,
+      "loss": 3.5578,
+      "step": 19050
+    },
+    {
+      "epoch": 2.055752879130341,
+      "grad_norm": 0.5865322947502136,
+      "learning_rate": 0.00047721366232087055,
+      "loss": 3.5501,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0611344311699495,
+      "grad_norm": 0.5691093802452087,
+      "learning_rate": 0.00047689042129080915,
+      "loss": 3.544,
+      "step": 19150
+    },
+    {
+      "epoch": 2.0665159832095576,
+      "grad_norm": 0.6482433080673218,
+      "learning_rate": 0.00047656718026074774,
+      "loss": 3.5627,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0718975352491658,
+      "grad_norm": 0.5731157064437866,
+      "learning_rate": 0.00047624393923068634,
+      "loss": 3.548,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0772790872887743,
+      "grad_norm": 0.5512408018112183,
+      "learning_rate": 0.0004759206982006249,
+      "loss": 3.5753,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0826606393283824,
+      "grad_norm": 0.5742067098617554,
+      "learning_rate": 0.00047559745717056347,
+      "loss": 3.5479,
+      "step": 19350
+    },
+    {
+      "epoch": 2.0880421913679905,
+      "grad_norm": 0.5789666175842285,
+      "learning_rate": 0.000475274216140502,
+      "loss": 3.5566,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0934237434075986,
+      "grad_norm": 0.631552517414093,
+      "learning_rate": 0.00047495097511044066,
+      "loss": 3.5637,
+      "step": 19450
+    },
+    {
+      "epoch": 2.098805295447207,
+      "grad_norm": 0.5673854351043701,
+      "learning_rate": 0.00047462773408037925,
+      "loss": 3.5659,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1041868474868153,
+      "grad_norm": 0.6008737683296204,
+      "learning_rate": 0.0004743044930503178,
+      "loss": 3.5755,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1095683995264234,
+      "grad_norm": 0.5127817392349243,
+      "learning_rate": 0.0004739812520202564,
+      "loss": 3.576,
+      "step": 19600
+    },
+    {
+      "epoch": 2.1149499515660315,
+      "grad_norm": 0.5636597871780396,
+      "learning_rate": 0.000473658010990195,
+      "loss": 3.5646,
+      "step": 19650
+    },
+    {
+      "epoch": 2.12033150360564,
+      "grad_norm": 0.566287636756897,
+      "learning_rate": 0.0004733347699601336,
+      "loss": 3.5496,
+      "step": 19700
+    },
+    {
+      "epoch": 2.125713055645248,
+      "grad_norm": 0.5994743704795837,
+      "learning_rate": 0.0004730115289300722,
+      "loss": 3.558,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1310946076848563,
+      "grad_norm": 0.5499207377433777,
+      "learning_rate": 0.00047268828790001077,
+      "loss": 3.5618,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1364761597244644,
+      "grad_norm": 0.5498175621032715,
+      "learning_rate": 0.0004723650468699493,
+      "loss": 3.5618,
+      "step": 19850
+    },
+    {
+      "epoch": 2.141857711764073,
+      "grad_norm": 0.5704463720321655,
+      "learning_rate": 0.0004720418058398879,
+      "loss": 3.5423,
+      "step": 19900
+    },
+    {
+      "epoch": 2.147239263803681,
+      "grad_norm": 0.5655784606933594,
+      "learning_rate": 0.00047171856480982644,
+      "loss": 3.5586,
+      "step": 19950
+    },
+    {
+      "epoch": 2.152620815843289,
+      "grad_norm": 0.5610454678535461,
+      "learning_rate": 0.0004713953237797651,
+      "loss": 3.5691,
+      "step": 20000
+    },
+    {
+      "epoch": 2.152620815843289,
+      "eval_accuracy": 0.36227817290607683,
+      "eval_loss": 3.5729849338531494,
+      "eval_runtime": 182.9497,
+      "eval_samples_per_second": 98.448,
+      "eval_steps_per_second": 6.155,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1580023678828972,
+      "grad_norm": 0.5196320414543152,
+      "learning_rate": 0.0004710720827497037,
+      "loss": 3.5614,
+      "step": 20050
+    },
+    {
+      "epoch": 2.163383919922506,
+      "grad_norm": 0.5847726464271545,
+      "learning_rate": 0.00047074884171964223,
+      "loss": 3.5468,
+      "step": 20100
+    },
+    {
+      "epoch": 2.168765471962114,
+      "grad_norm": 0.5680404901504517,
+      "learning_rate": 0.0004704256006895808,
+      "loss": 3.5578,
+      "step": 20150
+    },
+    {
+      "epoch": 2.174147024001722,
+      "grad_norm": 0.5938388705253601,
+      "learning_rate": 0.00047010235965951936,
+      "loss": 3.5858,
+      "step": 20200
+    },
+    {
+      "epoch": 2.1795285760413305,
+      "grad_norm": 0.6427321434020996,
+      "learning_rate": 0.00046977911862945796,
+      "loss": 3.5649,
+      "step": 20250
+    },
+    {
+      "epoch": 2.1849101280809387,
+      "grad_norm": 0.578403651714325,
+      "learning_rate": 0.0004694558775993966,
+      "loss": 3.5625,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1902916801205468,
+      "grad_norm": 0.5661942362785339,
+      "learning_rate": 0.00046913263656933515,
+      "loss": 3.5531,
+      "step": 20350
+    },
+    {
+      "epoch": 2.195673232160155,
+      "grad_norm": 0.603402853012085,
+      "learning_rate": 0.00046880939553927374,
+      "loss": 3.5608,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2010547841997634,
+      "grad_norm": 0.5676315426826477,
+      "learning_rate": 0.00046848615450921234,
+      "loss": 3.5555,
+      "step": 20450
+    },
+    {
+      "epoch": 2.2064363362393715,
+      "grad_norm": 0.592079222202301,
+      "learning_rate": 0.0004681629134791509,
+      "loss": 3.5641,
+      "step": 20500
+    },
+    {
+      "epoch": 2.2118178882789796,
+      "grad_norm": 0.5975874662399292,
+      "learning_rate": 0.00046783967244908947,
+      "loss": 3.5477,
+      "step": 20550
+    },
+    {
+      "epoch": 2.2171994403185877,
+      "grad_norm": 0.6059330701828003,
+      "learning_rate": 0.0004675164314190281,
+      "loss": 3.5673,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2225809923581963,
+      "grad_norm": 0.5840981602668762,
+      "learning_rate": 0.00046719319038896666,
+      "loss": 3.5702,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2279625443978044,
+      "grad_norm": 0.5514042377471924,
+      "learning_rate": 0.00046686994935890526,
+      "loss": 3.5504,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2333440964374125,
+      "grad_norm": 0.5639162659645081,
+      "learning_rate": 0.0004665467083288438,
+      "loss": 3.5866,
+      "step": 20750
+    },
+    {
+      "epoch": 2.2387256484770206,
+      "grad_norm": 0.5674893260002136,
+      "learning_rate": 0.0004662234672987824,
+      "loss": 3.5668,
+      "step": 20800
+    },
+    {
+      "epoch": 2.244107200516629,
+      "grad_norm": 0.6036685109138489,
+      "learning_rate": 0.00046590022626872104,
+      "loss": 3.5557,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2494887525562373,
+      "grad_norm": 0.5608086585998535,
+      "learning_rate": 0.0004655769852386596,
+      "loss": 3.5617,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2548703045958454,
+      "grad_norm": 0.6411144137382507,
+      "learning_rate": 0.0004652537442085982,
+      "loss": 3.5523,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2602518566354535,
+      "grad_norm": 0.5787631273269653,
+      "learning_rate": 0.00046493050317853677,
+      "loss": 3.5718,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2602518566354535,
+      "eval_accuracy": 0.3634697691267383,
+      "eval_loss": 3.564570665359497,
+      "eval_runtime": 183.029,
+      "eval_samples_per_second": 98.405,
+      "eval_steps_per_second": 6.152,
+      "step": 21000
+    },
+    {
+      "epoch": 2.265633408675062,
+      "grad_norm": 0.6253136992454529,
+      "learning_rate": 0.0004646072621484753,
+      "loss": 3.5701,
+      "step": 21050
+    },
+    {
+      "epoch": 2.27101496071467,
+      "grad_norm": 0.6177113652229309,
+      "learning_rate": 0.0004642840211184139,
+      "loss": 3.575,
+      "step": 21100
+    },
+    {
+      "epoch": 2.2763965127542782,
+      "grad_norm": 0.6227148771286011,
+      "learning_rate": 0.00046396078008835255,
+      "loss": 3.5739,
+      "step": 21150
+    },
+    {
+      "epoch": 2.281778064793887,
+      "grad_norm": 0.6199923157691956,
+      "learning_rate": 0.0004636375390582911,
+      "loss": 3.5579,
+      "step": 21200
+    },
+    {
+      "epoch": 2.287159616833495,
+      "grad_norm": 0.5631394982337952,
+      "learning_rate": 0.0004633142980282297,
+      "loss": 3.5603,
+      "step": 21250
+    },
+    {
+      "epoch": 2.292541168873103,
+      "grad_norm": 0.5839282870292664,
+      "learning_rate": 0.00046299105699816823,
+      "loss": 3.5685,
+      "step": 21300
+    },
+    {
+      "epoch": 2.297922720912711,
+      "grad_norm": 0.5705982446670532,
+      "learning_rate": 0.0004626678159681068,
+      "loss": 3.5749,
+      "step": 21350
+    },
+    {
+      "epoch": 2.303304272952319,
+      "grad_norm": 0.5807977914810181,
+      "learning_rate": 0.0004623445749380454,
+      "loss": 3.5513,
+      "step": 21400
+    },
+    {
+      "epoch": 2.3086858249919278,
+      "grad_norm": 0.5916268229484558,
+      "learning_rate": 0.000462021333907984,
+      "loss": 3.5706,
+      "step": 21450
+    },
+    {
+      "epoch": 2.314067377031536,
+      "grad_norm": 0.5938501358032227,
+      "learning_rate": 0.0004616980928779226,
+      "loss": 3.5709,
+      "step": 21500
+    },
+    {
+      "epoch": 2.319448929071144,
+      "grad_norm": 0.5628405809402466,
+      "learning_rate": 0.0004613748518478612,
+      "loss": 3.5559,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3248304811107525,
+      "grad_norm": 0.5705186128616333,
+      "learning_rate": 0.00046105161081779974,
+      "loss": 3.5768,
+      "step": 21600
+    },
+    {
+      "epoch": 2.3302120331503606,
+      "grad_norm": 0.5919448137283325,
+      "learning_rate": 0.00046072836978773834,
+      "loss": 3.5586,
+      "step": 21650
+    },
+    {
+      "epoch": 2.3355935851899687,
+      "grad_norm": 0.5722348690032959,
+      "learning_rate": 0.000460405128757677,
+      "loss": 3.5473,
+      "step": 21700
+    },
+    {
+      "epoch": 2.340975137229577,
+      "grad_norm": 0.5616813898086548,
+      "learning_rate": 0.0004600818877276155,
+      "loss": 3.5644,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3463566892691854,
+      "grad_norm": 0.5710437297821045,
+      "learning_rate": 0.0004597586466975541,
+      "loss": 3.5366,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3517382413087935,
+      "grad_norm": 0.5971595644950867,
+      "learning_rate": 0.00045943540566749266,
+      "loss": 3.5574,
+      "step": 21850
+    },
+    {
+      "epoch": 2.3571197933484016,
+      "grad_norm": 0.6181142330169678,
+      "learning_rate": 0.00045911216463743126,
+      "loss": 3.5667,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3625013453880097,
+      "grad_norm": 0.6226775646209717,
+      "learning_rate": 0.00045878892360736985,
+      "loss": 3.5504,
+      "step": 21950
+    },
+    {
+      "epoch": 2.3678828974276183,
+      "grad_norm": 0.6260030269622803,
+      "learning_rate": 0.00045846568257730845,
+      "loss": 3.5632,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3678828974276183,
+      "eval_accuracy": 0.3642475064977143,
+      "eval_loss": 3.552438497543335,
+      "eval_runtime": 183.1631,
+      "eval_samples_per_second": 98.333,
+      "eval_steps_per_second": 6.148,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3732644494672264,
+      "grad_norm": 0.5650054216384888,
+      "learning_rate": 0.00045814244154724704,
+      "loss": 3.5589,
+      "step": 22050
+    },
+    {
+      "epoch": 2.3786460015068345,
+      "grad_norm": 0.5674408078193665,
+      "learning_rate": 0.00045781920051718563,
+      "loss": 3.5569,
+      "step": 22100
+    },
+    {
+      "epoch": 2.384027553546443,
+      "grad_norm": 0.6000884175300598,
+      "learning_rate": 0.0004574959594871242,
+      "loss": 3.5644,
+      "step": 22150
+    },
+    {
+      "epoch": 2.389409105586051,
+      "grad_norm": 0.6201997399330139,
+      "learning_rate": 0.00045717271845706277,
+      "loss": 3.5686,
+      "step": 22200
+    },
+    {
+      "epoch": 2.3947906576256592,
+      "grad_norm": 0.5663855075836182,
+      "learning_rate": 0.0004568494774270013,
+      "loss": 3.5715,
+      "step": 22250
+    },
+    {
+      "epoch": 2.4001722096652673,
+      "grad_norm": 0.6110810041427612,
+      "learning_rate": 0.00045652623639693996,
+      "loss": 3.5599,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4055537617048754,
+      "grad_norm": 0.5673940181732178,
+      "learning_rate": 0.00045620299536687855,
+      "loss": 3.5529,
+      "step": 22350
+    },
+    {
+      "epoch": 2.410935313744484,
+      "grad_norm": 0.5478566288948059,
+      "learning_rate": 0.0004558797543368171,
+      "loss": 3.5546,
+      "step": 22400
+    },
+    {
+      "epoch": 2.416316865784092,
+      "grad_norm": 0.5602691769599915,
+      "learning_rate": 0.0004555565133067557,
+      "loss": 3.5553,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4216984178237,
+      "grad_norm": 0.6060113906860352,
+      "learning_rate": 0.0004552332722766943,
+      "loss": 3.5723,
+      "step": 22500
+    },
+    {
+      "epoch": 2.4270799698633088,
+      "grad_norm": 0.607200562953949,
+      "learning_rate": 0.0004549100312466328,
+      "loss": 3.5494,
+      "step": 22550
+    },
+    {
+      "epoch": 2.432461521902917,
+      "grad_norm": 0.5753543376922607,
+      "learning_rate": 0.0004545867902165715,
+      "loss": 3.5594,
+      "step": 22600
+    },
+    {
+      "epoch": 2.437843073942525,
+      "grad_norm": 1.2295790910720825,
+      "learning_rate": 0.00045426354918651007,
+      "loss": 3.5365,
+      "step": 22650
+    },
+    {
+      "epoch": 2.443224625982133,
+      "grad_norm": 0.6025099754333496,
+      "learning_rate": 0.0004539403081564486,
+      "loss": 3.5705,
+      "step": 22700
+    },
+    {
+      "epoch": 2.4486061780217416,
+      "grad_norm": 0.7182654142379761,
+      "learning_rate": 0.0004536170671263872,
+      "loss": 3.5582,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4539877300613497,
+      "grad_norm": 0.5721381306648254,
+      "learning_rate": 0.00045329382609632574,
+      "loss": 3.5563,
+      "step": 22800
+    },
+    {
+      "epoch": 2.459369282100958,
+      "grad_norm": 0.5419154763221741,
+      "learning_rate": 0.0004529705850662644,
+      "loss": 3.562,
+      "step": 22850
+    },
+    {
+      "epoch": 2.464750834140566,
+      "grad_norm": 0.5342304706573486,
+      "learning_rate": 0.000452647344036203,
+      "loss": 3.5658,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4701323861801745,
+      "grad_norm": 0.6328591108322144,
+      "learning_rate": 0.00045232410300614153,
+      "loss": 3.5421,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4755139382197826,
+      "grad_norm": 0.556786060333252,
+      "learning_rate": 0.0004520008619760801,
+      "loss": 3.5554,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4755139382197826,
+      "eval_accuracy": 0.3653008962450757,
+      "eval_loss": 3.5433034896850586,
+      "eval_runtime": 182.589,
+      "eval_samples_per_second": 98.642,
+      "eval_steps_per_second": 6.167,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4808954902593907,
+      "grad_norm": 0.5863376259803772,
+      "learning_rate": 0.00045168408576661993,
+      "loss": 3.5371,
+      "step": 23050
+    },
+    {
+      "epoch": 2.4862770422989993,
+      "grad_norm": 0.5524567365646362,
+      "learning_rate": 0.0004513608447365585,
+      "loss": 3.5489,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4916585943386074,
+      "grad_norm": 0.5922284722328186,
+      "learning_rate": 0.00045103760370649706,
+      "loss": 3.5695,
+      "step": 23150
+    },
+    {
+      "epoch": 2.4970401463782155,
+      "grad_norm": 0.5904964804649353,
+      "learning_rate": 0.00045071436267643566,
+      "loss": 3.551,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5024216984178236,
+      "grad_norm": 0.5709808468818665,
+      "learning_rate": 0.0004503911216463743,
+      "loss": 3.5497,
+      "step": 23250
+    },
+    {
+      "epoch": 2.5078032504574317,
+      "grad_norm": 0.6329324841499329,
+      "learning_rate": 0.00045006788061631285,
+      "loss": 3.546,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5131848024970402,
+      "grad_norm": 0.5848966836929321,
+      "learning_rate": 0.00044974463958625144,
+      "loss": 3.5475,
+      "step": 23350
+    },
+    {
+      "epoch": 2.5185663545366483,
+      "grad_norm": 0.6118052005767822,
+      "learning_rate": 0.00044942139855619004,
+      "loss": 3.5342,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5239479065762565,
+      "grad_norm": 0.5881749391555786,
+      "learning_rate": 0.0004490981575261286,
+      "loss": 3.5452,
+      "step": 23450
+    },
+    {
+      "epoch": 2.529329458615865,
+      "grad_norm": 0.5847213864326477,
+      "learning_rate": 0.0004487749164960672,
+      "loss": 3.5521,
+      "step": 23500
+    },
+    {
+      "epoch": 2.534711010655473,
+      "grad_norm": 0.6213960647583008,
+      "learning_rate": 0.0004484516754660058,
+      "loss": 3.5339,
+      "step": 23550
+    },
+    {
+      "epoch": 2.540092562695081,
+      "grad_norm": 0.6048945784568787,
+      "learning_rate": 0.00044812843443594436,
+      "loss": 3.5714,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5454741147346893,
+      "grad_norm": 0.6373373866081238,
+      "learning_rate": 0.00044780519340588296,
+      "loss": 3.5608,
+      "step": 23650
+    },
+    {
+      "epoch": 2.550855666774298,
+      "grad_norm": 0.5849354863166809,
+      "learning_rate": 0.0004474819523758215,
+      "loss": 3.5378,
+      "step": 23700
+    },
+    {
+      "epoch": 2.556237218813906,
+      "grad_norm": 0.5594618916511536,
+      "learning_rate": 0.0004471587113457601,
+      "loss": 3.5356,
+      "step": 23750
+    },
+    {
+      "epoch": 2.561618770853514,
+      "grad_norm": 0.6139757633209229,
+      "learning_rate": 0.00044683547031569874,
+      "loss": 3.5414,
+      "step": 23800
+    },
+    {
+      "epoch": 2.567000322893122,
+      "grad_norm": 0.5510112643241882,
+      "learning_rate": 0.0004465122292856373,
+      "loss": 3.5638,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5723818749327307,
+      "grad_norm": 0.6046518087387085,
+      "learning_rate": 0.0004461889882555759,
+      "loss": 3.5382,
+      "step": 23900
+    },
+    {
+      "epoch": 2.577763426972339,
+      "grad_norm": 0.5786885619163513,
+      "learning_rate": 0.00044586574722551447,
+      "loss": 3.571,
+      "step": 23950
+    },
+    {
+      "epoch": 2.583144979011947,
+      "grad_norm": 0.5536140203475952,
+      "learning_rate": 0.000445542506195453,
+      "loss": 3.5367,
+      "step": 24000
+    },
+    {
+      "epoch": 2.583144979011947,
+      "eval_accuracy": 0.36611677077967464,
+      "eval_loss": 3.534451484680176,
+      "eval_runtime": 183.1729,
+      "eval_samples_per_second": 98.328,
+      "eval_steps_per_second": 6.147,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5885265310515555,
+      "grad_norm": 0.5413175225257874,
+      "learning_rate": 0.0004452192651653916,
+      "loss": 3.5572,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5939080830911636,
+      "grad_norm": 0.614355206489563,
+      "learning_rate": 0.00044489602413533025,
+      "loss": 3.551,
+      "step": 24100
+    },
+    {
+      "epoch": 2.5992896351307717,
+      "grad_norm": 0.5778051018714905,
+      "learning_rate": 0.0004445727831052688,
+      "loss": 3.551,
+      "step": 24150
+    },
+    {
+      "epoch": 2.60467118717038,
+      "grad_norm": 0.5734228491783142,
+      "learning_rate": 0.0004442495420752074,
+      "loss": 3.5781,
+      "step": 24200
+    },
+    {
+      "epoch": 2.610052739209988,
+      "grad_norm": 0.6527189016342163,
+      "learning_rate": 0.00044392630104514593,
+      "loss": 3.5606,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6154342912495965,
+      "grad_norm": 0.5808097720146179,
+      "learning_rate": 0.0004436030600150845,
+      "loss": 3.539,
+      "step": 24300
+    },
+    {
+      "epoch": 2.6208158432892046,
+      "grad_norm": 0.578321099281311,
+      "learning_rate": 0.0004432798189850231,
+      "loss": 3.5517,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6261973953288127,
+      "grad_norm": 0.5915061831474304,
+      "learning_rate": 0.0004429565779549617,
+      "loss": 3.5437,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6315789473684212,
+      "grad_norm": 0.6535822749137878,
+      "learning_rate": 0.0004426333369249003,
+      "loss": 3.5527,
+      "step": 24450
+    },
+    {
+      "epoch": 2.6369604994080293,
+      "grad_norm": 0.5836544036865234,
+      "learning_rate": 0.0004423100958948389,
+      "loss": 3.5435,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6423420514476375,
+      "grad_norm": 0.6255151629447937,
+      "learning_rate": 0.00044198685486477744,
+      "loss": 3.5353,
+      "step": 24550
+    },
+    {
+      "epoch": 2.6477236034872456,
+      "grad_norm": 0.5904508829116821,
+      "learning_rate": 0.00044166361383471604,
+      "loss": 3.5546,
+      "step": 24600
+    },
+    {
+      "epoch": 2.653105155526854,
+      "grad_norm": 0.5963026285171509,
+      "learning_rate": 0.0004413403728046547,
+      "loss": 3.5417,
+      "step": 24650
+    },
+    {
+      "epoch": 2.658486707566462,
+      "grad_norm": 0.5733145475387573,
+      "learning_rate": 0.0004410171317745932,
+      "loss": 3.5371,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6638682596060703,
+      "grad_norm": 0.574026346206665,
+      "learning_rate": 0.0004406938907445318,
+      "loss": 3.5503,
+      "step": 24750
+    },
+    {
+      "epoch": 2.6692498116456784,
+      "grad_norm": 0.5410118103027344,
+      "learning_rate": 0.00044037064971447036,
+      "loss": 3.5505,
+      "step": 24800
+    },
+    {
+      "epoch": 2.674631363685287,
+      "grad_norm": 0.5905002951622009,
+      "learning_rate": 0.00044004740868440896,
+      "loss": 3.5496,
+      "step": 24850
+    },
+    {
+      "epoch": 2.680012915724895,
+      "grad_norm": 0.5615861415863037,
+      "learning_rate": 0.00043972416765434755,
+      "loss": 3.5361,
+      "step": 24900
+    },
+    {
+      "epoch": 2.685394467764503,
+      "grad_norm": 0.5930156111717224,
+      "learning_rate": 0.00043940092662428615,
+      "loss": 3.5658,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6907760198041117,
+      "grad_norm": 0.636413037776947,
+      "learning_rate": 0.00043907768559422474,
+      "loss": 3.543,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6907760198041117,
+      "eval_accuracy": 0.36713582621448665,
+      "eval_loss": 3.526466131210327,
+      "eval_runtime": 182.6832,
+      "eval_samples_per_second": 98.591,
+      "eval_steps_per_second": 6.164,
+      "step": 25000
+    },
+    {
+      "epoch": 2.69615757184372,
+      "grad_norm": 0.5980111360549927,
+      "learning_rate": 0.00043875444456416334,
+      "loss": 3.5362,
+      "step": 25050
+    },
+    {
+      "epoch": 2.701539123883328,
+      "grad_norm": 0.5899256467819214,
+      "learning_rate": 0.0004384312035341019,
+      "loss": 3.5305,
+      "step": 25100
+    },
+    {
+      "epoch": 2.706920675922936,
+      "grad_norm": 0.6567667722702026,
+      "learning_rate": 0.00043810796250404047,
+      "loss": 3.5296,
+      "step": 25150
+    },
+    {
+      "epoch": 2.712302227962544,
+      "grad_norm": 0.6192686557769775,
+      "learning_rate": 0.0004377911862945803,
+      "loss": 3.5487,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7176837800021527,
+      "grad_norm": 0.5786051154136658,
+      "learning_rate": 0.00043746794526451887,
+      "loss": 3.5591,
+      "step": 25250
+    },
+    {
+      "epoch": 2.723065332041761,
+      "grad_norm": 0.6026502847671509,
+      "learning_rate": 0.00043714470423445747,
+      "loss": 3.5304,
+      "step": 25300
+    },
+    {
+      "epoch": 2.728446884081369,
+      "grad_norm": 0.6267446279525757,
+      "learning_rate": 0.00043682146320439606,
+      "loss": 3.5329,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7338284361209775,
+      "grad_norm": 0.617989182472229,
+      "learning_rate": 0.00043649822217433466,
+      "loss": 3.5411,
+      "step": 25400
+    },
+    {
+      "epoch": 2.7392099881605856,
+      "grad_norm": 0.5602372288703918,
+      "learning_rate": 0.0004361749811442732,
+      "loss": 3.5651,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7445915402001937,
+      "grad_norm": 0.5438847541809082,
+      "learning_rate": 0.0004358517401142118,
+      "loss": 3.5549,
+      "step": 25500
+    },
+    {
+      "epoch": 2.749973092239802,
+      "grad_norm": 0.5827850103378296,
+      "learning_rate": 0.00043552849908415033,
+      "loss": 3.5294,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7553546442794103,
+      "grad_norm": 0.5704156160354614,
+      "learning_rate": 0.000435205258054089,
+      "loss": 3.5352,
+      "step": 25600
+    },
+    {
+      "epoch": 2.7607361963190185,
+      "grad_norm": 0.5868292450904846,
+      "learning_rate": 0.0004348820170240276,
+      "loss": 3.5492,
+      "step": 25650
+    },
+    {
+      "epoch": 2.7661177483586266,
+      "grad_norm": 0.5794882774353027,
+      "learning_rate": 0.0004345587759939661,
+      "loss": 3.5491,
+      "step": 25700
+    },
+    {
+      "epoch": 2.7714993003982347,
+      "grad_norm": 0.6541237235069275,
+      "learning_rate": 0.0004342355349639047,
+      "loss": 3.5236,
+      "step": 25750
+    },
+    {
+      "epoch": 2.776880852437843,
+      "grad_norm": 0.6571041345596313,
+      "learning_rate": 0.0004339122939338433,
+      "loss": 3.5395,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7822624044774513,
+      "grad_norm": 0.5382871627807617,
+      "learning_rate": 0.00043358905290378184,
+      "loss": 3.5449,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7876439565170594,
+      "grad_norm": 0.5958513021469116,
+      "learning_rate": 0.0004332658118737205,
+      "loss": 3.5299,
+      "step": 25900
+    },
+    {
+      "epoch": 2.793025508556668,
+      "grad_norm": 0.5741668939590454,
+      "learning_rate": 0.0004329425708436591,
+      "loss": 3.5432,
+      "step": 25950
+    },
+    {
+      "epoch": 2.798407060596276,
+      "grad_norm": 0.6117798686027527,
+      "learning_rate": 0.00043261932981359763,
+      "loss": 3.542,
+      "step": 26000
+    },
+    {
+      "epoch": 2.798407060596276,
+      "eval_accuracy": 0.3677167932056307,
+      "eval_loss": 3.520785093307495,
+      "eval_runtime": 182.753,
+      "eval_samples_per_second": 98.554,
+      "eval_steps_per_second": 6.161,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803788612635884,
+      "grad_norm": 0.5519882440567017,
+      "learning_rate": 0.0004322960887835362,
+      "loss": 3.5522,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8091701646754923,
+      "grad_norm": 0.5597295761108398,
+      "learning_rate": 0.00043197284775347476,
+      "loss": 3.5286,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8145517167151004,
+      "grad_norm": 0.6364133358001709,
+      "learning_rate": 0.00043164960672341336,
+      "loss": 3.5472,
+      "step": 26150
+    },
+    {
+      "epoch": 2.819933268754709,
+      "grad_norm": 0.6041819453239441,
+      "learning_rate": 0.000431326365693352,
+      "loss": 3.5536,
+      "step": 26200
+    },
+    {
+      "epoch": 2.825314820794317,
+      "grad_norm": 0.5981535911560059,
+      "learning_rate": 0.00043100312466329055,
+      "loss": 3.5325,
+      "step": 26250
+    },
+    {
+      "epoch": 2.830696372833925,
+      "grad_norm": 0.6039505004882812,
+      "learning_rate": 0.00043067988363322914,
+      "loss": 3.5477,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8360779248735337,
+      "grad_norm": 0.5583112835884094,
+      "learning_rate": 0.00043035664260316774,
+      "loss": 3.5385,
+      "step": 26350
+    },
+    {
+      "epoch": 2.841459476913142,
+      "grad_norm": 0.6340571045875549,
+      "learning_rate": 0.0004300334015731063,
+      "loss": 3.5228,
+      "step": 26400
+    },
+    {
+      "epoch": 2.84684102895275,
+      "grad_norm": 0.5811586380004883,
+      "learning_rate": 0.0004297101605430449,
+      "loss": 3.5302,
+      "step": 26450
+    },
+    {
+      "epoch": 2.852222580992358,
+      "grad_norm": 0.5881547331809998,
+      "learning_rate": 0.0004293869195129835,
+      "loss": 3.5351,
+      "step": 26500
+    },
+    {
+      "epoch": 2.857604133031966,
+      "grad_norm": 0.6151056289672852,
+      "learning_rate": 0.00042906367848292206,
+      "loss": 3.5292,
+      "step": 26550
+    },
+    {
+      "epoch": 2.8629856850715747,
+      "grad_norm": 0.606568455696106,
+      "learning_rate": 0.00042874043745286066,
+      "loss": 3.5441,
+      "step": 26600
+    },
+    {
+      "epoch": 2.868367237111183,
+      "grad_norm": 0.535579264163971,
+      "learning_rate": 0.0004284171964227992,
+      "loss": 3.5116,
+      "step": 26650
+    },
+    {
+      "epoch": 2.873748789150791,
+      "grad_norm": 0.5709394216537476,
+      "learning_rate": 0.0004280939553927378,
+      "loss": 3.5527,
+      "step": 26700
+    },
+    {
+      "epoch": 2.8791303411903995,
+      "grad_norm": 0.6123267412185669,
+      "learning_rate": 0.00042777071436267644,
+      "loss": 3.5198,
+      "step": 26750
+    },
+    {
+      "epoch": 2.8845118932300076,
+      "grad_norm": 0.6574706435203552,
+      "learning_rate": 0.000427447473332615,
+      "loss": 3.5197,
+      "step": 26800
+    },
+    {
+      "epoch": 2.8898934452696157,
+      "grad_norm": 0.5446913838386536,
+      "learning_rate": 0.0004271242323025536,
+      "loss": 3.5375,
+      "step": 26850
+    },
+    {
+      "epoch": 2.895274997309224,
+      "grad_norm": 0.5856238007545471,
+      "learning_rate": 0.00042680099127249217,
+      "loss": 3.5379,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9006565493488323,
+      "grad_norm": 0.5693891048431396,
+      "learning_rate": 0.0004264777502424307,
+      "loss": 3.5373,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9060381013884404,
+      "grad_norm": 0.5685888528823853,
+      "learning_rate": 0.0004261545092123693,
+      "loss": 3.5247,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9060381013884404,
+      "eval_accuracy": 0.3690562660066076,
+      "eval_loss": 3.5081608295440674,
+      "eval_runtime": 182.7742,
+      "eval_samples_per_second": 98.542,
+      "eval_steps_per_second": 6.161,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9114196534280485,
+      "grad_norm": 0.5613411664962769,
+      "learning_rate": 0.00042583126818230795,
+      "loss": 3.5421,
+      "step": 27050
+    },
+    {
+      "epoch": 2.9168012054676566,
+      "grad_norm": 0.5848773717880249,
+      "learning_rate": 0.0004255080271522465,
+      "loss": 3.5342,
+      "step": 27100
+    },
+    {
+      "epoch": 2.922182757507265,
+      "grad_norm": 0.6449905037879944,
+      "learning_rate": 0.0004251847861221851,
+      "loss": 3.5196,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9275643095468733,
+      "grad_norm": 0.6094990372657776,
+      "learning_rate": 0.00042486154509212363,
+      "loss": 3.5496,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9329458615864814,
+      "grad_norm": 0.5689534544944763,
+      "learning_rate": 0.0004245447688826635,
+      "loss": 3.5291,
+      "step": 27250
+    },
+    {
+      "epoch": 2.93832741362609,
+      "grad_norm": 0.6069585084915161,
+      "learning_rate": 0.00042422152785260203,
+      "loss": 3.5095,
+      "step": 27300
+    },
+    {
+      "epoch": 2.943708965665698,
+      "grad_norm": 0.6799548268318176,
+      "learning_rate": 0.0004238982868225406,
+      "loss": 3.5521,
+      "step": 27350
+    },
+    {
+      "epoch": 2.949090517705306,
+      "grad_norm": 0.5622816681861877,
+      "learning_rate": 0.0004235750457924793,
+      "loss": 3.5134,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9544720697449143,
+      "grad_norm": 0.5812619924545288,
+      "learning_rate": 0.0004232518047624178,
+      "loss": 3.534,
+      "step": 27450
+    },
+    {
+      "epoch": 2.9598536217845224,
+      "grad_norm": 0.5781611800193787,
+      "learning_rate": 0.0004229285637323564,
+      "loss": 3.5307,
+      "step": 27500
+    },
+    {
+      "epoch": 2.965235173824131,
+      "grad_norm": 0.6182583570480347,
+      "learning_rate": 0.00042260532270229495,
+      "loss": 3.5355,
+      "step": 27550
+    },
+    {
+      "epoch": 2.970616725863739,
+      "grad_norm": 0.6099269390106201,
+      "learning_rate": 0.00042228208167223354,
+      "loss": 3.5141,
+      "step": 27600
+    },
+    {
+      "epoch": 2.975998277903347,
+      "grad_norm": 0.6089410185813904,
+      "learning_rate": 0.00042195884064217214,
+      "loss": 3.5328,
+      "step": 27650
+    },
+    {
+      "epoch": 2.9813798299429557,
+      "grad_norm": 0.577688992023468,
+      "learning_rate": 0.00042163559961211073,
+      "loss": 3.5638,
+      "step": 27700
+    },
+    {
+      "epoch": 2.986761381982564,
+      "grad_norm": 0.6033281683921814,
+      "learning_rate": 0.00042131235858204933,
+      "loss": 3.524,
+      "step": 27750
+    },
+    {
+      "epoch": 2.992142934022172,
+      "grad_norm": 0.5998152494430542,
+      "learning_rate": 0.0004209891175519879,
+      "loss": 3.5123,
+      "step": 27800
+    },
+    {
+      "epoch": 2.9975244860617805,
+      "grad_norm": 0.676117479801178,
+      "learning_rate": 0.00042066587652192646,
+      "loss": 3.5311,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0029060381013886,
+      "grad_norm": 0.5953896641731262,
+      "learning_rate": 0.00042034263549186506,
+      "loss": 3.494,
+      "step": 27900
+    },
+    {
+      "epoch": 3.0082875901409967,
+      "grad_norm": 0.5766949653625488,
+      "learning_rate": 0.0004200193944618036,
+      "loss": 3.419,
+      "step": 27950
+    },
+    {
+      "epoch": 3.0136691421806048,
+      "grad_norm": 0.5596526265144348,
+      "learning_rate": 0.00041969615343174225,
+      "loss": 3.4332,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0136691421806048,
+      "eval_accuracy": 0.3699133199856882,
+      "eval_loss": 3.5029404163360596,
+      "eval_runtime": 182.639,
+      "eval_samples_per_second": 98.615,
+      "eval_steps_per_second": 6.165,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0190506942202133,
+      "grad_norm": 0.581177830696106,
+      "learning_rate": 0.00041937291240168084,
+      "loss": 3.4454,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0244322462598214,
+      "grad_norm": 0.5963121652603149,
+      "learning_rate": 0.0004190496713716194,
+      "loss": 3.4537,
+      "step": 28100
+    },
+    {
+      "epoch": 3.0298137982994295,
+      "grad_norm": 0.5734453201293945,
+      "learning_rate": 0.000418726430341558,
+      "loss": 3.4422,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0351953503390376,
+      "grad_norm": 0.6005439758300781,
+      "learning_rate": 0.00041840318931149657,
+      "loss": 3.4446,
+      "step": 28200
+    },
+    {
+      "epoch": 3.040576902378646,
+      "grad_norm": 0.6174030303955078,
+      "learning_rate": 0.00041807994828143517,
+      "loss": 3.4522,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0459584544182543,
+      "grad_norm": 0.6408053636550903,
+      "learning_rate": 0.00041775670725137376,
+      "loss": 3.4324,
+      "step": 28300
+    },
+    {
+      "epoch": 3.0513400064578624,
+      "grad_norm": 0.6011624336242676,
+      "learning_rate": 0.00041743346622131236,
+      "loss": 3.4556,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0567215584974705,
+      "grad_norm": 0.6361274719238281,
+      "learning_rate": 0.0004171102251912509,
+      "loss": 3.4448,
+      "step": 28400
+    },
+    {
+      "epoch": 3.062103110537079,
+      "grad_norm": 0.602817952632904,
+      "learning_rate": 0.0004167869841611895,
+      "loss": 3.4486,
+      "step": 28450
+    },
+    {
+      "epoch": 3.067484662576687,
+      "grad_norm": 0.5594189167022705,
+      "learning_rate": 0.00041646374313112803,
+      "loss": 3.4733,
+      "step": 28500
+    },
+    {
+      "epoch": 3.0728662146162953,
+      "grad_norm": 0.7374558448791504,
+      "learning_rate": 0.0004161405021010667,
+      "loss": 3.4482,
+      "step": 28550
+    },
+    {
+      "epoch": 3.0782477666559034,
+      "grad_norm": 0.6085382699966431,
+      "learning_rate": 0.0004158172610710053,
+      "loss": 3.4641,
+      "step": 28600
+    },
+    {
+      "epoch": 3.083629318695512,
+      "grad_norm": 0.5731576681137085,
+      "learning_rate": 0.0004154940200409438,
+      "loss": 3.4374,
+      "step": 28650
+    },
+    {
+      "epoch": 3.08901087073512,
+      "grad_norm": 0.5917696356773376,
+      "learning_rate": 0.0004151707790108824,
+      "loss": 3.4511,
+      "step": 28700
+    },
+    {
+      "epoch": 3.094392422774728,
+      "grad_norm": 0.5805858373641968,
+      "learning_rate": 0.00041484753798082095,
+      "loss": 3.4436,
+      "step": 28750
+    },
+    {
+      "epoch": 3.0997739748143363,
+      "grad_norm": 0.575162947177887,
+      "learning_rate": 0.00041452429695075955,
+      "loss": 3.4435,
+      "step": 28800
+    },
+    {
+      "epoch": 3.105155526853945,
+      "grad_norm": 0.5958883166313171,
+      "learning_rate": 0.0004142010559206982,
+      "loss": 3.4675,
+      "step": 28850
+    },
+    {
+      "epoch": 3.110537078893553,
+      "grad_norm": 0.6094611287117004,
+      "learning_rate": 0.00041387781489063673,
+      "loss": 3.4567,
+      "step": 28900
+    },
+    {
+      "epoch": 3.115918630933161,
+      "grad_norm": 0.5872287750244141,
+      "learning_rate": 0.00041355457386057533,
+      "loss": 3.439,
+      "step": 28950
+    },
+    {
+      "epoch": 3.121300182972769,
+      "grad_norm": 0.6196814775466919,
+      "learning_rate": 0.0004132313328305139,
+      "loss": 3.444,
+      "step": 29000
+    },
+    {
+      "epoch": 3.121300182972769,
+      "eval_accuracy": 0.3706220627729021,
+      "eval_loss": 3.4992904663085938,
+      "eval_runtime": 182.5954,
+      "eval_samples_per_second": 98.639,
+      "eval_steps_per_second": 6.167,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266817350123777,
+      "grad_norm": 0.619157612323761,
+      "learning_rate": 0.00041290809180045246,
+      "loss": 3.4508,
+      "step": 29050
+    },
+    {
+      "epoch": 3.132063287051986,
+      "grad_norm": 0.5508050918579102,
+      "learning_rate": 0.0004125848507703911,
+      "loss": 3.4648,
+      "step": 29100
+    },
+    {
+      "epoch": 3.137444839091594,
+      "grad_norm": 0.5980096459388733,
+      "learning_rate": 0.0004122616097403297,
+      "loss": 3.4646,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1428263911312024,
+      "grad_norm": 0.6267610192298889,
+      "learning_rate": 0.00041193836871026825,
+      "loss": 3.453,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1482079431708105,
+      "grad_norm": 0.6516902446746826,
+      "learning_rate": 0.00041161512768020684,
+      "loss": 3.441,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1535894952104186,
+      "grad_norm": 0.5599831938743591,
+      "learning_rate": 0.0004112918866501454,
+      "loss": 3.4536,
+      "step": 29300
+    },
+    {
+      "epoch": 3.1589710472500268,
+      "grad_norm": 0.5650069713592529,
+      "learning_rate": 0.00041097511044068524,
+      "loss": 3.462,
+      "step": 29350
+    },
+    {
+      "epoch": 3.1643525992896353,
+      "grad_norm": 0.597890317440033,
+      "learning_rate": 0.0004106518694106238,
+      "loss": 3.4729,
+      "step": 29400
+    },
+    {
+      "epoch": 3.1697341513292434,
+      "grad_norm": 0.6053642630577087,
+      "learning_rate": 0.0004103286283805624,
+      "loss": 3.447,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1751157033688515,
+      "grad_norm": 0.6315788626670837,
+      "learning_rate": 0.00041000538735050103,
+      "loss": 3.4505,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1804972554084596,
+      "grad_norm": 0.5334057211875916,
+      "learning_rate": 0.00040968214632043957,
+      "loss": 3.4588,
+      "step": 29550
+    },
+    {
+      "epoch": 3.185878807448068,
+      "grad_norm": 0.5461787581443787,
+      "learning_rate": 0.00040935890529037816,
+      "loss": 3.4605,
+      "step": 29600
+    },
+    {
+      "epoch": 3.1912603594876763,
+      "grad_norm": 0.6308553814888,
+      "learning_rate": 0.00040903566426031676,
+      "loss": 3.4494,
+      "step": 29650
+    },
+    {
+      "epoch": 3.1966419115272844,
+      "grad_norm": 0.5765513777732849,
+      "learning_rate": 0.0004087124232302553,
+      "loss": 3.4575,
+      "step": 29700
+    },
+    {
+      "epoch": 3.2020234635668925,
+      "grad_norm": 0.5815839171409607,
+      "learning_rate": 0.0004083891822001939,
+      "loss": 3.4578,
+      "step": 29750
+    },
+    {
+      "epoch": 3.207405015606501,
+      "grad_norm": 0.5885655879974365,
+      "learning_rate": 0.00040806594117013254,
+      "loss": 3.4475,
+      "step": 29800
+    },
+    {
+      "epoch": 3.212786567646109,
+      "grad_norm": 0.6800254583358765,
+      "learning_rate": 0.0004077427001400711,
+      "loss": 3.4626,
+      "step": 29850
+    },
+    {
+      "epoch": 3.2181681196857173,
+      "grad_norm": 0.5856930613517761,
+      "learning_rate": 0.0004074194591100097,
+      "loss": 3.454,
+      "step": 29900
+    },
+    {
+      "epoch": 3.2235496717253254,
+      "grad_norm": 0.6474602222442627,
+      "learning_rate": 0.0004070962180799482,
+      "loss": 3.4509,
+      "step": 29950
+    },
+    {
+      "epoch": 3.228931223764934,
+      "grad_norm": 0.5948309302330017,
+      "learning_rate": 0.0004067729770498868,
+      "loss": 3.4617,
+      "step": 30000
+    },
+    {
+      "epoch": 3.228931223764934,
+      "eval_accuracy": 0.3714459776211993,
+      "eval_loss": 3.4921934604644775,
+      "eval_runtime": 182.5164,
+      "eval_samples_per_second": 98.682,
+      "eval_steps_per_second": 6.169,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234312775804542,
+      "grad_norm": 0.5894820094108582,
+      "learning_rate": 0.00040644973601982546,
+      "loss": 3.4608,
+      "step": 30050
+    },
+    {
+      "epoch": 3.23969432784415,
+      "grad_norm": 0.5914416313171387,
+      "learning_rate": 0.000406126494989764,
+      "loss": 3.4591,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2450758798837587,
+      "grad_norm": 0.5802457928657532,
+      "learning_rate": 0.0004058032539597026,
+      "loss": 3.4654,
+      "step": 30150
+    },
+    {
+      "epoch": 3.250457431923367,
+      "grad_norm": 0.584543764591217,
+      "learning_rate": 0.00040548001292964114,
+      "loss": 3.4611,
+      "step": 30200
+    },
+    {
+      "epoch": 3.255838983962975,
+      "grad_norm": 0.5898339748382568,
+      "learning_rate": 0.00040515677189957973,
+      "loss": 3.4736,
+      "step": 30250
+    },
+    {
+      "epoch": 3.261220536002583,
+      "grad_norm": 0.5681132078170776,
+      "learning_rate": 0.0004048335308695183,
+      "loss": 3.4583,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2666020880421915,
+      "grad_norm": 0.6399067640304565,
+      "learning_rate": 0.0004045102898394569,
+      "loss": 3.4683,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2719836400817996,
+      "grad_norm": 0.6160270571708679,
+      "learning_rate": 0.0004041870488093955,
+      "loss": 3.4713,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2773651921214078,
+      "grad_norm": 0.6204001307487488,
+      "learning_rate": 0.0004038638077793341,
+      "loss": 3.4577,
+      "step": 30450
+    },
+    {
+      "epoch": 3.282746744161016,
+      "grad_norm": 0.5580645203590393,
+      "learning_rate": 0.00040354056674927265,
+      "loss": 3.4456,
+      "step": 30500
+    },
+    {
+      "epoch": 3.2881282962006244,
+      "grad_norm": 0.5606558322906494,
+      "learning_rate": 0.00040321732571921124,
+      "loss": 3.449,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2935098482402325,
+      "grad_norm": 0.6012607216835022,
+      "learning_rate": 0.0004028940846891498,
+      "loss": 3.4587,
+      "step": 30600
+    },
+    {
+      "epoch": 3.2988914002798406,
+      "grad_norm": 0.5980702638626099,
+      "learning_rate": 0.00040257084365908843,
+      "loss": 3.4565,
+      "step": 30650
+    },
+    {
+      "epoch": 3.304272952319449,
+      "grad_norm": 0.5933027267456055,
+      "learning_rate": 0.00040224760262902703,
+      "loss": 3.4612,
+      "step": 30700
+    },
+    {
+      "epoch": 3.3096545043590573,
+      "grad_norm": 0.6065646409988403,
+      "learning_rate": 0.00040192436159896557,
+      "loss": 3.4632,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3150360563986654,
+      "grad_norm": 0.6052440404891968,
+      "learning_rate": 0.00040160112056890416,
+      "loss": 3.4554,
+      "step": 30800
+    },
+    {
+      "epoch": 3.3204176084382735,
+      "grad_norm": 0.5899329781532288,
+      "learning_rate": 0.00040127787953884276,
+      "loss": 3.4619,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3257991604778816,
+      "grad_norm": 0.6436607837677002,
+      "learning_rate": 0.00040095463850878135,
+      "loss": 3.4441,
+      "step": 30900
+    },
+    {
+      "epoch": 3.33118071251749,
+      "grad_norm": 0.5700132250785828,
+      "learning_rate": 0.00040063139747871995,
+      "loss": 3.4669,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3365622645570983,
+      "grad_norm": 0.5744343996047974,
+      "learning_rate": 0.00040030815644865854,
+      "loss": 3.4675,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3365622645570983,
+      "eval_accuracy": 0.37178160639165825,
+      "eval_loss": 3.485490083694458,
+      "eval_runtime": 182.4931,
+      "eval_samples_per_second": 98.694,
+      "eval_steps_per_second": 6.17,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3419438165967064,
+      "grad_norm": 0.5764909982681274,
+      "learning_rate": 0.0003999849154185971,
+      "loss": 3.4516,
+      "step": 31050
+    },
+    {
+      "epoch": 3.347325368636315,
+      "grad_norm": 0.6199538111686707,
+      "learning_rate": 0.0003996616743885357,
+      "loss": 3.4754,
+      "step": 31100
+    },
+    {
+      "epoch": 3.352706920675923,
+      "grad_norm": 0.6050508618354797,
+      "learning_rate": 0.0003993384333584742,
+      "loss": 3.4551,
+      "step": 31150
+    },
+    {
+      "epoch": 3.358088472715531,
+      "grad_norm": 0.591788113117218,
+      "learning_rate": 0.00039901519232841287,
+      "loss": 3.4839,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3634700247551392,
+      "grad_norm": 0.6040419340133667,
+      "learning_rate": 0.00039869195129835146,
+      "loss": 3.4702,
+      "step": 31250
+    },
+    {
+      "epoch": 3.368851576794748,
+      "grad_norm": 0.680150032043457,
+      "learning_rate": 0.00039836871026829,
+      "loss": 3.4417,
+      "step": 31300
+    },
+    {
+      "epoch": 3.374233128834356,
+      "grad_norm": 0.5792616605758667,
+      "learning_rate": 0.0003980454692382286,
+      "loss": 3.4799,
+      "step": 31350
+    },
+    {
+      "epoch": 3.379614680873964,
+      "grad_norm": 0.5896565914154053,
+      "learning_rate": 0.0003977286930287684,
+      "loss": 3.4703,
+      "step": 31400
+    },
+    {
+      "epoch": 3.384996232913572,
+      "grad_norm": 0.5844624042510986,
+      "learning_rate": 0.000397405451998707,
+      "loss": 3.4721,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3903777849531807,
+      "grad_norm": 0.5530960559844971,
+      "learning_rate": 0.00039708221096864554,
+      "loss": 3.4579,
+      "step": 31500
+    },
+    {
+      "epoch": 3.3957593369927888,
+      "grad_norm": 0.5991605520248413,
+      "learning_rate": 0.00039675896993858413,
+      "loss": 3.477,
+      "step": 31550
+    },
+    {
+      "epoch": 3.401140889032397,
+      "grad_norm": 0.5415410399436951,
+      "learning_rate": 0.0003964357289085228,
+      "loss": 3.4624,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4065224410720054,
+      "grad_norm": 0.6166115403175354,
+      "learning_rate": 0.0003961124878784613,
+      "loss": 3.4492,
+      "step": 31650
+    },
+    {
+      "epoch": 3.4119039931116135,
+      "grad_norm": 0.5882399082183838,
+      "learning_rate": 0.0003957892468483999,
+      "loss": 3.4693,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4172855451512216,
+      "grad_norm": 0.6354446411132812,
+      "learning_rate": 0.0003954660058183385,
+      "loss": 3.4692,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4226670971908297,
+      "grad_norm": 0.5936651229858398,
+      "learning_rate": 0.00039514276478827705,
+      "loss": 3.4842,
+      "step": 31800
+    },
+    {
+      "epoch": 3.428048649230438,
+      "grad_norm": 0.7023192644119263,
+      "learning_rate": 0.0003948195237582157,
+      "loss": 3.4645,
+      "step": 31850
+    },
+    {
+      "epoch": 3.4334302012700464,
+      "grad_norm": 0.6037172675132751,
+      "learning_rate": 0.0003944962827281543,
+      "loss": 3.4535,
+      "step": 31900
+    },
+    {
+      "epoch": 3.4388117533096545,
+      "grad_norm": 0.6036386489868164,
+      "learning_rate": 0.00039417304169809284,
+      "loss": 3.48,
+      "step": 31950
+    },
+    {
+      "epoch": 3.4441933053492626,
+      "grad_norm": 0.6221677660942078,
+      "learning_rate": 0.00039384980066803143,
+      "loss": 3.4746,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4441933053492626,
+      "eval_accuracy": 0.37269016183956294,
+      "eval_loss": 3.4805102348327637,
+      "eval_runtime": 182.9051,
+      "eval_samples_per_second": 98.472,
+      "eval_steps_per_second": 6.156,
+      "step": 32000
+    },
+    {
+      "epoch": 3.449574857388871,
+      "grad_norm": 0.5549188852310181,
+      "learning_rate": 0.00039352655963796997,
+      "loss": 3.4655,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4549564094284793,
+      "grad_norm": 0.6282884478569031,
+      "learning_rate": 0.00039320331860790857,
+      "loss": 3.473,
+      "step": 32100
+    },
+    {
+      "epoch": 3.4603379614680874,
+      "grad_norm": 0.5992251634597778,
+      "learning_rate": 0.0003928800775778472,
+      "loss": 3.4684,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4657195135076955,
+      "grad_norm": 0.6112256050109863,
+      "learning_rate": 0.00039255683654778576,
+      "loss": 3.4635,
+      "step": 32200
+    },
+    {
+      "epoch": 3.471101065547304,
+      "grad_norm": 0.6498141288757324,
+      "learning_rate": 0.00039223359551772435,
+      "loss": 3.4595,
+      "step": 32250
+    },
+    {
+      "epoch": 3.476482617586912,
+      "grad_norm": 0.8926645517349243,
+      "learning_rate": 0.00039191035448766294,
+      "loss": 3.4693,
+      "step": 32300
+    },
+    {
+      "epoch": 3.4818641696265202,
+      "grad_norm": 0.5939327478408813,
+      "learning_rate": 0.0003915871134576015,
+      "loss": 3.4674,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4872457216661283,
+      "grad_norm": 0.5974373817443848,
+      "learning_rate": 0.0003912638724275401,
+      "loss": 3.4627,
+      "step": 32400
+    },
+    {
+      "epoch": 3.492627273705737,
+      "grad_norm": 0.6161805987358093,
+      "learning_rate": 0.00039094063139747873,
+      "loss": 3.4778,
+      "step": 32450
+    },
+    {
+      "epoch": 3.498008825745345,
+      "grad_norm": 0.626426637172699,
+      "learning_rate": 0.00039061739036741727,
+      "loss": 3.4582,
+      "step": 32500
+    },
+    {
+      "epoch": 3.503390377784953,
+      "grad_norm": 0.6194849014282227,
+      "learning_rate": 0.00039029414933735586,
+      "loss": 3.4596,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5087719298245617,
+      "grad_norm": 0.5703500509262085,
+      "learning_rate": 0.0003899709083072944,
+      "loss": 3.4511,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5141534818641698,
+      "grad_norm": 0.5987148880958557,
+      "learning_rate": 0.000389647667277233,
+      "loss": 3.4513,
+      "step": 32650
+    },
+    {
+      "epoch": 3.519535033903778,
+      "grad_norm": 0.6004769802093506,
+      "learning_rate": 0.00038932442624717165,
+      "loss": 3.4605,
+      "step": 32700
+    },
+    {
+      "epoch": 3.524916585943386,
+      "grad_norm": 0.6045653820037842,
+      "learning_rate": 0.0003890011852171102,
+      "loss": 3.4566,
+      "step": 32750
+    },
+    {
+      "epoch": 3.530298137982994,
+      "grad_norm": 0.6004464030265808,
+      "learning_rate": 0.0003886779441870488,
+      "loss": 3.4689,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5356796900226026,
+      "grad_norm": 0.6028228998184204,
+      "learning_rate": 0.0003883547031569874,
+      "loss": 3.4587,
+      "step": 32850
+    },
+    {
+      "epoch": 3.5410612420622107,
+      "grad_norm": 0.6320363879203796,
+      "learning_rate": 0.0003880314621269259,
+      "loss": 3.4762,
+      "step": 32900
+    },
+    {
+      "epoch": 3.546442794101819,
+      "grad_norm": 0.6261252164840698,
+      "learning_rate": 0.0003877082210968645,
+      "loss": 3.4774,
+      "step": 32950
+    },
+    {
+      "epoch": 3.5518243461414274,
+      "grad_norm": 0.5780692100524902,
+      "learning_rate": 0.00038738498006680316,
+      "loss": 3.468,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5518243461414274,
+      "eval_accuracy": 0.3728782399883698,
+      "eval_loss": 3.4731545448303223,
+      "eval_runtime": 182.9321,
+      "eval_samples_per_second": 98.457,
+      "eval_steps_per_second": 6.155,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5572058981810355,
+      "grad_norm": 0.5741873979568481,
+      "learning_rate": 0.0003870617390367417,
+      "loss": 3.4358,
+      "step": 33050
+    },
+    {
+      "epoch": 3.5625874502206436,
+      "grad_norm": 0.587976336479187,
+      "learning_rate": 0.0003867384980066803,
+      "loss": 3.4524,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5679690022602517,
+      "grad_norm": 0.5886296033859253,
+      "learning_rate": 0.00038641525697661884,
+      "loss": 3.4588,
+      "step": 33150
+    },
+    {
+      "epoch": 3.57335055429986,
+      "grad_norm": 0.6252454519271851,
+      "learning_rate": 0.00038609201594655743,
+      "loss": 3.4569,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5787321063394684,
+      "grad_norm": 0.6233946681022644,
+      "learning_rate": 0.000385768774916496,
+      "loss": 3.4602,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5841136583790765,
+      "grad_norm": 0.6176490783691406,
+      "learning_rate": 0.0003854455338864346,
+      "loss": 3.4731,
+      "step": 33300
+    },
+    {
+      "epoch": 3.5894952104186846,
+      "grad_norm": 0.585235059261322,
+      "learning_rate": 0.0003851222928563732,
+      "loss": 3.4542,
+      "step": 33350
+    },
+    {
+      "epoch": 3.594876762458293,
+      "grad_norm": 0.6597333550453186,
+      "learning_rate": 0.0003847990518263118,
+      "loss": 3.4511,
+      "step": 33400
+    },
+    {
+      "epoch": 3.6002583144979012,
+      "grad_norm": 0.5512192249298096,
+      "learning_rate": 0.0003844822756168516,
+      "loss": 3.471,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6056398665375093,
+      "grad_norm": 0.5989030003547668,
+      "learning_rate": 0.00038415903458679016,
+      "loss": 3.4706,
+      "step": 33500
+    },
+    {
+      "epoch": 3.611021418577118,
+      "grad_norm": 0.643202543258667,
+      "learning_rate": 0.00038383579355672875,
+      "loss": 3.4446,
+      "step": 33550
+    },
+    {
+      "epoch": 3.616402970616726,
+      "grad_norm": 0.6083823442459106,
+      "learning_rate": 0.00038351255252666735,
+      "loss": 3.4702,
+      "step": 33600
+    },
+    {
+      "epoch": 3.621784522656334,
+      "grad_norm": 0.6128966212272644,
+      "learning_rate": 0.00038318931149660594,
+      "loss": 3.4584,
+      "step": 33650
+    },
+    {
+      "epoch": 3.627166074695942,
+      "grad_norm": 0.6018299460411072,
+      "learning_rate": 0.00038286607046654454,
+      "loss": 3.4597,
+      "step": 33700
+    },
+    {
+      "epoch": 3.6325476267355503,
+      "grad_norm": 0.6082707643508911,
+      "learning_rate": 0.00038254282943648313,
+      "loss": 3.4459,
+      "step": 33750
+    },
+    {
+      "epoch": 3.637929178775159,
+      "grad_norm": 0.5940393805503845,
+      "learning_rate": 0.00038221958840642167,
+      "loss": 3.4514,
+      "step": 33800
+    },
+    {
+      "epoch": 3.643310730814767,
+      "grad_norm": 0.6136729121208191,
+      "learning_rate": 0.00038189634737636027,
+      "loss": 3.4703,
+      "step": 33850
+    },
+    {
+      "epoch": 3.648692282854375,
+      "grad_norm": 0.571419358253479,
+      "learning_rate": 0.0003815731063462988,
+      "loss": 3.4434,
+      "step": 33900
+    },
+    {
+      "epoch": 3.6540738348939836,
+      "grad_norm": 0.5937557220458984,
+      "learning_rate": 0.00038124986531623745,
+      "loss": 3.4589,
+      "step": 33950
+    },
+    {
+      "epoch": 3.6594553869335917,
+      "grad_norm": 0.6121259331703186,
+      "learning_rate": 0.00038092662428617605,
+      "loss": 3.4691,
+      "step": 34000
+    },
+    {
+      "epoch": 3.6594553869335917,
+      "eval_accuracy": 0.3735422177858041,
+      "eval_loss": 3.4664924144744873,
+      "eval_runtime": 183.1266,
+      "eval_samples_per_second": 98.353,
+      "eval_steps_per_second": 6.149,
+      "step": 34000
+    },
+    {
+      "epoch": 3.6648369389732,
+      "grad_norm": 0.6286929249763489,
+      "learning_rate": 0.0003806033832561146,
+      "loss": 3.4675,
+      "step": 34050
+    },
+    {
+      "epoch": 3.670218491012808,
+      "grad_norm": 0.6216891407966614,
+      "learning_rate": 0.0003802801422260532,
+      "loss": 3.4353,
+      "step": 34100
+    },
+    {
+      "epoch": 3.675600043052416,
+      "grad_norm": 0.5747387409210205,
+      "learning_rate": 0.0003799569011959918,
+      "loss": 3.4438,
+      "step": 34150
+    },
+    {
+      "epoch": 3.6809815950920246,
+      "grad_norm": 0.6043647527694702,
+      "learning_rate": 0.0003796336601659303,
+      "loss": 3.4652,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6863631471316327,
+      "grad_norm": 0.6015361547470093,
+      "learning_rate": 0.00037931041913586897,
+      "loss": 3.4511,
+      "step": 34250
+    },
+    {
+      "epoch": 3.691744699171241,
+      "grad_norm": 0.5804011225700378,
+      "learning_rate": 0.00037898717810580756,
+      "loss": 3.4713,
+      "step": 34300
+    },
+    {
+      "epoch": 3.6971262512108494,
+      "grad_norm": 0.5761457681655884,
+      "learning_rate": 0.0003786639370757461,
+      "loss": 3.4658,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7025078032504575,
+      "grad_norm": 0.5895703434944153,
+      "learning_rate": 0.0003783406960456847,
+      "loss": 3.4654,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7078893552900656,
+      "grad_norm": 0.6082561016082764,
+      "learning_rate": 0.00037801745501562324,
+      "loss": 3.4621,
+      "step": 34450
+    },
+    {
+      "epoch": 3.713270907329674,
+      "grad_norm": 0.6229493021965027,
+      "learning_rate": 0.0003776942139855619,
+      "loss": 3.4451,
+      "step": 34500
+    },
+    {
+      "epoch": 3.7186524593692822,
+      "grad_norm": 0.5866677165031433,
+      "learning_rate": 0.0003773709729555005,
+      "loss": 3.4595,
+      "step": 34550
+    },
+    {
+      "epoch": 3.7240340114088903,
+      "grad_norm": 0.6240717172622681,
+      "learning_rate": 0.000377047731925439,
+      "loss": 3.4571,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7294155634484984,
+      "grad_norm": 0.6172893047332764,
+      "learning_rate": 0.0003767244908953776,
+      "loss": 3.4676,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7347971154881066,
+      "grad_norm": 0.6289220452308655,
+      "learning_rate": 0.0003764012498653162,
+      "loss": 3.4464,
+      "step": 34700
+    },
+    {
+      "epoch": 3.740178667527715,
+      "grad_norm": 0.6364248991012573,
+      "learning_rate": 0.00037607800883525475,
+      "loss": 3.465,
+      "step": 34750
+    },
+    {
+      "epoch": 3.745560219567323,
+      "grad_norm": 0.6611288189888,
+      "learning_rate": 0.0003757547678051934,
+      "loss": 3.4465,
+      "step": 34800
+    },
+    {
+      "epoch": 3.7509417716069313,
+      "grad_norm": 0.5711609125137329,
+      "learning_rate": 0.000375431526775132,
+      "loss": 3.4531,
+      "step": 34850
+    },
+    {
+      "epoch": 3.75632332364654,
+      "grad_norm": 0.6151444315910339,
+      "learning_rate": 0.00037510828574507054,
+      "loss": 3.4685,
+      "step": 34900
+    },
+    {
+      "epoch": 3.761704875686148,
+      "grad_norm": 0.6230692267417908,
+      "learning_rate": 0.00037478504471500913,
+      "loss": 3.446,
+      "step": 34950
+    },
+    {
+      "epoch": 3.767086427725756,
+      "grad_norm": 0.6537262797355652,
+      "learning_rate": 0.00037446180368494767,
+      "loss": 3.4681,
+      "step": 35000
+    },
+    {
+      "epoch": 3.767086427725756,
+      "eval_accuracy": 0.37434853086627534,
+      "eval_loss": 3.460052728652954,
+      "eval_runtime": 182.9222,
+      "eval_samples_per_second": 98.463,
+      "eval_steps_per_second": 6.156,
+      "step": 35000
+    },
+    {
+      "epoch": 3.772467979765364,
+      "grad_norm": 0.6269912123680115,
+      "learning_rate": 0.00037413856265488627,
+      "loss": 3.4745,
+      "step": 35050
+    },
+    {
+      "epoch": 3.7778495318049723,
+      "grad_norm": 0.6322194337844849,
+      "learning_rate": 0.0003738153216248249,
+      "loss": 3.4693,
+      "step": 35100
+    },
+    {
+      "epoch": 3.783231083844581,
+      "grad_norm": 0.632526695728302,
+      "learning_rate": 0.00037349208059476346,
+      "loss": 3.4557,
+      "step": 35150
+    },
+    {
+      "epoch": 3.788612635884189,
+      "grad_norm": 0.5771685242652893,
+      "learning_rate": 0.00037316883956470205,
+      "loss": 3.4585,
+      "step": 35200
+    },
+    {
+      "epoch": 3.793994187923797,
+      "grad_norm": 0.5918084979057312,
+      "learning_rate": 0.00037284559853464064,
+      "loss": 3.4503,
+      "step": 35250
+    },
+    {
+      "epoch": 3.7993757399634056,
+      "grad_norm": 0.6038704514503479,
+      "learning_rate": 0.0003725223575045792,
+      "loss": 3.4631,
+      "step": 35300
+    },
+    {
+      "epoch": 3.8047572920030137,
+      "grad_norm": 0.5990985631942749,
+      "learning_rate": 0.0003721991164745178,
+      "loss": 3.4527,
+      "step": 35350
+    },
+    {
+      "epoch": 3.810138844042622,
+      "grad_norm": 0.6696856021881104,
+      "learning_rate": 0.00037187587544445643,
+      "loss": 3.4572,
+      "step": 35400
+    },
+    {
+      "epoch": 3.8155203960822304,
+      "grad_norm": 0.5810059309005737,
+      "learning_rate": 0.00037155909923499624,
+      "loss": 3.4439,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8209019481218385,
+      "grad_norm": 0.6126786470413208,
+      "learning_rate": 0.0003712358582049348,
+      "loss": 3.4521,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8262835001614466,
+      "grad_norm": 0.5759022235870361,
+      "learning_rate": 0.00037091261717487337,
+      "loss": 3.439,
+      "step": 35550
+    },
+    {
+      "epoch": 3.8316650522010547,
+      "grad_norm": 0.6269648671150208,
+      "learning_rate": 0.00037058937614481197,
+      "loss": 3.4444,
+      "step": 35600
+    },
+    {
+      "epoch": 3.837046604240663,
+      "grad_norm": 0.6104834079742432,
+      "learning_rate": 0.0003702661351147505,
+      "loss": 3.4494,
+      "step": 35650
+    },
+    {
+      "epoch": 3.8424281562802713,
+      "grad_norm": 0.623383641242981,
+      "learning_rate": 0.0003699428940846891,
+      "loss": 3.4771,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8478097083198795,
+      "grad_norm": 0.6111641526222229,
+      "learning_rate": 0.00036961965305462775,
+      "loss": 3.4553,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8531912603594876,
+      "grad_norm": 0.6193336844444275,
+      "learning_rate": 0.0003692964120245663,
+      "loss": 3.4714,
+      "step": 35800
+    },
+    {
+      "epoch": 3.858572812399096,
+      "grad_norm": 0.6333847641944885,
+      "learning_rate": 0.0003689731709945049,
+      "loss": 3.4516,
+      "step": 35850
+    },
+    {
+      "epoch": 3.863954364438704,
+      "grad_norm": 0.6365866661071777,
+      "learning_rate": 0.0003686499299644434,
+      "loss": 3.4439,
+      "step": 35900
+    },
+    {
+      "epoch": 3.8693359164783123,
+      "grad_norm": 0.6254480481147766,
+      "learning_rate": 0.000368326688934382,
+      "loss": 3.4289,
+      "step": 35950
+    },
+    {
+      "epoch": 3.8747174685179204,
+      "grad_norm": 0.6164137125015259,
+      "learning_rate": 0.0003680034479043206,
+      "loss": 3.4485,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8747174685179204,
+      "eval_accuracy": 0.37518298504469055,
+      "eval_loss": 3.4550580978393555,
+      "eval_runtime": 182.8899,
+      "eval_samples_per_second": 98.48,
+      "eval_steps_per_second": 6.157,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8800990205575285,
+      "grad_norm": 0.599606990814209,
+      "learning_rate": 0.0003676802068742592,
+      "loss": 3.4553,
+      "step": 36050
+    },
+    {
+      "epoch": 3.885480572597137,
+      "grad_norm": 0.5952869653701782,
+      "learning_rate": 0.0003673569658441978,
+      "loss": 3.4557,
+      "step": 36100
+    },
+    {
+      "epoch": 3.890862124636745,
+      "grad_norm": 0.6147683262825012,
+      "learning_rate": 0.0003670337248141364,
+      "loss": 3.4397,
+      "step": 36150
+    },
+    {
+      "epoch": 3.8962436766763533,
+      "grad_norm": 0.6182876229286194,
+      "learning_rate": 0.00036671048378407494,
+      "loss": 3.4621,
+      "step": 36200
+    },
+    {
+      "epoch": 3.901625228715962,
+      "grad_norm": 0.6036136746406555,
+      "learning_rate": 0.00036638724275401353,
+      "loss": 3.459,
+      "step": 36250
+    },
+    {
+      "epoch": 3.90700678075557,
+      "grad_norm": 0.6490514278411865,
+      "learning_rate": 0.0003660640017239522,
+      "loss": 3.4717,
+      "step": 36300
+    },
+    {
+      "epoch": 3.912388332795178,
+      "grad_norm": 0.5808064341545105,
+      "learning_rate": 0.0003657407606938907,
+      "loss": 3.4619,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9177698848347866,
+      "grad_norm": 0.6013696193695068,
+      "learning_rate": 0.0003654175196638293,
+      "loss": 3.4525,
+      "step": 36400
+    },
+    {
+      "epoch": 3.9231514368743947,
+      "grad_norm": 0.609664797782898,
+      "learning_rate": 0.00036509427863376786,
+      "loss": 3.4375,
+      "step": 36450
+    },
+    {
+      "epoch": 3.928532988914003,
+      "grad_norm": 0.6366161704063416,
+      "learning_rate": 0.00036477103760370645,
+      "loss": 3.4491,
+      "step": 36500
+    },
+    {
+      "epoch": 3.933914540953611,
+      "grad_norm": 0.5997065305709839,
+      "learning_rate": 0.00036444779657364505,
+      "loss": 3.4635,
+      "step": 36550
+    },
+    {
+      "epoch": 3.939296092993219,
+      "grad_norm": 0.6210324764251709,
+      "learning_rate": 0.00036412455554358364,
+      "loss": 3.4626,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9446776450328276,
+      "grad_norm": 0.6512171030044556,
+      "learning_rate": 0.00036380131451352224,
+      "loss": 3.4469,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9500591970724357,
+      "grad_norm": 0.6340328454971313,
+      "learning_rate": 0.00036347807348346083,
+      "loss": 3.4664,
+      "step": 36700
+    },
+    {
+      "epoch": 3.955440749112044,
+      "grad_norm": 0.5851263403892517,
+      "learning_rate": 0.00036315483245339937,
+      "loss": 3.4502,
+      "step": 36750
+    },
+    {
+      "epoch": 3.9608223011516523,
+      "grad_norm": 0.6642547845840454,
+      "learning_rate": 0.00036283159142333797,
+      "loss": 3.4394,
+      "step": 36800
+    },
+    {
+      "epoch": 3.9662038531912605,
+      "grad_norm": 0.6904765963554382,
+      "learning_rate": 0.0003625083503932765,
+      "loss": 3.4662,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9715854052308686,
+      "grad_norm": 0.6355055570602417,
+      "learning_rate": 0.00036218510936321516,
+      "loss": 3.4568,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9769669572704767,
+      "grad_norm": 0.585128664970398,
+      "learning_rate": 0.00036186186833315375,
+      "loss": 3.4772,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9823485093100848,
+      "grad_norm": 0.5850505232810974,
+      "learning_rate": 0.0003615386273030923,
+      "loss": 3.4428,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9823485093100848,
+      "eval_accuracy": 0.37551470231118816,
+      "eval_loss": 3.449396848678589,
+      "eval_runtime": 182.7917,
+      "eval_samples_per_second": 98.533,
+      "eval_steps_per_second": 6.16,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9877300613496933,
+      "grad_norm": 0.6555930376052856,
+      "learning_rate": 0.0003612153862730309,
+      "loss": 3.4316,
+      "step": 37050
+    },
+    {
+      "epoch": 3.9931116133893014,
+      "grad_norm": 0.6331802010536194,
+      "learning_rate": 0.0003608921452429695,
+      "loss": 3.4507,
+      "step": 37100
+    },
+    {
+      "epoch": 3.9984931654289095,
+      "grad_norm": 0.5921234488487244,
+      "learning_rate": 0.000360568904212908,
+      "loss": 3.4469,
+      "step": 37150
+    },
+    {
+      "epoch": 4.003874717468518,
+      "grad_norm": 0.627491295337677,
+      "learning_rate": 0.00036024566318284667,
+      "loss": 3.384,
+      "step": 37200
+    },
+    {
+      "epoch": 4.009256269508126,
+      "grad_norm": 0.6413177847862244,
+      "learning_rate": 0.00035992242215278526,
+      "loss": 3.367,
+      "step": 37250
+    },
+    {
+      "epoch": 4.014637821547734,
+      "grad_norm": 0.6309844255447388,
+      "learning_rate": 0.0003595991811227238,
+      "loss": 3.363,
+      "step": 37300
+    },
+    {
+      "epoch": 4.020019373587343,
+      "grad_norm": 0.6667311191558838,
+      "learning_rate": 0.0003592759400926624,
+      "loss": 3.3572,
+      "step": 37350
+    },
+    {
+      "epoch": 4.0254009256269505,
+      "grad_norm": 0.6064625382423401,
+      "learning_rate": 0.00035895269906260094,
+      "loss": 3.3597,
+      "step": 37400
+    },
+    {
+      "epoch": 4.030782477666559,
+      "grad_norm": 0.6433827877044678,
+      "learning_rate": 0.0003586359228531408,
+      "loss": 3.3492,
+      "step": 37450
+    },
+    {
+      "epoch": 4.036164029706168,
+      "grad_norm": 0.66212397813797,
+      "learning_rate": 0.00035831268182307934,
+      "loss": 3.3711,
+      "step": 37500
+    },
+    {
+      "epoch": 4.041545581745775,
+      "grad_norm": 0.6347793936729431,
+      "learning_rate": 0.000357989440793018,
+      "loss": 3.3516,
+      "step": 37550
+    },
+    {
+      "epoch": 4.046927133785384,
+      "grad_norm": 0.6682549715042114,
+      "learning_rate": 0.0003576661997629566,
+      "loss": 3.3482,
+      "step": 37600
+    },
+    {
+      "epoch": 4.0523086858249915,
+      "grad_norm": 1.5289701223373413,
+      "learning_rate": 0.0003573429587328951,
+      "loss": 3.3792,
+      "step": 37650
+    },
+    {
+      "epoch": 4.0576902378646,
+      "grad_norm": 0.6569393873214722,
+      "learning_rate": 0.0003570197177028337,
+      "loss": 3.3742,
+      "step": 37700
+    },
+    {
+      "epoch": 4.063071789904209,
+      "grad_norm": 0.630212128162384,
+      "learning_rate": 0.00035669647667277226,
+      "loss": 3.368,
+      "step": 37750
+    },
+    {
+      "epoch": 4.068453341943816,
+      "grad_norm": 0.6595259308815002,
+      "learning_rate": 0.00035637323564271085,
+      "loss": 3.3954,
+      "step": 37800
+    },
+    {
+      "epoch": 4.073834893983425,
+      "grad_norm": 0.6565824151039124,
+      "learning_rate": 0.0003560499946126495,
+      "loss": 3.3667,
+      "step": 37850
+    },
+    {
+      "epoch": 4.079216446023033,
+      "grad_norm": 0.6169015765190125,
+      "learning_rate": 0.00035572675358258804,
+      "loss": 3.38,
+      "step": 37900
+    },
+    {
+      "epoch": 4.084597998062641,
+      "grad_norm": 0.6400800347328186,
+      "learning_rate": 0.00035540351255252664,
+      "loss": 3.3704,
+      "step": 37950
+    },
+    {
+      "epoch": 4.08997955010225,
+      "grad_norm": 0.6978241801261902,
+      "learning_rate": 0.00035508027152246523,
+      "loss": 3.3682,
+      "step": 38000
+    },
+    {
+      "epoch": 4.08997955010225,
+      "eval_accuracy": 0.3757263581366508,
+      "eval_loss": 3.452324151992798,
+      "eval_runtime": 182.3279,
+      "eval_samples_per_second": 98.784,
+      "eval_steps_per_second": 6.176,
+      "step": 38000
+    },
+    {
+      "epoch": 4.095361102141858,
+      "grad_norm": 0.6283411383628845,
+      "learning_rate": 0.0003547570304924038,
+      "loss": 3.3862,
+      "step": 38050
+    },
+    {
+      "epoch": 4.100742654181466,
+      "grad_norm": 0.6307471394538879,
+      "learning_rate": 0.0003544337894623424,
+      "loss": 3.3747,
+      "step": 38100
+    },
+    {
+      "epoch": 4.106124206221074,
+      "grad_norm": 0.597294270992279,
+      "learning_rate": 0.000354110548432281,
+      "loss": 3.3646,
+      "step": 38150
+    },
+    {
+      "epoch": 4.111505758260682,
+      "grad_norm": 0.5953186750411987,
+      "learning_rate": 0.00035378730740221956,
+      "loss": 3.3824,
+      "step": 38200
+    },
+    {
+      "epoch": 4.1168873103002905,
+      "grad_norm": 0.624042272567749,
+      "learning_rate": 0.00035346406637215815,
+      "loss": 3.3978,
+      "step": 38250
+    },
+    {
+      "epoch": 4.122268862339899,
+      "grad_norm": 0.6373803615570068,
+      "learning_rate": 0.0003531408253420967,
+      "loss": 3.3895,
+      "step": 38300
+    },
+    {
+      "epoch": 4.127650414379507,
+      "grad_norm": 0.6241115927696228,
+      "learning_rate": 0.0003528175843120353,
+      "loss": 3.382,
+      "step": 38350
+    },
+    {
+      "epoch": 4.133031966419115,
+      "grad_norm": 0.6354626417160034,
+      "learning_rate": 0.00035249434328197394,
+      "loss": 3.3788,
+      "step": 38400
+    },
+    {
+      "epoch": 4.138413518458724,
+      "grad_norm": 0.6405816674232483,
+      "learning_rate": 0.0003521711022519125,
+      "loss": 3.4024,
+      "step": 38450
+    },
+    {
+      "epoch": 4.1437950704983315,
+      "grad_norm": 0.6165269613265991,
+      "learning_rate": 0.00035184786122185107,
+      "loss": 3.3723,
+      "step": 38500
+    },
+    {
+      "epoch": 4.14917662253794,
+      "grad_norm": 0.6947121024131775,
+      "learning_rate": 0.00035152462019178967,
+      "loss": 3.3761,
+      "step": 38550
+    },
+    {
+      "epoch": 4.154558174577549,
+      "grad_norm": 0.6098374724388123,
+      "learning_rate": 0.0003512013791617282,
+      "loss": 3.37,
+      "step": 38600
+    },
+    {
+      "epoch": 4.159939726617156,
+      "grad_norm": 0.6297115087509155,
+      "learning_rate": 0.0003508781381316668,
+      "loss": 3.3762,
+      "step": 38650
+    },
+    {
+      "epoch": 4.165321278656765,
+      "grad_norm": 0.6331884860992432,
+      "learning_rate": 0.00035055489710160545,
+      "loss": 3.371,
+      "step": 38700
+    },
+    {
+      "epoch": 4.1707028306963725,
+      "grad_norm": 0.693463146686554,
+      "learning_rate": 0.000350231656071544,
+      "loss": 3.3778,
+      "step": 38750
+    },
+    {
+      "epoch": 4.176084382735981,
+      "grad_norm": 0.6266399025917053,
+      "learning_rate": 0.0003499084150414826,
+      "loss": 3.3779,
+      "step": 38800
+    },
+    {
+      "epoch": 4.18146593477559,
+      "grad_norm": 0.6141369342803955,
+      "learning_rate": 0.0003495851740114211,
+      "loss": 3.3825,
+      "step": 38850
+    },
+    {
+      "epoch": 4.186847486815197,
+      "grad_norm": 0.632418155670166,
+      "learning_rate": 0.0003492619329813597,
+      "loss": 3.3927,
+      "step": 38900
+    },
+    {
+      "epoch": 4.192229038854806,
+      "grad_norm": 0.6271191239356995,
+      "learning_rate": 0.0003489386919512983,
+      "loss": 3.3714,
+      "step": 38950
+    },
+    {
+      "epoch": 4.197610590894414,
+      "grad_norm": 0.7061901688575745,
+      "learning_rate": 0.0003486154509212369,
+      "loss": 3.3809,
+      "step": 39000
+    },
+    {
+      "epoch": 4.197610590894414,
+      "eval_accuracy": 0.37637001784406376,
+      "eval_loss": 3.448011875152588,
+      "eval_runtime": 182.5317,
+      "eval_samples_per_second": 98.673,
+      "eval_steps_per_second": 6.169,
+      "step": 39000
+    },
+    {
+      "epoch": 4.202992142934022,
+      "grad_norm": 0.6355361342430115,
+      "learning_rate": 0.0003482922098911755,
+      "loss": 3.3757,
+      "step": 39050
+    },
+    {
+      "epoch": 4.208373694973631,
+      "grad_norm": 0.6252569556236267,
+      "learning_rate": 0.0003479689688611141,
+      "loss": 3.3928,
+      "step": 39100
+    },
+    {
+      "epoch": 4.213755247013238,
+      "grad_norm": 0.6373631358146667,
+      "learning_rate": 0.00034764572783105264,
+      "loss": 3.3829,
+      "step": 39150
+    },
+    {
+      "epoch": 4.219136799052847,
+      "grad_norm": 0.6141077280044556,
+      "learning_rate": 0.00034732248680099123,
+      "loss": 3.3808,
+      "step": 39200
+    },
+    {
+      "epoch": 4.224518351092455,
+      "grad_norm": 0.6714136004447937,
+      "learning_rate": 0.00034700571059153104,
+      "loss": 3.3783,
+      "step": 39250
+    },
+    {
+      "epoch": 4.229899903132063,
+      "grad_norm": 0.6982507705688477,
+      "learning_rate": 0.00034668246956146963,
+      "loss": 3.3836,
+      "step": 39300
+    },
+    {
+      "epoch": 4.2352814551716715,
+      "grad_norm": 0.6456509828567505,
+      "learning_rate": 0.00034635922853140823,
+      "loss": 3.382,
+      "step": 39350
+    },
+    {
+      "epoch": 4.24066300721128,
+      "grad_norm": 0.6480652689933777,
+      "learning_rate": 0.0003460359875013468,
+      "loss": 3.3935,
+      "step": 39400
+    },
+    {
+      "epoch": 4.246044559250888,
+      "grad_norm": 0.6829937696456909,
+      "learning_rate": 0.0003457127464712854,
+      "loss": 3.3913,
+      "step": 39450
+    },
+    {
+      "epoch": 4.251426111290496,
+      "grad_norm": 0.6466607451438904,
+      "learning_rate": 0.00034538950544122396,
+      "loss": 3.3826,
+      "step": 39500
+    },
+    {
+      "epoch": 4.256807663330104,
+      "grad_norm": 0.665008544921875,
+      "learning_rate": 0.00034506626441116255,
+      "loss": 3.3962,
+      "step": 39550
+    },
+    {
+      "epoch": 4.2621892153697125,
+      "grad_norm": 0.633686363697052,
+      "learning_rate": 0.0003447430233811011,
+      "loss": 3.3794,
+      "step": 39600
+    },
+    {
+      "epoch": 4.267570767409321,
+      "grad_norm": 0.706251859664917,
+      "learning_rate": 0.00034441978235103974,
+      "loss": 3.3995,
+      "step": 39650
+    },
+    {
+      "epoch": 4.272952319448929,
+      "grad_norm": 0.6631447076797485,
+      "learning_rate": 0.00034409654132097834,
+      "loss": 3.3963,
+      "step": 39700
+    },
+    {
+      "epoch": 4.278333871488537,
+      "grad_norm": 0.6600978374481201,
+      "learning_rate": 0.0003437733002909169,
+      "loss": 3.3984,
+      "step": 39750
+    },
+    {
+      "epoch": 4.283715423528146,
+      "grad_norm": 0.624860942363739,
+      "learning_rate": 0.00034345005926085547,
+      "loss": 3.3892,
+      "step": 39800
+    },
+    {
+      "epoch": 4.2890969755677535,
+      "grad_norm": 0.6655292510986328,
+      "learning_rate": 0.00034312681823079407,
+      "loss": 3.3873,
+      "step": 39850
+    },
+    {
+      "epoch": 4.294478527607362,
+      "grad_norm": 0.6067180037498474,
+      "learning_rate": 0.00034280357720073266,
+      "loss": 3.4034,
+      "step": 39900
+    },
+    {
+      "epoch": 4.299860079646971,
+      "grad_norm": 0.6818814873695374,
+      "learning_rate": 0.00034248033617067126,
+      "loss": 3.3754,
+      "step": 39950
+    },
+    {
+      "epoch": 4.305241631686578,
+      "grad_norm": 0.6350970268249512,
+      "learning_rate": 0.00034215709514060985,
+      "loss": 3.3874,
+      "step": 40000
+    },
+    {
+      "epoch": 4.305241631686578,
+      "eval_accuracy": 0.3769179543573122,
+      "eval_loss": 3.442443609237671,
+      "eval_runtime": 182.8947,
+      "eval_samples_per_second": 98.477,
+      "eval_steps_per_second": 6.157,
+      "step": 40000
+    },
+    {
+      "epoch": 4.310623183726187,
+      "grad_norm": 0.6428865790367126,
+      "learning_rate": 0.0003418338541105484,
+      "loss": 3.3958,
+      "step": 40050
+    },
+    {
+      "epoch": 4.3160047357657945,
+      "grad_norm": 0.6549261212348938,
+      "learning_rate": 0.000341510613080487,
+      "loss": 3.3934,
+      "step": 40100
+    },
+    {
+      "epoch": 4.321386287805403,
+      "grad_norm": 0.6633482575416565,
+      "learning_rate": 0.0003411873720504255,
+      "loss": 3.4035,
+      "step": 40150
+    },
+    {
+      "epoch": 4.326767839845012,
+      "grad_norm": 0.6681509613990784,
+      "learning_rate": 0.0003408641310203642,
+      "loss": 3.4035,
+      "step": 40200
+    },
+    {
+      "epoch": 4.332149391884619,
+      "grad_norm": 0.6671531200408936,
+      "learning_rate": 0.00034054088999030277,
+      "loss": 3.3972,
+      "step": 40250
+    },
+    {
+      "epoch": 4.337530943924228,
+      "grad_norm": 0.6722056865692139,
+      "learning_rate": 0.0003402176489602413,
+      "loss": 3.3925,
+      "step": 40300
+    },
+    {
+      "epoch": 4.342912495963836,
+      "grad_norm": 0.634778618812561,
+      "learning_rate": 0.0003398944079301799,
+      "loss": 3.3951,
+      "step": 40350
+    },
+    {
+      "epoch": 4.348294048003444,
+      "grad_norm": 0.6151086091995239,
+      "learning_rate": 0.0003395711669001185,
+      "loss": 3.4087,
+      "step": 40400
+    },
+    {
+      "epoch": 4.3536756000430525,
+      "grad_norm": 0.6367116570472717,
+      "learning_rate": 0.00033924792587005704,
+      "loss": 3.3993,
+      "step": 40450
+    },
+    {
+      "epoch": 4.359057152082661,
+      "grad_norm": 0.6156135201454163,
+      "learning_rate": 0.0003389246848399957,
+      "loss": 3.3765,
+      "step": 40500
+    },
+    {
+      "epoch": 4.364438704122269,
+      "grad_norm": 0.6550938487052917,
+      "learning_rate": 0.0003386014438099343,
+      "loss": 3.4011,
+      "step": 40550
+    },
+    {
+      "epoch": 4.369820256161877,
+      "grad_norm": 0.6099169254302979,
+      "learning_rate": 0.0003382782027798728,
+      "loss": 3.3948,
+      "step": 40600
+    },
+    {
+      "epoch": 4.375201808201485,
+      "grad_norm": 0.6787609457969666,
+      "learning_rate": 0.0003379549617498114,
+      "loss": 3.4118,
+      "step": 40650
+    },
+    {
+      "epoch": 4.3805833602410935,
+      "grad_norm": 0.6305999755859375,
+      "learning_rate": 0.00033763172071974996,
+      "loss": 3.4038,
+      "step": 40700
+    },
+    {
+      "epoch": 4.385964912280702,
+      "grad_norm": 0.6593831181526184,
+      "learning_rate": 0.00033730847968968855,
+      "loss": 3.4024,
+      "step": 40750
+    },
+    {
+      "epoch": 4.39134646432031,
+      "grad_norm": 0.6971632838249207,
+      "learning_rate": 0.0003369852386596272,
+      "loss": 3.3823,
+      "step": 40800
+    },
+    {
+      "epoch": 4.396728016359918,
+      "grad_norm": 0.6232396960258484,
+      "learning_rate": 0.00033666199762956574,
+      "loss": 3.4003,
+      "step": 40850
+    },
+    {
+      "epoch": 4.402109568399527,
+      "grad_norm": 0.6262663006782532,
+      "learning_rate": 0.00033633875659950434,
+      "loss": 3.391,
+      "step": 40900
+    },
+    {
+      "epoch": 4.4074911204391345,
+      "grad_norm": 0.652194082736969,
+      "learning_rate": 0.0003360155155694429,
+      "loss": 3.373,
+      "step": 40950
+    },
+    {
+      "epoch": 4.412872672478743,
+      "grad_norm": 0.645897388458252,
+      "learning_rate": 0.0003356922745393815,
+      "loss": 3.4029,
+      "step": 41000
+    },
+    {
+      "epoch": 4.412872672478743,
+      "eval_accuracy": 0.37729009049807677,
+      "eval_loss": 3.4402434825897217,
+      "eval_runtime": 182.7308,
+      "eval_samples_per_second": 98.566,
+      "eval_steps_per_second": 6.162,
+      "step": 41000
+    },
+    {
+      "epoch": 4.418254224518351,
+      "grad_norm": 0.6175161004066467,
+      "learning_rate": 0.0003353690335093201,
+      "loss": 3.3898,
+      "step": 41050
+    },
+    {
+      "epoch": 4.423635776557959,
+      "grad_norm": 0.6909878849983215,
+      "learning_rate": 0.0003350457924792587,
+      "loss": 3.3979,
+      "step": 41100
+    },
+    {
+      "epoch": 4.429017328597568,
+      "grad_norm": 0.6703479290008545,
+      "learning_rate": 0.00033472255144919726,
+      "loss": 3.403,
+      "step": 41150
+    },
+    {
+      "epoch": 4.4343988806371755,
+      "grad_norm": 0.635631799697876,
+      "learning_rate": 0.00033439931041913585,
+      "loss": 3.3764,
+      "step": 41200
+    },
+    {
+      "epoch": 4.439780432676784,
+      "grad_norm": 0.6716395616531372,
+      "learning_rate": 0.0003340760693890744,
+      "loss": 3.3872,
+      "step": 41250
+    },
+    {
+      "epoch": 4.445161984716393,
+      "grad_norm": 0.6503021717071533,
+      "learning_rate": 0.000333752828359013,
+      "loss": 3.3977,
+      "step": 41300
+    },
+    {
+      "epoch": 4.450543536756,
+      "grad_norm": 0.6269751191139221,
+      "learning_rate": 0.00033342958732895164,
+      "loss": 3.3794,
+      "step": 41350
+    },
+    {
+      "epoch": 4.455925088795609,
+      "grad_norm": 0.6848242282867432,
+      "learning_rate": 0.0003331063462988902,
+      "loss": 3.3979,
+      "step": 41400
+    },
+    {
+      "epoch": 4.461306640835216,
+      "grad_norm": 0.6538987755775452,
+      "learning_rate": 0.00033278310526882877,
+      "loss": 3.3977,
+      "step": 41450
+    },
+    {
+      "epoch": 4.466688192874825,
+      "grad_norm": 0.687504768371582,
+      "learning_rate": 0.0003324598642387673,
+      "loss": 3.388,
+      "step": 41500
+    },
+    {
+      "epoch": 4.4720697449144335,
+      "grad_norm": 0.6829115748405457,
+      "learning_rate": 0.0003321366232087059,
+      "loss": 3.3855,
+      "step": 41550
+    },
+    {
+      "epoch": 4.477451296954041,
+      "grad_norm": 0.6400046944618225,
+      "learning_rate": 0.0003318133821786445,
+      "loss": 3.3888,
+      "step": 41600
+    },
+    {
+      "epoch": 4.48283284899365,
+      "grad_norm": 0.7195608615875244,
+      "learning_rate": 0.0003314901411485831,
+      "loss": 3.386,
+      "step": 41650
+    },
+    {
+      "epoch": 4.488214401033258,
+      "grad_norm": 0.6601772904396057,
+      "learning_rate": 0.0003311669001185217,
+      "loss": 3.4025,
+      "step": 41700
+    },
+    {
+      "epoch": 4.493595953072866,
+      "grad_norm": 0.6627285480499268,
+      "learning_rate": 0.0003308436590884603,
+      "loss": 3.3845,
+      "step": 41750
+    },
+    {
+      "epoch": 4.4989775051124745,
+      "grad_norm": 0.6553217172622681,
+      "learning_rate": 0.0003305204180583988,
+      "loss": 3.3975,
+      "step": 41800
+    },
+    {
+      "epoch": 4.504359057152083,
+      "grad_norm": 0.6015192270278931,
+      "learning_rate": 0.0003302036418489387,
+      "loss": 3.4143,
+      "step": 41850
+    },
+    {
+      "epoch": 4.509740609191691,
+      "grad_norm": 0.709814727306366,
+      "learning_rate": 0.0003298804008188772,
+      "loss": 3.3908,
+      "step": 41900
+    },
+    {
+      "epoch": 4.515122161231299,
+      "grad_norm": 0.6477102637290955,
+      "learning_rate": 0.0003295571597888158,
+      "loss": 3.3887,
+      "step": 41950
+    },
+    {
+      "epoch": 4.520503713270907,
+      "grad_norm": 0.6681662201881409,
+      "learning_rate": 0.00032923391875875447,
+      "loss": 3.4236,
+      "step": 42000
+    },
+    {
+      "epoch": 4.520503713270907,
+      "eval_accuracy": 0.3778097772604935,
+      "eval_loss": 3.4332406520843506,
+      "eval_runtime": 182.6918,
+      "eval_samples_per_second": 98.587,
+      "eval_steps_per_second": 6.163,
+      "step": 42000
+    },
+    {
+      "epoch": 4.5258852653105155,
+      "grad_norm": 0.6888872385025024,
+      "learning_rate": 0.000328910677728693,
+      "loss": 3.399,
+      "step": 42050
+    },
+    {
+      "epoch": 4.531266817350124,
+      "grad_norm": 0.6245088577270508,
+      "learning_rate": 0.0003285874366986316,
+      "loss": 3.3807,
+      "step": 42100
+    },
+    {
+      "epoch": 4.536648369389732,
+      "grad_norm": 0.6789868474006653,
+      "learning_rate": 0.00032826419566857015,
+      "loss": 3.385,
+      "step": 42150
+    },
+    {
+      "epoch": 4.54202992142934,
+      "grad_norm": 0.678841769695282,
+      "learning_rate": 0.00032794095463850874,
+      "loss": 3.4002,
+      "step": 42200
+    },
+    {
+      "epoch": 4.547411473468949,
+      "grad_norm": 0.6410040855407715,
+      "learning_rate": 0.0003276177136084473,
+      "loss": 3.3948,
+      "step": 42250
+    },
+    {
+      "epoch": 4.5527930255085565,
+      "grad_norm": 0.6948608756065369,
+      "learning_rate": 0.00032729447257838593,
+      "loss": 3.4055,
+      "step": 42300
+    },
+    {
+      "epoch": 4.558174577548165,
+      "grad_norm": 0.6666476726531982,
+      "learning_rate": 0.0003269712315483245,
+      "loss": 3.4067,
+      "step": 42350
+    },
+    {
+      "epoch": 4.563556129587774,
+      "grad_norm": 0.6703059673309326,
+      "learning_rate": 0.00032664799051826306,
+      "loss": 3.3935,
+      "step": 42400
+    },
+    {
+      "epoch": 4.568937681627381,
+      "grad_norm": 0.6622447371482849,
+      "learning_rate": 0.00032632474948820166,
+      "loss": 3.3907,
+      "step": 42450
+    },
+    {
+      "epoch": 4.57431923366699,
+      "grad_norm": 0.6864114999771118,
+      "learning_rate": 0.00032600150845814025,
+      "loss": 3.3855,
+      "step": 42500
+    },
+    {
+      "epoch": 4.579700785706597,
+      "grad_norm": 0.6508845686912537,
+      "learning_rate": 0.0003256782674280788,
+      "loss": 3.3981,
+      "step": 42550
+    },
+    {
+      "epoch": 4.585082337746206,
+      "grad_norm": 0.6998566389083862,
+      "learning_rate": 0.00032535502639801744,
+      "loss": 3.3939,
+      "step": 42600
+    },
+    {
+      "epoch": 4.5904638897858145,
+      "grad_norm": 0.6499090194702148,
+      "learning_rate": 0.00032503178536795604,
+      "loss": 3.396,
+      "step": 42650
+    },
+    {
+      "epoch": 4.595845441825422,
+      "grad_norm": 0.6362233757972717,
+      "learning_rate": 0.0003247085443378946,
+      "loss": 3.394,
+      "step": 42700
+    },
+    {
+      "epoch": 4.601226993865031,
+      "grad_norm": 0.6605574488639832,
+      "learning_rate": 0.0003243853033078332,
+      "loss": 3.3787,
+      "step": 42750
+    },
+    {
+      "epoch": 4.606608545904638,
+      "grad_norm": 0.6438785195350647,
+      "learning_rate": 0.0003240620622777717,
+      "loss": 3.4107,
+      "step": 42800
+    },
+    {
+      "epoch": 4.611990097944247,
+      "grad_norm": 0.6508049368858337,
+      "learning_rate": 0.00032373882124771036,
+      "loss": 3.3849,
+      "step": 42850
+    },
+    {
+      "epoch": 4.6173716499838555,
+      "grad_norm": 0.651975691318512,
+      "learning_rate": 0.00032341558021764896,
+      "loss": 3.3751,
+      "step": 42900
+    },
+    {
+      "epoch": 4.622753202023463,
+      "grad_norm": 0.6845095157623291,
+      "learning_rate": 0.0003230923391875875,
+      "loss": 3.384,
+      "step": 42950
+    },
+    {
+      "epoch": 4.628134754063072,
+      "grad_norm": 0.6449900269508362,
+      "learning_rate": 0.0003227690981575261,
+      "loss": 3.3834,
+      "step": 43000
+    },
+    {
+      "epoch": 4.628134754063072,
+      "eval_accuracy": 0.37862836811728773,
+      "eval_loss": 3.428248643875122,
+      "eval_runtime": 182.9864,
+      "eval_samples_per_second": 98.428,
+      "eval_steps_per_second": 6.153,
+      "step": 43000
+    },
+    {
+      "epoch": 4.63351630610268,
+      "grad_norm": 0.7843633890151978,
+      "learning_rate": 0.0003224458571274647,
+      "loss": 3.394,
+      "step": 43050
+    },
+    {
+      "epoch": 4.638897858142288,
+      "grad_norm": 0.6628377437591553,
+      "learning_rate": 0.00032212261609740323,
+      "loss": 3.4261,
+      "step": 43100
+    },
+    {
+      "epoch": 4.6442794101818965,
+      "grad_norm": 0.6455616354942322,
+      "learning_rate": 0.0003217993750673419,
+      "loss": 3.3975,
+      "step": 43150
+    },
+    {
+      "epoch": 4.649660962221505,
+      "grad_norm": 0.6171681880950928,
+      "learning_rate": 0.00032147613403728047,
+      "loss": 3.3786,
+      "step": 43200
+    },
+    {
+      "epoch": 4.655042514261113,
+      "grad_norm": 0.6919983625411987,
+      "learning_rate": 0.000321152893007219,
+      "loss": 3.3941,
+      "step": 43250
+    },
+    {
+      "epoch": 4.660424066300721,
+      "grad_norm": 0.6645509004592896,
+      "learning_rate": 0.0003208296519771576,
+      "loss": 3.3736,
+      "step": 43300
+    },
+    {
+      "epoch": 4.665805618340329,
+      "grad_norm": 0.6968538165092468,
+      "learning_rate": 0.00032050641094709615,
+      "loss": 3.3966,
+      "step": 43350
+    },
+    {
+      "epoch": 4.6711871703799375,
+      "grad_norm": 0.6048688888549805,
+      "learning_rate": 0.00032018316991703474,
+      "loss": 3.3867,
+      "step": 43400
+    },
+    {
+      "epoch": 4.676568722419546,
+      "grad_norm": 0.638982892036438,
+      "learning_rate": 0.0003198599288869734,
+      "loss": 3.4118,
+      "step": 43450
+    },
+    {
+      "epoch": 4.681950274459154,
+      "grad_norm": 0.7195716500282288,
+      "learning_rate": 0.00031953668785691193,
+      "loss": 3.4058,
+      "step": 43500
+    },
+    {
+      "epoch": 4.687331826498762,
+      "grad_norm": 0.6764750480651855,
+      "learning_rate": 0.0003192134468268505,
+      "loss": 3.4037,
+      "step": 43550
+    },
+    {
+      "epoch": 4.692713378538371,
+      "grad_norm": 0.6833808422088623,
+      "learning_rate": 0.0003188902057967891,
+      "loss": 3.3974,
+      "step": 43600
+    },
+    {
+      "epoch": 4.6980949305779784,
+      "grad_norm": 0.6663907766342163,
+      "learning_rate": 0.00031856696476672766,
+      "loss": 3.3963,
+      "step": 43650
+    },
+    {
+      "epoch": 4.703476482617587,
+      "grad_norm": 0.661689281463623,
+      "learning_rate": 0.0003182437237366663,
+      "loss": 3.3907,
+      "step": 43700
+    },
+    {
+      "epoch": 4.7088580346571955,
+      "grad_norm": 0.6300943493843079,
+      "learning_rate": 0.0003179204827066049,
+      "loss": 3.3981,
+      "step": 43750
+    },
+    {
+      "epoch": 4.714239586696803,
+      "grad_norm": 0.6751649975776672,
+      "learning_rate": 0.00031759724167654344,
+      "loss": 3.4063,
+      "step": 43800
+    },
+    {
+      "epoch": 4.719621138736412,
+      "grad_norm": 0.6561940908432007,
+      "learning_rate": 0.00031727400064648204,
+      "loss": 3.4089,
+      "step": 43850
+    },
+    {
+      "epoch": 4.725002690776019,
+      "grad_norm": 0.6460406184196472,
+      "learning_rate": 0.0003169507596164206,
+      "loss": 3.4121,
+      "step": 43900
+    },
+    {
+      "epoch": 4.730384242815628,
+      "grad_norm": 0.7084411382675171,
+      "learning_rate": 0.0003166275185863592,
+      "loss": 3.397,
+      "step": 43950
+    },
+    {
+      "epoch": 4.7357657948552365,
+      "grad_norm": 0.6859269142150879,
+      "learning_rate": 0.0003163042775562978,
+      "loss": 3.3957,
+      "step": 44000
+    },
+    {
+      "epoch": 4.7357657948552365,
+      "eval_accuracy": 0.3787118135351293,
+      "eval_loss": 3.4233696460723877,
+      "eval_runtime": 182.724,
+      "eval_samples_per_second": 98.569,
+      "eval_steps_per_second": 6.162,
+      "step": 44000
+    },
+    {
+      "epoch": 4.741147346894844,
+      "grad_norm": 0.6808167695999146,
+      "learning_rate": 0.00031598103652623636,
+      "loss": 3.3983,
+      "step": 44050
+    },
+    {
+      "epoch": 4.746528898934453,
+      "grad_norm": 0.6858932375907898,
+      "learning_rate": 0.00031565779549617496,
+      "loss": 3.3932,
+      "step": 44100
+    },
+    {
+      "epoch": 4.751910450974061,
+      "grad_norm": 0.6018813252449036,
+      "learning_rate": 0.00031533455446611355,
+      "loss": 3.3963,
+      "step": 44150
+    },
+    {
+      "epoch": 4.757292003013669,
+      "grad_norm": 0.6454479098320007,
+      "learning_rate": 0.0003150113134360521,
+      "loss": 3.4044,
+      "step": 44200
+    },
+    {
+      "epoch": 4.7626735550532775,
+      "grad_norm": 0.68412846326828,
+      "learning_rate": 0.0003146880724059907,
+      "loss": 3.4075,
+      "step": 44250
+    },
+    {
+      "epoch": 4.768055107092886,
+      "grad_norm": 0.6248152852058411,
+      "learning_rate": 0.00031436483137592934,
+      "loss": 3.4068,
+      "step": 44300
+    },
+    {
+      "epoch": 4.773436659132494,
+      "grad_norm": 0.7138214111328125,
+      "learning_rate": 0.0003140415903458679,
+      "loss": 3.4007,
+      "step": 44350
+    },
+    {
+      "epoch": 4.778818211172102,
+      "grad_norm": 0.6977150440216064,
+      "learning_rate": 0.00031371834931580647,
+      "loss": 3.3887,
+      "step": 44400
+    },
+    {
+      "epoch": 4.78419976321171,
+      "grad_norm": 0.6793955564498901,
+      "learning_rate": 0.000313395108285745,
+      "loss": 3.3965,
+      "step": 44450
+    },
+    {
+      "epoch": 4.7895813152513185,
+      "grad_norm": 0.7022804617881775,
+      "learning_rate": 0.0003130718672556836,
+      "loss": 3.3973,
+      "step": 44500
+    },
+    {
+      "epoch": 4.794962867290927,
+      "grad_norm": 0.6722605228424072,
+      "learning_rate": 0.0003127486262256222,
+      "loss": 3.3857,
+      "step": 44550
+    },
+    {
+      "epoch": 4.800344419330535,
+      "grad_norm": 0.6196901798248291,
+      "learning_rate": 0.0003124253851955608,
+      "loss": 3.3995,
+      "step": 44600
+    },
+    {
+      "epoch": 4.805725971370143,
+      "grad_norm": 0.6370094418525696,
+      "learning_rate": 0.0003121021441654994,
+      "loss": 3.3821,
+      "step": 44650
+    },
+    {
+      "epoch": 4.811107523409751,
+      "grad_norm": 0.6328933238983154,
+      "learning_rate": 0.0003117853679560392,
+      "loss": 3.4183,
+      "step": 44700
+    },
+    {
+      "epoch": 4.8164890754493594,
+      "grad_norm": 0.7133774757385254,
+      "learning_rate": 0.0003114621269259778,
+      "loss": 3.376,
+      "step": 44750
+    },
+    {
+      "epoch": 4.821870627488968,
+      "grad_norm": 0.6690739393234253,
+      "learning_rate": 0.00031113888589591633,
+      "loss": 3.3853,
+      "step": 44800
+    },
+    {
+      "epoch": 4.827252179528576,
+      "grad_norm": 0.6346887350082397,
+      "learning_rate": 0.00031081564486585493,
+      "loss": 3.406,
+      "step": 44850
+    },
+    {
+      "epoch": 4.832633731568184,
+      "grad_norm": 0.7371925115585327,
+      "learning_rate": 0.0003104924038357935,
+      "loss": 3.3955,
+      "step": 44900
+    },
+    {
+      "epoch": 4.838015283607793,
+      "grad_norm": 0.651321530342102,
+      "learning_rate": 0.0003101691628057321,
+      "loss": 3.3972,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8433968356474,
+      "grad_norm": 0.6792730689048767,
+      "learning_rate": 0.0003098459217756707,
+      "loss": 3.3904,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8433968356474,
+      "eval_accuracy": 0.37944913203183833,
+      "eval_loss": 3.41748309135437,
+      "eval_runtime": 182.8169,
+      "eval_samples_per_second": 98.519,
+      "eval_steps_per_second": 6.159,
+      "step": 45000
+    },
+    {
+      "epoch": 4.848778387687009,
+      "grad_norm": 0.66081702709198,
+      "learning_rate": 0.0003095226807456093,
+      "loss": 3.3904,
+      "step": 45050
+    },
+    {
+      "epoch": 4.8541599397266175,
+      "grad_norm": 0.7229673266410828,
+      "learning_rate": 0.00030919943971554785,
+      "loss": 3.4034,
+      "step": 45100
+    },
+    {
+      "epoch": 4.859541491766225,
+      "grad_norm": 0.6608504056930542,
+      "learning_rate": 0.00030887619868548644,
+      "loss": 3.3975,
+      "step": 45150
+    },
+    {
+      "epoch": 4.864923043805834,
+      "grad_norm": 0.6587163805961609,
+      "learning_rate": 0.000308552957655425,
+      "loss": 3.3832,
+      "step": 45200
+    },
+    {
+      "epoch": 4.870304595845441,
+      "grad_norm": 0.6220030784606934,
+      "learning_rate": 0.00030822971662536363,
+      "loss": 3.3956,
+      "step": 45250
+    },
+    {
+      "epoch": 4.87568614788505,
+      "grad_norm": 0.6646186113357544,
+      "learning_rate": 0.0003079064755953022,
+      "loss": 3.4069,
+      "step": 45300
+    },
+    {
+      "epoch": 4.8810676999246585,
+      "grad_norm": 0.6473865509033203,
+      "learning_rate": 0.00030758323456524077,
+      "loss": 3.3964,
+      "step": 45350
+    },
+    {
+      "epoch": 4.886449251964266,
+      "grad_norm": 0.6625189781188965,
+      "learning_rate": 0.00030725999353517936,
+      "loss": 3.3962,
+      "step": 45400
+    },
+    {
+      "epoch": 4.891830804003875,
+      "grad_norm": 0.656329333782196,
+      "learning_rate": 0.00030693675250511795,
+      "loss": 3.3861,
+      "step": 45450
+    },
+    {
+      "epoch": 4.897212356043483,
+      "grad_norm": 0.6948346495628357,
+      "learning_rate": 0.00030661351147505655,
+      "loss": 3.392,
+      "step": 45500
+    },
+    {
+      "epoch": 4.902593908083091,
+      "grad_norm": 0.6399689316749573,
+      "learning_rate": 0.00030629027044499514,
+      "loss": 3.3961,
+      "step": 45550
+    },
+    {
+      "epoch": 4.9079754601226995,
+      "grad_norm": 0.6699696779251099,
+      "learning_rate": 0.00030596702941493374,
+      "loss": 3.389,
+      "step": 45600
+    },
+    {
+      "epoch": 4.913357012162308,
+      "grad_norm": 0.6273096799850464,
+      "learning_rate": 0.0003056437883848723,
+      "loss": 3.3964,
+      "step": 45650
+    },
+    {
+      "epoch": 4.918738564201916,
+      "grad_norm": 0.6409399509429932,
+      "learning_rate": 0.0003053205473548109,
+      "loss": 3.3961,
+      "step": 45700
+    },
+    {
+      "epoch": 4.924120116241524,
+      "grad_norm": 0.6525577306747437,
+      "learning_rate": 0.0003049973063247494,
+      "loss": 3.3739,
+      "step": 45750
+    },
+    {
+      "epoch": 4.929501668281132,
+      "grad_norm": 0.6267456412315369,
+      "learning_rate": 0.00030467406529468806,
+      "loss": 3.3891,
+      "step": 45800
+    },
+    {
+      "epoch": 4.9348832203207404,
+      "grad_norm": 0.610181987285614,
+      "learning_rate": 0.00030435082426462666,
+      "loss": 3.3808,
+      "step": 45850
+    },
+    {
+      "epoch": 4.940264772360349,
+      "grad_norm": 0.6637799739837646,
+      "learning_rate": 0.0003040275832345652,
+      "loss": 3.392,
+      "step": 45900
+    },
+    {
+      "epoch": 4.945646324399957,
+      "grad_norm": 0.7256836891174316,
+      "learning_rate": 0.0003037043422045038,
+      "loss": 3.3918,
+      "step": 45950
+    },
+    {
+      "epoch": 4.951027876439565,
+      "grad_norm": 0.6999911069869995,
+      "learning_rate": 0.0003033811011744424,
+      "loss": 3.3921,
+      "step": 46000
+    },
+    {
+      "epoch": 4.951027876439565,
+      "eval_accuracy": 0.37981605283398784,
+      "eval_loss": 3.412885904312134,
+      "eval_runtime": 183.1736,
+      "eval_samples_per_second": 98.327,
+      "eval_steps_per_second": 6.147,
+      "step": 46000
+    },
+    {
+      "epoch": 4.956409428479174,
+      "grad_norm": 0.6925266981124878,
+      "learning_rate": 0.00030305786014438093,
+      "loss": 3.4043,
+      "step": 46050
+    },
+    {
+      "epoch": 4.961790980518781,
+      "grad_norm": 0.6724205613136292,
+      "learning_rate": 0.0003027346191143196,
+      "loss": 3.3755,
+      "step": 46100
+    },
+    {
+      "epoch": 4.96717253255839,
+      "grad_norm": 0.6727187633514404,
+      "learning_rate": 0.00030241137808425817,
+      "loss": 3.3878,
+      "step": 46150
+    },
+    {
+      "epoch": 4.9725540845979985,
+      "grad_norm": 0.6427590250968933,
+      "learning_rate": 0.0003020881370541967,
+      "loss": 3.4044,
+      "step": 46200
+    },
+    {
+      "epoch": 4.977935636637606,
+      "grad_norm": 0.6517530083656311,
+      "learning_rate": 0.0003017648960241353,
+      "loss": 3.3773,
+      "step": 46250
+    },
+    {
+      "epoch": 4.983317188677215,
+      "grad_norm": 0.6581140160560608,
+      "learning_rate": 0.00030144165499407385,
+      "loss": 3.3772,
+      "step": 46300
+    },
+    {
+      "epoch": 4.988698740716822,
+      "grad_norm": 0.6741767525672913,
+      "learning_rate": 0.00030111841396401244,
+      "loss": 3.3952,
+      "step": 46350
+    },
+    {
+      "epoch": 4.994080292756431,
+      "grad_norm": 0.6367133259773254,
+      "learning_rate": 0.0003007951729339511,
+      "loss": 3.3955,
+      "step": 46400
+    },
+    {
+      "epoch": 4.9994618447960395,
+      "grad_norm": 0.6892203092575073,
+      "learning_rate": 0.00030047193190388963,
+      "loss": 3.3779,
+      "step": 46450
+    },
+    {
+      "epoch": 5.004843396835647,
+      "grad_norm": 0.6889623403549194,
+      "learning_rate": 0.0003001486908738282,
+      "loss": 3.3198,
+      "step": 46500
+    },
+    {
+      "epoch": 5.010224948875256,
+      "grad_norm": 0.6872815489768982,
+      "learning_rate": 0.0002998254498437668,
+      "loss": 3.3037,
+      "step": 46550
+    },
+    {
+      "epoch": 5.015606500914864,
+      "grad_norm": 0.6558205485343933,
+      "learning_rate": 0.0002995022088137054,
+      "loss": 3.2964,
+      "step": 46600
+    },
+    {
+      "epoch": 5.020988052954472,
+      "grad_norm": 0.6722169518470764,
+      "learning_rate": 0.00029917896778364396,
+      "loss": 3.2927,
+      "step": 46650
+    },
+    {
+      "epoch": 5.0263696049940805,
+      "grad_norm": 0.6730004549026489,
+      "learning_rate": 0.00029885572675358255,
+      "loss": 3.3064,
+      "step": 46700
+    },
+    {
+      "epoch": 5.031751157033688,
+      "grad_norm": 0.6149821281433105,
+      "learning_rate": 0.00029853248572352114,
+      "loss": 3.2977,
+      "step": 46750
+    },
+    {
+      "epoch": 5.037132709073297,
+      "grad_norm": 0.6445924639701843,
+      "learning_rate": 0.00029820924469345974,
+      "loss": 3.3056,
+      "step": 46800
+    },
+    {
+      "epoch": 5.042514261112905,
+      "grad_norm": 0.6724662780761719,
+      "learning_rate": 0.0002978860036633983,
+      "loss": 3.3142,
+      "step": 46850
+    },
+    {
+      "epoch": 5.047895813152513,
+      "grad_norm": 0.6918281316757202,
+      "learning_rate": 0.00029756276263333693,
+      "loss": 3.3051,
+      "step": 46900
+    },
+    {
+      "epoch": 5.0532773651921215,
+      "grad_norm": 0.6730775833129883,
+      "learning_rate": 0.00029723952160327547,
+      "loss": 3.3048,
+      "step": 46950
+    },
+    {
+      "epoch": 5.05865891723173,
+      "grad_norm": 0.6861401200294495,
+      "learning_rate": 0.00029691628057321406,
+      "loss": 3.3313,
+      "step": 47000
+    },
+    {
+      "epoch": 5.05865891723173,
+      "eval_accuracy": 0.3799867465207444,
+      "eval_loss": 3.4170210361480713,
+      "eval_runtime": 182.9491,
+      "eval_samples_per_second": 98.448,
+      "eval_steps_per_second": 6.155,
+      "step": 47000
+    },
+    {
+      "epoch": 5.064040469271338,
+      "grad_norm": 0.6703794002532959,
+      "learning_rate": 0.00029659303954315266,
+      "loss": 3.3135,
+      "step": 47050
+    },
+    {
+      "epoch": 5.069422021310946,
+      "grad_norm": 0.6781489849090576,
+      "learning_rate": 0.00029626979851309125,
+      "loss": 3.3085,
+      "step": 47100
+    },
+    {
+      "epoch": 5.074803573350554,
+      "grad_norm": 0.6689127683639526,
+      "learning_rate": 0.00029594655748302985,
+      "loss": 3.3014,
+      "step": 47150
+    },
+    {
+      "epoch": 5.080185125390162,
+      "grad_norm": 0.6758719682693481,
+      "learning_rate": 0.0002956233164529684,
+      "loss": 3.3099,
+      "step": 47200
+    },
+    {
+      "epoch": 5.085566677429771,
+      "grad_norm": 0.6829871535301208,
+      "learning_rate": 0.000295300075422907,
+      "loss": 3.3137,
+      "step": 47250
+    },
+    {
+      "epoch": 5.090948229469379,
+      "grad_norm": 0.7148377299308777,
+      "learning_rate": 0.0002949768343928456,
+      "loss": 3.3026,
+      "step": 47300
+    },
+    {
+      "epoch": 5.096329781508987,
+      "grad_norm": 0.706376314163208,
+      "learning_rate": 0.00029465359336278417,
+      "loss": 3.3088,
+      "step": 47350
+    },
+    {
+      "epoch": 5.101711333548596,
+      "grad_norm": 0.658332109451294,
+      "learning_rate": 0.0002943303523327227,
+      "loss": 3.3188,
+      "step": 47400
+    },
+    {
+      "epoch": 5.107092885588203,
+      "grad_norm": 0.6532283425331116,
+      "learning_rate": 0.00029400711130266136,
+      "loss": 3.3087,
+      "step": 47450
+    },
+    {
+      "epoch": 5.112474437627812,
+      "grad_norm": 0.670990526676178,
+      "learning_rate": 0.0002936838702725999,
+      "loss": 3.319,
+      "step": 47500
+    },
+    {
+      "epoch": 5.1178559896674205,
+      "grad_norm": 0.6792743802070618,
+      "learning_rate": 0.0002933606292425385,
+      "loss": 3.3171,
+      "step": 47550
+    },
+    {
+      "epoch": 5.123237541707028,
+      "grad_norm": 0.6803917288780212,
+      "learning_rate": 0.0002930373882124771,
+      "loss": 3.3264,
+      "step": 47600
+    },
+    {
+      "epoch": 5.128619093746637,
+      "grad_norm": 0.6759734749794006,
+      "learning_rate": 0.0002927141471824157,
+      "loss": 3.3128,
+      "step": 47650
+    },
+    {
+      "epoch": 5.134000645786244,
+      "grad_norm": 0.6762935519218445,
+      "learning_rate": 0.0002923909061523542,
+      "loss": 3.3171,
+      "step": 47700
+    },
+    {
+      "epoch": 5.139382197825853,
+      "grad_norm": 0.7197624444961548,
+      "learning_rate": 0.0002920676651222928,
+      "loss": 3.3111,
+      "step": 47750
+    },
+    {
+      "epoch": 5.1447637498654615,
+      "grad_norm": 0.6673755049705505,
+      "learning_rate": 0.0002917444240922314,
+      "loss": 3.3256,
+      "step": 47800
+    },
+    {
+      "epoch": 5.150145301905069,
+      "grad_norm": 0.6834902167320251,
+      "learning_rate": 0.00029142118306216996,
+      "loss": 3.3293,
+      "step": 47850
+    },
+    {
+      "epoch": 5.155526853944678,
+      "grad_norm": 0.6759807467460632,
+      "learning_rate": 0.0002910979420321086,
+      "loss": 3.3295,
+      "step": 47900
+    },
+    {
+      "epoch": 5.160908405984286,
+      "grad_norm": 0.7257971167564392,
+      "learning_rate": 0.00029077470100204715,
+      "loss": 3.3301,
+      "step": 47950
+    },
+    {
+      "epoch": 5.166289958023894,
+      "grad_norm": 0.7746021747589111,
+      "learning_rate": 0.00029045145997198574,
+      "loss": 3.333,
+      "step": 48000
+    },
+    {
+      "epoch": 5.166289958023894,
+      "eval_accuracy": 0.3801953600653482,
+      "eval_loss": 3.416738986968994,
+      "eval_runtime": 183.2917,
+      "eval_samples_per_second": 98.264,
+      "eval_steps_per_second": 6.143,
+      "step": 48000
+    },
+    {
+      "epoch": 5.1716715100635025,
+      "grad_norm": 0.6825886964797974,
+      "learning_rate": 0.00029012821894192433,
+      "loss": 3.3222,
+      "step": 48050
+    },
+    {
+      "epoch": 5.17705306210311,
+      "grad_norm": 0.6440089344978333,
+      "learning_rate": 0.00028980497791186293,
+      "loss": 3.3423,
+      "step": 48100
+    },
+    {
+      "epoch": 5.182434614142719,
+      "grad_norm": 0.6887645125389099,
+      "learning_rate": 0.0002894817368818015,
+      "loss": 3.3232,
+      "step": 48150
+    },
+    {
+      "epoch": 5.187816166182327,
+      "grad_norm": 0.646507978439331,
+      "learning_rate": 0.00028915849585174006,
+      "loss": 3.3021,
+      "step": 48200
+    },
+    {
+      "epoch": 5.193197718221935,
+      "grad_norm": 0.726085364818573,
+      "learning_rate": 0.00028883525482167866,
+      "loss": 3.3173,
+      "step": 48250
+    },
+    {
+      "epoch": 5.198579270261543,
+      "grad_norm": 0.6602137684822083,
+      "learning_rate": 0.00028851201379161725,
+      "loss": 3.3115,
+      "step": 48300
+    },
+    {
+      "epoch": 5.203960822301152,
+      "grad_norm": 0.65272456407547,
+      "learning_rate": 0.00028818877276155585,
+      "loss": 3.34,
+      "step": 48350
+    },
+    {
+      "epoch": 5.20934237434076,
+      "grad_norm": 0.6690257787704468,
+      "learning_rate": 0.0002878655317314944,
+      "loss": 3.3203,
+      "step": 48400
+    },
+    {
+      "epoch": 5.214723926380368,
+      "grad_norm": 0.6628714203834534,
+      "learning_rate": 0.00028754229070143304,
+      "loss": 3.3076,
+      "step": 48450
+    },
+    {
+      "epoch": 5.220105478419977,
+      "grad_norm": 0.6842119097709656,
+      "learning_rate": 0.0002872190496713716,
+      "loss": 3.331,
+      "step": 48500
+    },
+    {
+      "epoch": 5.225487030459584,
+      "grad_norm": 0.696015477180481,
+      "learning_rate": 0.0002868958086413102,
+      "loss": 3.3258,
+      "step": 48550
+    },
+    {
+      "epoch": 5.230868582499193,
+      "grad_norm": 0.6456036567687988,
+      "learning_rate": 0.00028657256761124877,
+      "loss": 3.3165,
+      "step": 48600
+    },
+    {
+      "epoch": 5.236250134538801,
+      "grad_norm": 0.6485376954078674,
+      "learning_rate": 0.00028624932658118736,
+      "loss": 3.3451,
+      "step": 48650
+    },
+    {
+      "epoch": 5.241631686578409,
+      "grad_norm": 0.659721314907074,
+      "learning_rate": 0.0002859260855511259,
+      "loss": 3.3305,
+      "step": 48700
+    },
+    {
+      "epoch": 5.247013238618018,
+      "grad_norm": 0.6913601160049438,
+      "learning_rate": 0.00028560930934166576,
+      "loss": 3.3227,
+      "step": 48750
+    },
+    {
+      "epoch": 5.252394790657625,
+      "grad_norm": 0.6780300736427307,
+      "learning_rate": 0.00028528606831160436,
+      "loss": 3.331,
+      "step": 48800
+    },
+    {
+      "epoch": 5.257776342697234,
+      "grad_norm": 0.6792374849319458,
+      "learning_rate": 0.0002849628272815429,
+      "loss": 3.3381,
+      "step": 48850
+    },
+    {
+      "epoch": 5.2631578947368425,
+      "grad_norm": 0.6867091655731201,
+      "learning_rate": 0.0002846395862514815,
+      "loss": 3.3138,
+      "step": 48900
+    },
+    {
+      "epoch": 5.26853944677645,
+      "grad_norm": 0.6632397770881653,
+      "learning_rate": 0.0002843163452214201,
+      "loss": 3.3316,
+      "step": 48950
+    },
+    {
+      "epoch": 5.273920998816059,
+      "grad_norm": 0.6680681705474854,
+      "learning_rate": 0.0002839931041913587,
+      "loss": 3.3374,
+      "step": 49000
+    },
+    {
+      "epoch": 5.273920998816059,
+      "eval_accuracy": 0.380897692332181,
+      "eval_loss": 3.412996768951416,
+      "eval_runtime": 182.5799,
+      "eval_samples_per_second": 98.647,
+      "eval_steps_per_second": 6.167,
+      "step": 49000
+    },
+    {
+      "epoch": 5.279302550855666,
+      "grad_norm": 0.7043352127075195,
+      "learning_rate": 0.0002836698631612972,
+      "loss": 3.3241,
+      "step": 49050
+    },
+    {
+      "epoch": 5.284684102895275,
+      "grad_norm": 0.6589981913566589,
+      "learning_rate": 0.00028334662213123587,
+      "loss": 3.3239,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2900656549348835,
+      "grad_norm": 0.6708270907402039,
+      "learning_rate": 0.0002830233811011744,
+      "loss": 3.3458,
+      "step": 49150
+    },
+    {
+      "epoch": 5.295447206974491,
+      "grad_norm": 0.7209701538085938,
+      "learning_rate": 0.000282700140071113,
+      "loss": 3.3305,
+      "step": 49200
+    },
+    {
+      "epoch": 5.3008287590141,
+      "grad_norm": 0.6668233871459961,
+      "learning_rate": 0.0002823768990410516,
+      "loss": 3.3244,
+      "step": 49250
+    },
+    {
+      "epoch": 5.306210311053708,
+      "grad_norm": 0.7002584934234619,
+      "learning_rate": 0.00028205365801099014,
+      "loss": 3.3261,
+      "step": 49300
+    },
+    {
+      "epoch": 5.311591863093316,
+      "grad_norm": 0.6543057560920715,
+      "learning_rate": 0.00028173041698092874,
+      "loss": 3.3427,
+      "step": 49350
+    },
+    {
+      "epoch": 5.316973415132924,
+      "grad_norm": 0.6252505779266357,
+      "learning_rate": 0.00028140717595086733,
+      "loss": 3.3378,
+      "step": 49400
+    },
+    {
+      "epoch": 5.322354967172533,
+      "grad_norm": 0.7060573101043701,
+      "learning_rate": 0.0002810839349208059,
+      "loss": 3.3324,
+      "step": 49450
+    },
+    {
+      "epoch": 5.327736519212141,
+      "grad_norm": 0.7204796075820923,
+      "learning_rate": 0.00028076069389074447,
+      "loss": 3.339,
+      "step": 49500
+    },
+    {
+      "epoch": 5.333118071251749,
+      "grad_norm": 0.7042475342750549,
+      "learning_rate": 0.0002804374528606831,
+      "loss": 3.3392,
+      "step": 49550
+    },
+    {
+      "epoch": 5.338499623291357,
+      "grad_norm": 0.6284716129302979,
+      "learning_rate": 0.00028011421183062166,
+      "loss": 3.3334,
+      "step": 49600
+    },
+    {
+      "epoch": 5.343881175330965,
+      "grad_norm": 0.7170406579971313,
+      "learning_rate": 0.00027979097080056025,
+      "loss": 3.3344,
+      "step": 49650
+    },
+    {
+      "epoch": 5.349262727370574,
+      "grad_norm": 0.6857984066009521,
+      "learning_rate": 0.00027946772977049885,
+      "loss": 3.3497,
+      "step": 49700
+    },
+    {
+      "epoch": 5.354644279410182,
+      "grad_norm": 0.6786603331565857,
+      "learning_rate": 0.00027914448874043744,
+      "loss": 3.3371,
+      "step": 49750
+    },
+    {
+      "epoch": 5.36002583144979,
+      "grad_norm": 0.6698256134986877,
+      "learning_rate": 0.00027882124771037603,
+      "loss": 3.3305,
+      "step": 49800
+    },
+    {
+      "epoch": 5.365407383489399,
+      "grad_norm": 0.6916489601135254,
+      "learning_rate": 0.0002784980066803146,
+      "loss": 3.3315,
+      "step": 49850
+    },
+    {
+      "epoch": 5.370788935529006,
+      "grad_norm": 0.6628431081771851,
+      "learning_rate": 0.00027817476565025317,
+      "loss": 3.3218,
+      "step": 49900
+    },
+    {
+      "epoch": 5.376170487568615,
+      "grad_norm": 0.6686885356903076,
+      "learning_rate": 0.00027785152462019176,
+      "loss": 3.344,
+      "step": 49950
+    },
+    {
+      "epoch": 5.3815520396082235,
+      "grad_norm": 0.6686023473739624,
+      "learning_rate": 0.00027752828359013036,
+      "loss": 3.3217,
+      "step": 50000
+    },
+    {
+      "epoch": 5.3815520396082235,
+      "eval_accuracy": 0.3814086868635725,
+      "eval_loss": 3.4041335582733154,
+      "eval_runtime": 182.7349,
+      "eval_samples_per_second": 98.564,
+      "eval_steps_per_second": 6.162,
+      "step": 50000
+    },
+    {
+      "epoch": 5.386933591647831,
+      "grad_norm": 0.7179530262947083,
+      "learning_rate": 0.0002772050425600689,
+      "loss": 3.3474,
+      "step": 50050
+    },
+    {
+      "epoch": 5.39231514368744,
+      "grad_norm": 0.7137858867645264,
+      "learning_rate": 0.00027688180153000755,
+      "loss": 3.342,
+      "step": 50100
+    },
+    {
+      "epoch": 5.397696695727047,
+      "grad_norm": 0.6810168027877808,
+      "learning_rate": 0.0002765585604999461,
+      "loss": 3.3531,
+      "step": 50150
+    },
+    {
+      "epoch": 5.403078247766656,
+      "grad_norm": 0.6813143491744995,
+      "learning_rate": 0.0002762353194698847,
+      "loss": 3.3448,
+      "step": 50200
+    },
+    {
+      "epoch": 5.4084597998062645,
+      "grad_norm": 0.7286673188209534,
+      "learning_rate": 0.0002759120784398233,
+      "loss": 3.351,
+      "step": 50250
+    },
+    {
+      "epoch": 5.413841351845872,
+      "grad_norm": 0.670843243598938,
+      "learning_rate": 0.00027558883740976187,
+      "loss": 3.3436,
+      "step": 50300
+    },
+    {
+      "epoch": 5.419222903885481,
+      "grad_norm": 0.7313069105148315,
+      "learning_rate": 0.0002752655963797004,
+      "loss": 3.3302,
+      "step": 50350
+    },
+    {
+      "epoch": 5.424604455925088,
+      "grad_norm": 0.7491310834884644,
+      "learning_rate": 0.000274942355349639,
+      "loss": 3.3428,
+      "step": 50400
+    },
+    {
+      "epoch": 5.429986007964697,
+      "grad_norm": 0.6367394924163818,
+      "learning_rate": 0.0002746191143195776,
+      "loss": 3.3479,
+      "step": 50450
+    },
+    {
+      "epoch": 5.435367560004305,
+      "grad_norm": 0.6688310503959656,
+      "learning_rate": 0.0002742958732895162,
+      "loss": 3.3284,
+      "step": 50500
+    },
+    {
+      "epoch": 5.440749112043913,
+      "grad_norm": 0.7176879644393921,
+      "learning_rate": 0.0002739726322594548,
+      "loss": 3.3527,
+      "step": 50550
+    },
+    {
+      "epoch": 5.446130664083522,
+      "grad_norm": 0.6491169333457947,
+      "learning_rate": 0.00027364939122939333,
+      "loss": 3.3275,
+      "step": 50600
+    },
+    {
+      "epoch": 5.45151221612313,
+      "grad_norm": 0.75172358751297,
+      "learning_rate": 0.0002733261501993319,
+      "loss": 3.3364,
+      "step": 50650
+    },
+    {
+      "epoch": 5.456893768162738,
+      "grad_norm": 0.6788306832313538,
+      "learning_rate": 0.0002730029091692705,
+      "loss": 3.3293,
+      "step": 50700
+    },
+    {
+      "epoch": 5.462275320202346,
+      "grad_norm": 0.6975774765014648,
+      "learning_rate": 0.0002726861329598104,
+      "loss": 3.3368,
+      "step": 50750
+    },
+    {
+      "epoch": 5.467656872241955,
+      "grad_norm": 0.6561357378959656,
+      "learning_rate": 0.0002723628919297489,
+      "loss": 3.3413,
+      "step": 50800
+    },
+    {
+      "epoch": 5.473038424281563,
+      "grad_norm": 0.6256112456321716,
+      "learning_rate": 0.0002720396508996875,
+      "loss": 3.3281,
+      "step": 50850
+    },
+    {
+      "epoch": 5.478419976321171,
+      "grad_norm": 0.721541166305542,
+      "learning_rate": 0.0002717164098696261,
+      "loss": 3.3276,
+      "step": 50900
+    },
+    {
+      "epoch": 5.483801528360779,
+      "grad_norm": 0.6738684773445129,
+      "learning_rate": 0.00027139316883956465,
+      "loss": 3.35,
+      "step": 50950
+    },
+    {
+      "epoch": 5.489183080400387,
+      "grad_norm": 0.6639395952224731,
+      "learning_rate": 0.00027106992780950325,
+      "loss": 3.3285,
+      "step": 51000
+    },
+    {
+      "epoch": 5.489183080400387,
+      "eval_accuracy": 0.381844710902372,
+      "eval_loss": 3.4026851654052734,
+      "eval_runtime": 182.4881,
+      "eval_samples_per_second": 98.697,
+      "eval_steps_per_second": 6.17,
+      "step": 51000
+    },
+    {
+      "epoch": 5.494564632439996,
+      "grad_norm": 0.6820451617240906,
+      "learning_rate": 0.00027074668677944184,
+      "loss": 3.3397,
+      "step": 51050
+    },
+    {
+      "epoch": 5.499946184479604,
+      "grad_norm": 0.6895923018455505,
+      "learning_rate": 0.00027042344574938044,
+      "loss": 3.3435,
+      "step": 51100
+    },
+    {
+      "epoch": 5.505327736519212,
+      "grad_norm": 0.6487290859222412,
+      "learning_rate": 0.000270100204719319,
+      "loss": 3.3394,
+      "step": 51150
+    },
+    {
+      "epoch": 5.510709288558821,
+      "grad_norm": 0.682824969291687,
+      "learning_rate": 0.0002697769636892576,
+      "loss": 3.3384,
+      "step": 51200
+    },
+    {
+      "epoch": 5.516090840598428,
+      "grad_norm": 0.6406248807907104,
+      "learning_rate": 0.00026945372265919617,
+      "loss": 3.3383,
+      "step": 51250
+    },
+    {
+      "epoch": 5.521472392638037,
+      "grad_norm": 0.6504181027412415,
+      "learning_rate": 0.00026913048162913476,
+      "loss": 3.3614,
+      "step": 51300
+    },
+    {
+      "epoch": 5.5268539446776455,
+      "grad_norm": 0.7287770509719849,
+      "learning_rate": 0.00026880724059907336,
+      "loss": 3.3374,
+      "step": 51350
+    },
+    {
+      "epoch": 5.532235496717253,
+      "grad_norm": 0.7049496173858643,
+      "learning_rate": 0.00026848399956901195,
+      "loss": 3.3374,
+      "step": 51400
+    },
+    {
+      "epoch": 5.537617048756862,
+      "grad_norm": 0.659416913986206,
+      "learning_rate": 0.0002681607585389505,
+      "loss": 3.3432,
+      "step": 51450
+    },
+    {
+      "epoch": 5.542998600796469,
+      "grad_norm": 0.6534185409545898,
+      "learning_rate": 0.0002678375175088891,
+      "loss": 3.3324,
+      "step": 51500
+    },
+    {
+      "epoch": 5.548380152836078,
+      "grad_norm": 0.6897971630096436,
+      "learning_rate": 0.0002675142764788277,
+      "loss": 3.3326,
+      "step": 51550
+    },
+    {
+      "epoch": 5.553761704875686,
+      "grad_norm": 0.6762552261352539,
+      "learning_rate": 0.0002671910354487663,
+      "loss": 3.3378,
+      "step": 51600
+    },
+    {
+      "epoch": 5.559143256915294,
+      "grad_norm": 0.6849751472473145,
+      "learning_rate": 0.00026686779441870487,
+      "loss": 3.3355,
+      "step": 51650
+    },
+    {
+      "epoch": 5.564524808954903,
+      "grad_norm": 0.6968287229537964,
+      "learning_rate": 0.0002665445533886434,
+      "loss": 3.344,
+      "step": 51700
+    },
+    {
+      "epoch": 5.569906360994511,
+      "grad_norm": 0.6949653029441833,
+      "learning_rate": 0.00026622131235858206,
+      "loss": 3.3416,
+      "step": 51750
+    },
+    {
+      "epoch": 5.575287913034119,
+      "grad_norm": 0.677852213382721,
+      "learning_rate": 0.0002658980713285206,
+      "loss": 3.3566,
+      "step": 51800
+    },
+    {
+      "epoch": 5.580669465073727,
+      "grad_norm": 0.6766265034675598,
+      "learning_rate": 0.0002655748302984592,
+      "loss": 3.3509,
+      "step": 51850
+    },
+    {
+      "epoch": 5.586051017113336,
+      "grad_norm": 0.6820150017738342,
+      "learning_rate": 0.0002652515892683978,
+      "loss": 3.3529,
+      "step": 51900
+    },
+    {
+      "epoch": 5.591432569152944,
+      "grad_norm": 0.7299242615699768,
+      "learning_rate": 0.0002649283482383364,
+      "loss": 3.3359,
+      "step": 51950
+    },
+    {
+      "epoch": 5.596814121192552,
+      "grad_norm": 0.67701256275177,
+      "learning_rate": 0.0002646051072082749,
+      "loss": 3.3538,
+      "step": 52000
+    },
+    {
+      "epoch": 5.596814121192552,
+      "eval_accuracy": 0.38225780918184266,
+      "eval_loss": 3.3965156078338623,
+      "eval_runtime": 182.6924,
+      "eval_samples_per_second": 98.586,
+      "eval_steps_per_second": 6.163,
+      "step": 52000
+    },
+    {
+      "epoch": 5.60219567323216,
+      "grad_norm": 0.6956893801689148,
+      "learning_rate": 0.0002642818661782135,
+      "loss": 3.3411,
+      "step": 52050
+    },
+    {
+      "epoch": 5.607577225271768,
+      "grad_norm": 0.6988450288772583,
+      "learning_rate": 0.0002639586251481521,
+      "loss": 3.3266,
+      "step": 52100
+    },
+    {
+      "epoch": 5.612958777311377,
+      "grad_norm": 0.6658537983894348,
+      "learning_rate": 0.0002636353841180907,
+      "loss": 3.3534,
+      "step": 52150
+    },
+    {
+      "epoch": 5.618340329350985,
+      "grad_norm": 0.7193296551704407,
+      "learning_rate": 0.0002633121430880293,
+      "loss": 3.3355,
+      "step": 52200
+    },
+    {
+      "epoch": 5.623721881390593,
+      "grad_norm": 0.76201993227005,
+      "learning_rate": 0.00026298890205796784,
+      "loss": 3.3333,
+      "step": 52250
+    },
+    {
+      "epoch": 5.629103433430201,
+      "grad_norm": 0.6642473340034485,
+      "learning_rate": 0.00026266566102790644,
+      "loss": 3.3183,
+      "step": 52300
+    },
+    {
+      "epoch": 5.634484985469809,
+      "grad_norm": 0.7340889573097229,
+      "learning_rate": 0.00026234241999784503,
+      "loss": 3.3426,
+      "step": 52350
+    },
+    {
+      "epoch": 5.639866537509418,
+      "grad_norm": 0.6511938571929932,
+      "learning_rate": 0.0002620191789677836,
+      "loss": 3.3428,
+      "step": 52400
+    },
+    {
+      "epoch": 5.645248089549026,
+      "grad_norm": 0.7516026496887207,
+      "learning_rate": 0.00026169593793772217,
+      "loss": 3.3299,
+      "step": 52450
+    },
+    {
+      "epoch": 5.650629641588634,
+      "grad_norm": 0.6919920444488525,
+      "learning_rate": 0.0002613726969076608,
+      "loss": 3.3476,
+      "step": 52500
+    },
+    {
+      "epoch": 5.656011193628243,
+      "grad_norm": 0.6615656018257141,
+      "learning_rate": 0.00026104945587759936,
+      "loss": 3.3536,
+      "step": 52550
+    },
+    {
+      "epoch": 5.66139274566785,
+      "grad_norm": 0.6720345616340637,
+      "learning_rate": 0.00026072621484753795,
+      "loss": 3.3274,
+      "step": 52600
+    },
+    {
+      "epoch": 5.666774297707459,
+      "grad_norm": 0.7008869647979736,
+      "learning_rate": 0.00026040297381747655,
+      "loss": 3.3585,
+      "step": 52650
+    },
+    {
+      "epoch": 5.672155849747067,
+      "grad_norm": 0.6563118696212769,
+      "learning_rate": 0.00026007973278741514,
+      "loss": 3.3435,
+      "step": 52700
+    },
+    {
+      "epoch": 5.677537401786675,
+      "grad_norm": 0.7016202211380005,
+      "learning_rate": 0.00025975649175735373,
+      "loss": 3.3271,
+      "step": 52750
+    },
+    {
+      "epoch": 5.682918953826284,
+      "grad_norm": 0.6965526938438416,
+      "learning_rate": 0.0002594397155478935,
+      "loss": 3.3241,
+      "step": 52800
+    },
+    {
+      "epoch": 5.688300505865891,
+      "grad_norm": 0.6786898374557495,
+      "learning_rate": 0.00025911647451783214,
+      "loss": 3.3389,
+      "step": 52850
+    },
+    {
+      "epoch": 5.6936820579055,
+      "grad_norm": 0.7544721364974976,
+      "learning_rate": 0.0002587932334877707,
+      "loss": 3.3585,
+      "step": 52900
+    },
+    {
+      "epoch": 5.699063609945108,
+      "grad_norm": 0.6955776810646057,
+      "learning_rate": 0.00025846999245770927,
+      "loss": 3.3371,
+      "step": 52950
+    },
+    {
+      "epoch": 5.704445161984716,
+      "grad_norm": 0.7224853038787842,
+      "learning_rate": 0.00025814675142764787,
+      "loss": 3.3236,
+      "step": 53000
+    },
+    {
+      "epoch": 5.704445161984716,
+      "eval_accuracy": 0.3824792437672086,
+      "eval_loss": 3.3930840492248535,
+      "eval_runtime": 182.5431,
+      "eval_samples_per_second": 98.667,
+      "eval_steps_per_second": 6.168,
+      "step": 53000
+    },
+    {
+      "epoch": 5.709826714024325,
+      "grad_norm": 0.6928134560585022,
+      "learning_rate": 0.00025782351039758646,
+      "loss": 3.3293,
+      "step": 53050
+    },
+    {
+      "epoch": 5.715208266063933,
+      "grad_norm": 0.7381559610366821,
+      "learning_rate": 0.000257500269367525,
+      "loss": 3.3342,
+      "step": 53100
+    },
+    {
+      "epoch": 5.720589818103541,
+      "grad_norm": 0.6981896162033081,
+      "learning_rate": 0.0002571770283374636,
+      "loss": 3.3432,
+      "step": 53150
+    },
+    {
+      "epoch": 5.725971370143149,
+      "grad_norm": 0.7084531188011169,
+      "learning_rate": 0.0002568537873074022,
+      "loss": 3.3424,
+      "step": 53200
+    },
+    {
+      "epoch": 5.731352922182758,
+      "grad_norm": 0.6834819316864014,
+      "learning_rate": 0.0002565305462773408,
+      "loss": 3.3341,
+      "step": 53250
+    },
+    {
+      "epoch": 5.736734474222366,
+      "grad_norm": 0.6777629256248474,
+      "learning_rate": 0.0002562073052472794,
+      "loss": 3.3627,
+      "step": 53300
+    },
+    {
+      "epoch": 5.742116026261974,
+      "grad_norm": 0.6871793270111084,
+      "learning_rate": 0.0002558840642172179,
+      "loss": 3.3307,
+      "step": 53350
+    },
+    {
+      "epoch": 5.747497578301582,
+      "grad_norm": 0.6921470165252686,
+      "learning_rate": 0.00025556082318715657,
+      "loss": 3.337,
+      "step": 53400
+    },
+    {
+      "epoch": 5.75287913034119,
+      "grad_norm": 0.6944039463996887,
+      "learning_rate": 0.0002552375821570951,
+      "loss": 3.357,
+      "step": 53450
+    },
+    {
+      "epoch": 5.758260682380799,
+      "grad_norm": 0.7080736756324768,
+      "learning_rate": 0.0002549143411270337,
+      "loss": 3.3487,
+      "step": 53500
+    },
+    {
+      "epoch": 5.763642234420407,
+      "grad_norm": 0.6794440746307373,
+      "learning_rate": 0.0002545911000969723,
+      "loss": 3.344,
+      "step": 53550
+    },
+    {
+      "epoch": 5.769023786460015,
+      "grad_norm": 0.7343289256095886,
+      "learning_rate": 0.0002542678590669109,
+      "loss": 3.3341,
+      "step": 53600
+    },
+    {
+      "epoch": 5.774405338499624,
+      "grad_norm": 0.7471538186073303,
+      "learning_rate": 0.00025394461803684943,
+      "loss": 3.3386,
+      "step": 53650
+    },
+    {
+      "epoch": 5.779786890539231,
+      "grad_norm": 0.7091389894485474,
+      "learning_rate": 0.00025362137700678803,
+      "loss": 3.3487,
+      "step": 53700
+    },
+    {
+      "epoch": 5.78516844257884,
+      "grad_norm": 0.7032609581947327,
+      "learning_rate": 0.0002532981359767266,
+      "loss": 3.3359,
+      "step": 53750
+    },
+    {
+      "epoch": 5.790549994618448,
+      "grad_norm": 0.7148309350013733,
+      "learning_rate": 0.0002529748949466652,
+      "loss": 3.3482,
+      "step": 53800
+    },
+    {
+      "epoch": 5.795931546658056,
+      "grad_norm": 0.6615278124809265,
+      "learning_rate": 0.0002526516539166038,
+      "loss": 3.3418,
+      "step": 53850
+    },
+    {
+      "epoch": 5.801313098697665,
+      "grad_norm": 0.6663516163825989,
+      "learning_rate": 0.00025232841288654235,
+      "loss": 3.3545,
+      "step": 53900
+    },
+    {
+      "epoch": 5.806694650737272,
+      "grad_norm": 0.680908739566803,
+      "learning_rate": 0.00025200517185648095,
+      "loss": 3.3334,
+      "step": 53950
+    },
+    {
+      "epoch": 5.812076202776881,
+      "grad_norm": 0.7598308324813843,
+      "learning_rate": 0.00025168193082641954,
+      "loss": 3.3326,
+      "step": 54000
+    },
+    {
+      "epoch": 5.812076202776881,
+      "eval_accuracy": 0.3831473503743798,
+      "eval_loss": 3.387981414794922,
+      "eval_runtime": 182.8436,
+      "eval_samples_per_second": 98.505,
+      "eval_steps_per_second": 6.158,
+      "step": 54000
+    },
+    {
+      "epoch": 5.817457754816489,
+      "grad_norm": 0.6939805746078491,
+      "learning_rate": 0.00025135868979635814,
+      "loss": 3.3218,
+      "step": 54050
+    },
+    {
+      "epoch": 5.822839306856097,
+      "grad_norm": 0.6918478608131409,
+      "learning_rate": 0.0002510354487662967,
+      "loss": 3.3544,
+      "step": 54100
+    },
+    {
+      "epoch": 5.828220858895706,
+      "grad_norm": 0.703644335269928,
+      "learning_rate": 0.0002507122077362353,
+      "loss": 3.3398,
+      "step": 54150
+    },
+    {
+      "epoch": 5.833602410935313,
+      "grad_norm": 0.6816033720970154,
+      "learning_rate": 0.00025038896670617387,
+      "loss": 3.3412,
+      "step": 54200
+    },
+    {
+      "epoch": 5.838983962974922,
+      "grad_norm": 0.7181455492973328,
+      "learning_rate": 0.00025006572567611246,
+      "loss": 3.3289,
+      "step": 54250
+    },
+    {
+      "epoch": 5.84436551501453,
+      "grad_norm": 0.6734956502914429,
+      "learning_rate": 0.00024974248464605106,
+      "loss": 3.3377,
+      "step": 54300
+    },
+    {
+      "epoch": 5.849747067054138,
+      "grad_norm": 0.703682005405426,
+      "learning_rate": 0.00024941924361598965,
+      "loss": 3.3415,
+      "step": 54350
+    },
+    {
+      "epoch": 5.855128619093747,
+      "grad_norm": 0.7012028694152832,
+      "learning_rate": 0.00024909600258592825,
+      "loss": 3.3428,
+      "step": 54400
+    },
+    {
+      "epoch": 5.860510171133355,
+      "grad_norm": 0.7386256456375122,
+      "learning_rate": 0.0002487727615558668,
+      "loss": 3.3364,
+      "step": 54450
+    },
+    {
+      "epoch": 5.865891723172963,
+      "grad_norm": 0.685370683670044,
+      "learning_rate": 0.0002484495205258054,
+      "loss": 3.3489,
+      "step": 54500
+    },
+    {
+      "epoch": 5.871273275212571,
+      "grad_norm": 0.7215905785560608,
+      "learning_rate": 0.0002481327443163452,
+      "loss": 3.3417,
+      "step": 54550
+    },
+    {
+      "epoch": 5.87665482725218,
+      "grad_norm": 0.6396303176879883,
+      "learning_rate": 0.0002478095032862838,
+      "loss": 3.3425,
+      "step": 54600
+    },
+    {
+      "epoch": 5.882036379291788,
+      "grad_norm": 0.7235729098320007,
+      "learning_rate": 0.0002474862622562224,
+      "loss": 3.3433,
+      "step": 54650
+    },
+    {
+      "epoch": 5.887417931331396,
+      "grad_norm": 0.6854497790336609,
+      "learning_rate": 0.00024716302122616097,
+      "loss": 3.3425,
+      "step": 54700
+    },
+    {
+      "epoch": 5.892799483371004,
+      "grad_norm": 0.7037459015846252,
+      "learning_rate": 0.0002468397801960995,
+      "loss": 3.343,
+      "step": 54750
+    },
+    {
+      "epoch": 5.898181035410612,
+      "grad_norm": 0.6938320398330688,
+      "learning_rate": 0.0002465165391660381,
+      "loss": 3.3392,
+      "step": 54800
+    },
+    {
+      "epoch": 5.903562587450221,
+      "grad_norm": 0.7277181148529053,
+      "learning_rate": 0.0002461932981359767,
+      "loss": 3.34,
+      "step": 54850
+    },
+    {
+      "epoch": 5.9089441394898286,
+      "grad_norm": 0.7088878750801086,
+      "learning_rate": 0.0002458700571059153,
+      "loss": 3.3342,
+      "step": 54900
+    },
+    {
+      "epoch": 5.914325691529437,
+      "grad_norm": 0.7254401445388794,
+      "learning_rate": 0.0002455468160758539,
+      "loss": 3.334,
+      "step": 54950
+    },
+    {
+      "epoch": 5.919707243569046,
+      "grad_norm": 0.6835173964500427,
+      "learning_rate": 0.00024522357504579243,
+      "loss": 3.3383,
+      "step": 55000
+    },
+    {
+      "epoch": 5.919707243569046,
+      "eval_accuracy": 0.38354132574559513,
+      "eval_loss": 3.384719133377075,
+      "eval_runtime": 183.0018,
+      "eval_samples_per_second": 98.42,
+      "eval_steps_per_second": 6.153,
+      "step": 55000
+    },
+    {
+      "epoch": 5.925088795608653,
+      "grad_norm": 0.6644113063812256,
+      "learning_rate": 0.000244900334015731,
+      "loss": 3.3335,
+      "step": 55050
+    },
+    {
+      "epoch": 5.930470347648262,
+      "grad_norm": 0.6974128484725952,
+      "learning_rate": 0.0002445770929856696,
+      "loss": 3.3146,
+      "step": 55100
+    },
+    {
+      "epoch": 5.93585189968787,
+      "grad_norm": 0.6761989593505859,
+      "learning_rate": 0.0002442538519556082,
+      "loss": 3.3408,
+      "step": 55150
+    },
+    {
+      "epoch": 5.941233451727478,
+      "grad_norm": 0.7247626185417175,
+      "learning_rate": 0.0002439306109255468,
+      "loss": 3.3309,
+      "step": 55200
+    },
+    {
+      "epoch": 5.946615003767087,
+      "grad_norm": 0.7041187882423401,
+      "learning_rate": 0.00024360736989548538,
+      "loss": 3.3517,
+      "step": 55250
+    },
+    {
+      "epoch": 5.951996555806694,
+      "grad_norm": 0.6591287851333618,
+      "learning_rate": 0.00024328412886542394,
+      "loss": 3.3237,
+      "step": 55300
+    },
+    {
+      "epoch": 5.957378107846303,
+      "grad_norm": 0.6712716817855835,
+      "learning_rate": 0.00024296088783536257,
+      "loss": 3.3449,
+      "step": 55350
+    },
+    {
+      "epoch": 5.962759659885911,
+      "grad_norm": 0.7153301239013672,
+      "learning_rate": 0.00024263764680530113,
+      "loss": 3.3493,
+      "step": 55400
+    },
+    {
+      "epoch": 5.968141211925519,
+      "grad_norm": 0.6891111731529236,
+      "learning_rate": 0.0002423144057752397,
+      "loss": 3.3462,
+      "step": 55450
+    },
+    {
+      "epoch": 5.973522763965128,
+      "grad_norm": 0.7062400579452515,
+      "learning_rate": 0.00024199116474517832,
+      "loss": 3.3469,
+      "step": 55500
+    },
+    {
+      "epoch": 5.978904316004736,
+      "grad_norm": 0.763090193271637,
+      "learning_rate": 0.0002416679237151169,
+      "loss": 3.3347,
+      "step": 55550
+    },
+    {
+      "epoch": 5.984285868044344,
+      "grad_norm": 0.7440812587738037,
+      "learning_rate": 0.00024134468268505546,
+      "loss": 3.3311,
+      "step": 55600
+    },
+    {
+      "epoch": 5.989667420083952,
+      "grad_norm": 0.7073535323143005,
+      "learning_rate": 0.00024102144165499405,
+      "loss": 3.3357,
+      "step": 55650
+    },
+    {
+      "epoch": 5.995048972123561,
+      "grad_norm": 0.7137259244918823,
+      "learning_rate": 0.00024069820062493265,
+      "loss": 3.3291,
+      "step": 55700
+    },
+    {
+      "epoch": 6.000430524163169,
+      "grad_norm": 0.7363245487213135,
+      "learning_rate": 0.00024037495959487121,
+      "loss": 3.3288,
+      "step": 55750
+    },
+    {
+      "epoch": 6.005812076202777,
+      "grad_norm": 0.6792331337928772,
+      "learning_rate": 0.0002400517185648098,
+      "loss": 3.2392,
+      "step": 55800
+    },
+    {
+      "epoch": 6.011193628242385,
+      "grad_norm": 0.7233385443687439,
+      "learning_rate": 0.00023972847753474838,
+      "loss": 3.2546,
+      "step": 55850
+    },
+    {
+      "epoch": 6.016575180281993,
+      "grad_norm": 0.721964418888092,
+      "learning_rate": 0.00023940523650468697,
+      "loss": 3.2373,
+      "step": 55900
+    },
+    {
+      "epoch": 6.021956732321602,
+      "grad_norm": 0.7094835638999939,
+      "learning_rate": 0.00023908199547462557,
+      "loss": 3.2454,
+      "step": 55950
+    },
+    {
+      "epoch": 6.0273382843612096,
+      "grad_norm": 0.9055956602096558,
+      "learning_rate": 0.00023875875444456413,
+      "loss": 3.2422,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0273382843612096,
+      "eval_accuracy": 0.3834130066850862,
+      "eval_loss": 3.3882665634155273,
+      "eval_runtime": 182.8427,
+      "eval_samples_per_second": 98.505,
+      "eval_steps_per_second": 6.158,
+      "step": 56000
+    },
+    {
+      "epoch": 6.032719836400818,
+      "grad_norm": 0.703880250453949,
+      "learning_rate": 0.0002384355134145027,
+      "loss": 3.2557,
+      "step": 56050
+    },
+    {
+      "epoch": 6.038101388440427,
+      "grad_norm": 0.7022672295570374,
+      "learning_rate": 0.00023811227238444132,
+      "loss": 3.2543,
+      "step": 56100
+    },
+    {
+      "epoch": 6.043482940480034,
+      "grad_norm": 0.6952441334724426,
+      "learning_rate": 0.0002377890313543799,
+      "loss": 3.2414,
+      "step": 56150
+    },
+    {
+      "epoch": 6.048864492519643,
+      "grad_norm": 0.6822983622550964,
+      "learning_rate": 0.00023746579032431849,
+      "loss": 3.2709,
+      "step": 56200
+    },
+    {
+      "epoch": 6.0542460445592505,
+      "grad_norm": 0.7127947807312012,
+      "learning_rate": 0.00023714254929425708,
+      "loss": 3.2604,
+      "step": 56250
+    },
+    {
+      "epoch": 6.059627596598859,
+      "grad_norm": 0.7314467430114746,
+      "learning_rate": 0.00023681930826419565,
+      "loss": 3.2558,
+      "step": 56300
+    },
+    {
+      "epoch": 6.065009148638468,
+      "grad_norm": 0.7139166593551636,
+      "learning_rate": 0.00023649606723413424,
+      "loss": 3.2292,
+      "step": 56350
+    },
+    {
+      "epoch": 6.070390700678075,
+      "grad_norm": 0.7124274969100952,
+      "learning_rate": 0.0002361728262040728,
+      "loss": 3.2714,
+      "step": 56400
+    },
+    {
+      "epoch": 6.075772252717684,
+      "grad_norm": 0.718776285648346,
+      "learning_rate": 0.00023584958517401138,
+      "loss": 3.2716,
+      "step": 56450
+    },
+    {
+      "epoch": 6.081153804757292,
+      "grad_norm": 0.6894996166229248,
+      "learning_rate": 0.00023552634414395,
+      "loss": 3.2668,
+      "step": 56500
+    },
+    {
+      "epoch": 6.0865353567969,
+      "grad_norm": 0.729155421257019,
+      "learning_rate": 0.00023520310311388857,
+      "loss": 3.2687,
+      "step": 56550
+    },
+    {
+      "epoch": 6.091916908836509,
+      "grad_norm": 0.7677587866783142,
+      "learning_rate": 0.00023487986208382713,
+      "loss": 3.2677,
+      "step": 56600
+    },
+    {
+      "epoch": 6.097298460876116,
+      "grad_norm": 0.7513812184333801,
+      "learning_rate": 0.00023455662105376576,
+      "loss": 3.2547,
+      "step": 56650
+    },
+    {
+      "epoch": 6.102680012915725,
+      "grad_norm": 0.691321611404419,
+      "learning_rate": 0.00023423338002370432,
+      "loss": 3.2662,
+      "step": 56700
+    },
+    {
+      "epoch": 6.108061564955333,
+      "grad_norm": 0.70525062084198,
+      "learning_rate": 0.0002339101389936429,
+      "loss": 3.2635,
+      "step": 56750
+    },
+    {
+      "epoch": 6.113443116994941,
+      "grad_norm": 0.6724135279655457,
+      "learning_rate": 0.00023358689796358149,
+      "loss": 3.271,
+      "step": 56800
+    },
+    {
+      "epoch": 6.11882466903455,
+      "grad_norm": 0.7082234025001526,
+      "learning_rate": 0.00023326365693352008,
+      "loss": 3.256,
+      "step": 56850
+    },
+    {
+      "epoch": 6.124206221074158,
+      "grad_norm": 0.7638316750526428,
+      "learning_rate": 0.00023294041590345865,
+      "loss": 3.2694,
+      "step": 56900
+    },
+    {
+      "epoch": 6.129587773113766,
+      "grad_norm": 0.7289136052131653,
+      "learning_rate": 0.00023261717487339724,
+      "loss": 3.2692,
+      "step": 56950
+    },
+    {
+      "epoch": 6.134969325153374,
+      "grad_norm": 0.7192388772964478,
+      "learning_rate": 0.0002322939338433358,
+      "loss": 3.2646,
+      "step": 57000
+    },
+    {
+      "epoch": 6.134969325153374,
+      "eval_accuracy": 0.38360651747828384,
+      "eval_loss": 3.388136863708496,
+      "eval_runtime": 182.7928,
+      "eval_samples_per_second": 98.532,
+      "eval_steps_per_second": 6.16,
+      "step": 57000
+    },
+    {
+      "epoch": 6.140350877192983,
+      "grad_norm": 0.7573813796043396,
+      "learning_rate": 0.0002319706928132744,
+      "loss": 3.2741,
+      "step": 57050
+    },
+    {
+      "epoch": 6.1457324292325906,
+      "grad_norm": 0.732204020023346,
+      "learning_rate": 0.000231647451783213,
+      "loss": 3.2918,
+      "step": 57100
+    },
+    {
+      "epoch": 6.151113981272199,
+      "grad_norm": 0.7195457816123962,
+      "learning_rate": 0.00023132421075315157,
+      "loss": 3.2779,
+      "step": 57150
+    },
+    {
+      "epoch": 6.156495533311807,
+      "grad_norm": 0.7368000745773315,
+      "learning_rate": 0.0002310009697230902,
+      "loss": 3.2607,
+      "step": 57200
+    },
+    {
+      "epoch": 6.161877085351415,
+      "grad_norm": 0.711700975894928,
+      "learning_rate": 0.00023067772869302876,
+      "loss": 3.2661,
+      "step": 57250
+    },
+    {
+      "epoch": 6.167258637391024,
+      "grad_norm": 0.6981373429298401,
+      "learning_rate": 0.00023035448766296732,
+      "loss": 3.2747,
+      "step": 57300
+    },
+    {
+      "epoch": 6.1726401894306315,
+      "grad_norm": 0.707578182220459,
+      "learning_rate": 0.00023003124663290592,
+      "loss": 3.2662,
+      "step": 57350
+    },
+    {
+      "epoch": 6.17802174147024,
+      "grad_norm": 0.7146908044815063,
+      "learning_rate": 0.0002297080056028445,
+      "loss": 3.2686,
+      "step": 57400
+    },
+    {
+      "epoch": 6.183403293509849,
+      "grad_norm": 0.7029908895492554,
+      "learning_rate": 0.00022938476457278308,
+      "loss": 3.263,
+      "step": 57450
+    },
+    {
+      "epoch": 6.188784845549456,
+      "grad_norm": 0.8150429725646973,
+      "learning_rate": 0.00022906152354272168,
+      "loss": 3.2707,
+      "step": 57500
+    },
+    {
+      "epoch": 6.194166397589065,
+      "grad_norm": 0.7352506518363953,
+      "learning_rate": 0.00022873828251266024,
+      "loss": 3.2762,
+      "step": 57550
+    },
+    {
+      "epoch": 6.1995479496286725,
+      "grad_norm": 0.6864310503005981,
+      "learning_rate": 0.00022841504148259884,
+      "loss": 3.27,
+      "step": 57600
+    },
+    {
+      "epoch": 6.204929501668281,
+      "grad_norm": 0.737712562084198,
+      "learning_rate": 0.00022809180045253743,
+      "loss": 3.2762,
+      "step": 57650
+    },
+    {
+      "epoch": 6.21031105370789,
+      "grad_norm": 0.7097147107124329,
+      "learning_rate": 0.000227768559422476,
+      "loss": 3.2642,
+      "step": 57700
+    },
+    {
+      "epoch": 6.215692605747497,
+      "grad_norm": 0.6888893842697144,
+      "learning_rate": 0.00022744531839241457,
+      "loss": 3.282,
+      "step": 57750
+    },
+    {
+      "epoch": 6.221074157787106,
+      "grad_norm": 0.739660382270813,
+      "learning_rate": 0.0002271220773623532,
+      "loss": 3.2819,
+      "step": 57800
+    },
+    {
+      "epoch": 6.226455709826714,
+      "grad_norm": 0.7385263442993164,
+      "learning_rate": 0.00022679883633229176,
+      "loss": 3.277,
+      "step": 57850
+    },
+    {
+      "epoch": 6.231837261866322,
+      "grad_norm": 0.751966655254364,
+      "learning_rate": 0.00022647559530223032,
+      "loss": 3.2734,
+      "step": 57900
+    },
+    {
+      "epoch": 6.237218813905931,
+      "grad_norm": 0.7751827239990234,
+      "learning_rate": 0.00022615881909277016,
+      "loss": 3.2823,
+      "step": 57950
+    },
+    {
+      "epoch": 6.242600365945538,
+      "grad_norm": 0.7112472057342529,
+      "learning_rate": 0.00022583557806270875,
+      "loss": 3.2708,
+      "step": 58000
+    },
+    {
+      "epoch": 6.242600365945538,
+      "eval_accuracy": 0.38413565704194036,
+      "eval_loss": 3.384307384490967,
+      "eval_runtime": 183.0119,
+      "eval_samples_per_second": 98.414,
+      "eval_steps_per_second": 6.153,
+      "step": 58000
+    },
+    {
+      "epoch": 6.247981917985147,
+      "grad_norm": 0.7275688648223877,
+      "learning_rate": 0.00022551233703264732,
+      "loss": 3.2724,
+      "step": 58050
+    },
+    {
+      "epoch": 6.253363470024755,
+      "grad_norm": 0.7110071778297424,
+      "learning_rate": 0.0002251890960025859,
+      "loss": 3.2628,
+      "step": 58100
+    },
+    {
+      "epoch": 6.258745022064363,
+      "grad_norm": 0.7214199900627136,
+      "learning_rate": 0.0002248658549725245,
+      "loss": 3.2907,
+      "step": 58150
+    },
+    {
+      "epoch": 6.264126574103972,
+      "grad_norm": 0.6868590116500854,
+      "learning_rate": 0.00022454261394246308,
+      "loss": 3.2706,
+      "step": 58200
+    },
+    {
+      "epoch": 6.26950812614358,
+      "grad_norm": 0.7192474007606506,
+      "learning_rate": 0.00022421937291240164,
+      "loss": 3.2852,
+      "step": 58250
+    },
+    {
+      "epoch": 6.274889678183188,
+      "grad_norm": 0.6799070239067078,
+      "learning_rate": 0.00022389613188234027,
+      "loss": 3.2645,
+      "step": 58300
+    },
+    {
+      "epoch": 6.280271230222796,
+      "grad_norm": 0.7178609371185303,
+      "learning_rate": 0.00022357289085227883,
+      "loss": 3.2952,
+      "step": 58350
+    },
+    {
+      "epoch": 6.285652782262405,
+      "grad_norm": 0.7094721794128418,
+      "learning_rate": 0.0002232496498222174,
+      "loss": 3.2788,
+      "step": 58400
+    },
+    {
+      "epoch": 6.2910343343020125,
+      "grad_norm": 0.7339088320732117,
+      "learning_rate": 0.000222926408792156,
+      "loss": 3.2805,
+      "step": 58450
+    },
+    {
+      "epoch": 6.296415886341621,
+      "grad_norm": 0.7479764223098755,
+      "learning_rate": 0.0002226031677620946,
+      "loss": 3.2833,
+      "step": 58500
+    },
+    {
+      "epoch": 6.301797438381229,
+      "grad_norm": 0.7498093843460083,
+      "learning_rate": 0.00022227992673203316,
+      "loss": 3.2691,
+      "step": 58550
+    },
+    {
+      "epoch": 6.307178990420837,
+      "grad_norm": 0.6673323512077332,
+      "learning_rate": 0.00022195668570197175,
+      "loss": 3.2946,
+      "step": 58600
+    },
+    {
+      "epoch": 6.312560542460446,
+      "grad_norm": 0.7104136347770691,
+      "learning_rate": 0.00022163344467191032,
+      "loss": 3.2845,
+      "step": 58650
+    },
+    {
+      "epoch": 6.3179420945000535,
+      "grad_norm": 0.6776658296585083,
+      "learning_rate": 0.00022131020364184891,
+      "loss": 3.2848,
+      "step": 58700
+    },
+    {
+      "epoch": 6.323323646539662,
+      "grad_norm": 0.8017224073410034,
+      "learning_rate": 0.0002209869626117875,
+      "loss": 3.2985,
+      "step": 58750
+    },
+    {
+      "epoch": 6.328705198579271,
+      "grad_norm": 0.7754116654396057,
+      "learning_rate": 0.00022066372158172608,
+      "loss": 3.2727,
+      "step": 58800
+    },
+    {
+      "epoch": 6.334086750618878,
+      "grad_norm": 0.6976533532142639,
+      "learning_rate": 0.00022034048055166464,
+      "loss": 3.2717,
+      "step": 58850
+    },
+    {
+      "epoch": 6.339468302658487,
+      "grad_norm": 0.7677433490753174,
+      "learning_rate": 0.00022001723952160327,
+      "loss": 3.27,
+      "step": 58900
+    },
+    {
+      "epoch": 6.344849854698095,
+      "grad_norm": 0.7851221561431885,
+      "learning_rate": 0.00021969399849154183,
+      "loss": 3.2634,
+      "step": 58950
+    },
+    {
+      "epoch": 6.350231406737703,
+      "grad_norm": 0.7342744469642639,
+      "learning_rate": 0.00021937075746148043,
+      "loss": 3.2901,
+      "step": 59000
+    },
+    {
+      "epoch": 6.350231406737703,
+      "eval_accuracy": 0.3843829510146061,
+      "eval_loss": 3.38128662109375,
+      "eval_runtime": 183.0034,
+      "eval_samples_per_second": 98.419,
+      "eval_steps_per_second": 6.153,
+      "step": 59000
+    },
+    {
+      "epoch": 6.355612958777312,
+      "grad_norm": 0.7511043548583984,
+      "learning_rate": 0.00021904751643141902,
+      "loss": 3.2897,
+      "step": 59050
+    },
+    {
+      "epoch": 6.360994510816919,
+      "grad_norm": 0.7121841907501221,
+      "learning_rate": 0.0002187242754013576,
+      "loss": 3.2602,
+      "step": 59100
+    },
+    {
+      "epoch": 6.366376062856528,
+      "grad_norm": 0.7471939325332642,
+      "learning_rate": 0.00021840103437129619,
+      "loss": 3.2703,
+      "step": 59150
+    },
+    {
+      "epoch": 6.371757614896136,
+      "grad_norm": 0.7568403482437134,
+      "learning_rate": 0.00021807779334123475,
+      "loss": 3.2817,
+      "step": 59200
+    },
+    {
+      "epoch": 6.377139166935744,
+      "grad_norm": 0.7506998181343079,
+      "learning_rate": 0.00021775455231117335,
+      "loss": 3.2724,
+      "step": 59250
+    },
+    {
+      "epoch": 6.382520718975353,
+      "grad_norm": 0.7650529146194458,
+      "learning_rate": 0.00021743131128111194,
+      "loss": 3.2874,
+      "step": 59300
+    },
+    {
+      "epoch": 6.387902271014961,
+      "grad_norm": 0.712727963924408,
+      "learning_rate": 0.0002171080702510505,
+      "loss": 3.2699,
+      "step": 59350
+    },
+    {
+      "epoch": 6.393283823054569,
+      "grad_norm": 0.7634322643280029,
+      "learning_rate": 0.00021678482922098908,
+      "loss": 3.2916,
+      "step": 59400
+    },
+    {
+      "epoch": 6.398665375094177,
+      "grad_norm": 0.7388198375701904,
+      "learning_rate": 0.0002164615881909277,
+      "loss": 3.2963,
+      "step": 59450
+    },
+    {
+      "epoch": 6.404046927133785,
+      "grad_norm": 0.7384008169174194,
+      "learning_rate": 0.00021613834716086627,
+      "loss": 3.2789,
+      "step": 59500
+    },
+    {
+      "epoch": 6.4094284791733935,
+      "grad_norm": 0.7477920651435852,
+      "learning_rate": 0.00021581510613080483,
+      "loss": 3.2972,
+      "step": 59550
+    },
+    {
+      "epoch": 6.414810031213002,
+      "grad_norm": 0.7716207504272461,
+      "learning_rate": 0.00021549186510074346,
+      "loss": 3.2811,
+      "step": 59600
+    },
+    {
+      "epoch": 6.42019158325261,
+      "grad_norm": 0.7266528010368347,
+      "learning_rate": 0.00021516862407068202,
+      "loss": 3.2955,
+      "step": 59650
+    },
+    {
+      "epoch": 6.425573135292218,
+      "grad_norm": 0.7093421220779419,
+      "learning_rate": 0.0002148453830406206,
+      "loss": 3.2774,
+      "step": 59700
+    },
+    {
+      "epoch": 6.430954687331827,
+      "grad_norm": 0.6861085891723633,
+      "learning_rate": 0.00021452214201055919,
+      "loss": 3.2935,
+      "step": 59750
+    },
+    {
+      "epoch": 6.4363362393714345,
+      "grad_norm": 0.6963001489639282,
+      "learning_rate": 0.00021419890098049778,
+      "loss": 3.2823,
+      "step": 59800
+    },
+    {
+      "epoch": 6.441717791411043,
+      "grad_norm": 0.7377711534500122,
+      "learning_rate": 0.00021387565995043638,
+      "loss": 3.2938,
+      "step": 59850
+    },
+    {
+      "epoch": 6.447099343450651,
+      "grad_norm": 0.7165238857269287,
+      "learning_rate": 0.00021355241892037494,
+      "loss": 3.296,
+      "step": 59900
+    },
+    {
+      "epoch": 6.452480895490259,
+      "grad_norm": 0.7209922671318054,
+      "learning_rate": 0.0002132291778903135,
+      "loss": 3.2764,
+      "step": 59950
+    },
+    {
+      "epoch": 6.457862447529868,
+      "grad_norm": 0.6959686875343323,
+      "learning_rate": 0.00021290593686025213,
+      "loss": 3.3013,
+      "step": 60000
+    },
+    {
+      "epoch": 6.457862447529868,
+      "eval_accuracy": 0.38479148587278855,
+      "eval_loss": 3.3754682540893555,
+      "eval_runtime": 182.7058,
+      "eval_samples_per_second": 98.579,
+      "eval_steps_per_second": 6.163,
+      "step": 60000
+    },
+    {
+      "epoch": 6.4632439995694755,
+      "grad_norm": 0.8473927974700928,
+      "learning_rate": 0.0002125826958301907,
+      "loss": 3.2835,
+      "step": 60050
+    },
+    {
+      "epoch": 6.468625551609084,
+      "grad_norm": 0.7251399755477905,
+      "learning_rate": 0.00021225945480012927,
+      "loss": 3.3025,
+      "step": 60100
+    },
+    {
+      "epoch": 6.474007103648693,
+      "grad_norm": 0.767999529838562,
+      "learning_rate": 0.00021193621377006786,
+      "loss": 3.2906,
+      "step": 60150
+    },
+    {
+      "epoch": 6.4793886556883,
+      "grad_norm": 0.7236109972000122,
+      "learning_rate": 0.00021161297274000646,
+      "loss": 3.2944,
+      "step": 60200
+    },
+    {
+      "epoch": 6.484770207727909,
+      "grad_norm": 0.7092922925949097,
+      "learning_rate": 0.00021128973170994502,
+      "loss": 3.2888,
+      "step": 60250
+    },
+    {
+      "epoch": 6.490151759767517,
+      "grad_norm": 0.7969743609428406,
+      "learning_rate": 0.00021096649067988362,
+      "loss": 3.2951,
+      "step": 60300
+    },
+    {
+      "epoch": 6.495533311807125,
+      "grad_norm": 0.724816620349884,
+      "learning_rate": 0.00021064324964982219,
+      "loss": 3.2937,
+      "step": 60350
+    },
+    {
+      "epoch": 6.500914863846734,
+      "grad_norm": 0.792242169380188,
+      "learning_rate": 0.00021032000861976078,
+      "loss": 3.2896,
+      "step": 60400
+    },
+    {
+      "epoch": 6.506296415886341,
+      "grad_norm": 0.7577232718467712,
+      "learning_rate": 0.00020999676758969938,
+      "loss": 3.2881,
+      "step": 60450
+    },
+    {
+      "epoch": 6.51167796792595,
+      "grad_norm": 0.7519006729125977,
+      "learning_rate": 0.00020967352655963794,
+      "loss": 3.2775,
+      "step": 60500
+    },
+    {
+      "epoch": 6.517059519965558,
+      "grad_norm": 0.7319377660751343,
+      "learning_rate": 0.0002093502855295765,
+      "loss": 3.2871,
+      "step": 60550
+    },
+    {
+      "epoch": 6.522441072005166,
+      "grad_norm": 0.7363983988761902,
+      "learning_rate": 0.00020902704449951513,
+      "loss": 3.2811,
+      "step": 60600
+    },
+    {
+      "epoch": 6.5278226240447745,
+      "grad_norm": 0.7572370171546936,
+      "learning_rate": 0.0002087038034694537,
+      "loss": 3.2906,
+      "step": 60650
+    },
+    {
+      "epoch": 6.533204176084383,
+      "grad_norm": 0.7479851841926575,
+      "learning_rate": 0.00020838056243939227,
+      "loss": 3.2969,
+      "step": 60700
+    },
+    {
+      "epoch": 6.538585728123991,
+      "grad_norm": 0.754562497138977,
+      "learning_rate": 0.0002080573214093309,
+      "loss": 3.2942,
+      "step": 60750
+    },
+    {
+      "epoch": 6.543967280163599,
+      "grad_norm": 0.7065774202346802,
+      "learning_rate": 0.00020773408037926946,
+      "loss": 3.2863,
+      "step": 60800
+    },
+    {
+      "epoch": 6.549348832203208,
+      "grad_norm": 0.7242980599403381,
+      "learning_rate": 0.00020741083934920805,
+      "loss": 3.2771,
+      "step": 60850
+    },
+    {
+      "epoch": 6.5547303842428155,
+      "grad_norm": 0.7278899550437927,
+      "learning_rate": 0.00020708759831914662,
+      "loss": 3.2715,
+      "step": 60900
+    },
+    {
+      "epoch": 6.560111936282424,
+      "grad_norm": 0.738157331943512,
+      "learning_rate": 0.00020676435728908521,
+      "loss": 3.2806,
+      "step": 60950
+    },
+    {
+      "epoch": 6.565493488322032,
+      "grad_norm": 0.7449216842651367,
+      "learning_rate": 0.0002064411162590238,
+      "loss": 3.2954,
+      "step": 61000
+    },
+    {
+      "epoch": 6.565493488322032,
+      "eval_accuracy": 0.3854739346611513,
+      "eval_loss": 3.3736610412597656,
+      "eval_runtime": 182.6023,
+      "eval_samples_per_second": 98.635,
+      "eval_steps_per_second": 6.166,
+      "step": 61000
+    },
+    {
+      "epoch": 6.57087504036164,
+      "grad_norm": 0.7469400763511658,
+      "learning_rate": 0.00020611787522896238,
+      "loss": 3.2986,
+      "step": 61050
+    },
+    {
+      "epoch": 6.576256592401249,
+      "grad_norm": 0.73382169008255,
+      "learning_rate": 0.00020579463419890094,
+      "loss": 3.2728,
+      "step": 61100
+    },
+    {
+      "epoch": 6.5816381444408565,
+      "grad_norm": 0.7447705864906311,
+      "learning_rate": 0.00020547139316883957,
+      "loss": 3.298,
+      "step": 61150
+    },
+    {
+      "epoch": 6.587019696480465,
+      "grad_norm": 0.7730774283409119,
+      "learning_rate": 0.00020514815213877813,
+      "loss": 3.2885,
+      "step": 61200
+    },
+    {
+      "epoch": 6.592401248520073,
+      "grad_norm": 0.764479398727417,
+      "learning_rate": 0.0002048249111087167,
+      "loss": 3.2927,
+      "step": 61250
+    },
+    {
+      "epoch": 6.597782800559681,
+      "grad_norm": 0.7595162987709045,
+      "learning_rate": 0.00020450167007865532,
+      "loss": 3.2942,
+      "step": 61300
+    },
+    {
+      "epoch": 6.60316435259929,
+      "grad_norm": 0.7465935945510864,
+      "learning_rate": 0.0002041784290485939,
+      "loss": 3.2903,
+      "step": 61350
+    },
+    {
+      "epoch": 6.608545904638898,
+      "grad_norm": 0.7768982648849487,
+      "learning_rate": 0.00020385518801853246,
+      "loss": 3.2867,
+      "step": 61400
+    },
+    {
+      "epoch": 6.613927456678506,
+      "grad_norm": 0.7525243759155273,
+      "learning_rate": 0.00020353194698847105,
+      "loss": 3.2923,
+      "step": 61450
+    },
+    {
+      "epoch": 6.619309008718115,
+      "grad_norm": 0.7230768799781799,
+      "learning_rate": 0.00020320870595840965,
+      "loss": 3.287,
+      "step": 61500
+    },
+    {
+      "epoch": 6.624690560757722,
+      "grad_norm": 0.7975799441337585,
+      "learning_rate": 0.00020288546492834821,
+      "loss": 3.3008,
+      "step": 61550
+    },
+    {
+      "epoch": 6.630072112797331,
+      "grad_norm": 0.7720596194267273,
+      "learning_rate": 0.0002025622238982868,
+      "loss": 3.2733,
+      "step": 61600
+    },
+    {
+      "epoch": 6.635453664836939,
+      "grad_norm": 0.7760968804359436,
+      "learning_rate": 0.00020223898286822538,
+      "loss": 3.2823,
+      "step": 61650
+    },
+    {
+      "epoch": 6.640835216876547,
+      "grad_norm": 0.7323305010795593,
+      "learning_rate": 0.00020191574183816397,
+      "loss": 3.2987,
+      "step": 61700
+    },
+    {
+      "epoch": 6.6462167689161555,
+      "grad_norm": 0.7435678839683533,
+      "learning_rate": 0.00020159250080810257,
+      "loss": 3.2919,
+      "step": 61750
+    },
+    {
+      "epoch": 6.651598320955763,
+      "grad_norm": 0.779816210269928,
+      "learning_rate": 0.00020126925977804113,
+      "loss": 3.3058,
+      "step": 61800
+    },
+    {
+      "epoch": 6.656979872995372,
+      "grad_norm": 0.7927149534225464,
+      "learning_rate": 0.00020094601874797976,
+      "loss": 3.2841,
+      "step": 61850
+    },
+    {
+      "epoch": 6.66236142503498,
+      "grad_norm": 0.7235358953475952,
+      "learning_rate": 0.00020062277771791832,
+      "loss": 3.3052,
+      "step": 61900
+    },
+    {
+      "epoch": 6.667742977074588,
+      "grad_norm": 0.7375259399414062,
+      "learning_rate": 0.00020030600150845813,
+      "loss": 3.2933,
+      "step": 61950
+    },
+    {
+      "epoch": 6.6731245291141965,
+      "grad_norm": 0.7493704557418823,
+      "learning_rate": 0.0001999827604783967,
+      "loss": 3.2993,
+      "step": 62000
+    },
+    {
+      "epoch": 6.6731245291141965,
+      "eval_accuracy": 0.3856817876355404,
+      "eval_loss": 3.367237091064453,
+      "eval_runtime": 182.8132,
+      "eval_samples_per_second": 98.521,
+      "eval_steps_per_second": 6.159,
+      "step": 62000
+    },
+    {
+      "epoch": 6.678506081153805,
+      "grad_norm": 0.7092652916908264,
+      "learning_rate": 0.0001996595194483353,
+      "loss": 3.2846,
+      "step": 62050
+    },
+    {
+      "epoch": 6.683887633193413,
+      "grad_norm": 0.712834894657135,
+      "learning_rate": 0.00019933627841827389,
+      "loss": 3.2926,
+      "step": 62100
+    },
+    {
+      "epoch": 6.689269185233021,
+      "grad_norm": 0.7052910923957825,
+      "learning_rate": 0.00019901303738821245,
+      "loss": 3.2788,
+      "step": 62150
+    },
+    {
+      "epoch": 6.69465073727263,
+      "grad_norm": 0.727817177772522,
+      "learning_rate": 0.00019868979635815102,
+      "loss": 3.2872,
+      "step": 62200
+    },
+    {
+      "epoch": 6.7000322893122375,
+      "grad_norm": 0.7619044780731201,
+      "learning_rate": 0.00019836655532808964,
+      "loss": 3.2951,
+      "step": 62250
+    },
+    {
+      "epoch": 6.705413841351846,
+      "grad_norm": 0.742910623550415,
+      "learning_rate": 0.0001980433142980282,
+      "loss": 3.2846,
+      "step": 62300
+    },
+    {
+      "epoch": 6.710795393391454,
+      "grad_norm": 0.8113905191421509,
+      "learning_rate": 0.00019772007326796678,
+      "loss": 3.2751,
+      "step": 62350
+    },
+    {
+      "epoch": 6.716176945431062,
+      "grad_norm": 0.8383527994155884,
+      "learning_rate": 0.0001973968322379054,
+      "loss": 3.2939,
+      "step": 62400
+    },
+    {
+      "epoch": 6.721558497470671,
+      "grad_norm": 0.7623720169067383,
+      "learning_rate": 0.00019707359120784397,
+      "loss": 3.3112,
+      "step": 62450
+    },
+    {
+      "epoch": 6.7269400495102785,
+      "grad_norm": 0.7660823464393616,
+      "learning_rate": 0.00019675035017778253,
+      "loss": 3.2797,
+      "step": 62500
+    },
+    {
+      "epoch": 6.732321601549887,
+      "grad_norm": 0.7665824890136719,
+      "learning_rate": 0.00019642710914772113,
+      "loss": 3.2917,
+      "step": 62550
+    },
+    {
+      "epoch": 6.737703153589496,
+      "grad_norm": 0.754959762096405,
+      "learning_rate": 0.00019610386811765972,
+      "loss": 3.3053,
+      "step": 62600
+    },
+    {
+      "epoch": 6.743084705629103,
+      "grad_norm": 0.7177004218101501,
+      "learning_rate": 0.00019578062708759832,
+      "loss": 3.285,
+      "step": 62650
+    },
+    {
+      "epoch": 6.748466257668712,
+      "grad_norm": 0.741148054599762,
+      "learning_rate": 0.00019545738605753689,
+      "loss": 3.284,
+      "step": 62700
+    },
+    {
+      "epoch": 6.75384780970832,
+      "grad_norm": 0.7509447932243347,
+      "learning_rate": 0.00019513414502747545,
+      "loss": 3.2928,
+      "step": 62750
+    },
+    {
+      "epoch": 6.759229361747928,
+      "grad_norm": 0.7940971255302429,
+      "learning_rate": 0.00019481090399741408,
+      "loss": 3.2915,
+      "step": 62800
+    },
+    {
+      "epoch": 6.7646109137875365,
+      "grad_norm": 0.7626696228981018,
+      "learning_rate": 0.00019448766296735264,
+      "loss": 3.2999,
+      "step": 62850
+    },
+    {
+      "epoch": 6.769992465827144,
+      "grad_norm": 0.7458931803703308,
+      "learning_rate": 0.0001941644219372912,
+      "loss": 3.2895,
+      "step": 62900
+    },
+    {
+      "epoch": 6.775374017866753,
+      "grad_norm": 0.7777038812637329,
+      "learning_rate": 0.00019384118090722983,
+      "loss": 3.2885,
+      "step": 62950
+    },
+    {
+      "epoch": 6.780755569906361,
+      "grad_norm": 0.7893502116203308,
+      "learning_rate": 0.0001935179398771684,
+      "loss": 3.2894,
+      "step": 63000
+    },
+    {
+      "epoch": 6.780755569906361,
+      "eval_accuracy": 0.3861680093085102,
+      "eval_loss": 3.364060163497925,
+      "eval_runtime": 182.8147,
+      "eval_samples_per_second": 98.521,
+      "eval_steps_per_second": 6.159,
+      "step": 63000
+    },
+    {
+      "epoch": 6.786137121945969,
+      "grad_norm": 0.732779324054718,
+      "learning_rate": 0.00019319469884710697,
+      "loss": 3.289,
+      "step": 63050
+    },
+    {
+      "epoch": 6.7915186739855775,
+      "grad_norm": 0.7249733209609985,
+      "learning_rate": 0.00019287145781704556,
+      "loss": 3.2795,
+      "step": 63100
+    },
+    {
+      "epoch": 6.796900226025185,
+      "grad_norm": 0.7652485966682434,
+      "learning_rate": 0.00019254821678698416,
+      "loss": 3.2788,
+      "step": 63150
+    },
+    {
+      "epoch": 6.802281778064794,
+      "grad_norm": 0.7800796031951904,
+      "learning_rate": 0.00019222497575692272,
+      "loss": 3.282,
+      "step": 63200
+    },
+    {
+      "epoch": 6.807663330104402,
+      "grad_norm": 0.7617146968841553,
+      "learning_rate": 0.00019190173472686132,
+      "loss": 3.2922,
+      "step": 63250
+    },
+    {
+      "epoch": 6.813044882144011,
+      "grad_norm": 0.753646969795227,
+      "learning_rate": 0.0001915784936967999,
+      "loss": 3.2944,
+      "step": 63300
+    },
+    {
+      "epoch": 6.8184264341836185,
+      "grad_norm": 0.7656055688858032,
+      "learning_rate": 0.00019125525266673845,
+      "loss": 3.2996,
+      "step": 63350
+    },
+    {
+      "epoch": 6.823807986223227,
+      "grad_norm": 0.7957121133804321,
+      "learning_rate": 0.00019093201163667708,
+      "loss": 3.2804,
+      "step": 63400
+    },
+    {
+      "epoch": 6.829189538262835,
+      "grad_norm": 0.7702323794364929,
+      "learning_rate": 0.00019060877060661564,
+      "loss": 3.2949,
+      "step": 63450
+    },
+    {
+      "epoch": 6.834571090302443,
+      "grad_norm": 0.8194876313209534,
+      "learning_rate": 0.0001902855295765542,
+      "loss": 3.2796,
+      "step": 63500
+    },
+    {
+      "epoch": 6.839952642342052,
+      "grad_norm": 0.75143963098526,
+      "learning_rate": 0.00018996228854649283,
+      "loss": 3.3038,
+      "step": 63550
+    },
+    {
+      "epoch": 6.8453341943816595,
+      "grad_norm": 0.7525258660316467,
+      "learning_rate": 0.0001896390475164314,
+      "loss": 3.2956,
+      "step": 63600
+    },
+    {
+      "epoch": 6.850715746421268,
+      "grad_norm": 0.7692186236381531,
+      "learning_rate": 0.00018931580648637,
+      "loss": 3.2958,
+      "step": 63650
+    },
+    {
+      "epoch": 6.856097298460876,
+      "grad_norm": 0.7735649943351746,
+      "learning_rate": 0.00018899256545630856,
+      "loss": 3.2939,
+      "step": 63700
+    },
+    {
+      "epoch": 6.861478850500484,
+      "grad_norm": 0.7386011481285095,
+      "learning_rate": 0.00018866932442624716,
+      "loss": 3.3071,
+      "step": 63750
+    },
+    {
+      "epoch": 6.866860402540093,
+      "grad_norm": 0.7658208608627319,
+      "learning_rate": 0.00018834608339618575,
+      "loss": 3.2894,
+      "step": 63800
+    },
+    {
+      "epoch": 6.8722419545797,
+      "grad_norm": 0.7646046280860901,
+      "learning_rate": 0.00018802284236612432,
+      "loss": 3.3067,
+      "step": 63850
+    },
+    {
+      "epoch": 6.877623506619309,
+      "grad_norm": 0.732725203037262,
+      "learning_rate": 0.0001876996013360629,
+      "loss": 3.2885,
+      "step": 63900
+    },
+    {
+      "epoch": 6.8830050586589175,
+      "grad_norm": 0.6860552430152893,
+      "learning_rate": 0.00018738282512660272,
+      "loss": 3.3071,
+      "step": 63950
+    },
+    {
+      "epoch": 6.888386610698525,
+      "grad_norm": 0.7384079694747925,
+      "learning_rate": 0.0001870595840965413,
+      "loss": 3.294,
+      "step": 64000
+    },
+    {
+      "epoch": 6.888386610698525,
+      "eval_accuracy": 0.38654601270521677,
+      "eval_loss": 3.361266851425171,
+      "eval_runtime": 182.8473,
+      "eval_samples_per_second": 98.503,
+      "eval_steps_per_second": 6.158,
+      "step": 64000
+    },
+    {
+      "epoch": 6.893768162738134,
+      "grad_norm": 0.7515199184417725,
+      "learning_rate": 0.0001867363430664799,
+      "loss": 3.2797,
+      "step": 64050
+    },
+    {
+      "epoch": 6.899149714777742,
+      "grad_norm": 0.7795135378837585,
+      "learning_rate": 0.00018641310203641848,
+      "loss": 3.2799,
+      "step": 64100
+    },
+    {
+      "epoch": 6.90453126681735,
+      "grad_norm": 0.7016361355781555,
+      "learning_rate": 0.00018608986100635705,
+      "loss": 3.2991,
+      "step": 64150
+    },
+    {
+      "epoch": 6.9099128188569585,
+      "grad_norm": 0.7909216284751892,
+      "learning_rate": 0.00018576661997629564,
+      "loss": 3.2868,
+      "step": 64200
+    },
+    {
+      "epoch": 6.915294370896566,
+      "grad_norm": 0.7275353074073792,
+      "learning_rate": 0.00018544337894623423,
+      "loss": 3.2848,
+      "step": 64250
+    },
+    {
+      "epoch": 6.920675922936175,
+      "grad_norm": 0.7585761547088623,
+      "learning_rate": 0.0001851201379161728,
+      "loss": 3.2827,
+      "step": 64300
+    },
+    {
+      "epoch": 6.926057474975783,
+      "grad_norm": 0.7263973355293274,
+      "learning_rate": 0.0001847968968861114,
+      "loss": 3.296,
+      "step": 64350
+    },
+    {
+      "epoch": 6.931439027015391,
+      "grad_norm": 0.751746654510498,
+      "learning_rate": 0.00018447365585604996,
+      "loss": 3.2963,
+      "step": 64400
+    },
+    {
+      "epoch": 6.9368205790549995,
+      "grad_norm": 0.748279333114624,
+      "learning_rate": 0.00018415041482598859,
+      "loss": 3.2842,
+      "step": 64450
+    },
+    {
+      "epoch": 6.942202131094608,
+      "grad_norm": 0.7945259213447571,
+      "learning_rate": 0.00018382717379592715,
+      "loss": 3.3071,
+      "step": 64500
+    },
+    {
+      "epoch": 6.947583683134216,
+      "grad_norm": 0.74016934633255,
+      "learning_rate": 0.00018350393276586572,
+      "loss": 3.2925,
+      "step": 64550
+    },
+    {
+      "epoch": 6.952965235173824,
+      "grad_norm": 0.7379736304283142,
+      "learning_rate": 0.00018318069173580434,
+      "loss": 3.2959,
+      "step": 64600
+    },
+    {
+      "epoch": 6.958346787213433,
+      "grad_norm": 0.7810225486755371,
+      "learning_rate": 0.0001828574507057429,
+      "loss": 3.3001,
+      "step": 64650
+    },
+    {
+      "epoch": 6.9637283392530405,
+      "grad_norm": 0.7810648083686829,
+      "learning_rate": 0.00018253420967568148,
+      "loss": 3.2961,
+      "step": 64700
+    },
+    {
+      "epoch": 6.969109891292649,
+      "grad_norm": 0.740482747554779,
+      "learning_rate": 0.00018221096864562007,
+      "loss": 3.3029,
+      "step": 64750
+    },
+    {
+      "epoch": 6.974491443332257,
+      "grad_norm": 0.7730101943016052,
+      "learning_rate": 0.00018188772761555867,
+      "loss": 3.2891,
+      "step": 64800
+    },
+    {
+      "epoch": 6.979872995371865,
+      "grad_norm": 0.8048725128173828,
+      "learning_rate": 0.00018156448658549723,
+      "loss": 3.2735,
+      "step": 64850
+    },
+    {
+      "epoch": 6.985254547411474,
+      "grad_norm": 0.76474529504776,
+      "learning_rate": 0.00018124124555543583,
+      "loss": 3.2922,
+      "step": 64900
+    },
+    {
+      "epoch": 6.990636099451081,
+      "grad_norm": 0.7955503463745117,
+      "learning_rate": 0.0001809180045253744,
+      "loss": 3.2788,
+      "step": 64950
+    },
+    {
+      "epoch": 6.99601765149069,
+      "grad_norm": 0.7674031257629395,
+      "learning_rate": 0.00018059476349531296,
+      "loss": 3.2997,
+      "step": 65000
+    },
+    {
+      "epoch": 6.99601765149069,
+      "eval_accuracy": 0.3869485716545694,
+      "eval_loss": 3.357651710510254,
+      "eval_runtime": 182.7956,
+      "eval_samples_per_second": 98.531,
+      "eval_steps_per_second": 6.16,
+      "step": 65000
+    },
+    {
+      "epoch": 7.0013992035302985,
+      "grad_norm": 0.7507967948913574,
+      "learning_rate": 0.00018027152246525159,
+      "loss": 3.2669,
+      "step": 65050
+    },
+    {
+      "epoch": 7.006780755569906,
+      "grad_norm": 0.7749801278114319,
+      "learning_rate": 0.00017994828143519015,
+      "loss": 3.2086,
+      "step": 65100
+    },
+    {
+      "epoch": 7.012162307609515,
+      "grad_norm": 0.751576840877533,
+      "learning_rate": 0.00017962504040512872,
+      "loss": 3.1952,
+      "step": 65150
+    },
+    {
+      "epoch": 7.017543859649122,
+      "grad_norm": 0.719531774520874,
+      "learning_rate": 0.00017930179937506734,
+      "loss": 3.2096,
+      "step": 65200
+    },
+    {
+      "epoch": 7.022925411688731,
+      "grad_norm": 0.7776346802711487,
+      "learning_rate": 0.0001789785583450059,
+      "loss": 3.1961,
+      "step": 65250
+    },
+    {
+      "epoch": 7.0283069637283395,
+      "grad_norm": 0.7524562478065491,
+      "learning_rate": 0.00017865531731494448,
+      "loss": 3.212,
+      "step": 65300
+    },
+    {
+      "epoch": 7.033688515767947,
+      "grad_norm": 0.7441275119781494,
+      "learning_rate": 0.00017833207628488307,
+      "loss": 3.2052,
+      "step": 65350
+    },
+    {
+      "epoch": 7.039070067807556,
+      "grad_norm": 0.7578763365745544,
+      "learning_rate": 0.00017800883525482167,
+      "loss": 3.2185,
+      "step": 65400
+    },
+    {
+      "epoch": 7.044451619847164,
+      "grad_norm": 0.7912497520446777,
+      "learning_rate": 0.00017768559422476026,
+      "loss": 3.2265,
+      "step": 65450
+    },
+    {
+      "epoch": 7.049833171886772,
+      "grad_norm": 0.7631890177726746,
+      "learning_rate": 0.00017736235319469883,
+      "loss": 3.2046,
+      "step": 65500
+    },
+    {
+      "epoch": 7.0552147239263805,
+      "grad_norm": 0.7851689457893372,
+      "learning_rate": 0.0001770391121646374,
+      "loss": 3.2074,
+      "step": 65550
+    },
+    {
+      "epoch": 7.060596275965988,
+      "grad_norm": 0.734584629535675,
+      "learning_rate": 0.00017671587113457602,
+      "loss": 3.2278,
+      "step": 65600
+    },
+    {
+      "epoch": 7.065977828005597,
+      "grad_norm": 0.7264954447746277,
+      "learning_rate": 0.0001763926301045146,
+      "loss": 3.2161,
+      "step": 65650
+    },
+    {
+      "epoch": 7.071359380045205,
+      "grad_norm": 0.8186044692993164,
+      "learning_rate": 0.00017606938907445315,
+      "loss": 3.1974,
+      "step": 65700
+    },
+    {
+      "epoch": 7.076740932084813,
+      "grad_norm": 0.7647484540939331,
+      "learning_rate": 0.00017574614804439178,
+      "loss": 3.1959,
+      "step": 65750
+    },
+    {
+      "epoch": 7.0821224841244215,
+      "grad_norm": 0.7596768736839294,
+      "learning_rate": 0.00017542290701433034,
+      "loss": 3.216,
+      "step": 65800
+    },
+    {
+      "epoch": 7.08750403616403,
+      "grad_norm": 0.7448734641075134,
+      "learning_rate": 0.0001750996659842689,
+      "loss": 3.2196,
+      "step": 65850
+    },
+    {
+      "epoch": 7.092885588203638,
+      "grad_norm": 0.7379791736602783,
+      "learning_rate": 0.0001747764249542075,
+      "loss": 3.2247,
+      "step": 65900
+    },
+    {
+      "epoch": 7.098267140243246,
+      "grad_norm": 0.7869905829429626,
+      "learning_rate": 0.0001744531839241461,
+      "loss": 3.2226,
+      "step": 65950
+    },
+    {
+      "epoch": 7.103648692282855,
+      "grad_norm": 0.7377243638038635,
+      "learning_rate": 0.0001741364077146859,
+      "loss": 3.238,
+      "step": 66000
+    },
+    {
+      "epoch": 7.103648692282855,
+      "eval_accuracy": 0.3870309305435328,
+      "eval_loss": 3.3637986183166504,
+      "eval_runtime": 182.81,
+      "eval_samples_per_second": 98.523,
+      "eval_steps_per_second": 6.159,
+      "step": 66000
+    },
+    {
+      "epoch": 7.109030244322462,
+      "grad_norm": 0.7684816122055054,
+      "learning_rate": 0.00017381316668462447,
+      "loss": 3.2179,
+      "step": 66050
+    },
+    {
+      "epoch": 7.114411796362071,
+      "grad_norm": 0.7641874551773071,
+      "learning_rate": 0.00017348992565456304,
+      "loss": 3.2054,
+      "step": 66100
+    },
+    {
+      "epoch": 7.119793348401679,
+      "grad_norm": 0.7933626174926758,
+      "learning_rate": 0.00017316668462450166,
+      "loss": 3.2072,
+      "step": 66150
+    },
+    {
+      "epoch": 7.125174900441287,
+      "grad_norm": 0.7817862629890442,
+      "learning_rate": 0.00017284344359444023,
+      "loss": 3.2125,
+      "step": 66200
+    },
+    {
+      "epoch": 7.130556452480896,
+      "grad_norm": 0.8065374493598938,
+      "learning_rate": 0.00017252020256437885,
+      "loss": 3.236,
+      "step": 66250
+    },
+    {
+      "epoch": 7.135938004520503,
+      "grad_norm": 0.7458828091621399,
+      "learning_rate": 0.00017219696153431742,
+      "loss": 3.219,
+      "step": 66300
+    },
+    {
+      "epoch": 7.141319556560112,
+      "grad_norm": 0.7864915132522583,
+      "learning_rate": 0.000171873720504256,
+      "loss": 3.2115,
+      "step": 66350
+    },
+    {
+      "epoch": 7.1467011085997205,
+      "grad_norm": 0.7758970260620117,
+      "learning_rate": 0.0001715569442947958,
+      "loss": 3.2207,
+      "step": 66400
+    },
+    {
+      "epoch": 7.152082660639328,
+      "grad_norm": 0.7451347708702087,
+      "learning_rate": 0.0001712337032647344,
+      "loss": 3.2267,
+      "step": 66450
+    },
+    {
+      "epoch": 7.157464212678937,
+      "grad_norm": 0.7944716215133667,
+      "learning_rate": 0.00017091046223467298,
+      "loss": 3.2292,
+      "step": 66500
+    },
+    {
+      "epoch": 7.162845764718545,
+      "grad_norm": 0.7795634269714355,
+      "learning_rate": 0.00017058722120461155,
+      "loss": 3.2236,
+      "step": 66550
+    },
+    {
+      "epoch": 7.168227316758153,
+      "grad_norm": 0.8057452440261841,
+      "learning_rate": 0.00017026398017455012,
+      "loss": 3.2259,
+      "step": 66600
+    },
+    {
+      "epoch": 7.1736088687977615,
+      "grad_norm": 0.7921217679977417,
+      "learning_rate": 0.00016994073914448874,
+      "loss": 3.2398,
+      "step": 66650
+    },
+    {
+      "epoch": 7.178990420837369,
+      "grad_norm": 0.7860040664672852,
+      "learning_rate": 0.0001696174981144273,
+      "loss": 3.2137,
+      "step": 66700
+    },
+    {
+      "epoch": 7.184371972876978,
+      "grad_norm": 0.7447194457054138,
+      "learning_rate": 0.00016929425708436588,
+      "loss": 3.2191,
+      "step": 66750
+    },
+    {
+      "epoch": 7.189753524916586,
+      "grad_norm": 0.757375180721283,
+      "learning_rate": 0.0001689710160543045,
+      "loss": 3.2208,
+      "step": 66800
+    },
+    {
+      "epoch": 7.195135076956194,
+      "grad_norm": 0.7810249924659729,
+      "learning_rate": 0.00016864777502424307,
+      "loss": 3.2391,
+      "step": 66850
+    },
+    {
+      "epoch": 7.2005166289958025,
+      "grad_norm": 0.8149625062942505,
+      "learning_rate": 0.00016832453399418163,
+      "loss": 3.2079,
+      "step": 66900
+    },
+    {
+      "epoch": 7.205898181035411,
+      "grad_norm": 0.7833424210548401,
+      "learning_rate": 0.00016800129296412023,
+      "loss": 3.2331,
+      "step": 66950
+    },
+    {
+      "epoch": 7.211279733075019,
+      "grad_norm": 0.8489099740982056,
+      "learning_rate": 0.00016767805193405882,
+      "loss": 3.2202,
+      "step": 67000
+    },
+    {
+      "epoch": 7.211279733075019,
+      "eval_accuracy": 0.38730658291991815,
+      "eval_loss": 3.3612101078033447,
+      "eval_runtime": 182.5912,
+      "eval_samples_per_second": 98.641,
+      "eval_steps_per_second": 6.167,
+      "step": 67000
+    },
+    {
+      "epoch": 7.216661285114627,
+      "grad_norm": 0.7586920857429504,
+      "learning_rate": 0.0001673548109039974,
+      "loss": 3.2246,
+      "step": 67050
+    },
+    {
+      "epoch": 7.222042837154235,
+      "grad_norm": 0.8181234002113342,
+      "learning_rate": 0.00016703156987393598,
+      "loss": 3.2262,
+      "step": 67100
+    },
+    {
+      "epoch": 7.2274243891938434,
+      "grad_norm": 0.7846294641494751,
+      "learning_rate": 0.00016670832884387455,
+      "loss": 3.2376,
+      "step": 67150
+    },
+    {
+      "epoch": 7.232805941233452,
+      "grad_norm": 0.7629982233047485,
+      "learning_rate": 0.00016638508781381317,
+      "loss": 3.2267,
+      "step": 67200
+    },
+    {
+      "epoch": 7.23818749327306,
+      "grad_norm": 0.8482091426849365,
+      "learning_rate": 0.00016606184678375174,
+      "loss": 3.237,
+      "step": 67250
+    },
+    {
+      "epoch": 7.243569045312668,
+      "grad_norm": 0.8140213489532471,
+      "learning_rate": 0.0001657386057536903,
+      "loss": 3.2355,
+      "step": 67300
+    },
+    {
+      "epoch": 7.248950597352277,
+      "grad_norm": 0.7859810590744019,
+      "learning_rate": 0.00016541536472362893,
+      "loss": 3.2088,
+      "step": 67350
+    },
+    {
+      "epoch": 7.254332149391884,
+      "grad_norm": 0.7926174998283386,
+      "learning_rate": 0.0001650921236935675,
+      "loss": 3.2284,
+      "step": 67400
+    },
+    {
+      "epoch": 7.259713701431493,
+      "grad_norm": 0.8039096593856812,
+      "learning_rate": 0.00016476888266350607,
+      "loss": 3.2463,
+      "step": 67450
+    },
+    {
+      "epoch": 7.265095253471101,
+      "grad_norm": 0.7596452832221985,
+      "learning_rate": 0.00016444564163344466,
+      "loss": 3.2319,
+      "step": 67500
+    },
+    {
+      "epoch": 7.270476805510709,
+      "grad_norm": 0.7452762126922607,
+      "learning_rate": 0.00016412240060338326,
+      "loss": 3.2248,
+      "step": 67550
+    },
+    {
+      "epoch": 7.275858357550318,
+      "grad_norm": 0.7983496189117432,
+      "learning_rate": 0.00016379915957332182,
+      "loss": 3.2319,
+      "step": 67600
+    },
+    {
+      "epoch": 7.281239909589925,
+      "grad_norm": 0.7758620977401733,
+      "learning_rate": 0.00016347591854326042,
+      "loss": 3.2352,
+      "step": 67650
+    },
+    {
+      "epoch": 7.286621461629534,
+      "grad_norm": 0.8212805986404419,
+      "learning_rate": 0.00016315267751319898,
+      "loss": 3.2403,
+      "step": 67700
+    },
+    {
+      "epoch": 7.2920030136691425,
+      "grad_norm": 0.8642499446868896,
+      "learning_rate": 0.00016282943648313755,
+      "loss": 3.2181,
+      "step": 67750
+    },
+    {
+      "epoch": 7.29738456570875,
+      "grad_norm": 0.7983525991439819,
+      "learning_rate": 0.00016250619545307617,
+      "loss": 3.2292,
+      "step": 67800
+    },
+    {
+      "epoch": 7.302766117748359,
+      "grad_norm": 0.7927777171134949,
+      "learning_rate": 0.00016218295442301474,
+      "loss": 3.2408,
+      "step": 67850
+    },
+    {
+      "epoch": 7.308147669787967,
+      "grad_norm": 0.7210763692855835,
+      "learning_rate": 0.0001618597133929533,
+      "loss": 3.2259,
+      "step": 67900
+    },
+    {
+      "epoch": 7.313529221827575,
+      "grad_norm": 0.8408624529838562,
+      "learning_rate": 0.00016153647236289193,
+      "loss": 3.2333,
+      "step": 67950
+    },
+    {
+      "epoch": 7.3189107738671835,
+      "grad_norm": 0.7780632376670837,
+      "learning_rate": 0.0001612132313328305,
+      "loss": 3.2489,
+      "step": 68000
+    },
+    {
+      "epoch": 7.3189107738671835,
+      "eval_accuracy": 0.38737057947084086,
+      "eval_loss": 3.356868028640747,
+      "eval_runtime": 182.8416,
+      "eval_samples_per_second": 98.506,
+      "eval_steps_per_second": 6.158,
+      "step": 68000
+    },
+    {
+      "epoch": 7.324292325906791,
+      "grad_norm": 0.7512384653091431,
+      "learning_rate": 0.0001608899903027691,
+      "loss": 3.2333,
+      "step": 68050
+    },
+    {
+      "epoch": 7.3296738779464,
+      "grad_norm": 0.8020547032356262,
+      "learning_rate": 0.00016056674927270766,
+      "loss": 3.2387,
+      "step": 68100
+    },
+    {
+      "epoch": 7.335055429986008,
+      "grad_norm": 0.7911236882209778,
+      "learning_rate": 0.00016024350824264626,
+      "loss": 3.238,
+      "step": 68150
+    },
+    {
+      "epoch": 7.340436982025616,
+      "grad_norm": 0.8028789758682251,
+      "learning_rate": 0.00015992026721258485,
+      "loss": 3.2412,
+      "step": 68200
+    },
+    {
+      "epoch": 7.3458185340652244,
+      "grad_norm": 0.8485779166221619,
+      "learning_rate": 0.00015959702618252342,
+      "loss": 3.2367,
+      "step": 68250
+    },
+    {
+      "epoch": 7.351200086104833,
+      "grad_norm": 0.7784638404846191,
+      "learning_rate": 0.00015927378515246199,
+      "loss": 3.2297,
+      "step": 68300
+    },
+    {
+      "epoch": 7.356581638144441,
+      "grad_norm": 0.7919899821281433,
+      "learning_rate": 0.0001589505441224006,
+      "loss": 3.2273,
+      "step": 68350
+    },
+    {
+      "epoch": 7.361963190184049,
+      "grad_norm": 0.7827087044715881,
+      "learning_rate": 0.00015862730309233917,
+      "loss": 3.2288,
+      "step": 68400
+    },
+    {
+      "epoch": 7.367344742223658,
+      "grad_norm": 0.7712171077728271,
+      "learning_rate": 0.00015830406206227774,
+      "loss": 3.2382,
+      "step": 68450
+    },
+    {
+      "epoch": 7.372726294263265,
+      "grad_norm": 0.7548463940620422,
+      "learning_rate": 0.00015798082103221636,
+      "loss": 3.2307,
+      "step": 68500
+    },
+    {
+      "epoch": 7.378107846302874,
+      "grad_norm": 0.7684069275856018,
+      "learning_rate": 0.00015765758000215493,
+      "loss": 3.2362,
+      "step": 68550
+    },
+    {
+      "epoch": 7.383489398342482,
+      "grad_norm": 0.7532280087471008,
+      "learning_rate": 0.0001573343389720935,
+      "loss": 3.2386,
+      "step": 68600
+    },
+    {
+      "epoch": 7.38887095038209,
+      "grad_norm": 0.7906218767166138,
+      "learning_rate": 0.0001570110979420321,
+      "loss": 3.2553,
+      "step": 68650
+    },
+    {
+      "epoch": 7.394252502421699,
+      "grad_norm": 0.7797097563743591,
+      "learning_rate": 0.0001566878569119707,
+      "loss": 3.2411,
+      "step": 68700
+    },
+    {
+      "epoch": 7.399634054461306,
+      "grad_norm": 0.8284671902656555,
+      "learning_rate": 0.00015636461588190926,
+      "loss": 3.2456,
+      "step": 68750
+    },
+    {
+      "epoch": 7.405015606500915,
+      "grad_norm": 0.7855088114738464,
+      "learning_rate": 0.00015604137485184785,
+      "loss": 3.2456,
+      "step": 68800
+    },
+    {
+      "epoch": 7.4103971585405235,
+      "grad_norm": 0.8041961193084717,
+      "learning_rate": 0.00015571813382178642,
+      "loss": 3.2236,
+      "step": 68850
+    },
+    {
+      "epoch": 7.415778710580131,
+      "grad_norm": 0.8667538166046143,
+      "learning_rate": 0.000155394892791725,
+      "loss": 3.2492,
+      "step": 68900
+    },
+    {
+      "epoch": 7.42116026261974,
+      "grad_norm": 0.7783315777778625,
+      "learning_rate": 0.0001550716517616636,
+      "loss": 3.2429,
+      "step": 68950
+    },
+    {
+      "epoch": 7.426541814659347,
+      "grad_norm": 0.7982115149497986,
+      "learning_rate": 0.00015474841073160217,
+      "loss": 3.2278,
+      "step": 69000
+    },
+    {
+      "epoch": 7.426541814659347,
+      "eval_accuracy": 0.3878293119631936,
+      "eval_loss": 3.3534276485443115,
+      "eval_runtime": 182.9528,
+      "eval_samples_per_second": 98.446,
+      "eval_steps_per_second": 6.155,
+      "step": 69000
+    },
+    {
+      "epoch": 7.431923366698956,
+      "grad_norm": 0.8143750429153442,
+      "learning_rate": 0.0001544251697015408,
+      "loss": 3.2492,
+      "step": 69050
+    },
+    {
+      "epoch": 7.4373049187385645,
+      "grad_norm": 0.8255060315132141,
+      "learning_rate": 0.00015410192867147936,
+      "loss": 3.2362,
+      "step": 69100
+    },
+    {
+      "epoch": 7.442686470778172,
+      "grad_norm": 0.8084985613822937,
+      "learning_rate": 0.00015377868764141793,
+      "loss": 3.2162,
+      "step": 69150
+    },
+    {
+      "epoch": 7.448068022817781,
+      "grad_norm": 0.8223511576652527,
+      "learning_rate": 0.00015345544661135653,
+      "loss": 3.2509,
+      "step": 69200
+    },
+    {
+      "epoch": 7.453449574857389,
+      "grad_norm": 0.7960752844810486,
+      "learning_rate": 0.00015313220558129512,
+      "loss": 3.2334,
+      "step": 69250
+    },
+    {
+      "epoch": 7.458831126896997,
+      "grad_norm": 0.8062809705734253,
+      "learning_rate": 0.0001528089645512337,
+      "loss": 3.2391,
+      "step": 69300
+    },
+    {
+      "epoch": 7.4642126789366054,
+      "grad_norm": 0.8449177742004395,
+      "learning_rate": 0.00015248572352117228,
+      "loss": 3.2271,
+      "step": 69350
+    },
+    {
+      "epoch": 7.469594230976213,
+      "grad_norm": 0.8337926864624023,
+      "learning_rate": 0.00015216248249111085,
+      "loss": 3.2184,
+      "step": 69400
+    },
+    {
+      "epoch": 7.474975783015822,
+      "grad_norm": 0.7521905899047852,
+      "learning_rate": 0.00015183924146104945,
+      "loss": 3.2245,
+      "step": 69450
+    },
+    {
+      "epoch": 7.48035733505543,
+      "grad_norm": 0.8167929649353027,
+      "learning_rate": 0.00015151600043098804,
+      "loss": 3.2562,
+      "step": 69500
+    },
+    {
+      "epoch": 7.485738887095038,
+      "grad_norm": 0.7685818076133728,
+      "learning_rate": 0.0001511927594009266,
+      "loss": 3.2436,
+      "step": 69550
+    },
+    {
+      "epoch": 7.491120439134646,
+      "grad_norm": 0.7431442141532898,
+      "learning_rate": 0.00015086951837086518,
+      "loss": 3.2319,
+      "step": 69600
+    },
+    {
+      "epoch": 7.496501991174255,
+      "grad_norm": 0.783828854560852,
+      "learning_rate": 0.0001505462773408038,
+      "loss": 3.2341,
+      "step": 69650
+    },
+    {
+      "epoch": 7.501883543213863,
+      "grad_norm": 0.7887234687805176,
+      "learning_rate": 0.00015022303631074236,
+      "loss": 3.2257,
+      "step": 69700
+    },
+    {
+      "epoch": 7.507265095253471,
+      "grad_norm": 0.8269175887107849,
+      "learning_rate": 0.00014989979528068096,
+      "loss": 3.2514,
+      "step": 69750
+    },
+    {
+      "epoch": 7.51264664729308,
+      "grad_norm": 0.8568757772445679,
+      "learning_rate": 0.00014957655425061953,
+      "loss": 3.2363,
+      "step": 69800
+    },
+    {
+      "epoch": 7.518028199332687,
+      "grad_norm": 0.7879490852355957,
+      "learning_rate": 0.00014925331322055812,
+      "loss": 3.2664,
+      "step": 69850
+    },
+    {
+      "epoch": 7.523409751372296,
+      "grad_norm": 0.8363874554634094,
+      "learning_rate": 0.0001489300721904967,
+      "loss": 3.2386,
+      "step": 69900
+    },
+    {
+      "epoch": 7.528791303411904,
+      "grad_norm": 0.7977311015129089,
+      "learning_rate": 0.00014860683116043528,
+      "loss": 3.2544,
+      "step": 69950
+    },
+    {
+      "epoch": 7.534172855451512,
+      "grad_norm": 0.7749576568603516,
+      "learning_rate": 0.00014828359013037385,
+      "loss": 3.2375,
+      "step": 70000
+    },
+    {
+      "epoch": 7.534172855451512,
+      "eval_accuracy": 0.38801554301290764,
+      "eval_loss": 3.349801540374756,
+      "eval_runtime": 182.841,
+      "eval_samples_per_second": 98.506,
+      "eval_steps_per_second": 6.158,
+      "step": 70000
+    },
+    {
+      "epoch": 7.539554407491121,
+      "grad_norm": 0.7948490381240845,
+      "learning_rate": 0.00014796034910031245,
+      "loss": 3.2395,
+      "step": 70050
+    },
+    {
+      "epoch": 7.544935959530728,
+      "grad_norm": 0.7794623374938965,
+      "learning_rate": 0.00014763710807025104,
+      "loss": 3.2374,
+      "step": 70100
+    },
+    {
+      "epoch": 7.550317511570337,
+      "grad_norm": 0.7769473791122437,
+      "learning_rate": 0.0001473138670401896,
+      "loss": 3.2361,
+      "step": 70150
+    },
+    {
+      "epoch": 7.5556990636099455,
+      "grad_norm": 0.8074938654899597,
+      "learning_rate": 0.0001469906260101282,
+      "loss": 3.2296,
+      "step": 70200
+    },
+    {
+      "epoch": 7.561080615649553,
+      "grad_norm": 0.7983949184417725,
+      "learning_rate": 0.0001466673849800668,
+      "loss": 3.2314,
+      "step": 70250
+    },
+    {
+      "epoch": 7.566462167689162,
+      "grad_norm": 0.8049135208129883,
+      "learning_rate": 0.0001463441439500054,
+      "loss": 3.2324,
+      "step": 70300
+    },
+    {
+      "epoch": 7.57184371972877,
+      "grad_norm": 0.8315684795379639,
+      "learning_rate": 0.00014602090291994396,
+      "loss": 3.256,
+      "step": 70350
+    },
+    {
+      "epoch": 7.577225271768378,
+      "grad_norm": 0.8284087181091309,
+      "learning_rate": 0.00014569766188988255,
+      "loss": 3.2303,
+      "step": 70400
+    },
+    {
+      "epoch": 7.5826068238079865,
+      "grad_norm": 0.8085214495658875,
+      "learning_rate": 0.00014538088568042236,
+      "loss": 3.2405,
+      "step": 70450
+    },
+    {
+      "epoch": 7.587988375847594,
+      "grad_norm": 0.8195266723632812,
+      "learning_rate": 0.00014505764465036093,
+      "loss": 3.2479,
+      "step": 70500
+    },
+    {
+      "epoch": 7.593369927887203,
+      "grad_norm": 0.7887190580368042,
+      "learning_rate": 0.00014473440362029952,
+      "loss": 3.229,
+      "step": 70550
+    },
+    {
+      "epoch": 7.598751479926811,
+      "grad_norm": 0.7637575268745422,
+      "learning_rate": 0.00014441116259023812,
+      "loss": 3.2514,
+      "step": 70600
+    },
+    {
+      "epoch": 7.604133031966419,
+      "grad_norm": 0.8185248970985413,
+      "learning_rate": 0.00014408792156017669,
+      "loss": 3.2379,
+      "step": 70650
+    },
+    {
+      "epoch": 7.609514584006027,
+      "grad_norm": 0.7940756678581238,
+      "learning_rate": 0.00014376468053011528,
+      "loss": 3.2618,
+      "step": 70700
+    },
+    {
+      "epoch": 7.614896136045635,
+      "grad_norm": 0.7927145957946777,
+      "learning_rate": 0.00014344143950005387,
+      "loss": 3.2455,
+      "step": 70750
+    },
+    {
+      "epoch": 7.620277688085244,
+      "grad_norm": 0.7812114953994751,
+      "learning_rate": 0.00014311819846999244,
+      "loss": 3.2428,
+      "step": 70800
+    },
+    {
+      "epoch": 7.625659240124852,
+      "grad_norm": 0.7778552174568176,
+      "learning_rate": 0.00014279495743993104,
+      "loss": 3.2367,
+      "step": 70850
+    },
+    {
+      "epoch": 7.63104079216446,
+      "grad_norm": 0.8149660229682922,
+      "learning_rate": 0.00014247171640986963,
+      "loss": 3.244,
+      "step": 70900
+    },
+    {
+      "epoch": 7.636422344204068,
+      "grad_norm": 0.8159542679786682,
+      "learning_rate": 0.0001421484753798082,
+      "loss": 3.234,
+      "step": 70950
+    },
+    {
+      "epoch": 7.641803896243677,
+      "grad_norm": 0.7451802492141724,
+      "learning_rate": 0.0001418252343497468,
+      "loss": 3.247,
+      "step": 71000
+    },
+    {
+      "epoch": 7.641803896243677,
+      "eval_accuracy": 0.38847731778611916,
+      "eval_loss": 3.345766305923462,
+      "eval_runtime": 182.6769,
+      "eval_samples_per_second": 98.595,
+      "eval_steps_per_second": 6.164,
+      "step": 71000
+    },
+    {
+      "epoch": 7.647185448283285,
+      "grad_norm": 0.8018648624420166,
+      "learning_rate": 0.00014150199331968536,
+      "loss": 3.2532,
+      "step": 71050
+    },
+    {
+      "epoch": 7.652567000322893,
+      "grad_norm": 0.8257297277450562,
+      "learning_rate": 0.00014118521711022517,
+      "loss": 3.2431,
+      "step": 71100
+    },
+    {
+      "epoch": 7.657948552362502,
+      "grad_norm": 0.8156144618988037,
+      "learning_rate": 0.00014086197608016376,
+      "loss": 3.2349,
+      "step": 71150
+    },
+    {
+      "epoch": 7.663330104402109,
+      "grad_norm": 0.792155385017395,
+      "learning_rate": 0.00014053873505010236,
+      "loss": 3.2245,
+      "step": 71200
+    },
+    {
+      "epoch": 7.668711656441718,
+      "grad_norm": 0.8484143018722534,
+      "learning_rate": 0.00014021549402004092,
+      "loss": 3.2349,
+      "step": 71250
+    },
+    {
+      "epoch": 7.674093208481326,
+      "grad_norm": 0.8315868377685547,
+      "learning_rate": 0.00013989225298997952,
+      "loss": 3.2703,
+      "step": 71300
+    },
+    {
+      "epoch": 7.679474760520934,
+      "grad_norm": 0.7925692200660706,
+      "learning_rate": 0.00013956901195991811,
+      "loss": 3.2345,
+      "step": 71350
+    },
+    {
+      "epoch": 7.684856312560543,
+      "grad_norm": 0.8163354396820068,
+      "learning_rate": 0.00013924577092985668,
+      "loss": 3.2444,
+      "step": 71400
+    },
+    {
+      "epoch": 7.69023786460015,
+      "grad_norm": 0.82204669713974,
+      "learning_rate": 0.00013892252989979528,
+      "loss": 3.2424,
+      "step": 71450
+    },
+    {
+      "epoch": 7.695619416639759,
+      "grad_norm": 0.839684247970581,
+      "learning_rate": 0.00013859928886973384,
+      "loss": 3.2359,
+      "step": 71500
+    },
+    {
+      "epoch": 7.7010009686793675,
+      "grad_norm": 0.8018124103546143,
+      "learning_rate": 0.00013827604783967244,
+      "loss": 3.2321,
+      "step": 71550
+    },
+    {
+      "epoch": 7.706382520718975,
+      "grad_norm": 0.7850207090377808,
+      "learning_rate": 0.000137952806809611,
+      "loss": 3.2435,
+      "step": 71600
+    },
+    {
+      "epoch": 7.711764072758584,
+      "grad_norm": 0.7935765981674194,
+      "learning_rate": 0.0001376295657795496,
+      "loss": 3.2332,
+      "step": 71650
+    },
+    {
+      "epoch": 7.717145624798192,
+      "grad_norm": 0.8368144631385803,
+      "learning_rate": 0.0001373063247494882,
+      "loss": 3.2416,
+      "step": 71700
+    },
+    {
+      "epoch": 7.7225271768378,
+      "grad_norm": 0.8086535930633545,
+      "learning_rate": 0.00013698308371942676,
+      "loss": 3.2458,
+      "step": 71750
+    },
+    {
+      "epoch": 7.727908728877408,
+      "grad_norm": 0.8280001282691956,
+      "learning_rate": 0.00013665984268936536,
+      "loss": 3.2307,
+      "step": 71800
+    },
+    {
+      "epoch": 7.733290280917016,
+      "grad_norm": 0.8143231272697449,
+      "learning_rate": 0.00013633660165930395,
+      "loss": 3.2272,
+      "step": 71850
+    },
+    {
+      "epoch": 7.738671832956625,
+      "grad_norm": 0.784056544303894,
+      "learning_rate": 0.00013601336062924255,
+      "loss": 3.2341,
+      "step": 71900
+    },
+    {
+      "epoch": 7.744053384996233,
+      "grad_norm": 0.8832378387451172,
+      "learning_rate": 0.00013569011959918111,
+      "loss": 3.2553,
+      "step": 71950
+    },
+    {
+      "epoch": 7.749434937035841,
+      "grad_norm": 0.806605875492096,
+      "learning_rate": 0.0001353668785691197,
+      "loss": 3.2471,
+      "step": 72000
+    },
+    {
+      "epoch": 7.749434937035841,
+      "eval_accuracy": 0.3890513309924431,
+      "eval_loss": 3.342980146408081,
+      "eval_runtime": 182.5155,
+      "eval_samples_per_second": 98.682,
+      "eval_steps_per_second": 6.169,
+      "step": 72000
+    },
+    {
+      "epoch": 7.754816489075449,
+      "grad_norm": 0.8063587546348572,
+      "learning_rate": 0.00013504363753905828,
+      "loss": 3.2495,
+      "step": 72050
+    },
+    {
+      "epoch": 7.760198041115058,
+      "grad_norm": 0.8236513137817383,
+      "learning_rate": 0.00013472039650899687,
+      "loss": 3.2403,
+      "step": 72100
+    },
+    {
+      "epoch": 7.765579593154666,
+      "grad_norm": 0.801658034324646,
+      "learning_rate": 0.00013439715547893544,
+      "loss": 3.2375,
+      "step": 72150
+    },
+    {
+      "epoch": 7.770961145194274,
+      "grad_norm": 0.7991564273834229,
+      "learning_rate": 0.00013407391444887403,
+      "loss": 3.2362,
+      "step": 72200
+    },
+    {
+      "epoch": 7.776342697233883,
+      "grad_norm": 0.780725359916687,
+      "learning_rate": 0.0001337506734188126,
+      "loss": 3.2467,
+      "step": 72250
+    },
+    {
+      "epoch": 7.78172424927349,
+      "grad_norm": 0.7829933762550354,
+      "learning_rate": 0.0001334274323887512,
+      "loss": 3.2352,
+      "step": 72300
+    },
+    {
+      "epoch": 7.787105801313099,
+      "grad_norm": 0.8279498815536499,
+      "learning_rate": 0.0001331041913586898,
+      "loss": 3.2495,
+      "step": 72350
+    },
+    {
+      "epoch": 7.792487353352707,
+      "grad_norm": 0.8042828440666199,
+      "learning_rate": 0.00013278095032862838,
+      "loss": 3.2352,
+      "step": 72400
+    },
+    {
+      "epoch": 7.797868905392315,
+      "grad_norm": 0.8500546813011169,
+      "learning_rate": 0.00013245770929856695,
+      "loss": 3.248,
+      "step": 72450
+    },
+    {
+      "epoch": 7.803250457431924,
+      "grad_norm": 0.8876932859420776,
+      "learning_rate": 0.00013213446826850555,
+      "loss": 3.2425,
+      "step": 72500
+    },
+    {
+      "epoch": 7.808632009471531,
+      "grad_norm": 0.7689300775527954,
+      "learning_rate": 0.00013181122723844411,
+      "loss": 3.2308,
+      "step": 72550
+    },
+    {
+      "epoch": 7.81401356151114,
+      "grad_norm": 0.7971476912498474,
+      "learning_rate": 0.0001314879862083827,
+      "loss": 3.2349,
+      "step": 72600
+    },
+    {
+      "epoch": 7.819395113550748,
+      "grad_norm": 0.817987859249115,
+      "learning_rate": 0.00013116474517832128,
+      "loss": 3.2602,
+      "step": 72650
+    },
+    {
+      "epoch": 7.824776665590356,
+      "grad_norm": 0.819493293762207,
+      "learning_rate": 0.00013084150414825987,
+      "loss": 3.2283,
+      "step": 72700
+    },
+    {
+      "epoch": 7.830158217629965,
+      "grad_norm": 0.8192011117935181,
+      "learning_rate": 0.00013051826311819844,
+      "loss": 3.2344,
+      "step": 72750
+    },
+    {
+      "epoch": 7.835539769669572,
+      "grad_norm": 0.7826513051986694,
+      "learning_rate": 0.00013019502208813703,
+      "loss": 3.2499,
+      "step": 72800
+    },
+    {
+      "epoch": 7.840921321709181,
+      "grad_norm": 0.8338637351989746,
+      "learning_rate": 0.00012987178105807563,
+      "loss": 3.2288,
+      "step": 72850
+    },
+    {
+      "epoch": 7.846302873748789,
+      "grad_norm": 0.7698312997817993,
+      "learning_rate": 0.00012954854002801422,
+      "loss": 3.2517,
+      "step": 72900
+    },
+    {
+      "epoch": 7.851684425788397,
+      "grad_norm": 0.8017268180847168,
+      "learning_rate": 0.0001292252989979528,
+      "loss": 3.2544,
+      "step": 72950
+    },
+    {
+      "epoch": 7.857065977828006,
+      "grad_norm": 0.7961682677268982,
+      "learning_rate": 0.00012890205796789139,
+      "loss": 3.2406,
+      "step": 73000
+    },
+    {
+      "epoch": 7.857065977828006,
+      "eval_accuracy": 0.38943215936423287,
+      "eval_loss": 3.3387157917022705,
+      "eval_runtime": 182.7982,
+      "eval_samples_per_second": 98.529,
+      "eval_steps_per_second": 6.16,
+      "step": 73000
+    },
+    {
+      "epoch": 7.862447529867614,
+      "grad_norm": 0.8738141655921936,
+      "learning_rate": 0.00012857881693782998,
+      "loss": 3.2274,
+      "step": 73050
+    },
+    {
+      "epoch": 7.867829081907222,
+      "grad_norm": 0.7610816955566406,
+      "learning_rate": 0.00012825557590776855,
+      "loss": 3.2318,
+      "step": 73100
+    },
+    {
+      "epoch": 7.87321063394683,
+      "grad_norm": 0.8317003846168518,
+      "learning_rate": 0.00012793233487770714,
+      "loss": 3.2434,
+      "step": 73150
+    },
+    {
+      "epoch": 7.878592185986438,
+      "grad_norm": 0.8011530041694641,
+      "learning_rate": 0.0001276090938476457,
+      "loss": 3.2545,
+      "step": 73200
+    },
+    {
+      "epoch": 7.883973738026047,
+      "grad_norm": 0.8046674132347107,
+      "learning_rate": 0.0001272858528175843,
+      "loss": 3.2302,
+      "step": 73250
+    },
+    {
+      "epoch": 7.889355290065655,
+      "grad_norm": 0.8095680475234985,
+      "learning_rate": 0.00012696261178752287,
+      "loss": 3.2464,
+      "step": 73300
+    },
+    {
+      "epoch": 7.894736842105263,
+      "grad_norm": 0.7865450978279114,
+      "learning_rate": 0.00012663937075746147,
+      "loss": 3.2599,
+      "step": 73350
+    },
+    {
+      "epoch": 7.900118394144871,
+      "grad_norm": 0.7817733287811279,
+      "learning_rate": 0.00012631612972740006,
+      "loss": 3.2339,
+      "step": 73400
+    },
+    {
+      "epoch": 7.90549994618448,
+      "grad_norm": 0.8104103207588196,
+      "learning_rate": 0.00012599288869733863,
+      "loss": 3.2297,
+      "step": 73450
+    },
+    {
+      "epoch": 7.910881498224088,
+      "grad_norm": 0.8111141324043274,
+      "learning_rate": 0.00012566964766727722,
+      "loss": 3.2322,
+      "step": 73500
+    },
+    {
+      "epoch": 7.916263050263696,
+      "grad_norm": 0.8497397899627686,
+      "learning_rate": 0.00012534640663721582,
+      "loss": 3.2276,
+      "step": 73550
+    },
+    {
+      "epoch": 7.921644602303305,
+      "grad_norm": 0.8326553106307983,
+      "learning_rate": 0.00012502316560715439,
+      "loss": 3.2472,
+      "step": 73600
+    },
+    {
+      "epoch": 7.927026154342912,
+      "grad_norm": 0.8586411476135254,
+      "learning_rate": 0.00012469992457709298,
+      "loss": 3.2429,
+      "step": 73650
+    },
+    {
+      "epoch": 7.932407706382521,
+      "grad_norm": 0.7935066819190979,
+      "learning_rate": 0.00012437668354703158,
+      "loss": 3.2422,
+      "step": 73700
+    },
+    {
+      "epoch": 7.937789258422129,
+      "grad_norm": 0.7909044623374939,
+      "learning_rate": 0.00012405344251697014,
+      "loss": 3.2395,
+      "step": 73750
+    },
+    {
+      "epoch": 7.943170810461737,
+      "grad_norm": 0.8109498023986816,
+      "learning_rate": 0.00012373020148690874,
+      "loss": 3.2482,
+      "step": 73800
+    },
+    {
+      "epoch": 7.948552362501346,
+      "grad_norm": 0.8576630353927612,
+      "learning_rate": 0.0001234069604568473,
+      "loss": 3.239,
+      "step": 73850
+    },
+    {
+      "epoch": 7.953933914540953,
+      "grad_norm": 0.7980335354804993,
+      "learning_rate": 0.0001230837194267859,
+      "loss": 3.2356,
+      "step": 73900
+    },
+    {
+      "epoch": 7.959315466580562,
+      "grad_norm": 0.7563722729682922,
+      "learning_rate": 0.00012276047839672447,
+      "loss": 3.2249,
+      "step": 73950
+    },
+    {
+      "epoch": 7.96469701862017,
+      "grad_norm": 0.7781243324279785,
+      "learning_rate": 0.00012243723736666306,
+      "loss": 3.2495,
+      "step": 74000
+    },
+    {
+      "epoch": 7.96469701862017,
+      "eval_accuracy": 0.38969031862568004,
+      "eval_loss": 3.334991216659546,
+      "eval_runtime": 182.9415,
+      "eval_samples_per_second": 98.452,
+      "eval_steps_per_second": 6.155,
+      "step": 74000
+    },
+    {
+      "epoch": 7.970078570659778,
+      "grad_norm": 0.8000725507736206,
+      "learning_rate": 0.00012211399633660166,
+      "loss": 3.2204,
+      "step": 74050
+    },
+    {
+      "epoch": 7.975460122699387,
+      "grad_norm": 0.8195854425430298,
+      "learning_rate": 0.00012179075530654022,
+      "loss": 3.2446,
+      "step": 74100
+    },
+    {
+      "epoch": 7.980841674738995,
+      "grad_norm": 0.8186859488487244,
+      "learning_rate": 0.00012146751427647882,
+      "loss": 3.2262,
+      "step": 74150
+    },
+    {
+      "epoch": 7.986223226778603,
+      "grad_norm": 0.8260078430175781,
+      "learning_rate": 0.0001211442732464174,
+      "loss": 3.2422,
+      "step": 74200
+    },
+    {
+      "epoch": 7.991604778818211,
+      "grad_norm": 0.7887690663337708,
+      "learning_rate": 0.00012082103221635598,
+      "loss": 3.2354,
+      "step": 74250
+    },
+    {
+      "epoch": 7.996986330857819,
+      "grad_norm": 0.8294403553009033,
+      "learning_rate": 0.00012049779118629456,
+      "loss": 3.252,
+      "step": 74300
+    },
+    {
+      "epoch": 8.002367882897428,
+      "grad_norm": 0.830507755279541,
+      "learning_rate": 0.00012017455015623316,
+      "loss": 3.2075,
+      "step": 74350
+    },
+    {
+      "epoch": 8.007749434937036,
+      "grad_norm": 0.8426108360290527,
+      "learning_rate": 0.00011985130912617175,
+      "loss": 3.1655,
+      "step": 74400
+    },
+    {
+      "epoch": 8.013130986976645,
+      "grad_norm": 0.8193655610084534,
+      "learning_rate": 0.00011952806809611032,
+      "loss": 3.1579,
+      "step": 74450
+    },
+    {
+      "epoch": 8.018512539016251,
+      "grad_norm": 0.8287805318832397,
+      "learning_rate": 0.00011920482706604891,
+      "loss": 3.1689,
+      "step": 74500
+    },
+    {
+      "epoch": 8.02389409105586,
+      "grad_norm": 0.8014912009239197,
+      "learning_rate": 0.0001188815860359875,
+      "loss": 3.1657,
+      "step": 74550
+    },
+    {
+      "epoch": 8.029275643095469,
+      "grad_norm": 0.7986035943031311,
+      "learning_rate": 0.00011855834500592608,
+      "loss": 3.161,
+      "step": 74600
+    },
+    {
+      "epoch": 8.034657195135077,
+      "grad_norm": 0.7975103855133057,
+      "learning_rate": 0.00011823510397586466,
+      "loss": 3.1607,
+      "step": 74650
+    },
+    {
+      "epoch": 8.040038747174686,
+      "grad_norm": 0.8071293234825134,
+      "learning_rate": 0.00011791186294580325,
+      "loss": 3.1557,
+      "step": 74700
+    },
+    {
+      "epoch": 8.045420299214294,
+      "grad_norm": 0.8412976264953613,
+      "learning_rate": 0.00011758862191574182,
+      "loss": 3.1663,
+      "step": 74750
+    },
+    {
+      "epoch": 8.050801851253901,
+      "grad_norm": 0.8266987204551697,
+      "learning_rate": 0.00011726538088568041,
+      "loss": 3.1748,
+      "step": 74800
+    },
+    {
+      "epoch": 8.05618340329351,
+      "grad_norm": 0.8250101804733276,
+      "learning_rate": 0.000116942139855619,
+      "loss": 3.1774,
+      "step": 74850
+    },
+    {
+      "epoch": 8.061564955333118,
+      "grad_norm": 0.808516800403595,
+      "learning_rate": 0.00011661889882555759,
+      "loss": 3.1703,
+      "step": 74900
+    },
+    {
+      "epoch": 8.066946507372727,
+      "grad_norm": 0.8228300213813782,
+      "learning_rate": 0.00011629565779549616,
+      "loss": 3.1582,
+      "step": 74950
+    },
+    {
+      "epoch": 8.072328059412335,
+      "grad_norm": 0.8521013855934143,
+      "learning_rate": 0.00011597241676543475,
+      "loss": 3.191,
+      "step": 75000
+    },
+    {
+      "epoch": 8.072328059412335,
+      "eval_accuracy": 0.3899119705168216,
+      "eval_loss": 3.341465711593628,
+      "eval_runtime": 182.5475,
+      "eval_samples_per_second": 98.665,
+      "eval_steps_per_second": 6.168,
+      "step": 75000
+    },
+    {
+      "epoch": 8.077709611451942,
+      "grad_norm": 0.834549069404602,
+      "learning_rate": 0.00011564917573537335,
+      "loss": 3.1519,
+      "step": 75050
+    },
+    {
+      "epoch": 8.08309116349155,
+      "grad_norm": 0.8692988157272339,
+      "learning_rate": 0.00011533239952591314,
+      "loss": 3.1654,
+      "step": 75100
+    },
+    {
+      "epoch": 8.088472715531159,
+      "grad_norm": 0.8068045973777771,
+      "learning_rate": 0.00011500915849585173,
+      "loss": 3.1802,
+      "step": 75150
+    },
+    {
+      "epoch": 8.093854267570768,
+      "grad_norm": 0.8539791107177734,
+      "learning_rate": 0.00011468591746579033,
+      "loss": 3.1812,
+      "step": 75200
+    },
+    {
+      "epoch": 8.099235819610376,
+      "grad_norm": 0.8729422092437744,
+      "learning_rate": 0.0001143626764357289,
+      "loss": 3.1766,
+      "step": 75250
+    },
+    {
+      "epoch": 8.104617371649983,
+      "grad_norm": 0.8850834369659424,
+      "learning_rate": 0.00011403943540566749,
+      "loss": 3.1687,
+      "step": 75300
+    },
+    {
+      "epoch": 8.109998923689592,
+      "grad_norm": 0.8430905938148499,
+      "learning_rate": 0.00011371619437560607,
+      "loss": 3.1833,
+      "step": 75350
+    },
+    {
+      "epoch": 8.1153804757292,
+      "grad_norm": 0.7951804399490356,
+      "learning_rate": 0.00011339295334554464,
+      "loss": 3.1759,
+      "step": 75400
+    },
+    {
+      "epoch": 8.120762027768809,
+      "grad_norm": 0.8502519130706787,
+      "learning_rate": 0.00011306971231548323,
+      "loss": 3.1873,
+      "step": 75450
+    },
+    {
+      "epoch": 8.126143579808417,
+      "grad_norm": 1.6292310953140259,
+      "learning_rate": 0.00011274647128542183,
+      "loss": 3.1808,
+      "step": 75500
+    },
+    {
+      "epoch": 8.131525131848026,
+      "grad_norm": 0.8119076490402222,
+      "learning_rate": 0.0001124232302553604,
+      "loss": 3.1815,
+      "step": 75550
+    },
+    {
+      "epoch": 8.136906683887632,
+      "grad_norm": 0.8140049576759338,
+      "learning_rate": 0.00011209998922529899,
+      "loss": 3.1838,
+      "step": 75600
+    },
+    {
+      "epoch": 8.142288235927241,
+      "grad_norm": 0.8442477583885193,
+      "learning_rate": 0.00011177674819523757,
+      "loss": 3.1647,
+      "step": 75650
+    },
+    {
+      "epoch": 8.14766978796685,
+      "grad_norm": 0.8645812273025513,
+      "learning_rate": 0.00011145350716517617,
+      "loss": 3.1821,
+      "step": 75700
+    },
+    {
+      "epoch": 8.153051340006458,
+      "grad_norm": 0.778192400932312,
+      "learning_rate": 0.00011113026613511473,
+      "loss": 3.19,
+      "step": 75750
+    },
+    {
+      "epoch": 8.158432892046067,
+      "grad_norm": 0.8154362440109253,
+      "learning_rate": 0.00011080702510505333,
+      "loss": 3.1737,
+      "step": 75800
+    },
+    {
+      "epoch": 8.163814444085673,
+      "grad_norm": 0.8348283767700195,
+      "learning_rate": 0.00011048378407499191,
+      "loss": 3.1763,
+      "step": 75850
+    },
+    {
+      "epoch": 8.169195996125282,
+      "grad_norm": 0.8607150316238403,
+      "learning_rate": 0.00011016054304493049,
+      "loss": 3.1867,
+      "step": 75900
+    },
+    {
+      "epoch": 8.17457754816489,
+      "grad_norm": 0.8294938206672668,
+      "learning_rate": 0.00010983730201486907,
+      "loss": 3.179,
+      "step": 75950
+    },
+    {
+      "epoch": 8.1799591002045,
+      "grad_norm": 0.8422661423683167,
+      "learning_rate": 0.00010951406098480767,
+      "loss": 3.1735,
+      "step": 76000
+    },
+    {
+      "epoch": 8.1799591002045,
+      "eval_accuracy": 0.38991914160741736,
+      "eval_loss": 3.339107036590576,
+      "eval_runtime": 182.7233,
+      "eval_samples_per_second": 98.57,
+      "eval_steps_per_second": 6.162,
+      "step": 76000
+    },
+    {
+      "epoch": 8.185340652244108,
+      "grad_norm": 0.8388949036598206,
+      "learning_rate": 0.00010919081995474623,
+      "loss": 3.1938,
+      "step": 76050
+    },
+    {
+      "epoch": 8.190722204283716,
+      "grad_norm": 0.786255419254303,
+      "learning_rate": 0.00010886757892468483,
+      "loss": 3.1737,
+      "step": 76100
+    },
+    {
+      "epoch": 8.196103756323323,
+      "grad_norm": 0.8635239005088806,
+      "learning_rate": 0.00010854433789462342,
+      "loss": 3.1797,
+      "step": 76150
+    },
+    {
+      "epoch": 8.201485308362932,
+      "grad_norm": 0.8113918304443359,
+      "learning_rate": 0.000108221096864562,
+      "loss": 3.1891,
+      "step": 76200
+    },
+    {
+      "epoch": 8.20686686040254,
+      "grad_norm": 0.823483407497406,
+      "learning_rate": 0.00010789785583450059,
+      "loss": 3.1757,
+      "step": 76250
+    },
+    {
+      "epoch": 8.212248412442149,
+      "grad_norm": 0.7900793552398682,
+      "learning_rate": 0.00010757461480443917,
+      "loss": 3.1828,
+      "step": 76300
+    },
+    {
+      "epoch": 8.217629964481757,
+      "grad_norm": 0.8567901253700256,
+      "learning_rate": 0.00010725137377437776,
+      "loss": 3.1645,
+      "step": 76350
+    },
+    {
+      "epoch": 8.223011516521364,
+      "grad_norm": 0.8328141570091248,
+      "learning_rate": 0.00010692813274431633,
+      "loss": 3.1711,
+      "step": 76400
+    },
+    {
+      "epoch": 8.228393068560973,
+      "grad_norm": 0.8192028403282166,
+      "learning_rate": 0.00010660489171425492,
+      "loss": 3.1744,
+      "step": 76450
+    },
+    {
+      "epoch": 8.233774620600581,
+      "grad_norm": 0.8761533498764038,
+      "learning_rate": 0.0001062816506841935,
+      "loss": 3.1959,
+      "step": 76500
+    },
+    {
+      "epoch": 8.23915617264019,
+      "grad_norm": 0.8476336002349854,
+      "learning_rate": 0.00010595840965413209,
+      "loss": 3.1788,
+      "step": 76550
+    },
+    {
+      "epoch": 8.244537724679798,
+      "grad_norm": 0.8509985208511353,
+      "learning_rate": 0.00010563516862407067,
+      "loss": 3.1825,
+      "step": 76600
+    },
+    {
+      "epoch": 8.249919276719407,
+      "grad_norm": 0.8411309123039246,
+      "learning_rate": 0.00010531192759400926,
+      "loss": 3.1826,
+      "step": 76650
+    },
+    {
+      "epoch": 8.255300828759013,
+      "grad_norm": 0.8168168067932129,
+      "learning_rate": 0.00010498868656394784,
+      "loss": 3.185,
+      "step": 76700
+    },
+    {
+      "epoch": 8.260682380798622,
+      "grad_norm": 0.8766189813613892,
+      "learning_rate": 0.00010466544553388642,
+      "loss": 3.1826,
+      "step": 76750
+    },
+    {
+      "epoch": 8.26606393283823,
+      "grad_norm": 0.7768109440803528,
+      "learning_rate": 0.000104342204503825,
+      "loss": 3.193,
+      "step": 76800
+    },
+    {
+      "epoch": 8.27144548487784,
+      "grad_norm": 0.8481165170669556,
+      "learning_rate": 0.0001040189634737636,
+      "loss": 3.1845,
+      "step": 76850
+    },
+    {
+      "epoch": 8.276827036917448,
+      "grad_norm": 0.83498215675354,
+      "learning_rate": 0.00010369572244370217,
+      "loss": 3.1856,
+      "step": 76900
+    },
+    {
+      "epoch": 8.282208588957054,
+      "grad_norm": 0.8582146167755127,
+      "learning_rate": 0.00010337248141364076,
+      "loss": 3.1902,
+      "step": 76950
+    },
+    {
+      "epoch": 8.287590140996663,
+      "grad_norm": 0.8045734167098999,
+      "learning_rate": 0.00010304924038357936,
+      "loss": 3.1844,
+      "step": 77000
+    },
+    {
+      "epoch": 8.287590140996663,
+      "eval_accuracy": 0.3902144601564971,
+      "eval_loss": 3.3361928462982178,
+      "eval_runtime": 182.7919,
+      "eval_samples_per_second": 98.533,
+      "eval_steps_per_second": 6.16,
+      "step": 77000
+    },
+    {
+      "epoch": 8.292971693036272,
+      "grad_norm": 0.854202926158905,
+      "learning_rate": 0.00010272599935351792,
+      "loss": 3.1828,
+      "step": 77050
+    },
+    {
+      "epoch": 8.29835324507588,
+      "grad_norm": 0.8577749729156494,
+      "learning_rate": 0.00010240922314405774,
+      "loss": 3.1808,
+      "step": 77100
+    },
+    {
+      "epoch": 8.303734797115489,
+      "grad_norm": 0.8859609365463257,
+      "learning_rate": 0.00010208598211399634,
+      "loss": 3.1855,
+      "step": 77150
+    },
+    {
+      "epoch": 8.309116349155097,
+      "grad_norm": 0.8407235145568848,
+      "learning_rate": 0.0001017627410839349,
+      "loss": 3.1775,
+      "step": 77200
+    },
+    {
+      "epoch": 8.314497901194704,
+      "grad_norm": 0.8498949408531189,
+      "learning_rate": 0.0001014395000538735,
+      "loss": 3.1787,
+      "step": 77250
+    },
+    {
+      "epoch": 8.319879453234313,
+      "grad_norm": 0.8562375903129578,
+      "learning_rate": 0.00010111625902381208,
+      "loss": 3.1898,
+      "step": 77300
+    },
+    {
+      "epoch": 8.325261005273921,
+      "grad_norm": 0.8841971755027771,
+      "learning_rate": 0.00010079301799375066,
+      "loss": 3.2001,
+      "step": 77350
+    },
+    {
+      "epoch": 8.33064255731353,
+      "grad_norm": 0.8151336312294006,
+      "learning_rate": 0.00010046977696368924,
+      "loss": 3.1902,
+      "step": 77400
+    },
+    {
+      "epoch": 8.336024109353138,
+      "grad_norm": 0.8526255488395691,
+      "learning_rate": 0.00010014653593362784,
+      "loss": 3.1803,
+      "step": 77450
+    },
+    {
+      "epoch": 8.341405661392745,
+      "grad_norm": 0.8695437908172607,
+      "learning_rate": 9.982329490356642e-05,
+      "loss": 3.1975,
+      "step": 77500
+    },
+    {
+      "epoch": 8.346787213432354,
+      "grad_norm": 0.9299212694168091,
+      "learning_rate": 9.9500053873505e-05,
+      "loss": 3.1894,
+      "step": 77550
+    },
+    {
+      "epoch": 8.352168765471962,
+      "grad_norm": 0.8620630502700806,
+      "learning_rate": 9.917681284344358e-05,
+      "loss": 3.1769,
+      "step": 77600
+    },
+    {
+      "epoch": 8.35755031751157,
+      "grad_norm": 0.8638424277305603,
+      "learning_rate": 9.885357181338218e-05,
+      "loss": 3.1993,
+      "step": 77650
+    },
+    {
+      "epoch": 8.36293186955118,
+      "grad_norm": 0.832884669303894,
+      "learning_rate": 9.853033078332074e-05,
+      "loss": 3.1877,
+      "step": 77700
+    },
+    {
+      "epoch": 8.368313421590786,
+      "grad_norm": 0.857495129108429,
+      "learning_rate": 9.820708975325934e-05,
+      "loss": 3.1844,
+      "step": 77750
+    },
+    {
+      "epoch": 8.373694973630395,
+      "grad_norm": 0.8742907047271729,
+      "learning_rate": 9.788384872319793e-05,
+      "loss": 3.1762,
+      "step": 77800
+    },
+    {
+      "epoch": 8.379076525670003,
+      "grad_norm": 0.8619995713233948,
+      "learning_rate": 9.75606076931365e-05,
+      "loss": 3.1828,
+      "step": 77850
+    },
+    {
+      "epoch": 8.384458077709612,
+      "grad_norm": 0.8461150527000427,
+      "learning_rate": 9.723736666307508e-05,
+      "loss": 3.2053,
+      "step": 77900
+    },
+    {
+      "epoch": 8.38983962974922,
+      "grad_norm": 0.809444785118103,
+      "learning_rate": 9.691412563301368e-05,
+      "loss": 3.1613,
+      "step": 77950
+    },
+    {
+      "epoch": 8.395221181788829,
+      "grad_norm": 0.8414546251296997,
+      "learning_rate": 9.659088460295227e-05,
+      "loss": 3.1819,
+      "step": 78000
+    },
+    {
+      "epoch": 8.395221181788829,
+      "eval_accuracy": 0.3904907644502093,
+      "eval_loss": 3.3329615592956543,
+      "eval_runtime": 182.3978,
+      "eval_samples_per_second": 98.746,
+      "eval_steps_per_second": 6.173,
+      "step": 78000
+    },
+    {
+      "epoch": 8.400602733828435,
+      "grad_norm": 0.8904616832733154,
+      "learning_rate": 9.626764357289084e-05,
+      "loss": 3.1935,
+      "step": 78050
+    },
+    {
+      "epoch": 8.405984285868044,
+      "grad_norm": 0.8500921726226807,
+      "learning_rate": 9.594440254282943e-05,
+      "loss": 3.201,
+      "step": 78100
+    },
+    {
+      "epoch": 8.411365837907653,
+      "grad_norm": 0.8491564989089966,
+      "learning_rate": 9.562116151276802e-05,
+      "loss": 3.18,
+      "step": 78150
+    },
+    {
+      "epoch": 8.416747389947261,
+      "grad_norm": 0.8127497434616089,
+      "learning_rate": 9.52979204827066e-05,
+      "loss": 3.1789,
+      "step": 78200
+    },
+    {
+      "epoch": 8.42212894198687,
+      "grad_norm": 0.8613414764404297,
+      "learning_rate": 9.497467945264518e-05,
+      "loss": 3.1947,
+      "step": 78250
+    },
+    {
+      "epoch": 8.427510494026476,
+      "grad_norm": 0.8296908140182495,
+      "learning_rate": 9.465143842258377e-05,
+      "loss": 3.174,
+      "step": 78300
+    },
+    {
+      "epoch": 8.432892046066085,
+      "grad_norm": 0.8717524409294128,
+      "learning_rate": 9.432819739252234e-05,
+      "loss": 3.1924,
+      "step": 78350
+    },
+    {
+      "epoch": 8.438273598105694,
+      "grad_norm": 0.8354899287223816,
+      "learning_rate": 9.400495636246093e-05,
+      "loss": 3.1854,
+      "step": 78400
+    },
+    {
+      "epoch": 8.443655150145302,
+      "grad_norm": 0.8695220351219177,
+      "learning_rate": 9.368171533239952e-05,
+      "loss": 3.1859,
+      "step": 78450
+    },
+    {
+      "epoch": 8.44903670218491,
+      "grad_norm": 0.8374029994010925,
+      "learning_rate": 9.335847430233811e-05,
+      "loss": 3.1781,
+      "step": 78500
+    },
+    {
+      "epoch": 8.45441825422452,
+      "grad_norm": 0.8391903638839722,
+      "learning_rate": 9.303523327227668e-05,
+      "loss": 3.1786,
+      "step": 78550
+    },
+    {
+      "epoch": 8.459799806264126,
+      "grad_norm": 0.8173428177833557,
+      "learning_rate": 9.271199224221527e-05,
+      "loss": 3.1855,
+      "step": 78600
+    },
+    {
+      "epoch": 8.465181358303735,
+      "grad_norm": 0.8497678637504578,
+      "learning_rate": 9.238875121215387e-05,
+      "loss": 3.2016,
+      "step": 78650
+    },
+    {
+      "epoch": 8.470562910343343,
+      "grad_norm": 0.8355598449707031,
+      "learning_rate": 9.206551018209243e-05,
+      "loss": 3.1891,
+      "step": 78700
+    },
+    {
+      "epoch": 8.475944462382952,
+      "grad_norm": 0.9021919965744019,
+      "learning_rate": 9.174226915203103e-05,
+      "loss": 3.1931,
+      "step": 78750
+    },
+    {
+      "epoch": 8.48132601442256,
+      "grad_norm": 0.8534740805625916,
+      "learning_rate": 9.141902812196961e-05,
+      "loss": 3.1779,
+      "step": 78800
+    },
+    {
+      "epoch": 8.486707566462167,
+      "grad_norm": 0.8480976819992065,
+      "learning_rate": 9.109578709190818e-05,
+      "loss": 3.1862,
+      "step": 78850
+    },
+    {
+      "epoch": 8.492089118501776,
+      "grad_norm": 0.8286227583885193,
+      "learning_rate": 9.077254606184677e-05,
+      "loss": 3.214,
+      "step": 78900
+    },
+    {
+      "epoch": 8.497470670541384,
+      "grad_norm": 0.8237711787223816,
+      "learning_rate": 9.044930503178537e-05,
+      "loss": 3.1884,
+      "step": 78950
+    },
+    {
+      "epoch": 8.502852222580993,
+      "grad_norm": 0.8402367234230042,
+      "learning_rate": 9.012606400172395e-05,
+      "loss": 3.1854,
+      "step": 79000
+    },
+    {
+      "epoch": 8.502852222580993,
+      "eval_accuracy": 0.39106488630942104,
+      "eval_loss": 3.32970929145813,
+      "eval_runtime": 182.9601,
+      "eval_samples_per_second": 98.442,
+      "eval_steps_per_second": 6.154,
+      "step": 79000
+    },
+    {
+      "epoch": 8.508233774620601,
+      "grad_norm": 0.8203577995300293,
+      "learning_rate": 8.980282297166253e-05,
+      "loss": 3.1807,
+      "step": 79050
+    },
+    {
+      "epoch": 8.513615326660208,
+      "grad_norm": 0.8231856822967529,
+      "learning_rate": 8.947958194160111e-05,
+      "loss": 3.1975,
+      "step": 79100
+    },
+    {
+      "epoch": 8.518996878699816,
+      "grad_norm": 0.8475301861763,
+      "learning_rate": 8.916280573214092e-05,
+      "loss": 3.2022,
+      "step": 79150
+    },
+    {
+      "epoch": 8.524378430739425,
+      "grad_norm": 0.9068256616592407,
+      "learning_rate": 8.883956470207951e-05,
+      "loss": 3.1841,
+      "step": 79200
+    },
+    {
+      "epoch": 8.529759982779034,
+      "grad_norm": 0.8666347861289978,
+      "learning_rate": 8.851632367201809e-05,
+      "loss": 3.1918,
+      "step": 79250
+    },
+    {
+      "epoch": 8.535141534818642,
+      "grad_norm": 0.8289416432380676,
+      "learning_rate": 8.819308264195669e-05,
+      "loss": 3.1962,
+      "step": 79300
+    },
+    {
+      "epoch": 8.54052308685825,
+      "grad_norm": 0.8145918846130371,
+      "learning_rate": 8.786984161189526e-05,
+      "loss": 3.1855,
+      "step": 79350
+    },
+    {
+      "epoch": 8.545904638897857,
+      "grad_norm": 0.8508062958717346,
+      "learning_rate": 8.754660058183385e-05,
+      "loss": 3.1653,
+      "step": 79400
+    },
+    {
+      "epoch": 8.551286190937466,
+      "grad_norm": 0.8114455342292786,
+      "learning_rate": 8.722335955177243e-05,
+      "loss": 3.1935,
+      "step": 79450
+    },
+    {
+      "epoch": 8.556667742977075,
+      "grad_norm": 0.8585551977157593,
+      "learning_rate": 8.690658334231224e-05,
+      "loss": 3.2131,
+      "step": 79500
+    },
+    {
+      "epoch": 8.562049295016683,
+      "grad_norm": 0.8328741192817688,
+      "learning_rate": 8.658334231225083e-05,
+      "loss": 3.1715,
+      "step": 79550
+    },
+    {
+      "epoch": 8.567430847056292,
+      "grad_norm": 0.8506522178649902,
+      "learning_rate": 8.626010128218943e-05,
+      "loss": 3.2035,
+      "step": 79600
+    },
+    {
+      "epoch": 8.572812399095898,
+      "grad_norm": 0.8559414148330688,
+      "learning_rate": 8.5936860252128e-05,
+      "loss": 3.198,
+      "step": 79650
+    },
+    {
+      "epoch": 8.578193951135507,
+      "grad_norm": 0.8611463904380798,
+      "learning_rate": 8.561361922206658e-05,
+      "loss": 3.1907,
+      "step": 79700
+    },
+    {
+      "epoch": 8.583575503175116,
+      "grad_norm": 0.7964751720428467,
+      "learning_rate": 8.529037819200517e-05,
+      "loss": 3.1686,
+      "step": 79750
+    },
+    {
+      "epoch": 8.588957055214724,
+      "grad_norm": 0.8190116882324219,
+      "learning_rate": 8.496713716194374e-05,
+      "loss": 3.1866,
+      "step": 79800
+    },
+    {
+      "epoch": 8.594338607254333,
+      "grad_norm": 0.9009842872619629,
+      "learning_rate": 8.464389613188233e-05,
+      "loss": 3.1879,
+      "step": 79850
+    },
+    {
+      "epoch": 8.599720159293941,
+      "grad_norm": 0.7971447706222534,
+      "learning_rate": 8.432065510182093e-05,
+      "loss": 3.2059,
+      "step": 79900
+    },
+    {
+      "epoch": 8.605101711333548,
+      "grad_norm": 0.8355123996734619,
+      "learning_rate": 8.39974140717595e-05,
+      "loss": 3.1917,
+      "step": 79950
+    },
+    {
+      "epoch": 8.610483263373157,
+      "grad_norm": 0.9042732119560242,
+      "learning_rate": 8.367417304169809e-05,
+      "loss": 3.1847,
+      "step": 80000
+    },
+    {
+      "epoch": 8.610483263373157,
+      "eval_accuracy": 0.3913339108596497,
+      "eval_loss": 3.327451705932617,
+      "eval_runtime": 182.9461,
+      "eval_samples_per_second": 98.45,
+      "eval_steps_per_second": 6.155,
+      "step": 80000
+    },
+    {
+      "epoch": 8.615864815412765,
+      "grad_norm": 0.8184645175933838,
+      "learning_rate": 8.335093201163667e-05,
+      "loss": 3.177,
+      "step": 80050
+    },
+    {
+      "epoch": 8.621246367452374,
+      "grad_norm": 0.8244253396987915,
+      "learning_rate": 8.302769098157526e-05,
+      "loss": 3.1778,
+      "step": 80100
+    },
+    {
+      "epoch": 8.626627919491982,
+      "grad_norm": 0.934569239616394,
+      "learning_rate": 8.270444995151383e-05,
+      "loss": 3.2007,
+      "step": 80150
+    },
+    {
+      "epoch": 8.632009471531589,
+      "grad_norm": 0.798776388168335,
+      "learning_rate": 8.238120892145243e-05,
+      "loss": 3.1926,
+      "step": 80200
+    },
+    {
+      "epoch": 8.637391023571197,
+      "grad_norm": 0.8425439596176147,
+      "learning_rate": 8.205796789139101e-05,
+      "loss": 3.2076,
+      "step": 80250
+    },
+    {
+      "epoch": 8.642772575610806,
+      "grad_norm": 0.8503188490867615,
+      "learning_rate": 8.173472686132959e-05,
+      "loss": 3.1863,
+      "step": 80300
+    },
+    {
+      "epoch": 8.648154127650415,
+      "grad_norm": 0.8668044209480286,
+      "learning_rate": 8.141148583126817e-05,
+      "loss": 3.1889,
+      "step": 80350
+    },
+    {
+      "epoch": 8.653535679690023,
+      "grad_norm": 0.8459951877593994,
+      "learning_rate": 8.108824480120676e-05,
+      "loss": 3.1957,
+      "step": 80400
+    },
+    {
+      "epoch": 8.658917231729632,
+      "grad_norm": 0.911152720451355,
+      "learning_rate": 8.076500377114533e-05,
+      "loss": 3.1835,
+      "step": 80450
+    },
+    {
+      "epoch": 8.664298783769238,
+      "grad_norm": 0.8265844583511353,
+      "learning_rate": 8.044176274108393e-05,
+      "loss": 3.2005,
+      "step": 80500
+    },
+    {
+      "epoch": 8.669680335808847,
+      "grad_norm": 0.8671901226043701,
+      "learning_rate": 8.011852171102252e-05,
+      "loss": 3.2096,
+      "step": 80550
+    },
+    {
+      "epoch": 8.675061887848456,
+      "grad_norm": 0.8355448842048645,
+      "learning_rate": 7.97952806809611e-05,
+      "loss": 3.187,
+      "step": 80600
+    },
+    {
+      "epoch": 8.680443439888064,
+      "grad_norm": 0.8255912065505981,
+      "learning_rate": 7.947203965089967e-05,
+      "loss": 3.1844,
+      "step": 80650
+    },
+    {
+      "epoch": 8.685824991927673,
+      "grad_norm": 0.8755044341087341,
+      "learning_rate": 7.914879862083827e-05,
+      "loss": 3.1869,
+      "step": 80700
+    },
+    {
+      "epoch": 8.69120654396728,
+      "grad_norm": 0.8891885876655579,
+      "learning_rate": 7.882555759077686e-05,
+      "loss": 3.1692,
+      "step": 80750
+    },
+    {
+      "epoch": 8.696588096006888,
+      "grad_norm": 0.8529818058013916,
+      "learning_rate": 7.850231656071543e-05,
+      "loss": 3.188,
+      "step": 80800
+    },
+    {
+      "epoch": 8.701969648046497,
+      "grad_norm": 0.8659524321556091,
+      "learning_rate": 7.817907553065402e-05,
+      "loss": 3.1952,
+      "step": 80850
+    },
+    {
+      "epoch": 8.707351200086105,
+      "grad_norm": 0.8727506995201111,
+      "learning_rate": 7.78558345005926e-05,
+      "loss": 3.1946,
+      "step": 80900
+    },
+    {
+      "epoch": 8.712732752125714,
+      "grad_norm": 0.8414073586463928,
+      "learning_rate": 7.753259347053118e-05,
+      "loss": 3.1998,
+      "step": 80950
+    },
+    {
+      "epoch": 8.718114304165322,
+      "grad_norm": 0.8435977101325989,
+      "learning_rate": 7.720935244046977e-05,
+      "loss": 3.1935,
+      "step": 81000
+    },
+    {
+      "epoch": 8.718114304165322,
+      "eval_accuracy": 0.3915950037490679,
+      "eval_loss": 3.323737382888794,
+      "eval_runtime": 183.0449,
+      "eval_samples_per_second": 98.397,
+      "eval_steps_per_second": 6.151,
+      "step": 81000
+    },
+    {
+      "epoch": 8.723495856204929,
+      "grad_norm": 0.8313448429107666,
+      "learning_rate": 7.688611141040836e-05,
+      "loss": 3.1835,
+      "step": 81050
+    },
+    {
+      "epoch": 8.728877408244538,
+      "grad_norm": 0.8675030469894409,
+      "learning_rate": 7.656287038034694e-05,
+      "loss": 3.1836,
+      "step": 81100
+    },
+    {
+      "epoch": 8.734258960284146,
+      "grad_norm": 0.8448699116706848,
+      "learning_rate": 7.623962935028552e-05,
+      "loss": 3.2062,
+      "step": 81150
+    },
+    {
+      "epoch": 8.739640512323755,
+      "grad_norm": 0.8596765995025635,
+      "learning_rate": 7.59163883202241e-05,
+      "loss": 3.1863,
+      "step": 81200
+    },
+    {
+      "epoch": 8.745022064363363,
+      "grad_norm": 0.8727709054946899,
+      "learning_rate": 7.55931472901627e-05,
+      "loss": 3.1742,
+      "step": 81250
+    },
+    {
+      "epoch": 8.75040361640297,
+      "grad_norm": 0.8483537435531616,
+      "learning_rate": 7.526990626010127e-05,
+      "loss": 3.1818,
+      "step": 81300
+    },
+    {
+      "epoch": 8.755785168442578,
+      "grad_norm": 0.8239384293556213,
+      "learning_rate": 7.494666523003986e-05,
+      "loss": 3.1747,
+      "step": 81350
+    },
+    {
+      "epoch": 8.761166720482187,
+      "grad_norm": 0.8851037621498108,
+      "learning_rate": 7.462342419997844e-05,
+      "loss": 3.1878,
+      "step": 81400
+    },
+    {
+      "epoch": 8.766548272521796,
+      "grad_norm": 0.870302677154541,
+      "learning_rate": 7.430018316991704e-05,
+      "loss": 3.1834,
+      "step": 81450
+    },
+    {
+      "epoch": 8.771929824561404,
+      "grad_norm": 0.8676862120628357,
+      "learning_rate": 7.397694213985562e-05,
+      "loss": 3.1805,
+      "step": 81500
+    },
+    {
+      "epoch": 8.777311376601011,
+      "grad_norm": 0.8838236331939697,
+      "learning_rate": 7.36537011097942e-05,
+      "loss": 3.1878,
+      "step": 81550
+    },
+    {
+      "epoch": 8.78269292864062,
+      "grad_norm": 0.9052426218986511,
+      "learning_rate": 7.333046007973278e-05,
+      "loss": 3.176,
+      "step": 81600
+    },
+    {
+      "epoch": 8.788074480680228,
+      "grad_norm": 0.827214241027832,
+      "learning_rate": 7.300721904967136e-05,
+      "loss": 3.1943,
+      "step": 81650
+    },
+    {
+      "epoch": 8.793456032719837,
+      "grad_norm": 0.8450335264205933,
+      "learning_rate": 7.268397801960996e-05,
+      "loss": 3.1906,
+      "step": 81700
+    },
+    {
+      "epoch": 8.798837584759445,
+      "grad_norm": 0.8375703692436218,
+      "learning_rate": 7.236073698954854e-05,
+      "loss": 3.1844,
+      "step": 81750
+    },
+    {
+      "epoch": 8.804219136799054,
+      "grad_norm": 0.816857635974884,
+      "learning_rate": 7.203749595948712e-05,
+      "loss": 3.1712,
+      "step": 81800
+    },
+    {
+      "epoch": 8.80960068883866,
+      "grad_norm": 0.8428307771682739,
+      "learning_rate": 7.17142549294257e-05,
+      "loss": 3.1939,
+      "step": 81850
+    },
+    {
+      "epoch": 8.814982240878269,
+      "grad_norm": 0.8175262212753296,
+      "learning_rate": 7.139101389936428e-05,
+      "loss": 3.1919,
+      "step": 81900
+    },
+    {
+      "epoch": 8.820363792917878,
+      "grad_norm": 0.8650425672531128,
+      "learning_rate": 7.106777286930287e-05,
+      "loss": 3.2,
+      "step": 81950
+    },
+    {
+      "epoch": 8.825745344957486,
+      "grad_norm": 0.866899311542511,
+      "learning_rate": 7.074453183924146e-05,
+      "loss": 3.1881,
+      "step": 82000
+    },
+    {
+      "epoch": 8.825745344957486,
+      "eval_accuracy": 0.39193965070921544,
+      "eval_loss": 3.3191041946411133,
+      "eval_runtime": 182.564,
+      "eval_samples_per_second": 98.656,
+      "eval_steps_per_second": 6.168,
+      "step": 82000
+    },
+    {
+      "epoch": 8.831126896997095,
+      "grad_norm": 0.8628833889961243,
+      "learning_rate": 7.042129080918004e-05,
+      "loss": 3.1915,
+      "step": 82050
+    },
+    {
+      "epoch": 8.836508449036701,
+      "grad_norm": 0.9049327969551086,
+      "learning_rate": 7.009804977911862e-05,
+      "loss": 3.1922,
+      "step": 82100
+    },
+    {
+      "epoch": 8.84189000107631,
+      "grad_norm": 0.8244563937187195,
+      "learning_rate": 6.97748087490572e-05,
+      "loss": 3.1827,
+      "step": 82150
+    },
+    {
+      "epoch": 8.847271553115919,
+      "grad_norm": 0.8381883502006531,
+      "learning_rate": 6.945156771899579e-05,
+      "loss": 3.2009,
+      "step": 82200
+    },
+    {
+      "epoch": 8.852653105155527,
+      "grad_norm": 0.8337753415107727,
+      "learning_rate": 6.912832668893437e-05,
+      "loss": 3.1808,
+      "step": 82250
+    },
+    {
+      "epoch": 8.858034657195136,
+      "grad_norm": 0.8778477907180786,
+      "learning_rate": 6.880508565887297e-05,
+      "loss": 3.218,
+      "step": 82300
+    },
+    {
+      "epoch": 8.863416209234742,
+      "grad_norm": 0.8658258318901062,
+      "learning_rate": 6.848184462881155e-05,
+      "loss": 3.198,
+      "step": 82350
+    },
+    {
+      "epoch": 8.868797761274351,
+      "grad_norm": 0.861770749092102,
+      "learning_rate": 6.815860359875013e-05,
+      "loss": 3.1882,
+      "step": 82400
+    },
+    {
+      "epoch": 8.87417931331396,
+      "grad_norm": 1.062533974647522,
+      "learning_rate": 6.783536256868871e-05,
+      "loss": 3.2048,
+      "step": 82450
+    },
+    {
+      "epoch": 8.879560865353568,
+      "grad_norm": 0.878668487071991,
+      "learning_rate": 6.75121215386273e-05,
+      "loss": 3.1679,
+      "step": 82500
+    },
+    {
+      "epoch": 8.884942417393177,
+      "grad_norm": 0.8749240636825562,
+      "learning_rate": 6.718888050856589e-05,
+      "loss": 3.1839,
+      "step": 82550
+    },
+    {
+      "epoch": 8.890323969432785,
+      "grad_norm": 0.8733687996864319,
+      "learning_rate": 6.686563947850447e-05,
+      "loss": 3.1964,
+      "step": 82600
+    },
+    {
+      "epoch": 8.895705521472392,
+      "grad_norm": 0.8685417175292969,
+      "learning_rate": 6.654239844844305e-05,
+      "loss": 3.1968,
+      "step": 82650
+    },
+    {
+      "epoch": 8.901087073512,
+      "grad_norm": 0.8061568140983582,
+      "learning_rate": 6.621915741838163e-05,
+      "loss": 3.1726,
+      "step": 82700
+    },
+    {
+      "epoch": 8.906468625551609,
+      "grad_norm": 0.8258588314056396,
+      "learning_rate": 6.589591638832021e-05,
+      "loss": 3.1938,
+      "step": 82750
+    },
+    {
+      "epoch": 8.911850177591218,
+      "grad_norm": 0.8688393235206604,
+      "learning_rate": 6.557267535825881e-05,
+      "loss": 3.1881,
+      "step": 82800
+    },
+    {
+      "epoch": 8.917231729630826,
+      "grad_norm": 0.870529055595398,
+      "learning_rate": 6.524943432819739e-05,
+      "loss": 3.1997,
+      "step": 82850
+    },
+    {
+      "epoch": 8.922613281670433,
+      "grad_norm": 0.8257046341896057,
+      "learning_rate": 6.492619329813597e-05,
+      "loss": 3.1816,
+      "step": 82900
+    },
+    {
+      "epoch": 8.927994833710041,
+      "grad_norm": 0.858521580696106,
+      "learning_rate": 6.460295226807455e-05,
+      "loss": 3.1988,
+      "step": 82950
+    },
+    {
+      "epoch": 8.93337638574965,
+      "grad_norm": 0.8709513545036316,
+      "learning_rate": 6.427971123801313e-05,
+      "loss": 3.1831,
+      "step": 83000
+    },
+    {
+      "epoch": 8.93337638574965,
+      "eval_accuracy": 0.3923701334507364,
+      "eval_loss": 3.3174479007720947,
+      "eval_runtime": 182.9075,
+      "eval_samples_per_second": 98.471,
+      "eval_steps_per_second": 6.156,
+      "step": 83000
+    },
+    {
+      "epoch": 8.938757937789259,
+      "grad_norm": 0.8560492992401123,
+      "learning_rate": 6.395647020795173e-05,
+      "loss": 3.1856,
+      "step": 83050
+    },
+    {
+      "epoch": 8.944139489828867,
+      "grad_norm": 0.8757853507995605,
+      "learning_rate": 6.363322917789031e-05,
+      "loss": 3.1807,
+      "step": 83100
+    },
+    {
+      "epoch": 8.949521041868476,
+      "grad_norm": 0.852837860584259,
+      "learning_rate": 6.330998814782889e-05,
+      "loss": 3.1919,
+      "step": 83150
+    },
+    {
+      "epoch": 8.954902593908082,
+      "grad_norm": 0.8725079298019409,
+      "learning_rate": 6.298674711776748e-05,
+      "loss": 3.1904,
+      "step": 83200
+    },
+    {
+      "epoch": 8.960284145947691,
+      "grad_norm": 0.8584096431732178,
+      "learning_rate": 6.266350608770606e-05,
+      "loss": 3.1921,
+      "step": 83250
+    },
+    {
+      "epoch": 8.9656656979873,
+      "grad_norm": 0.8362786769866943,
+      "learning_rate": 6.234026505764465e-05,
+      "loss": 3.1988,
+      "step": 83300
+    },
+    {
+      "epoch": 8.971047250026908,
+      "grad_norm": 0.8816782832145691,
+      "learning_rate": 6.201702402758323e-05,
+      "loss": 3.1755,
+      "step": 83350
+    },
+    {
+      "epoch": 8.976428802066517,
+      "grad_norm": 0.8913627862930298,
+      "learning_rate": 6.169378299752181e-05,
+      "loss": 3.1949,
+      "step": 83400
+    },
+    {
+      "epoch": 8.981810354106123,
+      "grad_norm": 0.8645463585853577,
+      "learning_rate": 6.13705419674604e-05,
+      "loss": 3.1884,
+      "step": 83450
+    },
+    {
+      "epoch": 8.987191906145732,
+      "grad_norm": 0.8139950633049011,
+      "learning_rate": 6.104730093739898e-05,
+      "loss": 3.1769,
+      "step": 83500
+    },
+    {
+      "epoch": 8.99257345818534,
+      "grad_norm": 0.9117967486381531,
+      "learning_rate": 6.073052472793879e-05,
+      "loss": 3.1918,
+      "step": 83550
+    },
+    {
+      "epoch": 8.997955010224949,
+      "grad_norm": 0.8582161664962769,
+      "learning_rate": 6.0407283697877384e-05,
+      "loss": 3.1887,
+      "step": 83600
+    },
+    {
+      "epoch": 9.003336562264558,
+      "grad_norm": 0.8494151830673218,
+      "learning_rate": 6.0084042667815966e-05,
+      "loss": 3.1343,
+      "step": 83650
+    },
+    {
+      "epoch": 9.008718114304166,
+      "grad_norm": 0.8479065299034119,
+      "learning_rate": 5.976080163775455e-05,
+      "loss": 3.1253,
+      "step": 83700
+    },
+    {
+      "epoch": 9.014099666343773,
+      "grad_norm": 0.8485672473907471,
+      "learning_rate": 5.9437560607693135e-05,
+      "loss": 3.1436,
+      "step": 83750
+    },
+    {
+      "epoch": 9.019481218383381,
+      "grad_norm": 0.8757447004318237,
+      "learning_rate": 5.9114319577631716e-05,
+      "loss": 3.1274,
+      "step": 83800
+    },
+    {
+      "epoch": 9.02486277042299,
+      "grad_norm": 0.8620525598526001,
+      "learning_rate": 5.8791078547570304e-05,
+      "loss": 3.1237,
+      "step": 83850
+    },
+    {
+      "epoch": 9.030244322462599,
+      "grad_norm": 0.8854354619979858,
+      "learning_rate": 5.8467837517508885e-05,
+      "loss": 3.1347,
+      "step": 83900
+    },
+    {
+      "epoch": 9.035625874502207,
+      "grad_norm": 0.8926529884338379,
+      "learning_rate": 5.8144596487447466e-05,
+      "loss": 3.1083,
+      "step": 83950
+    },
+    {
+      "epoch": 9.041007426541814,
+      "grad_norm": 0.8925355076789856,
+      "learning_rate": 5.7821355457386054e-05,
+      "loss": 3.126,
+      "step": 84000
+    },
+    {
+      "epoch": 9.041007426541814,
+      "eval_accuracy": 0.3923292799649182,
+      "eval_loss": 3.3203017711639404,
+      "eval_runtime": 182.8076,
+      "eval_samples_per_second": 98.524,
+      "eval_steps_per_second": 6.159,
+      "step": 84000
+    },
+    {
+      "epoch": 9.046388978581422,
+      "grad_norm": 0.8431273698806763,
+      "learning_rate": 5.7498114427324635e-05,
+      "loss": 3.1304,
+      "step": 84050
+    },
+    {
+      "epoch": 9.051770530621031,
+      "grad_norm": 0.8672769665718079,
+      "learning_rate": 5.717487339726322e-05,
+      "loss": 3.1198,
+      "step": 84100
+    },
+    {
+      "epoch": 9.05715208266064,
+      "grad_norm": 0.8500916957855225,
+      "learning_rate": 5.6851632367201804e-05,
+      "loss": 3.1346,
+      "step": 84150
+    },
+    {
+      "epoch": 9.062533634700248,
+      "grad_norm": 0.8439621329307556,
+      "learning_rate": 5.6528391337140385e-05,
+      "loss": 3.1345,
+      "step": 84200
+    },
+    {
+      "epoch": 9.067915186739857,
+      "grad_norm": 0.8591272234916687,
+      "learning_rate": 5.620515030707897e-05,
+      "loss": 3.1324,
+      "step": 84250
+    },
+    {
+      "epoch": 9.073296738779463,
+      "grad_norm": 0.8671176433563232,
+      "learning_rate": 5.5881909277017554e-05,
+      "loss": 3.1263,
+      "step": 84300
+    },
+    {
+      "epoch": 9.078678290819072,
+      "grad_norm": 0.8660874962806702,
+      "learning_rate": 5.555866824695614e-05,
+      "loss": 3.1316,
+      "step": 84350
+    },
+    {
+      "epoch": 9.08405984285868,
+      "grad_norm": 0.8906688690185547,
+      "learning_rate": 5.523542721689472e-05,
+      "loss": 3.1293,
+      "step": 84400
+    },
+    {
+      "epoch": 9.089441394898289,
+      "grad_norm": 0.8381503820419312,
+      "learning_rate": 5.4912186186833304e-05,
+      "loss": 3.1144,
+      "step": 84450
+    },
+    {
+      "epoch": 9.094822946937898,
+      "grad_norm": 0.8566581606864929,
+      "learning_rate": 5.45889451567719e-05,
+      "loss": 3.1277,
+      "step": 84500
+    },
+    {
+      "epoch": 9.100204498977504,
+      "grad_norm": 0.8410578966140747,
+      "learning_rate": 5.426570412671048e-05,
+      "loss": 3.1333,
+      "step": 84550
+    },
+    {
+      "epoch": 9.105586051017113,
+      "grad_norm": 0.8373113870620728,
+      "learning_rate": 5.394246309664907e-05,
+      "loss": 3.1343,
+      "step": 84600
+    },
+    {
+      "epoch": 9.110967603056721,
+      "grad_norm": 0.8358331322669983,
+      "learning_rate": 5.361922206658765e-05,
+      "loss": 3.1296,
+      "step": 84650
+    },
+    {
+      "epoch": 9.11634915509633,
+      "grad_norm": 0.8469173908233643,
+      "learning_rate": 5.329598103652623e-05,
+      "loss": 3.1223,
+      "step": 84700
+    },
+    {
+      "epoch": 9.121730707135939,
+      "grad_norm": 0.8513491153717041,
+      "learning_rate": 5.297274000646482e-05,
+      "loss": 3.1239,
+      "step": 84750
+    },
+    {
+      "epoch": 9.127112259175545,
+      "grad_norm": 0.8536253571510315,
+      "learning_rate": 5.26494989764034e-05,
+      "loss": 3.1249,
+      "step": 84800
+    },
+    {
+      "epoch": 9.132493811215154,
+      "grad_norm": 0.9170660376548767,
+      "learning_rate": 5.232625794634199e-05,
+      "loss": 3.1293,
+      "step": 84850
+    },
+    {
+      "epoch": 9.137875363254762,
+      "grad_norm": 0.8490689992904663,
+      "learning_rate": 5.200301691628057e-05,
+      "loss": 3.1382,
+      "step": 84900
+    },
+    {
+      "epoch": 9.143256915294371,
+      "grad_norm": 0.8707136511802673,
+      "learning_rate": 5.167977588621915e-05,
+      "loss": 3.1361,
+      "step": 84950
+    },
+    {
+      "epoch": 9.14863846733398,
+      "grad_norm": 0.8578998446464539,
+      "learning_rate": 5.135653485615774e-05,
+      "loss": 3.1376,
+      "step": 85000
+    },
+    {
+      "epoch": 9.14863846733398,
+      "eval_accuracy": 0.39240283796996855,
+      "eval_loss": 3.3188514709472656,
+      "eval_runtime": 182.6043,
+      "eval_samples_per_second": 98.634,
+      "eval_steps_per_second": 6.166,
+      "step": 85000
+    },
+    {
+      "epoch": 9.154020019373588,
+      "grad_norm": 0.8587227463722229,
+      "learning_rate": 5.103329382609632e-05,
+      "loss": 3.1398,
+      "step": 85050
+    },
+    {
+      "epoch": 9.159401571413195,
+      "grad_norm": 0.8671402931213379,
+      "learning_rate": 5.0710052796034906e-05,
+      "loss": 3.1134,
+      "step": 85100
+    },
+    {
+      "epoch": 9.164783123452803,
+      "grad_norm": 0.8717479109764099,
+      "learning_rate": 5.038681176597349e-05,
+      "loss": 3.116,
+      "step": 85150
+    },
+    {
+      "epoch": 9.170164675492412,
+      "grad_norm": 0.8798679709434509,
+      "learning_rate": 5.006357073591207e-05,
+      "loss": 3.1406,
+      "step": 85200
+    },
+    {
+      "epoch": 9.17554622753202,
+      "grad_norm": 0.9084823727607727,
+      "learning_rate": 4.974032970585066e-05,
+      "loss": 3.144,
+      "step": 85250
+    },
+    {
+      "epoch": 9.180927779571629,
+      "grad_norm": 0.8945959210395813,
+      "learning_rate": 4.9417088675789244e-05,
+      "loss": 3.1382,
+      "step": 85300
+    },
+    {
+      "epoch": 9.186309331611236,
+      "grad_norm": 0.8824502229690552,
+      "learning_rate": 4.909384764572783e-05,
+      "loss": 3.1076,
+      "step": 85350
+    },
+    {
+      "epoch": 9.191690883650844,
+      "grad_norm": 0.8436898589134216,
+      "learning_rate": 4.877060661566641e-05,
+      "loss": 3.1517,
+      "step": 85400
+    },
+    {
+      "epoch": 9.197072435690453,
+      "grad_norm": 0.8343037366867065,
+      "learning_rate": 4.8447365585604994e-05,
+      "loss": 3.1237,
+      "step": 85450
+    },
+    {
+      "epoch": 9.202453987730062,
+      "grad_norm": 0.9071884751319885,
+      "learning_rate": 4.812412455554358e-05,
+      "loss": 3.1321,
+      "step": 85500
+    },
+    {
+      "epoch": 9.20783553976967,
+      "grad_norm": 0.8292504549026489,
+      "learning_rate": 4.780088352548216e-05,
+      "loss": 3.1412,
+      "step": 85550
+    },
+    {
+      "epoch": 9.213217091809279,
+      "grad_norm": 0.8744844794273376,
+      "learning_rate": 4.7484107316021976e-05,
+      "loss": 3.1387,
+      "step": 85600
+    },
+    {
+      "epoch": 9.218598643848885,
+      "grad_norm": 0.8166503310203552,
+      "learning_rate": 4.7160866285960564e-05,
+      "loss": 3.1446,
+      "step": 85650
+    },
+    {
+      "epoch": 9.223980195888494,
+      "grad_norm": 0.8761563897132874,
+      "learning_rate": 4.6837625255899145e-05,
+      "loss": 3.1335,
+      "step": 85700
+    },
+    {
+      "epoch": 9.229361747928102,
+      "grad_norm": 0.9411799907684326,
+      "learning_rate": 4.6514384225837726e-05,
+      "loss": 3.1534,
+      "step": 85750
+    },
+    {
+      "epoch": 9.234743299967711,
+      "grad_norm": 0.9132449626922607,
+      "learning_rate": 4.6191143195776314e-05,
+      "loss": 3.127,
+      "step": 85800
+    },
+    {
+      "epoch": 9.24012485200732,
+      "grad_norm": 0.8338202834129333,
+      "learning_rate": 4.5867902165714895e-05,
+      "loss": 3.129,
+      "step": 85850
+    },
+    {
+      "epoch": 9.245506404046926,
+      "grad_norm": 0.8570502400398254,
+      "learning_rate": 4.554466113565348e-05,
+      "loss": 3.1349,
+      "step": 85900
+    },
+    {
+      "epoch": 9.250887956086535,
+      "grad_norm": 0.8650547862052917,
+      "learning_rate": 4.5221420105592064e-05,
+      "loss": 3.1229,
+      "step": 85950
+    },
+    {
+      "epoch": 9.256269508126143,
+      "grad_norm": 0.8735968470573425,
+      "learning_rate": 4.4898179075530645e-05,
+      "loss": 3.1377,
+      "step": 86000
+    },
+    {
+      "epoch": 9.256269508126143,
+      "eval_accuracy": 0.3926530655706053,
+      "eval_loss": 3.316767692565918,
+      "eval_runtime": 182.8243,
+      "eval_samples_per_second": 98.515,
+      "eval_steps_per_second": 6.159,
+      "step": 86000
+    },
+    {
+      "epoch": 9.261651060165752,
+      "grad_norm": 0.8486570715904236,
+      "learning_rate": 4.457493804546923e-05,
+      "loss": 3.1238,
+      "step": 86050
+    },
+    {
+      "epoch": 9.26703261220536,
+      "grad_norm": 0.8732266426086426,
+      "learning_rate": 4.4251697015407814e-05,
+      "loss": 3.1398,
+      "step": 86100
+    },
+    {
+      "epoch": 9.272414164244967,
+      "grad_norm": 0.8636999726295471,
+      "learning_rate": 4.392845598534641e-05,
+      "loss": 3.1303,
+      "step": 86150
+    },
+    {
+      "epoch": 9.277795716284576,
+      "grad_norm": 0.8534356355667114,
+      "learning_rate": 4.360521495528499e-05,
+      "loss": 3.156,
+      "step": 86200
+    },
+    {
+      "epoch": 9.283177268324184,
+      "grad_norm": 0.8626139163970947,
+      "learning_rate": 4.3281973925223564e-05,
+      "loss": 3.1201,
+      "step": 86250
+    },
+    {
+      "epoch": 9.288558820363793,
+      "grad_norm": 0.8573952317237854,
+      "learning_rate": 4.295873289516216e-05,
+      "loss": 3.1281,
+      "step": 86300
+    },
+    {
+      "epoch": 9.293940372403402,
+      "grad_norm": 0.8308889269828796,
+      "learning_rate": 4.263549186510074e-05,
+      "loss": 3.1453,
+      "step": 86350
+    },
+    {
+      "epoch": 9.29932192444301,
+      "grad_norm": 0.813922643661499,
+      "learning_rate": 4.231225083503933e-05,
+      "loss": 3.1252,
+      "step": 86400
+    },
+    {
+      "epoch": 9.304703476482617,
+      "grad_norm": 0.8177182078361511,
+      "learning_rate": 4.198900980497791e-05,
+      "loss": 3.1361,
+      "step": 86450
+    },
+    {
+      "epoch": 9.310085028522225,
+      "grad_norm": 0.8975428938865662,
+      "learning_rate": 4.167223359551772e-05,
+      "loss": 3.1266,
+      "step": 86500
+    },
+    {
+      "epoch": 9.315466580561834,
+      "grad_norm": 0.8596468567848206,
+      "learning_rate": 4.1348992565456303e-05,
+      "loss": 3.1277,
+      "step": 86550
+    },
+    {
+      "epoch": 9.320848132601443,
+      "grad_norm": 0.8630208969116211,
+      "learning_rate": 4.102575153539489e-05,
+      "loss": 3.1461,
+      "step": 86600
+    },
+    {
+      "epoch": 9.326229684641051,
+      "grad_norm": 0.9110468029975891,
+      "learning_rate": 4.070251050533347e-05,
+      "loss": 3.1253,
+      "step": 86650
+    },
+    {
+      "epoch": 9.331611236680658,
+      "grad_norm": 0.8691954612731934,
+      "learning_rate": 4.037926947527206e-05,
+      "loss": 3.1465,
+      "step": 86700
+    },
+    {
+      "epoch": 9.336992788720266,
+      "grad_norm": 0.8495018482208252,
+      "learning_rate": 4.005602844521064e-05,
+      "loss": 3.129,
+      "step": 86750
+    },
+    {
+      "epoch": 9.342374340759875,
+      "grad_norm": 0.8317940831184387,
+      "learning_rate": 3.973278741514922e-05,
+      "loss": 3.1439,
+      "step": 86800
+    },
+    {
+      "epoch": 9.347755892799483,
+      "grad_norm": 0.8267626166343689,
+      "learning_rate": 3.940954638508781e-05,
+      "loss": 3.1323,
+      "step": 86850
+    },
+    {
+      "epoch": 9.353137444839092,
+      "grad_norm": 0.8663368225097656,
+      "learning_rate": 3.908630535502639e-05,
+      "loss": 3.1397,
+      "step": 86900
+    },
+    {
+      "epoch": 9.3585189968787,
+      "grad_norm": 0.8476428389549255,
+      "learning_rate": 3.876306432496498e-05,
+      "loss": 3.132,
+      "step": 86950
+    },
+    {
+      "epoch": 9.363900548918307,
+      "grad_norm": 0.8747643232345581,
+      "learning_rate": 3.843982329490356e-05,
+      "loss": 3.1389,
+      "step": 87000
+    },
+    {
+      "epoch": 9.363900548918307,
+      "eval_accuracy": 0.3930124893234956,
+      "eval_loss": 3.3138458728790283,
+      "eval_runtime": 182.6258,
+      "eval_samples_per_second": 98.622,
+      "eval_steps_per_second": 6.166,
+      "step": 87000
+    },
+    {
+      "epoch": 9.369282100957916,
+      "grad_norm": 0.922262966632843,
+      "learning_rate": 3.811658226484214e-05,
+      "loss": 3.1331,
+      "step": 87050
+    },
+    {
+      "epoch": 9.374663652997524,
+      "grad_norm": 0.8760948181152344,
+      "learning_rate": 3.7793341234780736e-05,
+      "loss": 3.1383,
+      "step": 87100
+    },
+    {
+      "epoch": 9.380045205037133,
+      "grad_norm": 0.8842629790306091,
+      "learning_rate": 3.747010020471931e-05,
+      "loss": 3.1441,
+      "step": 87150
+    },
+    {
+      "epoch": 9.385426757076742,
+      "grad_norm": 0.885867178440094,
+      "learning_rate": 3.71468591746579e-05,
+      "loss": 3.1412,
+      "step": 87200
+    },
+    {
+      "epoch": 9.390808309116348,
+      "grad_norm": 0.814979076385498,
+      "learning_rate": 3.6823618144596486e-05,
+      "loss": 3.1407,
+      "step": 87250
+    },
+    {
+      "epoch": 9.396189861155957,
+      "grad_norm": 0.8274505734443665,
+      "learning_rate": 3.650037711453507e-05,
+      "loss": 3.1296,
+      "step": 87300
+    },
+    {
+      "epoch": 9.401571413195565,
+      "grad_norm": 0.8518266081809998,
+      "learning_rate": 3.6177136084473655e-05,
+      "loss": 3.115,
+      "step": 87350
+    },
+    {
+      "epoch": 9.406952965235174,
+      "grad_norm": 0.8622663021087646,
+      "learning_rate": 3.5853895054412236e-05,
+      "loss": 3.1405,
+      "step": 87400
+    },
+    {
+      "epoch": 9.412334517274783,
+      "grad_norm": 0.8241939544677734,
+      "learning_rate": 3.553065402435082e-05,
+      "loss": 3.1449,
+      "step": 87450
+    },
+    {
+      "epoch": 9.417716069314391,
+      "grad_norm": 0.8843784928321838,
+      "learning_rate": 3.5207412994289405e-05,
+      "loss": 3.1569,
+      "step": 87500
+    },
+    {
+      "epoch": 9.423097621353998,
+      "grad_norm": 0.906825840473175,
+      "learning_rate": 3.488417196422799e-05,
+      "loss": 3.1392,
+      "step": 87550
+    },
+    {
+      "epoch": 9.428479173393606,
+      "grad_norm": 0.8589437007904053,
+      "learning_rate": 3.4560930934166574e-05,
+      "loss": 3.1291,
+      "step": 87600
+    },
+    {
+      "epoch": 9.433860725433215,
+      "grad_norm": 0.8674544095993042,
+      "learning_rate": 3.4237689904105156e-05,
+      "loss": 3.1476,
+      "step": 87650
+    },
+    {
+      "epoch": 9.439242277472824,
+      "grad_norm": 0.8766100406646729,
+      "learning_rate": 3.3914448874043743e-05,
+      "loss": 3.1407,
+      "step": 87700
+    },
+    {
+      "epoch": 9.444623829512432,
+      "grad_norm": 0.8393718600273132,
+      "learning_rate": 3.3591207843982325e-05,
+      "loss": 3.1521,
+      "step": 87750
+    },
+    {
+      "epoch": 9.450005381552039,
+      "grad_norm": 0.8990697860717773,
+      "learning_rate": 3.326796681392091e-05,
+      "loss": 3.1242,
+      "step": 87800
+    },
+    {
+      "epoch": 9.455386933591647,
+      "grad_norm": 0.8910731077194214,
+      "learning_rate": 3.2944725783859494e-05,
+      "loss": 3.1614,
+      "step": 87850
+    },
+    {
+      "epoch": 9.460768485631256,
+      "grad_norm": 0.8416497111320496,
+      "learning_rate": 3.2621484753798075e-05,
+      "loss": 3.1444,
+      "step": 87900
+    },
+    {
+      "epoch": 9.466150037670864,
+      "grad_norm": 0.8401027321815491,
+      "learning_rate": 3.229824372373666e-05,
+      "loss": 3.1381,
+      "step": 87950
+    },
+    {
+      "epoch": 9.471531589710473,
+      "grad_norm": 0.8736851215362549,
+      "learning_rate": 3.197500269367525e-05,
+      "loss": 3.1323,
+      "step": 88000
+    },
+    {
+      "epoch": 9.471531589710473,
+      "eval_accuracy": 0.39324837474294083,
+      "eval_loss": 3.312713384628296,
+      "eval_runtime": 182.6668,
+      "eval_samples_per_second": 98.6,
+      "eval_steps_per_second": 6.164,
+      "step": 88000
+    },
+    {
+      "epoch": 9.476913141750082,
+      "grad_norm": 0.8908987045288086,
+      "learning_rate": 3.165176166361383e-05,
+      "loss": 3.1391,
+      "step": 88050
+    },
+    {
+      "epoch": 9.482294693789688,
+      "grad_norm": 0.8490933179855347,
+      "learning_rate": 3.132852063355242e-05,
+      "loss": 3.1512,
+      "step": 88100
+    },
+    {
+      "epoch": 9.487676245829297,
+      "grad_norm": 0.9195178151130676,
+      "learning_rate": 3.1005279603491e-05,
+      "loss": 3.1404,
+      "step": 88150
+    },
+    {
+      "epoch": 9.493057797868905,
+      "grad_norm": 0.8848748803138733,
+      "learning_rate": 3.068203857342958e-05,
+      "loss": 3.149,
+      "step": 88200
+    },
+    {
+      "epoch": 9.498439349908514,
+      "grad_norm": 0.86981201171875,
+      "learning_rate": 3.035879754336817e-05,
+      "loss": 3.1165,
+      "step": 88250
+    },
+    {
+      "epoch": 9.503820901948123,
+      "grad_norm": 0.9147974848747253,
+      "learning_rate": 3.0035556513306754e-05,
+      "loss": 3.1424,
+      "step": 88300
+    },
+    {
+      "epoch": 9.50920245398773,
+      "grad_norm": 0.8942081332206726,
+      "learning_rate": 2.971231548324534e-05,
+      "loss": 3.1351,
+      "step": 88350
+    },
+    {
+      "epoch": 9.514584006027338,
+      "grad_norm": 0.8392196893692017,
+      "learning_rate": 2.938907445318392e-05,
+      "loss": 3.1381,
+      "step": 88400
+    },
+    {
+      "epoch": 9.519965558066946,
+      "grad_norm": 0.8911840915679932,
+      "learning_rate": 2.9065833423122504e-05,
+      "loss": 3.1417,
+      "step": 88450
+    },
+    {
+      "epoch": 9.525347110106555,
+      "grad_norm": 0.8524636626243591,
+      "learning_rate": 2.874259239306109e-05,
+      "loss": 3.1324,
+      "step": 88500
+    },
+    {
+      "epoch": 9.530728662146164,
+      "grad_norm": 0.8815006017684937,
+      "learning_rate": 2.8419351362999673e-05,
+      "loss": 3.1403,
+      "step": 88550
+    },
+    {
+      "epoch": 9.536110214185772,
+      "grad_norm": 0.8821752667427063,
+      "learning_rate": 2.809611033293826e-05,
+      "loss": 3.1315,
+      "step": 88600
+    },
+    {
+      "epoch": 9.541491766225379,
+      "grad_norm": 0.9411100745201111,
+      "learning_rate": 2.7772869302876845e-05,
+      "loss": 3.1516,
+      "step": 88650
+    },
+    {
+      "epoch": 9.546873318264987,
+      "grad_norm": 0.851620078086853,
+      "learning_rate": 2.7449628272815427e-05,
+      "loss": 3.1355,
+      "step": 88700
+    },
+    {
+      "epoch": 9.552254870304596,
+      "grad_norm": 0.8492972254753113,
+      "learning_rate": 2.712638724275401e-05,
+      "loss": 3.1426,
+      "step": 88750
+    },
+    {
+      "epoch": 9.557636422344205,
+      "grad_norm": 0.8770946860313416,
+      "learning_rate": 2.6803146212692596e-05,
+      "loss": 3.1521,
+      "step": 88800
+    },
+    {
+      "epoch": 9.563017974383813,
+      "grad_norm": 0.8662365078926086,
+      "learning_rate": 2.647990518263118e-05,
+      "loss": 3.1437,
+      "step": 88850
+    },
+    {
+      "epoch": 9.56839952642342,
+      "grad_norm": 0.8437976837158203,
+      "learning_rate": 2.6156664152569765e-05,
+      "loss": 3.1468,
+      "step": 88900
+    },
+    {
+      "epoch": 9.573781078463028,
+      "grad_norm": 0.8942129015922546,
+      "learning_rate": 2.5833423122508346e-05,
+      "loss": 3.1457,
+      "step": 88950
+    },
+    {
+      "epoch": 9.579162630502637,
+      "grad_norm": 0.9466322064399719,
+      "learning_rate": 2.5510182092446934e-05,
+      "loss": 3.1385,
+      "step": 89000
+    },
+    {
+      "epoch": 9.579162630502637,
+      "eval_accuracy": 0.39364148089105366,
+      "eval_loss": 3.309281587600708,
+      "eval_runtime": 182.7896,
+      "eval_samples_per_second": 98.534,
+      "eval_steps_per_second": 6.16,
+      "step": 89000
+    },
+    {
+      "epoch": 9.584544182542245,
+      "grad_norm": 0.9119086265563965,
+      "learning_rate": 2.5186941062385518e-05,
+      "loss": 3.1391,
+      "step": 89050
+    },
+    {
+      "epoch": 9.589925734581854,
+      "grad_norm": 0.9008535146713257,
+      "learning_rate": 2.4863700032324103e-05,
+      "loss": 3.1422,
+      "step": 89100
+    },
+    {
+      "epoch": 9.59530728662146,
+      "grad_norm": 0.8503806591033936,
+      "learning_rate": 2.4540459002262687e-05,
+      "loss": 3.151,
+      "step": 89150
+    },
+    {
+      "epoch": 9.60068883866107,
+      "grad_norm": 0.9381389617919922,
+      "learning_rate": 2.4217217972201268e-05,
+      "loss": 3.1439,
+      "step": 89200
+    },
+    {
+      "epoch": 9.606070390700678,
+      "grad_norm": 0.8869844079017639,
+      "learning_rate": 2.3893976942139853e-05,
+      "loss": 3.1347,
+      "step": 89250
+    },
+    {
+      "epoch": 9.611451942740286,
+      "grad_norm": 0.8782930374145508,
+      "learning_rate": 2.3570735912078437e-05,
+      "loss": 3.1562,
+      "step": 89300
+    },
+    {
+      "epoch": 9.616833494779895,
+      "grad_norm": 0.9496018290519714,
+      "learning_rate": 2.324749488201702e-05,
+      "loss": 3.1202,
+      "step": 89350
+    },
+    {
+      "epoch": 9.622215046819504,
+      "grad_norm": 0.8809753060340881,
+      "learning_rate": 2.292425385195561e-05,
+      "loss": 3.136,
+      "step": 89400
+    },
+    {
+      "epoch": 9.62759659885911,
+      "grad_norm": 0.8729392290115356,
+      "learning_rate": 2.260101282189419e-05,
+      "loss": 3.1371,
+      "step": 89450
+    },
+    {
+      "epoch": 9.632978150898719,
+      "grad_norm": 0.881260871887207,
+      "learning_rate": 2.2277771791832775e-05,
+      "loss": 3.1288,
+      "step": 89500
+    },
+    {
+      "epoch": 9.638359702938327,
+      "grad_norm": 0.8699634671211243,
+      "learning_rate": 2.195453076177136e-05,
+      "loss": 3.1467,
+      "step": 89550
+    },
+    {
+      "epoch": 9.643741254977936,
+      "grad_norm": 0.8865028619766235,
+      "learning_rate": 2.1631289731709944e-05,
+      "loss": 3.133,
+      "step": 89600
+    },
+    {
+      "epoch": 9.649122807017545,
+      "grad_norm": 0.8839409351348877,
+      "learning_rate": 2.130804870164853e-05,
+      "loss": 3.1223,
+      "step": 89650
+    },
+    {
+      "epoch": 9.654504359057151,
+      "grad_norm": 0.9245901703834534,
+      "learning_rate": 2.098480767158711e-05,
+      "loss": 3.1552,
+      "step": 89700
+    },
+    {
+      "epoch": 9.65988591109676,
+      "grad_norm": 0.8770548105239868,
+      "learning_rate": 2.0661566641525694e-05,
+      "loss": 3.131,
+      "step": 89750
+    },
+    {
+      "epoch": 9.665267463136368,
+      "grad_norm": 0.8477882146835327,
+      "learning_rate": 2.0338325611464282e-05,
+      "loss": 3.1385,
+      "step": 89800
+    },
+    {
+      "epoch": 9.670649015175977,
+      "grad_norm": 0.8711922764778137,
+      "learning_rate": 2.0015084581402867e-05,
+      "loss": 3.146,
+      "step": 89850
+    },
+    {
+      "epoch": 9.676030567215586,
+      "grad_norm": 0.8607136607170105,
+      "learning_rate": 1.969184355134145e-05,
+      "loss": 3.1414,
+      "step": 89900
+    },
+    {
+      "epoch": 9.681412119255192,
+      "grad_norm": 0.8953176736831665,
+      "learning_rate": 1.9368602521280032e-05,
+      "loss": 3.1304,
+      "step": 89950
+    },
+    {
+      "epoch": 9.6867936712948,
+      "grad_norm": 0.8567274212837219,
+      "learning_rate": 1.9045361491218617e-05,
+      "loss": 3.1358,
+      "step": 90000
+    },
+    {
+      "epoch": 9.6867936712948,
+      "eval_accuracy": 0.3937655624889378,
+      "eval_loss": 3.3076210021972656,
+      "eval_runtime": 182.9314,
+      "eval_samples_per_second": 98.458,
+      "eval_steps_per_second": 6.155,
+      "step": 90000
+    },
+    {
+      "epoch": 9.69217522333441,
+      "grad_norm": 0.8794586062431335,
+      "learning_rate": 1.87221204611572e-05,
+      "loss": 3.1459,
+      "step": 90050
+    },
+    {
+      "epoch": 9.697556775374018,
+      "grad_norm": 0.8762345314025879,
+      "learning_rate": 1.8398879431095786e-05,
+      "loss": 3.1327,
+      "step": 90100
+    },
+    {
+      "epoch": 9.702938327413626,
+      "grad_norm": 0.9184940457344055,
+      "learning_rate": 1.807563840103437e-05,
+      "loss": 3.1297,
+      "step": 90150
+    },
+    {
+      "epoch": 9.708319879453235,
+      "grad_norm": 0.8900254964828491,
+      "learning_rate": 1.7752397370972955e-05,
+      "loss": 3.1323,
+      "step": 90200
+    },
+    {
+      "epoch": 9.713701431492842,
+      "grad_norm": 0.8552128076553345,
+      "learning_rate": 1.742915634091154e-05,
+      "loss": 3.1414,
+      "step": 90250
+    },
+    {
+      "epoch": 9.71908298353245,
+      "grad_norm": 0.84034264087677,
+      "learning_rate": 1.7105915310850124e-05,
+      "loss": 3.1346,
+      "step": 90300
+    },
+    {
+      "epoch": 9.724464535572059,
+      "grad_norm": 0.8617538213729858,
+      "learning_rate": 1.6782674280788708e-05,
+      "loss": 3.1243,
+      "step": 90350
+    },
+    {
+      "epoch": 9.729846087611667,
+      "grad_norm": 0.9199374318122864,
+      "learning_rate": 1.645943325072729e-05,
+      "loss": 3.1422,
+      "step": 90400
+    },
+    {
+      "epoch": 9.735227639651276,
+      "grad_norm": 0.8754702210426331,
+      "learning_rate": 1.6136192220665877e-05,
+      "loss": 3.1472,
+      "step": 90450
+    },
+    {
+      "epoch": 9.740609191690883,
+      "grad_norm": 0.9364842772483826,
+      "learning_rate": 1.5819416011205687e-05,
+      "loss": 3.1309,
+      "step": 90500
+    },
+    {
+      "epoch": 9.745990743730491,
+      "grad_norm": 0.8289980292320251,
+      "learning_rate": 1.549617498114427e-05,
+      "loss": 3.1472,
+      "step": 90550
+    },
+    {
+      "epoch": 9.7513722957701,
+      "grad_norm": 0.8661265969276428,
+      "learning_rate": 1.5172933951082856e-05,
+      "loss": 3.1371,
+      "step": 90600
+    },
+    {
+      "epoch": 9.756753847809708,
+      "grad_norm": 0.8728821873664856,
+      "learning_rate": 1.4849692921021439e-05,
+      "loss": 3.1438,
+      "step": 90650
+    },
+    {
+      "epoch": 9.762135399849317,
+      "grad_norm": 0.8967341780662537,
+      "learning_rate": 1.4526451890960025e-05,
+      "loss": 3.1279,
+      "step": 90700
+    },
+    {
+      "epoch": 9.767516951888926,
+      "grad_norm": 0.8679084181785583,
+      "learning_rate": 1.420321086089861e-05,
+      "loss": 3.1178,
+      "step": 90750
+    },
+    {
+      "epoch": 9.772898503928532,
+      "grad_norm": 0.8503843545913696,
+      "learning_rate": 1.3879969830837192e-05,
+      "loss": 3.1446,
+      "step": 90800
+    },
+    {
+      "epoch": 9.77828005596814,
+      "grad_norm": 0.918580949306488,
+      "learning_rate": 1.3556728800775778e-05,
+      "loss": 3.13,
+      "step": 90850
+    },
+    {
+      "epoch": 9.78366160800775,
+      "grad_norm": 0.8387547135353088,
+      "learning_rate": 1.3233487770714361e-05,
+      "loss": 3.1473,
+      "step": 90900
+    },
+    {
+      "epoch": 9.789043160047358,
+      "grad_norm": 0.8897818326950073,
+      "learning_rate": 1.2910246740652946e-05,
+      "loss": 3.1323,
+      "step": 90950
+    },
+    {
+      "epoch": 9.794424712086967,
+      "grad_norm": 0.8836583495140076,
+      "learning_rate": 1.258700571059153e-05,
+      "loss": 3.1303,
+      "step": 91000
+    },
+    {
+      "epoch": 9.794424712086967,
+      "eval_accuracy": 0.3940082930403153,
+      "eval_loss": 3.3052875995635986,
+      "eval_runtime": 182.9499,
+      "eval_samples_per_second": 98.448,
+      "eval_steps_per_second": 6.155,
+      "step": 91000
+    },
+    {
+      "epoch": 9.799806264126573,
+      "grad_norm": 0.8632960319519043,
+      "learning_rate": 1.2263764680530113e-05,
+      "loss": 3.1429,
+      "step": 91050
+    },
+    {
+      "epoch": 9.805187816166182,
+      "grad_norm": 0.8691814541816711,
+      "learning_rate": 1.19405236504687e-05,
+      "loss": 3.1271,
+      "step": 91100
+    },
+    {
+      "epoch": 9.81056936820579,
+      "grad_norm": 0.8277129530906677,
+      "learning_rate": 1.1617282620407282e-05,
+      "loss": 3.1307,
+      "step": 91150
+    },
+    {
+      "epoch": 9.815950920245399,
+      "grad_norm": 0.8648361563682556,
+      "learning_rate": 1.1294041590345867e-05,
+      "loss": 3.1471,
+      "step": 91200
+    },
+    {
+      "epoch": 9.821332472285007,
+      "grad_norm": 0.8396468758583069,
+      "learning_rate": 1.0970800560284453e-05,
+      "loss": 3.1444,
+      "step": 91250
+    },
+    {
+      "epoch": 9.826714024324616,
+      "grad_norm": 0.8713970184326172,
+      "learning_rate": 1.0647559530223035e-05,
+      "loss": 3.1379,
+      "step": 91300
+    },
+    {
+      "epoch": 9.832095576364223,
+      "grad_norm": 0.8426269292831421,
+      "learning_rate": 1.032431850016162e-05,
+      "loss": 3.1476,
+      "step": 91350
+    },
+    {
+      "epoch": 9.837477128403831,
+      "grad_norm": 0.9147149324417114,
+      "learning_rate": 1.0001077470100203e-05,
+      "loss": 3.1378,
+      "step": 91400
+    },
+    {
+      "epoch": 9.84285868044344,
+      "grad_norm": 0.8848015069961548,
+      "learning_rate": 9.677836440038787e-06,
+      "loss": 3.1497,
+      "step": 91450
+    },
+    {
+      "epoch": 9.848240232483048,
+      "grad_norm": 0.9187755584716797,
+      "learning_rate": 9.354595409977372e-06,
+      "loss": 3.1406,
+      "step": 91500
+    },
+    {
+      "epoch": 9.853621784522657,
+      "grad_norm": 0.8338111639022827,
+      "learning_rate": 9.031354379915956e-06,
+      "loss": 3.1277,
+      "step": 91550
+    },
+    {
+      "epoch": 9.859003336562264,
+      "grad_norm": 0.880181074142456,
+      "learning_rate": 8.70811334985454e-06,
+      "loss": 3.1482,
+      "step": 91600
+    },
+    {
+      "epoch": 9.864384888601872,
+      "grad_norm": 0.8883482217788696,
+      "learning_rate": 8.384872319793125e-06,
+      "loss": 3.1256,
+      "step": 91650
+    },
+    {
+      "epoch": 9.869766440641481,
+      "grad_norm": 0.8639237880706787,
+      "learning_rate": 8.06163128973171e-06,
+      "loss": 3.1349,
+      "step": 91700
+    },
+    {
+      "epoch": 9.87514799268109,
+      "grad_norm": 0.8801106810569763,
+      "learning_rate": 7.738390259670293e-06,
+      "loss": 3.1469,
+      "step": 91750
+    },
+    {
+      "epoch": 9.880529544720698,
+      "grad_norm": 0.9389435648918152,
+      "learning_rate": 7.415149229608878e-06,
+      "loss": 3.1366,
+      "step": 91800
+    },
+    {
+      "epoch": 9.885911096760307,
+      "grad_norm": 0.8664288520812988,
+      "learning_rate": 7.091908199547462e-06,
+      "loss": 3.1287,
+      "step": 91850
+    },
+    {
+      "epoch": 9.891292648799913,
+      "grad_norm": 0.8594654202461243,
+      "learning_rate": 6.768667169486046e-06,
+      "loss": 3.1367,
+      "step": 91900
+    },
+    {
+      "epoch": 9.896674200839522,
+      "grad_norm": 0.852776288986206,
+      "learning_rate": 6.4454261394246305e-06,
+      "loss": 3.15,
+      "step": 91950
+    },
+    {
+      "epoch": 9.90205575287913,
+      "grad_norm": 0.871873676776886,
+      "learning_rate": 6.122185109363214e-06,
+      "loss": 3.1412,
+      "step": 92000
+    },
+    {
+      "epoch": 9.90205575287913,
+      "eval_accuracy": 0.3942072364779036,
+      "eval_loss": 3.303832769393921,
+      "eval_runtime": 183.128,
+      "eval_samples_per_second": 98.352,
+      "eval_steps_per_second": 6.149,
+      "step": 92000
+    },
+    {
+      "epoch": 9.907437304918739,
+      "grad_norm": 0.8564249277114868,
+      "learning_rate": 5.7989440793017995e-06,
+      "loss": 3.1272,
+      "step": 92050
+    },
+    {
+      "epoch": 9.912818856958348,
+      "grad_norm": 0.8567243814468384,
+      "learning_rate": 5.475703049240383e-06,
+      "loss": 3.1432,
+      "step": 92100
+    },
+    {
+      "epoch": 9.918200408997954,
+      "grad_norm": 0.8726806044578552,
+      "learning_rate": 5.152462019178968e-06,
+      "loss": 3.1318,
+      "step": 92150
+    },
+    {
+      "epoch": 9.923581961037563,
+      "grad_norm": 0.881909191608429,
+      "learning_rate": 4.829220989117551e-06,
+      "loss": 3.1186,
+      "step": 92200
+    },
+    {
+      "epoch": 9.928963513077171,
+      "grad_norm": 0.8314427137374878,
+      "learning_rate": 4.505979959056136e-06,
+      "loss": 3.1237,
+      "step": 92250
+    },
+    {
+      "epoch": 9.93434506511678,
+      "grad_norm": 0.8411848545074463,
+      "learning_rate": 4.18273892899472e-06,
+      "loss": 3.1282,
+      "step": 92300
+    },
+    {
+      "epoch": 9.939726617156388,
+      "grad_norm": 0.9116420745849609,
+      "learning_rate": 3.859497898933305e-06,
+      "loss": 3.1365,
+      "step": 92350
+    },
+    {
+      "epoch": 9.945108169195997,
+      "grad_norm": 0.8482275009155273,
+      "learning_rate": 3.5362568688718884e-06,
+      "loss": 3.1371,
+      "step": 92400
+    },
+    {
+      "epoch": 9.950489721235604,
+      "grad_norm": 0.8526927828788757,
+      "learning_rate": 3.2194806594117013e-06,
+      "loss": 3.1521,
+      "step": 92450
+    },
+    {
+      "epoch": 9.955871273275212,
+      "grad_norm": 0.8895639777183533,
+      "learning_rate": 2.8962396293502854e-06,
+      "loss": 3.1432,
+      "step": 92500
+    },
+    {
+      "epoch": 9.961252825314821,
+      "grad_norm": 0.8692976832389832,
+      "learning_rate": 2.5729985992888694e-06,
+      "loss": 3.148,
+      "step": 92550
+    },
+    {
+      "epoch": 9.96663437735443,
+      "grad_norm": 0.8431194424629211,
+      "learning_rate": 2.249757569227454e-06,
+      "loss": 3.1493,
+      "step": 92600
+    },
+    {
+      "epoch": 9.972015929394038,
+      "grad_norm": 0.8579978346824646,
+      "learning_rate": 1.926516539166038e-06,
+      "loss": 3.1416,
+      "step": 92650
+    },
+    {
+      "epoch": 9.977397481433645,
+      "grad_norm": 0.8592678308486938,
+      "learning_rate": 1.603275509104622e-06,
+      "loss": 3.1405,
+      "step": 92700
+    },
+    {
+      "epoch": 9.982779033473253,
+      "grad_norm": 0.8317117691040039,
+      "learning_rate": 1.2800344790432064e-06,
+      "loss": 3.1257,
+      "step": 92750
+    },
+    {
+      "epoch": 9.988160585512862,
+      "grad_norm": 0.8965446352958679,
+      "learning_rate": 9.567934489817906e-07,
+      "loss": 3.1337,
+      "step": 92800
+    },
+    {
+      "epoch": 9.99354213755247,
+      "grad_norm": 0.8839283585548401,
+      "learning_rate": 6.335524189203748e-07,
+      "loss": 3.1165,
+      "step": 92850
+    },
+    {
+      "epoch": 9.998923689592079,
+      "grad_norm": 0.8802606463432312,
+      "learning_rate": 3.103113888589591e-07,
+      "loss": 3.1291,
+      "step": 92900
+    },
+    {
+      "epoch": 10.0,
+      "step": 92910,
+      "total_flos": 7.7681859821568e+17,
+      "train_loss": 3.460850506426251,
+      "train_runtime": 80822.0047,
+      "train_samples_per_second": 36.784,
+      "train_steps_per_second": 1.15
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 92910,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.7681859821568e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}