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": 3.307018518447876,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M_1208/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": 1.6288660764694214,
+      "learning_rate": 0.000276,
+      "loss": 8.759,
+      "step": 50
+    },
+    {
+      "epoch": 0.010763104079216447,
+      "grad_norm": 8.2901611328125,
+      "learning_rate": 0.0005759999999999999,
+      "loss": 7.0168,
+      "step": 100
+    },
+    {
+      "epoch": 0.01614465611882467,
+      "grad_norm": 1.3438377380371094,
+      "learning_rate": 0.0005997026182523434,
+      "loss": 6.6009,
+      "step": 150
+    },
+    {
+      "epoch": 0.021526208158432893,
+      "grad_norm": 1.8783726692199707,
+      "learning_rate": 0.000599379377222282,
+      "loss": 6.3134,
+      "step": 200
+    },
+    {
+      "epoch": 0.026907760198041114,
+      "grad_norm": 1.0426092147827148,
+      "learning_rate": 0.0005990561361922206,
+      "loss": 6.1356,
+      "step": 250
+    },
+    {
+      "epoch": 0.03228931223764934,
+      "grad_norm": 1.021669864654541,
+      "learning_rate": 0.0005987328951621592,
+      "loss": 6.024,
+      "step": 300
+    },
+    {
+      "epoch": 0.03767086427725756,
+      "grad_norm": 1.6005719900131226,
+      "learning_rate": 0.0005984096541320978,
+      "loss": 5.9093,
+      "step": 350
+    },
+    {
+      "epoch": 0.04305241631686579,
+      "grad_norm": 1.0307564735412598,
+      "learning_rate": 0.0005980864131020364,
+      "loss": 5.8065,
+      "step": 400
+    },
+    {
+      "epoch": 0.048433968356474004,
+      "grad_norm": 1.7200894355773926,
+      "learning_rate": 0.0005977631720719749,
+      "loss": 5.7429,
+      "step": 450
+    },
+    {
+      "epoch": 0.05381552039608223,
+      "grad_norm": 1.1326223611831665,
+      "learning_rate": 0.0005974399310419136,
+      "loss": 5.6593,
+      "step": 500
+    },
+    {
+      "epoch": 0.05919707243569045,
+      "grad_norm": 1.3042094707489014,
+      "learning_rate": 0.0005971166900118521,
+      "loss": 5.587,
+      "step": 550
+    },
+    {
+      "epoch": 0.06457862447529868,
+      "grad_norm": 0.9970183372497559,
+      "learning_rate": 0.0005967934489817907,
+      "loss": 5.5242,
+      "step": 600
+    },
+    {
+      "epoch": 0.0699601765149069,
+      "grad_norm": 0.8722121119499207,
+      "learning_rate": 0.0005964702079517293,
+      "loss": 5.4456,
+      "step": 650
+    },
+    {
+      "epoch": 0.07534172855451512,
+      "grad_norm": 1.1744099855422974,
+      "learning_rate": 0.0005961469669216679,
+      "loss": 5.409,
+      "step": 700
+    },
+    {
+      "epoch": 0.08072328059412334,
+      "grad_norm": 1.2246752977371216,
+      "learning_rate": 0.0005958237258916065,
+      "loss": 5.3477,
+      "step": 750
+    },
+    {
+      "epoch": 0.08610483263373157,
+      "grad_norm": 1.0104173421859741,
+      "learning_rate": 0.000595500484861545,
+      "loss": 5.2821,
+      "step": 800
+    },
+    {
+      "epoch": 0.09148638467333979,
+      "grad_norm": 0.8681843876838684,
+      "learning_rate": 0.0005951772438314836,
+      "loss": 5.217,
+      "step": 850
+    },
+    {
+      "epoch": 0.09686793671294801,
+      "grad_norm": 1.285335898399353,
+      "learning_rate": 0.0005948540028014222,
+      "loss": 5.164,
+      "step": 900
+    },
+    {
+      "epoch": 0.10224948875255624,
+      "grad_norm": 1.4134410619735718,
+      "learning_rate": 0.0005945307617713608,
+      "loss": 5.1207,
+      "step": 950
+    },
+    {
+      "epoch": 0.10763104079216446,
+      "grad_norm": 1.1055577993392944,
+      "learning_rate": 0.0005942075207412994,
+      "loss": 5.081,
+      "step": 1000
+    },
+    {
+      "epoch": 0.10763104079216446,
+      "eval_accuracy": 0.2257252879057058,
+      "eval_loss": 5.034178256988525,
+      "eval_runtime": 185.9242,
+      "eval_samples_per_second": 96.873,
+      "eval_steps_per_second": 6.056,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11301259283177269,
+      "grad_norm": 1.1473057270050049,
+      "learning_rate": 0.0005938842797112379,
+      "loss": 5.051,
+      "step": 1050
+    },
+    {
+      "epoch": 0.1183941448713809,
+      "grad_norm": 1.2688162326812744,
+      "learning_rate": 0.0005935610386811766,
+      "loss": 5.0342,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12377569691098914,
+      "grad_norm": 1.471405029296875,
+      "learning_rate": 0.0005932377976511151,
+      "loss": 4.9927,
+      "step": 1150
+    },
+    {
+      "epoch": 0.12915724895059735,
+      "grad_norm": 1.2830071449279785,
+      "learning_rate": 0.0005929145566210538,
+      "loss": 4.9706,
+      "step": 1200
+    },
+    {
+      "epoch": 0.13453880099020557,
+      "grad_norm": 1.042227029800415,
+      "learning_rate": 0.0005925913155909923,
+      "loss": 4.9304,
+      "step": 1250
+    },
+    {
+      "epoch": 0.1399203530298138,
+      "grad_norm": 1.1886645555496216,
+      "learning_rate": 0.0005922680745609308,
+      "loss": 4.91,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14530190506942203,
+      "grad_norm": 1.1165701150894165,
+      "learning_rate": 0.0005919448335308695,
+      "loss": 4.8792,
+      "step": 1350
+    },
+    {
+      "epoch": 0.15068345710903025,
+      "grad_norm": 1.2330048084259033,
+      "learning_rate": 0.000591621592500808,
+      "loss": 4.8518,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15606500914863847,
+      "grad_norm": 1.231316089630127,
+      "learning_rate": 0.0005912983514707467,
+      "loss": 4.8285,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16144656118824668,
+      "grad_norm": 1.064316749572754,
+      "learning_rate": 0.0005909751104406852,
+      "loss": 4.8077,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1668281132278549,
+      "grad_norm": 0.7813340425491333,
+      "learning_rate": 0.0005906518694106238,
+      "loss": 4.7872,
+      "step": 1550
+    },
+    {
+      "epoch": 0.17220966526746315,
+      "grad_norm": 0.8941433429718018,
+      "learning_rate": 0.0005903286283805624,
+      "loss": 4.7446,
+      "step": 1600
+    },
+    {
+      "epoch": 0.17759121730707136,
+      "grad_norm": 0.838981568813324,
+      "learning_rate": 0.000590005387350501,
+      "loss": 4.7486,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18297276934667958,
+      "grad_norm": 1.0024216175079346,
+      "learning_rate": 0.0005896821463204395,
+      "loss": 4.7169,
+      "step": 1700
+    },
+    {
+      "epoch": 0.1883543213862878,
+      "grad_norm": 1.0789501667022705,
+      "learning_rate": 0.0005893589052903781,
+      "loss": 4.6844,
+      "step": 1750
+    },
+    {
+      "epoch": 0.19373587342589602,
+      "grad_norm": 0.9427714347839355,
+      "learning_rate": 0.0005890356642603167,
+      "loss": 4.7106,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19911742546550426,
+      "grad_norm": 1.179642677307129,
+      "learning_rate": 0.0005887124232302553,
+      "loss": 4.6463,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20449897750511248,
+      "grad_norm": 0.905066192150116,
+      "learning_rate": 0.0005883891822001939,
+      "loss": 4.6574,
+      "step": 1900
+    },
+    {
+      "epoch": 0.2098805295447207,
+      "grad_norm": 1.0582997798919678,
+      "learning_rate": 0.0005880659411701324,
+      "loss": 4.5862,
+      "step": 1950
+    },
+    {
+      "epoch": 0.2152620815843289,
+      "grad_norm": 0.9089963436126709,
+      "learning_rate": 0.0005877427001400711,
+      "loss": 4.5993,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2152620815843289,
+      "eval_accuracy": 0.26972286233863824,
+      "eval_loss": 4.515867233276367,
+      "eval_runtime": 185.6561,
+      "eval_samples_per_second": 97.013,
+      "eval_steps_per_second": 6.065,
+      "step": 2000
+    },
+    {
+      "epoch": 0.22064363362393713,
+      "grad_norm": 0.8239651918411255,
+      "learning_rate": 0.0005874194591100097,
+      "loss": 4.5754,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22602518566354537,
+      "grad_norm": 0.9209245443344116,
+      "learning_rate": 0.0005870962180799483,
+      "loss": 4.527,
+      "step": 2100
+    },
+    {
+      "epoch": 0.2314067377031536,
+      "grad_norm": 0.7858166694641113,
+      "learning_rate": 0.0005867729770498868,
+      "loss": 4.5283,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2367882897427618,
+      "grad_norm": 0.8506563901901245,
+      "learning_rate": 0.0005864497360198253,
+      "loss": 4.5017,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24216984178237003,
+      "grad_norm": 0.903475284576416,
+      "learning_rate": 0.000586126494989764,
+      "loss": 4.5116,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24755139382197827,
+      "grad_norm": 0.8945424556732178,
+      "learning_rate": 0.0005858032539597025,
+      "loss": 4.5007,
+      "step": 2300
+    },
+    {
+      "epoch": 0.2529329458615865,
+      "grad_norm": 0.6636348366737366,
+      "learning_rate": 0.0005854800129296412,
+      "loss": 4.4701,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2583144979011947,
+      "grad_norm": 1.0701271295547485,
+      "learning_rate": 0.0005851567718995797,
+      "loss": 4.4403,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2636960499408029,
+      "grad_norm": 1.0419641733169556,
+      "learning_rate": 0.0005848335308695184,
+      "loss": 4.4431,
+      "step": 2450
+    },
+    {
+      "epoch": 0.26907760198041114,
+      "grad_norm": 0.9337476491928101,
+      "learning_rate": 0.0005845102898394569,
+      "loss": 4.4188,
+      "step": 2500
+    },
+    {
+      "epoch": 0.27445915402001936,
+      "grad_norm": 0.9284446835517883,
+      "learning_rate": 0.0005841870488093954,
+      "loss": 4.3951,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2798407060596276,
+      "grad_norm": 1.0020089149475098,
+      "learning_rate": 0.0005838638077793341,
+      "loss": 4.3854,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2852222580992358,
+      "grad_norm": 0.8462651371955872,
+      "learning_rate": 0.0005835405667492726,
+      "loss": 4.379,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29060381013884407,
+      "grad_norm": 0.8200508952140808,
+      "learning_rate": 0.0005832173257192113,
+      "loss": 4.3916,
+      "step": 2700
+    },
+    {
+      "epoch": 0.2959853621784523,
+      "grad_norm": 0.9552410840988159,
+      "learning_rate": 0.0005828940846891498,
+      "loss": 4.3722,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3013669142180605,
+      "grad_norm": 1.0096497535705566,
+      "learning_rate": 0.0005825708436590884,
+      "loss": 4.3725,
+      "step": 2800
+    },
+    {
+      "epoch": 0.3067484662576687,
+      "grad_norm": 0.9831181168556213,
+      "learning_rate": 0.000582247602629027,
+      "loss": 4.3666,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31213001829727693,
+      "grad_norm": 1.0847501754760742,
+      "learning_rate": 0.0005819243615989656,
+      "loss": 4.3497,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31751157033688515,
+      "grad_norm": 0.7545516490936279,
+      "learning_rate": 0.0005816011205689042,
+      "loss": 4.3086,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32289312237649337,
+      "grad_norm": 1.1041558980941772,
+      "learning_rate": 0.0005812778795388427,
+      "loss": 4.2969,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32289312237649337,
+      "eval_accuracy": 0.29862963718301744,
+      "eval_loss": 4.24181604385376,
+      "eval_runtime": 185.692,
+      "eval_samples_per_second": 96.994,
+      "eval_steps_per_second": 6.064,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3282746744161016,
+      "grad_norm": 1.3289992809295654,
+      "learning_rate": 0.0005809546385087813,
+      "loss": 4.3089,
+      "step": 3050
+    },
+    {
+      "epoch": 0.3336562264557098,
+      "grad_norm": 0.8491426110267639,
+      "learning_rate": 0.0005806313974787199,
+      "loss": 4.2836,
+      "step": 3100
+    },
+    {
+      "epoch": 0.3390377784953181,
+      "grad_norm": 0.7015953660011292,
+      "learning_rate": 0.0005803081564486585,
+      "loss": 4.2754,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3444193305349263,
+      "grad_norm": 0.7845497131347656,
+      "learning_rate": 0.0005799849154185971,
+      "loss": 4.2675,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3498008825745345,
+      "grad_norm": 0.7819005250930786,
+      "learning_rate": 0.0005796616743885357,
+      "loss": 4.281,
+      "step": 3250
+    },
+    {
+      "epoch": 0.35518243461414273,
+      "grad_norm": 0.7934615015983582,
+      "learning_rate": 0.0005793384333584742,
+      "loss": 4.253,
+      "step": 3300
+    },
+    {
+      "epoch": 0.36056398665375095,
+      "grad_norm": 0.8142829537391663,
+      "learning_rate": 0.0005790151923284129,
+      "loss": 4.2436,
+      "step": 3350
+    },
+    {
+      "epoch": 0.36594553869335916,
+      "grad_norm": 0.875823438167572,
+      "learning_rate": 0.0005786919512983514,
+      "loss": 4.2468,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3713270907329674,
+      "grad_norm": 0.816535472869873,
+      "learning_rate": 0.00057836871026829,
+      "loss": 4.2693,
+      "step": 3450
+    },
+    {
+      "epoch": 0.3767086427725756,
+      "grad_norm": 0.7482356429100037,
+      "learning_rate": 0.0005780454692382286,
+      "loss": 4.2367,
+      "step": 3500
+    },
+    {
+      "epoch": 0.3820901948121838,
+      "grad_norm": 0.8074214458465576,
+      "learning_rate": 0.0005777222282081672,
+      "loss": 4.2383,
+      "step": 3550
+    },
+    {
+      "epoch": 0.38747174685179203,
+      "grad_norm": 1.1420109272003174,
+      "learning_rate": 0.0005773989871781058,
+      "loss": 4.2114,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3928532988914003,
+      "grad_norm": 0.9202913641929626,
+      "learning_rate": 0.0005770757461480443,
+      "loss": 4.2323,
+      "step": 3650
+    },
+    {
+      "epoch": 0.3982348509310085,
+      "grad_norm": 0.6997593641281128,
+      "learning_rate": 0.0005767525051179829,
+      "loss": 4.2006,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40361640297061674,
+      "grad_norm": 0.6690722703933716,
+      "learning_rate": 0.0005764292640879215,
+      "loss": 4.1984,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40899795501022496,
+      "grad_norm": 0.7391257882118225,
+      "learning_rate": 0.0005761060230578601,
+      "loss": 4.1973,
+      "step": 3800
+    },
+    {
+      "epoch": 0.4143795070498332,
+      "grad_norm": 0.7893011569976807,
+      "learning_rate": 0.0005757827820277987,
+      "loss": 4.1819,
+      "step": 3850
+    },
+    {
+      "epoch": 0.4197610590894414,
+      "grad_norm": 0.792421817779541,
+      "learning_rate": 0.0005754595409977372,
+      "loss": 4.1899,
+      "step": 3900
+    },
+    {
+      "epoch": 0.4251426111290496,
+      "grad_norm": 0.7149876952171326,
+      "learning_rate": 0.0005751362999676759,
+      "loss": 4.1678,
+      "step": 3950
+    },
+    {
+      "epoch": 0.4305241631686578,
+      "grad_norm": 0.6952950358390808,
+      "learning_rate": 0.0005748130589376144,
+      "loss": 4.1676,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4305241631686578,
+      "eval_accuracy": 0.3120642408026145,
+      "eval_loss": 4.0956950187683105,
+      "eval_runtime": 185.4003,
+      "eval_samples_per_second": 97.147,
+      "eval_steps_per_second": 6.073,
+      "step": 4000
+    },
+    {
+      "epoch": 0.43590571520826604,
+      "grad_norm": 0.7889477014541626,
+      "learning_rate": 0.0005744898179075531,
+      "loss": 4.1673,
+      "step": 4050
+    },
+    {
+      "epoch": 0.44128726724787426,
+      "grad_norm": 0.8199712038040161,
+      "learning_rate": 0.0005741665768774916,
+      "loss": 4.1462,
+      "step": 4100
+    },
+    {
+      "epoch": 0.44666881928748253,
+      "grad_norm": 0.6692017912864685,
+      "learning_rate": 0.0005738433358474301,
+      "loss": 4.1454,
+      "step": 4150
+    },
+    {
+      "epoch": 0.45205037132709075,
+      "grad_norm": 0.7322275042533875,
+      "learning_rate": 0.0005735200948173688,
+      "loss": 4.1325,
+      "step": 4200
+    },
+    {
+      "epoch": 0.45743192336669897,
+      "grad_norm": 0.8394356369972229,
+      "learning_rate": 0.0005731968537873073,
+      "loss": 4.1269,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4628134754063072,
+      "grad_norm": 0.6714458465576172,
+      "learning_rate": 0.0005728736127572459,
+      "loss": 4.1348,
+      "step": 4300
+    },
+    {
+      "epoch": 0.4681950274459154,
+      "grad_norm": 0.6486440300941467,
+      "learning_rate": 0.0005725503717271845,
+      "loss": 4.1298,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4735765794855236,
+      "grad_norm": 0.7148385643959045,
+      "learning_rate": 0.0005722271306971231,
+      "loss": 4.1185,
+      "step": 4400
+    },
+    {
+      "epoch": 0.47895813152513184,
+      "grad_norm": 0.7207582592964172,
+      "learning_rate": 0.0005719038896670617,
+      "loss": 4.1166,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48433968356474005,
+      "grad_norm": 0.6919761896133423,
+      "learning_rate": 0.0005715806486370003,
+      "loss": 4.1272,
+      "step": 4500
+    },
+    {
+      "epoch": 0.48972123560434827,
+      "grad_norm": 0.6202664375305176,
+      "learning_rate": 0.0005712574076069388,
+      "loss": 4.1198,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49510278764395654,
+      "grad_norm": 0.6014512777328491,
+      "learning_rate": 0.0005709341665768774,
+      "loss": 4.1071,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5004843396835648,
+      "grad_norm": 0.6499958634376526,
+      "learning_rate": 0.0005706109255468161,
+      "loss": 4.0796,
+      "step": 4650
+    },
+    {
+      "epoch": 0.505865891723173,
+      "grad_norm": 0.7313818335533142,
+      "learning_rate": 0.0005702876845167546,
+      "loss": 4.0998,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5112474437627812,
+      "grad_norm": 0.6985046863555908,
+      "learning_rate": 0.0005699644434866932,
+      "loss": 4.0918,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5166289958023894,
+      "grad_norm": 0.5906838178634644,
+      "learning_rate": 0.0005696412024566317,
+      "loss": 4.0921,
+      "step": 4800
+    },
+    {
+      "epoch": 0.5220105478419976,
+      "grad_norm": 0.8641616106033325,
+      "learning_rate": 0.0005693179614265704,
+      "loss": 4.0702,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5273920998816058,
+      "grad_norm": 0.7014225721359253,
+      "learning_rate": 0.000568994720396509,
+      "loss": 4.0881,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5327736519212141,
+      "grad_norm": 0.6952133774757385,
+      "learning_rate": 0.0005686714793664476,
+      "loss": 4.0599,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5381552039608223,
+      "grad_norm": 0.75461745262146,
+      "learning_rate": 0.0005683482383363861,
+      "loss": 4.0642,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5381552039608223,
+      "eval_accuracy": 0.3205846916121383,
+      "eval_loss": 4.000431060791016,
+      "eval_runtime": 185.3844,
+      "eval_samples_per_second": 97.155,
+      "eval_steps_per_second": 6.074,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5435367560004305,
+      "grad_norm": 0.6787461042404175,
+      "learning_rate": 0.0005680249973063247,
+      "loss": 4.0693,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5489183080400387,
+      "grad_norm": 0.7117706537246704,
+      "learning_rate": 0.0005677017562762633,
+      "loss": 4.056,
+      "step": 5100
+    },
+    {
+      "epoch": 0.5542998600796469,
+      "grad_norm": 0.6522045135498047,
+      "learning_rate": 0.0005673785152462018,
+      "loss": 4.039,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5596814121192552,
+      "grad_norm": 0.6263989806175232,
+      "learning_rate": 0.0005670552742161405,
+      "loss": 4.0767,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5650629641588634,
+      "grad_norm": 0.7219600081443787,
+      "learning_rate": 0.000566732033186079,
+      "loss": 4.0563,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5704445161984716,
+      "grad_norm": 0.6352048516273499,
+      "learning_rate": 0.0005664087921560177,
+      "loss": 4.0474,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5758260682380799,
+      "grad_norm": 0.6418537497520447,
+      "learning_rate": 0.0005660855511259562,
+      "loss": 4.0516,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5812076202776881,
+      "grad_norm": 0.7082635164260864,
+      "learning_rate": 0.0005657623100958947,
+      "loss": 4.0268,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5865891723172963,
+      "grad_norm": 0.6689324378967285,
+      "learning_rate": 0.0005654390690658334,
+      "loss": 4.024,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5919707243569046,
+      "grad_norm": 0.5856683850288391,
+      "learning_rate": 0.0005651158280357719,
+      "loss": 4.0459,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5973522763965128,
+      "grad_norm": 0.6519558429718018,
+      "learning_rate": 0.0005647925870057106,
+      "loss": 4.0324,
+      "step": 5550
+    },
+    {
+      "epoch": 0.602733828436121,
+      "grad_norm": 0.6687760353088379,
+      "learning_rate": 0.0005644693459756491,
+      "loss": 4.035,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6081153804757292,
+      "grad_norm": 0.6159760355949402,
+      "learning_rate": 0.0005641461049455877,
+      "loss": 4.0104,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.6416439414024353,
+      "learning_rate": 0.0005638228639155263,
+      "loss": 4.0014,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6188784845549457,
+      "grad_norm": 0.7194835543632507,
+      "learning_rate": 0.000563499622885465,
+      "loss": 4.0274,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6242600365945539,
+      "grad_norm": 0.5478219389915466,
+      "learning_rate": 0.0005631763818554035,
+      "loss": 4.0012,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6296415886341621,
+      "grad_norm": 1.0783581733703613,
+      "learning_rate": 0.000562853140825342,
+      "loss": 3.9946,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6350231406737703,
+      "grad_norm": 0.6307653188705444,
+      "learning_rate": 0.0005625298997952806,
+      "loss": 3.9929,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6404046927133785,
+      "grad_norm": 0.5995813608169556,
+      "learning_rate": 0.0005622066587652192,
+      "loss": 3.994,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6457862447529867,
+      "grad_norm": 0.6409510970115662,
+      "learning_rate": 0.0005618834177351578,
+      "loss": 4.0038,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6457862447529867,
+      "eval_accuracy": 0.3277830540827355,
+      "eval_loss": 3.9248440265655518,
+      "eval_runtime": 185.5096,
+      "eval_samples_per_second": 97.089,
+      "eval_steps_per_second": 6.07,
+      "step": 6000
+    },
+    {
+      "epoch": 0.651167796792595,
+      "grad_norm": 0.6729882955551147,
+      "learning_rate": 0.0005615601767050964,
+      "loss": 3.9844,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6565493488322032,
+      "grad_norm": 0.6977255940437317,
+      "learning_rate": 0.000561236935675035,
+      "loss": 3.9991,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6619309008718114,
+      "grad_norm": 0.5876686573028564,
+      "learning_rate": 0.0005609136946449735,
+      "loss": 3.9916,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6673124529114196,
+      "grad_norm": 0.6658395528793335,
+      "learning_rate": 0.0005605904536149122,
+      "loss": 3.9722,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6726940049510278,
+      "grad_norm": 0.6371570825576782,
+      "learning_rate": 0.0005602672125848507,
+      "loss": 3.9791,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6780755569906362,
+      "grad_norm": 0.6571714282035828,
+      "learning_rate": 0.0005599439715547892,
+      "loss": 3.9892,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6834571090302444,
+      "grad_norm": 0.5972535610198975,
+      "learning_rate": 0.0005596207305247279,
+      "loss": 3.9656,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6888386610698526,
+      "grad_norm": 0.6958330273628235,
+      "learning_rate": 0.0005592974894946665,
+      "loss": 3.9722,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6942202131094608,
+      "grad_norm": 0.6522466540336609,
+      "learning_rate": 0.0005589742484646051,
+      "loss": 3.9688,
+      "step": 6450
+    },
+    {
+      "epoch": 0.699601765149069,
+      "grad_norm": 0.6060879826545715,
+      "learning_rate": 0.0005586510074345436,
+      "loss": 3.9568,
+      "step": 6500
+    },
+    {
+      "epoch": 0.7049833171886772,
+      "grad_norm": 0.5448448657989502,
+      "learning_rate": 0.0005583277664044822,
+      "loss": 3.9701,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7103648692282855,
+      "grad_norm": 0.6579198837280273,
+      "learning_rate": 0.0005580045253744208,
+      "loss": 3.9474,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7157464212678937,
+      "grad_norm": 0.7126155495643616,
+      "learning_rate": 0.0005576812843443594,
+      "loss": 3.9606,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7211279733075019,
+      "grad_norm": 0.5947892665863037,
+      "learning_rate": 0.000557358043314298,
+      "loss": 3.9326,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7265095253471101,
+      "grad_norm": 0.5703105330467224,
+      "learning_rate": 0.0005570348022842365,
+      "loss": 3.9387,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7318910773867183,
+      "grad_norm": 0.6345952153205872,
+      "learning_rate": 0.0005567115612541752,
+      "loss": 3.9528,
+      "step": 6800
+    },
+    {
+      "epoch": 0.7372726294263265,
+      "grad_norm": 0.6352825164794922,
+      "learning_rate": 0.0005563883202241137,
+      "loss": 3.9621,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7426541814659348,
+      "grad_norm": 0.605293333530426,
+      "learning_rate": 0.0005560650791940524,
+      "loss": 3.9476,
+      "step": 6900
+    },
+    {
+      "epoch": 0.748035733505543,
+      "grad_norm": 0.629027247428894,
+      "learning_rate": 0.0005557418381639909,
+      "loss": 3.9496,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7534172855451512,
+      "grad_norm": 0.5393297672271729,
+      "learning_rate": 0.0005554185971339294,
+      "loss": 3.9256,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7534172855451512,
+      "eval_accuracy": 0.3326138701278551,
+      "eval_loss": 3.8684041500091553,
+      "eval_runtime": 185.676,
+      "eval_samples_per_second": 97.002,
+      "eval_steps_per_second": 6.064,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7587988375847594,
+      "grad_norm": 0.6152430772781372,
+      "learning_rate": 0.0005550953561038681,
+      "loss": 3.9355,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7641803896243676,
+      "grad_norm": 0.5815154910087585,
+      "learning_rate": 0.0005547721150738066,
+      "loss": 3.9276,
+      "step": 7100
+    },
+    {
+      "epoch": 0.7695619416639758,
+      "grad_norm": 0.5519508123397827,
+      "learning_rate": 0.0005544488740437452,
+      "loss": 3.9182,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7749434937035841,
+      "grad_norm": 0.5207048654556274,
+      "learning_rate": 0.0005541256330136838,
+      "loss": 3.9456,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7803250457431924,
+      "grad_norm": 0.670373797416687,
+      "learning_rate": 0.0005538023919836224,
+      "loss": 3.9291,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7857065977828006,
+      "grad_norm": 0.5639854073524475,
+      "learning_rate": 0.000553479150953561,
+      "loss": 3.9201,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7910881498224088,
+      "grad_norm": 0.6556904315948486,
+      "learning_rate": 0.0005531559099234996,
+      "loss": 3.9347,
+      "step": 7350
+    },
+    {
+      "epoch": 0.796469701862017,
+      "grad_norm": 0.5908870100975037,
+      "learning_rate": 0.0005528326688934381,
+      "loss": 3.9179,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8018512539016253,
+      "grad_norm": 0.5798718333244324,
+      "learning_rate": 0.0005525094278633767,
+      "loss": 3.9148,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8072328059412335,
+      "grad_norm": 0.6623952388763428,
+      "learning_rate": 0.0005521861868333154,
+      "loss": 3.9189,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8126143579808417,
+      "grad_norm": 0.7111433744430542,
+      "learning_rate": 0.0005518629458032539,
+      "loss": 3.8909,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8179959100204499,
+      "grad_norm": 0.6083288192749023,
+      "learning_rate": 0.0005515397047731925,
+      "loss": 3.9062,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8233774620600581,
+      "grad_norm": 0.6324785351753235,
+      "learning_rate": 0.000551216463743131,
+      "loss": 3.8885,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8287590140996663,
+      "grad_norm": 0.6778486371040344,
+      "learning_rate": 0.0005508932227130697,
+      "loss": 3.9184,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8341405661392746,
+      "grad_norm": 0.5886597037315369,
+      "learning_rate": 0.0005505699816830083,
+      "loss": 3.8933,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8395221181788828,
+      "grad_norm": 0.6386252641677856,
+      "learning_rate": 0.0005502467406529469,
+      "loss": 3.9005,
+      "step": 7800
+    },
+    {
+      "epoch": 0.844903670218491,
+      "grad_norm": 0.581160306930542,
+      "learning_rate": 0.0005499234996228854,
+      "loss": 3.8939,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8502852222580992,
+      "grad_norm": 0.6273126006126404,
+      "learning_rate": 0.000549600258592824,
+      "loss": 3.8867,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8556667742977074,
+      "grad_norm": 0.6495140790939331,
+      "learning_rate": 0.0005492770175627626,
+      "loss": 3.9011,
+      "step": 7950
+    },
+    {
+      "epoch": 0.8610483263373157,
+      "grad_norm": 0.6411253213882446,
+      "learning_rate": 0.0005489537765327011,
+      "loss": 3.8889,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8610483263373157,
+      "eval_accuracy": 0.33802684834588476,
+      "eval_loss": 3.8203883171081543,
+      "eval_runtime": 185.6274,
+      "eval_samples_per_second": 97.028,
+      "eval_steps_per_second": 6.066,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8664298783769239,
+      "grad_norm": 0.5747044086456299,
+      "learning_rate": 0.0005486305355026398,
+      "loss": 3.8982,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8718114304165321,
+      "grad_norm": 0.545897364616394,
+      "learning_rate": 0.0005483072944725783,
+      "loss": 3.8823,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8771929824561403,
+      "grad_norm": 0.6497473120689392,
+      "learning_rate": 0.000547984053442517,
+      "loss": 3.8708,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8825745344957485,
+      "grad_norm": 0.5869677066802979,
+      "learning_rate": 0.0005476608124124555,
+      "loss": 3.8915,
+      "step": 8200
+    },
+    {
+      "epoch": 0.8879560865353568,
+      "grad_norm": 0.6175825595855713,
+      "learning_rate": 0.000547337571382394,
+      "loss": 3.8862,
+      "step": 8250
+    },
+    {
+      "epoch": 0.8933376385749651,
+      "grad_norm": 0.6012660264968872,
+      "learning_rate": 0.0005470143303523327,
+      "loss": 3.8825,
+      "step": 8300
+    },
+    {
+      "epoch": 0.8987191906145733,
+      "grad_norm": 0.5652008652687073,
+      "learning_rate": 0.0005466910893222712,
+      "loss": 3.8759,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9041007426541815,
+      "grad_norm": 0.5548220872879028,
+      "learning_rate": 0.0005463678482922099,
+      "loss": 3.8608,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9094822946937897,
+      "grad_norm": 0.6385374069213867,
+      "learning_rate": 0.0005460446072621484,
+      "loss": 3.8615,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9148638467333979,
+      "grad_norm": 0.660474419593811,
+      "learning_rate": 0.000545721366232087,
+      "loss": 3.8685,
+      "step": 8500
+    },
+    {
+      "epoch": 0.9202453987730062,
+      "grad_norm": 0.6575385928153992,
+      "learning_rate": 0.0005453981252020256,
+      "loss": 3.8541,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9256269508126144,
+      "grad_norm": 0.5555447936058044,
+      "learning_rate": 0.0005450748841719643,
+      "loss": 3.877,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9310085028522226,
+      "grad_norm": 0.6523729562759399,
+      "learning_rate": 0.0005447516431419028,
+      "loss": 3.8495,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9363900548918308,
+      "grad_norm": 0.5873777270317078,
+      "learning_rate": 0.0005444284021118413,
+      "loss": 3.8549,
+      "step": 8700
+    },
+    {
+      "epoch": 0.941771606931439,
+      "grad_norm": 0.5743074417114258,
+      "learning_rate": 0.0005441051610817799,
+      "loss": 3.8675,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9471531589710472,
+      "grad_norm": 0.6072008013725281,
+      "learning_rate": 0.0005437819200517185,
+      "loss": 3.8561,
+      "step": 8800
+    },
+    {
+      "epoch": 0.9525347110106555,
+      "grad_norm": 0.5983343124389648,
+      "learning_rate": 0.0005434586790216571,
+      "loss": 3.8494,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9579162630502637,
+      "grad_norm": 0.6041466593742371,
+      "learning_rate": 0.0005431354379915957,
+      "loss": 3.8476,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9632978150898719,
+      "grad_norm": 0.5659998059272766,
+      "learning_rate": 0.0005428121969615343,
+      "loss": 3.8525,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9686793671294801,
+      "grad_norm": 0.6054639220237732,
+      "learning_rate": 0.0005424889559314729,
+      "loss": 3.8508,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9686793671294801,
+      "eval_accuracy": 0.34104316116450256,
+      "eval_loss": 3.7828726768493652,
+      "eval_runtime": 185.41,
+      "eval_samples_per_second": 97.141,
+      "eval_steps_per_second": 6.073,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9740609191690883,
+      "grad_norm": 0.5490997433662415,
+      "learning_rate": 0.0005421657149014114,
+      "loss": 3.8243,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9794424712086965,
+      "grad_norm": 0.5536127090454102,
+      "learning_rate": 0.00054184247387135,
+      "loss": 3.8392,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9848240232483048,
+      "grad_norm": 0.6085931658744812,
+      "learning_rate": 0.0005415192328412885,
+      "loss": 3.8462,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9902055752879131,
+      "grad_norm": 0.6172447800636292,
+      "learning_rate": 0.0005411959918112272,
+      "loss": 3.8469,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9955871273275213,
+      "grad_norm": 0.589678943157196,
+      "learning_rate": 0.0005408727507811658,
+      "loss": 3.8478,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0009686793671295,
+      "grad_norm": 0.5819297432899475,
+      "learning_rate": 0.0005405495097511044,
+      "loss": 3.8261,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0063502314067376,
+      "grad_norm": 0.6677573323249817,
+      "learning_rate": 0.0005402262687210429,
+      "loss": 3.7952,
+      "step": 9350
+    },
+    {
+      "epoch": 1.011731783446346,
+      "grad_norm": 0.6139549016952515,
+      "learning_rate": 0.0005399030276909816,
+      "loss": 3.786,
+      "step": 9400
+    },
+    {
+      "epoch": 1.017113335485954,
+      "grad_norm": 0.5575100183486938,
+      "learning_rate": 0.0005395797866609201,
+      "loss": 3.7745,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0224948875255624,
+      "grad_norm": 0.5324420928955078,
+      "learning_rate": 0.0005392565456308587,
+      "loss": 3.7737,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0278764395651705,
+      "grad_norm": 0.5157521367073059,
+      "learning_rate": 0.0005389333046007973,
+      "loss": 3.767,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0332579916047788,
+      "grad_norm": 0.5273111462593079,
+      "learning_rate": 0.0005386100635707358,
+      "loss": 3.782,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0386395436443872,
+      "grad_norm": 0.6705301403999329,
+      "learning_rate": 0.0005382868225406745,
+      "loss": 3.7519,
+      "step": 9650
+    },
+    {
+      "epoch": 1.0440210956839953,
+      "grad_norm": 0.6219047904014587,
+      "learning_rate": 0.000537963581510613,
+      "loss": 3.762,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0494026477236036,
+      "grad_norm": 0.644806981086731,
+      "learning_rate": 0.0005376403404805517,
+      "loss": 3.7834,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0547841997632117,
+      "grad_norm": 0.6287768483161926,
+      "learning_rate": 0.0005373170994504902,
+      "loss": 3.7877,
+      "step": 9800
+    },
+    {
+      "epoch": 1.06016575180282,
+      "grad_norm": 0.5888068079948425,
+      "learning_rate": 0.0005369938584204287,
+      "loss": 3.7641,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0655473038424281,
+      "grad_norm": 0.5658965110778809,
+      "learning_rate": 0.0005366706173903674,
+      "loss": 3.7561,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0709288558820365,
+      "grad_norm": 0.5976064801216125,
+      "learning_rate": 0.0005363473763603059,
+      "loss": 3.7677,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0763104079216446,
+      "grad_norm": 0.6668145060539246,
+      "learning_rate": 0.0005360241353302445,
+      "loss": 3.7758,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0763104079216446,
+      "eval_accuracy": 0.3436720177851739,
+      "eval_loss": 3.7506744861602783,
+      "eval_runtime": 185.8788,
+      "eval_samples_per_second": 96.896,
+      "eval_steps_per_second": 6.058,
+      "step": 10000
+    },
+    {
+      "epoch": 1.081691959961253,
+      "grad_norm": 0.6088933348655701,
+      "learning_rate": 0.0005357008943001831,
+      "loss": 3.7735,
+      "step": 10050
+    },
+    {
+      "epoch": 1.087073512000861,
+      "grad_norm": 0.5470409989356995,
+      "learning_rate": 0.0005353776532701217,
+      "loss": 3.7779,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0924550640404693,
+      "grad_norm": 0.5765193700790405,
+      "learning_rate": 0.0005350544122400603,
+      "loss": 3.7704,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0978366160800774,
+      "grad_norm": 0.6331609487533569,
+      "learning_rate": 0.0005347311712099989,
+      "loss": 3.7748,
+      "step": 10200
+    },
+    {
+      "epoch": 1.1032181681196858,
+      "grad_norm": 0.5721049308776855,
+      "learning_rate": 0.0005344079301799374,
+      "loss": 3.7828,
+      "step": 10250
+    },
+    {
+      "epoch": 1.1085997201592939,
+      "grad_norm": 0.6166802048683167,
+      "learning_rate": 0.000534084689149876,
+      "loss": 3.787,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1139812721989022,
+      "grad_norm": 0.7366760969161987,
+      "learning_rate": 0.0005337614481198147,
+      "loss": 3.7451,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1193628242385103,
+      "grad_norm": 0.5707156658172607,
+      "learning_rate": 0.0005334382070897532,
+      "loss": 3.7546,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1247443762781186,
+      "grad_norm": 0.5790042281150818,
+      "learning_rate": 0.0005331149660596918,
+      "loss": 3.7697,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1301259283177267,
+      "grad_norm": 0.5638990998268127,
+      "learning_rate": 0.0005327917250296303,
+      "loss": 3.7433,
+      "step": 10500
+    },
+    {
+      "epoch": 1.135507480357335,
+      "grad_norm": 0.858950674533844,
+      "learning_rate": 0.000532468483999569,
+      "loss": 3.751,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1408890323969434,
+      "grad_norm": 0.6017559766769409,
+      "learning_rate": 0.0005321452429695076,
+      "loss": 3.7675,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1462705844365515,
+      "grad_norm": 0.5325124859809875,
+      "learning_rate": 0.0005318220019394462,
+      "loss": 3.7637,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1516521364761596,
+      "grad_norm": 0.6113856434822083,
+      "learning_rate": 0.0005314987609093847,
+      "loss": 3.7691,
+      "step": 10700
+    },
+    {
+      "epoch": 1.157033688515768,
+      "grad_norm": 0.5921090841293335,
+      "learning_rate": 0.0005311755198793233,
+      "loss": 3.7483,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1624152405553763,
+      "grad_norm": 0.5955826044082642,
+      "learning_rate": 0.0005308522788492619,
+      "loss": 3.7757,
+      "step": 10800
+    },
+    {
+      "epoch": 1.1677967925949844,
+      "grad_norm": 0.5726268291473389,
+      "learning_rate": 0.0005305290378192004,
+      "loss": 3.7494,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1731783446345927,
+      "grad_norm": 0.7779279351234436,
+      "learning_rate": 0.0005302057967891391,
+      "loss": 3.7592,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1785598966742008,
+      "grad_norm": 0.539819598197937,
+      "learning_rate": 0.0005298825557590776,
+      "loss": 3.7457,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1839414487138091,
+      "grad_norm": 0.5312587022781372,
+      "learning_rate": 0.0005295593147290163,
+      "loss": 3.7585,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1839414487138091,
+      "eval_accuracy": 0.34616266793254524,
+      "eval_loss": 3.728276252746582,
+      "eval_runtime": 185.4986,
+      "eval_samples_per_second": 97.095,
+      "eval_steps_per_second": 6.07,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1893230007534172,
+      "grad_norm": 0.5671839118003845,
+      "learning_rate": 0.0005292360736989548,
+      "loss": 3.755,
+      "step": 11050
+    },
+    {
+      "epoch": 1.1947045527930256,
+      "grad_norm": 0.7498019933700562,
+      "learning_rate": 0.0005289128326688933,
+      "loss": 3.7819,
+      "step": 11100
+    },
+    {
+      "epoch": 1.2000861048326337,
+      "grad_norm": 0.6134175658226013,
+      "learning_rate": 0.000528589591638832,
+      "loss": 3.7623,
+      "step": 11150
+    },
+    {
+      "epoch": 1.205467656872242,
+      "grad_norm": 0.5958433747291565,
+      "learning_rate": 0.0005282663506087705,
+      "loss": 3.7607,
+      "step": 11200
+    },
+    {
+      "epoch": 1.21084920891185,
+      "grad_norm": 0.5764477252960205,
+      "learning_rate": 0.0005279431095787092,
+      "loss": 3.75,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2162307609514584,
+      "grad_norm": 0.6179909706115723,
+      "learning_rate": 0.0005276198685486477,
+      "loss": 3.7357,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2216123129910665,
+      "grad_norm": 0.6080899238586426,
+      "learning_rate": 0.0005272966275185863,
+      "loss": 3.7309,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2269938650306749,
+      "grad_norm": 0.5407081842422485,
+      "learning_rate": 0.0005269733864885249,
+      "loss": 3.7506,
+      "step": 11400
+    },
+    {
+      "epoch": 1.232375417070283,
+      "grad_norm": 0.5968424677848816,
+      "learning_rate": 0.0005266501454584636,
+      "loss": 3.74,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2377569691098913,
+      "grad_norm": 0.5409317016601562,
+      "learning_rate": 0.0005263269044284021,
+      "loss": 3.7471,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2431385211494996,
+      "grad_norm": 0.5300924777984619,
+      "learning_rate": 0.0005260036633983406,
+      "loss": 3.7419,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2485200731891077,
+      "grad_norm": 0.5615507960319519,
+      "learning_rate": 0.0005256804223682792,
+      "loss": 3.7444,
+      "step": 11600
+    },
+    {
+      "epoch": 1.2539016252287158,
+      "grad_norm": 0.5873222351074219,
+      "learning_rate": 0.0005253571813382178,
+      "loss": 3.7409,
+      "step": 11650
+    },
+    {
+      "epoch": 1.2592831772683242,
+      "grad_norm": 0.5365596413612366,
+      "learning_rate": 0.0005250339403081564,
+      "loss": 3.7514,
+      "step": 11700
+    },
+    {
+      "epoch": 1.2646647293079325,
+      "grad_norm": 0.5360502004623413,
+      "learning_rate": 0.000524710699278095,
+      "loss": 3.7524,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2700462813475406,
+      "grad_norm": 0.5745640993118286,
+      "learning_rate": 0.0005243874582480336,
+      "loss": 3.7194,
+      "step": 11800
+    },
+    {
+      "epoch": 1.275427833387149,
+      "grad_norm": 0.5777296423912048,
+      "learning_rate": 0.0005240642172179722,
+      "loss": 3.7302,
+      "step": 11850
+    },
+    {
+      "epoch": 1.280809385426757,
+      "grad_norm": 0.5392170548439026,
+      "learning_rate": 0.0005237409761879107,
+      "loss": 3.7295,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2861909374663654,
+      "grad_norm": 0.5685666799545288,
+      "learning_rate": 0.0005234177351578493,
+      "loss": 3.7478,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2915724895059735,
+      "grad_norm": 0.5274864435195923,
+      "learning_rate": 0.0005230944941277878,
+      "loss": 3.7348,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2915724895059735,
+      "eval_accuracy": 0.34907706434239305,
+      "eval_loss": 3.6997456550598145,
+      "eval_runtime": 185.8407,
+      "eval_samples_per_second": 96.916,
+      "eval_steps_per_second": 6.059,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2969540415455818,
+      "grad_norm": 0.5861959457397461,
+      "learning_rate": 0.0005227712530977265,
+      "loss": 3.7457,
+      "step": 12050
+    },
+    {
+      "epoch": 1.30233559358519,
+      "grad_norm": 0.6341837048530579,
+      "learning_rate": 0.0005224480120676651,
+      "loss": 3.7246,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3077171456247982,
+      "grad_norm": 0.555399477481842,
+      "learning_rate": 0.0005221247710376037,
+      "loss": 3.7246,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3130986976644063,
+      "grad_norm": 0.5940098762512207,
+      "learning_rate": 0.0005218015300075422,
+      "loss": 3.7392,
+      "step": 12200
+    },
+    {
+      "epoch": 1.3184802497040147,
+      "grad_norm": 0.567061185836792,
+      "learning_rate": 0.0005214782889774809,
+      "loss": 3.7269,
+      "step": 12250
+    },
+    {
+      "epoch": 1.3238618017436228,
+      "grad_norm": 0.5938940048217773,
+      "learning_rate": 0.0005211550479474194,
+      "loss": 3.7178,
+      "step": 12300
+    },
+    {
+      "epoch": 1.329243353783231,
+      "grad_norm": 0.5960969924926758,
+      "learning_rate": 0.000520831806917358,
+      "loss": 3.7543,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3346249058228392,
+      "grad_norm": 0.6383436322212219,
+      "learning_rate": 0.0005205085658872966,
+      "loss": 3.7231,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3400064578624475,
+      "grad_norm": 0.5827093720436096,
+      "learning_rate": 0.0005201853248572351,
+      "loss": 3.7422,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3453880099020559,
+      "grad_norm": 0.6017980575561523,
+      "learning_rate": 0.0005198620838271738,
+      "loss": 3.7215,
+      "step": 12500
+    },
+    {
+      "epoch": 1.350769561941664,
+      "grad_norm": 0.578345000743866,
+      "learning_rate": 0.0005195388427971123,
+      "loss": 3.7248,
+      "step": 12550
+    },
+    {
+      "epoch": 1.356151113981272,
+      "grad_norm": 0.5523024797439575,
+      "learning_rate": 0.000519215601767051,
+      "loss": 3.7439,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3615326660208804,
+      "grad_norm": 0.6129439473152161,
+      "learning_rate": 0.0005188923607369895,
+      "loss": 3.7299,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3669142180604887,
+      "grad_norm": 0.5610187649726868,
+      "learning_rate": 0.000518569119706928,
+      "loss": 3.7104,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3722957701000968,
+      "grad_norm": 0.569215714931488,
+      "learning_rate": 0.0005182458786768667,
+      "loss": 3.7232,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3776773221397052,
+      "grad_norm": 0.5174156427383423,
+      "learning_rate": 0.0005179226376468052,
+      "loss": 3.7279,
+      "step": 12800
+    },
+    {
+      "epoch": 1.3830588741793133,
+      "grad_norm": 0.5434882044792175,
+      "learning_rate": 0.0005175993966167438,
+      "loss": 3.7047,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3884404262189216,
+      "grad_norm": 0.5322872996330261,
+      "learning_rate": 0.0005172761555866824,
+      "loss": 3.6972,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3938219782585297,
+      "grad_norm": 0.587121307849884,
+      "learning_rate": 0.0005169529145566211,
+      "loss": 3.741,
+      "step": 12950
+    },
+    {
+      "epoch": 1.399203530298138,
+      "grad_norm": 0.605234682559967,
+      "learning_rate": 0.0005166296735265596,
+      "loss": 3.7217,
+      "step": 13000
+    },
+    {
+      "epoch": 1.399203530298138,
+      "eval_accuracy": 0.351000437762485,
+      "eval_loss": 3.6810786724090576,
+      "eval_runtime": 185.658,
+      "eval_samples_per_second": 97.012,
+      "eval_steps_per_second": 6.065,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4045850823377461,
+      "grad_norm": 0.6180337071418762,
+      "learning_rate": 0.0005163064324964982,
+      "loss": 3.6971,
+      "step": 13050
+    },
+    {
+      "epoch": 1.4099666343773545,
+      "grad_norm": 0.5797611474990845,
+      "learning_rate": 0.0005159831914664367,
+      "loss": 3.738,
+      "step": 13100
+    },
+    {
+      "epoch": 1.4153481864169626,
+      "grad_norm": 0.5675348043441772,
+      "learning_rate": 0.0005156599504363753,
+      "loss": 3.7153,
+      "step": 13150
+    },
+    {
+      "epoch": 1.420729738456571,
+      "grad_norm": 0.5436713695526123,
+      "learning_rate": 0.000515336709406314,
+      "loss": 3.7266,
+      "step": 13200
+    },
+    {
+      "epoch": 1.426111290496179,
+      "grad_norm": 0.5975011587142944,
+      "learning_rate": 0.0005150134683762525,
+      "loss": 3.7214,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4314928425357873,
+      "grad_norm": 0.5936241745948792,
+      "learning_rate": 0.0005146902273461911,
+      "loss": 3.7073,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4368743945753955,
+      "grad_norm": 0.5979006290435791,
+      "learning_rate": 0.0005143669863161297,
+      "loss": 3.7124,
+      "step": 13350
+    },
+    {
+      "epoch": 1.4422559466150038,
+      "grad_norm": 0.5488461852073669,
+      "learning_rate": 0.0005140437452860683,
+      "loss": 3.7312,
+      "step": 13400
+    },
+    {
+      "epoch": 1.447637498654612,
+      "grad_norm": 0.5289202928543091,
+      "learning_rate": 0.0005137205042560069,
+      "loss": 3.7225,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4530190506942202,
+      "grad_norm": 0.5346786975860596,
+      "learning_rate": 0.0005133972632259455,
+      "loss": 3.7079,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4584006027338283,
+      "grad_norm": 0.540759801864624,
+      "learning_rate": 0.000513074022195884,
+      "loss": 3.7041,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4637821547734367,
+      "grad_norm": 0.5486022233963013,
+      "learning_rate": 0.0005127507811658226,
+      "loss": 3.7191,
+      "step": 13600
+    },
+    {
+      "epoch": 1.469163706813045,
+      "grad_norm": 0.5345197916030884,
+      "learning_rate": 0.0005124275401357612,
+      "loss": 3.7005,
+      "step": 13650
+    },
+    {
+      "epoch": 1.474545258852653,
+      "grad_norm": 0.5897067785263062,
+      "learning_rate": 0.0005121042991056997,
+      "loss": 3.6899,
+      "step": 13700
+    },
+    {
+      "epoch": 1.4799268108922612,
+      "grad_norm": 0.6418835520744324,
+      "learning_rate": 0.0005117810580756384,
+      "loss": 3.7033,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4853083629318695,
+      "grad_norm": 0.5940849781036377,
+      "learning_rate": 0.0005114578170455769,
+      "loss": 3.7171,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4906899149714778,
+      "grad_norm": 0.5546824932098389,
+      "learning_rate": 0.0005111345760155156,
+      "loss": 3.6979,
+      "step": 13850
+    },
+    {
+      "epoch": 1.496071467011086,
+      "grad_norm": 0.5385919809341431,
+      "learning_rate": 0.0005108113349854541,
+      "loss": 3.6956,
+      "step": 13900
+    },
+    {
+      "epoch": 1.501453019050694,
+      "grad_norm": 0.55533367395401,
+      "learning_rate": 0.0005104880939553926,
+      "loss": 3.6939,
+      "step": 13950
+    },
+    {
+      "epoch": 1.5068345710903024,
+      "grad_norm": 0.5384886264801025,
+      "learning_rate": 0.0005101648529253313,
+      "loss": 3.7135,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5068345710903024,
+      "eval_accuracy": 0.35304756682179766,
+      "eval_loss": 3.6596641540527344,
+      "eval_runtime": 185.7342,
+      "eval_samples_per_second": 96.972,
+      "eval_steps_per_second": 6.062,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5122161231299107,
+      "grad_norm": 0.5481053590774536,
+      "learning_rate": 0.0005098416118952699,
+      "loss": 3.6949,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5175976751695188,
+      "grad_norm": 0.537710428237915,
+      "learning_rate": 0.0005095248356858097,
+      "loss": 3.6944,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5229792272091272,
+      "grad_norm": 0.5734174251556396,
+      "learning_rate": 0.0005092015946557483,
+      "loss": 3.6972,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5283607792487355,
+      "grad_norm": 0.6006892323493958,
+      "learning_rate": 0.0005088783536256868,
+      "loss": 3.7064,
+      "step": 14200
+    },
+    {
+      "epoch": 1.5337423312883436,
+      "grad_norm": 0.5982166528701782,
+      "learning_rate": 0.0005085551125956255,
+      "loss": 3.7011,
+      "step": 14250
+    },
+    {
+      "epoch": 1.5391238833279517,
+      "grad_norm": 0.5659240484237671,
+      "learning_rate": 0.000508231871565564,
+      "loss": 3.7004,
+      "step": 14300
+    },
+    {
+      "epoch": 1.54450543536756,
+      "grad_norm": 0.5931114554405212,
+      "learning_rate": 0.0005079086305355026,
+      "loss": 3.6446,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5498869874071683,
+      "grad_norm": 0.5821999311447144,
+      "learning_rate": 0.0005075853895054412,
+      "loss": 3.7035,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5552685394467765,
+      "grad_norm": 0.5994320511817932,
+      "learning_rate": 0.0005072621484753797,
+      "loss": 3.6743,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5606500914863846,
+      "grad_norm": 0.6112990379333496,
+      "learning_rate": 0.0005069389074453184,
+      "loss": 3.6926,
+      "step": 14500
+    },
+    {
+      "epoch": 1.566031643525993,
+      "grad_norm": 0.5438514351844788,
+      "learning_rate": 0.0005066156664152569,
+      "loss": 3.6814,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5714131955656012,
+      "grad_norm": 0.5742460489273071,
+      "learning_rate": 0.0005062924253851955,
+      "loss": 3.7011,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5767947476052093,
+      "grad_norm": 0.59683758020401,
+      "learning_rate": 0.0005059691843551341,
+      "loss": 3.6865,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5821762996448174,
+      "grad_norm": 0.584264874458313,
+      "learning_rate": 0.0005056459433250727,
+      "loss": 3.6839,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5875578516844258,
+      "grad_norm": 0.5256335735321045,
+      "learning_rate": 0.0005053227022950113,
+      "loss": 3.6917,
+      "step": 14750
+    },
+    {
+      "epoch": 1.592939403724034,
+      "grad_norm": 0.5729854702949524,
+      "learning_rate": 0.0005049994612649499,
+      "loss": 3.6842,
+      "step": 14800
+    },
+    {
+      "epoch": 1.5983209557636422,
+      "grad_norm": 0.5756853818893433,
+      "learning_rate": 0.0005046762202348884,
+      "loss": 3.6701,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6037025078032503,
+      "grad_norm": 0.5055978894233704,
+      "learning_rate": 0.0005043594440254282,
+      "loss": 3.6961,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6090840598428586,
+      "grad_norm": 0.5643936395645142,
+      "learning_rate": 0.0005040362029953669,
+      "loss": 3.6823,
+      "step": 14950
+    },
+    {
+      "epoch": 1.614465611882467,
+      "grad_norm": 0.5897239446640015,
+      "learning_rate": 0.0005037129619653054,
+      "loss": 3.6762,
+      "step": 15000
+    },
+    {
+      "epoch": 1.614465611882467,
+      "eval_accuracy": 0.35501114181038096,
+      "eval_loss": 3.641000747680664,
+      "eval_runtime": 185.4628,
+      "eval_samples_per_second": 97.114,
+      "eval_steps_per_second": 6.071,
+      "step": 15000
+    },
+    {
+      "epoch": 1.619847163922075,
+      "grad_norm": 0.5518121123313904,
+      "learning_rate": 0.000503389720935244,
+      "loss": 3.6836,
+      "step": 15050
+    },
+    {
+      "epoch": 1.6252287159616834,
+      "grad_norm": 0.6155235171318054,
+      "learning_rate": 0.0005030664799051826,
+      "loss": 3.6857,
+      "step": 15100
+    },
+    {
+      "epoch": 1.6306102680012917,
+      "grad_norm": 0.5442000031471252,
+      "learning_rate": 0.0005027432388751212,
+      "loss": 3.6784,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6359918200408998,
+      "grad_norm": 0.5509122610092163,
+      "learning_rate": 0.0005024199978450598,
+      "loss": 3.6753,
+      "step": 15200
+    },
+    {
+      "epoch": 1.641373372080508,
+      "grad_norm": 0.5513061285018921,
+      "learning_rate": 0.0005020967568149983,
+      "loss": 3.692,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6467549241201163,
+      "grad_norm": 0.5306194424629211,
+      "learning_rate": 0.0005017735157849369,
+      "loss": 3.6766,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6521364761597246,
+      "grad_norm": 0.583339273929596,
+      "learning_rate": 0.0005014502747548755,
+      "loss": 3.6923,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6575180281993327,
+      "grad_norm": 0.573192834854126,
+      "learning_rate": 0.0005011270337248141,
+      "loss": 3.6866,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6628995802389408,
+      "grad_norm": 0.5406928062438965,
+      "learning_rate": 0.0005008037926947527,
+      "loss": 3.6833,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6682811322785491,
+      "grad_norm": 0.5753325819969177,
+      "learning_rate": 0.0005004805516646912,
+      "loss": 3.6824,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6736626843181575,
+      "grad_norm": 0.6574186086654663,
+      "learning_rate": 0.0005001573106346299,
+      "loss": 3.6587,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6790442363577656,
+      "grad_norm": 0.6001209020614624,
+      "learning_rate": 0.0004998340696045684,
+      "loss": 3.6763,
+      "step": 15600
+    },
+    {
+      "epoch": 1.6844257883973737,
+      "grad_norm": 0.5472638010978699,
+      "learning_rate": 0.0004995108285745071,
+      "loss": 3.6817,
+      "step": 15650
+    },
+    {
+      "epoch": 1.689807340436982,
+      "grad_norm": 0.5288031697273254,
+      "learning_rate": 0.0004991875875444456,
+      "loss": 3.6681,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6951888924765903,
+      "grad_norm": 0.535642147064209,
+      "learning_rate": 0.0004988643465143841,
+      "loss": 3.6757,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7005704445161984,
+      "grad_norm": 0.5962195992469788,
+      "learning_rate": 0.0004985411054843228,
+      "loss": 3.665,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7059519965558065,
+      "grad_norm": 0.5484269261360168,
+      "learning_rate": 0.0004982178644542613,
+      "loss": 3.6723,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7113335485954149,
+      "grad_norm": 0.5585965514183044,
+      "learning_rate": 0.0004978946234242,
+      "loss": 3.6663,
+      "step": 15900
+    },
+    {
+      "epoch": 1.7167151006350232,
+      "grad_norm": 0.5554015040397644,
+      "learning_rate": 0.0004975713823941385,
+      "loss": 3.659,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7220966526746313,
+      "grad_norm": 0.5840094089508057,
+      "learning_rate": 0.0004972481413640771,
+      "loss": 3.6543,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7220966526746313,
+      "eval_accuracy": 0.35623272622807917,
+      "eval_loss": 3.624523639678955,
+      "eval_runtime": 185.5469,
+      "eval_samples_per_second": 97.07,
+      "eval_steps_per_second": 6.069,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7274782047142396,
+      "grad_norm": 0.5376983880996704,
+      "learning_rate": 0.0004969249003340157,
+      "loss": 3.6673,
+      "step": 16050
+    },
+    {
+      "epoch": 1.732859756753848,
+      "grad_norm": 0.5641462802886963,
+      "learning_rate": 0.0004966016593039543,
+      "loss": 3.6716,
+      "step": 16100
+    },
+    {
+      "epoch": 1.738241308793456,
+      "grad_norm": 0.5286644697189331,
+      "learning_rate": 0.0004962784182738928,
+      "loss": 3.6528,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7436228608330642,
+      "grad_norm": 0.5174794793128967,
+      "learning_rate": 0.0004959551772438314,
+      "loss": 3.6776,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7490044128726725,
+      "grad_norm": 0.5619024634361267,
+      "learning_rate": 0.0004956319362137701,
+      "loss": 3.6783,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7543859649122808,
+      "grad_norm": 0.5532479286193848,
+      "learning_rate": 0.0004953086951837086,
+      "loss": 3.6593,
+      "step": 16300
+    },
+    {
+      "epoch": 1.759767516951889,
+      "grad_norm": 0.5429633855819702,
+      "learning_rate": 0.0004949854541536472,
+      "loss": 3.6784,
+      "step": 16350
+    },
+    {
+      "epoch": 1.765149068991497,
+      "grad_norm": 0.5421962141990662,
+      "learning_rate": 0.0004946622131235857,
+      "loss": 3.6647,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7705306210311054,
+      "grad_norm": 0.6100990772247314,
+      "learning_rate": 0.0004943389720935244,
+      "loss": 3.6593,
+      "step": 16450
+    },
+    {
+      "epoch": 1.7759121730707137,
+      "grad_norm": 0.6250017881393433,
+      "learning_rate": 0.000494015731063463,
+      "loss": 3.6382,
+      "step": 16500
+    },
+    {
+      "epoch": 1.7812937251103218,
+      "grad_norm": 0.5683926343917847,
+      "learning_rate": 0.0004936924900334016,
+      "loss": 3.6588,
+      "step": 16550
+    },
+    {
+      "epoch": 1.78667527714993,
+      "grad_norm": 0.57091224193573,
+      "learning_rate": 0.0004933692490033401,
+      "loss": 3.6582,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7920568291895382,
+      "grad_norm": 0.6777468919754028,
+      "learning_rate": 0.0004930460079732786,
+      "loss": 3.6675,
+      "step": 16650
+    },
+    {
+      "epoch": 1.7974383812291466,
+      "grad_norm": 0.5705229640007019,
+      "learning_rate": 0.0004927227669432173,
+      "loss": 3.6536,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8028199332687547,
+      "grad_norm": 0.5894872546195984,
+      "learning_rate": 0.0004923995259131558,
+      "loss": 3.6827,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8082014853083628,
+      "grad_norm": 0.542786717414856,
+      "learning_rate": 0.0004920762848830945,
+      "loss": 3.6567,
+      "step": 16800
+    },
+    {
+      "epoch": 1.813583037347971,
+      "grad_norm": 0.6244315505027771,
+      "learning_rate": 0.000491753043853033,
+      "loss": 3.6561,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8189645893875794,
+      "grad_norm": 0.5931727290153503,
+      "learning_rate": 0.0004914298028229717,
+      "loss": 3.6583,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8243461414271875,
+      "grad_norm": 0.6103002429008484,
+      "learning_rate": 0.0004911065617929102,
+      "loss": 3.6632,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8297276934667959,
+      "grad_norm": 0.5281617641448975,
+      "learning_rate": 0.0004907833207628487,
+      "loss": 3.6543,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8297276934667959,
+      "eval_accuracy": 0.35833309520242085,
+      "eval_loss": 3.60933256149292,
+      "eval_runtime": 186.5425,
+      "eval_samples_per_second": 96.552,
+      "eval_steps_per_second": 6.036,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8351092455064042,
+      "grad_norm": 0.5481086373329163,
+      "learning_rate": 0.0004904600797327874,
+      "loss": 3.6467,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8404907975460123,
+      "grad_norm": 0.5665550231933594,
+      "learning_rate": 0.0004901368387027259,
+      "loss": 3.6423,
+      "step": 17100
+    },
+    {
+      "epoch": 1.8458723495856204,
+      "grad_norm": 0.5502485036849976,
+      "learning_rate": 0.0004898135976726646,
+      "loss": 3.657,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8512539016252287,
+      "grad_norm": 0.6010493040084839,
+      "learning_rate": 0.0004894903566426031,
+      "loss": 3.6452,
+      "step": 17200
+    },
+    {
+      "epoch": 1.856635453664837,
+      "grad_norm": 0.5746400952339172,
+      "learning_rate": 0.0004891671156125417,
+      "loss": 3.653,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8620170057044452,
+      "grad_norm": 0.6138567924499512,
+      "learning_rate": 0.0004888438745824803,
+      "loss": 3.6655,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8673985577440533,
+      "grad_norm": 0.583207905292511,
+      "learning_rate": 0.0004885206335524189,
+      "loss": 3.6324,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8727801097836616,
+      "grad_norm": 0.6441901922225952,
+      "learning_rate": 0.0004881973925223575,
+      "loss": 3.6564,
+      "step": 17400
+    },
+    {
+      "epoch": 1.87816166182327,
+      "grad_norm": 0.5731293559074402,
+      "learning_rate": 0.00048787415149229604,
+      "loss": 3.6478,
+      "step": 17450
+    },
+    {
+      "epoch": 1.883543213862878,
+      "grad_norm": 0.5457957983016968,
+      "learning_rate": 0.00048755091046223464,
+      "loss": 3.6492,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8889247659024861,
+      "grad_norm": 0.5309653282165527,
+      "learning_rate": 0.0004872276694321732,
+      "loss": 3.6396,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8943063179420945,
+      "grad_norm": 0.5328410267829895,
+      "learning_rate": 0.00048690442840211177,
+      "loss": 3.6489,
+      "step": 17600
+    },
+    {
+      "epoch": 1.8996878699817028,
+      "grad_norm": 0.5412832498550415,
+      "learning_rate": 0.0004865811873720504,
+      "loss": 3.6635,
+      "step": 17650
+    },
+    {
+      "epoch": 1.905069422021311,
+      "grad_norm": 0.5539383292198181,
+      "learning_rate": 0.00048625794634198896,
+      "loss": 3.629,
+      "step": 17700
+    },
+    {
+      "epoch": 1.910450974060919,
+      "grad_norm": 0.5800532698631287,
+      "learning_rate": 0.00048593470531192756,
+      "loss": 3.6246,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9158325261005273,
+      "grad_norm": 0.5952960848808289,
+      "learning_rate": 0.00048561146428186615,
+      "loss": 3.6437,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9212140781401357,
+      "grad_norm": 0.5447590947151184,
+      "learning_rate": 0.0004852882232518047,
+      "loss": 3.6428,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9265956301797438,
+      "grad_norm": 0.5065428018569946,
+      "learning_rate": 0.00048496498222174334,
+      "loss": 3.6448,
+      "step": 17900
+    },
+    {
+      "epoch": 1.931977182219352,
+      "grad_norm": 0.5957828760147095,
+      "learning_rate": 0.00048464174119168193,
+      "loss": 3.6365,
+      "step": 17950
+    },
+    {
+      "epoch": 1.9373587342589604,
+      "grad_norm": 0.5402979254722595,
+      "learning_rate": 0.0004843185001616205,
+      "loss": 3.6548,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9373587342589604,
+      "eval_accuracy": 0.3598062110554096,
+      "eval_loss": 3.5946199893951416,
+      "eval_runtime": 186.0184,
+      "eval_samples_per_second": 96.824,
+      "eval_steps_per_second": 6.053,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9427402862985685,
+      "grad_norm": 0.533379852771759,
+      "learning_rate": 0.00048399525913155907,
+      "loss": 3.6385,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9481218383381766,
+      "grad_norm": 0.5543186664581299,
+      "learning_rate": 0.0004836720181014976,
+      "loss": 3.6358,
+      "step": 18100
+    },
+    {
+      "epoch": 1.953503390377785,
+      "grad_norm": 0.6201730966567993,
+      "learning_rate": 0.0004833487770714362,
+      "loss": 3.6436,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9588849424173933,
+      "grad_norm": 0.5487646460533142,
+      "learning_rate": 0.00048302553604137485,
+      "loss": 3.6347,
+      "step": 18200
+    },
+    {
+      "epoch": 1.9642664944570014,
+      "grad_norm": 0.5432608723640442,
+      "learning_rate": 0.0004827022950113134,
+      "loss": 3.6549,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9696480464966095,
+      "grad_norm": 0.5554027557373047,
+      "learning_rate": 0.000482379053981252,
+      "loss": 3.6396,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9750295985362178,
+      "grad_norm": 0.5545364618301392,
+      "learning_rate": 0.0004820558129511906,
+      "loss": 3.6589,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9804111505758262,
+      "grad_norm": 0.6176003813743591,
+      "learning_rate": 0.0004817325719211291,
+      "loss": 3.6463,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9857927026154343,
+      "grad_norm": 0.5684056878089905,
+      "learning_rate": 0.0004814093308910677,
+      "loss": 3.6373,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9911742546550424,
+      "grad_norm": 0.5712719559669495,
+      "learning_rate": 0.00048108608986100637,
+      "loss": 3.6374,
+      "step": 18500
+    },
+    {
+      "epoch": 1.9965558066946507,
+      "grad_norm": 0.5630627870559692,
+      "learning_rate": 0.0004807628488309449,
+      "loss": 3.6451,
+      "step": 18550
+    },
+    {
+      "epoch": 2.001937358734259,
+      "grad_norm": 0.5038061738014221,
+      "learning_rate": 0.0004804396078008835,
+      "loss": 3.5965,
+      "step": 18600
+    },
+    {
+      "epoch": 2.007318910773867,
+      "grad_norm": 0.5746957659721375,
+      "learning_rate": 0.00048011636677082204,
+      "loss": 3.5246,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0127004628134753,
+      "grad_norm": 0.5593902468681335,
+      "learning_rate": 0.00047979312574076064,
+      "loss": 3.5567,
+      "step": 18700
+    },
+    {
+      "epoch": 2.018082014853084,
+      "grad_norm": 0.5422948598861694,
+      "learning_rate": 0.0004794698847106992,
+      "loss": 3.5636,
+      "step": 18750
+    },
+    {
+      "epoch": 2.023463566892692,
+      "grad_norm": 0.6284714341163635,
+      "learning_rate": 0.0004791466436806378,
+      "loss": 3.5386,
+      "step": 18800
+    },
+    {
+      "epoch": 2.0288451189323,
+      "grad_norm": 0.6103267669677734,
+      "learning_rate": 0.0004788234026505764,
+      "loss": 3.5498,
+      "step": 18850
+    },
+    {
+      "epoch": 2.034226670971908,
+      "grad_norm": 0.5950919985771179,
+      "learning_rate": 0.00047850016162051496,
+      "loss": 3.5738,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0396082230115167,
+      "grad_norm": 0.5741006135940552,
+      "learning_rate": 0.00047817692059045356,
+      "loss": 3.5594,
+      "step": 18950
+    },
+    {
+      "epoch": 2.044989775051125,
+      "grad_norm": 0.6038844585418701,
+      "learning_rate": 0.00047785367956039215,
+      "loss": 3.5527,
+      "step": 19000
+    },
+    {
+      "epoch": 2.044989775051125,
+      "eval_accuracy": 0.36119903242430346,
+      "eval_loss": 3.583731174468994,
+      "eval_runtime": 185.686,
+      "eval_samples_per_second": 96.997,
+      "eval_steps_per_second": 6.064,
+      "step": 19000
+    },
+    {
+      "epoch": 2.050371327090733,
+      "grad_norm": 0.5832013487815857,
+      "learning_rate": 0.00047753043853033075,
+      "loss": 3.5533,
+      "step": 19050
+    },
+    {
+      "epoch": 2.055752879130341,
+      "grad_norm": 0.6129773855209351,
+      "learning_rate": 0.00047720719750026934,
+      "loss": 3.5654,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0611344311699495,
+      "grad_norm": Infinity,
+      "learning_rate": 0.00047689042129080915,
+      "loss": 3.5587,
+      "step": 19150
+    },
+    {
+      "epoch": 2.0665159832095576,
+      "grad_norm": 0.6074003577232361,
+      "learning_rate": 0.00047656718026074774,
+      "loss": 3.5633,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0718975352491658,
+      "grad_norm": 0.6134311556816101,
+      "learning_rate": 0.00047624393923068634,
+      "loss": 3.5469,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0772790872887743,
+      "grad_norm": 0.5661240220069885,
+      "learning_rate": 0.0004759206982006249,
+      "loss": 3.5722,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0826606393283824,
+      "grad_norm": 0.5343161821365356,
+      "learning_rate": 0.00047559745717056347,
+      "loss": 3.5582,
+      "step": 19350
+    },
+    {
+      "epoch": 2.0880421913679905,
+      "grad_norm": 0.592681348323822,
+      "learning_rate": 0.000475274216140502,
+      "loss": 3.5487,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0934237434075986,
+      "grad_norm": 0.5857892036437988,
+      "learning_rate": 0.00047495097511044066,
+      "loss": 3.5669,
+      "step": 19450
+    },
+    {
+      "epoch": 2.098805295447207,
+      "grad_norm": 0.6342927813529968,
+      "learning_rate": 0.00047462773408037925,
+      "loss": 3.5761,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1041868474868153,
+      "grad_norm": 0.5716348886489868,
+      "learning_rate": 0.0004743044930503178,
+      "loss": 3.5475,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1095683995264234,
+      "grad_norm": 0.5774434208869934,
+      "learning_rate": 0.0004739812520202564,
+      "loss": 3.5398,
+      "step": 19600
+    },
+    {
+      "epoch": 2.1149499515660315,
+      "grad_norm": 0.6075140833854675,
+      "learning_rate": 0.000473658010990195,
+      "loss": 3.5544,
+      "step": 19650
+    },
+    {
+      "epoch": 2.12033150360564,
+      "grad_norm": 0.6039460301399231,
+      "learning_rate": 0.0004733347699601336,
+      "loss": 3.5715,
+      "step": 19700
+    },
+    {
+      "epoch": 2.125713055645248,
+      "grad_norm": 0.5807799100875854,
+      "learning_rate": 0.0004730115289300722,
+      "loss": 3.5565,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1310946076848563,
+      "grad_norm": 0.5377194285392761,
+      "learning_rate": 0.00047268828790001077,
+      "loss": 3.5585,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1364761597244644,
+      "grad_norm": 0.6161070466041565,
+      "learning_rate": 0.0004723650468699493,
+      "loss": 3.5665,
+      "step": 19850
+    },
+    {
+      "epoch": 2.141857711764073,
+      "grad_norm": 0.633151113986969,
+      "learning_rate": 0.0004720418058398879,
+      "loss": 3.5622,
+      "step": 19900
+    },
+    {
+      "epoch": 2.147239263803681,
+      "grad_norm": 0.538801908493042,
+      "learning_rate": 0.00047171856480982644,
+      "loss": 3.5557,
+      "step": 19950
+    },
+    {
+      "epoch": 2.152620815843289,
+      "grad_norm": 0.5712424516677856,
+      "learning_rate": 0.0004713953237797651,
+      "loss": 3.57,
+      "step": 20000
+    },
+    {
+      "epoch": 2.152620815843289,
+      "eval_accuracy": 0.3620634747997554,
+      "eval_loss": 3.5744924545288086,
+      "eval_runtime": 185.6525,
+      "eval_samples_per_second": 97.015,
+      "eval_steps_per_second": 6.065,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1580023678828972,
+      "grad_norm": 0.5541605353355408,
+      "learning_rate": 0.0004710720827497037,
+      "loss": 3.5565,
+      "step": 20050
+    },
+    {
+      "epoch": 2.163383919922506,
+      "grad_norm": 0.5375374555587769,
+      "learning_rate": 0.00047074884171964223,
+      "loss": 3.5493,
+      "step": 20100
+    },
+    {
+      "epoch": 2.168765471962114,
+      "grad_norm": 0.6366614103317261,
+      "learning_rate": 0.0004704256006895808,
+      "loss": 3.5769,
+      "step": 20150
+    },
+    {
+      "epoch": 2.174147024001722,
+      "grad_norm": 0.5753631591796875,
+      "learning_rate": 0.00047010235965951936,
+      "loss": 3.5631,
+      "step": 20200
+    },
+    {
+      "epoch": 2.1795285760413305,
+      "grad_norm": 0.5457363128662109,
+      "learning_rate": 0.00046977911862945796,
+      "loss": 3.5514,
+      "step": 20250
+    },
+    {
+      "epoch": 2.1849101280809387,
+      "grad_norm": 0.5075417160987854,
+      "learning_rate": 0.0004694558775993966,
+      "loss": 3.567,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1902916801205468,
+      "grad_norm": 0.5688516497612,
+      "learning_rate": 0.00046913263656933515,
+      "loss": 3.5478,
+      "step": 20350
+    },
+    {
+      "epoch": 2.195673232160155,
+      "grad_norm": 0.6135866641998291,
+      "learning_rate": 0.00046880939553927374,
+      "loss": 3.557,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2010547841997634,
+      "grad_norm": 0.5512149930000305,
+      "learning_rate": 0.00046848615450921234,
+      "loss": 3.5791,
+      "step": 20450
+    },
+    {
+      "epoch": 2.2064363362393715,
+      "grad_norm": 0.5562562942504883,
+      "learning_rate": 0.0004681629134791509,
+      "loss": 3.5703,
+      "step": 20500
+    },
+    {
+      "epoch": 2.2118178882789796,
+      "grad_norm": 0.5920165181159973,
+      "learning_rate": 0.00046783967244908947,
+      "loss": 3.5676,
+      "step": 20550
+    },
+    {
+      "epoch": 2.2171994403185877,
+      "grad_norm": 0.5704071521759033,
+      "learning_rate": 0.0004675164314190281,
+      "loss": 3.568,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2225809923581963,
+      "grad_norm": 0.5668667554855347,
+      "learning_rate": 0.00046719319038896666,
+      "loss": 3.5723,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2279625443978044,
+      "grad_norm": 0.6024516224861145,
+      "learning_rate": 0.00046686994935890526,
+      "loss": 3.5786,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2333440964374125,
+      "grad_norm": 0.5603658556938171,
+      "learning_rate": 0.0004665467083288438,
+      "loss": 3.5562,
+      "step": 20750
+    },
+    {
+      "epoch": 2.2387256484770206,
+      "grad_norm": 0.6055840253829956,
+      "learning_rate": 0.0004662234672987824,
+      "loss": 3.5717,
+      "step": 20800
+    },
+    {
+      "epoch": 2.244107200516629,
+      "grad_norm": 0.5434353947639465,
+      "learning_rate": 0.00046590022626872104,
+      "loss": 3.5666,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2494887525562373,
+      "grad_norm": 0.5617761611938477,
+      "learning_rate": 0.0004655769852386596,
+      "loss": 3.5527,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2548703045958454,
+      "grad_norm": 0.5419901013374329,
+      "learning_rate": 0.0004652537442085982,
+      "loss": 3.5604,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2602518566354535,
+      "grad_norm": 0.5438917279243469,
+      "learning_rate": 0.00046493050317853677,
+      "loss": 3.5501,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2602518566354535,
+      "eval_accuracy": 0.3635552789494483,
+      "eval_loss": 3.563424587249756,
+      "eval_runtime": 185.7944,
+      "eval_samples_per_second": 96.94,
+      "eval_steps_per_second": 6.06,
+      "step": 21000
+    },
+    {
+      "epoch": 2.265633408675062,
+      "grad_norm": 0.58818119764328,
+      "learning_rate": 0.0004646072621484753,
+      "loss": 3.5582,
+      "step": 21050
+    },
+    {
+      "epoch": 2.27101496071467,
+      "grad_norm": 0.5365327000617981,
+      "learning_rate": 0.0004642840211184139,
+      "loss": 3.5603,
+      "step": 21100
+    },
+    {
+      "epoch": 2.2763965127542782,
+      "grad_norm": 0.5736598968505859,
+      "learning_rate": 0.00046396078008835255,
+      "loss": 3.5621,
+      "step": 21150
+    },
+    {
+      "epoch": 2.281778064793887,
+      "grad_norm": 0.5941365361213684,
+      "learning_rate": 0.0004636440038788923,
+      "loss": 3.5627,
+      "step": 21200
+    },
+    {
+      "epoch": 2.287159616833495,
+      "grad_norm": 0.5709973573684692,
+      "learning_rate": 0.00046332076284883095,
+      "loss": 3.5505,
+      "step": 21250
+    },
+    {
+      "epoch": 2.292541168873103,
+      "grad_norm": 0.6074051856994629,
+      "learning_rate": 0.0004629975218187695,
+      "loss": 3.554,
+      "step": 21300
+    },
+    {
+      "epoch": 2.297922720912711,
+      "grad_norm": 0.546488344669342,
+      "learning_rate": 0.0004626742807887081,
+      "loss": 3.5745,
+      "step": 21350
+    },
+    {
+      "epoch": 2.303304272952319,
+      "grad_norm": 0.5935389399528503,
+      "learning_rate": 0.00046235103975864663,
+      "loss": 3.5596,
+      "step": 21400
+    },
+    {
+      "epoch": 2.3086858249919278,
+      "grad_norm": 0.6030110120773315,
+      "learning_rate": 0.0004620277987285852,
+      "loss": 3.5549,
+      "step": 21450
+    },
+    {
+      "epoch": 2.314067377031536,
+      "grad_norm": 0.6092215776443481,
+      "learning_rate": 0.0004617045576985239,
+      "loss": 3.5534,
+      "step": 21500
+    },
+    {
+      "epoch": 2.319448929071144,
+      "grad_norm": 0.557925820350647,
+      "learning_rate": 0.0004613813166684624,
+      "loss": 3.5498,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3248304811107525,
+      "grad_norm": 0.5842859745025635,
+      "learning_rate": 0.000461058075638401,
+      "loss": 3.5463,
+      "step": 21600
+    },
+    {
+      "epoch": 2.3302120331503606,
+      "grad_norm": 0.5733357071876526,
+      "learning_rate": 0.00046073483460833955,
+      "loss": 3.5722,
+      "step": 21650
+    },
+    {
+      "epoch": 2.3355935851899687,
+      "grad_norm": 0.5481617450714111,
+      "learning_rate": 0.00046041159357827814,
+      "loss": 3.552,
+      "step": 21700
+    },
+    {
+      "epoch": 2.340975137229577,
+      "grad_norm": 0.5121580362319946,
+      "learning_rate": 0.00046008835254821674,
+      "loss": 3.568,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3463566892691854,
+      "grad_norm": 0.5733767747879028,
+      "learning_rate": 0.0004597651115181554,
+      "loss": 3.5764,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3517382413087935,
+      "grad_norm": 0.5663204193115234,
+      "learning_rate": 0.00045944187048809393,
+      "loss": 3.5502,
+      "step": 21850
+    },
+    {
+      "epoch": 2.3571197933484016,
+      "grad_norm": 0.6015146970748901,
+      "learning_rate": 0.0004591186294580325,
+      "loss": 3.5489,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3625013453880097,
+      "grad_norm": 0.5521755814552307,
+      "learning_rate": 0.00045879538842797106,
+      "loss": 3.5512,
+      "step": 21950
+    },
+    {
+      "epoch": 2.3678828974276183,
+      "grad_norm": 0.587854266166687,
+      "learning_rate": 0.00045847214739790966,
+      "loss": 3.5595,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3678828974276183,
+      "eval_accuracy": 0.36471601774997037,
+      "eval_loss": 3.5532937049865723,
+      "eval_runtime": 185.6564,
+      "eval_samples_per_second": 97.013,
+      "eval_steps_per_second": 6.065,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3732644494672264,
+      "grad_norm": 0.5953104496002197,
+      "learning_rate": 0.0004581489063678482,
+      "loss": 3.5564,
+      "step": 22050
+    },
+    {
+      "epoch": 2.3786460015068345,
+      "grad_norm": 0.6211328506469727,
+      "learning_rate": 0.00045782566533778685,
+      "loss": 3.5638,
+      "step": 22100
+    },
+    {
+      "epoch": 2.384027553546443,
+      "grad_norm": 0.5673883557319641,
+      "learning_rate": 0.00045750242430772544,
+      "loss": 3.5464,
+      "step": 22150
+    },
+    {
+      "epoch": 2.389409105586051,
+      "grad_norm": 0.5393223166465759,
+      "learning_rate": 0.000457179183277664,
+      "loss": 3.5546,
+      "step": 22200
+    },
+    {
+      "epoch": 2.3947906576256592,
+      "grad_norm": 0.6029070615768433,
+      "learning_rate": 0.0004568559422476026,
+      "loss": 3.564,
+      "step": 22250
+    },
+    {
+      "epoch": 2.4001722096652673,
+      "grad_norm": 0.6550480127334595,
+      "learning_rate": 0.00045653270121754117,
+      "loss": 3.555,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4055537617048754,
+      "grad_norm": 0.5430505275726318,
+      "learning_rate": 0.0004562094601874797,
+      "loss": 3.5629,
+      "step": 22350
+    },
+    {
+      "epoch": 2.410935313744484,
+      "grad_norm": 0.5960400700569153,
+      "learning_rate": 0.00045588621915741836,
+      "loss": 3.5488,
+      "step": 22400
+    },
+    {
+      "epoch": 2.416316865784092,
+      "grad_norm": 0.5665459036827087,
+      "learning_rate": 0.00045556297812735696,
+      "loss": 3.5488,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4216984178237,
+      "grad_norm": 0.5476963520050049,
+      "learning_rate": 0.0004552397370972955,
+      "loss": 3.5555,
+      "step": 22500
+    },
+    {
+      "epoch": 2.4270799698633088,
+      "grad_norm": 0.6745426058769226,
+      "learning_rate": 0.0004549164960672341,
+      "loss": 3.5587,
+      "step": 22550
+    },
+    {
+      "epoch": 2.432461521902917,
+      "grad_norm": 0.6348827481269836,
+      "learning_rate": 0.00045459325503717263,
+      "loss": 3.5536,
+      "step": 22600
+    },
+    {
+      "epoch": 2.437843073942525,
+      "grad_norm": 0.5526021122932434,
+      "learning_rate": 0.0004542700140071113,
+      "loss": 3.5577,
+      "step": 22650
+    },
+    {
+      "epoch": 2.443224625982133,
+      "grad_norm": 0.617853045463562,
+      "learning_rate": 0.0004539467729770499,
+      "loss": 3.5433,
+      "step": 22700
+    },
+    {
+      "epoch": 2.4486061780217416,
+      "grad_norm": 0.5563679337501526,
+      "learning_rate": 0.0004536235319469884,
+      "loss": 3.5379,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4539877300613497,
+      "grad_norm": 0.6133175492286682,
+      "learning_rate": 0.000453300290916927,
+      "loss": 3.5545,
+      "step": 22800
+    },
+    {
+      "epoch": 2.459369282100958,
+      "grad_norm": 0.5517181754112244,
+      "learning_rate": 0.0004529770498868656,
+      "loss": 3.5394,
+      "step": 22850
+    },
+    {
+      "epoch": 2.464750834140566,
+      "grad_norm": 0.5468781590461731,
+      "learning_rate": 0.00045265380885680414,
+      "loss": 3.5555,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4701323861801745,
+      "grad_norm": 0.5771084427833557,
+      "learning_rate": 0.0004523305678267428,
+      "loss": 3.5494,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4755139382197826,
+      "grad_norm": 0.5483697652816772,
+      "learning_rate": 0.0004520073267966814,
+      "loss": 3.5527,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4755139382197826,
+      "eval_accuracy": 0.3653397939789133,
+      "eval_loss": 3.541123628616333,
+      "eval_runtime": 189.0274,
+      "eval_samples_per_second": 95.282,
+      "eval_steps_per_second": 5.957,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4808954902593907,
+      "grad_norm": 0.5761868357658386,
+      "learning_rate": 0.00045168408576661993,
+      "loss": 3.5461,
+      "step": 23050
+    },
+    {
+      "epoch": 2.4862770422989993,
+      "grad_norm": 0.5387217998504639,
+      "learning_rate": 0.0004513608447365585,
+      "loss": 3.5381,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4916585943386074,
+      "grad_norm": 0.5650376677513123,
+      "learning_rate": 0.00045103760370649706,
+      "loss": 3.5525,
+      "step": 23150
+    },
+    {
+      "epoch": 2.4970401463782155,
+      "grad_norm": 0.5395632386207581,
+      "learning_rate": 0.00045071436267643566,
+      "loss": 3.5574,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5024216984178236,
+      "grad_norm": 0.5282270312309265,
+      "learning_rate": 0.00045039758646697546,
+      "loss": 3.5661,
+      "step": 23250
+    },
+    {
+      "epoch": 2.5078032504574317,
+      "grad_norm": 0.5455463528633118,
+      "learning_rate": 0.0004500743454369141,
+      "loss": 3.5415,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5131848024970402,
+      "grad_norm": 0.5784099698066711,
+      "learning_rate": 0.0004497511044068527,
+      "loss": 3.545,
+      "step": 23350
+    },
+    {
+      "epoch": 2.5185663545366483,
+      "grad_norm": 0.5365017652511597,
+      "learning_rate": 0.00044942786337679125,
+      "loss": 3.5503,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5239479065762565,
+      "grad_norm": 0.6023415327072144,
+      "learning_rate": 0.00044910462234672984,
+      "loss": 3.5382,
+      "step": 23450
+    },
+    {
+      "epoch": 2.529329458615865,
+      "grad_norm": 0.6116933226585388,
+      "learning_rate": 0.0004487813813166684,
+      "loss": 3.5496,
+      "step": 23500
+    },
+    {
+      "epoch": 2.534711010655473,
+      "grad_norm": 0.5484300851821899,
+      "learning_rate": 0.000448458140286607,
+      "loss": 3.5543,
+      "step": 23550
+    },
+    {
+      "epoch": 2.540092562695081,
+      "grad_norm": 0.5980265140533447,
+      "learning_rate": 0.00044813489925654563,
+      "loss": 3.5318,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5454741147346893,
+      "grad_norm": 0.6223058700561523,
+      "learning_rate": 0.00044781165822648417,
+      "loss": 3.5388,
+      "step": 23650
+    },
+    {
+      "epoch": 2.550855666774298,
+      "grad_norm": 0.5941548943519592,
+      "learning_rate": 0.00044748841719642276,
+      "loss": 3.5611,
+      "step": 23700
+    },
+    {
+      "epoch": 2.556237218813906,
+      "grad_norm": 0.5907602310180664,
+      "learning_rate": 0.00044716517616636136,
+      "loss": 3.5495,
+      "step": 23750
+    },
+    {
+      "epoch": 2.561618770853514,
+      "grad_norm": 0.6576862931251526,
+      "learning_rate": 0.0004468419351362999,
+      "loss": 3.549,
+      "step": 23800
+    },
+    {
+      "epoch": 2.567000322893122,
+      "grad_norm": 0.6121534109115601,
+      "learning_rate": 0.0004465186941062385,
+      "loss": 3.5534,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5723818749327307,
+      "grad_norm": 0.5545932054519653,
+      "learning_rate": 0.00044619545307617714,
+      "loss": 3.5396,
+      "step": 23900
+    },
+    {
+      "epoch": 2.577763426972339,
+      "grad_norm": 0.5737384557723999,
+      "learning_rate": 0.0004458722120461157,
+      "loss": 3.5413,
+      "step": 23950
+    },
+    {
+      "epoch": 2.583144979011947,
+      "grad_norm": 0.5477451086044312,
+      "learning_rate": 0.0004455489710160543,
+      "loss": 3.5441,
+      "step": 24000
+    },
+    {
+      "epoch": 2.583144979011947,
+      "eval_accuracy": 0.36705824805258713,
+      "eval_loss": 3.531026601791382,
+      "eval_runtime": 185.3537,
+      "eval_samples_per_second": 97.171,
+      "eval_steps_per_second": 6.075,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5885265310515555,
+      "grad_norm": 0.5542248487472534,
+      "learning_rate": 0.0004452257299859928,
+      "loss": 3.5473,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5939080830911636,
+      "grad_norm": 0.5204565525054932,
+      "learning_rate": 0.0004449024889559314,
+      "loss": 3.5464,
+      "step": 24100
+    },
+    {
+      "epoch": 2.5992896351307717,
+      "grad_norm": 0.5798118114471436,
+      "learning_rate": 0.00044457924792587,
+      "loss": 3.5467,
+      "step": 24150
+    },
+    {
+      "epoch": 2.60467118717038,
+      "grad_norm": 0.620180070400238,
+      "learning_rate": 0.0004442560068958086,
+      "loss": 3.5442,
+      "step": 24200
+    },
+    {
+      "epoch": 2.610052739209988,
+      "grad_norm": 0.5598777532577515,
+      "learning_rate": 0.0004439327658657472,
+      "loss": 3.5455,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6154342912495965,
+      "grad_norm": 0.6063926815986633,
+      "learning_rate": 0.0004436095248356858,
+      "loss": 3.5301,
+      "step": 24300
+    },
+    {
+      "epoch": 2.6208158432892046,
+      "grad_norm": 0.5656163096427917,
+      "learning_rate": 0.00044328628380562433,
+      "loss": 3.5517,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6261973953288127,
+      "grad_norm": 0.5375902056694031,
+      "learning_rate": 0.0004429630427755629,
+      "loss": 3.5537,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6315789473684212,
+      "grad_norm": 0.584174633026123,
+      "learning_rate": 0.0004426398017455016,
+      "loss": 3.5609,
+      "step": 24450
+    },
+    {
+      "epoch": 2.6369604994080293,
+      "grad_norm": 0.559461236000061,
+      "learning_rate": 0.0004423165607154401,
+      "loss": 3.5418,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6423420514476375,
+      "grad_norm": 0.5810232162475586,
+      "learning_rate": 0.0004419933196853787,
+      "loss": 3.5508,
+      "step": 24550
+    },
+    {
+      "epoch": 2.6477236034872456,
+      "grad_norm": 0.5881146788597107,
+      "learning_rate": 0.00044167007865531725,
+      "loss": 3.5519,
+      "step": 24600
+    },
+    {
+      "epoch": 2.653105155526854,
+      "grad_norm": 0.5854257345199585,
+      "learning_rate": 0.00044134683762525584,
+      "loss": 3.5627,
+      "step": 24650
+    },
+    {
+      "epoch": 2.658486707566462,
+      "grad_norm": 0.5748598575592041,
+      "learning_rate": 0.00044102359659519444,
+      "loss": 3.5312,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6638682596060703,
+      "grad_norm": 0.5678013563156128,
+      "learning_rate": 0.00044070035556513303,
+      "loss": 3.5591,
+      "step": 24750
+    },
+    {
+      "epoch": 2.6692498116456784,
+      "grad_norm": 0.5374163389205933,
+      "learning_rate": 0.00044037711453507163,
+      "loss": 3.5553,
+      "step": 24800
+    },
+    {
+      "epoch": 2.674631363685287,
+      "grad_norm": 0.5951510667800903,
+      "learning_rate": 0.0004400538735050102,
+      "loss": 3.5548,
+      "step": 24850
+    },
+    {
+      "epoch": 2.680012915724895,
+      "grad_norm": 0.6498360633850098,
+      "learning_rate": 0.00043973063247494876,
+      "loss": 3.538,
+      "step": 24900
+    },
+    {
+      "epoch": 2.685394467764503,
+      "grad_norm": 0.5752738118171692,
+      "learning_rate": 0.00043940739144488736,
+      "loss": 3.5468,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6907760198041117,
+      "grad_norm": 0.5516283512115479,
+      "learning_rate": 0.0004390841504148259,
+      "loss": 3.5373,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6907760198041117,
+      "eval_accuracy": 0.3670333665412776,
+      "eval_loss": 3.524658441543579,
+      "eval_runtime": 185.8525,
+      "eval_samples_per_second": 96.91,
+      "eval_steps_per_second": 6.059,
+      "step": 25000
+    },
+    {
+      "epoch": 2.69615757184372,
+      "grad_norm": 0.5754989981651306,
+      "learning_rate": 0.00043876090938476455,
+      "loss": 3.5408,
+      "step": 25050
+    },
+    {
+      "epoch": 2.701539123883328,
+      "grad_norm": 0.5642285346984863,
+      "learning_rate": 0.00043843766835470314,
+      "loss": 3.5392,
+      "step": 25100
+    },
+    {
+      "epoch": 2.706920675922936,
+      "grad_norm": 0.5629568099975586,
+      "learning_rate": 0.00043812089214524295,
+      "loss": 3.5367,
+      "step": 25150
+    },
+    {
+      "epoch": 2.712302227962544,
+      "grad_norm": 0.5710545778274536,
+      "learning_rate": 0.00043779765111518154,
+      "loss": 3.5399,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7176837800021527,
+      "grad_norm": 0.5741075873374939,
+      "learning_rate": 0.0004374744100851201,
+      "loss": 3.5165,
+      "step": 25250
+    },
+    {
+      "epoch": 2.723065332041761,
+      "grad_norm": 0.5557492971420288,
+      "learning_rate": 0.0004371511690550587,
+      "loss": 3.538,
+      "step": 25300
+    },
+    {
+      "epoch": 2.728446884081369,
+      "grad_norm": 0.5741593241691589,
+      "learning_rate": 0.0004368279280249972,
+      "loss": 3.5379,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7338284361209775,
+      "grad_norm": 0.5418490767478943,
+      "learning_rate": 0.00043650468699493587,
+      "loss": 3.5393,
+      "step": 25400
+    },
+    {
+      "epoch": 2.7392099881605856,
+      "grad_norm": 0.5548985600471497,
+      "learning_rate": 0.00043618144596487446,
+      "loss": 3.5508,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7445915402001937,
+      "grad_norm": 0.60408616065979,
+      "learning_rate": 0.000435858204934813,
+      "loss": 3.5481,
+      "step": 25500
+    },
+    {
+      "epoch": 2.749973092239802,
+      "grad_norm": 0.5776184797286987,
+      "learning_rate": 0.0004355349639047516,
+      "loss": 3.5232,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7553546442794103,
+      "grad_norm": 0.6261433959007263,
+      "learning_rate": 0.0004352117228746902,
+      "loss": 3.5382,
+      "step": 25600
+    },
+    {
+      "epoch": 2.7607361963190185,
+      "grad_norm": 0.6292518973350525,
+      "learning_rate": 0.00043488848184462873,
+      "loss": 3.557,
+      "step": 25650
+    },
+    {
+      "epoch": 2.7661177483586266,
+      "grad_norm": 0.5952281951904297,
+      "learning_rate": 0.0004345652408145674,
+      "loss": 3.5446,
+      "step": 25700
+    },
+    {
+      "epoch": 2.7714993003982347,
+      "grad_norm": 0.5846927165985107,
+      "learning_rate": 0.000434241999784506,
+      "loss": 3.5261,
+      "step": 25750
+    },
+    {
+      "epoch": 2.776880852437843,
+      "grad_norm": 0.6028293967247009,
+      "learning_rate": 0.0004339187587544445,
+      "loss": 3.552,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7822624044774513,
+      "grad_norm": 0.6186114549636841,
+      "learning_rate": 0.0004335955177243831,
+      "loss": 3.5343,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7876439565170594,
+      "grad_norm": 0.5727343559265137,
+      "learning_rate": 0.00043327227669432165,
+      "loss": 3.5524,
+      "step": 25900
+    },
+    {
+      "epoch": 2.793025508556668,
+      "grad_norm": 0.6375524401664734,
+      "learning_rate": 0.00043294903566426025,
+      "loss": 3.5281,
+      "step": 25950
+    },
+    {
+      "epoch": 2.798407060596276,
+      "grad_norm": 0.5392520427703857,
+      "learning_rate": 0.0004326257946341989,
+      "loss": 3.5324,
+      "step": 26000
+    },
+    {
+      "epoch": 2.798407060596276,
+      "eval_accuracy": 0.3684580232063011,
+      "eval_loss": 3.5162575244903564,
+      "eval_runtime": 185.3426,
+      "eval_samples_per_second": 97.177,
+      "eval_steps_per_second": 6.075,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803788612635884,
+      "grad_norm": 0.5517733693122864,
+      "learning_rate": 0.00043230255360413744,
+      "loss": 3.5109,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8091701646754923,
+      "grad_norm": 0.5315096378326416,
+      "learning_rate": 0.00043197931257407603,
+      "loss": 3.5242,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8145517167151004,
+      "grad_norm": 0.5741550326347351,
+      "learning_rate": 0.0004316560715440146,
+      "loss": 3.5564,
+      "step": 26150
+    },
+    {
+      "epoch": 2.819933268754709,
+      "grad_norm": 0.6133660674095154,
+      "learning_rate": 0.00043133283051395317,
+      "loss": 3.5415,
+      "step": 26200
+    },
+    {
+      "epoch": 2.825314820794317,
+      "grad_norm": 0.5422095060348511,
+      "learning_rate": 0.0004310095894838918,
+      "loss": 3.4956,
+      "step": 26250
+    },
+    {
+      "epoch": 2.830696372833925,
+      "grad_norm": 0.5928894281387329,
+      "learning_rate": 0.0004306863484538304,
+      "loss": 3.5246,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8360779248735337,
+      "grad_norm": 0.5736707448959351,
+      "learning_rate": 0.00043036310742376895,
+      "loss": 3.5306,
+      "step": 26350
+    },
+    {
+      "epoch": 2.841459476913142,
+      "grad_norm": 0.611720621585846,
+      "learning_rate": 0.00043003986639370754,
+      "loss": 3.5317,
+      "step": 26400
+    },
+    {
+      "epoch": 2.84684102895275,
+      "grad_norm": 0.5900068283081055,
+      "learning_rate": 0.0004297166253636461,
+      "loss": 3.5433,
+      "step": 26450
+    },
+    {
+      "epoch": 2.852222580992358,
+      "grad_norm": 0.5795263051986694,
+      "learning_rate": 0.0004293933843335847,
+      "loss": 3.5589,
+      "step": 26500
+    },
+    {
+      "epoch": 2.857604133031966,
+      "grad_norm": 0.6196674108505249,
+      "learning_rate": 0.00042907014330352333,
+      "loss": 3.5409,
+      "step": 26550
+    },
+    {
+      "epoch": 2.8629856850715747,
+      "grad_norm": 0.5739900469779968,
+      "learning_rate": 0.00042874690227346187,
+      "loss": 3.5487,
+      "step": 26600
+    },
+    {
+      "epoch": 2.868367237111183,
+      "grad_norm": 0.6473361849784851,
+      "learning_rate": 0.00042842366124340046,
+      "loss": 3.5262,
+      "step": 26650
+    },
+    {
+      "epoch": 2.873748789150791,
+      "grad_norm": 0.5803670883178711,
+      "learning_rate": 0.00042810042021333906,
+      "loss": 3.5425,
+      "step": 26700
+    },
+    {
+      "epoch": 2.8791303411903995,
+      "grad_norm": 0.6125975847244263,
+      "learning_rate": 0.0004277771791832776,
+      "loss": 3.5435,
+      "step": 26750
+    },
+    {
+      "epoch": 2.8845118932300076,
+      "grad_norm": 0.5625960230827332,
+      "learning_rate": 0.0004274539381532162,
+      "loss": 3.5398,
+      "step": 26800
+    },
+    {
+      "epoch": 2.8898934452696157,
+      "grad_norm": 0.59404057264328,
+      "learning_rate": 0.00042713069712315484,
+      "loss": 3.5151,
+      "step": 26850
+    },
+    {
+      "epoch": 2.895274997309224,
+      "grad_norm": 0.6678456664085388,
+      "learning_rate": 0.0004268074560930934,
+      "loss": 3.5463,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9006565493488323,
+      "grad_norm": 0.667339563369751,
+      "learning_rate": 0.000426484215063032,
+      "loss": 3.5609,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9060381013884404,
+      "grad_norm": 0.603562593460083,
+      "learning_rate": 0.0004261609740329705,
+      "loss": 3.5449,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9060381013884404,
+      "eval_accuracy": 0.3694574124684187,
+      "eval_loss": 3.5066471099853516,
+      "eval_runtime": 185.8574,
+      "eval_samples_per_second": 96.908,
+      "eval_steps_per_second": 6.058,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9114196534280485,
+      "grad_norm": 0.5556718111038208,
+      "learning_rate": 0.0004258377330029091,
+      "loss": 3.5379,
+      "step": 27050
+    },
+    {
+      "epoch": 2.9168012054676566,
+      "grad_norm": 0.5727481842041016,
+      "learning_rate": 0.00042551449197284776,
+      "loss": 3.5364,
+      "step": 27100
+    },
+    {
+      "epoch": 2.922182757507265,
+      "grad_norm": 0.6576462388038635,
+      "learning_rate": 0.0004251912509427863,
+      "loss": 3.5465,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9275643095468733,
+      "grad_norm": 0.5705495476722717,
+      "learning_rate": 0.0004248680099127249,
+      "loss": 3.5189,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9329458615864814,
+      "grad_norm": 0.7621397376060486,
+      "learning_rate": 0.0004245447688826635,
+      "loss": 3.5286,
+      "step": 27250
+    },
+    {
+      "epoch": 2.93832741362609,
+      "grad_norm": 0.5782792568206787,
+      "learning_rate": 0.00042422152785260203,
+      "loss": 3.5186,
+      "step": 27300
+    },
+    {
+      "epoch": 2.943708965665698,
+      "grad_norm": 0.5814041495323181,
+      "learning_rate": 0.0004238982868225406,
+      "loss": 3.5288,
+      "step": 27350
+    },
+    {
+      "epoch": 2.949090517705306,
+      "grad_norm": 0.5303227305412292,
+      "learning_rate": 0.0004235750457924793,
+      "loss": 3.5119,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9544720697449143,
+      "grad_norm": 0.6215120553970337,
+      "learning_rate": 0.0004232518047624178,
+      "loss": 3.5261,
+      "step": 27450
+    },
+    {
+      "epoch": 2.9598536217845224,
+      "grad_norm": 0.6129084229469299,
+      "learning_rate": 0.0004229285637323564,
+      "loss": 3.5163,
+      "step": 27500
+    },
+    {
+      "epoch": 2.965235173824131,
+      "grad_norm": 0.5589991211891174,
+      "learning_rate": 0.00042260532270229495,
+      "loss": 3.5225,
+      "step": 27550
+    },
+    {
+      "epoch": 2.970616725863739,
+      "grad_norm": 0.6103320717811584,
+      "learning_rate": 0.00042228208167223354,
+      "loss": 3.5083,
+      "step": 27600
+    },
+    {
+      "epoch": 2.975998277903347,
+      "grad_norm": 0.6005710363388062,
+      "learning_rate": 0.00042195884064217214,
+      "loss": 3.5232,
+      "step": 27650
+    },
+    {
+      "epoch": 2.9813798299429557,
+      "grad_norm": 0.5742517709732056,
+      "learning_rate": 0.00042163559961211073,
+      "loss": 3.5162,
+      "step": 27700
+    },
+    {
+      "epoch": 2.986761381982564,
+      "grad_norm": 0.6228864789009094,
+      "learning_rate": 0.00042131235858204933,
+      "loss": 3.5152,
+      "step": 27750
+    },
+    {
+      "epoch": 2.992142934022172,
+      "grad_norm": 0.5636780261993408,
+      "learning_rate": 0.0004209891175519879,
+      "loss": 3.5466,
+      "step": 27800
+    },
+    {
+      "epoch": 2.9975244860617805,
+      "grad_norm": 0.5658657550811768,
+      "learning_rate": 0.00042066587652192646,
+      "loss": 3.5243,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0029060381013886,
+      "grad_norm": 0.6473289728164673,
+      "learning_rate": 0.00042034263549186506,
+      "loss": 3.4837,
+      "step": 27900
+    },
+    {
+      "epoch": 3.0082875901409967,
+      "grad_norm": 0.5536689758300781,
+      "learning_rate": 0.0004200193944618036,
+      "loss": 3.4297,
+      "step": 27950
+    },
+    {
+      "epoch": 3.0136691421806048,
+      "grad_norm": 0.608737051486969,
+      "learning_rate": 0.00041969615343174225,
+      "loss": 3.4313,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0136691421806048,
+      "eval_accuracy": 0.3700978125891972,
+      "eval_loss": 3.50144624710083,
+      "eval_runtime": 185.6059,
+      "eval_samples_per_second": 97.039,
+      "eval_steps_per_second": 6.067,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0190506942202133,
+      "grad_norm": 0.5768493413925171,
+      "learning_rate": 0.00041937291240168084,
+      "loss": 3.4541,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0244322462598214,
+      "grad_norm": 0.5930345058441162,
+      "learning_rate": 0.0004190496713716194,
+      "loss": 3.4356,
+      "step": 28100
+    },
+    {
+      "epoch": 3.0298137982994295,
+      "grad_norm": 0.6201900243759155,
+      "learning_rate": 0.000418726430341558,
+      "loss": 3.4185,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0351953503390376,
+      "grad_norm": 0.5873003005981445,
+      "learning_rate": 0.00041840318931149657,
+      "loss": 3.4312,
+      "step": 28200
+    },
+    {
+      "epoch": 3.040576902378646,
+      "grad_norm": 0.6229293346405029,
+      "learning_rate": 0.00041807994828143517,
+      "loss": 3.4371,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0459584544182543,
+      "grad_norm": 0.6742472052574158,
+      "learning_rate": 0.00041775670725137376,
+      "loss": 3.4465,
+      "step": 28300
+    },
+    {
+      "epoch": 3.0513400064578624,
+      "grad_norm": 0.56894850730896,
+      "learning_rate": 0.00041743346622131236,
+      "loss": 3.4321,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0567215584974705,
+      "grad_norm": 0.5871414542198181,
+      "learning_rate": 0.0004171102251912509,
+      "loss": 3.4307,
+      "step": 28400
+    },
+    {
+      "epoch": 3.062103110537079,
+      "grad_norm": 0.6387337446212769,
+      "learning_rate": 0.0004167869841611895,
+      "loss": 3.4524,
+      "step": 28450
+    },
+    {
+      "epoch": 3.067484662576687,
+      "grad_norm": 0.589634120464325,
+      "learning_rate": 0.00041646374313112803,
+      "loss": 3.4438,
+      "step": 28500
+    },
+    {
+      "epoch": 3.0728662146162953,
+      "grad_norm": 0.5635669231414795,
+      "learning_rate": 0.0004161405021010667,
+      "loss": 3.4422,
+      "step": 28550
+    },
+    {
+      "epoch": 3.0782477666559034,
+      "grad_norm": 0.7147294878959656,
+      "learning_rate": 0.0004158172610710053,
+      "loss": 3.4463,
+      "step": 28600
+    },
+    {
+      "epoch": 3.083629318695512,
+      "grad_norm": 0.6518290042877197,
+      "learning_rate": 0.0004154940200409438,
+      "loss": 3.4587,
+      "step": 28650
+    },
+    {
+      "epoch": 3.08901087073512,
+      "grad_norm": 0.5564368963241577,
+      "learning_rate": 0.0004151707790108824,
+      "loss": 3.4545,
+      "step": 28700
+    },
+    {
+      "epoch": 3.094392422774728,
+      "grad_norm": 0.5743803977966309,
+      "learning_rate": 0.00041484753798082095,
+      "loss": 3.4632,
+      "step": 28750
+    },
+    {
+      "epoch": 3.0997739748143363,
+      "grad_norm": 0.6107898950576782,
+      "learning_rate": 0.00041452429695075955,
+      "loss": 3.4346,
+      "step": 28800
+    },
+    {
+      "epoch": 3.105155526853945,
+      "grad_norm": 0.5537232160568237,
+      "learning_rate": 0.0004142010559206982,
+      "loss": 3.4531,
+      "step": 28850
+    },
+    {
+      "epoch": 3.110537078893553,
+      "grad_norm": 0.5820683836936951,
+      "learning_rate": 0.00041387781489063673,
+      "loss": 3.4577,
+      "step": 28900
+    },
+    {
+      "epoch": 3.115918630933161,
+      "grad_norm": 0.638676106929779,
+      "learning_rate": 0.00041355457386057533,
+      "loss": 3.4499,
+      "step": 28950
+    },
+    {
+      "epoch": 3.121300182972769,
+      "grad_norm": 0.616300642490387,
+      "learning_rate": 0.0004132313328305139,
+      "loss": 3.4477,
+      "step": 29000
+    },
+    {
+      "epoch": 3.121300182972769,
+      "eval_accuracy": 0.370598702402022,
+      "eval_loss": 3.5001132488250732,
+      "eval_runtime": 185.7876,
+      "eval_samples_per_second": 96.944,
+      "eval_steps_per_second": 6.061,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266817350123777,
+      "grad_norm": 0.6168830394744873,
+      "learning_rate": 0.00041290809180045246,
+      "loss": 3.4421,
+      "step": 29050
+    },
+    {
+      "epoch": 3.132063287051986,
+      "grad_norm": 0.639548659324646,
+      "learning_rate": 0.0004125848507703911,
+      "loss": 3.441,
+      "step": 29100
+    },
+    {
+      "epoch": 3.137444839091594,
+      "grad_norm": 0.5794745087623596,
+      "learning_rate": 0.00041226807456093087,
+      "loss": 3.4578,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1428263911312024,
+      "grad_norm": 0.5894873142242432,
+      "learning_rate": 0.0004119448335308695,
+      "loss": 3.455,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1482079431708105,
+      "grad_norm": 0.637736976146698,
+      "learning_rate": 0.0004116215925008081,
+      "loss": 3.4645,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1535894952104186,
+      "grad_norm": 0.5710036158561707,
+      "learning_rate": 0.00041129835147074665,
+      "loss": 3.4569,
+      "step": 29300
+    },
+    {
+      "epoch": 3.1589710472500268,
+      "grad_norm": 0.6197898983955383,
+      "learning_rate": 0.00041097511044068524,
+      "loss": 3.4523,
+      "step": 29350
+    },
+    {
+      "epoch": 3.1643525992896353,
+      "grad_norm": 0.5881309509277344,
+      "learning_rate": 0.0004106518694106238,
+      "loss": 3.4346,
+      "step": 29400
+    },
+    {
+      "epoch": 3.1697341513292434,
+      "grad_norm": 0.6068123579025269,
+      "learning_rate": 0.0004103286283805624,
+      "loss": 3.4547,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1751157033688515,
+      "grad_norm": 0.6294363141059875,
+      "learning_rate": 0.00041000538735050103,
+      "loss": 3.4433,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1804972554084596,
+      "grad_norm": 0.6260930299758911,
+      "learning_rate": 0.00040968214632043957,
+      "loss": 3.4388,
+      "step": 29550
+    },
+    {
+      "epoch": 3.185878807448068,
+      "grad_norm": 0.6103339791297913,
+      "learning_rate": 0.00040935890529037816,
+      "loss": 3.4646,
+      "step": 29600
+    },
+    {
+      "epoch": 3.1912603594876763,
+      "grad_norm": 0.6059884428977966,
+      "learning_rate": 0.00040903566426031676,
+      "loss": 3.4592,
+      "step": 29650
+    },
+    {
+      "epoch": 3.1966419115272844,
+      "grad_norm": 0.6250834465026855,
+      "learning_rate": 0.0004087124232302553,
+      "loss": 3.4505,
+      "step": 29700
+    },
+    {
+      "epoch": 3.2020234635668925,
+      "grad_norm": 0.6510362029075623,
+      "learning_rate": 0.0004083891822001939,
+      "loss": 3.4728,
+      "step": 29750
+    },
+    {
+      "epoch": 3.207405015606501,
+      "grad_norm": 0.6216087341308594,
+      "learning_rate": 0.00040806594117013254,
+      "loss": 3.4618,
+      "step": 29800
+    },
+    {
+      "epoch": 3.212786567646109,
+      "grad_norm": 0.6058763861656189,
+      "learning_rate": 0.0004077427001400711,
+      "loss": 3.4673,
+      "step": 29850
+    },
+    {
+      "epoch": 3.2181681196857173,
+      "grad_norm": 0.6276309490203857,
+      "learning_rate": 0.0004074194591100097,
+      "loss": 3.4602,
+      "step": 29900
+    },
+    {
+      "epoch": 3.2235496717253254,
+      "grad_norm": 0.5934738516807556,
+      "learning_rate": 0.0004070962180799482,
+      "loss": 3.4645,
+      "step": 29950
+    },
+    {
+      "epoch": 3.228931223764934,
+      "grad_norm": 0.6110907793045044,
+      "learning_rate": 0.0004067729770498868,
+      "loss": 3.4502,
+      "step": 30000
+    },
+    {
+      "epoch": 3.228931223764934,
+      "eval_accuracy": 0.37125866004260716,
+      "eval_loss": 3.492161750793457,
+      "eval_runtime": 185.333,
+      "eval_samples_per_second": 97.182,
+      "eval_steps_per_second": 6.076,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234312775804542,
+      "grad_norm": 0.5893679261207581,
+      "learning_rate": 0.00040644973601982546,
+      "loss": 3.4496,
+      "step": 30050
+    },
+    {
+      "epoch": 3.23969432784415,
+      "grad_norm": 0.604153573513031,
+      "learning_rate": 0.000406126494989764,
+      "loss": 3.479,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2450758798837587,
+      "grad_norm": 0.6097456216812134,
+      "learning_rate": 0.0004058032539597026,
+      "loss": 3.4635,
+      "step": 30150
+    },
+    {
+      "epoch": 3.250457431923367,
+      "grad_norm": 0.6210144758224487,
+      "learning_rate": 0.00040548001292964114,
+      "loss": 3.4591,
+      "step": 30200
+    },
+    {
+      "epoch": 3.255838983962975,
+      "grad_norm": 0.6128044724464417,
+      "learning_rate": 0.00040515677189957973,
+      "loss": 3.4472,
+      "step": 30250
+    },
+    {
+      "epoch": 3.261220536002583,
+      "grad_norm": 0.6151296496391296,
+      "learning_rate": 0.0004048335308695183,
+      "loss": 3.455,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2666020880421915,
+      "grad_norm": 0.641426146030426,
+      "learning_rate": 0.0004045102898394569,
+      "loss": 3.4667,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2719836400817996,
+      "grad_norm": 0.6009118556976318,
+      "learning_rate": 0.0004041870488093955,
+      "loss": 3.4466,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2773651921214078,
+      "grad_norm": 0.6047423481941223,
+      "learning_rate": 0.0004038638077793341,
+      "loss": 3.476,
+      "step": 30450
+    },
+    {
+      "epoch": 3.282746744161016,
+      "grad_norm": 0.590869128704071,
+      "learning_rate": 0.00040354056674927265,
+      "loss": 3.4808,
+      "step": 30500
+    },
+    {
+      "epoch": 3.2881282962006244,
+      "grad_norm": 0.6050732731819153,
+      "learning_rate": 0.00040321732571921124,
+      "loss": 3.4611,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2935098482402325,
+      "grad_norm": 0.661838710308075,
+      "learning_rate": 0.0004028940846891498,
+      "loss": 3.4573,
+      "step": 30600
+    },
+    {
+      "epoch": 3.2988914002798406,
+      "grad_norm": 0.6257044076919556,
+      "learning_rate": 0.00040257084365908843,
+      "loss": 3.454,
+      "step": 30650
+    },
+    {
+      "epoch": 3.304272952319449,
+      "grad_norm": 0.6638705730438232,
+      "learning_rate": 0.00040224760262902703,
+      "loss": 3.4776,
+      "step": 30700
+    },
+    {
+      "epoch": 3.3096545043590573,
+      "grad_norm": 0.6136927604675293,
+      "learning_rate": 0.00040192436159896557,
+      "loss": 3.4531,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3150360563986654,
+      "grad_norm": 0.6285628080368042,
+      "learning_rate": 0.00040160112056890416,
+      "loss": 3.4652,
+      "step": 30800
+    },
+    {
+      "epoch": 3.3204176084382735,
+      "grad_norm": 0.6483205556869507,
+      "learning_rate": 0.00040127787953884276,
+      "loss": 3.4742,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3257991604778816,
+      "grad_norm": 0.6225518584251404,
+      "learning_rate": 0.00040095463850878135,
+      "loss": 3.4749,
+      "step": 30900
+    },
+    {
+      "epoch": 3.33118071251749,
+      "grad_norm": 0.6386218070983887,
+      "learning_rate": 0.00040063139747871995,
+      "loss": 3.4612,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3365622645570983,
+      "grad_norm": 0.5485774278640747,
+      "learning_rate": 0.00040030815644865854,
+      "loss": 3.4542,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3365622645570983,
+      "eval_accuracy": 0.3717821496560973,
+      "eval_loss": 3.4868366718292236,
+      "eval_runtime": 185.675,
+      "eval_samples_per_second": 97.003,
+      "eval_steps_per_second": 6.064,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3419438165967064,
+      "grad_norm": 0.6492092609405518,
+      "learning_rate": 0.0003999849154185971,
+      "loss": 3.4707,
+      "step": 31050
+    },
+    {
+      "epoch": 3.347325368636315,
+      "grad_norm": 0.6124265789985657,
+      "learning_rate": 0.0003996616743885357,
+      "loss": 3.4668,
+      "step": 31100
+    },
+    {
+      "epoch": 3.352706920675923,
+      "grad_norm": 0.5960283279418945,
+      "learning_rate": 0.0003993384333584742,
+      "loss": 3.4744,
+      "step": 31150
+    },
+    {
+      "epoch": 3.358088472715531,
+      "grad_norm": 0.618414580821991,
+      "learning_rate": 0.0003990216571490141,
+      "loss": 3.4619,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3634700247551392,
+      "grad_norm": 0.6832355260848999,
+      "learning_rate": 0.0003986984161189526,
+      "loss": 3.4388,
+      "step": 31250
+    },
+    {
+      "epoch": 3.368851576794748,
+      "grad_norm": 0.6463454961776733,
+      "learning_rate": 0.00039837517508889127,
+      "loss": 3.4524,
+      "step": 31300
+    },
+    {
+      "epoch": 3.374233128834356,
+      "grad_norm": 0.6139732599258423,
+      "learning_rate": 0.00039805193405882986,
+      "loss": 3.4568,
+      "step": 31350
+    },
+    {
+      "epoch": 3.379614680873964,
+      "grad_norm": 0.6096798181533813,
+      "learning_rate": 0.0003977286930287684,
+      "loss": 3.448,
+      "step": 31400
+    },
+    {
+      "epoch": 3.384996232913572,
+      "grad_norm": 0.6143977046012878,
+      "learning_rate": 0.000397405451998707,
+      "loss": 3.4444,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3903777849531807,
+      "grad_norm": 0.6604296565055847,
+      "learning_rate": 0.00039708221096864554,
+      "loss": 3.4501,
+      "step": 31500
+    },
+    {
+      "epoch": 3.3957593369927888,
+      "grad_norm": 0.6364625096321106,
+      "learning_rate": 0.00039675896993858413,
+      "loss": 3.4474,
+      "step": 31550
+    },
+    {
+      "epoch": 3.401140889032397,
+      "grad_norm": 0.6189888715744019,
+      "learning_rate": 0.0003964357289085228,
+      "loss": 3.462,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4065224410720054,
+      "grad_norm": 0.5984906554222107,
+      "learning_rate": 0.0003961124878784613,
+      "loss": 3.4532,
+      "step": 31650
+    },
+    {
+      "epoch": 3.4119039931116135,
+      "grad_norm": 0.5987201929092407,
+      "learning_rate": 0.0003957892468483999,
+      "loss": 3.4789,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4172855451512216,
+      "grad_norm": 0.631290078163147,
+      "learning_rate": 0.0003954660058183385,
+      "loss": 3.4595,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4226670971908297,
+      "grad_norm": 0.6193310618400574,
+      "learning_rate": 0.00039514276478827705,
+      "loss": 3.4578,
+      "step": 31800
+    },
+    {
+      "epoch": 3.428048649230438,
+      "grad_norm": 0.6147798299789429,
+      "learning_rate": 0.0003948195237582157,
+      "loss": 3.4604,
+      "step": 31850
+    },
+    {
+      "epoch": 3.4334302012700464,
+      "grad_norm": 0.5727043747901917,
+      "learning_rate": 0.0003944962827281543,
+      "loss": 3.4634,
+      "step": 31900
+    },
+    {
+      "epoch": 3.4388117533096545,
+      "grad_norm": 0.5698534846305847,
+      "learning_rate": 0.00039417304169809284,
+      "loss": 3.4719,
+      "step": 31950
+    },
+    {
+      "epoch": 3.4441933053492626,
+      "grad_norm": 0.6117491126060486,
+      "learning_rate": 0.00039384980066803143,
+      "loss": 3.4608,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4441933053492626,
+      "eval_accuracy": 0.3729168117635439,
+      "eval_loss": 3.478281021118164,
+      "eval_runtime": 185.3871,
+      "eval_samples_per_second": 97.153,
+      "eval_steps_per_second": 6.074,
+      "step": 32000
+    },
+    {
+      "epoch": 3.449574857388871,
+      "grad_norm": 0.6180813312530518,
+      "learning_rate": 0.00039352655963796997,
+      "loss": 3.4454,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4549564094284793,
+      "grad_norm": 0.6341642737388611,
+      "learning_rate": 0.00039320331860790857,
+      "loss": 3.4648,
+      "step": 32100
+    },
+    {
+      "epoch": 3.4603379614680874,
+      "grad_norm": 0.6050530672073364,
+      "learning_rate": 0.0003928800775778472,
+      "loss": 3.4475,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4657195135076955,
+      "grad_norm": 0.6123150587081909,
+      "learning_rate": 0.00039255683654778576,
+      "loss": 3.472,
+      "step": 32200
+    },
+    {
+      "epoch": 3.471101065547304,
+      "grad_norm": 0.6094672083854675,
+      "learning_rate": 0.00039223359551772435,
+      "loss": 3.4547,
+      "step": 32250
+    },
+    {
+      "epoch": 3.476482617586912,
+      "grad_norm": 0.6197632551193237,
+      "learning_rate": 0.00039191035448766294,
+      "loss": 3.4647,
+      "step": 32300
+    },
+    {
+      "epoch": 3.4818641696265202,
+      "grad_norm": 0.5918575525283813,
+      "learning_rate": 0.0003915871134576015,
+      "loss": 3.452,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4872457216661283,
+      "grad_norm": 0.5822188854217529,
+      "learning_rate": 0.0003912638724275401,
+      "loss": 3.4672,
+      "step": 32400
+    },
+    {
+      "epoch": 3.492627273705737,
+      "grad_norm": 0.5970443487167358,
+      "learning_rate": 0.00039094063139747873,
+      "loss": 3.4638,
+      "step": 32450
+    },
+    {
+      "epoch": 3.498008825745345,
+      "grad_norm": 0.6250706911087036,
+      "learning_rate": 0.00039061739036741727,
+      "loss": 3.4542,
+      "step": 32500
+    },
+    {
+      "epoch": 3.503390377784953,
+      "grad_norm": 0.6195428967475891,
+      "learning_rate": 0.00039029414933735586,
+      "loss": 3.4535,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5087719298245617,
+      "grad_norm": 0.6094918847084045,
+      "learning_rate": 0.0003899709083072944,
+      "loss": 3.452,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5141534818641698,
+      "grad_norm": 0.6326495409011841,
+      "learning_rate": 0.000389647667277233,
+      "loss": 3.482,
+      "step": 32650
+    },
+    {
+      "epoch": 3.519535033903778,
+      "grad_norm": 0.5670629739761353,
+      "learning_rate": 0.00038932442624717165,
+      "loss": 3.4634,
+      "step": 32700
+    },
+    {
+      "epoch": 3.524916585943386,
+      "grad_norm": 0.5876345634460449,
+      "learning_rate": 0.0003890011852171102,
+      "loss": 3.4595,
+      "step": 32750
+    },
+    {
+      "epoch": 3.530298137982994,
+      "grad_norm": 0.6131546497344971,
+      "learning_rate": 0.0003886779441870488,
+      "loss": 3.4486,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5356796900226026,
+      "grad_norm": 0.6190427541732788,
+      "learning_rate": 0.0003883547031569874,
+      "loss": 3.4705,
+      "step": 32850
+    },
+    {
+      "epoch": 3.5410612420622107,
+      "grad_norm": 0.5621001124382019,
+      "learning_rate": 0.0003880379269475272,
+      "loss": 3.4763,
+      "step": 32900
+    },
+    {
+      "epoch": 3.546442794101819,
+      "grad_norm": 0.5962037444114685,
+      "learning_rate": 0.0003877146859174657,
+      "loss": 3.4385,
+      "step": 32950
+    },
+    {
+      "epoch": 3.5518243461414274,
+      "grad_norm": 0.6305086016654968,
+      "learning_rate": 0.0003873914448874043,
+      "loss": 3.4684,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5518243461414274,
+      "eval_accuracy": 0.37310097840838946,
+      "eval_loss": 3.4740188121795654,
+      "eval_runtime": 185.7362,
+      "eval_samples_per_second": 96.971,
+      "eval_steps_per_second": 6.062,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5572058981810355,
+      "grad_norm": 0.6291715502738953,
+      "learning_rate": 0.0003870682038573429,
+      "loss": 3.4518,
+      "step": 33050
+    },
+    {
+      "epoch": 3.5625874502206436,
+      "grad_norm": 0.6375169157981873,
+      "learning_rate": 0.0003867449628272815,
+      "loss": 3.4557,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5679690022602517,
+      "grad_norm": 0.6188652515411377,
+      "learning_rate": 0.0003864217217972201,
+      "loss": 3.4496,
+      "step": 33150
+    },
+    {
+      "epoch": 3.57335055429986,
+      "grad_norm": 0.6463585495948792,
+      "learning_rate": 0.0003860984807671587,
+      "loss": 3.4615,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5787321063394684,
+      "grad_norm": 0.6207422614097595,
+      "learning_rate": 0.00038577523973709724,
+      "loss": 3.4736,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5841136583790765,
+      "grad_norm": 0.6484593749046326,
+      "learning_rate": 0.00038545199870703583,
+      "loss": 3.4399,
+      "step": 33300
+    },
+    {
+      "epoch": 3.5894952104186846,
+      "grad_norm": 0.5928724408149719,
+      "learning_rate": 0.0003851287576769744,
+      "loss": 3.4583,
+      "step": 33350
+    },
+    {
+      "epoch": 3.594876762458293,
+      "grad_norm": 0.6784103512763977,
+      "learning_rate": 0.000384805516646913,
+      "loss": 3.4701,
+      "step": 33400
+    },
+    {
+      "epoch": 3.6002583144979012,
+      "grad_norm": 0.6526338458061218,
+      "learning_rate": 0.0003844822756168516,
+      "loss": 3.4453,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6056398665375093,
+      "grad_norm": 0.6782744526863098,
+      "learning_rate": 0.00038415903458679016,
+      "loss": 3.4555,
+      "step": 33500
+    },
+    {
+      "epoch": 3.611021418577118,
+      "grad_norm": 0.583367109298706,
+      "learning_rate": 0.00038383579355672875,
+      "loss": 3.4592,
+      "step": 33550
+    },
+    {
+      "epoch": 3.616402970616726,
+      "grad_norm": 0.558493971824646,
+      "learning_rate": 0.00038351255252666735,
+      "loss": 3.4465,
+      "step": 33600
+    },
+    {
+      "epoch": 3.621784522656334,
+      "grad_norm": 0.595182478427887,
+      "learning_rate": 0.00038318931149660594,
+      "loss": 3.46,
+      "step": 33650
+    },
+    {
+      "epoch": 3.627166074695942,
+      "grad_norm": 0.627747654914856,
+      "learning_rate": 0.00038286607046654454,
+      "loss": 3.4611,
+      "step": 33700
+    },
+    {
+      "epoch": 3.6325476267355503,
+      "grad_norm": 0.5847141146659851,
+      "learning_rate": 0.00038254282943648313,
+      "loss": 3.472,
+      "step": 33750
+    },
+    {
+      "epoch": 3.637929178775159,
+      "grad_norm": 0.6564003229141235,
+      "learning_rate": 0.00038221958840642167,
+      "loss": 3.4705,
+      "step": 33800
+    },
+    {
+      "epoch": 3.643310730814767,
+      "grad_norm": 0.5658453702926636,
+      "learning_rate": 0.00038189634737636027,
+      "loss": 3.4444,
+      "step": 33850
+    },
+    {
+      "epoch": 3.648692282854375,
+      "grad_norm": 0.6502963900566101,
+      "learning_rate": 0.0003815731063462988,
+      "loss": 3.4673,
+      "step": 33900
+    },
+    {
+      "epoch": 3.6540738348939836,
+      "grad_norm": 0.6217418313026428,
+      "learning_rate": 0.00038124986531623745,
+      "loss": 3.4785,
+      "step": 33950
+    },
+    {
+      "epoch": 3.6594553869335917,
+      "grad_norm": 0.5975374579429626,
+      "learning_rate": 0.00038092662428617605,
+      "loss": 3.4391,
+      "step": 34000
+    },
+    {
+      "epoch": 3.6594553869335917,
+      "eval_accuracy": 0.3740735304072169,
+      "eval_loss": 3.4671518802642822,
+      "eval_runtime": 185.7436,
+      "eval_samples_per_second": 96.967,
+      "eval_steps_per_second": 6.062,
+      "step": 34000
+    },
+    {
+      "epoch": 3.6648369389732,
+      "grad_norm": 0.605006992816925,
+      "learning_rate": 0.0003806033832561146,
+      "loss": 3.4693,
+      "step": 34050
+    },
+    {
+      "epoch": 3.670218491012808,
+      "grad_norm": 0.6553669571876526,
+      "learning_rate": 0.0003802801422260532,
+      "loss": 3.4804,
+      "step": 34100
+    },
+    {
+      "epoch": 3.675600043052416,
+      "grad_norm": 0.6531872749328613,
+      "learning_rate": 0.0003799569011959918,
+      "loss": 3.4517,
+      "step": 34150
+    },
+    {
+      "epoch": 3.6809815950920246,
+      "grad_norm": 0.5793938040733337,
+      "learning_rate": 0.0003796336601659303,
+      "loss": 3.4502,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6863631471316327,
+      "grad_norm": 0.6113420128822327,
+      "learning_rate": 0.00037931041913586897,
+      "loss": 3.4412,
+      "step": 34250
+    },
+    {
+      "epoch": 3.691744699171241,
+      "grad_norm": 0.6133025884628296,
+      "learning_rate": 0.00037898717810580756,
+      "loss": 3.4735,
+      "step": 34300
+    },
+    {
+      "epoch": 3.6971262512108494,
+      "grad_norm": 0.596380889415741,
+      "learning_rate": 0.0003786639370757461,
+      "loss": 3.4553,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7025078032504575,
+      "grad_norm": 0.6523725986480713,
+      "learning_rate": 0.0003783406960456847,
+      "loss": 3.4543,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7078893552900656,
+      "grad_norm": 0.6609398722648621,
+      "learning_rate": 0.00037801745501562324,
+      "loss": 3.45,
+      "step": 34450
+    },
+    {
+      "epoch": 3.713270907329674,
+      "grad_norm": 0.6276677846908569,
+      "learning_rate": 0.0003776942139855619,
+      "loss": 3.452,
+      "step": 34500
+    },
+    {
+      "epoch": 3.7186524593692822,
+      "grad_norm": 0.6465655565261841,
+      "learning_rate": 0.0003773709729555005,
+      "loss": 3.457,
+      "step": 34550
+    },
+    {
+      "epoch": 3.7240340114088903,
+      "grad_norm": 0.6079691052436829,
+      "learning_rate": 0.000377047731925439,
+      "loss": 3.4567,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7294155634484984,
+      "grad_norm": 0.5890680551528931,
+      "learning_rate": 0.0003767244908953776,
+      "loss": 3.4682,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7347971154881066,
+      "grad_norm": 0.6379863023757935,
+      "learning_rate": 0.0003764012498653162,
+      "loss": 3.4567,
+      "step": 34700
+    },
+    {
+      "epoch": 3.740178667527715,
+      "grad_norm": 0.6383402943611145,
+      "learning_rate": 0.00037607800883525475,
+      "loss": 3.4574,
+      "step": 34750
+    },
+    {
+      "epoch": 3.745560219567323,
+      "grad_norm": 0.5963407158851624,
+      "learning_rate": 0.0003757547678051934,
+      "loss": 3.4615,
+      "step": 34800
+    },
+    {
+      "epoch": 3.7509417716069313,
+      "grad_norm": 0.610348641872406,
+      "learning_rate": 0.000375431526775132,
+      "loss": 3.4562,
+      "step": 34850
+    },
+    {
+      "epoch": 3.75632332364654,
+      "grad_norm": 0.6593793034553528,
+      "learning_rate": 0.00037510828574507054,
+      "loss": 3.4678,
+      "step": 34900
+    },
+    {
+      "epoch": 3.761704875686148,
+      "grad_norm": 0.6169629693031311,
+      "learning_rate": 0.00037478504471500913,
+      "loss": 3.4711,
+      "step": 34950
+    },
+    {
+      "epoch": 3.767086427725756,
+      "grad_norm": 0.6133807301521301,
+      "learning_rate": 0.00037446180368494767,
+      "loss": 3.4523,
+      "step": 35000
+    },
+    {
+      "epoch": 3.767086427725756,
+      "eval_accuracy": 0.37458495955015964,
+      "eval_loss": 3.461313486099243,
+      "eval_runtime": 185.9117,
+      "eval_samples_per_second": 96.879,
+      "eval_steps_per_second": 6.057,
+      "step": 35000
+    },
+    {
+      "epoch": 3.772467979765364,
+      "grad_norm": 0.6491101980209351,
+      "learning_rate": 0.00037413856265488627,
+      "loss": 3.4394,
+      "step": 35050
+    },
+    {
+      "epoch": 3.7778495318049723,
+      "grad_norm": 0.5951021313667297,
+      "learning_rate": 0.0003738153216248249,
+      "loss": 3.4476,
+      "step": 35100
+    },
+    {
+      "epoch": 3.783231083844581,
+      "grad_norm": 0.5845459699630737,
+      "learning_rate": 0.00037349208059476346,
+      "loss": 3.4787,
+      "step": 35150
+    },
+    {
+      "epoch": 3.788612635884189,
+      "grad_norm": 0.6395171880722046,
+      "learning_rate": 0.00037316883956470205,
+      "loss": 3.4353,
+      "step": 35200
+    },
+    {
+      "epoch": 3.793994187923797,
+      "grad_norm": 0.6753062009811401,
+      "learning_rate": 0.00037284559853464064,
+      "loss": 3.469,
+      "step": 35250
+    },
+    {
+      "epoch": 3.7993757399634056,
+      "grad_norm": 0.6756654977798462,
+      "learning_rate": 0.0003725223575045792,
+      "loss": 3.461,
+      "step": 35300
+    },
+    {
+      "epoch": 3.8047572920030137,
+      "grad_norm": 0.6024090051651001,
+      "learning_rate": 0.0003721991164745178,
+      "loss": 3.4493,
+      "step": 35350
+    },
+    {
+      "epoch": 3.810138844042622,
+      "grad_norm": 0.6586195230484009,
+      "learning_rate": 0.00037187587544445643,
+      "loss": 3.4611,
+      "step": 35400
+    },
+    {
+      "epoch": 3.8155203960822304,
+      "grad_norm": 0.6518551111221313,
+      "learning_rate": 0.00037155263441439497,
+      "loss": 3.4629,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8209019481218385,
+      "grad_norm": 0.6372223496437073,
+      "learning_rate": 0.00037122939338433356,
+      "loss": 3.4726,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8262835001614466,
+      "grad_norm": 0.6337680220603943,
+      "learning_rate": 0.0003709061523542721,
+      "loss": 3.4411,
+      "step": 35550
+    },
+    {
+      "epoch": 3.8316650522010547,
+      "grad_norm": 0.5795866250991821,
+      "learning_rate": 0.0003705829113242107,
+      "loss": 3.4343,
+      "step": 35600
+    },
+    {
+      "epoch": 3.837046604240663,
+      "grad_norm": 0.6412902474403381,
+      "learning_rate": 0.00037025967029414935,
+      "loss": 3.4705,
+      "step": 35650
+    },
+    {
+      "epoch": 3.8424281562802713,
+      "grad_norm": 0.6131452918052673,
+      "learning_rate": 0.0003699364292640879,
+      "loss": 3.4557,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8478097083198795,
+      "grad_norm": 0.6585981249809265,
+      "learning_rate": 0.0003696131882340265,
+      "loss": 3.4658,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8531912603594876,
+      "grad_norm": 0.6409697532653809,
+      "learning_rate": 0.0003692899472039651,
+      "loss": 3.4591,
+      "step": 35800
+    },
+    {
+      "epoch": 3.858572812399096,
+      "grad_norm": 0.6385776996612549,
+      "learning_rate": 0.0003689667061739036,
+      "loss": 3.4378,
+      "step": 35850
+    },
+    {
+      "epoch": 3.863954364438704,
+      "grad_norm": 0.5871413350105286,
+      "learning_rate": 0.0003686434651438422,
+      "loss": 3.4511,
+      "step": 35900
+    },
+    {
+      "epoch": 3.8693359164783123,
+      "grad_norm": 0.6817817687988281,
+      "learning_rate": 0.00036832022411378086,
+      "loss": 3.4607,
+      "step": 35950
+    },
+    {
+      "epoch": 3.8747174685179204,
+      "grad_norm": 0.5841816663742065,
+      "learning_rate": 0.0003679969830837194,
+      "loss": 3.4294,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8747174685179204,
+      "eval_accuracy": 0.37546569985878386,
+      "eval_loss": 3.453197717666626,
+      "eval_runtime": 185.5043,
+      "eval_samples_per_second": 97.092,
+      "eval_steps_per_second": 6.07,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8800990205575285,
+      "grad_norm": 0.632652997970581,
+      "learning_rate": 0.000367673742053658,
+      "loss": 3.4522,
+      "step": 36050
+    },
+    {
+      "epoch": 3.885480572597137,
+      "grad_norm": 0.5645267367362976,
+      "learning_rate": 0.00036735050102359654,
+      "loss": 3.4583,
+      "step": 36100
+    },
+    {
+      "epoch": 3.890862124636745,
+      "grad_norm": 0.639022707939148,
+      "learning_rate": 0.00036702725999353513,
+      "loss": 3.4552,
+      "step": 36150
+    },
+    {
+      "epoch": 3.8962436766763533,
+      "grad_norm": 0.7405444979667664,
+      "learning_rate": 0.0003667040189634737,
+      "loss": 3.4483,
+      "step": 36200
+    },
+    {
+      "epoch": 3.901625228715962,
+      "grad_norm": 0.6363329887390137,
+      "learning_rate": 0.0003663807779334123,
+      "loss": 3.4391,
+      "step": 36250
+    },
+    {
+      "epoch": 3.90700678075557,
+      "grad_norm": 0.6380723714828491,
+      "learning_rate": 0.0003660575369033509,
+      "loss": 3.4614,
+      "step": 36300
+    },
+    {
+      "epoch": 3.912388332795178,
+      "grad_norm": 0.6373493075370789,
+      "learning_rate": 0.0003657342958732895,
+      "loss": 3.429,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9177698848347866,
+      "grad_norm": 0.5919589400291443,
+      "learning_rate": 0.00036541105484322805,
+      "loss": 3.4594,
+      "step": 36400
+    },
+    {
+      "epoch": 3.9231514368743947,
+      "grad_norm": 0.6067710518836975,
+      "learning_rate": 0.00036508781381316665,
+      "loss": 3.4619,
+      "step": 36450
+    },
+    {
+      "epoch": 3.928532988914003,
+      "grad_norm": 0.6119357347488403,
+      "learning_rate": 0.0003647645727831053,
+      "loss": 3.46,
+      "step": 36500
+    },
+    {
+      "epoch": 3.933914540953611,
+      "grad_norm": 0.6200315952301025,
+      "learning_rate": 0.00036444133175304384,
+      "loss": 3.4414,
+      "step": 36550
+    },
+    {
+      "epoch": 3.939296092993219,
+      "grad_norm": 0.6138097643852234,
+      "learning_rate": 0.00036411809072298243,
+      "loss": 3.4413,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9446776450328276,
+      "grad_norm": 0.6153903007507324,
+      "learning_rate": 0.00036379484969292097,
+      "loss": 3.4583,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9500591970724357,
+      "grad_norm": 0.6427406668663025,
+      "learning_rate": 0.00036347160866285956,
+      "loss": 3.4608,
+      "step": 36700
+    },
+    {
+      "epoch": 3.955440749112044,
+      "grad_norm": 0.6901724338531494,
+      "learning_rate": 0.0003631483676327981,
+      "loss": 3.4419,
+      "step": 36750
+    },
+    {
+      "epoch": 3.9608223011516523,
+      "grad_norm": 0.625988781452179,
+      "learning_rate": 0.00036282512660273675,
+      "loss": 3.4407,
+      "step": 36800
+    },
+    {
+      "epoch": 3.9662038531912605,
+      "grad_norm": 0.5625736117362976,
+      "learning_rate": 0.00036250188557267535,
+      "loss": 3.4409,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9715854052308686,
+      "grad_norm": 0.5924127697944641,
+      "learning_rate": 0.0003621786445426139,
+      "loss": 3.4583,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9769669572704767,
+      "grad_norm": 0.6781776547431946,
+      "learning_rate": 0.00036186186833315375,
+      "loss": 3.4488,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9823485093100848,
+      "grad_norm": 0.6270511150360107,
+      "learning_rate": 0.0003615386273030923,
+      "loss": 3.4562,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9823485093100848,
+      "eval_accuracy": 0.37570354103020975,
+      "eval_loss": 3.450030565261841,
+      "eval_runtime": 185.9889,
+      "eval_samples_per_second": 96.839,
+      "eval_steps_per_second": 6.054,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9877300613496933,
+      "grad_norm": 0.6131006479263306,
+      "learning_rate": 0.0003612153862730309,
+      "loss": 3.4413,
+      "step": 37050
+    },
+    {
+      "epoch": 3.9931116133893014,
+      "grad_norm": 0.6621198058128357,
+      "learning_rate": 0.0003608921452429695,
+      "loss": 3.4412,
+      "step": 37100
+    },
+    {
+      "epoch": 3.9984931654289095,
+      "grad_norm": 0.6415898203849792,
+      "learning_rate": 0.000360568904212908,
+      "loss": 3.4632,
+      "step": 37150
+    },
+    {
+      "epoch": 4.003874717468518,
+      "grad_norm": 0.6411539912223816,
+      "learning_rate": 0.00036024566318284667,
+      "loss": 3.3788,
+      "step": 37200
+    },
+    {
+      "epoch": 4.009256269508126,
+      "grad_norm": 0.6376098990440369,
+      "learning_rate": 0.00035992242215278526,
+      "loss": 3.3306,
+      "step": 37250
+    },
+    {
+      "epoch": 4.014637821547734,
+      "grad_norm": 0.635223388671875,
+      "learning_rate": 0.0003595991811227238,
+      "loss": 3.3473,
+      "step": 37300
+    },
+    {
+      "epoch": 4.020019373587343,
+      "grad_norm": 0.6792619228363037,
+      "learning_rate": 0.0003592759400926624,
+      "loss": 3.3541,
+      "step": 37350
+    },
+    {
+      "epoch": 4.0254009256269505,
+      "grad_norm": 0.6403281688690186,
+      "learning_rate": 0.00035895269906260094,
+      "loss": 3.3703,
+      "step": 37400
+    },
+    {
+      "epoch": 4.030782477666559,
+      "grad_norm": 0.6219056844711304,
+      "learning_rate": 0.0003586294580325396,
+      "loss": 3.349,
+      "step": 37450
+    },
+    {
+      "epoch": 4.036164029706168,
+      "grad_norm": 0.6096396446228027,
+      "learning_rate": 0.0003583062170024782,
+      "loss": 3.3572,
+      "step": 37500
+    },
+    {
+      "epoch": 4.041545581745775,
+      "grad_norm": 0.6848268508911133,
+      "learning_rate": 0.0003579829759724167,
+      "loss": 3.3605,
+      "step": 37550
+    },
+    {
+      "epoch": 4.046927133785384,
+      "grad_norm": 0.7016686201095581,
+      "learning_rate": 0.0003576597349423553,
+      "loss": 3.3493,
+      "step": 37600
+    },
+    {
+      "epoch": 4.0523086858249915,
+      "grad_norm": 0.685344934463501,
+      "learning_rate": 0.0003573364939122939,
+      "loss": 3.3565,
+      "step": 37650
+    },
+    {
+      "epoch": 4.0576902378646,
+      "grad_norm": 0.6827184557914734,
+      "learning_rate": 0.00035701325288223245,
+      "loss": 3.3671,
+      "step": 37700
+    },
+    {
+      "epoch": 4.063071789904209,
+      "grad_norm": 0.696050763130188,
+      "learning_rate": 0.0003566900118521711,
+      "loss": 3.3704,
+      "step": 37750
+    },
+    {
+      "epoch": 4.068453341943816,
+      "grad_norm": 0.7115525007247925,
+      "learning_rate": 0.0003563667708221097,
+      "loss": 3.358,
+      "step": 37800
+    },
+    {
+      "epoch": 4.073834893983425,
+      "grad_norm": 0.6731694340705872,
+      "learning_rate": 0.00035604352979204824,
+      "loss": 3.3879,
+      "step": 37850
+    },
+    {
+      "epoch": 4.079216446023033,
+      "grad_norm": 0.65992271900177,
+      "learning_rate": 0.00035572028876198683,
+      "loss": 3.3807,
+      "step": 37900
+    },
+    {
+      "epoch": 4.084597998062641,
+      "grad_norm": 0.6533928513526917,
+      "learning_rate": 0.00035539704773192537,
+      "loss": 3.3757,
+      "step": 37950
+    },
+    {
+      "epoch": 4.08997955010225,
+      "grad_norm": 0.6514235734939575,
+      "learning_rate": 0.00035507380670186397,
+      "loss": 3.3673,
+      "step": 38000
+    },
+    {
+      "epoch": 4.08997955010225,
+      "eval_accuracy": 0.37609534334366873,
+      "eval_loss": 3.4507036209106445,
+      "eval_runtime": 185.4683,
+      "eval_samples_per_second": 97.111,
+      "eval_steps_per_second": 6.071,
+      "step": 38000
+    },
+    {
+      "epoch": 4.095361102141858,
+      "grad_norm": 0.6657882928848267,
+      "learning_rate": 0.0003547505656718026,
+      "loss": 3.3763,
+      "step": 38050
+    },
+    {
+      "epoch": 4.100742654181466,
+      "grad_norm": 0.6397776007652283,
+      "learning_rate": 0.00035442732464174116,
+      "loss": 3.3646,
+      "step": 38100
+    },
+    {
+      "epoch": 4.106124206221074,
+      "grad_norm": 0.598796546459198,
+      "learning_rate": 0.00035410408361167975,
+      "loss": 3.3813,
+      "step": 38150
+    },
+    {
+      "epoch": 4.111505758260682,
+      "grad_norm": 0.7074088454246521,
+      "learning_rate": 0.0003537808425816183,
+      "loss": 3.3853,
+      "step": 38200
+    },
+    {
+      "epoch": 4.1168873103002905,
+      "grad_norm": 0.6053439378738403,
+      "learning_rate": 0.0003534576015515569,
+      "loss": 3.403,
+      "step": 38250
+    },
+    {
+      "epoch": 4.122268862339899,
+      "grad_norm": 0.6753089427947998,
+      "learning_rate": 0.00035313436052149553,
+      "loss": 3.3886,
+      "step": 38300
+    },
+    {
+      "epoch": 4.127650414379507,
+      "grad_norm": 0.6874846816062927,
+      "learning_rate": 0.0003528111194914341,
+      "loss": 3.3819,
+      "step": 38350
+    },
+    {
+      "epoch": 4.133031966419115,
+      "grad_norm": 0.6692848801612854,
+      "learning_rate": 0.00035248787846137267,
+      "loss": 3.3649,
+      "step": 38400
+    },
+    {
+      "epoch": 4.138413518458724,
+      "grad_norm": 0.6712144017219543,
+      "learning_rate": 0.00035216463743131126,
+      "loss": 3.3761,
+      "step": 38450
+    },
+    {
+      "epoch": 4.1437950704983315,
+      "grad_norm": 0.6473860144615173,
+      "learning_rate": 0.0003518413964012498,
+      "loss": 3.381,
+      "step": 38500
+    },
+    {
+      "epoch": 4.14917662253794,
+      "grad_norm": 0.641226589679718,
+      "learning_rate": 0.0003515181553711884,
+      "loss": 3.3771,
+      "step": 38550
+    },
+    {
+      "epoch": 4.154558174577549,
+      "grad_norm": 0.6594576835632324,
+      "learning_rate": 0.00035119491434112705,
+      "loss": 3.3875,
+      "step": 38600
+    },
+    {
+      "epoch": 4.159939726617156,
+      "grad_norm": 0.6367489695549011,
+      "learning_rate": 0.0003508716733110656,
+      "loss": 3.3879,
+      "step": 38650
+    },
+    {
+      "epoch": 4.165321278656765,
+      "grad_norm": 0.6333279609680176,
+      "learning_rate": 0.0003505484322810042,
+      "loss": 3.3843,
+      "step": 38700
+    },
+    {
+      "epoch": 4.1707028306963725,
+      "grad_norm": 0.6406641602516174,
+      "learning_rate": 0.0003502251912509427,
+      "loss": 3.3794,
+      "step": 38750
+    },
+    {
+      "epoch": 4.176084382735981,
+      "grad_norm": 0.6480084657669067,
+      "learning_rate": 0.0003499019502208813,
+      "loss": 3.3903,
+      "step": 38800
+    },
+    {
+      "epoch": 4.18146593477559,
+      "grad_norm": 0.6586027145385742,
+      "learning_rate": 0.0003495787091908199,
+      "loss": 3.3829,
+      "step": 38850
+    },
+    {
+      "epoch": 4.186847486815197,
+      "grad_norm": 0.6331707239151001,
+      "learning_rate": 0.0003492554681607585,
+      "loss": 3.3871,
+      "step": 38900
+    },
+    {
+      "epoch": 4.192229038854806,
+      "grad_norm": 0.6946048736572266,
+      "learning_rate": 0.0003489322271306971,
+      "loss": 3.3931,
+      "step": 38950
+    },
+    {
+      "epoch": 4.197610590894414,
+      "grad_norm": 0.6540223360061646,
+      "learning_rate": 0.0003486154509212369,
+      "loss": 3.3763,
+      "step": 39000
+    },
+    {
+      "epoch": 4.197610590894414,
+      "eval_accuracy": 0.3766727247895149,
+      "eval_loss": 3.4469048976898193,
+      "eval_runtime": 185.81,
+      "eval_samples_per_second": 96.932,
+      "eval_steps_per_second": 6.06,
+      "step": 39000
+    },
+    {
+      "epoch": 4.202992142934022,
+      "grad_norm": 0.6523948311805725,
+      "learning_rate": 0.0003482922098911755,
+      "loss": 3.3845,
+      "step": 39050
+    },
+    {
+      "epoch": 4.208373694973631,
+      "grad_norm": 0.6401236653327942,
+      "learning_rate": 0.0003479689688611141,
+      "loss": 3.3822,
+      "step": 39100
+    },
+    {
+      "epoch": 4.213755247013238,
+      "grad_norm": 0.6079734563827515,
+      "learning_rate": 0.00034764572783105264,
+      "loss": 3.3897,
+      "step": 39150
+    },
+    {
+      "epoch": 4.219136799052847,
+      "grad_norm": 0.6498140096664429,
+      "learning_rate": 0.00034732248680099123,
+      "loss": 3.3885,
+      "step": 39200
+    },
+    {
+      "epoch": 4.224518351092455,
+      "grad_norm": 0.6580759882926941,
+      "learning_rate": 0.0003469992457709299,
+      "loss": 3.3853,
+      "step": 39250
+    },
+    {
+      "epoch": 4.229899903132063,
+      "grad_norm": 0.636887788772583,
+      "learning_rate": 0.0003466760047408684,
+      "loss": 3.3872,
+      "step": 39300
+    },
+    {
+      "epoch": 4.2352814551716715,
+      "grad_norm": 0.6532869338989258,
+      "learning_rate": 0.000346352763710807,
+      "loss": 3.3741,
+      "step": 39350
+    },
+    {
+      "epoch": 4.24066300721128,
+      "grad_norm": 0.6606283187866211,
+      "learning_rate": 0.00034602952268074556,
+      "loss": 3.3828,
+      "step": 39400
+    },
+    {
+      "epoch": 4.246044559250888,
+      "grad_norm": 0.6259838342666626,
+      "learning_rate": 0.00034570628165068415,
+      "loss": 3.3853,
+      "step": 39450
+    },
+    {
+      "epoch": 4.251426111290496,
+      "grad_norm": 0.6093770265579224,
+      "learning_rate": 0.0003453830406206227,
+      "loss": 3.3766,
+      "step": 39500
+    },
+    {
+      "epoch": 4.256807663330104,
+      "grad_norm": 0.6336487531661987,
+      "learning_rate": 0.00034505979959056134,
+      "loss": 3.4046,
+      "step": 39550
+    },
+    {
+      "epoch": 4.2621892153697125,
+      "grad_norm": 0.6288484334945679,
+      "learning_rate": 0.00034473655856049994,
+      "loss": 3.3898,
+      "step": 39600
+    },
+    {
+      "epoch": 4.267570767409321,
+      "grad_norm": 0.6407979726791382,
+      "learning_rate": 0.0003444133175304385,
+      "loss": 3.3789,
+      "step": 39650
+    },
+    {
+      "epoch": 4.272952319448929,
+      "grad_norm": 0.6025500297546387,
+      "learning_rate": 0.00034409007650037707,
+      "loss": 3.377,
+      "step": 39700
+    },
+    {
+      "epoch": 4.278333871488537,
+      "grad_norm": 0.6767619252204895,
+      "learning_rate": 0.00034376683547031567,
+      "loss": 3.3676,
+      "step": 39750
+    },
+    {
+      "epoch": 4.283715423528146,
+      "grad_norm": 0.6752068996429443,
+      "learning_rate": 0.00034345005926085547,
+      "loss": 3.3805,
+      "step": 39800
+    },
+    {
+      "epoch": 4.2890969755677535,
+      "grad_norm": 0.6374906301498413,
+      "learning_rate": 0.00034312681823079407,
+      "loss": 3.3838,
+      "step": 39850
+    },
+    {
+      "epoch": 4.294478527607362,
+      "grad_norm": 0.6149640679359436,
+      "learning_rate": 0.00034280357720073266,
+      "loss": 3.3925,
+      "step": 39900
+    },
+    {
+      "epoch": 4.299860079646971,
+      "grad_norm": 0.762861967086792,
+      "learning_rate": 0.00034248033617067126,
+      "loss": 3.3599,
+      "step": 39950
+    },
+    {
+      "epoch": 4.305241631686578,
+      "grad_norm": 0.6646538376808167,
+      "learning_rate": 0.00034215709514060985,
+      "loss": 3.4159,
+      "step": 40000
+    },
+    {
+      "epoch": 4.305241631686578,
+      "eval_accuracy": 0.3771906731057265,
+      "eval_loss": 3.441789388656616,
+      "eval_runtime": 185.3693,
+      "eval_samples_per_second": 97.163,
+      "eval_steps_per_second": 6.074,
+      "step": 40000
+    },
+    {
+      "epoch": 4.310623183726187,
+      "grad_norm": 0.6618566513061523,
+      "learning_rate": 0.0003418338541105484,
+      "loss": 3.3963,
+      "step": 40050
+    },
+    {
+      "epoch": 4.3160047357657945,
+      "grad_norm": 0.6328936815261841,
+      "learning_rate": 0.000341510613080487,
+      "loss": 3.3607,
+      "step": 40100
+    },
+    {
+      "epoch": 4.321386287805403,
+      "grad_norm": 0.6682262420654297,
+      "learning_rate": 0.0003411873720504255,
+      "loss": 3.3899,
+      "step": 40150
+    },
+    {
+      "epoch": 4.326767839845012,
+      "grad_norm": 0.6178170442581177,
+      "learning_rate": 0.0003408641310203642,
+      "loss": 3.3719,
+      "step": 40200
+    },
+    {
+      "epoch": 4.332149391884619,
+      "grad_norm": 0.7068307399749756,
+      "learning_rate": 0.00034054088999030277,
+      "loss": 3.3932,
+      "step": 40250
+    },
+    {
+      "epoch": 4.337530943924228,
+      "grad_norm": 0.6037469506263733,
+      "learning_rate": 0.0003402176489602413,
+      "loss": 3.3895,
+      "step": 40300
+    },
+    {
+      "epoch": 4.342912495963836,
+      "grad_norm": 0.7188079357147217,
+      "learning_rate": 0.0003398944079301799,
+      "loss": 3.407,
+      "step": 40350
+    },
+    {
+      "epoch": 4.348294048003444,
+      "grad_norm": 0.6433869004249573,
+      "learning_rate": 0.0003395711669001185,
+      "loss": 3.3842,
+      "step": 40400
+    },
+    {
+      "epoch": 4.3536756000430525,
+      "grad_norm": 0.6392279863357544,
+      "learning_rate": 0.00033924792587005704,
+      "loss": 3.3886,
+      "step": 40450
+    },
+    {
+      "epoch": 4.359057152082661,
+      "grad_norm": 0.6604293584823608,
+      "learning_rate": 0.0003389246848399957,
+      "loss": 3.3957,
+      "step": 40500
+    },
+    {
+      "epoch": 4.364438704122269,
+      "grad_norm": 0.609643816947937,
+      "learning_rate": 0.0003386014438099343,
+      "loss": 3.4041,
+      "step": 40550
+    },
+    {
+      "epoch": 4.369820256161877,
+      "grad_norm": 0.6582847237586975,
+      "learning_rate": 0.0003382782027798728,
+      "loss": 3.3732,
+      "step": 40600
+    },
+    {
+      "epoch": 4.375201808201485,
+      "grad_norm": 0.6692153215408325,
+      "learning_rate": 0.0003379549617498114,
+      "loss": 3.3829,
+      "step": 40650
+    },
+    {
+      "epoch": 4.3805833602410935,
+      "grad_norm": 0.6821966171264648,
+      "learning_rate": 0.00033763172071974996,
+      "loss": 3.3999,
+      "step": 40700
+    },
+    {
+      "epoch": 4.385964912280702,
+      "grad_norm": 0.6171023845672607,
+      "learning_rate": 0.00033730847968968855,
+      "loss": 3.3994,
+      "step": 40750
+    },
+    {
+      "epoch": 4.39134646432031,
+      "grad_norm": 0.6396343111991882,
+      "learning_rate": 0.0003369852386596272,
+      "loss": 3.3971,
+      "step": 40800
+    },
+    {
+      "epoch": 4.396728016359918,
+      "grad_norm": 0.617160975933075,
+      "learning_rate": 0.00033666199762956574,
+      "loss": 3.4029,
+      "step": 40850
+    },
+    {
+      "epoch": 4.402109568399527,
+      "grad_norm": 0.6241002082824707,
+      "learning_rate": 0.00033633875659950434,
+      "loss": 3.3956,
+      "step": 40900
+    },
+    {
+      "epoch": 4.4074911204391345,
+      "grad_norm": 0.6754607558250427,
+      "learning_rate": 0.0003360155155694429,
+      "loss": 3.3941,
+      "step": 40950
+    },
+    {
+      "epoch": 4.412872672478743,
+      "grad_norm": 0.662516713142395,
+      "learning_rate": 0.0003356922745393815,
+      "loss": 3.3961,
+      "step": 41000
+    },
+    {
+      "epoch": 4.412872672478743,
+      "eval_accuracy": 0.3775866042289225,
+      "eval_loss": 3.434943437576294,
+      "eval_runtime": 185.6144,
+      "eval_samples_per_second": 97.035,
+      "eval_steps_per_second": 6.066,
+      "step": 41000
+    },
+    {
+      "epoch": 4.418254224518351,
+      "grad_norm": 0.7693639397621155,
+      "learning_rate": 0.0003353690335093201,
+      "loss": 3.3931,
+      "step": 41050
+    },
+    {
+      "epoch": 4.423635776557959,
+      "grad_norm": 0.6286409497261047,
+      "learning_rate": 0.0003350457924792587,
+      "loss": 3.3912,
+      "step": 41100
+    },
+    {
+      "epoch": 4.429017328597568,
+      "grad_norm": 0.6330351233482361,
+      "learning_rate": 0.00033472255144919726,
+      "loss": 3.4037,
+      "step": 41150
+    },
+    {
+      "epoch": 4.4343988806371755,
+      "grad_norm": 0.6741186380386353,
+      "learning_rate": 0.00033439931041913585,
+      "loss": 3.3872,
+      "step": 41200
+    },
+    {
+      "epoch": 4.439780432676784,
+      "grad_norm": 0.7080098986625671,
+      "learning_rate": 0.0003340760693890744,
+      "loss": 3.3859,
+      "step": 41250
+    },
+    {
+      "epoch": 4.445161984716393,
+      "grad_norm": 0.6425966024398804,
+      "learning_rate": 0.000333752828359013,
+      "loss": 3.3843,
+      "step": 41300
+    },
+    {
+      "epoch": 4.450543536756,
+      "grad_norm": 0.6547057032585144,
+      "learning_rate": 0.00033342958732895164,
+      "loss": 3.3959,
+      "step": 41350
+    },
+    {
+      "epoch": 4.455925088795609,
+      "grad_norm": 0.7627517580986023,
+      "learning_rate": 0.0003331063462988902,
+      "loss": 3.3875,
+      "step": 41400
+    },
+    {
+      "epoch": 4.461306640835216,
+      "grad_norm": 0.6568228602409363,
+      "learning_rate": 0.00033278310526882877,
+      "loss": 3.3955,
+      "step": 41450
+    },
+    {
+      "epoch": 4.466688192874825,
+      "grad_norm": 0.6827290654182434,
+      "learning_rate": 0.0003324598642387673,
+      "loss": 3.3893,
+      "step": 41500
+    },
+    {
+      "epoch": 4.4720697449144335,
+      "grad_norm": 0.6292122006416321,
+      "learning_rate": 0.0003321366232087059,
+      "loss": 3.3808,
+      "step": 41550
+    },
+    {
+      "epoch": 4.477451296954041,
+      "grad_norm": 0.6430679559707642,
+      "learning_rate": 0.0003318133821786445,
+      "loss": 3.4038,
+      "step": 41600
+    },
+    {
+      "epoch": 4.48283284899365,
+      "grad_norm": 0.6482014060020447,
+      "learning_rate": 0.0003314901411485831,
+      "loss": 3.3828,
+      "step": 41650
+    },
+    {
+      "epoch": 4.488214401033258,
+      "grad_norm": 0.621187150478363,
+      "learning_rate": 0.0003311669001185217,
+      "loss": 3.4006,
+      "step": 41700
+    },
+    {
+      "epoch": 4.493595953072866,
+      "grad_norm": 0.6572025418281555,
+      "learning_rate": 0.0003308436590884603,
+      "loss": 3.3966,
+      "step": 41750
+    },
+    {
+      "epoch": 4.4989775051124745,
+      "grad_norm": 0.7006889581680298,
+      "learning_rate": 0.0003305204180583988,
+      "loss": 3.3933,
+      "step": 41800
+    },
+    {
+      "epoch": 4.504359057152083,
+      "grad_norm": 0.6104776263237,
+      "learning_rate": 0.0003301971770283374,
+      "loss": 3.3785,
+      "step": 41850
+    },
+    {
+      "epoch": 4.509740609191691,
+      "grad_norm": 0.6377589106559753,
+      "learning_rate": 0.00032987393599827607,
+      "loss": 3.3773,
+      "step": 41900
+    },
+    {
+      "epoch": 4.515122161231299,
+      "grad_norm": 0.7028476595878601,
+      "learning_rate": 0.0003295506949682146,
+      "loss": 3.3933,
+      "step": 41950
+    },
+    {
+      "epoch": 4.520503713270907,
+      "grad_norm": 0.6473916172981262,
+      "learning_rate": 0.0003292274539381532,
+      "loss": 3.3853,
+      "step": 42000
+    },
+    {
+      "epoch": 4.520503713270907,
+      "eval_accuracy": 0.3784860328342508,
+      "eval_loss": 3.429154872894287,
+      "eval_runtime": 185.4968,
+      "eval_samples_per_second": 97.096,
+      "eval_steps_per_second": 6.07,
+      "step": 42000
+    },
+    {
+      "epoch": 4.5258852653105155,
+      "grad_norm": 0.7115603089332581,
+      "learning_rate": 0.000328910677728693,
+      "loss": 3.3936,
+      "step": 42050
+    },
+    {
+      "epoch": 4.531266817350124,
+      "grad_norm": 0.7045222520828247,
+      "learning_rate": 0.0003285874366986316,
+      "loss": 3.4024,
+      "step": 42100
+    },
+    {
+      "epoch": 4.536648369389732,
+      "grad_norm": 0.7149264812469482,
+      "learning_rate": 0.00032826419566857015,
+      "loss": 3.3964,
+      "step": 42150
+    },
+    {
+      "epoch": 4.54202992142934,
+      "grad_norm": 0.6592740416526794,
+      "learning_rate": 0.00032794095463850874,
+      "loss": 3.391,
+      "step": 42200
+    },
+    {
+      "epoch": 4.547411473468949,
+      "grad_norm": 0.642037034034729,
+      "learning_rate": 0.0003276177136084473,
+      "loss": 3.3929,
+      "step": 42250
+    },
+    {
+      "epoch": 4.5527930255085565,
+      "grad_norm": 0.6554846167564392,
+      "learning_rate": 0.00032729447257838593,
+      "loss": 3.3642,
+      "step": 42300
+    },
+    {
+      "epoch": 4.558174577548165,
+      "grad_norm": 0.6893788576126099,
+      "learning_rate": 0.0003269712315483245,
+      "loss": 3.3735,
+      "step": 42350
+    },
+    {
+      "epoch": 4.563556129587774,
+      "grad_norm": 0.6151097416877747,
+      "learning_rate": 0.00032664799051826306,
+      "loss": 3.398,
+      "step": 42400
+    },
+    {
+      "epoch": 4.568937681627381,
+      "grad_norm": 0.6453260183334351,
+      "learning_rate": 0.00032632474948820166,
+      "loss": 3.3841,
+      "step": 42450
+    },
+    {
+      "epoch": 4.57431923366699,
+      "grad_norm": 0.6441733241081238,
+      "learning_rate": 0.00032600150845814025,
+      "loss": 3.3916,
+      "step": 42500
+    },
+    {
+      "epoch": 4.579700785706597,
+      "grad_norm": 0.6358645558357239,
+      "learning_rate": 0.0003256782674280788,
+      "loss": 3.404,
+      "step": 42550
+    },
+    {
+      "epoch": 4.585082337746206,
+      "grad_norm": 0.6901255249977112,
+      "learning_rate": 0.00032535502639801744,
+      "loss": 3.4148,
+      "step": 42600
+    },
+    {
+      "epoch": 4.5904638897858145,
+      "grad_norm": 0.6465464234352112,
+      "learning_rate": 0.00032503178536795604,
+      "loss": 3.383,
+      "step": 42650
+    },
+    {
+      "epoch": 4.595845441825422,
+      "grad_norm": 0.6677713990211487,
+      "learning_rate": 0.0003247085443378946,
+      "loss": 3.3959,
+      "step": 42700
+    },
+    {
+      "epoch": 4.601226993865031,
+      "grad_norm": 0.6432662606239319,
+      "learning_rate": 0.0003243853033078332,
+      "loss": 3.3735,
+      "step": 42750
+    },
+    {
+      "epoch": 4.606608545904638,
+      "grad_norm": 0.6490405201911926,
+      "learning_rate": 0.0003240620622777717,
+      "loss": 3.4025,
+      "step": 42800
+    },
+    {
+      "epoch": 4.611990097944247,
+      "grad_norm": 0.662980318069458,
+      "learning_rate": 0.00032373882124771036,
+      "loss": 3.3855,
+      "step": 42850
+    },
+    {
+      "epoch": 4.6173716499838555,
+      "grad_norm": 0.6549072265625,
+      "learning_rate": 0.00032341558021764896,
+      "loss": 3.3841,
+      "step": 42900
+    },
+    {
+      "epoch": 4.622753202023463,
+      "grad_norm": 0.6647734642028809,
+      "learning_rate": 0.0003230923391875875,
+      "loss": 3.409,
+      "step": 42950
+    },
+    {
+      "epoch": 4.628134754063072,
+      "grad_norm": 0.6487623453140259,
+      "learning_rate": 0.0003227690981575261,
+      "loss": 3.3831,
+      "step": 43000
+    },
+    {
+      "epoch": 4.628134754063072,
+      "eval_accuracy": 0.37883632974456466,
+      "eval_loss": 3.42627215385437,
+      "eval_runtime": 185.8241,
+      "eval_samples_per_second": 96.925,
+      "eval_steps_per_second": 6.059,
+      "step": 43000
+    },
+    {
+      "epoch": 4.63351630610268,
+      "grad_norm": 0.6644917130470276,
+      "learning_rate": 0.0003224458571274647,
+      "loss": 3.393,
+      "step": 43050
+    },
+    {
+      "epoch": 4.638897858142288,
+      "grad_norm": 0.6768592596054077,
+      "learning_rate": 0.00032212261609740323,
+      "loss": 3.4032,
+      "step": 43100
+    },
+    {
+      "epoch": 4.6442794101818965,
+      "grad_norm": 0.6610867977142334,
+      "learning_rate": 0.0003217993750673419,
+      "loss": 3.3878,
+      "step": 43150
+    },
+    {
+      "epoch": 4.649660962221505,
+      "grad_norm": 0.6174038648605347,
+      "learning_rate": 0.00032147613403728047,
+      "loss": 3.3748,
+      "step": 43200
+    },
+    {
+      "epoch": 4.655042514261113,
+      "grad_norm": 0.691074550151825,
+      "learning_rate": 0.000321152893007219,
+      "loss": 3.4017,
+      "step": 43250
+    },
+    {
+      "epoch": 4.660424066300721,
+      "grad_norm": 0.676375150680542,
+      "learning_rate": 0.0003208296519771576,
+      "loss": 3.404,
+      "step": 43300
+    },
+    {
+      "epoch": 4.665805618340329,
+      "grad_norm": 0.6594715714454651,
+      "learning_rate": 0.00032050641094709615,
+      "loss": 3.392,
+      "step": 43350
+    },
+    {
+      "epoch": 4.6711871703799375,
+      "grad_norm": 0.6825419068336487,
+      "learning_rate": 0.00032018316991703474,
+      "loss": 3.4002,
+      "step": 43400
+    },
+    {
+      "epoch": 4.676568722419546,
+      "grad_norm": 0.6380799412727356,
+      "learning_rate": 0.0003198599288869734,
+      "loss": 3.3901,
+      "step": 43450
+    },
+    {
+      "epoch": 4.681950274459154,
+      "grad_norm": 0.6442010998725891,
+      "learning_rate": 0.00031953668785691193,
+      "loss": 3.3935,
+      "step": 43500
+    },
+    {
+      "epoch": 4.687331826498762,
+      "grad_norm": 0.6987720131874084,
+      "learning_rate": 0.0003192134468268505,
+      "loss": 3.3945,
+      "step": 43550
+    },
+    {
+      "epoch": 4.692713378538371,
+      "grad_norm": 0.6554795503616333,
+      "learning_rate": 0.0003188902057967891,
+      "loss": 3.386,
+      "step": 43600
+    },
+    {
+      "epoch": 4.6980949305779784,
+      "grad_norm": 0.6956591606140137,
+      "learning_rate": 0.00031856696476672766,
+      "loss": 3.3868,
+      "step": 43650
+    },
+    {
+      "epoch": 4.703476482617587,
+      "grad_norm": 0.6553178429603577,
+      "learning_rate": 0.0003182437237366663,
+      "loss": 3.3892,
+      "step": 43700
+    },
+    {
+      "epoch": 4.7088580346571955,
+      "grad_norm": 0.6245745420455933,
+      "learning_rate": 0.0003179204827066049,
+      "loss": 3.3923,
+      "step": 43750
+    },
+    {
+      "epoch": 4.714239586696803,
+      "grad_norm": 0.6830365061759949,
+      "learning_rate": 0.00031759724167654344,
+      "loss": 3.3573,
+      "step": 43800
+    },
+    {
+      "epoch": 4.719621138736412,
+      "grad_norm": 0.6748046875,
+      "learning_rate": 0.00031727400064648204,
+      "loss": 3.3908,
+      "step": 43850
+    },
+    {
+      "epoch": 4.725002690776019,
+      "grad_norm": 0.6480768918991089,
+      "learning_rate": 0.0003169507596164206,
+      "loss": 3.3871,
+      "step": 43900
+    },
+    {
+      "epoch": 4.730384242815628,
+      "grad_norm": 0.6577687859535217,
+      "learning_rate": 0.0003166275185863592,
+      "loss": 3.392,
+      "step": 43950
+    },
+    {
+      "epoch": 4.7357657948552365,
+      "grad_norm": 0.6757226586341858,
+      "learning_rate": 0.0003163042775562978,
+      "loss": 3.4007,
+      "step": 44000
+    },
+    {
+      "epoch": 4.7357657948552365,
+      "eval_accuracy": 0.37909340247713375,
+      "eval_loss": 3.422922372817993,
+      "eval_runtime": 185.4325,
+      "eval_samples_per_second": 97.13,
+      "eval_steps_per_second": 6.072,
+      "step": 44000
+    },
+    {
+      "epoch": 4.741147346894844,
+      "grad_norm": 0.7063640356063843,
+      "learning_rate": 0.00031598103652623636,
+      "loss": 3.3987,
+      "step": 44050
+    },
+    {
+      "epoch": 4.746528898934453,
+      "grad_norm": 0.6377637982368469,
+      "learning_rate": 0.00031565779549617496,
+      "loss": 3.3771,
+      "step": 44100
+    },
+    {
+      "epoch": 4.751910450974061,
+      "grad_norm": 0.6924796104431152,
+      "learning_rate": 0.00031533455446611355,
+      "loss": 3.3958,
+      "step": 44150
+    },
+    {
+      "epoch": 4.757292003013669,
+      "grad_norm": 0.6684271097183228,
+      "learning_rate": 0.0003150113134360521,
+      "loss": 3.3933,
+      "step": 44200
+    },
+    {
+      "epoch": 4.7626735550532775,
+      "grad_norm": 0.6748309135437012,
+      "learning_rate": 0.0003146880724059907,
+      "loss": 3.3961,
+      "step": 44250
+    },
+    {
+      "epoch": 4.768055107092886,
+      "grad_norm": 0.6288613677024841,
+      "learning_rate": 0.00031436483137592934,
+      "loss": 3.3982,
+      "step": 44300
+    },
+    {
+      "epoch": 4.773436659132494,
+      "grad_norm": 0.6949737071990967,
+      "learning_rate": 0.0003140415903458679,
+      "loss": 3.3737,
+      "step": 44350
+    },
+    {
+      "epoch": 4.778818211172102,
+      "grad_norm": 0.6416726112365723,
+      "learning_rate": 0.00031371834931580647,
+      "loss": 3.3954,
+      "step": 44400
+    },
+    {
+      "epoch": 4.78419976321171,
+      "grad_norm": 0.6573190689086914,
+      "learning_rate": 0.000313395108285745,
+      "loss": 3.3948,
+      "step": 44450
+    },
+    {
+      "epoch": 4.7895813152513185,
+      "grad_norm": 0.6981139779090881,
+      "learning_rate": 0.0003130718672556836,
+      "loss": 3.3971,
+      "step": 44500
+    },
+    {
+      "epoch": 4.794962867290927,
+      "grad_norm": 0.6780474185943604,
+      "learning_rate": 0.0003127486262256222,
+      "loss": 3.3948,
+      "step": 44550
+    },
+    {
+      "epoch": 4.800344419330535,
+      "grad_norm": 0.7529133558273315,
+      "learning_rate": 0.0003124253851955608,
+      "loss": 3.4077,
+      "step": 44600
+    },
+    {
+      "epoch": 4.805725971370143,
+      "grad_norm": 0.6594308018684387,
+      "learning_rate": 0.0003121021441654994,
+      "loss": 3.393,
+      "step": 44650
+    },
+    {
+      "epoch": 4.811107523409751,
+      "grad_norm": 0.6422306299209595,
+      "learning_rate": 0.0003117853679560392,
+      "loss": 3.3853,
+      "step": 44700
+    },
+    {
+      "epoch": 4.8164890754493594,
+      "grad_norm": 0.6755315661430359,
+      "learning_rate": 0.0003114621269259778,
+      "loss": 3.3917,
+      "step": 44750
+    },
+    {
+      "epoch": 4.821870627488968,
+      "grad_norm": 0.6830017566680908,
+      "learning_rate": 0.00031113888589591633,
+      "loss": 3.3829,
+      "step": 44800
+    },
+    {
+      "epoch": 4.827252179528576,
+      "grad_norm": 0.7093849182128906,
+      "learning_rate": 0.00031081564486585493,
+      "loss": 3.3923,
+      "step": 44850
+    },
+    {
+      "epoch": 4.832633731568184,
+      "grad_norm": 0.692285418510437,
+      "learning_rate": 0.0003104924038357935,
+      "loss": 3.3846,
+      "step": 44900
+    },
+    {
+      "epoch": 4.838015283607793,
+      "grad_norm": 0.687765896320343,
+      "learning_rate": 0.0003101691628057321,
+      "loss": 3.418,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8433968356474,
+      "grad_norm": 0.6697891354560852,
+      "learning_rate": 0.0003098459217756707,
+      "loss": 3.3982,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8433968356474,
+      "eval_accuracy": 0.3797329333748098,
+      "eval_loss": 3.416762351989746,
+      "eval_runtime": 185.8849,
+      "eval_samples_per_second": 96.893,
+      "eval_steps_per_second": 6.058,
+      "step": 45000
+    },
+    {
+      "epoch": 4.848778387687009,
+      "grad_norm": 0.6329440474510193,
+      "learning_rate": 0.0003095226807456093,
+      "loss": 3.3803,
+      "step": 45050
+    },
+    {
+      "epoch": 4.8541599397266175,
+      "grad_norm": 0.6872698664665222,
+      "learning_rate": 0.00030919943971554785,
+      "loss": 3.4032,
+      "step": 45100
+    },
+    {
+      "epoch": 4.859541491766225,
+      "grad_norm": 0.697952389717102,
+      "learning_rate": 0.00030887619868548644,
+      "loss": 3.3849,
+      "step": 45150
+    },
+    {
+      "epoch": 4.864923043805834,
+      "grad_norm": 0.6626936197280884,
+      "learning_rate": 0.00030855942247602625,
+      "loss": 3.396,
+      "step": 45200
+    },
+    {
+      "epoch": 4.870304595845441,
+      "grad_norm": 0.7497137784957886,
+      "learning_rate": 0.00030823618144596484,
+      "loss": 3.3844,
+      "step": 45250
+    },
+    {
+      "epoch": 4.87568614788505,
+      "grad_norm": 0.7056050896644592,
+      "learning_rate": 0.0003079129404159035,
+      "loss": 3.3903,
+      "step": 45300
+    },
+    {
+      "epoch": 4.8810676999246585,
+      "grad_norm": 0.6733185648918152,
+      "learning_rate": 0.00030758969938584203,
+      "loss": 3.3851,
+      "step": 45350
+    },
+    {
+      "epoch": 4.886449251964266,
+      "grad_norm": 0.7165461778640747,
+      "learning_rate": 0.0003072664583557806,
+      "loss": 3.3738,
+      "step": 45400
+    },
+    {
+      "epoch": 4.891830804003875,
+      "grad_norm": 0.672849178314209,
+      "learning_rate": 0.00030694321732571917,
+      "loss": 3.3913,
+      "step": 45450
+    },
+    {
+      "epoch": 4.897212356043483,
+      "grad_norm": 0.6391351222991943,
+      "learning_rate": 0.00030661997629565776,
+      "loss": 3.3746,
+      "step": 45500
+    },
+    {
+      "epoch": 4.902593908083091,
+      "grad_norm": 0.6835765838623047,
+      "learning_rate": 0.0003062967352655963,
+      "loss": 3.4118,
+      "step": 45550
+    },
+    {
+      "epoch": 4.9079754601226995,
+      "grad_norm": 0.7272644639015198,
+      "learning_rate": 0.00030597349423553495,
+      "loss": 3.3987,
+      "step": 45600
+    },
+    {
+      "epoch": 4.913357012162308,
+      "grad_norm": 0.6670034527778625,
+      "learning_rate": 0.00030565025320547355,
+      "loss": 3.3732,
+      "step": 45650
+    },
+    {
+      "epoch": 4.918738564201916,
+      "grad_norm": 0.6669803857803345,
+      "learning_rate": 0.0003053270121754121,
+      "loss": 3.3941,
+      "step": 45700
+    },
+    {
+      "epoch": 4.924120116241524,
+      "grad_norm": 0.6653757095336914,
+      "learning_rate": 0.0003050037711453507,
+      "loss": 3.3928,
+      "step": 45750
+    },
+    {
+      "epoch": 4.929501668281132,
+      "grad_norm": 0.6554093360900879,
+      "learning_rate": 0.0003046805301152893,
+      "loss": 3.4065,
+      "step": 45800
+    },
+    {
+      "epoch": 4.9348832203207404,
+      "grad_norm": 0.6358471512794495,
+      "learning_rate": 0.0003043572890852278,
+      "loss": 3.3935,
+      "step": 45850
+    },
+    {
+      "epoch": 4.940264772360349,
+      "grad_norm": 0.6480586528778076,
+      "learning_rate": 0.00030403404805516646,
+      "loss": 3.3703,
+      "step": 45900
+    },
+    {
+      "epoch": 4.945646324399957,
+      "grad_norm": 0.6871649026870728,
+      "learning_rate": 0.00030371080702510506,
+      "loss": 3.4116,
+      "step": 45950
+    },
+    {
+      "epoch": 4.951027876439565,
+      "grad_norm": 0.6699846982955933,
+      "learning_rate": 0.0003033875659950436,
+      "loss": 3.4117,
+      "step": 46000
+    },
+    {
+      "epoch": 4.951027876439565,
+      "eval_accuracy": 0.380043789286847,
+      "eval_loss": 3.411285877227783,
+      "eval_runtime": 185.4407,
+      "eval_samples_per_second": 97.125,
+      "eval_steps_per_second": 6.072,
+      "step": 46000
+    },
+    {
+      "epoch": 4.956409428479174,
+      "grad_norm": 0.6856145262718201,
+      "learning_rate": 0.0003030643249649822,
+      "loss": 3.3878,
+      "step": 46050
+    },
+    {
+      "epoch": 4.961790980518781,
+      "grad_norm": 0.6917241215705872,
+      "learning_rate": 0.00030274108393492073,
+      "loss": 3.3869,
+      "step": 46100
+    },
+    {
+      "epoch": 4.96717253255839,
+      "grad_norm": 0.6799637079238892,
+      "learning_rate": 0.00030241784290485933,
+      "loss": 3.3783,
+      "step": 46150
+    },
+    {
+      "epoch": 4.9725540845979985,
+      "grad_norm": 0.7519565224647522,
+      "learning_rate": 0.000302094601874798,
+      "loss": 3.3942,
+      "step": 46200
+    },
+    {
+      "epoch": 4.977935636637606,
+      "grad_norm": 0.6828088760375977,
+      "learning_rate": 0.0003017713608447365,
+      "loss": 3.3822,
+      "step": 46250
+    },
+    {
+      "epoch": 4.983317188677215,
+      "grad_norm": 0.6727144718170166,
+      "learning_rate": 0.0003014481198146751,
+      "loss": 3.4129,
+      "step": 46300
+    },
+    {
+      "epoch": 4.988698740716822,
+      "grad_norm": 0.652026355266571,
+      "learning_rate": 0.0003011248787846137,
+      "loss": 3.3789,
+      "step": 46350
+    },
+    {
+      "epoch": 4.994080292756431,
+      "grad_norm": 0.6903791427612305,
+      "learning_rate": 0.00030080163775455225,
+      "loss": 3.4116,
+      "step": 46400
+    },
+    {
+      "epoch": 4.9994618447960395,
+      "grad_norm": 0.6479547619819641,
+      "learning_rate": 0.0003004783967244909,
+      "loss": 3.3898,
+      "step": 46450
+    },
+    {
+      "epoch": 5.004843396835647,
+      "grad_norm": 0.6606223583221436,
+      "learning_rate": 0.0003001551556944295,
+      "loss": 3.2857,
+      "step": 46500
+    },
+    {
+      "epoch": 5.010224948875256,
+      "grad_norm": 0.7087961435317993,
+      "learning_rate": 0.00029983191466436803,
+      "loss": 3.3002,
+      "step": 46550
+    },
+    {
+      "epoch": 5.015606500914864,
+      "grad_norm": 0.6644490361213684,
+      "learning_rate": 0.0002995086736343066,
+      "loss": 3.309,
+      "step": 46600
+    },
+    {
+      "epoch": 5.020988052954472,
+      "grad_norm": 0.6827025413513184,
+      "learning_rate": 0.00029918543260424517,
+      "loss": 3.2999,
+      "step": 46650
+    },
+    {
+      "epoch": 5.0263696049940805,
+      "grad_norm": 0.7336397767066956,
+      "learning_rate": 0.0002988621915741838,
+      "loss": 3.3061,
+      "step": 46700
+    },
+    {
+      "epoch": 5.031751157033688,
+      "grad_norm": 0.7139917016029358,
+      "learning_rate": 0.00029853895054412236,
+      "loss": 3.3064,
+      "step": 46750
+    },
+    {
+      "epoch": 5.037132709073297,
+      "grad_norm": 0.7830448746681213,
+      "learning_rate": 0.00029821570951406095,
+      "loss": 3.2965,
+      "step": 46800
+    },
+    {
+      "epoch": 5.042514261112905,
+      "grad_norm": 0.6794210076332092,
+      "learning_rate": 0.00029789246848399955,
+      "loss": 3.29,
+      "step": 46850
+    },
+    {
+      "epoch": 5.047895813152513,
+      "grad_norm": 0.6791298389434814,
+      "learning_rate": 0.00029756922745393814,
+      "loss": 3.3002,
+      "step": 46900
+    },
+    {
+      "epoch": 5.0532773651921215,
+      "grad_norm": 0.684412956237793,
+      "learning_rate": 0.00029724598642387674,
+      "loss": 3.3228,
+      "step": 46950
+    },
+    {
+      "epoch": 5.05865891723173,
+      "grad_norm": 0.635731041431427,
+      "learning_rate": 0.0002969227453938153,
+      "loss": 3.2918,
+      "step": 47000
+    },
+    {
+      "epoch": 5.05865891723173,
+      "eval_accuracy": 0.38074449176036257,
+      "eval_loss": 3.415309429168701,
+      "eval_runtime": 185.4384,
+      "eval_samples_per_second": 97.127,
+      "eval_steps_per_second": 6.072,
+      "step": 47000
+    },
+    {
+      "epoch": 5.064040469271338,
+      "grad_norm": 0.7232507467269897,
+      "learning_rate": 0.00029659950436375387,
+      "loss": 3.3026,
+      "step": 47050
+    },
+    {
+      "epoch": 5.069422021310946,
+      "grad_norm": 0.6754257082939148,
+      "learning_rate": 0.00029627626333369246,
+      "loss": 3.3133,
+      "step": 47100
+    },
+    {
+      "epoch": 5.074803573350554,
+      "grad_norm": 0.6911904215812683,
+      "learning_rate": 0.00029595302230363106,
+      "loss": 3.3112,
+      "step": 47150
+    },
+    {
+      "epoch": 5.080185125390162,
+      "grad_norm": 0.6793113350868225,
+      "learning_rate": 0.0002956297812735696,
+      "loss": 3.3096,
+      "step": 47200
+    },
+    {
+      "epoch": 5.085566677429771,
+      "grad_norm": 0.7221434116363525,
+      "learning_rate": 0.00029530654024350825,
+      "loss": 3.3074,
+      "step": 47250
+    },
+    {
+      "epoch": 5.090948229469379,
+      "grad_norm": 0.6742022633552551,
+      "learning_rate": 0.0002949832992134468,
+      "loss": 3.315,
+      "step": 47300
+    },
+    {
+      "epoch": 5.096329781508987,
+      "grad_norm": 0.7125201225280762,
+      "learning_rate": 0.0002946600581833854,
+      "loss": 3.3104,
+      "step": 47350
+    },
+    {
+      "epoch": 5.101711333548596,
+      "grad_norm": 0.6573224067687988,
+      "learning_rate": 0.000294336817153324,
+      "loss": 3.3168,
+      "step": 47400
+    },
+    {
+      "epoch": 5.107092885588203,
+      "grad_norm": 0.7398542165756226,
+      "learning_rate": 0.0002940135761232626,
+      "loss": 3.3053,
+      "step": 47450
+    },
+    {
+      "epoch": 5.112474437627812,
+      "grad_norm": 0.7006282806396484,
+      "learning_rate": 0.0002936903350932011,
+      "loss": 3.3085,
+      "step": 47500
+    },
+    {
+      "epoch": 5.1178559896674205,
+      "grad_norm": 0.6444790363311768,
+      "learning_rate": 0.0002933670940631397,
+      "loss": 3.3023,
+      "step": 47550
+    },
+    {
+      "epoch": 5.123237541707028,
+      "grad_norm": 0.6432087421417236,
+      "learning_rate": 0.0002930438530330783,
+      "loss": 3.3043,
+      "step": 47600
+    },
+    {
+      "epoch": 5.128619093746637,
+      "grad_norm": 0.663712739944458,
+      "learning_rate": 0.0002927206120030169,
+      "loss": 3.3056,
+      "step": 47650
+    },
+    {
+      "epoch": 5.134000645786244,
+      "grad_norm": 0.6301243305206299,
+      "learning_rate": 0.0002923973709729555,
+      "loss": 3.3138,
+      "step": 47700
+    },
+    {
+      "epoch": 5.139382197825853,
+      "grad_norm": 0.6672236919403076,
+      "learning_rate": 0.00029207412994289403,
+      "loss": 3.3112,
+      "step": 47750
+    },
+    {
+      "epoch": 5.1447637498654615,
+      "grad_norm": 0.7072641253471375,
+      "learning_rate": 0.0002917508889128327,
+      "loss": 3.3353,
+      "step": 47800
+    },
+    {
+      "epoch": 5.150145301905069,
+      "grad_norm": 0.6850908994674683,
+      "learning_rate": 0.0002914276478827712,
+      "loss": 3.3226,
+      "step": 47850
+    },
+    {
+      "epoch": 5.155526853944678,
+      "grad_norm": 0.6990956664085388,
+      "learning_rate": 0.0002911044068527098,
+      "loss": 3.3258,
+      "step": 47900
+    },
+    {
+      "epoch": 5.160908405984286,
+      "grad_norm": 0.6988109350204468,
+      "learning_rate": 0.0002907811658226484,
+      "loss": 3.3262,
+      "step": 47950
+    },
+    {
+      "epoch": 5.166289958023894,
+      "grad_norm": 0.6829333305358887,
+      "learning_rate": 0.000290457924792587,
+      "loss": 3.3228,
+      "step": 48000
+    },
+    {
+      "epoch": 5.166289958023894,
+      "eval_accuracy": 0.38084434376426407,
+      "eval_loss": 3.4127867221832275,
+      "eval_runtime": 185.4858,
+      "eval_samples_per_second": 97.102,
+      "eval_steps_per_second": 6.071,
+      "step": 48000
+    },
+    {
+      "epoch": 5.1716715100635025,
+      "grad_norm": 0.7065103650093079,
+      "learning_rate": 0.00029013468376252555,
+      "loss": 3.3291,
+      "step": 48050
+    },
+    {
+      "epoch": 5.17705306210311,
+      "grad_norm": 0.6891149282455444,
+      "learning_rate": 0.00028981144273246414,
+      "loss": 3.3102,
+      "step": 48100
+    },
+    {
+      "epoch": 5.182434614142719,
+      "grad_norm": 0.6729560494422913,
+      "learning_rate": 0.00028948820170240274,
+      "loss": 3.3144,
+      "step": 48150
+    },
+    {
+      "epoch": 5.187816166182327,
+      "grad_norm": 0.715303897857666,
+      "learning_rate": 0.00028916496067234133,
+      "loss": 3.3287,
+      "step": 48200
+    },
+    {
+      "epoch": 5.193197718221935,
+      "grad_norm": 0.6873436570167542,
+      "learning_rate": 0.0002888417196422799,
+      "loss": 3.3151,
+      "step": 48250
+    },
+    {
+      "epoch": 5.198579270261543,
+      "grad_norm": 0.7039670348167419,
+      "learning_rate": 0.00028851847861221847,
+      "loss": 3.3218,
+      "step": 48300
+    },
+    {
+      "epoch": 5.203960822301152,
+      "grad_norm": 0.6787541508674622,
+      "learning_rate": 0.00028819523758215706,
+      "loss": 3.3226,
+      "step": 48350
+    },
+    {
+      "epoch": 5.20934237434076,
+      "grad_norm": 0.7561299204826355,
+      "learning_rate": 0.00028787199655209566,
+      "loss": 3.3334,
+      "step": 48400
+    },
+    {
+      "epoch": 5.214723926380368,
+      "grad_norm": 0.6767032146453857,
+      "learning_rate": 0.00028754875552203425,
+      "loss": 3.3192,
+      "step": 48450
+    },
+    {
+      "epoch": 5.220105478419977,
+      "grad_norm": 0.7488175630569458,
+      "learning_rate": 0.0002872255144919728,
+      "loss": 3.3447,
+      "step": 48500
+    },
+    {
+      "epoch": 5.225487030459584,
+      "grad_norm": 0.7246031761169434,
+      "learning_rate": 0.00028690227346191144,
+      "loss": 3.3417,
+      "step": 48550
+    },
+    {
+      "epoch": 5.230868582499193,
+      "grad_norm": 0.6832336783409119,
+      "learning_rate": 0.00028657903243185,
+      "loss": 3.3203,
+      "step": 48600
+    },
+    {
+      "epoch": 5.236250134538801,
+      "grad_norm": 0.6544204354286194,
+      "learning_rate": 0.0002862557914017886,
+      "loss": 3.3245,
+      "step": 48650
+    },
+    {
+      "epoch": 5.241631686578409,
+      "grad_norm": 0.6821257472038269,
+      "learning_rate": 0.00028593255037172717,
+      "loss": 3.3221,
+      "step": 48700
+    },
+    {
+      "epoch": 5.247013238618018,
+      "grad_norm": 0.6964419484138489,
+      "learning_rate": 0.00028560930934166576,
+      "loss": 3.322,
+      "step": 48750
+    },
+    {
+      "epoch": 5.252394790657625,
+      "grad_norm": 0.670827329158783,
+      "learning_rate": 0.00028528606831160436,
+      "loss": 3.3382,
+      "step": 48800
+    },
+    {
+      "epoch": 5.257776342697234,
+      "grad_norm": 0.7108796834945679,
+      "learning_rate": 0.0002849628272815429,
+      "loss": 3.3093,
+      "step": 48850
+    },
+    {
+      "epoch": 5.2631578947368425,
+      "grad_norm": 0.7239094972610474,
+      "learning_rate": 0.0002846395862514815,
+      "loss": 3.3269,
+      "step": 48900
+    },
+    {
+      "epoch": 5.26853944677645,
+      "grad_norm": 0.6504336595535278,
+      "learning_rate": 0.0002843163452214201,
+      "loss": 3.3456,
+      "step": 48950
+    },
+    {
+      "epoch": 5.273920998816059,
+      "grad_norm": 0.7608343362808228,
+      "learning_rate": 0.0002839931041913587,
+      "loss": 3.3358,
+      "step": 49000
+    },
+    {
+      "epoch": 5.273920998816059,
+      "eval_accuracy": 0.3811183663473322,
+      "eval_loss": 3.411630153656006,
+      "eval_runtime": 185.2565,
+      "eval_samples_per_second": 97.222,
+      "eval_steps_per_second": 6.078,
+      "step": 49000
+    },
+    {
+      "epoch": 5.279302550855666,
+      "grad_norm": 0.7475415468215942,
+      "learning_rate": 0.0002836698631612972,
+      "loss": 3.3314,
+      "step": 49050
+    },
+    {
+      "epoch": 5.284684102895275,
+      "grad_norm": 0.6391634941101074,
+      "learning_rate": 0.00028334662213123587,
+      "loss": 3.3364,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2900656549348835,
+      "grad_norm": 0.7369605302810669,
+      "learning_rate": 0.0002830233811011744,
+      "loss": 3.3333,
+      "step": 49150
+    },
+    {
+      "epoch": 5.295447206974491,
+      "grad_norm": 0.6933754682540894,
+      "learning_rate": 0.000282700140071113,
+      "loss": 3.3551,
+      "step": 49200
+    },
+    {
+      "epoch": 5.3008287590141,
+      "grad_norm": 0.7052974700927734,
+      "learning_rate": 0.0002823768990410516,
+      "loss": 3.3382,
+      "step": 49250
+    },
+    {
+      "epoch": 5.306210311053708,
+      "grad_norm": 0.7149296402931213,
+      "learning_rate": 0.00028205365801099014,
+      "loss": 3.3241,
+      "step": 49300
+    },
+    {
+      "epoch": 5.311591863093316,
+      "grad_norm": 0.7133325338363647,
+      "learning_rate": 0.00028173041698092874,
+      "loss": 3.3421,
+      "step": 49350
+    },
+    {
+      "epoch": 5.316973415132924,
+      "grad_norm": 0.7447312474250793,
+      "learning_rate": 0.00028140717595086733,
+      "loss": 3.3281,
+      "step": 49400
+    },
+    {
+      "epoch": 5.322354967172533,
+      "grad_norm": 0.6949333548545837,
+      "learning_rate": 0.0002810839349208059,
+      "loss": 3.331,
+      "step": 49450
+    },
+    {
+      "epoch": 5.327736519212141,
+      "grad_norm": 0.6988310813903809,
+      "learning_rate": 0.00028076069389074447,
+      "loss": 3.3313,
+      "step": 49500
+    },
+    {
+      "epoch": 5.333118071251749,
+      "grad_norm": 0.6758862733840942,
+      "learning_rate": 0.0002804374528606831,
+      "loss": 3.337,
+      "step": 49550
+    },
+    {
+      "epoch": 5.338499623291357,
+      "grad_norm": 0.6512594819068909,
+      "learning_rate": 0.00028011421183062166,
+      "loss": 3.3432,
+      "step": 49600
+    },
+    {
+      "epoch": 5.343881175330965,
+      "grad_norm": 0.701414942741394,
+      "learning_rate": 0.00027979097080056025,
+      "loss": 3.3197,
+      "step": 49650
+    },
+    {
+      "epoch": 5.349262727370574,
+      "grad_norm": 0.6790468692779541,
+      "learning_rate": 0.00027946772977049885,
+      "loss": 3.3326,
+      "step": 49700
+    },
+    {
+      "epoch": 5.354644279410182,
+      "grad_norm": 0.699934184551239,
+      "learning_rate": 0.00027914448874043744,
+      "loss": 3.3334,
+      "step": 49750
+    },
+    {
+      "epoch": 5.36002583144979,
+      "grad_norm": 0.7084718346595764,
+      "learning_rate": 0.00027882124771037603,
+      "loss": 3.3459,
+      "step": 49800
+    },
+    {
+      "epoch": 5.365407383489399,
+      "grad_norm": 0.6838877201080322,
+      "learning_rate": 0.0002784980066803146,
+      "loss": 3.3359,
+      "step": 49850
+    },
+    {
+      "epoch": 5.370788935529006,
+      "grad_norm": 0.7035666704177856,
+      "learning_rate": 0.00027817476565025317,
+      "loss": 3.3315,
+      "step": 49900
+    },
+    {
+      "epoch": 5.376170487568615,
+      "grad_norm": 0.6787278652191162,
+      "learning_rate": 0.00027785152462019176,
+      "loss": 3.3403,
+      "step": 49950
+    },
+    {
+      "epoch": 5.3815520396082235,
+      "grad_norm": 0.6613674163818359,
+      "learning_rate": 0.00027752828359013036,
+      "loss": 3.3283,
+      "step": 50000
+    },
+    {
+      "epoch": 5.3815520396082235,
+      "eval_accuracy": 0.38175344247660786,
+      "eval_loss": 3.403359889984131,
+      "eval_runtime": 185.9116,
+      "eval_samples_per_second": 96.879,
+      "eval_steps_per_second": 6.057,
+      "step": 50000
+    },
+    {
+      "epoch": 5.386933591647831,
+      "grad_norm": 0.733452558517456,
+      "learning_rate": 0.0002772050425600689,
+      "loss": 3.3362,
+      "step": 50050
+    },
+    {
+      "epoch": 5.39231514368744,
+      "grad_norm": 0.7132042050361633,
+      "learning_rate": 0.00027688180153000755,
+      "loss": 3.3338,
+      "step": 50100
+    },
+    {
+      "epoch": 5.397696695727047,
+      "grad_norm": 0.7193716764450073,
+      "learning_rate": 0.0002765585604999461,
+      "loss": 3.3303,
+      "step": 50150
+    },
+    {
+      "epoch": 5.403078247766656,
+      "grad_norm": 0.663294792175293,
+      "learning_rate": 0.0002762353194698847,
+      "loss": 3.3411,
+      "step": 50200
+    },
+    {
+      "epoch": 5.4084597998062645,
+      "grad_norm": 0.682615339756012,
+      "learning_rate": 0.0002759120784398233,
+      "loss": 3.3325,
+      "step": 50250
+    },
+    {
+      "epoch": 5.413841351845872,
+      "grad_norm": 0.6866773366928101,
+      "learning_rate": 0.00027558883740976187,
+      "loss": 3.3444,
+      "step": 50300
+    },
+    {
+      "epoch": 5.419222903885481,
+      "grad_norm": 0.7201193571090698,
+      "learning_rate": 0.0002752655963797004,
+      "loss": 3.3305,
+      "step": 50350
+    },
+    {
+      "epoch": 5.424604455925088,
+      "grad_norm": 0.7807847261428833,
+      "learning_rate": 0.0002749488201702403,
+      "loss": 3.3466,
+      "step": 50400
+    },
+    {
+      "epoch": 5.429986007964697,
+      "grad_norm": 0.7107503414154053,
+      "learning_rate": 0.0002746255791401788,
+      "loss": 3.3247,
+      "step": 50450
+    },
+    {
+      "epoch": 5.435367560004305,
+      "grad_norm": 0.7089921236038208,
+      "learning_rate": 0.0002743023381101174,
+      "loss": 3.3145,
+      "step": 50500
+    },
+    {
+      "epoch": 5.440749112043913,
+      "grad_norm": 0.7293418645858765,
+      "learning_rate": 0.000273979097080056,
+      "loss": 3.3432,
+      "step": 50550
+    },
+    {
+      "epoch": 5.446130664083522,
+      "grad_norm": 0.6610980033874512,
+      "learning_rate": 0.0002736558560499946,
+      "loss": 3.3245,
+      "step": 50600
+    },
+    {
+      "epoch": 5.45151221612313,
+      "grad_norm": 0.9671128392219543,
+      "learning_rate": 0.0002733326150199332,
+      "loss": 3.3421,
+      "step": 50650
+    },
+    {
+      "epoch": 5.456893768162738,
+      "grad_norm": 0.7456415295600891,
+      "learning_rate": 0.00027300937398987173,
+      "loss": 3.3256,
+      "step": 50700
+    },
+    {
+      "epoch": 5.462275320202346,
+      "grad_norm": 0.7163481712341309,
+      "learning_rate": 0.0002726861329598104,
+      "loss": 3.3545,
+      "step": 50750
+    },
+    {
+      "epoch": 5.467656872241955,
+      "grad_norm": 0.6905017495155334,
+      "learning_rate": 0.0002723628919297489,
+      "loss": 3.3196,
+      "step": 50800
+    },
+    {
+      "epoch": 5.473038424281563,
+      "grad_norm": 0.6870000958442688,
+      "learning_rate": 0.0002720396508996875,
+      "loss": 3.3231,
+      "step": 50850
+    },
+    {
+      "epoch": 5.478419976321171,
+      "grad_norm": 0.7138904333114624,
+      "learning_rate": 0.0002717164098696261,
+      "loss": 3.3389,
+      "step": 50900
+    },
+    {
+      "epoch": 5.483801528360779,
+      "grad_norm": 0.725881040096283,
+      "learning_rate": 0.00027139316883956465,
+      "loss": 3.3138,
+      "step": 50950
+    },
+    {
+      "epoch": 5.489183080400387,
+      "grad_norm": 0.7014544010162354,
+      "learning_rate": 0.00027106992780950325,
+      "loss": 3.3431,
+      "step": 51000
+    },
+    {
+      "epoch": 5.489183080400387,
+      "eval_accuracy": 0.38186068287688074,
+      "eval_loss": 3.402552604675293,
+      "eval_runtime": 185.4576,
+      "eval_samples_per_second": 97.117,
+      "eval_steps_per_second": 6.071,
+      "step": 51000
+    },
+    {
+      "epoch": 5.494564632439996,
+      "grad_norm": 0.758173942565918,
+      "learning_rate": 0.00027074668677944184,
+      "loss": 3.3294,
+      "step": 51050
+    },
+    {
+      "epoch": 5.499946184479604,
+      "grad_norm": 0.6955305933952332,
+      "learning_rate": 0.00027042344574938044,
+      "loss": 3.3494,
+      "step": 51100
+    },
+    {
+      "epoch": 5.505327736519212,
+      "grad_norm": 0.7428892850875854,
+      "learning_rate": 0.000270100204719319,
+      "loss": 3.3361,
+      "step": 51150
+    },
+    {
+      "epoch": 5.510709288558821,
+      "grad_norm": 0.7045648097991943,
+      "learning_rate": 0.0002697769636892576,
+      "loss": 3.3276,
+      "step": 51200
+    },
+    {
+      "epoch": 5.516090840598428,
+      "grad_norm": 0.7683273553848267,
+      "learning_rate": 0.00026945372265919617,
+      "loss": 3.34,
+      "step": 51250
+    },
+    {
+      "epoch": 5.521472392638037,
+      "grad_norm": 0.6977889537811279,
+      "learning_rate": 0.00026913048162913476,
+      "loss": 3.3156,
+      "step": 51300
+    },
+    {
+      "epoch": 5.5268539446776455,
+      "grad_norm": 0.7473436594009399,
+      "learning_rate": 0.00026880724059907336,
+      "loss": 3.3388,
+      "step": 51350
+    },
+    {
+      "epoch": 5.532235496717253,
+      "grad_norm": 0.7338312268257141,
+      "learning_rate": 0.00026848399956901195,
+      "loss": 3.3393,
+      "step": 51400
+    },
+    {
+      "epoch": 5.537617048756862,
+      "grad_norm": 0.6995713710784912,
+      "learning_rate": 0.0002681607585389505,
+      "loss": 3.3201,
+      "step": 51450
+    },
+    {
+      "epoch": 5.542998600796469,
+      "grad_norm": 0.684768795967102,
+      "learning_rate": 0.0002678375175088891,
+      "loss": 3.321,
+      "step": 51500
+    },
+    {
+      "epoch": 5.548380152836078,
+      "grad_norm": 0.7504066228866577,
+      "learning_rate": 0.0002675142764788277,
+      "loss": 3.3305,
+      "step": 51550
+    },
+    {
+      "epoch": 5.553761704875686,
+      "grad_norm": 0.687995433807373,
+      "learning_rate": 0.0002671910354487663,
+      "loss": 3.3544,
+      "step": 51600
+    },
+    {
+      "epoch": 5.559143256915294,
+      "grad_norm": 0.7170462012290955,
+      "learning_rate": 0.00026686779441870487,
+      "loss": 3.3206,
+      "step": 51650
+    },
+    {
+      "epoch": 5.564524808954903,
+      "grad_norm": 0.7041773796081543,
+      "learning_rate": 0.0002665445533886434,
+      "loss": 3.3604,
+      "step": 51700
+    },
+    {
+      "epoch": 5.569906360994511,
+      "grad_norm": 0.7011191248893738,
+      "learning_rate": 0.00026622131235858206,
+      "loss": 3.3446,
+      "step": 51750
+    },
+    {
+      "epoch": 5.575287913034119,
+      "grad_norm": 0.8032184839248657,
+      "learning_rate": 0.0002658980713285206,
+      "loss": 3.3336,
+      "step": 51800
+    },
+    {
+      "epoch": 5.580669465073727,
+      "grad_norm": 0.6872266530990601,
+      "learning_rate": 0.0002655748302984592,
+      "loss": 3.3379,
+      "step": 51850
+    },
+    {
+      "epoch": 5.586051017113336,
+      "grad_norm": 0.7369017004966736,
+      "learning_rate": 0.0002652515892683978,
+      "loss": 3.3358,
+      "step": 51900
+    },
+    {
+      "epoch": 5.591432569152944,
+      "grad_norm": 0.7350853085517883,
+      "learning_rate": 0.0002649283482383364,
+      "loss": 3.3207,
+      "step": 51950
+    },
+    {
+      "epoch": 5.596814121192552,
+      "grad_norm": 0.8044530749320984,
+      "learning_rate": 0.0002646051072082749,
+      "loss": 3.3383,
+      "step": 52000
+    },
+    {
+      "epoch": 5.596814121192552,
+      "eval_accuracy": 0.38261190894323005,
+      "eval_loss": 3.393895149230957,
+      "eval_runtime": 185.5111,
+      "eval_samples_per_second": 97.089,
+      "eval_steps_per_second": 6.07,
+      "step": 52000
+    },
+    {
+      "epoch": 5.60219567323216,
+      "grad_norm": 0.6660870909690857,
+      "learning_rate": 0.0002642818661782135,
+      "loss": 3.321,
+      "step": 52050
+    },
+    {
+      "epoch": 5.607577225271768,
+      "grad_norm": 0.6830106973648071,
+      "learning_rate": 0.0002639586251481521,
+      "loss": 3.3228,
+      "step": 52100
+    },
+    {
+      "epoch": 5.612958777311377,
+      "grad_norm": 0.6999291777610779,
+      "learning_rate": 0.0002636353841180907,
+      "loss": 3.3562,
+      "step": 52150
+    },
+    {
+      "epoch": 5.618340329350985,
+      "grad_norm": 0.7501614093780518,
+      "learning_rate": 0.0002633121430880293,
+      "loss": 3.3432,
+      "step": 52200
+    },
+    {
+      "epoch": 5.623721881390593,
+      "grad_norm": 0.6782894134521484,
+      "learning_rate": 0.00026298890205796784,
+      "loss": 3.3359,
+      "step": 52250
+    },
+    {
+      "epoch": 5.629103433430201,
+      "grad_norm": 0.6987687945365906,
+      "learning_rate": 0.00026266566102790644,
+      "loss": 3.3359,
+      "step": 52300
+    },
+    {
+      "epoch": 5.634484985469809,
+      "grad_norm": 0.6932606101036072,
+      "learning_rate": 0.00026234241999784503,
+      "loss": 3.3345,
+      "step": 52350
+    },
+    {
+      "epoch": 5.639866537509418,
+      "grad_norm": 0.7144378423690796,
+      "learning_rate": 0.0002620191789677836,
+      "loss": 3.3264,
+      "step": 52400
+    },
+    {
+      "epoch": 5.645248089549026,
+      "grad_norm": 0.7118236422538757,
+      "learning_rate": 0.00026169593793772217,
+      "loss": 3.3348,
+      "step": 52450
+    },
+    {
+      "epoch": 5.650629641588634,
+      "grad_norm": 0.7649123072624207,
+      "learning_rate": 0.0002613726969076608,
+      "loss": 3.327,
+      "step": 52500
+    },
+    {
+      "epoch": 5.656011193628243,
+      "grad_norm": 0.7233064770698547,
+      "learning_rate": 0.00026104945587759936,
+      "loss": 3.3222,
+      "step": 52550
+    },
+    {
+      "epoch": 5.66139274566785,
+      "grad_norm": 0.7189932465553284,
+      "learning_rate": 0.00026072621484753795,
+      "loss": 3.3528,
+      "step": 52600
+    },
+    {
+      "epoch": 5.666774297707459,
+      "grad_norm": 0.7032601237297058,
+      "learning_rate": 0.00026040297381747655,
+      "loss": 3.3345,
+      "step": 52650
+    },
+    {
+      "epoch": 5.672155849747067,
+      "grad_norm": 0.7438763976097107,
+      "learning_rate": 0.00026007973278741514,
+      "loss": 3.3398,
+      "step": 52700
+    },
+    {
+      "epoch": 5.677537401786675,
+      "grad_norm": 0.7075427174568176,
+      "learning_rate": 0.00025975649175735373,
+      "loss": 3.3387,
+      "step": 52750
+    },
+    {
+      "epoch": 5.682918953826284,
+      "grad_norm": 0.7556169033050537,
+      "learning_rate": 0.0002594332507272923,
+      "loss": 3.3379,
+      "step": 52800
+    },
+    {
+      "epoch": 5.688300505865891,
+      "grad_norm": 0.7540663480758667,
+      "learning_rate": 0.00025911647451783214,
+      "loss": 3.3285,
+      "step": 52850
+    },
+    {
+      "epoch": 5.6936820579055,
+      "grad_norm": 0.6908034682273865,
+      "learning_rate": 0.0002587932334877707,
+      "loss": 3.3309,
+      "step": 52900
+    },
+    {
+      "epoch": 5.699063609945108,
+      "grad_norm": 0.7552903890609741,
+      "learning_rate": 0.00025846999245770927,
+      "loss": 3.3334,
+      "step": 52950
+    },
+    {
+      "epoch": 5.704445161984716,
+      "grad_norm": 0.7142480611801147,
+      "learning_rate": 0.00025814675142764787,
+      "loss": 3.3608,
+      "step": 53000
+    },
+    {
+      "epoch": 5.704445161984716,
+      "eval_accuracy": 0.3832693675673955,
+      "eval_loss": 3.3909168243408203,
+      "eval_runtime": 185.7589,
+      "eval_samples_per_second": 96.959,
+      "eval_steps_per_second": 6.062,
+      "step": 53000
+    },
+    {
+      "epoch": 5.709826714024325,
+      "grad_norm": 0.7215988636016846,
+      "learning_rate": 0.00025782351039758646,
+      "loss": 3.3398,
+      "step": 53050
+    },
+    {
+      "epoch": 5.715208266063933,
+      "grad_norm": 0.7397409081459045,
+      "learning_rate": 0.000257500269367525,
+      "loss": 3.3299,
+      "step": 53100
+    },
+    {
+      "epoch": 5.720589818103541,
+      "grad_norm": 0.7068989872932434,
+      "learning_rate": 0.0002571770283374636,
+      "loss": 3.3416,
+      "step": 53150
+    },
+    {
+      "epoch": 5.725971370143149,
+      "grad_norm": 0.7299404144287109,
+      "learning_rate": 0.0002568537873074022,
+      "loss": 3.3325,
+      "step": 53200
+    },
+    {
+      "epoch": 5.731352922182758,
+      "grad_norm": 0.6692743301391602,
+      "learning_rate": 0.0002565305462773408,
+      "loss": 3.3225,
+      "step": 53250
+    },
+    {
+      "epoch": 5.736734474222366,
+      "grad_norm": 0.7195047736167908,
+      "learning_rate": 0.0002562073052472794,
+      "loss": 3.339,
+      "step": 53300
+    },
+    {
+      "epoch": 5.742116026261974,
+      "grad_norm": 0.6635109186172485,
+      "learning_rate": 0.0002558840642172179,
+      "loss": 3.3364,
+      "step": 53350
+    },
+    {
+      "epoch": 5.747497578301582,
+      "grad_norm": 0.8553937673568726,
+      "learning_rate": 0.00025556082318715657,
+      "loss": 3.3108,
+      "step": 53400
+    },
+    {
+      "epoch": 5.75287913034119,
+      "grad_norm": 0.7185947299003601,
+      "learning_rate": 0.0002552375821570951,
+      "loss": 3.3427,
+      "step": 53450
+    },
+    {
+      "epoch": 5.758260682380799,
+      "grad_norm": 0.712882936000824,
+      "learning_rate": 0.0002549143411270337,
+      "loss": 3.3366,
+      "step": 53500
+    },
+    {
+      "epoch": 5.763642234420407,
+      "grad_norm": 0.7325207591056824,
+      "learning_rate": 0.0002545911000969723,
+      "loss": 3.3438,
+      "step": 53550
+    },
+    {
+      "epoch": 5.769023786460015,
+      "grad_norm": 0.731934130191803,
+      "learning_rate": 0.0002542678590669109,
+      "loss": 3.3219,
+      "step": 53600
+    },
+    {
+      "epoch": 5.774405338499624,
+      "grad_norm": 0.7562335133552551,
+      "learning_rate": 0.00025394461803684943,
+      "loss": 3.3353,
+      "step": 53650
+    },
+    {
+      "epoch": 5.779786890539231,
+      "grad_norm": 0.717185378074646,
+      "learning_rate": 0.00025362137700678803,
+      "loss": 3.3252,
+      "step": 53700
+    },
+    {
+      "epoch": 5.78516844257884,
+      "grad_norm": 0.698867917060852,
+      "learning_rate": 0.0002532981359767266,
+      "loss": 3.3297,
+      "step": 53750
+    },
+    {
+      "epoch": 5.790549994618448,
+      "grad_norm": 0.6668176054954529,
+      "learning_rate": 0.0002529748949466652,
+      "loss": 3.3428,
+      "step": 53800
+    },
+    {
+      "epoch": 5.795931546658056,
+      "grad_norm": 0.7497868537902832,
+      "learning_rate": 0.0002526516539166038,
+      "loss": 3.3363,
+      "step": 53850
+    },
+    {
+      "epoch": 5.801313098697665,
+      "grad_norm": 0.704843282699585,
+      "learning_rate": 0.00025232841288654235,
+      "loss": 3.3362,
+      "step": 53900
+    },
+    {
+      "epoch": 5.806694650737272,
+      "grad_norm": 0.7422738075256348,
+      "learning_rate": 0.00025200517185648095,
+      "loss": 3.3463,
+      "step": 53950
+    },
+    {
+      "epoch": 5.812076202776881,
+      "grad_norm": 0.7051922082901001,
+      "learning_rate": 0.00025168193082641954,
+      "loss": 3.3379,
+      "step": 54000
+    },
+    {
+      "epoch": 5.812076202776881,
+      "eval_accuracy": 0.38340464041272454,
+      "eval_loss": 3.38680100440979,
+      "eval_runtime": 185.4763,
+      "eval_samples_per_second": 97.107,
+      "eval_steps_per_second": 6.071,
+      "step": 54000
+    },
+    {
+      "epoch": 5.817457754816489,
+      "grad_norm": 0.7407611608505249,
+      "learning_rate": 0.00025135868979635814,
+      "loss": 3.3502,
+      "step": 54050
+    },
+    {
+      "epoch": 5.822839306856097,
+      "grad_norm": 0.7295721769332886,
+      "learning_rate": 0.0002510354487662967,
+      "loss": 3.337,
+      "step": 54100
+    },
+    {
+      "epoch": 5.828220858895706,
+      "grad_norm": 0.7691437005996704,
+      "learning_rate": 0.0002507122077362353,
+      "loss": 3.3366,
+      "step": 54150
+    },
+    {
+      "epoch": 5.833602410935313,
+      "grad_norm": 0.7400103211402893,
+      "learning_rate": 0.00025038896670617387,
+      "loss": 3.344,
+      "step": 54200
+    },
+    {
+      "epoch": 5.838983962974922,
+      "grad_norm": 0.7032003402709961,
+      "learning_rate": 0.00025006572567611246,
+      "loss": 3.3371,
+      "step": 54250
+    },
+    {
+      "epoch": 5.84436551501453,
+      "grad_norm": 0.7601761817932129,
+      "learning_rate": 0.00024974248464605106,
+      "loss": 3.3301,
+      "step": 54300
+    },
+    {
+      "epoch": 5.849747067054138,
+      "grad_norm": 0.7390912771224976,
+      "learning_rate": 0.00024941924361598965,
+      "loss": 3.3324,
+      "step": 54350
+    },
+    {
+      "epoch": 5.855128619093747,
+      "grad_norm": 0.7346526384353638,
+      "learning_rate": 0.00024909600258592825,
+      "loss": 3.3404,
+      "step": 54400
+    },
+    {
+      "epoch": 5.860510171133355,
+      "grad_norm": 0.711598813533783,
+      "learning_rate": 0.0002487727615558668,
+      "loss": 3.3352,
+      "step": 54450
+    },
+    {
+      "epoch": 5.865891723172963,
+      "grad_norm": 0.7088502049446106,
+      "learning_rate": 0.0002484495205258054,
+      "loss": 3.3254,
+      "step": 54500
+    },
+    {
+      "epoch": 5.871273275212571,
+      "grad_norm": 0.742213249206543,
+      "learning_rate": 0.000248126279495744,
+      "loss": 3.3508,
+      "step": 54550
+    },
+    {
+      "epoch": 5.87665482725218,
+      "grad_norm": 0.7339125275611877,
+      "learning_rate": 0.00024780303846568257,
+      "loss": 3.3257,
+      "step": 54600
+    },
+    {
+      "epoch": 5.882036379291788,
+      "grad_norm": 0.718253493309021,
+      "learning_rate": 0.0002474797974356211,
+      "loss": 3.3375,
+      "step": 54650
+    },
+    {
+      "epoch": 5.887417931331396,
+      "grad_norm": 0.7570130228996277,
+      "learning_rate": 0.00024715655640555976,
+      "loss": 3.3329,
+      "step": 54700
+    },
+    {
+      "epoch": 5.892799483371004,
+      "grad_norm": 0.7305703163146973,
+      "learning_rate": 0.0002468333153754983,
+      "loss": 3.3268,
+      "step": 54750
+    },
+    {
+      "epoch": 5.898181035410612,
+      "grad_norm": 0.7191780209541321,
+      "learning_rate": 0.0002465100743454369,
+      "loss": 3.3346,
+      "step": 54800
+    },
+    {
+      "epoch": 5.903562587450221,
+      "grad_norm": 0.703218936920166,
+      "learning_rate": 0.0002461868333153755,
+      "loss": 3.359,
+      "step": 54850
+    },
+    {
+      "epoch": 5.9089441394898286,
+      "grad_norm": 0.7352034449577332,
+      "learning_rate": 0.0002458635922853141,
+      "loss": 3.3591,
+      "step": 54900
+    },
+    {
+      "epoch": 5.914325691529437,
+      "grad_norm": 0.7736100554466248,
+      "learning_rate": 0.0002455403512552526,
+      "loss": 3.3425,
+      "step": 54950
+    },
+    {
+      "epoch": 5.919707243569046,
+      "grad_norm": 0.7174371480941772,
+      "learning_rate": 0.0002452171102251912,
+      "loss": 3.335,
+      "step": 55000
+    },
+    {
+      "epoch": 5.919707243569046,
+      "eval_accuracy": 0.3835768552399105,
+      "eval_loss": 3.3835244178771973,
+      "eval_runtime": 185.5374,
+      "eval_samples_per_second": 97.075,
+      "eval_steps_per_second": 6.069,
+      "step": 55000
+    },
+    {
+      "epoch": 5.925088795608653,
+      "grad_norm": 0.7378474473953247,
+      "learning_rate": 0.0002448938691951298,
+      "loss": 3.3232,
+      "step": 55050
+    },
+    {
+      "epoch": 5.930470347648262,
+      "grad_norm": 0.766350269317627,
+      "learning_rate": 0.0002445706281650684,
+      "loss": 3.3273,
+      "step": 55100
+    },
+    {
+      "epoch": 5.93585189968787,
+      "grad_norm": 0.6800521612167358,
+      "learning_rate": 0.000244247387135007,
+      "loss": 3.3414,
+      "step": 55150
+    },
+    {
+      "epoch": 5.941233451727478,
+      "grad_norm": 0.7534164190292358,
+      "learning_rate": 0.0002439306109255468,
+      "loss": 3.3353,
+      "step": 55200
+    },
+    {
+      "epoch": 5.946615003767087,
+      "grad_norm": 0.703264594078064,
+      "learning_rate": 0.00024360736989548538,
+      "loss": 3.3429,
+      "step": 55250
+    },
+    {
+      "epoch": 5.951996555806694,
+      "grad_norm": 0.7734049558639526,
+      "learning_rate": 0.00024328412886542394,
+      "loss": 3.3348,
+      "step": 55300
+    },
+    {
+      "epoch": 5.957378107846303,
+      "grad_norm": 0.7409945726394653,
+      "learning_rate": 0.00024296088783536257,
+      "loss": 3.3396,
+      "step": 55350
+    },
+    {
+      "epoch": 5.962759659885911,
+      "grad_norm": 0.783704400062561,
+      "learning_rate": 0.00024263764680530113,
+      "loss": 3.3395,
+      "step": 55400
+    },
+    {
+      "epoch": 5.968141211925519,
+      "grad_norm": 0.7635119557380676,
+      "learning_rate": 0.0002423144057752397,
+      "loss": 3.3409,
+      "step": 55450
+    },
+    {
+      "epoch": 5.973522763965128,
+      "grad_norm": 0.703266978263855,
+      "learning_rate": 0.00024199116474517832,
+      "loss": 3.3156,
+      "step": 55500
+    },
+    {
+      "epoch": 5.978904316004736,
+      "grad_norm": 0.7971009612083435,
+      "learning_rate": 0.0002416679237151169,
+      "loss": 3.3301,
+      "step": 55550
+    },
+    {
+      "epoch": 5.984285868044344,
+      "grad_norm": 0.778295636177063,
+      "learning_rate": 0.00024134468268505546,
+      "loss": 3.3348,
+      "step": 55600
+    },
+    {
+      "epoch": 5.989667420083952,
+      "grad_norm": 0.7256489396095276,
+      "learning_rate": 0.00024102144165499405,
+      "loss": 3.35,
+      "step": 55650
+    },
+    {
+      "epoch": 5.995048972123561,
+      "grad_norm": 0.6783924102783203,
+      "learning_rate": 0.00024069820062493265,
+      "loss": 3.3272,
+      "step": 55700
+    },
+    {
+      "epoch": 6.000430524163169,
+      "grad_norm": 0.7053371071815491,
+      "learning_rate": 0.00024037495959487121,
+      "loss": 3.3418,
+      "step": 55750
+    },
+    {
+      "epoch": 6.005812076202777,
+      "grad_norm": 0.7383301854133606,
+      "learning_rate": 0.0002400517185648098,
+      "loss": 3.2308,
+      "step": 55800
+    },
+    {
+      "epoch": 6.011193628242385,
+      "grad_norm": 0.7670775651931763,
+      "learning_rate": 0.00023972847753474838,
+      "loss": 3.2593,
+      "step": 55850
+    },
+    {
+      "epoch": 6.016575180281993,
+      "grad_norm": 0.78084397315979,
+      "learning_rate": 0.00023940523650468697,
+      "loss": 3.2608,
+      "step": 55900
+    },
+    {
+      "epoch": 6.021956732321602,
+      "grad_norm": 0.7219025492668152,
+      "learning_rate": 0.00023908199547462557,
+      "loss": 3.2278,
+      "step": 55950
+    },
+    {
+      "epoch": 6.0273382843612096,
+      "grad_norm": 0.7354509234428406,
+      "learning_rate": 0.00023875875444456413,
+      "loss": 3.2537,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0273382843612096,
+      "eval_accuracy": 0.3841397858516773,
+      "eval_loss": 3.3841917514801025,
+      "eval_runtime": 185.4923,
+      "eval_samples_per_second": 97.098,
+      "eval_steps_per_second": 6.07,
+      "step": 56000
+    },
+    {
+      "epoch": 6.032719836400818,
+      "grad_norm": 0.7289640307426453,
+      "learning_rate": 0.0002384355134145027,
+      "loss": 3.2428,
+      "step": 56050
+    },
+    {
+      "epoch": 6.038101388440427,
+      "grad_norm": 0.8031316995620728,
+      "learning_rate": 0.00023811227238444132,
+      "loss": 3.2619,
+      "step": 56100
+    },
+    {
+      "epoch": 6.043482940480034,
+      "grad_norm": 0.7483075261116028,
+      "learning_rate": 0.0002377890313543799,
+      "loss": 3.277,
+      "step": 56150
+    },
+    {
+      "epoch": 6.048864492519643,
+      "grad_norm": 0.7377139925956726,
+      "learning_rate": 0.00023746579032431849,
+      "loss": 3.251,
+      "step": 56200
+    },
+    {
+      "epoch": 6.0542460445592505,
+      "grad_norm": 0.8221212029457092,
+      "learning_rate": 0.00023714254929425708,
+      "loss": 3.2444,
+      "step": 56250
+    },
+    {
+      "epoch": 6.059627596598859,
+      "grad_norm": 0.7545392513275146,
+      "learning_rate": 0.00023681930826419565,
+      "loss": 3.2673,
+      "step": 56300
+    },
+    {
+      "epoch": 6.065009148638468,
+      "grad_norm": 0.7426929473876953,
+      "learning_rate": 0.00023649606723413424,
+      "loss": 3.241,
+      "step": 56350
+    },
+    {
+      "epoch": 6.070390700678075,
+      "grad_norm": 0.7875537872314453,
+      "learning_rate": 0.0002361728262040728,
+      "loss": 3.2486,
+      "step": 56400
+    },
+    {
+      "epoch": 6.075772252717684,
+      "grad_norm": 0.7852635979652405,
+      "learning_rate": 0.00023584958517401138,
+      "loss": 3.2515,
+      "step": 56450
+    },
+    {
+      "epoch": 6.081153804757292,
+      "grad_norm": 0.7957573533058167,
+      "learning_rate": 0.00023552634414395,
+      "loss": 3.2585,
+      "step": 56500
+    },
+    {
+      "epoch": 6.0865353567969,
+      "grad_norm": 0.774386465549469,
+      "learning_rate": 0.00023520310311388857,
+      "loss": 3.2489,
+      "step": 56550
+    },
+    {
+      "epoch": 6.091916908836509,
+      "grad_norm": 0.758723795413971,
+      "learning_rate": 0.00023487986208382713,
+      "loss": 3.2669,
+      "step": 56600
+    },
+    {
+      "epoch": 6.097298460876116,
+      "grad_norm": 0.767099916934967,
+      "learning_rate": 0.00023455662105376576,
+      "loss": 3.2598,
+      "step": 56650
+    },
+    {
+      "epoch": 6.102680012915725,
+      "grad_norm": 0.7581093907356262,
+      "learning_rate": 0.00023423338002370432,
+      "loss": 3.2556,
+      "step": 56700
+    },
+    {
+      "epoch": 6.108061564955333,
+      "grad_norm": 0.7787352800369263,
+      "learning_rate": 0.0002339101389936429,
+      "loss": 3.2607,
+      "step": 56750
+    },
+    {
+      "epoch": 6.113443116994941,
+      "grad_norm": 0.7579735517501831,
+      "learning_rate": 0.00023358689796358149,
+      "loss": 3.2616,
+      "step": 56800
+    },
+    {
+      "epoch": 6.11882466903455,
+      "grad_norm": 0.7276836633682251,
+      "learning_rate": 0.00023326365693352008,
+      "loss": 3.248,
+      "step": 56850
+    },
+    {
+      "epoch": 6.124206221074158,
+      "grad_norm": 0.7455276250839233,
+      "learning_rate": 0.00023294041590345865,
+      "loss": 3.2639,
+      "step": 56900
+    },
+    {
+      "epoch": 6.129587773113766,
+      "grad_norm": 0.7798923850059509,
+      "learning_rate": 0.00023261717487339724,
+      "loss": 3.2627,
+      "step": 56950
+    },
+    {
+      "epoch": 6.134969325153374,
+      "grad_norm": 0.7457557916641235,
+      "learning_rate": 0.0002322939338433358,
+      "loss": 3.2644,
+      "step": 57000
+    },
+    {
+      "epoch": 6.134969325153374,
+      "eval_accuracy": 0.3839235666049265,
+      "eval_loss": 3.386077404022217,
+      "eval_runtime": 185.7152,
+      "eval_samples_per_second": 96.982,
+      "eval_steps_per_second": 6.063,
+      "step": 57000
+    },
+    {
+      "epoch": 6.140350877192983,
+      "grad_norm": 0.7664401531219482,
+      "learning_rate": 0.0002319706928132744,
+      "loss": 3.266,
+      "step": 57050
+    },
+    {
+      "epoch": 6.1457324292325906,
+      "grad_norm": 0.8032768368721008,
+      "learning_rate": 0.000231647451783213,
+      "loss": 3.2536,
+      "step": 57100
+    },
+    {
+      "epoch": 6.151113981272199,
+      "grad_norm": 0.7738435864448547,
+      "learning_rate": 0.00023132421075315157,
+      "loss": 3.2583,
+      "step": 57150
+    },
+    {
+      "epoch": 6.156495533311807,
+      "grad_norm": 0.7568543553352356,
+      "learning_rate": 0.0002310009697230902,
+      "loss": 3.2865,
+      "step": 57200
+    },
+    {
+      "epoch": 6.161877085351415,
+      "grad_norm": 0.7438731789588928,
+      "learning_rate": 0.00023067772869302876,
+      "loss": 3.2716,
+      "step": 57250
+    },
+    {
+      "epoch": 6.167258637391024,
+      "grad_norm": 0.7933516502380371,
+      "learning_rate": 0.00023035448766296732,
+      "loss": 3.2693,
+      "step": 57300
+    },
+    {
+      "epoch": 6.1726401894306315,
+      "grad_norm": 0.7111642956733704,
+      "learning_rate": 0.00023003124663290592,
+      "loss": 3.2594,
+      "step": 57350
+    },
+    {
+      "epoch": 6.17802174147024,
+      "grad_norm": 0.773277759552002,
+      "learning_rate": 0.0002297080056028445,
+      "loss": 3.2683,
+      "step": 57400
+    },
+    {
+      "epoch": 6.183403293509849,
+      "grad_norm": 0.7642556428909302,
+      "learning_rate": 0.00022938476457278308,
+      "loss": 3.2932,
+      "step": 57450
+    },
+    {
+      "epoch": 6.188784845549456,
+      "grad_norm": 0.7450052499771118,
+      "learning_rate": 0.00022906152354272168,
+      "loss": 3.281,
+      "step": 57500
+    },
+    {
+      "epoch": 6.194166397589065,
+      "grad_norm": 0.7851862907409668,
+      "learning_rate": 0.00022873828251266024,
+      "loss": 3.2602,
+      "step": 57550
+    },
+    {
+      "epoch": 6.1995479496286725,
+      "grad_norm": 0.7528820633888245,
+      "learning_rate": 0.00022841504148259884,
+      "loss": 3.2719,
+      "step": 57600
+    },
+    {
+      "epoch": 6.204929501668281,
+      "grad_norm": 0.7508034706115723,
+      "learning_rate": 0.00022809180045253743,
+      "loss": 3.2716,
+      "step": 57650
+    },
+    {
+      "epoch": 6.21031105370789,
+      "grad_norm": 0.7565895318984985,
+      "learning_rate": 0.000227768559422476,
+      "loss": 3.2751,
+      "step": 57700
+    },
+    {
+      "epoch": 6.215692605747497,
+      "grad_norm": 0.6810287833213806,
+      "learning_rate": 0.00022744531839241457,
+      "loss": 3.2633,
+      "step": 57750
+    },
+    {
+      "epoch": 6.221074157787106,
+      "grad_norm": 0.7242906093597412,
+      "learning_rate": 0.0002271220773623532,
+      "loss": 3.2827,
+      "step": 57800
+    },
+    {
+      "epoch": 6.226455709826714,
+      "grad_norm": 0.7394772171974182,
+      "learning_rate": 0.00022679883633229176,
+      "loss": 3.2501,
+      "step": 57850
+    },
+    {
+      "epoch": 6.231837261866322,
+      "grad_norm": 0.7512344121932983,
+      "learning_rate": 0.00022647559530223032,
+      "loss": 3.2657,
+      "step": 57900
+    },
+    {
+      "epoch": 6.237218813905931,
+      "grad_norm": 0.79515141248703,
+      "learning_rate": 0.00022615235427216895,
+      "loss": 3.2681,
+      "step": 57950
+    },
+    {
+      "epoch": 6.242600365945538,
+      "grad_norm": 0.7382014989852905,
+      "learning_rate": 0.0002258291132421075,
+      "loss": 3.2737,
+      "step": 58000
+    },
+    {
+      "epoch": 6.242600365945538,
+      "eval_accuracy": 0.38454593034632784,
+      "eval_loss": 3.382006883621216,
+      "eval_runtime": 185.7297,
+      "eval_samples_per_second": 96.974,
+      "eval_steps_per_second": 6.063,
+      "step": 58000
+    },
+    {
+      "epoch": 6.247981917985147,
+      "grad_norm": 0.8108021020889282,
+      "learning_rate": 0.0002255058722120461,
+      "loss": 3.2884,
+      "step": 58050
+    },
+    {
+      "epoch": 6.253363470024755,
+      "grad_norm": 0.7890158891677856,
+      "learning_rate": 0.00022518263118198468,
+      "loss": 3.2859,
+      "step": 58100
+    },
+    {
+      "epoch": 6.258745022064363,
+      "grad_norm": 0.7976743578910828,
+      "learning_rate": 0.00022485939015192327,
+      "loss": 3.2755,
+      "step": 58150
+    },
+    {
+      "epoch": 6.264126574103972,
+      "grad_norm": 0.748257040977478,
+      "learning_rate": 0.00022454261394246308,
+      "loss": 3.2899,
+      "step": 58200
+    },
+    {
+      "epoch": 6.26950812614358,
+      "grad_norm": 0.7588167190551758,
+      "learning_rate": 0.00022421937291240164,
+      "loss": 3.2758,
+      "step": 58250
+    },
+    {
+      "epoch": 6.274889678183188,
+      "grad_norm": 0.7655159831047058,
+      "learning_rate": 0.00022389613188234027,
+      "loss": 3.269,
+      "step": 58300
+    },
+    {
+      "epoch": 6.280271230222796,
+      "grad_norm": 0.7243340611457825,
+      "learning_rate": 0.00022357289085227883,
+      "loss": 3.2706,
+      "step": 58350
+    },
+    {
+      "epoch": 6.285652782262405,
+      "grad_norm": 0.7392916083335876,
+      "learning_rate": 0.0002232496498222174,
+      "loss": 3.2707,
+      "step": 58400
+    },
+    {
+      "epoch": 6.2910343343020125,
+      "grad_norm": 0.7228137850761414,
+      "learning_rate": 0.000222926408792156,
+      "loss": 3.2608,
+      "step": 58450
+    },
+    {
+      "epoch": 6.296415886341621,
+      "grad_norm": 0.7461840510368347,
+      "learning_rate": 0.0002226031677620946,
+      "loss": 3.2923,
+      "step": 58500
+    },
+    {
+      "epoch": 6.301797438381229,
+      "grad_norm": 0.7727910280227661,
+      "learning_rate": 0.00022227992673203316,
+      "loss": 3.2879,
+      "step": 58550
+    },
+    {
+      "epoch": 6.307178990420837,
+      "grad_norm": 0.7870025634765625,
+      "learning_rate": 0.00022195668570197175,
+      "loss": 3.282,
+      "step": 58600
+    },
+    {
+      "epoch": 6.312560542460446,
+      "grad_norm": 0.7641369104385376,
+      "learning_rate": 0.00022163344467191032,
+      "loss": 3.2693,
+      "step": 58650
+    },
+    {
+      "epoch": 6.3179420945000535,
+      "grad_norm": 0.776572585105896,
+      "learning_rate": 0.00022131020364184891,
+      "loss": 3.2868,
+      "step": 58700
+    },
+    {
+      "epoch": 6.323323646539662,
+      "grad_norm": 0.8025811314582825,
+      "learning_rate": 0.0002209869626117875,
+      "loss": 3.2701,
+      "step": 58750
+    },
+    {
+      "epoch": 6.328705198579271,
+      "grad_norm": 0.780965268611908,
+      "learning_rate": 0.00022066372158172608,
+      "loss": 3.2675,
+      "step": 58800
+    },
+    {
+      "epoch": 6.334086750618878,
+      "grad_norm": 0.775322675704956,
+      "learning_rate": 0.00022034048055166464,
+      "loss": 3.2817,
+      "step": 58850
+    },
+    {
+      "epoch": 6.339468302658487,
+      "grad_norm": 0.7998670935630798,
+      "learning_rate": 0.00022001723952160327,
+      "loss": 3.2744,
+      "step": 58900
+    },
+    {
+      "epoch": 6.344849854698095,
+      "grad_norm": 0.8096821904182434,
+      "learning_rate": 0.00021969399849154183,
+      "loss": 3.2793,
+      "step": 58950
+    },
+    {
+      "epoch": 6.350231406737703,
+      "grad_norm": 0.7768423557281494,
+      "learning_rate": 0.00021937075746148043,
+      "loss": 3.2764,
+      "step": 59000
+    },
+    {
+      "epoch": 6.350231406737703,
+      "eval_accuracy": 0.38469174252177485,
+      "eval_loss": 3.37949800491333,
+      "eval_runtime": 185.7739,
+      "eval_samples_per_second": 96.951,
+      "eval_steps_per_second": 6.061,
+      "step": 59000
+    },
+    {
+      "epoch": 6.355612958777312,
+      "grad_norm": 0.7386652231216431,
+      "learning_rate": 0.00021904751643141902,
+      "loss": 3.2864,
+      "step": 59050
+    },
+    {
+      "epoch": 6.360994510816919,
+      "grad_norm": 0.7582374811172485,
+      "learning_rate": 0.0002187242754013576,
+      "loss": 3.2706,
+      "step": 59100
+    },
+    {
+      "epoch": 6.366376062856528,
+      "grad_norm": 0.7769702672958374,
+      "learning_rate": 0.00021840103437129619,
+      "loss": 3.2884,
+      "step": 59150
+    },
+    {
+      "epoch": 6.371757614896136,
+      "grad_norm": 0.8279547691345215,
+      "learning_rate": 0.00021807779334123475,
+      "loss": 3.2635,
+      "step": 59200
+    },
+    {
+      "epoch": 6.377139166935744,
+      "grad_norm": 0.7678849697113037,
+      "learning_rate": 0.00021775455231117335,
+      "loss": 3.2879,
+      "step": 59250
+    },
+    {
+      "epoch": 6.382520718975353,
+      "grad_norm": 0.7861189842224121,
+      "learning_rate": 0.00021743131128111194,
+      "loss": 3.2774,
+      "step": 59300
+    },
+    {
+      "epoch": 6.387902271014961,
+      "grad_norm": 0.7797603607177734,
+      "learning_rate": 0.0002171080702510505,
+      "loss": 3.2787,
+      "step": 59350
+    },
+    {
+      "epoch": 6.393283823054569,
+      "grad_norm": 0.7491734027862549,
+      "learning_rate": 0.00021678482922098908,
+      "loss": 3.3011,
+      "step": 59400
+    },
+    {
+      "epoch": 6.398665375094177,
+      "grad_norm": 0.7296600341796875,
+      "learning_rate": 0.0002164615881909277,
+      "loss": 3.2786,
+      "step": 59450
+    },
+    {
+      "epoch": 6.404046927133785,
+      "grad_norm": 0.7515764236450195,
+      "learning_rate": 0.00021613834716086627,
+      "loss": 3.2813,
+      "step": 59500
+    },
+    {
+      "epoch": 6.4094284791733935,
+      "grad_norm": 0.7807029485702515,
+      "learning_rate": 0.00021581510613080483,
+      "loss": 3.2723,
+      "step": 59550
+    },
+    {
+      "epoch": 6.414810031213002,
+      "grad_norm": 0.8298711776733398,
+      "learning_rate": 0.00021549186510074346,
+      "loss": 3.3014,
+      "step": 59600
+    },
+    {
+      "epoch": 6.42019158325261,
+      "grad_norm": 0.7524072527885437,
+      "learning_rate": 0.00021516862407068202,
+      "loss": 3.2686,
+      "step": 59650
+    },
+    {
+      "epoch": 6.425573135292218,
+      "grad_norm": 0.7501651048660278,
+      "learning_rate": 0.0002148453830406206,
+      "loss": 3.2749,
+      "step": 59700
+    },
+    {
+      "epoch": 6.430954687331827,
+      "grad_norm": 0.8404172658920288,
+      "learning_rate": 0.00021452214201055919,
+      "loss": 3.2816,
+      "step": 59750
+    },
+    {
+      "epoch": 6.4363362393714345,
+      "grad_norm": 0.7685465812683105,
+      "learning_rate": 0.00021419890098049778,
+      "loss": 3.3003,
+      "step": 59800
+    },
+    {
+      "epoch": 6.441717791411043,
+      "grad_norm": 0.756592333316803,
+      "learning_rate": 0.00021387565995043638,
+      "loss": 3.2818,
+      "step": 59850
+    },
+    {
+      "epoch": 6.447099343450651,
+      "grad_norm": 0.7764129638671875,
+      "learning_rate": 0.00021355241892037494,
+      "loss": 3.3023,
+      "step": 59900
+    },
+    {
+      "epoch": 6.452480895490259,
+      "grad_norm": 0.7589853405952454,
+      "learning_rate": 0.0002132291778903135,
+      "loss": 3.2805,
+      "step": 59950
+    },
+    {
+      "epoch": 6.457862447529868,
+      "grad_norm": 0.7987534403800964,
+      "learning_rate": 0.00021290593686025213,
+      "loss": 3.2804,
+      "step": 60000
+    },
+    {
+      "epoch": 6.457862447529868,
+      "eval_accuracy": 0.38529639584246245,
+      "eval_loss": 3.374647855758667,
+      "eval_runtime": 185.834,
+      "eval_samples_per_second": 96.92,
+      "eval_steps_per_second": 6.059,
+      "step": 60000
+    },
+    {
+      "epoch": 6.4632439995694755,
+      "grad_norm": 0.8642520308494568,
+      "learning_rate": 0.0002125826958301907,
+      "loss": 3.2897,
+      "step": 60050
+    },
+    {
+      "epoch": 6.468625551609084,
+      "grad_norm": 0.873009443283081,
+      "learning_rate": 0.00021225945480012927,
+      "loss": 3.2711,
+      "step": 60100
+    },
+    {
+      "epoch": 6.474007103648693,
+      "grad_norm": 0.7259575128555298,
+      "learning_rate": 0.00021193621377006786,
+      "loss": 3.3043,
+      "step": 60150
+    },
+    {
+      "epoch": 6.4793886556883,
+      "grad_norm": 0.7962141633033752,
+      "learning_rate": 0.00021161943756060767,
+      "loss": 3.277,
+      "step": 60200
+    },
+    {
+      "epoch": 6.484770207727909,
+      "grad_norm": 0.7820010781288147,
+      "learning_rate": 0.00021129619653054626,
+      "loss": 3.2984,
+      "step": 60250
+    },
+    {
+      "epoch": 6.490151759767517,
+      "grad_norm": 0.7885971069335938,
+      "learning_rate": 0.00021097295550048483,
+      "loss": 3.2794,
+      "step": 60300
+    },
+    {
+      "epoch": 6.495533311807125,
+      "grad_norm": 0.7316437363624573,
+      "learning_rate": 0.00021064971447042343,
+      "loss": 3.2962,
+      "step": 60350
+    },
+    {
+      "epoch": 6.500914863846734,
+      "grad_norm": 0.8525658249855042,
+      "learning_rate": 0.00021032647344036202,
+      "loss": 3.2721,
+      "step": 60400
+    },
+    {
+      "epoch": 6.506296415886341,
+      "grad_norm": 0.7880814671516418,
+      "learning_rate": 0.0002100032324103006,
+      "loss": 3.3057,
+      "step": 60450
+    },
+    {
+      "epoch": 6.51167796792595,
+      "grad_norm": 0.778484582901001,
+      "learning_rate": 0.00020967999138023916,
+      "loss": 3.2887,
+      "step": 60500
+    },
+    {
+      "epoch": 6.517059519965558,
+      "grad_norm": 0.8094730973243713,
+      "learning_rate": 0.00020935675035017778,
+      "loss": 3.2875,
+      "step": 60550
+    },
+    {
+      "epoch": 6.522441072005166,
+      "grad_norm": 0.8072361350059509,
+      "learning_rate": 0.00020903350932011634,
+      "loss": 3.2965,
+      "step": 60600
+    },
+    {
+      "epoch": 6.5278226240447745,
+      "grad_norm": 0.7884482741355896,
+      "learning_rate": 0.0002087102682900549,
+      "loss": 3.276,
+      "step": 60650
+    },
+    {
+      "epoch": 6.533204176084383,
+      "grad_norm": 0.7414411902427673,
+      "learning_rate": 0.00020838702725999353,
+      "loss": 3.2788,
+      "step": 60700
+    },
+    {
+      "epoch": 6.538585728123991,
+      "grad_norm": 0.7233083844184875,
+      "learning_rate": 0.0002080637862299321,
+      "loss": 3.2865,
+      "step": 60750
+    },
+    {
+      "epoch": 6.543967280163599,
+      "grad_norm": 0.7610272765159607,
+      "learning_rate": 0.0002077405451998707,
+      "loss": 3.2619,
+      "step": 60800
+    },
+    {
+      "epoch": 6.549348832203208,
+      "grad_norm": 0.8353431820869446,
+      "learning_rate": 0.00020741730416980926,
+      "loss": 3.277,
+      "step": 60850
+    },
+    {
+      "epoch": 6.5547303842428155,
+      "grad_norm": 0.78453129529953,
+      "learning_rate": 0.00020709406313974786,
+      "loss": 3.2924,
+      "step": 60900
+    },
+    {
+      "epoch": 6.560111936282424,
+      "grad_norm": 0.7540761828422546,
+      "learning_rate": 0.00020677082210968645,
+      "loss": 3.2709,
+      "step": 60950
+    },
+    {
+      "epoch": 6.565493488322032,
+      "grad_norm": 0.8066650629043579,
+      "learning_rate": 0.00020644758107962502,
+      "loss": 3.2691,
+      "step": 61000
+    },
+    {
+      "epoch": 6.565493488322032,
+      "eval_accuracy": 0.38553358509656144,
+      "eval_loss": 3.3686017990112305,
+      "eval_runtime": 185.35,
+      "eval_samples_per_second": 97.173,
+      "eval_steps_per_second": 6.075,
+      "step": 61000
+    },
+    {
+      "epoch": 6.57087504036164,
+      "grad_norm": 0.8115589022636414,
+      "learning_rate": 0.0002061243400495636,
+      "loss": 3.2867,
+      "step": 61050
+    },
+    {
+      "epoch": 6.576256592401249,
+      "grad_norm": 0.7685365676879883,
+      "learning_rate": 0.0002058010990195022,
+      "loss": 3.2744,
+      "step": 61100
+    },
+    {
+      "epoch": 6.5816381444408565,
+      "grad_norm": 0.8063514828681946,
+      "learning_rate": 0.00020547785798944078,
+      "loss": 3.2983,
+      "step": 61150
+    },
+    {
+      "epoch": 6.587019696480465,
+      "grad_norm": 0.7861371636390686,
+      "learning_rate": 0.00020515461695937934,
+      "loss": 3.2795,
+      "step": 61200
+    },
+    {
+      "epoch": 6.592401248520073,
+      "grad_norm": 0.7573912739753723,
+      "learning_rate": 0.00020483137592931797,
+      "loss": 3.2956,
+      "step": 61250
+    },
+    {
+      "epoch": 6.597782800559681,
+      "grad_norm": 0.7656623125076294,
+      "learning_rate": 0.00020450813489925653,
+      "loss": 3.2842,
+      "step": 61300
+    },
+    {
+      "epoch": 6.60316435259929,
+      "grad_norm": 0.7915394306182861,
+      "learning_rate": 0.0002041848938691951,
+      "loss": 3.2873,
+      "step": 61350
+    },
+    {
+      "epoch": 6.608545904638898,
+      "grad_norm": 0.7657784223556519,
+      "learning_rate": 0.0002038616528391337,
+      "loss": 3.2952,
+      "step": 61400
+    },
+    {
+      "epoch": 6.613927456678506,
+      "grad_norm": 0.7530330419540405,
+      "learning_rate": 0.00020353841180907226,
+      "loss": 3.2698,
+      "step": 61450
+    },
+    {
+      "epoch": 6.619309008718115,
+      "grad_norm": 0.8081464171409607,
+      "learning_rate": 0.00020321517077901086,
+      "loss": 3.2847,
+      "step": 61500
+    },
+    {
+      "epoch": 6.624690560757722,
+      "grad_norm": 0.7701252102851868,
+      "learning_rate": 0.00020289192974894945,
+      "loss": 3.2788,
+      "step": 61550
+    },
+    {
+      "epoch": 6.630072112797331,
+      "grad_norm": 0.7802005410194397,
+      "learning_rate": 0.00020256868871888802,
+      "loss": 3.2673,
+      "step": 61600
+    },
+    {
+      "epoch": 6.635453664836939,
+      "grad_norm": 0.7802078127861023,
+      "learning_rate": 0.00020224544768882664,
+      "loss": 3.2762,
+      "step": 61650
+    },
+    {
+      "epoch": 6.640835216876547,
+      "grad_norm": 0.7452589273452759,
+      "learning_rate": 0.0002019222066587652,
+      "loss": 3.2844,
+      "step": 61700
+    },
+    {
+      "epoch": 6.6462167689161555,
+      "grad_norm": 0.8053356409072876,
+      "learning_rate": 0.00020159896562870378,
+      "loss": 3.2961,
+      "step": 61750
+    },
+    {
+      "epoch": 6.651598320955763,
+      "grad_norm": 0.7690718770027161,
+      "learning_rate": 0.00020127572459864237,
+      "loss": 3.2916,
+      "step": 61800
+    },
+    {
+      "epoch": 6.656979872995372,
+      "grad_norm": 0.8138352036476135,
+      "learning_rate": 0.00020095248356858097,
+      "loss": 3.2882,
+      "step": 61850
+    },
+    {
+      "epoch": 6.66236142503498,
+      "grad_norm": 0.7424994111061096,
+      "learning_rate": 0.00020062924253851953,
+      "loss": 3.2886,
+      "step": 61900
+    },
+    {
+      "epoch": 6.667742977074588,
+      "grad_norm": 0.7463759183883667,
+      "learning_rate": 0.00020030600150845813,
+      "loss": 3.2918,
+      "step": 61950
+    },
+    {
+      "epoch": 6.6731245291141965,
+      "grad_norm": 0.8001076579093933,
+      "learning_rate": 0.0001999827604783967,
+      "loss": 3.2979,
+      "step": 62000
+    },
+    {
+      "epoch": 6.6731245291141965,
+      "eval_accuracy": 0.3860651150237499,
+      "eval_loss": 3.3661580085754395,
+      "eval_runtime": 185.7923,
+      "eval_samples_per_second": 96.942,
+      "eval_steps_per_second": 6.061,
+      "step": 62000
+    },
+    {
+      "epoch": 6.678506081153805,
+      "grad_norm": 0.7533683776855469,
+      "learning_rate": 0.0001996595194483353,
+      "loss": 3.2703,
+      "step": 62050
+    },
+    {
+      "epoch": 6.683887633193413,
+      "grad_norm": 0.7832164168357849,
+      "learning_rate": 0.00019933627841827389,
+      "loss": 3.2881,
+      "step": 62100
+    },
+    {
+      "epoch": 6.689269185233021,
+      "grad_norm": 0.8090530633926392,
+      "learning_rate": 0.00019901303738821245,
+      "loss": 3.2888,
+      "step": 62150
+    },
+    {
+      "epoch": 6.69465073727263,
+      "grad_norm": 0.8002354502677917,
+      "learning_rate": 0.00019868979635815102,
+      "loss": 3.2732,
+      "step": 62200
+    },
+    {
+      "epoch": 6.7000322893122375,
+      "grad_norm": 0.7758427262306213,
+      "learning_rate": 0.00019837302014869085,
+      "loss": 3.2879,
+      "step": 62250
+    },
+    {
+      "epoch": 6.705413841351846,
+      "grad_norm": 0.770083487033844,
+      "learning_rate": 0.00019804977911862942,
+      "loss": 3.3033,
+      "step": 62300
+    },
+    {
+      "epoch": 6.710795393391454,
+      "grad_norm": 0.8458510041236877,
+      "learning_rate": 0.00019772653808856804,
+      "loss": 3.295,
+      "step": 62350
+    },
+    {
+      "epoch": 6.716176945431062,
+      "grad_norm": 0.775764524936676,
+      "learning_rate": 0.0001974032970585066,
+      "loss": 3.2913,
+      "step": 62400
+    },
+    {
+      "epoch": 6.721558497470671,
+      "grad_norm": 0.7734060287475586,
+      "learning_rate": 0.00019708005602844518,
+      "loss": 3.2845,
+      "step": 62450
+    },
+    {
+      "epoch": 6.7269400495102785,
+      "grad_norm": 0.8082960247993469,
+      "learning_rate": 0.00019675681499838377,
+      "loss": 3.2689,
+      "step": 62500
+    },
+    {
+      "epoch": 6.732321601549887,
+      "grad_norm": 0.8059088587760925,
+      "learning_rate": 0.00019643357396832237,
+      "loss": 3.2937,
+      "step": 62550
+    },
+    {
+      "epoch": 6.737703153589496,
+      "grad_norm": 0.7840763330459595,
+      "learning_rate": 0.00019611033293826096,
+      "loss": 3.2818,
+      "step": 62600
+    },
+    {
+      "epoch": 6.743084705629103,
+      "grad_norm": 0.8507702350616455,
+      "learning_rate": 0.00019578709190819953,
+      "loss": 3.2839,
+      "step": 62650
+    },
+    {
+      "epoch": 6.748466257668712,
+      "grad_norm": 0.8038814067840576,
+      "learning_rate": 0.0001954638508781381,
+      "loss": 3.2967,
+      "step": 62700
+    },
+    {
+      "epoch": 6.75384780970832,
+      "grad_norm": 0.8040089011192322,
+      "learning_rate": 0.00019514060984807672,
+      "loss": 3.2696,
+      "step": 62750
+    },
+    {
+      "epoch": 6.759229361747928,
+      "grad_norm": 0.7637272477149963,
+      "learning_rate": 0.0001948173688180153,
+      "loss": 3.2666,
+      "step": 62800
+    },
+    {
+      "epoch": 6.7646109137875365,
+      "grad_norm": 0.817797839641571,
+      "learning_rate": 0.00019449412778795386,
+      "loss": 3.2741,
+      "step": 62850
+    },
+    {
+      "epoch": 6.769992465827144,
+      "grad_norm": 0.770376443862915,
+      "learning_rate": 0.00019417088675789245,
+      "loss": 3.2783,
+      "step": 62900
+    },
+    {
+      "epoch": 6.775374017866753,
+      "grad_norm": 0.7574276328086853,
+      "learning_rate": 0.00019384764572783104,
+      "loss": 3.2922,
+      "step": 62950
+    },
+    {
+      "epoch": 6.780755569906361,
+      "grad_norm": 0.789776086807251,
+      "learning_rate": 0.0001935244046977696,
+      "loss": 3.2994,
+      "step": 63000
+    },
+    {
+      "epoch": 6.780755569906361,
+      "eval_accuracy": 0.3864262772228452,
+      "eval_loss": 3.362417221069336,
+      "eval_runtime": 185.7937,
+      "eval_samples_per_second": 96.941,
+      "eval_steps_per_second": 6.06,
+      "step": 63000
+    },
+    {
+      "epoch": 6.786137121945969,
+      "grad_norm": 0.8060975074768066,
+      "learning_rate": 0.0001932011636677082,
+      "loss": 3.2821,
+      "step": 63050
+    },
+    {
+      "epoch": 6.7915186739855775,
+      "grad_norm": 0.7560842633247375,
+      "learning_rate": 0.00019287792263764677,
+      "loss": 3.2596,
+      "step": 63100
+    },
+    {
+      "epoch": 6.796900226025185,
+      "grad_norm": 0.7956615090370178,
+      "learning_rate": 0.00019255468160758537,
+      "loss": 3.2651,
+      "step": 63150
+    },
+    {
+      "epoch": 6.802281778064794,
+      "grad_norm": 0.7850049138069153,
+      "learning_rate": 0.00019223144057752396,
+      "loss": 3.2838,
+      "step": 63200
+    },
+    {
+      "epoch": 6.807663330104402,
+      "grad_norm": 0.7724599242210388,
+      "learning_rate": 0.00019190819954746253,
+      "loss": 3.2904,
+      "step": 63250
+    },
+    {
+      "epoch": 6.813044882144011,
+      "grad_norm": 0.7807796001434326,
+      "learning_rate": 0.0001915849585174011,
+      "loss": 3.2708,
+      "step": 63300
+    },
+    {
+      "epoch": 6.8184264341836185,
+      "grad_norm": 0.7677724957466125,
+      "learning_rate": 0.00019126171748733972,
+      "loss": 3.2991,
+      "step": 63350
+    },
+    {
+      "epoch": 6.823807986223227,
+      "grad_norm": 0.8082730770111084,
+      "learning_rate": 0.0001909384764572783,
+      "loss": 3.292,
+      "step": 63400
+    },
+    {
+      "epoch": 6.829189538262835,
+      "grad_norm": 0.8102218508720398,
+      "learning_rate": 0.00019061523542721688,
+      "loss": 3.2824,
+      "step": 63450
+    },
+    {
+      "epoch": 6.834571090302443,
+      "grad_norm": 0.7808842658996582,
+      "learning_rate": 0.00019029199439715548,
+      "loss": 3.2874,
+      "step": 63500
+    },
+    {
+      "epoch": 6.839952642342052,
+      "grad_norm": 0.821823000907898,
+      "learning_rate": 0.00018996875336709404,
+      "loss": 3.2905,
+      "step": 63550
+    },
+    {
+      "epoch": 6.8453341943816595,
+      "grad_norm": 0.7801973223686218,
+      "learning_rate": 0.00018964551233703264,
+      "loss": 3.2922,
+      "step": 63600
+    },
+    {
+      "epoch": 6.850715746421268,
+      "grad_norm": 1.0729844570159912,
+      "learning_rate": 0.0001893222713069712,
+      "loss": 3.288,
+      "step": 63650
+    },
+    {
+      "epoch": 6.856097298460876,
+      "grad_norm": 0.8157760500907898,
+      "learning_rate": 0.0001889990302769098,
+      "loss": 3.3039,
+      "step": 63700
+    },
+    {
+      "epoch": 6.861478850500484,
+      "grad_norm": 0.7870354652404785,
+      "learning_rate": 0.0001886757892468484,
+      "loss": 3.2865,
+      "step": 63750
+    },
+    {
+      "epoch": 6.866860402540093,
+      "grad_norm": 0.7796850204467773,
+      "learning_rate": 0.00018835254821678696,
+      "loss": 3.2731,
+      "step": 63800
+    },
+    {
+      "epoch": 6.8722419545797,
+      "grad_norm": 0.7909936904907227,
+      "learning_rate": 0.00018802930718672553,
+      "loss": 3.2906,
+      "step": 63850
+    },
+    {
+      "epoch": 6.877623506619309,
+      "grad_norm": 0.8563127517700195,
+      "learning_rate": 0.00018770606615666415,
+      "loss": 3.29,
+      "step": 63900
+    },
+    {
+      "epoch": 6.8830050586589175,
+      "grad_norm": 0.7505841255187988,
+      "learning_rate": 0.00018738282512660272,
+      "loss": 3.3153,
+      "step": 63950
+    },
+    {
+      "epoch": 6.888386610698525,
+      "grad_norm": 0.801104724407196,
+      "learning_rate": 0.0001870595840965413,
+      "loss": 3.2885,
+      "step": 64000
+    },
+    {
+      "epoch": 6.888386610698525,
+      "eval_accuracy": 0.38679113362012624,
+      "eval_loss": 3.3589389324188232,
+      "eval_runtime": 185.6402,
+      "eval_samples_per_second": 97.021,
+      "eval_steps_per_second": 6.065,
+      "step": 64000
+    },
+    {
+      "epoch": 6.893768162738134,
+      "grad_norm": 0.813956618309021,
+      "learning_rate": 0.0001867363430664799,
+      "loss": 3.2965,
+      "step": 64050
+    },
+    {
+      "epoch": 6.899149714777742,
+      "grad_norm": 0.8185772895812988,
+      "learning_rate": 0.00018641310203641848,
+      "loss": 3.293,
+      "step": 64100
+    },
+    {
+      "epoch": 6.90453126681735,
+      "grad_norm": 0.8417222499847412,
+      "learning_rate": 0.00018608986100635705,
+      "loss": 3.2929,
+      "step": 64150
+    },
+    {
+      "epoch": 6.9099128188569585,
+      "grad_norm": 0.7140278816223145,
+      "learning_rate": 0.00018576661997629564,
+      "loss": 3.297,
+      "step": 64200
+    },
+    {
+      "epoch": 6.915294370896566,
+      "grad_norm": 0.7850991487503052,
+      "learning_rate": 0.00018544337894623423,
+      "loss": 3.2798,
+      "step": 64250
+    },
+    {
+      "epoch": 6.920675922936175,
+      "grad_norm": 0.7667518854141235,
+      "learning_rate": 0.00018512660273677404,
+      "loss": 3.2776,
+      "step": 64300
+    },
+    {
+      "epoch": 6.926057474975783,
+      "grad_norm": 0.7750104665756226,
+      "learning_rate": 0.0001848033617067126,
+      "loss": 3.2649,
+      "step": 64350
+    },
+    {
+      "epoch": 6.931439027015391,
+      "grad_norm": 0.8614099621772766,
+      "learning_rate": 0.00018448012067665123,
+      "loss": 3.2924,
+      "step": 64400
+    },
+    {
+      "epoch": 6.9368205790549995,
+      "grad_norm": 0.7528159022331238,
+      "learning_rate": 0.0001841568796465898,
+      "loss": 3.3094,
+      "step": 64450
+    },
+    {
+      "epoch": 6.942202131094608,
+      "grad_norm": 0.7764745950698853,
+      "learning_rate": 0.00018383363861652837,
+      "loss": 3.2851,
+      "step": 64500
+    },
+    {
+      "epoch": 6.947583683134216,
+      "grad_norm": 0.778229832649231,
+      "learning_rate": 0.00018351039758646696,
+      "loss": 3.2955,
+      "step": 64550
+    },
+    {
+      "epoch": 6.952965235173824,
+      "grad_norm": 0.8336729407310486,
+      "learning_rate": 0.00018318715655640555,
+      "loss": 3.2736,
+      "step": 64600
+    },
+    {
+      "epoch": 6.958346787213433,
+      "grad_norm": 0.853729784488678,
+      "learning_rate": 0.00018286391552634412,
+      "loss": 3.2976,
+      "step": 64650
+    },
+    {
+      "epoch": 6.9637283392530405,
+      "grad_norm": 0.7838858366012573,
+      "learning_rate": 0.00018254067449628272,
+      "loss": 3.2957,
+      "step": 64700
+    },
+    {
+      "epoch": 6.969109891292649,
+      "grad_norm": 0.7707374691963196,
+      "learning_rate": 0.00018221743346622128,
+      "loss": 3.2768,
+      "step": 64750
+    },
+    {
+      "epoch": 6.974491443332257,
+      "grad_norm": 0.8833007216453552,
+      "learning_rate": 0.00018189419243615988,
+      "loss": 3.2889,
+      "step": 64800
+    },
+    {
+      "epoch": 6.979872995371865,
+      "grad_norm": 0.7651616334915161,
+      "learning_rate": 0.00018157095140609847,
+      "loss": 3.2859,
+      "step": 64850
+    },
+    {
+      "epoch": 6.985254547411474,
+      "grad_norm": 0.8112615346908569,
+      "learning_rate": 0.00018124771037603704,
+      "loss": 3.2847,
+      "step": 64900
+    },
+    {
+      "epoch": 6.990636099451081,
+      "grad_norm": 0.8491316437721252,
+      "learning_rate": 0.0001809244693459756,
+      "loss": 3.2857,
+      "step": 64950
+    },
+    {
+      "epoch": 6.99601765149069,
+      "grad_norm": 0.8059223890304565,
+      "learning_rate": 0.00018060122831591423,
+      "loss": 3.2941,
+      "step": 65000
+    },
+    {
+      "epoch": 6.99601765149069,
+      "eval_accuracy": 0.3872998464408737,
+      "eval_loss": 3.3537228107452393,
+      "eval_runtime": 185.6093,
+      "eval_samples_per_second": 97.037,
+      "eval_steps_per_second": 6.067,
+      "step": 65000
+    },
+    {
+      "epoch": 7.0013992035302985,
+      "grad_norm": 0.8688104748725891,
+      "learning_rate": 0.0001802779872858528,
+      "loss": 3.2449,
+      "step": 65050
+    },
+    {
+      "epoch": 7.006780755569906,
+      "grad_norm": 0.8001312613487244,
+      "learning_rate": 0.00017995474625579137,
+      "loss": 3.2098,
+      "step": 65100
+    },
+    {
+      "epoch": 7.012162307609515,
+      "grad_norm": 0.8134597539901733,
+      "learning_rate": 0.00017963150522573,
+      "loss": 3.2169,
+      "step": 65150
+    },
+    {
+      "epoch": 7.017543859649122,
+      "grad_norm": 0.8403487205505371,
+      "learning_rate": 0.00017930826419566856,
+      "loss": 3.2073,
+      "step": 65200
+    },
+    {
+      "epoch": 7.022925411688731,
+      "grad_norm": 0.8222055435180664,
+      "learning_rate": 0.00017898502316560712,
+      "loss": 3.2029,
+      "step": 65250
+    },
+    {
+      "epoch": 7.0283069637283395,
+      "grad_norm": 0.7753844857215881,
+      "learning_rate": 0.00017866178213554572,
+      "loss": 3.2085,
+      "step": 65300
+    },
+    {
+      "epoch": 7.033688515767947,
+      "grad_norm": 0.742112934589386,
+      "learning_rate": 0.0001783385411054843,
+      "loss": 3.2248,
+      "step": 65350
+    },
+    {
+      "epoch": 7.039070067807556,
+      "grad_norm": 0.8007506132125854,
+      "learning_rate": 0.0001780153000754229,
+      "loss": 3.1953,
+      "step": 65400
+    },
+    {
+      "epoch": 7.044451619847164,
+      "grad_norm": 0.7830198407173157,
+      "learning_rate": 0.00017769205904536147,
+      "loss": 3.2213,
+      "step": 65450
+    },
+    {
+      "epoch": 7.049833171886772,
+      "grad_norm": 0.8625771403312683,
+      "learning_rate": 0.00017736881801530004,
+      "loss": 3.198,
+      "step": 65500
+    },
+    {
+      "epoch": 7.0552147239263805,
+      "grad_norm": 0.8744086027145386,
+      "learning_rate": 0.00017704557698523866,
+      "loss": 3.2183,
+      "step": 65550
+    },
+    {
+      "epoch": 7.060596275965988,
+      "grad_norm": 0.8134574890136719,
+      "learning_rate": 0.00017672233595517723,
+      "loss": 3.22,
+      "step": 65600
+    },
+    {
+      "epoch": 7.065977828005597,
+      "grad_norm": 0.8503029942512512,
+      "learning_rate": 0.0001763990949251158,
+      "loss": 3.2134,
+      "step": 65650
+    },
+    {
+      "epoch": 7.071359380045205,
+      "grad_norm": 0.8678685426712036,
+      "learning_rate": 0.00017607585389505442,
+      "loss": 3.2003,
+      "step": 65700
+    },
+    {
+      "epoch": 7.076740932084813,
+      "grad_norm": 0.7783809900283813,
+      "learning_rate": 0.000175752612864993,
+      "loss": 3.2171,
+      "step": 65750
+    },
+    {
+      "epoch": 7.0821224841244215,
+      "grad_norm": 0.7878561615943909,
+      "learning_rate": 0.00017542937183493156,
+      "loss": 3.201,
+      "step": 65800
+    },
+    {
+      "epoch": 7.08750403616403,
+      "grad_norm": 0.7991673946380615,
+      "learning_rate": 0.00017510613080487015,
+      "loss": 3.2115,
+      "step": 65850
+    },
+    {
+      "epoch": 7.092885588203638,
+      "grad_norm": 0.7442201375961304,
+      "learning_rate": 0.00017478288977480874,
+      "loss": 3.2065,
+      "step": 65900
+    },
+    {
+      "epoch": 7.098267140243246,
+      "grad_norm": 0.8097749352455139,
+      "learning_rate": 0.0001744596487447473,
+      "loss": 3.2114,
+      "step": 65950
+    },
+    {
+      "epoch": 7.103648692282855,
+      "grad_norm": 0.8379082679748535,
+      "learning_rate": 0.0001741364077146859,
+      "loss": 3.2105,
+      "step": 66000
+    },
+    {
+      "epoch": 7.103648692282855,
+      "eval_accuracy": 0.3868561080470393,
+      "eval_loss": 3.363415002822876,
+      "eval_runtime": 185.6189,
+      "eval_samples_per_second": 97.032,
+      "eval_steps_per_second": 6.066,
+      "step": 66000
+    },
+    {
+      "epoch": 7.109030244322462,
+      "grad_norm": 0.866179347038269,
+      "learning_rate": 0.00017381316668462447,
+      "loss": 3.2038,
+      "step": 66050
+    },
+    {
+      "epoch": 7.114411796362071,
+      "grad_norm": 0.7942879796028137,
+      "learning_rate": 0.00017348992565456304,
+      "loss": 3.2041,
+      "step": 66100
+    },
+    {
+      "epoch": 7.119793348401679,
+      "grad_norm": 0.7950063347816467,
+      "learning_rate": 0.00017316668462450166,
+      "loss": 3.2214,
+      "step": 66150
+    },
+    {
+      "epoch": 7.125174900441287,
+      "grad_norm": 0.8257644176483154,
+      "learning_rate": 0.00017284344359444023,
+      "loss": 3.2397,
+      "step": 66200
+    },
+    {
+      "epoch": 7.130556452480896,
+      "grad_norm": 0.8185299634933472,
+      "learning_rate": 0.00017252020256437885,
+      "loss": 3.2046,
+      "step": 66250
+    },
+    {
+      "epoch": 7.135938004520503,
+      "grad_norm": 0.8815109133720398,
+      "learning_rate": 0.00017219696153431742,
+      "loss": 3.2194,
+      "step": 66300
+    },
+    {
+      "epoch": 7.141319556560112,
+      "grad_norm": 0.8232885599136353,
+      "learning_rate": 0.00017188018532485723,
+      "loss": 3.222,
+      "step": 66350
+    },
+    {
+      "epoch": 7.1467011085997205,
+      "grad_norm": 0.8150448799133301,
+      "learning_rate": 0.0001715569442947958,
+      "loss": 3.212,
+      "step": 66400
+    },
+    {
+      "epoch": 7.152082660639328,
+      "grad_norm": 0.8206075429916382,
+      "learning_rate": 0.0001712337032647344,
+      "loss": 3.2137,
+      "step": 66450
+    },
+    {
+      "epoch": 7.157464212678937,
+      "grad_norm": 0.8803149461746216,
+      "learning_rate": 0.00017091046223467298,
+      "loss": 3.2161,
+      "step": 66500
+    },
+    {
+      "epoch": 7.162845764718545,
+      "grad_norm": 0.8381869792938232,
+      "learning_rate": 0.00017058722120461155,
+      "loss": 3.2219,
+      "step": 66550
+    },
+    {
+      "epoch": 7.168227316758153,
+      "grad_norm": 0.8405774831771851,
+      "learning_rate": 0.00017026398017455012,
+      "loss": 3.2131,
+      "step": 66600
+    },
+    {
+      "epoch": 7.1736088687977615,
+      "grad_norm": 0.800108790397644,
+      "learning_rate": 0.00016994073914448874,
+      "loss": 3.2177,
+      "step": 66650
+    },
+    {
+      "epoch": 7.178990420837369,
+      "grad_norm": 0.7671277523040771,
+      "learning_rate": 0.0001696174981144273,
+      "loss": 3.2221,
+      "step": 66700
+    },
+    {
+      "epoch": 7.184371972876978,
+      "grad_norm": 0.8463557362556458,
+      "learning_rate": 0.00016929425708436588,
+      "loss": 3.2389,
+      "step": 66750
+    },
+    {
+      "epoch": 7.189753524916586,
+      "grad_norm": 0.8206205368041992,
+      "learning_rate": 0.0001689710160543045,
+      "loss": 3.2278,
+      "step": 66800
+    },
+    {
+      "epoch": 7.195135076956194,
+      "grad_norm": 0.8687519431114197,
+      "learning_rate": 0.00016864777502424307,
+      "loss": 3.2203,
+      "step": 66850
+    },
+    {
+      "epoch": 7.2005166289958025,
+      "grad_norm": 0.840397834777832,
+      "learning_rate": 0.00016832453399418163,
+      "loss": 3.1994,
+      "step": 66900
+    },
+    {
+      "epoch": 7.205898181035411,
+      "grad_norm": 0.8231752514839172,
+      "learning_rate": 0.00016800129296412023,
+      "loss": 3.2327,
+      "step": 66950
+    },
+    {
+      "epoch": 7.211279733075019,
+      "grad_norm": 0.8082619309425354,
+      "learning_rate": 0.00016767805193405882,
+      "loss": 3.2258,
+      "step": 67000
+    },
+    {
+      "epoch": 7.211279733075019,
+      "eval_accuracy": 0.38720238480050406,
+      "eval_loss": 3.359046459197998,
+      "eval_runtime": 185.3339,
+      "eval_samples_per_second": 97.181,
+      "eval_steps_per_second": 6.076,
+      "step": 67000
+    },
+    {
+      "epoch": 7.216661285114627,
+      "grad_norm": 0.879897952079773,
+      "learning_rate": 0.0001673548109039974,
+      "loss": 3.2243,
+      "step": 67050
+    },
+    {
+      "epoch": 7.222042837154235,
+      "grad_norm": 0.8425320982933044,
+      "learning_rate": 0.00016703156987393598,
+      "loss": 3.2285,
+      "step": 67100
+    },
+    {
+      "epoch": 7.2274243891938434,
+      "grad_norm": 0.8390564322471619,
+      "learning_rate": 0.00016670832884387455,
+      "loss": 3.2255,
+      "step": 67150
+    },
+    {
+      "epoch": 7.232805941233452,
+      "grad_norm": 0.7988588809967041,
+      "learning_rate": 0.00016638508781381317,
+      "loss": 3.2167,
+      "step": 67200
+    },
+    {
+      "epoch": 7.23818749327306,
+      "grad_norm": 0.8039420247077942,
+      "learning_rate": 0.00016606184678375174,
+      "loss": 3.2235,
+      "step": 67250
+    },
+    {
+      "epoch": 7.243569045312668,
+      "grad_norm": 0.818771243095398,
+      "learning_rate": 0.0001657386057536903,
+      "loss": 3.2446,
+      "step": 67300
+    },
+    {
+      "epoch": 7.248950597352277,
+      "grad_norm": 0.8249024748802185,
+      "learning_rate": 0.00016541536472362893,
+      "loss": 3.2306,
+      "step": 67350
+    },
+    {
+      "epoch": 7.254332149391884,
+      "grad_norm": 0.8478593230247498,
+      "learning_rate": 0.0001650921236935675,
+      "loss": 3.228,
+      "step": 67400
+    },
+    {
+      "epoch": 7.259713701431493,
+      "grad_norm": 0.8184794187545776,
+      "learning_rate": 0.00016476888266350607,
+      "loss": 3.2159,
+      "step": 67450
+    },
+    {
+      "epoch": 7.265095253471101,
+      "grad_norm": 0.7827985882759094,
+      "learning_rate": 0.00016444564163344466,
+      "loss": 3.2198,
+      "step": 67500
+    },
+    {
+      "epoch": 7.270476805510709,
+      "grad_norm": 0.8600140810012817,
+      "learning_rate": 0.00016412240060338326,
+      "loss": 3.2158,
+      "step": 67550
+    },
+    {
+      "epoch": 7.275858357550318,
+      "grad_norm": 0.8089606761932373,
+      "learning_rate": 0.00016379915957332182,
+      "loss": 3.2221,
+      "step": 67600
+    },
+    {
+      "epoch": 7.281239909589925,
+      "grad_norm": 0.8981764316558838,
+      "learning_rate": 0.00016347591854326042,
+      "loss": 3.2302,
+      "step": 67650
+    },
+    {
+      "epoch": 7.286621461629534,
+      "grad_norm": 0.7918499708175659,
+      "learning_rate": 0.00016315267751319898,
+      "loss": 3.2141,
+      "step": 67700
+    },
+    {
+      "epoch": 7.2920030136691425,
+      "grad_norm": 0.8603319525718689,
+      "learning_rate": 0.00016282943648313755,
+      "loss": 3.2435,
+      "step": 67750
+    },
+    {
+      "epoch": 7.29738456570875,
+      "grad_norm": 0.8360230326652527,
+      "learning_rate": 0.00016250619545307617,
+      "loss": 3.2135,
+      "step": 67800
+    },
+    {
+      "epoch": 7.302766117748359,
+      "grad_norm": 0.8522823452949524,
+      "learning_rate": 0.00016218295442301474,
+      "loss": 3.23,
+      "step": 67850
+    },
+    {
+      "epoch": 7.308147669787967,
+      "grad_norm": 0.844285786151886,
+      "learning_rate": 0.0001618597133929533,
+      "loss": 3.2286,
+      "step": 67900
+    },
+    {
+      "epoch": 7.313529221827575,
+      "grad_norm": 0.8203629851341248,
+      "learning_rate": 0.00016153647236289193,
+      "loss": 3.2121,
+      "step": 67950
+    },
+    {
+      "epoch": 7.3189107738671835,
+      "grad_norm": 0.7987516522407532,
+      "learning_rate": 0.0001612132313328305,
+      "loss": 3.2207,
+      "step": 68000
+    },
+    {
+      "epoch": 7.3189107738671835,
+      "eval_accuracy": 0.38765394620226107,
+      "eval_loss": 3.3566348552703857,
+      "eval_runtime": 185.6747,
+      "eval_samples_per_second": 97.003,
+      "eval_steps_per_second": 6.064,
+      "step": 68000
+    },
+    {
+      "epoch": 7.324292325906791,
+      "grad_norm": 0.8273880481719971,
+      "learning_rate": 0.0001608899903027691,
+      "loss": 3.2247,
+      "step": 68050
+    },
+    {
+      "epoch": 7.3296738779464,
+      "grad_norm": 0.8238916993141174,
+      "learning_rate": 0.00016056674927270766,
+      "loss": 3.2411,
+      "step": 68100
+    },
+    {
+      "epoch": 7.335055429986008,
+      "grad_norm": 0.8411089777946472,
+      "learning_rate": 0.00016024350824264626,
+      "loss": 3.2264,
+      "step": 68150
+    },
+    {
+      "epoch": 7.340436982025616,
+      "grad_norm": 0.8254903554916382,
+      "learning_rate": 0.00015992026721258485,
+      "loss": 3.2129,
+      "step": 68200
+    },
+    {
+      "epoch": 7.3458185340652244,
+      "grad_norm": 0.8240244388580322,
+      "learning_rate": 0.00015959702618252342,
+      "loss": 3.2247,
+      "step": 68250
+    },
+    {
+      "epoch": 7.351200086104833,
+      "grad_norm": 0.8349695801734924,
+      "learning_rate": 0.00015927378515246199,
+      "loss": 3.2313,
+      "step": 68300
+    },
+    {
+      "epoch": 7.356581638144441,
+      "grad_norm": 0.8326880931854248,
+      "learning_rate": 0.0001589505441224006,
+      "loss": 3.2266,
+      "step": 68350
+    },
+    {
+      "epoch": 7.361963190184049,
+      "grad_norm": 0.7652195692062378,
+      "learning_rate": 0.00015862730309233917,
+      "loss": 3.219,
+      "step": 68400
+    },
+    {
+      "epoch": 7.367344742223658,
+      "grad_norm": 0.8464148044586182,
+      "learning_rate": 0.00015830406206227774,
+      "loss": 3.2397,
+      "step": 68450
+    },
+    {
+      "epoch": 7.372726294263265,
+      "grad_norm": 0.8529747128486633,
+      "learning_rate": 0.00015798082103221636,
+      "loss": 3.2218,
+      "step": 68500
+    },
+    {
+      "epoch": 7.378107846302874,
+      "grad_norm": 0.8056176900863647,
+      "learning_rate": 0.00015765758000215493,
+      "loss": 3.2313,
+      "step": 68550
+    },
+    {
+      "epoch": 7.383489398342482,
+      "grad_norm": 0.870385468006134,
+      "learning_rate": 0.00015734080379269474,
+      "loss": 3.2297,
+      "step": 68600
+    },
+    {
+      "epoch": 7.38887095038209,
+      "grad_norm": 0.7906653881072998,
+      "learning_rate": 0.00015701756276263333,
+      "loss": 3.2335,
+      "step": 68650
+    },
+    {
+      "epoch": 7.394252502421699,
+      "grad_norm": 0.8384215831756592,
+      "learning_rate": 0.0001566943217325719,
+      "loss": 3.2413,
+      "step": 68700
+    },
+    {
+      "epoch": 7.399634054461306,
+      "grad_norm": 0.7999741435050964,
+      "learning_rate": 0.0001563710807025105,
+      "loss": 3.2276,
+      "step": 68750
+    },
+    {
+      "epoch": 7.405015606500915,
+      "grad_norm": 0.7932824492454529,
+      "learning_rate": 0.00015604783967244906,
+      "loss": 3.2345,
+      "step": 68800
+    },
+    {
+      "epoch": 7.4103971585405235,
+      "grad_norm": 0.7963410019874573,
+      "learning_rate": 0.00015572459864238763,
+      "loss": 3.2353,
+      "step": 68850
+    },
+    {
+      "epoch": 7.415778710580131,
+      "grad_norm": 0.8788455128669739,
+      "learning_rate": 0.00015540135761232625,
+      "loss": 3.2389,
+      "step": 68900
+    },
+    {
+      "epoch": 7.42116026261974,
+      "grad_norm": 0.8035975098609924,
+      "learning_rate": 0.00015507811658226482,
+      "loss": 3.2232,
+      "step": 68950
+    },
+    {
+      "epoch": 7.426541814659347,
+      "grad_norm": 0.8268063068389893,
+      "learning_rate": 0.00015475487555220344,
+      "loss": 3.229,
+      "step": 69000
+    },
+    {
+      "epoch": 7.426541814659347,
+      "eval_accuracy": 0.38789928442294613,
+      "eval_loss": 3.3524458408355713,
+      "eval_runtime": 185.8971,
+      "eval_samples_per_second": 96.887,
+      "eval_steps_per_second": 6.057,
+      "step": 69000
+    },
+    {
+      "epoch": 7.431923366698956,
+      "grad_norm": 0.8097623586654663,
+      "learning_rate": 0.000154431634522142,
+      "loss": 3.2243,
+      "step": 69050
+    },
+    {
+      "epoch": 7.4373049187385645,
+      "grad_norm": 0.8671795129776001,
+      "learning_rate": 0.00015410839349208058,
+      "loss": 3.2357,
+      "step": 69100
+    },
+    {
+      "epoch": 7.442686470778172,
+      "grad_norm": 0.8262737989425659,
+      "learning_rate": 0.00015378515246201917,
+      "loss": 3.2358,
+      "step": 69150
+    },
+    {
+      "epoch": 7.448068022817781,
+      "grad_norm": 0.913790762424469,
+      "learning_rate": 0.00015346191143195774,
+      "loss": 3.2301,
+      "step": 69200
+    },
+    {
+      "epoch": 7.453449574857389,
+      "grad_norm": 0.8186403512954712,
+      "learning_rate": 0.00015313867040189633,
+      "loss": 3.2385,
+      "step": 69250
+    },
+    {
+      "epoch": 7.458831126896997,
+      "grad_norm": 0.7833678126335144,
+      "learning_rate": 0.00015281542937183493,
+      "loss": 3.2463,
+      "step": 69300
+    },
+    {
+      "epoch": 7.4642126789366054,
+      "grad_norm": 0.8381041884422302,
+      "learning_rate": 0.0001524921883417735,
+      "loss": 3.2222,
+      "step": 69350
+    },
+    {
+      "epoch": 7.469594230976213,
+      "grad_norm": 0.88124018907547,
+      "learning_rate": 0.00015216894731171206,
+      "loss": 3.2236,
+      "step": 69400
+    },
+    {
+      "epoch": 7.474975783015822,
+      "grad_norm": 0.8232921361923218,
+      "learning_rate": 0.00015184570628165068,
+      "loss": 3.2535,
+      "step": 69450
+    },
+    {
+      "epoch": 7.48035733505543,
+      "grad_norm": 0.8925224542617798,
+      "learning_rate": 0.00015152246525158925,
+      "loss": 3.2339,
+      "step": 69500
+    },
+    {
+      "epoch": 7.485738887095038,
+      "grad_norm": 0.8126924633979797,
+      "learning_rate": 0.00015119922422152782,
+      "loss": 3.2205,
+      "step": 69550
+    },
+    {
+      "epoch": 7.491120439134646,
+      "grad_norm": 0.83549565076828,
+      "learning_rate": 0.00015087598319146644,
+      "loss": 3.2345,
+      "step": 69600
+    },
+    {
+      "epoch": 7.496501991174255,
+      "grad_norm": 0.855729341506958,
+      "learning_rate": 0.000150552742161405,
+      "loss": 3.2305,
+      "step": 69650
+    },
+    {
+      "epoch": 7.501883543213863,
+      "grad_norm": 0.849579930305481,
+      "learning_rate": 0.00015022950113134358,
+      "loss": 3.2176,
+      "step": 69700
+    },
+    {
+      "epoch": 7.507265095253471,
+      "grad_norm": 0.8337715268135071,
+      "learning_rate": 0.00014990626010128217,
+      "loss": 3.2484,
+      "step": 69750
+    },
+    {
+      "epoch": 7.51264664729308,
+      "grad_norm": 0.8676668405532837,
+      "learning_rate": 0.00014958301907122077,
+      "loss": 3.2371,
+      "step": 69800
+    },
+    {
+      "epoch": 7.518028199332687,
+      "grad_norm": 0.8204949498176575,
+      "learning_rate": 0.00014925977804115933,
+      "loss": 3.2365,
+      "step": 69850
+    },
+    {
+      "epoch": 7.523409751372296,
+      "grad_norm": 0.8827537894248962,
+      "learning_rate": 0.00014893653701109793,
+      "loss": 3.2383,
+      "step": 69900
+    },
+    {
+      "epoch": 7.528791303411904,
+      "grad_norm": 0.8607305288314819,
+      "learning_rate": 0.0001486132959810365,
+      "loss": 3.2327,
+      "step": 69950
+    },
+    {
+      "epoch": 7.534172855451512,
+      "grad_norm": 0.812737226486206,
+      "learning_rate": 0.0001482900549509751,
+      "loss": 3.2185,
+      "step": 70000
+    },
+    {
+      "epoch": 7.534172855451512,
+      "eval_accuracy": 0.38844483057266266,
+      "eval_loss": 3.349090099334717,
+      "eval_runtime": 185.3781,
+      "eval_samples_per_second": 97.158,
+      "eval_steps_per_second": 6.074,
+      "step": 70000
+    },
+    {
+      "epoch": 7.539554407491121,
+      "grad_norm": 0.8333404660224915,
+      "learning_rate": 0.00014796681392091368,
+      "loss": 3.2335,
+      "step": 70050
+    },
+    {
+      "epoch": 7.544935959530728,
+      "grad_norm": 0.840907633304596,
+      "learning_rate": 0.00014764357289085228,
+      "loss": 3.2342,
+      "step": 70100
+    },
+    {
+      "epoch": 7.550317511570337,
+      "grad_norm": 0.8278952836990356,
+      "learning_rate": 0.00014732033186079085,
+      "loss": 3.2332,
+      "step": 70150
+    },
+    {
+      "epoch": 7.5556990636099455,
+      "grad_norm": 0.8081293106079102,
+      "learning_rate": 0.00014699709083072944,
+      "loss": 3.2356,
+      "step": 70200
+    },
+    {
+      "epoch": 7.561080615649553,
+      "grad_norm": 0.8069691061973572,
+      "learning_rate": 0.00014667384980066804,
+      "loss": 3.2448,
+      "step": 70250
+    },
+    {
+      "epoch": 7.566462167689162,
+      "grad_norm": 0.8184692859649658,
+      "learning_rate": 0.0001463506087706066,
+      "loss": 3.2432,
+      "step": 70300
+    },
+    {
+      "epoch": 7.57184371972877,
+      "grad_norm": 0.8225573897361755,
+      "learning_rate": 0.0001460273677405452,
+      "loss": 3.252,
+      "step": 70350
+    },
+    {
+      "epoch": 7.577225271768378,
+      "grad_norm": 0.8349873423576355,
+      "learning_rate": 0.00014570412671048377,
+      "loss": 3.2395,
+      "step": 70400
+    },
+    {
+      "epoch": 7.5826068238079865,
+      "grad_norm": 0.8561345934867859,
+      "learning_rate": 0.00014538088568042236,
+      "loss": 3.2493,
+      "step": 70450
+    },
+    {
+      "epoch": 7.587988375847594,
+      "grad_norm": 0.7891167998313904,
+      "learning_rate": 0.00014505764465036093,
+      "loss": 3.2417,
+      "step": 70500
+    },
+    {
+      "epoch": 7.593369927887203,
+      "grad_norm": 0.8202583193778992,
+      "learning_rate": 0.00014473440362029952,
+      "loss": 3.2534,
+      "step": 70550
+    },
+    {
+      "epoch": 7.598751479926811,
+      "grad_norm": 0.8512953519821167,
+      "learning_rate": 0.00014441116259023812,
+      "loss": 3.215,
+      "step": 70600
+    },
+    {
+      "epoch": 7.604133031966419,
+      "grad_norm": 0.8122953176498413,
+      "learning_rate": 0.00014409438638077792,
+      "loss": 3.2393,
+      "step": 70650
+    },
+    {
+      "epoch": 7.609514584006027,
+      "grad_norm": 0.8483431935310364,
+      "learning_rate": 0.00014377114535071652,
+      "loss": 3.2375,
+      "step": 70700
+    },
+    {
+      "epoch": 7.614896136045635,
+      "grad_norm": 0.8433930277824402,
+      "learning_rate": 0.0001434479043206551,
+      "loss": 3.2332,
+      "step": 70750
+    },
+    {
+      "epoch": 7.620277688085244,
+      "grad_norm": 0.8184301257133484,
+      "learning_rate": 0.00014312466329059368,
+      "loss": 3.2189,
+      "step": 70800
+    },
+    {
+      "epoch": 7.625659240124852,
+      "grad_norm": 0.884261429309845,
+      "learning_rate": 0.00014280142226053225,
+      "loss": 3.2276,
+      "step": 70850
+    },
+    {
+      "epoch": 7.63104079216446,
+      "grad_norm": 0.8433180451393127,
+      "learning_rate": 0.00014247818123047084,
+      "loss": 3.2411,
+      "step": 70900
+    },
+    {
+      "epoch": 7.636422344204068,
+      "grad_norm": 0.8581477403640747,
+      "learning_rate": 0.0001421549402004094,
+      "loss": 3.2187,
+      "step": 70950
+    },
+    {
+      "epoch": 7.641803896243677,
+      "grad_norm": 0.8302317261695862,
+      "learning_rate": 0.000141831699170348,
+      "loss": 3.2483,
+      "step": 71000
+    },
+    {
+      "epoch": 7.641803896243677,
+      "eval_accuracy": 0.3884902474797691,
+      "eval_loss": 3.346506357192993,
+      "eval_runtime": 185.5331,
+      "eval_samples_per_second": 97.077,
+      "eval_steps_per_second": 6.069,
+      "step": 71000
+    },
+    {
+      "epoch": 7.647185448283285,
+      "grad_norm": 0.8029564619064331,
+      "learning_rate": 0.0001415084581402866,
+      "loss": 3.2419,
+      "step": 71050
+    },
+    {
+      "epoch": 7.652567000322893,
+      "grad_norm": 0.9303278923034668,
+      "learning_rate": 0.00014118521711022517,
+      "loss": 3.2487,
+      "step": 71100
+    },
+    {
+      "epoch": 7.657948552362502,
+      "grad_norm": 0.8449538350105286,
+      "learning_rate": 0.00014086197608016376,
+      "loss": 3.2347,
+      "step": 71150
+    },
+    {
+      "epoch": 7.663330104402109,
+      "grad_norm": 0.8532068133354187,
+      "learning_rate": 0.00014053873505010236,
+      "loss": 3.2476,
+      "step": 71200
+    },
+    {
+      "epoch": 7.668711656441718,
+      "grad_norm": 0.8297402858734131,
+      "learning_rate": 0.00014021549402004092,
+      "loss": 3.2303,
+      "step": 71250
+    },
+    {
+      "epoch": 7.674093208481326,
+      "grad_norm": 0.8572767376899719,
+      "learning_rate": 0.00013989225298997952,
+      "loss": 3.2308,
+      "step": 71300
+    },
+    {
+      "epoch": 7.679474760520934,
+      "grad_norm": 0.859466552734375,
+      "learning_rate": 0.00013956901195991811,
+      "loss": 3.2333,
+      "step": 71350
+    },
+    {
+      "epoch": 7.684856312560543,
+      "grad_norm": 0.8073216676712036,
+      "learning_rate": 0.00013924577092985668,
+      "loss": 3.235,
+      "step": 71400
+    },
+    {
+      "epoch": 7.69023786460015,
+      "grad_norm": 0.8267530798912048,
+      "learning_rate": 0.00013892252989979528,
+      "loss": 3.2256,
+      "step": 71450
+    },
+    {
+      "epoch": 7.695619416639759,
+      "grad_norm": 0.7940675020217896,
+      "learning_rate": 0.00013859928886973384,
+      "loss": 3.252,
+      "step": 71500
+    },
+    {
+      "epoch": 7.7010009686793675,
+      "grad_norm": 0.8497976064682007,
+      "learning_rate": 0.00013827604783967244,
+      "loss": 3.2478,
+      "step": 71550
+    },
+    {
+      "epoch": 7.706382520718975,
+      "grad_norm": 0.8075435757637024,
+      "learning_rate": 0.000137952806809611,
+      "loss": 3.2527,
+      "step": 71600
+    },
+    {
+      "epoch": 7.711764072758584,
+      "grad_norm": 0.9132469892501831,
+      "learning_rate": 0.0001376295657795496,
+      "loss": 3.2292,
+      "step": 71650
+    },
+    {
+      "epoch": 7.717145624798192,
+      "grad_norm": 0.8953233957290649,
+      "learning_rate": 0.0001373063247494882,
+      "loss": 3.2451,
+      "step": 71700
+    },
+    {
+      "epoch": 7.7225271768378,
+      "grad_norm": 0.8485434651374817,
+      "learning_rate": 0.00013698308371942676,
+      "loss": 3.2117,
+      "step": 71750
+    },
+    {
+      "epoch": 7.727908728877408,
+      "grad_norm": 0.8661373853683472,
+      "learning_rate": 0.00013665984268936536,
+      "loss": 3.2466,
+      "step": 71800
+    },
+    {
+      "epoch": 7.733290280917016,
+      "grad_norm": 0.836283802986145,
+      "learning_rate": 0.00013633660165930395,
+      "loss": 3.2389,
+      "step": 71850
+    },
+    {
+      "epoch": 7.738671832956625,
+      "grad_norm": 0.807246208190918,
+      "learning_rate": 0.00013601336062924255,
+      "loss": 3.231,
+      "step": 71900
+    },
+    {
+      "epoch": 7.744053384996233,
+      "grad_norm": 0.8651958107948303,
+      "learning_rate": 0.00013569011959918111,
+      "loss": 3.2336,
+      "step": 71950
+    },
+    {
+      "epoch": 7.749434937035841,
+      "grad_norm": 0.8273272514343262,
+      "learning_rate": 0.0001353668785691197,
+      "loss": 3.2444,
+      "step": 72000
+    },
+    {
+      "epoch": 7.749434937035841,
+      "eval_accuracy": 0.38930840372501213,
+      "eval_loss": 3.340482234954834,
+      "eval_runtime": 185.405,
+      "eval_samples_per_second": 97.144,
+      "eval_steps_per_second": 6.073,
+      "step": 72000
+    },
+    {
+      "epoch": 7.754816489075449,
+      "grad_norm": 0.8395845890045166,
+      "learning_rate": 0.00013504363753905828,
+      "loss": 3.2439,
+      "step": 72050
+    },
+    {
+      "epoch": 7.760198041115058,
+      "grad_norm": 0.8324407935142517,
+      "learning_rate": 0.00013472039650899687,
+      "loss": 3.2334,
+      "step": 72100
+    },
+    {
+      "epoch": 7.765579593154666,
+      "grad_norm": 0.8480279445648193,
+      "learning_rate": 0.00013439715547893544,
+      "loss": 3.2353,
+      "step": 72150
+    },
+    {
+      "epoch": 7.770961145194274,
+      "grad_norm": 0.8069897890090942,
+      "learning_rate": 0.00013407391444887403,
+      "loss": 3.2422,
+      "step": 72200
+    },
+    {
+      "epoch": 7.776342697233883,
+      "grad_norm": 0.8504781126976013,
+      "learning_rate": 0.0001337506734188126,
+      "loss": 3.2324,
+      "step": 72250
+    },
+    {
+      "epoch": 7.78172424927349,
+      "grad_norm": 0.8418586254119873,
+      "learning_rate": 0.0001334274323887512,
+      "loss": 3.2312,
+      "step": 72300
+    },
+    {
+      "epoch": 7.787105801313099,
+      "grad_norm": 0.8298173546791077,
+      "learning_rate": 0.0001331041913586898,
+      "loss": 3.2409,
+      "step": 72350
+    },
+    {
+      "epoch": 7.792487353352707,
+      "grad_norm": 0.8165563941001892,
+      "learning_rate": 0.00013278095032862838,
+      "loss": 3.2535,
+      "step": 72400
+    },
+    {
+      "epoch": 7.797868905392315,
+      "grad_norm": 0.8215430974960327,
+      "learning_rate": 0.00013245770929856695,
+      "loss": 3.2498,
+      "step": 72450
+    },
+    {
+      "epoch": 7.803250457431924,
+      "grad_norm": 0.8466613292694092,
+      "learning_rate": 0.00013213446826850555,
+      "loss": 3.2202,
+      "step": 72500
+    },
+    {
+      "epoch": 7.808632009471531,
+      "grad_norm": 0.8652186989784241,
+      "learning_rate": 0.00013181122723844411,
+      "loss": 3.2449,
+      "step": 72550
+    },
+    {
+      "epoch": 7.81401356151114,
+      "grad_norm": 0.8607468008995056,
+      "learning_rate": 0.0001314879862083827,
+      "loss": 3.2233,
+      "step": 72600
+    },
+    {
+      "epoch": 7.819395113550748,
+      "grad_norm": 0.9004116654396057,
+      "learning_rate": 0.00013116474517832128,
+      "loss": 3.2111,
+      "step": 72650
+    },
+    {
+      "epoch": 7.824776665590356,
+      "grad_norm": 0.8476096391677856,
+      "learning_rate": 0.00013084796896886108,
+      "loss": 3.2389,
+      "step": 72700
+    },
+    {
+      "epoch": 7.830158217629965,
+      "grad_norm": 0.8326628804206848,
+      "learning_rate": 0.00013052472793879968,
+      "loss": 3.2499,
+      "step": 72750
+    },
+    {
+      "epoch": 7.835539769669572,
+      "grad_norm": 0.8324896693229675,
+      "learning_rate": 0.00013020148690873827,
+      "loss": 3.2302,
+      "step": 72800
+    },
+    {
+      "epoch": 7.840921321709181,
+      "grad_norm": 0.8407576680183411,
+      "learning_rate": 0.00012987824587867687,
+      "loss": 3.2296,
+      "step": 72850
+    },
+    {
+      "epoch": 7.846302873748789,
+      "grad_norm": 0.8449313044548035,
+      "learning_rate": 0.00012955500484861543,
+      "loss": 3.2321,
+      "step": 72900
+    },
+    {
+      "epoch": 7.851684425788397,
+      "grad_norm": 0.8302655816078186,
+      "learning_rate": 0.00012923176381855403,
+      "loss": 3.2465,
+      "step": 72950
+    },
+    {
+      "epoch": 7.857065977828006,
+      "grad_norm": 0.860231339931488,
+      "learning_rate": 0.00012890852278849262,
+      "loss": 3.2322,
+      "step": 73000
+    },
+    {
+      "epoch": 7.857065977828006,
+      "eval_accuracy": 0.38935349467345515,
+      "eval_loss": 3.340813159942627,
+      "eval_runtime": 185.8493,
+      "eval_samples_per_second": 96.912,
+      "eval_steps_per_second": 6.059,
+      "step": 73000
+    },
+    {
+      "epoch": 7.862447529867614,
+      "grad_norm": 0.8611194491386414,
+      "learning_rate": 0.0001285852817584312,
+      "loss": 3.2398,
+      "step": 73050
+    },
+    {
+      "epoch": 7.867829081907222,
+      "grad_norm": 0.8002138137817383,
+      "learning_rate": 0.0001282620407283698,
+      "loss": 3.2443,
+      "step": 73100
+    },
+    {
+      "epoch": 7.87321063394683,
+      "grad_norm": 0.84026700258255,
+      "learning_rate": 0.00012793879969830835,
+      "loss": 3.235,
+      "step": 73150
+    },
+    {
+      "epoch": 7.878592185986438,
+      "grad_norm": 0.8398573994636536,
+      "learning_rate": 0.00012761555866824695,
+      "loss": 3.2354,
+      "step": 73200
+    },
+    {
+      "epoch": 7.883973738026047,
+      "grad_norm": 0.8527116179466248,
+      "learning_rate": 0.00012729231763818552,
+      "loss": 3.2332,
+      "step": 73250
+    },
+    {
+      "epoch": 7.889355290065655,
+      "grad_norm": 0.7653493285179138,
+      "learning_rate": 0.0001269690766081241,
+      "loss": 3.2396,
+      "step": 73300
+    },
+    {
+      "epoch": 7.894736842105263,
+      "grad_norm": 0.8176438808441162,
+      "learning_rate": 0.0001266458355780627,
+      "loss": 3.2318,
+      "step": 73350
+    },
+    {
+      "epoch": 7.900118394144871,
+      "grad_norm": 0.8960065245628357,
+      "learning_rate": 0.00012632259454800127,
+      "loss": 3.2519,
+      "step": 73400
+    },
+    {
+      "epoch": 7.90549994618448,
+      "grad_norm": 0.8434229493141174,
+      "learning_rate": 0.00012599935351793987,
+      "loss": 3.2446,
+      "step": 73450
+    },
+    {
+      "epoch": 7.910881498224088,
+      "grad_norm": 0.8432302474975586,
+      "learning_rate": 0.00012567611248787846,
+      "loss": 3.2352,
+      "step": 73500
+    },
+    {
+      "epoch": 7.916263050263696,
+      "grad_norm": 0.8683959245681763,
+      "learning_rate": 0.00012535287145781703,
+      "loss": 3.2382,
+      "step": 73550
+    },
+    {
+      "epoch": 7.921644602303305,
+      "grad_norm": 0.8744735717773438,
+      "learning_rate": 0.00012502963042775562,
+      "loss": 3.2391,
+      "step": 73600
+    },
+    {
+      "epoch": 7.927026154342912,
+      "grad_norm": 0.8530826568603516,
+      "learning_rate": 0.00012470638939769422,
+      "loss": 3.2288,
+      "step": 73650
+    },
+    {
+      "epoch": 7.932407706382521,
+      "grad_norm": 1.1098095178604126,
+      "learning_rate": 0.0001243831483676328,
+      "loss": 3.2351,
+      "step": 73700
+    },
+    {
+      "epoch": 7.937789258422129,
+      "grad_norm": 0.8341307640075684,
+      "learning_rate": 0.00012405990733757138,
+      "loss": 3.2396,
+      "step": 73750
+    },
+    {
+      "epoch": 7.943170810461737,
+      "grad_norm": 0.7903709411621094,
+      "learning_rate": 0.00012373666630750995,
+      "loss": 3.2357,
+      "step": 73800
+    },
+    {
+      "epoch": 7.948552362501346,
+      "grad_norm": 0.8577542304992676,
+      "learning_rate": 0.00012341342527744854,
+      "loss": 3.2336,
+      "step": 73850
+    },
+    {
+      "epoch": 7.953933914540953,
+      "grad_norm": 0.8331533670425415,
+      "learning_rate": 0.0001230901842473871,
+      "loss": 3.2377,
+      "step": 73900
+    },
+    {
+      "epoch": 7.959315466580562,
+      "grad_norm": 0.7993036508560181,
+      "learning_rate": 0.0001227669432173257,
+      "loss": 3.2171,
+      "step": 73950
+    },
+    {
+      "epoch": 7.96469701862017,
+      "grad_norm": 0.8579351902008057,
+      "learning_rate": 0.0001224437021872643,
+      "loss": 3.2364,
+      "step": 74000
+    },
+    {
+      "epoch": 7.96469701862017,
+      "eval_accuracy": 0.3899118618639338,
+      "eval_loss": 3.3331358432769775,
+      "eval_runtime": 185.509,
+      "eval_samples_per_second": 97.09,
+      "eval_steps_per_second": 6.07,
+      "step": 74000
+    },
+    {
+      "epoch": 7.970078570659778,
+      "grad_norm": 0.8279479146003723,
+      "learning_rate": 0.00012212046115720287,
+      "loss": 3.2597,
+      "step": 74050
+    },
+    {
+      "epoch": 7.975460122699387,
+      "grad_norm": 0.8555736541748047,
+      "learning_rate": 0.00012179722012714146,
+      "loss": 3.2439,
+      "step": 74100
+    },
+    {
+      "epoch": 7.980841674738995,
+      "grad_norm": 0.9351614117622375,
+      "learning_rate": 0.00012147397909708004,
+      "loss": 3.2259,
+      "step": 74150
+    },
+    {
+      "epoch": 7.986223226778603,
+      "grad_norm": 0.8798420429229736,
+      "learning_rate": 0.00012115073806701864,
+      "loss": 3.21,
+      "step": 74200
+    },
+    {
+      "epoch": 7.991604778818211,
+      "grad_norm": 0.937835693359375,
+      "learning_rate": 0.0001208274970369572,
+      "loss": 3.2358,
+      "step": 74250
+    },
+    {
+      "epoch": 7.996986330857819,
+      "grad_norm": 0.8014573454856873,
+      "learning_rate": 0.0001205042560068958,
+      "loss": 3.2228,
+      "step": 74300
+    },
+    {
+      "epoch": 8.002367882897428,
+      "grad_norm": 0.8197176456451416,
+      "learning_rate": 0.0001201810149768344,
+      "loss": 3.1932,
+      "step": 74350
+    },
+    {
+      "epoch": 8.007749434937036,
+      "grad_norm": 0.835379421710968,
+      "learning_rate": 0.00011985777394677296,
+      "loss": 3.1667,
+      "step": 74400
+    },
+    {
+      "epoch": 8.013130986976645,
+      "grad_norm": 0.828336775302887,
+      "learning_rate": 0.00011953453291671154,
+      "loss": 3.152,
+      "step": 74450
+    },
+    {
+      "epoch": 8.018512539016251,
+      "grad_norm": 0.8461458683013916,
+      "learning_rate": 0.00011921129188665014,
+      "loss": 3.1727,
+      "step": 74500
+    },
+    {
+      "epoch": 8.02389409105586,
+      "grad_norm": 0.8270170092582703,
+      "learning_rate": 0.0001188880508565887,
+      "loss": 3.1569,
+      "step": 74550
+    },
+    {
+      "epoch": 8.029275643095469,
+      "grad_norm": 0.8645870685577393,
+      "learning_rate": 0.0001185648098265273,
+      "loss": 3.1714,
+      "step": 74600
+    },
+    {
+      "epoch": 8.034657195135077,
+      "grad_norm": 0.8919357061386108,
+      "learning_rate": 0.0001182415687964659,
+      "loss": 3.1602,
+      "step": 74650
+    },
+    {
+      "epoch": 8.040038747174686,
+      "grad_norm": 0.8075178265571594,
+      "learning_rate": 0.00011792479258700569,
+      "loss": 3.1649,
+      "step": 74700
+    },
+    {
+      "epoch": 8.045420299214294,
+      "grad_norm": 0.8222769498825073,
+      "learning_rate": 0.00011760155155694428,
+      "loss": 3.1784,
+      "step": 74750
+    },
+    {
+      "epoch": 8.050801851253901,
+      "grad_norm": 0.8642512559890747,
+      "learning_rate": 0.00011727831052688288,
+      "loss": 3.1766,
+      "step": 74800
+    },
+    {
+      "epoch": 8.05618340329351,
+      "grad_norm": 0.8777766227722168,
+      "learning_rate": 0.00011695506949682145,
+      "loss": 3.18,
+      "step": 74850
+    },
+    {
+      "epoch": 8.061564955333118,
+      "grad_norm": 0.9250959157943726,
+      "learning_rate": 0.00011663182846676004,
+      "loss": 3.1601,
+      "step": 74900
+    },
+    {
+      "epoch": 8.066946507372727,
+      "grad_norm": 0.8130394816398621,
+      "learning_rate": 0.00011630858743669862,
+      "loss": 3.1705,
+      "step": 74950
+    },
+    {
+      "epoch": 8.072328059412335,
+      "grad_norm": 0.8571843504905701,
+      "learning_rate": 0.0001159853464066372,
+      "loss": 3.1419,
+      "step": 75000
+    },
+    {
+      "epoch": 8.072328059412335,
+      "eval_accuracy": 0.3896375133222022,
+      "eval_loss": 3.3418679237365723,
+      "eval_runtime": 185.5319,
+      "eval_samples_per_second": 97.078,
+      "eval_steps_per_second": 6.069,
+      "step": 75000
+    },
+    {
+      "epoch": 8.077709611451942,
+      "grad_norm": 0.8765535950660706,
+      "learning_rate": 0.00011566210537657578,
+      "loss": 3.1765,
+      "step": 75050
+    },
+    {
+      "epoch": 8.08309116349155,
+      "grad_norm": 0.8717322945594788,
+      "learning_rate": 0.00011533886434651438,
+      "loss": 3.164,
+      "step": 75100
+    },
+    {
+      "epoch": 8.088472715531159,
+      "grad_norm": 0.8551101684570312,
+      "learning_rate": 0.00011501562331645296,
+      "loss": 3.1698,
+      "step": 75150
+    },
+    {
+      "epoch": 8.093854267570768,
+      "grad_norm": 0.8140628933906555,
+      "learning_rate": 0.00011469238228639154,
+      "loss": 3.156,
+      "step": 75200
+    },
+    {
+      "epoch": 8.099235819610376,
+      "grad_norm": 0.8646162152290344,
+      "learning_rate": 0.00011436914125633012,
+      "loss": 3.1699,
+      "step": 75250
+    },
+    {
+      "epoch": 8.104617371649983,
+      "grad_norm": 0.9125895500183105,
+      "learning_rate": 0.00011404590022626872,
+      "loss": 3.1562,
+      "step": 75300
+    },
+    {
+      "epoch": 8.109998923689592,
+      "grad_norm": 0.9186262488365173,
+      "learning_rate": 0.00011372265919620728,
+      "loss": 3.1651,
+      "step": 75350
+    },
+    {
+      "epoch": 8.1153804757292,
+      "grad_norm": 0.8632041811943054,
+      "learning_rate": 0.00011339941816614588,
+      "loss": 3.1789,
+      "step": 75400
+    },
+    {
+      "epoch": 8.120762027768809,
+      "grad_norm": 0.8314523696899414,
+      "learning_rate": 0.00011307617713608447,
+      "loss": 3.1708,
+      "step": 75450
+    },
+    {
+      "epoch": 8.126143579808417,
+      "grad_norm": 0.8675382137298584,
+      "learning_rate": 0.00011275293610602305,
+      "loss": 3.1817,
+      "step": 75500
+    },
+    {
+      "epoch": 8.131525131848026,
+      "grad_norm": 0.8234276175498962,
+      "learning_rate": 0.00011242969507596164,
+      "loss": 3.1704,
+      "step": 75550
+    },
+    {
+      "epoch": 8.136906683887632,
+      "grad_norm": 0.8563738465309143,
+      "learning_rate": 0.00011210645404590022,
+      "loss": 3.166,
+      "step": 75600
+    },
+    {
+      "epoch": 8.142288235927241,
+      "grad_norm": 0.889087975025177,
+      "learning_rate": 0.00011178321301583881,
+      "loss": 3.16,
+      "step": 75650
+    },
+    {
+      "epoch": 8.14766978796685,
+      "grad_norm": 0.8593072891235352,
+      "learning_rate": 0.00011145997198577738,
+      "loss": 3.1734,
+      "step": 75700
+    },
+    {
+      "epoch": 8.153051340006458,
+      "grad_norm": 0.8361025452613831,
+      "learning_rate": 0.00011113673095571597,
+      "loss": 3.1776,
+      "step": 75750
+    },
+    {
+      "epoch": 8.158432892046067,
+      "grad_norm": 0.8463952541351318,
+      "learning_rate": 0.00011081348992565455,
+      "loss": 3.1732,
+      "step": 75800
+    },
+    {
+      "epoch": 8.163814444085673,
+      "grad_norm": 0.8684695959091187,
+      "learning_rate": 0.00011049024889559314,
+      "loss": 3.1761,
+      "step": 75850
+    },
+    {
+      "epoch": 8.169195996125282,
+      "grad_norm": 0.8610966801643372,
+      "learning_rate": 0.00011016700786553172,
+      "loss": 3.1876,
+      "step": 75900
+    },
+    {
+      "epoch": 8.17457754816489,
+      "grad_norm": 0.8567910194396973,
+      "learning_rate": 0.00010984376683547031,
+      "loss": 3.181,
+      "step": 75950
+    },
+    {
+      "epoch": 8.1799591002045,
+      "grad_norm": 0.8930727243423462,
+      "learning_rate": 0.00010952052580540889,
+      "loss": 3.1676,
+      "step": 76000
+    },
+    {
+      "epoch": 8.1799591002045,
+      "eval_accuracy": 0.39042416022997906,
+      "eval_loss": 3.335977077484131,
+      "eval_runtime": 185.6456,
+      "eval_samples_per_second": 97.018,
+      "eval_steps_per_second": 6.065,
+      "step": 76000
+    },
+    {
+      "epoch": 8.185340652244108,
+      "grad_norm": 0.8385201096534729,
+      "learning_rate": 0.00010919728477534747,
+      "loss": 3.1758,
+      "step": 76050
+    },
+    {
+      "epoch": 8.190722204283716,
+      "grad_norm": 0.8283881545066833,
+      "learning_rate": 0.00010887404374528605,
+      "loss": 3.1731,
+      "step": 76100
+    },
+    {
+      "epoch": 8.196103756323323,
+      "grad_norm": 0.9429885149002075,
+      "learning_rate": 0.00010855080271522465,
+      "loss": 3.1752,
+      "step": 76150
+    },
+    {
+      "epoch": 8.201485308362932,
+      "grad_norm": 0.8766368627548218,
+      "learning_rate": 0.00010822756168516322,
+      "loss": 3.1624,
+      "step": 76200
+    },
+    {
+      "epoch": 8.20686686040254,
+      "grad_norm": 0.9043298959732056,
+      "learning_rate": 0.00010790432065510181,
+      "loss": 3.1825,
+      "step": 76250
+    },
+    {
+      "epoch": 8.212248412442149,
+      "grad_norm": 0.9231343269348145,
+      "learning_rate": 0.0001075810796250404,
+      "loss": 3.1755,
+      "step": 76300
+    },
+    {
+      "epoch": 8.217629964481757,
+      "grad_norm": 0.8669461607933044,
+      "learning_rate": 0.00010725783859497897,
+      "loss": 3.1646,
+      "step": 76350
+    },
+    {
+      "epoch": 8.223011516521364,
+      "grad_norm": 0.9116910099983215,
+      "learning_rate": 0.00010693459756491757,
+      "loss": 3.1826,
+      "step": 76400
+    },
+    {
+      "epoch": 8.228393068560973,
+      "grad_norm": 0.9033715128898621,
+      "learning_rate": 0.00010661135653485615,
+      "loss": 3.1876,
+      "step": 76450
+    },
+    {
+      "epoch": 8.233774620600581,
+      "grad_norm": 0.9512905478477478,
+      "learning_rate": 0.00010628811550479474,
+      "loss": 3.1752,
+      "step": 76500
+    },
+    {
+      "epoch": 8.23915617264019,
+      "grad_norm": 0.84199458360672,
+      "learning_rate": 0.00010596487447473331,
+      "loss": 3.1633,
+      "step": 76550
+    },
+    {
+      "epoch": 8.244537724679798,
+      "grad_norm": 1.1381860971450806,
+      "learning_rate": 0.0001056416334446719,
+      "loss": 3.1744,
+      "step": 76600
+    },
+    {
+      "epoch": 8.249919276719407,
+      "grad_norm": 0.8594551682472229,
+      "learning_rate": 0.00010531839241461049,
+      "loss": 3.1727,
+      "step": 76650
+    },
+    {
+      "epoch": 8.255300828759013,
+      "grad_norm": 0.9258723258972168,
+      "learning_rate": 0.00010499515138454907,
+      "loss": 3.1999,
+      "step": 76700
+    },
+    {
+      "epoch": 8.260682380798622,
+      "grad_norm": 0.8554587960243225,
+      "learning_rate": 0.00010467191035448765,
+      "loss": 3.1899,
+      "step": 76750
+    },
+    {
+      "epoch": 8.26606393283823,
+      "grad_norm": 0.8644444942474365,
+      "learning_rate": 0.00010434866932442624,
+      "loss": 3.1766,
+      "step": 76800
+    },
+    {
+      "epoch": 8.27144548487784,
+      "grad_norm": 0.8380032181739807,
+      "learning_rate": 0.00010403189311496605,
+      "loss": 3.1812,
+      "step": 76850
+    },
+    {
+      "epoch": 8.276827036917448,
+      "grad_norm": 0.8622545003890991,
+      "learning_rate": 0.00010370865208490463,
+      "loss": 3.1949,
+      "step": 76900
+    },
+    {
+      "epoch": 8.282208588957054,
+      "grad_norm": 0.8545438647270203,
+      "learning_rate": 0.00010338541105484323,
+      "loss": 3.1748,
+      "step": 76950
+    },
+    {
+      "epoch": 8.287590140996663,
+      "grad_norm": 0.8771522641181946,
+      "learning_rate": 0.0001030621700247818,
+      "loss": 3.1775,
+      "step": 77000
+    },
+    {
+      "epoch": 8.287590140996663,
+      "eval_accuracy": 0.3902832374344837,
+      "eval_loss": 3.336022138595581,
+      "eval_runtime": 185.8491,
+      "eval_samples_per_second": 96.912,
+      "eval_steps_per_second": 6.059,
+      "step": 77000
+    },
+    {
+      "epoch": 8.292971693036272,
+      "grad_norm": 0.8804690837860107,
+      "learning_rate": 0.00010273892899472039,
+      "loss": 3.174,
+      "step": 77050
+    },
+    {
+      "epoch": 8.29835324507588,
+      "grad_norm": 0.8765997290611267,
+      "learning_rate": 0.00010241568796465898,
+      "loss": 3.183,
+      "step": 77100
+    },
+    {
+      "epoch": 8.303734797115489,
+      "grad_norm": 0.8380710482597351,
+      "learning_rate": 0.00010209244693459755,
+      "loss": 3.1927,
+      "step": 77150
+    },
+    {
+      "epoch": 8.309116349155097,
+      "grad_norm": 0.886776864528656,
+      "learning_rate": 0.00010176920590453613,
+      "loss": 3.1668,
+      "step": 77200
+    },
+    {
+      "epoch": 8.314497901194704,
+      "grad_norm": 0.8976190090179443,
+      "learning_rate": 0.00010144596487447473,
+      "loss": 3.1838,
+      "step": 77250
+    },
+    {
+      "epoch": 8.319879453234313,
+      "grad_norm": 0.8447895050048828,
+      "learning_rate": 0.00010112272384441332,
+      "loss": 3.1865,
+      "step": 77300
+    },
+    {
+      "epoch": 8.325261005273921,
+      "grad_norm": 0.8776809573173523,
+      "learning_rate": 0.00010079948281435189,
+      "loss": 3.1652,
+      "step": 77350
+    },
+    {
+      "epoch": 8.33064255731353,
+      "grad_norm": 0.863149106502533,
+      "learning_rate": 0.00010047624178429048,
+      "loss": 3.1782,
+      "step": 77400
+    },
+    {
+      "epoch": 8.336024109353138,
+      "grad_norm": 0.8674944043159485,
+      "learning_rate": 0.00010015300075422906,
+      "loss": 3.1791,
+      "step": 77450
+    },
+    {
+      "epoch": 8.341405661392745,
+      "grad_norm": 0.8574560284614563,
+      "learning_rate": 9.982975972416765e-05,
+      "loss": 3.1608,
+      "step": 77500
+    },
+    {
+      "epoch": 8.346787213432354,
+      "grad_norm": 0.8597567081451416,
+      "learning_rate": 9.950651869410623e-05,
+      "loss": 3.1781,
+      "step": 77550
+    },
+    {
+      "epoch": 8.352168765471962,
+      "grad_norm": 0.9663563370704651,
+      "learning_rate": 9.918327766404482e-05,
+      "loss": 3.1775,
+      "step": 77600
+    },
+    {
+      "epoch": 8.35755031751157,
+      "grad_norm": 0.8977883458137512,
+      "learning_rate": 9.886003663398339e-05,
+      "loss": 3.1797,
+      "step": 77650
+    },
+    {
+      "epoch": 8.36293186955118,
+      "grad_norm": 0.8807985782623291,
+      "learning_rate": 9.853679560392198e-05,
+      "loss": 3.1731,
+      "step": 77700
+    },
+    {
+      "epoch": 8.368313421590786,
+      "grad_norm": 0.8611074090003967,
+      "learning_rate": 9.821355457386056e-05,
+      "loss": 3.1659,
+      "step": 77750
+    },
+    {
+      "epoch": 8.373694973630395,
+      "grad_norm": 0.8290106654167175,
+      "learning_rate": 9.789031354379916e-05,
+      "loss": 3.197,
+      "step": 77800
+    },
+    {
+      "epoch": 8.379076525670003,
+      "grad_norm": 0.8541063070297241,
+      "learning_rate": 9.757353733433897e-05,
+      "loss": 3.1953,
+      "step": 77850
+    },
+    {
+      "epoch": 8.384458077709612,
+      "grad_norm": 0.8111417889595032,
+      "learning_rate": 9.725029630427755e-05,
+      "loss": 3.1913,
+      "step": 77900
+    },
+    {
+      "epoch": 8.38983962974922,
+      "grad_norm": 0.8581722974777222,
+      "learning_rate": 9.692705527421613e-05,
+      "loss": 3.1858,
+      "step": 77950
+    },
+    {
+      "epoch": 8.395221181788829,
+      "grad_norm": 0.847231388092041,
+      "learning_rate": 9.660381424415471e-05,
+      "loss": 3.1892,
+      "step": 78000
+    },
+    {
+      "epoch": 8.395221181788829,
+      "eval_accuracy": 0.3906853617722851,
+      "eval_loss": 3.3310365676879883,
+      "eval_runtime": 185.5025,
+      "eval_samples_per_second": 97.093,
+      "eval_steps_per_second": 6.07,
+      "step": 78000
+    },
+    {
+      "epoch": 8.400602733828435,
+      "grad_norm": 0.878339946269989,
+      "learning_rate": 9.62805732140933e-05,
+      "loss": 3.1815,
+      "step": 78050
+    },
+    {
+      "epoch": 8.405984285868044,
+      "grad_norm": 0.9056654572486877,
+      "learning_rate": 9.595733218403187e-05,
+      "loss": 3.1581,
+      "step": 78100
+    },
+    {
+      "epoch": 8.411365837907653,
+      "grad_norm": 0.8745837211608887,
+      "learning_rate": 9.563409115397047e-05,
+      "loss": 3.1974,
+      "step": 78150
+    },
+    {
+      "epoch": 8.416747389947261,
+      "grad_norm": 0.8908196091651917,
+      "learning_rate": 9.531085012390906e-05,
+      "loss": 3.1726,
+      "step": 78200
+    },
+    {
+      "epoch": 8.42212894198687,
+      "grad_norm": 0.8629767894744873,
+      "learning_rate": 9.498760909384764e-05,
+      "loss": 3.1835,
+      "step": 78250
+    },
+    {
+      "epoch": 8.427510494026476,
+      "grad_norm": 0.8738199472427368,
+      "learning_rate": 9.466436806378622e-05,
+      "loss": 3.181,
+      "step": 78300
+    },
+    {
+      "epoch": 8.432892046066085,
+      "grad_norm": 0.836013913154602,
+      "learning_rate": 9.43411270337248e-05,
+      "loss": 3.1909,
+      "step": 78350
+    },
+    {
+      "epoch": 8.438273598105694,
+      "grad_norm": 0.8754242062568665,
+      "learning_rate": 9.40178860036634e-05,
+      "loss": 3.185,
+      "step": 78400
+    },
+    {
+      "epoch": 8.443655150145302,
+      "grad_norm": 0.875544011592865,
+      "learning_rate": 9.369464497360197e-05,
+      "loss": 3.1792,
+      "step": 78450
+    },
+    {
+      "epoch": 8.44903670218491,
+      "grad_norm": 0.9628906846046448,
+      "learning_rate": 9.337140394354056e-05,
+      "loss": 3.174,
+      "step": 78500
+    },
+    {
+      "epoch": 8.45441825422452,
+      "grad_norm": 0.9128474593162537,
+      "learning_rate": 9.304816291347914e-05,
+      "loss": 3.1968,
+      "step": 78550
+    },
+    {
+      "epoch": 8.459799806264126,
+      "grad_norm": 0.8624871373176575,
+      "learning_rate": 9.272492188341772e-05,
+      "loss": 3.2002,
+      "step": 78600
+    },
+    {
+      "epoch": 8.465181358303735,
+      "grad_norm": 0.8904179334640503,
+      "learning_rate": 9.24016808533563e-05,
+      "loss": 3.1815,
+      "step": 78650
+    },
+    {
+      "epoch": 8.470562910343343,
+      "grad_norm": 0.9286843538284302,
+      "learning_rate": 9.20784398232949e-05,
+      "loss": 3.1722,
+      "step": 78700
+    },
+    {
+      "epoch": 8.475944462382952,
+      "grad_norm": 0.8553939461708069,
+      "learning_rate": 9.175519879323348e-05,
+      "loss": 3.1986,
+      "step": 78750
+    },
+    {
+      "epoch": 8.48132601442256,
+      "grad_norm": 0.8739849328994751,
+      "learning_rate": 9.143195776317206e-05,
+      "loss": 3.1898,
+      "step": 78800
+    },
+    {
+      "epoch": 8.486707566462167,
+      "grad_norm": 0.8722026348114014,
+      "learning_rate": 9.110871673311064e-05,
+      "loss": 3.1976,
+      "step": 78850
+    },
+    {
+      "epoch": 8.492089118501776,
+      "grad_norm": 0.894361138343811,
+      "learning_rate": 9.078547570304924e-05,
+      "loss": 3.2006,
+      "step": 78900
+    },
+    {
+      "epoch": 8.497470670541384,
+      "grad_norm": 0.8858276009559631,
+      "learning_rate": 9.04622346729878e-05,
+      "loss": 3.1861,
+      "step": 78950
+    },
+    {
+      "epoch": 8.502852222580993,
+      "grad_norm": 0.9259702563285828,
+      "learning_rate": 9.01389936429264e-05,
+      "loss": 3.1954,
+      "step": 79000
+    },
+    {
+      "epoch": 8.502852222580993,
+      "eval_accuracy": 0.39098730814752153,
+      "eval_loss": 3.3294010162353516,
+      "eval_runtime": 185.7955,
+      "eval_samples_per_second": 96.94,
+      "eval_steps_per_second": 6.06,
+      "step": 79000
+    },
+    {
+      "epoch": 8.508233774620601,
+      "grad_norm": 0.8744315505027771,
+      "learning_rate": 8.9815752612865e-05,
+      "loss": 3.1921,
+      "step": 79050
+    },
+    {
+      "epoch": 8.513615326660208,
+      "grad_norm": 0.8402537703514099,
+      "learning_rate": 8.949251158280356e-05,
+      "loss": 3.1806,
+      "step": 79100
+    },
+    {
+      "epoch": 8.518996878699816,
+      "grad_norm": 0.869113564491272,
+      "learning_rate": 8.916927055274216e-05,
+      "loss": 3.1682,
+      "step": 79150
+    },
+    {
+      "epoch": 8.524378430739425,
+      "grad_norm": 0.8778077363967896,
+      "learning_rate": 8.884602952268074e-05,
+      "loss": 3.184,
+      "step": 79200
+    },
+    {
+      "epoch": 8.529759982779034,
+      "grad_norm": 0.9248849153518677,
+      "learning_rate": 8.852278849261933e-05,
+      "loss": 3.185,
+      "step": 79250
+    },
+    {
+      "epoch": 8.535141534818642,
+      "grad_norm": 0.9163711071014404,
+      "learning_rate": 8.81995474625579e-05,
+      "loss": 3.197,
+      "step": 79300
+    },
+    {
+      "epoch": 8.54052308685825,
+      "grad_norm": 0.8865178227424622,
+      "learning_rate": 8.78763064324965e-05,
+      "loss": 3.181,
+      "step": 79350
+    },
+    {
+      "epoch": 8.545904638897857,
+      "grad_norm": 0.8790997862815857,
+      "learning_rate": 8.755306540243508e-05,
+      "loss": 3.187,
+      "step": 79400
+    },
+    {
+      "epoch": 8.551286190937466,
+      "grad_norm": 0.9154719114303589,
+      "learning_rate": 8.722982437237366e-05,
+      "loss": 3.1881,
+      "step": 79450
+    },
+    {
+      "epoch": 8.556667742977075,
+      "grad_norm": 0.8903060555458069,
+      "learning_rate": 8.690658334231224e-05,
+      "loss": 3.1816,
+      "step": 79500
+    },
+    {
+      "epoch": 8.562049295016683,
+      "grad_norm": 0.9519715309143066,
+      "learning_rate": 8.658334231225083e-05,
+      "loss": 3.1716,
+      "step": 79550
+    },
+    {
+      "epoch": 8.567430847056292,
+      "grad_norm": 0.8465832471847534,
+      "learning_rate": 8.626010128218943e-05,
+      "loss": 3.1964,
+      "step": 79600
+    },
+    {
+      "epoch": 8.572812399095898,
+      "grad_norm": 0.8901507258415222,
+      "learning_rate": 8.5936860252128e-05,
+      "loss": 3.2031,
+      "step": 79650
+    },
+    {
+      "epoch": 8.578193951135507,
+      "grad_norm": 0.9377862811088562,
+      "learning_rate": 8.561361922206658e-05,
+      "loss": 3.2035,
+      "step": 79700
+    },
+    {
+      "epoch": 8.583575503175116,
+      "grad_norm": 0.9061275124549866,
+      "learning_rate": 8.529037819200517e-05,
+      "loss": 3.1873,
+      "step": 79750
+    },
+    {
+      "epoch": 8.588957055214724,
+      "grad_norm": 0.8865317106246948,
+      "learning_rate": 8.496713716194374e-05,
+      "loss": 3.1799,
+      "step": 79800
+    },
+    {
+      "epoch": 8.594338607254333,
+      "grad_norm": 0.895861804485321,
+      "learning_rate": 8.464389613188233e-05,
+      "loss": 3.1745,
+      "step": 79850
+    },
+    {
+      "epoch": 8.599720159293941,
+      "grad_norm": 0.9048733115196228,
+      "learning_rate": 8.432065510182093e-05,
+      "loss": 3.1968,
+      "step": 79900
+    },
+    {
+      "epoch": 8.605101711333548,
+      "grad_norm": 0.8376519680023193,
+      "learning_rate": 8.39974140717595e-05,
+      "loss": 3.1901,
+      "step": 79950
+    },
+    {
+      "epoch": 8.610483263373157,
+      "grad_norm": 0.8688256144523621,
+      "learning_rate": 8.367417304169809e-05,
+      "loss": 3.2003,
+      "step": 80000
+    },
+    {
+      "epoch": 8.610483263373157,
+      "eval_accuracy": 0.3915412205695997,
+      "eval_loss": 3.324683904647827,
+      "eval_runtime": 185.4409,
+      "eval_samples_per_second": 97.125,
+      "eval_steps_per_second": 6.072,
+      "step": 80000
+    },
+    {
+      "epoch": 8.615864815412765,
+      "grad_norm": 0.9323558211326599,
+      "learning_rate": 8.335093201163667e-05,
+      "loss": 3.2,
+      "step": 80050
+    },
+    {
+      "epoch": 8.621246367452374,
+      "grad_norm": 0.8564095497131348,
+      "learning_rate": 8.302769098157526e-05,
+      "loss": 3.1817,
+      "step": 80100
+    },
+    {
+      "epoch": 8.626627919491982,
+      "grad_norm": 0.8850892186164856,
+      "learning_rate": 8.270444995151383e-05,
+      "loss": 3.1796,
+      "step": 80150
+    },
+    {
+      "epoch": 8.632009471531589,
+      "grad_norm": 0.8858954906463623,
+      "learning_rate": 8.238120892145243e-05,
+      "loss": 3.1884,
+      "step": 80200
+    },
+    {
+      "epoch": 8.637391023571197,
+      "grad_norm": 0.896980881690979,
+      "learning_rate": 8.205796789139101e-05,
+      "loss": 3.1824,
+      "step": 80250
+    },
+    {
+      "epoch": 8.642772575610806,
+      "grad_norm": 0.8858007788658142,
+      "learning_rate": 8.173472686132959e-05,
+      "loss": 3.1748,
+      "step": 80300
+    },
+    {
+      "epoch": 8.648154127650415,
+      "grad_norm": 0.8943002820014954,
+      "learning_rate": 8.141148583126817e-05,
+      "loss": 3.1638,
+      "step": 80350
+    },
+    {
+      "epoch": 8.653535679690023,
+      "grad_norm": 0.8572128415107727,
+      "learning_rate": 8.108824480120676e-05,
+      "loss": 3.1858,
+      "step": 80400
+    },
+    {
+      "epoch": 8.658917231729632,
+      "grad_norm": 0.8594452142715454,
+      "learning_rate": 8.076500377114533e-05,
+      "loss": 3.1896,
+      "step": 80450
+    },
+    {
+      "epoch": 8.664298783769238,
+      "grad_norm": 0.8973984718322754,
+      "learning_rate": 8.044176274108393e-05,
+      "loss": 3.1764,
+      "step": 80500
+    },
+    {
+      "epoch": 8.669680335808847,
+      "grad_norm": 0.8921085596084595,
+      "learning_rate": 8.011852171102252e-05,
+      "loss": 3.1859,
+      "step": 80550
+    },
+    {
+      "epoch": 8.675061887848456,
+      "grad_norm": 0.8630329370498657,
+      "learning_rate": 7.97952806809611e-05,
+      "loss": 3.1843,
+      "step": 80600
+    },
+    {
+      "epoch": 8.680443439888064,
+      "grad_norm": 0.9078119397163391,
+      "learning_rate": 7.947203965089967e-05,
+      "loss": 3.1967,
+      "step": 80650
+    },
+    {
+      "epoch": 8.685824991927673,
+      "grad_norm": 0.8860991597175598,
+      "learning_rate": 7.914879862083827e-05,
+      "loss": 3.1859,
+      "step": 80700
+    },
+    {
+      "epoch": 8.69120654396728,
+      "grad_norm": 0.8678221106529236,
+      "learning_rate": 7.882555759077686e-05,
+      "loss": 3.1855,
+      "step": 80750
+    },
+    {
+      "epoch": 8.696588096006888,
+      "grad_norm": 0.9255539774894714,
+      "learning_rate": 7.850231656071543e-05,
+      "loss": 3.1857,
+      "step": 80800
+    },
+    {
+      "epoch": 8.701969648046497,
+      "grad_norm": 0.8874573111534119,
+      "learning_rate": 7.817907553065402e-05,
+      "loss": 3.1878,
+      "step": 80850
+    },
+    {
+      "epoch": 8.707351200086105,
+      "grad_norm": 0.8450101017951965,
+      "learning_rate": 7.78558345005926e-05,
+      "loss": 3.1748,
+      "step": 80900
+    },
+    {
+      "epoch": 8.712732752125714,
+      "grad_norm": 0.8771765232086182,
+      "learning_rate": 7.753259347053118e-05,
+      "loss": 3.201,
+      "step": 80950
+    },
+    {
+      "epoch": 8.718114304165322,
+      "grad_norm": 0.8859253525733948,
+      "learning_rate": 7.720935244046977e-05,
+      "loss": 3.1731,
+      "step": 81000
+    },
+    {
+      "epoch": 8.718114304165322,
+      "eval_accuracy": 0.39170767679373153,
+      "eval_loss": 3.321019172668457,
+      "eval_runtime": 185.439,
+      "eval_samples_per_second": 97.126,
+      "eval_steps_per_second": 6.072,
+      "step": 81000
+    },
+    {
+      "epoch": 8.723495856204929,
+      "grad_norm": 0.8705030083656311,
+      "learning_rate": 7.688611141040836e-05,
+      "loss": 3.1827,
+      "step": 81050
+    },
+    {
+      "epoch": 8.728877408244538,
+      "grad_norm": 0.8424695730209351,
+      "learning_rate": 7.656287038034694e-05,
+      "loss": 3.1839,
+      "step": 81100
+    },
+    {
+      "epoch": 8.734258960284146,
+      "grad_norm": 0.9107112884521484,
+      "learning_rate": 7.623962935028552e-05,
+      "loss": 3.1932,
+      "step": 81150
+    },
+    {
+      "epoch": 8.739640512323755,
+      "grad_norm": 0.9014825820922852,
+      "learning_rate": 7.59163883202241e-05,
+      "loss": 3.1895,
+      "step": 81200
+    },
+    {
+      "epoch": 8.745022064363363,
+      "grad_norm": 0.9136543273925781,
+      "learning_rate": 7.55931472901627e-05,
+      "loss": 3.1694,
+      "step": 81250
+    },
+    {
+      "epoch": 8.75040361640297,
+      "grad_norm": 0.8371633887290955,
+      "learning_rate": 7.526990626010127e-05,
+      "loss": 3.1862,
+      "step": 81300
+    },
+    {
+      "epoch": 8.755785168442578,
+      "grad_norm": 0.9018141031265259,
+      "learning_rate": 7.494666523003986e-05,
+      "loss": 3.1812,
+      "step": 81350
+    },
+    {
+      "epoch": 8.761166720482187,
+      "grad_norm": 0.9236919283866882,
+      "learning_rate": 7.462342419997844e-05,
+      "loss": 3.1872,
+      "step": 81400
+    },
+    {
+      "epoch": 8.766548272521796,
+      "grad_norm": 0.8943373560905457,
+      "learning_rate": 7.430018316991704e-05,
+      "loss": 3.1721,
+      "step": 81450
+    },
+    {
+      "epoch": 8.771929824561404,
+      "grad_norm": 0.9408779144287109,
+      "learning_rate": 7.397694213985562e-05,
+      "loss": 3.1712,
+      "step": 81500
+    },
+    {
+      "epoch": 8.777311376601011,
+      "grad_norm": 0.9225397706031799,
+      "learning_rate": 7.36537011097942e-05,
+      "loss": 3.1913,
+      "step": 81550
+    },
+    {
+      "epoch": 8.78269292864062,
+      "grad_norm": 0.8760676980018616,
+      "learning_rate": 7.333046007973278e-05,
+      "loss": 3.1892,
+      "step": 81600
+    },
+    {
+      "epoch": 8.788074480680228,
+      "grad_norm": 0.8328770995140076,
+      "learning_rate": 7.300721904967136e-05,
+      "loss": 3.1736,
+      "step": 81650
+    },
+    {
+      "epoch": 8.793456032719837,
+      "grad_norm": 0.8863219618797302,
+      "learning_rate": 7.268397801960996e-05,
+      "loss": 3.1748,
+      "step": 81700
+    },
+    {
+      "epoch": 8.798837584759445,
+      "grad_norm": 0.8880018591880798,
+      "learning_rate": 7.236073698954854e-05,
+      "loss": 3.1689,
+      "step": 81750
+    },
+    {
+      "epoch": 8.804219136799054,
+      "grad_norm": 0.9081376194953918,
+      "learning_rate": 7.203749595948712e-05,
+      "loss": 3.174,
+      "step": 81800
+    },
+    {
+      "epoch": 8.80960068883866,
+      "grad_norm": 0.8762065768241882,
+      "learning_rate": 7.17142549294257e-05,
+      "loss": 3.1811,
+      "step": 81850
+    },
+    {
+      "epoch": 8.814982240878269,
+      "grad_norm": 0.8693548440933228,
+      "learning_rate": 7.139747871996552e-05,
+      "loss": 3.187,
+      "step": 81900
+    },
+    {
+      "epoch": 8.820363792917878,
+      "grad_norm": 0.856499195098877,
+      "learning_rate": 7.10742376899041e-05,
+      "loss": 3.1885,
+      "step": 81950
+    },
+    {
+      "epoch": 8.825745344957486,
+      "grad_norm": 0.8709934949874878,
+      "learning_rate": 7.075099665984268e-05,
+      "loss": 3.1837,
+      "step": 82000
+    },
+    {
+      "epoch": 8.825745344957486,
+      "eval_accuracy": 0.3921330528495252,
+      "eval_loss": 3.3179738521575928,
+      "eval_runtime": 185.7425,
+      "eval_samples_per_second": 96.968,
+      "eval_steps_per_second": 6.062,
+      "step": 82000
+    },
+    {
+      "epoch": 8.831126896997095,
+      "grad_norm": 0.9127514958381653,
+      "learning_rate": 7.042775562978126e-05,
+      "loss": 3.1907,
+      "step": 82050
+    },
+    {
+      "epoch": 8.836508449036701,
+      "grad_norm": 0.9038019776344299,
+      "learning_rate": 7.010451459971986e-05,
+      "loss": 3.1822,
+      "step": 82100
+    },
+    {
+      "epoch": 8.84189000107631,
+      "grad_norm": 0.828102707862854,
+      "learning_rate": 6.978127356965844e-05,
+      "loss": 3.1783,
+      "step": 82150
+    },
+    {
+      "epoch": 8.847271553115919,
+      "grad_norm": 0.9361454844474792,
+      "learning_rate": 6.945803253959702e-05,
+      "loss": 3.1988,
+      "step": 82200
+    },
+    {
+      "epoch": 8.852653105155527,
+      "grad_norm": 0.911879301071167,
+      "learning_rate": 6.91347915095356e-05,
+      "loss": 3.1646,
+      "step": 82250
+    },
+    {
+      "epoch": 8.858034657195136,
+      "grad_norm": 0.8919516801834106,
+      "learning_rate": 6.881155047947418e-05,
+      "loss": 3.202,
+      "step": 82300
+    },
+    {
+      "epoch": 8.863416209234742,
+      "grad_norm": 0.8728047609329224,
+      "learning_rate": 6.848830944941278e-05,
+      "loss": 3.1533,
+      "step": 82350
+    },
+    {
+      "epoch": 8.868797761274351,
+      "grad_norm": 0.8462620377540588,
+      "learning_rate": 6.816506841935136e-05,
+      "loss": 3.1775,
+      "step": 82400
+    },
+    {
+      "epoch": 8.87417931331396,
+      "grad_norm": 0.8610141277313232,
+      "learning_rate": 6.784182738928994e-05,
+      "loss": 3.1925,
+      "step": 82450
+    },
+    {
+      "epoch": 8.879560865353568,
+      "grad_norm": 0.9105156064033508,
+      "learning_rate": 6.751858635922853e-05,
+      "loss": 3.1855,
+      "step": 82500
+    },
+    {
+      "epoch": 8.884942417393177,
+      "grad_norm": 0.885654628276825,
+      "learning_rate": 6.719534532916711e-05,
+      "loss": 3.1936,
+      "step": 82550
+    },
+    {
+      "epoch": 8.890323969432785,
+      "grad_norm": 0.882387638092041,
+      "learning_rate": 6.68721042991057e-05,
+      "loss": 3.1939,
+      "step": 82600
+    },
+    {
+      "epoch": 8.895705521472392,
+      "grad_norm": 0.874205470085144,
+      "learning_rate": 6.654886326904428e-05,
+      "loss": 3.2006,
+      "step": 82650
+    },
+    {
+      "epoch": 8.901087073512,
+      "grad_norm": 0.8664303421974182,
+      "learning_rate": 6.622562223898286e-05,
+      "loss": 3.1921,
+      "step": 82700
+    },
+    {
+      "epoch": 8.906468625551609,
+      "grad_norm": 0.936762809753418,
+      "learning_rate": 6.590238120892145e-05,
+      "loss": 3.1831,
+      "step": 82750
+    },
+    {
+      "epoch": 8.911850177591218,
+      "grad_norm": 0.9175882935523987,
+      "learning_rate": 6.557914017886003e-05,
+      "loss": 3.1757,
+      "step": 82800
+    },
+    {
+      "epoch": 8.917231729630826,
+      "grad_norm": 0.853027880191803,
+      "learning_rate": 6.525589914879861e-05,
+      "loss": 3.1803,
+      "step": 82850
+    },
+    {
+      "epoch": 8.922613281670433,
+      "grad_norm": 0.9543153047561646,
+      "learning_rate": 6.49326581187372e-05,
+      "loss": 3.1711,
+      "step": 82900
+    },
+    {
+      "epoch": 8.927994833710041,
+      "grad_norm": 0.8535247445106506,
+      "learning_rate": 6.460941708867578e-05,
+      "loss": 3.1663,
+      "step": 82950
+    },
+    {
+      "epoch": 8.93337638574965,
+      "grad_norm": 0.8562341332435608,
+      "learning_rate": 6.428617605861437e-05,
+      "loss": 3.1898,
+      "step": 83000
+    },
+    {
+      "epoch": 8.93337638574965,
+      "eval_accuracy": 0.3924914987264252,
+      "eval_loss": 3.3148152828216553,
+      "eval_runtime": 185.3394,
+      "eval_samples_per_second": 97.178,
+      "eval_steps_per_second": 6.075,
+      "step": 83000
+    },
+    {
+      "epoch": 8.938757937789259,
+      "grad_norm": 0.8555806875228882,
+      "learning_rate": 6.396293502855295e-05,
+      "loss": 3.1725,
+      "step": 83050
+    },
+    {
+      "epoch": 8.944139489828867,
+      "grad_norm": 0.9101053476333618,
+      "learning_rate": 6.363969399849153e-05,
+      "loss": 3.1803,
+      "step": 83100
+    },
+    {
+      "epoch": 8.949521041868476,
+      "grad_norm": 0.8834825754165649,
+      "learning_rate": 6.331645296843011e-05,
+      "loss": 3.1802,
+      "step": 83150
+    },
+    {
+      "epoch": 8.954902593908082,
+      "grad_norm": 0.9319151639938354,
+      "learning_rate": 6.29932119383687e-05,
+      "loss": 3.1824,
+      "step": 83200
+    },
+    {
+      "epoch": 8.960284145947691,
+      "grad_norm": 0.9192076325416565,
+      "learning_rate": 6.266997090830729e-05,
+      "loss": 3.1869,
+      "step": 83250
+    },
+    {
+      "epoch": 8.9656656979873,
+      "grad_norm": 0.8794592618942261,
+      "learning_rate": 6.234672987824587e-05,
+      "loss": 3.1779,
+      "step": 83300
+    },
+    {
+      "epoch": 8.971047250026908,
+      "grad_norm": 0.9025070667266846,
+      "learning_rate": 6.202348884818447e-05,
+      "loss": 3.1928,
+      "step": 83350
+    },
+    {
+      "epoch": 8.976428802066517,
+      "grad_norm": 0.872613787651062,
+      "learning_rate": 6.170024781812305e-05,
+      "loss": 3.1747,
+      "step": 83400
+    },
+    {
+      "epoch": 8.981810354106123,
+      "grad_norm": 0.9203582406044006,
+      "learning_rate": 6.137700678806163e-05,
+      "loss": 3.1928,
+      "step": 83450
+    },
+    {
+      "epoch": 8.987191906145732,
+      "grad_norm": 0.9269907474517822,
+      "learning_rate": 6.105376575800021e-05,
+      "loss": 3.1746,
+      "step": 83500
+    },
+    {
+      "epoch": 8.99257345818534,
+      "grad_norm": 0.927282989025116,
+      "learning_rate": 6.073052472793879e-05,
+      "loss": 3.1934,
+      "step": 83550
+    },
+    {
+      "epoch": 8.997955010224949,
+      "grad_norm": 0.8827540278434753,
+      "learning_rate": 6.0407283697877384e-05,
+      "loss": 3.1596,
+      "step": 83600
+    },
+    {
+      "epoch": 9.003336562264558,
+      "grad_norm": 0.9697520136833191,
+      "learning_rate": 6.0084042667815966e-05,
+      "loss": 3.1432,
+      "step": 83650
+    },
+    {
+      "epoch": 9.008718114304166,
+      "grad_norm": 0.8449240922927856,
+      "learning_rate": 5.976080163775455e-05,
+      "loss": 3.1176,
+      "step": 83700
+    },
+    {
+      "epoch": 9.014099666343773,
+      "grad_norm": 0.9064879417419434,
+      "learning_rate": 5.9437560607693135e-05,
+      "loss": 3.1424,
+      "step": 83750
+    },
+    {
+      "epoch": 9.019481218383381,
+      "grad_norm": 0.8806599378585815,
+      "learning_rate": 5.9114319577631716e-05,
+      "loss": 3.1106,
+      "step": 83800
+    },
+    {
+      "epoch": 9.02486277042299,
+      "grad_norm": 0.849968433380127,
+      "learning_rate": 5.8791078547570304e-05,
+      "loss": 3.1202,
+      "step": 83850
+    },
+    {
+      "epoch": 9.030244322462599,
+      "grad_norm": 0.9165353775024414,
+      "learning_rate": 5.8467837517508885e-05,
+      "loss": 3.1276,
+      "step": 83900
+    },
+    {
+      "epoch": 9.035625874502207,
+      "grad_norm": 0.9038225412368774,
+      "learning_rate": 5.8144596487447466e-05,
+      "loss": 3.122,
+      "step": 83950
+    },
+    {
+      "epoch": 9.041007426541814,
+      "grad_norm": 0.9322003722190857,
+      "learning_rate": 5.7821355457386054e-05,
+      "loss": 3.1178,
+      "step": 84000
+    },
+    {
+      "epoch": 9.041007426541814,
+      "eval_accuracy": 0.3923702421036242,
+      "eval_loss": 3.31853985786438,
+      "eval_runtime": 185.4856,
+      "eval_samples_per_second": 97.102,
+      "eval_steps_per_second": 6.071,
+      "step": 84000
+    },
+    {
+      "epoch": 9.046388978581422,
+      "grad_norm": 0.8468679785728455,
+      "learning_rate": 5.7498114427324635e-05,
+      "loss": 3.1195,
+      "step": 84050
+    },
+    {
+      "epoch": 9.051770530621031,
+      "grad_norm": 0.9400200843811035,
+      "learning_rate": 5.717487339726322e-05,
+      "loss": 3.1386,
+      "step": 84100
+    },
+    {
+      "epoch": 9.05715208266064,
+      "grad_norm": 0.8910375237464905,
+      "learning_rate": 5.6851632367201804e-05,
+      "loss": 3.1324,
+      "step": 84150
+    },
+    {
+      "epoch": 9.062533634700248,
+      "grad_norm": 0.8627359867095947,
+      "learning_rate": 5.653485615774162e-05,
+      "loss": 3.1302,
+      "step": 84200
+    },
+    {
+      "epoch": 9.067915186739857,
+      "grad_norm": 0.8794528245925903,
+      "learning_rate": 5.62116151276802e-05,
+      "loss": 3.1333,
+      "step": 84250
+    },
+    {
+      "epoch": 9.073296738779463,
+      "grad_norm": 0.909424901008606,
+      "learning_rate": 5.5888374097618786e-05,
+      "loss": 3.1135,
+      "step": 84300
+    },
+    {
+      "epoch": 9.078678290819072,
+      "grad_norm": 0.8574796915054321,
+      "learning_rate": 5.556513306755737e-05,
+      "loss": 3.1427,
+      "step": 84350
+    },
+    {
+      "epoch": 9.08405984285868,
+      "grad_norm": 0.8929221034049988,
+      "learning_rate": 5.5241892037495955e-05,
+      "loss": 3.1352,
+      "step": 84400
+    },
+    {
+      "epoch": 9.089441394898289,
+      "grad_norm": 0.9075905084609985,
+      "learning_rate": 5.4918651007434536e-05,
+      "loss": 3.1296,
+      "step": 84450
+    },
+    {
+      "epoch": 9.094822946937898,
+      "grad_norm": 0.8548242449760437,
+      "learning_rate": 5.459540997737312e-05,
+      "loss": 3.1353,
+      "step": 84500
+    },
+    {
+      "epoch": 9.100204498977504,
+      "grad_norm": 0.9040631055831909,
+      "learning_rate": 5.427216894731171e-05,
+      "loss": 3.1112,
+      "step": 84550
+    },
+    {
+      "epoch": 9.105586051017113,
+      "grad_norm": 0.9202734231948853,
+      "learning_rate": 5.394892791725029e-05,
+      "loss": 3.1364,
+      "step": 84600
+    },
+    {
+      "epoch": 9.110967603056721,
+      "grad_norm": 0.8696184754371643,
+      "learning_rate": 5.362568688718888e-05,
+      "loss": 3.1482,
+      "step": 84650
+    },
+    {
+      "epoch": 9.11634915509633,
+      "grad_norm": 0.889745831489563,
+      "learning_rate": 5.330244585712746e-05,
+      "loss": 3.1262,
+      "step": 84700
+    },
+    {
+      "epoch": 9.121730707135939,
+      "grad_norm": 0.8935040235519409,
+      "learning_rate": 5.297920482706604e-05,
+      "loss": 3.1359,
+      "step": 84750
+    },
+    {
+      "epoch": 9.127112259175545,
+      "grad_norm": 0.9295973181724548,
+      "learning_rate": 5.265596379700463e-05,
+      "loss": 3.1164,
+      "step": 84800
+    },
+    {
+      "epoch": 9.132493811215154,
+      "grad_norm": 0.8489879369735718,
+      "learning_rate": 5.233272276694321e-05,
+      "loss": 3.1082,
+      "step": 84850
+    },
+    {
+      "epoch": 9.137875363254762,
+      "grad_norm": 0.8829606771469116,
+      "learning_rate": 5.20094817368818e-05,
+      "loss": 3.1167,
+      "step": 84900
+    },
+    {
+      "epoch": 9.143256915294371,
+      "grad_norm": 0.8823826909065247,
+      "learning_rate": 5.168624070682038e-05,
+      "loss": 3.128,
+      "step": 84950
+    },
+    {
+      "epoch": 9.14863846733398,
+      "grad_norm": 0.8622445464134216,
+      "learning_rate": 5.136299967675896e-05,
+      "loss": 3.138,
+      "step": 85000
+    },
+    {
+      "epoch": 9.14863846733398,
+      "eval_accuracy": 0.39261253804345053,
+      "eval_loss": 3.318570375442505,
+      "eval_runtime": 185.8038,
+      "eval_samples_per_second": 96.936,
+      "eval_steps_per_second": 6.06,
+      "step": 85000
+    },
+    {
+      "epoch": 9.154020019373588,
+      "grad_norm": 0.9152718186378479,
+      "learning_rate": 5.103975864669755e-05,
+      "loss": 3.1287,
+      "step": 85050
+    },
+    {
+      "epoch": 9.159401571413195,
+      "grad_norm": 0.8725242018699646,
+      "learning_rate": 5.071651761663613e-05,
+      "loss": 3.1311,
+      "step": 85100
+    },
+    {
+      "epoch": 9.164783123452803,
+      "grad_norm": 0.9095432758331299,
+      "learning_rate": 5.039327658657472e-05,
+      "loss": 3.1305,
+      "step": 85150
+    },
+    {
+      "epoch": 9.170164675492412,
+      "grad_norm": 0.8616788983345032,
+      "learning_rate": 5.00700355565133e-05,
+      "loss": 3.1236,
+      "step": 85200
+    },
+    {
+      "epoch": 9.17554622753202,
+      "grad_norm": 0.9051676988601685,
+      "learning_rate": 4.974679452645188e-05,
+      "loss": 3.1343,
+      "step": 85250
+    },
+    {
+      "epoch": 9.180927779571629,
+      "grad_norm": 0.9643937349319458,
+      "learning_rate": 4.942355349639047e-05,
+      "loss": 3.1188,
+      "step": 85300
+    },
+    {
+      "epoch": 9.186309331611236,
+      "grad_norm": 0.88005530834198,
+      "learning_rate": 4.910031246632905e-05,
+      "loss": 3.1162,
+      "step": 85350
+    },
+    {
+      "epoch": 9.191690883650844,
+      "grad_norm": 0.9087311029434204,
+      "learning_rate": 4.8777071436267645e-05,
+      "loss": 3.1419,
+      "step": 85400
+    },
+    {
+      "epoch": 9.197072435690453,
+      "grad_norm": 0.8891034126281738,
+      "learning_rate": 4.8453830406206226e-05,
+      "loss": 3.1348,
+      "step": 85450
+    },
+    {
+      "epoch": 9.202453987730062,
+      "grad_norm": 0.8857452869415283,
+      "learning_rate": 4.813058937614481e-05,
+      "loss": 3.1174,
+      "step": 85500
+    },
+    {
+      "epoch": 9.20783553976967,
+      "grad_norm": 0.8889995217323303,
+      "learning_rate": 4.7807348346083395e-05,
+      "loss": 3.1239,
+      "step": 85550
+    },
+    {
+      "epoch": 9.213217091809279,
+      "grad_norm": 0.8715901374816895,
+      "learning_rate": 4.7484107316021976e-05,
+      "loss": 3.1142,
+      "step": 85600
+    },
+    {
+      "epoch": 9.218598643848885,
+      "grad_norm": 0.9088683128356934,
+      "learning_rate": 4.7160866285960564e-05,
+      "loss": 3.13,
+      "step": 85650
+    },
+    {
+      "epoch": 9.223980195888494,
+      "grad_norm": 0.9194056987762451,
+      "learning_rate": 4.6837625255899145e-05,
+      "loss": 3.1327,
+      "step": 85700
+    },
+    {
+      "epoch": 9.229361747928102,
+      "grad_norm": 0.9201943278312683,
+      "learning_rate": 4.6514384225837726e-05,
+      "loss": 3.1354,
+      "step": 85750
+    },
+    {
+      "epoch": 9.234743299967711,
+      "grad_norm": 0.901462197303772,
+      "learning_rate": 4.6191143195776314e-05,
+      "loss": 3.1575,
+      "step": 85800
+    },
+    {
+      "epoch": 9.24012485200732,
+      "grad_norm": 0.9084486365318298,
+      "learning_rate": 4.5867902165714895e-05,
+      "loss": 3.1434,
+      "step": 85850
+    },
+    {
+      "epoch": 9.245506404046926,
+      "grad_norm": 0.8696123957633972,
+      "learning_rate": 4.554466113565348e-05,
+      "loss": 3.1319,
+      "step": 85900
+    },
+    {
+      "epoch": 9.250887956086535,
+      "grad_norm": 0.8794529438018799,
+      "learning_rate": 4.5221420105592064e-05,
+      "loss": 3.1328,
+      "step": 85950
+    },
+    {
+      "epoch": 9.256269508126143,
+      "grad_norm": 0.9295294880867004,
+      "learning_rate": 4.4898179075530645e-05,
+      "loss": 3.1277,
+      "step": 86000
+    },
+    {
+      "epoch": 9.256269508126143,
+      "eval_accuracy": 0.3928442946531588,
+      "eval_loss": 3.3159546852111816,
+      "eval_runtime": 185.3781,
+      "eval_samples_per_second": 97.158,
+      "eval_steps_per_second": 6.074,
+      "step": 86000
+    },
+    {
+      "epoch": 9.261651060165752,
+      "grad_norm": 0.8791679739952087,
+      "learning_rate": 4.457493804546923e-05,
+      "loss": 3.1304,
+      "step": 86050
+    },
+    {
+      "epoch": 9.26703261220536,
+      "grad_norm": 0.8820205926895142,
+      "learning_rate": 4.4251697015407814e-05,
+      "loss": 3.133,
+      "step": 86100
+    },
+    {
+      "epoch": 9.272414164244967,
+      "grad_norm": 0.8805775046348572,
+      "learning_rate": 4.392845598534641e-05,
+      "loss": 3.1279,
+      "step": 86150
+    },
+    {
+      "epoch": 9.277795716284576,
+      "grad_norm": 0.85921710729599,
+      "learning_rate": 4.360521495528499e-05,
+      "loss": 3.1439,
+      "step": 86200
+    },
+    {
+      "epoch": 9.283177268324184,
+      "grad_norm": 0.8640028834342957,
+      "learning_rate": 4.3281973925223564e-05,
+      "loss": 3.1315,
+      "step": 86250
+    },
+    {
+      "epoch": 9.288558820363793,
+      "grad_norm": 0.9068918824195862,
+      "learning_rate": 4.295873289516216e-05,
+      "loss": 3.1313,
+      "step": 86300
+    },
+    {
+      "epoch": 9.293940372403402,
+      "grad_norm": 0.8862238526344299,
+      "learning_rate": 4.264195668570197e-05,
+      "loss": 3.1322,
+      "step": 86350
+    },
+    {
+      "epoch": 9.29932192444301,
+      "grad_norm": 0.8842107057571411,
+      "learning_rate": 4.231871565564055e-05,
+      "loss": 3.1263,
+      "step": 86400
+    },
+    {
+      "epoch": 9.304703476482617,
+      "grad_norm": 0.8852851986885071,
+      "learning_rate": 4.199547462557914e-05,
+      "loss": 3.1363,
+      "step": 86450
+    },
+    {
+      "epoch": 9.310085028522225,
+      "grad_norm": 0.9032968878746033,
+      "learning_rate": 4.167223359551772e-05,
+      "loss": 3.1351,
+      "step": 86500
+    },
+    {
+      "epoch": 9.315466580561834,
+      "grad_norm": 0.8953644633293152,
+      "learning_rate": 4.1348992565456303e-05,
+      "loss": 3.1332,
+      "step": 86550
+    },
+    {
+      "epoch": 9.320848132601443,
+      "grad_norm": 0.9063191413879395,
+      "learning_rate": 4.102575153539489e-05,
+      "loss": 3.1393,
+      "step": 86600
+    },
+    {
+      "epoch": 9.326229684641051,
+      "grad_norm": 0.8734878897666931,
+      "learning_rate": 4.070251050533347e-05,
+      "loss": 3.1192,
+      "step": 86650
+    },
+    {
+      "epoch": 9.331611236680658,
+      "grad_norm": 0.9429718255996704,
+      "learning_rate": 4.037926947527206e-05,
+      "loss": 3.135,
+      "step": 86700
+    },
+    {
+      "epoch": 9.336992788720266,
+      "grad_norm": 0.889334499835968,
+      "learning_rate": 4.005602844521064e-05,
+      "loss": 3.1323,
+      "step": 86750
+    },
+    {
+      "epoch": 9.342374340759875,
+      "grad_norm": 0.9378020167350769,
+      "learning_rate": 3.973278741514922e-05,
+      "loss": 3.1342,
+      "step": 86800
+    },
+    {
+      "epoch": 9.347755892799483,
+      "grad_norm": 0.8967880606651306,
+      "learning_rate": 3.940954638508781e-05,
+      "loss": 3.1498,
+      "step": 86850
+    },
+    {
+      "epoch": 9.353137444839092,
+      "grad_norm": 0.8839611411094666,
+      "learning_rate": 3.908630535502639e-05,
+      "loss": 3.1252,
+      "step": 86900
+    },
+    {
+      "epoch": 9.3585189968787,
+      "grad_norm": 0.9033184051513672,
+      "learning_rate": 3.876306432496498e-05,
+      "loss": 3.145,
+      "step": 86950
+    },
+    {
+      "epoch": 9.363900548918307,
+      "grad_norm": 0.9021019339561462,
+      "learning_rate": 3.843982329490356e-05,
+      "loss": 3.1482,
+      "step": 87000
+    },
+    {
+      "epoch": 9.363900548918307,
+      "eval_accuracy": 0.3931308123183256,
+      "eval_loss": 3.313615083694458,
+      "eval_runtime": 185.5293,
+      "eval_samples_per_second": 97.079,
+      "eval_steps_per_second": 6.069,
+      "step": 87000
+    },
+    {
+      "epoch": 9.369282100957916,
+      "grad_norm": 0.8933983445167542,
+      "learning_rate": 3.811658226484214e-05,
+      "loss": 3.1526,
+      "step": 87050
+    },
+    {
+      "epoch": 9.374663652997524,
+      "grad_norm": 0.8870809078216553,
+      "learning_rate": 3.7793341234780736e-05,
+      "loss": 3.1295,
+      "step": 87100
+    },
+    {
+      "epoch": 9.380045205037133,
+      "grad_norm": 0.9845548868179321,
+      "learning_rate": 3.747010020471931e-05,
+      "loss": 3.1501,
+      "step": 87150
+    },
+    {
+      "epoch": 9.385426757076742,
+      "grad_norm": 0.9062376618385315,
+      "learning_rate": 3.71468591746579e-05,
+      "loss": 3.1186,
+      "step": 87200
+    },
+    {
+      "epoch": 9.390808309116348,
+      "grad_norm": 0.8852116465568542,
+      "learning_rate": 3.6823618144596486e-05,
+      "loss": 3.1221,
+      "step": 87250
+    },
+    {
+      "epoch": 9.396189861155957,
+      "grad_norm": 0.8883943557739258,
+      "learning_rate": 3.650037711453507e-05,
+      "loss": 3.1418,
+      "step": 87300
+    },
+    {
+      "epoch": 9.401571413195565,
+      "grad_norm": 0.8981691598892212,
+      "learning_rate": 3.6177136084473655e-05,
+      "loss": 3.135,
+      "step": 87350
+    },
+    {
+      "epoch": 9.406952965235174,
+      "grad_norm": 0.8964318037033081,
+      "learning_rate": 3.5853895054412236e-05,
+      "loss": 3.142,
+      "step": 87400
+    },
+    {
+      "epoch": 9.412334517274783,
+      "grad_norm": 0.8862378597259521,
+      "learning_rate": 3.553065402435082e-05,
+      "loss": 3.1275,
+      "step": 87450
+    },
+    {
+      "epoch": 9.417716069314391,
+      "grad_norm": 0.9136159420013428,
+      "learning_rate": 3.5207412994289405e-05,
+      "loss": 3.1339,
+      "step": 87500
+    },
+    {
+      "epoch": 9.423097621353998,
+      "grad_norm": 0.8476567268371582,
+      "learning_rate": 3.488417196422799e-05,
+      "loss": 3.1351,
+      "step": 87550
+    },
+    {
+      "epoch": 9.428479173393606,
+      "grad_norm": 0.8648849129676819,
+      "learning_rate": 3.4560930934166574e-05,
+      "loss": 3.1264,
+      "step": 87600
+    },
+    {
+      "epoch": 9.433860725433215,
+      "grad_norm": 0.9079493284225464,
+      "learning_rate": 3.4237689904105156e-05,
+      "loss": 3.1469,
+      "step": 87650
+    },
+    {
+      "epoch": 9.439242277472824,
+      "grad_norm": 0.9165453910827637,
+      "learning_rate": 3.3914448874043743e-05,
+      "loss": 3.1164,
+      "step": 87700
+    },
+    {
+      "epoch": 9.444623829512432,
+      "grad_norm": 0.8907744288444519,
+      "learning_rate": 3.3591207843982325e-05,
+      "loss": 3.1385,
+      "step": 87750
+    },
+    {
+      "epoch": 9.450005381552039,
+      "grad_norm": 0.9528211355209351,
+      "learning_rate": 3.326796681392091e-05,
+      "loss": 3.1527,
+      "step": 87800
+    },
+    {
+      "epoch": 9.455386933591647,
+      "grad_norm": 0.8839420080184937,
+      "learning_rate": 3.2944725783859494e-05,
+      "loss": 3.1181,
+      "step": 87850
+    },
+    {
+      "epoch": 9.460768485631256,
+      "grad_norm": 0.8799306750297546,
+      "learning_rate": 3.2621484753798075e-05,
+      "loss": 3.1246,
+      "step": 87900
+    },
+    {
+      "epoch": 9.466150037670864,
+      "grad_norm": 0.9193063378334045,
+      "learning_rate": 3.229824372373666e-05,
+      "loss": 3.1133,
+      "step": 87950
+    },
+    {
+      "epoch": 9.471531589710473,
+      "grad_norm": 0.8744547367095947,
+      "learning_rate": 3.197500269367525e-05,
+      "loss": 3.1287,
+      "step": 88000
+    },
+    {
+      "epoch": 9.471531589710473,
+      "eval_accuracy": 0.39340396567829117,
+      "eval_loss": 3.3119935989379883,
+      "eval_runtime": 185.5092,
+      "eval_samples_per_second": 97.09,
+      "eval_steps_per_second": 6.07,
+      "step": 88000
+    },
+    {
+      "epoch": 9.476913141750082,
+      "grad_norm": 0.8958976864814758,
+      "learning_rate": 3.165176166361383e-05,
+      "loss": 3.1506,
+      "step": 88050
+    },
+    {
+      "epoch": 9.482294693789688,
+      "grad_norm": 0.9113842248916626,
+      "learning_rate": 3.132852063355242e-05,
+      "loss": 3.1137,
+      "step": 88100
+    },
+    {
+      "epoch": 9.487676245829297,
+      "grad_norm": 0.8699529767036438,
+      "learning_rate": 3.1005279603491e-05,
+      "loss": 3.1334,
+      "step": 88150
+    },
+    {
+      "epoch": 9.493057797868905,
+      "grad_norm": 0.9267193078994751,
+      "learning_rate": 3.068203857342958e-05,
+      "loss": 3.125,
+      "step": 88200
+    },
+    {
+      "epoch": 9.498439349908514,
+      "grad_norm": 0.8742485046386719,
+      "learning_rate": 3.035879754336817e-05,
+      "loss": 3.1351,
+      "step": 88250
+    },
+    {
+      "epoch": 9.503820901948123,
+      "grad_norm": 0.9233673810958862,
+      "learning_rate": 3.0035556513306754e-05,
+      "loss": 3.1264,
+      "step": 88300
+    },
+    {
+      "epoch": 9.50920245398773,
+      "grad_norm": 0.9509491920471191,
+      "learning_rate": 2.971231548324534e-05,
+      "loss": 3.1236,
+      "step": 88350
+    },
+    {
+      "epoch": 9.514584006027338,
+      "grad_norm": 0.8872981071472168,
+      "learning_rate": 2.938907445318392e-05,
+      "loss": 3.1261,
+      "step": 88400
+    },
+    {
+      "epoch": 9.519965558066946,
+      "grad_norm": 0.8559231758117676,
+      "learning_rate": 2.9065833423122504e-05,
+      "loss": 3.1436,
+      "step": 88450
+    },
+    {
+      "epoch": 9.525347110106555,
+      "grad_norm": 0.8599565625190735,
+      "learning_rate": 2.8749057213662317e-05,
+      "loss": 3.1164,
+      "step": 88500
+    },
+    {
+      "epoch": 9.530728662146164,
+      "grad_norm": 0.9169127345085144,
+      "learning_rate": 2.8425816183600902e-05,
+      "loss": 3.1414,
+      "step": 88550
+    },
+    {
+      "epoch": 9.536110214185772,
+      "grad_norm": 0.8855370283126831,
+      "learning_rate": 2.8102575153539486e-05,
+      "loss": 3.1357,
+      "step": 88600
+    },
+    {
+      "epoch": 9.541491766225379,
+      "grad_norm": 0.8835952281951904,
+      "learning_rate": 2.777933412347807e-05,
+      "loss": 3.1505,
+      "step": 88650
+    },
+    {
+      "epoch": 9.546873318264987,
+      "grad_norm": 0.9065445065498352,
+      "learning_rate": 2.7456093093416652e-05,
+      "loss": 3.145,
+      "step": 88700
+    },
+    {
+      "epoch": 9.552254870304596,
+      "grad_norm": 0.9225653409957886,
+      "learning_rate": 2.713285206335524e-05,
+      "loss": 3.1268,
+      "step": 88750
+    },
+    {
+      "epoch": 9.557636422344205,
+      "grad_norm": 0.8876903653144836,
+      "learning_rate": 2.6809611033293824e-05,
+      "loss": 3.1237,
+      "step": 88800
+    },
+    {
+      "epoch": 9.563017974383813,
+      "grad_norm": 0.9166932106018066,
+      "learning_rate": 2.648637000323241e-05,
+      "loss": 3.1285,
+      "step": 88850
+    },
+    {
+      "epoch": 9.56839952642342,
+      "grad_norm": 0.8732258677482605,
+      "learning_rate": 2.6163128973170993e-05,
+      "loss": 3.1483,
+      "step": 88900
+    },
+    {
+      "epoch": 9.573781078463028,
+      "grad_norm": 0.8783851265907288,
+      "learning_rate": 2.5839887943109574e-05,
+      "loss": 3.1164,
+      "step": 88950
+    },
+    {
+      "epoch": 9.579162630502637,
+      "grad_norm": 0.8867563605308533,
+      "learning_rate": 2.551664691304816e-05,
+      "loss": 3.1232,
+      "step": 89000
+    },
+    {
+      "epoch": 9.579162630502637,
+      "eval_accuracy": 0.3936484346758738,
+      "eval_loss": 3.3092377185821533,
+      "eval_runtime": 185.6477,
+      "eval_samples_per_second": 97.017,
+      "eval_steps_per_second": 6.065,
+      "step": 89000
+    },
+    {
+      "epoch": 9.584544182542245,
+      "grad_norm": 0.9191532135009766,
+      "learning_rate": 2.5193405882986743e-05,
+      "loss": 3.1409,
+      "step": 89050
+    },
+    {
+      "epoch": 9.589925734581854,
+      "grad_norm": 0.9038717150688171,
+      "learning_rate": 2.487016485292533e-05,
+      "loss": 3.1408,
+      "step": 89100
+    },
+    {
+      "epoch": 9.59530728662146,
+      "grad_norm": 0.8949322700500488,
+      "learning_rate": 2.4546923822863916e-05,
+      "loss": 3.1347,
+      "step": 89150
+    },
+    {
+      "epoch": 9.60068883866107,
+      "grad_norm": 0.95352703332901,
+      "learning_rate": 2.4223682792802497e-05,
+      "loss": 3.1245,
+      "step": 89200
+    },
+    {
+      "epoch": 9.606070390700678,
+      "grad_norm": 0.8739637136459351,
+      "learning_rate": 2.390044176274108e-05,
+      "loss": 3.1143,
+      "step": 89250
+    },
+    {
+      "epoch": 9.611451942740286,
+      "grad_norm": 0.9231200814247131,
+      "learning_rate": 2.3577200732679666e-05,
+      "loss": 3.1369,
+      "step": 89300
+    },
+    {
+      "epoch": 9.616833494779895,
+      "grad_norm": 0.9198866486549377,
+      "learning_rate": 2.325395970261825e-05,
+      "loss": 3.1352,
+      "step": 89350
+    },
+    {
+      "epoch": 9.622215046819504,
+      "grad_norm": 0.925963819026947,
+      "learning_rate": 2.2930718672556835e-05,
+      "loss": 3.1413,
+      "step": 89400
+    },
+    {
+      "epoch": 9.62759659885911,
+      "grad_norm": 0.8934301137924194,
+      "learning_rate": 2.2607477642495416e-05,
+      "loss": 3.1424,
+      "step": 89450
+    },
+    {
+      "epoch": 9.632978150898719,
+      "grad_norm": 0.8845741748809814,
+      "learning_rate": 2.2284236612434e-05,
+      "loss": 3.1236,
+      "step": 89500
+    },
+    {
+      "epoch": 9.638359702938327,
+      "grad_norm": 0.8709970116615295,
+      "learning_rate": 2.196099558237259e-05,
+      "loss": 3.1353,
+      "step": 89550
+    },
+    {
+      "epoch": 9.643741254977936,
+      "grad_norm": 0.8816275000572205,
+      "learning_rate": 2.1637754552311173e-05,
+      "loss": 3.1203,
+      "step": 89600
+    },
+    {
+      "epoch": 9.649122807017545,
+      "grad_norm": 0.9300759434700012,
+      "learning_rate": 2.1314513522249757e-05,
+      "loss": 3.1219,
+      "step": 89650
+    },
+    {
+      "epoch": 9.654504359057151,
+      "grad_norm": 0.8957176804542542,
+      "learning_rate": 2.099127249218834e-05,
+      "loss": 3.1457,
+      "step": 89700
+    },
+    {
+      "epoch": 9.65988591109676,
+      "grad_norm": 0.898618221282959,
+      "learning_rate": 2.0668031462126923e-05,
+      "loss": 3.1321,
+      "step": 89750
+    },
+    {
+      "epoch": 9.665267463136368,
+      "grad_norm": 0.8680826425552368,
+      "learning_rate": 2.0344790432065507e-05,
+      "loss": 3.1332,
+      "step": 89800
+    },
+    {
+      "epoch": 9.670649015175977,
+      "grad_norm": 0.8869649767875671,
+      "learning_rate": 2.0021549402004092e-05,
+      "loss": 3.1427,
+      "step": 89850
+    },
+    {
+      "epoch": 9.676030567215586,
+      "grad_norm": 0.9026452898979187,
+      "learning_rate": 1.969830837194268e-05,
+      "loss": 3.1494,
+      "step": 89900
+    },
+    {
+      "epoch": 9.681412119255192,
+      "grad_norm": 0.9441297054290771,
+      "learning_rate": 1.9375067341881258e-05,
+      "loss": 3.1428,
+      "step": 89950
+    },
+    {
+      "epoch": 9.6867936712948,
+      "grad_norm": 0.9195820093154907,
+      "learning_rate": 1.9051826311819845e-05,
+      "loss": 3.124,
+      "step": 90000
+    },
+    {
+      "epoch": 9.6867936712948,
+      "eval_accuracy": 0.39382314851947947,
+      "eval_loss": 3.307018518447876,
+      "eval_runtime": 185.4223,
+      "eval_samples_per_second": 97.135,
+      "eval_steps_per_second": 6.073,
+      "step": 90000
+    },
+    {
+      "epoch": 9.69217522333441,
+      "grad_norm": 0.8820453882217407,
+      "learning_rate": 1.872858528175843e-05,
+      "loss": 3.1495,
+      "step": 90050
+    },
+    {
+      "epoch": 9.697556775374018,
+      "grad_norm": 0.8979201912879944,
+      "learning_rate": 1.8405344251697014e-05,
+      "loss": 3.1279,
+      "step": 90100
+    },
+    {
+      "epoch": 9.702938327413626,
+      "grad_norm": 0.8804288506507874,
+      "learning_rate": 1.80821032216356e-05,
+      "loss": 3.1391,
+      "step": 90150
+    },
+    {
+      "epoch": 9.708319879453235,
+      "grad_norm": 0.8932719826698303,
+      "learning_rate": 1.7758862191574183e-05,
+      "loss": 3.1305,
+      "step": 90200
+    },
+    {
+      "epoch": 9.713701431492842,
+      "grad_norm": 0.8876745104789734,
+      "learning_rate": 1.7435621161512768e-05,
+      "loss": 3.1155,
+      "step": 90250
+    },
+    {
+      "epoch": 9.71908298353245,
+      "grad_norm": 0.9139533638954163,
+      "learning_rate": 1.711238013145135e-05,
+      "loss": 3.1393,
+      "step": 90300
+    },
+    {
+      "epoch": 9.724464535572059,
+      "grad_norm": 0.8662781715393066,
+      "learning_rate": 1.6789139101389937e-05,
+      "loss": 3.1385,
+      "step": 90350
+    },
+    {
+      "epoch": 9.729846087611667,
+      "grad_norm": 0.865672767162323,
+      "learning_rate": 1.6465898071328518e-05,
+      "loss": 3.1382,
+      "step": 90400
+    },
+    {
+      "epoch": 9.735227639651276,
+      "grad_norm": 0.86024010181427,
+      "learning_rate": 1.6142657041267102e-05,
+      "loss": 3.1191,
+      "step": 90450
+    },
+    {
+      "epoch": 9.740609191690883,
+      "grad_norm": 0.8847818374633789,
+      "learning_rate": 1.5819416011205687e-05,
+      "loss": 3.1262,
+      "step": 90500
+    },
+    {
+      "epoch": 9.745990743730491,
+      "grad_norm": 0.9043201804161072,
+      "learning_rate": 1.549617498114427e-05,
+      "loss": 3.1345,
+      "step": 90550
+    },
+    {
+      "epoch": 9.7513722957701,
+      "grad_norm": 0.8859184980392456,
+      "learning_rate": 1.5172933951082856e-05,
+      "loss": 3.1337,
+      "step": 90600
+    },
+    {
+      "epoch": 9.756753847809708,
+      "grad_norm": 0.9086697101593018,
+      "learning_rate": 1.4849692921021439e-05,
+      "loss": 3.1363,
+      "step": 90650
+    },
+    {
+      "epoch": 9.762135399849317,
+      "grad_norm": 0.9162412285804749,
+      "learning_rate": 1.4526451890960025e-05,
+      "loss": 3.1585,
+      "step": 90700
+    },
+    {
+      "epoch": 9.767516951888926,
+      "grad_norm": 0.8920602202415466,
+      "learning_rate": 1.4209675681499837e-05,
+      "loss": 3.1296,
+      "step": 90750
+    },
+    {
+      "epoch": 9.772898503928532,
+      "grad_norm": 0.8730294704437256,
+      "learning_rate": 1.3886434651438423e-05,
+      "loss": 3.1243,
+      "step": 90800
+    },
+    {
+      "epoch": 9.77828005596814,
+      "grad_norm": 0.8961200714111328,
+      "learning_rate": 1.3563193621377006e-05,
+      "loss": 3.1312,
+      "step": 90850
+    },
+    {
+      "epoch": 9.78366160800775,
+      "grad_norm": 0.9190976619720459,
+      "learning_rate": 1.323995259131559e-05,
+      "loss": 3.1369,
+      "step": 90900
+    },
+    {
+      "epoch": 9.789043160047358,
+      "grad_norm": 0.845563530921936,
+      "learning_rate": 1.2916711561254173e-05,
+      "loss": 3.1341,
+      "step": 90950
+    },
+    {
+      "epoch": 9.794424712086967,
+      "grad_norm": 0.8940364122390747,
+      "learning_rate": 1.2593470531192759e-05,
+      "loss": 3.1261,
+      "step": 91000
+    },
+    {
+      "epoch": 9.794424712086967,
+      "eval_accuracy": 0.3941041248873677,
+      "eval_loss": 3.304793119430542,
+      "eval_runtime": 185.5725,
+      "eval_samples_per_second": 97.056,
+      "eval_steps_per_second": 6.068,
+      "step": 91000
+    },
+    {
+      "epoch": 9.799806264126573,
+      "grad_norm": 0.9099761247634888,
+      "learning_rate": 1.2270229501131344e-05,
+      "loss": 3.116,
+      "step": 91050
+    },
+    {
+      "epoch": 9.805187816166182,
+      "grad_norm": 0.9464150667190552,
+      "learning_rate": 1.1946988471069926e-05,
+      "loss": 3.1343,
+      "step": 91100
+    },
+    {
+      "epoch": 9.81056936820579,
+      "grad_norm": 0.9240645170211792,
+      "learning_rate": 1.162374744100851e-05,
+      "loss": 3.141,
+      "step": 91150
+    },
+    {
+      "epoch": 9.815950920245399,
+      "grad_norm": 0.8982186317443848,
+      "learning_rate": 1.1300506410947095e-05,
+      "loss": 3.1414,
+      "step": 91200
+    },
+    {
+      "epoch": 9.821332472285007,
+      "grad_norm": 0.8952885270118713,
+      "learning_rate": 1.097726538088568e-05,
+      "loss": 3.1253,
+      "step": 91250
+    },
+    {
+      "epoch": 9.826714024324616,
+      "grad_norm": 0.8702488541603088,
+      "learning_rate": 1.0654024350824264e-05,
+      "loss": 3.1269,
+      "step": 91300
+    },
+    {
+      "epoch": 9.832095576364223,
+      "grad_norm": 0.8995033502578735,
+      "learning_rate": 1.0330783320762847e-05,
+      "loss": 3.1303,
+      "step": 91350
+    },
+    {
+      "epoch": 9.837477128403831,
+      "grad_norm": 0.9203200340270996,
+      "learning_rate": 1.0007542290701433e-05,
+      "loss": 3.132,
+      "step": 91400
+    },
+    {
+      "epoch": 9.84285868044344,
+      "grad_norm": 0.8515362739562988,
+      "learning_rate": 9.684301260640016e-06,
+      "loss": 3.1253,
+      "step": 91450
+    },
+    {
+      "epoch": 9.848240232483048,
+      "grad_norm": 0.8821361064910889,
+      "learning_rate": 9.3610602305786e-06,
+      "loss": 3.1126,
+      "step": 91500
+    },
+    {
+      "epoch": 9.853621784522657,
+      "grad_norm": 0.9096932411193848,
+      "learning_rate": 9.037819200517185e-06,
+      "loss": 3.1399,
+      "step": 91550
+    },
+    {
+      "epoch": 9.859003336562264,
+      "grad_norm": 0.8763339519500732,
+      "learning_rate": 8.71457817045577e-06,
+      "loss": 3.121,
+      "step": 91600
+    },
+    {
+      "epoch": 9.864384888601872,
+      "grad_norm": 0.8977196216583252,
+      "learning_rate": 8.391337140394354e-06,
+      "loss": 3.1172,
+      "step": 91650
+    },
+    {
+      "epoch": 9.869766440641481,
+      "grad_norm": 0.8750319480895996,
+      "learning_rate": 8.068096110332939e-06,
+      "loss": 3.1382,
+      "step": 91700
+    },
+    {
+      "epoch": 9.87514799268109,
+      "grad_norm": 0.8793190121650696,
+      "learning_rate": 7.744855080271521e-06,
+      "loss": 3.13,
+      "step": 91750
+    },
+    {
+      "epoch": 9.880529544720698,
+      "grad_norm": 0.8860962986946106,
+      "learning_rate": 7.421614050210106e-06,
+      "loss": 3.1217,
+      "step": 91800
+    },
+    {
+      "epoch": 9.885911096760307,
+      "grad_norm": 0.8924028277397156,
+      "learning_rate": 7.09837302014869e-06,
+      "loss": 3.1385,
+      "step": 91850
+    },
+    {
+      "epoch": 9.891292648799913,
+      "grad_norm": 0.9022184610366821,
+      "learning_rate": 6.775131990087275e-06,
+      "loss": 3.1347,
+      "step": 91900
+    },
+    {
+      "epoch": 9.896674200839522,
+      "grad_norm": 0.8693097829818726,
+      "learning_rate": 6.451890960025859e-06,
+      "loss": 3.1366,
+      "step": 91950
+    },
+    {
+      "epoch": 9.90205575287913,
+      "grad_norm": 0.9073081016540527,
+      "learning_rate": 6.128649929964443e-06,
+      "loss": 3.1251,
+      "step": 92000
+    },
+    {
+      "epoch": 9.90205575287913,
+      "eval_accuracy": 0.39426732152486504,
+      "eval_loss": 3.3039650917053223,
+      "eval_runtime": 185.5627,
+      "eval_samples_per_second": 97.062,
+      "eval_steps_per_second": 6.068,
+      "step": 92000
+    },
+    {
+      "epoch": 9.907437304918739,
+      "grad_norm": 0.8901609182357788,
+      "learning_rate": 5.805408899903027e-06,
+      "loss": 3.1469,
+      "step": 92050
+    },
+    {
+      "epoch": 9.912818856958348,
+      "grad_norm": 0.9514927268028259,
+      "learning_rate": 5.482167869841611e-06,
+      "loss": 3.127,
+      "step": 92100
+    },
+    {
+      "epoch": 9.918200408997954,
+      "grad_norm": 0.8460099101066589,
+      "learning_rate": 5.1589268397801965e-06,
+      "loss": 3.1446,
+      "step": 92150
+    },
+    {
+      "epoch": 9.923581961037563,
+      "grad_norm": 0.8752443790435791,
+      "learning_rate": 4.83568580971878e-06,
+      "loss": 3.1296,
+      "step": 92200
+    },
+    {
+      "epoch": 9.928963513077171,
+      "grad_norm": 0.8948190212249756,
+      "learning_rate": 4.512444779657364e-06,
+      "loss": 3.1376,
+      "step": 92250
+    },
+    {
+      "epoch": 9.93434506511678,
+      "grad_norm": 0.9106724858283997,
+      "learning_rate": 4.189203749595948e-06,
+      "loss": 3.1204,
+      "step": 92300
+    },
+    {
+      "epoch": 9.939726617156388,
+      "grad_norm": 0.879543662071228,
+      "learning_rate": 3.865962719534533e-06,
+      "loss": 3.1356,
+      "step": 92350
+    },
+    {
+      "epoch": 9.945108169195997,
+      "grad_norm": 0.8651831150054932,
+      "learning_rate": 3.542721689473117e-06,
+      "loss": 3.1342,
+      "step": 92400
+    },
+    {
+      "epoch": 9.950489721235604,
+      "grad_norm": 0.8711269497871399,
+      "learning_rate": 3.2194806594117013e-06,
+      "loss": 3.1307,
+      "step": 92450
+    },
+    {
+      "epoch": 9.955871273275212,
+      "grad_norm": 0.8695480823516846,
+      "learning_rate": 2.8962396293502854e-06,
+      "loss": 3.1078,
+      "step": 92500
+    },
+    {
+      "epoch": 9.961252825314821,
+      "grad_norm": 0.8748586177825928,
+      "learning_rate": 2.5729985992888694e-06,
+      "loss": 3.1396,
+      "step": 92550
+    },
+    {
+      "epoch": 9.96663437735443,
+      "grad_norm": 0.8648610711097717,
+      "learning_rate": 2.249757569227454e-06,
+      "loss": 3.1406,
+      "step": 92600
+    },
+    {
+      "epoch": 9.972015929394038,
+      "grad_norm": 0.9035518765449524,
+      "learning_rate": 1.926516539166038e-06,
+      "loss": 3.1359,
+      "step": 92650
+    },
+    {
+      "epoch": 9.977397481433645,
+      "grad_norm": 0.9193885922431946,
+      "learning_rate": 1.603275509104622e-06,
+      "loss": 3.1331,
+      "step": 92700
+    },
+    {
+      "epoch": 9.982779033473253,
+      "grad_norm": 0.8790634870529175,
+      "learning_rate": 1.2800344790432064e-06,
+      "loss": 3.1437,
+      "step": 92750
+    },
+    {
+      "epoch": 9.988160585512862,
+      "grad_norm": 0.8983802795410156,
+      "learning_rate": 9.567934489817906e-07,
+      "loss": 3.1235,
+      "step": 92800
+    },
+    {
+      "epoch": 9.99354213755247,
+      "grad_norm": 0.8843927979469299,
+      "learning_rate": 6.335524189203748e-07,
+      "loss": 3.1244,
+      "step": 92850
+    },
+    {
+      "epoch": 9.998923689592079,
+      "grad_norm": 0.8855726718902588,
+      "learning_rate": 3.103113888589591e-07,
+      "loss": 3.13,
+      "step": 92900
+    },
+    {
+      "epoch": 10.0,
+      "step": 92910,
+      "total_flos": 7.76821211136e+17,
+      "train_loss": 3.458438355611428,
+      "train_runtime": 80109.3033,
+      "train_samples_per_second": 37.112,
+      "train_steps_per_second": 1.16
+    }
+  ],
+  "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": 7.76821211136e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}