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": 3.3019161224365234,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M_8397/checkpoint-90000",
+  "epoch": 10.0,
+  "eval_steps": 1000,
+  "global_step": 92910,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005381552039608223,
+      "grad_norm": 1.3847299814224243,
+      "learning_rate": 0.0003,
+      "loss": 8.4655,
+      "step": 50
+    },
+    {
+      "epoch": 0.010763104079216447,
+      "grad_norm": 1.0632425546646118,
+      "learning_rate": 0.0006,
+      "loss": 6.8249,
+      "step": 100
+    },
+    {
+      "epoch": 0.01614465611882467,
+      "grad_norm": 2.088347911834717,
+      "learning_rate": 0.0005996767589699385,
+      "loss": 6.4221,
+      "step": 150
+    },
+    {
+      "epoch": 0.021526208158432893,
+      "grad_norm": 1.0710954666137695,
+      "learning_rate": 0.0005993535179398771,
+      "loss": 6.1909,
+      "step": 200
+    },
+    {
+      "epoch": 0.026907760198041114,
+      "grad_norm": 1.5236896276474,
+      "learning_rate": 0.0005990302769098158,
+      "loss": 6.0442,
+      "step": 250
+    },
+    {
+      "epoch": 0.03228931223764934,
+      "grad_norm": 1.8247500658035278,
+      "learning_rate": 0.0005987070358797543,
+      "loss": 5.9513,
+      "step": 300
+    },
+    {
+      "epoch": 0.03767086427725756,
+      "grad_norm": 1.2985302209854126,
+      "learning_rate": 0.0005983837948496929,
+      "loss": 5.8643,
+      "step": 350
+    },
+    {
+      "epoch": 0.04305241631686579,
+      "grad_norm": 1.256401538848877,
+      "learning_rate": 0.0005980605538196314,
+      "loss": 5.8018,
+      "step": 400
+    },
+    {
+      "epoch": 0.048433968356474004,
+      "grad_norm": 0.8569295406341553,
+      "learning_rate": 0.0005977373127895701,
+      "loss": 5.7053,
+      "step": 450
+    },
+    {
+      "epoch": 0.05381552039608223,
+      "grad_norm": 0.860461413860321,
+      "learning_rate": 0.0005974140717595086,
+      "loss": 5.6413,
+      "step": 500
+    },
+    {
+      "epoch": 0.05919707243569045,
+      "grad_norm": 1.6800360679626465,
+      "learning_rate": 0.0005970908307294472,
+      "loss": 5.5789,
+      "step": 550
+    },
+    {
+      "epoch": 0.06457862447529868,
+      "grad_norm": 1.408766508102417,
+      "learning_rate": 0.0005967675896993858,
+      "loss": 5.4957,
+      "step": 600
+    },
+    {
+      "epoch": 0.0699601765149069,
+      "grad_norm": 1.263355016708374,
+      "learning_rate": 0.0005964443486693243,
+      "loss": 5.421,
+      "step": 650
+    },
+    {
+      "epoch": 0.07534172855451512,
+      "grad_norm": 1.617833137512207,
+      "learning_rate": 0.000596121107639263,
+      "loss": 5.334,
+      "step": 700
+    },
+    {
+      "epoch": 0.08072328059412334,
+      "grad_norm": 1.2163816690444946,
+      "learning_rate": 0.0005957978666092015,
+      "loss": 5.2925,
+      "step": 750
+    },
+    {
+      "epoch": 0.08610483263373157,
+      "grad_norm": 1.4054591655731201,
+      "learning_rate": 0.0005954746255791401,
+      "loss": 5.222,
+      "step": 800
+    },
+    {
+      "epoch": 0.09148638467333979,
+      "grad_norm": 1.0674018859863281,
+      "learning_rate": 0.0005951513845490787,
+      "loss": 5.2071,
+      "step": 850
+    },
+    {
+      "epoch": 0.09686793671294801,
+      "grad_norm": 1.4989951848983765,
+      "learning_rate": 0.0005948281435190174,
+      "loss": 5.1526,
+      "step": 900
+    },
+    {
+      "epoch": 0.10224948875255624,
+      "grad_norm": 1.028799057006836,
+      "learning_rate": 0.0005945049024889559,
+      "loss": 5.1098,
+      "step": 950
+    },
+    {
+      "epoch": 0.10763104079216446,
+      "grad_norm": 0.955634355545044,
+      "learning_rate": 0.0005941816614588944,
+      "loss": 5.0633,
+      "step": 1000
+    },
+    {
+      "epoch": 0.10763104079216446,
+      "eval_accuracy": 0.22915491630956988,
+      "eval_loss": 4.996466636657715,
+      "eval_runtime": 202.5064,
+      "eval_samples_per_second": 88.94,
+      "eval_steps_per_second": 5.56,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11301259283177269,
+      "grad_norm": 1.2216582298278809,
+      "learning_rate": 0.000593858420428833,
+      "loss": 5.0219,
+      "step": 1050
+    },
+    {
+      "epoch": 0.1183941448713809,
+      "grad_norm": 1.2919893264770508,
+      "learning_rate": 0.0005935351793987716,
+      "loss": 4.9962,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12377569691098914,
+      "grad_norm": 1.842529058456421,
+      "learning_rate": 0.0005932119383687103,
+      "loss": 4.9736,
+      "step": 1150
+    },
+    {
+      "epoch": 0.12915724895059735,
+      "grad_norm": 0.9416630268096924,
+      "learning_rate": 0.0005928886973386488,
+      "loss": 4.9687,
+      "step": 1200
+    },
+    {
+      "epoch": 0.13453880099020557,
+      "grad_norm": 1.340468406677246,
+      "learning_rate": 0.0005925654563085874,
+      "loss": 4.9251,
+      "step": 1250
+    },
+    {
+      "epoch": 0.1399203530298138,
+      "grad_norm": 1.4057867527008057,
+      "learning_rate": 0.000592242215278526,
+      "loss": 4.8782,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14530190506942203,
+      "grad_norm": 1.3462789058685303,
+      "learning_rate": 0.0005919189742484645,
+      "loss": 4.8431,
+      "step": 1350
+    },
+    {
+      "epoch": 0.15068345710903025,
+      "grad_norm": 0.9291279911994934,
+      "learning_rate": 0.0005915957332184032,
+      "loss": 4.8316,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15606500914863847,
+      "grad_norm": 0.7789126634597778,
+      "learning_rate": 0.0005912724921883417,
+      "loss": 4.8028,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16144656118824668,
+      "grad_norm": 1.2111430168151855,
+      "learning_rate": 0.0005909492511582803,
+      "loss": 4.7837,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1668281132278549,
+      "grad_norm": 1.1703121662139893,
+      "learning_rate": 0.0005906260101282189,
+      "loss": 4.7456,
+      "step": 1550
+    },
+    {
+      "epoch": 0.17220966526746315,
+      "grad_norm": 0.9755436182022095,
+      "learning_rate": 0.0005903027690981575,
+      "loss": 4.7303,
+      "step": 1600
+    },
+    {
+      "epoch": 0.17759121730707136,
+      "grad_norm": 0.9121830463409424,
+      "learning_rate": 0.000589979528068096,
+      "loss": 4.7437,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18297276934667958,
+      "grad_norm": 1.2337779998779297,
+      "learning_rate": 0.0005896562870380347,
+      "loss": 4.6991,
+      "step": 1700
+    },
+    {
+      "epoch": 0.1883543213862878,
+      "grad_norm": 0.7923302054405212,
+      "learning_rate": 0.0005893330460079732,
+      "loss": 4.6819,
+      "step": 1750
+    },
+    {
+      "epoch": 0.19373587342589602,
+      "grad_norm": 0.9414947032928467,
+      "learning_rate": 0.0005890098049779118,
+      "loss": 4.6401,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19911742546550426,
+      "grad_norm": 0.8575401306152344,
+      "learning_rate": 0.0005886865639478504,
+      "loss": 4.6503,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20449897750511248,
+      "grad_norm": 0.874113917350769,
+      "learning_rate": 0.0005883633229177889,
+      "loss": 4.6262,
+      "step": 1900
+    },
+    {
+      "epoch": 0.2098805295447207,
+      "grad_norm": 0.7579442262649536,
+      "learning_rate": 0.0005880400818877276,
+      "loss": 4.5946,
+      "step": 1950
+    },
+    {
+      "epoch": 0.2152620815843289,
+      "grad_norm": 0.8062272071838379,
+      "learning_rate": 0.0005877168408576662,
+      "loss": 4.579,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2152620815843289,
+      "eval_accuracy": 0.2712283567521957,
+      "eval_loss": 4.50425910949707,
+      "eval_runtime": 211.1194,
+      "eval_samples_per_second": 85.312,
+      "eval_steps_per_second": 5.333,
+      "step": 2000
+    },
+    {
+      "epoch": 0.22064363362393713,
+      "grad_norm": 1.1572167873382568,
+      "learning_rate": 0.0005873935998276048,
+      "loss": 4.5564,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22602518566354537,
+      "grad_norm": 1.1494579315185547,
+      "learning_rate": 0.0005870703587975433,
+      "loss": 4.5425,
+      "step": 2100
+    },
+    {
+      "epoch": 0.2314067377031536,
+      "grad_norm": 0.9105241298675537,
+      "learning_rate": 0.0005867471177674818,
+      "loss": 4.5083,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2367882897427618,
+      "grad_norm": 0.9526411890983582,
+      "learning_rate": 0.0005864238767374205,
+      "loss": 4.5128,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24216984178237003,
+      "grad_norm": 0.7726457715034485,
+      "learning_rate": 0.0005861006357073591,
+      "loss": 4.4899,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24755139382197827,
+      "grad_norm": 1.0442702770233154,
+      "learning_rate": 0.0005857773946772977,
+      "loss": 4.4716,
+      "step": 2300
+    },
+    {
+      "epoch": 0.2529329458615865,
+      "grad_norm": 1.1448233127593994,
+      "learning_rate": 0.0005854541536472362,
+      "loss": 4.4607,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2583144979011947,
+      "grad_norm": 0.9515467286109924,
+      "learning_rate": 0.0005851309126171749,
+      "loss": 4.4544,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2636960499408029,
+      "grad_norm": 0.7929104566574097,
+      "learning_rate": 0.0005848076715871134,
+      "loss": 4.4392,
+      "step": 2450
+    },
+    {
+      "epoch": 0.26907760198041114,
+      "grad_norm": 1.1213116645812988,
+      "learning_rate": 0.000584484430557052,
+      "loss": 4.4407,
+      "step": 2500
+    },
+    {
+      "epoch": 0.27445915402001936,
+      "grad_norm": 1.006108045578003,
+      "learning_rate": 0.0005841611895269906,
+      "loss": 4.3997,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2798407060596276,
+      "grad_norm": 0.7756773829460144,
+      "learning_rate": 0.0005838379484969291,
+      "loss": 4.3982,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2852222580992358,
+      "grad_norm": 0.7442255020141602,
+      "learning_rate": 0.0005835147074668678,
+      "loss": 4.3751,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29060381013884407,
+      "grad_norm": 0.7645350694656372,
+      "learning_rate": 0.0005831914664368063,
+      "loss": 4.3873,
+      "step": 2700
+    },
+    {
+      "epoch": 0.2959853621784523,
+      "grad_norm": 0.7028666734695435,
+      "learning_rate": 0.0005828682254067449,
+      "loss": 4.3659,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3013669142180605,
+      "grad_norm": 0.8504071831703186,
+      "learning_rate": 0.0005825449843766835,
+      "loss": 4.3617,
+      "step": 2800
+    },
+    {
+      "epoch": 0.3067484662576687,
+      "grad_norm": 0.880102276802063,
+      "learning_rate": 0.0005822217433466221,
+      "loss": 4.3115,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31213001829727693,
+      "grad_norm": 0.7290977835655212,
+      "learning_rate": 0.0005818985023165607,
+      "loss": 4.3215,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31751157033688515,
+      "grad_norm": 0.8019910454750061,
+      "learning_rate": 0.0005815752612864992,
+      "loss": 4.3149,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32289312237649337,
+      "grad_norm": 0.7739085555076599,
+      "learning_rate": 0.0005812520202564378,
+      "loss": 4.322,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32289312237649337,
+      "eval_accuracy": 0.2992396144952079,
+      "eval_loss": 4.226711750030518,
+      "eval_runtime": 199.2176,
+      "eval_samples_per_second": 90.409,
+      "eval_steps_per_second": 5.652,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3282746744161016,
+      "grad_norm": 1.0869840383529663,
+      "learning_rate": 0.0005809287792263764,
+      "loss": 4.3072,
+      "step": 3050
+    },
+    {
+      "epoch": 0.3336562264557098,
+      "grad_norm": 0.7337630391120911,
+      "learning_rate": 0.0005806055381963151,
+      "loss": 4.2759,
+      "step": 3100
+    },
+    {
+      "epoch": 0.3390377784953181,
+      "grad_norm": 0.7336916923522949,
+      "learning_rate": 0.0005802822971662536,
+      "loss": 4.2838,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3444193305349263,
+      "grad_norm": 0.7901497483253479,
+      "learning_rate": 0.0005799590561361922,
+      "loss": 4.2757,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3498008825745345,
+      "grad_norm": 0.9568474888801575,
+      "learning_rate": 0.0005796358151061307,
+      "loss": 4.2574,
+      "step": 3250
+    },
+    {
+      "epoch": 0.35518243461414273,
+      "grad_norm": 0.9219122529029846,
+      "learning_rate": 0.0005793125740760694,
+      "loss": 4.2423,
+      "step": 3300
+    },
+    {
+      "epoch": 0.36056398665375095,
+      "grad_norm": 0.6552030444145203,
+      "learning_rate": 0.0005789893330460079,
+      "loss": 4.2339,
+      "step": 3350
+    },
+    {
+      "epoch": 0.36594553869335916,
+      "grad_norm": 0.6318998336791992,
+      "learning_rate": 0.0005786660920159465,
+      "loss": 4.2253,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3713270907329674,
+      "grad_norm": 0.7610264420509338,
+      "learning_rate": 0.0005783428509858851,
+      "loss": 4.2369,
+      "step": 3450
+    },
+    {
+      "epoch": 0.3767086427725756,
+      "grad_norm": 0.6418978571891785,
+      "learning_rate": 0.0005780196099558237,
+      "loss": 4.2239,
+      "step": 3500
+    },
+    {
+      "epoch": 0.3820901948121838,
+      "grad_norm": 0.7791884541511536,
+      "learning_rate": 0.0005776963689257623,
+      "loss": 4.2253,
+      "step": 3550
+    },
+    {
+      "epoch": 0.38747174685179203,
+      "grad_norm": 0.7014104127883911,
+      "learning_rate": 0.0005773731278957008,
+      "loss": 4.1888,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3928532988914003,
+      "grad_norm": 0.7691745758056641,
+      "learning_rate": 0.0005770498868656394,
+      "loss": 4.2055,
+      "step": 3650
+    },
+    {
+      "epoch": 0.3982348509310085,
+      "grad_norm": 0.7330165505409241,
+      "learning_rate": 0.000576726645835578,
+      "loss": 4.1917,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40361640297061674,
+      "grad_norm": 0.6409156918525696,
+      "learning_rate": 0.0005764034048055167,
+      "loss": 4.1739,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40899795501022496,
+      "grad_norm": 0.751600444316864,
+      "learning_rate": 0.0005760801637754552,
+      "loss": 4.1815,
+      "step": 3800
+    },
+    {
+      "epoch": 0.4143795070498332,
+      "grad_norm": 0.6341384649276733,
+      "learning_rate": 0.0005757569227453937,
+      "loss": 4.1752,
+      "step": 3850
+    },
+    {
+      "epoch": 0.4197610590894414,
+      "grad_norm": 0.8716601729393005,
+      "learning_rate": 0.0005754336817153324,
+      "loss": 4.1772,
+      "step": 3900
+    },
+    {
+      "epoch": 0.4251426111290496,
+      "grad_norm": 0.7268335223197937,
+      "learning_rate": 0.0005751104406852709,
+      "loss": 4.1579,
+      "step": 3950
+    },
+    {
+      "epoch": 0.4305241631686578,
+      "grad_norm": 0.7490660548210144,
+      "learning_rate": 0.0005747871996552096,
+      "loss": 4.1439,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4305241631686578,
+      "eval_accuracy": 0.31218462820231296,
+      "eval_loss": 4.092499256134033,
+      "eval_runtime": 199.4467,
+      "eval_samples_per_second": 90.305,
+      "eval_steps_per_second": 5.646,
+      "step": 4000
+    },
+    {
+      "epoch": 0.43590571520826604,
+      "grad_norm": 0.9424974918365479,
+      "learning_rate": 0.0005744639586251481,
+      "loss": 4.1592,
+      "step": 4050
+    },
+    {
+      "epoch": 0.44128726724787426,
+      "grad_norm": 0.771391749382019,
+      "learning_rate": 0.0005741407175950867,
+      "loss": 4.1455,
+      "step": 4100
+    },
+    {
+      "epoch": 0.44666881928748253,
+      "grad_norm": 0.7916195392608643,
+      "learning_rate": 0.0005738174765650253,
+      "loss": 4.1329,
+      "step": 4150
+    },
+    {
+      "epoch": 0.45205037132709075,
+      "grad_norm": 0.7015873789787292,
+      "learning_rate": 0.0005734942355349638,
+      "loss": 4.1338,
+      "step": 4200
+    },
+    {
+      "epoch": 0.45743192336669897,
+      "grad_norm": 0.62185138463974,
+      "learning_rate": 0.0005731709945049025,
+      "loss": 4.1517,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4628134754063072,
+      "grad_norm": 0.6499165296554565,
+      "learning_rate": 0.000572847753474841,
+      "loss": 4.1249,
+      "step": 4300
+    },
+    {
+      "epoch": 0.4681950274459154,
+      "grad_norm": 0.8142028450965881,
+      "learning_rate": 0.0005725245124447796,
+      "loss": 4.1094,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4735765794855236,
+      "grad_norm": 0.686265230178833,
+      "learning_rate": 0.0005722012714147182,
+      "loss": 4.1156,
+      "step": 4400
+    },
+    {
+      "epoch": 0.47895813152513184,
+      "grad_norm": 0.6861261129379272,
+      "learning_rate": 0.0005718780303846568,
+      "loss": 4.0911,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48433968356474005,
+      "grad_norm": 0.717899739742279,
+      "learning_rate": 0.0005715547893545953,
+      "loss": 4.0984,
+      "step": 4500
+    },
+    {
+      "epoch": 0.48972123560434827,
+      "grad_norm": 0.635183572769165,
+      "learning_rate": 0.000571231548324534,
+      "loss": 4.0915,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49510278764395654,
+      "grad_norm": 0.6602552533149719,
+      "learning_rate": 0.0005709083072944725,
+      "loss": 4.0886,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5004843396835648,
+      "grad_norm": 0.7243141531944275,
+      "learning_rate": 0.0005705850662644111,
+      "loss": 4.0928,
+      "step": 4650
+    },
+    {
+      "epoch": 0.505865891723173,
+      "grad_norm": 0.6310231685638428,
+      "learning_rate": 0.0005702618252343497,
+      "loss": 4.0986,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5112474437627812,
+      "grad_norm": 0.5868064761161804,
+      "learning_rate": 0.0005699385842042882,
+      "loss": 4.0687,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5166289958023894,
+      "grad_norm": 0.6389384269714355,
+      "learning_rate": 0.0005696153431742269,
+      "loss": 4.0789,
+      "step": 4800
+    },
+    {
+      "epoch": 0.5220105478419976,
+      "grad_norm": 0.7776073217391968,
+      "learning_rate": 0.0005692921021441655,
+      "loss": 4.0977,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5273920998816058,
+      "grad_norm": 0.6234896779060364,
+      "learning_rate": 0.0005689688611141041,
+      "loss": 4.0547,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5327736519212141,
+      "grad_norm": 0.8750380277633667,
+      "learning_rate": 0.0005686456200840426,
+      "loss": 4.0718,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5381552039608223,
+      "grad_norm": 0.7700529098510742,
+      "learning_rate": 0.0005683223790539811,
+      "loss": 4.043,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5381552039608223,
+      "eval_accuracy": 0.3214325100957547,
+      "eval_loss": 3.9926369190216064,
+      "eval_runtime": 225.5059,
+      "eval_samples_per_second": 79.869,
+      "eval_steps_per_second": 4.993,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5435367560004305,
+      "grad_norm": 0.7643857002258301,
+      "learning_rate": 0.0005679991380239198,
+      "loss": 4.0605,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5489183080400387,
+      "grad_norm": 0.554442286491394,
+      "learning_rate": 0.0005676758969938584,
+      "loss": 4.0446,
+      "step": 5100
+    },
+    {
+      "epoch": 0.5542998600796469,
+      "grad_norm": 0.7216308116912842,
+      "learning_rate": 0.000567352655963797,
+      "loss": 4.0313,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5596814121192552,
+      "grad_norm": 0.6057455539703369,
+      "learning_rate": 0.0005670294149337355,
+      "loss": 4.0337,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5650629641588634,
+      "grad_norm": 0.535860002040863,
+      "learning_rate": 0.0005667061739036742,
+      "loss": 4.0392,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5704445161984716,
+      "grad_norm": 0.647204577922821,
+      "learning_rate": 0.0005663829328736127,
+      "loss": 4.0421,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5758260682380799,
+      "grad_norm": 0.5126392841339111,
+      "learning_rate": 0.0005660596918435512,
+      "loss": 4.0318,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5812076202776881,
+      "grad_norm": 0.6040687561035156,
+      "learning_rate": 0.0005657364508134899,
+      "loss": 4.0117,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5865891723172963,
+      "grad_norm": 0.6971921324729919,
+      "learning_rate": 0.0005654132097834284,
+      "loss": 4.025,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5919707243569046,
+      "grad_norm": 0.5343753099441528,
+      "learning_rate": 0.0005650899687533671,
+      "loss": 4.015,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5973522763965128,
+      "grad_norm": 0.5600801706314087,
+      "learning_rate": 0.0005647667277233056,
+      "loss": 4.0087,
+      "step": 5550
+    },
+    {
+      "epoch": 0.602733828436121,
+      "grad_norm": 0.6542683243751526,
+      "learning_rate": 0.0005644434866932442,
+      "loss": 3.9957,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6081153804757292,
+      "grad_norm": 0.6249606609344482,
+      "learning_rate": 0.0005641202456631828,
+      "loss": 4.0136,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.6572969555854797,
+      "learning_rate": 0.0005637970046331214,
+      "loss": 4.0046,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6188784845549457,
+      "grad_norm": 0.6926669478416443,
+      "learning_rate": 0.00056347376360306,
+      "loss": 3.9869,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6242600365945539,
+      "grad_norm": 0.6491366624832153,
+      "learning_rate": 0.0005631505225729985,
+      "loss": 4.0004,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6296415886341621,
+      "grad_norm": 0.6138956546783447,
+      "learning_rate": 0.0005628272815429371,
+      "loss": 3.9814,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6350231406737703,
+      "grad_norm": 0.6467788219451904,
+      "learning_rate": 0.0005625040405128757,
+      "loss": 3.9771,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6404046927133785,
+      "grad_norm": 0.6469590067863464,
+      "learning_rate": 0.0005621807994828143,
+      "loss": 3.9882,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6457862447529867,
+      "grad_norm": 0.7396846413612366,
+      "learning_rate": 0.0005618575584527529,
+      "loss": 3.9759,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6457862447529867,
+      "eval_accuracy": 0.32833664054615025,
+      "eval_loss": 3.919649839401245,
+      "eval_runtime": 200.1239,
+      "eval_samples_per_second": 89.999,
+      "eval_steps_per_second": 5.627,
+      "step": 6000
+    },
+    {
+      "epoch": 0.651167796792595,
+      "grad_norm": 0.7368170619010925,
+      "learning_rate": 0.0005615343174226915,
+      "loss": 3.9789,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6565493488322032,
+      "grad_norm": 0.593928873538971,
+      "learning_rate": 0.00056121107639263,
+      "loss": 3.962,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6619309008718114,
+      "grad_norm": 0.5571395754814148,
+      "learning_rate": 0.0005608878353625687,
+      "loss": 3.9732,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6673124529114196,
+      "grad_norm": 0.6184373497962952,
+      "learning_rate": 0.0005605645943325072,
+      "loss": 3.9618,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6726940049510278,
+      "grad_norm": 0.6731172800064087,
+      "learning_rate": 0.0005602413533024458,
+      "loss": 3.9629,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6780755569906362,
+      "grad_norm": Infinity,
+      "learning_rate": 0.0005599245770929855,
+      "loss": 3.9661,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6834571090302444,
+      "grad_norm": 0.6674894690513611,
+      "learning_rate": 0.0005596013360629242,
+      "loss": 3.9598,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6888386610698526,
+      "grad_norm": 0.6270789504051208,
+      "learning_rate": 0.0005592780950328628,
+      "loss": 3.9443,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6942202131094608,
+      "grad_norm": 0.6163814067840576,
+      "learning_rate": 0.0005589548540028014,
+      "loss": 3.9506,
+      "step": 6450
+    },
+    {
+      "epoch": 0.699601765149069,
+      "grad_norm": 0.6060782670974731,
+      "learning_rate": 0.0005586316129727399,
+      "loss": 3.9645,
+      "step": 6500
+    },
+    {
+      "epoch": 0.7049833171886772,
+      "grad_norm": 0.6246522068977356,
+      "learning_rate": 0.0005583083719426786,
+      "loss": 3.9614,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7103648692282855,
+      "grad_norm": 0.603921115398407,
+      "learning_rate": 0.0005579851309126171,
+      "loss": 3.94,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7157464212678937,
+      "grad_norm": 0.5655505061149597,
+      "learning_rate": 0.0005576618898825558,
+      "loss": 3.9318,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7211279733075019,
+      "grad_norm": 0.604542076587677,
+      "learning_rate": 0.0005573386488524943,
+      "loss": 3.9307,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7265095253471101,
+      "grad_norm": 0.600004255771637,
+      "learning_rate": 0.0005570154078224328,
+      "loss": 3.9442,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7318910773867183,
+      "grad_norm": 0.6960250735282898,
+      "learning_rate": 0.0005566921667923715,
+      "loss": 3.9425,
+      "step": 6800
+    },
+    {
+      "epoch": 0.7372726294263265,
+      "grad_norm": 0.5619109869003296,
+      "learning_rate": 0.00055636892576231,
+      "loss": 3.9073,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7426541814659348,
+      "grad_norm": 0.6036113500595093,
+      "learning_rate": 0.0005560456847322487,
+      "loss": 3.9305,
+      "step": 6900
+    },
+    {
+      "epoch": 0.748035733505543,
+      "grad_norm": 0.6234127283096313,
+      "learning_rate": 0.0005557224437021872,
+      "loss": 3.9285,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7534172855451512,
+      "grad_norm": 0.6131523251533508,
+      "learning_rate": 0.0005553992026721258,
+      "loss": 3.922,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7534172855451512,
+      "eval_accuracy": 0.3334209437785411,
+      "eval_loss": 3.861107349395752,
+      "eval_runtime": 205.1306,
+      "eval_samples_per_second": 87.803,
+      "eval_steps_per_second": 5.489,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7587988375847594,
+      "grad_norm": 0.6058487296104431,
+      "learning_rate": 0.0005550759616420644,
+      "loss": 3.9083,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7641803896243676,
+      "grad_norm": 0.624620258808136,
+      "learning_rate": 0.000554752720612003,
+      "loss": 3.9349,
+      "step": 7100
+    },
+    {
+      "epoch": 0.7695619416639758,
+      "grad_norm": 0.5525732040405273,
+      "learning_rate": 0.0005544294795819415,
+      "loss": 3.9261,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7749434937035841,
+      "grad_norm": 0.5950748324394226,
+      "learning_rate": 0.0005541062385518801,
+      "loss": 3.9067,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7803250457431924,
+      "grad_norm": 0.5052813291549683,
+      "learning_rate": 0.0005537829975218188,
+      "loss": 3.9004,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7857065977828006,
+      "grad_norm": 0.5610913038253784,
+      "learning_rate": 0.0005534662213123586,
+      "loss": 3.8911,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7910881498224088,
+      "grad_norm": 0.5639758110046387,
+      "learning_rate": 0.0005531429802822971,
+      "loss": 3.931,
+      "step": 7350
+    },
+    {
+      "epoch": 0.796469701862017,
+      "grad_norm": 0.6221727728843689,
+      "learning_rate": 0.0005528197392522357,
+      "loss": 3.9288,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8018512539016253,
+      "grad_norm": 0.5807225108146667,
+      "learning_rate": 0.0005524964982221743,
+      "loss": 3.9206,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8072328059412335,
+      "grad_norm": 0.6214258074760437,
+      "learning_rate": 0.0005521732571921129,
+      "loss": 3.9216,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8126143579808417,
+      "grad_norm": 0.6088549494743347,
+      "learning_rate": 0.0005518500161620514,
+      "loss": 3.9102,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8179959100204499,
+      "grad_norm": 0.615315318107605,
+      "learning_rate": 0.00055152677513199,
+      "loss": 3.8932,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8233774620600581,
+      "grad_norm": 0.5932325124740601,
+      "learning_rate": 0.0005512035341019286,
+      "loss": 3.9037,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8287590140996663,
+      "grad_norm": 0.632127583026886,
+      "learning_rate": 0.0005508802930718672,
+      "loss": 3.9074,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8341405661392746,
+      "grad_norm": 0.5906286239624023,
+      "learning_rate": 0.0005505570520418058,
+      "loss": 3.8969,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8395221181788828,
+      "grad_norm": 0.7419958114624023,
+      "learning_rate": 0.0005502338110117443,
+      "loss": 3.911,
+      "step": 7800
+    },
+    {
+      "epoch": 0.844903670218491,
+      "grad_norm": 0.5959650874137878,
+      "learning_rate": 0.000549910569981683,
+      "loss": 3.885,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8502852222580992,
+      "grad_norm": 0.5370450615882874,
+      "learning_rate": 0.0005495873289516215,
+      "loss": 3.8883,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8556667742977074,
+      "grad_norm": 0.6142526865005493,
+      "learning_rate": 0.0005492640879215602,
+      "loss": 3.8891,
+      "step": 7950
+    },
+    {
+      "epoch": 0.8610483263373157,
+      "grad_norm": 0.5250054001808167,
+      "learning_rate": 0.0005489408468914987,
+      "loss": 3.8731,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8610483263373157,
+      "eval_accuracy": 0.3380041398923315,
+      "eval_loss": 3.814573049545288,
+      "eval_runtime": 218.7029,
+      "eval_samples_per_second": 82.354,
+      "eval_steps_per_second": 5.149,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8664298783769239,
+      "grad_norm": 0.5592741370201111,
+      "learning_rate": 0.0005486176058614372,
+      "loss": 3.8661,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8718114304165321,
+      "grad_norm": 0.558358371257782,
+      "learning_rate": 0.0005482943648313759,
+      "loss": 3.8665,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8771929824561403,
+      "grad_norm": 0.6065065860748291,
+      "learning_rate": 0.0005479711238013145,
+      "loss": 3.8815,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8825745344957485,
+      "grad_norm": 0.566906750202179,
+      "learning_rate": 0.0005476478827712531,
+      "loss": 3.8788,
+      "step": 8200
+    },
+    {
+      "epoch": 0.8879560865353568,
+      "grad_norm": 0.5542231202125549,
+      "learning_rate": 0.0005473246417411916,
+      "loss": 3.8757,
+      "step": 8250
+    },
+    {
+      "epoch": 0.8933376385749651,
+      "grad_norm": 0.5650736093521118,
+      "learning_rate": 0.0005470014007111302,
+      "loss": 3.867,
+      "step": 8300
+    },
+    {
+      "epoch": 0.8987191906145733,
+      "grad_norm": 0.6359168887138367,
+      "learning_rate": 0.0005466781596810688,
+      "loss": 3.8704,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9041007426541815,
+      "grad_norm": 0.5609973669052124,
+      "learning_rate": 0.0005463549186510073,
+      "loss": 3.8704,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9094822946937897,
+      "grad_norm": 0.5966047048568726,
+      "learning_rate": 0.000546031677620946,
+      "loss": 3.8649,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9148638467333979,
+      "grad_norm": 0.555415689945221,
+      "learning_rate": 0.0005457084365908845,
+      "loss": 3.8775,
+      "step": 8500
+    },
+    {
+      "epoch": 0.9202453987730062,
+      "grad_norm": 0.5719990134239197,
+      "learning_rate": 0.0005453851955608232,
+      "loss": 3.8601,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9256269508126144,
+      "grad_norm": 0.5464230179786682,
+      "learning_rate": 0.0005450619545307617,
+      "loss": 3.8599,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9310085028522226,
+      "grad_norm": 0.6069023013114929,
+      "learning_rate": 0.0005447387135007003,
+      "loss": 3.8584,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9363900548918308,
+      "grad_norm": 0.556610643863678,
+      "learning_rate": 0.0005444154724706389,
+      "loss": 3.8636,
+      "step": 8700
+    },
+    {
+      "epoch": 0.941771606931439,
+      "grad_norm": 0.5988168120384216,
+      "learning_rate": 0.0005440922314405775,
+      "loss": 3.8366,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9471531589710472,
+      "grad_norm": 0.5634347200393677,
+      "learning_rate": 0.0005437689904105161,
+      "loss": 3.8604,
+      "step": 8800
+    },
+    {
+      "epoch": 0.9525347110106555,
+      "grad_norm": 0.5349180102348328,
+      "learning_rate": 0.0005434457493804546,
+      "loss": 3.8395,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9579162630502637,
+      "grad_norm": 0.5537461042404175,
+      "learning_rate": 0.0005431225083503932,
+      "loss": 3.8644,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9632978150898719,
+      "grad_norm": 0.5711461901664734,
+      "learning_rate": 0.0005427992673203318,
+      "loss": 3.8352,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9686793671294801,
+      "grad_norm": 0.6128414273262024,
+      "learning_rate": 0.0005424760262902704,
+      "loss": 3.8405,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9686793671294801,
+      "eval_accuracy": 0.3417073562677125,
+      "eval_loss": 3.779106378555298,
+      "eval_runtime": 210.7209,
+      "eval_samples_per_second": 85.473,
+      "eval_steps_per_second": 5.344,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9740609191690883,
+      "grad_norm": 0.6156434416770935,
+      "learning_rate": 0.000542152785260209,
+      "loss": 3.8333,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9794424712086965,
+      "grad_norm": 0.5652556419372559,
+      "learning_rate": 0.0005418295442301476,
+      "loss": 3.8327,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9848240232483048,
+      "grad_norm": 0.6443074941635132,
+      "learning_rate": 0.0005415063032000861,
+      "loss": 3.842,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9902055752879131,
+      "grad_norm": 0.632297158241272,
+      "learning_rate": 0.0005411830621700248,
+      "loss": 3.8312,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9955871273275213,
+      "grad_norm": 0.5863285064697266,
+      "learning_rate": 0.0005408598211399633,
+      "loss": 3.8298,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0009686793671295,
+      "grad_norm": 0.5393086671829224,
+      "learning_rate": 0.0005405365801099019,
+      "loss": 3.8069,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0063502314067376,
+      "grad_norm": 0.5285312533378601,
+      "learning_rate": 0.0005402133390798405,
+      "loss": 3.7619,
+      "step": 9350
+    },
+    {
+      "epoch": 1.011731783446346,
+      "grad_norm": 0.5921708941459656,
+      "learning_rate": 0.000539890098049779,
+      "loss": 3.7701,
+      "step": 9400
+    },
+    {
+      "epoch": 1.017113335485954,
+      "grad_norm": 0.5384525060653687,
+      "learning_rate": 0.0005395668570197177,
+      "loss": 3.7657,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0224948875255624,
+      "grad_norm": 0.5624929666519165,
+      "learning_rate": 0.0005392436159896562,
+      "loss": 3.7754,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0278764395651705,
+      "grad_norm": 0.5882208943367004,
+      "learning_rate": 0.0005389203749595948,
+      "loss": 3.7848,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0332579916047788,
+      "grad_norm": 0.5588924884796143,
+      "learning_rate": 0.0005385971339295334,
+      "loss": 3.7603,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0386395436443872,
+      "grad_norm": 0.5837873220443726,
+      "learning_rate": 0.000538273892899472,
+      "loss": 3.7785,
+      "step": 9650
+    },
+    {
+      "epoch": 1.0440210956839953,
+      "grad_norm": 0.5880899429321289,
+      "learning_rate": 0.0005379506518694106,
+      "loss": 3.7506,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0494026477236036,
+      "grad_norm": 0.5579642653465271,
+      "learning_rate": 0.0005376274108393491,
+      "loss": 3.7714,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0547841997632117,
+      "grad_norm": 0.576766312122345,
+      "learning_rate": 0.0005373041698092877,
+      "loss": 3.7653,
+      "step": 9800
+    },
+    {
+      "epoch": 1.06016575180282,
+      "grad_norm": 0.5471925735473633,
+      "learning_rate": 0.0005369809287792263,
+      "loss": 3.7686,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0655473038424281,
+      "grad_norm": 0.59092116355896,
+      "learning_rate": 0.000536657687749165,
+      "loss": 3.7667,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0709288558820365,
+      "grad_norm": 0.638380765914917,
+      "learning_rate": 0.0005363344467191035,
+      "loss": 3.7722,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0763104079216446,
+      "grad_norm": 0.5594536066055298,
+      "learning_rate": 0.000536011205689042,
+      "loss": 3.7655,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0763104079216446,
+      "eval_accuracy": 0.3451049320696713,
+      "eval_loss": 3.7489354610443115,
+      "eval_runtime": 202.8767,
+      "eval_samples_per_second": 88.778,
+      "eval_steps_per_second": 5.55,
+      "step": 10000
+    },
+    {
+      "epoch": 1.081691959961253,
+      "grad_norm": 0.640766441822052,
+      "learning_rate": 0.0005356879646589807,
+      "loss": 3.7646,
+      "step": 10050
+    },
+    {
+      "epoch": 1.087073512000861,
+      "grad_norm": 0.6311773657798767,
+      "learning_rate": 0.0005353647236289192,
+      "loss": 3.7643,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0924550640404693,
+      "grad_norm": 0.5654868483543396,
+      "learning_rate": 0.0005350414825988579,
+      "loss": 3.7592,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0978366160800774,
+      "grad_norm": 0.5341681838035583,
+      "learning_rate": 0.0005347182415687964,
+      "loss": 3.7609,
+      "step": 10200
+    },
+    {
+      "epoch": 1.1032181681196858,
+      "grad_norm": 0.8984493613243103,
+      "learning_rate": 0.000534395000538735,
+      "loss": 3.7588,
+      "step": 10250
+    },
+    {
+      "epoch": 1.1085997201592939,
+      "grad_norm": 0.6130673885345459,
+      "learning_rate": 0.0005340717595086736,
+      "loss": 3.7725,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1139812721989022,
+      "grad_norm": 0.5171802043914795,
+      "learning_rate": 0.0005337485184786122,
+      "loss": 3.7686,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1193628242385103,
+      "grad_norm": 0.6430924534797668,
+      "learning_rate": 0.0005334252774485507,
+      "loss": 3.7493,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1247443762781186,
+      "grad_norm": 0.5376786589622498,
+      "learning_rate": 0.0005331020364184894,
+      "loss": 3.7686,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1301259283177267,
+      "grad_norm": 0.5798326730728149,
+      "learning_rate": 0.0005327787953884279,
+      "loss": 3.7654,
+      "step": 10500
+    },
+    {
+      "epoch": 1.135507480357335,
+      "grad_norm": 0.5384038090705872,
+      "learning_rate": 0.0005324555543583665,
+      "loss": 3.7485,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1408890323969434,
+      "grad_norm": 0.5550758838653564,
+      "learning_rate": 0.0005321323133283051,
+      "loss": 3.748,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1462705844365515,
+      "grad_norm": 0.5656107664108276,
+      "learning_rate": 0.0005318090722982436,
+      "loss": 3.7485,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1516521364761596,
+      "grad_norm": 0.5888227224349976,
+      "learning_rate": 0.0005314858312681823,
+      "loss": 3.7438,
+      "step": 10700
+    },
+    {
+      "epoch": 1.157033688515768,
+      "grad_norm": 0.5515899658203125,
+      "learning_rate": 0.0005311625902381209,
+      "loss": 3.7721,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1624152405553763,
+      "grad_norm": 0.6210424900054932,
+      "learning_rate": 0.0005308393492080595,
+      "loss": 3.7652,
+      "step": 10800
+    },
+    {
+      "epoch": 1.1677967925949844,
+      "grad_norm": 0.5752713084220886,
+      "learning_rate": 0.000530516108177998,
+      "loss": 3.7486,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1731783446345927,
+      "grad_norm": 0.6572223901748657,
+      "learning_rate": 0.0005301928671479365,
+      "loss": 3.7272,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1785598966742008,
+      "grad_norm": 0.6310757994651794,
+      "learning_rate": 0.0005298696261178752,
+      "loss": 3.7563,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1839414487138091,
+      "grad_norm": 0.5674658417701721,
+      "learning_rate": 0.0005295463850878138,
+      "loss": 3.7441,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1839414487138091,
+      "eval_accuracy": 0.34708784727228553,
+      "eval_loss": 3.7200334072113037,
+      "eval_runtime": 208.4996,
+      "eval_samples_per_second": 86.384,
+      "eval_steps_per_second": 5.4,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1893230007534172,
+      "grad_norm": 0.5541161298751831,
+      "learning_rate": 0.0005292231440577524,
+      "loss": 3.7499,
+      "step": 11050
+    },
+    {
+      "epoch": 1.1947045527930256,
+      "grad_norm": 0.5533831715583801,
+      "learning_rate": 0.0005288999030276909,
+      "loss": 3.7418,
+      "step": 11100
+    },
+    {
+      "epoch": 1.2000861048326337,
+      "grad_norm": 0.5780303478240967,
+      "learning_rate": 0.0005285766619976295,
+      "loss": 3.7409,
+      "step": 11150
+    },
+    {
+      "epoch": 1.205467656872242,
+      "grad_norm": 0.6292614340782166,
+      "learning_rate": 0.0005282534209675681,
+      "loss": 3.7435,
+      "step": 11200
+    },
+    {
+      "epoch": 1.21084920891185,
+      "grad_norm": 0.5940732359886169,
+      "learning_rate": 0.0005279301799375066,
+      "loss": 3.7362,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2162307609514584,
+      "grad_norm": 0.5676620006561279,
+      "learning_rate": 0.0005276134037280465,
+      "loss": 3.7241,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2216123129910665,
+      "grad_norm": 0.5294714570045471,
+      "learning_rate": 0.0005272901626979851,
+      "loss": 3.7192,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2269938650306749,
+      "grad_norm": 0.5521119832992554,
+      "learning_rate": 0.0005269669216679236,
+      "loss": 3.734,
+      "step": 11400
+    },
+    {
+      "epoch": 1.232375417070283,
+      "grad_norm": 0.7975606918334961,
+      "learning_rate": 0.0005266436806378623,
+      "loss": 3.7188,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2377569691098913,
+      "grad_norm": 0.5794736742973328,
+      "learning_rate": 0.0005263204396078008,
+      "loss": 3.7279,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2431385211494996,
+      "grad_norm": 0.5361841917037964,
+      "learning_rate": 0.0005259971985777394,
+      "loss": 3.7432,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2485200731891077,
+      "grad_norm": 0.5248964428901672,
+      "learning_rate": 0.000525673957547678,
+      "loss": 3.7447,
+      "step": 11600
+    },
+    {
+      "epoch": 1.2539016252287158,
+      "grad_norm": 0.5553768873214722,
+      "learning_rate": 0.0005253507165176167,
+      "loss": 3.7477,
+      "step": 11650
+    },
+    {
+      "epoch": 1.2592831772683242,
+      "grad_norm": 0.5761224627494812,
+      "learning_rate": 0.0005250274754875552,
+      "loss": 3.7211,
+      "step": 11700
+    },
+    {
+      "epoch": 1.2646647293079325,
+      "grad_norm": 0.607130229473114,
+      "learning_rate": 0.0005247042344574938,
+      "loss": 3.7359,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2700462813475406,
+      "grad_norm": 0.5540531873703003,
+      "learning_rate": 0.0005243809934274323,
+      "loss": 3.7217,
+      "step": 11800
+    },
+    {
+      "epoch": 1.275427833387149,
+      "grad_norm": 0.6098884344100952,
+      "learning_rate": 0.0005240577523973709,
+      "loss": 3.7431,
+      "step": 11850
+    },
+    {
+      "epoch": 1.280809385426757,
+      "grad_norm": 0.5816884636878967,
+      "learning_rate": 0.0005237345113673095,
+      "loss": 3.7404,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2861909374663654,
+      "grad_norm": 0.5769429802894592,
+      "learning_rate": 0.0005234112703372481,
+      "loss": 3.7404,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2915724895059735,
+      "grad_norm": 0.563298761844635,
+      "learning_rate": 0.0005230880293071867,
+      "loss": 3.7293,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2915724895059735,
+      "eval_accuracy": 0.34958697234490643,
+      "eval_loss": 3.6987245082855225,
+      "eval_runtime": 217.4267,
+      "eval_samples_per_second": 82.837,
+      "eval_steps_per_second": 5.179,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2969540415455818,
+      "grad_norm": 0.5788484811782837,
+      "learning_rate": 0.0005227647882771253,
+      "loss": 3.7336,
+      "step": 12050
+    },
+    {
+      "epoch": 1.30233559358519,
+      "grad_norm": 0.6204023957252502,
+      "learning_rate": 0.0005224415472470639,
+      "loss": 3.7243,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3077171456247982,
+      "grad_norm": 0.5986481308937073,
+      "learning_rate": 0.0005221183062170024,
+      "loss": 3.7316,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3130986976644063,
+      "grad_norm": 0.6356789469718933,
+      "learning_rate": 0.0005217950651869409,
+      "loss": 3.7352,
+      "step": 12200
+    },
+    {
+      "epoch": 1.3184802497040147,
+      "grad_norm": 0.555164098739624,
+      "learning_rate": 0.0005214718241568796,
+      "loss": 3.7149,
+      "step": 12250
+    },
+    {
+      "epoch": 1.3238618017436228,
+      "grad_norm": 0.6156308054924011,
+      "learning_rate": 0.0005211485831268182,
+      "loss": 3.6991,
+      "step": 12300
+    },
+    {
+      "epoch": 1.329243353783231,
+      "grad_norm": 0.5790920853614807,
+      "learning_rate": 0.0005208253420967568,
+      "loss": 3.7206,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3346249058228392,
+      "grad_norm": 0.5524982810020447,
+      "learning_rate": 0.0005205021010666953,
+      "loss": 3.7344,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3400064578624475,
+      "grad_norm": 0.5789136290550232,
+      "learning_rate": 0.0005201788600366339,
+      "loss": 3.714,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3453880099020559,
+      "grad_norm": 0.5845145583152771,
+      "learning_rate": 0.0005198556190065725,
+      "loss": 3.7083,
+      "step": 12500
+    },
+    {
+      "epoch": 1.350769561941664,
+      "grad_norm": 0.6391981840133667,
+      "learning_rate": 0.0005195323779765112,
+      "loss": 3.7103,
+      "step": 12550
+    },
+    {
+      "epoch": 1.356151113981272,
+      "grad_norm": 0.543088972568512,
+      "learning_rate": 0.0005192091369464497,
+      "loss": 3.7256,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3615326660208804,
+      "grad_norm": 0.5750318169593811,
+      "learning_rate": 0.0005188858959163882,
+      "loss": 3.7289,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3669142180604887,
+      "grad_norm": 0.6135967969894409,
+      "learning_rate": 0.0005185626548863269,
+      "loss": 3.7014,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3722957701000968,
+      "grad_norm": 0.5833747982978821,
+      "learning_rate": 0.0005182394138562654,
+      "loss": 3.7292,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3776773221397052,
+      "grad_norm": 0.571729838848114,
+      "learning_rate": 0.0005179161728262041,
+      "loss": 3.7158,
+      "step": 12800
+    },
+    {
+      "epoch": 1.3830588741793133,
+      "grad_norm": 0.6265289783477783,
+      "learning_rate": 0.0005175929317961426,
+      "loss": 3.7261,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3884404262189216,
+      "grad_norm": 0.6207230091094971,
+      "learning_rate": 0.0005172696907660812,
+      "loss": 3.699,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3938219782585297,
+      "grad_norm": 0.5325709581375122,
+      "learning_rate": 0.0005169464497360198,
+      "loss": 3.7059,
+      "step": 12950
+    },
+    {
+      "epoch": 1.399203530298138,
+      "grad_norm": 0.5781683325767517,
+      "learning_rate": 0.0005166232087059583,
+      "loss": 3.7029,
+      "step": 13000
+    },
+    {
+      "epoch": 1.399203530298138,
+      "eval_accuracy": 0.3522667871699628,
+      "eval_loss": 3.674454927444458,
+      "eval_runtime": 217.915,
+      "eval_samples_per_second": 82.651,
+      "eval_steps_per_second": 5.167,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4045850823377461,
+      "grad_norm": 0.565986156463623,
+      "learning_rate": 0.0005162999676758969,
+      "loss": 3.7075,
+      "step": 13050
+    },
+    {
+      "epoch": 1.4099666343773545,
+      "grad_norm": 0.5653162002563477,
+      "learning_rate": 0.0005159767266458355,
+      "loss": 3.6983,
+      "step": 13100
+    },
+    {
+      "epoch": 1.4153481864169626,
+      "grad_norm": 0.5863538384437561,
+      "learning_rate": 0.0005156534856157741,
+      "loss": 3.7083,
+      "step": 13150
+    },
+    {
+      "epoch": 1.420729738456571,
+      "grad_norm": 0.5486993789672852,
+      "learning_rate": 0.0005153302445857127,
+      "loss": 3.7251,
+      "step": 13200
+    },
+    {
+      "epoch": 1.426111290496179,
+      "grad_norm": 0.6212597489356995,
+      "learning_rate": 0.0005150070035556513,
+      "loss": 3.7196,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4314928425357873,
+      "grad_norm": 0.6025354862213135,
+      "learning_rate": 0.0005146837625255898,
+      "loss": 3.7195,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4368743945753955,
+      "grad_norm": 0.5640348792076111,
+      "learning_rate": 0.0005143669863161297,
+      "loss": 3.7035,
+      "step": 13350
+    },
+    {
+      "epoch": 1.4422559466150038,
+      "grad_norm": 0.5984680652618408,
+      "learning_rate": 0.0005140437452860683,
+      "loss": 3.7052,
+      "step": 13400
+    },
+    {
+      "epoch": 1.447637498654612,
+      "grad_norm": 0.5823872685432434,
+      "learning_rate": 0.0005137205042560069,
+      "loss": 3.695,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4530190506942202,
+      "grad_norm": 0.5364094972610474,
+      "learning_rate": 0.0005133972632259455,
+      "loss": 3.7199,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4584006027338283,
+      "grad_norm": 0.5685153603553772,
+      "learning_rate": 0.000513074022195884,
+      "loss": 3.6992,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4637821547734367,
+      "grad_norm": 0.5787658095359802,
+      "learning_rate": 0.0005127507811658226,
+      "loss": 3.6923,
+      "step": 13600
+    },
+    {
+      "epoch": 1.469163706813045,
+      "grad_norm": 0.5486370325088501,
+      "learning_rate": 0.0005124275401357612,
+      "loss": 3.7024,
+      "step": 13650
+    },
+    {
+      "epoch": 1.474545258852653,
+      "grad_norm": 0.5473746061325073,
+      "learning_rate": 0.0005121042991056997,
+      "loss": 3.6833,
+      "step": 13700
+    },
+    {
+      "epoch": 1.4799268108922612,
+      "grad_norm": 0.5517615675926208,
+      "learning_rate": 0.0005117810580756384,
+      "loss": 3.6902,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4853083629318695,
+      "grad_norm": 0.5971811413764954,
+      "learning_rate": 0.0005114578170455769,
+      "loss": 3.6845,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4906899149714778,
+      "grad_norm": 0.5672309398651123,
+      "learning_rate": 0.0005111345760155156,
+      "loss": 3.6854,
+      "step": 13850
+    },
+    {
+      "epoch": 1.496071467011086,
+      "grad_norm": 0.5523454546928406,
+      "learning_rate": 0.0005108113349854541,
+      "loss": 3.6961,
+      "step": 13900
+    },
+    {
+      "epoch": 1.501453019050694,
+      "grad_norm": 0.5691514611244202,
+      "learning_rate": 0.0005104880939553926,
+      "loss": 3.6803,
+      "step": 13950
+    },
+    {
+      "epoch": 1.5068345710903024,
+      "grad_norm": 0.5720673203468323,
+      "learning_rate": 0.0005101648529253313,
+      "loss": 3.6962,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5068345710903024,
+      "eval_accuracy": 0.35364048563060124,
+      "eval_loss": 3.655740261077881,
+      "eval_runtime": 204.9718,
+      "eval_samples_per_second": 87.871,
+      "eval_steps_per_second": 5.493,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5122161231299107,
+      "grad_norm": 0.5242102742195129,
+      "learning_rate": 0.0005098480767158711,
+      "loss": 3.697,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5175976751695188,
+      "grad_norm": 0.5905733108520508,
+      "learning_rate": 0.0005095248356858097,
+      "loss": 3.6869,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5229792272091272,
+      "grad_norm": 0.6295709609985352,
+      "learning_rate": 0.0005092015946557483,
+      "loss": 3.6839,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5283607792487355,
+      "grad_norm": 0.6163989901542664,
+      "learning_rate": 0.0005088783536256868,
+      "loss": 3.6979,
+      "step": 14200
+    },
+    {
+      "epoch": 1.5337423312883436,
+      "grad_norm": 0.5677395462989807,
+      "learning_rate": 0.0005085551125956255,
+      "loss": 3.698,
+      "step": 14250
+    },
+    {
+      "epoch": 1.5391238833279517,
+      "grad_norm": 0.5294774174690247,
+      "learning_rate": 0.000508231871565564,
+      "loss": 3.6858,
+      "step": 14300
+    },
+    {
+      "epoch": 1.54450543536756,
+      "grad_norm": 0.585757315158844,
+      "learning_rate": 0.0005079086305355026,
+      "loss": 3.7028,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5498869874071683,
+      "grad_norm": 0.5384769439697266,
+      "learning_rate": 0.0005075853895054412,
+      "loss": 3.686,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5552685394467765,
+      "grad_norm": 0.5806359052658081,
+      "learning_rate": 0.0005072621484753797,
+      "loss": 3.689,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5606500914863846,
+      "grad_norm": 0.5702711939811707,
+      "learning_rate": 0.0005069389074453184,
+      "loss": 3.675,
+      "step": 14500
+    },
+    {
+      "epoch": 1.566031643525993,
+      "grad_norm": 0.5366608500480652,
+      "learning_rate": 0.0005066156664152569,
+      "loss": 3.6914,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5714131955656012,
+      "grad_norm": 0.60133296251297,
+      "learning_rate": 0.0005062924253851955,
+      "loss": 3.693,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5767947476052093,
+      "grad_norm": 0.5777215361595154,
+      "learning_rate": 0.0005059691843551341,
+      "loss": 3.6786,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5821762996448174,
+      "grad_norm": 0.5946151614189148,
+      "learning_rate": 0.0005056459433250727,
+      "loss": 3.6849,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5875578516844258,
+      "grad_norm": 0.601273775100708,
+      "learning_rate": 0.0005053227022950113,
+      "loss": 3.6833,
+      "step": 14750
+    },
+    {
+      "epoch": 1.592939403724034,
+      "grad_norm": 0.5582096576690674,
+      "learning_rate": 0.0005049994612649499,
+      "loss": 3.681,
+      "step": 14800
+    },
+    {
+      "epoch": 1.5983209557636422,
+      "grad_norm": 0.5676015615463257,
+      "learning_rate": 0.0005046762202348884,
+      "loss": 3.6892,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6037025078032503,
+      "grad_norm": 0.5280758738517761,
+      "learning_rate": 0.000504352979204827,
+      "loss": 3.6736,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6090840598428586,
+      "grad_norm": 0.6497607827186584,
+      "learning_rate": 0.0005040297381747656,
+      "loss": 3.6761,
+      "step": 14950
+    },
+    {
+      "epoch": 1.614465611882467,
+      "grad_norm": 0.6543067693710327,
+      "learning_rate": 0.0005037064971447042,
+      "loss": 3.6601,
+      "step": 15000
+    },
+    {
+      "epoch": 1.614465611882467,
+      "eval_accuracy": 0.3556327449815676,
+      "eval_loss": 3.6367475986480713,
+      "eval_runtime": 205.3772,
+      "eval_samples_per_second": 87.697,
+      "eval_steps_per_second": 5.483,
+      "step": 15000
+    },
+    {
+      "epoch": 1.619847163922075,
+      "grad_norm": 0.5906841158866882,
+      "learning_rate": 0.0005033832561146428,
+      "loss": 3.6741,
+      "step": 15050
+    },
+    {
+      "epoch": 1.6252287159616834,
+      "grad_norm": 0.5810515880584717,
+      "learning_rate": 0.0005030600150845813,
+      "loss": 3.6827,
+      "step": 15100
+    },
+    {
+      "epoch": 1.6306102680012917,
+      "grad_norm": 0.6275368928909302,
+      "learning_rate": 0.00050273677405452,
+      "loss": 3.673,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6359918200408998,
+      "grad_norm": 0.5321255922317505,
+      "learning_rate": 0.0005024135330244585,
+      "loss": 3.6789,
+      "step": 15200
+    },
+    {
+      "epoch": 1.641373372080508,
+      "grad_norm": 0.5455909371376038,
+      "learning_rate": 0.0005020902919943972,
+      "loss": 3.6691,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6467549241201163,
+      "grad_norm": 0.5684463977813721,
+      "learning_rate": 0.0005017670509643357,
+      "loss": 3.656,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6521364761597246,
+      "grad_norm": 0.5891856551170349,
+      "learning_rate": 0.0005014438099342743,
+      "loss": 3.666,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6575180281993327,
+      "grad_norm": 0.5768228769302368,
+      "learning_rate": 0.0005011205689042129,
+      "loss": 3.6802,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6628995802389408,
+      "grad_norm": 0.5384355187416077,
+      "learning_rate": 0.0005007973278741514,
+      "loss": 3.6677,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6682811322785491,
+      "grad_norm": 0.5776270031929016,
+      "learning_rate": 0.00050047408684409,
+      "loss": 3.6913,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6736626843181575,
+      "grad_norm": 0.5457106232643127,
+      "learning_rate": 0.0005001508458140286,
+      "loss": 3.6747,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6790442363577656,
+      "grad_norm": 0.5584999322891235,
+      "learning_rate": 0.0004998276047839673,
+      "loss": 3.6656,
+      "step": 15600
+    },
+    {
+      "epoch": 1.6844257883973737,
+      "grad_norm": 0.5563948154449463,
+      "learning_rate": 0.0004995043637539058,
+      "loss": 3.6699,
+      "step": 15650
+    },
+    {
+      "epoch": 1.689807340436982,
+      "grad_norm": 0.5881028175354004,
+      "learning_rate": 0.0004991811227238443,
+      "loss": 3.6704,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6951888924765903,
+      "grad_norm": 0.5460503101348877,
+      "learning_rate": 0.0004988578816937829,
+      "loss": 3.6895,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7005704445161984,
+      "grad_norm": 0.6377148032188416,
+      "learning_rate": 0.0004985346406637215,
+      "loss": 3.6985,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7059519965558065,
+      "grad_norm": 0.5566238760948181,
+      "learning_rate": 0.0004982113996336602,
+      "loss": 3.6537,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7113335485954149,
+      "grad_norm": 0.5709816813468933,
+      "learning_rate": 0.0004978881586035987,
+      "loss": 3.6434,
+      "step": 15900
+    },
+    {
+      "epoch": 1.7167151006350232,
+      "grad_norm": 0.5734738111495972,
+      "learning_rate": 0.0004975649175735373,
+      "loss": 3.6449,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7220966526746313,
+      "grad_norm": 0.5572895407676697,
+      "learning_rate": 0.0004972416765434759,
+      "loss": 3.6579,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7220966526746313,
+      "eval_accuracy": 0.35718387360800713,
+      "eval_loss": 3.6182830333709717,
+      "eval_runtime": 214.6803,
+      "eval_samples_per_second": 83.897,
+      "eval_steps_per_second": 5.245,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7274782047142396,
+      "grad_norm": 0.5495673418045044,
+      "learning_rate": 0.0004969184355134145,
+      "loss": 3.6629,
+      "step": 16050
+    },
+    {
+      "epoch": 1.732859756753848,
+      "grad_norm": 0.6197149753570557,
+      "learning_rate": 0.0004965951944833531,
+      "loss": 3.6439,
+      "step": 16100
+    },
+    {
+      "epoch": 1.738241308793456,
+      "grad_norm": 0.5787531137466431,
+      "learning_rate": 0.0004962719534532916,
+      "loss": 3.6703,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7436228608330642,
+      "grad_norm": 0.5500142574310303,
+      "learning_rate": 0.0004959487124232302,
+      "loss": 3.656,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7490044128726725,
+      "grad_norm": 0.5271518230438232,
+      "learning_rate": 0.0004956254713931688,
+      "loss": 3.6594,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7543859649122808,
+      "grad_norm": 0.5560300350189209,
+      "learning_rate": 0.0004953022303631074,
+      "loss": 3.6622,
+      "step": 16300
+    },
+    {
+      "epoch": 1.759767516951889,
+      "grad_norm": 0.5545780062675476,
+      "learning_rate": 0.0004949789893330459,
+      "loss": 3.6657,
+      "step": 16350
+    },
+    {
+      "epoch": 1.765149068991497,
+      "grad_norm": 0.5637168884277344,
+      "learning_rate": 0.0004946557483029846,
+      "loss": 3.6701,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7705306210311054,
+      "grad_norm": 0.6499923467636108,
+      "learning_rate": 0.0004943325072729231,
+      "loss": 3.6591,
+      "step": 16450
+    },
+    {
+      "epoch": 1.7759121730707137,
+      "grad_norm": 0.5951055884361267,
+      "learning_rate": 0.0004940092662428617,
+      "loss": 3.6589,
+      "step": 16500
+    },
+    {
+      "epoch": 1.7812937251103218,
+      "grad_norm": 0.5989054441452026,
+      "learning_rate": 0.0004936860252128003,
+      "loss": 3.6615,
+      "step": 16550
+    },
+    {
+      "epoch": 1.78667527714993,
+      "grad_norm": 0.6071681976318359,
+      "learning_rate": 0.0004933627841827388,
+      "loss": 3.6413,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7920568291895382,
+      "grad_norm": 0.5401762127876282,
+      "learning_rate": 0.0004930395431526775,
+      "loss": 3.6646,
+      "step": 16650
+    },
+    {
+      "epoch": 1.7974383812291466,
+      "grad_norm": 0.6720755100250244,
+      "learning_rate": 0.0004927163021226161,
+      "loss": 3.6535,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8028199332687547,
+      "grad_norm": 0.5372287034988403,
+      "learning_rate": 0.0004923930610925547,
+      "loss": 3.6387,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8082014853083628,
+      "grad_norm": 0.5767584443092346,
+      "learning_rate": 0.0004920698200624932,
+      "loss": 3.6509,
+      "step": 16800
+    },
+    {
+      "epoch": 1.813583037347971,
+      "grad_norm": 0.619735598564148,
+      "learning_rate": 0.0004917465790324317,
+      "loss": 3.6593,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8189645893875794,
+      "grad_norm": 0.612182080745697,
+      "learning_rate": 0.0004914233380023704,
+      "loss": 3.669,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8243461414271875,
+      "grad_norm": 0.5973613262176514,
+      "learning_rate": 0.0004911000969723089,
+      "loss": 3.6581,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8297276934667959,
+      "grad_norm": 0.5377869606018066,
+      "learning_rate": 0.0004907768559422476,
+      "loss": 3.6582,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8297276934667959,
+      "eval_accuracy": 0.3589950085949867,
+      "eval_loss": 3.60500168800354,
+      "eval_runtime": 203.9944,
+      "eval_samples_per_second": 88.292,
+      "eval_steps_per_second": 5.52,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8351092455064042,
+      "grad_norm": 0.5382079482078552,
+      "learning_rate": 0.0004904536149121861,
+      "loss": 3.6536,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8404907975460123,
+      "grad_norm": 0.5734342336654663,
+      "learning_rate": 0.0004901303738821248,
+      "loss": 3.6325,
+      "step": 17100
+    },
+    {
+      "epoch": 1.8458723495856204,
+      "grad_norm": 0.6593245267868042,
+      "learning_rate": 0.0004898071328520633,
+      "loss": 3.6733,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8512539016252287,
+      "grad_norm": 0.5794579982757568,
+      "learning_rate": 0.0004894838918220019,
+      "loss": 3.6634,
+      "step": 17200
+    },
+    {
+      "epoch": 1.856635453664837,
+      "grad_norm": 0.5953862071037292,
+      "learning_rate": 0.0004891606507919405,
+      "loss": 3.6544,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8620170057044452,
+      "grad_norm": 0.6200346946716309,
+      "learning_rate": 0.000488837409761879,
+      "loss": 3.6443,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8673985577440533,
+      "grad_norm": 0.585228681564331,
+      "learning_rate": 0.0004885141687318177,
+      "loss": 3.6478,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8727801097836616,
+      "grad_norm": 0.6394591927528381,
+      "learning_rate": 0.00048819092770175623,
+      "loss": 3.6416,
+      "step": 17400
+    },
+    {
+      "epoch": 1.87816166182327,
+      "grad_norm": 0.5823236107826233,
+      "learning_rate": 0.0004878676866716948,
+      "loss": 3.6572,
+      "step": 17450
+    },
+    {
+      "epoch": 1.883543213862878,
+      "grad_norm": 0.5914448499679565,
+      "learning_rate": 0.00048754444564163337,
+      "loss": 3.6479,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8889247659024861,
+      "grad_norm": 0.6287000775337219,
+      "learning_rate": 0.000487221204611572,
+      "loss": 3.6431,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8943063179420945,
+      "grad_norm": 0.5507499575614929,
+      "learning_rate": 0.00048689796358151056,
+      "loss": 3.6444,
+      "step": 17600
+    },
+    {
+      "epoch": 1.8996878699817028,
+      "grad_norm": 0.5838987827301025,
+      "learning_rate": 0.00048657472255144915,
+      "loss": 3.649,
+      "step": 17650
+    },
+    {
+      "epoch": 1.905069422021311,
+      "grad_norm": 0.6239995360374451,
+      "learning_rate": 0.00048625148152138775,
+      "loss": 3.6385,
+      "step": 17700
+    },
+    {
+      "epoch": 1.910450974060919,
+      "grad_norm": 0.5718323588371277,
+      "learning_rate": 0.0004859282404913263,
+      "loss": 3.6248,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9158325261005273,
+      "grad_norm": 0.5826128125190735,
+      "learning_rate": 0.0004856049994612649,
+      "loss": 3.6441,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9212140781401357,
+      "grad_norm": 0.6261082291603088,
+      "learning_rate": 0.00048528175843120353,
+      "loss": 3.6379,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9265956301797438,
+      "grad_norm": 0.5729628205299377,
+      "learning_rate": 0.0004849585174011421,
+      "loss": 3.6403,
+      "step": 17900
+    },
+    {
+      "epoch": 1.931977182219352,
+      "grad_norm": 0.6479345560073853,
+      "learning_rate": 0.00048464174119168193,
+      "loss": 3.6513,
+      "step": 17950
+    },
+    {
+      "epoch": 1.9373587342589604,
+      "grad_norm": 0.6019191145896912,
+      "learning_rate": 0.0004843185001616205,
+      "loss": 3.6427,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9373587342589604,
+      "eval_accuracy": 0.3606186086975985,
+      "eval_loss": 3.589313268661499,
+      "eval_runtime": 204.8701,
+      "eval_samples_per_second": 87.914,
+      "eval_steps_per_second": 5.496,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9427402862985685,
+      "grad_norm": 0.5419387817382812,
+      "learning_rate": 0.00048399525913155907,
+      "loss": 3.6361,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9481218383381766,
+      "grad_norm": 0.6220253109931946,
+      "learning_rate": 0.0004836720181014976,
+      "loss": 3.6207,
+      "step": 18100
+    },
+    {
+      "epoch": 1.953503390377785,
+      "grad_norm": 0.5842125415802002,
+      "learning_rate": 0.0004833487770714362,
+      "loss": 3.6405,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9588849424173933,
+      "grad_norm": 0.5908029675483704,
+      "learning_rate": 0.00048302553604137485,
+      "loss": 3.6445,
+      "step": 18200
+    },
+    {
+      "epoch": 1.9642664944570014,
+      "grad_norm": 0.5424453020095825,
+      "learning_rate": 0.0004827022950113134,
+      "loss": 3.6419,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9696480464966095,
+      "grad_norm": 0.6032727360725403,
+      "learning_rate": 0.000482379053981252,
+      "loss": 3.6507,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9750295985362178,
+      "grad_norm": 0.6096407771110535,
+      "learning_rate": 0.0004820558129511906,
+      "loss": 3.6364,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9804111505758262,
+      "grad_norm": 0.6347204446792603,
+      "learning_rate": 0.0004817325719211291,
+      "loss": 3.6451,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9857927026154343,
+      "grad_norm": 0.5634959936141968,
+      "learning_rate": 0.0004814093308910677,
+      "loss": 3.641,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9911742546550424,
+      "grad_norm": 0.5860807299613953,
+      "learning_rate": 0.00048108608986100637,
+      "loss": 3.6336,
+      "step": 18500
+    },
+    {
+      "epoch": 1.9965558066946507,
+      "grad_norm": 0.5338708162307739,
+      "learning_rate": 0.0004807628488309449,
+      "loss": 3.6211,
+      "step": 18550
+    },
+    {
+      "epoch": 2.001937358734259,
+      "grad_norm": 0.5974145531654358,
+      "learning_rate": 0.0004804396078008835,
+      "loss": 3.5958,
+      "step": 18600
+    },
+    {
+      "epoch": 2.007318910773867,
+      "grad_norm": 0.5562605261802673,
+      "learning_rate": 0.00048011636677082204,
+      "loss": 3.5582,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0127004628134753,
+      "grad_norm": 0.6376757025718689,
+      "learning_rate": 0.00047979312574076064,
+      "loss": 3.5377,
+      "step": 18700
+    },
+    {
+      "epoch": 2.018082014853084,
+      "grad_norm": 0.5702477097511292,
+      "learning_rate": 0.0004794698847106992,
+      "loss": 3.5431,
+      "step": 18750
+    },
+    {
+      "epoch": 2.023463566892692,
+      "grad_norm": 0.5655612945556641,
+      "learning_rate": 0.0004791466436806378,
+      "loss": 3.5661,
+      "step": 18800
+    },
+    {
+      "epoch": 2.0288451189323,
+      "grad_norm": 0.5684214234352112,
+      "learning_rate": 0.0004788234026505764,
+      "loss": 3.5378,
+      "step": 18850
+    },
+    {
+      "epoch": 2.034226670971908,
+      "grad_norm": 0.595730185508728,
+      "learning_rate": 0.00047850016162051496,
+      "loss": 3.5512,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0396082230115167,
+      "grad_norm": 0.602304220199585,
+      "learning_rate": 0.00047817692059045356,
+      "loss": 3.5471,
+      "step": 18950
+    },
+    {
+      "epoch": 2.044989775051125,
+      "grad_norm": 0.6218283176422119,
+      "learning_rate": 0.00047785367956039215,
+      "loss": 3.56,
+      "step": 19000
+    },
+    {
+      "epoch": 2.044989775051125,
+      "eval_accuracy": 0.3616626542966078,
+      "eval_loss": 3.5785436630249023,
+      "eval_runtime": 202.2767,
+      "eval_samples_per_second": 89.041,
+      "eval_steps_per_second": 5.567,
+      "step": 19000
+    },
+    {
+      "epoch": 2.050371327090733,
+      "grad_norm": 0.5897380113601685,
+      "learning_rate": 0.00047753043853033075,
+      "loss": 3.5425,
+      "step": 19050
+    },
+    {
+      "epoch": 2.055752879130341,
+      "grad_norm": 0.5932830572128296,
+      "learning_rate": 0.00047720719750026934,
+      "loss": 3.5699,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0611344311699495,
+      "grad_norm": 0.5921191573143005,
+      "learning_rate": 0.00047688395647020793,
+      "loss": 3.5431,
+      "step": 19150
+    },
+    {
+      "epoch": 2.0665159832095576,
+      "grad_norm": 0.6052051186561584,
+      "learning_rate": 0.0004765607154401465,
+      "loss": 3.5734,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0718975352491658,
+      "grad_norm": 0.562523603439331,
+      "learning_rate": 0.00047623747441008507,
+      "loss": 3.543,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0772790872887743,
+      "grad_norm": 0.546420156955719,
+      "learning_rate": 0.0004759142333800236,
+      "loss": 3.5441,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0826606393283824,
+      "grad_norm": 0.590813159942627,
+      "learning_rate": 0.00047559099234996226,
+      "loss": 3.5624,
+      "step": 19350
+    },
+    {
+      "epoch": 2.0880421913679905,
+      "grad_norm": 0.6105584502220154,
+      "learning_rate": 0.00047526775131990085,
+      "loss": 3.557,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0934237434075986,
+      "grad_norm": 0.6170618534088135,
+      "learning_rate": 0.0004749445102898394,
+      "loss": 3.5373,
+      "step": 19450
+    },
+    {
+      "epoch": 2.098805295447207,
+      "grad_norm": 0.5814986228942871,
+      "learning_rate": 0.000474621269259778,
+      "loss": 3.5693,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1041868474868153,
+      "grad_norm": 0.6113174557685852,
+      "learning_rate": 0.0004742980282297166,
+      "loss": 3.5589,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1095683995264234,
+      "grad_norm": 0.6312413811683655,
+      "learning_rate": 0.0004739747871996551,
+      "loss": 3.5398,
+      "step": 19600
+    },
+    {
+      "epoch": 2.1149499515660315,
+      "grad_norm": 0.6352087259292603,
+      "learning_rate": 0.00047365154616959377,
+      "loss": 3.5541,
+      "step": 19650
+    },
+    {
+      "epoch": 2.12033150360564,
+      "grad_norm": 0.6081045269966125,
+      "learning_rate": 0.00047332830513953237,
+      "loss": 3.576,
+      "step": 19700
+    },
+    {
+      "epoch": 2.125713055645248,
+      "grad_norm": 0.5704367160797119,
+      "learning_rate": 0.0004730050641094709,
+      "loss": 3.5563,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1310946076848563,
+      "grad_norm": 0.6367883682250977,
+      "learning_rate": 0.0004726818230794095,
+      "loss": 3.5358,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1364761597244644,
+      "grad_norm": 0.5543844699859619,
+      "learning_rate": 0.00047235858204934804,
+      "loss": 3.5561,
+      "step": 19850
+    },
+    {
+      "epoch": 2.141857711764073,
+      "grad_norm": 0.6325891017913818,
+      "learning_rate": 0.0004720353410192867,
+      "loss": 3.5459,
+      "step": 19900
+    },
+    {
+      "epoch": 2.147239263803681,
+      "grad_norm": 0.5876049995422363,
+      "learning_rate": 0.0004717120999892253,
+      "loss": 3.5568,
+      "step": 19950
+    },
+    {
+      "epoch": 2.152620815843289,
+      "grad_norm": 0.5623601675033569,
+      "learning_rate": 0.0004713888589591638,
+      "loss": 3.5527,
+      "step": 20000
+    },
+    {
+      "epoch": 2.152620815843289,
+      "eval_accuracy": 0.3627479879929867,
+      "eval_loss": 3.5705673694610596,
+      "eval_runtime": 206.2591,
+      "eval_samples_per_second": 87.322,
+      "eval_steps_per_second": 5.459,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1580023678828972,
+      "grad_norm": 0.5725038051605225,
+      "learning_rate": 0.0004710656179291024,
+      "loss": 3.5565,
+      "step": 20050
+    },
+    {
+      "epoch": 2.163383919922506,
+      "grad_norm": 0.6287389993667603,
+      "learning_rate": 0.000470742376899041,
+      "loss": 3.5528,
+      "step": 20100
+    },
+    {
+      "epoch": 2.168765471962114,
+      "grad_norm": 0.5884525179862976,
+      "learning_rate": 0.00047041913586897956,
+      "loss": 3.5597,
+      "step": 20150
+    },
+    {
+      "epoch": 2.174147024001722,
+      "grad_norm": 0.5764286518096924,
+      "learning_rate": 0.0004700958948389182,
+      "loss": 3.5426,
+      "step": 20200
+    },
+    {
+      "epoch": 2.1795285760413305,
+      "grad_norm": 0.6041747331619263,
+      "learning_rate": 0.0004697726538088568,
+      "loss": 3.5471,
+      "step": 20250
+    },
+    {
+      "epoch": 2.1849101280809387,
+      "grad_norm": 0.5907628536224365,
+      "learning_rate": 0.00046944941277879534,
+      "loss": 3.5453,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1902916801205468,
+      "grad_norm": 0.5632272958755493,
+      "learning_rate": 0.00046912617174873394,
+      "loss": 3.5558,
+      "step": 20350
+    },
+    {
+      "epoch": 2.195673232160155,
+      "grad_norm": 0.6226711869239807,
+      "learning_rate": 0.0004688029307186725,
+      "loss": 3.5865,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2010547841997634,
+      "grad_norm": 0.5935404896736145,
+      "learning_rate": 0.00046847968968861107,
+      "loss": 3.5594,
+      "step": 20450
+    },
+    {
+      "epoch": 2.2064363362393715,
+      "grad_norm": 0.5879222750663757,
+      "learning_rate": 0.0004681564486585497,
+      "loss": 3.5637,
+      "step": 20500
+    },
+    {
+      "epoch": 2.2118178882789796,
+      "grad_norm": 0.5888842940330505,
+      "learning_rate": 0.00046783320762848826,
+      "loss": 3.5438,
+      "step": 20550
+    },
+    {
+      "epoch": 2.2171994403185877,
+      "grad_norm": 0.5803065299987793,
+      "learning_rate": 0.00046750996659842685,
+      "loss": 3.5429,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2225809923581963,
+      "grad_norm": 0.5627387166023254,
+      "learning_rate": 0.00046718672556836545,
+      "loss": 3.5601,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2279625443978044,
+      "grad_norm": 0.5622759461402893,
+      "learning_rate": 0.000466863484538304,
+      "loss": 3.5676,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2333440964374125,
+      "grad_norm": 0.5685352087020874,
+      "learning_rate": 0.0004665402435082426,
+      "loss": 3.5572,
+      "step": 20750
+    },
+    {
+      "epoch": 2.2387256484770206,
+      "grad_norm": 0.6102375388145447,
+      "learning_rate": 0.00046621700247818123,
+      "loss": 3.5546,
+      "step": 20800
+    },
+    {
+      "epoch": 2.244107200516629,
+      "grad_norm": 0.5915598273277283,
+      "learning_rate": 0.0004658937614481198,
+      "loss": 3.5519,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2494887525562373,
+      "grad_norm": 0.5531293153762817,
+      "learning_rate": 0.00046557052041805837,
+      "loss": 3.5491,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2548703045958454,
+      "grad_norm": 0.5677109956741333,
+      "learning_rate": 0.0004652472793879969,
+      "loss": 3.5562,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2602518566354535,
+      "grad_norm": 0.5943840742111206,
+      "learning_rate": 0.0004649240383579355,
+      "loss": 3.5487,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2602518566354535,
+      "eval_accuracy": 0.36394262649450687,
+      "eval_loss": 3.560176372528076,
+      "eval_runtime": 204.7535,
+      "eval_samples_per_second": 87.964,
+      "eval_steps_per_second": 5.499,
+      "step": 21000
+    },
+    {
+      "epoch": 2.265633408675062,
+      "grad_norm": 0.5741783380508423,
+      "learning_rate": 0.00046460079732787415,
+      "loss": 3.5606,
+      "step": 21050
+    },
+    {
+      "epoch": 2.27101496071467,
+      "grad_norm": 0.6168876886367798,
+      "learning_rate": 0.0004642775562978127,
+      "loss": 3.5662,
+      "step": 21100
+    },
+    {
+      "epoch": 2.2763965127542782,
+      "grad_norm": 0.6492966413497925,
+      "learning_rate": 0.0004639543152677513,
+      "loss": 3.5546,
+      "step": 21150
+    },
+    {
+      "epoch": 2.281778064793887,
+      "grad_norm": 0.634781539440155,
+      "learning_rate": 0.0004636310742376899,
+      "loss": 3.5674,
+      "step": 21200
+    },
+    {
+      "epoch": 2.287159616833495,
+      "grad_norm": 0.5752384066581726,
+      "learning_rate": 0.0004633078332076284,
+      "loss": 3.5561,
+      "step": 21250
+    },
+    {
+      "epoch": 2.292541168873103,
+      "grad_norm": 0.5751746296882629,
+      "learning_rate": 0.000462984592177567,
+      "loss": 3.5505,
+      "step": 21300
+    },
+    {
+      "epoch": 2.297922720912711,
+      "grad_norm": 0.5733442902565002,
+      "learning_rate": 0.00046266135114750567,
+      "loss": 3.5489,
+      "step": 21350
+    },
+    {
+      "epoch": 2.303304272952319,
+      "grad_norm": 0.6262395977973938,
+      "learning_rate": 0.0004623381101174442,
+      "loss": 3.552,
+      "step": 21400
+    },
+    {
+      "epoch": 2.3086858249919278,
+      "grad_norm": 0.6084133982658386,
+      "learning_rate": 0.0004620148690873828,
+      "loss": 3.5515,
+      "step": 21450
+    },
+    {
+      "epoch": 2.314067377031536,
+      "grad_norm": 0.558005690574646,
+      "learning_rate": 0.00046169162805732134,
+      "loss": 3.5501,
+      "step": 21500
+    },
+    {
+      "epoch": 2.319448929071144,
+      "grad_norm": 0.656355619430542,
+      "learning_rate": 0.00046136838702725994,
+      "loss": 3.5677,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3248304811107525,
+      "grad_norm": 0.6125053763389587,
+      "learning_rate": 0.00046104514599719853,
+      "loss": 3.5496,
+      "step": 21600
+    },
+    {
+      "epoch": 2.3302120331503606,
+      "grad_norm": 0.708257257938385,
+      "learning_rate": 0.0004607219049671371,
+      "loss": 3.594,
+      "step": 21650
+    },
+    {
+      "epoch": 2.3355935851899687,
+      "grad_norm": 0.6216398477554321,
+      "learning_rate": 0.0004603986639370757,
+      "loss": 3.5749,
+      "step": 21700
+    },
+    {
+      "epoch": 2.340975137229577,
+      "grad_norm": 0.5760979056358337,
+      "learning_rate": 0.0004600754229070143,
+      "loss": 3.5546,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3463566892691854,
+      "grad_norm": 0.5885084867477417,
+      "learning_rate": 0.00045975218187695286,
+      "loss": 3.5654,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3517382413087935,
+      "grad_norm": 0.6174579858779907,
+      "learning_rate": 0.00045942894084689145,
+      "loss": 3.5543,
+      "step": 21850
+    },
+    {
+      "epoch": 2.3571197933484016,
+      "grad_norm": 0.6659450531005859,
+      "learning_rate": 0.0004591056998168301,
+      "loss": 3.553,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3625013453880097,
+      "grad_norm": 0.5789748430252075,
+      "learning_rate": 0.00045878245878676864,
+      "loss": 3.5611,
+      "step": 21950
+    },
+    {
+      "epoch": 2.3678828974276183,
+      "grad_norm": 0.5881466865539551,
+      "learning_rate": 0.00045846568257730845,
+      "loss": 3.5664,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3678828974276183,
+      "eval_accuracy": 0.36495016472321057,
+      "eval_loss": 3.5474905967712402,
+      "eval_runtime": 194.8012,
+      "eval_samples_per_second": 92.458,
+      "eval_steps_per_second": 5.78,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3732644494672264,
+      "grad_norm": 0.6417264342308044,
+      "learning_rate": 0.00045814244154724704,
+      "loss": 3.5509,
+      "step": 22050
+    },
+    {
+      "epoch": 2.3786460015068345,
+      "grad_norm": 0.6134101748466492,
+      "learning_rate": 0.00045781920051718563,
+      "loss": 3.5482,
+      "step": 22100
+    },
+    {
+      "epoch": 2.384027553546443,
+      "grad_norm": 0.8174729943275452,
+      "learning_rate": 0.0004574959594871242,
+      "loss": 3.5538,
+      "step": 22150
+    },
+    {
+      "epoch": 2.389409105586051,
+      "grad_norm": 0.6091681122779846,
+      "learning_rate": 0.00045717271845706277,
+      "loss": 3.5516,
+      "step": 22200
+    },
+    {
+      "epoch": 2.3947906576256592,
+      "grad_norm": 0.5839729905128479,
+      "learning_rate": 0.0004568494774270013,
+      "loss": 3.5545,
+      "step": 22250
+    },
+    {
+      "epoch": 2.4001722096652673,
+      "grad_norm": 0.620452344417572,
+      "learning_rate": 0.00045652623639693996,
+      "loss": 3.5427,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4055537617048754,
+      "grad_norm": 0.6439001560211182,
+      "learning_rate": 0.00045620299536687855,
+      "loss": 3.554,
+      "step": 22350
+    },
+    {
+      "epoch": 2.410935313744484,
+      "grad_norm": 0.6319786906242371,
+      "learning_rate": 0.0004558797543368171,
+      "loss": 3.5723,
+      "step": 22400
+    },
+    {
+      "epoch": 2.416316865784092,
+      "grad_norm": 0.6055615544319153,
+      "learning_rate": 0.0004555565133067557,
+      "loss": 3.5568,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4216984178237,
+      "grad_norm": 0.6199345588684082,
+      "learning_rate": 0.0004552332722766943,
+      "loss": 3.549,
+      "step": 22500
+    },
+    {
+      "epoch": 2.4270799698633088,
+      "grad_norm": 0.587843656539917,
+      "learning_rate": 0.0004549100312466328,
+      "loss": 3.5311,
+      "step": 22550
+    },
+    {
+      "epoch": 2.432461521902917,
+      "grad_norm": 0.5868039727210999,
+      "learning_rate": 0.0004545867902165715,
+      "loss": 3.5624,
+      "step": 22600
+    },
+    {
+      "epoch": 2.437843073942525,
+      "grad_norm": 0.5922572016716003,
+      "learning_rate": 0.00045426354918651007,
+      "loss": 3.5538,
+      "step": 22650
+    },
+    {
+      "epoch": 2.443224625982133,
+      "grad_norm": 0.6361838579177856,
+      "learning_rate": 0.0004539403081564486,
+      "loss": 3.5364,
+      "step": 22700
+    },
+    {
+      "epoch": 2.4486061780217416,
+      "grad_norm": 0.5849791765213013,
+      "learning_rate": 0.0004536170671263872,
+      "loss": 3.5606,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4539877300613497,
+      "grad_norm": 0.6513404250144958,
+      "learning_rate": 0.00045329382609632574,
+      "loss": 3.535,
+      "step": 22800
+    },
+    {
+      "epoch": 2.459369282100958,
+      "grad_norm": 0.5861430168151855,
+      "learning_rate": 0.0004529705850662644,
+      "loss": 3.5432,
+      "step": 22850
+    },
+    {
+      "epoch": 2.464750834140566,
+      "grad_norm": 0.6476746797561646,
+      "learning_rate": 0.000452647344036203,
+      "loss": 3.5454,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4701323861801745,
+      "grad_norm": 0.6115290522575378,
+      "learning_rate": 0.00045232410300614153,
+      "loss": 3.5439,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4755139382197826,
+      "grad_norm": 0.5903562903404236,
+      "learning_rate": 0.0004520008619760801,
+      "loss": 3.5464,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4755139382197826,
+      "eval_accuracy": 0.3655437354493411,
+      "eval_loss": 3.5417819023132324,
+      "eval_runtime": 206.588,
+      "eval_samples_per_second": 87.183,
+      "eval_steps_per_second": 5.45,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4808954902593907,
+      "grad_norm": 0.5942795276641846,
+      "learning_rate": 0.0004516776209460187,
+      "loss": 3.5293,
+      "step": 23050
+    },
+    {
+      "epoch": 2.4862770422989993,
+      "grad_norm": 0.5559934973716736,
+      "learning_rate": 0.00045135437991595726,
+      "loss": 3.5569,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4916585943386074,
+      "grad_norm": 0.6209527254104614,
+      "learning_rate": 0.0004510311388858959,
+      "loss": 3.5374,
+      "step": 23150
+    },
+    {
+      "epoch": 2.4970401463782155,
+      "grad_norm": 0.5962895750999451,
+      "learning_rate": 0.0004507078978558345,
+      "loss": 3.5538,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5024216984178236,
+      "grad_norm": 0.6051533222198486,
+      "learning_rate": 0.00045038465682577304,
+      "loss": 3.552,
+      "step": 23250
+    },
+    {
+      "epoch": 2.5078032504574317,
+      "grad_norm": 0.5640018582344055,
+      "learning_rate": 0.00045006141579571164,
+      "loss": 3.5596,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5131848024970402,
+      "grad_norm": 0.6006429195404053,
+      "learning_rate": 0.0004497381747656502,
+      "loss": 3.5692,
+      "step": 23350
+    },
+    {
+      "epoch": 2.5185663545366483,
+      "grad_norm": 0.6432710886001587,
+      "learning_rate": 0.00044941493373558877,
+      "loss": 3.5302,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5239479065762565,
+      "grad_norm": 0.5585759282112122,
+      "learning_rate": 0.0004490916927055274,
+      "loss": 3.5559,
+      "step": 23450
+    },
+    {
+      "epoch": 2.529329458615865,
+      "grad_norm": 0.6355800032615662,
+      "learning_rate": 0.00044876845167546596,
+      "loss": 3.5332,
+      "step": 23500
+    },
+    {
+      "epoch": 2.534711010655473,
+      "grad_norm": 0.5865529179573059,
+      "learning_rate": 0.00044844521064540455,
+      "loss": 3.5558,
+      "step": 23550
+    },
+    {
+      "epoch": 2.540092562695081,
+      "grad_norm": 0.6548492312431335,
+      "learning_rate": 0.00044812196961534315,
+      "loss": 3.5339,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5454741147346893,
+      "grad_norm": 0.6574897766113281,
+      "learning_rate": 0.0004477987285852817,
+      "loss": 3.5498,
+      "step": 23650
+    },
+    {
+      "epoch": 2.550855666774298,
+      "grad_norm": 0.5814394950866699,
+      "learning_rate": 0.00044747548755522034,
+      "loss": 3.5552,
+      "step": 23700
+    },
+    {
+      "epoch": 2.556237218813906,
+      "grad_norm": 0.5807026028633118,
+      "learning_rate": 0.00044715224652515893,
+      "loss": 3.5408,
+      "step": 23750
+    },
+    {
+      "epoch": 2.561618770853514,
+      "grad_norm": 0.6143361926078796,
+      "learning_rate": 0.0004468290054950975,
+      "loss": 3.5664,
+      "step": 23800
+    },
+    {
+      "epoch": 2.567000322893122,
+      "grad_norm": 0.5572286248207092,
+      "learning_rate": 0.00044650576446503607,
+      "loss": 3.5545,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5723818749327307,
+      "grad_norm": 0.5634745955467224,
+      "learning_rate": 0.0004461825234349746,
+      "loss": 3.5302,
+      "step": 23900
+    },
+    {
+      "epoch": 2.577763426972339,
+      "grad_norm": 0.5858882665634155,
+      "learning_rate": 0.00044586574722551447,
+      "loss": 3.5528,
+      "step": 23950
+    },
+    {
+      "epoch": 2.583144979011947,
+      "grad_norm": 0.6945074200630188,
+      "learning_rate": 0.000445542506195453,
+      "loss": 3.5619,
+      "step": 24000
+    },
+    {
+      "epoch": 2.583144979011947,
+      "eval_accuracy": 0.36655561979355733,
+      "eval_loss": 3.5325984954833984,
+      "eval_runtime": 195.3908,
+      "eval_samples_per_second": 92.179,
+      "eval_steps_per_second": 5.763,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5885265310515555,
+      "grad_norm": 0.5496438145637512,
+      "learning_rate": 0.0004452192651653916,
+      "loss": 3.5522,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5939080830911636,
+      "grad_norm": 0.5970087051391602,
+      "learning_rate": 0.00044489602413533025,
+      "loss": 3.5635,
+      "step": 24100
+    },
+    {
+      "epoch": 2.5992896351307717,
+      "grad_norm": 0.6501139998435974,
+      "learning_rate": 0.0004445727831052688,
+      "loss": 3.533,
+      "step": 24150
+    },
+    {
+      "epoch": 2.60467118717038,
+      "grad_norm": 0.6015759110450745,
+      "learning_rate": 0.0004442495420752074,
+      "loss": 3.5451,
+      "step": 24200
+    },
+    {
+      "epoch": 2.610052739209988,
+      "grad_norm": 0.5871186852455139,
+      "learning_rate": 0.00044392630104514593,
+      "loss": 3.5508,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6154342912495965,
+      "grad_norm": 0.5801131725311279,
+      "learning_rate": 0.0004436030600150845,
+      "loss": 3.5321,
+      "step": 24300
+    },
+    {
+      "epoch": 2.6208158432892046,
+      "grad_norm": 0.7078419327735901,
+      "learning_rate": 0.0004432798189850231,
+      "loss": 3.5526,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6261973953288127,
+      "grad_norm": 0.6336981654167175,
+      "learning_rate": 0.0004429565779549617,
+      "loss": 3.553,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6315789473684212,
+      "grad_norm": 0.621759831905365,
+      "learning_rate": 0.0004426333369249003,
+      "loss": 3.5336,
+      "step": 24450
+    },
+    {
+      "epoch": 2.6369604994080293,
+      "grad_norm": 0.5916746854782104,
+      "learning_rate": 0.0004423100958948389,
+      "loss": 3.5525,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6423420514476375,
+      "grad_norm": 0.6469952464103699,
+      "learning_rate": 0.00044198685486477744,
+      "loss": 3.5264,
+      "step": 24550
+    },
+    {
+      "epoch": 2.6477236034872456,
+      "grad_norm": 0.6205400824546814,
+      "learning_rate": 0.00044166361383471604,
+      "loss": 3.5513,
+      "step": 24600
+    },
+    {
+      "epoch": 2.653105155526854,
+      "grad_norm": 0.6007969975471497,
+      "learning_rate": 0.0004413403728046547,
+      "loss": 3.5303,
+      "step": 24650
+    },
+    {
+      "epoch": 2.658486707566462,
+      "grad_norm": 0.6586825847625732,
+      "learning_rate": 0.0004410171317745932,
+      "loss": 3.5627,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6638682596060703,
+      "grad_norm": 0.5926357507705688,
+      "learning_rate": 0.0004406938907445318,
+      "loss": 3.5663,
+      "step": 24750
+    },
+    {
+      "epoch": 2.6692498116456784,
+      "grad_norm": 0.5873600840568542,
+      "learning_rate": 0.00044037064971447036,
+      "loss": 3.5241,
+      "step": 24800
+    },
+    {
+      "epoch": 2.674631363685287,
+      "grad_norm": 0.592735767364502,
+      "learning_rate": 0.00044004740868440896,
+      "loss": 3.5567,
+      "step": 24850
+    },
+    {
+      "epoch": 2.680012915724895,
+      "grad_norm": 0.626936674118042,
+      "learning_rate": 0.00043972416765434755,
+      "loss": 3.5204,
+      "step": 24900
+    },
+    {
+      "epoch": 2.685394467764503,
+      "grad_norm": 0.5691058039665222,
+      "learning_rate": 0.00043940092662428615,
+      "loss": 3.5214,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6907760198041117,
+      "grad_norm": 0.6450164914131165,
+      "learning_rate": 0.00043907768559422474,
+      "loss": 3.5487,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6907760198041117,
+      "eval_accuracy": 0.36812326365894465,
+      "eval_loss": 3.521350145339966,
+      "eval_runtime": 197.2244,
+      "eval_samples_per_second": 91.322,
+      "eval_steps_per_second": 5.709,
+      "step": 25000
+    },
+    {
+      "epoch": 2.69615757184372,
+      "grad_norm": 0.6423863768577576,
+      "learning_rate": 0.00043875444456416334,
+      "loss": 3.5373,
+      "step": 25050
+    },
+    {
+      "epoch": 2.701539123883328,
+      "grad_norm": 0.6400243043899536,
+      "learning_rate": 0.0004384312035341019,
+      "loss": 3.5428,
+      "step": 25100
+    },
+    {
+      "epoch": 2.706920675922936,
+      "grad_norm": 0.6252772212028503,
+      "learning_rate": 0.00043810796250404047,
+      "loss": 3.5469,
+      "step": 25150
+    },
+    {
+      "epoch": 2.712302227962544,
+      "grad_norm": 0.5882490873336792,
+      "learning_rate": 0.000437784721473979,
+      "loss": 3.5376,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7176837800021527,
+      "grad_norm": 0.604751467704773,
+      "learning_rate": 0.00043746148044391766,
+      "loss": 3.5409,
+      "step": 25250
+    },
+    {
+      "epoch": 2.723065332041761,
+      "grad_norm": 0.6292399168014526,
+      "learning_rate": 0.00043713823941385625,
+      "loss": 3.5313,
+      "step": 25300
+    },
+    {
+      "epoch": 2.728446884081369,
+      "grad_norm": 0.6144165396690369,
+      "learning_rate": 0.0004368149983837948,
+      "loss": 3.5484,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7338284361209775,
+      "grad_norm": 0.6086465716362,
+      "learning_rate": 0.0004364917573537334,
+      "loss": 3.5381,
+      "step": 25400
+    },
+    {
+      "epoch": 2.7392099881605856,
+      "grad_norm": 0.6256065964698792,
+      "learning_rate": 0.00043616851632367193,
+      "loss": 3.557,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7445915402001937,
+      "grad_norm": 0.6041662096977234,
+      "learning_rate": 0.0004358452752936106,
+      "loss": 3.5347,
+      "step": 25500
+    },
+    {
+      "epoch": 2.749973092239802,
+      "grad_norm": 0.6548029780387878,
+      "learning_rate": 0.0004355220342635492,
+      "loss": 3.5374,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7553546442794103,
+      "grad_norm": 0.6124310493469238,
+      "learning_rate": 0.00043519879323348777,
+      "loss": 3.529,
+      "step": 25600
+    },
+    {
+      "epoch": 2.7607361963190185,
+      "grad_norm": 0.596762478351593,
+      "learning_rate": 0.0004348755522034263,
+      "loss": 3.5273,
+      "step": 25650
+    },
+    {
+      "epoch": 2.7661177483586266,
+      "grad_norm": 0.5779131650924683,
+      "learning_rate": 0.0004345523111733649,
+      "loss": 3.5448,
+      "step": 25700
+    },
+    {
+      "epoch": 2.7714993003982347,
+      "grad_norm": 0.6550736427307129,
+      "learning_rate": 0.00043422907014330344,
+      "loss": 3.5405,
+      "step": 25750
+    },
+    {
+      "epoch": 2.776880852437843,
+      "grad_norm": 0.7348501682281494,
+      "learning_rate": 0.0004339058291132421,
+      "loss": 3.5329,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7822624044774513,
+      "grad_norm": 0.6231963038444519,
+      "learning_rate": 0.0004335825880831807,
+      "loss": 3.5465,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7876439565170594,
+      "grad_norm": 0.5558022856712341,
+      "learning_rate": 0.00043325934705311923,
+      "loss": 3.5381,
+      "step": 25900
+    },
+    {
+      "epoch": 2.793025508556668,
+      "grad_norm": 0.6114339232444763,
+      "learning_rate": 0.0004329361060230578,
+      "loss": 3.5492,
+      "step": 25950
+    },
+    {
+      "epoch": 2.798407060596276,
+      "grad_norm": 0.621716320514679,
+      "learning_rate": 0.00043261286499299636,
+      "loss": 3.5277,
+      "step": 26000
+    },
+    {
+      "epoch": 2.798407060596276,
+      "eval_accuracy": 0.36862752171129165,
+      "eval_loss": 3.51253342628479,
+      "eval_runtime": 212.7222,
+      "eval_samples_per_second": 84.669,
+      "eval_steps_per_second": 5.293,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803788612635884,
+      "grad_norm": 0.6198678016662598,
+      "learning_rate": 0.00043228962396293496,
+      "loss": 3.5241,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8091701646754923,
+      "grad_norm": 0.5953556299209595,
+      "learning_rate": 0.0004319663829328736,
+      "loss": 3.5257,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8145517167151004,
+      "grad_norm": 0.6188264489173889,
+      "learning_rate": 0.00043164314190281215,
+      "loss": 3.5268,
+      "step": 26150
+    },
+    {
+      "epoch": 2.819933268754709,
+      "grad_norm": 0.5980226993560791,
+      "learning_rate": 0.00043131990087275074,
+      "loss": 3.5478,
+      "step": 26200
+    },
+    {
+      "epoch": 2.825314820794317,
+      "grad_norm": 0.6212344169616699,
+      "learning_rate": 0.00043099665984268934,
+      "loss": 3.5297,
+      "step": 26250
+    },
+    {
+      "epoch": 2.830696372833925,
+      "grad_norm": 0.5705240964889526,
+      "learning_rate": 0.0004306734188126279,
+      "loss": 3.5324,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8360779248735337,
+      "grad_norm": 0.6547979712486267,
+      "learning_rate": 0.00043035017778256647,
+      "loss": 3.533,
+      "step": 26350
+    },
+    {
+      "epoch": 2.841459476913142,
+      "grad_norm": 0.5784933567047119,
+      "learning_rate": 0.0004300269367525051,
+      "loss": 3.5297,
+      "step": 26400
+    },
+    {
+      "epoch": 2.84684102895275,
+      "grad_norm": 0.7146517038345337,
+      "learning_rate": 0.00042970369572244366,
+      "loss": 3.5382,
+      "step": 26450
+    },
+    {
+      "epoch": 2.852222580992358,
+      "grad_norm": 0.6623178124427795,
+      "learning_rate": 0.00042938045469238226,
+      "loss": 3.5374,
+      "step": 26500
+    },
+    {
+      "epoch": 2.857604133031966,
+      "grad_norm": 0.5954160094261169,
+      "learning_rate": 0.0004290572136623208,
+      "loss": 3.5509,
+      "step": 26550
+    },
+    {
+      "epoch": 2.8629856850715747,
+      "grad_norm": 0.6484636068344116,
+      "learning_rate": 0.0004287339726322594,
+      "loss": 3.5399,
+      "step": 26600
+    },
+    {
+      "epoch": 2.868367237111183,
+      "grad_norm": 0.6219885349273682,
+      "learning_rate": 0.00042841073160219804,
+      "loss": 3.5334,
+      "step": 26650
+    },
+    {
+      "epoch": 2.873748789150791,
+      "grad_norm": 0.6355128288269043,
+      "learning_rate": 0.0004280874905721366,
+      "loss": 3.5135,
+      "step": 26700
+    },
+    {
+      "epoch": 2.8791303411903995,
+      "grad_norm": 0.5998795032501221,
+      "learning_rate": 0.0004277642495420752,
+      "loss": 3.5442,
+      "step": 26750
+    },
+    {
+      "epoch": 2.8845118932300076,
+      "grad_norm": 0.5889739394187927,
+      "learning_rate": 0.00042744100851201377,
+      "loss": 3.5133,
+      "step": 26800
+    },
+    {
+      "epoch": 2.8898934452696157,
+      "grad_norm": 0.6281293630599976,
+      "learning_rate": 0.0004271177674819523,
+      "loss": 3.5056,
+      "step": 26850
+    },
+    {
+      "epoch": 2.895274997309224,
+      "grad_norm": 0.5909113883972168,
+      "learning_rate": 0.0004267945264518909,
+      "loss": 3.5157,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9006565493488323,
+      "grad_norm": 0.6297270059585571,
+      "learning_rate": 0.00042647128542182955,
+      "loss": 3.5354,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9060381013884404,
+      "grad_norm": 0.6224883794784546,
+      "learning_rate": 0.0004261480443917681,
+      "loss": 3.5346,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9060381013884404,
+      "eval_accuracy": 0.3694928333098462,
+      "eval_loss": 3.5037596225738525,
+      "eval_runtime": 206.2343,
+      "eval_samples_per_second": 87.333,
+      "eval_steps_per_second": 5.46,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9114196534280485,
+      "grad_norm": 0.6158651113510132,
+      "learning_rate": 0.00042583126818230795,
+      "loss": 3.5352,
+      "step": 27050
+    },
+    {
+      "epoch": 2.9168012054676566,
+      "grad_norm": 0.6259952187538147,
+      "learning_rate": 0.0004255080271522465,
+      "loss": 3.5156,
+      "step": 27100
+    },
+    {
+      "epoch": 2.922182757507265,
+      "grad_norm": 0.5961571335792542,
+      "learning_rate": 0.0004251847861221851,
+      "loss": 3.5362,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9275643095468733,
+      "grad_norm": 0.62592613697052,
+      "learning_rate": 0.00042486154509212363,
+      "loss": 3.5094,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9329458615864814,
+      "grad_norm": 0.5974050760269165,
+      "learning_rate": 0.0004245383040620622,
+      "loss": 3.5258,
+      "step": 27250
+    },
+    {
+      "epoch": 2.93832741362609,
+      "grad_norm": 0.5732911229133606,
+      "learning_rate": 0.0004242150630320009,
+      "loss": 3.5374,
+      "step": 27300
+    },
+    {
+      "epoch": 2.943708965665698,
+      "grad_norm": 0.6453855633735657,
+      "learning_rate": 0.0004238918220019394,
+      "loss": 3.527,
+      "step": 27350
+    },
+    {
+      "epoch": 2.949090517705306,
+      "grad_norm": 0.5842043161392212,
+      "learning_rate": 0.000423568580971878,
+      "loss": 3.5214,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9544720697449143,
+      "grad_norm": 0.6021690964698792,
+      "learning_rate": 0.00042324533994181655,
+      "loss": 3.5026,
+      "step": 27450
+    },
+    {
+      "epoch": 2.9598536217845224,
+      "grad_norm": 0.6042524576187134,
+      "learning_rate": 0.00042292209891175514,
+      "loss": 3.5169,
+      "step": 27500
+    },
+    {
+      "epoch": 2.965235173824131,
+      "grad_norm": 0.6471766829490662,
+      "learning_rate": 0.00042259885788169374,
+      "loss": 3.5289,
+      "step": 27550
+    },
+    {
+      "epoch": 2.970616725863739,
+      "grad_norm": 0.6328997611999512,
+      "learning_rate": 0.00042227561685163233,
+      "loss": 3.521,
+      "step": 27600
+    },
+    {
+      "epoch": 2.975998277903347,
+      "grad_norm": 0.6381612420082092,
+      "learning_rate": 0.00042195237582157093,
+      "loss": 3.5258,
+      "step": 27650
+    },
+    {
+      "epoch": 2.9813798299429557,
+      "grad_norm": 0.6035014986991882,
+      "learning_rate": 0.0004216291347915095,
+      "loss": 3.5295,
+      "step": 27700
+    },
+    {
+      "epoch": 2.986761381982564,
+      "grad_norm": 0.6743372678756714,
+      "learning_rate": 0.00042130589376144806,
+      "loss": 3.5261,
+      "step": 27750
+    },
+    {
+      "epoch": 2.992142934022172,
+      "grad_norm": 0.6017929315567017,
+      "learning_rate": 0.00042098265273138666,
+      "loss": 3.5172,
+      "step": 27800
+    },
+    {
+      "epoch": 2.9975244860617805,
+      "grad_norm": 0.6554933786392212,
+      "learning_rate": 0.0004206594117013252,
+      "loss": 3.5245,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0029060381013886,
+      "grad_norm": 0.6437206864356995,
+      "learning_rate": 0.00042033617067126385,
+      "loss": 3.4742,
+      "step": 27900
+    },
+    {
+      "epoch": 3.0082875901409967,
+      "grad_norm": 0.6490093469619751,
+      "learning_rate": 0.00042001292964120244,
+      "loss": 3.4372,
+      "step": 27950
+    },
+    {
+      "epoch": 3.0136691421806048,
+      "grad_norm": 0.6507807374000549,
+      "learning_rate": 0.000419689688611141,
+      "loss": 3.4461,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0136691421806048,
+      "eval_accuracy": 0.3706709565724186,
+      "eval_loss": 3.497730016708374,
+      "eval_runtime": 215.1239,
+      "eval_samples_per_second": 83.724,
+      "eval_steps_per_second": 5.234,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0190506942202133,
+      "grad_norm": 0.6191187500953674,
+      "learning_rate": 0.0004193664475810796,
+      "loss": 3.4181,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0244322462598214,
+      "grad_norm": 0.62087082862854,
+      "learning_rate": 0.00041904320655101817,
+      "loss": 3.4181,
+      "step": 28100
+    },
+    {
+      "epoch": 3.0298137982994295,
+      "grad_norm": 0.6341911554336548,
+      "learning_rate": 0.0004187199655209567,
+      "loss": 3.4286,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0351953503390376,
+      "grad_norm": 0.6108666062355042,
+      "learning_rate": 0.00041839672449089536,
+      "loss": 3.4278,
+      "step": 28200
+    },
+    {
+      "epoch": 3.040576902378646,
+      "grad_norm": 0.5951593518257141,
+      "learning_rate": 0.00041807348346083395,
+      "loss": 3.441,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0459584544182543,
+      "grad_norm": 0.6405830383300781,
+      "learning_rate": 0.0004177502424307725,
+      "loss": 3.4542,
+      "step": 28300
+    },
+    {
+      "epoch": 3.0513400064578624,
+      "grad_norm": 0.6463958024978638,
+      "learning_rate": 0.0004174270014007111,
+      "loss": 3.4252,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0567215584974705,
+      "grad_norm": 0.5980671048164368,
+      "learning_rate": 0.00041710376037064963,
+      "loss": 3.4458,
+      "step": 28400
+    },
+    {
+      "epoch": 3.062103110537079,
+      "grad_norm": 0.6921752691268921,
+      "learning_rate": 0.0004167805193405883,
+      "loss": 3.4406,
+      "step": 28450
+    },
+    {
+      "epoch": 3.067484662576687,
+      "grad_norm": 0.6020042896270752,
+      "learning_rate": 0.0004164572783105269,
+      "loss": 3.4446,
+      "step": 28500
+    },
+    {
+      "epoch": 3.0728662146162953,
+      "grad_norm": 0.6640591025352478,
+      "learning_rate": 0.0004161340372804654,
+      "loss": 3.4635,
+      "step": 28550
+    },
+    {
+      "epoch": 3.0782477666559034,
+      "grad_norm": 0.6354085206985474,
+      "learning_rate": 0.000415810796250404,
+      "loss": 3.4555,
+      "step": 28600
+    },
+    {
+      "epoch": 3.083629318695512,
+      "grad_norm": 0.636901319026947,
+      "learning_rate": 0.0004154875552203426,
+      "loss": 3.4515,
+      "step": 28650
+    },
+    {
+      "epoch": 3.08901087073512,
+      "grad_norm": 0.6186822056770325,
+      "learning_rate": 0.00041516431419028114,
+      "loss": 3.4345,
+      "step": 28700
+    },
+    {
+      "epoch": 3.094392422774728,
+      "grad_norm": 0.6416160464286804,
+      "learning_rate": 0.0004148410731602198,
+      "loss": 3.45,
+      "step": 28750
+    },
+    {
+      "epoch": 3.0997739748143363,
+      "grad_norm": 0.6019704341888428,
+      "learning_rate": 0.0004145178321301584,
+      "loss": 3.4452,
+      "step": 28800
+    },
+    {
+      "epoch": 3.105155526853945,
+      "grad_norm": 0.6140844225883484,
+      "learning_rate": 0.00041419459110009693,
+      "loss": 3.4474,
+      "step": 28850
+    },
+    {
+      "epoch": 3.110537078893553,
+      "grad_norm": 0.6584010720252991,
+      "learning_rate": 0.0004138713500700355,
+      "loss": 3.442,
+      "step": 28900
+    },
+    {
+      "epoch": 3.115918630933161,
+      "grad_norm": 0.6763238906860352,
+      "learning_rate": 0.00041354810903997406,
+      "loss": 3.4475,
+      "step": 28950
+    },
+    {
+      "epoch": 3.121300182972769,
+      "grad_norm": 0.6160973906517029,
+      "learning_rate": 0.00041322486800991266,
+      "loss": 3.4491,
+      "step": 29000
+    },
+    {
+      "epoch": 3.121300182972769,
+      "eval_accuracy": 0.3712210661434233,
+      "eval_loss": 3.494593858718872,
+      "eval_runtime": 209.8562,
+      "eval_samples_per_second": 85.825,
+      "eval_steps_per_second": 5.366,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266817350123777,
+      "grad_norm": 0.6309645771980286,
+      "learning_rate": 0.0004129016269798513,
+      "loss": 3.452,
+      "step": 29050
+    },
+    {
+      "epoch": 3.132063287051986,
+      "grad_norm": 0.616537868976593,
+      "learning_rate": 0.00041257838594978985,
+      "loss": 3.4708,
+      "step": 29100
+    },
+    {
+      "epoch": 3.137444839091594,
+      "grad_norm": 0.6185223460197449,
+      "learning_rate": 0.00041225514491972844,
+      "loss": 3.455,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1428263911312024,
+      "grad_norm": 0.6138353943824768,
+      "learning_rate": 0.00041193190388966704,
+      "loss": 3.4406,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1482079431708105,
+      "grad_norm": 0.608535647392273,
+      "learning_rate": 0.0004116086628596056,
+      "loss": 3.4461,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1535894952104186,
+      "grad_norm": 0.6112627387046814,
+      "learning_rate": 0.0004112854218295442,
+      "loss": 3.4395,
+      "step": 29300
+    },
+    {
+      "epoch": 3.1589710472500268,
+      "grad_norm": 0.5768312811851501,
+      "learning_rate": 0.0004109621807994828,
+      "loss": 3.4302,
+      "step": 29350
+    },
+    {
+      "epoch": 3.1643525992896353,
+      "grad_norm": 0.6325687170028687,
+      "learning_rate": 0.00041063893976942136,
+      "loss": 3.4465,
+      "step": 29400
+    },
+    {
+      "epoch": 3.1697341513292434,
+      "grad_norm": 0.6342477202415466,
+      "learning_rate": 0.00041031569873935996,
+      "loss": 3.442,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1751157033688515,
+      "grad_norm": 0.6336598992347717,
+      "learning_rate": 0.0004099924577092985,
+      "loss": 3.4462,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1804972554084596,
+      "grad_norm": 0.607629656791687,
+      "learning_rate": 0.0004096692166792371,
+      "loss": 3.4299,
+      "step": 29550
+    },
+    {
+      "epoch": 3.185878807448068,
+      "grad_norm": 0.6477324366569519,
+      "learning_rate": 0.00040934597564917574,
+      "loss": 3.448,
+      "step": 29600
+    },
+    {
+      "epoch": 3.1912603594876763,
+      "grad_norm": 0.6741542220115662,
+      "learning_rate": 0.0004090227346191143,
+      "loss": 3.4596,
+      "step": 29650
+    },
+    {
+      "epoch": 3.1966419115272844,
+      "grad_norm": 0.6519025564193726,
+      "learning_rate": 0.0004086994935890529,
+      "loss": 3.4715,
+      "step": 29700
+    },
+    {
+      "epoch": 3.2020234635668925,
+      "grad_norm": 0.6172560453414917,
+      "learning_rate": 0.00040837625255899147,
+      "loss": 3.4494,
+      "step": 29750
+    },
+    {
+      "epoch": 3.207405015606501,
+      "grad_norm": 0.6829916834831238,
+      "learning_rate": 0.00040805301152893,
+      "loss": 3.4398,
+      "step": 29800
+    },
+    {
+      "epoch": 3.212786567646109,
+      "grad_norm": 0.6081319451332092,
+      "learning_rate": 0.0004077297704988686,
+      "loss": 3.446,
+      "step": 29850
+    },
+    {
+      "epoch": 3.2181681196857173,
+      "grad_norm": 0.6863629817962646,
+      "learning_rate": 0.00040740652946880725,
+      "loss": 3.4471,
+      "step": 29900
+    },
+    {
+      "epoch": 3.2235496717253254,
+      "grad_norm": 0.6253100633621216,
+      "learning_rate": 0.0004070832884387458,
+      "loss": 3.4593,
+      "step": 29950
+    },
+    {
+      "epoch": 3.228931223764934,
+      "grad_norm": 0.61070317029953,
+      "learning_rate": 0.0004067600474086844,
+      "loss": 3.4684,
+      "step": 30000
+    },
+    {
+      "epoch": 3.228931223764934,
+      "eval_accuracy": 0.3713381939564873,
+      "eval_loss": 3.4922268390655518,
+      "eval_runtime": 208.2583,
+      "eval_samples_per_second": 86.484,
+      "eval_steps_per_second": 5.407,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234312775804542,
+      "grad_norm": 0.6275219321250916,
+      "learning_rate": 0.00040643680637862293,
+      "loss": 3.4539,
+      "step": 30050
+    },
+    {
+      "epoch": 3.23969432784415,
+      "grad_norm": 0.6307212710380554,
+      "learning_rate": 0.0004061135653485615,
+      "loss": 3.483,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2450758798837587,
+      "grad_norm": 0.6696796417236328,
+      "learning_rate": 0.0004057903243185001,
+      "loss": 3.4521,
+      "step": 30150
+    },
+    {
+      "epoch": 3.250457431923367,
+      "grad_norm": 0.6193971037864685,
+      "learning_rate": 0.0004054670832884387,
+      "loss": 3.4574,
+      "step": 30200
+    },
+    {
+      "epoch": 3.255838983962975,
+      "grad_norm": 0.6391464471817017,
+      "learning_rate": 0.0004051438422583773,
+      "loss": 3.4539,
+      "step": 30250
+    },
+    {
+      "epoch": 3.261220536002583,
+      "grad_norm": 0.641261100769043,
+      "learning_rate": 0.0004048206012283159,
+      "loss": 3.4772,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2666020880421915,
+      "grad_norm": 0.6269187331199646,
+      "learning_rate": 0.00040449736019825444,
+      "loss": 3.4534,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2719836400817996,
+      "grad_norm": 0.6142138242721558,
+      "learning_rate": 0.00040417411916819304,
+      "loss": 3.4688,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2773651921214078,
+      "grad_norm": 0.6109180450439453,
+      "learning_rate": 0.0004038508781381317,
+      "loss": 3.4442,
+      "step": 30450
+    },
+    {
+      "epoch": 3.282746744161016,
+      "grad_norm": 0.6190860867500305,
+      "learning_rate": 0.0004035276371080702,
+      "loss": 3.4456,
+      "step": 30500
+    },
+    {
+      "epoch": 3.2881282962006244,
+      "grad_norm": 0.6348389983177185,
+      "learning_rate": 0.0004032043960780088,
+      "loss": 3.4503,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2935098482402325,
+      "grad_norm": 0.669296145439148,
+      "learning_rate": 0.00040288115504794736,
+      "loss": 3.4602,
+      "step": 30600
+    },
+    {
+      "epoch": 3.2988914002798406,
+      "grad_norm": 0.6249747276306152,
+      "learning_rate": 0.00040255791401788596,
+      "loss": 3.4655,
+      "step": 30650
+    },
+    {
+      "epoch": 3.304272952319449,
+      "grad_norm": 0.577464759349823,
+      "learning_rate": 0.00040223467298782455,
+      "loss": 3.4727,
+      "step": 30700
+    },
+    {
+      "epoch": 3.3096545043590573,
+      "grad_norm": 0.6019587516784668,
+      "learning_rate": 0.00040191143195776315,
+      "loss": 3.4585,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3150360563986654,
+      "grad_norm": 0.601098895072937,
+      "learning_rate": 0.00040158819092770174,
+      "loss": 3.4467,
+      "step": 30800
+    },
+    {
+      "epoch": 3.3204176084382735,
+      "grad_norm": 0.6603884696960449,
+      "learning_rate": 0.00040127141471824155,
+      "loss": 3.4478,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3257991604778816,
+      "grad_norm": 0.6579861044883728,
+      "learning_rate": 0.00040094817368818014,
+      "loss": 3.4784,
+      "step": 30900
+    },
+    {
+      "epoch": 3.33118071251749,
+      "grad_norm": 0.622153103351593,
+      "learning_rate": 0.0004006249326581187,
+      "loss": 3.469,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3365622645570983,
+      "grad_norm": 0.6163874864578247,
+      "learning_rate": 0.0004003016916280573,
+      "loss": 3.4662,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3365622645570983,
+      "eval_accuracy": 0.37212973024421586,
+      "eval_loss": 3.4844980239868164,
+      "eval_runtime": 207.7673,
+      "eval_samples_per_second": 86.688,
+      "eval_steps_per_second": 5.42,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3419438165967064,
+      "grad_norm": 0.6363682746887207,
+      "learning_rate": 0.00039997845059799587,
+      "loss": 3.4479,
+      "step": 31050
+    },
+    {
+      "epoch": 3.347325368636315,
+      "grad_norm": 0.6347747445106506,
+      "learning_rate": 0.00039965520956793447,
+      "loss": 3.4736,
+      "step": 31100
+    },
+    {
+      "epoch": 3.352706920675923,
+      "grad_norm": 0.6099669933319092,
+      "learning_rate": 0.00039933196853787306,
+      "loss": 3.4788,
+      "step": 31150
+    },
+    {
+      "epoch": 3.358088472715531,
+      "grad_norm": 0.6515194177627563,
+      "learning_rate": 0.00039900872750781166,
+      "loss": 3.4613,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3634700247551392,
+      "grad_norm": 0.6574184894561768,
+      "learning_rate": 0.0003986854864777502,
+      "loss": 3.4714,
+      "step": 31250
+    },
+    {
+      "epoch": 3.368851576794748,
+      "grad_norm": 0.6157483458518982,
+      "learning_rate": 0.0003983622454476888,
+      "loss": 3.4752,
+      "step": 31300
+    },
+    {
+      "epoch": 3.374233128834356,
+      "grad_norm": 0.632178544998169,
+      "learning_rate": 0.00039803900441762733,
+      "loss": 3.4579,
+      "step": 31350
+    },
+    {
+      "epoch": 3.379614680873964,
+      "grad_norm": 0.6561319231987,
+      "learning_rate": 0.000397715763387566,
+      "loss": 3.4659,
+      "step": 31400
+    },
+    {
+      "epoch": 3.384996232913572,
+      "grad_norm": 0.6884437203407288,
+      "learning_rate": 0.0003973925223575046,
+      "loss": 3.4761,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3903777849531807,
+      "grad_norm": 0.6156170964241028,
+      "learning_rate": 0.0003970692813274431,
+      "loss": 3.4604,
+      "step": 31500
+    },
+    {
+      "epoch": 3.3957593369927888,
+      "grad_norm": 0.7083854675292969,
+      "learning_rate": 0.0003967460402973817,
+      "loss": 3.465,
+      "step": 31550
+    },
+    {
+      "epoch": 3.401140889032397,
+      "grad_norm": 0.6180551648139954,
+      "learning_rate": 0.0003964227992673203,
+      "loss": 3.4453,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4065224410720054,
+      "grad_norm": 0.6462782621383667,
+      "learning_rate": 0.00039609955823725884,
+      "loss": 3.4589,
+      "step": 31650
+    },
+    {
+      "epoch": 3.4119039931116135,
+      "grad_norm": 0.6157835721969604,
+      "learning_rate": 0.0003957763172071975,
+      "loss": 3.462,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4172855451512216,
+      "grad_norm": 0.6632956862449646,
+      "learning_rate": 0.0003954530761771361,
+      "loss": 3.4574,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4226670971908297,
+      "grad_norm": 0.5914563536643982,
+      "learning_rate": 0.00039512983514707463,
+      "loss": 3.4395,
+      "step": 31800
+    },
+    {
+      "epoch": 3.428048649230438,
+      "grad_norm": 0.6110662817955017,
+      "learning_rate": 0.0003948065941170132,
+      "loss": 3.4704,
+      "step": 31850
+    },
+    {
+      "epoch": 3.4334302012700464,
+      "grad_norm": 0.6194736361503601,
+      "learning_rate": 0.00039448335308695176,
+      "loss": 3.4669,
+      "step": 31900
+    },
+    {
+      "epoch": 3.4388117533096545,
+      "grad_norm": 0.5982261896133423,
+      "learning_rate": 0.00039416011205689036,
+      "loss": 3.4541,
+      "step": 31950
+    },
+    {
+      "epoch": 3.4441933053492626,
+      "grad_norm": 0.5901939868927002,
+      "learning_rate": 0.000393836871026829,
+      "loss": 3.4454,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4441933053492626,
+      "eval_accuracy": 0.3730818555001341,
+      "eval_loss": 3.4764976501464844,
+      "eval_runtime": 194.7932,
+      "eval_samples_per_second": 92.462,
+      "eval_steps_per_second": 5.78,
+      "step": 32000
+    },
+    {
+      "epoch": 3.449574857388871,
+      "grad_norm": 0.633826732635498,
+      "learning_rate": 0.00039351362999676755,
+      "loss": 3.4546,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4549564094284793,
+      "grad_norm": 0.6311871409416199,
+      "learning_rate": 0.00039319038896670614,
+      "loss": 3.4664,
+      "step": 32100
+    },
+    {
+      "epoch": 3.4603379614680874,
+      "grad_norm": 0.6177763342857361,
+      "learning_rate": 0.00039286714793664474,
+      "loss": 3.4497,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4657195135076955,
+      "grad_norm": 0.6959267258644104,
+      "learning_rate": 0.0003925439069065833,
+      "loss": 3.461,
+      "step": 32200
+    },
+    {
+      "epoch": 3.471101065547304,
+      "grad_norm": 0.6201044321060181,
+      "learning_rate": 0.0003922206658765219,
+      "loss": 3.4577,
+      "step": 32250
+    },
+    {
+      "epoch": 3.476482617586912,
+      "grad_norm": 0.6961297988891602,
+      "learning_rate": 0.0003918974248464605,
+      "loss": 3.4482,
+      "step": 32300
+    },
+    {
+      "epoch": 3.4818641696265202,
+      "grad_norm": 0.6520218253135681,
+      "learning_rate": 0.00039157418381639906,
+      "loss": 3.4618,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4872457216661283,
+      "grad_norm": 0.6764187216758728,
+      "learning_rate": 0.00039125094278633766,
+      "loss": 3.4571,
+      "step": 32400
+    },
+    {
+      "epoch": 3.492627273705737,
+      "grad_norm": 0.6045281291007996,
+      "learning_rate": 0.0003909277017562762,
+      "loss": 3.464,
+      "step": 32450
+    },
+    {
+      "epoch": 3.498008825745345,
+      "grad_norm": 0.6529583930969238,
+      "learning_rate": 0.0003906044607262148,
+      "loss": 3.4465,
+      "step": 32500
+    },
+    {
+      "epoch": 3.503390377784953,
+      "grad_norm": 0.6828312277793884,
+      "learning_rate": 0.00039028121969615344,
+      "loss": 3.4478,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5087719298245617,
+      "grad_norm": 0.6279124021530151,
+      "learning_rate": 0.000389957978666092,
+      "loss": 3.4532,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5141534818641698,
+      "grad_norm": 0.6408252120018005,
+      "learning_rate": 0.0003896347376360306,
+      "loss": 3.461,
+      "step": 32650
+    },
+    {
+      "epoch": 3.519535033903778,
+      "grad_norm": 0.6262297034263611,
+      "learning_rate": 0.0003893114966059691,
+      "loss": 3.4685,
+      "step": 32700
+    },
+    {
+      "epoch": 3.524916585943386,
+      "grad_norm": 0.6279938220977783,
+      "learning_rate": 0.0003889882555759077,
+      "loss": 3.4681,
+      "step": 32750
+    },
+    {
+      "epoch": 3.530298137982994,
+      "grad_norm": 0.6286149621009827,
+      "learning_rate": 0.0003886650145458463,
+      "loss": 3.4685,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5356796900226026,
+      "grad_norm": 0.6136440634727478,
+      "learning_rate": 0.0003883417735157849,
+      "loss": 3.4395,
+      "step": 32850
+    },
+    {
+      "epoch": 3.5410612420622107,
+      "grad_norm": 0.6475735902786255,
+      "learning_rate": 0.0003880185324857235,
+      "loss": 3.4601,
+      "step": 32900
+    },
+    {
+      "epoch": 3.546442794101819,
+      "grad_norm": 0.6448686718940735,
+      "learning_rate": 0.0003877017562762633,
+      "loss": 3.4557,
+      "step": 32950
+    },
+    {
+      "epoch": 3.5518243461414274,
+      "grad_norm": 0.6235514283180237,
+      "learning_rate": 0.0003873785152462019,
+      "loss": 3.4605,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5518243461414274,
+      "eval_accuracy": 0.3741357885119346,
+      "eval_loss": 3.4696691036224365,
+      "eval_runtime": 204.336,
+      "eval_samples_per_second": 88.144,
+      "eval_steps_per_second": 5.511,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5572058981810355,
+      "grad_norm": 0.6031110882759094,
+      "learning_rate": 0.0003870552742161405,
+      "loss": 3.4576,
+      "step": 33050
+    },
+    {
+      "epoch": 3.5625874502206436,
+      "grad_norm": 0.6580427289009094,
+      "learning_rate": 0.00038673203318607903,
+      "loss": 3.4585,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5679690022602517,
+      "grad_norm": 0.6172491908073425,
+      "learning_rate": 0.0003864087921560176,
+      "loss": 3.445,
+      "step": 33150
+    },
+    {
+      "epoch": 3.57335055429986,
+      "grad_norm": 0.6818593144416809,
+      "learning_rate": 0.0003860855511259563,
+      "loss": 3.4474,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5787321063394684,
+      "grad_norm": 0.6435357332229614,
+      "learning_rate": 0.0003857623100958948,
+      "loss": 3.4501,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5841136583790765,
+      "grad_norm": 0.639147937297821,
+      "learning_rate": 0.0003854390690658334,
+      "loss": 3.481,
+      "step": 33300
+    },
+    {
+      "epoch": 3.5894952104186846,
+      "grad_norm": 0.6366181373596191,
+      "learning_rate": 0.00038511582803577195,
+      "loss": 3.4775,
+      "step": 33350
+    },
+    {
+      "epoch": 3.594876762458293,
+      "grad_norm": 0.6355386972427368,
+      "learning_rate": 0.00038479258700571054,
+      "loss": 3.4591,
+      "step": 33400
+    },
+    {
+      "epoch": 3.6002583144979012,
+      "grad_norm": 0.6315577030181885,
+      "learning_rate": 0.00038446934597564914,
+      "loss": 3.4535,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6056398665375093,
+      "grad_norm": 0.6854292750358582,
+      "learning_rate": 0.00038414610494558773,
+      "loss": 3.4665,
+      "step": 33500
+    },
+    {
+      "epoch": 3.611021418577118,
+      "grad_norm": 0.680067777633667,
+      "learning_rate": 0.00038382286391552633,
+      "loss": 3.4476,
+      "step": 33550
+    },
+    {
+      "epoch": 3.616402970616726,
+      "grad_norm": 0.6212747693061829,
+      "learning_rate": 0.0003834996228854649,
+      "loss": 3.4603,
+      "step": 33600
+    },
+    {
+      "epoch": 3.621784522656334,
+      "grad_norm": 0.6352766156196594,
+      "learning_rate": 0.00038317638185540346,
+      "loss": 3.452,
+      "step": 33650
+    },
+    {
+      "epoch": 3.627166074695942,
+      "grad_norm": 0.6111792922019958,
+      "learning_rate": 0.00038285314082534206,
+      "loss": 3.4636,
+      "step": 33700
+    },
+    {
+      "epoch": 3.6325476267355503,
+      "grad_norm": 0.6384958028793335,
+      "learning_rate": 0.0003825298997952806,
+      "loss": 3.4458,
+      "step": 33750
+    },
+    {
+      "epoch": 3.637929178775159,
+      "grad_norm": 0.6627931594848633,
+      "learning_rate": 0.00038220665876521925,
+      "loss": 3.4675,
+      "step": 33800
+    },
+    {
+      "epoch": 3.643310730814767,
+      "grad_norm": 0.7328792810440063,
+      "learning_rate": 0.00038188341773515784,
+      "loss": 3.434,
+      "step": 33850
+    },
+    {
+      "epoch": 3.648692282854375,
+      "grad_norm": 0.7406517863273621,
+      "learning_rate": 0.0003815601767050964,
+      "loss": 3.4399,
+      "step": 33900
+    },
+    {
+      "epoch": 3.6540738348939836,
+      "grad_norm": 0.6318268775939941,
+      "learning_rate": 0.000381236935675035,
+      "loss": 3.4448,
+      "step": 33950
+    },
+    {
+      "epoch": 3.6594553869335917,
+      "grad_norm": 0.6655885577201843,
+      "learning_rate": 0.0003809136946449735,
+      "loss": 3.4428,
+      "step": 34000
+    },
+    {
+      "epoch": 3.6594553869335917,
+      "eval_accuracy": 0.3745497560145078,
+      "eval_loss": 3.463956594467163,
+      "eval_runtime": 207.206,
+      "eval_samples_per_second": 86.923,
+      "eval_steps_per_second": 5.434,
+      "step": 34000
+    },
+    {
+      "epoch": 3.6648369389732,
+      "grad_norm": 0.6424694061279297,
+      "learning_rate": 0.00038059045361491217,
+      "loss": 3.4602,
+      "step": 34050
+    },
+    {
+      "epoch": 3.670218491012808,
+      "grad_norm": 0.6229196190834045,
+      "learning_rate": 0.00038026721258485076,
+      "loss": 3.463,
+      "step": 34100
+    },
+    {
+      "epoch": 3.675600043052416,
+      "grad_norm": 0.6844025254249573,
+      "learning_rate": 0.0003799439715547893,
+      "loss": 3.4696,
+      "step": 34150
+    },
+    {
+      "epoch": 3.6809815950920246,
+      "grad_norm": 0.707100510597229,
+      "learning_rate": 0.0003796207305247279,
+      "loss": 3.465,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6863631471316327,
+      "grad_norm": 0.6834603548049927,
+      "learning_rate": 0.0003792974894946665,
+      "loss": 3.4657,
+      "step": 34250
+    },
+    {
+      "epoch": 3.691744699171241,
+      "grad_norm": 0.6267171502113342,
+      "learning_rate": 0.00037897424846460503,
+      "loss": 3.4584,
+      "step": 34300
+    },
+    {
+      "epoch": 3.6971262512108494,
+      "grad_norm": 0.6426784992218018,
+      "learning_rate": 0.0003786510074345437,
+      "loss": 3.4654,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7025078032504575,
+      "grad_norm": 0.697043776512146,
+      "learning_rate": 0.0003783277664044823,
+      "loss": 3.451,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7078893552900656,
+      "grad_norm": 0.6688522100448608,
+      "learning_rate": 0.0003780045253744208,
+      "loss": 3.4714,
+      "step": 34450
+    },
+    {
+      "epoch": 3.713270907329674,
+      "grad_norm": 0.6643778085708618,
+      "learning_rate": 0.0003776812843443594,
+      "loss": 3.4316,
+      "step": 34500
+    },
+    {
+      "epoch": 3.7186524593692822,
+      "grad_norm": 0.6475136280059814,
+      "learning_rate": 0.00037735804331429795,
+      "loss": 3.4301,
+      "step": 34550
+    },
+    {
+      "epoch": 3.7240340114088903,
+      "grad_norm": 0.6495978236198425,
+      "learning_rate": 0.00037703480228423654,
+      "loss": 3.4566,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7294155634484984,
+      "grad_norm": 0.6580095887184143,
+      "learning_rate": 0.0003767115612541752,
+      "loss": 3.4636,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7347971154881066,
+      "grad_norm": 0.6825853586196899,
+      "learning_rate": 0.00037638832022411373,
+      "loss": 3.459,
+      "step": 34700
+    },
+    {
+      "epoch": 3.740178667527715,
+      "grad_norm": 0.6098463535308838,
+      "learning_rate": 0.00037606507919405233,
+      "loss": 3.4392,
+      "step": 34750
+    },
+    {
+      "epoch": 3.745560219567323,
+      "grad_norm": 0.7137957811355591,
+      "learning_rate": 0.0003757418381639909,
+      "loss": 3.4407,
+      "step": 34800
+    },
+    {
+      "epoch": 3.7509417716069313,
+      "grad_norm": 0.6484520435333252,
+      "learning_rate": 0.00037541859713392946,
+      "loss": 3.4611,
+      "step": 34850
+    },
+    {
+      "epoch": 3.75632332364654,
+      "grad_norm": 0.7095451354980469,
+      "learning_rate": 0.0003750953561038681,
+      "loss": 3.4435,
+      "step": 34900
+    },
+    {
+      "epoch": 3.761704875686148,
+      "grad_norm": 0.625594973564148,
+      "learning_rate": 0.0003747721150738067,
+      "loss": 3.4527,
+      "step": 34950
+    },
+    {
+      "epoch": 3.767086427725756,
+      "grad_norm": 0.6271294355392456,
+      "learning_rate": 0.00037444887404374525,
+      "loss": 3.458,
+      "step": 35000
+    },
+    {
+      "epoch": 3.767086427725756,
+      "eval_accuracy": 0.37491146147804216,
+      "eval_loss": 3.45908522605896,
+      "eval_runtime": 199.5792,
+      "eval_samples_per_second": 90.245,
+      "eval_steps_per_second": 5.642,
+      "step": 35000
+    },
+    {
+      "epoch": 3.772467979765364,
+      "grad_norm": 0.6571464538574219,
+      "learning_rate": 0.00037412563301368384,
+      "loss": 3.4627,
+      "step": 35050
+    },
+    {
+      "epoch": 3.7778495318049723,
+      "grad_norm": 0.6304094791412354,
+      "learning_rate": 0.0003738023919836224,
+      "loss": 3.4734,
+      "step": 35100
+    },
+    {
+      "epoch": 3.783231083844581,
+      "grad_norm": 0.7219865322113037,
+      "learning_rate": 0.000373479150953561,
+      "loss": 3.4536,
+      "step": 35150
+    },
+    {
+      "epoch": 3.788612635884189,
+      "grad_norm": 0.6388824582099915,
+      "learning_rate": 0.0003731559099234996,
+      "loss": 3.4584,
+      "step": 35200
+    },
+    {
+      "epoch": 3.793994187923797,
+      "grad_norm": 0.6524561047554016,
+      "learning_rate": 0.00037283266889343817,
+      "loss": 3.4621,
+      "step": 35250
+    },
+    {
+      "epoch": 3.7993757399634056,
+      "grad_norm": 0.6496708989143372,
+      "learning_rate": 0.00037250942786337676,
+      "loss": 3.4551,
+      "step": 35300
+    },
+    {
+      "epoch": 3.8047572920030137,
+      "grad_norm": 0.668373167514801,
+      "learning_rate": 0.00037218618683331536,
+      "loss": 3.4609,
+      "step": 35350
+    },
+    {
+      "epoch": 3.810138844042622,
+      "grad_norm": 0.6569227576255798,
+      "learning_rate": 0.0003718629458032539,
+      "loss": 3.4436,
+      "step": 35400
+    },
+    {
+      "epoch": 3.8155203960822304,
+      "grad_norm": 0.6292812824249268,
+      "learning_rate": 0.0003715397047731925,
+      "loss": 3.4514,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8209019481218385,
+      "grad_norm": 0.6967467665672302,
+      "learning_rate": 0.00037121646374313114,
+      "loss": 3.4759,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8262835001614466,
+      "grad_norm": 0.6586191058158875,
+      "learning_rate": 0.0003708932227130697,
+      "loss": 3.46,
+      "step": 35550
+    },
+    {
+      "epoch": 3.8316650522010547,
+      "grad_norm": 0.6478423476219177,
+      "learning_rate": 0.0003705699816830083,
+      "loss": 3.4529,
+      "step": 35600
+    },
+    {
+      "epoch": 3.837046604240663,
+      "grad_norm": 0.6029471755027771,
+      "learning_rate": 0.0003702467406529468,
+      "loss": 3.4454,
+      "step": 35650
+    },
+    {
+      "epoch": 3.8424281562802713,
+      "grad_norm": 0.6288083791732788,
+      "learning_rate": 0.0003699234996228854,
+      "loss": 3.4531,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8478097083198795,
+      "grad_norm": 0.6120226383209229,
+      "learning_rate": 0.000369600258592824,
+      "loss": 3.4688,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8531912603594876,
+      "grad_norm": 0.6600449085235596,
+      "learning_rate": 0.0003692770175627626,
+      "loss": 3.4735,
+      "step": 35800
+    },
+    {
+      "epoch": 3.858572812399096,
+      "grad_norm": 0.6423937678337097,
+      "learning_rate": 0.0003689537765327012,
+      "loss": 3.4451,
+      "step": 35850
+    },
+    {
+      "epoch": 3.863954364438704,
+      "grad_norm": 0.644229531288147,
+      "learning_rate": 0.000368637000323241,
+      "loss": 3.468,
+      "step": 35900
+    },
+    {
+      "epoch": 3.8693359164783123,
+      "grad_norm": 0.6580808162689209,
+      "learning_rate": 0.0003683137592931796,
+      "loss": 3.4742,
+      "step": 35950
+    },
+    {
+      "epoch": 3.8747174685179204,
+      "grad_norm": 0.6204085350036621,
+      "learning_rate": 0.00036799051826311814,
+      "loss": 3.4501,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8747174685179204,
+      "eval_accuracy": 0.37576525587048837,
+      "eval_loss": 3.4518706798553467,
+      "eval_runtime": 205.4042,
+      "eval_samples_per_second": 87.686,
+      "eval_steps_per_second": 5.482,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8800990205575285,
+      "grad_norm": 0.6474447250366211,
+      "learning_rate": 0.00036766727723305673,
+      "loss": 3.4434,
+      "step": 36050
+    },
+    {
+      "epoch": 3.885480572597137,
+      "grad_norm": 0.6868647336959839,
+      "learning_rate": 0.0003673440362029953,
+      "loss": 3.4569,
+      "step": 36100
+    },
+    {
+      "epoch": 3.890862124636745,
+      "grad_norm": 0.6220848560333252,
+      "learning_rate": 0.0003670207951729339,
+      "loss": 3.4423,
+      "step": 36150
+    },
+    {
+      "epoch": 3.8962436766763533,
+      "grad_norm": 0.7336555123329163,
+      "learning_rate": 0.0003666975541428725,
+      "loss": 3.4505,
+      "step": 36200
+    },
+    {
+      "epoch": 3.901625228715962,
+      "grad_norm": 0.6814396381378174,
+      "learning_rate": 0.0003663743131128111,
+      "loss": 3.453,
+      "step": 36250
+    },
+    {
+      "epoch": 3.90700678075557,
+      "grad_norm": 0.6469142436981201,
+      "learning_rate": 0.00036605107208274965,
+      "loss": 3.4619,
+      "step": 36300
+    },
+    {
+      "epoch": 3.912388332795178,
+      "grad_norm": 1.2596371173858643,
+      "learning_rate": 0.00036572783105268824,
+      "loss": 3.4511,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9177698848347866,
+      "grad_norm": 0.6139094233512878,
+      "learning_rate": 0.0003654045900226268,
+      "loss": 3.4416,
+      "step": 36400
+    },
+    {
+      "epoch": 3.9231514368743947,
+      "grad_norm": 0.6947947144508362,
+      "learning_rate": 0.00036508134899256543,
+      "loss": 3.4434,
+      "step": 36450
+    },
+    {
+      "epoch": 3.928532988914003,
+      "grad_norm": 0.6977193355560303,
+      "learning_rate": 0.00036475810796250403,
+      "loss": 3.4436,
+      "step": 36500
+    },
+    {
+      "epoch": 3.933914540953611,
+      "grad_norm": 0.6614829301834106,
+      "learning_rate": 0.00036443486693244257,
+      "loss": 3.4628,
+      "step": 36550
+    },
+    {
+      "epoch": 3.939296092993219,
+      "grad_norm": 0.629019558429718,
+      "learning_rate": 0.00036411162590238116,
+      "loss": 3.446,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9446776450328276,
+      "grad_norm": 0.6478345394134521,
+      "learning_rate": 0.00036378838487231976,
+      "loss": 3.46,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9500591970724357,
+      "grad_norm": 0.7331146597862244,
+      "learning_rate": 0.00036346514384225835,
+      "loss": 3.4329,
+      "step": 36700
+    },
+    {
+      "epoch": 3.955440749112044,
+      "grad_norm": 0.6980277299880981,
+      "learning_rate": 0.00036314190281219695,
+      "loss": 3.4525,
+      "step": 36750
+    },
+    {
+      "epoch": 3.9608223011516523,
+      "grad_norm": 0.6497023105621338,
+      "learning_rate": 0.00036281866178213554,
+      "loss": 3.4378,
+      "step": 36800
+    },
+    {
+      "epoch": 3.9662038531912605,
+      "grad_norm": 0.6118669509887695,
+      "learning_rate": 0.0003624954207520741,
+      "loss": 3.4387,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9715854052308686,
+      "grad_norm": 0.6508960127830505,
+      "learning_rate": 0.0003621721797220127,
+      "loss": 3.4462,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9769669572704767,
+      "grad_norm": 0.6341844201087952,
+      "learning_rate": 0.0003618489386919512,
+      "loss": 3.4397,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9823485093100848,
+      "grad_norm": 0.6382229328155518,
+      "learning_rate": 0.00036152569766188987,
+      "loss": 3.4582,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9823485093100848,
+      "eval_accuracy": 0.37578046727478237,
+      "eval_loss": 3.4466753005981445,
+      "eval_runtime": 203.9996,
+      "eval_samples_per_second": 88.289,
+      "eval_steps_per_second": 5.52,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9877300613496933,
+      "grad_norm": 0.6708056926727295,
+      "learning_rate": 0.00036120245663182846,
+      "loss": 3.4533,
+      "step": 37050
+    },
+    {
+      "epoch": 3.9931116133893014,
+      "grad_norm": 0.7044811248779297,
+      "learning_rate": 0.000360879215601767,
+      "loss": 3.4423,
+      "step": 37100
+    },
+    {
+      "epoch": 3.9984931654289095,
+      "grad_norm": 0.6734232306480408,
+      "learning_rate": 0.0003605559745717056,
+      "loss": 3.4398,
+      "step": 37150
+    },
+    {
+      "epoch": 4.003874717468518,
+      "grad_norm": 0.624794602394104,
+      "learning_rate": 0.0003602327335416442,
+      "loss": 3.3979,
+      "step": 37200
+    },
+    {
+      "epoch": 4.009256269508126,
+      "grad_norm": 0.6523617506027222,
+      "learning_rate": 0.00035990949251158273,
+      "loss": 3.3628,
+      "step": 37250
+    },
+    {
+      "epoch": 4.014637821547734,
+      "grad_norm": 0.6801915764808655,
+      "learning_rate": 0.0003595862514815214,
+      "loss": 3.3562,
+      "step": 37300
+    },
+    {
+      "epoch": 4.020019373587343,
+      "grad_norm": 0.6472498178482056,
+      "learning_rate": 0.00035926301045146,
+      "loss": 3.3433,
+      "step": 37350
+    },
+    {
+      "epoch": 4.0254009256269505,
+      "grad_norm": 0.6406301259994507,
+      "learning_rate": 0.0003589397694213985,
+      "loss": 3.3586,
+      "step": 37400
+    },
+    {
+      "epoch": 4.030782477666559,
+      "grad_norm": 0.635485827922821,
+      "learning_rate": 0.0003586165283913371,
+      "loss": 3.3601,
+      "step": 37450
+    },
+    {
+      "epoch": 4.036164029706168,
+      "grad_norm": 0.6421555876731873,
+      "learning_rate": 0.00035829328736127565,
+      "loss": 3.3533,
+      "step": 37500
+    },
+    {
+      "epoch": 4.041545581745775,
+      "grad_norm": 0.6644110679626465,
+      "learning_rate": 0.00035797004633121425,
+      "loss": 3.3551,
+      "step": 37550
+    },
+    {
+      "epoch": 4.046927133785384,
+      "grad_norm": 0.6391462683677673,
+      "learning_rate": 0.0003576468053011529,
+      "loss": 3.3478,
+      "step": 37600
+    },
+    {
+      "epoch": 4.0523086858249915,
+      "grad_norm": 0.7233789563179016,
+      "learning_rate": 0.00035732356427109143,
+      "loss": 3.3716,
+      "step": 37650
+    },
+    {
+      "epoch": 4.0576902378646,
+      "grad_norm": 0.6202532052993774,
+      "learning_rate": 0.00035700032324103003,
+      "loss": 3.375,
+      "step": 37700
+    },
+    {
+      "epoch": 4.063071789904209,
+      "grad_norm": 0.6700003743171692,
+      "learning_rate": 0.0003566770822109686,
+      "loss": 3.3763,
+      "step": 37750
+    },
+    {
+      "epoch": 4.068453341943816,
+      "grad_norm": 0.6778683066368103,
+      "learning_rate": 0.00035635384118090716,
+      "loss": 3.3668,
+      "step": 37800
+    },
+    {
+      "epoch": 4.073834893983425,
+      "grad_norm": 0.6656447649002075,
+      "learning_rate": 0.0003560306001508458,
+      "loss": 3.3624,
+      "step": 37850
+    },
+    {
+      "epoch": 4.079216446023033,
+      "grad_norm": 0.6423748135566711,
+      "learning_rate": 0.0003557073591207844,
+      "loss": 3.3872,
+      "step": 37900
+    },
+    {
+      "epoch": 4.084597998062641,
+      "grad_norm": 0.6550871729850769,
+      "learning_rate": 0.00035538411809072295,
+      "loss": 3.3689,
+      "step": 37950
+    },
+    {
+      "epoch": 4.08997955010225,
+      "grad_norm": 0.7397666573524475,
+      "learning_rate": 0.00035506087706066154,
+      "loss": 3.3691,
+      "step": 38000
+    },
+    {
+      "epoch": 4.08997955010225,
+      "eval_accuracy": 0.37662122332069087,
+      "eval_loss": 3.448406934738159,
+      "eval_runtime": 209.7019,
+      "eval_samples_per_second": 85.889,
+      "eval_steps_per_second": 5.37,
+      "step": 38000
+    },
+    {
+      "epoch": 4.095361102141858,
+      "grad_norm": 0.6562075614929199,
+      "learning_rate": 0.00035474410085120135,
+      "loss": 3.3729,
+      "step": 38050
+    },
+    {
+      "epoch": 4.100742654181466,
+      "grad_norm": 0.663729190826416,
+      "learning_rate": 0.00035442085982113994,
+      "loss": 3.3756,
+      "step": 38100
+    },
+    {
+      "epoch": 4.106124206221074,
+      "grad_norm": 0.6242133378982544,
+      "learning_rate": 0.0003540976187910785,
+      "loss": 3.3834,
+      "step": 38150
+    },
+    {
+      "epoch": 4.111505758260682,
+      "grad_norm": 0.7042984962463379,
+      "learning_rate": 0.0003537743777610171,
+      "loss": 3.3753,
+      "step": 38200
+    },
+    {
+      "epoch": 4.1168873103002905,
+      "grad_norm": 0.6397444605827332,
+      "learning_rate": 0.00035345113673095573,
+      "loss": 3.3607,
+      "step": 38250
+    },
+    {
+      "epoch": 4.122268862339899,
+      "grad_norm": 0.6845241189002991,
+      "learning_rate": 0.00035312789570089427,
+      "loss": 3.3678,
+      "step": 38300
+    },
+    {
+      "epoch": 4.127650414379507,
+      "grad_norm": 0.6499135494232178,
+      "learning_rate": 0.00035280465467083286,
+      "loss": 3.3847,
+      "step": 38350
+    },
+    {
+      "epoch": 4.133031966419115,
+      "grad_norm": 0.6635656356811523,
+      "learning_rate": 0.0003524814136407714,
+      "loss": 3.3735,
+      "step": 38400
+    },
+    {
+      "epoch": 4.138413518458724,
+      "grad_norm": 0.6924629211425781,
+      "learning_rate": 0.00035215817261071,
+      "loss": 3.3783,
+      "step": 38450
+    },
+    {
+      "epoch": 4.1437950704983315,
+      "grad_norm": 0.6163055300712585,
+      "learning_rate": 0.00035183493158064865,
+      "loss": 3.3716,
+      "step": 38500
+    },
+    {
+      "epoch": 4.14917662253794,
+      "grad_norm": 0.6056356430053711,
+      "learning_rate": 0.0003515116905505872,
+      "loss": 3.3781,
+      "step": 38550
+    },
+    {
+      "epoch": 4.154558174577549,
+      "grad_norm": 0.6951189041137695,
+      "learning_rate": 0.0003511884495205258,
+      "loss": 3.3743,
+      "step": 38600
+    },
+    {
+      "epoch": 4.159939726617156,
+      "grad_norm": 0.6552825570106506,
+      "learning_rate": 0.0003508652084904644,
+      "loss": 3.3885,
+      "step": 38650
+    },
+    {
+      "epoch": 4.165321278656765,
+      "grad_norm": 0.6377124786376953,
+      "learning_rate": 0.0003505419674604029,
+      "loss": 3.3997,
+      "step": 38700
+    },
+    {
+      "epoch": 4.1707028306963725,
+      "grad_norm": 0.6340206861495972,
+      "learning_rate": 0.0003502187264303415,
+      "loss": 3.3831,
+      "step": 38750
+    },
+    {
+      "epoch": 4.176084382735981,
+      "grad_norm": 0.6724329590797424,
+      "learning_rate": 0.00034989548540028016,
+      "loss": 3.4015,
+      "step": 38800
+    },
+    {
+      "epoch": 4.18146593477559,
+      "grad_norm": 0.6731334924697876,
+      "learning_rate": 0.0003495722443702187,
+      "loss": 3.3766,
+      "step": 38850
+    },
+    {
+      "epoch": 4.186847486815197,
+      "grad_norm": 0.6341095566749573,
+      "learning_rate": 0.0003492490033401573,
+      "loss": 3.3741,
+      "step": 38900
+    },
+    {
+      "epoch": 4.192229038854806,
+      "grad_norm": 0.6485639214515686,
+      "learning_rate": 0.00034892576231009584,
+      "loss": 3.3827,
+      "step": 38950
+    },
+    {
+      "epoch": 4.197610590894414,
+      "grad_norm": 0.6944514513015747,
+      "learning_rate": 0.00034860252128003443,
+      "loss": 3.3847,
+      "step": 39000
+    },
+    {
+      "epoch": 4.197610590894414,
+      "eval_accuracy": 0.3769026343001303,
+      "eval_loss": 3.4444234371185303,
+      "eval_runtime": 194.3118,
+      "eval_samples_per_second": 92.691,
+      "eval_steps_per_second": 5.795,
+      "step": 39000
+    },
+    {
+      "epoch": 4.202992142934022,
+      "grad_norm": 0.6771103143692017,
+      "learning_rate": 0.000348279280249973,
+      "loss": 3.3822,
+      "step": 39050
+    },
+    {
+      "epoch": 4.208373694973631,
+      "grad_norm": 0.6469762325286865,
+      "learning_rate": 0.0003479560392199116,
+      "loss": 3.3847,
+      "step": 39100
+    },
+    {
+      "epoch": 4.213755247013238,
+      "grad_norm": 0.6654626727104187,
+      "learning_rate": 0.0003476327981898502,
+      "loss": 3.3962,
+      "step": 39150
+    },
+    {
+      "epoch": 4.219136799052847,
+      "grad_norm": 0.6649177074432373,
+      "learning_rate": 0.0003473095571597888,
+      "loss": 3.3866,
+      "step": 39200
+    },
+    {
+      "epoch": 4.224518351092455,
+      "grad_norm": 0.639963686466217,
+      "learning_rate": 0.00034698631612972735,
+      "loss": 3.3654,
+      "step": 39250
+    },
+    {
+      "epoch": 4.229899903132063,
+      "grad_norm": 0.7196988463401794,
+      "learning_rate": 0.00034666307509966594,
+      "loss": 3.3669,
+      "step": 39300
+    },
+    {
+      "epoch": 4.2352814551716715,
+      "grad_norm": 0.6603704690933228,
+      "learning_rate": 0.0003463398340696045,
+      "loss": 3.3879,
+      "step": 39350
+    },
+    {
+      "epoch": 4.24066300721128,
+      "grad_norm": 0.7158796191215515,
+      "learning_rate": 0.00034601659303954313,
+      "loss": 3.3925,
+      "step": 39400
+    },
+    {
+      "epoch": 4.246044559250888,
+      "grad_norm": 0.6354226469993591,
+      "learning_rate": 0.00034569335200948173,
+      "loss": 3.3954,
+      "step": 39450
+    },
+    {
+      "epoch": 4.251426111290496,
+      "grad_norm": 0.6703064441680908,
+      "learning_rate": 0.00034537011097942027,
+      "loss": 3.4026,
+      "step": 39500
+    },
+    {
+      "epoch": 4.256807663330104,
+      "grad_norm": 0.6066316962242126,
+      "learning_rate": 0.00034504686994935886,
+      "loss": 3.4001,
+      "step": 39550
+    },
+    {
+      "epoch": 4.2621892153697125,
+      "grad_norm": 0.6286178827285767,
+      "learning_rate": 0.00034472362891929746,
+      "loss": 3.3713,
+      "step": 39600
+    },
+    {
+      "epoch": 4.267570767409321,
+      "grad_norm": 0.6771307587623596,
+      "learning_rate": 0.00034440038788923605,
+      "loss": 3.3985,
+      "step": 39650
+    },
+    {
+      "epoch": 4.272952319448929,
+      "grad_norm": 0.6318666338920593,
+      "learning_rate": 0.00034407714685917465,
+      "loss": 3.381,
+      "step": 39700
+    },
+    {
+      "epoch": 4.278333871488537,
+      "grad_norm": 0.6529156565666199,
+      "learning_rate": 0.00034375390582911324,
+      "loss": 3.3848,
+      "step": 39750
+    },
+    {
+      "epoch": 4.283715423528146,
+      "grad_norm": 0.6864298582077026,
+      "learning_rate": 0.0003434306647990518,
+      "loss": 3.378,
+      "step": 39800
+    },
+    {
+      "epoch": 4.2890969755677535,
+      "grad_norm": 0.7262616753578186,
+      "learning_rate": 0.0003431074237689904,
+      "loss": 3.3899,
+      "step": 39850
+    },
+    {
+      "epoch": 4.294478527607362,
+      "grad_norm": 0.7095841765403748,
+      "learning_rate": 0.0003427841827389289,
+      "loss": 3.3793,
+      "step": 39900
+    },
+    {
+      "epoch": 4.299860079646971,
+      "grad_norm": 0.6717966794967651,
+      "learning_rate": 0.00034246094170886757,
+      "loss": 3.3844,
+      "step": 39950
+    },
+    {
+      "epoch": 4.305241631686578,
+      "grad_norm": 0.6544067859649658,
+      "learning_rate": 0.00034213770067880616,
+      "loss": 3.3714,
+      "step": 40000
+    },
+    {
+      "epoch": 4.305241631686578,
+      "eval_accuracy": 0.3771877394777555,
+      "eval_loss": 3.439746141433716,
+      "eval_runtime": 210.3742,
+      "eval_samples_per_second": 85.614,
+      "eval_steps_per_second": 5.352,
+      "step": 40000
+    },
+    {
+      "epoch": 4.310623183726187,
+      "grad_norm": 0.7182822227478027,
+      "learning_rate": 0.0003418144596487447,
+      "loss": 3.3831,
+      "step": 40050
+    },
+    {
+      "epoch": 4.3160047357657945,
+      "grad_norm": 0.6663074493408203,
+      "learning_rate": 0.0003414912186186833,
+      "loss": 3.3781,
+      "step": 40100
+    },
+    {
+      "epoch": 4.321386287805403,
+      "grad_norm": 0.7040979862213135,
+      "learning_rate": 0.0003411679775886219,
+      "loss": 3.3946,
+      "step": 40150
+    },
+    {
+      "epoch": 4.326767839845012,
+      "grad_norm": 0.6793686151504517,
+      "learning_rate": 0.00034084473655856043,
+      "loss": 3.3925,
+      "step": 40200
+    },
+    {
+      "epoch": 4.332149391884619,
+      "grad_norm": 0.6753449440002441,
+      "learning_rate": 0.0003405214955284991,
+      "loss": 3.3899,
+      "step": 40250
+    },
+    {
+      "epoch": 4.337530943924228,
+      "grad_norm": 0.6695329546928406,
+      "learning_rate": 0.0003401982544984377,
+      "loss": 3.3848,
+      "step": 40300
+    },
+    {
+      "epoch": 4.342912495963836,
+      "grad_norm": 1.0501422882080078,
+      "learning_rate": 0.0003398750134683762,
+      "loss": 3.3769,
+      "step": 40350
+    },
+    {
+      "epoch": 4.348294048003444,
+      "grad_norm": 0.6779940128326416,
+      "learning_rate": 0.0003395517724383148,
+      "loss": 3.3852,
+      "step": 40400
+    },
+    {
+      "epoch": 4.3536756000430525,
+      "grad_norm": 0.6622795462608337,
+      "learning_rate": 0.0003392349962288546,
+      "loss": 3.3954,
+      "step": 40450
+    },
+    {
+      "epoch": 4.359057152082661,
+      "grad_norm": 0.6417733430862427,
+      "learning_rate": 0.0003389117551987932,
+      "loss": 3.3999,
+      "step": 40500
+    },
+    {
+      "epoch": 4.364438704122269,
+      "grad_norm": 0.6586073040962219,
+      "learning_rate": 0.00033858851416873175,
+      "loss": 3.3942,
+      "step": 40550
+    },
+    {
+      "epoch": 4.369820256161877,
+      "grad_norm": 0.6641150116920471,
+      "learning_rate": 0.0003382652731386704,
+      "loss": 3.3892,
+      "step": 40600
+    },
+    {
+      "epoch": 4.375201808201485,
+      "grad_norm": 0.6859793066978455,
+      "learning_rate": 0.000337942032108609,
+      "loss": 3.3936,
+      "step": 40650
+    },
+    {
+      "epoch": 4.3805833602410935,
+      "grad_norm": 0.6642147302627563,
+      "learning_rate": 0.00033761879107854754,
+      "loss": 3.3869,
+      "step": 40700
+    },
+    {
+      "epoch": 4.385964912280702,
+      "grad_norm": 0.6813688278198242,
+      "learning_rate": 0.00033729555004848613,
+      "loss": 3.3792,
+      "step": 40750
+    },
+    {
+      "epoch": 4.39134646432031,
+      "grad_norm": 0.6705623865127563,
+      "learning_rate": 0.00033697230901842467,
+      "loss": 3.3931,
+      "step": 40800
+    },
+    {
+      "epoch": 4.396728016359918,
+      "grad_norm": 0.6627140045166016,
+      "learning_rate": 0.00033664906798836327,
+      "loss": 3.3971,
+      "step": 40850
+    },
+    {
+      "epoch": 4.402109568399527,
+      "grad_norm": 0.6509461998939514,
+      "learning_rate": 0.0003363258269583019,
+      "loss": 3.3875,
+      "step": 40900
+    },
+    {
+      "epoch": 4.4074911204391345,
+      "grad_norm": 0.7194676399230957,
+      "learning_rate": 0.00033600258592824046,
+      "loss": 3.3982,
+      "step": 40950
+    },
+    {
+      "epoch": 4.412872672478743,
+      "grad_norm": 0.6531248092651367,
+      "learning_rate": 0.00033567934489817905,
+      "loss": 3.3768,
+      "step": 41000
+    },
+    {
+      "epoch": 4.412872672478743,
+      "eval_accuracy": 0.3779571105763699,
+      "eval_loss": 3.437781810760498,
+      "eval_runtime": 205.598,
+      "eval_samples_per_second": 87.603,
+      "eval_steps_per_second": 5.477,
+      "step": 41000
+    },
+    {
+      "epoch": 4.418254224518351,
+      "grad_norm": 0.6822459101676941,
+      "learning_rate": 0.00033535610386811764,
+      "loss": 3.4082,
+      "step": 41050
+    },
+    {
+      "epoch": 4.423635776557959,
+      "grad_norm": 0.7048671841621399,
+      "learning_rate": 0.0003350328628380562,
+      "loss": 3.4037,
+      "step": 41100
+    },
+    {
+      "epoch": 4.429017328597568,
+      "grad_norm": 0.6404821276664734,
+      "learning_rate": 0.0003347096218079948,
+      "loss": 3.377,
+      "step": 41150
+    },
+    {
+      "epoch": 4.4343988806371755,
+      "grad_norm": 0.7114750742912292,
+      "learning_rate": 0.00033438638077793343,
+      "loss": 3.3984,
+      "step": 41200
+    },
+    {
+      "epoch": 4.439780432676784,
+      "grad_norm": 0.6833235025405884,
+      "learning_rate": 0.00033406313974787197,
+      "loss": 3.3948,
+      "step": 41250
+    },
+    {
+      "epoch": 4.445161984716393,
+      "grad_norm": 0.671991765499115,
+      "learning_rate": 0.00033373989871781056,
+      "loss": 3.3905,
+      "step": 41300
+    },
+    {
+      "epoch": 4.450543536756,
+      "grad_norm": 0.6879648566246033,
+      "learning_rate": 0.0003334166576877491,
+      "loss": 3.4125,
+      "step": 41350
+    },
+    {
+      "epoch": 4.455925088795609,
+      "grad_norm": 0.6425567865371704,
+      "learning_rate": 0.0003330934166576877,
+      "loss": 3.3877,
+      "step": 41400
+    },
+    {
+      "epoch": 4.461306640835216,
+      "grad_norm": 0.6497949361801147,
+      "learning_rate": 0.00033277017562762635,
+      "loss": 3.3787,
+      "step": 41450
+    },
+    {
+      "epoch": 4.466688192874825,
+      "grad_norm": 0.6711752414703369,
+      "learning_rate": 0.0003324469345975649,
+      "loss": 3.4,
+      "step": 41500
+    },
+    {
+      "epoch": 4.4720697449144335,
+      "grad_norm": 0.6858512163162231,
+      "learning_rate": 0.0003321236935675035,
+      "loss": 3.3806,
+      "step": 41550
+    },
+    {
+      "epoch": 4.477451296954041,
+      "grad_norm": 0.6588053107261658,
+      "learning_rate": 0.0003318004525374421,
+      "loss": 3.3769,
+      "step": 41600
+    },
+    {
+      "epoch": 4.48283284899365,
+      "grad_norm": 0.6885084509849548,
+      "learning_rate": 0.0003314772115073806,
+      "loss": 3.3833,
+      "step": 41650
+    },
+    {
+      "epoch": 4.488214401033258,
+      "grad_norm": 0.6567678451538086,
+      "learning_rate": 0.0003311539704773192,
+      "loss": 3.3795,
+      "step": 41700
+    },
+    {
+      "epoch": 4.493595953072866,
+      "grad_norm": 0.644980251789093,
+      "learning_rate": 0.00033083072944725786,
+      "loss": 3.3881,
+      "step": 41750
+    },
+    {
+      "epoch": 4.4989775051124745,
+      "grad_norm": 0.6543886065483093,
+      "learning_rate": 0.0003305074884171964,
+      "loss": 3.3881,
+      "step": 41800
+    },
+    {
+      "epoch": 4.504359057152083,
+      "grad_norm": 0.6759121417999268,
+      "learning_rate": 0.000330184247387135,
+      "loss": 3.3842,
+      "step": 41850
+    },
+    {
+      "epoch": 4.509740609191691,
+      "grad_norm": 0.68775475025177,
+      "learning_rate": 0.00032986100635707354,
+      "loss": 3.3953,
+      "step": 41900
+    },
+    {
+      "epoch": 4.515122161231299,
+      "grad_norm": 0.6535020470619202,
+      "learning_rate": 0.00032953776532701213,
+      "loss": 3.3988,
+      "step": 41950
+    },
+    {
+      "epoch": 4.520503713270907,
+      "grad_norm": 0.720953643321991,
+      "learning_rate": 0.00032921452429695067,
+      "loss": 3.393,
+      "step": 42000
+    },
+    {
+      "epoch": 4.520503713270907,
+      "eval_accuracy": 0.3784747329339181,
+      "eval_loss": 3.4299232959747314,
+      "eval_runtime": 204.2689,
+      "eval_samples_per_second": 88.173,
+      "eval_steps_per_second": 5.512,
+      "step": 42000
+    },
+    {
+      "epoch": 4.5258852653105155,
+      "grad_norm": 0.6605381965637207,
+      "learning_rate": 0.0003288912832668893,
+      "loss": 3.4042,
+      "step": 42050
+    },
+    {
+      "epoch": 4.531266817350124,
+      "grad_norm": 0.6638337969779968,
+      "learning_rate": 0.0003285680422368279,
+      "loss": 3.4088,
+      "step": 42100
+    },
+    {
+      "epoch": 4.536648369389732,
+      "grad_norm": 0.6967762112617493,
+      "learning_rate": 0.0003282448012067665,
+      "loss": 3.3833,
+      "step": 42150
+    },
+    {
+      "epoch": 4.54202992142934,
+      "grad_norm": 0.6514131426811218,
+      "learning_rate": 0.00032792156017670505,
+      "loss": 3.3887,
+      "step": 42200
+    },
+    {
+      "epoch": 4.547411473468949,
+      "grad_norm": 0.6850041747093201,
+      "learning_rate": 0.00032759831914664365,
+      "loss": 3.3844,
+      "step": 42250
+    },
+    {
+      "epoch": 4.5527930255085565,
+      "grad_norm": 0.7080932855606079,
+      "learning_rate": 0.0003272750781165823,
+      "loss": 3.3936,
+      "step": 42300
+    },
+    {
+      "epoch": 4.558174577548165,
+      "grad_norm": 0.6961115598678589,
+      "learning_rate": 0.00032695183708652083,
+      "loss": 3.3951,
+      "step": 42350
+    },
+    {
+      "epoch": 4.563556129587774,
+      "grad_norm": 0.6541742086410522,
+      "learning_rate": 0.00032662859605645943,
+      "loss": 3.3929,
+      "step": 42400
+    },
+    {
+      "epoch": 4.568937681627381,
+      "grad_norm": 0.6856427788734436,
+      "learning_rate": 0.00032630535502639797,
+      "loss": 3.3886,
+      "step": 42450
+    },
+    {
+      "epoch": 4.57431923366699,
+      "grad_norm": 0.684547483921051,
+      "learning_rate": 0.00032598211399633656,
+      "loss": 3.3862,
+      "step": 42500
+    },
+    {
+      "epoch": 4.579700785706597,
+      "grad_norm": 0.6282208561897278,
+      "learning_rate": 0.0003256588729662751,
+      "loss": 3.394,
+      "step": 42550
+    },
+    {
+      "epoch": 4.585082337746206,
+      "grad_norm": 0.6854305863380432,
+      "learning_rate": 0.00032533563193621375,
+      "loss": 3.3815,
+      "step": 42600
+    },
+    {
+      "epoch": 4.5904638897858145,
+      "grad_norm": 0.693110466003418,
+      "learning_rate": 0.00032501239090615235,
+      "loss": 3.3888,
+      "step": 42650
+    },
+    {
+      "epoch": 4.595845441825422,
+      "grad_norm": 0.7092149257659912,
+      "learning_rate": 0.0003246891498760909,
+      "loss": 3.3926,
+      "step": 42700
+    },
+    {
+      "epoch": 4.601226993865031,
+      "grad_norm": 0.6959360241889954,
+      "learning_rate": 0.0003243659088460295,
+      "loss": 3.394,
+      "step": 42750
+    },
+    {
+      "epoch": 4.606608545904638,
+      "grad_norm": 0.6965814828872681,
+      "learning_rate": 0.0003240426678159681,
+      "loss": 3.3794,
+      "step": 42800
+    },
+    {
+      "epoch": 4.611990097944247,
+      "grad_norm": 0.700532078742981,
+      "learning_rate": 0.0003237194267859066,
+      "loss": 3.401,
+      "step": 42850
+    },
+    {
+      "epoch": 4.6173716499838555,
+      "grad_norm": 0.6489181518554688,
+      "learning_rate": 0.00032339618575584527,
+      "loss": 3.4006,
+      "step": 42900
+    },
+    {
+      "epoch": 4.622753202023463,
+      "grad_norm": 0.6220369935035706,
+      "learning_rate": 0.000323079409546385,
+      "loss": 3.3956,
+      "step": 42950
+    },
+    {
+      "epoch": 4.628134754063072,
+      "grad_norm": 0.7331081032752991,
+      "learning_rate": 0.00032275616851632367,
+      "loss": 3.3915,
+      "step": 43000
+    },
+    {
+      "epoch": 4.628134754063072,
+      "eval_accuracy": 0.3789644314992979,
+      "eval_loss": 3.4236953258514404,
+      "eval_runtime": 203.2493,
+      "eval_samples_per_second": 88.615,
+      "eval_steps_per_second": 5.54,
+      "step": 43000
+    },
+    {
+      "epoch": 4.63351630610268,
+      "grad_norm": 0.8542162179946899,
+      "learning_rate": 0.00032243292748626226,
+      "loss": 3.3898,
+      "step": 43050
+    },
+    {
+      "epoch": 4.638897858142288,
+      "grad_norm": 0.6471918821334839,
+      "learning_rate": 0.0003221096864562008,
+      "loss": 3.4054,
+      "step": 43100
+    },
+    {
+      "epoch": 4.6442794101818965,
+      "grad_norm": 0.654157280921936,
+      "learning_rate": 0.0003217864454261394,
+      "loss": 3.396,
+      "step": 43150
+    },
+    {
+      "epoch": 4.649660962221505,
+      "grad_norm": 0.6689890623092651,
+      "learning_rate": 0.00032146320439607794,
+      "loss": 3.3881,
+      "step": 43200
+    },
+    {
+      "epoch": 4.655042514261113,
+      "grad_norm": 0.6910762786865234,
+      "learning_rate": 0.0003211399633660166,
+      "loss": 3.3807,
+      "step": 43250
+    },
+    {
+      "epoch": 4.660424066300721,
+      "grad_norm": 0.6583350300788879,
+      "learning_rate": 0.0003208167223359552,
+      "loss": 3.3996,
+      "step": 43300
+    },
+    {
+      "epoch": 4.665805618340329,
+      "grad_norm": 0.7289586663246155,
+      "learning_rate": 0.0003204934813058937,
+      "loss": 3.398,
+      "step": 43350
+    },
+    {
+      "epoch": 4.6711871703799375,
+      "grad_norm": 0.7323175668716431,
+      "learning_rate": 0.0003201702402758323,
+      "loss": 3.3893,
+      "step": 43400
+    },
+    {
+      "epoch": 4.676568722419546,
+      "grad_norm": 0.7023487687110901,
+      "learning_rate": 0.0003198469992457709,
+      "loss": 3.381,
+      "step": 43450
+    },
+    {
+      "epoch": 4.681950274459154,
+      "grad_norm": 0.6924834847450256,
+      "learning_rate": 0.00031952375821570945,
+      "loss": 3.3818,
+      "step": 43500
+    },
+    {
+      "epoch": 4.687331826498762,
+      "grad_norm": 0.6525893211364746,
+      "learning_rate": 0.0003192005171856481,
+      "loss": 3.3878,
+      "step": 43550
+    },
+    {
+      "epoch": 4.692713378538371,
+      "grad_norm": 0.6708674430847168,
+      "learning_rate": 0.0003188772761555867,
+      "loss": 3.3879,
+      "step": 43600
+    },
+    {
+      "epoch": 4.6980949305779784,
+      "grad_norm": 0.6731662750244141,
+      "learning_rate": 0.00031855403512552524,
+      "loss": 3.3815,
+      "step": 43650
+    },
+    {
+      "epoch": 4.703476482617587,
+      "grad_norm": 0.7286627888679504,
+      "learning_rate": 0.00031823079409546383,
+      "loss": 3.396,
+      "step": 43700
+    },
+    {
+      "epoch": 4.7088580346571955,
+      "grad_norm": 0.6887483596801758,
+      "learning_rate": 0.00031790755306540237,
+      "loss": 3.3951,
+      "step": 43750
+    },
+    {
+      "epoch": 4.714239586696803,
+      "grad_norm": 0.6878728270530701,
+      "learning_rate": 0.00031758431203534097,
+      "loss": 3.382,
+      "step": 43800
+    },
+    {
+      "epoch": 4.719621138736412,
+      "grad_norm": 0.7092850208282471,
+      "learning_rate": 0.0003172610710052796,
+      "loss": 3.3839,
+      "step": 43850
+    },
+    {
+      "epoch": 4.725002690776019,
+      "grad_norm": 0.6616544723510742,
+      "learning_rate": 0.00031693782997521816,
+      "loss": 3.3777,
+      "step": 43900
+    },
+    {
+      "epoch": 4.730384242815628,
+      "grad_norm": 0.6741201281547546,
+      "learning_rate": 0.000316621053765758,
+      "loss": 3.3813,
+      "step": 43950
+    },
+    {
+      "epoch": 4.7357657948552365,
+      "grad_norm": 0.6615583896636963,
+      "learning_rate": 0.00031629781273569656,
+      "loss": 3.3875,
+      "step": 44000
+    },
+    {
+      "epoch": 4.7357657948552365,
+      "eval_accuracy": 0.3794617357668248,
+      "eval_loss": 3.419157028198242,
+      "eval_runtime": 196.8203,
+      "eval_samples_per_second": 91.51,
+      "eval_steps_per_second": 5.721,
+      "step": 44000
+    },
+    {
+      "epoch": 4.741147346894844,
+      "grad_norm": 0.6903731226921082,
+      "learning_rate": 0.00031597457170563515,
+      "loss": 3.391,
+      "step": 44050
+    },
+    {
+      "epoch": 4.746528898934453,
+      "grad_norm": 0.6820841431617737,
+      "learning_rate": 0.0003156513306755737,
+      "loss": 3.3847,
+      "step": 44100
+    },
+    {
+      "epoch": 4.751910450974061,
+      "grad_norm": 0.6453799605369568,
+      "learning_rate": 0.0003153280896455123,
+      "loss": 3.3879,
+      "step": 44150
+    },
+    {
+      "epoch": 4.757292003013669,
+      "grad_norm": 0.7102324962615967,
+      "learning_rate": 0.00031500484861545094,
+      "loss": 3.4133,
+      "step": 44200
+    },
+    {
+      "epoch": 4.7626735550532775,
+      "grad_norm": 0.6926384568214417,
+      "learning_rate": 0.0003146816075853895,
+      "loss": 3.3997,
+      "step": 44250
+    },
+    {
+      "epoch": 4.768055107092886,
+      "grad_norm": 0.6980499029159546,
+      "learning_rate": 0.00031435836655532807,
+      "loss": 3.3938,
+      "step": 44300
+    },
+    {
+      "epoch": 4.773436659132494,
+      "grad_norm": 0.6628013253211975,
+      "learning_rate": 0.00031403512552526667,
+      "loss": 3.3952,
+      "step": 44350
+    },
+    {
+      "epoch": 4.778818211172102,
+      "grad_norm": 0.7179418206214905,
+      "learning_rate": 0.0003137118844952052,
+      "loss": 3.3886,
+      "step": 44400
+    },
+    {
+      "epoch": 4.78419976321171,
+      "grad_norm": 0.6799003481864929,
+      "learning_rate": 0.0003133886434651438,
+      "loss": 3.385,
+      "step": 44450
+    },
+    {
+      "epoch": 4.7895813152513185,
+      "grad_norm": 0.6856629848480225,
+      "learning_rate": 0.00031306540243508245,
+      "loss": 3.3767,
+      "step": 44500
+    },
+    {
+      "epoch": 4.794962867290927,
+      "grad_norm": 0.689735472202301,
+      "learning_rate": 0.000312742161405021,
+      "loss": 3.4087,
+      "step": 44550
+    },
+    {
+      "epoch": 4.800344419330535,
+      "grad_norm": 0.685651957988739,
+      "learning_rate": 0.0003124189203749596,
+      "loss": 3.3955,
+      "step": 44600
+    },
+    {
+      "epoch": 4.805725971370143,
+      "grad_norm": 0.6903262138366699,
+      "learning_rate": 0.0003120956793448981,
+      "loss": 3.3996,
+      "step": 44650
+    },
+    {
+      "epoch": 4.811107523409751,
+      "grad_norm": 0.7058689594268799,
+      "learning_rate": 0.0003117724383148367,
+      "loss": 3.4022,
+      "step": 44700
+    },
+    {
+      "epoch": 4.8164890754493594,
+      "grad_norm": 0.6723160147666931,
+      "learning_rate": 0.00031144919728477526,
+      "loss": 3.3906,
+      "step": 44750
+    },
+    {
+      "epoch": 4.821870627488968,
+      "grad_norm": 0.6740667223930359,
+      "learning_rate": 0.0003111259562547139,
+      "loss": 3.3985,
+      "step": 44800
+    },
+    {
+      "epoch": 4.827252179528576,
+      "grad_norm": 0.6879459023475647,
+      "learning_rate": 0.0003108027152246525,
+      "loss": 3.379,
+      "step": 44850
+    },
+    {
+      "epoch": 4.832633731568184,
+      "grad_norm": 0.672219455242157,
+      "learning_rate": 0.0003104794741945911,
+      "loss": 3.3935,
+      "step": 44900
+    },
+    {
+      "epoch": 4.838015283607793,
+      "grad_norm": 0.680568277835846,
+      "learning_rate": 0.00031015623316452964,
+      "loss": 3.3993,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8433968356474,
+      "grad_norm": 0.6724383234977722,
+      "learning_rate": 0.00030983299213446823,
+      "loss": 3.4045,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8433968356474,
+      "eval_accuracy": 0.3801395124810116,
+      "eval_loss": 3.414053440093994,
+      "eval_runtime": 206.488,
+      "eval_samples_per_second": 87.225,
+      "eval_steps_per_second": 5.453,
+      "step": 45000
+    },
+    {
+      "epoch": 4.848778387687009,
+      "grad_norm": 0.7093207240104675,
+      "learning_rate": 0.0003095097511044069,
+      "loss": 3.3656,
+      "step": 45050
+    },
+    {
+      "epoch": 4.8541599397266175,
+      "grad_norm": 0.6884026527404785,
+      "learning_rate": 0.0003091865100743454,
+      "loss": 3.3815,
+      "step": 45100
+    },
+    {
+      "epoch": 4.859541491766225,
+      "grad_norm": 0.6394287943840027,
+      "learning_rate": 0.000308863269044284,
+      "loss": 3.4021,
+      "step": 45150
+    },
+    {
+      "epoch": 4.864923043805834,
+      "grad_norm": 0.656760573387146,
+      "learning_rate": 0.00030854002801422256,
+      "loss": 3.3942,
+      "step": 45200
+    },
+    {
+      "epoch": 4.870304595845441,
+      "grad_norm": 0.7228603363037109,
+      "learning_rate": 0.00030821678698416115,
+      "loss": 3.3831,
+      "step": 45250
+    },
+    {
+      "epoch": 4.87568614788505,
+      "grad_norm": 0.6911720633506775,
+      "learning_rate": 0.0003078935459540997,
+      "loss": 3.3941,
+      "step": 45300
+    },
+    {
+      "epoch": 4.8810676999246585,
+      "grad_norm": 0.6649526357650757,
+      "learning_rate": 0.00030757030492403834,
+      "loss": 3.3868,
+      "step": 45350
+    },
+    {
+      "epoch": 4.886449251964266,
+      "grad_norm": 0.6843016743659973,
+      "learning_rate": 0.00030724706389397694,
+      "loss": 3.3774,
+      "step": 45400
+    },
+    {
+      "epoch": 4.891830804003875,
+      "grad_norm": 0.6798348426818848,
+      "learning_rate": 0.0003069238228639155,
+      "loss": 3.3856,
+      "step": 45450
+    },
+    {
+      "epoch": 4.897212356043483,
+      "grad_norm": 0.6509679555892944,
+      "learning_rate": 0.00030660058183385407,
+      "loss": 3.3992,
+      "step": 45500
+    },
+    {
+      "epoch": 4.902593908083091,
+      "grad_norm": 0.6958284378051758,
+      "learning_rate": 0.00030627734080379267,
+      "loss": 3.3722,
+      "step": 45550
+    },
+    {
+      "epoch": 4.9079754601226995,
+      "grad_norm": 0.6654974222183228,
+      "learning_rate": 0.0003059540997737312,
+      "loss": 3.3861,
+      "step": 45600
+    },
+    {
+      "epoch": 4.913357012162308,
+      "grad_norm": 0.7664541602134705,
+      "learning_rate": 0.00030563085874366986,
+      "loss": 3.4079,
+      "step": 45650
+    },
+    {
+      "epoch": 4.918738564201916,
+      "grad_norm": 0.7275959849357605,
+      "learning_rate": 0.00030530761771360845,
+      "loss": 3.405,
+      "step": 45700
+    },
+    {
+      "epoch": 4.924120116241524,
+      "grad_norm": 0.6668584942817688,
+      "learning_rate": 0.000304984376683547,
+      "loss": 3.4006,
+      "step": 45750
+    },
+    {
+      "epoch": 4.929501668281132,
+      "grad_norm": 0.6965872645378113,
+      "learning_rate": 0.0003046611356534856,
+      "loss": 3.3963,
+      "step": 45800
+    },
+    {
+      "epoch": 4.9348832203207404,
+      "grad_norm": 0.7191141247749329,
+      "learning_rate": 0.0003043378946234241,
+      "loss": 3.3976,
+      "step": 45850
+    },
+    {
+      "epoch": 4.940264772360349,
+      "grad_norm": 0.6690873503684998,
+      "learning_rate": 0.0003040146535933628,
+      "loss": 3.4008,
+      "step": 45900
+    },
+    {
+      "epoch": 4.945646324399957,
+      "grad_norm": 0.6937520503997803,
+      "learning_rate": 0.00030369141256330137,
+      "loss": 3.3907,
+      "step": 45950
+    },
+    {
+      "epoch": 4.951027876439565,
+      "grad_norm": 0.6963992118835449,
+      "learning_rate": 0.0003033681715332399,
+      "loss": 3.3761,
+      "step": 46000
+    },
+    {
+      "epoch": 4.951027876439565,
+      "eval_accuracy": 0.3804870930691301,
+      "eval_loss": 3.4095349311828613,
+      "eval_runtime": 205.4933,
+      "eval_samples_per_second": 87.648,
+      "eval_steps_per_second": 5.479,
+      "step": 46000
+    },
+    {
+      "epoch": 4.956409428479174,
+      "grad_norm": 0.7414904832839966,
+      "learning_rate": 0.0003030449305031785,
+      "loss": 3.3747,
+      "step": 46050
+    },
+    {
+      "epoch": 4.961790980518781,
+      "grad_norm": 0.7259211540222168,
+      "learning_rate": 0.0003027216894731171,
+      "loss": 3.3819,
+      "step": 46100
+    },
+    {
+      "epoch": 4.96717253255839,
+      "grad_norm": 0.7090394496917725,
+      "learning_rate": 0.00030239844844305564,
+      "loss": 3.3889,
+      "step": 46150
+    },
+    {
+      "epoch": 4.9725540845979985,
+      "grad_norm": 0.7188470363616943,
+      "learning_rate": 0.0003020752074129943,
+      "loss": 3.3881,
+      "step": 46200
+    },
+    {
+      "epoch": 4.977935636637606,
+      "grad_norm": 0.6995404362678528,
+      "learning_rate": 0.0003017519663829329,
+      "loss": 3.3543,
+      "step": 46250
+    },
+    {
+      "epoch": 4.983317188677215,
+      "grad_norm": 0.6940346956253052,
+      "learning_rate": 0.0003014287253528714,
+      "loss": 3.378,
+      "step": 46300
+    },
+    {
+      "epoch": 4.988698740716822,
+      "grad_norm": 0.717013955116272,
+      "learning_rate": 0.00030110548432281,
+      "loss": 3.3916,
+      "step": 46350
+    },
+    {
+      "epoch": 4.994080292756431,
+      "grad_norm": 0.7562733888626099,
+      "learning_rate": 0.00030078224329274856,
+      "loss": 3.3738,
+      "step": 46400
+    },
+    {
+      "epoch": 4.9994618447960395,
+      "grad_norm": 0.6929606199264526,
+      "learning_rate": 0.00030045900226268715,
+      "loss": 3.3991,
+      "step": 46450
+    },
+    {
+      "epoch": 5.004843396835647,
+      "grad_norm": 0.684966504573822,
+      "learning_rate": 0.0003001357612326258,
+      "loss": 3.3138,
+      "step": 46500
+    },
+    {
+      "epoch": 5.010224948875256,
+      "grad_norm": 0.6948528289794922,
+      "learning_rate": 0.00029981252020256434,
+      "loss": 3.2973,
+      "step": 46550
+    },
+    {
+      "epoch": 5.015606500914864,
+      "grad_norm": 0.7069754004478455,
+      "learning_rate": 0.00029948927917250294,
+      "loss": 3.2938,
+      "step": 46600
+    },
+    {
+      "epoch": 5.020988052954472,
+      "grad_norm": 0.6681404113769531,
+      "learning_rate": 0.00029916603814244153,
+      "loss": 3.3031,
+      "step": 46650
+    },
+    {
+      "epoch": 5.0263696049940805,
+      "grad_norm": 0.7222424745559692,
+      "learning_rate": 0.0002988427971123801,
+      "loss": 3.2958,
+      "step": 46700
+    },
+    {
+      "epoch": 5.031751157033688,
+      "grad_norm": 0.6528456211090088,
+      "learning_rate": 0.00029851955608231867,
+      "loss": 3.2972,
+      "step": 46750
+    },
+    {
+      "epoch": 5.037132709073297,
+      "grad_norm": 0.7115670442581177,
+      "learning_rate": 0.00029819631505225726,
+      "loss": 3.3177,
+      "step": 46800
+    },
+    {
+      "epoch": 5.042514261112905,
+      "grad_norm": 0.7243658304214478,
+      "learning_rate": 0.00029787307402219586,
+      "loss": 3.3053,
+      "step": 46850
+    },
+    {
+      "epoch": 5.047895813152513,
+      "grad_norm": 0.6976185441017151,
+      "learning_rate": 0.00029754983299213445,
+      "loss": 3.3041,
+      "step": 46900
+    },
+    {
+      "epoch": 5.0532773651921215,
+      "grad_norm": 0.6580340266227722,
+      "learning_rate": 0.000297226591962073,
+      "loss": 3.2969,
+      "step": 46950
+    },
+    {
+      "epoch": 5.05865891723173,
+      "grad_norm": 0.6749905943870544,
+      "learning_rate": 0.00029690335093201164,
+      "loss": 3.3004,
+      "step": 47000
+    },
+    {
+      "epoch": 5.05865891723173,
+      "eval_accuracy": 0.380867812788032,
+      "eval_loss": 3.412269353866577,
+      "eval_runtime": 196.4805,
+      "eval_samples_per_second": 91.668,
+      "eval_steps_per_second": 5.731,
+      "step": 47000
+    },
+    {
+      "epoch": 5.064040469271338,
+      "grad_norm": 0.7184104919433594,
+      "learning_rate": 0.0002965801099019502,
+      "loss": 3.3063,
+      "step": 47050
+    },
+    {
+      "epoch": 5.069422021310946,
+      "grad_norm": 0.6803567409515381,
+      "learning_rate": 0.0002962568688718888,
+      "loss": 3.2995,
+      "step": 47100
+    },
+    {
+      "epoch": 5.074803573350554,
+      "grad_norm": 0.6682900190353394,
+      "learning_rate": 0.00029593362784182737,
+      "loss": 3.291,
+      "step": 47150
+    },
+    {
+      "epoch": 5.080185125390162,
+      "grad_norm": 0.6613481640815735,
+      "learning_rate": 0.00029561038681176596,
+      "loss": 3.2996,
+      "step": 47200
+    },
+    {
+      "epoch": 5.085566677429771,
+      "grad_norm": 0.6656726598739624,
+      "learning_rate": 0.00029528714578170456,
+      "loss": 3.3157,
+      "step": 47250
+    },
+    {
+      "epoch": 5.090948229469379,
+      "grad_norm": 0.6987922191619873,
+      "learning_rate": 0.0002949639047516431,
+      "loss": 3.3214,
+      "step": 47300
+    },
+    {
+      "epoch": 5.096329781508987,
+      "grad_norm": 0.6765496134757996,
+      "learning_rate": 0.0002946406637215817,
+      "loss": 3.3238,
+      "step": 47350
+    },
+    {
+      "epoch": 5.101711333548596,
+      "grad_norm": 0.722225546836853,
+      "learning_rate": 0.0002943174226915203,
+      "loss": 3.3189,
+      "step": 47400
+    },
+    {
+      "epoch": 5.107092885588203,
+      "grad_norm": 0.6817000508308411,
+      "learning_rate": 0.0002939941816614589,
+      "loss": 3.3209,
+      "step": 47450
+    },
+    {
+      "epoch": 5.112474437627812,
+      "grad_norm": 0.7062390446662903,
+      "learning_rate": 0.0002936709406313974,
+      "loss": 3.315,
+      "step": 47500
+    },
+    {
+      "epoch": 5.1178559896674205,
+      "grad_norm": 0.6629499197006226,
+      "learning_rate": 0.0002933476996013361,
+      "loss": 3.319,
+      "step": 47550
+    },
+    {
+      "epoch": 5.123237541707028,
+      "grad_norm": 0.6865987777709961,
+      "learning_rate": 0.0002930244585712746,
+      "loss": 3.3216,
+      "step": 47600
+    },
+    {
+      "epoch": 5.128619093746637,
+      "grad_norm": 0.7235592007637024,
+      "learning_rate": 0.0002927012175412132,
+      "loss": 3.3115,
+      "step": 47650
+    },
+    {
+      "epoch": 5.134000645786244,
+      "grad_norm": 0.7191505432128906,
+      "learning_rate": 0.0002923779765111518,
+      "loss": 3.3012,
+      "step": 47700
+    },
+    {
+      "epoch": 5.139382197825853,
+      "grad_norm": 0.706909716129303,
+      "learning_rate": 0.0002920547354810904,
+      "loss": 3.3227,
+      "step": 47750
+    },
+    {
+      "epoch": 5.1447637498654615,
+      "grad_norm": 0.7347841262817383,
+      "learning_rate": 0.00029173149445102894,
+      "loss": 3.3219,
+      "step": 47800
+    },
+    {
+      "epoch": 5.150145301905069,
+      "grad_norm": 0.7402267456054688,
+      "learning_rate": 0.00029140825342096753,
+      "loss": 3.3008,
+      "step": 47850
+    },
+    {
+      "epoch": 5.155526853944678,
+      "grad_norm": 0.678464949131012,
+      "learning_rate": 0.0002910850123909061,
+      "loss": 3.3218,
+      "step": 47900
+    },
+    {
+      "epoch": 5.160908405984286,
+      "grad_norm": 0.7275342345237732,
+      "learning_rate": 0.0002907617713608447,
+      "loss": 3.3034,
+      "step": 47950
+    },
+    {
+      "epoch": 5.166289958023894,
+      "grad_norm": 0.706177294254303,
+      "learning_rate": 0.0002904385303307833,
+      "loss": 3.3289,
+      "step": 48000
+    },
+    {
+      "epoch": 5.166289958023894,
+      "eval_accuracy": 0.3814682286460948,
+      "eval_loss": 3.4115307331085205,
+      "eval_runtime": 209.8199,
+      "eval_samples_per_second": 85.84,
+      "eval_steps_per_second": 5.367,
+      "step": 48000
+    },
+    {
+      "epoch": 5.1716715100635025,
+      "grad_norm": 0.6725974678993225,
+      "learning_rate": 0.00029011528930072186,
+      "loss": 3.3346,
+      "step": 48050
+    },
+    {
+      "epoch": 5.17705306210311,
+      "grad_norm": 0.726219117641449,
+      "learning_rate": 0.00028979204827066045,
+      "loss": 3.3088,
+      "step": 48100
+    },
+    {
+      "epoch": 5.182434614142719,
+      "grad_norm": 0.6821277141571045,
+      "learning_rate": 0.00028946880724059905,
+      "loss": 3.3257,
+      "step": 48150
+    },
+    {
+      "epoch": 5.187816166182327,
+      "grad_norm": 0.695120096206665,
+      "learning_rate": 0.00028914556621053764,
+      "loss": 3.3328,
+      "step": 48200
+    },
+    {
+      "epoch": 5.193197718221935,
+      "grad_norm": 0.694706380367279,
+      "learning_rate": 0.00028882232518047624,
+      "loss": 3.3343,
+      "step": 48250
+    },
+    {
+      "epoch": 5.198579270261543,
+      "grad_norm": 0.7311663031578064,
+      "learning_rate": 0.00028849908415041483,
+      "loss": 3.3231,
+      "step": 48300
+    },
+    {
+      "epoch": 5.203960822301152,
+      "grad_norm": 0.7439178228378296,
+      "learning_rate": 0.00028817584312035337,
+      "loss": 3.314,
+      "step": 48350
+    },
+    {
+      "epoch": 5.20934237434076,
+      "grad_norm": 0.7027797698974609,
+      "learning_rate": 0.00028785260209029197,
+      "loss": 3.3172,
+      "step": 48400
+    },
+    {
+      "epoch": 5.214723926380368,
+      "grad_norm": 0.7429949045181274,
+      "learning_rate": 0.00028752936106023056,
+      "loss": 3.3185,
+      "step": 48450
+    },
+    {
+      "epoch": 5.220105478419977,
+      "grad_norm": 0.6877169013023376,
+      "learning_rate": 0.00028720612003016915,
+      "loss": 3.3372,
+      "step": 48500
+    },
+    {
+      "epoch": 5.225487030459584,
+      "grad_norm": 0.6922637224197388,
+      "learning_rate": 0.00028688287900010775,
+      "loss": 3.3232,
+      "step": 48550
+    },
+    {
+      "epoch": 5.230868582499193,
+      "grad_norm": 0.7518928647041321,
+      "learning_rate": 0.0002865596379700463,
+      "loss": 3.3261,
+      "step": 48600
+    },
+    {
+      "epoch": 5.236250134538801,
+      "grad_norm": 0.7136549353599548,
+      "learning_rate": 0.0002862363969399849,
+      "loss": 3.3301,
+      "step": 48650
+    },
+    {
+      "epoch": 5.241631686578409,
+      "grad_norm": 0.7302454710006714,
+      "learning_rate": 0.0002859131559099235,
+      "loss": 3.3319,
+      "step": 48700
+    },
+    {
+      "epoch": 5.247013238618018,
+      "grad_norm": 0.6988076567649841,
+      "learning_rate": 0.0002855899148798621,
+      "loss": 3.3271,
+      "step": 48750
+    },
+    {
+      "epoch": 5.252394790657625,
+      "grad_norm": 0.6556980609893799,
+      "learning_rate": 0.0002852666738498006,
+      "loss": 3.3376,
+      "step": 48800
+    },
+    {
+      "epoch": 5.257776342697234,
+      "grad_norm": 0.7035303115844727,
+      "learning_rate": 0.00028494343281973926,
+      "loss": 3.3302,
+      "step": 48850
+    },
+    {
+      "epoch": 5.2631578947368425,
+      "grad_norm": 0.7211655974388123,
+      "learning_rate": 0.0002846201917896778,
+      "loss": 3.3299,
+      "step": 48900
+    },
+    {
+      "epoch": 5.26853944677645,
+      "grad_norm": 0.7089269161224365,
+      "learning_rate": 0.0002842969507596164,
+      "loss": 3.3217,
+      "step": 48950
+    },
+    {
+      "epoch": 5.273920998816059,
+      "grad_norm": 0.6631143689155579,
+      "learning_rate": 0.000283973709729555,
+      "loss": 3.3151,
+      "step": 49000
+    },
+    {
+      "epoch": 5.273920998816059,
+      "eval_accuracy": 0.3817304080643912,
+      "eval_loss": 3.4057440757751465,
+      "eval_runtime": 211.9424,
+      "eval_samples_per_second": 84.981,
+      "eval_steps_per_second": 5.313,
+      "step": 49000
+    },
+    {
+      "epoch": 5.279302550855666,
+      "grad_norm": 0.682040810585022,
+      "learning_rate": 0.0002836504686994936,
+      "loss": 3.3137,
+      "step": 49050
+    },
+    {
+      "epoch": 5.284684102895275,
+      "grad_norm": 0.6834739446640015,
+      "learning_rate": 0.00028332722766943213,
+      "loss": 3.3403,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2900656549348835,
+      "grad_norm": 0.7186492681503296,
+      "learning_rate": 0.0002830039866393707,
+      "loss": 3.3335,
+      "step": 49150
+    },
+    {
+      "epoch": 5.295447206974491,
+      "grad_norm": 0.6996015310287476,
+      "learning_rate": 0.0002826807456093093,
+      "loss": 3.3068,
+      "step": 49200
+    },
+    {
+      "epoch": 5.3008287590141,
+      "grad_norm": 0.7867818474769592,
+      "learning_rate": 0.0002823575045792479,
+      "loss": 3.3175,
+      "step": 49250
+    },
+    {
+      "epoch": 5.306210311053708,
+      "grad_norm": 0.6740609407424927,
+      "learning_rate": 0.0002820342635491865,
+      "loss": 3.307,
+      "step": 49300
+    },
+    {
+      "epoch": 5.311591863093316,
+      "grad_norm": 0.6935590505599976,
+      "learning_rate": 0.00028171102251912505,
+      "loss": 3.3232,
+      "step": 49350
+    },
+    {
+      "epoch": 5.316973415132924,
+      "grad_norm": 0.690018892288208,
+      "learning_rate": 0.00028138778148906364,
+      "loss": 3.3367,
+      "step": 49400
+    },
+    {
+      "epoch": 5.322354967172533,
+      "grad_norm": 0.7360737919807434,
+      "learning_rate": 0.00028106454045900224,
+      "loss": 3.3176,
+      "step": 49450
+    },
+    {
+      "epoch": 5.327736519212141,
+      "grad_norm": 0.6747078895568848,
+      "learning_rate": 0.00028074129942894083,
+      "loss": 3.3376,
+      "step": 49500
+    },
+    {
+      "epoch": 5.333118071251749,
+      "grad_norm": 0.7133691310882568,
+      "learning_rate": 0.0002804180583988794,
+      "loss": 3.3254,
+      "step": 49550
+    },
+    {
+      "epoch": 5.338499623291357,
+      "grad_norm": 0.8014851212501526,
+      "learning_rate": 0.00028009481736881797,
+      "loss": 3.321,
+      "step": 49600
+    },
+    {
+      "epoch": 5.343881175330965,
+      "grad_norm": 0.7052977085113525,
+      "learning_rate": 0.00027977157633875656,
+      "loss": 3.3394,
+      "step": 49650
+    },
+    {
+      "epoch": 5.349262727370574,
+      "grad_norm": 0.7522420883178711,
+      "learning_rate": 0.00027944833530869516,
+      "loss": 3.3236,
+      "step": 49700
+    },
+    {
+      "epoch": 5.354644279410182,
+      "grad_norm": 0.7162737846374512,
+      "learning_rate": 0.00027913155909923496,
+      "loss": 3.3151,
+      "step": 49750
+    },
+    {
+      "epoch": 5.36002583144979,
+      "grad_norm": 0.7829045653343201,
+      "learning_rate": 0.00027880831806917356,
+      "loss": 3.3139,
+      "step": 49800
+    },
+    {
+      "epoch": 5.365407383489399,
+      "grad_norm": 0.6889187693595886,
+      "learning_rate": 0.00027848507703911215,
+      "loss": 3.3473,
+      "step": 49850
+    },
+    {
+      "epoch": 5.370788935529006,
+      "grad_norm": 0.7904999256134033,
+      "learning_rate": 0.0002781618360090507,
+      "loss": 3.3207,
+      "step": 49900
+    },
+    {
+      "epoch": 5.376170487568615,
+      "grad_norm": 0.6826810836791992,
+      "learning_rate": 0.00027783859497898934,
+      "loss": 3.338,
+      "step": 49950
+    },
+    {
+      "epoch": 5.3815520396082235,
+      "grad_norm": 0.7005066871643066,
+      "learning_rate": 0.0002775153539489279,
+      "loss": 3.3528,
+      "step": 50000
+    },
+    {
+      "epoch": 5.3815520396082235,
+      "eval_accuracy": 0.3818176563333062,
+      "eval_loss": 3.402635097503662,
+      "eval_runtime": 195.8066,
+      "eval_samples_per_second": 91.984,
+      "eval_steps_per_second": 5.751,
+      "step": 50000
+    },
+    {
+      "epoch": 5.386933591647831,
+      "grad_norm": 0.7351107001304626,
+      "learning_rate": 0.0002771921129188665,
+      "loss": 3.3345,
+      "step": 50050
+    },
+    {
+      "epoch": 5.39231514368744,
+      "grad_norm": 0.6939401626586914,
+      "learning_rate": 0.00027686887188880507,
+      "loss": 3.3426,
+      "step": 50100
+    },
+    {
+      "epoch": 5.397696695727047,
+      "grad_norm": 0.7426904439926147,
+      "learning_rate": 0.00027654563085874366,
+      "loss": 3.3367,
+      "step": 50150
+    },
+    {
+      "epoch": 5.403078247766656,
+      "grad_norm": 0.7072582840919495,
+      "learning_rate": 0.00027622238982868226,
+      "loss": 3.3355,
+      "step": 50200
+    },
+    {
+      "epoch": 5.4084597998062645,
+      "grad_norm": 0.829706609249115,
+      "learning_rate": 0.0002758991487986208,
+      "loss": 3.3383,
+      "step": 50250
+    },
+    {
+      "epoch": 5.413841351845872,
+      "grad_norm": 0.8014844655990601,
+      "learning_rate": 0.0002755759077685594,
+      "loss": 3.3264,
+      "step": 50300
+    },
+    {
+      "epoch": 5.419222903885481,
+      "grad_norm": 0.7258954644203186,
+      "learning_rate": 0.000275252666738498,
+      "loss": 3.3297,
+      "step": 50350
+    },
+    {
+      "epoch": 5.424604455925088,
+      "grad_norm": 0.706342339515686,
+      "learning_rate": 0.0002749294257084366,
+      "loss": 3.332,
+      "step": 50400
+    },
+    {
+      "epoch": 5.429986007964697,
+      "grad_norm": 0.7490754723548889,
+      "learning_rate": 0.0002746061846783751,
+      "loss": 3.3274,
+      "step": 50450
+    },
+    {
+      "epoch": 5.435367560004305,
+      "grad_norm": 0.6892290115356445,
+      "learning_rate": 0.0002742829436483138,
+      "loss": 3.3248,
+      "step": 50500
+    },
+    {
+      "epoch": 5.440749112043913,
+      "grad_norm": 0.7274524569511414,
+      "learning_rate": 0.0002739597026182523,
+      "loss": 3.3216,
+      "step": 50550
+    },
+    {
+      "epoch": 5.446130664083522,
+      "grad_norm": 0.7700362801551819,
+      "learning_rate": 0.0002736364615881909,
+      "loss": 3.3392,
+      "step": 50600
+    },
+    {
+      "epoch": 5.45151221612313,
+      "grad_norm": 0.6922512650489807,
+      "learning_rate": 0.0002733132205581295,
+      "loss": 3.3374,
+      "step": 50650
+    },
+    {
+      "epoch": 5.456893768162738,
+      "grad_norm": 0.7054064273834229,
+      "learning_rate": 0.00027298997952806804,
+      "loss": 3.3416,
+      "step": 50700
+    },
+    {
+      "epoch": 5.462275320202346,
+      "grad_norm": 0.7377570867538452,
+      "learning_rate": 0.00027266673849800664,
+      "loss": 3.3335,
+      "step": 50750
+    },
+    {
+      "epoch": 5.467656872241955,
+      "grad_norm": 0.7138562202453613,
+      "learning_rate": 0.00027234349746794523,
+      "loss": 3.3428,
+      "step": 50800
+    },
+    {
+      "epoch": 5.473038424281563,
+      "grad_norm": 0.7018982172012329,
+      "learning_rate": 0.00027202025643788383,
+      "loss": 3.3298,
+      "step": 50850
+    },
+    {
+      "epoch": 5.478419976321171,
+      "grad_norm": 0.7368566393852234,
+      "learning_rate": 0.00027169701540782237,
+      "loss": 3.3449,
+      "step": 50900
+    },
+    {
+      "epoch": 5.483801528360779,
+      "grad_norm": 0.6727787852287292,
+      "learning_rate": 0.000271373774377761,
+      "loss": 3.3402,
+      "step": 50950
+    },
+    {
+      "epoch": 5.489183080400387,
+      "grad_norm": 0.7162392139434814,
+      "learning_rate": 0.00027105053334769956,
+      "loss": 3.3271,
+      "step": 51000
+    },
+    {
+      "epoch": 5.489183080400387,
+      "eval_accuracy": 0.38230561645248107,
+      "eval_loss": 3.398566722869873,
+      "eval_runtime": 200.4209,
+      "eval_samples_per_second": 89.866,
+      "eval_steps_per_second": 5.618,
+      "step": 51000
+    },
+    {
+      "epoch": 5.494564632439996,
+      "grad_norm": 0.7033275365829468,
+      "learning_rate": 0.00027072729231763815,
+      "loss": 3.3318,
+      "step": 51050
+    },
+    {
+      "epoch": 5.499946184479604,
+      "grad_norm": 0.6969659328460693,
+      "learning_rate": 0.00027040405128757675,
+      "loss": 3.34,
+      "step": 51100
+    },
+    {
+      "epoch": 5.505327736519212,
+      "grad_norm": 0.827999472618103,
+      "learning_rate": 0.00027008081025751534,
+      "loss": 3.3444,
+      "step": 51150
+    },
+    {
+      "epoch": 5.510709288558821,
+      "grad_norm": 0.7825914025306702,
+      "learning_rate": 0.00026975756922745394,
+      "loss": 3.3355,
+      "step": 51200
+    },
+    {
+      "epoch": 5.516090840598428,
+      "grad_norm": 0.7196942567825317,
+      "learning_rate": 0.0002694343281973925,
+      "loss": 3.3247,
+      "step": 51250
+    },
+    {
+      "epoch": 5.521472392638037,
+      "grad_norm": 0.6952111721038818,
+      "learning_rate": 0.00026911108716733107,
+      "loss": 3.3484,
+      "step": 51300
+    },
+    {
+      "epoch": 5.5268539446776455,
+      "grad_norm": 0.7286564111709595,
+      "learning_rate": 0.00026878784613726967,
+      "loss": 3.3394,
+      "step": 51350
+    },
+    {
+      "epoch": 5.532235496717253,
+      "grad_norm": 0.7400366067886353,
+      "learning_rate": 0.00026846460510720826,
+      "loss": 3.3425,
+      "step": 51400
+    },
+    {
+      "epoch": 5.537617048756862,
+      "grad_norm": 0.6877460479736328,
+      "learning_rate": 0.0002681413640771468,
+      "loss": 3.3296,
+      "step": 51450
+    },
+    {
+      "epoch": 5.542998600796469,
+      "grad_norm": 0.7166876196861267,
+      "learning_rate": 0.00026781812304708545,
+      "loss": 3.3404,
+      "step": 51500
+    },
+    {
+      "epoch": 5.548380152836078,
+      "grad_norm": 0.7324938178062439,
+      "learning_rate": 0.000267494882017024,
+      "loss": 3.3422,
+      "step": 51550
+    },
+    {
+      "epoch": 5.553761704875686,
+      "grad_norm": 0.7295821309089661,
+      "learning_rate": 0.0002671716409869626,
+      "loss": 3.3304,
+      "step": 51600
+    },
+    {
+      "epoch": 5.559143256915294,
+      "grad_norm": 0.7150875926017761,
+      "learning_rate": 0.0002668483999569012,
+      "loss": 3.3561,
+      "step": 51650
+    },
+    {
+      "epoch": 5.564524808954903,
+      "grad_norm": 0.7677608728408813,
+      "learning_rate": 0.0002665251589268398,
+      "loss": 3.3307,
+      "step": 51700
+    },
+    {
+      "epoch": 5.569906360994511,
+      "grad_norm": 0.6857333183288574,
+      "learning_rate": 0.0002662019178967783,
+      "loss": 3.3412,
+      "step": 51750
+    },
+    {
+      "epoch": 5.575287913034119,
+      "grad_norm": 0.6856406927108765,
+      "learning_rate": 0.0002658786768667169,
+      "loss": 3.3386,
+      "step": 51800
+    },
+    {
+      "epoch": 5.580669465073727,
+      "grad_norm": 0.7286893129348755,
+      "learning_rate": 0.0002655554358366555,
+      "loss": 3.3352,
+      "step": 51850
+    },
+    {
+      "epoch": 5.586051017113336,
+      "grad_norm": 0.7269234657287598,
+      "learning_rate": 0.0002652321948065941,
+      "loss": 3.3422,
+      "step": 51900
+    },
+    {
+      "epoch": 5.591432569152944,
+      "grad_norm": 0.7077545523643494,
+      "learning_rate": 0.0002649089537765327,
+      "loss": 3.3353,
+      "step": 51950
+    },
+    {
+      "epoch": 5.596814121192552,
+      "grad_norm": 0.7062633633613586,
+      "learning_rate": 0.00026458571274647123,
+      "loss": 3.3457,
+      "step": 52000
+    },
+    {
+      "epoch": 5.596814121192552,
+      "eval_accuracy": 0.38252248761655877,
+      "eval_loss": 3.3943653106689453,
+      "eval_runtime": 202.3361,
+      "eval_samples_per_second": 89.015,
+      "eval_steps_per_second": 5.565,
+      "step": 52000
+    },
+    {
+      "epoch": 5.60219567323216,
+      "grad_norm": 0.7586387991905212,
+      "learning_rate": 0.0002642624717164099,
+      "loss": 3.3286,
+      "step": 52050
+    },
+    {
+      "epoch": 5.607577225271768,
+      "grad_norm": 0.6984559297561646,
+      "learning_rate": 0.0002639392306863484,
+      "loss": 3.3364,
+      "step": 52100
+    },
+    {
+      "epoch": 5.612958777311377,
+      "grad_norm": 0.7536754608154297,
+      "learning_rate": 0.000263615989656287,
+      "loss": 3.3353,
+      "step": 52150
+    },
+    {
+      "epoch": 5.618340329350985,
+      "grad_norm": 0.7495378851890564,
+      "learning_rate": 0.0002632927486262256,
+      "loss": 3.3361,
+      "step": 52200
+    },
+    {
+      "epoch": 5.623721881390593,
+      "grad_norm": 0.684282124042511,
+      "learning_rate": 0.0002629695075961642,
+      "loss": 3.3353,
+      "step": 52250
+    },
+    {
+      "epoch": 5.629103433430201,
+      "grad_norm": 0.7329372763633728,
+      "learning_rate": 0.00026264626656610275,
+      "loss": 3.3291,
+      "step": 52300
+    },
+    {
+      "epoch": 5.634484985469809,
+      "grad_norm": 0.8108019232749939,
+      "learning_rate": 0.00026232302553604134,
+      "loss": 3.3548,
+      "step": 52350
+    },
+    {
+      "epoch": 5.639866537509418,
+      "grad_norm": 0.7560662627220154,
+      "learning_rate": 0.00026199978450597994,
+      "loss": 3.325,
+      "step": 52400
+    },
+    {
+      "epoch": 5.645248089549026,
+      "grad_norm": 0.7306268215179443,
+      "learning_rate": 0.00026167654347591853,
+      "loss": 3.334,
+      "step": 52450
+    },
+    {
+      "epoch": 5.650629641588634,
+      "grad_norm": 0.7203580737113953,
+      "learning_rate": 0.0002613533024458571,
+      "loss": 3.3221,
+      "step": 52500
+    },
+    {
+      "epoch": 5.656011193628243,
+      "grad_norm": 0.6746958494186401,
+      "learning_rate": 0.00026103006141579567,
+      "loss": 3.3364,
+      "step": 52550
+    },
+    {
+      "epoch": 5.66139274566785,
+      "grad_norm": 0.7283848524093628,
+      "learning_rate": 0.00026070682038573426,
+      "loss": 3.3309,
+      "step": 52600
+    },
+    {
+      "epoch": 5.666774297707459,
+      "grad_norm": 0.7701716423034668,
+      "learning_rate": 0.00026038357935567286,
+      "loss": 3.3383,
+      "step": 52650
+    },
+    {
+      "epoch": 5.672155849747067,
+      "grad_norm": 0.6934614777565002,
+      "learning_rate": 0.00026006033832561145,
+      "loss": 3.3301,
+      "step": 52700
+    },
+    {
+      "epoch": 5.677537401786675,
+      "grad_norm": 0.7119128108024597,
+      "learning_rate": 0.00025973709729555,
+      "loss": 3.3367,
+      "step": 52750
+    },
+    {
+      "epoch": 5.682918953826284,
+      "grad_norm": 0.7460734248161316,
+      "learning_rate": 0.00025941385626548864,
+      "loss": 3.3561,
+      "step": 52800
+    },
+    {
+      "epoch": 5.688300505865891,
+      "grad_norm": 0.7418410181999207,
+      "learning_rate": 0.0002590906152354272,
+      "loss": 3.324,
+      "step": 52850
+    },
+    {
+      "epoch": 5.6936820579055,
+      "grad_norm": 0.7116801142692566,
+      "learning_rate": 0.0002587673742053658,
+      "loss": 3.3337,
+      "step": 52900
+    },
+    {
+      "epoch": 5.699063609945108,
+      "grad_norm": 0.7011289000511169,
+      "learning_rate": 0.00025844413317530437,
+      "loss": 3.3404,
+      "step": 52950
+    },
+    {
+      "epoch": 5.704445161984716,
+      "grad_norm": 0.7183201909065247,
+      "learning_rate": 0.00025812089214524296,
+      "loss": 3.3489,
+      "step": 53000
+    },
+    {
+      "epoch": 5.704445161984716,
+      "eval_accuracy": 0.3831644088777667,
+      "eval_loss": 3.3893256187438965,
+      "eval_runtime": 213.8188,
+      "eval_samples_per_second": 84.235,
+      "eval_steps_per_second": 5.266,
+      "step": 53000
+    },
+    {
+      "epoch": 5.709826714024325,
+      "grad_norm": 0.7784079313278198,
+      "learning_rate": 0.00025779765111518156,
+      "loss": 3.3336,
+      "step": 53050
+    },
+    {
+      "epoch": 5.715208266063933,
+      "grad_norm": 0.7234706282615662,
+      "learning_rate": 0.0002574744100851201,
+      "loss": 3.3246,
+      "step": 53100
+    },
+    {
+      "epoch": 5.720589818103541,
+      "grad_norm": 0.7097893953323364,
+      "learning_rate": 0.0002571511690550587,
+      "loss": 3.3239,
+      "step": 53150
+    },
+    {
+      "epoch": 5.725971370143149,
+      "grad_norm": 0.6980921030044556,
+      "learning_rate": 0.0002568279280249973,
+      "loss": 3.3451,
+      "step": 53200
+    },
+    {
+      "epoch": 5.731352922182758,
+      "grad_norm": 0.7374033331871033,
+      "learning_rate": 0.0002565046869949359,
+      "loss": 3.345,
+      "step": 53250
+    },
+    {
+      "epoch": 5.736734474222366,
+      "grad_norm": 0.7064562439918518,
+      "learning_rate": 0.0002561814459648744,
+      "loss": 3.3339,
+      "step": 53300
+    },
+    {
+      "epoch": 5.742116026261974,
+      "grad_norm": 0.7073032259941101,
+      "learning_rate": 0.00025585820493481307,
+      "loss": 3.3426,
+      "step": 53350
+    },
+    {
+      "epoch": 5.747497578301582,
+      "grad_norm": 0.6804144382476807,
+      "learning_rate": 0.0002555349639047516,
+      "loss": 3.3255,
+      "step": 53400
+    },
+    {
+      "epoch": 5.75287913034119,
+      "grad_norm": 0.6917299032211304,
+      "learning_rate": 0.0002552117228746902,
+      "loss": 3.3281,
+      "step": 53450
+    },
+    {
+      "epoch": 5.758260682380799,
+      "grad_norm": 0.7830952405929565,
+      "learning_rate": 0.0002548884818446288,
+      "loss": 3.3367,
+      "step": 53500
+    },
+    {
+      "epoch": 5.763642234420407,
+      "grad_norm": 0.6828339099884033,
+      "learning_rate": 0.0002545652408145674,
+      "loss": 3.3337,
+      "step": 53550
+    },
+    {
+      "epoch": 5.769023786460015,
+      "grad_norm": 0.7331600785255432,
+      "learning_rate": 0.00025424199978450594,
+      "loss": 3.336,
+      "step": 53600
+    },
+    {
+      "epoch": 5.774405338499624,
+      "grad_norm": 0.6974768042564392,
+      "learning_rate": 0.00025391875875444453,
+      "loss": 3.3284,
+      "step": 53650
+    },
+    {
+      "epoch": 5.779786890539231,
+      "grad_norm": 0.7154784798622131,
+      "learning_rate": 0.0002535955177243831,
+      "loss": 3.3236,
+      "step": 53700
+    },
+    {
+      "epoch": 5.78516844257884,
+      "grad_norm": 0.7374145984649658,
+      "learning_rate": 0.0002532722766943217,
+      "loss": 3.3258,
+      "step": 53750
+    },
+    {
+      "epoch": 5.790549994618448,
+      "grad_norm": 0.6925607323646545,
+      "learning_rate": 0.00025295550048486153,
+      "loss": 3.3431,
+      "step": 53800
+    },
+    {
+      "epoch": 5.795931546658056,
+      "grad_norm": 0.6958518028259277,
+      "learning_rate": 0.0002526322594548001,
+      "loss": 3.3338,
+      "step": 53850
+    },
+    {
+      "epoch": 5.801313098697665,
+      "grad_norm": 0.6767904758453369,
+      "learning_rate": 0.0002523090184247387,
+      "loss": 3.3531,
+      "step": 53900
+    },
+    {
+      "epoch": 5.806694650737272,
+      "grad_norm": 0.7042453289031982,
+      "learning_rate": 0.00025198577739467726,
+      "loss": 3.3333,
+      "step": 53950
+    },
+    {
+      "epoch": 5.812076202776881,
+      "grad_norm": 0.7105914354324341,
+      "learning_rate": 0.00025166253636461585,
+      "loss": 3.3353,
+      "step": 54000
+    },
+    {
+      "epoch": 5.812076202776881,
+      "eval_accuracy": 0.38365334687293184,
+      "eval_loss": 3.386536121368408,
+      "eval_runtime": 195.2212,
+      "eval_samples_per_second": 92.259,
+      "eval_steps_per_second": 5.768,
+      "step": 54000
+    },
+    {
+      "epoch": 5.817457754816489,
+      "grad_norm": 0.7248342037200928,
+      "learning_rate": 0.00025133929533455445,
+      "loss": 3.3423,
+      "step": 54050
+    },
+    {
+      "epoch": 5.822839306856097,
+      "grad_norm": 0.7134623527526855,
+      "learning_rate": 0.00025101605430449304,
+      "loss": 3.3469,
+      "step": 54100
+    },
+    {
+      "epoch": 5.828220858895706,
+      "grad_norm": 0.7342662215232849,
+      "learning_rate": 0.00025069281327443164,
+      "loss": 3.3406,
+      "step": 54150
+    },
+    {
+      "epoch": 5.833602410935313,
+      "grad_norm": 0.6971819996833801,
+      "learning_rate": 0.0002503695722443702,
+      "loss": 3.3296,
+      "step": 54200
+    },
+    {
+      "epoch": 5.838983962974922,
+      "grad_norm": 0.6914299130439758,
+      "learning_rate": 0.00025004633121430877,
+      "loss": 3.335,
+      "step": 54250
+    },
+    {
+      "epoch": 5.84436551501453,
+      "grad_norm": 0.6908444762229919,
+      "learning_rate": 0.00024972309018424737,
+      "loss": 3.3218,
+      "step": 54300
+    },
+    {
+      "epoch": 5.849747067054138,
+      "grad_norm": 0.7167321443557739,
+      "learning_rate": 0.00024939984915418596,
+      "loss": 3.3309,
+      "step": 54350
+    },
+    {
+      "epoch": 5.855128619093747,
+      "grad_norm": 0.7270901799201965,
+      "learning_rate": 0.0002490766081241245,
+      "loss": 3.3447,
+      "step": 54400
+    },
+    {
+      "epoch": 5.860510171133355,
+      "grad_norm": 0.7137880325317383,
+      "learning_rate": 0.00024875336709406315,
+      "loss": 3.3156,
+      "step": 54450
+    },
+    {
+      "epoch": 5.865891723172963,
+      "grad_norm": 0.6899557709693909,
+      "learning_rate": 0.0002484301260640017,
+      "loss": 3.3324,
+      "step": 54500
+    },
+    {
+      "epoch": 5.871273275212571,
+      "grad_norm": 0.7073144316673279,
+      "learning_rate": 0.0002481068850339403,
+      "loss": 3.3377,
+      "step": 54550
+    },
+    {
+      "epoch": 5.87665482725218,
+      "grad_norm": 0.7121581435203552,
+      "learning_rate": 0.0002477836440038789,
+      "loss": 3.3358,
+      "step": 54600
+    },
+    {
+      "epoch": 5.882036379291788,
+      "grad_norm": 0.7033504843711853,
+      "learning_rate": 0.0002474604029738175,
+      "loss": 3.3309,
+      "step": 54650
+    },
+    {
+      "epoch": 5.887417931331396,
+      "grad_norm": 0.7366927862167358,
+      "learning_rate": 0.000247137161943756,
+      "loss": 3.3426,
+      "step": 54700
+    },
+    {
+      "epoch": 5.892799483371004,
+      "grad_norm": 0.6953083276748657,
+      "learning_rate": 0.0002468139209136946,
+      "loss": 3.3304,
+      "step": 54750
+    },
+    {
+      "epoch": 5.898181035410612,
+      "grad_norm": 0.707262396812439,
+      "learning_rate": 0.0002464906798836332,
+      "loss": 3.3298,
+      "step": 54800
+    },
+    {
+      "epoch": 5.903562587450221,
+      "grad_norm": 0.6863567233085632,
+      "learning_rate": 0.0002461674388535718,
+      "loss": 3.3294,
+      "step": 54850
+    },
+    {
+      "epoch": 5.9089441394898286,
+      "grad_norm": 0.7244144678115845,
+      "learning_rate": 0.0002458441978235104,
+      "loss": 3.3334,
+      "step": 54900
+    },
+    {
+      "epoch": 5.914325691529437,
+      "grad_norm": 0.7477735877037048,
+      "learning_rate": 0.00024552095679344893,
+      "loss": 3.3464,
+      "step": 54950
+    },
+    {
+      "epoch": 5.919707243569046,
+      "grad_norm": 0.7203809022903442,
+      "learning_rate": 0.0002451977157633876,
+      "loss": 3.3306,
+      "step": 55000
+    },
+    {
+      "epoch": 5.919707243569046,
+      "eval_accuracy": 0.3839801747594778,
+      "eval_loss": 3.3797736167907715,
+      "eval_runtime": 195.5869,
+      "eval_samples_per_second": 92.087,
+      "eval_steps_per_second": 5.757,
+      "step": 55000
+    },
+    {
+      "epoch": 5.925088795608653,
+      "grad_norm": 0.7410472631454468,
+      "learning_rate": 0.0002448744747333261,
+      "loss": 3.3449,
+      "step": 55050
+    },
+    {
+      "epoch": 5.930470347648262,
+      "grad_norm": 0.6759241223335266,
+      "learning_rate": 0.0002445512337032647,
+      "loss": 3.3361,
+      "step": 55100
+    },
+    {
+      "epoch": 5.93585189968787,
+      "grad_norm": 0.6957202553749084,
+      "learning_rate": 0.0002442279926732033,
+      "loss": 3.3363,
+      "step": 55150
+    },
+    {
+      "epoch": 5.941233451727478,
+      "grad_norm": 0.7306824922561646,
+      "learning_rate": 0.00024390475164314188,
+      "loss": 3.3315,
+      "step": 55200
+    },
+    {
+      "epoch": 5.946615003767087,
+      "grad_norm": 0.760296106338501,
+      "learning_rate": 0.00024358151061308045,
+      "loss": 3.3511,
+      "step": 55250
+    },
+    {
+      "epoch": 5.951996555806694,
+      "grad_norm": 0.7307605743408203,
+      "learning_rate": 0.00024325826958301907,
+      "loss": 3.3263,
+      "step": 55300
+    },
+    {
+      "epoch": 5.957378107846303,
+      "grad_norm": 0.6909796595573425,
+      "learning_rate": 0.00024293502855295764,
+      "loss": 3.3515,
+      "step": 55350
+    },
+    {
+      "epoch": 5.962759659885911,
+      "grad_norm": 0.7553929686546326,
+      "learning_rate": 0.0002426117875228962,
+      "loss": 3.3455,
+      "step": 55400
+    },
+    {
+      "epoch": 5.968141211925519,
+      "grad_norm": 0.7335132360458374,
+      "learning_rate": 0.0002422885464928348,
+      "loss": 3.3374,
+      "step": 55450
+    },
+    {
+      "epoch": 5.973522763965128,
+      "grad_norm": 0.7116664052009583,
+      "learning_rate": 0.0002419653054627734,
+      "loss": 3.3307,
+      "step": 55500
+    },
+    {
+      "epoch": 5.978904316004736,
+      "grad_norm": 0.7170029282569885,
+      "learning_rate": 0.00024164206443271196,
+      "loss": 3.355,
+      "step": 55550
+    },
+    {
+      "epoch": 5.984285868044344,
+      "grad_norm": 0.6848576068878174,
+      "learning_rate": 0.00024131882340265056,
+      "loss": 3.3344,
+      "step": 55600
+    },
+    {
+      "epoch": 5.989667420083952,
+      "grad_norm": 0.7043144702911377,
+      "learning_rate": 0.00024099558237258912,
+      "loss": 3.3398,
+      "step": 55650
+    },
+    {
+      "epoch": 5.995048972123561,
+      "grad_norm": 0.7126272320747375,
+      "learning_rate": 0.00024067234134252772,
+      "loss": 3.3339,
+      "step": 55700
+    },
+    {
+      "epoch": 6.000430524163169,
+      "grad_norm": 0.7398678064346313,
+      "learning_rate": 0.00024035556513306752,
+      "loss": 3.3327,
+      "step": 55750
+    },
+    {
+      "epoch": 6.005812076202777,
+      "grad_norm": 0.745004415512085,
+      "learning_rate": 0.00024003232410300615,
+      "loss": 3.2401,
+      "step": 55800
+    },
+    {
+      "epoch": 6.011193628242385,
+      "grad_norm": 0.7168341875076294,
+      "learning_rate": 0.00023970908307294471,
+      "loss": 3.254,
+      "step": 55850
+    },
+    {
+      "epoch": 6.016575180281993,
+      "grad_norm": 0.7345452904701233,
+      "learning_rate": 0.00023938584204288328,
+      "loss": 3.2373,
+      "step": 55900
+    },
+    {
+      "epoch": 6.021956732321602,
+      "grad_norm": 0.7234600186347961,
+      "learning_rate": 0.00023906260101282188,
+      "loss": 3.2584,
+      "step": 55950
+    },
+    {
+      "epoch": 6.0273382843612096,
+      "grad_norm": 0.7159119248390198,
+      "learning_rate": 0.00023873935998276047,
+      "loss": 3.2266,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0273382843612096,
+      "eval_accuracy": 0.3844155468809504,
+      "eval_loss": 3.382514715194702,
+      "eval_runtime": 205.1809,
+      "eval_samples_per_second": 87.781,
+      "eval_steps_per_second": 5.488,
+      "step": 56000
+    },
+    {
+      "epoch": 6.032719836400818,
+      "grad_norm": 0.6958070397377014,
+      "learning_rate": 0.00023841611895269904,
+      "loss": 3.2395,
+      "step": 56050
+    },
+    {
+      "epoch": 6.038101388440427,
+      "grad_norm": 0.7117956280708313,
+      "learning_rate": 0.00023809287792263763,
+      "loss": 3.2614,
+      "step": 56100
+    },
+    {
+      "epoch": 6.043482940480034,
+      "grad_norm": 0.7864623069763184,
+      "learning_rate": 0.0002377696368925762,
+      "loss": 3.2521,
+      "step": 56150
+    },
+    {
+      "epoch": 6.048864492519643,
+      "grad_norm": 0.715238630771637,
+      "learning_rate": 0.0002374463958625148,
+      "loss": 3.2547,
+      "step": 56200
+    },
+    {
+      "epoch": 6.0542460445592505,
+      "grad_norm": 0.6912676692008972,
+      "learning_rate": 0.0002371231548324534,
+      "loss": 3.253,
+      "step": 56250
+    },
+    {
+      "epoch": 6.059627596598859,
+      "grad_norm": 0.7019177675247192,
+      "learning_rate": 0.00023679991380239196,
+      "loss": 3.2513,
+      "step": 56300
+    },
+    {
+      "epoch": 6.065009148638468,
+      "grad_norm": 0.8073034286499023,
+      "learning_rate": 0.00023647667277233053,
+      "loss": 3.2551,
+      "step": 56350
+    },
+    {
+      "epoch": 6.070390700678075,
+      "grad_norm": 0.716173529624939,
+      "learning_rate": 0.00023615343174226915,
+      "loss": 3.2406,
+      "step": 56400
+    },
+    {
+      "epoch": 6.075772252717684,
+      "grad_norm": 0.7097351551055908,
+      "learning_rate": 0.00023583019071220771,
+      "loss": 3.2585,
+      "step": 56450
+    },
+    {
+      "epoch": 6.081153804757292,
+      "grad_norm": 0.7224815487861633,
+      "learning_rate": 0.00023550694968214628,
+      "loss": 3.2563,
+      "step": 56500
+    },
+    {
+      "epoch": 6.0865353567969,
+      "grad_norm": 0.7451142072677612,
+      "learning_rate": 0.00023518370865208488,
+      "loss": 3.2523,
+      "step": 56550
+    },
+    {
+      "epoch": 6.091916908836509,
+      "grad_norm": 0.7142341732978821,
+      "learning_rate": 0.00023486046762202347,
+      "loss": 3.2581,
+      "step": 56600
+    },
+    {
+      "epoch": 6.097298460876116,
+      "grad_norm": 0.699760913848877,
+      "learning_rate": 0.00023453722659196207,
+      "loss": 3.2556,
+      "step": 56650
+    },
+    {
+      "epoch": 6.102680012915725,
+      "grad_norm": 0.7422381639480591,
+      "learning_rate": 0.00023421398556190063,
+      "loss": 3.2764,
+      "step": 56700
+    },
+    {
+      "epoch": 6.108061564955333,
+      "grad_norm": 0.8956770896911621,
+      "learning_rate": 0.0002338907445318392,
+      "loss": 3.2671,
+      "step": 56750
+    },
+    {
+      "epoch": 6.113443116994941,
+      "grad_norm": 0.7111889123916626,
+      "learning_rate": 0.00023356750350177782,
+      "loss": 3.2572,
+      "step": 56800
+    },
+    {
+      "epoch": 6.11882466903455,
+      "grad_norm": 0.9345306158065796,
+      "learning_rate": 0.0002332442624717164,
+      "loss": 3.2521,
+      "step": 56850
+    },
+    {
+      "epoch": 6.124206221074158,
+      "grad_norm": 0.7105541825294495,
+      "learning_rate": 0.00023292102144165496,
+      "loss": 3.2668,
+      "step": 56900
+    },
+    {
+      "epoch": 6.129587773113766,
+      "grad_norm": 0.7327850461006165,
+      "learning_rate": 0.00023259778041159358,
+      "loss": 3.2601,
+      "step": 56950
+    },
+    {
+      "epoch": 6.134969325153374,
+      "grad_norm": 0.7219511866569519,
+      "learning_rate": 0.00023227453938153215,
+      "loss": 3.2555,
+      "step": 57000
+    },
+    {
+      "epoch": 6.134969325153374,
+      "eval_accuracy": 0.38421410442694237,
+      "eval_loss": 3.3847293853759766,
+      "eval_runtime": 196.0722,
+      "eval_samples_per_second": 91.859,
+      "eval_steps_per_second": 5.743,
+      "step": 57000
+    },
+    {
+      "epoch": 6.140350877192983,
+      "grad_norm": 0.7666511535644531,
+      "learning_rate": 0.00023195129835147071,
+      "loss": 3.2573,
+      "step": 57050
+    },
+    {
+      "epoch": 6.1457324292325906,
+      "grad_norm": 0.8009618520736694,
+      "learning_rate": 0.0002316280573214093,
+      "loss": 3.2679,
+      "step": 57100
+    },
+    {
+      "epoch": 6.151113981272199,
+      "grad_norm": 0.6945069432258606,
+      "learning_rate": 0.0002313048162913479,
+      "loss": 3.2667,
+      "step": 57150
+    },
+    {
+      "epoch": 6.156495533311807,
+      "grad_norm": 0.750709593296051,
+      "learning_rate": 0.00023098157526128647,
+      "loss": 3.2475,
+      "step": 57200
+    },
+    {
+      "epoch": 6.161877085351415,
+      "grad_norm": 0.7243725657463074,
+      "learning_rate": 0.00023065833423122507,
+      "loss": 3.2824,
+      "step": 57250
+    },
+    {
+      "epoch": 6.167258637391024,
+      "grad_norm": 0.7340604066848755,
+      "learning_rate": 0.00023033509320116363,
+      "loss": 3.2589,
+      "step": 57300
+    },
+    {
+      "epoch": 6.1726401894306315,
+      "grad_norm": 0.7163415551185608,
+      "learning_rate": 0.00023001185217110223,
+      "loss": 3.2712,
+      "step": 57350
+    },
+    {
+      "epoch": 6.17802174147024,
+      "grad_norm": 0.7414380311965942,
+      "learning_rate": 0.00022968861114104082,
+      "loss": 3.2653,
+      "step": 57400
+    },
+    {
+      "epoch": 6.183403293509849,
+      "grad_norm": 0.7300434112548828,
+      "learning_rate": 0.0002293653701109794,
+      "loss": 3.274,
+      "step": 57450
+    },
+    {
+      "epoch": 6.188784845549456,
+      "grad_norm": 0.7358796000480652,
+      "learning_rate": 0.00022904212908091796,
+      "loss": 3.2541,
+      "step": 57500
+    },
+    {
+      "epoch": 6.194166397589065,
+      "grad_norm": 0.7499557137489319,
+      "learning_rate": 0.00022871888805085658,
+      "loss": 3.2607,
+      "step": 57550
+    },
+    {
+      "epoch": 6.1995479496286725,
+      "grad_norm": 0.77440345287323,
+      "learning_rate": 0.00022839564702079515,
+      "loss": 3.2597,
+      "step": 57600
+    },
+    {
+      "epoch": 6.204929501668281,
+      "grad_norm": 0.759096622467041,
+      "learning_rate": 0.00022807240599073374,
+      "loss": 3.2771,
+      "step": 57650
+    },
+    {
+      "epoch": 6.21031105370789,
+      "grad_norm": 0.7167288064956665,
+      "learning_rate": 0.00022774916496067234,
+      "loss": 3.2498,
+      "step": 57700
+    },
+    {
+      "epoch": 6.215692605747497,
+      "grad_norm": 0.7278882265090942,
+      "learning_rate": 0.0002274259239306109,
+      "loss": 3.2825,
+      "step": 57750
+    },
+    {
+      "epoch": 6.221074157787106,
+      "grad_norm": 0.7202818989753723,
+      "learning_rate": 0.0002271026829005495,
+      "loss": 3.2604,
+      "step": 57800
+    },
+    {
+      "epoch": 6.226455709826714,
+      "grad_norm": 0.7370897531509399,
+      "learning_rate": 0.00022677944187048807,
+      "loss": 3.2528,
+      "step": 57850
+    },
+    {
+      "epoch": 6.231837261866322,
+      "grad_norm": 0.7153465151786804,
+      "learning_rate": 0.00022645620084042666,
+      "loss": 3.2778,
+      "step": 57900
+    },
+    {
+      "epoch": 6.237218813905931,
+      "grad_norm": 0.7415947914123535,
+      "learning_rate": 0.00022613295981036526,
+      "loss": 3.2753,
+      "step": 57950
+    },
+    {
+      "epoch": 6.242600365945538,
+      "grad_norm": 0.734861433506012,
+      "learning_rate": 0.00022580971878030382,
+      "loss": 3.2849,
+      "step": 58000
+    },
+    {
+      "epoch": 6.242600365945538,
+      "eval_accuracy": 0.3847145596282159,
+      "eval_loss": 3.380995988845825,
+      "eval_runtime": 200.7754,
+      "eval_samples_per_second": 89.707,
+      "eval_steps_per_second": 5.608,
+      "step": 58000
+    },
+    {
+      "epoch": 6.247981917985147,
+      "grad_norm": 0.7297642827033997,
+      "learning_rate": 0.00022549294257084366,
+      "loss": 3.2695,
+      "step": 58050
+    },
+    {
+      "epoch": 6.253363470024755,
+      "grad_norm": 0.7514773011207581,
+      "learning_rate": 0.00022516970154078222,
+      "loss": 3.2576,
+      "step": 58100
+    },
+    {
+      "epoch": 6.258745022064363,
+      "grad_norm": 0.7224261164665222,
+      "learning_rate": 0.0002248464605107208,
+      "loss": 3.2726,
+      "step": 58150
+    },
+    {
+      "epoch": 6.264126574103972,
+      "grad_norm": 0.7479972839355469,
+      "learning_rate": 0.0002245232194806594,
+      "loss": 3.2817,
+      "step": 58200
+    },
+    {
+      "epoch": 6.26950812614358,
+      "grad_norm": 0.7621592879295349,
+      "learning_rate": 0.00022419997845059798,
+      "loss": 3.2763,
+      "step": 58250
+    },
+    {
+      "epoch": 6.274889678183188,
+      "grad_norm": 0.7210646271705627,
+      "learning_rate": 0.00022387673742053655,
+      "loss": 3.277,
+      "step": 58300
+    },
+    {
+      "epoch": 6.280271230222796,
+      "grad_norm": 0.7708672881126404,
+      "learning_rate": 0.00022355349639047514,
+      "loss": 3.2824,
+      "step": 58350
+    },
+    {
+      "epoch": 6.285652782262405,
+      "grad_norm": 0.7061500549316406,
+      "learning_rate": 0.0002232302553604137,
+      "loss": 3.2647,
+      "step": 58400
+    },
+    {
+      "epoch": 6.2910343343020125,
+      "grad_norm": 0.7927210927009583,
+      "learning_rate": 0.00022290701433035233,
+      "loss": 3.2727,
+      "step": 58450
+    },
+    {
+      "epoch": 6.296415886341621,
+      "grad_norm": 0.7816336154937744,
+      "learning_rate": 0.0002225837733002909,
+      "loss": 3.2825,
+      "step": 58500
+    },
+    {
+      "epoch": 6.301797438381229,
+      "grad_norm": 0.7397032380104065,
+      "learning_rate": 0.00022226053227022947,
+      "loss": 3.2754,
+      "step": 58550
+    },
+    {
+      "epoch": 6.307178990420837,
+      "grad_norm": 0.7220003008842468,
+      "learning_rate": 0.0002219372912401681,
+      "loss": 3.2776,
+      "step": 58600
+    },
+    {
+      "epoch": 6.312560542460446,
+      "grad_norm": 0.7121575474739075,
+      "learning_rate": 0.00022161405021010666,
+      "loss": 3.2601,
+      "step": 58650
+    },
+    {
+      "epoch": 6.3179420945000535,
+      "grad_norm": 0.7781762480735779,
+      "learning_rate": 0.00022129080918004523,
+      "loss": 3.2796,
+      "step": 58700
+    },
+    {
+      "epoch": 6.323323646539662,
+      "grad_norm": 0.7582905888557434,
+      "learning_rate": 0.00022096756814998382,
+      "loss": 3.2895,
+      "step": 58750
+    },
+    {
+      "epoch": 6.328705198579271,
+      "grad_norm": 0.7163707613945007,
+      "learning_rate": 0.00022064432711992241,
+      "loss": 3.2841,
+      "step": 58800
+    },
+    {
+      "epoch": 6.334086750618878,
+      "grad_norm": 0.7681860327720642,
+      "learning_rate": 0.00022032108608986098,
+      "loss": 3.2909,
+      "step": 58850
+    },
+    {
+      "epoch": 6.339468302658487,
+      "grad_norm": 0.7766373157501221,
+      "learning_rate": 0.00021999784505979958,
+      "loss": 3.2916,
+      "step": 58900
+    },
+    {
+      "epoch": 6.344849854698095,
+      "grad_norm": 0.739279568195343,
+      "learning_rate": 0.00021967460402973814,
+      "loss": 3.2823,
+      "step": 58950
+    },
+    {
+      "epoch": 6.350231406737703,
+      "grad_norm": 0.7410812377929688,
+      "learning_rate": 0.00021935136299967674,
+      "loss": 3.2801,
+      "step": 59000
+    },
+    {
+      "epoch": 6.350231406737703,
+      "eval_accuracy": 0.38521621001125533,
+      "eval_loss": 3.3765883445739746,
+      "eval_runtime": 194.9686,
+      "eval_samples_per_second": 92.379,
+      "eval_steps_per_second": 5.775,
+      "step": 59000
+    },
+    {
+      "epoch": 6.355612958777312,
+      "grad_norm": 0.7452576756477356,
+      "learning_rate": 0.00021902812196961533,
+      "loss": 3.2732,
+      "step": 59050
+    },
+    {
+      "epoch": 6.360994510816919,
+      "grad_norm": 0.7552573680877686,
+      "learning_rate": 0.0002187048809395539,
+      "loss": 3.266,
+      "step": 59100
+    },
+    {
+      "epoch": 6.366376062856528,
+      "grad_norm": 0.7315678596496582,
+      "learning_rate": 0.00021838163990949247,
+      "loss": 3.2772,
+      "step": 59150
+    },
+    {
+      "epoch": 6.371757614896136,
+      "grad_norm": 0.7016833424568176,
+      "learning_rate": 0.0002180583988794311,
+      "loss": 3.2874,
+      "step": 59200
+    },
+    {
+      "epoch": 6.377139166935744,
+      "grad_norm": 0.7350359559059143,
+      "learning_rate": 0.00021773515784936966,
+      "loss": 3.2959,
+      "step": 59250
+    },
+    {
+      "epoch": 6.382520718975353,
+      "grad_norm": 0.7227166891098022,
+      "learning_rate": 0.00021741191681930823,
+      "loss": 3.281,
+      "step": 59300
+    },
+    {
+      "epoch": 6.387902271014961,
+      "grad_norm": 0.7294051051139832,
+      "learning_rate": 0.00021708867578924685,
+      "loss": 3.298,
+      "step": 59350
+    },
+    {
+      "epoch": 6.393283823054569,
+      "grad_norm": 0.718024730682373,
+      "learning_rate": 0.00021676543475918541,
+      "loss": 3.2599,
+      "step": 59400
+    },
+    {
+      "epoch": 6.398665375094177,
+      "grad_norm": 0.7561959624290466,
+      "learning_rate": 0.000216442193729124,
+      "loss": 3.2947,
+      "step": 59450
+    },
+    {
+      "epoch": 6.404046927133785,
+      "grad_norm": 0.7592124938964844,
+      "learning_rate": 0.00021611895269906258,
+      "loss": 3.2771,
+      "step": 59500
+    },
+    {
+      "epoch": 6.4094284791733935,
+      "grad_norm": 0.750074565410614,
+      "learning_rate": 0.00021579571166900117,
+      "loss": 3.26,
+      "step": 59550
+    },
+    {
+      "epoch": 6.414810031213002,
+      "grad_norm": 0.7350614666938782,
+      "learning_rate": 0.00021547247063893977,
+      "loss": 3.2863,
+      "step": 59600
+    },
+    {
+      "epoch": 6.42019158325261,
+      "grad_norm": 0.7412868738174438,
+      "learning_rate": 0.00021514922960887833,
+      "loss": 3.269,
+      "step": 59650
+    },
+    {
+      "epoch": 6.425573135292218,
+      "grad_norm": 0.7642093300819397,
+      "learning_rate": 0.0002148259885788169,
+      "loss": 3.2749,
+      "step": 59700
+    },
+    {
+      "epoch": 6.430954687331827,
+      "grad_norm": 0.7282441258430481,
+      "learning_rate": 0.00021450274754875552,
+      "loss": 3.269,
+      "step": 59750
+    },
+    {
+      "epoch": 6.4363362393714345,
+      "grad_norm": 0.7450237274169922,
+      "learning_rate": 0.0002141795065186941,
+      "loss": 3.2695,
+      "step": 59800
+    },
+    {
+      "epoch": 6.441717791411043,
+      "grad_norm": 0.7366235852241516,
+      "learning_rate": 0.00021385626548863266,
+      "loss": 3.2876,
+      "step": 59850
+    },
+    {
+      "epoch": 6.447099343450651,
+      "grad_norm": 0.7377769947052002,
+      "learning_rate": 0.00021353302445857128,
+      "loss": 3.2793,
+      "step": 59900
+    },
+    {
+      "epoch": 6.452480895490259,
+      "grad_norm": 0.771925151348114,
+      "learning_rate": 0.00021320978342850985,
+      "loss": 3.2807,
+      "step": 59950
+    },
+    {
+      "epoch": 6.457862447529868,
+      "grad_norm": 0.7846429944038391,
+      "learning_rate": 0.00021288654239844842,
+      "loss": 3.2773,
+      "step": 60000
+    },
+    {
+      "epoch": 6.457862447529868,
+      "eval_accuracy": 0.385375712450567,
+      "eval_loss": 3.3726203441619873,
+      "eval_runtime": 198.7032,
+      "eval_samples_per_second": 90.643,
+      "eval_steps_per_second": 5.667,
+      "step": 60000
+    },
+    {
+      "epoch": 6.4632439995694755,
+      "grad_norm": 0.7637952566146851,
+      "learning_rate": 0.000212563301368387,
+      "loss": 3.2915,
+      "step": 60050
+    },
+    {
+      "epoch": 6.468625551609084,
+      "grad_norm": 0.7368430495262146,
+      "learning_rate": 0.00021224006033832558,
+      "loss": 3.2963,
+      "step": 60100
+    },
+    {
+      "epoch": 6.474007103648693,
+      "grad_norm": 0.7520938515663147,
+      "learning_rate": 0.0002119232841288654,
+      "loss": 3.2928,
+      "step": 60150
+    },
+    {
+      "epoch": 6.4793886556883,
+      "grad_norm": 0.739131510257721,
+      "learning_rate": 0.00021160004309880398,
+      "loss": 3.2874,
+      "step": 60200
+    },
+    {
+      "epoch": 6.484770207727909,
+      "grad_norm": 0.7567487359046936,
+      "learning_rate": 0.0002112768020687426,
+      "loss": 3.2851,
+      "step": 60250
+    },
+    {
+      "epoch": 6.490151759767517,
+      "grad_norm": 0.7027845978736877,
+      "learning_rate": 0.00021095356103868117,
+      "loss": 3.2906,
+      "step": 60300
+    },
+    {
+      "epoch": 6.495533311807125,
+      "grad_norm": 0.7968448400497437,
+      "learning_rate": 0.00021063032000861974,
+      "loss": 3.2912,
+      "step": 60350
+    },
+    {
+      "epoch": 6.500914863846734,
+      "grad_norm": 0.7443140745162964,
+      "learning_rate": 0.00021030707897855833,
+      "loss": 3.2914,
+      "step": 60400
+    },
+    {
+      "epoch": 6.506296415886341,
+      "grad_norm": 0.7305548787117004,
+      "learning_rate": 0.00020998383794849692,
+      "loss": 3.2959,
+      "step": 60450
+    },
+    {
+      "epoch": 6.51167796792595,
+      "grad_norm": 0.7129182815551758,
+      "learning_rate": 0.0002096605969184355,
+      "loss": 3.275,
+      "step": 60500
+    },
+    {
+      "epoch": 6.517059519965558,
+      "grad_norm": 0.697638213634491,
+      "learning_rate": 0.0002093373558883741,
+      "loss": 3.3068,
+      "step": 60550
+    },
+    {
+      "epoch": 6.522441072005166,
+      "grad_norm": 0.7496232986450195,
+      "learning_rate": 0.00020901411485831265,
+      "loss": 3.2661,
+      "step": 60600
+    },
+    {
+      "epoch": 6.5278226240447745,
+      "grad_norm": 0.800989031791687,
+      "learning_rate": 0.00020869087382825125,
+      "loss": 3.2937,
+      "step": 60650
+    },
+    {
+      "epoch": 6.533204176084383,
+      "grad_norm": 0.7812984585762024,
+      "learning_rate": 0.00020836763279818984,
+      "loss": 3.2756,
+      "step": 60700
+    },
+    {
+      "epoch": 6.538585728123991,
+      "grad_norm": 0.7700937986373901,
+      "learning_rate": 0.0002080443917681284,
+      "loss": 3.2949,
+      "step": 60750
+    },
+    {
+      "epoch": 6.543967280163599,
+      "grad_norm": 0.7460906505584717,
+      "learning_rate": 0.00020772115073806698,
+      "loss": 3.2751,
+      "step": 60800
+    },
+    {
+      "epoch": 6.549348832203208,
+      "grad_norm": 0.7481933832168579,
+      "learning_rate": 0.0002073979097080056,
+      "loss": 3.2899,
+      "step": 60850
+    },
+    {
+      "epoch": 6.5547303842428155,
+      "grad_norm": 0.7331646680831909,
+      "learning_rate": 0.00020707466867794417,
+      "loss": 3.306,
+      "step": 60900
+    },
+    {
+      "epoch": 6.560111936282424,
+      "grad_norm": 0.7299278974533081,
+      "learning_rate": 0.00020675142764788274,
+      "loss": 3.2903,
+      "step": 60950
+    },
+    {
+      "epoch": 6.565493488322032,
+      "grad_norm": 0.7748175859451294,
+      "learning_rate": 0.00020642818661782136,
+      "loss": 3.3014,
+      "step": 61000
+    },
+    {
+      "epoch": 6.565493488322032,
+      "eval_accuracy": 0.38595026892133,
+      "eval_loss": 3.3667492866516113,
+      "eval_runtime": 196.5736,
+      "eval_samples_per_second": 91.625,
+      "eval_steps_per_second": 5.728,
+      "step": 61000
+    },
+    {
+      "epoch": 6.57087504036164,
+      "grad_norm": 0.7791305184364319,
+      "learning_rate": 0.00020610494558775993,
+      "loss": 3.2782,
+      "step": 61050
+    },
+    {
+      "epoch": 6.576256592401249,
+      "grad_norm": 0.7687615752220154,
+      "learning_rate": 0.0002057817045576985,
+      "loss": 3.292,
+      "step": 61100
+    },
+    {
+      "epoch": 6.5816381444408565,
+      "grad_norm": 0.7428541779518127,
+      "learning_rate": 0.0002054584635276371,
+      "loss": 3.2634,
+      "step": 61150
+    },
+    {
+      "epoch": 6.587019696480465,
+      "grad_norm": 0.7501689195632935,
+      "learning_rate": 0.00020513522249757568,
+      "loss": 3.2993,
+      "step": 61200
+    },
+    {
+      "epoch": 6.592401248520073,
+      "grad_norm": 0.7482814192771912,
+      "learning_rate": 0.00020481198146751428,
+      "loss": 3.2895,
+      "step": 61250
+    },
+    {
+      "epoch": 6.597782800559681,
+      "grad_norm": 0.7531591653823853,
+      "learning_rate": 0.00020448874043745284,
+      "loss": 3.2832,
+      "step": 61300
+    },
+    {
+      "epoch": 6.60316435259929,
+      "grad_norm": 0.8052058815956116,
+      "learning_rate": 0.0002041654994073914,
+      "loss": 3.3006,
+      "step": 61350
+    },
+    {
+      "epoch": 6.608545904638898,
+      "grad_norm": 0.7585821151733398,
+      "learning_rate": 0.00020384225837733003,
+      "loss": 3.2867,
+      "step": 61400
+    },
+    {
+      "epoch": 6.613927456678506,
+      "grad_norm": 0.7599231004714966,
+      "learning_rate": 0.0002035190173472686,
+      "loss": 3.2948,
+      "step": 61450
+    },
+    {
+      "epoch": 6.619309008718115,
+      "grad_norm": 0.7827469110488892,
+      "learning_rate": 0.00020319577631720717,
+      "loss": 3.2863,
+      "step": 61500
+    },
+    {
+      "epoch": 6.624690560757722,
+      "grad_norm": 0.7314283847808838,
+      "learning_rate": 0.00020287253528714576,
+      "loss": 3.2803,
+      "step": 61550
+    },
+    {
+      "epoch": 6.630072112797331,
+      "grad_norm": 0.7610256671905518,
+      "learning_rate": 0.00020254929425708436,
+      "loss": 3.2964,
+      "step": 61600
+    },
+    {
+      "epoch": 6.635453664836939,
+      "grad_norm": 0.7744382619857788,
+      "learning_rate": 0.00020222605322702293,
+      "loss": 3.3023,
+      "step": 61650
+    },
+    {
+      "epoch": 6.640835216876547,
+      "grad_norm": 0.7310662269592285,
+      "learning_rate": 0.00020190281219696152,
+      "loss": 3.2774,
+      "step": 61700
+    },
+    {
+      "epoch": 6.6462167689161555,
+      "grad_norm": 0.7460449934005737,
+      "learning_rate": 0.0002015795711669001,
+      "loss": 3.2858,
+      "step": 61750
+    },
+    {
+      "epoch": 6.651598320955763,
+      "grad_norm": 0.790534496307373,
+      "learning_rate": 0.00020125633013683868,
+      "loss": 3.2875,
+      "step": 61800
+    },
+    {
+      "epoch": 6.656979872995372,
+      "grad_norm": 0.7475619316101074,
+      "learning_rate": 0.00020093308910677728,
+      "loss": 3.2868,
+      "step": 61850
+    },
+    {
+      "epoch": 6.66236142503498,
+      "grad_norm": 0.7804775834083557,
+      "learning_rate": 0.00020060984807671584,
+      "loss": 3.2854,
+      "step": 61900
+    },
+    {
+      "epoch": 6.667742977074588,
+      "grad_norm": 0.7852612733840942,
+      "learning_rate": 0.0002002866070466544,
+      "loss": 3.2824,
+      "step": 61950
+    },
+    {
+      "epoch": 6.6731245291141965,
+      "grad_norm": 0.7747927904129028,
+      "learning_rate": 0.00019996336601659303,
+      "loss": 3.2774,
+      "step": 62000
+    },
+    {
+      "epoch": 6.6731245291141965,
+      "eval_accuracy": 0.3861736592586766,
+      "eval_loss": 3.363661050796509,
+      "eval_runtime": 194.5926,
+      "eval_samples_per_second": 92.557,
+      "eval_steps_per_second": 5.786,
+      "step": 62000
+    },
+    {
+      "epoch": 6.678506081153805,
+      "grad_norm": 0.7766702175140381,
+      "learning_rate": 0.0001996401249865316,
+      "loss": 3.2727,
+      "step": 62050
+    },
+    {
+      "epoch": 6.683887633193413,
+      "grad_norm": 0.759154200553894,
+      "learning_rate": 0.00019931688395647017,
+      "loss": 3.2908,
+      "step": 62100
+    },
+    {
+      "epoch": 6.689269185233021,
+      "grad_norm": 0.7269113063812256,
+      "learning_rate": 0.0001989936429264088,
+      "loss": 3.2694,
+      "step": 62150
+    },
+    {
+      "epoch": 6.69465073727263,
+      "grad_norm": 0.7533604502677917,
+      "learning_rate": 0.00019867040189634736,
+      "loss": 3.2954,
+      "step": 62200
+    },
+    {
+      "epoch": 6.7000322893122375,
+      "grad_norm": 0.7951676845550537,
+      "learning_rate": 0.00019834716086628595,
+      "loss": 3.291,
+      "step": 62250
+    },
+    {
+      "epoch": 6.705413841351846,
+      "grad_norm": 0.7553754448890686,
+      "learning_rate": 0.00019802391983622452,
+      "loss": 3.2993,
+      "step": 62300
+    },
+    {
+      "epoch": 6.710795393391454,
+      "grad_norm": 0.7384345531463623,
+      "learning_rate": 0.00019770714362676435,
+      "loss": 3.3098,
+      "step": 62350
+    },
+    {
+      "epoch": 6.716176945431062,
+      "grad_norm": 0.7837821841239929,
+      "learning_rate": 0.00019738390259670292,
+      "loss": 3.2873,
+      "step": 62400
+    },
+    {
+      "epoch": 6.721558497470671,
+      "grad_norm": 0.8101162910461426,
+      "learning_rate": 0.0001970606615666415,
+      "loss": 3.2793,
+      "step": 62450
+    },
+    {
+      "epoch": 6.7269400495102785,
+      "grad_norm": 0.7428802847862244,
+      "learning_rate": 0.0001967374205365801,
+      "loss": 3.2889,
+      "step": 62500
+    },
+    {
+      "epoch": 6.732321601549887,
+      "grad_norm": 0.7610371708869934,
+      "learning_rate": 0.00019641417950651868,
+      "loss": 3.2868,
+      "step": 62550
+    },
+    {
+      "epoch": 6.737703153589496,
+      "grad_norm": 0.7788252830505371,
+      "learning_rate": 0.00019609093847645725,
+      "loss": 3.2811,
+      "step": 62600
+    },
+    {
+      "epoch": 6.743084705629103,
+      "grad_norm": 0.7891542911529541,
+      "learning_rate": 0.00019576769744639587,
+      "loss": 3.2871,
+      "step": 62650
+    },
+    {
+      "epoch": 6.748466257668712,
+      "grad_norm": 0.7551553845405579,
+      "learning_rate": 0.00019544445641633444,
+      "loss": 3.2824,
+      "step": 62700
+    },
+    {
+      "epoch": 6.75384780970832,
+      "grad_norm": 0.7407161593437195,
+      "learning_rate": 0.000195121215386273,
+      "loss": 3.2785,
+      "step": 62750
+    },
+    {
+      "epoch": 6.759229361747928,
+      "grad_norm": 0.7767391204833984,
+      "learning_rate": 0.0001947979743562116,
+      "loss": 3.2957,
+      "step": 62800
+    },
+    {
+      "epoch": 6.7646109137875365,
+      "grad_norm": 0.7716450691223145,
+      "learning_rate": 0.00019447473332615017,
+      "loss": 3.2908,
+      "step": 62850
+    },
+    {
+      "epoch": 6.769992465827144,
+      "grad_norm": 0.7589283585548401,
+      "learning_rate": 0.00019415149229608876,
+      "loss": 3.2708,
+      "step": 62900
+    },
+    {
+      "epoch": 6.775374017866753,
+      "grad_norm": 0.7275950312614441,
+      "learning_rate": 0.00019382825126602735,
+      "loss": 3.283,
+      "step": 62950
+    },
+    {
+      "epoch": 6.780755569906361,
+      "grad_norm": 0.7468757629394531,
+      "learning_rate": 0.00019350501023596592,
+      "loss": 3.2826,
+      "step": 63000
+    },
+    {
+      "epoch": 6.780755569906361,
+      "eval_accuracy": 0.38676788190213396,
+      "eval_loss": 3.3597800731658936,
+      "eval_runtime": 200.2482,
+      "eval_samples_per_second": 89.943,
+      "eval_steps_per_second": 5.623,
+      "step": 63000
+    },
+    {
+      "epoch": 6.786137121945969,
+      "grad_norm": 0.745414137840271,
+      "learning_rate": 0.00019318176920590454,
+      "loss": 3.273,
+      "step": 63050
+    },
+    {
+      "epoch": 6.7915186739855775,
+      "grad_norm": 0.756697952747345,
+      "learning_rate": 0.0001928585281758431,
+      "loss": 3.2861,
+      "step": 63100
+    },
+    {
+      "epoch": 6.796900226025185,
+      "grad_norm": 0.7485399842262268,
+      "learning_rate": 0.00019253528714578168,
+      "loss": 3.2847,
+      "step": 63150
+    },
+    {
+      "epoch": 6.802281778064794,
+      "grad_norm": 0.7797415256500244,
+      "learning_rate": 0.00019221204611572027,
+      "loss": 3.2908,
+      "step": 63200
+    },
+    {
+      "epoch": 6.807663330104402,
+      "grad_norm": 0.818111002445221,
+      "learning_rate": 0.00019188880508565887,
+      "loss": 3.2909,
+      "step": 63250
+    },
+    {
+      "epoch": 6.813044882144011,
+      "grad_norm": 0.7770239114761353,
+      "learning_rate": 0.00019156556405559744,
+      "loss": 3.2904,
+      "step": 63300
+    },
+    {
+      "epoch": 6.8184264341836185,
+      "grad_norm": 0.7970494627952576,
+      "learning_rate": 0.00019124232302553603,
+      "loss": 3.2932,
+      "step": 63350
+    },
+    {
+      "epoch": 6.823807986223227,
+      "grad_norm": 0.7903392910957336,
+      "learning_rate": 0.0001909190819954746,
+      "loss": 3.2974,
+      "step": 63400
+    },
+    {
+      "epoch": 6.829189538262835,
+      "grad_norm": 0.8986805081367493,
+      "learning_rate": 0.0001905958409654132,
+      "loss": 3.2837,
+      "step": 63450
+    },
+    {
+      "epoch": 6.834571090302443,
+      "grad_norm": 0.7686243057250977,
+      "learning_rate": 0.0001902725999353518,
+      "loss": 3.2957,
+      "step": 63500
+    },
+    {
+      "epoch": 6.839952642342052,
+      "grad_norm": 0.7945030927658081,
+      "learning_rate": 0.00018994935890529036,
+      "loss": 3.275,
+      "step": 63550
+    },
+    {
+      "epoch": 6.8453341943816595,
+      "grad_norm": 0.8145069479942322,
+      "learning_rate": 0.00018962611787522892,
+      "loss": 3.287,
+      "step": 63600
+    },
+    {
+      "epoch": 6.850715746421268,
+      "grad_norm": 0.7765173316001892,
+      "learning_rate": 0.00018930287684516754,
+      "loss": 3.2964,
+      "step": 63650
+    },
+    {
+      "epoch": 6.856097298460876,
+      "grad_norm": 0.8863644599914551,
+      "learning_rate": 0.0001889796358151061,
+      "loss": 3.2674,
+      "step": 63700
+    },
+    {
+      "epoch": 6.861478850500484,
+      "grad_norm": 0.7589664459228516,
+      "learning_rate": 0.00018865639478504468,
+      "loss": 3.2718,
+      "step": 63750
+    },
+    {
+      "epoch": 6.866860402540093,
+      "grad_norm": 0.7652623057365417,
+      "learning_rate": 0.0001883331537549833,
+      "loss": 3.2736,
+      "step": 63800
+    },
+    {
+      "epoch": 6.8722419545797,
+      "grad_norm": 0.7880659103393555,
+      "learning_rate": 0.00018800991272492187,
+      "loss": 3.2994,
+      "step": 63850
+    },
+    {
+      "epoch": 6.877623506619309,
+      "grad_norm": 0.7556933760643005,
+      "learning_rate": 0.00018768667169486044,
+      "loss": 3.2908,
+      "step": 63900
+    },
+    {
+      "epoch": 6.8830050586589175,
+      "grad_norm": 0.778171956539154,
+      "learning_rate": 0.00018736343066479903,
+      "loss": 3.2839,
+      "step": 63950
+    },
+    {
+      "epoch": 6.888386610698525,
+      "grad_norm": 0.7967923879623413,
+      "learning_rate": 0.00018704018963473763,
+      "loss": 3.2734,
+      "step": 64000
+    },
+    {
+      "epoch": 6.888386610698525,
+      "eval_accuracy": 0.38728550425968217,
+      "eval_loss": 3.3551573753356934,
+      "eval_runtime": 203.474,
+      "eval_samples_per_second": 88.517,
+      "eval_steps_per_second": 5.534,
+      "step": 64000
+    },
+    {
+      "epoch": 6.893768162738134,
+      "grad_norm": 0.761313796043396,
+      "learning_rate": 0.00018671694860467622,
+      "loss": 3.2713,
+      "step": 64050
+    },
+    {
+      "epoch": 6.899149714777742,
+      "grad_norm": 0.7254140377044678,
+      "learning_rate": 0.0001863937075746148,
+      "loss": 3.2736,
+      "step": 64100
+    },
+    {
+      "epoch": 6.90453126681735,
+      "grad_norm": 0.8009861707687378,
+      "learning_rate": 0.00018607046654455336,
+      "loss": 3.2954,
+      "step": 64150
+    },
+    {
+      "epoch": 6.9099128188569585,
+      "grad_norm": 0.7890373468399048,
+      "learning_rate": 0.00018574722551449198,
+      "loss": 3.2827,
+      "step": 64200
+    },
+    {
+      "epoch": 6.915294370896566,
+      "grad_norm": 0.7626961469650269,
+      "learning_rate": 0.00018542398448443054,
+      "loss": 3.2909,
+      "step": 64250
+    },
+    {
+      "epoch": 6.920675922936175,
+      "grad_norm": 0.8804692029953003,
+      "learning_rate": 0.0001851007434543691,
+      "loss": 3.289,
+      "step": 64300
+    },
+    {
+      "epoch": 6.926057474975783,
+      "grad_norm": 0.7699129581451416,
+      "learning_rate": 0.00018477750242430773,
+      "loss": 3.2924,
+      "step": 64350
+    },
+    {
+      "epoch": 6.931439027015391,
+      "grad_norm": 0.773622989654541,
+      "learning_rate": 0.0001844542613942463,
+      "loss": 3.2788,
+      "step": 64400
+    },
+    {
+      "epoch": 6.9368205790549995,
+      "grad_norm": 0.7274935245513916,
+      "learning_rate": 0.00018413102036418487,
+      "loss": 3.2926,
+      "step": 64450
+    },
+    {
+      "epoch": 6.942202131094608,
+      "grad_norm": 0.7527948021888733,
+      "learning_rate": 0.00018380777933412346,
+      "loss": 3.2984,
+      "step": 64500
+    },
+    {
+      "epoch": 6.947583683134216,
+      "grad_norm": 0.8110983371734619,
+      "learning_rate": 0.00018348453830406206,
+      "loss": 3.2862,
+      "step": 64550
+    },
+    {
+      "epoch": 6.952965235173824,
+      "grad_norm": 0.7983884215354919,
+      "learning_rate": 0.00018316129727400063,
+      "loss": 3.286,
+      "step": 64600
+    },
+    {
+      "epoch": 6.958346787213433,
+      "grad_norm": 0.772709310054779,
+      "learning_rate": 0.00018283805624393922,
+      "loss": 3.3001,
+      "step": 64650
+    },
+    {
+      "epoch": 6.9637283392530405,
+      "grad_norm": 0.7958887815475464,
+      "learning_rate": 0.0001825148152138778,
+      "loss": 3.2976,
+      "step": 64700
+    },
+    {
+      "epoch": 6.969109891292649,
+      "grad_norm": 0.7898955941200256,
+      "learning_rate": 0.00018219157418381636,
+      "loss": 3.2906,
+      "step": 64750
+    },
+    {
+      "epoch": 6.974491443332257,
+      "grad_norm": 0.7751755118370056,
+      "learning_rate": 0.00018186833315375498,
+      "loss": 3.2859,
+      "step": 64800
+    },
+    {
+      "epoch": 6.979872995371865,
+      "grad_norm": 0.7762618660926819,
+      "learning_rate": 0.00018154509212369355,
+      "loss": 3.2973,
+      "step": 64850
+    },
+    {
+      "epoch": 6.985254547411474,
+      "grad_norm": 0.7949537038803101,
+      "learning_rate": 0.0001812218510936321,
+      "loss": 3.2853,
+      "step": 64900
+    },
+    {
+      "epoch": 6.990636099451081,
+      "grad_norm": 0.7892162799835205,
+      "learning_rate": 0.00018089861006357073,
+      "loss": 3.2872,
+      "step": 64950
+    },
+    {
+      "epoch": 6.99601765149069,
+      "grad_norm": 0.7522981762886047,
+      "learning_rate": 0.0001805753690335093,
+      "loss": 3.2676,
+      "step": 65000
+    },
+    {
+      "epoch": 6.99601765149069,
+      "eval_accuracy": 0.3876287387322881,
+      "eval_loss": 3.352322578430176,
+      "eval_runtime": 215.3455,
+      "eval_samples_per_second": 83.638,
+      "eval_steps_per_second": 5.229,
+      "step": 65000
+    },
+    {
+      "epoch": 7.0013992035302985,
+      "grad_norm": 0.7797788381576538,
+      "learning_rate": 0.0001802521280034479,
+      "loss": 3.2643,
+      "step": 65050
+    },
+    {
+      "epoch": 7.006780755569906,
+      "grad_norm": 0.7758247256278992,
+      "learning_rate": 0.00017992888697338646,
+      "loss": 3.2046,
+      "step": 65100
+    },
+    {
+      "epoch": 7.012162307609515,
+      "grad_norm": 0.8069040775299072,
+      "learning_rate": 0.00017960564594332506,
+      "loss": 3.2005,
+      "step": 65150
+    },
+    {
+      "epoch": 7.017543859649122,
+      "grad_norm": 0.7784696817398071,
+      "learning_rate": 0.00017928240491326365,
+      "loss": 3.2059,
+      "step": 65200
+    },
+    {
+      "epoch": 7.022925411688731,
+      "grad_norm": 0.7886298298835754,
+      "learning_rate": 0.00017895916388320222,
+      "loss": 3.2059,
+      "step": 65250
+    },
+    {
+      "epoch": 7.0283069637283395,
+      "grad_norm": 0.810814380645752,
+      "learning_rate": 0.0001786359228531408,
+      "loss": 3.2004,
+      "step": 65300
+    },
+    {
+      "epoch": 7.033688515767947,
+      "grad_norm": 0.8277598023414612,
+      "learning_rate": 0.0001783126818230794,
+      "loss": 3.1962,
+      "step": 65350
+    },
+    {
+      "epoch": 7.039070067807556,
+      "grad_norm": 0.7609949707984924,
+      "learning_rate": 0.00017798944079301798,
+      "loss": 3.2008,
+      "step": 65400
+    },
+    {
+      "epoch": 7.044451619847164,
+      "grad_norm": 0.7364971041679382,
+      "learning_rate": 0.00017766619976295655,
+      "loss": 3.2008,
+      "step": 65450
+    },
+    {
+      "epoch": 7.049833171886772,
+      "grad_norm": 0.7349782586097717,
+      "learning_rate": 0.00017734295873289517,
+      "loss": 3.1923,
+      "step": 65500
+    },
+    {
+      "epoch": 7.0552147239263805,
+      "grad_norm": 0.8025264739990234,
+      "learning_rate": 0.00017702618252343495,
+      "loss": 3.2097,
+      "step": 65550
+    },
+    {
+      "epoch": 7.060596275965988,
+      "grad_norm": 0.7878516316413879,
+      "learning_rate": 0.00017670294149337354,
+      "loss": 3.1973,
+      "step": 65600
+    },
+    {
+      "epoch": 7.065977828005597,
+      "grad_norm": 0.7897598147392273,
+      "learning_rate": 0.00017637970046331214,
+      "loss": 3.196,
+      "step": 65650
+    },
+    {
+      "epoch": 7.071359380045205,
+      "grad_norm": 0.7731457352638245,
+      "learning_rate": 0.0001760564594332507,
+      "loss": 3.2343,
+      "step": 65700
+    },
+    {
+      "epoch": 7.076740932084813,
+      "grad_norm": 1.2706592082977295,
+      "learning_rate": 0.0001757332184031893,
+      "loss": 3.2107,
+      "step": 65750
+    },
+    {
+      "epoch": 7.0821224841244215,
+      "grad_norm": 0.8142618536949158,
+      "learning_rate": 0.00017540997737312787,
+      "loss": 3.2197,
+      "step": 65800
+    },
+    {
+      "epoch": 7.08750403616403,
+      "grad_norm": 0.7961957454681396,
+      "learning_rate": 0.0001750867363430665,
+      "loss": 3.2175,
+      "step": 65850
+    },
+    {
+      "epoch": 7.092885588203638,
+      "grad_norm": 0.8617919683456421,
+      "learning_rate": 0.00017476349531300506,
+      "loss": 3.2236,
+      "step": 65900
+    },
+    {
+      "epoch": 7.098267140243246,
+      "grad_norm": 0.7776252031326294,
+      "learning_rate": 0.00017444025428294362,
+      "loss": 3.2133,
+      "step": 65950
+    },
+    {
+      "epoch": 7.103648692282855,
+      "grad_norm": 0.7629320621490479,
+      "learning_rate": 0.00017411701325288224,
+      "loss": 3.2141,
+      "step": 66000
+    },
+    {
+      "epoch": 7.103648692282855,
+      "eval_accuracy": 0.3874447893932182,
+      "eval_loss": 3.357611894607544,
+      "eval_runtime": 211.7423,
+      "eval_samples_per_second": 85.061,
+      "eval_steps_per_second": 5.318,
+      "step": 66000
+    },
+    {
+      "epoch": 7.109030244322462,
+      "grad_norm": 0.8089470863342285,
+      "learning_rate": 0.0001737937722228208,
+      "loss": 3.2077,
+      "step": 66050
+    },
+    {
+      "epoch": 7.114411796362071,
+      "grad_norm": 0.7836511135101318,
+      "learning_rate": 0.00017347053119275938,
+      "loss": 3.2124,
+      "step": 66100
+    },
+    {
+      "epoch": 7.119793348401679,
+      "grad_norm": 0.7699388861656189,
+      "learning_rate": 0.00017314729016269797,
+      "loss": 3.2208,
+      "step": 66150
+    },
+    {
+      "epoch": 7.125174900441287,
+      "grad_norm": 0.7971923351287842,
+      "learning_rate": 0.00017282404913263654,
+      "loss": 3.2245,
+      "step": 66200
+    },
+    {
+      "epoch": 7.130556452480896,
+      "grad_norm": 0.8028796911239624,
+      "learning_rate": 0.00017250080810257514,
+      "loss": 3.2306,
+      "step": 66250
+    },
+    {
+      "epoch": 7.135938004520503,
+      "grad_norm": 0.7812018394470215,
+      "learning_rate": 0.00017217756707251373,
+      "loss": 3.1989,
+      "step": 66300
+    },
+    {
+      "epoch": 7.141319556560112,
+      "grad_norm": 0.7548637986183167,
+      "learning_rate": 0.0001718543260424523,
+      "loss": 3.211,
+      "step": 66350
+    },
+    {
+      "epoch": 7.1467011085997205,
+      "grad_norm": 0.8434266448020935,
+      "learning_rate": 0.00017153108501239087,
+      "loss": 3.2338,
+      "step": 66400
+    },
+    {
+      "epoch": 7.152082660639328,
+      "grad_norm": 0.8046761155128479,
+      "learning_rate": 0.0001712078439823295,
+      "loss": 3.2238,
+      "step": 66450
+    },
+    {
+      "epoch": 7.157464212678937,
+      "grad_norm": 0.7850152850151062,
+      "learning_rate": 0.00017088460295226806,
+      "loss": 3.2273,
+      "step": 66500
+    },
+    {
+      "epoch": 7.162845764718545,
+      "grad_norm": 0.8294812440872192,
+      "learning_rate": 0.00017056136192220662,
+      "loss": 3.2112,
+      "step": 66550
+    },
+    {
+      "epoch": 7.168227316758153,
+      "grad_norm": 0.744547426700592,
+      "learning_rate": 0.00017023812089214524,
+      "loss": 3.2101,
+      "step": 66600
+    },
+    {
+      "epoch": 7.1736088687977615,
+      "grad_norm": 0.8168179988861084,
+      "learning_rate": 0.0001699148798620838,
+      "loss": 3.2147,
+      "step": 66650
+    },
+    {
+      "epoch": 7.178990420837369,
+      "grad_norm": 0.8229835033416748,
+      "learning_rate": 0.00016959163883202238,
+      "loss": 3.2302,
+      "step": 66700
+    },
+    {
+      "epoch": 7.184371972876978,
+      "grad_norm": 0.8025431632995605,
+      "learning_rate": 0.00016926839780196097,
+      "loss": 3.2207,
+      "step": 66750
+    },
+    {
+      "epoch": 7.189753524916586,
+      "grad_norm": 0.782611608505249,
+      "learning_rate": 0.00016894515677189957,
+      "loss": 3.241,
+      "step": 66800
+    },
+    {
+      "epoch": 7.195135076956194,
+      "grad_norm": 0.8258253335952759,
+      "learning_rate": 0.00016862191574183816,
+      "loss": 3.2092,
+      "step": 66850
+    },
+    {
+      "epoch": 7.2005166289958025,
+      "grad_norm": 0.8076246976852417,
+      "learning_rate": 0.00016829867471177673,
+      "loss": 3.2358,
+      "step": 66900
+    },
+    {
+      "epoch": 7.205898181035411,
+      "grad_norm": 0.8116209506988525,
+      "learning_rate": 0.0001679754336817153,
+      "loss": 3.2076,
+      "step": 66950
+    },
+    {
+      "epoch": 7.211279733075019,
+      "grad_norm": 0.8048698306083679,
+      "learning_rate": 0.00016765219265165392,
+      "loss": 3.2343,
+      "step": 67000
+    },
+    {
+      "epoch": 7.211279733075019,
+      "eval_accuracy": 0.3876256964514293,
+      "eval_loss": 3.3560845851898193,
+      "eval_runtime": 211.2944,
+      "eval_samples_per_second": 85.241,
+      "eval_steps_per_second": 5.329,
+      "step": 67000
+    },
+    {
+      "epoch": 7.216661285114627,
+      "grad_norm": 0.8519846796989441,
+      "learning_rate": 0.0001673289516215925,
+      "loss": 3.2344,
+      "step": 67050
+    },
+    {
+      "epoch": 7.222042837154235,
+      "grad_norm": 0.7423728108406067,
+      "learning_rate": 0.00016700571059153106,
+      "loss": 3.2322,
+      "step": 67100
+    },
+    {
+      "epoch": 7.2274243891938434,
+      "grad_norm": 0.7732436060905457,
+      "learning_rate": 0.00016668246956146968,
+      "loss": 3.2076,
+      "step": 67150
+    },
+    {
+      "epoch": 7.232805941233452,
+      "grad_norm": 0.7401405572891235,
+      "learning_rate": 0.00016635922853140825,
+      "loss": 3.2221,
+      "step": 67200
+    },
+    {
+      "epoch": 7.23818749327306,
+      "grad_norm": 0.7472262382507324,
+      "learning_rate": 0.0001660359875013468,
+      "loss": 3.2246,
+      "step": 67250
+    },
+    {
+      "epoch": 7.243569045312668,
+      "grad_norm": 0.7432873845100403,
+      "learning_rate": 0.0001657127464712854,
+      "loss": 3.209,
+      "step": 67300
+    },
+    {
+      "epoch": 7.248950597352277,
+      "grad_norm": 0.7662221789360046,
+      "learning_rate": 0.000165389505441224,
+      "loss": 3.2232,
+      "step": 67350
+    },
+    {
+      "epoch": 7.254332149391884,
+      "grad_norm": 0.804426372051239,
+      "learning_rate": 0.00016506626441116257,
+      "loss": 3.2239,
+      "step": 67400
+    },
+    {
+      "epoch": 7.259713701431493,
+      "grad_norm": 0.7635779976844788,
+      "learning_rate": 0.00016474302338110116,
+      "loss": 3.2356,
+      "step": 67450
+    },
+    {
+      "epoch": 7.265095253471101,
+      "grad_norm": 0.7849301695823669,
+      "learning_rate": 0.00016441978235103973,
+      "loss": 3.2353,
+      "step": 67500
+    },
+    {
+      "epoch": 7.270476805510709,
+      "grad_norm": 0.7454454302787781,
+      "learning_rate": 0.00016409654132097833,
+      "loss": 3.2238,
+      "step": 67550
+    },
+    {
+      "epoch": 7.275858357550318,
+      "grad_norm": 0.8164418339729309,
+      "learning_rate": 0.00016377330029091692,
+      "loss": 3.2453,
+      "step": 67600
+    },
+    {
+      "epoch": 7.281239909589925,
+      "grad_norm": 0.7791739702224731,
+      "learning_rate": 0.0001634500592608555,
+      "loss": 3.2339,
+      "step": 67650
+    },
+    {
+      "epoch": 7.286621461629534,
+      "grad_norm": 0.7832018733024597,
+      "learning_rate": 0.00016312681823079406,
+      "loss": 3.2339,
+      "step": 67700
+    },
+    {
+      "epoch": 7.2920030136691425,
+      "grad_norm": 0.8332245945930481,
+      "learning_rate": 0.00016280357720073268,
+      "loss": 3.2326,
+      "step": 67750
+    },
+    {
+      "epoch": 7.29738456570875,
+      "grad_norm": 0.7885503172874451,
+      "learning_rate": 0.00016248033617067125,
+      "loss": 3.2424,
+      "step": 67800
+    },
+    {
+      "epoch": 7.302766117748359,
+      "grad_norm": 0.7824217677116394,
+      "learning_rate": 0.00016215709514060984,
+      "loss": 3.218,
+      "step": 67850
+    },
+    {
+      "epoch": 7.308147669787967,
+      "grad_norm": 0.8223897814750671,
+      "learning_rate": 0.00016183385411054843,
+      "loss": 3.2276,
+      "step": 67900
+    },
+    {
+      "epoch": 7.313529221827575,
+      "grad_norm": 0.7712852358818054,
+      "learning_rate": 0.000161510613080487,
+      "loss": 3.2069,
+      "step": 67950
+    },
+    {
+      "epoch": 7.3189107738671835,
+      "grad_norm": 0.7743715643882751,
+      "learning_rate": 0.0001611873720504256,
+      "loss": 3.2277,
+      "step": 68000
+    },
+    {
+      "epoch": 7.3189107738671835,
+      "eval_accuracy": 0.3880603080026872,
+      "eval_loss": 3.3522820472717285,
+      "eval_runtime": 210.2908,
+      "eval_samples_per_second": 85.648,
+      "eval_steps_per_second": 5.354,
+      "step": 68000
+    },
+    {
+      "epoch": 7.324292325906791,
+      "grad_norm": 0.7747835516929626,
+      "learning_rate": 0.00016086413102036416,
+      "loss": 3.2336,
+      "step": 68050
+    },
+    {
+      "epoch": 7.3296738779464,
+      "grad_norm": 0.8009321093559265,
+      "learning_rate": 0.00016054088999030273,
+      "loss": 3.2293,
+      "step": 68100
+    },
+    {
+      "epoch": 7.335055429986008,
+      "grad_norm": 0.8049393892288208,
+      "learning_rate": 0.00016021764896024135,
+      "loss": 3.2359,
+      "step": 68150
+    },
+    {
+      "epoch": 7.340436982025616,
+      "grad_norm": 0.8045018315315247,
+      "learning_rate": 0.00015990087275078113,
+      "loss": 3.2419,
+      "step": 68200
+    },
+    {
+      "epoch": 7.3458185340652244,
+      "grad_norm": 0.8061736226081848,
+      "learning_rate": 0.00015957763172071976,
+      "loss": 3.217,
+      "step": 68250
+    },
+    {
+      "epoch": 7.351200086104833,
+      "grad_norm": 0.8088467717170715,
+      "learning_rate": 0.00015925439069065832,
+      "loss": 3.2275,
+      "step": 68300
+    },
+    {
+      "epoch": 7.356581638144441,
+      "grad_norm": 0.8118233680725098,
+      "learning_rate": 0.0001589311496605969,
+      "loss": 3.2308,
+      "step": 68350
+    },
+    {
+      "epoch": 7.361963190184049,
+      "grad_norm": 0.7720018625259399,
+      "learning_rate": 0.00015860790863053548,
+      "loss": 3.231,
+      "step": 68400
+    },
+    {
+      "epoch": 7.367344742223658,
+      "grad_norm": 0.7653403878211975,
+      "learning_rate": 0.00015828466760047408,
+      "loss": 3.2414,
+      "step": 68450
+    },
+    {
+      "epoch": 7.372726294263265,
+      "grad_norm": 0.7920792102813721,
+      "learning_rate": 0.00015796142657041265,
+      "loss": 3.2353,
+      "step": 68500
+    },
+    {
+      "epoch": 7.378107846302874,
+      "grad_norm": 0.8059729337692261,
+      "learning_rate": 0.00015763818554035124,
+      "loss": 3.2339,
+      "step": 68550
+    },
+    {
+      "epoch": 7.383489398342482,
+      "grad_norm": 0.7684776186943054,
+      "learning_rate": 0.0001573149445102898,
+      "loss": 3.2355,
+      "step": 68600
+    },
+    {
+      "epoch": 7.38887095038209,
+      "grad_norm": 0.8400054574012756,
+      "learning_rate": 0.00015699170348022843,
+      "loss": 3.2392,
+      "step": 68650
+    },
+    {
+      "epoch": 7.394252502421699,
+      "grad_norm": 0.7758475542068481,
+      "learning_rate": 0.000156668462450167,
+      "loss": 3.2384,
+      "step": 68700
+    },
+    {
+      "epoch": 7.399634054461306,
+      "grad_norm": 0.8178544640541077,
+      "learning_rate": 0.00015634522142010557,
+      "loss": 3.2373,
+      "step": 68750
+    },
+    {
+      "epoch": 7.405015606500915,
+      "grad_norm": 0.7898211479187012,
+      "learning_rate": 0.0001560219803900442,
+      "loss": 3.2331,
+      "step": 68800
+    },
+    {
+      "epoch": 7.4103971585405235,
+      "grad_norm": 0.7738881707191467,
+      "learning_rate": 0.00015569873935998276,
+      "loss": 3.2337,
+      "step": 68850
+    },
+    {
+      "epoch": 7.415778710580131,
+      "grad_norm": 0.7856907844543457,
+      "learning_rate": 0.00015537549832992132,
+      "loss": 3.2221,
+      "step": 68900
+    },
+    {
+      "epoch": 7.42116026261974,
+      "grad_norm": 0.7582386136054993,
+      "learning_rate": 0.00015505225729985992,
+      "loss": 3.2397,
+      "step": 68950
+    },
+    {
+      "epoch": 7.426541814659347,
+      "grad_norm": 0.8215800523757935,
+      "learning_rate": 0.0001547290162697985,
+      "loss": 3.236,
+      "step": 69000
+    },
+    {
+      "epoch": 7.426541814659347,
+      "eval_accuracy": 0.3885883610374656,
+      "eval_loss": 3.349855661392212,
+      "eval_runtime": 210.9678,
+      "eval_samples_per_second": 85.373,
+      "eval_steps_per_second": 5.337,
+      "step": 69000
+    },
+    {
+      "epoch": 7.431923366698956,
+      "grad_norm": 0.8055285811424255,
+      "learning_rate": 0.00015440577523973708,
+      "loss": 3.2277,
+      "step": 69050
+    },
+    {
+      "epoch": 7.4373049187385645,
+      "grad_norm": 0.7805681824684143,
+      "learning_rate": 0.00015408253420967567,
+      "loss": 3.2534,
+      "step": 69100
+    },
+    {
+      "epoch": 7.442686470778172,
+      "grad_norm": 0.8025108575820923,
+      "learning_rate": 0.00015375929317961424,
+      "loss": 3.2343,
+      "step": 69150
+    },
+    {
+      "epoch": 7.448068022817781,
+      "grad_norm": 0.8178415894508362,
+      "learning_rate": 0.00015343605214955284,
+      "loss": 3.2282,
+      "step": 69200
+    },
+    {
+      "epoch": 7.453449574857389,
+      "grad_norm": 0.7723139524459839,
+      "learning_rate": 0.00015311281111949143,
+      "loss": 3.2373,
+      "step": 69250
+    },
+    {
+      "epoch": 7.458831126896997,
+      "grad_norm": 0.8103027939796448,
+      "learning_rate": 0.00015278957008943,
+      "loss": 3.2253,
+      "step": 69300
+    },
+    {
+      "epoch": 7.4642126789366054,
+      "grad_norm": 0.7889986038208008,
+      "learning_rate": 0.00015246632905936857,
+      "loss": 3.2271,
+      "step": 69350
+    },
+    {
+      "epoch": 7.469594230976213,
+      "grad_norm": 0.7937176823616028,
+      "learning_rate": 0.0001521430880293072,
+      "loss": 3.2375,
+      "step": 69400
+    },
+    {
+      "epoch": 7.474975783015822,
+      "grad_norm": 0.8158497214317322,
+      "learning_rate": 0.00015181984699924576,
+      "loss": 3.2368,
+      "step": 69450
+    },
+    {
+      "epoch": 7.48035733505543,
+      "grad_norm": 0.8363879323005676,
+      "learning_rate": 0.00015149660596918432,
+      "loss": 3.2239,
+      "step": 69500
+    },
+    {
+      "epoch": 7.485738887095038,
+      "grad_norm": 0.7895720601081848,
+      "learning_rate": 0.00015117336493912295,
+      "loss": 3.2266,
+      "step": 69550
+    },
+    {
+      "epoch": 7.491120439134646,
+      "grad_norm": 0.7915844917297363,
+      "learning_rate": 0.0001508501239090615,
+      "loss": 3.2448,
+      "step": 69600
+    },
+    {
+      "epoch": 7.496501991174255,
+      "grad_norm": 0.7852685451507568,
+      "learning_rate": 0.0001505268828790001,
+      "loss": 3.2139,
+      "step": 69650
+    },
+    {
+      "epoch": 7.501883543213863,
+      "grad_norm": 0.7899736166000366,
+      "learning_rate": 0.00015020364184893867,
+      "loss": 3.2317,
+      "step": 69700
+    },
+    {
+      "epoch": 7.507265095253471,
+      "grad_norm": 0.7891963124275208,
+      "learning_rate": 0.00014988040081887724,
+      "loss": 3.2278,
+      "step": 69750
+    },
+    {
+      "epoch": 7.51264664729308,
+      "grad_norm": 0.7811875343322754,
+      "learning_rate": 0.00014955715978881584,
+      "loss": 3.2406,
+      "step": 69800
+    },
+    {
+      "epoch": 7.518028199332687,
+      "grad_norm": 0.7988965511322021,
+      "learning_rate": 0.00014923391875875443,
+      "loss": 3.2298,
+      "step": 69850
+    },
+    {
+      "epoch": 7.523409751372296,
+      "grad_norm": 0.8040280342102051,
+      "learning_rate": 0.00014891067772869303,
+      "loss": 3.2465,
+      "step": 69900
+    },
+    {
+      "epoch": 7.528791303411904,
+      "grad_norm": 0.8975268602371216,
+      "learning_rate": 0.0001485874366986316,
+      "loss": 3.2452,
+      "step": 69950
+    },
+    {
+      "epoch": 7.534172855451512,
+      "grad_norm": 0.8084391951560974,
+      "learning_rate": 0.0001482641956685702,
+      "loss": 3.2201,
+      "step": 70000
+    },
+    {
+      "epoch": 7.534172855451512,
+      "eval_accuracy": 0.38874221352661087,
+      "eval_loss": 3.3477556705474854,
+      "eval_runtime": 211.1497,
+      "eval_samples_per_second": 85.3,
+      "eval_steps_per_second": 5.333,
+      "step": 70000
+    },
+    {
+      "epoch": 7.539554407491121,
+      "grad_norm": 0.8021398186683655,
+      "learning_rate": 0.00014794095463850878,
+      "loss": 3.2476,
+      "step": 70050
+    },
+    {
+      "epoch": 7.544935959530728,
+      "grad_norm": 0.796826183795929,
+      "learning_rate": 0.00014761771360844735,
+      "loss": 3.2431,
+      "step": 70100
+    },
+    {
+      "epoch": 7.550317511570337,
+      "grad_norm": 0.8226954340934753,
+      "learning_rate": 0.00014729447257838595,
+      "loss": 3.2146,
+      "step": 70150
+    },
+    {
+      "epoch": 7.5556990636099455,
+      "grad_norm": 0.7751065492630005,
+      "learning_rate": 0.0001469712315483245,
+      "loss": 3.2377,
+      "step": 70200
+    },
+    {
+      "epoch": 7.561080615649553,
+      "grad_norm": 0.8223302960395813,
+      "learning_rate": 0.0001466479905182631,
+      "loss": 3.2459,
+      "step": 70250
+    },
+    {
+      "epoch": 7.566462167689162,
+      "grad_norm": 0.7794532179832458,
+      "learning_rate": 0.00014632474948820168,
+      "loss": 3.2225,
+      "step": 70300
+    },
+    {
+      "epoch": 7.57184371972877,
+      "grad_norm": 0.8701120615005493,
+      "learning_rate": 0.00014600150845814027,
+      "loss": 3.2331,
+      "step": 70350
+    },
+    {
+      "epoch": 7.577225271768378,
+      "grad_norm": 0.8439012169837952,
+      "learning_rate": 0.00014567826742807886,
+      "loss": 3.2266,
+      "step": 70400
+    },
+    {
+      "epoch": 7.5826068238079865,
+      "grad_norm": 0.82621169090271,
+      "learning_rate": 0.00014535502639801743,
+      "loss": 3.2267,
+      "step": 70450
+    },
+    {
+      "epoch": 7.587988375847594,
+      "grad_norm": 0.8370195627212524,
+      "learning_rate": 0.00014503178536795603,
+      "loss": 3.2329,
+      "step": 70500
+    },
+    {
+      "epoch": 7.593369927887203,
+      "grad_norm": 0.8157381415367126,
+      "learning_rate": 0.00014470854433789462,
+      "loss": 3.2428,
+      "step": 70550
+    },
+    {
+      "epoch": 7.598751479926811,
+      "grad_norm": 0.8180111050605774,
+      "learning_rate": 0.0001443853033078332,
+      "loss": 3.2468,
+      "step": 70600
+    },
+    {
+      "epoch": 7.604133031966419,
+      "grad_norm": 0.8646537065505981,
+      "learning_rate": 0.00014406206227777178,
+      "loss": 3.22,
+      "step": 70650
+    },
+    {
+      "epoch": 7.609514584006027,
+      "grad_norm": 0.8048348426818848,
+      "learning_rate": 0.0001437452860683116,
+      "loss": 3.2472,
+      "step": 70700
+    },
+    {
+      "epoch": 7.614896136045635,
+      "grad_norm": 0.7684217691421509,
+      "learning_rate": 0.00014342204503825018,
+      "loss": 3.2327,
+      "step": 70750
+    },
+    {
+      "epoch": 7.620277688085244,
+      "grad_norm": 0.8237805962562561,
+      "learning_rate": 0.00014309880400818875,
+      "loss": 3.2486,
+      "step": 70800
+    },
+    {
+      "epoch": 7.625659240124852,
+      "grad_norm": 0.8122904896736145,
+      "learning_rate": 0.00014277556297812735,
+      "loss": 3.2419,
+      "step": 70850
+    },
+    {
+      "epoch": 7.63104079216446,
+      "grad_norm": 0.8078117966651917,
+      "learning_rate": 0.00014245232194806591,
+      "loss": 3.2286,
+      "step": 70900
+    },
+    {
+      "epoch": 7.636422344204068,
+      "grad_norm": 0.8151229619979858,
+      "learning_rate": 0.0001421290809180045,
+      "loss": 3.2208,
+      "step": 70950
+    },
+    {
+      "epoch": 7.641803896243677,
+      "grad_norm": 0.8114475011825562,
+      "learning_rate": 0.0001418058398879431,
+      "loss": 3.2337,
+      "step": 71000
+    },
+    {
+      "epoch": 7.641803896243677,
+      "eval_accuracy": 0.38909826903997896,
+      "eval_loss": 3.3436226844787598,
+      "eval_runtime": 207.9913,
+      "eval_samples_per_second": 86.595,
+      "eval_steps_per_second": 5.414,
+      "step": 71000
+    },
+    {
+      "epoch": 7.647185448283285,
+      "grad_norm": 0.8878672122955322,
+      "learning_rate": 0.00014148259885788167,
+      "loss": 3.2354,
+      "step": 71050
+    },
+    {
+      "epoch": 7.652567000322893,
+      "grad_norm": 0.8463232517242432,
+      "learning_rate": 0.00014115935782782027,
+      "loss": 3.2492,
+      "step": 71100
+    },
+    {
+      "epoch": 7.657948552362502,
+      "grad_norm": 0.8358443975448608,
+      "learning_rate": 0.00014083611679775886,
+      "loss": 3.2434,
+      "step": 71150
+    },
+    {
+      "epoch": 7.663330104402109,
+      "grad_norm": 0.8259372115135193,
+      "learning_rate": 0.00014051287576769743,
+      "loss": 3.2452,
+      "step": 71200
+    },
+    {
+      "epoch": 7.668711656441718,
+      "grad_norm": 0.8233510851860046,
+      "learning_rate": 0.00014018963473763602,
+      "loss": 3.2524,
+      "step": 71250
+    },
+    {
+      "epoch": 7.674093208481326,
+      "grad_norm": 0.7944367527961731,
+      "learning_rate": 0.0001398663937075746,
+      "loss": 3.2322,
+      "step": 71300
+    },
+    {
+      "epoch": 7.679474760520934,
+      "grad_norm": 0.818215012550354,
+      "learning_rate": 0.00013954315267751319,
+      "loss": 3.242,
+      "step": 71350
+    },
+    {
+      "epoch": 7.684856312560543,
+      "grad_norm": 0.8027128577232361,
+      "learning_rate": 0.00013921991164745175,
+      "loss": 3.2194,
+      "step": 71400
+    },
+    {
+      "epoch": 7.69023786460015,
+      "grad_norm": 0.7849347591400146,
+      "learning_rate": 0.00013889667061739035,
+      "loss": 3.2435,
+      "step": 71450
+    },
+    {
+      "epoch": 7.695619416639759,
+      "grad_norm": 0.8004842400550842,
+      "learning_rate": 0.00013857342958732894,
+      "loss": 3.2417,
+      "step": 71500
+    },
+    {
+      "epoch": 7.7010009686793675,
+      "grad_norm": 0.7999834418296814,
+      "learning_rate": 0.0001382501885572675,
+      "loss": 3.2345,
+      "step": 71550
+    },
+    {
+      "epoch": 7.706382520718975,
+      "grad_norm": 0.813932478427887,
+      "learning_rate": 0.0001379269475272061,
+      "loss": 3.2388,
+      "step": 71600
+    },
+    {
+      "epoch": 7.711764072758584,
+      "grad_norm": 0.8163520693778992,
+      "learning_rate": 0.0001376037064971447,
+      "loss": 3.2463,
+      "step": 71650
+    },
+    {
+      "epoch": 7.717145624798192,
+      "grad_norm": 0.8218525648117065,
+      "learning_rate": 0.0001372804654670833,
+      "loss": 3.2376,
+      "step": 71700
+    },
+    {
+      "epoch": 7.7225271768378,
+      "grad_norm": 0.8164900541305542,
+      "learning_rate": 0.00013695722443702186,
+      "loss": 3.2415,
+      "step": 71750
+    },
+    {
+      "epoch": 7.727908728877408,
+      "grad_norm": 0.8270570039749146,
+      "learning_rate": 0.00013663398340696046,
+      "loss": 3.2334,
+      "step": 71800
+    },
+    {
+      "epoch": 7.733290280917016,
+      "grad_norm": 0.8093637824058533,
+      "learning_rate": 0.00013631074237689902,
+      "loss": 3.2408,
+      "step": 71850
+    },
+    {
+      "epoch": 7.738671832956625,
+      "grad_norm": 0.8747385740280151,
+      "learning_rate": 0.00013598750134683762,
+      "loss": 3.2449,
+      "step": 71900
+    },
+    {
+      "epoch": 7.744053384996233,
+      "grad_norm": 0.7810999751091003,
+      "learning_rate": 0.00013566426031677619,
+      "loss": 3.2159,
+      "step": 71950
+    },
+    {
+      "epoch": 7.749434937035841,
+      "grad_norm": 0.8121122717857361,
+      "learning_rate": 0.00013534748410731602,
+      "loss": 3.2174,
+      "step": 72000
+    },
+    {
+      "epoch": 7.749434937035841,
+      "eval_accuracy": 0.3896095895300339,
+      "eval_loss": 3.3376104831695557,
+      "eval_runtime": 189.9062,
+      "eval_samples_per_second": 94.842,
+      "eval_steps_per_second": 5.929,
+      "step": 72000
+    },
+    {
+      "epoch": 7.754816489075449,
+      "grad_norm": 0.8262868523597717,
+      "learning_rate": 0.0001350242430772546,
+      "loss": 3.2463,
+      "step": 72050
+    },
+    {
+      "epoch": 7.760198041115058,
+      "grad_norm": 0.8187814354896545,
+      "learning_rate": 0.00013470100204719318,
+      "loss": 3.2571,
+      "step": 72100
+    },
+    {
+      "epoch": 7.765579593154666,
+      "grad_norm": 0.8290274739265442,
+      "learning_rate": 0.00013437776101713178,
+      "loss": 3.2556,
+      "step": 72150
+    },
+    {
+      "epoch": 7.770961145194274,
+      "grad_norm": 0.8006277680397034,
+      "learning_rate": 0.00013405451998707034,
+      "loss": 3.2402,
+      "step": 72200
+    },
+    {
+      "epoch": 7.776342697233883,
+      "grad_norm": 0.8297838568687439,
+      "learning_rate": 0.00013373127895700894,
+      "loss": 3.2314,
+      "step": 72250
+    },
+    {
+      "epoch": 7.78172424927349,
+      "grad_norm": 0.8442043662071228,
+      "learning_rate": 0.00013340803792694753,
+      "loss": 3.2356,
+      "step": 72300
+    },
+    {
+      "epoch": 7.787105801313099,
+      "grad_norm": 0.8034444451332092,
+      "learning_rate": 0.0001330847968968861,
+      "loss": 3.2229,
+      "step": 72350
+    },
+    {
+      "epoch": 7.792487353352707,
+      "grad_norm": 0.8173365592956543,
+      "learning_rate": 0.00013276155586682467,
+      "loss": 3.2286,
+      "step": 72400
+    },
+    {
+      "epoch": 7.797868905392315,
+      "grad_norm": 0.777065098285675,
+      "learning_rate": 0.00013243831483676326,
+      "loss": 3.2472,
+      "step": 72450
+    },
+    {
+      "epoch": 7.803250457431924,
+      "grad_norm": 0.8149591088294983,
+      "learning_rate": 0.00013211507380670186,
+      "loss": 3.2358,
+      "step": 72500
+    },
+    {
+      "epoch": 7.808632009471531,
+      "grad_norm": 0.8245164752006531,
+      "learning_rate": 0.00013179183277664042,
+      "loss": 3.2379,
+      "step": 72550
+    },
+    {
+      "epoch": 7.81401356151114,
+      "grad_norm": 0.858371376991272,
+      "learning_rate": 0.00013146859174657902,
+      "loss": 3.2399,
+      "step": 72600
+    },
+    {
+      "epoch": 7.819395113550748,
+      "grad_norm": 0.8818589448928833,
+      "learning_rate": 0.00013114535071651761,
+      "loss": 3.2313,
+      "step": 72650
+    },
+    {
+      "epoch": 7.824776665590356,
+      "grad_norm": 0.7951632738113403,
+      "learning_rate": 0.00013082210968645618,
+      "loss": 3.2385,
+      "step": 72700
+    },
+    {
+      "epoch": 7.830158217629965,
+      "grad_norm": 0.8036244511604309,
+      "learning_rate": 0.00013049886865639478,
+      "loss": 3.2455,
+      "step": 72750
+    },
+    {
+      "epoch": 7.835539769669572,
+      "grad_norm": 0.7847592830657959,
+      "learning_rate": 0.00013017562762633337,
+      "loss": 3.2634,
+      "step": 72800
+    },
+    {
+      "epoch": 7.840921321709181,
+      "grad_norm": 0.8040270805358887,
+      "learning_rate": 0.00012985238659627194,
+      "loss": 3.2263,
+      "step": 72850
+    },
+    {
+      "epoch": 7.846302873748789,
+      "grad_norm": 0.8288953900337219,
+      "learning_rate": 0.00012952914556621053,
+      "loss": 3.252,
+      "step": 72900
+    },
+    {
+      "epoch": 7.851684425788397,
+      "grad_norm": 0.8158117532730103,
+      "learning_rate": 0.0001292059045361491,
+      "loss": 3.2296,
+      "step": 72950
+    },
+    {
+      "epoch": 7.857065977828006,
+      "grad_norm": 0.8312733173370361,
+      "learning_rate": 0.0001288826635060877,
+      "loss": 3.2182,
+      "step": 73000
+    },
+    {
+      "epoch": 7.857065977828006,
+      "eval_accuracy": 0.38989512931921033,
+      "eval_loss": 3.334813117980957,
+      "eval_runtime": 209.8317,
+      "eval_samples_per_second": 85.835,
+      "eval_steps_per_second": 5.366,
+      "step": 73000
+    },
+    {
+      "epoch": 7.862447529867614,
+      "grad_norm": 0.8196859359741211,
+      "learning_rate": 0.00012855942247602626,
+      "loss": 3.2287,
+      "step": 73050
+    },
+    {
+      "epoch": 7.867829081907222,
+      "grad_norm": 0.810441792011261,
+      "learning_rate": 0.00012823618144596486,
+      "loss": 3.2479,
+      "step": 73100
+    },
+    {
+      "epoch": 7.87321063394683,
+      "grad_norm": 0.7915072441101074,
+      "learning_rate": 0.00012791294041590345,
+      "loss": 3.2385,
+      "step": 73150
+    },
+    {
+      "epoch": 7.878592185986438,
+      "grad_norm": 0.8137247562408447,
+      "learning_rate": 0.00012758969938584202,
+      "loss": 3.2249,
+      "step": 73200
+    },
+    {
+      "epoch": 7.883973738026047,
+      "grad_norm": 0.7836380004882812,
+      "learning_rate": 0.00012726645835578061,
+      "loss": 3.2342,
+      "step": 73250
+    },
+    {
+      "epoch": 7.889355290065655,
+      "grad_norm": 0.8190814852714539,
+      "learning_rate": 0.0001269432173257192,
+      "loss": 3.223,
+      "step": 73300
+    },
+    {
+      "epoch": 7.894736842105263,
+      "grad_norm": 0.8309457302093506,
+      "learning_rate": 0.00012661997629565778,
+      "loss": 3.2194,
+      "step": 73350
+    },
+    {
+      "epoch": 7.900118394144871,
+      "grad_norm": 0.7745057344436646,
+      "learning_rate": 0.00012629673526559637,
+      "loss": 3.2503,
+      "step": 73400
+    },
+    {
+      "epoch": 7.90549994618448,
+      "grad_norm": 0.8173632621765137,
+      "learning_rate": 0.00012597349423553497,
+      "loss": 3.2438,
+      "step": 73450
+    },
+    {
+      "epoch": 7.910881498224088,
+      "grad_norm": 0.8117119669914246,
+      "learning_rate": 0.00012565025320547353,
+      "loss": 3.2514,
+      "step": 73500
+    },
+    {
+      "epoch": 7.916263050263696,
+      "grad_norm": 0.7812601923942566,
+      "learning_rate": 0.00012532701217541213,
+      "loss": 3.2333,
+      "step": 73550
+    },
+    {
+      "epoch": 7.921644602303305,
+      "grad_norm": 0.882896363735199,
+      "learning_rate": 0.0001250037711453507,
+      "loss": 3.223,
+      "step": 73600
+    },
+    {
+      "epoch": 7.927026154342912,
+      "grad_norm": 0.7868736982345581,
+      "learning_rate": 0.0001246805301152893,
+      "loss": 3.2356,
+      "step": 73650
+    },
+    {
+      "epoch": 7.932407706382521,
+      "grad_norm": 0.8104776740074158,
+      "learning_rate": 0.00012435728908522786,
+      "loss": 3.2451,
+      "step": 73700
+    },
+    {
+      "epoch": 7.937789258422129,
+      "grad_norm": 0.7862167954444885,
+      "learning_rate": 0.00012403404805516645,
+      "loss": 3.2351,
+      "step": 73750
+    },
+    {
+      "epoch": 7.943170810461737,
+      "grad_norm": 0.832856297492981,
+      "learning_rate": 0.00012371080702510505,
+      "loss": 3.2329,
+      "step": 73800
+    },
+    {
+      "epoch": 7.948552362501346,
+      "grad_norm": 0.8277978301048279,
+      "learning_rate": 0.00012338756599504362,
+      "loss": 3.2403,
+      "step": 73850
+    },
+    {
+      "epoch": 7.953933914540953,
+      "grad_norm": 0.8304550051689148,
+      "learning_rate": 0.0001230643249649822,
+      "loss": 3.2312,
+      "step": 73900
+    },
+    {
+      "epoch": 7.959315466580562,
+      "grad_norm": 0.844760000705719,
+      "learning_rate": 0.0001227410839349208,
+      "loss": 3.2351,
+      "step": 73950
+    },
+    {
+      "epoch": 7.96469701862017,
+      "grad_norm": 0.8154044151306152,
+      "learning_rate": 0.0001224178429048594,
+      "loss": 3.2393,
+      "step": 74000
+    },
+    {
+      "epoch": 7.96469701862017,
+      "eval_accuracy": 0.3899715122993439,
+      "eval_loss": 3.331357002258301,
+      "eval_runtime": 211.3692,
+      "eval_samples_per_second": 85.211,
+      "eval_steps_per_second": 5.327,
+      "step": 74000
+    },
+    {
+      "epoch": 7.970078570659778,
+      "grad_norm": 0.8356589078903198,
+      "learning_rate": 0.00012209460187479797,
+      "loss": 3.2136,
+      "step": 74050
+    },
+    {
+      "epoch": 7.975460122699387,
+      "grad_norm": 0.8147774338722229,
+      "learning_rate": 0.00012177136084473655,
+      "loss": 3.2281,
+      "step": 74100
+    },
+    {
+      "epoch": 7.980841674738995,
+      "grad_norm": 0.8949999809265137,
+      "learning_rate": 0.00012144811981467514,
+      "loss": 3.2365,
+      "step": 74150
+    },
+    {
+      "epoch": 7.986223226778603,
+      "grad_norm": 0.7900296449661255,
+      "learning_rate": 0.00012112487878461371,
+      "loss": 3.2091,
+      "step": 74200
+    },
+    {
+      "epoch": 7.991604778818211,
+      "grad_norm": 0.844578742980957,
+      "learning_rate": 0.0001208016377545523,
+      "loss": 3.247,
+      "step": 74250
+    },
+    {
+      "epoch": 7.996986330857819,
+      "grad_norm": 0.7930771708488464,
+      "learning_rate": 0.00012047839672449089,
+      "loss": 3.2137,
+      "step": 74300
+    },
+    {
+      "epoch": 8.002367882897428,
+      "grad_norm": 0.8258639574050903,
+      "learning_rate": 0.00012015515569442947,
+      "loss": 3.1959,
+      "step": 74350
+    },
+    {
+      "epoch": 8.007749434937036,
+      "grad_norm": 0.8070922493934631,
+      "learning_rate": 0.00011983191466436805,
+      "loss": 3.1728,
+      "step": 74400
+    },
+    {
+      "epoch": 8.013130986976645,
+      "grad_norm": 0.8373987078666687,
+      "learning_rate": 0.00011950867363430664,
+      "loss": 3.1395,
+      "step": 74450
+    },
+    {
+      "epoch": 8.018512539016251,
+      "grad_norm": 0.8283763527870178,
+      "learning_rate": 0.00011918543260424522,
+      "loss": 3.1552,
+      "step": 74500
+    },
+    {
+      "epoch": 8.02389409105586,
+      "grad_norm": 0.8418523669242859,
+      "learning_rate": 0.0001188621915741838,
+      "loss": 3.1531,
+      "step": 74550
+    },
+    {
+      "epoch": 8.029275643095469,
+      "grad_norm": 0.7724992632865906,
+      "learning_rate": 0.00011853895054412239,
+      "loss": 3.1875,
+      "step": 74600
+    },
+    {
+      "epoch": 8.034657195135077,
+      "grad_norm": 0.8510676622390747,
+      "learning_rate": 0.00011821570951406098,
+      "loss": 3.1649,
+      "step": 74650
+    },
+    {
+      "epoch": 8.040038747174686,
+      "grad_norm": 0.8318743109703064,
+      "learning_rate": 0.00011789246848399955,
+      "loss": 3.164,
+      "step": 74700
+    },
+    {
+      "epoch": 8.045420299214294,
+      "grad_norm": 0.8723147511482239,
+      "learning_rate": 0.00011756922745393814,
+      "loss": 3.1604,
+      "step": 74750
+    },
+    {
+      "epoch": 8.050801851253901,
+      "grad_norm": 0.8099079132080078,
+      "learning_rate": 0.00011724598642387674,
+      "loss": 3.1651,
+      "step": 74800
+    },
+    {
+      "epoch": 8.05618340329351,
+      "grad_norm": 0.8209378123283386,
+      "learning_rate": 0.0001169227453938153,
+      "loss": 3.1699,
+      "step": 74850
+    },
+    {
+      "epoch": 8.061564955333118,
+      "grad_norm": 0.8424363136291504,
+      "learning_rate": 0.0001165995043637539,
+      "loss": 3.1525,
+      "step": 74900
+    },
+    {
+      "epoch": 8.066946507372727,
+      "grad_norm": 0.822147786617279,
+      "learning_rate": 0.00011627626333369248,
+      "loss": 3.1864,
+      "step": 74950
+    },
+    {
+      "epoch": 8.072328059412335,
+      "grad_norm": 0.8202683329582214,
+      "learning_rate": 0.00011595302230363108,
+      "loss": 3.1517,
+      "step": 75000
+    },
+    {
+      "epoch": 8.072328059412335,
+      "eval_accuracy": 0.3900216012806264,
+      "eval_loss": 3.336542844772339,
+      "eval_runtime": 191.5515,
+      "eval_samples_per_second": 94.027,
+      "eval_steps_per_second": 5.878,
+      "step": 75000
+    },
+    {
+      "epoch": 8.077709611451942,
+      "grad_norm": 0.8051620721817017,
+      "learning_rate": 0.00011562978127356964,
+      "loss": 3.1685,
+      "step": 75050
+    },
+    {
+      "epoch": 8.08309116349155,
+      "grad_norm": 0.9138336777687073,
+      "learning_rate": 0.00011530654024350824,
+      "loss": 3.1629,
+      "step": 75100
+    },
+    {
+      "epoch": 8.088472715531159,
+      "grad_norm": 0.8304498195648193,
+      "learning_rate": 0.00011498329921344682,
+      "loss": 3.1683,
+      "step": 75150
+    },
+    {
+      "epoch": 8.093854267570768,
+      "grad_norm": 0.8430973887443542,
+      "learning_rate": 0.0001146600581833854,
+      "loss": 3.1618,
+      "step": 75200
+    },
+    {
+      "epoch": 8.099235819610376,
+      "grad_norm": 0.8061507344245911,
+      "learning_rate": 0.00011433681715332398,
+      "loss": 3.1781,
+      "step": 75250
+    },
+    {
+      "epoch": 8.104617371649983,
+      "grad_norm": 0.8056333065032959,
+      "learning_rate": 0.00011401357612326258,
+      "loss": 3.1757,
+      "step": 75300
+    },
+    {
+      "epoch": 8.109998923689592,
+      "grad_norm": 0.8301107883453369,
+      "learning_rate": 0.00011369033509320114,
+      "loss": 3.1759,
+      "step": 75350
+    },
+    {
+      "epoch": 8.1153804757292,
+      "grad_norm": 0.8385125398635864,
+      "learning_rate": 0.00011336709406313974,
+      "loss": 3.163,
+      "step": 75400
+    },
+    {
+      "epoch": 8.120762027768809,
+      "grad_norm": 0.8474127054214478,
+      "learning_rate": 0.00011304385303307832,
+      "loss": 3.1968,
+      "step": 75450
+    },
+    {
+      "epoch": 8.126143579808417,
+      "grad_norm": 0.8145176768302917,
+      "learning_rate": 0.00011272061200301691,
+      "loss": 3.1802,
+      "step": 75500
+    },
+    {
+      "epoch": 8.131525131848026,
+      "grad_norm": 0.823329508304596,
+      "learning_rate": 0.00011239737097295548,
+      "loss": 3.1642,
+      "step": 75550
+    },
+    {
+      "epoch": 8.136906683887632,
+      "grad_norm": 0.8719600439071655,
+      "learning_rate": 0.00011207412994289408,
+      "loss": 3.1745,
+      "step": 75600
+    },
+    {
+      "epoch": 8.142288235927241,
+      "grad_norm": 0.8388107419013977,
+      "learning_rate": 0.00011175088891283267,
+      "loss": 3.1866,
+      "step": 75650
+    },
+    {
+      "epoch": 8.14766978796685,
+      "grad_norm": 0.8055160641670227,
+      "learning_rate": 0.00011142764788277124,
+      "loss": 3.1704,
+      "step": 75700
+    },
+    {
+      "epoch": 8.153051340006458,
+      "grad_norm": 0.8534181714057922,
+      "learning_rate": 0.00011110440685270983,
+      "loss": 3.1663,
+      "step": 75750
+    },
+    {
+      "epoch": 8.158432892046067,
+      "grad_norm": 0.8355813026428223,
+      "learning_rate": 0.00011078116582264841,
+      "loss": 3.1759,
+      "step": 75800
+    },
+    {
+      "epoch": 8.163814444085673,
+      "grad_norm": 0.8218345046043396,
+      "learning_rate": 0.000110457924792587,
+      "loss": 3.1865,
+      "step": 75850
+    },
+    {
+      "epoch": 8.169195996125282,
+      "grad_norm": 0.8139203190803528,
+      "learning_rate": 0.00011013468376252558,
+      "loss": 3.1649,
+      "step": 75900
+    },
+    {
+      "epoch": 8.17457754816489,
+      "grad_norm": 0.8463836908340454,
+      "learning_rate": 0.00010981144273246417,
+      "loss": 3.1948,
+      "step": 75950
+    },
+    {
+      "epoch": 8.1799591002045,
+      "grad_norm": 0.7959445714950562,
+      "learning_rate": 0.00010948820170240275,
+      "loss": 3.1824,
+      "step": 76000
+    },
+    {
+      "epoch": 8.1799591002045,
+      "eval_accuracy": 0.3903759183477894,
+      "eval_loss": 3.334186553955078,
+      "eval_runtime": 221.1785,
+      "eval_samples_per_second": 81.432,
+      "eval_steps_per_second": 5.091,
+      "step": 76000
+    },
+    {
+      "epoch": 8.185340652244108,
+      "grad_norm": 0.8416066765785217,
+      "learning_rate": 0.00010916496067234133,
+      "loss": 3.1709,
+      "step": 76050
+    },
+    {
+      "epoch": 8.190722204283716,
+      "grad_norm": 0.8329446315765381,
+      "learning_rate": 0.00010884171964227991,
+      "loss": 3.1742,
+      "step": 76100
+    },
+    {
+      "epoch": 8.196103756323323,
+      "grad_norm": 0.8333694934844971,
+      "learning_rate": 0.00010851847861221851,
+      "loss": 3.183,
+      "step": 76150
+    },
+    {
+      "epoch": 8.201485308362932,
+      "grad_norm": 0.8373638987541199,
+      "learning_rate": 0.00010819523758215708,
+      "loss": 3.1723,
+      "step": 76200
+    },
+    {
+      "epoch": 8.20686686040254,
+      "grad_norm": 0.8664990067481995,
+      "learning_rate": 0.00010787199655209567,
+      "loss": 3.182,
+      "step": 76250
+    },
+    {
+      "epoch": 8.212248412442149,
+      "grad_norm": 0.8091091513633728,
+      "learning_rate": 0.00010754875552203425,
+      "loss": 3.1728,
+      "step": 76300
+    },
+    {
+      "epoch": 8.217629964481757,
+      "grad_norm": 0.8922072052955627,
+      "learning_rate": 0.00010723197931257406,
+      "loss": 3.1758,
+      "step": 76350
+    },
+    {
+      "epoch": 8.223011516521364,
+      "grad_norm": 0.8110160231590271,
+      "learning_rate": 0.00010690873828251265,
+      "loss": 3.1921,
+      "step": 76400
+    },
+    {
+      "epoch": 8.228393068560973,
+      "grad_norm": 0.8317387104034424,
+      "learning_rate": 0.00010658549725245123,
+      "loss": 3.1855,
+      "step": 76450
+    },
+    {
+      "epoch": 8.233774620600581,
+      "grad_norm": 0.8507430553436279,
+      "learning_rate": 0.00010626225622238982,
+      "loss": 3.1745,
+      "step": 76500
+    },
+    {
+      "epoch": 8.23915617264019,
+      "grad_norm": 0.8063206672668457,
+      "learning_rate": 0.0001059390151923284,
+      "loss": 3.1763,
+      "step": 76550
+    },
+    {
+      "epoch": 8.244537724679798,
+      "grad_norm": 0.8336623907089233,
+      "learning_rate": 0.00010561577416226699,
+      "loss": 3.1586,
+      "step": 76600
+    },
+    {
+      "epoch": 8.249919276719407,
+      "grad_norm": 0.844421923160553,
+      "learning_rate": 0.00010529253313220556,
+      "loss": 3.1821,
+      "step": 76650
+    },
+    {
+      "epoch": 8.255300828759013,
+      "grad_norm": 0.8457732200622559,
+      "learning_rate": 0.00010496929210214415,
+      "loss": 3.1778,
+      "step": 76700
+    },
+    {
+      "epoch": 8.260682380798622,
+      "grad_norm": 0.8985422849655151,
+      "learning_rate": 0.00010464605107208275,
+      "loss": 3.1718,
+      "step": 76750
+    },
+    {
+      "epoch": 8.26606393283823,
+      "grad_norm": 0.8263887166976929,
+      "learning_rate": 0.00010432281004202133,
+      "loss": 3.1938,
+      "step": 76800
+    },
+    {
+      "epoch": 8.27144548487784,
+      "grad_norm": 0.8684414625167847,
+      "learning_rate": 0.00010399956901195991,
+      "loss": 3.1709,
+      "step": 76850
+    },
+    {
+      "epoch": 8.276827036917448,
+      "grad_norm": 0.8270609378814697,
+      "learning_rate": 0.00010367632798189849,
+      "loss": 3.1768,
+      "step": 76900
+    },
+    {
+      "epoch": 8.282208588957054,
+      "grad_norm": 0.8352575302124023,
+      "learning_rate": 0.00010335308695183709,
+      "loss": 3.1952,
+      "step": 76950
+    },
+    {
+      "epoch": 8.287590140996663,
+      "grad_norm": 0.804658055305481,
+      "learning_rate": 0.00010302984592177565,
+      "loss": 3.1814,
+      "step": 77000
+    },
+    {
+      "epoch": 8.287590140996663,
+      "eval_accuracy": 0.3905953971811747,
+      "eval_loss": 3.3317322731018066,
+      "eval_runtime": 190.2074,
+      "eval_samples_per_second": 94.691,
+      "eval_steps_per_second": 5.92,
+      "step": 77000
+    },
+    {
+      "epoch": 8.292971693036272,
+      "grad_norm": 0.7943740487098694,
+      "learning_rate": 0.00010270660489171425,
+      "loss": 3.1838,
+      "step": 77050
+    },
+    {
+      "epoch": 8.29835324507588,
+      "grad_norm": 0.8035009503364563,
+      "learning_rate": 0.00010238336386165283,
+      "loss": 3.1595,
+      "step": 77100
+    },
+    {
+      "epoch": 8.303734797115489,
+      "grad_norm": 0.8640129566192627,
+      "learning_rate": 0.00010206012283159141,
+      "loss": 3.1757,
+      "step": 77150
+    },
+    {
+      "epoch": 8.309116349155097,
+      "grad_norm": 0.858672559261322,
+      "learning_rate": 0.00010173688180152999,
+      "loss": 3.1889,
+      "step": 77200
+    },
+    {
+      "epoch": 8.314497901194704,
+      "grad_norm": 0.8206643462181091,
+      "learning_rate": 0.00010141364077146859,
+      "loss": 3.1814,
+      "step": 77250
+    },
+    {
+      "epoch": 8.319879453234313,
+      "grad_norm": 0.8095281720161438,
+      "learning_rate": 0.00010109039974140718,
+      "loss": 3.1806,
+      "step": 77300
+    },
+    {
+      "epoch": 8.325261005273921,
+      "grad_norm": 0.8287177085876465,
+      "learning_rate": 0.00010076715871134575,
+      "loss": 3.1703,
+      "step": 77350
+    },
+    {
+      "epoch": 8.33064255731353,
+      "grad_norm": 0.8322443962097168,
+      "learning_rate": 0.00010044391768128433,
+      "loss": 3.187,
+      "step": 77400
+    },
+    {
+      "epoch": 8.336024109353138,
+      "grad_norm": 0.8651958107948303,
+      "learning_rate": 0.00010012067665122292,
+      "loss": 3.1705,
+      "step": 77450
+    },
+    {
+      "epoch": 8.341405661392745,
+      "grad_norm": 0.842995822429657,
+      "learning_rate": 9.979743562116149e-05,
+      "loss": 3.1877,
+      "step": 77500
+    },
+    {
+      "epoch": 8.346787213432354,
+      "grad_norm": 0.8423287868499756,
+      "learning_rate": 9.947419459110009e-05,
+      "loss": 3.2024,
+      "step": 77550
+    },
+    {
+      "epoch": 8.352168765471962,
+      "grad_norm": 0.830844521522522,
+      "learning_rate": 9.915095356103868e-05,
+      "loss": 3.1762,
+      "step": 77600
+    },
+    {
+      "epoch": 8.35755031751157,
+      "grad_norm": 0.8280004262924194,
+      "learning_rate": 9.882771253097725e-05,
+      "loss": 3.1796,
+      "step": 77650
+    },
+    {
+      "epoch": 8.36293186955118,
+      "grad_norm": 0.8298020362854004,
+      "learning_rate": 9.850447150091584e-05,
+      "loss": 3.1803,
+      "step": 77700
+    },
+    {
+      "epoch": 8.368313421590786,
+      "grad_norm": 0.883962869644165,
+      "learning_rate": 9.818123047085442e-05,
+      "loss": 3.1974,
+      "step": 77750
+    },
+    {
+      "epoch": 8.373694973630395,
+      "grad_norm": 0.7966394424438477,
+      "learning_rate": 9.785798944079302e-05,
+      "loss": 3.1821,
+      "step": 77800
+    },
+    {
+      "epoch": 8.379076525670003,
+      "grad_norm": 0.8396945595741272,
+      "learning_rate": 9.753474841073159e-05,
+      "loss": 3.1992,
+      "step": 77850
+    },
+    {
+      "epoch": 8.384458077709612,
+      "grad_norm": 0.8430613279342651,
+      "learning_rate": 9.721150738067018e-05,
+      "loss": 3.186,
+      "step": 77900
+    },
+    {
+      "epoch": 8.38983962974922,
+      "grad_norm": 0.7840559482574463,
+      "learning_rate": 9.688826635060876e-05,
+      "loss": 3.1821,
+      "step": 77950
+    },
+    {
+      "epoch": 8.395221181788829,
+      "grad_norm": 0.8649613261222839,
+      "learning_rate": 9.656502532054734e-05,
+      "loss": 3.1945,
+      "step": 78000
+    },
+    {
+      "epoch": 8.395221181788829,
+      "eval_accuracy": 0.39121287154262435,
+      "eval_loss": 3.3277697563171387,
+      "eval_runtime": 189.8853,
+      "eval_samples_per_second": 94.852,
+      "eval_steps_per_second": 5.93,
+      "step": 78000
+    },
+    {
+      "epoch": 8.400602733828435,
+      "grad_norm": 0.8213173151016235,
+      "learning_rate": 9.624178429048592e-05,
+      "loss": 3.1755,
+      "step": 78050
+    },
+    {
+      "epoch": 8.405984285868044,
+      "grad_norm": 0.8209165930747986,
+      "learning_rate": 9.591854326042452e-05,
+      "loss": 3.18,
+      "step": 78100
+    },
+    {
+      "epoch": 8.411365837907653,
+      "grad_norm": 0.8557418584823608,
+      "learning_rate": 9.559530223036309e-05,
+      "loss": 3.169,
+      "step": 78150
+    },
+    {
+      "epoch": 8.416747389947261,
+      "grad_norm": 0.8438737988471985,
+      "learning_rate": 9.527206120030168e-05,
+      "loss": 3.2073,
+      "step": 78200
+    },
+    {
+      "epoch": 8.42212894198687,
+      "grad_norm": 0.8334610462188721,
+      "learning_rate": 9.494882017024028e-05,
+      "loss": 3.1781,
+      "step": 78250
+    },
+    {
+      "epoch": 8.427510494026476,
+      "grad_norm": 0.8696321845054626,
+      "learning_rate": 9.462557914017886e-05,
+      "loss": 3.1714,
+      "step": 78300
+    },
+    {
+      "epoch": 8.432892046066085,
+      "grad_norm": 0.9134912490844727,
+      "learning_rate": 9.430233811011742e-05,
+      "loss": 3.177,
+      "step": 78350
+    },
+    {
+      "epoch": 8.438273598105694,
+      "grad_norm": 0.8383008241653442,
+      "learning_rate": 9.397909708005602e-05,
+      "loss": 3.173,
+      "step": 78400
+    },
+    {
+      "epoch": 8.443655150145302,
+      "grad_norm": 0.86134934425354,
+      "learning_rate": 9.365585604999461e-05,
+      "loss": 3.1812,
+      "step": 78450
+    },
+    {
+      "epoch": 8.44903670218491,
+      "grad_norm": 0.8300356268882751,
+      "learning_rate": 9.333261501993318e-05,
+      "loss": 3.1625,
+      "step": 78500
+    },
+    {
+      "epoch": 8.45441825422452,
+      "grad_norm": 0.832892119884491,
+      "learning_rate": 9.300937398987178e-05,
+      "loss": 3.1952,
+      "step": 78550
+    },
+    {
+      "epoch": 8.459799806264126,
+      "grad_norm": 0.9564111828804016,
+      "learning_rate": 9.268613295981036e-05,
+      "loss": 3.1878,
+      "step": 78600
+    },
+    {
+      "epoch": 8.465181358303735,
+      "grad_norm": 0.8386510014533997,
+      "learning_rate": 9.236289192974894e-05,
+      "loss": 3.202,
+      "step": 78650
+    },
+    {
+      "epoch": 8.470562910343343,
+      "grad_norm": 0.8402494192123413,
+      "learning_rate": 9.203965089968752e-05,
+      "loss": 3.1691,
+      "step": 78700
+    },
+    {
+      "epoch": 8.475944462382952,
+      "grad_norm": 0.8221197128295898,
+      "learning_rate": 9.171640986962611e-05,
+      "loss": 3.181,
+      "step": 78750
+    },
+    {
+      "epoch": 8.48132601442256,
+      "grad_norm": 0.8528578877449036,
+      "learning_rate": 9.13931688395647e-05,
+      "loss": 3.1747,
+      "step": 78800
+    },
+    {
+      "epoch": 8.486707566462167,
+      "grad_norm": 0.846588134765625,
+      "learning_rate": 9.106992780950328e-05,
+      "loss": 3.1856,
+      "step": 78850
+    },
+    {
+      "epoch": 8.492089118501776,
+      "grad_norm": 0.8404663801193237,
+      "learning_rate": 9.074668677944186e-05,
+      "loss": 3.1811,
+      "step": 78900
+    },
+    {
+      "epoch": 8.497470670541384,
+      "grad_norm": 0.8380141854286194,
+      "learning_rate": 9.042344574938045e-05,
+      "loss": 3.1771,
+      "step": 78950
+    },
+    {
+      "epoch": 8.502852222580993,
+      "grad_norm": 0.8275344371795654,
+      "learning_rate": 9.010020471931902e-05,
+      "loss": 3.1847,
+      "step": 79000
+    },
+    {
+      "epoch": 8.502852222580993,
+      "eval_accuracy": 0.3914904796709904,
+      "eval_loss": 3.3259856700897217,
+      "eval_runtime": 189.3897,
+      "eval_samples_per_second": 95.1,
+      "eval_steps_per_second": 5.945,
+      "step": 79000
+    },
+    {
+      "epoch": 8.508233774620601,
+      "grad_norm": 0.8297053575515747,
+      "learning_rate": 8.977696368925761e-05,
+      "loss": 3.1785,
+      "step": 79050
+    },
+    {
+      "epoch": 8.513615326660208,
+      "grad_norm": 0.8653829097747803,
+      "learning_rate": 8.945372265919621e-05,
+      "loss": 3.2013,
+      "step": 79100
+    },
+    {
+      "epoch": 8.518996878699816,
+      "grad_norm": 0.8570589423179626,
+      "learning_rate": 8.913048162913478e-05,
+      "loss": 3.1734,
+      "step": 79150
+    },
+    {
+      "epoch": 8.524378430739425,
+      "grad_norm": 0.8229767680168152,
+      "learning_rate": 8.880724059907337e-05,
+      "loss": 3.1873,
+      "step": 79200
+    },
+    {
+      "epoch": 8.529759982779034,
+      "grad_norm": 0.8804620504379272,
+      "learning_rate": 8.848399956901195e-05,
+      "loss": 3.1884,
+      "step": 79250
+    },
+    {
+      "epoch": 8.535141534818642,
+      "grad_norm": 0.8399264216423035,
+      "learning_rate": 8.816075853895055e-05,
+      "loss": 3.1759,
+      "step": 79300
+    },
+    {
+      "epoch": 8.54052308685825,
+      "grad_norm": 0.8268064260482788,
+      "learning_rate": 8.783751750888911e-05,
+      "loss": 3.1752,
+      "step": 79350
+    },
+    {
+      "epoch": 8.545904638897857,
+      "grad_norm": 0.8142411112785339,
+      "learning_rate": 8.751427647882771e-05,
+      "loss": 3.1912,
+      "step": 79400
+    },
+    {
+      "epoch": 8.551286190937466,
+      "grad_norm": 0.8386333584785461,
+      "learning_rate": 8.719103544876629e-05,
+      "loss": 3.1949,
+      "step": 79450
+    },
+    {
+      "epoch": 8.556667742977075,
+      "grad_norm": 0.8724285364151001,
+      "learning_rate": 8.686779441870487e-05,
+      "loss": 3.1779,
+      "step": 79500
+    },
+    {
+      "epoch": 8.562049295016683,
+      "grad_norm": 0.8269312977790833,
+      "learning_rate": 8.654455338864345e-05,
+      "loss": 3.1914,
+      "step": 79550
+    },
+    {
+      "epoch": 8.567430847056292,
+      "grad_norm": 0.8587058186531067,
+      "learning_rate": 8.622131235858205e-05,
+      "loss": 3.1748,
+      "step": 79600
+    },
+    {
+      "epoch": 8.572812399095898,
+      "grad_norm": 0.8489319086074829,
+      "learning_rate": 8.589807132852061e-05,
+      "loss": 3.1859,
+      "step": 79650
+    },
+    {
+      "epoch": 8.578193951135507,
+      "grad_norm": 0.8191342949867249,
+      "learning_rate": 8.557483029845921e-05,
+      "loss": 3.1851,
+      "step": 79700
+    },
+    {
+      "epoch": 8.583575503175116,
+      "grad_norm": 0.859576404094696,
+      "learning_rate": 8.525158926839779e-05,
+      "loss": 3.1702,
+      "step": 79750
+    },
+    {
+      "epoch": 8.588957055214724,
+      "grad_norm": 0.8672488927841187,
+      "learning_rate": 8.492834823833639e-05,
+      "loss": 3.196,
+      "step": 79800
+    },
+    {
+      "epoch": 8.594338607254333,
+      "grad_norm": 0.8281680941581726,
+      "learning_rate": 8.460510720827495e-05,
+      "loss": 3.2033,
+      "step": 79850
+    },
+    {
+      "epoch": 8.599720159293941,
+      "grad_norm": 0.8545637726783752,
+      "learning_rate": 8.428186617821355e-05,
+      "loss": 3.1826,
+      "step": 79900
+    },
+    {
+      "epoch": 8.605101711333548,
+      "grad_norm": 0.8453432321548462,
+      "learning_rate": 8.395862514815214e-05,
+      "loss": 3.1924,
+      "step": 79950
+    },
+    {
+      "epoch": 8.610483263373157,
+      "grad_norm": 0.8370970487594604,
+      "learning_rate": 8.363538411809071e-05,
+      "loss": 3.1648,
+      "step": 80000
+    },
+    {
+      "epoch": 8.610483263373157,
+      "eval_accuracy": 0.39175287639506234,
+      "eval_loss": 3.322192668914795,
+      "eval_runtime": 189.7221,
+      "eval_samples_per_second": 94.934,
+      "eval_steps_per_second": 5.935,
+      "step": 80000
+    },
+    {
+      "epoch": 8.615864815412765,
+      "grad_norm": 0.8636527061462402,
+      "learning_rate": 8.33121430880293e-05,
+      "loss": 3.1804,
+      "step": 80050
+    },
+    {
+      "epoch": 8.621246367452374,
+      "grad_norm": 0.8623721599578857,
+      "learning_rate": 8.298890205796789e-05,
+      "loss": 3.1682,
+      "step": 80100
+    },
+    {
+      "epoch": 8.626627919491982,
+      "grad_norm": 0.8467229008674622,
+      "learning_rate": 8.266566102790647e-05,
+      "loss": 3.1673,
+      "step": 80150
+    },
+    {
+      "epoch": 8.632009471531589,
+      "grad_norm": 0.8296520113945007,
+      "learning_rate": 8.234241999784505e-05,
+      "loss": 3.1888,
+      "step": 80200
+    },
+    {
+      "epoch": 8.637391023571197,
+      "grad_norm": 0.8253055214881897,
+      "learning_rate": 8.201917896778364e-05,
+      "loss": 3.1978,
+      "step": 80250
+    },
+    {
+      "epoch": 8.642772575610806,
+      "grad_norm": 0.8557524085044861,
+      "learning_rate": 8.169593793772222e-05,
+      "loss": 3.1835,
+      "step": 80300
+    },
+    {
+      "epoch": 8.648154127650415,
+      "grad_norm": 0.8345833420753479,
+      "learning_rate": 8.137916172826203e-05,
+      "loss": 3.171,
+      "step": 80350
+    },
+    {
+      "epoch": 8.653535679690023,
+      "grad_norm": 0.8740684986114502,
+      "learning_rate": 8.105592069820062e-05,
+      "loss": 3.1772,
+      "step": 80400
+    },
+    {
+      "epoch": 8.658917231729632,
+      "grad_norm": 0.8588552474975586,
+      "learning_rate": 8.073267966813919e-05,
+      "loss": 3.1802,
+      "step": 80450
+    },
+    {
+      "epoch": 8.664298783769238,
+      "grad_norm": 0.8434314727783203,
+      "learning_rate": 8.040943863807779e-05,
+      "loss": 3.1869,
+      "step": 80500
+    },
+    {
+      "epoch": 8.669680335808847,
+      "grad_norm": 0.8790923953056335,
+      "learning_rate": 8.008619760801637e-05,
+      "loss": 3.1811,
+      "step": 80550
+    },
+    {
+      "epoch": 8.675061887848456,
+      "grad_norm": 0.8365694880485535,
+      "learning_rate": 7.976295657795496e-05,
+      "loss": 3.1876,
+      "step": 80600
+    },
+    {
+      "epoch": 8.680443439888064,
+      "grad_norm": 0.8803290724754333,
+      "learning_rate": 7.943971554789353e-05,
+      "loss": 3.1946,
+      "step": 80650
+    },
+    {
+      "epoch": 8.685824991927673,
+      "grad_norm": 0.8246529698371887,
+      "learning_rate": 7.911647451783212e-05,
+      "loss": 3.1868,
+      "step": 80700
+    },
+    {
+      "epoch": 8.69120654396728,
+      "grad_norm": 0.8908007144927979,
+      "learning_rate": 7.879323348777072e-05,
+      "loss": 3.1883,
+      "step": 80750
+    },
+    {
+      "epoch": 8.696588096006888,
+      "grad_norm": 0.8227264285087585,
+      "learning_rate": 7.846999245770929e-05,
+      "loss": 3.1765,
+      "step": 80800
+    },
+    {
+      "epoch": 8.701969648046497,
+      "grad_norm": 0.8562191724777222,
+      "learning_rate": 7.814675142764787e-05,
+      "loss": 3.2047,
+      "step": 80850
+    },
+    {
+      "epoch": 8.707351200086105,
+      "grad_norm": 0.840075671672821,
+      "learning_rate": 7.782351039758646e-05,
+      "loss": 3.1723,
+      "step": 80900
+    },
+    {
+      "epoch": 8.712732752125714,
+      "grad_norm": 0.8150270581245422,
+      "learning_rate": 7.750026936752503e-05,
+      "loss": 3.1816,
+      "step": 80950
+    },
+    {
+      "epoch": 8.718114304165322,
+      "grad_norm": 0.835625410079956,
+      "learning_rate": 7.717702833746362e-05,
+      "loss": 3.179,
+      "step": 81000
+    },
+    {
+      "epoch": 8.718114304165322,
+      "eval_accuracy": 0.3919582303530317,
+      "eval_loss": 3.3187644481658936,
+      "eval_runtime": 189.4444,
+      "eval_samples_per_second": 95.073,
+      "eval_steps_per_second": 5.944,
+      "step": 81000
+    },
+    {
+      "epoch": 8.723495856204929,
+      "grad_norm": 0.8230716586112976,
+      "learning_rate": 7.685378730740222e-05,
+      "loss": 3.1832,
+      "step": 81050
+    },
+    {
+      "epoch": 8.728877408244538,
+      "grad_norm": 0.8545243740081787,
+      "learning_rate": 7.65305462773408e-05,
+      "loss": 3.1767,
+      "step": 81100
+    },
+    {
+      "epoch": 8.734258960284146,
+      "grad_norm": 0.8454192280769348,
+      "learning_rate": 7.620730524727938e-05,
+      "loss": 3.1891,
+      "step": 81150
+    },
+    {
+      "epoch": 8.739640512323755,
+      "grad_norm": 0.8239614367485046,
+      "learning_rate": 7.588406421721796e-05,
+      "loss": 3.1903,
+      "step": 81200
+    },
+    {
+      "epoch": 8.745022064363363,
+      "grad_norm": 0.8292789459228516,
+      "learning_rate": 7.556082318715656e-05,
+      "loss": 3.1815,
+      "step": 81250
+    },
+    {
+      "epoch": 8.75040361640297,
+      "grad_norm": 0.897819995880127,
+      "learning_rate": 7.523758215709513e-05,
+      "loss": 3.2009,
+      "step": 81300
+    },
+    {
+      "epoch": 8.755785168442578,
+      "grad_norm": 0.8491601943969727,
+      "learning_rate": 7.491434112703372e-05,
+      "loss": 3.1827,
+      "step": 81350
+    },
+    {
+      "epoch": 8.761166720482187,
+      "grad_norm": 0.8508619666099548,
+      "learning_rate": 7.45911000969723e-05,
+      "loss": 3.1863,
+      "step": 81400
+    },
+    {
+      "epoch": 8.766548272521796,
+      "grad_norm": 0.835036039352417,
+      "learning_rate": 7.426785906691088e-05,
+      "loss": 3.1758,
+      "step": 81450
+    },
+    {
+      "epoch": 8.771929824561404,
+      "grad_norm": 0.8110843300819397,
+      "learning_rate": 7.394461803684946e-05,
+      "loss": 3.2087,
+      "step": 81500
+    },
+    {
+      "epoch": 8.777311376601011,
+      "grad_norm": 0.9202770590782166,
+      "learning_rate": 7.362137700678806e-05,
+      "loss": 3.171,
+      "step": 81550
+    },
+    {
+      "epoch": 8.78269292864062,
+      "grad_norm": 0.8067617416381836,
+      "learning_rate": 7.329813597672664e-05,
+      "loss": 3.2149,
+      "step": 81600
+    },
+    {
+      "epoch": 8.788074480680228,
+      "grad_norm": 0.8563730120658875,
+      "learning_rate": 7.297489494666523e-05,
+      "loss": 3.1909,
+      "step": 81650
+    },
+    {
+      "epoch": 8.793456032719837,
+      "grad_norm": 0.8543161153793335,
+      "learning_rate": 7.265165391660381e-05,
+      "loss": 3.1819,
+      "step": 81700
+    },
+    {
+      "epoch": 8.798837584759445,
+      "grad_norm": 0.8220990300178528,
+      "learning_rate": 7.23284128865424e-05,
+      "loss": 3.2023,
+      "step": 81750
+    },
+    {
+      "epoch": 8.804219136799054,
+      "grad_norm": 0.8801039457321167,
+      "learning_rate": 7.200517185648098e-05,
+      "loss": 3.187,
+      "step": 81800
+    },
+    {
+      "epoch": 8.80960068883866,
+      "grad_norm": 0.8284077644348145,
+      "learning_rate": 7.168193082641956e-05,
+      "loss": 3.1896,
+      "step": 81850
+    },
+    {
+      "epoch": 8.814982240878269,
+      "grad_norm": 0.8685952425003052,
+      "learning_rate": 7.135868979635815e-05,
+      "loss": 3.179,
+      "step": 81900
+    },
+    {
+      "epoch": 8.820363792917878,
+      "grad_norm": 0.8335699439048767,
+      "learning_rate": 7.103544876629673e-05,
+      "loss": 3.1853,
+      "step": 81950
+    },
+    {
+      "epoch": 8.825745344957486,
+      "grad_norm": 0.8425479531288147,
+      "learning_rate": 7.071220773623531e-05,
+      "loss": 3.1809,
+      "step": 82000
+    },
+    {
+      "epoch": 8.825745344957486,
+      "eval_accuracy": 0.39258222388775027,
+      "eval_loss": 3.3150174617767334,
+      "eval_runtime": 190.1288,
+      "eval_samples_per_second": 94.731,
+      "eval_steps_per_second": 5.922,
+      "step": 82000
+    },
+    {
+      "epoch": 8.831126896997095,
+      "grad_norm": 0.88204026222229,
+      "learning_rate": 7.03889667061739e-05,
+      "loss": 3.1956,
+      "step": 82050
+    },
+    {
+      "epoch": 8.836508449036701,
+      "grad_norm": 0.8744454979896545,
+      "learning_rate": 7.006572567611248e-05,
+      "loss": 3.1756,
+      "step": 82100
+    },
+    {
+      "epoch": 8.84189000107631,
+      "grad_norm": 0.8653980493545532,
+      "learning_rate": 6.974248464605107e-05,
+      "loss": 3.1819,
+      "step": 82150
+    },
+    {
+      "epoch": 8.847271553115919,
+      "grad_norm": 0.8841899037361145,
+      "learning_rate": 6.941924361598965e-05,
+      "loss": 3.2043,
+      "step": 82200
+    },
+    {
+      "epoch": 8.852653105155527,
+      "grad_norm": 0.8339793086051941,
+      "learning_rate": 6.909600258592823e-05,
+      "loss": 3.1922,
+      "step": 82250
+    },
+    {
+      "epoch": 8.858034657195136,
+      "grad_norm": 0.8752954006195068,
+      "learning_rate": 6.877276155586681e-05,
+      "loss": 3.1984,
+      "step": 82300
+    },
+    {
+      "epoch": 8.863416209234742,
+      "grad_norm": 0.839018702507019,
+      "learning_rate": 6.84495205258054e-05,
+      "loss": 3.1782,
+      "step": 82350
+    },
+    {
+      "epoch": 8.868797761274351,
+      "grad_norm": 0.8543075323104858,
+      "learning_rate": 6.813274431634522e-05,
+      "loss": 3.1868,
+      "step": 82400
+    },
+    {
+      "epoch": 8.87417931331396,
+      "grad_norm": 0.8411060571670532,
+      "learning_rate": 6.78095032862838e-05,
+      "loss": 3.1812,
+      "step": 82450
+    },
+    {
+      "epoch": 8.879560865353568,
+      "grad_norm": 0.8576669692993164,
+      "learning_rate": 6.748626225622238e-05,
+      "loss": 3.1995,
+      "step": 82500
+    },
+    {
+      "epoch": 8.884942417393177,
+      "grad_norm": 0.8171575665473938,
+      "learning_rate": 6.716302122616096e-05,
+      "loss": 3.1939,
+      "step": 82550
+    },
+    {
+      "epoch": 8.890323969432785,
+      "grad_norm": 0.8460285067558289,
+      "learning_rate": 6.683978019609955e-05,
+      "loss": 3.1678,
+      "step": 82600
+    },
+    {
+      "epoch": 8.895705521472392,
+      "grad_norm": 0.8574362993240356,
+      "learning_rate": 6.651653916603814e-05,
+      "loss": 3.1833,
+      "step": 82650
+    },
+    {
+      "epoch": 8.901087073512,
+      "grad_norm": 0.8909504413604736,
+      "learning_rate": 6.619329813597673e-05,
+      "loss": 3.1862,
+      "step": 82700
+    },
+    {
+      "epoch": 8.906468625551609,
+      "grad_norm": 0.8539124131202698,
+      "learning_rate": 6.587005710591531e-05,
+      "loss": 3.1692,
+      "step": 82750
+    },
+    {
+      "epoch": 8.911850177591218,
+      "grad_norm": 0.8660421967506409,
+      "learning_rate": 6.554681607585389e-05,
+      "loss": 3.1867,
+      "step": 82800
+    },
+    {
+      "epoch": 8.917231729630826,
+      "grad_norm": 0.8633134365081787,
+      "learning_rate": 6.522357504579247e-05,
+      "loss": 3.1872,
+      "step": 82850
+    },
+    {
+      "epoch": 8.922613281670433,
+      "grad_norm": 0.841986894607544,
+      "learning_rate": 6.490033401573105e-05,
+      "loss": 3.1805,
+      "step": 82900
+    },
+    {
+      "epoch": 8.927994833710041,
+      "grad_norm": 0.87624591588974,
+      "learning_rate": 6.457709298566965e-05,
+      "loss": 3.1753,
+      "step": 82950
+    },
+    {
+      "epoch": 8.93337638574965,
+      "grad_norm": 0.8404093384742737,
+      "learning_rate": 6.425385195560823e-05,
+      "loss": 3.1939,
+      "step": 83000
+    },
+    {
+      "epoch": 8.93337638574965,
+      "eval_accuracy": 0.3927638915161761,
+      "eval_loss": 3.311628580093384,
+      "eval_runtime": 189.6303,
+      "eval_samples_per_second": 94.98,
+      "eval_steps_per_second": 5.938,
+      "step": 83000
+    },
+    {
+      "epoch": 8.938757937789259,
+      "grad_norm": 0.8670982122421265,
+      "learning_rate": 6.393061092554681e-05,
+      "loss": 3.1851,
+      "step": 83050
+    },
+    {
+      "epoch": 8.944139489828867,
+      "grad_norm": 0.8687721490859985,
+      "learning_rate": 6.360736989548539e-05,
+      "loss": 3.1744,
+      "step": 83100
+    },
+    {
+      "epoch": 8.949521041868476,
+      "grad_norm": 0.8958396911621094,
+      "learning_rate": 6.328412886542397e-05,
+      "loss": 3.2024,
+      "step": 83150
+    },
+    {
+      "epoch": 8.954902593908082,
+      "grad_norm": 0.8612350225448608,
+      "learning_rate": 6.296088783536257e-05,
+      "loss": 3.184,
+      "step": 83200
+    },
+    {
+      "epoch": 8.960284145947691,
+      "grad_norm": 0.8626738786697388,
+      "learning_rate": 6.263764680530115e-05,
+      "loss": 3.172,
+      "step": 83250
+    },
+    {
+      "epoch": 8.9656656979873,
+      "grad_norm": 0.8333890438079834,
+      "learning_rate": 6.231440577523973e-05,
+      "loss": 3.1774,
+      "step": 83300
+    },
+    {
+      "epoch": 8.971047250026908,
+      "grad_norm": 0.8436484932899475,
+      "learning_rate": 6.199116474517831e-05,
+      "loss": 3.1899,
+      "step": 83350
+    },
+    {
+      "epoch": 8.976428802066517,
+      "grad_norm": 0.863152265548706,
+      "learning_rate": 6.166792371511689e-05,
+      "loss": 3.1785,
+      "step": 83400
+    },
+    {
+      "epoch": 8.981810354106123,
+      "grad_norm": 0.8480373620986938,
+      "learning_rate": 6.134468268505549e-05,
+      "loss": 3.1859,
+      "step": 83450
+    },
+    {
+      "epoch": 8.987191906145732,
+      "grad_norm": 0.8849746584892273,
+      "learning_rate": 6.10279064755953e-05,
+      "loss": 3.1647,
+      "step": 83500
+    },
+    {
+      "epoch": 8.99257345818534,
+      "grad_norm": 0.8656249642372131,
+      "learning_rate": 6.070466544553388e-05,
+      "loss": 3.1873,
+      "step": 83550
+    },
+    {
+      "epoch": 8.997955010224949,
+      "grad_norm": 0.8052066564559937,
+      "learning_rate": 6.038142441547246e-05,
+      "loss": 3.189,
+      "step": 83600
+    },
+    {
+      "epoch": 9.003336562264558,
+      "grad_norm": 0.8620032668113708,
+      "learning_rate": 6.005818338541105e-05,
+      "loss": 3.147,
+      "step": 83650
+    },
+    {
+      "epoch": 9.008718114304166,
+      "grad_norm": 0.8975006937980652,
+      "learning_rate": 5.973494235534963e-05,
+      "loss": 3.119,
+      "step": 83700
+    },
+    {
+      "epoch": 9.014099666343773,
+      "grad_norm": 0.8371387720108032,
+      "learning_rate": 5.941170132528822e-05,
+      "loss": 3.1233,
+      "step": 83750
+    },
+    {
+      "epoch": 9.019481218383381,
+      "grad_norm": 0.8606297373771667,
+      "learning_rate": 5.90884602952268e-05,
+      "loss": 3.1383,
+      "step": 83800
+    },
+    {
+      "epoch": 9.02486277042299,
+      "grad_norm": 0.8248050212860107,
+      "learning_rate": 5.876521926516538e-05,
+      "loss": 3.1193,
+      "step": 83850
+    },
+    {
+      "epoch": 9.030244322462599,
+      "grad_norm": 0.9047676920890808,
+      "learning_rate": 5.844197823510397e-05,
+      "loss": 3.1173,
+      "step": 83900
+    },
+    {
+      "epoch": 9.035625874502207,
+      "grad_norm": 0.8255630731582642,
+      "learning_rate": 5.811873720504255e-05,
+      "loss": 3.1424,
+      "step": 83950
+    },
+    {
+      "epoch": 9.041007426541814,
+      "grad_norm": 0.8557283878326416,
+      "learning_rate": 5.7795496174981145e-05,
+      "loss": 3.1185,
+      "step": 84000
+    },
+    {
+      "epoch": 9.041007426541814,
+      "eval_accuracy": 0.39279083743235405,
+      "eval_loss": 3.3150100708007812,
+      "eval_runtime": 189.3544,
+      "eval_samples_per_second": 95.118,
+      "eval_steps_per_second": 5.947,
+      "step": 84000
+    },
+    {
+      "epoch": 9.046388978581422,
+      "grad_norm": 0.8470101952552795,
+      "learning_rate": 5.7472255144919726e-05,
+      "loss": 3.1401,
+      "step": 84050
+    },
+    {
+      "epoch": 9.051770530621031,
+      "grad_norm": 0.8201737999916077,
+      "learning_rate": 5.714901411485831e-05,
+      "loss": 3.1163,
+      "step": 84100
+    },
+    {
+      "epoch": 9.05715208266064,
+      "grad_norm": 0.8869134187698364,
+      "learning_rate": 5.6825773084796895e-05,
+      "loss": 3.1184,
+      "step": 84150
+    },
+    {
+      "epoch": 9.062533634700248,
+      "grad_norm": 0.8410740494728088,
+      "learning_rate": 5.6502532054735476e-05,
+      "loss": 3.1237,
+      "step": 84200
+    },
+    {
+      "epoch": 9.067915186739857,
+      "grad_norm": 0.8496425747871399,
+      "learning_rate": 5.6179291024674064e-05,
+      "loss": 3.1181,
+      "step": 84250
+    },
+    {
+      "epoch": 9.073296738779463,
+      "grad_norm": 0.8405110239982605,
+      "learning_rate": 5.5856049994612645e-05,
+      "loss": 3.1268,
+      "step": 84300
+    },
+    {
+      "epoch": 9.078678290819072,
+      "grad_norm": 0.8484364748001099,
+      "learning_rate": 5.5532808964551227e-05,
+      "loss": 3.1385,
+      "step": 84350
+    },
+    {
+      "epoch": 9.08405984285868,
+      "grad_norm": 0.8305612206459045,
+      "learning_rate": 5.5209567934489814e-05,
+      "loss": 3.1149,
+      "step": 84400
+    },
+    {
+      "epoch": 9.089441394898289,
+      "grad_norm": 0.8964749574661255,
+      "learning_rate": 5.4886326904428396e-05,
+      "loss": 3.1238,
+      "step": 84450
+    },
+    {
+      "epoch": 9.094822946937898,
+      "grad_norm": 0.8592599630355835,
+      "learning_rate": 5.4563085874366983e-05,
+      "loss": 3.1313,
+      "step": 84500
+    },
+    {
+      "epoch": 9.100204498977504,
+      "grad_norm": 0.8475759625434875,
+      "learning_rate": 5.4239844844305565e-05,
+      "loss": 3.1228,
+      "step": 84550
+    },
+    {
+      "epoch": 9.105586051017113,
+      "grad_norm": 0.8360592126846313,
+      "learning_rate": 5.3916603814244146e-05,
+      "loss": 3.1551,
+      "step": 84600
+    },
+    {
+      "epoch": 9.110967603056721,
+      "grad_norm": 0.8481616973876953,
+      "learning_rate": 5.3593362784182734e-05,
+      "loss": 3.1232,
+      "step": 84650
+    },
+    {
+      "epoch": 9.11634915509633,
+      "grad_norm": 0.8523578643798828,
+      "learning_rate": 5.3270121754121315e-05,
+      "loss": 3.1473,
+      "step": 84700
+    },
+    {
+      "epoch": 9.121730707135939,
+      "grad_norm": 0.828809916973114,
+      "learning_rate": 5.294688072405991e-05,
+      "loss": 3.1185,
+      "step": 84750
+    },
+    {
+      "epoch": 9.127112259175545,
+      "grad_norm": 0.8841274380683899,
+      "learning_rate": 5.262363969399849e-05,
+      "loss": 3.1403,
+      "step": 84800
+    },
+    {
+      "epoch": 9.132493811215154,
+      "grad_norm": 0.8152642250061035,
+      "learning_rate": 5.230039866393707e-05,
+      "loss": 3.1395,
+      "step": 84850
+    },
+    {
+      "epoch": 9.137875363254762,
+      "grad_norm": 0.8742380738258362,
+      "learning_rate": 5.197715763387566e-05,
+      "loss": 3.1114,
+      "step": 84900
+    },
+    {
+      "epoch": 9.143256915294371,
+      "grad_norm": 0.8858259916305542,
+      "learning_rate": 5.165391660381424e-05,
+      "loss": 3.1252,
+      "step": 84950
+    },
+    {
+      "epoch": 9.14863846733398,
+      "grad_norm": 0.8560479879379272,
+      "learning_rate": 5.133067557375283e-05,
+      "loss": 3.1242,
+      "step": 85000
+    },
+    {
+      "epoch": 9.14863846733398,
+      "eval_accuracy": 0.39295196966498297,
+      "eval_loss": 3.3140878677368164,
+      "eval_runtime": 189.6918,
+      "eval_samples_per_second": 94.949,
+      "eval_steps_per_second": 5.936,
+      "step": 85000
+    },
+    {
+      "epoch": 9.154020019373588,
+      "grad_norm": 0.9070020318031311,
+      "learning_rate": 5.100743454369141e-05,
+      "loss": 3.1242,
+      "step": 85050
+    },
+    {
+      "epoch": 9.159401571413195,
+      "grad_norm": 0.8739373087882996,
+      "learning_rate": 5.068419351362999e-05,
+      "loss": 3.1321,
+      "step": 85100
+    },
+    {
+      "epoch": 9.164783123452803,
+      "grad_norm": 0.8586959838867188,
+      "learning_rate": 5.036095248356858e-05,
+      "loss": 3.1283,
+      "step": 85150
+    },
+    {
+      "epoch": 9.170164675492412,
+      "grad_norm": 0.8272525668144226,
+      "learning_rate": 5.003771145350716e-05,
+      "loss": 3.1298,
+      "step": 85200
+    },
+    {
+      "epoch": 9.17554622753202,
+      "grad_norm": 0.8490020036697388,
+      "learning_rate": 4.971447042344575e-05,
+      "loss": 3.1514,
+      "step": 85250
+    },
+    {
+      "epoch": 9.180927779571629,
+      "grad_norm": 0.8110401630401611,
+      "learning_rate": 4.939122939338433e-05,
+      "loss": 3.1387,
+      "step": 85300
+    },
+    {
+      "epoch": 9.186309331611236,
+      "grad_norm": 0.8785157203674316,
+      "learning_rate": 4.906798836332291e-05,
+      "loss": 3.1274,
+      "step": 85350
+    },
+    {
+      "epoch": 9.191690883650844,
+      "grad_norm": 0.8156988024711609,
+      "learning_rate": 4.87447473332615e-05,
+      "loss": 3.1198,
+      "step": 85400
+    },
+    {
+      "epoch": 9.197072435690453,
+      "grad_norm": 0.8705768585205078,
+      "learning_rate": 4.842150630320008e-05,
+      "loss": 3.1515,
+      "step": 85450
+    },
+    {
+      "epoch": 9.202453987730062,
+      "grad_norm": 0.8462039232254028,
+      "learning_rate": 4.8098265273138667e-05,
+      "loss": 3.1292,
+      "step": 85500
+    },
+    {
+      "epoch": 9.20783553976967,
+      "grad_norm": 0.859347939491272,
+      "learning_rate": 4.777502424307725e-05,
+      "loss": 3.1349,
+      "step": 85550
+    },
+    {
+      "epoch": 9.213217091809279,
+      "grad_norm": 0.8954098224639893,
+      "learning_rate": 4.745178321301583e-05,
+      "loss": 3.1338,
+      "step": 85600
+    },
+    {
+      "epoch": 9.218598643848885,
+      "grad_norm": 0.8277496099472046,
+      "learning_rate": 4.7128542182954423e-05,
+      "loss": 3.1399,
+      "step": 85650
+    },
+    {
+      "epoch": 9.223980195888494,
+      "grad_norm": 0.8506129384040833,
+      "learning_rate": 4.6805301152893005e-05,
+      "loss": 3.1263,
+      "step": 85700
+    },
+    {
+      "epoch": 9.229361747928102,
+      "grad_norm": 0.8558870553970337,
+      "learning_rate": 4.648206012283159e-05,
+      "loss": 3.115,
+      "step": 85750
+    },
+    {
+      "epoch": 9.234743299967711,
+      "grad_norm": 0.8409327268600464,
+      "learning_rate": 4.6158819092770174e-05,
+      "loss": 3.1123,
+      "step": 85800
+    },
+    {
+      "epoch": 9.24012485200732,
+      "grad_norm": 0.8239386677742004,
+      "learning_rate": 4.5835578062708755e-05,
+      "loss": 3.1228,
+      "step": 85850
+    },
+    {
+      "epoch": 9.245506404046926,
+      "grad_norm": 0.863751232624054,
+      "learning_rate": 4.551233703264734e-05,
+      "loss": 3.1394,
+      "step": 85900
+    },
+    {
+      "epoch": 9.250887956086535,
+      "grad_norm": 0.8437860012054443,
+      "learning_rate": 4.5189096002585924e-05,
+      "loss": 3.1205,
+      "step": 85950
+    },
+    {
+      "epoch": 9.256269508126143,
+      "grad_norm": 0.8293090462684631,
+      "learning_rate": 4.486585497252451e-05,
+      "loss": 3.1324,
+      "step": 86000
+    },
+    {
+      "epoch": 9.256269508126143,
+      "eval_accuracy": 0.3932856426834612,
+      "eval_loss": 3.3111629486083984,
+      "eval_runtime": 189.3869,
+      "eval_samples_per_second": 95.102,
+      "eval_steps_per_second": 5.945,
+      "step": 86000
+    },
+    {
+      "epoch": 9.261651060165752,
+      "grad_norm": 0.8659238815307617,
+      "learning_rate": 4.4549078763064325e-05,
+      "loss": 3.1371,
+      "step": 86050
+    },
+    {
+      "epoch": 9.26703261220536,
+      "grad_norm": 0.8517236113548279,
+      "learning_rate": 4.4225837733002906e-05,
+      "loss": 3.1217,
+      "step": 86100
+    },
+    {
+      "epoch": 9.272414164244967,
+      "grad_norm": 0.8780714273452759,
+      "learning_rate": 4.390259670294149e-05,
+      "loss": 3.1419,
+      "step": 86150
+    },
+    {
+      "epoch": 9.277795716284576,
+      "grad_norm": 0.8498876690864563,
+      "learning_rate": 4.3579355672880075e-05,
+      "loss": 3.128,
+      "step": 86200
+    },
+    {
+      "epoch": 9.283177268324184,
+      "grad_norm": 0.8308305740356445,
+      "learning_rate": 4.3256114642818656e-05,
+      "loss": 3.1307,
+      "step": 86250
+    },
+    {
+      "epoch": 9.288558820363793,
+      "grad_norm": 0.8509485125541687,
+      "learning_rate": 4.2932873612757244e-05,
+      "loss": 3.1287,
+      "step": 86300
+    },
+    {
+      "epoch": 9.293940372403402,
+      "grad_norm": 0.8401972055435181,
+      "learning_rate": 4.2609632582695825e-05,
+      "loss": 3.1373,
+      "step": 86350
+    },
+    {
+      "epoch": 9.29932192444301,
+      "grad_norm": 0.8404197096824646,
+      "learning_rate": 4.2286391552634406e-05,
+      "loss": 3.1289,
+      "step": 86400
+    },
+    {
+      "epoch": 9.304703476482617,
+      "grad_norm": 0.8347073197364807,
+      "learning_rate": 4.1963150522572994e-05,
+      "loss": 3.1325,
+      "step": 86450
+    },
+    {
+      "epoch": 9.310085028522225,
+      "grad_norm": 0.8774033188819885,
+      "learning_rate": 4.1639909492511575e-05,
+      "loss": 3.1243,
+      "step": 86500
+    },
+    {
+      "epoch": 9.315466580561834,
+      "grad_norm": 0.9243816137313843,
+      "learning_rate": 4.131666846245017e-05,
+      "loss": 3.1181,
+      "step": 86550
+    },
+    {
+      "epoch": 9.320848132601443,
+      "grad_norm": 0.8504797220230103,
+      "learning_rate": 4.099342743238875e-05,
+      "loss": 3.1397,
+      "step": 86600
+    },
+    {
+      "epoch": 9.326229684641051,
+      "grad_norm": 0.8603664636611938,
+      "learning_rate": 4.067018640232733e-05,
+      "loss": 3.1403,
+      "step": 86650
+    },
+    {
+      "epoch": 9.331611236680658,
+      "grad_norm": 0.8422338366508484,
+      "learning_rate": 4.034694537226592e-05,
+      "loss": 3.1305,
+      "step": 86700
+    },
+    {
+      "epoch": 9.336992788720266,
+      "grad_norm": 0.8984904885292053,
+      "learning_rate": 4.00237043422045e-05,
+      "loss": 3.1268,
+      "step": 86750
+    },
+    {
+      "epoch": 9.342374340759875,
+      "grad_norm": 0.8661045432090759,
+      "learning_rate": 3.970046331214309e-05,
+      "loss": 3.1122,
+      "step": 86800
+    },
+    {
+      "epoch": 9.347755892799483,
+      "grad_norm": 0.8606688380241394,
+      "learning_rate": 3.937722228208167e-05,
+      "loss": 3.1295,
+      "step": 86850
+    },
+    {
+      "epoch": 9.353137444839092,
+      "grad_norm": 0.8438649773597717,
+      "learning_rate": 3.905398125202025e-05,
+      "loss": 3.1299,
+      "step": 86900
+    },
+    {
+      "epoch": 9.3585189968787,
+      "grad_norm": 0.8938318490982056,
+      "learning_rate": 3.873074022195884e-05,
+      "loss": 3.1416,
+      "step": 86950
+    },
+    {
+      "epoch": 9.363900548918307,
+      "grad_norm": 0.8666595816612244,
+      "learning_rate": 3.840749919189742e-05,
+      "loss": 3.1224,
+      "step": 87000
+    },
+    {
+      "epoch": 9.363900548918307,
+      "eval_accuracy": 0.3935082724505931,
+      "eval_loss": 3.309908151626587,
+      "eval_runtime": 189.3766,
+      "eval_samples_per_second": 95.107,
+      "eval_steps_per_second": 5.946,
+      "step": 87000
+    },
+    {
+      "epoch": 9.369282100957916,
+      "grad_norm": 0.8648133873939514,
+      "learning_rate": 3.808425816183601e-05,
+      "loss": 3.1365,
+      "step": 87050
+    },
+    {
+      "epoch": 9.374663652997524,
+      "grad_norm": 0.8668041825294495,
+      "learning_rate": 3.776101713177459e-05,
+      "loss": 3.1471,
+      "step": 87100
+    },
+    {
+      "epoch": 9.380045205037133,
+      "grad_norm": 0.8217251300811768,
+      "learning_rate": 3.743777610171318e-05,
+      "loss": 3.131,
+      "step": 87150
+    },
+    {
+      "epoch": 9.385426757076742,
+      "grad_norm": 0.8287473320960999,
+      "learning_rate": 3.711453507165176e-05,
+      "loss": 3.1532,
+      "step": 87200
+    },
+    {
+      "epoch": 9.390808309116348,
+      "grad_norm": 0.8670193552970886,
+      "learning_rate": 3.679129404159034e-05,
+      "loss": 3.1399,
+      "step": 87250
+    },
+    {
+      "epoch": 9.396189861155957,
+      "grad_norm": 0.8458430171012878,
+      "learning_rate": 3.646805301152893e-05,
+      "loss": 3.1549,
+      "step": 87300
+    },
+    {
+      "epoch": 9.401571413195565,
+      "grad_norm": 0.8716211318969727,
+      "learning_rate": 3.614481198146751e-05,
+      "loss": 3.1314,
+      "step": 87350
+    },
+    {
+      "epoch": 9.406952965235174,
+      "grad_norm": 0.8429900407791138,
+      "learning_rate": 3.5821570951406096e-05,
+      "loss": 3.128,
+      "step": 87400
+    },
+    {
+      "epoch": 9.412334517274783,
+      "grad_norm": 0.8358698487281799,
+      "learning_rate": 3.5498329921344684e-05,
+      "loss": 3.1314,
+      "step": 87450
+    },
+    {
+      "epoch": 9.417716069314391,
+      "grad_norm": 0.8356906175613403,
+      "learning_rate": 3.5175088891283265e-05,
+      "loss": 3.1225,
+      "step": 87500
+    },
+    {
+      "epoch": 9.423097621353998,
+      "grad_norm": 0.8311634063720703,
+      "learning_rate": 3.4851847861221846e-05,
+      "loss": 3.1125,
+      "step": 87550
+    },
+    {
+      "epoch": 9.428479173393606,
+      "grad_norm": 0.8509854674339294,
+      "learning_rate": 3.4528606831160434e-05,
+      "loss": 3.1331,
+      "step": 87600
+    },
+    {
+      "epoch": 9.433860725433215,
+      "grad_norm": 0.890250563621521,
+      "learning_rate": 3.4205365801099015e-05,
+      "loss": 3.1257,
+      "step": 87650
+    },
+    {
+      "epoch": 9.439242277472824,
+      "grad_norm": 0.8621748089790344,
+      "learning_rate": 3.38821247710376e-05,
+      "loss": 3.1427,
+      "step": 87700
+    },
+    {
+      "epoch": 9.444623829512432,
+      "grad_norm": 0.8697271347045898,
+      "learning_rate": 3.3558883740976184e-05,
+      "loss": 3.1607,
+      "step": 87750
+    },
+    {
+      "epoch": 9.450005381552039,
+      "grad_norm": 0.8986278772354126,
+      "learning_rate": 3.323564271091477e-05,
+      "loss": 3.1247,
+      "step": 87800
+    },
+    {
+      "epoch": 9.455386933591647,
+      "grad_norm": 0.8373309373855591,
+      "learning_rate": 3.291240168085335e-05,
+      "loss": 3.1418,
+      "step": 87850
+    },
+    {
+      "epoch": 9.460768485631256,
+      "grad_norm": 0.8242829442024231,
+      "learning_rate": 3.258916065079194e-05,
+      "loss": 3.1379,
+      "step": 87900
+    },
+    {
+      "epoch": 9.466150037670864,
+      "grad_norm": 0.848978579044342,
+      "learning_rate": 3.226591962073052e-05,
+      "loss": 3.1385,
+      "step": 87950
+    },
+    {
+      "epoch": 9.471531589710473,
+      "grad_norm": 0.8514211773872375,
+      "learning_rate": 3.19426785906691e-05,
+      "loss": 3.1341,
+      "step": 88000
+    },
+    {
+      "epoch": 9.471531589710473,
+      "eval_accuracy": 0.39374882994421434,
+      "eval_loss": 3.307783603668213,
+      "eval_runtime": 189.7402,
+      "eval_samples_per_second": 94.925,
+      "eval_steps_per_second": 5.934,
+      "step": 88000
+    },
+    {
+      "epoch": 9.476913141750082,
+      "grad_norm": 0.8711098432540894,
+      "learning_rate": 3.161943756060769e-05,
+      "loss": 3.126,
+      "step": 88050
+    },
+    {
+      "epoch": 9.482294693789688,
+      "grad_norm": 0.8469831943511963,
+      "learning_rate": 3.129619653054627e-05,
+      "loss": 3.1443,
+      "step": 88100
+    },
+    {
+      "epoch": 9.487676245829297,
+      "grad_norm": 0.8410546183586121,
+      "learning_rate": 3.097295550048486e-05,
+      "loss": 3.1266,
+      "step": 88150
+    },
+    {
+      "epoch": 9.493057797868905,
+      "grad_norm": 0.8234712481498718,
+      "learning_rate": 3.064971447042345e-05,
+      "loss": 3.1295,
+      "step": 88200
+    },
+    {
+      "epoch": 9.498439349908514,
+      "grad_norm": 0.8497011661529541,
+      "learning_rate": 3.0326473440362026e-05,
+      "loss": 3.131,
+      "step": 88250
+    },
+    {
+      "epoch": 9.503820901948123,
+      "grad_norm": 0.8871855735778809,
+      "learning_rate": 3.000323241030061e-05,
+      "loss": 3.1225,
+      "step": 88300
+    },
+    {
+      "epoch": 9.50920245398773,
+      "grad_norm": 0.8837377429008484,
+      "learning_rate": 2.9679991380239195e-05,
+      "loss": 3.1358,
+      "step": 88350
+    },
+    {
+      "epoch": 9.514584006027338,
+      "grad_norm": 0.8673506379127502,
+      "learning_rate": 2.9356750350177782e-05,
+      "loss": 3.1208,
+      "step": 88400
+    },
+    {
+      "epoch": 9.519965558066946,
+      "grad_norm": 0.8431509733200073,
+      "learning_rate": 2.9033509320116367e-05,
+      "loss": 3.1308,
+      "step": 88450
+    },
+    {
+      "epoch": 9.525347110106555,
+      "grad_norm": 0.8733773827552795,
+      "learning_rate": 2.8710268290054948e-05,
+      "loss": 3.1542,
+      "step": 88500
+    },
+    {
+      "epoch": 9.530728662146164,
+      "grad_norm": 0.9029594659805298,
+      "learning_rate": 2.8387027259993533e-05,
+      "loss": 3.133,
+      "step": 88550
+    },
+    {
+      "epoch": 9.536110214185772,
+      "grad_norm": 0.8783562183380127,
+      "learning_rate": 2.8063786229932117e-05,
+      "loss": 3.1165,
+      "step": 88600
+    },
+    {
+      "epoch": 9.541491766225379,
+      "grad_norm": 0.8718871474266052,
+      "learning_rate": 2.77405451998707e-05,
+      "loss": 3.1176,
+      "step": 88650
+    },
+    {
+      "epoch": 9.546873318264987,
+      "grad_norm": 0.8462399244308472,
+      "learning_rate": 2.7417304169809286e-05,
+      "loss": 3.1366,
+      "step": 88700
+    },
+    {
+      "epoch": 9.552254870304596,
+      "grad_norm": 0.8613501191139221,
+      "learning_rate": 2.7094063139747867e-05,
+      "loss": 3.1189,
+      "step": 88750
+    },
+    {
+      "epoch": 9.557636422344205,
+      "grad_norm": 0.8645268678665161,
+      "learning_rate": 2.677082210968645e-05,
+      "loss": 3.1303,
+      "step": 88800
+    },
+    {
+      "epoch": 9.563017974383813,
+      "grad_norm": 0.861893892288208,
+      "learning_rate": 2.644758107962504e-05,
+      "loss": 3.1502,
+      "step": 88850
+    },
+    {
+      "epoch": 9.56839952642342,
+      "grad_norm": 0.828042209148407,
+      "learning_rate": 2.6124340049563624e-05,
+      "loss": 3.1388,
+      "step": 88900
+    },
+    {
+      "epoch": 9.573781078463028,
+      "grad_norm": 0.8182950019836426,
+      "learning_rate": 2.580109901950221e-05,
+      "loss": 3.1395,
+      "step": 88950
+    },
+    {
+      "epoch": 9.579162630502637,
+      "grad_norm": 0.8487094640731812,
+      "learning_rate": 2.547785798944079e-05,
+      "loss": 3.1354,
+      "step": 89000
+    },
+    {
+      "epoch": 9.579162630502637,
+      "eval_accuracy": 0.3941069498624509,
+      "eval_loss": 3.3047800064086914,
+      "eval_runtime": 189.7111,
+      "eval_samples_per_second": 94.939,
+      "eval_steps_per_second": 5.935,
+      "step": 89000
+    },
+    {
+      "epoch": 9.584544182542245,
+      "grad_norm": 0.8931549787521362,
+      "learning_rate": 2.5154616959379374e-05,
+      "loss": 3.138,
+      "step": 89050
+    },
+    {
+      "epoch": 9.589925734581854,
+      "grad_norm": 0.8639717102050781,
+      "learning_rate": 2.483137592931796e-05,
+      "loss": 3.15,
+      "step": 89100
+    },
+    {
+      "epoch": 9.59530728662146,
+      "grad_norm": 0.8543230295181274,
+      "learning_rate": 2.4508134899256543e-05,
+      "loss": 3.1383,
+      "step": 89150
+    },
+    {
+      "epoch": 9.60068883866107,
+      "grad_norm": 0.9166204929351807,
+      "learning_rate": 2.418489386919513e-05,
+      "loss": 3.1253,
+      "step": 89200
+    },
+    {
+      "epoch": 9.606070390700678,
+      "grad_norm": 0.8401595950126648,
+      "learning_rate": 2.386165283913371e-05,
+      "loss": 3.1514,
+      "step": 89250
+    },
+    {
+      "epoch": 9.611451942740286,
+      "grad_norm": 0.8688633441925049,
+      "learning_rate": 2.3544876629673522e-05,
+      "loss": 3.1363,
+      "step": 89300
+    },
+    {
+      "epoch": 9.616833494779895,
+      "grad_norm": 0.8411018252372742,
+      "learning_rate": 2.3221635599612106e-05,
+      "loss": 3.1256,
+      "step": 89350
+    },
+    {
+      "epoch": 9.622215046819504,
+      "grad_norm": 0.8895683288574219,
+      "learning_rate": 2.2898394569550694e-05,
+      "loss": 3.1329,
+      "step": 89400
+    },
+    {
+      "epoch": 9.62759659885911,
+      "grad_norm": 0.8607434034347534,
+      "learning_rate": 2.257515353948928e-05,
+      "loss": 3.1494,
+      "step": 89450
+    },
+    {
+      "epoch": 9.632978150898719,
+      "grad_norm": 0.8818168640136719,
+      "learning_rate": 2.2251912509427863e-05,
+      "loss": 3.1479,
+      "step": 89500
+    },
+    {
+      "epoch": 9.638359702938327,
+      "grad_norm": 0.8703548312187195,
+      "learning_rate": 2.1928671479366444e-05,
+      "loss": 3.1207,
+      "step": 89550
+    },
+    {
+      "epoch": 9.643741254977936,
+      "grad_norm": 0.8854098916053772,
+      "learning_rate": 2.160543044930503e-05,
+      "loss": 3.1324,
+      "step": 89600
+    },
+    {
+      "epoch": 9.649122807017545,
+      "grad_norm": 0.8342804908752441,
+      "learning_rate": 2.1282189419243613e-05,
+      "loss": 3.1383,
+      "step": 89650
+    },
+    {
+      "epoch": 9.654504359057151,
+      "grad_norm": 0.8617409467697144,
+      "learning_rate": 2.0958948389182198e-05,
+      "loss": 3.1366,
+      "step": 89700
+    },
+    {
+      "epoch": 9.65988591109676,
+      "grad_norm": 0.8699505925178528,
+      "learning_rate": 2.0635707359120786e-05,
+      "loss": 3.1416,
+      "step": 89750
+    },
+    {
+      "epoch": 9.665267463136368,
+      "grad_norm": 0.8576476573944092,
+      "learning_rate": 2.0312466329059367e-05,
+      "loss": 3.1251,
+      "step": 89800
+    },
+    {
+      "epoch": 9.670649015175977,
+      "grad_norm": 0.848074197769165,
+      "learning_rate": 1.998922529899795e-05,
+      "loss": 3.1331,
+      "step": 89850
+    },
+    {
+      "epoch": 9.676030567215586,
+      "grad_norm": 0.8670492768287659,
+      "learning_rate": 1.9665984268936536e-05,
+      "loss": 3.1474,
+      "step": 89900
+    },
+    {
+      "epoch": 9.681412119255192,
+      "grad_norm": 0.8532721996307373,
+      "learning_rate": 1.934274323887512e-05,
+      "loss": 3.1403,
+      "step": 89950
+    },
+    {
+      "epoch": 9.6867936712948,
+      "grad_norm": 0.8756169080734253,
+      "learning_rate": 1.9019502208813705e-05,
+      "loss": 3.1262,
+      "step": 90000
+    },
+    {
+      "epoch": 9.6867936712948,
+      "eval_accuracy": 0.3943953146267105,
+      "eval_loss": 3.3019161224365234,
+      "eval_runtime": 189.5494,
+      "eval_samples_per_second": 95.02,
+      "eval_steps_per_second": 5.94,
+      "step": 90000
+    },
+    {
+      "epoch": 9.69217522333441,
+      "grad_norm": 0.8606821894645691,
+      "learning_rate": 1.869626117875229e-05,
+      "loss": 3.1353,
+      "step": 90050
+    },
+    {
+      "epoch": 9.697556775374018,
+      "grad_norm": 0.8910887241363525,
+      "learning_rate": 1.837302014869087e-05,
+      "loss": 3.1403,
+      "step": 90100
+    },
+    {
+      "epoch": 9.702938327413626,
+      "grad_norm": 0.9061274528503418,
+      "learning_rate": 1.8049779118629455e-05,
+      "loss": 3.14,
+      "step": 90150
+    },
+    {
+      "epoch": 9.708319879453235,
+      "grad_norm": 0.8429542183876038,
+      "learning_rate": 1.7726538088568043e-05,
+      "loss": 3.1161,
+      "step": 90200
+    },
+    {
+      "epoch": 9.713701431492842,
+      "grad_norm": 0.8305377960205078,
+      "learning_rate": 1.7403297058506624e-05,
+      "loss": 3.1368,
+      "step": 90250
+    },
+    {
+      "epoch": 9.71908298353245,
+      "grad_norm": 0.837834894657135,
+      "learning_rate": 1.708005602844521e-05,
+      "loss": 3.1374,
+      "step": 90300
+    },
+    {
+      "epoch": 9.724464535572059,
+      "grad_norm": 0.852417528629303,
+      "learning_rate": 1.6756814998383793e-05,
+      "loss": 3.1363,
+      "step": 90350
+    },
+    {
+      "epoch": 9.729846087611667,
+      "grad_norm": 0.87115079164505,
+      "learning_rate": 1.6433573968322377e-05,
+      "loss": 3.1364,
+      "step": 90400
+    },
+    {
+      "epoch": 9.735227639651276,
+      "grad_norm": 0.8634170293807983,
+      "learning_rate": 1.6110332938260962e-05,
+      "loss": 3.1237,
+      "step": 90450
+    },
+    {
+      "epoch": 9.740609191690883,
+      "grad_norm": 0.871769905090332,
+      "learning_rate": 1.5787091908199546e-05,
+      "loss": 3.156,
+      "step": 90500
+    },
+    {
+      "epoch": 9.745990743730491,
+      "grad_norm": 0.8181995749473572,
+      "learning_rate": 1.546385087813813e-05,
+      "loss": 3.1309,
+      "step": 90550
+    },
+    {
+      "epoch": 9.7513722957701,
+      "grad_norm": 0.8241446614265442,
+      "learning_rate": 1.5140609848076714e-05,
+      "loss": 3.1164,
+      "step": 90600
+    },
+    {
+      "epoch": 9.756753847809708,
+      "grad_norm": 0.8035761117935181,
+      "learning_rate": 1.4817368818015298e-05,
+      "loss": 3.1295,
+      "step": 90650
+    },
+    {
+      "epoch": 9.762135399849317,
+      "grad_norm": 0.8078320622444153,
+      "learning_rate": 1.4494127787953884e-05,
+      "loss": 3.1354,
+      "step": 90700
+    },
+    {
+      "epoch": 9.767516951888926,
+      "grad_norm": 0.9000764489173889,
+      "learning_rate": 1.4170886757892467e-05,
+      "loss": 3.1242,
+      "step": 90750
+    },
+    {
+      "epoch": 9.772898503928532,
+      "grad_norm": 0.8568733334541321,
+      "learning_rate": 1.3847645727831052e-05,
+      "loss": 3.117,
+      "step": 90800
+    },
+    {
+      "epoch": 9.77828005596814,
+      "grad_norm": 0.8243940472602844,
+      "learning_rate": 1.3524404697769635e-05,
+      "loss": 3.1242,
+      "step": 90850
+    },
+    {
+      "epoch": 9.78366160800775,
+      "grad_norm": 0.8713697195053101,
+      "learning_rate": 1.320116366770822e-05,
+      "loss": 3.13,
+      "step": 90900
+    },
+    {
+      "epoch": 9.789043160047358,
+      "grad_norm": 0.837845504283905,
+      "learning_rate": 1.2877922637646805e-05,
+      "loss": 3.1421,
+      "step": 90950
+    },
+    {
+      "epoch": 9.794424712086967,
+      "grad_norm": 0.8562050461769104,
+      "learning_rate": 1.2554681607585388e-05,
+      "loss": 3.129,
+      "step": 91000
+    },
+    {
+      "epoch": 9.794424712086967,
+      "eval_accuracy": 0.3945871956265909,
+      "eval_loss": 3.300642490386963,
+      "eval_runtime": 190.2591,
+      "eval_samples_per_second": 94.666,
+      "eval_steps_per_second": 5.918,
+      "step": 91000
+    },
+    {
+      "epoch": 9.799806264126573,
+      "grad_norm": 0.8212922811508179,
+      "learning_rate": 1.2231440577523972e-05,
+      "loss": 3.1433,
+      "step": 91050
+    },
+    {
+      "epoch": 9.805187816166182,
+      "grad_norm": 0.9242744445800781,
+      "learning_rate": 1.1908199547462555e-05,
+      "loss": 3.1234,
+      "step": 91100
+    },
+    {
+      "epoch": 9.81056936820579,
+      "grad_norm": 0.8359904289245605,
+      "learning_rate": 1.1584958517401141e-05,
+      "loss": 3.1384,
+      "step": 91150
+    },
+    {
+      "epoch": 9.815950920245399,
+      "grad_norm": 0.8217136263847351,
+      "learning_rate": 1.1261717487339726e-05,
+      "loss": 3.1144,
+      "step": 91200
+    },
+    {
+      "epoch": 9.821332472285007,
+      "grad_norm": 0.8538930416107178,
+      "learning_rate": 1.0938476457278309e-05,
+      "loss": 3.1489,
+      "step": 91250
+    },
+    {
+      "epoch": 9.826714024324616,
+      "grad_norm": 0.8195740580558777,
+      "learning_rate": 1.0615235427216895e-05,
+      "loss": 3.1369,
+      "step": 91300
+    },
+    {
+      "epoch": 9.832095576364223,
+      "grad_norm": 0.8804702162742615,
+      "learning_rate": 1.0291994397155478e-05,
+      "loss": 3.1469,
+      "step": 91350
+    },
+    {
+      "epoch": 9.837477128403831,
+      "grad_norm": 0.8412031531333923,
+      "learning_rate": 9.968753367094062e-06,
+      "loss": 3.1415,
+      "step": 91400
+    },
+    {
+      "epoch": 9.84285868044344,
+      "grad_norm": 0.8647484183311462,
+      "learning_rate": 9.645512337032647e-06,
+      "loss": 3.1224,
+      "step": 91450
+    },
+    {
+      "epoch": 9.848240232483048,
+      "grad_norm": 0.8878872990608215,
+      "learning_rate": 9.32227130697123e-06,
+      "loss": 3.1219,
+      "step": 91500
+    },
+    {
+      "epoch": 9.853621784522657,
+      "grad_norm": 0.839019775390625,
+      "learning_rate": 8.999030276909816e-06,
+      "loss": 3.1497,
+      "step": 91550
+    },
+    {
+      "epoch": 9.859003336562264,
+      "grad_norm": 0.8272830247879028,
+      "learning_rate": 8.675789246848399e-06,
+      "loss": 3.1342,
+      "step": 91600
+    },
+    {
+      "epoch": 9.864384888601872,
+      "grad_norm": 0.8204966187477112,
+      "learning_rate": 8.352548216786983e-06,
+      "loss": 3.1181,
+      "step": 91650
+    },
+    {
+      "epoch": 9.869766440641481,
+      "grad_norm": 0.8158440589904785,
+      "learning_rate": 8.029307186725568e-06,
+      "loss": 3.1251,
+      "step": 91700
+    },
+    {
+      "epoch": 9.87514799268109,
+      "grad_norm": 0.9214332103729248,
+      "learning_rate": 7.706066156664152e-06,
+      "loss": 3.1234,
+      "step": 91750
+    },
+    {
+      "epoch": 9.880529544720698,
+      "grad_norm": 0.8618704676628113,
+      "learning_rate": 7.3828251266027365e-06,
+      "loss": 3.1344,
+      "step": 91800
+    },
+    {
+      "epoch": 9.885911096760307,
+      "grad_norm": 0.8866255879402161,
+      "learning_rate": 7.05958409654132e-06,
+      "loss": 3.1481,
+      "step": 91850
+    },
+    {
+      "epoch": 9.891292648799913,
+      "grad_norm": 0.8363102674484253,
+      "learning_rate": 6.736343066479905e-06,
+      "loss": 3.1315,
+      "step": 91900
+    },
+    {
+      "epoch": 9.896674200839522,
+      "grad_norm": 0.8203515410423279,
+      "learning_rate": 6.413102036418488e-06,
+      "loss": 3.1338,
+      "step": 91950
+    },
+    {
+      "epoch": 9.90205575287913,
+      "grad_norm": 0.8286318778991699,
+      "learning_rate": 6.089861006357073e-06,
+      "loss": 3.1392,
+      "step": 92000
+    },
+    {
+      "epoch": 9.90205575287913,
+      "eval_accuracy": 0.3947092128196065,
+      "eval_loss": 3.29917311668396,
+      "eval_runtime": 189.5412,
+      "eval_samples_per_second": 95.024,
+      "eval_steps_per_second": 5.941,
+      "step": 92000
+    },
+    {
+      "epoch": 9.907437304918739,
+      "grad_norm": 0.8603388071060181,
+      "learning_rate": 5.766619976295657e-06,
+      "loss": 3.1372,
+      "step": 92050
+    },
+    {
+      "epoch": 9.912818856958348,
+      "grad_norm": 0.9106926321983337,
+      "learning_rate": 5.443378946234242e-06,
+      "loss": 3.1309,
+      "step": 92100
+    },
+    {
+      "epoch": 9.918200408997954,
+      "grad_norm": 0.8053128123283386,
+      "learning_rate": 5.1201379161728254e-06,
+      "loss": 3.1273,
+      "step": 92150
+    },
+    {
+      "epoch": 9.923581961037563,
+      "grad_norm": 0.8366634249687195,
+      "learning_rate": 4.79689688611141e-06,
+      "loss": 3.1282,
+      "step": 92200
+    },
+    {
+      "epoch": 9.928963513077171,
+      "grad_norm": 0.8284056782722473,
+      "learning_rate": 4.4736558560499944e-06,
+      "loss": 3.141,
+      "step": 92250
+    },
+    {
+      "epoch": 9.93434506511678,
+      "grad_norm": 0.8485735058784485,
+      "learning_rate": 4.150414825988578e-06,
+      "loss": 3.1262,
+      "step": 92300
+    },
+    {
+      "epoch": 9.939726617156388,
+      "grad_norm": 0.859318196773529,
+      "learning_rate": 3.8271737959271626e-06,
+      "loss": 3.1151,
+      "step": 92350
+    },
+    {
+      "epoch": 9.945108169195997,
+      "grad_norm": 0.8240682482719421,
+      "learning_rate": 3.5039327658657466e-06,
+      "loss": 3.1192,
+      "step": 92400
+    },
+    {
+      "epoch": 9.950489721235604,
+      "grad_norm": 0.850426971912384,
+      "learning_rate": 3.180691735804331e-06,
+      "loss": 3.1372,
+      "step": 92450
+    },
+    {
+      "epoch": 9.955871273275212,
+      "grad_norm": 0.8800275921821594,
+      "learning_rate": 2.857450705742915e-06,
+      "loss": 3.1292,
+      "step": 92500
+    },
+    {
+      "epoch": 9.961252825314821,
+      "grad_norm": 0.8482245802879333,
+      "learning_rate": 2.5342096756814993e-06,
+      "loss": 3.141,
+      "step": 92550
+    },
+    {
+      "epoch": 9.96663437735443,
+      "grad_norm": 0.8627044558525085,
+      "learning_rate": 2.2109686456200838e-06,
+      "loss": 3.1262,
+      "step": 92600
+    },
+    {
+      "epoch": 9.972015929394038,
+      "grad_norm": 0.9186998009681702,
+      "learning_rate": 1.8877276155586683e-06,
+      "loss": 3.1342,
+      "step": 92650
+    },
+    {
+      "epoch": 9.977397481433645,
+      "grad_norm": 0.9178165197372437,
+      "learning_rate": 1.5644865854972521e-06,
+      "loss": 3.1263,
+      "step": 92700
+    },
+    {
+      "epoch": 9.982779033473253,
+      "grad_norm": 0.8731764554977417,
+      "learning_rate": 1.2412455554358364e-06,
+      "loss": 3.1389,
+      "step": 92750
+    },
+    {
+      "epoch": 9.988160585512862,
+      "grad_norm": 0.8474687337875366,
+      "learning_rate": 9.180045253744208e-07,
+      "loss": 3.1368,
+      "step": 92800
+    },
+    {
+      "epoch": 9.99354213755247,
+      "grad_norm": 0.8867104649543762,
+      "learning_rate": 5.947634953130051e-07,
+      "loss": 3.153,
+      "step": 92850
+    },
+    {
+      "epoch": 9.998923689592079,
+      "grad_norm": 0.8560895919799805,
+      "learning_rate": 2.7152246525158925e-07,
+      "loss": 3.1305,
+      "step": 92900
+    },
+    {
+      "epoch": 10.0,
+      "step": 92910,
+      "total_flos": 7.76821211136e+17,
+      "train_loss": 3.4556877505752324,
+      "train_runtime": 82552.4271,
+      "train_samples_per_second": 36.013,
+      "train_steps_per_second": 1.125
+    }
+  ],
+  "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
+}