diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,6139 @@
+{
+  "best_global_step": 5300,
+  "best_metric": 1.168828010559082,
+  "best_model_checkpoint": "/workspace/project_2026_1/checkpoints/sft/checkpoint-5300",
+  "epoch": 1.6181229773462782,
+  "eval_steps": 100,
+  "global_step": 5500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.1349310517311095,
+      "epoch": 0.0029420417769932335,
+      "grad_norm": 5.712224006652832,
+      "learning_rate": 3.5294117647058825e-06,
+      "loss": 3.9359054565429688,
+      "mean_token_accuracy": 0.45907036662101747,
+      "num_tokens": 40265.0,
+      "step": 10
+    },
+    {
+      "entropy": 1.1904624342918395,
+      "epoch": 0.005884083553986467,
+      "grad_norm": 3.6260242462158203,
+      "learning_rate": 7.450980392156863e-06,
+      "loss": 3.447366714477539,
+      "mean_token_accuracy": 0.4740333199501038,
+      "num_tokens": 80768.0,
+      "step": 20
+    },
+    {
+      "entropy": 1.641614854335785,
+      "epoch": 0.0088261253309797,
+      "grad_norm": 1.758035659790039,
+      "learning_rate": 1.1372549019607843e-05,
+      "loss": 2.9342424392700197,
+      "mean_token_accuracy": 0.4756320804357529,
+      "num_tokens": 121287.0,
+      "step": 30
+    },
+    {
+      "entropy": 2.0289795875549315,
+      "epoch": 0.011768167107972934,
+      "grad_norm": 0.8422009348869324,
+      "learning_rate": 1.5294117647058826e-05,
+      "loss": 2.4612199783325197,
+      "mean_token_accuracy": 0.5210394144058228,
+      "num_tokens": 161772.0,
+      "step": 40
+    },
+    {
+      "entropy": 2.1589465618133543,
+      "epoch": 0.014710208884966167,
+      "grad_norm": 0.6909259557723999,
+      "learning_rate": 1.9215686274509807e-05,
+      "loss": 2.1433944702148438,
+      "mean_token_accuracy": 0.5696752369403839,
+      "num_tokens": 202309.0,
+      "step": 50
+    },
+    {
+      "entropy": 1.932080829143524,
+      "epoch": 0.0176522506619594,
+      "grad_norm": 0.7911986112594604,
+      "learning_rate": 2.3137254901960788e-05,
+      "loss": 1.8811990737915039,
+      "mean_token_accuracy": 0.6150015771389008,
+      "num_tokens": 242490.0,
+      "step": 60
+    },
+    {
+      "entropy": 1.6605080842971802,
+      "epoch": 0.020594292438952635,
+      "grad_norm": 0.790881872177124,
+      "learning_rate": 2.7058823529411766e-05,
+      "loss": 1.6363672256469726,
+      "mean_token_accuracy": 0.6599553287029266,
+      "num_tokens": 282845.0,
+      "step": 70
+    },
+    {
+      "entropy": 1.5992008209228517,
+      "epoch": 0.023536334215945868,
+      "grad_norm": 0.4468841850757599,
+      "learning_rate": 3.098039215686275e-05,
+      "loss": 1.6219806671142578,
+      "mean_token_accuracy": 0.6577221751213074,
+      "num_tokens": 323245.0,
+      "step": 80
+    },
+    {
+      "entropy": 1.5317162871360779,
+      "epoch": 0.0264783759929391,
+      "grad_norm": 0.49570420384407043,
+      "learning_rate": 3.4901960784313725e-05,
+      "loss": 1.5623438835144043,
+      "mean_token_accuracy": 0.6620652556419373,
+      "num_tokens": 363709.0,
+      "step": 90
+    },
+    {
+      "entropy": 1.5458295822143555,
+      "epoch": 0.029420417769932334,
+      "grad_norm": 0.4536747634410858,
+      "learning_rate": 3.882352941176471e-05,
+      "loss": 1.577082633972168,
+      "mean_token_accuracy": 0.6642140686511994,
+      "num_tokens": 404093.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.029420417769932334,
+      "eval_entropy": 1.519468894467892,
+      "eval_loss": 1.5212204456329346,
+      "eval_mean_token_accuracy": 0.6755449191478919,
+      "eval_num_tokens": 404093.0,
+      "eval_runtime": 116.8522,
+      "eval_samples_per_second": 26.059,
+      "eval_steps_per_second": 3.261,
+      "step": 100
+    },
+    {
+      "entropy": 1.5117838263511658,
+      "epoch": 0.032362459546925564,
+      "grad_norm": 0.4160189628601074,
+      "learning_rate": 4.274509803921569e-05,
+      "loss": 1.5348424911499023,
+      "mean_token_accuracy": 0.6751707136631012,
+      "num_tokens": 444665.0,
+      "step": 110
+    },
+    {
+      "entropy": 1.4816083550453185,
+      "epoch": 0.0353045013239188,
+      "grad_norm": 0.4050695300102234,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.5027993202209473,
+      "mean_token_accuracy": 0.6793931305408478,
+      "num_tokens": 485247.0,
+      "step": 120
+    },
+    {
+      "entropy": 1.5376260161399842,
+      "epoch": 0.03824654310091203,
+      "grad_norm": 0.4373331367969513,
+      "learning_rate": 5.058823529411765e-05,
+      "loss": 1.5507566452026367,
+      "mean_token_accuracy": 0.6680718421936035,
+      "num_tokens": 525640.0,
+      "step": 130
+    },
+    {
+      "entropy": 1.568453598022461,
+      "epoch": 0.04118858487790527,
+      "grad_norm": 0.41970527172088623,
+      "learning_rate": 5.450980392156863e-05,
+      "loss": 1.5769514083862304,
+      "mean_token_accuracy": 0.6546712338924408,
+      "num_tokens": 565941.0,
+      "step": 140
+    },
+    {
+      "entropy": 1.4061901330947877,
+      "epoch": 0.0441306266548985,
+      "grad_norm": 0.43462952971458435,
+      "learning_rate": 5.843137254901961e-05,
+      "loss": 1.4184735298156739,
+      "mean_token_accuracy": 0.6973597705364227,
+      "num_tokens": 606491.0,
+      "step": 150
+    },
+    {
+      "entropy": 1.4754432439804077,
+      "epoch": 0.047072668431891736,
+      "grad_norm": 0.42188599705696106,
+      "learning_rate": 6.23529411764706e-05,
+      "loss": 1.5123021125793457,
+      "mean_token_accuracy": 0.6780355155467988,
+      "num_tokens": 646949.0,
+      "step": 160
+    },
+    {
+      "entropy": 1.4351889610290527,
+      "epoch": 0.05001471020888497,
+      "grad_norm": 0.4059581458568573,
+      "learning_rate": 6.627450980392157e-05,
+      "loss": 1.4382680892944335,
+      "mean_token_accuracy": 0.6908938407897949,
+      "num_tokens": 687299.0,
+      "step": 170
+    },
+    {
+      "entropy": 1.4386033058166503,
+      "epoch": 0.0529567519858782,
+      "grad_norm": 0.4081089198589325,
+      "learning_rate": 7.019607843137255e-05,
+      "loss": 1.4630658149719238,
+      "mean_token_accuracy": 0.6837093889713287,
+      "num_tokens": 727817.0,
+      "step": 180
+    },
+    {
+      "entropy": 1.4710845828056336,
+      "epoch": 0.055898793762871435,
+      "grad_norm": 0.4348011612892151,
+      "learning_rate": 7.411764705882354e-05,
+      "loss": 1.483832550048828,
+      "mean_token_accuracy": 0.6794860780239105,
+      "num_tokens": 767982.0,
+      "step": 190
+    },
+    {
+      "entropy": 1.4718186855316162,
+      "epoch": 0.05884083553986467,
+      "grad_norm": 0.3977382779121399,
+      "learning_rate": 7.803921568627451e-05,
+      "loss": 1.4843897819519043,
+      "mean_token_accuracy": 0.6810027420520782,
+      "num_tokens": 808206.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.05884083553986467,
+      "eval_entropy": 1.4348028574402876,
+      "eval_loss": 1.4701354503631592,
+      "eval_mean_token_accuracy": 0.6813244935408664,
+      "eval_num_tokens": 808206.0,
+      "eval_runtime": 116.9082,
+      "eval_samples_per_second": 26.046,
+      "eval_steps_per_second": 3.259,
+      "step": 200
+    },
+    {
+      "entropy": 1.4585287928581239,
+      "epoch": 0.0617828773168579,
+      "grad_norm": 0.41666489839553833,
+      "learning_rate": 8.196078431372549e-05,
+      "loss": 1.4701197624206543,
+      "mean_token_accuracy": 0.6891506373882293,
+      "num_tokens": 848584.0,
+      "step": 210
+    },
+    {
+      "entropy": 1.404802179336548,
+      "epoch": 0.06472491909385113,
+      "grad_norm": 0.4331190288066864,
+      "learning_rate": 8.588235294117646e-05,
+      "loss": 1.4301957130432128,
+      "mean_token_accuracy": 0.6959148466587066,
+      "num_tokens": 889067.0,
+      "step": 220
+    },
+    {
+      "entropy": 1.419868004322052,
+      "epoch": 0.06766696087084437,
+      "grad_norm": 0.4185863435268402,
+      "learning_rate": 8.980392156862746e-05,
+      "loss": 1.4252424240112305,
+      "mean_token_accuracy": 0.6949863255023956,
+      "num_tokens": 929411.0,
+      "step": 230
+    },
+    {
+      "entropy": 1.4240824103355407,
+      "epoch": 0.0706090026478376,
+      "grad_norm": 0.43419042229652405,
+      "learning_rate": 9.372549019607843e-05,
+      "loss": 1.4376652717590332,
+      "mean_token_accuracy": 0.6870559632778168,
+      "num_tokens": 970055.0,
+      "step": 240
+    },
+    {
+      "entropy": 1.4307694911956788,
+      "epoch": 0.07355104442483083,
+      "grad_norm": 0.3987244963645935,
+      "learning_rate": 9.764705882352942e-05,
+      "loss": 1.4383943557739258,
+      "mean_token_accuracy": 0.685227632522583,
+      "num_tokens": 1010437.0,
+      "step": 250
+    },
+    {
+      "entropy": 1.50078626871109,
+      "epoch": 0.07649308620182406,
+      "grad_norm": 0.3530856668949127,
+      "learning_rate": 0.00010156862745098039,
+      "loss": 1.5160396575927735,
+      "mean_token_accuracy": 0.6708036184310913,
+      "num_tokens": 1051087.0,
+      "step": 260
+    },
+    {
+      "entropy": 1.4214369535446167,
+      "epoch": 0.0794351279788173,
+      "grad_norm": 0.35215267539024353,
+      "learning_rate": 0.00010549019607843139,
+      "loss": 1.4516315460205078,
+      "mean_token_accuracy": 0.6868359208106994,
+      "num_tokens": 1091565.0,
+      "step": 270
+    },
+    {
+      "entropy": 1.4430809259414672,
+      "epoch": 0.08237716975581054,
+      "grad_norm": 0.3903290629386902,
+      "learning_rate": 0.00010941176470588237,
+      "loss": 1.4475218772888183,
+      "mean_token_accuracy": 0.686697655916214,
+      "num_tokens": 1131001.0,
+      "step": 280
+    },
+    {
+      "entropy": 1.4334508419036864,
+      "epoch": 0.08531921153280377,
+      "grad_norm": 0.3959214985370636,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 1.4455389976501465,
+      "mean_token_accuracy": 0.6854739308357238,
+      "num_tokens": 1171541.0,
+      "step": 290
+    },
+    {
+      "entropy": 1.5002384662628174,
+      "epoch": 0.088261253309797,
+      "grad_norm": 0.3723829984664917,
+      "learning_rate": 0.00011725490196078433,
+      "loss": 1.5168241500854491,
+      "mean_token_accuracy": 0.66768679022789,
+      "num_tokens": 1211498.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.088261253309797,
+      "eval_entropy": 1.4693221822185467,
+      "eval_loss": 1.448572039604187,
+      "eval_mean_token_accuracy": 0.6847088332251301,
+      "eval_num_tokens": 1211498.0,
+      "eval_runtime": 116.8305,
+      "eval_samples_per_second": 26.063,
+      "eval_steps_per_second": 3.261,
+      "step": 300
+    },
+    {
+      "entropy": 1.3928457140922545,
+      "epoch": 0.09120329508679023,
+      "grad_norm": 0.4283956289291382,
+      "learning_rate": 0.0001211764705882353,
+      "loss": 1.4159896850585938,
+      "mean_token_accuracy": 0.6940667152404785,
+      "num_tokens": 1251562.0,
+      "step": 310
+    },
+    {
+      "entropy": 1.3655309796333313,
+      "epoch": 0.09414533686378347,
+      "grad_norm": 0.34452083706855774,
+      "learning_rate": 0.00012509803921568628,
+      "loss": 1.3777886390686036,
+      "mean_token_accuracy": 0.7042169332504272,
+      "num_tokens": 1292029.0,
+      "step": 320
+    },
+    {
+      "entropy": 1.4647573351860046,
+      "epoch": 0.0970873786407767,
+      "grad_norm": 0.3209940493106842,
+      "learning_rate": 0.00012901960784313728,
+      "loss": 1.4774354934692382,
+      "mean_token_accuracy": 0.6791389882564545,
+      "num_tokens": 1332530.0,
+      "step": 330
+    },
+    {
+      "entropy": 1.4148489713668824,
+      "epoch": 0.10002942041776994,
+      "grad_norm": 0.4031716585159302,
+      "learning_rate": 0.00013294117647058822,
+      "loss": 1.4249235153198243,
+      "mean_token_accuracy": 0.692065304517746,
+      "num_tokens": 1372940.0,
+      "step": 340
+    },
+    {
+      "entropy": 1.3905927181243896,
+      "epoch": 0.10297146219476316,
+      "grad_norm": 0.40045231580734253,
+      "learning_rate": 0.00013686274509803922,
+      "loss": 1.4047340393066405,
+      "mean_token_accuracy": 0.6952064216136933,
+      "num_tokens": 1413313.0,
+      "step": 350
+    },
+    {
+      "entropy": 1.393438732624054,
+      "epoch": 0.1059135039717564,
+      "grad_norm": 0.3352883458137512,
+      "learning_rate": 0.00014078431372549022,
+      "loss": 1.4050509452819824,
+      "mean_token_accuracy": 0.6900130629539489,
+      "num_tokens": 1453217.0,
+      "step": 360
+    },
+    {
+      "entropy": 1.3930254817008971,
+      "epoch": 0.10885554574874963,
+      "grad_norm": 0.31767141819000244,
+      "learning_rate": 0.0001447058823529412,
+      "loss": 1.3919607162475587,
+      "mean_token_accuracy": 0.6927550494670868,
+      "num_tokens": 1493584.0,
+      "step": 370
+    },
+    {
+      "entropy": 1.3573615312576295,
+      "epoch": 0.11179758752574287,
+      "grad_norm": 0.32946810126304626,
+      "learning_rate": 0.00014862745098039216,
+      "loss": 1.3821157455444335,
+      "mean_token_accuracy": 0.6986469984054565,
+      "num_tokens": 1534100.0,
+      "step": 380
+    },
+    {
+      "entropy": 1.4580389261245728,
+      "epoch": 0.1147396293027361,
+      "grad_norm": 0.3781375288963318,
+      "learning_rate": 0.00015254901960784313,
+      "loss": 1.4741416931152345,
+      "mean_token_accuracy": 0.6809489071369171,
+      "num_tokens": 1574623.0,
+      "step": 390
+    },
+    {
+      "entropy": 1.4030352354049682,
+      "epoch": 0.11768167107972934,
+      "grad_norm": 0.3613905906677246,
+      "learning_rate": 0.00015647058823529413,
+      "loss": 1.4134941101074219,
+      "mean_token_accuracy": 0.6923583388328552,
+      "num_tokens": 1615045.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.11768167107972934,
+      "eval_entropy": 1.3998811366989856,
+      "eval_loss": 1.432859182357788,
+      "eval_mean_token_accuracy": 0.6876164205430999,
+      "eval_num_tokens": 1615045.0,
+      "eval_runtime": 116.8804,
+      "eval_samples_per_second": 26.052,
+      "eval_steps_per_second": 3.26,
+      "step": 400
+    },
+    {
+      "entropy": 1.3861081957817079,
+      "epoch": 0.12062371285672256,
+      "grad_norm": 0.32785341143608093,
+      "learning_rate": 0.0001603921568627451,
+      "loss": 1.4069743156433105,
+      "mean_token_accuracy": 0.6965118229389191,
+      "num_tokens": 1655438.0,
+      "step": 410
+    },
+    {
+      "entropy": 1.3972581624984741,
+      "epoch": 0.1235657546337158,
+      "grad_norm": 0.3461436331272125,
+      "learning_rate": 0.00016431372549019607,
+      "loss": 1.4036929130554199,
+      "mean_token_accuracy": 0.6954509198665619,
+      "num_tokens": 1696005.0,
+      "step": 420
+    },
+    {
+      "entropy": 1.3920337677001953,
+      "epoch": 0.12650779641070903,
+      "grad_norm": 0.330691933631897,
+      "learning_rate": 0.00016823529411764707,
+      "loss": 1.4209896087646485,
+      "mean_token_accuracy": 0.6902368903160095,
+      "num_tokens": 1736450.0,
+      "step": 430
+    },
+    {
+      "entropy": 1.447186005115509,
+      "epoch": 0.12944983818770225,
+      "grad_norm": 0.3359420895576477,
+      "learning_rate": 0.00017215686274509807,
+      "loss": 1.4491106986999511,
+      "mean_token_accuracy": 0.6825862407684327,
+      "num_tokens": 1776731.0,
+      "step": 440
+    },
+    {
+      "entropy": 1.3688698172569276,
+      "epoch": 0.1323918799646955,
+      "grad_norm": 0.31416866183280945,
+      "learning_rate": 0.000176078431372549,
+      "loss": 1.3883934020996094,
+      "mean_token_accuracy": 0.6974851131439209,
+      "num_tokens": 1816994.0,
+      "step": 450
+    },
+    {
+      "entropy": 1.4649083375930787,
+      "epoch": 0.13533392174168873,
+      "grad_norm": 0.3449016511440277,
+      "learning_rate": 0.00018,
+      "loss": 1.4666101455688476,
+      "mean_token_accuracy": 0.6721475541591644,
+      "num_tokens": 1857622.0,
+      "step": 460
+    },
+    {
+      "entropy": 1.394101870059967,
+      "epoch": 0.13827596351868196,
+      "grad_norm": 0.32241928577423096,
+      "learning_rate": 0.00018392156862745098,
+      "loss": 1.4229127883911132,
+      "mean_token_accuracy": 0.6893933176994324,
+      "num_tokens": 1898085.0,
+      "step": 470
+    },
+    {
+      "entropy": 1.4726597905158996,
+      "epoch": 0.1412180052956752,
+      "grad_norm": 0.3263033926486969,
+      "learning_rate": 0.00018784313725490198,
+      "loss": 1.495950412750244,
+      "mean_token_accuracy": 0.6736144661903382,
+      "num_tokens": 1938687.0,
+      "step": 480
+    },
+    {
+      "entropy": 1.4047853350639343,
+      "epoch": 0.14416004707266844,
+      "grad_norm": 0.32949718832969666,
+      "learning_rate": 0.00019176470588235295,
+      "loss": 1.4045245170593261,
+      "mean_token_accuracy": 0.6931259393692016,
+      "num_tokens": 1978716.0,
+      "step": 490
+    },
+    {
+      "entropy": 1.4191704154014588,
+      "epoch": 0.14710208884966167,
+      "grad_norm": 0.31867218017578125,
+      "learning_rate": 0.00019568627450980392,
+      "loss": 1.4353232383728027,
+      "mean_token_accuracy": 0.68796147108078,
+      "num_tokens": 2019187.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.14710208884966167,
+      "eval_entropy": 1.3842839742582926,
+      "eval_loss": 1.4221439361572266,
+      "eval_mean_token_accuracy": 0.6893131504534423,
+      "eval_num_tokens": 2019187.0,
+      "eval_runtime": 116.869,
+      "eval_samples_per_second": 26.055,
+      "eval_steps_per_second": 3.26,
+      "step": 500
+    },
+    {
+      "entropy": 1.4995567321777343,
+      "epoch": 0.1500441306266549,
+      "grad_norm": 0.32378971576690674,
+      "learning_rate": 0.00019960784313725492,
+      "loss": 1.5118988037109375,
+      "mean_token_accuracy": 0.6660271644592285,
+      "num_tokens": 2059838.0,
+      "step": 510
+    },
+    {
+      "entropy": 1.3857444763183593,
+      "epoch": 0.15298617240364812,
+      "grad_norm": 0.32856202125549316,
+      "learning_rate": 0.00019999957403309267,
+      "loss": 1.3925944328308106,
+      "mean_token_accuracy": 0.6954998314380646,
+      "num_tokens": 2099887.0,
+      "step": 520
+    },
+    {
+      "entropy": 1.444947075843811,
+      "epoch": 0.15592821418064137,
+      "grad_norm": 0.34244897961616516,
+      "learning_rate": 0.00019999810155955347,
+      "loss": 1.4644223213195802,
+      "mean_token_accuracy": 0.6775131762027741,
+      "num_tokens": 2139863.0,
+      "step": 530
+    },
+    {
+      "entropy": 1.4525998115539551,
+      "epoch": 0.1588702559576346,
+      "grad_norm": 0.3342621624469757,
+      "learning_rate": 0.00019999557733601503,
+      "loss": 1.4777566909790039,
+      "mean_token_accuracy": 0.6743626773357392,
+      "num_tokens": 2180117.0,
+      "step": 540
+    },
+    {
+      "entropy": 1.3970038414001464,
+      "epoch": 0.16181229773462782,
+      "grad_norm": 0.33053645491600037,
+      "learning_rate": 0.00019999200138902642,
+      "loss": 1.4096016883850098,
+      "mean_token_accuracy": 0.6931852400302887,
+      "num_tokens": 2219829.0,
+      "step": 550
+    },
+    {
+      "entropy": 1.4066221356391906,
+      "epoch": 0.16475433951162108,
+      "grad_norm": 0.33736753463745117,
+      "learning_rate": 0.00019998737375619835,
+      "loss": 1.4260540008544922,
+      "mean_token_accuracy": 0.6908943474292755,
+      "num_tokens": 2260238.0,
+      "step": 560
+    },
+    {
+      "entropy": 1.3926284670829774,
+      "epoch": 0.1676963812886143,
+      "grad_norm": 0.3630967140197754,
+      "learning_rate": 0.0001999816944862029,
+      "loss": 1.408204936981201,
+      "mean_token_accuracy": 0.6962983906269073,
+      "num_tokens": 2300497.0,
+      "step": 570
+    },
+    {
+      "entropy": 1.3583880305290221,
+      "epoch": 0.17063842306560753,
+      "grad_norm": 0.3344590365886688,
+      "learning_rate": 0.0001999749636387729,
+      "loss": 1.3612471580505372,
+      "mean_token_accuracy": 0.7032208800315857,
+      "num_tokens": 2340896.0,
+      "step": 580
+    },
+    {
+      "entropy": 1.3552134156227111,
+      "epoch": 0.17358046484260076,
+      "grad_norm": 0.3601242005825043,
+      "learning_rate": 0.00019996718128470137,
+      "loss": 1.378493595123291,
+      "mean_token_accuracy": 0.6997579276561737,
+      "num_tokens": 2381277.0,
+      "step": 590
+    },
+    {
+      "entropy": 1.4107160449028016,
+      "epoch": 0.176522506619594,
+      "grad_norm": 0.38008254766464233,
+      "learning_rate": 0.00019995834750584078,
+      "loss": 1.4177864074707032,
+      "mean_token_accuracy": 0.6893383502960205,
+      "num_tokens": 2421617.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.176522506619594,
+      "eval_entropy": 1.3129558118935332,
+      "eval_loss": 1.4119162559509277,
+      "eval_mean_token_accuracy": 0.6910557604524408,
+      "eval_num_tokens": 2421617.0,
+      "eval_runtime": 117.0182,
+      "eval_samples_per_second": 26.022,
+      "eval_steps_per_second": 3.256,
+      "step": 600
+    },
+    {
+      "entropy": 1.3507792115211488,
+      "epoch": 0.17946454839658724,
+      "grad_norm": 0.3318082392215729,
+      "learning_rate": 0.00019994846239510216,
+      "loss": 1.3647551536560059,
+      "mean_token_accuracy": 0.7037387728691101,
+      "num_tokens": 2462059.0,
+      "step": 610
+    },
+    {
+      "entropy": 1.395737397670746,
+      "epoch": 0.18240659017358046,
+      "grad_norm": 0.34794482588768005,
+      "learning_rate": 0.00019993752605645417,
+      "loss": 1.4275907516479491,
+      "mean_token_accuracy": 0.6850803017616272,
+      "num_tokens": 2502114.0,
+      "step": 620
+    },
+    {
+      "entropy": 1.4139577507972718,
+      "epoch": 0.1853486319505737,
+      "grad_norm": 0.3424963653087616,
+      "learning_rate": 0.00019992553860492191,
+      "loss": 1.4030399322509766,
+      "mean_token_accuracy": 0.6936035215854645,
+      "num_tokens": 2542391.0,
+      "step": 630
+    },
+    {
+      "entropy": 1.3600899696350097,
+      "epoch": 0.18829067372756694,
+      "grad_norm": 0.4584237039089203,
+      "learning_rate": 0.00019991250016658578,
+      "loss": 1.3970141410827637,
+      "mean_token_accuracy": 0.6953531980514527,
+      "num_tokens": 2582892.0,
+      "step": 640
+    },
+    {
+      "entropy": 1.4006186485290528,
+      "epoch": 0.19123271550456017,
+      "grad_norm": 0.3958089053630829,
+      "learning_rate": 0.00019989841087858019,
+      "loss": 1.4342127799987794,
+      "mean_token_accuracy": 0.6836777746677398,
+      "num_tokens": 2623322.0,
+      "step": 650
+    },
+    {
+      "entropy": 1.4262026309967042,
+      "epoch": 0.1941747572815534,
+      "grad_norm": 0.38685840368270874,
+      "learning_rate": 0.000199883270889092,
+      "loss": 1.4162178993225099,
+      "mean_token_accuracy": 0.6810121357440948,
+      "num_tokens": 2664008.0,
+      "step": 660
+    },
+    {
+      "entropy": 1.3658849358558656,
+      "epoch": 0.19711679905854662,
+      "grad_norm": 0.37710094451904297,
+      "learning_rate": 0.00019986708035735914,
+      "loss": 1.390056037902832,
+      "mean_token_accuracy": 0.6923020124435425,
+      "num_tokens": 2704594.0,
+      "step": 670
+    },
+    {
+      "entropy": 1.363602066040039,
+      "epoch": 0.20005884083553988,
+      "grad_norm": 0.32848283648490906,
+      "learning_rate": 0.0001998498394536687,
+      "loss": 1.4025785446166992,
+      "mean_token_accuracy": 0.6898211300373077,
+      "num_tokens": 2744928.0,
+      "step": 680
+    },
+    {
+      "entropy": 1.2932772994041444,
+      "epoch": 0.2030008826125331,
+      "grad_norm": 0.375255286693573,
+      "learning_rate": 0.00019983154835935535,
+      "loss": 1.271113681793213,
+      "mean_token_accuracy": 0.72461918592453,
+      "num_tokens": 2785405.0,
+      "step": 690
+    },
+    {
+      "entropy": 1.31307555437088,
+      "epoch": 0.20594292438952633,
+      "grad_norm": 0.3716314733028412,
+      "learning_rate": 0.0001998122072667993,
+      "loss": 1.3366676330566407,
+      "mean_token_accuracy": 0.7082294344902038,
+      "num_tokens": 2825853.0,
+      "step": 700
+    },
+    {
+      "epoch": 0.20594292438952633,
+      "eval_entropy": 1.43291619208854,
+      "eval_loss": 1.3930107355117798,
+      "eval_mean_token_accuracy": 0.6941823344531022,
+      "eval_num_tokens": 2825853.0,
+      "eval_runtime": 117.0332,
+      "eval_samples_per_second": 26.018,
+      "eval_steps_per_second": 3.255,
+      "step": 700
+    },
+    {
+      "entropy": 1.3790567636489868,
+      "epoch": 0.20888496616651955,
+      "grad_norm": 0.3361744284629822,
+      "learning_rate": 0.0001997918163794244,
+      "loss": 1.4195181846618652,
+      "mean_token_accuracy": 0.6918542444705963,
+      "num_tokens": 2866199.0,
+      "step": 710
+    },
+    {
+      "entropy": 1.4847285509109498,
+      "epoch": 0.2118270079435128,
+      "grad_norm": 0.5025491118431091,
+      "learning_rate": 0.00019977037591169583,
+      "loss": 1.479151153564453,
+      "mean_token_accuracy": 0.6787452459335327,
+      "num_tokens": 2906632.0,
+      "step": 720
+    },
+    {
+      "entropy": 1.3509209036827088,
+      "epoch": 0.21476904972050603,
+      "grad_norm": 0.3572346866130829,
+      "learning_rate": 0.00019974788608911802,
+      "loss": 1.3688506126403808,
+      "mean_token_accuracy": 0.6984890639781952,
+      "num_tokens": 2947203.0,
+      "step": 730
+    },
+    {
+      "entropy": 1.3631734371185302,
+      "epoch": 0.21771109149749926,
+      "grad_norm": 0.33028289675712585,
+      "learning_rate": 0.00019972434714823217,
+      "loss": 1.389684009552002,
+      "mean_token_accuracy": 0.7006505787372589,
+      "num_tokens": 2987399.0,
+      "step": 740
+    },
+    {
+      "entropy": 1.3604696154594422,
+      "epoch": 0.22065313327449249,
+      "grad_norm": 0.368522584438324,
+      "learning_rate": 0.00019969975933661378,
+      "loss": 1.376063919067383,
+      "mean_token_accuracy": 0.6953622698783875,
+      "num_tokens": 3027379.0,
+      "step": 750
+    },
+    {
+      "entropy": 1.4122770190238954,
+      "epoch": 0.22359517505148574,
+      "grad_norm": 0.3348753750324249,
+      "learning_rate": 0.00019967412291287007,
+      "loss": 1.406267261505127,
+      "mean_token_accuracy": 0.6962361812591553,
+      "num_tokens": 3067608.0,
+      "step": 760
+    },
+    {
+      "entropy": 1.357818615436554,
+      "epoch": 0.22653721682847897,
+      "grad_norm": 0.34611567854881287,
+      "learning_rate": 0.00019964743814663725,
+      "loss": 1.3942573547363282,
+      "mean_token_accuracy": 0.6927765250205994,
+      "num_tokens": 3108050.0,
+      "step": 770
+    },
+    {
+      "entropy": 1.3579653978347779,
+      "epoch": 0.2294792586054722,
+      "grad_norm": 0.42795485258102417,
+      "learning_rate": 0.0001996197053185777,
+      "loss": 1.369248104095459,
+      "mean_token_accuracy": 0.70080486536026,
+      "num_tokens": 3148448.0,
+      "step": 780
+    },
+    {
+      "entropy": 1.3725898623466493,
+      "epoch": 0.23242130038246542,
+      "grad_norm": 0.37439125776290894,
+      "learning_rate": 0.00019959092472037688,
+      "loss": 1.3844552040100098,
+      "mean_token_accuracy": 0.6901408195495605,
+      "num_tokens": 3189104.0,
+      "step": 790
+    },
+    {
+      "entropy": 1.3633452892303466,
+      "epoch": 0.23536334215945867,
+      "grad_norm": 0.3426375091075897,
+      "learning_rate": 0.0001995610966547406,
+      "loss": 1.3828603744506835,
+      "mean_token_accuracy": 0.699894517660141,
+      "num_tokens": 3229594.0,
+      "step": 800
+    },
+    {
+      "epoch": 0.23536334215945867,
+      "eval_entropy": 1.3796262137220288,
+      "eval_loss": 1.3804020881652832,
+      "eval_mean_token_accuracy": 0.6970919832157025,
+      "eval_num_tokens": 3229594.0,
+      "eval_runtime": 117.0459,
+      "eval_samples_per_second": 26.015,
+      "eval_steps_per_second": 3.255,
+      "step": 800
+    },
+    {
+      "entropy": 1.3366303324699402,
+      "epoch": 0.2383053839364519,
+      "grad_norm": 0.31953537464141846,
+      "learning_rate": 0.00019953022143539144,
+      "loss": 1.3653806686401366,
+      "mean_token_accuracy": 0.7049518287181854,
+      "num_tokens": 3269982.0,
+      "step": 810
+    },
+    {
+      "entropy": 1.3933040380477906,
+      "epoch": 0.24124742571344512,
+      "grad_norm": 0.39059045910835266,
+      "learning_rate": 0.00019949829938706567,
+      "loss": 1.4077239990234376,
+      "mean_token_accuracy": 0.6878176212310791,
+      "num_tokens": 3310501.0,
+      "step": 820
+    },
+    {
+      "entropy": 1.4102402210235596,
+      "epoch": 0.24418946749043838,
+      "grad_norm": 0.40408599376678467,
+      "learning_rate": 0.00019946533084550983,
+      "loss": 1.42384614944458,
+      "mean_token_accuracy": 0.6878543615341186,
+      "num_tokens": 3350905.0,
+      "step": 830
+    },
+    {
+      "entropy": 1.3664647340774536,
+      "epoch": 0.2471315092674316,
+      "grad_norm": 0.33828359842300415,
+      "learning_rate": 0.00019943131615747715,
+      "loss": 1.3716326713562013,
+      "mean_token_accuracy": 0.7006340861320496,
+      "num_tokens": 3391405.0,
+      "step": 840
+    },
+    {
+      "entropy": 1.427152693271637,
+      "epoch": 0.25007355104442486,
+      "grad_norm": 0.48738211393356323,
+      "learning_rate": 0.00019939625568072387,
+      "loss": 1.4487317085266114,
+      "mean_token_accuracy": 0.6779101848602295,
+      "num_tokens": 3431764.0,
+      "step": 850
+    },
+    {
+      "entropy": 1.3256952285766601,
+      "epoch": 0.25301559282141806,
+      "grad_norm": 0.4037957191467285,
+      "learning_rate": 0.00019936014978400558,
+      "loss": 1.341677474975586,
+      "mean_token_accuracy": 0.7102087676525116,
+      "num_tokens": 3472134.0,
+      "step": 860
+    },
+    {
+      "entropy": 1.3219331979751587,
+      "epoch": 0.2559576345984113,
+      "grad_norm": 0.3991295099258423,
+      "learning_rate": 0.00019932299884707324,
+      "loss": 1.3323281288146973,
+      "mean_token_accuracy": 0.7040995419025421,
+      "num_tokens": 3512618.0,
+      "step": 870
+    },
+    {
+      "entropy": 1.327842366695404,
+      "epoch": 0.2588996763754045,
+      "grad_norm": 0.37586092948913574,
+      "learning_rate": 0.00019928480326066925,
+      "loss": 1.3401626586914062,
+      "mean_token_accuracy": 0.7077195107936859,
+      "num_tokens": 3553039.0,
+      "step": 880
+    },
+    {
+      "entropy": 1.4176750898361206,
+      "epoch": 0.26184171815239776,
+      "grad_norm": 0.3479557931423187,
+      "learning_rate": 0.00019924556342652334,
+      "loss": 1.435785961151123,
+      "mean_token_accuracy": 0.6839641869068146,
+      "num_tokens": 3593551.0,
+      "step": 890
+    },
+    {
+      "entropy": 1.3467723608016968,
+      "epoch": 0.264783759929391,
+      "grad_norm": 0.3519476056098938,
+      "learning_rate": 0.00019920527975734827,
+      "loss": 1.3625640869140625,
+      "mean_token_accuracy": 0.6986214816570282,
+      "num_tokens": 3633984.0,
+      "step": 900
+    },
+    {
+      "epoch": 0.264783759929391,
+      "eval_entropy": 1.327493928861743,
+      "eval_loss": 1.3682215213775635,
+      "eval_mean_token_accuracy": 0.6995677048452883,
+      "eval_num_tokens": 3633984.0,
+      "eval_runtime": 117.0553,
+      "eval_samples_per_second": 26.013,
+      "eval_steps_per_second": 3.255,
+      "step": 900
+    },
+    {
+      "entropy": 1.377519929409027,
+      "epoch": 0.2677258017063842,
+      "grad_norm": 0.3842776417732239,
+      "learning_rate": 0.0001991639526768356,
+      "loss": 1.4039417266845704,
+      "mean_token_accuracy": 0.6875424087047577,
+      "num_tokens": 3674545.0,
+      "step": 910
+    },
+    {
+      "entropy": 1.314482867717743,
+      "epoch": 0.27066784348337747,
+      "grad_norm": 0.4236571192741394,
+      "learning_rate": 0.00019912158261965114,
+      "loss": 1.3185301780700684,
+      "mean_token_accuracy": 0.7092587351799011,
+      "num_tokens": 3714993.0,
+      "step": 920
+    },
+    {
+      "entropy": 1.3224342346191407,
+      "epoch": 0.2736098852603707,
+      "grad_norm": 0.3825512230396271,
+      "learning_rate": 0.0001990781700314304,
+      "loss": 1.3342713356018066,
+      "mean_token_accuracy": 0.7061914503574371,
+      "num_tokens": 3755443.0,
+      "step": 930
+    },
+    {
+      "entropy": 1.3411121726036073,
+      "epoch": 0.2765519270373639,
+      "grad_norm": 0.3913728594779968,
+      "learning_rate": 0.00019903371536877397,
+      "loss": 1.372488784790039,
+      "mean_token_accuracy": 0.6965407311916352,
+      "num_tokens": 3795944.0,
+      "step": 940
+    },
+    {
+      "entropy": 1.3303414344787599,
+      "epoch": 0.2794939688143572,
+      "grad_norm": 0.34784528613090515,
+      "learning_rate": 0.00019898821909924267,
+      "loss": 1.329643440246582,
+      "mean_token_accuracy": 0.7043894708156586,
+      "num_tokens": 3836572.0,
+      "step": 950
+    },
+    {
+      "entropy": 1.3415798902511598,
+      "epoch": 0.2824360105913504,
+      "grad_norm": 0.3249607980251312,
+      "learning_rate": 0.00019894168170135252,
+      "loss": 1.3672316551208497,
+      "mean_token_accuracy": 0.7032464861869812,
+      "num_tokens": 3876943.0,
+      "step": 960
+    },
+    {
+      "entropy": 1.3220824122428894,
+      "epoch": 0.2853780523683436,
+      "grad_norm": 0.3845159411430359,
+      "learning_rate": 0.00019889410366456995,
+      "loss": 1.326685905456543,
+      "mean_token_accuracy": 0.7063803553581238,
+      "num_tokens": 3917514.0,
+      "step": 970
+    },
+    {
+      "entropy": 1.3524356603622436,
+      "epoch": 0.2883200941453369,
+      "grad_norm": 0.49597394466400146,
+      "learning_rate": 0.00019884548548930648,
+      "loss": 1.3823152542114259,
+      "mean_token_accuracy": 0.688988733291626,
+      "num_tokens": 3958016.0,
+      "step": 980
+    },
+    {
+      "entropy": 1.3366973161697389,
+      "epoch": 0.2912621359223301,
+      "grad_norm": 0.3381548523902893,
+      "learning_rate": 0.00019879582768691343,
+      "loss": 1.3448416709899902,
+      "mean_token_accuracy": 0.7067325949668884,
+      "num_tokens": 3998636.0,
+      "step": 990
+    },
+    {
+      "entropy": 1.342777681350708,
+      "epoch": 0.29420417769932333,
+      "grad_norm": 0.3816724717617035,
+      "learning_rate": 0.0001987451307796767,
+      "loss": 1.3505562782287597,
+      "mean_token_accuracy": 0.7011382758617402,
+      "num_tokens": 4039116.0,
+      "step": 1000
+    },
+    {
+      "epoch": 0.29420417769932333,
+      "eval_entropy": 1.3210531396502898,
+      "eval_loss": 1.3546180725097656,
+      "eval_mean_token_accuracy": 0.7016779092040275,
+      "eval_num_tokens": 4039116.0,
+      "eval_runtime": 116.9387,
+      "eval_samples_per_second": 26.039,
+      "eval_steps_per_second": 3.258,
+      "step": 1000
+    },
+    {
+      "entropy": 1.366915476322174,
+      "epoch": 0.2971462194763166,
+      "grad_norm": 0.35182058811187744,
+      "learning_rate": 0.00019869339530081105,
+      "loss": 1.3859369277954101,
+      "mean_token_accuracy": 0.6906876146793366,
+      "num_tokens": 4079633.0,
+      "step": 1010
+    },
+    {
+      "entropy": 1.3304717183113097,
+      "epoch": 0.3000882612533098,
+      "grad_norm": 0.3552508056163788,
+      "learning_rate": 0.0001986406217944548,
+      "loss": 1.3530879974365235,
+      "mean_token_accuracy": 0.703288596868515,
+      "num_tokens": 4120199.0,
+      "step": 1020
+    },
+    {
+      "entropy": 1.3517386436462402,
+      "epoch": 0.30303030303030304,
+      "grad_norm": 0.3985072076320648,
+      "learning_rate": 0.0001985868108156638,
+      "loss": 1.3472537040710448,
+      "mean_token_accuracy": 0.7062719881534576,
+      "num_tokens": 4160213.0,
+      "step": 1030
+    },
+    {
+      "entropy": 1.2475824117660523,
+      "epoch": 0.30597234480729624,
+      "grad_norm": 0.35150817036628723,
+      "learning_rate": 0.00019853196293040577,
+      "loss": 1.2721343040466309,
+      "mean_token_accuracy": 0.7261709928512573,
+      "num_tokens": 4200515.0,
+      "step": 1040
+    },
+    {
+      "entropy": 1.3748682379722594,
+      "epoch": 0.3089143865842895,
+      "grad_norm": 0.3976305425167084,
+      "learning_rate": 0.00019847607871555426,
+      "loss": 1.3997004508972168,
+      "mean_token_accuracy": 0.6935756504535675,
+      "num_tokens": 4240452.0,
+      "step": 1050
+    },
+    {
+      "entropy": 1.250860321521759,
+      "epoch": 0.31185642836128274,
+      "grad_norm": 0.4553461968898773,
+      "learning_rate": 0.00019841915875888272,
+      "loss": 1.2498172760009765,
+      "mean_token_accuracy": 0.7244091928005219,
+      "num_tokens": 4280962.0,
+      "step": 1060
+    },
+    {
+      "entropy": 1.3760404348373414,
+      "epoch": 0.31479847013827594,
+      "grad_norm": 0.489533394575119,
+      "learning_rate": 0.00019836120365905813,
+      "loss": 1.3888651847839355,
+      "mean_token_accuracy": 0.693582957983017,
+      "num_tokens": 4321434.0,
+      "step": 1070
+    },
+    {
+      "entropy": 1.3051153898239136,
+      "epoch": 0.3177405119152692,
+      "grad_norm": 0.3459921181201935,
+      "learning_rate": 0.0001983022140256348,
+      "loss": 1.3199047088623046,
+      "mean_token_accuracy": 0.7050609290599823,
+      "num_tokens": 4361714.0,
+      "step": 1080
+    },
+    {
+      "entropy": 1.2985354661941528,
+      "epoch": 0.32068255369226245,
+      "grad_norm": 0.33598437905311584,
+      "learning_rate": 0.00019824219047904804,
+      "loss": 1.3020724296569823,
+      "mean_token_accuracy": 0.713257223367691,
+      "num_tokens": 4401949.0,
+      "step": 1090
+    },
+    {
+      "entropy": 1.276042139530182,
+      "epoch": 0.32362459546925565,
+      "grad_norm": 0.43967100977897644,
+      "learning_rate": 0.00019818113365060742,
+      "loss": 1.3166107177734374,
+      "mean_token_accuracy": 0.7159667015075684,
+      "num_tokens": 4442174.0,
+      "step": 1100
+    },
+    {
+      "epoch": 0.32362459546925565,
+      "eval_entropy": 1.3324664305514238,
+      "eval_loss": 1.3422644138336182,
+      "eval_mean_token_accuracy": 0.7045243512301307,
+      "eval_num_tokens": 4442174.0,
+      "eval_runtime": 116.854,
+      "eval_samples_per_second": 26.058,
+      "eval_steps_per_second": 3.26,
+      "step": 1100
+    },
+    {
+      "entropy": 1.3644051671028137,
+      "epoch": 0.3265666372462489,
+      "grad_norm": 0.3831937313079834,
+      "learning_rate": 0.0001981190441824903,
+      "loss": 1.3389662742614745,
+      "mean_token_accuracy": 0.7077198505401612,
+      "num_tokens": 4482638.0,
+      "step": 1110
+    },
+    {
+      "entropy": 1.3949143767356873,
+      "epoch": 0.32950867902324216,
+      "grad_norm": 0.4152087867259979,
+      "learning_rate": 0.0001980559227277352,
+      "loss": 1.4393955230712892,
+      "mean_token_accuracy": 0.6806177437305451,
+      "num_tokens": 4523231.0,
+      "step": 1120
+    },
+    {
+      "entropy": 1.5005442142486571,
+      "epoch": 0.33245072080023536,
+      "grad_norm": 0.3600977957248688,
+      "learning_rate": 0.00019799176995023446,
+      "loss": 1.5087374687194823,
+      "mean_token_accuracy": 0.6630822360515595,
+      "num_tokens": 4563875.0,
+      "step": 1130
+    },
+    {
+      "entropy": 1.3872592091560363,
+      "epoch": 0.3353927625772286,
+      "grad_norm": 0.41946983337402344,
+      "learning_rate": 0.00019792658652472784,
+      "loss": 1.3950308799743651,
+      "mean_token_accuracy": 0.6971809685230255,
+      "num_tokens": 4604493.0,
+      "step": 1140
+    },
+    {
+      "entropy": 1.2849906086921692,
+      "epoch": 0.3383348043542218,
+      "grad_norm": 0.44172239303588867,
+      "learning_rate": 0.00019786037313679496,
+      "loss": 1.3040260314941405,
+      "mean_token_accuracy": 0.7124627947807312,
+      "num_tokens": 4644971.0,
+      "step": 1150
+    },
+    {
+      "entropy": 1.3872546672821044,
+      "epoch": 0.34127684613121506,
+      "grad_norm": 0.42291128635406494,
+      "learning_rate": 0.0001977931304828484,
+      "loss": 1.3932037353515625,
+      "mean_token_accuracy": 0.6873931646347046,
+      "num_tokens": 4685402.0,
+      "step": 1160
+    },
+    {
+      "entropy": 1.3018505930900575,
+      "epoch": 0.3442188879082083,
+      "grad_norm": 0.40709924697875977,
+      "learning_rate": 0.00019772485927012617,
+      "loss": 1.3295734405517579,
+      "mean_token_accuracy": 0.708010071516037,
+      "num_tokens": 4725827.0,
+      "step": 1170
+    },
+    {
+      "entropy": 1.306389570236206,
+      "epoch": 0.3471609296852015,
+      "grad_norm": 0.3591320514678955,
+      "learning_rate": 0.00019765556021668438,
+      "loss": 1.2882349014282226,
+      "mean_token_accuracy": 0.7196535110473633,
+      "num_tokens": 4766299.0,
+      "step": 1180
+    },
+    {
+      "entropy": 1.274339497089386,
+      "epoch": 0.35010297146219477,
+      "grad_norm": 0.49734798073768616,
+      "learning_rate": 0.0001975852340513897,
+      "loss": 1.322108268737793,
+      "mean_token_accuracy": 0.7085251092910767,
+      "num_tokens": 4806716.0,
+      "step": 1190
+    },
+    {
+      "entropy": 1.3069317936897278,
+      "epoch": 0.353045013239188,
+      "grad_norm": 0.38912537693977356,
+      "learning_rate": 0.00019751388151391153,
+      "loss": 1.2966851234436034,
+      "mean_token_accuracy": 0.7151984930038452,
+      "num_tokens": 4846722.0,
+      "step": 1200
+    },
+    {
+      "epoch": 0.353045013239188,
+      "eval_entropy": 1.2881334237852122,
+      "eval_loss": 1.3329390287399292,
+      "eval_mean_token_accuracy": 0.7064482895095204,
+      "eval_num_tokens": 4846722.0,
+      "eval_runtime": 116.8679,
+      "eval_samples_per_second": 26.055,
+      "eval_steps_per_second": 3.26,
+      "step": 1200
+    },
+    {
+      "entropy": 1.323479461669922,
+      "epoch": 0.3559870550161812,
+      "grad_norm": 0.3935072720050812,
+      "learning_rate": 0.0001974415033547145,
+      "loss": 1.3559471130371095,
+      "mean_token_accuracy": 0.6980177283287048,
+      "num_tokens": 4887382.0,
+      "step": 1210
+    },
+    {
+      "entropy": 1.3160091280937194,
+      "epoch": 0.3589290967931745,
+      "grad_norm": 0.402215838432312,
+      "learning_rate": 0.00019736810033505037,
+      "loss": 1.3169782638549805,
+      "mean_token_accuracy": 0.7115081787109375,
+      "num_tokens": 4927725.0,
+      "step": 1220
+    },
+    {
+      "entropy": 1.3298014521598815,
+      "epoch": 0.36187113857016767,
+      "grad_norm": 0.4016437530517578,
+      "learning_rate": 0.00019729367322695,
+      "loss": 1.340937900543213,
+      "mean_token_accuracy": 0.7020221769809722,
+      "num_tokens": 4968297.0,
+      "step": 1230
+    },
+    {
+      "entropy": 1.3313292741775513,
+      "epoch": 0.3648131803471609,
+      "grad_norm": 0.4023797810077667,
+      "learning_rate": 0.00019721822281321537,
+      "loss": 1.3570178031921387,
+      "mean_token_accuracy": 0.695677000284195,
+      "num_tokens": 5008063.0,
+      "step": 1240
+    },
+    {
+      "entropy": 1.3236193537712098,
+      "epoch": 0.3677552221241542,
+      "grad_norm": 0.39722758531570435,
+      "learning_rate": 0.00019714174988741127,
+      "loss": 1.3275541305541991,
+      "mean_token_accuracy": 0.7125855982303619,
+      "num_tokens": 5048336.0,
+      "step": 1250
+    },
+    {
+      "entropy": 1.3190148949623108,
+      "epoch": 0.3706972639011474,
+      "grad_norm": 0.35919976234436035,
+      "learning_rate": 0.000197064255253857,
+      "loss": 1.3260068893432617,
+      "mean_token_accuracy": 0.7106278300285339,
+      "num_tokens": 5088942.0,
+      "step": 1260
+    },
+    {
+      "entropy": 1.3633437037467957,
+      "epoch": 0.37363930567814063,
+      "grad_norm": 0.41243675351142883,
+      "learning_rate": 0.0001969857397276178,
+      "loss": 1.3687942504882813,
+      "mean_token_accuracy": 0.6920778334140778,
+      "num_tokens": 5129372.0,
+      "step": 1270
+    },
+    {
+      "entropy": 1.2383417725563048,
+      "epoch": 0.3765813474551339,
+      "grad_norm": 0.40572160482406616,
+      "learning_rate": 0.00019690620413449642,
+      "loss": 1.2481071472167968,
+      "mean_token_accuracy": 0.7250577747821808,
+      "num_tokens": 5169845.0,
+      "step": 1280
+    },
+    {
+      "entropy": 1.2865507364273072,
+      "epoch": 0.3795233892321271,
+      "grad_norm": 0.42980971932411194,
+      "learning_rate": 0.00019682564931102435,
+      "loss": 1.3222503662109375,
+      "mean_token_accuracy": 0.7078245043754577,
+      "num_tokens": 5210494.0,
+      "step": 1290
+    },
+    {
+      "entropy": 1.3427109718322754,
+      "epoch": 0.38246543100912034,
+      "grad_norm": 0.39987629652023315,
+      "learning_rate": 0.000196744076104453,
+      "loss": 1.3458648681640626,
+      "mean_token_accuracy": 0.7039949476718903,
+      "num_tokens": 5250876.0,
+      "step": 1300
+    },
+    {
+      "epoch": 0.38246543100912034,
+      "eval_entropy": 1.3054086448639397,
+      "eval_loss": 1.328322172164917,
+      "eval_mean_token_accuracy": 0.7073736231470984,
+      "eval_num_tokens": 5250876.0,
+      "eval_runtime": 116.8194,
+      "eval_samples_per_second": 26.066,
+      "eval_steps_per_second": 3.261,
+      "step": 1300
+    },
+    {
+      "entropy": 1.3619997262954713,
+      "epoch": 0.38540747278611354,
+      "grad_norm": 0.39274510741233826,
+      "learning_rate": 0.00019666148537274486,
+      "loss": 1.393809986114502,
+      "mean_token_accuracy": 0.6958752512931824,
+      "num_tokens": 5291408.0,
+      "step": 1310
+    },
+    {
+      "entropy": 1.313580584526062,
+      "epoch": 0.3883495145631068,
+      "grad_norm": 0.4472017288208008,
+      "learning_rate": 0.00019657787798456447,
+      "loss": 1.3207698822021485,
+      "mean_token_accuracy": 0.7056093811988831,
+      "num_tokens": 5331902.0,
+      "step": 1320
+    },
+    {
+      "entropy": 1.3660961747169496,
+      "epoch": 0.39129155634010004,
+      "grad_norm": 0.3704398572444916,
+      "learning_rate": 0.00019649325481926918,
+      "loss": 1.3579423904418946,
+      "mean_token_accuracy": 0.6998582005500793,
+      "num_tokens": 5372344.0,
+      "step": 1330
+    },
+    {
+      "entropy": 1.2838171362876891,
+      "epoch": 0.39423359811709324,
+      "grad_norm": 0.42561763525009155,
+      "learning_rate": 0.0001964076167669001,
+      "loss": 1.3317262649536132,
+      "mean_token_accuracy": 0.707372397184372,
+      "num_tokens": 5412588.0,
+      "step": 1340
+    },
+    {
+      "entropy": 1.3983584880828857,
+      "epoch": 0.3971756398940865,
+      "grad_norm": 0.39428314566612244,
+      "learning_rate": 0.00019632096472817247,
+      "loss": 1.3956350326538085,
+      "mean_token_accuracy": 0.6877002000808716,
+      "num_tokens": 5452540.0,
+      "step": 1350
+    },
+    {
+      "entropy": 1.2444414377212525,
+      "epoch": 0.40011768167107975,
+      "grad_norm": 0.46192431449890137,
+      "learning_rate": 0.00019623329961446646,
+      "loss": 1.2581539154052734,
+      "mean_token_accuracy": 0.7236546576023102,
+      "num_tokens": 5492713.0,
+      "step": 1360
+    },
+    {
+      "entropy": 1.3872360110282898,
+      "epoch": 0.40305972344807295,
+      "grad_norm": 0.4279949367046356,
+      "learning_rate": 0.0001961446223478174,
+      "loss": 1.3998414993286132,
+      "mean_token_accuracy": 0.6933107733726501,
+      "num_tokens": 5532960.0,
+      "step": 1370
+    },
+    {
+      "entropy": 1.3212559223175049,
+      "epoch": 0.4060017652250662,
+      "grad_norm": 0.5924211740493774,
+      "learning_rate": 0.0001960549338609061,
+      "loss": 1.3529240608215332,
+      "mean_token_accuracy": 0.7018321037292481,
+      "num_tokens": 5573284.0,
+      "step": 1380
+    },
+    {
+      "entropy": 1.2787784337997437,
+      "epoch": 0.40894380700205946,
+      "grad_norm": 0.4120098948478699,
+      "learning_rate": 0.00019596423509704916,
+      "loss": 1.2666643142700196,
+      "mean_token_accuracy": 0.7207745730876922,
+      "num_tokens": 5613401.0,
+      "step": 1390
+    },
+    {
+      "entropy": 1.2354493618011475,
+      "epoch": 0.41188584877905265,
+      "grad_norm": 0.3948381245136261,
+      "learning_rate": 0.00019587252701018897,
+      "loss": 1.258436393737793,
+      "mean_token_accuracy": 0.7286782383918762,
+      "num_tokens": 5653527.0,
+      "step": 1400
+    },
+    {
+      "epoch": 0.41188584877905265,
+      "eval_entropy": 1.3250519281297217,
+      "eval_loss": 1.317603588104248,
+      "eval_mean_token_accuracy": 0.709270513902499,
+      "eval_num_tokens": 5653527.0,
+      "eval_runtime": 117.0125,
+      "eval_samples_per_second": 26.023,
+      "eval_steps_per_second": 3.256,
+      "step": 1400
+    },
+    {
+      "entropy": 1.2924874544143676,
+      "epoch": 0.4148278905560459,
+      "grad_norm": 0.3901238739490509,
+      "learning_rate": 0.0001957798105648836,
+      "loss": 1.3112905502319336,
+      "mean_token_accuracy": 0.7130284488201142,
+      "num_tokens": 5693830.0,
+      "step": 1410
+    },
+    {
+      "entropy": 1.3326545000076293,
+      "epoch": 0.4177699323330391,
+      "grad_norm": 0.4452548623085022,
+      "learning_rate": 0.0001956860867362968,
+      "loss": 1.3477538108825684,
+      "mean_token_accuracy": 0.6966541707515717,
+      "num_tokens": 5734423.0,
+      "step": 1420
+    },
+    {
+      "entropy": 1.3726879000663756,
+      "epoch": 0.42071197411003236,
+      "grad_norm": 0.41140350699424744,
+      "learning_rate": 0.00019559135651018764,
+      "loss": 1.364595603942871,
+      "mean_token_accuracy": 0.6968912184238434,
+      "num_tokens": 5774991.0,
+      "step": 1430
+    },
+    {
+      "entropy": 1.2218275964260101,
+      "epoch": 0.4236540158870256,
+      "grad_norm": 0.34981605410575867,
+      "learning_rate": 0.0001954956208829002,
+      "loss": 1.255105686187744,
+      "mean_token_accuracy": 0.7247668862342834,
+      "num_tokens": 5815447.0,
+      "step": 1440
+    },
+    {
+      "entropy": 1.277131152153015,
+      "epoch": 0.4265960576640188,
+      "grad_norm": 0.4556473195552826,
+      "learning_rate": 0.00019539888086135302,
+      "loss": 1.2920147895812988,
+      "mean_token_accuracy": 0.7172133564949036,
+      "num_tokens": 5855958.0,
+      "step": 1450
+    },
+    {
+      "entropy": 1.2765482187271118,
+      "epoch": 0.42953809944101207,
+      "grad_norm": 0.3995361924171448,
+      "learning_rate": 0.00019530113746302864,
+      "loss": 1.2754140853881837,
+      "mean_token_accuracy": 0.7180281519889832,
+      "num_tokens": 5896270.0,
+      "step": 1460
+    },
+    {
+      "entropy": 1.3740562319755554,
+      "epoch": 0.4324801412180053,
+      "grad_norm": 0.42559972405433655,
+      "learning_rate": 0.00019520239171596276,
+      "loss": 1.4036828994750976,
+      "mean_token_accuracy": 0.6909762322902679,
+      "num_tokens": 5936780.0,
+      "step": 1470
+    },
+    {
+      "entropy": 1.3417500495910644,
+      "epoch": 0.4354221829949985,
+      "grad_norm": 0.36076462268829346,
+      "learning_rate": 0.00019510264465873344,
+      "loss": 1.3366243362426757,
+      "mean_token_accuracy": 0.7046165943145752,
+      "num_tokens": 5977191.0,
+      "step": 1480
+    },
+    {
+      "entropy": 1.2735178232192994,
+      "epoch": 0.4383642247719918,
+      "grad_norm": 0.4245460033416748,
+      "learning_rate": 0.0001950018973404503,
+      "loss": 1.296212387084961,
+      "mean_token_accuracy": 0.7112137854099274,
+      "num_tokens": 6017786.0,
+      "step": 1490
+    },
+    {
+      "entropy": 1.2875243425369263,
+      "epoch": 0.44130626654898497,
+      "grad_norm": 0.39781293272972107,
+      "learning_rate": 0.00019490015082074342,
+      "loss": 1.2946128845214844,
+      "mean_token_accuracy": 0.7150110900402069,
+      "num_tokens": 6058171.0,
+      "step": 1500
+    },
+    {
+      "epoch": 0.44130626654898497,
+      "eval_entropy": 1.2906748921539526,
+      "eval_loss": 1.3084396123886108,
+      "eval_mean_token_accuracy": 0.711328562789076,
+      "eval_num_tokens": 6058171.0,
+      "eval_runtime": 116.9727,
+      "eval_samples_per_second": 26.032,
+      "eval_steps_per_second": 3.257,
+      "step": 1500
+    },
+    {
+      "entropy": 1.335238778591156,
+      "epoch": 0.4442483083259782,
+      "grad_norm": 0.46100911498069763,
+      "learning_rate": 0.00019479740616975207,
+      "loss": 1.3694096565246583,
+      "mean_token_accuracy": 0.6983109295368195,
+      "num_tokens": 6098236.0,
+      "step": 1510
+    },
+    {
+      "entropy": 1.3258445739746094,
+      "epoch": 0.4471903501029715,
+      "grad_norm": 0.4548512399196625,
+      "learning_rate": 0.00019469366446811368,
+      "loss": 1.3223464012145996,
+      "mean_token_accuracy": 0.7081819117069245,
+      "num_tokens": 6138323.0,
+      "step": 1520
+    },
+    {
+      "entropy": 1.4055490851402284,
+      "epoch": 0.4501323918799647,
+      "grad_norm": 0.5857915282249451,
+      "learning_rate": 0.0001945889268069523,
+      "loss": 1.4265625,
+      "mean_token_accuracy": 0.6870070040225983,
+      "num_tokens": 6178759.0,
+      "step": 1530
+    },
+    {
+      "entropy": 1.3535587549209596,
+      "epoch": 0.45307443365695793,
+      "grad_norm": 0.4886157810688019,
+      "learning_rate": 0.00019448319428786714,
+      "loss": 1.3616491317749024,
+      "mean_token_accuracy": 0.6958298921585083,
+      "num_tokens": 6219216.0,
+      "step": 1540
+    },
+    {
+      "entropy": 1.2599135279655456,
+      "epoch": 0.4560164754339512,
+      "grad_norm": 0.4452349543571472,
+      "learning_rate": 0.00019437646802292116,
+      "loss": 1.2533982276916504,
+      "mean_token_accuracy": 0.7242700159549713,
+      "num_tokens": 6259618.0,
+      "step": 1550
+    },
+    {
+      "entropy": 1.247275459766388,
+      "epoch": 0.4589585172109444,
+      "grad_norm": 0.43969377875328064,
+      "learning_rate": 0.0001942687491346291,
+      "loss": 1.3022977828979492,
+      "mean_token_accuracy": 0.7166135847568512,
+      "num_tokens": 6300125.0,
+      "step": 1560
+    },
+    {
+      "entropy": 1.2933772325515747,
+      "epoch": 0.46190055898793764,
+      "grad_norm": 0.4098397195339203,
+      "learning_rate": 0.0001941600387559459,
+      "loss": 1.2829959869384766,
+      "mean_token_accuracy": 0.7151767909526825,
+      "num_tokens": 6340600.0,
+      "step": 1570
+    },
+    {
+      "entropy": 1.3236228585243226,
+      "epoch": 0.46484260076493084,
+      "grad_norm": 0.9286667108535767,
+      "learning_rate": 0.0001940503380302547,
+      "loss": 1.3246389389038087,
+      "mean_token_accuracy": 0.7080819308757782,
+      "num_tokens": 6381016.0,
+      "step": 1580
+    },
+    {
+      "entropy": 1.3100301146507263,
+      "epoch": 0.4677846425419241,
+      "grad_norm": 0.44687584042549133,
+      "learning_rate": 0.00019393964811135475,
+      "loss": 1.3230223655700684,
+      "mean_token_accuracy": 0.7061593413352967,
+      "num_tokens": 6421104.0,
+      "step": 1590
+    },
+    {
+      "entropy": 1.3405604600906371,
+      "epoch": 0.47072668431891734,
+      "grad_norm": 0.4198276400566101,
+      "learning_rate": 0.00019382797016344937,
+      "loss": 1.372106170654297,
+      "mean_token_accuracy": 0.6942126870155334,
+      "num_tokens": 6461464.0,
+      "step": 1600
+    },
+    {
+      "epoch": 0.47072668431891734,
+      "eval_entropy": 1.3327079477898405,
+      "eval_loss": 1.3015934228897095,
+      "eval_mean_token_accuracy": 0.7128373644170486,
+      "eval_num_tokens": 6461464.0,
+      "eval_runtime": 116.8886,
+      "eval_samples_per_second": 26.05,
+      "eval_steps_per_second": 3.26,
+      "step": 1600
+    },
+    {
+      "entropy": 1.2775195240974426,
+      "epoch": 0.47366872609591054,
+      "grad_norm": 0.36669495701789856,
+      "learning_rate": 0.00019371530536113372,
+      "loss": 1.2723214149475097,
+      "mean_token_accuracy": 0.7203551054000854,
+      "num_tokens": 6501717.0,
+      "step": 1610
+    },
+    {
+      "entropy": 1.263567042350769,
+      "epoch": 0.4766107678729038,
+      "grad_norm": 0.3709637522697449,
+      "learning_rate": 0.00019360165488938228,
+      "loss": 1.2917292594909668,
+      "mean_token_accuracy": 0.7129571974277497,
+      "num_tokens": 6542157.0,
+      "step": 1620
+    },
+    {
+      "entropy": 1.2828770697116851,
+      "epoch": 0.47955280964989705,
+      "grad_norm": 0.40589454770088196,
+      "learning_rate": 0.00019348701994353662,
+      "loss": 1.2924720764160156,
+      "mean_token_accuracy": 0.7142295658588409,
+      "num_tokens": 6582169.0,
+      "step": 1630
+    },
+    {
+      "entropy": 1.304066789150238,
+      "epoch": 0.48249485142689025,
+      "grad_norm": 0.41759273409843445,
+      "learning_rate": 0.0001933714017292927,
+      "loss": 1.3207477569580077,
+      "mean_token_accuracy": 0.7090053021907806,
+      "num_tokens": 6622526.0,
+      "step": 1640
+    },
+    {
+      "entropy": 1.2956494092941284,
+      "epoch": 0.4854368932038835,
+      "grad_norm": 0.49899545311927795,
+      "learning_rate": 0.00019325480146268812,
+      "loss": 1.3091160774230957,
+      "mean_token_accuracy": 0.7111666679382325,
+      "num_tokens": 6663165.0,
+      "step": 1650
+    },
+    {
+      "entropy": 1.2508437275886535,
+      "epoch": 0.48837893498087676,
+      "grad_norm": 0.4445253610610962,
+      "learning_rate": 0.0001931372203700895,
+      "loss": 1.2658663749694825,
+      "mean_token_accuracy": 0.7174579679965973,
+      "num_tokens": 6703494.0,
+      "step": 1660
+    },
+    {
+      "entropy": 1.287984347343445,
+      "epoch": 0.49132097675786995,
+      "grad_norm": 0.4067074656486511,
+      "learning_rate": 0.00019301865968817948,
+      "loss": 1.317389678955078,
+      "mean_token_accuracy": 0.7164031744003296,
+      "num_tokens": 6743946.0,
+      "step": 1670
+    },
+    {
+      "entropy": 1.272541528940201,
+      "epoch": 0.4942630185348632,
+      "grad_norm": 0.36161571741104126,
+      "learning_rate": 0.0001928991206639436,
+      "loss": 1.2725687980651856,
+      "mean_token_accuracy": 0.7200010478496551,
+      "num_tokens": 6784135.0,
+      "step": 1680
+    },
+    {
+      "entropy": 1.2077086687088012,
+      "epoch": 0.4972050603118564,
+      "grad_norm": 0.45158880949020386,
+      "learning_rate": 0.00019277860455465753,
+      "loss": 1.2226747512817382,
+      "mean_token_accuracy": 0.7305562674999238,
+      "num_tokens": 6824602.0,
+      "step": 1690
+    },
+    {
+      "entropy": 1.3152117371559142,
+      "epoch": 0.5001471020888497,
+      "grad_norm": 0.4278320074081421,
+      "learning_rate": 0.00019265711262787347,
+      "loss": 1.3219596862792968,
+      "mean_token_accuracy": 0.7065966606140137,
+      "num_tokens": 6864339.0,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5001471020888497,
+      "eval_entropy": 1.2670136558727956,
+      "eval_loss": 1.2904326915740967,
+      "eval_mean_token_accuracy": 0.7152716207692004,
+      "eval_num_tokens": 6864339.0,
+      "eval_runtime": 116.9086,
+      "eval_samples_per_second": 26.046,
+      "eval_steps_per_second": 3.259,
+      "step": 1700
+    },
+    {
+      "entropy": 1.3221548914909362,
+      "epoch": 0.5030891438658429,
+      "grad_norm": 0.3975633382797241,
+      "learning_rate": 0.00019253464616140702,
+      "loss": 1.349279022216797,
+      "mean_token_accuracy": 0.7058489739894866,
+      "num_tokens": 6904961.0,
+      "step": 1710
+    },
+    {
+      "entropy": 1.3088136434555053,
+      "epoch": 0.5060311856428361,
+      "grad_norm": 0.40347951650619507,
+      "learning_rate": 0.00019241120644332367,
+      "loss": 1.3145333290100099,
+      "mean_token_accuracy": 0.7059141278266907,
+      "num_tokens": 6945507.0,
+      "step": 1720
+    },
+    {
+      "entropy": 1.1808736383914948,
+      "epoch": 0.5089732274198293,
+      "grad_norm": 0.36107468605041504,
+      "learning_rate": 0.00019228679477192534,
+      "loss": 1.1965153694152832,
+      "mean_token_accuracy": 0.7326464653015137,
+      "num_tokens": 6985888.0,
+      "step": 1730
+    },
+    {
+      "entropy": 1.360763430595398,
+      "epoch": 0.5119152691968226,
+      "grad_norm": 0.4324724078178406,
+      "learning_rate": 0.0001921614124557366,
+      "loss": 1.3564807891845703,
+      "mean_token_accuracy": 0.7025493502616882,
+      "num_tokens": 7026252.0,
+      "step": 1740
+    },
+    {
+      "entropy": 1.2095451593399047,
+      "epoch": 0.5148573109738158,
+      "grad_norm": 0.39905378222465515,
+      "learning_rate": 0.000192035060813491,
+      "loss": 1.267704391479492,
+      "mean_token_accuracy": 0.7238605260848999,
+      "num_tokens": 7066705.0,
+      "step": 1750
+    },
+    {
+      "entropy": 1.379474115371704,
+      "epoch": 0.517799352750809,
+      "grad_norm": 0.44654110074043274,
+      "learning_rate": 0.00019190774117411717,
+      "loss": 1.352774715423584,
+      "mean_token_accuracy": 0.6996113300323487,
+      "num_tokens": 7106781.0,
+      "step": 1760
+    },
+    {
+      "entropy": 1.2537951588630676,
+      "epoch": 0.5207413945278023,
+      "grad_norm": 0.4148572087287903,
+      "learning_rate": 0.00019177945487672482,
+      "loss": 1.3051136970520019,
+      "mean_token_accuracy": 0.708976149559021,
+      "num_tokens": 7147220.0,
+      "step": 1770
+    },
+    {
+      "entropy": 1.2442652940750123,
+      "epoch": 0.5236834363047955,
+      "grad_norm": 0.45965835452079773,
+      "learning_rate": 0.00019165020327059073,
+      "loss": 1.2358969688415526,
+      "mean_token_accuracy": 0.7264375448226928,
+      "num_tokens": 7187657.0,
+      "step": 1780
+    },
+    {
+      "entropy": 1.2980039596557618,
+      "epoch": 0.5266254780817887,
+      "grad_norm": 0.4910111725330353,
+      "learning_rate": 0.00019151998771514442,
+      "loss": 1.3084582328796386,
+      "mean_token_accuracy": 0.7066324114799499,
+      "num_tokens": 7228181.0,
+      "step": 1790
+    },
+    {
+      "entropy": 1.2434914827346801,
+      "epoch": 0.529567519858782,
+      "grad_norm": 0.3658753037452698,
+      "learning_rate": 0.0001913888095799541,
+      "loss": 1.2698859214782714,
+      "mean_token_accuracy": 0.7207267701625824,
+      "num_tokens": 7268538.0,
+      "step": 1800
+    },
+    {
+      "epoch": 0.529567519858782,
+      "eval_entropy": 1.2639452193978578,
+      "eval_loss": 1.281688928604126,
+      "eval_mean_token_accuracy": 0.7168223152323345,
+      "eval_num_tokens": 7268538.0,
+      "eval_runtime": 116.9626,
+      "eval_samples_per_second": 26.034,
+      "eval_steps_per_second": 3.257,
+      "step": 1800
+    },
+    {
+      "entropy": 1.3006227135658264,
+      "epoch": 0.5325095616357752,
+      "grad_norm": 0.4497034549713135,
+      "learning_rate": 0.00019125667024471186,
+      "loss": 1.3108736991882324,
+      "mean_token_accuracy": 0.7042722702026367,
+      "num_tokens": 7308724.0,
+      "step": 1810
+    },
+    {
+      "entropy": 1.275818121433258,
+      "epoch": 0.5354516034127684,
+      "grad_norm": 0.4820014536380768,
+      "learning_rate": 0.00019112357109921964,
+      "loss": 1.2958525657653808,
+      "mean_token_accuracy": 0.7129449069499969,
+      "num_tokens": 7349172.0,
+      "step": 1820
+    },
+    {
+      "entropy": 1.245762574672699,
+      "epoch": 0.5383936451897617,
+      "grad_norm": 0.45138782262802124,
+      "learning_rate": 0.00019098951354337422,
+      "loss": 1.2449359893798828,
+      "mean_token_accuracy": 0.7270717859268189,
+      "num_tokens": 7389600.0,
+      "step": 1830
+    },
+    {
+      "entropy": 1.1886737942695618,
+      "epoch": 0.5413356869667549,
+      "grad_norm": 0.42880910634994507,
+      "learning_rate": 0.00019085449898715274,
+      "loss": 1.2165825843811036,
+      "mean_token_accuracy": 0.7328237056732178,
+      "num_tokens": 7429786.0,
+      "step": 1840
+    },
+    {
+      "entropy": 1.261668312549591,
+      "epoch": 0.5442777287437481,
+      "grad_norm": 0.41189083456993103,
+      "learning_rate": 0.0001907185288505978,
+      "loss": 1.2767746925354004,
+      "mean_token_accuracy": 0.718351137638092,
+      "num_tokens": 7470252.0,
+      "step": 1850
+    },
+    {
+      "entropy": 1.2039696455001831,
+      "epoch": 0.5472197705207414,
+      "grad_norm": 0.384387344121933,
+      "learning_rate": 0.0001905816045638024,
+      "loss": 1.2115536689758302,
+      "mean_token_accuracy": 0.7361586034297943,
+      "num_tokens": 7510473.0,
+      "step": 1860
+    },
+    {
+      "entropy": 1.2600649833679198,
+      "epoch": 0.5501618122977346,
+      "grad_norm": 0.4238559901714325,
+      "learning_rate": 0.00019044372756689504,
+      "loss": 1.2758872985839844,
+      "mean_token_accuracy": 0.7137023746967316,
+      "num_tokens": 7550932.0,
+      "step": 1870
+    },
+    {
+      "entropy": 1.3069803357124328,
+      "epoch": 0.5531038540747278,
+      "grad_norm": 0.44645845890045166,
+      "learning_rate": 0.00019030489931002461,
+      "loss": 1.3207664489746094,
+      "mean_token_accuracy": 0.7104817628860474,
+      "num_tokens": 7591311.0,
+      "step": 1880
+    },
+    {
+      "entropy": 1.3120111107826233,
+      "epoch": 0.556045895851721,
+      "grad_norm": 0.46638140082359314,
+      "learning_rate": 0.00019016512125334502,
+      "loss": 1.3248316764831543,
+      "mean_token_accuracy": 0.705955958366394,
+      "num_tokens": 7631397.0,
+      "step": 1890
+    },
+    {
+      "entropy": 1.279408586025238,
+      "epoch": 0.5589879376287143,
+      "grad_norm": 0.4149499535560608,
+      "learning_rate": 0.00019002439486699987,
+      "loss": 1.2773524284362794,
+      "mean_token_accuracy": 0.7147800862789154,
+      "num_tokens": 7671968.0,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5589879376287143,
+      "eval_entropy": 1.2283503402249394,
+      "eval_loss": 1.2789214849472046,
+      "eval_mean_token_accuracy": 0.7170038914743058,
+      "eval_num_tokens": 7671968.0,
+      "eval_runtime": 117.0781,
+      "eval_samples_per_second": 26.008,
+      "eval_steps_per_second": 3.254,
+      "step": 1900
+    },
+    {
+      "entropy": 1.230725634098053,
+      "epoch": 0.5619299794057075,
+      "grad_norm": 0.4699569046497345,
+      "learning_rate": 0.00018988272163110703,
+      "loss": 1.2412220001220704,
+      "mean_token_accuracy": 0.7230030238628388,
+      "num_tokens": 7712399.0,
+      "step": 1910
+    },
+    {
+      "entropy": 1.3014813184738159,
+      "epoch": 0.5648720211827007,
+      "grad_norm": 0.4014744162559509,
+      "learning_rate": 0.0001897401030357431,
+      "loss": 1.3222554206848145,
+      "mean_token_accuracy": 0.7079461216926575,
+      "num_tokens": 7752805.0,
+      "step": 1920
+    },
+    {
+      "entropy": 1.2195597231388091,
+      "epoch": 0.567814062959694,
+      "grad_norm": 0.4482584297657013,
+      "learning_rate": 0.00018959654058092753,
+      "loss": 1.2318772315979003,
+      "mean_token_accuracy": 0.7294041752815247,
+      "num_tokens": 7793144.0,
+      "step": 1930
+    },
+    {
+      "entropy": 1.1877473592758179,
+      "epoch": 0.5707561047366873,
+      "grad_norm": 0.44251689314842224,
+      "learning_rate": 0.00018945203577660718,
+      "loss": 1.2138689041137696,
+      "mean_token_accuracy": 0.7385743677616119,
+      "num_tokens": 7833447.0,
+      "step": 1940
+    },
+    {
+      "entropy": 1.289076042175293,
+      "epoch": 0.5736981465136805,
+      "grad_norm": 0.39147278666496277,
+      "learning_rate": 0.00018930659014264017,
+      "loss": 1.2974119186401367,
+      "mean_token_accuracy": 0.712678074836731,
+      "num_tokens": 7873974.0,
+      "step": 1950
+    },
+    {
+      "entropy": 1.208195787668228,
+      "epoch": 0.5766401882906738,
+      "grad_norm": 0.5337726473808289,
+      "learning_rate": 0.00018916020520877994,
+      "loss": 1.2156153678894044,
+      "mean_token_accuracy": 0.726024466753006,
+      "num_tokens": 7914528.0,
+      "step": 1960
+    },
+    {
+      "entropy": 1.2428096413612366,
+      "epoch": 0.579582230067667,
+      "grad_norm": 0.41402509808540344,
+      "learning_rate": 0.00018901288251465937,
+      "loss": 1.2540960311889648,
+      "mean_token_accuracy": 0.7271076798439026,
+      "num_tokens": 7954995.0,
+      "step": 1970
+    },
+    {
+      "entropy": 1.3336944222450255,
+      "epoch": 0.5825242718446602,
+      "grad_norm": 0.47286257147789,
+      "learning_rate": 0.00018886462360977418,
+      "loss": 1.3511391639709474,
+      "mean_token_accuracy": 0.7000328660011291,
+      "num_tokens": 7995173.0,
+      "step": 1980
+    },
+    {
+      "entropy": 1.210561752319336,
+      "epoch": 0.5854663136216535,
+      "grad_norm": 0.4485618472099304,
+      "learning_rate": 0.00018871543005346712,
+      "loss": 1.222616958618164,
+      "mean_token_accuracy": 0.7286027550697327,
+      "num_tokens": 8035513.0,
+      "step": 1990
+    },
+    {
+      "entropy": 1.2922529458999634,
+      "epoch": 0.5884083553986467,
+      "grad_norm": 0.4895835220813751,
+      "learning_rate": 0.0001885653034149111,
+      "loss": 1.3031671524047852,
+      "mean_token_accuracy": 0.7130830705165863,
+      "num_tokens": 8075879.0,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5884083553986467,
+      "eval_entropy": 1.2732777583004609,
+      "eval_loss": 1.2699540853500366,
+      "eval_mean_token_accuracy": 0.7192083870957843,
+      "eval_num_tokens": 8075879.0,
+      "eval_runtime": 116.9367,
+      "eval_samples_per_second": 26.04,
+      "eval_steps_per_second": 3.258,
+      "step": 2000
+    },
+    {
+      "entropy": 1.2049331665039062,
+      "epoch": 0.5913503971756399,
+      "grad_norm": 0.42137858271598816,
+      "learning_rate": 0.00018841424527309312,
+      "loss": 1.220776081085205,
+      "mean_token_accuracy": 0.7255268514156341,
+      "num_tokens": 8116494.0,
+      "step": 2010
+    },
+    {
+      "entropy": 1.2747172474861146,
+      "epoch": 0.5942924389526332,
+      "grad_norm": 0.47128528356552124,
+      "learning_rate": 0.0001882622572167973,
+      "loss": 1.2953474044799804,
+      "mean_token_accuracy": 0.7114850461483002,
+      "num_tokens": 8156917.0,
+      "step": 2020
+    },
+    {
+      "entropy": 1.2040826320648192,
+      "epoch": 0.5972344807296264,
+      "grad_norm": 0.477896124124527,
+      "learning_rate": 0.0001881093408445884,
+      "loss": 1.2177928924560546,
+      "mean_token_accuracy": 0.7306720972061157,
+      "num_tokens": 8197335.0,
+      "step": 2030
+    },
+    {
+      "entropy": 1.2654430389404296,
+      "epoch": 0.6001765225066196,
+      "grad_norm": 0.5165997743606567,
+      "learning_rate": 0.00018795549776479478,
+      "loss": 1.275616455078125,
+      "mean_token_accuracy": 0.7176172435283661,
+      "num_tokens": 8237564.0,
+      "step": 2040
+    },
+    {
+      "entropy": 1.2124343276023866,
+      "epoch": 0.6031185642836129,
+      "grad_norm": 0.47672176361083984,
+      "learning_rate": 0.0001878007295954919,
+      "loss": 1.2235189437866212,
+      "mean_token_accuracy": 0.7277295291423798,
+      "num_tokens": 8277586.0,
+      "step": 2050
+    },
+    {
+      "entropy": 1.2670048475265503,
+      "epoch": 0.6060606060606061,
+      "grad_norm": 0.42278149724006653,
+      "learning_rate": 0.00018764503796448478,
+      "loss": 1.284127426147461,
+      "mean_token_accuracy": 0.7204706132411957,
+      "num_tokens": 8318120.0,
+      "step": 2060
+    },
+    {
+      "entropy": 1.2286874294281005,
+      "epoch": 0.6090026478375993,
+      "grad_norm": 0.4719674587249756,
+      "learning_rate": 0.0001874884245092913,
+      "loss": 1.2510007858276366,
+      "mean_token_accuracy": 0.7220764279365539,
+      "num_tokens": 8358689.0,
+      "step": 2070
+    },
+    {
+      "entropy": 1.298528516292572,
+      "epoch": 0.6119446896145925,
+      "grad_norm": 0.46403807401657104,
+      "learning_rate": 0.00018733089087712469,
+      "loss": 1.2856470108032227,
+      "mean_token_accuracy": 0.7156751930713654,
+      "num_tokens": 8399274.0,
+      "step": 2080
+    },
+    {
+      "entropy": 1.268327033519745,
+      "epoch": 0.6148867313915858,
+      "grad_norm": 0.4383772611618042,
+      "learning_rate": 0.00018717243872487643,
+      "loss": 1.3167724609375,
+      "mean_token_accuracy": 0.7053386807441712,
+      "num_tokens": 8439732.0,
+      "step": 2090
+    },
+    {
+      "entropy": 1.3164120435714721,
+      "epoch": 0.617828773168579,
+      "grad_norm": 0.5242652893066406,
+      "learning_rate": 0.00018701306971909864,
+      "loss": 1.3042461395263671,
+      "mean_token_accuracy": 0.7102492094039917,
+      "num_tokens": 8479716.0,
+      "step": 2100
+    },
+    {
+      "epoch": 0.617828773168579,
+      "eval_entropy": 1.2351457424989836,
+      "eval_loss": 1.2649626731872559,
+      "eval_mean_token_accuracy": 0.7205018502833649,
+      "eval_num_tokens": 8479716.0,
+      "eval_runtime": 117.0816,
+      "eval_samples_per_second": 26.008,
+      "eval_steps_per_second": 3.254,
+      "step": 2100
+    },
+    {
+      "entropy": 1.2392801344394684,
+      "epoch": 0.6207708149455722,
+      "grad_norm": 0.47463512420654297,
+      "learning_rate": 0.00018685278553598665,
+      "loss": 1.2612558364868165,
+      "mean_token_accuracy": 0.7227189362049102,
+      "num_tokens": 8520008.0,
+      "step": 2110
+    },
+    {
+      "entropy": 1.2834307312965394,
+      "epoch": 0.6237128567225655,
+      "grad_norm": 0.4995603859424591,
+      "learning_rate": 0.0001866915878613614,
+      "loss": 1.2821990966796875,
+      "mean_token_accuracy": 0.7140654444694519,
+      "num_tokens": 8560586.0,
+      "step": 2120
+    },
+    {
+      "entropy": 1.2452851295471192,
+      "epoch": 0.6266548984995587,
+      "grad_norm": 0.40243977308273315,
+      "learning_rate": 0.00018652947839065159,
+      "loss": 1.2768383026123047,
+      "mean_token_accuracy": 0.7194826602935791,
+      "num_tokens": 8601116.0,
+      "step": 2130
+    },
+    {
+      "entropy": 1.2355848252773285,
+      "epoch": 0.6295969402765519,
+      "grad_norm": 0.3945181369781494,
+      "learning_rate": 0.00018636645882887592,
+      "loss": 1.2494622230529786,
+      "mean_token_accuracy": 0.7192557215690613,
+      "num_tokens": 8641094.0,
+      "step": 2140
+    },
+    {
+      "entropy": 1.243088138103485,
+      "epoch": 0.6325389820535452,
+      "grad_norm": 0.43109068274497986,
+      "learning_rate": 0.0001862025308906252,
+      "loss": 1.2423103332519532,
+      "mean_token_accuracy": 0.7271890878677368,
+      "num_tokens": 8681458.0,
+      "step": 2150
+    },
+    {
+      "entropy": 1.2035149574279784,
+      "epoch": 0.6354810238305384,
+      "grad_norm": 0.4727862477302551,
+      "learning_rate": 0.0001860376963000443,
+      "loss": 1.2361689567565919,
+      "mean_token_accuracy": 0.7284120082855224,
+      "num_tokens": 8722066.0,
+      "step": 2160
+    },
+    {
+      "entropy": 1.2721379399299622,
+      "epoch": 0.6384230656075316,
+      "grad_norm": 0.5057058930397034,
+      "learning_rate": 0.00018587195679081386,
+      "loss": 1.2708622932434082,
+      "mean_token_accuracy": 0.7155144691467286,
+      "num_tokens": 8762672.0,
+      "step": 2170
+    },
+    {
+      "entropy": 1.2714126229286193,
+      "epoch": 0.6413651073845249,
+      "grad_norm": 0.4464847147464752,
+      "learning_rate": 0.0001857053141061323,
+      "loss": 1.2917202949523925,
+      "mean_token_accuracy": 0.7170847177505493,
+      "num_tokens": 8802819.0,
+      "step": 2180
+    },
+    {
+      "entropy": 1.2385571956634522,
+      "epoch": 0.6443071491615181,
+      "grad_norm": 0.4425857365131378,
+      "learning_rate": 0.00018553776999869737,
+      "loss": 1.252675437927246,
+      "mean_token_accuracy": 0.7212274014949799,
+      "num_tokens": 8842977.0,
+      "step": 2190
+    },
+    {
+      "entropy": 1.2919608235359192,
+      "epoch": 0.6472491909385113,
+      "grad_norm": 0.43156683444976807,
+      "learning_rate": 0.00018536932623068757,
+      "loss": 1.2964573860168458,
+      "mean_token_accuracy": 0.7152363657951355,
+      "num_tokens": 8883579.0,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6472491909385113,
+      "eval_entropy": 1.2316402986919472,
+      "eval_loss": 1.2606068849563599,
+      "eval_mean_token_accuracy": 0.7208151789162103,
+      "eval_num_tokens": 8883579.0,
+      "eval_runtime": 117.0496,
+      "eval_samples_per_second": 26.015,
+      "eval_steps_per_second": 3.255,
+      "step": 2200
+    },
+    {
+      "entropy": 1.3093620181083678,
+      "epoch": 0.6501912327155046,
+      "grad_norm": 0.4637244641780853,
+      "learning_rate": 0.00018519998457374395,
+      "loss": 1.34036865234375,
+      "mean_token_accuracy": 0.7034331321716308,
+      "num_tokens": 8923903.0,
+      "step": 2210
+    },
+    {
+      "entropy": 1.214417290687561,
+      "epoch": 0.6531332744924978,
+      "grad_norm": 0.5183310508728027,
+      "learning_rate": 0.00018502974680895115,
+      "loss": 1.230905532836914,
+      "mean_token_accuracy": 0.7264467597007751,
+      "num_tokens": 8963848.0,
+      "step": 2220
+    },
+    {
+      "entropy": 1.2735732913017273,
+      "epoch": 0.656075316269491,
+      "grad_norm": 0.6038491725921631,
+      "learning_rate": 0.00018485861472681888,
+      "loss": 1.2594423294067383,
+      "mean_token_accuracy": 0.7218607306480408,
+      "num_tokens": 9003224.0,
+      "step": 2230
+    },
+    {
+      "entropy": 1.278822934627533,
+      "epoch": 0.6590173580464843,
+      "grad_norm": 0.47614696621894836,
+      "learning_rate": 0.00018468659012726301,
+      "loss": 1.3205299377441406,
+      "mean_token_accuracy": 0.7092170417308807,
+      "num_tokens": 9043662.0,
+      "step": 2240
+    },
+    {
+      "entropy": 1.2612668752670289,
+      "epoch": 0.6619593998234775,
+      "grad_norm": 0.4333754777908325,
+      "learning_rate": 0.00018451367481958655,
+      "loss": 1.269089412689209,
+      "mean_token_accuracy": 0.7188269674777985,
+      "num_tokens": 9084132.0,
+      "step": 2250
+    },
+    {
+      "entropy": 1.2136012673377992,
+      "epoch": 0.6649014416004707,
+      "grad_norm": 0.44086429476737976,
+      "learning_rate": 0.0001843398706224608,
+      "loss": 1.222397518157959,
+      "mean_token_accuracy": 0.7309954643249512,
+      "num_tokens": 9124313.0,
+      "step": 2260
+    },
+    {
+      "entropy": 1.2528056859970094,
+      "epoch": 0.6678434833774639,
+      "grad_norm": 0.45429208874702454,
+      "learning_rate": 0.0001841651793639061,
+      "loss": 1.2543872833251952,
+      "mean_token_accuracy": 0.7223168253898621,
+      "num_tokens": 9164680.0,
+      "step": 2270
+    },
+    {
+      "entropy": 1.2683658719062805,
+      "epoch": 0.6707855251544572,
+      "grad_norm": 0.4545646011829376,
+      "learning_rate": 0.00018398960288127264,
+      "loss": 1.3083954811096192,
+      "mean_token_accuracy": 0.7077820837497711,
+      "num_tokens": 9205179.0,
+      "step": 2280
+    },
+    {
+      "entropy": 1.267857301235199,
+      "epoch": 0.6737275669314504,
+      "grad_norm": 0.43586161732673645,
+      "learning_rate": 0.00018381314302122115,
+      "loss": 1.2671592712402344,
+      "mean_token_accuracy": 0.7227232694625855,
+      "num_tokens": 9245707.0,
+      "step": 2290
+    },
+    {
+      "entropy": 1.191249167919159,
+      "epoch": 0.6766696087084436,
+      "grad_norm": 0.5259418487548828,
+      "learning_rate": 0.00018363580163970343,
+      "loss": 1.1978882789611816,
+      "mean_token_accuracy": 0.7335732400417327,
+      "num_tokens": 9286200.0,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6766696087084436,
+      "eval_entropy": 1.1682455519365826,
+      "eval_loss": 1.257144808769226,
+      "eval_mean_token_accuracy": 0.7223104308909318,
+      "eval_num_tokens": 9286200.0,
+      "eval_runtime": 116.9696,
+      "eval_samples_per_second": 26.032,
+      "eval_steps_per_second": 3.257,
+      "step": 2300
+    },
+    {
+      "entropy": 1.2716636419296266,
+      "epoch": 0.6796116504854369,
+      "grad_norm": 0.415056049823761,
+      "learning_rate": 0.00018345758060194287,
+      "loss": 1.2905259132385254,
+      "mean_token_accuracy": 0.710951566696167,
+      "num_tokens": 9326847.0,
+      "step": 2310
+    },
+    {
+      "entropy": 1.1874773681163788,
+      "epoch": 0.6825536922624301,
+      "grad_norm": 0.49722516536712646,
+      "learning_rate": 0.00018327848178241481,
+      "loss": 1.217663288116455,
+      "mean_token_accuracy": 0.7317953050136566,
+      "num_tokens": 9367084.0,
+      "step": 2320
+    },
+    {
+      "entropy": 1.2677314758300782,
+      "epoch": 0.6854957340394233,
+      "grad_norm": 0.4097212255001068,
+      "learning_rate": 0.00018309850706482687,
+      "loss": 1.2633783340454101,
+      "mean_token_accuracy": 0.7190278351306916,
+      "num_tokens": 9407189.0,
+      "step": 2330
+    },
+    {
+      "entropy": 1.2582902312278748,
+      "epoch": 0.6884377758164166,
+      "grad_norm": 0.40571051836013794,
+      "learning_rate": 0.00018291765834209907,
+      "loss": 1.2858672142028809,
+      "mean_token_accuracy": 0.7170897841453552,
+      "num_tokens": 9447880.0,
+      "step": 2340
+    },
+    {
+      "entropy": 1.327420747280121,
+      "epoch": 0.6913798175934098,
+      "grad_norm": 0.48363062739372253,
+      "learning_rate": 0.0001827359375163439,
+      "loss": 1.327678108215332,
+      "mean_token_accuracy": 0.7030242502689361,
+      "num_tokens": 9488502.0,
+      "step": 2350
+    },
+    {
+      "entropy": 1.1727963089942932,
+      "epoch": 0.694321859370403,
+      "grad_norm": 0.6001095175743103,
+      "learning_rate": 0.00018255334649884653,
+      "loss": 1.2080462455749512,
+      "mean_token_accuracy": 0.7356064558029175,
+      "num_tokens": 9529082.0,
+      "step": 2360
+    },
+    {
+      "entropy": 1.2679656863212585,
+      "epoch": 0.6972639011473963,
+      "grad_norm": 0.4376157522201538,
+      "learning_rate": 0.00018236988721004435,
+      "loss": 1.2510211944580079,
+      "mean_token_accuracy": 0.7209162175655365,
+      "num_tokens": 9569520.0,
+      "step": 2370
+    },
+    {
+      "entropy": 1.192594301700592,
+      "epoch": 0.7002059429243895,
+      "grad_norm": 0.4717561900615692,
+      "learning_rate": 0.00018218556157950712,
+      "loss": 1.2164905548095704,
+      "mean_token_accuracy": 0.7315115988254547,
+      "num_tokens": 9609903.0,
+      "step": 2380
+    },
+    {
+      "entropy": 1.3375505089759827,
+      "epoch": 0.7031479847013827,
+      "grad_norm": 0.5176673531532288,
+      "learning_rate": 0.00018200037154591643,
+      "loss": 1.3507174491882323,
+      "mean_token_accuracy": 0.69825981259346,
+      "num_tokens": 9650434.0,
+      "step": 2390
+    },
+    {
+      "entropy": 1.2428280234336853,
+      "epoch": 0.706090026478376,
+      "grad_norm": 0.4390230178833008,
+      "learning_rate": 0.00018181431905704546,
+      "loss": 1.254446029663086,
+      "mean_token_accuracy": 0.7244620621204376,
+      "num_tokens": 9690991.0,
+      "step": 2400
+    },
+    {
+      "epoch": 0.706090026478376,
+      "eval_entropy": 1.2351404309585652,
+      "eval_loss": 1.2472655773162842,
+      "eval_mean_token_accuracy": 0.7242527881006556,
+      "eval_num_tokens": 9690991.0,
+      "eval_runtime": 117.0968,
+      "eval_samples_per_second": 26.004,
+      "eval_steps_per_second": 3.254,
+      "step": 2400
+    },
+    {
+      "entropy": 1.1983890414237977,
+      "epoch": 0.7090320682553692,
+      "grad_norm": 0.44956910610198975,
+      "learning_rate": 0.0001816274060697384,
+      "loss": 1.2177659034729005,
+      "mean_token_accuracy": 0.731769061088562,
+      "num_tokens": 9731440.0,
+      "step": 2410
+    },
+    {
+      "entropy": 1.2803593873977661,
+      "epoch": 0.7119741100323624,
+      "grad_norm": 0.4463217854499817,
+      "learning_rate": 0.00018143963454988994,
+      "loss": 1.2735008239746093,
+      "mean_token_accuracy": 0.7187759101390838,
+      "num_tokens": 9771061.0,
+      "step": 2420
+    },
+    {
+      "entropy": 1.22296462059021,
+      "epoch": 0.7149161518093556,
+      "grad_norm": 0.4359455406665802,
+      "learning_rate": 0.0001812510064724245,
+      "loss": 1.260395622253418,
+      "mean_token_accuracy": 0.7190466344356536,
+      "num_tokens": 9811550.0,
+      "step": 2430
+    },
+    {
+      "entropy": 1.2306616604328156,
+      "epoch": 0.717858193586349,
+      "grad_norm": 0.4980376660823822,
+      "learning_rate": 0.0001810615238212755,
+      "loss": 1.227048110961914,
+      "mean_token_accuracy": 0.7294842720031738,
+      "num_tokens": 9852102.0,
+      "step": 2440
+    },
+    {
+      "entropy": 1.2669499397277832,
+      "epoch": 0.7208002353633421,
+      "grad_norm": 0.4520755708217621,
+      "learning_rate": 0.00018087118858936462,
+      "loss": 1.2932658195495605,
+      "mean_token_accuracy": 0.7133265674114228,
+      "num_tokens": 9892746.0,
+      "step": 2450
+    },
+    {
+      "entropy": 1.2530406713485718,
+      "epoch": 0.7237422771403353,
+      "grad_norm": 0.6179884076118469,
+      "learning_rate": 0.00018068000277858065,
+      "loss": 1.2789620399475097,
+      "mean_token_accuracy": 0.7135308802127838,
+      "num_tokens": 9933185.0,
+      "step": 2460
+    },
+    {
+      "entropy": 1.2204147577285767,
+      "epoch": 0.7266843189173287,
+      "grad_norm": 0.4976007044315338,
+      "learning_rate": 0.00018048796839975856,
+      "loss": 1.2141535758972168,
+      "mean_token_accuracy": 0.7293932437896729,
+      "num_tokens": 9973384.0,
+      "step": 2470
+    },
+    {
+      "entropy": 1.1704169631004333,
+      "epoch": 0.7296263606943219,
+      "grad_norm": 0.44038140773773193,
+      "learning_rate": 0.0001802950874726582,
+      "loss": 1.1908206939697266,
+      "mean_token_accuracy": 0.7376551747322082,
+      "num_tokens": 10013764.0,
+      "step": 2480
+    },
+    {
+      "entropy": 1.2523088693618774,
+      "epoch": 0.732568402471315,
+      "grad_norm": 0.46554332971572876,
+      "learning_rate": 0.00018010136202594332,
+      "loss": 1.2656194686889648,
+      "mean_token_accuracy": 0.7242594540119172,
+      "num_tokens": 10054063.0,
+      "step": 2490
+    },
+    {
+      "entropy": 1.3448559761047363,
+      "epoch": 0.7355104442483084,
+      "grad_norm": 0.4372340142726898,
+      "learning_rate": 0.00017990679409715993,
+      "loss": 1.3519328117370606,
+      "mean_token_accuracy": 0.698470801115036,
+      "num_tokens": 10094687.0,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7355104442483084,
+      "eval_entropy": 1.2676614907782848,
+      "eval_loss": 1.2452576160430908,
+      "eval_mean_token_accuracy": 0.7236289200507436,
+      "eval_num_tokens": 10094687.0,
+      "eval_runtime": 116.8804,
+      "eval_samples_per_second": 26.052,
+      "eval_steps_per_second": 3.26,
+      "step": 2500
+    },
+    {
+      "entropy": 1.2084831714630127,
+      "epoch": 0.7384524860253016,
+      "grad_norm": 0.4193192422389984,
+      "learning_rate": 0.00017971138573271507,
+      "loss": 1.218832778930664,
+      "mean_token_accuracy": 0.7289343297481536,
+      "num_tokens": 10135026.0,
+      "step": 2510
+    },
+    {
+      "entropy": 1.1639327347278594,
+      "epoch": 0.7413945278022948,
+      "grad_norm": 0.48731788992881775,
+      "learning_rate": 0.0001795151389878552,
+      "loss": 1.1885252952575684,
+      "mean_token_accuracy": 0.7402738213539124,
+      "num_tokens": 10175310.0,
+      "step": 2520
+    },
+    {
+      "entropy": 1.2578992068767547,
+      "epoch": 0.7443365695792881,
+      "grad_norm": 0.434038370847702,
+      "learning_rate": 0.00017931805592664472,
+      "loss": 1.26021728515625,
+      "mean_token_accuracy": 0.7184948623180389,
+      "num_tokens": 10215765.0,
+      "step": 2530
+    },
+    {
+      "entropy": 1.2038448989391326,
+      "epoch": 0.7472786113562813,
+      "grad_norm": 0.48660168051719666,
+      "learning_rate": 0.00017912013862194404,
+      "loss": 1.211390781402588,
+      "mean_token_accuracy": 0.7284208476543427,
+      "num_tokens": 10256371.0,
+      "step": 2540
+    },
+    {
+      "entropy": 1.1407162606716157,
+      "epoch": 0.7502206531332745,
+      "grad_norm": 0.5016793608665466,
+      "learning_rate": 0.0001789213891553879,
+      "loss": 1.1628236770629883,
+      "mean_token_accuracy": 0.7418089389801026,
+      "num_tokens": 10296590.0,
+      "step": 2550
+    },
+    {
+      "entropy": 1.2328106760978699,
+      "epoch": 0.7531626949102678,
+      "grad_norm": 0.42070743441581726,
+      "learning_rate": 0.00017872180961736356,
+      "loss": 1.245603656768799,
+      "mean_token_accuracy": 0.7262676537036896,
+      "num_tokens": 10337069.0,
+      "step": 2560
+    },
+    {
+      "entropy": 1.2932706713676452,
+      "epoch": 0.756104736687261,
+      "grad_norm": 0.46600809693336487,
+      "learning_rate": 0.00017852140210698858,
+      "loss": 1.2860488891601562,
+      "mean_token_accuracy": 0.7088023841381073,
+      "num_tokens": 10377428.0,
+      "step": 2570
+    },
+    {
+      "entropy": 1.147745430469513,
+      "epoch": 0.7590467784642542,
+      "grad_norm": 0.4609155058860779,
+      "learning_rate": 0.00017832016873208905,
+      "loss": 1.1787425994873046,
+      "mean_token_accuracy": 0.7361489832401276,
+      "num_tokens": 10417739.0,
+      "step": 2580
+    },
+    {
+      "entropy": 1.2378079295158386,
+      "epoch": 0.7619888202412475,
+      "grad_norm": 0.4880058467388153,
+      "learning_rate": 0.00017811811160917712,
+      "loss": 1.2569812774658202,
+      "mean_token_accuracy": 0.7204902648925782,
+      "num_tokens": 10458215.0,
+      "step": 2590
+    },
+    {
+      "entropy": 1.2285701274871825,
+      "epoch": 0.7649308620182407,
+      "grad_norm": 0.5326588153839111,
+      "learning_rate": 0.0001779152328634289,
+      "loss": 1.221341609954834,
+      "mean_token_accuracy": 0.7287818729877472,
+      "num_tokens": 10498647.0,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7649308620182407,
+      "eval_entropy": 1.2003295410649357,
+      "eval_loss": 1.236678957939148,
+      "eval_mean_token_accuracy": 0.7259835525760501,
+      "eval_num_tokens": 10498647.0,
+      "eval_runtime": 116.9331,
+      "eval_samples_per_second": 26.041,
+      "eval_steps_per_second": 3.258,
+      "step": 2600
+    },
+    {
+      "entropy": 1.2428824484348298,
+      "epoch": 0.7678729037952339,
+      "grad_norm": 0.45255497097969055,
+      "learning_rate": 0.00017771153462866216,
+      "loss": 1.2709949493408204,
+      "mean_token_accuracy": 0.7177605211734772,
+      "num_tokens": 10539009.0,
+      "step": 2610
+    },
+    {
+      "entropy": 1.277386212348938,
+      "epoch": 0.7708149455722271,
+      "grad_norm": 0.48587530851364136,
+      "learning_rate": 0.00017750701904731373,
+      "loss": 1.2778440475463868,
+      "mean_token_accuracy": 0.7140256285667419,
+      "num_tokens": 10579502.0,
+      "step": 2620
+    },
+    {
+      "entropy": 1.2311038613319396,
+      "epoch": 0.7737569873492204,
+      "grad_norm": 0.6130372285842896,
+      "learning_rate": 0.00017730168827041708,
+      "loss": 1.2585201263427734,
+      "mean_token_accuracy": 0.7184097468852997,
+      "num_tokens": 10619385.0,
+      "step": 2630
+    },
+    {
+      "entropy": 1.2518787860870362,
+      "epoch": 0.7766990291262136,
+      "grad_norm": 0.43607237935066223,
+      "learning_rate": 0.00017709554445757966,
+      "loss": 1.2394478797912598,
+      "mean_token_accuracy": 0.726053637266159,
+      "num_tokens": 10659852.0,
+      "step": 2640
+    },
+    {
+      "entropy": 1.2289941668510438,
+      "epoch": 0.7796410709032068,
+      "grad_norm": 0.4687124788761139,
+      "learning_rate": 0.00017688858977696014,
+      "loss": 1.2466455459594727,
+      "mean_token_accuracy": 0.7200910389423371,
+      "num_tokens": 10700123.0,
+      "step": 2650
+    },
+    {
+      "entropy": 1.232028889656067,
+      "epoch": 0.7825831126802001,
+      "grad_norm": 0.45022356510162354,
+      "learning_rate": 0.00017668082640524574,
+      "loss": 1.2503207206726075,
+      "mean_token_accuracy": 0.7186446607112884,
+      "num_tokens": 10740503.0,
+      "step": 2660
+    },
+    {
+      "entropy": 1.204378592967987,
+      "epoch": 0.7855251544571933,
+      "grad_norm": 0.41389408707618713,
+      "learning_rate": 0.0001764722565276292,
+      "loss": 1.211115264892578,
+      "mean_token_accuracy": 0.7328132688999176,
+      "num_tokens": 10780655.0,
+      "step": 2670
+    },
+    {
+      "entropy": 1.2730337023735045,
+      "epoch": 0.7884671962341865,
+      "grad_norm": 0.4794485569000244,
+      "learning_rate": 0.00017626288233778582,
+      "loss": 1.2866595268249512,
+      "mean_token_accuracy": 0.7171245098114014,
+      "num_tokens": 10821171.0,
+      "step": 2680
+    },
+    {
+      "entropy": 1.2131190776824952,
+      "epoch": 0.7914092380111798,
+      "grad_norm": 0.4519226551055908,
+      "learning_rate": 0.00017605270603785047,
+      "loss": 1.228813934326172,
+      "mean_token_accuracy": 0.7272532522678375,
+      "num_tokens": 10861594.0,
+      "step": 2690
+    },
+    {
+      "entropy": 1.2511601805686952,
+      "epoch": 0.794351279788173,
+      "grad_norm": 0.4799201488494873,
+      "learning_rate": 0.00017584172983839435,
+      "loss": 1.2575819969177247,
+      "mean_token_accuracy": 0.7226161539554596,
+      "num_tokens": 10902043.0,
+      "step": 2700
+    },
+    {
+      "epoch": 0.794351279788173,
+      "eval_entropy": 1.2509275682642078,
+      "eval_loss": 1.2339112758636475,
+      "eval_mean_token_accuracy": 0.7265911623248904,
+      "eval_num_tokens": 10902043.0,
+      "eval_runtime": 116.9442,
+      "eval_samples_per_second": 26.038,
+      "eval_steps_per_second": 3.258,
+      "step": 2700
+    },
+    {
+      "entropy": 1.2342944502830506,
+      "epoch": 0.7972933215651662,
+      "grad_norm": 0.49187448620796204,
+      "learning_rate": 0.00017562995595840178,
+      "loss": 1.2416543006896972,
+      "mean_token_accuracy": 0.7259420096874237,
+      "num_tokens": 10942392.0,
+      "step": 2710
+    },
+    {
+      "entropy": 1.1988507807254791,
+      "epoch": 0.8002353633421595,
+      "grad_norm": 0.4817524254322052,
+      "learning_rate": 0.00017541738662524677,
+      "loss": 1.2237167358398438,
+      "mean_token_accuracy": 0.7307404637336731,
+      "num_tokens": 10982924.0,
+      "step": 2720
+    },
+    {
+      "entropy": 1.236850619316101,
+      "epoch": 0.8031774051191527,
+      "grad_norm": 0.4729112684726715,
+      "learning_rate": 0.0001752040240746698,
+      "loss": 1.2266542434692382,
+      "mean_token_accuracy": 0.7280332922935486,
+      "num_tokens": 11023305.0,
+      "step": 2730
+    },
+    {
+      "entropy": 1.1383702993392943,
+      "epoch": 0.8061194468961459,
+      "grad_norm": 0.42324507236480713,
+      "learning_rate": 0.00017498987055075403,
+      "loss": 1.1699549674987793,
+      "mean_token_accuracy": 0.7414192616939544,
+      "num_tokens": 11063084.0,
+      "step": 2740
+    },
+    {
+      "entropy": 1.1376792788505554,
+      "epoch": 0.8090614886731392,
+      "grad_norm": 0.5435130000114441,
+      "learning_rate": 0.00017477492830590192,
+      "loss": 1.1505720138549804,
+      "mean_token_accuracy": 0.7437731683254242,
+      "num_tokens": 11103518.0,
+      "step": 2750
+    },
+    {
+      "entropy": 1.2045920014381408,
+      "epoch": 0.8120035304501324,
+      "grad_norm": 0.4564155340194702,
+      "learning_rate": 0.00017455919960081149,
+      "loss": 1.1978718757629394,
+      "mean_token_accuracy": 0.7353939712047577,
+      "num_tokens": 11143802.0,
+      "step": 2760
+    },
+    {
+      "entropy": 1.197198224067688,
+      "epoch": 0.8149455722271256,
+      "grad_norm": 0.457720011472702,
+      "learning_rate": 0.0001743426867044524,
+      "loss": 1.2358501434326172,
+      "mean_token_accuracy": 0.7267086863517761,
+      "num_tokens": 11184158.0,
+      "step": 2770
+    },
+    {
+      "entropy": 1.2718565106391906,
+      "epoch": 0.8178876140041189,
+      "grad_norm": 0.4250863194465637,
+      "learning_rate": 0.00017412539189404233,
+      "loss": 1.2535717010498046,
+      "mean_token_accuracy": 0.7206644594669342,
+      "num_tokens": 11224574.0,
+      "step": 2780
+    },
+    {
+      "entropy": 1.172040694952011,
+      "epoch": 0.8208296557811121,
+      "grad_norm": 0.5394912958145142,
+      "learning_rate": 0.00017390731745502283,
+      "loss": 1.2080710411071778,
+      "mean_token_accuracy": 0.7338366210460663,
+      "num_tokens": 11265027.0,
+      "step": 2790
+    },
+    {
+      "entropy": 1.187865948677063,
+      "epoch": 0.8237716975581053,
+      "grad_norm": 0.49136948585510254,
+      "learning_rate": 0.00017368846568103529,
+      "loss": 1.1811614990234376,
+      "mean_token_accuracy": 0.7365565001964569,
+      "num_tokens": 11305432.0,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8237716975581053,
+      "eval_entropy": 1.2183332227346466,
+      "eval_loss": 1.2257474660873413,
+      "eval_mean_token_accuracy": 0.7285947153574526,
+      "eval_num_tokens": 11305432.0,
+      "eval_runtime": 116.8432,
+      "eval_samples_per_second": 26.061,
+      "eval_steps_per_second": 3.261,
+      "step": 2800
+    },
+    {
+      "entropy": 1.1511970162391663,
+      "epoch": 0.8267137393350985,
+      "grad_norm": 0.487531840801239,
+      "learning_rate": 0.00017346883887389702,
+      "loss": 1.1708711624145507,
+      "mean_token_accuracy": 0.7420612633228302,
+      "num_tokens": 11345702.0,
+      "step": 2810
+    },
+    {
+      "entropy": 1.2634961485862732,
+      "epoch": 0.8296557811120918,
+      "grad_norm": 0.4640207886695862,
+      "learning_rate": 0.00017324843934357674,
+      "loss": 1.277150821685791,
+      "mean_token_accuracy": 0.7138958215713501,
+      "num_tokens": 11386240.0,
+      "step": 2820
+    },
+    {
+      "entropy": 1.1891654789447785,
+      "epoch": 0.832597822889085,
+      "grad_norm": 0.47752827405929565,
+      "learning_rate": 0.0001730272694081706,
+      "loss": 1.193849754333496,
+      "mean_token_accuracy": 0.7351171731948852,
+      "num_tokens": 11426663.0,
+      "step": 2830
+    },
+    {
+      "entropy": 1.2108049154281617,
+      "epoch": 0.8355398646660782,
+      "grad_norm": 0.514695942401886,
+      "learning_rate": 0.0001728053313938775,
+      "loss": 1.2478459358215332,
+      "mean_token_accuracy": 0.721454119682312,
+      "num_tokens": 11467195.0,
+      "step": 2840
+    },
+    {
+      "entropy": 1.2818018913269043,
+      "epoch": 0.8384819064430715,
+      "grad_norm": 0.517238438129425,
+      "learning_rate": 0.00017258262763497482,
+      "loss": 1.2610112190246583,
+      "mean_token_accuracy": 0.7156127452850342,
+      "num_tokens": 11507742.0,
+      "step": 2850
+    },
+    {
+      "entropy": 1.1540677964687347,
+      "epoch": 0.8414239482200647,
+      "grad_norm": 0.4621984362602234,
+      "learning_rate": 0.00017235916047379383,
+      "loss": 1.187222385406494,
+      "mean_token_accuracy": 0.7345345914363861,
+      "num_tokens": 11548215.0,
+      "step": 2860
+    },
+    {
+      "entropy": 1.2128564953804015,
+      "epoch": 0.8443659899970579,
+      "grad_norm": 0.4317253530025482,
+      "learning_rate": 0.000172134932260695,
+      "loss": 1.2142438888549805,
+      "mean_token_accuracy": 0.7298630118370056,
+      "num_tokens": 11588711.0,
+      "step": 2870
+    },
+    {
+      "entropy": 1.1962445259094239,
+      "epoch": 0.8473080317740512,
+      "grad_norm": 0.6231359839439392,
+      "learning_rate": 0.00017190994535404332,
+      "loss": 1.221367359161377,
+      "mean_token_accuracy": 0.7313773334026337,
+      "num_tokens": 11629162.0,
+      "step": 2880
+    },
+    {
+      "entropy": 1.2900652885437012,
+      "epoch": 0.8502500735510444,
+      "grad_norm": 0.4637579321861267,
+      "learning_rate": 0.00017168420212018354,
+      "loss": 1.2854097366333008,
+      "mean_token_accuracy": 0.7109606087207794,
+      "num_tokens": 11669650.0,
+      "step": 2890
+    },
+    {
+      "entropy": 1.2192873358726501,
+      "epoch": 0.8531921153280376,
+      "grad_norm": 0.4562380909919739,
+      "learning_rate": 0.00017145770493341518,
+      "loss": 1.2436570167541503,
+      "mean_token_accuracy": 0.7263434827327728,
+      "num_tokens": 11710292.0,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8531921153280376,
+      "eval_entropy": 1.2169157832939168,
+      "eval_loss": 1.219910979270935,
+      "eval_mean_token_accuracy": 0.7299123456471861,
+      "eval_num_tokens": 11710292.0,
+      "eval_runtime": 116.9129,
+      "eval_samples_per_second": 26.045,
+      "eval_steps_per_second": 3.259,
+      "step": 2900
+    },
+    {
+      "entropy": 1.2365688323974608,
+      "epoch": 0.8561341571050309,
+      "grad_norm": 0.46011191606521606,
+      "learning_rate": 0.00017123045617596763,
+      "loss": 1.2509427070617676,
+      "mean_token_accuracy": 0.7250486254692078,
+      "num_tokens": 11750649.0,
+      "step": 2910
+    },
+    {
+      "entropy": 1.2337135434150697,
+      "epoch": 0.8590761988820241,
+      "grad_norm": 0.5224452018737793,
+      "learning_rate": 0.00017100245823797503,
+      "loss": 1.2394111633300782,
+      "mean_token_accuracy": 0.729231595993042,
+      "num_tokens": 11791097.0,
+      "step": 2920
+    },
+    {
+      "entropy": 1.2292932152748108,
+      "epoch": 0.8620182406590173,
+      "grad_norm": 0.48644399642944336,
+      "learning_rate": 0.00017077371351745124,
+      "loss": 1.2508816719055176,
+      "mean_token_accuracy": 0.718151307106018,
+      "num_tokens": 11831392.0,
+      "step": 2930
+    },
+    {
+      "entropy": 1.1478191256523131,
+      "epoch": 0.8649602824360106,
+      "grad_norm": 0.4293639063835144,
+      "learning_rate": 0.00017054422442026456,
+      "loss": 1.1457470893859862,
+      "mean_token_accuracy": 0.7468528985977173,
+      "num_tokens": 11871925.0,
+      "step": 2940
+    },
+    {
+      "entropy": 1.2405227303504944,
+      "epoch": 0.8679023242130038,
+      "grad_norm": 0.6136884689331055,
+      "learning_rate": 0.00017031399336011238,
+      "loss": 1.2617270469665527,
+      "mean_token_accuracy": 0.716576772928238,
+      "num_tokens": 11912605.0,
+      "step": 2950
+    },
+    {
+      "entropy": 1.2667588710784912,
+      "epoch": 0.870844365989997,
+      "grad_norm": 0.4269845485687256,
+      "learning_rate": 0.00017008302275849582,
+      "loss": 1.283921241760254,
+      "mean_token_accuracy": 0.7173857808113098,
+      "num_tokens": 11953010.0,
+      "step": 2960
+    },
+    {
+      "entropy": 1.3015403628349305,
+      "epoch": 0.8737864077669902,
+      "grad_norm": 0.46401771903038025,
+      "learning_rate": 0.0001698513150446943,
+      "loss": 1.3019817352294922,
+      "mean_token_accuracy": 0.7130701661109924,
+      "num_tokens": 11993545.0,
+      "step": 2970
+    },
+    {
+      "entropy": 1.2569140315055847,
+      "epoch": 0.8767284495439835,
+      "grad_norm": 0.47312793135643005,
+      "learning_rate": 0.00016961887265574,
+      "loss": 1.279769515991211,
+      "mean_token_accuracy": 0.7160746216773987,
+      "num_tokens": 12033961.0,
+      "step": 2980
+    },
+    {
+      "entropy": 1.253390657901764,
+      "epoch": 0.8796704913209767,
+      "grad_norm": 0.5449343323707581,
+      "learning_rate": 0.0001693856980363921,
+      "loss": 1.259514045715332,
+      "mean_token_accuracy": 0.7175013661384583,
+      "num_tokens": 12074553.0,
+      "step": 2990
+    },
+    {
+      "entropy": 1.3062225699424743,
+      "epoch": 0.8826125330979699,
+      "grad_norm": 0.4499056041240692,
+      "learning_rate": 0.00016915179363911125,
+      "loss": 1.3181246757507323,
+      "mean_token_accuracy": 0.7042843997478485,
+      "num_tokens": 12115075.0,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8826125330979699,
+      "eval_entropy": 1.2066716964789264,
+      "eval_loss": 1.2160207033157349,
+      "eval_mean_token_accuracy": 0.7302276686107706,
+      "eval_num_tokens": 12115075.0,
+      "eval_runtime": 117.081,
+      "eval_samples_per_second": 26.008,
+      "eval_steps_per_second": 3.254,
+      "step": 3000
+    },
+    {
+      "entropy": 1.1849471926689148,
+      "epoch": 0.8855545748749633,
+      "grad_norm": 0.45680174231529236,
+      "learning_rate": 0.00016891716192403365,
+      "loss": 1.18974027633667,
+      "mean_token_accuracy": 0.735182011127472,
+      "num_tokens": 12155441.0,
+      "step": 3010
+    },
+    {
+      "entropy": 1.218670165538788,
+      "epoch": 0.8884966166519565,
+      "grad_norm": 0.4255404770374298,
+      "learning_rate": 0.0001686818053589452,
+      "loss": 1.2337156295776368,
+      "mean_token_accuracy": 0.7265771627426147,
+      "num_tokens": 12196036.0,
+      "step": 3020
+    },
+    {
+      "entropy": 1.2351657152175903,
+      "epoch": 0.8914386584289496,
+      "grad_norm": 0.49140819907188416,
+      "learning_rate": 0.0001684457264192556,
+      "loss": 1.2371573448181152,
+      "mean_token_accuracy": 0.7276750206947327,
+      "num_tokens": 12236509.0,
+      "step": 3030
+    },
+    {
+      "entropy": 1.1386435210704804,
+      "epoch": 0.894380700205943,
+      "grad_norm": 0.4721025824546814,
+      "learning_rate": 0.00016820892758797218,
+      "loss": 1.1611692428588867,
+      "mean_token_accuracy": 0.7427328288555145,
+      "num_tokens": 12276991.0,
+      "step": 3040
+    },
+    {
+      "entropy": 1.2254295825958252,
+      "epoch": 0.8973227419829362,
+      "grad_norm": 0.5237665772438049,
+      "learning_rate": 0.000167971411355674,
+      "loss": 1.2355279922485352,
+      "mean_token_accuracy": 0.7242857456207276,
+      "num_tokens": 12317371.0,
+      "step": 3050
+    },
+    {
+      "entropy": 1.2108076691627503,
+      "epoch": 0.9002647837599294,
+      "grad_norm": 0.46400943398475647,
+      "learning_rate": 0.00016773318022048536,
+      "loss": 1.210099983215332,
+      "mean_token_accuracy": 0.7319884955883026,
+      "num_tokens": 12357779.0,
+      "step": 3060
+    },
+    {
+      "entropy": 1.1315430402755737,
+      "epoch": 0.9032068255369227,
+      "grad_norm": 0.46516865491867065,
+      "learning_rate": 0.00016749423668804988,
+      "loss": 1.160158634185791,
+      "mean_token_accuracy": 0.7439006865024567,
+      "num_tokens": 12398230.0,
+      "step": 3070
+    },
+    {
+      "entropy": 1.2317523241043091,
+      "epoch": 0.9061488673139159,
+      "grad_norm": 0.45237967371940613,
+      "learning_rate": 0.00016725458327150383,
+      "loss": 1.228554630279541,
+      "mean_token_accuracy": 0.728976035118103,
+      "num_tokens": 12438762.0,
+      "step": 3080
+    },
+    {
+      "entropy": 1.1924173712730408,
+      "epoch": 0.9090909090909091,
+      "grad_norm": 0.4868202805519104,
+      "learning_rate": 0.00016701422249144985,
+      "loss": 1.2131217956542968,
+      "mean_token_accuracy": 0.7263190448284149,
+      "num_tokens": 12479195.0,
+      "step": 3090
+    },
+    {
+      "entropy": 1.159307700395584,
+      "epoch": 0.9120329508679024,
+      "grad_norm": 0.44851839542388916,
+      "learning_rate": 0.00016677315687593048,
+      "loss": 1.1793177604675293,
+      "mean_token_accuracy": 0.7422453939914704,
+      "num_tokens": 12518938.0,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9120329508679024,
+      "eval_entropy": 1.193232403965447,
+      "eval_loss": 1.2098932266235352,
+      "eval_mean_token_accuracy": 0.7319249039872737,
+      "eval_num_tokens": 12518938.0,
+      "eval_runtime": 117.0019,
+      "eval_samples_per_second": 26.025,
+      "eval_steps_per_second": 3.256,
+      "step": 3100
+    },
+    {
+      "entropy": 1.2515848875045776,
+      "epoch": 0.9149749926448956,
+      "grad_norm": 0.46064046025276184,
+      "learning_rate": 0.00016653138896040144,
+      "loss": 1.24728364944458,
+      "mean_token_accuracy": 0.7212517559528351,
+      "num_tokens": 12559307.0,
+      "step": 3110
+    },
+    {
+      "entropy": 1.2065839052200318,
+      "epoch": 0.9179170344218888,
+      "grad_norm": 0.4273196756839752,
+      "learning_rate": 0.00016628892128770506,
+      "loss": 1.2376407623291015,
+      "mean_token_accuracy": 0.7283597230911255,
+      "num_tokens": 12599752.0,
+      "step": 3120
+    },
+    {
+      "entropy": 1.245788073539734,
+      "epoch": 0.9208590761988821,
+      "grad_norm": 0.4527672529220581,
+      "learning_rate": 0.0001660457564080435,
+      "loss": 1.2551823616027833,
+      "mean_token_accuracy": 0.7254024922847748,
+      "num_tokens": 12640368.0,
+      "step": 3130
+    },
+    {
+      "entropy": 1.2258686184883119,
+      "epoch": 0.9238011179758753,
+      "grad_norm": 0.4471310079097748,
+      "learning_rate": 0.00016580189687895192,
+      "loss": 1.2315049171447754,
+      "mean_token_accuracy": 0.7266620457172394,
+      "num_tokens": 12681011.0,
+      "step": 3140
+    },
+    {
+      "entropy": 1.2006858110427856,
+      "epoch": 0.9267431597528685,
+      "grad_norm": 0.5084729790687561,
+      "learning_rate": 0.00016555734526527163,
+      "loss": 1.20996150970459,
+      "mean_token_accuracy": 0.733843994140625,
+      "num_tokens": 12721291.0,
+      "step": 3150
+    },
+    {
+      "entropy": 1.240491509437561,
+      "epoch": 0.9296852015298617,
+      "grad_norm": 0.5322201251983643,
+      "learning_rate": 0.000165312104139123,
+      "loss": 1.2419602394104003,
+      "mean_token_accuracy": 0.7259994149208069,
+      "num_tokens": 12761832.0,
+      "step": 3160
+    },
+    {
+      "entropy": 1.2303740501403808,
+      "epoch": 0.932627243306855,
+      "grad_norm": 0.48525163531303406,
+      "learning_rate": 0.00016506617607987863,
+      "loss": 1.266739845275879,
+      "mean_token_accuracy": 0.7198700964450836,
+      "num_tokens": 12802050.0,
+      "step": 3170
+    },
+    {
+      "entropy": 1.2406162559986114,
+      "epoch": 0.9355692850838482,
+      "grad_norm": 0.5300395488739014,
+      "learning_rate": 0.0001648195636741359,
+      "loss": 1.236351776123047,
+      "mean_token_accuracy": 0.7296032607555389,
+      "num_tokens": 12842323.0,
+      "step": 3180
+    },
+    {
+      "entropy": 1.1737658739089967,
+      "epoch": 0.9385113268608414,
+      "grad_norm": 0.580947756767273,
+      "learning_rate": 0.0001645722695156901,
+      "loss": 1.1983850479125977,
+      "mean_token_accuracy": 0.7319401502609253,
+      "num_tokens": 12882948.0,
+      "step": 3190
+    },
+    {
+      "entropy": 1.2107102632522584,
+      "epoch": 0.9414533686378347,
+      "grad_norm": 0.4780050814151764,
+      "learning_rate": 0.00016432429620550688,
+      "loss": 1.1875343322753906,
+      "mean_token_accuracy": 0.7371236264705658,
+      "num_tokens": 12923316.0,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9414533686378347,
+      "eval_entropy": 1.192403563051399,
+      "eval_loss": 1.204405426979065,
+      "eval_mean_token_accuracy": 0.7331006780384094,
+      "eval_num_tokens": 12923316.0,
+      "eval_runtime": 117.0112,
+      "eval_samples_per_second": 26.023,
+      "eval_steps_per_second": 3.256,
+      "step": 3200
+    },
+    {
+      "entropy": 1.1371994316577911,
+      "epoch": 0.9443954104148279,
+      "grad_norm": 0.5067015290260315,
+      "learning_rate": 0.00016407564635169503,
+      "loss": 1.186887264251709,
+      "mean_token_accuracy": 0.7382100522518158,
+      "num_tokens": 12963817.0,
+      "step": 3210
+    },
+    {
+      "entropy": 1.217560636997223,
+      "epoch": 0.9473374521918211,
+      "grad_norm": 0.4357713460922241,
+      "learning_rate": 0.00016382632256947908,
+      "loss": 1.2167092323303224,
+      "mean_token_accuracy": 0.7259755432605743,
+      "num_tokens": 13004277.0,
+      "step": 3220
+    },
+    {
+      "entropy": 1.2078778982162475,
+      "epoch": 0.9502794939688144,
+      "grad_norm": 0.45559161901474,
+      "learning_rate": 0.0001635763274811716,
+      "loss": 1.2228084564208985,
+      "mean_token_accuracy": 0.7331404030323029,
+      "num_tokens": 13044683.0,
+      "step": 3230
+    },
+    {
+      "entropy": 1.210153341293335,
+      "epoch": 0.9532215357458076,
+      "grad_norm": 0.48570674657821655,
+      "learning_rate": 0.00016332566371614595,
+      "loss": 1.2366246223449706,
+      "mean_token_accuracy": 0.7288424909114838,
+      "num_tokens": 13085217.0,
+      "step": 3240
+    },
+    {
+      "entropy": 1.200468325614929,
+      "epoch": 0.9561635775228008,
+      "grad_norm": 0.43572553992271423,
+      "learning_rate": 0.0001630743339108083,
+      "loss": 1.1855230331420898,
+      "mean_token_accuracy": 0.7426514804363251,
+      "num_tokens": 13125762.0,
+      "step": 3250
+    },
+    {
+      "entropy": 1.192023515701294,
+      "epoch": 0.9591056192997941,
+      "grad_norm": 0.6283465623855591,
+      "learning_rate": 0.00016282234070857,
+      "loss": 1.2304601669311523,
+      "mean_token_accuracy": 0.7263703107833862,
+      "num_tokens": 13166141.0,
+      "step": 3260
+    },
+    {
+      "entropy": 1.3121253371238708,
+      "epoch": 0.9620476610767873,
+      "grad_norm": 0.4841027855873108,
+      "learning_rate": 0.0001625696867598199,
+      "loss": 1.3120348930358887,
+      "mean_token_accuracy": 0.7042850077152252,
+      "num_tokens": 13206360.0,
+      "step": 3270
+    },
+    {
+      "entropy": 1.2325164914131164,
+      "epoch": 0.9649897028537805,
+      "grad_norm": 0.5523087978363037,
+      "learning_rate": 0.0001623163747218964,
+      "loss": 1.2477660179138184,
+      "mean_token_accuracy": 0.7236345887184144,
+      "num_tokens": 13246935.0,
+      "step": 3280
+    },
+    {
+      "entropy": 1.2571437239646912,
+      "epoch": 0.9679317446307738,
+      "grad_norm": 0.4331943690776825,
+      "learning_rate": 0.00016206240725905938,
+      "loss": 1.263328742980957,
+      "mean_token_accuracy": 0.7202905654907227,
+      "num_tokens": 13287422.0,
+      "step": 3290
+    },
+    {
+      "entropy": 1.132285052537918,
+      "epoch": 0.970873786407767,
+      "grad_norm": 0.4219360053539276,
+      "learning_rate": 0.00016180778704246238,
+      "loss": 1.1526874542236327,
+      "mean_token_accuracy": 0.745033609867096,
+      "num_tokens": 13327893.0,
+      "step": 3300
+    },
+    {
+      "epoch": 0.970873786407767,
+      "eval_entropy": 1.2065895927233958,
+      "eval_loss": 1.2001736164093018,
+      "eval_mean_token_accuracy": 0.7332249573209467,
+      "eval_num_tokens": 13327893.0,
+      "eval_runtime": 117.023,
+      "eval_samples_per_second": 26.021,
+      "eval_steps_per_second": 3.256,
+      "step": 3300
+    },
+    {
+      "entropy": 1.2010907173156737,
+      "epoch": 0.9738158281847602,
+      "grad_norm": 0.45471927523612976,
+      "learning_rate": 0.00016155251675012433,
+      "loss": 1.2022515296936036,
+      "mean_token_accuracy": 0.7346913456916809,
+      "num_tokens": 13368297.0,
+      "step": 3310
+    },
+    {
+      "entropy": 1.235162889957428,
+      "epoch": 0.9767578699617535,
+      "grad_norm": 0.4965246319770813,
+      "learning_rate": 0.0001612965990669015,
+      "loss": 1.2688727378845215,
+      "mean_token_accuracy": 0.7194641828536987,
+      "num_tokens": 13408602.0,
+      "step": 3320
+    },
+    {
+      "entropy": 1.3277331352233888,
+      "epoch": 0.9796999117387467,
+      "grad_norm": 0.5565961599349976,
+      "learning_rate": 0.00016104003668445925,
+      "loss": 1.3193525314331054,
+      "mean_token_accuracy": 0.7008399486541748,
+      "num_tokens": 13448838.0,
+      "step": 3330
+    },
+    {
+      "entropy": 1.1913153290748597,
+      "epoch": 0.9826419535157399,
+      "grad_norm": 0.49679502844810486,
+      "learning_rate": 0.00016078283230124365,
+      "loss": 1.2235237121582032,
+      "mean_token_accuracy": 0.7243378221988678,
+      "num_tokens": 13489366.0,
+      "step": 3340
+    },
+    {
+      "entropy": 1.2956088483333588,
+      "epoch": 0.9855839952927331,
+      "grad_norm": 0.49954238533973694,
+      "learning_rate": 0.00016052498862245313,
+      "loss": 1.2841950416564942,
+      "mean_token_accuracy": 0.7132591784000397,
+      "num_tokens": 13529414.0,
+      "step": 3350
+    },
+    {
+      "entropy": 1.182485854625702,
+      "epoch": 0.9885260370697264,
+      "grad_norm": 0.4281366169452667,
+      "learning_rate": 0.00016026650836001012,
+      "loss": 1.2035736083984374,
+      "mean_token_accuracy": 0.7331153571605682,
+      "num_tokens": 13569904.0,
+      "step": 3360
+    },
+    {
+      "entropy": 1.2611793637275697,
+      "epoch": 0.9914680788467196,
+      "grad_norm": 0.4736562669277191,
+      "learning_rate": 0.0001600073942325323,
+      "loss": 1.2855722427368164,
+      "mean_token_accuracy": 0.7184321641921997,
+      "num_tokens": 13610365.0,
+      "step": 3370
+    },
+    {
+      "entropy": 1.2419211864471436,
+      "epoch": 0.9944101206237128,
+      "grad_norm": 0.45400166511535645,
+      "learning_rate": 0.00015974764896530433,
+      "loss": 1.2359369277954102,
+      "mean_token_accuracy": 0.7274314403533936,
+      "num_tokens": 13650738.0,
+      "step": 3380
+    },
+    {
+      "entropy": 1.16963592171669,
+      "epoch": 0.9973521624007061,
+      "grad_norm": 0.5148010849952698,
+      "learning_rate": 0.0001594872752902489,
+      "loss": 1.1966312408447266,
+      "mean_token_accuracy": 0.7361967086791992,
+      "num_tokens": 13691179.0,
+      "step": 3390
+    },
+    {
+      "entropy": 1.263899064064026,
+      "epoch": 1.0002942041776994,
+      "grad_norm": 0.47132888436317444,
+      "learning_rate": 0.0001592262759458981,
+      "loss": 1.2513002395629882,
+      "mean_token_accuracy": 0.7206339836120605,
+      "num_tokens": 13729142.0,
+      "step": 3400
+    },
+    {
+      "epoch": 1.0002942041776994,
+      "eval_entropy": 1.1728849974204236,
+      "eval_loss": 1.196743369102478,
+      "eval_mean_token_accuracy": 0.7349204848131795,
+      "eval_num_tokens": 13729142.0,
+      "eval_runtime": 116.9564,
+      "eval_samples_per_second": 26.035,
+      "eval_steps_per_second": 3.258,
+      "step": 3400
+    },
+    {
+      "entropy": 1.0056753396987914,
+      "epoch": 1.0032362459546926,
+      "grad_norm": 0.48898911476135254,
+      "learning_rate": 0.00015896465367736467,
+      "loss": 0.9896906852722168,
+      "mean_token_accuracy": 0.7708241283893585,
+      "num_tokens": 13769748.0,
+      "step": 3410
+    },
+    {
+      "entropy": 0.9479109048843384,
+      "epoch": 1.0061782877316858,
+      "grad_norm": 0.5644809603691101,
+      "learning_rate": 0.00015870241123631303,
+      "loss": 0.969275951385498,
+      "mean_token_accuracy": 0.7748298406600952,
+      "num_tokens": 13810433.0,
+      "step": 3420
+    },
+    {
+      "entropy": 1.0169321179389954,
+      "epoch": 1.009120329508679,
+      "grad_norm": 0.5554332733154297,
+      "learning_rate": 0.00015843955138093043,
+      "loss": 1.0067197799682617,
+      "mean_token_accuracy": 0.7651751101016998,
+      "num_tokens": 13850895.0,
+      "step": 3430
+    },
+    {
+      "entropy": 0.9419119358062744,
+      "epoch": 1.0120623712856722,
+      "grad_norm": 0.5447728037834167,
+      "learning_rate": 0.00015817607687589787,
+      "loss": 0.9617524147033691,
+      "mean_token_accuracy": 0.771773511171341,
+      "num_tokens": 13890924.0,
+      "step": 3440
+    },
+    {
+      "entropy": 0.9768797099590302,
+      "epoch": 1.0150044130626654,
+      "grad_norm": 0.5598154664039612,
+      "learning_rate": 0.00015791199049236106,
+      "loss": 0.984192180633545,
+      "mean_token_accuracy": 0.7716309785842895,
+      "num_tokens": 13931279.0,
+      "step": 3450
+    },
+    {
+      "entropy": 0.9213016629219055,
+      "epoch": 1.0179464548396586,
+      "grad_norm": 0.5594082474708557,
+      "learning_rate": 0.00015764729500790132,
+      "loss": 0.925694465637207,
+      "mean_token_accuracy": 0.7817917823791504,
+      "num_tokens": 13971714.0,
+      "step": 3460
+    },
+    {
+      "entropy": 0.9738560080528259,
+      "epoch": 1.020888496616652,
+      "grad_norm": 0.576998233795166,
+      "learning_rate": 0.00015738199320650622,
+      "loss": 0.9819319725036622,
+      "mean_token_accuracy": 0.7748230636119843,
+      "num_tokens": 14012038.0,
+      "step": 3470
+    },
+    {
+      "entropy": 0.9529994606971741,
+      "epoch": 1.0238305383936452,
+      "grad_norm": 0.507428765296936,
+      "learning_rate": 0.00015711608787854041,
+      "loss": 0.968116569519043,
+      "mean_token_accuracy": 0.7762803137302399,
+      "num_tokens": 14052608.0,
+      "step": 3480
+    },
+    {
+      "entropy": 0.9343097984790802,
+      "epoch": 1.0267725801706384,
+      "grad_norm": 0.48517364263534546,
+      "learning_rate": 0.0001568495818207163,
+      "loss": 0.9297596931457519,
+      "mean_token_accuracy": 0.783079195022583,
+      "num_tokens": 14093119.0,
+      "step": 3490
+    },
+    {
+      "entropy": 0.9860086500644684,
+      "epoch": 1.0297146219476316,
+      "grad_norm": 0.4717691242694855,
+      "learning_rate": 0.00015658247783606455,
+      "loss": 1.004935073852539,
+      "mean_token_accuracy": 0.765727037191391,
+      "num_tokens": 14133569.0,
+      "step": 3500
+    },
+    {
+      "epoch": 1.0297146219476316,
+      "eval_entropy": 1.0715967291609196,
+      "eval_loss": 1.2108769416809082,
+      "eval_mean_token_accuracy": 0.7342690184047529,
+      "eval_num_tokens": 14133569.0,
+      "eval_runtime": 116.9449,
+      "eval_samples_per_second": 26.038,
+      "eval_steps_per_second": 3.258,
+      "step": 3500
+    },
+    {
+      "entropy": 0.9845475435256958,
+      "epoch": 1.0326566637246248,
+      "grad_norm": 0.6196256875991821,
+      "learning_rate": 0.00015631477873390463,
+      "loss": 0.9737442970275879,
+      "mean_token_accuracy": 0.7720641791820526,
+      "num_tokens": 14174112.0,
+      "step": 3510
+    },
+    {
+      "entropy": 0.9606890559196473,
+      "epoch": 1.035598705501618,
+      "grad_norm": 0.5352163910865784,
+      "learning_rate": 0.00015604648732981535,
+      "loss": 0.9936755180358887,
+      "mean_token_accuracy": 0.767872554063797,
+      "num_tokens": 14214732.0,
+      "step": 3520
+    },
+    {
+      "entropy": 0.9683861494064331,
+      "epoch": 1.0385407472786115,
+      "grad_norm": 0.5286763906478882,
+      "learning_rate": 0.00015577760644560506,
+      "loss": 0.9670245170593261,
+      "mean_token_accuracy": 0.7790086328983307,
+      "num_tokens": 14255029.0,
+      "step": 3530
+    },
+    {
+      "entropy": 0.9281998634338379,
+      "epoch": 1.0414827890556047,
+      "grad_norm": 0.5544895529747009,
+      "learning_rate": 0.0001555081389092822,
+      "loss": 0.9275782585144043,
+      "mean_token_accuracy": 0.7800623893737793,
+      "num_tokens": 14294859.0,
+      "step": 3540
+    },
+    {
+      "entropy": 0.9516816318035126,
+      "epoch": 1.0444248308325979,
+      "grad_norm": 0.5528038144111633,
+      "learning_rate": 0.0001552380875550253,
+      "loss": 0.971556282043457,
+      "mean_token_accuracy": 0.7744876265525817,
+      "num_tokens": 14335364.0,
+      "step": 3550
+    },
+    {
+      "entropy": 1.0375867664813996,
+      "epoch": 1.047366872609591,
+      "grad_norm": 0.5946918725967407,
+      "learning_rate": 0.00015496745522315352,
+      "loss": 1.036135482788086,
+      "mean_token_accuracy": 0.7563816487789154,
+      "num_tokens": 14375785.0,
+      "step": 3560
+    },
+    {
+      "entropy": 0.9602185845375061,
+      "epoch": 1.0503089143865842,
+      "grad_norm": 0.55278080701828,
+      "learning_rate": 0.00015469624476009637,
+      "loss": 0.9790426254272461,
+      "mean_token_accuracy": 0.7711789608001709,
+      "num_tokens": 14416298.0,
+      "step": 3570
+    },
+    {
+      "entropy": 0.9802630722522736,
+      "epoch": 1.0532509561635774,
+      "grad_norm": 0.5366263389587402,
+      "learning_rate": 0.00015442445901836407,
+      "loss": 0.9828317642211915,
+      "mean_token_accuracy": 0.7732825756072998,
+      "num_tokens": 14456758.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.9827463984489441,
+      "epoch": 1.0561929979405709,
+      "grad_norm": 0.5232464075088501,
+      "learning_rate": 0.0001541521008565174,
+      "loss": 0.9965373039245605,
+      "mean_token_accuracy": 0.7664778053760528,
+      "num_tokens": 14497079.0,
+      "step": 3590
+    },
+    {
+      "entropy": 0.9974855959415436,
+      "epoch": 1.059135039717564,
+      "grad_norm": 0.5640583634376526,
+      "learning_rate": 0.0001538791731391377,
+      "loss": 1.001423168182373,
+      "mean_token_accuracy": 0.7696728587150574,
+      "num_tokens": 14537442.0,
+      "step": 3600
+    },
+    {
+      "epoch": 1.059135039717564,
+      "eval_entropy": 1.059611619926813,
+      "eval_loss": 1.2121630907058716,
+      "eval_mean_token_accuracy": 0.7346383677379978,
+      "eval_num_tokens": 14537442.0,
+      "eval_runtime": 116.9964,
+      "eval_samples_per_second": 26.026,
+      "eval_steps_per_second": 3.257,
+      "step": 3600
+    },
+    {
+      "entropy": 1.019939649105072,
+      "epoch": 1.0620770814945573,
+      "grad_norm": 0.6510360836982727,
+      "learning_rate": 0.00015360567873679682,
+      "loss": 1.0335427284240724,
+      "mean_token_accuracy": 0.7557238221168519,
+      "num_tokens": 14577903.0,
+      "step": 3610
+    },
+    {
+      "entropy": 0.9929381251335144,
+      "epoch": 1.0650191232715505,
+      "grad_norm": 0.4985535740852356,
+      "learning_rate": 0.00015333162052602663,
+      "loss": 0.9860858917236328,
+      "mean_token_accuracy": 0.7693518400192261,
+      "num_tokens": 14618362.0,
+      "step": 3620
+    },
+    {
+      "entropy": 1.0021346986293793,
+      "epoch": 1.0679611650485437,
+      "grad_norm": 0.5036218762397766,
+      "learning_rate": 0.00015305700138928914,
+      "loss": 1.0195841789245605,
+      "mean_token_accuracy": 0.7599446833133697,
+      "num_tokens": 14658524.0,
+      "step": 3630
+    },
+    {
+      "entropy": 0.9681568443775177,
+      "epoch": 1.0709032068255369,
+      "grad_norm": 0.685461699962616,
+      "learning_rate": 0.00015278182421494597,
+      "loss": 0.9650713920593261,
+      "mean_token_accuracy": 0.7742046892642975,
+      "num_tokens": 14699023.0,
+      "step": 3640
+    },
+    {
+      "entropy": 0.9974366188049316,
+      "epoch": 1.07384524860253,
+      "grad_norm": 0.5122579336166382,
+      "learning_rate": 0.0001525060918972279,
+      "loss": 1.010261631011963,
+      "mean_token_accuracy": 0.7668596982955933,
+      "num_tokens": 14739433.0,
+      "step": 3650
+    },
+    {
+      "entropy": 0.9449751615524292,
+      "epoch": 1.0767872903795235,
+      "grad_norm": 0.5020188689231873,
+      "learning_rate": 0.00015222980733620473,
+      "loss": 0.9527727127075195,
+      "mean_token_accuracy": 0.7792690694332123,
+      "num_tokens": 14779729.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.9212399244308471,
+      "epoch": 1.0797293321565167,
+      "grad_norm": 0.4754965901374817,
+      "learning_rate": 0.0001519529734377545,
+      "loss": 0.931304931640625,
+      "mean_token_accuracy": 0.780491977930069,
+      "num_tokens": 14819988.0,
+      "step": 3670
+    },
+    {
+      "entropy": 0.9701619386672974,
+      "epoch": 1.0826713739335099,
+      "grad_norm": 0.51582270860672,
+      "learning_rate": 0.0001516755931135329,
+      "loss": 0.9845802307128906,
+      "mean_token_accuracy": 0.7685158431529999,
+      "num_tokens": 14860489.0,
+      "step": 3680
+    },
+    {
+      "entropy": 0.9930408954620361,
+      "epoch": 1.085613415710503,
+      "grad_norm": 0.5440912246704102,
+      "learning_rate": 0.00015139766928094303,
+      "loss": 0.9996217727661133,
+      "mean_token_accuracy": 0.7675297498703003,
+      "num_tokens": 14901041.0,
+      "step": 3690
+    },
+    {
+      "entropy": 0.9450788199901581,
+      "epoch": 1.0885554574874963,
+      "grad_norm": 0.5000481605529785,
+      "learning_rate": 0.00015111920486310417,
+      "loss": 0.9332949638366699,
+      "mean_token_accuracy": 0.7813515663146973,
+      "num_tokens": 14941532.0,
+      "step": 3700
+    },
+    {
+      "epoch": 1.0885554574874963,
+      "eval_entropy": 1.0368518566522074,
+      "eval_loss": 1.209765076637268,
+      "eval_mean_token_accuracy": 0.7356539055744181,
+      "eval_num_tokens": 14941532.0,
+      "eval_runtime": 117.1062,
+      "eval_samples_per_second": 26.002,
+      "eval_steps_per_second": 3.253,
+      "step": 3700
+    },
+    {
+      "entropy": 0.9497840762138366,
+      "epoch": 1.0914974992644895,
+      "grad_norm": 0.5802638530731201,
+      "learning_rate": 0.00015084020278882153,
+      "loss": 0.9746996879577636,
+      "mean_token_accuracy": 0.7703352689743042,
+      "num_tokens": 14981818.0,
+      "step": 3710
+    },
+    {
+      "entropy": 0.9469231545925141,
+      "epoch": 1.0944395410414829,
+      "grad_norm": 0.5750518441200256,
+      "learning_rate": 0.00015056066599255502,
+      "loss": 0.9662343025207519,
+      "mean_token_accuracy": 0.7726805984973908,
+      "num_tokens": 15022307.0,
+      "step": 3720
+    },
+    {
+      "entropy": 1.0127877593040466,
+      "epoch": 1.097381582818476,
+      "grad_norm": 0.5340924859046936,
+      "learning_rate": 0.0001502805974143888,
+      "loss": 0.9845216751098633,
+      "mean_token_accuracy": 0.7723982453346252,
+      "num_tokens": 15062810.0,
+      "step": 3730
+    },
+    {
+      "entropy": 0.974898761510849,
+      "epoch": 1.1003236245954693,
+      "grad_norm": 0.5145857334136963,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 1.0081477165222168,
+      "mean_token_accuracy": 0.7643806636333466,
+      "num_tokens": 15103409.0,
+      "step": 3740
+    },
+    {
+      "entropy": 0.9510980069637298,
+      "epoch": 1.1032656663724625,
+      "grad_norm": 0.5267722606658936,
+      "learning_rate": 0.00014971887670062802,
+      "loss": 0.9365800857543946,
+      "mean_token_accuracy": 0.780946570634842,
+      "num_tokens": 15143947.0,
+      "step": 3750
+    },
+    {
+      "entropy": 1.0326339960098267,
+      "epoch": 1.1062077081494557,
+      "grad_norm": 0.517667293548584,
+      "learning_rate": 0.0001494372304730432,
+      "loss": 1.0606879234313964,
+      "mean_token_accuracy": 0.7520448863506317,
+      "num_tokens": 15184466.0,
+      "step": 3760
+    },
+    {
+      "entropy": 0.9095244884490967,
+      "epoch": 1.1091497499264489,
+      "grad_norm": 0.5059592127799988,
+      "learning_rate": 0.00014915506427951605,
+      "loss": 0.9141671180725097,
+      "mean_token_accuracy": 0.7883690059185028,
+      "num_tokens": 15224876.0,
+      "step": 3770
+    },
+    {
+      "entropy": 0.922091954946518,
+      "epoch": 1.1120917917034423,
+      "grad_norm": 0.5040479898452759,
+      "learning_rate": 0.0001488723810877858,
+      "loss": 0.9369124412536621,
+      "mean_token_accuracy": 0.7827723801136017,
+      "num_tokens": 15265334.0,
+      "step": 3780
+    },
+    {
+      "entropy": 0.9985373139381408,
+      "epoch": 1.1150338334804355,
+      "grad_norm": 0.593623697757721,
+      "learning_rate": 0.00014858918387102943,
+      "loss": 1.003388023376465,
+      "mean_token_accuracy": 0.7662831544876099,
+      "num_tokens": 15305821.0,
+      "step": 3790
+    },
+    {
+      "entropy": 1.0036361336708068,
+      "epoch": 1.1179758752574287,
+      "grad_norm": 0.46527841687202454,
+      "learning_rate": 0.00014830547560783013,
+      "loss": 1.0003900527954102,
+      "mean_token_accuracy": 0.7689646422863007,
+      "num_tokens": 15346212.0,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1179758752574287,
+      "eval_entropy": 1.058281412587704,
+      "eval_loss": 1.207934856414795,
+      "eval_mean_token_accuracy": 0.7360553802467706,
+      "eval_num_tokens": 15346212.0,
+      "eval_runtime": 117.1795,
+      "eval_samples_per_second": 25.986,
+      "eval_steps_per_second": 3.251,
+      "step": 3800
+    },
+    {
+      "entropy": 1.0539171755313874,
+      "epoch": 1.120917917034422,
+      "grad_norm": 0.5680721402168274,
+      "learning_rate": 0.00014802125928214626,
+      "loss": 1.0743833541870118,
+      "mean_token_accuracy": 0.7493933796882629,
+      "num_tokens": 15386724.0,
+      "step": 3810
+    },
+    {
+      "entropy": 0.916973739862442,
+      "epoch": 1.123859958811415,
+      "grad_norm": 0.5414533019065857,
+      "learning_rate": 0.0001477365378832797,
+      "loss": 0.9400577545166016,
+      "mean_token_accuracy": 0.7826773941516876,
+      "num_tokens": 15427047.0,
+      "step": 3820
+    },
+    {
+      "entropy": 1.044243198633194,
+      "epoch": 1.1268020005884083,
+      "grad_norm": 0.5607292056083679,
+      "learning_rate": 0.0001474513144058447,
+      "loss": 1.0428730964660644,
+      "mean_token_accuracy": 0.7559836447238922,
+      "num_tokens": 15467253.0,
+      "step": 3830
+    },
+    {
+      "entropy": 0.9271435976028443,
+      "epoch": 1.1297440423654015,
+      "grad_norm": 0.5749255418777466,
+      "learning_rate": 0.0001471655918497361,
+      "loss": 0.9598716735839844,
+      "mean_token_accuracy": 0.7791375935077667,
+      "num_tokens": 15507806.0,
+      "step": 3840
+    },
+    {
+      "entropy": 0.9984059453010559,
+      "epoch": 1.132686084142395,
+      "grad_norm": 0.671238362789154,
+      "learning_rate": 0.00014687937322009793,
+      "loss": 0.9737402915954589,
+      "mean_token_accuracy": 0.7700112521648407,
+      "num_tokens": 15548383.0,
+      "step": 3850
+    },
+    {
+      "entropy": 0.9276021301746369,
+      "epoch": 1.135628125919388,
+      "grad_norm": 0.5724707841873169,
+      "learning_rate": 0.00014659266152729176,
+      "loss": 0.9540791511535645,
+      "mean_token_accuracy": 0.7781682848930359,
+      "num_tokens": 15588893.0,
+      "step": 3860
+    },
+    {
+      "entropy": 0.9864838421344757,
+      "epoch": 1.1385701676963813,
+      "grad_norm": 0.5926504731178284,
+      "learning_rate": 0.0001463054597868651,
+      "loss": 0.9837197303771973,
+      "mean_token_accuracy": 0.7700894057750702,
+      "num_tokens": 15629429.0,
+      "step": 3870
+    },
+    {
+      "entropy": 1.044099175930023,
+      "epoch": 1.1415122094733745,
+      "grad_norm": 0.5998210906982422,
+      "learning_rate": 0.00014601777101951957,
+      "loss": 1.064276123046875,
+      "mean_token_accuracy": 0.7496702373027802,
+      "num_tokens": 15669112.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.9448180139064789,
+      "epoch": 1.1444542512503677,
+      "grad_norm": 0.5697433948516846,
+      "learning_rate": 0.00014572959825107922,
+      "loss": 0.939006233215332,
+      "mean_token_accuracy": 0.7818219542503357,
+      "num_tokens": 15709418.0,
+      "step": 3890
+    },
+    {
+      "entropy": 0.9624788880348205,
+      "epoch": 1.147396293027361,
+      "grad_norm": 0.5536630749702454,
+      "learning_rate": 0.0001454409445124587,
+      "loss": 0.9612356185913086,
+      "mean_token_accuracy": 0.7735930144786834,
+      "num_tokens": 15749926.0,
+      "step": 3900
+    },
+    {
+      "epoch": 1.147396293027361,
+      "eval_entropy": 1.002416886056815,
+      "eval_loss": 1.2163069248199463,
+      "eval_mean_token_accuracy": 0.736245130147208,
+      "eval_num_tokens": 15749926.0,
+      "eval_runtime": 117.0848,
+      "eval_samples_per_second": 26.007,
+      "eval_steps_per_second": 3.254,
+      "step": 3900
+    },
+    {
+      "entropy": 0.9970716178417206,
+      "epoch": 1.150338334804354,
+      "grad_norm": 0.551414966583252,
+      "learning_rate": 0.00014515181283963132,
+      "loss": 1.024658489227295,
+      "mean_token_accuracy": 0.7571725130081177,
+      "num_tokens": 15790231.0,
+      "step": 3910
+    },
+    {
+      "entropy": 1.0260444402694702,
+      "epoch": 1.1532803765813475,
+      "grad_norm": 0.565596878528595,
+      "learning_rate": 0.0001448622062735972,
+      "loss": 1.0220839500427246,
+      "mean_token_accuracy": 0.7611811697483063,
+      "num_tokens": 15830791.0,
+      "step": 3920
+    },
+    {
+      "entropy": 0.9745386719703675,
+      "epoch": 1.1562224183583407,
+      "grad_norm": 0.522496223449707,
+      "learning_rate": 0.00014457212786035122,
+      "loss": 0.9898022651672364,
+      "mean_token_accuracy": 0.7702635705471039,
+      "num_tokens": 15871330.0,
+      "step": 3930
+    },
+    {
+      "entropy": 1.00519802570343,
+      "epoch": 1.159164460135334,
+      "grad_norm": 0.579716145992279,
+      "learning_rate": 0.00014428158065085098,
+      "loss": 1.0124700546264649,
+      "mean_token_accuracy": 0.7646716058254241,
+      "num_tokens": 15911828.0,
+      "step": 3940
+    },
+    {
+      "entropy": 0.9504435777664184,
+      "epoch": 1.1621065019123271,
+      "grad_norm": 0.5688005089759827,
+      "learning_rate": 0.00014399056770098478,
+      "loss": 0.9538630485534668,
+      "mean_token_accuracy": 0.776702755689621,
+      "num_tokens": 15952424.0,
+      "step": 3950
+    },
+    {
+      "entropy": 0.972985816001892,
+      "epoch": 1.1650485436893203,
+      "grad_norm": 0.5217563509941101,
+      "learning_rate": 0.00014369909207153947,
+      "loss": 0.9882010459899903,
+      "mean_token_accuracy": 0.7707372605800629,
+      "num_tokens": 15992995.0,
+      "step": 3960
+    },
+    {
+      "entropy": 1.0266568660736084,
+      "epoch": 1.1679905854663137,
+      "grad_norm": 0.607991099357605,
+      "learning_rate": 0.00014340715682816806,
+      "loss": 1.0269791603088378,
+      "mean_token_accuracy": 0.7627450406551362,
+      "num_tokens": 16033357.0,
+      "step": 3970
+    },
+    {
+      "entropy": 1.023914647102356,
+      "epoch": 1.170932627243307,
+      "grad_norm": 0.5779170393943787,
+      "learning_rate": 0.00014311476504135794,
+      "loss": 1.0473231315612792,
+      "mean_token_accuracy": 0.7546798884868622,
+      "num_tokens": 16073799.0,
+      "step": 3980
+    },
+    {
+      "entropy": 0.9414635598659515,
+      "epoch": 1.1738746690203001,
+      "grad_norm": 0.5319372415542603,
+      "learning_rate": 0.00014282191978639799,
+      "loss": 0.9280409812927246,
+      "mean_token_accuracy": 0.7800300478935241,
+      "num_tokens": 16114313.0,
+      "step": 3990
+    },
+    {
+      "entropy": 0.9413889229297638,
+      "epoch": 1.1768167107972933,
+      "grad_norm": 0.6013742089271545,
+      "learning_rate": 0.00014252862414334665,
+      "loss": 0.9760286331176757,
+      "mean_token_accuracy": 0.7707565903663636,
+      "num_tokens": 16154854.0,
+      "step": 4000
+    },
+    {
+      "epoch": 1.1768167107972933,
+      "eval_entropy": 1.0498265117954395,
+      "eval_loss": 1.2053319215774536,
+      "eval_mean_token_accuracy": 0.7371757869019596,
+      "eval_num_tokens": 16154854.0,
+      "eval_runtime": 116.9953,
+      "eval_samples_per_second": 26.027,
+      "eval_steps_per_second": 3.257,
+      "step": 4000
+    },
+    {
+      "entropy": 1.01264528632164,
+      "epoch": 1.1797587525742865,
+      "grad_norm": 0.5052332282066345,
+      "learning_rate": 0.00014223488119699944,
+      "loss": 0.9941823959350586,
+      "mean_token_accuracy": 0.7712079107761383,
+      "num_tokens": 16195404.0,
+      "step": 4010
+    },
+    {
+      "entropy": 0.9542388617992401,
+      "epoch": 1.1827007943512797,
+      "grad_norm": 0.6624171733856201,
+      "learning_rate": 0.00014194069403685643,
+      "loss": 0.9699134826660156,
+      "mean_token_accuracy": 0.7701153516769409,
+      "num_tokens": 16235933.0,
+      "step": 4020
+    },
+    {
+      "entropy": 0.9835664987564087,
+      "epoch": 1.185642836128273,
+      "grad_norm": 0.6638393998146057,
+      "learning_rate": 0.00014164606575708984,
+      "loss": 1.004053497314453,
+      "mean_token_accuracy": 0.7658522069454193,
+      "num_tokens": 16276402.0,
+      "step": 4030
+    },
+    {
+      "entropy": 0.9395282685756683,
+      "epoch": 1.1885848779052663,
+      "grad_norm": 0.5497994422912598,
+      "learning_rate": 0.0001413509994565114,
+      "loss": 0.9349452972412109,
+      "mean_token_accuracy": 0.7839109897613525,
+      "num_tokens": 16316751.0,
+      "step": 4040
+    },
+    {
+      "entropy": 0.9585356175899505,
+      "epoch": 1.1915269196822595,
+      "grad_norm": 0.5739843845367432,
+      "learning_rate": 0.00014105549823853987,
+      "loss": 0.9759317398071289,
+      "mean_token_accuracy": 0.7685169577598572,
+      "num_tokens": 16357253.0,
+      "step": 4050
+    },
+    {
+      "entropy": 0.9427942156791687,
+      "epoch": 1.1944689614592527,
+      "grad_norm": 0.5962517261505127,
+      "learning_rate": 0.00014075956521116827,
+      "loss": 0.9541123390197754,
+      "mean_token_accuracy": 0.7789060473442078,
+      "num_tokens": 16397470.0,
+      "step": 4060
+    },
+    {
+      "entropy": 0.9317695260047912,
+      "epoch": 1.197411003236246,
+      "grad_norm": 0.5742291212081909,
+      "learning_rate": 0.00014046320348693134,
+      "loss": 0.9425789833068847,
+      "mean_token_accuracy": 0.7804486751556396,
+      "num_tokens": 16438035.0,
+      "step": 4070
+    },
+    {
+      "entropy": 0.9994288563728333,
+      "epoch": 1.2003530450132391,
+      "grad_norm": 0.628220796585083,
+      "learning_rate": 0.00014016641618287264,
+      "loss": 1.0083752632141114,
+      "mean_token_accuracy": 0.7678903639316559,
+      "num_tokens": 16478563.0,
+      "step": 4080
+    },
+    {
+      "entropy": 0.9661390006542205,
+      "epoch": 1.2032950867902323,
+      "grad_norm": 0.5717418789863586,
+      "learning_rate": 0.00013986920642051196,
+      "loss": 0.956269645690918,
+      "mean_token_accuracy": 0.7756177723407746,
+      "num_tokens": 16519059.0,
+      "step": 4090
+    },
+    {
+      "entropy": 0.9637384474277496,
+      "epoch": 1.2062371285672255,
+      "grad_norm": 0.5957377552986145,
+      "learning_rate": 0.00013957157732581227,
+      "loss": 0.9740910530090332,
+      "mean_token_accuracy": 0.7705820441246033,
+      "num_tokens": 16559585.0,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2062371285672255,
+      "eval_entropy": 1.0279845802646297,
+      "eval_loss": 1.204230546951294,
+      "eval_mean_token_accuracy": 0.7378412169108554,
+      "eval_num_tokens": 16559585.0,
+      "eval_runtime": 116.9158,
+      "eval_samples_per_second": 26.044,
+      "eval_steps_per_second": 3.259,
+      "step": 4100
+    },
+    {
+      "entropy": 0.9299012839794158,
+      "epoch": 1.209179170344219,
+      "grad_norm": 0.5047076344490051,
+      "learning_rate": 0.00013927353202914704,
+      "loss": 0.9325620651245117,
+      "mean_token_accuracy": 0.7856215178966522,
+      "num_tokens": 16599897.0,
+      "step": 4110
+    },
+    {
+      "entropy": 0.9579800009727478,
+      "epoch": 1.2121212121212122,
+      "grad_norm": 0.6236160397529602,
+      "learning_rate": 0.00013897507366526717,
+      "loss": 0.9825181007385254,
+      "mean_token_accuracy": 0.771970808506012,
+      "num_tokens": 16640445.0,
+      "step": 4120
+    },
+    {
+      "entropy": 0.9490958392620087,
+      "epoch": 1.2150632538982054,
+      "grad_norm": 0.5913335084915161,
+      "learning_rate": 0.00013867620537326807,
+      "loss": 0.9593384742736817,
+      "mean_token_accuracy": 0.7774133384227753,
+      "num_tokens": 16681074.0,
+      "step": 4130
+    },
+    {
+      "entropy": 0.9436961114406586,
+      "epoch": 1.2180052956751986,
+      "grad_norm": 0.5566667318344116,
+      "learning_rate": 0.00013837693029655673,
+      "loss": 0.9432112693786621,
+      "mean_token_accuracy": 0.7793719172477722,
+      "num_tokens": 16721515.0,
+      "step": 4140
+    },
+    {
+      "entropy": 0.9555536270141601,
+      "epoch": 1.2209473374521917,
+      "grad_norm": 0.5549019575119019,
+      "learning_rate": 0.00013807725158281845,
+      "loss": 0.964748764038086,
+      "mean_token_accuracy": 0.7745252251625061,
+      "num_tokens": 16761971.0,
+      "step": 4150
+    },
+    {
+      "entropy": 1.0964530289173127,
+      "epoch": 1.2238893792291852,
+      "grad_norm": 0.5607911944389343,
+      "learning_rate": 0.000137777172383984,
+      "loss": 1.1134785652160644,
+      "mean_token_accuracy": 0.7422545254230499,
+      "num_tokens": 16801992.0,
+      "step": 4160
+    },
+    {
+      "entropy": 0.991803640127182,
+      "epoch": 1.2268314210061784,
+      "grad_norm": 0.5912790894508362,
+      "learning_rate": 0.00013747669585619621,
+      "loss": 0.9909832954406739,
+      "mean_token_accuracy": 0.7725964307785034,
+      "num_tokens": 16842361.0,
+      "step": 4170
+    },
+    {
+      "entropy": 0.9360088646411896,
+      "epoch": 1.2297734627831716,
+      "grad_norm": 0.5747997760772705,
+      "learning_rate": 0.00013717582515977703,
+      "loss": 0.9449604988098145,
+      "mean_token_accuracy": 0.7815338909626007,
+      "num_tokens": 16882838.0,
+      "step": 4180
+    },
+    {
+      "entropy": 0.9765557944774628,
+      "epoch": 1.2327155045601648,
+      "grad_norm": 0.6022299528121948,
+      "learning_rate": 0.000136874563459194,
+      "loss": 0.9870802879333496,
+      "mean_token_accuracy": 0.7712435305118561,
+      "num_tokens": 16923293.0,
+      "step": 4190
+    },
+    {
+      "entropy": 0.9462984323501586,
+      "epoch": 1.235657546337158,
+      "grad_norm": 0.6731317639350891,
+      "learning_rate": 0.0001365729139230273,
+      "loss": 0.9600887298583984,
+      "mean_token_accuracy": 0.776117742061615,
+      "num_tokens": 16963785.0,
+      "step": 4200
+    },
+    {
+      "epoch": 1.235657546337158,
+      "eval_entropy": 1.0305638493984703,
+      "eval_loss": 1.2021441459655762,
+      "eval_mean_token_accuracy": 0.7379450076834111,
+      "eval_num_tokens": 16963785.0,
+      "eval_runtime": 117.0903,
+      "eval_samples_per_second": 26.006,
+      "eval_steps_per_second": 3.254,
+      "step": 4200
+    },
+    {
+      "entropy": 0.8939768195152282,
+      "epoch": 1.2385995881141512,
+      "grad_norm": 0.6204389333724976,
+      "learning_rate": 0.00013627087972393605,
+      "loss": 0.8930576324462891,
+      "mean_token_accuracy": 0.7922165811061859,
+      "num_tokens": 17004178.0,
+      "step": 4210
+    },
+    {
+      "entropy": 0.9507308840751648,
+      "epoch": 1.2415416298911444,
+      "grad_norm": 0.6163705587387085,
+      "learning_rate": 0.00013596846403862535,
+      "loss": 0.9666755676269532,
+      "mean_token_accuracy": 0.7786275684833527,
+      "num_tokens": 17044627.0,
+      "step": 4220
+    },
+    {
+      "entropy": 0.975046980381012,
+      "epoch": 1.2444836716681378,
+      "grad_norm": 0.5746111273765564,
+      "learning_rate": 0.00013566567004781246,
+      "loss": 0.9781759262084961,
+      "mean_token_accuracy": 0.7753113329410553,
+      "num_tokens": 17084732.0,
+      "step": 4230
+    },
+    {
+      "entropy": 1.0526322185993195,
+      "epoch": 1.247425713445131,
+      "grad_norm": 0.5716729164123535,
+      "learning_rate": 0.00013536250093619369,
+      "loss": 1.072258472442627,
+      "mean_token_accuracy": 0.7464600443840027,
+      "num_tokens": 17125248.0,
+      "step": 4240
+    },
+    {
+      "entropy": 1.0479264855384827,
+      "epoch": 1.2503677552221242,
+      "grad_norm": 0.493936151266098,
+      "learning_rate": 0.0001350589598924107,
+      "loss": 1.0373758316040038,
+      "mean_token_accuracy": 0.755853122472763,
+      "num_tokens": 17165782.0,
+      "step": 4250
+    },
+    {
+      "entropy": 1.0287334680557252,
+      "epoch": 1.2533097969991174,
+      "grad_norm": 0.5333849191665649,
+      "learning_rate": 0.000134755050109017,
+      "loss": 1.04658203125,
+      "mean_token_accuracy": 0.7597375035285949,
+      "num_tokens": 17206071.0,
+      "step": 4260
+    },
+    {
+      "entropy": 1.031979387998581,
+      "epoch": 1.2562518387761106,
+      "grad_norm": 0.5658190250396729,
+      "learning_rate": 0.00013445077478244443,
+      "loss": 1.043376350402832,
+      "mean_token_accuracy": 0.7583476483821869,
+      "num_tokens": 17246566.0,
+      "step": 4270
+    },
+    {
+      "entropy": 0.9648872256278992,
+      "epoch": 1.2591938805531038,
+      "grad_norm": 0.5289658308029175,
+      "learning_rate": 0.00013414613711296952,
+      "loss": 0.9618735313415527,
+      "mean_token_accuracy": 0.7800238966941834,
+      "num_tokens": 17286973.0,
+      "step": 4280
+    },
+    {
+      "entropy": 1.0292047560214996,
+      "epoch": 1.262135922330097,
+      "grad_norm": 0.5604351162910461,
+      "learning_rate": 0.0001338411403046797,
+      "loss": 1.0540773391723632,
+      "mean_token_accuracy": 0.7540550827980042,
+      "num_tokens": 17327524.0,
+      "step": 4290
+    },
+    {
+      "entropy": 0.9784096240997314,
+      "epoch": 1.2650779641070904,
+      "grad_norm": 0.4879390597343445,
+      "learning_rate": 0.0001335357875654399,
+      "loss": 0.9683723449707031,
+      "mean_token_accuracy": 0.7743236780166626,
+      "num_tokens": 17367944.0,
+      "step": 4300
+    },
+    {
+      "epoch": 1.2650779641070904,
+      "eval_entropy": 1.024107585976443,
+      "eval_loss": 1.2039283514022827,
+      "eval_mean_token_accuracy": 0.7379744303195183,
+      "eval_num_tokens": 17367944.0,
+      "eval_runtime": 117.345,
+      "eval_samples_per_second": 25.949,
+      "eval_steps_per_second": 3.247,
+      "step": 4300
+    },
+    {
+      "entropy": 0.9147277295589447,
+      "epoch": 1.2680200058840836,
+      "grad_norm": 0.6553487181663513,
+      "learning_rate": 0.00013323008210685847,
+      "loss": 0.9428836822509765,
+      "mean_token_accuracy": 0.7800225138664245,
+      "num_tokens": 17408180.0,
+      "step": 4310
+    },
+    {
+      "entropy": 0.9862501919269562,
+      "epoch": 1.2709620476610768,
+      "grad_norm": 0.5519722700119019,
+      "learning_rate": 0.00013292402714425362,
+      "loss": 0.986777400970459,
+      "mean_token_accuracy": 0.7709893763065339,
+      "num_tokens": 17448767.0,
+      "step": 4320
+    },
+    {
+      "entropy": 0.984834861755371,
+      "epoch": 1.27390408943807,
+      "grad_norm": 0.6772357821464539,
+      "learning_rate": 0.0001326176258966195,
+      "loss": 1.0062361717224122,
+      "mean_token_accuracy": 0.7673668503761292,
+      "num_tokens": 17488778.0,
+      "step": 4330
+    },
+    {
+      "entropy": 1.0079510390758515,
+      "epoch": 1.2768461312150632,
+      "grad_norm": 0.6486600637435913,
+      "learning_rate": 0.00013231088158659245,
+      "loss": 1.0013746261596679,
+      "mean_token_accuracy": 0.7689732432365417,
+      "num_tokens": 17529357.0,
+      "step": 4340
+    },
+    {
+      "entropy": 0.930513882637024,
+      "epoch": 1.2797881729920566,
+      "grad_norm": 0.5784080028533936,
+      "learning_rate": 0.000132003797440417,
+      "loss": 0.941129207611084,
+      "mean_token_accuracy": 0.7789569735527039,
+      "num_tokens": 17569912.0,
+      "step": 4350
+    },
+    {
+      "entropy": 0.9446238994598388,
+      "epoch": 1.2827302147690498,
+      "grad_norm": 0.5628758072853088,
+      "learning_rate": 0.00013169637668791192,
+      "loss": 0.9538597106933594,
+      "mean_token_accuracy": 0.77459996342659,
+      "num_tokens": 17609948.0,
+      "step": 4360
+    },
+    {
+      "entropy": 0.9073866546154022,
+      "epoch": 1.285672256546043,
+      "grad_norm": 0.530838131904602,
+      "learning_rate": 0.0001313886225624364,
+      "loss": 0.9122503280639649,
+      "mean_token_accuracy": 0.7910377562046051,
+      "num_tokens": 17650396.0,
+      "step": 4370
+    },
+    {
+      "entropy": 0.9176636934280396,
+      "epoch": 1.2886142983230362,
+      "grad_norm": 0.5510920286178589,
+      "learning_rate": 0.00013108053830085585,
+      "loss": 0.9334745407104492,
+      "mean_token_accuracy": 0.7811925649642945,
+      "num_tokens": 17690414.0,
+      "step": 4380
+    },
+    {
+      "entropy": 0.9487575829029083,
+      "epoch": 1.2915563401000294,
+      "grad_norm": 0.5764872431755066,
+      "learning_rate": 0.00013077212714350807,
+      "loss": 0.9546429634094238,
+      "mean_token_accuracy": 0.7754026472568512,
+      "num_tokens": 17730995.0,
+      "step": 4390
+    },
+    {
+      "entropy": 0.9580233573913575,
+      "epoch": 1.2944983818770226,
+      "grad_norm": 0.510247528553009,
+      "learning_rate": 0.00013046339233416896,
+      "loss": 0.9728780746459961,
+      "mean_token_accuracy": 0.7761204540729523,
+      "num_tokens": 17771461.0,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2944983818770226,
+      "eval_entropy": 1.0582224565541025,
+      "eval_loss": 1.1942919492721558,
+      "eval_mean_token_accuracy": 0.7387891846691842,
+      "eval_num_tokens": 17771461.0,
+      "eval_runtime": 116.9989,
+      "eval_samples_per_second": 26.026,
+      "eval_steps_per_second": 3.256,
+      "step": 4400
+    },
+    {
+      "entropy": 1.009384435415268,
+      "epoch": 1.2974404236540158,
+      "grad_norm": 0.6525319218635559,
+      "learning_rate": 0.00013015433712001853,
+      "loss": 1.0117576599121094,
+      "mean_token_accuracy": 0.7641228914260865,
+      "num_tokens": 17811743.0,
+      "step": 4410
+    },
+    {
+      "entropy": 1.0025621831417084,
+      "epoch": 1.3003824654310092,
+      "grad_norm": 0.573704719543457,
+      "learning_rate": 0.00012984496475160667,
+      "loss": 1.0045761108398437,
+      "mean_token_accuracy": 0.7632306814193726,
+      "num_tokens": 17852150.0,
+      "step": 4420
+    },
+    {
+      "entropy": 0.889021772146225,
+      "epoch": 1.3033245072080024,
+      "grad_norm": 0.5899218916893005,
+      "learning_rate": 0.00012953527848281907,
+      "loss": 0.9004623413085937,
+      "mean_token_accuracy": 0.7905885875225067,
+      "num_tokens": 17892636.0,
+      "step": 4430
+    },
+    {
+      "entropy": 0.9145092189311981,
+      "epoch": 1.3062665489849956,
+      "grad_norm": 0.5533415675163269,
+      "learning_rate": 0.00012922528157084288,
+      "loss": 0.9265445709228516,
+      "mean_token_accuracy": 0.7852813005447388,
+      "num_tokens": 17933064.0,
+      "step": 4440
+    },
+    {
+      "entropy": 0.9506058990955353,
+      "epoch": 1.3092085907619888,
+      "grad_norm": 0.6637270450592041,
+      "learning_rate": 0.00012891497727613254,
+      "loss": 0.9665675163269043,
+      "mean_token_accuracy": 0.7746671974658966,
+      "num_tokens": 17973649.0,
+      "step": 4450
+    },
+    {
+      "entropy": 0.9762279152870178,
+      "epoch": 1.312150632538982,
+      "grad_norm": 0.5837135314941406,
+      "learning_rate": 0.0001286043688623754,
+      "loss": 0.9735669136047364,
+      "mean_token_accuracy": 0.7707210063934327,
+      "num_tokens": 18013745.0,
+      "step": 4460
+    },
+    {
+      "entropy": 0.9729628086090087,
+      "epoch": 1.3150926743159752,
+      "grad_norm": 0.5597095489501953,
+      "learning_rate": 0.00012829345959645744,
+      "loss": 0.983339500427246,
+      "mean_token_accuracy": 0.7697983980178833,
+      "num_tokens": 18054069.0,
+      "step": 4470
+    },
+    {
+      "entropy": 0.9756720840930939,
+      "epoch": 1.3180347160929684,
+      "grad_norm": 0.5869277119636536,
+      "learning_rate": 0.00012798225274842902,
+      "loss": 0.9763286590576172,
+      "mean_token_accuracy": 0.772778332233429,
+      "num_tokens": 18094600.0,
+      "step": 4480
+    },
+    {
+      "entropy": 0.9129612624645234,
+      "epoch": 1.3209767578699618,
+      "grad_norm": 0.5384027361869812,
+      "learning_rate": 0.00012767075159147022,
+      "loss": 0.9275237083435058,
+      "mean_token_accuracy": 0.7822975754737854,
+      "num_tokens": 18135046.0,
+      "step": 4490
+    },
+    {
+      "entropy": 1.018626469373703,
+      "epoch": 1.323918799646955,
+      "grad_norm": 0.6402458548545837,
+      "learning_rate": 0.0001273589594018567,
+      "loss": 1.0211774826049804,
+      "mean_token_accuracy": 0.7581977605819702,
+      "num_tokens": 18175538.0,
+      "step": 4500
+    },
+    {
+      "epoch": 1.323918799646955,
+      "eval_entropy": 1.0286849455570612,
+      "eval_loss": 1.1943385601043701,
+      "eval_mean_token_accuracy": 0.7395750786375812,
+      "eval_num_tokens": 18175538.0,
+      "eval_runtime": 117.1237,
+      "eval_samples_per_second": 25.998,
+      "eval_steps_per_second": 3.253,
+      "step": 4500
+    },
+    {
+      "entropy": 0.9917169988155365,
+      "epoch": 1.3268608414239482,
+      "grad_norm": 0.5651717185974121,
+      "learning_rate": 0.00012704687945892505,
+      "loss": 0.9929667472839355,
+      "mean_token_accuracy": 0.7704780220985412,
+      "num_tokens": 18215849.0,
+      "step": 4510
+    },
+    {
+      "entropy": 1.033211100101471,
+      "epoch": 1.3298028832009414,
+      "grad_norm": 0.6076104640960693,
+      "learning_rate": 0.00012673451504503842,
+      "loss": 1.053134059906006,
+      "mean_token_accuracy": 0.7530263483524322,
+      "num_tokens": 18256385.0,
+      "step": 4520
+    },
+    {
+      "entropy": 0.9446504592895508,
+      "epoch": 1.3327449249779346,
+      "grad_norm": 0.6524848341941833,
+      "learning_rate": 0.00012642186944555186,
+      "loss": 0.9612871170043945,
+      "mean_token_accuracy": 0.7765836179256439,
+      "num_tokens": 18296852.0,
+      "step": 4530
+    },
+    {
+      "entropy": 0.95798819065094,
+      "epoch": 1.335686966754928,
+      "grad_norm": 0.5617368817329407,
+      "learning_rate": 0.00012610894594877788,
+      "loss": 0.945002555847168,
+      "mean_token_accuracy": 0.7801140964031219,
+      "num_tokens": 18337142.0,
+      "step": 4540
+    },
+    {
+      "entropy": 0.9328159093856812,
+      "epoch": 1.338629008531921,
+      "grad_norm": 0.5558400750160217,
+      "learning_rate": 0.00012579574784595188,
+      "loss": 0.9508322715759278,
+      "mean_token_accuracy": 0.7804535567760468,
+      "num_tokens": 18377435.0,
+      "step": 4550
+    },
+    {
+      "entropy": 0.9236328303813934,
+      "epoch": 1.3415710503089144,
+      "grad_norm": 0.711991548538208,
+      "learning_rate": 0.00012548227843119743,
+      "loss": 0.9464892387390137,
+      "mean_token_accuracy": 0.781309175491333,
+      "num_tokens": 18418008.0,
+      "step": 4560
+    },
+    {
+      "entropy": 0.9790382087230682,
+      "epoch": 1.3445130920859076,
+      "grad_norm": 0.6019552946090698,
+      "learning_rate": 0.00012516854100149164,
+      "loss": 0.9772819519042969,
+      "mean_token_accuracy": 0.7732869625091553,
+      "num_tokens": 18458454.0,
+      "step": 4570
+    },
+    {
+      "entropy": 0.9923386096954345,
+      "epoch": 1.3474551338629008,
+      "grad_norm": 0.5897583365440369,
+      "learning_rate": 0.00012485453885663063,
+      "loss": 0.9893753051757812,
+      "mean_token_accuracy": 0.7679760038852692,
+      "num_tokens": 18498893.0,
+      "step": 4580
+    },
+    {
+      "entropy": 0.975125765800476,
+      "epoch": 1.350397175639894,
+      "grad_norm": 0.5963767766952515,
+      "learning_rate": 0.00012454027529919462,
+      "loss": 0.9868227958679199,
+      "mean_token_accuracy": 0.7681001186370849,
+      "num_tokens": 18539290.0,
+      "step": 4590
+    },
+    {
+      "entropy": 0.9119367241859436,
+      "epoch": 1.3533392174168872,
+      "grad_norm": 0.6379720568656921,
+      "learning_rate": 0.00012422575363451335,
+      "loss": 0.9250626564025879,
+      "mean_token_accuracy": 0.7873954355716706,
+      "num_tokens": 18579755.0,
+      "step": 4600
+    },
+    {
+      "epoch": 1.3533392174168872,
+      "eval_entropy": 1.045042129717474,
+      "eval_loss": 1.1902120113372803,
+      "eval_mean_token_accuracy": 0.739525359915936,
+      "eval_num_tokens": 18579755.0,
+      "eval_runtime": 117.0159,
+      "eval_samples_per_second": 26.022,
+      "eval_steps_per_second": 3.256,
+      "step": 4600
+    },
+    {
+      "entropy": 0.9590096414089203,
+      "epoch": 1.3562812591938807,
+      "grad_norm": 0.6441698670387268,
+      "learning_rate": 0.00012391097717063117,
+      "loss": 0.9705069541931153,
+      "mean_token_accuracy": 0.772186666727066,
+      "num_tokens": 18620196.0,
+      "step": 4610
+    },
+    {
+      "entropy": 0.9852827608585357,
+      "epoch": 1.3592233009708738,
+      "grad_norm": 0.6008490920066833,
+      "learning_rate": 0.00012359594921827245,
+      "loss": 0.9953920364379882,
+      "mean_token_accuracy": 0.7687745451927185,
+      "num_tokens": 18660133.0,
+      "step": 4620
+    },
+    {
+      "entropy": 1.0110926747322082,
+      "epoch": 1.362165342747867,
+      "grad_norm": 0.5604993104934692,
+      "learning_rate": 0.00012328067309080653,
+      "loss": 1.0147642135620116,
+      "mean_token_accuracy": 0.764192932844162,
+      "num_tokens": 18700534.0,
+      "step": 4630
+    },
+    {
+      "entropy": 0.9773748695850373,
+      "epoch": 1.3651073845248602,
+      "grad_norm": 0.5655143857002258,
+      "learning_rate": 0.0001229651521042131,
+      "loss": 1.0055460929870605,
+      "mean_token_accuracy": 0.7645916283130646,
+      "num_tokens": 18741011.0,
+      "step": 4640
+    },
+    {
+      "entropy": 0.912693589925766,
+      "epoch": 1.3680494263018534,
+      "grad_norm": 0.5726847648620605,
+      "learning_rate": 0.00012264938957704707,
+      "loss": 0.9029450416564941,
+      "mean_token_accuracy": 0.7917129874229432,
+      "num_tokens": 18781532.0,
+      "step": 4650
+    },
+    {
+      "entropy": 0.9796160280704498,
+      "epoch": 1.3709914680788466,
+      "grad_norm": 0.5168410539627075,
+      "learning_rate": 0.00012233338883040385,
+      "loss": 0.994998550415039,
+      "mean_token_accuracy": 0.7707984328269959,
+      "num_tokens": 18821962.0,
+      "step": 4660
+    },
+    {
+      "entropy": 0.9919624567031861,
+      "epoch": 1.3739335098558398,
+      "grad_norm": 0.5887457728385925,
+      "learning_rate": 0.00012201715318788445,
+      "loss": 0.9917054176330566,
+      "mean_token_accuracy": 0.7673246085643768,
+      "num_tokens": 18862330.0,
+      "step": 4670
+    },
+    {
+      "entropy": 1.0523385763168336,
+      "epoch": 1.3768755516328333,
+      "grad_norm": 0.5428098440170288,
+      "learning_rate": 0.00012170068597556035,
+      "loss": 1.077983283996582,
+      "mean_token_accuracy": 0.753084135055542,
+      "num_tokens": 18902801.0,
+      "step": 4680
+    },
+    {
+      "entropy": 0.9840884447097779,
+      "epoch": 1.3798175934098265,
+      "grad_norm": 0.6384422183036804,
+      "learning_rate": 0.00012138399052193867,
+      "loss": 0.9722138404846191,
+      "mean_token_accuracy": 0.7726257860660553,
+      "num_tokens": 18943381.0,
+      "step": 4690
+    },
+    {
+      "entropy": 0.9885792315006257,
+      "epoch": 1.3827596351868197,
+      "grad_norm": 0.5873745679855347,
+      "learning_rate": 0.00012106707015792702,
+      "loss": 1.0171488761901855,
+      "mean_token_accuracy": 0.7622905492782592,
+      "num_tokens": 18983723.0,
+      "step": 4700
+    },
+    {
+      "epoch": 1.3827596351868197,
+      "eval_entropy": 1.0410449004235856,
+      "eval_loss": 1.1852467060089111,
+      "eval_mean_token_accuracy": 0.7411133794021105,
+      "eval_num_tokens": 18983723.0,
+      "eval_runtime": 117.0373,
+      "eval_samples_per_second": 26.017,
+      "eval_steps_per_second": 3.255,
+      "step": 4700
+    },
+    {
+      "entropy": 0.9628095984458923,
+      "epoch": 1.3857016769638129,
+      "grad_norm": 0.6936764121055603,
+      "learning_rate": 0.00012074992821679866,
+      "loss": 0.9454580307006836,
+      "mean_token_accuracy": 0.7766146242618561,
+      "num_tokens": 19024209.0,
+      "step": 4710
+    },
+    {
+      "entropy": 0.9361280083656311,
+      "epoch": 1.388643718740806,
+      "grad_norm": 0.6228808164596558,
+      "learning_rate": 0.00012043256803415723,
+      "loss": 0.9670154571533203,
+      "mean_token_accuracy": 0.7755923092365264,
+      "num_tokens": 19064478.0,
+      "step": 4720
+    },
+    {
+      "entropy": 0.9650130271911621,
+      "epoch": 1.3915857605177995,
+      "grad_norm": 0.6201843619346619,
+      "learning_rate": 0.00012011499294790188,
+      "loss": 0.9677058219909668,
+      "mean_token_accuracy": 0.7731058478355408,
+      "num_tokens": 19104877.0,
+      "step": 4730
+    },
+    {
+      "entropy": 0.9925507187843323,
+      "epoch": 1.3945278022947925,
+      "grad_norm": 0.6081472635269165,
+      "learning_rate": 0.00011979720629819195,
+      "loss": 0.9994287490844727,
+      "mean_token_accuracy": 0.7684627115726471,
+      "num_tokens": 19145110.0,
+      "step": 4740
+    },
+    {
+      "entropy": 0.9522078454494476,
+      "epoch": 1.3974698440717859,
+      "grad_norm": 0.5446240305900574,
+      "learning_rate": 0.00011947921142741197,
+      "loss": 0.9563077926635742,
+      "mean_token_accuracy": 0.7776003420352936,
+      "num_tokens": 19185735.0,
+      "step": 4750
+    },
+    {
+      "entropy": 0.9502395629882813,
+      "epoch": 1.400411885848779,
+      "grad_norm": 0.6443742513656616,
+      "learning_rate": 0.00011916101168013649,
+      "loss": 0.9714900016784668,
+      "mean_token_accuracy": 0.7751095175743103,
+      "num_tokens": 19226033.0,
+      "step": 4760
+    },
+    {
+      "entropy": 0.9543413400650025,
+      "epoch": 1.4033539276257723,
+      "grad_norm": 0.60639488697052,
+      "learning_rate": 0.0001188426104030949,
+      "loss": 0.9422737121582031,
+      "mean_token_accuracy": 0.77926025390625,
+      "num_tokens": 19266072.0,
+      "step": 4770
+    },
+    {
+      "entropy": 0.8988826811313629,
+      "epoch": 1.4062959694027655,
+      "grad_norm": 0.6495632529258728,
+      "learning_rate": 0.00011852401094513621,
+      "loss": 0.9340031623840332,
+      "mean_token_accuracy": 0.7852738976478577,
+      "num_tokens": 19306236.0,
+      "step": 4780
+    },
+    {
+      "entropy": 1.0078811585903167,
+      "epoch": 1.4092380111797587,
+      "grad_norm": 0.6039602160453796,
+      "learning_rate": 0.00011820521665719377,
+      "loss": 1.0128409385681152,
+      "mean_token_accuracy": 0.7650022029876709,
+      "num_tokens": 19346514.0,
+      "step": 4790
+    },
+    {
+      "entropy": 1.002649539709091,
+      "epoch": 1.412180052956752,
+      "grad_norm": 0.5959292054176331,
+      "learning_rate": 0.00011788623089225024,
+      "loss": 0.9958960533142089,
+      "mean_token_accuracy": 0.7670526385307312,
+      "num_tokens": 19387036.0,
+      "step": 4800
+    },
+    {
+      "epoch": 1.412180052956752,
+      "eval_entropy": 1.0362290414612436,
+      "eval_loss": 1.1831785440444946,
+      "eval_mean_token_accuracy": 0.741490173058247,
+      "eval_num_tokens": 19387036.0,
+      "eval_runtime": 116.8979,
+      "eval_samples_per_second": 26.048,
+      "eval_steps_per_second": 3.259,
+      "step": 4800
+    },
+    {
+      "entropy": 0.9328351199626923,
+      "epoch": 1.4151220947337453,
+      "grad_norm": 0.5852963924407959,
+      "learning_rate": 0.00011756705700530206,
+      "loss": 0.9353754043579101,
+      "mean_token_accuracy": 0.7827379524707794,
+      "num_tokens": 19427397.0,
+      "step": 4810
+    },
+    {
+      "entropy": 0.9303468823432922,
+      "epoch": 1.4180641365107385,
+      "grad_norm": 0.5715077519416809,
+      "learning_rate": 0.0001172476983533243,
+      "loss": 0.9525899887084961,
+      "mean_token_accuracy": 0.7789464175701142,
+      "num_tokens": 19467357.0,
+      "step": 4820
+    },
+    {
+      "entropy": 0.9718841493129731,
+      "epoch": 1.4210061782877317,
+      "grad_norm": 0.5573017597198486,
+      "learning_rate": 0.00011692815829523536,
+      "loss": 0.974174690246582,
+      "mean_token_accuracy": 0.7711581230163574,
+      "num_tokens": 19507831.0,
+      "step": 4830
+    },
+    {
+      "entropy": 1.0064192593097687,
+      "epoch": 1.4239482200647249,
+      "grad_norm": 0.5526002645492554,
+      "learning_rate": 0.00011660844019186159,
+      "loss": 1.024794101715088,
+      "mean_token_accuracy": 0.7663941740989685,
+      "num_tokens": 19548408.0,
+      "step": 4840
+    },
+    {
+      "entropy": 1.0319006383419036,
+      "epoch": 1.426890261841718,
+      "grad_norm": 0.5624271035194397,
+      "learning_rate": 0.000116288547405902,
+      "loss": 1.0340099334716797,
+      "mean_token_accuracy": 0.760799127817154,
+      "num_tokens": 19588730.0,
+      "step": 4850
+    },
+    {
+      "entropy": 0.9651375532150268,
+      "epoch": 1.4298323036187113,
+      "grad_norm": 0.5835928320884705,
+      "learning_rate": 0.00011596848330189282,
+      "loss": 0.9745287895202637,
+      "mean_token_accuracy": 0.770064502954483,
+      "num_tokens": 19629242.0,
+      "step": 4860
+    },
+    {
+      "entropy": 0.9604414582252503,
+      "epoch": 1.4327743453957047,
+      "grad_norm": 0.6139530539512634,
+      "learning_rate": 0.00011564825124617218,
+      "loss": 0.967037582397461,
+      "mean_token_accuracy": 0.7793013870716095,
+      "num_tokens": 19669879.0,
+      "step": 4870
+    },
+    {
+      "entropy": 0.9346582233905792,
+      "epoch": 1.435716387172698,
+      "grad_norm": 0.6224908232688904,
+      "learning_rate": 0.00011532785460684466,
+      "loss": 0.9508832931518555,
+      "mean_token_accuracy": 0.7788917005062104,
+      "num_tokens": 19710258.0,
+      "step": 4880
+    },
+    {
+      "entropy": 0.9697970807552337,
+      "epoch": 1.438658428949691,
+      "grad_norm": 0.6372175216674805,
+      "learning_rate": 0.00011500729675374589,
+      "loss": 0.9690608978271484,
+      "mean_token_accuracy": 0.773440134525299,
+      "num_tokens": 19750412.0,
+      "step": 4890
+    },
+    {
+      "entropy": 0.9443894863128662,
+      "epoch": 1.4416004707266843,
+      "grad_norm": 0.6615795493125916,
+      "learning_rate": 0.00011468658105840706,
+      "loss": 0.9526325225830078,
+      "mean_token_accuracy": 0.7742028653621673,
+      "num_tokens": 19790931.0,
+      "step": 4900
+    },
+    {
+      "epoch": 1.4416004707266843,
+      "eval_entropy": 1.0257837637828717,
+      "eval_loss": 1.1847437620162964,
+      "eval_mean_token_accuracy": 0.7422285342779685,
+      "eval_num_tokens": 19790931.0,
+      "eval_runtime": 117.0067,
+      "eval_samples_per_second": 26.024,
+      "eval_steps_per_second": 3.256,
+      "step": 4900
+    },
+    {
+      "entropy": 0.9378713190555572,
+      "epoch": 1.4445425125036775,
+      "grad_norm": 0.5865938663482666,
+      "learning_rate": 0.0001143657108940196,
+      "loss": 0.9491632461547852,
+      "mean_token_accuracy": 0.7826192319393158,
+      "num_tokens": 19830998.0,
+      "step": 4910
+    },
+    {
+      "entropy": 0.8981156468391418,
+      "epoch": 1.447484554280671,
+      "grad_norm": 0.5529859066009521,
+      "learning_rate": 0.00011404468963539945,
+      "loss": 0.9059307098388671,
+      "mean_token_accuracy": 0.7881741523742676,
+      "num_tokens": 19871496.0,
+      "step": 4920
+    },
+    {
+      "entropy": 0.9764446496963501,
+      "epoch": 1.4504265960576639,
+      "grad_norm": 0.5437819361686707,
+      "learning_rate": 0.00011372352065895185,
+      "loss": 0.9850223541259766,
+      "mean_token_accuracy": 0.7667160212993622,
+      "num_tokens": 19912139.0,
+      "step": 4930
+    },
+    {
+      "entropy": 1.009915566444397,
+      "epoch": 1.4533686378346573,
+      "grad_norm": 0.48672956228256226,
+      "learning_rate": 0.00011340220734263562,
+      "loss": 1.009783935546875,
+      "mean_token_accuracy": 0.765262508392334,
+      "num_tokens": 19952597.0,
+      "step": 4940
+    },
+    {
+      "entropy": 0.9249416530132294,
+      "epoch": 1.4563106796116505,
+      "grad_norm": 0.5625399351119995,
+      "learning_rate": 0.00011308075306592771,
+      "loss": 0.9417881011962891,
+      "mean_token_accuracy": 0.7811116933822632,
+      "num_tokens": 19992805.0,
+      "step": 4950
+    },
+    {
+      "entropy": 0.9421619713306427,
+      "epoch": 1.4592527213886437,
+      "grad_norm": 0.5547005534172058,
+      "learning_rate": 0.00011275916120978769,
+      "loss": 0.9452463150024414,
+      "mean_token_accuracy": 0.7800073266029358,
+      "num_tokens": 20033035.0,
+      "step": 4960
+    },
+    {
+      "entropy": 0.9370833516120911,
+      "epoch": 1.462194763165637,
+      "grad_norm": 0.6140998601913452,
+      "learning_rate": 0.00011243743515662209,
+      "loss": 0.9688581466674805,
+      "mean_token_accuracy": 0.7777487993240356,
+      "num_tokens": 20073484.0,
+      "step": 4970
+    },
+    {
+      "entropy": 1.0171671450138091,
+      "epoch": 1.46513680494263,
+      "grad_norm": 0.6281487345695496,
+      "learning_rate": 0.00011211557829024892,
+      "loss": 0.9987648963928223,
+      "mean_token_accuracy": 0.7690559566020966,
+      "num_tokens": 20113884.0,
+      "step": 4980
+    },
+    {
+      "entropy": 0.9113238871097564,
+      "epoch": 1.4680788467196235,
+      "grad_norm": 0.5871062278747559,
+      "learning_rate": 0.00011179359399586202,
+      "loss": 0.9162681579589844,
+      "mean_token_accuracy": 0.7859906852245331,
+      "num_tokens": 20154035.0,
+      "step": 4990
+    },
+    {
+      "entropy": 0.9293498694896698,
+      "epoch": 1.4710208884966167,
+      "grad_norm": 0.5837708115577698,
+      "learning_rate": 0.00011147148565999553,
+      "loss": 0.9455188751220703,
+      "mean_token_accuracy": 0.7774161994457245,
+      "num_tokens": 20194398.0,
+      "step": 5000
+    },
+    {
+      "epoch": 1.4710208884966167,
+      "eval_entropy": 1.0215178381583196,
+      "eval_loss": 1.182082176208496,
+      "eval_mean_token_accuracy": 0.7424087036313034,
+      "eval_num_tokens": 20194398.0,
+      "eval_runtime": 116.9583,
+      "eval_samples_per_second": 26.035,
+      "eval_steps_per_second": 3.258,
+      "step": 5000
+    },
+    {
+      "entropy": 0.8972196578979492,
+      "epoch": 1.47396293027361,
+      "grad_norm": 0.6664556264877319,
+      "learning_rate": 0.00011114925667048814,
+      "loss": 0.897000789642334,
+      "mean_token_accuracy": 0.7910800576210022,
+      "num_tokens": 20234842.0,
+      "step": 5010
+    },
+    {
+      "entropy": 0.9366280138492584,
+      "epoch": 1.4769049720506031,
+      "grad_norm": 0.5105628967285156,
+      "learning_rate": 0.00011082691041644762,
+      "loss": 0.9634222984313965,
+      "mean_token_accuracy": 0.7767743766307831,
+      "num_tokens": 20275050.0,
+      "step": 5020
+    },
+    {
+      "entropy": 1.0160198926925659,
+      "epoch": 1.4798470138275963,
+      "grad_norm": 0.6369785666465759,
+      "learning_rate": 0.00011050445028821504,
+      "loss": 1.0192986488342286,
+      "mean_token_accuracy": 0.763582181930542,
+      "num_tokens": 20315449.0,
+      "step": 5030
+    },
+    {
+      "entropy": 1.007281619310379,
+      "epoch": 1.4827890556045895,
+      "grad_norm": 0.6209942698478699,
+      "learning_rate": 0.00011018187967732918,
+      "loss": 0.9973898887634277,
+      "mean_token_accuracy": 0.7695520281791687,
+      "num_tokens": 20355896.0,
+      "step": 5040
+    },
+    {
+      "entropy": 0.9635446310043335,
+      "epoch": 1.4857310973815827,
+      "grad_norm": 0.6402096748352051,
+      "learning_rate": 0.00010985920197649086,
+      "loss": 1.0030365943908692,
+      "mean_token_accuracy": 0.7655979931354523,
+      "num_tokens": 20396451.0,
+      "step": 5050
+    },
+    {
+      "entropy": 0.9760343492031097,
+      "epoch": 1.4886731391585761,
+      "grad_norm": 0.658566415309906,
+      "learning_rate": 0.00010953642057952722,
+      "loss": 0.9697424888610839,
+      "mean_token_accuracy": 0.7760293245315552,
+      "num_tokens": 20436615.0,
+      "step": 5060
+    },
+    {
+      "entropy": 0.9795664429664612,
+      "epoch": 1.4916151809355693,
+      "grad_norm": 0.5999899506568909,
+      "learning_rate": 0.00010921353888135605,
+      "loss": 0.9814806938171386,
+      "mean_token_accuracy": 0.7703676760196686,
+      "num_tokens": 20477070.0,
+      "step": 5070
+    },
+    {
+      "entropy": 0.9684806585311889,
+      "epoch": 1.4945572227125625,
+      "grad_norm": 0.6002531051635742,
+      "learning_rate": 0.00010889056027795009,
+      "loss": 0.9861810684204102,
+      "mean_token_accuracy": 0.7723352909088135,
+      "num_tokens": 20517284.0,
+      "step": 5080
+    },
+    {
+      "entropy": 0.9738320171833038,
+      "epoch": 1.4974992644895557,
+      "grad_norm": 0.6089113354682922,
+      "learning_rate": 0.00010856748816630127,
+      "loss": 0.984062385559082,
+      "mean_token_accuracy": 0.772194218635559,
+      "num_tokens": 20557266.0,
+      "step": 5090
+    },
+    {
+      "entropy": 0.9736992299556733,
+      "epoch": 1.500441306266549,
+      "grad_norm": 0.5777798295021057,
+      "learning_rate": 0.00010824432594438505,
+      "loss": 0.9862478256225586,
+      "mean_token_accuracy": 0.7714533090591431,
+      "num_tokens": 20597675.0,
+      "step": 5100
+    },
+    {
+      "epoch": 1.500441306266549,
+      "eval_entropy": 1.0292494011519775,
+      "eval_loss": 1.1772292852401733,
+      "eval_mean_token_accuracy": 0.7431638131304363,
+      "eval_num_tokens": 20597675.0,
+      "eval_runtime": 116.9139,
+      "eval_samples_per_second": 26.045,
+      "eval_steps_per_second": 3.259,
+      "step": 5100
+    },
+    {
+      "entropy": 0.902402263879776,
+      "epoch": 1.5033833480435423,
+      "grad_norm": 0.6047548055648804,
+      "learning_rate": 0.0001079210770111246,
+      "loss": 0.9013402938842774,
+      "mean_token_accuracy": 0.7890763878822327,
+      "num_tokens": 20638028.0,
+      "step": 5110
+    },
+    {
+      "entropy": 0.9433676958084106,
+      "epoch": 1.5063253898205353,
+      "grad_norm": 0.7120152711868286,
+      "learning_rate": 0.00010759774476635513,
+      "loss": 0.9629843711853028,
+      "mean_token_accuracy": 0.7767750382423401,
+      "num_tokens": 20677777.0,
+      "step": 5120
+    },
+    {
+      "entropy": 1.028594321012497,
+      "epoch": 1.5092674315975287,
+      "grad_norm": 0.5008658766746521,
+      "learning_rate": 0.00010727433261078808,
+      "loss": 1.0339035987854004,
+      "mean_token_accuracy": 0.757422798871994,
+      "num_tokens": 20718022.0,
+      "step": 5130
+    },
+    {
+      "entropy": 1.0165768265724182,
+      "epoch": 1.512209473374522,
+      "grad_norm": 0.5533917546272278,
+      "learning_rate": 0.00010695084394597537,
+      "loss": 1.0265631675720215,
+      "mean_token_accuracy": 0.7609834551811219,
+      "num_tokens": 20758730.0,
+      "step": 5140
+    },
+    {
+      "entropy": 1.011590701341629,
+      "epoch": 1.5151515151515151,
+      "grad_norm": 0.6278111934661865,
+      "learning_rate": 0.00010662728217427362,
+      "loss": 1.0116978645324708,
+      "mean_token_accuracy": 0.7609262108802796,
+      "num_tokens": 20799202.0,
+      "step": 5150
+    },
+    {
+      "entropy": 0.9122114419937134,
+      "epoch": 1.5180935569285083,
+      "grad_norm": 0.5556227564811707,
+      "learning_rate": 0.00010630365069880837,
+      "loss": 0.9219463348388672,
+      "mean_token_accuracy": 0.7841397285461426,
+      "num_tokens": 20839777.0,
+      "step": 5160
+    },
+    {
+      "entropy": 0.9289700865745545,
+      "epoch": 1.5210355987055015,
+      "grad_norm": 0.656017005443573,
+      "learning_rate": 0.00010597995292343827,
+      "loss": 0.9393370628356934,
+      "mean_token_accuracy": 0.7785055756568908,
+      "num_tokens": 20880324.0,
+      "step": 5170
+    },
+    {
+      "entropy": 0.9671324849128723,
+      "epoch": 1.523977640482495,
+      "grad_norm": 0.5946719646453857,
+      "learning_rate": 0.00010565619225271934,
+      "loss": 0.9726341247558594,
+      "mean_token_accuracy": 0.772115957736969,
+      "num_tokens": 20920903.0,
+      "step": 5180
+    },
+    {
+      "entropy": 0.9615132927894592,
+      "epoch": 1.526919682259488,
+      "grad_norm": 0.5949153304100037,
+      "learning_rate": 0.00010533237209186904,
+      "loss": 0.9622700691223145,
+      "mean_token_accuracy": 0.7711909949779511,
+      "num_tokens": 20960934.0,
+      "step": 5190
+    },
+    {
+      "entropy": 0.9015084564685821,
+      "epoch": 1.5298617240364814,
+      "grad_norm": 0.594940721988678,
+      "learning_rate": 0.00010500849584673059,
+      "loss": 0.9115975379943848,
+      "mean_token_accuracy": 0.7880061626434326,
+      "num_tokens": 21001387.0,
+      "step": 5200
+    },
+    {
+      "epoch": 1.5298617240364814,
+      "eval_entropy": 1.0002246155669996,
+      "eval_loss": 1.1774756908416748,
+      "eval_mean_token_accuracy": 0.7437332308511408,
+      "eval_num_tokens": 21001387.0,
+      "eval_runtime": 116.917,
+      "eval_samples_per_second": 26.044,
+      "eval_steps_per_second": 3.259,
+      "step": 5200
+    },
+    {
+      "entropy": 0.9508689403533935,
+      "epoch": 1.5328037658134746,
+      "grad_norm": 0.5927796959877014,
+      "learning_rate": 0.00010468456692373703,
+      "loss": 0.9860681533813477,
+      "mean_token_accuracy": 0.7731941938400269,
+      "num_tokens": 21041244.0,
+      "step": 5210
+    },
+    {
+      "entropy": 0.9849535644054412,
+      "epoch": 1.5357458075904677,
+      "grad_norm": 0.6247168183326721,
+      "learning_rate": 0.0001043605887298755,
+      "loss": 0.9714067459106446,
+      "mean_token_accuracy": 0.7714132785797119,
+      "num_tokens": 21081657.0,
+      "step": 5220
+    },
+    {
+      "entropy": 0.9713167667388916,
+      "epoch": 1.5386878493674612,
+      "grad_norm": 0.677769124507904,
+      "learning_rate": 0.00010403656467265138,
+      "loss": 0.9748648643493653,
+      "mean_token_accuracy": 0.7746425211429596,
+      "num_tokens": 21122109.0,
+      "step": 5230
+    },
+    {
+      "entropy": 0.9380859136581421,
+      "epoch": 1.5416298911444541,
+      "grad_norm": 0.6017981767654419,
+      "learning_rate": 0.00010371249816005235,
+      "loss": 0.950676441192627,
+      "mean_token_accuracy": 0.7762335240840912,
+      "num_tokens": 21162652.0,
+      "step": 5240
+    },
+    {
+      "entropy": 0.9800868451595306,
+      "epoch": 1.5445719329214476,
+      "grad_norm": 0.6067262291908264,
+      "learning_rate": 0.00010338839260051265,
+      "loss": 0.9775652885437012,
+      "mean_token_accuracy": 0.7736145675182342,
+      "num_tokens": 21203018.0,
+      "step": 5250
+    },
+    {
+      "entropy": 0.9683707654476166,
+      "epoch": 1.5475139746984408,
+      "grad_norm": 0.5600547790527344,
+      "learning_rate": 0.00010306425140287724,
+      "loss": 0.9947422027587891,
+      "mean_token_accuracy": 0.768592232465744,
+      "num_tokens": 21243193.0,
+      "step": 5260
+    },
+    {
+      "entropy": 0.9973979473114014,
+      "epoch": 1.550456016475434,
+      "grad_norm": 0.6263974905014038,
+      "learning_rate": 0.00010274007797636589,
+      "loss": 1.0072562217712402,
+      "mean_token_accuracy": 0.7686869263648987,
+      "num_tokens": 21283229.0,
+      "step": 5270
+    },
+    {
+      "entropy": 0.9464682042598724,
+      "epoch": 1.5533980582524272,
+      "grad_norm": 0.6058053970336914,
+      "learning_rate": 0.00010241587573053732,
+      "loss": 0.9497817993164063,
+      "mean_token_accuracy": 0.7785175144672394,
+      "num_tokens": 21323783.0,
+      "step": 5280
+    },
+    {
+      "entropy": 1.0010394990444182,
+      "epoch": 1.5563401000294204,
+      "grad_norm": 0.6609871983528137,
+      "learning_rate": 0.0001020916480752534,
+      "loss": 1.0204454421997071,
+      "mean_token_accuracy": 0.761556738615036,
+      "num_tokens": 21364201.0,
+      "step": 5290
+    },
+    {
+      "entropy": 0.9770765423774719,
+      "epoch": 1.5592821418064138,
+      "grad_norm": 0.5533031225204468,
+      "learning_rate": 0.00010176739842064323,
+      "loss": 0.9723864555358886,
+      "mean_token_accuracy": 0.7763941168785096,
+      "num_tokens": 21404698.0,
+      "step": 5300
+    },
+    {
+      "epoch": 1.5592821418064138,
+      "eval_entropy": 1.0454554280896826,
+      "eval_loss": 1.168828010559082,
+      "eval_mean_token_accuracy": 0.744419585844976,
+      "eval_num_tokens": 21404698.0,
+      "eval_runtime": 116.9992,
+      "eval_samples_per_second": 26.026,
+      "eval_steps_per_second": 3.256,
+      "step": 5300
+    },
+    {
+      "entropy": 0.9656676173210144,
+      "epoch": 1.5622241835834068,
+      "grad_norm": 0.6528343558311462,
+      "learning_rate": 0.00010144313017706726,
+      "loss": 0.9678720474243164,
+      "mean_token_accuracy": 0.779698771238327,
+      "num_tokens": 21445187.0,
+      "step": 5310
+    },
+    {
+      "entropy": 0.9781722486019134,
+      "epoch": 1.5651662253604002,
+      "grad_norm": 0.5709927678108215,
+      "learning_rate": 0.00010111884675508151,
+      "loss": 1.010976505279541,
+      "mean_token_accuracy": 0.7641195952892303,
+      "num_tokens": 21485680.0,
+      "step": 5320
+    },
+    {
+      "entropy": 0.9678059935569763,
+      "epoch": 1.5681082671373934,
+      "grad_norm": 0.5990722179412842,
+      "learning_rate": 0.00010079455156540163,
+      "loss": 0.9739880561828613,
+      "mean_token_accuracy": 0.777262145280838,
+      "num_tokens": 21526255.0,
+      "step": 5330
+    },
+    {
+      "entropy": 0.934472793340683,
+      "epoch": 1.5710503089143866,
+      "grad_norm": 0.5266041159629822,
+      "learning_rate": 0.00010047024801886702,
+      "loss": 0.9320767402648926,
+      "mean_token_accuracy": 0.7839205145835877,
+      "num_tokens": 21566600.0,
+      "step": 5340
+    },
+    {
+      "entropy": 0.9387928783893585,
+      "epoch": 1.5739923506913798,
+      "grad_norm": 0.5499687790870667,
+      "learning_rate": 0.00010014593952640494,
+      "loss": 0.9497169494628906,
+      "mean_token_accuracy": 0.7800655484199523,
+      "num_tokens": 21607156.0,
+      "step": 5350
+    },
+    {
+      "entropy": 1.0074778258800507,
+      "epoch": 1.576934392468373,
+      "grad_norm": 0.6578675508499146,
+      "learning_rate": 9.982162949899479e-05,
+      "loss": 1.020614242553711,
+      "mean_token_accuracy": 0.765311861038208,
+      "num_tokens": 21647444.0,
+      "step": 5360
+    },
+    {
+      "entropy": 0.9886070728302002,
+      "epoch": 1.5798764342453664,
+      "grad_norm": 0.6587820053100586,
+      "learning_rate": 9.949732134763199e-05,
+      "loss": 0.993044662475586,
+      "mean_token_accuracy": 0.7681563913822174,
+      "num_tokens": 21688001.0,
+      "step": 5370
+    },
+    {
+      "entropy": 0.899987381696701,
+      "epoch": 1.5828184760223594,
+      "grad_norm": 0.5484562516212463,
+      "learning_rate": 9.917301848329231e-05,
+      "loss": 0.9030593872070313,
+      "mean_token_accuracy": 0.7870559990406036,
+      "num_tokens": 21727882.0,
+      "step": 5380
+    },
+    {
+      "entropy": 0.9620799243450164,
+      "epoch": 1.5857605177993528,
+      "grad_norm": 0.619149386882782,
+      "learning_rate": 9.884872431689581e-05,
+      "loss": 0.9912420272827148,
+      "mean_token_accuracy": 0.7723720014095307,
+      "num_tokens": 21767753.0,
+      "step": 5390
+    },
+    {
+      "entropy": 1.000234466791153,
+      "epoch": 1.588702559576346,
+      "grad_norm": 0.5575194954872131,
+      "learning_rate": 9.852444225927122e-05,
+      "loss": 0.9978320121765136,
+      "mean_token_accuracy": 0.7687974095344543,
+      "num_tokens": 21808266.0,
+      "step": 5400
+    },
+    {
+      "epoch": 1.588702559576346,
+      "eval_entropy": 1.0398932406752128,
+      "eval_loss": 1.1691675186157227,
+      "eval_mean_token_accuracy": 0.7446287646694133,
+      "eval_num_tokens": 21808266.0,
+      "eval_runtime": 116.9348,
+      "eval_samples_per_second": 26.04,
+      "eval_steps_per_second": 3.258,
+      "step": 5400
+    },
+    {
+      "entropy": 0.9786124050617218,
+      "epoch": 1.5916446013533392,
+      "grad_norm": 0.5668993592262268,
+      "learning_rate": 9.820017572111973e-05,
+      "loss": 0.9736597061157226,
+      "mean_token_accuracy": 0.768933230638504,
+      "num_tokens": 21848699.0,
+      "step": 5410
+    },
+    {
+      "entropy": 0.9540457367897034,
+      "epoch": 1.5945866431303326,
+      "grad_norm": 0.5640490651130676,
+      "learning_rate": 9.787592811297946e-05,
+      "loss": 0.9902207374572753,
+      "mean_token_accuracy": 0.7725074052810669,
+      "num_tokens": 21888992.0,
+      "step": 5420
+    },
+    {
+      "entropy": 1.0212572634220123,
+      "epoch": 1.5975286849073256,
+      "grad_norm": 0.5303104519844055,
+      "learning_rate": 9.755170284518941e-05,
+      "loss": 1.0194875717163085,
+      "mean_token_accuracy": 0.7608138382434845,
+      "num_tokens": 21929626.0,
+      "step": 5430
+    },
+    {
+      "entropy": 0.9701859831809998,
+      "epoch": 1.600470726684319,
+      "grad_norm": 0.5962478518486023,
+      "learning_rate": 9.722750332785349e-05,
+      "loss": 0.9606605529785156,
+      "mean_token_accuracy": 0.7774016797542572,
+      "num_tokens": 21970234.0,
+      "step": 5440
+    },
+    {
+      "entropy": 0.9712904334068299,
+      "epoch": 1.6034127684613122,
+      "grad_norm": 0.6927991509437561,
+      "learning_rate": 9.690333297080493e-05,
+      "loss": 0.9966094970703125,
+      "mean_token_accuracy": 0.7730933606624604,
+      "num_tokens": 22010748.0,
+      "step": 5450
+    },
+    {
+      "entropy": 0.9969470083713532,
+      "epoch": 1.6063548102383054,
+      "grad_norm": 0.48069873452186584,
+      "learning_rate": 9.657919518357008e-05,
+      "loss": 1.0035072326660157,
+      "mean_token_accuracy": 0.7701967000961304,
+      "num_tokens": 22051287.0,
+      "step": 5460
+    },
+    {
+      "entropy": 0.9618437588214874,
+      "epoch": 1.6092968520152986,
+      "grad_norm": 0.5836319923400879,
+      "learning_rate": 9.625509337533296e-05,
+      "loss": 0.9715272903442382,
+      "mean_token_accuracy": 0.7741812229156494,
+      "num_tokens": 22091789.0,
+      "step": 5470
+    },
+    {
+      "entropy": 0.92412930727005,
+      "epoch": 1.6122388937922918,
+      "grad_norm": 0.5578837990760803,
+      "learning_rate": 9.593103095489895e-05,
+      "loss": 0.9371700286865234,
+      "mean_token_accuracy": 0.780053973197937,
+      "num_tokens": 22132202.0,
+      "step": 5480
+    },
+    {
+      "entropy": 0.9944902658462524,
+      "epoch": 1.6151809355692852,
+      "grad_norm": 0.6106790900230408,
+      "learning_rate": 9.560701133065932e-05,
+      "loss": 0.9925059318542481,
+      "mean_token_accuracy": 0.771945059299469,
+      "num_tokens": 22172268.0,
+      "step": 5490
+    },
+    {
+      "entropy": 0.9864640951156616,
+      "epoch": 1.6181229773462782,
+      "grad_norm": 0.6734248399734497,
+      "learning_rate": 9.528303791055511e-05,
+      "loss": 1.001873779296875,
+      "mean_token_accuracy": 0.7679209113121033,
+      "num_tokens": 22212768.0,
+      "step": 5500
+    },
+    {
+      "epoch": 1.6181229773462782,
+      "eval_entropy": 1.017620304985622,
+      "eval_loss": 1.1692047119140625,
+      "eval_mean_token_accuracy": 0.7452891367314056,
+      "eval_num_tokens": 22212768.0,
+      "eval_runtime": 116.9536,
+      "eval_samples_per_second": 26.036,
+      "eval_steps_per_second": 3.258,
+      "step": 5500
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 10197,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0319598971846e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}