diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,7117 @@
+{
+  "best_global_step": 1000,
+  "best_metric": 0.9796417355537415,
+  "best_model_checkpoint": "/data6/jongbinwon/evolving_MoE/checkpoints/sft_llama31_8b_lbox_shared-generalist_fallback_ffn_only_5ep_eval500/checkpoint-1000",
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 6925,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.3488953799009322,
+      "epoch": 0.007224128589488893,
+      "grad_norm": 12.38263988494873,
+      "learning_rate": 8.653846153846154e-07,
+      "loss": 1.2809778213500977,
+      "mean_token_accuracy": 0.7182683728635311,
+      "num_tokens": 29664.0,
+      "step": 10
+    },
+    {
+      "entropy": 2.2779021948575973,
+      "epoch": 0.014448257178977787,
+      "grad_norm": 6.46359395980835,
+      "learning_rate": 1.826923076923077e-06,
+      "loss": 1.4152549743652343,
+      "mean_token_accuracy": 0.7039315856993198,
+      "num_tokens": 65995.0,
+      "step": 20
+    },
+    {
+      "entropy": 2.2851940393447876,
+      "epoch": 0.021672385768466678,
+      "grad_norm": 7.5366010665893555,
+      "learning_rate": 2.7884615384615386e-06,
+      "loss": 1.3896544456481934,
+      "mean_token_accuracy": 0.7117640927433968,
+      "num_tokens": 99531.0,
+      "step": 30
+    },
+    {
+      "entropy": 2.295980241894722,
+      "epoch": 0.028896514357955573,
+      "grad_norm": 3.707631826400757,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 1.3731188774108887,
+      "mean_token_accuracy": 0.7028925865888596,
+      "num_tokens": 135081.0,
+      "step": 40
+    },
+    {
+      "entropy": 2.38641714155674,
+      "epoch": 0.036120642947444465,
+      "grad_norm": 4.440042495727539,
+      "learning_rate": 4.711538461538462e-06,
+      "loss": 1.3516225814819336,
+      "mean_token_accuracy": 0.704179621487856,
+      "num_tokens": 168474.0,
+      "step": 50
+    },
+    {
+      "entropy": 2.3031668663024902,
+      "epoch": 0.043344771536933356,
+      "grad_norm": 3.3444595336914062,
+      "learning_rate": 5.6730769230769235e-06,
+      "loss": 0.7931435108184814,
+      "mean_token_accuracy": 0.8210788562893867,
+      "num_tokens": 199656.0,
+      "step": 60
+    },
+    {
+      "entropy": 2.2916590481996537,
+      "epoch": 0.05056890012642225,
+      "grad_norm": 3.439544439315796,
+      "learning_rate": 6.6346153846153846e-06,
+      "loss": 0.8380290985107421,
+      "mean_token_accuracy": 0.8113071389496327,
+      "num_tokens": 237137.0,
+      "step": 70
+    },
+    {
+      "entropy": 2.3342412978410723,
+      "epoch": 0.057793028715911146,
+      "grad_norm": 3.820638656616211,
+      "learning_rate": 7.5961538461538465e-06,
+      "loss": 0.5829432487487793,
+      "mean_token_accuracy": 0.8551584139466286,
+      "num_tokens": 269916.0,
+      "step": 80
+    },
+    {
+      "entropy": 2.3470212489366533,
+      "epoch": 0.06501715730540003,
+      "grad_norm": 5.16990852355957,
+      "learning_rate": 8.557692307692308e-06,
+      "loss": 0.7105390548706054,
+      "mean_token_accuracy": 0.8280642315745353,
+      "num_tokens": 302577.0,
+      "step": 90
+    },
+    {
+      "entropy": 2.3362986624240873,
+      "epoch": 0.07224128589488893,
+      "grad_norm": 5.412623405456543,
+      "learning_rate": 9.51923076923077e-06,
+      "loss": 0.5351422786712646,
+      "mean_token_accuracy": 0.8582781746983528,
+      "num_tokens": 333925.0,
+      "step": 100
+    },
+    {
+      "entropy": 2.3203985303640366,
+      "epoch": 0.07946541448437783,
+      "grad_norm": 5.008820056915283,
+      "learning_rate": 1.0480769230769232e-05,
+      "loss": 0.42360854148864746,
+      "mean_token_accuracy": 0.8884407944977284,
+      "num_tokens": 366680.0,
+      "step": 110
+    },
+    {
+      "entropy": 2.3535674154758452,
+      "epoch": 0.08668954307386671,
+      "grad_norm": 3.0379772186279297,
+      "learning_rate": 1.1442307692307693e-05,
+      "loss": 0.41391968727111816,
+      "mean_token_accuracy": 0.9000807195901871,
+      "num_tokens": 397501.0,
+      "step": 120
+    },
+    {
+      "entropy": 2.324735555052757,
+      "epoch": 0.09391367166335561,
+      "grad_norm": 3.118257761001587,
+      "learning_rate": 1.2403846153846156e-05,
+      "loss": 0.3381591558456421,
+      "mean_token_accuracy": 0.9182203456759452,
+      "num_tokens": 429067.0,
+      "step": 130
+    },
+    {
+      "entropy": 2.3521973222494124,
+      "epoch": 0.1011378002528445,
+      "grad_norm": 4.433260917663574,
+      "learning_rate": 1.3365384615384615e-05,
+      "loss": 0.3163116931915283,
+      "mean_token_accuracy": 0.9162691205739975,
+      "num_tokens": 458216.0,
+      "step": 140
+    },
+    {
+      "entropy": 2.2347404181957247,
+      "epoch": 0.1083619288423334,
+      "grad_norm": 1.7790768146514893,
+      "learning_rate": 1.4326923076923078e-05,
+      "loss": 0.19349080324172974,
+      "mean_token_accuracy": 0.9514048635959625,
+      "num_tokens": 495714.0,
+      "step": 150
+    },
+    {
+      "entropy": 2.21967813372612,
+      "epoch": 0.11558605743182229,
+      "grad_norm": 5.751720905303955,
+      "learning_rate": 1.528846153846154e-05,
+      "loss": 0.3013446569442749,
+      "mean_token_accuracy": 0.927564126253128,
+      "num_tokens": 532706.0,
+      "step": 160
+    },
+    {
+      "entropy": 2.209682157635689,
+      "epoch": 0.12281018602131118,
+      "grad_norm": 4.959426403045654,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 0.15157370567321776,
+      "mean_token_accuracy": 0.9651991903781891,
+      "num_tokens": 569522.0,
+      "step": 170
+    },
+    {
+      "entropy": 2.2797712594270707,
+      "epoch": 0.13003431461080006,
+      "grad_norm": 5.981288433074951,
+      "learning_rate": 1.7211538461538465e-05,
+      "loss": 0.19515550136566162,
+      "mean_token_accuracy": 0.9430472493171692,
+      "num_tokens": 600249.0,
+      "step": 180
+    },
+    {
+      "entropy": 2.282535618543625,
+      "epoch": 0.13725844320028896,
+      "grad_norm": 2.61159348487854,
+      "learning_rate": 1.8173076923076924e-05,
+      "loss": 0.12393771409988404,
+      "mean_token_accuracy": 0.9631432116031646,
+      "num_tokens": 633357.0,
+      "step": 190
+    },
+    {
+      "entropy": 2.1924006402492524,
+      "epoch": 0.14448257178977786,
+      "grad_norm": 2.5875096321105957,
+      "learning_rate": 1.9134615384615387e-05,
+      "loss": 0.1760977864265442,
+      "mean_token_accuracy": 0.9507284834980965,
+      "num_tokens": 669069.0,
+      "step": 200
+    },
+    {
+      "entropy": 2.210735687613487,
+      "epoch": 0.15170670037926676,
+      "grad_norm": 1.9026196002960205,
+      "learning_rate": 1.9999998906247536e-05,
+      "loss": 0.13357192277908325,
+      "mean_token_accuracy": 0.9658363074064255,
+      "num_tokens": 702442.0,
+      "step": 210
+    },
+    {
+      "entropy": 2.1797422796487806,
+      "epoch": 0.15893082896875566,
+      "grad_norm": 7.958958148956299,
+      "learning_rate": 1.9999867656241115e-05,
+      "loss": 0.1266096353530884,
+      "mean_token_accuracy": 0.960417041182518,
+      "num_tokens": 734380.0,
+      "step": 220
+    },
+    {
+      "entropy": 2.2051376909017564,
+      "epoch": 0.16615495755824453,
+      "grad_norm": 6.823901176452637,
+      "learning_rate": 1.999951765903129e-05,
+      "loss": 0.1252610445022583,
+      "mean_token_accuracy": 0.9647611796855926,
+      "num_tokens": 771815.0,
+      "step": 230
+    },
+    {
+      "entropy": 2.2458199709653854,
+      "epoch": 0.17337908614773342,
+      "grad_norm": 4.231349945068359,
+      "learning_rate": 1.9998948922274254e-05,
+      "loss": 0.19357305765151978,
+      "mean_token_accuracy": 0.9423901841044426,
+      "num_tokens": 801392.0,
+      "step": 240
+    },
+    {
+      "entropy": 2.2267865359783174,
+      "epoch": 0.18060321473722232,
+      "grad_norm": 4.5076093673706055,
+      "learning_rate": 1.9998161458411122e-05,
+      "loss": 0.1421981930732727,
+      "mean_token_accuracy": 0.9549527898430824,
+      "num_tokens": 835126.0,
+      "step": 250
+    },
+    {
+      "entropy": 2.2681135296821595,
+      "epoch": 0.18782734332671122,
+      "grad_norm": 4.918508529663086,
+      "learning_rate": 1.9997155284667678e-05,
+      "loss": 0.11410770416259766,
+      "mean_token_accuracy": 0.9669175207614898,
+      "num_tokens": 865543.0,
+      "step": 260
+    },
+    {
+      "entropy": 2.1435286939144134,
+      "epoch": 0.19505147191620012,
+      "grad_norm": 3.8047282695770264,
+      "learning_rate": 1.9995930423053985e-05,
+      "loss": 0.16196959018707274,
+      "mean_token_accuracy": 0.9546934857964515,
+      "num_tokens": 905535.0,
+      "step": 270
+    },
+    {
+      "entropy": 2.221021184325218,
+      "epoch": 0.202275600505689,
+      "grad_norm": 4.958682060241699,
+      "learning_rate": 1.9994486900363898e-05,
+      "loss": 0.15060049295425415,
+      "mean_token_accuracy": 0.9543882220983505,
+      "num_tokens": 942677.0,
+      "step": 280
+    },
+    {
+      "entropy": 2.188411471247673,
+      "epoch": 0.2094997290951779,
+      "grad_norm": 2.7439119815826416,
+      "learning_rate": 1.9992824748174494e-05,
+      "loss": 0.09933483004570007,
+      "mean_token_accuracy": 0.969770023226738,
+      "num_tokens": 976377.0,
+      "step": 290
+    },
+    {
+      "entropy": 2.2444622576236726,
+      "epoch": 0.2167238576846668,
+      "grad_norm": 5.669027328491211,
+      "learning_rate": 1.9990944002845366e-05,
+      "loss": 0.08762626647949219,
+      "mean_token_accuracy": 0.9665105521678925,
+      "num_tokens": 1008228.0,
+      "step": 300
+    },
+    {
+      "entropy": 2.157886192202568,
+      "epoch": 0.22394798627415569,
+      "grad_norm": 3.5573644638061523,
+      "learning_rate": 1.9988844705517843e-05,
+      "loss": 0.07699022889137268,
+      "mean_token_accuracy": 0.9739585548639298,
+      "num_tokens": 1038551.0,
+      "step": 310
+    },
+    {
+      "entropy": 2.131389778852463,
+      "epoch": 0.23117211486364458,
+      "grad_norm": 1.3280601501464844,
+      "learning_rate": 1.9986526902114064e-05,
+      "loss": 0.09419863224029541,
+      "mean_token_accuracy": 0.9755162551999093,
+      "num_tokens": 1071092.0,
+      "step": 320
+    },
+    {
+      "entropy": 2.16340401917696,
+      "epoch": 0.23839624345313346,
+      "grad_norm": 0.6447048783302307,
+      "learning_rate": 1.9983990643336007e-05,
+      "loss": 0.10957064628601074,
+      "mean_token_accuracy": 0.9793596610426902,
+      "num_tokens": 1110011.0,
+      "step": 330
+    },
+    {
+      "entropy": 2.1304306954145433,
+      "epoch": 0.24562037204262235,
+      "grad_norm": 5.598320007324219,
+      "learning_rate": 1.9981235984664363e-05,
+      "loss": 0.07113409042358398,
+      "mean_token_accuracy": 0.9785689562559128,
+      "num_tokens": 1145119.0,
+      "step": 340
+    },
+    {
+      "entropy": 2.156084191799164,
+      "epoch": 0.2528445006321112,
+      "grad_norm": 0.9510290026664734,
+      "learning_rate": 1.9978262986357305e-05,
+      "loss": 0.1030649185180664,
+      "mean_token_accuracy": 0.9748681783676147,
+      "num_tokens": 1178189.0,
+      "step": 350
+    },
+    {
+      "entropy": 2.195864510536194,
+      "epoch": 0.2600686292216001,
+      "grad_norm": 4.438415050506592,
+      "learning_rate": 1.997507171344921e-05,
+      "loss": 0.07983499765396118,
+      "mean_token_accuracy": 0.9773692816495896,
+      "num_tokens": 1211561.0,
+      "step": 360
+    },
+    {
+      "entropy": 2.2007741034030914,
+      "epoch": 0.267292757811089,
+      "grad_norm": 0.4064837396144867,
+      "learning_rate": 1.99716622357492e-05,
+      "loss": 0.060518258810043336,
+      "mean_token_accuracy": 0.9889198035001755,
+      "num_tokens": 1240872.0,
+      "step": 370
+    },
+    {
+      "entropy": 2.1124327674508097,
+      "epoch": 0.2745168864005779,
+      "grad_norm": 1.925007939338684,
+      "learning_rate": 1.9968034627839637e-05,
+      "loss": 0.07150581479072571,
+      "mean_token_accuracy": 0.9760589763522148,
+      "num_tokens": 1276234.0,
+      "step": 380
+    },
+    {
+      "entropy": 2.139196202158928,
+      "epoch": 0.2817410149900668,
+      "grad_norm": 2.570542335510254,
+      "learning_rate": 1.996418896907448e-05,
+      "loss": 0.08276339173316956,
+      "mean_token_accuracy": 0.9852717742323875,
+      "num_tokens": 1309890.0,
+      "step": 390
+    },
+    {
+      "entropy": 2.1617587089538572,
+      "epoch": 0.2889651435795557,
+      "grad_norm": 4.684780120849609,
+      "learning_rate": 1.9960125343577547e-05,
+      "loss": 0.06835437417030335,
+      "mean_token_accuracy": 0.9801086381077766,
+      "num_tokens": 1343597.0,
+      "step": 400
+    },
+    {
+      "entropy": 2.117720311880112,
+      "epoch": 0.2961892721690446,
+      "grad_norm": 0.42047184705734253,
+      "learning_rate": 1.995584384024069e-05,
+      "loss": 0.06545560359954834,
+      "mean_token_accuracy": 0.9876575991511345,
+      "num_tokens": 1377821.0,
+      "step": 410
+    },
+    {
+      "entropy": 2.1088694870471953,
+      "epoch": 0.3034134007585335,
+      "grad_norm": 3.0441112518310547,
+      "learning_rate": 1.995134455272183e-05,
+      "loss": 0.04034406542778015,
+      "mean_token_accuracy": 0.9879964560270309,
+      "num_tokens": 1412577.0,
+      "step": 420
+    },
+    {
+      "entropy": 2.1040756702423096,
+      "epoch": 0.3106375293480224,
+      "grad_norm": 1.2212275266647339,
+      "learning_rate": 1.994662757944293e-05,
+      "loss": 0.05455552339553833,
+      "mean_token_accuracy": 0.981954412162304,
+      "num_tokens": 1449199.0,
+      "step": 430
+    },
+    {
+      "entropy": 2.1250744581222536,
+      "epoch": 0.3178616579375113,
+      "grad_norm": 1.0598998069763184,
+      "learning_rate": 1.9941693023587835e-05,
+      "loss": 0.06651412844657897,
+      "mean_token_accuracy": 0.9793136209249497,
+      "num_tokens": 1481363.0,
+      "step": 440
+    },
+    {
+      "entropy": 2.1289766818284988,
+      "epoch": 0.32508578652700015,
+      "grad_norm": 1.3468321561813354,
+      "learning_rate": 1.9936540993099992e-05,
+      "loss": 0.02549194097518921,
+      "mean_token_accuracy": 0.9882893145084382,
+      "num_tokens": 1511239.0,
+      "step": 450
+    },
+    {
+      "entropy": 2.148378947377205,
+      "epoch": 0.33230991511648905,
+      "grad_norm": 0.3286466896533966,
+      "learning_rate": 1.9931171600680127e-05,
+      "loss": 0.05644198060035706,
+      "mean_token_accuracy": 0.9872142806649208,
+      "num_tokens": 1547292.0,
+      "step": 460
+    },
+    {
+      "entropy": 2.106003427505493,
+      "epoch": 0.33953404370597795,
+      "grad_norm": 0.10986531525850296,
+      "learning_rate": 1.992558496378374e-05,
+      "loss": 0.08314967751502991,
+      "mean_token_accuracy": 0.9774487376213074,
+      "num_tokens": 1581706.0,
+      "step": 470
+    },
+    {
+      "entropy": 2.180526411533356,
+      "epoch": 0.34675817229546685,
+      "grad_norm": 4.3229169845581055,
+      "learning_rate": 1.991978120461858e-05,
+      "loss": 0.055954158306121826,
+      "mean_token_accuracy": 0.9919549554586411,
+      "num_tokens": 1612881.0,
+      "step": 480
+    },
+    {
+      "entropy": 2.1669235289096833,
+      "epoch": 0.35398230088495575,
+      "grad_norm": 1.8098896741867065,
+      "learning_rate": 1.9913760450141934e-05,
+      "loss": 0.079369056224823,
+      "mean_token_accuracy": 0.9778457656502724,
+      "num_tokens": 1642879.0,
+      "step": 490
+    },
+    {
+      "entropy": 2.1620357751846315,
+      "epoch": 0.36120642947444465,
+      "grad_norm": 0.16477970778942108,
+      "learning_rate": 1.9907522832057858e-05,
+      "loss": 0.06561388969421386,
+      "mean_token_accuracy": 0.9901054993271827,
+      "num_tokens": 1676452.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.36120642947444465,
+      "eval_entropy": 2.0017294202354137,
+      "eval_loss": 1.0213959217071533,
+      "eval_mean_token_accuracy": 0.843235912928387,
+      "eval_num_tokens": 1676452.0,
+      "eval_runtime": 1007.0825,
+      "eval_samples_per_second": 7.597,
+      "eval_steps_per_second": 0.95,
+      "step": 500
+    },
+    {
+      "entropy": 2.121370476484299,
+      "epoch": 0.36843055806393354,
+      "grad_norm": 5.161382675170898,
+      "learning_rate": 1.9901068486814315e-05,
+      "loss": 0.09131132364273072,
+      "mean_token_accuracy": 0.9809340670704841,
+      "num_tokens": 1711347.0,
+      "step": 510
+    },
+    {
+      "entropy": 2.1376258194446565,
+      "epoch": 0.37565468665342244,
+      "grad_norm": 0.20808692276477814,
+      "learning_rate": 1.9894397555600168e-05,
+      "loss": 0.05019047260284424,
+      "mean_token_accuracy": 0.9852594628930091,
+      "num_tokens": 1748117.0,
+      "step": 520
+    },
+    {
+      "entropy": 2.1469887882471084,
+      "epoch": 0.38287881524291134,
+      "grad_norm": 0.4585625231266022,
+      "learning_rate": 1.9887510184342108e-05,
+      "loss": 0.021144218742847443,
+      "mean_token_accuracy": 0.9943749994039536,
+      "num_tokens": 1779161.0,
+      "step": 530
+    },
+    {
+      "entropy": 2.1128806680440904,
+      "epoch": 0.39010294383240024,
+      "grad_norm": 0.17531993985176086,
+      "learning_rate": 1.9880406523701444e-05,
+      "loss": 0.037897059321403505,
+      "mean_token_accuracy": 0.9935355380177497,
+      "num_tokens": 1817476.0,
+      "step": 540
+    },
+    {
+      "entropy": 2.1039916306734083,
+      "epoch": 0.3973270724218891,
+      "grad_norm": 12.807392120361328,
+      "learning_rate": 1.9873086729070822e-05,
+      "loss": 0.04961663484573364,
+      "mean_token_accuracy": 0.9899179771542549,
+      "num_tokens": 1850699.0,
+      "step": 550
+    },
+    {
+      "entropy": 2.08780357837677,
+      "epoch": 0.404551201011378,
+      "grad_norm": 1.5689797401428223,
+      "learning_rate": 1.9865550960570824e-05,
+      "loss": 0.09309354424476624,
+      "mean_token_accuracy": 0.9854865536093712,
+      "num_tokens": 1880712.0,
+      "step": 560
+    },
+    {
+      "entropy": 2.079355040192604,
+      "epoch": 0.4117753296008669,
+      "grad_norm": 1.0645219087600708,
+      "learning_rate": 1.9857799383046465e-05,
+      "loss": 0.035767218470573424,
+      "mean_token_accuracy": 0.9891987189650535,
+      "num_tokens": 1919614.0,
+      "step": 570
+    },
+    {
+      "entropy": 2.1122144132852556,
+      "epoch": 0.4189994581903558,
+      "grad_norm": 0.2414686679840088,
+      "learning_rate": 1.984983216606357e-05,
+      "loss": 0.03828282654285431,
+      "mean_token_accuracy": 0.9935470074415207,
+      "num_tokens": 1950236.0,
+      "step": 580
+    },
+    {
+      "entropy": 2.1520024478435515,
+      "epoch": 0.4262235867798447,
+      "grad_norm": 0.025072164833545685,
+      "learning_rate": 1.9841649483905092e-05,
+      "loss": 0.026039117574691774,
+      "mean_token_accuracy": 0.9961486741900444,
+      "num_tokens": 1979716.0,
+      "step": 590
+    },
+    {
+      "entropy": 2.1203277856111526,
+      "epoch": 0.4334477153693336,
+      "grad_norm": 0.03449135646224022,
+      "learning_rate": 1.983325151556729e-05,
+      "loss": 0.06812232732772827,
+      "mean_token_accuracy": 0.9831810146570206,
+      "num_tokens": 2014478.0,
+      "step": 600
+    },
+    {
+      "entropy": 2.0620292097330095,
+      "epoch": 0.4406718439588225,
+      "grad_norm": 1.8488832712173462,
+      "learning_rate": 1.9824638444755795e-05,
+      "loss": 0.012539258599281311,
+      "mean_token_accuracy": 0.9972435891628265,
+      "num_tokens": 2051436.0,
+      "step": 610
+    },
+    {
+      "entropy": 2.1081482648849486,
+      "epoch": 0.44789597254831137,
+      "grad_norm": 0.1524757593870163,
+      "learning_rate": 1.981581045988162e-05,
+      "loss": 0.061668336391448975,
+      "mean_token_accuracy": 0.9885178819298744,
+      "num_tokens": 2086848.0,
+      "step": 620
+    },
+    {
+      "entropy": 2.153286537528038,
+      "epoch": 0.45512010113780027,
+      "grad_norm": 0.20357979834079742,
+      "learning_rate": 1.9806767754057022e-05,
+      "loss": 0.024441687762737273,
+      "mean_token_accuracy": 0.9957340195775032,
+      "num_tokens": 2119688.0,
+      "step": 630
+    },
+    {
+      "entropy": 2.109396052360535,
+      "epoch": 0.46234422972728917,
+      "grad_norm": 0.03233234956860542,
+      "learning_rate": 1.9797510525091278e-05,
+      "loss": 0.004525944963097573,
+      "mean_token_accuracy": 0.9989999994635582,
+      "num_tokens": 2155633.0,
+      "step": 640
+    },
+    {
+      "entropy": 2.1455644100904463,
+      "epoch": 0.469568358316778,
+      "grad_norm": 0.5513462424278259,
+      "learning_rate": 1.9788038975486363e-05,
+      "loss": 0.018023833632469177,
+      "mean_token_accuracy": 0.9958193480968476,
+      "num_tokens": 2190543.0,
+      "step": 650
+    },
+    {
+      "entropy": 2.0951803505420683,
+      "epoch": 0.4767924869062669,
+      "grad_norm": 0.05144621431827545,
+      "learning_rate": 1.977835331243252e-05,
+      "loss": 0.011193668842315674,
+      "mean_token_accuracy": 0.9985294118523598,
+      "num_tokens": 2225664.0,
+      "step": 660
+    },
+    {
+      "entropy": 2.05837681889534,
+      "epoch": 0.4840166154957558,
+      "grad_norm": 1.9019652605056763,
+      "learning_rate": 1.9768453747803722e-05,
+      "loss": 0.045949330925941466,
+      "mean_token_accuracy": 0.9922287076711654,
+      "num_tokens": 2263929.0,
+      "step": 670
+    },
+    {
+      "entropy": 2.1198449566960336,
+      "epoch": 0.4912407440852447,
+      "grad_norm": 2.392086982727051,
+      "learning_rate": 1.975834049815304e-05,
+      "loss": 0.04401857256889343,
+      "mean_token_accuracy": 0.9881693422794342,
+      "num_tokens": 2300646.0,
+      "step": 680
+    },
+    {
+      "entropy": 2.137920472025871,
+      "epoch": 0.4984648726747336,
+      "grad_norm": 0.7350146770477295,
+      "learning_rate": 1.9748013784707914e-05,
+      "loss": 0.04097718000411987,
+      "mean_token_accuracy": 0.9914300963282585,
+      "num_tokens": 2337419.0,
+      "step": 690
+    },
+    {
+      "entropy": 2.152908131480217,
+      "epoch": 0.5056890012642224,
+      "grad_norm": 2.7877094745635986,
+      "learning_rate": 1.9737473833365303e-05,
+      "loss": 0.06197558641433716,
+      "mean_token_accuracy": 0.9885002598166466,
+      "num_tokens": 2370156.0,
+      "step": 700
+    },
+    {
+      "entropy": 2.114130911231041,
+      "epoch": 0.5129131298537114,
+      "grad_norm": 4.815999984741211,
+      "learning_rate": 1.972672087468674e-05,
+      "loss": 0.036697188019752504,
+      "mean_token_accuracy": 0.9890773802995682,
+      "num_tokens": 2403350.0,
+      "step": 710
+    },
+    {
+      "entropy": 2.1276988863945006,
+      "epoch": 0.5201372584432002,
+      "grad_norm": 0.09675746411085129,
+      "learning_rate": 1.971575514389331e-05,
+      "loss": 0.025695490837097167,
+      "mean_token_accuracy": 0.9903780594468117,
+      "num_tokens": 2439829.0,
+      "step": 720
+    },
+    {
+      "entropy": 2.1967761904001235,
+      "epoch": 0.5273613870326892,
+      "grad_norm": 0.11605588346719742,
+      "learning_rate": 1.9704576880860474e-05,
+      "loss": 0.03519304394721985,
+      "mean_token_accuracy": 0.9883782863616943,
+      "num_tokens": 2474018.0,
+      "step": 730
+    },
+    {
+      "entropy": 2.1506566435098646,
+      "epoch": 0.534585515622178,
+      "grad_norm": 2.496562957763672,
+      "learning_rate": 1.969318633011285e-05,
+      "loss": 0.02293294370174408,
+      "mean_token_accuracy": 0.9894841268658638,
+      "num_tokens": 2511615.0,
+      "step": 740
+    },
+    {
+      "entropy": 2.1656083434820177,
+      "epoch": 0.541809644211667,
+      "grad_norm": 0.48594969511032104,
+      "learning_rate": 1.968158374081885e-05,
+      "loss": 0.05291303396224976,
+      "mean_token_accuracy": 0.9936918020248413,
+      "num_tokens": 2546286.0,
+      "step": 750
+    },
+    {
+      "entropy": 2.2043658822774885,
+      "epoch": 0.5490337728011558,
+      "grad_norm": 6.537981986999512,
+      "learning_rate": 1.9669769366785223e-05,
+      "loss": 0.026339799165725708,
+      "mean_token_accuracy": 0.9959500461816788,
+      "num_tokens": 2579815.0,
+      "step": 760
+    },
+    {
+      "entropy": 2.187710216641426,
+      "epoch": 0.5562579013906448,
+      "grad_norm": 0.06361158192157745,
+      "learning_rate": 1.9657743466451524e-05,
+      "loss": 0.03191069066524506,
+      "mean_token_accuracy": 0.9880474433302879,
+      "num_tokens": 2608498.0,
+      "step": 770
+    },
+    {
+      "entropy": 2.130194824934006,
+      "epoch": 0.5634820299801336,
+      "grad_norm": 5.55103063583374,
+      "learning_rate": 1.964550630288444e-05,
+      "loss": 0.0548154354095459,
+      "mean_token_accuracy": 0.986955127120018,
+      "num_tokens": 2642930.0,
+      "step": 780
+    },
+    {
+      "entropy": 2.127747783064842,
+      "epoch": 0.5707061585696226,
+      "grad_norm": 2.8869001865386963,
+      "learning_rate": 1.9633058143772037e-05,
+      "loss": 0.016472166776657103,
+      "mean_token_accuracy": 0.9959523811936378,
+      "num_tokens": 2672358.0,
+      "step": 790
+    },
+    {
+      "entropy": 2.1343603670597076,
+      "epoch": 0.5779302871591114,
+      "grad_norm": 6.028234481811523,
+      "learning_rate": 1.9620399261417925e-05,
+      "loss": 0.0534799337387085,
+      "mean_token_accuracy": 0.9870725125074387,
+      "num_tokens": 2712309.0,
+      "step": 800
+    },
+    {
+      "entropy": 2.1584979057312013,
+      "epoch": 0.5851544157486003,
+      "grad_norm": 0.13096173107624054,
+      "learning_rate": 1.960752993273528e-05,
+      "loss": 0.039956134557723996,
+      "mean_token_accuracy": 0.9930102661252022,
+      "num_tokens": 2750765.0,
+      "step": 810
+    },
+    {
+      "entropy": 2.0611986249685286,
+      "epoch": 0.5923785443380892,
+      "grad_norm": 0.45354947447776794,
+      "learning_rate": 1.9594450439240793e-05,
+      "loss": 0.049787935614585874,
+      "mean_token_accuracy": 0.9887179493904114,
+      "num_tokens": 2787510.0,
+      "step": 820
+    },
+    {
+      "entropy": 2.092721477150917,
+      "epoch": 0.5996026729275781,
+      "grad_norm": 0.033594291657209396,
+      "learning_rate": 1.958116106704851e-05,
+      "loss": 0.013710010051727294,
+      "mean_token_accuracy": 0.9969642847776413,
+      "num_tokens": 2820507.0,
+      "step": 830
+    },
+    {
+      "entropy": 2.110324975848198,
+      "epoch": 0.606826801517067,
+      "grad_norm": 1.94284987449646,
+      "learning_rate": 1.9567662106863583e-05,
+      "loss": 0.05006279349327088,
+      "mean_token_accuracy": 0.9929195806384087,
+      "num_tokens": 2856832.0,
+      "step": 840
+    },
+    {
+      "entropy": 2.146149677038193,
+      "epoch": 0.6140509301065559,
+      "grad_norm": 0.025745930150151253,
+      "learning_rate": 1.9553953853975894e-05,
+      "loss": 0.049893584847450254,
+      "mean_token_accuracy": 0.9861560463905334,
+      "num_tokens": 2886911.0,
+      "step": 850
+    },
+    {
+      "entropy": 2.2159456998109817,
+      "epoch": 0.6212750586960448,
+      "grad_norm": 0.22102245688438416,
+      "learning_rate": 1.9540036608253615e-05,
+      "loss": 0.011719506233930588,
+      "mean_token_accuracy": 0.9975980386137963,
+      "num_tokens": 2918768.0,
+      "step": 860
+    },
+    {
+      "entropy": 2.240058869123459,
+      "epoch": 0.6284991872855337,
+      "grad_norm": 3.7185215950012207,
+      "learning_rate": 1.952591067413663e-05,
+      "loss": 0.02348371744155884,
+      "mean_token_accuracy": 0.9929980143904686,
+      "num_tokens": 2951922.0,
+      "step": 870
+    },
+    {
+      "entropy": 2.1650724798440932,
+      "epoch": 0.6357233158750226,
+      "grad_norm": 0.26696911454200745,
+      "learning_rate": 1.9511576360629884e-05,
+      "loss": 0.013108362257480622,
+      "mean_token_accuracy": 0.9952360153198242,
+      "num_tokens": 2980067.0,
+      "step": 880
+    },
+    {
+      "entropy": 2.1337937146425245,
+      "epoch": 0.6429474444645115,
+      "grad_norm": 0.017593568190932274,
+      "learning_rate": 1.9497033981296635e-05,
+      "loss": 0.0056190770119428635,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 3011282.0,
+      "step": 890
+    },
+    {
+      "entropy": 2.104154738783836,
+      "epoch": 0.6501715730540003,
+      "grad_norm": 0.03523773327469826,
+      "learning_rate": 1.9482283854251564e-05,
+      "loss": 0.05853831768035889,
+      "mean_token_accuracy": 0.988778468966484,
+      "num_tokens": 3046943.0,
+      "step": 900
+    },
+    {
+      "entropy": 2.071388727426529,
+      "epoch": 0.6573957016434893,
+      "grad_norm": 2.63916015625,
+      "learning_rate": 1.9467326302153854e-05,
+      "loss": 0.015439648926258088,
+      "mean_token_accuracy": 0.9967436969280243,
+      "num_tokens": 3086686.0,
+      "step": 910
+    },
+    {
+      "entropy": 2.1768156826496123,
+      "epoch": 0.6646198302329781,
+      "grad_norm": 7.410099029541016,
+      "learning_rate": 1.9452161652200096e-05,
+      "loss": 0.035445448756217954,
+      "mean_token_accuracy": 0.9946778699755668,
+      "num_tokens": 3116964.0,
+      "step": 920
+    },
+    {
+      "entropy": 2.113570731878281,
+      "epoch": 0.6718439588224671,
+      "grad_norm": 12.763428688049316,
+      "learning_rate": 1.9436790236117162e-05,
+      "loss": 0.05357277989387512,
+      "mean_token_accuracy": 0.989736121892929,
+      "num_tokens": 3151640.0,
+      "step": 930
+    },
+    {
+      "entropy": 2.151062309741974,
+      "epoch": 0.6790680874119559,
+      "grad_norm": 4.3073811531066895,
+      "learning_rate": 1.942121239015493e-05,
+      "loss": 0.02954191863536835,
+      "mean_token_accuracy": 0.9931784182786941,
+      "num_tokens": 3183172.0,
+      "step": 940
+    },
+    {
+      "entropy": 2.140563413500786,
+      "epoch": 0.6862922160014449,
+      "grad_norm": 0.015436495654284954,
+      "learning_rate": 1.9405428455078933e-05,
+      "loss": 0.021537141501903535,
+      "mean_token_accuracy": 0.9967385903000832,
+      "num_tokens": 3215196.0,
+      "step": 950
+    },
+    {
+      "entropy": 2.1755530834198,
+      "epoch": 0.6935163445909337,
+      "grad_norm": 0.1176186203956604,
+      "learning_rate": 1.938943877616291e-05,
+      "loss": 0.041554605960845946,
+      "mean_token_accuracy": 0.9923611104488372,
+      "num_tokens": 3250033.0,
+      "step": 960
+    },
+    {
+      "entropy": 2.160807254910469,
+      "epoch": 0.7007404731804227,
+      "grad_norm": 0.4536968469619751,
+      "learning_rate": 1.9373243703181234e-05,
+      "loss": 0.022222673892974852,
+      "mean_token_accuracy": 0.9943934962153435,
+      "num_tokens": 3284275.0,
+      "step": 970
+    },
+    {
+      "entropy": 2.118019551038742,
+      "epoch": 0.7079646017699115,
+      "grad_norm": 0.028539419174194336,
+      "learning_rate": 1.9356843590401295e-05,
+      "loss": 0.03024633228778839,
+      "mean_token_accuracy": 0.993554450571537,
+      "num_tokens": 3318920.0,
+      "step": 980
+    },
+    {
+      "entropy": 2.115340527892113,
+      "epoch": 0.7151887303594004,
+      "grad_norm": 0.022816361859440804,
+      "learning_rate": 1.9340238796575722e-05,
+      "loss": 0.029753103852272034,
+      "mean_token_accuracy": 0.991157278418541,
+      "num_tokens": 3355066.0,
+      "step": 990
+    },
+    {
+      "entropy": 2.0531561613082885,
+      "epoch": 0.7224128589488893,
+      "grad_norm": 5.430459499359131,
+      "learning_rate": 1.932342968493454e-05,
+      "loss": 0.03846350014209747,
+      "mean_token_accuracy": 0.9928767189383507,
+      "num_tokens": 3396801.0,
+      "step": 1000
+    },
+    {
+      "epoch": 0.7224128589488893,
+      "eval_entropy": 2.015599282557688,
+      "eval_loss": 0.9796417355537415,
+      "eval_mean_token_accuracy": 0.8523120447261455,
+      "eval_num_tokens": 3396801.0,
+      "eval_runtime": 1007.5014,
+      "eval_samples_per_second": 7.594,
+      "eval_steps_per_second": 0.95,
+      "step": 1000
+    },
+    {
+      "entropy": 2.2311429172754287,
+      "epoch": 0.7296369875383781,
+      "grad_norm": 0.3221210539340973,
+      "learning_rate": 1.9306416623177233e-05,
+      "loss": 0.02508760690689087,
+      "mean_token_accuracy": 0.9961589574813843,
+      "num_tokens": 3424673.0,
+      "step": 1010
+    },
+    {
+      "entropy": 2.1565969973802566,
+      "epoch": 0.7368611161278671,
+      "grad_norm": 0.07839540392160416,
+      "learning_rate": 1.928919998346469e-05,
+      "loss": 0.02089208662509918,
+      "mean_token_accuracy": 0.9951127827167511,
+      "num_tokens": 3460589.0,
+      "step": 1020
+    },
+    {
+      "entropy": 2.1316963762044905,
+      "epoch": 0.7440852447173559,
+      "grad_norm": 2.571152925491333,
+      "learning_rate": 1.927178014241108e-05,
+      "loss": 0.01810687780380249,
+      "mean_token_accuracy": 0.9948151633143425,
+      "num_tokens": 3492356.0,
+      "step": 1030
+    },
+    {
+      "entropy": 2.131281113624573,
+      "epoch": 0.7513093733068449,
+      "grad_norm": 1.0611048936843872,
+      "learning_rate": 1.9254157481075593e-05,
+      "loss": 0.011593367159366607,
+      "mean_token_accuracy": 0.9969405591487884,
+      "num_tokens": 3528933.0,
+      "step": 1040
+    },
+    {
+      "entropy": 2.1135970920324327,
+      "epoch": 0.7585335018963337,
+      "grad_norm": 0.06862345337867737,
+      "learning_rate": 1.923633238495412e-05,
+      "loss": 0.03084944188594818,
+      "mean_token_accuracy": 0.9917931839823723,
+      "num_tokens": 3562754.0,
+      "step": 1050
+    },
+    {
+      "entropy": 2.135847768187523,
+      "epoch": 0.7657576304858227,
+      "grad_norm": 3.3710029125213623,
+      "learning_rate": 1.9218305243970824e-05,
+      "loss": 0.03355285227298736,
+      "mean_token_accuracy": 0.9917197063565254,
+      "num_tokens": 3596291.0,
+      "step": 1060
+    },
+    {
+      "entropy": 2.115863263607025,
+      "epoch": 0.7729817590753115,
+      "grad_norm": 1.0154095888137817,
+      "learning_rate": 1.9200076452469575e-05,
+      "loss": 0.027915269136428833,
+      "mean_token_accuracy": 0.9920019909739495,
+      "num_tokens": 3634202.0,
+      "step": 1070
+    },
+    {
+      "entropy": 2.180663511157036,
+      "epoch": 0.7802058876648005,
+      "grad_norm": 0.008630992844700813,
+      "learning_rate": 1.9181646409205376e-05,
+      "loss": 0.01911990940570831,
+      "mean_token_accuracy": 0.9977481618523598,
+      "num_tokens": 3665654.0,
+      "step": 1080
+    },
+    {
+      "entropy": 2.066094806790352,
+      "epoch": 0.7874300162542893,
+      "grad_norm": 0.016932925209403038,
+      "learning_rate": 1.9163015517335605e-05,
+      "loss": 0.0038531143218278885,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 3701414.0,
+      "step": 1090
+    },
+    {
+      "entropy": 2.0516481667757036,
+      "epoch": 0.7946541448437782,
+      "grad_norm": 0.9654312133789062,
+      "learning_rate": 1.9144184184411204e-05,
+      "loss": 0.023603864014148712,
+      "mean_token_accuracy": 0.9940719693899155,
+      "num_tokens": 3738650.0,
+      "step": 1100
+    },
+    {
+      "entropy": 2.1311365902423858,
+      "epoch": 0.8018782734332671,
+      "grad_norm": 0.07597270607948303,
+      "learning_rate": 1.9125152822367764e-05,
+      "loss": 0.008270335197448731,
+      "mean_token_accuracy": 0.997916667163372,
+      "num_tokens": 3769387.0,
+      "step": 1110
+    },
+    {
+      "entropy": 2.1482642620801924,
+      "epoch": 0.809102402022756,
+      "grad_norm": 0.05144130066037178,
+      "learning_rate": 1.910592184751652e-05,
+      "loss": 0.008368145674467087,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 3795728.0,
+      "step": 1120
+    },
+    {
+      "entropy": 2.1289342790842056,
+      "epoch": 0.8163265306122449,
+      "grad_norm": 4.872783660888672,
+      "learning_rate": 1.908649168053523e-05,
+      "loss": 0.043704470992088316,
+      "mean_token_accuracy": 0.9917654454708099,
+      "num_tokens": 3824859.0,
+      "step": 1130
+    },
+    {
+      "entropy": 2.0587215080857275,
+      "epoch": 0.8235506592017338,
+      "grad_norm": 0.3473055064678192,
+      "learning_rate": 1.9066862746458998e-05,
+      "loss": 0.01864089071750641,
+      "mean_token_accuracy": 0.9959340661764144,
+      "num_tokens": 3860239.0,
+      "step": 1140
+    },
+    {
+      "entropy": 2.077919548749924,
+      "epoch": 0.8307747877912227,
+      "grad_norm": 1.1250643730163574,
+      "learning_rate": 1.904703547467094e-05,
+      "loss": 0.047300055623054504,
+      "mean_token_accuracy": 0.9930105939507484,
+      "num_tokens": 3893211.0,
+      "step": 1150
+    },
+    {
+      "entropy": 2.1111993461847307,
+      "epoch": 0.8379989163807116,
+      "grad_norm": 0.04652652144432068,
+      "learning_rate": 1.902701029889283e-05,
+      "loss": 0.042614486813545224,
+      "mean_token_accuracy": 0.9891733899712563,
+      "num_tokens": 3923474.0,
+      "step": 1160
+    },
+    {
+      "entropy": 2.133824098110199,
+      "epoch": 0.8452230449702005,
+      "grad_norm": 4.08588981628418,
+      "learning_rate": 1.900678765717558e-05,
+      "loss": 0.03261755108833313,
+      "mean_token_accuracy": 0.9877159208059311,
+      "num_tokens": 3952262.0,
+      "step": 1170
+    },
+    {
+      "entropy": 2.1812889873981476,
+      "epoch": 0.8524471735596894,
+      "grad_norm": 0.01818382926285267,
+      "learning_rate": 1.8986367991889677e-05,
+      "loss": 0.004279704764485359,
+      "mean_token_accuracy": 0.9988872766494751,
+      "num_tokens": 3982302.0,
+      "step": 1180
+    },
+    {
+      "entropy": 2.111463187634945,
+      "epoch": 0.8596713021491783,
+      "grad_norm": 3.6590194702148438,
+      "learning_rate": 1.8965751749715502e-05,
+      "loss": 0.04557798206806183,
+      "mean_token_accuracy": 0.9872519835829735,
+      "num_tokens": 4017674.0,
+      "step": 1190
+    },
+    {
+      "entropy": 2.1546395272016525,
+      "epoch": 0.8668954307386671,
+      "grad_norm": 0.011192182078957558,
+      "learning_rate": 1.8944939381633553e-05,
+      "loss": 0.015546497702598572,
+      "mean_token_accuracy": 0.9963636353611947,
+      "num_tokens": 4049489.0,
+      "step": 1200
+    },
+    {
+      "entropy": 2.061059150099754,
+      "epoch": 0.874119559328156,
+      "grad_norm": 0.2434389889240265,
+      "learning_rate": 1.8923931342914592e-05,
+      "loss": 0.0059484150260686874,
+      "mean_token_accuracy": 0.9977272734045982,
+      "num_tokens": 4085651.0,
+      "step": 1210
+    },
+    {
+      "entropy": 2.11140855550766,
+      "epoch": 0.881343687917645,
+      "grad_norm": 5.586461544036865,
+      "learning_rate": 1.8902728093109668e-05,
+      "loss": 0.018709875643253326,
+      "mean_token_accuracy": 0.9961805552244186,
+      "num_tokens": 4115626.0,
+      "step": 1220
+    },
+    {
+      "entropy": 2.0685029208660124,
+      "epoch": 0.8885678165071338,
+      "grad_norm": 6.326735973358154,
+      "learning_rate": 1.888133009604008e-05,
+      "loss": 0.024366311728954315,
+      "mean_token_accuracy": 0.9944709703326226,
+      "num_tokens": 4148992.0,
+      "step": 1230
+    },
+    {
+      "entropy": 2.104837965965271,
+      "epoch": 0.8957919450966227,
+      "grad_norm": 14.592947006225586,
+      "learning_rate": 1.8859737819787224e-05,
+      "loss": 0.023436115682125093,
+      "mean_token_accuracy": 0.9899750709533691,
+      "num_tokens": 4181714.0,
+      "step": 1240
+    },
+    {
+      "entropy": 2.055359047651291,
+      "epoch": 0.9030160736861116,
+      "grad_norm": 0.028958171606063843,
+      "learning_rate": 1.8837951736682353e-05,
+      "loss": 0.0017842631787061691,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 4220903.0,
+      "step": 1250
+    },
+    {
+      "entropy": 2.0872839361429216,
+      "epoch": 0.9102402022756005,
+      "grad_norm": 0.036007024347782135,
+      "learning_rate": 1.8815972323296258e-05,
+      "loss": 0.01379597932100296,
+      "mean_token_accuracy": 0.9942857146263122,
+      "num_tokens": 4251549.0,
+      "step": 1260
+    },
+    {
+      "entropy": 2.0923456490039825,
+      "epoch": 0.9174643308650894,
+      "grad_norm": 0.7058573365211487,
+      "learning_rate": 1.8793800060428814e-05,
+      "loss": 0.02520722448825836,
+      "mean_token_accuracy": 0.9965818896889687,
+      "num_tokens": 4287420.0,
+      "step": 1270
+    },
+    {
+      "entropy": 2.1157235205173492,
+      "epoch": 0.9246884594545783,
+      "grad_norm": 0.39984047412872314,
+      "learning_rate": 1.8771435433098493e-05,
+      "loss": 0.03538358211517334,
+      "mean_token_accuracy": 0.9956959694623947,
+      "num_tokens": 4318898.0,
+      "step": 1280
+    },
+    {
+      "entropy": 2.108782610297203,
+      "epoch": 0.9319125880440672,
+      "grad_norm": 0.09862552583217621,
+      "learning_rate": 1.8748878930531743e-05,
+      "loss": 0.033087140321731566,
+      "mean_token_accuracy": 0.9920297145843506,
+      "num_tokens": 4351356.0,
+      "step": 1290
+    },
+    {
+      "entropy": 2.093110492825508,
+      "epoch": 0.939136716633556,
+      "grad_norm": 0.00461851991713047,
+      "learning_rate": 1.8726131046152272e-05,
+      "loss": 0.010160531848669052,
+      "mean_token_accuracy": 0.996524153649807,
+      "num_tokens": 4384553.0,
+      "step": 1300
+    },
+    {
+      "entropy": 2.032593864202499,
+      "epoch": 0.946360845223045,
+      "grad_norm": 0.5353080034255981,
+      "learning_rate": 1.8703192277570277e-05,
+      "loss": 0.013478608429431915,
+      "mean_token_accuracy": 0.996477273106575,
+      "num_tokens": 4418996.0,
+      "step": 1310
+    },
+    {
+      "entropy": 2.1145312935113907,
+      "epoch": 0.9535849738125338,
+      "grad_norm": 0.10275590419769287,
+      "learning_rate": 1.8680063126571547e-05,
+      "loss": 0.002380320429801941,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 4450751.0,
+      "step": 1320
+    },
+    {
+      "entropy": 2.1038990378379823,
+      "epoch": 0.9608091024020228,
+      "grad_norm": 2.1970865726470947,
+      "learning_rate": 1.865674409910649e-05,
+      "loss": 0.01084609180688858,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 4481621.0,
+      "step": 1330
+    },
+    {
+      "entropy": 2.012085109949112,
+      "epoch": 0.9680332309915116,
+      "grad_norm": 0.035805828869342804,
+      "learning_rate": 1.863323570527906e-05,
+      "loss": 0.0015195850282907486,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 4518380.0,
+      "step": 1340
+    },
+    {
+      "entropy": 2.0171980261802673,
+      "epoch": 0.9752573595810006,
+      "grad_norm": 0.13926967978477478,
+      "learning_rate": 1.8609538459335595e-05,
+      "loss": 0.03831966817378998,
+      "mean_token_accuracy": 0.9956393390893936,
+      "num_tokens": 4553464.0,
+      "step": 1350
+    },
+    {
+      "entropy": 2.0884960144758224,
+      "epoch": 0.9824814881704894,
+      "grad_norm": 0.05643892660737038,
+      "learning_rate": 1.8585652879653598e-05,
+      "loss": 0.003447505459189415,
+      "mean_token_accuracy": 0.9991379305720329,
+      "num_tokens": 4585078.0,
+      "step": 1360
+    },
+    {
+      "entropy": 2.0204695969820023,
+      "epoch": 0.9897056167599784,
+      "grad_norm": 0.152903750538826,
+      "learning_rate": 1.856157948873035e-05,
+      "loss": 0.008481259644031524,
+      "mean_token_accuracy": 0.9986805558204651,
+      "num_tokens": 4620089.0,
+      "step": 1370
+    },
+    {
+      "entropy": 2.0601888835430144,
+      "epoch": 0.9969297453494672,
+      "grad_norm": 5.249927997589111,
+      "learning_rate": 1.8537318813171516e-05,
+      "loss": 0.014892134070396423,
+      "mean_token_accuracy": 0.9947420626878738,
+      "num_tokens": 4653211.0,
+      "step": 1380
+    },
+    {
+      "entropy": 2.111465578143661,
+      "epoch": 1.0036120642947444,
+      "grad_norm": 0.030989494174718857,
+      "learning_rate": 1.8512871383679617e-05,
+      "loss": 0.02316879332065582,
+      "mean_token_accuracy": 0.9923184062983539,
+      "num_tokens": 4684069.0,
+      "step": 1390
+    },
+    {
+      "entropy": 2.1355521261692045,
+      "epoch": 1.0108361928842333,
+      "grad_norm": 0.013790454715490341,
+      "learning_rate": 1.8488237735042406e-05,
+      "loss": 0.0007782953325659037,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 4716475.0,
+      "step": 1400
+    },
+    {
+      "entropy": 2.0239915043115615,
+      "epoch": 1.0180603214737223,
+      "grad_norm": 0.015894528478384018,
+      "learning_rate": 1.8463418406121198e-05,
+      "loss": 0.008045677840709687,
+      "mean_token_accuracy": 0.9967708334326744,
+      "num_tokens": 4754383.0,
+      "step": 1410
+    },
+    {
+      "entropy": 1.9871837973594666,
+      "epoch": 1.0252844500632112,
+      "grad_norm": 0.017707591876387596,
+      "learning_rate": 1.843841393983905e-05,
+      "loss": 0.02116468399763107,
+      "mean_token_accuracy": 0.995119047164917,
+      "num_tokens": 4794361.0,
+      "step": 1420
+    },
+    {
+      "entropy": 2.0534275233745576,
+      "epoch": 1.0325085786527,
+      "grad_norm": 0.20345048606395721,
+      "learning_rate": 1.8413224883168912e-05,
+      "loss": 0.0025019023567438127,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 4836053.0,
+      "step": 1430
+    },
+    {
+      "entropy": 2.057610747218132,
+      "epoch": 1.039732707242189,
+      "grad_norm": 0.009533772245049477,
+      "learning_rate": 1.8387851787121646e-05,
+      "loss": 0.021535003185272218,
+      "mean_token_accuracy": 0.9924398139119148,
+      "num_tokens": 4868865.0,
+      "step": 1440
+    },
+    {
+      "entropy": 2.119044467806816,
+      "epoch": 1.0469568358316779,
+      "grad_norm": 2.236619472503662,
+      "learning_rate": 1.8362295206733978e-05,
+      "loss": 0.017270827293395997,
+      "mean_token_accuracy": 0.9924002185463905,
+      "num_tokens": 4904908.0,
+      "step": 1450
+    },
+    {
+      "entropy": 2.06851644217968,
+      "epoch": 1.0541809644211666,
+      "grad_norm": 0.06309136003255844,
+      "learning_rate": 1.833655570105635e-05,
+      "loss": 0.0062024291604757305,
+      "mean_token_accuracy": 0.9989130437374115,
+      "num_tokens": 4938538.0,
+      "step": 1460
+    },
+    {
+      "entropy": 2.0950154691934584,
+      "epoch": 1.0614050930106556,
+      "grad_norm": 0.18221551179885864,
+      "learning_rate": 1.8310633833140713e-05,
+      "loss": 0.021643282473087312,
+      "mean_token_accuracy": 0.9981763914227486,
+      "num_tokens": 4971145.0,
+      "step": 1470
+    },
+    {
+      "entropy": 2.0922484278678892,
+      "epoch": 1.0686292216001445,
+      "grad_norm": 0.05216178297996521,
+      "learning_rate": 1.8284530170028176e-05,
+      "loss": 0.0033873818814754487,
+      "mean_token_accuracy": 0.9992424249649048,
+      "num_tokens": 4999892.0,
+      "step": 1480
+    },
+    {
+      "entropy": 2.0764939337968826,
+      "epoch": 1.0758533501896335,
+      "grad_norm": 0.025218123570084572,
+      "learning_rate": 1.8258245282736627e-05,
+      "loss": 0.02744237780570984,
+      "mean_token_accuracy": 0.9955294117331505,
+      "num_tokens": 5034282.0,
+      "step": 1490
+    },
+    {
+      "entropy": 2.0980449438095095,
+      "epoch": 1.0830774787791222,
+      "grad_norm": 0.0042065721936523914,
+      "learning_rate": 1.8231779746248234e-05,
+      "loss": 0.018268114328384398,
+      "mean_token_accuracy": 0.9975520834326744,
+      "num_tokens": 5064549.0,
+      "step": 1500
+    },
+    {
+      "epoch": 1.0830774787791222,
+      "eval_entropy": 1.9504261345698917,
+      "eval_loss": 1.0086095333099365,
+      "eval_mean_token_accuracy": 0.8621797298189241,
+      "eval_num_tokens": 5064549.0,
+      "eval_runtime": 1006.9488,
+      "eval_samples_per_second": 7.598,
+      "eval_steps_per_second": 0.95,
+      "step": 1500
+    },
+    {
+      "entropy": 2.0826266705989838,
+      "epoch": 1.0903016073686111,
+      "grad_norm": 0.23226939141750336,
+      "learning_rate": 1.8205134139496866e-05,
+      "loss": 0.0053476162254810335,
+      "mean_token_accuracy": 0.9989583328366279,
+      "num_tokens": 5098381.0,
+      "step": 1510
+    },
+    {
+      "entropy": 2.1424541413784026,
+      "epoch": 1.0975257359581,
+      "grad_norm": 0.005317744333297014,
+      "learning_rate": 1.817830904535544e-05,
+      "loss": 0.005732352659106254,
+      "mean_token_accuracy": 0.996690820157528,
+      "num_tokens": 5128572.0,
+      "step": 1520
+    },
+    {
+      "entropy": 2.1302927315235136,
+      "epoch": 1.104749864547589,
+      "grad_norm": 0.02699355222284794,
+      "learning_rate": 1.8151305050623148e-05,
+      "loss": 0.011630669981241227,
+      "mean_token_accuracy": 0.9959165081381798,
+      "num_tokens": 5162862.0,
+      "step": 1530
+    },
+    {
+      "entropy": 2.103237381577492,
+      "epoch": 1.1119739931370778,
+      "grad_norm": 0.010451219044625759,
+      "learning_rate": 1.8124122746012642e-05,
+      "loss": 0.012390993535518646,
+      "mean_token_accuracy": 0.9969166666269302,
+      "num_tokens": 5195249.0,
+      "step": 1540
+    },
+    {
+      "entropy": 2.134079247713089,
+      "epoch": 1.1191981217265667,
+      "grad_norm": 1.3870619535446167,
+      "learning_rate": 1.8096762726137106e-05,
+      "loss": 0.0017786189913749696,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5226438.0,
+      "step": 1550
+    },
+    {
+      "entropy": 2.117493250966072,
+      "epoch": 1.1264222503160557,
+      "grad_norm": 1.1784181594848633,
+      "learning_rate": 1.806922558949724e-05,
+      "loss": 0.005632025003433227,
+      "mean_token_accuracy": 0.9968269228935241,
+      "num_tokens": 5260179.0,
+      "step": 1560
+    },
+    {
+      "entropy": 2.1169673800468445,
+      "epoch": 1.1336463789055444,
+      "grad_norm": 0.005867530591785908,
+      "learning_rate": 1.804151193846818e-05,
+      "loss": 0.0018024042248725892,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5295898.0,
+      "step": 1570
+    },
+    {
+      "entropy": 2.154021382331848,
+      "epoch": 1.1408705074950334,
+      "grad_norm": 2.601689577102661,
+      "learning_rate": 1.8013622379286317e-05,
+      "loss": 0.02105346620082855,
+      "mean_token_accuracy": 0.9956855490803719,
+      "num_tokens": 5327650.0,
+      "step": 1580
+    },
+    {
+      "entropy": 2.041078993678093,
+      "epoch": 1.1480946360845223,
+      "grad_norm": 0.16076812148094177,
+      "learning_rate": 1.798555752203603e-05,
+      "loss": 0.0056386619806289675,
+      "mean_token_accuracy": 0.9974999994039535,
+      "num_tokens": 5365261.0,
+      "step": 1590
+    },
+    {
+      "entropy": 2.1064918786287308,
+      "epoch": 1.1553187646740113,
+      "grad_norm": 0.13275979459285736,
+      "learning_rate": 1.795731798063635e-05,
+      "loss": 0.0002915527205914259,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5395638.0,
+      "step": 1600
+    },
+    {
+      "entropy": 2.1877143383026123,
+      "epoch": 1.1625428932635,
+      "grad_norm": 0.021166006103157997,
+      "learning_rate": 1.7928904372827523e-05,
+      "loss": 0.002695506438612938,
+      "mean_token_accuracy": 0.999074074625969,
+      "num_tokens": 5427822.0,
+      "step": 1610
+    },
+    {
+      "entropy": 2.1708212316036226,
+      "epoch": 1.169767021852989,
+      "grad_norm": 0.012815513648092747,
+      "learning_rate": 1.7900317320157497e-05,
+      "loss": 0.012587207555770873,
+      "mean_token_accuracy": 0.999074074625969,
+      "num_tokens": 5457509.0,
+      "step": 1620
+    },
+    {
+      "entropy": 2.1449846118688582,
+      "epoch": 1.176991150442478,
+      "grad_norm": 0.1441592127084732,
+      "learning_rate": 1.7871557447968332e-05,
+      "loss": 0.014236865937709809,
+      "mean_token_accuracy": 0.9972222223877907,
+      "num_tokens": 5497324.0,
+      "step": 1630
+    },
+    {
+      "entropy": 2.148208647966385,
+      "epoch": 1.1842152790319669,
+      "grad_norm": 0.0048500122502446175,
+      "learning_rate": 1.7842625385382514e-05,
+      "loss": 0.00042399438098073006,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5530447.0,
+      "step": 1640
+    },
+    {
+      "entropy": 2.20017766058445,
+      "epoch": 1.1914394076214556,
+      "grad_norm": 0.011082557030022144,
+      "learning_rate": 1.78135217652892e-05,
+      "loss": 0.02469281405210495,
+      "mean_token_accuracy": 0.9926190480589867,
+      "num_tokens": 5559519.0,
+      "step": 1650
+    },
+    {
+      "entropy": 2.181740128993988,
+      "epoch": 1.1986635362109446,
+      "grad_norm": 0.13058708608150482,
+      "learning_rate": 1.7784247224330364e-05,
+      "loss": 0.002104618772864342,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5590509.0,
+      "step": 1660
+    },
+    {
+      "entropy": 2.1344178140163423,
+      "epoch": 1.2058876648004335,
+      "grad_norm": 3.456376075744629,
+      "learning_rate": 1.7754802402886878e-05,
+      "loss": 0.005177551507949829,
+      "mean_token_accuracy": 0.9957516342401505,
+      "num_tokens": 5624522.0,
+      "step": 1670
+    },
+    {
+      "entropy": 2.199016976356506,
+      "epoch": 1.2131117933899223,
+      "grad_norm": 0.011777219362556934,
+      "learning_rate": 1.77251879450645e-05,
+      "loss": 0.0028427716344594954,
+      "mean_token_accuracy": 0.9994565218687057,
+      "num_tokens": 5655676.0,
+      "step": 1680
+    },
+    {
+      "entropy": 2.232581838965416,
+      "epoch": 1.2203359219794112,
+      "grad_norm": 7.441551208496094,
+      "learning_rate": 1.7695404498679783e-05,
+      "loss": 0.016997954249382018,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 5690286.0,
+      "step": 1690
+    },
+    {
+      "entropy": 2.197023203969002,
+      "epoch": 1.2275600505689002,
+      "grad_norm": 0.01528174802660942,
+      "learning_rate": 1.7665452715245918e-05,
+      "loss": 0.020658569037914278,
+      "mean_token_accuracy": 0.9934895828366279,
+      "num_tokens": 5720068.0,
+      "step": 1700
+    },
+    {
+      "entropy": 2.1230575770139692,
+      "epoch": 1.234784179158389,
+      "grad_norm": 0.012365175411105156,
+      "learning_rate": 1.7635333249958458e-05,
+      "loss": 0.006468945741653442,
+      "mean_token_accuracy": 0.998863635957241,
+      "num_tokens": 5752141.0,
+      "step": 1710
+    },
+    {
+      "entropy": 2.1280272156000137,
+      "epoch": 1.2420083077478778,
+      "grad_norm": 0.07447945326566696,
+      "learning_rate": 1.7605046761681e-05,
+      "loss": 0.009069929271936417,
+      "mean_token_accuracy": 0.998504464328289,
+      "num_tokens": 5784292.0,
+      "step": 1720
+    },
+    {
+      "entropy": 2.049823743104935,
+      "epoch": 1.2492324363373668,
+      "grad_norm": 0.010746085084974766,
+      "learning_rate": 1.757459391293078e-05,
+      "loss": 0.00027041586581617595,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5822036.0,
+      "step": 1730
+    },
+    {
+      "entropy": 2.197540044784546,
+      "epoch": 1.2564565649268558,
+      "grad_norm": 0.2788766026496887,
+      "learning_rate": 1.7543975369864158e-05,
+      "loss": 0.012624423205852508,
+      "mean_token_accuracy": 0.9980769231915474,
+      "num_tokens": 5857543.0,
+      "step": 1740
+    },
+    {
+      "entropy": 2.1441560611128807,
+      "epoch": 1.2636806935163447,
+      "grad_norm": 0.07280537486076355,
+      "learning_rate": 1.7513191802262068e-05,
+      "loss": 0.0011107440106570721,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5890577.0,
+      "step": 1750
+    },
+    {
+      "entropy": 2.144422373175621,
+      "epoch": 1.2709048221058334,
+      "grad_norm": 0.0037534120492637157,
+      "learning_rate": 1.7482243883515352e-05,
+      "loss": 0.00637623593211174,
+      "mean_token_accuracy": 0.9977481618523598,
+      "num_tokens": 5926746.0,
+      "step": 1760
+    },
+    {
+      "entropy": 2.105022558569908,
+      "epoch": 1.2781289506953224,
+      "grad_norm": 0.6015522480010986,
+      "learning_rate": 1.7451132290610035e-05,
+      "loss": 0.0020936239510774613,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 5958942.0,
+      "step": 1770
+    },
+    {
+      "entropy": 2.161861351132393,
+      "epoch": 1.2853530792848114,
+      "grad_norm": 0.014276999980211258,
+      "learning_rate": 1.7419857704112527e-05,
+      "loss": 0.0010045192204415799,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 5989151.0,
+      "step": 1780
+    },
+    {
+      "entropy": 1.981706216931343,
+      "epoch": 1.2925772078743,
+      "grad_norm": 5.441474437713623,
+      "learning_rate": 1.738842080815471e-05,
+      "loss": 0.0060867641121149065,
+      "mean_token_accuracy": 0.9974999994039535,
+      "num_tokens": 6026197.0,
+      "step": 1790
+    },
+    {
+      "entropy": 2.1758767783641817,
+      "epoch": 1.299801336463789,
+      "grad_norm": 2.714914083480835,
+      "learning_rate": 1.735682229041899e-05,
+      "loss": 0.012385857105255128,
+      "mean_token_accuracy": 0.9976098895072937,
+      "num_tokens": 6061662.0,
+      "step": 1800
+    },
+    {
+      "entropy": 2.1634930104017256,
+      "epoch": 1.307025465053278,
+      "grad_norm": 0.019172759726643562,
+      "learning_rate": 1.7325062842123257e-05,
+      "loss": 0.0014975195750594138,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6094422.0,
+      "step": 1810
+    },
+    {
+      "entropy": 2.1602396339178087,
+      "epoch": 1.3142495936427667,
+      "grad_norm": 0.007157791871577501,
+      "learning_rate": 1.7293143158005753e-05,
+      "loss": 0.002255440875887871,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6124176.0,
+      "step": 1820
+    },
+    {
+      "entropy": 2.1500022053718566,
+      "epoch": 1.3214737222322557,
+      "grad_norm": 0.03131004422903061,
+      "learning_rate": 1.7261063936309886e-05,
+      "loss": 0.0033995889127254484,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 6158848.0,
+      "step": 1830
+    },
+    {
+      "entropy": 2.1492036104202272,
+      "epoch": 1.3286978508217446,
+      "grad_norm": 0.061348363757133484,
+      "learning_rate": 1.722882587876894e-05,
+      "loss": 0.007053826749324798,
+      "mean_token_accuracy": 0.9971022740006447,
+      "num_tokens": 6191610.0,
+      "step": 1840
+    },
+    {
+      "entropy": 2.182556915283203,
+      "epoch": 1.3359219794112336,
+      "grad_norm": 0.00448696780949831,
+      "learning_rate": 1.7196429690590745e-05,
+      "loss": 0.0005504220258444548,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6220927.0,
+      "step": 1850
+    },
+    {
+      "entropy": 2.105365613102913,
+      "epoch": 1.3431461080007225,
+      "grad_norm": 0.04105228930711746,
+      "learning_rate": 1.7163876080442236e-05,
+      "loss": 0.0006919845938682557,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6253925.0,
+      "step": 1860
+    },
+    {
+      "entropy": 2.1027141720056535,
+      "epoch": 1.3503702365902113,
+      "grad_norm": 0.003532874397933483,
+      "learning_rate": 1.713116576043396e-05,
+      "loss": 0.00023068697191774845,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6286725.0,
+      "step": 1870
+    },
+    {
+      "entropy": 2.059778553247452,
+      "epoch": 1.3575943651797002,
+      "grad_norm": 0.004922214429825544,
+      "learning_rate": 1.709829944610448e-05,
+      "loss": 0.00023504530545324088,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6326607.0,
+      "step": 1880
+    },
+    {
+      "entropy": 2.213117519021034,
+      "epoch": 1.3648184937691892,
+      "grad_norm": 0.005468189250677824,
+      "learning_rate": 1.706527785640475e-05,
+      "loss": 0.007529944181442261,
+      "mean_token_accuracy": 0.9987499997019768,
+      "num_tokens": 6363050.0,
+      "step": 1890
+    },
+    {
+      "entropy": 2.1393338203430177,
+      "epoch": 1.372042622358678,
+      "grad_norm": 0.013368732295930386,
+      "learning_rate": 1.703210171368237e-05,
+      "loss": 0.00037029776722192765,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6394373.0,
+      "step": 1900
+    },
+    {
+      "entropy": 2.0972392201423644,
+      "epoch": 1.3792667509481669,
+      "grad_norm": 0.0500776469707489,
+      "learning_rate": 1.699877174366579e-05,
+      "loss": 0.0019076000899076462,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6430285.0,
+      "step": 1910
+    },
+    {
+      "entropy": 2.116664645075798,
+      "epoch": 1.3864908795376558,
+      "grad_norm": 2.305840492248535,
+      "learning_rate": 1.696528867544843e-05,
+      "loss": 0.026842620968818665,
+      "mean_token_accuracy": 0.9895500689744949,
+      "num_tokens": 6464191.0,
+      "step": 1920
+    },
+    {
+      "entropy": 2.1112887859344482,
+      "epoch": 1.3937150081271446,
+      "grad_norm": 0.027304597198963165,
+      "learning_rate": 1.6931653241472736e-05,
+      "loss": 0.0007224025204777718,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6503510.0,
+      "step": 1930
+    },
+    {
+      "entropy": 2.1380940586328507,
+      "epoch": 1.4009391367166335,
+      "grad_norm": 0.012063352391123772,
+      "learning_rate": 1.689786617751416e-05,
+      "loss": 0.0008827324956655503,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6536930.0,
+      "step": 1940
+    },
+    {
+      "entropy": 2.0396142452955246,
+      "epoch": 1.4081632653061225,
+      "grad_norm": 0.001350369886495173,
+      "learning_rate": 1.6863928222665053e-05,
+      "loss": 0.004736289381980896,
+      "mean_token_accuracy": 0.999074074625969,
+      "num_tokens": 6572556.0,
+      "step": 1950
+    },
+    {
+      "entropy": 2.072943928837776,
+      "epoch": 1.4153873938956114,
+      "grad_norm": 0.5194405913352966,
+      "learning_rate": 1.6829840119318518e-05,
+      "loss": 0.0009896421805024147,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6604208.0,
+      "step": 1960
+    },
+    {
+      "entropy": 2.130077823996544,
+      "epoch": 1.4226115224851004,
+      "grad_norm": 2.9718973636627197,
+      "learning_rate": 1.6795602613152152e-05,
+      "loss": 0.025895681977272034,
+      "mean_token_accuracy": 0.9964082792401314,
+      "num_tokens": 6635729.0,
+      "step": 1970
+    },
+    {
+      "entropy": 2.0558677911758423,
+      "epoch": 1.429835651074589,
+      "grad_norm": 0.01987330988049507,
+      "learning_rate": 1.6761216453111732e-05,
+      "loss": 0.005741150304675102,
+      "mean_token_accuracy": 0.9987557873129844,
+      "num_tokens": 6671807.0,
+      "step": 1980
+    },
+    {
+      "entropy": 2.1100833773612977,
+      "epoch": 1.437059779664078,
+      "grad_norm": 0.0030209566466510296,
+      "learning_rate": 1.6726682391394854e-05,
+      "loss": 0.0012206203304231168,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6701449.0,
+      "step": 1990
+    },
+    {
+      "entropy": 2.119494986534119,
+      "epoch": 1.444283908253567,
+      "grad_norm": 0.0030340475495904684,
+      "learning_rate": 1.6692001183434453e-05,
+      "loss": 0.007272050529718399,
+      "mean_token_accuracy": 0.9970016330480576,
+      "num_tokens": 6736102.0,
+      "step": 2000
+    },
+    {
+      "epoch": 1.444283908253567,
+      "eval_entropy": 1.9451654732165042,
+      "eval_loss": 1.0858334302902222,
+      "eval_mean_token_accuracy": 0.8618772350633157,
+      "eval_num_tokens": 6736102.0,
+      "eval_runtime": 1006.5211,
+      "eval_samples_per_second": 7.601,
+      "eval_steps_per_second": 0.951,
+      "step": 2000
+    },
+    {
+      "entropy": 2.159361606836319,
+      "epoch": 1.4515080368430557,
+      "grad_norm": 0.003348611295223236,
+      "learning_rate": 1.66571735878823e-05,
+      "loss": 0.008503157645463943,
+      "mean_token_accuracy": 0.9962714686989784,
+      "num_tokens": 6765417.0,
+      "step": 2010
+    },
+    {
+      "entropy": 2.082682377099991,
+      "epoch": 1.4587321654325447,
+      "grad_norm": 0.45656055212020874,
+      "learning_rate": 1.6622200366592385e-05,
+      "loss": 0.002443484589457512,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 6802913.0,
+      "step": 2020
+    },
+    {
+      "entropy": 2.0617480158805845,
+      "epoch": 1.4659562940220336,
+      "grad_norm": 0.006900084670633078,
+      "learning_rate": 1.658708228460427e-05,
+      "loss": 0.0298701673746109,
+      "mean_token_accuracy": 0.997139498591423,
+      "num_tokens": 6835849.0,
+      "step": 2030
+    },
+    {
+      "entropy": 2.1594283044338227,
+      "epoch": 1.4731804226115224,
+      "grad_norm": 0.29801133275032043,
+      "learning_rate": 1.6551820110126346e-05,
+      "loss": 0.0018545888364315034,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 6866237.0,
+      "step": 2040
+    },
+    {
+      "entropy": 2.069524332880974,
+      "epoch": 1.4804045512010113,
+      "grad_norm": 0.017278041690587997,
+      "learning_rate": 1.6516414614519023e-05,
+      "loss": 0.003127451241016388,
+      "mean_token_accuracy": 0.9973416283726693,
+      "num_tokens": 6905968.0,
+      "step": 2050
+    },
+    {
+      "entropy": 2.027516800165176,
+      "epoch": 1.4876286797905003,
+      "grad_norm": 6.259517669677734,
+      "learning_rate": 1.6480866572277875e-05,
+      "loss": 0.003948203101754188,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 6937825.0,
+      "step": 2060
+    },
+    {
+      "entropy": 2.1773772150278092,
+      "epoch": 1.4948528083799892,
+      "grad_norm": 0.426023006439209,
+      "learning_rate": 1.6445176761016664e-05,
+      "loss": 0.009390848875045776,
+      "mean_token_accuracy": 0.9991666659712791,
+      "num_tokens": 6965652.0,
+      "step": 2070
+    },
+    {
+      "entropy": 2.128205436468124,
+      "epoch": 1.5020769369694782,
+      "grad_norm": 0.014904591254889965,
+      "learning_rate": 1.6409345961450366e-05,
+      "loss": 0.00557580292224884,
+      "mean_token_accuracy": 0.9975241541862487,
+      "num_tokens": 6999159.0,
+      "step": 2080
+    },
+    {
+      "entropy": 2.0940597355365753,
+      "epoch": 1.509301065558967,
+      "grad_norm": 0.025805262848734856,
+      "learning_rate": 1.6373374957378078e-05,
+      "loss": 0.0005395350977778435,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 7032483.0,
+      "step": 2090
+    },
+    {
+      "entropy": 2.1171302914619448,
+      "epoch": 1.5165251941484559,
+      "grad_norm": 3.1691181659698486,
+      "learning_rate": 1.6337264535665864e-05,
+      "loss": 0.006728671491146088,
+      "mean_token_accuracy": 0.9943749994039536,
+      "num_tokens": 7065864.0,
+      "step": 2100
+    },
+    {
+      "entropy": 2.122132509946823,
+      "epoch": 1.5237493227379448,
+      "grad_norm": 3.207775115966797,
+      "learning_rate": 1.630101548622956e-05,
+      "loss": 0.042887577414512636,
+      "mean_token_accuracy": 0.994207875430584,
+      "num_tokens": 7099331.0,
+      "step": 2110
+    },
+    {
+      "entropy": 2.1320498794317246,
+      "epoch": 1.5309734513274336,
+      "grad_norm": 0.05691314488649368,
+      "learning_rate": 1.6264628602017473e-05,
+      "loss": 0.04059517979621887,
+      "mean_token_accuracy": 0.9966666668653488,
+      "num_tokens": 7134548.0,
+      "step": 2120
+    },
+    {
+      "entropy": 2.0526298761367796,
+      "epoch": 1.5381975799169225,
+      "grad_norm": 0.032850734889507294,
+      "learning_rate": 1.6228104678993055e-05,
+      "loss": 0.012917523086071015,
+      "mean_token_accuracy": 0.9980769231915474,
+      "num_tokens": 7167221.0,
+      "step": 2130
+    },
+    {
+      "entropy": 2.0888993412256243,
+      "epoch": 1.5454217085064115,
+      "grad_norm": 0.03867359459400177,
+      "learning_rate": 1.6191444516117486e-05,
+      "loss": 0.01570902019739151,
+      "mean_token_accuracy": 0.9949736505746841,
+      "num_tokens": 7201449.0,
+      "step": 2140
+    },
+    {
+      "entropy": 2.079554834961891,
+      "epoch": 1.5526458370959002,
+      "grad_norm": 0.08862973004579544,
+      "learning_rate": 1.6154648915332183e-05,
+      "loss": 0.02136228382587433,
+      "mean_token_accuracy": 0.9924007937312126,
+      "num_tokens": 7234108.0,
+      "step": 2150
+    },
+    {
+      "entropy": 2.0425251334905625,
+      "epoch": 1.5598699656853892,
+      "grad_norm": 0.01745772175490856,
+      "learning_rate": 1.6117718681541282e-05,
+      "loss": 0.007442861795425415,
+      "mean_token_accuracy": 0.9947222217917442,
+      "num_tokens": 7267952.0,
+      "step": 2160
+    },
+    {
+      "entropy": 2.0624156802892686,
+      "epoch": 1.5670940942748781,
+      "grad_norm": 0.012653755955398083,
+      "learning_rate": 1.6080654622594005e-05,
+      "loss": 0.0031592480838298796,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 7302191.0,
+      "step": 2170
+    },
+    {
+      "entropy": 2.0713628053665163,
+      "epoch": 1.5743182228643668,
+      "grad_norm": 0.17270442843437195,
+      "learning_rate": 1.6043457549267008e-05,
+      "loss": 0.000261211208999157,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 7333553.0,
+      "step": 2180
+    },
+    {
+      "entropy": 2.0779812574386596,
+      "epoch": 1.581542351453856,
+      "grad_norm": 0.11458373814821243,
+      "learning_rate": 1.600612827524664e-05,
+      "loss": 0.01028701737523079,
+      "mean_token_accuracy": 0.9985294118523598,
+      "num_tokens": 7364794.0,
+      "step": 2190
+    },
+    {
+      "entropy": 2.0878961592912675,
+      "epoch": 1.5887664800433448,
+      "grad_norm": 1.063446044921875,
+      "learning_rate": 1.596866761711113e-05,
+      "loss": 0.0034152865409851074,
+      "mean_token_accuracy": 0.9991935476660728,
+      "num_tokens": 7395934.0,
+      "step": 2200
+    },
+    {
+      "entropy": 2.160181927680969,
+      "epoch": 1.5959906086328337,
+      "grad_norm": 2.0979878902435303,
+      "learning_rate": 1.593107639431275e-05,
+      "loss": 0.016500005125999452,
+      "mean_token_accuracy": 0.9973897039890289,
+      "num_tokens": 7425221.0,
+      "step": 2210
+    },
+    {
+      "entropy": 2.0582812547683718,
+      "epoch": 1.6032147372223227,
+      "grad_norm": 0.02503189444541931,
+      "learning_rate": 1.5893355429159867e-05,
+      "loss": 0.006939806044101715,
+      "mean_token_accuracy": 0.9991666659712791,
+      "num_tokens": 7463793.0,
+      "step": 2220
+    },
+    {
+      "entropy": 2.062930929660797,
+      "epoch": 1.6104388658118114,
+      "grad_norm": 3.7741119861602783,
+      "learning_rate": 1.5855505546798963e-05,
+      "loss": 0.03627827763557434,
+      "mean_token_accuracy": 0.9965909093618393,
+      "num_tokens": 7500363.0,
+      "step": 2230
+    },
+    {
+      "entropy": 2.1142319977283477,
+      "epoch": 1.6176629944013003,
+      "grad_norm": 0.14156651496887207,
+      "learning_rate": 1.581752757519659e-05,
+      "loss": 0.0020683687180280685,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 7532751.0,
+      "step": 2240
+    },
+    {
+      "entropy": 2.047053983807564,
+      "epoch": 1.6248871229907893,
+      "grad_norm": 0.05737530440092087,
+      "learning_rate": 1.577942234512125e-05,
+      "loss": 0.0005965373944491148,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 7571820.0,
+      "step": 2250
+    },
+    {
+      "entropy": 2.068295753002167,
+      "epoch": 1.632111251580278,
+      "grad_norm": 0.0038673433009535074,
+      "learning_rate": 1.5741190690125224e-05,
+      "loss": 0.03620594441890716,
+      "mean_token_accuracy": 0.9943652525544167,
+      "num_tokens": 7605275.0,
+      "step": 2260
+    },
+    {
+      "entropy": 2.135337767004967,
+      "epoch": 1.639335380169767,
+      "grad_norm": 0.007885460741817951,
+      "learning_rate": 1.5702833446526342e-05,
+      "loss": 0.0054336290806531904,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 7635492.0,
+      "step": 2270
+    },
+    {
+      "entropy": 2.1049996852874755,
+      "epoch": 1.646559508759256,
+      "grad_norm": 0.12335117161273956,
+      "learning_rate": 1.5664351453389687e-05,
+      "loss": 0.00200213510543108,
+      "mean_token_accuracy": 0.995833332836628,
+      "num_tokens": 7671880.0,
+      "step": 2280
+    },
+    {
+      "entropy": 2.091116154193878,
+      "epoch": 1.6537836373487447,
+      "grad_norm": 0.0021040288265794516,
+      "learning_rate": 1.5625745552509236e-05,
+      "loss": 0.002798403427004814,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 7705881.0,
+      "step": 2290
+    },
+    {
+      "entropy": 2.1217050731182097,
+      "epoch": 1.6610077659382338,
+      "grad_norm": 0.014709223993122578,
+      "learning_rate": 1.558701658838945e-05,
+      "loss": 0.005910974368453026,
+      "mean_token_accuracy": 0.9959374994039536,
+      "num_tokens": 7739861.0,
+      "step": 2300
+    },
+    {
+      "entropy": 2.0269686609506605,
+      "epoch": 1.6682318945277226,
+      "grad_norm": 0.015317005105316639,
+      "learning_rate": 1.5548165408226806e-05,
+      "loss": 0.014388790726661682,
+      "mean_token_accuracy": 0.995833332836628,
+      "num_tokens": 7774542.0,
+      "step": 2310
+    },
+    {
+      "entropy": 2.013978272676468,
+      "epoch": 1.6754560231172115,
+      "grad_norm": 0.0025651876349002123,
+      "learning_rate": 1.5509192861891245e-05,
+      "loss": 0.011645805835723878,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 7811748.0,
+      "step": 2320
+    },
+    {
+      "entropy": 2.104548341035843,
+      "epoch": 1.6826801517067005,
+      "grad_norm": 0.2633267641067505,
+      "learning_rate": 1.5470099801907603e-05,
+      "loss": 0.01602161079645157,
+      "mean_token_accuracy": 0.9939585506916047,
+      "num_tokens": 7846954.0,
+      "step": 2330
+    },
+    {
+      "entropy": 2.1113144189119337,
+      "epoch": 1.6899042802961892,
+      "grad_norm": 0.010862859897315502,
+      "learning_rate": 1.543088708343696e-05,
+      "loss": 0.0021166058257222177,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 7880033.0,
+      "step": 2340
+    },
+    {
+      "entropy": 2.1317659109830855,
+      "epoch": 1.6971284088856782,
+      "grad_norm": 0.0034101735800504684,
+      "learning_rate": 1.5391555564257907e-05,
+      "loss": 0.0007758287712931633,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 7914051.0,
+      "step": 2350
+    },
+    {
+      "entropy": 2.08573197722435,
+      "epoch": 1.7043525374751671,
+      "grad_norm": 0.00433329539373517,
+      "learning_rate": 1.5352106104747816e-05,
+      "loss": 0.005741733685135841,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 7950879.0,
+      "step": 2360
+    },
+    {
+      "entropy": 2.145374095439911,
+      "epoch": 1.7115766660646559,
+      "grad_norm": 1.4243534803390503,
+      "learning_rate": 1.531253956786401e-05,
+      "loss": 0.001701824553310871,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 7983672.0,
+      "step": 2370
+    },
+    {
+      "entropy": 2.154380649328232,
+      "epoch": 1.7188007946541448,
+      "grad_norm": 0.0023556272499263287,
+      "learning_rate": 1.5272856819124862e-05,
+      "loss": 0.008817800879478454,
+      "mean_token_accuracy": 0.9947222217917442,
+      "num_tokens": 8012964.0,
+      "step": 2380
+    },
+    {
+      "entropy": 2.031127080321312,
+      "epoch": 1.7260249232436338,
+      "grad_norm": 0.004914262797683477,
+      "learning_rate": 1.5233058726590896e-05,
+      "loss": 0.0030698470771312715,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 8049896.0,
+      "step": 2390
+    },
+    {
+      "entropy": 2.0777418464422226,
+      "epoch": 1.7332490518331225,
+      "grad_norm": 0.0020385284442454576,
+      "learning_rate": 1.5193146160845783e-05,
+      "loss": 0.00504988357424736,
+      "mean_token_accuracy": 0.9974358975887299,
+      "num_tokens": 8080833.0,
+      "step": 2400
+    },
+    {
+      "entropy": 2.0629676073789596,
+      "epoch": 1.7404731804226117,
+      "grad_norm": 0.004871215671300888,
+      "learning_rate": 1.5153119994977286e-05,
+      "loss": 0.006580570340156555,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 8115944.0,
+      "step": 2410
+    },
+    {
+      "entropy": 2.146797752380371,
+      "epoch": 1.7476973090121004,
+      "grad_norm": 0.005745346192270517,
+      "learning_rate": 1.5112981104558176e-05,
+      "loss": 0.011366159468889237,
+      "mean_token_accuracy": 0.9981481477618217,
+      "num_tokens": 8146140.0,
+      "step": 2420
+    },
+    {
+      "entropy": 2.124235710501671,
+      "epoch": 1.7549214376015894,
+      "grad_norm": 0.05242975428700447,
+      "learning_rate": 1.5072730367627079e-05,
+      "loss": 0.011133132874965668,
+      "mean_token_accuracy": 0.9965277776122093,
+      "num_tokens": 8178690.0,
+      "step": 2430
+    },
+    {
+      "entropy": 2.0248345702886583,
+      "epoch": 1.7621455661910783,
+      "grad_norm": 0.032173994928598404,
+      "learning_rate": 1.5032368664669262e-05,
+      "loss": 0.019655060768127442,
+      "mean_token_accuracy": 0.9976576581597328,
+      "num_tokens": 8215889.0,
+      "step": 2440
+    },
+    {
+      "entropy": 2.077134358882904,
+      "epoch": 1.769369694780567,
+      "grad_norm": 0.029686057940125465,
+      "learning_rate": 1.4991896878597371e-05,
+      "loss": 0.00024895952083170414,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8252786.0,
+      "step": 2450
+    },
+    {
+      "entropy": 2.0941765367984773,
+      "epoch": 1.776593823370056,
+      "grad_norm": 0.0018677343614399433,
+      "learning_rate": 1.4951315894732128e-05,
+      "loss": 0.0016262145712971688,
+      "mean_token_accuracy": 0.9982142850756646,
+      "num_tokens": 8285439.0,
+      "step": 2460
+    },
+    {
+      "entropy": 2.0770112931728364,
+      "epoch": 1.783817951959545,
+      "grad_norm": 0.030317634344100952,
+      "learning_rate": 1.4910626600782953e-05,
+      "loss": 0.021285098791122437,
+      "mean_token_accuracy": 0.9956127449870109,
+      "num_tokens": 8320214.0,
+      "step": 2470
+    },
+    {
+      "entropy": 2.1631770461797712,
+      "epoch": 1.7910420805490337,
+      "grad_norm": 0.0071309832856059074,
+      "learning_rate": 1.4869829886828555e-05,
+      "loss": 0.0012782313860952854,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8351086.0,
+      "step": 2480
+    },
+    {
+      "entropy": 2.08559827208519,
+      "epoch": 1.7982662091385226,
+      "grad_norm": 0.009854475036263466,
+      "learning_rate": 1.4828926645297447e-05,
+      "loss": 0.0008186939172446728,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8385744.0,
+      "step": 2490
+    },
+    {
+      "entropy": 2.1026084780693055,
+      "epoch": 1.8054903377280116,
+      "grad_norm": 0.07382749766111374,
+      "learning_rate": 1.4787917770948444e-05,
+      "loss": 0.0014898085966706276,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8421729.0,
+      "step": 2500
+    },
+    {
+      "epoch": 1.8054903377280116,
+      "eval_entropy": 1.9514784755627066,
+      "eval_loss": 0.9975342750549316,
+      "eval_mean_token_accuracy": 0.8620556939233682,
+      "eval_num_tokens": 8421729.0,
+      "eval_runtime": 1006.2961,
+      "eval_samples_per_second": 7.603,
+      "eval_steps_per_second": 0.951,
+      "step": 2500
+    },
+    {
+      "entropy": 2.1760506868362426,
+      "epoch": 1.8127144663175003,
+      "grad_norm": 0.006193229462951422,
+      "learning_rate": 1.4746804160851076e-05,
+      "loss": 0.011964889615774155,
+      "mean_token_accuracy": 0.9964714959263802,
+      "num_tokens": 8455688.0,
+      "step": 2510
+    },
+    {
+      "entropy": 2.0681952387094498,
+      "epoch": 1.8199385949069895,
+      "grad_norm": 0.011737292632460594,
+      "learning_rate": 1.4705586714365967e-05,
+      "loss": 0.0016200032085180282,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8489613.0,
+      "step": 2520
+    },
+    {
+      "entropy": 2.083423137664795,
+      "epoch": 1.8271627234964782,
+      "grad_norm": 0.0025027799420058727,
+      "learning_rate": 1.4664266333125166e-05,
+      "loss": 0.0005176069214940071,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8520801.0,
+      "step": 2530
+    },
+    {
+      "entropy": 2.1015573650598527,
+      "epoch": 1.834386852085967,
+      "grad_norm": 0.005616495851427317,
+      "learning_rate": 1.4622843921012418e-05,
+      "loss": 0.0005612615030258894,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8552679.0,
+      "step": 2540
+    },
+    {
+      "entropy": 2.0993423074483872,
+      "epoch": 1.8416109806754561,
+      "grad_norm": 0.013693253509700298,
+      "learning_rate": 1.4581320384143397e-05,
+      "loss": 0.029371032118797304,
+      "mean_token_accuracy": 0.9958041965961456,
+      "num_tokens": 8585601.0,
+      "step": 2550
+    },
+    {
+      "entropy": 2.134834015369415,
+      "epoch": 1.8488351092649449,
+      "grad_norm": 0.004303743597120047,
+      "learning_rate": 1.4539696630845882e-05,
+      "loss": 0.001381959579885006,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8614091.0,
+      "step": 2560
+    },
+    {
+      "entropy": 2.08456112742424,
+      "epoch": 1.8560592378544338,
+      "grad_norm": 5.2712721824646,
+      "learning_rate": 1.4497973571639883e-05,
+      "loss": 0.003928203508257866,
+      "mean_token_accuracy": 0.997916667163372,
+      "num_tokens": 8647175.0,
+      "step": 2570
+    },
+    {
+      "entropy": 2.019808714091778,
+      "epoch": 1.8632833664439228,
+      "grad_norm": 0.046853918582201004,
+      "learning_rate": 1.4456152119217732e-05,
+      "loss": 0.0012758148834109307,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8689485.0,
+      "step": 2580
+    },
+    {
+      "entropy": 2.084204414486885,
+      "epoch": 1.8705074950334115,
+      "grad_norm": 0.06356920301914215,
+      "learning_rate": 1.441423318842411e-05,
+      "loss": 0.009153851121664048,
+      "mean_token_accuracy": 0.9972222223877907,
+      "num_tokens": 8727433.0,
+      "step": 2590
+    },
+    {
+      "entropy": 2.1179255068302156,
+      "epoch": 1.8777316236229005,
+      "grad_norm": 1.0097019672393799,
+      "learning_rate": 1.4372217696236045e-05,
+      "loss": 0.030431166291236877,
+      "mean_token_accuracy": 0.9957663685083389,
+      "num_tokens": 8757201.0,
+      "step": 2600
+    },
+    {
+      "entropy": 2.073572924733162,
+      "epoch": 1.8849557522123894,
+      "grad_norm": 0.18690069019794464,
+      "learning_rate": 1.4330106561742842e-05,
+      "loss": 0.0184125691652298,
+      "mean_token_accuracy": 0.9870535716414451,
+      "num_tokens": 8788479.0,
+      "step": 2610
+    },
+    {
+      "entropy": 2.077366715669632,
+      "epoch": 1.8921798808018782,
+      "grad_norm": 0.005627399776130915,
+      "learning_rate": 1.4287900706125982e-05,
+      "loss": 0.0002856994513422251,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8823250.0,
+      "step": 2620
+    },
+    {
+      "entropy": 2.1045290410518644,
+      "epoch": 1.8994040093913673,
+      "grad_norm": 0.019973022863268852,
+      "learning_rate": 1.4245601052638967e-05,
+      "loss": 0.0014879602007567883,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8859128.0,
+      "step": 2630
+    },
+    {
+      "entropy": 2.088423106074333,
+      "epoch": 1.906628137980856,
+      "grad_norm": 0.00223484612070024,
+      "learning_rate": 1.4203208526587137e-05,
+      "loss": 0.002584621496498585,
+      "mean_token_accuracy": 0.9993055552244187,
+      "num_tokens": 8898952.0,
+      "step": 2640
+    },
+    {
+      "entropy": 2.06300730407238,
+      "epoch": 1.9138522665703448,
+      "grad_norm": 0.2910439074039459,
+      "learning_rate": 1.416072405530742e-05,
+      "loss": 0.0012499621137976647,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 8932753.0,
+      "step": 2650
+    },
+    {
+      "entropy": 2.0773635655641556,
+      "epoch": 1.921076395159834,
+      "grad_norm": 0.0037493736017495394,
+      "learning_rate": 1.4118148568148038e-05,
+      "loss": 0.010332270711660384,
+      "mean_token_accuracy": 0.9993243247270585,
+      "num_tokens": 8961577.0,
+      "step": 2660
+    },
+    {
+      "entropy": 2.1508059173822405,
+      "epoch": 1.9283005237493227,
+      "grad_norm": 0.009645289741456509,
+      "learning_rate": 1.4075482996448196e-05,
+      "loss": 0.003071390464901924,
+      "mean_token_accuracy": 0.997916667163372,
+      "num_tokens": 8990569.0,
+      "step": 2670
+    },
+    {
+      "entropy": 2.1294310480356216,
+      "epoch": 1.9355246523388117,
+      "grad_norm": 5.6811723709106445,
+      "learning_rate": 1.4032728273517693e-05,
+      "loss": 0.011133387684822083,
+      "mean_token_accuracy": 0.9980962634086609,
+      "num_tokens": 9021975.0,
+      "step": 2680
+    },
+    {
+      "entropy": 2.03683922290802,
+      "epoch": 1.9427487809283006,
+      "grad_norm": 0.08936524391174316,
+      "learning_rate": 1.398988533461651e-05,
+      "loss": 0.05962691307067871,
+      "mean_token_accuracy": 0.9925133690237999,
+      "num_tokens": 9055502.0,
+      "step": 2690
+    },
+    {
+      "entropy": 2.0664506271481513,
+      "epoch": 1.9499729095177893,
+      "grad_norm": 0.09585876017808914,
+      "learning_rate": 1.3946955116934366e-05,
+      "loss": 0.0051217213273048404,
+      "mean_token_accuracy": 0.996875,
+      "num_tokens": 9086035.0,
+      "step": 2700
+    },
+    {
+      "entropy": 2.126824939250946,
+      "epoch": 1.9571970381072783,
+      "grad_norm": 0.08990640938282013,
+      "learning_rate": 1.3903938559570187e-05,
+      "loss": 0.0011291544884443282,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9116140.0,
+      "step": 2710
+    },
+    {
+      "entropy": 2.0418795704841615,
+      "epoch": 1.9644211666967673,
+      "grad_norm": 0.10705948621034622,
+      "learning_rate": 1.3860836603511595e-05,
+      "loss": 0.0012339632026851176,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9151933.0,
+      "step": 2720
+    },
+    {
+      "entropy": 2.0204968392848968,
+      "epoch": 1.971645295286256,
+      "grad_norm": 0.0022592502646148205,
+      "learning_rate": 1.3817650191614296e-05,
+      "loss": 0.009728147834539413,
+      "mean_token_accuracy": 0.9977865785360336,
+      "num_tokens": 9186649.0,
+      "step": 2730
+    },
+    {
+      "entropy": 2.01169753074646,
+      "epoch": 1.978869423875745,
+      "grad_norm": 0.0048128338530659676,
+      "learning_rate": 1.3774380268581483e-05,
+      "loss": 0.0025031518191099165,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9220783.0,
+      "step": 2740
+    },
+    {
+      "entropy": 1.9962417513132096,
+      "epoch": 1.986093552465234,
+      "grad_norm": 0.008226979523897171,
+      "learning_rate": 1.3731027780943142e-05,
+      "loss": 0.005799595266580582,
+      "mean_token_accuracy": 0.9980769231915474,
+      "num_tokens": 9260611.0,
+      "step": 2750
+    },
+    {
+      "entropy": 2.0178030014038084,
+      "epoch": 1.9933176810547226,
+      "grad_norm": 0.05092993378639221,
+      "learning_rate": 1.368759367703537e-05,
+      "loss": 0.009776970744132996,
+      "mean_token_accuracy": 0.9982142850756646,
+      "num_tokens": 9297499.0,
+      "step": 2760
+    },
+    {
+      "entropy": 2.0301727668659106,
+      "epoch": 2.0,
+      "grad_norm": 0.0022351753432303667,
+      "learning_rate": 1.3644078906979617e-05,
+      "loss": 0.007662636786699295,
+      "mean_token_accuracy": 0.9968399163838979,
+      "num_tokens": 9331880.0,
+      "step": 2770
+    },
+    {
+      "entropy": 1.9933802992105485,
+      "epoch": 2.0072241285894887,
+      "grad_norm": 2.35412335395813,
+      "learning_rate": 1.3600484422661908e-05,
+      "loss": 0.0009718227200210094,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9371782.0,
+      "step": 2780
+    },
+    {
+      "entropy": 1.9624912858009338,
+      "epoch": 2.014448257178978,
+      "grad_norm": 1.6708158254623413,
+      "learning_rate": 1.3556811177712012e-05,
+      "loss": 0.005323154479265213,
+      "mean_token_accuracy": 0.9985294118523598,
+      "num_tokens": 9410776.0,
+      "step": 2790
+    },
+    {
+      "entropy": 2.0888388842344283,
+      "epoch": 2.0216723857684666,
+      "grad_norm": 0.06923649460077286,
+      "learning_rate": 1.3513060127482595e-05,
+      "loss": 0.001955091394484043,
+      "mean_token_accuracy": 0.9985294118523598,
+      "num_tokens": 9442042.0,
+      "step": 2800
+    },
+    {
+      "entropy": 2.0619905948638917,
+      "epoch": 2.0288965143579554,
+      "grad_norm": 0.004627038724720478,
+      "learning_rate": 1.3469232229028318e-05,
+      "loss": 0.002406453527510166,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 9479069.0,
+      "step": 2810
+    },
+    {
+      "entropy": 2.10693601667881,
+      "epoch": 2.0361206429474445,
+      "grad_norm": 0.16550160944461823,
+      "learning_rate": 1.3425328441084877e-05,
+      "loss": 0.0013710076920688153,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9511919.0,
+      "step": 2820
+    },
+    {
+      "entropy": 2.0103971630334856,
+      "epoch": 2.0433447715369333,
+      "grad_norm": 0.38094449043273926,
+      "learning_rate": 1.3381349724048074e-05,
+      "loss": 0.00045806909911334516,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9545770.0,
+      "step": 2830
+    },
+    {
+      "entropy": 2.0798803091049196,
+      "epoch": 2.0505689001264225,
+      "grad_norm": 3.134385108947754,
+      "learning_rate": 1.3337297039952778e-05,
+      "loss": 0.0038271814584732054,
+      "mean_token_accuracy": 0.9950000002980233,
+      "num_tokens": 9578858.0,
+      "step": 2840
+    },
+    {
+      "entropy": 2.0519185990095137,
+      "epoch": 2.057793028715911,
+      "grad_norm": 0.0022963378578424454,
+      "learning_rate": 1.329317135245188e-05,
+      "loss": 0.000910833477973938,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9611522.0,
+      "step": 2850
+    },
+    {
+      "entropy": 2.1064377218484878,
+      "epoch": 2.0650171573054,
+      "grad_norm": 0.06155133619904518,
+      "learning_rate": 1.3248973626795239e-05,
+      "loss": 0.0014734589494764805,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9647641.0,
+      "step": 2860
+    },
+    {
+      "entropy": 2.083224031329155,
+      "epoch": 2.072241285894889,
+      "grad_norm": 0.005543859675526619,
+      "learning_rate": 1.320470482980853e-05,
+      "loss": 0.004852643609046936,
+      "mean_token_accuracy": 0.999074074625969,
+      "num_tokens": 9680790.0,
+      "step": 2870
+    },
+    {
+      "entropy": 2.0704041182994843,
+      "epoch": 2.079465414484378,
+      "grad_norm": 0.005386595148593187,
+      "learning_rate": 1.3160365929872127e-05,
+      "loss": 0.0003251228015869856,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9715564.0,
+      "step": 2880
+    },
+    {
+      "entropy": 2.0982192888855935,
+      "epoch": 2.0866895430738666,
+      "grad_norm": 0.0017691103275865316,
+      "learning_rate": 1.31159578968999e-05,
+      "loss": 0.004996376112103462,
+      "mean_token_accuracy": 0.9980000004172325,
+      "num_tokens": 9747350.0,
+      "step": 2890
+    },
+    {
+      "entropy": 1.9677607417106628,
+      "epoch": 2.0939136716633557,
+      "grad_norm": 0.0013132777530699968,
+      "learning_rate": 1.3071481702318013e-05,
+      "loss": 0.0020328672602772714,
+      "mean_token_accuracy": 0.9994047611951828,
+      "num_tokens": 9777462.0,
+      "step": 2900
+    },
+    {
+      "entropy": 2.0211497992277145,
+      "epoch": 2.1011378002528445,
+      "grad_norm": 0.04309982806444168,
+      "learning_rate": 1.3026938319043653e-05,
+      "loss": 0.010742837935686112,
+      "mean_token_accuracy": 0.9973127111792565,
+      "num_tokens": 9810045.0,
+      "step": 2910
+    },
+    {
+      "entropy": 2.0541851848363875,
+      "epoch": 2.108361928842333,
+      "grad_norm": 0.8860397934913635,
+      "learning_rate": 1.2982328721463771e-05,
+      "loss": 0.0005173175595700741,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9841633.0,
+      "step": 2920
+    },
+    {
+      "entropy": 2.0442396640777587,
+      "epoch": 2.1155860574318224,
+      "grad_norm": 0.01298177894204855,
+      "learning_rate": 1.2937653885413753e-05,
+      "loss": 0.0019559381529688836,
+      "mean_token_accuracy": 0.9991379305720329,
+      "num_tokens": 9875065.0,
+      "step": 2930
+    },
+    {
+      "entropy": 2.0949876844882964,
+      "epoch": 2.122810186021311,
+      "grad_norm": 0.08153260499238968,
+      "learning_rate": 1.2892914788156077e-05,
+      "loss": 0.011897164583206176,
+      "mean_token_accuracy": 0.9970238104462623,
+      "num_tokens": 9908190.0,
+      "step": 2940
+    },
+    {
+      "entropy": 2.068952742218971,
+      "epoch": 2.1300343146108,
+      "grad_norm": 0.0025051895063370466,
+      "learning_rate": 1.2848112408358935e-05,
+      "loss": 0.0026764221489429473,
+      "mean_token_accuracy": 0.995833332836628,
+      "num_tokens": 9940211.0,
+      "step": 2950
+    },
+    {
+      "entropy": 2.0292742609977723,
+      "epoch": 2.137258443200289,
+      "grad_norm": 0.0027262514922767878,
+      "learning_rate": 1.2803247726074828e-05,
+      "loss": 0.0006647636648267508,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 9976844.0,
+      "step": 2960
+    },
+    {
+      "entropy": 2.104836627840996,
+      "epoch": 2.1444825717897777,
+      "grad_norm": 0.08476550132036209,
+      "learning_rate": 1.2758321722719121e-05,
+      "loss": 0.0007632578257471323,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10009418.0,
+      "step": 2970
+    },
+    {
+      "entropy": 2.0324369966983795,
+      "epoch": 2.151706700379267,
+      "grad_norm": 0.016819585114717484,
+      "learning_rate": 1.271333538104858e-05,
+      "loss": 0.00019808937795460225,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10044535.0,
+      "step": 2980
+    },
+    {
+      "entropy": 2.0965832740068437,
+      "epoch": 2.1589308289687557,
+      "grad_norm": 0.0015572819393128157,
+      "learning_rate": 1.2668289685139872e-05,
+      "loss": 0.00022766462061554193,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10075829.0,
+      "step": 2990
+    },
+    {
+      "entropy": 2.0006484925746917,
+      "epoch": 2.1661549575582444,
+      "grad_norm": 0.0032824883237481117,
+      "learning_rate": 1.2623185620368039e-05,
+      "loss": 0.00015145333018153905,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10112270.0,
+      "step": 3000
+    },
+    {
+      "epoch": 2.1661549575582444,
+      "eval_entropy": 1.9313233175247815,
+      "eval_loss": 1.0283830165863037,
+      "eval_mean_token_accuracy": 0.8682382702204625,
+      "eval_num_tokens": 10112270.0,
+      "eval_runtime": 1006.1426,
+      "eval_samples_per_second": 7.604,
+      "eval_steps_per_second": 0.951,
+      "step": 3000
+    },
+    {
+      "entropy": 2.071103125810623,
+      "epoch": 2.1733790861477336,
+      "grad_norm": 0.0022774692624807358,
+      "learning_rate": 1.257802417338494e-05,
+      "loss": 0.00011760392226278782,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10146218.0,
+      "step": 3010
+    },
+    {
+      "entropy": 2.0262765318155287,
+      "epoch": 2.1806032147372223,
+      "grad_norm": 0.06287238746881485,
+      "learning_rate": 1.2532806332097674e-05,
+      "loss": 0.004662457481026649,
+      "mean_token_accuracy": 0.9982142850756646,
+      "num_tokens": 10177685.0,
+      "step": 3020
+    },
+    {
+      "entropy": 2.037755089998245,
+      "epoch": 2.187827343326711,
+      "grad_norm": 0.002215301152318716,
+      "learning_rate": 1.2487533085646963e-05,
+      "loss": 0.0005286297295242548,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10209037.0,
+      "step": 3030
+    },
+    {
+      "entropy": 2.0680573701858522,
+      "epoch": 2.1950514719162,
+      "grad_norm": 0.025813493877649307,
+      "learning_rate": 1.244220542438552e-05,
+      "loss": 0.011255993694067,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 10240793.0,
+      "step": 3040
+    },
+    {
+      "entropy": 2.039670634269714,
+      "epoch": 2.202275600505689,
+      "grad_norm": 1.8143988847732544,
+      "learning_rate": 1.2396824339856382e-05,
+      "loss": 0.0010964684188365936,
+      "mean_token_accuracy": 0.9990384608507157,
+      "num_tokens": 10279065.0,
+      "step": 3050
+    },
+    {
+      "entropy": 2.144818735122681,
+      "epoch": 2.209499729095178,
+      "grad_norm": 0.006999373901635408,
+      "learning_rate": 1.2351390824771216e-05,
+      "loss": 0.0004216683562844992,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10306803.0,
+      "step": 3060
+    },
+    {
+      "entropy": 1.9957463964819908,
+      "epoch": 2.216723857684667,
+      "grad_norm": 0.011525344103574753,
+      "learning_rate": 1.2305905872988619e-05,
+      "loss": 0.004228273034095764,
+      "mean_token_accuracy": 0.9991071432828903,
+      "num_tokens": 10343140.0,
+      "step": 3070
+    },
+    {
+      "entropy": 2.0431026369333267,
+      "epoch": 2.2239479862741556,
+      "grad_norm": 0.008179805241525173,
+      "learning_rate": 1.226037047949235e-05,
+      "loss": 0.002482306584715843,
+      "mean_token_accuracy": 0.9982142850756646,
+      "num_tokens": 10380486.0,
+      "step": 3080
+    },
+    {
+      "entropy": 2.155712330341339,
+      "epoch": 2.2311721148636448,
+      "grad_norm": 0.0033373201731592417,
+      "learning_rate": 1.2214785640369593e-05,
+      "loss": 0.00017562623834237455,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10413499.0,
+      "step": 3090
+    },
+    {
+      "entropy": 2.0313654780387878,
+      "epoch": 2.2383962434531335,
+      "grad_norm": 0.005848964676260948,
+      "learning_rate": 1.2169152352789155e-05,
+      "loss": 0.0036901481449604034,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 10446135.0,
+      "step": 3100
+    },
+    {
+      "entropy": 2.0065710693597794,
+      "epoch": 2.245620372042622,
+      "grad_norm": 0.008071194402873516,
+      "learning_rate": 1.212347161497965e-05,
+      "loss": 0.0026075785979628565,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 10477985.0,
+      "step": 3110
+    },
+    {
+      "entropy": 1.9910441845655442,
+      "epoch": 2.2528445006321114,
+      "grad_norm": 0.06832147389650345,
+      "learning_rate": 1.2077744426207665e-05,
+      "loss": 0.0002928957110270858,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10514883.0,
+      "step": 3120
+    },
+    {
+      "entropy": 2.0544280052185058,
+      "epoch": 2.2600686292216,
+      "grad_norm": 0.0016180315287783742,
+      "learning_rate": 1.2031971786755908e-05,
+      "loss": 0.0034284554421901703,
+      "mean_token_accuracy": 0.997916667163372,
+      "num_tokens": 10547163.0,
+      "step": 3130
+    },
+    {
+      "entropy": 2.0648707509040833,
+      "epoch": 2.267292757811089,
+      "grad_norm": 0.002448343439027667,
+      "learning_rate": 1.1986154697901313e-05,
+      "loss": 0.0013522141613066196,
+      "mean_token_accuracy": 0.99921875,
+      "num_tokens": 10583986.0,
+      "step": 3140
+    },
+    {
+      "entropy": 2.1439515709877015,
+      "epoch": 2.274516886400578,
+      "grad_norm": 0.006475028581917286,
+      "learning_rate": 1.1940294161893153e-05,
+      "loss": 0.0001555228722281754,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10617384.0,
+      "step": 3150
+    },
+    {
+      "entropy": 2.0301067173480987,
+      "epoch": 2.2817410149900668,
+      "grad_norm": 0.002055932767689228,
+      "learning_rate": 1.1894391181931104e-05,
+      "loss": 0.00035485513508319854,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10649547.0,
+      "step": 3160
+    },
+    {
+      "entropy": 2.082839173078537,
+      "epoch": 2.2889651435795555,
+      "grad_norm": 0.0013027176028117537,
+      "learning_rate": 1.1848446762143304e-05,
+      "loss": 0.00797640010714531,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 10682210.0,
+      "step": 3170
+    },
+    {
+      "entropy": 2.0413184225559236,
+      "epoch": 2.2961892721690447,
+      "grad_norm": 0.012554303742945194,
+      "learning_rate": 1.180246190756439e-05,
+      "loss": 0.00010339131113141775,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10714579.0,
+      "step": 3180
+    },
+    {
+      "entropy": 2.1068718791007996,
+      "epoch": 2.3034134007585334,
+      "grad_norm": 0.001185836736112833,
+      "learning_rate": 1.1756437624113506e-05,
+      "loss": 0.00328139141201973,
+      "mean_token_accuracy": 0.9972222223877907,
+      "num_tokens": 10747908.0,
+      "step": 3190
+    },
+    {
+      "entropy": 2.050973677635193,
+      "epoch": 2.3106375293480226,
+      "grad_norm": 0.0016986754490062594,
+      "learning_rate": 1.1710374918572308e-05,
+      "loss": 0.005435990169644356,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 10781175.0,
+      "step": 3200
+    },
+    {
+      "entropy": 2.0618736773729323,
+      "epoch": 2.3178616579375113,
+      "grad_norm": 0.010860864073038101,
+      "learning_rate": 1.1664274798562927e-05,
+      "loss": 0.011393396556377411,
+      "mean_token_accuracy": 0.9954861104488373,
+      "num_tokens": 10816087.0,
+      "step": 3210
+    },
+    {
+      "entropy": 2.156007653474808,
+      "epoch": 2.325085786527,
+      "grad_norm": 0.003099059220403433,
+      "learning_rate": 1.161813827252595e-05,
+      "loss": 0.0001721467007882893,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10848858.0,
+      "step": 3220
+    },
+    {
+      "entropy": 2.0750046342611315,
+      "epoch": 2.332309915116489,
+      "grad_norm": 0.0023590456694364548,
+      "learning_rate": 1.1571966349698334e-05,
+      "loss": 0.0007454160135239362,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 10882504.0,
+      "step": 3230
+    },
+    {
+      "entropy": 2.155749836564064,
+      "epoch": 2.339534043705978,
+      "grad_norm": 0.007863319478929043,
+      "learning_rate": 1.1525760040091345e-05,
+      "loss": 0.006022463366389274,
+      "mean_token_accuracy": 0.999285714328289,
+      "num_tokens": 10912504.0,
+      "step": 3240
+    },
+    {
+      "entropy": 2.0390058249235152,
+      "epoch": 2.3467581722954667,
+      "grad_norm": 0.0026729318778961897,
+      "learning_rate": 1.1479520354468462e-05,
+      "loss": 0.011591699719429017,
+      "mean_token_accuracy": 0.9992424249649048,
+      "num_tokens": 10948673.0,
+      "step": 3250
+    },
+    {
+      "entropy": 2.007386389374733,
+      "epoch": 2.353982300884956,
+      "grad_norm": 11.629199981689453,
+      "learning_rate": 1.1433248304323265e-05,
+      "loss": 0.008497416973114014,
+      "mean_token_accuracy": 0.9974519237875938,
+      "num_tokens": 10981856.0,
+      "step": 3260
+    },
+    {
+      "entropy": 2.0721147865056992,
+      "epoch": 2.3612064294744446,
+      "grad_norm": 0.0032983580604195595,
+      "learning_rate": 1.1386944901857302e-05,
+      "loss": 0.0001707537448965013,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11016701.0,
+      "step": 3270
+    },
+    {
+      "entropy": 2.0985424160957336,
+      "epoch": 2.3684305580639338,
+      "grad_norm": 0.028967533260583878,
+      "learning_rate": 1.134061115995796e-05,
+      "loss": 0.001917528174817562,
+      "mean_token_accuracy": 0.9991666659712791,
+      "num_tokens": 11053496.0,
+      "step": 3280
+    },
+    {
+      "entropy": 2.109357753396034,
+      "epoch": 2.3756546866534225,
+      "grad_norm": 0.025716641917824745,
+      "learning_rate": 1.1294248092176296e-05,
+      "loss": 0.0006847483571618795,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11085915.0,
+      "step": 3290
+    },
+    {
+      "entropy": 2.0609450221061705,
+      "epoch": 2.3828788152429112,
+      "grad_norm": 0.012144936248660088,
+      "learning_rate": 1.1247856712704878e-05,
+      "loss": 0.00018151594558730723,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11119602.0,
+      "step": 3300
+    },
+    {
+      "entropy": 2.1143811851739884,
+      "epoch": 2.3901029438324004,
+      "grad_norm": 0.007466345094144344,
+      "learning_rate": 1.1201438036355589e-05,
+      "loss": 0.003900020942091942,
+      "mean_token_accuracy": 0.9982142850756646,
+      "num_tokens": 11152938.0,
+      "step": 3310
+    },
+    {
+      "entropy": 2.0662292540073395,
+      "epoch": 2.397327072421889,
+      "grad_norm": 0.0035816922318190336,
+      "learning_rate": 1.115499307853743e-05,
+      "loss": 0.0002706871833652258,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11188975.0,
+      "step": 3320
+    },
+    {
+      "entropy": 2.0358931571245193,
+      "epoch": 2.404551201011378,
+      "grad_norm": 0.0022427074145525694,
+      "learning_rate": 1.1108522855234307e-05,
+      "loss": 0.00035625442396849396,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11224655.0,
+      "step": 3330
+    },
+    {
+      "entropy": 2.1082998305559157,
+      "epoch": 2.411775329600867,
+      "grad_norm": 0.005758251529186964,
+      "learning_rate": 1.1062028382982817e-05,
+      "loss": 0.00044987592846155164,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11255165.0,
+      "step": 3340
+    },
+    {
+      "entropy": 2.061891108751297,
+      "epoch": 2.418999458190356,
+      "grad_norm": 0.008551395498216152,
+      "learning_rate": 1.1015510678849994e-05,
+      "loss": 0.004435742273926735,
+      "mean_token_accuracy": 0.9986111104488373,
+      "num_tokens": 11291478.0,
+      "step": 3350
+    },
+    {
+      "entropy": 2.0787433266639708,
+      "epoch": 2.4262235867798445,
+      "grad_norm": 0.01050745602697134,
+      "learning_rate": 1.0968970760411078e-05,
+      "loss": 0.00033552371896803377,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11324107.0,
+      "step": 3360
+    },
+    {
+      "entropy": 2.0784036934375765,
+      "epoch": 2.4334477153693337,
+      "grad_norm": 0.003264373168349266,
+      "learning_rate": 1.092240964572724e-05,
+      "loss": 0.00031879942398518326,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11355874.0,
+      "step": 3370
+    },
+    {
+      "entropy": 2.032439711689949,
+      "epoch": 2.4406718439588224,
+      "grad_norm": 0.013324176892638206,
+      "learning_rate": 1.0875828353323321e-05,
+      "loss": 0.0003661923110485077,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11391783.0,
+      "step": 3380
+    },
+    {
+      "entropy": 2.0780230134725572,
+      "epoch": 2.447895972548311,
+      "grad_norm": 0.0024669126141816378,
+      "learning_rate": 1.0829227902165553e-05,
+      "loss": 0.0009197730571031571,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11425236.0,
+      "step": 3390
+    },
+    {
+      "entropy": 2.031269261240959,
+      "epoch": 2.4551201011378003,
+      "grad_norm": 0.004380874801427126,
+      "learning_rate": 1.0782609311639263e-05,
+      "loss": 0.000289306603372097,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11458086.0,
+      "step": 3400
+    },
+    {
+      "entropy": 2.0453762233257295,
+      "epoch": 2.462344229727289,
+      "grad_norm": 0.24470053613185883,
+      "learning_rate": 1.073597360152658e-05,
+      "loss": 0.0004137265030294657,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11493591.0,
+      "step": 3410
+    },
+    {
+      "entropy": 2.063965144753456,
+      "epoch": 2.469568358316778,
+      "grad_norm": 0.008266739547252655,
+      "learning_rate": 1.0689321791984117e-05,
+      "loss": 0.00012792153283953668,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11526615.0,
+      "step": 3420
+    },
+    {
+      "entropy": 2.083864006400108,
+      "epoch": 2.476792486906267,
+      "grad_norm": 0.029528291895985603,
+      "learning_rate": 1.0642654903520671e-05,
+      "loss": 0.004928890988230706,
+      "mean_token_accuracy": 0.999479167163372,
+      "num_tokens": 11560627.0,
+      "step": 3430
+    },
+    {
+      "entropy": 2.0178954720497133,
+      "epoch": 2.4840166154957557,
+      "grad_norm": 0.004495162982493639,
+      "learning_rate": 1.0595973956974886e-05,
+      "loss": 0.0003031428437680006,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11598679.0,
+      "step": 3440
+    },
+    {
+      "entropy": 2.068479546904564,
+      "epoch": 2.491240744085245,
+      "grad_norm": 0.002551234792917967,
+      "learning_rate": 1.0549279973492923e-05,
+      "loss": 0.000176075950730592,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11628974.0,
+      "step": 3450
+    },
+    {
+      "entropy": 2.120567148923874,
+      "epoch": 2.4984648726747336,
+      "grad_norm": 0.008339786902070045,
+      "learning_rate": 1.0502573974506137e-05,
+      "loss": 0.00029307235963642596,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11658626.0,
+      "step": 3460
+    },
+    {
+      "entropy": 2.097331014275551,
+      "epoch": 2.5056890012642223,
+      "grad_norm": 0.02314414270222187,
+      "learning_rate": 1.045585698170871e-05,
+      "loss": 0.005216201767325401,
+      "mean_token_accuracy": 0.9977272734045982,
+      "num_tokens": 11697099.0,
+      "step": 3470
+    },
+    {
+      "entropy": 2.0958898663520813,
+      "epoch": 2.5129131298537115,
+      "grad_norm": 0.056762658059597015,
+      "learning_rate": 1.040913001703532e-05,
+      "loss": 0.0002604601439088583,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11727638.0,
+      "step": 3480
+    },
+    {
+      "entropy": 2.055971249938011,
+      "epoch": 2.5201372584432002,
+      "grad_norm": 0.005196870304644108,
+      "learning_rate": 1.0362394102638782e-05,
+      "loss": 0.00024930061772465706,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11763716.0,
+      "step": 3490
+    },
+    {
+      "entropy": 2.0658002614974977,
+      "epoch": 2.5273613870326894,
+      "grad_norm": 0.002534567378461361,
+      "learning_rate": 1.0315650260867683e-05,
+      "loss": 0.00019605269189924002,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11799642.0,
+      "step": 3500
+    },
+    {
+      "epoch": 2.5273613870326894,
+      "eval_entropy": 1.9229904009630687,
+      "eval_loss": 1.0553957223892212,
+      "eval_mean_token_accuracy": 0.8690185485215023,
+      "eval_num_tokens": 11799642.0,
+      "eval_runtime": 1006.5888,
+      "eval_samples_per_second": 7.601,
+      "eval_steps_per_second": 0.951,
+      "step": 3500
+    },
+    {
+      "entropy": 2.1393352538347243,
+      "epoch": 2.534585515622178,
+      "grad_norm": 0.011761508882045746,
+      "learning_rate": 1.0268899514244019e-05,
+      "loss": 0.0009708752855658531,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11829891.0,
+      "step": 3510
+    },
+    {
+      "entropy": 2.1366262555122377,
+      "epoch": 2.541809644211667,
+      "grad_norm": 0.0419895239174366,
+      "learning_rate": 1.0222142885440836e-05,
+      "loss": 0.0002197347581386566,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11860479.0,
+      "step": 3520
+    },
+    {
+      "entropy": 2.0491745442152025,
+      "epoch": 2.549033772801156,
+      "grad_norm": 0.007780018728226423,
+      "learning_rate": 1.0175381397259845e-05,
+      "loss": 0.0007411201484501361,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11895990.0,
+      "step": 3530
+    },
+    {
+      "entropy": 2.0275103002786636,
+      "epoch": 2.556257901390645,
+      "grad_norm": 0.02360602281987667,
+      "learning_rate": 1.0128616072609064e-05,
+      "loss": 0.005948469042778015,
+      "mean_token_accuracy": 0.997916667163372,
+      "num_tokens": 11931817.0,
+      "step": 3540
+    },
+    {
+      "entropy": 2.0498583540320396,
+      "epoch": 2.5634820299801335,
+      "grad_norm": 0.0018535489216446877,
+      "learning_rate": 1.0081847934480428e-05,
+      "loss": 0.00025608809664845467,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 11965008.0,
+      "step": 3550
+    },
+    {
+      "entropy": 2.044562515616417,
+      "epoch": 2.5707061585696227,
+      "grad_norm": 0.002041611587628722,
+      "learning_rate": 1.0035078005927421e-05,
+      "loss": 8.137651020660997e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12001243.0,
+      "step": 3560
+    },
+    {
+      "entropy": 2.1808446675539015,
+      "epoch": 2.5779302871591114,
+      "grad_norm": 0.0018940613372251391,
+      "learning_rate": 9.98830731004269e-06,
+      "loss": 0.0004888596944510937,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12030657.0,
+      "step": 3570
+    },
+    {
+      "entropy": 2.029064553976059,
+      "epoch": 2.5851544157486,
+      "grad_norm": 0.3856402039527893,
+      "learning_rate": 9.941536869935666e-06,
+      "loss": 0.0004893512930721044,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12070942.0,
+      "step": 3580
+    },
+    {
+      "entropy": 2.067329040169716,
+      "epoch": 2.5923785443380893,
+      "grad_norm": 0.011225709691643715,
+      "learning_rate": 9.894767708710187e-06,
+      "loss": 0.0002632188843563199,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12103626.0,
+      "step": 3590
+    },
+    {
+      "entropy": 2.114652419090271,
+      "epoch": 2.599602672927578,
+      "grad_norm": 0.0012341891415417194,
+      "learning_rate": 9.848000849442115e-06,
+      "loss": 0.00045462558045983315,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12140496.0,
+      "step": 3600
+    },
+    {
+      "entropy": 1.9965310275554657,
+      "epoch": 2.606826801517067,
+      "grad_norm": 0.0031538368202745914,
+      "learning_rate": 9.801237315156958e-06,
+      "loss": 0.00018033559899777174,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12174214.0,
+      "step": 3610
+    },
+    {
+      "entropy": 2.0614873960614206,
+      "epoch": 2.614050930106556,
+      "grad_norm": 0.0019219790119677782,
+      "learning_rate": 9.754478128807488e-06,
+      "loss": 0.00020585947204381226,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12204992.0,
+      "step": 3620
+    },
+    {
+      "entropy": 2.1520297437906266,
+      "epoch": 2.6212750586960447,
+      "grad_norm": 0.003325063269585371,
+      "learning_rate": 9.707724313251367e-06,
+      "loss": 0.0006235354579985142,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12236300.0,
+      "step": 3630
+    },
+    {
+      "entropy": 1.9406953275203704,
+      "epoch": 2.6284991872855334,
+      "grad_norm": 0.001424752757884562,
+      "learning_rate": 9.660976891228772e-06,
+      "loss": 0.0002312889089807868,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12280554.0,
+      "step": 3640
+    },
+    {
+      "entropy": 2.05175538957119,
+      "epoch": 2.6357233158750226,
+      "grad_norm": 0.0013753697276115417,
+      "learning_rate": 9.614236885340018e-06,
+      "loss": 0.00021411580964922906,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12315314.0,
+      "step": 3650
+    },
+    {
+      "entropy": 2.0753579437732697,
+      "epoch": 2.6429474444645114,
+      "grad_norm": 0.030707193538546562,
+      "learning_rate": 9.567505318023192e-06,
+      "loss": 0.00024836480151861905,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12350002.0,
+      "step": 3660
+    },
+    {
+      "entropy": 2.0509102582931518,
+      "epoch": 2.650171573054,
+      "grad_norm": 0.026452315971255302,
+      "learning_rate": 9.520783211531792e-06,
+      "loss": 0.0026783553883433344,
+      "mean_token_accuracy": 0.9989130437374115,
+      "num_tokens": 12387908.0,
+      "step": 3670
+    },
+    {
+      "entropy": 2.0522196829319,
+      "epoch": 2.6573957016434893,
+      "grad_norm": 0.0006438027485273778,
+      "learning_rate": 9.474071587912359e-06,
+      "loss": 5.928742466494441e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12420677.0,
+      "step": 3680
+    },
+    {
+      "entropy": 2.0196869641542436,
+      "epoch": 2.664619830232978,
+      "grad_norm": 0.0011145708849653602,
+      "learning_rate": 9.427371468982115e-06,
+      "loss": 0.0005806103348731994,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12457408.0,
+      "step": 3690
+    },
+    {
+      "entropy": 2.0822910487651827,
+      "epoch": 2.671843958822467,
+      "grad_norm": 0.09432120621204376,
+      "learning_rate": 9.380683876306623e-06,
+      "loss": 0.00019203609554097056,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12493796.0,
+      "step": 3700
+    },
+    {
+      "entropy": 2.031005597114563,
+      "epoch": 2.679068087411956,
+      "grad_norm": 0.3591781556606293,
+      "learning_rate": 9.334009831177433e-06,
+      "loss": 0.014735683798789978,
+      "mean_token_accuracy": 0.9961842104792595,
+      "num_tokens": 12527387.0,
+      "step": 3710
+    },
+    {
+      "entropy": 2.040946564078331,
+      "epoch": 2.686292216001445,
+      "grad_norm": 0.0015764066483825445,
+      "learning_rate": 9.287350354589742e-06,
+      "loss": 0.00015463890740647913,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12556284.0,
+      "step": 3720
+    },
+    {
+      "entropy": 2.0359088510274885,
+      "epoch": 2.693516344590934,
+      "grad_norm": 0.29461848735809326,
+      "learning_rate": 9.240706467220056e-06,
+      "loss": 0.00022188578732311725,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12589295.0,
+      "step": 3730
+    },
+    {
+      "entropy": 2.060906508564949,
+      "epoch": 2.7007404731804225,
+      "grad_norm": 0.39082056283950806,
+      "learning_rate": 9.194079189403875e-06,
+      "loss": 0.0038298744708299636,
+      "mean_token_accuracy": 0.9986842110753059,
+      "num_tokens": 12627227.0,
+      "step": 3740
+    },
+    {
+      "entropy": 1.9807851999998092,
+      "epoch": 2.7079646017699117,
+      "grad_norm": 0.006283764727413654,
+      "learning_rate": 9.147469541113358e-06,
+      "loss": 0.00011564189335331321,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12670866.0,
+      "step": 3750
+    },
+    {
+      "entropy": 2.022774323821068,
+      "epoch": 2.7151887303594004,
+      "grad_norm": 0.0015482399612665176,
+      "learning_rate": 9.10087854193502e-06,
+      "loss": 7.613286725245416e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12706262.0,
+      "step": 3760
+    },
+    {
+      "entropy": 2.0236416876316072,
+      "epoch": 2.722412858948889,
+      "grad_norm": 0.0006946782814338803,
+      "learning_rate": 9.054307211047427e-06,
+      "loss": 0.00010619325330480933,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12741193.0,
+      "step": 3770
+    },
+    {
+      "entropy": 2.1197534084320067,
+      "epoch": 2.7296369875383784,
+      "grad_norm": 0.0009746414725668728,
+      "learning_rate": 9.007756567198901e-06,
+      "loss": 9.164836956188083e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12772725.0,
+      "step": 3780
+    },
+    {
+      "entropy": 2.100919196009636,
+      "epoch": 2.736861116127867,
+      "grad_norm": 0.06009179726243019,
+      "learning_rate": 8.961227628685233e-06,
+      "loss": 0.0018002629280090333,
+      "mean_token_accuracy": 0.999479167163372,
+      "num_tokens": 12802118.0,
+      "step": 3790
+    },
+    {
+      "entropy": 2.085750639438629,
+      "epoch": 2.744085244717356,
+      "grad_norm": 0.012048608623445034,
+      "learning_rate": 8.914721413327413e-06,
+      "loss": 0.00016923286020755767,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12836979.0,
+      "step": 3800
+    },
+    {
+      "entropy": 2.058210352063179,
+      "epoch": 2.751309373306845,
+      "grad_norm": 0.0028724810108542442,
+      "learning_rate": 8.868238938449359e-06,
+      "loss": 0.0001348356483504176,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12870957.0,
+      "step": 3810
+    },
+    {
+      "entropy": 2.060334774851799,
+      "epoch": 2.7585335018963337,
+      "grad_norm": 0.008394302800297737,
+      "learning_rate": 8.821781220855664e-06,
+      "loss": 0.004397500678896904,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 12904043.0,
+      "step": 3820
+    },
+    {
+      "entropy": 2.094987612962723,
+      "epoch": 2.7657576304858225,
+      "grad_norm": 0.11255097389221191,
+      "learning_rate": 8.775349276809362e-06,
+      "loss": 0.000713779591023922,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12935070.0,
+      "step": 3830
+    },
+    {
+      "entropy": 2.0855062276124956,
+      "epoch": 2.7729817590753116,
+      "grad_norm": 0.0009015518589876592,
+      "learning_rate": 8.728944122009681e-06,
+      "loss": 0.005582309514284134,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 12967168.0,
+      "step": 3840
+    },
+    {
+      "entropy": 2.0803301841020585,
+      "epoch": 2.7802058876648004,
+      "grad_norm": 0.04961608350276947,
+      "learning_rate": 8.682566771569844e-06,
+      "loss": 0.00015667194966226817,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 12998789.0,
+      "step": 3850
+    },
+    {
+      "entropy": 2.0449639528989794,
+      "epoch": 2.787430016254289,
+      "grad_norm": 0.0016990930307656527,
+      "learning_rate": 8.636218239994845e-06,
+      "loss": 0.003403344750404358,
+      "mean_token_accuracy": 0.999285714328289,
+      "num_tokens": 13031593.0,
+      "step": 3860
+    },
+    {
+      "entropy": 2.077379512786865,
+      "epoch": 2.7946541448437783,
+      "grad_norm": 0.0010355949634686112,
+      "learning_rate": 8.58989954115927e-06,
+      "loss": 0.0002980584278702736,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13065699.0,
+      "step": 3870
+    },
+    {
+      "entropy": 2.063441056013107,
+      "epoch": 2.801878273433267,
+      "grad_norm": 0.03642053157091141,
+      "learning_rate": 8.543611688285111e-06,
+      "loss": 9.34006820898503e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13092554.0,
+      "step": 3880
+    },
+    {
+      "entropy": 2.0276421874761583,
+      "epoch": 2.8091024020227557,
+      "grad_norm": 0.014171849004924297,
+      "learning_rate": 8.497355693919605e-06,
+      "loss": 0.00023828567937016486,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13128152.0,
+      "step": 3890
+    },
+    {
+      "entropy": 2.05616457760334,
+      "epoch": 2.816326530612245,
+      "grad_norm": 0.011352301575243473,
+      "learning_rate": 8.451132569913085e-06,
+      "loss": 0.00010320036672055722,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13156902.0,
+      "step": 3900
+    },
+    {
+      "entropy": 2.0568599939346313,
+      "epoch": 2.8235506592017336,
+      "grad_norm": 0.0081400815397501,
+      "learning_rate": 8.404943327396842e-06,
+      "loss": 0.00039729168638587,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13193343.0,
+      "step": 3910
+    },
+    {
+      "entropy": 2.0809281915426254,
+      "epoch": 2.830774787791223,
+      "grad_norm": 0.0030671602580696344,
+      "learning_rate": 8.358788976761014e-06,
+      "loss": 0.000372506445273757,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13228389.0,
+      "step": 3920
+    },
+    {
+      "entropy": 2.0903642386198045,
+      "epoch": 2.8379989163807116,
+      "grad_norm": 0.002352460753172636,
+      "learning_rate": 8.312670527632473e-06,
+      "loss": 4.645454173441976e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13256457.0,
+      "step": 3930
+    },
+    {
+      "entropy": 2.0301788434386254,
+      "epoch": 2.8452230449702007,
+      "grad_norm": 0.0036281412467360497,
+      "learning_rate": 8.266588988852748e-06,
+      "loss": 0.0001243971986696124,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13289681.0,
+      "step": 3940
+    },
+    {
+      "entropy": 2.029602548480034,
+      "epoch": 2.8524471735596895,
+      "grad_norm": 0.002834299812093377,
+      "learning_rate": 8.220545368455956e-06,
+      "loss": 0.014585791528224945,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 13328416.0,
+      "step": 3950
+    },
+    {
+      "entropy": 2.063823339343071,
+      "epoch": 2.859671302149178,
+      "grad_norm": 0.003287531668320298,
+      "learning_rate": 8.17454067364674e-06,
+      "loss": 0.003944706171751022,
+      "mean_token_accuracy": 0.995833332836628,
+      "num_tokens": 13357159.0,
+      "step": 3960
+    },
+    {
+      "entropy": 2.078008511662483,
+      "epoch": 2.8668954307386674,
+      "grad_norm": 0.012140554375946522,
+      "learning_rate": 8.128575910778257e-06,
+      "loss": 0.0005927850026637316,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13391270.0,
+      "step": 3970
+    },
+    {
+      "entropy": 2.0515424638986586,
+      "epoch": 2.874119559328156,
+      "grad_norm": 0.005324928089976311,
+      "learning_rate": 8.082652085330145e-06,
+      "loss": 4.898167098872363e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13426130.0,
+      "step": 3980
+    },
+    {
+      "entropy": 2.086466285586357,
+      "epoch": 2.881343687917645,
+      "grad_norm": 0.0037718871608376503,
+      "learning_rate": 8.036770201886537e-06,
+      "loss": 9.53549169935286e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13457128.0,
+      "step": 3990
+    },
+    {
+      "entropy": 2.065127009153366,
+      "epoch": 2.888567816507134,
+      "grad_norm": 0.0024332841858267784,
+      "learning_rate": 7.990931264114087e-06,
+      "loss": 6.472773966379464e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13488834.0,
+      "step": 4000
+    },
+    {
+      "epoch": 2.888567816507134,
+      "eval_entropy": 1.8939840909455636,
+      "eval_loss": 1.0826828479766846,
+      "eval_mean_token_accuracy": 0.8724846162267862,
+      "eval_num_tokens": 13488834.0,
+      "eval_runtime": 1005.9483,
+      "eval_samples_per_second": 7.606,
+      "eval_steps_per_second": 0.951,
+      "step": 4000
+    },
+    {
+      "entropy": 2.081470862030983,
+      "epoch": 2.8957919450966227,
+      "grad_norm": 0.11557070910930634,
+      "learning_rate": 7.94513627474001e-06,
+      "loss": 0.0001834964146837592,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13517745.0,
+      "step": 4010
+    },
+    {
+      "entropy": 2.0741114974021913,
+      "epoch": 2.9030160736861115,
+      "grad_norm": 0.011433808133006096,
+      "learning_rate": 7.899386235530148e-06,
+      "loss": 7.996972417458891e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13545439.0,
+      "step": 4020
+    },
+    {
+      "entropy": 2.0428703770041468,
+      "epoch": 2.9102402022756007,
+      "grad_norm": 0.0010173572227358818,
+      "learning_rate": 7.853682147267059e-06,
+      "loss": 0.004161747172474861,
+      "mean_token_accuracy": 0.9989583328366279,
+      "num_tokens": 13582414.0,
+      "step": 4030
+    },
+    {
+      "entropy": 1.9939154922962188,
+      "epoch": 2.9174643308650894,
+      "grad_norm": 0.00418558344244957,
+      "learning_rate": 7.808025009728125e-06,
+      "loss": 0.0070745997130870816,
+      "mean_token_accuracy": 0.9985294118523598,
+      "num_tokens": 13612293.0,
+      "step": 4040
+    },
+    {
+      "entropy": 2.051701617240906,
+      "epoch": 2.924688459454578,
+      "grad_norm": 0.3029954731464386,
+      "learning_rate": 7.762415821663677e-06,
+      "loss": 0.0010162684135138988,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13645598.0,
+      "step": 4050
+    },
+    {
+      "entropy": 2.0056424707174303,
+      "epoch": 2.9319125880440673,
+      "grad_norm": 0.35396990180015564,
+      "learning_rate": 7.716855580775155e-06,
+      "loss": 0.0003054623492062092,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13677891.0,
+      "step": 4060
+    },
+    {
+      "entropy": 2.0558212995529175,
+      "epoch": 2.939136716633556,
+      "grad_norm": 0.014027765952050686,
+      "learning_rate": 7.671345283693274e-06,
+      "loss": 0.00015732048777863383,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13713298.0,
+      "step": 4070
+    },
+    {
+      "entropy": 2.057107946276665,
+      "epoch": 2.9463608452230448,
+      "grad_norm": 0.0469052754342556,
+      "learning_rate": 7.625885925956235e-06,
+      "loss": 0.00019727607723325492,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13741705.0,
+      "step": 4080
+    },
+    {
+      "entropy": 1.9607576489448548,
+      "epoch": 2.953584973812534,
+      "grad_norm": 0.00442711403593421,
+      "learning_rate": 7.580478501987932e-06,
+      "loss": 0.00901368036866188,
+      "mean_token_accuracy": 0.9989583328366279,
+      "num_tokens": 13773371.0,
+      "step": 4090
+    },
+    {
+      "entropy": 1.98673774600029,
+      "epoch": 2.9608091024020227,
+      "grad_norm": 0.015278922393918037,
+      "learning_rate": 7.535124005076215e-06,
+      "loss": 0.0002138084266334772,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13808274.0,
+      "step": 4100
+    },
+    {
+      "entropy": 2.096168154478073,
+      "epoch": 2.9680332309915114,
+      "grad_norm": 0.0005756252794526517,
+      "learning_rate": 7.489823427351146e-06,
+      "loss": 0.00018357214285060763,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13843522.0,
+      "step": 4110
+    },
+    {
+      "entropy": 2.060718294978142,
+      "epoch": 2.9752573595810006,
+      "grad_norm": 0.0011303877690806985,
+      "learning_rate": 7.4445777597633115e-06,
+      "loss": 0.009400687366724014,
+      "mean_token_accuracy": 0.9980592116713524,
+      "num_tokens": 13872279.0,
+      "step": 4120
+    },
+    {
+      "entropy": 2.013058941066265,
+      "epoch": 2.9824814881704893,
+      "grad_norm": 0.00043935581925325096,
+      "learning_rate": 7.3993879920621336e-06,
+      "loss": 6.445502513088285e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13909552.0,
+      "step": 4130
+    },
+    {
+      "entropy": 1.9740059912204742,
+      "epoch": 2.9897056167599785,
+      "grad_norm": 0.0009029169450514019,
+      "learning_rate": 7.354255112774227e-06,
+      "loss": 0.00041622747667133806,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13950502.0,
+      "step": 4140
+    },
+    {
+      "entropy": 2.062269702553749,
+      "epoch": 2.996929745349467,
+      "grad_norm": 0.002542090369388461,
+      "learning_rate": 7.309180109181769e-06,
+      "loss": 0.0012434286065399647,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 13985325.0,
+      "step": 4150
+    },
+    {
+      "entropy": 2.0545924515337557,
+      "epoch": 3.0036120642947446,
+      "grad_norm": 0.0019471081905066967,
+      "learning_rate": 7.264163967300908e-06,
+      "loss": 6.45760737825185e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14020377.0,
+      "step": 4160
+    },
+    {
+      "entropy": 2.053666239976883,
+      "epoch": 3.0108361928842333,
+      "grad_norm": 0.01275340560823679,
+      "learning_rate": 7.2192076718601914e-06,
+      "loss": 5.218477454036474e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14054964.0,
+      "step": 4170
+    },
+    {
+      "entropy": 2.05324681699276,
+      "epoch": 3.018060321473722,
+      "grad_norm": 0.003125807736068964,
+      "learning_rate": 7.174312206279022e-06,
+      "loss": 0.002049664966762066,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 14086930.0,
+      "step": 4180
+    },
+    {
+      "entropy": 2.0771075263619423,
+      "epoch": 3.0252844500632112,
+      "grad_norm": 0.0010819229064509273,
+      "learning_rate": 7.129478552646153e-06,
+      "loss": 0.0001039954018779099,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14124095.0,
+      "step": 4190
+    },
+    {
+      "entropy": 2.0247433304786684,
+      "epoch": 3.0325085786527,
+      "grad_norm": 0.0011280939215794206,
+      "learning_rate": 7.084707691698198e-06,
+      "loss": 7.054000161588192e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14159941.0,
+      "step": 4200
+    },
+    {
+      "entropy": 2.043915817141533,
+      "epoch": 3.039732707242189,
+      "grad_norm": 0.018443454056978226,
+      "learning_rate": 7.040000602798183e-06,
+      "loss": 0.0099873349070549,
+      "mean_token_accuracy": 0.9983333334326744,
+      "num_tokens": 14192414.0,
+      "step": 4210
+    },
+    {
+      "entropy": 2.0046940714120867,
+      "epoch": 3.046956835831678,
+      "grad_norm": 0.0009632164146751165,
+      "learning_rate": 6.995358263914115e-06,
+      "loss": 0.0003674726001918316,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14227241.0,
+      "step": 4220
+    },
+    {
+      "entropy": 2.033160814642906,
+      "epoch": 3.0541809644211666,
+      "grad_norm": 0.0031618638895452023,
+      "learning_rate": 6.950781651597598e-06,
+      "loss": 4.04816324589774e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14256661.0,
+      "step": 4230
+    },
+    {
+      "entropy": 2.0661414325237275,
+      "epoch": 3.0614050930106558,
+      "grad_norm": 0.0018059182912111282,
+      "learning_rate": 6.906271740962465e-06,
+      "loss": 0.004481592774391174,
+      "mean_token_accuracy": 0.9977272734045982,
+      "num_tokens": 14292544.0,
+      "step": 4240
+    },
+    {
+      "entropy": 2.0076374232769014,
+      "epoch": 3.0686292216001445,
+      "grad_norm": 0.0018631487619131804,
+      "learning_rate": 6.861829505663449e-06,
+      "loss": 5.830081645399332e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14324498.0,
+      "step": 4250
+    },
+    {
+      "entropy": 2.0799363285303114,
+      "epoch": 3.0758533501896332,
+      "grad_norm": 0.0013724598102271557,
+      "learning_rate": 6.817455917874885e-06,
+      "loss": 0.00023024685215204953,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14354881.0,
+      "step": 4260
+    },
+    {
+      "entropy": 2.1223824709653853,
+      "epoch": 3.0830774787791224,
+      "grad_norm": 0.0013472167775034904,
+      "learning_rate": 6.773151948269442e-06,
+      "loss": 0.0002774083754047751,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14386303.0,
+      "step": 4270
+    },
+    {
+      "entropy": 2.040089511871338,
+      "epoch": 3.090301607368611,
+      "grad_norm": 0.0028757641557604074,
+      "learning_rate": 6.728918565996891e-06,
+      "loss": 0.0002688025590032339,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14422045.0,
+      "step": 4280
+    },
+    {
+      "entropy": 2.03407336473465,
+      "epoch": 3.0975257359581,
+      "grad_norm": 0.0008419603109359741,
+      "learning_rate": 6.684756738662908e-06,
+      "loss": 0.00011189808137714862,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14455751.0,
+      "step": 4290
+    },
+    {
+      "entropy": 2.0355051040649412,
+      "epoch": 3.104749864547589,
+      "grad_norm": 0.0015740561066195369,
+      "learning_rate": 6.640667432307894e-06,
+      "loss": 5.130659556016326e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14493691.0,
+      "step": 4300
+    },
+    {
+      "entropy": 2.000814378261566,
+      "epoch": 3.111973993137078,
+      "grad_norm": 0.0007698965491726995,
+      "learning_rate": 6.596651611385865e-06,
+      "loss": 5.027313018217683e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14529860.0,
+      "step": 4310
+    },
+    {
+      "entropy": 2.077588292956352,
+      "epoch": 3.1191981217265665,
+      "grad_norm": 0.0005067693418823183,
+      "learning_rate": 6.552710238743336e-06,
+      "loss": 0.00013087192783132196,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14555834.0,
+      "step": 4320
+    },
+    {
+      "entropy": 2.0925796955823897,
+      "epoch": 3.1264222503160557,
+      "grad_norm": 0.006536214146763086,
+      "learning_rate": 6.508844275598263e-06,
+      "loss": 0.00019935504533350468,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14587367.0,
+      "step": 4330
+    },
+    {
+      "entropy": 2.0527830511331557,
+      "epoch": 3.1336463789055444,
+      "grad_norm": 0.04956198111176491,
+      "learning_rate": 6.465054681519025e-06,
+      "loss": 9.359649848192931e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14622396.0,
+      "step": 4340
+    },
+    {
+      "entropy": 1.9672885686159134,
+      "epoch": 3.1408705074950336,
+      "grad_norm": 0.0006884578033350408,
+      "learning_rate": 6.421342414403422e-06,
+      "loss": 0.00013261778512969614,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14656692.0,
+      "step": 4350
+    },
+    {
+      "entropy": 2.072387820482254,
+      "epoch": 3.1480946360845223,
+      "grad_norm": 0.018161380663514137,
+      "learning_rate": 6.377708430457728e-06,
+      "loss": 0.00018474491080269217,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14689145.0,
+      "step": 4360
+    },
+    {
+      "entropy": 2.024577698111534,
+      "epoch": 3.155318764674011,
+      "grad_norm": 0.0033226576633751392,
+      "learning_rate": 6.334153684175771e-06,
+      "loss": 0.0005758251994848251,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14722326.0,
+      "step": 4370
+    },
+    {
+      "entropy": 2.049726849794388,
+      "epoch": 3.1625428932635002,
+      "grad_norm": 0.001250626053661108,
+      "learning_rate": 6.290679128318054e-06,
+      "loss": 8.391140145249664e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14752330.0,
+      "step": 4380
+    },
+    {
+      "entropy": 2.0609421640634538,
+      "epoch": 3.169767021852989,
+      "grad_norm": 0.002631265204399824,
+      "learning_rate": 6.247285713890918e-06,
+      "loss": 0.00014461235841736197,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14789123.0,
+      "step": 4390
+    },
+    {
+      "entropy": 1.967462283372879,
+      "epoch": 3.1769911504424777,
+      "grad_norm": 0.017055487260222435,
+      "learning_rate": 6.203974390125724e-06,
+      "loss": 9.934637928381563e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14824386.0,
+      "step": 4400
+    },
+    {
+      "entropy": 1.9794509530067443,
+      "epoch": 3.184215279031967,
+      "grad_norm": 0.011362893506884575,
+      "learning_rate": 6.160746104458111e-06,
+      "loss": 5.548715707845986e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14861051.0,
+      "step": 4410
+    },
+    {
+      "entropy": 2.071438956260681,
+      "epoch": 3.1914394076214556,
+      "grad_norm": 0.0026533612981438637,
+      "learning_rate": 6.117601802507249e-06,
+      "loss": 0.00016078923363238573,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14892203.0,
+      "step": 4420
+    },
+    {
+      "entropy": 2.0377372980117796,
+      "epoch": 3.1986635362109443,
+      "grad_norm": 0.01846657320857048,
+      "learning_rate": 6.0745424280551725e-06,
+      "loss": 0.00072297896258533,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14924515.0,
+      "step": 4430
+    },
+    {
+      "entropy": 2.0567080438137055,
+      "epoch": 3.2058876648004335,
+      "grad_norm": 0.005206569097936153,
+      "learning_rate": 6.031568923026119e-06,
+      "loss": 0.00012273758184164763,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14957918.0,
+      "step": 4440
+    },
+    {
+      "entropy": 2.028385517001152,
+      "epoch": 3.2131117933899223,
+      "grad_norm": 0.01966731622815132,
+      "learning_rate": 5.988682227465934e-06,
+      "loss": 0.00015165915247052907,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 14990794.0,
+      "step": 4450
+    },
+    {
+      "entropy": 1.9939093351364137,
+      "epoch": 3.2203359219794114,
+      "grad_norm": 0.007970397360622883,
+      "learning_rate": 5.945883279521508e-06,
+      "loss": 0.00010207220911979675,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15032262.0,
+      "step": 4460
+    },
+    {
+      "entropy": 2.0098903119564056,
+      "epoch": 3.2275600505689,
+      "grad_norm": 0.006879137828946114,
+      "learning_rate": 5.903173015420244e-06,
+      "loss": 8.936674566939473e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15064056.0,
+      "step": 4470
+    },
+    {
+      "entropy": 2.026275262236595,
+      "epoch": 3.234784179158389,
+      "grad_norm": 0.0008903327980078757,
+      "learning_rate": 5.860552369449591e-06,
+      "loss": 0.00019860574975609778,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15101750.0,
+      "step": 4480
+    },
+    {
+      "entropy": 2.035792517662048,
+      "epoch": 3.242008307747878,
+      "grad_norm": 0.0385742224752903,
+      "learning_rate": 5.8180222739366014e-06,
+      "loss": 0.00011005952255800366,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15132217.0,
+      "step": 4490
+    },
+    {
+      "entropy": 2.0055119007825852,
+      "epoch": 3.249232436337367,
+      "grad_norm": 0.014958804473280907,
+      "learning_rate": 5.77558365922753e-06,
+      "loss": 8.74112593010068e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15165100.0,
+      "step": 4500
+    },
+    {
+      "epoch": 3.249232436337367,
+      "eval_entropy": 1.8884758904436165,
+      "eval_loss": 1.0809279680252075,
+      "eval_mean_token_accuracy": 0.8737331791980388,
+      "eval_num_tokens": 15165100.0,
+      "eval_runtime": 1004.9032,
+      "eval_samples_per_second": 7.614,
+      "eval_steps_per_second": 0.952,
+      "step": 4500
+    },
+    {
+      "entropy": 2.0021767020225525,
+      "epoch": 3.2564565649268555,
+      "grad_norm": 0.0017272443510591984,
+      "learning_rate": 5.7332374536674914e-06,
+      "loss": 0.00018979933811351656,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15203811.0,
+      "step": 4510
+    },
+    {
+      "entropy": 2.0449478060007094,
+      "epoch": 3.2636806935163447,
+      "grad_norm": 0.0016825678758323193,
+      "learning_rate": 5.690984583580155e-06,
+      "loss": 0.0001557380543090403,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15235220.0,
+      "step": 4520
+    },
+    {
+      "entropy": 2.0222400814294814,
+      "epoch": 3.2709048221058334,
+      "grad_norm": 0.24223871529102325,
+      "learning_rate": 5.648825973247459e-06,
+      "loss": 0.00020689407829195262,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15271037.0,
+      "step": 4530
+    },
+    {
+      "entropy": 2.000539484620094,
+      "epoch": 3.278128950695322,
+      "grad_norm": 0.0028660078532993793,
+      "learning_rate": 5.60676254488943e-06,
+      "loss": 0.00011368633713573217,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15305922.0,
+      "step": 4540
+    },
+    {
+      "entropy": 2.029051575064659,
+      "epoch": 3.2853530792848114,
+      "grad_norm": 0.00038728417712263763,
+      "learning_rate": 5.56479521864397e-06,
+      "loss": 7.007254171185196e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15337389.0,
+      "step": 4550
+    },
+    {
+      "entropy": 2.0867973893880842,
+      "epoch": 3.2925772078743,
+      "grad_norm": 0.0017691091634333134,
+      "learning_rate": 5.522924912546761e-06,
+      "loss": 2.9183743754401804e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15367632.0,
+      "step": 4560
+    },
+    {
+      "entropy": 2.038247913122177,
+      "epoch": 3.299801336463789,
+      "grad_norm": 0.001519956742413342,
+      "learning_rate": 5.481152542511152e-06,
+      "loss": 6.189012783579529e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15398365.0,
+      "step": 4570
+    },
+    {
+      "entropy": 2.0492154121398927,
+      "epoch": 3.307025465053278,
+      "grad_norm": 0.05697038024663925,
+      "learning_rate": 5.439479022308156e-06,
+      "loss": 0.00017389660933986306,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15429567.0,
+      "step": 4580
+    },
+    {
+      "entropy": 1.9483360826969147,
+      "epoch": 3.3142495936427667,
+      "grad_norm": 0.0008817281923256814,
+      "learning_rate": 5.3979052635464325e-06,
+      "loss": 7.576268981210888e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15463743.0,
+      "step": 4590
+    },
+    {
+      "entropy": 2.0008713573217394,
+      "epoch": 3.321473722232256,
+      "grad_norm": 0.010526027530431747,
+      "learning_rate": 5.35643217565237e-06,
+      "loss": 4.54952591098845e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15491685.0,
+      "step": 4600
+    },
+    {
+      "entropy": 2.0740255534648897,
+      "epoch": 3.3286978508217446,
+      "grad_norm": 0.0018182476051151752,
+      "learning_rate": 5.315060665850165e-06,
+      "loss": 5.49975608009845e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15522064.0,
+      "step": 4610
+    },
+    {
+      "entropy": 2.0470830738544463,
+      "epoch": 3.3359219794112334,
+      "grad_norm": 0.0010773467365652323,
+      "learning_rate": 5.273791639142012e-06,
+      "loss": 7.785457419231533e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15554349.0,
+      "step": 4620
+    },
+    {
+      "entropy": 2.075201654434204,
+      "epoch": 3.3431461080007225,
+      "grad_norm": 0.0010672385105863214,
+      "learning_rate": 5.2326259982882675e-06,
+      "loss": 7.268890622071921e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15587755.0,
+      "step": 4630
+    },
+    {
+      "entropy": 1.974511331319809,
+      "epoch": 3.3503702365902113,
+      "grad_norm": 0.02245929464697838,
+      "learning_rate": 5.191564643787739e-06,
+      "loss": 8.146609761752188e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15619055.0,
+      "step": 4640
+    },
+    {
+      "entropy": 2.1137306481599807,
+      "epoch": 3.3575943651797004,
+      "grad_norm": 0.02009975165128708,
+      "learning_rate": 5.150608473857951e-06,
+      "loss": 9.43507591728121e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15651956.0,
+      "step": 4650
+    },
+    {
+      "entropy": 2.054837018251419,
+      "epoch": 3.364818493769189,
+      "grad_norm": 0.0010314187966287136,
+      "learning_rate": 5.109758384415533e-06,
+      "loss": 0.00012439355487003922,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15686287.0,
+      "step": 4660
+    },
+    {
+      "entropy": 2.0372716188430786,
+      "epoch": 3.372042622358678,
+      "grad_norm": 0.003698289394378662,
+      "learning_rate": 5.069015269056587e-06,
+      "loss": 9.244094835594296e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15719188.0,
+      "step": 4670
+    },
+    {
+      "entropy": 2.055976676940918,
+      "epoch": 3.379266750948167,
+      "grad_norm": 0.03741403669118881,
+      "learning_rate": 5.028380019037165e-06,
+      "loss": 0.00013077817857265472,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15752790.0,
+      "step": 4680
+    },
+    {
+      "entropy": 2.085843190550804,
+      "epoch": 3.386490879537656,
+      "grad_norm": 0.0007527766865678132,
+      "learning_rate": 4.98785352325376e-06,
+      "loss": 5.083156866021454e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15787893.0,
+      "step": 4690
+    },
+    {
+      "entropy": 2.032337984442711,
+      "epoch": 3.3937150081271446,
+      "grad_norm": 0.001193483010865748,
+      "learning_rate": 4.9474366682238664e-06,
+      "loss": 0.00010785561753436922,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15823196.0,
+      "step": 4700
+    },
+    {
+      "entropy": 2.017588832974434,
+      "epoch": 3.4009391367166337,
+      "grad_norm": 0.000920341641176492,
+      "learning_rate": 4.907130338066589e-06,
+      "loss": 6.847251788713037e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15858171.0,
+      "step": 4710
+    },
+    {
+      "entropy": 2.014309874176979,
+      "epoch": 3.4081632653061225,
+      "grad_norm": 0.0028325754683464766,
+      "learning_rate": 4.866935414483291e-06,
+      "loss": 6.0408358694985506e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15888184.0,
+      "step": 4720
+    },
+    {
+      "entropy": 1.9666501939296723,
+      "epoch": 3.415387393895611,
+      "grad_norm": 0.0004978736978955567,
+      "learning_rate": 4.8268527767383266e-06,
+      "loss": 6.155979936011136e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15925857.0,
+      "step": 4730
+    },
+    {
+      "entropy": 2.1072573363780975,
+      "epoch": 3.4226115224851004,
+      "grad_norm": 0.0013934965245425701,
+      "learning_rate": 4.786883301639786e-06,
+      "loss": 9.986537043005228e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15955224.0,
+      "step": 4740
+    },
+    {
+      "entropy": 2.01272608935833,
+      "epoch": 3.429835651074589,
+      "grad_norm": 0.004395121242851019,
+      "learning_rate": 4.747027863520337e-06,
+      "loss": 8.816031040623785e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 15987442.0,
+      "step": 4750
+    },
+    {
+      "entropy": 1.9714742422103881,
+      "epoch": 3.437059779664078,
+      "grad_norm": 0.0011384258978068829,
+      "learning_rate": 4.7072873342180735e-06,
+      "loss": 0.0002765763783827424,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16024444.0,
+      "step": 4760
+    },
+    {
+      "entropy": 2.0219936460256576,
+      "epoch": 3.444283908253567,
+      "grad_norm": 0.0011205687187612057,
+      "learning_rate": 4.667662583057473e-06,
+      "loss": 5.4150784853845836e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16061081.0,
+      "step": 4770
+    },
+    {
+      "entropy": 1.9706620872020721,
+      "epoch": 3.4515080368430557,
+      "grad_norm": 0.008661802858114243,
+      "learning_rate": 4.628154476830348e-06,
+      "loss": 5.808487185277045e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16101282.0,
+      "step": 4780
+    },
+    {
+      "entropy": 2.010636579990387,
+      "epoch": 3.4587321654325445,
+      "grad_norm": 0.016732508316636086,
+      "learning_rate": 4.58876387977692e-06,
+      "loss": 0.00012360099935904145,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16138042.0,
+      "step": 4790
+    },
+    {
+      "entropy": 2.015199688076973,
+      "epoch": 3.4659562940220336,
+      "grad_norm": 0.0014761262573301792,
+      "learning_rate": 4.549491653566879e-06,
+      "loss": 4.3076850124634804e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16171761.0,
+      "step": 4800
+    },
+    {
+      "entropy": 1.9639876455068588,
+      "epoch": 3.4731804226115224,
+      "grad_norm": 0.07578303664922714,
+      "learning_rate": 4.5103386572805676e-06,
+      "loss": 0.0001370429410599172,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16208331.0,
+      "step": 4810
+    },
+    {
+      "entropy": 2.0875407248735427,
+      "epoch": 3.4804045512010116,
+      "grad_norm": 0.0006163098732940853,
+      "learning_rate": 4.471305747390159e-06,
+      "loss": 8.360695792362094e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16239154.0,
+      "step": 4820
+    },
+    {
+      "entropy": 2.053487077355385,
+      "epoch": 3.4876286797905003,
+      "grad_norm": 0.0032585824374109507,
+      "learning_rate": 4.432393777740945e-06,
+      "loss": 6.120784091763198e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16273072.0,
+      "step": 4830
+    },
+    {
+      "entropy": 2.1002973705530166,
+      "epoch": 3.494852808379989,
+      "grad_norm": 0.004883012268692255,
+      "learning_rate": 4.393603599532642e-06,
+      "loss": 6.172609282657504e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16305328.0,
+      "step": 4840
+    },
+    {
+      "entropy": 2.0459698259830477,
+      "epoch": 3.502076936969478,
+      "grad_norm": 0.0010911468416452408,
+      "learning_rate": 4.354936061300784e-06,
+      "loss": 8.416592027060688e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16336726.0,
+      "step": 4850
+    },
+    {
+      "entropy": 1.9999713480472565,
+      "epoch": 3.509301065558967,
+      "grad_norm": 0.0006581201450899243,
+      "learning_rate": 4.3163920088981434e-06,
+      "loss": 5.7820964138954876e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16374779.0,
+      "step": 4860
+    },
+    {
+      "entropy": 2.0465083062648772,
+      "epoch": 3.516525194148456,
+      "grad_norm": 0.00047667152830399573,
+      "learning_rate": 4.277972285476255e-06,
+      "loss": 4.597740771714598e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16404204.0,
+      "step": 4870
+    },
+    {
+      "entropy": 1.9927714973688126,
+      "epoch": 3.523749322737945,
+      "grad_norm": 0.007273843511939049,
+      "learning_rate": 4.239677731466939e-06,
+      "loss": 0.00034961386118084194,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16439756.0,
+      "step": 4880
+    },
+    {
+      "entropy": 2.025309592485428,
+      "epoch": 3.5309734513274336,
+      "grad_norm": 0.0002263276546727866,
+      "learning_rate": 4.201509184563947e-06,
+      "loss": 8.668734808452428e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16475447.0,
+      "step": 4890
+    },
+    {
+      "entropy": 2.0977191269397735,
+      "epoch": 3.5381975799169227,
+      "grad_norm": 0.0010200958931818604,
+      "learning_rate": 4.163467479704612e-06,
+      "loss": 4.953600000590086e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16504583.0,
+      "step": 4900
+    },
+    {
+      "entropy": 2.088295194506645,
+      "epoch": 3.5454217085064115,
+      "grad_norm": 0.0011138464324176311,
+      "learning_rate": 4.1255534490516105e-06,
+      "loss": 4.0841943700797856e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16539294.0,
+      "step": 4910
+    },
+    {
+      "entropy": 2.0325693815946577,
+      "epoch": 3.5526458370959,
+      "grad_norm": 0.10155083239078522,
+      "learning_rate": 4.087767921974727e-06,
+      "loss": 0.0002867079107090831,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16572157.0,
+      "step": 4920
+    },
+    {
+      "entropy": 2.084858962893486,
+      "epoch": 3.5598699656853894,
+      "grad_norm": 0.0021113206166774035,
+      "learning_rate": 4.050111725032746e-06,
+      "loss": 5.9655355289578435e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16605161.0,
+      "step": 4930
+    },
+    {
+      "entropy": 1.975836205482483,
+      "epoch": 3.567094094274878,
+      "grad_norm": 0.0003153013240080327,
+      "learning_rate": 4.012585681955339e-06,
+      "loss": 0.00014247578801587223,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16638719.0,
+      "step": 4940
+    },
+    {
+      "entropy": 1.9811270862817765,
+      "epoch": 3.574318222864367,
+      "grad_norm": 0.0007030869019217789,
+      "learning_rate": 3.975190613625076e-06,
+      "loss": 0.0007503291126340628,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16677651.0,
+      "step": 4950
+    },
+    {
+      "entropy": 2.093290412425995,
+      "epoch": 3.581542351453856,
+      "grad_norm": 0.0005874686175957322,
+      "learning_rate": 3.93792733805944e-06,
+      "loss": 8.813977474346757e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16710154.0,
+      "step": 4960
+    },
+    {
+      "entropy": 2.02408949136734,
+      "epoch": 3.5887664800433448,
+      "grad_norm": 0.0016752072842791677,
+      "learning_rate": 3.900796670392952e-06,
+      "loss": 5.5695866467431185e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16744349.0,
+      "step": 4970
+    },
+    {
+      "entropy": 2.050968898832798,
+      "epoch": 3.5959906086328335,
+      "grad_norm": 0.0057442523539066315,
+      "learning_rate": 3.863799422859328e-06,
+      "loss": 5.2600842900574206e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16779567.0,
+      "step": 4980
+    },
+    {
+      "entropy": 2.062971496582031,
+      "epoch": 3.6032147372223227,
+      "grad_norm": 0.01761898770928383,
+      "learning_rate": 3.826936404773729e-06,
+      "loss": 0.00018251293804496526,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16814218.0,
+      "step": 4990
+    },
+    {
+      "entropy": 2.0626443684101106,
+      "epoch": 3.6104388658118114,
+      "grad_norm": 0.022460924461483955,
+      "learning_rate": 3.7902084225150325e-06,
+      "loss": 0.00012502372264862062,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16846468.0,
+      "step": 5000
+    },
+    {
+      "epoch": 3.6104388658118114,
+      "eval_entropy": 1.8869691470573688,
+      "eval_loss": 1.1117266416549683,
+      "eval_mean_token_accuracy": 0.8723597162195882,
+      "eval_num_tokens": 16846468.0,
+      "eval_runtime": 1005.0317,
+      "eval_samples_per_second": 7.613,
+      "eval_steps_per_second": 0.952,
+      "step": 5000
+    },
+    {
+      "entropy": 2.0077989429235457,
+      "epoch": 3.6176629944013,
+      "grad_norm": 0.0006305546266958117,
+      "learning_rate": 3.7536162795082175e-06,
+      "loss": 5.892690387554467e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16878696.0,
+      "step": 5010
+    },
+    {
+      "entropy": 2.09415146112442,
+      "epoch": 3.6248871229907893,
+      "grad_norm": 0.0026127202436327934,
+      "learning_rate": 3.7171607762067664e-06,
+      "loss": 3.7291095941327514e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16911183.0,
+      "step": 5020
+    },
+    {
+      "entropy": 2.0023372024297714,
+      "epoch": 3.632111251580278,
+      "grad_norm": 0.0015028188936412334,
+      "learning_rate": 3.6808427100751785e-06,
+      "loss": 5.9176504146307705e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16943059.0,
+      "step": 5030
+    },
+    {
+      "entropy": 2.0355609893798827,
+      "epoch": 3.6393353801697668,
+      "grad_norm": 0.0012083424953743815,
+      "learning_rate": 3.644662875571503e-06,
+      "loss": 2.9069939046166837e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 16973792.0,
+      "step": 5040
+    },
+    {
+      "entropy": 2.065312069654465,
+      "epoch": 3.646559508759256,
+      "grad_norm": 0.002507750876247883,
+      "learning_rate": 3.608622064129982e-06,
+      "loss": 4.3749200995080176e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17009089.0,
+      "step": 5050
+    },
+    {
+      "entropy": 2.0725962966680527,
+      "epoch": 3.6537836373487447,
+      "grad_norm": 0.010749811306595802,
+      "learning_rate": 3.5727210641437126e-06,
+      "loss": 5.394321633502841e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17046624.0,
+      "step": 5060
+    },
+    {
+      "entropy": 2.055601179599762,
+      "epoch": 3.661007765938234,
+      "grad_norm": 0.021961471065878868,
+      "learning_rate": 3.5369606609474307e-06,
+      "loss": 6.937128491699696e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17082420.0,
+      "step": 5070
+    },
+    {
+      "entropy": 1.9774365901947022,
+      "epoch": 3.6682318945277226,
+      "grad_norm": 0.004600843880325556,
+      "learning_rate": 3.5013416368003005e-06,
+      "loss": 3.830065543297678e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17116764.0,
+      "step": 5080
+    },
+    {
+      "entropy": 2.1132187873125075,
+      "epoch": 3.6754560231172118,
+      "grad_norm": 0.0004788452060893178,
+      "learning_rate": 3.465864770868833e-06,
+      "loss": 2.8599039069376886e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17144115.0,
+      "step": 5090
+    },
+    {
+      "entropy": 1.9877906680107116,
+      "epoch": 3.6826801517067005,
+      "grad_norm": 0.000511924852617085,
+      "learning_rate": 3.430530839209817e-06,
+      "loss": 4.964641411788762e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17177077.0,
+      "step": 5100
+    },
+    {
+      "entropy": 1.9881062805652618,
+      "epoch": 3.689904280296189,
+      "grad_norm": 0.0019181350944563746,
+      "learning_rate": 3.3953406147533508e-06,
+      "loss": 4.389421083033085e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17213359.0,
+      "step": 5110
+    },
+    {
+      "entropy": 2.070028159022331,
+      "epoch": 3.6971284088856784,
+      "grad_norm": 0.005504785571247339,
+      "learning_rate": 3.360294867285948e-06,
+      "loss": 4.010663251392543e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17244540.0,
+      "step": 5120
+    },
+    {
+      "entropy": 1.9767438888549804,
+      "epoch": 3.704352537475167,
+      "grad_norm": 0.0003027736092917621,
+      "learning_rate": 3.325394363433674e-06,
+      "loss": 3.884226316586137e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17279513.0,
+      "step": 5130
+    },
+    {
+      "entropy": 2.043398728966713,
+      "epoch": 3.711576666064656,
+      "grad_norm": 0.0009300747769884765,
+      "learning_rate": 3.2906398666453975e-06,
+      "loss": 5.342182121239603e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17308432.0,
+      "step": 5140
+    },
+    {
+      "entropy": 2.0269203215837477,
+      "epoch": 3.718800794654145,
+      "grad_norm": 0.0010025069350376725,
+      "learning_rate": 3.256032137176075e-06,
+      "loss": 4.8141475417651236e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17341587.0,
+      "step": 5150
+    },
+    {
+      "entropy": 2.045577418804169,
+      "epoch": 3.7260249232436338,
+      "grad_norm": 0.007715600077062845,
+      "learning_rate": 3.2215719320701322e-06,
+      "loss": 0.005004372447729111,
+      "mean_token_accuracy": 0.9984375,
+      "num_tokens": 17373509.0,
+      "step": 5160
+    },
+    {
+      "entropy": 2.0684278547763824,
+      "epoch": 3.7332490518331225,
+      "grad_norm": 0.0007976724882610142,
+      "learning_rate": 3.1872600051448922e-06,
+      "loss": 5.152486264705658e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17406403.0,
+      "step": 5170
+    },
+    {
+      "entropy": 2.023306903243065,
+      "epoch": 3.7404731804226117,
+      "grad_norm": 0.0015424734447151423,
+      "learning_rate": 3.1530971069740966e-06,
+      "loss": 0.00023578524123877287,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17443555.0,
+      "step": 5180
+    },
+    {
+      "entropy": 1.9953484445810319,
+      "epoch": 3.7476973090121004,
+      "grad_norm": 0.0004028415132779628,
+      "learning_rate": 3.119083984871476e-06,
+      "loss": 0.00022710610646754504,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17480119.0,
+      "step": 5190
+    },
+    {
+      "entropy": 2.009831804037094,
+      "epoch": 3.754921437601589,
+      "grad_norm": 0.003044330282136798,
+      "learning_rate": 3.085221382874417e-06,
+      "loss": 6.885324837639928e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17513720.0,
+      "step": 5200
+    },
+    {
+      "entropy": 1.9898712247610093,
+      "epoch": 3.7621455661910783,
+      "grad_norm": 0.0006311017787083983,
+      "learning_rate": 3.0515100417276642e-06,
+      "loss": 0.00010676863603293895,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17548810.0,
+      "step": 5210
+    },
+    {
+      "entropy": 2.031584769487381,
+      "epoch": 3.769369694780567,
+      "grad_norm": 0.005535018630325794,
+      "learning_rate": 3.0179506988671435e-06,
+      "loss": 0.001260968390852213,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17585065.0,
+      "step": 5220
+    },
+    {
+      "entropy": 2.011755895614624,
+      "epoch": 3.776593823370056,
+      "grad_norm": 0.0042891171760857105,
+      "learning_rate": 2.9845440884038046e-06,
+      "loss": 0.00010934327729046344,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17618270.0,
+      "step": 5230
+    },
+    {
+      "entropy": 2.02085902094841,
+      "epoch": 3.783817951959545,
+      "grad_norm": 0.0007354042027145624,
+      "learning_rate": 2.951290941107585e-06,
+      "loss": 0.00011397586204111576,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17655698.0,
+      "step": 5240
+    },
+    {
+      "entropy": 2.014476954936981,
+      "epoch": 3.7910420805490337,
+      "grad_norm": 0.0014770718989893794,
+      "learning_rate": 2.9181919843914055e-06,
+      "loss": 0.00023540158290416002,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17686863.0,
+      "step": 5250
+    },
+    {
+      "entropy": 2.0327656120061874,
+      "epoch": 3.7982662091385224,
+      "grad_norm": 0.007617010269314051,
+      "learning_rate": 2.885247942295274e-06,
+      "loss": 9.638697374612094e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17724248.0,
+      "step": 5260
+    },
+    {
+      "entropy": 1.9850782930850983,
+      "epoch": 3.8054903377280116,
+      "grad_norm": 0.005483856424689293,
+      "learning_rate": 2.852459535470431e-06,
+      "loss": 7.874672883190215e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17758437.0,
+      "step": 5270
+    },
+    {
+      "entropy": 2.0032659009099008,
+      "epoch": 3.8127144663175003,
+      "grad_norm": 0.00028377847047522664,
+      "learning_rate": 2.819827481163603e-06,
+      "loss": 3.7973991129547355e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17790728.0,
+      "step": 5280
+    },
+    {
+      "entropy": 1.978261086344719,
+      "epoch": 3.8199385949069895,
+      "grad_norm": 0.0018890398787334561,
+      "learning_rate": 2.787352493201294e-06,
+      "loss": 7.107729325070978e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17829297.0,
+      "step": 5290
+    },
+    {
+      "entropy": 2.003551349043846,
+      "epoch": 3.8271627234964782,
+      "grad_norm": 0.0006683628889732063,
+      "learning_rate": 2.755035281974191e-06,
+      "loss": 4.195647488813847e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17861481.0,
+      "step": 5300
+    },
+    {
+      "entropy": 2.0011202305555345,
+      "epoch": 3.834386852085967,
+      "grad_norm": 0.0010510816937312484,
+      "learning_rate": 2.722876554421603e-06,
+      "loss": 0.0023531623184680937,
+      "mean_token_accuracy": 0.9980769231915474,
+      "num_tokens": 17899030.0,
+      "step": 5310
+    },
+    {
+      "entropy": 1.981515035033226,
+      "epoch": 3.841610980675456,
+      "grad_norm": 0.0009371911291964352,
+      "learning_rate": 2.6908770140160155e-06,
+      "loss": 3.624772652983665e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17936142.0,
+      "step": 5320
+    },
+    {
+      "entropy": 1.966176024079323,
+      "epoch": 3.848835109264945,
+      "grad_norm": 0.011589719913899899,
+      "learning_rate": 2.659037360747686e-06,
+      "loss": 4.514714237302542e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 17969355.0,
+      "step": 5330
+    },
+    {
+      "entropy": 1.9954922467470169,
+      "epoch": 3.856059237854434,
+      "grad_norm": 0.0006365366862155497,
+      "learning_rate": 2.6273582911093478e-06,
+      "loss": 5.152818048372865e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18006001.0,
+      "step": 5340
+    },
+    {
+      "entropy": 2.0939167737960815,
+      "epoch": 3.863283366443923,
+      "grad_norm": 0.0013512835139408708,
+      "learning_rate": 2.5958404980809547e-06,
+      "loss": 0.0001905821030959487,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18039417.0,
+      "step": 5350
+    },
+    {
+      "entropy": 2.0816452115774156,
+      "epoch": 3.8705074950334115,
+      "grad_norm": 0.004600442014634609,
+      "learning_rate": 2.5644846711145465e-06,
+      "loss": 5.767461261712015e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18075820.0,
+      "step": 5360
+    },
+    {
+      "entropy": 2.0448678851127626,
+      "epoch": 3.8777316236229007,
+      "grad_norm": 0.0038580053951591253,
+      "learning_rate": 2.533291496119141e-06,
+      "loss": 3.333230270072818e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18107244.0,
+      "step": 5370
+    },
+    {
+      "entropy": 1.9495148241519928,
+      "epoch": 3.8849557522123894,
+      "grad_norm": 0.008305185474455357,
+      "learning_rate": 2.5022616554457533e-06,
+      "loss": 0.000500003295019269,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18146835.0,
+      "step": 5380
+    },
+    {
+      "entropy": 2.11137031018734,
+      "epoch": 3.892179880801878,
+      "grad_norm": 0.0028492852579802275,
+      "learning_rate": 2.4713958278724503e-06,
+      "loss": 8.689896203577519e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18175825.0,
+      "step": 5390
+    },
+    {
+      "entropy": 2.0402226716279985,
+      "epoch": 3.8994040093913673,
+      "grad_norm": 0.008078484795987606,
+      "learning_rate": 2.440694688589512e-06,
+      "loss": 5.0963758258149025e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18209999.0,
+      "step": 5400
+    },
+    {
+      "entropy": 1.984485951066017,
+      "epoch": 3.906628137980856,
+      "grad_norm": 0.10858892649412155,
+      "learning_rate": 2.4101589091846677e-06,
+      "loss": 6.12112577073276e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18244424.0,
+      "step": 5410
+    },
+    {
+      "entropy": 2.01485433280468,
+      "epoch": 3.913852266570345,
+      "grad_norm": 0.0008091669878922403,
+      "learning_rate": 2.379789157628387e-06,
+      "loss": 3.532860428094864e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18277674.0,
+      "step": 5420
+    },
+    {
+      "entropy": 2.0380670696496965,
+      "epoch": 3.921076395159834,
+      "grad_norm": 0.0016133576864376664,
+      "learning_rate": 2.3495860982592887e-06,
+      "loss": 5.4576556431129573e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18312666.0,
+      "step": 5430
+    },
+    {
+      "entropy": 2.0862816482782365,
+      "epoch": 3.9283005237493227,
+      "grad_norm": 0.0010300560388714075,
+      "learning_rate": 2.3195503917695903e-06,
+      "loss": 0.0007599231321364641,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18341922.0,
+      "step": 5440
+    },
+    {
+      "entropy": 2.0342622667551042,
+      "epoch": 3.9355246523388114,
+      "grad_norm": 0.0007309906068257987,
+      "learning_rate": 2.2896826951906704e-06,
+      "loss": 6.794939399696886e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18373576.0,
+      "step": 5450
+    },
+    {
+      "entropy": 2.0291885286569595,
+      "epoch": 3.9427487809283006,
+      "grad_norm": 0.000551744771655649,
+      "learning_rate": 2.259983661878681e-06,
+      "loss": 0.001429925113916397,
+      "mean_token_accuracy": 0.9985294118523598,
+      "num_tokens": 18408254.0,
+      "step": 5460
+    },
+    {
+      "entropy": 2.054783394932747,
+      "epoch": 3.9499729095177893,
+      "grad_norm": 0.0015953091206029058,
+      "learning_rate": 2.2304539415002735e-06,
+      "loss": 5.455065984278917e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18440474.0,
+      "step": 5470
+    },
+    {
+      "entropy": 2.1094057828187944,
+      "epoch": 3.957197038107278,
+      "grad_norm": 0.012093332596123219,
+      "learning_rate": 2.2010941800183692e-06,
+      "loss": 6.816875538788736e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18469821.0,
+      "step": 5480
+    },
+    {
+      "entropy": 2.0123610883951186,
+      "epoch": 3.9644211666967673,
+      "grad_norm": 0.0009424237650819123,
+      "learning_rate": 2.1719050196780368e-06,
+      "loss": 3.265676496084779e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18508180.0,
+      "step": 5490
+    },
+    {
+      "entropy": 2.1019891053438187,
+      "epoch": 3.971645295286256,
+      "grad_norm": 0.0016745254397392273,
+      "learning_rate": 2.1428870989924523e-06,
+      "loss": 4.188414022792131e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18537306.0,
+      "step": 5500
+    },
+    {
+      "epoch": 3.971645295286256,
+      "eval_entropy": 1.8881443387662356,
+      "eval_loss": 1.1148877143859863,
+      "eval_mean_token_accuracy": 0.8733496793634341,
+      "eval_num_tokens": 18537306.0,
+      "eval_runtime": 1003.2062,
+      "eval_samples_per_second": 7.627,
+      "eval_steps_per_second": 0.954,
+      "step": 5500
+    },
+    {
+      "entropy": 2.1032094329595568,
+      "epoch": 3.9788694238757447,
+      "grad_norm": 0.006147739477455616,
+      "learning_rate": 2.114041052728911e-06,
+      "loss": 5.7927443413063885e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18565742.0,
+      "step": 5510
+    },
+    {
+      "entropy": 2.017053109407425,
+      "epoch": 3.986093552465234,
+      "grad_norm": 0.0012200704077258706,
+      "learning_rate": 2.085367511894959e-06,
+      "loss": 3.577861934900284e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18599478.0,
+      "step": 5520
+    },
+    {
+      "entropy": 2.119554433226585,
+      "epoch": 3.9933176810547226,
+      "grad_norm": 0.0014397584600374103,
+      "learning_rate": 2.056867103724588e-06,
+      "loss": 7.5902248499915e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18628278.0,
+      "step": 5530
+    },
+    {
+      "entropy": 2.0237791248269983,
+      "epoch": 4.0,
+      "grad_norm": 0.0003623274969868362,
+      "learning_rate": 2.028540451664501e-06,
+      "loss": 3.263393009547144e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18663760.0,
+      "step": 5540
+    },
+    {
+      "entropy": 2.0520517587661744,
+      "epoch": 4.007224128589489,
+      "grad_norm": 0.005471021868288517,
+      "learning_rate": 2.0003881753604993e-06,
+      "loss": 0.0001229748479090631,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18697411.0,
+      "step": 5550
+    },
+    {
+      "entropy": 1.9183115169405938,
+      "epoch": 4.0144482571789775,
+      "grad_norm": 0.002424974460154772,
+      "learning_rate": 1.9724108906438965e-06,
+      "loss": 0.00012065081391483545,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18738405.0,
+      "step": 5560
+    },
+    {
+      "entropy": 1.9771938562393188,
+      "epoch": 4.021672385768467,
+      "grad_norm": 0.01849484071135521,
+      "learning_rate": 1.9446092095180757e-06,
+      "loss": 0.00019705821759998798,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18777815.0,
+      "step": 5570
+    },
+    {
+      "entropy": 2.000231164693832,
+      "epoch": 4.028896514357956,
+      "grad_norm": 0.0007417634478770196,
+      "learning_rate": 1.916983740145082e-06,
+      "loss": 4.214680520817637e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18811678.0,
+      "step": 5580
+    },
+    {
+      "entropy": 2.08649323284626,
+      "epoch": 4.036120642947444,
+      "grad_norm": 0.0005442023393698037,
+      "learning_rate": 1.8895350868323336e-06,
+      "loss": 5.348328268155456e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18841150.0,
+      "step": 5590
+    },
+    {
+      "entropy": 2.052967756986618,
+      "epoch": 4.043344771536933,
+      "grad_norm": 0.029023820534348488,
+      "learning_rate": 1.8622638500193878e-06,
+      "loss": 7.951875450089574e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18872327.0,
+      "step": 5600
+    },
+    {
+      "entropy": 1.9854425325989724,
+      "epoch": 4.0505689001264225,
+      "grad_norm": 0.001789949252270162,
+      "learning_rate": 1.8351706262648217e-06,
+      "loss": 0.00010194053174927831,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18908521.0,
+      "step": 5610
+    },
+    {
+      "entropy": 2.0517650842666626,
+      "epoch": 4.057793028715911,
+      "grad_norm": 0.00034689446329139173,
+      "learning_rate": 1.808256008233169e-06,
+      "loss": 3.732289478648454e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18942866.0,
+      "step": 5620
+    },
+    {
+      "entropy": 2.042857128381729,
+      "epoch": 4.0650171573054,
+      "grad_norm": 0.001570753171108663,
+      "learning_rate": 1.781520584681966e-06,
+      "loss": 8.434284245595336e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 18978591.0,
+      "step": 5630
+    },
+    {
+      "entropy": 2.0445887804031373,
+      "epoch": 4.072241285894889,
+      "grad_norm": 0.002613665768876672,
+      "learning_rate": 1.7549649404488644e-06,
+      "loss": 8.788983686827123e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19011261.0,
+      "step": 5640
+    },
+    {
+      "entropy": 2.0116062104701995,
+      "epoch": 4.079465414484378,
+      "grad_norm": 0.0019005907233804464,
+      "learning_rate": 1.728589656438846e-06,
+      "loss": 2.841704699676484e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19050831.0,
+      "step": 5650
+    },
+    {
+      "entropy": 2.0562655299901964,
+      "epoch": 4.086689543073867,
+      "grad_norm": 0.0002598224382381886,
+      "learning_rate": 1.7023953096115032e-06,
+      "loss": 3.73309594579041e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19081443.0,
+      "step": 5660
+    },
+    {
+      "entropy": 2.0815482050180436,
+      "epoch": 4.093913671663356,
+      "grad_norm": 0.004718273412436247,
+      "learning_rate": 1.6763824729684364e-06,
+      "loss": 6.015845574438572e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19116035.0,
+      "step": 5670
+    },
+    {
+      "entropy": 2.017623296380043,
+      "epoch": 4.101137800252845,
+      "grad_norm": 0.0005751053686253726,
+      "learning_rate": 1.650551715540698e-06,
+      "loss": 4.4731769594363866e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19148343.0,
+      "step": 5680
+    },
+    {
+      "entropy": 1.9582554414868354,
+      "epoch": 4.108361928842333,
+      "grad_norm": 0.0016292211366817355,
+      "learning_rate": 1.6249036023763654e-06,
+      "loss": 3.706713323481381e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19187626.0,
+      "step": 5690
+    },
+    {
+      "entropy": 2.0343296408653258,
+      "epoch": 4.115586057431822,
+      "grad_norm": 0.0009523277985863388,
+      "learning_rate": 1.5994386945281626e-06,
+      "loss": 3.8216341636143625e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19221929.0,
+      "step": 5700
+    },
+    {
+      "entropy": 2.0681279003620148,
+      "epoch": 4.1228101860213116,
+      "grad_norm": 0.0008883064147084951,
+      "learning_rate": 1.5741575490412043e-06,
+      "loss": 2.718248579185456e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19254907.0,
+      "step": 5710
+    },
+    {
+      "entropy": 2.011669307947159,
+      "epoch": 4.1300343146108,
+      "grad_norm": 0.0008795072790235281,
+      "learning_rate": 1.5490607189407903e-06,
+      "loss": 3.153630532324314e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19287150.0,
+      "step": 5720
+    },
+    {
+      "entropy": 2.0814460307359695,
+      "epoch": 4.137258443200289,
+      "grad_norm": 0.001299955532886088,
+      "learning_rate": 1.5241487532203324e-06,
+      "loss": 0.00039165206253528596,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19320643.0,
+      "step": 5730
+    },
+    {
+      "entropy": 2.053331843018532,
+      "epoch": 4.144482571789778,
+      "grad_norm": 0.003343022894114256,
+      "learning_rate": 1.4994221968293209e-06,
+      "loss": 3.842532460112125e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19353163.0,
+      "step": 5740
+    },
+    {
+      "entropy": 2.051739862561226,
+      "epoch": 4.1517067003792665,
+      "grad_norm": 0.0025682623963803053,
+      "learning_rate": 1.4748815906614256e-06,
+      "loss": 4.806564247701317e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19381773.0,
+      "step": 5750
+    },
+    {
+      "entropy": 1.9667051672935485,
+      "epoch": 4.158930828968756,
+      "grad_norm": 0.000453523505711928,
+      "learning_rate": 1.450527471542641e-06,
+      "loss": 7.420713664032518e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19417822.0,
+      "step": 5760
+    },
+    {
+      "entropy": 2.061727538704872,
+      "epoch": 4.166154957558245,
+      "grad_norm": 0.0005199933657422662,
+      "learning_rate": 1.4263603722195696e-06,
+      "loss": 7.046694518066942e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19449523.0,
+      "step": 5770
+    },
+    {
+      "entropy": 1.944933032989502,
+      "epoch": 4.173379086147733,
+      "grad_norm": 0.0017413697205483913,
+      "learning_rate": 1.4023808213477419e-06,
+      "loss": 3.615806344896555e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19492156.0,
+      "step": 5780
+    },
+    {
+      "entropy": 2.0748004257678985,
+      "epoch": 4.180603214737222,
+      "grad_norm": 0.0007003610371612012,
+      "learning_rate": 1.3785893434800735e-06,
+      "loss": 4.728134081233293e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19526200.0,
+      "step": 5790
+    },
+    {
+      "entropy": 2.0519924253225326,
+      "epoch": 4.1878273433267115,
+      "grad_norm": 0.004808162804692984,
+      "learning_rate": 1.354986459055374e-06,
+      "loss": 5.692571285180748e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19556309.0,
+      "step": 5800
+    },
+    {
+      "entropy": 2.0686589539051057,
+      "epoch": 4.1950514719162,
+      "grad_norm": 0.005478905979543924,
+      "learning_rate": 1.3315726843869703e-06,
+      "loss": 5.530117196030915e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19586633.0,
+      "step": 5810
+    },
+    {
+      "entropy": 1.9959229797124862,
+      "epoch": 4.202275600505689,
+      "grad_norm": 0.0006617996259592474,
+      "learning_rate": 1.30834853165142e-06,
+      "loss": 4.879834596067667e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19622993.0,
+      "step": 5820
+    },
+    {
+      "entropy": 2.063421201705933,
+      "epoch": 4.209499729095178,
+      "grad_norm": 0.0005491877673193812,
+      "learning_rate": 1.2853145088772867e-06,
+      "loss": 3.759705286938697e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19650842.0,
+      "step": 5830
+    },
+    {
+      "entropy": 2.0779141038656235,
+      "epoch": 4.216723857684666,
+      "grad_norm": 0.0042537664994597435,
+      "learning_rate": 1.26247111993405e-06,
+      "loss": 5.043308483436704e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19683451.0,
+      "step": 5840
+    },
+    {
+      "entropy": 2.0339601665735243,
+      "epoch": 4.223947986274156,
+      "grad_norm": 0.000250296201556921,
+      "learning_rate": 1.239818864521063e-06,
+      "loss": 3.1771216890774666e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19713433.0,
+      "step": 5850
+    },
+    {
+      "entropy": 2.0060952216386796,
+      "epoch": 4.231172114863645,
+      "grad_norm": 0.00018487044144421816,
+      "learning_rate": 1.2173582381566406e-06,
+      "loss": 4.990113084204495e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19745784.0,
+      "step": 5860
+    },
+    {
+      "entropy": 1.9054115578532218,
+      "epoch": 4.238396243453133,
+      "grad_norm": 0.008540745824575424,
+      "learning_rate": 1.1950897321672028e-06,
+      "loss": 3.5091297468170525e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19788454.0,
+      "step": 5870
+    },
+    {
+      "entropy": 2.07164244055748,
+      "epoch": 4.245620372042622,
+      "grad_norm": 0.005256814416497946,
+      "learning_rate": 1.1730138336765395e-06,
+      "loss": 4.395857686176896e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19819428.0,
+      "step": 5880
+    },
+    {
+      "entropy": 2.037181797623634,
+      "epoch": 4.252844500632111,
+      "grad_norm": 0.0005752044962719083,
+      "learning_rate": 1.1511310255951436e-06,
+      "loss": 3.161632921546698e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19858243.0,
+      "step": 5890
+    },
+    {
+      "entropy": 2.064584118127823,
+      "epoch": 4.2600686292216,
+      "grad_norm": 0.002521161688491702,
+      "learning_rate": 1.1294417866096618e-06,
+      "loss": 3.7657920620404185e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19889760.0,
+      "step": 5900
+    },
+    {
+      "entropy": 2.0095575898885727,
+      "epoch": 4.267292757811089,
+      "grad_norm": 0.0036700242199003696,
+      "learning_rate": 1.1079465911724063e-06,
+      "loss": 4.4873528531752525e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19923286.0,
+      "step": 5910
+    },
+    {
+      "entropy": 2.090346944332123,
+      "epoch": 4.274516886400578,
+      "grad_norm": 0.008253049105405807,
+      "learning_rate": 1.0866459094909931e-06,
+      "loss": 4.445763770490885e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19951812.0,
+      "step": 5920
+    },
+    {
+      "entropy": 2.0771326392889025,
+      "epoch": 4.281741014990067,
+      "grad_norm": 0.0017921896651387215,
+      "learning_rate": 1.06554020751804e-06,
+      "loss": 3.478447324596345e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 19986840.0,
+      "step": 5930
+    },
+    {
+      "entropy": 2.088083365559578,
+      "epoch": 4.2889651435795555,
+      "grad_norm": 0.0014664095360785723,
+      "learning_rate": 1.044629946940987e-06,
+      "loss": 4.977693315595388e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20027096.0,
+      "step": 5940
+    },
+    {
+      "entropy": 2.1036684960126877,
+      "epoch": 4.296189272169045,
+      "grad_norm": 0.0010140404338017106,
+      "learning_rate": 1.0239155851719896e-06,
+      "loss": 2.367785637034103e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20059239.0,
+      "step": 5950
+    },
+    {
+      "entropy": 2.0400282979011535,
+      "epoch": 4.303413400758534,
+      "grad_norm": 0.0007224315195344388,
+      "learning_rate": 1.0033975753379166e-06,
+      "loss": 3.497881989460439e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20096547.0,
+      "step": 5960
+    },
+    {
+      "entropy": 1.9657370626926423,
+      "epoch": 4.310637529348022,
+      "grad_norm": 0.011448175646364689,
+      "learning_rate": 9.830763662704345e-07,
+      "loss": 7.57243949919939e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20136655.0,
+      "step": 5970
+    },
+    {
+      "entropy": 2.081816816329956,
+      "epoch": 4.317861657937511,
+      "grad_norm": 0.0038148025050759315,
+      "learning_rate": 9.629524024961911e-07,
+      "loss": 5.8540661120787264e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20170189.0,
+      "step": 5980
+    },
+    {
+      "entropy": 2.0057587683200837,
+      "epoch": 4.3250857865270005,
+      "grad_norm": 0.006282564718276262,
+      "learning_rate": 9.4302612422709e-07,
+      "loss": 4.116616037208587e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20205120.0,
+      "step": 5990
+    },
+    {
+      "entropy": 2.0236620873212816,
+      "epoch": 4.332309915116489,
+      "grad_norm": 0.00033325853291898966,
+      "learning_rate": 9.232979673506693e-07,
+      "loss": 4.479981726035476e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20234412.0,
+      "step": 6000
+    },
+    {
+      "epoch": 4.332309915116489,
+      "eval_entropy": 1.8867571529549865,
+      "eval_loss": 1.1205323934555054,
+      "eval_mean_token_accuracy": 0.8733835436458249,
+      "eval_num_tokens": 20234412.0,
+      "eval_runtime": 1004.0478,
+      "eval_samples_per_second": 7.62,
+      "eval_steps_per_second": 0.953,
+      "step": 6000
+    },
+    {
+      "entropy": 2.038132134079933,
+      "epoch": 4.339534043705978,
+      "grad_norm": 0.0035877148620784283,
+      "learning_rate": 9.037683634205497e-07,
+      "loss": 3.72432143194601e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20269947.0,
+      "step": 6010
+    },
+    {
+      "entropy": 2.037680074572563,
+      "epoch": 4.346758172295467,
+      "grad_norm": 0.0020296431612223387,
+      "learning_rate": 8.844377396470139e-07,
+      "loss": 5.266892840154469e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20302664.0,
+      "step": 6020
+    },
+    {
+      "entropy": 2.0102846056222914,
+      "epoch": 4.353982300884955,
+      "grad_norm": 0.001758980448357761,
+      "learning_rate": 8.653065188876442e-07,
+      "loss": 5.401996313594282e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20338650.0,
+      "step": 6030
+    },
+    {
+      "entropy": 2.0160276383161544,
+      "epoch": 4.361206429474445,
+      "grad_norm": 0.0009282240644097328,
+      "learning_rate": 8.463751196380842e-07,
+      "loss": 0.00013109834399074315,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20376622.0,
+      "step": 6040
+    },
+    {
+      "entropy": 2.0747092813253403,
+      "epoch": 4.368430558063934,
+      "grad_norm": 0.01705658808350563,
+      "learning_rate": 8.276439560228777e-07,
+      "loss": 4.950579605065286e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20410205.0,
+      "step": 6050
+    },
+    {
+      "entropy": 1.9550865411758422,
+      "epoch": 4.375654686653422,
+      "grad_norm": 0.0005268985987640917,
+      "learning_rate": 8.091134377864162e-07,
+      "loss": 3.07281588902697e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20445702.0,
+      "step": 6060
+    },
+    {
+      "entropy": 2.09949953854084,
+      "epoch": 4.382878815242911,
+      "grad_norm": 0.0009493506513535976,
+      "learning_rate": 7.907839702839648e-07,
+      "loss": 5.5659073404967784e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20474828.0,
+      "step": 6070
+    },
+    {
+      "entropy": 2.0183721661567686,
+      "epoch": 4.3901029438324,
+      "grad_norm": 0.00525283720344305,
+      "learning_rate": 7.726559544728074e-07,
+      "loss": 6.567650707438588e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20508159.0,
+      "step": 6080
+    },
+    {
+      "entropy": 2.074488589167595,
+      "epoch": 4.397327072421889,
+      "grad_norm": 0.007316200062632561,
+      "learning_rate": 7.547297869034676e-07,
+      "loss": 4.217610985506326e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20541886.0,
+      "step": 6090
+    },
+    {
+      "entropy": 2.0580724850296974,
+      "epoch": 4.404551201011378,
+      "grad_norm": 0.0010083678644150496,
+      "learning_rate": 7.370058597110397e-07,
+      "loss": 3.816606185864657e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20574514.0,
+      "step": 6100
+    },
+    {
+      "entropy": 2.1109296530485153,
+      "epoch": 4.411775329600867,
+      "grad_norm": 0.001863549230620265,
+      "learning_rate": 7.194845606066014e-07,
+      "loss": 6.602524663321674e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20604343.0,
+      "step": 6110
+    },
+    {
+      "entropy": 2.0420098900794983,
+      "epoch": 4.418999458190356,
+      "grad_norm": 0.0010925420792773366,
+      "learning_rate": 7.021662728687472e-07,
+      "loss": 4.863352223765105e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20634121.0,
+      "step": 6120
+    },
+    {
+      "entropy": 2.092097768187523,
+      "epoch": 4.4262235867798445,
+      "grad_norm": 0.0007851732079870999,
+      "learning_rate": 6.850513753351884e-07,
+      "loss": 0.00010627959854900837,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20672421.0,
+      "step": 6130
+    },
+    {
+      "entropy": 2.050906154513359,
+      "epoch": 4.433447715369334,
+      "grad_norm": 0.004971630405634642,
+      "learning_rate": 6.681402423944794e-07,
+      "loss": 6.272855098359287e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20703713.0,
+      "step": 6140
+    },
+    {
+      "entropy": 2.061757782101631,
+      "epoch": 4.440671843958823,
+      "grad_norm": 0.006796205881983042,
+      "learning_rate": 6.514332439778181e-07,
+      "loss": 6.277162465266884e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20737059.0,
+      "step": 6150
+    },
+    {
+      "entropy": 2.0285993814468384,
+      "epoch": 4.447895972548311,
+      "grad_norm": 0.0004371571121737361,
+      "learning_rate": 6.349307455509612e-07,
+      "loss": 7.777039427310228e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20767549.0,
+      "step": 6160
+    },
+    {
+      "entropy": 2.0497070878744124,
+      "epoch": 4.4551201011378,
+      "grad_norm": 0.0009208957781083882,
+      "learning_rate": 6.186331081062213e-07,
+      "loss": 4.10045322496444e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20802465.0,
+      "step": 6170
+    },
+    {
+      "entropy": 2.029895249009132,
+      "epoch": 4.4623442297272895,
+      "grad_norm": 0.004188739228993654,
+      "learning_rate": 6.025406881545825e-07,
+      "loss": 5.5925117339938876e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20837550.0,
+      "step": 6180
+    },
+    {
+      "entropy": 2.0028230965137483,
+      "epoch": 4.469568358316778,
+      "grad_norm": 0.0003794363292399794,
+      "learning_rate": 5.866538377178843e-07,
+      "loss": 4.9718539230525494e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20876929.0,
+      "step": 6190
+    },
+    {
+      "entropy": 2.0583949118852614,
+      "epoch": 4.476792486906267,
+      "grad_norm": 0.00045059359399601817,
+      "learning_rate": 5.709729043211398e-07,
+      "loss": 3.476354759186506e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20910047.0,
+      "step": 6200
+    },
+    {
+      "entropy": 1.9930611312389375,
+      "epoch": 4.484016615495756,
+      "grad_norm": 0.000457485526567325,
+      "learning_rate": 5.554982309849199e-07,
+      "loss": 4.2988642235286534e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20943601.0,
+      "step": 6210
+    },
+    {
+      "entropy": 2.026224583387375,
+      "epoch": 4.491240744085244,
+      "grad_norm": 0.0005757972248829901,
+      "learning_rate": 5.40230156217858e-07,
+      "loss": 4.6183171798475085e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 20973770.0,
+      "step": 6220
+    },
+    {
+      "entropy": 2.013220492005348,
+      "epoch": 4.498464872674734,
+      "grad_norm": 0.0014384216628968716,
+      "learning_rate": 5.251690140092391e-07,
+      "loss": 4.870914272032678e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21011933.0,
+      "step": 6230
+    },
+    {
+      "entropy": 2.058389586210251,
+      "epoch": 4.505689001264223,
+      "grad_norm": 0.0012885585892945528,
+      "learning_rate": 5.10315133821695e-07,
+      "loss": 5.40612090844661e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21046830.0,
+      "step": 6240
+    },
+    {
+      "entropy": 2.0383670747280123,
+      "epoch": 4.512913129853711,
+      "grad_norm": 0.004921869840472937,
+      "learning_rate": 4.956688405840049e-07,
+      "loss": 3.439287829678506e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21078696.0,
+      "step": 6250
+    },
+    {
+      "entropy": 2.075214546918869,
+      "epoch": 4.5201372584432,
+      "grad_norm": 0.002247129799798131,
+      "learning_rate": 4.81230454683973e-07,
+      "loss": 0.00020004326943308116,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21111401.0,
+      "step": 6260
+    },
+    {
+      "entropy": 2.0556839615106584,
+      "epoch": 4.527361387032689,
+      "grad_norm": 0.0010530236177146435,
+      "learning_rate": 4.670002919614369e-07,
+      "loss": 3.903879842255264e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21142013.0,
+      "step": 6270
+    },
+    {
+      "entropy": 2.088409799337387,
+      "epoch": 4.534585515622178,
+      "grad_norm": 0.00486265541985631,
+      "learning_rate": 4.529786637013445e-07,
+      "loss": 7.261448190547526e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21173184.0,
+      "step": 6280
+    },
+    {
+      "entropy": 2.0148551791906355,
+      "epoch": 4.541809644211667,
+      "grad_norm": 0.0006567241507582366,
+      "learning_rate": 4.391658766269524e-07,
+      "loss": 4.8429635353386405e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21204028.0,
+      "step": 6290
+    },
+    {
+      "entropy": 2.000632828474045,
+      "epoch": 4.549033772801156,
+      "grad_norm": 0.001035205670632422,
+      "learning_rate": 4.25562232893113e-07,
+      "loss": 8.508394821546972e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21237889.0,
+      "step": 6300
+    },
+    {
+      "entropy": 1.9882080256938934,
+      "epoch": 4.556257901390644,
+      "grad_norm": 0.00034841979504562914,
+      "learning_rate": 4.1216803007966823e-07,
+      "loss": 0.0003705588635057211,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21272623.0,
+      "step": 6310
+    },
+    {
+      "entropy": 2.045357859134674,
+      "epoch": 4.5634820299801335,
+      "grad_norm": 0.0020742975175380707,
+      "learning_rate": 3.9898356118493465e-07,
+      "loss": 4.6494670095853506e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21305594.0,
+      "step": 6320
+    },
+    {
+      "entropy": 1.970546191930771,
+      "epoch": 4.570706158569623,
+      "grad_norm": 0.01608274318277836,
+      "learning_rate": 3.8600911461930233e-07,
+      "loss": 3.988950047641993e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21343984.0,
+      "step": 6330
+    },
+    {
+      "entropy": 2.0401349663734436,
+      "epoch": 4.577930287159111,
+      "grad_norm": 0.030885402113199234,
+      "learning_rate": 3.7324497419891305e-07,
+      "loss": 0.00012581488117575647,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21376303.0,
+      "step": 6340
+    },
+    {
+      "entropy": 2.0059654384851457,
+      "epoch": 4.5851544157486,
+      "grad_norm": 0.000837457540910691,
+      "learning_rate": 3.6069141913946836e-07,
+      "loss": 2.7812441112473606e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21408533.0,
+      "step": 6350
+    },
+    {
+      "entropy": 2.0227129757404327,
+      "epoch": 4.592378544338089,
+      "grad_norm": 0.0016140820225700736,
+      "learning_rate": 3.48348724050106e-07,
+      "loss": 4.559661610983312e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21443948.0,
+      "step": 6360
+    },
+    {
+      "entropy": 2.042063796520233,
+      "epoch": 4.599602672927578,
+      "grad_norm": 0.002120249206200242,
+      "learning_rate": 3.3621715892740435e-07,
+      "loss": 4.2640851461328566e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21473536.0,
+      "step": 6370
+    },
+    {
+      "entropy": 1.9752472653985023,
+      "epoch": 4.606826801517067,
+      "grad_norm": 0.002964077051728964,
+      "learning_rate": 3.242969891494685e-07,
+      "loss": 3.5691412631422284e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21507221.0,
+      "step": 6380
+    },
+    {
+      "entropy": 2.0139128088951113,
+      "epoch": 4.614050930106556,
+      "grad_norm": 0.030351942405104637,
+      "learning_rate": 3.125884754701303e-07,
+      "loss": 7.241095299832523e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21541769.0,
+      "step": 6390
+    },
+    {
+      "entropy": 2.0333428621292113,
+      "epoch": 4.621275058696045,
+      "grad_norm": 0.000693982932716608,
+      "learning_rate": 3.010918740132407e-07,
+      "loss": 4.2272888822481033e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21574865.0,
+      "step": 6400
+    },
+    {
+      "entropy": 2.0549756467342375,
+      "epoch": 4.6284991872855334,
+      "grad_norm": 0.0026722848415374756,
+      "learning_rate": 2.898074362670722e-07,
+      "loss": 3.962405025959015e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21609017.0,
+      "step": 6410
+    },
+    {
+      "entropy": 1.9579277276992797,
+      "epoch": 4.635723315875023,
+      "grad_norm": 0.00036720273783430457,
+      "learning_rate": 2.7873540907880857e-07,
+      "loss": 2.7999217854812743e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21645079.0,
+      "step": 6420
+    },
+    {
+      "entropy": 2.038615897297859,
+      "epoch": 4.642947444464512,
+      "grad_norm": 0.0006429091445170343,
+      "learning_rate": 2.678760346491571e-07,
+      "loss": 2.7544971089810133e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21676845.0,
+      "step": 6430
+    },
+    {
+      "entropy": 2.0673672020435334,
+      "epoch": 4.650171573054,
+      "grad_norm": 0.0004574761842377484,
+      "learning_rate": 2.572295505270361e-07,
+      "loss": 4.758274299092591e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21708826.0,
+      "step": 6440
+    },
+    {
+      "entropy": 2.0488554000854493,
+      "epoch": 4.657395701643489,
+      "grad_norm": 0.002278710948303342,
+      "learning_rate": 2.4679618960439354e-07,
+      "loss": 3.772459167521447e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21743886.0,
+      "step": 6450
+    },
+    {
+      "entropy": 2.039225164055824,
+      "epoch": 4.664619830232978,
+      "grad_norm": 0.01752655953168869,
+      "learning_rate": 2.3657618011110105e-07,
+      "loss": 5.816233460791409e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21775228.0,
+      "step": 6460
+    },
+    {
+      "entropy": 2.059047856926918,
+      "epoch": 4.671843958822467,
+      "grad_norm": 0.0005467822193168104,
+      "learning_rate": 2.2656974560996804e-07,
+      "loss": 7.531020673923194e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21808593.0,
+      "step": 6470
+    },
+    {
+      "entropy": 2.1210929095745086,
+      "epoch": 4.679068087411956,
+      "grad_norm": 0.0012726244749501348,
+      "learning_rate": 2.1677710499184656e-07,
+      "loss": 6.718997610732913e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21841076.0,
+      "step": 6480
+    },
+    {
+      "entropy": 2.0896143525838853,
+      "epoch": 4.686292216001445,
+      "grad_norm": 0.003853370202705264,
+      "learning_rate": 2.0719847247084755e-07,
+      "loss": 3.789499460253865e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21875809.0,
+      "step": 6490
+    },
+    {
+      "entropy": 2.037109050154686,
+      "epoch": 4.693516344590933,
+      "grad_norm": 0.00031700858380645514,
+      "learning_rate": 1.9783405757965002e-07,
+      "loss": 3.2565766014158724e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21909724.0,
+      "step": 6500
+    },
+    {
+      "epoch": 4.693516344590933,
+      "eval_entropy": 1.8861580940125007,
+      "eval_loss": 1.1229685544967651,
+      "eval_mean_token_accuracy": 0.8733846908328177,
+      "eval_num_tokens": 21909724.0,
+      "eval_runtime": 1004.3195,
+      "eval_samples_per_second": 7.618,
+      "eval_steps_per_second": 0.953,
+      "step": 6500
+    },
+    {
+      "entropy": 2.0284633457660677,
+      "epoch": 4.7007404731804225,
+      "grad_norm": 0.020248375833034515,
+      "learning_rate": 1.8868406516492244e-07,
+      "loss": 5.51959325093776e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21940685.0,
+      "step": 6510
+    },
+    {
+      "entropy": 1.9785596251487731,
+      "epoch": 4.707964601769912,
+      "grad_norm": 0.0038106630090624094,
+      "learning_rate": 1.797486953828398e-07,
+      "loss": 4.6884844778105617e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 21980865.0,
+      "step": 6520
+    },
+    {
+      "entropy": 1.988435271382332,
+      "epoch": 4.715188730359401,
+      "grad_norm": 0.0015697875060141087,
+      "learning_rate": 1.7102814369470366e-07,
+      "loss": 3.299066156614572e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22015086.0,
+      "step": 6530
+    },
+    {
+      "entropy": 1.9778899431228638,
+      "epoch": 4.722412858948889,
+      "grad_norm": 0.00027815153589472175,
+      "learning_rate": 1.6252260086266792e-07,
+      "loss": 2.8657037182711066e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22051125.0,
+      "step": 6540
+    },
+    {
+      "entropy": 2.040974870324135,
+      "epoch": 4.729636987538378,
+      "grad_norm": 0.003017911920323968,
+      "learning_rate": 1.5423225294556643e-07,
+      "loss": 5.40966575499624e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22082404.0,
+      "step": 6550
+    },
+    {
+      "entropy": 2.0276382893323897,
+      "epoch": 4.7368611161278675,
+      "grad_norm": 0.0021381389815360308,
+      "learning_rate": 1.4615728129483974e-07,
+      "loss": 5.566022009588778e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22115266.0,
+      "step": 6560
+    },
+    {
+      "entropy": 2.060152867436409,
+      "epoch": 4.744085244717356,
+      "grad_norm": 0.0016187896253541112,
+      "learning_rate": 1.3829786255057487e-07,
+      "loss": 2.9083571280352772e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22149791.0,
+      "step": 6570
+    },
+    {
+      "entropy": 2.051058277487755,
+      "epoch": 4.751309373306845,
+      "grad_norm": 0.00024818177917040884,
+      "learning_rate": 1.3065416863763057e-07,
+      "loss": 2.7069501811638474e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22179029.0,
+      "step": 6580
+    },
+    {
+      "entropy": 2.0647223949432374,
+      "epoch": 4.758533501896334,
+      "grad_norm": 0.0028722593560814857,
+      "learning_rate": 1.232263667618905e-07,
+      "loss": 3.091147227678448e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22212576.0,
+      "step": 6590
+    },
+    {
+      "entropy": 2.0464107930660247,
+      "epoch": 4.7657576304858225,
+      "grad_norm": 0.0008732263813726604,
+      "learning_rate": 1.1601461940659031e-07,
+      "loss": 5.022935220040381e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22242912.0,
+      "step": 6600
+    },
+    {
+      "entropy": 2.0649017661809923,
+      "epoch": 4.772981759075312,
+      "grad_norm": 0.0048006330616772175,
+      "learning_rate": 1.0901908432877639e-07,
+      "loss": 5.4197025019675496e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22281513.0,
+      "step": 6610
+    },
+    {
+      "entropy": 1.984046956896782,
+      "epoch": 4.780205887664801,
+      "grad_norm": 0.0030445149168372154,
+      "learning_rate": 1.0223991455584614e-07,
+      "loss": 2.9524989076890052e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22317667.0,
+      "step": 6620
+    },
+    {
+      "entropy": 2.0595791935920715,
+      "epoch": 4.787430016254289,
+      "grad_norm": 0.0011976293753832579,
+      "learning_rate": 9.567725838220743e-08,
+      "loss": 3.321574768051505e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22346682.0,
+      "step": 6630
+    },
+    {
+      "entropy": 2.026864069700241,
+      "epoch": 4.794654144843778,
+      "grad_norm": 0.00043818942504003644,
+      "learning_rate": 8.933125936602671e-08,
+      "loss": 3.484268963802606e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22385013.0,
+      "step": 6640
+    },
+    {
+      "entropy": 2.046583649516106,
+      "epoch": 4.8018782734332675,
+      "grad_norm": 0.0025603007525205612,
+      "learning_rate": 8.32020563260938e-08,
+      "loss": 5.122844595462084e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22418400.0,
+      "step": 6650
+    },
+    {
+      "entropy": 1.975029507279396,
+      "epoch": 4.809102402022756,
+      "grad_norm": 0.0005151774967089295,
+      "learning_rate": 7.728978333878867e-08,
+      "loss": 2.6047686696983872e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22451151.0,
+      "step": 6660
+    },
+    {
+      "entropy": 2.067280852794647,
+      "epoch": 4.816326530612245,
+      "grad_norm": 0.0006476499256677926,
+      "learning_rate": 7.159456973513834e-08,
+      "loss": 3.575811279006302e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22478905.0,
+      "step": 6670
+    },
+    {
+      "entropy": 2.03638633787632,
+      "epoch": 4.823550659201734,
+      "grad_norm": 0.0013524072710424662,
+      "learning_rate": 6.611654009799795e-08,
+      "loss": 4.682386352214962e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22515537.0,
+      "step": 6680
+    },
+    {
+      "entropy": 2.051380980014801,
+      "epoch": 4.830774787791222,
+      "grad_norm": 0.0002707350067794323,
+      "learning_rate": 6.085581425931853e-08,
+      "loss": 3.42914805514738e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22546886.0,
+      "step": 6690
+    },
+    {
+      "entropy": 2.0533095479011534,
+      "epoch": 4.837998916380712,
+      "grad_norm": 0.0015662594232708216,
+      "learning_rate": 5.581250729753129e-08,
+      "loss": 4.324914189055562e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22581235.0,
+      "step": 6700
+    },
+    {
+      "entropy": 2.029346165060997,
+      "epoch": 4.845223044970201,
+      "grad_norm": 0.001764796208590269,
+      "learning_rate": 5.098672953502415e-08,
+      "loss": 3.3904198789969084e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22610108.0,
+      "step": 6710
+    },
+    {
+      "entropy": 2.1031780898571015,
+      "epoch": 4.852447173559689,
+      "grad_norm": 0.0018990206299349666,
+      "learning_rate": 4.637858653573468e-08,
+      "loss": 4.420768236741423e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22643481.0,
+      "step": 6720
+    },
+    {
+      "entropy": 2.064891684055328,
+      "epoch": 4.859671302149178,
+      "grad_norm": 0.0023969768080860376,
+      "learning_rate": 4.198817910283426e-08,
+      "loss": 4.414340655785054e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22672931.0,
+      "step": 6730
+    },
+    {
+      "entropy": 2.0357902348041534,
+      "epoch": 4.866895430738667,
+      "grad_norm": 0.003920447546988726,
+      "learning_rate": 3.781560327653089e-08,
+      "loss": 5.700000911019742e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22705766.0,
+      "step": 6740
+    },
+    {
+      "entropy": 2.039839392900467,
+      "epoch": 4.874119559328156,
+      "grad_norm": 0.00022526641259901226,
+      "learning_rate": 3.3860950331959794e-08,
+      "loss": 5.020995740778744e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22739415.0,
+      "step": 6750
+    },
+    {
+      "entropy": 2.04675834774971,
+      "epoch": 4.881343687917645,
+      "grad_norm": 0.01613064855337143,
+      "learning_rate": 3.0124306777195025e-08,
+      "loss": 5.47548639588058e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22770833.0,
+      "step": 6760
+    },
+    {
+      "entropy": 1.9878700971603394,
+      "epoch": 4.888567816507134,
+      "grad_norm": 0.0018846408929675817,
+      "learning_rate": 2.660575435134982e-08,
+      "loss": 2.85835936665535e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22808032.0,
+      "step": 6770
+    },
+    {
+      "entropy": 2.0023221403360365,
+      "epoch": 4.895791945096622,
+      "grad_norm": 0.002222092356532812,
+      "learning_rate": 2.3305370022793648e-08,
+      "loss": 0.00022039401810616254,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22841123.0,
+      "step": 6780
+    },
+    {
+      "entropy": 2.0070317834615707,
+      "epoch": 4.9030160736861115,
+      "grad_norm": 0.009242741391062737,
+      "learning_rate": 2.022322598746462e-08,
+      "loss": 8.193657849915325e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22876442.0,
+      "step": 6790
+    },
+    {
+      "entropy": 1.9670007824897766,
+      "epoch": 4.910240202275601,
+      "grad_norm": 0.002695423783734441,
+      "learning_rate": 1.7359389667296333e-08,
+      "loss": 2.583371242508292e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22908640.0,
+      "step": 6800
+    },
+    {
+      "entropy": 2.0583061963319778,
+      "epoch": 4.917464330865089,
+      "grad_norm": 0.0005823559476993978,
+      "learning_rate": 1.47139237087357e-08,
+      "loss": 3.0780312954448164e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22940536.0,
+      "step": 6810
+    },
+    {
+      "entropy": 2.0565018385648726,
+      "epoch": 4.924688459454578,
+      "grad_norm": 0.0013910674024373293,
+      "learning_rate": 1.2286885981376284e-08,
+      "loss": 5.821611266583204e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 22974965.0,
+      "step": 6820
+    },
+    {
+      "entropy": 2.0042783856391906,
+      "epoch": 4.931912588044067,
+      "grad_norm": 0.011966842226684093,
+      "learning_rate": 1.0078329576691526e-08,
+      "loss": 0.0006500833202153445,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23010525.0,
+      "step": 6830
+    },
+    {
+      "entropy": 2.040247473120689,
+      "epoch": 4.939136716633556,
+      "grad_norm": 0.0010858806781470776,
+      "learning_rate": 8.088302806875669e-09,
+      "loss": 5.4765178356319665e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23043869.0,
+      "step": 6840
+    },
+    {
+      "entropy": 1.992214022576809,
+      "epoch": 4.946360845223045,
+      "grad_norm": 0.01539916917681694,
+      "learning_rate": 6.316849203783504e-09,
+      "loss": 5.2837148541584614e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23083979.0,
+      "step": 6850
+    },
+    {
+      "entropy": 2.040625846385956,
+      "epoch": 4.953584973812534,
+      "grad_norm": 0.006538680754601955,
+      "learning_rate": 4.764007517978897e-09,
+      "loss": 4.6846986515447496e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23114925.0,
+      "step": 6860
+    },
+    {
+      "entropy": 2.0183849722146987,
+      "epoch": 4.960809102402023,
+      "grad_norm": 0.01923307776451111,
+      "learning_rate": 3.4298117178910294e-09,
+      "loss": 6.437179399654269e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23149219.0,
+      "step": 6870
+    },
+    {
+      "entropy": 2.0850543707609175,
+      "epoch": 4.968033230991511,
+      "grad_norm": 0.0006016053375788033,
+      "learning_rate": 2.3142909890661035e-09,
+      "loss": 4.915224853903055e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23180291.0,
+      "step": 6880
+    },
+    {
+      "entropy": 2.112953135371208,
+      "epoch": 4.975257359581001,
+      "grad_norm": 0.0011069178581237793,
+      "learning_rate": 1.4174697335289644e-09,
+      "loss": 3.848375054076314e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23211797.0,
+      "step": 6890
+    },
+    {
+      "entropy": 2.022666311264038,
+      "epoch": 4.98248148817049,
+      "grad_norm": 0.0002605836489237845,
+      "learning_rate": 7.393675692546343e-10,
+      "loss": 7.313743117265403e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23245939.0,
+      "step": 6900
+    },
+    {
+      "entropy": 2.052492305636406,
+      "epoch": 4.989705616759978,
+      "grad_norm": 0.0013098756317049265,
+      "learning_rate": 2.799993297353254e-10,
+      "loss": 2.9006070690229535e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23277936.0,
+      "step": 6910
+    },
+    {
+      "entropy": 2.017625370621681,
+      "epoch": 4.996929745349467,
+      "grad_norm": 0.0027501226868480444,
+      "learning_rate": 3.9375063654034654e-11,
+      "loss": 3.442804736550897e-05,
+      "mean_token_accuracy": 1.0,
+      "num_tokens": 23314979.0,
+      "step": 6920
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 1.8847647032144799,
+      "eval_loss": 1.1231932640075684,
+      "eval_mean_token_accuracy": 0.8734663190512821,
+      "eval_num_tokens": 23329700.0,
+      "eval_runtime": 1004.8237,
+      "eval_samples_per_second": 7.614,
+      "eval_steps_per_second": 0.952,
+      "step": 6925
+    },
+    {
+      "epoch": 5.0,
+      "step": 6925,
+      "total_flos": 1.3802440307700695e+18,
+      "train_loss": 0.02861043398070783,
+      "train_runtime": 25037.5703,
+      "train_samples_per_second": 2.211,
+      "train_steps_per_second": 0.277
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 6925,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3802440307700695e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}