diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -2,3118 +2,6468 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 10.0,
+  "epoch": 5.0,
   "eval_steps": 500,
-  "global_step": 310,
+  "global_step": 645,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "entropy": 1.2108377814292908,
-      "epoch": 0.03225806451612903,
-      "grad_norm": 0.6928424835205078,
+      "entropy": 1.1461779922246933,
+      "epoch": 0.007774538386783284,
+      "grad_norm": 0.2348785102367401,
       "learning_rate": 0.0002,
-      "loss": 2.6145451068878174,
-      "mean_token_accuracy": 0.5257208347320557,
-      "num_tokens": 77915.0,
+      "loss": 2.5822062492370605,
+      "mean_token_accuracy": 0.5106955841183662,
+      "num_tokens": 37728.0,
       "step": 1
     },
     {
-      "entropy": 1.2280026972293854,
-      "epoch": 0.06451612903225806,
-      "grad_norm": 0.40360888838768005,
+      "entropy": 1.2709465026855469,
+      "epoch": 0.015549076773566569,
+      "grad_norm": 0.19847582280635834,
       "learning_rate": 0.0002,
-      "loss": 2.12216854095459,
-      "mean_token_accuracy": 0.5751811861991882,
-      "num_tokens": 155609.0,
+      "loss": 2.256243944168091,
+      "mean_token_accuracy": 0.5385559350252151,
+      "num_tokens": 74823.0,
       "step": 2
     },
     {
-      "entropy": 1.3157195746898651,
-      "epoch": 0.0967741935483871,
-      "grad_norm": 0.37724417448043823,
+      "entropy": 1.457642063498497,
+      "epoch": 0.023323615160349854,
+      "grad_norm": 0.14325936138629913,
       "learning_rate": 0.0002,
-      "loss": 1.7939939498901367,
-      "mean_token_accuracy": 0.5949544161558151,
-      "num_tokens": 234747.0,
+      "loss": 1.8135013580322266,
+      "mean_token_accuracy": 0.5658686235547066,
+      "num_tokens": 112223.0,
       "step": 3
     },
     {
-      "entropy": 1.3692472279071808,
-      "epoch": 0.12903225806451613,
-      "grad_norm": 0.4072861969470978,
+      "entropy": 1.4453874379396439,
+      "epoch": 0.031098153547133137,
+      "grad_norm": 0.11537832021713257,
       "learning_rate": 0.0002,
-      "loss": 1.5643707513809204,
-      "mean_token_accuracy": 0.6154418289661407,
-      "num_tokens": 314203.0,
+      "loss": 1.4907238483428955,
+      "mean_token_accuracy": 0.6095475852489471,
+      "num_tokens": 149646.0,
       "step": 4
     },
     {
-      "entropy": 1.3363037109375,
-      "epoch": 0.16129032258064516,
-      "grad_norm": 0.25536584854125977,
+      "entropy": 1.4222392737865448,
+      "epoch": 0.038872691933916424,
+      "grad_norm": 0.14347851276397705,
       "learning_rate": 0.0002,
-      "loss": 1.3993436098098755,
-      "mean_token_accuracy": 0.6389602273702621,
-      "num_tokens": 393638.0,
+      "loss": 1.3716599941253662,
+      "mean_token_accuracy": 0.614239789545536,
+      "num_tokens": 186807.0,
       "step": 5
     },
     {
-      "entropy": 1.3661861419677734,
-      "epoch": 0.1935483870967742,
-      "grad_norm": 0.16949856281280518,
+      "entropy": 1.3676332831382751,
+      "epoch": 0.04664723032069971,
+      "grad_norm": 0.07458607107400894,
       "learning_rate": 0.0002,
-      "loss": 1.2982456684112549,
-      "mean_token_accuracy": 0.649403303861618,
-      "num_tokens": 472178.0,
+      "loss": 1.2618626356124878,
+      "mean_token_accuracy": 0.6355780363082886,
+      "num_tokens": 224243.0,
       "step": 6
     },
     {
-      "entropy": 1.3246012330055237,
-      "epoch": 0.22580645161290322,
-      "grad_norm": 0.1267392337322235,
+      "entropy": 1.2903779596090317,
+      "epoch": 0.05442176870748299,
+      "grad_norm": 0.050369806587696075,
       "learning_rate": 0.0002,
-      "loss": 1.2008121013641357,
-      "mean_token_accuracy": 0.6618895977735519,
-      "num_tokens": 550442.0,
+      "loss": 1.1705503463745117,
+      "mean_token_accuracy": 0.6446279659867287,
+      "num_tokens": 261678.0,
       "step": 7
     },
     {
-      "entropy": 1.2534887194633484,
-      "epoch": 0.25806451612903225,
-      "grad_norm": 0.12660343945026398,
+      "entropy": 1.231130599975586,
+      "epoch": 0.062196307094266275,
+      "grad_norm": 0.05470636859536171,
       "learning_rate": 0.0002,
-      "loss": 1.0963741540908813,
-      "mean_token_accuracy": 0.6759665459394455,
-      "num_tokens": 631057.0,
+      "loss": 1.0978062152862549,
+      "mean_token_accuracy": 0.6580947414040565,
+      "num_tokens": 298703.0,
       "step": 8
     },
     {
-      "entropy": 1.2272652387619019,
-      "epoch": 0.2903225806451613,
-      "grad_norm": 0.1357613056898117,
+      "entropy": 1.1396304965019226,
+      "epoch": 0.06997084548104957,
+      "grad_norm": 0.05549474433064461,
       "learning_rate": 0.0002,
-      "loss": 1.0560389757156372,
-      "mean_token_accuracy": 0.6746651232242584,
-      "num_tokens": 709785.0,
+      "loss": 1.026996374130249,
+      "mean_token_accuracy": 0.6708663776516914,
+      "num_tokens": 336240.0,
       "step": 9
     },
     {
-      "entropy": 1.1543474793434143,
-      "epoch": 0.3225806451612903,
-      "grad_norm": 0.1262923926115036,
+      "entropy": 1.0528622567653656,
+      "epoch": 0.07774538386783285,
+      "grad_norm": 0.05926097184419632,
       "learning_rate": 0.0002,
-      "loss": 0.9836021661758423,
-      "mean_token_accuracy": 0.693942204117775,
-      "num_tokens": 788705.0,
+      "loss": 0.9768505096435547,
+      "mean_token_accuracy": 0.6747338622808456,
+      "num_tokens": 373808.0,
       "step": 10
     },
     {
-      "entropy": 1.0659774839878082,
-      "epoch": 0.3548387096774194,
-      "grad_norm": 0.10476168245077133,
+      "entropy": 0.978929877281189,
+      "epoch": 0.08551992225461613,
+      "grad_norm": 0.08427942544221878,
       "learning_rate": 0.0002,
-      "loss": 0.9247925877571106,
-      "mean_token_accuracy": 0.7010035812854767,
-      "num_tokens": 866898.0,
+      "loss": 0.9218605756759644,
+      "mean_token_accuracy": 0.6868897452950478,
+      "num_tokens": 410822.0,
       "step": 11
     },
     {
-      "entropy": 0.9727843105792999,
-      "epoch": 0.3870967741935484,
-      "grad_norm": 0.09436453878879547,
+      "entropy": 0.9272410869598389,
+      "epoch": 0.09329446064139942,
+      "grad_norm": 0.0502631776034832,
       "learning_rate": 0.0002,
-      "loss": 0.8839961290359497,
-      "mean_token_accuracy": 0.7076074182987213,
-      "num_tokens": 945336.0,
+      "loss": 0.8876349925994873,
+      "mean_token_accuracy": 0.688661016523838,
+      "num_tokens": 447975.0,
       "step": 12
     },
     {
-      "entropy": 0.8906200975179672,
-      "epoch": 0.41935483870967744,
-      "grad_norm": 0.0898677408695221,
+      "entropy": 0.860072985291481,
+      "epoch": 0.1010689990281827,
+      "grad_norm": 0.04337044060230255,
       "learning_rate": 0.0002,
-      "loss": 0.8381521105766296,
-      "mean_token_accuracy": 0.7165039479732513,
-      "num_tokens": 1023470.0,
+      "loss": 0.8350162506103516,
+      "mean_token_accuracy": 0.7001973986625671,
+      "num_tokens": 485476.0,
       "step": 13
     },
     {
-      "entropy": 0.8297237753868103,
-      "epoch": 0.45161290322580644,
-      "grad_norm": 0.09061617404222488,
+      "entropy": 0.7882575914263725,
+      "epoch": 0.10884353741496598,
+      "grad_norm": 0.06162057816982269,
       "learning_rate": 0.0002,
-      "loss": 0.8075975179672241,
-      "mean_token_accuracy": 0.7183874249458313,
-      "num_tokens": 1102366.0,
+      "loss": 0.8057988882064819,
+      "mean_token_accuracy": 0.7050592303276062,
+      "num_tokens": 522792.0,
       "step": 14
     },
     {
-      "entropy": 0.7862683832645416,
-      "epoch": 0.4838709677419355,
-      "grad_norm": 0.09677388519048691,
+      "entropy": 0.7824503108859062,
+      "epoch": 0.11661807580174927,
+      "grad_norm": 0.03932056203484535,
       "learning_rate": 0.0002,
-      "loss": 0.7915093898773193,
-      "mean_token_accuracy": 0.7166366279125214,
-      "num_tokens": 1181463.0,
+      "loss": 0.7716900110244751,
+      "mean_token_accuracy": 0.7123773470520973,
+      "num_tokens": 559845.0,
       "step": 15
     },
     {
-      "entropy": 0.7366400808095932,
-      "epoch": 0.5161290322580645,
-      "grad_norm": 0.08824242651462555,
+      "entropy": 0.7571921274065971,
+      "epoch": 0.12439261418853255,
+      "grad_norm": 0.03887049853801727,
       "learning_rate": 0.0002,
-      "loss": 0.7438744306564331,
-      "mean_token_accuracy": 0.7302631139755249,
-      "num_tokens": 1260006.0,
+      "loss": 0.7427914142608643,
+      "mean_token_accuracy": 0.721781499683857,
+      "num_tokens": 596752.0,
       "step": 16
     },
     {
-      "entropy": 0.7312893569469452,
-      "epoch": 0.5483870967741935,
-      "grad_norm": 0.07547659426927567,
+      "entropy": 0.7445609271526337,
+      "epoch": 0.13216715257531583,
+      "grad_norm": 0.037259574979543686,
       "learning_rate": 0.0002,
-      "loss": 0.7313656806945801,
-      "mean_token_accuracy": 0.7311705946922302,
-      "num_tokens": 1338935.0,
+      "loss": 0.7245944738388062,
+      "mean_token_accuracy": 0.7208689078688622,
+      "num_tokens": 634143.0,
       "step": 17
     },
     {
-      "entropy": 0.7072197049856186,
-      "epoch": 0.5806451612903226,
-      "grad_norm": 0.07940181344747543,
+      "entropy": 0.7450662180781364,
+      "epoch": 0.13994169096209913,
+      "grad_norm": 0.03612779080867767,
       "learning_rate": 0.0002,
-      "loss": 0.6973146796226501,
-      "mean_token_accuracy": 0.7380757033824921,
-      "num_tokens": 1418014.0,
+      "loss": 0.7255409955978394,
+      "mean_token_accuracy": 0.7214401289820671,
+      "num_tokens": 671585.0,
       "step": 18
     },
     {
-      "entropy": 0.7047975361347198,
-      "epoch": 0.6129032258064516,
-      "grad_norm": 0.08228685706853867,
+      "entropy": 0.7271910980343819,
+      "epoch": 0.1477162293488824,
+      "grad_norm": 0.03141465783119202,
       "learning_rate": 0.0002,
-      "loss": 0.6983991861343384,
-      "mean_token_accuracy": 0.7352693974971771,
-      "num_tokens": 1497045.0,
+      "loss": 0.7076310515403748,
+      "mean_token_accuracy": 0.7255471646785736,
+      "num_tokens": 708756.0,
       "step": 19
     },
     {
-      "entropy": 0.6795883923768997,
-      "epoch": 0.6451612903225806,
-      "grad_norm": 0.0756707489490509,
+      "entropy": 0.7060252502560616,
+      "epoch": 0.1554907677356657,
+      "grad_norm": 0.035933561623096466,
       "learning_rate": 0.0002,
-      "loss": 0.6704490184783936,
-      "mean_token_accuracy": 0.7453291118144989,
-      "num_tokens": 1576440.0,
+      "loss": 0.6859346628189087,
+      "mean_token_accuracy": 0.732830710709095,
+      "num_tokens": 746139.0,
       "step": 20
     },
     {
-      "entropy": 0.6742499768733978,
-      "epoch": 0.6774193548387096,
-      "grad_norm": 0.06836266070604324,
+      "entropy": 0.6829683035612106,
+      "epoch": 0.16326530612244897,
+      "grad_norm": 0.03822736069560051,
       "learning_rate": 0.0002,
-      "loss": 0.6623807549476624,
-      "mean_token_accuracy": 0.7448357492685318,
-      "num_tokens": 1656085.0,
+      "loss": 0.6772962212562561,
+      "mean_token_accuracy": 0.7338771298527718,
+      "num_tokens": 783417.0,
       "step": 21
     },
     {
-      "entropy": 0.6770815700292587,
-      "epoch": 0.7096774193548387,
-      "grad_norm": 0.0641896054148674,
+      "entropy": 0.6782366037368774,
+      "epoch": 0.17103984450923226,
+      "grad_norm": 0.031850460916757584,
       "learning_rate": 0.0002,
-      "loss": 0.6653863787651062,
-      "mean_token_accuracy": 0.742030069231987,
-      "num_tokens": 1733785.0,
+      "loss": 0.6737133264541626,
+      "mean_token_accuracy": 0.7363682463765144,
+      "num_tokens": 820745.0,
       "step": 22
     },
     {
-      "entropy": 0.6622436791658401,
-      "epoch": 0.7419354838709677,
-      "grad_norm": 0.061193544417619705,
+      "entropy": 0.6792467907071114,
+      "epoch": 0.17881438289601556,
+      "grad_norm": 0.031343962997198105,
       "learning_rate": 0.0002,
-      "loss": 0.6471956968307495,
-      "mean_token_accuracy": 0.7478197664022446,
-      "num_tokens": 1811732.0,
+      "loss": 0.6833145618438721,
+      "mean_token_accuracy": 0.7314189001917839,
+      "num_tokens": 858477.0,
       "step": 23
     },
     {
-      "entropy": 0.6454922705888748,
-      "epoch": 0.7741935483870968,
-      "grad_norm": 0.05598059669137001,
+      "entropy": 0.6493787094950676,
+      "epoch": 0.18658892128279883,
+      "grad_norm": 0.028227699920535088,
       "learning_rate": 0.0002,
-      "loss": 0.6315531134605408,
-      "mean_token_accuracy": 0.7541234940290451,
-      "num_tokens": 1890421.0,
+      "loss": 0.6512086987495422,
+      "mean_token_accuracy": 0.7448033019900322,
+      "num_tokens": 895754.0,
       "step": 24
     },
     {
-      "entropy": 0.6487089246511459,
-      "epoch": 0.8064516129032258,
-      "grad_norm": 0.0565122552216053,
+      "entropy": 0.6527747735381126,
+      "epoch": 0.19436345966958213,
+      "grad_norm": 0.02940857782959938,
       "learning_rate": 0.0002,
-      "loss": 0.6389895677566528,
-      "mean_token_accuracy": 0.7514772266149521,
-      "num_tokens": 1967043.0,
+      "loss": 0.6458977460861206,
+      "mean_token_accuracy": 0.7454173043370247,
+      "num_tokens": 933200.0,
       "step": 25
     },
     {
-      "entropy": 0.6330023109912872,
-      "epoch": 0.8387096774193549,
-      "grad_norm": 0.06039460375905037,
+      "entropy": 0.6647381633520126,
+      "epoch": 0.2021379980563654,
+      "grad_norm": 0.03015323169529438,
       "learning_rate": 0.0002,
-      "loss": 0.6244106292724609,
-      "mean_token_accuracy": 0.7557308524847031,
-      "num_tokens": 2045932.0,
+      "loss": 0.6591448783874512,
+      "mean_token_accuracy": 0.7374408692121506,
+      "num_tokens": 970838.0,
       "step": 26
     },
     {
-      "entropy": 0.624711200594902,
-      "epoch": 0.8709677419354839,
-      "grad_norm": 0.05815982446074486,
+      "entropy": 0.6438521891832352,
+      "epoch": 0.2099125364431487,
+      "grad_norm": 0.035360876470804214,
       "learning_rate": 0.0002,
-      "loss": 0.6193004846572876,
-      "mean_token_accuracy": 0.7580623775720596,
-      "num_tokens": 2125352.0,
+      "loss": 0.6528849601745605,
+      "mean_token_accuracy": 0.7403199598193169,
+      "num_tokens": 1008094.0,
       "step": 27
     },
     {
-      "entropy": 0.6006695479154587,
-      "epoch": 0.9032258064516129,
-      "grad_norm": 0.06260675936937332,
+      "entropy": 0.6399626433849335,
+      "epoch": 0.21768707482993196,
+      "grad_norm": 0.02637004852294922,
       "learning_rate": 0.0002,
-      "loss": 0.6044747829437256,
-      "mean_token_accuracy": 0.7639114111661911,
-      "num_tokens": 2203090.0,
+      "loss": 0.6530839204788208,
+      "mean_token_accuracy": 0.7398090362548828,
+      "num_tokens": 1045463.0,
       "step": 28
     },
     {
-      "entropy": 0.5943089425563812,
-      "epoch": 0.9354838709677419,
-      "grad_norm": 0.060202307999134064,
+      "entropy": 0.6436624526977539,
+      "epoch": 0.22546161321671526,
+      "grad_norm": 0.02592352218925953,
       "learning_rate": 0.0002,
-      "loss": 0.5964701771736145,
-      "mean_token_accuracy": 0.7653262317180634,
-      "num_tokens": 2281537.0,
+      "loss": 0.6487458944320679,
+      "mean_token_accuracy": 0.7412822172045708,
+      "num_tokens": 1083139.0,
       "step": 29
     },
     {
-      "entropy": 0.5986111164093018,
-      "epoch": 0.967741935483871,
-      "grad_norm": 0.0541241429746151,
+      "entropy": 0.653114914894104,
+      "epoch": 0.23323615160349853,
+      "grad_norm": 0.018322035670280457,
       "learning_rate": 0.0002,
-      "loss": 0.5991954803466797,
-      "mean_token_accuracy": 0.7634078860282898,
-      "num_tokens": 2360516.0,
+      "loss": 0.6511994004249573,
+      "mean_token_accuracy": 0.7411110699176788,
+      "num_tokens": 1120783.0,
       "step": 30
     },
     {
-      "entropy": 0.5967210531234741,
-      "epoch": 1.0,
-      "grad_norm": 0.05269192159175873,
+      "entropy": 0.642485961318016,
+      "epoch": 0.24101068999028183,
+      "grad_norm": 0.021689819172024727,
       "learning_rate": 0.0002,
-      "loss": 0.5975545048713684,
-      "mean_token_accuracy": 0.7649656534194946,
-      "num_tokens": 2438564.0,
+      "loss": 0.644116222858429,
+      "mean_token_accuracy": 0.7440010756254196,
+      "num_tokens": 1158207.0,
       "step": 31
     },
     {
-      "entropy": 0.5890886783599854,
-      "epoch": 1.032258064516129,
-      "grad_norm": 0.05449746176600456,
+      "entropy": 0.6343450918793678,
+      "epoch": 0.2487852283770651,
+      "grad_norm": 0.02071012370288372,
       "learning_rate": 0.0002,
-      "loss": 0.5800559520721436,
-      "mean_token_accuracy": 0.7717925757169724,
-      "num_tokens": 2516742.0,
+      "loss": 0.6319076418876648,
+      "mean_token_accuracy": 0.7487388849258423,
+      "num_tokens": 1195845.0,
       "step": 32
     },
     {
-      "entropy": 0.5956283509731293,
-      "epoch": 1.064516129032258,
-      "grad_norm": 0.05351584777235985,
+      "entropy": 0.6410091295838356,
+      "epoch": 0.2565597667638484,
+      "grad_norm": 0.025183040648698807,
       "learning_rate": 0.0002,
-      "loss": 0.5879819393157959,
-      "mean_token_accuracy": 0.7676646113395691,
-      "num_tokens": 2595672.0,
+      "loss": 0.6358401775360107,
+      "mean_token_accuracy": 0.7465788945555687,
+      "num_tokens": 1233366.0,
       "step": 33
     },
     {
-      "entropy": 0.5949074476957321,
-      "epoch": 1.096774193548387,
-      "grad_norm": 0.05134779214859009,
+      "entropy": 0.6519959643483162,
+      "epoch": 0.26433430515063167,
+      "grad_norm": 0.017367947846651077,
       "learning_rate": 0.0002,
-      "loss": 0.593163251876831,
-      "mean_token_accuracy": 0.7667708247900009,
-      "num_tokens": 2674992.0,
+      "loss": 0.6433607339859009,
+      "mean_token_accuracy": 0.7429918646812439,
+      "num_tokens": 1270710.0,
       "step": 34
     },
     {
-      "entropy": 0.5895005166530609,
-      "epoch": 1.129032258064516,
-      "grad_norm": 0.04475667327642441,
+      "entropy": 0.6490929946303368,
+      "epoch": 0.272108843537415,
+      "grad_norm": 0.023137487471103668,
       "learning_rate": 0.0002,
-      "loss": 0.5876166820526123,
-      "mean_token_accuracy": 0.7677939385175705,
-      "num_tokens": 2754315.0,
+      "loss": 0.6468774676322937,
+      "mean_token_accuracy": 0.7406889051198959,
+      "num_tokens": 1308095.0,
       "step": 35
     },
     {
-      "entropy": 0.5794888734817505,
-      "epoch": 1.1612903225806452,
-      "grad_norm": 0.03956792876124382,
+      "entropy": 0.6438097059726715,
+      "epoch": 0.27988338192419826,
+      "grad_norm": 0.016993502154946327,
       "learning_rate": 0.0002,
-      "loss": 0.5770071744918823,
-      "mean_token_accuracy": 0.7725034654140472,
-      "num_tokens": 2834196.0,
+      "loss": 0.6424656510353088,
+      "mean_token_accuracy": 0.7413171753287315,
+      "num_tokens": 1345646.0,
       "step": 36
     },
     {
-      "entropy": 0.5960649400949478,
-      "epoch": 1.1935483870967742,
-      "grad_norm": 0.04742363467812538,
+      "entropy": 0.6314259544014931,
+      "epoch": 0.28765792031098153,
+      "grad_norm": 0.016146596521139145,
       "learning_rate": 0.0002,
-      "loss": 0.5952359437942505,
-      "mean_token_accuracy": 0.7638276815414429,
-      "num_tokens": 2912297.0,
+      "loss": 0.6370364427566528,
+      "mean_token_accuracy": 0.7465247958898544,
+      "num_tokens": 1382928.0,
       "step": 37
     },
     {
-      "entropy": 0.588940218091011,
-      "epoch": 1.2258064516129032,
-      "grad_norm": 0.04095599055290222,
+      "entropy": 0.6270746886730194,
+      "epoch": 0.2954324586977648,
+      "grad_norm": 0.019162772223353386,
       "learning_rate": 0.0002,
-      "loss": 0.5888320207595825,
-      "mean_token_accuracy": 0.7661104947328568,
-      "num_tokens": 2989829.0,
+      "loss": 0.6330348253250122,
+      "mean_token_accuracy": 0.747043713927269,
+      "num_tokens": 1420089.0,
       "step": 38
     },
     {
-      "entropy": 0.5950354784727097,
-      "epoch": 1.2580645161290323,
-      "grad_norm": 0.041268933564424515,
+      "entropy": 0.6293222159147263,
+      "epoch": 0.3032069970845481,
+      "grad_norm": 0.01702946238219738,
       "learning_rate": 0.0002,
-      "loss": 0.5957392454147339,
-      "mean_token_accuracy": 0.762763723731041,
-      "num_tokens": 3067135.0,
+      "loss": 0.6317635178565979,
+      "mean_token_accuracy": 0.7488930821418762,
+      "num_tokens": 1457216.0,
       "step": 39
     },
     {
-      "entropy": 0.5728808343410492,
-      "epoch": 1.2903225806451613,
-      "grad_norm": 0.030816541984677315,
+      "entropy": 0.6111311987042427,
+      "epoch": 0.3109815354713314,
+      "grad_norm": 0.015137894079089165,
       "learning_rate": 0.0002,
-      "loss": 0.5696955919265747,
-      "mean_token_accuracy": 0.7730816453695297,
-      "num_tokens": 3147099.0,
+      "loss": 0.6177330017089844,
+      "mean_token_accuracy": 0.753462478518486,
+      "num_tokens": 1494415.0,
       "step": 40
     },
     {
-      "entropy": 0.5765672028064728,
-      "epoch": 1.3225806451612903,
-      "grad_norm": 0.04333742335438728,
+      "entropy": 0.6119553595781326,
+      "epoch": 0.31875607385811466,
+      "grad_norm": 0.014271042309701443,
       "learning_rate": 0.0002,
-      "loss": 0.577884316444397,
-      "mean_token_accuracy": 0.7716145664453506,
-      "num_tokens": 3225549.0,
+      "loss": 0.6184912919998169,
+      "mean_token_accuracy": 0.7530097812414169,
+      "num_tokens": 1531665.0,
       "step": 41
     },
     {
-      "entropy": 0.573496475815773,
-      "epoch": 1.3548387096774195,
-      "grad_norm": 0.03675653412938118,
+      "entropy": 0.627587340772152,
+      "epoch": 0.32653061224489793,
+      "grad_norm": 0.01676957495510578,
       "learning_rate": 0.0002,
-      "loss": 0.5731475949287415,
-      "mean_token_accuracy": 0.7725466936826706,
-      "num_tokens": 3302551.0,
+      "loss": 0.628797173500061,
+      "mean_token_accuracy": 0.7489041686058044,
+      "num_tokens": 1568905.0,
       "step": 42
     },
     {
-      "entropy": 0.5770469456911087,
-      "epoch": 1.3870967741935485,
-      "grad_norm": 0.0320008248090744,
+      "entropy": 0.6380143910646439,
+      "epoch": 0.33430515063168126,
+      "grad_norm": 0.013997296802699566,
       "learning_rate": 0.0002,
-      "loss": 0.5700103044509888,
-      "mean_token_accuracy": 0.7736899256706238,
-      "num_tokens": 3381541.0,
+      "loss": 0.6401110291481018,
+      "mean_token_accuracy": 0.7437692731618881,
+      "num_tokens": 1606022.0,
       "step": 43
     },
     {
-      "entropy": 0.5819686949253082,
-      "epoch": 1.4193548387096775,
-      "grad_norm": 0.03123253397643566,
+      "entropy": 0.6236186027526855,
+      "epoch": 0.34207968901846453,
+      "grad_norm": 0.013737027533352375,
       "learning_rate": 0.0002,
-      "loss": 0.5789672136306763,
-      "mean_token_accuracy": 0.7693924754858017,
-      "num_tokens": 3460837.0,
+      "loss": 0.622249960899353,
+      "mean_token_accuracy": 0.7506565973162651,
+      "num_tokens": 1643817.0,
       "step": 44
     },
     {
-      "entropy": 0.580963209271431,
-      "epoch": 1.4516129032258065,
-      "grad_norm": 0.04066276177763939,
+      "entropy": 0.6351634711027145,
+      "epoch": 0.3498542274052478,
+      "grad_norm": 0.013750840909779072,
       "learning_rate": 0.0002,
-      "loss": 0.5793372988700867,
-      "mean_token_accuracy": 0.7687239795923233,
-      "num_tokens": 3538644.0,
+      "loss": 0.6342694163322449,
+      "mean_token_accuracy": 0.7469271644949913,
+      "num_tokens": 1681417.0,
       "step": 45
     },
     {
-      "entropy": 0.5791344493627548,
-      "epoch": 1.4838709677419355,
-      "grad_norm": 0.031351737678050995,
+      "entropy": 0.6226040348410606,
+      "epoch": 0.3576287657920311,
+      "grad_norm": 0.015259931795299053,
       "learning_rate": 0.0002,
-      "loss": 0.5801466107368469,
-      "mean_token_accuracy": 0.7692239880561829,
-      "num_tokens": 3616245.0,
+      "loss": 0.6237019896507263,
+      "mean_token_accuracy": 0.7505070269107819,
+      "num_tokens": 1718574.0,
       "step": 46
     },
     {
-      "entropy": 0.5724106431007385,
-      "epoch": 1.5161290322580645,
-      "grad_norm": 0.033206913620233536,
+      "entropy": 0.6210184469819069,
+      "epoch": 0.3654033041788144,
+      "grad_norm": 0.013225192204117775,
       "learning_rate": 0.0002,
-      "loss": 0.5713744163513184,
-      "mean_token_accuracy": 0.7738502323627472,
-      "num_tokens": 3695574.0,
+      "loss": 0.6187986135482788,
+      "mean_token_accuracy": 0.7520730867981911,
+      "num_tokens": 1756095.0,
       "step": 47
     },
     {
-      "entropy": 0.5739834606647491,
-      "epoch": 1.5483870967741935,
-      "grad_norm": 0.03075285442173481,
+      "entropy": 0.6205598041415215,
+      "epoch": 0.37317784256559766,
+      "grad_norm": 0.012887167744338512,
       "learning_rate": 0.0002,
-      "loss": 0.5729138255119324,
-      "mean_token_accuracy": 0.7724349647760391,
-      "num_tokens": 3774001.0,
+      "loss": 0.6181526780128479,
+      "mean_token_accuracy": 0.7511478438973427,
+      "num_tokens": 1793392.0,
       "step": 48
     },
     {
-      "entropy": 0.5661155879497528,
-      "epoch": 1.5806451612903225,
-      "grad_norm": 0.026813488453626633,
+      "entropy": 0.629549115896225,
+      "epoch": 0.38095238095238093,
+      "grad_norm": 0.013227855786681175,
       "learning_rate": 0.0002,
-      "loss": 0.5657498240470886,
-      "mean_token_accuracy": 0.7752608209848404,
-      "num_tokens": 3852374.0,
+      "loss": 0.6304742097854614,
+      "mean_token_accuracy": 0.7472723796963692,
+      "num_tokens": 1830606.0,
       "step": 49
     },
     {
-      "entropy": 0.5684164315462112,
-      "epoch": 1.6129032258064515,
-      "grad_norm": 0.0284164696931839,
+      "entropy": 0.6296898201107979,
+      "epoch": 0.38872691933916426,
+      "grad_norm": 0.015021155588328838,
       "learning_rate": 0.0002,
-      "loss": 0.5668526887893677,
-      "mean_token_accuracy": 0.774157777428627,
-      "num_tokens": 3930849.0,
+      "loss": 0.6288135647773743,
+      "mean_token_accuracy": 0.7465364933013916,
+      "num_tokens": 1868232.0,
       "step": 50
     },
     {
-      "entropy": 0.5668631792068481,
-      "epoch": 1.6451612903225805,
-      "grad_norm": 0.029652748256921768,
+      "entropy": 0.6357625275850296,
+      "epoch": 0.3965014577259475,
+      "grad_norm": 0.011794226244091988,
       "learning_rate": 0.0002,
-      "loss": 0.5632265210151672,
-      "mean_token_accuracy": 0.7760342657566071,
-      "num_tokens": 4010738.0,
+      "loss": 0.6371973156929016,
+      "mean_token_accuracy": 0.7450965940952301,
+      "num_tokens": 1905598.0,
       "step": 51
     },
     {
-      "entropy": 0.5639189630746841,
-      "epoch": 1.6774193548387095,
-      "grad_norm": 0.02636871486902237,
+      "entropy": 0.6295278668403625,
+      "epoch": 0.4042759961127308,
+      "grad_norm": 0.012368254363536835,
       "learning_rate": 0.0002,
-      "loss": 0.56150221824646,
-      "mean_token_accuracy": 0.7754913568496704,
-      "num_tokens": 4089609.0,
+      "loss": 0.6367925405502319,
+      "mean_token_accuracy": 0.7448450028896332,
+      "num_tokens": 1942785.0,
       "step": 52
     },
     {
-      "entropy": 0.5659375041723251,
-      "epoch": 1.7096774193548387,
-      "grad_norm": 0.030338820070028305,
+      "entropy": 0.6321664452552795,
+      "epoch": 0.41205053449951407,
+      "grad_norm": 0.011660202406346798,
       "learning_rate": 0.0002,
-      "loss": 0.5647764205932617,
-      "mean_token_accuracy": 0.7749795466661453,
-      "num_tokens": 4169556.0,
+      "loss": 0.6337926983833313,
+      "mean_token_accuracy": 0.7433984354138374,
+      "num_tokens": 1980646.0,
       "step": 53
     },
     {
-      "entropy": 0.5694165378808975,
-      "epoch": 1.7419354838709677,
-      "grad_norm": 0.02605626732110977,
+      "entropy": 0.6304126009345055,
+      "epoch": 0.4198250728862974,
+      "grad_norm": 0.012000083923339844,
       "learning_rate": 0.0002,
-      "loss": 0.5678636431694031,
-      "mean_token_accuracy": 0.7740114033222198,
-      "num_tokens": 4247901.0,
+      "loss": 0.6315821409225464,
+      "mean_token_accuracy": 0.7455745488405228,
+      "num_tokens": 2018148.0,
       "step": 54
     },
     {
-      "entropy": 0.5724839717149734,
-      "epoch": 1.7741935483870968,
-      "grad_norm": 0.023751629516482353,
+      "entropy": 0.6312398687005043,
+      "epoch": 0.42759961127308066,
+      "grad_norm": 0.012478172779083252,
       "learning_rate": 0.0002,
-      "loss": 0.570723295211792,
-      "mean_token_accuracy": 0.7715318500995636,
-      "num_tokens": 4325711.0,
+      "loss": 0.6293996572494507,
+      "mean_token_accuracy": 0.7464256510138512,
+      "num_tokens": 2055123.0,
       "step": 55
     },
     {
-      "entropy": 0.5751885026693344,
-      "epoch": 1.8064516129032258,
-      "grad_norm": 0.029801664873957634,
+      "entropy": 0.6144075095653534,
+      "epoch": 0.43537414965986393,
+      "grad_norm": 0.010203132405877113,
       "learning_rate": 0.0002,
-      "loss": 0.5713039636611938,
-      "mean_token_accuracy": 0.7714657932519913,
-      "num_tokens": 4404074.0,
+      "loss": 0.6101412773132324,
+      "mean_token_accuracy": 0.7551510035991669,
+      "num_tokens": 2092759.0,
       "step": 56
     },
     {
-      "entropy": 0.5683879405260086,
-      "epoch": 1.838709677419355,
-      "grad_norm": 0.024153606966137886,
+      "entropy": 0.6334135234355927,
+      "epoch": 0.44314868804664725,
+      "grad_norm": 0.012491037137806416,
       "learning_rate": 0.0002,
-      "loss": 0.5647911429405212,
-      "mean_token_accuracy": 0.7757914811372757,
-      "num_tokens": 4484176.0,
+      "loss": 0.6366456747055054,
+      "mean_token_accuracy": 0.7441563606262207,
+      "num_tokens": 2130468.0,
       "step": 57
     },
     {
-      "entropy": 0.5598143339157104,
-      "epoch": 1.870967741935484,
-      "grad_norm": 0.02074209600687027,
+      "entropy": 0.6204067915678024,
+      "epoch": 0.4509232264334305,
+      "grad_norm": 0.012662500143051147,
       "learning_rate": 0.0002,
-      "loss": 0.5574623942375183,
-      "mean_token_accuracy": 0.7780470103025436,
-      "num_tokens": 4563496.0,
+      "loss": 0.6225067377090454,
+      "mean_token_accuracy": 0.7498065009713173,
+      "num_tokens": 2167896.0,
       "step": 58
     },
     {
-      "entropy": 0.5672164112329483,
-      "epoch": 1.903225806451613,
-      "grad_norm": 0.024077970534563065,
+      "entropy": 0.626205176115036,
+      "epoch": 0.4586977648202138,
+      "grad_norm": 0.011919913813471794,
       "learning_rate": 0.0002,
-      "loss": 0.5635547041893005,
-      "mean_token_accuracy": 0.7746394872665405,
-      "num_tokens": 4642865.0,
+      "loss": 0.6297718286514282,
+      "mean_token_accuracy": 0.7457590103149414,
+      "num_tokens": 2204991.0,
       "step": 59
     },
     {
-      "entropy": 0.5679460912942886,
-      "epoch": 1.935483870967742,
-      "grad_norm": 0.02324177511036396,
+      "entropy": 0.6314920708537102,
+      "epoch": 0.46647230320699706,
+      "grad_norm": 0.011844294145703316,
       "learning_rate": 0.0002,
-      "loss": 0.5665385723114014,
-      "mean_token_accuracy": 0.7735758572816849,
-      "num_tokens": 4720857.0,
+      "loss": 0.6283063888549805,
+      "mean_token_accuracy": 0.7473507225513458,
+      "num_tokens": 2242421.0,
       "step": 60
     },
     {
-      "entropy": 0.5664634257555008,
-      "epoch": 1.967741935483871,
-      "grad_norm": 0.020476961508393288,
+      "entropy": 0.629265584051609,
+      "epoch": 0.4742468415937804,
+      "grad_norm": 0.012998970225453377,
       "learning_rate": 0.0002,
-      "loss": 0.5626745820045471,
-      "mean_token_accuracy": 0.7753059267997742,
-      "num_tokens": 4799418.0,
+      "loss": 0.6303939819335938,
+      "mean_token_accuracy": 0.7454146966338158,
+      "num_tokens": 2279756.0,
       "step": 61
     },
     {
-      "entropy": 0.5632900893688202,
-      "epoch": 2.0,
-      "grad_norm": 0.022128164768218994,
+      "entropy": 0.6266416981816292,
+      "epoch": 0.48202137998056366,
+      "grad_norm": 0.010719172656536102,
       "learning_rate": 0.0002,
-      "loss": 0.5620392560958862,
-      "mean_token_accuracy": 0.7761431336402893,
-      "num_tokens": 4877822.0,
+      "loss": 0.6256165504455566,
+      "mean_token_accuracy": 0.7477178424596786,
+      "num_tokens": 2317119.0,
       "step": 62
     },
     {
-      "entropy": 0.5647830814123154,
-      "epoch": 2.032258064516129,
-      "grad_norm": 0.022481486201286316,
+      "entropy": 0.6031938791275024,
+      "epoch": 0.4897959183673469,
+      "grad_norm": 0.011259288527071476,
       "learning_rate": 0.0002,
-      "loss": 0.5607410669326782,
-      "mean_token_accuracy": 0.775324359536171,
-      "num_tokens": 4956332.0,
+      "loss": 0.5979391932487488,
+      "mean_token_accuracy": 0.7593775168061256,
+      "num_tokens": 2354598.0,
       "step": 63
     },
     {
-      "entropy": 0.5720101296901703,
-      "epoch": 2.064516129032258,
-      "grad_norm": 0.019787590950727463,
+      "entropy": 0.6359111741185188,
+      "epoch": 0.4975704567541302,
+      "grad_norm": 0.01182929240167141,
       "learning_rate": 0.0002,
-      "loss": 0.5699939131736755,
-      "mean_token_accuracy": 0.7723715603351593,
-      "num_tokens": 5034188.0,
+      "loss": 0.6360020041465759,
+      "mean_token_accuracy": 0.7429406046867371,
+      "num_tokens": 2391824.0,
       "step": 64
     },
     {
-      "entropy": 0.550421878695488,
-      "epoch": 2.096774193548387,
-      "grad_norm": 0.020370641723275185,
+      "entropy": 0.6270973086357117,
+      "epoch": 0.5053449951409135,
+      "grad_norm": 0.01080579124391079,
       "learning_rate": 0.0002,
-      "loss": 0.5468724966049194,
-      "mean_token_accuracy": 0.7817957699298859,
-      "num_tokens": 5114558.0,
+      "loss": 0.6288405060768127,
+      "mean_token_accuracy": 0.7458023801445961,
+      "num_tokens": 2428949.0,
       "step": 65
     },
     {
-      "entropy": 0.5751384943723679,
-      "epoch": 2.129032258064516,
-      "grad_norm": 0.023148294538259506,
+      "entropy": 0.6371860653162003,
+      "epoch": 0.5131195335276968,
+      "grad_norm": 0.012401307933032513,
       "learning_rate": 0.0002,
-      "loss": 0.574794590473175,
-      "mean_token_accuracy": 0.7702242583036423,
-      "num_tokens": 5192598.0,
+      "loss": 0.6441806554794312,
+      "mean_token_accuracy": 0.7399793341755867,
+      "num_tokens": 2466079.0,
       "step": 66
     },
     {
-      "entropy": 0.5702789425849915,
-      "epoch": 2.161290322580645,
-      "grad_norm": 0.02145976386964321,
+      "entropy": 0.6057624816894531,
+      "epoch": 0.5208940719144801,
+      "grad_norm": 0.012216067872941494,
       "learning_rate": 0.0002,
-      "loss": 0.5721745491027832,
-      "mean_token_accuracy": 0.7724328637123108,
-      "num_tokens": 5272044.0,
+      "loss": 0.6104533672332764,
+      "mean_token_accuracy": 0.754143625497818,
+      "num_tokens": 2503686.0,
       "step": 67
     },
     {
-      "entropy": 0.5614835023880005,
-      "epoch": 2.193548387096774,
-      "grad_norm": 0.02582446113228798,
+      "entropy": 0.6143222749233246,
+      "epoch": 0.5286686103012633,
+      "grad_norm": 0.010598300956189632,
       "learning_rate": 0.0002,
-      "loss": 0.56424880027771,
-      "mean_token_accuracy": 0.7733140736818314,
-      "num_tokens": 5350931.0,
+      "loss": 0.617759108543396,
+      "mean_token_accuracy": 0.7507391050457954,
+      "num_tokens": 2541428.0,
       "step": 68
     },
     {
-      "entropy": 0.5537518709897995,
-      "epoch": 2.225806451612903,
-      "grad_norm": 0.020391393452882767,
+      "entropy": 0.6160529032349586,
+      "epoch": 0.5364431486880467,
+      "grad_norm": 0.010992376133799553,
       "learning_rate": 0.0002,
-      "loss": 0.5535145401954651,
-      "mean_token_accuracy": 0.779021218419075,
-      "num_tokens": 5429751.0,
+      "loss": 0.6156395673751831,
+      "mean_token_accuracy": 0.7503774240612984,
+      "num_tokens": 2578762.0,
       "step": 69
     },
     {
-      "entropy": 0.5642409324645996,
-      "epoch": 2.258064516129032,
-      "grad_norm": 0.023080818355083466,
+      "entropy": 0.6199713721871376,
+      "epoch": 0.54421768707483,
+      "grad_norm": 0.01134777907282114,
       "learning_rate": 0.0002,
-      "loss": 0.5602237582206726,
-      "mean_token_accuracy": 0.7764452248811722,
-      "num_tokens": 5508522.0,
+      "loss": 0.6140978932380676,
+      "mean_token_accuracy": 0.7530878558754921,
+      "num_tokens": 2616254.0,
       "step": 70
     },
     {
-      "entropy": 0.5553158819675446,
-      "epoch": 2.2903225806451615,
-      "grad_norm": 0.027256183326244354,
+      "entropy": 0.6425308436155319,
+      "epoch": 0.5519922254616132,
+      "grad_norm": 0.011413667351007462,
       "learning_rate": 0.0002,
-      "loss": 0.5490841269493103,
-      "mean_token_accuracy": 0.7800116837024689,
-      "num_tokens": 5587349.0,
+      "loss": 0.63960862159729,
+      "mean_token_accuracy": 0.7409021556377411,
+      "num_tokens": 2653615.0,
       "step": 71
     },
     {
-      "entropy": 0.5612549781799316,
-      "epoch": 2.3225806451612905,
-      "grad_norm": 0.019826354458928108,
+      "entropy": 0.6269439905881882,
+      "epoch": 0.5597667638483965,
+      "grad_norm": 0.011452693492174149,
       "learning_rate": 0.0002,
-      "loss": 0.5554416179656982,
-      "mean_token_accuracy": 0.7773576229810715,
-      "num_tokens": 5665374.0,
+      "loss": 0.6268940567970276,
+      "mean_token_accuracy": 0.7482407689094543,
+      "num_tokens": 2691144.0,
       "step": 72
     },
     {
-      "entropy": 0.5597121864557266,
-      "epoch": 2.3548387096774195,
-      "grad_norm": 0.021395916119217873,
+      "entropy": 0.6223431453108788,
+      "epoch": 0.5675413022351797,
+      "grad_norm": 0.011051226407289505,
       "learning_rate": 0.0002,
-      "loss": 0.5550926923751831,
-      "mean_token_accuracy": 0.7764492034912109,
-      "num_tokens": 5742628.0,
+      "loss": 0.627192497253418,
+      "mean_token_accuracy": 0.7496485412120819,
+      "num_tokens": 2728868.0,
       "step": 73
     },
     {
-      "entropy": 0.5639219135046005,
-      "epoch": 2.3870967741935485,
-      "grad_norm": 0.021380137652158737,
+      "entropy": 0.6142039522528648,
+      "epoch": 0.5753158406219631,
+      "grad_norm": 0.010462508536875248,
       "learning_rate": 0.0002,
-      "loss": 0.5646259784698486,
-      "mean_token_accuracy": 0.7729929685592651,
-      "num_tokens": 5823005.0,
+      "loss": 0.6176329851150513,
+      "mean_token_accuracy": 0.7508676871657372,
+      "num_tokens": 2766300.0,
       "step": 74
     },
     {
-      "entropy": 0.5548926442861557,
-      "epoch": 2.4193548387096775,
-      "grad_norm": 0.02109702117741108,
+      "entropy": 0.6121213287115097,
+      "epoch": 0.5830903790087464,
+      "grad_norm": 0.01290794089436531,
       "learning_rate": 0.0002,
-      "loss": 0.5567398071289062,
-      "mean_token_accuracy": 0.7773918360471725,
-      "num_tokens": 5901142.0,
+      "loss": 0.6207513809204102,
+      "mean_token_accuracy": 0.7499961778521538,
+      "num_tokens": 2803656.0,
       "step": 75
     },
     {
-      "entropy": 0.5468960106372833,
-      "epoch": 2.4516129032258065,
-      "grad_norm": 0.020396072417497635,
+      "entropy": 0.6135914027690887,
+      "epoch": 0.5908649173955296,
+      "grad_norm": 0.009645558893680573,
       "learning_rate": 0.0002,
-      "loss": 0.55022794008255,
-      "mean_token_accuracy": 0.7789339423179626,
-      "num_tokens": 5980524.0,
+      "loss": 0.6158081293106079,
+      "mean_token_accuracy": 0.7523107752203941,
+      "num_tokens": 2841367.0,
       "step": 76
     },
     {
-      "entropy": 0.5607923120260239,
-      "epoch": 2.4838709677419355,
-      "grad_norm": 0.023684872314333916,
+      "entropy": 0.6079287081956863,
+      "epoch": 0.5986394557823129,
+      "grad_norm": 0.010133286938071251,
       "learning_rate": 0.0002,
-      "loss": 0.5640513896942139,
-      "mean_token_accuracy": 0.7740097939968109,
-      "num_tokens": 6058950.0,
+      "loss": 0.6089348196983337,
+      "mean_token_accuracy": 0.7539728805422783,
+      "num_tokens": 2878743.0,
       "step": 77
     },
     {
-      "entropy": 0.5643693804740906,
-      "epoch": 2.5161290322580645,
-      "grad_norm": 0.01775151677429676,
+      "entropy": 0.619053803384304,
+      "epoch": 0.6064139941690962,
+      "grad_norm": 0.011026635766029358,
       "learning_rate": 0.0002,
-      "loss": 0.5636025071144104,
-      "mean_token_accuracy": 0.7733512371778488,
-      "num_tokens": 6137977.0,
+      "loss": 0.6179240942001343,
+      "mean_token_accuracy": 0.752836562693119,
+      "num_tokens": 2916083.0,
       "step": 78
     },
     {
-      "entropy": 0.5629870891571045,
-      "epoch": 2.5483870967741935,
-      "grad_norm": 0.020683595910668373,
+      "entropy": 0.6179199442267418,
+      "epoch": 0.6141885325558795,
+      "grad_norm": 0.009910841472446918,
       "learning_rate": 0.0002,
-      "loss": 0.5582836866378784,
-      "mean_token_accuracy": 0.7763189226388931,
-      "num_tokens": 6216345.0,
+      "loss": 0.6233879923820496,
+      "mean_token_accuracy": 0.7482481822371483,
+      "num_tokens": 2953599.0,
       "step": 79
     },
     {
-      "entropy": 0.5666602998971939,
-      "epoch": 2.5806451612903225,
-      "grad_norm": 0.019191065803170204,
+      "entropy": 0.6186161413788795,
+      "epoch": 0.6219630709426628,
+      "grad_norm": 0.00901505071669817,
       "learning_rate": 0.0002,
-      "loss": 0.5626470446586609,
-      "mean_token_accuracy": 0.7741801738739014,
-      "num_tokens": 6294649.0,
+      "loss": 0.6190614700317383,
+      "mean_token_accuracy": 0.7501069083809853,
+      "num_tokens": 2990903.0,
       "step": 80
     },
     {
-      "entropy": 0.5715147405862808,
-      "epoch": 2.6129032258064515,
-      "grad_norm": 0.019432390108704567,
+      "entropy": 0.6112105250358582,
+      "epoch": 0.6297376093294461,
+      "grad_norm": 0.011348889209330082,
       "learning_rate": 0.0002,
-      "loss": 0.5688481330871582,
-      "mean_token_accuracy": 0.773037850856781,
-      "num_tokens": 6373316.0,
+      "loss": 0.6137362718582153,
+      "mean_token_accuracy": 0.7515707835555077,
+      "num_tokens": 3028013.0,
       "step": 81
     },
     {
-      "entropy": 0.5707144886255264,
-      "epoch": 2.6451612903225805,
-      "grad_norm": 0.023160185664892197,
+      "entropy": 0.631052277982235,
+      "epoch": 0.6375121477162293,
+      "grad_norm": 0.012229959480464458,
       "learning_rate": 0.0002,
-      "loss": 0.5679908990859985,
-      "mean_token_accuracy": 0.7726998925209045,
-      "num_tokens": 6451529.0,
+      "loss": 0.6278471946716309,
+      "mean_token_accuracy": 0.7475830912590027,
+      "num_tokens": 3065444.0,
       "step": 82
     },
     {
-      "entropy": 0.5647523552179337,
-      "epoch": 2.6774193548387095,
-      "grad_norm": 0.022113436833024025,
+      "entropy": 0.6256950199604034,
+      "epoch": 0.6452866861030127,
+      "grad_norm": 0.009951086714863777,
       "learning_rate": 0.0002,
-      "loss": 0.5626997351646423,
-      "mean_token_accuracy": 0.7755681276321411,
-      "num_tokens": 6530936.0,
+      "loss": 0.6218305826187134,
+      "mean_token_accuracy": 0.7465297132730484,
+      "num_tokens": 3102683.0,
       "step": 83
     },
     {
-      "entropy": 0.5687842816114426,
-      "epoch": 2.709677419354839,
-      "grad_norm": 0.023050479590892792,
+      "entropy": 0.6196548789739609,
+      "epoch": 0.6530612244897959,
+      "grad_norm": 0.010236121714115143,
       "learning_rate": 0.0002,
-      "loss": 0.5685533881187439,
-      "mean_token_accuracy": 0.7720850706100464,
-      "num_tokens": 6609190.0,
+      "loss": 0.6180955767631531,
+      "mean_token_accuracy": 0.7518158033490181,
+      "num_tokens": 3140271.0,
       "step": 84
     },
     {
-      "entropy": 0.5653624683618546,
-      "epoch": 2.741935483870968,
-      "grad_norm": 0.020719164982438087,
+      "entropy": 0.6243495345115662,
+      "epoch": 0.6608357628765792,
+      "grad_norm": 0.011498089879751205,
       "learning_rate": 0.0002,
-      "loss": 0.5654714703559875,
-      "mean_token_accuracy": 0.7718778252601624,
-      "num_tokens": 6687358.0,
+      "loss": 0.6245218515396118,
+      "mean_token_accuracy": 0.7478918880224228,
+      "num_tokens": 3177703.0,
       "step": 85
     },
     {
-      "entropy": 0.557345911860466,
-      "epoch": 2.774193548387097,
-      "grad_norm": 0.02426682785153389,
+      "entropy": 0.619852252304554,
+      "epoch": 0.6686103012633625,
+      "grad_norm": 0.009944644756615162,
       "learning_rate": 0.0002,
-      "loss": 0.5572041869163513,
-      "mean_token_accuracy": 0.7766713351011276,
-      "num_tokens": 6765326.0,
+      "loss": 0.6247724294662476,
+      "mean_token_accuracy": 0.7476306930184364,
+      "num_tokens": 3215349.0,
       "step": 86
     },
     {
-      "entropy": 0.5582546591758728,
-      "epoch": 2.806451612903226,
-      "grad_norm": 0.020220208913087845,
+      "entropy": 0.6298213005065918,
+      "epoch": 0.6763848396501457,
+      "grad_norm": 0.009104517288506031,
       "learning_rate": 0.0002,
-      "loss": 0.5565258264541626,
-      "mean_token_accuracy": 0.7767817080020905,
-      "num_tokens": 6843956.0,
+      "loss": 0.6307124495506287,
+      "mean_token_accuracy": 0.7444773614406586,
+      "num_tokens": 3252965.0,
       "step": 87
     },
     {
-      "entropy": 0.5673591643571854,
-      "epoch": 2.838709677419355,
-      "grad_norm": 0.020146643742918968,
+      "entropy": 0.617066778242588,
+      "epoch": 0.6841593780369291,
+      "grad_norm": 0.009981841780245304,
       "learning_rate": 0.0002,
-      "loss": 0.5709323883056641,
-      "mean_token_accuracy": 0.7711061388254166,
-      "num_tokens": 6921812.0,
+      "loss": 0.6200389862060547,
+      "mean_token_accuracy": 0.7506762072443962,
+      "num_tokens": 3290319.0,
       "step": 88
     },
     {
-      "entropy": 0.556847408413887,
-      "epoch": 2.870967741935484,
-      "grad_norm": 0.01990182138979435,
+      "entropy": 0.6173762008547783,
+      "epoch": 0.6919339164237124,
+      "grad_norm": 0.010542662814259529,
       "learning_rate": 0.0002,
-      "loss": 0.5572229027748108,
-      "mean_token_accuracy": 0.7751472145318985,
-      "num_tokens": 7000840.0,
+      "loss": 0.6204540729522705,
+      "mean_token_accuracy": 0.7490114718675613,
+      "num_tokens": 3328270.0,
       "step": 89
     },
     {
-      "entropy": 0.563188761472702,
-      "epoch": 2.903225806451613,
-      "grad_norm": 0.02342582680284977,
+      "entropy": 0.6311885267496109,
+      "epoch": 0.6997084548104956,
+      "grad_norm": 0.009542804211378098,
       "learning_rate": 0.0002,
-      "loss": 0.5629247426986694,
-      "mean_token_accuracy": 0.7743992209434509,
-      "num_tokens": 7080445.0,
+      "loss": 0.630376935005188,
+      "mean_token_accuracy": 0.7458378821611404,
+      "num_tokens": 3365973.0,
       "step": 90
     },
     {
-      "entropy": 0.5722239166498184,
-      "epoch": 2.935483870967742,
-      "grad_norm": 0.019702041521668434,
+      "entropy": 0.6197073757648468,
+      "epoch": 0.7074829931972789,
+      "grad_norm": 0.009216145612299442,
       "learning_rate": 0.0002,
-      "loss": 0.5663557052612305,
-      "mean_token_accuracy": 0.7710109949111938,
-      "num_tokens": 7158666.0,
+      "loss": 0.6229982376098633,
+      "mean_token_accuracy": 0.7472847774624825,
+      "num_tokens": 3403209.0,
       "step": 91
     },
     {
-      "entropy": 0.5657145082950592,
-      "epoch": 2.967741935483871,
-      "grad_norm": 0.02240707166492939,
+      "entropy": 0.6161768138408661,
+      "epoch": 0.7152575315840622,
+      "grad_norm": 0.0102296257391572,
       "learning_rate": 0.0002,
-      "loss": 0.5609397888183594,
-      "mean_token_accuracy": 0.7747763097286224,
-      "num_tokens": 7238514.0,
+      "loss": 0.6171408891677856,
+      "mean_token_accuracy": 0.7507107704877853,
+      "num_tokens": 3440047.0,
       "step": 92
     },
     {
-      "entropy": 0.5848636776208878,
-      "epoch": 3.0,
-      "grad_norm": 0.023333914577960968,
+      "entropy": 0.6355379894375801,
+      "epoch": 0.7230320699708455,
+      "grad_norm": 0.011433378793299198,
       "learning_rate": 0.0002,
-      "loss": 0.5794190168380737,
-      "mean_token_accuracy": 0.7670062482357025,
-      "num_tokens": 7316444.0,
+      "loss": 0.6368508338928223,
+      "mean_token_accuracy": 0.7425181716680527,
+      "num_tokens": 3477583.0,
       "step": 93
     },
     {
-      "entropy": 0.5561475306749344,
-      "epoch": 3.032258064516129,
-      "grad_norm": 0.022262820973992348,
+      "entropy": 0.6173990294337273,
+      "epoch": 0.7308066083576288,
+      "grad_norm": 0.009371397085487843,
       "learning_rate": 0.0002,
-      "loss": 0.5519885420799255,
-      "mean_token_accuracy": 0.7779376059770584,
-      "num_tokens": 7394992.0,
+      "loss": 0.6161597371101379,
+      "mean_token_accuracy": 0.752352699637413,
+      "num_tokens": 3515443.0,
       "step": 94
     },
     {
-      "entropy": 0.5668191462755203,
-      "epoch": 3.064516129032258,
-      "grad_norm": 0.025911103934049606,
+      "entropy": 0.6139659285545349,
+      "epoch": 0.738581146744412,
+      "grad_norm": 0.00992497242987156,
       "learning_rate": 0.0002,
-      "loss": 0.5712538957595825,
-      "mean_token_accuracy": 0.770626112818718,
-      "num_tokens": 7472729.0,
+      "loss": 0.6122086644172668,
+      "mean_token_accuracy": 0.7519859671592712,
+      "num_tokens": 3552895.0,
       "step": 95
     },
     {
-      "entropy": 0.547975942492485,
-      "epoch": 3.096774193548387,
-      "grad_norm": 0.023955151438713074,
+      "entropy": 0.6270382553339005,
+      "epoch": 0.7463556851311953,
+      "grad_norm": 0.009038747288286686,
       "learning_rate": 0.0002,
-      "loss": 0.5533621311187744,
-      "mean_token_accuracy": 0.778180256485939,
-      "num_tokens": 7552213.0,
+      "loss": 0.6263892650604248,
+      "mean_token_accuracy": 0.7476321458816528,
+      "num_tokens": 3590558.0,
       "step": 96
     },
     {
-      "entropy": 0.5508177727460861,
-      "epoch": 3.129032258064516,
-      "grad_norm": 0.020201170817017555,
+      "entropy": 0.6252292916178703,
+      "epoch": 0.7541302235179786,
+      "grad_norm": 0.008724549785256386,
       "learning_rate": 0.0002,
-      "loss": 0.5523043870925903,
-      "mean_token_accuracy": 0.7791523486375809,
-      "num_tokens": 7631359.0,
+      "loss": 0.6235740184783936,
+      "mean_token_accuracy": 0.7499061301350594,
+      "num_tokens": 3628009.0,
       "step": 97
     },
     {
-      "entropy": 0.5587449967861176,
-      "epoch": 3.161290322580645,
-      "grad_norm": 0.022456398233771324,
+      "entropy": 0.6164776980876923,
+      "epoch": 0.7619047619047619,
+      "grad_norm": 0.009640969336032867,
       "learning_rate": 0.0002,
-      "loss": 0.5554713606834412,
-      "mean_token_accuracy": 0.7772288173437119,
-      "num_tokens": 7711191.0,
+      "loss": 0.6190721988677979,
+      "mean_token_accuracy": 0.7499102726578712,
+      "num_tokens": 3665592.0,
       "step": 98
     },
     {
-      "entropy": 0.5667514055967331,
-      "epoch": 3.193548387096774,
-      "grad_norm": 0.02376161515712738,
+      "entropy": 0.6058400347828865,
+      "epoch": 0.7696793002915452,
+      "grad_norm": 0.010068013332784176,
       "learning_rate": 0.0002,
-      "loss": 0.560375988483429,
-      "mean_token_accuracy": 0.7743281573057175,
-      "num_tokens": 7789741.0,
+      "loss": 0.6085440516471863,
+      "mean_token_accuracy": 0.7550365477800369,
+      "num_tokens": 3702702.0,
       "step": 99
     },
     {
-      "entropy": 0.5590975731611252,
-      "epoch": 3.225806451612903,
-      "grad_norm": 0.02175886370241642,
+      "entropy": 0.6062669232487679,
+      "epoch": 0.7774538386783285,
+      "grad_norm": 0.009750640951097012,
       "learning_rate": 0.0002,
-      "loss": 0.5548145771026611,
-      "mean_token_accuracy": 0.7771512418985367,
-      "num_tokens": 7867872.0,
+      "loss": 0.6106675267219543,
+      "mean_token_accuracy": 0.753579393029213,
+      "num_tokens": 3739974.0,
       "step": 100
     },
     {
-      "entropy": 0.5560945719480515,
-      "epoch": 3.258064516129032,
-      "grad_norm": 0.021177250891923904,
+      "entropy": 0.6228908747434616,
+      "epoch": 0.7852283770651117,
+      "grad_norm": 0.010727898217737675,
       "learning_rate": 0.0002,
-      "loss": 0.5567014217376709,
-      "mean_token_accuracy": 0.7768864035606384,
-      "num_tokens": 7946723.0,
+      "loss": 0.6273207664489746,
+      "mean_token_accuracy": 0.7462224960327148,
+      "num_tokens": 3777563.0,
       "step": 101
     },
     {
-      "entropy": 0.5489799231290817,
-      "epoch": 3.2903225806451615,
-      "grad_norm": 0.02086879126727581,
+      "entropy": 0.623473547399044,
+      "epoch": 0.793002915451895,
+      "grad_norm": 0.00916969496756792,
       "learning_rate": 0.0002,
-      "loss": 0.5482127070426941,
-      "mean_token_accuracy": 0.7793612778186798,
-      "num_tokens": 8025383.0,
+      "loss": 0.6267216205596924,
+      "mean_token_accuracy": 0.7489331811666489,
+      "num_tokens": 3815209.0,
       "step": 102
     },
     {
-      "entropy": 0.5634010285139084,
-      "epoch": 3.3225806451612905,
-      "grad_norm": 0.024660678580403328,
+      "entropy": 0.6183424741029739,
+      "epoch": 0.8007774538386784,
+      "grad_norm": 0.009042995050549507,
       "learning_rate": 0.0002,
-      "loss": 0.5666055083274841,
-      "mean_token_accuracy": 0.7726119607686996,
-      "num_tokens": 8103693.0,
+      "loss": 0.6215161681175232,
+      "mean_token_accuracy": 0.7501572594046593,
+      "num_tokens": 3853109.0,
       "step": 103
     },
     {
-      "entropy": 0.5655943602323532,
-      "epoch": 3.3548387096774195,
-      "grad_norm": 0.020694825798273087,
+      "entropy": 0.6086031794548035,
+      "epoch": 0.8085519922254616,
+      "grad_norm": 0.010368067771196365,
       "learning_rate": 0.0002,
-      "loss": 0.5662607550621033,
-      "mean_token_accuracy": 0.7717288136482239,
-      "num_tokens": 8180427.0,
+      "loss": 0.6092633008956909,
+      "mean_token_accuracy": 0.7548685073852539,
+      "num_tokens": 3890248.0,
       "step": 104
     },
     {
-      "entropy": 0.5499924272298813,
-      "epoch": 3.3870967741935485,
-      "grad_norm": 0.02108810283243656,
+      "entropy": 0.6166428253054619,
+      "epoch": 0.8163265306122449,
+      "grad_norm": 0.009941854514181614,
       "learning_rate": 0.0002,
-      "loss": 0.5465772151947021,
-      "mean_token_accuracy": 0.7803959846496582,
-      "num_tokens": 8260871.0,
+      "loss": 0.6201497316360474,
+      "mean_token_accuracy": 0.746953047811985,
+      "num_tokens": 3927797.0,
       "step": 105
     },
     {
-      "entropy": 0.557305708527565,
-      "epoch": 3.4193548387096775,
-      "grad_norm": 0.023198287934064865,
+      "entropy": 0.6270147785544395,
+      "epoch": 0.8241010689990281,
+      "grad_norm": 0.0085823442786932,
       "learning_rate": 0.0002,
-      "loss": 0.5546095371246338,
-      "mean_token_accuracy": 0.7774775624275208,
-      "num_tokens": 8339637.0,
+      "loss": 0.6264973878860474,
+      "mean_token_accuracy": 0.746023453772068,
+      "num_tokens": 3965467.0,
       "step": 106
     },
     {
-      "entropy": 0.5631329417228699,
-      "epoch": 3.4516129032258065,
-      "grad_norm": 0.025266552343964577,
+      "entropy": 0.6165590658783913,
+      "epoch": 0.8318756073858115,
+      "grad_norm": 0.009966393932700157,
       "learning_rate": 0.0002,
-      "loss": 0.5597484707832336,
-      "mean_token_accuracy": 0.7742401659488678,
-      "num_tokens": 8418700.0,
+      "loss": 0.6136085987091064,
+      "mean_token_accuracy": 0.7522286921739578,
+      "num_tokens": 4002647.0,
       "step": 107
     },
     {
-      "entropy": 0.5628319978713989,
-      "epoch": 3.4838709677419355,
-      "grad_norm": 0.02215176448225975,
+      "entropy": 0.6224516704678535,
+      "epoch": 0.8396501457725948,
+      "grad_norm": 0.01067335158586502,
       "learning_rate": 0.0002,
-      "loss": 0.5613549947738647,
-      "mean_token_accuracy": 0.7735034227371216,
-      "num_tokens": 8498009.0,
+      "loss": 0.6225627660751343,
+      "mean_token_accuracy": 0.7488038167357445,
+      "num_tokens": 4039990.0,
       "step": 108
     },
     {
-      "entropy": 0.55545274913311,
-      "epoch": 3.5161290322580645,
-      "grad_norm": 0.02309485711157322,
+      "entropy": 0.6210483014583588,
+      "epoch": 0.847424684159378,
+      "grad_norm": 0.010811593383550644,
       "learning_rate": 0.0002,
-      "loss": 0.5577254295349121,
-      "mean_token_accuracy": 0.7765019983053207,
-      "num_tokens": 8575886.0,
+      "loss": 0.6203762292861938,
+      "mean_token_accuracy": 0.7501420751214027,
+      "num_tokens": 4077461.0,
       "step": 109
     },
     {
-      "entropy": 0.5634028613567352,
-      "epoch": 3.5483870967741935,
-      "grad_norm": 0.025781216099858284,
+      "entropy": 0.6205326095223427,
+      "epoch": 0.8551992225461613,
+      "grad_norm": 0.009125830605626106,
       "learning_rate": 0.0002,
-      "loss": 0.566740870475769,
-      "mean_token_accuracy": 0.7714875340461731,
-      "num_tokens": 8654161.0,
+      "loss": 0.623924732208252,
+      "mean_token_accuracy": 0.7463032528758049,
+      "num_tokens": 4115100.0,
       "step": 110
     },
     {
-      "entropy": 0.5486367046833038,
-      "epoch": 3.5806451612903225,
-      "grad_norm": 0.024618223309516907,
+      "entropy": 0.6294923424720764,
+      "epoch": 0.8629737609329446,
+      "grad_norm": 0.008677372708916664,
       "learning_rate": 0.0002,
-      "loss": 0.5475247502326965,
-      "mean_token_accuracy": 0.7789197564125061,
-      "num_tokens": 8733226.0,
+      "loss": 0.6358845233917236,
+      "mean_token_accuracy": 0.7433422952890396,
+      "num_tokens": 4152751.0,
       "step": 111
     },
     {
-      "entropy": 0.5631587356328964,
-      "epoch": 3.6129032258064515,
-      "grad_norm": 0.024299612268805504,
+      "entropy": 0.615186795592308,
+      "epoch": 0.8707482993197279,
+      "grad_norm": 0.012182718142867088,
       "learning_rate": 0.0002,
-      "loss": 0.5573891401290894,
-      "mean_token_accuracy": 0.7757365703582764,
-      "num_tokens": 8810909.0,
+      "loss": 0.6257232427597046,
+      "mean_token_accuracy": 0.7469102591276169,
+      "num_tokens": 4189807.0,
       "step": 112
     },
     {
-      "entropy": 0.5608923137187958,
-      "epoch": 3.6451612903225805,
-      "grad_norm": 0.02582414820790291,
+      "entropy": 0.6161832436919212,
+      "epoch": 0.8785228377065112,
+      "grad_norm": 0.009921083226799965,
       "learning_rate": 0.0002,
-      "loss": 0.5579054951667786,
-      "mean_token_accuracy": 0.7754584103822708,
-      "num_tokens": 8890850.0,
+      "loss": 0.6149104237556458,
+      "mean_token_accuracy": 0.7501650750637054,
+      "num_tokens": 4227148.0,
       "step": 113
     },
     {
-      "entropy": 0.5561199635267258,
-      "epoch": 3.6774193548387095,
-      "grad_norm": 0.022211294621229172,
+      "entropy": 0.6069425120949745,
+      "epoch": 0.8862973760932945,
+      "grad_norm": 0.010401617735624313,
       "learning_rate": 0.0002,
-      "loss": 0.5566504001617432,
-      "mean_token_accuracy": 0.7767579257488251,
-      "num_tokens": 8969877.0,
+      "loss": 0.6072876453399658,
+      "mean_token_accuracy": 0.7546068653464317,
+      "num_tokens": 4264665.0,
       "step": 114
     },
     {
-      "entropy": 0.5588024407625198,
-      "epoch": 3.709677419354839,
-      "grad_norm": 0.022309575229883194,
+      "entropy": 0.6115981489419937,
+      "epoch": 0.8940719144800777,
+      "grad_norm": 0.009178240783512592,
       "learning_rate": 0.0002,
-      "loss": 0.5632297992706299,
-      "mean_token_accuracy": 0.7747749090194702,
-      "num_tokens": 9048037.0,
+      "loss": 0.6142789125442505,
+      "mean_token_accuracy": 0.7516015693545341,
+      "num_tokens": 4301705.0,
       "step": 115
     },
     {
-      "entropy": 0.551807701587677,
-      "epoch": 3.741935483870968,
-      "grad_norm": 0.025719881057739258,
+      "entropy": 0.6232319623231888,
+      "epoch": 0.901846452866861,
+      "grad_norm": 0.011550409719347954,
       "learning_rate": 0.0002,
-      "loss": 0.5533503293991089,
-      "mean_token_accuracy": 0.7774915248155594,
-      "num_tokens": 9126988.0,
+      "loss": 0.6232004165649414,
+      "mean_token_accuracy": 0.7479738518595695,
+      "num_tokens": 4339005.0,
       "step": 116
     },
     {
-      "entropy": 0.5595156997442245,
-      "epoch": 3.774193548387097,
-      "grad_norm": 0.024725891649723053,
+      "entropy": 0.6050816774368286,
+      "epoch": 0.9096209912536443,
+      "grad_norm": 0.009149852208793163,
       "learning_rate": 0.0002,
-      "loss": 0.555779218673706,
-      "mean_token_accuracy": 0.7771975845098495,
-      "num_tokens": 9206269.0,
+      "loss": 0.6073260307312012,
+      "mean_token_accuracy": 0.7553000226616859,
+      "num_tokens": 4376492.0,
       "step": 117
     },
     {
-      "entropy": 0.5715308636426926,
-      "epoch": 3.806451612903226,
-      "grad_norm": 0.02310684695839882,
+      "entropy": 0.6323671862483025,
+      "epoch": 0.9173955296404276,
+      "grad_norm": 0.01042769942432642,
       "learning_rate": 0.0002,
-      "loss": 0.5639746189117432,
-      "mean_token_accuracy": 0.772603914141655,
-      "num_tokens": 9284589.0,
+      "loss": 0.6354271769523621,
+      "mean_token_accuracy": 0.742473192512989,
+      "num_tokens": 4414516.0,
       "step": 118
     },
     {
-      "entropy": 0.5701656937599182,
-      "epoch": 3.838709677419355,
-      "grad_norm": 0.024147434160113335,
+      "entropy": 0.6279568076133728,
+      "epoch": 0.9251700680272109,
+      "grad_norm": 0.009754139930009842,
       "learning_rate": 0.0002,
-      "loss": 0.5665973424911499,
-      "mean_token_accuracy": 0.7718524187803268,
-      "num_tokens": 9362727.0,
+      "loss": 0.6289409399032593,
+      "mean_token_accuracy": 0.7444217056035995,
+      "num_tokens": 4452210.0,
       "step": 119
     },
     {
-      "entropy": 0.567099928855896,
-      "epoch": 3.870967741935484,
-      "grad_norm": 0.027005858719348907,
+      "entropy": 0.6090738251805305,
+      "epoch": 0.9329446064139941,
+      "grad_norm": 0.011155272834002972,
       "learning_rate": 0.0002,
-      "loss": 0.5683907270431519,
-      "mean_token_accuracy": 0.7712606638669968,
-      "num_tokens": 9441192.0,
+      "loss": 0.61219322681427,
+      "mean_token_accuracy": 0.7522945404052734,
+      "num_tokens": 4489567.0,
       "step": 120
     },
     {
-      "entropy": 0.5548799186944962,
-      "epoch": 3.903225806451613,
-      "grad_norm": 0.022680699825286865,
+      "entropy": 0.6002820879220963,
+      "epoch": 0.9407191448007775,
+      "grad_norm": 0.009737227112054825,
       "learning_rate": 0.0002,
-      "loss": 0.5592993497848511,
-      "mean_token_accuracy": 0.7751470804214478,
-      "num_tokens": 9519671.0,
+      "loss": 0.6018539071083069,
+      "mean_token_accuracy": 0.757467195391655,
+      "num_tokens": 4527111.0,
       "step": 121
     },
     {
-      "entropy": 0.5616104304790497,
-      "epoch": 3.935483870967742,
-      "grad_norm": 0.02054096944630146,
+      "entropy": 0.6134138256311417,
+      "epoch": 0.9484936831875608,
+      "grad_norm": 0.010466666892170906,
       "learning_rate": 0.0002,
-      "loss": 0.5620638132095337,
-      "mean_token_accuracy": 0.7740459442138672,
-      "num_tokens": 9597676.0,
+      "loss": 0.6216229200363159,
+      "mean_token_accuracy": 0.7486356720328331,
+      "num_tokens": 4564450.0,
       "step": 122
     },
     {
-      "entropy": 0.5476446449756622,
-      "epoch": 3.967741935483871,
-      "grad_norm": 0.020764634013175964,
+      "entropy": 0.6079460605978966,
+      "epoch": 0.956268221574344,
+      "grad_norm": 0.008756682276725769,
       "learning_rate": 0.0002,
-      "loss": 0.5447226762771606,
-      "mean_token_accuracy": 0.781438797712326,
-      "num_tokens": 9677971.0,
+      "loss": 0.6086549758911133,
+      "mean_token_accuracy": 0.7539011463522911,
+      "num_tokens": 4601781.0,
       "step": 123
     },
     {
-      "entropy": 0.5660000741481781,
-      "epoch": 4.0,
-      "grad_norm": 0.021668102592229843,
+      "entropy": 0.612805612385273,
+      "epoch": 0.9640427599611273,
+      "grad_norm": 0.009290359914302826,
       "learning_rate": 0.0002,
-      "loss": 0.5623193979263306,
-      "mean_token_accuracy": 0.7729711085557938,
-      "num_tokens": 9755431.0,
+      "loss": 0.6137915849685669,
+      "mean_token_accuracy": 0.752314880490303,
+      "num_tokens": 4639256.0,
       "step": 124
     },
     {
-      "entropy": 0.5558377206325531,
-      "epoch": 4.032258064516129,
-      "grad_norm": 0.0236333180218935,
+      "entropy": 0.6211065128445625,
+      "epoch": 0.9718172983479106,
+      "grad_norm": 0.010166316293179989,
       "learning_rate": 0.0002,
-      "loss": 0.5508249998092651,
-      "mean_token_accuracy": 0.7782040685415268,
-      "num_tokens": 9833682.0,
+      "loss": 0.6211485266685486,
+      "mean_token_accuracy": 0.7476831749081612,
+      "num_tokens": 4676689.0,
       "step": 125
     },
     {
-      "entropy": 0.5649599432945251,
-      "epoch": 4.064516129032258,
-      "grad_norm": 0.019735056906938553,
+      "entropy": 0.6219009906053543,
+      "epoch": 0.9795918367346939,
+      "grad_norm": 0.009910349734127522,
       "learning_rate": 0.0002,
-      "loss": 0.5595760941505432,
-      "mean_token_accuracy": 0.7747052013874054,
-      "num_tokens": 9911701.0,
+      "loss": 0.6177939176559448,
+      "mean_token_accuracy": 0.7506504207849503,
+      "num_tokens": 4714123.0,
       "step": 126
     },
     {
-      "entropy": 0.5603556483983994,
-      "epoch": 4.096774193548387,
-      "grad_norm": 0.019977176561951637,
+      "entropy": 0.6336647123098373,
+      "epoch": 0.9873663751214772,
+      "grad_norm": 0.00899409968405962,
       "learning_rate": 0.0002,
-      "loss": 0.560118556022644,
-      "mean_token_accuracy": 0.774644523859024,
-      "num_tokens": 9990441.0,
+      "loss": 0.6283681392669678,
+      "mean_token_accuracy": 0.7467537075281143,
+      "num_tokens": 4751822.0,
       "step": 127
     },
     {
-      "entropy": 0.556746780872345,
-      "epoch": 4.129032258064516,
-      "grad_norm": 0.02315964549779892,
+      "entropy": 0.6106422170996666,
+      "epoch": 0.9951409135082604,
+      "grad_norm": 0.009350291453301907,
       "learning_rate": 0.0002,
-      "loss": 0.5585101842880249,
-      "mean_token_accuracy": 0.7743638008832932,
-      "num_tokens": 10068737.0,
+      "loss": 0.6105437278747559,
+      "mean_token_accuracy": 0.7529440075159073,
+      "num_tokens": 4789003.0,
       "step": 128
     },
     {
-      "entropy": 0.5695119947195053,
-      "epoch": 4.161290322580645,
-      "grad_norm": 0.023623613640666008,
+      "entropy": 0.6007849931716919,
+      "epoch": 1.0,
+      "grad_norm": 0.011687238700687885,
       "learning_rate": 0.0002,
-      "loss": 0.573641300201416,
-      "mean_token_accuracy": 0.7688134759664536,
-      "num_tokens": 10146119.0,
+      "loss": 0.6050734519958496,
+      "mean_token_accuracy": 0.7555755734443664,
+      "num_tokens": 4812561.0,
       "step": 129
     },
     {
-      "entropy": 0.5613479018211365,
-      "epoch": 4.193548387096774,
-      "grad_norm": 0.02121862955391407,
+      "entropy": 0.6145111918449402,
+      "epoch": 1.0077745383867833,
+      "grad_norm": 0.00996165256947279,
       "learning_rate": 0.0002,
-      "loss": 0.5623332262039185,
-      "mean_token_accuracy": 0.773470863699913,
-      "num_tokens": 10223963.0,
+      "loss": 0.6200628280639648,
+      "mean_token_accuracy": 0.7501887455582619,
+      "num_tokens": 4849863.0,
       "step": 130
     },
     {
-      "entropy": 0.5604702234268188,
-      "epoch": 4.225806451612903,
-      "grad_norm": 0.020838554948568344,
+      "entropy": 0.6117109283804893,
+      "epoch": 1.0155490767735667,
+      "grad_norm": 0.009831785224378109,
       "learning_rate": 0.0002,
-      "loss": 0.5538380146026611,
-      "mean_token_accuracy": 0.7777739018201828,
-      "num_tokens": 10303313.0,
+      "loss": 0.6182730793952942,
+      "mean_token_accuracy": 0.7466977462172508,
+      "num_tokens": 4887693.0,
       "step": 131
     },
     {
-      "entropy": 0.5683186054229736,
-      "epoch": 4.258064516129032,
-      "grad_norm": 0.02106897532939911,
+      "entropy": 0.6117515116930008,
+      "epoch": 1.0233236151603498,
+      "grad_norm": 0.007678937632590532,
       "learning_rate": 0.0002,
-      "loss": 0.5649935603141785,
-      "mean_token_accuracy": 0.7719052284955978,
-      "num_tokens": 10381934.0,
+      "loss": 0.6126781105995178,
+      "mean_token_accuracy": 0.7521011829376221,
+      "num_tokens": 4925266.0,
       "step": 132
     },
     {
-      "entropy": 0.5469363033771515,
-      "epoch": 4.290322580645161,
-      "grad_norm": 0.0212655458599329,
+      "entropy": 0.6145991012454033,
+      "epoch": 1.031098153547133,
+      "grad_norm": 0.00878220982849598,
       "learning_rate": 0.0002,
-      "loss": 0.5432569980621338,
-      "mean_token_accuracy": 0.7824582010507584,
-      "num_tokens": 10461631.0,
+      "loss": 0.6143695116043091,
+      "mean_token_accuracy": 0.7531941831111908,
+      "num_tokens": 4962307.0,
       "step": 133
     },
     {
-      "entropy": 0.5437141954898834,
-      "epoch": 4.32258064516129,
-      "grad_norm": 0.02336573600769043,
+      "entropy": 0.6105677559971809,
+      "epoch": 1.0388726919339164,
+      "grad_norm": 0.009893382899463177,
       "learning_rate": 0.0002,
-      "loss": 0.5471414923667908,
-      "mean_token_accuracy": 0.7800082862377167,
-      "num_tokens": 10540427.0,
+      "loss": 0.6073343753814697,
+      "mean_token_accuracy": 0.7548379600048065,
+      "num_tokens": 4999709.0,
       "step": 134
     },
     {
-      "entropy": 0.5527804344892502,
-      "epoch": 4.354838709677419,
-      "grad_norm": 0.020969511941075325,
+      "entropy": 0.6226058155298233,
+      "epoch": 1.0466472303206997,
+      "grad_norm": 0.01014266349375248,
       "learning_rate": 0.0002,
-      "loss": 0.5548494458198547,
-      "mean_token_accuracy": 0.7759415507316589,
-      "num_tokens": 10618610.0,
+      "loss": 0.6212986707687378,
+      "mean_token_accuracy": 0.7492796331644058,
+      "num_tokens": 5036991.0,
       "step": 135
     },
     {
-      "entropy": 0.5538617074489594,
-      "epoch": 4.387096774193548,
-      "grad_norm": 0.02194584719836712,
+      "entropy": 0.6036651879549026,
+      "epoch": 1.054421768707483,
+      "grad_norm": 0.008363187313079834,
       "learning_rate": 0.0002,
-      "loss": 0.55549556016922,
-      "mean_token_accuracy": 0.7759937942028046,
-      "num_tokens": 10696364.0,
+      "loss": 0.6068825721740723,
+      "mean_token_accuracy": 0.753925733268261,
+      "num_tokens": 5074058.0,
       "step": 136
     },
     {
-      "entropy": 0.5536889731884003,
-      "epoch": 4.419354838709677,
-      "grad_norm": 0.020031817257404327,
+      "entropy": 0.6133922636508942,
+      "epoch": 1.0621963070942664,
+      "grad_norm": 0.00871324259787798,
       "learning_rate": 0.0002,
-      "loss": 0.5526796579360962,
-      "mean_token_accuracy": 0.7770518809556961,
-      "num_tokens": 10774927.0,
+      "loss": 0.6173220276832581,
+      "mean_token_accuracy": 0.7509826496243477,
+      "num_tokens": 5111393.0,
       "step": 137
     },
     {
-      "entropy": 0.5647106319665909,
-      "epoch": 4.451612903225806,
-      "grad_norm": 0.023352017626166344,
+      "entropy": 0.622228629887104,
+      "epoch": 1.0699708454810495,
+      "grad_norm": 0.010760784149169922,
       "learning_rate": 0.0002,
-      "loss": 0.5584677457809448,
-      "mean_token_accuracy": 0.7754687964916229,
-      "num_tokens": 10853063.0,
+      "loss": 0.629308819770813,
+      "mean_token_accuracy": 0.7456969246268272,
+      "num_tokens": 5148886.0,
       "step": 138
     },
     {
-      "entropy": 0.5529246926307678,
-      "epoch": 4.483870967741936,
-      "grad_norm": 0.02302417904138565,
+      "entropy": 0.6098063141107559,
+      "epoch": 1.0777453838678328,
+      "grad_norm": 0.010240385308861732,
       "learning_rate": 0.0002,
-      "loss": 0.5511515736579895,
-      "mean_token_accuracy": 0.777482196688652,
-      "num_tokens": 10932630.0,
+      "loss": 0.6158695220947266,
+      "mean_token_accuracy": 0.7498040646314621,
+      "num_tokens": 5186054.0,
       "step": 139
     },
     {
-      "entropy": 0.5594740360975266,
-      "epoch": 4.516129032258064,
-      "grad_norm": 0.023653097450733185,
+      "entropy": 0.6184578239917755,
+      "epoch": 1.0855199222546161,
+      "grad_norm": 0.010335118509829044,
       "learning_rate": 0.0002,
-      "loss": 0.5600106716156006,
-      "mean_token_accuracy": 0.7750049978494644,
-      "num_tokens": 11010608.0,
+      "loss": 0.6153397560119629,
+      "mean_token_accuracy": 0.7500413805246353,
+      "num_tokens": 5223497.0,
       "step": 140
     },
     {
-      "entropy": 0.5542867481708527,
-      "epoch": 4.548387096774194,
-      "grad_norm": 0.020881926640868187,
+      "entropy": 0.626154899597168,
+      "epoch": 1.0932944606413995,
+      "grad_norm": 0.009573339484632015,
       "learning_rate": 0.0002,
-      "loss": 0.5544250011444092,
-      "mean_token_accuracy": 0.7768481224775314,
-      "num_tokens": 11090001.0,
+      "loss": 0.6198714971542358,
+      "mean_token_accuracy": 0.7480529472231865,
+      "num_tokens": 5261284.0,
       "step": 141
     },
     {
-      "entropy": 0.5386150777339935,
-      "epoch": 4.580645161290323,
-      "grad_norm": 0.020752349868416786,
+      "entropy": 0.6239576488733292,
+      "epoch": 1.1010689990281828,
+      "grad_norm": 0.01078983023762703,
       "learning_rate": 0.0002,
-      "loss": 0.5364447832107544,
-      "mean_token_accuracy": 0.7837993949651718,
-      "num_tokens": 11170786.0,
+      "loss": 0.6222984790802002,
+      "mean_token_accuracy": 0.7482637241482735,
+      "num_tokens": 5298391.0,
       "step": 142
     },
     {
-      "entropy": 0.5630347430706024,
-      "epoch": 4.612903225806452,
-      "grad_norm": 0.02071365900337696,
+      "entropy": 0.6052972301840782,
+      "epoch": 1.1088435374149659,
+      "grad_norm": 0.009443830698728561,
       "learning_rate": 0.0002,
-      "loss": 0.5641552209854126,
-      "mean_token_accuracy": 0.7728276401758194,
-      "num_tokens": 11249199.0,
+      "loss": 0.6056097745895386,
+      "mean_token_accuracy": 0.755977600812912,
+      "num_tokens": 5335665.0,
       "step": 143
     },
     {
-      "entropy": 0.5528657734394073,
-      "epoch": 4.645161290322581,
-      "grad_norm": 0.023699482902884483,
+      "entropy": 0.6082469522953033,
+      "epoch": 1.1166180758017492,
+      "grad_norm": 0.008718990720808506,
       "learning_rate": 0.0002,
-      "loss": 0.5528600215911865,
-      "mean_token_accuracy": 0.7769691050052643,
-      "num_tokens": 11326014.0,
+      "loss": 0.6119707822799683,
+      "mean_token_accuracy": 0.7540762051939964,
+      "num_tokens": 5373467.0,
       "step": 144
     },
     {
-      "entropy": 0.5494723469018936,
-      "epoch": 4.67741935483871,
-      "grad_norm": 0.023178691044449806,
+      "entropy": 0.6099165305495262,
+      "epoch": 1.1243926141885325,
+      "grad_norm": 0.00886059831827879,
       "learning_rate": 0.0002,
-      "loss": 0.5483137369155884,
-      "mean_token_accuracy": 0.778669074177742,
-      "num_tokens": 11406477.0,
+      "loss": 0.6144919395446777,
+      "mean_token_accuracy": 0.7509082928299904,
+      "num_tokens": 5410699.0,
       "step": 145
     },
     {
-      "entropy": 0.5606386959552765,
-      "epoch": 4.709677419354839,
-      "grad_norm": 0.024073773995041847,
+      "entropy": 0.606502428650856,
+      "epoch": 1.1321671525753159,
+      "grad_norm": 0.009656739421188831,
       "learning_rate": 0.0002,
-      "loss": 0.5597125291824341,
-      "mean_token_accuracy": 0.7748508900403976,
-      "num_tokens": 11484581.0,
+      "loss": 0.6092766523361206,
+      "mean_token_accuracy": 0.7524644657969475,
+      "num_tokens": 5448145.0,
       "step": 146
     },
     {
-      "entropy": 0.5582331120967865,
-      "epoch": 4.741935483870968,
-      "grad_norm": 0.02165294997394085,
+      "entropy": 0.6240331009030342,
+      "epoch": 1.1399416909620992,
+      "grad_norm": 0.009215152822434902,
       "learning_rate": 0.0002,
-      "loss": 0.5575276017189026,
-      "mean_token_accuracy": 0.7765386253595352,
-      "num_tokens": 11563894.0,
+      "loss": 0.6283767819404602,
+      "mean_token_accuracy": 0.7434803545475006,
+      "num_tokens": 5485424.0,
       "step": 147
     },
     {
-      "entropy": 0.5467294007539749,
-      "epoch": 4.774193548387097,
-      "grad_norm": 0.024752194061875343,
+      "entropy": 0.6154414415359497,
+      "epoch": 1.1477162293488825,
+      "grad_norm": 0.009418639354407787,
       "learning_rate": 0.0002,
-      "loss": 0.5452256202697754,
-      "mean_token_accuracy": 0.779546245932579,
-      "num_tokens": 11643114.0,
+      "loss": 0.6141241192817688,
+      "mean_token_accuracy": 0.752318486571312,
+      "num_tokens": 5522963.0,
       "step": 148
     },
     {
-      "entropy": 0.5514553785324097,
-      "epoch": 4.806451612903226,
-      "grad_norm": 0.023302922025322914,
+      "entropy": 0.6115295439958572,
+      "epoch": 1.1554907677356656,
+      "grad_norm": 0.008995896205306053,
       "learning_rate": 0.0002,
-      "loss": 0.5512630343437195,
-      "mean_token_accuracy": 0.7783655822277069,
-      "num_tokens": 11721689.0,
+      "loss": 0.6093723773956299,
+      "mean_token_accuracy": 0.7540601193904877,
+      "num_tokens": 5560173.0,
       "step": 149
     },
     {
-      "entropy": 0.5526825040578842,
-      "epoch": 4.838709677419355,
-      "grad_norm": 0.023031046614050865,
+      "entropy": 0.6168172955513,
+      "epoch": 1.163265306122449,
+      "grad_norm": 0.007887676358222961,
       "learning_rate": 0.0002,
-      "loss": 0.5490496754646301,
-      "mean_token_accuracy": 0.7788337171077728,
-      "num_tokens": 11800889.0,
+      "loss": 0.6149598360061646,
+      "mean_token_accuracy": 0.7500165104866028,
+      "num_tokens": 5597781.0,
       "step": 150
     },
     {
-      "entropy": 0.5668125748634338,
-      "epoch": 4.870967741935484,
-      "grad_norm": 0.021988127380609512,
+      "entropy": 0.6069798469543457,
+      "epoch": 1.1710398445092323,
+      "grad_norm": 0.008911027573049068,
       "learning_rate": 0.0002,
-      "loss": 0.5608407258987427,
-      "mean_token_accuracy": 0.7742439359426498,
-      "num_tokens": 11880825.0,
+      "loss": 0.6064385771751404,
+      "mean_token_accuracy": 0.7551223114132881,
+      "num_tokens": 5635360.0,
       "step": 151
     },
     {
-      "entropy": 0.5554933696985245,
-      "epoch": 4.903225806451613,
-      "grad_norm": 0.020031005144119263,
+      "entropy": 0.6170337721705437,
+      "epoch": 1.1788143828960156,
+      "grad_norm": 0.008784991689026356,
       "learning_rate": 0.0002,
-      "loss": 0.5538957118988037,
-      "mean_token_accuracy": 0.7775003910064697,
-      "num_tokens": 11959141.0,
+      "loss": 0.6199371814727783,
+      "mean_token_accuracy": 0.7501093670725822,
+      "num_tokens": 5673236.0,
       "step": 152
     },
     {
-      "entropy": 0.5566612184047699,
-      "epoch": 4.935483870967742,
-      "grad_norm": 0.023369155824184418,
+      "entropy": 0.6042322665452957,
+      "epoch": 1.186588921282799,
+      "grad_norm": 0.008859240449965,
       "learning_rate": 0.0002,
-      "loss": 0.5597338676452637,
-      "mean_token_accuracy": 0.7750471085309982,
-      "num_tokens": 12036910.0,
+      "loss": 0.6099147200584412,
+      "mean_token_accuracy": 0.7529143393039703,
+      "num_tokens": 5710364.0,
       "step": 153
     },
     {
-      "entropy": 0.5446057766675949,
-      "epoch": 4.967741935483871,
-      "grad_norm": 0.02630646713078022,
+      "entropy": 0.6032149121165276,
+      "epoch": 1.194363459669582,
+      "grad_norm": 0.008698609657585621,
       "learning_rate": 0.0002,
-      "loss": 0.5490501523017883,
-      "mean_token_accuracy": 0.7789898663759232,
-      "num_tokens": 12115573.0,
+      "loss": 0.6091040372848511,
+      "mean_token_accuracy": 0.7545271888375282,
+      "num_tokens": 5747783.0,
       "step": 154
     },
     {
-      "entropy": 0.5549590289592743,
-      "epoch": 5.0,
-      "grad_norm": 0.02291618101298809,
+      "entropy": 0.5934446603059769,
+      "epoch": 1.2021379980563653,
+      "grad_norm": 0.009365703910589218,
       "learning_rate": 0.0002,
-      "loss": 0.5535373687744141,
-      "mean_token_accuracy": 0.7768057286739349,
-      "num_tokens": 12193899.0,
+      "loss": 0.6000581979751587,
+      "mean_token_accuracy": 0.7570822536945343,
+      "num_tokens": 5784550.0,
       "step": 155
     },
     {
-      "entropy": 0.5561019033193588,
-      "epoch": 5.032258064516129,
-      "grad_norm": 0.026937829330563545,
+      "entropy": 0.6240461468696594,
+      "epoch": 1.2099125364431487,
+      "grad_norm": 0.00959563348442316,
       "learning_rate": 0.0002,
-      "loss": 0.5483195781707764,
-      "mean_token_accuracy": 0.7804572731256485,
-      "num_tokens": 12272768.0,
+      "loss": 0.6229224801063538,
+      "mean_token_accuracy": 0.7504049837589264,
+      "num_tokens": 5822051.0,
       "step": 156
     },
     {
-      "entropy": 0.567187175154686,
-      "epoch": 5.064516129032258,
-      "grad_norm": 0.024019401520490646,
+      "entropy": 0.6372882649302483,
+      "epoch": 1.217687074829932,
+      "grad_norm": 0.008253706619143486,
       "learning_rate": 0.0002,
-      "loss": 0.5632338523864746,
-      "mean_token_accuracy": 0.7725232094526291,
-      "num_tokens": 12350931.0,
+      "loss": 0.6363227367401123,
+      "mean_token_accuracy": 0.7415706738829613,
+      "num_tokens": 5859606.0,
       "step": 157
     },
     {
-      "entropy": 0.5438211113214493,
-      "epoch": 5.096774193548387,
-      "grad_norm": 0.03028087317943573,
+      "entropy": 0.6115364283323288,
+      "epoch": 1.2254616132167153,
+      "grad_norm": 0.009423714131116867,
       "learning_rate": 0.0002,
-      "loss": 0.5464832782745361,
-      "mean_token_accuracy": 0.7778522670269012,
-      "num_tokens": 12429717.0,
+      "loss": 0.6082147359848022,
+      "mean_token_accuracy": 0.7525753974914551,
+      "num_tokens": 5897144.0,
       "step": 158
     },
     {
-      "entropy": 0.5475486516952515,
-      "epoch": 5.129032258064516,
-      "grad_norm": 0.02604476921260357,
+      "entropy": 0.6259980425238609,
+      "epoch": 1.2332361516034984,
+      "grad_norm": 0.008684671483933926,
       "learning_rate": 0.0002,
-      "loss": 0.549872875213623,
-      "mean_token_accuracy": 0.777956560254097,
-      "num_tokens": 12509565.0,
+      "loss": 0.6272211074829102,
+      "mean_token_accuracy": 0.7443205043673515,
+      "num_tokens": 5934271.0,
       "step": 159
     },
     {
-      "entropy": 0.5558717250823975,
-      "epoch": 5.161290322580645,
-      "grad_norm": 0.0321776308119297,
+      "entropy": 0.6230324283242226,
+      "epoch": 1.2410106899902817,
+      "grad_norm": 0.008543766103684902,
       "learning_rate": 0.0002,
-      "loss": 0.5513166189193726,
-      "mean_token_accuracy": 0.7780381888151169,
-      "num_tokens": 12588412.0,
+      "loss": 0.6271595358848572,
+      "mean_token_accuracy": 0.7448511347174644,
+      "num_tokens": 5971696.0,
       "step": 160
     },
     {
-      "entropy": 0.5628859251737595,
-      "epoch": 5.193548387096774,
-      "grad_norm": 0.02652231976389885,
+      "entropy": 0.6094089895486832,
+      "epoch": 1.248785228377065,
+      "grad_norm": 0.008484925143420696,
       "learning_rate": 0.0002,
-      "loss": 0.5562256574630737,
-      "mean_token_accuracy": 0.7748238295316696,
-      "num_tokens": 12665848.0,
+      "loss": 0.611902117729187,
+      "mean_token_accuracy": 0.751502238214016,
+      "num_tokens": 6009264.0,
       "step": 161
     },
     {
-      "entropy": 0.5413103997707367,
-      "epoch": 5.225806451612903,
-      "grad_norm": 0.03387759253382683,
+      "entropy": 0.6298199594020844,
+      "epoch": 1.2565597667638484,
+      "grad_norm": 0.009336225688457489,
       "learning_rate": 0.0002,
-      "loss": 0.5417234301567078,
-      "mean_token_accuracy": 0.7818101793527603,
-      "num_tokens": 12745448.0,
+      "loss": 0.6334519386291504,
+      "mean_token_accuracy": 0.7432255297899246,
+      "num_tokens": 6046764.0,
       "step": 162
     },
     {
-      "entropy": 0.5551229268312454,
-      "epoch": 5.258064516129032,
-      "grad_norm": 0.023084145039319992,
+      "entropy": 0.6082314997911453,
+      "epoch": 1.2643343051506317,
+      "grad_norm": 0.008643310517072678,
       "learning_rate": 0.0002,
-      "loss": 0.5586496591567993,
-      "mean_token_accuracy": 0.774695560336113,
-      "num_tokens": 12824082.0,
+      "loss": 0.6062188148498535,
+      "mean_token_accuracy": 0.7549923211336136,
+      "num_tokens": 6084129.0,
       "step": 163
     },
     {
-      "entropy": 0.5464837104082108,
-      "epoch": 5.290322580645161,
-      "grad_norm": 0.0308171845972538,
+      "entropy": 0.6305336207151413,
+      "epoch": 1.272108843537415,
+      "grad_norm": 0.009341283701360226,
       "learning_rate": 0.0002,
-      "loss": 0.5477242469787598,
-      "mean_token_accuracy": 0.7798135727643967,
-      "num_tokens": 12904238.0,
+      "loss": 0.6319139003753662,
+      "mean_token_accuracy": 0.7440996170043945,
+      "num_tokens": 6121588.0,
       "step": 164
     },
     {
-      "entropy": 0.5441529452800751,
-      "epoch": 5.32258064516129,
-      "grad_norm": 0.0238366536796093,
+      "entropy": 0.6267635151743889,
+      "epoch": 1.2798833819241984,
+      "grad_norm": 0.00801732949912548,
       "learning_rate": 0.0002,
-      "loss": 0.5439585447311401,
-      "mean_token_accuracy": 0.7792287319898605,
-      "num_tokens": 12983526.0,
+      "loss": 0.6295123100280762,
+      "mean_token_accuracy": 0.7461825907230377,
+      "num_tokens": 6158885.0,
       "step": 165
     },
     {
-      "entropy": 0.5672517120838165,
-      "epoch": 5.354838709677419,
-      "grad_norm": 0.03152715414762497,
+      "entropy": 0.6146969944238663,
+      "epoch": 1.2876579203109815,
+      "grad_norm": 0.009981551207602024,
       "learning_rate": 0.0002,
-      "loss": 0.5655900835990906,
-      "mean_token_accuracy": 0.772352933883667,
-      "num_tokens": 13060468.0,
+      "loss": 0.611693263053894,
+      "mean_token_accuracy": 0.7518708109855652,
+      "num_tokens": 6196224.0,
       "step": 166
     },
     {
-      "entropy": 0.5608488917350769,
-      "epoch": 5.387096774193548,
-      "grad_norm": 0.02373046800494194,
+      "entropy": 0.61357232183218,
+      "epoch": 1.2954324586977648,
+      "grad_norm": 0.009115675464272499,
       "learning_rate": 0.0002,
-      "loss": 0.5555386543273926,
-      "mean_token_accuracy": 0.7747525721788406,
-      "num_tokens": 13138649.0,
+      "loss": 0.6161482930183411,
+      "mean_token_accuracy": 0.7504228800535202,
+      "num_tokens": 6233989.0,
       "step": 167
     },
     {
-      "entropy": 0.5586176365613937,
-      "epoch": 5.419354838709677,
-      "grad_norm": 0.03595823422074318,
+      "entropy": 0.6055504828691483,
+      "epoch": 1.3032069970845481,
+      "grad_norm": 0.009081711992621422,
       "learning_rate": 0.0002,
-      "loss": 0.5541322827339172,
-      "mean_token_accuracy": 0.7748987674713135,
-      "num_tokens": 13216338.0,
+      "loss": 0.6060025691986084,
+      "mean_token_accuracy": 0.7522977739572525,
+      "num_tokens": 6271394.0,
       "step": 168
     },
     {
-      "entropy": 0.5443743169307709,
-      "epoch": 5.451612903225806,
-      "grad_norm": 0.023366756737232208,
+      "entropy": 0.6105418056249619,
+      "epoch": 1.3109815354713314,
+      "grad_norm": 0.009730422869324684,
       "learning_rate": 0.0002,
-      "loss": 0.5441352725028992,
-      "mean_token_accuracy": 0.780906617641449,
-      "num_tokens": 13295756.0,
+      "loss": 0.6106258630752563,
+      "mean_token_accuracy": 0.753718800842762,
+      "num_tokens": 6308814.0,
       "step": 169
     },
     {
-      "entropy": 0.5391481816768646,
-      "epoch": 5.483870967741936,
-      "grad_norm": 0.030348939821124077,
+      "entropy": 0.6137099266052246,
+      "epoch": 1.3187560738581148,
+      "grad_norm": 0.009750008583068848,
       "learning_rate": 0.0002,
-      "loss": 0.5443218350410461,
-      "mean_token_accuracy": 0.7819477766752243,
-      "num_tokens": 13374481.0,
+      "loss": 0.6176519393920898,
+      "mean_token_accuracy": 0.7496867552399635,
+      "num_tokens": 6346282.0,
       "step": 170
     },
     {
-      "entropy": 0.5512904077768326,
-      "epoch": 5.516129032258064,
-      "grad_norm": 0.02274961955845356,
+      "entropy": 0.6076096817851067,
+      "epoch": 1.3265306122448979,
+      "grad_norm": 0.00966191291809082,
       "learning_rate": 0.0002,
-      "loss": 0.551249623298645,
-      "mean_token_accuracy": 0.7771816253662109,
-      "num_tokens": 13451677.0,
+      "loss": 0.6177955865859985,
+      "mean_token_accuracy": 0.7480223625898361,
+      "num_tokens": 6383602.0,
       "step": 171
     },
     {
-      "entropy": 0.5613770186901093,
-      "epoch": 5.548387096774194,
-      "grad_norm": 0.028490891680121422,
+      "entropy": 0.6195629611611366,
+      "epoch": 1.3343051506316812,
+      "grad_norm": 0.008418413810431957,
       "learning_rate": 0.0002,
-      "loss": 0.5578492879867554,
-      "mean_token_accuracy": 0.776411771774292,
-      "num_tokens": 13529937.0,
+      "loss": 0.6202820539474487,
+      "mean_token_accuracy": 0.7488253712654114,
+      "num_tokens": 6421035.0,
       "step": 172
     },
     {
-      "entropy": 0.5444934517145157,
-      "epoch": 5.580645161290323,
-      "grad_norm": 0.02613925375044346,
+      "entropy": 0.6313579976558685,
+      "epoch": 1.3420796890184645,
+      "grad_norm": 0.009207581169903278,
       "learning_rate": 0.0002,
-      "loss": 0.5422302484512329,
-      "mean_token_accuracy": 0.7822085916996002,
-      "num_tokens": 13609443.0,
+      "loss": 0.6277562379837036,
+      "mean_token_accuracy": 0.7459008172154427,
+      "num_tokens": 6458322.0,
       "step": 173
     },
     {
-      "entropy": 0.5491741746664047,
-      "epoch": 5.612903225806452,
-      "grad_norm": 0.027057429775595665,
+      "entropy": 0.6184266805648804,
+      "epoch": 1.3498542274052479,
+      "grad_norm": 0.009536071680486202,
       "learning_rate": 0.0002,
-      "loss": 0.5511897206306458,
-      "mean_token_accuracy": 0.7783340960741043,
-      "num_tokens": 13689772.0,
+      "loss": 0.6161496639251709,
+      "mean_token_accuracy": 0.7486988604068756,
+      "num_tokens": 6495621.0,
       "step": 174
     },
     {
-      "entropy": 0.5478880554437637,
-      "epoch": 5.645161290322581,
-      "grad_norm": 0.026018880307674408,
+      "entropy": 0.6189115419983864,
+      "epoch": 1.3576287657920312,
+      "grad_norm": 0.0097253592684865,
       "learning_rate": 0.0002,
-      "loss": 0.5485314130783081,
-      "mean_token_accuracy": 0.7789710909128189,
-      "num_tokens": 13767989.0,
+      "loss": 0.6185729503631592,
+      "mean_token_accuracy": 0.7469287514686584,
+      "num_tokens": 6532892.0,
       "step": 175
     },
     {
-      "entropy": 0.5589290112257004,
-      "epoch": 5.67741935483871,
-      "grad_norm": 0.029512794688344002,
+      "entropy": 0.6204129755496979,
+      "epoch": 1.3654033041788143,
+      "grad_norm": 0.008999030105769634,
       "learning_rate": 0.0002,
-      "loss": 0.5614359378814697,
-      "mean_token_accuracy": 0.7731590867042542,
-      "num_tokens": 13845810.0,
+      "loss": 0.6241609454154968,
+      "mean_token_accuracy": 0.7467619553208351,
+      "num_tokens": 6570420.0,
       "step": 176
     },
     {
-      "entropy": 0.5505948513746262,
-      "epoch": 5.709677419354839,
-      "grad_norm": 0.024590788409113884,
+      "entropy": 0.6034655645489693,
+      "epoch": 1.3731778425655976,
+      "grad_norm": 0.009542280808091164,
       "learning_rate": 0.0002,
-      "loss": 0.5468696355819702,
-      "mean_token_accuracy": 0.780251756310463,
-      "num_tokens": 13924126.0,
+      "loss": 0.6092945337295532,
+      "mean_token_accuracy": 0.7532089725136757,
+      "num_tokens": 6607782.0,
       "step": 177
     },
     {
-      "entropy": 0.5619102567434311,
-      "epoch": 5.741935483870968,
-      "grad_norm": 0.024399949237704277,
+      "entropy": 0.5901695042848587,
+      "epoch": 1.380952380952381,
+      "grad_norm": 0.009766815230250359,
       "learning_rate": 0.0002,
-      "loss": 0.5583929419517517,
-      "mean_token_accuracy": 0.7739093750715256,
-      "num_tokens": 14002635.0,
+      "loss": 0.5972245335578918,
+      "mean_token_accuracy": 0.7562905699014664,
+      "num_tokens": 6644876.0,
       "step": 178
     },
     {
-      "entropy": 0.5550903379917145,
-      "epoch": 5.774193548387097,
-      "grad_norm": 0.024970386177301407,
+      "entropy": 0.6139540746808052,
+      "epoch": 1.3887269193391643,
+      "grad_norm": 0.00885457918047905,
       "learning_rate": 0.0002,
-      "loss": 0.552678108215332,
-      "mean_token_accuracy": 0.777456060051918,
-      "num_tokens": 14080946.0,
+      "loss": 0.6145105361938477,
+      "mean_token_accuracy": 0.7494652420282364,
+      "num_tokens": 6682153.0,
       "step": 179
     },
     {
-      "entropy": 0.5512722134590149,
-      "epoch": 5.806451612903226,
-      "grad_norm": 0.025160595774650574,
+      "entropy": 0.6128789633512497,
+      "epoch": 1.3965014577259476,
+      "grad_norm": 0.008380657061934471,
       "learning_rate": 0.0002,
-      "loss": 0.549900472164154,
-      "mean_token_accuracy": 0.7770860642194748,
-      "num_tokens": 14159665.0,
+      "loss": 0.6103215217590332,
+      "mean_token_accuracy": 0.7528893128037453,
+      "num_tokens": 6719589.0,
       "step": 180
     },
     {
-      "entropy": 0.5473987311124802,
-      "epoch": 5.838709677419355,
-      "grad_norm": 0.023530514910817146,
+      "entropy": 0.6158252954483032,
+      "epoch": 1.4042759961127307,
+      "grad_norm": 0.008612860925495625,
       "learning_rate": 0.0002,
-      "loss": 0.5512905120849609,
-      "mean_token_accuracy": 0.7771292179822922,
-      "num_tokens": 14238752.0,
+      "loss": 0.6157952547073364,
+      "mean_token_accuracy": 0.7490102499723434,
+      "num_tokens": 6757256.0,
       "step": 181
     },
     {
-      "entropy": 0.5468312054872513,
-      "epoch": 5.870967741935484,
-      "grad_norm": 0.02680164761841297,
+      "entropy": 0.6120009869337082,
+      "epoch": 1.412050534499514,
+      "grad_norm": 0.008242498151957989,
       "learning_rate": 0.0002,
-      "loss": 0.5493057370185852,
-      "mean_token_accuracy": 0.7771333903074265,
-      "num_tokens": 14318306.0,
+      "loss": 0.610822319984436,
+      "mean_token_accuracy": 0.7527074217796326,
+      "num_tokens": 6794427.0,
       "step": 182
     },
     {
-      "entropy": 0.5578427314758301,
-      "epoch": 5.903225806451613,
-      "grad_norm": 0.021048691123723984,
+      "entropy": 0.607716366648674,
+      "epoch": 1.4198250728862973,
+      "grad_norm": 0.00903896614909172,
       "learning_rate": 0.0002,
-      "loss": 0.5549115538597107,
-      "mean_token_accuracy": 0.775988444685936,
-      "num_tokens": 14396279.0,
+      "loss": 0.6105165481567383,
+      "mean_token_accuracy": 0.7544176280498505,
+      "num_tokens": 6831592.0,
       "step": 183
     },
     {
-      "entropy": 0.5536772906780243,
-      "epoch": 5.935483870967742,
-      "grad_norm": 0.022526104003190994,
+      "entropy": 0.6083493009209633,
+      "epoch": 1.4275996112730807,
+      "grad_norm": 0.007915884256362915,
       "learning_rate": 0.0002,
-      "loss": 0.5495197176933289,
-      "mean_token_accuracy": 0.7781153470277786,
-      "num_tokens": 14476180.0,
+      "loss": 0.611156165599823,
+      "mean_token_accuracy": 0.7522151321172714,
+      "num_tokens": 6869182.0,
       "step": 184
     },
     {
-      "entropy": 0.5544522255659103,
-      "epoch": 5.967741935483871,
-      "grad_norm": 0.024115389212965965,
+      "entropy": 0.6225397288799286,
+      "epoch": 1.435374149659864,
+      "grad_norm": 0.009225807152688503,
       "learning_rate": 0.0002,
-      "loss": 0.5527682304382324,
-      "mean_token_accuracy": 0.7773118168115616,
-      "num_tokens": 14554359.0,
+      "loss": 0.6281070709228516,
+      "mean_token_accuracy": 0.7460536956787109,
+      "num_tokens": 6906740.0,
       "step": 185
     },
     {
-      "entropy": 0.5480682402849197,
-      "epoch": 6.0,
-      "grad_norm": 0.02979295328259468,
+      "entropy": 0.6144149005413055,
+      "epoch": 1.4431486880466473,
+      "grad_norm": 0.010468881577253342,
       "learning_rate": 0.0002,
-      "loss": 0.5507486462593079,
-      "mean_token_accuracy": 0.7767946571111679,
-      "num_tokens": 14632526.0,
+      "loss": 0.6134935617446899,
+      "mean_token_accuracy": 0.7510944902896881,
+      "num_tokens": 6944215.0,
       "step": 186
     },
     {
-      "entropy": 0.5580382943153381,
-      "epoch": 6.032258064516129,
-      "grad_norm": 0.02307111956179142,
+      "entropy": 0.6219450011849403,
+      "epoch": 1.4509232264334306,
+      "grad_norm": 0.008004755713045597,
       "learning_rate": 0.0002,
-      "loss": 0.5539210438728333,
-      "mean_token_accuracy": 0.7763322442770004,
-      "num_tokens": 14711118.0,
+      "loss": 0.6226309537887573,
+      "mean_token_accuracy": 0.7474055960774422,
+      "num_tokens": 6981266.0,
       "step": 187
     },
     {
-      "entropy": 0.5532259047031403,
-      "epoch": 6.064516129032258,
-      "grad_norm": 0.028182433918118477,
+      "entropy": 0.6094852611422539,
+      "epoch": 1.4586977648202137,
+      "grad_norm": 0.00825564842671156,
       "learning_rate": 0.0002,
-      "loss": 0.547465443611145,
-      "mean_token_accuracy": 0.7796971946954727,
-      "num_tokens": 14789744.0,
+      "loss": 0.6140962839126587,
+      "mean_token_accuracy": 0.7508443966507912,
+      "num_tokens": 7018562.0,
       "step": 188
     },
     {
-      "entropy": 0.5554233491420746,
-      "epoch": 6.096774193548387,
-      "grad_norm": 0.021296130493283272,
+      "entropy": 0.6202999800443649,
+      "epoch": 1.466472303206997,
+      "grad_norm": 0.00858406163752079,
       "learning_rate": 0.0002,
-      "loss": 0.553041934967041,
-      "mean_token_accuracy": 0.7760283201932907,
-      "num_tokens": 14868371.0,
+      "loss": 0.6255123615264893,
+      "mean_token_accuracy": 0.744349479675293,
+      "num_tokens": 7055995.0,
       "step": 189
     },
     {
-      "entropy": 0.5400276184082031,
-      "epoch": 6.129032258064516,
-      "grad_norm": 0.029270071536302567,
+      "entropy": 0.620690606534481,
+      "epoch": 1.4742468415937804,
+      "grad_norm": 0.008920849300920963,
       "learning_rate": 0.0002,
-      "loss": 0.5436077117919922,
-      "mean_token_accuracy": 0.7790288478136063,
-      "num_tokens": 14946769.0,
+      "loss": 0.6179178357124329,
+      "mean_token_accuracy": 0.7499353438615799,
+      "num_tokens": 7093265.0,
       "step": 190
     },
     {
-      "entropy": 0.5573062002658844,
-      "epoch": 6.161290322580645,
-      "grad_norm": 0.02297142706811428,
+      "entropy": 0.6239896044135094,
+      "epoch": 1.4820213799805637,
+      "grad_norm": 0.00956004112958908,
       "learning_rate": 0.0002,
-      "loss": 0.5577552318572998,
-      "mean_token_accuracy": 0.7743619382381439,
-      "num_tokens": 15024271.0,
+      "loss": 0.6228440999984741,
+      "mean_token_accuracy": 0.7463066428899765,
+      "num_tokens": 7130758.0,
       "step": 191
     },
     {
-      "entropy": 0.5473188757896423,
-      "epoch": 6.193548387096774,
-      "grad_norm": 0.03082077018916607,
+      "entropy": 0.6123484745621681,
+      "epoch": 1.489795918367347,
+      "grad_norm": 0.009270581416785717,
       "learning_rate": 0.0002,
-      "loss": 0.5461888313293457,
-      "mean_token_accuracy": 0.7801126092672348,
-      "num_tokens": 15103815.0,
+      "loss": 0.6113156080245972,
+      "mean_token_accuracy": 0.752521239221096,
+      "num_tokens": 7168712.0,
       "step": 192
     },
     {
-      "entropy": 0.5432057976722717,
-      "epoch": 6.225806451612903,
-      "grad_norm": 0.0229332372546196,
+      "entropy": 0.616000160574913,
+      "epoch": 1.4975704567541301,
+      "grad_norm": 0.010770791210234165,
       "learning_rate": 0.0002,
-      "loss": 0.5412731170654297,
-      "mean_token_accuracy": 0.7815204709768295,
-      "num_tokens": 15182419.0,
+      "loss": 0.6215670108795166,
+      "mean_token_accuracy": 0.7469163537025452,
+      "num_tokens": 7205815.0,
       "step": 193
     },
     {
-      "entropy": 0.5486985445022583,
-      "epoch": 6.258064516129032,
-      "grad_norm": 0.02622872032225132,
+      "entropy": 0.6254132762551308,
+      "epoch": 1.5053449951409135,
+      "grad_norm": 0.008934563025832176,
       "learning_rate": 0.0002,
-      "loss": 0.5463793277740479,
-      "mean_token_accuracy": 0.7784279584884644,
-      "num_tokens": 15260644.0,
+      "loss": 0.627918541431427,
+      "mean_token_accuracy": 0.7457368895411491,
+      "num_tokens": 7243319.0,
       "step": 194
     },
     {
-      "entropy": 0.551539808511734,
-      "epoch": 6.290322580645161,
-      "grad_norm": 0.025613069534301758,
+      "entropy": 0.6185151413083076,
+      "epoch": 1.5131195335276968,
+      "grad_norm": 0.008593689650297165,
       "learning_rate": 0.0002,
-      "loss": 0.5521561503410339,
-      "mean_token_accuracy": 0.7762308567762375,
-      "num_tokens": 15338884.0,
+      "loss": 0.6180027723312378,
+      "mean_token_accuracy": 0.7505958154797554,
+      "num_tokens": 7280988.0,
       "step": 195
     },
     {
-      "entropy": 0.5487957298755646,
-      "epoch": 6.32258064516129,
-      "grad_norm": 0.02660960517823696,
+      "entropy": 0.6063526794314384,
+      "epoch": 1.5208940719144801,
+      "grad_norm": 0.01172675471752882,
       "learning_rate": 0.0002,
-      "loss": 0.5461074113845825,
-      "mean_token_accuracy": 0.7797946333885193,
-      "num_tokens": 15417356.0,
+      "loss": 0.6113625764846802,
+      "mean_token_accuracy": 0.7508470490574837,
+      "num_tokens": 7318322.0,
       "step": 196
     },
     {
-      "entropy": 0.5484287142753601,
-      "epoch": 6.354838709677419,
-      "grad_norm": 0.025937844067811966,
+      "entropy": 0.6110472977161407,
+      "epoch": 1.5286686103012634,
+      "grad_norm": 0.009914939291775227,
       "learning_rate": 0.0002,
-      "loss": 0.5483534336090088,
-      "mean_token_accuracy": 0.7784149944782257,
-      "num_tokens": 15496287.0,
+      "loss": 0.6066443920135498,
+      "mean_token_accuracy": 0.7531256228685379,
+      "num_tokens": 7355926.0,
       "step": 197
     },
     {
-      "entropy": 0.5399597436189651,
-      "epoch": 6.387096774193548,
-      "grad_norm": 0.02921709045767784,
+      "entropy": 0.6154029294848442,
+      "epoch": 1.5364431486880465,
+      "grad_norm": 0.008429116569459438,
       "learning_rate": 0.0002,
-      "loss": 0.5408337116241455,
-      "mean_token_accuracy": 0.7813759595155716,
-      "num_tokens": 15575856.0,
+      "loss": 0.6149218678474426,
+      "mean_token_accuracy": 0.7508518844842911,
+      "num_tokens": 7393433.0,
       "step": 198
     },
     {
-      "entropy": 0.5393888503313065,
-      "epoch": 6.419354838709677,
-      "grad_norm": 0.028209807351231575,
+      "entropy": 0.622289851307869,
+      "epoch": 1.54421768707483,
+      "grad_norm": 0.06439514458179474,
       "learning_rate": 0.0002,
-      "loss": 0.537765622138977,
-      "mean_token_accuracy": 0.7822040915489197,
-      "num_tokens": 15656049.0,
+      "loss": 0.631239116191864,
+      "mean_token_accuracy": 0.7464239746332169,
+      "num_tokens": 7431062.0,
       "step": 199
     },
     {
-      "entropy": 0.5398154109716415,
-      "epoch": 6.451612903225806,
-      "grad_norm": 0.02456795424222946,
+      "entropy": 0.619748018682003,
+      "epoch": 1.5519922254616132,
+      "grad_norm": 0.014919525012373924,
       "learning_rate": 0.0002,
-      "loss": 0.5377130508422852,
-      "mean_token_accuracy": 0.7834392338991165,
-      "num_tokens": 15736009.0,
+      "loss": 0.6189074516296387,
+      "mean_token_accuracy": 0.7492179349064827,
+      "num_tokens": 7468637.0,
       "step": 200
     },
     {
-      "entropy": 0.5495229363441467,
-      "epoch": 6.483870967741936,
-      "grad_norm": 0.024359729140996933,
+      "entropy": 0.6060447245836258,
+      "epoch": 1.5597667638483965,
+      "grad_norm": 0.019268253818154335,
       "learning_rate": 0.0002,
-      "loss": 0.5489087700843811,
-      "mean_token_accuracy": 0.7771141827106476,
-      "num_tokens": 15814393.0,
+      "loss": 0.6034001708030701,
+      "mean_token_accuracy": 0.7538621947169304,
+      "num_tokens": 7506043.0,
       "step": 201
     },
     {
-      "entropy": 0.5561530888080597,
-      "epoch": 6.516129032258064,
-      "grad_norm": 0.025875220075249672,
+      "entropy": 0.6062331944704056,
+      "epoch": 1.5675413022351798,
+      "grad_norm": 0.013310298323631287,
       "learning_rate": 0.0002,
-      "loss": 0.5526772141456604,
-      "mean_token_accuracy": 0.7763231992721558,
-      "num_tokens": 15893219.0,
+      "loss": 0.6041976809501648,
+      "mean_token_accuracy": 0.7555137276649475,
+      "num_tokens": 7543352.0,
       "step": 202
     },
     {
-      "entropy": 0.5465659946203232,
-      "epoch": 6.548387096774194,
-      "grad_norm": 0.024376418441534042,
+      "entropy": 0.6137887313961983,
+      "epoch": 1.575315840621963,
+      "grad_norm": 0.010207262821495533,
       "learning_rate": 0.0002,
-      "loss": 0.5443940758705139,
-      "mean_token_accuracy": 0.7796058654785156,
-      "num_tokens": 15971928.0,
+      "loss": 0.6102935671806335,
+      "mean_token_accuracy": 0.7525052726268768,
+      "num_tokens": 7580867.0,
       "step": 203
     },
     {
-      "entropy": 0.5460532903671265,
-      "epoch": 6.580645161290323,
-      "grad_norm": 0.02943381294608116,
+      "entropy": 0.6222132593393326,
+      "epoch": 1.5830903790087465,
+      "grad_norm": 0.013623848557472229,
       "learning_rate": 0.0002,
-      "loss": 0.5474627614021301,
-      "mean_token_accuracy": 0.7784818559885025,
-      "num_tokens": 16049771.0,
+      "loss": 0.6228193044662476,
+      "mean_token_accuracy": 0.7474973797798157,
+      "num_tokens": 7618160.0,
       "step": 204
     },
     {
-      "entropy": 0.5501101911067963,
-      "epoch": 6.612903225806452,
-      "grad_norm": 0.03332793340086937,
+      "entropy": 0.6147656589746475,
+      "epoch": 1.5908649173955296,
+      "grad_norm": 0.011189316399395466,
       "learning_rate": 0.0002,
-      "loss": 0.5518757700920105,
-      "mean_token_accuracy": 0.7763904929161072,
-      "num_tokens": 16127561.0,
+      "loss": 0.6171211004257202,
+      "mean_token_accuracy": 0.750194676220417,
+      "num_tokens": 7655454.0,
       "step": 205
     },
     {
-      "entropy": 0.5436413139104843,
-      "epoch": 6.645161290322581,
-      "grad_norm": 0.02456376701593399,
+      "entropy": 0.6172639206051826,
+      "epoch": 1.598639455782313,
+      "grad_norm": 0.010384263470768929,
       "learning_rate": 0.0002,
-      "loss": 0.5429570078849792,
-      "mean_token_accuracy": 0.7813099920749664,
-      "num_tokens": 16206759.0,
+      "loss": 0.6146183013916016,
+      "mean_token_accuracy": 0.7520541921257973,
+      "num_tokens": 7693399.0,
       "step": 206
     },
     {
-      "entropy": 0.5561850965023041,
-      "epoch": 6.67741935483871,
-      "grad_norm": 0.029962576925754547,
+      "entropy": 0.6170154958963394,
+      "epoch": 1.6064139941690962,
+      "grad_norm": 0.01584717258810997,
       "learning_rate": 0.0002,
-      "loss": 0.5575984716415405,
-      "mean_token_accuracy": 0.7756843119859695,
-      "num_tokens": 16284941.0,
+      "loss": 0.6188840866088867,
+      "mean_token_accuracy": 0.7489955052733421,
+      "num_tokens": 7730827.0,
       "step": 207
     },
     {
-      "entropy": 0.5501436293125153,
-      "epoch": 6.709677419354839,
-      "grad_norm": 0.024612169712781906,
+      "entropy": 0.6163731887936592,
+      "epoch": 1.6141885325558794,
+      "grad_norm": 0.018742689862847328,
       "learning_rate": 0.0002,
-      "loss": 0.5489822626113892,
-      "mean_token_accuracy": 0.7770082056522369,
-      "num_tokens": 16363484.0,
+      "loss": 0.6233645677566528,
+      "mean_token_accuracy": 0.7474886327981949,
+      "num_tokens": 7768375.0,
       "step": 208
     },
     {
-      "entropy": 0.5469796806573868,
-      "epoch": 6.741935483870968,
-      "grad_norm": 0.032858025282621384,
+      "entropy": 0.6161412820219994,
+      "epoch": 1.621963070942663,
+      "grad_norm": 0.00972844660282135,
       "learning_rate": 0.0002,
-      "loss": 0.5432774424552917,
-      "mean_token_accuracy": 0.7797504663467407,
-      "num_tokens": 16443224.0,
+      "loss": 0.622969388961792,
+      "mean_token_accuracy": 0.747713029384613,
+      "num_tokens": 7806087.0,
       "step": 209
     },
     {
-      "entropy": 0.5524753332138062,
-      "epoch": 6.774193548387097,
-      "grad_norm": 0.027275629341602325,
+      "entropy": 0.6096204742789268,
+      "epoch": 1.629737609329446,
+      "grad_norm": 0.008969136513769627,
       "learning_rate": 0.0002,
-      "loss": 0.5505518913269043,
-      "mean_token_accuracy": 0.7788595259189606,
-      "num_tokens": 16522163.0,
+      "loss": 0.6155728697776794,
+      "mean_token_accuracy": 0.7519758120179176,
+      "num_tokens": 7843539.0,
       "step": 210
     },
     {
-      "entropy": 0.5436955541372299,
-      "epoch": 6.806451612903226,
-      "grad_norm": 0.033380795270204544,
+      "entropy": 0.6264154762029648,
+      "epoch": 1.6375121477162293,
+      "grad_norm": 0.009598297998309135,
       "learning_rate": 0.0002,
-      "loss": 0.5442633628845215,
-      "mean_token_accuracy": 0.7801192104816437,
-      "num_tokens": 16600824.0,
+      "loss": 0.6284589767456055,
+      "mean_token_accuracy": 0.7461679801344872,
+      "num_tokens": 7881097.0,
       "step": 211
     },
     {
-      "entropy": 0.5455887615680695,
-      "epoch": 6.838709677419355,
-      "grad_norm": 0.029686089605093002,
+      "entropy": 0.6087732166051865,
+      "epoch": 1.6452866861030127,
+      "grad_norm": 0.012110014446079731,
       "learning_rate": 0.0002,
-      "loss": 0.5451895594596863,
-      "mean_token_accuracy": 0.7799460589885712,
-      "num_tokens": 16679996.0,
+      "loss": 0.6086750626564026,
+      "mean_token_accuracy": 0.7552923634648323,
+      "num_tokens": 7918507.0,
       "step": 212
     },
     {
-      "entropy": 0.5443432927131653,
-      "epoch": 6.870967741935484,
-      "grad_norm": 0.034105002880096436,
+      "entropy": 0.6052073761820793,
+      "epoch": 1.6530612244897958,
+      "grad_norm": 0.009252856485545635,
       "learning_rate": 0.0002,
-      "loss": 0.5448198914527893,
-      "mean_token_accuracy": 0.7806573361158371,
-      "num_tokens": 16757525.0,
+      "loss": 0.6080942153930664,
+      "mean_token_accuracy": 0.7508447393774986,
+      "num_tokens": 7955564.0,
       "step": 213
     },
     {
-      "entropy": 0.5445855557918549,
-      "epoch": 6.903225806451613,
-      "grad_norm": 0.03775215521454811,
+      "entropy": 0.6120294705033302,
+      "epoch": 1.6608357628765793,
+      "grad_norm": 0.00877399556338787,
       "learning_rate": 0.0002,
-      "loss": 0.5394209623336792,
-      "mean_token_accuracy": 0.7814014852046967,
-      "num_tokens": 16837120.0,
+      "loss": 0.614532470703125,
+      "mean_token_accuracy": 0.7511213645339012,
+      "num_tokens": 7992945.0,
       "step": 214
     },
     {
-      "entropy": 0.5457793027162552,
-      "epoch": 6.935483870967742,
-      "grad_norm": 0.0267504770308733,
+      "entropy": 0.6241516917943954,
+      "epoch": 1.6686103012633624,
+      "grad_norm": 0.01109123881906271,
       "learning_rate": 0.0002,
-      "loss": 0.5454181432723999,
-      "mean_token_accuracy": 0.7807471007108688,
-      "num_tokens": 16915568.0,
+      "loss": 0.626361608505249,
+      "mean_token_accuracy": 0.7474541217088699,
+      "num_tokens": 8030734.0,
       "step": 215
     },
     {
-      "entropy": 0.5482135564088821,
-      "epoch": 6.967741935483871,
-      "grad_norm": 0.038814593106508255,
+      "entropy": 0.6194194927811623,
+      "epoch": 1.6763848396501457,
+      "grad_norm": 0.009091328829526901,
       "learning_rate": 0.0002,
-      "loss": 0.552478551864624,
-      "mean_token_accuracy": 0.7767772674560547,
-      "num_tokens": 16993690.0,
+      "loss": 0.6160167455673218,
+      "mean_token_accuracy": 0.7495900243520737,
+      "num_tokens": 8067927.0,
       "step": 216
     },
     {
-      "entropy": 0.5654344111680984,
-      "epoch": 7.0,
-      "grad_norm": 0.031003009527921677,
+      "entropy": 0.6229095980525017,
+      "epoch": 1.684159378036929,
+      "grad_norm": 0.01083504967391491,
       "learning_rate": 0.0002,
-      "loss": 0.5617949366569519,
-      "mean_token_accuracy": 0.7726786136627197,
-      "num_tokens": 17071218.0,
+      "loss": 0.6173299551010132,
+      "mean_token_accuracy": 0.7501125037670135,
+      "num_tokens": 8105627.0,
       "step": 217
     },
     {
-      "entropy": 0.5554635226726532,
-      "epoch": 7.032258064516129,
-      "grad_norm": 0.03204672038555145,
+      "entropy": 0.6194786503911018,
+      "epoch": 1.6919339164237124,
+      "grad_norm": 0.008784794248640537,
       "learning_rate": 0.0002,
-      "loss": 0.5454404950141907,
-      "mean_token_accuracy": 0.7804635167121887,
-      "num_tokens": 17151568.0,
+      "loss": 0.6191180944442749,
+      "mean_token_accuracy": 0.748157188296318,
+      "num_tokens": 8143034.0,
       "step": 218
     },
     {
-      "entropy": 0.530587300658226,
-      "epoch": 7.064516129032258,
-      "grad_norm": 0.02964143082499504,
+      "entropy": 0.6127319037914276,
+      "epoch": 1.6997084548104957,
+      "grad_norm": 0.008322956040501595,
       "learning_rate": 0.0002,
-      "loss": 0.5306113958358765,
-      "mean_token_accuracy": 0.7852308750152588,
-      "num_tokens": 17230148.0,
+      "loss": 0.6156037449836731,
+      "mean_token_accuracy": 0.7487869411706924,
+      "num_tokens": 8180596.0,
       "step": 219
     },
     {
-      "entropy": 0.5281653106212616,
-      "epoch": 7.096774193548387,
-      "grad_norm": 0.027958055958151817,
+      "entropy": 0.6223913356661797,
+      "epoch": 1.7074829931972788,
+      "grad_norm": 0.008480758406221867,
       "learning_rate": 0.0002,
-      "loss": 0.5308482646942139,
-      "mean_token_accuracy": 0.7845461368560791,
-      "num_tokens": 17309751.0,
+      "loss": 0.6275689601898193,
+      "mean_token_accuracy": 0.7448212057352066,
+      "num_tokens": 8218297.0,
       "step": 220
     },
     {
-      "entropy": 0.548245832324028,
-      "epoch": 7.129032258064516,
-      "grad_norm": 0.0297231525182724,
+      "entropy": 0.6217603012919426,
+      "epoch": 1.7152575315840624,
+      "grad_norm": 0.008550974540412426,
       "learning_rate": 0.0002,
-      "loss": 0.5488801598548889,
-      "mean_token_accuracy": 0.7783880233764648,
-      "num_tokens": 17388675.0,
+      "loss": 0.6251344680786133,
+      "mean_token_accuracy": 0.74727413803339,
+      "num_tokens": 8255632.0,
       "step": 221
     },
     {
-      "entropy": 0.5462783575057983,
-      "epoch": 7.161290322580645,
-      "grad_norm": 0.02872973121702671,
+      "entropy": 0.6164599433541298,
+      "epoch": 1.7230320699708455,
+      "grad_norm": 0.008577285334467888,
       "learning_rate": 0.0002,
-      "loss": 0.5400034189224243,
-      "mean_token_accuracy": 0.7810834497213364,
-      "num_tokens": 17466936.0,
+      "loss": 0.6166675090789795,
+      "mean_token_accuracy": 0.7512230649590492,
+      "num_tokens": 8292769.0,
       "step": 222
     },
     {
-      "entropy": 0.5517232865095139,
-      "epoch": 7.193548387096774,
-      "grad_norm": 0.03185023367404938,
+      "entropy": 0.619569830596447,
+      "epoch": 1.7308066083576288,
+      "grad_norm": 0.008122924715280533,
       "learning_rate": 0.0002,
-      "loss": 0.5485620498657227,
-      "mean_token_accuracy": 0.7789378315210342,
-      "num_tokens": 17547010.0,
+      "loss": 0.6211944818496704,
+      "mean_token_accuracy": 0.7477053627371788,
+      "num_tokens": 8330358.0,
       "step": 223
     },
     {
-      "entropy": 0.5413918197154999,
-      "epoch": 7.225806451612903,
-      "grad_norm": 0.03178510442376137,
+      "entropy": 0.6090706288814545,
+      "epoch": 1.738581146744412,
+      "grad_norm": 0.008728940039873123,
       "learning_rate": 0.0002,
-      "loss": 0.5440365076065063,
-      "mean_token_accuracy": 0.7805528491735458,
-      "num_tokens": 17625499.0,
+      "loss": 0.6092976927757263,
+      "mean_token_accuracy": 0.7530095875263214,
+      "num_tokens": 8367398.0,
       "step": 224
     },
     {
-      "entropy": 0.5458857864141464,
-      "epoch": 7.258064516129032,
-      "grad_norm": 0.03427484259009361,
+      "entropy": 0.617340199649334,
+      "epoch": 1.7463556851311952,
+      "grad_norm": 0.007776155136525631,
       "learning_rate": 0.0002,
-      "loss": 0.545717179775238,
-      "mean_token_accuracy": 0.7795657068490982,
-      "num_tokens": 17704661.0,
+      "loss": 0.6156238913536072,
+      "mean_token_accuracy": 0.7495973780751228,
+      "num_tokens": 8404907.0,
       "step": 225
     },
     {
-      "entropy": 0.5347914099693298,
-      "epoch": 7.290322580645161,
-      "grad_norm": 0.04105692729353905,
+      "entropy": 0.6192676723003387,
+      "epoch": 1.7541302235179788,
+      "grad_norm": 0.008142861537635326,
       "learning_rate": 0.0002,
-      "loss": 0.537568211555481,
-      "mean_token_accuracy": 0.7824617475271225,
-      "num_tokens": 17783737.0,
+      "loss": 0.6212935447692871,
+      "mean_token_accuracy": 0.7502573132514954,
+      "num_tokens": 8442551.0,
       "step": 226
     },
     {
-      "entropy": 0.5410804450511932,
-      "epoch": 7.32258064516129,
-      "grad_norm": 0.027142923325300217,
+      "entropy": 0.6181206256151199,
+      "epoch": 1.7619047619047619,
+      "grad_norm": 0.009585011750459671,
       "learning_rate": 0.0002,
-      "loss": 0.5402100086212158,
-      "mean_token_accuracy": 0.7800760716199875,
-      "num_tokens": 17861910.0,
+      "loss": 0.625989556312561,
+      "mean_token_accuracy": 0.7449588850140572,
+      "num_tokens": 8480048.0,
       "step": 227
     },
     {
-      "entropy": 0.551323264837265,
-      "epoch": 7.354838709677419,
-      "grad_norm": 0.040414851158857346,
+      "entropy": 0.6096160188317299,
+      "epoch": 1.7696793002915452,
+      "grad_norm": 0.008407268673181534,
       "learning_rate": 0.0002,
-      "loss": 0.5469727516174316,
-      "mean_token_accuracy": 0.778721958398819,
-      "num_tokens": 17939724.0,
+      "loss": 0.6082819700241089,
+      "mean_token_accuracy": 0.7547600343823433,
+      "num_tokens": 8517427.0,
       "step": 228
     },
     {
-      "entropy": 0.5368226915597916,
-      "epoch": 7.387096774193548,
-      "grad_norm": 0.037495292723178864,
+      "entropy": 0.6192163527011871,
+      "epoch": 1.7774538386783285,
+      "grad_norm": 0.007702583912760019,
       "learning_rate": 0.0002,
-      "loss": 0.5379670858383179,
-      "mean_token_accuracy": 0.781633272767067,
-      "num_tokens": 18018788.0,
+      "loss": 0.6174746751785278,
+      "mean_token_accuracy": 0.7493149042129517,
+      "num_tokens": 8554570.0,
       "step": 229
     },
     {
-      "entropy": 0.5478527843952179,
-      "epoch": 7.419354838709677,
-      "grad_norm": 0.03409694880247116,
+      "entropy": 0.6089940667152405,
+      "epoch": 1.7852283770651116,
+      "grad_norm": 0.007490647025406361,
       "learning_rate": 0.0002,
-      "loss": 0.5442065000534058,
-      "mean_token_accuracy": 0.7787181437015533,
-      "num_tokens": 18098265.0,
+      "loss": 0.6111437082290649,
+      "mean_token_accuracy": 0.7528351470828056,
+      "num_tokens": 8591761.0,
       "step": 230
     },
     {
-      "entropy": 0.5493320971727371,
-      "epoch": 7.451612903225806,
-      "grad_norm": 0.03359943628311157,
+      "entropy": 0.6070573329925537,
+      "epoch": 1.7930029154518952,
+      "grad_norm": 0.008878695778548717,
       "learning_rate": 0.0002,
-      "loss": 0.5459405183792114,
-      "mean_token_accuracy": 0.7783495485782623,
-      "num_tokens": 18175715.0,
+      "loss": 0.6092367172241211,
+      "mean_token_accuracy": 0.7542654424905777,
+      "num_tokens": 8628522.0,
       "step": 231
     },
     {
-      "entropy": 0.5432014316320419,
-      "epoch": 7.483870967741936,
-      "grad_norm": 0.03568321838974953,
+      "entropy": 0.617328479886055,
+      "epoch": 1.8007774538386783,
+      "grad_norm": 0.008331574499607086,
       "learning_rate": 0.0002,
-      "loss": 0.5388095378875732,
-      "mean_token_accuracy": 0.7820334285497665,
-      "num_tokens": 18254428.0,
+      "loss": 0.6187005043029785,
+      "mean_token_accuracy": 0.7479601725935936,
+      "num_tokens": 8666140.0,
       "step": 232
     },
     {
-      "entropy": 0.5381491780281067,
-      "epoch": 7.516129032258064,
-      "grad_norm": 0.03510123863816261,
+      "entropy": 0.6115391552448273,
+      "epoch": 1.8085519922254616,
+      "grad_norm": 0.0087031414732337,
       "learning_rate": 0.0002,
-      "loss": 0.5440586805343628,
-      "mean_token_accuracy": 0.7799699455499649,
-      "num_tokens": 18333475.0,
+      "loss": 0.6169438362121582,
+      "mean_token_accuracy": 0.7492516785860062,
+      "num_tokens": 8703399.0,
       "step": 233
     },
     {
-      "entropy": 0.5427374690771103,
-      "epoch": 7.548387096774194,
-      "grad_norm": 0.04146914184093475,
+      "entropy": 0.6166737154126167,
+      "epoch": 1.816326530612245,
+      "grad_norm": 0.008778571151196957,
       "learning_rate": 0.0002,
-      "loss": 0.5435429215431213,
-      "mean_token_accuracy": 0.779713585972786,
-      "num_tokens": 18411870.0,
+      "loss": 0.6192659139633179,
+      "mean_token_accuracy": 0.7489692941308022,
+      "num_tokens": 8741141.0,
       "step": 234
     },
     {
-      "entropy": 0.5446290969848633,
-      "epoch": 7.580645161290323,
-      "grad_norm": 0.04291892424225807,
+      "entropy": 0.6278815269470215,
+      "epoch": 1.824101068999028,
+      "grad_norm": 0.008317695930600166,
       "learning_rate": 0.0002,
-      "loss": 0.5438494682312012,
-      "mean_token_accuracy": 0.7800731658935547,
-      "num_tokens": 18490201.0,
+      "loss": 0.6227933168411255,
+      "mean_token_accuracy": 0.7477347627282143,
+      "num_tokens": 8778706.0,
       "step": 235
     },
     {
-      "entropy": 0.5558793991804123,
-      "epoch": 7.612903225806452,
-      "grad_norm": 0.03633000701665878,
+      "entropy": 0.612150214612484,
+      "epoch": 1.8318756073858116,
+      "grad_norm": 0.008878265507519245,
       "learning_rate": 0.0002,
-      "loss": 0.5556504726409912,
-      "mean_token_accuracy": 0.7759376913309097,
-      "num_tokens": 18568345.0,
+      "loss": 0.6094038486480713,
+      "mean_token_accuracy": 0.7539248242974281,
+      "num_tokens": 8815790.0,
       "step": 236
     },
     {
-      "entropy": 0.5467730015516281,
-      "epoch": 7.645161290322581,
-      "grad_norm": 0.037996433675289154,
+      "entropy": 0.637272298336029,
+      "epoch": 1.8396501457725947,
+      "grad_norm": 0.007240319158881903,
       "learning_rate": 0.0002,
-      "loss": 0.5422417521476746,
-      "mean_token_accuracy": 0.7819823771715164,
-      "num_tokens": 18646506.0,
+      "loss": 0.6381018757820129,
+      "mean_token_accuracy": 0.7397879138588905,
+      "num_tokens": 8853228.0,
       "step": 237
     },
     {
-      "entropy": 0.5440531671047211,
-      "epoch": 7.67741935483871,
-      "grad_norm": 0.033006228506565094,
+      "entropy": 0.6079713776707649,
+      "epoch": 1.847424684159378,
+      "grad_norm": 0.008403577841818333,
       "learning_rate": 0.0002,
-      "loss": 0.5415350198745728,
-      "mean_token_accuracy": 0.7817762196063995,
-      "num_tokens": 18725716.0,
+      "loss": 0.6132397651672363,
+      "mean_token_accuracy": 0.750219389796257,
+      "num_tokens": 8890524.0,
       "step": 238
     },
     {
-      "entropy": 0.5402092933654785,
-      "epoch": 7.709677419354839,
-      "grad_norm": 0.04646630957722664,
+      "entropy": 0.6004362255334854,
+      "epoch": 1.8551992225461613,
+      "grad_norm": 0.008654654026031494,
       "learning_rate": 0.0002,
-      "loss": 0.5435824990272522,
-      "mean_token_accuracy": 0.7818535268306732,
-      "num_tokens": 18803274.0,
+      "loss": 0.6059733629226685,
+      "mean_token_accuracy": 0.7544897198677063,
+      "num_tokens": 8927518.0,
       "step": 239
     },
     {
-      "entropy": 0.5443805754184723,
-      "epoch": 7.741935483870968,
-      "grad_norm": 0.03801609203219414,
+      "entropy": 0.6205713227391243,
+      "epoch": 1.8629737609329446,
+      "grad_norm": 0.008852572180330753,
       "learning_rate": 0.0002,
-      "loss": 0.5479947328567505,
-      "mean_token_accuracy": 0.778268352150917,
-      "num_tokens": 18881044.0,
+      "loss": 0.6241973638534546,
+      "mean_token_accuracy": 0.7456497400999069,
+      "num_tokens": 8964770.0,
       "step": 240
     },
     {
-      "entropy": 0.5582472383975983,
-      "epoch": 7.774193548387097,
-      "grad_norm": 0.036031145602464676,
+      "entropy": 0.6309987902641296,
+      "epoch": 1.870748299319728,
+      "grad_norm": 0.00919515173882246,
       "learning_rate": 0.0002,
-      "loss": 0.5539592504501343,
-      "mean_token_accuracy": 0.7765996158123016,
-      "num_tokens": 18958606.0,
+      "loss": 0.6267882585525513,
+      "mean_token_accuracy": 0.7451199591159821,
+      "num_tokens": 9002039.0,
       "step": 241
     },
     {
-      "entropy": 0.5502520501613617,
-      "epoch": 7.806451612903226,
-      "grad_norm": 0.03642534092068672,
+      "entropy": 0.6201007068157196,
+      "epoch": 1.878522837706511,
+      "grad_norm": 0.008951977826654911,
       "learning_rate": 0.0002,
-      "loss": 0.5436160564422607,
-      "mean_token_accuracy": 0.7795243710279465,
-      "num_tokens": 19038252.0,
+      "loss": 0.6155404448509216,
+      "mean_token_accuracy": 0.7500432655215263,
+      "num_tokens": 9040052.0,
       "step": 242
     },
     {
-      "entropy": 0.5315112918615341,
-      "epoch": 7.838709677419355,
-      "grad_norm": 0.031813643872737885,
+      "entropy": 0.6137275472283363,
+      "epoch": 1.8862973760932946,
+      "grad_norm": 0.008249848149716854,
       "learning_rate": 0.0002,
-      "loss": 0.5328270196914673,
-      "mean_token_accuracy": 0.7842982560396194,
-      "num_tokens": 19116407.0,
+      "loss": 0.6162374019622803,
+      "mean_token_accuracy": 0.7486926540732384,
+      "num_tokens": 9077331.0,
       "step": 243
     },
     {
-      "entropy": 0.5394415557384491,
-      "epoch": 7.870967741935484,
-      "grad_norm": 0.03788164630532265,
+      "entropy": 0.6179945692420006,
+      "epoch": 1.8940719144800777,
+      "grad_norm": 0.008306370116770267,
       "learning_rate": 0.0002,
-      "loss": 0.5444040298461914,
-      "mean_token_accuracy": 0.7799783796072006,
-      "num_tokens": 19195585.0,
+      "loss": 0.6247057914733887,
+      "mean_token_accuracy": 0.7457103058695793,
+      "num_tokens": 9114582.0,
       "step": 244
     },
     {
-      "entropy": 0.5485299974679947,
-      "epoch": 7.903225806451613,
-      "grad_norm": 0.044309306889772415,
+      "entropy": 0.6145003139972687,
+      "epoch": 1.901846452866861,
+      "grad_norm": 0.009250648319721222,
       "learning_rate": 0.0002,
-      "loss": 0.5483869314193726,
-      "mean_token_accuracy": 0.7791053205728531,
-      "num_tokens": 19273583.0,
+      "loss": 0.6231353282928467,
+      "mean_token_accuracy": 0.7480179741978645,
+      "num_tokens": 9152485.0,
       "step": 245
     },
     {
-      "entropy": 0.542279988527298,
-      "epoch": 7.935483870967742,
-      "grad_norm": 0.03189428895711899,
+      "entropy": 0.6100682318210602,
+      "epoch": 1.9096209912536444,
+      "grad_norm": 0.008305463939905167,
       "learning_rate": 0.0002,
-      "loss": 0.5314847826957703,
-      "mean_token_accuracy": 0.7860700339078903,
-      "num_tokens": 19353829.0,
+      "loss": 0.61110520362854,
+      "mean_token_accuracy": 0.7516937106847763,
+      "num_tokens": 9189875.0,
       "step": 246
     },
     {
-      "entropy": 0.5488948225975037,
-      "epoch": 7.967741935483871,
-      "grad_norm": 0.03813633695244789,
+      "entropy": 0.6206101104617119,
+      "epoch": 1.9173955296404275,
+      "grad_norm": 0.009909824468195438,
       "learning_rate": 0.0002,
-      "loss": 0.5496359467506409,
-      "mean_token_accuracy": 0.7778913676738739,
-      "num_tokens": 19431439.0,
+      "loss": 0.6180837154388428,
+      "mean_token_accuracy": 0.7495366632938385,
+      "num_tokens": 9227450.0,
       "step": 247
     },
     {
-      "entropy": 0.5444356352090836,
-      "epoch": 8.0,
-      "grad_norm": 0.028016965836286545,
+      "entropy": 0.6217730790376663,
+      "epoch": 1.925170068027211,
+      "grad_norm": 0.008246448822319508,
       "learning_rate": 0.0002,
-      "loss": 0.5397998690605164,
-      "mean_token_accuracy": 0.7827391028404236,
-      "num_tokens": 19510222.0,
+      "loss": 0.6189583539962769,
+      "mean_token_accuracy": 0.7493142858147621,
+      "num_tokens": 9265223.0,
       "step": 248
     },
     {
-      "entropy": 0.5289585143327713,
-      "epoch": 8.03225806451613,
-      "grad_norm": 0.030331183224916458,
+      "entropy": 0.6133992001414299,
+      "epoch": 1.9329446064139941,
+      "grad_norm": 0.007729528471827507,
       "learning_rate": 0.0002,
-      "loss": 0.5258626937866211,
-      "mean_token_accuracy": 0.7890947312116623,
-      "num_tokens": 19589365.0,
+      "loss": 0.6164335608482361,
+      "mean_token_accuracy": 0.7500994428992271,
+      "num_tokens": 9302597.0,
       "step": 249
     },
     {
-      "entropy": 0.5283683687448502,
-      "epoch": 8.064516129032258,
-      "grad_norm": 0.041665561497211456,
+      "entropy": 0.596331886947155,
+      "epoch": 1.9407191448007775,
+      "grad_norm": 0.00885640550404787,
       "learning_rate": 0.0002,
-      "loss": 0.5343490242958069,
-      "mean_token_accuracy": 0.7842156440019608,
-      "num_tokens": 19667843.0,
+      "loss": 0.6054399013519287,
+      "mean_token_accuracy": 0.7546022981405258,
+      "num_tokens": 9339665.0,
       "step": 250
     },
     {
-      "entropy": 0.5371775776147842,
-      "epoch": 8.096774193548388,
-      "grad_norm": 0.031220240518450737,
+      "entropy": 0.6010145470499992,
+      "epoch": 1.9484936831875608,
+      "grad_norm": 0.00908851157873869,
       "learning_rate": 0.0002,
-      "loss": 0.5383769273757935,
-      "mean_token_accuracy": 0.7812917083501816,
-      "num_tokens": 19745970.0,
+      "loss": 0.6051974296569824,
+      "mean_token_accuracy": 0.7543318867683411,
+      "num_tokens": 9376514.0,
       "step": 251
     },
     {
-      "entropy": 0.535509005188942,
-      "epoch": 8.129032258064516,
-      "grad_norm": 0.034565962851047516,
+      "entropy": 0.6160777136683464,
+      "epoch": 1.9562682215743439,
+      "grad_norm": 0.008100342005491257,
       "learning_rate": 0.0002,
-      "loss": 0.5311164855957031,
-      "mean_token_accuracy": 0.7854022979736328,
-      "num_tokens": 19823505.0,
+      "loss": 0.6146577000617981,
+      "mean_token_accuracy": 0.7485847100615501,
+      "num_tokens": 9414214.0,
       "step": 252
     },
     {
-      "entropy": 0.5520065724849701,
-      "epoch": 8.161290322580646,
-      "grad_norm": 0.037510838359594345,
+      "entropy": 0.6268075257539749,
+      "epoch": 1.9640427599611274,
+      "grad_norm": 0.009163864888250828,
       "learning_rate": 0.0002,
-      "loss": 0.5506926774978638,
-      "mean_token_accuracy": 0.7766683101654053,
-      "num_tokens": 19902880.0,
+      "loss": 0.6204258799552917,
+      "mean_token_accuracy": 0.7494765147566795,
+      "num_tokens": 9451946.0,
       "step": 253
     },
     {
-      "entropy": 0.5412134379148483,
-      "epoch": 8.193548387096774,
-      "grad_norm": 0.03456275537610054,
+      "entropy": 0.6106092482805252,
+      "epoch": 1.9718172983479105,
+      "grad_norm": 0.007819678634405136,
       "learning_rate": 0.0002,
-      "loss": 0.5416734218597412,
-      "mean_token_accuracy": 0.780981183052063,
-      "num_tokens": 19981523.0,
+      "loss": 0.6087175011634827,
+      "mean_token_accuracy": 0.7527309507131577,
+      "num_tokens": 9489076.0,
       "step": 254
     },
     {
-      "entropy": 0.5260696411132812,
-      "epoch": 8.225806451612904,
-      "grad_norm": 0.04149680584669113,
+      "entropy": 0.6168656274676323,
+      "epoch": 1.9795918367346939,
+      "grad_norm": 0.007515368517488241,
       "learning_rate": 0.0002,
-      "loss": 0.5258564949035645,
-      "mean_token_accuracy": 0.7856866270303726,
-      "num_tokens": 20060057.0,
+      "loss": 0.6177273392677307,
+      "mean_token_accuracy": 0.751363955438137,
+      "num_tokens": 9526620.0,
       "step": 255
     },
     {
-      "entropy": 0.5432638376951218,
-      "epoch": 8.258064516129032,
-      "grad_norm": 0.03363949432969093,
+      "entropy": 0.6079138219356537,
+      "epoch": 1.9873663751214772,
+      "grad_norm": 0.009815288707613945,
       "learning_rate": 0.0002,
-      "loss": 0.5418345928192139,
-      "mean_token_accuracy": 0.7798439264297485,
-      "num_tokens": 20138196.0,
+      "loss": 0.6155093312263489,
+      "mean_token_accuracy": 0.7523162961006165,
+      "num_tokens": 9564466.0,
       "step": 256
     },
     {
-      "entropy": 0.5362787991762161,
-      "epoch": 8.290322580645162,
-      "grad_norm": 0.039886943995952606,
+      "entropy": 0.6036990880966187,
+      "epoch": 1.9951409135082603,
+      "grad_norm": 0.008685542270541191,
       "learning_rate": 0.0002,
-      "loss": 0.5321515798568726,
-      "mean_token_accuracy": 0.7850725948810577,
-      "num_tokens": 20217064.0,
+      "loss": 0.6101571917533875,
+      "mean_token_accuracy": 0.7522860541939735,
+      "num_tokens": 9601565.0,
       "step": 257
     },
     {
-      "entropy": 0.5206107646226883,
-      "epoch": 8.32258064516129,
-      "grad_norm": 0.03769670054316521,
+      "entropy": 0.6164417505264282,
+      "epoch": 2.0,
+      "grad_norm": 0.010608273558318615,
       "learning_rate": 0.0002,
-      "loss": 0.5199801921844482,
-      "mean_token_accuracy": 0.789309486746788,
-      "num_tokens": 20296824.0,
+      "loss": 0.6168809533119202,
+      "mean_token_accuracy": 0.7516962647438049,
+      "num_tokens": 9625165.0,
       "step": 258
     },
     {
-      "entropy": 0.533658817410469,
-      "epoch": 8.35483870967742,
-      "grad_norm": 0.044177986681461334,
+      "entropy": 0.6057095304131508,
+      "epoch": 2.007774538386783,
+      "grad_norm": 0.009618077427148819,
       "learning_rate": 0.0002,
-      "loss": 0.5321055054664612,
-      "mean_token_accuracy": 0.7848184555768967,
-      "num_tokens": 20375664.0,
+      "loss": 0.6003708839416504,
+      "mean_token_accuracy": 0.7543603405356407,
+      "num_tokens": 9662658.0,
       "step": 259
     },
     {
-      "entropy": 0.5468465983867645,
-      "epoch": 8.387096774193548,
-      "grad_norm": 0.04525522515177727,
+      "entropy": 0.6160204112529755,
+      "epoch": 2.0155490767735667,
+      "grad_norm": 0.007366312202066183,
       "learning_rate": 0.0002,
-      "loss": 0.5517196655273438,
-      "mean_token_accuracy": 0.7764200270175934,
-      "num_tokens": 20452975.0,
+      "loss": 0.6107323169708252,
+      "mean_token_accuracy": 0.751634031534195,
+      "num_tokens": 9700423.0,
       "step": 260
     },
     {
-      "entropy": 0.5353173762559891,
-      "epoch": 8.419354838709678,
-      "grad_norm": 0.03942183405160904,
+      "entropy": 0.6157180666923523,
+      "epoch": 2.0233236151603498,
+      "grad_norm": 0.008381242863833904,
       "learning_rate": 0.0002,
-      "loss": 0.5361676216125488,
-      "mean_token_accuracy": 0.7836414724588394,
-      "num_tokens": 20529996.0,
+      "loss": 0.6190871596336365,
+      "mean_token_accuracy": 0.7481335178017616,
+      "num_tokens": 9737692.0,
       "step": 261
     },
     {
-      "entropy": 0.5416340827941895,
-      "epoch": 8.451612903225806,
-      "grad_norm": 0.0363653190433979,
+      "entropy": 0.6075473129749298,
+      "epoch": 2.0310981535471333,
+      "grad_norm": 0.010082116350531578,
       "learning_rate": 0.0002,
-      "loss": 0.5382126569747925,
-      "mean_token_accuracy": 0.7817256599664688,
-      "num_tokens": 20608526.0,
+      "loss": 0.6141625642776489,
+      "mean_token_accuracy": 0.7516499608755112,
+      "num_tokens": 9775062.0,
       "step": 262
     },
     {
-      "entropy": 0.5331941545009613,
-      "epoch": 8.483870967741936,
-      "grad_norm": 0.03704048693180084,
+      "entropy": 0.6151047423481941,
+      "epoch": 2.0388726919339164,
+      "grad_norm": 0.009529951959848404,
       "learning_rate": 0.0002,
-      "loss": 0.5287242531776428,
-      "mean_token_accuracy": 0.7847851067781448,
-      "num_tokens": 20688137.0,
+      "loss": 0.6205469965934753,
+      "mean_token_accuracy": 0.7464399412274361,
+      "num_tokens": 9812526.0,
       "step": 263
     },
     {
-      "entropy": 0.5385544896125793,
-      "epoch": 8.516129032258064,
-      "grad_norm": 0.03550460934638977,
+      "entropy": 0.6053090617060661,
+      "epoch": 2.0466472303206995,
+      "grad_norm": 0.008532468229532242,
       "learning_rate": 0.0002,
-      "loss": 0.5389328002929688,
-      "mean_token_accuracy": 0.7807945758104324,
-      "num_tokens": 20766079.0,
+      "loss": 0.604971170425415,
+      "mean_token_accuracy": 0.7544130459427834,
+      "num_tokens": 9849685.0,
       "step": 264
     },
     {
-      "entropy": 0.5371405631303787,
-      "epoch": 8.548387096774194,
-      "grad_norm": 0.034064196050167084,
+      "entropy": 0.6305021941661835,
+      "epoch": 2.054421768707483,
+      "grad_norm": 0.009080994874238968,
       "learning_rate": 0.0002,
-      "loss": 0.5364264249801636,
-      "mean_token_accuracy": 0.7841684371232986,
-      "num_tokens": 20845237.0,
+      "loss": 0.6241769790649414,
+      "mean_token_accuracy": 0.7455881908535957,
+      "num_tokens": 9887441.0,
       "step": 265
     },
     {
-      "entropy": 0.5327372252941132,
-      "epoch": 8.580645161290322,
-      "grad_norm": 0.04183376953005791,
+      "entropy": 0.6111467704176903,
+      "epoch": 2.062196307094266,
+      "grad_norm": 0.008834858424961567,
       "learning_rate": 0.0002,
-      "loss": 0.5354538559913635,
-      "mean_token_accuracy": 0.7825859487056732,
-      "num_tokens": 20923137.0,
+      "loss": 0.6073428392410278,
+      "mean_token_accuracy": 0.7530370578169823,
+      "num_tokens": 9924611.0,
       "step": 266
     },
     {
-      "entropy": 0.550455629825592,
-      "epoch": 8.612903225806452,
-      "grad_norm": 0.04953514039516449,
+      "entropy": 0.6218099594116211,
+      "epoch": 2.0699708454810497,
+      "grad_norm": 0.009297652170062065,
       "learning_rate": 0.0002,
-      "loss": 0.5466119050979614,
-      "mean_token_accuracy": 0.7782289534807205,
-      "num_tokens": 21001585.0,
+      "loss": 0.6261637806892395,
+      "mean_token_accuracy": 0.7442486062645912,
+      "num_tokens": 9962409.0,
       "step": 267
     },
     {
-      "entropy": 0.5444563925266266,
-      "epoch": 8.64516129032258,
-      "grad_norm": 0.03682374581694603,
+      "entropy": 0.620900422334671,
+      "epoch": 2.077745383867833,
+      "grad_norm": 0.012807542458176613,
       "learning_rate": 0.0002,
-      "loss": 0.5486843585968018,
-      "mean_token_accuracy": 0.7768545746803284,
-      "num_tokens": 21079550.0,
+      "loss": 0.6319394111633301,
+      "mean_token_accuracy": 0.7447366267442703,
+      "num_tokens": 9999911.0,
       "step": 268
     },
     {
-      "entropy": 0.5407392978668213,
-      "epoch": 8.67741935483871,
-      "grad_norm": 0.03509804606437683,
+      "entropy": 0.6219062060117722,
+      "epoch": 2.0855199222546164,
+      "grad_norm": 0.008311600424349308,
       "learning_rate": 0.0002,
-      "loss": 0.5428568124771118,
-      "mean_token_accuracy": 0.7798480093479156,
-      "num_tokens": 21158113.0,
+      "loss": 0.6196457147598267,
+      "mean_token_accuracy": 0.7454339265823364,
+      "num_tokens": 10037309.0,
       "step": 269
     },
     {
-      "entropy": 0.5410716235637665,
-      "epoch": 8.709677419354838,
-      "grad_norm": 0.05400121584534645,
+      "entropy": 0.6179447770118713,
+      "epoch": 2.0932944606413995,
+      "grad_norm": 0.010762435384094715,
       "learning_rate": 0.0002,
-      "loss": 0.5374223589897156,
-      "mean_token_accuracy": 0.7816896289587021,
-      "num_tokens": 21236915.0,
+      "loss": 0.6135507822036743,
+      "mean_token_accuracy": 0.7513665333390236,
+      "num_tokens": 10074802.0,
       "step": 270
     },
     {
-      "entropy": 0.5346012562513351,
-      "epoch": 8.741935483870968,
-      "grad_norm": 0.05865916237235069,
+      "entropy": 0.6114057898521423,
+      "epoch": 2.1010689990281826,
+      "grad_norm": 0.009498962201178074,
       "learning_rate": 0.0002,
-      "loss": 0.5330258011817932,
-      "mean_token_accuracy": 0.7844430059194565,
-      "num_tokens": 21316149.0,
+      "loss": 0.6119563579559326,
+      "mean_token_accuracy": 0.7511925473809242,
+      "num_tokens": 10112087.0,
       "step": 271
     },
     {
-      "entropy": 0.5349076390266418,
-      "epoch": 8.774193548387096,
-      "grad_norm": 0.04287338629364967,
+      "entropy": 0.5985657572746277,
+      "epoch": 2.108843537414966,
+      "grad_norm": 0.00927242822945118,
       "learning_rate": 0.0002,
-      "loss": 0.5328216552734375,
-      "mean_token_accuracy": 0.7840270400047302,
-      "num_tokens": 21395461.0,
+      "loss": 0.6077044010162354,
+      "mean_token_accuracy": 0.751846119761467,
+      "num_tokens": 10149470.0,
       "step": 272
     },
     {
-      "entropy": 0.5366696864366531,
-      "epoch": 8.806451612903226,
-      "grad_norm": 0.03183649480342865,
+      "entropy": 0.5983473360538483,
+      "epoch": 2.116618075801749,
+      "grad_norm": 0.011125227436423302,
       "learning_rate": 0.0002,
-      "loss": 0.5360084772109985,
-      "mean_token_accuracy": 0.7832410633563995,
-      "num_tokens": 21474171.0,
+      "loss": 0.6070785522460938,
+      "mean_token_accuracy": 0.7521045431494713,
+      "num_tokens": 10187018.0,
       "step": 273
     },
     {
-      "entropy": 0.5488806217908859,
-      "epoch": 8.838709677419354,
-      "grad_norm": 0.03919295594096184,
+      "entropy": 0.6010436862707138,
+      "epoch": 2.1243926141885328,
+      "grad_norm": 0.008612287230789661,
       "learning_rate": 0.0002,
-      "loss": 0.5487998723983765,
-      "mean_token_accuracy": 0.7769176214933395,
-      "num_tokens": 21553568.0,
+      "loss": 0.6019182205200195,
+      "mean_token_accuracy": 0.7563453242182732,
+      "num_tokens": 10224534.0,
       "step": 274
     },
     {
-      "entropy": 0.5357145518064499,
-      "epoch": 8.870967741935484,
-      "grad_norm": 0.02984362095594406,
+      "entropy": 0.6097739785909653,
+      "epoch": 2.132167152575316,
+      "grad_norm": 0.009082912467420101,
       "learning_rate": 0.0002,
-      "loss": 0.5326351523399353,
-      "mean_token_accuracy": 0.7843077331781387,
-      "num_tokens": 21632757.0,
+      "loss": 0.6092417240142822,
+      "mean_token_accuracy": 0.7527365237474442,
+      "num_tokens": 10261606.0,
       "step": 275
     },
     {
-      "entropy": 0.5395043194293976,
-      "epoch": 8.903225806451612,
-      "grad_norm": 0.033653538674116135,
+      "entropy": 0.6300967186689377,
+      "epoch": 2.139941690962099,
+      "grad_norm": 0.00942782312631607,
       "learning_rate": 0.0002,
-      "loss": 0.536690354347229,
-      "mean_token_accuracy": 0.7822536379098892,
-      "num_tokens": 21712646.0,
+      "loss": 0.6258421540260315,
+      "mean_token_accuracy": 0.7469290718436241,
+      "num_tokens": 10299180.0,
       "step": 276
     },
     {
-      "entropy": 0.5327845886349678,
-      "epoch": 8.935483870967742,
-      "grad_norm": 0.035915497690439224,
+      "entropy": 0.6028410047292709,
+      "epoch": 2.1477162293488825,
+      "grad_norm": 0.008522949181497097,
       "learning_rate": 0.0002,
-      "loss": 0.5334969758987427,
-      "mean_token_accuracy": 0.7827408462762833,
-      "num_tokens": 21791364.0,
+      "loss": 0.6008127927780151,
+      "mean_token_accuracy": 0.7564279735088348,
+      "num_tokens": 10336360.0,
       "step": 277
     },
     {
-      "entropy": 0.5245150029659271,
-      "epoch": 8.967741935483872,
-      "grad_norm": 0.03674691170454025,
+      "entropy": 0.6102243736386299,
+      "epoch": 2.1554907677356656,
+      "grad_norm": 0.010031149722635746,
       "learning_rate": 0.0002,
-      "loss": 0.5283703804016113,
-      "mean_token_accuracy": 0.7849698960781097,
-      "num_tokens": 21869790.0,
+      "loss": 0.6199367046356201,
+      "mean_token_accuracy": 0.7469272315502167,
+      "num_tokens": 10373565.0,
       "step": 278
     },
     {
-      "entropy": 0.5436734557151794,
-      "epoch": 9.0,
-      "grad_norm": 0.03945446386933327,
+      "entropy": 0.6027829498052597,
+      "epoch": 2.163265306122449,
+      "grad_norm": 0.008753525093197823,
       "learning_rate": 0.0002,
-      "loss": 0.5423560738563538,
-      "mean_token_accuracy": 0.7805917263031006,
-      "num_tokens": 21949007.0,
+      "loss": 0.6074038147926331,
+      "mean_token_accuracy": 0.7534594535827637,
+      "num_tokens": 10411197.0,
       "step": 279
     },
     {
-      "entropy": 0.5403507351875305,
-      "epoch": 9.03225806451613,
-      "grad_norm": 0.040384337306022644,
+      "entropy": 0.6299364790320396,
+      "epoch": 2.1710398445092323,
+      "grad_norm": 0.007901565171778202,
       "learning_rate": 0.0002,
-      "loss": 0.5335516929626465,
-      "mean_token_accuracy": 0.784522145986557,
-      "num_tokens": 22026156.0,
+      "loss": 0.6327069401741028,
+      "mean_token_accuracy": 0.7429891973733902,
+      "num_tokens": 10448851.0,
       "step": 280
     },
     {
-      "entropy": 0.5205833613872528,
-      "epoch": 9.064516129032258,
-      "grad_norm": 0.03210374340415001,
+      "entropy": 0.623109444975853,
+      "epoch": 2.1788143828960154,
+      "grad_norm": 0.008401944302022457,
       "learning_rate": 0.0002,
-      "loss": 0.5193948745727539,
-      "mean_token_accuracy": 0.7887756377458572,
-      "num_tokens": 22104932.0,
+      "loss": 0.6217602491378784,
+      "mean_token_accuracy": 0.7446876764297485,
+      "num_tokens": 10486285.0,
       "step": 281
     },
     {
-      "entropy": 0.5186300128698349,
-      "epoch": 9.096774193548388,
-      "grad_norm": 0.043942417949438095,
+      "entropy": 0.5951671376824379,
+      "epoch": 2.186588921282799,
+      "grad_norm": 0.009740160778164864,
       "learning_rate": 0.0002,
-      "loss": 0.5202578902244568,
-      "mean_token_accuracy": 0.7897594273090363,
-      "num_tokens": 22182744.0,
+      "loss": 0.5930252075195312,
+      "mean_token_accuracy": 0.7590748071670532,
+      "num_tokens": 10523187.0,
       "step": 282
     },
     {
-      "entropy": 0.513887882232666,
-      "epoch": 9.129032258064516,
-      "grad_norm": 0.052492085844278336,
+      "entropy": 0.6063435301184654,
+      "epoch": 2.194363459669582,
+      "grad_norm": 0.008115135133266449,
       "learning_rate": 0.0002,
-      "loss": 0.5176864862442017,
-      "mean_token_accuracy": 0.790604293346405,
-      "num_tokens": 22262340.0,
+      "loss": 0.6058975458145142,
+      "mean_token_accuracy": 0.7546751350164413,
+      "num_tokens": 10560619.0,
       "step": 283
     },
     {
-      "entropy": 0.5307815670967102,
-      "epoch": 9.161290322580646,
-      "grad_norm": 0.050507497042417526,
+      "entropy": 0.6197424605488777,
+      "epoch": 2.2021379980563656,
+      "grad_norm": 0.007982614450156689,
       "learning_rate": 0.0002,
-      "loss": 0.5187040567398071,
-      "mean_token_accuracy": 0.7905398905277252,
-      "num_tokens": 22342518.0,
+      "loss": 0.6196328401565552,
+      "mean_token_accuracy": 0.7473800256848335,
+      "num_tokens": 10597952.0,
       "step": 284
     },
     {
-      "entropy": 0.5250707566738129,
-      "epoch": 9.193548387096774,
-      "grad_norm": 0.04122259095311165,
+      "entropy": 0.5975362882018089,
+      "epoch": 2.2099125364431487,
+      "grad_norm": 0.008934210054576397,
       "learning_rate": 0.0002,
-      "loss": 0.5248624086380005,
-      "mean_token_accuracy": 0.787539929151535,
-      "num_tokens": 22421363.0,
+      "loss": 0.6019341945648193,
+      "mean_token_accuracy": 0.7551193311810493,
+      "num_tokens": 10635474.0,
       "step": 285
     },
     {
-      "entropy": 0.5179962068796158,
-      "epoch": 9.225806451612904,
-      "grad_norm": 0.062456656247377396,
+      "entropy": 0.6057508885860443,
+      "epoch": 2.2176870748299318,
+      "grad_norm": 0.007894222624599934,
       "learning_rate": 0.0002,
-      "loss": 0.5286165475845337,
-      "mean_token_accuracy": 0.7862128764390945,
-      "num_tokens": 22500411.0,
+      "loss": 0.6054433584213257,
+      "mean_token_accuracy": 0.7555015385150909,
+      "num_tokens": 10672866.0,
       "step": 286
     },
     {
-      "entropy": 0.5425870567560196,
-      "epoch": 9.258064516129032,
-      "grad_norm": 0.05237641558051109,
+      "entropy": 0.6011307016015053,
+      "epoch": 2.2254616132167153,
+      "grad_norm": 0.008009341545403004,
       "learning_rate": 0.0002,
-      "loss": 0.5357320308685303,
-      "mean_token_accuracy": 0.7818474918603897,
-      "num_tokens": 22578954.0,
+      "loss": 0.6015121936798096,
+      "mean_token_accuracy": 0.7533531039953232,
+      "num_tokens": 10710119.0,
       "step": 287
     },
     {
-      "entropy": 0.5489015877246857,
-      "epoch": 9.290322580645162,
-      "grad_norm": 0.04592309519648552,
+      "entropy": 0.6083035543560982,
+      "epoch": 2.2332361516034984,
+      "grad_norm": 0.008571778424084187,
       "learning_rate": 0.0002,
-      "loss": 0.5391058921813965,
-      "mean_token_accuracy": 0.781953975558281,
-      "num_tokens": 22656865.0,
+      "loss": 0.6072158813476562,
+      "mean_token_accuracy": 0.7525660693645477,
+      "num_tokens": 10747533.0,
       "step": 288
     },
     {
-      "entropy": 0.5262634605169296,
-      "epoch": 9.32258064516129,
-      "grad_norm": 0.04665505513548851,
+      "entropy": 0.596297912299633,
+      "epoch": 2.241010689990282,
+      "grad_norm": 0.007835134863853455,
       "learning_rate": 0.0002,
-      "loss": 0.5303994417190552,
-      "mean_token_accuracy": 0.784704178571701,
-      "num_tokens": 22734848.0,
+      "loss": 0.6007227897644043,
+      "mean_token_accuracy": 0.75400510430336,
+      "num_tokens": 10784600.0,
       "step": 289
     },
     {
-      "entropy": 0.5207262188196182,
-      "epoch": 9.35483870967742,
-      "grad_norm": 0.06092476472258568,
+      "entropy": 0.6027623787522316,
+      "epoch": 2.248785228377065,
+      "grad_norm": 0.00971238687634468,
       "learning_rate": 0.0002,
-      "loss": 0.5245224237442017,
-      "mean_token_accuracy": 0.7855150699615479,
-      "num_tokens": 22813242.0,
+      "loss": 0.6095942258834839,
+      "mean_token_accuracy": 0.753618136048317,
+      "num_tokens": 10822085.0,
       "step": 290
     },
     {
-      "entropy": 0.5316435396671295,
-      "epoch": 9.387096774193548,
-      "grad_norm": 0.05380527302622795,
+      "entropy": 0.596187099814415,
+      "epoch": 2.256559766763848,
+      "grad_norm": 0.008036565966904163,
       "learning_rate": 0.0002,
-      "loss": 0.5261737108230591,
-      "mean_token_accuracy": 0.7874870747327805,
-      "num_tokens": 22892873.0,
+      "loss": 0.5973517298698425,
+      "mean_token_accuracy": 0.7577640637755394,
+      "num_tokens": 10859362.0,
       "step": 291
     },
     {
-      "entropy": 0.5344062894582748,
-      "epoch": 9.419354838709678,
-      "grad_norm": 0.04123701527714729,
+      "entropy": 0.6243500411510468,
+      "epoch": 2.2643343051506317,
+      "grad_norm": 0.00934487022459507,
       "learning_rate": 0.0002,
-      "loss": 0.5327101945877075,
-      "mean_token_accuracy": 0.7833229005336761,
-      "num_tokens": 22971116.0,
+      "loss": 0.6226072311401367,
+      "mean_token_accuracy": 0.7454966679215431,
+      "num_tokens": 10896757.0,
       "step": 292
     },
     {
-      "entropy": 0.524382010102272,
-      "epoch": 9.451612903225806,
-      "grad_norm": 0.08812697976827621,
+      "entropy": 0.6136296838521957,
+      "epoch": 2.272108843537415,
+      "grad_norm": 0.008677436038851738,
       "learning_rate": 0.0002,
-      "loss": 0.5324304103851318,
-      "mean_token_accuracy": 0.7840967774391174,
-      "num_tokens": 23049380.0,
+      "loss": 0.615119457244873,
+      "mean_token_accuracy": 0.750088632106781,
+      "num_tokens": 10934016.0,
       "step": 293
     },
     {
-      "entropy": 0.5306137502193451,
-      "epoch": 9.483870967741936,
-      "grad_norm": 0.08938439935445786,
+      "entropy": 0.5983113646507263,
+      "epoch": 2.2798833819241984,
+      "grad_norm": 0.007704727817326784,
       "learning_rate": 0.0002,
-      "loss": 0.5269825458526611,
-      "mean_token_accuracy": 0.7875665426254272,
-      "num_tokens": 23128218.0,
+      "loss": 0.5992428064346313,
+      "mean_token_accuracy": 0.7575259953737259,
+      "num_tokens": 10971732.0,
       "step": 294
     },
     {
-      "entropy": 0.5367314219474792,
-      "epoch": 9.516129032258064,
-      "grad_norm": 0.05447512865066528,
+      "entropy": 0.602768175303936,
+      "epoch": 2.2876579203109815,
+      "grad_norm": 0.008800746873021126,
       "learning_rate": 0.0002,
-      "loss": 0.5286938548088074,
-      "mean_token_accuracy": 0.7865005433559418,
-      "num_tokens": 23207525.0,
+      "loss": 0.6059098243713379,
+      "mean_token_accuracy": 0.7514860853552818,
+      "num_tokens": 11009363.0,
       "step": 295
     },
     {
-      "entropy": 0.5252849012613297,
-      "epoch": 9.548387096774194,
-      "grad_norm": 0.06832025945186615,
+      "entropy": 0.6180300787091255,
+      "epoch": 2.295432458697765,
+      "grad_norm": 0.009863065555691719,
       "learning_rate": 0.0002,
-      "loss": 0.5257875323295593,
-      "mean_token_accuracy": 0.7862931191921234,
-      "num_tokens": 23287210.0,
+      "loss": 0.6232578754425049,
+      "mean_token_accuracy": 0.7460011914372444,
+      "num_tokens": 11046850.0,
       "step": 296
     },
     {
-      "entropy": 0.5178247690200806,
-      "epoch": 9.580645161290322,
-      "grad_norm": 0.09167063236236572,
+      "entropy": 0.615628756582737,
+      "epoch": 2.303206997084548,
+      "grad_norm": 0.00794578343629837,
       "learning_rate": 0.0002,
-      "loss": 0.5258931517601013,
-      "mean_token_accuracy": 0.7874697148799896,
-      "num_tokens": 23366831.0,
+      "loss": 0.6193504929542542,
+      "mean_token_accuracy": 0.7466345354914665,
+      "num_tokens": 11084028.0,
       "step": 297
     },
     {
-      "entropy": 0.5381314009428024,
-      "epoch": 9.612903225806452,
-      "grad_norm": 0.05428946390748024,
+      "entropy": 0.6079032048583031,
+      "epoch": 2.3109815354713312,
+      "grad_norm": 0.008673022501170635,
       "learning_rate": 0.0002,
-      "loss": 0.5342376232147217,
-      "mean_token_accuracy": 0.7819546610116959,
-      "num_tokens": 23445938.0,
+      "loss": 0.6060619354248047,
+      "mean_token_accuracy": 0.7532483711838722,
+      "num_tokens": 11121691.0,
       "step": 298
     },
     {
-      "entropy": 0.5478924512863159,
-      "epoch": 9.64516129032258,
-      "grad_norm": 0.06756260246038437,
+      "entropy": 0.6126295253634453,
+      "epoch": 2.3187560738581148,
+      "grad_norm": 0.010211586020886898,
       "learning_rate": 0.0002,
-      "loss": 0.5411292314529419,
-      "mean_token_accuracy": 0.781807467341423,
-      "num_tokens": 23523910.0,
+      "loss": 0.6097185611724854,
+      "mean_token_accuracy": 0.7503807097673416,
+      "num_tokens": 11159118.0,
       "step": 299
     },
     {
-      "entropy": 0.5383109152317047,
-      "epoch": 9.67741935483871,
-      "grad_norm": 0.07821021974086761,
+      "entropy": 0.6223906055092812,
+      "epoch": 2.326530612244898,
+      "grad_norm": 0.007925095036625862,
       "learning_rate": 0.0002,
-      "loss": 0.5393465757369995,
-      "mean_token_accuracy": 0.7826394438743591,
-      "num_tokens": 23602935.0,
+      "loss": 0.6250431537628174,
+      "mean_token_accuracy": 0.7459972500801086,
+      "num_tokens": 11196522.0,
       "step": 300
     },
     {
-      "entropy": 0.5300856381654739,
-      "epoch": 9.709677419354838,
-      "grad_norm": 0.05791868641972542,
+      "entropy": 0.6105447709560394,
+      "epoch": 2.3343051506316814,
+      "grad_norm": 0.008913103491067886,
       "learning_rate": 0.0002,
-      "loss": 0.5366538763046265,
-      "mean_token_accuracy": 0.7823646813631058,
-      "num_tokens": 23680706.0,
+      "loss": 0.616736114025116,
+      "mean_token_accuracy": 0.749770350754261,
+      "num_tokens": 11234173.0,
       "step": 301
     },
     {
-      "entropy": 0.5394613444805145,
-      "epoch": 9.741935483870968,
-      "grad_norm": 0.07757086306810379,
+      "entropy": 0.6050237938761711,
+      "epoch": 2.3420796890184645,
+      "grad_norm": 0.00913191493600607,
       "learning_rate": 0.0002,
-      "loss": 0.5367089509963989,
-      "mean_token_accuracy": 0.78324094414711,
-      "num_tokens": 23759268.0,
+      "loss": 0.6078106164932251,
+      "mean_token_accuracy": 0.7534833922982216,
+      "num_tokens": 11271528.0,
       "step": 302
     },
     {
-      "entropy": 0.5447310507297516,
-      "epoch": 9.774193548387096,
-      "grad_norm": 0.049349840730428696,
+      "entropy": 0.624389261007309,
+      "epoch": 2.3498542274052476,
+      "grad_norm": 0.00800058338791132,
       "learning_rate": 0.0002,
-      "loss": 0.5364006757736206,
-      "mean_token_accuracy": 0.7828964740037918,
-      "num_tokens": 23837981.0,
+      "loss": 0.624384880065918,
+      "mean_token_accuracy": 0.7484247088432312,
+      "num_tokens": 11309275.0,
       "step": 303
     },
     {
-      "entropy": 0.5333449840545654,
-      "epoch": 9.806451612903226,
-      "grad_norm": 0.06826647371053696,
+      "entropy": 0.621875673532486,
+      "epoch": 2.357628765792031,
+      "grad_norm": 0.009797596372663975,
       "learning_rate": 0.0002,
-      "loss": 0.53536057472229,
-      "mean_token_accuracy": 0.7834214121103287,
-      "num_tokens": 23915246.0,
+      "loss": 0.6190812587738037,
+      "mean_token_accuracy": 0.7491643279790878,
+      "num_tokens": 11346386.0,
       "step": 304
     },
     {
-      "entropy": 0.5236239582300186,
-      "epoch": 9.838709677419354,
-      "grad_norm": 0.06063172593712807,
+      "entropy": 0.6205748915672302,
+      "epoch": 2.3654033041788143,
+      "grad_norm": 0.009043901227414608,
       "learning_rate": 0.0002,
-      "loss": 0.526973307132721,
-      "mean_token_accuracy": 0.7862582057714462,
-      "num_tokens": 23993749.0,
+      "loss": 0.616840124130249,
+      "mean_token_accuracy": 0.7496623322367668,
+      "num_tokens": 11383273.0,
       "step": 305
     },
     {
-      "entropy": 0.5442259162664413,
-      "epoch": 9.870967741935484,
-      "grad_norm": 0.0450434610247612,
+      "entropy": 0.5943257734179497,
+      "epoch": 2.373177842565598,
+      "grad_norm": 0.009680100716650486,
       "learning_rate": 0.0002,
-      "loss": 0.5387926697731018,
-      "mean_token_accuracy": 0.7815881818532944,
-      "num_tokens": 24072030.0,
+      "loss": 0.6005733013153076,
+      "mean_token_accuracy": 0.7566492632031441,
+      "num_tokens": 11420226.0,
       "step": 306
     },
     {
-      "entropy": 0.5275344997644424,
-      "epoch": 9.903225806451612,
-      "grad_norm": 0.04751954600214958,
+      "entropy": 0.5875989943742752,
+      "epoch": 2.380952380952381,
+      "grad_norm": 0.009831924922764301,
       "learning_rate": 0.0002,
-      "loss": 0.5248628258705139,
-      "mean_token_accuracy": 0.7874411344528198,
-      "num_tokens": 24152020.0,
+      "loss": 0.5920431613922119,
+      "mean_token_accuracy": 0.758441336452961,
+      "num_tokens": 11457658.0,
       "step": 307
     },
     {
-      "entropy": 0.5380201190710068,
-      "epoch": 9.935483870967742,
-      "grad_norm": 0.04302695393562317,
+      "entropy": 0.605907216668129,
+      "epoch": 2.388726919339164,
+      "grad_norm": 0.008310995064675808,
       "learning_rate": 0.0002,
-      "loss": 0.538284420967102,
-      "mean_token_accuracy": 0.7823643535375595,
-      "num_tokens": 24230082.0,
+      "loss": 0.606610894203186,
+      "mean_token_accuracy": 0.7539205178618431,
+      "num_tokens": 11495093.0,
       "step": 308
     },
     {
-      "entropy": 0.5229910910129547,
-      "epoch": 9.967741935483872,
-      "grad_norm": 0.0525338277220726,
+      "entropy": 0.6085820719599724,
+      "epoch": 2.3965014577259476,
+      "grad_norm": 0.008422457613050938,
       "learning_rate": 0.0002,
-      "loss": 0.5264443159103394,
-      "mean_token_accuracy": 0.786424458026886,
-      "num_tokens": 24309618.0,
+      "loss": 0.6112357378005981,
+      "mean_token_accuracy": 0.7495599314570427,
+      "num_tokens": 11532305.0,
       "step": 309
     },
     {
-      "entropy": 0.5351384729146957,
-      "epoch": 10.0,
-      "grad_norm": 0.04153383523225784,
+      "entropy": 0.6038077920675278,
+      "epoch": 2.4042759961127307,
+      "grad_norm": 0.009313938207924366,
       "learning_rate": 0.0002,
-      "loss": 0.5313504934310913,
-      "mean_token_accuracy": 0.7846148610115051,
-      "num_tokens": 24387187.0,
+      "loss": 0.6037260293960571,
+      "mean_token_accuracy": 0.7537373602390289,
+      "num_tokens": 11569956.0,
       "step": 310
+    },
+    {
+      "entropy": 0.6088597327470779,
+      "epoch": 2.4120505344995142,
+      "grad_norm": 0.008129115216434002,
+      "learning_rate": 0.0002,
+      "loss": 0.6097284555435181,
+      "mean_token_accuracy": 0.752282939851284,
+      "num_tokens": 11607580.0,
+      "step": 311
+    },
+    {
+      "entropy": 0.6035630702972412,
+      "epoch": 2.4198250728862973,
+      "grad_norm": 0.010189153254032135,
+      "learning_rate": 0.0002,
+      "loss": 0.6098878383636475,
+      "mean_token_accuracy": 0.7509205341339111,
+      "num_tokens": 11645065.0,
+      "step": 312
+    },
+    {
+      "entropy": 0.5988744720816612,
+      "epoch": 2.427599611273081,
+      "grad_norm": 0.008477658964693546,
+      "learning_rate": 0.0002,
+      "loss": 0.600048840045929,
+      "mean_token_accuracy": 0.7566671743988991,
+      "num_tokens": 11682523.0,
+      "step": 313
+    },
+    {
+      "entropy": 0.604704961180687,
+      "epoch": 2.435374149659864,
+      "grad_norm": 0.009996007196605206,
+      "learning_rate": 0.0002,
+      "loss": 0.6124827861785889,
+      "mean_token_accuracy": 0.7483595088124275,
+      "num_tokens": 11719829.0,
+      "step": 314
+    },
+    {
+      "entropy": 0.6142484471201897,
+      "epoch": 2.443148688046647,
+      "grad_norm": 0.008275930769741535,
+      "learning_rate": 0.0002,
+      "loss": 0.6168549060821533,
+      "mean_token_accuracy": 0.7509273141622543,
+      "num_tokens": 11757130.0,
+      "step": 315
+    },
+    {
+      "entropy": 0.618528425693512,
+      "epoch": 2.4509232264334306,
+      "grad_norm": 0.00913267582654953,
+      "learning_rate": 0.0002,
+      "loss": 0.6144739985466003,
+      "mean_token_accuracy": 0.7495302185416222,
+      "num_tokens": 11794646.0,
+      "step": 316
+    },
+    {
+      "entropy": 0.6221663355827332,
+      "epoch": 2.4586977648202137,
+      "grad_norm": 0.008833467960357666,
+      "learning_rate": 0.0002,
+      "loss": 0.6160144805908203,
+      "mean_token_accuracy": 0.7488524913787842,
+      "num_tokens": 11832667.0,
+      "step": 317
+    },
+    {
+      "entropy": 0.6151534616947174,
+      "epoch": 2.466472303206997,
+      "grad_norm": 0.008367245085537434,
+      "learning_rate": 0.0002,
+      "loss": 0.6178625226020813,
+      "mean_token_accuracy": 0.7469210624694824,
+      "num_tokens": 11870196.0,
+      "step": 318
+    },
+    {
+      "entropy": 0.599577471613884,
+      "epoch": 2.4742468415937804,
+      "grad_norm": 0.008229793980717659,
+      "learning_rate": 0.0002,
+      "loss": 0.6013622283935547,
+      "mean_token_accuracy": 0.7555016428232193,
+      "num_tokens": 11907738.0,
+      "step": 319
+    },
+    {
+      "entropy": 0.6037501096725464,
+      "epoch": 2.4820213799805635,
+      "grad_norm": 0.009975764900445938,
+      "learning_rate": 0.0002,
+      "loss": 0.6080362796783447,
+      "mean_token_accuracy": 0.754118837416172,
+      "num_tokens": 11945154.0,
+      "step": 320
+    },
+    {
+      "entropy": 0.6234212592244148,
+      "epoch": 2.489795918367347,
+      "grad_norm": 0.009158926084637642,
+      "learning_rate": 0.0002,
+      "loss": 0.6259138584136963,
+      "mean_token_accuracy": 0.7461888268589973,
+      "num_tokens": 11983030.0,
+      "step": 321
+    },
+    {
+      "entropy": 0.6105703264474869,
+      "epoch": 2.49757045675413,
+      "grad_norm": 0.009549647569656372,
+      "learning_rate": 0.0002,
+      "loss": 0.6138617992401123,
+      "mean_token_accuracy": 0.7500449195504189,
+      "num_tokens": 12020106.0,
+      "step": 322
+    },
+    {
+      "entropy": 0.5886913314461708,
+      "epoch": 2.5053449951409137,
+      "grad_norm": 0.009142185561358929,
+      "learning_rate": 0.0002,
+      "loss": 0.5951993465423584,
+      "mean_token_accuracy": 0.7568985298275948,
+      "num_tokens": 12057274.0,
+      "step": 323
+    },
+    {
+      "entropy": 0.6163628548383713,
+      "epoch": 2.513119533527697,
+      "grad_norm": 0.008757554925978184,
+      "learning_rate": 0.0002,
+      "loss": 0.6170852184295654,
+      "mean_token_accuracy": 0.749226838350296,
+      "num_tokens": 12094665.0,
+      "step": 324
+    },
+    {
+      "entropy": 0.6229220703244209,
+      "epoch": 2.52089407191448,
+      "grad_norm": 0.007944419980049133,
+      "learning_rate": 0.0002,
+      "loss": 0.6211696267127991,
+      "mean_token_accuracy": 0.749010942876339,
+      "num_tokens": 12132302.0,
+      "step": 325
+    },
+    {
+      "entropy": 0.6257840767502785,
+      "epoch": 2.5286686103012634,
+      "grad_norm": 0.008713958784937859,
+      "learning_rate": 0.0002,
+      "loss": 0.6223682165145874,
+      "mean_token_accuracy": 0.7470938488841057,
+      "num_tokens": 12169510.0,
+      "step": 326
+    },
+    {
+      "entropy": 0.6103069111704826,
+      "epoch": 2.5364431486880465,
+      "grad_norm": 0.008112641051411629,
+      "learning_rate": 0.0002,
+      "loss": 0.6112810969352722,
+      "mean_token_accuracy": 0.7517933994531631,
+      "num_tokens": 12206878.0,
+      "step": 327
+    },
+    {
+      "entropy": 0.6295960918068886,
+      "epoch": 2.54421768707483,
+      "grad_norm": 0.009597997181117535,
+      "learning_rate": 0.0002,
+      "loss": 0.6329588890075684,
+      "mean_token_accuracy": 0.7417808771133423,
+      "num_tokens": 12244230.0,
+      "step": 328
+    },
+    {
+      "entropy": 0.6144092008471489,
+      "epoch": 2.551992225461613,
+      "grad_norm": 0.007173395249992609,
+      "learning_rate": 0.0002,
+      "loss": 0.6167792081832886,
+      "mean_token_accuracy": 0.750128723680973,
+      "num_tokens": 12281944.0,
+      "step": 329
+    },
+    {
+      "entropy": 0.6102811768651009,
+      "epoch": 2.5597667638483967,
+      "grad_norm": 0.00874980166554451,
+      "learning_rate": 0.0002,
+      "loss": 0.6081819534301758,
+      "mean_token_accuracy": 0.7545690611004829,
+      "num_tokens": 12319144.0,
+      "step": 330
+    },
+    {
+      "entropy": 0.6088513135910034,
+      "epoch": 2.56754130223518,
+      "grad_norm": 0.007437287364155054,
+      "learning_rate": 0.0002,
+      "loss": 0.610903799533844,
+      "mean_token_accuracy": 0.7525679916143417,
+      "num_tokens": 12356752.0,
+      "step": 331
+    },
+    {
+      "entropy": 0.605910450220108,
+      "epoch": 2.575315840621963,
+      "grad_norm": 0.008948191069066525,
+      "learning_rate": 0.0002,
+      "loss": 0.6137778759002686,
+      "mean_token_accuracy": 0.7502530664205551,
+      "num_tokens": 12394388.0,
+      "step": 332
+    },
+    {
+      "entropy": 0.6148117780685425,
+      "epoch": 2.5830903790087465,
+      "grad_norm": 0.008310764096677303,
+      "learning_rate": 0.0002,
+      "loss": 0.6203396320343018,
+      "mean_token_accuracy": 0.7478612065315247,
+      "num_tokens": 12431583.0,
+      "step": 333
+    },
+    {
+      "entropy": 0.6049426943063736,
+      "epoch": 2.5908649173955296,
+      "grad_norm": 0.00928961019963026,
+      "learning_rate": 0.0002,
+      "loss": 0.6102631688117981,
+      "mean_token_accuracy": 0.7530224993824959,
+      "num_tokens": 12469057.0,
+      "step": 334
+    },
+    {
+      "entropy": 0.6105730906128883,
+      "epoch": 2.5986394557823127,
+      "grad_norm": 0.008061802014708519,
+      "learning_rate": 0.0002,
+      "loss": 0.6099789142608643,
+      "mean_token_accuracy": 0.7517432495951653,
+      "num_tokens": 12506375.0,
+      "step": 335
+    },
+    {
+      "entropy": 0.6109907403588295,
+      "epoch": 2.6064139941690962,
+      "grad_norm": 0.008131072856485844,
+      "learning_rate": 0.0002,
+      "loss": 0.6083254814147949,
+      "mean_token_accuracy": 0.7539641037583351,
+      "num_tokens": 12543683.0,
+      "step": 336
+    },
+    {
+      "entropy": 0.60856644064188,
+      "epoch": 2.6141885325558794,
+      "grad_norm": 0.008039598353207111,
+      "learning_rate": 0.0002,
+      "loss": 0.6059796810150146,
+      "mean_token_accuracy": 0.7539034485816956,
+      "num_tokens": 12581164.0,
+      "step": 337
+    },
+    {
+      "entropy": 0.6075717508792877,
+      "epoch": 2.621963070942663,
+      "grad_norm": 0.007622460834681988,
+      "learning_rate": 0.0002,
+      "loss": 0.6058254241943359,
+      "mean_token_accuracy": 0.755204826593399,
+      "num_tokens": 12618385.0,
+      "step": 338
+    },
+    {
+      "entropy": 0.6058209240436554,
+      "epoch": 2.629737609329446,
+      "grad_norm": 0.009571490809321404,
+      "learning_rate": 0.0002,
+      "loss": 0.6132853031158447,
+      "mean_token_accuracy": 0.7517998144030571,
+      "num_tokens": 12656198.0,
+      "step": 339
+    },
+    {
+      "entropy": 0.609808899462223,
+      "epoch": 2.6375121477162295,
+      "grad_norm": 0.009121098555624485,
+      "learning_rate": 0.0002,
+      "loss": 0.6116859912872314,
+      "mean_token_accuracy": 0.7522552087903023,
+      "num_tokens": 12693619.0,
+      "step": 340
+    },
+    {
+      "entropy": 0.5982947275042534,
+      "epoch": 2.6452866861030127,
+      "grad_norm": 0.007683332543820143,
+      "learning_rate": 0.0002,
+      "loss": 0.6035135984420776,
+      "mean_token_accuracy": 0.7554004937410355,
+      "num_tokens": 12731247.0,
+      "step": 341
+    },
+    {
+      "entropy": 0.6199963241815567,
+      "epoch": 2.6530612244897958,
+      "grad_norm": 0.008116503246128559,
+      "learning_rate": 0.0002,
+      "loss": 0.6209834814071655,
+      "mean_token_accuracy": 0.7469918876886368,
+      "num_tokens": 12768639.0,
+      "step": 342
+    },
+    {
+      "entropy": 0.611749030649662,
+      "epoch": 2.6608357628765793,
+      "grad_norm": 0.009202837944030762,
+      "learning_rate": 0.0002,
+      "loss": 0.609596848487854,
+      "mean_token_accuracy": 0.7527265250682831,
+      "num_tokens": 12805843.0,
+      "step": 343
+    },
+    {
+      "entropy": 0.6049032211303711,
+      "epoch": 2.6686103012633624,
+      "grad_norm": 0.008622320368885994,
+      "learning_rate": 0.0002,
+      "loss": 0.6062831282615662,
+      "mean_token_accuracy": 0.7529996708035469,
+      "num_tokens": 12843334.0,
+      "step": 344
+    },
+    {
+      "entropy": 0.6080296412110329,
+      "epoch": 2.6763848396501455,
+      "grad_norm": 0.010344683192670345,
+      "learning_rate": 0.0002,
+      "loss": 0.6155232787132263,
+      "mean_token_accuracy": 0.7492252364754677,
+      "num_tokens": 12881114.0,
+      "step": 345
+    },
+    {
+      "entropy": 0.6105146557092667,
+      "epoch": 2.684159378036929,
+      "grad_norm": 0.00811974797397852,
+      "learning_rate": 0.0002,
+      "loss": 0.61601722240448,
+      "mean_token_accuracy": 0.7480410560965538,
+      "num_tokens": 12918628.0,
+      "step": 346
+    },
+    {
+      "entropy": 0.616872251033783,
+      "epoch": 2.6919339164237126,
+      "grad_norm": 0.009087933227419853,
+      "learning_rate": 0.0002,
+      "loss": 0.6140284538269043,
+      "mean_token_accuracy": 0.7496198862791061,
+      "num_tokens": 12955856.0,
+      "step": 347
+    },
+    {
+      "entropy": 0.6135741174221039,
+      "epoch": 2.6997084548104957,
+      "grad_norm": 0.008728091605007648,
+      "learning_rate": 0.0002,
+      "loss": 0.6116735935211182,
+      "mean_token_accuracy": 0.7527862265706062,
+      "num_tokens": 12993282.0,
+      "step": 348
+    },
+    {
+      "entropy": 0.6206527948379517,
+      "epoch": 2.707482993197279,
+      "grad_norm": 0.008598288521170616,
+      "learning_rate": 0.0002,
+      "loss": 0.6233333349227905,
+      "mean_token_accuracy": 0.7463840544223785,
+      "num_tokens": 13030919.0,
+      "step": 349
+    },
+    {
+      "entropy": 0.608021192252636,
+      "epoch": 2.7152575315840624,
+      "grad_norm": 0.008716718293726444,
+      "learning_rate": 0.0002,
+      "loss": 0.612204909324646,
+      "mean_token_accuracy": 0.7511651292443275,
+      "num_tokens": 13068022.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.6055907234549522,
+      "epoch": 2.7230320699708455,
+      "grad_norm": 0.007125901058316231,
+      "learning_rate": 0.0002,
+      "loss": 0.6055992841720581,
+      "mean_token_accuracy": 0.753873273730278,
+      "num_tokens": 13105324.0,
+      "step": 351
+    },
+    {
+      "entropy": 0.6134061738848686,
+      "epoch": 2.7308066083576286,
+      "grad_norm": 0.007964730262756348,
+      "learning_rate": 0.0002,
+      "loss": 0.6122632026672363,
+      "mean_token_accuracy": 0.7516495808959007,
+      "num_tokens": 13143067.0,
+      "step": 352
+    },
+    {
+      "entropy": 0.6035361513495445,
+      "epoch": 2.738581146744412,
+      "grad_norm": 0.008376477286219597,
+      "learning_rate": 0.0002,
+      "loss": 0.6055692434310913,
+      "mean_token_accuracy": 0.7536868900060654,
+      "num_tokens": 13180307.0,
+      "step": 353
+    },
+    {
+      "entropy": 0.6108582690358162,
+      "epoch": 2.746355685131195,
+      "grad_norm": 0.008256555534899235,
+      "learning_rate": 0.0002,
+      "loss": 0.6159192323684692,
+      "mean_token_accuracy": 0.7500140145421028,
+      "num_tokens": 13217651.0,
+      "step": 354
+    },
+    {
+      "entropy": 0.6065188273787498,
+      "epoch": 2.7541302235179788,
+      "grad_norm": 0.007910750806331635,
+      "learning_rate": 0.0002,
+      "loss": 0.6087648868560791,
+      "mean_token_accuracy": 0.7539975792169571,
+      "num_tokens": 13255245.0,
+      "step": 355
+    },
+    {
+      "entropy": 0.6058445647358894,
+      "epoch": 2.761904761904762,
+      "grad_norm": 0.007995028980076313,
+      "learning_rate": 0.0002,
+      "loss": 0.6082870960235596,
+      "mean_token_accuracy": 0.7520255744457245,
+      "num_tokens": 13293244.0,
+      "step": 356
+    },
+    {
+      "entropy": 0.6188160851597786,
+      "epoch": 2.7696793002915454,
+      "grad_norm": 0.008255942724645138,
+      "learning_rate": 0.0002,
+      "loss": 0.615801990032196,
+      "mean_token_accuracy": 0.7498086243867874,
+      "num_tokens": 13330234.0,
+      "step": 357
+    },
+    {
+      "entropy": 0.6236889883875847,
+      "epoch": 2.7774538386783285,
+      "grad_norm": 0.007726432289928198,
+      "learning_rate": 0.0002,
+      "loss": 0.6215789914131165,
+      "mean_token_accuracy": 0.7462954670190811,
+      "num_tokens": 13367753.0,
+      "step": 358
+    },
+    {
+      "entropy": 0.6057674139738083,
+      "epoch": 2.7852283770651116,
+      "grad_norm": 0.007486944552510977,
+      "learning_rate": 0.0002,
+      "loss": 0.6107175946235657,
+      "mean_token_accuracy": 0.7536005079746246,
+      "num_tokens": 13405367.0,
+      "step": 359
+    },
+    {
+      "entropy": 0.6108546778559685,
+      "epoch": 2.793002915451895,
+      "grad_norm": 0.00800468772649765,
+      "learning_rate": 0.0002,
+      "loss": 0.6151923537254333,
+      "mean_token_accuracy": 0.7498300299048424,
+      "num_tokens": 13442458.0,
+      "step": 360
+    },
+    {
+      "entropy": 0.611615277826786,
+      "epoch": 2.8007774538386783,
+      "grad_norm": 0.008358453400433064,
+      "learning_rate": 0.0002,
+      "loss": 0.6143432855606079,
+      "mean_token_accuracy": 0.7498634308576584,
+      "num_tokens": 13479706.0,
+      "step": 361
+    },
+    {
+      "entropy": 0.6097937971353531,
+      "epoch": 2.8085519922254614,
+      "grad_norm": 0.008056551218032837,
+      "learning_rate": 0.0002,
+      "loss": 0.6140632629394531,
+      "mean_token_accuracy": 0.7493433877825737,
+      "num_tokens": 13517106.0,
+      "step": 362
+    },
+    {
+      "entropy": 0.6036530286073685,
+      "epoch": 2.816326530612245,
+      "grad_norm": 0.008669313974678516,
+      "learning_rate": 0.0002,
+      "loss": 0.604174792766571,
+      "mean_token_accuracy": 0.755392961204052,
+      "num_tokens": 13554029.0,
+      "step": 363
+    },
+    {
+      "entropy": 0.6074316874146461,
+      "epoch": 2.824101068999028,
+      "grad_norm": 0.008313254453241825,
+      "learning_rate": 0.0002,
+      "loss": 0.6034117937088013,
+      "mean_token_accuracy": 0.7538974210619926,
+      "num_tokens": 13591496.0,
+      "step": 364
+    },
+    {
+      "entropy": 0.625335082411766,
+      "epoch": 2.8318756073858116,
+      "grad_norm": 0.00826285034418106,
+      "learning_rate": 0.0002,
+      "loss": 0.6179195642471313,
+      "mean_token_accuracy": 0.7473107278347015,
+      "num_tokens": 13629175.0,
+      "step": 365
+    },
+    {
+      "entropy": 0.6023752987384796,
+      "epoch": 2.8396501457725947,
+      "grad_norm": 0.008760266937315464,
+      "learning_rate": 0.0002,
+      "loss": 0.6035041809082031,
+      "mean_token_accuracy": 0.752547599375248,
+      "num_tokens": 13666378.0,
+      "step": 366
+    },
+    {
+      "entropy": 0.6186700463294983,
+      "epoch": 2.847424684159378,
+      "grad_norm": 0.009422390721738338,
+      "learning_rate": 0.0002,
+      "loss": 0.6255558133125305,
+      "mean_token_accuracy": 0.7451329976320267,
+      "num_tokens": 13704147.0,
+      "step": 367
+    },
+    {
+      "entropy": 0.6027014032006264,
+      "epoch": 2.8551992225461613,
+      "grad_norm": 0.007943754084408283,
+      "learning_rate": 0.0002,
+      "loss": 0.6077042818069458,
+      "mean_token_accuracy": 0.7518687173724174,
+      "num_tokens": 13741252.0,
+      "step": 368
+    },
+    {
+      "entropy": 0.6081340312957764,
+      "epoch": 2.8629737609329444,
+      "grad_norm": 0.008472657762467861,
+      "learning_rate": 0.0002,
+      "loss": 0.6082110404968262,
+      "mean_token_accuracy": 0.7499295845627785,
+      "num_tokens": 13778637.0,
+      "step": 369
+    },
+    {
+      "entropy": 0.6041673645377159,
+      "epoch": 2.870748299319728,
+      "grad_norm": 0.008525537326931953,
+      "learning_rate": 0.0002,
+      "loss": 0.6059847474098206,
+      "mean_token_accuracy": 0.7531944662332535,
+      "num_tokens": 13816076.0,
+      "step": 370
+    },
+    {
+      "entropy": 0.6156945377588272,
+      "epoch": 2.878522837706511,
+      "grad_norm": 0.007397054228931665,
+      "learning_rate": 0.0002,
+      "loss": 0.6175694465637207,
+      "mean_token_accuracy": 0.7489196062088013,
+      "num_tokens": 13853700.0,
+      "step": 371
+    },
+    {
+      "entropy": 0.6109963580965996,
+      "epoch": 2.8862973760932946,
+      "grad_norm": 0.008632549084722996,
+      "learning_rate": 0.0002,
+      "loss": 0.6189043521881104,
+      "mean_token_accuracy": 0.7480105683207512,
+      "num_tokens": 13890909.0,
+      "step": 372
+    },
+    {
+      "entropy": 0.60771045088768,
+      "epoch": 2.8940719144800777,
+      "grad_norm": 0.008767201565206051,
+      "learning_rate": 0.0002,
+      "loss": 0.6133548021316528,
+      "mean_token_accuracy": 0.7522530779242516,
+      "num_tokens": 13928292.0,
+      "step": 373
+    },
+    {
+      "entropy": 0.6067496612668037,
+      "epoch": 2.9018464528668613,
+      "grad_norm": 0.007676385343074799,
+      "learning_rate": 0.0002,
+      "loss": 0.60523521900177,
+      "mean_token_accuracy": 0.754976786673069,
+      "num_tokens": 13965803.0,
+      "step": 374
+    },
+    {
+      "entropy": 0.6119726821780205,
+      "epoch": 2.9096209912536444,
+      "grad_norm": 0.007059477269649506,
+      "learning_rate": 0.0002,
+      "loss": 0.6106954216957092,
+      "mean_token_accuracy": 0.7522552087903023,
+      "num_tokens": 14002852.0,
+      "step": 375
+    },
+    {
+      "entropy": 0.6248558238148689,
+      "epoch": 2.9173955296404275,
+      "grad_norm": 0.008261643350124359,
+      "learning_rate": 0.0002,
+      "loss": 0.6251908540725708,
+      "mean_token_accuracy": 0.7473416924476624,
+      "num_tokens": 14040214.0,
+      "step": 376
+    },
+    {
+      "entropy": 0.6125365421175957,
+      "epoch": 2.925170068027211,
+      "grad_norm": 0.00803174264729023,
+      "learning_rate": 0.0002,
+      "loss": 0.6121209859848022,
+      "mean_token_accuracy": 0.7514154240489006,
+      "num_tokens": 14077708.0,
+      "step": 377
+    },
+    {
+      "entropy": 0.6050901785492897,
+      "epoch": 2.932944606413994,
+      "grad_norm": 0.007550297304987907,
+      "learning_rate": 0.0002,
+      "loss": 0.6031768321990967,
+      "mean_token_accuracy": 0.7547935023903847,
+      "num_tokens": 14115081.0,
+      "step": 378
+    },
+    {
+      "entropy": 0.5867772176861763,
+      "epoch": 2.9407191448007772,
+      "grad_norm": 0.007504226174205542,
+      "learning_rate": 0.0002,
+      "loss": 0.586522102355957,
+      "mean_token_accuracy": 0.7607723250985146,
+      "num_tokens": 14152707.0,
+      "step": 379
+    },
+    {
+      "entropy": 0.6053405031561852,
+      "epoch": 2.9484936831875608,
+      "grad_norm": 0.007515515200793743,
+      "learning_rate": 0.0002,
+      "loss": 0.6078545451164246,
+      "mean_token_accuracy": 0.7521200627088547,
+      "num_tokens": 14190102.0,
+      "step": 380
+    },
+    {
+      "entropy": 0.6189891770482063,
+      "epoch": 2.956268221574344,
+      "grad_norm": 0.009104753844439983,
+      "learning_rate": 0.0002,
+      "loss": 0.6249637603759766,
+      "mean_token_accuracy": 0.745013989508152,
+      "num_tokens": 14227592.0,
+      "step": 381
+    },
+    {
+      "entropy": 0.6138845458626747,
+      "epoch": 2.9640427599611274,
+      "grad_norm": 0.008113187737762928,
+      "learning_rate": 0.0002,
+      "loss": 0.6206336617469788,
+      "mean_token_accuracy": 0.7456697672605515,
+      "num_tokens": 14264742.0,
+      "step": 382
+    },
+    {
+      "entropy": 0.6130691468715668,
+      "epoch": 2.9718172983479105,
+      "grad_norm": 0.009299568831920624,
+      "learning_rate": 0.0002,
+      "loss": 0.6096789836883545,
+      "mean_token_accuracy": 0.7523899152874947,
+      "num_tokens": 14302195.0,
+      "step": 383
+    },
+    {
+      "entropy": 0.6205710843205452,
+      "epoch": 2.979591836734694,
+      "grad_norm": 0.009512864053249359,
+      "learning_rate": 0.0002,
+      "loss": 0.6156758069992065,
+      "mean_token_accuracy": 0.7508921846747398,
+      "num_tokens": 14339245.0,
+      "step": 384
+    },
+    {
+      "entropy": 0.6069811135530472,
+      "epoch": 2.987366375121477,
+      "grad_norm": 0.007910305634140968,
+      "learning_rate": 0.0002,
+      "loss": 0.6085284948348999,
+      "mean_token_accuracy": 0.753339558839798,
+      "num_tokens": 14376587.0,
+      "step": 385
+    },
+    {
+      "entropy": 0.6119454056024551,
+      "epoch": 2.9951409135082603,
+      "grad_norm": 0.009647355414927006,
+      "learning_rate": 0.0002,
+      "loss": 0.6212526559829712,
+      "mean_token_accuracy": 0.7483848258852959,
+      "num_tokens": 14414269.0,
+      "step": 386
+    },
+    {
+      "entropy": 0.5991271257400512,
+      "epoch": 3.0,
+      "grad_norm": 0.010158240795135498,
+      "learning_rate": 0.0002,
+      "loss": 0.6065676212310791,
+      "mean_token_accuracy": 0.7547077178955078,
+      "num_tokens": 14437732.0,
+      "step": 387
+    },
+    {
+      "entropy": 0.6155622974038124,
+      "epoch": 3.007774538386783,
+      "grad_norm": 0.007469667587429285,
+      "learning_rate": 0.0002,
+      "loss": 0.610929012298584,
+      "mean_token_accuracy": 0.7500728666782379,
+      "num_tokens": 14475215.0,
+      "step": 388
+    },
+    {
+      "entropy": 0.6059615164995193,
+      "epoch": 3.0155490767735667,
+      "grad_norm": 0.008237253874540329,
+      "learning_rate": 0.0002,
+      "loss": 0.6052074432373047,
+      "mean_token_accuracy": 0.75154148042202,
+      "num_tokens": 14512098.0,
+      "step": 389
+    },
+    {
+      "entropy": 0.6095065400004387,
+      "epoch": 3.0233236151603498,
+      "grad_norm": 0.008823649026453495,
+      "learning_rate": 0.0002,
+      "loss": 0.6052320599555969,
+      "mean_token_accuracy": 0.7540639191865921,
+      "num_tokens": 14549796.0,
+      "step": 390
+    },
+    {
+      "entropy": 0.5965104550123215,
+      "epoch": 3.0310981535471333,
+      "grad_norm": 0.008173096925020218,
+      "learning_rate": 0.0002,
+      "loss": 0.5943379998207092,
+      "mean_token_accuracy": 0.7575219720602036,
+      "num_tokens": 14587236.0,
+      "step": 391
+    },
+    {
+      "entropy": 0.6153795942664146,
+      "epoch": 3.0388726919339164,
+      "grad_norm": 0.011041256599128246,
+      "learning_rate": 0.0002,
+      "loss": 0.6242270469665527,
+      "mean_token_accuracy": 0.7458753809332848,
+      "num_tokens": 14624782.0,
+      "step": 392
+    },
+    {
+      "entropy": 0.5972734019160271,
+      "epoch": 3.0466472303206995,
+      "grad_norm": 0.008677808567881584,
+      "learning_rate": 0.0002,
+      "loss": 0.5998678207397461,
+      "mean_token_accuracy": 0.7576502561569214,
+      "num_tokens": 14662174.0,
+      "step": 393
+    },
+    {
+      "entropy": 0.6102932617068291,
+      "epoch": 3.054421768707483,
+      "grad_norm": 0.010088201612234116,
+      "learning_rate": 0.0002,
+      "loss": 0.6051709651947021,
+      "mean_token_accuracy": 0.7535885870456696,
+      "num_tokens": 14699746.0,
+      "step": 394
+    },
+    {
+      "entropy": 0.6008228436112404,
+      "epoch": 3.062196307094266,
+      "grad_norm": 0.008312424644827843,
+      "learning_rate": 0.0002,
+      "loss": 0.6032264232635498,
+      "mean_token_accuracy": 0.7517057359218597,
+      "num_tokens": 14737247.0,
+      "step": 395
+    },
+    {
+      "entropy": 0.5989790633320808,
+      "epoch": 3.0699708454810497,
+      "grad_norm": 0.008239499293267727,
+      "learning_rate": 0.0002,
+      "loss": 0.6043699979782104,
+      "mean_token_accuracy": 0.7529819533228874,
+      "num_tokens": 14774702.0,
+      "step": 396
+    },
+    {
+      "entropy": 0.6045716553926468,
+      "epoch": 3.077745383867833,
+      "grad_norm": 0.01003638468682766,
+      "learning_rate": 0.0002,
+      "loss": 0.6113682985305786,
+      "mean_token_accuracy": 0.7505319342017174,
+      "num_tokens": 14811886.0,
+      "step": 397
+    },
+    {
+      "entropy": 0.6115910932421684,
+      "epoch": 3.0855199222546164,
+      "grad_norm": 0.009529519826173782,
+      "learning_rate": 0.0002,
+      "loss": 0.6107526421546936,
+      "mean_token_accuracy": 0.7507694736123085,
+      "num_tokens": 14848988.0,
+      "step": 398
+    },
+    {
+      "entropy": 0.6084096133708954,
+      "epoch": 3.0932944606413995,
+      "grad_norm": 0.010617745108902454,
+      "learning_rate": 0.0002,
+      "loss": 0.6034584045410156,
+      "mean_token_accuracy": 0.753932997584343,
+      "num_tokens": 14886459.0,
+      "step": 399
+    },
+    {
+      "entropy": 0.6054683178663254,
+      "epoch": 3.1010689990281826,
+      "grad_norm": 0.008427645079791546,
+      "learning_rate": 0.0002,
+      "loss": 0.603972852230072,
+      "mean_token_accuracy": 0.7547962665557861,
+      "num_tokens": 14924071.0,
+      "step": 400
+    },
+    {
+      "entropy": 0.5898149386048317,
+      "epoch": 3.108843537414966,
+      "grad_norm": 0.009828660637140274,
+      "learning_rate": 0.0002,
+      "loss": 0.5968987941741943,
+      "mean_token_accuracy": 0.7556750178337097,
+      "num_tokens": 14961693.0,
+      "step": 401
+    },
+    {
+      "entropy": 0.5877138301730156,
+      "epoch": 3.116618075801749,
+      "grad_norm": 0.010062897577881813,
+      "learning_rate": 0.0002,
+      "loss": 0.5967484712600708,
+      "mean_token_accuracy": 0.7571568712592125,
+      "num_tokens": 14998846.0,
+      "step": 402
+    },
+    {
+      "entropy": 0.5948435962200165,
+      "epoch": 3.1243926141885328,
+      "grad_norm": 0.008929714560508728,
+      "learning_rate": 0.0002,
+      "loss": 0.596827507019043,
+      "mean_token_accuracy": 0.7577486112713814,
+      "num_tokens": 15036731.0,
+      "step": 403
+    },
+    {
+      "entropy": 0.6053501963615417,
+      "epoch": 3.132167152575316,
+      "grad_norm": 0.009441581554710865,
+      "learning_rate": 0.0002,
+      "loss": 0.6060553789138794,
+      "mean_token_accuracy": 0.7532980218529701,
+      "num_tokens": 15074609.0,
+      "step": 404
+    },
+    {
+      "entropy": 0.6139290183782578,
+      "epoch": 3.139941690962099,
+      "grad_norm": 0.009152664802968502,
+      "learning_rate": 0.0002,
+      "loss": 0.6124223470687866,
+      "mean_token_accuracy": 0.7500443160533905,
+      "num_tokens": 15112062.0,
+      "step": 405
+    },
+    {
+      "entropy": 0.5967981591820717,
+      "epoch": 3.1477162293488825,
+      "grad_norm": 0.009577469900250435,
+      "learning_rate": 0.0002,
+      "loss": 0.60134357213974,
+      "mean_token_accuracy": 0.75638198107481,
+      "num_tokens": 15149260.0,
+      "step": 406
+    },
+    {
+      "entropy": 0.6026358082890511,
+      "epoch": 3.1554907677356656,
+      "grad_norm": 0.008992240764200687,
+      "learning_rate": 0.0002,
+      "loss": 0.6071676015853882,
+      "mean_token_accuracy": 0.7526509240269661,
+      "num_tokens": 15186342.0,
+      "step": 407
+    },
+    {
+      "entropy": 0.6130121350288391,
+      "epoch": 3.163265306122449,
+      "grad_norm": 0.009787038899958134,
+      "learning_rate": 0.0002,
+      "loss": 0.616807222366333,
+      "mean_token_accuracy": 0.7485663071274757,
+      "num_tokens": 15224037.0,
+      "step": 408
+    },
+    {
+      "entropy": 0.6062626764178276,
+      "epoch": 3.1710398445092323,
+      "grad_norm": 0.009504587389528751,
+      "learning_rate": 0.0002,
+      "loss": 0.6050899624824524,
+      "mean_token_accuracy": 0.7536401823163033,
+      "num_tokens": 15261579.0,
+      "step": 409
+    },
+    {
+      "entropy": 0.6065879017114639,
+      "epoch": 3.1788143828960154,
+      "grad_norm": 0.008364197798073292,
+      "learning_rate": 0.0002,
+      "loss": 0.6022209525108337,
+      "mean_token_accuracy": 0.7534467577934265,
+      "num_tokens": 15299024.0,
+      "step": 410
+    },
+    {
+      "entropy": 0.6233559027314186,
+      "epoch": 3.186588921282799,
+      "grad_norm": 0.010281233116984367,
+      "learning_rate": 0.0002,
+      "loss": 0.6214974522590637,
+      "mean_token_accuracy": 0.7470389008522034,
+      "num_tokens": 15336558.0,
+      "step": 411
+    },
+    {
+      "entropy": 0.6087061986327171,
+      "epoch": 3.194363459669582,
+      "grad_norm": 0.00957780983299017,
+      "learning_rate": 0.0002,
+      "loss": 0.6129252910614014,
+      "mean_token_accuracy": 0.7495343014597893,
+      "num_tokens": 15374032.0,
+      "step": 412
+    },
+    {
+      "entropy": 0.596970833837986,
+      "epoch": 3.2021379980563656,
+      "grad_norm": 0.00984536949545145,
+      "learning_rate": 0.0002,
+      "loss": 0.5985467433929443,
+      "mean_token_accuracy": 0.7565372809767723,
+      "num_tokens": 15411483.0,
+      "step": 413
+    },
+    {
+      "entropy": 0.5959193110466003,
+      "epoch": 3.2099125364431487,
+      "grad_norm": 0.010170173831284046,
+      "learning_rate": 0.0002,
+      "loss": 0.599077045917511,
+      "mean_token_accuracy": 0.756831057369709,
+      "num_tokens": 15448812.0,
+      "step": 414
+    },
+    {
+      "entropy": 0.6122774854302406,
+      "epoch": 3.2176870748299318,
+      "grad_norm": 0.00988403894007206,
+      "learning_rate": 0.0002,
+      "loss": 0.6123433113098145,
+      "mean_token_accuracy": 0.7474271655082703,
+      "num_tokens": 15486225.0,
+      "step": 415
+    },
+    {
+      "entropy": 0.5967384874820709,
+      "epoch": 3.2254616132167153,
+      "grad_norm": 0.010967456735670567,
+      "learning_rate": 0.0002,
+      "loss": 0.60502028465271,
+      "mean_token_accuracy": 0.75384970754385,
+      "num_tokens": 15523563.0,
+      "step": 416
+    },
+    {
+      "entropy": 0.5971691235899925,
+      "epoch": 3.2332361516034984,
+      "grad_norm": 0.009652617387473583,
+      "learning_rate": 0.0002,
+      "loss": 0.596883237361908,
+      "mean_token_accuracy": 0.7563904896378517,
+      "num_tokens": 15561284.0,
+      "step": 417
+    },
+    {
+      "entropy": 0.5993461608886719,
+      "epoch": 3.241010689990282,
+      "grad_norm": 0.01028553955256939,
+      "learning_rate": 0.0002,
+      "loss": 0.5969662666320801,
+      "mean_token_accuracy": 0.7561450004577637,
+      "num_tokens": 15598518.0,
+      "step": 418
+    },
+    {
+      "entropy": 0.6052126884460449,
+      "epoch": 3.248785228377065,
+      "grad_norm": 0.009491597302258015,
+      "learning_rate": 0.0002,
+      "loss": 0.6018233299255371,
+      "mean_token_accuracy": 0.754707433283329,
+      "num_tokens": 15636077.0,
+      "step": 419
+    },
+    {
+      "entropy": 0.5944898948073387,
+      "epoch": 3.256559766763848,
+      "grad_norm": 0.009509161114692688,
+      "learning_rate": 0.0002,
+      "loss": 0.598441481590271,
+      "mean_token_accuracy": 0.7549791261553764,
+      "num_tokens": 15673148.0,
+      "step": 420
+    },
+    {
+      "entropy": 0.5968391299247742,
+      "epoch": 3.2643343051506317,
+      "grad_norm": 0.012232691049575806,
+      "learning_rate": 0.0002,
+      "loss": 0.6061251163482666,
+      "mean_token_accuracy": 0.7526469007134438,
+      "num_tokens": 15710778.0,
+      "step": 421
+    },
+    {
+      "entropy": 0.6134240105748177,
+      "epoch": 3.272108843537415,
+      "grad_norm": 0.00959104672074318,
+      "learning_rate": 0.0002,
+      "loss": 0.6182757616043091,
+      "mean_token_accuracy": 0.7455755323171616,
+      "num_tokens": 15748509.0,
+      "step": 422
+    },
+    {
+      "entropy": 0.6232677772641182,
+      "epoch": 3.2798833819241984,
+      "grad_norm": 0.008630241267383099,
+      "learning_rate": 0.0002,
+      "loss": 0.6197680234909058,
+      "mean_token_accuracy": 0.7470053881406784,
+      "num_tokens": 15786160.0,
+      "step": 423
+    },
+    {
+      "entropy": 0.613219253718853,
+      "epoch": 3.2876579203109815,
+      "grad_norm": 0.0128620695322752,
+      "learning_rate": 0.0002,
+      "loss": 0.607075572013855,
+      "mean_token_accuracy": 0.7511464059352875,
+      "num_tokens": 15823826.0,
+      "step": 424
+    },
+    {
+      "entropy": 0.6041038259863853,
+      "epoch": 3.295432458697765,
+      "grad_norm": 0.011236878111958504,
+      "learning_rate": 0.0002,
+      "loss": 0.6055868864059448,
+      "mean_token_accuracy": 0.7537136226892471,
+      "num_tokens": 15861102.0,
+      "step": 425
+    },
+    {
+      "entropy": 0.6037536263465881,
+      "epoch": 3.303206997084548,
+      "grad_norm": 0.01190877053886652,
+      "learning_rate": 0.0002,
+      "loss": 0.6142691373825073,
+      "mean_token_accuracy": 0.7493078485131264,
+      "num_tokens": 15898770.0,
+      "step": 426
+    },
+    {
+      "entropy": 0.5997900441288948,
+      "epoch": 3.3109815354713312,
+      "grad_norm": 0.01027680840343237,
+      "learning_rate": 0.0002,
+      "loss": 0.606103777885437,
+      "mean_token_accuracy": 0.7539091035723686,
+      "num_tokens": 15936500.0,
+      "step": 427
+    },
+    {
+      "entropy": 0.611216127872467,
+      "epoch": 3.3187560738581148,
+      "grad_norm": 0.011017821729183197,
+      "learning_rate": 0.0002,
+      "loss": 0.6174614429473877,
+      "mean_token_accuracy": 0.7461550533771515,
+      "num_tokens": 15974513.0,
+      "step": 428
+    },
+    {
+      "entropy": 0.6048758924007416,
+      "epoch": 3.326530612244898,
+      "grad_norm": 0.010105214081704617,
+      "learning_rate": 0.0002,
+      "loss": 0.6031618118286133,
+      "mean_token_accuracy": 0.7546434104442596,
+      "num_tokens": 16012014.0,
+      "step": 429
+    },
+    {
+      "entropy": 0.6172537207603455,
+      "epoch": 3.3343051506316814,
+      "grad_norm": 0.009731764905154705,
+      "learning_rate": 0.0002,
+      "loss": 0.6138004064559937,
+      "mean_token_accuracy": 0.7473105862736702,
+      "num_tokens": 16049595.0,
+      "step": 430
+    },
+    {
+      "entropy": 0.601245328783989,
+      "epoch": 3.3420796890184645,
+      "grad_norm": 0.010152102448046207,
+      "learning_rate": 0.0002,
+      "loss": 0.6028028726577759,
+      "mean_token_accuracy": 0.7536987066268921,
+      "num_tokens": 16087175.0,
+      "step": 431
+    },
+    {
+      "entropy": 0.6049877479672432,
+      "epoch": 3.3498542274052476,
+      "grad_norm": 0.008624416776001453,
+      "learning_rate": 0.0002,
+      "loss": 0.6070842742919922,
+      "mean_token_accuracy": 0.7543686851859093,
+      "num_tokens": 16124434.0,
+      "step": 432
+    },
+    {
+      "entropy": 0.6113510951399803,
+      "epoch": 3.357628765792031,
+      "grad_norm": 0.00979162473231554,
+      "learning_rate": 0.0002,
+      "loss": 0.614505410194397,
+      "mean_token_accuracy": 0.7483974099159241,
+      "num_tokens": 16161797.0,
+      "step": 433
+    },
+    {
+      "entropy": 0.6149813681840897,
+      "epoch": 3.3654033041788143,
+      "grad_norm": 0.009648281149566174,
+      "learning_rate": 0.0002,
+      "loss": 0.6138153076171875,
+      "mean_token_accuracy": 0.7505435049533844,
+      "num_tokens": 16199386.0,
+      "step": 434
+    },
+    {
+      "entropy": 0.6114403232932091,
+      "epoch": 3.373177842565598,
+      "grad_norm": 0.00991707295179367,
+      "learning_rate": 0.0002,
+      "loss": 0.6108669638633728,
+      "mean_token_accuracy": 0.7510098665952682,
+      "num_tokens": 16236951.0,
+      "step": 435
+    },
+    {
+      "entropy": 0.6059002429246902,
+      "epoch": 3.380952380952381,
+      "grad_norm": 0.009901657700538635,
+      "learning_rate": 0.0002,
+      "loss": 0.6064650416374207,
+      "mean_token_accuracy": 0.7527595311403275,
+      "num_tokens": 16274203.0,
+      "step": 436
+    },
+    {
+      "entropy": 0.6086044162511826,
+      "epoch": 3.388726919339164,
+      "grad_norm": 0.011115623638033867,
+      "learning_rate": 0.0002,
+      "loss": 0.609772801399231,
+      "mean_token_accuracy": 0.7524074539542198,
+      "num_tokens": 16311856.0,
+      "step": 437
+    },
+    {
+      "entropy": 0.6069784238934517,
+      "epoch": 3.3965014577259476,
+      "grad_norm": 0.009795066900551319,
+      "learning_rate": 0.0002,
+      "loss": 0.60982745885849,
+      "mean_token_accuracy": 0.7514877542853355,
+      "num_tokens": 16349625.0,
+      "step": 438
+    },
+    {
+      "entropy": 0.602266825735569,
+      "epoch": 3.4042759961127307,
+      "grad_norm": 0.009143724106252193,
+      "learning_rate": 0.0002,
+      "loss": 0.6032094359397888,
+      "mean_token_accuracy": 0.7547830194234848,
+      "num_tokens": 16387358.0,
+      "step": 439
+    },
+    {
+      "entropy": 0.5899055674672127,
+      "epoch": 3.4120505344995142,
+      "grad_norm": 0.009034923277795315,
+      "learning_rate": 0.0002,
+      "loss": 0.5931638479232788,
+      "mean_token_accuracy": 0.7577911838889122,
+      "num_tokens": 16424225.0,
+      "step": 440
+    },
+    {
+      "entropy": 0.5923598855733871,
+      "epoch": 3.4198250728862973,
+      "grad_norm": 0.00976620800793171,
+      "learning_rate": 0.0002,
+      "loss": 0.5963320732116699,
+      "mean_token_accuracy": 0.7556646093726158,
+      "num_tokens": 16461363.0,
+      "step": 441
+    },
+    {
+      "entropy": 0.6050904020667076,
+      "epoch": 3.427599611273081,
+      "grad_norm": 0.010530819185078144,
+      "learning_rate": 0.0002,
+      "loss": 0.6116576194763184,
+      "mean_token_accuracy": 0.7500759586691856,
+      "num_tokens": 16498597.0,
+      "step": 442
+    },
+    {
+      "entropy": 0.6166655048727989,
+      "epoch": 3.435374149659864,
+      "grad_norm": 0.009784851223230362,
+      "learning_rate": 0.0002,
+      "loss": 0.6191204786300659,
+      "mean_token_accuracy": 0.7476971745491028,
+      "num_tokens": 16535886.0,
+      "step": 443
+    },
+    {
+      "entropy": 0.6059679165482521,
+      "epoch": 3.443148688046647,
+      "grad_norm": 0.009777678176760674,
+      "learning_rate": 0.0002,
+      "loss": 0.600814700126648,
+      "mean_token_accuracy": 0.7545428648591042,
+      "num_tokens": 16573204.0,
+      "step": 444
+    },
+    {
+      "entropy": 0.6082785129547119,
+      "epoch": 3.4509232264334306,
+      "grad_norm": 0.009606909938156605,
+      "learning_rate": 0.0002,
+      "loss": 0.6063122153282166,
+      "mean_token_accuracy": 0.7535424381494522,
+      "num_tokens": 16610341.0,
+      "step": 445
+    },
+    {
+      "entropy": 0.6118370518088341,
+      "epoch": 3.4586977648202137,
+      "grad_norm": 0.009755875915288925,
+      "learning_rate": 0.0002,
+      "loss": 0.6115654706954956,
+      "mean_token_accuracy": 0.7514548450708389,
+      "num_tokens": 16648078.0,
+      "step": 446
+    },
+    {
+      "entropy": 0.5936368778347969,
+      "epoch": 3.466472303206997,
+      "grad_norm": 0.011094003915786743,
+      "learning_rate": 0.0002,
+      "loss": 0.6031495332717896,
+      "mean_token_accuracy": 0.75296650826931,
+      "num_tokens": 16684979.0,
+      "step": 447
+    },
+    {
+      "entropy": 0.5946005135774612,
+      "epoch": 3.4742468415937804,
+      "grad_norm": 0.00817946158349514,
+      "learning_rate": 0.0002,
+      "loss": 0.597072184085846,
+      "mean_token_accuracy": 0.756522886455059,
+      "num_tokens": 16722537.0,
+      "step": 448
+    },
+    {
+      "entropy": 0.6108994483947754,
+      "epoch": 3.4820213799805635,
+      "grad_norm": 0.0094683188945055,
+      "learning_rate": 0.0002,
+      "loss": 0.6126965284347534,
+      "mean_token_accuracy": 0.750556543469429,
+      "num_tokens": 16759490.0,
+      "step": 449
+    },
+    {
+      "entropy": 0.6224265471100807,
+      "epoch": 3.489795918367347,
+      "grad_norm": 0.009895196184515953,
+      "learning_rate": 0.0002,
+      "loss": 0.6202942132949829,
+      "mean_token_accuracy": 0.7480411902070045,
+      "num_tokens": 16797135.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.597562313079834,
+      "epoch": 3.49757045675413,
+      "grad_norm": 0.00880007166415453,
+      "learning_rate": 0.0002,
+      "loss": 0.5973268151283264,
+      "mean_token_accuracy": 0.7553020343184471,
+      "num_tokens": 16834270.0,
+      "step": 451
+    },
+    {
+      "entropy": 0.5932627841830254,
+      "epoch": 3.5053449951409137,
+      "grad_norm": 0.009338297881186008,
+      "learning_rate": 0.0002,
+      "loss": 0.5983443856239319,
+      "mean_token_accuracy": 0.7540717199444771,
+      "num_tokens": 16871670.0,
+      "step": 452
+    },
+    {
+      "entropy": 0.603311225771904,
+      "epoch": 3.513119533527697,
+      "grad_norm": 0.009173530153930187,
+      "learning_rate": 0.0002,
+      "loss": 0.6072041988372803,
+      "mean_token_accuracy": 0.752394326031208,
+      "num_tokens": 16909221.0,
+      "step": 453
+    },
+    {
+      "entropy": 0.6008370220661163,
+      "epoch": 3.52089407191448,
+      "grad_norm": 0.009413264691829681,
+      "learning_rate": 0.0002,
+      "loss": 0.6025704145431519,
+      "mean_token_accuracy": 0.7541562095284462,
+      "num_tokens": 16946802.0,
+      "step": 454
+    },
+    {
+      "entropy": 0.6156598478555679,
+      "epoch": 3.5286686103012634,
+      "grad_norm": 0.009066940285265446,
+      "learning_rate": 0.0002,
+      "loss": 0.6163049936294556,
+      "mean_token_accuracy": 0.7471535354852676,
+      "num_tokens": 16983926.0,
+      "step": 455
+    },
+    {
+      "entropy": 0.6066097542643547,
+      "epoch": 3.5364431486880465,
+      "grad_norm": 0.008060373365879059,
+      "learning_rate": 0.0002,
+      "loss": 0.6057984828948975,
+      "mean_token_accuracy": 0.753935232758522,
+      "num_tokens": 17021676.0,
+      "step": 456
+    },
+    {
+      "entropy": 0.5982315316796303,
+      "epoch": 3.54421768707483,
+      "grad_norm": 0.008438881486654282,
+      "learning_rate": 0.0002,
+      "loss": 0.6000136137008667,
+      "mean_token_accuracy": 0.7557498961687088,
+      "num_tokens": 17059698.0,
+      "step": 457
+    },
+    {
+      "entropy": 0.6007475033402443,
+      "epoch": 3.551992225461613,
+      "grad_norm": 0.008833416737616062,
+      "learning_rate": 0.0002,
+      "loss": 0.6044740676879883,
+      "mean_token_accuracy": 0.7540414854884148,
+      "num_tokens": 17096952.0,
+      "step": 458
+    },
+    {
+      "entropy": 0.6120948642492294,
+      "epoch": 3.5597667638483967,
+      "grad_norm": 0.010080665349960327,
+      "learning_rate": 0.0002,
+      "loss": 0.6188148260116577,
+      "mean_token_accuracy": 0.7483313381671906,
+      "num_tokens": 17134663.0,
+      "step": 459
+    },
+    {
+      "entropy": 0.6083438396453857,
+      "epoch": 3.56754130223518,
+      "grad_norm": 0.009289734996855259,
+      "learning_rate": 0.0002,
+      "loss": 0.6083492636680603,
+      "mean_token_accuracy": 0.7500124499201775,
+      "num_tokens": 17172169.0,
+      "step": 460
+    },
+    {
+      "entropy": 0.6143280863761902,
+      "epoch": 3.575315840621963,
+      "grad_norm": 0.010136788710951805,
+      "learning_rate": 0.0002,
+      "loss": 0.6146470308303833,
+      "mean_token_accuracy": 0.749954454600811,
+      "num_tokens": 17209481.0,
+      "step": 461
+    },
+    {
+      "entropy": 0.6075241789221764,
+      "epoch": 3.5830903790087465,
+      "grad_norm": 0.008481534197926521,
+      "learning_rate": 0.0002,
+      "loss": 0.6094347238540649,
+      "mean_token_accuracy": 0.7524155974388123,
+      "num_tokens": 17246466.0,
+      "step": 462
+    },
+    {
+      "entropy": 0.6080920621752739,
+      "epoch": 3.5908649173955296,
+      "grad_norm": 0.009619004093110561,
+      "learning_rate": 0.0002,
+      "loss": 0.6094264984130859,
+      "mean_token_accuracy": 0.752539336681366,
+      "num_tokens": 17284014.0,
+      "step": 463
+    },
+    {
+      "entropy": 0.5951874181628227,
+      "epoch": 3.5986394557823127,
+      "grad_norm": 0.010971908457577229,
+      "learning_rate": 0.0002,
+      "loss": 0.5968849062919617,
+      "mean_token_accuracy": 0.7571029588580132,
+      "num_tokens": 17321447.0,
+      "step": 464
+    },
+    {
+      "entropy": 0.600244015455246,
+      "epoch": 3.6064139941690962,
+      "grad_norm": 0.009995541535317898,
+      "learning_rate": 0.0002,
+      "loss": 0.6033259630203247,
+      "mean_token_accuracy": 0.7534294128417969,
+      "num_tokens": 17358802.0,
+      "step": 465
+    },
+    {
+      "entropy": 0.6006019562482834,
+      "epoch": 3.6141885325558794,
+      "grad_norm": 0.008752602152526379,
+      "learning_rate": 0.0002,
+      "loss": 0.6017124652862549,
+      "mean_token_accuracy": 0.756244920194149,
+      "num_tokens": 17396188.0,
+      "step": 466
+    },
+    {
+      "entropy": 0.6169783622026443,
+      "epoch": 3.621963070942663,
+      "grad_norm": 0.010406076908111572,
+      "learning_rate": 0.0002,
+      "loss": 0.6129188537597656,
+      "mean_token_accuracy": 0.7500476241111755,
+      "num_tokens": 17433577.0,
+      "step": 467
+    },
+    {
+      "entropy": 0.6030160784721375,
+      "epoch": 3.629737609329446,
+      "grad_norm": 0.009500440210103989,
+      "learning_rate": 0.0002,
+      "loss": 0.6009221076965332,
+      "mean_token_accuracy": 0.7557852789759636,
+      "num_tokens": 17471102.0,
+      "step": 468
+    },
+    {
+      "entropy": 0.5835254713892937,
+      "epoch": 3.6375121477162295,
+      "grad_norm": 0.009339805692434311,
+      "learning_rate": 0.0002,
+      "loss": 0.5876451730728149,
+      "mean_token_accuracy": 0.7601824253797531,
+      "num_tokens": 17508438.0,
+      "step": 469
+    },
+    {
+      "entropy": 0.6057174652814865,
+      "epoch": 3.6452866861030127,
+      "grad_norm": 0.010681871324777603,
+      "learning_rate": 0.0002,
+      "loss": 0.6165672540664673,
+      "mean_token_accuracy": 0.7476890683174133,
+      "num_tokens": 17546176.0,
+      "step": 470
+    },
+    {
+      "entropy": 0.6023501753807068,
+      "epoch": 3.6530612244897958,
+      "grad_norm": 0.009493213146924973,
+      "learning_rate": 0.0002,
+      "loss": 0.6091412305831909,
+      "mean_token_accuracy": 0.7539057731628418,
+      "num_tokens": 17583501.0,
+      "step": 471
+    },
+    {
+      "entropy": 0.6135850101709366,
+      "epoch": 3.6608357628765793,
+      "grad_norm": 0.009873135015368462,
+      "learning_rate": 0.0002,
+      "loss": 0.6167327165603638,
+      "mean_token_accuracy": 0.7497850060462952,
+      "num_tokens": 17620694.0,
+      "step": 472
+    },
+    {
+      "entropy": 0.6023069024085999,
+      "epoch": 3.6686103012633624,
+      "grad_norm": 0.009261439554393291,
+      "learning_rate": 0.0002,
+      "loss": 0.5992604494094849,
+      "mean_token_accuracy": 0.7568845972418785,
+      "num_tokens": 17657846.0,
+      "step": 473
+    },
+    {
+      "entropy": 0.6220817118883133,
+      "epoch": 3.6763848396501455,
+      "grad_norm": 0.008002927526831627,
+      "learning_rate": 0.0002,
+      "loss": 0.6206220984458923,
+      "mean_token_accuracy": 0.7471845746040344,
+      "num_tokens": 17695493.0,
+      "step": 474
+    },
+    {
+      "entropy": 0.6081802695989609,
+      "epoch": 3.684159378036929,
+      "grad_norm": 0.008587166666984558,
+      "learning_rate": 0.0002,
+      "loss": 0.607234001159668,
+      "mean_token_accuracy": 0.755810484290123,
+      "num_tokens": 17732856.0,
+      "step": 475
+    },
+    {
+      "entropy": 0.6025589928030968,
+      "epoch": 3.6919339164237126,
+      "grad_norm": 0.009315953589975834,
+      "learning_rate": 0.0002,
+      "loss": 0.6054584383964539,
+      "mean_token_accuracy": 0.7516716048121452,
+      "num_tokens": 17769974.0,
+      "step": 476
+    },
+    {
+      "entropy": 0.6041041016578674,
+      "epoch": 3.6997084548104957,
+      "grad_norm": 0.009252023883163929,
+      "learning_rate": 0.0002,
+      "loss": 0.6035861968994141,
+      "mean_token_accuracy": 0.754462331533432,
+      "num_tokens": 17807234.0,
+      "step": 477
+    },
+    {
+      "entropy": 0.6150568872690201,
+      "epoch": 3.707482993197279,
+      "grad_norm": 0.009806321002542973,
+      "learning_rate": 0.0002,
+      "loss": 0.6143267154693604,
+      "mean_token_accuracy": 0.7500679716467857,
+      "num_tokens": 17844713.0,
+      "step": 478
+    },
+    {
+      "entropy": 0.6017017439007759,
+      "epoch": 3.7152575315840624,
+      "grad_norm": 0.009355618618428707,
+      "learning_rate": 0.0002,
+      "loss": 0.6050205230712891,
+      "mean_token_accuracy": 0.7535286843776703,
+      "num_tokens": 17882167.0,
+      "step": 479
+    },
+    {
+      "entropy": 0.6020700857043266,
+      "epoch": 3.7230320699708455,
+      "grad_norm": 0.010259266011416912,
+      "learning_rate": 0.0002,
+      "loss": 0.6069679856300354,
+      "mean_token_accuracy": 0.7521628364920616,
+      "num_tokens": 17919373.0,
+      "step": 480
+    },
+    {
+      "entropy": 0.6041392460465431,
+      "epoch": 3.7308066083576286,
+      "grad_norm": 0.008957406505942345,
+      "learning_rate": 0.0002,
+      "loss": 0.6065245270729065,
+      "mean_token_accuracy": 0.7511508986353874,
+      "num_tokens": 17956544.0,
+      "step": 481
+    },
+    {
+      "entropy": 0.6143322214484215,
+      "epoch": 3.738581146744412,
+      "grad_norm": 0.010185849852859974,
+      "learning_rate": 0.0002,
+      "loss": 0.6167340278625488,
+      "mean_token_accuracy": 0.7486473768949509,
+      "num_tokens": 17993752.0,
+      "step": 482
+    },
+    {
+      "entropy": 0.597200371325016,
+      "epoch": 3.746355685131195,
+      "grad_norm": 0.009210821241140366,
+      "learning_rate": 0.0002,
+      "loss": 0.5978806018829346,
+      "mean_token_accuracy": 0.7557302191853523,
+      "num_tokens": 18031025.0,
+      "step": 483
+    },
+    {
+      "entropy": 0.6069202646613121,
+      "epoch": 3.7541302235179788,
+      "grad_norm": 0.011117368005216122,
+      "learning_rate": 0.0002,
+      "loss": 0.6108845472335815,
+      "mean_token_accuracy": 0.7496053129434586,
+      "num_tokens": 18067927.0,
+      "step": 484
+    },
+    {
+      "entropy": 0.6041139885783195,
+      "epoch": 3.761904761904762,
+      "grad_norm": 0.00938522256910801,
+      "learning_rate": 0.0002,
+      "loss": 0.6076387166976929,
+      "mean_token_accuracy": 0.7511148899793625,
+      "num_tokens": 18105521.0,
+      "step": 485
+    },
+    {
+      "entropy": 0.5986451655626297,
+      "epoch": 3.7696793002915454,
+      "grad_norm": 0.009973375126719475,
+      "learning_rate": 0.0002,
+      "loss": 0.6053519248962402,
+      "mean_token_accuracy": 0.7534126862883568,
+      "num_tokens": 18142635.0,
+      "step": 486
+    },
+    {
+      "entropy": 0.6009933426976204,
+      "epoch": 3.7774538386783285,
+      "grad_norm": 0.008744049817323685,
+      "learning_rate": 0.0002,
+      "loss": 0.6009329557418823,
+      "mean_token_accuracy": 0.7565900459885597,
+      "num_tokens": 18179737.0,
+      "step": 487
+    },
+    {
+      "entropy": 0.6145603656768799,
+      "epoch": 3.7852283770651116,
+      "grad_norm": 0.008431525900959969,
+      "learning_rate": 0.0002,
+      "loss": 0.6102414727210999,
+      "mean_token_accuracy": 0.7497778162360191,
+      "num_tokens": 18216997.0,
+      "step": 488
+    },
+    {
+      "entropy": 0.6055081114172935,
+      "epoch": 3.793002915451895,
+      "grad_norm": 0.00999277364462614,
+      "learning_rate": 0.0002,
+      "loss": 0.6034255623817444,
+      "mean_token_accuracy": 0.7561624273657799,
+      "num_tokens": 18254319.0,
+      "step": 489
+    },
+    {
+      "entropy": 0.5926950201392174,
+      "epoch": 3.8007774538386783,
+      "grad_norm": 0.008595049381256104,
+      "learning_rate": 0.0002,
+      "loss": 0.5922105312347412,
+      "mean_token_accuracy": 0.7585625573992729,
+      "num_tokens": 18291224.0,
+      "step": 490
+    },
+    {
+      "entropy": 0.6234352439641953,
+      "epoch": 3.8085519922254614,
+      "grad_norm": 0.009778706356883049,
+      "learning_rate": 0.0002,
+      "loss": 0.628953754901886,
+      "mean_token_accuracy": 0.743121363222599,
+      "num_tokens": 18328744.0,
+      "step": 491
+    },
+    {
+      "entropy": 0.6095947250723839,
+      "epoch": 3.816326530612245,
+      "grad_norm": 0.009355447255074978,
+      "learning_rate": 0.0002,
+      "loss": 0.6148620843887329,
+      "mean_token_accuracy": 0.7512015551328659,
+      "num_tokens": 18366284.0,
+      "step": 492
+    },
+    {
+      "entropy": 0.6000761389732361,
+      "epoch": 3.824101068999028,
+      "grad_norm": 0.00857703946530819,
+      "learning_rate": 0.0002,
+      "loss": 0.604392409324646,
+      "mean_token_accuracy": 0.7549523860216141,
+      "num_tokens": 18403721.0,
+      "step": 493
+    },
+    {
+      "entropy": 0.5995704010128975,
+      "epoch": 3.8318756073858116,
+      "grad_norm": 0.010483372025191784,
+      "learning_rate": 0.0002,
+      "loss": 0.6008812785148621,
+      "mean_token_accuracy": 0.7550973892211914,
+      "num_tokens": 18440931.0,
+      "step": 494
+    },
+    {
+      "entropy": 0.6065568849444389,
+      "epoch": 3.8396501457725947,
+      "grad_norm": 0.00975153036415577,
+      "learning_rate": 0.0002,
+      "loss": 0.6107536554336548,
+      "mean_token_accuracy": 0.7495853081345558,
+      "num_tokens": 18478285.0,
+      "step": 495
+    },
+    {
+      "entropy": 0.6092389151453972,
+      "epoch": 3.847424684159378,
+      "grad_norm": 0.008589878678321838,
+      "learning_rate": 0.0002,
+      "loss": 0.6096630096435547,
+      "mean_token_accuracy": 0.7536827027797699,
+      "num_tokens": 18515709.0,
+      "step": 496
+    },
+    {
+      "entropy": 0.5968381017446518,
+      "epoch": 3.8551992225461613,
+      "grad_norm": 0.011193851009011269,
+      "learning_rate": 0.0002,
+      "loss": 0.6000962257385254,
+      "mean_token_accuracy": 0.7573477849364281,
+      "num_tokens": 18553032.0,
+      "step": 497
+    },
+    {
+      "entropy": 0.6045369654893875,
+      "epoch": 3.8629737609329444,
+      "grad_norm": 0.009280527010560036,
+      "learning_rate": 0.0002,
+      "loss": 0.6028830409049988,
+      "mean_token_accuracy": 0.7542475759983063,
+      "num_tokens": 18590308.0,
+      "step": 498
+    },
+    {
+      "entropy": 0.6031981706619263,
+      "epoch": 3.870748299319728,
+      "grad_norm": 0.009086862206459045,
+      "learning_rate": 0.0002,
+      "loss": 0.6040035486221313,
+      "mean_token_accuracy": 0.7532067000865936,
+      "num_tokens": 18627503.0,
+      "step": 499
+    },
+    {
+      "entropy": 0.6000340953469276,
+      "epoch": 3.878522837706511,
+      "grad_norm": 0.008906064555048943,
+      "learning_rate": 0.0002,
+      "loss": 0.6016078591346741,
+      "mean_token_accuracy": 0.7553933039307594,
+      "num_tokens": 18664999.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.613756813108921,
+      "epoch": 3.8862973760932946,
+      "grad_norm": 0.00961924996227026,
+      "learning_rate": 0.0002,
+      "loss": 0.6131439208984375,
+      "mean_token_accuracy": 0.7474770769476891,
+      "num_tokens": 18702736.0,
+      "step": 501
+    },
+    {
+      "entropy": 0.6073272302746773,
+      "epoch": 3.8940719144800777,
+      "grad_norm": 0.008611987344920635,
+      "learning_rate": 0.0002,
+      "loss": 0.6086411476135254,
+      "mean_token_accuracy": 0.7537659332156181,
+      "num_tokens": 18740002.0,
+      "step": 502
+    },
+    {
+      "entropy": 0.5979632884263992,
+      "epoch": 3.9018464528668613,
+      "grad_norm": 0.008601277135312557,
+      "learning_rate": 0.0002,
+      "loss": 0.5986984968185425,
+      "mean_token_accuracy": 0.7558256238698959,
+      "num_tokens": 18777409.0,
+      "step": 503
+    },
+    {
+      "entropy": 0.5990572273731232,
+      "epoch": 3.9096209912536444,
+      "grad_norm": 0.009787015616893768,
+      "learning_rate": 0.0002,
+      "loss": 0.5969674587249756,
+      "mean_token_accuracy": 0.7572850733995438,
+      "num_tokens": 18814832.0,
+      "step": 504
+    },
+    {
+      "entropy": 0.5952424928545952,
+      "epoch": 3.9173955296404275,
+      "grad_norm": 0.009869958274066448,
+      "learning_rate": 0.0002,
+      "loss": 0.600977897644043,
+      "mean_token_accuracy": 0.7555051743984222,
+      "num_tokens": 18852288.0,
+      "step": 505
+    },
+    {
+      "entropy": 0.5953851416707039,
+      "epoch": 3.925170068027211,
+      "grad_norm": 0.009635512717068195,
+      "learning_rate": 0.0002,
+      "loss": 0.5988470315933228,
+      "mean_token_accuracy": 0.7552962228655815,
+      "num_tokens": 18889544.0,
+      "step": 506
+    },
+    {
+      "entropy": 0.6188566163182259,
+      "epoch": 3.932944606413994,
+      "grad_norm": 0.00981516670435667,
+      "learning_rate": 0.0002,
+      "loss": 0.6207889914512634,
+      "mean_token_accuracy": 0.748513400554657,
+      "num_tokens": 18927252.0,
+      "step": 507
+    },
+    {
+      "entropy": 0.6106720492243767,
+      "epoch": 3.9407191448007772,
+      "grad_norm": 0.011930068954825401,
+      "learning_rate": 0.0002,
+      "loss": 0.6130895614624023,
+      "mean_token_accuracy": 0.7500510811805725,
+      "num_tokens": 18964681.0,
+      "step": 508
+    },
+    {
+      "entropy": 0.6134953126311302,
+      "epoch": 3.9484936831875608,
+      "grad_norm": 0.009485729970037937,
+      "learning_rate": 0.0002,
+      "loss": 0.6170802116394043,
+      "mean_token_accuracy": 0.7490968182682991,
+      "num_tokens": 19002066.0,
+      "step": 509
+    },
+    {
+      "entropy": 0.6209791451692581,
+      "epoch": 3.956268221574344,
+      "grad_norm": 0.009478907100856304,
+      "learning_rate": 0.0002,
+      "loss": 0.6176908612251282,
+      "mean_token_accuracy": 0.7488641366362572,
+      "num_tokens": 19039553.0,
+      "step": 510
+    },
+    {
+      "entropy": 0.6174532473087311,
+      "epoch": 3.9640427599611274,
+      "grad_norm": 0.010146988555788994,
+      "learning_rate": 0.0002,
+      "loss": 0.6160500049591064,
+      "mean_token_accuracy": 0.7503278478980064,
+      "num_tokens": 19076777.0,
+      "step": 511
+    },
+    {
+      "entropy": 0.6139644980430603,
+      "epoch": 3.9718172983479105,
+      "grad_norm": 0.009649143554270267,
+      "learning_rate": 0.0002,
+      "loss": 0.6151527166366577,
+      "mean_token_accuracy": 0.7500214949250221,
+      "num_tokens": 19114295.0,
+      "step": 512
+    },
+    {
+      "entropy": 0.5903387889266014,
+      "epoch": 3.979591836734694,
+      "grad_norm": 0.009517780505120754,
+      "learning_rate": 0.0002,
+      "loss": 0.5949221253395081,
+      "mean_token_accuracy": 0.7585930302739143,
+      "num_tokens": 19151608.0,
+      "step": 513
+    },
+    {
+      "entropy": 0.60362558811903,
+      "epoch": 3.987366375121477,
+      "grad_norm": 0.008633180521428585,
+      "learning_rate": 0.0002,
+      "loss": 0.606721818447113,
+      "mean_token_accuracy": 0.7524262145161629,
+      "num_tokens": 19189092.0,
+      "step": 514
+    },
+    {
+      "entropy": 0.6161174476146698,
+      "epoch": 3.9951409135082603,
+      "grad_norm": 0.00904986634850502,
+      "learning_rate": 0.0002,
+      "loss": 0.6170212030410767,
+      "mean_token_accuracy": 0.7481908574700356,
+      "num_tokens": 19226930.0,
+      "step": 515
+    },
+    {
+      "entropy": 0.6157908320426941,
+      "epoch": 4.0,
+      "grad_norm": 0.011043760925531387,
+      "learning_rate": 0.0002,
+      "loss": 0.6197408437728882,
+      "mean_token_accuracy": 0.7473827481269837,
+      "num_tokens": 19250261.0,
+      "step": 516
+    },
+    {
+      "entropy": 0.5984989553689957,
+      "epoch": 4.0077745383867835,
+      "grad_norm": 0.010571295395493507,
+      "learning_rate": 0.0002,
+      "loss": 0.5962023138999939,
+      "mean_token_accuracy": 0.7551586180925369,
+      "num_tokens": 19287781.0,
+      "step": 517
+    },
+    {
+      "entropy": 0.5755362883210182,
+      "epoch": 4.015549076773566,
+      "grad_norm": 0.01085303071886301,
+      "learning_rate": 0.0002,
+      "loss": 0.5781339406967163,
+      "mean_token_accuracy": 0.763518862426281,
+      "num_tokens": 19324763.0,
+      "step": 518
+    },
+    {
+      "entropy": 0.5975531563162804,
+      "epoch": 4.02332361516035,
+      "grad_norm": 0.009798814542591572,
+      "learning_rate": 0.0002,
+      "loss": 0.6017999649047852,
+      "mean_token_accuracy": 0.7532066255807877,
+      "num_tokens": 19362451.0,
+      "step": 519
+    },
+    {
+      "entropy": 0.5838107392191887,
+      "epoch": 4.031098153547133,
+      "grad_norm": 0.011370974592864513,
+      "learning_rate": 0.0002,
+      "loss": 0.5857977867126465,
+      "mean_token_accuracy": 0.7589215487241745,
+      "num_tokens": 19399597.0,
+      "step": 520
+    },
+    {
+      "entropy": 0.586229719221592,
+      "epoch": 4.038872691933917,
+      "grad_norm": 0.010519232600927353,
+      "learning_rate": 0.0002,
+      "loss": 0.5870205760002136,
+      "mean_token_accuracy": 0.7586957290768623,
+      "num_tokens": 19436947.0,
+      "step": 521
+    },
+    {
+      "entropy": 0.5909631699323654,
+      "epoch": 4.0466472303206995,
+      "grad_norm": 0.011675548739731312,
+      "learning_rate": 0.0002,
+      "loss": 0.5927350521087646,
+      "mean_token_accuracy": 0.7586767747998238,
+      "num_tokens": 19474464.0,
+      "step": 522
+    },
+    {
+      "entropy": 0.602616548538208,
+      "epoch": 4.054421768707483,
+      "grad_norm": 0.01183371152728796,
+      "learning_rate": 0.0002,
+      "loss": 0.6010576486587524,
+      "mean_token_accuracy": 0.753461204469204,
+      "num_tokens": 19512364.0,
+      "step": 523
+    },
+    {
+      "entropy": 0.5877483561635017,
+      "epoch": 4.062196307094267,
+      "grad_norm": 0.011604937724769115,
+      "learning_rate": 0.0002,
+      "loss": 0.5869612693786621,
+      "mean_token_accuracy": 0.7605117782950401,
+      "num_tokens": 19549629.0,
+      "step": 524
+    },
+    {
+      "entropy": 0.5913394093513489,
+      "epoch": 4.069970845481049,
+      "grad_norm": 0.010066917166113853,
+      "learning_rate": 0.0002,
+      "loss": 0.5947997570037842,
+      "mean_token_accuracy": 0.755763366818428,
+      "num_tokens": 19586996.0,
+      "step": 525
+    },
+    {
+      "entropy": 0.5896845906972885,
+      "epoch": 4.077745383867833,
+      "grad_norm": 0.013865236192941666,
+      "learning_rate": 0.0002,
+      "loss": 0.5975737571716309,
+      "mean_token_accuracy": 0.756596714258194,
+      "num_tokens": 19624454.0,
+      "step": 526
+    },
+    {
+      "entropy": 0.6071945577859879,
+      "epoch": 4.085519922254616,
+      "grad_norm": 0.01125173456966877,
+      "learning_rate": 0.0002,
+      "loss": 0.6075310707092285,
+      "mean_token_accuracy": 0.7516418322920799,
+      "num_tokens": 19662369.0,
+      "step": 527
+    },
+    {
+      "entropy": 0.6004839465022087,
+      "epoch": 4.093294460641399,
+      "grad_norm": 0.012244274839758873,
+      "learning_rate": 0.0002,
+      "loss": 0.5988795161247253,
+      "mean_token_accuracy": 0.7533307895064354,
+      "num_tokens": 19699833.0,
+      "step": 528
+    },
+    {
+      "entropy": 0.5805928781628609,
+      "epoch": 4.101068999028183,
+      "grad_norm": 0.012553567998111248,
+      "learning_rate": 0.0002,
+      "loss": 0.5836637616157532,
+      "mean_token_accuracy": 0.7613400369882584,
+      "num_tokens": 19736626.0,
+      "step": 529
+    },
+    {
+      "entropy": 0.5802028924226761,
+      "epoch": 4.108843537414966,
+      "grad_norm": 0.013091178610920906,
+      "learning_rate": 0.0002,
+      "loss": 0.5839430093765259,
+      "mean_token_accuracy": 0.7619325369596481,
+      "num_tokens": 19773772.0,
+      "step": 530
+    },
+    {
+      "entropy": 0.5931153893470764,
+      "epoch": 4.11661807580175,
+      "grad_norm": 0.0117764538154006,
+      "learning_rate": 0.0002,
+      "loss": 0.598299503326416,
+      "mean_token_accuracy": 0.755822129547596,
+      "num_tokens": 19810974.0,
+      "step": 531
+    },
+    {
+      "entropy": 0.6081870496273041,
+      "epoch": 4.124392614188532,
+      "grad_norm": 0.013240421190857887,
+      "learning_rate": 0.0002,
+      "loss": 0.6058256030082703,
+      "mean_token_accuracy": 0.7531123831868172,
+      "num_tokens": 19848433.0,
+      "step": 532
+    },
+    {
+      "entropy": 0.6002762466669083,
+      "epoch": 4.132167152575316,
+      "grad_norm": 0.010874156840145588,
+      "learning_rate": 0.0002,
+      "loss": 0.5934455990791321,
+      "mean_token_accuracy": 0.7567582130432129,
+      "num_tokens": 19885930.0,
+      "step": 533
+    },
+    {
+      "entropy": 0.604529857635498,
+      "epoch": 4.139941690962099,
+      "grad_norm": 0.011289739049971104,
+      "learning_rate": 0.0002,
+      "loss": 0.6058647632598877,
+      "mean_token_accuracy": 0.7521885484457016,
+      "num_tokens": 19923130.0,
+      "step": 534
+    },
+    {
+      "entropy": 0.5959961265325546,
+      "epoch": 4.147716229348882,
+      "grad_norm": 0.0105770044028759,
+      "learning_rate": 0.0002,
+      "loss": 0.5998325347900391,
+      "mean_token_accuracy": 0.7545555010437965,
+      "num_tokens": 19960651.0,
+      "step": 535
+    },
+    {
+      "entropy": 0.6115074902772903,
+      "epoch": 4.155490767735666,
+      "grad_norm": 0.01272744033485651,
+      "learning_rate": 0.0002,
+      "loss": 0.6133629083633423,
+      "mean_token_accuracy": 0.7504753991961479,
+      "num_tokens": 19998458.0,
+      "step": 536
+    },
+    {
+      "entropy": 0.5959336087107658,
+      "epoch": 4.163265306122449,
+      "grad_norm": 0.01114166434854269,
+      "learning_rate": 0.0002,
+      "loss": 0.5921066403388977,
+      "mean_token_accuracy": 0.7599503472447395,
+      "num_tokens": 20035742.0,
+      "step": 537
+    },
+    {
+      "entropy": 0.6055325567722321,
+      "epoch": 4.171039844509233,
+      "grad_norm": 0.01163046807050705,
+      "learning_rate": 0.0002,
+      "loss": 0.6048938035964966,
+      "mean_token_accuracy": 0.7535717412829399,
+      "num_tokens": 20073263.0,
+      "step": 538
+    },
+    {
+      "entropy": 0.5937555730342865,
+      "epoch": 4.178814382896015,
+      "grad_norm": 0.010925386101007462,
+      "learning_rate": 0.0002,
+      "loss": 0.5978741645812988,
+      "mean_token_accuracy": 0.7580423355102539,
+      "num_tokens": 20110930.0,
+      "step": 539
+    },
+    {
+      "entropy": 0.6025584638118744,
+      "epoch": 4.186588921282799,
+      "grad_norm": 0.013231265358626842,
+      "learning_rate": 0.0002,
+      "loss": 0.6098158955574036,
+      "mean_token_accuracy": 0.7506869807839394,
+      "num_tokens": 20148878.0,
+      "step": 540
+    },
+    {
+      "entropy": 0.5955850780010223,
+      "epoch": 4.1943634596695825,
+      "grad_norm": 0.01381654478609562,
+      "learning_rate": 0.0002,
+      "loss": 0.6050733327865601,
+      "mean_token_accuracy": 0.7540115043520927,
+      "num_tokens": 20186590.0,
+      "step": 541
+    },
+    {
+      "entropy": 0.6100573241710663,
+      "epoch": 4.202137998056365,
+      "grad_norm": 0.01217405591160059,
+      "learning_rate": 0.0002,
+      "loss": 0.6091620922088623,
+      "mean_token_accuracy": 0.7501495629549026,
+      "num_tokens": 20224026.0,
+      "step": 542
+    },
+    {
+      "entropy": 0.6055784001946449,
+      "epoch": 4.209912536443149,
+      "grad_norm": 0.01034702267497778,
+      "learning_rate": 0.0002,
+      "loss": 0.6047437191009521,
+      "mean_token_accuracy": 0.7533470839262009,
+      "num_tokens": 20261538.0,
+      "step": 543
+    },
+    {
+      "entropy": 0.5968894883990288,
+      "epoch": 4.217687074829932,
+      "grad_norm": 0.013704034499824047,
+      "learning_rate": 0.0002,
+      "loss": 0.6015293002128601,
+      "mean_token_accuracy": 0.7537187412381172,
+      "num_tokens": 20298484.0,
+      "step": 544
+    },
+    {
+      "entropy": 0.5940377935767174,
+      "epoch": 4.225461613216715,
+      "grad_norm": 0.011667119339108467,
+      "learning_rate": 0.0002,
+      "loss": 0.5925427079200745,
+      "mean_token_accuracy": 0.7584238648414612,
+      "num_tokens": 20336353.0,
+      "step": 545
+    },
+    {
+      "entropy": 0.6106893792748451,
+      "epoch": 4.233236151603498,
+      "grad_norm": 0.01224418543279171,
+      "learning_rate": 0.0002,
+      "loss": 0.611664891242981,
+      "mean_token_accuracy": 0.7507496625185013,
+      "num_tokens": 20374019.0,
+      "step": 546
+    },
+    {
+      "entropy": 0.6042328551411629,
+      "epoch": 4.241010689990282,
+      "grad_norm": 0.01316764671355486,
+      "learning_rate": 0.0002,
+      "loss": 0.6062799096107483,
+      "mean_token_accuracy": 0.7534623667597771,
+      "num_tokens": 20411398.0,
+      "step": 547
+    },
+    {
+      "entropy": 0.6105416268110275,
+      "epoch": 4.2487852283770655,
+      "grad_norm": 0.012853113003075123,
+      "learning_rate": 0.0002,
+      "loss": 0.6092797517776489,
+      "mean_token_accuracy": 0.7522982135415077,
+      "num_tokens": 20448672.0,
+      "step": 548
+    },
+    {
+      "entropy": 0.5981989204883575,
+      "epoch": 4.256559766763848,
+      "grad_norm": 0.011458336375653744,
+      "learning_rate": 0.0002,
+      "loss": 0.5959842205047607,
+      "mean_token_accuracy": 0.7573426440358162,
+      "num_tokens": 20486126.0,
+      "step": 549
+    },
+    {
+      "entropy": 0.5939232558012009,
+      "epoch": 4.264334305150632,
+      "grad_norm": 0.013793280348181725,
+      "learning_rate": 0.0002,
+      "loss": 0.6013151407241821,
+      "mean_token_accuracy": 0.7532796338200569,
+      "num_tokens": 20523335.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.5925507992506027,
+      "epoch": 4.272108843537415,
+      "grad_norm": 0.012333216145634651,
+      "learning_rate": 0.0002,
+      "loss": 0.5946025848388672,
+      "mean_token_accuracy": 0.7598577961325645,
+      "num_tokens": 20560754.0,
+      "step": 551
+    },
+    {
+      "entropy": 0.6068093478679657,
+      "epoch": 4.279883381924198,
+      "grad_norm": 0.012082048691809177,
+      "learning_rate": 0.0002,
+      "loss": 0.6070762872695923,
+      "mean_token_accuracy": 0.7520242631435394,
+      "num_tokens": 20598383.0,
+      "step": 552
+    },
+    {
+      "entropy": 0.5919818729162216,
+      "epoch": 4.2876579203109815,
+      "grad_norm": 0.011891715228557587,
+      "learning_rate": 0.0002,
+      "loss": 0.5928791761398315,
+      "mean_token_accuracy": 0.7572562247514725,
+      "num_tokens": 20635386.0,
+      "step": 553
+    },
+    {
+      "entropy": 0.6074100211262703,
+      "epoch": 4.295432458697765,
+      "grad_norm": 0.014123337343335152,
+      "learning_rate": 0.0002,
+      "loss": 0.6073712706565857,
+      "mean_token_accuracy": 0.751992791891098,
+      "num_tokens": 20672892.0,
+      "step": 554
+    },
+    {
+      "entropy": 0.5886855572462082,
+      "epoch": 4.303206997084548,
+      "grad_norm": 0.0101819122210145,
+      "learning_rate": 0.0002,
+      "loss": 0.5912100076675415,
+      "mean_token_accuracy": 0.7583736777305603,
+      "num_tokens": 20710466.0,
+      "step": 555
+    },
+    {
+      "entropy": 0.6005367860198021,
+      "epoch": 4.310981535471331,
+      "grad_norm": 0.014812132343649864,
+      "learning_rate": 0.0002,
+      "loss": 0.6055353879928589,
+      "mean_token_accuracy": 0.7521958500146866,
+      "num_tokens": 20747846.0,
+      "step": 556
+    },
+    {
+      "entropy": 0.5981654673814774,
+      "epoch": 4.318756073858115,
+      "grad_norm": 0.011666829697787762,
+      "learning_rate": 0.0002,
+      "loss": 0.597842276096344,
+      "mean_token_accuracy": 0.7551755681633949,
+      "num_tokens": 20784741.0,
+      "step": 557
+    },
+    {
+      "entropy": 0.5991752445697784,
+      "epoch": 4.326530612244898,
+      "grad_norm": 0.012378910556435585,
+      "learning_rate": 0.0002,
+      "loss": 0.5976471900939941,
+      "mean_token_accuracy": 0.7547527700662613,
+      "num_tokens": 20822385.0,
+      "step": 558
+    },
+    {
+      "entropy": 0.6060795336961746,
+      "epoch": 4.334305150631681,
+      "grad_norm": 0.012490339577198029,
+      "learning_rate": 0.0002,
+      "loss": 0.6054461002349854,
+      "mean_token_accuracy": 0.7526181265711784,
+      "num_tokens": 20860006.0,
+      "step": 559
+    },
+    {
+      "entropy": 0.6072399467229843,
+      "epoch": 4.3420796890184645,
+      "grad_norm": 0.014051525853574276,
+      "learning_rate": 0.0002,
+      "loss": 0.6056400537490845,
+      "mean_token_accuracy": 0.7516866251826286,
+      "num_tokens": 20897385.0,
+      "step": 560
+    },
+    {
+      "entropy": 0.5980757176876068,
+      "epoch": 4.349854227405248,
+      "grad_norm": 0.011723170056939125,
+      "learning_rate": 0.0002,
+      "loss": 0.5963304042816162,
+      "mean_token_accuracy": 0.758495531976223,
+      "num_tokens": 20934914.0,
+      "step": 561
+    },
+    {
+      "entropy": 0.5847646594047546,
+      "epoch": 4.357628765792031,
+      "grad_norm": 0.013849250972270966,
+      "learning_rate": 0.0002,
+      "loss": 0.5895295143127441,
+      "mean_token_accuracy": 0.7575425952672958,
+      "num_tokens": 20972211.0,
+      "step": 562
+    },
+    {
+      "entropy": 0.5863577350974083,
+      "epoch": 4.365403304178814,
+      "grad_norm": 0.01223720796406269,
+      "learning_rate": 0.0002,
+      "loss": 0.5942325592041016,
+      "mean_token_accuracy": 0.7572259679436684,
+      "num_tokens": 21009384.0,
+      "step": 563
+    },
+    {
+      "entropy": 0.5997140482068062,
+      "epoch": 4.373177842565598,
+      "grad_norm": 0.014335358515381813,
+      "learning_rate": 0.0002,
+      "loss": 0.6024245619773865,
+      "mean_token_accuracy": 0.7510812431573868,
+      "num_tokens": 21046739.0,
+      "step": 564
+    },
+    {
+      "entropy": 0.6076342239975929,
+      "epoch": 4.380952380952381,
+      "grad_norm": 0.012497556395828724,
+      "learning_rate": 0.0002,
+      "loss": 0.6062162518501282,
+      "mean_token_accuracy": 0.7522109970450401,
+      "num_tokens": 21084263.0,
+      "step": 565
+    },
+    {
+      "entropy": 0.6001847013831139,
+      "epoch": 4.388726919339164,
+      "grad_norm": 0.010952596552670002,
+      "learning_rate": 0.0002,
+      "loss": 0.6020297408103943,
+      "mean_token_accuracy": 0.753420390188694,
+      "num_tokens": 21121894.0,
+      "step": 566
+    },
+    {
+      "entropy": 0.6051000133156776,
+      "epoch": 4.396501457725948,
+      "grad_norm": 0.013345146551728249,
+      "learning_rate": 0.0002,
+      "loss": 0.6103641986846924,
+      "mean_token_accuracy": 0.7505576461553574,
+      "num_tokens": 21159434.0,
+      "step": 567
+    },
+    {
+      "entropy": 0.5888576209545135,
+      "epoch": 4.404275996112731,
+      "grad_norm": 0.011659773997962475,
+      "learning_rate": 0.0002,
+      "loss": 0.5904802680015564,
+      "mean_token_accuracy": 0.7587999328970909,
+      "num_tokens": 21197109.0,
+      "step": 568
+    },
+    {
+      "entropy": 0.6007365435361862,
+      "epoch": 4.412050534499514,
+      "grad_norm": 0.010827448219060898,
+      "learning_rate": 0.0002,
+      "loss": 0.6000881195068359,
+      "mean_token_accuracy": 0.7543788328766823,
+      "num_tokens": 21234413.0,
+      "step": 569
+    },
+    {
+      "entropy": 0.5974342674016953,
+      "epoch": 4.419825072886297,
+      "grad_norm": 0.012830649502575397,
+      "learning_rate": 0.0002,
+      "loss": 0.6020520925521851,
+      "mean_token_accuracy": 0.7534354031085968,
+      "num_tokens": 21271758.0,
+      "step": 570
+    },
+    {
+      "entropy": 0.6090180650353432,
+      "epoch": 4.427599611273081,
+      "grad_norm": 0.012071631848812103,
+      "learning_rate": 0.0002,
+      "loss": 0.6067869663238525,
+      "mean_token_accuracy": 0.7518825978040695,
+      "num_tokens": 21309387.0,
+      "step": 571
+    },
+    {
+      "entropy": 0.5914840847253799,
+      "epoch": 4.4353741496598635,
+      "grad_norm": 0.013224054127931595,
+      "learning_rate": 0.0002,
+      "loss": 0.595637857913971,
+      "mean_token_accuracy": 0.7551431059837341,
+      "num_tokens": 21346475.0,
+      "step": 572
+    },
+    {
+      "entropy": 0.5997625440359116,
+      "epoch": 4.443148688046647,
+      "grad_norm": 0.011799960397183895,
+      "learning_rate": 0.0002,
+      "loss": 0.602319598197937,
+      "mean_token_accuracy": 0.7539779245853424,
+      "num_tokens": 21383957.0,
+      "step": 573
+    },
+    {
+      "entropy": 0.6102292612195015,
+      "epoch": 4.450923226433431,
+      "grad_norm": 0.01201384887099266,
+      "learning_rate": 0.0002,
+      "loss": 0.6108826398849487,
+      "mean_token_accuracy": 0.7517506182193756,
+      "num_tokens": 21421332.0,
+      "step": 574
+    },
+    {
+      "entropy": 0.5991413816809654,
+      "epoch": 4.458697764820214,
+      "grad_norm": 0.010899091139435768,
+      "learning_rate": 0.0002,
+      "loss": 0.596094012260437,
+      "mean_token_accuracy": 0.7565351948142052,
+      "num_tokens": 21458719.0,
+      "step": 575
+    },
+    {
+      "entropy": 0.5933232456445694,
+      "epoch": 4.466472303206997,
+      "grad_norm": 0.010439079254865646,
+      "learning_rate": 0.0002,
+      "loss": 0.5930327773094177,
+      "mean_token_accuracy": 0.7583755627274513,
+      "num_tokens": 21496264.0,
+      "step": 576
+    },
+    {
+      "entropy": 0.6057801619172096,
+      "epoch": 4.47424684159378,
+      "grad_norm": 0.011234867386519909,
+      "learning_rate": 0.0002,
+      "loss": 0.6073678135871887,
+      "mean_token_accuracy": 0.7530662715435028,
+      "num_tokens": 21533621.0,
+      "step": 577
+    },
+    {
+      "entropy": 0.6064127087593079,
+      "epoch": 4.482021379980564,
+      "grad_norm": 0.01104860007762909,
+      "learning_rate": 0.0002,
+      "loss": 0.6121537685394287,
+      "mean_token_accuracy": 0.7498331218957901,
+      "num_tokens": 21570804.0,
+      "step": 578
+    },
+    {
+      "entropy": 0.5948187559843063,
+      "epoch": 4.489795918367347,
+      "grad_norm": 0.01251922082155943,
+      "learning_rate": 0.0002,
+      "loss": 0.598996102809906,
+      "mean_token_accuracy": 0.7545123025774956,
+      "num_tokens": 21607975.0,
+      "step": 579
+    },
+    {
+      "entropy": 0.614658959209919,
+      "epoch": 4.49757045675413,
+      "grad_norm": 0.010701852850615978,
+      "learning_rate": 0.0002,
+      "loss": 0.6137805581092834,
+      "mean_token_accuracy": 0.7465793862938881,
+      "num_tokens": 21645423.0,
+      "step": 580
+    },
+    {
+      "entropy": 0.5894118696451187,
+      "epoch": 4.505344995140914,
+      "grad_norm": 0.012251303531229496,
+      "learning_rate": 0.0002,
+      "loss": 0.5877728462219238,
+      "mean_token_accuracy": 0.75811368227005,
+      "num_tokens": 21682411.0,
+      "step": 581
+    },
+    {
+      "entropy": 0.6109742075204849,
+      "epoch": 4.513119533527696,
+      "grad_norm": 0.011500229127705097,
+      "learning_rate": 0.0002,
+      "loss": 0.6115055084228516,
+      "mean_token_accuracy": 0.7521626874804497,
+      "num_tokens": 21720167.0,
+      "step": 582
+    },
+    {
+      "entropy": 0.591577798128128,
+      "epoch": 4.52089407191448,
+      "grad_norm": 0.011309963651001453,
+      "learning_rate": 0.0002,
+      "loss": 0.5981800556182861,
+      "mean_token_accuracy": 0.7563935667276382,
+      "num_tokens": 21758012.0,
+      "step": 583
+    },
+    {
+      "entropy": 0.6063844040036201,
+      "epoch": 4.528668610301263,
+      "grad_norm": 0.014559866860508919,
+      "learning_rate": 0.0002,
+      "loss": 0.6146761178970337,
+      "mean_token_accuracy": 0.7492925748229027,
+      "num_tokens": 21795528.0,
+      "step": 584
+    },
+    {
+      "entropy": 0.5918325483798981,
+      "epoch": 4.536443148688047,
+      "grad_norm": 0.011580277234315872,
+      "learning_rate": 0.0002,
+      "loss": 0.5908594131469727,
+      "mean_token_accuracy": 0.7565957456827164,
+      "num_tokens": 21833000.0,
+      "step": 585
+    },
+    {
+      "entropy": 0.612893208861351,
+      "epoch": 4.54421768707483,
+      "grad_norm": 0.01330646313726902,
+      "learning_rate": 0.0002,
+      "loss": 0.6055165529251099,
+      "mean_token_accuracy": 0.7545250579714775,
+      "num_tokens": 21870418.0,
+      "step": 586
+    },
+    {
+      "entropy": 0.5842209830880165,
+      "epoch": 4.551992225461613,
+      "grad_norm": 0.012110188603401184,
+      "learning_rate": 0.0002,
+      "loss": 0.5849249362945557,
+      "mean_token_accuracy": 0.7618754804134369,
+      "num_tokens": 21907507.0,
+      "step": 587
+    },
+    {
+      "entropy": 0.6068574264645576,
+      "epoch": 4.559766763848397,
+      "grad_norm": 0.012672025710344315,
+      "learning_rate": 0.0002,
+      "loss": 0.6145685911178589,
+      "mean_token_accuracy": 0.750059075653553,
+      "num_tokens": 21944988.0,
+      "step": 588
+    },
+    {
+      "entropy": 0.5868222415447235,
+      "epoch": 4.567541302235179,
+      "grad_norm": 0.011733310297131538,
+      "learning_rate": 0.0002,
+      "loss": 0.5891362428665161,
+      "mean_token_accuracy": 0.7605921477079391,
+      "num_tokens": 21982183.0,
+      "step": 589
+    },
+    {
+      "entropy": 0.6040458679199219,
+      "epoch": 4.575315840621963,
+      "grad_norm": 0.010982259176671505,
+      "learning_rate": 0.0002,
+      "loss": 0.6011037230491638,
+      "mean_token_accuracy": 0.7543822079896927,
+      "num_tokens": 22019570.0,
+      "step": 590
+    },
+    {
+      "entropy": 0.5989029258489609,
+      "epoch": 4.5830903790087465,
+      "grad_norm": 0.012942496687173843,
+      "learning_rate": 0.0002,
+      "loss": 0.5922666788101196,
+      "mean_token_accuracy": 0.7587461397051811,
+      "num_tokens": 22056831.0,
+      "step": 591
+    },
+    {
+      "entropy": 0.5930005833506584,
+      "epoch": 4.59086491739553,
+      "grad_norm": 0.010408308357000351,
+      "learning_rate": 0.0002,
+      "loss": 0.5919668078422546,
+      "mean_token_accuracy": 0.7574074566364288,
+      "num_tokens": 22094356.0,
+      "step": 592
+    },
+    {
+      "entropy": 0.5869537368416786,
+      "epoch": 4.598639455782313,
+      "grad_norm": 0.013977079652249813,
+      "learning_rate": 0.0002,
+      "loss": 0.5976507663726807,
+      "mean_token_accuracy": 0.7580091804265976,
+      "num_tokens": 22131838.0,
+      "step": 593
+    },
+    {
+      "entropy": 0.5787304788827896,
+      "epoch": 4.606413994169096,
+      "grad_norm": 0.013945071026682854,
+      "learning_rate": 0.0002,
+      "loss": 0.5908917188644409,
+      "mean_token_accuracy": 0.7597628980875015,
+      "num_tokens": 22169036.0,
+      "step": 594
+    },
+    {
+      "entropy": 0.6027527749538422,
+      "epoch": 4.61418853255588,
+      "grad_norm": 0.00982499960809946,
+      "learning_rate": 0.0002,
+      "loss": 0.6021907925605774,
+      "mean_token_accuracy": 0.7546107098460197,
+      "num_tokens": 22206352.0,
+      "step": 595
+    },
+    {
+      "entropy": 0.6077359542250633,
+      "epoch": 4.6219630709426625,
+      "grad_norm": 0.011836841702461243,
+      "learning_rate": 0.0002,
+      "loss": 0.6049139499664307,
+      "mean_token_accuracy": 0.7528544068336487,
+      "num_tokens": 22243663.0,
+      "step": 596
+    },
+    {
+      "entropy": 0.5822278633713722,
+      "epoch": 4.629737609329446,
+      "grad_norm": 0.011577391996979713,
+      "learning_rate": 0.0002,
+      "loss": 0.5828737020492554,
+      "mean_token_accuracy": 0.7616033405065536,
+      "num_tokens": 22281074.0,
+      "step": 597
+    },
+    {
+      "entropy": 0.5920702815055847,
+      "epoch": 4.6375121477162295,
+      "grad_norm": 0.01129237376153469,
+      "learning_rate": 0.0002,
+      "loss": 0.5975418090820312,
+      "mean_token_accuracy": 0.7549193054437637,
+      "num_tokens": 22318190.0,
+      "step": 598
+    },
+    {
+      "entropy": 0.5989422127604485,
+      "epoch": 4.645286686103013,
+      "grad_norm": 0.014368323609232903,
+      "learning_rate": 0.0002,
+      "loss": 0.6041890978813171,
+      "mean_token_accuracy": 0.7541765421628952,
+      "num_tokens": 22355560.0,
+      "step": 599
+    },
+    {
+      "entropy": 0.602604515850544,
+      "epoch": 4.653061224489796,
+      "grad_norm": 0.012781691737473011,
+      "learning_rate": 0.0002,
+      "loss": 0.6051725149154663,
+      "mean_token_accuracy": 0.7530051618814468,
+      "num_tokens": 22393652.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.6060863882303238,
+      "epoch": 4.660835762876579,
+      "grad_norm": 0.013375566340982914,
+      "learning_rate": 0.0002,
+      "loss": 0.6032074689865112,
+      "mean_token_accuracy": 0.7527770921587944,
+      "num_tokens": 22431345.0,
+      "step": 601
+    },
+    {
+      "entropy": 0.5950343683362007,
+      "epoch": 4.668610301263363,
+      "grad_norm": 0.012800985015928745,
+      "learning_rate": 0.0002,
+      "loss": 0.5933507084846497,
+      "mean_token_accuracy": 0.7560804039239883,
+      "num_tokens": 22469042.0,
+      "step": 602
+    },
+    {
+      "entropy": 0.5962737649679184,
+      "epoch": 4.6763848396501455,
+      "grad_norm": 0.014961260370910168,
+      "learning_rate": 0.0002,
+      "loss": 0.6006712913513184,
+      "mean_token_accuracy": 0.7546053975820541,
+      "num_tokens": 22506485.0,
+      "step": 603
+    },
+    {
+      "entropy": 0.606819398701191,
+      "epoch": 4.684159378036929,
+      "grad_norm": 0.011082889512181282,
+      "learning_rate": 0.0002,
+      "loss": 0.6065730452537537,
+      "mean_token_accuracy": 0.7529779449105263,
+      "num_tokens": 22544222.0,
+      "step": 604
+    },
+    {
+      "entropy": 0.577197976410389,
+      "epoch": 4.691933916423713,
+      "grad_norm": 0.014075525104999542,
+      "learning_rate": 0.0002,
+      "loss": 0.5835320949554443,
+      "mean_token_accuracy": 0.7612094879150391,
+      "num_tokens": 22581327.0,
+      "step": 605
+    },
+    {
+      "entropy": 0.6152481734752655,
+      "epoch": 4.699708454810495,
+      "grad_norm": 0.012122808955609798,
+      "learning_rate": 0.0002,
+      "loss": 0.6139184236526489,
+      "mean_token_accuracy": 0.7484366744756699,
+      "num_tokens": 22619089.0,
+      "step": 606
+    },
+    {
+      "entropy": 0.6120219603180885,
+      "epoch": 4.707482993197279,
+      "grad_norm": 0.011670667678117752,
+      "learning_rate": 0.0002,
+      "loss": 0.6125247478485107,
+      "mean_token_accuracy": 0.7505181655287743,
+      "num_tokens": 22656719.0,
+      "step": 607
+    },
+    {
+      "entropy": 0.5946462079882622,
+      "epoch": 4.715257531584062,
+      "grad_norm": 0.011501600034534931,
+      "learning_rate": 0.0002,
+      "loss": 0.600022554397583,
+      "mean_token_accuracy": 0.7565812543034554,
+      "num_tokens": 22694298.0,
+      "step": 608
+    },
+    {
+      "entropy": 0.5967997536063194,
+      "epoch": 4.723032069970845,
+      "grad_norm": 0.011824545450508595,
+      "learning_rate": 0.0002,
+      "loss": 0.6004019975662231,
+      "mean_token_accuracy": 0.7516064792871475,
+      "num_tokens": 22731785.0,
+      "step": 609
+    },
+    {
+      "entropy": 0.6012083142995834,
+      "epoch": 4.730806608357629,
+      "grad_norm": 0.009814469143748283,
+      "learning_rate": 0.0002,
+      "loss": 0.6001886129379272,
+      "mean_token_accuracy": 0.7529330998659134,
+      "num_tokens": 22769434.0,
+      "step": 610
+    },
+    {
+      "entropy": 0.6195448860526085,
+      "epoch": 4.738581146744412,
+      "grad_norm": 0.012913862243294716,
+      "learning_rate": 0.0002,
+      "loss": 0.6205934286117554,
+      "mean_token_accuracy": 0.7455445975065231,
+      "num_tokens": 22806676.0,
+      "step": 611
+    },
+    {
+      "entropy": 0.5957802012562752,
+      "epoch": 4.746355685131196,
+      "grad_norm": 0.011553781107068062,
+      "learning_rate": 0.0002,
+      "loss": 0.5980523824691772,
+      "mean_token_accuracy": 0.7560853660106659,
+      "num_tokens": 22843869.0,
+      "step": 612
+    },
+    {
+      "entropy": 0.6090810522437096,
+      "epoch": 4.754130223517978,
+      "grad_norm": 0.01271018385887146,
+      "learning_rate": 0.0002,
+      "loss": 0.6106667518615723,
+      "mean_token_accuracy": 0.7504242658615112,
+      "num_tokens": 22881762.0,
+      "step": 613
+    },
+    {
+      "entropy": 0.599302776157856,
+      "epoch": 4.761904761904762,
+      "grad_norm": 0.012581666931509972,
+      "learning_rate": 0.0002,
+      "loss": 0.5982247591018677,
+      "mean_token_accuracy": 0.7549748122692108,
+      "num_tokens": 22918922.0,
+      "step": 614
+    },
+    {
+      "entropy": 0.5844490081071854,
+      "epoch": 4.769679300291545,
+      "grad_norm": 0.011192350648343563,
+      "learning_rate": 0.0002,
+      "loss": 0.5844826698303223,
+      "mean_token_accuracy": 0.7625714018940926,
+      "num_tokens": 22956332.0,
+      "step": 615
+    },
+    {
+      "entropy": 0.5874569863080978,
+      "epoch": 4.777453838678328,
+      "grad_norm": 0.014751655049622059,
+      "learning_rate": 0.0002,
+      "loss": 0.5924921035766602,
+      "mean_token_accuracy": 0.7579960823059082,
+      "num_tokens": 22993598.0,
+      "step": 616
+    },
+    {
+      "entropy": 0.5977986827492714,
+      "epoch": 4.785228377065112,
+      "grad_norm": 0.011030304245650768,
+      "learning_rate": 0.0002,
+      "loss": 0.6010499000549316,
+      "mean_token_accuracy": 0.7557829320430756,
+      "num_tokens": 23031001.0,
+      "step": 617
+    },
+    {
+      "entropy": 0.599865585565567,
+      "epoch": 4.793002915451895,
+      "grad_norm": 0.011046946048736572,
+      "learning_rate": 0.0002,
+      "loss": 0.6028144359588623,
+      "mean_token_accuracy": 0.7549864649772644,
+      "num_tokens": 23068467.0,
+      "step": 618
+    },
+    {
+      "entropy": 0.6039430871605873,
+      "epoch": 4.800777453838679,
+      "grad_norm": 0.011677506379783154,
+      "learning_rate": 0.0002,
+      "loss": 0.6007229089736938,
+      "mean_token_accuracy": 0.7543792948126793,
+      "num_tokens": 23105954.0,
+      "step": 619
+    },
+    {
+      "entropy": 0.5956612825393677,
+      "epoch": 4.808551992225461,
+      "grad_norm": 0.011992014944553375,
+      "learning_rate": 0.0002,
+      "loss": 0.5964670181274414,
+      "mean_token_accuracy": 0.7563618123531342,
+      "num_tokens": 23143320.0,
+      "step": 620
+    },
+    {
+      "entropy": 0.5978101342916489,
+      "epoch": 4.816326530612245,
+      "grad_norm": 0.011897146701812744,
+      "learning_rate": 0.0002,
+      "loss": 0.6034868359565735,
+      "mean_token_accuracy": 0.7537665143609047,
+      "num_tokens": 23180916.0,
+      "step": 621
+    },
+    {
+      "entropy": 0.5931472703814507,
+      "epoch": 4.8241010689990285,
+      "grad_norm": 0.013459539040923119,
+      "learning_rate": 0.0002,
+      "loss": 0.6040017604827881,
+      "mean_token_accuracy": 0.7539473846554756,
+      "num_tokens": 23218627.0,
+      "step": 622
+    },
+    {
+      "entropy": 0.5916698947548866,
+      "epoch": 4.831875607385811,
+      "grad_norm": 0.011180682107806206,
+      "learning_rate": 0.0002,
+      "loss": 0.5919395685195923,
+      "mean_token_accuracy": 0.7593255490064621,
+      "num_tokens": 23255772.0,
+      "step": 623
+    },
+    {
+      "entropy": 0.6087042316794395,
+      "epoch": 4.839650145772595,
+      "grad_norm": 0.010437374003231525,
+      "learning_rate": 0.0002,
+      "loss": 0.6082122325897217,
+      "mean_token_accuracy": 0.7515282183885574,
+      "num_tokens": 23293061.0,
+      "step": 624
+    },
+    {
+      "entropy": 0.5966414213180542,
+      "epoch": 4.847424684159378,
+      "grad_norm": 0.011942134238779545,
+      "learning_rate": 0.0002,
+      "loss": 0.595150351524353,
+      "mean_token_accuracy": 0.7563839182257652,
+      "num_tokens": 23330109.0,
+      "step": 625
+    },
+    {
+      "entropy": 0.5881121456623077,
+      "epoch": 4.855199222546162,
+      "grad_norm": 0.011955403722822666,
+      "learning_rate": 0.0002,
+      "loss": 0.590496301651001,
+      "mean_token_accuracy": 0.7592649832367897,
+      "num_tokens": 23367496.0,
+      "step": 626
+    },
+    {
+      "entropy": 0.6133239194750786,
+      "epoch": 4.862973760932944,
+      "grad_norm": 0.012963500805199146,
+      "learning_rate": 0.0002,
+      "loss": 0.6236002445220947,
+      "mean_token_accuracy": 0.7459099292755127,
+      "num_tokens": 23404854.0,
+      "step": 627
+    },
+    {
+      "entropy": 0.5968378111720085,
+      "epoch": 4.870748299319728,
+      "grad_norm": 0.013238959014415741,
+      "learning_rate": 0.0002,
+      "loss": 0.5976919531822205,
+      "mean_token_accuracy": 0.7568426579236984,
+      "num_tokens": 23442227.0,
+      "step": 628
+    },
+    {
+      "entropy": 0.609023816883564,
+      "epoch": 4.8785228377065115,
+      "grad_norm": 0.012166466563940048,
+      "learning_rate": 0.0002,
+      "loss": 0.6067010164260864,
+      "mean_token_accuracy": 0.7532210573554039,
+      "num_tokens": 23479346.0,
+      "step": 629
+    },
+    {
+      "entropy": 0.6018581166863441,
+      "epoch": 4.886297376093294,
+      "grad_norm": 0.011096199974417686,
+      "learning_rate": 0.0002,
+      "loss": 0.5967071652412415,
+      "mean_token_accuracy": 0.7587796524167061,
+      "num_tokens": 23516484.0,
+      "step": 630
+    },
+    {
+      "entropy": 0.6068798899650574,
+      "epoch": 4.894071914480078,
+      "grad_norm": 0.0127568943426013,
+      "learning_rate": 0.0002,
+      "loss": 0.6059014797210693,
+      "mean_token_accuracy": 0.7524421662092209,
+      "num_tokens": 23553907.0,
+      "step": 631
+    },
+    {
+      "entropy": 0.5919808447360992,
+      "epoch": 4.901846452866861,
+      "grad_norm": 0.011789746582508087,
+      "learning_rate": 0.0002,
+      "loss": 0.596630334854126,
+      "mean_token_accuracy": 0.757590651512146,
+      "num_tokens": 23591693.0,
+      "step": 632
+    },
+    {
+      "entropy": 0.6100347265601158,
+      "epoch": 4.909620991253644,
+      "grad_norm": 0.013258693739771843,
+      "learning_rate": 0.0002,
+      "loss": 0.6171900033950806,
+      "mean_token_accuracy": 0.7499294355511665,
+      "num_tokens": 23629030.0,
+      "step": 633
+    },
+    {
+      "entropy": 0.5936124697327614,
+      "epoch": 4.9173955296404275,
+      "grad_norm": 0.011930654756724834,
+      "learning_rate": 0.0002,
+      "loss": 0.5924524664878845,
+      "mean_token_accuracy": 0.7581655383110046,
+      "num_tokens": 23666601.0,
+      "step": 634
+    },
+    {
+      "entropy": 0.6037433966994286,
+      "epoch": 4.925170068027211,
+      "grad_norm": 0.010547666810452938,
+      "learning_rate": 0.0002,
+      "loss": 0.6054027080535889,
+      "mean_token_accuracy": 0.7513353079557419,
+      "num_tokens": 23703928.0,
+      "step": 635
+    },
+    {
+      "entropy": 0.6012375876307487,
+      "epoch": 4.932944606413994,
+      "grad_norm": 0.013904466293752193,
+      "learning_rate": 0.0002,
+      "loss": 0.6056373119354248,
+      "mean_token_accuracy": 0.7527650520205498,
+      "num_tokens": 23741413.0,
+      "step": 636
+    },
+    {
+      "entropy": 0.6080401092767715,
+      "epoch": 4.940719144800777,
+      "grad_norm": 0.012218053452670574,
+      "learning_rate": 0.0002,
+      "loss": 0.6113067865371704,
+      "mean_token_accuracy": 0.7506906762719154,
+      "num_tokens": 23778797.0,
+      "step": 637
+    },
+    {
+      "entropy": 0.6042793691158295,
+      "epoch": 4.948493683187561,
+      "grad_norm": 0.012100595980882645,
+      "learning_rate": 0.0002,
+      "loss": 0.6038735508918762,
+      "mean_token_accuracy": 0.7550994455814362,
+      "num_tokens": 23816409.0,
+      "step": 638
+    },
+    {
+      "entropy": 0.5889951586723328,
+      "epoch": 4.956268221574344,
+      "grad_norm": 0.012997129000723362,
+      "learning_rate": 0.0002,
+      "loss": 0.5903059840202332,
+      "mean_token_accuracy": 0.7594538182020187,
+      "num_tokens": 23853723.0,
+      "step": 639
+    },
+    {
+      "entropy": 0.5925157070159912,
+      "epoch": 4.964042759961127,
+      "grad_norm": 0.015738792717456818,
+      "learning_rate": 0.0002,
+      "loss": 0.5936194658279419,
+      "mean_token_accuracy": 0.7593093439936638,
+      "num_tokens": 23891029.0,
+      "step": 640
+    },
+    {
+      "entropy": 0.5936921015381813,
+      "epoch": 4.9718172983479105,
+      "grad_norm": 0.01318391039967537,
+      "learning_rate": 0.0002,
+      "loss": 0.5958787202835083,
+      "mean_token_accuracy": 0.7549382373690605,
+      "num_tokens": 23928148.0,
+      "step": 641
+    },
+    {
+      "entropy": 0.6125310584902763,
+      "epoch": 4.979591836734694,
+      "grad_norm": 0.013194733299314976,
+      "learning_rate": 0.0002,
+      "loss": 0.6129195094108582,
+      "mean_token_accuracy": 0.7493861764669418,
+      "num_tokens": 23965085.0,
+      "step": 642
+    },
+    {
+      "entropy": 0.5964741930365562,
+      "epoch": 4.987366375121477,
+      "grad_norm": 0.013308845460414886,
+      "learning_rate": 0.0002,
+      "loss": 0.5979577302932739,
+      "mean_token_accuracy": 0.7562628239393234,
+      "num_tokens": 24002145.0,
+      "step": 643
+    },
+    {
+      "entropy": 0.5872825607657433,
+      "epoch": 4.99514091350826,
+      "grad_norm": 0.010709763504564762,
+      "learning_rate": 0.0002,
+      "loss": 0.5851190686225891,
+      "mean_token_accuracy": 0.7606633305549622,
+      "num_tokens": 24039501.0,
+      "step": 644
+    },
+    {
+      "entropy": 0.6111570835113526,
+      "epoch": 5.0,
+      "grad_norm": 0.012978802435100079,
+      "learning_rate": 0.0002,
+      "loss": 0.6106741428375244,
+      "mean_token_accuracy": 0.749509060382843,
+      "num_tokens": 24062815.0,
+      "step": 645
     }
   ],
   "logging_steps": 1,
-  "max_steps": 310,
+  "max_steps": 645,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 10,
+  "num_train_epochs": 5,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -3127,8 +6477,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.632026661790679e+18,
-  "train_batch_size": 16,
+  "total_flos": 2.2502112888923095e+18,
+  "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null
 }