diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -2,5418 +2,7768 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 10.0,
+  "epoch": 5.0,
   "eval_steps": 500,
-  "global_step": 540,
+  "global_step": 775,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "entropy": 1.2066851258277893,
-      "epoch": 0.018691588785046728,
-      "grad_norm": 0.7179942727088928,
+      "entropy": 1.1549250930547714,
+      "epoch": 0.006488240064882401,
+      "grad_norm": 0.24168352782726288,
       "learning_rate": 0.0002,
-      "loss": 2.6056714057922363,
-      "mean_token_accuracy": 0.5238651186227798,
-      "num_tokens": 78763.0,
+      "loss": 2.622039794921875,
+      "mean_token_accuracy": 0.5027666613459587,
+      "num_tokens": 36790.0,
       "step": 1
     },
     {
-      "entropy": 1.230730652809143,
-      "epoch": 0.037383177570093455,
-      "grad_norm": 0.39783212542533875,
+      "entropy": 1.274625539779663,
+      "epoch": 0.012976480129764802,
+      "grad_norm": 0.1995907425880432,
       "learning_rate": 0.0002,
-      "loss": 2.1227054595947266,
-      "mean_token_accuracy": 0.569368377327919,
-      "num_tokens": 158094.0,
+      "loss": 2.257164239883423,
+      "mean_token_accuracy": 0.5388679280877113,
+      "num_tokens": 74141.0,
       "step": 2
     },
     {
-      "entropy": 1.331688642501831,
-      "epoch": 0.056074766355140186,
-      "grad_norm": 0.37228652834892273,
+      "entropy": 1.4621597826480865,
+      "epoch": 0.019464720194647202,
+      "grad_norm": 0.14427633583545685,
       "learning_rate": 0.0002,
-      "loss": 1.8091474771499634,
-      "mean_token_accuracy": 0.592425987124443,
-      "num_tokens": 237343.0,
+      "loss": 1.8207428455352783,
+      "mean_token_accuracy": 0.5618570819497108,
+      "num_tokens": 111279.0,
       "step": 3
     },
     {
-      "entropy": 1.4335883259773254,
-      "epoch": 0.07476635514018691,
-      "grad_norm": 0.41868939995765686,
+      "entropy": 1.432983249425888,
+      "epoch": 0.025952960259529603,
+      "grad_norm": 0.11290092021226883,
       "learning_rate": 0.0002,
-      "loss": 1.6234431266784668,
-      "mean_token_accuracy": 0.5964363217353821,
-      "num_tokens": 314944.0,
+      "loss": 1.4740469455718994,
+      "mean_token_accuracy": 0.6100572720170021,
+      "num_tokens": 148749.0,
       "step": 4
     },
     {
-      "entropy": 1.409546136856079,
-      "epoch": 0.09345794392523364,
-      "grad_norm": 0.29376551508903503,
+      "entropy": 1.4418732523918152,
+      "epoch": 0.032441200324412,
+      "grad_norm": 0.13801905512809753,
       "learning_rate": 0.0002,
-      "loss": 1.4345183372497559,
-      "mean_token_accuracy": 0.6207377016544342,
-      "num_tokens": 395410.0,
+      "loss": 1.3789974451065063,
+      "mean_token_accuracy": 0.6047125160694122,
+      "num_tokens": 186530.0,
       "step": 5
     },
     {
-      "entropy": 1.398401916027069,
-      "epoch": 0.11214953271028037,
-      "grad_norm": 0.17153914272785187,
+      "entropy": 1.3603367805480957,
+      "epoch": 0.038929440389294405,
+      "grad_norm": 0.07707933336496353,
       "learning_rate": 0.0002,
-      "loss": 1.318544864654541,
-      "mean_token_accuracy": 0.6380362659692764,
-      "num_tokens": 474666.0,
+      "loss": 1.2482553720474243,
+      "mean_token_accuracy": 0.6377034857869148,
+      "num_tokens": 224090.0,
       "step": 6
     },
     {
-      "entropy": 1.3791473805904388,
-      "epoch": 0.1308411214953271,
-      "grad_norm": 0.13288195431232452,
+      "entropy": 1.2907896488904953,
+      "epoch": 0.0454176804541768,
+      "grad_norm": 0.05054641142487526,
       "learning_rate": 0.0002,
-      "loss": 1.2380807399749756,
-      "mean_token_accuracy": 0.646199494600296,
-      "num_tokens": 552886.0,
+      "loss": 1.1776490211486816,
+      "mean_token_accuracy": 0.6423852369189262,
+      "num_tokens": 261358.0,
       "step": 7
     },
     {
-      "entropy": 1.3407225608825684,
-      "epoch": 0.14953271028037382,
-      "grad_norm": 0.1418880969285965,
+      "entropy": 1.2394904792308807,
+      "epoch": 0.05190592051905921,
+      "grad_norm": 0.05160299688577652,
       "learning_rate": 0.0002,
-      "loss": 1.1673367023468018,
-      "mean_token_accuracy": 0.6485770791769028,
-      "num_tokens": 630368.0,
+      "loss": 1.1141722202301025,
+      "mean_token_accuracy": 0.6483801528811455,
+      "num_tokens": 298387.0,
       "step": 8
     },
     {
-      "entropy": 1.2557969391345978,
-      "epoch": 0.16822429906542055,
-      "grad_norm": 0.14714179933071136,
+      "entropy": 1.140880137681961,
+      "epoch": 0.058394160583941604,
+      "grad_norm": 0.05524353310465813,
       "learning_rate": 0.0002,
-      "loss": 1.0843346118927002,
-      "mean_token_accuracy": 0.6631819307804108,
-      "num_tokens": 710219.0,
+      "loss": 1.0428144931793213,
+      "mean_token_accuracy": 0.6648038625717163,
+      "num_tokens": 335681.0,
       "step": 9
     },
     {
-      "entropy": 1.1940588653087616,
-      "epoch": 0.18691588785046728,
-      "grad_norm": 0.14930284023284912,
+      "entropy": 1.0597130507230759,
+      "epoch": 0.064882400648824,
+      "grad_norm": 0.0610174834728241,
       "learning_rate": 0.0002,
-      "loss": 1.015075445175171,
-      "mean_token_accuracy": 0.679719015955925,
-      "num_tokens": 788388.0,
+      "loss": 0.9854335784912109,
+      "mean_token_accuracy": 0.6730685979127884,
+      "num_tokens": 373108.0,
       "step": 10
     },
     {
-      "entropy": 1.0720389187335968,
-      "epoch": 0.205607476635514,
-      "grad_norm": 0.10242773592472076,
+      "entropy": 0.9984031766653061,
+      "epoch": 0.07137064071370641,
+      "grad_norm": 0.05013473331928253,
       "learning_rate": 0.0002,
-      "loss": 0.937964141368866,
-      "mean_token_accuracy": 0.6943712085485458,
-      "num_tokens": 870578.0,
+      "loss": 0.9311432838439941,
+      "mean_token_accuracy": 0.6808731183409691,
+      "num_tokens": 410188.0,
       "step": 11
     },
     {
-      "entropy": 0.9909108728170395,
-      "epoch": 0.22429906542056074,
-      "grad_norm": 0.09138211607933044,
+      "entropy": 0.9519696608185768,
+      "epoch": 0.07785888077858881,
+      "grad_norm": 0.7003989219665527,
       "learning_rate": 0.0002,
-      "loss": 0.9100936651229858,
-      "mean_token_accuracy": 0.6939836889505386,
-      "num_tokens": 950766.0,
+      "loss": 0.9051742553710938,
+      "mean_token_accuracy": 0.6799700632691383,
+      "num_tokens": 447666.0,
       "step": 12
     },
     {
-      "entropy": 0.9221479296684265,
-      "epoch": 0.24299065420560748,
-      "grad_norm": 0.10024455189704895,
+      "entropy": 0.905988834798336,
+      "epoch": 0.08434712084347121,
+      "grad_norm": 0.05167749896645546,
       "learning_rate": 0.0002,
-      "loss": 0.8748308420181274,
-      "mean_token_accuracy": 0.6973191946744919,
-      "num_tokens": 1029253.0,
+      "loss": 0.8597391843795776,
+      "mean_token_accuracy": 0.6946859285235405,
+      "num_tokens": 484844.0,
       "step": 13
     },
     {
-      "entropy": 0.8531733900308609,
-      "epoch": 0.2616822429906542,
-      "grad_norm": 0.09876618534326553,
+      "entropy": 0.8613085150718689,
+      "epoch": 0.0908353609083536,
+      "grad_norm": 0.0768115222454071,
       "learning_rate": 0.0002,
-      "loss": 0.8316620588302612,
-      "mean_token_accuracy": 0.705431193113327,
-      "num_tokens": 1107714.0,
+      "loss": 0.8413710594177246,
+      "mean_token_accuracy": 0.6983600705862045,
+      "num_tokens": 521975.0,
       "step": 14
     },
     {
-      "entropy": 0.8102366924285889,
-      "epoch": 0.2803738317757009,
-      "grad_norm": 0.09547463059425354,
+      "entropy": 0.8102021664381027,
+      "epoch": 0.09732360097323602,
+      "grad_norm": 0.5994402170181274,
       "learning_rate": 0.0002,
-      "loss": 0.8144597411155701,
-      "mean_token_accuracy": 0.7082843035459518,
-      "num_tokens": 1187423.0,
+      "loss": 0.81607985496521,
+      "mean_token_accuracy": 0.7021179124712944,
+      "num_tokens": 558889.0,
       "step": 15
     },
     {
-      "entropy": 0.7830685824155807,
-      "epoch": 0.29906542056074764,
-      "grad_norm": 0.08728648722171783,
+      "entropy": 0.8057567626237869,
+      "epoch": 0.10381184103811841,
+      "grad_norm": 0.03823373466730118,
       "learning_rate": 0.0002,
-      "loss": 0.7878029346466064,
-      "mean_token_accuracy": 0.7123377323150635,
-      "num_tokens": 1266800.0,
+      "loss": 0.7905861139297485,
+      "mean_token_accuracy": 0.7055389359593391,
+      "num_tokens": 596455.0,
       "step": 16
     },
     {
-      "entropy": 0.771837130188942,
-      "epoch": 0.3177570093457944,
-      "grad_norm": 0.07688825577497482,
+      "entropy": 0.7923594415187836,
+      "epoch": 0.11030008110300081,
+      "grad_norm": 0.0739479511976242,
       "learning_rate": 0.0002,
-      "loss": 0.7640396356582642,
-      "mean_token_accuracy": 0.7166268229484558,
-      "num_tokens": 1345669.0,
+      "loss": 0.769114077091217,
+      "mean_token_accuracy": 0.7118282616138458,
+      "num_tokens": 634110.0,
       "step": 17
     },
     {
-      "entropy": 0.7662298083305359,
-      "epoch": 0.3364485981308411,
-      "grad_norm": 0.08384118229150772,
+      "entropy": 0.7812566831707954,
+      "epoch": 0.11678832116788321,
+      "grad_norm": 0.03668873757123947,
       "learning_rate": 0.0002,
-      "loss": 0.7458442449569702,
-      "mean_token_accuracy": 0.7201817631721497,
-      "num_tokens": 1424919.0,
+      "loss": 0.7668346762657166,
+      "mean_token_accuracy": 0.7075815051794052,
+      "num_tokens": 671655.0,
       "step": 18
     },
     {
-      "entropy": 0.7514005601406097,
-      "epoch": 0.35514018691588783,
-      "grad_norm": 0.08126289397478104,
+      "entropy": 0.7474964931607246,
+      "epoch": 0.12327656123276562,
+      "grad_norm": 0.041200920939445496,
       "learning_rate": 0.0002,
-      "loss": 0.7338656187057495,
-      "mean_token_accuracy": 0.7214008867740631,
-      "num_tokens": 1503286.0,
+      "loss": 0.7393766641616821,
+      "mean_token_accuracy": 0.716952882707119,
+      "num_tokens": 708824.0,
       "step": 19
     },
     {
-      "entropy": 0.7352934032678604,
-      "epoch": 0.37383177570093457,
-      "grad_norm": 0.0703548714518547,
+      "entropy": 0.7383133918046951,
+      "epoch": 0.129764801297648,
+      "grad_norm": 0.0367242768406868,
       "learning_rate": 0.0002,
-      "loss": 0.7204664945602417,
-      "mean_token_accuracy": 0.7259397804737091,
-      "num_tokens": 1582654.0,
+      "loss": 0.728330135345459,
+      "mean_token_accuracy": 0.7191555127501488,
+      "num_tokens": 746274.0,
       "step": 20
     },
     {
-      "entropy": 0.704582080245018,
-      "epoch": 0.3925233644859813,
-      "grad_norm": 0.0680006816983223,
+      "entropy": 0.7354703769087791,
+      "epoch": 0.1362530413625304,
+      "grad_norm": 0.039265185594558716,
       "learning_rate": 0.0002,
-      "loss": 0.6956260800361633,
-      "mean_token_accuracy": 0.7327138334512711,
-      "num_tokens": 1662345.0,
+      "loss": 0.7205026149749756,
+      "mean_token_accuracy": 0.7199463173747063,
+      "num_tokens": 783720.0,
       "step": 21
     },
     {
-      "entropy": 0.7041595876216888,
-      "epoch": 0.411214953271028,
-      "grad_norm": 0.07259424775838852,
+      "entropy": 0.7178014516830444,
+      "epoch": 0.14274128142741282,
+      "grad_norm": 0.03647130727767944,
       "learning_rate": 0.0002,
-      "loss": 0.7061991691589355,
-      "mean_token_accuracy": 0.7257621735334396,
-      "num_tokens": 1740618.0,
+      "loss": 0.7003612518310547,
+      "mean_token_accuracy": 0.7242974415421486,
+      "num_tokens": 821241.0,
       "step": 22
     },
     {
-      "entropy": 0.6996002197265625,
-      "epoch": 0.42990654205607476,
-      "grad_norm": 0.05930614843964577,
+      "entropy": 0.7019922286272049,
+      "epoch": 0.1492295214922952,
+      "grad_norm": 0.03146970644593239,
       "learning_rate": 0.0002,
-      "loss": 0.6944414377212524,
-      "mean_token_accuracy": 0.7298829406499863,
-      "num_tokens": 1818731.0,
+      "loss": 0.6875081062316895,
+      "mean_token_accuracy": 0.7302066311240196,
+      "num_tokens": 858171.0,
       "step": 23
     },
     {
-      "entropy": 0.6977217197418213,
-      "epoch": 0.4485981308411215,
-      "grad_norm": 0.059824682772159576,
+      "entropy": 0.691544882953167,
+      "epoch": 0.15571776155717762,
+      "grad_norm": 0.03450683504343033,
       "learning_rate": 0.0002,
-      "loss": 0.68711256980896,
-      "mean_token_accuracy": 0.7326766103506088,
-      "num_tokens": 1896889.0,
+      "loss": 0.6847634315490723,
+      "mean_token_accuracy": 0.7311759144067764,
+      "num_tokens": 895368.0,
       "step": 24
     },
     {
-      "entropy": 0.6951911151409149,
-      "epoch": 0.4672897196261682,
-      "grad_norm": 0.060844291001558304,
+      "entropy": 0.6791905760765076,
+      "epoch": 0.16220600162206,
+      "grad_norm": 0.029409531503915787,
       "learning_rate": 0.0002,
-      "loss": 0.681706428527832,
-      "mean_token_accuracy": 0.731402724981308,
-      "num_tokens": 1974795.0,
+      "loss": 0.672599732875824,
+      "mean_token_accuracy": 0.7371302843093872,
+      "num_tokens": 932589.0,
       "step": 25
     },
     {
-      "entropy": 0.6741717159748077,
-      "epoch": 0.48598130841121495,
-      "grad_norm": 0.05977668985724449,
+      "entropy": 0.6795228570699692,
+      "epoch": 0.16869424168694241,
+      "grad_norm": 0.029320087283849716,
       "learning_rate": 0.0002,
-      "loss": 0.6651660203933716,
-      "mean_token_accuracy": 0.7382005453109741,
-      "num_tokens": 2052366.0,
+      "loss": 0.6811176538467407,
+      "mean_token_accuracy": 0.7309242188930511,
+      "num_tokens": 969558.0,
       "step": 26
     },
     {
-      "entropy": 0.658872976899147,
-      "epoch": 0.5046728971962616,
-      "grad_norm": 0.05988093093037605,
+      "entropy": 0.6630020141601562,
+      "epoch": 0.17518248175182483,
+      "grad_norm": 0.025668611750006676,
       "learning_rate": 0.0002,
-      "loss": 0.6478220224380493,
-      "mean_token_accuracy": 0.7451245039701462,
-      "num_tokens": 2132623.0,
+      "loss": 0.6629281044006348,
+      "mean_token_accuracy": 0.7399500384926796,
+      "num_tokens": 1007155.0,
       "step": 27
     },
     {
-      "entropy": 0.6636136025190353,
-      "epoch": 0.5233644859813084,
-      "grad_norm": 0.05302215740084648,
+      "entropy": 0.6723117381334305,
+      "epoch": 0.1816707218167072,
+      "grad_norm": 0.025801967829465866,
       "learning_rate": 0.0002,
-      "loss": 0.6561492681503296,
-      "mean_token_accuracy": 0.742748498916626,
-      "num_tokens": 2212171.0,
+      "loss": 0.672906756401062,
+      "mean_token_accuracy": 0.7337809279561043,
+      "num_tokens": 1044459.0,
       "step": 28
     },
     {
-      "entropy": 0.6493570953607559,
-      "epoch": 0.5420560747663551,
-      "grad_norm": 0.058367419987916946,
+      "entropy": 0.671604260802269,
+      "epoch": 0.18815896188158962,
+      "grad_norm": 0.026239318773150444,
       "learning_rate": 0.0002,
-      "loss": 0.6493097543716431,
-      "mean_token_accuracy": 0.7458167970180511,
-      "num_tokens": 2290737.0,
+      "loss": 0.6697608828544617,
+      "mean_token_accuracy": 0.7337322384119034,
+      "num_tokens": 1081975.0,
       "step": 29
     },
     {
-      "entropy": 0.6432235687971115,
-      "epoch": 0.5607476635514018,
-      "grad_norm": 0.05448915436863899,
+      "entropy": 0.646067701280117,
+      "epoch": 0.19464720194647203,
+      "grad_norm": 0.02587837167084217,
       "learning_rate": 0.0002,
-      "loss": 0.6421418190002441,
-      "mean_token_accuracy": 0.7468905597925186,
-      "num_tokens": 2369888.0,
+      "loss": 0.6468892097473145,
+      "mean_token_accuracy": 0.7421439811587334,
+      "num_tokens": 1119148.0,
       "step": 30
     },
     {
-      "entropy": 0.641756996512413,
-      "epoch": 0.5794392523364486,
-      "grad_norm": 0.05883530527353287,
+      "entropy": 0.6522824168205261,
+      "epoch": 0.20113544201135442,
+      "grad_norm": 0.0248495414853096,
       "learning_rate": 0.0002,
-      "loss": 0.6386466026306152,
-      "mean_token_accuracy": 0.7485450953245163,
-      "num_tokens": 2450096.0,
+      "loss": 0.6558438539505005,
+      "mean_token_accuracy": 0.7381908968091011,
+      "num_tokens": 1156351.0,
       "step": 31
     },
     {
-      "entropy": 0.6456558853387833,
-      "epoch": 0.5981308411214953,
-      "grad_norm": 0.043611567467451096,
+      "entropy": 0.6425730809569359,
+      "epoch": 0.20762368207623683,
+      "grad_norm": 0.02851717919111252,
       "learning_rate": 0.0002,
-      "loss": 0.6375504732131958,
-      "mean_token_accuracy": 0.7480568140745163,
-      "num_tokens": 2529348.0,
+      "loss": 0.6539334654808044,
+      "mean_token_accuracy": 0.7376680299639702,
+      "num_tokens": 1193707.0,
       "step": 32
     },
     {
-      "entropy": 0.6468172371387482,
-      "epoch": 0.616822429906542,
-      "grad_norm": 0.05578931048512459,
+      "entropy": 0.6417841091752052,
+      "epoch": 0.2141119221411192,
+      "grad_norm": 0.02549365535378456,
       "learning_rate": 0.0002,
-      "loss": 0.6400123238563538,
-      "mean_token_accuracy": 0.7488409727811813,
-      "num_tokens": 2608327.0,
+      "loss": 0.6482184529304504,
+      "mean_token_accuracy": 0.7417171224951744,
+      "num_tokens": 1230700.0,
       "step": 33
     },
     {
-      "entropy": 0.6378584653139114,
-      "epoch": 0.6355140186915887,
-      "grad_norm": 0.04630431905388832,
+      "entropy": 0.6465194374322891,
+      "epoch": 0.22060016220600162,
+      "grad_norm": 0.025215625762939453,
       "learning_rate": 0.0002,
-      "loss": 0.6317487955093384,
-      "mean_token_accuracy": 0.7513478249311447,
-      "num_tokens": 2688604.0,
+      "loss": 0.6494179964065552,
+      "mean_token_accuracy": 0.7402999773621559,
+      "num_tokens": 1267782.0,
       "step": 34
     },
     {
-      "entropy": 0.6285719871520996,
-      "epoch": 0.6542056074766355,
-      "grad_norm": 0.04269322752952576,
+      "entropy": 0.6392467468976974,
+      "epoch": 0.22708840227088403,
+      "grad_norm": 0.023007551208138466,
       "learning_rate": 0.0002,
-      "loss": 0.6267668008804321,
-      "mean_token_accuracy": 0.7520004361867905,
-      "num_tokens": 2767770.0,
+      "loss": 0.6377967000007629,
+      "mean_token_accuracy": 0.7462463900446892,
+      "num_tokens": 1304976.0,
       "step": 35
     },
     {
-      "entropy": 0.613593116402626,
-      "epoch": 0.6728971962616822,
-      "grad_norm": 0.04195692762732506,
+      "entropy": 0.6491241455078125,
+      "epoch": 0.23357664233576642,
+      "grad_norm": 0.022631216794252396,
       "learning_rate": 0.0002,
-      "loss": 0.6151101589202881,
-      "mean_token_accuracy": 0.7558344155550003,
-      "num_tokens": 2846844.0,
+      "loss": 0.6419650316238403,
+      "mean_token_accuracy": 0.7427782192826271,
+      "num_tokens": 1342405.0,
       "step": 36
     },
     {
-      "entropy": 0.619237095117569,
-      "epoch": 0.6915887850467289,
-      "grad_norm": 0.04547283425927162,
+      "entropy": 0.6426859274506569,
+      "epoch": 0.24006488240064883,
+      "grad_norm": 0.020762791857123375,
       "learning_rate": 0.0002,
-      "loss": 0.6201531291007996,
-      "mean_token_accuracy": 0.7537788301706314,
-      "num_tokens": 2924106.0,
+      "loss": 0.638580322265625,
+      "mean_token_accuracy": 0.74485033005476,
+      "num_tokens": 1379588.0,
       "step": 37
     },
     {
-      "entropy": 0.6138854026794434,
-      "epoch": 0.7102803738317757,
-      "grad_norm": 0.045413266867399216,
+      "entropy": 0.6500778123736382,
+      "epoch": 0.24655312246553124,
+      "grad_norm": 0.01766084134578705,
       "learning_rate": 0.0002,
-      "loss": 0.6134829521179199,
-      "mean_token_accuracy": 0.7565835565328598,
-      "num_tokens": 3003024.0,
+      "loss": 0.6436338424682617,
+      "mean_token_accuracy": 0.7424680069088936,
+      "num_tokens": 1417229.0,
       "step": 38
     },
     {
-      "entropy": 0.620908111333847,
-      "epoch": 0.7289719626168224,
-      "grad_norm": 0.03639939799904823,
+      "entropy": 0.654863677918911,
+      "epoch": 0.25304136253041365,
+      "grad_norm": 0.018832355737686157,
       "learning_rate": 0.0002,
-      "loss": 0.6201748847961426,
-      "mean_token_accuracy": 0.7546153366565704,
-      "num_tokens": 3081301.0,
+      "loss": 0.6571300029754639,
+      "mean_token_accuracy": 0.7378578186035156,
+      "num_tokens": 1454802.0,
       "step": 39
     },
     {
-      "entropy": 0.6099251806735992,
-      "epoch": 0.7476635514018691,
-      "grad_norm": 0.03840401768684387,
+      "entropy": 0.6454419046640396,
+      "epoch": 0.259529602595296,
+      "grad_norm": 0.018613073974847794,
       "learning_rate": 0.0002,
-      "loss": 0.61121666431427,
-      "mean_token_accuracy": 0.7575128823518753,
-      "num_tokens": 3159456.0,
+      "loss": 0.6524553298950195,
+      "mean_token_accuracy": 0.7370505034923553,
+      "num_tokens": 1492073.0,
       "step": 40
     },
     {
-      "entropy": 0.6060190051794052,
-      "epoch": 0.7663551401869159,
-      "grad_norm": 0.03731069341301918,
+      "entropy": 0.6527602002024651,
+      "epoch": 0.2660178426601784,
+      "grad_norm": 0.01693454198539257,
       "learning_rate": 0.0002,
-      "loss": 0.6043084859848022,
-      "mean_token_accuracy": 0.7605673968791962,
-      "num_tokens": 3237706.0,
+      "loss": 0.6585220098495483,
+      "mean_token_accuracy": 0.73494041711092,
+      "num_tokens": 1529432.0,
       "step": 41
     },
     {
-      "entropy": 0.6159602850675583,
-      "epoch": 0.7850467289719626,
-      "grad_norm": 0.036441072821617126,
+      "entropy": 0.6482313647866249,
+      "epoch": 0.2725060827250608,
+      "grad_norm": 0.015154066495597363,
       "learning_rate": 0.0002,
-      "loss": 0.6172773838043213,
-      "mean_token_accuracy": 0.7549006193876266,
-      "num_tokens": 3316458.0,
+      "loss": 0.6487522125244141,
+      "mean_token_accuracy": 0.7386852130293846,
+      "num_tokens": 1566849.0,
       "step": 42
     },
     {
-      "entropy": 0.6140847951173782,
-      "epoch": 0.8037383177570093,
-      "grad_norm": 0.03407798334956169,
+      "entropy": 0.6501431986689568,
+      "epoch": 0.27899432278994324,
+      "grad_norm": 0.015506139025092125,
       "learning_rate": 0.0002,
-      "loss": 0.6130833625793457,
-      "mean_token_accuracy": 0.7558080703020096,
-      "num_tokens": 3395531.0,
+      "loss": 0.6508647203445435,
+      "mean_token_accuracy": 0.7396439164876938,
+      "num_tokens": 1603797.0,
       "step": 43
     },
     {
-      "entropy": 0.6078946739435196,
-      "epoch": 0.822429906542056,
-      "grad_norm": 0.03341364115476608,
+      "entropy": 0.6399750784039497,
+      "epoch": 0.28548256285482565,
+      "grad_norm": 0.014903928153216839,
       "learning_rate": 0.0002,
-      "loss": 0.6066131591796875,
-      "mean_token_accuracy": 0.7582335621118546,
-      "num_tokens": 3473820.0,
+      "loss": 0.639168381690979,
+      "mean_token_accuracy": 0.743517704308033,
+      "num_tokens": 1641043.0,
       "step": 44
     },
     {
-      "entropy": 0.6053218245506287,
-      "epoch": 0.8411214953271028,
-      "grad_norm": 0.02799089439213276,
+      "entropy": 0.6295603513717651,
+      "epoch": 0.291970802919708,
+      "grad_norm": 0.014574805274605751,
       "learning_rate": 0.0002,
-      "loss": 0.600709855556488,
-      "mean_token_accuracy": 0.7598105370998383,
-      "num_tokens": 3553354.0,
+      "loss": 0.6305776834487915,
+      "mean_token_accuracy": 0.746990330517292,
+      "num_tokens": 1677698.0,
       "step": 45
     },
     {
-      "entropy": 0.6008974015712738,
-      "epoch": 0.8598130841121495,
-      "grad_norm": 0.03266238048672676,
+      "entropy": 0.6334760338068008,
+      "epoch": 0.2984590429845904,
+      "grad_norm": 0.016435733065009117,
       "learning_rate": 0.0002,
-      "loss": 0.5975630879402161,
-      "mean_token_accuracy": 0.7631259560585022,
-      "num_tokens": 3632012.0,
+      "loss": 0.631117582321167,
+      "mean_token_accuracy": 0.7475783005356789,
+      "num_tokens": 1714954.0,
       "step": 46
     },
     {
-      "entropy": 0.6067604422569275,
-      "epoch": 0.8785046728971962,
-      "grad_norm": 0.03338483348488808,
+      "entropy": 0.6437651589512825,
+      "epoch": 0.30494728304947283,
+      "grad_norm": 0.016290629282593727,
       "learning_rate": 0.0002,
-      "loss": 0.6048734188079834,
-      "mean_token_accuracy": 0.7590593695640564,
-      "num_tokens": 3711139.0,
+      "loss": 0.6405509114265442,
+      "mean_token_accuracy": 0.7430335581302643,
+      "num_tokens": 1751913.0,
       "step": 47
     },
     {
-      "entropy": 0.5967728197574615,
-      "epoch": 0.897196261682243,
-      "grad_norm": 0.025535311549901962,
+      "entropy": 0.6363071575760841,
+      "epoch": 0.31143552311435524,
+      "grad_norm": 0.01479713711887598,
       "learning_rate": 0.0002,
-      "loss": 0.5967084765434265,
-      "mean_token_accuracy": 0.7615634948015213,
-      "num_tokens": 3790547.0,
+      "loss": 0.6342425346374512,
+      "mean_token_accuracy": 0.7451304048299789,
+      "num_tokens": 1789211.0,
       "step": 48
     },
     {
-      "entropy": 0.6170180141925812,
-      "epoch": 0.9158878504672897,
-      "grad_norm": 0.0306242685765028,
+      "entropy": 0.6442478150129318,
+      "epoch": 0.31792376317923765,
+      "grad_norm": 0.014846265316009521,
       "learning_rate": 0.0002,
-      "loss": 0.6141654253005981,
-      "mean_token_accuracy": 0.7549907118082047,
-      "num_tokens": 3868857.0,
+      "loss": 0.641524612903595,
+      "mean_token_accuracy": 0.7414352521300316,
+      "num_tokens": 1826606.0,
       "step": 49
     },
     {
-      "entropy": 0.6021094918251038,
-      "epoch": 0.9345794392523364,
-      "grad_norm": 0.03037024661898613,
+      "entropy": 0.6363928020000458,
+      "epoch": 0.32441200324412,
+      "grad_norm": 0.01481616124510765,
       "learning_rate": 0.0002,
-      "loss": 0.6032395362854004,
-      "mean_token_accuracy": 0.7590876370668411,
-      "num_tokens": 3948653.0,
+      "loss": 0.6363080739974976,
+      "mean_token_accuracy": 0.7444929406046867,
+      "num_tokens": 1863966.0,
       "step": 50
     },
     {
-      "entropy": 0.6060637980699539,
-      "epoch": 0.9532710280373832,
-      "grad_norm": 0.030959416180849075,
+      "entropy": 0.6335690394043922,
+      "epoch": 0.3309002433090024,
+      "grad_norm": 0.015328066423535347,
       "learning_rate": 0.0002,
-      "loss": 0.6047207117080688,
-      "mean_token_accuracy": 0.7576227784156799,
-      "num_tokens": 4027651.0,
+      "loss": 0.6387162208557129,
+      "mean_token_accuracy": 0.7436776608228683,
+      "num_tokens": 1900997.0,
       "step": 51
     },
     {
-      "entropy": 0.5999785810709,
-      "epoch": 0.9719626168224299,
-      "grad_norm": 0.03629245609045029,
+      "entropy": 0.6325219571590424,
+      "epoch": 0.33738848337388483,
+      "grad_norm": 0.015631280839443207,
       "learning_rate": 0.0002,
-      "loss": 0.6036806106567383,
-      "mean_token_accuracy": 0.7583939582109451,
-      "num_tokens": 4106355.0,
+      "loss": 0.6400341987609863,
+      "mean_token_accuracy": 0.7412212640047073,
+      "num_tokens": 1937974.0,
       "step": 52
     },
     {
-      "entropy": 0.6017800122499466,
-      "epoch": 0.9906542056074766,
-      "grad_norm": 0.038867365568876266,
+      "entropy": 0.63508290797472,
+      "epoch": 0.34387672343876724,
+      "grad_norm": 0.014866264536976814,
       "learning_rate": 0.0002,
-      "loss": 0.6031535863876343,
-      "mean_token_accuracy": 0.7586866319179535,
-      "num_tokens": 4185523.0,
+      "loss": 0.6396467089653015,
+      "mean_token_accuracy": 0.7431643679738045,
+      "num_tokens": 1975204.0,
       "step": 53
     },
     {
-      "entropy": 0.6058991253376007,
-      "epoch": 1.0,
-      "grad_norm": 0.02517516352236271,
+      "entropy": 0.6382915005087852,
+      "epoch": 0.35036496350364965,
+      "grad_norm": 0.013517456129193306,
       "learning_rate": 0.0002,
-      "loss": 0.6025224924087524,
-      "mean_token_accuracy": 0.7598057687282562,
-      "num_tokens": 4225197.0,
+      "loss": 0.6404691338539124,
+      "mean_token_accuracy": 0.7433327361941338,
+      "num_tokens": 2012218.0,
       "step": 54
     },
     {
-      "entropy": 0.6096542477607727,
-      "epoch": 1.0186915887850467,
-      "grad_norm": 0.03110332600772381,
+      "entropy": 0.6375824585556984,
+      "epoch": 0.35685320356853206,
+      "grad_norm": 0.015061435289680958,
       "learning_rate": 0.0002,
-      "loss": 0.6018860340118408,
-      "mean_token_accuracy": 0.760483518242836,
-      "num_tokens": 4305195.0,
+      "loss": 0.6400342583656311,
+      "mean_token_accuracy": 0.7421080023050308,
+      "num_tokens": 2049462.0,
       "step": 55
     },
     {
-      "entropy": 0.6131730824708939,
-      "epoch": 1.0373831775700935,
-      "grad_norm": 0.03143934905529022,
+      "entropy": 0.6377979964017868,
+      "epoch": 0.3633414436334144,
+      "grad_norm": 0.016312120482325554,
       "learning_rate": 0.0002,
-      "loss": 0.6093547344207764,
-      "mean_token_accuracy": 0.7554158419370651,
-      "num_tokens": 4383674.0,
+      "loss": 0.635837197303772,
+      "mean_token_accuracy": 0.7426418736577034,
+      "num_tokens": 2086665.0,
       "step": 56
     },
     {
-      "entropy": 0.6045864671468735,
-      "epoch": 1.0560747663551402,
-      "grad_norm": 0.02568645216524601,
+      "entropy": 0.6413707137107849,
+      "epoch": 0.36982968369829683,
+      "grad_norm": 0.012621643021702766,
       "learning_rate": 0.0002,
-      "loss": 0.6015181541442871,
-      "mean_token_accuracy": 0.758747324347496,
-      "num_tokens": 4462590.0,
+      "loss": 0.6389581561088562,
+      "mean_token_accuracy": 0.7432198747992516,
+      "num_tokens": 2124115.0,
       "step": 57
     },
     {
-      "entropy": 0.5954300314188004,
-      "epoch": 1.074766355140187,
-      "grad_norm": 0.02599039301276207,
+      "entropy": 0.6355349719524384,
+      "epoch": 0.37631792376317924,
+      "grad_norm": 0.01422160305082798,
       "learning_rate": 0.0002,
-      "loss": 0.59382164478302,
-      "mean_token_accuracy": 0.7613071352243423,
-      "num_tokens": 4542209.0,
+      "loss": 0.6343538165092468,
+      "mean_token_accuracy": 0.743286557495594,
+      "num_tokens": 2161345.0,
       "step": 58
     },
     {
-      "entropy": 0.5987758338451385,
-      "epoch": 1.0934579439252337,
-      "grad_norm": 0.026119234040379524,
+      "entropy": 0.6376762092113495,
+      "epoch": 0.38280616382806165,
+      "grad_norm": 0.014769963920116425,
       "learning_rate": 0.0002,
-      "loss": 0.5977290868759155,
-      "mean_token_accuracy": 0.7600577920675278,
-      "num_tokens": 4621289.0,
+      "loss": 0.6412167549133301,
+      "mean_token_accuracy": 0.7418163046240807,
+      "num_tokens": 2198636.0,
       "step": 59
     },
     {
-      "entropy": 0.6043910086154938,
-      "epoch": 1.1121495327102804,
-      "grad_norm": 0.027874266728758812,
+      "entropy": 0.6334660202264786,
+      "epoch": 0.38929440389294406,
+      "grad_norm": 0.014948038384318352,
       "learning_rate": 0.0002,
-      "loss": 0.5998106598854065,
-      "mean_token_accuracy": 0.7593644857406616,
-      "num_tokens": 4702264.0,
+      "loss": 0.6379479169845581,
+      "mean_token_accuracy": 0.7443428039550781,
+      "num_tokens": 2235377.0,
       "step": 60
     },
     {
-      "entropy": 0.6029744148254395,
-      "epoch": 1.1308411214953271,
-      "grad_norm": 0.024287983775138855,
+      "entropy": 0.632827527821064,
+      "epoch": 0.3957826439578264,
+      "grad_norm": 0.01332375779747963,
       "learning_rate": 0.0002,
-      "loss": 0.601834237575531,
-      "mean_token_accuracy": 0.759440615773201,
-      "num_tokens": 4781921.0,
+      "loss": 0.633344292640686,
+      "mean_token_accuracy": 0.7467714250087738,
+      "num_tokens": 2272588.0,
       "step": 61
     },
     {
-      "entropy": 0.6114215701818466,
-      "epoch": 1.1495327102803738,
-      "grad_norm": 0.029389025643467903,
+      "entropy": 0.632963553071022,
+      "epoch": 0.40227088402270883,
+      "grad_norm": 0.013366672210395336,
       "learning_rate": 0.0002,
-      "loss": 0.6087201237678528,
-      "mean_token_accuracy": 0.7564449906349182,
-      "num_tokens": 4861920.0,
+      "loss": 0.6344192028045654,
+      "mean_token_accuracy": 0.7444779649376869,
+      "num_tokens": 2310205.0,
       "step": 62
     },
     {
-      "entropy": 0.6049663573503494,
-      "epoch": 1.1682242990654206,
-      "grad_norm": 0.02526937425136566,
+      "entropy": 0.6468651965260506,
+      "epoch": 0.40875912408759124,
+      "grad_norm": 0.012876944616436958,
       "learning_rate": 0.0002,
-      "loss": 0.6054017543792725,
-      "mean_token_accuracy": 0.7583751827478409,
-      "num_tokens": 4940914.0,
+      "loss": 0.6487270593643188,
+      "mean_token_accuracy": 0.7404673770070076,
+      "num_tokens": 2347612.0,
       "step": 63
     },
     {
-      "entropy": 0.605153813958168,
-      "epoch": 1.1869158878504673,
-      "grad_norm": 0.023482007905840874,
+      "entropy": 0.6336022540926933,
+      "epoch": 0.41524736415247365,
+      "grad_norm": 0.011953601613640785,
       "learning_rate": 0.0002,
-      "loss": 0.6047523021697998,
-      "mean_token_accuracy": 0.7565990537405014,
-      "num_tokens": 5020328.0,
+      "loss": 0.6323578357696533,
+      "mean_token_accuracy": 0.7435265332460403,
+      "num_tokens": 2385324.0,
       "step": 64
     },
     {
-      "entropy": 0.6064505577087402,
-      "epoch": 1.205607476635514,
-      "grad_norm": 0.028204405680298805,
+      "entropy": 0.6421113535761833,
+      "epoch": 0.42173560421735606,
+      "grad_norm": 0.011578400619328022,
       "learning_rate": 0.0002,
-      "loss": 0.602495551109314,
-      "mean_token_accuracy": 0.758506640791893,
-      "num_tokens": 5099429.0,
+      "loss": 0.640021800994873,
+      "mean_token_accuracy": 0.7417890876531601,
+      "num_tokens": 2422386.0,
       "step": 65
     },
     {
-      "entropy": 0.6176208853721619,
-      "epoch": 1.2242990654205608,
-      "grad_norm": 0.0220986045897007,
+      "entropy": 0.6243318468332291,
+      "epoch": 0.4282238442822384,
+      "grad_norm": 0.01271536760032177,
       "learning_rate": 0.0002,
-      "loss": 0.6151657104492188,
-      "mean_token_accuracy": 0.7535989731550217,
-      "num_tokens": 5176684.0,
+      "loss": 0.6278913617134094,
+      "mean_token_accuracy": 0.7464011386036873,
+      "num_tokens": 2459184.0,
       "step": 66
     },
     {
-      "entropy": 0.5883279144763947,
-      "epoch": 1.2429906542056075,
-      "grad_norm": 0.02462647296488285,
+      "entropy": 0.6474096700549126,
+      "epoch": 0.43471208434712083,
+      "grad_norm": 0.011297891847789288,
       "learning_rate": 0.0002,
-      "loss": 0.5873193144798279,
-      "mean_token_accuracy": 0.7656737864017487,
-      "num_tokens": 5256058.0,
+      "loss": 0.649469256401062,
+      "mean_token_accuracy": 0.7372781112790108,
+      "num_tokens": 2496769.0,
       "step": 67
     },
     {
-      "entropy": 0.5964405983686447,
-      "epoch": 1.2616822429906542,
-      "grad_norm": 0.027081497013568878,
+      "entropy": 0.629207618534565,
+      "epoch": 0.44120032441200324,
+      "grad_norm": 0.01338763814419508,
       "learning_rate": 0.0002,
-      "loss": 0.5986899137496948,
-      "mean_token_accuracy": 0.7598204612731934,
-      "num_tokens": 5335338.0,
+      "loss": 0.6260238885879517,
+      "mean_token_accuracy": 0.7460990995168686,
+      "num_tokens": 2533774.0,
       "step": 68
     },
     {
-      "entropy": 0.595260500907898,
-      "epoch": 1.280373831775701,
-      "grad_norm": 0.02198946289718151,
+      "entropy": 0.6363127902150154,
+      "epoch": 0.44768856447688565,
+      "grad_norm": 0.01232822984457016,
       "learning_rate": 0.0002,
-      "loss": 0.5952171087265015,
-      "mean_token_accuracy": 0.7604967951774597,
-      "num_tokens": 5413582.0,
+      "loss": 0.6326106786727905,
+      "mean_token_accuracy": 0.7443196326494217,
+      "num_tokens": 2570775.0,
       "step": 69
     },
     {
-      "entropy": 0.5949370414018631,
-      "epoch": 1.2990654205607477,
-      "grad_norm": 0.027311844751238823,
+      "entropy": 0.632358968257904,
+      "epoch": 0.45417680454176806,
+      "grad_norm": 0.012171111069619656,
       "learning_rate": 0.0002,
-      "loss": 0.5914739370346069,
-      "mean_token_accuracy": 0.763220340013504,
-      "num_tokens": 5492704.0,
+      "loss": 0.6312543749809265,
+      "mean_token_accuracy": 0.745995283126831,
+      "num_tokens": 2608121.0,
       "step": 70
     },
     {
-      "entropy": 0.6006052494049072,
-      "epoch": 1.3177570093457944,
-      "grad_norm": 0.021869085729122162,
+      "entropy": 0.6332346796989441,
+      "epoch": 0.4606650446066504,
+      "grad_norm": 0.015379011631011963,
       "learning_rate": 0.0002,
-      "loss": 0.5958817601203918,
-      "mean_token_accuracy": 0.759731113910675,
-      "num_tokens": 5571133.0,
+      "loss": 0.6291565299034119,
+      "mean_token_accuracy": 0.7465488314628601,
+      "num_tokens": 2645664.0,
       "step": 71
     },
     {
-      "entropy": 0.6128104329109192,
-      "epoch": 1.3364485981308412,
-      "grad_norm": 0.026876263320446014,
+      "entropy": 0.6404105871915817,
+      "epoch": 0.46715328467153283,
+      "grad_norm": 0.011988981626927853,
       "learning_rate": 0.0002,
-      "loss": 0.612541913986206,
-      "mean_token_accuracy": 0.7541105449199677,
-      "num_tokens": 5648947.0,
+      "loss": 0.642942488193512,
+      "mean_token_accuracy": 0.7407120391726494,
+      "num_tokens": 2683394.0,
       "step": 72
     },
     {
-      "entropy": 0.6033170968294144,
-      "epoch": 1.355140186915888,
-      "grad_norm": 0.026047907769680023,
+      "entropy": 0.6460321098566055,
+      "epoch": 0.47364152473641524,
+      "grad_norm": 0.013498514890670776,
       "learning_rate": 0.0002,
-      "loss": 0.6022995114326477,
-      "mean_token_accuracy": 0.7580565810203552,
-      "num_tokens": 5727865.0,
+      "loss": 0.6516697406768799,
+      "mean_token_accuracy": 0.7381351664662361,
+      "num_tokens": 2720891.0,
       "step": 73
     },
     {
-      "entropy": 0.597144678235054,
-      "epoch": 1.3738317757009346,
-      "grad_norm": 0.02528836391866207,
+      "entropy": 0.6352040842175484,
+      "epoch": 0.48012976480129765,
+      "grad_norm": 0.013542759232223034,
       "learning_rate": 0.0002,
-      "loss": 0.5937597751617432,
-      "mean_token_accuracy": 0.761174738407135,
-      "num_tokens": 5806694.0,
+      "loss": 0.6443178653717041,
+      "mean_token_accuracy": 0.7414710223674774,
+      "num_tokens": 2758174.0,
       "step": 74
     },
     {
-      "entropy": 0.601003110408783,
-      "epoch": 1.3925233644859814,
-      "grad_norm": 0.024352161213755608,
+      "entropy": 0.6283697411417961,
+      "epoch": 0.48661800486618007,
+      "grad_norm": 0.014966949820518494,
       "learning_rate": 0.0002,
-      "loss": 0.5970538258552551,
-      "mean_token_accuracy": 0.7603916376829147,
-      "num_tokens": 5886675.0,
+      "loss": 0.6394721269607544,
+      "mean_token_accuracy": 0.7419639229774475,
+      "num_tokens": 2795106.0,
       "step": 75
     },
     {
-      "entropy": 0.6055264323949814,
-      "epoch": 1.411214953271028,
-      "grad_norm": 0.02295341156423092,
+      "entropy": 0.642236091196537,
+      "epoch": 0.4931062449310625,
+      "grad_norm": 0.015388374216854572,
       "learning_rate": 0.0002,
-      "loss": 0.6047403812408447,
-      "mean_token_accuracy": 0.7562543004751205,
-      "num_tokens": 5966272.0,
+      "loss": 0.6414813995361328,
+      "mean_token_accuracy": 0.7409433200955391,
+      "num_tokens": 2832247.0,
       "step": 76
     },
     {
-      "entropy": 0.6091946959495544,
-      "epoch": 1.4299065420560748,
-      "grad_norm": 0.027446286752820015,
+      "entropy": 0.6422772705554962,
+      "epoch": 0.49959448499594483,
+      "grad_norm": 0.013754413463175297,
       "learning_rate": 0.0002,
-      "loss": 0.6084959506988525,
-      "mean_token_accuracy": 0.7555825561285019,
-      "num_tokens": 6044286.0,
+      "loss": 0.6367785930633545,
+      "mean_token_accuracy": 0.7442163750529289,
+      "num_tokens": 2869526.0,
       "step": 77
     },
     {
-      "entropy": 0.6051345616579056,
-      "epoch": 1.4485981308411215,
-      "grad_norm": 0.02242988348007202,
+      "entropy": 0.653985932469368,
+      "epoch": 0.5060827250608273,
+      "grad_norm": 0.012936330400407314,
       "learning_rate": 0.0002,
-      "loss": 0.6049442291259766,
-      "mean_token_accuracy": 0.7556533664464951,
-      "num_tokens": 6123208.0,
+      "loss": 0.6488639712333679,
+      "mean_token_accuracy": 0.7398064211010933,
+      "num_tokens": 2906821.0,
       "step": 78
     },
     {
-      "entropy": 0.6011074483394623,
-      "epoch": 1.4672897196261683,
-      "grad_norm": 0.023782888427376747,
+      "entropy": 0.6418487578630447,
+      "epoch": 0.5125709651257097,
+      "grad_norm": 0.011183886788785458,
       "learning_rate": 0.0002,
-      "loss": 0.596901535987854,
-      "mean_token_accuracy": 0.7608011811971664,
-      "num_tokens": 6202897.0,
+      "loss": 0.639202356338501,
+      "mean_token_accuracy": 0.7434348165988922,
+      "num_tokens": 2944032.0,
       "step": 79
     },
     {
-      "entropy": 0.5916998535394669,
-      "epoch": 1.485981308411215,
-      "grad_norm": 0.022254550829529762,
+      "entropy": 0.6251209825277328,
+      "epoch": 0.519059205190592,
+      "grad_norm": 0.013851741328835487,
       "learning_rate": 0.0002,
-      "loss": 0.5927384495735168,
-      "mean_token_accuracy": 0.7623648941516876,
-      "num_tokens": 6281706.0,
+      "loss": 0.6242102384567261,
+      "mean_token_accuracy": 0.7505335137248039,
+      "num_tokens": 2981414.0,
       "step": 80
     },
     {
-      "entropy": 0.6035054922103882,
-      "epoch": 1.5046728971962615,
-      "grad_norm": 0.02131548523902893,
+      "entropy": 0.6139597967267036,
+      "epoch": 0.5255474452554745,
+      "grad_norm": 0.013575786724686623,
       "learning_rate": 0.0002,
-      "loss": 0.6067797541618347,
-      "mean_token_accuracy": 0.755115196108818,
-      "num_tokens": 6361239.0,
+      "loss": 0.6168844699859619,
+      "mean_token_accuracy": 0.7502986714243889,
+      "num_tokens": 3018494.0,
       "step": 81
     },
     {
-      "entropy": 0.6077107489109039,
-      "epoch": 1.5233644859813085,
-      "grad_norm": 0.025049913674592972,
+      "entropy": 0.6357163935899734,
+      "epoch": 0.5320356853203568,
+      "grad_norm": 0.012106247246265411,
       "learning_rate": 0.0002,
-      "loss": 0.6130092144012451,
-      "mean_token_accuracy": 0.7551602274179459,
-      "num_tokens": 6439682.0,
+      "loss": 0.6393176317214966,
+      "mean_token_accuracy": 0.743524856865406,
+      "num_tokens": 3055667.0,
       "step": 82
     },
     {
-      "entropy": 0.602842390537262,
-      "epoch": 1.542056074766355,
-      "grad_norm": 0.02274921163916588,
+      "entropy": 0.630703866481781,
+      "epoch": 0.5385239253852393,
+      "grad_norm": 0.011411616578698158,
       "learning_rate": 0.0002,
-      "loss": 0.6033390164375305,
-      "mean_token_accuracy": 0.7566068917512894,
-      "num_tokens": 6517795.0,
+      "loss": 0.6349452137947083,
+      "mean_token_accuracy": 0.7417918294668198,
+      "num_tokens": 3092980.0,
       "step": 83
     },
     {
-      "entropy": 0.5934896320104599,
-      "epoch": 1.560747663551402,
-      "grad_norm": 0.019730787724256516,
+      "entropy": 0.6312713250517845,
+      "epoch": 0.5450121654501217,
+      "grad_norm": 0.013269394636154175,
       "learning_rate": 0.0002,
-      "loss": 0.5883506536483765,
-      "mean_token_accuracy": 0.763173297047615,
-      "num_tokens": 6597109.0,
+      "loss": 0.6335923075675964,
+      "mean_token_accuracy": 0.7427873983979225,
+      "num_tokens": 3129991.0,
       "step": 84
     },
     {
-      "entropy": 0.5969595015048981,
-      "epoch": 1.5794392523364484,
-      "grad_norm": 0.023908721283078194,
+      "entropy": 0.6429375261068344,
+      "epoch": 0.551500405515004,
+      "grad_norm": 0.01290512178093195,
       "learning_rate": 0.0002,
-      "loss": 0.5954324007034302,
-      "mean_token_accuracy": 0.7612140029668808,
-      "num_tokens": 6676983.0,
+      "loss": 0.6390902996063232,
+      "mean_token_accuracy": 0.7440147623419762,
+      "num_tokens": 3167600.0,
       "step": 85
     },
     {
-      "entropy": 0.583121731877327,
-      "epoch": 1.5981308411214954,
-      "grad_norm": 0.020361311733722687,
+      "entropy": 0.6209964007139206,
+      "epoch": 0.5579886455798865,
+      "grad_norm": 0.011449260637164116,
       "learning_rate": 0.0002,
-      "loss": 0.5792651176452637,
-      "mean_token_accuracy": 0.7673831284046173,
-      "num_tokens": 6756180.0,
+      "loss": 0.6205590963363647,
+      "mean_token_accuracy": 0.7478290274739265,
+      "num_tokens": 3204616.0,
       "step": 86
     },
     {
-      "entropy": 0.5962767004966736,
-      "epoch": 1.616822429906542,
-      "grad_norm": 0.021907050162553787,
+      "entropy": 0.6387689039111137,
+      "epoch": 0.5644768856447688,
+      "grad_norm": 0.012288929894566536,
       "learning_rate": 0.0002,
-      "loss": 0.5951491594314575,
-      "mean_token_accuracy": 0.7604141533374786,
-      "num_tokens": 6835203.0,
+      "loss": 0.6413038372993469,
+      "mean_token_accuracy": 0.7402090132236481,
+      "num_tokens": 3241598.0,
       "step": 87
     },
     {
-      "entropy": 0.5975628942251205,
-      "epoch": 1.6355140186915889,
-      "grad_norm": 0.025554567575454712,
+      "entropy": 0.6296097934246063,
+      "epoch": 0.5709651257096513,
+      "grad_norm": 0.012685527093708515,
       "learning_rate": 0.0002,
-      "loss": 0.5959036350250244,
-      "mean_token_accuracy": 0.7604859173297882,
-      "num_tokens": 6914015.0,
+      "loss": 0.6300569772720337,
+      "mean_token_accuracy": 0.745740607380867,
+      "num_tokens": 3278904.0,
       "step": 88
     },
     {
-      "entropy": 0.5989606827497482,
-      "epoch": 1.6542056074766354,
-      "grad_norm": 0.01990366540849209,
+      "entropy": 0.6304027885198593,
+      "epoch": 0.5774533657745337,
+      "grad_norm": 0.012135171331465244,
       "learning_rate": 0.0002,
-      "loss": 0.5967363119125366,
-      "mean_token_accuracy": 0.7600581496953964,
-      "num_tokens": 6993444.0,
+      "loss": 0.6322764158248901,
+      "mean_token_accuracy": 0.7428491413593292,
+      "num_tokens": 3315916.0,
       "step": 89
     },
     {
-      "entropy": 0.5878342539072037,
-      "epoch": 1.6728971962616823,
-      "grad_norm": 0.0246729776263237,
+      "entropy": 0.6270208656787872,
+      "epoch": 0.583941605839416,
+      "grad_norm": 0.012634403072297573,
       "learning_rate": 0.0002,
-      "loss": 0.5855717062950134,
-      "mean_token_accuracy": 0.7647435963153839,
-      "num_tokens": 7073123.0,
+      "loss": 0.6289095878601074,
+      "mean_token_accuracy": 0.74432173371315,
+      "num_tokens": 3352975.0,
       "step": 90
     },
     {
-      "entropy": 0.5907899141311646,
-      "epoch": 1.6915887850467288,
-      "grad_norm": 0.01860217936336994,
+      "entropy": 0.6426985636353493,
+      "epoch": 0.5904298459042985,
+      "grad_norm": 0.012899120338261127,
       "learning_rate": 0.0002,
-      "loss": 0.5894824266433716,
-      "mean_token_accuracy": 0.7620080858469009,
-      "num_tokens": 7152412.0,
+      "loss": 0.6412180662155151,
+      "mean_token_accuracy": 0.7413002401590347,
+      "num_tokens": 3390460.0,
       "step": 91
     },
     {
-      "entropy": 0.5949963480234146,
-      "epoch": 1.7102803738317758,
-      "grad_norm": 0.02079867199063301,
+      "entropy": 0.6386194601655006,
+      "epoch": 0.5969180859691808,
+      "grad_norm": 0.011871455237269402,
       "learning_rate": 0.0002,
-      "loss": 0.5942832231521606,
-      "mean_token_accuracy": 0.7613794505596161,
-      "num_tokens": 7231659.0,
+      "loss": 0.6381049156188965,
+      "mean_token_accuracy": 0.7410541325807571,
+      "num_tokens": 3427650.0,
       "step": 92
     },
     {
-      "entropy": 0.603457123041153,
-      "epoch": 1.7289719626168223,
-      "grad_norm": 0.021803148090839386,
+      "entropy": 0.6219398155808449,
+      "epoch": 0.6034063260340633,
+      "grad_norm": 0.013389634899795055,
       "learning_rate": 0.0002,
-      "loss": 0.6034700274467468,
-      "mean_token_accuracy": 0.7573578655719757,
-      "num_tokens": 7308587.0,
+      "loss": 0.628428041934967,
+      "mean_token_accuracy": 0.7469312027096748,
+      "num_tokens": 3464664.0,
       "step": 93
     },
     {
-      "entropy": 0.6119105815887451,
-      "epoch": 1.7476635514018692,
-      "grad_norm": 0.025299299508333206,
+      "entropy": 0.6363367736339569,
+      "epoch": 0.6098945660989457,
+      "grad_norm": 0.010384561493992805,
       "learning_rate": 0.0002,
-      "loss": 0.6091461777687073,
-      "mean_token_accuracy": 0.7555771619081497,
-      "num_tokens": 7387591.0,
+      "loss": 0.6355550289154053,
+      "mean_token_accuracy": 0.742337591946125,
+      "num_tokens": 3502010.0,
       "step": 94
     },
     {
-      "entropy": 0.6056051701307297,
-      "epoch": 1.7663551401869158,
-      "grad_norm": 0.021394768729805946,
+      "entropy": 0.6357248276472092,
+      "epoch": 0.616382806163828,
+      "grad_norm": 0.010791895911097527,
       "learning_rate": 0.0002,
-      "loss": 0.5999703407287598,
-      "mean_token_accuracy": 0.7588638365268707,
-      "num_tokens": 7466317.0,
+      "loss": 0.6378734111785889,
+      "mean_token_accuracy": 0.7428740784525871,
+      "num_tokens": 3539472.0,
       "step": 95
     },
     {
-      "entropy": 0.5954627841711044,
-      "epoch": 1.7850467289719627,
-      "grad_norm": 0.02461361326277256,
+      "entropy": 0.6360966116189957,
+      "epoch": 0.6228710462287105,
+      "grad_norm": 0.010037368163466454,
       "learning_rate": 0.0002,
-      "loss": 0.5979853272438049,
-      "mean_token_accuracy": 0.75945945084095,
-      "num_tokens": 7546769.0,
+      "loss": 0.6382291316986084,
+      "mean_token_accuracy": 0.7439633831381798,
+      "num_tokens": 3576863.0,
       "step": 96
     },
     {
-      "entropy": 0.5966174304485321,
-      "epoch": 1.8037383177570092,
-      "grad_norm": 0.025569939985871315,
+      "entropy": 0.6299866884946823,
+      "epoch": 0.6293592862935928,
+      "grad_norm": 0.011554226279258728,
       "learning_rate": 0.0002,
-      "loss": 0.6006815433502197,
-      "mean_token_accuracy": 0.7578661143779755,
-      "num_tokens": 7624681.0,
+      "loss": 0.6333507895469666,
+      "mean_token_accuracy": 0.7430082336068153,
+      "num_tokens": 3613859.0,
       "step": 97
     },
     {
-      "entropy": 0.594436764717102,
-      "epoch": 1.8224299065420562,
-      "grad_norm": 0.020345669239759445,
+      "entropy": 0.6372612193226814,
+      "epoch": 0.6358475263584753,
+      "grad_norm": 0.012070685625076294,
       "learning_rate": 0.0002,
-      "loss": 0.597395122051239,
-      "mean_token_accuracy": 0.7583171874284744,
-      "num_tokens": 7703257.0,
+      "loss": 0.6347473859786987,
+      "mean_token_accuracy": 0.7440797835588455,
+      "num_tokens": 3651376.0,
       "step": 98
     },
     {
-      "entropy": 0.6004362404346466,
-      "epoch": 1.8411214953271027,
-      "grad_norm": 0.026212656870484352,
+      "entropy": 0.6408236473798752,
+      "epoch": 0.6423357664233577,
+      "grad_norm": 0.010936851613223553,
       "learning_rate": 0.0002,
-      "loss": 0.5976188778877258,
-      "mean_token_accuracy": 0.7598545700311661,
-      "num_tokens": 7781455.0,
+      "loss": 0.6386737823486328,
+      "mean_token_accuracy": 0.7431015446782112,
+      "num_tokens": 3688847.0,
       "step": 99
     },
     {
-      "entropy": 0.6042353212833405,
-      "epoch": 1.8598130841121496,
-      "grad_norm": 0.02437760680913925,
+      "entropy": 0.6442330852150917,
+      "epoch": 0.64882400648824,
+      "grad_norm": 0.01024035457521677,
       "learning_rate": 0.0002,
-      "loss": 0.6001909971237183,
-      "mean_token_accuracy": 0.7570216208696365,
-      "num_tokens": 7860292.0,
+      "loss": 0.6463708877563477,
+      "mean_token_accuracy": 0.7389701902866364,
+      "num_tokens": 3726347.0,
       "step": 100
     },
     {
-      "entropy": 0.6039856970310211,
-      "epoch": 1.8785046728971961,
-      "grad_norm": 0.02607814036309719,
+      "entropy": 0.6199506223201752,
+      "epoch": 0.6553122465531225,
+      "grad_norm": 0.01070485170930624,
       "learning_rate": 0.0002,
-      "loss": 0.5982059240341187,
-      "mean_token_accuracy": 0.7589994370937347,
-      "num_tokens": 7939144.0,
+      "loss": 0.6225665807723999,
+      "mean_token_accuracy": 0.7488372325897217,
+      "num_tokens": 3763687.0,
       "step": 101
     },
     {
-      "entropy": 0.595164492726326,
-      "epoch": 1.897196261682243,
-      "grad_norm": 0.020943354815244675,
+      "entropy": 0.6321059763431549,
+      "epoch": 0.6618004866180048,
+      "grad_norm": 0.010497628711163998,
       "learning_rate": 0.0002,
-      "loss": 0.593543291091919,
-      "mean_token_accuracy": 0.7612017840147018,
-      "num_tokens": 8018850.0,
+      "loss": 0.6386170387268066,
+      "mean_token_accuracy": 0.7398883476853371,
+      "num_tokens": 3801390.0,
       "step": 102
     },
     {
-      "entropy": 0.5922124683856964,
-      "epoch": 1.9158878504672896,
-      "grad_norm": 0.021710926666855812,
+      "entropy": 0.6317892819643021,
+      "epoch": 0.6682887266828873,
+      "grad_norm": 0.009558591060340405,
       "learning_rate": 0.0002,
-      "loss": 0.5924748182296753,
-      "mean_token_accuracy": 0.7613015025854111,
-      "num_tokens": 8097651.0,
+      "loss": 0.6320483684539795,
+      "mean_token_accuracy": 0.744752362370491,
+      "num_tokens": 3838836.0,
       "step": 103
     },
     {
-      "entropy": 0.5902406722307205,
-      "epoch": 1.9345794392523366,
-      "grad_norm": 0.02682112529873848,
+      "entropy": 0.6251348331570625,
+      "epoch": 0.6747769667477697,
+      "grad_norm": 0.012168657965958118,
       "learning_rate": 0.0002,
-      "loss": 0.5906952023506165,
-      "mean_token_accuracy": 0.7610596120357513,
-      "num_tokens": 8175432.0,
+      "loss": 0.624641478061676,
+      "mean_token_accuracy": 0.7455502301454544,
+      "num_tokens": 3876059.0,
       "step": 104
     },
     {
-      "entropy": 0.5985020995140076,
-      "epoch": 1.953271028037383,
-      "grad_norm": 0.021625829860568047,
+      "entropy": 0.6264195889234543,
+      "epoch": 0.681265206812652,
+      "grad_norm": 0.01135180238634348,
       "learning_rate": 0.0002,
-      "loss": 0.5948777198791504,
-      "mean_token_accuracy": 0.7605670243501663,
-      "num_tokens": 8254517.0,
+      "loss": 0.6254705190658569,
+      "mean_token_accuracy": 0.7437941655516624,
+      "num_tokens": 3913140.0,
       "step": 105
     },
     {
-      "entropy": 0.592091292142868,
-      "epoch": 1.97196261682243,
-      "grad_norm": 0.026733042672276497,
+      "entropy": 0.6271661967039108,
+      "epoch": 0.6877534468775345,
+      "grad_norm": 0.011404785327613354,
       "learning_rate": 0.0002,
-      "loss": 0.5928829908370972,
-      "mean_token_accuracy": 0.7605357766151428,
-      "num_tokens": 8332655.0,
+      "loss": 0.627234160900116,
+      "mean_token_accuracy": 0.7443915233016014,
+      "num_tokens": 3950197.0,
       "step": 106
     },
     {
-      "entropy": 0.602226734161377,
-      "epoch": 1.9906542056074765,
-      "grad_norm": 0.024320000782608986,
+      "entropy": 0.6316048800945282,
+      "epoch": 0.6942416869424168,
+      "grad_norm": 0.012731475755572319,
       "learning_rate": 0.0002,
-      "loss": 0.6039963960647583,
-      "mean_token_accuracy": 0.7566654533147812,
-      "num_tokens": 8411757.0,
+      "loss": 0.631871223449707,
+      "mean_token_accuracy": 0.7436823844909668,
+      "num_tokens": 3987880.0,
       "step": 107
     },
     {
-      "entropy": 0.5967253148555756,
-      "epoch": 2.0,
-      "grad_norm": 0.024751760065555573,
+      "entropy": 0.6351677849888802,
+      "epoch": 0.7007299270072993,
+      "grad_norm": 0.011057685129344463,
       "learning_rate": 0.0002,
-      "loss": 0.5935255885124207,
-      "mean_token_accuracy": 0.7621923387050629,
-      "num_tokens": 8450525.0,
+      "loss": 0.634377121925354,
+      "mean_token_accuracy": 0.7431640774011612,
+      "num_tokens": 4025490.0,
       "step": 108
     },
     {
-      "entropy": 0.5986887663602829,
-      "epoch": 2.0186915887850465,
-      "grad_norm": 0.023288823664188385,
+      "entropy": 0.6258664429187775,
+      "epoch": 0.7072181670721817,
+      "grad_norm": 0.012130603194236755,
       "learning_rate": 0.0002,
-      "loss": 0.5900849103927612,
-      "mean_token_accuracy": 0.7628316581249237,
-      "num_tokens": 8529518.0,
+      "loss": 0.628970742225647,
+      "mean_token_accuracy": 0.7431703954935074,
+      "num_tokens": 4062925.0,
       "step": 109
     },
     {
-      "entropy": 0.6044819355010986,
-      "epoch": 2.0373831775700935,
-      "grad_norm": 0.028756342828273773,
+      "entropy": 0.6371812373399734,
+      "epoch": 0.7137064071370641,
+      "grad_norm": 0.01057880837470293,
       "learning_rate": 0.0002,
-      "loss": 0.5983761548995972,
-      "mean_token_accuracy": 0.7610821276903152,
-      "num_tokens": 8608171.0,
+      "loss": 0.6419240236282349,
+      "mean_token_accuracy": 0.7392893806099892,
+      "num_tokens": 4100176.0,
       "step": 110
     },
     {
-      "entropy": 0.5922483056783676,
-      "epoch": 2.05607476635514,
-      "grad_norm": 0.0210813470184803,
+      "entropy": 0.6297428086400032,
+      "epoch": 0.7201946472019465,
+      "grad_norm": 0.01054214034229517,
       "learning_rate": 0.0002,
-      "loss": 0.5954931974411011,
-      "mean_token_accuracy": 0.7605789005756378,
-      "num_tokens": 8687290.0,
+      "loss": 0.6334270238876343,
+      "mean_token_accuracy": 0.7451630085706711,
+      "num_tokens": 4137649.0,
       "step": 111
     },
     {
-      "entropy": 0.5983303636312485,
-      "epoch": 2.074766355140187,
-      "grad_norm": 0.028979478403925896,
+      "entropy": 0.6174112483859062,
+      "epoch": 0.7266828872668288,
+      "grad_norm": 0.01192370243370533,
       "learning_rate": 0.0002,
-      "loss": 0.603399395942688,
-      "mean_token_accuracy": 0.7568083107471466,
-      "num_tokens": 8766472.0,
+      "loss": 0.6196002960205078,
+      "mean_token_accuracy": 0.748643770813942,
+      "num_tokens": 4174992.0,
       "step": 112
     },
     {
-      "entropy": 0.5887072533369064,
-      "epoch": 2.0934579439252334,
-      "grad_norm": 0.026534179225564003,
+      "entropy": 0.6409988701343536,
+      "epoch": 0.7331711273317113,
+      "grad_norm": 0.012110228650271893,
       "learning_rate": 0.0002,
-      "loss": 0.5949929356575012,
-      "mean_token_accuracy": 0.7597325891256332,
-      "num_tokens": 8846053.0,
+      "loss": 0.6420451402664185,
+      "mean_token_accuracy": 0.7404038235545158,
+      "num_tokens": 4212392.0,
       "step": 113
     },
     {
-      "entropy": 0.5866035968065262,
-      "epoch": 2.1121495327102804,
-      "grad_norm": 0.022542569786310196,
+      "entropy": 0.6189580634236336,
+      "epoch": 0.7396593673965937,
+      "grad_norm": 0.010546376928687096,
       "learning_rate": 0.0002,
-      "loss": 0.5825520157814026,
-      "mean_token_accuracy": 0.7647050321102142,
-      "num_tokens": 8924770.0,
+      "loss": 0.6227763891220093,
+      "mean_token_accuracy": 0.7471738830208778,
+      "num_tokens": 4249387.0,
       "step": 114
     },
     {
-      "entropy": 0.6025423407554626,
-      "epoch": 2.130841121495327,
-      "grad_norm": 0.02788623608648777,
+      "entropy": 0.6477556005120277,
+      "epoch": 0.7461476074614761,
+      "grad_norm": 0.012936307117342949,
       "learning_rate": 0.0002,
-      "loss": 0.5966792106628418,
-      "mean_token_accuracy": 0.7595364600419998,
-      "num_tokens": 9003585.0,
+      "loss": 0.6444486379623413,
+      "mean_token_accuracy": 0.7407133355736732,
+      "num_tokens": 4286813.0,
       "step": 115
     },
     {
-      "entropy": 0.5885701924562454,
-      "epoch": 2.149532710280374,
-      "grad_norm": 0.02178538590669632,
+      "entropy": 0.6388062611222267,
+      "epoch": 0.7526358475263585,
+      "grad_norm": 0.011020737700164318,
       "learning_rate": 0.0002,
-      "loss": 0.58498215675354,
-      "mean_token_accuracy": 0.7650727927684784,
-      "num_tokens": 9082999.0,
+      "loss": 0.6358926892280579,
+      "mean_token_accuracy": 0.7421080470085144,
+      "num_tokens": 4324068.0,
       "step": 116
     },
     {
-      "entropy": 0.600003108382225,
-      "epoch": 2.1682242990654204,
-      "grad_norm": 0.026978863403201103,
+      "entropy": 0.623978279531002,
+      "epoch": 0.7591240875912408,
+      "grad_norm": 0.011272276751697063,
       "learning_rate": 0.0002,
-      "loss": 0.6022127866744995,
-      "mean_token_accuracy": 0.7581324279308319,
-      "num_tokens": 9161688.0,
+      "loss": 0.6239657402038574,
+      "mean_token_accuracy": 0.7491330280900002,
+      "num_tokens": 4361117.0,
       "step": 117
     },
     {
-      "entropy": 0.5930540412664413,
-      "epoch": 2.1869158878504673,
-      "grad_norm": 0.025554828345775604,
+      "entropy": 0.6343322321772575,
+      "epoch": 0.7656123276561233,
+      "grad_norm": 0.011995693668723106,
       "learning_rate": 0.0002,
-      "loss": 0.5951005816459656,
-      "mean_token_accuracy": 0.7604358941316605,
-      "num_tokens": 9240469.0,
+      "loss": 0.635836124420166,
+      "mean_token_accuracy": 0.7406070083379745,
+      "num_tokens": 4398800.0,
       "step": 118
     },
     {
-      "entropy": 0.5965612530708313,
-      "epoch": 2.205607476635514,
-      "grad_norm": 0.020386407151818275,
+      "entropy": 0.6236653178930283,
+      "epoch": 0.7721005677210057,
+      "grad_norm": 0.012621116824448109,
       "learning_rate": 0.0002,
-      "loss": 0.5925002098083496,
-      "mean_token_accuracy": 0.7609757632017136,
-      "num_tokens": 9319306.0,
+      "loss": 0.6293167471885681,
+      "mean_token_accuracy": 0.7453652620315552,
+      "num_tokens": 4435941.0,
       "step": 119
     },
     {
-      "entropy": 0.5974278450012207,
-      "epoch": 2.2242990654205608,
-      "grad_norm": 0.02594870701432228,
+      "entropy": 0.6156119927763939,
+      "epoch": 0.7785888077858881,
+      "grad_norm": 0.010849852114915848,
       "learning_rate": 0.0002,
-      "loss": 0.5941160917282104,
-      "mean_token_accuracy": 0.760003000497818,
-      "num_tokens": 9399792.0,
+      "loss": 0.619170069694519,
+      "mean_token_accuracy": 0.7481161803007126,
+      "num_tokens": 4473532.0,
       "step": 120
     },
     {
-      "entropy": 0.6072829514741898,
-      "epoch": 2.2429906542056073,
-      "grad_norm": 0.022801121696829796,
+      "entropy": 0.6325643733143806,
+      "epoch": 0.7850770478507705,
+      "grad_norm": 0.012033924460411072,
       "learning_rate": 0.0002,
-      "loss": 0.6069557070732117,
-      "mean_token_accuracy": 0.7553433626890182,
-      "num_tokens": 9477182.0,
+      "loss": 0.6320570707321167,
+      "mean_token_accuracy": 0.745756022632122,
+      "num_tokens": 4510663.0,
       "step": 121
     },
     {
-      "entropy": 0.5936979353427887,
-      "epoch": 2.2616822429906542,
-      "grad_norm": 0.019751207903027534,
+      "entropy": 0.6439174935221672,
+      "epoch": 0.7915652879156528,
+      "grad_norm": 0.011160428635776043,
       "learning_rate": 0.0002,
-      "loss": 0.5905026197433472,
-      "mean_token_accuracy": 0.7626521587371826,
-      "num_tokens": 9556043.0,
+      "loss": 0.6456828117370605,
+      "mean_token_accuracy": 0.7372104227542877,
+      "num_tokens": 4548366.0,
       "step": 122
     },
     {
-      "entropy": 0.5964563935995102,
-      "epoch": 2.2803738317757007,
-      "grad_norm": 0.021466070786118507,
+      "entropy": 0.614606499671936,
+      "epoch": 0.7980535279805353,
+      "grad_norm": 0.012264038436114788,
       "learning_rate": 0.0002,
-      "loss": 0.5965335369110107,
-      "mean_token_accuracy": 0.7616595923900604,
-      "num_tokens": 9636288.0,
+      "loss": 0.6181074976921082,
+      "mean_token_accuracy": 0.7488836497068405,
+      "num_tokens": 4585220.0,
       "step": 123
     },
     {
-      "entropy": 0.5857243090867996,
-      "epoch": 2.2990654205607477,
-      "grad_norm": 0.02454223297536373,
+      "entropy": 0.6238071173429489,
+      "epoch": 0.8045417680454177,
+      "grad_norm": 0.010948721319437027,
       "learning_rate": 0.0002,
-      "loss": 0.5892512798309326,
-      "mean_token_accuracy": 0.762162521481514,
-      "num_tokens": 9715788.0,
+      "loss": 0.6282123327255249,
+      "mean_token_accuracy": 0.7447341307997704,
+      "num_tokens": 4622594.0,
       "step": 124
     },
     {
-      "entropy": 0.5936813354492188,
-      "epoch": 2.317757009345794,
-      "grad_norm": 0.019078049808740616,
+      "entropy": 0.6207244768738747,
+      "epoch": 0.8110300081103001,
+      "grad_norm": 0.011111956089735031,
       "learning_rate": 0.0002,
-      "loss": 0.5922213792800903,
-      "mean_token_accuracy": 0.7600427120923996,
-      "num_tokens": 9793512.0,
+      "loss": 0.6224273443222046,
+      "mean_token_accuracy": 0.748368538916111,
+      "num_tokens": 4659890.0,
       "step": 125
     },
     {
-      "entropy": 0.5969529002904892,
-      "epoch": 2.336448598130841,
-      "grad_norm": 0.019933439791202545,
+      "entropy": 0.6334651708602905,
+      "epoch": 0.8175182481751825,
+      "grad_norm": 0.01036700326949358,
       "learning_rate": 0.0002,
-      "loss": 0.5948631763458252,
-      "mean_token_accuracy": 0.7603116780519485,
-      "num_tokens": 9871266.0,
+      "loss": 0.6354086399078369,
+      "mean_token_accuracy": 0.742103323340416,
+      "num_tokens": 4697214.0,
       "step": 126
     },
     {
-      "entropy": 0.6052633076906204,
-      "epoch": 2.3551401869158877,
-      "grad_norm": 0.02256905473768711,
+      "entropy": 0.6321412995457649,
+      "epoch": 0.8240064882400648,
+      "grad_norm": 0.011659913696348667,
       "learning_rate": 0.0002,
-      "loss": 0.6023162603378296,
-      "mean_token_accuracy": 0.7571097612380981,
-      "num_tokens": 9951361.0,
+      "loss": 0.632038950920105,
+      "mean_token_accuracy": 0.7430339977145195,
+      "num_tokens": 4734306.0,
       "step": 127
     },
     {
-      "entropy": 0.5992947816848755,
-      "epoch": 2.3738317757009346,
-      "grad_norm": 0.019124582409858704,
+      "entropy": 0.6252134963870049,
+      "epoch": 0.8304947283049473,
+      "grad_norm": 0.010766416788101196,
       "learning_rate": 0.0002,
-      "loss": 0.59776771068573,
-      "mean_token_accuracy": 0.7580797523260117,
-      "num_tokens": 10030704.0,
+      "loss": 0.6210082173347473,
+      "mean_token_accuracy": 0.747887596487999,
+      "num_tokens": 4772005.0,
       "step": 128
     },
     {
-      "entropy": 0.5913878232240677,
-      "epoch": 2.392523364485981,
-      "grad_norm": 0.020428184419870377,
+      "entropy": 0.632231742143631,
+      "epoch": 0.8369829683698297,
+      "grad_norm": 0.009685291908681393,
       "learning_rate": 0.0002,
-      "loss": 0.5932009220123291,
-      "mean_token_accuracy": 0.7593762278556824,
-      "num_tokens": 10109097.0,
+      "loss": 0.6338882446289062,
+      "mean_token_accuracy": 0.7436560764908791,
+      "num_tokens": 4809261.0,
       "step": 129
     },
     {
-      "entropy": 0.5881628692150116,
-      "epoch": 2.411214953271028,
-      "grad_norm": 0.019032983109354973,
+      "entropy": 0.6244007870554924,
+      "epoch": 0.8434712084347121,
+      "grad_norm": 0.011300327256321907,
       "learning_rate": 0.0002,
-      "loss": 0.5873616933822632,
-      "mean_token_accuracy": 0.7626058608293533,
-      "num_tokens": 10189819.0,
+      "loss": 0.6297156810760498,
+      "mean_token_accuracy": 0.7441424131393433,
+      "num_tokens": 4846773.0,
       "step": 130
     },
     {
-      "entropy": 0.5958539843559265,
-      "epoch": 2.4299065420560746,
-      "grad_norm": 0.022081410512328148,
+      "entropy": 0.6212883591651917,
+      "epoch": 0.8499594484995945,
+      "grad_norm": 0.011122049763798714,
       "learning_rate": 0.0002,
-      "loss": 0.5977240800857544,
-      "mean_token_accuracy": 0.7586192786693573,
-      "num_tokens": 10268190.0,
+      "loss": 0.6272852420806885,
+      "mean_token_accuracy": 0.745480865240097,
+      "num_tokens": 4884087.0,
       "step": 131
     },
     {
-      "entropy": 0.5923604220151901,
-      "epoch": 2.4485981308411215,
-      "grad_norm": 0.018967753276228905,
+      "entropy": 0.6275059729814529,
+      "epoch": 0.8564476885644768,
+      "grad_norm": 0.011759513057768345,
       "learning_rate": 0.0002,
-      "loss": 0.5907920598983765,
-      "mean_token_accuracy": 0.7627393752336502,
-      "num_tokens": 10345719.0,
+      "loss": 0.6352666020393372,
+      "mean_token_accuracy": 0.741997979581356,
+      "num_tokens": 4921034.0,
       "step": 132
     },
     {
-      "entropy": 0.5843045115470886,
-      "epoch": 2.467289719626168,
-      "grad_norm": 0.02031179890036583,
+      "entropy": 0.6292549446225166,
+      "epoch": 0.8629359286293593,
+      "grad_norm": 0.010103589855134487,
       "learning_rate": 0.0002,
-      "loss": 0.5829470157623291,
-      "mean_token_accuracy": 0.7647084146738052,
-      "num_tokens": 10424933.0,
+      "loss": 0.6285660862922668,
+      "mean_token_accuracy": 0.7458189874887466,
+      "num_tokens": 4958394.0,
       "step": 133
     },
     {
-      "entropy": 0.596602737903595,
-      "epoch": 2.485981308411215,
-      "grad_norm": 0.019796643406152725,
+      "entropy": 0.6254143863916397,
+      "epoch": 0.8694241686942417,
+      "grad_norm": 0.009165394119918346,
       "learning_rate": 0.0002,
-      "loss": 0.595852792263031,
-      "mean_token_accuracy": 0.7595040202140808,
-      "num_tokens": 10504010.0,
+      "loss": 0.6281845569610596,
+      "mean_token_accuracy": 0.7441290989518166,
+      "num_tokens": 4995560.0,
       "step": 134
     },
     {
-      "entropy": 0.5987847000360489,
-      "epoch": 2.5046728971962615,
-      "grad_norm": 0.01805124618113041,
+      "entropy": 0.6326478272676468,
+      "epoch": 0.8759124087591241,
+      "grad_norm": 0.009857391007244587,
       "learning_rate": 0.0002,
-      "loss": 0.5960158109664917,
-      "mean_token_accuracy": 0.7604788392782211,
-      "num_tokens": 10582414.0,
+      "loss": 0.6341642141342163,
+      "mean_token_accuracy": 0.7431726306676865,
+      "num_tokens": 5032908.0,
       "step": 135
     },
     {
-      "entropy": 0.5917297601699829,
-      "epoch": 2.5233644859813085,
-      "grad_norm": 0.021082131192088127,
+      "entropy": 0.6411951705813408,
+      "epoch": 0.8824006488240065,
+      "grad_norm": 0.009912555105984211,
       "learning_rate": 0.0002,
-      "loss": 0.5902372598648071,
-      "mean_token_accuracy": 0.7612239867448807,
-      "num_tokens": 10661151.0,
+      "loss": 0.6402624845504761,
+      "mean_token_accuracy": 0.7418247908353806,
+      "num_tokens": 5070353.0,
       "step": 136
     },
     {
-      "entropy": 0.5957841575145721,
-      "epoch": 2.542056074766355,
-      "grad_norm": 0.019563838839530945,
+      "entropy": 0.6335436180233955,
+      "epoch": 0.8888888888888888,
+      "grad_norm": 0.010203655809164047,
       "learning_rate": 0.0002,
-      "loss": 0.5957234501838684,
-      "mean_token_accuracy": 0.758367732167244,
-      "num_tokens": 10739185.0,
+      "loss": 0.6348065137863159,
+      "mean_token_accuracy": 0.7417963817715645,
+      "num_tokens": 5107488.0,
       "step": 137
     },
     {
-      "entropy": 0.6024006307125092,
-      "epoch": 2.560747663551402,
-      "grad_norm": 0.01933375746011734,
+      "entropy": 0.6298110708594322,
+      "epoch": 0.8953771289537713,
+      "grad_norm": 0.009738304652273655,
       "learning_rate": 0.0002,
-      "loss": 0.6043826937675476,
-      "mean_token_accuracy": 0.7564682364463806,
-      "num_tokens": 10817785.0,
+      "loss": 0.6292046904563904,
+      "mean_token_accuracy": 0.7428107112646103,
+      "num_tokens": 5144774.0,
       "step": 138
     },
     {
-      "entropy": 0.5953152179718018,
-      "epoch": 2.5794392523364484,
-      "grad_norm": 0.020373884588479996,
+      "entropy": 0.621233694255352,
+      "epoch": 0.9018653690186537,
+      "grad_norm": 0.009310460649430752,
       "learning_rate": 0.0002,
-      "loss": 0.5951474905014038,
-      "mean_token_accuracy": 0.759823739528656,
-      "num_tokens": 10895828.0,
+      "loss": 0.6167433261871338,
+      "mean_token_accuracy": 0.7501207813620567,
+      "num_tokens": 5181712.0,
       "step": 139
     },
     {
-      "entropy": 0.5937645733356476,
-      "epoch": 2.5981308411214954,
-      "grad_norm": 0.019647756591439247,
+      "entropy": 0.6066861748695374,
+      "epoch": 0.9083536090835361,
+      "grad_norm": 0.010622376576066017,
       "learning_rate": 0.0002,
-      "loss": 0.5909141302108765,
-      "mean_token_accuracy": 0.7605070471763611,
-      "num_tokens": 10974768.0,
+      "loss": 0.6090973615646362,
+      "mean_token_accuracy": 0.7524907812476158,
+      "num_tokens": 5219096.0,
       "step": 140
     },
     {
-      "entropy": 0.6062350273132324,
-      "epoch": 2.616822429906542,
-      "grad_norm": 0.02389589697122574,
+      "entropy": 0.6264965310692787,
+      "epoch": 0.9148418491484185,
+      "grad_norm": 0.011424071155488491,
       "learning_rate": 0.0002,
-      "loss": 0.6018738746643066,
-      "mean_token_accuracy": 0.7562691420316696,
-      "num_tokens": 11055326.0,
+      "loss": 0.6337799429893494,
+      "mean_token_accuracy": 0.7431259453296661,
+      "num_tokens": 5256221.0,
       "step": 141
     },
     {
-      "entropy": 0.5986295640468597,
-      "epoch": 2.635514018691589,
-      "grad_norm": 0.018676772713661194,
+      "entropy": 0.6391952782869339,
+      "epoch": 0.9213300892133008,
+      "grad_norm": 0.009158595465123653,
       "learning_rate": 0.0002,
-      "loss": 0.5976943969726562,
-      "mean_token_accuracy": 0.758969247341156,
-      "num_tokens": 11133311.0,
+      "loss": 0.6384302973747253,
+      "mean_token_accuracy": 0.7389715909957886,
+      "num_tokens": 5293696.0,
       "step": 142
     },
     {
-      "entropy": 0.5892178267240524,
-      "epoch": 2.6542056074766354,
-      "grad_norm": 0.01959327794611454,
+      "entropy": 0.6427849903702736,
+      "epoch": 0.9278183292781833,
+      "grad_norm": 0.011200455017387867,
       "learning_rate": 0.0002,
-      "loss": 0.5930218696594238,
-      "mean_token_accuracy": 0.7599967420101166,
-      "num_tokens": 11212924.0,
+      "loss": 0.6386270523071289,
+      "mean_token_accuracy": 0.7409881576895714,
+      "num_tokens": 5330767.0,
       "step": 143
     },
     {
-      "entropy": 0.5876894295215607,
-      "epoch": 2.6728971962616823,
-      "grad_norm": 0.02165973000228405,
+      "entropy": 0.6417384147644043,
+      "epoch": 0.9343065693430657,
+      "grad_norm": 0.010727403685450554,
       "learning_rate": 0.0002,
-      "loss": 0.5930266976356506,
-      "mean_token_accuracy": 0.7615959346294403,
-      "num_tokens": 11292441.0,
+      "loss": 0.63932865858078,
+      "mean_token_accuracy": 0.7433461174368858,
+      "num_tokens": 5368288.0,
       "step": 144
     },
     {
-      "entropy": 0.5921325981616974,
-      "epoch": 2.691588785046729,
-      "grad_norm": 0.019541792571544647,
+      "entropy": 0.6349349692463875,
+      "epoch": 0.9407948094079481,
+      "grad_norm": 0.01087568886578083,
       "learning_rate": 0.0002,
-      "loss": 0.5935483574867249,
-      "mean_token_accuracy": 0.7591264247894287,
-      "num_tokens": 11370065.0,
+      "loss": 0.6332679986953735,
+      "mean_token_accuracy": 0.7431394383311272,
+      "num_tokens": 5405743.0,
       "step": 145
     },
     {
-      "entropy": 0.5976132452487946,
-      "epoch": 2.710280373831776,
-      "grad_norm": 0.018777787685394287,
+      "entropy": 0.6268604248762131,
+      "epoch": 0.9472830494728305,
+      "grad_norm": 0.009905870072543621,
       "learning_rate": 0.0002,
-      "loss": 0.5930612087249756,
-      "mean_token_accuracy": 0.7606300711631775,
-      "num_tokens": 11448141.0,
+      "loss": 0.6278096437454224,
+      "mean_token_accuracy": 0.7460607141256332,
+      "num_tokens": 5442824.0,
       "step": 146
     },
     {
-      "entropy": 0.5865790247917175,
-      "epoch": 2.7289719626168223,
-      "grad_norm": 0.019476981833577156,
+      "entropy": 0.6252990365028381,
+      "epoch": 0.9537712895377128,
+      "grad_norm": 0.011337473057210445,
       "learning_rate": 0.0002,
-      "loss": 0.584563136100769,
-      "mean_token_accuracy": 0.7639805823564529,
-      "num_tokens": 11527975.0,
+      "loss": 0.6313953399658203,
+      "mean_token_accuracy": 0.7449029609560966,
+      "num_tokens": 5480060.0,
       "step": 147
     },
     {
-      "entropy": 0.601652979850769,
-      "epoch": 2.7476635514018692,
-      "grad_norm": 0.020754467695951462,
+      "entropy": 0.6279519572854042,
+      "epoch": 0.9602595296025953,
+      "grad_norm": 0.011046580970287323,
       "learning_rate": 0.0002,
-      "loss": 0.5971537828445435,
-      "mean_token_accuracy": 0.7591513991355896,
-      "num_tokens": 11606011.0,
+      "loss": 0.6361185312271118,
+      "mean_token_accuracy": 0.7435570135712624,
+      "num_tokens": 5517276.0,
       "step": 148
     },
     {
-      "entropy": 0.5934069901704788,
-      "epoch": 2.7663551401869158,
-      "grad_norm": 0.0185554176568985,
+      "entropy": 0.6244647055864334,
+      "epoch": 0.9667477696674777,
+      "grad_norm": 0.010344677604734898,
       "learning_rate": 0.0002,
-      "loss": 0.5921209454536438,
-      "mean_token_accuracy": 0.7602465599775314,
-      "num_tokens": 11685000.0,
+      "loss": 0.6286739110946655,
+      "mean_token_accuracy": 0.7459346577525139,
+      "num_tokens": 5554772.0,
       "step": 149
     },
     {
-      "entropy": 0.5966046750545502,
-      "epoch": 2.7850467289719627,
-      "grad_norm": 0.022011004388332367,
+      "entropy": 0.625552199780941,
+      "epoch": 0.9732360097323601,
+      "grad_norm": 0.010485424660146236,
       "learning_rate": 0.0002,
-      "loss": 0.6005374789237976,
-      "mean_token_accuracy": 0.7573995739221573,
-      "num_tokens": 11763461.0,
+      "loss": 0.6246562600135803,
+      "mean_token_accuracy": 0.7459065243601799,
+      "num_tokens": 5592235.0,
       "step": 150
     },
     {
-      "entropy": 0.5875833630561829,
-      "epoch": 2.803738317757009,
-      "grad_norm": 0.021797338500618935,
+      "entropy": 0.650402158498764,
+      "epoch": 0.9797242497972425,
+      "grad_norm": 0.01180343609303236,
       "learning_rate": 0.0002,
-      "loss": 0.5878199338912964,
-      "mean_token_accuracy": 0.7617918699979782,
-      "num_tokens": 11841433.0,
+      "loss": 0.6489013433456421,
+      "mean_token_accuracy": 0.7372798472642899,
+      "num_tokens": 5629879.0,
       "step": 151
     },
     {
-      "entropy": 0.5691852122545242,
-      "epoch": 2.822429906542056,
-      "grad_norm": 0.019778162240982056,
+      "entropy": 0.625130906701088,
+      "epoch": 0.986212489862125,
+      "grad_norm": 0.010784060694277287,
       "learning_rate": 0.0002,
-      "loss": 0.5664920210838318,
-      "mean_token_accuracy": 0.7711170166730881,
-      "num_tokens": 11923054.0,
+      "loss": 0.6217543482780457,
+      "mean_token_accuracy": 0.7487893849611282,
+      "num_tokens": 5667174.0,
       "step": 152
     },
     {
-      "entropy": 0.5948211699724197,
-      "epoch": 2.8411214953271027,
-      "grad_norm": 0.025879517197608948,
+      "entropy": 0.6295798048377037,
+      "epoch": 0.9927007299270073,
+      "grad_norm": 0.00907177198678255,
       "learning_rate": 0.0002,
-      "loss": 0.5868144035339355,
-      "mean_token_accuracy": 0.7634373158216476,
-      "num_tokens": 12004145.0,
+      "loss": 0.6333097219467163,
+      "mean_token_accuracy": 0.7429074198007584,
+      "num_tokens": 5704388.0,
       "step": 153
     },
     {
-      "entropy": 0.5877972543239594,
-      "epoch": 2.8598130841121496,
-      "grad_norm": 0.021945485845208168,
+      "entropy": 0.6149610579013824,
+      "epoch": 0.9991889699918897,
+      "grad_norm": 0.010923929512500763,
       "learning_rate": 0.0002,
-      "loss": 0.5912054777145386,
-      "mean_token_accuracy": 0.7609079629182816,
-      "num_tokens": 12083572.0,
+      "loss": 0.6188989877700806,
+      "mean_token_accuracy": 0.7496026828885078,
+      "num_tokens": 5741236.0,
       "step": 154
     },
     {
-      "entropy": 0.596399575471878,
-      "epoch": 2.878504672897196,
-      "grad_norm": 0.02310393564403057,
+      "entropy": 0.6364340782165527,
+      "epoch": 1.0,
+      "grad_norm": 0.023092100396752357,
       "learning_rate": 0.0002,
-      "loss": 0.5992023944854736,
-      "mean_token_accuracy": 0.7590157091617584,
-      "num_tokens": 12163010.0,
+      "loss": 0.6314291954040527,
+      "mean_token_accuracy": 0.7433016300201416,
+      "num_tokens": 5745896.0,
       "step": 155
     },
     {
-      "entropy": 0.6026438027620316,
-      "epoch": 2.897196261682243,
-      "grad_norm": 0.02354845404624939,
+      "entropy": 0.6255532875657082,
+      "epoch": 1.0064882400648825,
+      "grad_norm": 0.010001299902796745,
       "learning_rate": 0.0002,
-      "loss": 0.6004827618598938,
-      "mean_token_accuracy": 0.7573492974042892,
-      "num_tokens": 12242021.0,
+      "loss": 0.6274809837341309,
+      "mean_token_accuracy": 0.7447079941630363,
+      "num_tokens": 5782887.0,
       "step": 156
     },
     {
-      "entropy": 0.6044338345527649,
-      "epoch": 2.9158878504672896,
-      "grad_norm": 0.017202777788043022,
+      "entropy": 0.6249469369649887,
+      "epoch": 1.0129764801297647,
+      "grad_norm": 0.010014750063419342,
       "learning_rate": 0.0002,
-      "loss": 0.6010818481445312,
-      "mean_token_accuracy": 0.7568353414535522,
-      "num_tokens": 12320543.0,
+      "loss": 0.6293044090270996,
+      "mean_token_accuracy": 0.743991531431675,
+      "num_tokens": 5820452.0,
       "step": 157
     },
     {
-      "entropy": 0.5929509401321411,
-      "epoch": 2.9345794392523366,
-      "grad_norm": 0.02025710418820381,
+      "entropy": 0.6250087693333626,
+      "epoch": 1.0194647201946472,
+      "grad_norm": 0.009554821066558361,
       "learning_rate": 0.0002,
-      "loss": 0.5914134979248047,
-      "mean_token_accuracy": 0.7599974423646927,
-      "num_tokens": 12399529.0,
+      "loss": 0.6216879487037659,
+      "mean_token_accuracy": 0.7472942620515823,
+      "num_tokens": 5857511.0,
       "step": 158
     },
     {
-      "entropy": 0.5920598953962326,
-      "epoch": 2.953271028037383,
-      "grad_norm": 0.0207651536911726,
+      "entropy": 0.6345709040760994,
+      "epoch": 1.0259529602595296,
+      "grad_norm": 0.010529364459216595,
       "learning_rate": 0.0002,
-      "loss": 0.5891057252883911,
-      "mean_token_accuracy": 0.7613909542560577,
-      "num_tokens": 12479234.0,
+      "loss": 0.6336644887924194,
+      "mean_token_accuracy": 0.7428694814443588,
+      "num_tokens": 5895426.0,
       "step": 159
     },
     {
-      "entropy": 0.599518284201622,
-      "epoch": 2.97196261682243,
-      "grad_norm": 0.021147435531020164,
+      "entropy": 0.6269640624523163,
+      "epoch": 1.0324412003244121,
+      "grad_norm": 0.010084465146064758,
       "learning_rate": 0.0002,
-      "loss": 0.600294828414917,
-      "mean_token_accuracy": 0.7578493654727936,
-      "num_tokens": 12557675.0,
+      "loss": 0.6298044919967651,
+      "mean_token_accuracy": 0.7432648167014122,
+      "num_tokens": 5932788.0,
       "step": 160
     },
     {
-      "entropy": 0.5950981378555298,
-      "epoch": 2.9906542056074765,
-      "grad_norm": 0.020992575213313103,
+      "entropy": 0.6268706321716309,
+      "epoch": 1.0389294403892944,
+      "grad_norm": 0.009462922811508179,
       "learning_rate": 0.0002,
-      "loss": 0.5942000150680542,
-      "mean_token_accuracy": 0.7601594924926758,
-      "num_tokens": 12636883.0,
+      "loss": 0.6329535245895386,
+      "mean_token_accuracy": 0.7419963404536247,
+      "num_tokens": 5970211.0,
       "step": 161
     },
     {
-      "entropy": 0.5946215093135834,
-      "epoch": 3.0,
-      "grad_norm": 0.025768527761101723,
+      "entropy": 0.6167154312133789,
+      "epoch": 1.0454176804541768,
+      "grad_norm": 0.009687564335763454,
       "learning_rate": 0.0002,
-      "loss": 0.5949423313140869,
-      "mean_token_accuracy": 0.7577188909053802,
-      "num_tokens": 12675677.0,
+      "loss": 0.6214160919189453,
+      "mean_token_accuracy": 0.7477982342243195,
+      "num_tokens": 6007676.0,
       "step": 162
     },
     {
-      "entropy": 0.5952535271644592,
-      "epoch": 3.0186915887850465,
-      "grad_norm": 0.019218550994992256,
+      "entropy": 0.6232415363192558,
+      "epoch": 1.0519059205190593,
+      "grad_norm": 0.011512194760143757,
       "learning_rate": 0.0002,
-      "loss": 0.5950205326080322,
-      "mean_token_accuracy": 0.7591802030801773,
-      "num_tokens": 12754544.0,
+      "loss": 0.6291061639785767,
+      "mean_token_accuracy": 0.743974357843399,
+      "num_tokens": 6044881.0,
       "step": 163
     },
     {
-      "entropy": 0.5869429856538773,
-      "epoch": 3.0373831775700935,
-      "grad_norm": 0.021725568920373917,
+      "entropy": 0.6384411826729774,
+      "epoch": 1.0583941605839415,
+      "grad_norm": 0.009664908982813358,
       "learning_rate": 0.0002,
-      "loss": 0.5894550681114197,
-      "mean_token_accuracy": 0.7605041265487671,
-      "num_tokens": 12834662.0,
+      "loss": 0.6418322920799255,
+      "mean_token_accuracy": 0.7393775209784508,
+      "num_tokens": 6081959.0,
       "step": 164
     },
     {
-      "entropy": 0.5999327003955841,
-      "epoch": 3.05607476635514,
-      "grad_norm": 0.02341764234006405,
+      "entropy": 0.6312756985425949,
+      "epoch": 1.064882400648824,
+      "grad_norm": 0.009909010492265224,
       "learning_rate": 0.0002,
-      "loss": 0.6024653911590576,
-      "mean_token_accuracy": 0.7566392123699188,
-      "num_tokens": 12913130.0,
+      "loss": 0.6294491291046143,
+      "mean_token_accuracy": 0.7447398081421852,
+      "num_tokens": 6119502.0,
       "step": 165
     },
     {
-      "entropy": 0.5968442261219025,
-      "epoch": 3.074766355140187,
-      "grad_norm": 0.01957128569483757,
+      "entropy": 0.6399073377251625,
+      "epoch": 1.0713706407137065,
+      "grad_norm": 0.009380994364619255,
       "learning_rate": 0.0002,
-      "loss": 0.5923704504966736,
-      "mean_token_accuracy": 0.7603843212127686,
-      "num_tokens": 12993868.0,
+      "loss": 0.6359134912490845,
+      "mean_token_accuracy": 0.7427374646067619,
+      "num_tokens": 6157187.0,
       "step": 166
     },
     {
-      "entropy": 0.5876279473304749,
-      "epoch": 3.0934579439252334,
-      "grad_norm": 0.025587067008018494,
+      "entropy": 0.6151667386293411,
+      "epoch": 1.0778588807785887,
+      "grad_norm": 0.008555393666028976,
       "learning_rate": 0.0002,
-      "loss": 0.5829859972000122,
-      "mean_token_accuracy": 0.7648214399814606,
-      "num_tokens": 13074131.0,
+      "loss": 0.6155805587768555,
+      "mean_token_accuracy": 0.7496763169765472,
+      "num_tokens": 6194222.0,
       "step": 167
     },
     {
-      "entropy": 0.5873458981513977,
-      "epoch": 3.1121495327102804,
-      "grad_norm": 0.02201727218925953,
+      "entropy": 0.6240349486470222,
+      "epoch": 1.0843471208434712,
+      "grad_norm": 0.008739444427192211,
       "learning_rate": 0.0002,
-      "loss": 0.5889235734939575,
-      "mean_token_accuracy": 0.7630662024021149,
-      "num_tokens": 13152894.0,
+      "loss": 0.6246569156646729,
+      "mean_token_accuracy": 0.7454928457736969,
+      "num_tokens": 6231423.0,
       "step": 168
     },
     {
-      "entropy": 0.5871318429708481,
-      "epoch": 3.130841121495327,
-      "grad_norm": 0.027795247733592987,
+      "entropy": 0.6332473009824753,
+      "epoch": 1.0908353609083536,
+      "grad_norm": 0.009752689860761166,
       "learning_rate": 0.0002,
-      "loss": 0.5916054844856262,
-      "mean_token_accuracy": 0.7616733014583588,
-      "num_tokens": 13232178.0,
+      "loss": 0.6364842653274536,
+      "mean_token_accuracy": 0.7410347387194633,
+      "num_tokens": 6268835.0,
       "step": 169
     },
     {
-      "entropy": 0.5886196792125702,
-      "epoch": 3.149532710280374,
-      "grad_norm": 0.01805286295711994,
+      "entropy": 0.6179100349545479,
+      "epoch": 1.0973236009732361,
+      "grad_norm": 0.010130702517926693,
       "learning_rate": 0.0002,
-      "loss": 0.5864452123641968,
-      "mean_token_accuracy": 0.7620666176080704,
-      "num_tokens": 13310344.0,
+      "loss": 0.6235646605491638,
+      "mean_token_accuracy": 0.7480539605021477,
+      "num_tokens": 6306169.0,
       "step": 170
     },
     {
-      "entropy": 0.6028583645820618,
-      "epoch": 3.1682242990654204,
-      "grad_norm": 0.024633541703224182,
+      "entropy": 0.6232842579483986,
+      "epoch": 1.1038118410381184,
+      "grad_norm": 0.00973625760525465,
       "learning_rate": 0.0002,
-      "loss": 0.5942882299423218,
-      "mean_token_accuracy": 0.7610559463500977,
-      "num_tokens": 13389650.0,
+      "loss": 0.6288238763809204,
+      "mean_token_accuracy": 0.7431181073188782,
+      "num_tokens": 6343047.0,
       "step": 171
     },
     {
-      "entropy": 0.6023786067962646,
-      "epoch": 3.1869158878504673,
-      "grad_norm": 0.024393081665039062,
+      "entropy": 0.6252638548612595,
+      "epoch": 1.1103000811030008,
+      "grad_norm": 0.009265839122235775,
       "learning_rate": 0.0002,
-      "loss": 0.5962838530540466,
-      "mean_token_accuracy": 0.759291335940361,
-      "num_tokens": 13467561.0,
+      "loss": 0.6260301470756531,
+      "mean_token_accuracy": 0.7449894994497299,
+      "num_tokens": 6380248.0,
       "step": 172
     },
     {
-      "entropy": 0.5901230871677399,
-      "epoch": 3.205607476635514,
-      "grad_norm": 0.023710792884230614,
+      "entropy": 0.6273873671889305,
+      "epoch": 1.1167883211678833,
+      "grad_norm": 0.009544978849589825,
       "learning_rate": 0.0002,
-      "loss": 0.5943902730941772,
-      "mean_token_accuracy": 0.7589634358882904,
-      "num_tokens": 13546021.0,
+      "loss": 0.624110996723175,
+      "mean_token_accuracy": 0.7455884218215942,
+      "num_tokens": 6417799.0,
       "step": 173
     },
     {
-      "entropy": 0.590122401714325,
-      "epoch": 3.2242990654205608,
-      "grad_norm": 0.02311398833990097,
+      "entropy": 0.6261554434895515,
+      "epoch": 1.1232765612327655,
+      "grad_norm": 0.009425261989235878,
       "learning_rate": 0.0002,
-      "loss": 0.596176266670227,
-      "mean_token_accuracy": 0.7589921206235886,
-      "num_tokens": 13624893.0,
+      "loss": 0.6238696575164795,
+      "mean_token_accuracy": 0.747870922088623,
+      "num_tokens": 6454969.0,
       "step": 174
     },
     {
-      "entropy": 0.5887411236763,
-      "epoch": 3.2429906542056073,
-      "grad_norm": 0.0215385053306818,
+      "entropy": 0.6361829712986946,
+      "epoch": 1.129764801297648,
+      "grad_norm": 0.009325449354946613,
       "learning_rate": 0.0002,
-      "loss": 0.5897305607795715,
-      "mean_token_accuracy": 0.7616126239299774,
-      "num_tokens": 13703582.0,
+      "loss": 0.6336718201637268,
+      "mean_token_accuracy": 0.7412231788039207,
+      "num_tokens": 6493115.0,
       "step": 175
     },
     {
-      "entropy": 0.5897439420223236,
-      "epoch": 3.2616822429906542,
-      "grad_norm": 0.02364741452038288,
+      "entropy": 0.626146711409092,
+      "epoch": 1.1362530413625305,
+      "grad_norm": 0.010177000425755978,
       "learning_rate": 0.0002,
-      "loss": 0.5889256000518799,
-      "mean_token_accuracy": 0.7620260119438171,
-      "num_tokens": 13781679.0,
+      "loss": 0.6264243125915527,
+      "mean_token_accuracy": 0.7452349737286568,
+      "num_tokens": 6530141.0,
       "step": 176
     },
     {
-      "entropy": 0.6066396534442902,
-      "epoch": 3.2803738317757007,
-      "grad_norm": 0.024549856781959534,
+      "entropy": 0.620332308113575,
+      "epoch": 1.142741281427413,
+      "grad_norm": 0.010463408194482327,
       "learning_rate": 0.0002,
-      "loss": 0.6063922643661499,
-      "mean_token_accuracy": 0.7544234395027161,
-      "num_tokens": 13860430.0,
+      "loss": 0.6263006329536438,
+      "mean_token_accuracy": 0.7455399706959724,
+      "num_tokens": 6567506.0,
       "step": 177
     },
     {
-      "entropy": 0.5902666449546814,
-      "epoch": 3.2990654205607477,
-      "grad_norm": 0.02692057006061077,
+      "entropy": 0.6137627214193344,
+      "epoch": 1.1492295214922952,
+      "grad_norm": 0.011479280889034271,
       "learning_rate": 0.0002,
-      "loss": 0.5886927843093872,
-      "mean_token_accuracy": 0.7610446214675903,
-      "num_tokens": 13939321.0,
+      "loss": 0.6208648681640625,
+      "mean_token_accuracy": 0.7491373345255852,
+      "num_tokens": 6604409.0,
       "step": 178
     },
     {
-      "entropy": 0.5902948081493378,
-      "epoch": 3.317757009345794,
-      "grad_norm": 0.019016796723008156,
+      "entropy": 0.6161536052823067,
+      "epoch": 1.1557177615571776,
+      "grad_norm": 0.010295857675373554,
       "learning_rate": 0.0002,
-      "loss": 0.584492027759552,
-      "mean_token_accuracy": 0.7628366202116013,
-      "num_tokens": 14019328.0,
+      "loss": 0.6196001768112183,
+      "mean_token_accuracy": 0.7471510693430901,
+      "num_tokens": 6641779.0,
       "step": 179
     },
     {
-      "entropy": 0.5843053162097931,
-      "epoch": 3.336448598130841,
-      "grad_norm": 0.02076794020831585,
+      "entropy": 0.615701287984848,
+      "epoch": 1.1622060016220601,
+      "grad_norm": 0.00977401528507471,
       "learning_rate": 0.0002,
-      "loss": 0.5817679166793823,
-      "mean_token_accuracy": 0.76412633061409,
-      "num_tokens": 14098406.0,
+      "loss": 0.6152433156967163,
+      "mean_token_accuracy": 0.7513743862509727,
+      "num_tokens": 6678811.0,
       "step": 180
     },
     {
-      "entropy": 0.5929713398218155,
-      "epoch": 3.3551401869158877,
-      "grad_norm": 0.022940950468182564,
+      "entropy": 0.6276117786765099,
+      "epoch": 1.1686942416869424,
+      "grad_norm": 0.009978534653782845,
       "learning_rate": 0.0002,
-      "loss": 0.5940371155738831,
-      "mean_token_accuracy": 0.7611016482114792,
-      "num_tokens": 14177564.0,
+      "loss": 0.6257454752922058,
+      "mean_token_accuracy": 0.7457039654254913,
+      "num_tokens": 6715971.0,
       "step": 181
     },
     {
-      "entropy": 0.5819283872842789,
-      "epoch": 3.3738317757009346,
-      "grad_norm": 0.01927620731294155,
+      "entropy": 0.622728981077671,
+      "epoch": 1.1751824817518248,
+      "grad_norm": 0.010105178691446781,
       "learning_rate": 0.0002,
-      "loss": 0.5849300622940063,
-      "mean_token_accuracy": 0.7633251845836639,
-      "num_tokens": 14255531.0,
+      "loss": 0.6200466156005859,
+      "mean_token_accuracy": 0.7463524416089058,
+      "num_tokens": 6753100.0,
       "step": 182
     },
     {
-      "entropy": 0.578275516629219,
-      "epoch": 3.392523364485981,
-      "grad_norm": 0.019883625209331512,
+      "entropy": 0.6278671473264694,
+      "epoch": 1.1816707218167073,
+      "grad_norm": 0.00981152057647705,
       "learning_rate": 0.0002,
-      "loss": 0.5792434215545654,
-      "mean_token_accuracy": 0.7661247700452805,
-      "num_tokens": 14334891.0,
+      "loss": 0.629209041595459,
+      "mean_token_accuracy": 0.7440850958228111,
+      "num_tokens": 6790606.0,
       "step": 183
     },
     {
-      "entropy": 0.5787920653820038,
-      "epoch": 3.411214953271028,
-      "grad_norm": 0.020030220970511436,
+      "entropy": 0.6340474411845207,
+      "epoch": 1.1881589618815895,
+      "grad_norm": 0.010966816917061806,
       "learning_rate": 0.0002,
-      "loss": 0.5766516923904419,
-      "mean_token_accuracy": 0.7661706358194351,
-      "num_tokens": 14413858.0,
+      "loss": 0.6343464851379395,
+      "mean_token_accuracy": 0.7423162013292313,
+      "num_tokens": 6827773.0,
       "step": 184
     },
     {
-      "entropy": 0.5973383039236069,
-      "epoch": 3.4299065420560746,
-      "grad_norm": 0.020308909937739372,
+      "entropy": 0.627871498465538,
+      "epoch": 1.194647201946472,
+      "grad_norm": 0.01046378631144762,
       "learning_rate": 0.0002,
-      "loss": 0.5918881297111511,
-      "mean_token_accuracy": 0.7606652081012726,
-      "num_tokens": 14492504.0,
+      "loss": 0.6280568838119507,
+      "mean_token_accuracy": 0.7452474683523178,
+      "num_tokens": 6865004.0,
       "step": 185
     },
     {
-      "entropy": 0.583291083574295,
-      "epoch": 3.4485981308411215,
-      "grad_norm": 0.020716240629553795,
+      "entropy": 0.632583849132061,
+      "epoch": 1.2011354420113545,
+      "grad_norm": 0.011151660233736038,
       "learning_rate": 0.0002,
-      "loss": 0.5781140327453613,
-      "mean_token_accuracy": 0.765817254781723,
-      "num_tokens": 14572853.0,
+      "loss": 0.6399388313293457,
+      "mean_token_accuracy": 0.7401903197169304,
+      "num_tokens": 6902199.0,
       "step": 186
     },
     {
-      "entropy": 0.5810463130474091,
-      "epoch": 3.467289719626168,
-      "grad_norm": 0.020133642479777336,
+      "entropy": 0.6207470819354057,
+      "epoch": 1.2076236820762367,
+      "grad_norm": 0.009539226070046425,
       "learning_rate": 0.0002,
-      "loss": 0.5826877355575562,
-      "mean_token_accuracy": 0.7634966522455215,
-      "num_tokens": 14651962.0,
+      "loss": 0.626314640045166,
+      "mean_token_accuracy": 0.7447792664170265,
+      "num_tokens": 6939345.0,
       "step": 187
     },
     {
-      "entropy": 0.57536381483078,
-      "epoch": 3.485981308411215,
-      "grad_norm": 0.020142007619142532,
+      "entropy": 0.6265325620770454,
+      "epoch": 1.2141119221411192,
+      "grad_norm": 0.010582523420453072,
       "learning_rate": 0.0002,
-      "loss": 0.5788486003875732,
-      "mean_token_accuracy": 0.7644489705562592,
-      "num_tokens": 14733151.0,
+      "loss": 0.6258584260940552,
+      "mean_token_accuracy": 0.7450206950306892,
+      "num_tokens": 6976749.0,
       "step": 188
     },
     {
-      "entropy": 0.5983182191848755,
-      "epoch": 3.5046728971962615,
-      "grad_norm": 0.018696345388889313,
+      "entropy": 0.6242929175496101,
+      "epoch": 1.2206001622060016,
+      "grad_norm": 0.010438770987093449,
       "learning_rate": 0.0002,
-      "loss": 0.5981752872467041,
-      "mean_token_accuracy": 0.7576424926519394,
-      "num_tokens": 14813025.0,
+      "loss": 0.6282113790512085,
+      "mean_token_accuracy": 0.7453068420290947,
+      "num_tokens": 7013986.0,
       "step": 189
     },
     {
-      "entropy": 0.6043048650026321,
-      "epoch": 3.5233644859813085,
-      "grad_norm": 0.022231722250580788,
+      "entropy": 0.6362300962209702,
+      "epoch": 1.2270884022708841,
+      "grad_norm": 0.009592331014573574,
       "learning_rate": 0.0002,
-      "loss": 0.5974507331848145,
-      "mean_token_accuracy": 0.7578679323196411,
-      "num_tokens": 14891780.0,
+      "loss": 0.6338801980018616,
+      "mean_token_accuracy": 0.7425767108798027,
+      "num_tokens": 7051122.0,
       "step": 190
     },
     {
-      "entropy": 0.6004075407981873,
-      "epoch": 3.542056074766355,
-      "grad_norm": 0.021045353263616562,
+      "entropy": 0.6142800450325012,
+      "epoch": 1.2335766423357664,
+      "grad_norm": 0.008985689841210842,
       "learning_rate": 0.0002,
-      "loss": 0.5976691246032715,
-      "mean_token_accuracy": 0.7593707591295242,
-      "num_tokens": 14970339.0,
+      "loss": 0.6143558025360107,
+      "mean_token_accuracy": 0.750966027379036,
+      "num_tokens": 7088050.0,
       "step": 191
     },
     {
-      "entropy": 0.5853922069072723,
-      "epoch": 3.560747663551402,
-      "grad_norm": 0.02037801779806614,
+      "entropy": 0.6302971392869949,
+      "epoch": 1.2400648824006488,
+      "grad_norm": 0.009001200087368488,
       "learning_rate": 0.0002,
-      "loss": 0.5871971845626831,
-      "mean_token_accuracy": 0.7610767930746078,
-      "num_tokens": 15050034.0,
+      "loss": 0.6324248313903809,
+      "mean_token_accuracy": 0.7426666542887688,
+      "num_tokens": 7125652.0,
       "step": 192
     },
     {
-      "entropy": 0.5899361073970795,
-      "epoch": 3.5794392523364484,
-      "grad_norm": 0.021168449893593788,
+      "entropy": 0.6278156340122223,
+      "epoch": 1.2465531224655313,
+      "grad_norm": 0.009873399510979652,
       "learning_rate": 0.0002,
-      "loss": 0.5936985015869141,
-      "mean_token_accuracy": 0.759898915886879,
-      "num_tokens": 15128028.0,
+      "loss": 0.6288595795631409,
+      "mean_token_accuracy": 0.7446199506521225,
+      "num_tokens": 7163209.0,
       "step": 193
     },
     {
-      "entropy": 0.5894113481044769,
-      "epoch": 3.5981308411214954,
-      "grad_norm": 0.019629884511232376,
+      "entropy": 0.6151107102632523,
+      "epoch": 1.2530413625304138,
+      "grad_norm": 0.010955240577459335,
       "learning_rate": 0.0002,
-      "loss": 0.5924787521362305,
-      "mean_token_accuracy": 0.7594863474369049,
-      "num_tokens": 15207165.0,
+      "loss": 0.6196198463439941,
+      "mean_token_accuracy": 0.7487600594758987,
+      "num_tokens": 7200676.0,
       "step": 194
     },
     {
-      "entropy": 0.59434375166893,
-      "epoch": 3.616822429906542,
-      "grad_norm": 0.01782124675810337,
+      "entropy": 0.624420776963234,
+      "epoch": 1.259529602595296,
+      "grad_norm": 0.009569366462528706,
       "learning_rate": 0.0002,
-      "loss": 0.5949936509132385,
-      "mean_token_accuracy": 0.759057804942131,
-      "num_tokens": 15286394.0,
+      "loss": 0.6245594024658203,
+      "mean_token_accuracy": 0.7455159053206444,
+      "num_tokens": 7237914.0,
       "step": 195
     },
     {
-      "entropy": 0.5976666957139969,
-      "epoch": 3.635514018691589,
-      "grad_norm": 0.019005272537469864,
+      "entropy": 0.62844218313694,
+      "epoch": 1.2660178426601785,
+      "grad_norm": 0.00903658103197813,
       "learning_rate": 0.0002,
-      "loss": 0.5931147933006287,
-      "mean_token_accuracy": 0.7595807015895844,
-      "num_tokens": 15365716.0,
+      "loss": 0.6268140077590942,
+      "mean_token_accuracy": 0.7442804053425789,
+      "num_tokens": 7275515.0,
       "step": 196
     },
     {
-      "entropy": 0.5822330415248871,
-      "epoch": 3.6542056074766354,
-      "grad_norm": 0.021969236433506012,
+      "entropy": 0.6202637255191803,
+      "epoch": 1.272506082725061,
+      "grad_norm": 0.00881174299865961,
       "learning_rate": 0.0002,
-      "loss": 0.5772209763526917,
-      "mean_token_accuracy": 0.7657969892024994,
-      "num_tokens": 15445001.0,
+      "loss": 0.6185364723205566,
+      "mean_token_accuracy": 0.7477486506104469,
+      "num_tokens": 7313158.0,
       "step": 197
     },
     {
-      "entropy": 0.5981509387493134,
-      "epoch": 3.6728971962616823,
-      "grad_norm": 0.01774434745311737,
+      "entropy": 0.6143034398555756,
+      "epoch": 1.2789943227899432,
+      "grad_norm": 0.010709850117564201,
       "learning_rate": 0.0002,
-      "loss": 0.5937886834144592,
-      "mean_token_accuracy": 0.7590377032756805,
-      "num_tokens": 15523368.0,
+      "loss": 0.6169742345809937,
+      "mean_token_accuracy": 0.7487710192799568,
+      "num_tokens": 7350030.0,
       "step": 198
     },
     {
-      "entropy": 0.5803258270025253,
-      "epoch": 3.691588785046729,
-      "grad_norm": 0.017782161012291908,
+      "entropy": 0.6223605200648308,
+      "epoch": 1.2854825628548256,
+      "grad_norm": 0.010480264201760292,
       "learning_rate": 0.0002,
-      "loss": 0.5816370248794556,
-      "mean_token_accuracy": 0.7656546384096146,
-      "num_tokens": 15601373.0,
+      "loss": 0.6294817924499512,
+      "mean_token_accuracy": 0.7423868179321289,
+      "num_tokens": 7387006.0,
       "step": 199
     },
     {
-      "entropy": 0.5881159752607346,
-      "epoch": 3.710280373831776,
-      "grad_norm": 0.023958368226885796,
+      "entropy": 0.6272002533078194,
+      "epoch": 1.2919708029197081,
+      "grad_norm": 0.009464031085371971,
       "learning_rate": 0.0002,
-      "loss": 0.5951263904571533,
-      "mean_token_accuracy": 0.7590856105089188,
-      "num_tokens": 15678762.0,
+      "loss": 0.6319963932037354,
+      "mean_token_accuracy": 0.7425662279129028,
+      "num_tokens": 7424425.0,
       "step": 200
     },
     {
-      "entropy": 0.5951340943574905,
-      "epoch": 3.7289719626168223,
-      "grad_norm": 0.019875694066286087,
+      "entropy": 0.6301053911447525,
+      "epoch": 1.2984590429845904,
+      "grad_norm": 0.009957416914403439,
       "learning_rate": 0.0002,
-      "loss": 0.595394492149353,
-      "mean_token_accuracy": 0.7592459321022034,
-      "num_tokens": 15757310.0,
+      "loss": 0.6337630748748779,
+      "mean_token_accuracy": 0.7440446689724922,
+      "num_tokens": 7461695.0,
       "step": 201
     },
     {
-      "entropy": 0.5929934829473495,
-      "epoch": 3.7476635514018692,
-      "grad_norm": 0.018640384078025818,
+      "entropy": 0.6238225474953651,
+      "epoch": 1.3049472830494728,
+      "grad_norm": 0.00954597070813179,
       "learning_rate": 0.0002,
-      "loss": 0.5888406038284302,
-      "mean_token_accuracy": 0.7608652859926224,
-      "num_tokens": 15835991.0,
+      "loss": 0.6226916909217834,
+      "mean_token_accuracy": 0.7449641600251198,
+      "num_tokens": 7498774.0,
       "step": 202
     },
     {
-      "entropy": 0.5996460020542145,
-      "epoch": 3.7663551401869158,
-      "grad_norm": 0.021940160542726517,
+      "entropy": 0.630424328148365,
+      "epoch": 1.3114355231143553,
+      "grad_norm": 0.0086854612454772,
       "learning_rate": 0.0002,
-      "loss": 0.5948377847671509,
-      "mean_token_accuracy": 0.7593396604061127,
-      "num_tokens": 15915033.0,
+      "loss": 0.6267648935317993,
+      "mean_token_accuracy": 0.745999850332737,
+      "num_tokens": 7535974.0,
       "step": 203
     },
     {
-      "entropy": 0.5859507322311401,
-      "epoch": 3.7850467289719627,
-      "grad_norm": 0.019842160865664482,
+      "entropy": 0.6177534386515617,
+      "epoch": 1.3179237631792375,
+      "grad_norm": 0.009849452413618565,
       "learning_rate": 0.0002,
-      "loss": 0.5832802057266235,
-      "mean_token_accuracy": 0.7638317793607712,
-      "num_tokens": 15994090.0,
+      "loss": 0.6153725385665894,
+      "mean_token_accuracy": 0.7507055774331093,
+      "num_tokens": 7572957.0,
       "step": 204
     },
     {
-      "entropy": 0.588510274887085,
-      "epoch": 3.803738317757009,
-      "grad_norm": 0.02117988094687462,
+      "entropy": 0.622411735355854,
+      "epoch": 1.32441200324412,
+      "grad_norm": 0.009472545236349106,
       "learning_rate": 0.0002,
-      "loss": 0.5909595489501953,
-      "mean_token_accuracy": 0.7599914968013763,
-      "num_tokens": 16073191.0,
+      "loss": 0.625112771987915,
+      "mean_token_accuracy": 0.7435147687792778,
+      "num_tokens": 7610380.0,
       "step": 205
     },
     {
-      "entropy": 0.5994965881109238,
-      "epoch": 3.822429906542056,
-      "grad_norm": 0.024261780083179474,
+      "entropy": 0.6200075596570969,
+      "epoch": 1.3309002433090025,
+      "grad_norm": 0.010570540092885494,
       "learning_rate": 0.0002,
-      "loss": 0.6034358143806458,
-      "mean_token_accuracy": 0.7556708604097366,
-      "num_tokens": 16151732.0,
+      "loss": 0.6272220611572266,
+      "mean_token_accuracy": 0.7441560178995132,
+      "num_tokens": 7647843.0,
       "step": 206
     },
     {
-      "entropy": 0.5854174643754959,
-      "epoch": 3.8411214953271027,
-      "grad_norm": 0.021501438692212105,
+      "entropy": 0.6281143799424171,
+      "epoch": 1.3373884833738847,
+      "grad_norm": 0.008813613094389439,
       "learning_rate": 0.0002,
-      "loss": 0.5888152122497559,
-      "mean_token_accuracy": 0.7620731145143509,
-      "num_tokens": 16231390.0,
+      "loss": 0.6328994631767273,
+      "mean_token_accuracy": 0.742665521800518,
+      "num_tokens": 7685138.0,
       "step": 207
     },
     {
-      "entropy": 0.6051391512155533,
-      "epoch": 3.8598130841121496,
-      "grad_norm": 0.01915089599788189,
+      "entropy": 0.6417431458830833,
+      "epoch": 1.3438767234387672,
+      "grad_norm": 0.00971722137182951,
       "learning_rate": 0.0002,
-      "loss": 0.6023332476615906,
-      "mean_token_accuracy": 0.7562840282917023,
-      "num_tokens": 16309103.0,
+      "loss": 0.6397589445114136,
+      "mean_token_accuracy": 0.7406186908483505,
+      "num_tokens": 7722781.0,
       "step": 208
     },
     {
-      "entropy": 0.6067698895931244,
-      "epoch": 3.878504672897196,
-      "grad_norm": 0.021229425445199013,
+      "entropy": 0.6444113031029701,
+      "epoch": 1.3503649635036497,
+      "grad_norm": 0.009866150096058846,
       "learning_rate": 0.0002,
-      "loss": 0.599652886390686,
-      "mean_token_accuracy": 0.7584574669599533,
-      "num_tokens": 16387337.0,
+      "loss": 0.6412906646728516,
+      "mean_token_accuracy": 0.7412897050380707,
+      "num_tokens": 7760109.0,
       "step": 209
     },
     {
-      "entropy": 0.5951249748468399,
-      "epoch": 3.897196261682243,
-      "grad_norm": 0.020794285461306572,
+      "entropy": 0.6214337572455406,
+      "epoch": 1.3568532035685321,
+      "grad_norm": 0.009207811206579208,
       "learning_rate": 0.0002,
-      "loss": 0.589394211769104,
-      "mean_token_accuracy": 0.7610641866922379,
-      "num_tokens": 16465143.0,
+      "loss": 0.6234208345413208,
+      "mean_token_accuracy": 0.7459528893232346,
+      "num_tokens": 7797532.0,
       "step": 210
     },
     {
-      "entropy": 0.5834344774484634,
-      "epoch": 3.9158878504672896,
-      "grad_norm": 0.021171296015381813,
+      "entropy": 0.6157770529389381,
+      "epoch": 1.3633414436334144,
+      "grad_norm": 0.01025467086583376,
       "learning_rate": 0.0002,
-      "loss": 0.587925374507904,
-      "mean_token_accuracy": 0.7619622200727463,
-      "num_tokens": 16543880.0,
+      "loss": 0.6178190112113953,
+      "mean_token_accuracy": 0.7505671828985214,
+      "num_tokens": 7834554.0,
       "step": 211
     },
     {
-      "entropy": 0.5819447487592697,
-      "epoch": 3.9345794392523366,
-      "grad_norm": 0.021571574732661247,
+      "entropy": 0.619218721985817,
+      "epoch": 1.3698296836982968,
+      "grad_norm": 0.01020651962608099,
       "learning_rate": 0.0002,
-      "loss": 0.5871191024780273,
-      "mean_token_accuracy": 0.7640725821256638,
-      "num_tokens": 16624144.0,
+      "loss": 0.6230384111404419,
+      "mean_token_accuracy": 0.7464535236358643,
+      "num_tokens": 7871899.0,
       "step": 212
     },
     {
-      "entropy": 0.5863354504108429,
-      "epoch": 3.953271028037383,
-      "grad_norm": 0.018521364778280258,
+      "entropy": 0.6242957413196564,
+      "epoch": 1.3763179237631793,
+      "grad_norm": 0.009102249518036842,
       "learning_rate": 0.0002,
-      "loss": 0.5881295204162598,
-      "mean_token_accuracy": 0.7625996768474579,
-      "num_tokens": 16702818.0,
+      "loss": 0.625146210193634,
+      "mean_token_accuracy": 0.7436935156583786,
+      "num_tokens": 7908550.0,
       "step": 213
     },
     {
-      "entropy": 0.5845038741827011,
-      "epoch": 3.97196261682243,
-      "grad_norm": 0.01778440922498703,
+      "entropy": 0.6318832859396935,
+      "epoch": 1.3828061638280618,
+      "grad_norm": 0.008958768099546432,
       "learning_rate": 0.0002,
-      "loss": 0.5809848308563232,
-      "mean_token_accuracy": 0.7636913359165192,
-      "num_tokens": 16781881.0,
+      "loss": 0.6291754245758057,
+      "mean_token_accuracy": 0.7419761046767235,
+      "num_tokens": 7945528.0,
       "step": 214
     },
     {
-      "entropy": 0.6045233011245728,
-      "epoch": 3.9906542056074765,
-      "grad_norm": 0.02102098986506462,
+      "entropy": 0.622461624443531,
+      "epoch": 1.389294403892944,
+      "grad_norm": 0.010115247219800949,
       "learning_rate": 0.0002,
-      "loss": 0.5983396768569946,
-      "mean_token_accuracy": 0.7578406184911728,
-      "num_tokens": 16860588.0,
+      "loss": 0.6236110329627991,
+      "mean_token_accuracy": 0.7462903261184692,
+      "num_tokens": 7982511.0,
       "step": 215
     },
     {
-      "entropy": 0.586735337972641,
-      "epoch": 4.0,
-      "grad_norm": 0.021059105172753334,
+      "entropy": 0.6254477724432945,
+      "epoch": 1.3957826439578265,
+      "grad_norm": 0.008989102207124233,
       "learning_rate": 0.0002,
-      "loss": 0.5811917781829834,
-      "mean_token_accuracy": 0.7643606066703796,
-      "num_tokens": 16900889.0,
+      "loss": 0.6284307241439819,
+      "mean_token_accuracy": 0.7429582253098488,
+      "num_tokens": 8019795.0,
       "step": 216
     },
     {
-      "entropy": 0.5940107554197311,
-      "epoch": 4.018691588785047,
-      "grad_norm": 0.018036525696516037,
+      "entropy": 0.6165763810276985,
+      "epoch": 1.402270884022709,
+      "grad_norm": 0.010876733809709549,
       "learning_rate": 0.0002,
-      "loss": 0.5917088985443115,
-      "mean_token_accuracy": 0.7601685523986816,
-      "num_tokens": 16979059.0,
+      "loss": 0.6224349737167358,
+      "mean_token_accuracy": 0.7477402463555336,
+      "num_tokens": 8057519.0,
       "step": 217
     },
     {
-      "entropy": 0.5794219076633453,
-      "epoch": 4.037383177570093,
-      "grad_norm": 0.02091035433113575,
+      "entropy": 0.6111054793000221,
+      "epoch": 1.4087591240875912,
+      "grad_norm": 0.008845480158925056,
       "learning_rate": 0.0002,
-      "loss": 0.5814393162727356,
-      "mean_token_accuracy": 0.7636075466871262,
-      "num_tokens": 17058416.0,
+      "loss": 0.6134991645812988,
+      "mean_token_accuracy": 0.7499867528676987,
+      "num_tokens": 8094863.0,
       "step": 218
     },
     {
-      "entropy": 0.5681750178337097,
-      "epoch": 4.05607476635514,
-      "grad_norm": 0.02142472006380558,
+      "entropy": 0.6361298933625221,
+      "epoch": 1.4152473641524737,
+      "grad_norm": 0.009090340696275234,
       "learning_rate": 0.0002,
-      "loss": 0.5742322206497192,
-      "mean_token_accuracy": 0.7665071785449982,
-      "num_tokens": 17138894.0,
+      "loss": 0.6369084715843201,
+      "mean_token_accuracy": 0.7387163192033768,
+      "num_tokens": 8132597.0,
       "step": 219
     },
     {
-      "entropy": 0.5883741974830627,
-      "epoch": 4.074766355140187,
-      "grad_norm": 0.018591072410345078,
+      "entropy": 0.6406821832060814,
+      "epoch": 1.4217356042173561,
+      "grad_norm": 0.0092760706320405,
       "learning_rate": 0.0002,
-      "loss": 0.5917633771896362,
-      "mean_token_accuracy": 0.7602983862161636,
-      "num_tokens": 17217518.0,
+      "loss": 0.6413878202438354,
+      "mean_token_accuracy": 0.7377218827605247,
+      "num_tokens": 8170263.0,
       "step": 220
     },
     {
-      "entropy": 0.586336225271225,
-      "epoch": 4.093457943925234,
-      "grad_norm": 0.018653078004717827,
+      "entropy": 0.6221970841288567,
+      "epoch": 1.4282238442822384,
+      "grad_norm": 0.009138553403317928,
       "learning_rate": 0.0002,
-      "loss": 0.5852929949760437,
-      "mean_token_accuracy": 0.7621860057115555,
-      "num_tokens": 17297429.0,
+      "loss": 0.6239045858383179,
+      "mean_token_accuracy": 0.7459966242313385,
+      "num_tokens": 8207517.0,
       "step": 221
     },
     {
-      "entropy": 0.5893379747867584,
-      "epoch": 4.11214953271028,
-      "grad_norm": 0.02176232449710369,
+      "entropy": 0.6360906660556793,
+      "epoch": 1.4347120843471208,
+      "grad_norm": 0.008928137831389904,
       "learning_rate": 0.0002,
-      "loss": 0.5851144790649414,
-      "mean_token_accuracy": 0.7623254060745239,
-      "num_tokens": 17375993.0,
+      "loss": 0.6395260095596313,
+      "mean_token_accuracy": 0.7385511174798012,
+      "num_tokens": 8244836.0,
       "step": 222
     },
     {
-      "entropy": 0.6061349809169769,
-      "epoch": 4.130841121495327,
-      "grad_norm": 0.019815821200609207,
+      "entropy": 0.6243730783462524,
+      "epoch": 1.4412003244120033,
+      "grad_norm": 0.009026050567626953,
       "learning_rate": 0.0002,
-      "loss": 0.6047327518463135,
-      "mean_token_accuracy": 0.7547261565923691,
-      "num_tokens": 17454782.0,
+      "loss": 0.6269445419311523,
+      "mean_token_accuracy": 0.7457072138786316,
+      "num_tokens": 8281919.0,
       "step": 223
     },
     {
-      "entropy": 0.5827231258153915,
-      "epoch": 4.149532710280374,
-      "grad_norm": 0.020148327574133873,
+      "entropy": 0.6337901726365089,
+      "epoch": 1.4476885644768855,
+      "grad_norm": 0.00801246426999569,
       "learning_rate": 0.0002,
-      "loss": 0.5840202569961548,
-      "mean_token_accuracy": 0.7633996903896332,
-      "num_tokens": 17533535.0,
+      "loss": 0.6330679059028625,
+      "mean_token_accuracy": 0.7426828071475029,
+      "num_tokens": 8319228.0,
       "step": 224
     },
     {
-      "entropy": 0.5854267328977585,
-      "epoch": 4.168224299065421,
-      "grad_norm": 0.02046262100338936,
+      "entropy": 0.631360612809658,
+      "epoch": 1.454176804541768,
+      "grad_norm": 0.009467655792832375,
       "learning_rate": 0.0002,
-      "loss": 0.5860427618026733,
-      "mean_token_accuracy": 0.7614548951387405,
-      "num_tokens": 17612055.0,
+      "loss": 0.6300734281539917,
+      "mean_token_accuracy": 0.7453820630908012,
+      "num_tokens": 8356483.0,
       "step": 225
     },
     {
-      "entropy": 0.5952410101890564,
-      "epoch": 4.186915887850467,
-      "grad_norm": 0.02046579122543335,
+      "entropy": 0.6360229924321175,
+      "epoch": 1.4606650446066505,
+      "grad_norm": 0.009358935989439487,
       "learning_rate": 0.0002,
-      "loss": 0.5909503698348999,
-      "mean_token_accuracy": 0.7594263106584549,
-      "num_tokens": 17691143.0,
+      "loss": 0.6341854929924011,
+      "mean_token_accuracy": 0.7435640841722488,
+      "num_tokens": 8393723.0,
       "step": 226
     },
     {
-      "entropy": 0.5892802178859711,
-      "epoch": 4.205607476635514,
-      "grad_norm": 0.02167705073952675,
+      "entropy": 0.6200751587748528,
+      "epoch": 1.4671532846715327,
+      "grad_norm": 0.008350496180355549,
       "learning_rate": 0.0002,
-      "loss": 0.5853694677352905,
-      "mean_token_accuracy": 0.76229228079319,
-      "num_tokens": 17770094.0,
+      "loss": 0.6218277215957642,
+      "mean_token_accuracy": 0.7479682192206383,
+      "num_tokens": 8431281.0,
       "step": 227
     },
     {
-      "entropy": 0.5899169147014618,
-      "epoch": 4.224299065420561,
-      "grad_norm": 0.022735821083188057,
+      "entropy": 0.6275244578719139,
+      "epoch": 1.4736415247364152,
+      "grad_norm": 0.009802248328924179,
       "learning_rate": 0.0002,
-      "loss": 0.5906339883804321,
-      "mean_token_accuracy": 0.7605825811624527,
-      "num_tokens": 17848840.0,
+      "loss": 0.6335194706916809,
+      "mean_token_accuracy": 0.7433927357196808,
+      "num_tokens": 8468465.0,
       "step": 228
     },
     {
-      "entropy": 0.5718450546264648,
-      "epoch": 4.242990654205608,
-      "grad_norm": 0.021596962586045265,
+      "entropy": 0.6231343746185303,
+      "epoch": 1.4801297648012977,
+      "grad_norm": 0.010109310038387775,
       "learning_rate": 0.0002,
-      "loss": 0.5731650590896606,
-      "mean_token_accuracy": 0.7678278982639313,
-      "num_tokens": 17928543.0,
+      "loss": 0.6223199367523193,
+      "mean_token_accuracy": 0.7451789528131485,
+      "num_tokens": 8505713.0,
       "step": 229
     },
     {
-      "entropy": 0.5924698412418365,
-      "epoch": 4.261682242990654,
-      "grad_norm": 0.019548065960407257,
+      "entropy": 0.6206792145967484,
+      "epoch": 1.4866180048661801,
+      "grad_norm": 0.010297620669007301,
       "learning_rate": 0.0002,
-      "loss": 0.5934122800827026,
-      "mean_token_accuracy": 0.7606871724128723,
-      "num_tokens": 18007377.0,
+      "loss": 0.6225035190582275,
+      "mean_token_accuracy": 0.7460876628756523,
+      "num_tokens": 8542783.0,
       "step": 230
     },
     {
-      "entropy": 0.6042954921722412,
-      "epoch": 4.280373831775701,
-      "grad_norm": 0.021219445392489433,
+      "entropy": 0.6291577965021133,
+      "epoch": 1.4931062449310626,
+      "grad_norm": 0.009544001892209053,
       "learning_rate": 0.0002,
-      "loss": 0.6020859479904175,
-      "mean_token_accuracy": 0.7561117857694626,
-      "num_tokens": 18086305.0,
+      "loss": 0.6289199590682983,
+      "mean_token_accuracy": 0.7454885393381119,
+      "num_tokens": 8580029.0,
       "step": 231
     },
     {
-      "entropy": 0.5859596431255341,
-      "epoch": 4.299065420560748,
-      "grad_norm": 0.020383305847644806,
+      "entropy": 0.6250879839062691,
+      "epoch": 1.4995944849959448,
+      "grad_norm": 0.008444451726973057,
       "learning_rate": 0.0002,
-      "loss": 0.5832017064094543,
-      "mean_token_accuracy": 0.762650728225708,
-      "num_tokens": 18166007.0,
+      "loss": 0.628132700920105,
+      "mean_token_accuracy": 0.7435998693108559,
+      "num_tokens": 8617067.0,
       "step": 232
     },
     {
-      "entropy": 0.5846229046583176,
-      "epoch": 4.317757009345795,
-      "grad_norm": 0.01955682970583439,
+      "entropy": 0.6236734315752983,
+      "epoch": 1.5060827250608273,
+      "grad_norm": 0.009344187565147877,
       "learning_rate": 0.0002,
-      "loss": 0.5832371115684509,
-      "mean_token_accuracy": 0.7625042945146561,
-      "num_tokens": 18244884.0,
+      "loss": 0.6258345246315002,
+      "mean_token_accuracy": 0.7458996251225471,
+      "num_tokens": 8654420.0,
       "step": 233
     },
     {
-      "entropy": 0.5895711481571198,
-      "epoch": 4.336448598130841,
-      "grad_norm": 0.01915646530687809,
+      "entropy": 0.6138928905129433,
+      "epoch": 1.5125709651257098,
+      "grad_norm": 0.009892390109598637,
       "learning_rate": 0.0002,
-      "loss": 0.591120719909668,
-      "mean_token_accuracy": 0.7605114132165909,
-      "num_tokens": 18322523.0,
+      "loss": 0.6232701539993286,
+      "mean_token_accuracy": 0.7446143329143524,
+      "num_tokens": 8691786.0,
       "step": 234
     },
     {
-      "entropy": 0.5956000238656998,
-      "epoch": 4.355140186915888,
-      "grad_norm": 0.020273521542549133,
+      "entropy": 0.6233969107270241,
+      "epoch": 1.519059205190592,
+      "grad_norm": 0.010160520672798157,
       "learning_rate": 0.0002,
-      "loss": 0.5929043292999268,
-      "mean_token_accuracy": 0.7591133713722229,
-      "num_tokens": 18400838.0,
+      "loss": 0.6211707592010498,
+      "mean_token_accuracy": 0.7472252398729324,
+      "num_tokens": 8729028.0,
       "step": 235
     },
     {
-      "entropy": 0.5959005355834961,
-      "epoch": 4.373831775700935,
-      "grad_norm": 0.020820753648877144,
+      "entropy": 0.6429286673665047,
+      "epoch": 1.5255474452554745,
+      "grad_norm": 0.0098539674654603,
       "learning_rate": 0.0002,
-      "loss": 0.5935447216033936,
-      "mean_token_accuracy": 0.7600593417882919,
-      "num_tokens": 18479961.0,
+      "loss": 0.6372069120407104,
+      "mean_token_accuracy": 0.7402112632989883,
+      "num_tokens": 8766503.0,
       "step": 236
     },
     {
-      "entropy": 0.5854752659797668,
-      "epoch": 4.392523364485982,
-      "grad_norm": 0.019046681001782417,
+      "entropy": 0.6305690258741379,
+      "epoch": 1.532035685320357,
+      "grad_norm": 0.009324081242084503,
       "learning_rate": 0.0002,
-      "loss": 0.5836696624755859,
-      "mean_token_accuracy": 0.7631508558988571,
-      "num_tokens": 18558893.0,
+      "loss": 0.6331846714019775,
+      "mean_token_accuracy": 0.7420341670513153,
+      "num_tokens": 8803595.0,
       "step": 237
     },
     {
-      "entropy": 0.5805842876434326,
-      "epoch": 4.411214953271028,
-      "grad_norm": 0.02072669006884098,
+      "entropy": 0.6305363178253174,
+      "epoch": 1.5385239253852392,
+      "grad_norm": 0.009419843554496765,
       "learning_rate": 0.0002,
-      "loss": 0.5796903967857361,
-      "mean_token_accuracy": 0.7648639678955078,
-      "num_tokens": 18636920.0,
+      "loss": 0.6305243968963623,
+      "mean_token_accuracy": 0.7442135661840439,
+      "num_tokens": 8840805.0,
       "step": 238
     },
     {
-      "entropy": 0.5842614471912384,
-      "epoch": 4.429906542056075,
-      "grad_norm": 0.021100452169775963,
+      "entropy": 0.6271083503961563,
+      "epoch": 1.5450121654501217,
+      "grad_norm": 0.008481448516249657,
       "learning_rate": 0.0002,
-      "loss": 0.5860409736633301,
-      "mean_token_accuracy": 0.7626742720603943,
-      "num_tokens": 18716615.0,
+      "loss": 0.628629207611084,
+      "mean_token_accuracy": 0.7426930069923401,
+      "num_tokens": 8877765.0,
       "step": 239
     },
     {
-      "entropy": 0.5835539698600769,
-      "epoch": 4.4485981308411215,
-      "grad_norm": 0.01936553232371807,
+      "entropy": 0.6074612811207771,
+      "epoch": 1.5515004055150041,
+      "grad_norm": 0.008962183259427547,
       "learning_rate": 0.0002,
-      "loss": 0.5853263139724731,
-      "mean_token_accuracy": 0.7634955793619156,
-      "num_tokens": 18796073.0,
+      "loss": 0.6033654808998108,
+      "mean_token_accuracy": 0.7540537416934967,
+      "num_tokens": 8915298.0,
       "step": 240
     },
     {
-      "entropy": 0.5898151844739914,
-      "epoch": 4.4672897196261685,
-      "grad_norm": 0.02180354669690132,
+      "entropy": 0.6332517638802528,
+      "epoch": 1.5579886455798864,
+      "grad_norm": 0.009697218425571918,
       "learning_rate": 0.0002,
-      "loss": 0.588620126247406,
-      "mean_token_accuracy": 0.7620908468961716,
-      "num_tokens": 18874321.0,
+      "loss": 0.6378505229949951,
+      "mean_token_accuracy": 0.7404668927192688,
+      "num_tokens": 8952480.0,
       "step": 241
     },
     {
-      "entropy": 0.5985790193080902,
-      "epoch": 4.485981308411215,
-      "grad_norm": 0.019755367189645767,
+      "entropy": 0.6220754161477089,
+      "epoch": 1.5644768856447688,
+      "grad_norm": 0.00962862279266119,
       "learning_rate": 0.0002,
-      "loss": 0.5967073440551758,
-      "mean_token_accuracy": 0.7568804025650024,
-      "num_tokens": 18952908.0,
+      "loss": 0.6282638907432556,
+      "mean_token_accuracy": 0.7437562346458435,
+      "num_tokens": 8990098.0,
       "step": 242
     },
     {
-      "entropy": 0.5842784494161606,
-      "epoch": 4.5046728971962615,
-      "grad_norm": 0.023035677149891853,
+      "entropy": 0.6257785856723785,
+      "epoch": 1.5709651257096513,
+      "grad_norm": 0.009893639013171196,
       "learning_rate": 0.0002,
-      "loss": 0.5853761434555054,
-      "mean_token_accuracy": 0.7622469514608383,
-      "num_tokens": 19033127.0,
+      "loss": 0.6297152042388916,
+      "mean_token_accuracy": 0.7440835684537888,
+      "num_tokens": 9026984.0,
       "step": 243
     },
     {
-      "entropy": 0.5959431380033493,
-      "epoch": 4.5233644859813085,
-      "grad_norm": 0.01961715891957283,
+      "entropy": 0.6254734173417091,
+      "epoch": 1.5774533657745335,
+      "grad_norm": 0.009527175687253475,
       "learning_rate": 0.0002,
-      "loss": 0.5963428020477295,
-      "mean_token_accuracy": 0.7580619305372238,
-      "num_tokens": 19112318.0,
+      "loss": 0.62839674949646,
+      "mean_token_accuracy": 0.7452434748411179,
+      "num_tokens": 9064358.0,
       "step": 244
     },
     {
-      "entropy": 0.5870704054832458,
-      "epoch": 4.542056074766355,
-      "grad_norm": 0.022338932380080223,
+      "entropy": 0.6305152326822281,
+      "epoch": 1.583941605839416,
+      "grad_norm": 0.009316297248005867,
       "learning_rate": 0.0002,
-      "loss": 0.5821323394775391,
-      "mean_token_accuracy": 0.7648284286260605,
-      "num_tokens": 19191913.0,
+      "loss": 0.6323485374450684,
+      "mean_token_accuracy": 0.74256681650877,
+      "num_tokens": 9101356.0,
       "step": 245
     },
     {
-      "entropy": 0.5891195833683014,
-      "epoch": 4.5607476635514015,
-      "grad_norm": 0.020034756511449814,
+      "entropy": 0.6291619464755058,
+      "epoch": 1.5904298459042985,
+      "grad_norm": 0.008978272788226604,
       "learning_rate": 0.0002,
-      "loss": 0.586045503616333,
-      "mean_token_accuracy": 0.7623850703239441,
-      "num_tokens": 19271876.0,
+      "loss": 0.6273784637451172,
+      "mean_token_accuracy": 0.7450171038508415,
+      "num_tokens": 9138713.0,
       "step": 246
     },
     {
-      "entropy": 0.5894780457019806,
-      "epoch": 4.579439252336448,
-      "grad_norm": 0.019152812659740448,
+      "entropy": 0.6248873770236969,
+      "epoch": 1.5969180859691807,
+      "grad_norm": 0.00856445450335741,
       "learning_rate": 0.0002,
-      "loss": 0.5905551910400391,
-      "mean_token_accuracy": 0.7604437470436096,
-      "num_tokens": 19350369.0,
+      "loss": 0.6265925168991089,
+      "mean_token_accuracy": 0.7451888769865036,
+      "num_tokens": 9175681.0,
       "step": 247
     },
     {
-      "entropy": 0.591143861413002,
-      "epoch": 4.598130841121495,
-      "grad_norm": 0.027622630819678307,
+      "entropy": 0.6189131289720535,
+      "epoch": 1.6034063260340634,
+      "grad_norm": 0.00968277920037508,
       "learning_rate": 0.0002,
-      "loss": 0.59825199842453,
-      "mean_token_accuracy": 0.7581405639648438,
-      "num_tokens": 19428773.0,
+      "loss": 0.62480229139328,
+      "mean_token_accuracy": 0.7459377571940422,
+      "num_tokens": 9213377.0,
       "step": 248
     },
     {
-      "entropy": 0.5909495502710342,
-      "epoch": 4.616822429906542,
-      "grad_norm": 0.021174976602196693,
+      "entropy": 0.6176823750138283,
+      "epoch": 1.6098945660989457,
+      "grad_norm": 0.008562079630792141,
       "learning_rate": 0.0002,
-      "loss": 0.5904080271720886,
-      "mean_token_accuracy": 0.7613418698310852,
-      "num_tokens": 19507298.0,
+      "loss": 0.6216596961021423,
+      "mean_token_accuracy": 0.7459762021899223,
+      "num_tokens": 9249919.0,
       "step": 249
     },
     {
-      "entropy": 0.5994816869497299,
-      "epoch": 4.635514018691588,
-      "grad_norm": 0.02356564812362194,
+      "entropy": 0.6210065707564354,
+      "epoch": 1.616382806163828,
+      "grad_norm": 0.009402911178767681,
       "learning_rate": 0.0002,
-      "loss": 0.5948832035064697,
-      "mean_token_accuracy": 0.7591896653175354,
-      "num_tokens": 19585802.0,
+      "loss": 0.6242903470993042,
+      "mean_token_accuracy": 0.7469801977276802,
+      "num_tokens": 9287219.0,
       "step": 250
     },
     {
-      "entropy": 0.5748025625944138,
-      "epoch": 4.654205607476635,
-      "grad_norm": 0.021352779120206833,
+      "entropy": 0.6323156431317329,
+      "epoch": 1.6228710462287106,
+      "grad_norm": 0.009092130698263645,
       "learning_rate": 0.0002,
-      "loss": 0.5704630017280579,
-      "mean_token_accuracy": 0.7698894888162613,
-      "num_tokens": 19666028.0,
+      "loss": 0.6306654810905457,
+      "mean_token_accuracy": 0.7425123229622841,
+      "num_tokens": 9324174.0,
       "step": 251
     },
     {
-      "entropy": 0.5918545722961426,
-      "epoch": 4.672897196261682,
-      "grad_norm": 0.026110973209142685,
+      "entropy": 0.6303895488381386,
+      "epoch": 1.6293592862935928,
+      "grad_norm": 0.009144228883087635,
       "learning_rate": 0.0002,
-      "loss": 0.5959534645080566,
-      "mean_token_accuracy": 0.7576208114624023,
-      "num_tokens": 19744495.0,
+      "loss": 0.6268880367279053,
+      "mean_token_accuracy": 0.7460083961486816,
+      "num_tokens": 9361494.0,
       "step": 252
     },
     {
-      "entropy": 0.5881817638874054,
-      "epoch": 4.691588785046729,
-      "grad_norm": 0.023665674030780792,
+      "entropy": 0.6353992223739624,
+      "epoch": 1.6358475263584753,
+      "grad_norm": 0.00928189791738987,
       "learning_rate": 0.0002,
-      "loss": 0.5900920033454895,
-      "mean_token_accuracy": 0.75875024497509,
-      "num_tokens": 19823204.0,
+      "loss": 0.6352483034133911,
+      "mean_token_accuracy": 0.7439208328723907,
+      "num_tokens": 9398944.0,
       "step": 253
     },
     {
-      "entropy": 0.5882778912782669,
-      "epoch": 4.710280373831775,
-      "grad_norm": 0.022530725225806236,
+      "entropy": 0.6256071850657463,
+      "epoch": 1.6423357664233578,
+      "grad_norm": 0.008181072771549225,
       "learning_rate": 0.0002,
-      "loss": 0.5867466926574707,
-      "mean_token_accuracy": 0.7620512098073959,
-      "num_tokens": 19901758.0,
+      "loss": 0.6292101144790649,
+      "mean_token_accuracy": 0.743871308863163,
+      "num_tokens": 9436011.0,
       "step": 254
     },
     {
-      "entropy": 0.5733254551887512,
-      "epoch": 4.728971962616822,
-      "grad_norm": 0.024119891226291656,
+      "entropy": 0.6333504095673561,
+      "epoch": 1.64882400648824,
+      "grad_norm": 0.009949415922164917,
       "learning_rate": 0.0002,
-      "loss": 0.5707401037216187,
-      "mean_token_accuracy": 0.7695588320493698,
-      "num_tokens": 19981817.0,
+      "loss": 0.6347252130508423,
+      "mean_token_accuracy": 0.7401739284396172,
+      "num_tokens": 9473110.0,
       "step": 255
     },
     {
-      "entropy": 0.5853608995676041,
-      "epoch": 4.747663551401869,
-      "grad_norm": 0.02340366318821907,
+      "entropy": 0.6310138329863548,
+      "epoch": 1.6553122465531225,
+      "grad_norm": 0.00846829917281866,
       "learning_rate": 0.0002,
-      "loss": 0.5855814218521118,
-      "mean_token_accuracy": 0.7639720290899277,
-      "num_tokens": 20060590.0,
+      "loss": 0.6339915990829468,
+      "mean_token_accuracy": 0.7427053153514862,
+      "num_tokens": 9510350.0,
       "step": 256
     },
     {
-      "entropy": 0.5997171998023987,
-      "epoch": 4.766355140186916,
-      "grad_norm": 0.02151717059314251,
+      "entropy": 0.6202550306916237,
+      "epoch": 1.661800486618005,
+      "grad_norm": 0.009316318668425083,
       "learning_rate": 0.0002,
-      "loss": 0.6026393175125122,
-      "mean_token_accuracy": 0.7556241452693939,
-      "num_tokens": 20138939.0,
+      "loss": 0.6241627335548401,
+      "mean_token_accuracy": 0.7441129460930824,
+      "num_tokens": 9547581.0,
       "step": 257
     },
     {
-      "entropy": 0.5747672617435455,
-      "epoch": 4.785046728971962,
-      "grad_norm": 0.024703452363610268,
+      "entropy": 0.640548974275589,
+      "epoch": 1.6682887266828872,
+      "grad_norm": 0.008892681449651718,
       "learning_rate": 0.0002,
-      "loss": 0.575732409954071,
-      "mean_token_accuracy": 0.7657794505357742,
-      "num_tokens": 20219498.0,
+      "loss": 0.6457140445709229,
+      "mean_token_accuracy": 0.736494742333889,
+      "num_tokens": 9584913.0,
       "step": 258
     },
     {
-      "entropy": 0.5892387628555298,
-      "epoch": 4.803738317757009,
-      "grad_norm": 0.020495539531111717,
+      "entropy": 0.6206653639674187,
+      "epoch": 1.6747769667477697,
+      "grad_norm": 0.008709060959517956,
       "learning_rate": 0.0002,
-      "loss": 0.5848783254623413,
-      "mean_token_accuracy": 0.7626757174730301,
-      "num_tokens": 20297931.0,
+      "loss": 0.6203189492225647,
+      "mean_token_accuracy": 0.7498921826481819,
+      "num_tokens": 9622351.0,
       "step": 259
     },
     {
-      "entropy": 0.5842539519071579,
-      "epoch": 4.822429906542056,
-      "grad_norm": 0.024631142616271973,
+      "entropy": 0.6282004490494728,
+      "epoch": 1.6812652068126521,
+      "grad_norm": 0.0090560931712389,
       "learning_rate": 0.0002,
-      "loss": 0.5821443796157837,
-      "mean_token_accuracy": 0.762598842382431,
-      "num_tokens": 20377397.0,
+      "loss": 0.6246728897094727,
+      "mean_token_accuracy": 0.746510811150074,
+      "num_tokens": 9659652.0,
       "step": 260
     },
     {
-      "entropy": 0.5811708569526672,
-      "epoch": 4.841121495327103,
-      "grad_norm": 0.019834479317069054,
+      "entropy": 0.6318396255373955,
+      "epoch": 1.6877534468775344,
+      "grad_norm": 0.007608218118548393,
       "learning_rate": 0.0002,
-      "loss": 0.5774982571601868,
-      "mean_token_accuracy": 0.7656573057174683,
-      "num_tokens": 20457183.0,
+      "loss": 0.62848961353302,
+      "mean_token_accuracy": 0.7440729290246964,
+      "num_tokens": 9697284.0,
       "step": 261
     },
     {
-      "entropy": 0.584972158074379,
-      "epoch": 4.859813084112149,
-      "grad_norm": 0.02263413555920124,
+      "entropy": 0.6200717613101006,
+      "epoch": 1.6942416869424168,
+      "grad_norm": 0.009493952617049217,
       "learning_rate": 0.0002,
-      "loss": 0.5836901068687439,
-      "mean_token_accuracy": 0.7635170221328735,
-      "num_tokens": 20536220.0,
+      "loss": 0.6228904724121094,
+      "mean_token_accuracy": 0.7480449378490448,
+      "num_tokens": 9734627.0,
       "step": 262
     },
     {
-      "entropy": 0.5772289782762527,
-      "epoch": 4.878504672897196,
-      "grad_norm": 0.022467754781246185,
+      "entropy": 0.6331067234277725,
+      "epoch": 1.7007299270072993,
+      "grad_norm": 0.008472482673823833,
       "learning_rate": 0.0002,
-      "loss": 0.574739396572113,
-      "mean_token_accuracy": 0.7662009298801422,
-      "num_tokens": 20615923.0,
+      "loss": 0.6337133646011353,
+      "mean_token_accuracy": 0.7426456362009048,
+      "num_tokens": 9772063.0,
       "step": 263
     },
     {
-      "entropy": 0.5771598666906357,
-      "epoch": 4.897196261682243,
-      "grad_norm": 0.025707783177495003,
+      "entropy": 0.6172152981162071,
+      "epoch": 1.7072181670721815,
+      "grad_norm": 0.008520574308931828,
       "learning_rate": 0.0002,
-      "loss": 0.5843824148178101,
-      "mean_token_accuracy": 0.7631357461214066,
-      "num_tokens": 20694883.0,
+      "loss": 0.6212704181671143,
+      "mean_token_accuracy": 0.7456109151244164,
+      "num_tokens": 9809281.0,
       "step": 264
     },
     {
-      "entropy": 0.5870597958564758,
-      "epoch": 4.91588785046729,
-      "grad_norm": 0.021971244364976883,
+      "entropy": 0.6255315691232681,
+      "epoch": 1.7137064071370642,
+      "grad_norm": 0.008351411670446396,
       "learning_rate": 0.0002,
-      "loss": 0.589478611946106,
-      "mean_token_accuracy": 0.7609104216098785,
-      "num_tokens": 20773561.0,
+      "loss": 0.6286416053771973,
+      "mean_token_accuracy": 0.7446007430553436,
+      "num_tokens": 9846539.0,
       "step": 265
     },
     {
-      "entropy": 0.5848358869552612,
-      "epoch": 4.934579439252336,
-      "grad_norm": 0.0271756611764431,
+      "entropy": 0.6238176673650742,
+      "epoch": 1.7201946472019465,
+      "grad_norm": 0.007990843616425991,
       "learning_rate": 0.0002,
-      "loss": 0.5806835293769836,
-      "mean_token_accuracy": 0.7638518363237381,
-      "num_tokens": 20852649.0,
+      "loss": 0.6225998401641846,
+      "mean_token_accuracy": 0.7458446845412254,
+      "num_tokens": 9884004.0,
       "step": 266
     },
     {
-      "entropy": 0.589514821767807,
-      "epoch": 4.953271028037383,
-      "grad_norm": 0.02279243804514408,
+      "entropy": 0.6216712817549706,
+      "epoch": 1.7266828872668287,
+      "grad_norm": 0.00867629423737526,
       "learning_rate": 0.0002,
-      "loss": 0.5852782726287842,
-      "mean_token_accuracy": 0.7629085779190063,
-      "num_tokens": 20930316.0,
+      "loss": 0.6247957944869995,
+      "mean_token_accuracy": 0.7463371530175209,
+      "num_tokens": 9921115.0,
       "step": 267
     },
     {
-      "entropy": 0.5957552790641785,
-      "epoch": 4.97196261682243,
-      "grad_norm": 0.020448876544833183,
+      "entropy": 0.6401625499129295,
+      "epoch": 1.7331711273317114,
+      "grad_norm": 0.00958551187068224,
       "learning_rate": 0.0002,
-      "loss": 0.5981478691101074,
-      "mean_token_accuracy": 0.7577082216739655,
-      "num_tokens": 21008020.0,
+      "loss": 0.6359888911247253,
+      "mean_token_accuracy": 0.742260254919529,
+      "num_tokens": 9958978.0,
       "step": 268
     },
     {
-      "entropy": 0.5856837332248688,
-      "epoch": 4.990654205607477,
-      "grad_norm": 0.02667153999209404,
+      "entropy": 0.6297059804201126,
+      "epoch": 1.7396593673965937,
+      "grad_norm": 0.009358945302665234,
       "learning_rate": 0.0002,
-      "loss": 0.5909496545791626,
-      "mean_token_accuracy": 0.7603503912687302,
-      "num_tokens": 21087720.0,
+      "loss": 0.6260662078857422,
+      "mean_token_accuracy": 0.7483563721179962,
+      "num_tokens": 9996046.0,
       "step": 269
     },
     {
-      "entropy": 0.6056640446186066,
-      "epoch": 5.0,
-      "grad_norm": 0.02682621218264103,
+      "entropy": 0.6102557927370071,
+      "epoch": 1.7461476074614761,
+      "grad_norm": 0.008488922379910946,
       "learning_rate": 0.0002,
-      "loss": 0.603358268737793,
-      "mean_token_accuracy": 0.7550290524959564,
-      "num_tokens": 21125912.0,
+      "loss": 0.6104397773742676,
+      "mean_token_accuracy": 0.7516083940863609,
+      "num_tokens": 10033145.0,
       "step": 270
     },
     {
-      "entropy": 0.5837572664022446,
-      "epoch": 5.018691588785047,
-      "grad_norm": 0.026227472350001335,
+      "entropy": 0.6110394448041916,
+      "epoch": 1.7526358475263586,
+      "grad_norm": 0.009671971201896667,
       "learning_rate": 0.0002,
-      "loss": 0.5720741748809814,
-      "mean_token_accuracy": 0.7683691382408142,
-      "num_tokens": 21204387.0,
+      "loss": 0.6171742081642151,
+      "mean_token_accuracy": 0.7473302856087685,
+      "num_tokens": 10070347.0,
       "step": 271
     },
     {
-      "entropy": 0.578724667429924,
-      "epoch": 5.037383177570093,
-      "grad_norm": 0.023315778002142906,
+      "entropy": 0.6102030426263809,
+      "epoch": 1.7591240875912408,
+      "grad_norm": 0.01004206482321024,
       "learning_rate": 0.0002,
-      "loss": 0.5750924348831177,
-      "mean_token_accuracy": 0.7657794952392578,
-      "num_tokens": 21283728.0,
+      "loss": 0.6197723746299744,
+      "mean_token_accuracy": 0.7478247955441475,
+      "num_tokens": 10107543.0,
       "step": 272
     },
     {
-      "entropy": 0.5838986188173294,
-      "epoch": 5.05607476635514,
-      "grad_norm": 0.020290400832891464,
+      "entropy": 0.6324944272637367,
+      "epoch": 1.7656123276561233,
+      "grad_norm": 0.008376670069992542,
       "learning_rate": 0.0002,
-      "loss": 0.5844892263412476,
-      "mean_token_accuracy": 0.7619540095329285,
-      "num_tokens": 21362457.0,
+      "loss": 0.6334455609321594,
+      "mean_token_accuracy": 0.7418324947357178,
+      "num_tokens": 10144562.0,
       "step": 273
     },
     {
-      "entropy": 0.5746825039386749,
-      "epoch": 5.074766355140187,
-      "grad_norm": 0.023446552455425262,
+      "entropy": 0.6232613623142242,
+      "epoch": 1.7721005677210058,
+      "grad_norm": 0.009440955705940723,
       "learning_rate": 0.0002,
-      "loss": 0.5793564319610596,
-      "mean_token_accuracy": 0.7645887285470963,
-      "num_tokens": 21440203.0,
+      "loss": 0.6198002099990845,
+      "mean_token_accuracy": 0.750127449631691,
+      "num_tokens": 10181837.0,
       "step": 274
     },
     {
-      "entropy": 0.5728374719619751,
-      "epoch": 5.093457943925234,
-      "grad_norm": 0.020337004214525223,
+      "entropy": 0.6218618527054787,
+      "epoch": 1.778588807785888,
+      "grad_norm": 0.008727550506591797,
       "learning_rate": 0.0002,
-      "loss": 0.5744699239730835,
-      "mean_token_accuracy": 0.7651638239622116,
-      "num_tokens": 21519380.0,
+      "loss": 0.6228897571563721,
+      "mean_token_accuracy": 0.7451170310378075,
+      "num_tokens": 10218887.0,
       "step": 275
     },
     {
-      "entropy": 0.5869368463754654,
-      "epoch": 5.11214953271028,
-      "grad_norm": 0.024822693318128586,
+      "entropy": 0.6297050788998604,
+      "epoch": 1.7850770478507705,
+      "grad_norm": 0.008303189650177956,
       "learning_rate": 0.0002,
-      "loss": 0.5826805830001831,
-      "mean_token_accuracy": 0.7643714398145676,
-      "num_tokens": 21598439.0,
+      "loss": 0.6325977444648743,
+      "mean_token_accuracy": 0.7429015561938286,
+      "num_tokens": 10256316.0,
       "step": 276
     },
     {
-      "entropy": 0.5783119201660156,
-      "epoch": 5.130841121495327,
-      "grad_norm": 0.02055417001247406,
+      "entropy": 0.6150370612740517,
+      "epoch": 1.791565287915653,
+      "grad_norm": 0.009460131637752056,
       "learning_rate": 0.0002,
-      "loss": 0.5751723051071167,
-      "mean_token_accuracy": 0.7667541205883026,
-      "num_tokens": 21678545.0,
+      "loss": 0.6189665198326111,
+      "mean_token_accuracy": 0.7513208463788033,
+      "num_tokens": 10292958.0,
       "step": 277
     },
     {
-      "entropy": 0.5900953710079193,
-      "epoch": 5.149532710280374,
-      "grad_norm": 0.021529069170355797,
+      "entropy": 0.6380631476640701,
+      "epoch": 1.7980535279805352,
+      "grad_norm": 0.008382478728890419,
       "learning_rate": 0.0002,
-      "loss": 0.5940208435058594,
-      "mean_token_accuracy": 0.758448600769043,
-      "num_tokens": 21755725.0,
+      "loss": 0.6411799192428589,
+      "mean_token_accuracy": 0.7389534488320351,
+      "num_tokens": 10330143.0,
       "step": 278
     },
     {
-      "entropy": 0.5827012360095978,
-      "epoch": 5.168224299065421,
-      "grad_norm": 0.030911732465028763,
+      "entropy": 0.6122751757502556,
+      "epoch": 1.8045417680454177,
+      "grad_norm": 0.0077201686799526215,
       "learning_rate": 0.0002,
-      "loss": 0.5863416194915771,
-      "mean_token_accuracy": 0.7614113390445709,
-      "num_tokens": 21834983.0,
+      "loss": 0.6154634356498718,
+      "mean_token_accuracy": 0.7478373274207115,
+      "num_tokens": 10367348.0,
       "step": 279
     },
     {
-      "entropy": 0.5782018750905991,
-      "epoch": 5.186915887850467,
-      "grad_norm": 0.022593926638364792,
+      "entropy": 0.6171681880950928,
+      "epoch": 1.8110300081103001,
+      "grad_norm": 0.014772728085517883,
       "learning_rate": 0.0002,
-      "loss": 0.5739820003509521,
-      "mean_token_accuracy": 0.7669774144887924,
-      "num_tokens": 21914036.0,
+      "loss": 0.6175127625465393,
+      "mean_token_accuracy": 0.7471248880028725,
+      "num_tokens": 10404586.0,
       "step": 280
     },
     {
-      "entropy": 0.5947683751583099,
-      "epoch": 5.205607476635514,
-      "grad_norm": 0.02552945725619793,
+      "entropy": 0.6203941032290459,
+      "epoch": 1.8175182481751824,
+      "grad_norm": 0.00933381449431181,
       "learning_rate": 0.0002,
-      "loss": 0.5868436694145203,
-      "mean_token_accuracy": 0.7614100873470306,
-      "num_tokens": 21993146.0,
+      "loss": 0.6234451532363892,
+      "mean_token_accuracy": 0.7449408918619156,
+      "num_tokens": 10441660.0,
       "step": 281
     },
     {
-      "entropy": 0.5864996314048767,
-      "epoch": 5.224299065420561,
-      "grad_norm": 0.02433052286505699,
+      "entropy": 0.6171698048710823,
+      "epoch": 1.8240064882400648,
+      "grad_norm": 0.02071473002433777,
       "learning_rate": 0.0002,
-      "loss": 0.5833629369735718,
-      "mean_token_accuracy": 0.7621068209409714,
-      "num_tokens": 22070765.0,
+      "loss": 0.6131167411804199,
+      "mean_token_accuracy": 0.7504094913601875,
+      "num_tokens": 10478740.0,
       "step": 282
     },
     {
-      "entropy": 0.5739980936050415,
-      "epoch": 5.242990654205608,
-      "grad_norm": 0.03387480974197388,
+      "entropy": 0.6204886436462402,
+      "epoch": 1.8304947283049473,
+      "grad_norm": 0.008232798427343369,
       "learning_rate": 0.0002,
-      "loss": 0.5857794880867004,
-      "mean_token_accuracy": 0.7629742175340652,
-      "num_tokens": 22149016.0,
+      "loss": 0.6195346117019653,
+      "mean_token_accuracy": 0.7477771490812302,
+      "num_tokens": 10515739.0,
       "step": 283
     },
     {
-      "entropy": 0.5784127712249756,
-      "epoch": 5.261682242990654,
-      "grad_norm": 0.02454730123281479,
+      "entropy": 0.6222864389419556,
+      "epoch": 1.8369829683698295,
+      "grad_norm": 0.008826926350593567,
       "learning_rate": 0.0002,
-      "loss": 0.5814563632011414,
-      "mean_token_accuracy": 0.7645214051008224,
-      "num_tokens": 22228077.0,
+      "loss": 0.6246115565299988,
+      "mean_token_accuracy": 0.7467887252569199,
+      "num_tokens": 10552926.0,
       "step": 284
     },
     {
-      "entropy": 0.5910583734512329,
-      "epoch": 5.280373831775701,
-      "grad_norm": 0.023027142509818077,
+      "entropy": 0.6199471354484558,
+      "epoch": 1.8434712084347122,
+      "grad_norm": 0.031130356714129448,
       "learning_rate": 0.0002,
-      "loss": 0.5844653844833374,
-      "mean_token_accuracy": 0.7618297934532166,
-      "num_tokens": 22308791.0,
+      "loss": 0.6234173774719238,
+      "mean_token_accuracy": 0.7465962916612625,
+      "num_tokens": 10590404.0,
       "step": 285
     },
     {
-      "entropy": 0.5918671190738678,
-      "epoch": 5.299065420560748,
-      "grad_norm": 0.028409529477357864,
+      "entropy": 0.6229867041110992,
+      "epoch": 1.8499594484995945,
+      "grad_norm": 0.009124074131250381,
       "learning_rate": 0.0002,
-      "loss": 0.583986759185791,
-      "mean_token_accuracy": 0.7624140679836273,
-      "num_tokens": 22388553.0,
+      "loss": 0.6230523586273193,
+      "mean_token_accuracy": 0.7485576421022415,
+      "num_tokens": 10628169.0,
       "step": 286
     },
     {
-      "entropy": 0.5927205830812454,
-      "epoch": 5.317757009345795,
-      "grad_norm": 0.020732548087835312,
+      "entropy": 0.6245164126157761,
+      "epoch": 1.8564476885644767,
+      "grad_norm": 0.04299340024590492,
       "learning_rate": 0.0002,
-      "loss": 0.5924677848815918,
-      "mean_token_accuracy": 0.7593077719211578,
-      "num_tokens": 22466926.0,
+      "loss": 0.6296451091766357,
+      "mean_token_accuracy": 0.746208980679512,
+      "num_tokens": 10665578.0,
       "step": 287
     },
     {
-      "entropy": 0.5829800367355347,
-      "epoch": 5.336448598130841,
-      "grad_norm": 0.03212112933397293,
+      "entropy": 0.6367667019367218,
+      "epoch": 1.8629359286293594,
+      "grad_norm": 0.008064360357820988,
       "learning_rate": 0.0002,
-      "loss": 0.5898263454437256,
-      "mean_token_accuracy": 0.7594864368438721,
-      "num_tokens": 22545473.0,
+      "loss": 0.6342769861221313,
+      "mean_token_accuracy": 0.7435217127203941,
+      "num_tokens": 10703123.0,
       "step": 288
     },
     {
-      "entropy": 0.5854717344045639,
-      "epoch": 5.355140186915888,
-      "grad_norm": 0.023481527343392372,
+      "entropy": 0.6245287135243416,
+      "epoch": 1.8694241686942417,
+      "grad_norm": 0.021351464092731476,
       "learning_rate": 0.0002,
-      "loss": 0.5866089463233948,
-      "mean_token_accuracy": 0.7612689137458801,
-      "num_tokens": 22624084.0,
+      "loss": 0.6260511875152588,
+      "mean_token_accuracy": 0.7465523779392242,
+      "num_tokens": 10740314.0,
       "step": 289
     },
     {
-      "entropy": 0.594966858625412,
-      "epoch": 5.373831775700935,
-      "grad_norm": 0.025457587093114853,
+      "entropy": 0.6212104111909866,
+      "epoch": 1.8759124087591241,
+      "grad_norm": 0.01059524156153202,
       "learning_rate": 0.0002,
-      "loss": 0.5871824026107788,
-      "mean_token_accuracy": 0.7623147368431091,
-      "num_tokens": 22702606.0,
+      "loss": 0.6210169792175293,
+      "mean_token_accuracy": 0.7475718706846237,
+      "num_tokens": 10777825.0,
       "step": 290
     },
     {
-      "entropy": 0.5951681286096573,
-      "epoch": 5.392523364485982,
-      "grad_norm": 0.02658168412744999,
+      "entropy": 0.6373793482780457,
+      "epoch": 1.8824006488240066,
+      "grad_norm": 0.008884789422154427,
       "learning_rate": 0.0002,
-      "loss": 0.5921801328659058,
-      "mean_token_accuracy": 0.7598440647125244,
-      "num_tokens": 22781840.0,
+      "loss": 0.6366032361984253,
+      "mean_token_accuracy": 0.7391576170921326,
+      "num_tokens": 10815225.0,
       "step": 291
     },
     {
-      "entropy": 0.5740289241075516,
-      "epoch": 5.411214953271028,
-      "grad_norm": 0.025105727836489677,
+      "entropy": 0.6392252072691917,
+      "epoch": 1.8888888888888888,
+      "grad_norm": 0.013435903936624527,
       "learning_rate": 0.0002,
-      "loss": 0.5768394470214844,
-      "mean_token_accuracy": 0.7659056484699249,
-      "num_tokens": 22861158.0,
+      "loss": 0.6392862796783447,
+      "mean_token_accuracy": 0.7395146489143372,
+      "num_tokens": 10852456.0,
       "step": 292
     },
     {
-      "entropy": 0.5927090644836426,
-      "epoch": 5.429906542056075,
-      "grad_norm": 0.027616171166300774,
+      "entropy": 0.6242144107818604,
+      "epoch": 1.8953771289537713,
+      "grad_norm": 0.009207301773130894,
       "learning_rate": 0.0002,
-      "loss": 0.5969903469085693,
-      "mean_token_accuracy": 0.7579035460948944,
-      "num_tokens": 22939419.0,
+      "loss": 0.6249897480010986,
+      "mean_token_accuracy": 0.7450198903679848,
+      "num_tokens": 10889534.0,
       "step": 293
     },
     {
-      "entropy": 0.5913174897432327,
-      "epoch": 5.4485981308411215,
-      "grad_norm": 0.023430990055203438,
+      "entropy": 0.6158905401825905,
+      "epoch": 1.9018653690186538,
+      "grad_norm": 0.009233659133315086,
       "learning_rate": 0.0002,
-      "loss": 0.584871768951416,
-      "mean_token_accuracy": 0.7618055790662766,
-      "num_tokens": 23018229.0,
+      "loss": 0.6220048069953918,
+      "mean_token_accuracy": 0.7474906519055367,
+      "num_tokens": 10927138.0,
       "step": 294
     },
     {
-      "entropy": 0.6013437658548355,
-      "epoch": 5.4672897196261685,
-      "grad_norm": 0.02779349498450756,
+      "entropy": 0.6057498604059219,
+      "epoch": 1.908353609083536,
+      "grad_norm": 0.00886959582567215,
       "learning_rate": 0.0002,
-      "loss": 0.596491813659668,
-      "mean_token_accuracy": 0.7574046552181244,
-      "num_tokens": 23097420.0,
+      "loss": 0.6085807085037231,
+      "mean_token_accuracy": 0.7527724504470825,
+      "num_tokens": 10964121.0,
       "step": 295
     },
     {
-      "entropy": 0.5900115221738815,
-      "epoch": 5.485981308411215,
-      "grad_norm": 0.02505963295698166,
+      "entropy": 0.6314896494150162,
+      "epoch": 1.9148418491484185,
+      "grad_norm": 0.00863397866487503,
       "learning_rate": 0.0002,
-      "loss": 0.5895775556564331,
-      "mean_token_accuracy": 0.760588601231575,
-      "num_tokens": 23176053.0,
+      "loss": 0.6325403451919556,
+      "mean_token_accuracy": 0.7428522482514381,
+      "num_tokens": 11001557.0,
       "step": 296
     },
     {
-      "entropy": 0.5766851752996445,
-      "epoch": 5.5046728971962615,
-      "grad_norm": 0.029174979776144028,
+      "entropy": 0.6221454739570618,
+      "epoch": 1.921330089213301,
+      "grad_norm": 0.009161150082945824,
       "learning_rate": 0.0002,
-      "loss": 0.5806823968887329,
-      "mean_token_accuracy": 0.7644506692886353,
-      "num_tokens": 23256066.0,
+      "loss": 0.6208068132400513,
+      "mean_token_accuracy": 0.7462965175509453,
+      "num_tokens": 11038904.0,
       "step": 297
     },
     {
-      "entropy": 0.5836054682731628,
-      "epoch": 5.5233644859813085,
-      "grad_norm": 0.023760344833135605,
+      "entropy": 0.6189481094479561,
+      "epoch": 1.9278183292781832,
+      "grad_norm": 0.008889259770512581,
       "learning_rate": 0.0002,
-      "loss": 0.5859720706939697,
-      "mean_token_accuracy": 0.7620792984962463,
-      "num_tokens": 23335050.0,
+      "loss": 0.6184496283531189,
+      "mean_token_accuracy": 0.7480250671505928,
+      "num_tokens": 11076150.0,
       "step": 298
     },
     {
-      "entropy": 0.5844072699546814,
-      "epoch": 5.542056074766355,
-      "grad_norm": 0.022240791469812393,
+      "entropy": 0.6238919943571091,
+      "epoch": 1.9343065693430657,
+      "grad_norm": 0.0090845488011837,
       "learning_rate": 0.0002,
-      "loss": 0.581206202507019,
-      "mean_token_accuracy": 0.7644935250282288,
-      "num_tokens": 23413527.0,
+      "loss": 0.6274781227111816,
+      "mean_token_accuracy": 0.744391493499279,
+      "num_tokens": 11113447.0,
       "step": 299
     },
     {
-      "entropy": 0.5935729742050171,
-      "epoch": 5.5607476635514015,
-      "grad_norm": 0.024417495355010033,
+      "entropy": 0.6231639310717583,
+      "epoch": 1.9407948094079481,
+      "grad_norm": 0.009156333282589912,
       "learning_rate": 0.0002,
-      "loss": 0.589938759803772,
-      "mean_token_accuracy": 0.7614579200744629,
-      "num_tokens": 23492631.0,
+      "loss": 0.6222721934318542,
+      "mean_token_accuracy": 0.7496197745203972,
+      "num_tokens": 11150750.0,
       "step": 300
     },
     {
-      "entropy": 0.5919747650623322,
-      "epoch": 5.579439252336448,
-      "grad_norm": 0.023943627253174782,
+      "entropy": 0.6238841712474823,
+      "epoch": 1.9472830494728304,
+      "grad_norm": 0.008916804566979408,
       "learning_rate": 0.0002,
-      "loss": 0.5906035304069519,
-      "mean_token_accuracy": 0.7593590617179871,
-      "num_tokens": 23570833.0,
+      "loss": 0.6307159066200256,
+      "mean_token_accuracy": 0.7432639226317406,
+      "num_tokens": 11187972.0,
       "step": 301
     },
     {
-      "entropy": 0.5831288248300552,
-      "epoch": 5.598130841121495,
-      "grad_norm": 0.025227652862668037,
+      "entropy": 0.628197155892849,
+      "epoch": 1.9537712895377128,
+      "grad_norm": 0.009915877133607864,
       "learning_rate": 0.0002,
-      "loss": 0.583688497543335,
-      "mean_token_accuracy": 0.7631240636110306,
-      "num_tokens": 23650408.0,
+      "loss": 0.6339118480682373,
+      "mean_token_accuracy": 0.741386704146862,
+      "num_tokens": 11225108.0,
       "step": 302
     },
     {
-      "entropy": 0.5717721283435822,
-      "epoch": 5.616822429906542,
-      "grad_norm": 0.021833831444382668,
+      "entropy": 0.6199254244565964,
+      "epoch": 1.9602595296025953,
+      "grad_norm": 0.008810635656118393,
       "learning_rate": 0.0002,
-      "loss": 0.5739642381668091,
-      "mean_token_accuracy": 0.7671074420213699,
-      "num_tokens": 23729933.0,
+      "loss": 0.6184746026992798,
+      "mean_token_accuracy": 0.7482490092515945,
+      "num_tokens": 11262735.0,
       "step": 303
     },
     {
-      "entropy": 0.5738316625356674,
-      "epoch": 5.635514018691588,
-      "grad_norm": 0.027755092829465866,
+      "entropy": 0.6206595823168755,
+      "epoch": 1.9667477696674776,
+      "grad_norm": 0.008489571511745453,
       "learning_rate": 0.0002,
-      "loss": 0.5777672529220581,
-      "mean_token_accuracy": 0.7638746052980423,
-      "num_tokens": 23809992.0,
+      "loss": 0.6251119375228882,
+      "mean_token_accuracy": 0.7444824799895287,
+      "num_tokens": 11299837.0,
       "step": 304
     },
     {
-      "entropy": 0.5943573713302612,
-      "epoch": 5.654205607476635,
-      "grad_norm": 0.024508560076355934,
+      "entropy": 0.6231965124607086,
+      "epoch": 1.9732360097323602,
+      "grad_norm": 0.008548264391720295,
       "learning_rate": 0.0002,
-      "loss": 0.5914307236671448,
-      "mean_token_accuracy": 0.7582877576351166,
-      "num_tokens": 23888866.0,
+      "loss": 0.6258702874183655,
+      "mean_token_accuracy": 0.7451799884438515,
+      "num_tokens": 11337335.0,
       "step": 305
     },
     {
-      "entropy": 0.582805335521698,
-      "epoch": 5.672897196261682,
-      "grad_norm": 0.02408367209136486,
+      "entropy": 0.6233715191483498,
+      "epoch": 1.9797242497972425,
+      "grad_norm": 0.008984974585473537,
       "learning_rate": 0.0002,
-      "loss": 0.5777289867401123,
-      "mean_token_accuracy": 0.766036868095398,
-      "num_tokens": 23969206.0,
+      "loss": 0.6242036819458008,
+      "mean_token_accuracy": 0.7449030429124832,
+      "num_tokens": 11375037.0,
       "step": 306
     },
     {
-      "entropy": 0.5912429988384247,
-      "epoch": 5.691588785046729,
-      "grad_norm": 0.02204861491918564,
+      "entropy": 0.6145095080137253,
+      "epoch": 1.986212489862125,
+      "grad_norm": 0.008570129983127117,
       "learning_rate": 0.0002,
-      "loss": 0.5905025601387024,
-      "mean_token_accuracy": 0.7602925002574921,
-      "num_tokens": 24047777.0,
+      "loss": 0.6161348819732666,
+      "mean_token_accuracy": 0.7494730427861214,
+      "num_tokens": 11412372.0,
       "step": 307
     },
     {
-      "entropy": 0.582305058836937,
-      "epoch": 5.710280373831775,
-      "grad_norm": 0.02204098179936409,
+      "entropy": 0.6227882578969002,
+      "epoch": 1.9927007299270074,
+      "grad_norm": 0.009132519364356995,
       "learning_rate": 0.0002,
-      "loss": 0.5845706462860107,
-      "mean_token_accuracy": 0.762296050786972,
-      "num_tokens": 24125999.0,
+      "loss": 0.6249089241027832,
+      "mean_token_accuracy": 0.7440183088183403,
+      "num_tokens": 11449566.0,
       "step": 308
     },
     {
-      "entropy": 0.5866653472185135,
-      "epoch": 5.728971962616822,
-      "grad_norm": 0.025352345779538155,
+      "entropy": 0.6210556477308273,
+      "epoch": 1.9991889699918897,
+      "grad_norm": 0.007969100028276443,
       "learning_rate": 0.0002,
-      "loss": 0.5904455184936523,
-      "mean_token_accuracy": 0.7604480236768723,
-      "num_tokens": 24204762.0,
+      "loss": 0.6244759559631348,
+      "mean_token_accuracy": 0.746830090880394,
+      "num_tokens": 11487152.0,
       "step": 309
     },
     {
-      "entropy": 0.5811505317687988,
-      "epoch": 5.747663551401869,
-      "grad_norm": 0.02347879484295845,
+      "entropy": 0.6104058027267456,
+      "epoch": 2.0,
+      "grad_norm": 0.019919538870453835,
       "learning_rate": 0.0002,
-      "loss": 0.5761972665786743,
-      "mean_token_accuracy": 0.766374871134758,
-      "num_tokens": 24283469.0,
+      "loss": 0.6209793090820312,
+      "mean_token_accuracy": 0.7482735514640808,
+      "num_tokens": 11491673.0,
       "step": 310
     },
     {
-      "entropy": 0.5975845456123352,
-      "epoch": 5.766355140186916,
-      "grad_norm": 0.022250916808843613,
+      "entropy": 0.6369761228561401,
+      "epoch": 2.0064882400648822,
+      "grad_norm": 0.011179310269653797,
       "learning_rate": 0.0002,
-      "loss": 0.5930895805358887,
-      "mean_token_accuracy": 0.7588881254196167,
-      "num_tokens": 24362733.0,
+      "loss": 0.6280978918075562,
+      "mean_token_accuracy": 0.7433952018618584,
+      "num_tokens": 11528991.0,
       "step": 311
     },
     {
-      "entropy": 0.5830185413360596,
-      "epoch": 5.785046728971962,
-      "grad_norm": 0.02017797715961933,
+      "entropy": 0.6231134757399559,
+      "epoch": 2.012976480129765,
+      "grad_norm": 0.012395660392940044,
       "learning_rate": 0.0002,
-      "loss": 0.5834226608276367,
-      "mean_token_accuracy": 0.7631048262119293,
-      "num_tokens": 24441132.0,
+      "loss": 0.6168753504753113,
+      "mean_token_accuracy": 0.7469241917133331,
+      "num_tokens": 11566174.0,
       "step": 312
     },
     {
-      "entropy": 0.5777293294668198,
-      "epoch": 5.803738317757009,
-      "grad_norm": 0.026559006422758102,
+      "entropy": 0.6361245587468147,
+      "epoch": 2.019464720194647,
+      "grad_norm": 0.010151471942663193,
       "learning_rate": 0.0002,
-      "loss": 0.5827245712280273,
-      "mean_token_accuracy": 0.7643077671527863,
-      "num_tokens": 24519646.0,
+      "loss": 0.6396852731704712,
+      "mean_token_accuracy": 0.7390208393335342,
+      "num_tokens": 11603660.0,
       "step": 313
     },
     {
-      "entropy": 0.5818401873111725,
-      "epoch": 5.822429906542056,
-      "grad_norm": 0.025208717212080956,
+      "entropy": 0.6205078735947609,
+      "epoch": 2.0259529602595294,
+      "grad_norm": 0.011173305101692677,
       "learning_rate": 0.0002,
-      "loss": 0.5849472880363464,
-      "mean_token_accuracy": 0.7625377178192139,
-      "num_tokens": 24599240.0,
+      "loss": 0.6290929317474365,
+      "mean_token_accuracy": 0.7439283728599548,
+      "num_tokens": 11640777.0,
       "step": 314
     },
     {
-      "entropy": 0.578282043337822,
-      "epoch": 5.841121495327103,
-      "grad_norm": 0.021883705630898476,
+      "entropy": 0.6196452155709267,
+      "epoch": 2.032441200324412,
+      "grad_norm": 0.011216065846383572,
       "learning_rate": 0.0002,
-      "loss": 0.5731823444366455,
-      "mean_token_accuracy": 0.7665036618709564,
-      "num_tokens": 24678919.0,
+      "loss": 0.6270949840545654,
+      "mean_token_accuracy": 0.7424712553620338,
+      "num_tokens": 11677826.0,
       "step": 315
     },
     {
-      "entropy": 0.5882506668567657,
-      "epoch": 5.859813084112149,
-      "grad_norm": 0.027508528903126717,
+      "entropy": 0.6094615831971169,
+      "epoch": 2.0389294403892944,
+      "grad_norm": 0.009836558252573013,
       "learning_rate": 0.0002,
-      "loss": 0.5805738568305969,
-      "mean_token_accuracy": 0.7643076032400131,
-      "num_tokens": 24759765.0,
+      "loss": 0.6110177040100098,
+      "mean_token_accuracy": 0.7522123828530312,
+      "num_tokens": 11714782.0,
       "step": 316
     },
     {
-      "entropy": 0.5644465982913971,
-      "epoch": 5.878504672897196,
-      "grad_norm": 0.02333572506904602,
+      "entropy": 0.6228377372026443,
+      "epoch": 2.0454176804541766,
+      "grad_norm": 0.009827308356761932,
       "learning_rate": 0.0002,
-      "loss": 0.5681304931640625,
-      "mean_token_accuracy": 0.768625870347023,
-      "num_tokens": 24838278.0,
+      "loss": 0.6198099851608276,
+      "mean_token_accuracy": 0.7482241690158844,
+      "num_tokens": 11751901.0,
       "step": 317
     },
     {
-      "entropy": 0.571717843413353,
-      "epoch": 5.897196261682243,
-      "grad_norm": 0.027057446539402008,
+      "entropy": 0.6236361563205719,
+      "epoch": 2.0519059205190593,
+      "grad_norm": 0.010331663303077221,
       "learning_rate": 0.0002,
-      "loss": 0.5760421752929688,
-      "mean_token_accuracy": 0.7669651806354523,
-      "num_tokens": 24916847.0,
+      "loss": 0.6171518564224243,
+      "mean_token_accuracy": 0.747296430170536,
+      "num_tokens": 11789727.0,
       "step": 318
     },
     {
-      "entropy": 0.5824219137430191,
-      "epoch": 5.91588785046729,
-      "grad_norm": 0.024362891912460327,
+      "entropy": 0.6258771196007729,
+      "epoch": 2.0583941605839415,
+      "grad_norm": 0.010642528533935547,
       "learning_rate": 0.0002,
-      "loss": 0.5817481279373169,
-      "mean_token_accuracy": 0.7647458910942078,
-      "num_tokens": 24996241.0,
+      "loss": 0.6271529197692871,
+      "mean_token_accuracy": 0.7447842955589294,
+      "num_tokens": 11826898.0,
       "step": 319
     },
     {
-      "entropy": 0.5849971771240234,
-      "epoch": 5.934579439252336,
-      "grad_norm": 0.02463519386947155,
+      "entropy": 0.6150587797164917,
+      "epoch": 2.0648824006488242,
+      "grad_norm": 0.009091152809560299,
       "learning_rate": 0.0002,
-      "loss": 0.582144558429718,
-      "mean_token_accuracy": 0.764634981751442,
-      "num_tokens": 25076927.0,
+      "loss": 0.6166321039199829,
+      "mean_token_accuracy": 0.7505485117435455,
+      "num_tokens": 11863842.0,
       "step": 320
     },
     {
-      "entropy": 0.6033434122800827,
-      "epoch": 5.953271028037383,
-      "grad_norm": 0.02771107293665409,
+      "entropy": 0.6226691231131554,
+      "epoch": 2.0713706407137065,
+      "grad_norm": 0.008850189857184887,
       "learning_rate": 0.0002,
-      "loss": 0.599830150604248,
-      "mean_token_accuracy": 0.7559361606836319,
-      "num_tokens": 25155040.0,
+      "loss": 0.6213895082473755,
+      "mean_token_accuracy": 0.7468440607190132,
+      "num_tokens": 11901235.0,
       "step": 321
     },
     {
-      "entropy": 0.581831768155098,
-      "epoch": 5.97196261682243,
-      "grad_norm": 0.02171648107469082,
+      "entropy": 0.634496659040451,
+      "epoch": 2.0778588807785887,
+      "grad_norm": 0.009685751050710678,
       "learning_rate": 0.0002,
-      "loss": 0.5809007883071899,
-      "mean_token_accuracy": 0.7641598880290985,
-      "num_tokens": 25233942.0,
+      "loss": 0.6369626522064209,
+      "mean_token_accuracy": 0.7408950179815292,
+      "num_tokens": 11938336.0,
       "step": 322
     },
     {
-      "entropy": 0.5758271366357803,
-      "epoch": 5.990654205607477,
-      "grad_norm": 0.02509225904941559,
+      "entropy": 0.6164573132991791,
+      "epoch": 2.0843471208434714,
+      "grad_norm": 0.010890510864555836,
       "learning_rate": 0.0002,
-      "loss": 0.5794141292572021,
-      "mean_token_accuracy": 0.7650892734527588,
-      "num_tokens": 25312004.0,
+      "loss": 0.6235396265983582,
+      "mean_token_accuracy": 0.7454954236745834,
+      "num_tokens": 11975368.0,
       "step": 323
     },
     {
-      "entropy": 0.5852536261081696,
-      "epoch": 6.0,
-      "grad_norm": 0.031556036323308945,
+      "entropy": 0.6137611418962479,
+      "epoch": 2.0908353609083536,
+      "grad_norm": 0.009321420453488827,
       "learning_rate": 0.0002,
-      "loss": 0.5901615023612976,
-      "mean_token_accuracy": 0.7596581876277924,
-      "num_tokens": 25351251.0,
+      "loss": 0.6175767183303833,
+      "mean_token_accuracy": 0.7478668764233589,
+      "num_tokens": 12012459.0,
       "step": 324
     },
     {
-      "entropy": 0.5906165093183517,
-      "epoch": 6.018691588785047,
-      "grad_norm": 0.021236663684248924,
+      "entropy": 0.6179320439696312,
+      "epoch": 2.097323600973236,
+      "grad_norm": 0.009330744855105877,
       "learning_rate": 0.0002,
-      "loss": 0.5863031148910522,
-      "mean_token_accuracy": 0.7612903267145157,
-      "num_tokens": 25429823.0,
+      "loss": 0.6184362173080444,
+      "mean_token_accuracy": 0.748590737581253,
+      "num_tokens": 12049829.0,
       "step": 325
     },
     {
-      "entropy": 0.5902242064476013,
-      "epoch": 6.037383177570093,
-      "grad_norm": 0.02503519505262375,
+      "entropy": 0.6192868947982788,
+      "epoch": 2.1038118410381186,
+      "grad_norm": 0.008818828500807285,
       "learning_rate": 0.0002,
-      "loss": 0.5857101678848267,
-      "mean_token_accuracy": 0.7612328678369522,
-      "num_tokens": 25509032.0,
+      "loss": 0.6174098253250122,
+      "mean_token_accuracy": 0.7478644922375679,
+      "num_tokens": 12087727.0,
       "step": 326
     },
     {
-      "entropy": 0.5882705599069595,
-      "epoch": 6.05607476635514,
-      "grad_norm": 0.022178243845701218,
+      "entropy": 0.626248262822628,
+      "epoch": 2.110300081103001,
+      "grad_norm": 0.008749144151806831,
       "learning_rate": 0.0002,
-      "loss": 0.5829363465309143,
-      "mean_token_accuracy": 0.7635704278945923,
-      "num_tokens": 25588440.0,
+      "loss": 0.6212329268455505,
+      "mean_token_accuracy": 0.747315414249897,
+      "num_tokens": 12124835.0,
       "step": 327
     },
     {
-      "entropy": 0.5852685421705246,
-      "epoch": 6.074766355140187,
-      "grad_norm": 0.02558092586696148,
+      "entropy": 0.625133216381073,
+      "epoch": 2.116788321167883,
+      "grad_norm": 0.009766868315637112,
       "learning_rate": 0.0002,
-      "loss": 0.5829890370368958,
-      "mean_token_accuracy": 0.7630147635936737,
-      "num_tokens": 25666862.0,
+      "loss": 0.624916136264801,
+      "mean_token_accuracy": 0.7457389757037163,
+      "num_tokens": 12162010.0,
       "step": 328
     },
     {
-      "entropy": 0.5755240172147751,
-      "epoch": 6.093457943925234,
-      "grad_norm": 0.03008244000375271,
+      "entropy": 0.6180580258369446,
+      "epoch": 2.1232765612327658,
+      "grad_norm": 0.010169712826609612,
       "learning_rate": 0.0002,
-      "loss": 0.5809780955314636,
-      "mean_token_accuracy": 0.7630897909402847,
-      "num_tokens": 25745492.0,
+      "loss": 0.6185243129730225,
+      "mean_token_accuracy": 0.7475102469325066,
+      "num_tokens": 12199560.0,
       "step": 329
     },
     {
-      "entropy": 0.592608630657196,
-      "epoch": 6.11214953271028,
-      "grad_norm": 0.02237210050225258,
+      "entropy": 0.618090882897377,
+      "epoch": 2.129764801297648,
+      "grad_norm": 0.010965410619974136,
       "learning_rate": 0.0002,
-      "loss": 0.5901657342910767,
-      "mean_token_accuracy": 0.7589312344789505,
-      "num_tokens": 25824245.0,
+      "loss": 0.628595232963562,
+      "mean_token_accuracy": 0.742675431072712,
+      "num_tokens": 12236987.0,
       "step": 330
     },
     {
-      "entropy": 0.5749453604221344,
-      "epoch": 6.130841121495327,
-      "grad_norm": 0.024231985211372375,
+      "entropy": 0.6243258565664291,
+      "epoch": 2.1362530413625302,
+      "grad_norm": 0.009334998205304146,
       "learning_rate": 0.0002,
-      "loss": 0.5709967613220215,
-      "mean_token_accuracy": 0.7677164077758789,
-      "num_tokens": 25903416.0,
+      "loss": 0.6280615329742432,
+      "mean_token_accuracy": 0.7434326112270355,
+      "num_tokens": 12274464.0,
       "step": 331
     },
     {
-      "entropy": 0.5901312232017517,
-      "epoch": 6.149532710280374,
-      "grad_norm": 0.025874990969896317,
+      "entropy": 0.6221758723258972,
+      "epoch": 2.142741281427413,
+      "grad_norm": 0.00768797192722559,
       "learning_rate": 0.0002,
-      "loss": 0.59079909324646,
-      "mean_token_accuracy": 0.7597486227750778,
-      "num_tokens": 25981127.0,
+      "loss": 0.6218302845954895,
+      "mean_token_accuracy": 0.7480092272162437,
+      "num_tokens": 12312027.0,
       "step": 332
     },
     {
-      "entropy": 0.5772260576486588,
-      "epoch": 6.168224299065421,
-      "grad_norm": 0.031823720782995224,
+      "entropy": 0.6239754110574722,
+      "epoch": 2.149229521492295,
+      "grad_norm": 0.007848316803574562,
       "learning_rate": 0.0002,
-      "loss": 0.585727334022522,
-      "mean_token_accuracy": 0.7610679566860199,
-      "num_tokens": 26060209.0,
+      "loss": 0.6233847141265869,
+      "mean_token_accuracy": 0.7466081529855728,
+      "num_tokens": 12349207.0,
       "step": 333
     },
     {
-      "entropy": 0.5794196575880051,
-      "epoch": 6.186915887850467,
-      "grad_norm": 0.02636229805648327,
+      "entropy": 0.6159957125782967,
+      "epoch": 2.1557177615571774,
+      "grad_norm": 0.009600339457392693,
       "learning_rate": 0.0002,
-      "loss": 0.576323390007019,
-      "mean_token_accuracy": 0.7644224166870117,
-      "num_tokens": 26138697.0,
+      "loss": 0.6117550134658813,
+      "mean_token_accuracy": 0.7508383393287659,
+      "num_tokens": 12386312.0,
       "step": 334
     },
     {
-      "entropy": 0.5810049474239349,
-      "epoch": 6.205607476635514,
-      "grad_norm": 0.02574821375310421,
+      "entropy": 0.6285756379365921,
+      "epoch": 2.16220600162206,
+      "grad_norm": 0.009564606472849846,
       "learning_rate": 0.0002,
-      "loss": 0.5759303569793701,
-      "mean_token_accuracy": 0.7661094963550568,
-      "num_tokens": 26217081.0,
+      "loss": 0.6304960250854492,
+      "mean_token_accuracy": 0.7436012774705887,
+      "num_tokens": 12423969.0,
       "step": 335
     },
     {
-      "entropy": 0.577182799577713,
-      "epoch": 6.224299065420561,
-      "grad_norm": 0.02286566235125065,
+      "entropy": 0.6194324493408203,
+      "epoch": 2.1686942416869424,
+      "grad_norm": 0.009417028166353703,
       "learning_rate": 0.0002,
-      "loss": 0.5750881433486938,
-      "mean_token_accuracy": 0.7653518319129944,
-      "num_tokens": 26295788.0,
+      "loss": 0.622604250907898,
+      "mean_token_accuracy": 0.7460262328386307,
+      "num_tokens": 12461498.0,
       "step": 336
     },
     {
-      "entropy": 0.5861713141202927,
-      "epoch": 6.242990654205608,
-      "grad_norm": 0.028622545301914215,
+      "entropy": 0.6123020723462105,
+      "epoch": 2.1751824817518246,
+      "grad_norm": 0.00952741876244545,
       "learning_rate": 0.0002,
-      "loss": 0.5886875987052917,
-      "mean_token_accuracy": 0.7594505399465561,
-      "num_tokens": 26375047.0,
+      "loss": 0.6190693378448486,
+      "mean_token_accuracy": 0.7473057582974434,
+      "num_tokens": 12498772.0,
       "step": 337
     },
     {
-      "entropy": 0.5752665996551514,
-      "epoch": 6.261682242990654,
-      "grad_norm": 0.023522673174738884,
+      "entropy": 0.6101055964827538,
+      "epoch": 2.1816707218167073,
+      "grad_norm": 0.00931888073682785,
       "learning_rate": 0.0002,
-      "loss": 0.5748910903930664,
-      "mean_token_accuracy": 0.766600638628006,
-      "num_tokens": 26453546.0,
+      "loss": 0.6161482334136963,
+      "mean_token_accuracy": 0.7476179748773575,
+      "num_tokens": 12535894.0,
       "step": 338
     },
     {
-      "entropy": 0.5716692358255386,
-      "epoch": 6.280373831775701,
-      "grad_norm": 0.025497497990727425,
+      "entropy": 0.6213592365384102,
+      "epoch": 2.1881589618815895,
+      "grad_norm": 0.008865389041602612,
       "learning_rate": 0.0002,
-      "loss": 0.5707880258560181,
-      "mean_token_accuracy": 0.7675977051258087,
-      "num_tokens": 26532437.0,
+      "loss": 0.6219321489334106,
+      "mean_token_accuracy": 0.7449849396944046,
+      "num_tokens": 12573498.0,
       "step": 339
     },
     {
-      "entropy": 0.5823859572410583,
-      "epoch": 6.299065420560748,
-      "grad_norm": 0.02345922775566578,
+      "entropy": 0.614892452955246,
+      "epoch": 2.1946472019464722,
+      "grad_norm": 0.009812900796532631,
       "learning_rate": 0.0002,
-      "loss": 0.5806757211685181,
-      "mean_token_accuracy": 0.7633695006370544,
-      "num_tokens": 26611054.0,
+      "loss": 0.6124197840690613,
+      "mean_token_accuracy": 0.7499328628182411,
+      "num_tokens": 12610465.0,
       "step": 340
     },
     {
-      "entropy": 0.577152207493782,
-      "epoch": 6.317757009345795,
-      "grad_norm": 0.03670202195644379,
+      "entropy": 0.6235473528504372,
+      "epoch": 2.2011354420113545,
+      "grad_norm": 0.009034697897732258,
       "learning_rate": 0.0002,
-      "loss": 0.5789405107498169,
-      "mean_token_accuracy": 0.7646550685167313,
-      "num_tokens": 26690138.0,
+      "loss": 0.6235278248786926,
+      "mean_token_accuracy": 0.7453176379203796,
+      "num_tokens": 12647446.0,
       "step": 341
     },
     {
-      "entropy": 0.5890039503574371,
-      "epoch": 6.336448598130841,
-      "grad_norm": 0.03328755870461464,
+      "entropy": 0.6155827194452286,
+      "epoch": 2.2076236820762367,
+      "grad_norm": 0.011849821545183659,
       "learning_rate": 0.0002,
-      "loss": 0.588995635509491,
-      "mean_token_accuracy": 0.7606080025434494,
-      "num_tokens": 26769925.0,
+      "loss": 0.6141517758369446,
+      "mean_token_accuracy": 0.7493102103471756,
+      "num_tokens": 12684339.0,
       "step": 342
     },
     {
-      "entropy": 0.5640043765306473,
-      "epoch": 6.355140186915888,
-      "grad_norm": 0.027529461309313774,
+      "entropy": 0.6220799088478088,
+      "epoch": 2.2141119221411194,
+      "grad_norm": 0.011021377518773079,
       "learning_rate": 0.0002,
-      "loss": 0.5621784925460815,
-      "mean_token_accuracy": 0.7716686129570007,
-      "num_tokens": 26849502.0,
+      "loss": 0.6232143640518188,
+      "mean_token_accuracy": 0.7435769438743591,
+      "num_tokens": 12722039.0,
       "step": 343
     },
     {
-      "entropy": 0.5703081637620926,
-      "epoch": 6.373831775700935,
-      "grad_norm": 0.029232876375317574,
+      "entropy": 0.60894425958395,
+      "epoch": 2.2206001622060016,
+      "grad_norm": 0.00959057454019785,
       "learning_rate": 0.0002,
-      "loss": 0.5690990090370178,
-      "mean_token_accuracy": 0.7682793736457825,
-      "num_tokens": 26928682.0,
+      "loss": 0.614958643913269,
+      "mean_token_accuracy": 0.7492904141545296,
+      "num_tokens": 12758559.0,
       "step": 344
     },
     {
-      "entropy": 0.5833006352186203,
-      "epoch": 6.392523364485982,
-      "grad_norm": 0.029239878058433533,
+      "entropy": 0.61026880890131,
+      "epoch": 2.227088402270884,
+      "grad_norm": 0.010138140991330147,
       "learning_rate": 0.0002,
-      "loss": 0.5839906930923462,
-      "mean_token_accuracy": 0.7636394202709198,
-      "num_tokens": 27007594.0,
+      "loss": 0.616437554359436,
+      "mean_token_accuracy": 0.7471959367394447,
+      "num_tokens": 12795759.0,
       "step": 345
     },
     {
-      "entropy": 0.5740661174058914,
-      "epoch": 6.411214953271028,
-      "grad_norm": 0.03573233261704445,
+      "entropy": 0.6212582364678383,
+      "epoch": 2.2335766423357666,
+      "grad_norm": 0.013862723484635353,
       "learning_rate": 0.0002,
-      "loss": 0.5750782489776611,
-      "mean_token_accuracy": 0.7659066617488861,
-      "num_tokens": 27086262.0,
+      "loss": 0.6215099692344666,
+      "mean_token_accuracy": 0.7478090971708298,
+      "num_tokens": 12833075.0,
       "step": 346
     },
     {
-      "entropy": 0.5843356847763062,
-      "epoch": 6.429906542056075,
-      "grad_norm": 0.03266820311546326,
+      "entropy": 0.6183774024248123,
+      "epoch": 2.240064882400649,
+      "grad_norm": 0.00895402580499649,
       "learning_rate": 0.0002,
-      "loss": 0.5765944719314575,
-      "mean_token_accuracy": 0.7660123109817505,
-      "num_tokens": 27166205.0,
+      "loss": 0.6187215447425842,
+      "mean_token_accuracy": 0.7494830936193466,
+      "num_tokens": 12870571.0,
       "step": 347
     },
     {
-      "entropy": 0.5712840706110001,
-      "epoch": 6.4485981308411215,
-      "grad_norm": 0.028312938287854195,
+      "entropy": 0.6210155412554741,
+      "epoch": 2.246553122465531,
+      "grad_norm": 0.041780415922403336,
       "learning_rate": 0.0002,
-      "loss": 0.5677918791770935,
-      "mean_token_accuracy": 0.7700672447681427,
-      "num_tokens": 27245542.0,
+      "loss": 0.6219582557678223,
+      "mean_token_accuracy": 0.7483534067869186,
+      "num_tokens": 12908163.0,
       "step": 348
     },
     {
-      "entropy": 0.5784569084644318,
-      "epoch": 6.4672897196261685,
-      "grad_norm": 0.03176464885473251,
+      "entropy": 0.6203203722834587,
+      "epoch": 2.2530413625304138,
+      "grad_norm": 0.02689778059720993,
       "learning_rate": 0.0002,
-      "loss": 0.5817923545837402,
-      "mean_token_accuracy": 0.7636367082595825,
-      "num_tokens": 27324606.0,
+      "loss": 0.6272082328796387,
+      "mean_token_accuracy": 0.7450312674045563,
+      "num_tokens": 12945330.0,
       "step": 349
     },
     {
-      "entropy": 0.58133964240551,
-      "epoch": 6.485981308411215,
-      "grad_norm": 0.02975059486925602,
+      "entropy": 0.6245186030864716,
+      "epoch": 2.259529602595296,
+      "grad_norm": 0.012432226911187172,
       "learning_rate": 0.0002,
-      "loss": 0.5827153325080872,
-      "mean_token_accuracy": 0.7636205703020096,
-      "num_tokens": 27403022.0,
+      "loss": 0.6235542297363281,
+      "mean_token_accuracy": 0.7462796568870544,
+      "num_tokens": 12982731.0,
       "step": 350
     },
     {
-      "entropy": 0.5791384875774384,
-      "epoch": 6.5046728971962615,
-      "grad_norm": 0.037513330578804016,
+      "entropy": 0.6220052242279053,
+      "epoch": 2.2660178426601782,
+      "grad_norm": 0.010308763943612576,
       "learning_rate": 0.0002,
-      "loss": 0.5764743089675903,
-      "mean_token_accuracy": 0.7667028158903122,
-      "num_tokens": 27482564.0,
+      "loss": 0.624070405960083,
+      "mean_token_accuracy": 0.743579313158989,
+      "num_tokens": 13019541.0,
       "step": 351
     },
     {
-      "entropy": 0.572890043258667,
-      "epoch": 6.5233644859813085,
-      "grad_norm": 0.028714006766676903,
+      "entropy": 0.64016043394804,
+      "epoch": 2.272506082725061,
+      "grad_norm": 0.010659064166247845,
       "learning_rate": 0.0002,
-      "loss": 0.5721226930618286,
-      "mean_token_accuracy": 0.7661556899547577,
-      "num_tokens": 27561557.0,
+      "loss": 0.636542558670044,
+      "mean_token_accuracy": 0.7407100200653076,
+      "num_tokens": 13057158.0,
       "step": 352
     },
     {
-      "entropy": 0.5851928144693375,
-      "epoch": 6.542056074766355,
-      "grad_norm": 0.04256724938750267,
+      "entropy": 0.6278761327266693,
+      "epoch": 2.278994322789943,
+      "grad_norm": 0.010416081175208092,
       "learning_rate": 0.0002,
-      "loss": 0.5871050357818604,
-      "mean_token_accuracy": 0.762816771864891,
-      "num_tokens": 27639557.0,
+      "loss": 0.62436842918396,
+      "mean_token_accuracy": 0.7455000206828117,
+      "num_tokens": 13094262.0,
       "step": 353
     },
     {
-      "entropy": 0.5750937461853027,
-      "epoch": 6.5607476635514015,
-      "grad_norm": 0.030066121369600296,
+      "entropy": 0.6232447475194931,
+      "epoch": 2.285482562854826,
+      "grad_norm": 0.009887438267469406,
       "learning_rate": 0.0002,
-      "loss": 0.5751978754997253,
-      "mean_token_accuracy": 0.7659846991300583,
-      "num_tokens": 27718059.0,
+      "loss": 0.6241923570632935,
+      "mean_token_accuracy": 0.7437829151749611,
+      "num_tokens": 13131360.0,
       "step": 354
     },
     {
-      "entropy": 0.567540854215622,
-      "epoch": 6.579439252336448,
-      "grad_norm": 0.04286859929561615,
+      "entropy": 0.6319752037525177,
+      "epoch": 2.291970802919708,
+      "grad_norm": 0.009396783076226711,
       "learning_rate": 0.0002,
-      "loss": 0.57082200050354,
-      "mean_token_accuracy": 0.7673868238925934,
-      "num_tokens": 27797774.0,
+      "loss": 0.6283852458000183,
+      "mean_token_accuracy": 0.7450494468212128,
+      "num_tokens": 13168644.0,
       "step": 355
     },
     {
-      "entropy": 0.5699326694011688,
-      "epoch": 6.598130841121495,
-      "grad_norm": 0.03684304282069206,
+      "entropy": 0.6192774921655655,
+      "epoch": 2.2984590429845904,
+      "grad_norm": 0.009457712061703205,
       "learning_rate": 0.0002,
-      "loss": 0.5703436136245728,
-      "mean_token_accuracy": 0.7681662887334824,
-      "num_tokens": 27877878.0,
+      "loss": 0.6166653633117676,
+      "mean_token_accuracy": 0.7473447695374489,
+      "num_tokens": 13205900.0,
       "step": 356
     },
     {
-      "entropy": 0.5923850387334824,
-      "epoch": 6.616822429906542,
-      "grad_norm": 0.04039483144879341,
+      "entropy": 0.616138719022274,
+      "epoch": 2.304947283049473,
+      "grad_norm": 0.009487454779446125,
       "learning_rate": 0.0002,
-      "loss": 0.5857964754104614,
-      "mean_token_accuracy": 0.7607691735029221,
-      "num_tokens": 27956164.0,
+      "loss": 0.6193556189537048,
+      "mean_token_accuracy": 0.746711365878582,
+      "num_tokens": 13243403.0,
       "step": 357
     },
     {
-      "entropy": 0.5853666663169861,
-      "epoch": 6.635514018691588,
-      "grad_norm": 0.034810423851013184,
+      "entropy": 0.6114000976085663,
+      "epoch": 2.3114355231143553,
+      "grad_norm": 0.010738314129412174,
       "learning_rate": 0.0002,
-      "loss": 0.5785076022148132,
-      "mean_token_accuracy": 0.7639980912208557,
-      "num_tokens": 28036636.0,
+      "loss": 0.6193853616714478,
+      "mean_token_accuracy": 0.7472727373242378,
+      "num_tokens": 13280595.0,
       "step": 358
     },
     {
-      "entropy": 0.5655460357666016,
-      "epoch": 6.654205607476635,
-      "grad_norm": 0.04020382836461067,
+      "entropy": 0.601548932492733,
+      "epoch": 2.3179237631792375,
+      "grad_norm": 0.010257442481815815,
       "learning_rate": 0.0002,
-      "loss": 0.5703827738761902,
-      "mean_token_accuracy": 0.767535388469696,
-      "num_tokens": 28114415.0,
+      "loss": 0.6057910919189453,
+      "mean_token_accuracy": 0.7506168782711029,
+      "num_tokens": 13317413.0,
       "step": 359
     },
     {
-      "entropy": 0.5866734981536865,
-      "epoch": 6.672897196261682,
-      "grad_norm": 0.03967415541410446,
+      "entropy": 0.6248240396380424,
+      "epoch": 2.3244120032441202,
+      "grad_norm": 0.009444781579077244,
       "learning_rate": 0.0002,
-      "loss": 0.5894461870193481,
-      "mean_token_accuracy": 0.7609697133302689,
-      "num_tokens": 28193663.0,
+      "loss": 0.6265978217124939,
+      "mean_token_accuracy": 0.7447159513831139,
+      "num_tokens": 13354442.0,
       "step": 360
     },
     {
-      "entropy": 0.59061299264431,
-      "epoch": 6.691588785046729,
-      "grad_norm": 0.041625600308179855,
+      "entropy": 0.6169225797057152,
+      "epoch": 2.3309002433090025,
+      "grad_norm": 0.010277079418301582,
       "learning_rate": 0.0002,
-      "loss": 0.5840169191360474,
-      "mean_token_accuracy": 0.763459637761116,
-      "num_tokens": 28272596.0,
+      "loss": 0.6173827648162842,
+      "mean_token_accuracy": 0.7486959248781204,
+      "num_tokens": 13392357.0,
       "step": 361
     },
     {
-      "entropy": 0.5873990952968597,
-      "epoch": 6.710280373831775,
-      "grad_norm": 0.03666547313332558,
+      "entropy": 0.6302815526723862,
+      "epoch": 2.3373884833738847,
+      "grad_norm": 0.009320929646492004,
       "learning_rate": 0.0002,
-      "loss": 0.581563413143158,
-      "mean_token_accuracy": 0.7643765062093735,
-      "num_tokens": 28350671.0,
+      "loss": 0.6289012432098389,
+      "mean_token_accuracy": 0.7438602596521378,
+      "num_tokens": 13429444.0,
       "step": 362
     },
     {
-      "entropy": 0.5714496225118637,
-      "epoch": 6.728971962616822,
-      "grad_norm": 0.035765308886766434,
+      "entropy": 0.6252695918083191,
+      "epoch": 2.3438767234387674,
+      "grad_norm": 0.00795667339116335,
       "learning_rate": 0.0002,
-      "loss": 0.5783935189247131,
-      "mean_token_accuracy": 0.7644345313310623,
-      "num_tokens": 28429350.0,
+      "loss": 0.6236270666122437,
+      "mean_token_accuracy": 0.7462895065546036,
+      "num_tokens": 13466973.0,
       "step": 363
     },
     {
-      "entropy": 0.5763610005378723,
-      "epoch": 6.747663551401869,
-      "grad_norm": 0.04010038077831268,
+      "entropy": 0.6325998455286026,
+      "epoch": 2.3503649635036497,
+      "grad_norm": 0.009790928103029728,
       "learning_rate": 0.0002,
-      "loss": 0.5831538438796997,
-      "mean_token_accuracy": 0.7639245241880417,
-      "num_tokens": 28507878.0,
+      "loss": 0.6316322088241577,
+      "mean_token_accuracy": 0.7428558766841888,
+      "num_tokens": 13504219.0,
       "step": 364
     },
     {
-      "entropy": 0.5884422659873962,
-      "epoch": 6.766355140186916,
-      "grad_norm": 0.0327216275036335,
+      "entropy": 0.6235339045524597,
+      "epoch": 2.356853203568532,
+      "grad_norm": 0.009260588325560093,
       "learning_rate": 0.0002,
-      "loss": 0.5833562016487122,
-      "mean_token_accuracy": 0.7639438211917877,
-      "num_tokens": 28586473.0,
+      "loss": 0.6231226921081543,
+      "mean_token_accuracy": 0.7452265918254852,
+      "num_tokens": 13541359.0,
       "step": 365
     },
     {
-      "entropy": 0.578765258193016,
-      "epoch": 6.785046728971962,
-      "grad_norm": 0.03694374859333038,
+      "entropy": 0.6141555160284042,
+      "epoch": 2.3633414436334146,
+      "grad_norm": 0.010724055580794811,
       "learning_rate": 0.0002,
-      "loss": 0.5703417658805847,
-      "mean_token_accuracy": 0.7693156599998474,
-      "num_tokens": 28666176.0,
+      "loss": 0.6217249631881714,
+      "mean_token_accuracy": 0.7470070049166679,
+      "num_tokens": 13578759.0,
       "step": 366
     },
     {
-      "entropy": 0.5803310126066208,
-      "epoch": 6.803738317757009,
-      "grad_norm": 0.030944958329200745,
+      "entropy": 0.6315039545297623,
+      "epoch": 2.369829683698297,
+      "grad_norm": 0.009067324921488762,
       "learning_rate": 0.0002,
-      "loss": 0.580040693283081,
-      "mean_token_accuracy": 0.7660168260335922,
-      "num_tokens": 28745637.0,
+      "loss": 0.634614109992981,
+      "mean_token_accuracy": 0.7410515621304512,
+      "num_tokens": 13616296.0,
       "step": 367
     },
     {
-      "entropy": 0.5808713585138321,
-      "epoch": 6.822429906542056,
-      "grad_norm": 0.03630272299051285,
+      "entropy": 0.6196957975625992,
+      "epoch": 2.376317923763179,
+      "grad_norm": 0.009248278103768826,
       "learning_rate": 0.0002,
-      "loss": 0.5874212980270386,
-      "mean_token_accuracy": 0.7598031908273697,
-      "num_tokens": 28822415.0,
+      "loss": 0.6189215183258057,
+      "mean_token_accuracy": 0.7481505498290062,
+      "num_tokens": 13653415.0,
       "step": 368
     },
     {
-      "entropy": 0.5853964239358902,
-      "epoch": 6.841121495327103,
-      "grad_norm": 0.032110489904880524,
+      "entropy": 0.6168980821967125,
+      "epoch": 2.3828061638280618,
+      "grad_norm": 0.010181935504078865,
       "learning_rate": 0.0002,
-      "loss": 0.586665153503418,
-      "mean_token_accuracy": 0.7622753381729126,
-      "num_tokens": 28900295.0,
+      "loss": 0.6161677837371826,
+      "mean_token_accuracy": 0.7499739304184914,
+      "num_tokens": 13690618.0,
       "step": 369
     },
     {
-      "entropy": 0.5945642292499542,
-      "epoch": 6.859813084112149,
-      "grad_norm": 0.026387203484773636,
+      "entropy": 0.6258252412080765,
+      "epoch": 2.389294403892944,
+      "grad_norm": 0.008928108029067516,
       "learning_rate": 0.0002,
-      "loss": 0.5878679752349854,
-      "mean_token_accuracy": 0.7614142149686813,
-      "num_tokens": 28979417.0,
+      "loss": 0.6290237903594971,
+      "mean_token_accuracy": 0.7429782524704933,
+      "num_tokens": 13727901.0,
       "step": 370
     },
     {
-      "entropy": 0.5783472508192062,
-      "epoch": 6.878504672897196,
-      "grad_norm": 0.031560156494379044,
+      "entropy": 0.6094075217843056,
+      "epoch": 2.3957826439578263,
+      "grad_norm": 0.009067908860743046,
       "learning_rate": 0.0002,
-      "loss": 0.5714523792266846,
-      "mean_token_accuracy": 0.7680636942386627,
-      "num_tokens": 29060256.0,
+      "loss": 0.6124044060707092,
+      "mean_token_accuracy": 0.7487504705786705,
+      "num_tokens": 13765097.0,
       "step": 371
     },
     {
-      "entropy": 0.5836873352527618,
-      "epoch": 6.897196261682243,
-      "grad_norm": 0.026852963492274284,
+      "entropy": 0.6190275996923447,
+      "epoch": 2.402270884022709,
+      "grad_norm": 0.009943855926394463,
       "learning_rate": 0.0002,
-      "loss": 0.5827068090438843,
-      "mean_token_accuracy": 0.7619643807411194,
-      "num_tokens": 29140811.0,
+      "loss": 0.6234393119812012,
+      "mean_token_accuracy": 0.7450670152902603,
+      "num_tokens": 13802604.0,
       "step": 372
     },
     {
-      "entropy": 0.5799173414707184,
-      "epoch": 6.91588785046729,
-      "grad_norm": 0.029336581006646156,
+      "entropy": 0.6125053241848946,
+      "epoch": 2.408759124087591,
+      "grad_norm": 0.009481463581323624,
       "learning_rate": 0.0002,
-      "loss": 0.5870332717895508,
-      "mean_token_accuracy": 0.7611389607191086,
-      "num_tokens": 29218919.0,
+      "loss": 0.6187013387680054,
+      "mean_token_accuracy": 0.7489439323544502,
+      "num_tokens": 13839487.0,
       "step": 373
     },
     {
-      "entropy": 0.578309491276741,
-      "epoch": 6.934579439252336,
-      "grad_norm": 0.030016575008630753,
+      "entropy": 0.6130451932549477,
+      "epoch": 2.4152473641524734,
+      "grad_norm": 0.008458724245429039,
       "learning_rate": 0.0002,
-      "loss": 0.5815063118934631,
-      "mean_token_accuracy": 0.7626699954271317,
-      "num_tokens": 29299051.0,
+      "loss": 0.614911675453186,
+      "mean_token_accuracy": 0.7505282834172249,
+      "num_tokens": 13876935.0,
       "step": 374
     },
     {
-      "entropy": 0.5871727913618088,
-      "epoch": 6.953271028037383,
-      "grad_norm": 0.026473497971892357,
+      "entropy": 0.6462682038545609,
+      "epoch": 2.421735604217356,
+      "grad_norm": 0.010083320550620556,
       "learning_rate": 0.0002,
-      "loss": 0.5825446248054504,
-      "mean_token_accuracy": 0.7642439305782318,
-      "num_tokens": 29378648.0,
+      "loss": 0.6467190980911255,
+      "mean_token_accuracy": 0.7378857955336571,
+      "num_tokens": 13914429.0,
       "step": 375
     },
     {
-      "entropy": 0.5949303209781647,
-      "epoch": 6.97196261682243,
-      "grad_norm": 0.029133882373571396,
+      "entropy": 0.6287035867571831,
+      "epoch": 2.4282238442822384,
+      "grad_norm": 0.008747600950300694,
       "learning_rate": 0.0002,
-      "loss": 0.5897645950317383,
-      "mean_token_accuracy": 0.7598071992397308,
-      "num_tokens": 29456791.0,
+      "loss": 0.6284276247024536,
+      "mean_token_accuracy": 0.7447932213544846,
+      "num_tokens": 13951880.0,
       "step": 376
     },
     {
-      "entropy": 0.5799039900302887,
-      "epoch": 6.990654205607477,
-      "grad_norm": 0.02209646999835968,
+      "entropy": 0.6305522173643112,
+      "epoch": 2.4347120843471206,
+      "grad_norm": 0.010268261656165123,
       "learning_rate": 0.0002,
-      "loss": 0.5738506317138672,
-      "mean_token_accuracy": 0.7665348201990128,
-      "num_tokens": 29535767.0,
+      "loss": 0.6298754215240479,
+      "mean_token_accuracy": 0.7451391518115997,
+      "num_tokens": 13989473.0,
       "step": 377
     },
     {
-      "entropy": 0.5705015957355499,
-      "epoch": 7.0,
-      "grad_norm": 0.03486085683107376,
+      "entropy": 0.6239131912589073,
+      "epoch": 2.4412003244120033,
+      "grad_norm": 0.00940751750022173,
       "learning_rate": 0.0002,
-      "loss": 0.5747893452644348,
-      "mean_token_accuracy": 0.7658974528312683,
-      "num_tokens": 29576698.0,
+      "loss": 0.6264931559562683,
+      "mean_token_accuracy": 0.7422467172145844,
+      "num_tokens": 14026832.0,
       "step": 378
     },
     {
-      "entropy": 0.5749023258686066,
-      "epoch": 7.018691588785047,
-      "grad_norm": 0.029213452711701393,
+      "entropy": 0.6218756884336472,
+      "epoch": 2.4476885644768855,
+      "grad_norm": 0.008956913836300373,
       "learning_rate": 0.0002,
-      "loss": 0.5739940404891968,
-      "mean_token_accuracy": 0.7652797847986221,
-      "num_tokens": 29655453.0,
+      "loss": 0.6260896325111389,
+      "mean_token_accuracy": 0.7415326088666916,
+      "num_tokens": 14064159.0,
       "step": 379
     },
     {
-      "entropy": 0.5714626312255859,
-      "epoch": 7.037383177570093,
-      "grad_norm": 0.023325830698013306,
+      "entropy": 0.6187591478228569,
+      "epoch": 2.4541768045417682,
+      "grad_norm": 0.008552344515919685,
       "learning_rate": 0.0002,
-      "loss": 0.5672348737716675,
-      "mean_token_accuracy": 0.7701223790645599,
-      "num_tokens": 29735726.0,
+      "loss": 0.6216272115707397,
+      "mean_token_accuracy": 0.7469553202390671,
+      "num_tokens": 14101277.0,
       "step": 380
     },
     {
-      "entropy": 0.5690907686948776,
-      "epoch": 7.05607476635514,
-      "grad_norm": 0.031354986131191254,
+      "entropy": 0.6129069477319717,
+      "epoch": 2.4606650446066505,
+      "grad_norm": 0.009866154752671719,
       "learning_rate": 0.0002,
-      "loss": 0.5650231838226318,
-      "mean_token_accuracy": 0.7709572166204453,
-      "num_tokens": 29814948.0,
+      "loss": 0.6184383630752563,
+      "mean_token_accuracy": 0.7457786872982979,
+      "num_tokens": 14138545.0,
       "step": 381
     },
     {
-      "entropy": 0.5705015808343887,
-      "epoch": 7.074766355140187,
-      "grad_norm": 0.023136666044592857,
+      "entropy": 0.6332196220755577,
+      "epoch": 2.4671532846715327,
+      "grad_norm": 0.007945130579173565,
       "learning_rate": 0.0002,
-      "loss": 0.5665818452835083,
-      "mean_token_accuracy": 0.7688614875078201,
-      "num_tokens": 29894457.0,
+      "loss": 0.6312772631645203,
+      "mean_token_accuracy": 0.7440632656216621,
+      "num_tokens": 14176229.0,
       "step": 382
     },
     {
-      "entropy": 0.5640405267477036,
-      "epoch": 7.093457943925234,
-      "grad_norm": 0.033995371311903,
+      "entropy": 0.6162554249167442,
+      "epoch": 2.4736415247364154,
+      "grad_norm": 0.00908781960606575,
       "learning_rate": 0.0002,
-      "loss": 0.5620932579040527,
-      "mean_token_accuracy": 0.7714024633169174,
-      "num_tokens": 29975233.0,
+      "loss": 0.6104109883308411,
+      "mean_token_accuracy": 0.7492068335413933,
+      "num_tokens": 14213375.0,
       "step": 383
     },
     {
-      "entropy": 0.5753348916769028,
-      "epoch": 7.11214953271028,
-      "grad_norm": 0.03603081405162811,
+      "entropy": 0.6241936013102531,
+      "epoch": 2.4801297648012977,
+      "grad_norm": 0.009399542585015297,
       "learning_rate": 0.0002,
-      "loss": 0.581680417060852,
-      "mean_token_accuracy": 0.7630927562713623,
-      "num_tokens": 30053632.0,
+      "loss": 0.6161544919013977,
+      "mean_token_accuracy": 0.7491407096385956,
+      "num_tokens": 14250696.0,
       "step": 384
     },
     {
-      "entropy": 0.5801223963499069,
-      "epoch": 7.130841121495327,
-      "grad_norm": 0.029232708737254143,
+      "entropy": 0.6147832497954369,
+      "epoch": 2.48661800486618,
+      "grad_norm": 0.007576434873044491,
       "learning_rate": 0.0002,
-      "loss": 0.5759778618812561,
-      "mean_token_accuracy": 0.764848455786705,
-      "num_tokens": 30132300.0,
+      "loss": 0.6159515976905823,
+      "mean_token_accuracy": 0.7494903281331062,
+      "num_tokens": 14287863.0,
       "step": 385
     },
     {
-      "entropy": 0.5866703689098358,
-      "epoch": 7.149532710280374,
-      "grad_norm": 0.03570999205112457,
+      "entropy": 0.6227882355451584,
+      "epoch": 2.4931062449310626,
+      "grad_norm": 0.009297952987253666,
       "learning_rate": 0.0002,
-      "loss": 0.5798372030258179,
-      "mean_token_accuracy": 0.763422429561615,
-      "num_tokens": 30210403.0,
+      "loss": 0.6279860734939575,
+      "mean_token_accuracy": 0.745159700512886,
+      "num_tokens": 14325371.0,
       "step": 386
     },
     {
-      "entropy": 0.5626991838216782,
-      "epoch": 7.168224299065421,
-      "grad_norm": 0.03066566027700901,
+      "entropy": 0.6066790446639061,
+      "epoch": 2.499594484995945,
+      "grad_norm": 0.009967000223696232,
       "learning_rate": 0.0002,
-      "loss": 0.5639806985855103,
-      "mean_token_accuracy": 0.7704606354236603,
-      "num_tokens": 30289334.0,
+      "loss": 0.6151248216629028,
+      "mean_token_accuracy": 0.7479714751243591,
+      "num_tokens": 14362698.0,
       "step": 387
     },
     {
-      "entropy": 0.5659942924976349,
-      "epoch": 7.186915887850467,
-      "grad_norm": 0.03829355165362358,
+      "entropy": 0.6248459070920944,
+      "epoch": 2.5060827250608275,
+      "grad_norm": 0.009727658703923225,
       "learning_rate": 0.0002,
-      "loss": 0.5711922645568848,
-      "mean_token_accuracy": 0.7667083889245987,
-      "num_tokens": 30367788.0,
+      "loss": 0.6268327236175537,
+      "mean_token_accuracy": 0.7463877573609352,
+      "num_tokens": 14399544.0,
       "step": 388
     },
     {
-      "entropy": 0.5780204236507416,
-      "epoch": 7.205607476635514,
-      "grad_norm": 0.026930075138807297,
+      "entropy": 0.6305608153343201,
+      "epoch": 2.5125709651257098,
+      "grad_norm": 0.008822647854685783,
       "learning_rate": 0.0002,
-      "loss": 0.5756471157073975,
-      "mean_token_accuracy": 0.7649676948785782,
-      "num_tokens": 30446365.0,
+      "loss": 0.6284365653991699,
+      "mean_token_accuracy": 0.7454124987125397,
+      "num_tokens": 14436877.0,
       "step": 389
     },
     {
-      "entropy": 0.5732643306255341,
-      "epoch": 7.224299065420561,
-      "grad_norm": 0.036685965955257416,
+      "entropy": 0.6289694681763649,
+      "epoch": 2.519059205190592,
+      "grad_norm": 0.009071394801139832,
       "learning_rate": 0.0002,
-      "loss": 0.5662911534309387,
-      "mean_token_accuracy": 0.7686476409435272,
-      "num_tokens": 30525763.0,
+      "loss": 0.6264450550079346,
+      "mean_token_accuracy": 0.7433068230748177,
+      "num_tokens": 14473915.0,
       "step": 390
     },
     {
-      "entropy": 0.5636216700077057,
-      "epoch": 7.242990654205608,
-      "grad_norm": 0.028072474524378777,
+      "entropy": 0.6271463260054588,
+      "epoch": 2.5255474452554747,
+      "grad_norm": 0.008187719620764256,
       "learning_rate": 0.0002,
-      "loss": 0.5616763830184937,
-      "mean_token_accuracy": 0.7721907496452332,
-      "num_tokens": 30606020.0,
+      "loss": 0.6289336085319519,
+      "mean_token_accuracy": 0.7423637583851814,
+      "num_tokens": 14511402.0,
       "step": 391
     },
     {
-      "entropy": 0.5680046379566193,
-      "epoch": 7.261682242990654,
-      "grad_norm": 0.037153203040361404,
+      "entropy": 0.6084585338830948,
+      "epoch": 2.532035685320357,
+      "grad_norm": 0.008899841457605362,
       "learning_rate": 0.0002,
-      "loss": 0.5728225708007812,
-      "mean_token_accuracy": 0.7663372904062271,
-      "num_tokens": 30684916.0,
+      "loss": 0.6130256652832031,
+      "mean_token_accuracy": 0.7509424611926079,
+      "num_tokens": 14548753.0,
       "step": 392
     },
     {
-      "entropy": 0.5744252055883408,
-      "epoch": 7.280373831775701,
-      "grad_norm": 0.02831830456852913,
+      "entropy": 0.6104612424969673,
+      "epoch": 2.538523925385239,
+      "grad_norm": 0.00985915120691061,
       "learning_rate": 0.0002,
-      "loss": 0.5753362774848938,
-      "mean_token_accuracy": 0.7671765983104706,
-      "num_tokens": 30763260.0,
+      "loss": 0.6200201511383057,
+      "mean_token_accuracy": 0.7466698288917542,
+      "num_tokens": 14586042.0,
       "step": 393
     },
     {
-      "entropy": 0.5881899744272232,
-      "epoch": 7.299065420560748,
-      "grad_norm": 0.03482871875166893,
+      "entropy": 0.6256568655371666,
+      "epoch": 2.545012165450122,
+      "grad_norm": 0.00867030955851078,
       "learning_rate": 0.0002,
-      "loss": 0.5832597613334656,
-      "mean_token_accuracy": 0.7622921317815781,
-      "num_tokens": 30841943.0,
+      "loss": 0.6292595267295837,
+      "mean_token_accuracy": 0.7430211529135704,
+      "num_tokens": 14623448.0,
       "step": 394
     },
     {
-      "entropy": 0.5729344636201859,
-      "epoch": 7.317757009345795,
-      "grad_norm": 0.03387274965643883,
+      "entropy": 0.6263261586427689,
+      "epoch": 2.551500405515004,
+      "grad_norm": 0.010737903416156769,
       "learning_rate": 0.0002,
-      "loss": 0.5740589499473572,
-      "mean_token_accuracy": 0.7642743289470673,
-      "num_tokens": 30920447.0,
+      "loss": 0.6249234676361084,
+      "mean_token_accuracy": 0.7444213852286339,
+      "num_tokens": 14660585.0,
       "step": 395
     },
     {
-      "entropy": 0.5637217462062836,
-      "epoch": 7.336448598130841,
-      "grad_norm": 0.028617210686206818,
+      "entropy": 0.6266406700015068,
+      "epoch": 2.5579886455798864,
+      "grad_norm": 0.008770007640123367,
       "learning_rate": 0.0002,
-      "loss": 0.5685332417488098,
-      "mean_token_accuracy": 0.768056258559227,
-      "num_tokens": 30998691.0,
+      "loss": 0.6228189468383789,
+      "mean_token_accuracy": 0.7473640963435173,
+      "num_tokens": 14697839.0,
       "step": 396
     },
     {
-      "entropy": 0.5718225091695786,
-      "epoch": 7.355140186915888,
-      "grad_norm": 0.028454070910811424,
+      "entropy": 0.6101134717464447,
+      "epoch": 2.564476885644769,
+      "grad_norm": 0.008944439701735973,
       "learning_rate": 0.0002,
-      "loss": 0.5710096955299377,
-      "mean_token_accuracy": 0.767062708735466,
-      "num_tokens": 31077693.0,
+      "loss": 0.6120374798774719,
+      "mean_token_accuracy": 0.7512135058641434,
+      "num_tokens": 14735253.0,
       "step": 397
     },
     {
-      "entropy": 0.5885006636381149,
-      "epoch": 7.373831775700935,
-      "grad_norm": 0.026537394151091576,
+      "entropy": 0.6294003054499626,
+      "epoch": 2.5709651257096513,
+      "grad_norm": 0.010403397493064404,
       "learning_rate": 0.0002,
-      "loss": 0.5815306901931763,
-      "mean_token_accuracy": 0.7627963721752167,
-      "num_tokens": 31157987.0,
+      "loss": 0.6361932754516602,
+      "mean_token_accuracy": 0.7407518178224564,
+      "num_tokens": 14772533.0,
       "step": 398
     },
     {
-      "entropy": 0.5804588943719864,
-      "epoch": 7.392523364485982,
-      "grad_norm": 0.029504472389817238,
+      "entropy": 0.6084831207990646,
+      "epoch": 2.5774533657745335,
+      "grad_norm": 0.007994825951755047,
       "learning_rate": 0.0002,
-      "loss": 0.575842022895813,
-      "mean_token_accuracy": 0.7651453614234924,
-      "num_tokens": 31237052.0,
+      "loss": 0.6090534329414368,
+      "mean_token_accuracy": 0.7516237422823906,
+      "num_tokens": 14809746.0,
       "step": 399
     },
     {
-      "entropy": 0.5789126753807068,
-      "epoch": 7.411214953271028,
-      "grad_norm": 0.03179961070418358,
+      "entropy": 0.6314651742577553,
+      "epoch": 2.5839416058394162,
+      "grad_norm": 0.010198352858424187,
       "learning_rate": 0.0002,
-      "loss": 0.5805697441101074,
-      "mean_token_accuracy": 0.7626224607229233,
-      "num_tokens": 31316295.0,
+      "loss": 0.6245699524879456,
+      "mean_token_accuracy": 0.7476952150464058,
+      "num_tokens": 14847650.0,
       "step": 400
     },
     {
-      "entropy": 0.5770617574453354,
-      "epoch": 7.429906542056075,
-      "grad_norm": 0.028732778504490852,
+      "entropy": 0.6174981743097305,
+      "epoch": 2.5904298459042985,
+      "grad_norm": 0.009044546633958817,
       "learning_rate": 0.0002,
-      "loss": 0.5769131183624268,
-      "mean_token_accuracy": 0.7638264745473862,
-      "num_tokens": 31395245.0,
+      "loss": 0.615368664264679,
+      "mean_token_accuracy": 0.7484294474124908,
+      "num_tokens": 14884930.0,
       "step": 401
     },
     {
-      "entropy": 0.5773504972457886,
-      "epoch": 7.4485981308411215,
-      "grad_norm": 0.030201883986592293,
+      "entropy": 0.6164123043417931,
+      "epoch": 2.5969180859691807,
+      "grad_norm": 0.008100450970232487,
       "learning_rate": 0.0002,
-      "loss": 0.5783373117446899,
-      "mean_token_accuracy": 0.7651851773262024,
-      "num_tokens": 31473942.0,
+      "loss": 0.620932936668396,
+      "mean_token_accuracy": 0.7468846142292023,
+      "num_tokens": 14922467.0,
       "step": 402
     },
     {
-      "entropy": 0.572372242808342,
-      "epoch": 7.4672897196261685,
-      "grad_norm": 0.030262500047683716,
+      "entropy": 0.6131496354937553,
+      "epoch": 2.6034063260340634,
+      "grad_norm": 0.01058341097086668,
       "learning_rate": 0.0002,
-      "loss": 0.5714971423149109,
-      "mean_token_accuracy": 0.7668078988790512,
-      "num_tokens": 31551761.0,
+      "loss": 0.6206756830215454,
+      "mean_token_accuracy": 0.7483528330922127,
+      "num_tokens": 14959768.0,
       "step": 403
     },
     {
-      "entropy": 0.5828695595264435,
-      "epoch": 7.485981308411215,
-      "grad_norm": 0.031323399394750595,
+      "entropy": 0.61551932990551,
+      "epoch": 2.6098945660989457,
+      "grad_norm": 0.008704769425094128,
       "learning_rate": 0.0002,
-      "loss": 0.578177809715271,
-      "mean_token_accuracy": 0.7639882117509842,
-      "num_tokens": 31630091.0,
+      "loss": 0.6186975240707397,
+      "mean_token_accuracy": 0.7474388629198074,
+      "num_tokens": 14996628.0,
       "step": 404
     },
     {
-      "entropy": 0.5827517062425613,
-      "epoch": 7.5046728971962615,
-      "grad_norm": 0.030732523649930954,
+      "entropy": 0.6159948483109474,
+      "epoch": 2.616382806163828,
+      "grad_norm": 0.009592502377927303,
       "learning_rate": 0.0002,
-      "loss": 0.5837417840957642,
-      "mean_token_accuracy": 0.7618696689605713,
-      "num_tokens": 31708952.0,
+      "loss": 0.6187817454338074,
+      "mean_token_accuracy": 0.7487591952085495,
+      "num_tokens": 15033592.0,
       "step": 405
     },
     {
-      "entropy": 0.5743730962276459,
-      "epoch": 7.5233644859813085,
-      "grad_norm": 0.03165881335735321,
+      "entropy": 0.6296302676200867,
+      "epoch": 2.6228710462287106,
+      "grad_norm": 0.009406859055161476,
       "learning_rate": 0.0002,
-      "loss": 0.5784010887145996,
-      "mean_token_accuracy": 0.7655890434980392,
-      "num_tokens": 31788434.0,
+      "loss": 0.6244117617607117,
+      "mean_token_accuracy": 0.7439019158482552,
+      "num_tokens": 15071235.0,
       "step": 406
     },
     {
-      "entropy": 0.5767014771699905,
-      "epoch": 7.542056074766355,
-      "grad_norm": 0.033420149236917496,
+      "entropy": 0.6396452710032463,
+      "epoch": 2.629359286293593,
+      "grad_norm": 0.009445966221392155,
       "learning_rate": 0.0002,
-      "loss": 0.5742257833480835,
-      "mean_token_accuracy": 0.7655230015516281,
-      "num_tokens": 31868440.0,
+      "loss": 0.634600043296814,
+      "mean_token_accuracy": 0.7421920970082283,
+      "num_tokens": 15108591.0,
       "step": 407
     },
     {
-      "entropy": 0.5853272676467896,
-      "epoch": 7.5607476635514015,
-      "grad_norm": 0.036208223551511765,
+      "entropy": 0.6241939589381218,
+      "epoch": 2.635847526358475,
+      "grad_norm": 0.008459821343421936,
       "learning_rate": 0.0002,
-      "loss": 0.5799100399017334,
-      "mean_token_accuracy": 0.7635609358549118,
-      "num_tokens": 31947381.0,
+      "loss": 0.6214285492897034,
+      "mean_token_accuracy": 0.745953157544136,
+      "num_tokens": 15146394.0,
       "step": 408
     },
     {
-      "entropy": 0.5662664473056793,
-      "epoch": 7.579439252336448,
-      "grad_norm": 0.034790199249982834,
+      "entropy": 0.6104935929179192,
+      "epoch": 2.6423357664233578,
+      "grad_norm": 0.010156887583434582,
       "learning_rate": 0.0002,
-      "loss": 0.5666247606277466,
-      "mean_token_accuracy": 0.768843486905098,
-      "num_tokens": 32025790.0,
+      "loss": 0.6155494451522827,
+      "mean_token_accuracy": 0.7467065751552582,
+      "num_tokens": 15183954.0,
       "step": 409
     },
     {
-      "entropy": 0.5646423548460007,
-      "epoch": 7.598130841121495,
-      "grad_norm": 0.03210895508527756,
+      "entropy": 0.6200908496975899,
+      "epoch": 2.64882400648824,
+      "grad_norm": 0.01134799886494875,
       "learning_rate": 0.0002,
-      "loss": 0.5644558072090149,
-      "mean_token_accuracy": 0.7712739408016205,
-      "num_tokens": 32105937.0,
+      "loss": 0.6298259496688843,
+      "mean_token_accuracy": 0.7434926256537437,
+      "num_tokens": 15221318.0,
       "step": 410
     },
     {
-      "entropy": 0.5722718238830566,
-      "epoch": 7.616822429906542,
-      "grad_norm": 0.04400765895843506,
+      "entropy": 0.6163679510354996,
+      "epoch": 2.6553122465531223,
+      "grad_norm": 0.008384560234844685,
       "learning_rate": 0.0002,
-      "loss": 0.576938271522522,
-      "mean_token_accuracy": 0.766321137547493,
-      "num_tokens": 32184675.0,
+      "loss": 0.6196904182434082,
+      "mean_token_accuracy": 0.7484511360526085,
+      "num_tokens": 15258714.0,
       "step": 411
     },
     {
-      "entropy": 0.5691269785165787,
-      "epoch": 7.635514018691588,
-      "grad_norm": 0.026371125131845474,
+      "entropy": 0.6290042921900749,
+      "epoch": 2.661800486618005,
+      "grad_norm": 0.011763868853449821,
       "learning_rate": 0.0002,
-      "loss": 0.5647107362747192,
-      "mean_token_accuracy": 0.7685294896364212,
-      "num_tokens": 32264427.0,
+      "loss": 0.623403787612915,
+      "mean_token_accuracy": 0.7461108565330505,
+      "num_tokens": 15295950.0,
       "step": 412
     },
     {
-      "entropy": 0.5793853551149368,
-      "epoch": 7.654205607476635,
-      "grad_norm": 0.034230053424835205,
+      "entropy": 0.6304820030927658,
+      "epoch": 2.668288726682887,
+      "grad_norm": 0.008899668231606483,
       "learning_rate": 0.0002,
-      "loss": 0.579481303691864,
-      "mean_token_accuracy": 0.7630301266908646,
-      "num_tokens": 32342836.0,
+      "loss": 0.6314485669136047,
+      "mean_token_accuracy": 0.7429024130105972,
+      "num_tokens": 15333333.0,
       "step": 413
     },
     {
-      "entropy": 0.5742208808660507,
-      "epoch": 7.672897196261682,
-      "grad_norm": 0.031009649857878685,
+      "entropy": 0.6225186064839363,
+      "epoch": 2.6747769667477694,
+      "grad_norm": 0.008537418209016323,
       "learning_rate": 0.0002,
-      "loss": 0.5699969530105591,
-      "mean_token_accuracy": 0.7685294300317764,
-      "num_tokens": 32421621.0,
+      "loss": 0.6211932897567749,
+      "mean_token_accuracy": 0.747981533408165,
+      "num_tokens": 15370982.0,
       "step": 414
     },
     {
-      "entropy": 0.5670227259397507,
-      "epoch": 7.691588785046729,
-      "grad_norm": 0.03162868693470955,
+      "entropy": 0.6107104495167732,
+      "epoch": 2.681265206812652,
+      "grad_norm": 0.01061639841645956,
       "learning_rate": 0.0002,
-      "loss": 0.5665925741195679,
-      "mean_token_accuracy": 0.7693023234605789,
-      "num_tokens": 32500845.0,
+      "loss": 0.6177730560302734,
+      "mean_token_accuracy": 0.7482814937829971,
+      "num_tokens": 15408426.0,
       "step": 415
     },
     {
-      "entropy": 0.5678260177373886,
-      "epoch": 7.710280373831775,
-      "grad_norm": 0.033084671944379807,
+      "entropy": 0.6148537546396255,
+      "epoch": 2.6877534468775344,
+      "grad_norm": 0.011781369335949421,
       "learning_rate": 0.0002,
-      "loss": 0.5648590922355652,
-      "mean_token_accuracy": 0.7698398530483246,
-      "num_tokens": 32579287.0,
+      "loss": 0.6236268281936646,
+      "mean_token_accuracy": 0.7474222108721733,
+      "num_tokens": 15445617.0,
       "step": 416
     },
     {
-      "entropy": 0.5775247514247894,
-      "epoch": 7.728971962616822,
-      "grad_norm": 0.043284881860017776,
+      "entropy": 0.6345531940460205,
+      "epoch": 2.6942416869424166,
+      "grad_norm": 0.008664336055517197,
       "learning_rate": 0.0002,
-      "loss": 0.5803564786911011,
-      "mean_token_accuracy": 0.7633648663759232,
-      "num_tokens": 32659251.0,
+      "loss": 0.6367125511169434,
+      "mean_token_accuracy": 0.7417123690247536,
+      "num_tokens": 15483295.0,
       "step": 417
     },
     {
-      "entropy": 0.5801283568143845,
-      "epoch": 7.747663551401869,
-      "grad_norm": 0.033718738704919815,
+      "entropy": 0.632311999797821,
+      "epoch": 2.7007299270072993,
+      "grad_norm": 0.010981544852256775,
       "learning_rate": 0.0002,
-      "loss": 0.5818736553192139,
-      "mean_token_accuracy": 0.762788325548172,
-      "num_tokens": 32736756.0,
+      "loss": 0.6256700158119202,
+      "mean_token_accuracy": 0.7448865845799446,
+      "num_tokens": 15520324.0,
       "step": 418
     },
     {
-      "entropy": 0.5818484872579575,
-      "epoch": 7.766355140186916,
-      "grad_norm": 0.03887256979942322,
+      "entropy": 0.6247982755303383,
+      "epoch": 2.7072181670721815,
+      "grad_norm": 0.010701069608330727,
       "learning_rate": 0.0002,
-      "loss": 0.5807799100875854,
-      "mean_token_accuracy": 0.7634662240743637,
-      "num_tokens": 32815653.0,
+      "loss": 0.6191861629486084,
+      "mean_token_accuracy": 0.747975766658783,
+      "num_tokens": 15557534.0,
       "step": 419
     },
     {
-      "entropy": 0.5854622423648834,
-      "epoch": 7.785046728971962,
-      "grad_norm": 0.032060056924819946,
+      "entropy": 0.6309525445103645,
+      "epoch": 2.7137064071370642,
+      "grad_norm": 0.008924427442252636,
       "learning_rate": 0.0002,
-      "loss": 0.5819664001464844,
-      "mean_token_accuracy": 0.7622562795877457,
-      "num_tokens": 32893976.0,
+      "loss": 0.6328113079071045,
+      "mean_token_accuracy": 0.741542711853981,
+      "num_tokens": 15594763.0,
       "step": 420
     },
     {
-      "entropy": 0.5702288597822189,
-      "epoch": 7.803738317757009,
-      "grad_norm": 0.0487007200717926,
+      "entropy": 0.6043407768011093,
+      "epoch": 2.7201946472019465,
+      "grad_norm": 0.008746204897761345,
       "learning_rate": 0.0002,
-      "loss": 0.5694952011108398,
-      "mean_token_accuracy": 0.767988994717598,
-      "num_tokens": 32973443.0,
+      "loss": 0.6077402830123901,
+      "mean_token_accuracy": 0.7525621652603149,
+      "num_tokens": 15631963.0,
       "step": 421
     },
     {
-      "entropy": 0.5743753612041473,
-      "epoch": 7.822429906542056,
-      "grad_norm": 0.03143998235464096,
+      "entropy": 0.623171254992485,
+      "epoch": 2.7266828872668287,
+      "grad_norm": 0.012448329478502274,
       "learning_rate": 0.0002,
-      "loss": 0.5762680768966675,
-      "mean_token_accuracy": 0.765287771821022,
-      "num_tokens": 33051336.0,
+      "loss": 0.6337227821350098,
+      "mean_token_accuracy": 0.7424034029245377,
+      "num_tokens": 15669213.0,
       "step": 422
     },
     {
-      "entropy": 0.580144464969635,
-      "epoch": 7.841121495327103,
-      "grad_norm": 0.04133511334657669,
+      "entropy": 0.6182989627122879,
+      "epoch": 2.7331711273317114,
+      "grad_norm": 0.00980245042592287,
       "learning_rate": 0.0002,
-      "loss": 0.5772802829742432,
-      "mean_token_accuracy": 0.765201061964035,
-      "num_tokens": 33129798.0,
+      "loss": 0.6246788501739502,
+      "mean_token_accuracy": 0.7442185133695602,
+      "num_tokens": 15706502.0,
       "step": 423
     },
     {
-      "entropy": 0.5860956609249115,
-      "epoch": 7.859813084112149,
-      "grad_norm": 0.03347109630703926,
+      "entropy": 0.6341974511742592,
+      "epoch": 2.7396593673965937,
+      "grad_norm": 0.010192792862653732,
       "learning_rate": 0.0002,
-      "loss": 0.5789963006973267,
-      "mean_token_accuracy": 0.7634441554546356,
-      "num_tokens": 33209273.0,
+      "loss": 0.6291946172714233,
+      "mean_token_accuracy": 0.7423791959881783,
+      "num_tokens": 15744210.0,
       "step": 424
     },
     {
-      "entropy": 0.5873236507177353,
-      "epoch": 7.878504672897196,
-      "grad_norm": 0.037143707275390625,
+      "entropy": 0.6213679686188698,
+      "epoch": 2.7461476074614763,
+      "grad_norm": 0.010065656155347824,
       "learning_rate": 0.0002,
-      "loss": 0.5865265727043152,
-      "mean_token_accuracy": 0.7612786442041397,
-      "num_tokens": 33288771.0,
+      "loss": 0.6173956394195557,
+      "mean_token_accuracy": 0.7478613257408142,
+      "num_tokens": 15781645.0,
       "step": 425
     },
     {
-      "entropy": 0.5686286091804504,
-      "epoch": 7.897196261682243,
-      "grad_norm": 0.03085118532180786,
+      "entropy": 0.6221437901258469,
+      "epoch": 2.7526358475263586,
+      "grad_norm": 0.0095105255022645,
       "learning_rate": 0.0002,
-      "loss": 0.5708251595497131,
-      "mean_token_accuracy": 0.7655164450407028,
-      "num_tokens": 33368400.0,
+      "loss": 0.6207852363586426,
+      "mean_token_accuracy": 0.7459077090024948,
+      "num_tokens": 15818882.0,
       "step": 426
     },
     {
-      "entropy": 0.5756514668464661,
-      "epoch": 7.91588785046729,
-      "grad_norm": 0.031129810959100723,
+      "entropy": 0.6309037730097771,
+      "epoch": 2.759124087591241,
+      "grad_norm": 0.008715137839317322,
       "learning_rate": 0.0002,
-      "loss": 0.5749300718307495,
-      "mean_token_accuracy": 0.765411451458931,
-      "num_tokens": 33447115.0,
+      "loss": 0.635662317276001,
+      "mean_token_accuracy": 0.7393417879939079,
+      "num_tokens": 15856215.0,
       "step": 427
     },
     {
-      "entropy": 0.5711506605148315,
-      "epoch": 7.934579439252336,
-      "grad_norm": 0.030737726017832756,
+      "entropy": 0.6107242554426193,
+      "epoch": 2.7656123276561235,
+      "grad_norm": 0.01041481178253889,
       "learning_rate": 0.0002,
-      "loss": 0.5667474269866943,
-      "mean_token_accuracy": 0.7683475613594055,
-      "num_tokens": 33527195.0,
+      "loss": 0.6185387372970581,
+      "mean_token_accuracy": 0.7483986243605614,
+      "num_tokens": 15892879.0,
       "step": 428
     },
     {
-      "entropy": 0.5755315274000168,
-      "epoch": 7.953271028037383,
-      "grad_norm": 0.029060108587145805,
+      "entropy": 0.6216551586985588,
+      "epoch": 2.7721005677210058,
+      "grad_norm": 0.0099997129291296,
       "learning_rate": 0.0002,
-      "loss": 0.5743991732597351,
-      "mean_token_accuracy": 0.7665413618087769,
-      "num_tokens": 33606110.0,
+      "loss": 0.6273539066314697,
+      "mean_token_accuracy": 0.7425299435853958,
+      "num_tokens": 15930440.0,
       "step": 429
     },
     {
-      "entropy": 0.5762781202793121,
-      "epoch": 7.97196261682243,
-      "grad_norm": 0.03172822669148445,
+      "entropy": 0.6228181794285774,
+      "epoch": 2.778588807785888,
+      "grad_norm": 0.009126834571361542,
       "learning_rate": 0.0002,
-      "loss": 0.5783078074455261,
-      "mean_token_accuracy": 0.7643001526594162,
-      "num_tokens": 33685445.0,
+      "loss": 0.6193030476570129,
+      "mean_token_accuracy": 0.748077854514122,
+      "num_tokens": 15967547.0,
       "step": 430
     },
     {
-      "entropy": 0.5830086320638657,
-      "epoch": 7.990654205607477,
-      "grad_norm": 0.03025674633681774,
+      "entropy": 0.6386850997805595,
+      "epoch": 2.7850770478507707,
+      "grad_norm": 0.00938471220433712,
       "learning_rate": 0.0002,
-      "loss": 0.58307945728302,
-      "mean_token_accuracy": 0.7618844658136368,
-      "num_tokens": 33763163.0,
+      "loss": 0.6340270042419434,
+      "mean_token_accuracy": 0.7409438043832779,
+      "num_tokens": 16004947.0,
       "step": 431
     },
     {
-      "entropy": 0.5734254121780396,
-      "epoch": 8.0,
-      "grad_norm": 0.044769272208213806,
+      "entropy": 0.6217259094119072,
+      "epoch": 2.791565287915653,
+      "grad_norm": 0.010031217709183693,
       "learning_rate": 0.0002,
-      "loss": 0.5759371519088745,
-      "mean_token_accuracy": 0.7648670375347137,
-      "num_tokens": 33802024.0,
+      "loss": 0.6190569400787354,
+      "mean_token_accuracy": 0.7476614862680435,
+      "num_tokens": 16041946.0,
       "step": 432
     },
     {
-      "entropy": 0.5717909038066864,
-      "epoch": 8.018691588785046,
-      "grad_norm": 0.03209570422768593,
+      "entropy": 0.6218375042080879,
+      "epoch": 2.798053527980535,
+      "grad_norm": 0.008579161949455738,
       "learning_rate": 0.0002,
-      "loss": 0.564399242401123,
-      "mean_token_accuracy": 0.7706466615200043,
-      "num_tokens": 33881555.0,
+      "loss": 0.6240730285644531,
+      "mean_token_accuracy": 0.746677428483963,
+      "num_tokens": 16078788.0,
       "step": 433
     },
     {
-      "entropy": 0.5607770830392838,
-      "epoch": 8.037383177570094,
-      "grad_norm": 0.04142817482352257,
+      "entropy": 0.627423033118248,
+      "epoch": 2.804541768045418,
+      "grad_norm": 0.010432623326778412,
       "learning_rate": 0.0002,
-      "loss": 0.5611358880996704,
-      "mean_token_accuracy": 0.7706198245286942,
-      "num_tokens": 33960478.0,
+      "loss": 0.6334057450294495,
+      "mean_token_accuracy": 0.7425751611590385,
+      "num_tokens": 16116073.0,
       "step": 434
     },
     {
-      "entropy": 0.5713176429271698,
-      "epoch": 8.05607476635514,
-      "grad_norm": 0.03570501133799553,
+      "entropy": 0.6249776631593704,
+      "epoch": 2.8110300081103,
+      "grad_norm": 0.01080512534826994,
       "learning_rate": 0.0002,
-      "loss": 0.5715903639793396,
-      "mean_token_accuracy": 0.7666511684656143,
-      "num_tokens": 34039461.0,
+      "loss": 0.6291300058364868,
+      "mean_token_accuracy": 0.7412911430001259,
+      "num_tokens": 16153421.0,
       "step": 435
     },
     {
-      "entropy": 0.5835350900888443,
-      "epoch": 8.074766355140186,
-      "grad_norm": 0.042137570679187775,
+      "entropy": 0.6199780628085136,
+      "epoch": 2.8175182481751824,
+      "grad_norm": 0.008089151233434677,
       "learning_rate": 0.0002,
-      "loss": 0.578724205493927,
-      "mean_token_accuracy": 0.7641146183013916,
-      "num_tokens": 34119055.0,
+      "loss": 0.6166943311691284,
+      "mean_token_accuracy": 0.7502307966351509,
+      "num_tokens": 16190692.0,
       "step": 436
     },
     {
-      "entropy": 0.5712213516235352,
-      "epoch": 8.093457943925234,
-      "grad_norm": 0.0390690453350544,
+      "entropy": 0.6264552250504494,
+      "epoch": 2.824006488240065,
+      "grad_norm": 0.009256084449589252,
       "learning_rate": 0.0002,
-      "loss": 0.5641062259674072,
-      "mean_token_accuracy": 0.7685039341449738,
-      "num_tokens": 34197572.0,
+      "loss": 0.6239896416664124,
+      "mean_token_accuracy": 0.7462676838040352,
+      "num_tokens": 16227626.0,
       "step": 437
     },
     {
-      "entropy": 0.5626108348369598,
-      "epoch": 8.11214953271028,
-      "grad_norm": 0.05106540396809578,
+      "entropy": 0.6257441863417625,
+      "epoch": 2.8304947283049473,
+      "grad_norm": 0.010192189365625381,
       "learning_rate": 0.0002,
-      "loss": 0.5654831528663635,
-      "mean_token_accuracy": 0.7689186930656433,
-      "num_tokens": 34277266.0,
+      "loss": 0.6272855997085571,
+      "mean_token_accuracy": 0.7451135069131851,
+      "num_tokens": 16264951.0,
       "step": 438
     },
     {
-      "entropy": 0.5697167217731476,
-      "epoch": 8.130841121495328,
-      "grad_norm": 0.043526820838451385,
+      "entropy": 0.6197425052523613,
+      "epoch": 2.8369829683698295,
+      "grad_norm": 0.00909087248146534,
       "learning_rate": 0.0002,
-      "loss": 0.5732031464576721,
-      "mean_token_accuracy": 0.7650811076164246,
-      "num_tokens": 34355043.0,
+      "loss": 0.6246816515922546,
+      "mean_token_accuracy": 0.7462664917111397,
+      "num_tokens": 16302276.0,
       "step": 439
     },
     {
-      "entropy": 0.5768076628446579,
-      "epoch": 8.149532710280374,
-      "grad_norm": 0.04215633124113083,
+      "entropy": 0.6187947019934654,
+      "epoch": 2.8434712084347122,
+      "grad_norm": 0.010161105543375015,
       "learning_rate": 0.0002,
-      "loss": 0.5718085765838623,
-      "mean_token_accuracy": 0.766776368021965,
-      "num_tokens": 34435449.0,
+      "loss": 0.6317975521087646,
+      "mean_token_accuracy": 0.7427225112915039,
+      "num_tokens": 16339652.0,
       "step": 440
     },
     {
-      "entropy": 0.5621174275875092,
-      "epoch": 8.16822429906542,
-      "grad_norm": 0.03552766144275665,
+      "entropy": 0.6189513206481934,
+      "epoch": 2.8499594484995945,
+      "grad_norm": 0.009055238217115402,
       "learning_rate": 0.0002,
-      "loss": 0.5604408979415894,
-      "mean_token_accuracy": 0.7712910920381546,
-      "num_tokens": 34513638.0,
+      "loss": 0.6247466802597046,
+      "mean_token_accuracy": 0.7450408861041069,
+      "num_tokens": 16376803.0,
       "step": 441
     },
     {
-      "entropy": 0.5522593855857849,
-      "epoch": 8.186915887850468,
-      "grad_norm": 0.05295529216527939,
+      "entropy": 0.6144088804721832,
+      "epoch": 2.8564476885644767,
+      "grad_norm": 0.008574709296226501,
       "learning_rate": 0.0002,
-      "loss": 0.560666024684906,
-      "mean_token_accuracy": 0.770970806479454,
-      "num_tokens": 34591894.0,
+      "loss": 0.618188738822937,
+      "mean_token_accuracy": 0.7477873042225838,
+      "num_tokens": 16413555.0,
       "step": 442
     },
     {
-      "entropy": 0.5586026161909103,
-      "epoch": 8.205607476635514,
-      "grad_norm": 0.045166611671447754,
+      "entropy": 0.6285263672471046,
+      "epoch": 2.8629359286293594,
+      "grad_norm": 0.01058571133762598,
       "learning_rate": 0.0002,
-      "loss": 0.5580676794052124,
-      "mean_token_accuracy": 0.7712274938821793,
-      "num_tokens": 34671649.0,
+      "loss": 0.6210506558418274,
+      "mean_token_accuracy": 0.7466302737593651,
+      "num_tokens": 16450235.0,
       "step": 443
     },
     {
-      "entropy": 0.5645840466022491,
-      "epoch": 8.22429906542056,
-      "grad_norm": 0.043214891105890274,
+      "entropy": 0.6278195902705193,
+      "epoch": 2.8694241686942417,
+      "grad_norm": 0.008600706234574318,
       "learning_rate": 0.0002,
-      "loss": 0.5546973347663879,
-      "mean_token_accuracy": 0.772943839430809,
-      "num_tokens": 34750535.0,
+      "loss": 0.6228291988372803,
+      "mean_token_accuracy": 0.7465885654091835,
+      "num_tokens": 16487688.0,
       "step": 444
     },
     {
-      "entropy": 0.566862940788269,
-      "epoch": 8.242990654205608,
-      "grad_norm": 0.05362832173705101,
+      "entropy": 0.6128537505865097,
+      "epoch": 2.875912408759124,
+      "grad_norm": 0.009188942611217499,
       "learning_rate": 0.0002,
-      "loss": 0.5646588802337646,
-      "mean_token_accuracy": 0.7682545930147171,
-      "num_tokens": 34829176.0,
+      "loss": 0.6115404367446899,
+      "mean_token_accuracy": 0.7511069476604462,
+      "num_tokens": 16525056.0,
       "step": 445
     },
     {
-      "entropy": 0.5670454204082489,
-      "epoch": 8.261682242990654,
-      "grad_norm": 0.05154046416282654,
+      "entropy": 0.6209624111652374,
+      "epoch": 2.8824006488240066,
+      "grad_norm": 0.009248930029571056,
       "learning_rate": 0.0002,
-      "loss": 0.5703809261322021,
-      "mean_token_accuracy": 0.7675753831863403,
-      "num_tokens": 34908547.0,
+      "loss": 0.6210386753082275,
+      "mean_token_accuracy": 0.7469235137104988,
+      "num_tokens": 16562914.0,
       "step": 446
     },
     {
-      "entropy": 0.5630051344633102,
-      "epoch": 8.280373831775702,
-      "grad_norm": 0.046372540295124054,
+      "entropy": 0.5965787023305893,
+      "epoch": 2.888888888888889,
+      "grad_norm": 0.0094021987169981,
       "learning_rate": 0.0002,
-      "loss": 0.5637614727020264,
-      "mean_token_accuracy": 0.7699243128299713,
-      "num_tokens": 34986705.0,
+      "loss": 0.6033673286437988,
+      "mean_token_accuracy": 0.7514390125870705,
+      "num_tokens": 16599899.0,
       "step": 447
     },
     {
-      "entropy": 0.5612086057662964,
-      "epoch": 8.299065420560748,
-      "grad_norm": 0.04247089475393295,
+      "entropy": 0.6115651428699493,
+      "epoch": 2.895377128953771,
+      "grad_norm": 0.01325083989650011,
       "learning_rate": 0.0002,
-      "loss": 0.5558289289474487,
-      "mean_token_accuracy": 0.7724988609552383,
-      "num_tokens": 35066634.0,
+      "loss": 0.6262028813362122,
+      "mean_token_accuracy": 0.7461578771471977,
+      "num_tokens": 16637011.0,
       "step": 448
     },
     {
-      "entropy": 0.5652319639921188,
-      "epoch": 8.317757009345794,
-      "grad_norm": 0.038978833705186844,
+      "entropy": 0.6188539639115334,
+      "epoch": 2.9018653690186538,
+      "grad_norm": 0.00891121570020914,
       "learning_rate": 0.0002,
-      "loss": 0.5635871887207031,
-      "mean_token_accuracy": 0.7691775411367416,
-      "num_tokens": 35145714.0,
+      "loss": 0.6159206628799438,
+      "mean_token_accuracy": 0.7494730055332184,
+      "num_tokens": 16673697.0,
       "step": 449
     },
     {
-      "entropy": 0.5542118549346924,
-      "epoch": 8.336448598130842,
-      "grad_norm": 0.04010032117366791,
+      "entropy": 0.6208583936095238,
+      "epoch": 2.908353609083536,
+      "grad_norm": 0.009828220121562481,
       "learning_rate": 0.0002,
-      "loss": 0.5565598011016846,
-      "mean_token_accuracy": 0.7727922797203064,
-      "num_tokens": 35225145.0,
+      "loss": 0.6155424118041992,
+      "mean_token_accuracy": 0.7491355165839195,
+      "num_tokens": 16710885.0,
       "step": 450
     },
     {
-      "entropy": 0.5695662796497345,
-      "epoch": 8.355140186915888,
-      "grad_norm": 0.03877483680844307,
+      "entropy": 0.6197483390569687,
+      "epoch": 2.9148418491484183,
+      "grad_norm": 0.008900245651602745,
       "learning_rate": 0.0002,
-      "loss": 0.5652719140052795,
-      "mean_token_accuracy": 0.7696373909711838,
-      "num_tokens": 35304911.0,
+      "loss": 0.6142081618309021,
+      "mean_token_accuracy": 0.748231329023838,
+      "num_tokens": 16748182.0,
       "step": 451
     },
     {
-      "entropy": 0.566323772072792,
-      "epoch": 8.373831775700934,
-      "grad_norm": 0.03444257006049156,
+      "entropy": 0.6212092339992523,
+      "epoch": 2.921330089213301,
+      "grad_norm": 0.00849118735641241,
       "learning_rate": 0.0002,
-      "loss": 0.5614066123962402,
-      "mean_token_accuracy": 0.7709103524684906,
-      "num_tokens": 35386357.0,
+      "loss": 0.6236221790313721,
+      "mean_token_accuracy": 0.7436698824167252,
+      "num_tokens": 16785028.0,
       "step": 452
     },
     {
-      "entropy": 0.5655058771371841,
-      "epoch": 8.392523364485982,
-      "grad_norm": 0.04712752252817154,
+      "entropy": 0.6107748746871948,
+      "epoch": 2.927818329278183,
+      "grad_norm": 0.009662901051342487,
       "learning_rate": 0.0002,
-      "loss": 0.5687972903251648,
-      "mean_token_accuracy": 0.7682380080223083,
-      "num_tokens": 35464875.0,
+      "loss": 0.6197327375411987,
+      "mean_token_accuracy": 0.7483203411102295,
+      "num_tokens": 16822318.0,
       "step": 453
     },
     {
-      "entropy": 0.5632896423339844,
-      "epoch": 8.411214953271028,
-      "grad_norm": 0.0352141447365284,
+      "entropy": 0.6189533397555351,
+      "epoch": 2.9343065693430654,
+      "grad_norm": 0.009161571972072124,
       "learning_rate": 0.0002,
-      "loss": 0.5623925924301147,
-      "mean_token_accuracy": 0.7705592215061188,
-      "num_tokens": 35543398.0,
+      "loss": 0.6221146583557129,
+      "mean_token_accuracy": 0.7479159906506538,
+      "num_tokens": 16859984.0,
       "step": 454
     },
     {
-      "entropy": 0.5786634534597397,
-      "epoch": 8.429906542056075,
-      "grad_norm": 0.04121265932917595,
+      "entropy": 0.6262816786766052,
+      "epoch": 2.940794809407948,
+      "grad_norm": 0.008490455336868763,
       "learning_rate": 0.0002,
-      "loss": 0.5740705728530884,
-      "mean_token_accuracy": 0.7659155428409576,
-      "num_tokens": 35622893.0,
+      "loss": 0.6270855069160461,
+      "mean_token_accuracy": 0.7452046275138855,
+      "num_tokens": 16897277.0,
       "step": 455
     },
     {
-      "entropy": 0.572764053940773,
-      "epoch": 8.448598130841122,
-      "grad_norm": 0.03707192838191986,
+      "entropy": 0.6230187118053436,
+      "epoch": 2.9472830494728304,
+      "grad_norm": 0.00920241978019476,
       "learning_rate": 0.0002,
-      "loss": 0.5670691728591919,
-      "mean_token_accuracy": 0.7682500779628754,
-      "num_tokens": 35701401.0,
+      "loss": 0.6212553977966309,
+      "mean_token_accuracy": 0.7461629658937454,
+      "num_tokens": 16934423.0,
       "step": 456
     },
     {
-      "entropy": 0.5609558075666428,
-      "epoch": 8.467289719626168,
-      "grad_norm": 0.04114922136068344,
+      "entropy": 0.6259749531745911,
+      "epoch": 2.9537712895377126,
+      "grad_norm": 0.008131385780870914,
       "learning_rate": 0.0002,
-      "loss": 0.5627433061599731,
-      "mean_token_accuracy": 0.7700482457876205,
-      "num_tokens": 35779899.0,
+      "loss": 0.6244622468948364,
+      "mean_token_accuracy": 0.7466467097401619,
+      "num_tokens": 16971857.0,
       "step": 457
     },
     {
-      "entropy": 0.5678876638412476,
-      "epoch": 8.485981308411215,
-      "grad_norm": 0.03938848152756691,
+      "entropy": 0.6159861534833908,
+      "epoch": 2.9602595296025953,
+      "grad_norm": 0.008032194338738918,
       "learning_rate": 0.0002,
-      "loss": 0.5726190805435181,
-      "mean_token_accuracy": 0.7662196010351181,
-      "num_tokens": 35857585.0,
+      "loss": 0.6193053722381592,
+      "mean_token_accuracy": 0.7479442283511162,
+      "num_tokens": 17008930.0,
       "step": 458
     },
     {
-      "entropy": 0.5759584605693817,
-      "epoch": 8.504672897196262,
-      "grad_norm": 0.04363275319337845,
+      "entropy": 0.617520272731781,
+      "epoch": 2.9667477696674776,
+      "grad_norm": 0.009034648537635803,
       "learning_rate": 0.0002,
-      "loss": 0.5743913054466248,
-      "mean_token_accuracy": 0.7650963664054871,
-      "num_tokens": 35935911.0,
+      "loss": 0.6226905584335327,
+      "mean_token_accuracy": 0.7449267134070396,
+      "num_tokens": 17046294.0,
       "step": 459
     },
     {
-      "entropy": 0.5694950520992279,
-      "epoch": 8.523364485981308,
-      "grad_norm": 0.03605388104915619,
+      "entropy": 0.6264740154147148,
+      "epoch": 2.9732360097323602,
+      "grad_norm": 0.008523489348590374,
       "learning_rate": 0.0002,
-      "loss": 0.5647161602973938,
-      "mean_token_accuracy": 0.7692456245422363,
-      "num_tokens": 36013948.0,
+      "loss": 0.6288236379623413,
+      "mean_token_accuracy": 0.7429340779781342,
+      "num_tokens": 17083358.0,
       "step": 460
     },
     {
-      "entropy": 0.5708936899900436,
-      "epoch": 8.542056074766355,
-      "grad_norm": 0.04307771846652031,
+      "entropy": 0.6205747723579407,
+      "epoch": 2.9797242497972425,
+      "grad_norm": 0.008030165918171406,
       "learning_rate": 0.0002,
-      "loss": 0.572138249874115,
-      "mean_token_accuracy": 0.7665975391864777,
-      "num_tokens": 36092952.0,
+      "loss": 0.6183617115020752,
+      "mean_token_accuracy": 0.7463237047195435,
+      "num_tokens": 17120761.0,
       "step": 461
     },
     {
-      "entropy": 0.5669836699962616,
-      "epoch": 8.560747663551401,
-      "grad_norm": 0.03957648202776909,
+      "entropy": 0.6367032006382942,
+      "epoch": 2.986212489862125,
+      "grad_norm": 0.009663688018918037,
       "learning_rate": 0.0002,
-      "loss": 0.567891001701355,
-      "mean_token_accuracy": 0.7687968164682388,
-      "num_tokens": 36171019.0,
+      "loss": 0.6329810619354248,
+      "mean_token_accuracy": 0.7433002442121506,
+      "num_tokens": 17158231.0,
       "step": 462
     },
     {
-      "entropy": 0.5661501735448837,
-      "epoch": 8.57943925233645,
-      "grad_norm": 0.037156861275434494,
+      "entropy": 0.626891016960144,
+      "epoch": 2.9927007299270074,
+      "grad_norm": 0.009945257566869259,
       "learning_rate": 0.0002,
-      "loss": 0.5598872900009155,
-      "mean_token_accuracy": 0.7714900076389313,
-      "num_tokens": 36249848.0,
+      "loss": 0.6275311708450317,
+      "mean_token_accuracy": 0.7450283095240593,
+      "num_tokens": 17195764.0,
       "step": 463
     },
     {
-      "entropy": 0.5739653706550598,
-      "epoch": 8.598130841121495,
-      "grad_norm": 0.0407034270465374,
+      "entropy": 0.6275185570120811,
+      "epoch": 2.9991889699918897,
+      "grad_norm": 0.008921638131141663,
       "learning_rate": 0.0002,
-      "loss": 0.5717343688011169,
-      "mean_token_accuracy": 0.7671405524015427,
-      "num_tokens": 36328929.0,
+      "loss": 0.6317576169967651,
+      "mean_token_accuracy": 0.7413767129182816,
+      "num_tokens": 17232867.0,
       "step": 464
     },
     {
-      "entropy": 0.5699207782745361,
-      "epoch": 8.616822429906541,
-      "grad_norm": 0.036081064492464066,
+      "entropy": 0.5935274362564087,
+      "epoch": 3.0,
+      "grad_norm": 0.02002483420073986,
       "learning_rate": 0.0002,
-      "loss": 0.5727741718292236,
-      "mean_token_accuracy": 0.766769751906395,
-      "num_tokens": 36407646.0,
+      "loss": 0.5921852588653564,
+      "mean_token_accuracy": 0.7577869892120361,
+      "num_tokens": 17237490.0,
       "step": 465
     },
     {
-      "entropy": 0.5586220026016235,
-      "epoch": 8.63551401869159,
-      "grad_norm": 0.03620290383696556,
+      "entropy": 0.610297404229641,
+      "epoch": 3.0064882400648822,
+      "grad_norm": 0.010434089228510857,
       "learning_rate": 0.0002,
-      "loss": 0.5579659938812256,
-      "mean_token_accuracy": 0.771889865398407,
-      "num_tokens": 36487160.0,
+      "loss": 0.6096053123474121,
+      "mean_token_accuracy": 0.7505725026130676,
+      "num_tokens": 17274381.0,
       "step": 466
     },
     {
-      "entropy": 0.5877305865287781,
-      "epoch": 8.654205607476635,
-      "grad_norm": 0.033869534730911255,
+      "entropy": 0.6093652322888374,
+      "epoch": 3.012976480129765,
+      "grad_norm": 0.011252244934439659,
       "learning_rate": 0.0002,
-      "loss": 0.5841851234436035,
-      "mean_token_accuracy": 0.7620754092931747,
-      "num_tokens": 36567160.0,
+      "loss": 0.6079680919647217,
+      "mean_token_accuracy": 0.751014731824398,
+      "num_tokens": 17311247.0,
       "step": 467
     },
     {
-      "entropy": 0.5693440735340118,
-      "epoch": 8.672897196261681,
-      "grad_norm": 0.0372026227414608,
+      "entropy": 0.6135042458772659,
+      "epoch": 3.019464720194647,
+      "grad_norm": 0.011452090926468372,
       "learning_rate": 0.0002,
-      "loss": 0.5639486312866211,
-      "mean_token_accuracy": 0.7693230509757996,
-      "num_tokens": 36647567.0,
+      "loss": 0.621872067451477,
+      "mean_token_accuracy": 0.7463375702500343,
+      "num_tokens": 17348490.0,
       "step": 468
     },
     {
-      "entropy": 0.5630517154932022,
-      "epoch": 8.69158878504673,
-      "grad_norm": 0.036731891334056854,
+      "entropy": 0.6276230961084366,
+      "epoch": 3.0259529602595294,
+      "grad_norm": 0.011181695386767387,
       "learning_rate": 0.0002,
-      "loss": 0.5677603483200073,
-      "mean_token_accuracy": 0.767775371670723,
-      "num_tokens": 36726634.0,
+      "loss": 0.6287038326263428,
+      "mean_token_accuracy": 0.7455412894487381,
+      "num_tokens": 17385419.0,
       "step": 469
     },
     {
-      "entropy": 0.5637719482183456,
-      "epoch": 8.710280373831775,
-      "grad_norm": 0.03504373878240585,
+      "entropy": 0.6226534396409988,
+      "epoch": 3.032441200324412,
+      "grad_norm": 0.00998624972999096,
       "learning_rate": 0.0002,
-      "loss": 0.5635030269622803,
-      "mean_token_accuracy": 0.7712674289941788,
-      "num_tokens": 36805361.0,
+      "loss": 0.6229760646820068,
+      "mean_token_accuracy": 0.7449204847216606,
+      "num_tokens": 17423056.0,
       "step": 470
     },
     {
-      "entropy": 0.5734125971794128,
-      "epoch": 8.728971962616823,
-      "grad_norm": 0.03756216913461685,
+      "entropy": 0.6008478105068207,
+      "epoch": 3.0389294403892944,
+      "grad_norm": 0.0095088891685009,
       "learning_rate": 0.0002,
-      "loss": 0.57008957862854,
-      "mean_token_accuracy": 0.7670037597417831,
-      "num_tokens": 36883849.0,
+      "loss": 0.6020908951759338,
+      "mean_token_accuracy": 0.7544899210333824,
+      "num_tokens": 17460157.0,
       "step": 471
     },
     {
-      "entropy": 0.5627129673957825,
-      "epoch": 8.74766355140187,
-      "grad_norm": 0.03570760041475296,
+      "entropy": 0.6252934113144875,
+      "epoch": 3.0454176804541766,
+      "grad_norm": 0.010803421959280968,
       "learning_rate": 0.0002,
-      "loss": 0.5617348551750183,
-      "mean_token_accuracy": 0.7711147964000702,
-      "num_tokens": 36963050.0,
+      "loss": 0.6294186115264893,
+      "mean_token_accuracy": 0.7426271066069603,
+      "num_tokens": 17497579.0,
       "step": 472
     },
     {
-      "entropy": 0.5754802078008652,
-      "epoch": 8.766355140186915,
-      "grad_norm": 0.04743500053882599,
+      "entropy": 0.6106472909450531,
+      "epoch": 3.0519059205190593,
+      "grad_norm": 0.010041839443147182,
       "learning_rate": 0.0002,
-      "loss": 0.5773271322250366,
-      "mean_token_accuracy": 0.763431116938591,
-      "num_tokens": 37041165.0,
+      "loss": 0.6172924041748047,
+      "mean_token_accuracy": 0.7476223781704903,
+      "num_tokens": 17534955.0,
       "step": 473
     },
     {
-      "entropy": 0.5732788294553757,
-      "epoch": 8.785046728971963,
-      "grad_norm": 0.04425869137048721,
+      "entropy": 0.6221166178584099,
+      "epoch": 3.0583941605839415,
+      "grad_norm": 0.009123523719608784,
       "learning_rate": 0.0002,
-      "loss": 0.5683580636978149,
-      "mean_token_accuracy": 0.7678799331188202,
-      "num_tokens": 37120193.0,
+      "loss": 0.6244035363197327,
+      "mean_token_accuracy": 0.7448302879929543,
+      "num_tokens": 17572150.0,
       "step": 474
     },
     {
-      "entropy": 0.5810448527336121,
-      "epoch": 8.80373831775701,
-      "grad_norm": 0.03165487200021744,
+      "entropy": 0.6102956235408783,
+      "epoch": 3.0648824006488242,
+      "grad_norm": 0.008510733023285866,
       "learning_rate": 0.0002,
-      "loss": 0.5795904397964478,
-      "mean_token_accuracy": 0.7625752538442612,
-      "num_tokens": 37198754.0,
+      "loss": 0.6107138991355896,
+      "mean_token_accuracy": 0.7484675496816635,
+      "num_tokens": 17609343.0,
       "step": 475
     },
     {
-      "entropy": 0.5730142742395401,
-      "epoch": 8.822429906542055,
-      "grad_norm": 0.04261263459920883,
+      "entropy": 0.6124226823449135,
+      "epoch": 3.0713706407137065,
+      "grad_norm": 0.009730803780257702,
       "learning_rate": 0.0002,
-      "loss": 0.5765471458435059,
-      "mean_token_accuracy": 0.7660854458808899,
-      "num_tokens": 37276721.0,
+      "loss": 0.6136341691017151,
+      "mean_token_accuracy": 0.7490884363651276,
+      "num_tokens": 17646448.0,
       "step": 476
     },
     {
-      "entropy": 0.5741526335477829,
-      "epoch": 8.841121495327103,
-      "grad_norm": 0.032484255731105804,
+      "entropy": 0.6132366731762886,
+      "epoch": 3.0778588807785887,
+      "grad_norm": 0.008984670974314213,
       "learning_rate": 0.0002,
-      "loss": 0.5710293650627136,
-      "mean_token_accuracy": 0.7682300359010696,
-      "num_tokens": 37355931.0,
+      "loss": 0.6176835298538208,
+      "mean_token_accuracy": 0.7453758120536804,
+      "num_tokens": 17683285.0,
       "step": 477
     },
     {
-      "entropy": 0.5785616040229797,
-      "epoch": 8.85981308411215,
-      "grad_norm": 0.038904186338186264,
+      "entropy": 0.6138317957520485,
+      "epoch": 3.0843471208434714,
+      "grad_norm": 0.009527822025120258,
       "learning_rate": 0.0002,
-      "loss": 0.5774451494216919,
-      "mean_token_accuracy": 0.7645587772130966,
-      "num_tokens": 37434220.0,
+      "loss": 0.6128236055374146,
+      "mean_token_accuracy": 0.7498287856578827,
+      "num_tokens": 17720424.0,
       "step": 478
     },
     {
-      "entropy": 0.5649435520172119,
-      "epoch": 8.878504672897197,
-      "grad_norm": 0.03235228732228279,
+      "entropy": 0.6150398552417755,
+      "epoch": 3.0908353609083536,
+      "grad_norm": 0.012212819419801235,
       "learning_rate": 0.0002,
-      "loss": 0.5644275546073914,
-      "mean_token_accuracy": 0.7696262449026108,
-      "num_tokens": 37512955.0,
+      "loss": 0.6184934377670288,
+      "mean_token_accuracy": 0.74715456366539,
+      "num_tokens": 17757738.0,
       "step": 479
     },
     {
-      "entropy": 0.5670848190784454,
-      "epoch": 8.897196261682243,
-      "grad_norm": 0.03787115588784218,
+      "entropy": 0.6138227432966232,
+      "epoch": 3.097323600973236,
+      "grad_norm": 0.009692898020148277,
       "learning_rate": 0.0002,
-      "loss": 0.5672973990440369,
-      "mean_token_accuracy": 0.7695111483335495,
-      "num_tokens": 37592916.0,
+      "loss": 0.615317165851593,
+      "mean_token_accuracy": 0.7491249144077301,
+      "num_tokens": 17794819.0,
       "step": 480
     },
     {
-      "entropy": 0.5635072439908981,
-      "epoch": 8.91588785046729,
-      "grad_norm": 0.039154935628175735,
+      "entropy": 0.6220738738775253,
+      "epoch": 3.1038118410381186,
+      "grad_norm": 0.009081809781491756,
       "learning_rate": 0.0002,
-      "loss": 0.5630161762237549,
-      "mean_token_accuracy": 0.7698703557252884,
-      "num_tokens": 37672769.0,
+      "loss": 0.6233057975769043,
+      "mean_token_accuracy": 0.7442603260278702,
+      "num_tokens": 17832045.0,
       "step": 481
     },
     {
-      "entropy": 0.5751306414604187,
-      "epoch": 8.934579439252337,
-      "grad_norm": 0.03351837024092674,
+      "entropy": 0.6058575734496117,
+      "epoch": 3.110300081103001,
+      "grad_norm": 0.009930483065545559,
       "learning_rate": 0.0002,
-      "loss": 0.5736410617828369,
-      "mean_token_accuracy": 0.7663961499929428,
-      "num_tokens": 37751335.0,
+      "loss": 0.6012391448020935,
+      "mean_token_accuracy": 0.7559900879859924,
+      "num_tokens": 17869448.0,
       "step": 482
     },
     {
-      "entropy": 0.5660780221223831,
-      "epoch": 8.953271028037383,
-      "grad_norm": 0.05055345594882965,
+      "entropy": 0.6158036887645721,
+      "epoch": 3.116788321167883,
+      "grad_norm": 0.009399381466209888,
       "learning_rate": 0.0002,
-      "loss": 0.5660668611526489,
-      "mean_token_accuracy": 0.7696817517280579,
-      "num_tokens": 37831244.0,
+      "loss": 0.6149718761444092,
+      "mean_token_accuracy": 0.7499518990516663,
+      "num_tokens": 17906111.0,
       "step": 483
     },
     {
-      "entropy": 0.5755114257335663,
-      "epoch": 8.97196261682243,
-      "grad_norm": 0.029924951493740082,
+      "entropy": 0.6169976964592934,
+      "epoch": 3.1232765612327658,
+      "grad_norm": 0.009214673191308975,
       "learning_rate": 0.0002,
-      "loss": 0.5722489356994629,
-      "mean_token_accuracy": 0.7663017213344574,
-      "num_tokens": 37909173.0,
+      "loss": 0.6195358037948608,
+      "mean_token_accuracy": 0.744194358587265,
+      "num_tokens": 17943306.0,
       "step": 484
     },
     {
-      "entropy": 0.5656759142875671,
-      "epoch": 8.990654205607477,
-      "grad_norm": 0.05582307279109955,
+      "entropy": 0.620849184691906,
+      "epoch": 3.129764801297648,
+      "grad_norm": 0.00997876189649105,
       "learning_rate": 0.0002,
-      "loss": 0.5674799680709839,
-      "mean_token_accuracy": 0.7680423557758331,
-      "num_tokens": 37987942.0,
+      "loss": 0.6267833113670349,
+      "mean_token_accuracy": 0.744784764945507,
+      "num_tokens": 17980613.0,
       "step": 485
     },
     {
-      "entropy": 0.5792860984802246,
-      "epoch": 9.0,
-      "grad_norm": 0.04136984795331955,
+      "entropy": 0.6128402128815651,
+      "epoch": 3.1362530413625302,
+      "grad_norm": 0.008573702536523342,
       "learning_rate": 0.0002,
-      "loss": 0.5720219612121582,
-      "mean_token_accuracy": 0.7673911452293396,
-      "num_tokens": 38027415.0,
+      "loss": 0.6157655715942383,
+      "mean_token_accuracy": 0.7487166821956635,
+      "num_tokens": 18017947.0,
       "step": 486
     },
     {
-      "entropy": 0.5568317472934723,
-      "epoch": 9.018691588785046,
-      "grad_norm": 0.04542655497789383,
+      "entropy": 0.622900165617466,
+      "epoch": 3.142741281427413,
+      "grad_norm": 0.009601105935871601,
       "learning_rate": 0.0002,
-      "loss": 0.5488422513008118,
-      "mean_token_accuracy": 0.7752774208784103,
-      "num_tokens": 38106838.0,
+      "loss": 0.6263000965118408,
+      "mean_token_accuracy": 0.7445476576685905,
+      "num_tokens": 18055764.0,
       "step": 487
     },
     {
-      "entropy": 0.5644272416830063,
-      "epoch": 9.037383177570094,
-      "grad_norm": 0.039940495043992996,
+      "entropy": 0.6130729392170906,
+      "epoch": 3.149229521492295,
+      "grad_norm": 0.008899933658540249,
       "learning_rate": 0.0002,
-      "loss": 0.5659648776054382,
-      "mean_token_accuracy": 0.7679545432329178,
-      "num_tokens": 38185916.0,
+      "loss": 0.6115667819976807,
+      "mean_token_accuracy": 0.7505542188882828,
+      "num_tokens": 18092869.0,
       "step": 488
     },
     {
-      "entropy": 0.5693039298057556,
-      "epoch": 9.05607476635514,
-      "grad_norm": 0.054858870804309845,
+      "entropy": 0.615093469619751,
+      "epoch": 3.1557177615571774,
+      "grad_norm": 0.009630289860069752,
       "learning_rate": 0.0002,
-      "loss": 0.5734539031982422,
-      "mean_token_accuracy": 0.7664594948291779,
-      "num_tokens": 38264460.0,
+      "loss": 0.6127116084098816,
+      "mean_token_accuracy": 0.74873948097229,
+      "num_tokens": 18130032.0,
       "step": 489
     },
     {
-      "entropy": 0.5590891689062119,
-      "epoch": 9.074766355140186,
-      "grad_norm": 0.0514463409781456,
+      "entropy": 0.6184138208627701,
+      "epoch": 3.16220600162206,
+      "grad_norm": 0.009784260764718056,
       "learning_rate": 0.0002,
-      "loss": 0.5518264174461365,
-      "mean_token_accuracy": 0.7742762863636017,
-      "num_tokens": 38344456.0,
+      "loss": 0.6196454763412476,
+      "mean_token_accuracy": 0.7479289025068283,
+      "num_tokens": 18167189.0,
       "step": 490
     },
     {
-      "entropy": 0.5555678457021713,
-      "epoch": 9.093457943925234,
-      "grad_norm": 0.041370172053575516,
+      "entropy": 0.6225049868226051,
+      "epoch": 3.1686942416869424,
+      "grad_norm": 0.009817374870181084,
       "learning_rate": 0.0002,
-      "loss": 0.5566983222961426,
-      "mean_token_accuracy": 0.7714695483446121,
-      "num_tokens": 38423617.0,
+      "loss": 0.6254591345787048,
+      "mean_token_accuracy": 0.7443916201591492,
+      "num_tokens": 18204423.0,
       "step": 491
     },
     {
-      "entropy": 0.5446047633886337,
-      "epoch": 9.11214953271028,
-      "grad_norm": 0.06446854770183563,
+      "entropy": 0.599067859351635,
+      "epoch": 3.1751824817518246,
+      "grad_norm": 0.009707923047244549,
       "learning_rate": 0.0002,
-      "loss": 0.55574631690979,
-      "mean_token_accuracy": 0.7727518975734711,
-      "num_tokens": 38501911.0,
+      "loss": 0.603665292263031,
+      "mean_token_accuracy": 0.7512796595692635,
+      "num_tokens": 18241647.0,
       "step": 492
     },
     {
-      "entropy": 0.5527215301990509,
-      "epoch": 9.130841121495328,
-      "grad_norm": 0.045423898845911026,
+      "entropy": 0.6269692480564117,
+      "epoch": 3.1816707218167073,
+      "grad_norm": 0.009654807858169079,
       "learning_rate": 0.0002,
-      "loss": 0.5502952337265015,
-      "mean_token_accuracy": 0.7772093117237091,
-      "num_tokens": 38581657.0,
+      "loss": 0.6272069215774536,
+      "mean_token_accuracy": 0.7451600208878517,
+      "num_tokens": 18278874.0,
       "step": 493
     },
     {
-      "entropy": 0.5689537227153778,
-      "epoch": 9.149532710280374,
-      "grad_norm": 0.04693090170621872,
+      "entropy": 0.60984867811203,
+      "epoch": 3.1881589618815895,
+      "grad_norm": 0.00973904225975275,
       "learning_rate": 0.0002,
-      "loss": 0.5581166744232178,
-      "mean_token_accuracy": 0.7714518308639526,
-      "num_tokens": 38661327.0,
+      "loss": 0.6120602488517761,
+      "mean_token_accuracy": 0.7500757128000259,
+      "num_tokens": 18316077.0,
       "step": 494
     },
     {
-      "entropy": 0.5627695322036743,
-      "epoch": 9.16822429906542,
-      "grad_norm": 0.038687728345394135,
+      "entropy": 0.6112287938594818,
+      "epoch": 3.1946472019464722,
+      "grad_norm": 0.010687717236578465,
       "learning_rate": 0.0002,
-      "loss": 0.5605401992797852,
-      "mean_token_accuracy": 0.770438477396965,
-      "num_tokens": 38739599.0,
+      "loss": 0.607744038105011,
+      "mean_token_accuracy": 0.7535340562462807,
+      "num_tokens": 18353199.0,
       "step": 495
     },
     {
-      "entropy": 0.5462548285722733,
-      "epoch": 9.186915887850468,
-      "grad_norm": 0.0521576888859272,
+      "entropy": 0.6153931394219398,
+      "epoch": 3.2011354420113545,
+      "grad_norm": 0.009243862703442574,
       "learning_rate": 0.0002,
-      "loss": 0.5529406070709229,
-      "mean_token_accuracy": 0.7740911543369293,
-      "num_tokens": 38818199.0,
+      "loss": 0.6182031035423279,
+      "mean_token_accuracy": 0.7475122958421707,
+      "num_tokens": 18390725.0,
       "step": 496
     },
     {
-      "entropy": 0.5611866265535355,
-      "epoch": 9.205607476635514,
-      "grad_norm": 0.04400985687971115,
+      "entropy": 0.6184414252638817,
+      "epoch": 3.2076236820762367,
+      "grad_norm": 0.009725847281515598,
       "learning_rate": 0.0002,
-      "loss": 0.5519530773162842,
-      "mean_token_accuracy": 0.774472787976265,
-      "num_tokens": 38898216.0,
+      "loss": 0.6202661395072937,
+      "mean_token_accuracy": 0.7463328316807747,
+      "num_tokens": 18428083.0,
       "step": 497
     },
     {
-      "entropy": 0.5600290298461914,
-      "epoch": 9.22429906542056,
-      "grad_norm": 0.04186929762363434,
+      "entropy": 0.6206400170922279,
+      "epoch": 3.2141119221411194,
+      "grad_norm": 0.009717848151922226,
       "learning_rate": 0.0002,
-      "loss": 0.5568031072616577,
-      "mean_token_accuracy": 0.7704694867134094,
-      "num_tokens": 38975808.0,
+      "loss": 0.6237635612487793,
+      "mean_token_accuracy": 0.7449309974908829,
+      "num_tokens": 18465093.0,
       "step": 498
     },
     {
-      "entropy": 0.5411058068275452,
-      "epoch": 9.242990654205608,
-      "grad_norm": 0.055334728211164474,
+      "entropy": 0.6063303649425507,
+      "epoch": 3.2206001622060016,
+      "grad_norm": 0.009384922683238983,
       "learning_rate": 0.0002,
-      "loss": 0.5467251539230347,
-      "mean_token_accuracy": 0.7770728468894958,
-      "num_tokens": 39056858.0,
+      "loss": 0.6071447134017944,
+      "mean_token_accuracy": 0.750182680785656,
+      "num_tokens": 18502351.0,
       "step": 499
     },
     {
-      "entropy": 0.5553129017353058,
-      "epoch": 9.261682242990654,
-      "grad_norm": 0.040604446083307266,
+      "entropy": 0.6171220317482948,
+      "epoch": 3.227088402270884,
+      "grad_norm": 0.010402821935713291,
       "learning_rate": 0.0002,
-      "loss": 0.5534853935241699,
-      "mean_token_accuracy": 0.7741707265377045,
-      "num_tokens": 39137258.0,
+      "loss": 0.6183167695999146,
+      "mean_token_accuracy": 0.7468552514910698,
+      "num_tokens": 18540172.0,
       "step": 500
     },
     {
-      "entropy": 0.5694791674613953,
-      "epoch": 9.280373831775702,
-      "grad_norm": 0.05060841515660286,
+      "entropy": 0.6202204376459122,
+      "epoch": 3.2335766423357666,
+      "grad_norm": 0.008789432235062122,
       "learning_rate": 0.0002,
-      "loss": 0.5628358125686646,
-      "mean_token_accuracy": 0.769134595990181,
-      "num_tokens": 39217134.0,
+      "loss": 0.6224135160446167,
+      "mean_token_accuracy": 0.745467983186245,
+      "num_tokens": 18577734.0,
       "step": 501
     },
     {
-      "entropy": 0.5658796727657318,
-      "epoch": 9.299065420560748,
-      "grad_norm": 0.050324369221925735,
+      "entropy": 0.6132732257246971,
+      "epoch": 3.240064882400649,
+      "grad_norm": 0.010597198270261288,
       "learning_rate": 0.0002,
-      "loss": 0.5642348527908325,
-      "mean_token_accuracy": 0.7691941559314728,
-      "num_tokens": 39296226.0,
+      "loss": 0.6129106283187866,
+      "mean_token_accuracy": 0.7499266192317009,
+      "num_tokens": 18614944.0,
       "step": 502
     },
     {
-      "entropy": 0.5391237288713455,
-      "epoch": 9.317757009345794,
-      "grad_norm": 0.054539285600185394,
+      "entropy": 0.6337689533829689,
+      "epoch": 3.246553122465531,
+      "grad_norm": 0.009572361595928669,
       "learning_rate": 0.0002,
-      "loss": 0.5483332872390747,
-      "mean_token_accuracy": 0.7765073031187057,
-      "num_tokens": 39375266.0,
+      "loss": 0.6304764747619629,
+      "mean_token_accuracy": 0.7434330955147743,
+      "num_tokens": 18651997.0,
       "step": 503
     },
     {
-      "entropy": 0.5603165626525879,
-      "epoch": 9.336448598130842,
-      "grad_norm": 0.042973097413778305,
+      "entropy": 0.6075436845421791,
+      "epoch": 3.2530413625304138,
+      "grad_norm": 0.00990824680775404,
       "learning_rate": 0.0002,
-      "loss": 0.5604079961776733,
-      "mean_token_accuracy": 0.7713109701871872,
-      "num_tokens": 39453754.0,
+      "loss": 0.6082068681716919,
+      "mean_token_accuracy": 0.7516702488064766,
+      "num_tokens": 18688674.0,
       "step": 504
     },
     {
-      "entropy": 0.5705735981464386,
-      "epoch": 9.355140186915888,
-      "grad_norm": 0.047841478139162064,
+      "entropy": 0.6139610409736633,
+      "epoch": 3.259529602595296,
+      "grad_norm": 0.010237157344818115,
       "learning_rate": 0.0002,
-      "loss": 0.5631089806556702,
-      "mean_token_accuracy": 0.7698779404163361,
-      "num_tokens": 39532715.0,
+      "loss": 0.6186607480049133,
+      "mean_token_accuracy": 0.7454981952905655,
+      "num_tokens": 18726405.0,
       "step": 505
     },
     {
-      "entropy": 0.5585284233093262,
-      "epoch": 9.373831775700934,
-      "grad_norm": 0.04569873958826065,
+      "entropy": 0.6063608303666115,
+      "epoch": 3.2660178426601782,
+      "grad_norm": 0.010847094468772411,
       "learning_rate": 0.0002,
-      "loss": 0.5519518852233887,
-      "mean_token_accuracy": 0.7755726873874664,
-      "num_tokens": 39611746.0,
+      "loss": 0.6137136816978455,
+      "mean_token_accuracy": 0.7497691661119461,
+      "num_tokens": 18763668.0,
       "step": 506
     },
     {
-      "entropy": 0.5604812204837799,
-      "epoch": 9.392523364485982,
-      "grad_norm": 0.052518170326948166,
+      "entropy": 0.6163499504327774,
+      "epoch": 3.272506082725061,
+      "grad_norm": 0.010017551481723785,
       "learning_rate": 0.0002,
-      "loss": 0.5635286569595337,
-      "mean_token_accuracy": 0.770183876156807,
-      "num_tokens": 39691136.0,
+      "loss": 0.6188673973083496,
+      "mean_token_accuracy": 0.7473317757248878,
+      "num_tokens": 18800872.0,
       "step": 507
     },
     {
-      "entropy": 0.5538004338741302,
-      "epoch": 9.411214953271028,
-      "grad_norm": 0.044914163649082184,
+      "entropy": 0.6251264810562134,
+      "epoch": 3.278994322789943,
+      "grad_norm": 0.011264265514910221,
       "learning_rate": 0.0002,
-      "loss": 0.5522688627243042,
-      "mean_token_accuracy": 0.7753826528787613,
-      "num_tokens": 39769624.0,
+      "loss": 0.6177271604537964,
+      "mean_token_accuracy": 0.7485972419381142,
+      "num_tokens": 18838150.0,
       "step": 508
     },
     {
-      "entropy": 0.5605034232139587,
-      "epoch": 9.429906542056075,
-      "grad_norm": 0.04380393773317337,
+      "entropy": 0.6273062154650688,
+      "epoch": 3.285482562854826,
+      "grad_norm": 0.010041182860732079,
       "learning_rate": 0.0002,
-      "loss": 0.5553767681121826,
-      "mean_token_accuracy": 0.7720271348953247,
-      "num_tokens": 39849011.0,
+      "loss": 0.6206075549125671,
+      "mean_token_accuracy": 0.7470289915800095,
+      "num_tokens": 18875738.0,
       "step": 509
     },
     {
-      "entropy": 0.5566978305578232,
-      "epoch": 9.448598130841122,
-      "grad_norm": 0.04671022295951843,
+      "entropy": 0.6267990991473198,
+      "epoch": 3.291970802919708,
+      "grad_norm": 0.011449483223259449,
       "learning_rate": 0.0002,
-      "loss": 0.5574364066123962,
-      "mean_token_accuracy": 0.7712553292512894,
-      "num_tokens": 39927938.0,
+      "loss": 0.6346527934074402,
+      "mean_token_accuracy": 0.7385271340608597,
+      "num_tokens": 18913137.0,
       "step": 510
     },
     {
-      "entropy": 0.5586736798286438,
-      "epoch": 9.467289719626168,
-      "grad_norm": 0.056707724928855896,
+      "entropy": 0.6025072038173676,
+      "epoch": 3.2984590429845904,
+      "grad_norm": 0.010422894731163979,
       "learning_rate": 0.0002,
-      "loss": 0.5596231818199158,
-      "mean_token_accuracy": 0.7711254358291626,
-      "num_tokens": 40006081.0,
+      "loss": 0.6072125434875488,
+      "mean_token_accuracy": 0.7508590817451477,
+      "num_tokens": 18950470.0,
       "step": 511
     },
     {
-      "entropy": 0.566216766834259,
-      "epoch": 9.485981308411215,
-      "grad_norm": 0.04336009919643402,
+      "entropy": 0.6073831468820572,
+      "epoch": 3.304947283049473,
+      "grad_norm": 0.010642215609550476,
       "learning_rate": 0.0002,
-      "loss": 0.5624270439147949,
-      "mean_token_accuracy": 0.7705419957637787,
-      "num_tokens": 40084773.0,
+      "loss": 0.6133738160133362,
+      "mean_token_accuracy": 0.7492869570851326,
+      "num_tokens": 18988313.0,
       "step": 512
     },
     {
-      "entropy": 0.5668596029281616,
-      "epoch": 9.504672897196262,
-      "grad_norm": 0.07046239823102951,
+      "entropy": 0.5974380895495415,
+      "epoch": 3.3114355231143553,
+      "grad_norm": 0.010486697778105736,
       "learning_rate": 0.0002,
-      "loss": 0.5715997815132141,
-      "mean_token_accuracy": 0.7684156596660614,
-      "num_tokens": 40163757.0,
+      "loss": 0.6020730137825012,
+      "mean_token_accuracy": 0.7548841238021851,
+      "num_tokens": 19024956.0,
       "step": 513
     },
     {
-      "entropy": 0.572264775633812,
-      "epoch": 9.523364485981308,
-      "grad_norm": 0.06164439395070076,
+      "entropy": 0.6128261536359787,
+      "epoch": 3.3179237631792375,
+      "grad_norm": 0.011510336771607399,
       "learning_rate": 0.0002,
-      "loss": 0.5664000511169434,
-      "mean_token_accuracy": 0.7686414867639542,
-      "num_tokens": 40242815.0,
+      "loss": 0.6143049001693726,
+      "mean_token_accuracy": 0.7480410858988762,
+      "num_tokens": 19062179.0,
       "step": 514
     },
     {
-      "entropy": 0.5551411956548691,
-      "epoch": 9.542056074766355,
-      "grad_norm": 0.053951751440763474,
+      "entropy": 0.6077246218919754,
+      "epoch": 3.3244120032441202,
+      "grad_norm": 0.010107292793691158,
       "learning_rate": 0.0002,
-      "loss": 0.5564425587654114,
-      "mean_token_accuracy": 0.7743828147649765,
-      "num_tokens": 40321756.0,
+      "loss": 0.6046852469444275,
+      "mean_token_accuracy": 0.7533954977989197,
+      "num_tokens": 19099090.0,
       "step": 515
     },
     {
-      "entropy": 0.5653344392776489,
-      "epoch": 9.560747663551401,
-      "grad_norm": 0.05632255598902702,
+      "entropy": 0.6247019246220589,
+      "epoch": 3.3309002433090025,
+      "grad_norm": 0.010466467589139938,
       "learning_rate": 0.0002,
-      "loss": 0.5728057622909546,
-      "mean_token_accuracy": 0.7655349224805832,
-      "num_tokens": 40398936.0,
+      "loss": 0.6306489706039429,
+      "mean_token_accuracy": 0.741244375705719,
+      "num_tokens": 19136307.0,
       "step": 516
     },
     {
-      "entropy": 0.5666876137256622,
-      "epoch": 9.57943925233645,
-      "grad_norm": 0.0477413646876812,
+      "entropy": 0.6153485924005508,
+      "epoch": 3.3373884833738847,
+      "grad_norm": 0.009494316764175892,
       "learning_rate": 0.0002,
-      "loss": 0.5577602982521057,
-      "mean_token_accuracy": 0.7715498358011246,
-      "num_tokens": 40477696.0,
+      "loss": 0.6167474985122681,
+      "mean_token_accuracy": 0.7500151321291924,
+      "num_tokens": 19173544.0,
       "step": 517
     },
     {
-      "entropy": 0.5684141516685486,
-      "epoch": 9.598130841121495,
-      "grad_norm": 0.04057633876800537,
+      "entropy": 0.6086971685290337,
+      "epoch": 3.3438767234387674,
+      "grad_norm": 0.009192675352096558,
       "learning_rate": 0.0002,
-      "loss": 0.560270369052887,
-      "mean_token_accuracy": 0.7696860134601593,
-      "num_tokens": 40556332.0,
+      "loss": 0.6096853017807007,
+      "mean_token_accuracy": 0.7501685246825218,
+      "num_tokens": 19210386.0,
       "step": 518
     },
     {
-      "entropy": 0.55315001308918,
-      "epoch": 9.616822429906541,
-      "grad_norm": 0.07429754734039307,
+      "entropy": 0.6297651901841164,
+      "epoch": 3.3503649635036497,
+      "grad_norm": 0.009902825579047203,
       "learning_rate": 0.0002,
-      "loss": 0.5606566667556763,
-      "mean_token_accuracy": 0.7719446271657944,
-      "num_tokens": 40634410.0,
+      "loss": 0.6314697265625,
+      "mean_token_accuracy": 0.7417091131210327,
+      "num_tokens": 19247761.0,
       "step": 519
     },
     {
-      "entropy": 0.5610033273696899,
-      "epoch": 9.63551401869159,
-      "grad_norm": 0.06447485834360123,
+      "entropy": 0.6149463355541229,
+      "epoch": 3.356853203568532,
+      "grad_norm": 0.009715819731354713,
       "learning_rate": 0.0002,
-      "loss": 0.5673164129257202,
-      "mean_token_accuracy": 0.7681170105934143,
-      "num_tokens": 40712496.0,
+      "loss": 0.6137614250183105,
+      "mean_token_accuracy": 0.7476155012845993,
+      "num_tokens": 19284908.0,
       "step": 520
     },
     {
-      "entropy": 0.5906848162412643,
-      "epoch": 9.654205607476635,
-      "grad_norm": 0.051934584975242615,
+      "entropy": 0.6115437000989914,
+      "epoch": 3.3633414436334146,
+      "grad_norm": 0.009652029722929,
       "learning_rate": 0.0002,
-      "loss": 0.5791685581207275,
-      "mean_token_accuracy": 0.764894962310791,
-      "num_tokens": 40790409.0,
+      "loss": 0.6114376783370972,
+      "mean_token_accuracy": 0.7499561905860901,
+      "num_tokens": 19321727.0,
       "step": 521
     },
     {
-      "entropy": 0.5693906992673874,
-      "epoch": 9.672897196261681,
-      "grad_norm": 0.05814937874674797,
+      "entropy": 0.6143479123711586,
+      "epoch": 3.369829683698297,
+      "grad_norm": 0.010150844231247902,
       "learning_rate": 0.0002,
-      "loss": 0.5632771849632263,
-      "mean_token_accuracy": 0.7700486779212952,
-      "num_tokens": 40870640.0,
+      "loss": 0.6162387132644653,
+      "mean_token_accuracy": 0.7469704374670982,
+      "num_tokens": 19359065.0,
       "step": 522
     },
     {
-      "entropy": 0.5577443987131119,
-      "epoch": 9.69158878504673,
-      "grad_norm": 0.05073021724820137,
+      "entropy": 0.6042628288269043,
+      "epoch": 3.376317923763179,
+      "grad_norm": 0.010199622251093388,
       "learning_rate": 0.0002,
-      "loss": 0.5627031326293945,
-      "mean_token_accuracy": 0.7702322006225586,
-      "num_tokens": 40949427.0,
+      "loss": 0.607587456703186,
+      "mean_token_accuracy": 0.7510121315717697,
+      "num_tokens": 19396140.0,
       "step": 523
     },
     {
-      "entropy": 0.5583484172821045,
-      "epoch": 9.710280373831775,
-      "grad_norm": 0.06356833130121231,
+      "entropy": 0.6103994250297546,
+      "epoch": 3.3828061638280618,
+      "grad_norm": 0.009757442399859428,
       "learning_rate": 0.0002,
-      "loss": 0.5622925162315369,
-      "mean_token_accuracy": 0.7712053507566452,
-      "num_tokens": 41028516.0,
+      "loss": 0.6118942499160767,
+      "mean_token_accuracy": 0.7493396252393723,
+      "num_tokens": 19433282.0,
       "step": 524
     },
     {
-      "entropy": 0.5624096095561981,
-      "epoch": 9.728971962616823,
-      "grad_norm": 0.051840588450431824,
+      "entropy": 0.6115303710103035,
+      "epoch": 3.389294403892944,
+      "grad_norm": 0.009843776933848858,
       "learning_rate": 0.0002,
-      "loss": 0.557092010974884,
-      "mean_token_accuracy": 0.7728729248046875,
-      "num_tokens": 41108171.0,
+      "loss": 0.6144024133682251,
+      "mean_token_accuracy": 0.7491987869143486,
+      "num_tokens": 19470735.0,
       "step": 525
     },
     {
-      "entropy": 0.5580566078424454,
-      "epoch": 9.74766355140187,
-      "grad_norm": 0.054242074489593506,
+      "entropy": 0.6124155074357986,
+      "epoch": 3.3957826439578263,
+      "grad_norm": 0.012832233682274818,
       "learning_rate": 0.0002,
-      "loss": 0.558272123336792,
-      "mean_token_accuracy": 0.7722387164831161,
-      "num_tokens": 41186785.0,
+      "loss": 0.6174359321594238,
+      "mean_token_accuracy": 0.7484075799584389,
+      "num_tokens": 19508279.0,
       "step": 526
     },
     {
-      "entropy": 0.5527939796447754,
-      "epoch": 9.766355140186915,
-      "grad_norm": 0.05423031747341156,
+      "entropy": 0.621652752161026,
+      "epoch": 3.402270884022709,
+      "grad_norm": 0.010444359853863716,
       "learning_rate": 0.0002,
-      "loss": 0.5510618090629578,
-      "mean_token_accuracy": 0.7753184139728546,
-      "num_tokens": 41266511.0,
+      "loss": 0.6202374696731567,
+      "mean_token_accuracy": 0.747683048248291,
+      "num_tokens": 19545629.0,
       "step": 527
     },
     {
-      "entropy": 0.5567087680101395,
-      "epoch": 9.785046728971963,
-      "grad_norm": 0.059091318398714066,
+      "entropy": 0.6133333742618561,
+      "epoch": 3.408759124087591,
+      "grad_norm": 0.01157453190535307,
       "learning_rate": 0.0002,
-      "loss": 0.5601646900177002,
-      "mean_token_accuracy": 0.7721138298511505,
-      "num_tokens": 41344767.0,
+      "loss": 0.6190609931945801,
+      "mean_token_accuracy": 0.7481984347105026,
+      "num_tokens": 19583135.0,
       "step": 528
     },
     {
-      "entropy": 0.5563396960496902,
-      "epoch": 9.80373831775701,
-      "grad_norm": 0.041039131581783295,
+      "entropy": 0.6031940132379532,
+      "epoch": 3.4152473641524734,
+      "grad_norm": 0.01220928505063057,
       "learning_rate": 0.0002,
-      "loss": 0.5506303310394287,
-      "mean_token_accuracy": 0.7754775285720825,
-      "num_tokens": 41424990.0,
+      "loss": 0.60877525806427,
+      "mean_token_accuracy": 0.7505937144160271,
+      "num_tokens": 19620410.0,
       "step": 529
     },
     {
-      "entropy": 0.567491352558136,
-      "epoch": 9.822429906542055,
-      "grad_norm": 0.05270151048898697,
+      "entropy": 0.6026583164930344,
+      "epoch": 3.421735604217356,
+      "grad_norm": 0.009105565026402473,
       "learning_rate": 0.0002,
-      "loss": 0.5634373426437378,
-      "mean_token_accuracy": 0.770146906375885,
-      "num_tokens": 41503925.0,
+      "loss": 0.6003735661506653,
+      "mean_token_accuracy": 0.7538656741380692,
+      "num_tokens": 19657757.0,
       "step": 530
     },
     {
-      "entropy": 0.5676887929439545,
-      "epoch": 9.841121495327103,
-      "grad_norm": 0.044822655618190765,
+      "entropy": 0.6212241724133492,
+      "epoch": 3.4282238442822384,
+      "grad_norm": 0.012120554223656654,
       "learning_rate": 0.0002,
-      "loss": 0.568418025970459,
-      "mean_token_accuracy": 0.7675944864749908,
-      "num_tokens": 41583158.0,
+      "loss": 0.6169411540031433,
+      "mean_token_accuracy": 0.7489817962050438,
+      "num_tokens": 19694775.0,
       "step": 531
     },
     {
-      "entropy": 0.5546302497386932,
-      "epoch": 9.85981308411215,
-      "grad_norm": 0.055056799203157425,
+      "entropy": 0.611191414296627,
+      "epoch": 3.4347120843471206,
+      "grad_norm": 0.010209856554865837,
       "learning_rate": 0.0002,
-      "loss": 0.5574401021003723,
-      "mean_token_accuracy": 0.7734245508909225,
-      "num_tokens": 41661822.0,
+      "loss": 0.6133028268814087,
+      "mean_token_accuracy": 0.7480586767196655,
+      "num_tokens": 19731954.0,
       "step": 532
     },
     {
-      "entropy": 0.5644202679395676,
-      "epoch": 9.878504672897197,
-      "grad_norm": 0.048922717571258545,
+      "entropy": 0.6154542937874794,
+      "epoch": 3.4412003244120033,
+      "grad_norm": 0.009237438440322876,
       "learning_rate": 0.0002,
-      "loss": 0.5685901641845703,
-      "mean_token_accuracy": 0.7667403668165207,
-      "num_tokens": 41740079.0,
+      "loss": 0.6181926131248474,
+      "mean_token_accuracy": 0.7482445612549782,
+      "num_tokens": 19769550.0,
       "step": 533
     },
     {
-      "entropy": 0.5666642636060715,
-      "epoch": 9.897196261682243,
-      "grad_norm": 0.048663362860679626,
+      "entropy": 0.6260659247636795,
+      "epoch": 3.4476885644768855,
+      "grad_norm": 0.011564482003450394,
       "learning_rate": 0.0002,
-      "loss": 0.5589666962623596,
-      "mean_token_accuracy": 0.7732465118169785,
-      "num_tokens": 41818436.0,
+      "loss": 0.6311028599739075,
+      "mean_token_accuracy": 0.7445508390665054,
+      "num_tokens": 19806694.0,
       "step": 534
     },
     {
-      "entropy": 0.5761356949806213,
-      "epoch": 9.91588785046729,
-      "grad_norm": 0.042342960834503174,
+      "entropy": 0.6114797592163086,
+      "epoch": 3.4541768045417682,
+      "grad_norm": 0.011250223033130169,
       "learning_rate": 0.0002,
-      "loss": 0.570763349533081,
-      "mean_token_accuracy": 0.7665978372097015,
-      "num_tokens": 41896161.0,
+      "loss": 0.6188318133354187,
+      "mean_token_accuracy": 0.7451098337769508,
+      "num_tokens": 19843897.0,
       "step": 535
     },
     {
-      "entropy": 0.5440604090690613,
-      "epoch": 9.934579439252337,
-      "grad_norm": 0.043155740946531296,
+      "entropy": 0.6260335817933083,
+      "epoch": 3.4606650446066505,
+      "grad_norm": 0.00978784915059805,
       "learning_rate": 0.0002,
-      "loss": 0.549519419670105,
-      "mean_token_accuracy": 0.7761891782283783,
-      "num_tokens": 41976573.0,
+      "loss": 0.6255401372909546,
+      "mean_token_accuracy": 0.7468025237321854,
+      "num_tokens": 19881321.0,
       "step": 536
     },
     {
-      "entropy": 0.5546491295099258,
-      "epoch": 9.953271028037383,
-      "grad_norm": 0.0412793830037117,
+      "entropy": 0.6212326288223267,
+      "epoch": 3.4671532846715327,
+      "grad_norm": 0.013453345745801926,
       "learning_rate": 0.0002,
-      "loss": 0.5588980913162231,
-      "mean_token_accuracy": 0.7721990048885345,
-      "num_tokens": 42055586.0,
+      "loss": 0.611083984375,
+      "mean_token_accuracy": 0.7521479055285454,
+      "num_tokens": 19918768.0,
       "step": 537
     },
     {
-      "entropy": 0.5679792314767838,
-      "epoch": 9.97196261682243,
-      "grad_norm": 0.03634324297308922,
+      "entropy": 0.6172696650028229,
+      "epoch": 3.4736415247364154,
+      "grad_norm": 0.010021936148405075,
       "learning_rate": 0.0002,
-      "loss": 0.5597659945487976,
-      "mean_token_accuracy": 0.7712419778108597,
-      "num_tokens": 42134080.0,
+      "loss": 0.6165583729743958,
+      "mean_token_accuracy": 0.7488679885864258,
+      "num_tokens": 19956374.0,
       "step": 538
     },
     {
-      "entropy": 0.5654930472373962,
-      "epoch": 9.990654205607477,
-      "grad_norm": 0.034971874207258224,
+      "entropy": 0.5984770804643631,
+      "epoch": 3.4801297648012977,
+      "grad_norm": 0.010912574827671051,
       "learning_rate": 0.0002,
-      "loss": 0.5591872334480286,
-      "mean_token_accuracy": 0.7716458737850189,
-      "num_tokens": 42212830.0,
+      "loss": 0.6038742065429688,
+      "mean_token_accuracy": 0.752843476831913,
+      "num_tokens": 19993165.0,
       "step": 539
     },
     {
-      "entropy": 0.5517400205135345,
-      "epoch": 10.0,
-      "grad_norm": 0.04363137111067772,
+      "entropy": 0.6006497591733932,
+      "epoch": 3.48661800486618,
+      "grad_norm": 0.010195601731538773,
       "learning_rate": 0.0002,
-      "loss": 0.5499259233474731,
-      "mean_token_accuracy": 0.7756347954273224,
-      "num_tokens": 42252523.0,
+      "loss": 0.6058304905891418,
+      "mean_token_accuracy": 0.7511888965964317,
+      "num_tokens": 20030632.0,
       "step": 540
+    },
+    {
+      "entropy": 0.603807657957077,
+      "epoch": 3.4931062449310626,
+      "grad_norm": 0.010789190419018269,
+      "learning_rate": 0.0002,
+      "loss": 0.6101077198982239,
+      "mean_token_accuracy": 0.7498170882463455,
+      "num_tokens": 20067720.0,
+      "step": 541
+    },
+    {
+      "entropy": 0.6094614043831825,
+      "epoch": 3.499594484995945,
+      "grad_norm": 0.010353523306548595,
+      "learning_rate": 0.0002,
+      "loss": 0.6108995676040649,
+      "mean_token_accuracy": 0.7505486905574799,
+      "num_tokens": 20105092.0,
+      "step": 542
+    },
+    {
+      "entropy": 0.6127925515174866,
+      "epoch": 3.5060827250608275,
+      "grad_norm": 0.011608858592808247,
+      "learning_rate": 0.0002,
+      "loss": 0.6113302111625671,
+      "mean_token_accuracy": 0.7498401030898094,
+      "num_tokens": 20142060.0,
+      "step": 543
+    },
+    {
+      "entropy": 0.6288227662444115,
+      "epoch": 3.5125709651257098,
+      "grad_norm": 0.00944559182971716,
+      "learning_rate": 0.0002,
+      "loss": 0.6295347213745117,
+      "mean_token_accuracy": 0.7438011020421982,
+      "num_tokens": 20179321.0,
+      "step": 544
+    },
+    {
+      "entropy": 0.6155826225876808,
+      "epoch": 3.519059205190592,
+      "grad_norm": 0.010050074197351933,
+      "learning_rate": 0.0002,
+      "loss": 0.6170110106468201,
+      "mean_token_accuracy": 0.7460537254810333,
+      "num_tokens": 20216788.0,
+      "step": 545
+    },
+    {
+      "entropy": 0.6109997108578682,
+      "epoch": 3.5255474452554747,
+      "grad_norm": 0.010288557037711143,
+      "learning_rate": 0.0002,
+      "loss": 0.6153451204299927,
+      "mean_token_accuracy": 0.747114397585392,
+      "num_tokens": 20254482.0,
+      "step": 546
+    },
+    {
+      "entropy": 0.6220259815454483,
+      "epoch": 3.532035685320357,
+      "grad_norm": 0.010569143109023571,
+      "learning_rate": 0.0002,
+      "loss": 0.6245878338813782,
+      "mean_token_accuracy": 0.745466060936451,
+      "num_tokens": 20291323.0,
+      "step": 547
+    },
+    {
+      "entropy": 0.6251974999904633,
+      "epoch": 3.538523925385239,
+      "grad_norm": 0.010512638837099075,
+      "learning_rate": 0.0002,
+      "loss": 0.6209877729415894,
+      "mean_token_accuracy": 0.7460760995745659,
+      "num_tokens": 20328541.0,
+      "step": 548
+    },
+    {
+      "entropy": 0.6372380778193474,
+      "epoch": 3.545012165450122,
+      "grad_norm": 0.009989661164581776,
+      "learning_rate": 0.0002,
+      "loss": 0.63493812084198,
+      "mean_token_accuracy": 0.7409111261367798,
+      "num_tokens": 20366154.0,
+      "step": 549
+    },
+    {
+      "entropy": 0.616200253367424,
+      "epoch": 3.551500405515004,
+      "grad_norm": 0.009802714921534061,
+      "learning_rate": 0.0002,
+      "loss": 0.6198533773422241,
+      "mean_token_accuracy": 0.7493973523378372,
+      "num_tokens": 20403539.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.6109699085354805,
+      "epoch": 3.5579886455798864,
+      "grad_norm": 0.011730330996215343,
+      "learning_rate": 0.0002,
+      "loss": 0.6169507503509521,
+      "mean_token_accuracy": 0.7469825968146324,
+      "num_tokens": 20440857.0,
+      "step": 551
+    },
+    {
+      "entropy": 0.6107888966798782,
+      "epoch": 3.564476885644769,
+      "grad_norm": 0.010902747511863708,
+      "learning_rate": 0.0002,
+      "loss": 0.6169427633285522,
+      "mean_token_accuracy": 0.7473291158676147,
+      "num_tokens": 20478685.0,
+      "step": 552
+    },
+    {
+      "entropy": 0.6232053637504578,
+      "epoch": 3.5709651257096513,
+      "grad_norm": 0.01024220883846283,
+      "learning_rate": 0.0002,
+      "loss": 0.6223392486572266,
+      "mean_token_accuracy": 0.7471131235361099,
+      "num_tokens": 20516186.0,
+      "step": 553
+    },
+    {
+      "entropy": 0.6318716630339622,
+      "epoch": 3.5774533657745335,
+      "grad_norm": 0.00972492340952158,
+      "learning_rate": 0.0002,
+      "loss": 0.6295208930969238,
+      "mean_token_accuracy": 0.7421210557222366,
+      "num_tokens": 20553418.0,
+      "step": 554
+    },
+    {
+      "entropy": 0.6323020830750465,
+      "epoch": 3.5839416058394162,
+      "grad_norm": 0.009747047908604145,
+      "learning_rate": 0.0002,
+      "loss": 0.6291106343269348,
+      "mean_token_accuracy": 0.742633767426014,
+      "num_tokens": 20590921.0,
+      "step": 555
+    },
+    {
+      "entropy": 0.6193426549434662,
+      "epoch": 3.5904298459042985,
+      "grad_norm": 0.009529096074402332,
+      "learning_rate": 0.0002,
+      "loss": 0.622657299041748,
+      "mean_token_accuracy": 0.7454424574971199,
+      "num_tokens": 20628134.0,
+      "step": 556
+    },
+    {
+      "entropy": 0.606108270585537,
+      "epoch": 3.5969180859691807,
+      "grad_norm": 0.01090499758720398,
+      "learning_rate": 0.0002,
+      "loss": 0.6156042218208313,
+      "mean_token_accuracy": 0.7475811764597893,
+      "num_tokens": 20665529.0,
+      "step": 557
+    },
+    {
+      "entropy": 0.6024795696139336,
+      "epoch": 3.6034063260340634,
+      "grad_norm": 0.01122215110808611,
+      "learning_rate": 0.0002,
+      "loss": 0.6096351146697998,
+      "mean_token_accuracy": 0.7500594854354858,
+      "num_tokens": 20702623.0,
+      "step": 558
+    },
+    {
+      "entropy": 0.6050852239131927,
+      "epoch": 3.6098945660989457,
+      "grad_norm": 0.008911820128560066,
+      "learning_rate": 0.0002,
+      "loss": 0.6063804626464844,
+      "mean_token_accuracy": 0.750556543469429,
+      "num_tokens": 20739535.0,
+      "step": 559
+    },
+    {
+      "entropy": 0.6184356659650803,
+      "epoch": 3.616382806163828,
+      "grad_norm": 0.008947025053203106,
+      "learning_rate": 0.0002,
+      "loss": 0.6186281442642212,
+      "mean_token_accuracy": 0.7470304444432259,
+      "num_tokens": 20776900.0,
+      "step": 560
+    },
+    {
+      "entropy": 0.6115831136703491,
+      "epoch": 3.6228710462287106,
+      "grad_norm": 0.010605190880596638,
+      "learning_rate": 0.0002,
+      "loss": 0.6092396974563599,
+      "mean_token_accuracy": 0.750456377863884,
+      "num_tokens": 20814326.0,
+      "step": 561
+    },
+    {
+      "entropy": 0.6234035491943359,
+      "epoch": 3.629359286293593,
+      "grad_norm": 0.010619143024086952,
+      "learning_rate": 0.0002,
+      "loss": 0.622122585773468,
+      "mean_token_accuracy": 0.7449623718857765,
+      "num_tokens": 20851735.0,
+      "step": 562
+    },
+    {
+      "entropy": 0.6131063625216484,
+      "epoch": 3.635847526358475,
+      "grad_norm": 0.009914328344166279,
+      "learning_rate": 0.0002,
+      "loss": 0.6129999160766602,
+      "mean_token_accuracy": 0.7512071952223778,
+      "num_tokens": 20888959.0,
+      "step": 563
+    },
+    {
+      "entropy": 0.6103895530104637,
+      "epoch": 3.6423357664233578,
+      "grad_norm": 0.009590883739292622,
+      "learning_rate": 0.0002,
+      "loss": 0.6121896505355835,
+      "mean_token_accuracy": 0.7519041821360588,
+      "num_tokens": 20926880.0,
+      "step": 564
+    },
+    {
+      "entropy": 0.6107024624943733,
+      "epoch": 3.64882400648824,
+      "grad_norm": 0.010136992670595646,
+      "learning_rate": 0.0002,
+      "loss": 0.615648090839386,
+      "mean_token_accuracy": 0.7478983402252197,
+      "num_tokens": 20964278.0,
+      "step": 565
+    },
+    {
+      "entropy": 0.6019061878323555,
+      "epoch": 3.6553122465531223,
+      "grad_norm": 0.01197731215506792,
+      "learning_rate": 0.0002,
+      "loss": 0.6083622574806213,
+      "mean_token_accuracy": 0.752247080206871,
+      "num_tokens": 21001703.0,
+      "step": 566
+    },
+    {
+      "entropy": 0.6136230602860451,
+      "epoch": 3.661800486618005,
+      "grad_norm": 0.010383468121290207,
+      "learning_rate": 0.0002,
+      "loss": 0.6142981052398682,
+      "mean_token_accuracy": 0.750029981136322,
+      "num_tokens": 21039545.0,
+      "step": 567
+    },
+    {
+      "entropy": 0.6270170286297798,
+      "epoch": 3.668288726682887,
+      "grad_norm": 0.01222953386604786,
+      "learning_rate": 0.0002,
+      "loss": 0.622101902961731,
+      "mean_token_accuracy": 0.7463989406824112,
+      "num_tokens": 21076601.0,
+      "step": 568
+    },
+    {
+      "entropy": 0.635653018951416,
+      "epoch": 3.6747769667477694,
+      "grad_norm": 0.009029991924762726,
+      "learning_rate": 0.0002,
+      "loss": 0.6368217468261719,
+      "mean_token_accuracy": 0.739650160074234,
+      "num_tokens": 21113628.0,
+      "step": 569
+    },
+    {
+      "entropy": 0.6155265644192696,
+      "epoch": 3.681265206812652,
+      "grad_norm": 0.010669651441276073,
+      "learning_rate": 0.0002,
+      "loss": 0.6169142723083496,
+      "mean_token_accuracy": 0.7483588233590126,
+      "num_tokens": 21150841.0,
+      "step": 570
+    },
+    {
+      "entropy": 0.6179278939962387,
+      "epoch": 3.6877534468775344,
+      "grad_norm": 0.01090275775641203,
+      "learning_rate": 0.0002,
+      "loss": 0.620256781578064,
+      "mean_token_accuracy": 0.7459961548447609,
+      "num_tokens": 21188047.0,
+      "step": 571
+    },
+    {
+      "entropy": 0.6013471409678459,
+      "epoch": 3.6942416869424166,
+      "grad_norm": 0.013129356317222118,
+      "learning_rate": 0.0002,
+      "loss": 0.6093657612800598,
+      "mean_token_accuracy": 0.7505828440189362,
+      "num_tokens": 21224977.0,
+      "step": 572
+    },
+    {
+      "entropy": 0.6192966923117638,
+      "epoch": 3.7007299270072993,
+      "grad_norm": 0.01149990689009428,
+      "learning_rate": 0.0002,
+      "loss": 0.624458909034729,
+      "mean_token_accuracy": 0.7465271949768066,
+      "num_tokens": 21262229.0,
+      "step": 573
+    },
+    {
+      "entropy": 0.6305530145764351,
+      "epoch": 3.7072181670721815,
+      "grad_norm": 0.011260929517447948,
+      "learning_rate": 0.0002,
+      "loss": 0.6282562613487244,
+      "mean_token_accuracy": 0.7442766949534416,
+      "num_tokens": 21299934.0,
+      "step": 574
+    },
+    {
+      "entropy": 0.6216496974229813,
+      "epoch": 3.7137064071370642,
+      "grad_norm": 0.011454667896032333,
+      "learning_rate": 0.0002,
+      "loss": 0.6176272630691528,
+      "mean_token_accuracy": 0.7486661300063133,
+      "num_tokens": 21337298.0,
+      "step": 575
+    },
+    {
+      "entropy": 0.6251465007662773,
+      "epoch": 3.7201946472019465,
+      "grad_norm": 0.011989898979663849,
+      "learning_rate": 0.0002,
+      "loss": 0.6237755417823792,
+      "mean_token_accuracy": 0.7449339032173157,
+      "num_tokens": 21374338.0,
+      "step": 576
+    },
+    {
+      "entropy": 0.6160696297883987,
+      "epoch": 3.7266828872668287,
+      "grad_norm": 0.009978833608329296,
+      "learning_rate": 0.0002,
+      "loss": 0.6201364398002625,
+      "mean_token_accuracy": 0.7472834289073944,
+      "num_tokens": 21411412.0,
+      "step": 577
+    },
+    {
+      "entropy": 0.6086283698678017,
+      "epoch": 3.7331711273317114,
+      "grad_norm": 0.01127020176500082,
+      "learning_rate": 0.0002,
+      "loss": 0.6133439540863037,
+      "mean_token_accuracy": 0.7486971095204353,
+      "num_tokens": 21449184.0,
+      "step": 578
+    },
+    {
+      "entropy": 0.6155035272240639,
+      "epoch": 3.7396593673965937,
+      "grad_norm": 0.009915207512676716,
+      "learning_rate": 0.0002,
+      "loss": 0.6193184852600098,
+      "mean_token_accuracy": 0.7452678084373474,
+      "num_tokens": 21486452.0,
+      "step": 579
+    },
+    {
+      "entropy": 0.6227790787816048,
+      "epoch": 3.7461476074614763,
+      "grad_norm": 0.009025565348565578,
+      "learning_rate": 0.0002,
+      "loss": 0.6250243186950684,
+      "mean_token_accuracy": 0.7430055662989616,
+      "num_tokens": 21524049.0,
+      "step": 580
+    },
+    {
+      "entropy": 0.6145926639437675,
+      "epoch": 3.7526358475263586,
+      "grad_norm": 0.010546623729169369,
+      "learning_rate": 0.0002,
+      "loss": 0.6121389269828796,
+      "mean_token_accuracy": 0.7498128041625023,
+      "num_tokens": 21560910.0,
+      "step": 581
+    },
+    {
+      "entropy": 0.6132489815354347,
+      "epoch": 3.759124087591241,
+      "grad_norm": 0.01081460528075695,
+      "learning_rate": 0.0002,
+      "loss": 0.6121793389320374,
+      "mean_token_accuracy": 0.7495040893554688,
+      "num_tokens": 21598124.0,
+      "step": 582
+    },
+    {
+      "entropy": 0.6323688253760338,
+      "epoch": 3.7656123276561235,
+      "grad_norm": 0.010107534937560558,
+      "learning_rate": 0.0002,
+      "loss": 0.6294635534286499,
+      "mean_token_accuracy": 0.7433409690856934,
+      "num_tokens": 21635437.0,
+      "step": 583
+    },
+    {
+      "entropy": 0.6125660836696625,
+      "epoch": 3.7721005677210058,
+      "grad_norm": 0.00997902825474739,
+      "learning_rate": 0.0002,
+      "loss": 0.6156618595123291,
+      "mean_token_accuracy": 0.7480011209845543,
+      "num_tokens": 21672898.0,
+      "step": 584
+    },
+    {
+      "entropy": 0.6139216870069504,
+      "epoch": 3.778588807785888,
+      "grad_norm": 0.01199332345277071,
+      "learning_rate": 0.0002,
+      "loss": 0.6214488744735718,
+      "mean_token_accuracy": 0.748315691947937,
+      "num_tokens": 21710166.0,
+      "step": 585
+    },
+    {
+      "entropy": 0.5991447791457176,
+      "epoch": 3.7850770478507707,
+      "grad_norm": 0.01200946606695652,
+      "learning_rate": 0.0002,
+      "loss": 0.6080783605575562,
+      "mean_token_accuracy": 0.7511922717094421,
+      "num_tokens": 21746897.0,
+      "step": 586
+    },
+    {
+      "entropy": 0.6105977669358253,
+      "epoch": 3.791565287915653,
+      "grad_norm": 0.011675121262669563,
+      "learning_rate": 0.0002,
+      "loss": 0.6159127950668335,
+      "mean_token_accuracy": 0.7491589412093163,
+      "num_tokens": 21784354.0,
+      "step": 587
+    },
+    {
+      "entropy": 0.6225078999996185,
+      "epoch": 3.798053527980535,
+      "grad_norm": 0.00965933594852686,
+      "learning_rate": 0.0002,
+      "loss": 0.6214642524719238,
+      "mean_token_accuracy": 0.7455714717507362,
+      "num_tokens": 21821856.0,
+      "step": 588
+    },
+    {
+      "entropy": 0.627250537276268,
+      "epoch": 3.804541768045418,
+      "grad_norm": 0.010286493226885796,
+      "learning_rate": 0.0002,
+      "loss": 0.6219584345817566,
+      "mean_token_accuracy": 0.7479822859168053,
+      "num_tokens": 21858968.0,
+      "step": 589
+    },
+    {
+      "entropy": 0.6303782388567924,
+      "epoch": 3.8110300081103,
+      "grad_norm": 0.009684202261269093,
+      "learning_rate": 0.0002,
+      "loss": 0.629400372505188,
+      "mean_token_accuracy": 0.7433283776044846,
+      "num_tokens": 21896486.0,
+      "step": 590
+    },
+    {
+      "entropy": 0.6274826601147652,
+      "epoch": 3.8175182481751824,
+      "grad_norm": 0.00968723464757204,
+      "learning_rate": 0.0002,
+      "loss": 0.6308175325393677,
+      "mean_token_accuracy": 0.7409676983952522,
+      "num_tokens": 21933746.0,
+      "step": 591
+    },
+    {
+      "entropy": 0.607462614774704,
+      "epoch": 3.824006488240065,
+      "grad_norm": 0.010024085640907288,
+      "learning_rate": 0.0002,
+      "loss": 0.6111215353012085,
+      "mean_token_accuracy": 0.7507134452462196,
+      "num_tokens": 21970727.0,
+      "step": 592
+    },
+    {
+      "entropy": 0.6113528236746788,
+      "epoch": 3.8304947283049473,
+      "grad_norm": 0.009630702435970306,
+      "learning_rate": 0.0002,
+      "loss": 0.615958571434021,
+      "mean_token_accuracy": 0.7483637630939484,
+      "num_tokens": 22008126.0,
+      "step": 593
+    },
+    {
+      "entropy": 0.6118894517421722,
+      "epoch": 3.8369829683698295,
+      "grad_norm": 0.00891201663762331,
+      "learning_rate": 0.0002,
+      "loss": 0.6150209903717041,
+      "mean_token_accuracy": 0.7505509406328201,
+      "num_tokens": 22045058.0,
+      "step": 594
+    },
+    {
+      "entropy": 0.6181119605898857,
+      "epoch": 3.8434712084347122,
+      "grad_norm": 0.00870099663734436,
+      "learning_rate": 0.0002,
+      "loss": 0.6192715764045715,
+      "mean_token_accuracy": 0.7477136552333832,
+      "num_tokens": 22082485.0,
+      "step": 595
+    },
+    {
+      "entropy": 0.6250515654683113,
+      "epoch": 3.8499594484995945,
+      "grad_norm": 0.00869520753622055,
+      "learning_rate": 0.0002,
+      "loss": 0.6237080097198486,
+      "mean_token_accuracy": 0.7448958083987236,
+      "num_tokens": 22120186.0,
+      "step": 596
+    },
+    {
+      "entropy": 0.6200907677412033,
+      "epoch": 3.8564476885644767,
+      "grad_norm": 0.008723787032067776,
+      "learning_rate": 0.0002,
+      "loss": 0.6199079751968384,
+      "mean_token_accuracy": 0.7453882619738579,
+      "num_tokens": 22157538.0,
+      "step": 597
+    },
+    {
+      "entropy": 0.6176511943340302,
+      "epoch": 3.8629359286293594,
+      "grad_norm": 0.010193057358264923,
+      "learning_rate": 0.0002,
+      "loss": 0.6177176237106323,
+      "mean_token_accuracy": 0.7501931935548782,
+      "num_tokens": 22195003.0,
+      "step": 598
+    },
+    {
+      "entropy": 0.6236204281449318,
+      "epoch": 3.8694241686942417,
+      "grad_norm": 0.00979641918092966,
+      "learning_rate": 0.0002,
+      "loss": 0.6303654909133911,
+      "mean_token_accuracy": 0.7406158596277237,
+      "num_tokens": 22232607.0,
+      "step": 599
+    },
+    {
+      "entropy": 0.6202436164021492,
+      "epoch": 3.875912408759124,
+      "grad_norm": 0.010212398134171963,
+      "learning_rate": 0.0002,
+      "loss": 0.6213873028755188,
+      "mean_token_accuracy": 0.7464328482747078,
+      "num_tokens": 22269920.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.613631397485733,
+      "epoch": 3.8824006488240066,
+      "grad_norm": 0.009876083582639694,
+      "learning_rate": 0.0002,
+      "loss": 0.6182212829589844,
+      "mean_token_accuracy": 0.7480113580822945,
+      "num_tokens": 22306694.0,
+      "step": 601
+    },
+    {
+      "entropy": 0.6167808622121811,
+      "epoch": 3.888888888888889,
+      "grad_norm": 0.010310813784599304,
+      "learning_rate": 0.0002,
+      "loss": 0.6251655220985413,
+      "mean_token_accuracy": 0.7455293834209442,
+      "num_tokens": 22343827.0,
+      "step": 602
+    },
+    {
+      "entropy": 0.6132790520787239,
+      "epoch": 3.895377128953771,
+      "grad_norm": 0.009235870093107224,
+      "learning_rate": 0.0002,
+      "loss": 0.6129046678543091,
+      "mean_token_accuracy": 0.7505499720573425,
+      "num_tokens": 22381214.0,
+      "step": 603
+    },
+    {
+      "entropy": 0.6209843009710312,
+      "epoch": 3.9018653690186538,
+      "grad_norm": 0.009926384314894676,
+      "learning_rate": 0.0002,
+      "loss": 0.6192604899406433,
+      "mean_token_accuracy": 0.7472119927406311,
+      "num_tokens": 22418601.0,
+      "step": 604
+    },
+    {
+      "entropy": 0.6364298164844513,
+      "epoch": 3.908353609083536,
+      "grad_norm": 0.011292664334177971,
+      "learning_rate": 0.0002,
+      "loss": 0.6354763507843018,
+      "mean_token_accuracy": 0.742254801094532,
+      "num_tokens": 22455732.0,
+      "step": 605
+    },
+    {
+      "entropy": 0.6375604569911957,
+      "epoch": 3.9148418491484183,
+      "grad_norm": 0.009769946336746216,
+      "learning_rate": 0.0002,
+      "loss": 0.6385936737060547,
+      "mean_token_accuracy": 0.7394958734512329,
+      "num_tokens": 22492996.0,
+      "step": 606
+    },
+    {
+      "entropy": 0.6145768612623215,
+      "epoch": 3.921330089213301,
+      "grad_norm": 0.010400141589343548,
+      "learning_rate": 0.0002,
+      "loss": 0.6150961518287659,
+      "mean_token_accuracy": 0.7494016960263252,
+      "num_tokens": 22530887.0,
+      "step": 607
+    },
+    {
+      "entropy": 0.6110121458768845,
+      "epoch": 3.927818329278183,
+      "grad_norm": 0.01383623294532299,
+      "learning_rate": 0.0002,
+      "loss": 0.6156477332115173,
+      "mean_token_accuracy": 0.7485735863447189,
+      "num_tokens": 22568066.0,
+      "step": 608
+    },
+    {
+      "entropy": 0.6164045110344887,
+      "epoch": 3.9343065693430654,
+      "grad_norm": 0.011148249730467796,
+      "learning_rate": 0.0002,
+      "loss": 0.6243358850479126,
+      "mean_token_accuracy": 0.7457598373293877,
+      "num_tokens": 22605687.0,
+      "step": 609
+    },
+    {
+      "entropy": 0.6160785332322121,
+      "epoch": 3.940794809407948,
+      "grad_norm": 0.011384137906134129,
+      "learning_rate": 0.0002,
+      "loss": 0.6210045218467712,
+      "mean_token_accuracy": 0.7471992895007133,
+      "num_tokens": 22642646.0,
+      "step": 610
+    },
+    {
+      "entropy": 0.6149823144078255,
+      "epoch": 3.9472830494728304,
+      "grad_norm": 0.012957974337041378,
+      "learning_rate": 0.0002,
+      "loss": 0.6171556711196899,
+      "mean_token_accuracy": 0.7489227131009102,
+      "num_tokens": 22679842.0,
+      "step": 611
+    },
+    {
+      "entropy": 0.6145467907190323,
+      "epoch": 3.9537712895377126,
+      "grad_norm": 0.010933002457022667,
+      "learning_rate": 0.0002,
+      "loss": 0.6167636513710022,
+      "mean_token_accuracy": 0.7463917657732964,
+      "num_tokens": 22717084.0,
+      "step": 612
+    },
+    {
+      "entropy": 0.6232969015836716,
+      "epoch": 3.9602595296025953,
+      "grad_norm": 0.009922384284436703,
+      "learning_rate": 0.0002,
+      "loss": 0.6252729892730713,
+      "mean_token_accuracy": 0.743752971291542,
+      "num_tokens": 22754399.0,
+      "step": 613
+    },
+    {
+      "entropy": 0.6329927816987038,
+      "epoch": 3.9667477696674776,
+      "grad_norm": 0.012450586073100567,
+      "learning_rate": 0.0002,
+      "loss": 0.6310559511184692,
+      "mean_token_accuracy": 0.7443704977631569,
+      "num_tokens": 22791668.0,
+      "step": 614
+    },
+    {
+      "entropy": 0.6241239309310913,
+      "epoch": 3.9732360097323602,
+      "grad_norm": 0.010473434813320637,
+      "learning_rate": 0.0002,
+      "loss": 0.6205140352249146,
+      "mean_token_accuracy": 0.7461888715624809,
+      "num_tokens": 22829072.0,
+      "step": 615
+    },
+    {
+      "entropy": 0.6387766450643539,
+      "epoch": 3.9797242497972425,
+      "grad_norm": 0.011033239774405956,
+      "learning_rate": 0.0002,
+      "loss": 0.638136088848114,
+      "mean_token_accuracy": 0.7390930280089378,
+      "num_tokens": 22866252.0,
+      "step": 616
+    },
+    {
+      "entropy": 0.6195501312613487,
+      "epoch": 3.986212489862125,
+      "grad_norm": 0.011396627873182297,
+      "learning_rate": 0.0002,
+      "loss": 0.6249865293502808,
+      "mean_token_accuracy": 0.7435013577342033,
+      "num_tokens": 22903910.0,
+      "step": 617
+    },
+    {
+      "entropy": 0.6073991656303406,
+      "epoch": 3.9927007299270074,
+      "grad_norm": 0.011404155753552914,
+      "learning_rate": 0.0002,
+      "loss": 0.6144699454307556,
+      "mean_token_accuracy": 0.7486153542995453,
+      "num_tokens": 22941419.0,
+      "step": 618
+    },
+    {
+      "entropy": 0.6133335530757904,
+      "epoch": 3.9991889699918897,
+      "grad_norm": 0.012148975394666195,
+      "learning_rate": 0.0002,
+      "loss": 0.6239128112792969,
+      "mean_token_accuracy": 0.744051106274128,
+      "num_tokens": 22978672.0,
+      "step": 619
+    },
+    {
+      "entropy": 0.6109002828598022,
+      "epoch": 4.0,
+      "grad_norm": 0.02331351302564144,
+      "learning_rate": 0.0002,
+      "loss": 0.6144630312919617,
+      "mean_token_accuracy": 0.7544165849685669,
+      "num_tokens": 22983289.0,
+      "step": 620
+    },
+    {
+      "entropy": 0.6217539757490158,
+      "epoch": 4.006488240064883,
+      "grad_norm": 0.013456643559038639,
+      "learning_rate": 0.0002,
+      "loss": 0.613330602645874,
+      "mean_token_accuracy": 0.7499688565731049,
+      "num_tokens": 23020412.0,
+      "step": 621
+    },
+    {
+      "entropy": 0.6039147675037384,
+      "epoch": 4.0129764801297645,
+      "grad_norm": 0.013175777159631252,
+      "learning_rate": 0.0002,
+      "loss": 0.597152829170227,
+      "mean_token_accuracy": 0.7552643939852715,
+      "num_tokens": 23057448.0,
+      "step": 622
+    },
+    {
+      "entropy": 0.6142989620566368,
+      "epoch": 4.019464720194647,
+      "grad_norm": 0.014114204794168472,
+      "learning_rate": 0.0002,
+      "loss": 0.6228339672088623,
+      "mean_token_accuracy": 0.744938038289547,
+      "num_tokens": 23094545.0,
+      "step": 623
+    },
+    {
+      "entropy": 0.6015428304672241,
+      "epoch": 4.02595296025953,
+      "grad_norm": 0.015099803917109966,
+      "learning_rate": 0.0002,
+      "loss": 0.6131606698036194,
+      "mean_token_accuracy": 0.7470837756991386,
+      "num_tokens": 23131965.0,
+      "step": 624
+    },
+    {
+      "entropy": 0.5885332375764847,
+      "epoch": 4.032441200324412,
+      "grad_norm": 0.014588817954063416,
+      "learning_rate": 0.0002,
+      "loss": 0.593397855758667,
+      "mean_token_accuracy": 0.7566570267081261,
+      "num_tokens": 23168813.0,
+      "step": 625
+    },
+    {
+      "entropy": 0.6016655787825584,
+      "epoch": 4.038929440389294,
+      "grad_norm": 0.012094132602214813,
+      "learning_rate": 0.0002,
+      "loss": 0.6046090126037598,
+      "mean_token_accuracy": 0.7522159889340401,
+      "num_tokens": 23205779.0,
+      "step": 626
+    },
+    {
+      "entropy": 0.6151943281292915,
+      "epoch": 4.045417680454177,
+      "grad_norm": 0.012667203322052956,
+      "learning_rate": 0.0002,
+      "loss": 0.6111838221549988,
+      "mean_token_accuracy": 0.7486316040158272,
+      "num_tokens": 23242684.0,
+      "step": 627
+    },
+    {
+      "entropy": 0.6105204373598099,
+      "epoch": 4.051905920519059,
+      "grad_norm": 0.011044032871723175,
+      "learning_rate": 0.0002,
+      "loss": 0.6119928956031799,
+      "mean_token_accuracy": 0.7496168315410614,
+      "num_tokens": 23279684.0,
+      "step": 628
+    },
+    {
+      "entropy": 0.6080695539712906,
+      "epoch": 4.0583941605839415,
+      "grad_norm": 0.012295052409172058,
+      "learning_rate": 0.0002,
+      "loss": 0.6104212999343872,
+      "mean_token_accuracy": 0.7498067244887352,
+      "num_tokens": 23316599.0,
+      "step": 629
+    },
+    {
+      "entropy": 0.6009281948208809,
+      "epoch": 4.064882400648824,
+      "grad_norm": 0.012583774514496326,
+      "learning_rate": 0.0002,
+      "loss": 0.6041219234466553,
+      "mean_token_accuracy": 0.7518585100769997,
+      "num_tokens": 23353800.0,
+      "step": 630
+    },
+    {
+      "entropy": 0.6015147417783737,
+      "epoch": 4.071370640713706,
+      "grad_norm": 0.011909936554729939,
+      "learning_rate": 0.0002,
+      "loss": 0.6030286550521851,
+      "mean_token_accuracy": 0.7531652525067329,
+      "num_tokens": 23390939.0,
+      "step": 631
+    },
+    {
+      "entropy": 0.6044563055038452,
+      "epoch": 4.077858880778589,
+      "grad_norm": 0.011805822141468525,
+      "learning_rate": 0.0002,
+      "loss": 0.6060421466827393,
+      "mean_token_accuracy": 0.7509199753403664,
+      "num_tokens": 23427956.0,
+      "step": 632
+    },
+    {
+      "entropy": 0.6049995571374893,
+      "epoch": 4.084347120843471,
+      "grad_norm": 0.012876358814537525,
+      "learning_rate": 0.0002,
+      "loss": 0.6117682456970215,
+      "mean_token_accuracy": 0.7485300377011299,
+      "num_tokens": 23465445.0,
+      "step": 633
+    },
+    {
+      "entropy": 0.6048118397593498,
+      "epoch": 4.090835360908353,
+      "grad_norm": 0.012466302141547203,
+      "learning_rate": 0.0002,
+      "loss": 0.6076602339744568,
+      "mean_token_accuracy": 0.7526114881038666,
+      "num_tokens": 23502361.0,
+      "step": 634
+    },
+    {
+      "entropy": 0.6199515163898468,
+      "epoch": 4.097323600973236,
+      "grad_norm": 0.013892917893826962,
+      "learning_rate": 0.0002,
+      "loss": 0.6193369626998901,
+      "mean_token_accuracy": 0.7488663196563721,
+      "num_tokens": 23540086.0,
+      "step": 635
+    },
+    {
+      "entropy": 0.6109070330858231,
+      "epoch": 4.103811841038119,
+      "grad_norm": 0.012778576463460922,
+      "learning_rate": 0.0002,
+      "loss": 0.6099793910980225,
+      "mean_token_accuracy": 0.7521104216575623,
+      "num_tokens": 23577435.0,
+      "step": 636
+    },
+    {
+      "entropy": 0.6217869147658348,
+      "epoch": 4.110300081103,
+      "grad_norm": 0.011083797551691532,
+      "learning_rate": 0.0002,
+      "loss": 0.6223993301391602,
+      "mean_token_accuracy": 0.7441250160336494,
+      "num_tokens": 23614892.0,
+      "step": 637
+    },
+    {
+      "entropy": 0.6092150881886482,
+      "epoch": 4.116788321167883,
+      "grad_norm": 0.011457768268883228,
+      "learning_rate": 0.0002,
+      "loss": 0.6109668016433716,
+      "mean_token_accuracy": 0.7505389377474785,
+      "num_tokens": 23652404.0,
+      "step": 638
+    },
+    {
+      "entropy": 0.5892162844538689,
+      "epoch": 4.123276561232766,
+      "grad_norm": 0.011224270798265934,
+      "learning_rate": 0.0002,
+      "loss": 0.5923285484313965,
+      "mean_token_accuracy": 0.7561718374490738,
+      "num_tokens": 23689182.0,
+      "step": 639
+    },
+    {
+      "entropy": 0.6155631691217422,
+      "epoch": 4.1297648012976484,
+      "grad_norm": 0.011392487213015556,
+      "learning_rate": 0.0002,
+      "loss": 0.6169704794883728,
+      "mean_token_accuracy": 0.7484367191791534,
+      "num_tokens": 23726827.0,
+      "step": 640
+    },
+    {
+      "entropy": 0.6134647503495216,
+      "epoch": 4.13625304136253,
+      "grad_norm": 0.01305895671248436,
+      "learning_rate": 0.0002,
+      "loss": 0.6227377653121948,
+      "mean_token_accuracy": 0.7455505281686783,
+      "num_tokens": 23764215.0,
+      "step": 641
+    },
+    {
+      "entropy": 0.6127767860889435,
+      "epoch": 4.142741281427413,
+      "grad_norm": 0.011744504794478416,
+      "learning_rate": 0.0002,
+      "loss": 0.6154167056083679,
+      "mean_token_accuracy": 0.7480334118008614,
+      "num_tokens": 23801609.0,
+      "step": 642
+    },
+    {
+      "entropy": 0.6113294437527657,
+      "epoch": 4.149229521492296,
+      "grad_norm": 0.013036824762821198,
+      "learning_rate": 0.0002,
+      "loss": 0.6141045093536377,
+      "mean_token_accuracy": 0.7493610754609108,
+      "num_tokens": 23839155.0,
+      "step": 643
+    },
+    {
+      "entropy": 0.6171691864728928,
+      "epoch": 4.155717761557177,
+      "grad_norm": 0.012765160761773586,
+      "learning_rate": 0.0002,
+      "loss": 0.6207258701324463,
+      "mean_token_accuracy": 0.7462699934840202,
+      "num_tokens": 23876532.0,
+      "step": 644
+    },
+    {
+      "entropy": 0.6020215824246407,
+      "epoch": 4.16220600162206,
+      "grad_norm": 0.011224629357457161,
+      "learning_rate": 0.0002,
+      "loss": 0.6040517091751099,
+      "mean_token_accuracy": 0.7520366311073303,
+      "num_tokens": 23913624.0,
+      "step": 645
+    },
+    {
+      "entropy": 0.6073898002505302,
+      "epoch": 4.168694241686943,
+      "grad_norm": 0.012669643387198448,
+      "learning_rate": 0.0002,
+      "loss": 0.6089828610420227,
+      "mean_token_accuracy": 0.751623198390007,
+      "num_tokens": 23951134.0,
+      "step": 646
+    },
+    {
+      "entropy": 0.6056643575429916,
+      "epoch": 4.175182481751825,
+      "grad_norm": 0.013099576346576214,
+      "learning_rate": 0.0002,
+      "loss": 0.6066282391548157,
+      "mean_token_accuracy": 0.7509062737226486,
+      "num_tokens": 23988712.0,
+      "step": 647
+    },
+    {
+      "entropy": 0.6190872266888618,
+      "epoch": 4.181670721816707,
+      "grad_norm": 0.013141883537173271,
+      "learning_rate": 0.0002,
+      "loss": 0.6172468662261963,
+      "mean_token_accuracy": 0.7470818608999252,
+      "num_tokens": 24025925.0,
+      "step": 648
+    },
+    {
+      "entropy": 0.6018052324652672,
+      "epoch": 4.18815896188159,
+      "grad_norm": 0.01188298873603344,
+      "learning_rate": 0.0002,
+      "loss": 0.6035501956939697,
+      "mean_token_accuracy": 0.7526876479387283,
+      "num_tokens": 24063159.0,
+      "step": 649
+    },
+    {
+      "entropy": 0.6047950685024261,
+      "epoch": 4.194647201946472,
+      "grad_norm": 0.014647013507783413,
+      "learning_rate": 0.0002,
+      "loss": 0.6088840961456299,
+      "mean_token_accuracy": 0.7501959800720215,
+      "num_tokens": 24100259.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.6092503741383553,
+      "epoch": 4.2011354420113545,
+      "grad_norm": 0.012967122718691826,
+      "learning_rate": 0.0002,
+      "loss": 0.6154978275299072,
+      "mean_token_accuracy": 0.7469052001833916,
+      "num_tokens": 24137670.0,
+      "step": 651
+    },
+    {
+      "entropy": 0.6076086908578873,
+      "epoch": 4.207623682076237,
+      "grad_norm": 0.015226679854094982,
+      "learning_rate": 0.0002,
+      "loss": 0.6063575744628906,
+      "mean_token_accuracy": 0.7520931884646416,
+      "num_tokens": 24174990.0,
+      "step": 652
+    },
+    {
+      "entropy": 0.5980376899242401,
+      "epoch": 4.214111922141119,
+      "grad_norm": 0.01129455491900444,
+      "learning_rate": 0.0002,
+      "loss": 0.5961977243423462,
+      "mean_token_accuracy": 0.7564675956964493,
+      "num_tokens": 24212331.0,
+      "step": 653
+    },
+    {
+      "entropy": 0.6016443520784378,
+      "epoch": 4.220600162206002,
+      "grad_norm": 0.014820009469985962,
+      "learning_rate": 0.0002,
+      "loss": 0.6029236912727356,
+      "mean_token_accuracy": 0.7525514140725136,
+      "num_tokens": 24249375.0,
+      "step": 654
+    },
+    {
+      "entropy": 0.6087661758065224,
+      "epoch": 4.227088402270884,
+      "grad_norm": 0.0128677599132061,
+      "learning_rate": 0.0002,
+      "loss": 0.6112372279167175,
+      "mean_token_accuracy": 0.748601570725441,
+      "num_tokens": 24286871.0,
+      "step": 655
+    },
+    {
+      "entropy": 0.615173451602459,
+      "epoch": 4.233576642335766,
+      "grad_norm": 0.013609779067337513,
+      "learning_rate": 0.0002,
+      "loss": 0.6189801096916199,
+      "mean_token_accuracy": 0.7472759932279587,
+      "num_tokens": 24323996.0,
+      "step": 656
+    },
+    {
+      "entropy": 0.6089212968945503,
+      "epoch": 4.240064882400649,
+      "grad_norm": 0.013383638113737106,
+      "learning_rate": 0.0002,
+      "loss": 0.6078221797943115,
+      "mean_token_accuracy": 0.7503775134682655,
+      "num_tokens": 24361162.0,
+      "step": 657
+    },
+    {
+      "entropy": 0.6125769466161728,
+      "epoch": 4.2465531224655315,
+      "grad_norm": 0.01211923360824585,
+      "learning_rate": 0.0002,
+      "loss": 0.6157176494598389,
+      "mean_token_accuracy": 0.7472077012062073,
+      "num_tokens": 24398308.0,
+      "step": 658
+    },
+    {
+      "entropy": 0.6129145845770836,
+      "epoch": 4.253041362530413,
+      "grad_norm": 0.013655205257236958,
+      "learning_rate": 0.0002,
+      "loss": 0.6173056364059448,
+      "mean_token_accuracy": 0.7478220462799072,
+      "num_tokens": 24435361.0,
+      "step": 659
+    },
+    {
+      "entropy": 0.6286280378699303,
+      "epoch": 4.259529602595296,
+      "grad_norm": 0.012913207523524761,
+      "learning_rate": 0.0002,
+      "loss": 0.6232912540435791,
+      "mean_token_accuracy": 0.7459053471684456,
+      "num_tokens": 24472862.0,
+      "step": 660
+    },
+    {
+      "entropy": 0.6073895320296288,
+      "epoch": 4.266017842660179,
+      "grad_norm": 0.012715890072286129,
+      "learning_rate": 0.0002,
+      "loss": 0.6063985824584961,
+      "mean_token_accuracy": 0.7502154931426048,
+      "num_tokens": 24510378.0,
+      "step": 661
+    },
+    {
+      "entropy": 0.5987715944647789,
+      "epoch": 4.2725060827250605,
+      "grad_norm": 0.01144983246922493,
+      "learning_rate": 0.0002,
+      "loss": 0.603581964969635,
+      "mean_token_accuracy": 0.7529052719473839,
+      "num_tokens": 24547450.0,
+      "step": 662
+    },
+    {
+      "entropy": 0.6105594113469124,
+      "epoch": 4.278994322789943,
+      "grad_norm": 0.01477118767797947,
+      "learning_rate": 0.0002,
+      "loss": 0.6208908557891846,
+      "mean_token_accuracy": 0.7453452572226524,
+      "num_tokens": 24584414.0,
+      "step": 663
+    },
+    {
+      "entropy": 0.6013319417834282,
+      "epoch": 4.285482562854826,
+      "grad_norm": 0.011751738376915455,
+      "learning_rate": 0.0002,
+      "loss": 0.6031821966171265,
+      "mean_token_accuracy": 0.7529275268316269,
+      "num_tokens": 24621522.0,
+      "step": 664
+    },
+    {
+      "entropy": 0.6105660572648048,
+      "epoch": 4.291970802919708,
+      "grad_norm": 0.011309951543807983,
+      "learning_rate": 0.0002,
+      "loss": 0.6077959537506104,
+      "mean_token_accuracy": 0.7497444152832031,
+      "num_tokens": 24658985.0,
+      "step": 665
+    },
+    {
+      "entropy": 0.6143631562590599,
+      "epoch": 4.29845904298459,
+      "grad_norm": 0.01235566008836031,
+      "learning_rate": 0.0002,
+      "loss": 0.6165258884429932,
+      "mean_token_accuracy": 0.7466364279389381,
+      "num_tokens": 24696556.0,
+      "step": 666
+    },
+    {
+      "entropy": 0.6108560711145401,
+      "epoch": 4.304947283049473,
+      "grad_norm": 0.012649192474782467,
+      "learning_rate": 0.0002,
+      "loss": 0.6156973838806152,
+      "mean_token_accuracy": 0.7493721544742584,
+      "num_tokens": 24733563.0,
+      "step": 667
+    },
+    {
+      "entropy": 0.5972470939159393,
+      "epoch": 4.311435523114355,
+      "grad_norm": 0.012324381619691849,
+      "learning_rate": 0.0002,
+      "loss": 0.593143880367279,
+      "mean_token_accuracy": 0.7569256648421288,
+      "num_tokens": 24770521.0,
+      "step": 668
+    },
+    {
+      "entropy": 0.6238643601536751,
+      "epoch": 4.3179237631792375,
+      "grad_norm": 0.013767421245574951,
+      "learning_rate": 0.0002,
+      "loss": 0.6210561990737915,
+      "mean_token_accuracy": 0.7456039562821388,
+      "num_tokens": 24807746.0,
+      "step": 669
+    },
+    {
+      "entropy": 0.6074958592653275,
+      "epoch": 4.32441200324412,
+      "grad_norm": 0.012642712332308292,
+      "learning_rate": 0.0002,
+      "loss": 0.6104567646980286,
+      "mean_token_accuracy": 0.7498242110013962,
+      "num_tokens": 24844948.0,
+      "step": 670
+    },
+    {
+      "entropy": 0.6030944362282753,
+      "epoch": 4.330900243309002,
+      "grad_norm": 0.014455183409154415,
+      "learning_rate": 0.0002,
+      "loss": 0.6104896068572998,
+      "mean_token_accuracy": 0.751430056989193,
+      "num_tokens": 24882330.0,
+      "step": 671
+    },
+    {
+      "entropy": 0.6085335910320282,
+      "epoch": 4.337388483373885,
+      "grad_norm": 0.012203236110508442,
+      "learning_rate": 0.0002,
+      "loss": 0.6149730682373047,
+      "mean_token_accuracy": 0.748404935002327,
+      "num_tokens": 24919894.0,
+      "step": 672
+    },
+    {
+      "entropy": 0.6215960904955864,
+      "epoch": 4.343876723438767,
+      "grad_norm": 0.01517926249653101,
+      "learning_rate": 0.0002,
+      "loss": 0.6182515621185303,
+      "mean_token_accuracy": 0.7460620552301407,
+      "num_tokens": 24957338.0,
+      "step": 673
+    },
+    {
+      "entropy": 0.6206786409020424,
+      "epoch": 4.350364963503649,
+      "grad_norm": 0.012405015528202057,
+      "learning_rate": 0.0002,
+      "loss": 0.6114768385887146,
+      "mean_token_accuracy": 0.749580554664135,
+      "num_tokens": 24994430.0,
+      "step": 674
+    },
+    {
+      "entropy": 0.6176847890019417,
+      "epoch": 4.356853203568532,
+      "grad_norm": 0.012959569692611694,
+      "learning_rate": 0.0002,
+      "loss": 0.6179355382919312,
+      "mean_token_accuracy": 0.7443158105015755,
+      "num_tokens": 25031819.0,
+      "step": 675
+    },
+    {
+      "entropy": 0.6137174069881439,
+      "epoch": 4.363341443633415,
+      "grad_norm": 0.01568513549864292,
+      "learning_rate": 0.0002,
+      "loss": 0.6264492273330688,
+      "mean_token_accuracy": 0.7434774041175842,
+      "num_tokens": 25069321.0,
+      "step": 676
+    },
+    {
+      "entropy": 0.6073234900832176,
+      "epoch": 4.369829683698297,
+      "grad_norm": 0.01254733745008707,
+      "learning_rate": 0.0002,
+      "loss": 0.613095760345459,
+      "mean_token_accuracy": 0.7486145123839378,
+      "num_tokens": 25106724.0,
+      "step": 677
+    },
+    {
+      "entropy": 0.6180970370769501,
+      "epoch": 4.376317923763179,
+      "grad_norm": 0.01219653245061636,
+      "learning_rate": 0.0002,
+      "loss": 0.6162555813789368,
+      "mean_token_accuracy": 0.7453421652317047,
+      "num_tokens": 25144301.0,
+      "step": 678
+    },
+    {
+      "entropy": 0.6155680492520332,
+      "epoch": 4.382806163828062,
+      "grad_norm": 0.013878269121050835,
+      "learning_rate": 0.0002,
+      "loss": 0.6091170907020569,
+      "mean_token_accuracy": 0.750255212187767,
+      "num_tokens": 25181210.0,
+      "step": 679
+    },
+    {
+      "entropy": 0.6227680519223213,
+      "epoch": 4.3892944038929445,
+      "grad_norm": 0.01349555142223835,
+      "learning_rate": 0.0002,
+      "loss": 0.6250801086425781,
+      "mean_token_accuracy": 0.743384949862957,
+      "num_tokens": 25218262.0,
+      "step": 680
+    },
+    {
+      "entropy": 0.5870469063520432,
+      "epoch": 4.395782643957826,
+      "grad_norm": 0.012459780089557171,
+      "learning_rate": 0.0002,
+      "loss": 0.5922443270683289,
+      "mean_token_accuracy": 0.75776407122612,
+      "num_tokens": 25255586.0,
+      "step": 681
+    },
+    {
+      "entropy": 0.6126704216003418,
+      "epoch": 4.402270884022709,
+      "grad_norm": 0.012495280243456364,
+      "learning_rate": 0.0002,
+      "loss": 0.6199891567230225,
+      "mean_token_accuracy": 0.7477240189909935,
+      "num_tokens": 25293110.0,
+      "step": 682
+    },
+    {
+      "entropy": 0.6052056401968002,
+      "epoch": 4.408759124087592,
+      "grad_norm": 0.011480255052447319,
+      "learning_rate": 0.0002,
+      "loss": 0.6054994463920593,
+      "mean_token_accuracy": 0.7517327517271042,
+      "num_tokens": 25330240.0,
+      "step": 683
+    },
+    {
+      "entropy": 0.6134005859494209,
+      "epoch": 4.415247364152473,
+      "grad_norm": 0.013378560543060303,
+      "learning_rate": 0.0002,
+      "loss": 0.6032570600509644,
+      "mean_token_accuracy": 0.7544484660029411,
+      "num_tokens": 25367930.0,
+      "step": 684
+    },
+    {
+      "entropy": 0.6132281497120857,
+      "epoch": 4.421735604217356,
+      "grad_norm": 0.011817892082035542,
+      "learning_rate": 0.0002,
+      "loss": 0.6155675649642944,
+      "mean_token_accuracy": 0.7474397569894791,
+      "num_tokens": 25404927.0,
+      "step": 685
+    },
+    {
+      "entropy": 0.6157776713371277,
+      "epoch": 4.428223844282239,
+      "grad_norm": 0.013845066539943218,
+      "learning_rate": 0.0002,
+      "loss": 0.6208671927452087,
+      "mean_token_accuracy": 0.7460346668958664,
+      "num_tokens": 25441808.0,
+      "step": 686
+    },
+    {
+      "entropy": 0.6034616231918335,
+      "epoch": 4.434712084347121,
+      "grad_norm": 0.011209644377231598,
+      "learning_rate": 0.0002,
+      "loss": 0.6051755547523499,
+      "mean_token_accuracy": 0.751818560063839,
+      "num_tokens": 25478982.0,
+      "step": 687
+    },
+    {
+      "entropy": 0.6197097897529602,
+      "epoch": 4.441200324412003,
+      "grad_norm": 0.01221329253166914,
+      "learning_rate": 0.0002,
+      "loss": 0.624691367149353,
+      "mean_token_accuracy": 0.743960902094841,
+      "num_tokens": 25516230.0,
+      "step": 688
+    },
+    {
+      "entropy": 0.6036801412701607,
+      "epoch": 4.447688564476886,
+      "grad_norm": 0.011895528994500637,
+      "learning_rate": 0.0002,
+      "loss": 0.604611873626709,
+      "mean_token_accuracy": 0.7565013989806175,
+      "num_tokens": 25553987.0,
+      "step": 689
+    },
+    {
+      "entropy": 0.6126270368695259,
+      "epoch": 4.454176804541768,
+      "grad_norm": 0.014394382014870644,
+      "learning_rate": 0.0002,
+      "loss": 0.6157338619232178,
+      "mean_token_accuracy": 0.7499115914106369,
+      "num_tokens": 25591126.0,
+      "step": 690
+    },
+    {
+      "entropy": 0.6108303889632225,
+      "epoch": 4.4606650446066505,
+      "grad_norm": 0.014680239371955395,
+      "learning_rate": 0.0002,
+      "loss": 0.6087052822113037,
+      "mean_token_accuracy": 0.7516598179936409,
+      "num_tokens": 25628123.0,
+      "step": 691
+    },
+    {
+      "entropy": 0.6043496802449226,
+      "epoch": 4.467153284671533,
+      "grad_norm": 0.013057287782430649,
+      "learning_rate": 0.0002,
+      "loss": 0.6059702038764954,
+      "mean_token_accuracy": 0.7521848082542419,
+      "num_tokens": 25665273.0,
+      "step": 692
+    },
+    {
+      "entropy": 0.6091088205575943,
+      "epoch": 4.473641524736415,
+      "grad_norm": 0.012656234204769135,
+      "learning_rate": 0.0002,
+      "loss": 0.6140547394752502,
+      "mean_token_accuracy": 0.7490741834044456,
+      "num_tokens": 25702835.0,
+      "step": 693
+    },
+    {
+      "entropy": 0.6054038554430008,
+      "epoch": 4.480129764801298,
+      "grad_norm": 0.01144889835268259,
+      "learning_rate": 0.0002,
+      "loss": 0.6068828105926514,
+      "mean_token_accuracy": 0.7513371407985687,
+      "num_tokens": 25740321.0,
+      "step": 694
+    },
+    {
+      "entropy": 0.6075622960925102,
+      "epoch": 4.48661800486618,
+      "grad_norm": 0.014180710539221764,
+      "learning_rate": 0.0002,
+      "loss": 0.6113091707229614,
+      "mean_token_accuracy": 0.7500150203704834,
+      "num_tokens": 25777354.0,
+      "step": 695
+    },
+    {
+      "entropy": 0.6083768233656883,
+      "epoch": 4.493106244931062,
+      "grad_norm": 0.01257058884948492,
+      "learning_rate": 0.0002,
+      "loss": 0.613832950592041,
+      "mean_token_accuracy": 0.7493138536810875,
+      "num_tokens": 25814766.0,
+      "step": 696
+    },
+    {
+      "entropy": 0.6096158251166344,
+      "epoch": 4.499594484995945,
+      "grad_norm": 0.01403698232024908,
+      "learning_rate": 0.0002,
+      "loss": 0.606455385684967,
+      "mean_token_accuracy": 0.751399964094162,
+      "num_tokens": 25852032.0,
+      "step": 697
+    },
+    {
+      "entropy": 0.61270372569561,
+      "epoch": 4.5060827250608275,
+      "grad_norm": 0.013846375048160553,
+      "learning_rate": 0.0002,
+      "loss": 0.6081539988517761,
+      "mean_token_accuracy": 0.7516685873270035,
+      "num_tokens": 25889272.0,
+      "step": 698
+    },
+    {
+      "entropy": 0.6115692555904388,
+      "epoch": 4.512570965125709,
+      "grad_norm": 0.011561271734535694,
+      "learning_rate": 0.0002,
+      "loss": 0.6132485866546631,
+      "mean_token_accuracy": 0.7496950924396515,
+      "num_tokens": 25926376.0,
+      "step": 699
+    },
+    {
+      "entropy": 0.6165875196456909,
+      "epoch": 4.519059205190592,
+      "grad_norm": 0.014013632200658321,
+      "learning_rate": 0.0002,
+      "loss": 0.6210501194000244,
+      "mean_token_accuracy": 0.7466816902160645,
+      "num_tokens": 25964071.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.6069151386618614,
+      "epoch": 4.525547445255475,
+      "grad_norm": 0.01162275206297636,
+      "learning_rate": 0.0002,
+      "loss": 0.6112617254257202,
+      "mean_token_accuracy": 0.7507427483797073,
+      "num_tokens": 26001770.0,
+      "step": 701
+    },
+    {
+      "entropy": 0.5982604473829269,
+      "epoch": 4.5320356853203565,
+      "grad_norm": 0.010527896694839,
+      "learning_rate": 0.0002,
+      "loss": 0.5975767374038696,
+      "mean_token_accuracy": 0.7562049925327301,
+      "num_tokens": 26038790.0,
+      "step": 702
+    },
+    {
+      "entropy": 0.6178539171814919,
+      "epoch": 4.538523925385239,
+      "grad_norm": 0.011301129125058651,
+      "learning_rate": 0.0002,
+      "loss": 0.6221362352371216,
+      "mean_token_accuracy": 0.743792898952961,
+      "num_tokens": 26076263.0,
+      "step": 703
+    },
+    {
+      "entropy": 0.6195997893810272,
+      "epoch": 4.545012165450122,
+      "grad_norm": 0.01180302631109953,
+      "learning_rate": 0.0002,
+      "loss": 0.618148922920227,
+      "mean_token_accuracy": 0.7460778802633286,
+      "num_tokens": 26113372.0,
+      "step": 704
+    },
+    {
+      "entropy": 0.6094764620065689,
+      "epoch": 4.551500405515004,
+      "grad_norm": 0.011488985270261765,
+      "learning_rate": 0.0002,
+      "loss": 0.6123408675193787,
+      "mean_token_accuracy": 0.7495540827512741,
+      "num_tokens": 26150197.0,
+      "step": 705
+    },
+    {
+      "entropy": 0.6152739748358727,
+      "epoch": 4.557988645579886,
+      "grad_norm": 0.011894487775862217,
+      "learning_rate": 0.0002,
+      "loss": 0.6184768676757812,
+      "mean_token_accuracy": 0.7447682395577431,
+      "num_tokens": 26187580.0,
+      "step": 706
+    },
+    {
+      "entropy": 0.6004127934575081,
+      "epoch": 4.564476885644769,
+      "grad_norm": 0.01042366586625576,
+      "learning_rate": 0.0002,
+      "loss": 0.6032600402832031,
+      "mean_token_accuracy": 0.7507253810763359,
+      "num_tokens": 26224583.0,
+      "step": 707
+    },
+    {
+      "entropy": 0.6170988976955414,
+      "epoch": 4.570965125709652,
+      "grad_norm": 0.012767941690981388,
+      "learning_rate": 0.0002,
+      "loss": 0.6203943490982056,
+      "mean_token_accuracy": 0.7446292862296104,
+      "num_tokens": 26261715.0,
+      "step": 708
+    },
+    {
+      "entropy": 0.6010770201683044,
+      "epoch": 4.5774533657745335,
+      "grad_norm": 0.010196649469435215,
+      "learning_rate": 0.0002,
+      "loss": 0.6022849082946777,
+      "mean_token_accuracy": 0.7530830800533295,
+      "num_tokens": 26298743.0,
+      "step": 709
+    },
+    {
+      "entropy": 0.6049724221229553,
+      "epoch": 4.583941605839416,
+      "grad_norm": 0.012646735645830631,
+      "learning_rate": 0.0002,
+      "loss": 0.6061745882034302,
+      "mean_token_accuracy": 0.7506730481982231,
+      "num_tokens": 26335489.0,
+      "step": 710
+    },
+    {
+      "entropy": 0.6168033257126808,
+      "epoch": 4.590429845904298,
+      "grad_norm": 0.013492587953805923,
+      "learning_rate": 0.0002,
+      "loss": 0.616999626159668,
+      "mean_token_accuracy": 0.7479605078697205,
+      "num_tokens": 26372489.0,
+      "step": 711
+    },
+    {
+      "entropy": 0.6182409450411797,
+      "epoch": 4.596918085969181,
+      "grad_norm": 0.010737297125160694,
+      "learning_rate": 0.0002,
+      "loss": 0.6131489276885986,
+      "mean_token_accuracy": 0.7496219426393509,
+      "num_tokens": 26409535.0,
+      "step": 712
+    },
+    {
+      "entropy": 0.6057656034827232,
+      "epoch": 4.603406326034063,
+      "grad_norm": 0.014007823541760445,
+      "learning_rate": 0.0002,
+      "loss": 0.6030014753341675,
+      "mean_token_accuracy": 0.7548240646719933,
+      "num_tokens": 26446279.0,
+      "step": 713
+    },
+    {
+      "entropy": 0.6160694062709808,
+      "epoch": 4.609894566098946,
+      "grad_norm": 0.012945987284183502,
+      "learning_rate": 0.0002,
+      "loss": 0.620917558670044,
+      "mean_token_accuracy": 0.7472490146756172,
+      "num_tokens": 26483525.0,
+      "step": 714
+    },
+    {
+      "entropy": 0.608958512544632,
+      "epoch": 4.616382806163828,
+      "grad_norm": 0.014964601024985313,
+      "learning_rate": 0.0002,
+      "loss": 0.6204575300216675,
+      "mean_token_accuracy": 0.7449859008193016,
+      "num_tokens": 26520720.0,
+      "step": 715
+    },
+    {
+      "entropy": 0.6118902638554573,
+      "epoch": 4.622871046228711,
+      "grad_norm": 0.011436744593083858,
+      "learning_rate": 0.0002,
+      "loss": 0.6165215969085693,
+      "mean_token_accuracy": 0.7477826401591301,
+      "num_tokens": 26557982.0,
+      "step": 716
+    },
+    {
+      "entropy": 0.6222316101193428,
+      "epoch": 4.629359286293592,
+      "grad_norm": 0.011866914108395576,
+      "learning_rate": 0.0002,
+      "loss": 0.6159400939941406,
+      "mean_token_accuracy": 0.7480059117078781,
+      "num_tokens": 26595389.0,
+      "step": 717
+    },
+    {
+      "entropy": 0.6267769560217857,
+      "epoch": 4.635847526358475,
+      "grad_norm": 0.012834588065743446,
+      "learning_rate": 0.0002,
+      "loss": 0.6215125322341919,
+      "mean_token_accuracy": 0.7467430606484413,
+      "num_tokens": 26632558.0,
+      "step": 718
+    },
+    {
+      "entropy": 0.6167991682887077,
+      "epoch": 4.642335766423358,
+      "grad_norm": 0.010781707242131233,
+      "learning_rate": 0.0002,
+      "loss": 0.6117191314697266,
+      "mean_token_accuracy": 0.7489187940955162,
+      "num_tokens": 26669534.0,
+      "step": 719
+    },
+    {
+      "entropy": 0.608165793120861,
+      "epoch": 4.6488240064882405,
+      "grad_norm": 0.013234053738415241,
+      "learning_rate": 0.0002,
+      "loss": 0.6105374097824097,
+      "mean_token_accuracy": 0.7508261948823929,
+      "num_tokens": 26707068.0,
+      "step": 720
+    },
+    {
+      "entropy": 0.6050869300961494,
+      "epoch": 4.655312246553122,
+      "grad_norm": 0.015265488997101784,
+      "learning_rate": 0.0002,
+      "loss": 0.6178677082061768,
+      "mean_token_accuracy": 0.7447132244706154,
+      "num_tokens": 26744214.0,
+      "step": 721
+    },
+    {
+      "entropy": 0.606209434568882,
+      "epoch": 4.661800486618005,
+      "grad_norm": 0.012334835715591908,
+      "learning_rate": 0.0002,
+      "loss": 0.6131197214126587,
+      "mean_token_accuracy": 0.7471205219626427,
+      "num_tokens": 26781504.0,
+      "step": 722
+    },
+    {
+      "entropy": 0.6154639571905136,
+      "epoch": 4.668288726682888,
+      "grad_norm": 0.011331942863762379,
+      "learning_rate": 0.0002,
+      "loss": 0.6135961413383484,
+      "mean_token_accuracy": 0.750006228685379,
+      "num_tokens": 26818648.0,
+      "step": 723
+    },
+    {
+      "entropy": 0.6127299889922142,
+      "epoch": 4.674776966747769,
+      "grad_norm": 0.013016032055020332,
+      "learning_rate": 0.0002,
+      "loss": 0.6065088510513306,
+      "mean_token_accuracy": 0.7515178844332695,
+      "num_tokens": 26856269.0,
+      "step": 724
+    },
+    {
+      "entropy": 0.6198037564754486,
+      "epoch": 4.681265206812652,
+      "grad_norm": 0.010600600391626358,
+      "learning_rate": 0.0002,
+      "loss": 0.6175575256347656,
+      "mean_token_accuracy": 0.748474046587944,
+      "num_tokens": 26893819.0,
+      "step": 725
+    },
+    {
+      "entropy": 0.6088383719325066,
+      "epoch": 4.687753446877535,
+      "grad_norm": 0.012035705149173737,
+      "learning_rate": 0.0002,
+      "loss": 0.6124040484428406,
+      "mean_token_accuracy": 0.7508236169815063,
+      "num_tokens": 26931388.0,
+      "step": 726
+    },
+    {
+      "entropy": 0.6097267419099808,
+      "epoch": 4.694241686942417,
+      "grad_norm": 0.013257896527647972,
+      "learning_rate": 0.0002,
+      "loss": 0.6185274720191956,
+      "mean_token_accuracy": 0.7469557821750641,
+      "num_tokens": 26968862.0,
+      "step": 727
+    },
+    {
+      "entropy": 0.6031776890158653,
+      "epoch": 4.700729927007299,
+      "grad_norm": 0.012637577950954437,
+      "learning_rate": 0.0002,
+      "loss": 0.6094907522201538,
+      "mean_token_accuracy": 0.7504973784089088,
+      "num_tokens": 27006297.0,
+      "step": 728
+    },
+    {
+      "entropy": 0.6267594248056412,
+      "epoch": 4.707218167072182,
+      "grad_norm": 0.012524669989943504,
+      "learning_rate": 0.0002,
+      "loss": 0.6284726858139038,
+      "mean_token_accuracy": 0.74198117852211,
+      "num_tokens": 27043660.0,
+      "step": 729
+    },
+    {
+      "entropy": 0.6199127435684204,
+      "epoch": 4.713706407137064,
+      "grad_norm": 0.011993559077382088,
+      "learning_rate": 0.0002,
+      "loss": 0.6179155707359314,
+      "mean_token_accuracy": 0.7490969151258469,
+      "num_tokens": 27080997.0,
+      "step": 730
+    },
+    {
+      "entropy": 0.619316965341568,
+      "epoch": 4.7201946472019465,
+      "grad_norm": 0.011862427927553654,
+      "learning_rate": 0.0002,
+      "loss": 0.6162521839141846,
+      "mean_token_accuracy": 0.7466591447591782,
+      "num_tokens": 27118582.0,
+      "step": 731
+    },
+    {
+      "entropy": 0.617161475121975,
+      "epoch": 4.726682887266829,
+      "grad_norm": 0.010898209176957607,
+      "learning_rate": 0.0002,
+      "loss": 0.6218090057373047,
+      "mean_token_accuracy": 0.7456279322504997,
+      "num_tokens": 27155770.0,
+      "step": 732
+    },
+    {
+      "entropy": 0.6097361296415329,
+      "epoch": 4.733171127331711,
+      "grad_norm": 0.012140274979174137,
+      "learning_rate": 0.0002,
+      "loss": 0.6126585602760315,
+      "mean_token_accuracy": 0.7512409463524818,
+      "num_tokens": 27193021.0,
+      "step": 733
+    },
+    {
+      "entropy": 0.5991591587662697,
+      "epoch": 4.739659367396594,
+      "grad_norm": 0.011335095390677452,
+      "learning_rate": 0.0002,
+      "loss": 0.6005336046218872,
+      "mean_token_accuracy": 0.7536215856671333,
+      "num_tokens": 27230104.0,
+      "step": 734
+    },
+    {
+      "entropy": 0.6187688335776329,
+      "epoch": 4.746147607461476,
+      "grad_norm": 0.011418522335588932,
+      "learning_rate": 0.0002,
+      "loss": 0.619892418384552,
+      "mean_token_accuracy": 0.7460336685180664,
+      "num_tokens": 27267697.0,
+      "step": 735
+    },
+    {
+      "entropy": 0.6177256479859352,
+      "epoch": 4.752635847526358,
+      "grad_norm": 0.01126981433480978,
+      "learning_rate": 0.0002,
+      "loss": 0.6184297204017639,
+      "mean_token_accuracy": 0.7480708360671997,
+      "num_tokens": 27304935.0,
+      "step": 736
+    },
+    {
+      "entropy": 0.6365424171090126,
+      "epoch": 4.759124087591241,
+      "grad_norm": 0.012852554209530354,
+      "learning_rate": 0.0002,
+      "loss": 0.6335437297821045,
+      "mean_token_accuracy": 0.7411114126443863,
+      "num_tokens": 27342734.0,
+      "step": 737
+    },
+    {
+      "entropy": 0.6231958642601967,
+      "epoch": 4.7656123276561235,
+      "grad_norm": 0.01126949768513441,
+      "learning_rate": 0.0002,
+      "loss": 0.6242541670799255,
+      "mean_token_accuracy": 0.7444687336683273,
+      "num_tokens": 27380129.0,
+      "step": 738
+    },
+    {
+      "entropy": 0.6135790795087814,
+      "epoch": 4.772100567721005,
+      "grad_norm": 0.011775590479373932,
+      "learning_rate": 0.0002,
+      "loss": 0.6142154335975647,
+      "mean_token_accuracy": 0.7484701201319695,
+      "num_tokens": 27417574.0,
+      "step": 739
+    },
+    {
+      "entropy": 0.5981535315513611,
+      "epoch": 4.778588807785888,
+      "grad_norm": 0.012303443625569344,
+      "learning_rate": 0.0002,
+      "loss": 0.6043405532836914,
+      "mean_token_accuracy": 0.7530733942985535,
+      "num_tokens": 27454673.0,
+      "step": 740
+    },
+    {
+      "entropy": 0.6199762895703316,
+      "epoch": 4.785077047850771,
+      "grad_norm": 0.011197846382856369,
+      "learning_rate": 0.0002,
+      "loss": 0.6214004158973694,
+      "mean_token_accuracy": 0.7454070076346397,
+      "num_tokens": 27492327.0,
+      "step": 741
+    },
+    {
+      "entropy": 0.6030875891447067,
+      "epoch": 4.7915652879156525,
+      "grad_norm": 0.012394090183079243,
+      "learning_rate": 0.0002,
+      "loss": 0.6037598848342896,
+      "mean_token_accuracy": 0.7523747533559799,
+      "num_tokens": 27529616.0,
+      "step": 742
+    },
+    {
+      "entropy": 0.613655112683773,
+      "epoch": 4.798053527980535,
+      "grad_norm": 0.010351658798754215,
+      "learning_rate": 0.0002,
+      "loss": 0.6146197319030762,
+      "mean_token_accuracy": 0.7493305876851082,
+      "num_tokens": 27567467.0,
+      "step": 743
+    },
+    {
+      "entropy": 0.609668679535389,
+      "epoch": 4.804541768045418,
+      "grad_norm": 0.010698414407670498,
+      "learning_rate": 0.0002,
+      "loss": 0.6130056381225586,
+      "mean_token_accuracy": 0.749318391084671,
+      "num_tokens": 27604340.0,
+      "step": 744
+    },
+    {
+      "entropy": 0.6055747121572495,
+      "epoch": 4.811030008110301,
+      "grad_norm": 0.011774055659770966,
+      "learning_rate": 0.0002,
+      "loss": 0.6064777374267578,
+      "mean_token_accuracy": 0.7513219490647316,
+      "num_tokens": 27641869.0,
+      "step": 745
+    },
+    {
+      "entropy": 0.6063344553112984,
+      "epoch": 4.817518248175182,
+      "grad_norm": 0.010040192864835262,
+      "learning_rate": 0.0002,
+      "loss": 0.6085660457611084,
+      "mean_token_accuracy": 0.7483800128102303,
+      "num_tokens": 27678992.0,
+      "step": 746
+    },
+    {
+      "entropy": 0.6175437271595001,
+      "epoch": 4.824006488240065,
+      "grad_norm": 0.011712496168911457,
+      "learning_rate": 0.0002,
+      "loss": 0.6189186573028564,
+      "mean_token_accuracy": 0.7486408054828644,
+      "num_tokens": 27716703.0,
+      "step": 747
+    },
+    {
+      "entropy": 0.6142439544200897,
+      "epoch": 4.830494728304947,
+      "grad_norm": 0.012662764638662338,
+      "learning_rate": 0.0002,
+      "loss": 0.614926815032959,
+      "mean_token_accuracy": 0.747495986521244,
+      "num_tokens": 27754004.0,
+      "step": 748
+    },
+    {
+      "entropy": 0.6087512150406837,
+      "epoch": 4.8369829683698295,
+      "grad_norm": 0.010618063621222973,
+      "learning_rate": 0.0002,
+      "loss": 0.6098265647888184,
+      "mean_token_accuracy": 0.7527805492281914,
+      "num_tokens": 27791212.0,
+      "step": 749
+    },
+    {
+      "entropy": 0.6073944419622421,
+      "epoch": 4.843471208434712,
+      "grad_norm": 0.011879456229507923,
+      "learning_rate": 0.0002,
+      "loss": 0.6084223389625549,
+      "mean_token_accuracy": 0.7535420805215836,
+      "num_tokens": 27828439.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6072632223367691,
+      "epoch": 4.849959448499595,
+      "grad_norm": 0.01333905290812254,
+      "learning_rate": 0.0002,
+      "loss": 0.6154837608337402,
+      "mean_token_accuracy": 0.7462009564042091,
+      "num_tokens": 27865908.0,
+      "step": 751
+    },
+    {
+      "entropy": 0.6182335615158081,
+      "epoch": 4.856447688564477,
+      "grad_norm": 0.012253145687282085,
+      "learning_rate": 0.0002,
+      "loss": 0.6162886619567871,
+      "mean_token_accuracy": 0.7482575252652168,
+      "num_tokens": 27903553.0,
+      "step": 752
+    },
+    {
+      "entropy": 0.623457059264183,
+      "epoch": 4.862935928629359,
+      "grad_norm": 0.012347408570349216,
+      "learning_rate": 0.0002,
+      "loss": 0.6175649166107178,
+      "mean_token_accuracy": 0.7451874390244484,
+      "num_tokens": 27940745.0,
+      "step": 753
+    },
+    {
+      "entropy": 0.6147228851914406,
+      "epoch": 4.869424168694241,
+      "grad_norm": 0.013636295683681965,
+      "learning_rate": 0.0002,
+      "loss": 0.6182140111923218,
+      "mean_token_accuracy": 0.7459222003817558,
+      "num_tokens": 27977798.0,
+      "step": 754
+    },
+    {
+      "entropy": 0.6009573265910149,
+      "epoch": 4.875912408759124,
+      "grad_norm": 0.011285821907222271,
+      "learning_rate": 0.0002,
+      "loss": 0.6073520183563232,
+      "mean_token_accuracy": 0.7508896961808205,
+      "num_tokens": 28014880.0,
+      "step": 755
+    },
+    {
+      "entropy": 0.5950382575392723,
+      "epoch": 4.882400648824007,
+      "grad_norm": 0.011523284018039703,
+      "learning_rate": 0.0002,
+      "loss": 0.6006320714950562,
+      "mean_token_accuracy": 0.7559562996029854,
+      "num_tokens": 28052148.0,
+      "step": 756
+    },
+    {
+      "entropy": 0.6160282120108604,
+      "epoch": 4.888888888888889,
+      "grad_norm": 0.011825107038021088,
+      "learning_rate": 0.0002,
+      "loss": 0.6167672872543335,
+      "mean_token_accuracy": 0.7469073683023453,
+      "num_tokens": 28089658.0,
+      "step": 757
+    },
+    {
+      "entropy": 0.6171636879444122,
+      "epoch": 4.895377128953771,
+      "grad_norm": 0.011352675966918468,
+      "learning_rate": 0.0002,
+      "loss": 0.6145755052566528,
+      "mean_token_accuracy": 0.746620424091816,
+      "num_tokens": 28127170.0,
+      "step": 758
+    },
+    {
+      "entropy": 0.608257532119751,
+      "epoch": 4.901865369018654,
+      "grad_norm": 0.012769036926329136,
+      "learning_rate": 0.0002,
+      "loss": 0.609970211982727,
+      "mean_token_accuracy": 0.7516561150550842,
+      "num_tokens": 28164347.0,
+      "step": 759
+    },
+    {
+      "entropy": 0.6164683550596237,
+      "epoch": 4.9083536090835365,
+      "grad_norm": 0.012305669486522675,
+      "learning_rate": 0.0002,
+      "loss": 0.6191939115524292,
+      "mean_token_accuracy": 0.7453163638710976,
+      "num_tokens": 28201799.0,
+      "step": 760
+    },
+    {
+      "entropy": 0.614104337990284,
+      "epoch": 4.914841849148418,
+      "grad_norm": 0.010976862162351608,
+      "learning_rate": 0.0002,
+      "loss": 0.6142949461936951,
+      "mean_token_accuracy": 0.7480729147791862,
+      "num_tokens": 28239441.0,
+      "step": 761
+    },
+    {
+      "entropy": 0.5957873240113258,
+      "epoch": 4.921330089213301,
+      "grad_norm": 0.011151162907481194,
+      "learning_rate": 0.0002,
+      "loss": 0.5960466265678406,
+      "mean_token_accuracy": 0.7578283548355103,
+      "num_tokens": 28276966.0,
+      "step": 762
+    },
+    {
+      "entropy": 0.6149696856737137,
+      "epoch": 4.927818329278184,
+      "grad_norm": 0.01262627262622118,
+      "learning_rate": 0.0002,
+      "loss": 0.6141815185546875,
+      "mean_token_accuracy": 0.7478419542312622,
+      "num_tokens": 28314493.0,
+      "step": 763
+    },
+    {
+      "entropy": 0.6079921796917915,
+      "epoch": 4.934306569343065,
+      "grad_norm": 0.012012857012450695,
+      "learning_rate": 0.0002,
+      "loss": 0.6085329055786133,
+      "mean_token_accuracy": 0.7516164556145668,
+      "num_tokens": 28352110.0,
+      "step": 764
+    },
+    {
+      "entropy": 0.606193870306015,
+      "epoch": 4.940794809407948,
+      "grad_norm": 0.012743561528623104,
+      "learning_rate": 0.0002,
+      "loss": 0.6102321147918701,
+      "mean_token_accuracy": 0.7491661831736565,
+      "num_tokens": 28389448.0,
+      "step": 765
+    },
+    {
+      "entropy": 0.6021564230322838,
+      "epoch": 4.947283049472831,
+      "grad_norm": 0.011576979421079159,
+      "learning_rate": 0.0002,
+      "loss": 0.6045505404472351,
+      "mean_token_accuracy": 0.7516537830233574,
+      "num_tokens": 28427005.0,
+      "step": 766
+    },
+    {
+      "entropy": 0.6197597831487656,
+      "epoch": 4.953771289537713,
+      "grad_norm": 0.013263656757771969,
+      "learning_rate": 0.0002,
+      "loss": 0.6234422922134399,
+      "mean_token_accuracy": 0.7442535609006882,
+      "num_tokens": 28464143.0,
+      "step": 767
+    },
+    {
+      "entropy": 0.5995957404375076,
+      "epoch": 4.960259529602595,
+      "grad_norm": 0.013597379438579082,
+      "learning_rate": 0.0002,
+      "loss": 0.608078122138977,
+      "mean_token_accuracy": 0.7510829046368599,
+      "num_tokens": 28501245.0,
+      "step": 768
+    },
+    {
+      "entropy": 0.5990025177598,
+      "epoch": 4.966747769667478,
+      "grad_norm": 0.011187846772372723,
+      "learning_rate": 0.0002,
+      "loss": 0.6021981835365295,
+      "mean_token_accuracy": 0.751527689397335,
+      "num_tokens": 28538395.0,
+      "step": 769
+    },
+    {
+      "entropy": 0.6094980016350746,
+      "epoch": 4.97323600973236,
+      "grad_norm": 0.011942482553422451,
+      "learning_rate": 0.0002,
+      "loss": 0.6086335182189941,
+      "mean_token_accuracy": 0.7511740922927856,
+      "num_tokens": 28575389.0,
+      "step": 770
+    },
+    {
+      "entropy": 0.607696995139122,
+      "epoch": 4.9797242497972425,
+      "grad_norm": 0.010719696059823036,
+      "learning_rate": 0.0002,
+      "loss": 0.6058928370475769,
+      "mean_token_accuracy": 0.7512983754277229,
+      "num_tokens": 28612614.0,
+      "step": 771
+    },
+    {
+      "entropy": 0.6127375960350037,
+      "epoch": 4.986212489862125,
+      "grad_norm": 0.011552160605788231,
+      "learning_rate": 0.0002,
+      "loss": 0.610999345779419,
+      "mean_token_accuracy": 0.7520032823085785,
+      "num_tokens": 28649626.0,
+      "step": 772
+    },
+    {
+      "entropy": 0.6094198897480965,
+      "epoch": 4.992700729927007,
+      "grad_norm": 0.010641835629940033,
+      "learning_rate": 0.0002,
+      "loss": 0.609566330909729,
+      "mean_token_accuracy": 0.7494478672742844,
+      "num_tokens": 28687121.0,
+      "step": 773
+    },
+    {
+      "entropy": 0.6135342344641685,
+      "epoch": 4.99918896999189,
+      "grad_norm": 0.010831725783646107,
+      "learning_rate": 0.0002,
+      "loss": 0.6144300103187561,
+      "mean_token_accuracy": 0.7496655061841011,
+      "num_tokens": 28724466.0,
+      "step": 774
+    },
+    {
+      "entropy": 0.6133913993835449,
+      "epoch": 5.0,
+      "grad_norm": 0.025664443150162697,
+      "learning_rate": 0.0002,
+      "loss": 0.6142559051513672,
+      "mean_token_accuracy": 0.7518454194068909,
+      "num_tokens": 28729104.0,
+      "step": 775
     }
   ],
   "logging_steps": 1,
-  "max_steps": 540,
+  "max_steps": 775,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 10,
+  "num_train_epochs": 5,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -5427,8 +7777,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 2.7784501160281375e+18,
-  "train_batch_size": 16,
+  "total_flos": 2.6824700471332045e+18,
+  "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null
 }