diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,10043 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 6252,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0023992322456813818,
+      "grad_norm": 98.49387406847768,
+      "learning_rate": 3.9936102236421723e-07,
+      "loss": 3.4946,
+      "mean_token_accuracy": 0.4661314785480499,
+      "step": 5
+    },
+    {
+      "epoch": 0.0047984644913627635,
+      "grad_norm": 99.1562953039722,
+      "learning_rate": 7.987220447284345e-07,
+      "loss": 3.3942,
+      "mean_token_accuracy": 0.4688503980636597,
+      "step": 10
+    },
+    {
+      "epoch": 0.007197696737044146,
+      "grad_norm": 103.59800633052092,
+      "learning_rate": 1.1980830670926517e-06,
+      "loss": 2.8201,
+      "mean_token_accuracy": 0.5040203750133514,
+      "step": 15
+    },
+    {
+      "epoch": 0.009596928982725527,
+      "grad_norm": 41.28134143567553,
+      "learning_rate": 1.597444089456869e-06,
+      "loss": 2.2437,
+      "mean_token_accuracy": 0.5471958994865418,
+      "step": 20
+    },
+    {
+      "epoch": 0.01199616122840691,
+      "grad_norm": 7.008155501822939,
+      "learning_rate": 1.9968051118210863e-06,
+      "loss": 1.9265,
+      "mean_token_accuracy": 0.5786412835121155,
+      "step": 25
+    },
+    {
+      "epoch": 0.014395393474088292,
+      "grad_norm": 3.7304961794538944,
+      "learning_rate": 2.3961661341853035e-06,
+      "loss": 1.6877,
+      "mean_token_accuracy": 0.6062699258327484,
+      "step": 30
+    },
+    {
+      "epoch": 0.016794625719769675,
+      "grad_norm": 3.161125047418359,
+      "learning_rate": 2.7955271565495207e-06,
+      "loss": 1.5141,
+      "mean_token_accuracy": 0.6290818691253662,
+      "step": 35
+    },
+    {
+      "epoch": 0.019193857965451054,
+      "grad_norm": 3.2545444446941714,
+      "learning_rate": 3.194888178913738e-06,
+      "loss": 1.3254,
+      "mean_token_accuracy": 0.6603264272212982,
+      "step": 40
+    },
+    {
+      "epoch": 0.021593090211132437,
+      "grad_norm": 2.5316219364038686,
+      "learning_rate": 3.5942492012779555e-06,
+      "loss": 1.1635,
+      "mean_token_accuracy": 0.692139345407486,
+      "step": 45
+    },
+    {
+      "epoch": 0.02399232245681382,
+      "grad_norm": 2.401613446779219,
+      "learning_rate": 3.993610223642173e-06,
+      "loss": 1.0714,
+      "mean_token_accuracy": 0.7093232333660126,
+      "step": 50
+    },
+    {
+      "epoch": 0.026391554702495202,
+      "grad_norm": 3.26330685077421,
+      "learning_rate": 4.39297124600639e-06,
+      "loss": 0.9826,
+      "mean_token_accuracy": 0.7275775074958801,
+      "step": 55
+    },
+    {
+      "epoch": 0.028790786948176585,
+      "grad_norm": 2.618739831701825,
+      "learning_rate": 4.792332268370607e-06,
+      "loss": 0.9195,
+      "mean_token_accuracy": 0.7401251554489136,
+      "step": 60
+    },
+    {
+      "epoch": 0.031190019193857964,
+      "grad_norm": 2.398659345240197,
+      "learning_rate": 5.191693290734825e-06,
+      "loss": 0.8725,
+      "mean_token_accuracy": 0.7501980125904083,
+      "step": 65
+    },
+    {
+      "epoch": 0.03358925143953935,
+      "grad_norm": 1.9361958724739154,
+      "learning_rate": 5.591054313099041e-06,
+      "loss": 0.8286,
+      "mean_token_accuracy": 0.760456132888794,
+      "step": 70
+    },
+    {
+      "epoch": 0.03598848368522073,
+      "grad_norm": 1.888737531555722,
+      "learning_rate": 5.990415335463259e-06,
+      "loss": 0.8147,
+      "mean_token_accuracy": 0.7615372240543365,
+      "step": 75
+    },
+    {
+      "epoch": 0.03838771593090211,
+      "grad_norm": 1.7614713374922835,
+      "learning_rate": 6.389776357827476e-06,
+      "loss": 0.7833,
+      "mean_token_accuracy": 0.7687996089458465,
+      "step": 80
+    },
+    {
+      "epoch": 0.040786948176583494,
+      "grad_norm": 1.5424851045060726,
+      "learning_rate": 6.789137380191693e-06,
+      "loss": 0.7614,
+      "mean_token_accuracy": 0.7746296703815461,
+      "step": 85
+    },
+    {
+      "epoch": 0.04318618042226487,
+      "grad_norm": 1.4479842150627094,
+      "learning_rate": 7.188498402555911e-06,
+      "loss": 0.7418,
+      "mean_token_accuracy": 0.7788853943347931,
+      "step": 90
+    },
+    {
+      "epoch": 0.04558541266794626,
+      "grad_norm": 1.573368555060965,
+      "learning_rate": 7.5878594249201285e-06,
+      "loss": 0.7137,
+      "mean_token_accuracy": 0.7856944262981415,
+      "step": 95
+    },
+    {
+      "epoch": 0.04798464491362764,
+      "grad_norm": 1.4141414526158402,
+      "learning_rate": 7.987220447284345e-06,
+      "loss": 0.7205,
+      "mean_token_accuracy": 0.783971655368805,
+      "step": 100
+    },
+    {
+      "epoch": 0.05038387715930902,
+      "grad_norm": 1.9824310652518375,
+      "learning_rate": 8.386581469648563e-06,
+      "loss": 0.71,
+      "mean_token_accuracy": 0.7859264612197876,
+      "step": 105
+    },
+    {
+      "epoch": 0.052783109404990404,
+      "grad_norm": 1.9827753371234091,
+      "learning_rate": 8.78594249201278e-06,
+      "loss": 0.708,
+      "mean_token_accuracy": 0.7866788566112518,
+      "step": 110
+    },
+    {
+      "epoch": 0.05518234165067178,
+      "grad_norm": 1.2728746495222643,
+      "learning_rate": 9.185303514376996e-06,
+      "loss": 0.6948,
+      "mean_token_accuracy": 0.7887039840221405,
+      "step": 115
+    },
+    {
+      "epoch": 0.05758157389635317,
+      "grad_norm": 1.5593829761863378,
+      "learning_rate": 9.584664536741214e-06,
+      "loss": 0.693,
+      "mean_token_accuracy": 0.7892837047576904,
+      "step": 120
+    },
+    {
+      "epoch": 0.05998080614203455,
+      "grad_norm": 1.33665011351666,
+      "learning_rate": 9.984025559105432e-06,
+      "loss": 0.6705,
+      "mean_token_accuracy": 0.79561066031456,
+      "step": 125
+    },
+    {
+      "epoch": 0.06238003838771593,
+      "grad_norm": 1.650657003715394,
+      "learning_rate": 1.038338658146965e-05,
+      "loss": 0.6588,
+      "mean_token_accuracy": 0.7983006715774537,
+      "step": 130
+    },
+    {
+      "epoch": 0.0647792706333973,
+      "grad_norm": 1.2501851092297778,
+      "learning_rate": 1.0782747603833867e-05,
+      "loss": 0.6706,
+      "mean_token_accuracy": 0.7948801577091217,
+      "step": 135
+    },
+    {
+      "epoch": 0.0671785028790787,
+      "grad_norm": 1.489237643326096,
+      "learning_rate": 1.1182108626198083e-05,
+      "loss": 0.669,
+      "mean_token_accuracy": 0.7941152632236481,
+      "step": 140
+    },
+    {
+      "epoch": 0.06957773512476008,
+      "grad_norm": 1.1801378516581407,
+      "learning_rate": 1.1581469648562302e-05,
+      "loss": 0.6692,
+      "mean_token_accuracy": 0.7941246449947357,
+      "step": 145
+    },
+    {
+      "epoch": 0.07197696737044146,
+      "grad_norm": 1.3234370393319912,
+      "learning_rate": 1.1980830670926518e-05,
+      "loss": 0.6601,
+      "mean_token_accuracy": 0.7972006022930145,
+      "step": 150
+    },
+    {
+      "epoch": 0.07437619961612284,
+      "grad_norm": 1.3266843826076535,
+      "learning_rate": 1.2380191693290735e-05,
+      "loss": 0.6431,
+      "mean_token_accuracy": 0.8009279251098633,
+      "step": 155
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 1.171676614531739,
+      "learning_rate": 1.2779552715654951e-05,
+      "loss": 0.6441,
+      "mean_token_accuracy": 0.8012164771556854,
+      "step": 160
+    },
+    {
+      "epoch": 0.07917466410748561,
+      "grad_norm": 1.191195633173334,
+      "learning_rate": 1.3178913738019169e-05,
+      "loss": 0.6313,
+      "mean_token_accuracy": 0.8047502994537353,
+      "step": 165
+    },
+    {
+      "epoch": 0.08157389635316699,
+      "grad_norm": 1.2319321923794087,
+      "learning_rate": 1.3578274760383387e-05,
+      "loss": 0.6375,
+      "mean_token_accuracy": 0.8019321501255036,
+      "step": 170
+    },
+    {
+      "epoch": 0.08397312859884837,
+      "grad_norm": 1.4534520664061426,
+      "learning_rate": 1.3977635782747606e-05,
+      "loss": 0.6436,
+      "mean_token_accuracy": 0.8006344079971314,
+      "step": 175
+    },
+    {
+      "epoch": 0.08637236084452975,
+      "grad_norm": 1.2484396717014798,
+      "learning_rate": 1.4376996805111822e-05,
+      "loss": 0.6298,
+      "mean_token_accuracy": 0.8044987142086029,
+      "step": 180
+    },
+    {
+      "epoch": 0.08877159309021113,
+      "grad_norm": 1.401816255311233,
+      "learning_rate": 1.477635782747604e-05,
+      "loss": 0.6412,
+      "mean_token_accuracy": 0.8011281907558441,
+      "step": 185
+    },
+    {
+      "epoch": 0.09117082533589252,
+      "grad_norm": 1.3511684768747914,
+      "learning_rate": 1.5175718849840257e-05,
+      "loss": 0.621,
+      "mean_token_accuracy": 0.8077153265476227,
+      "step": 190
+    },
+    {
+      "epoch": 0.0935700575815739,
+      "grad_norm": 1.108573103056433,
+      "learning_rate": 1.5575079872204475e-05,
+      "loss": 0.6172,
+      "mean_token_accuracy": 0.8076364099979401,
+      "step": 195
+    },
+    {
+      "epoch": 0.09596928982725528,
+      "grad_norm": 1.3243400421217433,
+      "learning_rate": 1.597444089456869e-05,
+      "loss": 0.6385,
+      "mean_token_accuracy": 0.802015745639801,
+      "step": 200
+    },
+    {
+      "epoch": 0.09836852207293666,
+      "grad_norm": 1.5147183355913882,
+      "learning_rate": 1.6373801916932906e-05,
+      "loss": 0.6273,
+      "mean_token_accuracy": 0.8052714228630066,
+      "step": 205
+    },
+    {
+      "epoch": 0.10076775431861804,
+      "grad_norm": 1.0814564338138692,
+      "learning_rate": 1.6773162939297126e-05,
+      "loss": 0.6159,
+      "mean_token_accuracy": 0.8079419076442719,
+      "step": 210
+    },
+    {
+      "epoch": 0.10316698656429943,
+      "grad_norm": 1.1275338205488943,
+      "learning_rate": 1.7172523961661345e-05,
+      "loss": 0.6347,
+      "mean_token_accuracy": 0.801912808418274,
+      "step": 215
+    },
+    {
+      "epoch": 0.10556621880998081,
+      "grad_norm": 1.0329542633437603,
+      "learning_rate": 1.757188498402556e-05,
+      "loss": 0.6151,
+      "mean_token_accuracy": 0.8077832996845246,
+      "step": 220
+    },
+    {
+      "epoch": 0.10796545105566219,
+      "grad_norm": 1.0527318791464246,
+      "learning_rate": 1.7971246006389777e-05,
+      "loss": 0.5926,
+      "mean_token_accuracy": 0.8138438284397125,
+      "step": 225
+    },
+    {
+      "epoch": 0.11036468330134357,
+      "grad_norm": 1.0376015582992664,
+      "learning_rate": 1.8370607028753993e-05,
+      "loss": 0.6067,
+      "mean_token_accuracy": 0.8101162552833557,
+      "step": 230
+    },
+    {
+      "epoch": 0.11276391554702495,
+      "grad_norm": 1.2251681217622086,
+      "learning_rate": 1.8769968051118212e-05,
+      "loss": 0.6252,
+      "mean_token_accuracy": 0.8043080806732178,
+      "step": 235
+    },
+    {
+      "epoch": 0.11516314779270634,
+      "grad_norm": 1.1885968023406797,
+      "learning_rate": 1.9169329073482428e-05,
+      "loss": 0.6126,
+      "mean_token_accuracy": 0.8087013304233551,
+      "step": 240
+    },
+    {
+      "epoch": 0.11756238003838772,
+      "grad_norm": 1.2212514526533986,
+      "learning_rate": 1.9568690095846644e-05,
+      "loss": 0.6055,
+      "mean_token_accuracy": 0.8101045370101929,
+      "step": 245
+    },
+    {
+      "epoch": 0.1199616122840691,
+      "grad_norm": 1.1976425853454404,
+      "learning_rate": 1.9968051118210863e-05,
+      "loss": 0.6204,
+      "mean_token_accuracy": 0.8052823603153229,
+      "step": 250
+    },
+    {
+      "epoch": 0.12236084452975048,
+      "grad_norm": 1.0624500234563632,
+      "learning_rate": 2.0367412140575082e-05,
+      "loss": 0.6026,
+      "mean_token_accuracy": 0.8114176630973816,
+      "step": 255
+    },
+    {
+      "epoch": 0.12476007677543186,
+      "grad_norm": 1.3455443254267732,
+      "learning_rate": 2.07667731629393e-05,
+      "loss": 0.6042,
+      "mean_token_accuracy": 0.8102307736873626,
+      "step": 260
+    },
+    {
+      "epoch": 0.12715930902111325,
+      "grad_norm": 1.1145770451465071,
+      "learning_rate": 2.1166134185303514e-05,
+      "loss": 0.6104,
+      "mean_token_accuracy": 0.8091982364654541,
+      "step": 265
+    },
+    {
+      "epoch": 0.1295585412667946,
+      "grad_norm": 0.9619063430955215,
+      "learning_rate": 2.1565495207667734e-05,
+      "loss": 0.6051,
+      "mean_token_accuracy": 0.8094959020614624,
+      "step": 270
+    },
+    {
+      "epoch": 0.131957773512476,
+      "grad_norm": 1.0999050169577322,
+      "learning_rate": 2.196485623003195e-05,
+      "loss": 0.6129,
+      "mean_token_accuracy": 0.807866895198822,
+      "step": 275
+    },
+    {
+      "epoch": 0.1343570057581574,
+      "grad_norm": 1.383699374554589,
+      "learning_rate": 2.2364217252396165e-05,
+      "loss": 0.6185,
+      "mean_token_accuracy": 0.8059761583805084,
+      "step": 280
+    },
+    {
+      "epoch": 0.13675623800383876,
+      "grad_norm": 0.8522287990957228,
+      "learning_rate": 2.2763578274760385e-05,
+      "loss": 0.6093,
+      "mean_token_accuracy": 0.8090029060840607,
+      "step": 285
+    },
+    {
+      "epoch": 0.13915547024952016,
+      "grad_norm": 0.9617614265565335,
+      "learning_rate": 2.3162939297124604e-05,
+      "loss": 0.6076,
+      "mean_token_accuracy": 0.8090137720108033,
+      "step": 290
+    },
+    {
+      "epoch": 0.14155470249520152,
+      "grad_norm": 1.0820740523297105,
+      "learning_rate": 2.356230031948882e-05,
+      "loss": 0.6108,
+      "mean_token_accuracy": 0.8079468786716462,
+      "step": 295
+    },
+    {
+      "epoch": 0.14395393474088292,
+      "grad_norm": 1.208085906040005,
+      "learning_rate": 2.3961661341853036e-05,
+      "loss": 0.6137,
+      "mean_token_accuracy": 0.8066973686218262,
+      "step": 300
+    },
+    {
+      "epoch": 0.1463531669865643,
+      "grad_norm": 1.2867186093418053,
+      "learning_rate": 2.4361022364217255e-05,
+      "loss": 0.6165,
+      "mean_token_accuracy": 0.806743037700653,
+      "step": 305
+    },
+    {
+      "epoch": 0.14875239923224567,
+      "grad_norm": 1.2335001938431094,
+      "learning_rate": 2.476038338658147e-05,
+      "loss": 0.6184,
+      "mean_token_accuracy": 0.8057464599609375,
+      "step": 310
+    },
+    {
+      "epoch": 0.15115163147792707,
+      "grad_norm": 1.1353843569347062,
+      "learning_rate": 2.515974440894569e-05,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.8137832462787629,
+      "step": 315
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 0.9520234538262042,
+      "learning_rate": 2.5559105431309903e-05,
+      "loss": 0.6065,
+      "mean_token_accuracy": 0.8100951492786408,
+      "step": 320
+    },
+    {
+      "epoch": 0.15595009596928983,
+      "grad_norm": 1.1027272126900591,
+      "learning_rate": 2.5958466453674125e-05,
+      "loss": 0.6046,
+      "mean_token_accuracy": 0.8104814648628235,
+      "step": 325
+    },
+    {
+      "epoch": 0.15834932821497122,
+      "grad_norm": 0.9494045507221671,
+      "learning_rate": 2.6357827476038338e-05,
+      "loss": 0.6021,
+      "mean_token_accuracy": 0.8098224759101867,
+      "step": 330
+    },
+    {
+      "epoch": 0.16074856046065258,
+      "grad_norm": 1.064270767277033,
+      "learning_rate": 2.6757188498402557e-05,
+      "loss": 0.6028,
+      "mean_token_accuracy": 0.8100285053253173,
+      "step": 335
+    },
+    {
+      "epoch": 0.16314779270633398,
+      "grad_norm": 0.9515730626704475,
+      "learning_rate": 2.7156549520766773e-05,
+      "loss": 0.6034,
+      "mean_token_accuracy": 0.809994375705719,
+      "step": 340
+    },
+    {
+      "epoch": 0.16554702495201534,
+      "grad_norm": 1.1169315271788858,
+      "learning_rate": 2.7555910543130992e-05,
+      "loss": 0.617,
+      "mean_token_accuracy": 0.8061847150325775,
+      "step": 345
+    },
+    {
+      "epoch": 0.16794625719769674,
+      "grad_norm": 1.0117750051838819,
+      "learning_rate": 2.7955271565495212e-05,
+      "loss": 0.5877,
+      "mean_token_accuracy": 0.8143430769443512,
+      "step": 350
+    },
+    {
+      "epoch": 0.17034548944337813,
+      "grad_norm": 1.011758660828968,
+      "learning_rate": 2.8354632587859424e-05,
+      "loss": 0.5902,
+      "mean_token_accuracy": 0.813703978061676,
+      "step": 355
+    },
+    {
+      "epoch": 0.1727447216890595,
+      "grad_norm": 0.991649818964084,
+      "learning_rate": 2.8753993610223644e-05,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8137360095977784,
+      "step": 360
+    },
+    {
+      "epoch": 0.1751439539347409,
+      "grad_norm": 0.9886966147061703,
+      "learning_rate": 2.915335463258786e-05,
+      "loss": 0.5835,
+      "mean_token_accuracy": 0.8160845398902893,
+      "step": 365
+    },
+    {
+      "epoch": 0.17754318618042225,
+      "grad_norm": 0.9333152812140807,
+      "learning_rate": 2.955271565495208e-05,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.8131695687770844,
+      "step": 370
+    },
+    {
+      "epoch": 0.17994241842610365,
+      "grad_norm": 0.8650635350560877,
+      "learning_rate": 2.9952076677316295e-05,
+      "loss": 0.5897,
+      "mean_token_accuracy": 0.8132461369037628,
+      "step": 375
+    },
+    {
+      "epoch": 0.18234165067178504,
+      "grad_norm": 1.0526197764669132,
+      "learning_rate": 3.0351437699680514e-05,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8123046815395355,
+      "step": 380
+    },
+    {
+      "epoch": 0.1847408829174664,
+      "grad_norm": 0.8702811072498366,
+      "learning_rate": 3.075079872204473e-05,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.815815019607544,
+      "step": 385
+    },
+    {
+      "epoch": 0.1871401151631478,
+      "grad_norm": 1.0530301566850875,
+      "learning_rate": 3.115015974440895e-05,
+      "loss": 0.5886,
+      "mean_token_accuracy": 0.8134219169616699,
+      "step": 390
+    },
+    {
+      "epoch": 0.18953934740882916,
+      "grad_norm": 1.0645206955111999,
+      "learning_rate": 3.154952076677317e-05,
+      "loss": 0.6064,
+      "mean_token_accuracy": 0.8076598584651947,
+      "step": 395
+    },
+    {
+      "epoch": 0.19193857965451055,
+      "grad_norm": 0.928351802093768,
+      "learning_rate": 3.194888178913738e-05,
+      "loss": 0.5986,
+      "mean_token_accuracy": 0.8101225137710572,
+      "step": 400
+    },
+    {
+      "epoch": 0.19433781190019195,
+      "grad_norm": 0.8939754967147308,
+      "learning_rate": 3.23482428115016e-05,
+      "loss": 0.5881,
+      "mean_token_accuracy": 0.8137291073799133,
+      "step": 405
+    },
+    {
+      "epoch": 0.1967370441458733,
+      "grad_norm": 0.8998458172125146,
+      "learning_rate": 3.274760383386581e-05,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.815822857618332,
+      "step": 410
+    },
+    {
+      "epoch": 0.1991362763915547,
+      "grad_norm": 0.9045669856998589,
+      "learning_rate": 3.314696485623003e-05,
+      "loss": 0.5965,
+      "mean_token_accuracy": 0.810960841178894,
+      "step": 415
+    },
+    {
+      "epoch": 0.20153550863723607,
+      "grad_norm": 1.2366817121637053,
+      "learning_rate": 3.354632587859425e-05,
+      "loss": 0.6007,
+      "mean_token_accuracy": 0.8106834948062897,
+      "step": 420
+    },
+    {
+      "epoch": 0.20393474088291746,
+      "grad_norm": 1.1241688540561303,
+      "learning_rate": 3.394568690095847e-05,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8151392281055451,
+      "step": 425
+    },
+    {
+      "epoch": 0.20633397312859886,
+      "grad_norm": 0.8871443151549863,
+      "learning_rate": 3.434504792332269e-05,
+      "loss": 0.5939,
+      "mean_token_accuracy": 0.8126257479190826,
+      "step": 430
+    },
+    {
+      "epoch": 0.20873320537428022,
+      "grad_norm": 0.8424395176147732,
+      "learning_rate": 3.47444089456869e-05,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8188099205493927,
+      "step": 435
+    },
+    {
+      "epoch": 0.21113243761996162,
+      "grad_norm": 0.7827214371353761,
+      "learning_rate": 3.514376996805112e-05,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8164414882659912,
+      "step": 440
+    },
+    {
+      "epoch": 0.21353166986564298,
+      "grad_norm": 0.9316075347810953,
+      "learning_rate": 3.5543130990415334e-05,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8147446632385253,
+      "step": 445
+    },
+    {
+      "epoch": 0.21593090211132437,
+      "grad_norm": 1.0158313500744018,
+      "learning_rate": 3.5942492012779554e-05,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.8138821184635162,
+      "step": 450
+    },
+    {
+      "epoch": 0.21833013435700577,
+      "grad_norm": 0.8530663228237867,
+      "learning_rate": 3.6341853035143766e-05,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8160447537899017,
+      "step": 455
+    },
+    {
+      "epoch": 0.22072936660268713,
+      "grad_norm": 0.8480787301010322,
+      "learning_rate": 3.6741214057507985e-05,
+      "loss": 0.5842,
+      "mean_token_accuracy": 0.8153697609901428,
+      "step": 460
+    },
+    {
+      "epoch": 0.22312859884836853,
+      "grad_norm": 0.7935009865546706,
+      "learning_rate": 3.714057507987221e-05,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8143532335758209,
+      "step": 465
+    },
+    {
+      "epoch": 0.2255278310940499,
+      "grad_norm": 0.9176259936940574,
+      "learning_rate": 3.7539936102236424e-05,
+      "loss": 0.5954,
+      "mean_token_accuracy": 0.8111553907394409,
+      "step": 470
+    },
+    {
+      "epoch": 0.22792706333973128,
+      "grad_norm": 0.9937759416482129,
+      "learning_rate": 3.793929712460064e-05,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8177870512008667,
+      "step": 475
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 0.9396616647288063,
+      "learning_rate": 3.8338658146964856e-05,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8213005423545837,
+      "step": 480
+    },
+    {
+      "epoch": 0.23272552783109404,
+      "grad_norm": 0.9673726655600524,
+      "learning_rate": 3.8738019169329075e-05,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8171282410621643,
+      "step": 485
+    },
+    {
+      "epoch": 0.23512476007677544,
+      "grad_norm": 1.397789609885595,
+      "learning_rate": 3.913738019169329e-05,
+      "loss": 0.5965,
+      "mean_token_accuracy": 0.8120984196662903,
+      "step": 490
+    },
+    {
+      "epoch": 0.2375239923224568,
+      "grad_norm": 1.4129747649477833,
+      "learning_rate": 3.953674121405751e-05,
+      "loss": 0.579,
+      "mean_token_accuracy": 0.8163260221481323,
+      "step": 495
+    },
+    {
+      "epoch": 0.2399232245681382,
+      "grad_norm": 0.912512990318671,
+      "learning_rate": 3.9936102236421726e-05,
+      "loss": 0.6037,
+      "mean_token_accuracy": 0.8085700690746307,
+      "step": 500
+    },
+    {
+      "epoch": 0.2423224568138196,
+      "grad_norm": 1.1301806240169505,
+      "learning_rate": 4.0335463258785946e-05,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8150687873363495,
+      "step": 505
+    },
+    {
+      "epoch": 0.24472168905950095,
+      "grad_norm": 0.7765974643502626,
+      "learning_rate": 4.0734824281150165e-05,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.811828863620758,
+      "step": 510
+    },
+    {
+      "epoch": 0.24712092130518235,
+      "grad_norm": 1.3548826892769203,
+      "learning_rate": 4.113418530351438e-05,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8138845384120941,
+      "step": 515
+    },
+    {
+      "epoch": 0.2495201535508637,
+      "grad_norm": 1.1163949052296775,
+      "learning_rate": 4.15335463258786e-05,
+      "loss": 0.5879,
+      "mean_token_accuracy": 0.81446653008461,
+      "step": 520
+    },
+    {
+      "epoch": 0.2519193857965451,
+      "grad_norm": 0.9714485046081821,
+      "learning_rate": 4.193290734824281e-05,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.8142540156841278,
+      "step": 525
+    },
+    {
+      "epoch": 0.2543186180422265,
+      "grad_norm": 0.7557409936221269,
+      "learning_rate": 4.233226837060703e-05,
+      "loss": 0.5851,
+      "mean_token_accuracy": 0.8139719665050507,
+      "step": 530
+    },
+    {
+      "epoch": 0.2567178502879079,
+      "grad_norm": 0.7496854833800273,
+      "learning_rate": 4.273162939297125e-05,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.8175175011157989,
+      "step": 535
+    },
+    {
+      "epoch": 0.2591170825335892,
+      "grad_norm": 0.7888537580262366,
+      "learning_rate": 4.313099041533547e-05,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8156923830509186,
+      "step": 540
+    },
+    {
+      "epoch": 0.2615163147792706,
+      "grad_norm": 0.785622281172855,
+      "learning_rate": 4.3530351437699686e-05,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8164088368415833,
+      "step": 545
+    },
+    {
+      "epoch": 0.263915547024952,
+      "grad_norm": 0.852186691849217,
+      "learning_rate": 4.39297124600639e-05,
+      "loss": 0.6022,
+      "mean_token_accuracy": 0.8093033075332642,
+      "step": 550
+    },
+    {
+      "epoch": 0.2663147792706334,
+      "grad_norm": 0.7784512087734883,
+      "learning_rate": 4.432907348242812e-05,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8160103797912598,
+      "step": 555
+    },
+    {
+      "epoch": 0.2687140115163148,
+      "grad_norm": 0.8320565617519823,
+      "learning_rate": 4.472843450479233e-05,
+      "loss": 0.5939,
+      "mean_token_accuracy": 0.8122792363166809,
+      "step": 560
+    },
+    {
+      "epoch": 0.27111324376199614,
+      "grad_norm": 0.8877223112366365,
+      "learning_rate": 4.512779552715655e-05,
+      "loss": 0.594,
+      "mean_token_accuracy": 0.8112381875514985,
+      "step": 565
+    },
+    {
+      "epoch": 0.27351247600767753,
+      "grad_norm": 0.8868733480541511,
+      "learning_rate": 4.552715654952077e-05,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8121663808822632,
+      "step": 570
+    },
+    {
+      "epoch": 0.2759117082533589,
+      "grad_norm": 1.2963706262591022,
+      "learning_rate": 4.592651757188499e-05,
+      "loss": 0.6031,
+      "mean_token_accuracy": 0.8092474520206452,
+      "step": 575
+    },
+    {
+      "epoch": 0.2783109404990403,
+      "grad_norm": 1.310894898713945,
+      "learning_rate": 4.632587859424921e-05,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8162494540214539,
+      "step": 580
+    },
+    {
+      "epoch": 0.2807101727447217,
+      "grad_norm": 0.8988385867516254,
+      "learning_rate": 4.672523961661342e-05,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8153525233268738,
+      "step": 585
+    },
+    {
+      "epoch": 0.28310940499040305,
+      "grad_norm": 0.751447563005443,
+      "learning_rate": 4.712460063897764e-05,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.816132253408432,
+      "step": 590
+    },
+    {
+      "epoch": 0.28550863723608444,
+      "grad_norm": 0.6807754316677679,
+      "learning_rate": 4.752396166134185e-05,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8204606473445892,
+      "step": 595
+    },
+    {
+      "epoch": 0.28790786948176583,
+      "grad_norm": 0.8236607801744058,
+      "learning_rate": 4.792332268370607e-05,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8139038801193237,
+      "step": 600
+    },
+    {
+      "epoch": 0.2903071017274472,
+      "grad_norm": 1.0126301458862197,
+      "learning_rate": 4.832268370607029e-05,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8181136310100555,
+      "step": 605
+    },
+    {
+      "epoch": 0.2927063339731286,
+      "grad_norm": 0.8019914579553031,
+      "learning_rate": 4.872204472843451e-05,
+      "loss": 0.5542,
+      "mean_token_accuracy": 0.8230095028877258,
+      "step": 610
+    },
+    {
+      "epoch": 0.29510556621880996,
+      "grad_norm": 0.7062463323534514,
+      "learning_rate": 4.912140575079873e-05,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8183253645896912,
+      "step": 615
+    },
+    {
+      "epoch": 0.29750479846449135,
+      "grad_norm": 0.6855738394736139,
+      "learning_rate": 4.952076677316294e-05,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8156778931617736,
+      "step": 620
+    },
+    {
+      "epoch": 0.29990403071017274,
+      "grad_norm": 0.6936388947928334,
+      "learning_rate": 4.992012779552716e-05,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8175960659980774,
+      "step": 625
+    },
+    {
+      "epoch": 0.30230326295585414,
+      "grad_norm": 1.0037748634842334,
+      "learning_rate": 4.996445076430857e-05,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.815348619222641,
+      "step": 630
+    },
+    {
+      "epoch": 0.30470249520153553,
+      "grad_norm": 6.244854046134436,
+      "learning_rate": 4.992001421969428e-05,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8144954383373261,
+      "step": 635
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 0.8969783520250975,
+      "learning_rate": 4.9875577675079985e-05,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8159963071346283,
+      "step": 640
+    },
+    {
+      "epoch": 0.30950095969289826,
+      "grad_norm": 0.7734920429461225,
+      "learning_rate": 4.9831141130465697e-05,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8177722036838532,
+      "step": 645
+    },
+    {
+      "epoch": 0.31190019193857965,
+      "grad_norm": 0.7142126577980644,
+      "learning_rate": 4.978670458585141e-05,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8191527545452117,
+      "step": 650
+    },
+    {
+      "epoch": 0.31429942418426104,
+      "grad_norm": 0.7382008749131507,
+      "learning_rate": 4.9742268041237114e-05,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8193980932235718,
+      "step": 655
+    },
+    {
+      "epoch": 0.31669865642994244,
+      "grad_norm": 0.7169900376811519,
+      "learning_rate": 4.9697831496622825e-05,
+      "loss": 0.5543,
+      "mean_token_accuracy": 0.8232217609882355,
+      "step": 660
+    },
+    {
+      "epoch": 0.3190978886756238,
+      "grad_norm": 0.694035172942871,
+      "learning_rate": 4.965339495200854e-05,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8184644281864166,
+      "step": 665
+    },
+    {
+      "epoch": 0.32149712092130517,
+      "grad_norm": 0.6463126718358011,
+      "learning_rate": 4.960895840739425e-05,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8177339375019074,
+      "step": 670
+    },
+    {
+      "epoch": 0.32389635316698656,
+      "grad_norm": 0.829236370161891,
+      "learning_rate": 4.956452186277995e-05,
+      "loss": 0.5516,
+      "mean_token_accuracy": 0.823012375831604,
+      "step": 675
+    },
+    {
+      "epoch": 0.32629558541266795,
+      "grad_norm": 0.7806704010520713,
+      "learning_rate": 4.952008531816566e-05,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8154712855815888,
+      "step": 680
+    },
+    {
+      "epoch": 0.32869481765834935,
+      "grad_norm": 0.6873888024143939,
+      "learning_rate": 4.947564877355137e-05,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8201543867588044,
+      "step": 685
+    },
+    {
+      "epoch": 0.3310940499040307,
+      "grad_norm": 1.2696445158394363,
+      "learning_rate": 4.9431212228937076e-05,
+      "loss": 0.5527,
+      "mean_token_accuracy": 0.8234491348266602,
+      "step": 690
+    },
+    {
+      "epoch": 0.3334932821497121,
+      "grad_norm": 0.8417323314066051,
+      "learning_rate": 4.938677568432279e-05,
+      "loss": 0.5681,
+      "mean_token_accuracy": 0.81938636302948,
+      "step": 695
+    },
+    {
+      "epoch": 0.33589251439539347,
+      "grad_norm": 1.03191767907223,
+      "learning_rate": 4.93423391397085e-05,
+      "loss": 0.5659,
+      "mean_token_accuracy": 0.8202950179576873,
+      "step": 700
+    },
+    {
+      "epoch": 0.33829174664107486,
+      "grad_norm": 0.838214828399018,
+      "learning_rate": 4.929790259509421e-05,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8157517671585083,
+      "step": 705
+    },
+    {
+      "epoch": 0.34069097888675626,
+      "grad_norm": 0.9470259432449729,
+      "learning_rate": 4.925346605047992e-05,
+      "loss": 0.5572,
+      "mean_token_accuracy": 0.8220482587814331,
+      "step": 710
+    },
+    {
+      "epoch": 0.3430902111324376,
+      "grad_norm": 0.7616814758368874,
+      "learning_rate": 4.920902950586563e-05,
+      "loss": 0.5518,
+      "mean_token_accuracy": 0.8239187180995942,
+      "step": 715
+    },
+    {
+      "epoch": 0.345489443378119,
+      "grad_norm": 0.6928452392521073,
+      "learning_rate": 4.916459296125134e-05,
+      "loss": 0.5623,
+      "mean_token_accuracy": 0.8204801857471467,
+      "step": 720
+    },
+    {
+      "epoch": 0.3478886756238004,
+      "grad_norm": 0.5932890667190485,
+      "learning_rate": 4.912015641663704e-05,
+      "loss": 0.5549,
+      "mean_token_accuracy": 0.8232076942920685,
+      "step": 725
+    },
+    {
+      "epoch": 0.3502879078694818,
+      "grad_norm": 1.4632469671190833,
+      "learning_rate": 4.907571987202275e-05,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8213146209716797,
+      "step": 730
+    },
+    {
+      "epoch": 0.35268714011516317,
+      "grad_norm": 0.6762729658771497,
+      "learning_rate": 4.903128332740846e-05,
+      "loss": 0.555,
+      "mean_token_accuracy": 0.822946411371231,
+      "step": 735
+    },
+    {
+      "epoch": 0.3550863723608445,
+      "grad_norm": 0.6390018861279814,
+      "learning_rate": 4.8986846782794174e-05,
+      "loss": 0.5636,
+      "mean_token_accuracy": 0.8206958174705505,
+      "step": 740
+    },
+    {
+      "epoch": 0.3574856046065259,
+      "grad_norm": 0.6032245697331051,
+      "learning_rate": 4.894241023817988e-05,
+      "loss": 0.555,
+      "mean_token_accuracy": 0.8216525256633759,
+      "step": 745
+    },
+    {
+      "epoch": 0.3598848368522073,
+      "grad_norm": 0.6617068176509739,
+      "learning_rate": 4.889797369356559e-05,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8194816887378693,
+      "step": 750
+    },
+    {
+      "epoch": 0.3622840690978887,
+      "grad_norm": 0.5676852742206375,
+      "learning_rate": 4.88535371489513e-05,
+      "loss": 0.5605,
+      "mean_token_accuracy": 0.8203934669494629,
+      "step": 755
+    },
+    {
+      "epoch": 0.3646833013435701,
+      "grad_norm": 0.6372895506366388,
+      "learning_rate": 4.880910060433701e-05,
+      "loss": 0.5567,
+      "mean_token_accuracy": 0.8218388557434082,
+      "step": 760
+    },
+    {
+      "epoch": 0.3670825335892514,
+      "grad_norm": 0.6855330039826973,
+      "learning_rate": 4.876466405972272e-05,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8184073507785797,
+      "step": 765
+    },
+    {
+      "epoch": 0.3694817658349328,
+      "grad_norm": 0.6063969401795531,
+      "learning_rate": 4.872022751510843e-05,
+      "loss": 0.5608,
+      "mean_token_accuracy": 0.8208466172218323,
+      "step": 770
+    },
+    {
+      "epoch": 0.3718809980806142,
+      "grad_norm": 0.6318300543152913,
+      "learning_rate": 4.867579097049414e-05,
+      "loss": 0.5792,
+      "mean_token_accuracy": 0.8148290395736695,
+      "step": 775
+    },
+    {
+      "epoch": 0.3742802303262956,
+      "grad_norm": 0.6240652015135113,
+      "learning_rate": 4.863135442587984e-05,
+      "loss": 0.5573,
+      "mean_token_accuracy": 0.8221084117889405,
+      "step": 780
+    },
+    {
+      "epoch": 0.376679462571977,
+      "grad_norm": 0.6072419176079188,
+      "learning_rate": 4.8586917881265554e-05,
+      "loss": 0.5602,
+      "mean_token_accuracy": 0.820676177740097,
+      "step": 785
+    },
+    {
+      "epoch": 0.3790786948176583,
+      "grad_norm": 0.6916151373942763,
+      "learning_rate": 4.8542481336651266e-05,
+      "loss": 0.5618,
+      "mean_token_accuracy": 0.8195629417896271,
+      "step": 790
+    },
+    {
+      "epoch": 0.3814779270633397,
+      "grad_norm": 0.7558352618674623,
+      "learning_rate": 4.849804479203697e-05,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8190373599529266,
+      "step": 795
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 0.8372896189519592,
+      "learning_rate": 4.845360824742268e-05,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8195176303386689,
+      "step": 800
+    },
+    {
+      "epoch": 0.3862763915547025,
+      "grad_norm": 0.6238576830422211,
+      "learning_rate": 4.8409171702808395e-05,
+      "loss": 0.5574,
+      "mean_token_accuracy": 0.8217509090900421,
+      "step": 805
+    },
+    {
+      "epoch": 0.3886756238003839,
+      "grad_norm": 0.6698872279230852,
+      "learning_rate": 4.83647351581941e-05,
+      "loss": 0.5836,
+      "mean_token_accuracy": 0.8135609984397888,
+      "step": 810
+    },
+    {
+      "epoch": 0.39107485604606523,
+      "grad_norm": 0.6386814692049486,
+      "learning_rate": 4.832029861357981e-05,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8212653994560242,
+      "step": 815
+    },
+    {
+      "epoch": 0.3934740882917466,
+      "grad_norm": 0.75029553320145,
+      "learning_rate": 4.827586206896552e-05,
+      "loss": 0.5454,
+      "mean_token_accuracy": 0.8249085962772369,
+      "step": 820
+    },
+    {
+      "epoch": 0.395873320537428,
+      "grad_norm": 0.6082151973850165,
+      "learning_rate": 4.823142552435123e-05,
+      "loss": 0.5487,
+      "mean_token_accuracy": 0.8236030340194702,
+      "step": 825
+    },
+    {
+      "epoch": 0.3982725527831094,
+      "grad_norm": 0.5915943123457897,
+      "learning_rate": 4.8186988979736933e-05,
+      "loss": 0.552,
+      "mean_token_accuracy": 0.8226123511791229,
+      "step": 830
+    },
+    {
+      "epoch": 0.4006717850287908,
+      "grad_norm": 0.6071974690638059,
+      "learning_rate": 4.8142552435122645e-05,
+      "loss": 0.5601,
+      "mean_token_accuracy": 0.8202442348003387,
+      "step": 835
+    },
+    {
+      "epoch": 0.40307101727447214,
+      "grad_norm": 0.6125811367551213,
+      "learning_rate": 4.809811589050836e-05,
+      "loss": 0.5532,
+      "mean_token_accuracy": 0.8221201241016388,
+      "step": 840
+    },
+    {
+      "epoch": 0.40547024952015354,
+      "grad_norm": 0.6682594766455511,
+      "learning_rate": 4.805367934589406e-05,
+      "loss": 0.5399,
+      "mean_token_accuracy": 0.8263704001903533,
+      "step": 845
+    },
+    {
+      "epoch": 0.40786948176583493,
+      "grad_norm": 0.735519043559457,
+      "learning_rate": 4.8009242801279774e-05,
+      "loss": 0.548,
+      "mean_token_accuracy": 0.8243632435798645,
+      "step": 850
+    },
+    {
+      "epoch": 0.4102687140115163,
+      "grad_norm": 0.5423593788691745,
+      "learning_rate": 4.7964806256665486e-05,
+      "loss": 0.5551,
+      "mean_token_accuracy": 0.8216099441051483,
+      "step": 855
+    },
+    {
+      "epoch": 0.4126679462571977,
+      "grad_norm": 0.569128145619009,
+      "learning_rate": 4.792036971205119e-05,
+      "loss": 0.5374,
+      "mean_token_accuracy": 0.8269579291343689,
+      "step": 860
+    },
+    {
+      "epoch": 0.41506717850287905,
+      "grad_norm": 0.5621821735679401,
+      "learning_rate": 4.78759331674369e-05,
+      "loss": 0.5266,
+      "mean_token_accuracy": 0.8298987448215485,
+      "step": 865
+    },
+    {
+      "epoch": 0.41746641074856045,
+      "grad_norm": 0.5731699475882329,
+      "learning_rate": 4.7831496622822615e-05,
+      "loss": 0.5407,
+      "mean_token_accuracy": 0.8258492708206177,
+      "step": 870
+    },
+    {
+      "epoch": 0.41986564299424184,
+      "grad_norm": 0.5662106727038226,
+      "learning_rate": 4.7787060078208327e-05,
+      "loss": 0.5382,
+      "mean_token_accuracy": 0.8260211527347565,
+      "step": 875
+    },
+    {
+      "epoch": 0.42226487523992323,
+      "grad_norm": 0.5906976504172934,
+      "learning_rate": 4.7742623533594025e-05,
+      "loss": 0.5585,
+      "mean_token_accuracy": 0.8203856468200683,
+      "step": 880
+    },
+    {
+      "epoch": 0.4246641074856046,
+      "grad_norm": 0.5719381852725446,
+      "learning_rate": 4.769818698897974e-05,
+      "loss": 0.5298,
+      "mean_token_accuracy": 0.8297291934490204,
+      "step": 885
+    },
+    {
+      "epoch": 0.42706333973128596,
+      "grad_norm": 0.6104627055672889,
+      "learning_rate": 4.765375044436545e-05,
+      "loss": 0.5574,
+      "mean_token_accuracy": 0.8215497851371765,
+      "step": 890
+    },
+    {
+      "epoch": 0.42946257197696736,
+      "grad_norm": 0.5106967506472315,
+      "learning_rate": 4.7609313899751154e-05,
+      "loss": 0.5505,
+      "mean_token_accuracy": 0.8234421014785767,
+      "step": 895
+    },
+    {
+      "epoch": 0.43186180422264875,
+      "grad_norm": 0.5486180289895729,
+      "learning_rate": 4.7564877355136865e-05,
+      "loss": 0.5279,
+      "mean_token_accuracy": 0.8291941523551941,
+      "step": 900
+    },
+    {
+      "epoch": 0.43426103646833014,
+      "grad_norm": 0.8110944973636702,
+      "learning_rate": 4.752044081052258e-05,
+      "loss": 0.5438,
+      "mean_token_accuracy": 0.824740606546402,
+      "step": 905
+    },
+    {
+      "epoch": 0.43666026871401153,
+      "grad_norm": 0.5083722030175633,
+      "learning_rate": 4.747600426590829e-05,
+      "loss": 0.5371,
+      "mean_token_accuracy": 0.8267485558986664,
+      "step": 910
+    },
+    {
+      "epoch": 0.43905950095969287,
+      "grad_norm": 0.5450657681596905,
+      "learning_rate": 4.7431567721293994e-05,
+      "loss": 0.5331,
+      "mean_token_accuracy": 0.8280423700809478,
+      "step": 915
+    },
+    {
+      "epoch": 0.44145873320537427,
+      "grad_norm": 0.5587272673659367,
+      "learning_rate": 4.7387131176679706e-05,
+      "loss": 0.533,
+      "mean_token_accuracy": 0.8275462627410889,
+      "step": 920
+    },
+    {
+      "epoch": 0.44385796545105566,
+      "grad_norm": 0.597994173280171,
+      "learning_rate": 4.734269463206542e-05,
+      "loss": 0.5221,
+      "mean_token_accuracy": 0.8312480449676514,
+      "step": 925
+    },
+    {
+      "epoch": 0.44625719769673705,
+      "grad_norm": 0.5699731767132227,
+      "learning_rate": 4.7298258087451116e-05,
+      "loss": 0.5417,
+      "mean_token_accuracy": 0.8253937721252441,
+      "step": 930
+    },
+    {
+      "epoch": 0.44865642994241844,
+      "grad_norm": 0.530459601961475,
+      "learning_rate": 4.725382154283683e-05,
+      "loss": 0.5411,
+      "mean_token_accuracy": 0.8255383253097535,
+      "step": 935
+    },
+    {
+      "epoch": 0.4510556621880998,
+      "grad_norm": 0.5786015168216376,
+      "learning_rate": 4.720938499822254e-05,
+      "loss": 0.5367,
+      "mean_token_accuracy": 0.8274415552616119,
+      "step": 940
+    },
+    {
+      "epoch": 0.4534548944337812,
+      "grad_norm": 0.5352156988298586,
+      "learning_rate": 4.716494845360825e-05,
+      "loss": 0.5486,
+      "mean_token_accuracy": 0.8225701510906219,
+      "step": 945
+    },
+    {
+      "epoch": 0.45585412667946257,
+      "grad_norm": 0.5537066343867931,
+      "learning_rate": 4.712051190899396e-05,
+      "loss": 0.5332,
+      "mean_token_accuracy": 0.8273627698421478,
+      "step": 950
+    },
+    {
+      "epoch": 0.45825335892514396,
+      "grad_norm": 0.5496688141801068,
+      "learning_rate": 4.707607536437967e-05,
+      "loss": 0.5505,
+      "mean_token_accuracy": 0.8225115597248077,
+      "step": 955
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 0.5291584157449768,
+      "learning_rate": 4.703163881976538e-05,
+      "loss": 0.5284,
+      "mean_token_accuracy": 0.829184639453888,
+      "step": 960
+    },
+    {
+      "epoch": 0.4630518234165067,
+      "grad_norm": 0.6097079936773455,
+      "learning_rate": 4.6987202275151086e-05,
+      "loss": 0.5293,
+      "mean_token_accuracy": 0.8293408989906311,
+      "step": 965
+    },
+    {
+      "epoch": 0.4654510556621881,
+      "grad_norm": 0.6198873182908643,
+      "learning_rate": 4.69427657305368e-05,
+      "loss": 0.523,
+      "mean_token_accuracy": 0.8304136216640472,
+      "step": 970
+    },
+    {
+      "epoch": 0.4678502879078695,
+      "grad_norm": 0.5891656862471465,
+      "learning_rate": 4.689832918592251e-05,
+      "loss": 0.549,
+      "mean_token_accuracy": 0.8228592276573181,
+      "step": 975
+    },
+    {
+      "epoch": 0.47024952015355087,
+      "grad_norm": 0.6107839108650862,
+      "learning_rate": 4.6853892641308214e-05,
+      "loss": 0.5365,
+      "mean_token_accuracy": 0.8271978914737701,
+      "step": 980
+    },
+    {
+      "epoch": 0.47264875239923226,
+      "grad_norm": 0.571716205851,
+      "learning_rate": 4.680945609669392e-05,
+      "loss": 0.5394,
+      "mean_token_accuracy": 0.8253172099590301,
+      "step": 985
+    },
+    {
+      "epoch": 0.4750479846449136,
+      "grad_norm": 0.5542308753871588,
+      "learning_rate": 4.676501955207963e-05,
+      "loss": 0.5207,
+      "mean_token_accuracy": 0.8310206890106201,
+      "step": 990
+    },
+    {
+      "epoch": 0.477447216890595,
+      "grad_norm": 0.5483955302881889,
+      "learning_rate": 4.672058300746534e-05,
+      "loss": 0.5275,
+      "mean_token_accuracy": 0.8294080853462219,
+      "step": 995
+    },
+    {
+      "epoch": 0.4798464491362764,
+      "grad_norm": 0.580528097435676,
+      "learning_rate": 4.667614646285105e-05,
+      "loss": 0.5326,
+      "mean_token_accuracy": 0.8276298522949219,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4822456813819578,
+      "grad_norm": 0.5350717699048666,
+      "learning_rate": 4.663170991823676e-05,
+      "loss": 0.5315,
+      "mean_token_accuracy": 0.8278722941875458,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4846449136276392,
+      "grad_norm": 0.49218288607628896,
+      "learning_rate": 4.658727337362247e-05,
+      "loss": 0.5328,
+      "mean_token_accuracy": 0.8278626799583435,
+      "step": 1010
+    },
+    {
+      "epoch": 0.4870441458733205,
+      "grad_norm": 0.5134720273160304,
+      "learning_rate": 4.654283682900818e-05,
+      "loss": 0.524,
+      "mean_token_accuracy": 0.830355030298233,
+      "step": 1015
+    },
+    {
+      "epoch": 0.4894433781190019,
+      "grad_norm": 0.5251369142411848,
+      "learning_rate": 4.649840028439389e-05,
+      "loss": 0.5397,
+      "mean_token_accuracy": 0.8257737815380096,
+      "step": 1020
+    },
+    {
+      "epoch": 0.4918426103646833,
+      "grad_norm": 0.5157247220630689,
+      "learning_rate": 4.64539637397796e-05,
+      "loss": 0.5237,
+      "mean_token_accuracy": 0.8299675941467285,
+      "step": 1025
+    },
+    {
+      "epoch": 0.4942418426103647,
+      "grad_norm": 0.4859001572426086,
+      "learning_rate": 4.6409527195165306e-05,
+      "loss": 0.5277,
+      "mean_token_accuracy": 0.8294627845287323,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4966410748560461,
+      "grad_norm": 0.561175024557525,
+      "learning_rate": 4.636509065055101e-05,
+      "loss": 0.5455,
+      "mean_token_accuracy": 0.8240741550922394,
+      "step": 1035
+    },
+    {
+      "epoch": 0.4990403071017274,
+      "grad_norm": 0.5338562625947265,
+      "learning_rate": 4.632065410593672e-05,
+      "loss": 0.5374,
+      "mean_token_accuracy": 0.8271634161472321,
+      "step": 1040
+    },
+    {
+      "epoch": 0.5014395393474088,
+      "grad_norm": 0.515601873079243,
+      "learning_rate": 4.6276217561322435e-05,
+      "loss": 0.5345,
+      "mean_token_accuracy": 0.8272102892398834,
+      "step": 1045
+    },
+    {
+      "epoch": 0.5038387715930902,
+      "grad_norm": 0.6177024352288669,
+      "learning_rate": 4.623178101670814e-05,
+      "loss": 0.5333,
+      "mean_token_accuracy": 0.8278423607349396,
+      "step": 1050
+    },
+    {
+      "epoch": 0.5062380038387716,
+      "grad_norm": 0.5512310415321986,
+      "learning_rate": 4.618734447209385e-05,
+      "loss": 0.5242,
+      "mean_token_accuracy": 0.8300135910511017,
+      "step": 1055
+    },
+    {
+      "epoch": 0.508637236084453,
+      "grad_norm": 0.5663859189128105,
+      "learning_rate": 4.6142907927479563e-05,
+      "loss": 0.5154,
+      "mean_token_accuracy": 0.8328012764453888,
+      "step": 1060
+    },
+    {
+      "epoch": 0.5110364683301344,
+      "grad_norm": 0.5773949952752142,
+      "learning_rate": 4.609847138286527e-05,
+      "loss": 0.5349,
+      "mean_token_accuracy": 0.8270993530750275,
+      "step": 1065
+    },
+    {
+      "epoch": 0.5134357005758158,
+      "grad_norm": 0.6280922601470035,
+      "learning_rate": 4.605403483825098e-05,
+      "loss": 0.5348,
+      "mean_token_accuracy": 0.8269883990287781,
+      "step": 1070
+    },
+    {
+      "epoch": 0.5158349328214972,
+      "grad_norm": 0.5990994330045243,
+      "learning_rate": 4.600959829363669e-05,
+      "loss": 0.5293,
+      "mean_token_accuracy": 0.8284970939159393,
+      "step": 1075
+    },
+    {
+      "epoch": 0.5182341650671785,
+      "grad_norm": 0.5656451607094958,
+      "learning_rate": 4.5965161749022404e-05,
+      "loss": 0.5277,
+      "mean_token_accuracy": 0.8288775861263276,
+      "step": 1080
+    },
+    {
+      "epoch": 0.5206333973128598,
+      "grad_norm": 0.5397614018675521,
+      "learning_rate": 4.59207252044081e-05,
+      "loss": 0.5449,
+      "mean_token_accuracy": 0.8244773209095001,
+      "step": 1085
+    },
+    {
+      "epoch": 0.5230326295585412,
+      "grad_norm": 0.5669284914862057,
+      "learning_rate": 4.5876288659793814e-05,
+      "loss": 0.5148,
+      "mean_token_accuracy": 0.8320637166500091,
+      "step": 1090
+    },
+    {
+      "epoch": 0.5254318618042226,
+      "grad_norm": 0.5378169685314852,
+      "learning_rate": 4.5831852115179526e-05,
+      "loss": 0.5312,
+      "mean_token_accuracy": 0.8280978500843048,
+      "step": 1095
+    },
+    {
+      "epoch": 0.527831094049904,
+      "grad_norm": 0.4785055105443453,
+      "learning_rate": 4.578741557056523e-05,
+      "loss": 0.511,
+      "mean_token_accuracy": 0.8339739978313446,
+      "step": 1100
+    },
+    {
+      "epoch": 0.5302303262955854,
+      "grad_norm": 0.5033536918525533,
+      "learning_rate": 4.574297902595094e-05,
+      "loss": 0.5343,
+      "mean_token_accuracy": 0.8264274299144745,
+      "step": 1105
+    },
+    {
+      "epoch": 0.5326295585412668,
+      "grad_norm": 0.5617636032573375,
+      "learning_rate": 4.5698542481336655e-05,
+      "loss": 0.5251,
+      "mean_token_accuracy": 0.8295752882957459,
+      "step": 1110
+    },
+    {
+      "epoch": 0.5350287907869482,
+      "grad_norm": 0.47649118768675885,
+      "learning_rate": 4.565410593672237e-05,
+      "loss": 0.5224,
+      "mean_token_accuracy": 0.8308883607387543,
+      "step": 1115
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 0.5332406727760631,
+      "learning_rate": 4.560966939210807e-05,
+      "loss": 0.5321,
+      "mean_token_accuracy": 0.8277345418930053,
+      "step": 1120
+    },
+    {
+      "epoch": 0.539827255278311,
+      "grad_norm": 0.4782827735405752,
+      "learning_rate": 4.5565232847493784e-05,
+      "loss": 0.5268,
+      "mean_token_accuracy": 0.8292799592018127,
+      "step": 1125
+    },
+    {
+      "epoch": 0.5422264875239923,
+      "grad_norm": 0.5375370012352939,
+      "learning_rate": 4.5520796302879495e-05,
+      "loss": 0.5422,
+      "mean_token_accuracy": 0.8255633115768433,
+      "step": 1130
+    },
+    {
+      "epoch": 0.5446257197696737,
+      "grad_norm": 0.693752706192889,
+      "learning_rate": 4.5476359758265194e-05,
+      "loss": 0.5298,
+      "mean_token_accuracy": 0.8276696979999543,
+      "step": 1135
+    },
+    {
+      "epoch": 0.5470249520153551,
+      "grad_norm": 0.6665627567388126,
+      "learning_rate": 4.5431923213650906e-05,
+      "loss": 0.5266,
+      "mean_token_accuracy": 0.8296854555606842,
+      "step": 1140
+    },
+    {
+      "epoch": 0.5494241842610365,
+      "grad_norm": 0.5755677817276045,
+      "learning_rate": 4.538748666903662e-05,
+      "loss": 0.5178,
+      "mean_token_accuracy": 0.831487900018692,
+      "step": 1145
+    },
+    {
+      "epoch": 0.5518234165067178,
+      "grad_norm": 0.5462744381863546,
+      "learning_rate": 4.534305012442233e-05,
+      "loss": 0.5204,
+      "mean_token_accuracy": 0.8311738312244416,
+      "step": 1150
+    },
+    {
+      "epoch": 0.5542226487523992,
+      "grad_norm": 0.6500856819910776,
+      "learning_rate": 4.5298613579808034e-05,
+      "loss": 0.5245,
+      "mean_token_accuracy": 0.8298864960670471,
+      "step": 1155
+    },
+    {
+      "epoch": 0.5566218809980806,
+      "grad_norm": 0.6007744834984243,
+      "learning_rate": 4.5254177035193746e-05,
+      "loss": 0.5308,
+      "mean_token_accuracy": 0.8276986002922058,
+      "step": 1160
+    },
+    {
+      "epoch": 0.559021113243762,
+      "grad_norm": 0.4925324081774109,
+      "learning_rate": 4.520974049057946e-05,
+      "loss": 0.5206,
+      "mean_token_accuracy": 0.8303386211395264,
+      "step": 1165
+    },
+    {
+      "epoch": 0.5614203454894434,
+      "grad_norm": 0.5198373434963387,
+      "learning_rate": 4.516530394596516e-05,
+      "loss": 0.5214,
+      "mean_token_accuracy": 0.8302409529685975,
+      "step": 1170
+    },
+    {
+      "epoch": 0.5638195777351248,
+      "grad_norm": 0.5220759377556011,
+      "learning_rate": 4.5120867401350875e-05,
+      "loss": 0.5184,
+      "mean_token_accuracy": 0.8315722942352295,
+      "step": 1175
+    },
+    {
+      "epoch": 0.5662188099808061,
+      "grad_norm": 0.5114196157128489,
+      "learning_rate": 4.507643085673659e-05,
+      "loss": 0.5341,
+      "mean_token_accuracy": 0.8262094497680664,
+      "step": 1180
+    },
+    {
+      "epoch": 0.5686180422264875,
+      "grad_norm": 0.5068021513263434,
+      "learning_rate": 4.503199431212229e-05,
+      "loss": 0.5187,
+      "mean_token_accuracy": 0.8312457025051116,
+      "step": 1185
+    },
+    {
+      "epoch": 0.5710172744721689,
+      "grad_norm": 0.5613227478451318,
+      "learning_rate": 4.4987557767508e-05,
+      "loss": 0.5057,
+      "mean_token_accuracy": 0.8353295505046845,
+      "step": 1190
+    },
+    {
+      "epoch": 0.5734165067178503,
+      "grad_norm": 0.8284201861643404,
+      "learning_rate": 4.494312122289371e-05,
+      "loss": 0.5133,
+      "mean_token_accuracy": 0.8334067642688752,
+      "step": 1195
+    },
+    {
+      "epoch": 0.5758157389635317,
+      "grad_norm": 0.9210564168780322,
+      "learning_rate": 4.489868467827942e-05,
+      "loss": 0.5195,
+      "mean_token_accuracy": 0.8312300026416779,
+      "step": 1200
+    },
+    {
+      "epoch": 0.5782149712092131,
+      "grad_norm": 0.5997534401025089,
+      "learning_rate": 4.4854248133665126e-05,
+      "loss": 0.532,
+      "mean_token_accuracy": 0.8277329862117767,
+      "step": 1205
+    },
+    {
+      "epoch": 0.5806142034548945,
+      "grad_norm": 0.509102977352861,
+      "learning_rate": 4.480981158905084e-05,
+      "loss": 0.5229,
+      "mean_token_accuracy": 0.8301851987838745,
+      "step": 1210
+    },
+    {
+      "epoch": 0.5830134357005758,
+      "grad_norm": 0.5232298511871786,
+      "learning_rate": 4.476537504443655e-05,
+      "loss": 0.5115,
+      "mean_token_accuracy": 0.8335091173648834,
+      "step": 1215
+    },
+    {
+      "epoch": 0.5854126679462572,
+      "grad_norm": 1.4520334905055199,
+      "learning_rate": 4.4720938499822255e-05,
+      "loss": 0.5266,
+      "mean_token_accuracy": 0.8287900865077973,
+      "step": 1220
+    },
+    {
+      "epoch": 0.5878119001919386,
+      "grad_norm": 0.6774712243421297,
+      "learning_rate": 4.4676501955207966e-05,
+      "loss": 0.5423,
+      "mean_token_accuracy": 0.8255094051361084,
+      "step": 1225
+    },
+    {
+      "epoch": 0.5902111324376199,
+      "grad_norm": 1.5754303835971644,
+      "learning_rate": 4.463206541059368e-05,
+      "loss": 0.5188,
+      "mean_token_accuracy": 0.8306097269058228,
+      "step": 1230
+    },
+    {
+      "epoch": 0.5926103646833013,
+      "grad_norm": 0.7313851331688538,
+      "learning_rate": 4.458762886597938e-05,
+      "loss": 0.5361,
+      "mean_token_accuracy": 0.8274235844612121,
+      "step": 1235
+    },
+    {
+      "epoch": 0.5950095969289827,
+      "grad_norm": 0.7811366827057676,
+      "learning_rate": 4.454319232136509e-05,
+      "loss": 0.5376,
+      "mean_token_accuracy": 0.8265086829662323,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5974088291746641,
+      "grad_norm": 0.5852967202784879,
+      "learning_rate": 4.44987557767508e-05,
+      "loss": 0.5222,
+      "mean_token_accuracy": 0.829976099729538,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5998080614203455,
+      "grad_norm": 0.6500504682698087,
+      "learning_rate": 4.445431923213651e-05,
+      "loss": 0.524,
+      "mean_token_accuracy": 0.8303073704242706,
+      "step": 1250
+    },
+    {
+      "epoch": 0.6022072936660269,
+      "grad_norm": 0.5183436136927764,
+      "learning_rate": 4.440988268752222e-05,
+      "loss": 0.5219,
+      "mean_token_accuracy": 0.8303581476211548,
+      "step": 1255
+    },
+    {
+      "epoch": 0.6046065259117083,
+      "grad_norm": 1.1492046855836786,
+      "learning_rate": 4.436544614290793e-05,
+      "loss": 0.5332,
+      "mean_token_accuracy": 0.8281478404998779,
+      "step": 1260
+    },
+    {
+      "epoch": 0.6070057581573897,
+      "grad_norm": 0.752249891324009,
+      "learning_rate": 4.432100959829364e-05,
+      "loss": 0.512,
+      "mean_token_accuracy": 0.8334548771381378,
+      "step": 1265
+    },
+    {
+      "epoch": 0.6094049904030711,
+      "grad_norm": 0.5544631019636599,
+      "learning_rate": 4.4276573053679346e-05,
+      "loss": 0.5163,
+      "mean_token_accuracy": 0.8321223258972168,
+      "step": 1270
+    },
+    {
+      "epoch": 0.6118042226487524,
+      "grad_norm": 0.5153914235770953,
+      "learning_rate": 4.423213650906506e-05,
+      "loss": 0.5296,
+      "mean_token_accuracy": 0.8289190053939819,
+      "step": 1275
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 0.47633132901358466,
+      "learning_rate": 4.418769996445077e-05,
+      "loss": 0.5277,
+      "mean_token_accuracy": 0.8282978534698486,
+      "step": 1280
+    },
+    {
+      "epoch": 0.6166026871401151,
+      "grad_norm": 0.4932196916279109,
+      "learning_rate": 4.4143263419836475e-05,
+      "loss": 0.5213,
+      "mean_token_accuracy": 0.8302229821681977,
+      "step": 1285
+    },
+    {
+      "epoch": 0.6190019193857965,
+      "grad_norm": 0.5011121697532454,
+      "learning_rate": 4.409882687522218e-05,
+      "loss": 0.5225,
+      "mean_token_accuracy": 0.8293979346752167,
+      "step": 1290
+    },
+    {
+      "epoch": 0.6214011516314779,
+      "grad_norm": 0.4856904264094217,
+      "learning_rate": 4.405439033060789e-05,
+      "loss": 0.5219,
+      "mean_token_accuracy": 0.829535436630249,
+      "step": 1295
+    },
+    {
+      "epoch": 0.6238003838771593,
+      "grad_norm": 0.4820149931102137,
+      "learning_rate": 4.4009953785993604e-05,
+      "loss": 0.5138,
+      "mean_token_accuracy": 0.8327163219451904,
+      "step": 1300
+    },
+    {
+      "epoch": 0.6261996161228407,
+      "grad_norm": 0.4624881398808915,
+      "learning_rate": 4.396551724137931e-05,
+      "loss": 0.5157,
+      "mean_token_accuracy": 0.8317691743373871,
+      "step": 1305
+    },
+    {
+      "epoch": 0.6285988483685221,
+      "grad_norm": 0.5095347564346104,
+      "learning_rate": 4.392108069676502e-05,
+      "loss": 0.5158,
+      "mean_token_accuracy": 0.8313847720623017,
+      "step": 1310
+    },
+    {
+      "epoch": 0.6309980806142035,
+      "grad_norm": 0.44685463760378097,
+      "learning_rate": 4.387664415215073e-05,
+      "loss": 0.5273,
+      "mean_token_accuracy": 0.8286812126636505,
+      "step": 1315
+    },
+    {
+      "epoch": 0.6333973128598849,
+      "grad_norm": 0.4641733447876531,
+      "learning_rate": 4.383220760753644e-05,
+      "loss": 0.5146,
+      "mean_token_accuracy": 0.8313398063182831,
+      "step": 1320
+    },
+    {
+      "epoch": 0.6357965451055663,
+      "grad_norm": 0.4766102041605914,
+      "learning_rate": 4.378777106292215e-05,
+      "loss": 0.5109,
+      "mean_token_accuracy": 0.8354115903377533,
+      "step": 1325
+    },
+    {
+      "epoch": 0.6381957773512476,
+      "grad_norm": 0.546100262545878,
+      "learning_rate": 4.374333451830786e-05,
+      "loss": 0.5353,
+      "mean_token_accuracy": 0.8268633961677552,
+      "step": 1330
+    },
+    {
+      "epoch": 0.6405950095969289,
+      "grad_norm": 0.5356075705213353,
+      "learning_rate": 4.369889797369357e-05,
+      "loss": 0.5139,
+      "mean_token_accuracy": 0.8328723609447479,
+      "step": 1335
+    },
+    {
+      "epoch": 0.6429942418426103,
+      "grad_norm": 0.5350381311815007,
+      "learning_rate": 4.365446142907927e-05,
+      "loss": 0.5182,
+      "mean_token_accuracy": 0.8312582015991211,
+      "step": 1340
+    },
+    {
+      "epoch": 0.6453934740882917,
+      "grad_norm": 0.5202240202485383,
+      "learning_rate": 4.361002488446498e-05,
+      "loss": 0.5056,
+      "mean_token_accuracy": 0.8346873342990875,
+      "step": 1345
+    },
+    {
+      "epoch": 0.6477927063339731,
+      "grad_norm": 0.5784323915265774,
+      "learning_rate": 4.3565588339850695e-05,
+      "loss": 0.529,
+      "mean_token_accuracy": 0.8280220568180084,
+      "step": 1350
+    },
+    {
+      "epoch": 0.6501919385796545,
+      "grad_norm": 0.4835689781428135,
+      "learning_rate": 4.352115179523641e-05,
+      "loss": 0.5231,
+      "mean_token_accuracy": 0.8293167948722839,
+      "step": 1355
+    },
+    {
+      "epoch": 0.6525911708253359,
+      "grad_norm": 0.5757605413585991,
+      "learning_rate": 4.347671525062211e-05,
+      "loss": 0.5334,
+      "mean_token_accuracy": 0.8265172779560089,
+      "step": 1360
+    },
+    {
+      "epoch": 0.6549904030710173,
+      "grad_norm": 0.507486093941236,
+      "learning_rate": 4.3432278706007824e-05,
+      "loss": 0.5206,
+      "mean_token_accuracy": 0.8299190640449524,
+      "step": 1365
+    },
+    {
+      "epoch": 0.6573896353166987,
+      "grad_norm": 0.5249598517466133,
+      "learning_rate": 4.3387842161393536e-05,
+      "loss": 0.5195,
+      "mean_token_accuracy": 0.830981433391571,
+      "step": 1370
+    },
+    {
+      "epoch": 0.6597888675623801,
+      "grad_norm": 0.48631807389153725,
+      "learning_rate": 4.334340561677924e-05,
+      "loss": 0.5228,
+      "mean_token_accuracy": 0.829991239309311,
+      "step": 1375
+    },
+    {
+      "epoch": 0.6621880998080614,
+      "grad_norm": 0.486725909348472,
+      "learning_rate": 4.329896907216495e-05,
+      "loss": 0.5112,
+      "mean_token_accuracy": 0.8330169022083282,
+      "step": 1380
+    },
+    {
+      "epoch": 0.6645873320537428,
+      "grad_norm": 0.47067469083752866,
+      "learning_rate": 4.3254532527550664e-05,
+      "loss": 0.5139,
+      "mean_token_accuracy": 0.8317618012428284,
+      "step": 1385
+    },
+    {
+      "epoch": 0.6669865642994242,
+      "grad_norm": 0.4707320443459022,
+      "learning_rate": 4.321009598293637e-05,
+      "loss": 0.5033,
+      "mean_token_accuracy": 0.8353248655796051,
+      "step": 1390
+    },
+    {
+      "epoch": 0.6693857965451055,
+      "grad_norm": 0.5177653250480144,
+      "learning_rate": 4.3165659438322074e-05,
+      "loss": 0.5094,
+      "mean_token_accuracy": 0.8332755088806152,
+      "step": 1395
+    },
+    {
+      "epoch": 0.6717850287907869,
+      "grad_norm": 0.5351120723588554,
+      "learning_rate": 4.3121222893707786e-05,
+      "loss": 0.5226,
+      "mean_token_accuracy": 0.8294448137283326,
+      "step": 1400
+    },
+    {
+      "epoch": 0.6741842610364683,
+      "grad_norm": 0.4765578169939615,
+      "learning_rate": 4.30767863490935e-05,
+      "loss": 0.5044,
+      "mean_token_accuracy": 0.83488889336586,
+      "step": 1405
+    },
+    {
+      "epoch": 0.6765834932821497,
+      "grad_norm": 0.5188267403257218,
+      "learning_rate": 4.30323498044792e-05,
+      "loss": 0.5025,
+      "mean_token_accuracy": 0.8355053424835205,
+      "step": 1410
+    },
+    {
+      "epoch": 0.6789827255278311,
+      "grad_norm": 0.5033840380744201,
+      "learning_rate": 4.2987913259864915e-05,
+      "loss": 0.5019,
+      "mean_token_accuracy": 0.8350803136825562,
+      "step": 1415
+    },
+    {
+      "epoch": 0.6813819577735125,
+      "grad_norm": 0.5037590855559116,
+      "learning_rate": 4.294347671525063e-05,
+      "loss": 0.4861,
+      "mean_token_accuracy": 0.8396735787391663,
+      "step": 1420
+    },
+    {
+      "epoch": 0.6837811900191939,
+      "grad_norm": 0.4530621438574399,
+      "learning_rate": 4.289904017063633e-05,
+      "loss": 0.5233,
+      "mean_token_accuracy": 0.8285799026489258,
+      "step": 1425
+    },
+    {
+      "epoch": 0.6861804222648752,
+      "grad_norm": 0.4312748853048293,
+      "learning_rate": 4.2854603626022044e-05,
+      "loss": 0.5029,
+      "mean_token_accuracy": 0.8345647633075715,
+      "step": 1430
+    },
+    {
+      "epoch": 0.6885796545105566,
+      "grad_norm": 0.4273932940130073,
+      "learning_rate": 4.2810167081407756e-05,
+      "loss": 0.4906,
+      "mean_token_accuracy": 0.8378296196460724,
+      "step": 1435
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 0.44519718560433147,
+      "learning_rate": 4.276573053679346e-05,
+      "loss": 0.5146,
+      "mean_token_accuracy": 0.8315480768680572,
+      "step": 1440
+    },
+    {
+      "epoch": 0.6933781190019194,
+      "grad_norm": 2.050838607632764,
+      "learning_rate": 4.2721293992179166e-05,
+      "loss": 0.4954,
+      "mean_token_accuracy": 0.8377414226531983,
+      "step": 1445
+    },
+    {
+      "epoch": 0.6957773512476008,
+      "grad_norm": 0.5012731719663341,
+      "learning_rate": 4.267685744756488e-05,
+      "loss": 0.4955,
+      "mean_token_accuracy": 0.8370757579803467,
+      "step": 1450
+    },
+    {
+      "epoch": 0.6981765834932822,
+      "grad_norm": 0.46356735369736507,
+      "learning_rate": 4.263242090295059e-05,
+      "loss": 0.4939,
+      "mean_token_accuracy": 0.8377406358718872,
+      "step": 1455
+    },
+    {
+      "epoch": 0.7005758157389635,
+      "grad_norm": 0.4325835028161503,
+      "learning_rate": 4.2587984358336295e-05,
+      "loss": 0.5148,
+      "mean_token_accuracy": 0.8308480203151702,
+      "step": 1460
+    },
+    {
+      "epoch": 0.7029750479846449,
+      "grad_norm": 0.4595501874101042,
+      "learning_rate": 4.2543547813722007e-05,
+      "loss": 0.5,
+      "mean_token_accuracy": 0.835820984840393,
+      "step": 1465
+    },
+    {
+      "epoch": 0.7053742802303263,
+      "grad_norm": 0.4822053236469392,
+      "learning_rate": 4.249911126910772e-05,
+      "loss": 0.503,
+      "mean_token_accuracy": 0.83474200963974,
+      "step": 1470
+    },
+    {
+      "epoch": 0.7077735124760077,
+      "grad_norm": 0.47568643360076,
+      "learning_rate": 4.2454674724493423e-05,
+      "loss": 0.5097,
+      "mean_token_accuracy": 0.8334442794322967,
+      "step": 1475
+    },
+    {
+      "epoch": 0.710172744721689,
+      "grad_norm": 0.45809054187912474,
+      "learning_rate": 4.2410238179879135e-05,
+      "loss": 0.4993,
+      "mean_token_accuracy": 0.835784274339676,
+      "step": 1480
+    },
+    {
+      "epoch": 0.7125719769673704,
+      "grad_norm": 0.5960561618915279,
+      "learning_rate": 4.236580163526485e-05,
+      "loss": 0.5135,
+      "mean_token_accuracy": 0.8313300907611847,
+      "step": 1485
+    },
+    {
+      "epoch": 0.7149712092130518,
+      "grad_norm": 0.47511929182323426,
+      "learning_rate": 4.232136509065055e-05,
+      "loss": 0.5057,
+      "mean_token_accuracy": 0.8338849306106567,
+      "step": 1490
+    },
+    {
+      "epoch": 0.7173704414587332,
+      "grad_norm": 0.4763020223745529,
+      "learning_rate": 4.227692854603626e-05,
+      "loss": 0.4954,
+      "mean_token_accuracy": 0.837703138589859,
+      "step": 1495
+    },
+    {
+      "epoch": 0.7197696737044146,
+      "grad_norm": 0.4929704143972219,
+      "learning_rate": 4.223249200142197e-05,
+      "loss": 0.5148,
+      "mean_token_accuracy": 0.8312839865684509,
+      "step": 1500
+    },
+    {
+      "epoch": 0.722168905950096,
+      "grad_norm": 0.4777453194471662,
+      "learning_rate": 4.218805545680768e-05,
+      "loss": 0.5004,
+      "mean_token_accuracy": 0.8349537491798401,
+      "step": 1505
+    },
+    {
+      "epoch": 0.7245681381957774,
+      "grad_norm": 0.4990271169279016,
+      "learning_rate": 4.2143618912193386e-05,
+      "loss": 0.4995,
+      "mean_token_accuracy": 0.835792076587677,
+      "step": 1510
+    },
+    {
+      "epoch": 0.7269673704414588,
+      "grad_norm": 0.45068189297866923,
+      "learning_rate": 4.20991823675791e-05,
+      "loss": 0.4993,
+      "mean_token_accuracy": 0.8355522215366363,
+      "step": 1515
+    },
+    {
+      "epoch": 0.7293666026871402,
+      "grad_norm": 0.5488621493176185,
+      "learning_rate": 4.205474582296481e-05,
+      "loss": 0.5034,
+      "mean_token_accuracy": 0.8344599723815918,
+      "step": 1520
+    },
+    {
+      "epoch": 0.7317658349328215,
+      "grad_norm": 0.49110092403567085,
+      "learning_rate": 4.2010309278350515e-05,
+      "loss": 0.5049,
+      "mean_token_accuracy": 0.8338232159614563,
+      "step": 1525
+    },
+    {
+      "epoch": 0.7341650671785028,
+      "grad_norm": 0.44819619071534816,
+      "learning_rate": 4.196587273373623e-05,
+      "loss": 0.4989,
+      "mean_token_accuracy": 0.8356217563152313,
+      "step": 1530
+    },
+    {
+      "epoch": 0.7365642994241842,
+      "grad_norm": 0.4978704155039036,
+      "learning_rate": 4.192143618912194e-05,
+      "loss": 0.5041,
+      "mean_token_accuracy": 0.8346357643604279,
+      "step": 1535
+    },
+    {
+      "epoch": 0.7389635316698656,
+      "grad_norm": 0.4955027083886941,
+      "learning_rate": 4.187699964450765e-05,
+      "loss": 0.5015,
+      "mean_token_accuracy": 0.8347662448883056,
+      "step": 1540
+    },
+    {
+      "epoch": 0.741362763915547,
+      "grad_norm": 0.9777169453073922,
+      "learning_rate": 4.183256309989335e-05,
+      "loss": 0.5107,
+      "mean_token_accuracy": 0.832754397392273,
+      "step": 1545
+    },
+    {
+      "epoch": 0.7437619961612284,
+      "grad_norm": 0.5239229227325649,
+      "learning_rate": 4.178812655527906e-05,
+      "loss": 0.5038,
+      "mean_token_accuracy": 0.8346226036548614,
+      "step": 1550
+    },
+    {
+      "epoch": 0.7461612284069098,
+      "grad_norm": 0.49596215598318977,
+      "learning_rate": 4.174369001066477e-05,
+      "loss": 0.5026,
+      "mean_token_accuracy": 0.8348248362541199,
+      "step": 1555
+    },
+    {
+      "epoch": 0.7485604606525912,
+      "grad_norm": 0.48117516776556213,
+      "learning_rate": 4.169925346605048e-05,
+      "loss": 0.4875,
+      "mean_token_accuracy": 0.8402361094951629,
+      "step": 1560
+    },
+    {
+      "epoch": 0.7509596928982726,
+      "grad_norm": 0.4438969952921841,
+      "learning_rate": 4.165481692143619e-05,
+      "loss": 0.4977,
+      "mean_token_accuracy": 0.8357623934745788,
+      "step": 1565
+    },
+    {
+      "epoch": 0.753358925143954,
+      "grad_norm": 0.521505618398822,
+      "learning_rate": 4.16103803768219e-05,
+      "loss": 0.4887,
+      "mean_token_accuracy": 0.8391876101493836,
+      "step": 1570
+    },
+    {
+      "epoch": 0.7557581573896354,
+      "grad_norm": 0.4693636851519488,
+      "learning_rate": 4.156594383220761e-05,
+      "loss": 0.5063,
+      "mean_token_accuracy": 0.8331591010093689,
+      "step": 1575
+    },
+    {
+      "epoch": 0.7581573896353166,
+      "grad_norm": 0.496197058535834,
+      "learning_rate": 4.152150728759332e-05,
+      "loss": 0.5126,
+      "mean_token_accuracy": 0.8317332327365875,
+      "step": 1580
+    },
+    {
+      "epoch": 0.760556621880998,
+      "grad_norm": 0.6505659912029434,
+      "learning_rate": 4.147707074297903e-05,
+      "loss": 0.4984,
+      "mean_token_accuracy": 0.8353803277015686,
+      "step": 1585
+    },
+    {
+      "epoch": 0.7629558541266794,
+      "grad_norm": 0.47473838918744676,
+      "learning_rate": 4.143263419836474e-05,
+      "loss": 0.5106,
+      "mean_token_accuracy": 0.8329426884651184,
+      "step": 1590
+    },
+    {
+      "epoch": 0.7653550863723608,
+      "grad_norm": 0.5803134205385464,
+      "learning_rate": 4.138819765375045e-05,
+      "loss": 0.5169,
+      "mean_token_accuracy": 0.8308128654956818,
+      "step": 1595
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 0.5614686683573403,
+      "learning_rate": 4.134376110913615e-05,
+      "loss": 0.5109,
+      "mean_token_accuracy": 0.83311847448349,
+      "step": 1600
+    },
+    {
+      "epoch": 0.7701535508637236,
+      "grad_norm": 0.7065207576015031,
+      "learning_rate": 4.1299324564521864e-05,
+      "loss": 0.4999,
+      "mean_token_accuracy": 0.8362725853919983,
+      "step": 1605
+    },
+    {
+      "epoch": 0.772552783109405,
+      "grad_norm": 0.560451963494759,
+      "learning_rate": 4.1254888019907576e-05,
+      "loss": 0.4998,
+      "mean_token_accuracy": 0.8343302607536316,
+      "step": 1610
+    },
+    {
+      "epoch": 0.7749520153550864,
+      "grad_norm": 0.5244654481078436,
+      "learning_rate": 4.121045147529328e-05,
+      "loss": 0.5132,
+      "mean_token_accuracy": 0.8325715720653534,
+      "step": 1615
+    },
+    {
+      "epoch": 0.7773512476007678,
+      "grad_norm": 0.46348761758989515,
+      "learning_rate": 4.116601493067899e-05,
+      "loss": 0.5092,
+      "mean_token_accuracy": 0.8330723762512207,
+      "step": 1620
+    },
+    {
+      "epoch": 0.7797504798464492,
+      "grad_norm": 0.4481507057971629,
+      "learning_rate": 4.1121578386064704e-05,
+      "loss": 0.5002,
+      "mean_token_accuracy": 0.8361100614070892,
+      "step": 1625
+    },
+    {
+      "epoch": 0.7821497120921305,
+      "grad_norm": 0.581248195098889,
+      "learning_rate": 4.107714184145041e-05,
+      "loss": 0.4929,
+      "mean_token_accuracy": 0.8380164384841919,
+      "step": 1630
+    },
+    {
+      "epoch": 0.7845489443378119,
+      "grad_norm": 0.520687770629966,
+      "learning_rate": 4.103270529683612e-05,
+      "loss": 0.5056,
+      "mean_token_accuracy": 0.8334766566753388,
+      "step": 1635
+    },
+    {
+      "epoch": 0.7869481765834933,
+      "grad_norm": 0.49427636553199883,
+      "learning_rate": 4.098826875222183e-05,
+      "loss": 0.4806,
+      "mean_token_accuracy": 0.8409277200698853,
+      "step": 1640
+    },
+    {
+      "epoch": 0.7893474088291746,
+      "grad_norm": 0.4746622492122426,
+      "learning_rate": 4.094383220760754e-05,
+      "loss": 0.4978,
+      "mean_token_accuracy": 0.836303836107254,
+      "step": 1645
+    },
+    {
+      "epoch": 0.791746641074856,
+      "grad_norm": 0.45998552131958137,
+      "learning_rate": 4.089939566299324e-05,
+      "loss": 0.5055,
+      "mean_token_accuracy": 0.8346006035804748,
+      "step": 1650
+    },
+    {
+      "epoch": 0.7941458733205374,
+      "grad_norm": 0.7306011013931711,
+      "learning_rate": 4.0854959118378955e-05,
+      "loss": 0.4914,
+      "mean_token_accuracy": 0.8384609997272492,
+      "step": 1655
+    },
+    {
+      "epoch": 0.7965451055662188,
+      "grad_norm": 0.48801343160132227,
+      "learning_rate": 4.081052257376467e-05,
+      "loss": 0.5002,
+      "mean_token_accuracy": 0.8357686400413513,
+      "step": 1660
+    },
+    {
+      "epoch": 0.7989443378119002,
+      "grad_norm": 0.46142435250279384,
+      "learning_rate": 4.076608602915037e-05,
+      "loss": 0.5068,
+      "mean_token_accuracy": 0.8335560023784637,
+      "step": 1665
+    },
+    {
+      "epoch": 0.8013435700575816,
+      "grad_norm": 0.4953618366906397,
+      "learning_rate": 4.0721649484536084e-05,
+      "loss": 0.4981,
+      "mean_token_accuracy": 0.8363311767578125,
+      "step": 1670
+    },
+    {
+      "epoch": 0.803742802303263,
+      "grad_norm": 0.4470979902731658,
+      "learning_rate": 4.0677212939921796e-05,
+      "loss": 0.5018,
+      "mean_token_accuracy": 0.8343466818332672,
+      "step": 1675
+    },
+    {
+      "epoch": 0.8061420345489443,
+      "grad_norm": 0.5257607137210768,
+      "learning_rate": 4.06327763953075e-05,
+      "loss": 0.5272,
+      "mean_token_accuracy": 0.8283361494541168,
+      "step": 1680
+    },
+    {
+      "epoch": 0.8085412667946257,
+      "grad_norm": 0.5449197937622,
+      "learning_rate": 4.058833985069321e-05,
+      "loss": 0.5027,
+      "mean_token_accuracy": 0.8345248162746429,
+      "step": 1685
+    },
+    {
+      "epoch": 0.8109404990403071,
+      "grad_norm": 0.5272292325409751,
+      "learning_rate": 4.0543903306078925e-05,
+      "loss": 0.5003,
+      "mean_token_accuracy": 0.8353365778923034,
+      "step": 1690
+    },
+    {
+      "epoch": 0.8133397312859885,
+      "grad_norm": 0.49483975937460833,
+      "learning_rate": 4.049946676146463e-05,
+      "loss": 0.48,
+      "mean_token_accuracy": 0.841797149181366,
+      "step": 1695
+    },
+    {
+      "epoch": 0.8157389635316699,
+      "grad_norm": 0.455696435370605,
+      "learning_rate": 4.0455030216850335e-05,
+      "loss": 0.5031,
+      "mean_token_accuracy": 0.8344521641731262,
+      "step": 1700
+    },
+    {
+      "epoch": 0.8181381957773513,
+      "grad_norm": 0.809778240421179,
+      "learning_rate": 4.041059367223605e-05,
+      "loss": 0.4793,
+      "mean_token_accuracy": 0.8415903806686401,
+      "step": 1705
+    },
+    {
+      "epoch": 0.8205374280230326,
+      "grad_norm": 0.4581620367543629,
+      "learning_rate": 4.036615712762176e-05,
+      "loss": 0.5031,
+      "mean_token_accuracy": 0.8346677839756012,
+      "step": 1710
+    },
+    {
+      "epoch": 0.822936660268714,
+      "grad_norm": 0.4793760548775191,
+      "learning_rate": 4.0321720583007464e-05,
+      "loss": 0.5082,
+      "mean_token_accuracy": 0.8328176736831665,
+      "step": 1715
+    },
+    {
+      "epoch": 0.8253358925143954,
+      "grad_norm": 0.7488626062500724,
+      "learning_rate": 4.0277284038393175e-05,
+      "loss": 0.4992,
+      "mean_token_accuracy": 0.8357600927352905,
+      "step": 1720
+    },
+    {
+      "epoch": 0.8277351247600768,
+      "grad_norm": 0.4417556948390778,
+      "learning_rate": 4.023284749377889e-05,
+      "loss": 0.4785,
+      "mean_token_accuracy": 0.8417211413383484,
+      "step": 1725
+    },
+    {
+      "epoch": 0.8301343570057581,
+      "grad_norm": 0.4778636985339465,
+      "learning_rate": 4.018841094916459e-05,
+      "loss": 0.4986,
+      "mean_token_accuracy": 0.8356509983539582,
+      "step": 1730
+    },
+    {
+      "epoch": 0.8325335892514395,
+      "grad_norm": 0.41886367691680565,
+      "learning_rate": 4.0143974404550304e-05,
+      "loss": 0.4959,
+      "mean_token_accuracy": 0.8363343060016633,
+      "step": 1735
+    },
+    {
+      "epoch": 0.8349328214971209,
+      "grad_norm": 0.46794847168479464,
+      "learning_rate": 4.0099537859936016e-05,
+      "loss": 0.4882,
+      "mean_token_accuracy": 0.8390391588211059,
+      "step": 1740
+    },
+    {
+      "epoch": 0.8373320537428023,
+      "grad_norm": 0.45094719812416384,
+      "learning_rate": 4.005510131532173e-05,
+      "loss": 0.5056,
+      "mean_token_accuracy": 0.8326801776885986,
+      "step": 1745
+    },
+    {
+      "epoch": 0.8397312859884837,
+      "grad_norm": 0.4410114321911972,
+      "learning_rate": 4.001066477070743e-05,
+      "loss": 0.4915,
+      "mean_token_accuracy": 0.8374046802520752,
+      "step": 1750
+    },
+    {
+      "epoch": 0.8421305182341651,
+      "grad_norm": 0.5013043600759126,
+      "learning_rate": 3.996622822609314e-05,
+      "loss": 0.4987,
+      "mean_token_accuracy": 0.8353951811790467,
+      "step": 1755
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 0.43861309242180074,
+      "learning_rate": 3.992179168147885e-05,
+      "loss": 0.4899,
+      "mean_token_accuracy": 0.8380984723567962,
+      "step": 1760
+    },
+    {
+      "epoch": 0.8469289827255279,
+      "grad_norm": 0.41336946937588115,
+      "learning_rate": 3.9877355136864555e-05,
+      "loss": 0.4924,
+      "mean_token_accuracy": 0.8372499346733093,
+      "step": 1765
+    },
+    {
+      "epoch": 0.8493282149712092,
+      "grad_norm": 0.4545097883190453,
+      "learning_rate": 3.983291859225027e-05,
+      "loss": 0.5168,
+      "mean_token_accuracy": 0.8302761137485504,
+      "step": 1770
+    },
+    {
+      "epoch": 0.8517274472168906,
+      "grad_norm": 0.474966739957993,
+      "learning_rate": 3.978848204763598e-05,
+      "loss": 0.494,
+      "mean_token_accuracy": 0.8385133445262909,
+      "step": 1775
+    },
+    {
+      "epoch": 0.8541266794625719,
+      "grad_norm": 0.5865771257587197,
+      "learning_rate": 3.974404550302169e-05,
+      "loss": 0.5173,
+      "mean_token_accuracy": 0.8318293452262878,
+      "step": 1780
+    },
+    {
+      "epoch": 0.8565259117082533,
+      "grad_norm": 0.5147748010011185,
+      "learning_rate": 3.9699608958407396e-05,
+      "loss": 0.4901,
+      "mean_token_accuracy": 0.8380867540836334,
+      "step": 1785
+    },
+    {
+      "epoch": 0.8589251439539347,
+      "grad_norm": 0.6332747120400611,
+      "learning_rate": 3.965517241379311e-05,
+      "loss": 0.4877,
+      "mean_token_accuracy": 0.8399276256561279,
+      "step": 1790
+    },
+    {
+      "epoch": 0.8613243761996161,
+      "grad_norm": 0.47831786443131424,
+      "learning_rate": 3.961073586917882e-05,
+      "loss": 0.5068,
+      "mean_token_accuracy": 0.8339995861053466,
+      "step": 1795
+    },
+    {
+      "epoch": 0.8637236084452975,
+      "grad_norm": 0.4832626111095525,
+      "learning_rate": 3.9566299324564524e-05,
+      "loss": 0.4936,
+      "mean_token_accuracy": 0.8372804462909699,
+      "step": 1800
+    },
+    {
+      "epoch": 0.8661228406909789,
+      "grad_norm": 0.5261836705483144,
+      "learning_rate": 3.952186277995023e-05,
+      "loss": 0.4769,
+      "mean_token_accuracy": 0.8424659311771393,
+      "step": 1805
+    },
+    {
+      "epoch": 0.8685220729366603,
+      "grad_norm": 0.451251646000052,
+      "learning_rate": 3.947742623533594e-05,
+      "loss": 0.4892,
+      "mean_token_accuracy": 0.8388195514678956,
+      "step": 1810
+    },
+    {
+      "epoch": 0.8709213051823417,
+      "grad_norm": 0.48207059123147333,
+      "learning_rate": 3.943298969072165e-05,
+      "loss": 0.4983,
+      "mean_token_accuracy": 0.8356702029705048,
+      "step": 1815
+    },
+    {
+      "epoch": 0.8733205374280231,
+      "grad_norm": 0.47858523546672777,
+      "learning_rate": 3.938855314610736e-05,
+      "loss": 0.4831,
+      "mean_token_accuracy": 0.8398571729660034,
+      "step": 1820
+    },
+    {
+      "epoch": 0.8757197696737045,
+      "grad_norm": 0.4855780035679636,
+      "learning_rate": 3.934411660149307e-05,
+      "loss": 0.5054,
+      "mean_token_accuracy": 0.8337648630142211,
+      "step": 1825
+    },
+    {
+      "epoch": 0.8781190019193857,
+      "grad_norm": 0.4670757101333078,
+      "learning_rate": 3.929968005687878e-05,
+      "loss": 0.4824,
+      "mean_token_accuracy": 0.8406095683574677,
+      "step": 1830
+    },
+    {
+      "epoch": 0.8805182341650671,
+      "grad_norm": 0.4549267712242883,
+      "learning_rate": 3.925524351226449e-05,
+      "loss": 0.4921,
+      "mean_token_accuracy": 0.8366944909095764,
+      "step": 1835
+    },
+    {
+      "epoch": 0.8829174664107485,
+      "grad_norm": 0.45048453890973705,
+      "learning_rate": 3.92108069676502e-05,
+      "loss": 0.5092,
+      "mean_token_accuracy": 0.8323357820510864,
+      "step": 1840
+    },
+    {
+      "epoch": 0.8853166986564299,
+      "grad_norm": 0.4362188334466366,
+      "learning_rate": 3.916637042303591e-05,
+      "loss": 0.4818,
+      "mean_token_accuracy": 0.8405416011810303,
+      "step": 1845
+    },
+    {
+      "epoch": 0.8877159309021113,
+      "grad_norm": 0.4609804738877885,
+      "learning_rate": 3.9121933878421616e-05,
+      "loss": 0.4833,
+      "mean_token_accuracy": 0.8405681133270264,
+      "step": 1850
+    },
+    {
+      "epoch": 0.8901151631477927,
+      "grad_norm": 0.45727787424378313,
+      "learning_rate": 3.907749733380732e-05,
+      "loss": 0.4806,
+      "mean_token_accuracy": 0.8407939493656158,
+      "step": 1855
+    },
+    {
+      "epoch": 0.8925143953934741,
+      "grad_norm": 0.4517352190937965,
+      "learning_rate": 3.903306078919303e-05,
+      "loss": 0.4856,
+      "mean_token_accuracy": 0.8397642016410828,
+      "step": 1860
+    },
+    {
+      "epoch": 0.8949136276391555,
+      "grad_norm": 0.5410580134801548,
+      "learning_rate": 3.8988624244578745e-05,
+      "loss": 0.5012,
+      "mean_token_accuracy": 0.835526442527771,
+      "step": 1865
+    },
+    {
+      "epoch": 0.8973128598848369,
+      "grad_norm": 0.4255275614609763,
+      "learning_rate": 3.894418769996445e-05,
+      "loss": 0.4882,
+      "mean_token_accuracy": 0.8384945869445801,
+      "step": 1870
+    },
+    {
+      "epoch": 0.8997120921305183,
+      "grad_norm": 0.5016674928378082,
+      "learning_rate": 3.889975115535016e-05,
+      "loss": 0.4864,
+      "mean_token_accuracy": 0.8388539910316467,
+      "step": 1875
+    },
+    {
+      "epoch": 0.9021113243761996,
+      "grad_norm": 0.47185204205793313,
+      "learning_rate": 3.885531461073587e-05,
+      "loss": 0.5008,
+      "mean_token_accuracy": 0.834422481060028,
+      "step": 1880
+    },
+    {
+      "epoch": 0.904510556621881,
+      "grad_norm": 0.44149478527936975,
+      "learning_rate": 3.881087806612158e-05,
+      "loss": 0.4747,
+      "mean_token_accuracy": 0.8427581369876862,
+      "step": 1885
+    },
+    {
+      "epoch": 0.9069097888675623,
+      "grad_norm": 0.42788888304582523,
+      "learning_rate": 3.876644152150729e-05,
+      "loss": 0.5003,
+      "mean_token_accuracy": 0.835177195072174,
+      "step": 1890
+    },
+    {
+      "epoch": 0.9093090211132437,
+      "grad_norm": 0.44941832370477924,
+      "learning_rate": 3.8722004976893e-05,
+      "loss": 0.4783,
+      "mean_token_accuracy": 0.841586971282959,
+      "step": 1895
+    },
+    {
+      "epoch": 0.9117082533589251,
+      "grad_norm": 0.4427516271825846,
+      "learning_rate": 3.867756843227871e-05,
+      "loss": 0.491,
+      "mean_token_accuracy": 0.8373789012432098,
+      "step": 1900
+    },
+    {
+      "epoch": 0.9141074856046065,
+      "grad_norm": 0.4476423549355035,
+      "learning_rate": 3.863313188766441e-05,
+      "loss": 0.4995,
+      "mean_token_accuracy": 0.8350639104843139,
+      "step": 1905
+    },
+    {
+      "epoch": 0.9165067178502879,
+      "grad_norm": 0.4492630820771489,
+      "learning_rate": 3.8588695343050124e-05,
+      "loss": 0.4879,
+      "mean_token_accuracy": 0.8386876225471497,
+      "step": 1910
+    },
+    {
+      "epoch": 0.9189059500959693,
+      "grad_norm": 0.47882494003670634,
+      "learning_rate": 3.8544258798435836e-05,
+      "loss": 0.4833,
+      "mean_token_accuracy": 0.8402657985687256,
+      "step": 1915
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 0.4419811380727961,
+      "learning_rate": 3.849982225382154e-05,
+      "loss": 0.4936,
+      "mean_token_accuracy": 0.8363905549049377,
+      "step": 1920
+    },
+    {
+      "epoch": 0.9237044145873321,
+      "grad_norm": 0.39621957717208156,
+      "learning_rate": 3.845538570920725e-05,
+      "loss": 0.484,
+      "mean_token_accuracy": 0.8395649671554566,
+      "step": 1925
+    },
+    {
+      "epoch": 0.9261036468330134,
+      "grad_norm": 0.4529698155759262,
+      "learning_rate": 3.8410949164592965e-05,
+      "loss": 0.4686,
+      "mean_token_accuracy": 0.84321910738945,
+      "step": 1930
+    },
+    {
+      "epoch": 0.9285028790786948,
+      "grad_norm": 0.40480258139982583,
+      "learning_rate": 3.836651261997867e-05,
+      "loss": 0.4935,
+      "mean_token_accuracy": 0.8369616866111755,
+      "step": 1935
+    },
+    {
+      "epoch": 0.9309021113243762,
+      "grad_norm": 0.49493368299837737,
+      "learning_rate": 3.832207607536438e-05,
+      "loss": 0.4729,
+      "mean_token_accuracy": 0.8431230068206788,
+      "step": 1940
+    },
+    {
+      "epoch": 0.9333013435700576,
+      "grad_norm": 0.5013934672328659,
+      "learning_rate": 3.8277639530750094e-05,
+      "loss": 0.4843,
+      "mean_token_accuracy": 0.8394111692905426,
+      "step": 1945
+    },
+    {
+      "epoch": 0.935700575815739,
+      "grad_norm": 0.4330692332817214,
+      "learning_rate": 3.8233202986135805e-05,
+      "loss": 0.4876,
+      "mean_token_accuracy": 0.838521808385849,
+      "step": 1950
+    },
+    {
+      "epoch": 0.9380998080614203,
+      "grad_norm": 0.4518953171866384,
+      "learning_rate": 3.818876644152151e-05,
+      "loss": 0.4993,
+      "mean_token_accuracy": 0.8351330041885376,
+      "step": 1955
+    },
+    {
+      "epoch": 0.9404990403071017,
+      "grad_norm": 0.46867857792205087,
+      "learning_rate": 3.8144329896907216e-05,
+      "loss": 0.4836,
+      "mean_token_accuracy": 0.8396610736846923,
+      "step": 1960
+    },
+    {
+      "epoch": 0.9428982725527831,
+      "grad_norm": 0.4628778139252986,
+      "learning_rate": 3.809989335229293e-05,
+      "loss": 0.5,
+      "mean_token_accuracy": 0.834876400232315,
+      "step": 1965
+    },
+    {
+      "epoch": 0.9452975047984645,
+      "grad_norm": 0.4600077570542966,
+      "learning_rate": 3.805545680767863e-05,
+      "loss": 0.4935,
+      "mean_token_accuracy": 0.8370054304599762,
+      "step": 1970
+    },
+    {
+      "epoch": 0.9476967370441459,
+      "grad_norm": 0.4993993819155987,
+      "learning_rate": 3.8011020263064344e-05,
+      "loss": 0.4935,
+      "mean_token_accuracy": 0.8372502028942108,
+      "step": 1975
+    },
+    {
+      "epoch": 0.9500959692898272,
+      "grad_norm": 0.4439645062611687,
+      "learning_rate": 3.7966583718450056e-05,
+      "loss": 0.4891,
+      "mean_token_accuracy": 0.8377054810523987,
+      "step": 1980
+    },
+    {
+      "epoch": 0.9524952015355086,
+      "grad_norm": 0.5121282144414162,
+      "learning_rate": 3.792214717383577e-05,
+      "loss": 0.4832,
+      "mean_token_accuracy": 0.8399454653263092,
+      "step": 1985
+    },
+    {
+      "epoch": 0.95489443378119,
+      "grad_norm": 0.49717141863444153,
+      "learning_rate": 3.787771062922147e-05,
+      "loss": 0.4896,
+      "mean_token_accuracy": 0.837871116399765,
+      "step": 1990
+    },
+    {
+      "epoch": 0.9572936660268714,
+      "grad_norm": 0.4544349403730862,
+      "learning_rate": 3.7833274084607185e-05,
+      "loss": 0.5072,
+      "mean_token_accuracy": 0.8325957894325257,
+      "step": 1995
+    },
+    {
+      "epoch": 0.9596928982725528,
+      "grad_norm": 0.42858727516703904,
+      "learning_rate": 3.77888375399929e-05,
+      "loss": 0.4828,
+      "mean_token_accuracy": 0.8397642016410828,
+      "step": 2000
+    },
+    {
+      "epoch": 0.9620921305182342,
+      "grad_norm": 0.40754396210939614,
+      "learning_rate": 3.77444009953786e-05,
+      "loss": 0.4765,
+      "mean_token_accuracy": 0.8419104278087616,
+      "step": 2005
+    },
+    {
+      "epoch": 0.9644913627639156,
+      "grad_norm": 0.44726864462424454,
+      "learning_rate": 3.769996445076431e-05,
+      "loss": 0.4854,
+      "mean_token_accuracy": 0.8394436955451965,
+      "step": 2010
+    },
+    {
+      "epoch": 0.966890595009597,
+      "grad_norm": 0.47404065872867157,
+      "learning_rate": 3.765552790615002e-05,
+      "loss": 0.5063,
+      "mean_token_accuracy": 0.8324910879135132,
+      "step": 2015
+    },
+    {
+      "epoch": 0.9692898272552783,
+      "grad_norm": 0.410702396325773,
+      "learning_rate": 3.761109136153573e-05,
+      "loss": 0.477,
+      "mean_token_accuracy": 0.8416393160820007,
+      "step": 2020
+    },
+    {
+      "epoch": 0.9716890595009597,
+      "grad_norm": 0.4476680516713628,
+      "learning_rate": 3.7566654816921436e-05,
+      "loss": 0.4881,
+      "mean_token_accuracy": 0.8375460922718048,
+      "step": 2025
+    },
+    {
+      "epoch": 0.974088291746641,
+      "grad_norm": 0.44590317336698554,
+      "learning_rate": 3.752221827230715e-05,
+      "loss": 0.4827,
+      "mean_token_accuracy": 0.8395681023597718,
+      "step": 2030
+    },
+    {
+      "epoch": 0.9764875239923224,
+      "grad_norm": 0.5040536414096327,
+      "learning_rate": 3.747778172769286e-05,
+      "loss": 0.4922,
+      "mean_token_accuracy": 0.8366780757904053,
+      "step": 2035
+    },
+    {
+      "epoch": 0.9788867562380038,
+      "grad_norm": 0.40814566317367457,
+      "learning_rate": 3.7433345183078564e-05,
+      "loss": 0.4864,
+      "mean_token_accuracy": 0.8382266163825989,
+      "step": 2040
+    },
+    {
+      "epoch": 0.9812859884836852,
+      "grad_norm": 0.4511394234284313,
+      "learning_rate": 3.7388908638464276e-05,
+      "loss": 0.4893,
+      "mean_token_accuracy": 0.8377293705940246,
+      "step": 2045
+    },
+    {
+      "epoch": 0.9836852207293666,
+      "grad_norm": 0.436955498827051,
+      "learning_rate": 3.734447209384999e-05,
+      "loss": 0.4918,
+      "mean_token_accuracy": 0.8379468977451324,
+      "step": 2050
+    },
+    {
+      "epoch": 0.986084452975048,
+      "grad_norm": 0.5357720057793889,
+      "learning_rate": 3.730003554923569e-05,
+      "loss": 0.491,
+      "mean_token_accuracy": 0.8379468977451324,
+      "step": 2055
+    },
+    {
+      "epoch": 0.9884836852207294,
+      "grad_norm": 0.4217359911795722,
+      "learning_rate": 3.72555990046214e-05,
+      "loss": 0.4761,
+      "mean_token_accuracy": 0.8418299496173859,
+      "step": 2060
+    },
+    {
+      "epoch": 0.9908829174664108,
+      "grad_norm": 0.43096352852270436,
+      "learning_rate": 3.721116246000711e-05,
+      "loss": 0.4885,
+      "mean_token_accuracy": 0.8379851818084717,
+      "step": 2065
+    },
+    {
+      "epoch": 0.9932821497120922,
+      "grad_norm": 0.4297567148467833,
+      "learning_rate": 3.716672591539282e-05,
+      "loss": 0.4914,
+      "mean_token_accuracy": 0.8365148782730103,
+      "step": 2070
+    },
+    {
+      "epoch": 0.9956813819577736,
+      "grad_norm": 0.4804447999108489,
+      "learning_rate": 3.712228937077853e-05,
+      "loss": 0.4936,
+      "mean_token_accuracy": 0.836709177494049,
+      "step": 2075
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 0.4691591099165681,
+      "learning_rate": 3.707785282616424e-05,
+      "loss": 0.4919,
+      "mean_token_accuracy": 0.8369327783584595,
+      "step": 2080
+    },
+    {
+      "epoch": 1.0004798464491362,
+      "grad_norm": 0.5385056566589963,
+      "learning_rate": 3.703341628154995e-05,
+      "loss": 0.4725,
+      "mean_token_accuracy": 0.8429514586925506,
+      "step": 2085
+    },
+    {
+      "epoch": 1.0028790786948176,
+      "grad_norm": 0.5247407907209998,
+      "learning_rate": 3.6988979736935656e-05,
+      "loss": 0.3989,
+      "mean_token_accuracy": 0.8624344766139984,
+      "step": 2090
+    },
+    {
+      "epoch": 1.005278310940499,
+      "grad_norm": 0.48811929992401026,
+      "learning_rate": 3.694454319232137e-05,
+      "loss": 0.4044,
+      "mean_token_accuracy": 0.8603689730167389,
+      "step": 2095
+    },
+    {
+      "epoch": 1.0076775431861804,
+      "grad_norm": 0.4398358878276954,
+      "learning_rate": 3.690010664770708e-05,
+      "loss": 0.4098,
+      "mean_token_accuracy": 0.8592318236827851,
+      "step": 2100
+    },
+    {
+      "epoch": 1.0100767754318618,
+      "grad_norm": 0.5344316993297887,
+      "learning_rate": 3.6855670103092785e-05,
+      "loss": 0.4156,
+      "mean_token_accuracy": 0.8575473427772522,
+      "step": 2105
+    },
+    {
+      "epoch": 1.0124760076775432,
+      "grad_norm": 0.43873766298460454,
+      "learning_rate": 3.681123355847849e-05,
+      "loss": 0.4143,
+      "mean_token_accuracy": 0.8583036422729492,
+      "step": 2110
+    },
+    {
+      "epoch": 1.0148752399232246,
+      "grad_norm": 0.4431906254548146,
+      "learning_rate": 3.67667970138642e-05,
+      "loss": 0.4007,
+      "mean_token_accuracy": 0.8621528685092926,
+      "step": 2115
+    },
+    {
+      "epoch": 1.017274472168906,
+      "grad_norm": 0.46385051341702943,
+      "learning_rate": 3.6722360469249913e-05,
+      "loss": 0.4133,
+      "mean_token_accuracy": 0.8582450449466705,
+      "step": 2120
+    },
+    {
+      "epoch": 1.0196737044145874,
+      "grad_norm": 0.47509224818785306,
+      "learning_rate": 3.667792392463562e-05,
+      "loss": 0.4083,
+      "mean_token_accuracy": 0.8595705032348633,
+      "step": 2125
+    },
+    {
+      "epoch": 1.0220729366602688,
+      "grad_norm": 0.4552889927508356,
+      "learning_rate": 3.663348738002133e-05,
+      "loss": 0.4096,
+      "mean_token_accuracy": 0.8591662049293518,
+      "step": 2130
+    },
+    {
+      "epoch": 1.0244721689059502,
+      "grad_norm": 0.5509774885331505,
+      "learning_rate": 3.658905083540704e-05,
+      "loss": 0.421,
+      "mean_token_accuracy": 0.8578622102737427,
+      "step": 2135
+    },
+    {
+      "epoch": 1.0268714011516316,
+      "grad_norm": 0.4556583809000766,
+      "learning_rate": 3.654461429079275e-05,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8639024317264556,
+      "step": 2140
+    },
+    {
+      "epoch": 1.029270633397313,
+      "grad_norm": 0.5340933536827104,
+      "learning_rate": 3.650017774617846e-05,
+      "loss": 0.4122,
+      "mean_token_accuracy": 0.8583052039146424,
+      "step": 2145
+    },
+    {
+      "epoch": 1.0316698656429943,
+      "grad_norm": 0.42404563232605813,
+      "learning_rate": 3.645574120156417e-05,
+      "loss": 0.4011,
+      "mean_token_accuracy": 0.8615194737911225,
+      "step": 2150
+    },
+    {
+      "epoch": 1.0340690978886755,
+      "grad_norm": 0.5072208926542439,
+      "learning_rate": 3.6411304656949876e-05,
+      "loss": 0.4036,
+      "mean_token_accuracy": 0.8606564939022064,
+      "step": 2155
+    },
+    {
+      "epoch": 1.036468330134357,
+      "grad_norm": 0.4182179906134632,
+      "learning_rate": 3.636686811233559e-05,
+      "loss": 0.4248,
+      "mean_token_accuracy": 0.8546604514122009,
+      "step": 2160
+    },
+    {
+      "epoch": 1.0388675623800383,
+      "grad_norm": 0.44924145858745945,
+      "learning_rate": 3.632243156772129e-05,
+      "loss": 0.4063,
+      "mean_token_accuracy": 0.8609499096870422,
+      "step": 2165
+    },
+    {
+      "epoch": 1.0412667946257197,
+      "grad_norm": 0.48104691609354716,
+      "learning_rate": 3.6277995023107005e-05,
+      "loss": 0.434,
+      "mean_token_accuracy": 0.852498596906662,
+      "step": 2170
+    },
+    {
+      "epoch": 1.043666026871401,
+      "grad_norm": 0.48132504476906385,
+      "learning_rate": 3.623355847849271e-05,
+      "loss": 0.4011,
+      "mean_token_accuracy": 0.8613822042942048,
+      "step": 2175
+    },
+    {
+      "epoch": 1.0460652591170825,
+      "grad_norm": 0.46151017509763026,
+      "learning_rate": 3.618912193387842e-05,
+      "loss": 0.4006,
+      "mean_token_accuracy": 0.8626507818698883,
+      "step": 2180
+    },
+    {
+      "epoch": 1.0484644913627639,
+      "grad_norm": 0.4829126029300642,
+      "learning_rate": 3.6144685389264134e-05,
+      "loss": 0.4085,
+      "mean_token_accuracy": 0.8593409061431885,
+      "step": 2185
+    },
+    {
+      "epoch": 1.0508637236084453,
+      "grad_norm": 0.4719323814072435,
+      "learning_rate": 3.6100248844649845e-05,
+      "loss": 0.4071,
+      "mean_token_accuracy": 0.860792088508606,
+      "step": 2190
+    },
+    {
+      "epoch": 1.0532629558541267,
+      "grad_norm": 0.44618383248608906,
+      "learning_rate": 3.605581230003555e-05,
+      "loss": 0.4071,
+      "mean_token_accuracy": 0.8597249269485474,
+      "step": 2195
+    },
+    {
+      "epoch": 1.055662188099808,
+      "grad_norm": 0.484371830161018,
+      "learning_rate": 3.601137575542126e-05,
+      "loss": 0.4001,
+      "mean_token_accuracy": 0.8620155990123749,
+      "step": 2200
+    },
+    {
+      "epoch": 1.0580614203454894,
+      "grad_norm": 0.4726148631163947,
+      "learning_rate": 3.5966939210806974e-05,
+      "loss": 0.4083,
+      "mean_token_accuracy": 0.8604319036006928,
+      "step": 2205
+    },
+    {
+      "epoch": 1.0604606525911708,
+      "grad_norm": 0.5195452365929821,
+      "learning_rate": 3.592250266619268e-05,
+      "loss": 0.4177,
+      "mean_token_accuracy": 0.8567543268203736,
+      "step": 2210
+    },
+    {
+      "epoch": 1.0628598848368522,
+      "grad_norm": 0.5067336953723536,
+      "learning_rate": 3.5878066121578384e-05,
+      "loss": 0.4055,
+      "mean_token_accuracy": 0.860358464717865,
+      "step": 2215
+    },
+    {
+      "epoch": 1.0652591170825336,
+      "grad_norm": 0.44003137165691186,
+      "learning_rate": 3.5833629576964096e-05,
+      "loss": 0.4041,
+      "mean_token_accuracy": 0.8620176315307617,
+      "step": 2220
+    },
+    {
+      "epoch": 1.067658349328215,
+      "grad_norm": 0.47306018294891,
+      "learning_rate": 3.578919303234981e-05,
+      "loss": 0.4027,
+      "mean_token_accuracy": 0.8611225724220276,
+      "step": 2225
+    },
+    {
+      "epoch": 1.0700575815738964,
+      "grad_norm": 0.43149528893426176,
+      "learning_rate": 3.574475648773551e-05,
+      "loss": 0.3874,
+      "mean_token_accuracy": 0.8662525892257691,
+      "step": 2230
+    },
+    {
+      "epoch": 1.0724568138195778,
+      "grad_norm": 0.4558625625028902,
+      "learning_rate": 3.5700319943121225e-05,
+      "loss": 0.4056,
+      "mean_token_accuracy": 0.8602272033691406,
+      "step": 2235
+    },
+    {
+      "epoch": 1.0748560460652592,
+      "grad_norm": 0.45346032009736464,
+      "learning_rate": 3.565588339850694e-05,
+      "loss": 0.411,
+      "mean_token_accuracy": 0.8586599171161652,
+      "step": 2240
+    },
+    {
+      "epoch": 1.0772552783109406,
+      "grad_norm": 0.46919295313231174,
+      "learning_rate": 3.561144685389264e-05,
+      "loss": 0.3986,
+      "mean_token_accuracy": 0.8619990050792694,
+      "step": 2245
+    },
+    {
+      "epoch": 1.079654510556622,
+      "grad_norm": 0.48726010152230115,
+      "learning_rate": 3.5567010309278354e-05,
+      "loss": 0.4025,
+      "mean_token_accuracy": 0.8616835474967957,
+      "step": 2250
+    },
+    {
+      "epoch": 1.0820537428023034,
+      "grad_norm": 0.48673873249135896,
+      "learning_rate": 3.5522573764664066e-05,
+      "loss": 0.4256,
+      "mean_token_accuracy": 0.8545612275600434,
+      "step": 2255
+    },
+    {
+      "epoch": 1.0844529750479845,
+      "grad_norm": 0.4771285752641356,
+      "learning_rate": 3.547813722004977e-05,
+      "loss": 0.4115,
+      "mean_token_accuracy": 0.858768516778946,
+      "step": 2260
+    },
+    {
+      "epoch": 1.086852207293666,
+      "grad_norm": 0.47918813523195114,
+      "learning_rate": 3.5433700675435476e-05,
+      "loss": 0.3993,
+      "mean_token_accuracy": 0.8627258002758026,
+      "step": 2265
+    },
+    {
+      "epoch": 1.0892514395393473,
+      "grad_norm": 0.44180464504771916,
+      "learning_rate": 3.538926413082119e-05,
+      "loss": 0.4098,
+      "mean_token_accuracy": 0.8589477658271789,
+      "step": 2270
+    },
+    {
+      "epoch": 1.0916506717850287,
+      "grad_norm": 0.48558126860005485,
+      "learning_rate": 3.53448275862069e-05,
+      "loss": 0.4018,
+      "mean_token_accuracy": 0.8613408207893372,
+      "step": 2275
+    },
+    {
+      "epoch": 1.0940499040307101,
+      "grad_norm": 0.4830201657987925,
+      "learning_rate": 3.5300391041592605e-05,
+      "loss": 0.4012,
+      "mean_token_accuracy": 0.8613366425037384,
+      "step": 2280
+    },
+    {
+      "epoch": 1.0964491362763915,
+      "grad_norm": 0.45267780099531413,
+      "learning_rate": 3.5255954496978316e-05,
+      "loss": 0.3976,
+      "mean_token_accuracy": 0.8629164338111878,
+      "step": 2285
+    },
+    {
+      "epoch": 1.098848368522073,
+      "grad_norm": 0.4502807128522101,
+      "learning_rate": 3.521151795236403e-05,
+      "loss": 0.411,
+      "mean_token_accuracy": 0.8588341414928437,
+      "step": 2290
+    },
+    {
+      "epoch": 1.1012476007677543,
+      "grad_norm": 0.44440205971107866,
+      "learning_rate": 3.516708140774973e-05,
+      "loss": 0.4135,
+      "mean_token_accuracy": 0.8572691977024078,
+      "step": 2295
+    },
+    {
+      "epoch": 1.1036468330134357,
+      "grad_norm": 0.4979388904136079,
+      "learning_rate": 3.5122644863135445e-05,
+      "loss": 0.4094,
+      "mean_token_accuracy": 0.8590193212032318,
+      "step": 2300
+    },
+    {
+      "epoch": 1.106046065259117,
+      "grad_norm": 0.46332524132082875,
+      "learning_rate": 3.507820831852116e-05,
+      "loss": 0.3928,
+      "mean_token_accuracy": 0.864500904083252,
+      "step": 2305
+    },
+    {
+      "epoch": 1.1084452975047985,
+      "grad_norm": 0.43360291665533524,
+      "learning_rate": 3.503377177390686e-05,
+      "loss": 0.4196,
+      "mean_token_accuracy": 0.8552972078323364,
+      "step": 2310
+    },
+    {
+      "epoch": 1.1108445297504799,
+      "grad_norm": 0.4429935353816478,
+      "learning_rate": 3.498933522929257e-05,
+      "loss": 0.3985,
+      "mean_token_accuracy": 0.8629589080810547,
+      "step": 2315
+    },
+    {
+      "epoch": 1.1132437619961613,
+      "grad_norm": 0.46705306376162703,
+      "learning_rate": 3.494489868467828e-05,
+      "loss": 0.3979,
+      "mean_token_accuracy": 0.8631719291210175,
+      "step": 2320
+    },
+    {
+      "epoch": 1.1156429942418427,
+      "grad_norm": 0.4418967124373468,
+      "learning_rate": 3.490046214006399e-05,
+      "loss": 0.4008,
+      "mean_token_accuracy": 0.8622421741485595,
+      "step": 2325
+    },
+    {
+      "epoch": 1.118042226487524,
+      "grad_norm": 0.44881463295860696,
+      "learning_rate": 3.4856025595449696e-05,
+      "loss": 0.4055,
+      "mean_token_accuracy": 0.8604006469249725,
+      "step": 2330
+    },
+    {
+      "epoch": 1.1204414587332054,
+      "grad_norm": 0.4919264589965018,
+      "learning_rate": 3.481158905083541e-05,
+      "loss": 0.4082,
+      "mean_token_accuracy": 0.8602576732635498,
+      "step": 2335
+    },
+    {
+      "epoch": 1.1228406909788868,
+      "grad_norm": 0.4704160431070676,
+      "learning_rate": 3.476715250622112e-05,
+      "loss": 0.4207,
+      "mean_token_accuracy": 0.8553534567356109,
+      "step": 2340
+    },
+    {
+      "epoch": 1.1252399232245682,
+      "grad_norm": 0.444446496623664,
+      "learning_rate": 3.4722715961606825e-05,
+      "loss": 0.4048,
+      "mean_token_accuracy": 0.8606355607509613,
+      "step": 2345
+    },
+    {
+      "epoch": 1.1276391554702494,
+      "grad_norm": 0.4846769028223817,
+      "learning_rate": 3.467827941699254e-05,
+      "loss": 0.3908,
+      "mean_token_accuracy": 0.8652548551559448,
+      "step": 2350
+    },
+    {
+      "epoch": 1.1300383877159308,
+      "grad_norm": 0.43614245378491234,
+      "learning_rate": 3.463384287237825e-05,
+      "loss": 0.4088,
+      "mean_token_accuracy": 0.8596396625041962,
+      "step": 2355
+    },
+    {
+      "epoch": 1.1324376199616122,
+      "grad_norm": 0.4657677207483913,
+      "learning_rate": 3.4589406327763954e-05,
+      "loss": 0.4124,
+      "mean_token_accuracy": 0.85767782330513,
+      "step": 2360
+    },
+    {
+      "epoch": 1.1348368522072936,
+      "grad_norm": 0.4426745604094487,
+      "learning_rate": 3.4544969783149665e-05,
+      "loss": 0.3971,
+      "mean_token_accuracy": 0.8634883403778076,
+      "step": 2365
+    },
+    {
+      "epoch": 1.137236084452975,
+      "grad_norm": 0.46815018460988816,
+      "learning_rate": 3.450053323853537e-05,
+      "loss": 0.4039,
+      "mean_token_accuracy": 0.8610803782939911,
+      "step": 2370
+    },
+    {
+      "epoch": 1.1396353166986564,
+      "grad_norm": 0.438316624790037,
+      "learning_rate": 3.445609669392108e-05,
+      "loss": 0.4001,
+      "mean_token_accuracy": 0.8615439713001252,
+      "step": 2375
+    },
+    {
+      "epoch": 1.1420345489443378,
+      "grad_norm": 0.4418058595426201,
+      "learning_rate": 3.441166014930679e-05,
+      "loss": 0.4065,
+      "mean_token_accuracy": 0.8600826561450958,
+      "step": 2380
+    },
+    {
+      "epoch": 1.1444337811900192,
+      "grad_norm": 0.4847595999781224,
+      "learning_rate": 3.43672236046925e-05,
+      "loss": 0.4239,
+      "mean_token_accuracy": 0.855004221200943,
+      "step": 2385
+    },
+    {
+      "epoch": 1.1468330134357005,
+      "grad_norm": 1.5446867021881845,
+      "learning_rate": 3.432278706007821e-05,
+      "loss": 0.4038,
+      "mean_token_accuracy": 0.8607834935188293,
+      "step": 2390
+    },
+    {
+      "epoch": 1.149232245681382,
+      "grad_norm": 0.48658583367657177,
+      "learning_rate": 3.4278350515463916e-05,
+      "loss": 0.4075,
+      "mean_token_accuracy": 0.8598154127597809,
+      "step": 2395
+    },
+    {
+      "epoch": 1.1516314779270633,
+      "grad_norm": 0.5326892549587043,
+      "learning_rate": 3.423391397084963e-05,
+      "loss": 0.4114,
+      "mean_token_accuracy": 0.8590575993061066,
+      "step": 2400
+    },
+    {
+      "epoch": 1.1540307101727447,
+      "grad_norm": 0.4771899586929266,
+      "learning_rate": 3.418947742623534e-05,
+      "loss": 0.4136,
+      "mean_token_accuracy": 0.8581348896026612,
+      "step": 2405
+    },
+    {
+      "epoch": 1.1564299424184261,
+      "grad_norm": 0.455821737127857,
+      "learning_rate": 3.414504088162105e-05,
+      "loss": 0.4134,
+      "mean_token_accuracy": 0.8582122266292572,
+      "step": 2410
+    },
+    {
+      "epoch": 1.1588291746641075,
+      "grad_norm": 0.5094254888878418,
+      "learning_rate": 3.410060433700676e-05,
+      "loss": 0.4169,
+      "mean_token_accuracy": 0.8570551395416259,
+      "step": 2415
+    },
+    {
+      "epoch": 1.161228406909789,
+      "grad_norm": 0.4593576345470221,
+      "learning_rate": 3.405616779239246e-05,
+      "loss": 0.4108,
+      "mean_token_accuracy": 0.8585395991802216,
+      "step": 2420
+    },
+    {
+      "epoch": 1.1636276391554703,
+      "grad_norm": 0.45335113787411774,
+      "learning_rate": 3.4011731247778174e-05,
+      "loss": 0.4141,
+      "mean_token_accuracy": 0.858020031452179,
+      "step": 2425
+    },
+    {
+      "epoch": 1.1660268714011517,
+      "grad_norm": 0.43176314428761675,
+      "learning_rate": 3.396729470316388e-05,
+      "loss": 0.4048,
+      "mean_token_accuracy": 0.8606631636619568,
+      "step": 2430
+    },
+    {
+      "epoch": 1.168426103646833,
+      "grad_norm": 0.5078546819049952,
+      "learning_rate": 3.392285815854959e-05,
+      "loss": 0.4293,
+      "mean_token_accuracy": 0.8537072598934173,
+      "step": 2435
+    },
+    {
+      "epoch": 1.1708253358925145,
+      "grad_norm": 0.4502128911684478,
+      "learning_rate": 3.38784216139353e-05,
+      "loss": 0.4055,
+      "mean_token_accuracy": 0.860330331325531,
+      "step": 2440
+    },
+    {
+      "epoch": 1.1732245681381959,
+      "grad_norm": 0.45356485938149477,
+      "learning_rate": 3.3833985069321014e-05,
+      "loss": 0.4068,
+      "mean_token_accuracy": 0.8610100626945496,
+      "step": 2445
+    },
+    {
+      "epoch": 1.1756238003838773,
+      "grad_norm": 0.44546255582121363,
+      "learning_rate": 3.378954852470672e-05,
+      "loss": 0.4059,
+      "mean_token_accuracy": 0.8603936195373535,
+      "step": 2450
+    },
+    {
+      "epoch": 1.1780230326295587,
+      "grad_norm": 0.519034529576858,
+      "learning_rate": 3.374511198009243e-05,
+      "loss": 0.3996,
+      "mean_token_accuracy": 0.8620694935321808,
+      "step": 2455
+    },
+    {
+      "epoch": 1.18042226487524,
+      "grad_norm": 0.4306474312910174,
+      "learning_rate": 3.370067543547814e-05,
+      "loss": 0.4175,
+      "mean_token_accuracy": 0.8564855635166169,
+      "step": 2460
+    },
+    {
+      "epoch": 1.1828214971209212,
+      "grad_norm": 0.44331321051212536,
+      "learning_rate": 3.365623889086385e-05,
+      "loss": 0.4221,
+      "mean_token_accuracy": 0.8551550149917603,
+      "step": 2465
+    },
+    {
+      "epoch": 1.1852207293666026,
+      "grad_norm": 0.46799379104396804,
+      "learning_rate": 3.361180234624955e-05,
+      "loss": 0.4057,
+      "mean_token_accuracy": 0.8598842024803162,
+      "step": 2470
+    },
+    {
+      "epoch": 1.187619961612284,
+      "grad_norm": 0.4456420970776554,
+      "learning_rate": 3.3567365801635265e-05,
+      "loss": 0.4123,
+      "mean_token_accuracy": 0.8581973850727082,
+      "step": 2475
+    },
+    {
+      "epoch": 1.1900191938579654,
+      "grad_norm": 0.45848417063764557,
+      "learning_rate": 3.352292925702098e-05,
+      "loss": 0.4107,
+      "mean_token_accuracy": 0.8585059881210327,
+      "step": 2480
+    },
+    {
+      "epoch": 1.1924184261036468,
+      "grad_norm": 0.4316965925133896,
+      "learning_rate": 3.347849271240668e-05,
+      "loss": 0.4138,
+      "mean_token_accuracy": 0.8580934762954712,
+      "step": 2485
+    },
+    {
+      "epoch": 1.1948176583493282,
+      "grad_norm": 0.43525772194220896,
+      "learning_rate": 3.3434056167792394e-05,
+      "loss": 0.4149,
+      "mean_token_accuracy": 0.8575979828834533,
+      "step": 2490
+    },
+    {
+      "epoch": 1.1972168905950096,
+      "grad_norm": 0.46870966305994644,
+      "learning_rate": 3.3389619623178106e-05,
+      "loss": 0.4175,
+      "mean_token_accuracy": 0.856852000951767,
+      "step": 2495
+    },
+    {
+      "epoch": 1.199616122840691,
+      "grad_norm": 0.40911552344798013,
+      "learning_rate": 3.334518307856381e-05,
+      "loss": 0.4009,
+      "mean_token_accuracy": 0.8614593029022217,
+      "step": 2500
+    },
+    {
+      "epoch": 1.2020153550863724,
+      "grad_norm": 0.4340404405974866,
+      "learning_rate": 3.330074653394952e-05,
+      "loss": 0.3927,
+      "mean_token_accuracy": 0.8641118109226227,
+      "step": 2505
+    },
+    {
+      "epoch": 1.2044145873320538,
+      "grad_norm": 0.43787534396656413,
+      "learning_rate": 3.3256309989335235e-05,
+      "loss": 0.4085,
+      "mean_token_accuracy": 0.8591709673404694,
+      "step": 2510
+    },
+    {
+      "epoch": 1.2068138195777351,
+      "grad_norm": 0.44863502667369215,
+      "learning_rate": 3.321187344472094e-05,
+      "loss": 0.4144,
+      "mean_token_accuracy": 0.8567050993442535,
+      "step": 2515
+    },
+    {
+      "epoch": 1.2092130518234165,
+      "grad_norm": 0.4213872953371581,
+      "learning_rate": 3.316743690010665e-05,
+      "loss": 0.3892,
+      "mean_token_accuracy": 0.8648431181907654,
+      "step": 2520
+    },
+    {
+      "epoch": 1.211612284069098,
+      "grad_norm": 0.3923155586128721,
+      "learning_rate": 3.3123000355492357e-05,
+      "loss": 0.4038,
+      "mean_token_accuracy": 0.8607444286346435,
+      "step": 2525
+    },
+    {
+      "epoch": 1.2140115163147793,
+      "grad_norm": 0.433505967574003,
+      "learning_rate": 3.307856381087807e-05,
+      "loss": 0.4046,
+      "mean_token_accuracy": 0.8599982798099518,
+      "step": 2530
+    },
+    {
+      "epoch": 1.2164107485604607,
+      "grad_norm": 0.40196481275606427,
+      "learning_rate": 3.3034127266263773e-05,
+      "loss": 0.3952,
+      "mean_token_accuracy": 0.8635883510112763,
+      "step": 2535
+    },
+    {
+      "epoch": 1.2188099808061421,
+      "grad_norm": 0.4360372715597363,
+      "learning_rate": 3.2989690721649485e-05,
+      "loss": 0.4021,
+      "mean_token_accuracy": 0.8613421142101287,
+      "step": 2540
+    },
+    {
+      "epoch": 1.2212092130518235,
+      "grad_norm": 0.4316984682373569,
+      "learning_rate": 3.29452541770352e-05,
+      "loss": 0.4086,
+      "mean_token_accuracy": 0.8592318296432495,
+      "step": 2545
+    },
+    {
+      "epoch": 1.2236084452975047,
+      "grad_norm": 0.456502910992107,
+      "learning_rate": 3.29008176324209e-05,
+      "loss": 0.4066,
+      "mean_token_accuracy": 0.8597779631614685,
+      "step": 2550
+    },
+    {
+      "epoch": 1.226007677543186,
+      "grad_norm": 0.4444102807735176,
+      "learning_rate": 3.2856381087806614e-05,
+      "loss": 0.4074,
+      "mean_token_accuracy": 0.8595255970954895,
+      "step": 2555
+    },
+    {
+      "epoch": 1.2284069097888675,
+      "grad_norm": 0.4231888031917078,
+      "learning_rate": 3.2811944543192326e-05,
+      "loss": 0.4072,
+      "mean_token_accuracy": 0.8598849952220917,
+      "step": 2560
+    },
+    {
+      "epoch": 1.2308061420345489,
+      "grad_norm": 0.4570773325504898,
+      "learning_rate": 3.276750799857803e-05,
+      "loss": 0.4204,
+      "mean_token_accuracy": 0.85575150847435,
+      "step": 2565
+    },
+    {
+      "epoch": 1.2332053742802302,
+      "grad_norm": 0.411842115095972,
+      "learning_rate": 3.272307145396374e-05,
+      "loss": 0.4013,
+      "mean_token_accuracy": 0.8611139714717865,
+      "step": 2570
+    },
+    {
+      "epoch": 1.2356046065259116,
+      "grad_norm": 0.45644905287947346,
+      "learning_rate": 3.267863490934945e-05,
+      "loss": 0.4153,
+      "mean_token_accuracy": 0.8568707346916199,
+      "step": 2575
+    },
+    {
+      "epoch": 1.238003838771593,
+      "grad_norm": 0.4818186654409561,
+      "learning_rate": 3.263419836473516e-05,
+      "loss": 0.4013,
+      "mean_token_accuracy": 0.8617413580417633,
+      "step": 2580
+    },
+    {
+      "epoch": 1.2404030710172744,
+      "grad_norm": 0.438770764421927,
+      "learning_rate": 3.2589761820120865e-05,
+      "loss": 0.4122,
+      "mean_token_accuracy": 0.8579687118530274,
+      "step": 2585
+    },
+    {
+      "epoch": 1.2428023032629558,
+      "grad_norm": 0.4653378155990347,
+      "learning_rate": 3.254532527550658e-05,
+      "loss": 0.4183,
+      "mean_token_accuracy": 0.8564316511154175,
+      "step": 2590
+    },
+    {
+      "epoch": 1.2452015355086372,
+      "grad_norm": 0.4813156809808044,
+      "learning_rate": 3.250088873089229e-05,
+      "loss": 0.4283,
+      "mean_token_accuracy": 0.853442233800888,
+      "step": 2595
+    },
+    {
+      "epoch": 1.2476007677543186,
+      "grad_norm": 0.41203719170614067,
+      "learning_rate": 3.2456452186277994e-05,
+      "loss": 0.3965,
+      "mean_token_accuracy": 0.862660163640976,
+      "step": 2600
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 0.47609148798409084,
+      "learning_rate": 3.2412015641663706e-05,
+      "loss": 0.4175,
+      "mean_token_accuracy": 0.8569301187992096,
+      "step": 2605
+    },
+    {
+      "epoch": 1.2523992322456814,
+      "grad_norm": 0.42629307255923604,
+      "learning_rate": 3.236757909704942e-05,
+      "loss": 0.4138,
+      "mean_token_accuracy": 0.8578989326953887,
+      "step": 2610
+    },
+    {
+      "epoch": 1.2547984644913628,
+      "grad_norm": 0.4283572421423295,
+      "learning_rate": 3.232314255243513e-05,
+      "loss": 0.4105,
+      "mean_token_accuracy": 0.8592274427413941,
+      "step": 2615
+    },
+    {
+      "epoch": 1.2571976967370442,
+      "grad_norm": 0.43715619923246357,
+      "learning_rate": 3.2278706007820834e-05,
+      "loss": 0.4007,
+      "mean_token_accuracy": 0.8620129883289337,
+      "step": 2620
+    },
+    {
+      "epoch": 1.2595969289827256,
+      "grad_norm": 0.42096711462262465,
+      "learning_rate": 3.223426946320654e-05,
+      "loss": 0.3966,
+      "mean_token_accuracy": 0.8628343999385834,
+      "step": 2625
+    },
+    {
+      "epoch": 1.261996161228407,
+      "grad_norm": 0.41932841031562834,
+      "learning_rate": 3.218983291859225e-05,
+      "loss": 0.4051,
+      "mean_token_accuracy": 0.8604014277458191,
+      "step": 2630
+    },
+    {
+      "epoch": 1.2643953934740884,
+      "grad_norm": 0.5315771823488319,
+      "learning_rate": 3.2145396373977956e-05,
+      "loss": 0.4322,
+      "mean_token_accuracy": 0.8546823263168335,
+      "step": 2635
+    },
+    {
+      "epoch": 1.2667946257197698,
+      "grad_norm": 0.41897523637834844,
+      "learning_rate": 3.210095982936367e-05,
+      "loss": 0.3968,
+      "mean_token_accuracy": 0.86270392537117,
+      "step": 2640
+    },
+    {
+      "epoch": 1.2691938579654511,
+      "grad_norm": 0.40715830938846465,
+      "learning_rate": 3.205652328474938e-05,
+      "loss": 0.4068,
+      "mean_token_accuracy": 0.8598295211791992,
+      "step": 2645
+    },
+    {
+      "epoch": 1.2715930902111325,
+      "grad_norm": 0.8769975146104805,
+      "learning_rate": 3.201208674013509e-05,
+      "loss": 0.425,
+      "mean_token_accuracy": 0.854360431432724,
+      "step": 2650
+    },
+    {
+      "epoch": 1.273992322456814,
+      "grad_norm": 3.3831831887481756,
+      "learning_rate": 3.19676501955208e-05,
+      "loss": 0.4314,
+      "mean_token_accuracy": 0.852653294801712,
+      "step": 2655
+    },
+    {
+      "epoch": 1.2763915547024953,
+      "grad_norm": 0.45730751260407754,
+      "learning_rate": 3.192321365090651e-05,
+      "loss": 0.4195,
+      "mean_token_accuracy": 0.8559574007987976,
+      "step": 2660
+    },
+    {
+      "epoch": 1.2787907869481767,
+      "grad_norm": 0.4092466623798688,
+      "learning_rate": 3.187877710629222e-05,
+      "loss": 0.3926,
+      "mean_token_accuracy": 0.8644907474517822,
+      "step": 2665
+    },
+    {
+      "epoch": 1.2811900191938579,
+      "grad_norm": 0.4309016236914282,
+      "learning_rate": 3.1834340561677926e-05,
+      "loss": 0.4066,
+      "mean_token_accuracy": 0.8598381280899048,
+      "step": 2670
+    },
+    {
+      "epoch": 1.2835892514395393,
+      "grad_norm": 0.5020853714609004,
+      "learning_rate": 3.178990401706363e-05,
+      "loss": 0.4025,
+      "mean_token_accuracy": 0.8616749465465545,
+      "step": 2675
+    },
+    {
+      "epoch": 1.2859884836852207,
+      "grad_norm": 0.4564585894852153,
+      "learning_rate": 3.174546747244934e-05,
+      "loss": 0.4065,
+      "mean_token_accuracy": 0.8599982798099518,
+      "step": 2680
+    },
+    {
+      "epoch": 1.288387715930902,
+      "grad_norm": 0.4551031892701387,
+      "learning_rate": 3.1701030927835054e-05,
+      "loss": 0.4055,
+      "mean_token_accuracy": 0.8602951765060425,
+      "step": 2685
+    },
+    {
+      "epoch": 1.2907869481765835,
+      "grad_norm": 0.5107255154240907,
+      "learning_rate": 3.165659438322076e-05,
+      "loss": 0.423,
+      "mean_token_accuracy": 0.8549253046512604,
+      "step": 2690
+    },
+    {
+      "epoch": 1.2931861804222649,
+      "grad_norm": 0.500179788278748,
+      "learning_rate": 3.161215783860647e-05,
+      "loss": 0.4212,
+      "mean_token_accuracy": 0.8551692306995392,
+      "step": 2695
+    },
+    {
+      "epoch": 1.2955854126679462,
+      "grad_norm": 0.4230688657327201,
+      "learning_rate": 3.156772129399218e-05,
+      "loss": 0.4213,
+      "mean_token_accuracy": 0.8560464382171631,
+      "step": 2700
+    },
+    {
+      "epoch": 1.2979846449136276,
+      "grad_norm": 0.43479209014268144,
+      "learning_rate": 3.152328474937789e-05,
+      "loss": 0.3941,
+      "mean_token_accuracy": 0.8634872555732727,
+      "step": 2705
+    },
+    {
+      "epoch": 1.300383877159309,
+      "grad_norm": 0.43650872497538457,
+      "learning_rate": 3.14788482047636e-05,
+      "loss": 0.408,
+      "mean_token_accuracy": 0.8591680586338043,
+      "step": 2710
+    },
+    {
+      "epoch": 1.3027831094049904,
+      "grad_norm": 0.4493400523195851,
+      "learning_rate": 3.143441166014931e-05,
+      "loss": 0.4175,
+      "mean_token_accuracy": 0.8564364910125732,
+      "step": 2715
+    },
+    {
+      "epoch": 1.3051823416506718,
+      "grad_norm": 0.459115570189103,
+      "learning_rate": 3.138997511553502e-05,
+      "loss": 0.403,
+      "mean_token_accuracy": 0.860854583978653,
+      "step": 2720
+    },
+    {
+      "epoch": 1.3075815738963532,
+      "grad_norm": 0.45103286240520074,
+      "learning_rate": 3.134553857092073e-05,
+      "loss": 0.4134,
+      "mean_token_accuracy": 0.8586698293685913,
+      "step": 2725
+    },
+    {
+      "epoch": 1.3099808061420346,
+      "grad_norm": 0.4272327842623446,
+      "learning_rate": 3.1301102026306434e-05,
+      "loss": 0.3999,
+      "mean_token_accuracy": 0.8617429137229919,
+      "step": 2730
+    },
+    {
+      "epoch": 1.312380038387716,
+      "grad_norm": 0.45529413891054127,
+      "learning_rate": 3.1256665481692146e-05,
+      "loss": 0.4129,
+      "mean_token_accuracy": 0.8579215168952942,
+      "step": 2735
+    },
+    {
+      "epoch": 1.3147792706333974,
+      "grad_norm": 0.4086111236520043,
+      "learning_rate": 3.121222893707785e-05,
+      "loss": 0.4068,
+      "mean_token_accuracy": 0.8598678052425385,
+      "step": 2740
+    },
+    {
+      "epoch": 1.3171785028790786,
+      "grad_norm": 0.44772635264843236,
+      "learning_rate": 3.116779239246356e-05,
+      "loss": 0.4197,
+      "mean_token_accuracy": 0.8561370968818665,
+      "step": 2745
+    },
+    {
+      "epoch": 1.31957773512476,
+      "grad_norm": 0.4521226002995627,
+      "learning_rate": 3.1123355847849275e-05,
+      "loss": 0.4242,
+      "mean_token_accuracy": 0.8544846594333648,
+      "step": 2750
+    },
+    {
+      "epoch": 1.3219769673704413,
+      "grad_norm": 0.435967523632314,
+      "learning_rate": 3.107891930323498e-05,
+      "loss": 0.4005,
+      "mean_token_accuracy": 0.8615476310253143,
+      "step": 2755
+    },
+    {
+      "epoch": 1.3243761996161227,
+      "grad_norm": 0.45930032570272467,
+      "learning_rate": 3.103448275862069e-05,
+      "loss": 0.4041,
+      "mean_token_accuracy": 0.8609639704227448,
+      "step": 2760
+    },
+    {
+      "epoch": 1.3267754318618041,
+      "grad_norm": 0.44549351512795554,
+      "learning_rate": 3.0990046214006403e-05,
+      "loss": 0.3965,
+      "mean_token_accuracy": 0.8633203625679016,
+      "step": 2765
+    },
+    {
+      "epoch": 1.3291746641074855,
+      "grad_norm": 0.3934861435770103,
+      "learning_rate": 3.094560966939211e-05,
+      "loss": 0.4045,
+      "mean_token_accuracy": 0.859321677684784,
+      "step": 2770
+    },
+    {
+      "epoch": 1.331573896353167,
+      "grad_norm": 0.41342925620089666,
+      "learning_rate": 3.090117312477782e-05,
+      "loss": 0.4033,
+      "mean_token_accuracy": 0.8610045909881592,
+      "step": 2775
+    },
+    {
+      "epoch": 1.3339731285988483,
+      "grad_norm": 0.403635823706059,
+      "learning_rate": 3.0856736580163525e-05,
+      "loss": 0.3841,
+      "mean_token_accuracy": 0.8668143391609192,
+      "step": 2780
+    },
+    {
+      "epoch": 1.3363723608445297,
+      "grad_norm": 0.39217459524479503,
+      "learning_rate": 3.081230003554924e-05,
+      "loss": 0.4054,
+      "mean_token_accuracy": 0.8603837728500366,
+      "step": 2785
+    },
+    {
+      "epoch": 1.338771593090211,
+      "grad_norm": 0.4134206958266228,
+      "learning_rate": 3.076786349093494e-05,
+      "loss": 0.3956,
+      "mean_token_accuracy": 0.8637172639369964,
+      "step": 2790
+    },
+    {
+      "epoch": 1.3411708253358925,
+      "grad_norm": 0.41573581786552966,
+      "learning_rate": 3.0723426946320654e-05,
+      "loss": 0.4058,
+      "mean_token_accuracy": 0.8605648100376129,
+      "step": 2795
+    },
+    {
+      "epoch": 1.3435700575815739,
+      "grad_norm": 0.443513172094798,
+      "learning_rate": 3.0678990401706366e-05,
+      "loss": 0.4093,
+      "mean_token_accuracy": 0.858421903848648,
+      "step": 2800
+    },
+    {
+      "epoch": 1.3459692898272553,
+      "grad_norm": 0.43938493124740363,
+      "learning_rate": 3.063455385709207e-05,
+      "loss": 0.4065,
+      "mean_token_accuracy": 0.8605037808418274,
+      "step": 2805
+    },
+    {
+      "epoch": 1.3483685220729367,
+      "grad_norm": 0.4165424593181452,
+      "learning_rate": 3.059011731247778e-05,
+      "loss": 0.41,
+      "mean_token_accuracy": 0.8584935009479523,
+      "step": 2810
+    },
+    {
+      "epoch": 1.350767754318618,
+      "grad_norm": 0.43948093820333756,
+      "learning_rate": 3.0545680767863495e-05,
+      "loss": 0.4058,
+      "mean_token_accuracy": 0.8599857807159423,
+      "step": 2815
+    },
+    {
+      "epoch": 1.3531669865642995,
+      "grad_norm": 0.4340762126855867,
+      "learning_rate": 3.0501244223249203e-05,
+      "loss": 0.4151,
+      "mean_token_accuracy": 0.8570738732814789,
+      "step": 2820
+    },
+    {
+      "epoch": 1.3555662188099808,
+      "grad_norm": 0.4279976668729953,
+      "learning_rate": 3.0456807678634912e-05,
+      "loss": 0.4113,
+      "mean_token_accuracy": 0.8584268510341644,
+      "step": 2825
+    },
+    {
+      "epoch": 1.3579654510556622,
+      "grad_norm": 0.4643169392851994,
+      "learning_rate": 3.0412371134020617e-05,
+      "loss": 0.4082,
+      "mean_token_accuracy": 0.8594587981700897,
+      "step": 2830
+    },
+    {
+      "epoch": 1.3603646833013436,
+      "grad_norm": 0.4520395742541979,
+      "learning_rate": 3.036793458940633e-05,
+      "loss": 0.4152,
+      "mean_token_accuracy": 0.8566230773925781,
+      "step": 2835
+    },
+    {
+      "epoch": 1.362763915547025,
+      "grad_norm": 0.43104350994262125,
+      "learning_rate": 3.0323498044792037e-05,
+      "loss": 0.405,
+      "mean_token_accuracy": 0.8607537925243378,
+      "step": 2840
+    },
+    {
+      "epoch": 1.3651631477927064,
+      "grad_norm": 0.4946581783931828,
+      "learning_rate": 3.0279061500177746e-05,
+      "loss": 0.4108,
+      "mean_token_accuracy": 0.8585208475589752,
+      "step": 2845
+    },
+    {
+      "epoch": 1.3675623800383878,
+      "grad_norm": 0.43118015846032265,
+      "learning_rate": 3.0234624955563457e-05,
+      "loss": 0.4167,
+      "mean_token_accuracy": 0.8567094743251801,
+      "step": 2850
+    },
+    {
+      "epoch": 1.3699616122840692,
+      "grad_norm": 0.4205461305435935,
+      "learning_rate": 3.0190188410949166e-05,
+      "loss": 0.4069,
+      "mean_token_accuracy": 0.8604311168193817,
+      "step": 2855
+    },
+    {
+      "epoch": 1.3723608445297506,
+      "grad_norm": 0.45673558688969584,
+      "learning_rate": 3.0145751866334874e-05,
+      "loss": 0.401,
+      "mean_token_accuracy": 0.862135136127472,
+      "step": 2860
+    },
+    {
+      "epoch": 1.374760076775432,
+      "grad_norm": 0.49315897603113723,
+      "learning_rate": 3.0101315321720586e-05,
+      "loss": 0.41,
+      "mean_token_accuracy": 0.8588114857673645,
+      "step": 2865
+    },
+    {
+      "epoch": 1.3771593090211132,
+      "grad_norm": 0.4411223310245759,
+      "learning_rate": 3.0056878777106295e-05,
+      "loss": 0.3969,
+      "mean_token_accuracy": 0.8631333410739899,
+      "step": 2870
+    },
+    {
+      "epoch": 1.3795585412667946,
+      "grad_norm": 0.4129525665270095,
+      "learning_rate": 3.0012442232492007e-05,
+      "loss": 0.3946,
+      "mean_token_accuracy": 0.8636532902717591,
+      "step": 2875
+    },
+    {
+      "epoch": 1.381957773512476,
+      "grad_norm": 0.4103741557418468,
+      "learning_rate": 2.9968005687877708e-05,
+      "loss": 0.3983,
+      "mean_token_accuracy": 0.8622260808944702,
+      "step": 2880
+    },
+    {
+      "epoch": 1.3843570057581573,
+      "grad_norm": 0.40528936732169213,
+      "learning_rate": 2.992356914326342e-05,
+      "loss": 0.4044,
+      "mean_token_accuracy": 0.8614343047142029,
+      "step": 2885
+    },
+    {
+      "epoch": 1.3867562380038387,
+      "grad_norm": 0.4800728703117366,
+      "learning_rate": 2.987913259864913e-05,
+      "loss": 0.4201,
+      "mean_token_accuracy": 0.8567433834075928,
+      "step": 2890
+    },
+    {
+      "epoch": 1.3891554702495201,
+      "grad_norm": 0.4384817482064134,
+      "learning_rate": 2.9834696054034837e-05,
+      "loss": 0.4136,
+      "mean_token_accuracy": 0.8570279777050018,
+      "step": 2895
+    },
+    {
+      "epoch": 1.3915547024952015,
+      "grad_norm": 0.40788290325888715,
+      "learning_rate": 2.979025950942055e-05,
+      "loss": 0.4047,
+      "mean_token_accuracy": 0.8605920732021332,
+      "step": 2900
+    },
+    {
+      "epoch": 1.393953934740883,
+      "grad_norm": 0.4371937523140351,
+      "learning_rate": 2.9745822964806257e-05,
+      "loss": 0.4055,
+      "mean_token_accuracy": 0.8605655014514924,
+      "step": 2905
+    },
+    {
+      "epoch": 1.3963531669865643,
+      "grad_norm": 0.42445573104601925,
+      "learning_rate": 2.970138642019197e-05,
+      "loss": 0.4054,
+      "mean_token_accuracy": 0.8602490782737732,
+      "step": 2910
+    },
+    {
+      "epoch": 1.3987523992322457,
+      "grad_norm": 0.40880871706092814,
+      "learning_rate": 2.9656949875577678e-05,
+      "loss": 0.4028,
+      "mean_token_accuracy": 0.8603983104228974,
+      "step": 2915
+    },
+    {
+      "epoch": 1.401151631477927,
+      "grad_norm": 0.42266914941617284,
+      "learning_rate": 2.9612513330963386e-05,
+      "loss": 0.397,
+      "mean_token_accuracy": 0.8627976059913636,
+      "step": 2920
+    },
+    {
+      "epoch": 1.4035508637236085,
+      "grad_norm": 0.43438131430207655,
+      "learning_rate": 2.9568076786349098e-05,
+      "loss": 0.4179,
+      "mean_token_accuracy": 0.8561128735542297,
+      "step": 2925
+    },
+    {
+      "epoch": 1.4059500959692899,
+      "grad_norm": 0.41884217203981505,
+      "learning_rate": 2.9523640241734806e-05,
+      "loss": 0.4081,
+      "mean_token_accuracy": 0.8592755794525146,
+      "step": 2930
+    },
+    {
+      "epoch": 1.4083493282149713,
+      "grad_norm": 0.42071658115511323,
+      "learning_rate": 2.947920369712051e-05,
+      "loss": 0.4073,
+      "mean_token_accuracy": 0.858444768190384,
+      "step": 2935
+    },
+    {
+      "epoch": 1.4107485604606527,
+      "grad_norm": 0.4398724235442647,
+      "learning_rate": 2.943476715250622e-05,
+      "loss": 0.4189,
+      "mean_token_accuracy": 0.856487900018692,
+      "step": 2940
+    },
+    {
+      "epoch": 1.4131477927063338,
+      "grad_norm": 0.4153146639627856,
+      "learning_rate": 2.9390330607891932e-05,
+      "loss": 0.4054,
+      "mean_token_accuracy": 0.8598765075206757,
+      "step": 2945
+    },
+    {
+      "epoch": 1.4155470249520152,
+      "grad_norm": 0.4246136142452087,
+      "learning_rate": 2.934589406327764e-05,
+      "loss": 0.4058,
+      "mean_token_accuracy": 0.859784209728241,
+      "step": 2950
+    },
+    {
+      "epoch": 1.4179462571976966,
+      "grad_norm": 0.4368135342362413,
+      "learning_rate": 2.930145751866335e-05,
+      "loss": 0.4108,
+      "mean_token_accuracy": 0.8586536586284638,
+      "step": 2955
+    },
+    {
+      "epoch": 1.420345489443378,
+      "grad_norm": 0.42306966276186897,
+      "learning_rate": 2.925702097404906e-05,
+      "loss": 0.4027,
+      "mean_token_accuracy": 0.8606998920440674,
+      "step": 2960
+    },
+    {
+      "epoch": 1.4227447216890594,
+      "grad_norm": 0.40958147969244185,
+      "learning_rate": 2.921258442943477e-05,
+      "loss": 0.4083,
+      "mean_token_accuracy": 0.8588349342346191,
+      "step": 2965
+    },
+    {
+      "epoch": 1.4251439539347408,
+      "grad_norm": 0.4186234420817474,
+      "learning_rate": 2.9168147884820478e-05,
+      "loss": 0.4103,
+      "mean_token_accuracy": 0.8586184978485107,
+      "step": 2970
+    },
+    {
+      "epoch": 1.4275431861804222,
+      "grad_norm": 0.43568043114669147,
+      "learning_rate": 2.912371134020619e-05,
+      "loss": 0.42,
+      "mean_token_accuracy": 0.8555362820625305,
+      "step": 2975
+    },
+    {
+      "epoch": 1.4299424184261036,
+      "grad_norm": 0.37993098256654056,
+      "learning_rate": 2.9079274795591898e-05,
+      "loss": 0.3977,
+      "mean_token_accuracy": 0.8621952950954437,
+      "step": 2980
+    },
+    {
+      "epoch": 1.432341650671785,
+      "grad_norm": 0.40862306924190445,
+      "learning_rate": 2.9034838250977603e-05,
+      "loss": 0.4099,
+      "mean_token_accuracy": 0.8586802303791046,
+      "step": 2985
+    },
+    {
+      "epoch": 1.4347408829174664,
+      "grad_norm": 0.4256385901866345,
+      "learning_rate": 2.899040170636331e-05,
+      "loss": 0.3993,
+      "mean_token_accuracy": 0.8616679191589356,
+      "step": 2990
+    },
+    {
+      "epoch": 1.4371401151631478,
+      "grad_norm": 0.4444750312443054,
+      "learning_rate": 2.8945965161749023e-05,
+      "loss": 0.4138,
+      "mean_token_accuracy": 0.8581645727157593,
+      "step": 2995
+    },
+    {
+      "epoch": 1.4395393474088292,
+      "grad_norm": 0.42054008525573455,
+      "learning_rate": 2.8901528617134732e-05,
+      "loss": 0.3985,
+      "mean_token_accuracy": 0.8622124850749969,
+      "step": 3000
+    },
+    {
+      "epoch": 1.4419385796545106,
+      "grad_norm": 0.39199991797047684,
+      "learning_rate": 2.8857092072520444e-05,
+      "loss": 0.3978,
+      "mean_token_accuracy": 0.8625296950340271,
+      "step": 3005
+    },
+    {
+      "epoch": 1.444337811900192,
+      "grad_norm": 0.4252466446636856,
+      "learning_rate": 2.8812655527906152e-05,
+      "loss": 0.4053,
+      "mean_token_accuracy": 0.8601209402084351,
+      "step": 3010
+    },
+    {
+      "epoch": 1.4467370441458733,
+      "grad_norm": 0.4250623615104366,
+      "learning_rate": 2.876821898329186e-05,
+      "loss": 0.4018,
+      "mean_token_accuracy": 0.8616835534572601,
+      "step": 3015
+    },
+    {
+      "epoch": 1.4491362763915547,
+      "grad_norm": 0.4072236614562204,
+      "learning_rate": 2.8723782438677572e-05,
+      "loss": 0.4009,
+      "mean_token_accuracy": 0.8617772936820984,
+      "step": 3020
+    },
+    {
+      "epoch": 1.4515355086372361,
+      "grad_norm": 0.39611034940959056,
+      "learning_rate": 2.867934589406328e-05,
+      "loss": 0.3968,
+      "mean_token_accuracy": 0.8629406571388245,
+      "step": 3025
+    },
+    {
+      "epoch": 1.4539347408829175,
+      "grad_norm": 0.4027069569602149,
+      "learning_rate": 2.863490934944899e-05,
+      "loss": 0.4053,
+      "mean_token_accuracy": 0.8606762647628784,
+      "step": 3030
+    },
+    {
+      "epoch": 1.456333973128599,
+      "grad_norm": 0.43051154118237706,
+      "learning_rate": 2.8590472804834694e-05,
+      "loss": 0.4094,
+      "mean_token_accuracy": 0.859013843536377,
+      "step": 3035
+    },
+    {
+      "epoch": 1.4587332053742803,
+      "grad_norm": 0.42541461100511707,
+      "learning_rate": 2.8546036260220406e-05,
+      "loss": 0.4072,
+      "mean_token_accuracy": 0.860070937871933,
+      "step": 3040
+    },
+    {
+      "epoch": 1.4611324376199617,
+      "grad_norm": 0.41627586707069786,
+      "learning_rate": 2.8501599715606115e-05,
+      "loss": 0.4167,
+      "mean_token_accuracy": 0.8566191613674163,
+      "step": 3045
+    },
+    {
+      "epoch": 1.463531669865643,
+      "grad_norm": 0.41940607277568154,
+      "learning_rate": 2.8457163170991823e-05,
+      "loss": 0.401,
+      "mean_token_accuracy": 0.8612983524799347,
+      "step": 3050
+    },
+    {
+      "epoch": 1.4659309021113245,
+      "grad_norm": 0.3978647928798745,
+      "learning_rate": 2.8412726626377535e-05,
+      "loss": 0.4126,
+      "mean_token_accuracy": 0.8580301880836487,
+      "step": 3055
+    },
+    {
+      "epoch": 1.4683301343570059,
+      "grad_norm": 0.4688543778463637,
+      "learning_rate": 2.8368290081763243e-05,
+      "loss": 0.4012,
+      "mean_token_accuracy": 0.8619030773639679,
+      "step": 3060
+    },
+    {
+      "epoch": 1.4707293666026873,
+      "grad_norm": 0.4247011368186658,
+      "learning_rate": 2.8323853537148952e-05,
+      "loss": 0.4119,
+      "mean_token_accuracy": 0.8577637672424316,
+      "step": 3065
+    },
+    {
+      "epoch": 1.4731285988483684,
+      "grad_norm": 0.3864061246345447,
+      "learning_rate": 2.8279416992534664e-05,
+      "loss": 0.4062,
+      "mean_token_accuracy": 0.8601014137268066,
+      "step": 3070
+    },
+    {
+      "epoch": 1.4755278310940498,
+      "grad_norm": 0.4127338674644227,
+      "learning_rate": 2.8234980447920372e-05,
+      "loss": 0.4116,
+      "mean_token_accuracy": 0.8582200467586517,
+      "step": 3075
+    },
+    {
+      "epoch": 1.4779270633397312,
+      "grad_norm": 0.4242265641239629,
+      "learning_rate": 2.8190543903306084e-05,
+      "loss": 0.4235,
+      "mean_token_accuracy": 0.85434011220932,
+      "step": 3080
+    },
+    {
+      "epoch": 1.4803262955854126,
+      "grad_norm": 0.461521417802285,
+      "learning_rate": 2.8146107358691786e-05,
+      "loss": 0.4047,
+      "mean_token_accuracy": 0.860513162612915,
+      "step": 3085
+    },
+    {
+      "epoch": 1.482725527831094,
+      "grad_norm": 0.3917183172391874,
+      "learning_rate": 2.8101670814077498e-05,
+      "loss": 0.4089,
+      "mean_token_accuracy": 0.8591904163360595,
+      "step": 3090
+    },
+    {
+      "epoch": 1.4851247600767754,
+      "grad_norm": 0.43071306913770446,
+      "learning_rate": 2.8057234269463206e-05,
+      "loss": 0.4206,
+      "mean_token_accuracy": 0.8550276577472686,
+      "step": 3095
+    },
+    {
+      "epoch": 1.4875239923224568,
+      "grad_norm": 0.40999435056243977,
+      "learning_rate": 2.8012797724848915e-05,
+      "loss": 0.3958,
+      "mean_token_accuracy": 0.8640438497066498,
+      "step": 3100
+    },
+    {
+      "epoch": 1.4899232245681382,
+      "grad_norm": 0.4383534888886836,
+      "learning_rate": 2.7968361180234626e-05,
+      "loss": 0.4004,
+      "mean_token_accuracy": 0.8615210294723511,
+      "step": 3105
+    },
+    {
+      "epoch": 1.4923224568138196,
+      "grad_norm": 0.4018639584899305,
+      "learning_rate": 2.7923924635620335e-05,
+      "loss": 0.4,
+      "mean_token_accuracy": 0.8615460336208344,
+      "step": 3110
+    },
+    {
+      "epoch": 1.494721689059501,
+      "grad_norm": 0.4176219228220342,
+      "learning_rate": 2.7879488091006047e-05,
+      "loss": 0.4013,
+      "mean_token_accuracy": 0.8609389662742615,
+      "step": 3115
+    },
+    {
+      "epoch": 1.4971209213051824,
+      "grad_norm": 0.44302343248595116,
+      "learning_rate": 2.7835051546391755e-05,
+      "loss": 0.4106,
+      "mean_token_accuracy": 0.8591771364212036,
+      "step": 3120
+    },
+    {
+      "epoch": 1.4995201535508638,
+      "grad_norm": 0.4178864693802271,
+      "learning_rate": 2.7790615001777464e-05,
+      "loss": 0.4019,
+      "mean_token_accuracy": 0.861190801858902,
+      "step": 3125
+    },
+    {
+      "epoch": 1.5019193857965452,
+      "grad_norm": 0.44169993683798264,
+      "learning_rate": 2.7746178457163175e-05,
+      "loss": 0.411,
+      "mean_token_accuracy": 0.8583481729030609,
+      "step": 3130
+    },
+    {
+      "epoch": 1.5043186180422263,
+      "grad_norm": 0.4080953695790485,
+      "learning_rate": 2.7701741912548884e-05,
+      "loss": 0.3981,
+      "mean_token_accuracy": 0.8621671617031097,
+      "step": 3135
+    },
+    {
+      "epoch": 1.5067178502879077,
+      "grad_norm": 0.3993318929444135,
+      "learning_rate": 2.765730536793459e-05,
+      "loss": 0.3962,
+      "mean_token_accuracy": 0.8624468684196472,
+      "step": 3140
+    },
+    {
+      "epoch": 1.5091170825335891,
+      "grad_norm": 0.42653052791592105,
+      "learning_rate": 2.7612868823320297e-05,
+      "loss": 0.4071,
+      "mean_token_accuracy": 0.858930242061615,
+      "step": 3145
+    },
+    {
+      "epoch": 1.5115163147792705,
+      "grad_norm": 0.42799410610582433,
+      "learning_rate": 2.756843227870601e-05,
+      "loss": 0.3989,
+      "mean_token_accuracy": 0.8621241927146912,
+      "step": 3150
+    },
+    {
+      "epoch": 1.513915547024952,
+      "grad_norm": 0.4269578808768684,
+      "learning_rate": 2.7523995734091718e-05,
+      "loss": 0.3916,
+      "mean_token_accuracy": 0.8643751204013824,
+      "step": 3155
+    },
+    {
+      "epoch": 1.5163147792706333,
+      "grad_norm": 0.407062947790096,
+      "learning_rate": 2.7479559189477426e-05,
+      "loss": 0.4188,
+      "mean_token_accuracy": 0.8555765926837922,
+      "step": 3160
+    },
+    {
+      "epoch": 1.5187140115163147,
+      "grad_norm": 0.3994207604991086,
+      "learning_rate": 2.7435122644863138e-05,
+      "loss": 0.4142,
+      "mean_token_accuracy": 0.8575942277908325,
+      "step": 3165
+    },
+    {
+      "epoch": 1.521113243761996,
+      "grad_norm": 0.43620726489357947,
+      "learning_rate": 2.7390686100248847e-05,
+      "loss": 0.3989,
+      "mean_token_accuracy": 0.8623096764087677,
+      "step": 3170
+    },
+    {
+      "epoch": 1.5235124760076775,
+      "grad_norm": 0.4056020697430398,
+      "learning_rate": 2.7346249555634555e-05,
+      "loss": 0.4246,
+      "mean_token_accuracy": 0.8546995222568512,
+      "step": 3175
+    },
+    {
+      "epoch": 1.5259117082533589,
+      "grad_norm": 0.4289924980887499,
+      "learning_rate": 2.7301813011020267e-05,
+      "loss": 0.4002,
+      "mean_token_accuracy": 0.8619452834129333,
+      "step": 3180
+    },
+    {
+      "epoch": 1.5283109404990403,
+      "grad_norm": 0.4000591175003932,
+      "learning_rate": 2.7257376466405975e-05,
+      "loss": 0.4014,
+      "mean_token_accuracy": 0.8615616619586944,
+      "step": 3185
+    },
+    {
+      "epoch": 1.5307101727447217,
+      "grad_norm": 0.4298865662213465,
+      "learning_rate": 2.721293992179168e-05,
+      "loss": 0.4036,
+      "mean_token_accuracy": 0.8603943943977356,
+      "step": 3190
+    },
+    {
+      "epoch": 1.533109404990403,
+      "grad_norm": 0.4148912869232642,
+      "learning_rate": 2.716850337717739e-05,
+      "loss": 0.3952,
+      "mean_token_accuracy": 0.863661789894104,
+      "step": 3195
+    },
+    {
+      "epoch": 1.5355086372360844,
+      "grad_norm": 0.4005028765836154,
+      "learning_rate": 2.71240668325631e-05,
+      "loss": 0.4132,
+      "mean_token_accuracy": 0.8576488554477691,
+      "step": 3200
+    },
+    {
+      "epoch": 1.5379078694817658,
+      "grad_norm": 0.5863150189773588,
+      "learning_rate": 2.707963028794881e-05,
+      "loss": 0.4254,
+      "mean_token_accuracy": 0.855337244272232,
+      "step": 3205
+    },
+    {
+      "epoch": 1.5403071017274472,
+      "grad_norm": 0.4479332320371879,
+      "learning_rate": 2.7035193743334518e-05,
+      "loss": 0.4111,
+      "mean_token_accuracy": 0.8578731298446656,
+      "step": 3210
+    },
+    {
+      "epoch": 1.5427063339731286,
+      "grad_norm": 0.3975764092017198,
+      "learning_rate": 2.699075719872023e-05,
+      "loss": 0.4002,
+      "mean_token_accuracy": 0.8622437357902527,
+      "step": 3215
+    },
+    {
+      "epoch": 1.54510556621881,
+      "grad_norm": 0.40011161992288835,
+      "learning_rate": 2.6946320654105938e-05,
+      "loss": 0.4059,
+      "mean_token_accuracy": 0.8596208989620209,
+      "step": 3220
+    },
+    {
+      "epoch": 1.5475047984644914,
+      "grad_norm": 0.38016688564090934,
+      "learning_rate": 2.690188410949165e-05,
+      "loss": 0.3985,
+      "mean_token_accuracy": 0.8621473670005798,
+      "step": 3225
+    },
+    {
+      "epoch": 1.5499040307101728,
+      "grad_norm": 0.4138659075634796,
+      "learning_rate": 2.6857447564877358e-05,
+      "loss": 0.4058,
+      "mean_token_accuracy": 0.8596255898475647,
+      "step": 3230
+    },
+    {
+      "epoch": 1.5523032629558542,
+      "grad_norm": 0.422626771580196,
+      "learning_rate": 2.6813011020263067e-05,
+      "loss": 0.406,
+      "mean_token_accuracy": 0.8597724795341491,
+      "step": 3235
+    },
+    {
+      "epoch": 1.5547024952015356,
+      "grad_norm": 0.4170391560591557,
+      "learning_rate": 2.6768574475648772e-05,
+      "loss": 0.4004,
+      "mean_token_accuracy": 0.8617077708244324,
+      "step": 3240
+    },
+    {
+      "epoch": 1.557101727447217,
+      "grad_norm": 0.4275526743052804,
+      "learning_rate": 2.672413793103448e-05,
+      "loss": 0.3952,
+      "mean_token_accuracy": 0.8631601929664612,
+      "step": 3245
+    },
+    {
+      "epoch": 1.5595009596928984,
+      "grad_norm": 0.46752720154183036,
+      "learning_rate": 2.6679701386420192e-05,
+      "loss": 0.4044,
+      "mean_token_accuracy": 0.8601115763187408,
+      "step": 3250
+    },
+    {
+      "epoch": 1.5619001919385798,
+      "grad_norm": 0.48835763222260903,
+      "learning_rate": 2.66352648418059e-05,
+      "loss": 0.3944,
+      "mean_token_accuracy": 0.8636266410350799,
+      "step": 3255
+    },
+    {
+      "epoch": 1.5642994241842612,
+      "grad_norm": 0.4312067375707565,
+      "learning_rate": 2.6590828297191612e-05,
+      "loss": 0.4027,
+      "mean_token_accuracy": 0.8612389922142029,
+      "step": 3260
+    },
+    {
+      "epoch": 1.5666986564299425,
+      "grad_norm": 0.42201364947376946,
+      "learning_rate": 2.654639175257732e-05,
+      "loss": 0.4171,
+      "mean_token_accuracy": 0.8569965362548828,
+      "step": 3265
+    },
+    {
+      "epoch": 1.569097888675624,
+      "grad_norm": 0.41974298276995264,
+      "learning_rate": 2.650195520796303e-05,
+      "loss": 0.4067,
+      "mean_token_accuracy": 0.8597201347351074,
+      "step": 3270
+    },
+    {
+      "epoch": 1.5714971209213053,
+      "grad_norm": 0.39120429378823546,
+      "learning_rate": 2.645751866334874e-05,
+      "loss": 0.4077,
+      "mean_token_accuracy": 0.8595295011997223,
+      "step": 3275
+    },
+    {
+      "epoch": 1.5738963531669867,
+      "grad_norm": 0.3694200404366206,
+      "learning_rate": 2.641308211873445e-05,
+      "loss": 0.4008,
+      "mean_token_accuracy": 0.8617702662944794,
+      "step": 3280
+    },
+    {
+      "epoch": 1.576295585412668,
+      "grad_norm": 0.41337781650891703,
+      "learning_rate": 2.636864557412016e-05,
+      "loss": 0.4017,
+      "mean_token_accuracy": 0.8608327209949493,
+      "step": 3285
+    },
+    {
+      "epoch": 1.5786948176583493,
+      "grad_norm": 0.4072382286123216,
+      "learning_rate": 2.632420902950587e-05,
+      "loss": 0.3983,
+      "mean_token_accuracy": 0.862227326631546,
+      "step": 3290
+    },
+    {
+      "epoch": 1.5810940499040307,
+      "grad_norm": 0.4103556260194619,
+      "learning_rate": 2.6279772484891575e-05,
+      "loss": 0.3938,
+      "mean_token_accuracy": 0.8634500622749328,
+      "step": 3295
+    },
+    {
+      "epoch": 1.583493282149712,
+      "grad_norm": 0.38369554256372573,
+      "learning_rate": 2.6235335940277284e-05,
+      "loss": 0.3826,
+      "mean_token_accuracy": 0.8668561160564423,
+      "step": 3300
+    },
+    {
+      "epoch": 1.5858925143953935,
+      "grad_norm": 0.4153210125652736,
+      "learning_rate": 2.6190899395662992e-05,
+      "loss": 0.4074,
+      "mean_token_accuracy": 0.8587499022483825,
+      "step": 3305
+    },
+    {
+      "epoch": 1.5882917466410749,
+      "grad_norm": 0.3882161148344929,
+      "learning_rate": 2.6146462851048704e-05,
+      "loss": 0.405,
+      "mean_token_accuracy": 0.8604959726333619,
+      "step": 3310
+    },
+    {
+      "epoch": 1.5906909788867563,
+      "grad_norm": 0.41476299336957645,
+      "learning_rate": 2.6102026306434412e-05,
+      "loss": 0.4005,
+      "mean_token_accuracy": 0.8615757286548614,
+      "step": 3315
+    },
+    {
+      "epoch": 1.5930902111324377,
+      "grad_norm": 0.39628831496502565,
+      "learning_rate": 2.6057589761820124e-05,
+      "loss": 0.3976,
+      "mean_token_accuracy": 0.8621320128440857,
+      "step": 3320
+    },
+    {
+      "epoch": 1.595489443378119,
+      "grad_norm": 0.4286234719013056,
+      "learning_rate": 2.6013153217205833e-05,
+      "loss": 0.395,
+      "mean_token_accuracy": 0.8630506694316864,
+      "step": 3325
+    },
+    {
+      "epoch": 1.5978886756238004,
+      "grad_norm": 0.4238437642153525,
+      "learning_rate": 2.596871667259154e-05,
+      "loss": 0.4024,
+      "mean_token_accuracy": 0.8605756640434266,
+      "step": 3330
+    },
+    {
+      "epoch": 1.6002879078694816,
+      "grad_norm": 0.40038266554374713,
+      "learning_rate": 2.5924280127977253e-05,
+      "loss": 0.3903,
+      "mean_token_accuracy": 0.8641821384429932,
+      "step": 3335
+    },
+    {
+      "epoch": 1.602687140115163,
+      "grad_norm": 0.41971373829719233,
+      "learning_rate": 2.587984358336296e-05,
+      "loss": 0.4123,
+      "mean_token_accuracy": 0.8578661143779754,
+      "step": 3340
+    },
+    {
+      "epoch": 1.6050863723608444,
+      "grad_norm": 0.44399643165166497,
+      "learning_rate": 2.5835407038748667e-05,
+      "loss": 0.4126,
+      "mean_token_accuracy": 0.8577236354351043,
+      "step": 3345
+    },
+    {
+      "epoch": 1.6074856046065258,
+      "grad_norm": 0.39536435977240997,
+      "learning_rate": 2.5790970494134375e-05,
+      "loss": 0.4012,
+      "mean_token_accuracy": 0.8612061738967896,
+      "step": 3350
+    },
+    {
+      "epoch": 1.6098848368522072,
+      "grad_norm": 0.38863987975951847,
+      "learning_rate": 2.5746533949520087e-05,
+      "loss": 0.3984,
+      "mean_token_accuracy": 0.8619537353515625,
+      "step": 3355
+    },
+    {
+      "epoch": 1.6122840690978886,
+      "grad_norm": 0.39991351372949546,
+      "learning_rate": 2.5702097404905795e-05,
+      "loss": 0.3899,
+      "mean_token_accuracy": 0.8649181246757507,
+      "step": 3360
+    },
+    {
+      "epoch": 1.61468330134357,
+      "grad_norm": 0.398298780822706,
+      "learning_rate": 2.5657660860291504e-05,
+      "loss": 0.3999,
+      "mean_token_accuracy": 0.8612264811992645,
+      "step": 3365
+    },
+    {
+      "epoch": 1.6170825335892514,
+      "grad_norm": 0.4074770492939784,
+      "learning_rate": 2.5613224315677216e-05,
+      "loss": 0.4003,
+      "mean_token_accuracy": 0.8616210460662842,
+      "step": 3370
+    },
+    {
+      "epoch": 1.6194817658349328,
+      "grad_norm": 0.45214174453225614,
+      "learning_rate": 2.5568787771062924e-05,
+      "loss": 0.4167,
+      "mean_token_accuracy": 0.8581067621707916,
+      "step": 3375
+    },
+    {
+      "epoch": 1.6218809980806141,
+      "grad_norm": 0.4256452714311465,
+      "learning_rate": 2.5524351226448633e-05,
+      "loss": 0.4161,
+      "mean_token_accuracy": 0.8563941478729248,
+      "step": 3380
+    },
+    {
+      "epoch": 1.6242802303262955,
+      "grad_norm": 0.38160619565953624,
+      "learning_rate": 2.5479914681834344e-05,
+      "loss": 0.3894,
+      "mean_token_accuracy": 0.865361112356186,
+      "step": 3385
+    },
+    {
+      "epoch": 1.626679462571977,
+      "grad_norm": 0.3829043295701325,
+      "learning_rate": 2.5435478137220053e-05,
+      "loss": 0.4023,
+      "mean_token_accuracy": 0.8603506445884704,
+      "step": 3390
+    },
+    {
+      "epoch": 1.6290786948176583,
+      "grad_norm": 0.4223589108616364,
+      "learning_rate": 2.5391041592605758e-05,
+      "loss": 0.4002,
+      "mean_token_accuracy": 0.8612348198890686,
+      "step": 3395
+    },
+    {
+      "epoch": 1.6314779270633397,
+      "grad_norm": 0.4200284982519176,
+      "learning_rate": 2.5346605047991466e-05,
+      "loss": 0.3995,
+      "mean_token_accuracy": 0.8623109340667725,
+      "step": 3400
+    },
+    {
+      "epoch": 1.633877159309021,
+      "grad_norm": 0.4111996448465633,
+      "learning_rate": 2.5302168503377178e-05,
+      "loss": 0.4056,
+      "mean_token_accuracy": 0.8594146549701691,
+      "step": 3405
+    },
+    {
+      "epoch": 1.6362763915547025,
+      "grad_norm": 0.42908333059801745,
+      "learning_rate": 2.5257731958762887e-05,
+      "loss": 0.4044,
+      "mean_token_accuracy": 0.860127204656601,
+      "step": 3410
+    },
+    {
+      "epoch": 1.638675623800384,
+      "grad_norm": 0.42152377015778175,
+      "learning_rate": 2.5213295414148595e-05,
+      "loss": 0.3973,
+      "mean_token_accuracy": 0.8624828159809113,
+      "step": 3415
+    },
+    {
+      "epoch": 1.6410748560460653,
+      "grad_norm": 0.36981715731062414,
+      "learning_rate": 2.5168858869534307e-05,
+      "loss": 0.4041,
+      "mean_token_accuracy": 0.8604373574256897,
+      "step": 3420
+    },
+    {
+      "epoch": 1.6434740882917467,
+      "grad_norm": 0.394194627184814,
+      "learning_rate": 2.5124422324920015e-05,
+      "loss": 0.3878,
+      "mean_token_accuracy": 0.865257203578949,
+      "step": 3425
+    },
+    {
+      "epoch": 1.645873320537428,
+      "grad_norm": 0.39941467192632707,
+      "learning_rate": 2.5079985780305727e-05,
+      "loss": 0.3909,
+      "mean_token_accuracy": 0.8645266890525818,
+      "step": 3430
+    },
+    {
+      "epoch": 1.6482725527831095,
+      "grad_norm": 0.39808401212537314,
+      "learning_rate": 2.5035549235691436e-05,
+      "loss": 0.3944,
+      "mean_token_accuracy": 0.8638250768184662,
+      "step": 3435
+    },
+    {
+      "epoch": 1.6506717850287909,
+      "grad_norm": 0.4260497481390886,
+      "learning_rate": 2.4991112691077144e-05,
+      "loss": 0.4059,
+      "mean_token_accuracy": 0.8600857853889465,
+      "step": 3440
+    },
+    {
+      "epoch": 1.6530710172744723,
+      "grad_norm": 0.39232268275793675,
+      "learning_rate": 2.4946676146462853e-05,
+      "loss": 0.4011,
+      "mean_token_accuracy": 0.8610795974731446,
+      "step": 3445
+    },
+    {
+      "epoch": 1.6554702495201536,
+      "grad_norm": 0.3703252160647266,
+      "learning_rate": 2.490223960184856e-05,
+      "loss": 0.401,
+      "mean_token_accuracy": 0.8614936888217926,
+      "step": 3450
+    },
+    {
+      "epoch": 1.657869481765835,
+      "grad_norm": 0.38835790843763873,
+      "learning_rate": 2.4857803057234273e-05,
+      "loss": 0.3963,
+      "mean_token_accuracy": 0.8633289515972138,
+      "step": 3455
+    },
+    {
+      "epoch": 1.6602687140115164,
+      "grad_norm": 0.3792197463821151,
+      "learning_rate": 2.4813366512619978e-05,
+      "loss": 0.4005,
+      "mean_token_accuracy": 0.8607326924800873,
+      "step": 3460
+    },
+    {
+      "epoch": 1.6626679462571978,
+      "grad_norm": 0.4162862689767936,
+      "learning_rate": 2.476892996800569e-05,
+      "loss": 0.4097,
+      "mean_token_accuracy": 0.8588239908218384,
+      "step": 3465
+    },
+    {
+      "epoch": 1.6650671785028792,
+      "grad_norm": 0.4013668992861733,
+      "learning_rate": 2.47244934233914e-05,
+      "loss": 0.399,
+      "mean_token_accuracy": 0.8627789199352265,
+      "step": 3470
+    },
+    {
+      "epoch": 1.6674664107485606,
+      "grad_norm": 0.400314301487421,
+      "learning_rate": 2.4680056878777107e-05,
+      "loss": 0.4114,
+      "mean_token_accuracy": 0.8580051839351654,
+      "step": 3475
+    },
+    {
+      "epoch": 1.669865642994242,
+      "grad_norm": 0.3846055122002692,
+      "learning_rate": 2.463562033416282e-05,
+      "loss": 0.4023,
+      "mean_token_accuracy": 0.8605178356170654,
+      "step": 3480
+    },
+    {
+      "epoch": 1.6722648752399232,
+      "grad_norm": 0.4139839951325302,
+      "learning_rate": 2.4591183789548524e-05,
+      "loss": 0.4028,
+      "mean_token_accuracy": 0.8606350421905518,
+      "step": 3485
+    },
+    {
+      "epoch": 1.6746641074856046,
+      "grad_norm": 0.4053280087192033,
+      "learning_rate": 2.4546747244934236e-05,
+      "loss": 0.4203,
+      "mean_token_accuracy": 0.8548753023147583,
+      "step": 3490
+    },
+    {
+      "epoch": 1.677063339731286,
+      "grad_norm": 0.40101094344450555,
+      "learning_rate": 2.4502310700319944e-05,
+      "loss": 0.4026,
+      "mean_token_accuracy": 0.8610198259353637,
+      "step": 3495
+    },
+    {
+      "epoch": 1.6794625719769674,
+      "grad_norm": 0.42870363162924036,
+      "learning_rate": 2.4457874155705653e-05,
+      "loss": 0.4022,
+      "mean_token_accuracy": 0.8605069994926453,
+      "step": 3500
+    },
+    {
+      "epoch": 1.6818618042226487,
+      "grad_norm": 0.40157835655622814,
+      "learning_rate": 2.4413437611091364e-05,
+      "loss": 0.4045,
+      "mean_token_accuracy": 0.8600060880184174,
+      "step": 3505
+    },
+    {
+      "epoch": 1.6842610364683301,
+      "grad_norm": 0.49268372911579256,
+      "learning_rate": 2.436900106647707e-05,
+      "loss": 0.3982,
+      "mean_token_accuracy": 0.8626984477043151,
+      "step": 3510
+    },
+    {
+      "epoch": 1.6866602687140115,
+      "grad_norm": 0.4068686641753371,
+      "learning_rate": 2.432456452186278e-05,
+      "loss": 0.4064,
+      "mean_token_accuracy": 0.8597881138324738,
+      "step": 3515
+    },
+    {
+      "epoch": 1.689059500959693,
+      "grad_norm": 0.4204236115303968,
+      "learning_rate": 2.428012797724849e-05,
+      "loss": 0.4074,
+      "mean_token_accuracy": 0.8585520923137665,
+      "step": 3520
+    },
+    {
+      "epoch": 1.6914587332053743,
+      "grad_norm": 0.4231286443089349,
+      "learning_rate": 2.4235691432634198e-05,
+      "loss": 0.4112,
+      "mean_token_accuracy": 0.858112221956253,
+      "step": 3525
+    },
+    {
+      "epoch": 1.6938579654510557,
+      "grad_norm": 0.41874689386169434,
+      "learning_rate": 2.419125488801991e-05,
+      "loss": 0.3886,
+      "mean_token_accuracy": 0.8646687090396881,
+      "step": 3530
+    },
+    {
+      "epoch": 1.6962571976967369,
+      "grad_norm": 0.41289939460863173,
+      "learning_rate": 2.4146818343405615e-05,
+      "loss": 0.396,
+      "mean_token_accuracy": 0.8622913122177124,
+      "step": 3535
+    },
+    {
+      "epoch": 1.6986564299424183,
+      "grad_norm": 0.3700205825502841,
+      "learning_rate": 2.4102381798791327e-05,
+      "loss": 0.3928,
+      "mean_token_accuracy": 0.8635813176631928,
+      "step": 3540
+    },
+    {
+      "epoch": 1.7010556621880997,
+      "grad_norm": 0.40716705114833024,
+      "learning_rate": 2.4057945254177036e-05,
+      "loss": 0.3933,
+      "mean_token_accuracy": 0.863760232925415,
+      "step": 3545
+    },
+    {
+      "epoch": 1.703454894433781,
+      "grad_norm": 0.3969600924112917,
+      "learning_rate": 2.4013508709562747e-05,
+      "loss": 0.4017,
+      "mean_token_accuracy": 0.8610733449459076,
+      "step": 3550
+    },
+    {
+      "epoch": 1.7058541266794625,
+      "grad_norm": 0.40472357502298995,
+      "learning_rate": 2.3969072164948456e-05,
+      "loss": 0.3821,
+      "mean_token_accuracy": 0.8672096610069275,
+      "step": 3555
+    },
+    {
+      "epoch": 1.7082533589251438,
+      "grad_norm": 0.40035135249600823,
+      "learning_rate": 2.3924635620334164e-05,
+      "loss": 0.3981,
+      "mean_token_accuracy": 0.8622802495956421,
+      "step": 3560
+    },
+    {
+      "epoch": 1.7106525911708252,
+      "grad_norm": 0.4066364825773236,
+      "learning_rate": 2.3880199075719873e-05,
+      "loss": 0.4025,
+      "mean_token_accuracy": 0.8600334405899048,
+      "step": 3565
+    },
+    {
+      "epoch": 1.7130518234165066,
+      "grad_norm": 0.38018607577288216,
+      "learning_rate": 2.383576253110558e-05,
+      "loss": 0.4043,
+      "mean_token_accuracy": 0.8596990585327149,
+      "step": 3570
+    },
+    {
+      "epoch": 1.715451055662188,
+      "grad_norm": 0.42371739940852104,
+      "learning_rate": 2.3791325986491293e-05,
+      "loss": 0.3929,
+      "mean_token_accuracy": 0.8638055503368378,
+      "step": 3575
+    },
+    {
+      "epoch": 1.7178502879078694,
+      "grad_norm": 0.40407956177407894,
+      "learning_rate": 2.3746889441877e-05,
+      "loss": 0.386,
+      "mean_token_accuracy": 0.8652970552444458,
+      "step": 3580
+    },
+    {
+      "epoch": 1.7202495201535508,
+      "grad_norm": 0.3997203361126536,
+      "learning_rate": 2.370245289726271e-05,
+      "loss": 0.3971,
+      "mean_token_accuracy": 0.862407797574997,
+      "step": 3585
+    },
+    {
+      "epoch": 1.7226487523992322,
+      "grad_norm": 0.4177803154829972,
+      "learning_rate": 2.365801635264842e-05,
+      "loss": 0.3947,
+      "mean_token_accuracy": 0.8625789046287536,
+      "step": 3590
+    },
+    {
+      "epoch": 1.7250479846449136,
+      "grad_norm": 0.4182283380641395,
+      "learning_rate": 2.3613579808034127e-05,
+      "loss": 0.4049,
+      "mean_token_accuracy": 0.8597943723201752,
+      "step": 3595
+    },
+    {
+      "epoch": 1.727447216890595,
+      "grad_norm": 0.4010408090491947,
+      "learning_rate": 2.356914326341984e-05,
+      "loss": 0.386,
+      "mean_token_accuracy": 0.8659861564636231,
+      "step": 3600
+    },
+    {
+      "epoch": 1.7298464491362764,
+      "grad_norm": 0.3676558695389256,
+      "learning_rate": 2.3524706718805547e-05,
+      "loss": 0.3922,
+      "mean_token_accuracy": 0.8638743042945862,
+      "step": 3605
+    },
+    {
+      "epoch": 1.7322456813819578,
+      "grad_norm": 0.3902289844190933,
+      "learning_rate": 2.3480270174191256e-05,
+      "loss": 0.4022,
+      "mean_token_accuracy": 0.861260861158371,
+      "step": 3610
+    },
+    {
+      "epoch": 1.7346449136276392,
+      "grad_norm": 0.3997085585871422,
+      "learning_rate": 2.3435833629576964e-05,
+      "loss": 0.3978,
+      "mean_token_accuracy": 0.8619546592235565,
+      "step": 3615
+    },
+    {
+      "epoch": 1.7370441458733206,
+      "grad_norm": 0.3921537008472013,
+      "learning_rate": 2.3391397084962673e-05,
+      "loss": 0.3964,
+      "mean_token_accuracy": 0.8633320808410645,
+      "step": 3620
+    },
+    {
+      "epoch": 1.739443378119002,
+      "grad_norm": 0.4129934478129214,
+      "learning_rate": 2.3346960540348384e-05,
+      "loss": 0.3956,
+      "mean_token_accuracy": 0.8630765914916992,
+      "step": 3625
+    },
+    {
+      "epoch": 1.7418426103646834,
+      "grad_norm": 0.42221672710484526,
+      "learning_rate": 2.3302523995734093e-05,
+      "loss": 0.3999,
+      "mean_token_accuracy": 0.8610452234745025,
+      "step": 3630
+    },
+    {
+      "epoch": 1.7442418426103647,
+      "grad_norm": 0.38243888997259784,
+      "learning_rate": 2.3258087451119805e-05,
+      "loss": 0.3924,
+      "mean_token_accuracy": 0.8632264256477356,
+      "step": 3635
+    },
+    {
+      "epoch": 1.7466410748560461,
+      "grad_norm": 0.4191338371404725,
+      "learning_rate": 2.321365090650551e-05,
+      "loss": 0.4009,
+      "mean_token_accuracy": 0.8605326831340789,
+      "step": 3640
+    },
+    {
+      "epoch": 1.7490403071017275,
+      "grad_norm": 0.39265771608065647,
+      "learning_rate": 2.316921436189122e-05,
+      "loss": 0.3978,
+      "mean_token_accuracy": 0.8619015336036682,
+      "step": 3645
+    },
+    {
+      "epoch": 1.751439539347409,
+      "grad_norm": 1.3814131145219963,
+      "learning_rate": 2.312477781727693e-05,
+      "loss": 0.396,
+      "mean_token_accuracy": 0.8628222942352295,
+      "step": 3650
+    },
+    {
+      "epoch": 1.7538387715930903,
+      "grad_norm": 0.41013751416722777,
+      "learning_rate": 2.308034127266264e-05,
+      "loss": 0.3953,
+      "mean_token_accuracy": 0.8634633362293244,
+      "step": 3655
+    },
+    {
+      "epoch": 1.7562380038387717,
+      "grad_norm": 0.41357541362248335,
+      "learning_rate": 2.303590472804835e-05,
+      "loss": 0.3924,
+      "mean_token_accuracy": 0.8636547565460205,
+      "step": 3660
+    },
+    {
+      "epoch": 1.758637236084453,
+      "grad_norm": 0.39842235759316297,
+      "learning_rate": 2.2991468183434056e-05,
+      "loss": 0.4014,
+      "mean_token_accuracy": 0.8601654767990112,
+      "step": 3665
+    },
+    {
+      "epoch": 1.7610364683301345,
+      "grad_norm": 0.37838926308872844,
+      "learning_rate": 2.2947031638819767e-05,
+      "loss": 0.3843,
+      "mean_token_accuracy": 0.865840059518814,
+      "step": 3670
+    },
+    {
+      "epoch": 1.763435700575816,
+      "grad_norm": 0.3917883727663996,
+      "learning_rate": 2.2902595094205476e-05,
+      "loss": 0.3779,
+      "mean_token_accuracy": 0.8679964363574981,
+      "step": 3675
+    },
+    {
+      "epoch": 1.7658349328214973,
+      "grad_norm": 0.3898434426651832,
+      "learning_rate": 2.2858158549591184e-05,
+      "loss": 0.3997,
+      "mean_token_accuracy": 0.8615639984607697,
+      "step": 3680
+    },
+    {
+      "epoch": 1.7682341650671785,
+      "grad_norm": 0.38658220390707115,
+      "learning_rate": 2.2813722004976896e-05,
+      "loss": 0.3888,
+      "mean_token_accuracy": 0.8650773525238037,
+      "step": 3685
+    },
+    {
+      "epoch": 1.7706333973128598,
+      "grad_norm": 0.43035561137005984,
+      "learning_rate": 2.27692854603626e-05,
+      "loss": 0.4071,
+      "mean_token_accuracy": 0.859384971857071,
+      "step": 3690
+    },
+    {
+      "epoch": 1.7730326295585412,
+      "grad_norm": 0.38721362538552495,
+      "learning_rate": 2.2724848915748313e-05,
+      "loss": 0.3933,
+      "mean_token_accuracy": 0.863738352060318,
+      "step": 3695
+    },
+    {
+      "epoch": 1.7754318618042226,
+      "grad_norm": 0.4132259137622107,
+      "learning_rate": 2.268041237113402e-05,
+      "loss": 0.4064,
+      "mean_token_accuracy": 0.8590145349502564,
+      "step": 3700
+    },
+    {
+      "epoch": 1.777831094049904,
+      "grad_norm": 0.4299568515616119,
+      "learning_rate": 2.263597582651973e-05,
+      "loss": 0.3928,
+      "mean_token_accuracy": 0.863865715265274,
+      "step": 3705
+    },
+    {
+      "epoch": 1.7802303262955854,
+      "grad_norm": 2.9581425527056684,
+      "learning_rate": 2.2591539281905442e-05,
+      "loss": 0.3953,
+      "mean_token_accuracy": 0.8633383333683013,
+      "step": 3710
+    },
+    {
+      "epoch": 1.7826295585412668,
+      "grad_norm": 0.422515346351166,
+      "learning_rate": 2.2547102737291147e-05,
+      "loss": 0.4013,
+      "mean_token_accuracy": 0.8613436758518219,
+      "step": 3715
+    },
+    {
+      "epoch": 1.7850287907869482,
+      "grad_norm": 0.3982732965857703,
+      "learning_rate": 2.250266619267686e-05,
+      "loss": 0.3903,
+      "mean_token_accuracy": 0.8644165217876434,
+      "step": 3720
+    },
+    {
+      "epoch": 1.7874280230326296,
+      "grad_norm": 0.40233117601112617,
+      "learning_rate": 2.2458229648062567e-05,
+      "loss": 0.3915,
+      "mean_token_accuracy": 0.8637703895568848,
+      "step": 3725
+    },
+    {
+      "epoch": 1.789827255278311,
+      "grad_norm": 0.4025965574518156,
+      "learning_rate": 2.2413793103448276e-05,
+      "loss": 0.3855,
+      "mean_token_accuracy": 0.8658525586128235,
+      "step": 3730
+    },
+    {
+      "epoch": 1.7922264875239922,
+      "grad_norm": 0.45149891362450206,
+      "learning_rate": 2.2369356558833988e-05,
+      "loss": 0.4104,
+      "mean_token_accuracy": 0.8584145903587341,
+      "step": 3735
+    },
+    {
+      "epoch": 1.7946257197696736,
+      "grad_norm": 0.4391967468837121,
+      "learning_rate": 2.2324920014219693e-05,
+      "loss": 0.4004,
+      "mean_token_accuracy": 0.8613960266113281,
+      "step": 3740
+    },
+    {
+      "epoch": 1.797024952015355,
+      "grad_norm": 0.3780249057947582,
+      "learning_rate": 2.2280483469605405e-05,
+      "loss": 0.385,
+      "mean_token_accuracy": 0.8659556925296783,
+      "step": 3745
+    },
+    {
+      "epoch": 1.7994241842610363,
+      "grad_norm": 0.4164461266316941,
+      "learning_rate": 2.2236046924991113e-05,
+      "loss": 0.3908,
+      "mean_token_accuracy": 0.864171975851059,
+      "step": 3750
+    },
+    {
+      "epoch": 1.8018234165067177,
+      "grad_norm": 0.419648458986139,
+      "learning_rate": 2.2191610380376825e-05,
+      "loss": 0.402,
+      "mean_token_accuracy": 0.8604326784610749,
+      "step": 3755
+    },
+    {
+      "epoch": 1.8042226487523991,
+      "grad_norm": 0.40472603981661803,
+      "learning_rate": 2.2147173835762533e-05,
+      "loss": 0.3967,
+      "mean_token_accuracy": 0.8622312307357788,
+      "step": 3760
+    },
+    {
+      "epoch": 1.8066218809980805,
+      "grad_norm": 0.38537513710763316,
+      "learning_rate": 2.210273729114824e-05,
+      "loss": 0.3843,
+      "mean_token_accuracy": 0.8659494280815124,
+      "step": 3765
+    },
+    {
+      "epoch": 1.809021113243762,
+      "grad_norm": 0.40414961296090346,
+      "learning_rate": 2.205830074653395e-05,
+      "loss": 0.3896,
+      "mean_token_accuracy": 0.8649540543556213,
+      "step": 3770
+    },
+    {
+      "epoch": 1.8114203454894433,
+      "grad_norm": 0.40323221984828034,
+      "learning_rate": 2.201386420191966e-05,
+      "loss": 0.4042,
+      "mean_token_accuracy": 0.8601405024528503,
+      "step": 3775
+    },
+    {
+      "epoch": 1.8138195777351247,
+      "grad_norm": 0.398941675227608,
+      "learning_rate": 2.196942765730537e-05,
+      "loss": 0.4121,
+      "mean_token_accuracy": 0.8570371568202972,
+      "step": 3780
+    },
+    {
+      "epoch": 1.816218809980806,
+      "grad_norm": 0.37972257074705956,
+      "learning_rate": 2.192499111269108e-05,
+      "loss": 0.3927,
+      "mean_token_accuracy": 0.863745391368866,
+      "step": 3785
+    },
+    {
+      "epoch": 1.8186180422264875,
+      "grad_norm": 0.40411822515933343,
+      "learning_rate": 2.1880554568076787e-05,
+      "loss": 0.3853,
+      "mean_token_accuracy": 0.8656267642974853,
+      "step": 3790
+    },
+    {
+      "epoch": 1.8210172744721689,
+      "grad_norm": 0.3860353548447939,
+      "learning_rate": 2.1836118023462496e-05,
+      "loss": 0.4031,
+      "mean_token_accuracy": 0.8604975342750549,
+      "step": 3795
+    },
+    {
+      "epoch": 1.8234165067178503,
+      "grad_norm": 0.3993705247036203,
+      "learning_rate": 2.1791681478848204e-05,
+      "loss": 0.4001,
+      "mean_token_accuracy": 0.8609272360801696,
+      "step": 3800
+    },
+    {
+      "epoch": 1.8258157389635317,
+      "grad_norm": 0.4107616220909192,
+      "learning_rate": 2.1747244934233916e-05,
+      "loss": 0.3898,
+      "mean_token_accuracy": 0.8649306237697602,
+      "step": 3805
+    },
+    {
+      "epoch": 1.828214971209213,
+      "grad_norm": 0.42130544815400517,
+      "learning_rate": 2.1702808389619625e-05,
+      "loss": 0.3922,
+      "mean_token_accuracy": 0.8640716016292572,
+      "step": 3810
+    },
+    {
+      "epoch": 1.8306142034548945,
+      "grad_norm": 0.39983926601339675,
+      "learning_rate": 2.1658371845005333e-05,
+      "loss": 0.3876,
+      "mean_token_accuracy": 0.8645726799964905,
+      "step": 3815
+    },
+    {
+      "epoch": 1.8330134357005758,
+      "grad_norm": 0.43776571460399505,
+      "learning_rate": 2.161393530039104e-05,
+      "loss": 0.4095,
+      "mean_token_accuracy": 0.8584106743335724,
+      "step": 3820
+    },
+    {
+      "epoch": 1.8354126679462572,
+      "grad_norm": 0.398550236871868,
+      "learning_rate": 2.156949875577675e-05,
+      "loss": 0.3824,
+      "mean_token_accuracy": 0.8671510696411133,
+      "step": 3825
+    },
+    {
+      "epoch": 1.8378119001919386,
+      "grad_norm": 0.3929981521422064,
+      "learning_rate": 2.1525062211162462e-05,
+      "loss": 0.3909,
+      "mean_token_accuracy": 0.864390742778778,
+      "step": 3830
+    },
+    {
+      "epoch": 1.84021113243762,
+      "grad_norm": 0.40038946784997037,
+      "learning_rate": 2.148062566654817e-05,
+      "loss": 0.3879,
+      "mean_token_accuracy": 0.8649435222148896,
+      "step": 3835
+    },
+    {
+      "epoch": 1.8426103646833014,
+      "grad_norm": 0.4041167948196589,
+      "learning_rate": 2.1436189121933882e-05,
+      "loss": 0.3923,
+      "mean_token_accuracy": 0.863636976480484,
+      "step": 3840
+    },
+    {
+      "epoch": 1.8450095969289828,
+      "grad_norm": 0.40739126242654766,
+      "learning_rate": 2.1391752577319587e-05,
+      "loss": 0.3835,
+      "mean_token_accuracy": 0.8662947714328766,
+      "step": 3845
+    },
+    {
+      "epoch": 1.8474088291746642,
+      "grad_norm": 0.3818901707442913,
+      "learning_rate": 2.1347316032705296e-05,
+      "loss": 0.3974,
+      "mean_token_accuracy": 0.8623984336853028,
+      "step": 3850
+    },
+    {
+      "epoch": 1.8498080614203456,
+      "grad_norm": 0.4122122337260937,
+      "learning_rate": 2.1302879488091008e-05,
+      "loss": 0.3988,
+      "mean_token_accuracy": 0.8618648052215576,
+      "step": 3855
+    },
+    {
+      "epoch": 1.852207293666027,
+      "grad_norm": 0.3717712293387995,
+      "learning_rate": 2.1258442943476716e-05,
+      "loss": 0.3987,
+      "mean_token_accuracy": 0.8617163598537445,
+      "step": 3860
+    },
+    {
+      "epoch": 1.8546065259117084,
+      "grad_norm": 0.3947013818989466,
+      "learning_rate": 2.1214006398862428e-05,
+      "loss": 0.3855,
+      "mean_token_accuracy": 0.86544628739357,
+      "step": 3865
+    },
+    {
+      "epoch": 1.8570057581573898,
+      "grad_norm": 0.39602921518954387,
+      "learning_rate": 2.1169569854248133e-05,
+      "loss": 0.388,
+      "mean_token_accuracy": 0.8647743582725524,
+      "step": 3870
+    },
+    {
+      "epoch": 1.8594049904030712,
+      "grad_norm": 0.40195445490269366,
+      "learning_rate": 2.1125133309633845e-05,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8621335804462433,
+      "step": 3875
+    },
+    {
+      "epoch": 1.8618042226487526,
+      "grad_norm": 0.4440168571579167,
+      "learning_rate": 2.1080696765019553e-05,
+      "loss": 0.3855,
+      "mean_token_accuracy": 0.8659838140010834,
+      "step": 3880
+    },
+    {
+      "epoch": 1.8642034548944337,
+      "grad_norm": 0.3940893324231355,
+      "learning_rate": 2.1036260220405262e-05,
+      "loss": 0.3947,
+      "mean_token_accuracy": 0.8628605663776397,
+      "step": 3885
+    },
+    {
+      "epoch": 1.8666026871401151,
+      "grad_norm": 0.3930058952084085,
+      "learning_rate": 2.0991823675790974e-05,
+      "loss": 0.3828,
+      "mean_token_accuracy": 0.866622906923294,
+      "step": 3890
+    },
+    {
+      "epoch": 1.8690019193857965,
+      "grad_norm": 0.3749207822841214,
+      "learning_rate": 2.094738713117668e-05,
+      "loss": 0.3774,
+      "mean_token_accuracy": 0.8686128854751587,
+      "step": 3895
+    },
+    {
+      "epoch": 1.871401151631478,
+      "grad_norm": 0.3858339851657635,
+      "learning_rate": 2.090295058656239e-05,
+      "loss": 0.3888,
+      "mean_token_accuracy": 0.8645212113857269,
+      "step": 3900
+    },
+    {
+      "epoch": 1.8738003838771593,
+      "grad_norm": 0.3783986491624493,
+      "learning_rate": 2.08585140419481e-05,
+      "loss": 0.39,
+      "mean_token_accuracy": 0.8643775939941406,
+      "step": 3905
+    },
+    {
+      "epoch": 1.8761996161228407,
+      "grad_norm": 0.42063415379891794,
+      "learning_rate": 2.0814077497333808e-05,
+      "loss": 0.3908,
+      "mean_token_accuracy": 0.8639869809150695,
+      "step": 3910
+    },
+    {
+      "epoch": 1.878598848368522,
+      "grad_norm": 0.40149079913253105,
+      "learning_rate": 2.076964095271952e-05,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8618444919586181,
+      "step": 3915
+    },
+    {
+      "epoch": 1.8809980806142035,
+      "grad_norm": 0.39434822315885765,
+      "learning_rate": 2.0725204408105224e-05,
+      "loss": 0.3986,
+      "mean_token_accuracy": 0.8615225970745086,
+      "step": 3920
+    },
+    {
+      "epoch": 1.8833973128598849,
+      "grad_norm": 0.39080023400068203,
+      "learning_rate": 2.0680767863490936e-05,
+      "loss": 0.3931,
+      "mean_token_accuracy": 0.8635000646114349,
+      "step": 3925
+    },
+    {
+      "epoch": 1.8857965451055663,
+      "grad_norm": 0.41835303330036994,
+      "learning_rate": 2.0636331318876645e-05,
+      "loss": 0.3902,
+      "mean_token_accuracy": 0.8642172873020172,
+      "step": 3930
+    },
+    {
+      "epoch": 1.8881957773512474,
+      "grad_norm": 0.40460172529599076,
+      "learning_rate": 2.0591894774262353e-05,
+      "loss": 0.3838,
+      "mean_token_accuracy": 0.8665140569210052,
+      "step": 3935
+    },
+    {
+      "epoch": 1.8905950095969288,
+      "grad_norm": 0.40767639010951,
+      "learning_rate": 2.0547458229648065e-05,
+      "loss": 0.3914,
+      "mean_token_accuracy": 0.8642727673053742,
+      "step": 3940
+    },
+    {
+      "epoch": 1.8929942418426102,
+      "grad_norm": 0.3924246672999069,
+      "learning_rate": 2.050302168503377e-05,
+      "loss": 0.3876,
+      "mean_token_accuracy": 0.8654048681259155,
+      "step": 3945
+    },
+    {
+      "epoch": 1.8953934740882916,
+      "grad_norm": 0.39817152067892664,
+      "learning_rate": 2.0458585140419482e-05,
+      "loss": 0.4026,
+      "mean_token_accuracy": 0.8603006482124329,
+      "step": 3950
+    },
+    {
+      "epoch": 1.897792706333973,
+      "grad_norm": 0.40930635461715104,
+      "learning_rate": 2.041414859580519e-05,
+      "loss": 0.3962,
+      "mean_token_accuracy": 0.8623538970947265,
+      "step": 3955
+    },
+    {
+      "epoch": 1.9001919385796544,
+      "grad_norm": 0.39188090083012994,
+      "learning_rate": 2.03697120511909e-05,
+      "loss": 0.3873,
+      "mean_token_accuracy": 0.8655392467975617,
+      "step": 3960
+    },
+    {
+      "epoch": 1.9025911708253358,
+      "grad_norm": 0.3872694109948547,
+      "learning_rate": 2.032527550657661e-05,
+      "loss": 0.3803,
+      "mean_token_accuracy": 0.8677206397056579,
+      "step": 3965
+    },
+    {
+      "epoch": 1.9049904030710172,
+      "grad_norm": 0.38762477385935795,
+      "learning_rate": 2.028083896196232e-05,
+      "loss": 0.392,
+      "mean_token_accuracy": 0.8632125437259675,
+      "step": 3970
+    },
+    {
+      "epoch": 1.9073896353166986,
+      "grad_norm": 0.38142115013351746,
+      "learning_rate": 2.0236402417348028e-05,
+      "loss": 0.3987,
+      "mean_token_accuracy": 0.8616180002689362,
+      "step": 3975
+    },
+    {
+      "epoch": 1.90978886756238,
+      "grad_norm": 0.37033083914918996,
+      "learning_rate": 2.0191965872733736e-05,
+      "loss": 0.3943,
+      "mean_token_accuracy": 0.8636281967163086,
+      "step": 3980
+    },
+    {
+      "epoch": 1.9121880998080614,
+      "grad_norm": 0.3953126160258497,
+      "learning_rate": 2.0147529328119448e-05,
+      "loss": 0.3836,
+      "mean_token_accuracy": 0.8667151033878326,
+      "step": 3985
+    },
+    {
+      "epoch": 1.9145873320537428,
+      "grad_norm": 0.39082658638737106,
+      "learning_rate": 2.0103092783505157e-05,
+      "loss": 0.3824,
+      "mean_token_accuracy": 0.8666440069675445,
+      "step": 3990
+    },
+    {
+      "epoch": 1.9169865642994242,
+      "grad_norm": 0.4236031781517909,
+      "learning_rate": 2.0058656238890865e-05,
+      "loss": 0.4044,
+      "mean_token_accuracy": 0.8597177863121033,
+      "step": 3995
+    },
+    {
+      "epoch": 1.9193857965451055,
+      "grad_norm": 0.3914701005030144,
+      "learning_rate": 2.0014219694276573e-05,
+      "loss": 0.3926,
+      "mean_token_accuracy": 0.863865715265274,
+      "step": 4000
+    },
+    {
+      "epoch": 1.921785028790787,
+      "grad_norm": 0.4487506346824904,
+      "learning_rate": 1.9969783149662282e-05,
+      "loss": 0.391,
+      "mean_token_accuracy": 0.863703978061676,
+      "step": 4005
+    },
+    {
+      "epoch": 1.9241842610364683,
+      "grad_norm": 0.39004399723283967,
+      "learning_rate": 1.9925346605047994e-05,
+      "loss": 0.3873,
+      "mean_token_accuracy": 0.8656900286674499,
+      "step": 4010
+    },
+    {
+      "epoch": 1.9265834932821497,
+      "grad_norm": 0.41669329574673414,
+      "learning_rate": 1.9880910060433702e-05,
+      "loss": 0.4008,
+      "mean_token_accuracy": 0.8607615292072296,
+      "step": 4015
+    },
+    {
+      "epoch": 1.9289827255278311,
+      "grad_norm": 0.4374851172479612,
+      "learning_rate": 1.983647351581941e-05,
+      "loss": 0.3861,
+      "mean_token_accuracy": 0.8661767899990082,
+      "step": 4020
+    },
+    {
+      "epoch": 1.9313819577735125,
+      "grad_norm": 0.4107608827328522,
+      "learning_rate": 1.979203697120512e-05,
+      "loss": 0.3977,
+      "mean_token_accuracy": 0.8626461088657379,
+      "step": 4025
+    },
+    {
+      "epoch": 1.933781190019194,
+      "grad_norm": 0.3816761081097792,
+      "learning_rate": 1.9747600426590828e-05,
+      "loss": 0.3895,
+      "mean_token_accuracy": 0.8643610537052154,
+      "step": 4030
+    },
+    {
+      "epoch": 1.9361804222648753,
+      "grad_norm": 0.3900662876650821,
+      "learning_rate": 1.970316388197654e-05,
+      "loss": 0.3897,
+      "mean_token_accuracy": 0.8647509276866913,
+      "step": 4035
+    },
+    {
+      "epoch": 1.9385796545105567,
+      "grad_norm": 0.38289370485330754,
+      "learning_rate": 1.9658727337362248e-05,
+      "loss": 0.3727,
+      "mean_token_accuracy": 0.8701489210128784,
+      "step": 4040
+    },
+    {
+      "epoch": 1.940978886756238,
+      "grad_norm": 0.39717478282372676,
+      "learning_rate": 1.9614290792747956e-05,
+      "loss": 0.3975,
+      "mean_token_accuracy": 0.8616616666316986,
+      "step": 4045
+    },
+    {
+      "epoch": 1.9433781190019195,
+      "grad_norm": 0.48163705436697307,
+      "learning_rate": 1.9569854248133665e-05,
+      "loss": 0.3902,
+      "mean_token_accuracy": 0.8639610290527344,
+      "step": 4050
+    },
+    {
+      "epoch": 1.9457773512476009,
+      "grad_norm": 0.37838363527658575,
+      "learning_rate": 1.9525417703519373e-05,
+      "loss": 0.3884,
+      "mean_token_accuracy": 0.8645766973495483,
+      "step": 4055
+    },
+    {
+      "epoch": 1.9481765834932823,
+      "grad_norm": 0.38263168857995816,
+      "learning_rate": 1.9480981158905085e-05,
+      "loss": 0.3893,
+      "mean_token_accuracy": 0.8647235751152038,
+      "step": 4060
+    },
+    {
+      "epoch": 1.9505758157389637,
+      "grad_norm": 0.3930060747239377,
+      "learning_rate": 1.9436544614290794e-05,
+      "loss": 0.3869,
+      "mean_token_accuracy": 0.8648993670940399,
+      "step": 4065
+    },
+    {
+      "epoch": 1.952975047984645,
+      "grad_norm": 0.39704337841199966,
+      "learning_rate": 1.9392108069676505e-05,
+      "loss": 0.4068,
+      "mean_token_accuracy": 0.8590138375759124,
+      "step": 4070
+    },
+    {
+      "epoch": 1.9553742802303264,
+      "grad_norm": 0.3879254475886731,
+      "learning_rate": 1.934767152506221e-05,
+      "loss": 0.3883,
+      "mean_token_accuracy": 0.8654634654521942,
+      "step": 4075
+    },
+    {
+      "epoch": 1.9577735124760078,
+      "grad_norm": 0.38844784689750755,
+      "learning_rate": 1.930323498044792e-05,
+      "loss": 0.3798,
+      "mean_token_accuracy": 0.8672338962554932,
+      "step": 4080
+    },
+    {
+      "epoch": 1.960172744721689,
+      "grad_norm": 0.37446283650301254,
+      "learning_rate": 1.925879843583363e-05,
+      "loss": 0.3788,
+      "mean_token_accuracy": 0.867630785703659,
+      "step": 4085
+    },
+    {
+      "epoch": 1.9625719769673704,
+      "grad_norm": 0.4861745002358442,
+      "learning_rate": 1.921436189121934e-05,
+      "loss": 0.3884,
+      "mean_token_accuracy": 0.8643026351928711,
+      "step": 4090
+    },
+    {
+      "epoch": 1.9649712092130518,
+      "grad_norm": 0.46859343821993776,
+      "learning_rate": 1.916992534660505e-05,
+      "loss": 0.3938,
+      "mean_token_accuracy": 0.8629789352416992,
+      "step": 4095
+    },
+    {
+      "epoch": 1.9673704414587332,
+      "grad_norm": 0.3704611655318632,
+      "learning_rate": 1.9125488801990756e-05,
+      "loss": 0.3926,
+      "mean_token_accuracy": 0.8637086689472199,
+      "step": 4100
+    },
+    {
+      "epoch": 1.9697696737044146,
+      "grad_norm": 0.39545534380900776,
+      "learning_rate": 1.9081052257376468e-05,
+      "loss": 0.3893,
+      "mean_token_accuracy": 0.8637082576751709,
+      "step": 4105
+    },
+    {
+      "epoch": 1.972168905950096,
+      "grad_norm": 0.4112546795437855,
+      "learning_rate": 1.9036615712762177e-05,
+      "loss": 0.3933,
+      "mean_token_accuracy": 0.8629828453063965,
+      "step": 4110
+    },
+    {
+      "epoch": 1.9745681381957774,
+      "grad_norm": 0.3945209577425338,
+      "learning_rate": 1.8992179168147885e-05,
+      "loss": 0.3889,
+      "mean_token_accuracy": 0.8651759505271912,
+      "step": 4115
+    },
+    {
+      "epoch": 1.9769673704414588,
+      "grad_norm": 0.4004234809978467,
+      "learning_rate": 1.8947742623533597e-05,
+      "loss": 0.3943,
+      "mean_token_accuracy": 0.8631406724452972,
+      "step": 4120
+    },
+    {
+      "epoch": 1.9793666026871402,
+      "grad_norm": 0.4002978488761072,
+      "learning_rate": 1.8903306078919302e-05,
+      "loss": 0.4068,
+      "mean_token_accuracy": 0.8592036962509155,
+      "step": 4125
+    },
+    {
+      "epoch": 1.9817658349328215,
+      "grad_norm": 0.42559417968429664,
+      "learning_rate": 1.8858869534305014e-05,
+      "loss": 0.3995,
+      "mean_token_accuracy": 0.8614804029464722,
+      "step": 4130
+    },
+    {
+      "epoch": 1.9841650671785027,
+      "grad_norm": 0.4441111127890975,
+      "learning_rate": 1.8814432989690722e-05,
+      "loss": 0.3811,
+      "mean_token_accuracy": 0.8669618248939515,
+      "step": 4135
+    },
+    {
+      "epoch": 1.986564299424184,
+      "grad_norm": 0.37759355410482454,
+      "learning_rate": 1.876999644507643e-05,
+      "loss": 0.3955,
+      "mean_token_accuracy": 0.8630719184875488,
+      "step": 4140
+    },
+    {
+      "epoch": 1.9889635316698655,
+      "grad_norm": 0.38851753327004795,
+      "learning_rate": 1.8725559900462143e-05,
+      "loss": 0.3895,
+      "mean_token_accuracy": 0.8648181080818176,
+      "step": 4145
+    },
+    {
+      "epoch": 1.991362763915547,
+      "grad_norm": 0.39128512330716314,
+      "learning_rate": 1.8681123355847848e-05,
+      "loss": 0.3873,
+      "mean_token_accuracy": 0.8651157855987549,
+      "step": 4150
+    },
+    {
+      "epoch": 1.9937619961612283,
+      "grad_norm": 0.36186928153286846,
+      "learning_rate": 1.863668681123356e-05,
+      "loss": 0.3863,
+      "mean_token_accuracy": 0.8653822124004364,
+      "step": 4155
+    },
+    {
+      "epoch": 1.9961612284069097,
+      "grad_norm": 0.3955538160916505,
+      "learning_rate": 1.8592250266619268e-05,
+      "loss": 0.392,
+      "mean_token_accuracy": 0.8637274146080017,
+      "step": 4160
+    },
+    {
+      "epoch": 1.998560460652591,
+      "grad_norm": 0.43379145511335016,
+      "learning_rate": 1.8547813722004976e-05,
+      "loss": 0.3956,
+      "mean_token_accuracy": 0.8617413640022278,
+      "step": 4165
+    },
+    {
+      "epoch": 2.0009596928982725,
+      "grad_norm": 0.4990607339389488,
+      "learning_rate": 1.8503377177390688e-05,
+      "loss": 0.3482,
+      "mean_token_accuracy": 0.877407294511795,
+      "step": 4170
+    },
+    {
+      "epoch": 2.003358925143954,
+      "grad_norm": 0.48774642605648677,
+      "learning_rate": 1.8458940632776397e-05,
+      "loss": 0.2789,
+      "mean_token_accuracy": 0.8987186670303344,
+      "step": 4175
+    },
+    {
+      "epoch": 2.0057581573896353,
+      "grad_norm": 0.550368833299061,
+      "learning_rate": 1.8414504088162105e-05,
+      "loss": 0.2838,
+      "mean_token_accuracy": 0.8964677393436432,
+      "step": 4180
+    },
+    {
+      "epoch": 2.0081573896353166,
+      "grad_norm": 0.448658045217398,
+      "learning_rate": 1.8370067543547814e-05,
+      "loss": 0.2717,
+      "mean_token_accuracy": 0.9005320727825165,
+      "step": 4185
+    },
+    {
+      "epoch": 2.010556621880998,
+      "grad_norm": 0.4571573770888426,
+      "learning_rate": 1.8325630998933526e-05,
+      "loss": 0.2718,
+      "mean_token_accuracy": 0.9007344186306,
+      "step": 4190
+    },
+    {
+      "epoch": 2.0129558541266794,
+      "grad_norm": 0.4566936280537468,
+      "learning_rate": 1.8281194454319234e-05,
+      "loss": 0.2681,
+      "mean_token_accuracy": 0.9025704741477967,
+      "step": 4195
+    },
+    {
+      "epoch": 2.015355086372361,
+      "grad_norm": 0.5002619274348096,
+      "learning_rate": 1.8236757909704942e-05,
+      "loss": 0.2775,
+      "mean_token_accuracy": 0.8995882570743561,
+      "step": 4200
+    },
+    {
+      "epoch": 2.017754318618042,
+      "grad_norm": 0.45520437909491956,
+      "learning_rate": 1.819232136509065e-05,
+      "loss": 0.2545,
+      "mean_token_accuracy": 0.9068623960018158,
+      "step": 4205
+    },
+    {
+      "epoch": 2.0201535508637236,
+      "grad_norm": 0.44638235220575284,
+      "learning_rate": 1.814788482047636e-05,
+      "loss": 0.2755,
+      "mean_token_accuracy": 0.8995202898979187,
+      "step": 4210
+    },
+    {
+      "epoch": 2.022552783109405,
+      "grad_norm": 0.4321368591508329,
+      "learning_rate": 1.810344827586207e-05,
+      "loss": 0.2634,
+      "mean_token_accuracy": 0.9036517858505249,
+      "step": 4215
+    },
+    {
+      "epoch": 2.0249520153550864,
+      "grad_norm": 0.49384643713338167,
+      "learning_rate": 1.805901173124778e-05,
+      "loss": 0.283,
+      "mean_token_accuracy": 0.8976084411144256,
+      "step": 4220
+    },
+    {
+      "epoch": 2.027351247600768,
+      "grad_norm": 0.46123013151789993,
+      "learning_rate": 1.8014575186633488e-05,
+      "loss": 0.274,
+      "mean_token_accuracy": 0.9010094285011292,
+      "step": 4225
+    },
+    {
+      "epoch": 2.029750479846449,
+      "grad_norm": 0.46138647274505384,
+      "learning_rate": 1.7970138642019197e-05,
+      "loss": 0.2712,
+      "mean_token_accuracy": 0.9008758366107941,
+      "step": 4230
+    },
+    {
+      "epoch": 2.0321497120921306,
+      "grad_norm": 1.786349693464167,
+      "learning_rate": 1.7925702097404905e-05,
+      "loss": 0.2801,
+      "mean_token_accuracy": 0.9000257790088654,
+      "step": 4235
+    },
+    {
+      "epoch": 2.034548944337812,
+      "grad_norm": 0.44348259045087873,
+      "learning_rate": 1.7881265552790617e-05,
+      "loss": 0.2786,
+      "mean_token_accuracy": 0.8985202252864838,
+      "step": 4240
+    },
+    {
+      "epoch": 2.0369481765834934,
+      "grad_norm": 0.5000759402358198,
+      "learning_rate": 1.7836829008176325e-05,
+      "loss": 0.2858,
+      "mean_token_accuracy": 0.8969248056411743,
+      "step": 4245
+    },
+    {
+      "epoch": 2.0393474088291748,
+      "grad_norm": 0.448800907729344,
+      "learning_rate": 1.7792392463562034e-05,
+      "loss": 0.2738,
+      "mean_token_accuracy": 0.9001585960388183,
+      "step": 4250
+    },
+    {
+      "epoch": 2.041746641074856,
+      "grad_norm": 0.4571099322615939,
+      "learning_rate": 1.7747955918947742e-05,
+      "loss": 0.2701,
+      "mean_token_accuracy": 0.9012974858283996,
+      "step": 4255
+    },
+    {
+      "epoch": 2.0441458733205375,
+      "grad_norm": 0.4642114394657761,
+      "learning_rate": 1.770351937433345e-05,
+      "loss": 0.2672,
+      "mean_token_accuracy": 0.9021126329898834,
+      "step": 4260
+    },
+    {
+      "epoch": 2.046545105566219,
+      "grad_norm": 0.49976082645529357,
+      "learning_rate": 1.7659082829719163e-05,
+      "loss": 0.2798,
+      "mean_token_accuracy": 0.8982514500617981,
+      "step": 4265
+    },
+    {
+      "epoch": 2.0489443378119003,
+      "grad_norm": 0.4379654371585255,
+      "learning_rate": 1.761464628510487e-05,
+      "loss": 0.2668,
+      "mean_token_accuracy": 0.9023063957691193,
+      "step": 4270
+    },
+    {
+      "epoch": 2.0513435700575817,
+      "grad_norm": 0.5045991197144502,
+      "learning_rate": 1.7570209740490583e-05,
+      "loss": 0.2753,
+      "mean_token_accuracy": 0.8996968567371368,
+      "step": 4275
+    },
+    {
+      "epoch": 2.053742802303263,
+      "grad_norm": 0.4561539463036029,
+      "learning_rate": 1.7525773195876288e-05,
+      "loss": 0.2568,
+      "mean_token_accuracy": 0.906041008234024,
+      "step": 4280
+    },
+    {
+      "epoch": 2.0561420345489445,
+      "grad_norm": 0.5058292980069206,
+      "learning_rate": 1.7481336651261996e-05,
+      "loss": 0.2723,
+      "mean_token_accuracy": 0.9003937780857086,
+      "step": 4285
+    },
+    {
+      "epoch": 2.058541266794626,
+      "grad_norm": 0.4372928580482913,
+      "learning_rate": 1.743690010664771e-05,
+      "loss": 0.2803,
+      "mean_token_accuracy": 0.8972904622554779,
+      "step": 4290
+    },
+    {
+      "epoch": 2.0609404990403073,
+      "grad_norm": 0.4757830344041431,
+      "learning_rate": 1.7392463562033417e-05,
+      "loss": 0.2626,
+      "mean_token_accuracy": 0.9038049280643463,
+      "step": 4295
+    },
+    {
+      "epoch": 2.0633397312859887,
+      "grad_norm": 0.46326509822644846,
+      "learning_rate": 1.734802701741913e-05,
+      "loss": 0.2713,
+      "mean_token_accuracy": 0.9007812976837158,
+      "step": 4300
+    },
+    {
+      "epoch": 2.06573896353167,
+      "grad_norm": 0.43282465527513386,
+      "learning_rate": 1.7303590472804834e-05,
+      "loss": 0.2618,
+      "mean_token_accuracy": 0.9037814855575561,
+      "step": 4305
+    },
+    {
+      "epoch": 2.068138195777351,
+      "grad_norm": 0.47292568561468,
+      "learning_rate": 1.7259153928190546e-05,
+      "loss": 0.2741,
+      "mean_token_accuracy": 0.8997789025306702,
+      "step": 4310
+    },
+    {
+      "epoch": 2.0705374280230324,
+      "grad_norm": 0.47569236672519766,
+      "learning_rate": 1.7214717383576254e-05,
+      "loss": 0.2634,
+      "mean_token_accuracy": 0.9033852636814117,
+      "step": 4315
+    },
+    {
+      "epoch": 2.072936660268714,
+      "grad_norm": 0.48951867821674516,
+      "learning_rate": 1.7170280838961963e-05,
+      "loss": 0.2733,
+      "mean_token_accuracy": 0.9004883110523224,
+      "step": 4320
+    },
+    {
+      "epoch": 2.075335892514395,
+      "grad_norm": 0.473035784375584,
+      "learning_rate": 1.7125844294347674e-05,
+      "loss": 0.2699,
+      "mean_token_accuracy": 0.9013882339000702,
+      "step": 4325
+    },
+    {
+      "epoch": 2.0777351247600766,
+      "grad_norm": 0.46053056456143077,
+      "learning_rate": 1.708140774973338e-05,
+      "loss": 0.2544,
+      "mean_token_accuracy": 0.9071746647357941,
+      "step": 4330
+    },
+    {
+      "epoch": 2.080134357005758,
+      "grad_norm": 0.42859421990106655,
+      "learning_rate": 1.703697120511909e-05,
+      "loss": 0.2674,
+      "mean_token_accuracy": 0.9021602928638458,
+      "step": 4335
+    },
+    {
+      "epoch": 2.0825335892514394,
+      "grad_norm": 0.49007953604178717,
+      "learning_rate": 1.69925346605048e-05,
+      "loss": 0.2707,
+      "mean_token_accuracy": 0.900754737854004,
+      "step": 4340
+    },
+    {
+      "epoch": 2.084932821497121,
+      "grad_norm": 4.37790219062138,
+      "learning_rate": 1.6948098115890508e-05,
+      "loss": 0.2883,
+      "mean_token_accuracy": 0.8987155377864837,
+      "step": 4345
+    },
+    {
+      "epoch": 2.087332053742802,
+      "grad_norm": 0.49261366427321174,
+      "learning_rate": 1.690366157127622e-05,
+      "loss": 0.2783,
+      "mean_token_accuracy": 0.8987280428409576,
+      "step": 4350
+    },
+    {
+      "epoch": 2.0897312859884836,
+      "grad_norm": 0.42896597289174915,
+      "learning_rate": 1.685922502666193e-05,
+      "loss": 0.2713,
+      "mean_token_accuracy": 0.900896155834198,
+      "step": 4355
+    },
+    {
+      "epoch": 2.092130518234165,
+      "grad_norm": 0.45450827173150554,
+      "learning_rate": 1.6814788482047637e-05,
+      "loss": 0.273,
+      "mean_token_accuracy": 0.9007102012634277,
+      "step": 4360
+    },
+    {
+      "epoch": 2.0945297504798464,
+      "grad_norm": 0.4522864152085452,
+      "learning_rate": 1.6770351937433345e-05,
+      "loss": 0.2648,
+      "mean_token_accuracy": 0.9030673801898956,
+      "step": 4365
+    },
+    {
+      "epoch": 2.0969289827255277,
+      "grad_norm": 0.4633404403314624,
+      "learning_rate": 1.6725915392819054e-05,
+      "loss": 0.2781,
+      "mean_token_accuracy": 0.898474907875061,
+      "step": 4370
+    },
+    {
+      "epoch": 2.099328214971209,
+      "grad_norm": 0.46662108591262225,
+      "learning_rate": 1.6681478848204766e-05,
+      "loss": 0.2832,
+      "mean_token_accuracy": 0.8965177476406098,
+      "step": 4375
+    },
+    {
+      "epoch": 2.1017274472168905,
+      "grad_norm": 0.44562361030276315,
+      "learning_rate": 1.6637042303590474e-05,
+      "loss": 0.2681,
+      "mean_token_accuracy": 0.9022032618522644,
+      "step": 4380
+    },
+    {
+      "epoch": 2.104126679462572,
+      "grad_norm": 0.44918481180735653,
+      "learning_rate": 1.6592605758976183e-05,
+      "loss": 0.2642,
+      "mean_token_accuracy": 0.9034236431121826,
+      "step": 4385
+    },
+    {
+      "epoch": 2.1065259117082533,
+      "grad_norm": 0.4649130093432506,
+      "learning_rate": 1.654816921436189e-05,
+      "loss": 0.2687,
+      "mean_token_accuracy": 0.9015680313110351,
+      "step": 4390
+    },
+    {
+      "epoch": 2.1089251439539347,
+      "grad_norm": 0.47144552458086414,
+      "learning_rate": 1.6503732669747603e-05,
+      "loss": 0.2643,
+      "mean_token_accuracy": 0.9031572341918945,
+      "step": 4395
+    },
+    {
+      "epoch": 2.111324376199616,
+      "grad_norm": 0.45486321477616076,
+      "learning_rate": 1.645929612513331e-05,
+      "loss": 0.2732,
+      "mean_token_accuracy": 0.900248795747757,
+      "step": 4400
+    },
+    {
+      "epoch": 2.1137236084452975,
+      "grad_norm": 0.4759359634653368,
+      "learning_rate": 1.641485958051902e-05,
+      "loss": 0.2682,
+      "mean_token_accuracy": 0.9020274698734283,
+      "step": 4405
+    },
+    {
+      "epoch": 2.116122840690979,
+      "grad_norm": 0.4720824801877516,
+      "learning_rate": 1.637042303590473e-05,
+      "loss": 0.2701,
+      "mean_token_accuracy": 0.9012477397918701,
+      "step": 4410
+    },
+    {
+      "epoch": 2.1185220729366603,
+      "grad_norm": 0.45178559782138805,
+      "learning_rate": 1.6325986491290437e-05,
+      "loss": 0.2653,
+      "mean_token_accuracy": 0.9030517518520356,
+      "step": 4415
+    },
+    {
+      "epoch": 2.1209213051823417,
+      "grad_norm": 0.4477863492661275,
+      "learning_rate": 1.628154994667615e-05,
+      "loss": 0.2702,
+      "mean_token_accuracy": 0.9010055363178253,
+      "step": 4420
+    },
+    {
+      "epoch": 2.123320537428023,
+      "grad_norm": 0.4617479446804104,
+      "learning_rate": 1.6237113402061857e-05,
+      "loss": 0.2678,
+      "mean_token_accuracy": 0.9017586946487427,
+      "step": 4425
+    },
+    {
+      "epoch": 2.1257197696737045,
+      "grad_norm": 0.468851911945899,
+      "learning_rate": 1.6192676857447566e-05,
+      "loss": 0.2687,
+      "mean_token_accuracy": 0.9018571496009826,
+      "step": 4430
+    },
+    {
+      "epoch": 2.128119001919386,
+      "grad_norm": 0.4565575005949691,
+      "learning_rate": 1.6148240312833274e-05,
+      "loss": 0.2615,
+      "mean_token_accuracy": 0.9041065096855163,
+      "step": 4435
+    },
+    {
+      "epoch": 2.1305182341650672,
+      "grad_norm": 0.46814377259524864,
+      "learning_rate": 1.6103803768218983e-05,
+      "loss": 0.2743,
+      "mean_token_accuracy": 0.899460905790329,
+      "step": 4440
+    },
+    {
+      "epoch": 2.1329174664107486,
+      "grad_norm": 0.48726206079260836,
+      "learning_rate": 1.6059367223604694e-05,
+      "loss": 0.2626,
+      "mean_token_accuracy": 0.9039096117019654,
+      "step": 4445
+    },
+    {
+      "epoch": 2.13531669865643,
+      "grad_norm": 0.4420924963462651,
+      "learning_rate": 1.6014930678990403e-05,
+      "loss": 0.2765,
+      "mean_token_accuracy": 0.8989108800888062,
+      "step": 4450
+    },
+    {
+      "epoch": 2.1377159309021114,
+      "grad_norm": 0.4665040262296594,
+      "learning_rate": 1.597049413437611e-05,
+      "loss": 0.2746,
+      "mean_token_accuracy": 0.9000250041484833,
+      "step": 4455
+    },
+    {
+      "epoch": 2.140115163147793,
+      "grad_norm": 0.4106007778508939,
+      "learning_rate": 1.592605758976182e-05,
+      "loss": 0.2679,
+      "mean_token_accuracy": 0.9022345006465912,
+      "step": 4460
+    },
+    {
+      "epoch": 2.142514395393474,
+      "grad_norm": 0.4861737684759091,
+      "learning_rate": 1.5881621045147528e-05,
+      "loss": 0.2821,
+      "mean_token_accuracy": 0.8968492388725281,
+      "step": 4465
+    },
+    {
+      "epoch": 2.1449136276391556,
+      "grad_norm": 0.4714941781640915,
+      "learning_rate": 1.583718450053324e-05,
+      "loss": 0.2724,
+      "mean_token_accuracy": 0.9004875302314759,
+      "step": 4470
+    },
+    {
+      "epoch": 2.147312859884837,
+      "grad_norm": 0.47685731170785883,
+      "learning_rate": 1.579274795591895e-05,
+      "loss": 0.2658,
+      "mean_token_accuracy": 0.9028298735618592,
+      "step": 4475
+    },
+    {
+      "epoch": 2.1497120921305184,
+      "grad_norm": 0.48610023240176986,
+      "learning_rate": 1.5748311411304657e-05,
+      "loss": 0.272,
+      "mean_token_accuracy": 0.9006937980651856,
+      "step": 4480
+    },
+    {
+      "epoch": 2.1521113243762,
+      "grad_norm": 0.4710164843150768,
+      "learning_rate": 1.5703874866690366e-05,
+      "loss": 0.2683,
+      "mean_token_accuracy": 0.9017013192176819,
+      "step": 4485
+    },
+    {
+      "epoch": 2.154510556621881,
+      "grad_norm": 0.47511883246964265,
+      "learning_rate": 1.5659438322076074e-05,
+      "loss": 0.2725,
+      "mean_token_accuracy": 0.9000279366970062,
+      "step": 4490
+    },
+    {
+      "epoch": 2.1569097888675626,
+      "grad_norm": 0.4775014098408969,
+      "learning_rate": 1.5615001777461786e-05,
+      "loss": 0.2827,
+      "mean_token_accuracy": 0.896709942817688,
+      "step": 4495
+    },
+    {
+      "epoch": 2.159309021113244,
+      "grad_norm": 0.43665975325988027,
+      "learning_rate": 1.5570565232847494e-05,
+      "loss": 0.2755,
+      "mean_token_accuracy": 0.8993097960948944,
+      "step": 4500
+    },
+    {
+      "epoch": 2.161708253358925,
+      "grad_norm": 0.44223033656031446,
+      "learning_rate": 1.5526128688233206e-05,
+      "loss": 0.2651,
+      "mean_token_accuracy": 0.9031178951263428,
+      "step": 4505
+    },
+    {
+      "epoch": 2.1641074856046068,
+      "grad_norm": 0.4646989173708562,
+      "learning_rate": 1.548169214361891e-05,
+      "loss": 0.2657,
+      "mean_token_accuracy": 0.9026728272438049,
+      "step": 4510
+    },
+    {
+      "epoch": 2.1665067178502877,
+      "grad_norm": 0.45611006296440343,
+      "learning_rate": 1.543725559900462e-05,
+      "loss": 0.2733,
+      "mean_token_accuracy": 0.9000039041042328,
+      "step": 4515
+    },
+    {
+      "epoch": 2.168905950095969,
+      "grad_norm": 0.4910530067649774,
+      "learning_rate": 1.539281905439033e-05,
+      "loss": 0.2728,
+      "mean_token_accuracy": 0.9004085659980774,
+      "step": 4520
+    },
+    {
+      "epoch": 2.1713051823416505,
+      "grad_norm": 0.4902817246313605,
+      "learning_rate": 1.534838250977604e-05,
+      "loss": 0.2701,
+      "mean_token_accuracy": 0.9008141100406647,
+      "step": 4525
+    },
+    {
+      "epoch": 2.173704414587332,
+      "grad_norm": 0.4519098840938565,
+      "learning_rate": 1.5303945965161752e-05,
+      "loss": 0.275,
+      "mean_token_accuracy": 0.899957025051117,
+      "step": 4530
+    },
+    {
+      "epoch": 2.1761036468330133,
+      "grad_norm": 0.48915278087110464,
+      "learning_rate": 1.5259509420547457e-05,
+      "loss": 0.2657,
+      "mean_token_accuracy": 0.9028978407382965,
+      "step": 4535
+    },
+    {
+      "epoch": 2.1785028790786947,
+      "grad_norm": 0.45723186309703706,
+      "learning_rate": 1.5215072875933167e-05,
+      "loss": 0.2718,
+      "mean_token_accuracy": 0.9006772577762604,
+      "step": 4540
+    },
+    {
+      "epoch": 2.180902111324376,
+      "grad_norm": 0.4353868793708082,
+      "learning_rate": 1.5170636331318877e-05,
+      "loss": 0.2694,
+      "mean_token_accuracy": 0.9016266703605652,
+      "step": 4545
+    },
+    {
+      "epoch": 2.1833013435700575,
+      "grad_norm": 0.4600745274735988,
+      "learning_rate": 1.5126199786704587e-05,
+      "loss": 0.2704,
+      "mean_token_accuracy": 0.901211804151535,
+      "step": 4550
+    },
+    {
+      "epoch": 2.185700575815739,
+      "grad_norm": 0.46905406216354795,
+      "learning_rate": 1.5081763242090296e-05,
+      "loss": 0.2721,
+      "mean_token_accuracy": 0.900727391242981,
+      "step": 4555
+    },
+    {
+      "epoch": 2.1880998080614202,
+      "grad_norm": 0.46220184604712333,
+      "learning_rate": 1.5037326697476006e-05,
+      "loss": 0.2727,
+      "mean_token_accuracy": 0.9003375172615051,
+      "step": 4560
+    },
+    {
+      "epoch": 2.1904990403071016,
+      "grad_norm": 0.4828821698318418,
+      "learning_rate": 1.4992890152861713e-05,
+      "loss": 0.2708,
+      "mean_token_accuracy": 0.9010523974895477,
+      "step": 4565
+    },
+    {
+      "epoch": 2.192898272552783,
+      "grad_norm": 0.4640884012585103,
+      "learning_rate": 1.4948453608247423e-05,
+      "loss": 0.2638,
+      "mean_token_accuracy": 0.9035166263580322,
+      "step": 4570
+    },
+    {
+      "epoch": 2.1952975047984644,
+      "grad_norm": 0.47179862655018706,
+      "learning_rate": 1.4904017063633133e-05,
+      "loss": 0.2713,
+      "mean_token_accuracy": 0.9010164678096771,
+      "step": 4575
+    },
+    {
+      "epoch": 2.197696737044146,
+      "grad_norm": 0.46938698374728643,
+      "learning_rate": 1.4859580519018843e-05,
+      "loss": 0.2772,
+      "mean_token_accuracy": 0.8985358417034149,
+      "step": 4580
+    },
+    {
+      "epoch": 2.200095969289827,
+      "grad_norm": 0.6409860361891765,
+      "learning_rate": 1.4815143974404552e-05,
+      "loss": 0.2689,
+      "mean_token_accuracy": 0.9022345066070556,
+      "step": 4585
+    },
+    {
+      "epoch": 2.2024952015355086,
+      "grad_norm": 0.481090158096422,
+      "learning_rate": 1.4770707429790258e-05,
+      "loss": 0.2816,
+      "mean_token_accuracy": 0.8974818766117096,
+      "step": 4590
+    },
+    {
+      "epoch": 2.20489443378119,
+      "grad_norm": 0.46329506596417347,
+      "learning_rate": 1.4726270885175969e-05,
+      "loss": 0.278,
+      "mean_token_accuracy": 0.8984702050685882,
+      "step": 4595
+    },
+    {
+      "epoch": 2.2072936660268714,
+      "grad_norm": 0.4615303070585285,
+      "learning_rate": 1.4681834340561679e-05,
+      "loss": 0.2682,
+      "mean_token_accuracy": 0.9019219934940338,
+      "step": 4600
+    },
+    {
+      "epoch": 2.2096928982725528,
+      "grad_norm": 0.46163618974055615,
+      "learning_rate": 1.4637397795947389e-05,
+      "loss": 0.2649,
+      "mean_token_accuracy": 0.9029759705066681,
+      "step": 4605
+    },
+    {
+      "epoch": 2.212092130518234,
+      "grad_norm": 0.4882515005514511,
+      "learning_rate": 1.4592961251333097e-05,
+      "loss": 0.277,
+      "mean_token_accuracy": 0.8990098178386688,
+      "step": 4610
+    },
+    {
+      "epoch": 2.2144913627639156,
+      "grad_norm": 0.4646349415807869,
+      "learning_rate": 1.4548524706718806e-05,
+      "loss": 0.2653,
+      "mean_token_accuracy": 0.9031439483165741,
+      "step": 4615
+    },
+    {
+      "epoch": 2.216890595009597,
+      "grad_norm": 0.44449725057095435,
+      "learning_rate": 1.4504088162104514e-05,
+      "loss": 0.2684,
+      "mean_token_accuracy": 0.9020173132419587,
+      "step": 4620
+    },
+    {
+      "epoch": 2.2192898272552783,
+      "grad_norm": 0.4656768176749509,
+      "learning_rate": 1.4459651617490225e-05,
+      "loss": 0.2751,
+      "mean_token_accuracy": 0.8993194878101349,
+      "step": 4625
+    },
+    {
+      "epoch": 2.2216890595009597,
+      "grad_norm": 0.4615214335096717,
+      "learning_rate": 1.4415215072875935e-05,
+      "loss": 0.2676,
+      "mean_token_accuracy": 0.9016627192497253,
+      "step": 4630
+    },
+    {
+      "epoch": 2.224088291746641,
+      "grad_norm": 0.46166194098937074,
+      "learning_rate": 1.4370778528261645e-05,
+      "loss": 0.2761,
+      "mean_token_accuracy": 0.8995867013931275,
+      "step": 4635
+    },
+    {
+      "epoch": 2.2264875239923225,
+      "grad_norm": 0.4416530636695825,
+      "learning_rate": 1.4326341983647352e-05,
+      "loss": 0.2786,
+      "mean_token_accuracy": 0.8981483101844787,
+      "step": 4640
+    },
+    {
+      "epoch": 2.228886756238004,
+      "grad_norm": 0.46476362069241783,
+      "learning_rate": 1.428190543903306e-05,
+      "loss": 0.2661,
+      "mean_token_accuracy": 0.9023899972438812,
+      "step": 4645
+    },
+    {
+      "epoch": 2.2312859884836853,
+      "grad_norm": 0.4642652248450991,
+      "learning_rate": 1.423746889441877e-05,
+      "loss": 0.2717,
+      "mean_token_accuracy": 0.9007234871387482,
+      "step": 4650
+    },
+    {
+      "epoch": 2.2336852207293667,
+      "grad_norm": 0.4565894514180209,
+      "learning_rate": 1.419303234980448e-05,
+      "loss": 0.2732,
+      "mean_token_accuracy": 0.9001586079597473,
+      "step": 4655
+    },
+    {
+      "epoch": 2.236084452975048,
+      "grad_norm": 0.45828103387435565,
+      "learning_rate": 1.414859580519019e-05,
+      "loss": 0.2727,
+      "mean_token_accuracy": 0.9002468883991241,
+      "step": 4660
+    },
+    {
+      "epoch": 2.2384836852207295,
+      "grad_norm": 0.4371651994660012,
+      "learning_rate": 1.4104159260575897e-05,
+      "loss": 0.2649,
+      "mean_token_accuracy": 0.9028993964195251,
+      "step": 4665
+    },
+    {
+      "epoch": 2.240882917466411,
+      "grad_norm": 0.45449934507751794,
+      "learning_rate": 1.4059722715961607e-05,
+      "loss": 0.2809,
+      "mean_token_accuracy": 0.8974053084850311,
+      "step": 4670
+    },
+    {
+      "epoch": 2.2432821497120923,
+      "grad_norm": 0.49571433936890613,
+      "learning_rate": 1.4015286171347316e-05,
+      "loss": 0.2851,
+      "mean_token_accuracy": 0.8978420555591583,
+      "step": 4675
+    },
+    {
+      "epoch": 2.2456813819577737,
+      "grad_norm": 0.4795958379989915,
+      "learning_rate": 1.3970849626733026e-05,
+      "loss": 0.2647,
+      "mean_token_accuracy": 0.9032507300376892,
+      "step": 4680
+    },
+    {
+      "epoch": 2.248080614203455,
+      "grad_norm": 0.45039732495232765,
+      "learning_rate": 1.3926413082118736e-05,
+      "loss": 0.2694,
+      "mean_token_accuracy": 0.9013946115970611,
+      "step": 4685
+    },
+    {
+      "epoch": 2.2504798464491365,
+      "grad_norm": 0.44596801571533046,
+      "learning_rate": 1.3881976537504443e-05,
+      "loss": 0.27,
+      "mean_token_accuracy": 0.9010569810867309,
+      "step": 4690
+    },
+    {
+      "epoch": 2.252879078694818,
+      "grad_norm": 0.4571665567266272,
+      "learning_rate": 1.3837539992890153e-05,
+      "loss": 0.2639,
+      "mean_token_accuracy": 0.9037930130958557,
+      "step": 4695
+    },
+    {
+      "epoch": 2.255278310940499,
+      "grad_norm": 0.4525712169800606,
+      "learning_rate": 1.3793103448275863e-05,
+      "loss": 0.2639,
+      "mean_token_accuracy": 0.9033798992633819,
+      "step": 4700
+    },
+    {
+      "epoch": 2.2576775431861806,
+      "grad_norm": 0.4594097941262841,
+      "learning_rate": 1.3748666903661572e-05,
+      "loss": 0.2574,
+      "mean_token_accuracy": 0.9055361032485962,
+      "step": 4705
+    },
+    {
+      "epoch": 2.2600767754318616,
+      "grad_norm": 0.47646345360050185,
+      "learning_rate": 1.3704230359047282e-05,
+      "loss": 0.2716,
+      "mean_token_accuracy": 0.9010154664516449,
+      "step": 4710
+    },
+    {
+      "epoch": 2.2624760076775434,
+      "grad_norm": 0.49498838567379894,
+      "learning_rate": 1.3659793814432989e-05,
+      "loss": 0.2636,
+      "mean_token_accuracy": 0.9037136912345887,
+      "step": 4715
+    },
+    {
+      "epoch": 2.2648752399232244,
+      "grad_norm": 0.4790896027279024,
+      "learning_rate": 1.3615357269818699e-05,
+      "loss": 0.269,
+      "mean_token_accuracy": 0.9017230927944183,
+      "step": 4720
+    },
+    {
+      "epoch": 2.2672744721689058,
+      "grad_norm": 0.48992595431942215,
+      "learning_rate": 1.3570920725204409e-05,
+      "loss": 0.2746,
+      "mean_token_accuracy": 0.8995851278305054,
+      "step": 4725
+    },
+    {
+      "epoch": 2.269673704414587,
+      "grad_norm": 0.5359200002174315,
+      "learning_rate": 1.3526484180590117e-05,
+      "loss": 0.2669,
+      "mean_token_accuracy": 0.9026368260383606,
+      "step": 4730
+    },
+    {
+      "epoch": 2.2720729366602685,
+      "grad_norm": 0.4496883562391663,
+      "learning_rate": 1.3482047635975828e-05,
+      "loss": 0.2643,
+      "mean_token_accuracy": 0.903408020734787,
+      "step": 4735
+    },
+    {
+      "epoch": 2.27447216890595,
+      "grad_norm": 0.4589288131283035,
+      "learning_rate": 1.3437611091361538e-05,
+      "loss": 0.2764,
+      "mean_token_accuracy": 0.8986647486686706,
+      "step": 4740
+    },
+    {
+      "epoch": 2.2768714011516313,
+      "grad_norm": 0.5034973762490947,
+      "learning_rate": 1.3393174546747245e-05,
+      "loss": 0.2653,
+      "mean_token_accuracy": 0.9032142698764801,
+      "step": 4745
+    },
+    {
+      "epoch": 2.2792706333973127,
+      "grad_norm": 0.47904037777632885,
+      "learning_rate": 1.3348738002132955e-05,
+      "loss": 0.2573,
+      "mean_token_accuracy": 0.9054487824440003,
+      "step": 4750
+    },
+    {
+      "epoch": 2.281669865642994,
+      "grad_norm": 0.4503339806395172,
+      "learning_rate": 1.3304301457518665e-05,
+      "loss": 0.2764,
+      "mean_token_accuracy": 0.8993960499763489,
+      "step": 4755
+    },
+    {
+      "epoch": 2.2840690978886755,
+      "grad_norm": 0.4492074544524633,
+      "learning_rate": 1.3259864912904373e-05,
+      "loss": 0.2613,
+      "mean_token_accuracy": 0.9042619824409485,
+      "step": 4760
+    },
+    {
+      "epoch": 2.286468330134357,
+      "grad_norm": 0.46624116200742,
+      "learning_rate": 1.3215428368290083e-05,
+      "loss": 0.2743,
+      "mean_token_accuracy": 0.8995259821414947,
+      "step": 4765
+    },
+    {
+      "epoch": 2.2888675623800383,
+      "grad_norm": 0.48969471633508643,
+      "learning_rate": 1.317099182367579e-05,
+      "loss": 0.2773,
+      "mean_token_accuracy": 0.8987483620643616,
+      "step": 4770
+    },
+    {
+      "epoch": 2.2912667946257197,
+      "grad_norm": 0.461439163954989,
+      "learning_rate": 1.31265552790615e-05,
+      "loss": 0.2643,
+      "mean_token_accuracy": 0.9034291326999664,
+      "step": 4775
+    },
+    {
+      "epoch": 2.293666026871401,
+      "grad_norm": 0.4658475465623233,
+      "learning_rate": 1.308211873444721e-05,
+      "loss": 0.2774,
+      "mean_token_accuracy": 0.8987905502319335,
+      "step": 4780
+    },
+    {
+      "epoch": 2.2960652591170825,
+      "grad_norm": 0.48408792260508365,
+      "learning_rate": 1.3037682189832919e-05,
+      "loss": 0.2753,
+      "mean_token_accuracy": 0.8993827760219574,
+      "step": 4785
+    },
+    {
+      "epoch": 2.298464491362764,
+      "grad_norm": 0.4537574035724088,
+      "learning_rate": 1.299324564521863e-05,
+      "loss": 0.2699,
+      "mean_token_accuracy": 0.9014258801937103,
+      "step": 4790
+    },
+    {
+      "epoch": 2.3008637236084453,
+      "grad_norm": 0.4510379588751342,
+      "learning_rate": 1.2948809100604336e-05,
+      "loss": 0.2572,
+      "mean_token_accuracy": 0.9058745920658111,
+      "step": 4795
+    },
+    {
+      "epoch": 2.3032629558541267,
+      "grad_norm": 0.49423695496099035,
+      "learning_rate": 1.2904372555990046e-05,
+      "loss": 0.2674,
+      "mean_token_accuracy": 0.9029056489467621,
+      "step": 4800
+    },
+    {
+      "epoch": 2.305662188099808,
+      "grad_norm": 0.45087124194764544,
+      "learning_rate": 1.2859936011375756e-05,
+      "loss": 0.2705,
+      "mean_token_accuracy": 0.9013524353504181,
+      "step": 4805
+    },
+    {
+      "epoch": 2.3080614203454894,
+      "grad_norm": 0.45105794906636565,
+      "learning_rate": 1.2815499466761466e-05,
+      "loss": 0.258,
+      "mean_token_accuracy": 0.9049526512622833,
+      "step": 4810
+    },
+    {
+      "epoch": 2.310460652591171,
+      "grad_norm": 0.4573799419919393,
+      "learning_rate": 1.2771062922147175e-05,
+      "loss": 0.2609,
+      "mean_token_accuracy": 0.9045884072780609,
+      "step": 4815
+    },
+    {
+      "epoch": 2.3128598848368522,
+      "grad_norm": 0.45047866282195803,
+      "learning_rate": 1.2726626377532883e-05,
+      "loss": 0.2661,
+      "mean_token_accuracy": 0.9023790538311005,
+      "step": 4820
+    },
+    {
+      "epoch": 2.3152591170825336,
+      "grad_norm": 0.44484306126020534,
+      "learning_rate": 1.2682189832918592e-05,
+      "loss": 0.2697,
+      "mean_token_accuracy": 0.9012295663356781,
+      "step": 4825
+    },
+    {
+      "epoch": 2.317658349328215,
+      "grad_norm": 0.4570040655210912,
+      "learning_rate": 1.2637753288304302e-05,
+      "loss": 0.2716,
+      "mean_token_accuracy": 0.9005773723125458,
+      "step": 4830
+    },
+    {
+      "epoch": 2.3200575815738964,
+      "grad_norm": 0.44512252026288296,
+      "learning_rate": 1.2593316743690012e-05,
+      "loss": 0.2708,
+      "mean_token_accuracy": 0.9009211480617523,
+      "step": 4835
+    },
+    {
+      "epoch": 2.322456813819578,
+      "grad_norm": 0.4647677877596983,
+      "learning_rate": 1.2548880199075722e-05,
+      "loss": 0.2654,
+      "mean_token_accuracy": 0.9026337563991547,
+      "step": 4840
+    },
+    {
+      "epoch": 2.324856046065259,
+      "grad_norm": 0.4451962255029952,
+      "learning_rate": 1.2504443654461429e-05,
+      "loss": 0.2722,
+      "mean_token_accuracy": 0.901082444190979,
+      "step": 4845
+    },
+    {
+      "epoch": 2.3272552783109406,
+      "grad_norm": 0.464769462545177,
+      "learning_rate": 1.2460007109847138e-05,
+      "loss": 0.2673,
+      "mean_token_accuracy": 0.9020293533802033,
+      "step": 4850
+    },
+    {
+      "epoch": 2.329654510556622,
+      "grad_norm": 0.5852191961334264,
+      "learning_rate": 1.2415570565232848e-05,
+      "loss": 0.2708,
+      "mean_token_accuracy": 0.9010414719581604,
+      "step": 4855
+    },
+    {
+      "epoch": 2.3320537428023034,
+      "grad_norm": 0.4452112718520486,
+      "learning_rate": 1.2371134020618558e-05,
+      "loss": 0.267,
+      "mean_token_accuracy": 0.902037626504898,
+      "step": 4860
+    },
+    {
+      "epoch": 2.3344529750479848,
+      "grad_norm": 0.4660962453351574,
+      "learning_rate": 1.2326697476004266e-05,
+      "loss": 0.2628,
+      "mean_token_accuracy": 0.903652572631836,
+      "step": 4865
+    },
+    {
+      "epoch": 2.336852207293666,
+      "grad_norm": 0.4510731188676562,
+      "learning_rate": 1.2282260931389976e-05,
+      "loss": 0.2614,
+      "mean_token_accuracy": 0.9041078269481659,
+      "step": 4870
+    },
+    {
+      "epoch": 2.3392514395393476,
+      "grad_norm": 0.45955652339680203,
+      "learning_rate": 1.2237824386775685e-05,
+      "loss": 0.2632,
+      "mean_token_accuracy": 0.9035033404827117,
+      "step": 4875
+    },
+    {
+      "epoch": 2.341650671785029,
+      "grad_norm": 0.45395656931314454,
+      "learning_rate": 1.2193387842161393e-05,
+      "loss": 0.2648,
+      "mean_token_accuracy": 0.9031056582927703,
+      "step": 4880
+    },
+    {
+      "epoch": 2.3440499040307103,
+      "grad_norm": 0.48526237862405125,
+      "learning_rate": 1.2148951297547104e-05,
+      "loss": 0.2633,
+      "mean_token_accuracy": 0.9035713195800781,
+      "step": 4885
+    },
+    {
+      "epoch": 2.3464491362763917,
+      "grad_norm": 0.4446519439579417,
+      "learning_rate": 1.2104514752932812e-05,
+      "loss": 0.261,
+      "mean_token_accuracy": 0.9043677151203156,
+      "step": 4890
+    },
+    {
+      "epoch": 2.348848368522073,
+      "grad_norm": 0.460184455442555,
+      "learning_rate": 1.2060078208318522e-05,
+      "loss": 0.2618,
+      "mean_token_accuracy": 0.9043756604194642,
+      "step": 4895
+    },
+    {
+      "epoch": 2.3512476007677545,
+      "grad_norm": 0.46259357140850754,
+      "learning_rate": 1.2015641663704232e-05,
+      "loss": 0.2641,
+      "mean_token_accuracy": 0.9031814455986023,
+      "step": 4900
+    },
+    {
+      "epoch": 2.3536468330134355,
+      "grad_norm": 0.44949250560949733,
+      "learning_rate": 1.1971205119089939e-05,
+      "loss": 0.2573,
+      "mean_token_accuracy": 0.9055706620216369,
+      "step": 4905
+    },
+    {
+      "epoch": 2.3560460652591173,
+      "grad_norm": 0.4763595046478618,
+      "learning_rate": 1.192676857447565e-05,
+      "loss": 0.2703,
+      "mean_token_accuracy": 0.9010188162326813,
+      "step": 4910
+    },
+    {
+      "epoch": 2.3584452975047983,
+      "grad_norm": 0.4747501461498077,
+      "learning_rate": 1.1882332029861358e-05,
+      "loss": 0.2751,
+      "mean_token_accuracy": 0.8994937121868134,
+      "step": 4915
+    },
+    {
+      "epoch": 2.36084452975048,
+      "grad_norm": 0.434688069222164,
+      "learning_rate": 1.1837895485247068e-05,
+      "loss": 0.2611,
+      "mean_token_accuracy": 0.9040064930915832,
+      "step": 4920
+    },
+    {
+      "epoch": 2.363243761996161,
+      "grad_norm": 0.44744820965947996,
+      "learning_rate": 1.1793458940632778e-05,
+      "loss": 0.2663,
+      "mean_token_accuracy": 0.9024548411369324,
+      "step": 4925
+    },
+    {
+      "epoch": 2.3656429942418424,
+      "grad_norm": 0.45455409918577533,
+      "learning_rate": 1.1749022396018487e-05,
+      "loss": 0.2665,
+      "mean_token_accuracy": 0.9024829626083374,
+      "step": 4930
+    },
+    {
+      "epoch": 2.368042226487524,
+      "grad_norm": 0.4526421090000582,
+      "learning_rate": 1.1704585851404195e-05,
+      "loss": 0.267,
+      "mean_token_accuracy": 0.902225923538208,
+      "step": 4935
+    },
+    {
+      "epoch": 2.370441458733205,
+      "grad_norm": 0.4565988566092905,
+      "learning_rate": 1.1660149306789903e-05,
+      "loss": 0.2749,
+      "mean_token_accuracy": 0.8996913850307464,
+      "step": 4940
+    },
+    {
+      "epoch": 2.3728406909788866,
+      "grad_norm": 0.45706018927722675,
+      "learning_rate": 1.1615712762175614e-05,
+      "loss": 0.2646,
+      "mean_token_accuracy": 0.9030478537082672,
+      "step": 4945
+    },
+    {
+      "epoch": 2.375239923224568,
+      "grad_norm": 0.43642789516463637,
+      "learning_rate": 1.1571276217561324e-05,
+      "loss": 0.2688,
+      "mean_token_accuracy": 0.9015024363994598,
+      "step": 4950
+    },
+    {
+      "epoch": 2.3776391554702494,
+      "grad_norm": 0.47138729568799137,
+      "learning_rate": 1.1526839672947032e-05,
+      "loss": 0.2725,
+      "mean_token_accuracy": 0.9005305051803589,
+      "step": 4955
+    },
+    {
+      "epoch": 2.380038387715931,
+      "grad_norm": 0.46096707859577163,
+      "learning_rate": 1.1482403128332742e-05,
+      "loss": 0.2629,
+      "mean_token_accuracy": 0.9035275757312775,
+      "step": 4960
+    },
+    {
+      "epoch": 2.382437619961612,
+      "grad_norm": 0.46062913409984646,
+      "learning_rate": 1.1437966583718449e-05,
+      "loss": 0.2672,
+      "mean_token_accuracy": 0.9022954523563385,
+      "step": 4965
+    },
+    {
+      "epoch": 2.3848368522072936,
+      "grad_norm": 0.46322200081563997,
+      "learning_rate": 1.139353003910416e-05,
+      "loss": 0.2645,
+      "mean_token_accuracy": 0.9036223649978637,
+      "step": 4970
+    },
+    {
+      "epoch": 2.387236084452975,
+      "grad_norm": 0.46121487559901814,
+      "learning_rate": 1.134909349448987e-05,
+      "loss": 0.2632,
+      "mean_token_accuracy": 0.9033205091953278,
+      "step": 4975
+    },
+    {
+      "epoch": 2.3896353166986564,
+      "grad_norm": 0.4412637352222931,
+      "learning_rate": 1.1304656949875578e-05,
+      "loss": 0.2629,
+      "mean_token_accuracy": 0.9037455499172211,
+      "step": 4980
+    },
+    {
+      "epoch": 2.3920345489443378,
+      "grad_norm": 0.43743475515560587,
+      "learning_rate": 1.1260220405261288e-05,
+      "loss": 0.2639,
+      "mean_token_accuracy": 0.9039111793041229,
+      "step": 4985
+    },
+    {
+      "epoch": 2.394433781190019,
+      "grad_norm": 0.4702348583607359,
+      "learning_rate": 1.1215783860646997e-05,
+      "loss": 0.2705,
+      "mean_token_accuracy": 0.9011352300643921,
+      "step": 4990
+    },
+    {
+      "epoch": 2.3968330134357005,
+      "grad_norm": 0.4498054424074278,
+      "learning_rate": 1.1171347316032705e-05,
+      "loss": 0.2697,
+      "mean_token_accuracy": 0.90220405459404,
+      "step": 4995
+    },
+    {
+      "epoch": 2.399232245681382,
+      "grad_norm": 0.4514753499047538,
+      "learning_rate": 1.1126910771418415e-05,
+      "loss": 0.269,
+      "mean_token_accuracy": 0.901602441072464,
+      "step": 5000
+    },
+    {
+      "epoch": 2.4016314779270633,
+      "grad_norm": 0.43564513806024646,
+      "learning_rate": 1.1082474226804124e-05,
+      "loss": 0.2593,
+      "mean_token_accuracy": 0.9050846874713898,
+      "step": 5005
+    },
+    {
+      "epoch": 2.4040307101727447,
+      "grad_norm": 0.4490319843983511,
+      "learning_rate": 1.1038037682189834e-05,
+      "loss": 0.2631,
+      "mean_token_accuracy": 0.9036346018314362,
+      "step": 5010
+    },
+    {
+      "epoch": 2.406429942418426,
+      "grad_norm": 0.4862626649550366,
+      "learning_rate": 1.0993601137575544e-05,
+      "loss": 0.2656,
+      "mean_token_accuracy": 0.9032509863376618,
+      "step": 5015
+    },
+    {
+      "epoch": 2.4088291746641075,
+      "grad_norm": 0.43251142361474554,
+      "learning_rate": 1.0949164592961252e-05,
+      "loss": 0.2622,
+      "mean_token_accuracy": 0.9033509850502014,
+      "step": 5020
+    },
+    {
+      "epoch": 2.411228406909789,
+      "grad_norm": 0.457498561849623,
+      "learning_rate": 1.0904728048346961e-05,
+      "loss": 0.2636,
+      "mean_token_accuracy": 0.9033877074718475,
+      "step": 5025
+    },
+    {
+      "epoch": 2.4136276391554703,
+      "grad_norm": 0.4506931899290913,
+      "learning_rate": 1.086029150373267e-05,
+      "loss": 0.2677,
+      "mean_token_accuracy": 0.9021892070770263,
+      "step": 5030
+    },
+    {
+      "epoch": 2.4160268714011517,
+      "grad_norm": 0.4557896328818507,
+      "learning_rate": 1.081585495911838e-05,
+      "loss": 0.2644,
+      "mean_token_accuracy": 0.902996277809143,
+      "step": 5035
+    },
+    {
+      "epoch": 2.418426103646833,
+      "grad_norm": 0.49573202206092815,
+      "learning_rate": 1.077141841450409e-05,
+      "loss": 0.2786,
+      "mean_token_accuracy": 0.8987530529499054,
+      "step": 5040
+    },
+    {
+      "epoch": 2.4208253358925145,
+      "grad_norm": 0.47783860783254734,
+      "learning_rate": 1.0726981869889798e-05,
+      "loss": 0.2681,
+      "mean_token_accuracy": 0.9018009006977081,
+      "step": 5045
+    },
+    {
+      "epoch": 2.423224568138196,
+      "grad_norm": 0.4294184963917144,
+      "learning_rate": 1.0682545325275507e-05,
+      "loss": 0.2651,
+      "mean_token_accuracy": 0.9027284860610962,
+      "step": 5050
+    },
+    {
+      "epoch": 2.4256238003838773,
+      "grad_norm": 0.44114115756512584,
+      "learning_rate": 1.0638108780661215e-05,
+      "loss": 0.2592,
+      "mean_token_accuracy": 0.905122983455658,
+      "step": 5055
+    },
+    {
+      "epoch": 2.4280230326295587,
+      "grad_norm": 0.4552986094382972,
+      "learning_rate": 1.0593672236046925e-05,
+      "loss": 0.2713,
+      "mean_token_accuracy": 0.900846141576767,
+      "step": 5060
+    },
+    {
+      "epoch": 2.43042226487524,
+      "grad_norm": 0.44135073807729075,
+      "learning_rate": 1.0549235691432635e-05,
+      "loss": 0.2653,
+      "mean_token_accuracy": 0.9026603162288666,
+      "step": 5065
+    },
+    {
+      "epoch": 2.4328214971209214,
+      "grad_norm": 0.4658946339558394,
+      "learning_rate": 1.0504799146818344e-05,
+      "loss": 0.2617,
+      "mean_token_accuracy": 0.9037299215793609,
+      "step": 5070
+    },
+    {
+      "epoch": 2.435220729366603,
+      "grad_norm": 0.4600480372973933,
+      "learning_rate": 1.0460362602204054e-05,
+      "loss": 0.2659,
+      "mean_token_accuracy": 0.9025923550128937,
+      "step": 5075
+    },
+    {
+      "epoch": 2.4376199616122842,
+      "grad_norm": 0.46058445465120573,
+      "learning_rate": 1.0415926057589762e-05,
+      "loss": 0.2613,
+      "mean_token_accuracy": 0.9045198023319244,
+      "step": 5080
+    },
+    {
+      "epoch": 2.4400191938579656,
+      "grad_norm": 0.42788161382405593,
+      "learning_rate": 1.0371489512975471e-05,
+      "loss": 0.267,
+      "mean_token_accuracy": 0.9020305931568146,
+      "step": 5085
+    },
+    {
+      "epoch": 2.442418426103647,
+      "grad_norm": 0.44027187566248915,
+      "learning_rate": 1.0327052968361181e-05,
+      "loss": 0.2672,
+      "mean_token_accuracy": 0.9023063957691193,
+      "step": 5090
+    },
+    {
+      "epoch": 2.4448176583493284,
+      "grad_norm": 0.4420597215561833,
+      "learning_rate": 1.028261642374689e-05,
+      "loss": 0.2624,
+      "mean_token_accuracy": 0.9036471009254455,
+      "step": 5095
+    },
+    {
+      "epoch": 2.4472168905950094,
+      "grad_norm": 0.4501730001489318,
+      "learning_rate": 1.02381798791326e-05,
+      "loss": 0.2671,
+      "mean_token_accuracy": 0.902330607175827,
+      "step": 5100
+    },
+    {
+      "epoch": 2.449616122840691,
+      "grad_norm": 0.44641850092775204,
+      "learning_rate": 1.0193743334518308e-05,
+      "loss": 0.2679,
+      "mean_token_accuracy": 0.9021657586097718,
+      "step": 5105
+    },
+    {
+      "epoch": 2.452015355086372,
+      "grad_norm": 0.4490534199420156,
+      "learning_rate": 1.0149306789904017e-05,
+      "loss": 0.2638,
+      "mean_token_accuracy": 0.9032556712627411,
+      "step": 5110
+    },
+    {
+      "epoch": 2.454414587332054,
+      "grad_norm": 0.44072785923447616,
+      "learning_rate": 1.0104870245289727e-05,
+      "loss": 0.2672,
+      "mean_token_accuracy": 0.9025269448757172,
+      "step": 5115
+    },
+    {
+      "epoch": 2.456813819577735,
+      "grad_norm": 0.4349885463972589,
+      "learning_rate": 1.0060433700675435e-05,
+      "loss": 0.2563,
+      "mean_token_accuracy": 0.9060261607170105,
+      "step": 5120
+    },
+    {
+      "epoch": 2.4592130518234163,
+      "grad_norm": 0.43592396045568815,
+      "learning_rate": 1.0015997156061145e-05,
+      "loss": 0.2573,
+      "mean_token_accuracy": 0.9053417444229126,
+      "step": 5125
+    },
+    {
+      "epoch": 2.4616122840690977,
+      "grad_norm": 0.46686763325718933,
+      "learning_rate": 9.971560611446856e-06,
+      "loss": 0.2674,
+      "mean_token_accuracy": 0.9016594767570496,
+      "step": 5130
+    },
+    {
+      "epoch": 2.464011516314779,
+      "grad_norm": 0.46845074051085117,
+      "learning_rate": 9.927124066832564e-06,
+      "loss": 0.2669,
+      "mean_token_accuracy": 0.9023274838924408,
+      "step": 5135
+    },
+    {
+      "epoch": 2.4664107485604605,
+      "grad_norm": 0.43958699879094765,
+      "learning_rate": 9.882687522218272e-06,
+      "loss": 0.2618,
+      "mean_token_accuracy": 0.9038767993450165,
+      "step": 5140
+    },
+    {
+      "epoch": 2.468809980806142,
+      "grad_norm": 0.41776189535267333,
+      "learning_rate": 9.838250977603981e-06,
+      "loss": 0.2602,
+      "mean_token_accuracy": 0.9043119847774506,
+      "step": 5145
+    },
+    {
+      "epoch": 2.4712092130518233,
+      "grad_norm": 0.4538080255278082,
+      "learning_rate": 9.793814432989691e-06,
+      "loss": 0.2597,
+      "mean_token_accuracy": 0.90516517162323,
+      "step": 5150
+    },
+    {
+      "epoch": 2.4736084452975047,
+      "grad_norm": 0.45447861990535154,
+      "learning_rate": 9.749377888375401e-06,
+      "loss": 0.2608,
+      "mean_token_accuracy": 0.9045280575752258,
+      "step": 5155
+    },
+    {
+      "epoch": 2.476007677543186,
+      "grad_norm": 0.4444329474266233,
+      "learning_rate": 9.70494134376111e-06,
+      "loss": 0.2665,
+      "mean_token_accuracy": 0.9021993517875672,
+      "step": 5160
+    },
+    {
+      "epoch": 2.4784069097888675,
+      "grad_norm": 0.43102288802079947,
+      "learning_rate": 9.660504799146818e-06,
+      "loss": 0.2644,
+      "mean_token_accuracy": 0.9035103678703308,
+      "step": 5165
+    },
+    {
+      "epoch": 2.480806142034549,
+      "grad_norm": 0.4376707591162667,
+      "learning_rate": 9.616068254532528e-06,
+      "loss": 0.2625,
+      "mean_token_accuracy": 0.9035827338695526,
+      "step": 5170
+    },
+    {
+      "epoch": 2.4832053742802302,
+      "grad_norm": 0.462803097758271,
+      "learning_rate": 9.571631709918237e-06,
+      "loss": 0.2633,
+      "mean_token_accuracy": 0.9033353626728058,
+      "step": 5175
+    },
+    {
+      "epoch": 2.4856046065259116,
+      "grad_norm": 0.46402677328691877,
+      "learning_rate": 9.527195165303947e-06,
+      "loss": 0.2678,
+      "mean_token_accuracy": 0.9018805861473084,
+      "step": 5180
+    },
+    {
+      "epoch": 2.488003838771593,
+      "grad_norm": 0.4626165921167753,
+      "learning_rate": 9.482758620689655e-06,
+      "loss": 0.266,
+      "mean_token_accuracy": 0.9025485992431641,
+      "step": 5185
+    },
+    {
+      "epoch": 2.4904030710172744,
+      "grad_norm": 0.4405898286968182,
+      "learning_rate": 9.438322076075366e-06,
+      "loss": 0.2598,
+      "mean_token_accuracy": 0.9050846874713898,
+      "step": 5190
+    },
+    {
+      "epoch": 2.492802303262956,
+      "grad_norm": 0.4491809630944021,
+      "learning_rate": 9.393885531461074e-06,
+      "loss": 0.2581,
+      "mean_token_accuracy": 0.9051714241504669,
+      "step": 5195
+    },
+    {
+      "epoch": 2.495201535508637,
+      "grad_norm": 0.45982750692789043,
+      "learning_rate": 9.349448986846782e-06,
+      "loss": 0.2635,
+      "mean_token_accuracy": 0.9034877181053161,
+      "step": 5200
+    },
+    {
+      "epoch": 2.4976007677543186,
+      "grad_norm": 0.4469470317195643,
+      "learning_rate": 9.305012442232493e-06,
+      "loss": 0.2563,
+      "mean_token_accuracy": 0.9061425805091858,
+      "step": 5205
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 0.46660175179380375,
+      "learning_rate": 9.260575897618201e-06,
+      "loss": 0.2674,
+      "mean_token_accuracy": 0.9026032865047455,
+      "step": 5210
+    },
+    {
+      "epoch": 2.5023992322456814,
+      "grad_norm": 0.4504938084645796,
+      "learning_rate": 9.216139353003911e-06,
+      "loss": 0.2604,
+      "mean_token_accuracy": 0.9043690323829651,
+      "step": 5215
+    },
+    {
+      "epoch": 2.504798464491363,
+      "grad_norm": 0.4917779663483051,
+      "learning_rate": 9.171702808389621e-06,
+      "loss": 0.262,
+      "mean_token_accuracy": 0.9041010320186615,
+      "step": 5220
+    },
+    {
+      "epoch": 2.507197696737044,
+      "grad_norm": 0.44848121064946134,
+      "learning_rate": 9.127266263775328e-06,
+      "loss": 0.266,
+      "mean_token_accuracy": 0.9026572048664093,
+      "step": 5225
+    },
+    {
+      "epoch": 2.5095969289827256,
+      "grad_norm": 0.45719698912437934,
+      "learning_rate": 9.082829719161038e-06,
+      "loss": 0.2678,
+      "mean_token_accuracy": 0.9016813576221466,
+      "step": 5230
+    },
+    {
+      "epoch": 2.511996161228407,
+      "grad_norm": 0.45266032559301805,
+      "learning_rate": 9.038393174546747e-06,
+      "loss": 0.2664,
+      "mean_token_accuracy": 0.9022712409496307,
+      "step": 5235
+    },
+    {
+      "epoch": 2.5143953934740884,
+      "grad_norm": 0.4621304603744557,
+      "learning_rate": 8.993956629932457e-06,
+      "loss": 0.2701,
+      "mean_token_accuracy": 0.9014797747135163,
+      "step": 5240
+    },
+    {
+      "epoch": 2.5167946257197698,
+      "grad_norm": 0.42584001010410305,
+      "learning_rate": 8.949520085318167e-06,
+      "loss": 0.2581,
+      "mean_token_accuracy": 0.9049862444400787,
+      "step": 5245
+    },
+    {
+      "epoch": 2.519193857965451,
+      "grad_norm": 0.42966249908808063,
+      "learning_rate": 8.905083540703876e-06,
+      "loss": 0.2603,
+      "mean_token_accuracy": 0.9045995116233826,
+      "step": 5250
+    },
+    {
+      "epoch": 2.5215930902111325,
+      "grad_norm": 0.4319478750873697,
+      "learning_rate": 8.860646996089584e-06,
+      "loss": 0.2609,
+      "mean_token_accuracy": 0.9043649673461914,
+      "step": 5255
+    },
+    {
+      "epoch": 2.523992322456814,
+      "grad_norm": 0.42974453071915686,
+      "learning_rate": 8.816210451475294e-06,
+      "loss": 0.264,
+      "mean_token_accuracy": 0.902919715642929,
+      "step": 5260
+    },
+    {
+      "epoch": 2.5263915547024953,
+      "grad_norm": 0.4397803370543199,
+      "learning_rate": 8.771773906861003e-06,
+      "loss": 0.2568,
+      "mean_token_accuracy": 0.9055386304855346,
+      "step": 5265
+    },
+    {
+      "epoch": 2.5287907869481767,
+      "grad_norm": 0.43395124387170547,
+      "learning_rate": 8.727337362246713e-06,
+      "loss": 0.2545,
+      "mean_token_accuracy": 0.9064959287643433,
+      "step": 5270
+    },
+    {
+      "epoch": 2.531190019193858,
+      "grad_norm": 0.4568247587575843,
+      "learning_rate": 8.682900817632421e-06,
+      "loss": 0.259,
+      "mean_token_accuracy": 0.9051628172397613,
+      "step": 5275
+    },
+    {
+      "epoch": 2.5335892514395395,
+      "grad_norm": 0.4527876906013087,
+      "learning_rate": 8.638464273018131e-06,
+      "loss": 0.2529,
+      "mean_token_accuracy": 0.9071043610572815,
+      "step": 5280
+    },
+    {
+      "epoch": 2.535988483685221,
+      "grad_norm": 0.4546151972549622,
+      "learning_rate": 8.59402772840384e-06,
+      "loss": 0.258,
+      "mean_token_accuracy": 0.9052597045898437,
+      "step": 5285
+    },
+    {
+      "epoch": 2.5383877159309023,
+      "grad_norm": 0.4356879694330123,
+      "learning_rate": 8.549591183789548e-06,
+      "loss": 0.2619,
+      "mean_token_accuracy": 0.9034970939159394,
+      "step": 5290
+    },
+    {
+      "epoch": 2.5407869481765832,
+      "grad_norm": 0.44317873246471456,
+      "learning_rate": 8.505154639175259e-06,
+      "loss": 0.2665,
+      "mean_token_accuracy": 0.9028915882110595,
+      "step": 5295
+    },
+    {
+      "epoch": 2.543186180422265,
+      "grad_norm": 0.4465050891994833,
+      "learning_rate": 8.460718094560967e-06,
+      "loss": 0.2587,
+      "mean_token_accuracy": 0.9050167143344879,
+      "step": 5300
+    },
+    {
+      "epoch": 2.545585412667946,
+      "grad_norm": 0.41974202659893994,
+      "learning_rate": 8.416281549946677e-06,
+      "loss": 0.2541,
+      "mean_token_accuracy": 0.9067621350288391,
+      "step": 5305
+    },
+    {
+      "epoch": 2.547984644913628,
+      "grad_norm": 0.4539270877064396,
+      "learning_rate": 8.371845005332386e-06,
+      "loss": 0.27,
+      "mean_token_accuracy": 0.9010617911815644,
+      "step": 5310
+    },
+    {
+      "epoch": 2.550383877159309,
+      "grad_norm": 0.4516961601525185,
+      "learning_rate": 8.327408460718094e-06,
+      "loss": 0.255,
+      "mean_token_accuracy": 0.906544154882431,
+      "step": 5315
+    },
+    {
+      "epoch": 2.5527831094049906,
+      "grad_norm": 0.4313966938059142,
+      "learning_rate": 8.282971916103804e-06,
+      "loss": 0.2603,
+      "mean_token_accuracy": 0.9046440362930298,
+      "step": 5320
+    },
+    {
+      "epoch": 2.5551823416506716,
+      "grad_norm": 0.45263025739656687,
+      "learning_rate": 8.238535371489513e-06,
+      "loss": 0.258,
+      "mean_token_accuracy": 0.9054847240447998,
+      "step": 5325
+    },
+    {
+      "epoch": 2.5575815738963534,
+      "grad_norm": 0.4528658588049054,
+      "learning_rate": 8.194098826875223e-06,
+      "loss": 0.2574,
+      "mean_token_accuracy": 0.9053511083126068,
+      "step": 5330
+    },
+    {
+      "epoch": 2.5599808061420344,
+      "grad_norm": 0.4480117751583745,
+      "learning_rate": 8.149662282260933e-06,
+      "loss": 0.2634,
+      "mean_token_accuracy": 0.9041584849357605,
+      "step": 5335
+    },
+    {
+      "epoch": 2.5623800383877158,
+      "grad_norm": 0.4432714854943306,
+      "learning_rate": 8.10522573764664e-06,
+      "loss": 0.2583,
+      "mean_token_accuracy": 0.9051284432411194,
+      "step": 5340
+    },
+    {
+      "epoch": 2.564779270633397,
+      "grad_norm": 0.4713748453981665,
+      "learning_rate": 8.06078919303235e-06,
+      "loss": 0.2672,
+      "mean_token_accuracy": 0.9021687090396882,
+      "step": 5345
+    },
+    {
+      "epoch": 2.5671785028790786,
+      "grad_norm": 0.4292455291598537,
+      "learning_rate": 8.016352648418058e-06,
+      "loss": 0.2569,
+      "mean_token_accuracy": 0.9057573795318603,
+      "step": 5350
+    },
+    {
+      "epoch": 2.56957773512476,
+      "grad_norm": 0.4533127233186409,
+      "learning_rate": 7.971916103803769e-06,
+      "loss": 0.2681,
+      "mean_token_accuracy": 0.9021807849407196,
+      "step": 5355
+    },
+    {
+      "epoch": 2.5719769673704413,
+      "grad_norm": 0.45378175639273344,
+      "learning_rate": 7.927479559189479e-06,
+      "loss": 0.2581,
+      "mean_token_accuracy": 0.9051651895046234,
+      "step": 5360
+    },
+    {
+      "epoch": 2.5743761996161227,
+      "grad_norm": 0.4305214724468804,
+      "learning_rate": 7.883043014575187e-06,
+      "loss": 0.2603,
+      "mean_token_accuracy": 0.9044901371002197,
+      "step": 5365
+    },
+    {
+      "epoch": 2.576775431861804,
+      "grad_norm": 0.4338096404785789,
+      "learning_rate": 7.838606469960896e-06,
+      "loss": 0.2531,
+      "mean_token_accuracy": 0.9074086010456085,
+      "step": 5370
+    },
+    {
+      "epoch": 2.5791746641074855,
+      "grad_norm": 0.4489769755708051,
+      "learning_rate": 7.794169925346606e-06,
+      "loss": 0.2642,
+      "mean_token_accuracy": 0.9032134890556336,
+      "step": 5375
+    },
+    {
+      "epoch": 2.581573896353167,
+      "grad_norm": 0.4329446145757721,
+      "learning_rate": 7.749733380732314e-06,
+      "loss": 0.2535,
+      "mean_token_accuracy": 0.9067066788673401,
+      "step": 5380
+    },
+    {
+      "epoch": 2.5839731285988483,
+      "grad_norm": 0.4669192159499125,
+      "learning_rate": 7.705296836118024e-06,
+      "loss": 0.2605,
+      "mean_token_accuracy": 0.9048237323760986,
+      "step": 5385
+    },
+    {
+      "epoch": 2.5863723608445297,
+      "grad_norm": 0.4382912605817621,
+      "learning_rate": 7.660860291503733e-06,
+      "loss": 0.2607,
+      "mean_token_accuracy": 0.9042885541915894,
+      "step": 5390
+    },
+    {
+      "epoch": 2.588771593090211,
+      "grad_norm": 0.44945986639739094,
+      "learning_rate": 7.616423746889442e-06,
+      "loss": 0.2604,
+      "mean_token_accuracy": 0.9042424559593201,
+      "step": 5395
+    },
+    {
+      "epoch": 2.5911708253358925,
+      "grad_norm": 0.44732804884036653,
+      "learning_rate": 7.571987202275152e-06,
+      "loss": 0.2505,
+      "mean_token_accuracy": 0.907641065120697,
+      "step": 5400
+    },
+    {
+      "epoch": 2.593570057581574,
+      "grad_norm": 0.45128927801380114,
+      "learning_rate": 7.52755065766086e-06,
+      "loss": 0.2513,
+      "mean_token_accuracy": 0.9073582708835601,
+      "step": 5405
+    },
+    {
+      "epoch": 2.5959692898272553,
+      "grad_norm": 0.4836265077335699,
+      "learning_rate": 7.48311411304657e-06,
+      "loss": 0.2526,
+      "mean_token_accuracy": 0.9070647537708283,
+      "step": 5410
+    },
+    {
+      "epoch": 2.5983685220729367,
+      "grad_norm": 0.4242626701014422,
+      "learning_rate": 7.438677568432279e-06,
+      "loss": 0.2589,
+      "mean_token_accuracy": 0.905083417892456,
+      "step": 5415
+    },
+    {
+      "epoch": 2.600767754318618,
+      "grad_norm": 0.46111786679737476,
+      "learning_rate": 7.394241023817988e-06,
+      "loss": 0.2626,
+      "mean_token_accuracy": 0.9037265062332154,
+      "step": 5420
+    },
+    {
+      "epoch": 2.6031669865642995,
+      "grad_norm": 0.4339306870051137,
+      "learning_rate": 7.349804479203698e-06,
+      "loss": 0.2593,
+      "mean_token_accuracy": 0.9046136617660523,
+      "step": 5425
+    },
+    {
+      "epoch": 2.605566218809981,
+      "grad_norm": 0.4579505156147382,
+      "learning_rate": 7.3053679345894065e-06,
+      "loss": 0.2562,
+      "mean_token_accuracy": 0.9056503415107727,
+      "step": 5430
+    },
+    {
+      "epoch": 2.6079654510556622,
+      "grad_norm": 0.45657019011680416,
+      "learning_rate": 7.260931389975116e-06,
+      "loss": 0.2592,
+      "mean_token_accuracy": 0.9047815501689911,
+      "step": 5435
+    },
+    {
+      "epoch": 2.6103646833013436,
+      "grad_norm": 0.4304320724321448,
+      "learning_rate": 7.216494845360824e-06,
+      "loss": 0.2585,
+      "mean_token_accuracy": 0.9053276717662812,
+      "step": 5440
+    },
+    {
+      "epoch": 2.612763915547025,
+      "grad_norm": 0.4873685490263298,
+      "learning_rate": 7.1720583007465344e-06,
+      "loss": 0.2677,
+      "mean_token_accuracy": 0.9017115533351898,
+      "step": 5445
+    },
+    {
+      "epoch": 2.6151631477927064,
+      "grad_norm": 0.5584142725364183,
+      "learning_rate": 7.127621756132244e-06,
+      "loss": 0.2692,
+      "mean_token_accuracy": 0.901091468334198,
+      "step": 5450
+    },
+    {
+      "epoch": 2.617562380038388,
+      "grad_norm": 0.4393456506678057,
+      "learning_rate": 7.083185211517952e-06,
+      "loss": 0.2593,
+      "mean_token_accuracy": 0.9047893643379211,
+      "step": 5455
+    },
+    {
+      "epoch": 2.619961612284069,
+      "grad_norm": 0.46284001629573074,
+      "learning_rate": 7.038748666903662e-06,
+      "loss": 0.2621,
+      "mean_token_accuracy": 0.903766393661499,
+      "step": 5460
+    },
+    {
+      "epoch": 2.6223608445297506,
+      "grad_norm": 0.43277293461684013,
+      "learning_rate": 6.994312122289372e-06,
+      "loss": 0.2599,
+      "mean_token_accuracy": 0.9046470880508423,
+      "step": 5465
+    },
+    {
+      "epoch": 2.624760076775432,
+      "grad_norm": 0.49262225304162727,
+      "learning_rate": 6.94987557767508e-06,
+      "loss": 0.2537,
+      "mean_token_accuracy": 0.9066051006317138,
+      "step": 5470
+    },
+    {
+      "epoch": 2.6271593090211134,
+      "grad_norm": 0.4466173887837331,
+      "learning_rate": 6.9054390330607895e-06,
+      "loss": 0.2659,
+      "mean_token_accuracy": 0.9025446891784668,
+      "step": 5475
+    },
+    {
+      "epoch": 2.629558541266795,
+      "grad_norm": 0.48143348155877075,
+      "learning_rate": 6.861002488446498e-06,
+      "loss": 0.2669,
+      "mean_token_accuracy": 0.9022337436676026,
+      "step": 5480
+    },
+    {
+      "epoch": 2.631957773512476,
+      "grad_norm": 0.4569390008901632,
+      "learning_rate": 6.816565943832208e-06,
+      "loss": 0.262,
+      "mean_token_accuracy": 0.9040025889873504,
+      "step": 5485
+    },
+    {
+      "epoch": 2.634357005758157,
+      "grad_norm": 0.4324944567222657,
+      "learning_rate": 6.772129399217917e-06,
+      "loss": 0.2615,
+      "mean_token_accuracy": 0.9045298337936402,
+      "step": 5490
+    },
+    {
+      "epoch": 2.636756238003839,
+      "grad_norm": 0.45413612337003467,
+      "learning_rate": 6.727692854603626e-06,
+      "loss": 0.2589,
+      "mean_token_accuracy": 0.9049846887588501,
+      "step": 5495
+    },
+    {
+      "epoch": 2.63915547024952,
+      "grad_norm": 0.4402754826227816,
+      "learning_rate": 6.683256309989336e-06,
+      "loss": 0.254,
+      "mean_token_accuracy": 0.9072097957134246,
+      "step": 5500
+    },
+    {
+      "epoch": 2.6415547024952017,
+      "grad_norm": 0.43715463368337787,
+      "learning_rate": 6.6388197653750445e-06,
+      "loss": 0.2689,
+      "mean_token_accuracy": 0.9015414953231812,
+      "step": 5505
+    },
+    {
+      "epoch": 2.6439539347408827,
+      "grad_norm": 0.4827231208784812,
+      "learning_rate": 6.594383220760754e-06,
+      "loss": 0.2647,
+      "mean_token_accuracy": 0.9032431781291962,
+      "step": 5510
+    },
+    {
+      "epoch": 2.6463531669865645,
+      "grad_norm": 0.4402689194179168,
+      "learning_rate": 6.549946676146464e-06,
+      "loss": 0.2545,
+      "mean_token_accuracy": 0.907448124885559,
+      "step": 5515
+    },
+    {
+      "epoch": 2.6487523992322455,
+      "grad_norm": 0.4516566081943414,
+      "learning_rate": 6.505510131532172e-06,
+      "loss": 0.2618,
+      "mean_token_accuracy": 0.9039760291576385,
+      "step": 5520
+    },
+    {
+      "epoch": 2.6511516314779273,
+      "grad_norm": 0.45641350808044095,
+      "learning_rate": 6.461073586917882e-06,
+      "loss": 0.2681,
+      "mean_token_accuracy": 0.9020151495933533,
+      "step": 5525
+    },
+    {
+      "epoch": 2.6535508637236083,
+      "grad_norm": 0.45820168634061464,
+      "learning_rate": 6.41663704230359e-06,
+      "loss": 0.2594,
+      "mean_token_accuracy": 0.9047295212745666,
+      "step": 5530
+    },
+    {
+      "epoch": 2.65595009596929,
+      "grad_norm": 0.48078712024469483,
+      "learning_rate": 6.3722004976892995e-06,
+      "loss": 0.262,
+      "mean_token_accuracy": 0.9041463553905487,
+      "step": 5535
+    },
+    {
+      "epoch": 2.658349328214971,
+      "grad_norm": 0.4310848038599131,
+      "learning_rate": 6.32776395307501e-06,
+      "loss": 0.2614,
+      "mean_token_accuracy": 0.9039002537727356,
+      "step": 5540
+    },
+    {
+      "epoch": 2.6607485604606524,
+      "grad_norm": 0.45563225963393356,
+      "learning_rate": 6.283327408460718e-06,
+      "loss": 0.2589,
+      "mean_token_accuracy": 0.905088609457016,
+      "step": 5545
+    },
+    {
+      "epoch": 2.663147792706334,
+      "grad_norm": 0.4432971692450577,
+      "learning_rate": 6.238890863846427e-06,
+      "loss": 0.2539,
+      "mean_token_accuracy": 0.9068746447563172,
+      "step": 5550
+    },
+    {
+      "epoch": 2.6655470249520152,
+      "grad_norm": 0.45973148608425424,
+      "learning_rate": 6.194454319232137e-06,
+      "loss": 0.2636,
+      "mean_token_accuracy": 0.9035470664501191,
+      "step": 5555
+    },
+    {
+      "epoch": 2.6679462571976966,
+      "grad_norm": 0.42951757627124687,
+      "learning_rate": 6.150017774617846e-06,
+      "loss": 0.2559,
+      "mean_token_accuracy": 0.9060870885849,
+      "step": 5560
+    },
+    {
+      "epoch": 2.670345489443378,
+      "grad_norm": 0.4359773306703161,
+      "learning_rate": 6.1055812300035545e-06,
+      "loss": 0.2538,
+      "mean_token_accuracy": 0.9069285571575165,
+      "step": 5565
+    },
+    {
+      "epoch": 2.6727447216890594,
+      "grad_norm": 0.43446645300714337,
+      "learning_rate": 6.061144685389265e-06,
+      "loss": 0.2485,
+      "mean_token_accuracy": 0.9087942957878112,
+      "step": 5570
+    },
+    {
+      "epoch": 2.675143953934741,
+      "grad_norm": 0.45104511017647375,
+      "learning_rate": 6.016708140774974e-06,
+      "loss": 0.2641,
+      "mean_token_accuracy": 0.9027559340000153,
+      "step": 5575
+    },
+    {
+      "epoch": 2.677543186180422,
+      "grad_norm": 0.4522396662766918,
+      "learning_rate": 5.9722715961606824e-06,
+      "loss": 0.2585,
+      "mean_token_accuracy": 0.9057000160217286,
+      "step": 5580
+    },
+    {
+      "epoch": 2.6799424184261036,
+      "grad_norm": 0.4290452834201096,
+      "learning_rate": 5.927835051546392e-06,
+      "loss": 0.2581,
+      "mean_token_accuracy": 0.9052300214767456,
+      "step": 5585
+    },
+    {
+      "epoch": 2.682341650671785,
+      "grad_norm": 0.4232161059713672,
+      "learning_rate": 5.883398506932102e-06,
+      "loss": 0.2542,
+      "mean_token_accuracy": 0.906181687116623,
+      "step": 5590
+    },
+    {
+      "epoch": 2.6847408829174664,
+      "grad_norm": 0.43860671983027966,
+      "learning_rate": 5.83896196231781e-06,
+      "loss": 0.2612,
+      "mean_token_accuracy": 0.9042408883571624,
+      "step": 5595
+    },
+    {
+      "epoch": 2.6871401151631478,
+      "grad_norm": 0.4256076226233432,
+      "learning_rate": 5.79452541770352e-06,
+      "loss": 0.2573,
+      "mean_token_accuracy": 0.9054723083972931,
+      "step": 5600
+    },
+    {
+      "epoch": 2.689539347408829,
+      "grad_norm": 0.45573244253579825,
+      "learning_rate": 5.750088873089229e-06,
+      "loss": 0.2511,
+      "mean_token_accuracy": 0.9077927350997925,
+      "step": 5605
+    },
+    {
+      "epoch": 2.6919385796545106,
+      "grad_norm": 0.44045582178642395,
+      "learning_rate": 5.7056523284749374e-06,
+      "loss": 0.2567,
+      "mean_token_accuracy": 0.9058355093002319,
+      "step": 5610
+    },
+    {
+      "epoch": 2.694337811900192,
+      "grad_norm": 0.47977934654072885,
+      "learning_rate": 5.661215783860648e-06,
+      "loss": 0.2645,
+      "mean_token_accuracy": 0.9032580137252808,
+      "step": 5615
+    },
+    {
+      "epoch": 2.6967370441458733,
+      "grad_norm": 0.4379358865890731,
+      "learning_rate": 5.616779239246357e-06,
+      "loss": 0.2605,
+      "mean_token_accuracy": 0.9041158795356751,
+      "step": 5620
+    },
+    {
+      "epoch": 2.6991362763915547,
+      "grad_norm": 0.4445723936789973,
+      "learning_rate": 5.572342694632065e-06,
+      "loss": 0.262,
+      "mean_token_accuracy": 0.9042987108230591,
+      "step": 5625
+    },
+    {
+      "epoch": 2.701535508637236,
+      "grad_norm": 0.4325473400172011,
+      "learning_rate": 5.527906150017775e-06,
+      "loss": 0.2641,
+      "mean_token_accuracy": 0.9031509757041931,
+      "step": 5630
+    },
+    {
+      "epoch": 2.7039347408829175,
+      "grad_norm": 0.4145757455771518,
+      "learning_rate": 5.483469605403484e-06,
+      "loss": 0.249,
+      "mean_token_accuracy": 0.9083544373512268,
+      "step": 5635
+    },
+    {
+      "epoch": 2.706333973128599,
+      "grad_norm": 0.45663630022058826,
+      "learning_rate": 5.439033060789193e-06,
+      "loss": 0.2601,
+      "mean_token_accuracy": 0.904858124256134,
+      "step": 5640
+    },
+    {
+      "epoch": 2.7087332053742803,
+      "grad_norm": 0.43350528253109694,
+      "learning_rate": 5.394596516174903e-06,
+      "loss": 0.2486,
+      "mean_token_accuracy": 0.9084395945072175,
+      "step": 5645
+    },
+    {
+      "epoch": 2.7111324376199617,
+      "grad_norm": 0.43791759142392406,
+      "learning_rate": 5.350159971560612e-06,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.9072106003761291,
+      "step": 5650
+    },
+    {
+      "epoch": 2.713531669865643,
+      "grad_norm": 0.4479123601679622,
+      "learning_rate": 5.30572342694632e-06,
+      "loss": 0.256,
+      "mean_token_accuracy": 0.9059386551380157,
+      "step": 5655
+    },
+    {
+      "epoch": 2.7159309021113245,
+      "grad_norm": 0.49310385054485034,
+      "learning_rate": 5.2612868823320305e-06,
+      "loss": 0.2622,
+      "mean_token_accuracy": 0.9039049267768859,
+      "step": 5660
+    },
+    {
+      "epoch": 2.718330134357006,
+      "grad_norm": 0.4480477544426612,
+      "learning_rate": 5.216850337717739e-06,
+      "loss": 0.2678,
+      "mean_token_accuracy": 0.9020596444606781,
+      "step": 5665
+    },
+    {
+      "epoch": 2.7207293666026873,
+      "grad_norm": 0.44784864613557246,
+      "learning_rate": 5.172413793103448e-06,
+      "loss": 0.2602,
+      "mean_token_accuracy": 0.9044237077236176,
+      "step": 5670
+    },
+    {
+      "epoch": 2.7231285988483687,
+      "grad_norm": 0.4503609778786115,
+      "learning_rate": 5.127977248489158e-06,
+      "loss": 0.2537,
+      "mean_token_accuracy": 0.9068785607814789,
+      "step": 5675
+    },
+    {
+      "epoch": 2.72552783109405,
+      "grad_norm": 0.5298964325267139,
+      "learning_rate": 5.083540703874867e-06,
+      "loss": 0.26,
+      "mean_token_accuracy": 0.9050964057445526,
+      "step": 5680
+    },
+    {
+      "epoch": 2.7279270633397315,
+      "grad_norm": 0.44324275118851564,
+      "learning_rate": 5.039104159260576e-06,
+      "loss": 0.2575,
+      "mean_token_accuracy": 0.9054651856422424,
+      "step": 5685
+    },
+    {
+      "epoch": 2.730326295585413,
+      "grad_norm": 0.44159279875032903,
+      "learning_rate": 4.9946676146462856e-06,
+      "loss": 0.2568,
+      "mean_token_accuracy": 0.9059472501277923,
+      "step": 5690
+    },
+    {
+      "epoch": 2.732725527831094,
+      "grad_norm": 0.4500000199898264,
+      "learning_rate": 4.950231070031994e-06,
+      "loss": 0.2595,
+      "mean_token_accuracy": 0.9048760831356049,
+      "step": 5695
+    },
+    {
+      "epoch": 2.7351247600767756,
+      "grad_norm": 0.4399650800626675,
+      "learning_rate": 4.905794525417703e-06,
+      "loss": 0.2558,
+      "mean_token_accuracy": 0.9062402307987213,
+      "step": 5700
+    },
+    {
+      "epoch": 2.7375239923224566,
+      "grad_norm": 0.43485531521512283,
+      "learning_rate": 4.8613579808034135e-06,
+      "loss": 0.2583,
+      "mean_token_accuracy": 0.9057276964187622,
+      "step": 5705
+    },
+    {
+      "epoch": 2.7399232245681384,
+      "grad_norm": 0.4379171063459807,
+      "learning_rate": 4.816921436189122e-06,
+      "loss": 0.257,
+      "mean_token_accuracy": 0.9057183146476746,
+      "step": 5710
+    },
+    {
+      "epoch": 2.7423224568138194,
+      "grad_norm": 0.44593794361071326,
+      "learning_rate": 4.772484891574831e-06,
+      "loss": 0.2555,
+      "mean_token_accuracy": 0.9061355412006378,
+      "step": 5715
+    },
+    {
+      "epoch": 2.744721689059501,
+      "grad_norm": 0.4454660382415944,
+      "learning_rate": 4.7280483469605406e-06,
+      "loss": 0.2622,
+      "mean_token_accuracy": 0.9044401228427887,
+      "step": 5720
+    },
+    {
+      "epoch": 2.747120921305182,
+      "grad_norm": 0.466071797740869,
+      "learning_rate": 4.68361180234625e-06,
+      "loss": 0.2489,
+      "mean_token_accuracy": 0.9086145997047425,
+      "step": 5725
+    },
+    {
+      "epoch": 2.749520153550864,
+      "grad_norm": 0.4312082041812329,
+      "learning_rate": 4.639175257731959e-06,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.9071082472801208,
+      "step": 5730
+    },
+    {
+      "epoch": 2.751919385796545,
+      "grad_norm": 0.45469928675392296,
+      "learning_rate": 4.5947387131176685e-06,
+      "loss": 0.2573,
+      "mean_token_accuracy": 0.9056995809078217,
+      "step": 5735
+    },
+    {
+      "epoch": 2.7543186180422263,
+      "grad_norm": 0.4476907768840197,
+      "learning_rate": 4.550302168503377e-06,
+      "loss": 0.2521,
+      "mean_token_accuracy": 0.9072535812854767,
+      "step": 5740
+    },
+    {
+      "epoch": 2.7567178502879077,
+      "grad_norm": 0.42447627907739277,
+      "learning_rate": 4.505865623889086e-06,
+      "loss": 0.2523,
+      "mean_token_accuracy": 0.907138729095459,
+      "step": 5745
+    },
+    {
+      "epoch": 2.759117082533589,
+      "grad_norm": 0.45012533604922167,
+      "learning_rate": 4.4614290792747964e-06,
+      "loss": 0.2509,
+      "mean_token_accuracy": 0.9077301681041717,
+      "step": 5750
+    },
+    {
+      "epoch": 2.7615163147792705,
+      "grad_norm": 0.4409578164982194,
+      "learning_rate": 4.416992534660505e-06,
+      "loss": 0.2548,
+      "mean_token_accuracy": 0.906362110376358,
+      "step": 5755
+    },
+    {
+      "epoch": 2.763915547024952,
+      "grad_norm": 0.48147201127170397,
+      "learning_rate": 4.372555990046214e-06,
+      "loss": 0.2622,
+      "mean_token_accuracy": 0.9043830871582031,
+      "step": 5760
+    },
+    {
+      "epoch": 2.7663147792706333,
+      "grad_norm": 0.4497274342056709,
+      "learning_rate": 4.3281194454319235e-06,
+      "loss": 0.2612,
+      "mean_token_accuracy": 0.9043799638748169,
+      "step": 5765
+    },
+    {
+      "epoch": 2.7687140115163147,
+      "grad_norm": 0.4413348164130912,
+      "learning_rate": 4.283682900817633e-06,
+      "loss": 0.2536,
+      "mean_token_accuracy": 0.9070066928863525,
+      "step": 5770
+    },
+    {
+      "epoch": 2.771113243761996,
+      "grad_norm": 0.4573010381962229,
+      "learning_rate": 4.239246356203342e-06,
+      "loss": 0.2543,
+      "mean_token_accuracy": 0.9065881192684173,
+      "step": 5775
+    },
+    {
+      "epoch": 2.7735124760076775,
+      "grad_norm": 0.4370029883452461,
+      "learning_rate": 4.1948098115890514e-06,
+      "loss": 0.2448,
+      "mean_token_accuracy": 0.9098528385162353,
+      "step": 5780
+    },
+    {
+      "epoch": 2.775911708253359,
+      "grad_norm": 0.43272263858643206,
+      "learning_rate": 4.15037326697476e-06,
+      "loss": 0.2514,
+      "mean_token_accuracy": 0.9076957941055298,
+      "step": 5785
+    },
+    {
+      "epoch": 2.7783109404990403,
+      "grad_norm": 0.44030045712445476,
+      "learning_rate": 4.105936722360469e-06,
+      "loss": 0.2582,
+      "mean_token_accuracy": 0.9051995396614074,
+      "step": 5790
+    },
+    {
+      "epoch": 2.7807101727447217,
+      "grad_norm": 0.42445012441746516,
+      "learning_rate": 4.0615001777461785e-06,
+      "loss": 0.2557,
+      "mean_token_accuracy": 0.9057902097702026,
+      "step": 5795
+    },
+    {
+      "epoch": 2.783109404990403,
+      "grad_norm": 0.44008392900477256,
+      "learning_rate": 4.017063633131888e-06,
+      "loss": 0.2571,
+      "mean_token_accuracy": 0.9057230174541473,
+      "step": 5800
+    },
+    {
+      "epoch": 2.7855086372360844,
+      "grad_norm": 0.46535904070186546,
+      "learning_rate": 3.972627088517597e-06,
+      "loss": 0.2621,
+      "mean_token_accuracy": 0.9043276190757752,
+      "step": 5805
+    },
+    {
+      "epoch": 2.787907869481766,
+      "grad_norm": 0.4181165672832623,
+      "learning_rate": 3.9281905439033065e-06,
+      "loss": 0.2602,
+      "mean_token_accuracy": 0.9042987048625946,
+      "step": 5810
+    },
+    {
+      "epoch": 2.7903071017274472,
+      "grad_norm": 0.4522846803133769,
+      "learning_rate": 3.883753999289016e-06,
+      "loss": 0.2544,
+      "mean_token_accuracy": 0.9064529299736023,
+      "step": 5815
+    },
+    {
+      "epoch": 2.7927063339731286,
+      "grad_norm": 0.44815812391112714,
+      "learning_rate": 3.839317454674725e-06,
+      "loss": 0.2588,
+      "mean_token_accuracy": 0.9048435807228088,
+      "step": 5820
+    },
+    {
+      "epoch": 2.79510556621881,
+      "grad_norm": 0.46421235672539135,
+      "learning_rate": 3.794880910060434e-06,
+      "loss": 0.2592,
+      "mean_token_accuracy": 0.9051807999610901,
+      "step": 5825
+    },
+    {
+      "epoch": 2.7975047984644914,
+      "grad_norm": 3.709167924530769,
+      "learning_rate": 3.750444365446143e-06,
+      "loss": 0.2543,
+      "mean_token_accuracy": 0.9065676093101501,
+      "step": 5830
+    },
+    {
+      "epoch": 2.799904030710173,
+      "grad_norm": 0.4240367546919462,
+      "learning_rate": 3.706007820831852e-06,
+      "loss": 0.2544,
+      "mean_token_accuracy": 0.9064964950084686,
+      "step": 5835
+    },
+    {
+      "epoch": 2.802303262955854,
+      "grad_norm": 0.4456102248393658,
+      "learning_rate": 3.661571276217562e-06,
+      "loss": 0.2539,
+      "mean_token_accuracy": 0.906609308719635,
+      "step": 5840
+    },
+    {
+      "epoch": 2.8047024952015356,
+      "grad_norm": 0.44600145811223735,
+      "learning_rate": 3.617134731603271e-06,
+      "loss": 0.2582,
+      "mean_token_accuracy": 0.905487060546875,
+      "step": 5845
+    },
+    {
+      "epoch": 2.807101727447217,
+      "grad_norm": 0.4376857420314512,
+      "learning_rate": 3.5726981869889797e-06,
+      "loss": 0.2649,
+      "mean_token_accuracy": 0.9028728306293488,
+      "step": 5850
+    },
+    {
+      "epoch": 2.8095009596928984,
+      "grad_norm": 0.43309004835981246,
+      "learning_rate": 3.528261642374689e-06,
+      "loss": 0.255,
+      "mean_token_accuracy": 0.9064680218696595,
+      "step": 5855
+    },
+    {
+      "epoch": 2.8119001919385798,
+      "grad_norm": 0.45511351183444065,
+      "learning_rate": 3.483825097760398e-06,
+      "loss": 0.2535,
+      "mean_token_accuracy": 0.9068301141262054,
+      "step": 5860
+    },
+    {
+      "epoch": 2.814299424184261,
+      "grad_norm": 0.45459227347655273,
+      "learning_rate": 3.4393885531461076e-06,
+      "loss": 0.2542,
+      "mean_token_accuracy": 0.9065988540649415,
+      "step": 5865
+    },
+    {
+      "epoch": 2.8166986564299425,
+      "grad_norm": 0.4474470069462497,
+      "learning_rate": 3.394952008531817e-06,
+      "loss": 0.2528,
+      "mean_token_accuracy": 0.9072067022323609,
+      "step": 5870
+    },
+    {
+      "epoch": 2.819097888675624,
+      "grad_norm": 0.43956377189608403,
+      "learning_rate": 3.350515463917526e-06,
+      "loss": 0.2584,
+      "mean_token_accuracy": 0.9054464340209961,
+      "step": 5875
+    },
+    {
+      "epoch": 2.8214971209213053,
+      "grad_norm": 0.4369485147974561,
+      "learning_rate": 3.3060789193032347e-06,
+      "loss": 0.2533,
+      "mean_token_accuracy": 0.9070504426956176,
+      "step": 5880
+    },
+    {
+      "epoch": 2.8238963531669867,
+      "grad_norm": 0.4300540536602179,
+      "learning_rate": 3.2616423746889444e-06,
+      "loss": 0.2544,
+      "mean_token_accuracy": 0.9066879153251648,
+      "step": 5885
+    },
+    {
+      "epoch": 2.8262955854126677,
+      "grad_norm": 0.43603672723055475,
+      "learning_rate": 3.2172058300746537e-06,
+      "loss": 0.2585,
+      "mean_token_accuracy": 0.905434650182724,
+      "step": 5890
+    },
+    {
+      "epoch": 2.8286948176583495,
+      "grad_norm": 0.44584348089802384,
+      "learning_rate": 3.1727692854603626e-06,
+      "loss": 0.2431,
+      "mean_token_accuracy": 0.910099059343338,
+      "step": 5895
+    },
+    {
+      "epoch": 2.8310940499040305,
+      "grad_norm": 0.4255916066845973,
+      "learning_rate": 3.128332740846072e-06,
+      "loss": 0.2459,
+      "mean_token_accuracy": 0.9094623148441314,
+      "step": 5900
+    },
+    {
+      "epoch": 2.8334932821497123,
+      "grad_norm": 0.43813793048191685,
+      "learning_rate": 3.0838961962317812e-06,
+      "loss": 0.2567,
+      "mean_token_accuracy": 0.9055675387382507,
+      "step": 5905
+    },
+    {
+      "epoch": 2.8358925143953932,
+      "grad_norm": 0.4209655933645841,
+      "learning_rate": 3.03945965161749e-06,
+      "loss": 0.2541,
+      "mean_token_accuracy": 0.9070176243782043,
+      "step": 5910
+    },
+    {
+      "epoch": 2.838291746641075,
+      "grad_norm": 0.4428447675746139,
+      "learning_rate": 2.9950231070031994e-06,
+      "loss": 0.2586,
+      "mean_token_accuracy": 0.9049026608467102,
+      "step": 5915
+    },
+    {
+      "epoch": 2.840690978886756,
+      "grad_norm": 0.4404991474851633,
+      "learning_rate": 2.9505865623889087e-06,
+      "loss": 0.2584,
+      "mean_token_accuracy": 0.9048484981060028,
+      "step": 5920
+    },
+    {
+      "epoch": 2.843090211132438,
+      "grad_norm": 0.4440934912962013,
+      "learning_rate": 2.906150017774618e-06,
+      "loss": 0.2598,
+      "mean_token_accuracy": 0.9047917068004608,
+      "step": 5925
+    },
+    {
+      "epoch": 2.845489443378119,
+      "grad_norm": 0.4453569173499868,
+      "learning_rate": 2.861713473160327e-06,
+      "loss": 0.262,
+      "mean_token_accuracy": 0.9035783410072327,
+      "step": 5930
+    },
+    {
+      "epoch": 2.8478886756238007,
+      "grad_norm": 0.42075498659491584,
+      "learning_rate": 2.8172769285460367e-06,
+      "loss": 0.2535,
+      "mean_token_accuracy": 0.907467657327652,
+      "step": 5935
+    },
+    {
+      "epoch": 2.8502879078694816,
+      "grad_norm": 0.4463099961746632,
+      "learning_rate": 2.7728403839317456e-06,
+      "loss": 0.2523,
+      "mean_token_accuracy": 0.9073926568031311,
+      "step": 5940
+    },
+    {
+      "epoch": 2.852687140115163,
+      "grad_norm": 0.4270911922738369,
+      "learning_rate": 2.7284038393174545e-06,
+      "loss": 0.248,
+      "mean_token_accuracy": 0.9087224245071411,
+      "step": 5945
+    },
+    {
+      "epoch": 2.8550863723608444,
+      "grad_norm": 0.4420467470565063,
+      "learning_rate": 2.683967294703164e-06,
+      "loss": 0.2544,
+      "mean_token_accuracy": 0.9067691683769226,
+      "step": 5950
+    },
+    {
+      "epoch": 2.857485604606526,
+      "grad_norm": 0.45704237431731065,
+      "learning_rate": 2.639530750088873e-06,
+      "loss": 0.2532,
+      "mean_token_accuracy": 0.9069824755191803,
+      "step": 5955
+    },
+    {
+      "epoch": 2.859884836852207,
+      "grad_norm": 0.4293486256624201,
+      "learning_rate": 2.5950942054745824e-06,
+      "loss": 0.2571,
+      "mean_token_accuracy": 0.9057079017162323,
+      "step": 5960
+    },
+    {
+      "epoch": 2.8622840690978886,
+      "grad_norm": 0.4474552765618025,
+      "learning_rate": 2.5506576608602917e-06,
+      "loss": 0.2439,
+      "mean_token_accuracy": 0.9103707253932953,
+      "step": 5965
+    },
+    {
+      "epoch": 2.86468330134357,
+      "grad_norm": 0.4626661887370989,
+      "learning_rate": 2.506221116246001e-06,
+      "loss": 0.2588,
+      "mean_token_accuracy": 0.9054373741149903,
+      "step": 5970
+    },
+    {
+      "epoch": 2.8670825335892514,
+      "grad_norm": 0.44243947426860386,
+      "learning_rate": 2.46178457163171e-06,
+      "loss": 0.2511,
+      "mean_token_accuracy": 0.9077575266361236,
+      "step": 5975
+    },
+    {
+      "epoch": 2.8694817658349328,
+      "grad_norm": 0.4297876711361223,
+      "learning_rate": 2.417348027017419e-06,
+      "loss": 0.2598,
+      "mean_token_accuracy": 0.9047979533672332,
+      "step": 5980
+    },
+    {
+      "epoch": 2.871880998080614,
+      "grad_norm": 0.4522470971407196,
+      "learning_rate": 2.3729114824031285e-06,
+      "loss": 0.2515,
+      "mean_token_accuracy": 0.9076629757881165,
+      "step": 5985
+    },
+    {
+      "epoch": 2.8742802303262955,
+      "grad_norm": 0.4595793236712429,
+      "learning_rate": 2.3284749377888374e-06,
+      "loss": 0.2635,
+      "mean_token_accuracy": 0.903690081834793,
+      "step": 5990
+    },
+    {
+      "epoch": 2.876679462571977,
+      "grad_norm": 0.4346630104029628,
+      "learning_rate": 2.2840383931745467e-06,
+      "loss": 0.2511,
+      "mean_token_accuracy": 0.9076651513576508,
+      "step": 5995
+    },
+    {
+      "epoch": 2.8790786948176583,
+      "grad_norm": 0.45090891263086474,
+      "learning_rate": 2.239601848560256e-06,
+      "loss": 0.2529,
+      "mean_token_accuracy": 0.9068621635437012,
+      "step": 6000
+    },
+    {
+      "epoch": 2.8814779270633397,
+      "grad_norm": 0.4124115306908796,
+      "learning_rate": 2.1951653039459653e-06,
+      "loss": 0.2496,
+      "mean_token_accuracy": 0.9084739625453949,
+      "step": 6005
+    },
+    {
+      "epoch": 2.883877159309021,
+      "grad_norm": 0.447834623689059,
+      "learning_rate": 2.1507287593316742e-06,
+      "loss": 0.2546,
+      "mean_token_accuracy": 0.906530886888504,
+      "step": 6010
+    },
+    {
+      "epoch": 2.8862763915547025,
+      "grad_norm": 0.4196910908382261,
+      "learning_rate": 2.106292214717384e-06,
+      "loss": 0.2557,
+      "mean_token_accuracy": 0.9063535273075104,
+      "step": 6015
+    },
+    {
+      "epoch": 2.888675623800384,
+      "grad_norm": 0.42896879529787013,
+      "learning_rate": 2.061855670103093e-06,
+      "loss": 0.2504,
+      "mean_token_accuracy": 0.9081020653247833,
+      "step": 6020
+    },
+    {
+      "epoch": 2.8910748560460653,
+      "grad_norm": 0.4349599465103453,
+      "learning_rate": 2.017419125488802e-06,
+      "loss": 0.2555,
+      "mean_token_accuracy": 0.9059644281864166,
+      "step": 6025
+    },
+    {
+      "epoch": 2.8934740882917467,
+      "grad_norm": 0.41251060268351214,
+      "learning_rate": 1.9729825808745115e-06,
+      "loss": 0.2472,
+      "mean_token_accuracy": 0.9089505612850189,
+      "step": 6030
+    },
+    {
+      "epoch": 2.895873320537428,
+      "grad_norm": 0.437831283605074,
+      "learning_rate": 1.9285460362602203e-06,
+      "loss": 0.252,
+      "mean_token_accuracy": 0.9072934269905091,
+      "step": 6035
+    },
+    {
+      "epoch": 2.8982725527831095,
+      "grad_norm": 0.45750950782775357,
+      "learning_rate": 1.8841094916459297e-06,
+      "loss": 0.2538,
+      "mean_token_accuracy": 0.9064003944396972,
+      "step": 6040
+    },
+    {
+      "epoch": 2.900671785028791,
+      "grad_norm": 0.44178784137868615,
+      "learning_rate": 1.8396729470316388e-06,
+      "loss": 0.2493,
+      "mean_token_accuracy": 0.908105194568634,
+      "step": 6045
+    },
+    {
+      "epoch": 2.9030710172744723,
+      "grad_norm": 0.4285607000273115,
+      "learning_rate": 1.7952364024173483e-06,
+      "loss": 0.2509,
+      "mean_token_accuracy": 0.9079598605632782,
+      "step": 6050
+    },
+    {
+      "epoch": 2.9054702495201536,
+      "grad_norm": 0.47649370622698345,
+      "learning_rate": 1.7507998578030572e-06,
+      "loss": 0.2466,
+      "mean_token_accuracy": 0.9093623101711273,
+      "step": 6055
+    },
+    {
+      "epoch": 2.907869481765835,
+      "grad_norm": 0.42412747407093854,
+      "learning_rate": 1.7063633131887667e-06,
+      "loss": 0.2564,
+      "mean_token_accuracy": 0.906291025876999,
+      "step": 6060
+    },
+    {
+      "epoch": 2.9102687140115164,
+      "grad_norm": 0.436149863588771,
+      "learning_rate": 1.6619267685744758e-06,
+      "loss": 0.2541,
+      "mean_token_accuracy": 0.9071348130702972,
+      "step": 6065
+    },
+    {
+      "epoch": 2.912667946257198,
+      "grad_norm": 0.4369219283799936,
+      "learning_rate": 1.617490223960185e-06,
+      "loss": 0.2491,
+      "mean_token_accuracy": 0.9085583508014679,
+      "step": 6070
+    },
+    {
+      "epoch": 2.915067178502879,
+      "grad_norm": 0.437785169638069,
+      "learning_rate": 1.5730536793458942e-06,
+      "loss": 0.248,
+      "mean_token_accuracy": 0.9086403906345367,
+      "step": 6075
+    },
+    {
+      "epoch": 2.9174664107485606,
+      "grad_norm": 0.43690114298680655,
+      "learning_rate": 1.5286171347316033e-06,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.907256692647934,
+      "step": 6080
+    },
+    {
+      "epoch": 2.919865642994242,
+      "grad_norm": 0.4494829201863112,
+      "learning_rate": 1.4841805901173126e-06,
+      "loss": 0.2562,
+      "mean_token_accuracy": 0.9058855175971985,
+      "step": 6085
+    },
+    {
+      "epoch": 2.9222648752399234,
+      "grad_norm": 0.4209094239321652,
+      "learning_rate": 1.439744045503022e-06,
+      "loss": 0.2517,
+      "mean_token_accuracy": 0.9072973370552063,
+      "step": 6090
+    },
+    {
+      "epoch": 2.9246641074856043,
+      "grad_norm": 0.4382591677968272,
+      "learning_rate": 1.3953075008887308e-06,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.9073457717895508,
+      "step": 6095
+    },
+    {
+      "epoch": 2.927063339731286,
+      "grad_norm": 0.42827302601483624,
+      "learning_rate": 1.35087095627444e-06,
+      "loss": 0.2563,
+      "mean_token_accuracy": 0.9061714768409729,
+      "step": 6100
+    },
+    {
+      "epoch": 2.929462571976967,
+      "grad_norm": 0.447579716688551,
+      "learning_rate": 1.3064344116601494e-06,
+      "loss": 0.2553,
+      "mean_token_accuracy": 0.9065722763538361,
+      "step": 6105
+    },
+    {
+      "epoch": 2.931861804222649,
+      "grad_norm": 0.42827536998813603,
+      "learning_rate": 1.2619978670458585e-06,
+      "loss": 0.2476,
+      "mean_token_accuracy": 0.909209954738617,
+      "step": 6110
+    },
+    {
+      "epoch": 2.93426103646833,
+      "grad_norm": 0.4540116982203758,
+      "learning_rate": 1.2175613224315678e-06,
+      "loss": 0.2546,
+      "mean_token_accuracy": 0.9069316864013672,
+      "step": 6115
+    },
+    {
+      "epoch": 2.9366602687140118,
+      "grad_norm": 0.4374462373091291,
+      "learning_rate": 1.173124777817277e-06,
+      "loss": 0.2533,
+      "mean_token_accuracy": 0.9067832350730896,
+      "step": 6120
+    },
+    {
+      "epoch": 2.9390595009596927,
+      "grad_norm": 0.41988101865620747,
+      "learning_rate": 1.1286882332029862e-06,
+      "loss": 0.25,
+      "mean_token_accuracy": 0.9078817367553711,
+      "step": 6125
+    },
+    {
+      "epoch": 2.9414587332053745,
+      "grad_norm": 0.44991139970047894,
+      "learning_rate": 1.0842516885886955e-06,
+      "loss": 0.2526,
+      "mean_token_accuracy": 0.9072129487991333,
+      "step": 6130
+    },
+    {
+      "epoch": 2.9438579654510555,
+      "grad_norm": 0.4322545174476245,
+      "learning_rate": 1.0398151439744046e-06,
+      "loss": 0.2452,
+      "mean_token_accuracy": 0.909875613451004,
+      "step": 6135
+    },
+    {
+      "epoch": 2.946257197696737,
+      "grad_norm": 0.44606631531514007,
+      "learning_rate": 9.953785993601137e-07,
+      "loss": 0.2499,
+      "mean_token_accuracy": 0.9080965995788575,
+      "step": 6140
+    },
+    {
+      "epoch": 2.9486564299424183,
+      "grad_norm": 0.4174486331140099,
+      "learning_rate": 9.509420547458229e-07,
+      "loss": 0.2461,
+      "mean_token_accuracy": 0.9096888959407806,
+      "step": 6145
+    },
+    {
+      "epoch": 2.9510556621880997,
+      "grad_norm": 0.4314477007269236,
+      "learning_rate": 9.065055101315321e-07,
+      "loss": 0.254,
+      "mean_token_accuracy": 0.9069934070110321,
+      "step": 6150
+    },
+    {
+      "epoch": 2.953454894433781,
+      "grad_norm": 0.42487353340805534,
+      "learning_rate": 8.620689655172415e-07,
+      "loss": 0.242,
+      "mean_token_accuracy": 0.9112311661243438,
+      "step": 6155
+    },
+    {
+      "epoch": 2.9558541266794625,
+      "grad_norm": 0.45994818847778246,
+      "learning_rate": 8.176324209029507e-07,
+      "loss": 0.2509,
+      "mean_token_accuracy": 0.9082200407981873,
+      "step": 6160
+    },
+    {
+      "epoch": 2.958253358925144,
+      "grad_norm": 0.4334130334082495,
+      "learning_rate": 7.731958762886599e-07,
+      "loss": 0.247,
+      "mean_token_accuracy": 0.9093940913677215,
+      "step": 6165
+    },
+    {
+      "epoch": 2.9606525911708252,
+      "grad_norm": 0.44597082395725934,
+      "learning_rate": 7.28759331674369e-07,
+      "loss": 0.2533,
+      "mean_token_accuracy": 0.9066949546337127,
+      "step": 6170
+    },
+    {
+      "epoch": 2.9630518234165066,
+      "grad_norm": 0.4557245109030177,
+      "learning_rate": 6.843227870600783e-07,
+      "loss": 0.2553,
+      "mean_token_accuracy": 0.9065178632736206,
+      "step": 6175
+    },
+    {
+      "epoch": 2.965451055662188,
+      "grad_norm": 0.4478527557349485,
+      "learning_rate": 6.398862424457875e-07,
+      "loss": 0.2445,
+      "mean_token_accuracy": 0.9099810898303986,
+      "step": 6180
+    },
+    {
+      "epoch": 2.9678502879078694,
+      "grad_norm": 0.41646361122311465,
+      "learning_rate": 5.954496978314967e-07,
+      "loss": 0.2449,
+      "mean_token_accuracy": 0.9099904656410217,
+      "step": 6185
+    },
+    {
+      "epoch": 2.970249520153551,
+      "grad_norm": 0.40765652811684916,
+      "learning_rate": 5.510131532172059e-07,
+      "loss": 0.2459,
+      "mean_token_accuracy": 0.9095560610294342,
+      "step": 6190
+    },
+    {
+      "epoch": 2.972648752399232,
+      "grad_norm": 0.425191482050169,
+      "learning_rate": 5.06576608602915e-07,
+      "loss": 0.2544,
+      "mean_token_accuracy": 0.9066918194293976,
+      "step": 6195
+    },
+    {
+      "epoch": 2.9750479846449136,
+      "grad_norm": 0.43069330230599245,
+      "learning_rate": 4.6214006398862424e-07,
+      "loss": 0.2597,
+      "mean_token_accuracy": 0.9043260455131531,
+      "step": 6200
+    },
+    {
+      "epoch": 2.977447216890595,
+      "grad_norm": 0.4344718819179382,
+      "learning_rate": 4.177035193743335e-07,
+      "loss": 0.2452,
+      "mean_token_accuracy": 0.9098771750926972,
+      "step": 6205
+    },
+    {
+      "epoch": 2.9798464491362764,
+      "grad_norm": 0.4325317501106857,
+      "learning_rate": 3.7326697476004265e-07,
+      "loss": 0.2505,
+      "mean_token_accuracy": 0.907985383272171,
+      "step": 6210
+    },
+    {
+      "epoch": 2.982245681381958,
+      "grad_norm": 0.4126353892608848,
+      "learning_rate": 3.288304301457519e-07,
+      "loss": 0.2437,
+      "mean_token_accuracy": 0.9105557441711426,
+      "step": 6215
+    },
+    {
+      "epoch": 2.984644913627639,
+      "grad_norm": 0.4173259485121355,
+      "learning_rate": 2.8439388553146106e-07,
+      "loss": 0.2484,
+      "mean_token_accuracy": 0.9091325998306274,
+      "step": 6220
+    },
+    {
+      "epoch": 2.9870441458733206,
+      "grad_norm": 0.43216646266753433,
+      "learning_rate": 2.399573409171703e-07,
+      "loss": 0.2476,
+      "mean_token_accuracy": 0.9084903717041015,
+      "step": 6225
+    },
+    {
+      "epoch": 2.989443378119002,
+      "grad_norm": 0.44710775517043494,
+      "learning_rate": 1.955207963028795e-07,
+      "loss": 0.2503,
+      "mean_token_accuracy": 0.9080036342144012,
+      "step": 6230
+    },
+    {
+      "epoch": 2.9918426103646834,
+      "grad_norm": 0.45620131139782905,
+      "learning_rate": 1.510842516885887e-07,
+      "loss": 0.2535,
+      "mean_token_accuracy": 0.9065212190151215,
+      "step": 6235
+    },
+    {
+      "epoch": 2.9942418426103647,
+      "grad_norm": 0.4430030123252174,
+      "learning_rate": 1.0664770707429792e-07,
+      "loss": 0.2494,
+      "mean_token_accuracy": 0.908391398191452,
+      "step": 6240
+    },
+    {
+      "epoch": 2.996641074856046,
+      "grad_norm": 0.4180541397263229,
+      "learning_rate": 6.221116246000711e-08,
+      "loss": 0.2481,
+      "mean_token_accuracy": 0.9089529037475585,
+      "step": 6245
+    },
+    {
+      "epoch": 2.9990403071017275,
+      "grad_norm": 0.43586407533469335,
+      "learning_rate": 1.7774617845716316e-08,
+      "loss": 0.2476,
+      "mean_token_accuracy": 0.9093365132808685,
+      "step": 6250
+    },
+    {
+      "epoch": 3.0,
+      "mean_token_accuracy": 0.909225344657898,
+      "step": 6252,
+      "total_flos": 2553013234827264.0,
+      "train_loss": 0.4164313976191132,
+      "train_runtime": 41448.963,
+      "train_samples_per_second": 2.413,
+      "train_steps_per_second": 0.151
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 6252,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2553013234827264.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}