diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4606 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999074845036544,
+  "eval_steps": 100,
+  "global_step": 2702,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.001850309926912758,
+      "grad_norm": 1.5234375,
+      "learning_rate": 3.690036900369004e-07,
+      "loss": 1.0642,
+      "mean_token_accuracy": 0.7239072039072036,
+      "step": 5
+    },
+    {
+      "epoch": 0.003700619853825516,
+      "grad_norm": 1.3046875,
+      "learning_rate": 7.380073800738008e-07,
+      "loss": 1.1073,
+      "mean_token_accuracy": 0.7138827838827837,
+      "step": 10
+    },
+    {
+      "epoch": 0.005550929780738274,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.1070110701107011e-06,
+      "loss": 1.1028,
+      "mean_token_accuracy": 0.7148901098901097,
+      "step": 15
+    },
+    {
+      "epoch": 0.007401239707651032,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.4760147601476015e-06,
+      "loss": 1.0562,
+      "mean_token_accuracy": 0.7275457875457875,
+      "step": 20
+    },
+    {
+      "epoch": 0.00925154963456379,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.845018450184502e-06,
+      "loss": 1.0957,
+      "mean_token_accuracy": 0.7136141636141635,
+      "step": 25
+    },
+    {
+      "epoch": 0.011101859561476548,
+      "grad_norm": 1.3046875,
+      "learning_rate": 2.2140221402214023e-06,
+      "loss": 1.1284,
+      "mean_token_accuracy": 0.7092185592185591,
+      "step": 30
+    },
+    {
+      "epoch": 0.012952169488389306,
+      "grad_norm": 1.3984375,
+      "learning_rate": 2.5830258302583027e-06,
+      "loss": 1.1127,
+      "mean_token_accuracy": 0.7139438339438338,
+      "step": 35
+    },
+    {
+      "epoch": 0.014802479415302064,
+      "grad_norm": 1.2890625,
+      "learning_rate": 2.952029520295203e-06,
+      "loss": 1.0654,
+      "mean_token_accuracy": 0.7238278388278386,
+      "step": 40
+    },
+    {
+      "epoch": 0.01665278934221482,
+      "grad_norm": 1.234375,
+      "learning_rate": 3.3210332103321034e-06,
+      "loss": 1.1178,
+      "mean_token_accuracy": 0.7124236874236874,
+      "step": 45
+    },
+    {
+      "epoch": 0.01850309926912758,
+      "grad_norm": 1.2109375,
+      "learning_rate": 3.690036900369004e-06,
+      "loss": 1.084,
+      "mean_token_accuracy": 0.718971583971584,
+      "step": 50
+    },
+    {
+      "epoch": 0.020353409196040336,
+      "grad_norm": 1.1875,
+      "learning_rate": 4.059040590405905e-06,
+      "loss": 1.1139,
+      "mean_token_accuracy": 0.7134981684981684,
+      "step": 55
+    },
+    {
+      "epoch": 0.022203719122953096,
+      "grad_norm": 1.1640625,
+      "learning_rate": 4.428044280442805e-06,
+      "loss": 1.0929,
+      "mean_token_accuracy": 0.7163919413919414,
+      "step": 60
+    },
+    {
+      "epoch": 0.024054029049865852,
+      "grad_norm": 1.09375,
+      "learning_rate": 4.797047970479705e-06,
+      "loss": 1.096,
+      "mean_token_accuracy": 0.7136691086691085,
+      "step": 65
+    },
+    {
+      "epoch": 0.025904338976778612,
+      "grad_norm": 1.125,
+      "learning_rate": 5.166051660516605e-06,
+      "loss": 0.9822,
+      "mean_token_accuracy": 0.7407326007326007,
+      "step": 70
+    },
+    {
+      "epoch": 0.027754648903691368,
+      "grad_norm": 1.1171875,
+      "learning_rate": 5.535055350553506e-06,
+      "loss": 1.0772,
+      "mean_token_accuracy": 0.7176251526251525,
+      "step": 75
+    },
+    {
+      "epoch": 0.029604958830604128,
+      "grad_norm": 1.0703125,
+      "learning_rate": 5.904059040590406e-06,
+      "loss": 1.098,
+      "mean_token_accuracy": 0.7138461538461537,
+      "step": 80
+    },
+    {
+      "epoch": 0.031455268757516884,
+      "grad_norm": 1.109375,
+      "learning_rate": 6.273062730627307e-06,
+      "loss": 1.0866,
+      "mean_token_accuracy": 0.7150976800976799,
+      "step": 85
+    },
+    {
+      "epoch": 0.03330557868442964,
+      "grad_norm": 1.03125,
+      "learning_rate": 6.642066420664207e-06,
+      "loss": 1.0244,
+      "mean_token_accuracy": 0.7288766788766787,
+      "step": 90
+    },
+    {
+      "epoch": 0.0351558886113424,
+      "grad_norm": 1.0703125,
+      "learning_rate": 7.011070110701108e-06,
+      "loss": 1.0485,
+      "mean_token_accuracy": 0.7245970695970694,
+      "step": 95
+    },
+    {
+      "epoch": 0.03700619853825516,
+      "grad_norm": 0.95703125,
+      "learning_rate": 7.380073800738008e-06,
+      "loss": 1.0492,
+      "mean_token_accuracy": 0.7218559218559217,
+      "step": 100
+    },
+    {
+      "epoch": 0.03700619853825516,
+      "eval_loss": 1.0890889167785645,
+      "eval_mean_token_accuracy": 0.7098439686674981,
+      "eval_runtime": 26.1241,
+      "eval_samples_per_second": 4.938,
+      "eval_steps_per_second": 1.263,
+      "step": 100
+    },
+    {
+      "epoch": 0.038856508465167916,
+      "grad_norm": 0.90625,
+      "learning_rate": 7.749077490774908e-06,
+      "loss": 1.024,
+      "mean_token_accuracy": 0.7307387057387057,
+      "step": 105
+    },
+    {
+      "epoch": 0.04070681839208067,
+      "grad_norm": 0.91015625,
+      "learning_rate": 8.11808118081181e-06,
+      "loss": 1.0665,
+      "mean_token_accuracy": 0.7191514041514042,
+      "step": 110
+    },
+    {
+      "epoch": 0.04255712831899343,
+      "grad_norm": 0.95703125,
+      "learning_rate": 8.48708487084871e-06,
+      "loss": 1.0128,
+      "mean_token_accuracy": 0.7297191697191696,
+      "step": 115
+    },
+    {
+      "epoch": 0.04440743824590619,
+      "grad_norm": 0.94140625,
+      "learning_rate": 8.85608856088561e-06,
+      "loss": 1.0402,
+      "mean_token_accuracy": 0.7214346764346763,
+      "step": 120
+    },
+    {
+      "epoch": 0.04625774817281895,
+      "grad_norm": 0.89453125,
+      "learning_rate": 9.22509225092251e-06,
+      "loss": 1.0332,
+      "mean_token_accuracy": 0.7260256410256407,
+      "step": 125
+    },
+    {
+      "epoch": 0.048108058099731704,
+      "grad_norm": 0.83984375,
+      "learning_rate": 9.59409594095941e-06,
+      "loss": 0.9781,
+      "mean_token_accuracy": 0.7375274725274725,
+      "step": 130
+    },
+    {
+      "epoch": 0.04995836802664446,
+      "grad_norm": 0.94921875,
+      "learning_rate": 9.963099630996312e-06,
+      "loss": 1.007,
+      "mean_token_accuracy": 0.7275213675213674,
+      "step": 135
+    },
+    {
+      "epoch": 0.051808677953557224,
+      "grad_norm": 0.8125,
+      "learning_rate": 1.033210332103321e-05,
+      "loss": 1.0039,
+      "mean_token_accuracy": 0.7260073260073258,
+      "step": 140
+    },
+    {
+      "epoch": 0.05365898788046998,
+      "grad_norm": 0.828125,
+      "learning_rate": 1.0701107011070112e-05,
+      "loss": 1.0001,
+      "mean_token_accuracy": 0.7317399267399269,
+      "step": 145
+    },
+    {
+      "epoch": 0.055509297807382736,
+      "grad_norm": 0.703125,
+      "learning_rate": 1.1070110701107012e-05,
+      "loss": 1.0008,
+      "mean_token_accuracy": 0.7304945054945053,
+      "step": 150
+    },
+    {
+      "epoch": 0.05735960773429549,
+      "grad_norm": 0.8984375,
+      "learning_rate": 1.1439114391143913e-05,
+      "loss": 1.0005,
+      "mean_token_accuracy": 0.7269474969474969,
+      "step": 155
+    },
+    {
+      "epoch": 0.059209917661208256,
+      "grad_norm": 0.76171875,
+      "learning_rate": 1.1808118081180812e-05,
+      "loss": 0.9907,
+      "mean_token_accuracy": 0.7304334554334554,
+      "step": 160
+    },
+    {
+      "epoch": 0.06106022758812101,
+      "grad_norm": 0.6796875,
+      "learning_rate": 1.2177121771217713e-05,
+      "loss": 0.996,
+      "mean_token_accuracy": 0.7277838827838826,
+      "step": 165
+    },
+    {
+      "epoch": 0.06291053751503377,
+      "grad_norm": 0.68359375,
+      "learning_rate": 1.2546125461254614e-05,
+      "loss": 0.9898,
+      "mean_token_accuracy": 0.7293284493284492,
+      "step": 170
+    },
+    {
+      "epoch": 0.06476084744194653,
+      "grad_norm": 0.71875,
+      "learning_rate": 1.2915129151291515e-05,
+      "loss": 0.98,
+      "mean_token_accuracy": 0.7313614163614164,
+      "step": 175
+    },
+    {
+      "epoch": 0.06661115736885928,
+      "grad_norm": 0.7109375,
+      "learning_rate": 1.3284132841328414e-05,
+      "loss": 0.9618,
+      "mean_token_accuracy": 0.7372832722832724,
+      "step": 180
+    },
+    {
+      "epoch": 0.06846146729577204,
+      "grad_norm": 0.78515625,
+      "learning_rate": 1.3653136531365315e-05,
+      "loss": 0.9561,
+      "mean_token_accuracy": 0.7382661782661781,
+      "step": 185
+    },
+    {
+      "epoch": 0.0703117772226848,
+      "grad_norm": 0.671875,
+      "learning_rate": 1.4022140221402215e-05,
+      "loss": 0.9236,
+      "mean_token_accuracy": 0.7444200244200243,
+      "step": 190
+    },
+    {
+      "epoch": 0.07216208714959756,
+      "grad_norm": 0.7109375,
+      "learning_rate": 1.4391143911439116e-05,
+      "loss": 0.9678,
+      "mean_token_accuracy": 0.7315262515262515,
+      "step": 195
+    },
+    {
+      "epoch": 0.07401239707651032,
+      "grad_norm": 0.65625,
+      "learning_rate": 1.4760147601476015e-05,
+      "loss": 0.9769,
+      "mean_token_accuracy": 0.731929181929182,
+      "step": 200
+    },
+    {
+      "epoch": 0.07401239707651032,
+      "eval_loss": 0.9760187864303589,
+      "eval_mean_token_accuracy": 0.7278203278203277,
+      "eval_runtime": 26.1,
+      "eval_samples_per_second": 4.943,
+      "eval_steps_per_second": 1.264,
+      "step": 200
+    },
+    {
+      "epoch": 0.07586270700342307,
+      "grad_norm": 0.62890625,
+      "learning_rate": 1.5129151291512916e-05,
+      "loss": 0.9567,
+      "mean_token_accuracy": 0.7375702075702076,
+      "step": 205
+    },
+    {
+      "epoch": 0.07771301693033583,
+      "grad_norm": 0.671875,
+      "learning_rate": 1.5498154981549817e-05,
+      "loss": 0.9302,
+      "mean_token_accuracy": 0.741050061050061,
+      "step": 210
+    },
+    {
+      "epoch": 0.0795633268572486,
+      "grad_norm": 0.71484375,
+      "learning_rate": 1.5867158671586716e-05,
+      "loss": 0.9151,
+      "mean_token_accuracy": 0.7447313797313797,
+      "step": 215
+    },
+    {
+      "epoch": 0.08141363678416134,
+      "grad_norm": 0.86328125,
+      "learning_rate": 1.623616236162362e-05,
+      "loss": 1.0001,
+      "mean_token_accuracy": 0.7228144078144076,
+      "step": 220
+    },
+    {
+      "epoch": 0.08326394671107411,
+      "grad_norm": 0.6640625,
+      "learning_rate": 1.6605166051660518e-05,
+      "loss": 0.9472,
+      "mean_token_accuracy": 0.7352869352869352,
+      "step": 225
+    },
+    {
+      "epoch": 0.08511425663798686,
+      "grad_norm": 0.6484375,
+      "learning_rate": 1.697416974169742e-05,
+      "loss": 0.908,
+      "mean_token_accuracy": 0.746043956043956,
+      "step": 230
+    },
+    {
+      "epoch": 0.08696456656489962,
+      "grad_norm": 0.6640625,
+      "learning_rate": 1.734317343173432e-05,
+      "loss": 0.923,
+      "mean_token_accuracy": 0.7423565323565324,
+      "step": 235
+    },
+    {
+      "epoch": 0.08881487649181238,
+      "grad_norm": 0.62890625,
+      "learning_rate": 1.771217712177122e-05,
+      "loss": 0.9013,
+      "mean_token_accuracy": 0.7460927960927961,
+      "step": 240
+    },
+    {
+      "epoch": 0.09066518641872513,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.8081180811808117e-05,
+      "loss": 0.9575,
+      "mean_token_accuracy": 0.7327838827838828,
+      "step": 245
+    },
+    {
+      "epoch": 0.0925154963456379,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.845018450184502e-05,
+      "loss": 0.9318,
+      "mean_token_accuracy": 0.738046398046398,
+      "step": 250
+    },
+    {
+      "epoch": 0.09436580627255066,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.8819188191881922e-05,
+      "loss": 0.9054,
+      "mean_token_accuracy": 0.7462454212454211,
+      "step": 255
+    },
+    {
+      "epoch": 0.09621611619946341,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.918819188191882e-05,
+      "loss": 0.902,
+      "mean_token_accuracy": 0.7438034188034188,
+      "step": 260
+    },
+    {
+      "epoch": 0.09806642612637617,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.955719557195572e-05,
+      "loss": 0.9187,
+      "mean_token_accuracy": 0.7422344322344321,
+      "step": 265
+    },
+    {
+      "epoch": 0.09991673605328892,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.9926199261992623e-05,
+      "loss": 0.9635,
+      "mean_token_accuracy": 0.7290415140415138,
+      "step": 270
+    },
+    {
+      "epoch": 0.10176704598020168,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.9999866396188624e-05,
+      "loss": 0.9206,
+      "mean_token_accuracy": 0.7384615384615383,
+      "step": 275
+    },
+    {
+      "epoch": 0.10361735590711445,
+      "grad_norm": 0.625,
+      "learning_rate": 1.9999323636823398e-05,
+      "loss": 0.8948,
+      "mean_token_accuracy": 0.7485653235653235,
+      "step": 280
+    },
+    {
+      "epoch": 0.1054676658340272,
+      "grad_norm": 0.66015625,
+      "learning_rate": 1.9998363394309497e-05,
+      "loss": 0.8323,
+      "mean_token_accuracy": 0.7648534798534798,
+      "step": 285
+    },
+    {
+      "epoch": 0.10731797576093996,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.9996985708738146e-05,
+      "loss": 0.8243,
+      "mean_token_accuracy": 0.7656715506715509,
+      "step": 290
+    },
+    {
+      "epoch": 0.10916828568785271,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.999519063762928e-05,
+      "loss": 0.8986,
+      "mean_token_accuracy": 0.7459401709401711,
+      "step": 295
+    },
+    {
+      "epoch": 0.11101859561476547,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.9992978255929168e-05,
+      "loss": 0.9131,
+      "mean_token_accuracy": 0.7397191697191697,
+      "step": 300
+    },
+    {
+      "epoch": 0.11101859561476547,
+      "eval_loss": 0.9101393222808838,
+      "eval_mean_token_accuracy": 0.738495056142115,
+      "eval_runtime": 26.1023,
+      "eval_samples_per_second": 4.942,
+      "eval_steps_per_second": 1.264,
+      "step": 300
+    },
+    {
+      "epoch": 0.11286890554167824,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.999034865600726e-05,
+      "loss": 0.9541,
+      "mean_token_accuracy": 0.7295726495726494,
+      "step": 305
+    },
+    {
+      "epoch": 0.11471921546859098,
+      "grad_norm": 0.6484375,
+      "learning_rate": 1.9987301947652354e-05,
+      "loss": 0.8719,
+      "mean_token_accuracy": 0.7503113553113553,
+      "step": 310
+    },
+    {
+      "epoch": 0.11656952539550375,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.998383825806799e-05,
+      "loss": 0.8776,
+      "mean_token_accuracy": 0.74996336996337,
+      "step": 315
+    },
+    {
+      "epoch": 0.11841983532241651,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.9979957731867143e-05,
+      "loss": 0.8599,
+      "mean_token_accuracy": 0.7538766788766791,
+      "step": 320
+    },
+    {
+      "epoch": 0.12027014524932926,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.9975660531066215e-05,
+      "loss": 0.9302,
+      "mean_token_accuracy": 0.7337240537240539,
+      "step": 325
+    },
+    {
+      "epoch": 0.12212045517624202,
+      "grad_norm": 0.625,
+      "learning_rate": 1.9970946835078227e-05,
+      "loss": 0.8618,
+      "mean_token_accuracy": 0.7546825396825396,
+      "step": 330
+    },
+    {
+      "epoch": 0.12397076510315477,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.9965816840705355e-05,
+      "loss": 0.8338,
+      "mean_token_accuracy": 0.7603601953601955,
+      "step": 335
+    },
+    {
+      "epoch": 0.12582107503006754,
+      "grad_norm": 0.67578125,
+      "learning_rate": 1.9960270762130705e-05,
+      "loss": 0.8738,
+      "mean_token_accuracy": 0.7505334817039826,
+      "step": 340
+    },
+    {
+      "epoch": 0.1276713849569803,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.9954308830909372e-05,
+      "loss": 0.8405,
+      "mean_token_accuracy": 0.7583272283272284,
+      "step": 345
+    },
+    {
+      "epoch": 0.12952169488389306,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.9947931295958778e-05,
+      "loss": 0.8515,
+      "mean_token_accuracy": 0.7538400488400487,
+      "step": 350
+    },
+    {
+      "epoch": 0.1313720048108058,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.9941138423548266e-05,
+      "loss": 0.8548,
+      "mean_token_accuracy": 0.7522832722832723,
+      "step": 355
+    },
+    {
+      "epoch": 0.13322231473771856,
+      "grad_norm": 0.625,
+      "learning_rate": 1.9933930497287996e-05,
+      "loss": 0.8532,
+      "mean_token_accuracy": 0.7545360195360196,
+      "step": 360
+    },
+    {
+      "epoch": 0.13507262466463132,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.9926307818117098e-05,
+      "loss": 0.8515,
+      "mean_token_accuracy": 0.7543589743589741,
+      "step": 365
+    },
+    {
+      "epoch": 0.1369229345915441,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.9918270704291104e-05,
+      "loss": 0.8371,
+      "mean_token_accuracy": 0.7585164835164836,
+      "step": 370
+    },
+    {
+      "epoch": 0.13877324451845685,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.9909819491368677e-05,
+      "loss": 0.7795,
+      "mean_token_accuracy": 0.7738705738705741,
+      "step": 375
+    },
+    {
+      "epoch": 0.1406235544453696,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.990095453219757e-05,
+      "loss": 0.8354,
+      "mean_token_accuracy": 0.7600793650793649,
+      "step": 380
+    },
+    {
+      "epoch": 0.14247386437228235,
+      "grad_norm": 0.625,
+      "learning_rate": 1.989167619689993e-05,
+      "loss": 0.8644,
+      "mean_token_accuracy": 0.7514835164835164,
+      "step": 385
+    },
+    {
+      "epoch": 0.1443241742991951,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.988198487285682e-05,
+      "loss": 0.9114,
+      "mean_token_accuracy": 0.7397191697191696,
+      "step": 390
+    },
+    {
+      "epoch": 0.14617448422610788,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.9871880964692055e-05,
+      "loss": 0.8587,
+      "mean_token_accuracy": 0.7514285714285714,
+      "step": 395
+    },
+    {
+      "epoch": 0.14802479415302064,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.9861364894255306e-05,
+      "loss": 0.8299,
+      "mean_token_accuracy": 0.7600976800976802,
+      "step": 400
+    },
+    {
+      "epoch": 0.14802479415302064,
+      "eval_loss": 0.8789100050926208,
+      "eval_mean_token_accuracy": 0.7443620649503006,
+      "eval_runtime": 26.1347,
+      "eval_samples_per_second": 4.936,
+      "eval_steps_per_second": 1.263,
+      "step": 400
+    },
+    {
+      "epoch": 0.1498751040799334,
+      "grad_norm": 0.6171875,
+      "learning_rate": 1.985043710060449e-05,
+      "loss": 0.8761,
+      "mean_token_accuracy": 0.7473626373626374,
+      "step": 405
+    },
+    {
+      "epoch": 0.15172541400684614,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.9839098039987435e-05,
+      "loss": 0.827,
+      "mean_token_accuracy": 0.7622649572649576,
+      "step": 410
+    },
+    {
+      "epoch": 0.1535757239337589,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.9827348185822834e-05,
+      "loss": 0.8387,
+      "mean_token_accuracy": 0.7583516483516484,
+      "step": 415
+    },
+    {
+      "epoch": 0.15542603386067166,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.981518802868048e-05,
+      "loss": 0.8549,
+      "mean_token_accuracy": 0.7524542124542124,
+      "step": 420
+    },
+    {
+      "epoch": 0.15727634378758443,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.9802618076260784e-05,
+      "loss": 0.8172,
+      "mean_token_accuracy": 0.7641880341880342,
+      "step": 425
+    },
+    {
+      "epoch": 0.1591266537144972,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.9789638853373563e-05,
+      "loss": 0.8304,
+      "mean_token_accuracy": 0.7597374847374848,
+      "step": 430
+    },
+    {
+      "epoch": 0.16097696364140993,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.9776250901916168e-05,
+      "loss": 0.9115,
+      "mean_token_accuracy": 0.7372527472527475,
+      "step": 435
+    },
+    {
+      "epoch": 0.1628272735683227,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.9762454780850807e-05,
+      "loss": 0.8382,
+      "mean_token_accuracy": 0.7568253968253968,
+      "step": 440
+    },
+    {
+      "epoch": 0.16467758349523545,
+      "grad_norm": 0.6328125,
+      "learning_rate": 1.9748251066181247e-05,
+      "loss": 0.8118,
+      "mean_token_accuracy": 0.7624908424908425,
+      "step": 445
+    },
+    {
+      "epoch": 0.16652789342214822,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.973364035092875e-05,
+      "loss": 0.8812,
+      "mean_token_accuracy": 0.7446703296703295,
+      "step": 450
+    },
+    {
+      "epoch": 0.16837820334906098,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.971862324510732e-05,
+      "loss": 0.8931,
+      "mean_token_accuracy": 0.7403174603174604,
+      "step": 455
+    },
+    {
+      "epoch": 0.17022851327597371,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.9703200375698223e-05,
+      "loss": 0.8784,
+      "mean_token_accuracy": 0.7463064713064713,
+      "step": 460
+    },
+    {
+      "epoch": 0.17207882320288648,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.968737238662382e-05,
+      "loss": 0.7957,
+      "mean_token_accuracy": 0.7680525030525034,
+      "step": 465
+    },
+    {
+      "epoch": 0.17392913312979924,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.9671139938720678e-05,
+      "loss": 0.8147,
+      "mean_token_accuracy": 0.7610744810744812,
+      "step": 470
+    },
+    {
+      "epoch": 0.175779443056712,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.9654503709711984e-05,
+      "loss": 0.8562,
+      "mean_token_accuracy": 0.7506593406593407,
+      "step": 475
+    },
+    {
+      "epoch": 0.17762975298362477,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.963746439417924e-05,
+      "loss": 0.8001,
+      "mean_token_accuracy": 0.7670940170940171,
+      "step": 480
+    },
+    {
+      "epoch": 0.1794800629105375,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.962002270353328e-05,
+      "loss": 0.8008,
+      "mean_token_accuracy": 0.7670390720390722,
+      "step": 485
+    },
+    {
+      "epoch": 0.18133037283745027,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.960217936598454e-05,
+      "loss": 0.7907,
+      "mean_token_accuracy": 0.7678693528693529,
+      "step": 490
+    },
+    {
+      "epoch": 0.18318068276436303,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.958393512651269e-05,
+      "loss": 0.7641,
+      "mean_token_accuracy": 0.775854700854701,
+      "step": 495
+    },
+    {
+      "epoch": 0.1850309926912758,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.956529074683551e-05,
+      "loss": 0.8435,
+      "mean_token_accuracy": 0.7561477411477411,
+      "step": 500
+    },
+    {
+      "epoch": 0.1850309926912758,
+      "eval_loss": 0.8611128926277161,
+      "eval_mean_token_accuracy": 0.7471516500928265,
+      "eval_runtime": 26.0612,
+      "eval_samples_per_second": 4.95,
+      "eval_steps_per_second": 1.266,
+      "step": 500
+    },
+    {
+      "epoch": 0.18688130261818856,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.9546247005377065e-05,
+      "loss": 0.8145,
+      "mean_token_accuracy": 0.7607081807081808,
+      "step": 505
+    },
+    {
+      "epoch": 0.18873161254510132,
+      "grad_norm": 0.51953125,
+      "learning_rate": 1.952680469723526e-05,
+      "loss": 0.8148,
+      "mean_token_accuracy": 0.7623504273504274,
+      "step": 510
+    },
+    {
+      "epoch": 0.19058192247201405,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.9506964634148597e-05,
+      "loss": 0.8333,
+      "mean_token_accuracy": 0.7563797313797316,
+      "step": 515
+    },
+    {
+      "epoch": 0.19243223239892682,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.9486727644462306e-05,
+      "loss": 0.845,
+      "mean_token_accuracy": 0.7572832722832723,
+      "step": 520
+    },
+    {
+      "epoch": 0.19428254232583958,
+      "grad_norm": 0.62109375,
+      "learning_rate": 1.9466094573093744e-05,
+      "loss": 0.7821,
+      "mean_token_accuracy": 0.7752686202686203,
+      "step": 525
+    },
+    {
+      "epoch": 0.19613285225275234,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.9445066281497144e-05,
+      "loss": 0.7933,
+      "mean_token_accuracy": 0.7684310134310134,
+      "step": 530
+    },
+    {
+      "epoch": 0.1979831621796651,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.9423643647627625e-05,
+      "loss": 0.842,
+      "mean_token_accuracy": 0.7544505494505496,
+      "step": 535
+    },
+    {
+      "epoch": 0.19983347210657784,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.940182756590454e-05,
+      "loss": 0.816,
+      "mean_token_accuracy": 0.7597557997557999,
+      "step": 540
+    },
+    {
+      "epoch": 0.2016837820334906,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.9379618947174155e-05,
+      "loss": 0.8157,
+      "mean_token_accuracy": 0.7635531135531137,
+      "step": 545
+    },
+    {
+      "epoch": 0.20353409196040337,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.935701871867158e-05,
+      "loss": 0.8061,
+      "mean_token_accuracy": 0.7631379731379734,
+      "step": 550
+    },
+    {
+      "epoch": 0.20538440188731613,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.9334027823982103e-05,
+      "loss": 0.806,
+      "mean_token_accuracy": 0.7636942216942215,
+      "step": 555
+    },
+    {
+      "epoch": 0.2072347118142289,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.9310647223001752e-05,
+      "loss": 0.8684,
+      "mean_token_accuracy": 0.7475396825396826,
+      "step": 560
+    },
+    {
+      "epoch": 0.20908502174114163,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.9286877891897244e-05,
+      "loss": 0.7634,
+      "mean_token_accuracy": 0.776892551892552,
+      "step": 565
+    },
+    {
+      "epoch": 0.2109353316680544,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.9262720823065217e-05,
+      "loss": 0.8631,
+      "mean_token_accuracy": 0.7469902319902321,
+      "step": 570
+    },
+    {
+      "epoch": 0.21278564159496716,
+      "grad_norm": 0.51171875,
+      "learning_rate": 1.923817702509081e-05,
+      "loss": 0.8025,
+      "mean_token_accuracy": 0.7656715506715509,
+      "step": 575
+    },
+    {
+      "epoch": 0.21463595152187992,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.9213247522705532e-05,
+      "loss": 0.8464,
+      "mean_token_accuracy": 0.755897435897436,
+      "step": 580
+    },
+    {
+      "epoch": 0.21648626144879268,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.9187933356744504e-05,
+      "loss": 0.8072,
+      "mean_token_accuracy": 0.7652136752136752,
+      "step": 585
+    },
+    {
+      "epoch": 0.21833657137570542,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.9162235584102973e-05,
+      "loss": 0.8198,
+      "mean_token_accuracy": 0.76008547008547,
+      "step": 590
+    },
+    {
+      "epoch": 0.22018688130261818,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.9136155277692215e-05,
+      "loss": 0.8,
+      "mean_token_accuracy": 0.7653357753357756,
+      "step": 595
+    },
+    {
+      "epoch": 0.22203719122953094,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.9109693526394722e-05,
+      "loss": 0.8058,
+      "mean_token_accuracy": 0.7617155067155068,
+      "step": 600
+    },
+    {
+      "epoch": 0.22203719122953094,
+      "eval_loss": 0.8499741554260254,
+      "eval_mean_token_accuracy": 0.748883034177152,
+      "eval_runtime": 26.0615,
+      "eval_samples_per_second": 4.95,
+      "eval_steps_per_second": 1.266,
+      "step": 600
+    },
+    {
+      "epoch": 0.2238875011564437,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.9082851435018743e-05,
+      "loss": 0.8373,
+      "mean_token_accuracy": 0.7574297924297924,
+      "step": 605
+    },
+    {
+      "epoch": 0.22573781108335647,
+      "grad_norm": 0.5078125,
+      "learning_rate": 1.905563012425216e-05,
+      "loss": 0.8007,
+      "mean_token_accuracy": 0.7656776556776557,
+      "step": 610
+    },
+    {
+      "epoch": 0.2275881210102692,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.9028030730615696e-05,
+      "loss": 0.8416,
+      "mean_token_accuracy": 0.7550732600732599,
+      "step": 615
+    },
+    {
+      "epoch": 0.22943843093718197,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.9000054406415467e-05,
+      "loss": 0.8015,
+      "mean_token_accuracy": 0.766013431013431,
+      "step": 620
+    },
+    {
+      "epoch": 0.23128874086409473,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.897170231969486e-05,
+      "loss": 0.8184,
+      "mean_token_accuracy": 0.7606654456654457,
+      "step": 625
+    },
+    {
+      "epoch": 0.2331390507910075,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.8942975654185788e-05,
+      "loss": 0.8058,
+      "mean_token_accuracy": 0.7649084249084249,
+      "step": 630
+    },
+    {
+      "epoch": 0.23498936071792026,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.8913875609259246e-05,
+      "loss": 0.7989,
+      "mean_token_accuracy": 0.7661050061050061,
+      "step": 635
+    },
+    {
+      "epoch": 0.23683967064483302,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.8884403399875252e-05,
+      "loss": 0.8244,
+      "mean_token_accuracy": 0.7579548229548233,
+      "step": 640
+    },
+    {
+      "epoch": 0.23868998057174576,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.8854560256532098e-05,
+      "loss": 0.7568,
+      "mean_token_accuracy": 0.7783349428349429,
+      "step": 645
+    },
+    {
+      "epoch": 0.24054029049865852,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.8824347425215016e-05,
+      "loss": 0.8446,
+      "mean_token_accuracy": 0.7513308913308914,
+      "step": 650
+    },
+    {
+      "epoch": 0.24239060042557128,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.8793766167344115e-05,
+      "loss": 0.7974,
+      "mean_token_accuracy": 0.76482905982906,
+      "step": 655
+    },
+    {
+      "epoch": 0.24424091035248405,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.8762817759721735e-05,
+      "loss": 0.8104,
+      "mean_token_accuracy": 0.763052503052503,
+      "step": 660
+    },
+    {
+      "epoch": 0.2460912202793968,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.8731503494479132e-05,
+      "loss": 0.8352,
+      "mean_token_accuracy": 0.7563858363858367,
+      "step": 665
+    },
+    {
+      "epoch": 0.24794153020630955,
+      "grad_norm": 0.5,
+      "learning_rate": 1.869982467902255e-05,
+      "loss": 0.7835,
+      "mean_token_accuracy": 0.7682844932844934,
+      "step": 670
+    },
+    {
+      "epoch": 0.2497918401332223,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.8667782635978597e-05,
+      "loss": 0.8492,
+      "mean_token_accuracy": 0.7515811965811967,
+      "step": 675
+    },
+    {
+      "epoch": 0.2516421500601351,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.8635378703139066e-05,
+      "loss": 0.7594,
+      "mean_token_accuracy": 0.773962148962149,
+      "step": 680
+    },
+    {
+      "epoch": 0.25349245998704784,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.8602614233405047e-05,
+      "loss": 0.7986,
+      "mean_token_accuracy": 0.7644749694749697,
+      "step": 685
+    },
+    {
+      "epoch": 0.2553427699139606,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.8569490594730474e-05,
+      "loss": 0.8123,
+      "mean_token_accuracy": 0.7603785103785106,
+      "step": 690
+    },
+    {
+      "epoch": 0.25719307984087336,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.853600917006497e-05,
+      "loss": 0.7566,
+      "mean_token_accuracy": 0.7762026862026863,
+      "step": 695
+    },
+    {
+      "epoch": 0.2590433897677861,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.8502171357296144e-05,
+      "loss": 0.7554,
+      "mean_token_accuracy": 0.7761111111111113,
+      "step": 700
+    },
+    {
+      "epoch": 0.2590433897677861,
+      "eval_loss": 0.8421046137809753,
+      "eval_mean_token_accuracy": 0.7501704178174767,
+      "eval_runtime": 26.0623,
+      "eval_samples_per_second": 4.95,
+      "eval_steps_per_second": 1.266,
+      "step": 700
+    },
+    {
+      "epoch": 0.2608936996946989,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.8467978569191216e-05,
+      "loss": 0.815,
+      "mean_token_accuracy": 0.761892551892552,
+      "step": 705
+    },
+    {
+      "epoch": 0.2627440096216116,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.8433432233338027e-05,
+      "loss": 0.7738,
+      "mean_token_accuracy": 0.771825396825397,
+      "step": 710
+    },
+    {
+      "epoch": 0.26459431954852436,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.8398533792085436e-05,
+      "loss": 0.809,
+      "mean_token_accuracy": 0.7627655677655678,
+      "step": 715
+    },
+    {
+      "epoch": 0.2664446294754371,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.8363284702483106e-05,
+      "loss": 0.8306,
+      "mean_token_accuracy": 0.7560866910866911,
+      "step": 720
+    },
+    {
+      "epoch": 0.2682949394023499,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.832768643622067e-05,
+      "loss": 0.781,
+      "mean_token_accuracy": 0.7721184371184371,
+      "step": 725
+    },
+    {
+      "epoch": 0.27014524932926265,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.8291740479566286e-05,
+      "loss": 0.7894,
+      "mean_token_accuracy": 0.7696764346764348,
+      "step": 730
+    },
+    {
+      "epoch": 0.2719955592561754,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.825544833330457e-05,
+      "loss": 0.8008,
+      "mean_token_accuracy": 0.7653052503052502,
+      "step": 735
+    },
+    {
+      "epoch": 0.2738458691830882,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.8218811512673958e-05,
+      "loss": 0.8425,
+      "mean_token_accuracy": 0.7535836385836385,
+      "step": 740
+    },
+    {
+      "epoch": 0.27569617911000094,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.818183154730344e-05,
+      "loss": 0.807,
+      "mean_token_accuracy": 0.7621001221001222,
+      "step": 745
+    },
+    {
+      "epoch": 0.2775464890369137,
+      "grad_norm": 0.5078125,
+      "learning_rate": 1.8144509981148675e-05,
+      "loss": 0.7947,
+      "mean_token_accuracy": 0.7661050061050061,
+      "step": 750
+    },
+    {
+      "epoch": 0.27939679896382646,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.810684837242755e-05,
+      "loss": 0.8104,
+      "mean_token_accuracy": 0.7608913308913309,
+      "step": 755
+    },
+    {
+      "epoch": 0.2812471088907392,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.8068848293555118e-05,
+      "loss": 0.818,
+      "mean_token_accuracy": 0.7603174603174605,
+      "step": 760
+    },
+    {
+      "epoch": 0.28309741881765194,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.8030511331077945e-05,
+      "loss": 0.8027,
+      "mean_token_accuracy": 0.7637545787545791,
+      "step": 765
+    },
+    {
+      "epoch": 0.2849477287445647,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.799183908560787e-05,
+      "loss": 0.8525,
+      "mean_token_accuracy": 0.7526862026862027,
+      "step": 770
+    },
+    {
+      "epoch": 0.28679803867147746,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.795283317175518e-05,
+      "loss": 0.7676,
+      "mean_token_accuracy": 0.7719352869352871,
+      "step": 775
+    },
+    {
+      "epoch": 0.2886483485983902,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.7913495218061202e-05,
+      "loss": 0.8034,
+      "mean_token_accuracy": 0.7633028083028084,
+      "step": 780
+    },
+    {
+      "epoch": 0.290498658525303,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.787382686693029e-05,
+      "loss": 0.7901,
+      "mean_token_accuracy": 0.766959706959707,
+      "step": 785
+    },
+    {
+      "epoch": 0.29234896845221575,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.783382977456128e-05,
+      "loss": 0.7746,
+      "mean_token_accuracy": 0.7709951159951162,
+      "step": 790
+    },
+    {
+      "epoch": 0.2941992783791285,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.779350561087833e-05,
+      "loss": 0.839,
+      "mean_token_accuracy": 0.7553418803418805,
+      "step": 795
+    },
+    {
+      "epoch": 0.2960495883060413,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.775285605946119e-05,
+      "loss": 0.765,
+      "mean_token_accuracy": 0.7733821733821735,
+      "step": 800
+    },
+    {
+      "epoch": 0.2960495883060413,
+      "eval_loss": 0.8361122012138367,
+      "eval_mean_token_accuracy": 0.751261918908978,
+      "eval_runtime": 26.0689,
+      "eval_samples_per_second": 4.948,
+      "eval_steps_per_second": 1.266,
+      "step": 800
+    },
+    {
+      "epoch": 0.29789989823295404,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.7711882817474922e-05,
+      "loss": 0.7761,
+      "mean_token_accuracy": 0.7701587301587305,
+      "step": 805
+    },
+    {
+      "epoch": 0.2997502081598668,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.7670587595599034e-05,
+      "loss": 0.7763,
+      "mean_token_accuracy": 0.7681318681318681,
+      "step": 810
+    },
+    {
+      "epoch": 0.3016005180867795,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.762897211795607e-05,
+      "loss": 0.8317,
+      "mean_token_accuracy": 0.7544749694749694,
+      "step": 815
+    },
+    {
+      "epoch": 0.3034508280136923,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.758703812203961e-05,
+      "loss": 0.8387,
+      "mean_token_accuracy": 0.7542368742368742,
+      "step": 820
+    },
+    {
+      "epoch": 0.30530113794060504,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.7544787358641735e-05,
+      "loss": 0.881,
+      "mean_token_accuracy": 0.7437851037851038,
+      "step": 825
+    },
+    {
+      "epoch": 0.3071514478675178,
+      "grad_norm": 0.6328125,
+      "learning_rate": 1.7502221591779932e-05,
+      "loss": 0.8617,
+      "mean_token_accuracy": 0.7450671550671552,
+      "step": 830
+    },
+    {
+      "epoch": 0.30900175779443056,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.7459342598623438e-05,
+      "loss": 0.8028,
+      "mean_token_accuracy": 0.7642246642246643,
+      "step": 835
+    },
+    {
+      "epoch": 0.31085206772134333,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.741615216941905e-05,
+      "loss": 0.7574,
+      "mean_token_accuracy": 0.7761538461538462,
+      "step": 840
+    },
+    {
+      "epoch": 0.3127023776482561,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.7372652107416364e-05,
+      "loss": 0.8197,
+      "mean_token_accuracy": 0.7589438339438341,
+      "step": 845
+    },
+    {
+      "epoch": 0.31455268757516885,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.7328844228792513e-05,
+      "loss": 0.7363,
+      "mean_token_accuracy": 0.7819719169719171,
+      "step": 850
+    },
+    {
+      "epoch": 0.3164029975020816,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.7284730362576308e-05,
+      "loss": 0.8083,
+      "mean_token_accuracy": 0.7616605616605618,
+      "step": 855
+    },
+    {
+      "epoch": 0.3182533074289944,
+      "grad_norm": 0.671875,
+      "learning_rate": 1.7240312350571905e-05,
+      "loss": 0.7911,
+      "mean_token_accuracy": 0.7681623931623932,
+      "step": 860
+    },
+    {
+      "epoch": 0.3201036173559071,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.719559204728188e-05,
+      "loss": 0.7567,
+      "mean_token_accuracy": 0.7757326007326008,
+      "step": 865
+    },
+    {
+      "epoch": 0.32195392728281985,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.715057131982983e-05,
+      "loss": 0.8,
+      "mean_token_accuracy": 0.7663308913308913,
+      "step": 870
+    },
+    {
+      "epoch": 0.3238042372097326,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.710525204788239e-05,
+      "loss": 0.764,
+      "mean_token_accuracy": 0.772820512820513,
+      "step": 875
+    },
+    {
+      "epoch": 0.3256545471366454,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.7059636123570767e-05,
+      "loss": 0.8007,
+      "mean_token_accuracy": 0.7663186813186816,
+      "step": 880
+    },
+    {
+      "epoch": 0.32750485706355814,
+      "grad_norm": 0.5,
+      "learning_rate": 1.7013725451411757e-05,
+      "loss": 0.7417,
+      "mean_token_accuracy": 0.7798351648351649,
+      "step": 885
+    },
+    {
+      "epoch": 0.3293551669904709,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.696752194822819e-05,
+      "loss": 0.7639,
+      "mean_token_accuracy": 0.777094017094017,
+      "step": 890
+    },
+    {
+      "epoch": 0.33120547691738367,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.692102754306895e-05,
+      "loss": 0.7544,
+      "mean_token_accuracy": 0.776776556776557,
+      "step": 895
+    },
+    {
+      "epoch": 0.33305578684429643,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.6874244177128395e-05,
+      "loss": 0.768,
+      "mean_token_accuracy": 0.7750244200244201,
+      "step": 900
+    },
+    {
+      "epoch": 0.33305578684429643,
+      "eval_loss": 0.8318906426429749,
+      "eval_mean_token_accuracy": 0.7520444696915287,
+      "eval_runtime": 26.064,
+      "eval_samples_per_second": 4.949,
+      "eval_steps_per_second": 1.266,
+      "step": 900
+    },
+    {
+      "epoch": 0.3349060967712092,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.6827173803665328e-05,
+      "loss": 0.8461,
+      "mean_token_accuracy": 0.7512515262515264,
+      "step": 905
+    },
+    {
+      "epoch": 0.33675640669812196,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.677981838792144e-05,
+      "loss": 0.8129,
+      "mean_token_accuracy": 0.7606288156288159,
+      "step": 910
+    },
+    {
+      "epoch": 0.3386067166250347,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.6732179907039266e-05,
+      "loss": 0.7857,
+      "mean_token_accuracy": 0.7670573870573871,
+      "step": 915
+    },
+    {
+      "epoch": 0.34045702655194743,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.6684260349979637e-05,
+      "loss": 0.8006,
+      "mean_token_accuracy": 0.7629975579975581,
+      "step": 920
+    },
+    {
+      "epoch": 0.3423073364788602,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.6636061717438626e-05,
+      "loss": 0.7465,
+      "mean_token_accuracy": 0.7773809523809525,
+      "step": 925
+    },
+    {
+      "epoch": 0.34415764640577295,
+      "grad_norm": 0.5,
+      "learning_rate": 1.6587586021764022e-05,
+      "loss": 0.7637,
+      "mean_token_accuracy": 0.7725213675213676,
+      "step": 930
+    },
+    {
+      "epoch": 0.3460079563326857,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.653883528687133e-05,
+      "loss": 0.7932,
+      "mean_token_accuracy": 0.7646034264796542,
+      "step": 935
+    },
+    {
+      "epoch": 0.3478582662595985,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.6489811548159245e-05,
+      "loss": 0.8083,
+      "mean_token_accuracy": 0.7603724053724055,
+      "step": 940
+    },
+    {
+      "epoch": 0.34970857618651124,
+      "grad_norm": 0.51953125,
+      "learning_rate": 1.6440516852424678e-05,
+      "loss": 0.8193,
+      "mean_token_accuracy": 0.7595970695970699,
+      "step": 945
+    },
+    {
+      "epoch": 0.351558886113424,
+      "grad_norm": 0.50390625,
+      "learning_rate": 1.6390953257777324e-05,
+      "loss": 0.767,
+      "mean_token_accuracy": 0.7724481074481077,
+      "step": 950
+    },
+    {
+      "epoch": 0.35340919604033677,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.634112283355369e-05,
+      "loss": 0.8085,
+      "mean_token_accuracy": 0.7595360195360195,
+      "step": 955
+    },
+    {
+      "epoch": 0.35525950596724953,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.6291027660230735e-05,
+      "loss": 0.7847,
+      "mean_token_accuracy": 0.7684065934065933,
+      "step": 960
+    },
+    {
+      "epoch": 0.3571098158941623,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.6240669829338992e-05,
+      "loss": 0.8359,
+      "mean_token_accuracy": 0.7520329670329668,
+      "step": 965
+    },
+    {
+      "epoch": 0.358960125821075,
+      "grad_norm": 0.51953125,
+      "learning_rate": 1.6190051443375248e-05,
+      "loss": 0.7892,
+      "mean_token_accuracy": 0.767936507936508,
+      "step": 970
+    },
+    {
+      "epoch": 0.36081043574798777,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.6139174615714753e-05,
+      "loss": 0.8017,
+      "mean_token_accuracy": 0.7666666666666666,
+      "step": 975
+    },
+    {
+      "epoch": 0.36266074567490053,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.6088041470523005e-05,
+      "loss": 0.8326,
+      "mean_token_accuracy": 0.7542124542124545,
+      "step": 980
+    },
+    {
+      "epoch": 0.3645110556018133,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.6036654142667043e-05,
+      "loss": 0.8207,
+      "mean_token_accuracy": 0.7569352869352872,
+      "step": 985
+    },
+    {
+      "epoch": 0.36636136552872606,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.598501477762632e-05,
+      "loss": 0.8088,
+      "mean_token_accuracy": 0.7615811965811966,
+      "step": 990
+    },
+    {
+      "epoch": 0.3682116754556388,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.5933125531403135e-05,
+      "loss": 0.8056,
+      "mean_token_accuracy": 0.7620688113148681,
+      "step": 995
+    },
+    {
+      "epoch": 0.3700619853825516,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.5880988570432603e-05,
+      "loss": 0.7489,
+      "mean_token_accuracy": 0.7761599511599513,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3700619853825516,
+      "eval_loss": 0.828413724899292,
+      "eval_mean_token_accuracy": 0.7527104703575295,
+      "eval_runtime": 26.0583,
+      "eval_samples_per_second": 4.95,
+      "eval_steps_per_second": 1.266,
+      "step": 1000
+    },
+    {
+      "epoch": 0.37191229530946435,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.582860607149222e-05,
+      "loss": 0.7882,
+      "mean_token_accuracy": 0.7645482295482294,
+      "step": 1005
+    },
+    {
+      "epoch": 0.3737626052363771,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.5775980221610966e-05,
+      "loss": 0.8099,
+      "mean_token_accuracy": 0.7611477411477412,
+      "step": 1010
+    },
+    {
+      "epoch": 0.3756129151632899,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.5723113217978e-05,
+      "loss": 0.7989,
+      "mean_token_accuracy": 0.7649633699633701,
+      "step": 1015
+    },
+    {
+      "epoch": 0.37746322509020264,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.567000726785093e-05,
+      "loss": 0.7529,
+      "mean_token_accuracy": 0.7761538461538462,
+      "step": 1020
+    },
+    {
+      "epoch": 0.37931353501711534,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.561666458846365e-05,
+      "loss": 0.845,
+      "mean_token_accuracy": 0.7515079365079365,
+      "step": 1025
+    },
+    {
+      "epoch": 0.3811638449440281,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.5563087406933762e-05,
+      "loss": 0.8284,
+      "mean_token_accuracy": 0.7579242979242979,
+      "step": 1030
+    },
+    {
+      "epoch": 0.38301415487094087,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.550927796016961e-05,
+      "loss": 0.7804,
+      "mean_token_accuracy": 0.7690903540903541,
+      "step": 1035
+    },
+    {
+      "epoch": 0.38486446479785363,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.5455238494776876e-05,
+      "loss": 0.8273,
+      "mean_token_accuracy": 0.756892551892552,
+      "step": 1040
+    },
+    {
+      "epoch": 0.3867147747247664,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.5400971266964772e-05,
+      "loss": 0.7594,
+      "mean_token_accuracy": 0.7756166056166058,
+      "step": 1045
+    },
+    {
+      "epoch": 0.38856508465167916,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.5346478542451862e-05,
+      "loss": 0.8078,
+      "mean_token_accuracy": 0.7607142857142858,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3904153945785919,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.529176259637145e-05,
+      "loss": 0.806,
+      "mean_token_accuracy": 0.7628693528693529,
+      "step": 1055
+    },
+    {
+      "epoch": 0.3922657045055047,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.5236825713176584e-05,
+      "loss": 0.8176,
+      "mean_token_accuracy": 0.757783882783883,
+      "step": 1060
+    },
+    {
+      "epoch": 0.39411601443241745,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.5181670186544706e-05,
+      "loss": 0.8239,
+      "mean_token_accuracy": 0.7577289377289377,
+      "step": 1065
+    },
+    {
+      "epoch": 0.3959663243593302,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.5126298319281859e-05,
+      "loss": 0.7494,
+      "mean_token_accuracy": 0.7797496947496951,
+      "step": 1070
+    },
+    {
+      "epoch": 0.3978166342862429,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.5070712423226552e-05,
+      "loss": 0.7778,
+      "mean_token_accuracy": 0.7717277167277167,
+      "step": 1075
+    },
+    {
+      "epoch": 0.3996669442131557,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.5014914819153252e-05,
+      "loss": 0.8276,
+      "mean_token_accuracy": 0.7555494505494504,
+      "step": 1080
+    },
+    {
+      "epoch": 0.40151725414006845,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.4958907836675467e-05,
+      "loss": 0.8009,
+      "mean_token_accuracy": 0.7621550671550672,
+      "step": 1085
+    },
+    {
+      "epoch": 0.4033675640669812,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.490269381414849e-05,
+      "loss": 0.7902,
+      "mean_token_accuracy": 0.7648168498168497,
+      "step": 1090
+    },
+    {
+      "epoch": 0.405217873993894,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.484627509857178e-05,
+      "loss": 0.7866,
+      "mean_token_accuracy": 0.766733821733822,
+      "step": 1095
+    },
+    {
+      "epoch": 0.40706818392080674,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.4789654045490957e-05,
+      "loss": 0.7597,
+      "mean_token_accuracy": 0.7731868131868133,
+      "step": 1100
+    },
+    {
+      "epoch": 0.40706818392080674,
+      "eval_loss": 0.8258881568908691,
+      "eval_mean_token_accuracy": 0.7535510241392596,
+      "eval_runtime": 26.0571,
+      "eval_samples_per_second": 4.951,
+      "eval_steps_per_second": 1.266,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4089184938477195,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.4732833018899468e-05,
+      "loss": 0.8224,
+      "mean_token_accuracy": 0.7595115995115995,
+      "step": 1105
+    },
+    {
+      "epoch": 0.41076880377463226,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.4675814391139875e-05,
+      "loss": 0.804,
+      "mean_token_accuracy": 0.761831501831502,
+      "step": 1110
+    },
+    {
+      "epoch": 0.412619113701545,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.4618600542804819e-05,
+      "loss": 0.8687,
+      "mean_token_accuracy": 0.7440781440781439,
+      "step": 1115
+    },
+    {
+      "epoch": 0.4144694236284578,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.4561193862637621e-05,
+      "loss": 0.7881,
+      "mean_token_accuracy": 0.7670146520146519,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4163197335553705,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.4503596747432554e-05,
+      "loss": 0.7877,
+      "mean_token_accuracy": 0.7656837606837607,
+      "step": 1125
+    },
+    {
+      "epoch": 0.41817004348228326,
+      "grad_norm": 0.51953125,
+      "learning_rate": 1.4445811601934763e-05,
+      "loss": 0.776,
+      "mean_token_accuracy": 0.7679548229548232,
+      "step": 1130
+    },
+    {
+      "epoch": 0.420020353409196,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.4387840838739875e-05,
+      "loss": 0.8228,
+      "mean_token_accuracy": 0.7587057387057388,
+      "step": 1135
+    },
+    {
+      "epoch": 0.4218706633361088,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.4329686878193271e-05,
+      "loss": 0.8089,
+      "mean_token_accuracy": 0.758992673992674,
+      "step": 1140
+    },
+    {
+      "epoch": 0.42372097326302155,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.4271352148289025e-05,
+      "loss": 0.7588,
+      "mean_token_accuracy": 0.7771428571428574,
+      "step": 1145
+    },
+    {
+      "epoch": 0.4255712831899343,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.421283908456854e-05,
+      "loss": 0.7389,
+      "mean_token_accuracy": 0.7806227106227108,
+      "step": 1150
+    },
+    {
+      "epoch": 0.4274215931168471,
+      "grad_norm": 0.50390625,
+      "learning_rate": 1.4154150130018867e-05,
+      "loss": 0.7544,
+      "mean_token_accuracy": 0.7760378510378512,
+      "step": 1155
+    },
+    {
+      "epoch": 0.42927190304375984,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.4095287734970678e-05,
+      "loss": 0.78,
+      "mean_token_accuracy": 0.7688949938949939,
+      "step": 1160
+    },
+    {
+      "epoch": 0.4311222129706726,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.4036254356996004e-05,
+      "loss": 0.8118,
+      "mean_token_accuracy": 0.7598107448107447,
+      "step": 1165
+    },
+    {
+      "epoch": 0.43297252289758537,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.3977052460805597e-05,
+      "loss": 0.7781,
+      "mean_token_accuracy": 0.7703785103785107,
+      "step": 1170
+    },
+    {
+      "epoch": 0.43482283282449813,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.3917684518146044e-05,
+      "loss": 0.78,
+      "mean_token_accuracy": 0.7676862026862029,
+      "step": 1175
+    },
+    {
+      "epoch": 0.43667314275141084,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.3858153007696552e-05,
+      "loss": 0.788,
+      "mean_token_accuracy": 0.7661782661782661,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4385234526783236,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.3798460414965475e-05,
+      "loss": 0.7445,
+      "mean_token_accuracy": 0.7786141636141636,
+      "step": 1185
+    },
+    {
+      "epoch": 0.44037376260523636,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.3738609232186537e-05,
+      "loss": 0.8349,
+      "mean_token_accuracy": 0.7525091575091576,
+      "step": 1190
+    },
+    {
+      "epoch": 0.4422240725321491,
+      "grad_norm": 0.50390625,
+      "learning_rate": 1.3678601958214779e-05,
+      "loss": 0.7486,
+      "mean_token_accuracy": 0.7754212454212456,
+      "step": 1195
+    },
+    {
+      "epoch": 0.4440743824590619,
+      "grad_norm": 0.5234375,
+      "learning_rate": 1.3618441098422215e-05,
+      "loss": 0.7466,
+      "mean_token_accuracy": 0.7784615384615386,
+      "step": 1200
+    },
+    {
+      "epoch": 0.4440743824590619,
+      "eval_loss": 0.8240311145782471,
+      "eval_mean_token_accuracy": 0.7536634389575568,
+      "eval_runtime": 26.0628,
+      "eval_samples_per_second": 4.95,
+      "eval_steps_per_second": 1.266,
+      "step": 1200
+    },
+    {
+      "epoch": 0.44592469238597465,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.3558129164593256e-05,
+      "loss": 0.8022,
+      "mean_token_accuracy": 0.7632234432234433,
+      "step": 1205
+    },
+    {
+      "epoch": 0.4477750023128874,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.349766867481982e-05,
+      "loss": 0.8113,
+      "mean_token_accuracy": 0.7599938949938951,
+      "step": 1210
+    },
+    {
+      "epoch": 0.4496253122398002,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.3437062153396201e-05,
+      "loss": 0.8284,
+      "mean_token_accuracy": 0.7551892551892552,
+      "step": 1215
+    },
+    {
+      "epoch": 0.45147562216671294,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.337631213071369e-05,
+      "loss": 0.8252,
+      "mean_token_accuracy": 0.7550122100122101,
+      "step": 1220
+    },
+    {
+      "epoch": 0.4533259320936257,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.331542114315491e-05,
+      "loss": 0.7816,
+      "mean_token_accuracy": 0.7709096459096458,
+      "step": 1225
+    },
+    {
+      "epoch": 0.4551762420205384,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.325439173298793e-05,
+      "loss": 0.7404,
+      "mean_token_accuracy": 0.7791269841269844,
+      "step": 1230
+    },
+    {
+      "epoch": 0.4570265519474512,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.3193226448260128e-05,
+      "loss": 0.8103,
+      "mean_token_accuracy": 0.7608852258852259,
+      "step": 1235
+    },
+    {
+      "epoch": 0.45887686187436394,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.3131927842691793e-05,
+      "loss": 0.8181,
+      "mean_token_accuracy": 0.7588034188034188,
+      "step": 1240
+    },
+    {
+      "epoch": 0.4607271718012767,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.3070498475569507e-05,
+      "loss": 0.8619,
+      "mean_token_accuracy": 0.7451221001221,
+      "step": 1245
+    },
+    {
+      "epoch": 0.46257748172818947,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.3008940911639302e-05,
+      "loss": 0.82,
+      "mean_token_accuracy": 0.7582295482295478,
+      "step": 1250
+    },
+    {
+      "epoch": 0.46442779165510223,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.2947257720999577e-05,
+      "loss": 0.7679,
+      "mean_token_accuracy": 0.769072039072039,
+      "step": 1255
+    },
+    {
+      "epoch": 0.466278101582015,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.2885451478993777e-05,
+      "loss": 0.8327,
+      "mean_token_accuracy": 0.7543284493284494,
+      "step": 1260
+    },
+    {
+      "epoch": 0.46812841150892776,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.282352476610289e-05,
+      "loss": 0.8023,
+      "mean_token_accuracy": 0.7643589743589744,
+      "step": 1265
+    },
+    {
+      "epoch": 0.4699787214358405,
+      "grad_norm": 0.515625,
+      "learning_rate": 1.2761480167837705e-05,
+      "loss": 0.7547,
+      "mean_token_accuracy": 0.778888888888889,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4718290313627533,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.2699320274630847e-05,
+      "loss": 0.7727,
+      "mean_token_accuracy": 0.7712576312576314,
+      "step": 1275
+    },
+    {
+      "epoch": 0.47367934128966604,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.263704768172864e-05,
+      "loss": 0.8439,
+      "mean_token_accuracy": 0.7506593406593408,
+      "step": 1280
+    },
+    {
+      "epoch": 0.47552965121657875,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.257466498908276e-05,
+      "loss": 0.7718,
+      "mean_token_accuracy": 0.7702136752136753,
+      "step": 1285
+    },
+    {
+      "epoch": 0.4773799611434915,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.2512174801241657e-05,
+      "loss": 0.7446,
+      "mean_token_accuracy": 0.7791147741147741,
+      "step": 1290
+    },
+    {
+      "epoch": 0.4792302710704043,
+      "grad_norm": 0.51171875,
+      "learning_rate": 1.2449579727241834e-05,
+      "loss": 0.7829,
+      "mean_token_accuracy": 0.7663675213675214,
+      "step": 1295
+    },
+    {
+      "epoch": 0.48108058099731704,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.2386882380498918e-05,
+      "loss": 0.7929,
+      "mean_token_accuracy": 0.7648290598290599,
+      "step": 1300
+    },
+    {
+      "epoch": 0.48108058099731704,
+      "eval_loss": 0.822527289390564,
+      "eval_mean_token_accuracy": 0.7542538071949837,
+      "eval_runtime": 26.0765,
+      "eval_samples_per_second": 4.947,
+      "eval_steps_per_second": 1.266,
+      "step": 1300
+    },
+    {
+      "epoch": 0.4829308909242298,
+      "grad_norm": 0.51171875,
+      "learning_rate": 1.2324085378698529e-05,
+      "loss": 0.7185,
+      "mean_token_accuracy": 0.7862759462759464,
+      "step": 1305
+    },
+    {
+      "epoch": 0.48478120085114257,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.2261191343687e-05,
+      "loss": 0.7916,
+      "mean_token_accuracy": 0.7664346764346766,
+      "step": 1310
+    },
+    {
+      "epoch": 0.48663151077805533,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.219820290136192e-05,
+      "loss": 0.7397,
+      "mean_token_accuracy": 0.7816849816849819,
+      "step": 1315
+    },
+    {
+      "epoch": 0.4884818207049681,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.2135122681562481e-05,
+      "loss": 0.8033,
+      "mean_token_accuracy": 0.7621489621489622,
+      "step": 1320
+    },
+    {
+      "epoch": 0.49033213063188086,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.2071953317959692e-05,
+      "loss": 0.7802,
+      "mean_token_accuracy": 0.7701564101474896,
+      "step": 1325
+    },
+    {
+      "epoch": 0.4921824405587936,
+      "grad_norm": 0.515625,
+      "learning_rate": 1.2008697447946421e-05,
+      "loss": 0.783,
+      "mean_token_accuracy": 0.767771672771673,
+      "step": 1330
+    },
+    {
+      "epoch": 0.49403275048570633,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.1945357712527273e-05,
+      "loss": 0.7854,
+      "mean_token_accuracy": 0.7664041514041514,
+      "step": 1335
+    },
+    {
+      "epoch": 0.4958830604126191,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.1881936756208329e-05,
+      "loss": 0.7871,
+      "mean_token_accuracy": 0.7693589743589744,
+      "step": 1340
+    },
+    {
+      "epoch": 0.49773337033953186,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.1818437226886738e-05,
+      "loss": 0.7727,
+      "mean_token_accuracy": 0.7710500610500611,
+      "step": 1345
+    },
+    {
+      "epoch": 0.4995836802664446,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.1754861775740163e-05,
+      "loss": 0.7824,
+      "mean_token_accuracy": 0.7686996336996337,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5014339901933574,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.1691213057116082e-05,
+      "loss": 0.7581,
+      "mean_token_accuracy": 0.774822954822955,
+      "step": 1355
+    },
+    {
+      "epoch": 0.5032843001202701,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.1627493728420978e-05,
+      "loss": 0.7875,
+      "mean_token_accuracy": 0.768925518925519,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5051346100471829,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.1563706450009391e-05,
+      "loss": 0.7946,
+      "mean_token_accuracy": 0.766837606837607,
+      "step": 1365
+    },
+    {
+      "epoch": 0.5069849199740957,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.1499853885072827e-05,
+      "loss": 0.7333,
+      "mean_token_accuracy": 0.782954822954823,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5088352299010084,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.1435938699528586e-05,
+      "loss": 0.85,
+      "mean_token_accuracy": 0.7502014652014654,
+      "step": 1375
+    },
+    {
+      "epoch": 0.5106855398279212,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.137196356190845e-05,
+      "loss": 0.8028,
+      "mean_token_accuracy": 0.7618437118437119,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5125358497548339,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.1307931143247268e-05,
+      "loss": 0.7895,
+      "mean_token_accuracy": 0.765152625152625,
+      "step": 1385
+    },
+    {
+      "epoch": 0.5143861596817467,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.1243844116971433e-05,
+      "loss": 0.784,
+      "mean_token_accuracy": 0.7689743589743591,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5162364696086594,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.1179705158787276e-05,
+      "loss": 0.8092,
+      "mean_token_accuracy": 0.759847374847375,
+      "step": 1395
+    },
+    {
+      "epoch": 0.5180867795355723,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.1115516946569333e-05,
+      "loss": 0.7743,
+      "mean_token_accuracy": 0.7708180708180707,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5180867795355723,
+      "eval_loss": 0.8213639259338379,
+      "eval_mean_token_accuracy": 0.754418457359634,
+      "eval_runtime": 26.074,
+      "eval_samples_per_second": 4.947,
+      "eval_steps_per_second": 1.266,
+      "step": 1400
+    },
+    {
+      "epoch": 0.519937089462485,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.105128216024857e-05,
+      "loss": 0.7609,
+      "mean_token_accuracy": 0.7747985347985349,
+      "step": 1405
+    },
+    {
+      "epoch": 0.5217873993893978,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.0987003481700456e-05,
+      "loss": 0.7655,
+      "mean_token_accuracy": 0.7745360195360197,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5236377093163105,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.092268359463302e-05,
+      "loss": 0.7447,
+      "mean_token_accuracy": 0.7784698510154332,
+      "step": 1415
+    },
+    {
+      "epoch": 0.5254880192432232,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.0858325184474796e-05,
+      "loss": 0.7774,
+      "mean_token_accuracy": 0.7705372405372406,
+      "step": 1420
+    },
+    {
+      "epoch": 0.527338329170136,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.0793930938262689e-05,
+      "loss": 0.8122,
+      "mean_token_accuracy": 0.7594444444444445,
+      "step": 1425
+    },
+    {
+      "epoch": 0.5291886390970487,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.0729503544529814e-05,
+      "loss": 0.773,
+      "mean_token_accuracy": 0.7697496947496948,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5310389490239615,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.0665045693193226e-05,
+      "loss": 0.7922,
+      "mean_token_accuracy": 0.7678449328449327,
+      "step": 1435
+    },
+    {
+      "epoch": 0.5328892589508742,
+      "grad_norm": 0.47265625,
+      "learning_rate": 1.0600560075441617e-05,
+      "loss": 0.7439,
+      "mean_token_accuracy": 0.7811233211233213,
+      "step": 1440
+    },
+    {
+      "epoch": 0.5347395688777871,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.0536049383622966e-05,
+      "loss": 0.8202,
+      "mean_token_accuracy": 0.7574175824175824,
+      "step": 1445
+    },
+    {
+      "epoch": 0.5365898788046998,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.047151631113212e-05,
+      "loss": 0.762,
+      "mean_token_accuracy": 0.7733272283272286,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5384401887316126,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.0406963552298332e-05,
+      "loss": 0.7313,
+      "mean_token_accuracy": 0.781843711843712,
+      "step": 1455
+    },
+    {
+      "epoch": 0.5402904986585253,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.034239380227281e-05,
+      "loss": 0.7511,
+      "mean_token_accuracy": 0.778717948717949,
+      "step": 1460
+    },
+    {
+      "epoch": 0.542140808585438,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.0277809756916134e-05,
+      "loss": 0.8114,
+      "mean_token_accuracy": 0.7592735042735042,
+      "step": 1465
+    },
+    {
+      "epoch": 0.5439911185123508,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.0213214112685747e-05,
+      "loss": 0.7681,
+      "mean_token_accuracy": 0.7718192918192921,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5458414284392635,
+      "grad_norm": 0.51171875,
+      "learning_rate": 1.0148609566523358e-05,
+      "loss": 0.8011,
+      "mean_token_accuracy": 0.7625518925518927,
+      "step": 1475
+    },
+    {
+      "epoch": 0.5476917383661764,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.0083998815742335e-05,
+      "loss": 0.8135,
+      "mean_token_accuracy": 0.7618498168498167,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5495420482930891,
+      "grad_norm": 0.5546875,
+      "learning_rate": 1.0019384557915099e-05,
+      "loss": 0.78,
+      "mean_token_accuracy": 0.7669841269841271,
+      "step": 1485
+    },
+    {
+      "epoch": 0.5513923582200019,
+      "grad_norm": 0.52734375,
+      "learning_rate": 9.9547694907605e-06,
+      "loss": 0.7785,
+      "mean_token_accuracy": 0.7677655677655679,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5532426681469146,
+      "grad_norm": 0.55078125,
+      "learning_rate": 9.890156312031165e-06,
+      "loss": 0.7657,
+      "mean_token_accuracy": 0.7713492063492066,
+      "step": 1495
+    },
+    {
+      "epoch": 0.5550929780738274,
+      "grad_norm": 0.546875,
+      "learning_rate": 9.825547719400889e-06,
+      "loss": 0.7675,
+      "mean_token_accuracy": 0.7740781440781443,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5550929780738274,
+      "eval_loss": 0.8205817341804504,
+      "eval_mean_token_accuracy": 0.7543981073392838,
+      "eval_runtime": 26.0768,
+      "eval_samples_per_second": 4.947,
+      "eval_steps_per_second": 1.265,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5569432880007401,
+      "grad_norm": 0.54296875,
+      "learning_rate": 9.760946410351988e-06,
+      "loss": 0.8231,
+      "mean_token_accuracy": 0.756037851037851,
+      "step": 1505
+    },
+    {
+      "epoch": 0.5587935979276529,
+      "grad_norm": 0.65234375,
+      "learning_rate": 9.696355082062679e-06,
+      "loss": 0.7927,
+      "mean_token_accuracy": 0.7633943833943834,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5606439078545656,
+      "grad_norm": 0.4921875,
+      "learning_rate": 9.631776431294475e-06,
+      "loss": 0.7885,
+      "mean_token_accuracy": 0.767741147741148,
+      "step": 1515
+    },
+    {
+      "epoch": 0.5624942177814783,
+      "grad_norm": 0.51953125,
+      "learning_rate": 9.567213154279582e-06,
+      "loss": 0.7488,
+      "mean_token_accuracy": 0.775897435897436,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5643445277083912,
+      "grad_norm": 0.5546875,
+      "learning_rate": 9.502667946608332e-06,
+      "loss": 0.7858,
+      "mean_token_accuracy": 0.7679120879120881,
+      "step": 1525
+    },
+    {
+      "epoch": 0.5661948376353039,
+      "grad_norm": 0.5859375,
+      "learning_rate": 9.43814350311666e-06,
+      "loss": 0.7891,
+      "mean_token_accuracy": 0.7662637362637363,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5680451475622167,
+      "grad_norm": 0.5625,
+      "learning_rate": 9.37364251777355e-06,
+      "loss": 0.8253,
+      "mean_token_accuracy": 0.7550183150183152,
+      "step": 1535
+    },
+    {
+      "epoch": 0.5698954574891294,
+      "grad_norm": 0.5546875,
+      "learning_rate": 9.309167683568597e-06,
+      "loss": 0.8062,
+      "mean_token_accuracy": 0.7619536019536023,
+      "step": 1540
+    },
+    {
+      "epoch": 0.5717457674160422,
+      "grad_norm": 0.51171875,
+      "learning_rate": 9.244721692399545e-06,
+      "loss": 0.7547,
+      "mean_token_accuracy": 0.776001221001221,
+      "step": 1545
+    },
+    {
+      "epoch": 0.5735960773429549,
+      "grad_norm": 0.54296875,
+      "learning_rate": 9.180307234959918e-06,
+      "loss": 0.7877,
+      "mean_token_accuracy": 0.7657142857142858,
+      "step": 1550
+    },
+    {
+      "epoch": 0.5754463872698677,
+      "grad_norm": 0.546875,
+      "learning_rate": 9.115927000626665e-06,
+      "loss": 0.7956,
+      "mean_token_accuracy": 0.7652930402930405,
+      "step": 1555
+    },
+    {
+      "epoch": 0.5772966971967805,
+      "grad_norm": 0.515625,
+      "learning_rate": 9.051583677347879e-06,
+      "loss": 0.7882,
+      "mean_token_accuracy": 0.7654945054945055,
+      "step": 1560
+    },
+    {
+      "epoch": 0.5791470071236933,
+      "grad_norm": 0.5859375,
+      "learning_rate": 8.987279951530586e-06,
+      "loss": 0.7881,
+      "mean_token_accuracy": 0.7654822954822955,
+      "step": 1565
+    },
+    {
+      "epoch": 0.580997317050606,
+      "grad_norm": 0.53515625,
+      "learning_rate": 8.923018507928564e-06,
+      "loss": 0.8121,
+      "mean_token_accuracy": 0.7594566544566546,
+      "step": 1570
+    },
+    {
+      "epoch": 0.5828476269775187,
+      "grad_norm": 0.5390625,
+      "learning_rate": 8.85880202953026e-06,
+      "loss": 0.7848,
+      "mean_token_accuracy": 0.7694993894993898,
+      "step": 1575
+    },
+    {
+      "epoch": 0.5846979369044315,
+      "grad_norm": 0.56640625,
+      "learning_rate": 8.79463319744677e-06,
+      "loss": 0.8345,
+      "mean_token_accuracy": 0.7526739926739926,
+      "step": 1580
+    },
+    {
+      "epoch": 0.5865482468313442,
+      "grad_norm": 0.5625,
+      "learning_rate": 8.730514690799916e-06,
+      "loss": 0.7608,
+      "mean_token_accuracy": 0.770854700854701,
+      "step": 1585
+    },
+    {
+      "epoch": 0.588398556758257,
+      "grad_norm": 0.546875,
+      "learning_rate": 8.666449186610353e-06,
+      "loss": 0.7593,
+      "mean_token_accuracy": 0.7748901098901101,
+      "step": 1590
+    },
+    {
+      "epoch": 0.5902488666851697,
+      "grad_norm": 0.5078125,
+      "learning_rate": 8.60243935968585e-06,
+      "loss": 0.7608,
+      "mean_token_accuracy": 0.7717338217338219,
+      "step": 1595
+    },
+    {
+      "epoch": 0.5920991766120826,
+      "grad_norm": 0.56640625,
+      "learning_rate": 8.538487882509568e-06,
+      "loss": 0.7908,
+      "mean_token_accuracy": 0.7692551892551893,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5920991766120826,
+      "eval_loss": 0.8200157880783081,
+      "eval_mean_token_accuracy": 0.7547331753214106,
+      "eval_runtime": 26.0828,
+      "eval_samples_per_second": 4.946,
+      "eval_steps_per_second": 1.265,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5939494865389953,
+      "grad_norm": 0.5546875,
+      "learning_rate": 8.474597425128501e-06,
+      "loss": 0.7637,
+      "mean_token_accuracy": 0.7732722832722833,
+      "step": 1605
+    },
+    {
+      "epoch": 0.5957997964659081,
+      "grad_norm": 0.59765625,
+      "learning_rate": 8.410770655042003e-06,
+      "loss": 0.7728,
+      "mean_token_accuracy": 0.7693223443223444,
+      "step": 1610
+    },
+    {
+      "epoch": 0.5976501063928208,
+      "grad_norm": 0.52734375,
+      "learning_rate": 8.347010237090408e-06,
+      "loss": 0.8003,
+      "mean_token_accuracy": 0.7633760683760685,
+      "step": 1615
+    },
+    {
+      "epoch": 0.5995004163197336,
+      "grad_norm": 0.578125,
+      "learning_rate": 8.283318833343773e-06,
+      "loss": 0.8274,
+      "mean_token_accuracy": 0.757887667887668,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6013507262466463,
+      "grad_norm": 0.59375,
+      "learning_rate": 8.219699102990735e-06,
+      "loss": 0.8158,
+      "mean_token_accuracy": 0.7618517857655597,
+      "step": 1625
+    },
+    {
+      "epoch": 0.603201036173559,
+      "grad_norm": 0.5703125,
+      "learning_rate": 8.156153702227484e-06,
+      "loss": 0.775,
+      "mean_token_accuracy": 0.7701159951159952,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6050513461004718,
+      "grad_norm": 0.6171875,
+      "learning_rate": 8.092685284146865e-06,
+      "loss": 0.8372,
+      "mean_token_accuracy": 0.7498168498168496,
+      "step": 1635
+    },
+    {
+      "epoch": 0.6069016560273846,
+      "grad_norm": 0.54296875,
+      "learning_rate": 8.029296498627608e-06,
+      "loss": 0.7926,
+      "mean_token_accuracy": 0.7652503052503052,
+      "step": 1640
+    },
+    {
+      "epoch": 0.6087519659542974,
+      "grad_norm": 0.53125,
+      "learning_rate": 7.965989992223693e-06,
+      "loss": 0.7708,
+      "mean_token_accuracy": 0.7716910866910869,
+      "step": 1645
+    },
+    {
+      "epoch": 0.6106022758812101,
+      "grad_norm": 0.59375,
+      "learning_rate": 7.90276840805385e-06,
+      "loss": 0.8103,
+      "mean_token_accuracy": 0.7594139194139196,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6124525858081229,
+      "grad_norm": 0.5,
+      "learning_rate": 7.839634385691214e-06,
+      "loss": 0.7823,
+      "mean_token_accuracy": 0.7687240537240537,
+      "step": 1655
+    },
+    {
+      "epoch": 0.6143028957350356,
+      "grad_norm": 0.546875,
+      "learning_rate": 7.776590561053117e-06,
+      "loss": 0.811,
+      "mean_token_accuracy": 0.761892551892552,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6161532056619484,
+      "grad_norm": 0.53125,
+      "learning_rate": 7.713639566291028e-06,
+      "loss": 0.849,
+      "mean_token_accuracy": 0.7507020757020758,
+      "step": 1665
+    },
+    {
+      "epoch": 0.6180035155888611,
+      "grad_norm": 0.5390625,
+      "learning_rate": 7.650784029680662e-06,
+      "loss": 0.7242,
+      "mean_token_accuracy": 0.7835714285714287,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6198538255157738,
+      "grad_norm": 0.5234375,
+      "learning_rate": 7.58802657551225e-06,
+      "loss": 0.7465,
+      "mean_token_accuracy": 0.777753357753358,
+      "step": 1675
+    },
+    {
+      "epoch": 0.6217041354426867,
+      "grad_norm": 0.5703125,
+      "learning_rate": 7.52536982398097e-06,
+      "loss": 0.8063,
+      "mean_token_accuracy": 0.7636630036630038,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6235544453695994,
+      "grad_norm": 0.609375,
+      "learning_rate": 7.46281639107755e-06,
+      "loss": 0.8697,
+      "mean_token_accuracy": 0.7467540893006921,
+      "step": 1685
+    },
+    {
+      "epoch": 0.6254047552965122,
+      "grad_norm": 0.5234375,
+      "learning_rate": 7.400368888479048e-06,
+      "loss": 0.8231,
+      "mean_token_accuracy": 0.7551282051282052,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6272550652234249,
+      "grad_norm": 0.609375,
+      "learning_rate": 7.3380299234398076e-06,
+      "loss": 0.8251,
+      "mean_token_accuracy": 0.7577045177045176,
+      "step": 1695
+    },
+    {
+      "epoch": 0.6291053751503377,
+      "grad_norm": 0.56640625,
+      "learning_rate": 7.275802098682612e-06,
+      "loss": 0.8423,
+      "mean_token_accuracy": 0.7533882783882784,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6291053751503377,
+      "eval_loss": 0.8196747303009033,
+      "eval_mean_token_accuracy": 0.7543997396938573,
+      "eval_runtime": 26.0751,
+      "eval_samples_per_second": 4.947,
+      "eval_steps_per_second": 1.266,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6309556850772504,
+      "grad_norm": 0.578125,
+      "learning_rate": 7.213688012290004e-06,
+      "loss": 0.8027,
+      "mean_token_accuracy": 0.761959706959707,
+      "step": 1705
+    },
+    {
+      "epoch": 0.6328059950041632,
+      "grad_norm": 0.546875,
+      "learning_rate": 7.151690257595826e-06,
+      "loss": 0.7783,
+      "mean_token_accuracy": 0.7691391941391942,
+      "step": 1710
+    },
+    {
+      "epoch": 0.6346563049310759,
+      "grad_norm": 0.5625,
+      "learning_rate": 7.089811423076936e-06,
+      "loss": 0.7158,
+      "mean_token_accuracy": 0.7856288156288157,
+      "step": 1715
+    },
+    {
+      "epoch": 0.6365066148579888,
+      "grad_norm": 0.53515625,
+      "learning_rate": 7.028054092245134e-06,
+      "loss": 0.8086,
+      "mean_token_accuracy": 0.7597680097680097,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6383569247849015,
+      "grad_norm": 0.58984375,
+      "learning_rate": 6.966420843539321e-06,
+      "loss": 0.8179,
+      "mean_token_accuracy": 0.7561782661782661,
+      "step": 1725
+    },
+    {
+      "epoch": 0.6402072347118142,
+      "grad_norm": 0.498046875,
+      "learning_rate": 6.90491425021781e-06,
+      "loss": 0.8093,
+      "mean_token_accuracy": 0.759884004884005,
+      "step": 1730
+    },
+    {
+      "epoch": 0.642057544638727,
+      "grad_norm": 0.55078125,
+      "learning_rate": 6.843536880250914e-06,
+      "loss": 0.7813,
+      "mean_token_accuracy": 0.7674969474969477,
+      "step": 1735
+    },
+    {
+      "epoch": 0.6439078545656397,
+      "grad_norm": 0.57421875,
+      "learning_rate": 6.7822912962137225e-06,
+      "loss": 0.805,
+      "mean_token_accuracy": 0.7615506715506717,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6457581644925525,
+      "grad_norm": 0.53125,
+      "learning_rate": 6.721180055179113e-06,
+      "loss": 0.8015,
+      "mean_token_accuracy": 0.7642264861374811,
+      "step": 1745
+    },
+    {
+      "epoch": 0.6476084744194652,
+      "grad_norm": 0.609375,
+      "learning_rate": 6.660205708610987e-06,
+      "loss": 0.8335,
+      "mean_token_accuracy": 0.7567155067155065,
+      "step": 1750
+    },
+    {
+      "epoch": 0.649458784346378,
+      "grad_norm": 0.546875,
+      "learning_rate": 6.599370802257755e-06,
+      "loss": 0.788,
+      "mean_token_accuracy": 0.767722832722833,
+      "step": 1755
+    },
+    {
+      "epoch": 0.6513090942732908,
+      "grad_norm": 0.52734375,
+      "learning_rate": 6.5386778760460316e-06,
+      "loss": 0.7622,
+      "mean_token_accuracy": 0.7774420024420026,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6531594042002036,
+      "grad_norm": 0.58203125,
+      "learning_rate": 6.478129463974598e-06,
+      "loss": 0.8318,
+      "mean_token_accuracy": 0.7556715506715507,
+      "step": 1765
+    },
+    {
+      "epoch": 0.6550097141271163,
+      "grad_norm": 0.53125,
+      "learning_rate": 6.417728094008613e-06,
+      "loss": 0.7492,
+      "mean_token_accuracy": 0.7757814407814407,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6568600240540291,
+      "grad_norm": 0.5234375,
+      "learning_rate": 6.357476287974051e-06,
+      "loss": 0.789,
+      "mean_token_accuracy": 0.7653174603174605,
+      "step": 1775
+    },
+    {
+      "epoch": 0.6587103339809418,
+      "grad_norm": 0.52734375,
+      "learning_rate": 6.297376561452428e-06,
+      "loss": 0.77,
+      "mean_token_accuracy": 0.7738278388278391,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6605606439078545,
+      "grad_norm": 0.52734375,
+      "learning_rate": 6.237431423675764e-06,
+      "loss": 0.7891,
+      "mean_token_accuracy": 0.7663614163614165,
+      "step": 1785
+    },
+    {
+      "epoch": 0.6624109538347673,
+      "grad_norm": 0.54296875,
+      "learning_rate": 6.177643377421827e-06,
+      "loss": 0.7823,
+      "mean_token_accuracy": 0.7693284493284495,
+      "step": 1790
+    },
+    {
+      "epoch": 0.66426126376168,
+      "grad_norm": 0.5625,
+      "learning_rate": 6.118014918909633e-06,
+      "loss": 0.8053,
+      "mean_token_accuracy": 0.7628998778998779,
+      "step": 1795
+    },
+    {
+      "epoch": 0.6661115736885929,
+      "grad_norm": 0.55078125,
+      "learning_rate": 6.058548537695225e-06,
+      "loss": 0.7373,
+      "mean_token_accuracy": 0.7803357753357754,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6661115736885929,
+      "eval_loss": 0.8193041086196899,
+      "eval_mean_token_accuracy": 0.754612707553884,
+      "eval_runtime": 26.077,
+      "eval_samples_per_second": 4.947,
+      "eval_steps_per_second": 1.265,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6679618836155056,
+      "grad_norm": 0.546875,
+      "learning_rate": 5.999246716567737e-06,
+      "loss": 0.8142,
+      "mean_token_accuracy": 0.7571855921855921,
+      "step": 1805
+    },
+    {
+      "epoch": 0.6698121935424184,
+      "grad_norm": 0.60546875,
+      "learning_rate": 5.940111931445731e-06,
+      "loss": 0.8563,
+      "mean_token_accuracy": 0.7478327228327226,
+      "step": 1810
+    },
+    {
+      "epoch": 0.6716625034693311,
+      "grad_norm": 0.5,
+      "learning_rate": 5.881146651273825e-06,
+      "loss": 0.7408,
+      "mean_token_accuracy": 0.7795543345543348,
+      "step": 1815
+    },
+    {
+      "epoch": 0.6735128133962439,
+      "grad_norm": 0.5234375,
+      "learning_rate": 5.822353337919616e-06,
+      "loss": 0.7726,
+      "mean_token_accuracy": 0.7696886446886447,
+      "step": 1820
+    },
+    {
+      "epoch": 0.6753631233231566,
+      "grad_norm": 0.51171875,
+      "learning_rate": 5.763734446070892e-06,
+      "loss": 0.7316,
+      "mean_token_accuracy": 0.7817948717948717,
+      "step": 1825
+    },
+    {
+      "epoch": 0.6772134332500694,
+      "grad_norm": 0.515625,
+      "learning_rate": 5.705292423133133e-06,
+      "loss": 0.7376,
+      "mean_token_accuracy": 0.7823137973137975,
+      "step": 1830
+    },
+    {
+      "epoch": 0.6790637431769821,
+      "grad_norm": 0.58203125,
+      "learning_rate": 5.647029709127355e-06,
+      "loss": 0.8121,
+      "mean_token_accuracy": 0.7602686202686204,
+      "step": 1835
+    },
+    {
+      "epoch": 0.6809140531038949,
+      "grad_norm": 0.53125,
+      "learning_rate": 5.5889487365882065e-06,
+      "loss": 0.7721,
+      "mean_token_accuracy": 0.7713736263736264,
+      "step": 1840
+    },
+    {
+      "epoch": 0.6827643630308077,
+      "grad_norm": 0.546875,
+      "learning_rate": 5.531051930462437e-06,
+      "loss": 0.7729,
+      "mean_token_accuracy": 0.7722588522588524,
+      "step": 1845
+    },
+    {
+      "epoch": 0.6846146729577204,
+      "grad_norm": 0.53515625,
+      "learning_rate": 5.4733417080076325e-06,
+      "loss": 0.7598,
+      "mean_token_accuracy": 0.7739255189255192,
+      "step": 1850
+    },
+    {
+      "epoch": 0.6864649828846332,
+      "grad_norm": 0.57421875,
+      "learning_rate": 5.415820478691301e-06,
+      "loss": 0.7635,
+      "mean_token_accuracy": 0.7729120879120881,
+      "step": 1855
+    },
+    {
+      "epoch": 0.6883152928115459,
+      "grad_norm": 0.5546875,
+      "learning_rate": 5.358490644090263e-06,
+      "loss": 0.8096,
+      "mean_token_accuracy": 0.7614774114774117,
+      "step": 1860
+    },
+    {
+      "epoch": 0.6901656027384587,
+      "grad_norm": 0.5859375,
+      "learning_rate": 5.3013545977904005e-06,
+      "loss": 0.8301,
+      "mean_token_accuracy": 0.7540598290598292,
+      "step": 1865
+    },
+    {
+      "epoch": 0.6920159126653714,
+      "grad_norm": 0.5625,
+      "learning_rate": 5.244414725286717e-06,
+      "loss": 0.8301,
+      "mean_token_accuracy": 0.7557692307692309,
+      "step": 1870
+    },
+    {
+      "epoch": 0.6938662225922843,
+      "grad_norm": 0.51171875,
+      "learning_rate": 5.187673403883721e-06,
+      "loss": 0.7507,
+      "mean_token_accuracy": 0.776666666666667,
+      "step": 1875
+    },
+    {
+      "epoch": 0.695716532519197,
+      "grad_norm": 0.51171875,
+      "learning_rate": 5.131133002596199e-06,
+      "loss": 0.7309,
+      "mean_token_accuracy": 0.7812576312576314,
+      "step": 1880
+    },
+    {
+      "epoch": 0.6975668424461097,
+      "grad_norm": 0.5859375,
+      "learning_rate": 5.074795882050293e-06,
+      "loss": 0.8038,
+      "mean_token_accuracy": 0.7628693528693529,
+      "step": 1885
+    },
+    {
+      "epoch": 0.6994171523730225,
+      "grad_norm": 0.55078125,
+      "learning_rate": 5.018664394384942e-06,
+      "loss": 0.7564,
+      "mean_token_accuracy": 0.7764713064713066,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7012674622999352,
+      "grad_norm": 0.515625,
+      "learning_rate": 4.9627408831536705e-06,
+      "loss": 0.772,
+      "mean_token_accuracy": 0.7692429792429796,
+      "step": 1895
+    },
+    {
+      "epoch": 0.703117772226848,
+      "grad_norm": 0.6015625,
+      "learning_rate": 4.907027683226761e-06,
+      "loss": 0.7601,
+      "mean_token_accuracy": 0.7741116335089171,
+      "step": 1900
+    },
+    {
+      "epoch": 0.703117772226848,
+      "eval_loss": 0.8192035555839539,
+      "eval_mean_token_accuracy": 0.7546293575705342,
+      "eval_runtime": 26.0663,
+      "eval_samples_per_second": 4.949,
+      "eval_steps_per_second": 1.266,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7049680821537607,
+      "grad_norm": 0.5234375,
+      "learning_rate": 4.85152712069375e-06,
+      "loss": 0.7848,
+      "mean_token_accuracy": 0.7679609279609283,
+      "step": 1905
+    },
+    {
+      "epoch": 0.7068183920806735,
+      "grad_norm": 0.56640625,
+      "learning_rate": 4.7962415127663265e-06,
+      "loss": 0.7819,
+      "mean_token_accuracy": 0.7709890109890111,
+      "step": 1910
+    },
+    {
+      "epoch": 0.7086687020075862,
+      "grad_norm": 0.66796875,
+      "learning_rate": 4.74117316768158e-06,
+      "loss": 0.7899,
+      "mean_token_accuracy": 0.7628998778998778,
+      "step": 1915
+    },
+    {
+      "epoch": 0.7105190119344991,
+      "grad_norm": 0.55078125,
+      "learning_rate": 4.686324384605629e-06,
+      "loss": 0.8049,
+      "mean_token_accuracy": 0.7623748473748474,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7123693218614118,
+      "grad_norm": 0.5390625,
+      "learning_rate": 4.631697453537623e-06,
+      "loss": 0.7944,
+      "mean_token_accuracy": 0.7652136752136754,
+      "step": 1925
+    },
+    {
+      "epoch": 0.7142196317883246,
+      "grad_norm": 0.55078125,
+      "learning_rate": 4.577294655214144e-06,
+      "loss": 0.8065,
+      "mean_token_accuracy": 0.7621123321123322,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7160699417152373,
+      "grad_norm": 0.55859375,
+      "learning_rate": 4.523118261013969e-06,
+      "loss": 0.8069,
+      "mean_token_accuracy": 0.7615995115995114,
+      "step": 1935
+    },
+    {
+      "epoch": 0.71792025164215,
+      "grad_norm": 0.52734375,
+      "learning_rate": 4.469170532863254e-06,
+      "loss": 0.8386,
+      "mean_token_accuracy": 0.7511721611721612,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7197705615690628,
+      "grad_norm": 0.546875,
+      "learning_rate": 4.415453723141081e-06,
+      "loss": 0.8222,
+      "mean_token_accuracy": 0.7564896214896215,
+      "step": 1945
+    },
+    {
+      "epoch": 0.7216208714959755,
+      "grad_norm": 0.52734375,
+      "learning_rate": 4.361970074585426e-06,
+      "loss": 0.7626,
+      "mean_token_accuracy": 0.7755860805860808,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7234711814228884,
+      "grad_norm": 0.625,
+      "learning_rate": 4.308721820199529e-06,
+      "loss": 0.8138,
+      "mean_token_accuracy": 0.7604151404151402,
+      "step": 1955
+    },
+    {
+      "epoch": 0.7253214913498011,
+      "grad_norm": 0.55859375,
+      "learning_rate": 4.255711183158635e-06,
+      "loss": 0.773,
+      "mean_token_accuracy": 0.7697435897435898,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7271718012767139,
+      "grad_norm": 0.5546875,
+      "learning_rate": 4.2029403767172175e-06,
+      "loss": 0.7818,
+      "mean_token_accuracy": 0.7663492063492064,
+      "step": 1965
+    },
+    {
+      "epoch": 0.7290221112036266,
+      "grad_norm": 0.54296875,
+      "learning_rate": 4.150411604116531e-06,
+      "loss": 0.7982,
+      "mean_token_accuracy": 0.7639804639804639,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7308724211305394,
+      "grad_norm": 0.5234375,
+      "learning_rate": 4.098127058492652e-06,
+      "loss": 0.8013,
+      "mean_token_accuracy": 0.761813186813187,
+      "step": 1975
+    },
+    {
+      "epoch": 0.7327227310574521,
+      "grad_norm": 0.546875,
+      "learning_rate": 4.0460889227849e-06,
+      "loss": 0.8035,
+      "mean_token_accuracy": 0.7631562881562881,
+      "step": 1980
+    },
+    {
+      "epoch": 0.7345730409843649,
+      "grad_norm": 0.52734375,
+      "learning_rate": 3.9942993696447045e-06,
+      "loss": 0.7872,
+      "mean_token_accuracy": 0.7675457875457876,
+      "step": 1985
+    },
+    {
+      "epoch": 0.7364233509112776,
+      "grad_norm": 0.53515625,
+      "learning_rate": 3.942760561344877e-06,
+      "loss": 0.8127,
+      "mean_token_accuracy": 0.7607142857142856,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7382736608381903,
+      "grad_norm": 0.5390625,
+      "learning_rate": 3.891474649689362e-06,
+      "loss": 0.8181,
+      "mean_token_accuracy": 0.7563919413919414,
+      "step": 1995
+    },
+    {
+      "epoch": 0.7401239707651032,
+      "grad_norm": 0.56640625,
+      "learning_rate": 3.840443775923365e-06,
+      "loss": 0.7916,
+      "mean_token_accuracy": 0.7642673992673995,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7401239707651032,
+      "eval_loss": 0.819044828414917,
+      "eval_mean_token_accuracy": 0.7546460075871843,
+      "eval_runtime": 26.0834,
+      "eval_samples_per_second": 4.946,
+      "eval_steps_per_second": 1.265,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7419742806920159,
+      "grad_norm": 0.55078125,
+      "learning_rate": 3.7896700706439826e-06,
+      "loss": 0.7943,
+      "mean_token_accuracy": 0.7654884004884005,
+      "step": 2005
+    },
+    {
+      "epoch": 0.7438245906189287,
+      "grad_norm": 0.53125,
+      "learning_rate": 3.7391556537112282e-06,
+      "loss": 0.7535,
+      "mean_token_accuracy": 0.7750793650793653,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7456749005458414,
+      "grad_norm": 0.54296875,
+      "learning_rate": 3.6889026341595378e-06,
+      "loss": 0.75,
+      "mean_token_accuracy": 0.7773809523809525,
+      "step": 2015
+    },
+    {
+      "epoch": 0.7475252104727542,
+      "grad_norm": 0.54296875,
+      "learning_rate": 3.6389131101096953e-06,
+      "loss": 0.7725,
+      "mean_token_accuracy": 0.7696459096459096,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7493755203996669,
+      "grad_norm": 0.55078125,
+      "learning_rate": 3.5891891686812597e-06,
+      "loss": 0.8132,
+      "mean_token_accuracy": 0.758526040459889,
+      "step": 2025
+    },
+    {
+      "epoch": 0.7512258303265797,
+      "grad_norm": 0.5390625,
+      "learning_rate": 3.5397328859054138e-06,
+      "loss": 0.7967,
+      "mean_token_accuracy": 0.7621855921855921,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7530761402534925,
+      "grad_norm": 0.5625,
+      "learning_rate": 3.490546326638273e-06,
+      "loss": 0.8345,
+      "mean_token_accuracy": 0.7564122681613805,
+      "step": 2035
+    },
+    {
+      "epoch": 0.7549264501804053,
+      "grad_norm": 0.53515625,
+      "learning_rate": 3.441631544474705e-06,
+      "loss": 0.74,
+      "mean_token_accuracy": 0.7810012210012212,
+      "step": 2040
+    },
+    {
+      "epoch": 0.756776760107318,
+      "grad_norm": 0.56640625,
+      "learning_rate": 3.3929905816625653e-06,
+      "loss": 0.7891,
+      "mean_token_accuracy": 0.7667155067155067,
+      "step": 2045
+    },
+    {
+      "epoch": 0.7586270700342307,
+      "grad_norm": 0.6171875,
+      "learning_rate": 3.344625469017445e-06,
+      "loss": 0.8077,
+      "mean_token_accuracy": 0.7620695970695971,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7604773799611435,
+      "grad_norm": 0.5234375,
+      "learning_rate": 3.2965382258378674e-06,
+      "loss": 0.7406,
+      "mean_token_accuracy": 0.7823992673992677,
+      "step": 2055
+    },
+    {
+      "epoch": 0.7623276898880562,
+      "grad_norm": 0.5703125,
+      "learning_rate": 3.248730859821002e-06,
+      "loss": 0.7964,
+      "mean_token_accuracy": 0.7645726495726497,
+      "step": 2060
+    },
+    {
+      "epoch": 0.764177999814969,
+      "grad_norm": 0.6015625,
+      "learning_rate": 3.2012053669788136e-06,
+      "loss": 0.8087,
+      "mean_token_accuracy": 0.7617582417582419,
+      "step": 2065
+    },
+    {
+      "epoch": 0.7660283097418817,
+      "grad_norm": 0.54296875,
+      "learning_rate": 3.1539637315547524e-06,
+      "loss": 0.7816,
+      "mean_token_accuracy": 0.7692124542124544,
+      "step": 2070
+    },
+    {
+      "epoch": 0.7678786196687946,
+      "grad_norm": 0.546875,
+      "learning_rate": 3.1070079259408934e-06,
+      "loss": 0.8164,
+      "mean_token_accuracy": 0.7587851037851037,
+      "step": 2075
+    },
+    {
+      "epoch": 0.7697289295957073,
+      "grad_norm": 0.546875,
+      "learning_rate": 3.0603399105955966e-06,
+      "loss": 0.7959,
+      "mean_token_accuracy": 0.7655555555555555,
+      "step": 2080
+    },
+    {
+      "epoch": 0.7715792395226201,
+      "grad_norm": 0.57421875,
+      "learning_rate": 3.0139616339616394e-06,
+      "loss": 0.7766,
+      "mean_token_accuracy": 0.7690048840048841,
+      "step": 2085
+    },
+    {
+      "epoch": 0.7734295494495328,
+      "grad_norm": 0.5390625,
+      "learning_rate": 2.9678750323848893e-06,
+      "loss": 0.7716,
+      "mean_token_accuracy": 0.7713553113553113,
+      "step": 2090
+    },
+    {
+      "epoch": 0.7752798593764455,
+      "grad_norm": 0.58203125,
+      "learning_rate": 2.922082030033446e-06,
+      "loss": 0.8302,
+      "mean_token_accuracy": 0.7541452991452993,
+      "step": 2095
+    },
+    {
+      "epoch": 0.7771301693033583,
+      "grad_norm": 0.53515625,
+      "learning_rate": 2.8765845388172955e-06,
+      "loss": 0.7605,
+      "mean_token_accuracy": 0.7737423687423688,
+      "step": 2100
+    },
+    {
+      "epoch": 0.7771301693033583,
+      "eval_loss": 0.819015622138977,
+      "eval_mean_token_accuracy": 0.754966275554511,
+      "eval_runtime": 26.0914,
+      "eval_samples_per_second": 4.944,
+      "eval_steps_per_second": 1.265,
+      "step": 2100
+    },
+    {
+      "epoch": 0.778980479230271,
+      "grad_norm": 0.5234375,
+      "learning_rate": 2.831384458308518e-06,
+      "loss": 0.7411,
+      "mean_token_accuracy": 0.7796459096459099,
+      "step": 2105
+    },
+    {
+      "epoch": 0.7808307891571838,
+      "grad_norm": 0.5234375,
+      "learning_rate": 2.7864836756619407e-06,
+      "loss": 0.7773,
+      "mean_token_accuracy": 0.7681929181929184,
+      "step": 2110
+    },
+    {
+      "epoch": 0.7826810990840966,
+      "grad_norm": 0.55859375,
+      "learning_rate": 2.741884065536373e-06,
+      "loss": 0.7309,
+      "mean_token_accuracy": 0.780964590964591,
+      "step": 2115
+    },
+    {
+      "epoch": 0.7845314090110094,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.6975874900163223e-06,
+      "loss": 0.7862,
+      "mean_token_accuracy": 0.7671245421245422,
+      "step": 2120
+    },
+    {
+      "epoch": 0.7863817189379221,
+      "grad_norm": 0.53515625,
+      "learning_rate": 2.6535957985342653e-06,
+      "loss": 0.7644,
+      "mean_token_accuracy": 0.7734859584859586,
+      "step": 2125
+    },
+    {
+      "epoch": 0.7882320288648349,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.6099108277934105e-06,
+      "loss": 0.7323,
+      "mean_token_accuracy": 0.7814896214896216,
+      "step": 2130
+    },
+    {
+      "epoch": 0.7900823387917476,
+      "grad_norm": 0.51953125,
+      "learning_rate": 2.5665344016910367e-06,
+      "loss": 0.7531,
+      "mean_token_accuracy": 0.7742735042735043,
+      "step": 2135
+    },
+    {
+      "epoch": 0.7919326487186604,
+      "grad_norm": 0.546875,
+      "learning_rate": 2.523468331242329e-06,
+      "loss": 0.7908,
+      "mean_token_accuracy": 0.765952380952381,
+      "step": 2140
+    },
+    {
+      "epoch": 0.7937829586455731,
+      "grad_norm": 0.625,
+      "learning_rate": 2.4807144145047734e-06,
+      "loss": 0.834,
+      "mean_token_accuracy": 0.7536874236874237,
+      "step": 2145
+    },
+    {
+      "epoch": 0.7956332685724858,
+      "grad_norm": 0.625,
+      "learning_rate": 2.438274436503074e-06,
+      "loss": 0.7333,
+      "mean_token_accuracy": 0.7803418803418805,
+      "step": 2150
+    },
+    {
+      "epoch": 0.7974835784993987,
+      "grad_norm": 0.5546875,
+      "learning_rate": 2.396150169154644e-06,
+      "loss": 0.8079,
+      "mean_token_accuracy": 0.7606349206349208,
+      "step": 2155
+    },
+    {
+      "epoch": 0.7993338884263114,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.3543433711956197e-06,
+      "loss": 0.7997,
+      "mean_token_accuracy": 0.7624297924297926,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8011841983532242,
+      "grad_norm": 0.5078125,
+      "learning_rate": 2.3128557881074153e-06,
+      "loss": 0.7753,
+      "mean_token_accuracy": 0.772893772893773,
+      "step": 2165
+    },
+    {
+      "epoch": 0.8030345082801369,
+      "grad_norm": 0.55859375,
+      "learning_rate": 2.271689152043873e-06,
+      "loss": 0.7912,
+      "mean_token_accuracy": 0.7645604395604397,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8048848182070497,
+      "grad_norm": 0.59375,
+      "learning_rate": 2.230845181758928e-06,
+      "loss": 0.8491,
+      "mean_token_accuracy": 0.748827838827839,
+      "step": 2175
+    },
+    {
+      "epoch": 0.8067351281339624,
+      "grad_norm": 0.515625,
+      "learning_rate": 2.1903255825348533e-06,
+      "loss": 0.7627,
+      "mean_token_accuracy": 0.7751831501831503,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8085854380608752,
+      "grad_norm": 0.60546875,
+      "learning_rate": 2.150132046111054e-06,
+      "loss": 0.7893,
+      "mean_token_accuracy": 0.7648534798534798,
+      "step": 2185
+    },
+    {
+      "epoch": 0.810435747987788,
+      "grad_norm": 0.59765625,
+      "learning_rate": 2.1102662506134506e-06,
+      "loss": 0.8426,
+      "mean_token_accuracy": 0.7504273504273506,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8122860579147008,
+      "grad_norm": 0.55859375,
+      "learning_rate": 2.0707298604843964e-06,
+      "loss": 0.7826,
+      "mean_token_accuracy": 0.7685714285714287,
+      "step": 2195
+    },
+    {
+      "epoch": 0.8141363678416135,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.03152452641321e-06,
+      "loss": 0.7403,
+      "mean_token_accuracy": 0.778907203907204,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8141363678416135,
+      "eval_loss": 0.8189253807067871,
+      "eval_mean_token_accuracy": 0.7546959576371342,
+      "eval_runtime": 26.0836,
+      "eval_samples_per_second": 4.946,
+      "eval_steps_per_second": 1.265,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8159866777685262,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.9926518852672294e-06,
+      "loss": 0.7908,
+      "mean_token_accuracy": 0.7656471306471306,
+      "step": 2205
+    },
+    {
+      "epoch": 0.817836987695439,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.9541135600234917e-06,
+      "loss": 0.7584,
+      "mean_token_accuracy": 0.775720390720391,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8196872976223517,
+      "grad_norm": 0.51953125,
+      "learning_rate": 1.9159111597009584e-06,
+      "loss": 0.7581,
+      "mean_token_accuracy": 0.7746581196581198,
+      "step": 2215
+    },
+    {
+      "epoch": 0.8215376075492645,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.8780462792933473e-06,
+      "loss": 0.7515,
+      "mean_token_accuracy": 0.7767704517704519,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8233879174761772,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.8405204997025394e-06,
+      "loss": 0.8596,
+      "mean_token_accuracy": 0.7457570207570207,
+      "step": 2225
+    },
+    {
+      "epoch": 0.82523822740309,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.8033353876725578e-06,
+      "loss": 0.8478,
+      "mean_token_accuracy": 0.7503479853479854,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8270885373300028,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.766492495724178e-06,
+      "loss": 0.7654,
+      "mean_token_accuracy": 0.7733028083028085,
+      "step": 2235
+    },
+    {
+      "epoch": 0.8289388472569156,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.7299933620900945e-06,
+      "loss": 0.8072,
+      "mean_token_accuracy": 0.7613797313797315,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8307891571838283,
+      "grad_norm": 0.515625,
+      "learning_rate": 1.6938395106507034e-06,
+      "loss": 0.7069,
+      "mean_token_accuracy": 0.7902625152625156,
+      "step": 2245
+    },
+    {
+      "epoch": 0.832639467110741,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.658032450870467e-06,
+      "loss": 0.7755,
+      "mean_token_accuracy": 0.7686630036630037,
+      "step": 2250
+    },
+    {
+      "epoch": 0.8344897770376538,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.622573677734911e-06,
+      "loss": 0.7448,
+      "mean_token_accuracy": 0.7794871794871796,
+      "step": 2255
+    },
+    {
+      "epoch": 0.8363400869645665,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.587464671688187e-06,
+      "loss": 0.8006,
+      "mean_token_accuracy": 0.7608424908424909,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8381903968914793,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.552706898571288e-06,
+      "loss": 0.8142,
+      "mean_token_accuracy": 0.7587423687423687,
+      "step": 2265
+    },
+    {
+      "epoch": 0.840040706818392,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.5183018095608138e-06,
+      "loss": 0.8084,
+      "mean_token_accuracy": 0.7609201304433996,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8418910167453049,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.4842508411084145e-06,
+      "loss": 0.7561,
+      "mean_token_accuracy": 0.7751037851037853,
+      "step": 2275
+    },
+    {
+      "epoch": 0.8437413266722176,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.4505554148807954e-06,
+      "loss": 0.8169,
+      "mean_token_accuracy": 0.7584371184371185,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8455916365991304,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.4172169377003775e-06,
+      "loss": 0.7917,
+      "mean_token_accuracy": 0.7660012210012211,
+      "step": 2285
+    },
+    {
+      "epoch": 0.8474419465260431,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.3842368014865414e-06,
+      "loss": 0.7726,
+      "mean_token_accuracy": 0.7716788766788768,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8492922564529559,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.3516163831975337e-06,
+      "loss": 0.7783,
+      "mean_token_accuracy": 0.772661782661783,
+      "step": 2295
+    },
+    {
+      "epoch": 0.8511425663798686,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.3193570447729642e-06,
+      "loss": 0.8094,
+      "mean_token_accuracy": 0.7593650793650795,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8511425663798686,
+      "eval_loss": 0.8189005255699158,
+      "eval_mean_token_accuracy": 0.7546053075464842,
+      "eval_runtime": 26.0815,
+      "eval_samples_per_second": 4.946,
+      "eval_steps_per_second": 1.265,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8529928763067813,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.2874601330769488e-06,
+      "loss": 0.8282,
+      "mean_token_accuracy": 0.7572283272283271,
+      "step": 2305
+    },
+    {
+      "epoch": 0.8548431862336942,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.255926979841876e-06,
+      "loss": 0.7546,
+      "mean_token_accuracy": 0.7755677655677656,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8566934961606069,
+      "grad_norm": 0.53515625,
+      "learning_rate": 1.224758901612796e-06,
+      "loss": 0.773,
+      "mean_token_accuracy": 0.7707570207570209,
+      "step": 2315
+    },
+    {
+      "epoch": 0.8585438060875197,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.1939571996924738e-06,
+      "loss": 0.7457,
+      "mean_token_accuracy": 0.7784065934065935,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8603941160144324,
+      "grad_norm": 0.5234375,
+      "learning_rate": 1.1635231600870334e-06,
+      "loss": 0.8219,
+      "mean_token_accuracy": 0.7565201465201465,
+      "step": 2325
+    },
+    {
+      "epoch": 0.8622444259413452,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.1334580534522932e-06,
+      "loss": 0.7932,
+      "mean_token_accuracy": 0.7651282051282052,
+      "step": 2330
+    },
+    {
+      "epoch": 0.8640947358682579,
+      "grad_norm": 0.54296875,
+      "learning_rate": 1.1037631350406874e-06,
+      "loss": 0.6943,
+      "mean_token_accuracy": 0.7926190476190478,
+      "step": 2335
+    },
+    {
+      "epoch": 0.8659450457951707,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.0744396446488781e-06,
+      "loss": 0.7987,
+      "mean_token_accuracy": 0.7629731379731381,
+      "step": 2340
+    },
+    {
+      "epoch": 0.8677953557220834,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.0454888065659775e-06,
+      "loss": 0.84,
+      "mean_token_accuracy": 0.7513553113553114,
+      "step": 2345
+    },
+    {
+      "epoch": 0.8696456656489963,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.0169118295224488e-06,
+      "loss": 0.8309,
+      "mean_token_accuracy": 0.7559645909645909,
+      "step": 2350
+    },
+    {
+      "epoch": 0.871495975575909,
+      "grad_norm": 0.55859375,
+      "learning_rate": 9.887099066396178e-07,
+      "loss": 0.8251,
+      "mean_token_accuracy": 0.7557509157509159,
+      "step": 2355
+    },
+    {
+      "epoch": 0.8733462855028217,
+      "grad_norm": 0.5625,
+      "learning_rate": 9.608842153798903e-07,
+      "loss": 0.8459,
+      "mean_token_accuracy": 0.7503296703296702,
+      "step": 2360
+    },
+    {
+      "epoch": 0.8751965954297345,
+      "grad_norm": 0.56640625,
+      "learning_rate": 9.33435917497556e-07,
+      "loss": 0.7765,
+      "mean_token_accuracy": 0.7696031746031744,
+      "step": 2365
+    },
+    {
+      "epoch": 0.8770469053566472,
+      "grad_norm": 0.55859375,
+      "learning_rate": 9.063661589903116e-07,
+      "loss": 0.7871,
+      "mean_token_accuracy": 0.7675152625152626,
+      "step": 2370
+    },
+    {
+      "epoch": 0.87889721528356,
+      "grad_norm": 0.54296875,
+      "learning_rate": 8.796760700513984e-07,
+      "loss": 0.8339,
+      "mean_token_accuracy": 0.7530097680097679,
+      "step": 2375
+    },
+    {
+      "epoch": 0.8807475252104727,
+      "grad_norm": 0.546875,
+      "learning_rate": 8.533667650224253e-07,
+      "loss": 0.7915,
+      "mean_token_accuracy": 0.7669474969474972,
+      "step": 2380
+    },
+    {
+      "epoch": 0.8825978351373855,
+      "grad_norm": 0.60546875,
+      "learning_rate": 8.274393423468385e-07,
+      "loss": 0.8438,
+      "mean_token_accuracy": 0.7518070818070817,
+      "step": 2385
+    },
+    {
+      "epoch": 0.8844481450642983,
+      "grad_norm": 0.52734375,
+      "learning_rate": 8.018948845240538e-07,
+      "loss": 0.8147,
+      "mean_token_accuracy": 0.7604273504273504,
+      "step": 2390
+    },
+    {
+      "epoch": 0.8862984549912111,
+      "grad_norm": 0.58203125,
+      "learning_rate": 7.767344580642821e-07,
+      "loss": 0.785,
+      "mean_token_accuracy": 0.7690048840048841,
+      "step": 2395
+    },
+    {
+      "epoch": 0.8881487649181238,
+      "grad_norm": 0.546875,
+      "learning_rate": 7.519591134439753e-07,
+      "loss": 0.7843,
+      "mean_token_accuracy": 0.7686507936507938,
+      "step": 2400
+    },
+    {
+      "epoch": 0.8881487649181238,
+      "eval_loss": 0.8189650177955627,
+      "eval_mean_token_accuracy": 0.7548106577518344,
+      "eval_runtime": 26.072,
+      "eval_samples_per_second": 4.948,
+      "eval_steps_per_second": 1.266,
+      "step": 2400
+    },
+    {
+      "epoch": 0.8899990748450366,
+      "grad_norm": 0.546875,
+      "learning_rate": 7.275698850619861e-07,
+      "loss": 0.7779,
+      "mean_token_accuracy": 0.7692857142857144,
+      "step": 2405
+    },
+    {
+      "epoch": 0.8918493847719493,
+      "grad_norm": 0.57421875,
+      "learning_rate": 7.035677911963712e-07,
+      "loss": 0.7506,
+      "mean_token_accuracy": 0.7773748473748475,
+      "step": 2410
+    },
+    {
+      "epoch": 0.893699694698862,
+      "grad_norm": 0.5390625,
+      "learning_rate": 6.799538339618838e-07,
+      "loss": 0.796,
+      "mean_token_accuracy": 0.7610561660561661,
+      "step": 2415
+    },
+    {
+      "epoch": 0.8955500046257748,
+      "grad_norm": 0.49609375,
+      "learning_rate": 6.567289992681258e-07,
+      "loss": 0.7628,
+      "mean_token_accuracy": 0.7717032967032967,
+      "step": 2420
+    },
+    {
+      "epoch": 0.8974003145526875,
+      "grad_norm": 0.58203125,
+      "learning_rate": 6.33894256778399e-07,
+      "loss": 0.8022,
+      "mean_token_accuracy": 0.7613797313797314,
+      "step": 2425
+    },
+    {
+      "epoch": 0.8992506244796004,
+      "grad_norm": 0.6015625,
+      "learning_rate": 6.114505598692011e-07,
+      "loss": 0.7999,
+      "mean_token_accuracy": 0.7630158730158731,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9011009344065131,
+      "grad_norm": 0.57421875,
+      "learning_rate": 5.893988455904387e-07,
+      "loss": 0.743,
+      "mean_token_accuracy": 0.7777045177045179,
+      "step": 2435
+    },
+    {
+      "epoch": 0.9029512443334259,
+      "grad_norm": 0.494140625,
+      "learning_rate": 5.677400346262918e-07,
+      "loss": 0.7622,
+      "mean_token_accuracy": 0.7728266178266182,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9048015542603386,
+      "grad_norm": 0.578125,
+      "learning_rate": 5.464750312567835e-07,
+      "loss": 0.7049,
+      "mean_token_accuracy": 0.7893162393162396,
+      "step": 2445
+    },
+    {
+      "epoch": 0.9066518641872514,
+      "grad_norm": 0.55859375,
+      "learning_rate": 5.256047233200201e-07,
+      "loss": 0.7841,
+      "mean_token_accuracy": 0.7664774114774116,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9085021741141641,
+      "grad_norm": 0.5390625,
+      "learning_rate": 5.051299821751254e-07,
+      "loss": 0.7993,
+      "mean_token_accuracy": 0.7621306471306472,
+      "step": 2455
+    },
+    {
+      "epoch": 0.9103524840410768,
+      "grad_norm": 0.60546875,
+      "learning_rate": 4.850516626658585e-07,
+      "loss": 0.8352,
+      "mean_token_accuracy": 0.7551770451770453,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9122027939679896,
+      "grad_norm": 0.55859375,
+      "learning_rate": 4.653706030849214e-07,
+      "loss": 0.7927,
+      "mean_token_accuracy": 0.7655616605616608,
+      "step": 2465
+    },
+    {
+      "epoch": 0.9140531038949024,
+      "grad_norm": 0.59765625,
+      "learning_rate": 4.4608762513896455e-07,
+      "loss": 0.8552,
+      "mean_token_accuracy": 0.7490293040293041,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9159034138218152,
+      "grad_norm": 0.5390625,
+      "learning_rate": 4.2720353391427547e-07,
+      "loss": 0.7642,
+      "mean_token_accuracy": 0.7714896214896216,
+      "step": 2475
+    },
+    {
+      "epoch": 0.9177537237487279,
+      "grad_norm": 0.5234375,
+      "learning_rate": 4.087191178431682e-07,
+      "loss": 0.8003,
+      "mean_token_accuracy": 0.7616239316239317,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9196040336756407,
+      "grad_norm": 0.53515625,
+      "learning_rate": 3.9063514867105914e-07,
+      "loss": 0.8735,
+      "mean_token_accuracy": 0.7444444444444442,
+      "step": 2485
+    },
+    {
+      "epoch": 0.9214543436025534,
+      "grad_norm": 0.5546875,
+      "learning_rate": 3.729523814242608e-07,
+      "loss": 0.7967,
+      "mean_token_accuracy": 0.7642735042735043,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9233046535294662,
+      "grad_norm": 0.5390625,
+      "learning_rate": 3.5567155437843725e-07,
+      "loss": 0.7621,
+      "mean_token_accuracy": 0.7725457875457875,
+      "step": 2495
+    },
+    {
+      "epoch": 0.9251549634563789,
+      "grad_norm": 0.6640625,
+      "learning_rate": 3.3879338902779945e-07,
+      "loss": 0.7238,
+      "mean_token_accuracy": 0.7829304029304031,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9251549634563789,
+      "eval_loss": 0.8188990950584412,
+      "eval_mean_token_accuracy": 0.7546830076241844,
+      "eval_runtime": 26.0806,
+      "eval_samples_per_second": 4.946,
+      "eval_steps_per_second": 1.265,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9270052733832918,
+      "grad_norm": 0.58203125,
+      "learning_rate": 3.223185900549686e-07,
+      "loss": 0.8143,
+      "mean_token_accuracy": 0.7615140415140416,
+      "step": 2505
+    },
+    {
+      "epoch": 0.9288555833102045,
+      "grad_norm": 0.578125,
+      "learning_rate": 3.0624784530156384e-07,
+      "loss": 0.7935,
+      "mean_token_accuracy": 0.7645115995115995,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9307058932371172,
+      "grad_norm": 0.5234375,
+      "learning_rate": 2.905818257394799e-07,
+      "loss": 0.7468,
+      "mean_token_accuracy": 0.7803663003663006,
+      "step": 2515
+    },
+    {
+      "epoch": 0.93255620316403,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.753211854428728e-07,
+      "loss": 0.7384,
+      "mean_token_accuracy": 0.7796336996336997,
+      "step": 2520
+    },
+    {
+      "epoch": 0.9344065130909427,
+      "grad_norm": 0.51171875,
+      "learning_rate": 2.604665615608526e-07,
+      "loss": 0.8001,
+      "mean_token_accuracy": 0.7647697631640189,
+      "step": 2525
+    },
+    {
+      "epoch": 0.9362568230178555,
+      "grad_norm": 0.58203125,
+      "learning_rate": 2.460185742908816e-07,
+      "loss": 0.839,
+      "mean_token_accuracy": 0.7539621489621489,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9381071329447682,
+      "grad_norm": 0.56640625,
+      "learning_rate": 2.3197782685288385e-07,
+      "loss": 0.8144,
+      "mean_token_accuracy": 0.7606471306471306,
+      "step": 2535
+    },
+    {
+      "epoch": 0.939957442871681,
+      "grad_norm": 0.55078125,
+      "learning_rate": 2.1834490546405186e-07,
+      "loss": 0.7927,
+      "mean_token_accuracy": 0.7649450549450549,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9418077527985937,
+      "grad_norm": 0.55078125,
+      "learning_rate": 2.0512037931437855e-07,
+      "loss": 0.8167,
+      "mean_token_accuracy": 0.7583699633699635,
+      "step": 2545
+    },
+    {
+      "epoch": 0.9436580627255066,
+      "grad_norm": 0.51953125,
+      "learning_rate": 1.9230480054288958e-07,
+      "loss": 0.7274,
+      "mean_token_accuracy": 0.7822710622710624,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9455083726524193,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.7989870421459498e-07,
+      "loss": 0.7599,
+      "mean_token_accuracy": 0.7735409035409037,
+      "step": 2555
+    },
+    {
+      "epoch": 0.9473586825793321,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.6790260829814053e-07,
+      "loss": 0.8146,
+      "mean_token_accuracy": 0.758980463980464,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9492089925062448,
+      "grad_norm": 0.5390625,
+      "learning_rate": 1.5631701364419492e-07,
+      "loss": 0.7054,
+      "mean_token_accuracy": 0.7896398046398048,
+      "step": 2565
+    },
+    {
+      "epoch": 0.9510593024331575,
+      "grad_norm": 0.5078125,
+      "learning_rate": 1.4514240396452438e-07,
+      "loss": 0.7596,
+      "mean_token_accuracy": 0.7729059829059831,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9529096123600703,
+      "grad_norm": 0.51953125,
+      "learning_rate": 1.3437924581181205e-07,
+      "loss": 0.7282,
+      "mean_token_accuracy": 0.7823137973137976,
+      "step": 2575
+    },
+    {
+      "epoch": 0.954759922286983,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.2402798856016474e-07,
+      "loss": 0.8428,
+      "mean_token_accuracy": 0.7522832722832724,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9566102322138959,
+      "grad_norm": 0.55078125,
+      "learning_rate": 1.1408906438636236e-07,
+      "loss": 0.7615,
+      "mean_token_accuracy": 0.7756288156288157,
+      "step": 2585
+    },
+    {
+      "epoch": 0.9584605421408086,
+      "grad_norm": 0.53125,
+      "learning_rate": 1.045628882518046e-07,
+      "loss": 0.7397,
+      "mean_token_accuracy": 0.7781868131868135,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9603108520677214,
+      "grad_norm": 0.59375,
+      "learning_rate": 9.544985788519589e-08,
+      "loss": 0.8289,
+      "mean_token_accuracy": 0.7555555555555555,
+      "step": 2595
+    },
+    {
+      "epoch": 0.9621611619946341,
+      "grad_norm": 0.546875,
+      "learning_rate": 8.675035376593088e-08,
+      "loss": 0.7881,
+      "mean_token_accuracy": 0.766874236874237,
+      "step": 2600
+    },
+    {
+      "epoch": 0.9621611619946341,
+      "eval_loss": 0.8189713954925537,
+      "eval_mean_token_accuracy": 0.7548756254638613,
+      "eval_runtime": 26.0624,
+      "eval_samples_per_second": 4.95,
+      "eval_steps_per_second": 1.266,
+      "step": 2600
+    },
+    {
+      "epoch": 0.9640114719215469,
+      "grad_norm": 0.5546875,
+      "learning_rate": 7.846473910821162e-08,
+      "loss": 0.8085,
+      "mean_token_accuracy": 0.7612148962148964,
+      "step": 2605
+    },
+    {
+      "epoch": 0.9658617818484596,
+      "grad_norm": 0.53125,
+      "learning_rate": 7.059335984588634e-08,
+      "loss": 0.7267,
+      "mean_token_accuracy": 0.7843833943833947,
+      "step": 2610
+    },
+    {
+      "epoch": 0.9677120917753724,
+      "grad_norm": 0.61328125,
+      "learning_rate": 6.313654461800322e-08,
+      "loss": 0.8455,
+      "mean_token_accuracy": 0.7499328449328448,
+      "step": 2615
+    },
+    {
+      "epoch": 0.9695624017022851,
+      "grad_norm": 0.51953125,
+      "learning_rate": 5.609460475509032e-08,
+      "loss": 0.7448,
+      "mean_token_accuracy": 0.7773137973137976,
+      "step": 2620
+    },
+    {
+      "epoch": 0.9714127116291978,
+      "grad_norm": 0.53125,
+      "learning_rate": 4.9467834266154756e-08,
+      "loss": 0.8253,
+      "mean_token_accuracy": 0.7578510378510378,
+      "step": 2625
+    },
+    {
+      "epoch": 0.9732630215561107,
+      "grad_norm": 0.5546875,
+      "learning_rate": 4.325650982641039e-08,
+      "loss": 0.7935,
+      "mean_token_accuracy": 0.7637851037851039,
+      "step": 2630
+    },
+    {
+      "epoch": 0.9751133314830234,
+      "grad_norm": 0.53515625,
+      "learning_rate": 3.746089076572701e-08,
+      "loss": 0.7791,
+      "mean_token_accuracy": 0.7706654456654458,
+      "step": 2635
+    },
+    {
+      "epoch": 0.9769636414099362,
+      "grad_norm": 0.55078125,
+      "learning_rate": 3.208121905779904e-08,
+      "loss": 0.7357,
+      "mean_token_accuracy": 0.7803846153846157,
+      "step": 2640
+    },
+    {
+      "epoch": 0.9788139513368489,
+      "grad_norm": 0.578125,
+      "learning_rate": 2.711771931004692e-08,
+      "loss": 0.7612,
+      "mean_token_accuracy": 0.7714163614163615,
+      "step": 2645
+    },
+    {
+      "epoch": 0.9806642612637617,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.257059875423795e-08,
+      "loss": 0.8455,
+      "mean_token_accuracy": 0.748791208791209,
+      "step": 2650
+    },
+    {
+      "epoch": 0.9825145711906744,
+      "grad_norm": 0.546875,
+      "learning_rate": 1.8440047237832105e-08,
+      "loss": 0.6935,
+      "mean_token_accuracy": 0.7913308913308914,
+      "step": 2655
+    },
+    {
+      "epoch": 0.9843648811175872,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.472623721606059e-08,
+      "loss": 0.7697,
+      "mean_token_accuracy": 0.7722405372405374,
+      "step": 2660
+    },
+    {
+      "epoch": 0.9862151910445,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.1429323744720499e-08,
+      "loss": 0.7916,
+      "mean_token_accuracy": 0.7669413919413919,
+      "step": 2665
+    },
+    {
+      "epoch": 0.9880655009714127,
+      "grad_norm": 0.5546875,
+      "learning_rate": 8.549444473702207e-09,
+      "loss": 0.8311,
+      "mean_token_accuracy": 0.753003663003663,
+      "step": 2670
+    },
+    {
+      "epoch": 0.9899158108983255,
+      "grad_norm": 0.5625,
+      "learning_rate": 6.086719641246186e-09,
+      "loss": 0.7339,
+      "mean_token_accuracy": 0.7823321123321122,
+      "step": 2675
+    },
+    {
+      "epoch": 0.9917661208252382,
+      "grad_norm": 0.51953125,
+      "learning_rate": 4.041252068918145e-09,
+      "loss": 0.7361,
+      "mean_token_accuracy": 0.7801526251526253,
+      "step": 2680
+    },
+    {
+      "epoch": 0.993616430752151,
+      "grad_norm": 0.546875,
+      "learning_rate": 2.4131271573191172e-09,
+      "loss": 0.774,
+      "mean_token_accuracy": 0.7713064713064715,
+      "step": 2685
+    },
+    {
+      "epoch": 0.9954667406790637,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.2024128825172121e-09,
+      "loss": 0.7732,
+      "mean_token_accuracy": 0.7708485958485959,
+      "step": 2690
+    },
+    {
+      "epoch": 0.9973170506059765,
+      "grad_norm": 0.58984375,
+      "learning_rate": 4.0915979321320967e-10,
+      "loss": 0.7891,
+      "mean_token_accuracy": 0.7637240537240537,
+      "step": 2695
+    },
+    {
+      "epoch": 0.9991673605328892,
+      "grad_norm": 0.51953125,
+      "learning_rate": 3.3401008625588706e-11,
+      "loss": 0.7672,
+      "mean_token_accuracy": 0.7712332112332115,
+      "step": 2700
+    },
+    {
+      "epoch": 0.9991673605328892,
+      "eval_loss": 0.8189318776130676,
+      "eval_mean_token_accuracy": 0.7548460254342608,
+      "eval_runtime": 26.1059,
+      "eval_samples_per_second": 4.941,
+      "eval_steps_per_second": 1.264,
+      "step": 2700
+    },
+    {
+      "epoch": 0.9999074845036544,
+      "mean_token_accuracy": 0.7957417582417581,
+      "step": 2702,
+      "total_flos": 6.960985517882081e+17,
+      "train_loss": 0.8181797522337325,
+      "train_runtime": 16039.8645,
+      "train_samples_per_second": 1.348,
+      "train_steps_per_second": 0.168
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 2702,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.960985517882081e+17,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}