diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,25309 @@
+{
+  "best_global_step": 12500,
+  "best_metric": 0.012303678318858147,
+  "best_model_checkpoint": "./sft_model/checkpoint-12500",
+  "epoch": 7.054176072234763,
+  "eval_steps": 500,
+  "global_step": 12500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 0.7264264225959778,
+      "epoch": 0.0028216704288939053,
+      "grad_norm": 1.9348883628845215,
+      "learning_rate": 4.999999882129922e-06,
+      "loss": 0.7593,
+      "mean_token_accuracy": 0.8013904571533204,
+      "num_tokens": 40154.0,
+      "step": 5
+    },
+    {
+      "entropy": 0.8295215606689453,
+      "epoch": 0.0056433408577878106,
+      "grad_norm": 1.6994550228118896,
+      "learning_rate": 4.999999403282752e-06,
+      "loss": 0.6675,
+      "mean_token_accuracy": 0.8162339448928833,
+      "num_tokens": 80736.0,
+      "step": 10
+    },
+    {
+      "entropy": 0.8583701252937317,
+      "epoch": 0.008465011286681716,
+      "grad_norm": 1.4073373079299927,
+      "learning_rate": 4.999998556091706e-06,
+      "loss": 0.7051,
+      "mean_token_accuracy": 0.8052006244659424,
+      "num_tokens": 121191.0,
+      "step": 15
+    },
+    {
+      "entropy": 0.8272644519805908,
+      "epoch": 0.011286681715575621,
+      "grad_norm": 1.322540044784546,
+      "learning_rate": 4.999997340556951e-06,
+      "loss": 0.6492,
+      "mean_token_accuracy": 0.814669954776764,
+      "num_tokens": 161873.0,
+      "step": 20
+    },
+    {
+      "entropy": 0.7710798859596253,
+      "epoch": 0.014108352144469526,
+      "grad_norm": 1.4983023405075073,
+      "learning_rate": 4.999995756678724e-06,
+      "loss": 0.6253,
+      "mean_token_accuracy": 0.8207517623901367,
+      "num_tokens": 202530.0,
+      "step": 25
+    },
+    {
+      "entropy": 0.7607816934585572,
+      "epoch": 0.016930022573363433,
+      "grad_norm": 1.315825343132019,
+      "learning_rate": 4.999993804457336e-06,
+      "loss": 0.6278,
+      "mean_token_accuracy": 0.8195699453353882,
+      "num_tokens": 242963.0,
+      "step": 30
+    },
+    {
+      "entropy": 0.6939435482025147,
+      "epoch": 0.019751693002257337,
+      "grad_norm": 1.3920537233352661,
+      "learning_rate": 4.999991483893173e-06,
+      "loss": 0.5477,
+      "mean_token_accuracy": 0.8409396290779114,
+      "num_tokens": 283687.0,
+      "step": 35
+    },
+    {
+      "entropy": 0.7462215900421143,
+      "epoch": 0.022573363431151242,
+      "grad_norm": 1.3442474603652954,
+      "learning_rate": 4.999988794986688e-06,
+      "loss": 0.6124,
+      "mean_token_accuracy": 0.822938334941864,
+      "num_tokens": 324425.0,
+      "step": 40
+    },
+    {
+      "entropy": 0.7756533026695251,
+      "epoch": 0.025395033860045147,
+      "grad_norm": 1.3415160179138184,
+      "learning_rate": 4.999985737738411e-06,
+      "loss": 0.6479,
+      "mean_token_accuracy": 0.8175343155860901,
+      "num_tokens": 365286.0,
+      "step": 45
+    },
+    {
+      "entropy": 0.7056244015693665,
+      "epoch": 0.028216704288939052,
+      "grad_norm": 1.4092621803283691,
+      "learning_rate": 4.999982312148941e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8333210825920105,
+      "num_tokens": 405893.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.7671143412590027,
+      "epoch": 0.031038374717832957,
+      "grad_norm": 1.3218460083007812,
+      "learning_rate": 4.999978518218954e-06,
+      "loss": 0.6215,
+      "mean_token_accuracy": 0.8200177073478698,
+      "num_tokens": 446486.0,
+      "step": 55
+    },
+    {
+      "entropy": 0.71772620677948,
+      "epoch": 0.033860045146726865,
+      "grad_norm": 1.3232901096343994,
+      "learning_rate": 4.999974355949192e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.825872802734375,
+      "num_tokens": 487151.0,
+      "step": 60
+    },
+    {
+      "entropy": 0.8025589108467102,
+      "epoch": 0.03668171557562077,
+      "grad_norm": 1.3470555543899536,
+      "learning_rate": 4.999969825340475e-06,
+      "loss": 0.6601,
+      "mean_token_accuracy": 0.8110421895980835,
+      "num_tokens": 527748.0,
+      "step": 65
+    },
+    {
+      "entropy": 0.6747750401496887,
+      "epoch": 0.039503386004514675,
+      "grad_norm": 1.3361178636550903,
+      "learning_rate": 4.999964926393691e-06,
+      "loss": 0.5425,
+      "mean_token_accuracy": 0.8391481637954712,
+      "num_tokens": 568462.0,
+      "step": 70
+    },
+    {
+      "entropy": 0.7461091756820679,
+      "epoch": 0.04232505643340858,
+      "grad_norm": 1.4912503957748413,
+      "learning_rate": 4.999959659109804e-06,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.8242739796638489,
+      "num_tokens": 608879.0,
+      "step": 75
+    },
+    {
+      "entropy": 0.7261611700057984,
+      "epoch": 0.045146726862302484,
+      "grad_norm": 1.2540172338485718,
+      "learning_rate": 4.999954023489848e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8286155581474304,
+      "num_tokens": 649552.0,
+      "step": 80
+    },
+    {
+      "entropy": 0.7150743484497071,
+      "epoch": 0.04796839729119639,
+      "grad_norm": 1.4506585597991943,
+      "learning_rate": 4.99994801953493e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8319103837013244,
+      "num_tokens": 690358.0,
+      "step": 85
+    },
+    {
+      "entropy": 0.6064900994300843,
+      "epoch": 0.050790067720090294,
+      "grad_norm": 1.2554188966751099,
+      "learning_rate": 4.999941647246231e-06,
+      "loss": 0.4921,
+      "mean_token_accuracy": 0.851774251461029,
+      "num_tokens": 731078.0,
+      "step": 90
+    },
+    {
+      "entropy": 0.6873465299606323,
+      "epoch": 0.0536117381489842,
+      "grad_norm": 1.2316175699234009,
+      "learning_rate": 4.9999349066250014e-06,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8320832848548889,
+      "num_tokens": 771649.0,
+      "step": 95
+    },
+    {
+      "entropy": 0.7353047490119934,
+      "epoch": 0.056433408577878104,
+      "grad_norm": 2.000030994415283,
+      "learning_rate": 4.9999277976725655e-06,
+      "loss": 0.5994,
+      "mean_token_accuracy": 0.8255870938301086,
+      "num_tokens": 812286.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.7349668025970459,
+      "epoch": 0.05925507900677201,
+      "grad_norm": 1.4496697187423706,
+      "learning_rate": 4.999920320390319e-06,
+      "loss": 0.5832,
+      "mean_token_accuracy": 0.8332556962966919,
+      "num_tokens": 853019.0,
+      "step": 105
+    },
+    {
+      "entropy": 0.7148452520370483,
+      "epoch": 0.062076749435665914,
+      "grad_norm": 1.39479660987854,
+      "learning_rate": 4.999912474779733e-06,
+      "loss": 0.6143,
+      "mean_token_accuracy": 0.8211281180381775,
+      "num_tokens": 893817.0,
+      "step": 110
+    },
+    {
+      "entropy": 0.7357437491416932,
+      "epoch": 0.06489841986455983,
+      "grad_norm": 1.3685375452041626,
+      "learning_rate": 4.999904260842348e-06,
+      "loss": 0.5979,
+      "mean_token_accuracy": 0.8241528987884521,
+      "num_tokens": 934521.0,
+      "step": 115
+    },
+    {
+      "entropy": 0.6611983060836792,
+      "epoch": 0.06772009029345373,
+      "grad_norm": 1.266114354133606,
+      "learning_rate": 4.999895678579776e-06,
+      "loss": 0.5311,
+      "mean_token_accuracy": 0.8424633145332336,
+      "num_tokens": 975258.0,
+      "step": 120
+    },
+    {
+      "entropy": 0.6882431030273437,
+      "epoch": 0.07054176072234764,
+      "grad_norm": 1.4671064615249634,
+      "learning_rate": 4.999886727993704e-06,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.8341028809547424,
+      "num_tokens": 1015761.0,
+      "step": 125
+    },
+    {
+      "entropy": 0.7173007011413575,
+      "epoch": 0.07336343115124154,
+      "grad_norm": 1.4032313823699951,
+      "learning_rate": 4.999877409085892e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8307689428329468,
+      "num_tokens": 1056694.0,
+      "step": 130
+    },
+    {
+      "entropy": 0.6990100502967834,
+      "epoch": 0.07618510158013544,
+      "grad_norm": 1.3346563577651978,
+      "learning_rate": 4.999867721858168e-06,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8263980627059937,
+      "num_tokens": 1097488.0,
+      "step": 135
+    },
+    {
+      "entropy": 0.6879770636558533,
+      "epoch": 0.07900677200902935,
+      "grad_norm": 1.5278775691986084,
+      "learning_rate": 4.999857666312438e-06,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8352189302444458,
+      "num_tokens": 1138244.0,
+      "step": 140
+    },
+    {
+      "entropy": 0.6410947203636169,
+      "epoch": 0.08182844243792325,
+      "grad_norm": 1.5688629150390625,
+      "learning_rate": 4.999847242450674e-06,
+      "loss": 0.5137,
+      "mean_token_accuracy": 0.8461119532585144,
+      "num_tokens": 1179012.0,
+      "step": 145
+    },
+    {
+      "entropy": 0.669737708568573,
+      "epoch": 0.08465011286681716,
+      "grad_norm": 1.421505093574524,
+      "learning_rate": 4.999836450274926e-06,
+      "loss": 0.5398,
+      "mean_token_accuracy": 0.8409364104270936,
+      "num_tokens": 1219636.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.7268872022628784,
+      "epoch": 0.08747178329571106,
+      "grad_norm": 1.4139888286590576,
+      "learning_rate": 4.999825289787314e-06,
+      "loss": 0.5906,
+      "mean_token_accuracy": 0.8278973698616028,
+      "num_tokens": 1260358.0,
+      "step": 155
+    },
+    {
+      "entropy": 0.6793547749519349,
+      "epoch": 0.09029345372460497,
+      "grad_norm": 1.3994945287704468,
+      "learning_rate": 4.99981376099003e-06,
+      "loss": 0.5473,
+      "mean_token_accuracy": 0.8384710073471069,
+      "num_tokens": 1300160.0,
+      "step": 160
+    },
+    {
+      "entropy": 0.7041888356208801,
+      "epoch": 0.09311512415349887,
+      "grad_norm": 1.2893599271774292,
+      "learning_rate": 4.999801863885339e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8308470964431762,
+      "num_tokens": 1340909.0,
+      "step": 165
+    },
+    {
+      "entropy": 0.7476347208023071,
+      "epoch": 0.09593679458239278,
+      "grad_norm": 1.4180552959442139,
+      "learning_rate": 4.999789598475578e-06,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.8223087549209595,
+      "num_tokens": 1381254.0,
+      "step": 170
+    },
+    {
+      "entropy": 0.7091502547264099,
+      "epoch": 0.09875846501128668,
+      "grad_norm": 1.2832435369491577,
+      "learning_rate": 4.999776964763157e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8346888542175293,
+      "num_tokens": 1421956.0,
+      "step": 175
+    },
+    {
+      "entropy": 0.6569717288017273,
+      "epoch": 0.10158013544018059,
+      "grad_norm": 1.4216198921203613,
+      "learning_rate": 4.999763962750557e-06,
+      "loss": 0.5289,
+      "mean_token_accuracy": 0.8407601952552796,
+      "num_tokens": 1462724.0,
+      "step": 180
+    },
+    {
+      "entropy": 0.7254927158355713,
+      "epoch": 0.1044018058690745,
+      "grad_norm": 1.367143154144287,
+      "learning_rate": 4.999750592440333e-06,
+      "loss": 0.603,
+      "mean_token_accuracy": 0.8244819164276123,
+      "num_tokens": 1503032.0,
+      "step": 185
+    },
+    {
+      "entropy": 0.6174264311790466,
+      "epoch": 0.1072234762979684,
+      "grad_norm": 1.1871576309204102,
+      "learning_rate": 4.999736853835111e-06,
+      "loss": 0.5176,
+      "mean_token_accuracy": 0.8440237402915954,
+      "num_tokens": 1543678.0,
+      "step": 190
+    },
+    {
+      "entropy": 0.6621920347213746,
+      "epoch": 0.1100451467268623,
+      "grad_norm": 1.3519974946975708,
+      "learning_rate": 4.999722746937591e-06,
+      "loss": 0.5392,
+      "mean_token_accuracy": 0.8406000256538391,
+      "num_tokens": 1584469.0,
+      "step": 195
+    },
+    {
+      "entropy": 0.72998526096344,
+      "epoch": 0.11286681715575621,
+      "grad_norm": 1.5044736862182617,
+      "learning_rate": 4.999708271750544e-06,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8296970129013062,
+      "num_tokens": 1625117.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.620005464553833,
+      "epoch": 0.11568848758465011,
+      "grad_norm": 1.2598135471343994,
+      "learning_rate": 4.999693428276813e-06,
+      "loss": 0.5077,
+      "mean_token_accuracy": 0.8482168078422546,
+      "num_tokens": 1665890.0,
+      "step": 205
+    },
+    {
+      "entropy": 0.7349537253379822,
+      "epoch": 0.11851015801354402,
+      "grad_norm": 1.290712833404541,
+      "learning_rate": 4.999678216519314e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8309862017631531,
+      "num_tokens": 1706501.0,
+      "step": 210
+    },
+    {
+      "entropy": 0.6702666759490967,
+      "epoch": 0.12133182844243792,
+      "grad_norm": 1.1792408227920532,
+      "learning_rate": 4.999662636481035e-06,
+      "loss": 0.5404,
+      "mean_token_accuracy": 0.8420506477355957,
+      "num_tokens": 1747287.0,
+      "step": 215
+    },
+    {
+      "entropy": 0.7209498405456543,
+      "epoch": 0.12415349887133183,
+      "grad_norm": 1.4285236597061157,
+      "learning_rate": 4.999646688165039e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8266986727714538,
+      "num_tokens": 1787814.0,
+      "step": 220
+    },
+    {
+      "entropy": 0.6917775988578796,
+      "epoch": 0.12697516930022573,
+      "grad_norm": 1.4500813484191895,
+      "learning_rate": 4.999630371574457e-06,
+      "loss": 0.5451,
+      "mean_token_accuracy": 0.8405853033065795,
+      "num_tokens": 1828330.0,
+      "step": 225
+    },
+    {
+      "entropy": 0.7286237478256226,
+      "epoch": 0.12979683972911965,
+      "grad_norm": 1.197171688079834,
+      "learning_rate": 4.999613686712493e-06,
+      "loss": 0.5945,
+      "mean_token_accuracy": 0.8261087775230408,
+      "num_tokens": 1868916.0,
+      "step": 230
+    },
+    {
+      "entropy": 0.7036802411079407,
+      "epoch": 0.13261851015801354,
+      "grad_norm": 1.3373335599899292,
+      "learning_rate": 4.999596633582429e-06,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8307413101196289,
+      "num_tokens": 1909557.0,
+      "step": 235
+    },
+    {
+      "entropy": 0.6876640558242798,
+      "epoch": 0.13544018058690746,
+      "grad_norm": 1.2515480518341064,
+      "learning_rate": 4.999579212187611e-06,
+      "loss": 0.5421,
+      "mean_token_accuracy": 0.8372658848762512,
+      "num_tokens": 1949962.0,
+      "step": 240
+    },
+    {
+      "entropy": 0.6509540438652038,
+      "epoch": 0.13826185101580135,
+      "grad_norm": 1.2757313251495361,
+      "learning_rate": 4.999561422531464e-06,
+      "loss": 0.5263,
+      "mean_token_accuracy": 0.8403837084770203,
+      "num_tokens": 1990703.0,
+      "step": 245
+    },
+    {
+      "entropy": 0.6634021878242493,
+      "epoch": 0.14108352144469527,
+      "grad_norm": 1.3375166654586792,
+      "learning_rate": 4.9995432646174815e-06,
+      "loss": 0.5638,
+      "mean_token_accuracy": 0.831636929512024,
+      "num_tokens": 2031252.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7269237041473389,
+      "epoch": 0.14390519187358916,
+      "grad_norm": 1.3792425394058228,
+      "learning_rate": 4.9995247384492314e-06,
+      "loss": 0.5982,
+      "mean_token_accuracy": 0.8250751614570617,
+      "num_tokens": 2071662.0,
+      "step": 255
+    },
+    {
+      "entropy": 0.7198888540267945,
+      "epoch": 0.14672686230248308,
+      "grad_norm": 1.4608137607574463,
+      "learning_rate": 4.999505844030352e-06,
+      "loss": 0.5952,
+      "mean_token_accuracy": 0.8260551571846009,
+      "num_tokens": 2112264.0,
+      "step": 260
+    },
+    {
+      "entropy": 0.6873027205467224,
+      "epoch": 0.14954853273137697,
+      "grad_norm": 1.3915055990219116,
+      "learning_rate": 4.999486581364557e-06,
+      "loss": 0.5486,
+      "mean_token_accuracy": 0.837960147857666,
+      "num_tokens": 2152995.0,
+      "step": 265
+    },
+    {
+      "entropy": 0.7061834812164307,
+      "epoch": 0.1523702031602709,
+      "grad_norm": 1.3316701650619507,
+      "learning_rate": 4.999466950455628e-06,
+      "loss": 0.5651,
+      "mean_token_accuracy": 0.8314638137817383,
+      "num_tokens": 2193728.0,
+      "step": 270
+    },
+    {
+      "entropy": 0.6988749980926514,
+      "epoch": 0.15519187358916478,
+      "grad_norm": 1.4248734712600708,
+      "learning_rate": 4.999446951307424e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8328610777854919,
+      "num_tokens": 2234493.0,
+      "step": 275
+    },
+    {
+      "entropy": 0.7523432970046997,
+      "epoch": 0.1580135440180587,
+      "grad_norm": 1.5484312772750854,
+      "learning_rate": 4.999426583923873e-06,
+      "loss": 0.6236,
+      "mean_token_accuracy": 0.8170114874839782,
+      "num_tokens": 2274609.0,
+      "step": 280
+    },
+    {
+      "entropy": 0.6838536024093628,
+      "epoch": 0.1608352144469526,
+      "grad_norm": 1.367386817932129,
+      "learning_rate": 4.999405848308975e-06,
+      "loss": 0.5309,
+      "mean_token_accuracy": 0.842476773262024,
+      "num_tokens": 2315124.0,
+      "step": 285
+    },
+    {
+      "entropy": 0.6006429553031921,
+      "epoch": 0.1636568848758465,
+      "grad_norm": 1.3099923133850098,
+      "learning_rate": 4.999384744466805e-06,
+      "loss": 0.4904,
+      "mean_token_accuracy": 0.8527785658836364,
+      "num_tokens": 2355713.0,
+      "step": 290
+    },
+    {
+      "entropy": 0.6720806241035462,
+      "epoch": 0.1664785553047404,
+      "grad_norm": 1.2152323722839355,
+      "learning_rate": 4.999363272401508e-06,
+      "loss": 0.5263,
+      "mean_token_accuracy": 0.8428802371025086,
+      "num_tokens": 2395726.0,
+      "step": 295
+    },
+    {
+      "entropy": 0.730102264881134,
+      "epoch": 0.16930022573363432,
+      "grad_norm": 1.3833410739898682,
+      "learning_rate": 4.9993414321173014e-06,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.8238815784454345,
+      "num_tokens": 2436268.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7030096173286438,
+      "epoch": 0.1721218961625282,
+      "grad_norm": 1.3912793397903442,
+      "learning_rate": 4.9993192236184786e-06,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.829773461818695,
+      "num_tokens": 2476941.0,
+      "step": 305
+    },
+    {
+      "entropy": 0.692926001548767,
+      "epoch": 0.17494356659142213,
+      "grad_norm": 1.406414270401001,
+      "learning_rate": 4.9992966469094005e-06,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8338650107383728,
+      "num_tokens": 2517495.0,
+      "step": 310
+    },
+    {
+      "entropy": 0.7009713649749756,
+      "epoch": 0.17776523702031602,
+      "grad_norm": 1.4576942920684814,
+      "learning_rate": 4.999273701994501e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8266690850257874,
+      "num_tokens": 2558100.0,
+      "step": 315
+    },
+    {
+      "entropy": 0.6718906760215759,
+      "epoch": 0.18058690744920994,
+      "grad_norm": 1.4303754568099976,
+      "learning_rate": 4.999250388878291e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.830142343044281,
+      "num_tokens": 2598172.0,
+      "step": 320
+    },
+    {
+      "entropy": 0.6644892454147339,
+      "epoch": 0.18340857787810383,
+      "grad_norm": 1.3855103254318237,
+      "learning_rate": 4.999226707565348e-06,
+      "loss": 0.5289,
+      "mean_token_accuracy": 0.8401578664779663,
+      "num_tokens": 2638766.0,
+      "step": 325
+    },
+    {
+      "entropy": 0.6168012738227844,
+      "epoch": 0.18623024830699775,
+      "grad_norm": 1.3673598766326904,
+      "learning_rate": 4.999202658060324e-06,
+      "loss": 0.5022,
+      "mean_token_accuracy": 0.8519076347351074,
+      "num_tokens": 2679412.0,
+      "step": 330
+    },
+    {
+      "entropy": 0.7058925032615662,
+      "epoch": 0.18905191873589164,
+      "grad_norm": 1.2969095706939697,
+      "learning_rate": 4.9991782403679445e-06,
+      "loss": 0.5568,
+      "mean_token_accuracy": 0.8322369575500488,
+      "num_tokens": 2720202.0,
+      "step": 335
+    },
+    {
+      "entropy": 0.6692469239234924,
+      "epoch": 0.19187358916478556,
+      "grad_norm": 1.3060578107833862,
+      "learning_rate": 4.999153454493006e-06,
+      "loss": 0.5492,
+      "mean_token_accuracy": 0.837912917137146,
+      "num_tokens": 2760837.0,
+      "step": 340
+    },
+    {
+      "entropy": 0.6806850790977478,
+      "epoch": 0.19469525959367945,
+      "grad_norm": 1.3894520998001099,
+      "learning_rate": 4.999128300440377e-06,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.8271700620651246,
+      "num_tokens": 2801488.0,
+      "step": 345
+    },
+    {
+      "entropy": 0.6444134593009949,
+      "epoch": 0.19751693002257337,
+      "grad_norm": 1.2214727401733398,
+      "learning_rate": 4.9991027782150005e-06,
+      "loss": 0.4907,
+      "mean_token_accuracy": 0.8515629053115845,
+      "num_tokens": 2842036.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.720831036567688,
+      "epoch": 0.20033860045146726,
+      "grad_norm": 1.3418818712234497,
+      "learning_rate": 4.99907688782189e-06,
+      "loss": 0.5878,
+      "mean_token_accuracy": 0.8290072560310364,
+      "num_tokens": 2882628.0,
+      "step": 355
+    },
+    {
+      "entropy": 0.6308605074882507,
+      "epoch": 0.20316027088036118,
+      "grad_norm": 1.3402037620544434,
+      "learning_rate": 4.9990506292661315e-06,
+      "loss": 0.5121,
+      "mean_token_accuracy": 0.8474404573440552,
+      "num_tokens": 2923198.0,
+      "step": 360
+    },
+    {
+      "entropy": 0.6803479313850402,
+      "epoch": 0.20598194130925507,
+      "grad_norm": 1.4827245473861694,
+      "learning_rate": 4.9990240025528825e-06,
+      "loss": 0.5541,
+      "mean_token_accuracy": 0.8327592492103577,
+      "num_tokens": 2963798.0,
+      "step": 365
+    },
+    {
+      "entropy": 0.6540728807449341,
+      "epoch": 0.208803611738149,
+      "grad_norm": 1.3350443840026855,
+      "learning_rate": 4.998997007687375e-06,
+      "loss": 0.5089,
+      "mean_token_accuracy": 0.8447277069091796,
+      "num_tokens": 3004683.0,
+      "step": 370
+    },
+    {
+      "entropy": 0.7112419605255127,
+      "epoch": 0.21162528216704288,
+      "grad_norm": 1.3664087057113647,
+      "learning_rate": 4.998969644674911e-06,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8266646385192871,
+      "num_tokens": 3045439.0,
+      "step": 375
+    },
+    {
+      "entropy": 0.6424817204475403,
+      "epoch": 0.2144469525959368,
+      "grad_norm": 1.295762300491333,
+      "learning_rate": 4.998941913520867e-06,
+      "loss": 0.5321,
+      "mean_token_accuracy": 0.8423674941062927,
+      "num_tokens": 3086082.0,
+      "step": 380
+    },
+    {
+      "entropy": 0.6630952477455139,
+      "epoch": 0.2172686230248307,
+      "grad_norm": 1.189756989479065,
+      "learning_rate": 4.998913814230691e-06,
+      "loss": 0.5333,
+      "mean_token_accuracy": 0.839359712600708,
+      "num_tokens": 3126821.0,
+      "step": 385
+    },
+    {
+      "entropy": 0.7343137979507446,
+      "epoch": 0.2200902934537246,
+      "grad_norm": 1.3133877515792847,
+      "learning_rate": 4.998885346809902e-06,
+      "loss": 0.616,
+      "mean_token_accuracy": 0.820310366153717,
+      "num_tokens": 3167554.0,
+      "step": 390
+    },
+    {
+      "entropy": 0.7052167534828186,
+      "epoch": 0.2229119638826185,
+      "grad_norm": 1.2900844812393188,
+      "learning_rate": 4.998856511264094e-06,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.8320470333099366,
+      "num_tokens": 3208157.0,
+      "step": 395
+    },
+    {
+      "entropy": 0.7610344290733337,
+      "epoch": 0.22573363431151242,
+      "grad_norm": 1.5215799808502197,
+      "learning_rate": 4.99882730759893e-06,
+      "loss": 0.6134,
+      "mean_token_accuracy": 0.81887845993042,
+      "num_tokens": 3248261.0,
+      "step": 400
+    },
+    {
+      "entropy": 0.6336552977561951,
+      "epoch": 0.22855530474040633,
+      "grad_norm": 1.2715165615081787,
+      "learning_rate": 4.9987977358201475e-06,
+      "loss": 0.5034,
+      "mean_token_accuracy": 0.8473275661468506,
+      "num_tokens": 3288888.0,
+      "step": 405
+    },
+    {
+      "entropy": 0.6352953314781189,
+      "epoch": 0.23137697516930023,
+      "grad_norm": 1.3592379093170166,
+      "learning_rate": 4.998767795933557e-06,
+      "loss": 0.5094,
+      "mean_token_accuracy": 0.8488934755325317,
+      "num_tokens": 3329423.0,
+      "step": 410
+    },
+    {
+      "entropy": 0.6898619294166565,
+      "epoch": 0.23419864559819414,
+      "grad_norm": 1.5599400997161865,
+      "learning_rate": 4.998737487945039e-06,
+      "loss": 0.5475,
+      "mean_token_accuracy": 0.8362764358520508,
+      "num_tokens": 3369896.0,
+      "step": 415
+    },
+    {
+      "entropy": 0.6755191922187805,
+      "epoch": 0.23702031602708803,
+      "grad_norm": 1.3778337240219116,
+      "learning_rate": 4.998706811860548e-06,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8319776773452758,
+      "num_tokens": 3410527.0,
+      "step": 420
+    },
+    {
+      "entropy": 0.6867754459381104,
+      "epoch": 0.23984198645598195,
+      "grad_norm": 1.3699222803115845,
+      "learning_rate": 4.99867576768611e-06,
+      "loss": 0.5414,
+      "mean_token_accuracy": 0.839304780960083,
+      "num_tokens": 3451077.0,
+      "step": 425
+    },
+    {
+      "entropy": 0.683435583114624,
+      "epoch": 0.24266365688487584,
+      "grad_norm": 1.3843740224838257,
+      "learning_rate": 4.9986443554278244e-06,
+      "loss": 0.5439,
+      "mean_token_accuracy": 0.8384104609489441,
+      "num_tokens": 3491423.0,
+      "step": 430
+    },
+    {
+      "entropy": 0.6476885080337524,
+      "epoch": 0.24548532731376976,
+      "grad_norm": 1.2700644731521606,
+      "learning_rate": 4.998612575091861e-06,
+      "loss": 0.5217,
+      "mean_token_accuracy": 0.8458979368209839,
+      "num_tokens": 3532023.0,
+      "step": 435
+    },
+    {
+      "entropy": 0.6664133548736573,
+      "epoch": 0.24830699774266365,
+      "grad_norm": 1.3746997117996216,
+      "learning_rate": 4.998580426684464e-06,
+      "loss": 0.5236,
+      "mean_token_accuracy": 0.8449108362197876,
+      "num_tokens": 3572699.0,
+      "step": 440
+    },
+    {
+      "entropy": 0.6357210874557495,
+      "epoch": 0.25112866817155755,
+      "grad_norm": 1.2438454627990723,
+      "learning_rate": 4.99854791021195e-06,
+      "loss": 0.512,
+      "mean_token_accuracy": 0.8491321802139282,
+      "num_tokens": 3613440.0,
+      "step": 445
+    },
+    {
+      "entropy": 0.627473509311676,
+      "epoch": 0.25395033860045146,
+      "grad_norm": 1.1443346738815308,
+      "learning_rate": 4.998515025680703e-06,
+      "loss": 0.5025,
+      "mean_token_accuracy": 0.8488677740097046,
+      "num_tokens": 3653969.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.6401142597198486,
+      "epoch": 0.2567720090293454,
+      "grad_norm": 1.3476841449737549,
+      "learning_rate": 4.998481773097187e-06,
+      "loss": 0.503,
+      "mean_token_accuracy": 0.8511731863021851,
+      "num_tokens": 3693874.0,
+      "step": 455
+    },
+    {
+      "entropy": 0.645455002784729,
+      "epoch": 0.2595936794582393,
+      "grad_norm": 1.392980933189392,
+      "learning_rate": 4.998448152467933e-06,
+      "loss": 0.515,
+      "mean_token_accuracy": 0.8437379837036133,
+      "num_tokens": 3734684.0,
+      "step": 460
+    },
+    {
+      "entropy": 0.7039332389831543,
+      "epoch": 0.26241534988713316,
+      "grad_norm": 1.42012357711792,
+      "learning_rate": 4.998414163799545e-06,
+      "loss": 0.5646,
+      "mean_token_accuracy": 0.8330328106880188,
+      "num_tokens": 3775233.0,
+      "step": 465
+    },
+    {
+      "entropy": 0.6496623396873474,
+      "epoch": 0.2652370203160271,
+      "grad_norm": 1.168820858001709,
+      "learning_rate": 4.998379807098703e-06,
+      "loss": 0.512,
+      "mean_token_accuracy": 0.8448665142059326,
+      "num_tokens": 3815898.0,
+      "step": 470
+    },
+    {
+      "entropy": 0.6927712917327881,
+      "epoch": 0.268058690744921,
+      "grad_norm": 1.277652621269226,
+      "learning_rate": 4.998345082372153e-06,
+      "loss": 0.5485,
+      "mean_token_accuracy": 0.8371129631996155,
+      "num_tokens": 3856530.0,
+      "step": 475
+    },
+    {
+      "entropy": 0.6336887121200562,
+      "epoch": 0.2708803611738149,
+      "grad_norm": 1.3250688314437866,
+      "learning_rate": 4.998309989626718e-06,
+      "loss": 0.496,
+      "mean_token_accuracy": 0.8498493075370789,
+      "num_tokens": 3897409.0,
+      "step": 480
+    },
+    {
+      "entropy": 0.645351254940033,
+      "epoch": 0.2737020316027088,
+      "grad_norm": 1.1753090620040894,
+      "learning_rate": 4.998274528869292e-06,
+      "loss": 0.5256,
+      "mean_token_accuracy": 0.8436009883880615,
+      "num_tokens": 3938183.0,
+      "step": 485
+    },
+    {
+      "entropy": 0.6607337474822998,
+      "epoch": 0.2765237020316027,
+      "grad_norm": 1.386691927909851,
+      "learning_rate": 4.998238700106842e-06,
+      "loss": 0.5226,
+      "mean_token_accuracy": 0.8447944283485412,
+      "num_tokens": 3978774.0,
+      "step": 490
+    },
+    {
+      "entropy": 0.7330415606498718,
+      "epoch": 0.2793453724604966,
+      "grad_norm": 1.3130143880844116,
+      "learning_rate": 4.998202503346405e-06,
+      "loss": 0.5945,
+      "mean_token_accuracy": 0.8254420638084412,
+      "num_tokens": 4019301.0,
+      "step": 495
+    },
+    {
+      "entropy": 0.6656695961952209,
+      "epoch": 0.28216704288939054,
+      "grad_norm": 1.4055564403533936,
+      "learning_rate": 4.998165938595094e-06,
+      "loss": 0.5339,
+      "mean_token_accuracy": 0.8389903545379639,
+      "num_tokens": 4059906.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.28216704288939054,
+      "eval_entropy": 0.6397659778594971,
+      "eval_loss": 0.5099606513977051,
+      "eval_mean_token_accuracy": 0.857033371925354,
+      "eval_num_tokens": 4059906.0,
+      "eval_runtime": 0.1662,
+      "eval_samples_per_second": 24.067,
+      "eval_steps_per_second": 6.017,
+      "step": 500
+    },
+    {
+      "entropy": 0.6222930490970612,
+      "epoch": 0.2849887133182844,
+      "grad_norm": 1.3186826705932617,
+      "learning_rate": 4.99812900586009e-06,
+      "loss": 0.5115,
+      "mean_token_accuracy": 0.8475234031677246,
+      "num_tokens": 4100501.0,
+      "step": 505
+    },
+    {
+      "entropy": 0.6861591100692749,
+      "epoch": 0.2878103837471783,
+      "grad_norm": 1.2481554746627808,
+      "learning_rate": 4.998091705148649e-06,
+      "loss": 0.5386,
+      "mean_token_accuracy": 0.8370915293693543,
+      "num_tokens": 4141056.0,
+      "step": 510
+    },
+    {
+      "entropy": 0.6614163756370545,
+      "epoch": 0.29063205417607224,
+      "grad_norm": 1.3787559270858765,
+      "learning_rate": 4.998054036468099e-06,
+      "loss": 0.5331,
+      "mean_token_accuracy": 0.8423222064971924,
+      "num_tokens": 4181622.0,
+      "step": 515
+    },
+    {
+      "entropy": 0.6498169541358948,
+      "epoch": 0.29345372460496616,
+      "grad_norm": 1.140236735343933,
+      "learning_rate": 4.9980159998258406e-06,
+      "loss": 0.5086,
+      "mean_token_accuracy": 0.846357011795044,
+      "num_tokens": 4222377.0,
+      "step": 520
+    },
+    {
+      "entropy": 0.6879486322402955,
+      "epoch": 0.29627539503386,
+      "grad_norm": 1.4712022542953491,
+      "learning_rate": 4.997977595229346e-06,
+      "loss": 0.5573,
+      "mean_token_accuracy": 0.835738730430603,
+      "num_tokens": 4263122.0,
+      "step": 525
+    },
+    {
+      "entropy": 0.6771662116050721,
+      "epoch": 0.29909706546275394,
+      "grad_norm": 1.3491076231002808,
+      "learning_rate": 4.997938822686158e-06,
+      "loss": 0.547,
+      "mean_token_accuracy": 0.8368203639984131,
+      "num_tokens": 4303690.0,
+      "step": 530
+    },
+    {
+      "entropy": 0.614617896080017,
+      "epoch": 0.30191873589164786,
+      "grad_norm": 1.2441357374191284,
+      "learning_rate": 4.9978996822038964e-06,
+      "loss": 0.4844,
+      "mean_token_accuracy": 0.8524455547332763,
+      "num_tokens": 4344245.0,
+      "step": 535
+    },
+    {
+      "entropy": 0.6798507928848266,
+      "epoch": 0.3047404063205418,
+      "grad_norm": 1.3364735841751099,
+      "learning_rate": 4.997860173790247e-06,
+      "loss": 0.5325,
+      "mean_token_accuracy": 0.8381256461143494,
+      "num_tokens": 4384408.0,
+      "step": 540
+    },
+    {
+      "entropy": 0.6491321206092835,
+      "epoch": 0.30756207674943564,
+      "grad_norm": 1.3930151462554932,
+      "learning_rate": 4.997820297452975e-06,
+      "loss": 0.4999,
+      "mean_token_accuracy": 0.8447855472564697,
+      "num_tokens": 4424758.0,
+      "step": 545
+    },
+    {
+      "entropy": 0.6349780797958374,
+      "epoch": 0.31038374717832956,
+      "grad_norm": 1.406150460243225,
+      "learning_rate": 4.99778005319991e-06,
+      "loss": 0.51,
+      "mean_token_accuracy": 0.8480640172958374,
+      "num_tokens": 4465315.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.6838363170623779,
+      "epoch": 0.3132054176072235,
+      "grad_norm": 1.5294631719589233,
+      "learning_rate": 4.997739441038962e-06,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8342589735984802,
+      "num_tokens": 4505898.0,
+      "step": 555
+    },
+    {
+      "entropy": 0.7220178842544556,
+      "epoch": 0.3160270880361174,
+      "grad_norm": 1.3367938995361328,
+      "learning_rate": 4.997698460978107e-06,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8219369530677796,
+      "num_tokens": 4546647.0,
+      "step": 560
+    },
+    {
+      "entropy": 0.6441617131233215,
+      "epoch": 0.31884875846501126,
+      "grad_norm": 1.1907154321670532,
+      "learning_rate": 4.997657113025395e-06,
+      "loss": 0.5127,
+      "mean_token_accuracy": 0.8480484962463379,
+      "num_tokens": 4587353.0,
+      "step": 565
+    },
+    {
+      "entropy": 0.7197385430335999,
+      "epoch": 0.3216704288939052,
+      "grad_norm": 1.5142972469329834,
+      "learning_rate": 4.99761539718895e-06,
+      "loss": 0.5569,
+      "mean_token_accuracy": 0.8342020034790039,
+      "num_tokens": 4627673.0,
+      "step": 570
+    },
+    {
+      "entropy": 0.7443452477455139,
+      "epoch": 0.3244920993227991,
+      "grad_norm": 1.2722396850585938,
+      "learning_rate": 4.997573313476966e-06,
+      "loss": 0.6036,
+      "mean_token_accuracy": 0.8234388828277588,
+      "num_tokens": 4668389.0,
+      "step": 575
+    },
+    {
+      "entropy": 0.6642969369888305,
+      "epoch": 0.327313769751693,
+      "grad_norm": 1.3420699834823608,
+      "learning_rate": 4.997530861897713e-06,
+      "loss": 0.5121,
+      "mean_token_accuracy": 0.8478710532188416,
+      "num_tokens": 4708970.0,
+      "step": 580
+    },
+    {
+      "entropy": 0.6479909062385559,
+      "epoch": 0.3301354401805869,
+      "grad_norm": 1.422701358795166,
+      "learning_rate": 4.997488042459528e-06,
+      "loss": 0.5289,
+      "mean_token_accuracy": 0.8437910914421082,
+      "num_tokens": 4749674.0,
+      "step": 585
+    },
+    {
+      "entropy": 0.6523343324661255,
+      "epoch": 0.3329571106094808,
+      "grad_norm": 1.3672442436218262,
+      "learning_rate": 4.997444855170823e-06,
+      "loss": 0.531,
+      "mean_token_accuracy": 0.8426079154014587,
+      "num_tokens": 4790249.0,
+      "step": 590
+    },
+    {
+      "entropy": 0.6659448742866516,
+      "epoch": 0.3357787810383747,
+      "grad_norm": 1.267858624458313,
+      "learning_rate": 4.997401300040084e-06,
+      "loss": 0.5406,
+      "mean_token_accuracy": 0.8389504075050354,
+      "num_tokens": 4831050.0,
+      "step": 595
+    },
+    {
+      "entropy": 0.6824786901473999,
+      "epoch": 0.33860045146726864,
+      "grad_norm": 1.3464597463607788,
+      "learning_rate": 4.997357377075866e-06,
+      "loss": 0.5328,
+      "mean_token_accuracy": 0.8403973698616027,
+      "num_tokens": 4871225.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.6604692697525024,
+      "epoch": 0.34142212189616256,
+      "grad_norm": 1.4553182125091553,
+      "learning_rate": 4.997313086286797e-06,
+      "loss": 0.5313,
+      "mean_token_accuracy": 0.8407448887825012,
+      "num_tokens": 4911582.0,
+      "step": 605
+    },
+    {
+      "entropy": 0.6908664226531982,
+      "epoch": 0.3442437923250564,
+      "grad_norm": 1.3354190587997437,
+      "learning_rate": 4.997268427681579e-06,
+      "loss": 0.5585,
+      "mean_token_accuracy": 0.8351940751075745,
+      "num_tokens": 4952166.0,
+      "step": 610
+    },
+    {
+      "entropy": 0.6428348064422608,
+      "epoch": 0.34706546275395034,
+      "grad_norm": 1.2946590185165405,
+      "learning_rate": 4.997223401268985e-06,
+      "loss": 0.5154,
+      "mean_token_accuracy": 0.8446980834007263,
+      "num_tokens": 4992726.0,
+      "step": 615
+    },
+    {
+      "entropy": 0.6469574570655823,
+      "epoch": 0.34988713318284426,
+      "grad_norm": 1.3383957147598267,
+      "learning_rate": 4.9971780070578605e-06,
+      "loss": 0.5082,
+      "mean_token_accuracy": 0.8440989255905151,
+      "num_tokens": 5033375.0,
+      "step": 620
+    },
+    {
+      "entropy": 0.6753599047660828,
+      "epoch": 0.3527088036117382,
+      "grad_norm": 1.4977456331253052,
+      "learning_rate": 4.997132245057124e-06,
+      "loss": 0.5473,
+      "mean_token_accuracy": 0.8380128145217896,
+      "num_tokens": 5074101.0,
+      "step": 625
+    },
+    {
+      "entropy": 0.6009699106216431,
+      "epoch": 0.35553047404063204,
+      "grad_norm": 1.3025342226028442,
+      "learning_rate": 4.997086115275763e-06,
+      "loss": 0.4942,
+      "mean_token_accuracy": 0.8532807230949402,
+      "num_tokens": 5114789.0,
+      "step": 630
+    },
+    {
+      "entropy": 0.7020259857177734,
+      "epoch": 0.35835214446952596,
+      "grad_norm": 1.42304265499115,
+      "learning_rate": 4.997039617722843e-06,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8297113418579102,
+      "num_tokens": 5155557.0,
+      "step": 635
+    },
+    {
+      "entropy": 0.6381742715835571,
+      "epoch": 0.3611738148984199,
+      "grad_norm": 1.3583488464355469,
+      "learning_rate": 4.996992752407496e-06,
+      "loss": 0.4815,
+      "mean_token_accuracy": 0.8546654939651489,
+      "num_tokens": 5195254.0,
+      "step": 640
+    },
+    {
+      "entropy": 0.7070492506027222,
+      "epoch": 0.3639954853273138,
+      "grad_norm": 1.4853992462158203,
+      "learning_rate": 4.996945519338929e-06,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8343408942222595,
+      "num_tokens": 5235919.0,
+      "step": 645
+    },
+    {
+      "entropy": 0.7001669526100158,
+      "epoch": 0.36681715575620766,
+      "grad_norm": 1.263033390045166,
+      "learning_rate": 4.996897918526422e-06,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8342163324356079,
+      "num_tokens": 5276443.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.6899022579193115,
+      "epoch": 0.3696388261851016,
+      "grad_norm": 1.3277238607406616,
+      "learning_rate": 4.996849949979325e-06,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8299772024154664,
+      "num_tokens": 5317131.0,
+      "step": 655
+    },
+    {
+      "entropy": 0.6805570363998413,
+      "epoch": 0.3724604966139955,
+      "grad_norm": 1.2228710651397705,
+      "learning_rate": 4.996801613707063e-06,
+      "loss": 0.5515,
+      "mean_token_accuracy": 0.8347800970077515,
+      "num_tokens": 5357866.0,
+      "step": 660
+    },
+    {
+      "entropy": 0.6331149935722351,
+      "epoch": 0.3752821670428894,
+      "grad_norm": 1.3809823989868164,
+      "learning_rate": 4.9967529097191305e-06,
+      "loss": 0.5098,
+      "mean_token_accuracy": 0.8466669321060181,
+      "num_tokens": 5398573.0,
+      "step": 665
+    },
+    {
+      "entropy": 0.6736877083778381,
+      "epoch": 0.3781038374717833,
+      "grad_norm": 1.3853009939193726,
+      "learning_rate": 4.996703838025095e-06,
+      "loss": 0.5324,
+      "mean_token_accuracy": 0.839972174167633,
+      "num_tokens": 5439412.0,
+      "step": 670
+    },
+    {
+      "entropy": 0.6803161859512329,
+      "epoch": 0.3809255079006772,
+      "grad_norm": 1.4266542196273804,
+      "learning_rate": 4.996654398634597e-06,
+      "loss": 0.5506,
+      "mean_token_accuracy": 0.8333458185195923,
+      "num_tokens": 5479945.0,
+      "step": 675
+    },
+    {
+      "entropy": 0.7106911301612854,
+      "epoch": 0.3837471783295711,
+      "grad_norm": 1.482615351676941,
+      "learning_rate": 4.996604591557349e-06,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8281079649925231,
+      "num_tokens": 5520469.0,
+      "step": 680
+    },
+    {
+      "entropy": 0.6807895421981811,
+      "epoch": 0.38656884875846503,
+      "grad_norm": 1.2304539680480957,
+      "learning_rate": 4.996554416803137e-06,
+      "loss": 0.5497,
+      "mean_token_accuracy": 0.8346430540084839,
+      "num_tokens": 5561155.0,
+      "step": 685
+    },
+    {
+      "entropy": 0.6457465648651123,
+      "epoch": 0.3893905191873589,
+      "grad_norm": 1.470390796661377,
+      "learning_rate": 4.996503874381815e-06,
+      "loss": 0.5143,
+      "mean_token_accuracy": 0.846334969997406,
+      "num_tokens": 5601850.0,
+      "step": 690
+    },
+    {
+      "entropy": 0.5989644646644592,
+      "epoch": 0.3922121896162528,
+      "grad_norm": 1.3190268278121948,
+      "learning_rate": 4.996452964303315e-06,
+      "loss": 0.4601,
+      "mean_token_accuracy": 0.8595061659812927,
+      "num_tokens": 5642455.0,
+      "step": 695
+    },
+    {
+      "entropy": 0.6620395064353943,
+      "epoch": 0.39503386004514673,
+      "grad_norm": 1.4160014390945435,
+      "learning_rate": 4.996401686577636e-06,
+      "loss": 0.5315,
+      "mean_token_accuracy": 0.8426364421844482,
+      "num_tokens": 5683057.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.6351073563098908,
+      "epoch": 0.39785553047404065,
+      "grad_norm": 1.417966604232788,
+      "learning_rate": 4.996350041214852e-06,
+      "loss": 0.5148,
+      "mean_token_accuracy": 0.8437610864639282,
+      "num_tokens": 5723964.0,
+      "step": 705
+    },
+    {
+      "entropy": 0.7087972402572632,
+      "epoch": 0.4006772009029345,
+      "grad_norm": 1.4004472494125366,
+      "learning_rate": 4.996298028225111e-06,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8329956293106079,
+      "num_tokens": 5764573.0,
+      "step": 710
+    },
+    {
+      "entropy": 0.7214209794998169,
+      "epoch": 0.40349887133182843,
+      "grad_norm": 1.5199602842330933,
+      "learning_rate": 4.996245647618627e-06,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8271468758583069,
+      "num_tokens": 5805361.0,
+      "step": 715
+    },
+    {
+      "entropy": 0.6683865189552307,
+      "epoch": 0.40632054176072235,
+      "grad_norm": 1.3417999744415283,
+      "learning_rate": 4.996192899405693e-06,
+      "loss": 0.5318,
+      "mean_token_accuracy": 0.8423861384391784,
+      "num_tokens": 5846241.0,
+      "step": 720
+    },
+    {
+      "entropy": 0.7061811447143554,
+      "epoch": 0.40914221218961627,
+      "grad_norm": 1.456238031387329,
+      "learning_rate": 4.996139783596671e-06,
+      "loss": 0.5438,
+      "mean_token_accuracy": 0.8380335092544555,
+      "num_tokens": 5886839.0,
+      "step": 725
+    },
+    {
+      "entropy": 0.6855465412139893,
+      "epoch": 0.41196388261851014,
+      "grad_norm": 1.3961089849472046,
+      "learning_rate": 4.996086300201995e-06,
+      "loss": 0.5363,
+      "mean_token_accuracy": 0.838828194141388,
+      "num_tokens": 5927433.0,
+      "step": 730
+    },
+    {
+      "entropy": 0.6245934367179871,
+      "epoch": 0.41478555304740405,
+      "grad_norm": 1.252845048904419,
+      "learning_rate": 4.996032449232172e-06,
+      "loss": 0.5042,
+      "mean_token_accuracy": 0.8503703594207763,
+      "num_tokens": 5968213.0,
+      "step": 735
+    },
+    {
+      "entropy": 0.6127149283885955,
+      "epoch": 0.417607223476298,
+      "grad_norm": 1.207445502281189,
+      "learning_rate": 4.995978230697782e-06,
+      "loss": 0.4892,
+      "mean_token_accuracy": 0.8527963757514954,
+      "num_tokens": 6008851.0,
+      "step": 740
+    },
+    {
+      "entropy": 0.7167877793312073,
+      "epoch": 0.4204288939051919,
+      "grad_norm": 1.266051173210144,
+      "learning_rate": 4.995923644609474e-06,
+      "loss": 0.5714,
+      "mean_token_accuracy": 0.8294572830200195,
+      "num_tokens": 6049545.0,
+      "step": 745
+    },
+    {
+      "entropy": 0.5922774732112884,
+      "epoch": 0.42325056433408575,
+      "grad_norm": 1.3726112842559814,
+      "learning_rate": 4.995868690977974e-06,
+      "loss": 0.4756,
+      "mean_token_accuracy": 0.8563015341758728,
+      "num_tokens": 6090230.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6408256888389587,
+      "epoch": 0.4260722347629797,
+      "grad_norm": 1.3127083778381348,
+      "learning_rate": 4.995813369814075e-06,
+      "loss": 0.5208,
+      "mean_token_accuracy": 0.8428861260414123,
+      "num_tokens": 6130808.0,
+      "step": 755
+    },
+    {
+      "entropy": 0.6796544313430786,
+      "epoch": 0.4288939051918736,
+      "grad_norm": 1.409447431564331,
+      "learning_rate": 4.995757681128648e-06,
+      "loss": 0.5331,
+      "mean_token_accuracy": 0.8407068133354187,
+      "num_tokens": 6171110.0,
+      "step": 760
+    },
+    {
+      "entropy": 0.6506537318229675,
+      "epoch": 0.4317155756207675,
+      "grad_norm": 1.2559049129486084,
+      "learning_rate": 4.995701624932631e-06,
+      "loss": 0.4936,
+      "mean_token_accuracy": 0.8502960324287414,
+      "num_tokens": 6211953.0,
+      "step": 765
+    },
+    {
+      "entropy": 0.6637347579002381,
+      "epoch": 0.4345372460496614,
+      "grad_norm": 1.309942603111267,
+      "learning_rate": 4.995645201237036e-06,
+      "loss": 0.5321,
+      "mean_token_accuracy": 0.8399680137634278,
+      "num_tokens": 6252408.0,
+      "step": 770
+    },
+    {
+      "entropy": 0.6979188203811646,
+      "epoch": 0.4373589164785553,
+      "grad_norm": 1.5395498275756836,
+      "learning_rate": 4.995588410052948e-06,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8252558946609497,
+      "num_tokens": 6293047.0,
+      "step": 775
+    },
+    {
+      "entropy": 0.6070165634155273,
+      "epoch": 0.4401805869074492,
+      "grad_norm": 1.1005687713623047,
+      "learning_rate": 4.995531251391524e-06,
+      "loss": 0.488,
+      "mean_token_accuracy": 0.8509817957878113,
+      "num_tokens": 6333702.0,
+      "step": 780
+    },
+    {
+      "entropy": 0.6723381996154785,
+      "epoch": 0.44300225733634313,
+      "grad_norm": 1.27623450756073,
+      "learning_rate": 4.995473725263992e-06,
+      "loss": 0.5316,
+      "mean_token_accuracy": 0.8400357604026795,
+      "num_tokens": 6374319.0,
+      "step": 785
+    },
+    {
+      "entropy": 0.6453313708305359,
+      "epoch": 0.445823927765237,
+      "grad_norm": 1.3552383184432983,
+      "learning_rate": 4.995415831681654e-06,
+      "loss": 0.5294,
+      "mean_token_accuracy": 0.8450021266937255,
+      "num_tokens": 6415005.0,
+      "step": 790
+    },
+    {
+      "entropy": 0.7531963586807251,
+      "epoch": 0.4486455981941309,
+      "grad_norm": 1.303346037864685,
+      "learning_rate": 4.9953575706558835e-06,
+      "loss": 0.6133,
+      "mean_token_accuracy": 0.8193735122680664,
+      "num_tokens": 6455608.0,
+      "step": 795
+    },
+    {
+      "entropy": 0.6664739489555359,
+      "epoch": 0.45146726862302483,
+      "grad_norm": 1.3953213691711426,
+      "learning_rate": 4.9952989421981244e-06,
+      "loss": 0.5294,
+      "mean_token_accuracy": 0.8387627482414246,
+      "num_tokens": 6495958.0,
+      "step": 800
+    },
+    {
+      "entropy": 0.6475582718849182,
+      "epoch": 0.45428893905191875,
+      "grad_norm": 1.3377810716629028,
+      "learning_rate": 4.995239946319895e-06,
+      "loss": 0.5229,
+      "mean_token_accuracy": 0.8416747093200684,
+      "num_tokens": 6536670.0,
+      "step": 805
+    },
+    {
+      "entropy": 0.6350281596183777,
+      "epoch": 0.45711060948081267,
+      "grad_norm": 1.3273295164108276,
+      "learning_rate": 4.995180583032784e-06,
+      "loss": 0.513,
+      "mean_token_accuracy": 0.8475698709487915,
+      "num_tokens": 6577503.0,
+      "step": 810
+    },
+    {
+      "entropy": 0.6923723101615906,
+      "epoch": 0.45993227990970653,
+      "grad_norm": 1.4398008584976196,
+      "learning_rate": 4.9951208523484555e-06,
+      "loss": 0.5527,
+      "mean_token_accuracy": 0.834313976764679,
+      "num_tokens": 6617829.0,
+      "step": 815
+    },
+    {
+      "entropy": 0.6901078701019288,
+      "epoch": 0.46275395033860045,
+      "grad_norm": 1.2139265537261963,
+      "learning_rate": 4.995060754278642e-06,
+      "loss": 0.5589,
+      "mean_token_accuracy": 0.832787299156189,
+      "num_tokens": 6658320.0,
+      "step": 820
+    },
+    {
+      "entropy": 0.592526650428772,
+      "epoch": 0.46557562076749437,
+      "grad_norm": 1.34032142162323,
+      "learning_rate": 4.9950002888351514e-06,
+      "loss": 0.4736,
+      "mean_token_accuracy": 0.8565322160720825,
+      "num_tokens": 6699023.0,
+      "step": 825
+    },
+    {
+      "entropy": 0.713707959651947,
+      "epoch": 0.4683972911963883,
+      "grad_norm": 1.4716837406158447,
+      "learning_rate": 4.994939456029859e-06,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.8286639451980591,
+      "num_tokens": 6739446.0,
+      "step": 830
+    },
+    {
+      "entropy": 0.6702501654624939,
+      "epoch": 0.47121896162528215,
+      "grad_norm": 1.2845447063446045,
+      "learning_rate": 4.994878255874719e-06,
+      "loss": 0.5355,
+      "mean_token_accuracy": 0.8418485879898071,
+      "num_tokens": 6780190.0,
+      "step": 835
+    },
+    {
+      "entropy": 0.6634608149528504,
+      "epoch": 0.47404063205417607,
+      "grad_norm": 1.3624354600906372,
+      "learning_rate": 4.994816688381751e-06,
+      "loss": 0.5284,
+      "mean_token_accuracy": 0.8404200553894043,
+      "num_tokens": 6820877.0,
+      "step": 840
+    },
+    {
+      "entropy": 0.7302131295204163,
+      "epoch": 0.47686230248307,
+      "grad_norm": 1.4335999488830566,
+      "learning_rate": 4.994754753563054e-06,
+      "loss": 0.5943,
+      "mean_token_accuracy": 0.8234637618064881,
+      "num_tokens": 6861674.0,
+      "step": 845
+    },
+    {
+      "entropy": 0.6252121210098267,
+      "epoch": 0.4796839729119639,
+      "grad_norm": 1.423103928565979,
+      "learning_rate": 4.994692451430791e-06,
+      "loss": 0.5053,
+      "mean_token_accuracy": 0.8479559659957886,
+      "num_tokens": 6902472.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6240964293479919,
+      "epoch": 0.48250564334085777,
+      "grad_norm": 1.416658878326416,
+      "learning_rate": 4.9946297819972025e-06,
+      "loss": 0.509,
+      "mean_token_accuracy": 0.8452272534370422,
+      "num_tokens": 6942877.0,
+      "step": 855
+    },
+    {
+      "entropy": 0.6672571420669555,
+      "epoch": 0.4853273137697517,
+      "grad_norm": 1.3133835792541504,
+      "learning_rate": 4.994566745274601e-06,
+      "loss": 0.52,
+      "mean_token_accuracy": 0.8426130771636963,
+      "num_tokens": 6983426.0,
+      "step": 860
+    },
+    {
+      "entropy": 0.6581945300102234,
+      "epoch": 0.4881489841986456,
+      "grad_norm": 1.4337458610534668,
+      "learning_rate": 4.994503341275369e-06,
+      "loss": 0.5225,
+      "mean_token_accuracy": 0.8437890887260437,
+      "num_tokens": 7023732.0,
+      "step": 865
+    },
+    {
+      "entropy": 0.6551531553268433,
+      "epoch": 0.4909706546275395,
+      "grad_norm": 1.3199399709701538,
+      "learning_rate": 4.994439570011963e-06,
+      "loss": 0.518,
+      "mean_token_accuracy": 0.8422786116600036,
+      "num_tokens": 7064450.0,
+      "step": 870
+    },
+    {
+      "entropy": 0.6477535367012024,
+      "epoch": 0.4937923250564334,
+      "grad_norm": 1.2373626232147217,
+      "learning_rate": 4.99437543149691e-06,
+      "loss": 0.5278,
+      "mean_token_accuracy": 0.8424580574035645,
+      "num_tokens": 7105220.0,
+      "step": 875
+    },
+    {
+      "entropy": 0.6733549952507019,
+      "epoch": 0.4966139954853273,
+      "grad_norm": 1.3904764652252197,
+      "learning_rate": 4.994310925742811e-06,
+      "loss": 0.5219,
+      "mean_token_accuracy": 0.8425598025321961,
+      "num_tokens": 7145455.0,
+      "step": 880
+    },
+    {
+      "entropy": 0.6530658364295959,
+      "epoch": 0.4994356659142212,
+      "grad_norm": 1.387152075767517,
+      "learning_rate": 4.9942460527623374e-06,
+      "loss": 0.5397,
+      "mean_token_accuracy": 0.8402209639549255,
+      "num_tokens": 7185927.0,
+      "step": 885
+    },
+    {
+      "entropy": 0.6589832663536072,
+      "epoch": 0.5022573363431151,
+      "grad_norm": 1.2508260011672974,
+      "learning_rate": 4.9941808125682336e-06,
+      "loss": 0.5162,
+      "mean_token_accuracy": 0.8469531655311584,
+      "num_tokens": 7226531.0,
+      "step": 890
+    },
+    {
+      "entropy": 0.6467825770378113,
+      "epoch": 0.505079006772009,
+      "grad_norm": 1.2095719575881958,
+      "learning_rate": 4.994115205173317e-06,
+      "loss": 0.5325,
+      "mean_token_accuracy": 0.8432755708694458,
+      "num_tokens": 7267311.0,
+      "step": 895
+    },
+    {
+      "entropy": 0.6670600891113281,
+      "epoch": 0.5079006772009029,
+      "grad_norm": 1.3329170942306519,
+      "learning_rate": 4.994049230590474e-06,
+      "loss": 0.5145,
+      "mean_token_accuracy": 0.8473351120948791,
+      "num_tokens": 7308118.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.654244887828827,
+      "epoch": 0.5107223476297968,
+      "grad_norm": 1.3469221591949463,
+      "learning_rate": 4.993982888832667e-06,
+      "loss": 0.5262,
+      "mean_token_accuracy": 0.8424062848091125,
+      "num_tokens": 7348961.0,
+      "step": 905
+    },
+    {
+      "entropy": 0.6805335998535156,
+      "epoch": 0.5135440180586908,
+      "grad_norm": 1.4335166215896606,
+      "learning_rate": 4.993916179912929e-06,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8342411160469055,
+      "num_tokens": 7389295.0,
+      "step": 910
+    },
+    {
+      "entropy": 0.6766547679901123,
+      "epoch": 0.5163656884875847,
+      "grad_norm": 1.2276146411895752,
+      "learning_rate": 4.993849103844365e-06,
+      "loss": 0.5502,
+      "mean_token_accuracy": 0.8335286498069763,
+      "num_tokens": 7429891.0,
+      "step": 915
+    },
+    {
+      "entropy": 0.6441382646560669,
+      "epoch": 0.5191873589164786,
+      "grad_norm": 1.277137041091919,
+      "learning_rate": 4.9937816606401506e-06,
+      "loss": 0.5377,
+      "mean_token_accuracy": 0.8357362985610962,
+      "num_tokens": 7470461.0,
+      "step": 920
+    },
+    {
+      "entropy": 0.6156215190887451,
+      "epoch": 0.5220090293453724,
+      "grad_norm": 1.3269896507263184,
+      "learning_rate": 4.993713850313537e-06,
+      "loss": 0.4867,
+      "mean_token_accuracy": 0.8524217247962952,
+      "num_tokens": 7511054.0,
+      "step": 925
+    },
+    {
+      "entropy": 0.702682101726532,
+      "epoch": 0.5248306997742663,
+      "grad_norm": 1.4615741968154907,
+      "learning_rate": 4.993645672877843e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8279630184173584,
+      "num_tokens": 7551567.0,
+      "step": 930
+    },
+    {
+      "entropy": 0.6455079674720764,
+      "epoch": 0.5276523702031602,
+      "grad_norm": 1.3673226833343506,
+      "learning_rate": 4.993577128346465e-06,
+      "loss": 0.5157,
+      "mean_token_accuracy": 0.8442542672157287,
+      "num_tokens": 7591635.0,
+      "step": 935
+    },
+    {
+      "entropy": 0.6908176302909851,
+      "epoch": 0.5304740406320542,
+      "grad_norm": 1.4456593990325928,
+      "learning_rate": 4.993508216732867e-06,
+      "loss": 0.558,
+      "mean_token_accuracy": 0.832518482208252,
+      "num_tokens": 7631859.0,
+      "step": 940
+    },
+    {
+      "entropy": 0.7004367709159851,
+      "epoch": 0.5332957110609481,
+      "grad_norm": 1.3898345232009888,
+      "learning_rate": 4.993438938050587e-06,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8318912148475647,
+      "num_tokens": 7672520.0,
+      "step": 945
+    },
+    {
+      "entropy": 0.6737080454826355,
+      "epoch": 0.536117381489842,
+      "grad_norm": 1.398849606513977,
+      "learning_rate": 4.993369292313235e-06,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8320559740066529,
+      "num_tokens": 7713151.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6230229139328003,
+      "epoch": 0.5389390519187359,
+      "grad_norm": 1.264159917831421,
+      "learning_rate": 4.993299279534492e-06,
+      "loss": 0.4986,
+      "mean_token_accuracy": 0.8493775367736817,
+      "num_tokens": 7754018.0,
+      "step": 955
+    },
+    {
+      "entropy": 0.6292937159538269,
+      "epoch": 0.5417607223476298,
+      "grad_norm": 1.344561219215393,
+      "learning_rate": 4.993228899728113e-06,
+      "loss": 0.5148,
+      "mean_token_accuracy": 0.8411380529403687,
+      "num_tokens": 7794654.0,
+      "step": 960
+    },
+    {
+      "entropy": 0.6116863548755646,
+      "epoch": 0.5445823927765236,
+      "grad_norm": 1.1279850006103516,
+      "learning_rate": 4.993158152907923e-06,
+      "loss": 0.4921,
+      "mean_token_accuracy": 0.8529308915138245,
+      "num_tokens": 7835397.0,
+      "step": 965
+    },
+    {
+      "entropy": 0.6285845994949341,
+      "epoch": 0.5474040632054176,
+      "grad_norm": 1.1283153295516968,
+      "learning_rate": 4.993087039087823e-06,
+      "loss": 0.4815,
+      "mean_token_accuracy": 0.8528955459594727,
+      "num_tokens": 7876148.0,
+      "step": 970
+    },
+    {
+      "entropy": 0.6560633659362793,
+      "epoch": 0.5502257336343115,
+      "grad_norm": 1.213463306427002,
+      "learning_rate": 4.993015558281779e-06,
+      "loss": 0.5345,
+      "mean_token_accuracy": 0.8412879347801209,
+      "num_tokens": 7916823.0,
+      "step": 975
+    },
+    {
+      "entropy": 0.6548992276191712,
+      "epoch": 0.5530474040632054,
+      "grad_norm": 1.2896618843078613,
+      "learning_rate": 4.992943710503838e-06,
+      "loss": 0.5147,
+      "mean_token_accuracy": 0.8419336557388306,
+      "num_tokens": 7957342.0,
+      "step": 980
+    },
+    {
+      "entropy": 0.6178439378738403,
+      "epoch": 0.5558690744920993,
+      "grad_norm": 1.1932618618011475,
+      "learning_rate": 4.99287149576811e-06,
+      "loss": 0.4988,
+      "mean_token_accuracy": 0.850293779373169,
+      "num_tokens": 7998078.0,
+      "step": 985
+    },
+    {
+      "entropy": 0.7176673412322998,
+      "epoch": 0.5586907449209932,
+      "grad_norm": 1.495336651802063,
+      "learning_rate": 4.992798914088786e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8254365086555481,
+      "num_tokens": 8038148.0,
+      "step": 990
+    },
+    {
+      "entropy": 0.6931470155715942,
+      "epoch": 0.5615124153498872,
+      "grad_norm": 1.343912124633789,
+      "learning_rate": 4.992725965480121e-06,
+      "loss": 0.5707,
+      "mean_token_accuracy": 0.8297663688659668,
+      "num_tokens": 8078718.0,
+      "step": 995
+    },
+    {
+      "entropy": 0.7093847036361695,
+      "epoch": 0.5643340857787811,
+      "grad_norm": 1.2880916595458984,
+      "learning_rate": 4.992652649956448e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8289087891578675,
+      "num_tokens": 8119465.0,
+      "step": 1000
+    },
+    {
+      "epoch": 0.5643340857787811,
+      "eval_entropy": 0.6200407147407532,
+      "eval_loss": 0.49860483407974243,
+      "eval_mean_token_accuracy": 0.8631659746170044,
+      "eval_num_tokens": 8119465.0,
+      "eval_runtime": 0.1639,
+      "eval_samples_per_second": 24.405,
+      "eval_steps_per_second": 6.101,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6551719307899475,
+      "epoch": 0.5671557562076749,
+      "grad_norm": 1.444422721862793,
+      "learning_rate": 4.992578967532169e-06,
+      "loss": 0.5098,
+      "mean_token_accuracy": 0.8438852429389954,
+      "num_tokens": 8160136.0,
+      "step": 1005
+    },
+    {
+      "entropy": 0.6948068380355835,
+      "epoch": 0.5699774266365688,
+      "grad_norm": 1.4045120477676392,
+      "learning_rate": 4.992504918221759e-06,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8269060254096985,
+      "num_tokens": 8200865.0,
+      "step": 1010
+    },
+    {
+      "entropy": 0.6053695321083069,
+      "epoch": 0.5727990970654627,
+      "grad_norm": 1.2733221054077148,
+      "learning_rate": 4.9924305020397645e-06,
+      "loss": 0.4762,
+      "mean_token_accuracy": 0.8552789926528931,
+      "num_tokens": 8241635.0,
+      "step": 1015
+    },
+    {
+      "entropy": 0.6861397624015808,
+      "epoch": 0.5756207674943566,
+      "grad_norm": 1.3532050848007202,
+      "learning_rate": 4.992355719000805e-06,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8298478722572327,
+      "num_tokens": 8282118.0,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6331191897392273,
+      "epoch": 0.5784424379232506,
+      "grad_norm": 1.2071319818496704,
+      "learning_rate": 4.992280569119574e-06,
+      "loss": 0.5005,
+      "mean_token_accuracy": 0.850217878818512,
+      "num_tokens": 8322734.0,
+      "step": 1025
+    },
+    {
+      "entropy": 0.6948285818099975,
+      "epoch": 0.5812641083521445,
+      "grad_norm": 1.375131368637085,
+      "learning_rate": 4.99220505241083e-06,
+      "loss": 0.5409,
+      "mean_token_accuracy": 0.8394317030906677,
+      "num_tokens": 8363311.0,
+      "step": 1030
+    },
+    {
+      "entropy": 0.6704018592834473,
+      "epoch": 0.5840857787810384,
+      "grad_norm": 1.373281478881836,
+      "learning_rate": 4.992129168889412e-06,
+      "loss": 0.5439,
+      "mean_token_accuracy": 0.8385064721107482,
+      "num_tokens": 8403747.0,
+      "step": 1035
+    },
+    {
+      "entropy": 0.6323675155639649,
+      "epoch": 0.5869074492099323,
+      "grad_norm": 1.2113394737243652,
+      "learning_rate": 4.992052918570226e-06,
+      "loss": 0.495,
+      "mean_token_accuracy": 0.8503919124603272,
+      "num_tokens": 8444647.0,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6883712530136108,
+      "epoch": 0.5897291196388262,
+      "grad_norm": 1.329156517982483,
+      "learning_rate": 4.991976301468251e-06,
+      "loss": 0.5507,
+      "mean_token_accuracy": 0.8347142577171326,
+      "num_tokens": 8485437.0,
+      "step": 1045
+    },
+    {
+      "entropy": 0.6799022316932678,
+      "epoch": 0.59255079006772,
+      "grad_norm": 1.2880463600158691,
+      "learning_rate": 4.9918993175985384e-06,
+      "loss": 0.5525,
+      "mean_token_accuracy": 0.8357504963874817,
+      "num_tokens": 8526376.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6652753591537476,
+      "epoch": 0.595372460496614,
+      "grad_norm": 1.3654208183288574,
+      "learning_rate": 4.991821966976213e-06,
+      "loss": 0.5473,
+      "mean_token_accuracy": 0.8353073954582214,
+      "num_tokens": 8567085.0,
+      "step": 1055
+    },
+    {
+      "entropy": 0.6146788358688354,
+      "epoch": 0.5981941309255079,
+      "grad_norm": 1.140661597251892,
+      "learning_rate": 4.991744249616469e-06,
+      "loss": 0.4757,
+      "mean_token_accuracy": 0.8558377385139465,
+      "num_tokens": 8607488.0,
+      "step": 1060
+    },
+    {
+      "entropy": 0.710556972026825,
+      "epoch": 0.6010158013544018,
+      "grad_norm": 1.3625504970550537,
+      "learning_rate": 4.991666165534575e-06,
+      "loss": 0.5542,
+      "mean_token_accuracy": 0.8333917140960694,
+      "num_tokens": 8648210.0,
+      "step": 1065
+    },
+    {
+      "entropy": 0.6810308933258057,
+      "epoch": 0.6038374717832957,
+      "grad_norm": 1.393760323524475,
+      "learning_rate": 4.99158771474587e-06,
+      "loss": 0.5602,
+      "mean_token_accuracy": 0.8344960331916809,
+      "num_tokens": 8688725.0,
+      "step": 1070
+    },
+    {
+      "entropy": 0.6881044745445252,
+      "epoch": 0.6066591422121896,
+      "grad_norm": 1.4435288906097412,
+      "learning_rate": 4.991508897265766e-06,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8323047876358032,
+      "num_tokens": 8729344.0,
+      "step": 1075
+    },
+    {
+      "entropy": 0.6240251541137696,
+      "epoch": 0.6094808126410836,
+      "grad_norm": 1.3459124565124512,
+      "learning_rate": 4.991429713109746e-06,
+      "loss": 0.4968,
+      "mean_token_accuracy": 0.8495769262313843,
+      "num_tokens": 8770034.0,
+      "step": 1080
+    },
+    {
+      "entropy": 0.642047894001007,
+      "epoch": 0.6123024830699775,
+      "grad_norm": 1.186833143234253,
+      "learning_rate": 4.991350162293367e-06,
+      "loss": 0.5215,
+      "mean_token_accuracy": 0.8443755626678466,
+      "num_tokens": 8810541.0,
+      "step": 1085
+    },
+    {
+      "entropy": 0.6657473087310791,
+      "epoch": 0.6151241534988713,
+      "grad_norm": 1.4160339832305908,
+      "learning_rate": 4.991270244832256e-06,
+      "loss": 0.5373,
+      "mean_token_accuracy": 0.836929440498352,
+      "num_tokens": 8851054.0,
+      "step": 1090
+    },
+    {
+      "entropy": 0.664323914051056,
+      "epoch": 0.6179458239277652,
+      "grad_norm": 1.3085633516311646,
+      "learning_rate": 4.9911899607421116e-06,
+      "loss": 0.5336,
+      "mean_token_accuracy": 0.8405984282493592,
+      "num_tokens": 8891919.0,
+      "step": 1095
+    },
+    {
+      "entropy": 0.6161608934402466,
+      "epoch": 0.6207674943566591,
+      "grad_norm": 1.2209677696228027,
+      "learning_rate": 4.991109310038707e-06,
+      "loss": 0.4854,
+      "mean_token_accuracy": 0.8494919776916504,
+      "num_tokens": 8932520.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.633518648147583,
+      "epoch": 0.623589164785553,
+      "grad_norm": 1.422237515449524,
+      "learning_rate": 4.991028292737887e-06,
+      "loss": 0.5238,
+      "mean_token_accuracy": 0.8427853345870971,
+      "num_tokens": 8973154.0,
+      "step": 1105
+    },
+    {
+      "entropy": 0.7281236052513123,
+      "epoch": 0.626410835214447,
+      "grad_norm": 1.3925809860229492,
+      "learning_rate": 4.990946908855565e-06,
+      "loss": 0.6025,
+      "mean_token_accuracy": 0.8230817198753357,
+      "num_tokens": 9013989.0,
+      "step": 1110
+    },
+    {
+      "entropy": 0.6049828886985779,
+      "epoch": 0.6292325056433409,
+      "grad_norm": 1.2714378833770752,
+      "learning_rate": 4.990865158407731e-06,
+      "loss": 0.4823,
+      "mean_token_accuracy": 0.8520822525024414,
+      "num_tokens": 9054405.0,
+      "step": 1115
+    },
+    {
+      "entropy": 0.6474563360214234,
+      "epoch": 0.6320541760722348,
+      "grad_norm": 1.3699289560317993,
+      "learning_rate": 4.990783041410444e-06,
+      "loss": 0.5001,
+      "mean_token_accuracy": 0.8497814536094666,
+      "num_tokens": 9095024.0,
+      "step": 1120
+    },
+    {
+      "entropy": 0.665639317035675,
+      "epoch": 0.6348758465011287,
+      "grad_norm": 1.32370126247406,
+      "learning_rate": 4.9907005578798366e-06,
+      "loss": 0.5265,
+      "mean_token_accuracy": 0.8440038084983825,
+      "num_tokens": 9135786.0,
+      "step": 1125
+    },
+    {
+      "entropy": 0.6320675134658813,
+      "epoch": 0.6376975169300225,
+      "grad_norm": 1.2937510013580322,
+      "learning_rate": 4.990617707832111e-06,
+      "loss": 0.4979,
+      "mean_token_accuracy": 0.8501017332077027,
+      "num_tokens": 9176559.0,
+      "step": 1130
+    },
+    {
+      "entropy": 0.7049695611000061,
+      "epoch": 0.6405191873589164,
+      "grad_norm": 1.3041940927505493,
+      "learning_rate": 4.990534491283545e-06,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8350239872932435,
+      "num_tokens": 9217381.0,
+      "step": 1135
+    },
+    {
+      "entropy": 0.6608424663543702,
+      "epoch": 0.6433408577878104,
+      "grad_norm": 1.3915507793426514,
+      "learning_rate": 4.990450908250485e-06,
+      "loss": 0.5267,
+      "mean_token_accuracy": 0.8423561692237854,
+      "num_tokens": 9258041.0,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6109439551830291,
+      "epoch": 0.6461625282167043,
+      "grad_norm": 1.1623317003250122,
+      "learning_rate": 4.990366958749352e-06,
+      "loss": 0.4937,
+      "mean_token_accuracy": 0.8506023287773132,
+      "num_tokens": 9298558.0,
+      "step": 1145
+    },
+    {
+      "entropy": 0.6726685404777527,
+      "epoch": 0.6489841986455982,
+      "grad_norm": 1.4480568170547485,
+      "learning_rate": 4.990282642796638e-06,
+      "loss": 0.559,
+      "mean_token_accuracy": 0.8336881279945374,
+      "num_tokens": 9339287.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6401282668113708,
+      "epoch": 0.6518058690744921,
+      "grad_norm": 1.405791997909546,
+      "learning_rate": 4.9901979604089055e-06,
+      "loss": 0.5005,
+      "mean_token_accuracy": 0.8454666852951049,
+      "num_tokens": 9379965.0,
+      "step": 1155
+    },
+    {
+      "entropy": 0.616789698600769,
+      "epoch": 0.654627539503386,
+      "grad_norm": 1.2432202100753784,
+      "learning_rate": 4.990112911602792e-06,
+      "loss": 0.4809,
+      "mean_token_accuracy": 0.8562314510345459,
+      "num_tokens": 9420544.0,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6499378085136414,
+      "epoch": 0.65744920993228,
+      "grad_norm": 1.3318426609039307,
+      "learning_rate": 4.990027496395003e-06,
+      "loss": 0.5165,
+      "mean_token_accuracy": 0.8456689953804016,
+      "num_tokens": 9461170.0,
+      "step": 1165
+    },
+    {
+      "entropy": 0.6929373741149902,
+      "epoch": 0.6602708803611738,
+      "grad_norm": 1.501207709312439,
+      "learning_rate": 4.989941714802321e-06,
+      "loss": 0.5457,
+      "mean_token_accuracy": 0.8362460851669311,
+      "num_tokens": 9501863.0,
+      "step": 1170
+    },
+    {
+      "entropy": 0.6151013791561126,
+      "epoch": 0.6630925507900677,
+      "grad_norm": 1.291373372077942,
+      "learning_rate": 4.989855566841597e-06,
+      "loss": 0.4929,
+      "mean_token_accuracy": 0.8491938352584839,
+      "num_tokens": 9542539.0,
+      "step": 1175
+    },
+    {
+      "entropy": 0.6619673609733582,
+      "epoch": 0.6659142212189616,
+      "grad_norm": 1.8319789171218872,
+      "learning_rate": 4.989769052529754e-06,
+      "loss": 0.5406,
+      "mean_token_accuracy": 0.8366817712783814,
+      "num_tokens": 9583209.0,
+      "step": 1180
+    },
+    {
+      "entropy": 0.7023379445075989,
+      "epoch": 0.6687358916478555,
+      "grad_norm": 1.4245003461837769,
+      "learning_rate": 4.989682171883789e-06,
+      "loss": 0.5833,
+      "mean_token_accuracy": 0.8261257648468018,
+      "num_tokens": 9623904.0,
+      "step": 1185
+    },
+    {
+      "entropy": 0.6219159960746765,
+      "epoch": 0.6715575620767494,
+      "grad_norm": 1.5310255289077759,
+      "learning_rate": 4.9895949249207674e-06,
+      "loss": 0.4825,
+      "mean_token_accuracy": 0.8541441679000854,
+      "num_tokens": 9664582.0,
+      "step": 1190
+    },
+    {
+      "entropy": 0.6346824288368225,
+      "epoch": 0.6743792325056434,
+      "grad_norm": 1.3092063665390015,
+      "learning_rate": 4.989507311657832e-06,
+      "loss": 0.5112,
+      "mean_token_accuracy": 0.8455924034118653,
+      "num_tokens": 9705103.0,
+      "step": 1195
+    },
+    {
+      "entropy": 0.6392194151878356,
+      "epoch": 0.6772009029345373,
+      "grad_norm": 1.2018376588821411,
+      "learning_rate": 4.9894193321121915e-06,
+      "loss": 0.5101,
+      "mean_token_accuracy": 0.847294807434082,
+      "num_tokens": 9745685.0,
+      "step": 1200
+    },
+    {
+      "entropy": 0.7368967413902283,
+      "epoch": 0.6800225733634312,
+      "grad_norm": 1.4050233364105225,
+      "learning_rate": 4.989330986301131e-06,
+      "loss": 0.6216,
+      "mean_token_accuracy": 0.8151282906532288,
+      "num_tokens": 9786144.0,
+      "step": 1205
+    },
+    {
+      "entropy": 0.6174101948738098,
+      "epoch": 0.6828442437923251,
+      "grad_norm": 1.3221447467803955,
+      "learning_rate": 4.989242274242007e-06,
+      "loss": 0.4908,
+      "mean_token_accuracy": 0.8495320796966552,
+      "num_tokens": 9826811.0,
+      "step": 1210
+    },
+    {
+      "entropy": 0.6422483444213867,
+      "epoch": 0.6856659142212189,
+      "grad_norm": 1.1776975393295288,
+      "learning_rate": 4.989153195952246e-06,
+      "loss": 0.5173,
+      "mean_token_accuracy": 0.8441770195960998,
+      "num_tokens": 9867430.0,
+      "step": 1215
+    },
+    {
+      "entropy": 0.6633909225463868,
+      "epoch": 0.6884875846501128,
+      "grad_norm": 1.257130742073059,
+      "learning_rate": 4.989063751449346e-06,
+      "loss": 0.5406,
+      "mean_token_accuracy": 0.8380695223808289,
+      "num_tokens": 9907541.0,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6803221344947815,
+      "epoch": 0.6913092550790068,
+      "grad_norm": 1.227628231048584,
+      "learning_rate": 4.98897394075088e-06,
+      "loss": 0.5413,
+      "mean_token_accuracy": 0.8371790289878845,
+      "num_tokens": 9948303.0,
+      "step": 1225
+    },
+    {
+      "entropy": 0.6370683073997497,
+      "epoch": 0.6941309255079007,
+      "grad_norm": 1.1942147016525269,
+      "learning_rate": 4.9888837638744915e-06,
+      "loss": 0.5116,
+      "mean_token_accuracy": 0.8441259145736695,
+      "num_tokens": 9989051.0,
+      "step": 1230
+    },
+    {
+      "entropy": 0.6529434204101563,
+      "epoch": 0.6969525959367946,
+      "grad_norm": 1.3808612823486328,
+      "learning_rate": 4.988793220837895e-06,
+      "loss": 0.5046,
+      "mean_token_accuracy": 0.8458134651184082,
+      "num_tokens": 10029798.0,
+      "step": 1235
+    },
+    {
+      "entropy": 0.6289508223533631,
+      "epoch": 0.6997742663656885,
+      "grad_norm": 1.1567049026489258,
+      "learning_rate": 4.988702311658879e-06,
+      "loss": 0.5103,
+      "mean_token_accuracy": 0.8443259239196778,
+      "num_tokens": 10070451.0,
+      "step": 1240
+    },
+    {
+      "entropy": 0.6437416434288025,
+      "epoch": 0.7025959367945824,
+      "grad_norm": 1.3937760591506958,
+      "learning_rate": 4.9886110363553005e-06,
+      "loss": 0.5103,
+      "mean_token_accuracy": 0.846991765499115,
+      "num_tokens": 10111198.0,
+      "step": 1245
+    },
+    {
+      "entropy": 0.710420835018158,
+      "epoch": 0.7054176072234764,
+      "grad_norm": 1.3196383714675903,
+      "learning_rate": 4.988519394945092e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8278885602951049,
+      "num_tokens": 10151900.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.6752119541168213,
+      "epoch": 0.7082392776523702,
+      "grad_norm": 1.4347295761108398,
+      "learning_rate": 4.988427387446255e-06,
+      "loss": 0.5473,
+      "mean_token_accuracy": 0.8365179777145386,
+      "num_tokens": 10192756.0,
+      "step": 1255
+    },
+    {
+      "entropy": 0.6734273076057434,
+      "epoch": 0.7110609480812641,
+      "grad_norm": 1.4099767208099365,
+      "learning_rate": 4.988335013876867e-06,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8306289792060852,
+      "num_tokens": 10233478.0,
+      "step": 1260
+    },
+    {
+      "entropy": 0.7350514650344848,
+      "epoch": 0.713882618510158,
+      "grad_norm": 1.3657110929489136,
+      "learning_rate": 4.988242274255073e-06,
+      "loss": 0.6019,
+      "mean_token_accuracy": 0.8200275421142578,
+      "num_tokens": 10274135.0,
+      "step": 1265
+    },
+    {
+      "entropy": 0.6617794394493103,
+      "epoch": 0.7167042889390519,
+      "grad_norm": 1.4455223083496094,
+      "learning_rate": 4.988149168599092e-06,
+      "loss": 0.5323,
+      "mean_token_accuracy": 0.8396545290946961,
+      "num_tokens": 10314720.0,
+      "step": 1270
+    },
+    {
+      "entropy": 0.6266619205474854,
+      "epoch": 0.7195259593679458,
+      "grad_norm": 1.2425928115844727,
+      "learning_rate": 4.988055696927214e-06,
+      "loss": 0.5033,
+      "mean_token_accuracy": 0.8460469722747803,
+      "num_tokens": 10355383.0,
+      "step": 1275
+    },
+    {
+      "entropy": 0.6324691534042358,
+      "epoch": 0.7223476297968398,
+      "grad_norm": 1.2948311567306519,
+      "learning_rate": 4.987961859257803e-06,
+      "loss": 0.5149,
+      "mean_token_accuracy": 0.8449645042419434,
+      "num_tokens": 10395935.0,
+      "step": 1280
+    },
+    {
+      "entropy": 0.6536452293395996,
+      "epoch": 0.7251693002257337,
+      "grad_norm": 1.3648988008499146,
+      "learning_rate": 4.987867655609292e-06,
+      "loss": 0.5296,
+      "mean_token_accuracy": 0.8420405626296997,
+      "num_tokens": 10436654.0,
+      "step": 1285
+    },
+    {
+      "entropy": 0.6121232390403748,
+      "epoch": 0.7279909706546276,
+      "grad_norm": 1.2139304876327515,
+      "learning_rate": 4.987773086000188e-06,
+      "loss": 0.4986,
+      "mean_token_accuracy": 0.8504088521003723,
+      "num_tokens": 10477501.0,
+      "step": 1290
+    },
+    {
+      "entropy": 0.6448933601379394,
+      "epoch": 0.7308126410835214,
+      "grad_norm": 1.375567078590393,
+      "learning_rate": 4.987678150449069e-06,
+      "loss": 0.5122,
+      "mean_token_accuracy": 0.8450644254684448,
+      "num_tokens": 10518384.0,
+      "step": 1295
+    },
+    {
+      "entropy": 0.6476213932037354,
+      "epoch": 0.7336343115124153,
+      "grad_norm": 1.3577604293823242,
+      "learning_rate": 4.987582848974586e-06,
+      "loss": 0.5053,
+      "mean_token_accuracy": 0.8453767418861389,
+      "num_tokens": 10559194.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.6976300120353699,
+      "epoch": 0.7364559819413092,
+      "grad_norm": 1.3497973680496216,
+      "learning_rate": 4.987487181595459e-06,
+      "loss": 0.5519,
+      "mean_token_accuracy": 0.8350682735443116,
+      "num_tokens": 10599795.0,
+      "step": 1305
+    },
+    {
+      "entropy": 0.6458664059638977,
+      "epoch": 0.7392776523702032,
+      "grad_norm": 1.329574465751648,
+      "learning_rate": 4.987391148330485e-06,
+      "loss": 0.5169,
+      "mean_token_accuracy": 0.8440695524215698,
+      "num_tokens": 10640689.0,
+      "step": 1310
+    },
+    {
+      "entropy": 0.6906252384185791,
+      "epoch": 0.7420993227990971,
+      "grad_norm": 1.2950578927993774,
+      "learning_rate": 4.987294749198526e-06,
+      "loss": 0.5638,
+      "mean_token_accuracy": 0.8313373565673828,
+      "num_tokens": 10681276.0,
+      "step": 1315
+    },
+    {
+      "entropy": 0.6506387591362,
+      "epoch": 0.744920993227991,
+      "grad_norm": 1.5627875328063965,
+      "learning_rate": 4.987197984218522e-06,
+      "loss": 0.491,
+      "mean_token_accuracy": 0.8504640936851502,
+      "num_tokens": 10722061.0,
+      "step": 1320
+    },
+    {
+      "entropy": 0.6033547759056092,
+      "epoch": 0.7477426636568849,
+      "grad_norm": 1.4060436487197876,
+      "learning_rate": 4.9871008534094825e-06,
+      "loss": 0.4864,
+      "mean_token_accuracy": 0.8535768985748291,
+      "num_tokens": 10762663.0,
+      "step": 1325
+    },
+    {
+      "entropy": 0.6256420493125916,
+      "epoch": 0.7505643340857788,
+      "grad_norm": 1.2653220891952515,
+      "learning_rate": 4.987003356790487e-06,
+      "loss": 0.5119,
+      "mean_token_accuracy": 0.845154881477356,
+      "num_tokens": 10803082.0,
+      "step": 1330
+    },
+    {
+      "entropy": 0.627797293663025,
+      "epoch": 0.7533860045146726,
+      "grad_norm": 1.2059917449951172,
+      "learning_rate": 4.986905494380691e-06,
+      "loss": 0.483,
+      "mean_token_accuracy": 0.8541393876075745,
+      "num_tokens": 10843614.0,
+      "step": 1335
+    },
+    {
+      "entropy": 0.6704052805900573,
+      "epoch": 0.7562076749435666,
+      "grad_norm": 1.4258085489273071,
+      "learning_rate": 4.986807266199318e-06,
+      "loss": 0.5272,
+      "mean_token_accuracy": 0.8418290138244628,
+      "num_tokens": 10884087.0,
+      "step": 1340
+    },
+    {
+      "entropy": 0.7184508085250855,
+      "epoch": 0.7590293453724605,
+      "grad_norm": 1.3789485692977905,
+      "learning_rate": 4.986708672265667e-06,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8304919958114624,
+      "num_tokens": 10924781.0,
+      "step": 1345
+    },
+    {
+      "entropy": 0.6704004406929016,
+      "epoch": 0.7618510158013544,
+      "grad_norm": 1.290982961654663,
+      "learning_rate": 4.986609712599103e-06,
+      "loss": 0.5249,
+      "mean_token_accuracy": 0.8440407991409302,
+      "num_tokens": 10965534.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.6607297778129577,
+      "epoch": 0.7646726862302483,
+      "grad_norm": 1.294542670249939,
+      "learning_rate": 4.986510387219071e-06,
+      "loss": 0.5247,
+      "mean_token_accuracy": 0.8411241173744202,
+      "num_tokens": 11006238.0,
+      "step": 1355
+    },
+    {
+      "entropy": 0.6355977773666381,
+      "epoch": 0.7674943566591422,
+      "grad_norm": 1.3880748748779297,
+      "learning_rate": 4.98641069614508e-06,
+      "loss": 0.5017,
+      "mean_token_accuracy": 0.8476488828659058,
+      "num_tokens": 11046967.0,
+      "step": 1360
+    },
+    {
+      "entropy": 0.6151192545890808,
+      "epoch": 0.7703160270880361,
+      "grad_norm": 1.2665696144104004,
+      "learning_rate": 4.9863106393967165e-06,
+      "loss": 0.4792,
+      "mean_token_accuracy": 0.854477858543396,
+      "num_tokens": 11086831.0,
+      "step": 1365
+    },
+    {
+      "entropy": 0.6570278882980347,
+      "epoch": 0.7731376975169301,
+      "grad_norm": 1.4179819822311401,
+      "learning_rate": 4.986210216993636e-06,
+      "loss": 0.5266,
+      "mean_token_accuracy": 0.8439712643623352,
+      "num_tokens": 11127372.0,
+      "step": 1370
+    },
+    {
+      "entropy": 0.6928212285041809,
+      "epoch": 0.7759593679458239,
+      "grad_norm": 1.3699414730072021,
+      "learning_rate": 4.986109428955566e-06,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8342428803443909,
+      "num_tokens": 11167988.0,
+      "step": 1375
+    },
+    {
+      "entropy": 0.6448866724967957,
+      "epoch": 0.7787810383747178,
+      "grad_norm": 1.2906067371368408,
+      "learning_rate": 4.986008275302307e-06,
+      "loss": 0.5244,
+      "mean_token_accuracy": 0.8406425952911377,
+      "num_tokens": 11208797.0,
+      "step": 1380
+    },
+    {
+      "entropy": 0.6187806010246277,
+      "epoch": 0.7816027088036117,
+      "grad_norm": 1.25465726852417,
+      "learning_rate": 4.98590675605373e-06,
+      "loss": 0.4718,
+      "mean_token_accuracy": 0.8553349137306213,
+      "num_tokens": 11249307.0,
+      "step": 1385
+    },
+    {
+      "entropy": 0.6682080864906311,
+      "epoch": 0.7844243792325056,
+      "grad_norm": 1.278944969177246,
+      "learning_rate": 4.98580487122978e-06,
+      "loss": 0.5124,
+      "mean_token_accuracy": 0.8448834180831909,
+      "num_tokens": 11289926.0,
+      "step": 1390
+    },
+    {
+      "entropy": 0.6455815672874451,
+      "epoch": 0.7872460496613995,
+      "grad_norm": 1.4049651622772217,
+      "learning_rate": 4.98570262085047e-06,
+      "loss": 0.5146,
+      "mean_token_accuracy": 0.8430341005325317,
+      "num_tokens": 11330704.0,
+      "step": 1395
+    },
+    {
+      "entropy": 0.7064637899398803,
+      "epoch": 0.7900677200902935,
+      "grad_norm": 1.3357079029083252,
+      "learning_rate": 4.985600004935889e-06,
+      "loss": 0.565,
+      "mean_token_accuracy": 0.8299904108047486,
+      "num_tokens": 11371389.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.6495133399963379,
+      "epoch": 0.7928893905191874,
+      "grad_norm": 1.6325474977493286,
+      "learning_rate": 4.985497023506195e-06,
+      "loss": 0.5448,
+      "mean_token_accuracy": 0.8373395681381226,
+      "num_tokens": 11411887.0,
+      "step": 1405
+    },
+    {
+      "entropy": 0.6633385181427002,
+      "epoch": 0.7957110609480813,
+      "grad_norm": 1.4035420417785645,
+      "learning_rate": 4.985393676581619e-06,
+      "loss": 0.5252,
+      "mean_token_accuracy": 0.8406889915466309,
+      "num_tokens": 11452589.0,
+      "step": 1410
+    },
+    {
+      "entropy": 0.6448916912078857,
+      "epoch": 0.7985327313769752,
+      "grad_norm": 1.288009524345398,
+      "learning_rate": 4.985289964182463e-06,
+      "loss": 0.5105,
+      "mean_token_accuracy": 0.8462629556655884,
+      "num_tokens": 11493085.0,
+      "step": 1415
+    },
+    {
+      "entropy": 0.5965805172920227,
+      "epoch": 0.801354401805869,
+      "grad_norm": 1.2756577730178833,
+      "learning_rate": 4.985185886329101e-06,
+      "loss": 0.4867,
+      "mean_token_accuracy": 0.8524753332138062,
+      "num_tokens": 11533760.0,
+      "step": 1420
+    },
+    {
+      "entropy": 0.6665675163269043,
+      "epoch": 0.804176072234763,
+      "grad_norm": 1.3834420442581177,
+      "learning_rate": 4.985081443041981e-06,
+      "loss": 0.5285,
+      "mean_token_accuracy": 0.8393696427345276,
+      "num_tokens": 11574224.0,
+      "step": 1425
+    },
+    {
+      "entropy": 0.6934167623519898,
+      "epoch": 0.8069977426636569,
+      "grad_norm": 1.4740710258483887,
+      "learning_rate": 4.98497663434162e-06,
+      "loss": 0.5623,
+      "mean_token_accuracy": 0.830849575996399,
+      "num_tokens": 11614729.0,
+      "step": 1430
+    },
+    {
+      "entropy": 0.6643628239631653,
+      "epoch": 0.8098194130925508,
+      "grad_norm": 1.342612862586975,
+      "learning_rate": 4.984871460248607e-06,
+      "loss": 0.5361,
+      "mean_token_accuracy": 0.840022873878479,
+      "num_tokens": 11655327.0,
+      "step": 1435
+    },
+    {
+      "entropy": 0.6764454126358033,
+      "epoch": 0.8126410835214447,
+      "grad_norm": 1.5213992595672607,
+      "learning_rate": 4.984765920783604e-06,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8351798415184021,
+      "num_tokens": 11695967.0,
+      "step": 1440
+    },
+    {
+      "entropy": 0.6788670063018799,
+      "epoch": 0.8154627539503386,
+      "grad_norm": 1.2199022769927979,
+      "learning_rate": 4.984660015967343e-06,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8351686835289002,
+      "num_tokens": 11736108.0,
+      "step": 1445
+    },
+    {
+      "entropy": 0.6327372908592224,
+      "epoch": 0.8182844243792325,
+      "grad_norm": 1.4303865432739258,
+      "learning_rate": 4.984553745820631e-06,
+      "loss": 0.505,
+      "mean_token_accuracy": 0.8465925574302673,
+      "num_tokens": 11776854.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.6577797293663025,
+      "epoch": 0.8211060948081265,
+      "grad_norm": 1.3754053115844727,
+      "learning_rate": 4.984447110364343e-06,
+      "loss": 0.526,
+      "mean_token_accuracy": 0.8419236660003662,
+      "num_tokens": 11817612.0,
+      "step": 1455
+    },
+    {
+      "entropy": 0.6646520733833313,
+      "epoch": 0.8239277652370203,
+      "grad_norm": 1.260542631149292,
+      "learning_rate": 4.98434010961943e-06,
+      "loss": 0.533,
+      "mean_token_accuracy": 0.8385473966598511,
+      "num_tokens": 11858437.0,
+      "step": 1460
+    },
+    {
+      "entropy": 0.6954992532730102,
+      "epoch": 0.8267494356659142,
+      "grad_norm": 1.408316731452942,
+      "learning_rate": 4.9842327436069105e-06,
+      "loss": 0.5601,
+      "mean_token_accuracy": 0.8325203895568848,
+      "num_tokens": 11898873.0,
+      "step": 1465
+    },
+    {
+      "entropy": 0.6532683134078979,
+      "epoch": 0.8295711060948081,
+      "grad_norm": 1.1917073726654053,
+      "learning_rate": 4.984125012347876e-06,
+      "loss": 0.5346,
+      "mean_token_accuracy": 0.8390217423439026,
+      "num_tokens": 11939340.0,
+      "step": 1470
+    },
+    {
+      "entropy": 0.6418487191200256,
+      "epoch": 0.832392776523702,
+      "grad_norm": 1.243117332458496,
+      "learning_rate": 4.984016915863491e-06,
+      "loss": 0.4966,
+      "mean_token_accuracy": 0.8492981314659118,
+      "num_tokens": 11980119.0,
+      "step": 1475
+    },
+    {
+      "entropy": 0.6358325242996216,
+      "epoch": 0.835214446952596,
+      "grad_norm": 1.252548336982727,
+      "learning_rate": 4.983908454174993e-06,
+      "loss": 0.4919,
+      "mean_token_accuracy": 0.8524085760116578,
+      "num_tokens": 12020810.0,
+      "step": 1480
+    },
+    {
+      "entropy": 0.6454902291297913,
+      "epoch": 0.8380361173814899,
+      "grad_norm": 1.2286616563796997,
+      "learning_rate": 4.983799627303685e-06,
+      "loss": 0.4958,
+      "mean_token_accuracy": 0.8518875241279602,
+      "num_tokens": 12061121.0,
+      "step": 1485
+    },
+    {
+      "entropy": 0.6209899544715881,
+      "epoch": 0.8408577878103838,
+      "grad_norm": 1.2979341745376587,
+      "learning_rate": 4.98369043527095e-06,
+      "loss": 0.5001,
+      "mean_token_accuracy": 0.8472349762916564,
+      "num_tokens": 12101696.0,
+      "step": 1490
+    },
+    {
+      "entropy": 0.6519573330879211,
+      "epoch": 0.8436794582392777,
+      "grad_norm": 1.2703050374984741,
+      "learning_rate": 4.9835808780982375e-06,
+      "loss": 0.5115,
+      "mean_token_accuracy": 0.8439043760299683,
+      "num_tokens": 12142285.0,
+      "step": 1495
+    },
+    {
+      "entropy": 0.6138962626457214,
+      "epoch": 0.8465011286681715,
+      "grad_norm": 1.2608084678649902,
+      "learning_rate": 4.9834709558070695e-06,
+      "loss": 0.4987,
+      "mean_token_accuracy": 0.8513062238693238,
+      "num_tokens": 12183096.0,
+      "step": 1500
+    },
+    {
+      "epoch": 0.8465011286681715,
+      "eval_entropy": 0.6190634965896606,
+      "eval_loss": 0.450931191444397,
+      "eval_mean_token_accuracy": 0.8727481961250305,
+      "eval_num_tokens": 12183096.0,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 24.428,
+      "eval_steps_per_second": 6.107,
+      "step": 1500
+    },
+    {
+      "entropy": 0.6505842447280884,
+      "epoch": 0.8493227990970654,
+      "grad_norm": 1.3207699060440063,
+      "learning_rate": 4.983360668419041e-06,
+      "loss": 0.4894,
+      "mean_token_accuracy": 0.8519291162490845,
+      "num_tokens": 12223690.0,
+      "step": 1505
+    },
+    {
+      "entropy": 0.664458155632019,
+      "epoch": 0.8521444695259593,
+      "grad_norm": 1.2953647375106812,
+      "learning_rate": 4.983250015955818e-06,
+      "loss": 0.5346,
+      "mean_token_accuracy": 0.8392679691314697,
+      "num_tokens": 12264356.0,
+      "step": 1510
+    },
+    {
+      "entropy": 0.6668670177459717,
+      "epoch": 0.8549661399548533,
+      "grad_norm": 1.276229739189148,
+      "learning_rate": 4.983138998439137e-06,
+      "loss": 0.5242,
+      "mean_token_accuracy": 0.8406529307365418,
+      "num_tokens": 12305094.0,
+      "step": 1515
+    },
+    {
+      "entropy": 0.7231775879859924,
+      "epoch": 0.8577878103837472,
+      "grad_norm": 1.3792636394500732,
+      "learning_rate": 4.983027615890809e-06,
+      "loss": 0.5792,
+      "mean_token_accuracy": 0.827623724937439,
+      "num_tokens": 12345899.0,
+      "step": 1520
+    },
+    {
+      "entropy": 0.6197749614715576,
+      "epoch": 0.8606094808126411,
+      "grad_norm": 1.3481459617614746,
+      "learning_rate": 4.982915868332713e-06,
+      "loss": 0.4994,
+      "mean_token_accuracy": 0.8470370888710022,
+      "num_tokens": 12386462.0,
+      "step": 1525
+    },
+    {
+      "entropy": 0.7093599200248718,
+      "epoch": 0.863431151241535,
+      "grad_norm": 1.4356274604797363,
+      "learning_rate": 4.982803755786804e-06,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.825712525844574,
+      "num_tokens": 12427019.0,
+      "step": 1530
+    },
+    {
+      "entropy": 0.7157705545425415,
+      "epoch": 0.8662528216704289,
+      "grad_norm": 1.3810184001922607,
+      "learning_rate": 4.982691278275106e-06,
+      "loss": 0.5696,
+      "mean_token_accuracy": 0.8296974778175354,
+      "num_tokens": 12467761.0,
+      "step": 1535
+    },
+    {
+      "entropy": 0.6354712247848511,
+      "epoch": 0.8690744920993227,
+      "grad_norm": 1.4308044910430908,
+      "learning_rate": 4.982578435819714e-06,
+      "loss": 0.5066,
+      "mean_token_accuracy": 0.8471119284629822,
+      "num_tokens": 12508405.0,
+      "step": 1540
+    },
+    {
+      "entropy": 0.7147751808166504,
+      "epoch": 0.8718961625282167,
+      "grad_norm": 1.399246096611023,
+      "learning_rate": 4.982465228442797e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8266459822654724,
+      "num_tokens": 12549081.0,
+      "step": 1545
+    },
+    {
+      "entropy": 0.643587851524353,
+      "epoch": 0.8747178329571106,
+      "grad_norm": 1.2349791526794434,
+      "learning_rate": 4.982351656166595e-06,
+      "loss": 0.5062,
+      "mean_token_accuracy": 0.8451477646827698,
+      "num_tokens": 12589617.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.6347296714782715,
+      "epoch": 0.8775395033860045,
+      "grad_norm": 1.2148078680038452,
+      "learning_rate": 4.982237719013418e-06,
+      "loss": 0.5073,
+      "mean_token_accuracy": 0.8488546848297119,
+      "num_tokens": 12629403.0,
+      "step": 1555
+    },
+    {
+      "entropy": 0.6973551154136658,
+      "epoch": 0.8803611738148984,
+      "grad_norm": 1.2895756959915161,
+      "learning_rate": 4.982123417005651e-06,
+      "loss": 0.561,
+      "mean_token_accuracy": 0.8330285906791687,
+      "num_tokens": 12670150.0,
+      "step": 1560
+    },
+    {
+      "entropy": 0.7544346213340759,
+      "epoch": 0.8831828442437923,
+      "grad_norm": 1.3327959775924683,
+      "learning_rate": 4.982008750165746e-06,
+      "loss": 0.615,
+      "mean_token_accuracy": 0.818334436416626,
+      "num_tokens": 12710858.0,
+      "step": 1565
+    },
+    {
+      "entropy": 0.7033024072647095,
+      "epoch": 0.8860045146726863,
+      "grad_norm": 1.3447669744491577,
+      "learning_rate": 4.981893718516231e-06,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8303789258003235,
+      "num_tokens": 12751450.0,
+      "step": 1570
+    },
+    {
+      "entropy": 0.6676825523376465,
+      "epoch": 0.8888261851015802,
+      "grad_norm": 1.3791662454605103,
+      "learning_rate": 4.981778322079704e-06,
+      "loss": 0.5194,
+      "mean_token_accuracy": 0.8445156216621399,
+      "num_tokens": 12792048.0,
+      "step": 1575
+    },
+    {
+      "entropy": 0.6464922666549683,
+      "epoch": 0.891647855530474,
+      "grad_norm": 1.2325254678726196,
+      "learning_rate": 4.981662560878835e-06,
+      "loss": 0.5236,
+      "mean_token_accuracy": 0.8373061299324036,
+      "num_tokens": 12832714.0,
+      "step": 1580
+    },
+    {
+      "entropy": 0.6744041681289673,
+      "epoch": 0.8944695259593679,
+      "grad_norm": 1.4298688173294067,
+      "learning_rate": 4.981546434936363e-06,
+      "loss": 0.5531,
+      "mean_token_accuracy": 0.8334176301956177,
+      "num_tokens": 12873204.0,
+      "step": 1585
+    },
+    {
+      "entropy": 0.5900899410247803,
+      "epoch": 0.8972911963882618,
+      "grad_norm": 1.396270513534546,
+      "learning_rate": 4.981429944275103e-06,
+      "loss": 0.4826,
+      "mean_token_accuracy": 0.8552528858184815,
+      "num_tokens": 12912731.0,
+      "step": 1590
+    },
+    {
+      "entropy": 0.6489648938179016,
+      "epoch": 0.9001128668171557,
+      "grad_norm": 1.2255243062973022,
+      "learning_rate": 4.981313088917939e-06,
+      "loss": 0.5231,
+      "mean_token_accuracy": 0.8412474513053894,
+      "num_tokens": 12953377.0,
+      "step": 1595
+    },
+    {
+      "entropy": 0.7049468755722046,
+      "epoch": 0.9029345372460497,
+      "grad_norm": 1.6258138418197632,
+      "learning_rate": 4.9811958688878274e-06,
+      "loss": 0.5564,
+      "mean_token_accuracy": 0.8347102165222168,
+      "num_tokens": 12994028.0,
+      "step": 1600
+    },
+    {
+      "entropy": 0.6801372170448303,
+      "epoch": 0.9057562076749436,
+      "grad_norm": 1.2854009866714478,
+      "learning_rate": 4.981078284207797e-06,
+      "loss": 0.5349,
+      "mean_token_accuracy": 0.8424756646156311,
+      "num_tokens": 13034671.0,
+      "step": 1605
+    },
+    {
+      "entropy": 0.6932690858840942,
+      "epoch": 0.9085778781038375,
+      "grad_norm": 1.4718157052993774,
+      "learning_rate": 4.980960334900945e-06,
+      "loss": 0.5371,
+      "mean_token_accuracy": 0.8399105072021484,
+      "num_tokens": 13075220.0,
+      "step": 1610
+    },
+    {
+      "entropy": 0.6554085493087769,
+      "epoch": 0.9113995485327314,
+      "grad_norm": 1.1461048126220703,
+      "learning_rate": 4.980842020990444e-06,
+      "loss": 0.5251,
+      "mean_token_accuracy": 0.8404485821723938,
+      "num_tokens": 13115849.0,
+      "step": 1615
+    },
+    {
+      "entropy": 0.6477279543876648,
+      "epoch": 0.9142212189616253,
+      "grad_norm": 1.2576565742492676,
+      "learning_rate": 4.980723342499538e-06,
+      "loss": 0.5306,
+      "mean_token_accuracy": 0.8431344985961914,
+      "num_tokens": 13156403.0,
+      "step": 1620
+    },
+    {
+      "entropy": 0.6991305470466613,
+      "epoch": 0.9170428893905191,
+      "grad_norm": 1.3756641149520874,
+      "learning_rate": 4.9806042994515395e-06,
+      "loss": 0.5567,
+      "mean_token_accuracy": 0.8329641699790955,
+      "num_tokens": 13197015.0,
+      "step": 1625
+    },
+    {
+      "entropy": 0.6869478225708008,
+      "epoch": 0.9198645598194131,
+      "grad_norm": 1.265496015548706,
+      "learning_rate": 4.980484891869835e-06,
+      "loss": 0.5361,
+      "mean_token_accuracy": 0.8395046353340149,
+      "num_tokens": 13237500.0,
+      "step": 1630
+    },
+    {
+      "entropy": 0.6620127320289612,
+      "epoch": 0.922686230248307,
+      "grad_norm": 1.432976484298706,
+      "learning_rate": 4.980365119777882e-06,
+      "loss": 0.5251,
+      "mean_token_accuracy": 0.8438527345657348,
+      "num_tokens": 13277239.0,
+      "step": 1635
+    },
+    {
+      "entropy": 0.7288511991500854,
+      "epoch": 0.9255079006772009,
+      "grad_norm": 1.485297679901123,
+      "learning_rate": 4.980244983199211e-06,
+      "loss": 0.5528,
+      "mean_token_accuracy": 0.8336746454238891,
+      "num_tokens": 13317707.0,
+      "step": 1640
+    },
+    {
+      "entropy": 0.7077333211898804,
+      "epoch": 0.9283295711060948,
+      "grad_norm": 1.3443304300308228,
+      "learning_rate": 4.9801244821574216e-06,
+      "loss": 0.5556,
+      "mean_token_accuracy": 0.8335135579109192,
+      "num_tokens": 13358245.0,
+      "step": 1645
+    },
+    {
+      "entropy": 0.6646490931510926,
+      "epoch": 0.9311512415349887,
+      "grad_norm": 1.2345566749572754,
+      "learning_rate": 4.9800036166761866e-06,
+      "loss": 0.5416,
+      "mean_token_accuracy": 0.8396414875984192,
+      "num_tokens": 13398838.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.6034572184085846,
+      "epoch": 0.9339729119638827,
+      "grad_norm": 1.2768057584762573,
+      "learning_rate": 4.979882386779249e-06,
+      "loss": 0.463,
+      "mean_token_accuracy": 0.8570885419845581,
+      "num_tokens": 13439472.0,
+      "step": 1655
+    },
+    {
+      "entropy": 0.6616186738014221,
+      "epoch": 0.9367945823927766,
+      "grad_norm": 1.2942156791687012,
+      "learning_rate": 4.979760792490426e-06,
+      "loss": 0.5432,
+      "mean_token_accuracy": 0.8383080959320068,
+      "num_tokens": 13480194.0,
+      "step": 1660
+    },
+    {
+      "entropy": 0.6370395302772522,
+      "epoch": 0.9396162528216704,
+      "grad_norm": 1.384143352508545,
+      "learning_rate": 4.979638833833604e-06,
+      "loss": 0.4945,
+      "mean_token_accuracy": 0.8490473747253418,
+      "num_tokens": 13520923.0,
+      "step": 1665
+    },
+    {
+      "entropy": 0.648089063167572,
+      "epoch": 0.9424379232505643,
+      "grad_norm": 1.377890706062317,
+      "learning_rate": 4.979516510832743e-06,
+      "loss": 0.5125,
+      "mean_token_accuracy": 0.8464336395263672,
+      "num_tokens": 13561752.0,
+      "step": 1670
+    },
+    {
+      "entropy": 0.7131257057189941,
+      "epoch": 0.9452595936794582,
+      "grad_norm": 1.4829003810882568,
+      "learning_rate": 4.979393823511871e-06,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8287210106849671,
+      "num_tokens": 13602311.0,
+      "step": 1675
+    },
+    {
+      "entropy": 0.7149917125701905,
+      "epoch": 0.9480812641083521,
+      "grad_norm": 1.355993390083313,
+      "learning_rate": 4.979270771895093e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.827799940109253,
+      "num_tokens": 13642473.0,
+      "step": 1680
+    },
+    {
+      "entropy": 0.6665526628494263,
+      "epoch": 0.9509029345372461,
+      "grad_norm": 1.222419023513794,
+      "learning_rate": 4.979147356006579e-06,
+      "loss": 0.5285,
+      "mean_token_accuracy": 0.8411856889724731,
+      "num_tokens": 13683117.0,
+      "step": 1685
+    },
+    {
+      "entropy": 0.6773181915283203,
+      "epoch": 0.95372460496614,
+      "grad_norm": 1.3376954793930054,
+      "learning_rate": 4.979023575870577e-06,
+      "loss": 0.5327,
+      "mean_token_accuracy": 0.8385852217674256,
+      "num_tokens": 13723981.0,
+      "step": 1690
+    },
+    {
+      "entropy": 0.5895337462425232,
+      "epoch": 0.9565462753950339,
+      "grad_norm": 1.189743995666504,
+      "learning_rate": 4.978899431511401e-06,
+      "loss": 0.4602,
+      "mean_token_accuracy": 0.8576975584030151,
+      "num_tokens": 13764383.0,
+      "step": 1695
+    },
+    {
+      "entropy": 0.6576069235801697,
+      "epoch": 0.9593679458239278,
+      "grad_norm": 1.3030279874801636,
+      "learning_rate": 4.978774922953442e-06,
+      "loss": 0.5137,
+      "mean_token_accuracy": 0.847059428691864,
+      "num_tokens": 13805055.0,
+      "step": 1700
+    },
+    {
+      "entropy": 0.6367745637893677,
+      "epoch": 0.9621896162528216,
+      "grad_norm": 1.2039754390716553,
+      "learning_rate": 4.978650050221159e-06,
+      "loss": 0.5165,
+      "mean_token_accuracy": 0.8448570847511292,
+      "num_tokens": 13845904.0,
+      "step": 1705
+    },
+    {
+      "entropy": 0.6365428566932678,
+      "epoch": 0.9650112866817155,
+      "grad_norm": 1.157940149307251,
+      "learning_rate": 4.978524813339082e-06,
+      "loss": 0.5239,
+      "mean_token_accuracy": 0.8424256920814515,
+      "num_tokens": 13886682.0,
+      "step": 1710
+    },
+    {
+      "entropy": 0.66874258518219,
+      "epoch": 0.9678329571106095,
+      "grad_norm": 1.4614887237548828,
+      "learning_rate": 4.978399212331814e-06,
+      "loss": 0.5284,
+      "mean_token_accuracy": 0.8407912611961365,
+      "num_tokens": 13927181.0,
+      "step": 1715
+    },
+    {
+      "entropy": 0.643929636478424,
+      "epoch": 0.9706546275395034,
+      "grad_norm": 1.3897498846054077,
+      "learning_rate": 4.97827324722403e-06,
+      "loss": 0.5091,
+      "mean_token_accuracy": 0.8482011795043946,
+      "num_tokens": 13967885.0,
+      "step": 1720
+    },
+    {
+      "entropy": 0.6536713361740112,
+      "epoch": 0.9734762979683973,
+      "grad_norm": 1.3549083471298218,
+      "learning_rate": 4.978146918040476e-06,
+      "loss": 0.5065,
+      "mean_token_accuracy": 0.8463730335235595,
+      "num_tokens": 14008407.0,
+      "step": 1725
+    },
+    {
+      "entropy": 0.6258985161781311,
+      "epoch": 0.9762979683972912,
+      "grad_norm": 1.2369407415390015,
+      "learning_rate": 4.97802022480597e-06,
+      "loss": 0.4787,
+      "mean_token_accuracy": 0.8545310378074646,
+      "num_tokens": 14048919.0,
+      "step": 1730
+    },
+    {
+      "entropy": 0.6749670505523682,
+      "epoch": 0.9791196388261851,
+      "grad_norm": 1.5410014390945435,
+      "learning_rate": 4.977893167545398e-06,
+      "loss": 0.5411,
+      "mean_token_accuracy": 0.8386905789375305,
+      "num_tokens": 14089574.0,
+      "step": 1735
+    },
+    {
+      "entropy": 0.6478506565093994,
+      "epoch": 0.981941309255079,
+      "grad_norm": 1.3205047845840454,
+      "learning_rate": 4.977765746283724e-06,
+      "loss": 0.5138,
+      "mean_token_accuracy": 0.8434589385986329,
+      "num_tokens": 14130308.0,
+      "step": 1740
+    },
+    {
+      "entropy": 0.6606280326843261,
+      "epoch": 0.9847629796839729,
+      "grad_norm": 1.2447839975357056,
+      "learning_rate": 4.977637961045977e-06,
+      "loss": 0.5197,
+      "mean_token_accuracy": 0.8446960568428039,
+      "num_tokens": 14171031.0,
+      "step": 1745
+    },
+    {
+      "entropy": 0.6999424219131469,
+      "epoch": 0.9875846501128668,
+      "grad_norm": 1.2505875825881958,
+      "learning_rate": 4.977509811857263e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.828137469291687,
+      "num_tokens": 14211646.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.6732224225997925,
+      "epoch": 0.9904063205417607,
+      "grad_norm": 1.4964361190795898,
+      "learning_rate": 4.977381298742754e-06,
+      "loss": 0.5528,
+      "mean_token_accuracy": 0.8329331874847412,
+      "num_tokens": 14252319.0,
+      "step": 1755
+    },
+    {
+      "entropy": 0.6761491894721985,
+      "epoch": 0.9932279909706546,
+      "grad_norm": 1.426883339881897,
+      "learning_rate": 4.977252421727699e-06,
+      "loss": 0.5398,
+      "mean_token_accuracy": 0.8371304512023926,
+      "num_tokens": 14293055.0,
+      "step": 1760
+    },
+    {
+      "entropy": 0.6125996708869934,
+      "epoch": 0.9960496613995485,
+      "grad_norm": 1.3476899862289429,
+      "learning_rate": 4.977123180837416e-06,
+      "loss": 0.4946,
+      "mean_token_accuracy": 0.8501678228378295,
+      "num_tokens": 14333762.0,
+      "step": 1765
+    },
+    {
+      "entropy": 0.6156837463378906,
+      "epoch": 0.9988713318284425,
+      "grad_norm": 1.3276349306106567,
+      "learning_rate": 4.976993576097292e-06,
+      "loss": 0.5002,
+      "mean_token_accuracy": 0.8468546986579895,
+      "num_tokens": 14374541.0,
+      "step": 1770
+    },
+    {
+      "entropy": 0.5651689410209656,
+      "epoch": 1.0016930022573363,
+      "grad_norm": 0.9538065195083618,
+      "learning_rate": 4.97686360753279e-06,
+      "loss": 0.4177,
+      "mean_token_accuracy": 0.8691263794898987,
+      "num_tokens": 14409063.0,
+      "step": 1775
+    },
+    {
+      "entropy": 0.6223705530166626,
+      "epoch": 1.0045146726862302,
+      "grad_norm": 1.2451618909835815,
+      "learning_rate": 4.976733275169441e-06,
+      "loss": 0.438,
+      "mean_token_accuracy": 0.8648397326469421,
+      "num_tokens": 14449840.0,
+      "step": 1780
+    },
+    {
+      "entropy": 0.5574550032615662,
+      "epoch": 1.007336343115124,
+      "grad_norm": 1.2733477354049683,
+      "learning_rate": 4.976602579032849e-06,
+      "loss": 0.4043,
+      "mean_token_accuracy": 0.8714008688926697,
+      "num_tokens": 14490449.0,
+      "step": 1785
+    },
+    {
+      "entropy": 0.5262600839138031,
+      "epoch": 1.010158013544018,
+      "grad_norm": 1.3276273012161255,
+      "learning_rate": 4.976471519148691e-06,
+      "loss": 0.4036,
+      "mean_token_accuracy": 0.8747005462646484,
+      "num_tokens": 14531107.0,
+      "step": 1790
+    },
+    {
+      "entropy": 0.49118422865867617,
+      "epoch": 1.012979683972912,
+      "grad_norm": 1.2698107957839966,
+      "learning_rate": 4.976340095542711e-06,
+      "loss": 0.3685,
+      "mean_token_accuracy": 0.8826399683952332,
+      "num_tokens": 14571891.0,
+      "step": 1795
+    },
+    {
+      "entropy": 0.48148898482322694,
+      "epoch": 1.0158013544018059,
+      "grad_norm": 1.3123940229415894,
+      "learning_rate": 4.97620830824073e-06,
+      "loss": 0.3673,
+      "mean_token_accuracy": 0.8846578359603882,
+      "num_tokens": 14612632.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.5638070046901703,
+      "epoch": 1.0186230248306998,
+      "grad_norm": 1.2908910512924194,
+      "learning_rate": 4.976076157268636e-06,
+      "loss": 0.4132,
+      "mean_token_accuracy": 0.8727042317390442,
+      "num_tokens": 14653317.0,
+      "step": 1805
+    },
+    {
+      "entropy": 0.5422739446163177,
+      "epoch": 1.0214446952595937,
+      "grad_norm": 1.5749247074127197,
+      "learning_rate": 4.975943642652389e-06,
+      "loss": 0.4113,
+      "mean_token_accuracy": 0.8708185434341431,
+      "num_tokens": 14693958.0,
+      "step": 1810
+    },
+    {
+      "entropy": 0.5249835729599,
+      "epoch": 1.0242663656884876,
+      "grad_norm": 1.3571456670761108,
+      "learning_rate": 4.975810764418023e-06,
+      "loss": 0.3962,
+      "mean_token_accuracy": 0.8773154497146607,
+      "num_tokens": 14734520.0,
+      "step": 1815
+    },
+    {
+      "entropy": 0.4799295425415039,
+      "epoch": 1.0270880361173815,
+      "grad_norm": 1.2057713270187378,
+      "learning_rate": 4.975677522591642e-06,
+      "loss": 0.3698,
+      "mean_token_accuracy": 0.881498396396637,
+      "num_tokens": 14775267.0,
+      "step": 1820
+    },
+    {
+      "entropy": 0.5456785798072815,
+      "epoch": 1.0299097065462754,
+      "grad_norm": 1.4665296077728271,
+      "learning_rate": 4.975543917199422e-06,
+      "loss": 0.4268,
+      "mean_token_accuracy": 0.8673960208892822,
+      "num_tokens": 14816106.0,
+      "step": 1825
+    },
+    {
+      "entropy": 0.540210634469986,
+      "epoch": 1.0327313769751694,
+      "grad_norm": 1.2391974925994873,
+      "learning_rate": 4.975409948267608e-06,
+      "loss": 0.416,
+      "mean_token_accuracy": 0.8694531202316285,
+      "num_tokens": 14856910.0,
+      "step": 1830
+    },
+    {
+      "entropy": 0.5396132886409759,
+      "epoch": 1.0355530474040633,
+      "grad_norm": 1.2638976573944092,
+      "learning_rate": 4.97527561582252e-06,
+      "loss": 0.4162,
+      "mean_token_accuracy": 0.8690710544586182,
+      "num_tokens": 14897584.0,
+      "step": 1835
+    },
+    {
+      "entropy": 0.5132744669914245,
+      "epoch": 1.0383747178329572,
+      "grad_norm": 1.2913990020751953,
+      "learning_rate": 4.975140919890546e-06,
+      "loss": 0.4015,
+      "mean_token_accuracy": 0.873302161693573,
+      "num_tokens": 14938181.0,
+      "step": 1840
+    },
+    {
+      "entropy": 0.5024015605449677,
+      "epoch": 1.041196388261851,
+      "grad_norm": 1.1151163578033447,
+      "learning_rate": 4.975005860498148e-06,
+      "loss": 0.3876,
+      "mean_token_accuracy": 0.8748369932174682,
+      "num_tokens": 14978884.0,
+      "step": 1845
+    },
+    {
+      "entropy": 0.513664311170578,
+      "epoch": 1.0440180586907448,
+      "grad_norm": 1.298097014427185,
+      "learning_rate": 4.974870437671858e-06,
+      "loss": 0.4012,
+      "mean_token_accuracy": 0.8729772806167603,
+      "num_tokens": 15019646.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.5189130663871765,
+      "epoch": 1.0468397291196387,
+      "grad_norm": 1.3805726766586304,
+      "learning_rate": 4.97473465143828e-06,
+      "loss": 0.405,
+      "mean_token_accuracy": 0.8714963912963867,
+      "num_tokens": 15060355.0,
+      "step": 1855
+    },
+    {
+      "entropy": 0.5300206661224365,
+      "epoch": 1.0496613995485327,
+      "grad_norm": 1.3854206800460815,
+      "learning_rate": 4.9745985018240895e-06,
+      "loss": 0.4155,
+      "mean_token_accuracy": 0.8686167359352112,
+      "num_tokens": 15101164.0,
+      "step": 1860
+    },
+    {
+      "entropy": 0.49498083591461184,
+      "epoch": 1.0524830699774266,
+      "grad_norm": 1.3298227787017822,
+      "learning_rate": 4.974461988856033e-06,
+      "loss": 0.3802,
+      "mean_token_accuracy": 0.8796196699142456,
+      "num_tokens": 15141937.0,
+      "step": 1865
+    },
+    {
+      "entropy": 0.5611091613769531,
+      "epoch": 1.0553047404063205,
+      "grad_norm": 1.3966474533081055,
+      "learning_rate": 4.974325112560928e-06,
+      "loss": 0.4562,
+      "mean_token_accuracy": 0.8588791251182556,
+      "num_tokens": 15182669.0,
+      "step": 1870
+    },
+    {
+      "entropy": 0.5468938708305359,
+      "epoch": 1.0581264108352144,
+      "grad_norm": 1.2755643129348755,
+      "learning_rate": 4.974187872965665e-06,
+      "loss": 0.4084,
+      "mean_token_accuracy": 0.8714737057685852,
+      "num_tokens": 15223336.0,
+      "step": 1875
+    },
+    {
+      "entropy": 0.5215595126152038,
+      "epoch": 1.0609480812641083,
+      "grad_norm": 1.296976923942566,
+      "learning_rate": 4.974050270097203e-06,
+      "loss": 0.378,
+      "mean_token_accuracy": 0.8788340568542481,
+      "num_tokens": 15264013.0,
+      "step": 1880
+    },
+    {
+      "entropy": 0.5012482941150666,
+      "epoch": 1.0637697516930023,
+      "grad_norm": 1.3608322143554688,
+      "learning_rate": 4.973912303982575e-06,
+      "loss": 0.3661,
+      "mean_token_accuracy": 0.8823076605796814,
+      "num_tokens": 15304788.0,
+      "step": 1885
+    },
+    {
+      "entropy": 0.5157119750976562,
+      "epoch": 1.0665914221218962,
+      "grad_norm": 1.3526833057403564,
+      "learning_rate": 4.973773974648885e-06,
+      "loss": 0.4074,
+      "mean_token_accuracy": 0.8713153243064881,
+      "num_tokens": 15345095.0,
+      "step": 1890
+    },
+    {
+      "entropy": 0.5500873923301697,
+      "epoch": 1.06941309255079,
+      "grad_norm": 1.2834185361862183,
+      "learning_rate": 4.973635282123308e-06,
+      "loss": 0.4127,
+      "mean_token_accuracy": 0.8706364631652832,
+      "num_tokens": 15385312.0,
+      "step": 1895
+    },
+    {
+      "entropy": 0.5158918261528015,
+      "epoch": 1.072234762979684,
+      "grad_norm": 1.3259310722351074,
+      "learning_rate": 4.973496226433089e-06,
+      "loss": 0.395,
+      "mean_token_accuracy": 0.8755181312561036,
+      "num_tokens": 15426061.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.538950902223587,
+      "epoch": 1.075056433408578,
+      "grad_norm": 1.3279820680618286,
+      "learning_rate": 4.973356807605546e-06,
+      "loss": 0.4104,
+      "mean_token_accuracy": 0.8712123036384583,
+      "num_tokens": 15466630.0,
+      "step": 1905
+    },
+    {
+      "entropy": 0.49377835392951963,
+      "epoch": 1.0778781038374718,
+      "grad_norm": 1.1958774328231812,
+      "learning_rate": 4.973217025668068e-06,
+      "loss": 0.379,
+      "mean_token_accuracy": 0.8779303431510925,
+      "num_tokens": 15507113.0,
+      "step": 1910
+    },
+    {
+      "entropy": 0.5405571818351745,
+      "epoch": 1.0806997742663658,
+      "grad_norm": 1.2505043745040894,
+      "learning_rate": 4.973076880648115e-06,
+      "loss": 0.4102,
+      "mean_token_accuracy": 0.8709590196609497,
+      "num_tokens": 15547424.0,
+      "step": 1915
+    },
+    {
+      "entropy": 0.5551674902439118,
+      "epoch": 1.0835214446952597,
+      "grad_norm": 1.366829514503479,
+      "learning_rate": 4.972936372573218e-06,
+      "loss": 0.45,
+      "mean_token_accuracy": 0.8604497313499451,
+      "num_tokens": 15588125.0,
+      "step": 1920
+    },
+    {
+      "entropy": 0.5369654476642609,
+      "epoch": 1.0863431151241536,
+      "grad_norm": 1.367258906364441,
+      "learning_rate": 4.972795501470981e-06,
+      "loss": 0.4292,
+      "mean_token_accuracy": 0.8665614485740661,
+      "num_tokens": 15628472.0,
+      "step": 1925
+    },
+    {
+      "entropy": 0.5272303640842437,
+      "epoch": 1.0891647855530473,
+      "grad_norm": 1.4138174057006836,
+      "learning_rate": 4.972654267369078e-06,
+      "loss": 0.4193,
+      "mean_token_accuracy": 0.8680860280990601,
+      "num_tokens": 15669129.0,
+      "step": 1930
+    },
+    {
+      "entropy": 0.5395491600036622,
+      "epoch": 1.0919864559819412,
+      "grad_norm": 1.3633450269699097,
+      "learning_rate": 4.972512670295253e-06,
+      "loss": 0.4003,
+      "mean_token_accuracy": 0.8737997651100159,
+      "num_tokens": 15709941.0,
+      "step": 1935
+    },
+    {
+      "entropy": 0.5212275326251984,
+      "epoch": 1.0948081264108351,
+      "grad_norm": 1.37870454788208,
+      "learning_rate": 4.9723707102773235e-06,
+      "loss": 0.3997,
+      "mean_token_accuracy": 0.8710021495819091,
+      "num_tokens": 15750468.0,
+      "step": 1940
+    },
+    {
+      "entropy": 0.533467584848404,
+      "epoch": 1.097629796839729,
+      "grad_norm": 1.5006767511367798,
+      "learning_rate": 4.972228387343179e-06,
+      "loss": 0.4071,
+      "mean_token_accuracy": 0.8726310014724732,
+      "num_tokens": 15791223.0,
+      "step": 1945
+    },
+    {
+      "entropy": 0.5062848865985871,
+      "epoch": 1.100451467268623,
+      "grad_norm": 1.1996793746948242,
+      "learning_rate": 4.972085701520777e-06,
+      "loss": 0.3961,
+      "mean_token_accuracy": 0.8720014691352844,
+      "num_tokens": 15831746.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.5338364660739898,
+      "epoch": 1.103273137697517,
+      "grad_norm": 1.4911237955093384,
+      "learning_rate": 4.971942652838149e-06,
+      "loss": 0.4187,
+      "mean_token_accuracy": 0.8707549214363098,
+      "num_tokens": 15872459.0,
+      "step": 1955
+    },
+    {
+      "entropy": 0.5420314610004425,
+      "epoch": 1.1060948081264108,
+      "grad_norm": 1.4543707370758057,
+      "learning_rate": 4.971799241323397e-06,
+      "loss": 0.4279,
+      "mean_token_accuracy": 0.8651981711387634,
+      "num_tokens": 15913132.0,
+      "step": 1960
+    },
+    {
+      "entropy": 0.533877170085907,
+      "epoch": 1.1089164785553047,
+      "grad_norm": 1.4339110851287842,
+      "learning_rate": 4.971655467004693e-06,
+      "loss": 0.4169,
+      "mean_token_accuracy": 0.8666856288909912,
+      "num_tokens": 15953596.0,
+      "step": 1965
+    },
+    {
+      "entropy": 0.525646859407425,
+      "epoch": 1.1117381489841986,
+      "grad_norm": 1.370369791984558,
+      "learning_rate": 4.971511329910283e-06,
+      "loss": 0.4201,
+      "mean_token_accuracy": 0.8674192786216736,
+      "num_tokens": 15994327.0,
+      "step": 1970
+    },
+    {
+      "entropy": 0.5182268440723419,
+      "epoch": 1.1145598194130926,
+      "grad_norm": 1.2392444610595703,
+      "learning_rate": 4.971366830068483e-06,
+      "loss": 0.3987,
+      "mean_token_accuracy": 0.8725781679153443,
+      "num_tokens": 16035039.0,
+      "step": 1975
+    },
+    {
+      "entropy": 0.5197380304336547,
+      "epoch": 1.1173814898419865,
+      "grad_norm": 1.3075342178344727,
+      "learning_rate": 4.971221967507679e-06,
+      "loss": 0.4074,
+      "mean_token_accuracy": 0.8703087091445922,
+      "num_tokens": 16075828.0,
+      "step": 1980
+    },
+    {
+      "entropy": 0.5025268793106079,
+      "epoch": 1.1202031602708804,
+      "grad_norm": 1.4279546737670898,
+      "learning_rate": 4.9710767422563285e-06,
+      "loss": 0.3803,
+      "mean_token_accuracy": 0.8805047512054444,
+      "num_tokens": 16116500.0,
+      "step": 1985
+    },
+    {
+      "entropy": 0.533239609003067,
+      "epoch": 1.1230248306997743,
+      "grad_norm": 1.2557505369186401,
+      "learning_rate": 4.970931154342963e-06,
+      "loss": 0.4162,
+      "mean_token_accuracy": 0.8681887149810791,
+      "num_tokens": 16157026.0,
+      "step": 1990
+    },
+    {
+      "entropy": 0.5435242295265198,
+      "epoch": 1.1258465011286682,
+      "grad_norm": 1.3608372211456299,
+      "learning_rate": 4.970785203796182e-06,
+      "loss": 0.4237,
+      "mean_token_accuracy": 0.8656893134117126,
+      "num_tokens": 16197572.0,
+      "step": 1995
+    },
+    {
+      "entropy": 0.48926340937614443,
+      "epoch": 1.1286681715575622,
+      "grad_norm": 1.314292311668396,
+      "learning_rate": 4.970638890644658e-06,
+      "loss": 0.3675,
+      "mean_token_accuracy": 0.8834827184677124,
+      "num_tokens": 16238388.0,
+      "step": 2000
+    },
+    {
+      "epoch": 1.1286681715575622,
+      "eval_entropy": 0.5074299573898315,
+      "eval_loss": 0.3934297561645508,
+      "eval_mean_token_accuracy": 0.8853967189788818,
+      "eval_num_tokens": 16238388.0,
+      "eval_runtime": 0.1639,
+      "eval_samples_per_second": 24.402,
+      "eval_steps_per_second": 6.1,
+      "step": 2000
+    },
+    {
+      "entropy": 0.5785790205001831,
+      "epoch": 1.1314898419864559,
+      "grad_norm": 1.5585596561431885,
+      "learning_rate": 4.970492214917133e-06,
+      "loss": 0.4792,
+      "mean_token_accuracy": 0.8502647757530213,
+      "num_tokens": 16279114.0,
+      "step": 2005
+    },
+    {
+      "entropy": 0.5156497418880462,
+      "epoch": 1.13431151241535,
+      "grad_norm": 1.2416678667068481,
+      "learning_rate": 4.970345176642424e-06,
+      "loss": 0.409,
+      "mean_token_accuracy": 0.8698260307312011,
+      "num_tokens": 16319792.0,
+      "step": 2010
+    },
+    {
+      "entropy": 0.50312060713768,
+      "epoch": 1.1371331828442437,
+      "grad_norm": 1.343417763710022,
+      "learning_rate": 4.9701977758494135e-06,
+      "loss": 0.402,
+      "mean_token_accuracy": 0.8738492012023926,
+      "num_tokens": 16360729.0,
+      "step": 2015
+    },
+    {
+      "entropy": 0.47664583325386045,
+      "epoch": 1.1399548532731376,
+      "grad_norm": 1.3557952642440796,
+      "learning_rate": 4.970050012567061e-06,
+      "loss": 0.3691,
+      "mean_token_accuracy": 0.8808457016944885,
+      "num_tokens": 16401565.0,
+      "step": 2020
+    },
+    {
+      "entropy": 0.5471422076225281,
+      "epoch": 1.1427765237020315,
+      "grad_norm": 1.6078767776489258,
+      "learning_rate": 4.969901886824394e-06,
+      "loss": 0.4408,
+      "mean_token_accuracy": 0.861269211769104,
+      "num_tokens": 16442011.0,
+      "step": 2025
+    },
+    {
+      "entropy": 0.500029307603836,
+      "epoch": 1.1455981941309255,
+      "grad_norm": 1.2382087707519531,
+      "learning_rate": 4.969753398650511e-06,
+      "loss": 0.3831,
+      "mean_token_accuracy": 0.8787751078605652,
+      "num_tokens": 16482704.0,
+      "step": 2030
+    },
+    {
+      "entropy": 0.5860454142093658,
+      "epoch": 1.1484198645598194,
+      "grad_norm": 1.2965975999832153,
+      "learning_rate": 4.969604548074583e-06,
+      "loss": 0.4537,
+      "mean_token_accuracy": 0.858244788646698,
+      "num_tokens": 16523275.0,
+      "step": 2035
+    },
+    {
+      "entropy": 0.5472402393817901,
+      "epoch": 1.1512415349887133,
+      "grad_norm": 1.347503900527954,
+      "learning_rate": 4.969455335125852e-06,
+      "loss": 0.4212,
+      "mean_token_accuracy": 0.8660460352897644,
+      "num_tokens": 16563975.0,
+      "step": 2040
+    },
+    {
+      "entropy": 0.5726926565170288,
+      "epoch": 1.1540632054176072,
+      "grad_norm": 1.3004218339920044,
+      "learning_rate": 4.969305759833631e-06,
+      "loss": 0.4574,
+      "mean_token_accuracy": 0.8575977206230163,
+      "num_tokens": 16604607.0,
+      "step": 2045
+    },
+    {
+      "entropy": 0.554759293794632,
+      "epoch": 1.1568848758465011,
+      "grad_norm": 1.1861565113067627,
+      "learning_rate": 4.969155822227304e-06,
+      "loss": 0.4342,
+      "mean_token_accuracy": 0.8664988040924072,
+      "num_tokens": 16645243.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.5388511419296265,
+      "epoch": 1.159706546275395,
+      "grad_norm": 1.6149320602416992,
+      "learning_rate": 4.969005522336324e-06,
+      "loss": 0.4337,
+      "mean_token_accuracy": 0.8644424319267273,
+      "num_tokens": 16685870.0,
+      "step": 2055
+    },
+    {
+      "entropy": 0.5282904028892517,
+      "epoch": 1.162528216704289,
+      "grad_norm": 1.2274096012115479,
+      "learning_rate": 4.968854860190222e-06,
+      "loss": 0.427,
+      "mean_token_accuracy": 0.8676736235618592,
+      "num_tokens": 16726814.0,
+      "step": 2060
+    },
+    {
+      "entropy": 0.5028113007545472,
+      "epoch": 1.1653498871331829,
+      "grad_norm": 1.2310305833816528,
+      "learning_rate": 4.9687038358185904e-06,
+      "loss": 0.3945,
+      "mean_token_accuracy": 0.8749886631965638,
+      "num_tokens": 16767430.0,
+      "step": 2065
+    },
+    {
+      "entropy": 0.5054876983165741,
+      "epoch": 1.1681715575620768,
+      "grad_norm": 1.6121010780334473,
+      "learning_rate": 4.968552449251103e-06,
+      "loss": 0.4005,
+      "mean_token_accuracy": 0.8731933832168579,
+      "num_tokens": 16808070.0,
+      "step": 2070
+    },
+    {
+      "entropy": 0.510890108346939,
+      "epoch": 1.1709932279909707,
+      "grad_norm": 1.212408185005188,
+      "learning_rate": 4.968400700517496e-06,
+      "loss": 0.402,
+      "mean_token_accuracy": 0.8713488578796387,
+      "num_tokens": 16848790.0,
+      "step": 2075
+    },
+    {
+      "entropy": 0.5575174331665039,
+      "epoch": 1.1738148984198646,
+      "grad_norm": 1.3433505296707153,
+      "learning_rate": 4.968248589647582e-06,
+      "loss": 0.4241,
+      "mean_token_accuracy": 0.8673989772796631,
+      "num_tokens": 16888705.0,
+      "step": 2080
+    },
+    {
+      "entropy": 0.4738758623600006,
+      "epoch": 1.1766365688487586,
+      "grad_norm": 1.337335467338562,
+      "learning_rate": 4.968096116671243e-06,
+      "loss": 0.3633,
+      "mean_token_accuracy": 0.8840928077697754,
+      "num_tokens": 16929128.0,
+      "step": 2085
+    },
+    {
+      "entropy": 0.5100393950939178,
+      "epoch": 1.1794582392776523,
+      "grad_norm": 1.4914759397506714,
+      "learning_rate": 4.9679432816184316e-06,
+      "loss": 0.4018,
+      "mean_token_accuracy": 0.8735705256462097,
+      "num_tokens": 16969800.0,
+      "step": 2090
+    },
+    {
+      "entropy": 0.5905535221099854,
+      "epoch": 1.1822799097065464,
+      "grad_norm": 1.579995036125183,
+      "learning_rate": 4.967790084519174e-06,
+      "loss": 0.4902,
+      "mean_token_accuracy": 0.850010359287262,
+      "num_tokens": 17010171.0,
+      "step": 2095
+    },
+    {
+      "entropy": 0.5488203287124633,
+      "epoch": 1.18510158013544,
+      "grad_norm": 1.3214325904846191,
+      "learning_rate": 4.9676365254035645e-06,
+      "loss": 0.4295,
+      "mean_token_accuracy": 0.8675566554069519,
+      "num_tokens": 17050042.0,
+      "step": 2100
+    },
+    {
+      "entropy": 0.49664353728294375,
+      "epoch": 1.187923250564334,
+      "grad_norm": 1.3970224857330322,
+      "learning_rate": 4.96748260430177e-06,
+      "loss": 0.3776,
+      "mean_token_accuracy": 0.8784841656684875,
+      "num_tokens": 17090848.0,
+      "step": 2105
+    },
+    {
+      "entropy": 0.5250966966152191,
+      "epoch": 1.190744920993228,
+      "grad_norm": 1.3193445205688477,
+      "learning_rate": 4.967328321244028e-06,
+      "loss": 0.4011,
+      "mean_token_accuracy": 0.8732633233070374,
+      "num_tokens": 17131555.0,
+      "step": 2110
+    },
+    {
+      "entropy": 0.5157673597335816,
+      "epoch": 1.1935665914221218,
+      "grad_norm": 1.2861922979354858,
+      "learning_rate": 4.967173676260648e-06,
+      "loss": 0.4074,
+      "mean_token_accuracy": 0.873046362400055,
+      "num_tokens": 17172227.0,
+      "step": 2115
+    },
+    {
+      "entropy": 0.5405256807804107,
+      "epoch": 1.1963882618510158,
+      "grad_norm": 1.6772886514663696,
+      "learning_rate": 4.96701866938201e-06,
+      "loss": 0.4271,
+      "mean_token_accuracy": 0.8637170910835266,
+      "num_tokens": 17212727.0,
+      "step": 2120
+    },
+    {
+      "entropy": 0.5377147376537323,
+      "epoch": 1.1992099322799097,
+      "grad_norm": 1.3887752294540405,
+      "learning_rate": 4.9668633006385655e-06,
+      "loss": 0.4275,
+      "mean_token_accuracy": 0.8655383706092834,
+      "num_tokens": 17253478.0,
+      "step": 2125
+    },
+    {
+      "entropy": 0.5166448771953582,
+      "epoch": 1.2020316027088036,
+      "grad_norm": 1.2280800342559814,
+      "learning_rate": 4.966707570060835e-06,
+      "loss": 0.3918,
+      "mean_token_accuracy": 0.8748661637306213,
+      "num_tokens": 17294254.0,
+      "step": 2130
+    },
+    {
+      "entropy": 0.5040790915489197,
+      "epoch": 1.2048532731376975,
+      "grad_norm": 1.4066789150238037,
+      "learning_rate": 4.9665514776794145e-06,
+      "loss": 0.3728,
+      "mean_token_accuracy": 0.8822136998176575,
+      "num_tokens": 17335017.0,
+      "step": 2135
+    },
+    {
+      "entropy": 0.5264786601066589,
+      "epoch": 1.2076749435665914,
+      "grad_norm": 1.3595073223114014,
+      "learning_rate": 4.9663950235249655e-06,
+      "loss": 0.4151,
+      "mean_token_accuracy": 0.8696430683135986,
+      "num_tokens": 17375865.0,
+      "step": 2140
+    },
+    {
+      "entropy": 0.5130272388458252,
+      "epoch": 1.2104966139954854,
+      "grad_norm": 1.4424818754196167,
+      "learning_rate": 4.966238207628225e-06,
+      "loss": 0.4016,
+      "mean_token_accuracy": 0.8746838808059693,
+      "num_tokens": 17416458.0,
+      "step": 2145
+    },
+    {
+      "entropy": 0.49900220036506654,
+      "epoch": 1.2133182844243793,
+      "grad_norm": 1.3563543558120728,
+      "learning_rate": 4.966081030019999e-06,
+      "loss": 0.407,
+      "mean_token_accuracy": 0.872953200340271,
+      "num_tokens": 17457303.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.5183020770549774,
+      "epoch": 1.2161399548532732,
+      "grad_norm": 1.431388258934021,
+      "learning_rate": 4.965923490731166e-06,
+      "loss": 0.4001,
+      "mean_token_accuracy": 0.8713582038879395,
+      "num_tokens": 17498163.0,
+      "step": 2155
+    },
+    {
+      "entropy": 0.5679690599441528,
+      "epoch": 1.2189616252821671,
+      "grad_norm": 1.5881187915802002,
+      "learning_rate": 4.965765589792674e-06,
+      "loss": 0.4542,
+      "mean_token_accuracy": 0.8564951181411743,
+      "num_tokens": 17538730.0,
+      "step": 2160
+    },
+    {
+      "entropy": 0.5158195972442627,
+      "epoch": 1.221783295711061,
+      "grad_norm": 1.2861218452453613,
+      "learning_rate": 4.965607327235542e-06,
+      "loss": 0.3973,
+      "mean_token_accuracy": 0.8753223657608032,
+      "num_tokens": 17579319.0,
+      "step": 2165
+    },
+    {
+      "entropy": 0.5332009434700012,
+      "epoch": 1.224604966139955,
+      "grad_norm": 1.338793158531189,
+      "learning_rate": 4.965448703090861e-06,
+      "loss": 0.4198,
+      "mean_token_accuracy": 0.8673290014266968,
+      "num_tokens": 17620049.0,
+      "step": 2170
+    },
+    {
+      "entropy": 0.5594852328300476,
+      "epoch": 1.2274266365688487,
+      "grad_norm": 1.3682115077972412,
+      "learning_rate": 4.965289717389794e-06,
+      "loss": 0.4309,
+      "mean_token_accuracy": 0.8634846925735473,
+      "num_tokens": 17660479.0,
+      "step": 2175
+    },
+    {
+      "entropy": 0.5210170686244965,
+      "epoch": 1.2302483069977426,
+      "grad_norm": 1.4854122400283813,
+      "learning_rate": 4.965130370163572e-06,
+      "loss": 0.391,
+      "mean_token_accuracy": 0.8751443147659301,
+      "num_tokens": 17701156.0,
+      "step": 2180
+    },
+    {
+      "entropy": 0.5292206406593323,
+      "epoch": 1.2330699774266365,
+      "grad_norm": 1.3995417356491089,
+      "learning_rate": 4.9649706614435e-06,
+      "loss": 0.4127,
+      "mean_token_accuracy": 0.8673221230506897,
+      "num_tokens": 17742000.0,
+      "step": 2185
+    },
+    {
+      "entropy": 0.5528486132621765,
+      "epoch": 1.2358916478555304,
+      "grad_norm": 1.4038658142089844,
+      "learning_rate": 4.9648105912609525e-06,
+      "loss": 0.4354,
+      "mean_token_accuracy": 0.863236916065216,
+      "num_tokens": 17782813.0,
+      "step": 2190
+    },
+    {
+      "entropy": 0.520248967409134,
+      "epoch": 1.2387133182844243,
+      "grad_norm": 1.3488070964813232,
+      "learning_rate": 4.964650159647375e-06,
+      "loss": 0.4095,
+      "mean_token_accuracy": 0.8726580619812012,
+      "num_tokens": 17823603.0,
+      "step": 2195
+    },
+    {
+      "entropy": 0.5257691144943237,
+      "epoch": 1.2415349887133182,
+      "grad_norm": 1.3696351051330566,
+      "learning_rate": 4.964489366634285e-06,
+      "loss": 0.4045,
+      "mean_token_accuracy": 0.8737285494804382,
+      "num_tokens": 17864326.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.521408861875534,
+      "epoch": 1.2443566591422122,
+      "grad_norm": 1.2956690788269043,
+      "learning_rate": 4.964328212253269e-06,
+      "loss": 0.4046,
+      "mean_token_accuracy": 0.8745855331420899,
+      "num_tokens": 17904832.0,
+      "step": 2205
+    },
+    {
+      "entropy": 0.5116970539093018,
+      "epoch": 1.247178329571106,
+      "grad_norm": 1.1543620824813843,
+      "learning_rate": 4.9641666965359865e-06,
+      "loss": 0.4061,
+      "mean_token_accuracy": 0.871479332447052,
+      "num_tokens": 17945706.0,
+      "step": 2210
+    },
+    {
+      "entropy": 0.51253702044487,
+      "epoch": 1.25,
+      "grad_norm": 1.3165756464004517,
+      "learning_rate": 4.9640048195141685e-06,
+      "loss": 0.4001,
+      "mean_token_accuracy": 0.8732204794883728,
+      "num_tokens": 17986493.0,
+      "step": 2215
+    },
+    {
+      "entropy": 0.5361522495746612,
+      "epoch": 1.252821670428894,
+      "grad_norm": 1.3089770078659058,
+      "learning_rate": 4.9638425812196145e-06,
+      "loss": 0.3951,
+      "mean_token_accuracy": 0.8748211622238159,
+      "num_tokens": 18027243.0,
+      "step": 2220
+    },
+    {
+      "entropy": 0.4924753546714783,
+      "epoch": 1.2556433408577878,
+      "grad_norm": 1.3457053899765015,
+      "learning_rate": 4.963679981684195e-06,
+      "loss": 0.3846,
+      "mean_token_accuracy": 0.8790703415870667,
+      "num_tokens": 18067621.0,
+      "step": 2225
+    },
+    {
+      "entropy": 0.5386804461479187,
+      "epoch": 1.2584650112866818,
+      "grad_norm": 1.543764352798462,
+      "learning_rate": 4.963517020939855e-06,
+      "loss": 0.4387,
+      "mean_token_accuracy": 0.8642153739929199,
+      "num_tokens": 18108318.0,
+      "step": 2230
+    },
+    {
+      "entropy": 0.5258350789546966,
+      "epoch": 1.2612866817155757,
+      "grad_norm": 1.33335280418396,
+      "learning_rate": 4.963353699018607e-06,
+      "loss": 0.4202,
+      "mean_token_accuracy": 0.867599892616272,
+      "num_tokens": 18149098.0,
+      "step": 2235
+    },
+    {
+      "entropy": 0.4915396809577942,
+      "epoch": 1.2641083521444696,
+      "grad_norm": 1.3478151559829712,
+      "learning_rate": 4.963190015952536e-06,
+      "loss": 0.3941,
+      "mean_token_accuracy": 0.8742651224136353,
+      "num_tokens": 18189853.0,
+      "step": 2240
+    },
+    {
+      "entropy": 0.5037339508533478,
+      "epoch": 1.2669300225733635,
+      "grad_norm": 1.4654349088668823,
+      "learning_rate": 4.963025971773798e-06,
+      "loss": 0.3896,
+      "mean_token_accuracy": 0.8763032555580139,
+      "num_tokens": 18230648.0,
+      "step": 2245
+    },
+    {
+      "entropy": 0.5388745546340943,
+      "epoch": 1.2697516930022572,
+      "grad_norm": 1.3390324115753174,
+      "learning_rate": 4.962861566514618e-06,
+      "loss": 0.418,
+      "mean_token_accuracy": 0.865935492515564,
+      "num_tokens": 18271377.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.5285961091518402,
+      "epoch": 1.2725733634311513,
+      "grad_norm": 1.2784810066223145,
+      "learning_rate": 4.962696800207295e-06,
+      "loss": 0.4282,
+      "mean_token_accuracy": 0.8682160973548889,
+      "num_tokens": 18311967.0,
+      "step": 2255
+    },
+    {
+      "entropy": 0.5452647149562836,
+      "epoch": 1.275395033860045,
+      "grad_norm": 1.4500328302383423,
+      "learning_rate": 4.9625316728841966e-06,
+      "loss": 0.4185,
+      "mean_token_accuracy": 0.8690374493598938,
+      "num_tokens": 18352492.0,
+      "step": 2260
+    },
+    {
+      "entropy": 0.5347442328929901,
+      "epoch": 1.2782167042889392,
+      "grad_norm": 1.3491861820220947,
+      "learning_rate": 4.962366184577762e-06,
+      "loss": 0.3963,
+      "mean_token_accuracy": 0.8743094444274903,
+      "num_tokens": 18393148.0,
+      "step": 2265
+    },
+    {
+      "entropy": 0.5413298010826111,
+      "epoch": 1.2810383747178329,
+      "grad_norm": 1.3183116912841797,
+      "learning_rate": 4.962200335320502e-06,
+      "loss": 0.43,
+      "mean_token_accuracy": 0.8649118781089783,
+      "num_tokens": 18433888.0,
+      "step": 2270
+    },
+    {
+      "entropy": 0.49364901781082154,
+      "epoch": 1.2838600451467268,
+      "grad_norm": 1.4001456499099731,
+      "learning_rate": 4.962034125144997e-06,
+      "loss": 0.3979,
+      "mean_token_accuracy": 0.8754386663436889,
+      "num_tokens": 18474725.0,
+      "step": 2275
+    },
+    {
+      "entropy": 0.4968021988868713,
+      "epoch": 1.2866817155756207,
+      "grad_norm": 1.4670476913452148,
+      "learning_rate": 4.961867554083899e-06,
+      "loss": 0.3737,
+      "mean_token_accuracy": 0.8795897960662842,
+      "num_tokens": 18515515.0,
+      "step": 2280
+    },
+    {
+      "entropy": 0.49011672139167783,
+      "epoch": 1.2895033860045146,
+      "grad_norm": 1.5934028625488281,
+      "learning_rate": 4.961700622169931e-06,
+      "loss": 0.3733,
+      "mean_token_accuracy": 0.8795024633407593,
+      "num_tokens": 18556216.0,
+      "step": 2285
+    },
+    {
+      "entropy": 0.5093987703323364,
+      "epoch": 1.2923250564334086,
+      "grad_norm": 1.2995637655258179,
+      "learning_rate": 4.961533329435888e-06,
+      "loss": 0.4071,
+      "mean_token_accuracy": 0.8725569367408752,
+      "num_tokens": 18596786.0,
+      "step": 2290
+    },
+    {
+      "entropy": 0.5331693708896637,
+      "epoch": 1.2951467268623025,
+      "grad_norm": 1.4760665893554688,
+      "learning_rate": 4.9613656759146335e-06,
+      "loss": 0.4109,
+      "mean_token_accuracy": 0.869564437866211,
+      "num_tokens": 18637493.0,
+      "step": 2295
+    },
+    {
+      "entropy": 0.5001338601112366,
+      "epoch": 1.2979683972911964,
+      "grad_norm": 1.8200581073760986,
+      "learning_rate": 4.961197661639102e-06,
+      "loss": 0.3825,
+      "mean_token_accuracy": 0.8765197157859802,
+      "num_tokens": 18677593.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.5498757302761078,
+      "epoch": 1.3007900677200903,
+      "grad_norm": 1.2150392532348633,
+      "learning_rate": 4.9610292866423036e-06,
+      "loss": 0.4201,
+      "mean_token_accuracy": 0.8701222181320191,
+      "num_tokens": 18718374.0,
+      "step": 2305
+    },
+    {
+      "entropy": 0.5348264276981354,
+      "epoch": 1.3036117381489842,
+      "grad_norm": 1.4520456790924072,
+      "learning_rate": 4.960860550957311e-06,
+      "loss": 0.4204,
+      "mean_token_accuracy": 0.8689911484718322,
+      "num_tokens": 18758689.0,
+      "step": 2310
+    },
+    {
+      "entropy": 0.564541220664978,
+      "epoch": 1.3064334085778782,
+      "grad_norm": 1.4934542179107666,
+      "learning_rate": 4.960691454617276e-06,
+      "loss": 0.444,
+      "mean_token_accuracy": 0.8587620854377747,
+      "num_tokens": 18798953.0,
+      "step": 2315
+    },
+    {
+      "entropy": 0.5457278668880463,
+      "epoch": 1.309255079006772,
+      "grad_norm": 1.3648675680160522,
+      "learning_rate": 4.960521997655415e-06,
+      "loss": 0.4228,
+      "mean_token_accuracy": 0.8687914133071899,
+      "num_tokens": 18839547.0,
+      "step": 2320
+    },
+    {
+      "entropy": 0.5353664636611939,
+      "epoch": 1.312076749435666,
+      "grad_norm": 1.3305805921554565,
+      "learning_rate": 4.960352180105019e-06,
+      "loss": 0.4185,
+      "mean_token_accuracy": 0.8669357776641846,
+      "num_tokens": 18880107.0,
+      "step": 2325
+    },
+    {
+      "entropy": 0.5464560270309449,
+      "epoch": 1.31489841986456,
+      "grad_norm": 2.0227503776550293,
+      "learning_rate": 4.9601820019994495e-06,
+      "loss": 0.4141,
+      "mean_token_accuracy": 0.8719035506248474,
+      "num_tokens": 18920937.0,
+      "step": 2330
+    },
+    {
+      "entropy": 0.536427104473114,
+      "epoch": 1.3177200902934536,
+      "grad_norm": 1.3813787698745728,
+      "learning_rate": 4.960011463372136e-06,
+      "loss": 0.4265,
+      "mean_token_accuracy": 0.8644273400306701,
+      "num_tokens": 18961776.0,
+      "step": 2335
+    },
+    {
+      "entropy": 0.5107419192790985,
+      "epoch": 1.3205417607223477,
+      "grad_norm": 1.2994340658187866,
+      "learning_rate": 4.959840564256583e-06,
+      "loss": 0.4011,
+      "mean_token_accuracy": 0.873857831954956,
+      "num_tokens": 19002350.0,
+      "step": 2340
+    },
+    {
+      "entropy": 0.48056021332740784,
+      "epoch": 1.3233634311512414,
+      "grad_norm": 1.4537101984024048,
+      "learning_rate": 4.959669304686362e-06,
+      "loss": 0.3806,
+      "mean_token_accuracy": 0.8777095079421997,
+      "num_tokens": 19042920.0,
+      "step": 2345
+    },
+    {
+      "entropy": 0.5710622429847717,
+      "epoch": 1.3261851015801354,
+      "grad_norm": 1.362627625465393,
+      "learning_rate": 4.959497684695118e-06,
+      "loss": 0.4462,
+      "mean_token_accuracy": 0.8591257691383362,
+      "num_tokens": 19083640.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.521550840139389,
+      "epoch": 1.3290067720090293,
+      "grad_norm": 1.173241138458252,
+      "learning_rate": 4.959325704316565e-06,
+      "loss": 0.4025,
+      "mean_token_accuracy": 0.8729618310928344,
+      "num_tokens": 19123082.0,
+      "step": 2355
+    },
+    {
+      "entropy": 0.5515291333198548,
+      "epoch": 1.3318284424379232,
+      "grad_norm": 1.6536093950271606,
+      "learning_rate": 4.959153363584489e-06,
+      "loss": 0.4587,
+      "mean_token_accuracy": 0.8603331565856933,
+      "num_tokens": 19163643.0,
+      "step": 2360
+    },
+    {
+      "entropy": 0.5428348422050476,
+      "epoch": 1.3346501128668171,
+      "grad_norm": 1.6122465133666992,
+      "learning_rate": 4.958980662532747e-06,
+      "loss": 0.428,
+      "mean_token_accuracy": 0.8602743864059448,
+      "num_tokens": 19204151.0,
+      "step": 2365
+    },
+    {
+      "entropy": 0.558714485168457,
+      "epoch": 1.337471783295711,
+      "grad_norm": 1.321026086807251,
+      "learning_rate": 4.9588076011952655e-06,
+      "loss": 0.4514,
+      "mean_token_accuracy": 0.8587347507476807,
+      "num_tokens": 19244654.0,
+      "step": 2370
+    },
+    {
+      "entropy": 0.5276102185249328,
+      "epoch": 1.340293453724605,
+      "grad_norm": 1.311216950416565,
+      "learning_rate": 4.958634179606041e-06,
+      "loss": 0.4122,
+      "mean_token_accuracy": 0.8710083127021789,
+      "num_tokens": 19285304.0,
+      "step": 2375
+    },
+    {
+      "entropy": 0.5120981454849243,
+      "epoch": 1.3431151241534989,
+      "grad_norm": 1.2444289922714233,
+      "learning_rate": 4.9584603977991445e-06,
+      "loss": 0.3951,
+      "mean_token_accuracy": 0.876779317855835,
+      "num_tokens": 19326058.0,
+      "step": 2380
+    },
+    {
+      "entropy": 0.5315510213375092,
+      "epoch": 1.3459367945823928,
+      "grad_norm": 1.342822551727295,
+      "learning_rate": 4.958286255808714e-06,
+      "loss": 0.4153,
+      "mean_token_accuracy": 0.86748708486557,
+      "num_tokens": 19366695.0,
+      "step": 2385
+    },
+    {
+      "entropy": 0.5051746428012848,
+      "epoch": 1.3487584650112867,
+      "grad_norm": 1.3792295455932617,
+      "learning_rate": 4.958111753668962e-06,
+      "loss": 0.4084,
+      "mean_token_accuracy": 0.8715560555458068,
+      "num_tokens": 19407356.0,
+      "step": 2390
+    },
+    {
+      "entropy": 0.5739570260047913,
+      "epoch": 1.3515801354401806,
+      "grad_norm": 1.3936057090759277,
+      "learning_rate": 4.957936891414166e-06,
+      "loss": 0.443,
+      "mean_token_accuracy": 0.8616944670677185,
+      "num_tokens": 19446924.0,
+      "step": 2395
+    },
+    {
+      "entropy": 0.5593576908111573,
+      "epoch": 1.3544018058690745,
+      "grad_norm": 1.6129295825958252,
+      "learning_rate": 4.957761669078679e-06,
+      "loss": 0.453,
+      "mean_token_accuracy": 0.8601847767829895,
+      "num_tokens": 19487296.0,
+      "step": 2400
+    },
+    {
+      "entropy": 0.49403364062309263,
+      "epoch": 1.3572234762979685,
+      "grad_norm": 1.463240623474121,
+      "learning_rate": 4.957586086696925e-06,
+      "loss": 0.3719,
+      "mean_token_accuracy": 0.882617199420929,
+      "num_tokens": 19528192.0,
+      "step": 2405
+    },
+    {
+      "entropy": 0.5413905024528504,
+      "epoch": 1.3600451467268622,
+      "grad_norm": 1.3406250476837158,
+      "learning_rate": 4.957410144303396e-06,
+      "loss": 0.4149,
+      "mean_token_accuracy": 0.8708963632583618,
+      "num_tokens": 19568788.0,
+      "step": 2410
+    },
+    {
+      "entropy": 0.5129614770412445,
+      "epoch": 1.3628668171557563,
+      "grad_norm": 1.423504114151001,
+      "learning_rate": 4.957233841932655e-06,
+      "loss": 0.3972,
+      "mean_token_accuracy": 0.8748751640319824,
+      "num_tokens": 19609525.0,
+      "step": 2415
+    },
+    {
+      "entropy": 0.49986888766288756,
+      "epoch": 1.36568848758465,
+      "grad_norm": 1.3050624132156372,
+      "learning_rate": 4.957057179619339e-06,
+      "loss": 0.3815,
+      "mean_token_accuracy": 0.8794636130332947,
+      "num_tokens": 19650378.0,
+      "step": 2420
+    },
+    {
+      "entropy": 0.49531072974205015,
+      "epoch": 1.3685101580135441,
+      "grad_norm": 1.2032006978988647,
+      "learning_rate": 4.956880157398151e-06,
+      "loss": 0.3877,
+      "mean_token_accuracy": 0.8791410803794861,
+      "num_tokens": 19690935.0,
+      "step": 2425
+    },
+    {
+      "entropy": 0.5217344462871552,
+      "epoch": 1.3713318284424378,
+      "grad_norm": 1.3181284666061401,
+      "learning_rate": 4.956702775303868e-06,
+      "loss": 0.4113,
+      "mean_token_accuracy": 0.8702506303787232,
+      "num_tokens": 19731616.0,
+      "step": 2430
+    },
+    {
+      "entropy": 0.5071918427944183,
+      "epoch": 1.3741534988713318,
+      "grad_norm": 1.2046350240707397,
+      "learning_rate": 4.956525033371336e-06,
+      "loss": 0.4106,
+      "mean_token_accuracy": 0.8688374757766724,
+      "num_tokens": 19772129.0,
+      "step": 2435
+    },
+    {
+      "entropy": 0.5620993256568909,
+      "epoch": 1.3769751693002257,
+      "grad_norm": 1.3404937982559204,
+      "learning_rate": 4.956346931635474e-06,
+      "loss": 0.4299,
+      "mean_token_accuracy": 0.8631010413169861,
+      "num_tokens": 19812671.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.5249874234199524,
+      "epoch": 1.3797968397291196,
+      "grad_norm": 1.3918310403823853,
+      "learning_rate": 4.956168470131269e-06,
+      "loss": 0.4167,
+      "mean_token_accuracy": 0.8694810628890991,
+      "num_tokens": 19853386.0,
+      "step": 2445
+    },
+    {
+      "entropy": 0.48499825596809387,
+      "epoch": 1.3826185101580135,
+      "grad_norm": 1.3398149013519287,
+      "learning_rate": 4.95598964889378e-06,
+      "loss": 0.3775,
+      "mean_token_accuracy": 0.8827056169509888,
+      "num_tokens": 19893901.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.5267378866672516,
+      "epoch": 1.3854401805869074,
+      "grad_norm": 1.4285844564437866,
+      "learning_rate": 4.9558104679581366e-06,
+      "loss": 0.4089,
+      "mean_token_accuracy": 0.8731061100959778,
+      "num_tokens": 19934618.0,
+      "step": 2455
+    },
+    {
+      "entropy": 0.551910811662674,
+      "epoch": 1.3882618510158014,
+      "grad_norm": 1.462052822113037,
+      "learning_rate": 4.955630927359538e-06,
+      "loss": 0.4621,
+      "mean_token_accuracy": 0.8564296126365661,
+      "num_tokens": 19975257.0,
+      "step": 2460
+    },
+    {
+      "entropy": 0.5045246481895447,
+      "epoch": 1.3910835214446953,
+      "grad_norm": 1.2630540132522583,
+      "learning_rate": 4.9554510271332575e-06,
+      "loss": 0.3931,
+      "mean_token_accuracy": 0.8743220806121826,
+      "num_tokens": 20015796.0,
+      "step": 2465
+    },
+    {
+      "entropy": 0.4776579737663269,
+      "epoch": 1.3939051918735892,
+      "grad_norm": 1.3086442947387695,
+      "learning_rate": 4.955270767314633e-06,
+      "loss": 0.3594,
+      "mean_token_accuracy": 0.8837788105010986,
+      "num_tokens": 20056639.0,
+      "step": 2470
+    },
+    {
+      "entropy": 0.5198698997497558,
+      "epoch": 1.396726862302483,
+      "grad_norm": 1.1976763010025024,
+      "learning_rate": 4.955090147939079e-06,
+      "loss": 0.4014,
+      "mean_token_accuracy": 0.8731474876403809,
+      "num_tokens": 20097523.0,
+      "step": 2475
+    },
+    {
+      "entropy": 0.5507951974868774,
+      "epoch": 1.399548532731377,
+      "grad_norm": 1.595848798751831,
+      "learning_rate": 4.954909169042078e-06,
+      "loss": 0.4321,
+      "mean_token_accuracy": 0.8633155941963195,
+      "num_tokens": 20137741.0,
+      "step": 2480
+    },
+    {
+      "entropy": 0.5629419386386871,
+      "epoch": 1.402370203160271,
+      "grad_norm": 1.5942708253860474,
+      "learning_rate": 4.954727830659182e-06,
+      "loss": 0.4402,
+      "mean_token_accuracy": 0.8617273569107056,
+      "num_tokens": 20178295.0,
+      "step": 2485
+    },
+    {
+      "entropy": 0.5281494557857513,
+      "epoch": 1.4051918735891649,
+      "grad_norm": 1.2493270635604858,
+      "learning_rate": 4.954546132826017e-06,
+      "loss": 0.4143,
+      "mean_token_accuracy": 0.8682329654693604,
+      "num_tokens": 20219041.0,
+      "step": 2490
+    },
+    {
+      "entropy": 0.5261754095554352,
+      "epoch": 1.4080135440180586,
+      "grad_norm": 1.3487435579299927,
+      "learning_rate": 4.954364075578276e-06,
+      "loss": 0.4072,
+      "mean_token_accuracy": 0.8690428853034973,
+      "num_tokens": 20259747.0,
+      "step": 2495
+    },
+    {
+      "entropy": 0.5704293370246887,
+      "epoch": 1.4108352144469527,
+      "grad_norm": 1.5484710931777954,
+      "learning_rate": 4.954181658951725e-06,
+      "loss": 0.4643,
+      "mean_token_accuracy": 0.8576129674911499,
+      "num_tokens": 20300626.0,
+      "step": 2500
+    },
+    {
+      "epoch": 1.4108352144469527,
+      "eval_entropy": 0.49934637546539307,
+      "eval_loss": 0.3746188282966614,
+      "eval_mean_token_accuracy": 0.8903794288635254,
+      "eval_num_tokens": 20300626.0,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 24.539,
+      "eval_steps_per_second": 6.135,
+      "step": 2500
+    },
+    {
+      "entropy": 0.5491129279136657,
+      "epoch": 1.4136568848758464,
+      "grad_norm": 1.3182957172393799,
+      "learning_rate": 4.953998882982197e-06,
+      "loss": 0.4536,
+      "mean_token_accuracy": 0.8601503491401672,
+      "num_tokens": 20341333.0,
+      "step": 2505
+    },
+    {
+      "entropy": 0.5612141251564026,
+      "epoch": 1.4164785553047405,
+      "grad_norm": 1.5513429641723633,
+      "learning_rate": 4.9538157477056025e-06,
+      "loss": 0.4414,
+      "mean_token_accuracy": 0.8637246131896973,
+      "num_tokens": 20381822.0,
+      "step": 2510
+    },
+    {
+      "entropy": 0.5199264347553253,
+      "epoch": 1.4193002257336342,
+      "grad_norm": 1.3425822257995605,
+      "learning_rate": 4.953632253157916e-06,
+      "loss": 0.4135,
+      "mean_token_accuracy": 0.8695431590080261,
+      "num_tokens": 20422346.0,
+      "step": 2515
+    },
+    {
+      "entropy": 0.5060115993022919,
+      "epoch": 1.4221218961625282,
+      "grad_norm": 1.4134669303894043,
+      "learning_rate": 4.953448399375187e-06,
+      "loss": 0.387,
+      "mean_token_accuracy": 0.8781610608100892,
+      "num_tokens": 20462097.0,
+      "step": 2520
+    },
+    {
+      "entropy": 0.5111626803874969,
+      "epoch": 1.424943566591422,
+      "grad_norm": 1.3311017751693726,
+      "learning_rate": 4.953264186393531e-06,
+      "loss": 0.3961,
+      "mean_token_accuracy": 0.87530437707901,
+      "num_tokens": 20502526.0,
+      "step": 2525
+    },
+    {
+      "entropy": 0.5413293719291687,
+      "epoch": 1.427765237020316,
+      "grad_norm": 1.4240591526031494,
+      "learning_rate": 4.953079614249138e-06,
+      "loss": 0.4341,
+      "mean_token_accuracy": 0.8609416246414184,
+      "num_tokens": 20543142.0,
+      "step": 2530
+    },
+    {
+      "entropy": 0.5732130527496337,
+      "epoch": 1.43058690744921,
+      "grad_norm": 1.4825007915496826,
+      "learning_rate": 4.952894682978268e-06,
+      "loss": 0.4597,
+      "mean_token_accuracy": 0.8564027547836304,
+      "num_tokens": 20583174.0,
+      "step": 2535
+    },
+    {
+      "entropy": 0.4671943664550781,
+      "epoch": 1.4334085778781038,
+      "grad_norm": 1.1537166833877563,
+      "learning_rate": 4.952709392617248e-06,
+      "loss": 0.3627,
+      "mean_token_accuracy": 0.8833828091621398,
+      "num_tokens": 20623760.0,
+      "step": 2540
+    },
+    {
+      "entropy": 0.5528115510940552,
+      "epoch": 1.4362302483069977,
+      "grad_norm": 1.351750135421753,
+      "learning_rate": 4.952523743202482e-06,
+      "loss": 0.4395,
+      "mean_token_accuracy": 0.8619906187057496,
+      "num_tokens": 20664513.0,
+      "step": 2545
+    },
+    {
+      "entropy": 0.5158391892910004,
+      "epoch": 1.4390519187358917,
+      "grad_norm": 1.1706582307815552,
+      "learning_rate": 4.952337734770439e-06,
+      "loss": 0.4,
+      "mean_token_accuracy": 0.8752634882926941,
+      "num_tokens": 20705333.0,
+      "step": 2550
+    },
+    {
+      "entropy": 0.5181755602359772,
+      "epoch": 1.4418735891647856,
+      "grad_norm": 1.5158717632293701,
+      "learning_rate": 4.95215136735766e-06,
+      "loss": 0.3929,
+      "mean_token_accuracy": 0.8729309439659119,
+      "num_tokens": 20745887.0,
+      "step": 2555
+    },
+    {
+      "entropy": 0.5385544419288635,
+      "epoch": 1.4446952595936795,
+      "grad_norm": 1.2974624633789062,
+      "learning_rate": 4.951964641000757e-06,
+      "loss": 0.4326,
+      "mean_token_accuracy": 0.8661515951156616,
+      "num_tokens": 20786349.0,
+      "step": 2560
+    },
+    {
+      "entropy": 0.5237244606018067,
+      "epoch": 1.4475169300225734,
+      "grad_norm": 1.4556293487548828,
+      "learning_rate": 4.951777555736414e-06,
+      "loss": 0.415,
+      "mean_token_accuracy": 0.8732018351554871,
+      "num_tokens": 20826912.0,
+      "step": 2565
+    },
+    {
+      "entropy": 0.5504439949989319,
+      "epoch": 1.4503386004514673,
+      "grad_norm": 1.478190302848816,
+      "learning_rate": 4.951590111601381e-06,
+      "loss": 0.4226,
+      "mean_token_accuracy": 0.8666379451751709,
+      "num_tokens": 20867551.0,
+      "step": 2570
+    },
+    {
+      "entropy": 0.5655958175659179,
+      "epoch": 1.4531602708803613,
+      "grad_norm": 1.2426916360855103,
+      "learning_rate": 4.951402308632485e-06,
+      "loss": 0.4511,
+      "mean_token_accuracy": 0.8618389010429383,
+      "num_tokens": 20908206.0,
+      "step": 2575
+    },
+    {
+      "entropy": 0.5350692510604859,
+      "epoch": 1.455981941309255,
+      "grad_norm": 1.355296015739441,
+      "learning_rate": 4.951214146866617e-06,
+      "loss": 0.4225,
+      "mean_token_accuracy": 0.868348228931427,
+      "num_tokens": 20948945.0,
+      "step": 2580
+    },
+    {
+      "entropy": 0.5500153779983521,
+      "epoch": 1.458803611738149,
+      "grad_norm": 1.242108941078186,
+      "learning_rate": 4.951025626340743e-06,
+      "loss": 0.4279,
+      "mean_token_accuracy": 0.864187228679657,
+      "num_tokens": 20989560.0,
+      "step": 2585
+    },
+    {
+      "entropy": 0.5728635132312775,
+      "epoch": 1.4616252821670428,
+      "grad_norm": 1.4970265626907349,
+      "learning_rate": 4.950836747091896e-06,
+      "loss": 0.4321,
+      "mean_token_accuracy": 0.8666630983352661,
+      "num_tokens": 21030293.0,
+      "step": 2590
+    },
+    {
+      "entropy": 0.572597336769104,
+      "epoch": 1.4644469525959367,
+      "grad_norm": 1.5469214916229248,
+      "learning_rate": 4.950647509157184e-06,
+      "loss": 0.4535,
+      "mean_token_accuracy": 0.860745084285736,
+      "num_tokens": 21070816.0,
+      "step": 2595
+    },
+    {
+      "entropy": 0.4957478761672974,
+      "epoch": 1.4672686230248306,
+      "grad_norm": 1.1917364597320557,
+      "learning_rate": 4.950457912573781e-06,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8750966668128968,
+      "num_tokens": 21111621.0,
+      "step": 2600
+    },
+    {
+      "entropy": 0.5851558446884155,
+      "epoch": 1.4700902934537246,
+      "grad_norm": 1.4717788696289062,
+      "learning_rate": 4.950267957378934e-06,
+      "loss": 0.4614,
+      "mean_token_accuracy": 0.8556645750999451,
+      "num_tokens": 21152027.0,
+      "step": 2605
+    },
+    {
+      "entropy": 0.4879822790622711,
+      "epoch": 1.4729119638826185,
+      "grad_norm": 1.244563341140747,
+      "learning_rate": 4.950077643609959e-06,
+      "loss": 0.3706,
+      "mean_token_accuracy": 0.882515013217926,
+      "num_tokens": 21192521.0,
+      "step": 2610
+    },
+    {
+      "entropy": 0.5324353992938995,
+      "epoch": 1.4757336343115124,
+      "grad_norm": 1.4513381719589233,
+      "learning_rate": 4.949886971304245e-06,
+      "loss": 0.4337,
+      "mean_token_accuracy": 0.8638843178749085,
+      "num_tokens": 21233396.0,
+      "step": 2615
+    },
+    {
+      "entropy": 0.5465445816516876,
+      "epoch": 1.4785553047404063,
+      "grad_norm": 1.4188737869262695,
+      "learning_rate": 4.9496959404992475e-06,
+      "loss": 0.4304,
+      "mean_token_accuracy": 0.8674513459205627,
+      "num_tokens": 21274180.0,
+      "step": 2620
+    },
+    {
+      "entropy": 0.5312675893306732,
+      "epoch": 1.4813769751693002,
+      "grad_norm": 1.3827718496322632,
+      "learning_rate": 4.949504551232494e-06,
+      "loss": 0.424,
+      "mean_token_accuracy": 0.8689416646957397,
+      "num_tokens": 21314947.0,
+      "step": 2625
+    },
+    {
+      "entropy": 0.5221801578998566,
+      "epoch": 1.4841986455981941,
+      "grad_norm": 1.344963788986206,
+      "learning_rate": 4.949312803541586e-06,
+      "loss": 0.3887,
+      "mean_token_accuracy": 0.8764538764953613,
+      "num_tokens": 21355658.0,
+      "step": 2630
+    },
+    {
+      "entropy": 0.5304463326930999,
+      "epoch": 1.487020316027088,
+      "grad_norm": 1.5296121835708618,
+      "learning_rate": 4.94912069746419e-06,
+      "loss": 0.4037,
+      "mean_token_accuracy": 0.8726909160614014,
+      "num_tokens": 21396355.0,
+      "step": 2635
+    },
+    {
+      "entropy": 0.5112833499908447,
+      "epoch": 1.489841986455982,
+      "grad_norm": 1.4764835834503174,
+      "learning_rate": 4.948928233038046e-06,
+      "loss": 0.3963,
+      "mean_token_accuracy": 0.8752380728721618,
+      "num_tokens": 21437172.0,
+      "step": 2640
+    },
+    {
+      "entropy": 0.5136803925037384,
+      "epoch": 1.492663656884876,
+      "grad_norm": 1.5744765996932983,
+      "learning_rate": 4.948735410300964e-06,
+      "loss": 0.4151,
+      "mean_token_accuracy": 0.8673662304878235,
+      "num_tokens": 21477969.0,
+      "step": 2645
+    },
+    {
+      "entropy": 0.5488377809524536,
+      "epoch": 1.4954853273137698,
+      "grad_norm": 1.3822392225265503,
+      "learning_rate": 4.948542229290823e-06,
+      "loss": 0.4442,
+      "mean_token_accuracy": 0.8641608357429504,
+      "num_tokens": 21517644.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.48912598490715026,
+      "epoch": 1.4983069977426637,
+      "grad_norm": 1.3357199430465698,
+      "learning_rate": 4.948348690045574e-06,
+      "loss": 0.3948,
+      "mean_token_accuracy": 0.873920488357544,
+      "num_tokens": 21558283.0,
+      "step": 2655
+    },
+    {
+      "entropy": 0.5133354067802429,
+      "epoch": 1.5011286681715577,
+      "grad_norm": 1.3202459812164307,
+      "learning_rate": 4.948154792603237e-06,
+      "loss": 0.414,
+      "mean_token_accuracy": 0.8687997817993164,
+      "num_tokens": 21598905.0,
+      "step": 2660
+    },
+    {
+      "entropy": 0.5592216491699219,
+      "epoch": 1.5039503386004514,
+      "grad_norm": 1.515934944152832,
+      "learning_rate": 4.947960537001905e-06,
+      "loss": 0.4425,
+      "mean_token_accuracy": 0.8624489784240723,
+      "num_tokens": 21639436.0,
+      "step": 2665
+    },
+    {
+      "entropy": 0.5013287782669067,
+      "epoch": 1.5067720090293455,
+      "grad_norm": 1.3164470195770264,
+      "learning_rate": 4.947765923279738e-06,
+      "loss": 0.3836,
+      "mean_token_accuracy": 0.8810773968696595,
+      "num_tokens": 21680256.0,
+      "step": 2670
+    },
+    {
+      "entropy": 0.5708201169967652,
+      "epoch": 1.5095936794582392,
+      "grad_norm": 1.5717384815216064,
+      "learning_rate": 4.9475709514749695e-06,
+      "loss": 0.4513,
+      "mean_token_accuracy": 0.8578357696533203,
+      "num_tokens": 21720710.0,
+      "step": 2675
+    },
+    {
+      "entropy": 0.5323946833610534,
+      "epoch": 1.5124153498871333,
+      "grad_norm": 1.3646914958953857,
+      "learning_rate": 4.9473756216258996e-06,
+      "loss": 0.4152,
+      "mean_token_accuracy": 0.8670541882514954,
+      "num_tokens": 21761305.0,
+      "step": 2680
+    },
+    {
+      "entropy": 0.5191471993923187,
+      "epoch": 1.515237020316027,
+      "grad_norm": 1.259047508239746,
+      "learning_rate": 4.947179933770902e-06,
+      "loss": 0.397,
+      "mean_token_accuracy": 0.8745760560035706,
+      "num_tokens": 21801879.0,
+      "step": 2685
+    },
+    {
+      "entropy": 0.5398116707801819,
+      "epoch": 1.518058690744921,
+      "grad_norm": 1.3516515493392944,
+      "learning_rate": 4.94698388794842e-06,
+      "loss": 0.4344,
+      "mean_token_accuracy": 0.8646332740783691,
+      "num_tokens": 21842560.0,
+      "step": 2690
+    },
+    {
+      "entropy": 0.5416356801986695,
+      "epoch": 1.5208803611738149,
+      "grad_norm": 1.3463817834854126,
+      "learning_rate": 4.946787484196966e-06,
+      "loss": 0.4341,
+      "mean_token_accuracy": 0.8608946084976197,
+      "num_tokens": 21883327.0,
+      "step": 2695
+    },
+    {
+      "entropy": 0.555895859003067,
+      "epoch": 1.5237020316027088,
+      "grad_norm": 1.376157283782959,
+      "learning_rate": 4.9465907225551244e-06,
+      "loss": 0.4479,
+      "mean_token_accuracy": 0.8606306076049804,
+      "num_tokens": 21923780.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.5254535734653473,
+      "epoch": 1.5265237020316027,
+      "grad_norm": 1.310943365097046,
+      "learning_rate": 4.946393603061548e-06,
+      "loss": 0.4071,
+      "mean_token_accuracy": 0.8713202953338623,
+      "num_tokens": 21964188.0,
+      "step": 2705
+    },
+    {
+      "entropy": 0.5179416954517364,
+      "epoch": 1.5293453724604966,
+      "grad_norm": 1.361799955368042,
+      "learning_rate": 4.946196125754962e-06,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8728214979171753,
+      "num_tokens": 22004992.0,
+      "step": 2710
+    },
+    {
+      "entropy": 0.539003986120224,
+      "epoch": 1.5321670428893905,
+      "grad_norm": 1.5339001417160034,
+      "learning_rate": 4.9459982906741596e-06,
+      "loss": 0.4343,
+      "mean_token_accuracy": 0.8599718093872071,
+      "num_tokens": 22045651.0,
+      "step": 2715
+    },
+    {
+      "entropy": 0.48843835592269896,
+      "epoch": 1.5349887133182845,
+      "grad_norm": 1.1609723567962646,
+      "learning_rate": 4.945800097858007e-06,
+      "loss": 0.3754,
+      "mean_token_accuracy": 0.8784770965576172,
+      "num_tokens": 22085915.0,
+      "step": 2720
+    },
+    {
+      "entropy": 0.5748872756958008,
+      "epoch": 1.5378103837471784,
+      "grad_norm": 1.474261999130249,
+      "learning_rate": 4.945601547345439e-06,
+      "loss": 0.4581,
+      "mean_token_accuracy": 0.8565932989120484,
+      "num_tokens": 22126565.0,
+      "step": 2725
+    },
+    {
+      "entropy": 0.5497707307338715,
+      "epoch": 1.540632054176072,
+      "grad_norm": 1.328251600265503,
+      "learning_rate": 4.945402639175459e-06,
+      "loss": 0.4229,
+      "mean_token_accuracy": 0.8657533526420593,
+      "num_tokens": 22166897.0,
+      "step": 2730
+    },
+    {
+      "entropy": 0.5045753955841065,
+      "epoch": 1.5434537246049662,
+      "grad_norm": 1.4237067699432373,
+      "learning_rate": 4.945203373387145e-06,
+      "loss": 0.4037,
+      "mean_token_accuracy": 0.8722402095794678,
+      "num_tokens": 22207560.0,
+      "step": 2735
+    },
+    {
+      "entropy": 0.5575098991394043,
+      "epoch": 1.54627539503386,
+      "grad_norm": 1.7878057956695557,
+      "learning_rate": 4.945003750019641e-06,
+      "loss": 0.4381,
+      "mean_token_accuracy": 0.8603370666503907,
+      "num_tokens": 22247941.0,
+      "step": 2740
+    },
+    {
+      "entropy": 0.4867607891559601,
+      "epoch": 1.549097065462754,
+      "grad_norm": 1.526099443435669,
+      "learning_rate": 4.944803769112164e-06,
+      "loss": 0.401,
+      "mean_token_accuracy": 0.874083411693573,
+      "num_tokens": 22288838.0,
+      "step": 2745
+    },
+    {
+      "entropy": 0.5552621841430664,
+      "epoch": 1.5519187358916477,
+      "grad_norm": 1.3902403116226196,
+      "learning_rate": 4.944603430704e-06,
+      "loss": 0.4259,
+      "mean_token_accuracy": 0.8668664813041687,
+      "num_tokens": 22328956.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.5252302527427674,
+      "epoch": 1.554740406320542,
+      "grad_norm": 1.326379418373108,
+      "learning_rate": 4.944402734834506e-06,
+      "loss": 0.4095,
+      "mean_token_accuracy": 0.8710361838340759,
+      "num_tokens": 22369592.0,
+      "step": 2755
+    },
+    {
+      "entropy": 0.5309755086898804,
+      "epoch": 1.5575620767494356,
+      "grad_norm": 1.4432721138000488,
+      "learning_rate": 4.944201681543107e-06,
+      "loss": 0.423,
+      "mean_token_accuracy": 0.8664063453674317,
+      "num_tokens": 22410109.0,
+      "step": 2760
+    },
+    {
+      "entropy": 0.5249135076999665,
+      "epoch": 1.5603837471783297,
+      "grad_norm": 1.4632095098495483,
+      "learning_rate": 4.944000270869302e-06,
+      "loss": 0.3937,
+      "mean_token_accuracy": 0.8737531065940857,
+      "num_tokens": 22450758.0,
+      "step": 2765
+    },
+    {
+      "entropy": 0.509848439693451,
+      "epoch": 1.5632054176072234,
+      "grad_norm": 1.4338641166687012,
+      "learning_rate": 4.943798502852657e-06,
+      "loss": 0.4024,
+      "mean_token_accuracy": 0.8742005109786988,
+      "num_tokens": 22491593.0,
+      "step": 2770
+    },
+    {
+      "entropy": 0.5258462011814118,
+      "epoch": 1.5660270880361173,
+      "grad_norm": 1.3523592948913574,
+      "learning_rate": 4.94359637753281e-06,
+      "loss": 0.4196,
+      "mean_token_accuracy": 0.867405092716217,
+      "num_tokens": 22532315.0,
+      "step": 2775
+    },
+    {
+      "entropy": 0.5412358582019806,
+      "epoch": 1.5688487584650113,
+      "grad_norm": 1.4376283884048462,
+      "learning_rate": 4.943393894949469e-06,
+      "loss": 0.42,
+      "mean_token_accuracy": 0.8682662725448609,
+      "num_tokens": 22573023.0,
+      "step": 2780
+    },
+    {
+      "entropy": 0.530605137348175,
+      "epoch": 1.5716704288939052,
+      "grad_norm": 1.3493672609329224,
+      "learning_rate": 4.943191055142409e-06,
+      "loss": 0.4154,
+      "mean_token_accuracy": 0.8707961201667785,
+      "num_tokens": 22613743.0,
+      "step": 2785
+    },
+    {
+      "entropy": 0.5197601974010467,
+      "epoch": 1.574492099322799,
+      "grad_norm": 1.3924630880355835,
+      "learning_rate": 4.942987858151481e-06,
+      "loss": 0.4109,
+      "mean_token_accuracy": 0.8706324815750122,
+      "num_tokens": 22654339.0,
+      "step": 2790
+    },
+    {
+      "entropy": 0.48147222995758054,
+      "epoch": 1.577313769751693,
+      "grad_norm": 1.5284984111785889,
+      "learning_rate": 4.942784304016602e-06,
+      "loss": 0.3804,
+      "mean_token_accuracy": 0.8784740090370178,
+      "num_tokens": 22694670.0,
+      "step": 2795
+    },
+    {
+      "entropy": 0.46955564618110657,
+      "epoch": 1.580135440180587,
+      "grad_norm": 1.6057965755462646,
+      "learning_rate": 4.942580392777761e-06,
+      "loss": 0.3786,
+      "mean_token_accuracy": 0.8798289656639099,
+      "num_tokens": 22735192.0,
+      "step": 2800
+    },
+    {
+      "entropy": 0.5634661912918091,
+      "epoch": 1.5829571106094809,
+      "grad_norm": 1.395089030265808,
+      "learning_rate": 4.942376124475014e-06,
+      "loss": 0.4528,
+      "mean_token_accuracy": 0.8594593167304992,
+      "num_tokens": 22775738.0,
+      "step": 2805
+    },
+    {
+      "entropy": 0.6147814273834229,
+      "epoch": 1.5857787810383748,
+      "grad_norm": 1.4737554788589478,
+      "learning_rate": 4.942171499148492e-06,
+      "loss": 0.4841,
+      "mean_token_accuracy": 0.851655924320221,
+      "num_tokens": 22816401.0,
+      "step": 2810
+    },
+    {
+      "entropy": 0.581537914276123,
+      "epoch": 1.5886004514672685,
+      "grad_norm": 1.4444756507873535,
+      "learning_rate": 4.941966516838393e-06,
+      "loss": 0.4358,
+      "mean_token_accuracy": 0.8650146484375,
+      "num_tokens": 22857040.0,
+      "step": 2815
+    },
+    {
+      "entropy": 0.5177268862724305,
+      "epoch": 1.5914221218961626,
+      "grad_norm": 1.441490650177002,
+      "learning_rate": 4.941761177584985e-06,
+      "loss": 0.4075,
+      "mean_token_accuracy": 0.8717929363250733,
+      "num_tokens": 22897722.0,
+      "step": 2820
+    },
+    {
+      "entropy": 0.48996912837028506,
+      "epoch": 1.5942437923250563,
+      "grad_norm": 1.5225064754486084,
+      "learning_rate": 4.941555481428607e-06,
+      "loss": 0.3846,
+      "mean_token_accuracy": 0.8782232880592347,
+      "num_tokens": 22938229.0,
+      "step": 2825
+    },
+    {
+      "entropy": 0.5194081962108612,
+      "epoch": 1.5970654627539504,
+      "grad_norm": 1.3030446767807007,
+      "learning_rate": 4.94134942840967e-06,
+      "loss": 0.3979,
+      "mean_token_accuracy": 0.8713892698287964,
+      "num_tokens": 22978751.0,
+      "step": 2830
+    },
+    {
+      "entropy": 0.5596107244491577,
+      "epoch": 1.5998871331828441,
+      "grad_norm": 1.3859238624572754,
+      "learning_rate": 4.9411430185686505e-06,
+      "loss": 0.4545,
+      "mean_token_accuracy": 0.8581524729728699,
+      "num_tokens": 23019325.0,
+      "step": 2835
+    },
+    {
+      "entropy": 0.5840788006782531,
+      "epoch": 1.6027088036117383,
+      "grad_norm": 1.4645488262176514,
+      "learning_rate": 4.940936251946099e-06,
+      "loss": 0.466,
+      "mean_token_accuracy": 0.8551864385604858,
+      "num_tokens": 23060114.0,
+      "step": 2840
+    },
+    {
+      "entropy": 0.5338825523853302,
+      "epoch": 1.605530474040632,
+      "grad_norm": 1.3881251811981201,
+      "learning_rate": 4.940729128582636e-06,
+      "loss": 0.4135,
+      "mean_token_accuracy": 0.8701825737953186,
+      "num_tokens": 23100863.0,
+      "step": 2845
+    },
+    {
+      "entropy": 0.5128993272781373,
+      "epoch": 1.6083521444695261,
+      "grad_norm": 1.1398290395736694,
+      "learning_rate": 4.940521648518948e-06,
+      "loss": 0.3846,
+      "mean_token_accuracy": 0.8785142302513123,
+      "num_tokens": 23141333.0,
+      "step": 2850
+    },
+    {
+      "entropy": 0.5374301552772522,
+      "epoch": 1.6111738148984198,
+      "grad_norm": 1.459583044052124,
+      "learning_rate": 4.940313811795797e-06,
+      "loss": 0.4169,
+      "mean_token_accuracy": 0.8682045102119446,
+      "num_tokens": 23181557.0,
+      "step": 2855
+    },
+    {
+      "entropy": 0.530595988035202,
+      "epoch": 1.6139954853273137,
+      "grad_norm": 1.451833963394165,
+      "learning_rate": 4.9401056184540115e-06,
+      "loss": 0.4232,
+      "mean_token_accuracy": 0.8683579087257385,
+      "num_tokens": 23221979.0,
+      "step": 2860
+    },
+    {
+      "entropy": 0.5528306722640991,
+      "epoch": 1.6168171557562077,
+      "grad_norm": 1.3938990831375122,
+      "learning_rate": 4.939897068534491e-06,
+      "loss": 0.4351,
+      "mean_token_accuracy": 0.8650489091873169,
+      "num_tokens": 23262727.0,
+      "step": 2865
+    },
+    {
+      "entropy": 0.4935439705848694,
+      "epoch": 1.6196388261851016,
+      "grad_norm": 1.3947253227233887,
+      "learning_rate": 4.939688162078205e-06,
+      "loss": 0.3791,
+      "mean_token_accuracy": 0.8790611743927002,
+      "num_tokens": 23303214.0,
+      "step": 2870
+    },
+    {
+      "entropy": 0.5516513884067535,
+      "epoch": 1.6224604966139955,
+      "grad_norm": 1.4023181200027466,
+      "learning_rate": 4.939478899126196e-06,
+      "loss": 0.4402,
+      "mean_token_accuracy": 0.8614558100700378,
+      "num_tokens": 23343833.0,
+      "step": 2875
+    },
+    {
+      "entropy": 0.5269099056720734,
+      "epoch": 1.6252821670428894,
+      "grad_norm": 1.520416021347046,
+      "learning_rate": 4.939269279719569e-06,
+      "loss": 0.4226,
+      "mean_token_accuracy": 0.868548309803009,
+      "num_tokens": 23384399.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.5381790697574615,
+      "epoch": 1.6281038374717833,
+      "grad_norm": 1.252475380897522,
+      "learning_rate": 4.939059303899507e-06,
+      "loss": 0.4197,
+      "mean_token_accuracy": 0.8675281643867493,
+      "num_tokens": 23424933.0,
+      "step": 2885
+    },
+    {
+      "entropy": 0.5297176659107208,
+      "epoch": 1.6309255079006773,
+      "grad_norm": 1.3621821403503418,
+      "learning_rate": 4.93884897170726e-06,
+      "loss": 0.4147,
+      "mean_token_accuracy": 0.8692261695861816,
+      "num_tokens": 23465388.0,
+      "step": 2890
+    },
+    {
+      "entropy": 0.5208603262901306,
+      "epoch": 1.6337471783295712,
+      "grad_norm": 1.6305574178695679,
+      "learning_rate": 4.9386382831841455e-06,
+      "loss": 0.3942,
+      "mean_token_accuracy": 0.8741989612579346,
+      "num_tokens": 23506128.0,
+      "step": 2895
+    },
+    {
+      "entropy": 0.5832960605621338,
+      "epoch": 1.6365688487584649,
+      "grad_norm": 1.3824501037597656,
+      "learning_rate": 4.9384272383715535e-06,
+      "loss": 0.4412,
+      "mean_token_accuracy": 0.8618869662284852,
+      "num_tokens": 23546678.0,
+      "step": 2900
+    },
+    {
+      "entropy": 0.5429921388626099,
+      "epoch": 1.639390519187359,
+      "grad_norm": 1.4333373308181763,
+      "learning_rate": 4.938215837310947e-06,
+      "loss": 0.4234,
+      "mean_token_accuracy": 0.8673049211502075,
+      "num_tokens": 23587532.0,
+      "step": 2905
+    },
+    {
+      "entropy": 0.4961793005466461,
+      "epoch": 1.6422121896162527,
+      "grad_norm": 1.4071818590164185,
+      "learning_rate": 4.938004080043852e-06,
+      "loss": 0.3931,
+      "mean_token_accuracy": 0.8742106318473816,
+      "num_tokens": 23628159.0,
+      "step": 2910
+    },
+    {
+      "entropy": 0.5959875106811523,
+      "epoch": 1.6450338600451468,
+      "grad_norm": 1.366908073425293,
+      "learning_rate": 4.93779196661187e-06,
+      "loss": 0.4724,
+      "mean_token_accuracy": 0.8543134689331054,
+      "num_tokens": 23668932.0,
+      "step": 2915
+    },
+    {
+      "entropy": 0.48293390274047854,
+      "epoch": 1.6478555304740405,
+      "grad_norm": 1.2694140672683716,
+      "learning_rate": 4.937579497056671e-06,
+      "loss": 0.3597,
+      "mean_token_accuracy": 0.8866869568824768,
+      "num_tokens": 23709665.0,
+      "step": 2920
+    },
+    {
+      "entropy": 0.5429055094718933,
+      "epoch": 1.6506772009029347,
+      "grad_norm": 1.4968737363815308,
+      "learning_rate": 4.937366671419994e-06,
+      "loss": 0.4479,
+      "mean_token_accuracy": 0.8604987978935241,
+      "num_tokens": 23750495.0,
+      "step": 2925
+    },
+    {
+      "entropy": 0.49238319993019103,
+      "epoch": 1.6534988713318284,
+      "grad_norm": 1.4988641738891602,
+      "learning_rate": 4.937153489743649e-06,
+      "loss": 0.3955,
+      "mean_token_accuracy": 0.875276243686676,
+      "num_tokens": 23790793.0,
+      "step": 2930
+    },
+    {
+      "entropy": 0.5414620280265808,
+      "epoch": 1.6563205417607223,
+      "grad_norm": 1.4088878631591797,
+      "learning_rate": 4.936939952069515e-06,
+      "loss": 0.4192,
+      "mean_token_accuracy": 0.8669666767120361,
+      "num_tokens": 23831543.0,
+      "step": 2935
+    },
+    {
+      "entropy": 0.5236553966999054,
+      "epoch": 1.6591422121896162,
+      "grad_norm": 1.4377782344818115,
+      "learning_rate": 4.936726058439542e-06,
+      "loss": 0.4219,
+      "mean_token_accuracy": 0.8676124572753906,
+      "num_tokens": 23872266.0,
+      "step": 2940
+    },
+    {
+      "entropy": 0.5415829122066498,
+      "epoch": 1.6619638826185101,
+      "grad_norm": 1.4951739311218262,
+      "learning_rate": 4.936511808895751e-06,
+      "loss": 0.4225,
+      "mean_token_accuracy": 0.865878450870514,
+      "num_tokens": 23912731.0,
+      "step": 2945
+    },
+    {
+      "entropy": 0.5576195478439331,
+      "epoch": 1.664785553047404,
+      "grad_norm": 1.3496601581573486,
+      "learning_rate": 4.936297203480227e-06,
+      "loss": 0.4244,
+      "mean_token_accuracy": 0.8663969993591308,
+      "num_tokens": 23953363.0,
+      "step": 2950
+    },
+    {
+      "entropy": 0.5511267483234406,
+      "epoch": 1.667607223476298,
+      "grad_norm": 1.3088607788085938,
+      "learning_rate": 4.936082242235133e-06,
+      "loss": 0.4336,
+      "mean_token_accuracy": 0.8627111554145813,
+      "num_tokens": 23994093.0,
+      "step": 2955
+    },
+    {
+      "entropy": 0.5174029409885407,
+      "epoch": 1.670428893905192,
+      "grad_norm": 1.3591479063034058,
+      "learning_rate": 4.935866925202697e-06,
+      "loss": 0.3993,
+      "mean_token_accuracy": 0.8729613065719605,
+      "num_tokens": 24034859.0,
+      "step": 2960
+    },
+    {
+      "entropy": 0.5546496689319611,
+      "epoch": 1.6732505643340858,
+      "grad_norm": 1.4932035207748413,
+      "learning_rate": 4.935651252425219e-06,
+      "loss": 0.4278,
+      "mean_token_accuracy": 0.8636979460716248,
+      "num_tokens": 24075192.0,
+      "step": 2965
+    },
+    {
+      "entropy": 0.585486912727356,
+      "epoch": 1.6760722347629797,
+      "grad_norm": 1.5926809310913086,
+      "learning_rate": 4.935435223945066e-06,
+      "loss": 0.4578,
+      "mean_token_accuracy": 0.8578457713127137,
+      "num_tokens": 24115948.0,
+      "step": 2970
+    },
+    {
+      "entropy": 0.6288502216339111,
+      "epoch": 1.6788939051918734,
+      "grad_norm": 1.4562243223190308,
+      "learning_rate": 4.935218839804678e-06,
+      "loss": 0.5238,
+      "mean_token_accuracy": 0.8391013741493225,
+      "num_tokens": 24156616.0,
+      "step": 2975
+    },
+    {
+      "entropy": 0.518006545305252,
+      "epoch": 1.6817155756207676,
+      "grad_norm": 1.3895467519760132,
+      "learning_rate": 4.9350021000465645e-06,
+      "loss": 0.3946,
+      "mean_token_accuracy": 0.8746880412101745,
+      "num_tokens": 24196970.0,
+      "step": 2980
+    },
+    {
+      "entropy": 0.5556329131126404,
+      "epoch": 1.6845372460496613,
+      "grad_norm": 1.4481979608535767,
+      "learning_rate": 4.9347850047133025e-06,
+      "loss": 0.4324,
+      "mean_token_accuracy": 0.8628229856491089,
+      "num_tokens": 24237625.0,
+      "step": 2985
+    },
+    {
+      "entropy": 0.557821786403656,
+      "epoch": 1.6873589164785554,
+      "grad_norm": 1.4519360065460205,
+      "learning_rate": 4.934567553847541e-06,
+      "loss": 0.4377,
+      "mean_token_accuracy": 0.8617467164993287,
+      "num_tokens": 24278002.0,
+      "step": 2990
+    },
+    {
+      "entropy": 0.4816114008426666,
+      "epoch": 1.690180586907449,
+      "grad_norm": 1.354487657546997,
+      "learning_rate": 4.934349747491998e-06,
+      "loss": 0.3668,
+      "mean_token_accuracy": 0.883508825302124,
+      "num_tokens": 24318792.0,
+      "step": 2995
+    },
+    {
+      "entropy": 0.5235770165920257,
+      "epoch": 1.6930022573363432,
+      "grad_norm": 1.4343819618225098,
+      "learning_rate": 4.934131585689462e-06,
+      "loss": 0.4065,
+      "mean_token_accuracy": 0.8717373967170715,
+      "num_tokens": 24359410.0,
+      "step": 3000
+    },
+    {
+      "epoch": 1.6930022573363432,
+      "eval_entropy": 0.49325600266456604,
+      "eval_loss": 0.33416908979415894,
+      "eval_mean_token_accuracy": 0.8984284996986389,
+      "eval_num_tokens": 24359410.0,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 24.448,
+      "eval_steps_per_second": 6.112,
+      "step": 3000
+    },
+    {
+      "entropy": 0.5068280339241028,
+      "epoch": 1.695823927765237,
+      "grad_norm": 1.4546306133270264,
+      "learning_rate": 4.933913068482792e-06,
+      "loss": 0.4058,
+      "mean_token_accuracy": 0.8731536269187927,
+      "num_tokens": 24400072.0,
+      "step": 3005
+    },
+    {
+      "entropy": 0.5603242635726928,
+      "epoch": 1.698645598194131,
+      "grad_norm": 1.3821967840194702,
+      "learning_rate": 4.933694195914913e-06,
+      "loss": 0.4325,
+      "mean_token_accuracy": 0.8643564581871033,
+      "num_tokens": 24440775.0,
+      "step": 3010
+    },
+    {
+      "entropy": 0.5545867264270783,
+      "epoch": 1.7014672686230248,
+      "grad_norm": 1.3357994556427002,
+      "learning_rate": 4.9334749680288255e-06,
+      "loss": 0.4158,
+      "mean_token_accuracy": 0.8690497159957886,
+      "num_tokens": 24481543.0,
+      "step": 3015
+    },
+    {
+      "entropy": 0.5338212966918945,
+      "epoch": 1.7042889390519187,
+      "grad_norm": 1.5392085313796997,
+      "learning_rate": 4.9332553848675945e-06,
+      "loss": 0.4163,
+      "mean_token_accuracy": 0.867198383808136,
+      "num_tokens": 24522074.0,
+      "step": 3020
+    },
+    {
+      "entropy": 0.5320624589920044,
+      "epoch": 1.7071106094808126,
+      "grad_norm": 1.3222757577896118,
+      "learning_rate": 4.933035446474358e-06,
+      "loss": 0.4333,
+      "mean_token_accuracy": 0.8634262561798096,
+      "num_tokens": 24562720.0,
+      "step": 3025
+    },
+    {
+      "entropy": 0.5418852508068085,
+      "epoch": 1.7099322799097065,
+      "grad_norm": 1.3481212854385376,
+      "learning_rate": 4.932815152892323e-06,
+      "loss": 0.4225,
+      "mean_token_accuracy": 0.8684999108314514,
+      "num_tokens": 24603277.0,
+      "step": 3030
+    },
+    {
+      "entropy": 0.5434010863304138,
+      "epoch": 1.7127539503386005,
+      "grad_norm": 1.4969969987869263,
+      "learning_rate": 4.932594504164767e-06,
+      "loss": 0.4051,
+      "mean_token_accuracy": 0.8709228515625,
+      "num_tokens": 24643930.0,
+      "step": 3035
+    },
+    {
+      "entropy": 0.4718973636627197,
+      "epoch": 1.7155756207674944,
+      "grad_norm": 1.2923210859298706,
+      "learning_rate": 4.932373500335035e-06,
+      "loss": 0.3696,
+      "mean_token_accuracy": 0.8835036158561707,
+      "num_tokens": 24684556.0,
+      "step": 3040
+    },
+    {
+      "entropy": 0.5055824279785156,
+      "epoch": 1.7183972911963883,
+      "grad_norm": 1.2986489534378052,
+      "learning_rate": 4.932152141446545e-06,
+      "loss": 0.3948,
+      "mean_token_accuracy": 0.8733231663703919,
+      "num_tokens": 24725118.0,
+      "step": 3045
+    },
+    {
+      "entropy": 0.5701562583446502,
+      "epoch": 1.7212189616252822,
+      "grad_norm": 1.3613872528076172,
+      "learning_rate": 4.93193042754278e-06,
+      "loss": 0.4687,
+      "mean_token_accuracy": 0.8567159414291382,
+      "num_tokens": 24765797.0,
+      "step": 3050
+    },
+    {
+      "entropy": 0.5157757997512817,
+      "epoch": 1.7240406320541761,
+      "grad_norm": 1.4309362173080444,
+      "learning_rate": 4.931708358667299e-06,
+      "loss": 0.4019,
+      "mean_token_accuracy": 0.8720048069953918,
+      "num_tokens": 24806387.0,
+      "step": 3055
+    },
+    {
+      "entropy": 0.5428893506526947,
+      "epoch": 1.7268623024830698,
+      "grad_norm": 1.565901756286621,
+      "learning_rate": 4.9314859348637256e-06,
+      "loss": 0.4145,
+      "mean_token_accuracy": 0.8697570323944092,
+      "num_tokens": 24846979.0,
+      "step": 3060
+    },
+    {
+      "entropy": 0.5166443586349487,
+      "epoch": 1.729683972911964,
+      "grad_norm": 1.3636257648468018,
+      "learning_rate": 4.931263156175756e-06,
+      "loss": 0.4157,
+      "mean_token_accuracy": 0.8698770403862,
+      "num_tokens": 24887888.0,
+      "step": 3065
+    },
+    {
+      "entropy": 0.5039295554161072,
+      "epoch": 1.7325056433408577,
+      "grad_norm": 1.1980102062225342,
+      "learning_rate": 4.931040022647154e-06,
+      "loss": 0.3905,
+      "mean_token_accuracy": 0.8756434202194214,
+      "num_tokens": 24928370.0,
+      "step": 3070
+    },
+    {
+      "entropy": 0.510296243429184,
+      "epoch": 1.7353273137697518,
+      "grad_norm": 1.393905758857727,
+      "learning_rate": 4.930816534321755e-06,
+      "loss": 0.3965,
+      "mean_token_accuracy": 0.8748240351676941,
+      "num_tokens": 24969015.0,
+      "step": 3075
+    },
+    {
+      "entropy": 0.4927513003349304,
+      "epoch": 1.7381489841986455,
+      "grad_norm": 1.4032167196273804,
+      "learning_rate": 4.930592691243463e-06,
+      "loss": 0.3913,
+      "mean_token_accuracy": 0.8734770774841308,
+      "num_tokens": 25009646.0,
+      "step": 3080
+    },
+    {
+      "entropy": 0.5289837539196014,
+      "epoch": 1.7409706546275396,
+      "grad_norm": 1.3683279752731323,
+      "learning_rate": 4.930368493456252e-06,
+      "loss": 0.4062,
+      "mean_token_accuracy": 0.8745390772819519,
+      "num_tokens": 25050335.0,
+      "step": 3085
+    },
+    {
+      "entropy": 0.5285290896892547,
+      "epoch": 1.7437923250564333,
+      "grad_norm": 1.4092698097229004,
+      "learning_rate": 4.930143941004166e-06,
+      "loss": 0.4262,
+      "mean_token_accuracy": 0.8690651416778564,
+      "num_tokens": 25090875.0,
+      "step": 3090
+    },
+    {
+      "entropy": 0.5130545377731324,
+      "epoch": 1.7466139954853275,
+      "grad_norm": 1.300351619720459,
+      "learning_rate": 4.9299190339313186e-06,
+      "loss": 0.3881,
+      "mean_token_accuracy": 0.8777357459068298,
+      "num_tokens": 25131068.0,
+      "step": 3095
+    },
+    {
+      "entropy": 0.5526080071926117,
+      "epoch": 1.7494356659142212,
+      "grad_norm": 1.5255002975463867,
+      "learning_rate": 4.929693772281892e-06,
+      "loss": 0.4491,
+      "mean_token_accuracy": 0.859094786643982,
+      "num_tokens": 25171825.0,
+      "step": 3100
+    },
+    {
+      "entropy": 0.554760754108429,
+      "epoch": 1.752257336343115,
+      "grad_norm": 1.6149297952651978,
+      "learning_rate": 4.929468156100139e-06,
+      "loss": 0.4283,
+      "mean_token_accuracy": 0.8624596238136292,
+      "num_tokens": 25212450.0,
+      "step": 3105
+    },
+    {
+      "entropy": 0.5364702701568603,
+      "epoch": 1.755079006772009,
+      "grad_norm": 1.3212907314300537,
+      "learning_rate": 4.929242185430382e-06,
+      "loss": 0.4295,
+      "mean_token_accuracy": 0.8656898379325867,
+      "num_tokens": 25252979.0,
+      "step": 3110
+    },
+    {
+      "entropy": 0.5041682958602905,
+      "epoch": 1.757900677200903,
+      "grad_norm": 1.4357374906539917,
+      "learning_rate": 4.929015860317013e-06,
+      "loss": 0.3979,
+      "mean_token_accuracy": 0.8751791954040528,
+      "num_tokens": 25293782.0,
+      "step": 3115
+    },
+    {
+      "entropy": 0.5505750179290771,
+      "epoch": 1.7607223476297968,
+      "grad_norm": 1.3972926139831543,
+      "learning_rate": 4.928789180804494e-06,
+      "loss": 0.4544,
+      "mean_token_accuracy": 0.8585454225540161,
+      "num_tokens": 25334421.0,
+      "step": 3120
+    },
+    {
+      "entropy": 0.5151427447795868,
+      "epoch": 1.7635440180586908,
+      "grad_norm": 1.1562422513961792,
+      "learning_rate": 4.9285621469373565e-06,
+      "loss": 0.3976,
+      "mean_token_accuracy": 0.8717629671096802,
+      "num_tokens": 25374820.0,
+      "step": 3125
+    },
+    {
+      "entropy": 0.5338060319423675,
+      "epoch": 1.7663656884875847,
+      "grad_norm": 1.4789332151412964,
+      "learning_rate": 4.9283347587602e-06,
+      "loss": 0.4161,
+      "mean_token_accuracy": 0.8673757314682007,
+      "num_tokens": 25415684.0,
+      "step": 3130
+    },
+    {
+      "entropy": 0.5420262277126312,
+      "epoch": 1.7691873589164786,
+      "grad_norm": 1.34184730052948,
+      "learning_rate": 4.928107016317697e-06,
+      "loss": 0.4364,
+      "mean_token_accuracy": 0.8614213824272156,
+      "num_tokens": 25456228.0,
+      "step": 3135
+    },
+    {
+      "entropy": 0.5315328538417816,
+      "epoch": 1.7720090293453725,
+      "grad_norm": 1.4150985479354858,
+      "learning_rate": 4.927878919654585e-06,
+      "loss": 0.4275,
+      "mean_token_accuracy": 0.8663432002067566,
+      "num_tokens": 25497007.0,
+      "step": 3140
+    },
+    {
+      "entropy": 0.5486104905605316,
+      "epoch": 1.7748306997742662,
+      "grad_norm": 1.6440742015838623,
+      "learning_rate": 4.927650468815675e-06,
+      "loss": 0.4464,
+      "mean_token_accuracy": 0.8615983009338379,
+      "num_tokens": 25537572.0,
+      "step": 3145
+    },
+    {
+      "entropy": 0.5485983848571777,
+      "epoch": 1.7776523702031604,
+      "grad_norm": 1.3117893934249878,
+      "learning_rate": 4.927421663845847e-06,
+      "loss": 0.4287,
+      "mean_token_accuracy": 0.8655959486961364,
+      "num_tokens": 25578393.0,
+      "step": 3150
+    },
+    {
+      "entropy": 0.5457029461860656,
+      "epoch": 1.780474040632054,
+      "grad_norm": 1.4625993967056274,
+      "learning_rate": 4.927192504790048e-06,
+      "loss": 0.4178,
+      "mean_token_accuracy": 0.8674182772636414,
+      "num_tokens": 25619127.0,
+      "step": 3155
+    },
+    {
+      "entropy": 0.5470890522003173,
+      "epoch": 1.7832957110609482,
+      "grad_norm": 1.3004167079925537,
+      "learning_rate": 4.926962991693297e-06,
+      "loss": 0.4472,
+      "mean_token_accuracy": 0.8591425657272339,
+      "num_tokens": 25659696.0,
+      "step": 3160
+    },
+    {
+      "entropy": 0.5210420250892639,
+      "epoch": 1.786117381489842,
+      "grad_norm": 1.358970046043396,
+      "learning_rate": 4.926733124600682e-06,
+      "loss": 0.4098,
+      "mean_token_accuracy": 0.8688877105712891,
+      "num_tokens": 25700593.0,
+      "step": 3165
+    },
+    {
+      "entropy": 0.5520116329193115,
+      "epoch": 1.788939051918736,
+      "grad_norm": 1.470733880996704,
+      "learning_rate": 4.926502903557361e-06,
+      "loss": 0.44,
+      "mean_token_accuracy": 0.8626960635185241,
+      "num_tokens": 25741390.0,
+      "step": 3170
+    },
+    {
+      "entropy": 0.5124341726303101,
+      "epoch": 1.7917607223476297,
+      "grad_norm": 1.2576587200164795,
+      "learning_rate": 4.92627232860856e-06,
+      "loss": 0.4181,
+      "mean_token_accuracy": 0.8688732981681824,
+      "num_tokens": 25781767.0,
+      "step": 3175
+    },
+    {
+      "entropy": 0.48606478571891787,
+      "epoch": 1.7945823927765236,
+      "grad_norm": 1.2571027278900146,
+      "learning_rate": 4.926041399799576e-06,
+      "loss": 0.3849,
+      "mean_token_accuracy": 0.8780399203300476,
+      "num_tokens": 25822389.0,
+      "step": 3180
+    },
+    {
+      "entropy": 0.5661851406097412,
+      "epoch": 1.7974040632054176,
+      "grad_norm": 1.4934290647506714,
+      "learning_rate": 4.925810117175775e-06,
+      "loss": 0.4423,
+      "mean_token_accuracy": 0.8613282799720764,
+      "num_tokens": 25862747.0,
+      "step": 3185
+    },
+    {
+      "entropy": 0.5048659920692444,
+      "epoch": 1.8002257336343115,
+      "grad_norm": 1.2331528663635254,
+      "learning_rate": 4.925578480782593e-06,
+      "loss": 0.3776,
+      "mean_token_accuracy": 0.8817182898521423,
+      "num_tokens": 25903452.0,
+      "step": 3190
+    },
+    {
+      "entropy": 0.4995622456073761,
+      "epoch": 1.8030474040632054,
+      "grad_norm": 1.370931625366211,
+      "learning_rate": 4.925346490665533e-06,
+      "loss": 0.3816,
+      "mean_token_accuracy": 0.8787646174430848,
+      "num_tokens": 25943952.0,
+      "step": 3195
+    },
+    {
+      "entropy": 0.5398005902767181,
+      "epoch": 1.8058690744920993,
+      "grad_norm": 1.33382248878479,
+      "learning_rate": 4.925114146870172e-06,
+      "loss": 0.4339,
+      "mean_token_accuracy": 0.8646183371543884,
+      "num_tokens": 25984602.0,
+      "step": 3200
+    },
+    {
+      "entropy": 0.563685005903244,
+      "epoch": 1.8086907449209932,
+      "grad_norm": 1.504831314086914,
+      "learning_rate": 4.924881449442153e-06,
+      "loss": 0.4739,
+      "mean_token_accuracy": 0.8542726278305054,
+      "num_tokens": 26024952.0,
+      "step": 3205
+    },
+    {
+      "entropy": 0.5182668328285217,
+      "epoch": 1.8115124153498872,
+      "grad_norm": 1.46697199344635,
+      "learning_rate": 4.924648398427189e-06,
+      "loss": 0.4174,
+      "mean_token_accuracy": 0.8678057789802551,
+      "num_tokens": 26065459.0,
+      "step": 3210
+    },
+    {
+      "entropy": 0.5408389866352081,
+      "epoch": 1.814334085778781,
+      "grad_norm": 1.30887770652771,
+      "learning_rate": 4.924414993871063e-06,
+      "loss": 0.419,
+      "mean_token_accuracy": 0.8668316960334778,
+      "num_tokens": 26106028.0,
+      "step": 3215
+    },
+    {
+      "entropy": 0.542147308588028,
+      "epoch": 1.8171557562076748,
+      "grad_norm": 1.524593710899353,
+      "learning_rate": 4.924181235819627e-06,
+      "loss": 0.4275,
+      "mean_token_accuracy": 0.8657636642456055,
+      "num_tokens": 26146540.0,
+      "step": 3220
+    },
+    {
+      "entropy": 0.5389390230178833,
+      "epoch": 1.819977426636569,
+      "grad_norm": 1.4004186391830444,
+      "learning_rate": 4.923947124318804e-06,
+      "loss": 0.4232,
+      "mean_token_accuracy": 0.8676887512207031,
+      "num_tokens": 26186216.0,
+      "step": 3225
+    },
+    {
+      "entropy": 0.5603163003921509,
+      "epoch": 1.8227990970654626,
+      "grad_norm": 1.5067588090896606,
+      "learning_rate": 4.923712659414585e-06,
+      "loss": 0.4433,
+      "mean_token_accuracy": 0.8593494296073914,
+      "num_tokens": 26226650.0,
+      "step": 3230
+    },
+    {
+      "entropy": 0.5345525562763214,
+      "epoch": 1.8256207674943568,
+      "grad_norm": 1.264466643333435,
+      "learning_rate": 4.923477841153029e-06,
+      "loss": 0.424,
+      "mean_token_accuracy": 0.8678050875663758,
+      "num_tokens": 26267478.0,
+      "step": 3235
+    },
+    {
+      "entropy": 0.4989795684814453,
+      "epoch": 1.8284424379232505,
+      "grad_norm": 1.2940160036087036,
+      "learning_rate": 4.923242669580268e-06,
+      "loss": 0.3812,
+      "mean_token_accuracy": 0.8774576902389526,
+      "num_tokens": 26308103.0,
+      "step": 3240
+    },
+    {
+      "entropy": 0.5003684043884278,
+      "epoch": 1.8312641083521446,
+      "grad_norm": 1.4770820140838623,
+      "learning_rate": 4.923007144742501e-06,
+      "loss": 0.4049,
+      "mean_token_accuracy": 0.8739272236824036,
+      "num_tokens": 26348574.0,
+      "step": 3245
+    },
+    {
+      "entropy": 0.4989498794078827,
+      "epoch": 1.8340857787810383,
+      "grad_norm": 1.2541260719299316,
+      "learning_rate": 4.922771266685997e-06,
+      "loss": 0.3849,
+      "mean_token_accuracy": 0.8757034659385681,
+      "num_tokens": 26389200.0,
+      "step": 3250
+    },
+    {
+      "entropy": 0.5914472699165344,
+      "epoch": 1.8369074492099324,
+      "grad_norm": 1.3079047203063965,
+      "learning_rate": 4.922535035457094e-06,
+      "loss": 0.4896,
+      "mean_token_accuracy": 0.8483936309814453,
+      "num_tokens": 26429900.0,
+      "step": 3255
+    },
+    {
+      "entropy": 0.5169390797615051,
+      "epoch": 1.8397291196388261,
+      "grad_norm": 1.5444384813308716,
+      "learning_rate": 4.922298451102199e-06,
+      "loss": 0.3953,
+      "mean_token_accuracy": 0.8759800791740417,
+      "num_tokens": 26470465.0,
+      "step": 3260
+    },
+    {
+      "entropy": 0.536392205953598,
+      "epoch": 1.84255079006772,
+      "grad_norm": 1.189184546470642,
+      "learning_rate": 4.922061513667789e-06,
+      "loss": 0.434,
+      "mean_token_accuracy": 0.8635899901390076,
+      "num_tokens": 26511105.0,
+      "step": 3265
+    },
+    {
+      "entropy": 0.5151565790176391,
+      "epoch": 1.845372460496614,
+      "grad_norm": 1.5389307737350464,
+      "learning_rate": 4.921824223200412e-06,
+      "loss": 0.3973,
+      "mean_token_accuracy": 0.8712221384048462,
+      "num_tokens": 26551938.0,
+      "step": 3270
+    },
+    {
+      "entropy": 0.5374954879283905,
+      "epoch": 1.8481941309255079,
+      "grad_norm": 1.3970680236816406,
+      "learning_rate": 4.9215865797466826e-06,
+      "loss": 0.4253,
+      "mean_token_accuracy": 0.8682995200157165,
+      "num_tokens": 26592540.0,
+      "step": 3275
+    },
+    {
+      "entropy": 0.5597237467765808,
+      "epoch": 1.8510158013544018,
+      "grad_norm": 1.3503648042678833,
+      "learning_rate": 4.921348583353286e-06,
+      "loss": 0.4328,
+      "mean_token_accuracy": 0.8641868352890014,
+      "num_tokens": 26633368.0,
+      "step": 3280
+    },
+    {
+      "entropy": 0.5081552624702453,
+      "epoch": 1.8538374717832957,
+      "grad_norm": 1.3734320402145386,
+      "learning_rate": 4.921110234066977e-06,
+      "loss": 0.4065,
+      "mean_token_accuracy": 0.8727322340011596,
+      "num_tokens": 26674050.0,
+      "step": 3285
+    },
+    {
+      "entropy": 0.5667516946792602,
+      "epoch": 1.8566591422121896,
+      "grad_norm": 1.4739806652069092,
+      "learning_rate": 4.920871531934579e-06,
+      "loss": 0.4482,
+      "mean_token_accuracy": 0.8577787399291992,
+      "num_tokens": 26714528.0,
+      "step": 3290
+    },
+    {
+      "entropy": 0.5383511900901794,
+      "epoch": 1.8594808126410836,
+      "grad_norm": 1.450299620628357,
+      "learning_rate": 4.920632477002985e-06,
+      "loss": 0.4157,
+      "mean_token_accuracy": 0.8684499502182007,
+      "num_tokens": 26755102.0,
+      "step": 3295
+    },
+    {
+      "entropy": 0.5264916241168975,
+      "epoch": 1.8623024830699775,
+      "grad_norm": 1.339902400970459,
+      "learning_rate": 4.9203930693191575e-06,
+      "loss": 0.4283,
+      "mean_token_accuracy": 0.8646404504776001,
+      "num_tokens": 26795847.0,
+      "step": 3300
+    },
+    {
+      "entropy": 0.501398503780365,
+      "epoch": 1.8651241534988712,
+      "grad_norm": 1.3118853569030762,
+      "learning_rate": 4.920153308930128e-06,
+      "loss": 0.3929,
+      "mean_token_accuracy": 0.872932231426239,
+      "num_tokens": 26836356.0,
+      "step": 3305
+    },
+    {
+      "entropy": 0.5669954061508179,
+      "epoch": 1.8679458239277653,
+      "grad_norm": 1.4395278692245483,
+      "learning_rate": 4.919913195882997e-06,
+      "loss": 0.4449,
+      "mean_token_accuracy": 0.8604054570198059,
+      "num_tokens": 26876942.0,
+      "step": 3310
+    },
+    {
+      "entropy": 0.4821522831916809,
+      "epoch": 1.870767494356659,
+      "grad_norm": 1.3885427713394165,
+      "learning_rate": 4.919672730224936e-06,
+      "loss": 0.3866,
+      "mean_token_accuracy": 0.8788249969482422,
+      "num_tokens": 26917741.0,
+      "step": 3315
+    },
+    {
+      "entropy": 0.5123970150947571,
+      "epoch": 1.8735891647855532,
+      "grad_norm": 1.3463451862335205,
+      "learning_rate": 4.9194319120031836e-06,
+      "loss": 0.3925,
+      "mean_token_accuracy": 0.8763728380203247,
+      "num_tokens": 26957483.0,
+      "step": 3320
+    },
+    {
+      "entropy": 0.5375806808471679,
+      "epoch": 1.8764108352144468,
+      "grad_norm": 1.339368462562561,
+      "learning_rate": 4.91919074126505e-06,
+      "loss": 0.4255,
+      "mean_token_accuracy": 0.8660783290863037,
+      "num_tokens": 26998053.0,
+      "step": 3325
+    },
+    {
+      "entropy": 0.5030182540416718,
+      "epoch": 1.879232505643341,
+      "grad_norm": 1.2726454734802246,
+      "learning_rate": 4.91894921805791e-06,
+      "loss": 0.3762,
+      "mean_token_accuracy": 0.8820128679275513,
+      "num_tokens": 27038835.0,
+      "step": 3330
+    },
+    {
+      "entropy": 0.5537118554115296,
+      "epoch": 1.8820541760722347,
+      "grad_norm": 1.3522690534591675,
+      "learning_rate": 4.918707342429214e-06,
+      "loss": 0.4299,
+      "mean_token_accuracy": 0.8657578349113464,
+      "num_tokens": 27079464.0,
+      "step": 3335
+    },
+    {
+      "entropy": 0.5303456306457519,
+      "epoch": 1.8848758465011288,
+      "grad_norm": 1.423553705215454,
+      "learning_rate": 4.9184651144264776e-06,
+      "loss": 0.4217,
+      "mean_token_accuracy": 0.8663669109344483,
+      "num_tokens": 27120201.0,
+      "step": 3340
+    },
+    {
+      "entropy": 0.5344128489494324,
+      "epoch": 1.8876975169300225,
+      "grad_norm": 1.223803162574768,
+      "learning_rate": 4.918222534097286e-06,
+      "loss": 0.4155,
+      "mean_token_accuracy": 0.8672279477119446,
+      "num_tokens": 27160686.0,
+      "step": 3345
+    },
+    {
+      "entropy": 0.4743356049060822,
+      "epoch": 1.8905191873589164,
+      "grad_norm": 1.3717639446258545,
+      "learning_rate": 4.917979601489295e-06,
+      "loss": 0.3674,
+      "mean_token_accuracy": 0.8822289109230042,
+      "num_tokens": 27201250.0,
+      "step": 3350
+    },
+    {
+      "entropy": 0.5208858251571655,
+      "epoch": 1.8933408577878104,
+      "grad_norm": 1.2312960624694824,
+      "learning_rate": 4.917736316650228e-06,
+      "loss": 0.3954,
+      "mean_token_accuracy": 0.8755510807037353,
+      "num_tokens": 27241854.0,
+      "step": 3355
+    },
+    {
+      "entropy": 0.5430815577507019,
+      "epoch": 1.8961625282167043,
+      "grad_norm": 1.385640025138855,
+      "learning_rate": 4.917492679627879e-06,
+      "loss": 0.436,
+      "mean_token_accuracy": 0.8614830732345581,
+      "num_tokens": 27282703.0,
+      "step": 3360
+    },
+    {
+      "entropy": 0.5193690776824951,
+      "epoch": 1.8989841986455982,
+      "grad_norm": 1.6769845485687256,
+      "learning_rate": 4.917248690470109e-06,
+      "loss": 0.3994,
+      "mean_token_accuracy": 0.8753103852272034,
+      "num_tokens": 27323255.0,
+      "step": 3365
+    },
+    {
+      "entropy": 0.5402388453483582,
+      "epoch": 1.9018058690744921,
+      "grad_norm": 1.4692338705062866,
+      "learning_rate": 4.917004349224851e-06,
+      "loss": 0.4094,
+      "mean_token_accuracy": 0.8716601133346558,
+      "num_tokens": 27363961.0,
+      "step": 3370
+    },
+    {
+      "entropy": 0.5165061771869659,
+      "epoch": 1.904627539503386,
+      "grad_norm": 1.2910758256912231,
+      "learning_rate": 4.916759655940107e-06,
+      "loss": 0.3992,
+      "mean_token_accuracy": 0.8713883280754089,
+      "num_tokens": 27404297.0,
+      "step": 3375
+    },
+    {
+      "entropy": 0.5438097000122071,
+      "epoch": 1.90744920993228,
+      "grad_norm": 1.5428695678710938,
+      "learning_rate": 4.916514610663943e-06,
+      "loss": 0.4398,
+      "mean_token_accuracy": 0.8607651233673096,
+      "num_tokens": 27444966.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.5148679494857789,
+      "epoch": 1.9102708803611739,
+      "grad_norm": 1.370609164237976,
+      "learning_rate": 4.916269213444502e-06,
+      "loss": 0.3973,
+      "mean_token_accuracy": 0.8719611167907715,
+      "num_tokens": 27485807.0,
+      "step": 3385
+    },
+    {
+      "entropy": 0.552492767572403,
+      "epoch": 1.9130925507900676,
+      "grad_norm": 1.3460413217544556,
+      "learning_rate": 4.9160234643299935e-06,
+      "loss": 0.4232,
+      "mean_token_accuracy": 0.867039430141449,
+      "num_tokens": 27526548.0,
+      "step": 3390
+    },
+    {
+      "entropy": 0.5185729682445526,
+      "epoch": 1.9159142212189617,
+      "grad_norm": 1.2497392892837524,
+      "learning_rate": 4.91577736336869e-06,
+      "loss": 0.4138,
+      "mean_token_accuracy": 0.8712507843971252,
+      "num_tokens": 27566660.0,
+      "step": 3395
+    },
+    {
+      "entropy": 0.5251140117645263,
+      "epoch": 1.9187358916478554,
+      "grad_norm": 1.4792594909667969,
+      "learning_rate": 4.915530910608941e-06,
+      "loss": 0.4159,
+      "mean_token_accuracy": 0.8693275570869445,
+      "num_tokens": 27607302.0,
+      "step": 3400
+    },
+    {
+      "entropy": 0.5528305053710938,
+      "epoch": 1.9215575620767495,
+      "grad_norm": 1.4354878664016724,
+      "learning_rate": 4.915284106099162e-06,
+      "loss": 0.4363,
+      "mean_token_accuracy": 0.864462697505951,
+      "num_tokens": 27648058.0,
+      "step": 3405
+    },
+    {
+      "entropy": 0.5455857455730438,
+      "epoch": 1.9243792325056432,
+      "grad_norm": 1.3338799476623535,
+      "learning_rate": 4.915036949887838e-06,
+      "loss": 0.4146,
+      "mean_token_accuracy": 0.8690644383430481,
+      "num_tokens": 27688285.0,
+      "step": 3410
+    },
+    {
+      "entropy": 0.524651050567627,
+      "epoch": 1.9272009029345374,
+      "grad_norm": 1.3062248229980469,
+      "learning_rate": 4.914789442023523e-06,
+      "loss": 0.4082,
+      "mean_token_accuracy": 0.8696060180664062,
+      "num_tokens": 27728945.0,
+      "step": 3415
+    },
+    {
+      "entropy": 0.5397283554077148,
+      "epoch": 1.930022573363431,
+      "grad_norm": 1.3946961164474487,
+      "learning_rate": 4.914541582554838e-06,
+      "loss": 0.4419,
+      "mean_token_accuracy": 0.8649717569351196,
+      "num_tokens": 27769755.0,
+      "step": 3420
+    },
+    {
+      "entropy": 0.5519603669643403,
+      "epoch": 1.9328442437923252,
+      "grad_norm": 1.3521655797958374,
+      "learning_rate": 4.914293371530478e-06,
+      "loss": 0.448,
+      "mean_token_accuracy": 0.858772075176239,
+      "num_tokens": 27810361.0,
+      "step": 3425
+    },
+    {
+      "entropy": 0.5126847267150879,
+      "epoch": 1.935665914221219,
+      "grad_norm": 1.469565510749817,
+      "learning_rate": 4.914044808999202e-06,
+      "loss": 0.3962,
+      "mean_token_accuracy": 0.8749841451644897,
+      "num_tokens": 27851096.0,
+      "step": 3430
+    },
+    {
+      "entropy": 0.5041461646556854,
+      "epoch": 1.9384875846501128,
+      "grad_norm": 1.2633264064788818,
+      "learning_rate": 4.913795895009841e-06,
+      "loss": 0.3762,
+      "mean_token_accuracy": 0.8790737509727478,
+      "num_tokens": 27891803.0,
+      "step": 3435
+    },
+    {
+      "entropy": 0.5427121460437775,
+      "epoch": 1.9413092550790068,
+      "grad_norm": 1.4089435338974,
+      "learning_rate": 4.913546629611294e-06,
+      "loss": 0.4497,
+      "mean_token_accuracy": 0.8610092759132385,
+      "num_tokens": 27932403.0,
+      "step": 3440
+    },
+    {
+      "entropy": 0.5261597633361816,
+      "epoch": 1.9441309255079007,
+      "grad_norm": 1.4115110635757446,
+      "learning_rate": 4.913297012852528e-06,
+      "loss": 0.4186,
+      "mean_token_accuracy": 0.8676816701889039,
+      "num_tokens": 27973177.0,
+      "step": 3445
+    },
+    {
+      "entropy": 0.5300165891647339,
+      "epoch": 1.9469525959367946,
+      "grad_norm": 1.4206523895263672,
+      "learning_rate": 4.9130470447825816e-06,
+      "loss": 0.413,
+      "mean_token_accuracy": 0.8669154524803162,
+      "num_tokens": 28013878.0,
+      "step": 3450
+    },
+    {
+      "entropy": 0.5528669118881225,
+      "epoch": 1.9497742663656885,
+      "grad_norm": 1.4294172525405884,
+      "learning_rate": 4.912796725450562e-06,
+      "loss": 0.4264,
+      "mean_token_accuracy": 0.8657429218292236,
+      "num_tokens": 28054452.0,
+      "step": 3455
+    },
+    {
+      "entropy": 0.5473185420036316,
+      "epoch": 1.9525959367945824,
+      "grad_norm": 1.4492708444595337,
+      "learning_rate": 4.912546054905642e-06,
+      "loss": 0.4413,
+      "mean_token_accuracy": 0.8618425607681275,
+      "num_tokens": 28095117.0,
+      "step": 3460
+    },
+    {
+      "entropy": 0.5194451570510864,
+      "epoch": 1.9554176072234764,
+      "grad_norm": 1.4601637125015259,
+      "learning_rate": 4.912295033197068e-06,
+      "loss": 0.4066,
+      "mean_token_accuracy": 0.8705293416976929,
+      "num_tokens": 28135700.0,
+      "step": 3465
+    },
+    {
+      "entropy": 0.5036539554595947,
+      "epoch": 1.9582392776523703,
+      "grad_norm": 1.4262093305587769,
+      "learning_rate": 4.9120436603741515e-06,
+      "loss": 0.3792,
+      "mean_token_accuracy": 0.8786629676818848,
+      "num_tokens": 28176234.0,
+      "step": 3470
+    },
+    {
+      "entropy": 0.5227799832820892,
+      "epoch": 1.961060948081264,
+      "grad_norm": 1.4064563512802124,
+      "learning_rate": 4.911791936486276e-06,
+      "loss": 0.4158,
+      "mean_token_accuracy": 0.871609365940094,
+      "num_tokens": 28216567.0,
+      "step": 3475
+    },
+    {
+      "entropy": 0.49351215958595274,
+      "epoch": 1.963882618510158,
+      "grad_norm": 1.401670217514038,
+      "learning_rate": 4.911539861582893e-06,
+      "loss": 0.3878,
+      "mean_token_accuracy": 0.8766814708709717,
+      "num_tokens": 28257305.0,
+      "step": 3480
+    },
+    {
+      "entropy": 0.49786993861198425,
+      "epoch": 1.9667042889390518,
+      "grad_norm": 1.5233041048049927,
+      "learning_rate": 4.911287435713522e-06,
+      "loss": 0.3905,
+      "mean_token_accuracy": 0.8762277483940124,
+      "num_tokens": 28297776.0,
+      "step": 3485
+    },
+    {
+      "entropy": 0.4954915583133698,
+      "epoch": 1.969525959367946,
+      "grad_norm": 1.241149663925171,
+      "learning_rate": 4.911034658927751e-06,
+      "loss": 0.3957,
+      "mean_token_accuracy": 0.8773471236228942,
+      "num_tokens": 28338384.0,
+      "step": 3490
+    },
+    {
+      "entropy": 0.5027383029460907,
+      "epoch": 1.9723476297968396,
+      "grad_norm": 1.4190332889556885,
+      "learning_rate": 4.91078153127524e-06,
+      "loss": 0.3892,
+      "mean_token_accuracy": 0.8748612880706788,
+      "num_tokens": 28379137.0,
+      "step": 3495
+    },
+    {
+      "entropy": 0.5188802659511567,
+      "epoch": 1.9751693002257338,
+      "grad_norm": 1.4441595077514648,
+      "learning_rate": 4.910528052805714e-06,
+      "loss": 0.4141,
+      "mean_token_accuracy": 0.8701729416847229,
+      "num_tokens": 28419987.0,
+      "step": 3500
+    },
+    {
+      "epoch": 1.9751693002257338,
+      "eval_entropy": 0.5039494037628174,
+      "eval_loss": 0.32582059502601624,
+      "eval_mean_token_accuracy": 0.9022613763809204,
+      "eval_num_tokens": 28419987.0,
+      "eval_runtime": 0.1635,
+      "eval_samples_per_second": 24.464,
+      "eval_steps_per_second": 6.116,
+      "step": 3500
+    },
+    {
+      "entropy": 0.5050791263580322,
+      "epoch": 1.9779909706546275,
+      "grad_norm": 1.2525067329406738,
+      "learning_rate": 4.910274223568971e-06,
+      "loss": 0.4111,
+      "mean_token_accuracy": 0.8697703242301941,
+      "num_tokens": 28460369.0,
+      "step": 3505
+    },
+    {
+      "entropy": 0.5229144394397736,
+      "epoch": 1.9808126410835214,
+      "grad_norm": 1.4554386138916016,
+      "learning_rate": 4.9100200436148735e-06,
+      "loss": 0.4106,
+      "mean_token_accuracy": 0.8699278473854065,
+      "num_tokens": 28501101.0,
+      "step": 3510
+    },
+    {
+      "entropy": 0.5569809198379516,
+      "epoch": 1.9836343115124153,
+      "grad_norm": 1.4474997520446777,
+      "learning_rate": 4.909765512993357e-06,
+      "loss": 0.4563,
+      "mean_token_accuracy": 0.857849943637848,
+      "num_tokens": 28541729.0,
+      "step": 3515
+    },
+    {
+      "entropy": 0.5587629973888397,
+      "epoch": 1.9864559819413092,
+      "grad_norm": 1.3627238273620605,
+      "learning_rate": 4.909510631754425e-06,
+      "loss": 0.4268,
+      "mean_token_accuracy": 0.8645955443382263,
+      "num_tokens": 28582055.0,
+      "step": 3520
+    },
+    {
+      "entropy": 0.5122965276241302,
+      "epoch": 1.9892776523702032,
+      "grad_norm": 1.2392799854278564,
+      "learning_rate": 4.909255399948146e-06,
+      "loss": 0.3933,
+      "mean_token_accuracy": 0.8727590441703796,
+      "num_tokens": 28622596.0,
+      "step": 3525
+    },
+    {
+      "entropy": 0.5298355937004089,
+      "epoch": 1.992099322799097,
+      "grad_norm": 1.6109238862991333,
+      "learning_rate": 4.908999817624661e-06,
+      "loss": 0.4189,
+      "mean_token_accuracy": 0.8664365530014038,
+      "num_tokens": 28662503.0,
+      "step": 3530
+    },
+    {
+      "entropy": 0.5138360559940338,
+      "epoch": 1.994920993227991,
+      "grad_norm": 1.3417595624923706,
+      "learning_rate": 4.9087438848341806e-06,
+      "loss": 0.4158,
+      "mean_token_accuracy": 0.8667647480964661,
+      "num_tokens": 28702848.0,
+      "step": 3535
+    },
+    {
+      "entropy": 0.5267124712467194,
+      "epoch": 1.997742663656885,
+      "grad_norm": 1.2989228963851929,
+      "learning_rate": 4.908487601626983e-06,
+      "loss": 0.4113,
+      "mean_token_accuracy": 0.8676295042037964,
+      "num_tokens": 28743457.0,
+      "step": 3540
+    },
+    {
+      "entropy": 0.5033242404460907,
+      "epoch": 2.000564334085779,
+      "grad_norm": 1.0862282514572144,
+      "learning_rate": 4.9082309680534134e-06,
+      "loss": 0.3848,
+      "mean_token_accuracy": 0.8765107035636902,
+      "num_tokens": 28777728.0,
+      "step": 3545
+    },
+    {
+      "entropy": 0.4844656944274902,
+      "epoch": 2.0033860045146725,
+      "grad_norm": 1.1120542287826538,
+      "learning_rate": 4.907973984163888e-06,
+      "loss": 0.3023,
+      "mean_token_accuracy": 0.9055742740631103,
+      "num_tokens": 28818642.0,
+      "step": 3550
+    },
+    {
+      "entropy": 0.4910138726234436,
+      "epoch": 2.0062076749435667,
+      "grad_norm": 1.2199798822402954,
+      "learning_rate": 4.907716650008893e-06,
+      "loss": 0.3487,
+      "mean_token_accuracy": 0.8892881631851196,
+      "num_tokens": 28859021.0,
+      "step": 3555
+    },
+    {
+      "entropy": 0.4198545038700104,
+      "epoch": 2.0090293453724604,
+      "grad_norm": 1.348246693611145,
+      "learning_rate": 4.907458965638979e-06,
+      "loss": 0.3017,
+      "mean_token_accuracy": 0.9059156775474548,
+      "num_tokens": 28899517.0,
+      "step": 3560
+    },
+    {
+      "entropy": 0.42251318097114565,
+      "epoch": 2.0118510158013545,
+      "grad_norm": 1.4219539165496826,
+      "learning_rate": 4.90720093110477e-06,
+      "loss": 0.316,
+      "mean_token_accuracy": 0.8997484803199768,
+      "num_tokens": 28940360.0,
+      "step": 3565
+    },
+    {
+      "entropy": 0.4202295243740082,
+      "epoch": 2.014672686230248,
+      "grad_norm": 1.5008666515350342,
+      "learning_rate": 4.906942546456957e-06,
+      "loss": 0.3168,
+      "mean_token_accuracy": 0.8997807145118714,
+      "num_tokens": 28981158.0,
+      "step": 3570
+    },
+    {
+      "entropy": 0.39814443588256837,
+      "epoch": 2.0174943566591423,
+      "grad_norm": 1.4986308813095093,
+      "learning_rate": 4.906683811746298e-06,
+      "loss": 0.2957,
+      "mean_token_accuracy": 0.9047508955001831,
+      "num_tokens": 29021722.0,
+      "step": 3575
+    },
+    {
+      "entropy": 0.4490103781223297,
+      "epoch": 2.020316027088036,
+      "grad_norm": 1.7003601789474487,
+      "learning_rate": 4.9064247270236235e-06,
+      "loss": 0.3252,
+      "mean_token_accuracy": 0.8966346383094788,
+      "num_tokens": 29062372.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.440228271484375,
+      "epoch": 2.02313769751693,
+      "grad_norm": 1.8546236753463745,
+      "learning_rate": 4.906165292339828e-06,
+      "loss": 0.3164,
+      "mean_token_accuracy": 0.899936830997467,
+      "num_tokens": 29102844.0,
+      "step": 3585
+    },
+    {
+      "entropy": 0.39439972043037413,
+      "epoch": 2.025959367945824,
+      "grad_norm": 1.4125168323516846,
+      "learning_rate": 4.905905507745881e-06,
+      "loss": 0.2914,
+      "mean_token_accuracy": 0.9061366796493531,
+      "num_tokens": 29143188.0,
+      "step": 3590
+    },
+    {
+      "entropy": 0.40190274715423585,
+      "epoch": 2.028781038374718,
+      "grad_norm": 1.5160431861877441,
+      "learning_rate": 4.905645373292815e-06,
+      "loss": 0.2931,
+      "mean_token_accuracy": 0.9052773237228393,
+      "num_tokens": 29183656.0,
+      "step": 3595
+    },
+    {
+      "entropy": 0.4064796566963196,
+      "epoch": 2.0316027088036117,
+      "grad_norm": 1.6123907566070557,
+      "learning_rate": 4.905384889031734e-06,
+      "loss": 0.302,
+      "mean_token_accuracy": 0.9012960076332093,
+      "num_tokens": 29224451.0,
+      "step": 3600
+    },
+    {
+      "entropy": 0.4297832429409027,
+      "epoch": 2.034424379232506,
+      "grad_norm": 1.5297926664352417,
+      "learning_rate": 4.90512405501381e-06,
+      "loss": 0.3242,
+      "mean_token_accuracy": 0.8968385696411133,
+      "num_tokens": 29265091.0,
+      "step": 3605
+    },
+    {
+      "entropy": 0.41181485652923583,
+      "epoch": 2.0372460496613995,
+      "grad_norm": 1.4415594339370728,
+      "learning_rate": 4.904862871290285e-06,
+      "loss": 0.3107,
+      "mean_token_accuracy": 0.9039250016212463,
+      "num_tokens": 29305558.0,
+      "step": 3610
+    },
+    {
+      "entropy": 0.4206926107406616,
+      "epoch": 2.0400677200902932,
+      "grad_norm": 1.4016162157058716,
+      "learning_rate": 4.904601337912467e-06,
+      "loss": 0.3029,
+      "mean_token_accuracy": 0.9045567750930786,
+      "num_tokens": 29346298.0,
+      "step": 3615
+    },
+    {
+      "entropy": 0.3947125256061554,
+      "epoch": 2.0428893905191874,
+      "grad_norm": 1.505416989326477,
+      "learning_rate": 4.9043394549317345e-06,
+      "loss": 0.2789,
+      "mean_token_accuracy": 0.9085012078285217,
+      "num_tokens": 29387055.0,
+      "step": 3620
+    },
+    {
+      "entropy": 0.39603736996650696,
+      "epoch": 2.045711060948081,
+      "grad_norm": 1.6036330461502075,
+      "learning_rate": 4.904077222399534e-06,
+      "loss": 0.3023,
+      "mean_token_accuracy": 0.90137619972229,
+      "num_tokens": 29427763.0,
+      "step": 3625
+    },
+    {
+      "entropy": 0.3889478862285614,
+      "epoch": 2.0485327313769752,
+      "grad_norm": 1.672661304473877,
+      "learning_rate": 4.903814640367383e-06,
+      "loss": 0.2932,
+      "mean_token_accuracy": 0.9071918845176696,
+      "num_tokens": 29468411.0,
+      "step": 3630
+    },
+    {
+      "entropy": 0.39558014273643494,
+      "epoch": 2.051354401805869,
+      "grad_norm": 1.556510090827942,
+      "learning_rate": 4.903551708886865e-06,
+      "loss": 0.2879,
+      "mean_token_accuracy": 0.9074760317802429,
+      "num_tokens": 29509088.0,
+      "step": 3635
+    },
+    {
+      "entropy": 0.41549997925758364,
+      "epoch": 2.054176072234763,
+      "grad_norm": 1.4191838502883911,
+      "learning_rate": 4.903288428009632e-06,
+      "loss": 0.2912,
+      "mean_token_accuracy": 0.9053076505661011,
+      "num_tokens": 29549728.0,
+      "step": 3640
+    },
+    {
+      "entropy": 0.4345362961292267,
+      "epoch": 2.0569977426636568,
+      "grad_norm": 1.6131196022033691,
+      "learning_rate": 4.9030247977874064e-06,
+      "loss": 0.3214,
+      "mean_token_accuracy": 0.8968990206718445,
+      "num_tokens": 29590362.0,
+      "step": 3645
+    },
+    {
+      "entropy": 0.38488792777061465,
+      "epoch": 2.059819413092551,
+      "grad_norm": 1.391592025756836,
+      "learning_rate": 4.902760818271978e-06,
+      "loss": 0.2758,
+      "mean_token_accuracy": 0.9102860927581787,
+      "num_tokens": 29630543.0,
+      "step": 3650
+    },
+    {
+      "entropy": 0.4055006742477417,
+      "epoch": 2.0626410835214446,
+      "grad_norm": 1.5487695932388306,
+      "learning_rate": 4.902496489515206e-06,
+      "loss": 0.308,
+      "mean_token_accuracy": 0.8991694211959839,
+      "num_tokens": 29671134.0,
+      "step": 3655
+    },
+    {
+      "entropy": 0.4026013076305389,
+      "epoch": 2.0654627539503387,
+      "grad_norm": 1.4963382482528687,
+      "learning_rate": 4.902231811569016e-06,
+      "loss": 0.3016,
+      "mean_token_accuracy": 0.9023310422897339,
+      "num_tokens": 29711975.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.3941408574581146,
+      "epoch": 2.0682844243792324,
+      "grad_norm": 1.4817650318145752,
+      "learning_rate": 4.901966784485407e-06,
+      "loss": 0.2837,
+      "mean_token_accuracy": 0.9089763045310975,
+      "num_tokens": 29752694.0,
+      "step": 3665
+    },
+    {
+      "entropy": 0.4023356318473816,
+      "epoch": 2.0711060948081266,
+      "grad_norm": 1.6465110778808594,
+      "learning_rate": 4.901701408316443e-06,
+      "loss": 0.2836,
+      "mean_token_accuracy": 0.908521831035614,
+      "num_tokens": 29793192.0,
+      "step": 3670
+    },
+    {
+      "entropy": 0.38578653931617735,
+      "epoch": 2.0739277652370203,
+      "grad_norm": 1.5175875425338745,
+      "learning_rate": 4.901435683114255e-06,
+      "loss": 0.2886,
+      "mean_token_accuracy": 0.9055356621742249,
+      "num_tokens": 29834115.0,
+      "step": 3675
+    },
+    {
+      "entropy": 0.3811330258846283,
+      "epoch": 2.0767494356659144,
+      "grad_norm": 1.527234435081482,
+      "learning_rate": 4.901169608931046e-06,
+      "loss": 0.2779,
+      "mean_token_accuracy": 0.9086094379425049,
+      "num_tokens": 29874790.0,
+      "step": 3680
+    },
+    {
+      "entropy": 0.3947294354438782,
+      "epoch": 2.079571106094808,
+      "grad_norm": 1.5221364498138428,
+      "learning_rate": 4.900903185819088e-06,
+      "loss": 0.2842,
+      "mean_token_accuracy": 0.9065226435661315,
+      "num_tokens": 29915452.0,
+      "step": 3685
+    },
+    {
+      "entropy": 0.372803258895874,
+      "epoch": 2.082392776523702,
+      "grad_norm": 1.4854118824005127,
+      "learning_rate": 4.900636413830717e-06,
+      "loss": 0.2865,
+      "mean_token_accuracy": 0.9063424587249755,
+      "num_tokens": 29956198.0,
+      "step": 3690
+    },
+    {
+      "entropy": 0.3771813929080963,
+      "epoch": 2.085214446952596,
+      "grad_norm": 1.432308316230774,
+      "learning_rate": 4.900369293018342e-06,
+      "loss": 0.2663,
+      "mean_token_accuracy": 0.9135713815689087,
+      "num_tokens": 29996744.0,
+      "step": 3695
+    },
+    {
+      "entropy": 0.3829148352146149,
+      "epoch": 2.0880361173814896,
+      "grad_norm": 1.5369491577148438,
+      "learning_rate": 4.900101823434438e-06,
+      "loss": 0.2865,
+      "mean_token_accuracy": 0.9079620480537415,
+      "num_tokens": 30037615.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.40159552693367007,
+      "epoch": 2.090857787810384,
+      "grad_norm": 1.6757155656814575,
+      "learning_rate": 4.8998340051315515e-06,
+      "loss": 0.2867,
+      "mean_token_accuracy": 0.9064762830734253,
+      "num_tokens": 30078214.0,
+      "step": 3705
+    },
+    {
+      "entropy": 0.393253880739212,
+      "epoch": 2.0936794582392775,
+      "grad_norm": 1.5468506813049316,
+      "learning_rate": 4.899565838162292e-06,
+      "loss": 0.2948,
+      "mean_token_accuracy": 0.9056425452232361,
+      "num_tokens": 30118900.0,
+      "step": 3710
+    },
+    {
+      "entropy": 0.42414683699607847,
+      "epoch": 2.0965011286681716,
+      "grad_norm": 1.736525297164917,
+      "learning_rate": 4.899297322579345e-06,
+      "loss": 0.3287,
+      "mean_token_accuracy": 0.8957733750343323,
+      "num_tokens": 30159519.0,
+      "step": 3715
+    },
+    {
+      "entropy": 0.3917936205863953,
+      "epoch": 2.0993227990970653,
+      "grad_norm": 1.5324411392211914,
+      "learning_rate": 4.899028458435458e-06,
+      "loss": 0.2951,
+      "mean_token_accuracy": 0.9044302463531494,
+      "num_tokens": 30199519.0,
+      "step": 3720
+    },
+    {
+      "entropy": 0.38330122232437136,
+      "epoch": 2.1021444695259595,
+      "grad_norm": 1.4559762477874756,
+      "learning_rate": 4.898759245783449e-06,
+      "loss": 0.2755,
+      "mean_token_accuracy": 0.9099204063415527,
+      "num_tokens": 30239952.0,
+      "step": 3725
+    },
+    {
+      "entropy": 0.3951174974441528,
+      "epoch": 2.104966139954853,
+      "grad_norm": 1.5351479053497314,
+      "learning_rate": 4.898489684676205e-06,
+      "loss": 0.2907,
+      "mean_token_accuracy": 0.9066962122917175,
+      "num_tokens": 30280534.0,
+      "step": 3730
+    },
+    {
+      "entropy": 0.3635559678077698,
+      "epoch": 2.1077878103837473,
+      "grad_norm": 1.50784170627594,
+      "learning_rate": 4.898219775166683e-06,
+      "loss": 0.2621,
+      "mean_token_accuracy": 0.9133730411529541,
+      "num_tokens": 30321115.0,
+      "step": 3735
+    },
+    {
+      "entropy": 0.41765828132629396,
+      "epoch": 2.110609480812641,
+      "grad_norm": 1.554824709892273,
+      "learning_rate": 4.897949517307905e-06,
+      "loss": 0.3156,
+      "mean_token_accuracy": 0.898926043510437,
+      "num_tokens": 30361715.0,
+      "step": 3740
+    },
+    {
+      "entropy": 0.40577141046524046,
+      "epoch": 2.113431151241535,
+      "grad_norm": 1.543889045715332,
+      "learning_rate": 4.897678911152964e-06,
+      "loss": 0.302,
+      "mean_token_accuracy": 0.9033358812332153,
+      "num_tokens": 30402151.0,
+      "step": 3745
+    },
+    {
+      "entropy": 0.3970208168029785,
+      "epoch": 2.116252821670429,
+      "grad_norm": 1.4221984148025513,
+      "learning_rate": 4.897407956755021e-06,
+      "loss": 0.2984,
+      "mean_token_accuracy": 0.9049067020416259,
+      "num_tokens": 30442797.0,
+      "step": 3750
+    },
+    {
+      "entropy": 0.41688573360443115,
+      "epoch": 2.119074492099323,
+      "grad_norm": 1.6445451974868774,
+      "learning_rate": 4.897136654167304e-06,
+      "loss": 0.3151,
+      "mean_token_accuracy": 0.9019560813903809,
+      "num_tokens": 30483552.0,
+      "step": 3755
+    },
+    {
+      "entropy": 0.34632920026779174,
+      "epoch": 2.1218961625282167,
+      "grad_norm": 1.3843744993209839,
+      "learning_rate": 4.896865003443111e-06,
+      "loss": 0.2493,
+      "mean_token_accuracy": 0.9194693088531494,
+      "num_tokens": 30524141.0,
+      "step": 3760
+    },
+    {
+      "entropy": 0.41517892479896545,
+      "epoch": 2.124717832957111,
+      "grad_norm": 1.4717671871185303,
+      "learning_rate": 4.896593004635807e-06,
+      "loss": 0.2993,
+      "mean_token_accuracy": 0.9012321352958679,
+      "num_tokens": 30564727.0,
+      "step": 3765
+    },
+    {
+      "entropy": 0.3839627206325531,
+      "epoch": 2.1275395033860045,
+      "grad_norm": 1.5019539594650269,
+      "learning_rate": 4.896320657798828e-06,
+      "loss": 0.2671,
+      "mean_token_accuracy": 0.9121496915817261,
+      "num_tokens": 30605161.0,
+      "step": 3770
+    },
+    {
+      "entropy": 0.40346505045890807,
+      "epoch": 2.130361173814898,
+      "grad_norm": 1.8757257461547852,
+      "learning_rate": 4.896047962985676e-06,
+      "loss": 0.3044,
+      "mean_token_accuracy": 0.9000169515609742,
+      "num_tokens": 30645838.0,
+      "step": 3775
+    },
+    {
+      "entropy": 0.40536361932754517,
+      "epoch": 2.1331828442437923,
+      "grad_norm": 1.508545994758606,
+      "learning_rate": 4.89577492024992e-06,
+      "loss": 0.3083,
+      "mean_token_accuracy": 0.899243688583374,
+      "num_tokens": 30686256.0,
+      "step": 3780
+    },
+    {
+      "entropy": 0.4287443220615387,
+      "epoch": 2.136004514672686,
+      "grad_norm": 1.4562077522277832,
+      "learning_rate": 4.895501529645201e-06,
+      "loss": 0.3222,
+      "mean_token_accuracy": 0.8983871340751648,
+      "num_tokens": 30726917.0,
+      "step": 3785
+    },
+    {
+      "entropy": 0.434832763671875,
+      "epoch": 2.13882618510158,
+      "grad_norm": 1.5828227996826172,
+      "learning_rate": 4.895227791225228e-06,
+      "loss": 0.2999,
+      "mean_token_accuracy": 0.905066967010498,
+      "num_tokens": 30767481.0,
+      "step": 3790
+    },
+    {
+      "entropy": 0.4244759321212769,
+      "epoch": 2.141647855530474,
+      "grad_norm": 1.6923705339431763,
+      "learning_rate": 4.894953705043774e-06,
+      "loss": 0.3151,
+      "mean_token_accuracy": 0.901056170463562,
+      "num_tokens": 30808039.0,
+      "step": 3795
+    },
+    {
+      "entropy": 0.40609169006347656,
+      "epoch": 2.144469525959368,
+      "grad_norm": 1.470828652381897,
+      "learning_rate": 4.894679271154684e-06,
+      "loss": 0.3031,
+      "mean_token_accuracy": 0.9032929420471192,
+      "num_tokens": 30848528.0,
+      "step": 3800
+    },
+    {
+      "entropy": 0.38831552267074587,
+      "epoch": 2.1472911963882617,
+      "grad_norm": 1.4154300689697266,
+      "learning_rate": 4.894404489611872e-06,
+      "loss": 0.3071,
+      "mean_token_accuracy": 0.9009493231773377,
+      "num_tokens": 30889149.0,
+      "step": 3805
+    },
+    {
+      "entropy": 0.36803258061408994,
+      "epoch": 2.150112866817156,
+      "grad_norm": 1.5650067329406738,
+      "learning_rate": 4.894129360469317e-06,
+      "loss": 0.2586,
+      "mean_token_accuracy": 0.9152506470680237,
+      "num_tokens": 30929871.0,
+      "step": 3810
+    },
+    {
+      "entropy": 0.41122249364852903,
+      "epoch": 2.1529345372460496,
+      "grad_norm": 1.7758678197860718,
+      "learning_rate": 4.89385388378107e-06,
+      "loss": 0.3143,
+      "mean_token_accuracy": 0.9019395470619201,
+      "num_tokens": 30970228.0,
+      "step": 3815
+    },
+    {
+      "entropy": 0.3893428444862366,
+      "epoch": 2.1557562076749437,
+      "grad_norm": 1.581362009048462,
+      "learning_rate": 4.893578059601249e-06,
+      "loss": 0.2896,
+      "mean_token_accuracy": 0.9072283625602722,
+      "num_tokens": 31010919.0,
+      "step": 3820
+    },
+    {
+      "entropy": 0.3995703637599945,
+      "epoch": 2.1585778781038374,
+      "grad_norm": 1.6888242959976196,
+      "learning_rate": 4.893301887984036e-06,
+      "loss": 0.3082,
+      "mean_token_accuracy": 0.8979094982147217,
+      "num_tokens": 31051717.0,
+      "step": 3825
+    },
+    {
+      "entropy": 0.3909620702266693,
+      "epoch": 2.1613995485327315,
+      "grad_norm": 1.589010238647461,
+      "learning_rate": 4.893025368983688e-06,
+      "loss": 0.2876,
+      "mean_token_accuracy": 0.9069844603538513,
+      "num_tokens": 31092303.0,
+      "step": 3830
+    },
+    {
+      "entropy": 0.38756829500198364,
+      "epoch": 2.1642212189616252,
+      "grad_norm": 1.4219690561294556,
+      "learning_rate": 4.892748502654527e-06,
+      "loss": 0.2725,
+      "mean_token_accuracy": 0.9117750763893128,
+      "num_tokens": 31133212.0,
+      "step": 3835
+    },
+    {
+      "entropy": 0.39158605933189394,
+      "epoch": 2.1670428893905194,
+      "grad_norm": 1.5833606719970703,
+      "learning_rate": 4.892471289050942e-06,
+      "loss": 0.2781,
+      "mean_token_accuracy": 0.9098140478134156,
+      "num_tokens": 31173102.0,
+      "step": 3840
+    },
+    {
+      "entropy": 0.4036704897880554,
+      "epoch": 2.169864559819413,
+      "grad_norm": 1.4913930892944336,
+      "learning_rate": 4.892193728227393e-06,
+      "loss": 0.3055,
+      "mean_token_accuracy": 0.9021237850189209,
+      "num_tokens": 31213865.0,
+      "step": 3845
+    },
+    {
+      "entropy": 0.394330632686615,
+      "epoch": 2.172686230248307,
+      "grad_norm": 1.8555155992507935,
+      "learning_rate": 4.891915820238406e-06,
+      "loss": 0.3095,
+      "mean_token_accuracy": 0.8976067423820495,
+      "num_tokens": 31254408.0,
+      "step": 3850
+    },
+    {
+      "entropy": 0.4108275890350342,
+      "epoch": 2.175507900677201,
+      "grad_norm": 1.4580622911453247,
+      "learning_rate": 4.891637565138578e-06,
+      "loss": 0.3162,
+      "mean_token_accuracy": 0.8978480458259582,
+      "num_tokens": 31295211.0,
+      "step": 3855
+    },
+    {
+      "entropy": 0.42848613262176516,
+      "epoch": 2.1783295711060946,
+      "grad_norm": 1.54513680934906,
+      "learning_rate": 4.891358962982569e-06,
+      "loss": 0.308,
+      "mean_token_accuracy": 0.9015236139297486,
+      "num_tokens": 31335702.0,
+      "step": 3860
+    },
+    {
+      "entropy": 0.4215279221534729,
+      "epoch": 2.1811512415349887,
+      "grad_norm": 1.7928465604782104,
+      "learning_rate": 4.891080013825112e-06,
+      "loss": 0.3083,
+      "mean_token_accuracy": 0.9013218641281128,
+      "num_tokens": 31376407.0,
+      "step": 3865
+    },
+    {
+      "entropy": 0.4218919575214386,
+      "epoch": 2.1839729119638824,
+      "grad_norm": 1.5700446367263794,
+      "learning_rate": 4.890800717721007e-06,
+      "loss": 0.3067,
+      "mean_token_accuracy": 0.9020668745040894,
+      "num_tokens": 31416993.0,
+      "step": 3870
+    },
+    {
+      "entropy": 0.40297368764877317,
+      "epoch": 2.1867945823927766,
+      "grad_norm": 1.5355396270751953,
+      "learning_rate": 4.890521074725122e-06,
+      "loss": 0.2787,
+      "mean_token_accuracy": 0.9106267809867858,
+      "num_tokens": 31457735.0,
+      "step": 3875
+    },
+    {
+      "entropy": 0.4444594144821167,
+      "epoch": 2.1896162528216703,
+      "grad_norm": 1.762122631072998,
+      "learning_rate": 4.890241084892392e-06,
+      "loss": 0.324,
+      "mean_token_accuracy": 0.8961127758026123,
+      "num_tokens": 31498437.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.39649640321731566,
+      "epoch": 2.1924379232505644,
+      "grad_norm": 1.3777143955230713,
+      "learning_rate": 4.889960748277821e-06,
+      "loss": 0.2963,
+      "mean_token_accuracy": 0.9072598814964294,
+      "num_tokens": 31539261.0,
+      "step": 3885
+    },
+    {
+      "entropy": 0.36735162138938904,
+      "epoch": 2.195259593679458,
+      "grad_norm": 1.510568618774414,
+      "learning_rate": 4.889680064936483e-06,
+      "loss": 0.2806,
+      "mean_token_accuracy": 0.9091784954071045,
+      "num_tokens": 31580151.0,
+      "step": 3890
+    },
+    {
+      "entropy": 0.3932835698127747,
+      "epoch": 2.1980812641083523,
+      "grad_norm": 1.311471939086914,
+      "learning_rate": 4.889399034923515e-06,
+      "loss": 0.2901,
+      "mean_token_accuracy": 0.9054880023002625,
+      "num_tokens": 31620726.0,
+      "step": 3895
+    },
+    {
+      "entropy": 0.4081146061420441,
+      "epoch": 2.200902934537246,
+      "grad_norm": 1.7176251411437988,
+      "learning_rate": 4.889117658294128e-06,
+      "loss": 0.3022,
+      "mean_token_accuracy": 0.9020537614822388,
+      "num_tokens": 31661526.0,
+      "step": 3900
+    },
+    {
+      "entropy": 0.4014622807502747,
+      "epoch": 2.20372460496614,
+      "grad_norm": 1.5755354166030884,
+      "learning_rate": 4.888835935103598e-06,
+      "loss": 0.3051,
+      "mean_token_accuracy": 0.9023280262947082,
+      "num_tokens": 31702085.0,
+      "step": 3905
+    },
+    {
+      "entropy": 0.39336706399917604,
+      "epoch": 2.206546275395034,
+      "grad_norm": 1.734265923500061,
+      "learning_rate": 4.888553865407269e-06,
+      "loss": 0.2862,
+      "mean_token_accuracy": 0.905624508857727,
+      "num_tokens": 31742379.0,
+      "step": 3910
+    },
+    {
+      "entropy": 0.3690111577510834,
+      "epoch": 2.209367945823928,
+      "grad_norm": 1.5115963220596313,
+      "learning_rate": 4.888271449260554e-06,
+      "loss": 0.26,
+      "mean_token_accuracy": 0.9165234088897705,
+      "num_tokens": 31783016.0,
+      "step": 3915
+    },
+    {
+      "entropy": 0.3873672723770142,
+      "epoch": 2.2121896162528216,
+      "grad_norm": 1.7038564682006836,
+      "learning_rate": 4.887988686718933e-06,
+      "loss": 0.2887,
+      "mean_token_accuracy": 0.9066754698753356,
+      "num_tokens": 31823595.0,
+      "step": 3920
+    },
+    {
+      "entropy": 0.3954701006412506,
+      "epoch": 2.2150112866817158,
+      "grad_norm": 1.6473407745361328,
+      "learning_rate": 4.887705577837957e-06,
+      "loss": 0.2929,
+      "mean_token_accuracy": 0.9045971989631653,
+      "num_tokens": 31864135.0,
+      "step": 3925
+    },
+    {
+      "entropy": 0.38976659178733825,
+      "epoch": 2.2178329571106095,
+      "grad_norm": 1.4724841117858887,
+      "learning_rate": 4.88742212267324e-06,
+      "loss": 0.2907,
+      "mean_token_accuracy": 0.9068511605262757,
+      "num_tokens": 31904819.0,
+      "step": 3930
+    },
+    {
+      "entropy": 0.40335485339164734,
+      "epoch": 2.2206546275395036,
+      "grad_norm": 1.5783131122589111,
+      "learning_rate": 4.887138321280468e-06,
+      "loss": 0.3003,
+      "mean_token_accuracy": 0.9015839457511902,
+      "num_tokens": 31945530.0,
+      "step": 3935
+    },
+    {
+      "entropy": 0.3923260450363159,
+      "epoch": 2.2234762979683973,
+      "grad_norm": 1.4128057956695557,
+      "learning_rate": 4.886854173715393e-06,
+      "loss": 0.2819,
+      "mean_token_accuracy": 0.9102500557899476,
+      "num_tokens": 31986171.0,
+      "step": 3940
+    },
+    {
+      "entropy": 0.4288273572921753,
+      "epoch": 2.226297968397291,
+      "grad_norm": 1.6166179180145264,
+      "learning_rate": 4.886569680033837e-06,
+      "loss": 0.3106,
+      "mean_token_accuracy": 0.8994893431663513,
+      "num_tokens": 32026712.0,
+      "step": 3945
+    },
+    {
+      "entropy": 0.4336235702037811,
+      "epoch": 2.229119638826185,
+      "grad_norm": 1.6575360298156738,
+      "learning_rate": 4.886284840291689e-06,
+      "loss": 0.3289,
+      "mean_token_accuracy": 0.8974316120147705,
+      "num_tokens": 32067338.0,
+      "step": 3950
+    },
+    {
+      "entropy": 0.40959821343421937,
+      "epoch": 2.231941309255079,
+      "grad_norm": 1.4833348989486694,
+      "learning_rate": 4.885999654544904e-06,
+      "loss": 0.3031,
+      "mean_token_accuracy": 0.9015701293945313,
+      "num_tokens": 32108045.0,
+      "step": 3955
+    },
+    {
+      "entropy": 0.43567387461662294,
+      "epoch": 2.234762979683973,
+      "grad_norm": 1.6455271244049072,
+      "learning_rate": 4.885714122849509e-06,
+      "loss": 0.3236,
+      "mean_token_accuracy": 0.8950015425682067,
+      "num_tokens": 32148724.0,
+      "step": 3960
+    },
+    {
+      "entropy": 0.41589171886444093,
+      "epoch": 2.2375846501128667,
+      "grad_norm": 1.6400325298309326,
+      "learning_rate": 4.885428245261596e-06,
+      "loss": 0.3149,
+      "mean_token_accuracy": 0.8991590142250061,
+      "num_tokens": 32189079.0,
+      "step": 3965
+    },
+    {
+      "entropy": 0.4249677717685699,
+      "epoch": 2.240406320541761,
+      "grad_norm": 2.002429485321045,
+      "learning_rate": 4.885142021837323e-06,
+      "loss": 0.3116,
+      "mean_token_accuracy": 0.9002622365951538,
+      "num_tokens": 32229786.0,
+      "step": 3970
+    },
+    {
+      "entropy": 0.42671540975570676,
+      "epoch": 2.2432279909706545,
+      "grad_norm": 1.9370062351226807,
+      "learning_rate": 4.8848554526329236e-06,
+      "loss": 0.3381,
+      "mean_token_accuracy": 0.8927700161933899,
+      "num_tokens": 32270581.0,
+      "step": 3975
+    },
+    {
+      "entropy": 0.4029369592666626,
+      "epoch": 2.2460496613995486,
+      "grad_norm": 1.457308053970337,
+      "learning_rate": 4.88456853770469e-06,
+      "loss": 0.299,
+      "mean_token_accuracy": 0.9022388100624085,
+      "num_tokens": 32311159.0,
+      "step": 3980
+    },
+    {
+      "entropy": 0.4298437058925629,
+      "epoch": 2.2488713318284423,
+      "grad_norm": 1.4605598449707031,
+      "learning_rate": 4.88428127710899e-06,
+      "loss": 0.3317,
+      "mean_token_accuracy": 0.8953496932983398,
+      "num_tokens": 32351769.0,
+      "step": 3985
+    },
+    {
+      "entropy": 0.4445444464683533,
+      "epoch": 2.2516930022573365,
+      "grad_norm": 1.861961841583252,
+      "learning_rate": 4.883993670902254e-06,
+      "loss": 0.3314,
+      "mean_token_accuracy": 0.895211398601532,
+      "num_tokens": 32392495.0,
+      "step": 3990
+    },
+    {
+      "entropy": 0.4017652451992035,
+      "epoch": 2.25451467268623,
+      "grad_norm": 1.816008448600769,
+      "learning_rate": 4.883705719140982e-06,
+      "loss": 0.3018,
+      "mean_token_accuracy": 0.9027621269226074,
+      "num_tokens": 32432976.0,
+      "step": 3995
+    },
+    {
+      "entropy": 0.3940592765808105,
+      "epoch": 2.2573363431151243,
+      "grad_norm": 1.5475646257400513,
+      "learning_rate": 4.883417421881744e-06,
+      "loss": 0.294,
+      "mean_token_accuracy": 0.9063061594963073,
+      "num_tokens": 32473442.0,
+      "step": 4000
+    },
+    {
+      "epoch": 2.2573363431151243,
+      "eval_entropy": 0.402061402797699,
+      "eval_loss": 0.27405405044555664,
+      "eval_mean_token_accuracy": 0.9164431095123291,
+      "eval_num_tokens": 32473442.0,
+      "eval_runtime": 0.1642,
+      "eval_samples_per_second": 24.354,
+      "eval_steps_per_second": 6.089,
+      "step": 4000
+    },
+    {
+      "entropy": 0.4588034152984619,
+      "epoch": 2.260158013544018,
+      "grad_norm": 1.6592538356781006,
+      "learning_rate": 4.883128779181174e-06,
+      "loss": 0.3559,
+      "mean_token_accuracy": 0.889747953414917,
+      "num_tokens": 32514004.0,
+      "step": 4005
+    },
+    {
+      "entropy": 0.3681566655635834,
+      "epoch": 2.2629796839729117,
+      "grad_norm": 1.5150611400604248,
+      "learning_rate": 4.8828397910959766e-06,
+      "loss": 0.2613,
+      "mean_token_accuracy": 0.9151654362678527,
+      "num_tokens": 32554858.0,
+      "step": 4010
+    },
+    {
+      "entropy": 0.40542457699775697,
+      "epoch": 2.265801354401806,
+      "grad_norm": 1.5727524757385254,
+      "learning_rate": 4.882550457682924e-06,
+      "loss": 0.3099,
+      "mean_token_accuracy": 0.8985041737556457,
+      "num_tokens": 32595341.0,
+      "step": 4015
+    },
+    {
+      "entropy": 0.38482665419578554,
+      "epoch": 2.2686230248307,
+      "grad_norm": 1.5396711826324463,
+      "learning_rate": 4.8822607789988565e-06,
+      "loss": 0.2944,
+      "mean_token_accuracy": 0.9044581055641174,
+      "num_tokens": 32636025.0,
+      "step": 4020
+    },
+    {
+      "entropy": 0.4148333430290222,
+      "epoch": 2.2714446952595937,
+      "grad_norm": 1.5504050254821777,
+      "learning_rate": 4.881970755100679e-06,
+      "loss": 0.3041,
+      "mean_token_accuracy": 0.901849901676178,
+      "num_tokens": 32676078.0,
+      "step": 4025
+    },
+    {
+      "entropy": 0.4355687737464905,
+      "epoch": 2.2742663656884874,
+      "grad_norm": 3.5764992237091064,
+      "learning_rate": 4.88168038604537e-06,
+      "loss": 0.3248,
+      "mean_token_accuracy": 0.8947890400886536,
+      "num_tokens": 32716756.0,
+      "step": 4030
+    },
+    {
+      "entropy": 0.4102937400341034,
+      "epoch": 2.2770880361173815,
+      "grad_norm": 1.5980452299118042,
+      "learning_rate": 4.881389671889969e-06,
+      "loss": 0.3088,
+      "mean_token_accuracy": 0.9003346800804138,
+      "num_tokens": 32757375.0,
+      "step": 4035
+    },
+    {
+      "entropy": 0.4062139928340912,
+      "epoch": 2.2799097065462752,
+      "grad_norm": 1.4203381538391113,
+      "learning_rate": 4.881098612691589e-06,
+      "loss": 0.2957,
+      "mean_token_accuracy": 0.9053402423858643,
+      "num_tokens": 32797973.0,
+      "step": 4040
+    },
+    {
+      "entropy": 0.42087835669517515,
+      "epoch": 2.2827313769751694,
+      "grad_norm": 1.6059361696243286,
+      "learning_rate": 4.880807208507409e-06,
+      "loss": 0.3123,
+      "mean_token_accuracy": 0.8996002554893494,
+      "num_tokens": 32838742.0,
+      "step": 4045
+    },
+    {
+      "entropy": 0.3775734961032867,
+      "epoch": 2.285553047404063,
+      "grad_norm": 1.5873465538024902,
+      "learning_rate": 4.880515459394674e-06,
+      "loss": 0.2564,
+      "mean_token_accuracy": 0.9160468459129334,
+      "num_tokens": 32879298.0,
+      "step": 4050
+    },
+    {
+      "entropy": 0.3923003554344177,
+      "epoch": 2.288374717832957,
+      "grad_norm": 1.7001484632492065,
+      "learning_rate": 4.880223365410699e-06,
+      "loss": 0.3002,
+      "mean_token_accuracy": 0.9026227593421936,
+      "num_tokens": 32920082.0,
+      "step": 4055
+    },
+    {
+      "entropy": 0.41630412340164186,
+      "epoch": 2.291196388261851,
+      "grad_norm": 1.824363350868225,
+      "learning_rate": 4.879930926612866e-06,
+      "loss": 0.3176,
+      "mean_token_accuracy": 0.8960343599319458,
+      "num_tokens": 32960716.0,
+      "step": 4060
+    },
+    {
+      "entropy": 0.4242856979370117,
+      "epoch": 2.294018058690745,
+      "grad_norm": 1.5266849994659424,
+      "learning_rate": 4.879638143058625e-06,
+      "loss": 0.3006,
+      "mean_token_accuracy": 0.9029630541801452,
+      "num_tokens": 33001292.0,
+      "step": 4065
+    },
+    {
+      "entropy": 0.4204494059085846,
+      "epoch": 2.2968397291196387,
+      "grad_norm": 1.8229271173477173,
+      "learning_rate": 4.879345014805491e-06,
+      "loss": 0.3137,
+      "mean_token_accuracy": 0.9004954934120178,
+      "num_tokens": 33041946.0,
+      "step": 4070
+    },
+    {
+      "entropy": 0.40997507572174074,
+      "epoch": 2.299661399548533,
+      "grad_norm": 1.7101879119873047,
+      "learning_rate": 4.8790515419110516e-06,
+      "loss": 0.3036,
+      "mean_token_accuracy": 0.9026644349098205,
+      "num_tokens": 33082717.0,
+      "step": 4075
+    },
+    {
+      "entropy": 0.3944500207901001,
+      "epoch": 2.3024830699774266,
+      "grad_norm": 1.7713676691055298,
+      "learning_rate": 4.8787577244329585e-06,
+      "loss": 0.2867,
+      "mean_token_accuracy": 0.9076826214790344,
+      "num_tokens": 33123211.0,
+      "step": 4080
+    },
+    {
+      "entropy": 0.4161556899547577,
+      "epoch": 2.3053047404063207,
+      "grad_norm": 1.6283953189849854,
+      "learning_rate": 4.878463562428933e-06,
+      "loss": 0.317,
+      "mean_token_accuracy": 0.8988382816314697,
+      "num_tokens": 33164049.0,
+      "step": 4085
+    },
+    {
+      "entropy": 0.4177008092403412,
+      "epoch": 2.3081264108352144,
+      "grad_norm": 1.6907880306243896,
+      "learning_rate": 4.878169055956763e-06,
+      "loss": 0.3069,
+      "mean_token_accuracy": 0.8997668266296387,
+      "num_tokens": 33204881.0,
+      "step": 4090
+    },
+    {
+      "entropy": 0.3897433698177338,
+      "epoch": 2.310948081264108,
+      "grad_norm": 1.4863002300262451,
+      "learning_rate": 4.877874205074303e-06,
+      "loss": 0.2881,
+      "mean_token_accuracy": 0.9074371457099915,
+      "num_tokens": 33245356.0,
+      "step": 4095
+    },
+    {
+      "entropy": 0.41926553249359133,
+      "epoch": 2.3137697516930023,
+      "grad_norm": 1.7184562683105469,
+      "learning_rate": 4.877579009839478e-06,
+      "loss": 0.3263,
+      "mean_token_accuracy": 0.8960996985435485,
+      "num_tokens": 33285939.0,
+      "step": 4100
+    },
+    {
+      "entropy": 0.39500190019607545,
+      "epoch": 2.3165914221218964,
+      "grad_norm": 1.7541210651397705,
+      "learning_rate": 4.877283470310279e-06,
+      "loss": 0.2844,
+      "mean_token_accuracy": 0.9074315071105957,
+      "num_tokens": 33326588.0,
+      "step": 4105
+    },
+    {
+      "entropy": 0.40492220520973204,
+      "epoch": 2.31941309255079,
+      "grad_norm": 1.7052255868911743,
+      "learning_rate": 4.876987586544765e-06,
+      "loss": 0.3244,
+      "mean_token_accuracy": 0.8954932928085327,
+      "num_tokens": 33367169.0,
+      "step": 4110
+    },
+    {
+      "entropy": 0.38237152695655824,
+      "epoch": 2.322234762979684,
+      "grad_norm": 1.6368436813354492,
+      "learning_rate": 4.876691358601061e-06,
+      "loss": 0.2849,
+      "mean_token_accuracy": 0.9079111337661743,
+      "num_tokens": 33407822.0,
+      "step": 4115
+    },
+    {
+      "entropy": 0.42206219434738157,
+      "epoch": 2.325056433408578,
+      "grad_norm": 1.7534505128860474,
+      "learning_rate": 4.876394786537362e-06,
+      "loss": 0.3165,
+      "mean_token_accuracy": 0.899584424495697,
+      "num_tokens": 33448607.0,
+      "step": 4120
+    },
+    {
+      "entropy": 0.3744909644126892,
+      "epoch": 2.3278781038374716,
+      "grad_norm": 1.4802114963531494,
+      "learning_rate": 4.87609787041193e-06,
+      "loss": 0.278,
+      "mean_token_accuracy": 0.9111004710197449,
+      "num_tokens": 33489164.0,
+      "step": 4125
+    },
+    {
+      "entropy": 0.3757921814918518,
+      "epoch": 2.3306997742663658,
+      "grad_norm": 1.6543906927108765,
+      "learning_rate": 4.875800610283092e-06,
+      "loss": 0.2765,
+      "mean_token_accuracy": 0.9108646512031555,
+      "num_tokens": 33529990.0,
+      "step": 4130
+    },
+    {
+      "entropy": 0.42251219749450686,
+      "epoch": 2.3335214446952595,
+      "grad_norm": 1.6640608310699463,
+      "learning_rate": 4.875503006209249e-06,
+      "loss": 0.334,
+      "mean_token_accuracy": 0.8940789937973023,
+      "num_tokens": 33570747.0,
+      "step": 4135
+    },
+    {
+      "entropy": 0.3960157215595245,
+      "epoch": 2.3363431151241536,
+      "grad_norm": 1.8502579927444458,
+      "learning_rate": 4.875205058248861e-06,
+      "loss": 0.2835,
+      "mean_token_accuracy": 0.9076458573341369,
+      "num_tokens": 33611154.0,
+      "step": 4140
+    },
+    {
+      "entropy": 0.41452285647392273,
+      "epoch": 2.3391647855530473,
+      "grad_norm": 1.5899200439453125,
+      "learning_rate": 4.874906766460463e-06,
+      "loss": 0.314,
+      "mean_token_accuracy": 0.9006463170051575,
+      "num_tokens": 33651898.0,
+      "step": 4145
+    },
+    {
+      "entropy": 0.3915266990661621,
+      "epoch": 2.3419864559819414,
+      "grad_norm": 1.3926752805709839,
+      "learning_rate": 4.874608130902653e-06,
+      "loss": 0.2881,
+      "mean_token_accuracy": 0.9062711000442505,
+      "num_tokens": 33692461.0,
+      "step": 4150
+    },
+    {
+      "entropy": 0.380002897977829,
+      "epoch": 2.344808126410835,
+      "grad_norm": 1.4338423013687134,
+      "learning_rate": 4.874309151634098e-06,
+      "loss": 0.2729,
+      "mean_token_accuracy": 0.9096962571144104,
+      "num_tokens": 33732460.0,
+      "step": 4155
+    },
+    {
+      "entropy": 0.40470706224441527,
+      "epoch": 2.3476297968397293,
+      "grad_norm": 1.7443944215774536,
+      "learning_rate": 4.874009828713532e-06,
+      "loss": 0.2983,
+      "mean_token_accuracy": 0.9021994709968567,
+      "num_tokens": 33773109.0,
+      "step": 4160
+    },
+    {
+      "entropy": 0.3907041311264038,
+      "epoch": 2.350451467268623,
+      "grad_norm": 1.3750447034835815,
+      "learning_rate": 4.873710162199759e-06,
+      "loss": 0.2994,
+      "mean_token_accuracy": 0.9032772421836853,
+      "num_tokens": 33813658.0,
+      "step": 4165
+    },
+    {
+      "entropy": 0.42153160572052,
+      "epoch": 2.353273137697517,
+      "grad_norm": 1.5966558456420898,
+      "learning_rate": 4.873410152151648e-06,
+      "loss": 0.3312,
+      "mean_token_accuracy": 0.89634268283844,
+      "num_tokens": 33854046.0,
+      "step": 4170
+    },
+    {
+      "entropy": 0.39306110739707945,
+      "epoch": 2.356094808126411,
+      "grad_norm": 1.5900177955627441,
+      "learning_rate": 4.873109798628133e-06,
+      "loss": 0.3014,
+      "mean_token_accuracy": 0.902882993221283,
+      "num_tokens": 33894451.0,
+      "step": 4175
+    },
+    {
+      "entropy": 0.3619803309440613,
+      "epoch": 2.3589164785553045,
+      "grad_norm": 1.5183881521224976,
+      "learning_rate": 4.872809101688222e-06,
+      "loss": 0.2644,
+      "mean_token_accuracy": 0.9138602614402771,
+      "num_tokens": 33935166.0,
+      "step": 4180
+    },
+    {
+      "entropy": 0.40472997426986695,
+      "epoch": 2.3617381489841986,
+      "grad_norm": 1.5294315814971924,
+      "learning_rate": 4.872508061390986e-06,
+      "loss": 0.3044,
+      "mean_token_accuracy": 0.9031421303749084,
+      "num_tokens": 33975747.0,
+      "step": 4185
+    },
+    {
+      "entropy": 0.39990630745887756,
+      "epoch": 2.364559819413093,
+      "grad_norm": 1.6503218412399292,
+      "learning_rate": 4.872206677795564e-06,
+      "loss": 0.2919,
+      "mean_token_accuracy": 0.9047986268997192,
+      "num_tokens": 34016543.0,
+      "step": 4190
+    },
+    {
+      "entropy": 0.3896356463432312,
+      "epoch": 2.3673814898419865,
+      "grad_norm": 1.6634739637374878,
+      "learning_rate": 4.871904950961163e-06,
+      "loss": 0.2923,
+      "mean_token_accuracy": 0.9031400918960572,
+      "num_tokens": 34057312.0,
+      "step": 4195
+    },
+    {
+      "entropy": 0.41114925742149355,
+      "epoch": 2.37020316027088,
+      "grad_norm": 1.5319340229034424,
+      "learning_rate": 4.871602880947058e-06,
+      "loss": 0.3089,
+      "mean_token_accuracy": 0.9023614287376404,
+      "num_tokens": 34098081.0,
+      "step": 4200
+    },
+    {
+      "entropy": 0.4191340744495392,
+      "epoch": 2.3730248306997743,
+      "grad_norm": 1.60099196434021,
+      "learning_rate": 4.871300467812589e-06,
+      "loss": 0.3044,
+      "mean_token_accuracy": 0.9024178624153137,
+      "num_tokens": 34138698.0,
+      "step": 4205
+    },
+    {
+      "entropy": 0.4011139750480652,
+      "epoch": 2.375846501128668,
+      "grad_norm": 1.5579471588134766,
+      "learning_rate": 4.870997711617166e-06,
+      "loss": 0.3093,
+      "mean_token_accuracy": 0.9014039039611816,
+      "num_tokens": 34179294.0,
+      "step": 4210
+    },
+    {
+      "entropy": 0.38470744490623476,
+      "epoch": 2.378668171557562,
+      "grad_norm": 1.4612574577331543,
+      "learning_rate": 4.8706946124202666e-06,
+      "loss": 0.2772,
+      "mean_token_accuracy": 0.9105475664138794,
+      "num_tokens": 34219877.0,
+      "step": 4215
+    },
+    {
+      "entropy": 0.4085557281970978,
+      "epoch": 2.381489841986456,
+      "grad_norm": 1.6769572496414185,
+      "learning_rate": 4.8703911702814326e-06,
+      "loss": 0.3239,
+      "mean_token_accuracy": 0.8968191981315613,
+      "num_tokens": 34260392.0,
+      "step": 4220
+    },
+    {
+      "entropy": 0.4069161355495453,
+      "epoch": 2.38431151241535,
+      "grad_norm": 1.8266639709472656,
+      "learning_rate": 4.870087385260277e-06,
+      "loss": 0.294,
+      "mean_token_accuracy": 0.9052396297454834,
+      "num_tokens": 34301087.0,
+      "step": 4225
+    },
+    {
+      "entropy": 0.41574978828430176,
+      "epoch": 2.3871331828442437,
+      "grad_norm": 1.4496681690216064,
+      "learning_rate": 4.8697832574164786e-06,
+      "loss": 0.3123,
+      "mean_token_accuracy": 0.8990219950675964,
+      "num_tokens": 34341510.0,
+      "step": 4230
+    },
+    {
+      "entropy": 0.42293376922607423,
+      "epoch": 2.389954853273138,
+      "grad_norm": 1.5617249011993408,
+      "learning_rate": 4.86947878680978e-06,
+      "loss": 0.3288,
+      "mean_token_accuracy": 0.8938201785087585,
+      "num_tokens": 34382177.0,
+      "step": 4235
+    },
+    {
+      "entropy": 0.43548256158828735,
+      "epoch": 2.3927765237020315,
+      "grad_norm": 1.653110146522522,
+      "learning_rate": 4.869173973499999e-06,
+      "loss": 0.3322,
+      "mean_token_accuracy": 0.8943361163139343,
+      "num_tokens": 34422859.0,
+      "step": 4240
+    },
+    {
+      "entropy": 0.4331204891204834,
+      "epoch": 2.3955981941309257,
+      "grad_norm": 1.483518123626709,
+      "learning_rate": 4.868868817547013e-06,
+      "loss": 0.3349,
+      "mean_token_accuracy": 0.8963463306427002,
+      "num_tokens": 34463566.0,
+      "step": 4245
+    },
+    {
+      "entropy": 0.40947118401527405,
+      "epoch": 2.3984198645598194,
+      "grad_norm": 1.3235682249069214,
+      "learning_rate": 4.868563319010772e-06,
+      "loss": 0.2896,
+      "mean_token_accuracy": 0.9062234163284302,
+      "num_tokens": 34504064.0,
+      "step": 4250
+    },
+    {
+      "entropy": 0.4543120741844177,
+      "epoch": 2.4012415349887135,
+      "grad_norm": 1.6390918493270874,
+      "learning_rate": 4.86825747795129e-06,
+      "loss": 0.3415,
+      "mean_token_accuracy": 0.8885816693305969,
+      "num_tokens": 34544665.0,
+      "step": 4255
+    },
+    {
+      "entropy": 0.42363797426223754,
+      "epoch": 2.404063205417607,
+      "grad_norm": 1.629384160041809,
+      "learning_rate": 4.86795129442865e-06,
+      "loss": 0.3201,
+      "mean_token_accuracy": 0.8973617792129517,
+      "num_tokens": 34585268.0,
+      "step": 4260
+    },
+    {
+      "entropy": 0.3826207935810089,
+      "epoch": 2.406884875846501,
+      "grad_norm": 1.5705349445343018,
+      "learning_rate": 4.867644768503002e-06,
+      "loss": 0.2778,
+      "mean_token_accuracy": 0.9083486676216126,
+      "num_tokens": 34625843.0,
+      "step": 4265
+    },
+    {
+      "entropy": 0.38043850660324097,
+      "epoch": 2.409706546275395,
+      "grad_norm": 1.3238205909729004,
+      "learning_rate": 4.867337900234562e-06,
+      "loss": 0.28,
+      "mean_token_accuracy": 0.9097347378730773,
+      "num_tokens": 34666723.0,
+      "step": 4270
+    },
+    {
+      "entropy": 0.40482637882232664,
+      "epoch": 2.412528216704289,
+      "grad_norm": 1.4541096687316895,
+      "learning_rate": 4.867030689683615e-06,
+      "loss": 0.2967,
+      "mean_token_accuracy": 0.9034028768539428,
+      "num_tokens": 34707314.0,
+      "step": 4275
+    },
+    {
+      "entropy": 0.3765712082386017,
+      "epoch": 2.415349887133183,
+      "grad_norm": 1.694923996925354,
+      "learning_rate": 4.8667231369105126e-06,
+      "loss": 0.2784,
+      "mean_token_accuracy": 0.9113203883171082,
+      "num_tokens": 34748089.0,
+      "step": 4280
+    },
+    {
+      "entropy": 0.41379693150520325,
+      "epoch": 2.4181715575620766,
+      "grad_norm": 1.8324544429779053,
+      "learning_rate": 4.866415241975674e-06,
+      "loss": 0.3173,
+      "mean_token_accuracy": 0.8990246057510376,
+      "num_tokens": 34788615.0,
+      "step": 4285
+    },
+    {
+      "entropy": 0.405905282497406,
+      "epoch": 2.4209932279909707,
+      "grad_norm": 1.5715994834899902,
+      "learning_rate": 4.866107004939584e-06,
+      "loss": 0.2953,
+      "mean_token_accuracy": 0.9052024960517884,
+      "num_tokens": 34829247.0,
+      "step": 4290
+    },
+    {
+      "entropy": 0.38263545036315916,
+      "epoch": 2.4238148984198644,
+      "grad_norm": 1.4427995681762695,
+      "learning_rate": 4.865798425862797e-06,
+      "loss": 0.2802,
+      "mean_token_accuracy": 0.9089892625808715,
+      "num_tokens": 34869192.0,
+      "step": 4295
+    },
+    {
+      "entropy": 0.41300634741783143,
+      "epoch": 2.4266365688487586,
+      "grad_norm": 1.6823080778121948,
+      "learning_rate": 4.865489504805933e-06,
+      "loss": 0.3148,
+      "mean_token_accuracy": 0.8993620634078979,
+      "num_tokens": 34909855.0,
+      "step": 4300
+    },
+    {
+      "entropy": 0.3476841390132904,
+      "epoch": 2.4294582392776523,
+      "grad_norm": 1.5836542844772339,
+      "learning_rate": 4.865180241829679e-06,
+      "loss": 0.2606,
+      "mean_token_accuracy": 0.915683650970459,
+      "num_tokens": 34950497.0,
+      "step": 4305
+    },
+    {
+      "entropy": 0.40133320689201357,
+      "epoch": 2.4322799097065464,
+      "grad_norm": 1.5993763208389282,
+      "learning_rate": 4.8648706369947915e-06,
+      "loss": 0.3077,
+      "mean_token_accuracy": 0.9001999020576477,
+      "num_tokens": 34990946.0,
+      "step": 4310
+    },
+    {
+      "entropy": 0.43380234241485593,
+      "epoch": 2.43510158013544,
+      "grad_norm": 1.6708790063858032,
+      "learning_rate": 4.86456069036209e-06,
+      "loss": 0.3327,
+      "mean_token_accuracy": 0.8949172139167786,
+      "num_tokens": 35031584.0,
+      "step": 4315
+    },
+    {
+      "entropy": 0.36903883814811705,
+      "epoch": 2.4379232505643342,
+      "grad_norm": 1.705231785774231,
+      "learning_rate": 4.864250401992465e-06,
+      "loss": 0.2643,
+      "mean_token_accuracy": 0.9130355954170227,
+      "num_tokens": 35072066.0,
+      "step": 4320
+    },
+    {
+      "entropy": 0.4150381624698639,
+      "epoch": 2.440744920993228,
+      "grad_norm": 1.5706596374511719,
+      "learning_rate": 4.863939771946873e-06,
+      "loss": 0.3097,
+      "mean_token_accuracy": 0.8974992752075195,
+      "num_tokens": 35112703.0,
+      "step": 4325
+    },
+    {
+      "entropy": 0.4148720264434814,
+      "epoch": 2.443566591422122,
+      "grad_norm": 1.6442214250564575,
+      "learning_rate": 4.863628800286337e-06,
+      "loss": 0.3247,
+      "mean_token_accuracy": 0.8946601033210755,
+      "num_tokens": 35153539.0,
+      "step": 4330
+    },
+    {
+      "entropy": 0.40285355448722837,
+      "epoch": 2.4463882618510158,
+      "grad_norm": 1.4223581552505493,
+      "learning_rate": 4.863317487071946e-06,
+      "loss": 0.3108,
+      "mean_token_accuracy": 0.8984290480613708,
+      "num_tokens": 35194160.0,
+      "step": 4335
+    },
+    {
+      "entropy": 0.41776869893074037,
+      "epoch": 2.44920993227991,
+      "grad_norm": 1.5501775741577148,
+      "learning_rate": 4.8630058323648584e-06,
+      "loss": 0.3111,
+      "mean_token_accuracy": 0.8984562754631042,
+      "num_tokens": 35234676.0,
+      "step": 4340
+    },
+    {
+      "entropy": 0.4286969780921936,
+      "epoch": 2.4520316027088036,
+      "grad_norm": 1.6209558248519897,
+      "learning_rate": 4.862693836226301e-06,
+      "loss": 0.3231,
+      "mean_token_accuracy": 0.8965187311172486,
+      "num_tokens": 35275453.0,
+      "step": 4345
+    },
+    {
+      "entropy": 0.40539962649345396,
+      "epoch": 2.4548532731376973,
+      "grad_norm": 1.6065021753311157,
+      "learning_rate": 4.862381498717563e-06,
+      "loss": 0.3012,
+      "mean_token_accuracy": 0.9059406757354737,
+      "num_tokens": 35315750.0,
+      "step": 4350
+    },
+    {
+      "entropy": 0.38936737179756165,
+      "epoch": 2.4576749435665914,
+      "grad_norm": 1.5308164358139038,
+      "learning_rate": 4.862068819900003e-06,
+      "loss": 0.289,
+      "mean_token_accuracy": 0.905747401714325,
+      "num_tokens": 35356475.0,
+      "step": 4355
+    },
+    {
+      "entropy": 0.3944389283657074,
+      "epoch": 2.460496613995485,
+      "grad_norm": 1.6397894620895386,
+      "learning_rate": 4.8617557998350475e-06,
+      "loss": 0.2897,
+      "mean_token_accuracy": 0.9059877634048462,
+      "num_tokens": 35397000.0,
+      "step": 4360
+    },
+    {
+      "entropy": 0.3817551672458649,
+      "epoch": 2.4633182844243793,
+      "grad_norm": 1.590949296951294,
+      "learning_rate": 4.86144243858419e-06,
+      "loss": 0.2994,
+      "mean_token_accuracy": 0.9029469013214111,
+      "num_tokens": 35437662.0,
+      "step": 4365
+    },
+    {
+      "entropy": 0.43104063868522646,
+      "epoch": 2.466139954853273,
+      "grad_norm": 1.7022565603256226,
+      "learning_rate": 4.86112873620899e-06,
+      "loss": 0.346,
+      "mean_token_accuracy": 0.8878795146942139,
+      "num_tokens": 35478304.0,
+      "step": 4370
+    },
+    {
+      "entropy": 0.4168141961097717,
+      "epoch": 2.468961625282167,
+      "grad_norm": 1.5154368877410889,
+      "learning_rate": 4.860814692771072e-06,
+      "loss": 0.3225,
+      "mean_token_accuracy": 0.8962018251419067,
+      "num_tokens": 35518948.0,
+      "step": 4375
+    },
+    {
+      "entropy": 0.4418110430240631,
+      "epoch": 2.471783295711061,
+      "grad_norm": 1.6656417846679688,
+      "learning_rate": 4.860500308332134e-06,
+      "loss": 0.3292,
+      "mean_token_accuracy": 0.8942551970481872,
+      "num_tokens": 35559540.0,
+      "step": 4380
+    },
+    {
+      "entropy": 0.40012283325195314,
+      "epoch": 2.474604966139955,
+      "grad_norm": 1.5913735628128052,
+      "learning_rate": 4.8601855829539345e-06,
+      "loss": 0.3069,
+      "mean_token_accuracy": 0.8994509220123291,
+      "num_tokens": 35600363.0,
+      "step": 4385
+    },
+    {
+      "entropy": 0.4134869039058685,
+      "epoch": 2.4774266365688487,
+      "grad_norm": 1.574020504951477,
+      "learning_rate": 4.859870516698302e-06,
+      "loss": 0.3175,
+      "mean_token_accuracy": 0.8977964639663696,
+      "num_tokens": 35641089.0,
+      "step": 4390
+    },
+    {
+      "entropy": 0.39518179297447203,
+      "epoch": 2.480248306997743,
+      "grad_norm": 1.6124846935272217,
+      "learning_rate": 4.85955510962713e-06,
+      "loss": 0.2935,
+      "mean_token_accuracy": 0.9060514092445373,
+      "num_tokens": 35681822.0,
+      "step": 4395
+    },
+    {
+      "entropy": 0.403781658411026,
+      "epoch": 2.4830699774266365,
+      "grad_norm": 1.7911304235458374,
+      "learning_rate": 4.8592393618023816e-06,
+      "loss": 0.3146,
+      "mean_token_accuracy": 0.8975111484527588,
+      "num_tokens": 35722308.0,
+      "step": 4400
+    },
+    {
+      "entropy": 0.4161704480648041,
+      "epoch": 2.4858916478555306,
+      "grad_norm": 1.7675319910049438,
+      "learning_rate": 4.858923273286086e-06,
+      "loss": 0.3117,
+      "mean_token_accuracy": 0.8990342259407044,
+      "num_tokens": 35762753.0,
+      "step": 4405
+    },
+    {
+      "entropy": 0.3830337405204773,
+      "epoch": 2.4887133182844243,
+      "grad_norm": 1.3579264879226685,
+      "learning_rate": 4.858606844140337e-06,
+      "loss": 0.2877,
+      "mean_token_accuracy": 0.9072355508804322,
+      "num_tokens": 35803606.0,
+      "step": 4410
+    },
+    {
+      "entropy": 0.4230438947677612,
+      "epoch": 2.4915349887133185,
+      "grad_norm": 1.622119665145874,
+      "learning_rate": 4.8582900744272975e-06,
+      "loss": 0.323,
+      "mean_token_accuracy": 0.8985754609107971,
+      "num_tokens": 35844338.0,
+      "step": 4415
+    },
+    {
+      "entropy": 0.42835283279418945,
+      "epoch": 2.494356659142212,
+      "grad_norm": 1.5371416807174683,
+      "learning_rate": 4.857972964209199e-06,
+      "loss": 0.3226,
+      "mean_token_accuracy": 0.8958300590515137,
+      "num_tokens": 35885113.0,
+      "step": 4420
+    },
+    {
+      "entropy": 0.42627652883529665,
+      "epoch": 2.4971783295711063,
+      "grad_norm": 1.5911802053451538,
+      "learning_rate": 4.857655513548335e-06,
+      "loss": 0.3174,
+      "mean_token_accuracy": 0.8981908798217774,
+      "num_tokens": 35925643.0,
+      "step": 4425
+    },
+    {
+      "entropy": 0.37044663429260255,
+      "epoch": 2.5,
+      "grad_norm": 1.3842355012893677,
+      "learning_rate": 4.8573377225070715e-06,
+      "loss": 0.2614,
+      "mean_token_accuracy": 0.9132282495498657,
+      "num_tokens": 35966310.0,
+      "step": 4430
+    },
+    {
+      "entropy": 0.3875482201576233,
+      "epoch": 2.5028216704288937,
+      "grad_norm": 1.421795129776001,
+      "learning_rate": 4.857019591147836e-06,
+      "loss": 0.2989,
+      "mean_token_accuracy": 0.9054312467575073,
+      "num_tokens": 36007072.0,
+      "step": 4435
+    },
+    {
+      "entropy": 0.4442818224430084,
+      "epoch": 2.505643340857788,
+      "grad_norm": 1.6131327152252197,
+      "learning_rate": 4.856701119533128e-06,
+      "loss": 0.3393,
+      "mean_token_accuracy": 0.8924556851387024,
+      "num_tokens": 36047636.0,
+      "step": 4440
+    },
+    {
+      "entropy": 0.4388956606388092,
+      "epoch": 2.508465011286682,
+      "grad_norm": 1.5778138637542725,
+      "learning_rate": 4.856382307725509e-06,
+      "loss": 0.3225,
+      "mean_token_accuracy": 0.8961979508399963,
+      "num_tokens": 36088203.0,
+      "step": 4445
+    },
+    {
+      "entropy": 0.402739155292511,
+      "epoch": 2.5112866817155757,
+      "grad_norm": 1.5481038093566895,
+      "learning_rate": 4.856063155787611e-06,
+      "loss": 0.3023,
+      "mean_token_accuracy": 0.9025402426719665,
+      "num_tokens": 36128987.0,
+      "step": 4450
+    },
+    {
+      "entropy": 0.3847965598106384,
+      "epoch": 2.5141083521444694,
+      "grad_norm": 1.4700496196746826,
+      "learning_rate": 4.855743663782131e-06,
+      "loss": 0.2893,
+      "mean_token_accuracy": 0.9061774015426636,
+      "num_tokens": 36169817.0,
+      "step": 4455
+    },
+    {
+      "entropy": 0.39482749700546266,
+      "epoch": 2.5169300225733635,
+      "grad_norm": 1.6306626796722412,
+      "learning_rate": 4.855423831771832e-06,
+      "loss": 0.2963,
+      "mean_token_accuracy": 0.9046317577362061,
+      "num_tokens": 36210322.0,
+      "step": 4460
+    },
+    {
+      "entropy": 0.44486273527145387,
+      "epoch": 2.519751693002257,
+      "grad_norm": 1.7283015251159668,
+      "learning_rate": 4.8551036598195486e-06,
+      "loss": 0.3437,
+      "mean_token_accuracy": 0.8907332181930542,
+      "num_tokens": 36251034.0,
+      "step": 4465
+    },
+    {
+      "entropy": 0.42003960609436036,
+      "epoch": 2.5225733634311513,
+      "grad_norm": 1.4865223169326782,
+      "learning_rate": 4.8547831479881745e-06,
+      "loss": 0.3239,
+      "mean_token_accuracy": 0.8974966883659363,
+      "num_tokens": 36291628.0,
+      "step": 4470
+    },
+    {
+      "entropy": 0.39935733675956725,
+      "epoch": 2.525395033860045,
+      "grad_norm": 1.3825920820236206,
+      "learning_rate": 4.854462296340677e-06,
+      "loss": 0.2872,
+      "mean_token_accuracy": 0.9067023873329163,
+      "num_tokens": 36332435.0,
+      "step": 4475
+    },
+    {
+      "entropy": 0.36378782987594604,
+      "epoch": 2.528216704288939,
+      "grad_norm": 1.507226586341858,
+      "learning_rate": 4.854141104940087e-06,
+      "loss": 0.2582,
+      "mean_token_accuracy": 0.9157714605331421,
+      "num_tokens": 36373117.0,
+      "step": 4480
+    },
+    {
+      "entropy": 0.41645694971084596,
+      "epoch": 2.531038374717833,
+      "grad_norm": 1.6471916437149048,
+      "learning_rate": 4.853819573849502e-06,
+      "loss": 0.3073,
+      "mean_token_accuracy": 0.9004851222038269,
+      "num_tokens": 36413658.0,
+      "step": 4485
+    },
+    {
+      "entropy": 0.4097161889076233,
+      "epoch": 2.533860045146727,
+      "grad_norm": 1.7997804880142212,
+      "learning_rate": 4.8534977031320855e-06,
+      "loss": 0.3056,
+      "mean_token_accuracy": 0.901634693145752,
+      "num_tokens": 36454096.0,
+      "step": 4490
+    },
+    {
+      "entropy": 0.38941258788108823,
+      "epoch": 2.5366817155756207,
+      "grad_norm": 1.6591458320617676,
+      "learning_rate": 4.8531754928510725e-06,
+      "loss": 0.3015,
+      "mean_token_accuracy": 0.9032475352287292,
+      "num_tokens": 36494625.0,
+      "step": 4495
+    },
+    {
+      "entropy": 0.39249064326286315,
+      "epoch": 2.5395033860045144,
+      "grad_norm": 1.5502938032150269,
+      "learning_rate": 4.852852943069758e-06,
+      "loss": 0.2938,
+      "mean_token_accuracy": 0.9045561671257019,
+      "num_tokens": 36535372.0,
+      "step": 4500
+    },
+    {
+      "epoch": 2.5395033860045144,
+      "eval_entropy": 0.4075223505496979,
+      "eval_loss": 0.2745997905731201,
+      "eval_mean_token_accuracy": 0.918742835521698,
+      "eval_num_tokens": 36535372.0,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 24.395,
+      "eval_steps_per_second": 6.099,
+      "step": 4500
+    },
+    {
+      "entropy": 0.42699538469314574,
+      "epoch": 2.5423250564334086,
+      "grad_norm": 1.4848240613937378,
+      "learning_rate": 4.852530053851509e-06,
+      "loss": 0.3234,
+      "mean_token_accuracy": 0.8980291485786438,
+      "num_tokens": 36576121.0,
+      "step": 4505
+    },
+    {
+      "entropy": 0.4450218975543976,
+      "epoch": 2.5451467268623027,
+      "grad_norm": 1.2995103597640991,
+      "learning_rate": 4.852206825259756e-06,
+      "loss": 0.3528,
+      "mean_token_accuracy": 0.8892289876937867,
+      "num_tokens": 36616673.0,
+      "step": 4510
+    },
+    {
+      "entropy": 0.4333424806594849,
+      "epoch": 2.5479683972911964,
+      "grad_norm": 1.4600845575332642,
+      "learning_rate": 4.851883257357997e-06,
+      "loss": 0.3108,
+      "mean_token_accuracy": 0.9001733303070069,
+      "num_tokens": 36657491.0,
+      "step": 4515
+    },
+    {
+      "entropy": 0.4113191843032837,
+      "epoch": 2.55079006772009,
+      "grad_norm": 1.7340246438980103,
+      "learning_rate": 4.851559350209798e-06,
+      "loss": 0.3183,
+      "mean_token_accuracy": 0.8975558519363404,
+      "num_tokens": 36698183.0,
+      "step": 4520
+    },
+    {
+      "entropy": 0.4206542193889618,
+      "epoch": 2.5536117381489842,
+      "grad_norm": 1.6161917448043823,
+      "learning_rate": 4.85123510387879e-06,
+      "loss": 0.326,
+      "mean_token_accuracy": 0.8958797335624695,
+      "num_tokens": 36738687.0,
+      "step": 4525
+    },
+    {
+      "entropy": 0.42369606494903567,
+      "epoch": 2.5564334085778784,
+      "grad_norm": 1.4928462505340576,
+      "learning_rate": 4.850910518428672e-06,
+      "loss": 0.3245,
+      "mean_token_accuracy": 0.8953816771507264,
+      "num_tokens": 36779100.0,
+      "step": 4530
+    },
+    {
+      "entropy": 0.4580428898334503,
+      "epoch": 2.559255079006772,
+      "grad_norm": 1.6802058219909668,
+      "learning_rate": 4.850585593923209e-06,
+      "loss": 0.3362,
+      "mean_token_accuracy": 0.8919341087341308,
+      "num_tokens": 36819464.0,
+      "step": 4535
+    },
+    {
+      "entropy": 0.44548857808113096,
+      "epoch": 2.5620767494356658,
+      "grad_norm": 1.7012122869491577,
+      "learning_rate": 4.850260330426231e-06,
+      "loss": 0.3257,
+      "mean_token_accuracy": 0.8931902885437012,
+      "num_tokens": 36860115.0,
+      "step": 4540
+    },
+    {
+      "entropy": 0.42452192306518555,
+      "epoch": 2.56489841986456,
+      "grad_norm": 1.4587101936340332,
+      "learning_rate": 4.849934728001636e-06,
+      "loss": 0.3252,
+      "mean_token_accuracy": 0.8946157932281494,
+      "num_tokens": 36900844.0,
+      "step": 4545
+    },
+    {
+      "entropy": 0.4206928193569183,
+      "epoch": 2.5677200902934536,
+      "grad_norm": 1.6573847532272339,
+      "learning_rate": 4.84960878671339e-06,
+      "loss": 0.3297,
+      "mean_token_accuracy": 0.897615122795105,
+      "num_tokens": 36941461.0,
+      "step": 4550
+    },
+    {
+      "entropy": 0.40249053835868837,
+      "epoch": 2.5705417607223477,
+      "grad_norm": 1.6816022396087646,
+      "learning_rate": 4.849282506625525e-06,
+      "loss": 0.2879,
+      "mean_token_accuracy": 0.906862735748291,
+      "num_tokens": 36982267.0,
+      "step": 4555
+    },
+    {
+      "entropy": 0.3981044590473175,
+      "epoch": 2.5733634311512414,
+      "grad_norm": 1.6164435148239136,
+      "learning_rate": 4.848955887802135e-06,
+      "loss": 0.3043,
+      "mean_token_accuracy": 0.9023780941963195,
+      "num_tokens": 37023035.0,
+      "step": 4560
+    },
+    {
+      "entropy": 0.3921403527259827,
+      "epoch": 2.5761851015801356,
+      "grad_norm": 1.7836408615112305,
+      "learning_rate": 4.8486289303073884e-06,
+      "loss": 0.2822,
+      "mean_token_accuracy": 0.9081403493881226,
+      "num_tokens": 37063570.0,
+      "step": 4565
+    },
+    {
+      "entropy": 0.41735507249832154,
+      "epoch": 2.5790067720090293,
+      "grad_norm": 1.6461783647537231,
+      "learning_rate": 4.848301634205514e-06,
+      "loss": 0.3012,
+      "mean_token_accuracy": 0.9022216796875,
+      "num_tokens": 37104278.0,
+      "step": 4570
+    },
+    {
+      "entropy": 0.3928376615047455,
+      "epoch": 2.5818284424379234,
+      "grad_norm": 1.6805448532104492,
+      "learning_rate": 4.84797399956081e-06,
+      "loss": 0.2898,
+      "mean_token_accuracy": 0.9058070182800293,
+      "num_tokens": 37144698.0,
+      "step": 4575
+    },
+    {
+      "entropy": 0.41108508706092833,
+      "epoch": 2.584650112866817,
+      "grad_norm": 1.6354347467422485,
+      "learning_rate": 4.847646026437639e-06,
+      "loss": 0.3039,
+      "mean_token_accuracy": 0.9019859433174133,
+      "num_tokens": 37185166.0,
+      "step": 4580
+    },
+    {
+      "entropy": 0.39545953273773193,
+      "epoch": 2.587471783295711,
+      "grad_norm": 1.4855034351348877,
+      "learning_rate": 4.847317714900432e-06,
+      "loss": 0.2978,
+      "mean_token_accuracy": 0.9041112303733826,
+      "num_tokens": 37225882.0,
+      "step": 4585
+    },
+    {
+      "entropy": 0.378242689371109,
+      "epoch": 2.590293453724605,
+      "grad_norm": 1.4001781940460205,
+      "learning_rate": 4.846989065013687e-06,
+      "loss": 0.2667,
+      "mean_token_accuracy": 0.9122205972671509,
+      "num_tokens": 37266182.0,
+      "step": 4590
+    },
+    {
+      "entropy": 0.41660374999046323,
+      "epoch": 2.593115124153499,
+      "grad_norm": 1.6582450866699219,
+      "learning_rate": 4.846660076841966e-06,
+      "loss": 0.3095,
+      "mean_token_accuracy": 0.9031738400459289,
+      "num_tokens": 37306948.0,
+      "step": 4595
+    },
+    {
+      "entropy": 0.4056197464466095,
+      "epoch": 2.595936794582393,
+      "grad_norm": 1.9591373205184937,
+      "learning_rate": 4.8463307504498995e-06,
+      "loss": 0.3038,
+      "mean_token_accuracy": 0.9031343698501587,
+      "num_tokens": 37347762.0,
+      "step": 4600
+    },
+    {
+      "entropy": 0.43581578731536863,
+      "epoch": 2.5987584650112865,
+      "grad_norm": 1.5790077447891235,
+      "learning_rate": 4.846001085902182e-06,
+      "loss": 0.3175,
+      "mean_token_accuracy": 0.9001483678817749,
+      "num_tokens": 37388225.0,
+      "step": 4605
+    },
+    {
+      "entropy": 0.4038006603717804,
+      "epoch": 2.6015801354401806,
+      "grad_norm": 1.5071346759796143,
+      "learning_rate": 4.845671083263579e-06,
+      "loss": 0.2923,
+      "mean_token_accuracy": 0.9046383738517761,
+      "num_tokens": 37429028.0,
+      "step": 4610
+    },
+    {
+      "entropy": 0.4395131945610046,
+      "epoch": 2.6044018058690743,
+      "grad_norm": 1.651118516921997,
+      "learning_rate": 4.845340742598917e-06,
+      "loss": 0.3422,
+      "mean_token_accuracy": 0.8916622638702393,
+      "num_tokens": 37469231.0,
+      "step": 4615
+    },
+    {
+      "entropy": 0.43975943326950073,
+      "epoch": 2.6072234762979685,
+      "grad_norm": 1.7351619005203247,
+      "learning_rate": 4.8450100639730934e-06,
+      "loss": 0.327,
+      "mean_token_accuracy": 0.8946444272994996,
+      "num_tokens": 37509748.0,
+      "step": 4620
+    },
+    {
+      "entropy": 0.3946446657180786,
+      "epoch": 2.610045146726862,
+      "grad_norm": 1.5668245553970337,
+      "learning_rate": 4.844679047451068e-06,
+      "loss": 0.2897,
+      "mean_token_accuracy": 0.9054205894470215,
+      "num_tokens": 37549968.0,
+      "step": 4625
+    },
+    {
+      "entropy": 0.4205393612384796,
+      "epoch": 2.6128668171557563,
+      "grad_norm": 1.5956112146377563,
+      "learning_rate": 4.844347693097871e-06,
+      "loss": 0.3276,
+      "mean_token_accuracy": 0.8945238828659058,
+      "num_tokens": 37590500.0,
+      "step": 4630
+    },
+    {
+      "entropy": 0.4266462683677673,
+      "epoch": 2.61568848758465,
+      "grad_norm": 1.8106273412704468,
+      "learning_rate": 4.844016000978595e-06,
+      "loss": 0.3106,
+      "mean_token_accuracy": 0.9020758390426635,
+      "num_tokens": 37631127.0,
+      "step": 4635
+    },
+    {
+      "entropy": 0.3949384868144989,
+      "epoch": 2.618510158013544,
+      "grad_norm": 1.6231563091278076,
+      "learning_rate": 4.843683971158404e-06,
+      "loss": 0.2972,
+      "mean_token_accuracy": 0.9043740272521973,
+      "num_tokens": 37671522.0,
+      "step": 4640
+    },
+    {
+      "entropy": 0.427196079492569,
+      "epoch": 2.621331828442438,
+      "grad_norm": 1.8242963552474976,
+      "learning_rate": 4.843351603702522e-06,
+      "loss": 0.3276,
+      "mean_token_accuracy": 0.8958367943763733,
+      "num_tokens": 37711505.0,
+      "step": 4645
+    },
+    {
+      "entropy": 0.4056568145751953,
+      "epoch": 2.624153498871332,
+      "grad_norm": 1.767072081565857,
+      "learning_rate": 4.843018898676245e-06,
+      "loss": 0.2968,
+      "mean_token_accuracy": 0.9040299654006958,
+      "num_tokens": 37751970.0,
+      "step": 4650
+    },
+    {
+      "entropy": 0.4182514131069183,
+      "epoch": 2.6269751693002257,
+      "grad_norm": 1.381471872329712,
+      "learning_rate": 4.842685856144932e-06,
+      "loss": 0.3202,
+      "mean_token_accuracy": 0.8963116884231568,
+      "num_tokens": 37792712.0,
+      "step": 4655
+    },
+    {
+      "entropy": 0.3773132681846619,
+      "epoch": 2.62979683972912,
+      "grad_norm": 1.4395767450332642,
+      "learning_rate": 4.842352476174008e-06,
+      "loss": 0.2735,
+      "mean_token_accuracy": 0.9093364715576172,
+      "num_tokens": 37833341.0,
+      "step": 4660
+    },
+    {
+      "entropy": 0.38550790548324587,
+      "epoch": 2.6326185101580135,
+      "grad_norm": 1.4654998779296875,
+      "learning_rate": 4.842018758828968e-06,
+      "loss": 0.2897,
+      "mean_token_accuracy": 0.906914758682251,
+      "num_tokens": 37874036.0,
+      "step": 4665
+    },
+    {
+      "entropy": 0.42129603028297424,
+      "epoch": 2.635440180586907,
+      "grad_norm": 1.8509776592254639,
+      "learning_rate": 4.84168470417537e-06,
+      "loss": 0.3046,
+      "mean_token_accuracy": 0.9016175627708435,
+      "num_tokens": 37914771.0,
+      "step": 4670
+    },
+    {
+      "entropy": 0.4129085302352905,
+      "epoch": 2.6382618510158014,
+      "grad_norm": 1.6427429914474487,
+      "learning_rate": 4.841350312278838e-06,
+      "loss": 0.3373,
+      "mean_token_accuracy": 0.893082594871521,
+      "num_tokens": 37955424.0,
+      "step": 4675
+    },
+    {
+      "entropy": 0.455252867937088,
+      "epoch": 2.6410835214446955,
+      "grad_norm": 1.7675665616989136,
+      "learning_rate": 4.8410155832050635e-06,
+      "loss": 0.346,
+      "mean_token_accuracy": 0.8894224405288697,
+      "num_tokens": 37995924.0,
+      "step": 4680
+    },
+    {
+      "entropy": 0.4129503607749939,
+      "epoch": 2.643905191873589,
+      "grad_norm": 1.8195785284042358,
+      "learning_rate": 4.840680517019806e-06,
+      "loss": 0.3097,
+      "mean_token_accuracy": 0.9006987929344177,
+      "num_tokens": 38036393.0,
+      "step": 4685
+    },
+    {
+      "entropy": 0.41016621589660646,
+      "epoch": 2.646726862302483,
+      "grad_norm": 1.335083246231079,
+      "learning_rate": 4.840345113788887e-06,
+      "loss": 0.2876,
+      "mean_token_accuracy": 0.9052215456962586,
+      "num_tokens": 38077059.0,
+      "step": 4690
+    },
+    {
+      "entropy": 0.38619791269302367,
+      "epoch": 2.649548532731377,
+      "grad_norm": 1.5167460441589355,
+      "learning_rate": 4.840009373578197e-06,
+      "loss": 0.2864,
+      "mean_token_accuracy": 0.9078164935112,
+      "num_tokens": 38117855.0,
+      "step": 4695
+    },
+    {
+      "entropy": 0.40981505513191224,
+      "epoch": 2.6523702031602707,
+      "grad_norm": 1.4652817249298096,
+      "learning_rate": 4.839673296453694e-06,
+      "loss": 0.3079,
+      "mean_token_accuracy": 0.8995686173439026,
+      "num_tokens": 38158633.0,
+      "step": 4700
+    },
+    {
+      "entropy": 0.40746703147888186,
+      "epoch": 2.655191873589165,
+      "grad_norm": 1.5045616626739502,
+      "learning_rate": 4.839336882481398e-06,
+      "loss": 0.3076,
+      "mean_token_accuracy": 0.9018496155738831,
+      "num_tokens": 38199207.0,
+      "step": 4705
+    },
+    {
+      "entropy": 0.4251773953437805,
+      "epoch": 2.6580135440180586,
+      "grad_norm": 1.7509130239486694,
+      "learning_rate": 4.839000131727399e-06,
+      "loss": 0.3138,
+      "mean_token_accuracy": 0.8988733649253845,
+      "num_tokens": 38239853.0,
+      "step": 4710
+    },
+    {
+      "entropy": 0.4144779920578003,
+      "epoch": 2.6608352144469527,
+      "grad_norm": 1.6336438655853271,
+      "learning_rate": 4.8386630442578505e-06,
+      "loss": 0.3176,
+      "mean_token_accuracy": 0.8991317749023438,
+      "num_tokens": 38280445.0,
+      "step": 4715
+    },
+    {
+      "entropy": 0.42009794116020205,
+      "epoch": 2.6636568848758464,
+      "grad_norm": 1.679404854774475,
+      "learning_rate": 4.838325620138975e-06,
+      "loss": 0.3136,
+      "mean_token_accuracy": 0.8994911909103394,
+      "num_tokens": 38321216.0,
+      "step": 4720
+    },
+    {
+      "entropy": 0.3979401350021362,
+      "epoch": 2.6664785553047405,
+      "grad_norm": 1.5892705917358398,
+      "learning_rate": 4.837987859437058e-06,
+      "loss": 0.2973,
+      "mean_token_accuracy": 0.9033595561981201,
+      "num_tokens": 38361943.0,
+      "step": 4725
+    },
+    {
+      "entropy": 0.4081549823284149,
+      "epoch": 2.6693002257336342,
+      "grad_norm": 1.4048125743865967,
+      "learning_rate": 4.837649762218454e-06,
+      "loss": 0.3061,
+      "mean_token_accuracy": 0.9016985297203064,
+      "num_tokens": 38402444.0,
+      "step": 4730
+    },
+    {
+      "entropy": 0.40113077163696287,
+      "epoch": 2.6721218961625284,
+      "grad_norm": 1.590847134590149,
+      "learning_rate": 4.837311328549582e-06,
+      "loss": 0.3019,
+      "mean_token_accuracy": 0.9024832129478455,
+      "num_tokens": 38443055.0,
+      "step": 4735
+    },
+    {
+      "entropy": 0.4067776739597321,
+      "epoch": 2.674943566591422,
+      "grad_norm": 1.6257827281951904,
+      "learning_rate": 4.8369725584969265e-06,
+      "loss": 0.2947,
+      "mean_token_accuracy": 0.9031951904296875,
+      "num_tokens": 38483816.0,
+      "step": 4740
+    },
+    {
+      "entropy": 0.386566686630249,
+      "epoch": 2.677765237020316,
+      "grad_norm": 1.3581163883209229,
+      "learning_rate": 4.836633452127039e-06,
+      "loss": 0.2919,
+      "mean_token_accuracy": 0.9064470291137695,
+      "num_tokens": 38523703.0,
+      "step": 4745
+    },
+    {
+      "entropy": 0.40097097754478456,
+      "epoch": 2.68058690744921,
+      "grad_norm": 1.6537374258041382,
+      "learning_rate": 4.836294009506537e-06,
+      "loss": 0.2976,
+      "mean_token_accuracy": 0.9049440145492553,
+      "num_tokens": 38564385.0,
+      "step": 4750
+    },
+    {
+      "entropy": 0.3615904927253723,
+      "epoch": 2.6834085778781036,
+      "grad_norm": 1.5579919815063477,
+      "learning_rate": 4.835954230702105e-06,
+      "loss": 0.2632,
+      "mean_token_accuracy": 0.9136163473129273,
+      "num_tokens": 38605052.0,
+      "step": 4755
+    },
+    {
+      "entropy": 0.43404104113578795,
+      "epoch": 2.6862302483069977,
+      "grad_norm": 1.5290402173995972,
+      "learning_rate": 4.835614115780492e-06,
+      "loss": 0.3276,
+      "mean_token_accuracy": 0.8940969586372376,
+      "num_tokens": 38645892.0,
+      "step": 4760
+    },
+    {
+      "entropy": 0.4232151687145233,
+      "epoch": 2.689051918735892,
+      "grad_norm": 1.3929665088653564,
+      "learning_rate": 4.835273664808514e-06,
+      "loss": 0.3277,
+      "mean_token_accuracy": 0.8977612614631653,
+      "num_tokens": 38686583.0,
+      "step": 4765
+    },
+    {
+      "entropy": 0.399210125207901,
+      "epoch": 2.6918735891647856,
+      "grad_norm": 1.4951647520065308,
+      "learning_rate": 4.834932877853051e-06,
+      "loss": 0.2983,
+      "mean_token_accuracy": 0.9042705297470093,
+      "num_tokens": 38726923.0,
+      "step": 4770
+    },
+    {
+      "entropy": 0.42027135491371154,
+      "epoch": 2.6946952595936793,
+      "grad_norm": 1.3856654167175293,
+      "learning_rate": 4.834591754981053e-06,
+      "loss": 0.3092,
+      "mean_token_accuracy": 0.9001957535743713,
+      "num_tokens": 38767439.0,
+      "step": 4775
+    },
+    {
+      "entropy": 0.4031683325767517,
+      "epoch": 2.6975169300225734,
+      "grad_norm": 1.5139154195785522,
+      "learning_rate": 4.834250296259532e-06,
+      "loss": 0.2899,
+      "mean_token_accuracy": 0.9039840221405029,
+      "num_tokens": 38808039.0,
+      "step": 4780
+    },
+    {
+      "entropy": 0.40245692133903505,
+      "epoch": 2.700338600451467,
+      "grad_norm": 1.7345075607299805,
+      "learning_rate": 4.8339085017555685e-06,
+      "loss": 0.307,
+      "mean_token_accuracy": 0.9012099623680114,
+      "num_tokens": 38848491.0,
+      "step": 4785
+    },
+    {
+      "entropy": 0.4227768838405609,
+      "epoch": 2.7031602708803613,
+      "grad_norm": 1.7092951536178589,
+      "learning_rate": 4.833566371536307e-06,
+      "loss": 0.3181,
+      "mean_token_accuracy": 0.8983111381530762,
+      "num_tokens": 38889280.0,
+      "step": 4790
+    },
+    {
+      "entropy": 0.4330504536628723,
+      "epoch": 2.705981941309255,
+      "grad_norm": 1.6095880270004272,
+      "learning_rate": 4.83322390566896e-06,
+      "loss": 0.3443,
+      "mean_token_accuracy": 0.8882631182670593,
+      "num_tokens": 38929955.0,
+      "step": 4795
+    },
+    {
+      "entropy": 0.40083847045898435,
+      "epoch": 2.708803611738149,
+      "grad_norm": 1.877541184425354,
+      "learning_rate": 4.832881104220805e-06,
+      "loss": 0.2985,
+      "mean_token_accuracy": 0.9026248455047607,
+      "num_tokens": 38970610.0,
+      "step": 4800
+    },
+    {
+      "entropy": 0.43792678117752076,
+      "epoch": 2.711625282167043,
+      "grad_norm": 1.6264429092407227,
+      "learning_rate": 4.8325379672591845e-06,
+      "loss": 0.331,
+      "mean_token_accuracy": 0.8906769514083862,
+      "num_tokens": 39011465.0,
+      "step": 4805
+    },
+    {
+      "entropy": 0.40115252137184143,
+      "epoch": 2.714446952595937,
+      "grad_norm": 1.7116504907608032,
+      "learning_rate": 4.8321944948515085e-06,
+      "loss": 0.272,
+      "mean_token_accuracy": 0.9107529878616333,
+      "num_tokens": 39052019.0,
+      "step": 4810
+    },
+    {
+      "entropy": 0.3959204196929932,
+      "epoch": 2.7172686230248306,
+      "grad_norm": 1.5745071172714233,
+      "learning_rate": 4.831850687065253e-06,
+      "loss": 0.2908,
+      "mean_token_accuracy": 0.9042417287826539,
+      "num_tokens": 39092820.0,
+      "step": 4815
+    },
+    {
+      "entropy": 0.4313171863555908,
+      "epoch": 2.7200902934537243,
+      "grad_norm": 1.4608080387115479,
+      "learning_rate": 4.831506543967958e-06,
+      "loss": 0.3263,
+      "mean_token_accuracy": 0.8949733734130859,
+      "num_tokens": 39133601.0,
+      "step": 4820
+    },
+    {
+      "entropy": 0.4190512001514435,
+      "epoch": 2.7229119638826185,
+      "grad_norm": 1.6403937339782715,
+      "learning_rate": 4.831162065627229e-06,
+      "loss": 0.3379,
+      "mean_token_accuracy": 0.8949956893920898,
+      "num_tokens": 39174455.0,
+      "step": 4825
+    },
+    {
+      "entropy": 0.39245215654373167,
+      "epoch": 2.7257336343115126,
+      "grad_norm": 1.5303268432617188,
+      "learning_rate": 4.830817252110742e-06,
+      "loss": 0.2972,
+      "mean_token_accuracy": 0.9037534713745117,
+      "num_tokens": 39215087.0,
+      "step": 4830
+    },
+    {
+      "entropy": 0.40173652172088625,
+      "epoch": 2.7285553047404063,
+      "grad_norm": 1.5214855670928955,
+      "learning_rate": 4.830472103486233e-06,
+      "loss": 0.2968,
+      "mean_token_accuracy": 0.902808690071106,
+      "num_tokens": 39255856.0,
+      "step": 4835
+    },
+    {
+      "entropy": 0.4071968972682953,
+      "epoch": 2.7313769751693,
+      "grad_norm": 1.669796109199524,
+      "learning_rate": 4.830126619821508e-06,
+      "loss": 0.2997,
+      "mean_token_accuracy": 0.903080677986145,
+      "num_tokens": 39296108.0,
+      "step": 4840
+    },
+    {
+      "entropy": 0.4280484437942505,
+      "epoch": 2.734198645598194,
+      "grad_norm": 1.5314013957977295,
+      "learning_rate": 4.829780801184437e-06,
+      "loss": 0.322,
+      "mean_token_accuracy": 0.8953822731971741,
+      "num_tokens": 39336454.0,
+      "step": 4845
+    },
+    {
+      "entropy": 0.38799464106559756,
+      "epoch": 2.7370203160270883,
+      "grad_norm": 1.6457887887954712,
+      "learning_rate": 4.829434647642956e-06,
+      "loss": 0.3007,
+      "mean_token_accuracy": 0.902879273891449,
+      "num_tokens": 39377316.0,
+      "step": 4850
+    },
+    {
+      "entropy": 0.43819814920425415,
+      "epoch": 2.739841986455982,
+      "grad_norm": 1.7149285078048706,
+      "learning_rate": 4.829088159265067e-06,
+      "loss": 0.3406,
+      "mean_token_accuracy": 0.8905494093894959,
+      "num_tokens": 39418143.0,
+      "step": 4855
+    },
+    {
+      "entropy": 0.41923085451126096,
+      "epoch": 2.7426636568848757,
+      "grad_norm": 1.331289529800415,
+      "learning_rate": 4.828741336118837e-06,
+      "loss": 0.3192,
+      "mean_token_accuracy": 0.8980522155761719,
+      "num_tokens": 39457182.0,
+      "step": 4860
+    },
+    {
+      "entropy": 0.40077372193336486,
+      "epoch": 2.74548532731377,
+      "grad_norm": 1.6222182512283325,
+      "learning_rate": 4.828394178272401e-06,
+      "loss": 0.302,
+      "mean_token_accuracy": 0.9054031133651733,
+      "num_tokens": 39497879.0,
+      "step": 4865
+    },
+    {
+      "entropy": 0.39027564525604247,
+      "epoch": 2.7483069977426635,
+      "grad_norm": 1.7052563428878784,
+      "learning_rate": 4.828046685793957e-06,
+      "loss": 0.2949,
+      "mean_token_accuracy": 0.9051683902740478,
+      "num_tokens": 39538388.0,
+      "step": 4870
+    },
+    {
+      "entropy": 0.42039836645126344,
+      "epoch": 2.7511286681715577,
+      "grad_norm": 1.6202837228775024,
+      "learning_rate": 4.82769885875177e-06,
+      "loss": 0.3195,
+      "mean_token_accuracy": 0.8974061369895935,
+      "num_tokens": 39579248.0,
+      "step": 4875
+    },
+    {
+      "entropy": 0.42434735894203185,
+      "epoch": 2.7539503386004514,
+      "grad_norm": 1.6943541765213013,
+      "learning_rate": 4.8273506972141705e-06,
+      "loss": 0.3145,
+      "mean_token_accuracy": 0.8977732062339783,
+      "num_tokens": 39619680.0,
+      "step": 4880
+    },
+    {
+      "entropy": 0.4002332389354706,
+      "epoch": 2.7567720090293455,
+      "grad_norm": 1.5710278749465942,
+      "learning_rate": 4.827002201249556e-06,
+      "loss": 0.298,
+      "mean_token_accuracy": 0.9056402921676636,
+      "num_tokens": 39660296.0,
+      "step": 4885
+    },
+    {
+      "entropy": 0.39988999366760253,
+      "epoch": 2.759593679458239,
+      "grad_norm": 1.5473220348358154,
+      "learning_rate": 4.826653370926387e-06,
+      "loss": 0.2995,
+      "mean_token_accuracy": 0.9022600173950195,
+      "num_tokens": 39701090.0,
+      "step": 4890
+    },
+    {
+      "entropy": 0.41356533765792847,
+      "epoch": 2.7624153498871333,
+      "grad_norm": 1.5488544702529907,
+      "learning_rate": 4.826304206313193e-06,
+      "loss": 0.3091,
+      "mean_token_accuracy": 0.8972513794898986,
+      "num_tokens": 39741947.0,
+      "step": 4895
+    },
+    {
+      "entropy": 0.39104182124137876,
+      "epoch": 2.765237020316027,
+      "grad_norm": 1.5953575372695923,
+      "learning_rate": 4.825954707478565e-06,
+      "loss": 0.3018,
+      "mean_token_accuracy": 0.9020759582519531,
+      "num_tokens": 39782510.0,
+      "step": 4900
+    },
+    {
+      "entropy": 0.41210404634475706,
+      "epoch": 2.7680586907449207,
+      "grad_norm": 1.4687649011611938,
+      "learning_rate": 4.825604874491165e-06,
+      "loss": 0.3201,
+      "mean_token_accuracy": 0.8970984816551208,
+      "num_tokens": 39823334.0,
+      "step": 4905
+    },
+    {
+      "entropy": 0.38967686891555786,
+      "epoch": 2.770880361173815,
+      "grad_norm": 1.563970923423767,
+      "learning_rate": 4.825254707419716e-06,
+      "loss": 0.2888,
+      "mean_token_accuracy": 0.9057512521743775,
+      "num_tokens": 39864124.0,
+      "step": 4910
+    },
+    {
+      "entropy": 0.4446266949176788,
+      "epoch": 2.773702031602709,
+      "grad_norm": 1.8769588470458984,
+      "learning_rate": 4.824904206333009e-06,
+      "loss": 0.3441,
+      "mean_token_accuracy": 0.8903054833412171,
+      "num_tokens": 39904728.0,
+      "step": 4915
+    },
+    {
+      "entropy": 0.3861452877521515,
+      "epoch": 2.7765237020316027,
+      "grad_norm": 1.4496490955352783,
+      "learning_rate": 4.8245533712998995e-06,
+      "loss": 0.283,
+      "mean_token_accuracy": 0.9068530559539795,
+      "num_tokens": 39945454.0,
+      "step": 4920
+    },
+    {
+      "entropy": 0.407192063331604,
+      "epoch": 2.7793453724604964,
+      "grad_norm": 1.4505201578140259,
+      "learning_rate": 4.8242022023893095e-06,
+      "loss": 0.3053,
+      "mean_token_accuracy": 0.9021325826644897,
+      "num_tokens": 39986255.0,
+      "step": 4925
+    },
+    {
+      "entropy": 0.3858388364315033,
+      "epoch": 2.7821670428893905,
+      "grad_norm": 1.4844613075256348,
+      "learning_rate": 4.823850699670225e-06,
+      "loss": 0.2856,
+      "mean_token_accuracy": 0.9087419271469116,
+      "num_tokens": 40026935.0,
+      "step": 4930
+    },
+    {
+      "entropy": 0.3972056210041046,
+      "epoch": 2.7849887133182847,
+      "grad_norm": 1.8050625324249268,
+      "learning_rate": 4.823498863211701e-06,
+      "loss": 0.3016,
+      "mean_token_accuracy": 0.9033455014228821,
+      "num_tokens": 40067747.0,
+      "step": 4935
+    },
+    {
+      "entropy": 0.4366573691368103,
+      "epoch": 2.7878103837471784,
+      "grad_norm": 1.7209434509277344,
+      "learning_rate": 4.823146693082853e-06,
+      "loss": 0.331,
+      "mean_token_accuracy": 0.8931609511375427,
+      "num_tokens": 40107827.0,
+      "step": 4940
+    },
+    {
+      "entropy": 0.44152148365974425,
+      "epoch": 2.790632054176072,
+      "grad_norm": 1.6366186141967773,
+      "learning_rate": 4.822794189352867e-06,
+      "loss": 0.3618,
+      "mean_token_accuracy": 0.8887471437454224,
+      "num_tokens": 40148719.0,
+      "step": 4945
+    },
+    {
+      "entropy": 0.3947724997997284,
+      "epoch": 2.793453724604966,
+      "grad_norm": 1.3238067626953125,
+      "learning_rate": 4.822441352090992e-06,
+      "loss": 0.2977,
+      "mean_token_accuracy": 0.9030474781990051,
+      "num_tokens": 40189420.0,
+      "step": 4950
+    },
+    {
+      "entropy": 0.42072648406028745,
+      "epoch": 2.79627539503386,
+      "grad_norm": 1.553117036819458,
+      "learning_rate": 4.8220881813665435e-06,
+      "loss": 0.3086,
+      "mean_token_accuracy": 0.9006119012832642,
+      "num_tokens": 40230144.0,
+      "step": 4955
+    },
+    {
+      "entropy": 0.39687870144844056,
+      "epoch": 2.799097065462754,
+      "grad_norm": 1.5955865383148193,
+      "learning_rate": 4.8217346772489e-06,
+      "loss": 0.2926,
+      "mean_token_accuracy": 0.9029557347297669,
+      "num_tokens": 40270958.0,
+      "step": 4960
+    },
+    {
+      "entropy": 0.38045825362205504,
+      "epoch": 2.8019187358916477,
+      "grad_norm": 1.9875127077102661,
+      "learning_rate": 4.821380839807509e-06,
+      "loss": 0.2799,
+      "mean_token_accuracy": 0.9087697386741638,
+      "num_tokens": 40311657.0,
+      "step": 4965
+    },
+    {
+      "entropy": 0.42601910829544065,
+      "epoch": 2.804740406320542,
+      "grad_norm": 1.7846463918685913,
+      "learning_rate": 4.821026669111881e-06,
+      "loss": 0.3235,
+      "mean_token_accuracy": 0.8960994839668274,
+      "num_tokens": 40352256.0,
+      "step": 4970
+    },
+    {
+      "entropy": 0.40002630949020385,
+      "epoch": 2.8075620767494356,
+      "grad_norm": 1.5628581047058105,
+      "learning_rate": 4.820672165231595e-06,
+      "loss": 0.3004,
+      "mean_token_accuracy": 0.9042887330055237,
+      "num_tokens": 40392932.0,
+      "step": 4975
+    },
+    {
+      "entropy": 0.40157084465026854,
+      "epoch": 2.8103837471783297,
+      "grad_norm": 1.515154480934143,
+      "learning_rate": 4.8203173282362904e-06,
+      "loss": 0.3059,
+      "mean_token_accuracy": 0.900831151008606,
+      "num_tokens": 40433461.0,
+      "step": 4980
+    },
+    {
+      "entropy": 0.4204134941101074,
+      "epoch": 2.8132054176072234,
+      "grad_norm": 1.5249524116516113,
+      "learning_rate": 4.819962158195677e-06,
+      "loss": 0.3223,
+      "mean_token_accuracy": 0.8980574369430542,
+      "num_tokens": 40473869.0,
+      "step": 4985
+    },
+    {
+      "entropy": 0.3809087574481964,
+      "epoch": 2.816027088036117,
+      "grad_norm": 1.5786020755767822,
+      "learning_rate": 4.819606655179527e-06,
+      "loss": 0.2748,
+      "mean_token_accuracy": 0.9111071109771729,
+      "num_tokens": 40514570.0,
+      "step": 4990
+    },
+    {
+      "entropy": 0.4299299597740173,
+      "epoch": 2.8188487584650113,
+      "grad_norm": 1.6505275964736938,
+      "learning_rate": 4.819250819257679e-06,
+      "loss": 0.3196,
+      "mean_token_accuracy": 0.8989997148513794,
+      "num_tokens": 40555393.0,
+      "step": 4995
+    },
+    {
+      "entropy": 0.41674472093582154,
+      "epoch": 2.8216704288939054,
+      "grad_norm": 1.8646577596664429,
+      "learning_rate": 4.818894650500037e-06,
+      "loss": 0.3096,
+      "mean_token_accuracy": 0.9004587650299072,
+      "num_tokens": 40596103.0,
+      "step": 5000
+    },
+    {
+      "epoch": 2.8216704288939054,
+      "eval_entropy": 0.40231844782829285,
+      "eval_loss": 0.2420196682214737,
+      "eval_mean_token_accuracy": 0.9271751642227173,
+      "eval_num_tokens": 40596103.0,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 24.368,
+      "eval_steps_per_second": 6.092,
+      "step": 5000
+    },
+    {
+      "entropy": 0.4362106263637543,
+      "epoch": 2.824492099322799,
+      "grad_norm": 1.5646847486495972,
+      "learning_rate": 4.818538148976572e-06,
+      "loss": 0.3389,
+      "mean_token_accuracy": 0.8909627199172974,
+      "num_tokens": 40636724.0,
+      "step": 5005
+    },
+    {
+      "entropy": 0.41136063933372496,
+      "epoch": 2.827313769751693,
+      "grad_norm": 1.6143547296524048,
+      "learning_rate": 4.8181813147573166e-06,
+      "loss": 0.3059,
+      "mean_token_accuracy": 0.9023926854133606,
+      "num_tokens": 40677571.0,
+      "step": 5010
+    },
+    {
+      "entropy": 0.4300425052642822,
+      "epoch": 2.830135440180587,
+      "grad_norm": 1.538517713546753,
+      "learning_rate": 4.817824147912371e-06,
+      "loss": 0.3059,
+      "mean_token_accuracy": 0.901062273979187,
+      "num_tokens": 40717960.0,
+      "step": 5015
+    },
+    {
+      "entropy": 0.4204569935798645,
+      "epoch": 2.832957110609481,
+      "grad_norm": 1.51469886302948,
+      "learning_rate": 4.817466648511903e-06,
+      "loss": 0.317,
+      "mean_token_accuracy": 0.9010179877281189,
+      "num_tokens": 40758728.0,
+      "step": 5020
+    },
+    {
+      "entropy": 0.4019504964351654,
+      "epoch": 2.8357787810383748,
+      "grad_norm": 1.5685406923294067,
+      "learning_rate": 4.817108816626142e-06,
+      "loss": 0.3027,
+      "mean_token_accuracy": 0.9020584464073181,
+      "num_tokens": 40799222.0,
+      "step": 5025
+    },
+    {
+      "entropy": 0.4178150534629822,
+      "epoch": 2.8386004514672685,
+      "grad_norm": 1.5866230726242065,
+      "learning_rate": 4.816750652325382e-06,
+      "loss": 0.3083,
+      "mean_token_accuracy": 0.9016893863677978,
+      "num_tokens": 40839920.0,
+      "step": 5030
+    },
+    {
+      "entropy": 0.3792074918746948,
+      "epoch": 2.8414221218961626,
+      "grad_norm": 1.2347012758255005,
+      "learning_rate": 4.8163921556799885e-06,
+      "loss": 0.2941,
+      "mean_token_accuracy": 0.904798150062561,
+      "num_tokens": 40880474.0,
+      "step": 5035
+    },
+    {
+      "entropy": 0.42022097706794737,
+      "epoch": 2.8442437923250563,
+      "grad_norm": 2.0809385776519775,
+      "learning_rate": 4.816033326760384e-06,
+      "loss": 0.3125,
+      "mean_token_accuracy": 0.9000694155693054,
+      "num_tokens": 40920909.0,
+      "step": 5040
+    },
+    {
+      "entropy": 0.40716384053230287,
+      "epoch": 2.8470654627539504,
+      "grad_norm": 1.6952626705169678,
+      "learning_rate": 4.815674165637065e-06,
+      "loss": 0.3045,
+      "mean_token_accuracy": 0.9000877380371094,
+      "num_tokens": 40961474.0,
+      "step": 5045
+    },
+    {
+      "entropy": 0.4542030215263367,
+      "epoch": 2.849887133182844,
+      "grad_norm": 2.060497283935547,
+      "learning_rate": 4.815314672380586e-06,
+      "loss": 0.3413,
+      "mean_token_accuracy": 0.8906081914901733,
+      "num_tokens": 41002244.0,
+      "step": 5050
+    },
+    {
+      "entropy": 0.42580119967460633,
+      "epoch": 2.8527088036117383,
+      "grad_norm": 1.8272522687911987,
+      "learning_rate": 4.814954847061568e-06,
+      "loss": 0.3134,
+      "mean_token_accuracy": 0.8984528660774231,
+      "num_tokens": 41043006.0,
+      "step": 5055
+    },
+    {
+      "entropy": 0.39229719042778016,
+      "epoch": 2.855530474040632,
+      "grad_norm": 1.7011030912399292,
+      "learning_rate": 4.8145946897507026e-06,
+      "loss": 0.2907,
+      "mean_token_accuracy": 0.9052790999412537,
+      "num_tokens": 41083650.0,
+      "step": 5060
+    },
+    {
+      "entropy": 0.42324894666671753,
+      "epoch": 2.858352144469526,
+      "grad_norm": 1.7294552326202393,
+      "learning_rate": 4.814234200518741e-06,
+      "loss": 0.3193,
+      "mean_token_accuracy": 0.8964131712913513,
+      "num_tokens": 41124231.0,
+      "step": 5065
+    },
+    {
+      "entropy": 0.4176328182220459,
+      "epoch": 2.86117381489842,
+      "grad_norm": 1.8063850402832031,
+      "learning_rate": 4.813873379436499e-06,
+      "loss": 0.311,
+      "mean_token_accuracy": 0.8978959679603576,
+      "num_tokens": 41164837.0,
+      "step": 5070
+    },
+    {
+      "entropy": 0.4265771806240082,
+      "epoch": 2.8639954853273135,
+      "grad_norm": 1.4727050065994263,
+      "learning_rate": 4.813512226574863e-06,
+      "loss": 0.3381,
+      "mean_token_accuracy": 0.894121527671814,
+      "num_tokens": 41205500.0,
+      "step": 5075
+    },
+    {
+      "entropy": 0.4046603262424469,
+      "epoch": 2.8668171557562077,
+      "grad_norm": 1.5069383382797241,
+      "learning_rate": 4.813150742004782e-06,
+      "loss": 0.3129,
+      "mean_token_accuracy": 0.8987318396568298,
+      "num_tokens": 41246220.0,
+      "step": 5080
+    },
+    {
+      "entropy": 0.41487335562705996,
+      "epoch": 2.869638826185102,
+      "grad_norm": 1.5012837648391724,
+      "learning_rate": 4.812788925797267e-06,
+      "loss": 0.3078,
+      "mean_token_accuracy": 0.9006859183311462,
+      "num_tokens": 41287146.0,
+      "step": 5085
+    },
+    {
+      "entropy": 0.40536822080612184,
+      "epoch": 2.8724604966139955,
+      "grad_norm": 1.679571270942688,
+      "learning_rate": 4.812426778023398e-06,
+      "loss": 0.3058,
+      "mean_token_accuracy": 0.9021418213844299,
+      "num_tokens": 41327663.0,
+      "step": 5090
+    },
+    {
+      "entropy": 0.4111123144626617,
+      "epoch": 2.875282167042889,
+      "grad_norm": 1.7820204496383667,
+      "learning_rate": 4.812064298754319e-06,
+      "loss": 0.3202,
+      "mean_token_accuracy": 0.8966802835464478,
+      "num_tokens": 41367964.0,
+      "step": 5095
+    },
+    {
+      "entropy": 0.4519184172153473,
+      "epoch": 2.8781038374717833,
+      "grad_norm": 1.9354314804077148,
+      "learning_rate": 4.811701488061239e-06,
+      "loss": 0.3484,
+      "mean_token_accuracy": 0.8897565484046936,
+      "num_tokens": 41408264.0,
+      "step": 5100
+    },
+    {
+      "entropy": 0.3959288239479065,
+      "epoch": 2.8809255079006775,
+      "grad_norm": 1.4760626554489136,
+      "learning_rate": 4.811338346015434e-06,
+      "loss": 0.2947,
+      "mean_token_accuracy": 0.9035079836845398,
+      "num_tokens": 41448952.0,
+      "step": 5105
+    },
+    {
+      "entropy": 0.3818633139133453,
+      "epoch": 2.883747178329571,
+      "grad_norm": 1.3426601886749268,
+      "learning_rate": 4.81097487268824e-06,
+      "loss": 0.2856,
+      "mean_token_accuracy": 0.9082074284553527,
+      "num_tokens": 41489735.0,
+      "step": 5110
+    },
+    {
+      "entropy": 0.4096335232257843,
+      "epoch": 2.886568848758465,
+      "grad_norm": 1.6107584238052368,
+      "learning_rate": 4.810611068151064e-06,
+      "loss": 0.319,
+      "mean_token_accuracy": 0.898867416381836,
+      "num_tokens": 41530551.0,
+      "step": 5115
+    },
+    {
+      "entropy": 0.4616283357143402,
+      "epoch": 2.889390519187359,
+      "grad_norm": 1.9297223091125488,
+      "learning_rate": 4.810246932475374e-06,
+      "loss": 0.3473,
+      "mean_token_accuracy": 0.8871626138687134,
+      "num_tokens": 41571097.0,
+      "step": 5120
+    },
+    {
+      "entropy": 0.40603058934211733,
+      "epoch": 2.8922121896162527,
+      "grad_norm": 1.4726239442825317,
+      "learning_rate": 4.809882465732706e-06,
+      "loss": 0.3127,
+      "mean_token_accuracy": 0.8997641682624817,
+      "num_tokens": 41611601.0,
+      "step": 5125
+    },
+    {
+      "entropy": 0.4259068608283997,
+      "epoch": 2.895033860045147,
+      "grad_norm": 1.739429235458374,
+      "learning_rate": 4.809517667994657e-06,
+      "loss": 0.3129,
+      "mean_token_accuracy": 0.8975944757461548,
+      "num_tokens": 41651971.0,
+      "step": 5130
+    },
+    {
+      "entropy": 0.42120612859725953,
+      "epoch": 2.8978555304740405,
+      "grad_norm": 1.9967470169067383,
+      "learning_rate": 4.809152539332895e-06,
+      "loss": 0.3139,
+      "mean_token_accuracy": 0.9000747799873352,
+      "num_tokens": 41692648.0,
+      "step": 5135
+    },
+    {
+      "entropy": 0.4155091643333435,
+      "epoch": 2.9006772009029347,
+      "grad_norm": 1.7973077297210693,
+      "learning_rate": 4.808787079819147e-06,
+      "loss": 0.3242,
+      "mean_token_accuracy": 0.8949877023696899,
+      "num_tokens": 41733143.0,
+      "step": 5140
+    },
+    {
+      "entropy": 0.41151371598243713,
+      "epoch": 2.9034988713318284,
+      "grad_norm": 1.404732346534729,
+      "learning_rate": 4.808421289525208e-06,
+      "loss": 0.3154,
+      "mean_token_accuracy": 0.8984189987182617,
+      "num_tokens": 41773361.0,
+      "step": 5145
+    },
+    {
+      "entropy": 0.42156214118003843,
+      "epoch": 2.9063205417607225,
+      "grad_norm": 1.6763156652450562,
+      "learning_rate": 4.808055168522938e-06,
+      "loss": 0.3303,
+      "mean_token_accuracy": 0.8945571660995484,
+      "num_tokens": 41813917.0,
+      "step": 5150
+    },
+    {
+      "entropy": 0.4589288830757141,
+      "epoch": 2.909142212189616,
+      "grad_norm": 1.4989210367202759,
+      "learning_rate": 4.807688716884262e-06,
+      "loss": 0.359,
+      "mean_token_accuracy": 0.8893625497817993,
+      "num_tokens": 41854234.0,
+      "step": 5155
+    },
+    {
+      "entropy": 0.45793223977088926,
+      "epoch": 2.91196388261851,
+      "grad_norm": 1.8785529136657715,
+      "learning_rate": 4.807321934681168e-06,
+      "loss": 0.344,
+      "mean_token_accuracy": 0.8890782237052918,
+      "num_tokens": 41895042.0,
+      "step": 5160
+    },
+    {
+      "entropy": 0.42541446089744567,
+      "epoch": 2.914785553047404,
+      "grad_norm": 1.4923590421676636,
+      "learning_rate": 4.806954821985711e-06,
+      "loss": 0.3257,
+      "mean_token_accuracy": 0.8974828958511353,
+      "num_tokens": 41935666.0,
+      "step": 5165
+    },
+    {
+      "entropy": 0.411184823513031,
+      "epoch": 2.917607223476298,
+      "grad_norm": 1.8144638538360596,
+      "learning_rate": 4.806587378870011e-06,
+      "loss": 0.3096,
+      "mean_token_accuracy": 0.9010491013526917,
+      "num_tokens": 41976246.0,
+      "step": 5170
+    },
+    {
+      "entropy": 0.4226935863494873,
+      "epoch": 2.920428893905192,
+      "grad_norm": 1.6548048257827759,
+      "learning_rate": 4.806219605406253e-06,
+      "loss": 0.3246,
+      "mean_token_accuracy": 0.8946655511856079,
+      "num_tokens": 42016979.0,
+      "step": 5175
+    },
+    {
+      "entropy": 0.4352991938591003,
+      "epoch": 2.9232505643340856,
+      "grad_norm": 1.8966737985610962,
+      "learning_rate": 4.805851501666683e-06,
+      "loss": 0.3265,
+      "mean_token_accuracy": 0.8924818277359009,
+      "num_tokens": 42057536.0,
+      "step": 5180
+    },
+    {
+      "entropy": 0.43102755546569826,
+      "epoch": 2.9260722347629797,
+      "grad_norm": 1.7986955642700195,
+      "learning_rate": 4.805483067723618e-06,
+      "loss": 0.3331,
+      "mean_token_accuracy": 0.8937864661216736,
+      "num_tokens": 42097826.0,
+      "step": 5185
+    },
+    {
+      "entropy": 0.4175034463405609,
+      "epoch": 2.9288939051918734,
+      "grad_norm": 1.5185545682907104,
+      "learning_rate": 4.805114303649436e-06,
+      "loss": 0.3217,
+      "mean_token_accuracy": 0.8953788757324219,
+      "num_tokens": 42138400.0,
+      "step": 5190
+    },
+    {
+      "entropy": 0.4255950450897217,
+      "epoch": 2.9317155756207676,
+      "grad_norm": 1.8746482133865356,
+      "learning_rate": 4.80474520951658e-06,
+      "loss": 0.3113,
+      "mean_token_accuracy": 0.9003988265991211,
+      "num_tokens": 42178987.0,
+      "step": 5195
+    },
+    {
+      "entropy": 0.41336881518363955,
+      "epoch": 2.9345372460496613,
+      "grad_norm": 1.6806355714797974,
+      "learning_rate": 4.80437578539756e-06,
+      "loss": 0.3031,
+      "mean_token_accuracy": 0.9040932536125184,
+      "num_tokens": 42219657.0,
+      "step": 5200
+    },
+    {
+      "entropy": 0.4057203710079193,
+      "epoch": 2.9373589164785554,
+      "grad_norm": 1.564095377922058,
+      "learning_rate": 4.804006031364948e-06,
+      "loss": 0.2936,
+      "mean_token_accuracy": 0.9054961562156677,
+      "num_tokens": 42260473.0,
+      "step": 5205
+    },
+    {
+      "entropy": 0.39404407143592834,
+      "epoch": 2.940180586907449,
+      "grad_norm": 1.4518405199050903,
+      "learning_rate": 4.8036359474913826e-06,
+      "loss": 0.288,
+      "mean_token_accuracy": 0.9059139609336853,
+      "num_tokens": 42301055.0,
+      "step": 5210
+    },
+    {
+      "entropy": 0.3924104571342468,
+      "epoch": 2.9430022573363432,
+      "grad_norm": 1.319478154182434,
+      "learning_rate": 4.803265533849569e-06,
+      "loss": 0.3051,
+      "mean_token_accuracy": 0.9025569319725036,
+      "num_tokens": 42341730.0,
+      "step": 5215
+    },
+    {
+      "entropy": 0.3995777368545532,
+      "epoch": 2.945823927765237,
+      "grad_norm": 1.4966541528701782,
+      "learning_rate": 4.802894790512271e-06,
+      "loss": 0.3079,
+      "mean_token_accuracy": 0.901541805267334,
+      "num_tokens": 42382070.0,
+      "step": 5220
+    },
+    {
+      "entropy": 0.4068488717079163,
+      "epoch": 2.948645598194131,
+      "grad_norm": 1.8578685522079468,
+      "learning_rate": 4.8025237175523245e-06,
+      "loss": 0.2956,
+      "mean_token_accuracy": 0.9044872045516967,
+      "num_tokens": 42422271.0,
+      "step": 5225
+    },
+    {
+      "entropy": 0.40595067739486695,
+      "epoch": 2.9514672686230248,
+      "grad_norm": 1.6572576761245728,
+      "learning_rate": 4.8021523150426255e-06,
+      "loss": 0.3079,
+      "mean_token_accuracy": 0.9003460884094239,
+      "num_tokens": 42463045.0,
+      "step": 5230
+    },
+    {
+      "entropy": 0.4293433666229248,
+      "epoch": 2.954288939051919,
+      "grad_norm": 1.6312439441680908,
+      "learning_rate": 4.801780583056135e-06,
+      "loss": 0.3073,
+      "mean_token_accuracy": 0.9005551815032959,
+      "num_tokens": 42503743.0,
+      "step": 5235
+    },
+    {
+      "entropy": 0.43949187994003297,
+      "epoch": 2.9571106094808126,
+      "grad_norm": 1.5055447816848755,
+      "learning_rate": 4.8014085216658824e-06,
+      "loss": 0.3352,
+      "mean_token_accuracy": 0.8934733510017395,
+      "num_tokens": 42544394.0,
+      "step": 5240
+    },
+    {
+      "entropy": 0.3744525730609894,
+      "epoch": 2.9599322799097063,
+      "grad_norm": 1.7301849126815796,
+      "learning_rate": 4.801036130944957e-06,
+      "loss": 0.29,
+      "mean_token_accuracy": 0.9071328043937683,
+      "num_tokens": 42584858.0,
+      "step": 5245
+    },
+    {
+      "entropy": 0.41108238101005556,
+      "epoch": 2.9627539503386005,
+      "grad_norm": 1.419072151184082,
+      "learning_rate": 4.800663410966516e-06,
+      "loss": 0.3142,
+      "mean_token_accuracy": 0.8972790360450744,
+      "num_tokens": 42625252.0,
+      "step": 5250
+    },
+    {
+      "entropy": 0.4198480904102325,
+      "epoch": 2.9655756207674946,
+      "grad_norm": 1.5576399564743042,
+      "learning_rate": 4.80029036180378e-06,
+      "loss": 0.312,
+      "mean_token_accuracy": 0.9000542402267456,
+      "num_tokens": 42665780.0,
+      "step": 5255
+    },
+    {
+      "entropy": 0.4146255135536194,
+      "epoch": 2.9683972911963883,
+      "grad_norm": 1.5485845804214478,
+      "learning_rate": 4.799916983530035e-06,
+      "loss": 0.3035,
+      "mean_token_accuracy": 0.9038738012313843,
+      "num_tokens": 42706347.0,
+      "step": 5260
+    },
+    {
+      "entropy": 0.4407613933086395,
+      "epoch": 2.971218961625282,
+      "grad_norm": 1.6536972522735596,
+      "learning_rate": 4.7995432762186305e-06,
+      "loss": 0.3554,
+      "mean_token_accuracy": 0.8906416058540344,
+      "num_tokens": 42747026.0,
+      "step": 5265
+    },
+    {
+      "entropy": 0.4324296176433563,
+      "epoch": 2.974040632054176,
+      "grad_norm": 1.9314972162246704,
+      "learning_rate": 4.799169239942982e-06,
+      "loss": 0.323,
+      "mean_token_accuracy": 0.8953546643257141,
+      "num_tokens": 42787529.0,
+      "step": 5270
+    },
+    {
+      "entropy": 0.40229780673980714,
+      "epoch": 2.97686230248307,
+      "grad_norm": 1.437831163406372,
+      "learning_rate": 4.798794874776569e-06,
+      "loss": 0.301,
+      "mean_token_accuracy": 0.9024040460586548,
+      "num_tokens": 42828263.0,
+      "step": 5275
+    },
+    {
+      "entropy": 0.4193461239337921,
+      "epoch": 2.979683972911964,
+      "grad_norm": 1.6553016901016235,
+      "learning_rate": 4.798420180792934e-06,
+      "loss": 0.3335,
+      "mean_token_accuracy": 0.8927172064781189,
+      "num_tokens": 42868861.0,
+      "step": 5280
+    },
+    {
+      "entropy": 0.39387176036834715,
+      "epoch": 2.9825056433408577,
+      "grad_norm": 1.827269434928894,
+      "learning_rate": 4.7980451580656884e-06,
+      "loss": 0.3012,
+      "mean_token_accuracy": 0.9013678908348084,
+      "num_tokens": 42909719.0,
+      "step": 5285
+    },
+    {
+      "entropy": 0.4310999274253845,
+      "epoch": 2.985327313769752,
+      "grad_norm": 1.4325158596038818,
+      "learning_rate": 4.797669806668504e-06,
+      "loss": 0.3228,
+      "mean_token_accuracy": 0.8959625005722046,
+      "num_tokens": 42950331.0,
+      "step": 5290
+    },
+    {
+      "entropy": 0.3933727502822876,
+      "epoch": 2.9881489841986455,
+      "grad_norm": 1.5803815126419067,
+      "learning_rate": 4.797294126675117e-06,
+      "loss": 0.2916,
+      "mean_token_accuracy": 0.9042976379394532,
+      "num_tokens": 42991038.0,
+      "step": 5295
+    },
+    {
+      "entropy": 0.4161331236362457,
+      "epoch": 2.9909706546275396,
+      "grad_norm": 1.6079249382019043,
+      "learning_rate": 4.796918118159333e-06,
+      "loss": 0.3214,
+      "mean_token_accuracy": 0.8964751839637757,
+      "num_tokens": 43031524.0,
+      "step": 5300
+    },
+    {
+      "entropy": 0.41101468205451963,
+      "epoch": 2.9937923250564333,
+      "grad_norm": 1.6859304904937744,
+      "learning_rate": 4.796541781195018e-06,
+      "loss": 0.3066,
+      "mean_token_accuracy": 0.8995181202888489,
+      "num_tokens": 43072136.0,
+      "step": 5305
+    },
+    {
+      "entropy": 0.415380471944809,
+      "epoch": 2.9966139954853275,
+      "grad_norm": 1.5194308757781982,
+      "learning_rate": 4.796165115856101e-06,
+      "loss": 0.3134,
+      "mean_token_accuracy": 0.8985631585121154,
+      "num_tokens": 43111676.0,
+      "step": 5310
+    },
+    {
+      "entropy": 0.4073172271251678,
+      "epoch": 2.999435665914221,
+      "grad_norm": 1.6618201732635498,
+      "learning_rate": 4.79578812221658e-06,
+      "loss": 0.3191,
+      "mean_token_accuracy": 0.8969994068145752,
+      "num_tokens": 43152259.0,
+      "step": 5315
+    },
+    {
+      "entropy": 0.40902658700942995,
+      "epoch": 3.0022573363431153,
+      "grad_norm": 1.5331395864486694,
+      "learning_rate": 4.795410800350516e-06,
+      "loss": 0.2449,
+      "mean_token_accuracy": 0.9244343876838684,
+      "num_tokens": 43186514.0,
+      "step": 5320
+    },
+    {
+      "entropy": 0.3297205209732056,
+      "epoch": 3.005079006772009,
+      "grad_norm": 1.183812141418457,
+      "learning_rate": 4.795033150332033e-06,
+      "loss": 0.2,
+      "mean_token_accuracy": 0.93786141872406,
+      "num_tokens": 43227120.0,
+      "step": 5325
+    },
+    {
+      "entropy": 0.32953916788101195,
+      "epoch": 3.007900677200903,
+      "grad_norm": 1.991351842880249,
+      "learning_rate": 4.79465517223532e-06,
+      "loss": 0.2247,
+      "mean_token_accuracy": 0.9283014059066772,
+      "num_tokens": 43267656.0,
+      "step": 5330
+    },
+    {
+      "entropy": 0.3248750567436218,
+      "epoch": 3.010722347629797,
+      "grad_norm": 1.9206222295761108,
+      "learning_rate": 4.794276866134631e-06,
+      "loss": 0.2204,
+      "mean_token_accuracy": 0.9299774169921875,
+      "num_tokens": 43308512.0,
+      "step": 5335
+    },
+    {
+      "entropy": 0.29567849040031435,
+      "epoch": 3.0135440180586905,
+      "grad_norm": 1.8526611328125,
+      "learning_rate": 4.793898232104286e-06,
+      "loss": 0.2006,
+      "mean_token_accuracy": 0.9349032402038574,
+      "num_tokens": 43349318.0,
+      "step": 5340
+    },
+    {
+      "entropy": 0.29463449120521545,
+      "epoch": 3.0163656884875847,
+      "grad_norm": 1.8326590061187744,
+      "learning_rate": 4.7935192702186655e-06,
+      "loss": 0.1999,
+      "mean_token_accuracy": 0.936297345161438,
+      "num_tokens": 43389542.0,
+      "step": 5345
+    },
+    {
+      "entropy": 0.322478848695755,
+      "epoch": 3.0191873589164784,
+      "grad_norm": 1.6944438219070435,
+      "learning_rate": 4.793139980552218e-06,
+      "loss": 0.2056,
+      "mean_token_accuracy": 0.9352605819702149,
+      "num_tokens": 43430125.0,
+      "step": 5350
+    },
+    {
+      "entropy": 0.28133447766304015,
+      "epoch": 3.0220090293453725,
+      "grad_norm": 5.7940263748168945,
+      "learning_rate": 4.792760363179454e-06,
+      "loss": 0.1745,
+      "mean_token_accuracy": 0.9443145155906677,
+      "num_tokens": 43470808.0,
+      "step": 5355
+    },
+    {
+      "entropy": 0.31858010292053224,
+      "epoch": 3.024830699774266,
+      "grad_norm": 2.1480491161346436,
+      "learning_rate": 4.79238041817495e-06,
+      "loss": 0.1987,
+      "mean_token_accuracy": 0.936722457408905,
+      "num_tokens": 43511437.0,
+      "step": 5360
+    },
+    {
+      "entropy": 0.27969213724136355,
+      "epoch": 3.0276523702031604,
+      "grad_norm": 1.8360697031021118,
+      "learning_rate": 4.792000145613346e-06,
+      "loss": 0.1816,
+      "mean_token_accuracy": 0.9421907782554626,
+      "num_tokens": 43552179.0,
+      "step": 5365
+    },
+    {
+      "entropy": 0.3075882375240326,
+      "epoch": 3.030474040632054,
+      "grad_norm": 1.9553619623184204,
+      "learning_rate": 4.791619545569347e-06,
+      "loss": 0.2032,
+      "mean_token_accuracy": 0.9348996162414551,
+      "num_tokens": 43592847.0,
+      "step": 5370
+    },
+    {
+      "entropy": 0.3161883890628815,
+      "epoch": 3.033295711060948,
+      "grad_norm": 1.7996515035629272,
+      "learning_rate": 4.7912386181177216e-06,
+      "loss": 0.2017,
+      "mean_token_accuracy": 0.9368625760078431,
+      "num_tokens": 43633494.0,
+      "step": 5375
+    },
+    {
+      "entropy": 0.31576813459396363,
+      "epoch": 3.036117381489842,
+      "grad_norm": 1.7700996398925781,
+      "learning_rate": 4.790857363333303e-06,
+      "loss": 0.2138,
+      "mean_token_accuracy": 0.9302696347236633,
+      "num_tokens": 43674104.0,
+      "step": 5380
+    },
+    {
+      "entropy": 0.2922771334648132,
+      "epoch": 3.038939051918736,
+      "grad_norm": 2.0567281246185303,
+      "learning_rate": 4.790475781290988e-06,
+      "loss": 0.1944,
+      "mean_token_accuracy": 0.9376009941101074,
+      "num_tokens": 43714718.0,
+      "step": 5385
+    },
+    {
+      "entropy": 0.3182399868965149,
+      "epoch": 3.0417607223476297,
+      "grad_norm": 1.8016152381896973,
+      "learning_rate": 4.79009387206574e-06,
+      "loss": 0.2071,
+      "mean_token_accuracy": 0.9343981266021728,
+      "num_tokens": 43755355.0,
+      "step": 5390
+    },
+    {
+      "entropy": 0.27976890206336974,
+      "epoch": 3.044582392776524,
+      "grad_norm": 1.6723804473876953,
+      "learning_rate": 4.7897116357325844e-06,
+      "loss": 0.174,
+      "mean_token_accuracy": 0.9448309302330017,
+      "num_tokens": 43795791.0,
+      "step": 5395
+    },
+    {
+      "entropy": 0.30181885361671446,
+      "epoch": 3.0474040632054176,
+      "grad_norm": 1.6413596868515015,
+      "learning_rate": 4.7893290723666116e-06,
+      "loss": 0.192,
+      "mean_token_accuracy": 0.938207459449768,
+      "num_tokens": 43836503.0,
+      "step": 5400
+    },
+    {
+      "entropy": 0.2982407629489899,
+      "epoch": 3.0502257336343117,
+      "grad_norm": 1.594096064567566,
+      "learning_rate": 4.788946182042976e-06,
+      "loss": 0.1967,
+      "mean_token_accuracy": 0.9364789009094239,
+      "num_tokens": 43877133.0,
+      "step": 5405
+    },
+    {
+      "entropy": 0.34292480945587156,
+      "epoch": 3.0530474040632054,
+      "grad_norm": 1.880804181098938,
+      "learning_rate": 4.788562964836897e-06,
+      "loss": 0.2291,
+      "mean_token_accuracy": 0.9263935923576355,
+      "num_tokens": 43917640.0,
+      "step": 5410
+    },
+    {
+      "entropy": 0.32324235439300536,
+      "epoch": 3.055869074492099,
+      "grad_norm": 1.8121103048324585,
+      "learning_rate": 4.788179420823657e-06,
+      "loss": 0.2245,
+      "mean_token_accuracy": 0.9280650019645691,
+      "num_tokens": 43958238.0,
+      "step": 5415
+    },
+    {
+      "entropy": 0.31875272989273074,
+      "epoch": 3.0586907449209932,
+      "grad_norm": 1.8845757246017456,
+      "learning_rate": 4.787795550078603e-06,
+      "loss": 0.2127,
+      "mean_token_accuracy": 0.9321471333503724,
+      "num_tokens": 43999127.0,
+      "step": 5420
+    },
+    {
+      "entropy": 0.306513249874115,
+      "epoch": 3.061512415349887,
+      "grad_norm": 1.7889739274978638,
+      "learning_rate": 4.787411352677148e-06,
+      "loss": 0.2052,
+      "mean_token_accuracy": 0.9344831585884095,
+      "num_tokens": 44039844.0,
+      "step": 5425
+    },
+    {
+      "entropy": 0.3005764603614807,
+      "epoch": 3.064334085778781,
+      "grad_norm": 1.6967425346374512,
+      "learning_rate": 4.787026828694767e-06,
+      "loss": 0.1913,
+      "mean_token_accuracy": 0.9385549426078796,
+      "num_tokens": 44080390.0,
+      "step": 5430
+    },
+    {
+      "entropy": 0.28662807643413546,
+      "epoch": 3.0671557562076748,
+      "grad_norm": 1.8293821811676025,
+      "learning_rate": 4.786641978206999e-06,
+      "loss": 0.1713,
+      "mean_token_accuracy": 0.9455971002578736,
+      "num_tokens": 44120988.0,
+      "step": 5435
+    },
+    {
+      "entropy": 0.3011536240577698,
+      "epoch": 3.069977426636569,
+      "grad_norm": 1.7703379392623901,
+      "learning_rate": 4.786256801289449e-06,
+      "loss": 0.1839,
+      "mean_token_accuracy": 0.9414896726608276,
+      "num_tokens": 44161183.0,
+      "step": 5440
+    },
+    {
+      "entropy": 0.31452121734619143,
+      "epoch": 3.0727990970654626,
+      "grad_norm": 1.900452733039856,
+      "learning_rate": 4.785871298017783e-06,
+      "loss": 0.1999,
+      "mean_token_accuracy": 0.9365707755088806,
+      "num_tokens": 44201881.0,
+      "step": 5445
+    },
+    {
+      "entropy": 0.296010160446167,
+      "epoch": 3.0756207674943568,
+      "grad_norm": 2.206284761428833,
+      "learning_rate": 4.7854854684677345e-06,
+      "loss": 0.1944,
+      "mean_token_accuracy": 0.9396218180656433,
+      "num_tokens": 44242707.0,
+      "step": 5450
+    },
+    {
+      "entropy": 0.3006631314754486,
+      "epoch": 3.0784424379232505,
+      "grad_norm": 1.720755934715271,
+      "learning_rate": 4.785099312715101e-06,
+      "loss": 0.1956,
+      "mean_token_accuracy": 0.9376758575439453,
+      "num_tokens": 44283160.0,
+      "step": 5455
+    },
+    {
+      "entropy": 0.28210367560386657,
+      "epoch": 3.0812641083521446,
+      "grad_norm": 1.6647253036499023,
+      "learning_rate": 4.7847128308357414e-06,
+      "loss": 0.1767,
+      "mean_token_accuracy": 0.943335497379303,
+      "num_tokens": 44323781.0,
+      "step": 5460
+    },
+    {
+      "entropy": 0.32272166609764097,
+      "epoch": 3.0840857787810383,
+      "grad_norm": 2.004971742630005,
+      "learning_rate": 4.7843260229055805e-06,
+      "loss": 0.209,
+      "mean_token_accuracy": 0.9329189896583557,
+      "num_tokens": 44364364.0,
+      "step": 5465
+    },
+    {
+      "entropy": 0.3206708192825317,
+      "epoch": 3.0869074492099324,
+      "grad_norm": 2.0899529457092285,
+      "learning_rate": 4.7839388890006065e-06,
+      "loss": 0.2202,
+      "mean_token_accuracy": 0.9304483771324158,
+      "num_tokens": 44404981.0,
+      "step": 5470
+    },
+    {
+      "entropy": 0.33442134261131284,
+      "epoch": 3.089729119638826,
+      "grad_norm": 2.028860092163086,
+      "learning_rate": 4.783551429196872e-06,
+      "loss": 0.2262,
+      "mean_token_accuracy": 0.9287503361701965,
+      "num_tokens": 44445757.0,
+      "step": 5475
+    },
+    {
+      "entropy": 0.31830272674560545,
+      "epoch": 3.0925507900677203,
+      "grad_norm": 1.8910348415374756,
+      "learning_rate": 4.783163643570493e-06,
+      "loss": 0.2219,
+      "mean_token_accuracy": 0.9306220650672913,
+      "num_tokens": 44486469.0,
+      "step": 5480
+    },
+    {
+      "entropy": 0.2903516858816147,
+      "epoch": 3.095372460496614,
+      "grad_norm": 1.8281724452972412,
+      "learning_rate": 4.782775532197652e-06,
+      "loss": 0.2034,
+      "mean_token_accuracy": 0.9371914029121399,
+      "num_tokens": 44527281.0,
+      "step": 5485
+    },
+    {
+      "entropy": 0.2952991366386414,
+      "epoch": 3.098194130925508,
+      "grad_norm": 1.6057690382003784,
+      "learning_rate": 4.7823870951545924e-06,
+      "loss": 0.2059,
+      "mean_token_accuracy": 0.9358728647232055,
+      "num_tokens": 44567483.0,
+      "step": 5490
+    },
+    {
+      "entropy": 0.3091245710849762,
+      "epoch": 3.101015801354402,
+      "grad_norm": 1.645340919494629,
+      "learning_rate": 4.781998332517621e-06,
+      "loss": 0.2082,
+      "mean_token_accuracy": 0.9315654993057251,
+      "num_tokens": 44608171.0,
+      "step": 5495
+    },
+    {
+      "entropy": 0.28429390788078307,
+      "epoch": 3.1038374717832955,
+      "grad_norm": 1.748238205909729,
+      "learning_rate": 4.781609244363113e-06,
+      "loss": 0.1902,
+      "mean_token_accuracy": 0.9393861174583436,
+      "num_tokens": 44648818.0,
+      "step": 5500
+    },
+    {
+      "epoch": 3.1038374717832955,
+      "eval_entropy": 0.31511667370796204,
+      "eval_loss": 0.1905011534690857,
+      "eval_mean_token_accuracy": 0.9417401552200317,
+      "eval_num_tokens": 44648818.0,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 24.418,
+      "eval_steps_per_second": 6.104,
+      "step": 5500
+    },
+    {
+      "entropy": 0.291202050447464,
+      "epoch": 3.1066591422121896,
+      "grad_norm": 1.549842357635498,
+      "learning_rate": 4.781219830767503e-06,
+      "loss": 0.1874,
+      "mean_token_accuracy": 0.9398093223571777,
+      "num_tokens": 44689512.0,
+      "step": 5505
+    },
+    {
+      "entropy": 0.32962539196014407,
+      "epoch": 3.1094808126410833,
+      "grad_norm": 1.8189475536346436,
+      "learning_rate": 4.780830091807293e-06,
+      "loss": 0.2264,
+      "mean_token_accuracy": 0.9287977695465088,
+      "num_tokens": 44729392.0,
+      "step": 5510
+    },
+    {
+      "entropy": 0.3322260320186615,
+      "epoch": 3.1123024830699775,
+      "grad_norm": 1.7237313985824585,
+      "learning_rate": 4.780440027559045e-06,
+      "loss": 0.2275,
+      "mean_token_accuracy": 0.9315350294113159,
+      "num_tokens": 44770052.0,
+      "step": 5515
+    },
+    {
+      "entropy": 0.3075240433216095,
+      "epoch": 3.115124153498871,
+      "grad_norm": 1.6741464138031006,
+      "learning_rate": 4.780049638099389e-06,
+      "loss": 0.2057,
+      "mean_token_accuracy": 0.9344268202781677,
+      "num_tokens": 44810590.0,
+      "step": 5520
+    },
+    {
+      "entropy": 0.3267399907112122,
+      "epoch": 3.1179458239277653,
+      "grad_norm": 1.500756859779358,
+      "learning_rate": 4.779658923505016e-06,
+      "loss": 0.2024,
+      "mean_token_accuracy": 0.9370131492614746,
+      "num_tokens": 44851209.0,
+      "step": 5525
+    },
+    {
+      "entropy": 0.30828378200531004,
+      "epoch": 3.120767494356659,
+      "grad_norm": 1.5910139083862305,
+      "learning_rate": 4.779267883852683e-06,
+      "loss": 0.1998,
+      "mean_token_accuracy": 0.9370311021804809,
+      "num_tokens": 44891316.0,
+      "step": 5530
+    },
+    {
+      "entropy": 0.29669389128685,
+      "epoch": 3.123589164785553,
+      "grad_norm": 1.7703447341918945,
+      "learning_rate": 4.778876519219208e-06,
+      "loss": 0.1821,
+      "mean_token_accuracy": 0.9407304763793946,
+      "num_tokens": 44932141.0,
+      "step": 5535
+    },
+    {
+      "entropy": 0.29509271383285524,
+      "epoch": 3.126410835214447,
+      "grad_norm": 1.945721983909607,
+      "learning_rate": 4.778484829681477e-06,
+      "loss": 0.1991,
+      "mean_token_accuracy": 0.9373913645744324,
+      "num_tokens": 44972742.0,
+      "step": 5540
+    },
+    {
+      "entropy": 0.2876304090023041,
+      "epoch": 3.129232505643341,
+      "grad_norm": 1.8770701885223389,
+      "learning_rate": 4.778092815316436e-06,
+      "loss": 0.1877,
+      "mean_token_accuracy": 0.9409420013427734,
+      "num_tokens": 45013573.0,
+      "step": 5545
+    },
+    {
+      "entropy": 0.33961347937583924,
+      "epoch": 3.1320541760722347,
+      "grad_norm": 1.8638192415237427,
+      "learning_rate": 4.777700476201096e-06,
+      "loss": 0.2317,
+      "mean_token_accuracy": 0.9269558072090149,
+      "num_tokens": 45054408.0,
+      "step": 5550
+    },
+    {
+      "entropy": 0.3194903790950775,
+      "epoch": 3.134875846501129,
+      "grad_norm": 1.9824572801589966,
+      "learning_rate": 4.777307812412533e-06,
+      "loss": 0.2087,
+      "mean_token_accuracy": 0.9329106450080872,
+      "num_tokens": 45094995.0,
+      "step": 5555
+    },
+    {
+      "entropy": 0.3165232837200165,
+      "epoch": 3.1376975169300225,
+      "grad_norm": 1.8498613834381104,
+      "learning_rate": 4.776914824027885e-06,
+      "loss": 0.1959,
+      "mean_token_accuracy": 0.9351617097854614,
+      "num_tokens": 45135825.0,
+      "step": 5560
+    },
+    {
+      "entropy": 0.31093028783798216,
+      "epoch": 3.1405191873589167,
+      "grad_norm": 1.8327690362930298,
+      "learning_rate": 4.776521511124356e-06,
+      "loss": 0.2014,
+      "mean_token_accuracy": 0.9345197558403016,
+      "num_tokens": 45176378.0,
+      "step": 5565
+    },
+    {
+      "entropy": 0.2836467266082764,
+      "epoch": 3.1433408577878104,
+      "grad_norm": 1.7564665079116821,
+      "learning_rate": 4.7761278737792115e-06,
+      "loss": 0.1741,
+      "mean_token_accuracy": 0.9437781810760498,
+      "num_tokens": 45216787.0,
+      "step": 5570
+    },
+    {
+      "entropy": 0.2903373181819916,
+      "epoch": 3.1461625282167045,
+      "grad_norm": 1.730517029762268,
+      "learning_rate": 4.775733912069781e-06,
+      "loss": 0.1928,
+      "mean_token_accuracy": 0.9383692383766175,
+      "num_tokens": 45257435.0,
+      "step": 5575
+    },
+    {
+      "entropy": 0.27841404676437376,
+      "epoch": 3.148984198645598,
+      "grad_norm": 1.6484125852584839,
+      "learning_rate": 4.775339626073458e-06,
+      "loss": 0.1796,
+      "mean_token_accuracy": 0.9405380487442017,
+      "num_tokens": 45298209.0,
+      "step": 5580
+    },
+    {
+      "entropy": 0.30049493312835696,
+      "epoch": 3.151805869074492,
+      "grad_norm": 1.5576494932174683,
+      "learning_rate": 4.774945015867702e-06,
+      "loss": 0.1956,
+      "mean_token_accuracy": 0.9364784240722657,
+      "num_tokens": 45338825.0,
+      "step": 5585
+    },
+    {
+      "entropy": 0.31532043814659116,
+      "epoch": 3.154627539503386,
+      "grad_norm": 1.8203058242797852,
+      "learning_rate": 4.774550081530034e-06,
+      "loss": 0.1975,
+      "mean_token_accuracy": 0.9349407196044922,
+      "num_tokens": 45379636.0,
+      "step": 5590
+    },
+    {
+      "entropy": 0.29182214140892027,
+      "epoch": 3.1574492099322797,
+      "grad_norm": 1.9475021362304688,
+      "learning_rate": 4.774154823138037e-06,
+      "loss": 0.1936,
+      "mean_token_accuracy": 0.9374213933944702,
+      "num_tokens": 45420353.0,
+      "step": 5595
+    },
+    {
+      "entropy": 0.32046533823013307,
+      "epoch": 3.160270880361174,
+      "grad_norm": 1.7757936716079712,
+      "learning_rate": 4.773759240769361e-06,
+      "loss": 0.2002,
+      "mean_token_accuracy": 0.9345276355743408,
+      "num_tokens": 45460828.0,
+      "step": 5600
+    },
+    {
+      "entropy": 0.29744908809661863,
+      "epoch": 3.1630925507900676,
+      "grad_norm": 1.8185851573944092,
+      "learning_rate": 4.773363334501717e-06,
+      "loss": 0.1847,
+      "mean_token_accuracy": 0.939760959148407,
+      "num_tokens": 45501378.0,
+      "step": 5605
+    },
+    {
+      "entropy": 0.2820535182952881,
+      "epoch": 3.1659142212189617,
+      "grad_norm": 1.843806266784668,
+      "learning_rate": 4.772967104412882e-06,
+      "loss": 0.1958,
+      "mean_token_accuracy": 0.9386505484580994,
+      "num_tokens": 45542183.0,
+      "step": 5610
+    },
+    {
+      "entropy": 0.2983327269554138,
+      "epoch": 3.1687358916478554,
+      "grad_norm": 1.7788043022155762,
+      "learning_rate": 4.772570550580696e-06,
+      "loss": 0.1908,
+      "mean_token_accuracy": 0.939252245426178,
+      "num_tokens": 45582722.0,
+      "step": 5615
+    },
+    {
+      "entropy": 0.3242061614990234,
+      "epoch": 3.1715575620767495,
+      "grad_norm": 1.9367573261260986,
+      "learning_rate": 4.77217367308306e-06,
+      "loss": 0.2153,
+      "mean_token_accuracy": 0.9323940753936768,
+      "num_tokens": 45623151.0,
+      "step": 5620
+    },
+    {
+      "entropy": 0.32598050832748415,
+      "epoch": 3.1743792325056432,
+      "grad_norm": 1.8925623893737793,
+      "learning_rate": 4.7717764719979425e-06,
+      "loss": 0.2206,
+      "mean_token_accuracy": 0.9283890128135681,
+      "num_tokens": 45663702.0,
+      "step": 5625
+    },
+    {
+      "entropy": 0.296782386302948,
+      "epoch": 3.1772009029345374,
+      "grad_norm": 2.1091501712799072,
+      "learning_rate": 4.771378947403374e-06,
+      "loss": 0.2048,
+      "mean_token_accuracy": 0.9344670414924622,
+      "num_tokens": 45704366.0,
+      "step": 5630
+    },
+    {
+      "entropy": 0.3019926369190216,
+      "epoch": 3.180022573363431,
+      "grad_norm": 1.747178077697754,
+      "learning_rate": 4.770981099377445e-06,
+      "loss": 0.2092,
+      "mean_token_accuracy": 0.9341341137886048,
+      "num_tokens": 45744753.0,
+      "step": 5635
+    },
+    {
+      "entropy": 0.31195615530014037,
+      "epoch": 3.1828442437923252,
+      "grad_norm": 2.096071243286133,
+      "learning_rate": 4.7705829279983165e-06,
+      "loss": 0.2089,
+      "mean_token_accuracy": 0.9316715359687805,
+      "num_tokens": 45784713.0,
+      "step": 5640
+    },
+    {
+      "entropy": 0.3224526882171631,
+      "epoch": 3.185665914221219,
+      "grad_norm": 2.126277446746826,
+      "learning_rate": 4.770184433344207e-06,
+      "loss": 0.2126,
+      "mean_token_accuracy": 0.9330295085906982,
+      "num_tokens": 45825418.0,
+      "step": 5645
+    },
+    {
+      "entropy": 0.33095919489860537,
+      "epoch": 3.188487584650113,
+      "grad_norm": 1.8362377882003784,
+      "learning_rate": 4.769785615493403e-06,
+      "loss": 0.2234,
+      "mean_token_accuracy": 0.9276701092720032,
+      "num_tokens": 45866068.0,
+      "step": 5650
+    },
+    {
+      "entropy": 0.290876042842865,
+      "epoch": 3.1913092550790068,
+      "grad_norm": 1.6160290241241455,
+      "learning_rate": 4.76938647452425e-06,
+      "loss": 0.1988,
+      "mean_token_accuracy": 0.9370180249214173,
+      "num_tokens": 45906929.0,
+      "step": 5655
+    },
+    {
+      "entropy": 0.2966496706008911,
+      "epoch": 3.194130925507901,
+      "grad_norm": 1.9973173141479492,
+      "learning_rate": 4.76898701051516e-06,
+      "loss": 0.2004,
+      "mean_token_accuracy": 0.9346588611602783,
+      "num_tokens": 45947800.0,
+      "step": 5660
+    },
+    {
+      "entropy": 0.3274839758872986,
+      "epoch": 3.1969525959367946,
+      "grad_norm": 1.7757952213287354,
+      "learning_rate": 4.768587223544609e-06,
+      "loss": 0.2114,
+      "mean_token_accuracy": 0.9339724779129028,
+      "num_tokens": 45987994.0,
+      "step": 5665
+    },
+    {
+      "entropy": 0.31054744124412537,
+      "epoch": 3.1997742663656883,
+      "grad_norm": 1.8302743434906006,
+      "learning_rate": 4.768187113691134e-06,
+      "loss": 0.1925,
+      "mean_token_accuracy": 0.937913966178894,
+      "num_tokens": 46028630.0,
+      "step": 5670
+    },
+    {
+      "entropy": 0.296200168132782,
+      "epoch": 3.2025959367945824,
+      "grad_norm": 1.6775931119918823,
+      "learning_rate": 4.767786681033337e-06,
+      "loss": 0.1923,
+      "mean_token_accuracy": 0.9360583305358887,
+      "num_tokens": 46069224.0,
+      "step": 5675
+    },
+    {
+      "entropy": 0.31996251940727233,
+      "epoch": 3.205417607223476,
+      "grad_norm": 2.0050246715545654,
+      "learning_rate": 4.767385925649883e-06,
+      "loss": 0.2124,
+      "mean_token_accuracy": 0.9327707052230835,
+      "num_tokens": 46109724.0,
+      "step": 5680
+    },
+    {
+      "entropy": 0.2851140141487122,
+      "epoch": 3.2082392776523703,
+      "grad_norm": 2.052438497543335,
+      "learning_rate": 4.7669848476195005e-06,
+      "loss": 0.1954,
+      "mean_token_accuracy": 0.9375227808952331,
+      "num_tokens": 46150400.0,
+      "step": 5685
+    },
+    {
+      "entropy": 0.2947817206382751,
+      "epoch": 3.211060948081264,
+      "grad_norm": 1.9339162111282349,
+      "learning_rate": 4.766583447020981e-06,
+      "loss": 0.2079,
+      "mean_token_accuracy": 0.9334628462791443,
+      "num_tokens": 46191151.0,
+      "step": 5690
+    },
+    {
+      "entropy": 0.3035987734794617,
+      "epoch": 3.213882618510158,
+      "grad_norm": 1.9785780906677246,
+      "learning_rate": 4.76618172393318e-06,
+      "loss": 0.2027,
+      "mean_token_accuracy": 0.9366864800453186,
+      "num_tokens": 46231643.0,
+      "step": 5695
+    },
+    {
+      "entropy": 0.31024947464466096,
+      "epoch": 3.216704288939052,
+      "grad_norm": 1.4870526790618896,
+      "learning_rate": 4.765779678435016e-06,
+      "loss": 0.2126,
+      "mean_token_accuracy": 0.9308468461036682,
+      "num_tokens": 46272392.0,
+      "step": 5700
+    },
+    {
+      "entropy": 0.31926954388618467,
+      "epoch": 3.219525959367946,
+      "grad_norm": 1.933563470840454,
+      "learning_rate": 4.76537731060547e-06,
+      "loss": 0.219,
+      "mean_token_accuracy": 0.9304630041122437,
+      "num_tokens": 46313068.0,
+      "step": 5705
+    },
+    {
+      "entropy": 0.3053418666124344,
+      "epoch": 3.2223476297968396,
+      "grad_norm": 1.7980693578720093,
+      "learning_rate": 4.764974620523589e-06,
+      "loss": 0.1975,
+      "mean_token_accuracy": 0.9364609718322754,
+      "num_tokens": 46353615.0,
+      "step": 5710
+    },
+    {
+      "entropy": 0.29312550127506254,
+      "epoch": 3.225169300225734,
+      "grad_norm": 1.8937259912490845,
+      "learning_rate": 4.764571608268481e-06,
+      "loss": 0.1914,
+      "mean_token_accuracy": 0.9379542708396912,
+      "num_tokens": 46394082.0,
+      "step": 5715
+    },
+    {
+      "entropy": 0.3227978229522705,
+      "epoch": 3.2279909706546275,
+      "grad_norm": 1.9451820850372314,
+      "learning_rate": 4.764168273919317e-06,
+      "loss": 0.2264,
+      "mean_token_accuracy": 0.9285582900047302,
+      "num_tokens": 46434664.0,
+      "step": 5720
+    },
+    {
+      "entropy": 0.2956021040678024,
+      "epoch": 3.2308126410835216,
+      "grad_norm": 2.080122947692871,
+      "learning_rate": 4.7637646175553325e-06,
+      "loss": 0.2137,
+      "mean_token_accuracy": 0.9323472619056702,
+      "num_tokens": 46475192.0,
+      "step": 5725
+    },
+    {
+      "entropy": 0.32532267570495604,
+      "epoch": 3.2336343115124153,
+      "grad_norm": 1.788669466972351,
+      "learning_rate": 4.763360639255826e-06,
+      "loss": 0.2257,
+      "mean_token_accuracy": 0.9298007488250732,
+      "num_tokens": 46515676.0,
+      "step": 5730
+    },
+    {
+      "entropy": 0.3365131139755249,
+      "epoch": 3.2364559819413095,
+      "grad_norm": 1.9396251440048218,
+      "learning_rate": 4.762956339100158e-06,
+      "loss": 0.2328,
+      "mean_token_accuracy": 0.9281740784645081,
+      "num_tokens": 46556056.0,
+      "step": 5735
+    },
+    {
+      "entropy": 0.2854636192321777,
+      "epoch": 3.239277652370203,
+      "grad_norm": 1.900076150894165,
+      "learning_rate": 4.762551717167756e-06,
+      "loss": 0.1883,
+      "mean_token_accuracy": 0.9398433446884156,
+      "num_tokens": 46596732.0,
+      "step": 5740
+    },
+    {
+      "entropy": 0.3118133544921875,
+      "epoch": 3.2420993227990973,
+      "grad_norm": 2.0240774154663086,
+      "learning_rate": 4.762146773538105e-06,
+      "loss": 0.211,
+      "mean_token_accuracy": 0.9326440811157226,
+      "num_tokens": 46637432.0,
+      "step": 5745
+    },
+    {
+      "entropy": 0.293948894739151,
+      "epoch": 3.244920993227991,
+      "grad_norm": 1.7449288368225098,
+      "learning_rate": 4.7617415082907575e-06,
+      "loss": 0.198,
+      "mean_token_accuracy": 0.9358683586120605,
+      "num_tokens": 46678134.0,
+      "step": 5750
+    },
+    {
+      "entropy": 0.32562840580940244,
+      "epoch": 3.2477426636568847,
+      "grad_norm": 1.7750238180160522,
+      "learning_rate": 4.761335921505329e-06,
+      "loss": 0.2237,
+      "mean_token_accuracy": 0.929267966747284,
+      "num_tokens": 46718630.0,
+      "step": 5755
+    },
+    {
+      "entropy": 0.2856466591358185,
+      "epoch": 3.250564334085779,
+      "grad_norm": 1.5713127851486206,
+      "learning_rate": 4.760930013261495e-06,
+      "loss": 0.1906,
+      "mean_token_accuracy": 0.9386573076248169,
+      "num_tokens": 46759404.0,
+      "step": 5760
+    },
+    {
+      "entropy": 0.31055004596710206,
+      "epoch": 3.2533860045146725,
+      "grad_norm": 1.7997633218765259,
+      "learning_rate": 4.760523783638997e-06,
+      "loss": 0.1985,
+      "mean_token_accuracy": 0.9364279508590698,
+      "num_tokens": 46799993.0,
+      "step": 5765
+    },
+    {
+      "entropy": 0.317557692527771,
+      "epoch": 3.2562076749435667,
+      "grad_norm": 1.7583611011505127,
+      "learning_rate": 4.76011723271764e-06,
+      "loss": 0.2218,
+      "mean_token_accuracy": 0.9295771718025208,
+      "num_tokens": 46840518.0,
+      "step": 5770
+    },
+    {
+      "entropy": 0.3186119019985199,
+      "epoch": 3.2590293453724604,
+      "grad_norm": 2.132106065750122,
+      "learning_rate": 4.75971036057729e-06,
+      "loss": 0.2093,
+      "mean_token_accuracy": 0.9330814242362976,
+      "num_tokens": 46881236.0,
+      "step": 5775
+    },
+    {
+      "entropy": 0.2968993723392487,
+      "epoch": 3.2618510158013545,
+      "grad_norm": 1.693253993988037,
+      "learning_rate": 4.759303167297877e-06,
+      "loss": 0.1979,
+      "mean_token_accuracy": 0.9351974129676819,
+      "num_tokens": 46921984.0,
+      "step": 5780
+    },
+    {
+      "entropy": 0.2921249568462372,
+      "epoch": 3.264672686230248,
+      "grad_norm": 2.1287200450897217,
+      "learning_rate": 4.758895652959394e-06,
+      "loss": 0.198,
+      "mean_token_accuracy": 0.9361931085586548,
+      "num_tokens": 46962730.0,
+      "step": 5785
+    },
+    {
+      "entropy": 0.3355319321155548,
+      "epoch": 3.2674943566591423,
+      "grad_norm": 1.6830389499664307,
+      "learning_rate": 4.758487817641898e-06,
+      "loss": 0.2355,
+      "mean_token_accuracy": 0.9263626217842102,
+      "num_tokens": 47002800.0,
+      "step": 5790
+    },
+    {
+      "entropy": 0.32140293121337893,
+      "epoch": 3.270316027088036,
+      "grad_norm": 1.845828652381897,
+      "learning_rate": 4.758079661425508e-06,
+      "loss": 0.2129,
+      "mean_token_accuracy": 0.9305433511734009,
+      "num_tokens": 47043526.0,
+      "step": 5795
+    },
+    {
+      "entropy": 0.32089059948921206,
+      "epoch": 3.27313769751693,
+      "grad_norm": 1.3842658996582031,
+      "learning_rate": 4.757671184390406e-06,
+      "loss": 0.2152,
+      "mean_token_accuracy": 0.9322557806968689,
+      "num_tokens": 47084134.0,
+      "step": 5800
+    },
+    {
+      "entropy": 0.33169599771499636,
+      "epoch": 3.275959367945824,
+      "grad_norm": 1.901238203048706,
+      "learning_rate": 4.757262386616837e-06,
+      "loss": 0.2173,
+      "mean_token_accuracy": 0.9306012511253356,
+      "num_tokens": 47124773.0,
+      "step": 5805
+    },
+    {
+      "entropy": 0.3124689519405365,
+      "epoch": 3.278781038374718,
+      "grad_norm": 2.120180368423462,
+      "learning_rate": 4.75685326818511e-06,
+      "loss": 0.2199,
+      "mean_token_accuracy": 0.9284607768058777,
+      "num_tokens": 47165252.0,
+      "step": 5810
+    },
+    {
+      "entropy": 0.28955113887786865,
+      "epoch": 3.2816027088036117,
+      "grad_norm": 2.0301990509033203,
+      "learning_rate": 4.756443829175598e-06,
+      "loss": 0.1989,
+      "mean_token_accuracy": 0.9357622146606446,
+      "num_tokens": 47205788.0,
+      "step": 5815
+    },
+    {
+      "entropy": 0.33322848081588746,
+      "epoch": 3.2844243792325054,
+      "grad_norm": 1.8951489925384521,
+      "learning_rate": 4.756034069668732e-06,
+      "loss": 0.2272,
+      "mean_token_accuracy": 0.9254481196403503,
+      "num_tokens": 47246598.0,
+      "step": 5820
+    },
+    {
+      "entropy": 0.3263582348823547,
+      "epoch": 3.2872460496613995,
+      "grad_norm": 1.7902400493621826,
+      "learning_rate": 4.7556239897450116e-06,
+      "loss": 0.2064,
+      "mean_token_accuracy": 0.9346579432487487,
+      "num_tokens": 47287035.0,
+      "step": 5825
+    },
+    {
+      "entropy": 0.3200534522533417,
+      "epoch": 3.2900677200902937,
+      "grad_norm": 2.2464659214019775,
+      "learning_rate": 4.7552135894849965e-06,
+      "loss": 0.2172,
+      "mean_token_accuracy": 0.9306413531303406,
+      "num_tokens": 47327903.0,
+      "step": 5830
+    },
+    {
+      "entropy": 0.29628766179084776,
+      "epoch": 3.2928893905191874,
+      "grad_norm": 1.7990299463272095,
+      "learning_rate": 4.75480286896931e-06,
+      "loss": 0.1918,
+      "mean_token_accuracy": 0.9378930330276489,
+      "num_tokens": 47368326.0,
+      "step": 5835
+    },
+    {
+      "entropy": 0.29361318349838256,
+      "epoch": 3.295711060948081,
+      "grad_norm": 2.020881175994873,
+      "learning_rate": 4.754391828278638e-06,
+      "loss": 0.2086,
+      "mean_token_accuracy": 0.9333640217781067,
+      "num_tokens": 47409098.0,
+      "step": 5840
+    },
+    {
+      "entropy": 0.3207947850227356,
+      "epoch": 3.2985327313769752,
+      "grad_norm": 2.123802423477173,
+      "learning_rate": 4.753980467493729e-06,
+      "loss": 0.2166,
+      "mean_token_accuracy": 0.929839301109314,
+      "num_tokens": 47449708.0,
+      "step": 5845
+    },
+    {
+      "entropy": 0.314586877822876,
+      "epoch": 3.301354401805869,
+      "grad_norm": 1.9354177713394165,
+      "learning_rate": 4.753568786695395e-06,
+      "loss": 0.2091,
+      "mean_token_accuracy": 0.9324597477912903,
+      "num_tokens": 47490310.0,
+      "step": 5850
+    },
+    {
+      "entropy": 0.30399842858314513,
+      "epoch": 3.304176072234763,
+      "grad_norm": 1.9815207719802856,
+      "learning_rate": 4.753156785964512e-06,
+      "loss": 0.2005,
+      "mean_token_accuracy": 0.9356989383697509,
+      "num_tokens": 47530834.0,
+      "step": 5855
+    },
+    {
+      "entropy": 0.2951928973197937,
+      "epoch": 3.3069977426636568,
+      "grad_norm": 2.1382501125335693,
+      "learning_rate": 4.752744465382016e-06,
+      "loss": 0.1979,
+      "mean_token_accuracy": 0.9363240480422974,
+      "num_tokens": 47571446.0,
+      "step": 5860
+    },
+    {
+      "entropy": 0.30987902283668517,
+      "epoch": 3.309819413092551,
+      "grad_norm": 1.7661712169647217,
+      "learning_rate": 4.75233182502891e-06,
+      "loss": 0.1998,
+      "mean_token_accuracy": 0.9360772490501403,
+      "num_tokens": 47612058.0,
+      "step": 5865
+    },
+    {
+      "entropy": 0.29606603980064394,
+      "epoch": 3.3126410835214446,
+      "grad_norm": 1.722391963005066,
+      "learning_rate": 4.751918864986254e-06,
+      "loss": 0.1834,
+      "mean_token_accuracy": 0.9418554663658142,
+      "num_tokens": 47652000.0,
+      "step": 5870
+    },
+    {
+      "entropy": 0.29225061237812044,
+      "epoch": 3.3154627539503387,
+      "grad_norm": 2.109041929244995,
+      "learning_rate": 4.751505585335176e-06,
+      "loss": 0.2005,
+      "mean_token_accuracy": 0.9353228807449341,
+      "num_tokens": 47692508.0,
+      "step": 5875
+    },
+    {
+      "entropy": 0.29792939126491547,
+      "epoch": 3.3182844243792324,
+      "grad_norm": 1.6946020126342773,
+      "learning_rate": 4.751091986156864e-06,
+      "loss": 0.1992,
+      "mean_token_accuracy": 0.9354248166084289,
+      "num_tokens": 47733266.0,
+      "step": 5880
+    },
+    {
+      "entropy": 0.29692640602588655,
+      "epoch": 3.3211060948081266,
+      "grad_norm": 2.0159807205200195,
+      "learning_rate": 4.750678067532569e-06,
+      "loss": 0.1937,
+      "mean_token_accuracy": 0.9390650391578674,
+      "num_tokens": 47773918.0,
+      "step": 5885
+    },
+    {
+      "entropy": 0.29629198312759397,
+      "epoch": 3.3239277652370203,
+      "grad_norm": 2.0621113777160645,
+      "learning_rate": 4.750263829543608e-06,
+      "loss": 0.1907,
+      "mean_token_accuracy": 0.9389678955078125,
+      "num_tokens": 47814708.0,
+      "step": 5890
+    },
+    {
+      "entropy": 0.3075540721416473,
+      "epoch": 3.3267494356659144,
+      "grad_norm": 2.2669737339019775,
+      "learning_rate": 4.749849272271355e-06,
+      "loss": 0.2074,
+      "mean_token_accuracy": 0.9335337281227112,
+      "num_tokens": 47855163.0,
+      "step": 5895
+    },
+    {
+      "entropy": 0.3051905155181885,
+      "epoch": 3.329571106094808,
+      "grad_norm": 1.7983601093292236,
+      "learning_rate": 4.749434395797252e-06,
+      "loss": 0.2013,
+      "mean_token_accuracy": 0.9364488601684571,
+      "num_tokens": 47895912.0,
+      "step": 5900
+    },
+    {
+      "entropy": 0.30544620752334595,
+      "epoch": 3.332392776523702,
+      "grad_norm": 2.127479076385498,
+      "learning_rate": 4.749019200202801e-06,
+      "loss": 0.2216,
+      "mean_token_accuracy": 0.9293683290481567,
+      "num_tokens": 47936365.0,
+      "step": 5905
+    },
+    {
+      "entropy": 0.3163418352603912,
+      "epoch": 3.335214446952596,
+      "grad_norm": 1.8929691314697266,
+      "learning_rate": 4.748603685569566e-06,
+      "loss": 0.2188,
+      "mean_token_accuracy": 0.9315417289733887,
+      "num_tokens": 47977087.0,
+      "step": 5910
+    },
+    {
+      "entropy": 0.2961422860622406,
+      "epoch": 3.33803611738149,
+      "grad_norm": 1.8902735710144043,
+      "learning_rate": 4.7481878519791775e-06,
+      "loss": 0.2024,
+      "mean_token_accuracy": 0.9347827434539795,
+      "num_tokens": 48017600.0,
+      "step": 5915
+    },
+    {
+      "entropy": 0.3079638063907623,
+      "epoch": 3.340857787810384,
+      "grad_norm": 3.9220144748687744,
+      "learning_rate": 4.747771699513324e-06,
+      "loss": 0.204,
+      "mean_token_accuracy": 0.9359039187431335,
+      "num_tokens": 48058181.0,
+      "step": 5920
+    },
+    {
+      "entropy": 0.30697737336158754,
+      "epoch": 3.3436794582392775,
+      "grad_norm": 1.9922391176223755,
+      "learning_rate": 4.747355228253759e-06,
+      "loss": 0.2005,
+      "mean_token_accuracy": 0.9362172484397888,
+      "num_tokens": 48098969.0,
+      "step": 5925
+    },
+    {
+      "entropy": 0.29278679490089415,
+      "epoch": 3.3465011286681716,
+      "grad_norm": 1.6647714376449585,
+      "learning_rate": 4.746938438282297e-06,
+      "loss": 0.1868,
+      "mean_token_accuracy": 0.9388912916183472,
+      "num_tokens": 48139626.0,
+      "step": 5930
+    },
+    {
+      "entropy": 0.32587441205978396,
+      "epoch": 3.3493227990970653,
+      "grad_norm": 1.9658551216125488,
+      "learning_rate": 4.74652132968082e-06,
+      "loss": 0.2199,
+      "mean_token_accuracy": 0.931458306312561,
+      "num_tokens": 48180389.0,
+      "step": 5935
+    },
+    {
+      "entropy": 0.3118497312068939,
+      "epoch": 3.3521444695259595,
+      "grad_norm": 1.961451768875122,
+      "learning_rate": 4.746103902531266e-06,
+      "loss": 0.2137,
+      "mean_token_accuracy": 0.9307841777801513,
+      "num_tokens": 48221035.0,
+      "step": 5940
+    },
+    {
+      "entropy": 0.29938756227493285,
+      "epoch": 3.354966139954853,
+      "grad_norm": 1.9882782697677612,
+      "learning_rate": 4.7456861569156396e-06,
+      "loss": 0.2026,
+      "mean_token_accuracy": 0.9369753360748291,
+      "num_tokens": 48261699.0,
+      "step": 5945
+    },
+    {
+      "entropy": 0.297002974152565,
+      "epoch": 3.3577878103837473,
+      "grad_norm": 1.5427496433258057,
+      "learning_rate": 4.745268092916008e-06,
+      "loss": 0.1919,
+      "mean_token_accuracy": 0.9388345956802369,
+      "num_tokens": 48302567.0,
+      "step": 5950
+    },
+    {
+      "entropy": 0.31234695911407473,
+      "epoch": 3.360609480812641,
+      "grad_norm": 1.856497049331665,
+      "learning_rate": 4.744849710614498e-06,
+      "loss": 0.2081,
+      "mean_token_accuracy": 0.9339388251304627,
+      "num_tokens": 48343071.0,
+      "step": 5955
+    },
+    {
+      "entropy": 0.26803739964962003,
+      "epoch": 3.363431151241535,
+      "grad_norm": 1.5363012552261353,
+      "learning_rate": 4.744431010093302e-06,
+      "loss": 0.1604,
+      "mean_token_accuracy": 0.9466527342796326,
+      "num_tokens": 48383501.0,
+      "step": 5960
+    },
+    {
+      "entropy": 0.3034593999385834,
+      "epoch": 3.366252821670429,
+      "grad_norm": 1.9353678226470947,
+      "learning_rate": 4.744011991434673e-06,
+      "loss": 0.1963,
+      "mean_token_accuracy": 0.9345849275588989,
+      "num_tokens": 48424004.0,
+      "step": 5965
+    },
+    {
+      "entropy": 0.3189900636672974,
+      "epoch": 3.369074492099323,
+      "grad_norm": 1.8158825635910034,
+      "learning_rate": 4.743592654720929e-06,
+      "loss": 0.2096,
+      "mean_token_accuracy": 0.9319984436035156,
+      "num_tokens": 48464769.0,
+      "step": 5970
+    },
+    {
+      "entropy": 0.29853619933128356,
+      "epoch": 3.3718961625282167,
+      "grad_norm": 2.1045563220977783,
+      "learning_rate": 4.743173000034446e-06,
+      "loss": 0.1981,
+      "mean_token_accuracy": 0.9359387755393982,
+      "num_tokens": 48505548.0,
+      "step": 5975
+    },
+    {
+      "entropy": 0.2892568349838257,
+      "epoch": 3.374717832957111,
+      "grad_norm": 1.914672613143921,
+      "learning_rate": 4.7427530274576685e-06,
+      "loss": 0.184,
+      "mean_token_accuracy": 0.9403793454170227,
+      "num_tokens": 48546282.0,
+      "step": 5980
+    },
+    {
+      "entropy": 0.3369791269302368,
+      "epoch": 3.3775395033860045,
+      "grad_norm": 1.855833649635315,
+      "learning_rate": 4.742332737073098e-06,
+      "loss": 0.2296,
+      "mean_token_accuracy": 0.9269206643104553,
+      "num_tokens": 48586933.0,
+      "step": 5985
+    },
+    {
+      "entropy": 0.33269967436790465,
+      "epoch": 3.380361173814898,
+      "grad_norm": 1.9749621152877808,
+      "learning_rate": 4.741912128963301e-06,
+      "loss": 0.2292,
+      "mean_token_accuracy": 0.9259452581405639,
+      "num_tokens": 48627643.0,
+      "step": 5990
+    },
+    {
+      "entropy": 0.3066904067993164,
+      "epoch": 3.3831828442437923,
+      "grad_norm": 2.0527172088623047,
+      "learning_rate": 4.741491203210906e-06,
+      "loss": 0.2026,
+      "mean_token_accuracy": 0.9332493662834167,
+      "num_tokens": 48668300.0,
+      "step": 5995
+    },
+    {
+      "entropy": 0.3129414677619934,
+      "epoch": 3.386004514672686,
+      "grad_norm": 1.7080130577087402,
+      "learning_rate": 4.741069959898603e-06,
+      "loss": 0.2183,
+      "mean_token_accuracy": 0.929611599445343,
+      "num_tokens": 48708814.0,
+      "step": 6000
+    },
+    {
+      "epoch": 3.386004514672686,
+      "eval_entropy": 0.3240194022655487,
+      "eval_loss": 0.18876124918460846,
+      "eval_mean_token_accuracy": 0.9436565637588501,
+      "eval_num_tokens": 48708814.0,
+      "eval_runtime": 0.1645,
+      "eval_samples_per_second": 24.322,
+      "eval_steps_per_second": 6.08,
+      "step": 6000
+    },
+    {
+      "entropy": 0.31379024386405946,
+      "epoch": 3.38882618510158,
+      "grad_norm": 1.7492561340332031,
+      "learning_rate": 4.740648399109148e-06,
+      "loss": 0.2158,
+      "mean_token_accuracy": 0.9303012132644654,
+      "num_tokens": 48749515.0,
+      "step": 6005
+    },
+    {
+      "entropy": 0.27735379338264465,
+      "epoch": 3.391647855530474,
+      "grad_norm": 1.8119415044784546,
+      "learning_rate": 4.740226520925354e-06,
+      "loss": 0.1774,
+      "mean_token_accuracy": 0.9421341061592102,
+      "num_tokens": 48790013.0,
+      "step": 6010
+    },
+    {
+      "entropy": 0.27942020893096925,
+      "epoch": 3.394469525959368,
+      "grad_norm": 1.6023916006088257,
+      "learning_rate": 4.7398043254301e-06,
+      "loss": 0.1803,
+      "mean_token_accuracy": 0.94042888879776,
+      "num_tokens": 48830739.0,
+      "step": 6015
+    },
+    {
+      "entropy": 0.3403823971748352,
+      "epoch": 3.3972911963882617,
+      "grad_norm": 1.8019448518753052,
+      "learning_rate": 4.739381812706326e-06,
+      "loss": 0.2299,
+      "mean_token_accuracy": 0.9285380721092225,
+      "num_tokens": 48871272.0,
+      "step": 6020
+    },
+    {
+      "entropy": 0.30235469341278076,
+      "epoch": 3.400112866817156,
+      "grad_norm": 2.0274271965026855,
+      "learning_rate": 4.738958982837036e-06,
+      "loss": 0.208,
+      "mean_token_accuracy": 0.9346428751945496,
+      "num_tokens": 48911923.0,
+      "step": 6025
+    },
+    {
+      "entropy": 0.32721391320228577,
+      "epoch": 3.4029345372460496,
+      "grad_norm": 1.9175236225128174,
+      "learning_rate": 4.738535835905294e-06,
+      "loss": 0.2291,
+      "mean_token_accuracy": 0.9262001872062683,
+      "num_tokens": 48952801.0,
+      "step": 6030
+    },
+    {
+      "entropy": 0.3034025192260742,
+      "epoch": 3.4057562076749437,
+      "grad_norm": 1.9815678596496582,
+      "learning_rate": 4.738112371994227e-06,
+      "loss": 0.2018,
+      "mean_token_accuracy": 0.9346770763397216,
+      "num_tokens": 48993273.0,
+      "step": 6035
+    },
+    {
+      "entropy": 0.3161701142787933,
+      "epoch": 3.4085778781038374,
+      "grad_norm": 1.633199691772461,
+      "learning_rate": 4.737688591187024e-06,
+      "loss": 0.2118,
+      "mean_token_accuracy": 0.9302195191383362,
+      "num_tokens": 49033969.0,
+      "step": 6040
+    },
+    {
+      "entropy": 0.30224609375,
+      "epoch": 3.4113995485327315,
+      "grad_norm": 1.9072091579437256,
+      "learning_rate": 4.737264493566939e-06,
+      "loss": 0.2037,
+      "mean_token_accuracy": 0.933539628982544,
+      "num_tokens": 49074370.0,
+      "step": 6045
+    },
+    {
+      "entropy": 0.33060076236724856,
+      "epoch": 3.4142212189616252,
+      "grad_norm": 2.085325241088867,
+      "learning_rate": 4.736840079217284e-06,
+      "loss": 0.2364,
+      "mean_token_accuracy": 0.9252678632736206,
+      "num_tokens": 49114878.0,
+      "step": 6050
+    },
+    {
+      "entropy": 0.2963068068027496,
+      "epoch": 3.4170428893905194,
+      "grad_norm": 1.7405322790145874,
+      "learning_rate": 4.736415348221435e-06,
+      "loss": 0.1954,
+      "mean_token_accuracy": 0.9355825662612915,
+      "num_tokens": 49155589.0,
+      "step": 6055
+    },
+    {
+      "entropy": 0.3319820463657379,
+      "epoch": 3.419864559819413,
+      "grad_norm": 1.9911015033721924,
+      "learning_rate": 4.735990300662833e-06,
+      "loss": 0.2195,
+      "mean_token_accuracy": 0.9308295607566833,
+      "num_tokens": 49196038.0,
+      "step": 6060
+    },
+    {
+      "entropy": 0.29007573127746583,
+      "epoch": 3.422686230248307,
+      "grad_norm": 1.579049825668335,
+      "learning_rate": 4.7355649366249755e-06,
+      "loss": 0.1962,
+      "mean_token_accuracy": 0.9360047936439514,
+      "num_tokens": 49236253.0,
+      "step": 6065
+    },
+    {
+      "entropy": 0.28036363422870636,
+      "epoch": 3.425507900677201,
+      "grad_norm": 1.846561312675476,
+      "learning_rate": 4.735139256191428e-06,
+      "loss": 0.1917,
+      "mean_token_accuracy": 0.937602186203003,
+      "num_tokens": 49276934.0,
+      "step": 6070
+    },
+    {
+      "entropy": 0.30836820006370547,
+      "epoch": 3.4283295711060946,
+      "grad_norm": 2.027714967727661,
+      "learning_rate": 4.734713259445814e-06,
+      "loss": 0.2116,
+      "mean_token_accuracy": 0.9310431718826294,
+      "num_tokens": 49317527.0,
+      "step": 6075
+    },
+    {
+      "entropy": 0.2989709198474884,
+      "epoch": 3.4311512415349887,
+      "grad_norm": 1.928396224975586,
+      "learning_rate": 4.734286946471821e-06,
+      "loss": 0.2028,
+      "mean_token_accuracy": 0.9341145396232605,
+      "num_tokens": 49358187.0,
+      "step": 6080
+    },
+    {
+      "entropy": 0.29576932787895205,
+      "epoch": 3.4339729119638824,
+      "grad_norm": 1.8478814363479614,
+      "learning_rate": 4.733860317353198e-06,
+      "loss": 0.201,
+      "mean_token_accuracy": 0.9338787913322448,
+      "num_tokens": 49398870.0,
+      "step": 6085
+    },
+    {
+      "entropy": 0.34390089511871336,
+      "epoch": 3.4367945823927766,
+      "grad_norm": 2.0840272903442383,
+      "learning_rate": 4.733433372173756e-06,
+      "loss": 0.2391,
+      "mean_token_accuracy": 0.9246963381767273,
+      "num_tokens": 49439698.0,
+      "step": 6090
+    },
+    {
+      "entropy": 0.3055784046649933,
+      "epoch": 3.4396162528216703,
+      "grad_norm": 1.8206205368041992,
+      "learning_rate": 4.73300611101737e-06,
+      "loss": 0.2052,
+      "mean_token_accuracy": 0.933589768409729,
+      "num_tokens": 49480359.0,
+      "step": 6095
+    },
+    {
+      "entropy": 0.3226096034049988,
+      "epoch": 3.4424379232505644,
+      "grad_norm": 2.05794620513916,
+      "learning_rate": 4.732578533967974e-06,
+      "loss": 0.2108,
+      "mean_token_accuracy": 0.9320708394050599,
+      "num_tokens": 49521140.0,
+      "step": 6100
+    },
+    {
+      "entropy": 0.31547321677207946,
+      "epoch": 3.445259593679458,
+      "grad_norm": 2.0338547229766846,
+      "learning_rate": 4.732150641109566e-06,
+      "loss": 0.2055,
+      "mean_token_accuracy": 0.93296377658844,
+      "num_tokens": 49561801.0,
+      "step": 6105
+    },
+    {
+      "entropy": 0.32915824055671694,
+      "epoch": 3.4480812641083523,
+      "grad_norm": 1.9995898008346558,
+      "learning_rate": 4.731722432526206e-06,
+      "loss": 0.2213,
+      "mean_token_accuracy": 0.9298017382621765,
+      "num_tokens": 49602445.0,
+      "step": 6110
+    },
+    {
+      "entropy": 0.31122357249259947,
+      "epoch": 3.450902934537246,
+      "grad_norm": 1.9306919574737549,
+      "learning_rate": 4.731293908302014e-06,
+      "loss": 0.193,
+      "mean_token_accuracy": 0.9369074702262878,
+      "num_tokens": 49643049.0,
+      "step": 6115
+    },
+    {
+      "entropy": 0.283711576461792,
+      "epoch": 3.45372460496614,
+      "grad_norm": 1.7749558687210083,
+      "learning_rate": 4.730865068521177e-06,
+      "loss": 0.1865,
+      "mean_token_accuracy": 0.9403268456459045,
+      "num_tokens": 49683779.0,
+      "step": 6120
+    },
+    {
+      "entropy": 0.3096657872200012,
+      "epoch": 3.456546275395034,
+      "grad_norm": 1.9655077457427979,
+      "learning_rate": 4.730435913267937e-06,
+      "loss": 0.2029,
+      "mean_token_accuracy": 0.9331951022148133,
+      "num_tokens": 49724297.0,
+      "step": 6125
+    },
+    {
+      "entropy": 0.3291502416133881,
+      "epoch": 3.459367945823928,
+      "grad_norm": 1.9921224117279053,
+      "learning_rate": 4.7300064426266035e-06,
+      "loss": 0.2186,
+      "mean_token_accuracy": 0.9291182398796082,
+      "num_tokens": 49764911.0,
+      "step": 6130
+    },
+    {
+      "entropy": 0.31880820393562315,
+      "epoch": 3.4621896162528216,
+      "grad_norm": 1.8428974151611328,
+      "learning_rate": 4.729576656681545e-06,
+      "loss": 0.2154,
+      "mean_token_accuracy": 0.9307239055633545,
+      "num_tokens": 49805553.0,
+      "step": 6135
+    },
+    {
+      "entropy": 0.29867430329322814,
+      "epoch": 3.4650112866817158,
+      "grad_norm": 2.0829527378082275,
+      "learning_rate": 4.729146555517195e-06,
+      "loss": 0.2009,
+      "mean_token_accuracy": 0.9337917327880859,
+      "num_tokens": 49846334.0,
+      "step": 6140
+    },
+    {
+      "entropy": 0.3254325449466705,
+      "epoch": 3.4678329571106095,
+      "grad_norm": 1.9901976585388184,
+      "learning_rate": 4.728716139218045e-06,
+      "loss": 0.2165,
+      "mean_token_accuracy": 0.9306549906730652,
+      "num_tokens": 49886713.0,
+      "step": 6145
+    },
+    {
+      "entropy": 0.2868054747581482,
+      "epoch": 3.4706546275395036,
+      "grad_norm": 1.7727199792861938,
+      "learning_rate": 4.728285407868651e-06,
+      "loss": 0.1976,
+      "mean_token_accuracy": 0.936365807056427,
+      "num_tokens": 49927340.0,
+      "step": 6150
+    },
+    {
+      "entropy": 0.3261713206768036,
+      "epoch": 3.4734762979683973,
+      "grad_norm": 1.8373956680297852,
+      "learning_rate": 4.72785436155363e-06,
+      "loss": 0.2094,
+      "mean_token_accuracy": 0.9324244618415832,
+      "num_tokens": 49967868.0,
+      "step": 6155
+    },
+    {
+      "entropy": 0.3357150912284851,
+      "epoch": 3.476297968397291,
+      "grad_norm": 1.9818036556243896,
+      "learning_rate": 4.7274230003576625e-06,
+      "loss": 0.2388,
+      "mean_token_accuracy": 0.9242081761360168,
+      "num_tokens": 50008499.0,
+      "step": 6160
+    },
+    {
+      "entropy": 0.3065942347049713,
+      "epoch": 3.479119638826185,
+      "grad_norm": 1.6570138931274414,
+      "learning_rate": 4.726991324365487e-06,
+      "loss": 0.2021,
+      "mean_token_accuracy": 0.9351112723350525,
+      "num_tokens": 50049217.0,
+      "step": 6165
+    },
+    {
+      "entropy": 0.2773208677768707,
+      "epoch": 3.481941309255079,
+      "grad_norm": 1.8118728399276733,
+      "learning_rate": 4.726559333661908e-06,
+      "loss": 0.1871,
+      "mean_token_accuracy": 0.9404791593551636,
+      "num_tokens": 50090066.0,
+      "step": 6170
+    },
+    {
+      "entropy": 0.3141829252243042,
+      "epoch": 3.484762979683973,
+      "grad_norm": 1.8138505220413208,
+      "learning_rate": 4.726127028331789e-06,
+      "loss": 0.2124,
+      "mean_token_accuracy": 0.9332842707633973,
+      "num_tokens": 50130759.0,
+      "step": 6175
+    },
+    {
+      "entropy": 0.29586785435676577,
+      "epoch": 3.4875846501128667,
+      "grad_norm": 2.0417232513427734,
+      "learning_rate": 4.725694408460059e-06,
+      "loss": 0.1986,
+      "mean_token_accuracy": 0.9360351800918579,
+      "num_tokens": 50171307.0,
+      "step": 6180
+    },
+    {
+      "entropy": 0.3109571814537048,
+      "epoch": 3.490406320541761,
+      "grad_norm": 2.048214912414551,
+      "learning_rate": 4.725261474131703e-06,
+      "loss": 0.207,
+      "mean_token_accuracy": 0.9328826665878296,
+      "num_tokens": 50211913.0,
+      "step": 6185
+    },
+    {
+      "entropy": 0.3329827606678009,
+      "epoch": 3.4932279909706545,
+      "grad_norm": 1.681795597076416,
+      "learning_rate": 4.724828225431772e-06,
+      "loss": 0.2283,
+      "mean_token_accuracy": 0.9269816160202027,
+      "num_tokens": 50252502.0,
+      "step": 6190
+    },
+    {
+      "entropy": 0.29528833031654356,
+      "epoch": 3.4960496613995486,
+      "grad_norm": 1.5485318899154663,
+      "learning_rate": 4.72439466244538e-06,
+      "loss": 0.1888,
+      "mean_token_accuracy": 0.9390155076980591,
+      "num_tokens": 50293239.0,
+      "step": 6195
+    },
+    {
+      "entropy": 0.30525763630867003,
+      "epoch": 3.4988713318284423,
+      "grad_norm": 1.8242753744125366,
+      "learning_rate": 4.723960785257697e-06,
+      "loss": 0.207,
+      "mean_token_accuracy": 0.9336790919303894,
+      "num_tokens": 50333949.0,
+      "step": 6200
+    },
+    {
+      "entropy": 0.329301780462265,
+      "epoch": 3.5016930022573365,
+      "grad_norm": 1.7903170585632324,
+      "learning_rate": 4.72352659395396e-06,
+      "loss": 0.2453,
+      "mean_token_accuracy": 0.9249015927314759,
+      "num_tokens": 50374411.0,
+      "step": 6205
+    },
+    {
+      "entropy": 0.30608891248703,
+      "epoch": 3.50451467268623,
+      "grad_norm": 1.8584059476852417,
+      "learning_rate": 4.7230920886194655e-06,
+      "loss": 0.2178,
+      "mean_token_accuracy": 0.9296943068504333,
+      "num_tokens": 50415248.0,
+      "step": 6210
+    },
+    {
+      "entropy": 0.3110420048236847,
+      "epoch": 3.5073363431151243,
+      "grad_norm": 1.9278223514556885,
+      "learning_rate": 4.722657269339573e-06,
+      "loss": 0.2087,
+      "mean_token_accuracy": 0.9321151494979858,
+      "num_tokens": 50455903.0,
+      "step": 6215
+    },
+    {
+      "entropy": 0.29749436378479005,
+      "epoch": 3.510158013544018,
+      "grad_norm": 1.6023099422454834,
+      "learning_rate": 4.722222136199703e-06,
+      "loss": 0.1862,
+      "mean_token_accuracy": 0.940057122707367,
+      "num_tokens": 50496595.0,
+      "step": 6220
+    },
+    {
+      "entropy": 0.3107947289943695,
+      "epoch": 3.5129796839729117,
+      "grad_norm": 1.7268322706222534,
+      "learning_rate": 4.7217866892853355e-06,
+      "loss": 0.1982,
+      "mean_token_accuracy": 0.9365177512168884,
+      "num_tokens": 50537156.0,
+      "step": 6225
+    },
+    {
+      "entropy": 0.31932695508003234,
+      "epoch": 3.515801354401806,
+      "grad_norm": 1.9422742128372192,
+      "learning_rate": 4.721350928682017e-06,
+      "loss": 0.2254,
+      "mean_token_accuracy": 0.9268533945083618,
+      "num_tokens": 50577880.0,
+      "step": 6230
+    },
+    {
+      "entropy": 0.3243493378162384,
+      "epoch": 3.5186230248307,
+      "grad_norm": 1.8722442388534546,
+      "learning_rate": 4.720914854475349e-06,
+      "loss": 0.2181,
+      "mean_token_accuracy": 0.9315003275871276,
+      "num_tokens": 50618659.0,
+      "step": 6235
+    },
+    {
+      "entropy": 0.30051770210266116,
+      "epoch": 3.5214446952595937,
+      "grad_norm": 1.5247653722763062,
+      "learning_rate": 4.720478466751002e-06,
+      "loss": 0.1931,
+      "mean_token_accuracy": 0.9378492355346679,
+      "num_tokens": 50659414.0,
+      "step": 6240
+    },
+    {
+      "entropy": 0.3169930100440979,
+      "epoch": 3.5242663656884874,
+      "grad_norm": 2.1740469932556152,
+      "learning_rate": 4.720041765594701e-06,
+      "loss": 0.2171,
+      "mean_token_accuracy": 0.9304054498672485,
+      "num_tokens": 50700091.0,
+      "step": 6245
+    },
+    {
+      "entropy": 0.3257488250732422,
+      "epoch": 3.5270880361173815,
+      "grad_norm": 1.9426953792572021,
+      "learning_rate": 4.719604751092239e-06,
+      "loss": 0.221,
+      "mean_token_accuracy": 0.928559148311615,
+      "num_tokens": 50740627.0,
+      "step": 6250
+    },
+    {
+      "entropy": 0.31315135955810547,
+      "epoch": 3.5299097065462757,
+      "grad_norm": 1.9665589332580566,
+      "learning_rate": 4.719167423329467e-06,
+      "loss": 0.2139,
+      "mean_token_accuracy": 0.9316304087638855,
+      "num_tokens": 50781438.0,
+      "step": 6255
+    },
+    {
+      "entropy": 0.3040352761745453,
+      "epoch": 3.5327313769751694,
+      "grad_norm": 2.057408571243286,
+      "learning_rate": 4.718729782392297e-06,
+      "loss": 0.1961,
+      "mean_token_accuracy": 0.9354169130325317,
+      "num_tokens": 50822359.0,
+      "step": 6260
+    },
+    {
+      "entropy": 0.3147823929786682,
+      "epoch": 3.535553047404063,
+      "grad_norm": 2.11651349067688,
+      "learning_rate": 4.718291828366703e-06,
+      "loss": 0.2091,
+      "mean_token_accuracy": 0.9335868835449219,
+      "num_tokens": 50862733.0,
+      "step": 6265
+    },
+    {
+      "entropy": 0.3134510278701782,
+      "epoch": 3.538374717832957,
+      "grad_norm": 1.6482633352279663,
+      "learning_rate": 4.717853561338723e-06,
+      "loss": 0.2068,
+      "mean_token_accuracy": 0.9318931818008422,
+      "num_tokens": 50903490.0,
+      "step": 6270
+    },
+    {
+      "entropy": 0.3234701603651047,
+      "epoch": 3.541196388261851,
+      "grad_norm": 2.001359701156616,
+      "learning_rate": 4.717414981394454e-06,
+      "loss": 0.233,
+      "mean_token_accuracy": 0.9272037386894226,
+      "num_tokens": 50944012.0,
+      "step": 6275
+    },
+    {
+      "entropy": 0.2873412311077118,
+      "epoch": 3.544018058690745,
+      "grad_norm": 2.011967420578003,
+      "learning_rate": 4.716976088620055e-06,
+      "loss": 0.1917,
+      "mean_token_accuracy": 0.9380563020706176,
+      "num_tokens": 50984897.0,
+      "step": 6280
+    },
+    {
+      "entropy": 0.29418745040893557,
+      "epoch": 3.5468397291196387,
+      "grad_norm": 1.7138752937316895,
+      "learning_rate": 4.716536883101746e-06,
+      "loss": 0.1905,
+      "mean_token_accuracy": 0.9367462277412415,
+      "num_tokens": 51025455.0,
+      "step": 6285
+    },
+    {
+      "entropy": 0.3291135847568512,
+      "epoch": 3.549661399548533,
+      "grad_norm": 2.0384838581085205,
+      "learning_rate": 4.716097364925809e-06,
+      "loss": 0.2329,
+      "mean_token_accuracy": 0.9264655709266663,
+      "num_tokens": 51066205.0,
+      "step": 6290
+    },
+    {
+      "entropy": 0.31210439205169677,
+      "epoch": 3.5524830699774266,
+      "grad_norm": 1.871769905090332,
+      "learning_rate": 4.715657534178589e-06,
+      "loss": 0.2176,
+      "mean_token_accuracy": 0.9300808906555176,
+      "num_tokens": 51106895.0,
+      "step": 6295
+    },
+    {
+      "entropy": 0.2980155825614929,
+      "epoch": 3.5553047404063207,
+      "grad_norm": 1.6929700374603271,
+      "learning_rate": 4.715217390946489e-06,
+      "loss": 0.2014,
+      "mean_token_accuracy": 0.9350870847702026,
+      "num_tokens": 51147652.0,
+      "step": 6300
+    },
+    {
+      "entropy": 0.3304617702960968,
+      "epoch": 3.5581264108352144,
+      "grad_norm": 1.6658223867416382,
+      "learning_rate": 4.714776935315976e-06,
+      "loss": 0.2305,
+      "mean_token_accuracy": 0.9265113472938538,
+      "num_tokens": 51188269.0,
+      "step": 6305
+    },
+    {
+      "entropy": 0.30585951209068296,
+      "epoch": 3.560948081264108,
+      "grad_norm": 1.8482630252838135,
+      "learning_rate": 4.7143361673735774e-06,
+      "loss": 0.2076,
+      "mean_token_accuracy": 0.9323771357536316,
+      "num_tokens": 51228854.0,
+      "step": 6310
+    },
+    {
+      "entropy": 0.33988407254219055,
+      "epoch": 3.5637697516930023,
+      "grad_norm": 1.9136297702789307,
+      "learning_rate": 4.713895087205882e-06,
+      "loss": 0.2287,
+      "mean_token_accuracy": 0.9266070127487183,
+      "num_tokens": 51269073.0,
+      "step": 6315
+    },
+    {
+      "entropy": 0.3092219591140747,
+      "epoch": 3.5665914221218964,
+      "grad_norm": 1.8166900873184204,
+      "learning_rate": 4.71345369489954e-06,
+      "loss": 0.2046,
+      "mean_token_accuracy": 0.9348546147346497,
+      "num_tokens": 51309223.0,
+      "step": 6320
+    },
+    {
+      "entropy": 0.3356677830219269,
+      "epoch": 3.56941309255079,
+      "grad_norm": 1.9024507999420166,
+      "learning_rate": 4.7130119905412635e-06,
+      "loss": 0.2282,
+      "mean_token_accuracy": 0.927457618713379,
+      "num_tokens": 51349901.0,
+      "step": 6325
+    },
+    {
+      "entropy": 0.31340835690498353,
+      "epoch": 3.572234762979684,
+      "grad_norm": 1.8880701065063477,
+      "learning_rate": 4.712569974217826e-06,
+      "loss": 0.2102,
+      "mean_token_accuracy": 0.9327018618583679,
+      "num_tokens": 51390569.0,
+      "step": 6330
+    },
+    {
+      "entropy": 0.31486195921897886,
+      "epoch": 3.575056433408578,
+      "grad_norm": 2.056093215942383,
+      "learning_rate": 4.712127646016059e-06,
+      "loss": 0.211,
+      "mean_token_accuracy": 0.9327765345573426,
+      "num_tokens": 51431350.0,
+      "step": 6335
+    },
+    {
+      "entropy": 0.31965509057044983,
+      "epoch": 3.5778781038374716,
+      "grad_norm": 1.6555287837982178,
+      "learning_rate": 4.71168500602286e-06,
+      "loss": 0.2194,
+      "mean_token_accuracy": 0.9297954082489014,
+      "num_tokens": 51471728.0,
+      "step": 6340
+    },
+    {
+      "entropy": 0.34823336601257326,
+      "epoch": 3.5806997742663658,
+      "grad_norm": 1.9188921451568604,
+      "learning_rate": 4.7112420543251854e-06,
+      "loss": 0.2211,
+      "mean_token_accuracy": 0.9278290867805481,
+      "num_tokens": 51512451.0,
+      "step": 6345
+    },
+    {
+      "entropy": 0.27960309386253357,
+      "epoch": 3.5835214446952595,
+      "grad_norm": 1.7235885858535767,
+      "learning_rate": 4.710798791010054e-06,
+      "loss": 0.1939,
+      "mean_token_accuracy": 0.93757244348526,
+      "num_tokens": 51553127.0,
+      "step": 6350
+    },
+    {
+      "entropy": 0.30215221643447876,
+      "epoch": 3.5863431151241536,
+      "grad_norm": 1.8641575574874878,
+      "learning_rate": 4.710355216164543e-06,
+      "loss": 0.2143,
+      "mean_token_accuracy": 0.9315236330032348,
+      "num_tokens": 51593737.0,
+      "step": 6355
+    },
+    {
+      "entropy": 0.29333736300468444,
+      "epoch": 3.5891647855530473,
+      "grad_norm": 1.6100600957870483,
+      "learning_rate": 4.7099113298757934e-06,
+      "loss": 0.1993,
+      "mean_token_accuracy": 0.9377033352851868,
+      "num_tokens": 51634446.0,
+      "step": 6360
+    },
+    {
+      "entropy": 0.32711849808692933,
+      "epoch": 3.5919864559819414,
+      "grad_norm": 2.63490891456604,
+      "learning_rate": 4.709467132231007e-06,
+      "loss": 0.213,
+      "mean_token_accuracy": 0.9318007826805115,
+      "num_tokens": 51674890.0,
+      "step": 6365
+    },
+    {
+      "entropy": 0.3057248055934906,
+      "epoch": 3.594808126410835,
+      "grad_norm": 1.7993404865264893,
+      "learning_rate": 4.709022623317447e-06,
+      "loss": 0.2136,
+      "mean_token_accuracy": 0.9311740756034851,
+      "num_tokens": 51715328.0,
+      "step": 6370
+    },
+    {
+      "entropy": 0.32637726664543154,
+      "epoch": 3.5976297968397293,
+      "grad_norm": 1.7621092796325684,
+      "learning_rate": 4.708577803222437e-06,
+      "loss": 0.228,
+      "mean_token_accuracy": 0.926943325996399,
+      "num_tokens": 51755709.0,
+      "step": 6375
+    },
+    {
+      "entropy": 0.3093415260314941,
+      "epoch": 3.600451467268623,
+      "grad_norm": 1.7598847150802612,
+      "learning_rate": 4.708132672033361e-06,
+      "loss": 0.1968,
+      "mean_token_accuracy": 0.9351221680641174,
+      "num_tokens": 51796409.0,
+      "step": 6380
+    },
+    {
+      "entropy": 0.3067767798900604,
+      "epoch": 3.603273137697517,
+      "grad_norm": 1.894102931022644,
+      "learning_rate": 4.707687229837667e-06,
+      "loss": 0.2096,
+      "mean_token_accuracy": 0.9340128540992737,
+      "num_tokens": 51836874.0,
+      "step": 6385
+    },
+    {
+      "entropy": 0.31157588958740234,
+      "epoch": 3.606094808126411,
+      "grad_norm": 1.6844671964645386,
+      "learning_rate": 4.70724147672286e-06,
+      "loss": 0.2012,
+      "mean_token_accuracy": 0.9360531449317933,
+      "num_tokens": 51877598.0,
+      "step": 6390
+    },
+    {
+      "entropy": 0.30880475640296934,
+      "epoch": 3.6089164785553045,
+      "grad_norm": 1.8519737720489502,
+      "learning_rate": 4.706795412776509e-06,
+      "loss": 0.2139,
+      "mean_token_accuracy": 0.9301542401313782,
+      "num_tokens": 51918221.0,
+      "step": 6395
+    },
+    {
+      "entropy": 0.3159621596336365,
+      "epoch": 3.6117381489841986,
+      "grad_norm": 1.9711363315582275,
+      "learning_rate": 4.706349038086244e-06,
+      "loss": 0.218,
+      "mean_token_accuracy": 0.9309246063232421,
+      "num_tokens": 51959053.0,
+      "step": 6400
+    },
+    {
+      "entropy": 0.2915292739868164,
+      "epoch": 3.614559819413093,
+      "grad_norm": 1.8630083799362183,
+      "learning_rate": 4.7059023527397556e-06,
+      "loss": 0.1922,
+      "mean_token_accuracy": 0.9368671178817749,
+      "num_tokens": 51999218.0,
+      "step": 6405
+    },
+    {
+      "entropy": 0.3344090163707733,
+      "epoch": 3.6173814898419865,
+      "grad_norm": 1.93043851852417,
+      "learning_rate": 4.705455356824794e-06,
+      "loss": 0.2198,
+      "mean_token_accuracy": 0.9300200700759887,
+      "num_tokens": 52039875.0,
+      "step": 6410
+    },
+    {
+      "entropy": 0.3021121442317963,
+      "epoch": 3.62020316027088,
+      "grad_norm": 1.8353512287139893,
+      "learning_rate": 4.705008050429171e-06,
+      "loss": 0.2074,
+      "mean_token_accuracy": 0.9331298470497131,
+      "num_tokens": 52080363.0,
+      "step": 6415
+    },
+    {
+      "entropy": 0.30426907539367676,
+      "epoch": 3.6230248306997743,
+      "grad_norm": 1.7905832529067993,
+      "learning_rate": 4.704560433640762e-06,
+      "loss": 0.2117,
+      "mean_token_accuracy": 0.9322558403015136,
+      "num_tokens": 52121089.0,
+      "step": 6420
+    },
+    {
+      "entropy": 0.3197052538394928,
+      "epoch": 3.625846501128668,
+      "grad_norm": 1.7855573892593384,
+      "learning_rate": 4.7041125065475e-06,
+      "loss": 0.2154,
+      "mean_token_accuracy": 0.9314280033111573,
+      "num_tokens": 52161686.0,
+      "step": 6425
+    },
+    {
+      "entropy": 0.2890112340450287,
+      "epoch": 3.628668171557562,
+      "grad_norm": 2.0892858505249023,
+      "learning_rate": 4.703664269237381e-06,
+      "loss": 0.1937,
+      "mean_token_accuracy": 0.9377867579460144,
+      "num_tokens": 52202482.0,
+      "step": 6430
+    },
+    {
+      "entropy": 0.30858972668647766,
+      "epoch": 3.631489841986456,
+      "grad_norm": 1.9566696882247925,
+      "learning_rate": 4.703215721798462e-06,
+      "loss": 0.2021,
+      "mean_token_accuracy": 0.9352177262306214,
+      "num_tokens": 52243201.0,
+      "step": 6435
+    },
+    {
+      "entropy": 0.3409568965435028,
+      "epoch": 3.63431151241535,
+      "grad_norm": 1.81044602394104,
+      "learning_rate": 4.702766864318858e-06,
+      "loss": 0.2185,
+      "mean_token_accuracy": 0.9314499139785767,
+      "num_tokens": 52284031.0,
+      "step": 6440
+    },
+    {
+      "entropy": 0.3199546754360199,
+      "epoch": 3.6371331828442437,
+      "grad_norm": 1.9870702028274536,
+      "learning_rate": 4.70231769688675e-06,
+      "loss": 0.2192,
+      "mean_token_accuracy": 0.929648220539093,
+      "num_tokens": 52323965.0,
+      "step": 6445
+    },
+    {
+      "entropy": 0.3059984266757965,
+      "epoch": 3.639954853273138,
+      "grad_norm": 2.04311203956604,
+      "learning_rate": 4.701868219590374e-06,
+      "loss": 0.2123,
+      "mean_token_accuracy": 0.9337385654449463,
+      "num_tokens": 52363881.0,
+      "step": 6450
+    },
+    {
+      "entropy": 0.32826207876205443,
+      "epoch": 3.6427765237020315,
+      "grad_norm": 1.848895788192749,
+      "learning_rate": 4.701418432518032e-06,
+      "loss": 0.2206,
+      "mean_token_accuracy": 0.9300814032554626,
+      "num_tokens": 52404601.0,
+      "step": 6455
+    },
+    {
+      "entropy": 0.3011523485183716,
+      "epoch": 3.6455981941309257,
+      "grad_norm": 1.7186975479125977,
+      "learning_rate": 4.700968335758084e-06,
+      "loss": 0.1944,
+      "mean_token_accuracy": 0.9379884481430054,
+      "num_tokens": 52445278.0,
+      "step": 6460
+    },
+    {
+      "entropy": 0.3180157899856567,
+      "epoch": 3.6484198645598194,
+      "grad_norm": 2.0736236572265625,
+      "learning_rate": 4.700517929398951e-06,
+      "loss": 0.2125,
+      "mean_token_accuracy": 0.9314456939697265,
+      "num_tokens": 52486045.0,
+      "step": 6465
+    },
+    {
+      "entropy": 0.32199347019195557,
+      "epoch": 3.6512415349887135,
+      "grad_norm": 1.9433692693710327,
+      "learning_rate": 4.7000672135291166e-06,
+      "loss": 0.2133,
+      "mean_token_accuracy": 0.9318039417266846,
+      "num_tokens": 52525845.0,
+      "step": 6470
+    },
+    {
+      "entropy": 0.32936949729919435,
+      "epoch": 3.654063205417607,
+      "grad_norm": 1.8279980421066284,
+      "learning_rate": 4.699616188237123e-06,
+      "loss": 0.2376,
+      "mean_token_accuracy": 0.9232208609580994,
+      "num_tokens": 52566170.0,
+      "step": 6475
+    },
+    {
+      "entropy": 0.28730884194374084,
+      "epoch": 3.656884875846501,
+      "grad_norm": 1.848632574081421,
+      "learning_rate": 4.699164853611574e-06,
+      "loss": 0.1949,
+      "mean_token_accuracy": 0.9368890285491943,
+      "num_tokens": 52606933.0,
+      "step": 6480
+    },
+    {
+      "entropy": 0.2818956643342972,
+      "epoch": 3.659706546275395,
+      "grad_norm": 1.7565962076187134,
+      "learning_rate": 4.698713209741136e-06,
+      "loss": 0.1923,
+      "mean_token_accuracy": 0.9376680016517639,
+      "num_tokens": 52647547.0,
+      "step": 6485
+    },
+    {
+      "entropy": 0.30916563868522645,
+      "epoch": 3.662528216704289,
+      "grad_norm": 1.94307279586792,
+      "learning_rate": 4.698261256714533e-06,
+      "loss": 0.2076,
+      "mean_token_accuracy": 0.9336765766143799,
+      "num_tokens": 52688266.0,
+      "step": 6490
+    },
+    {
+      "entropy": 0.29969978928565977,
+      "epoch": 3.665349887133183,
+      "grad_norm": 1.9244630336761475,
+      "learning_rate": 4.697808994620552e-06,
+      "loss": 0.2085,
+      "mean_token_accuracy": 0.9322393655776977,
+      "num_tokens": 52728923.0,
+      "step": 6495
+    },
+    {
+      "entropy": 0.31762275099754333,
+      "epoch": 3.6681715575620766,
+      "grad_norm": 1.9523797035217285,
+      "learning_rate": 4.697356423548038e-06,
+      "loss": 0.2147,
+      "mean_token_accuracy": 0.9313531517982483,
+      "num_tokens": 52769425.0,
+      "step": 6500
+    },
+    {
+      "epoch": 3.6681715575620766,
+      "eval_entropy": 0.32103675603866577,
+      "eval_loss": 0.1689807027578354,
+      "eval_mean_token_accuracy": 0.9474894404411316,
+      "eval_num_tokens": 52769425.0,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 24.419,
+      "eval_steps_per_second": 6.105,
+      "step": 6500
+    },
+    {
+      "entropy": 0.3029796838760376,
+      "epoch": 3.6709932279909707,
+      "grad_norm": 1.7941051721572876,
+      "learning_rate": 4.696903543585902e-06,
+      "loss": 0.2128,
+      "mean_token_accuracy": 0.9323464274406433,
+      "num_tokens": 52810073.0,
+      "step": 6505
+    },
+    {
+      "entropy": 0.2974651575088501,
+      "epoch": 3.6738148984198644,
+      "grad_norm": 2.0666654109954834,
+      "learning_rate": 4.696450354823109e-06,
+      "loss": 0.2107,
+      "mean_token_accuracy": 0.9324890732765198,
+      "num_tokens": 52850886.0,
+      "step": 6510
+    },
+    {
+      "entropy": 0.3096803486347198,
+      "epoch": 3.6766365688487586,
+      "grad_norm": 1.9316201210021973,
+      "learning_rate": 4.69599685734869e-06,
+      "loss": 0.2055,
+      "mean_token_accuracy": 0.932162344455719,
+      "num_tokens": 52891548.0,
+      "step": 6515
+    },
+    {
+      "entropy": 0.307836776971817,
+      "epoch": 3.6794582392776523,
+      "grad_norm": 1.7384241819381714,
+      "learning_rate": 4.695543051251735e-06,
+      "loss": 0.2106,
+      "mean_token_accuracy": 0.930738651752472,
+      "num_tokens": 52932381.0,
+      "step": 6520
+    },
+    {
+      "entropy": 0.31080532670021055,
+      "epoch": 3.6822799097065464,
+      "grad_norm": 1.620614767074585,
+      "learning_rate": 4.695088936621393e-06,
+      "loss": 0.206,
+      "mean_token_accuracy": 0.9316131114959717,
+      "num_tokens": 52972931.0,
+      "step": 6525
+    },
+    {
+      "entropy": 0.3173711597919464,
+      "epoch": 3.68510158013544,
+      "grad_norm": 1.8724675178527832,
+      "learning_rate": 4.694634513546875e-06,
+      "loss": 0.2149,
+      "mean_token_accuracy": 0.9308315873146057,
+      "num_tokens": 53013578.0,
+      "step": 6530
+    },
+    {
+      "entropy": 0.29013689756393435,
+      "epoch": 3.6879232505643342,
+      "grad_norm": 1.7273049354553223,
+      "learning_rate": 4.6941797821174526e-06,
+      "loss": 0.1904,
+      "mean_token_accuracy": 0.9397516369819641,
+      "num_tokens": 53054348.0,
+      "step": 6535
+    },
+    {
+      "entropy": 0.30396153330802916,
+      "epoch": 3.690744920993228,
+      "grad_norm": 1.865898847579956,
+      "learning_rate": 4.693724742422458e-06,
+      "loss": 0.2176,
+      "mean_token_accuracy": 0.9278941035270691,
+      "num_tokens": 53094975.0,
+      "step": 6540
+    },
+    {
+      "entropy": 0.26694598495960237,
+      "epoch": 3.6935665914221216,
+      "grad_norm": 1.6677098274230957,
+      "learning_rate": 4.693269394551286e-06,
+      "loss": 0.1822,
+      "mean_token_accuracy": 0.9395427823066711,
+      "num_tokens": 53135477.0,
+      "step": 6545
+    },
+    {
+      "entropy": 0.3146726369857788,
+      "epoch": 3.6963882618510158,
+      "grad_norm": 1.8670984506607056,
+      "learning_rate": 4.6928137385933845e-06,
+      "loss": 0.2115,
+      "mean_token_accuracy": 0.9311745762825012,
+      "num_tokens": 53176141.0,
+      "step": 6550
+    },
+    {
+      "entropy": 0.32695006132125853,
+      "epoch": 3.69920993227991,
+      "grad_norm": 2.1717782020568848,
+      "learning_rate": 4.692357774638272e-06,
+      "loss": 0.2342,
+      "mean_token_accuracy": 0.9270057797431945,
+      "num_tokens": 53216706.0,
+      "step": 6555
+    },
+    {
+      "entropy": 0.3012891411781311,
+      "epoch": 3.7020316027088036,
+      "grad_norm": 1.886801838874817,
+      "learning_rate": 4.69190150277552e-06,
+      "loss": 0.2062,
+      "mean_token_accuracy": 0.9332509160041809,
+      "num_tokens": 53257505.0,
+      "step": 6560
+    },
+    {
+      "entropy": 0.316040962934494,
+      "epoch": 3.7048532731376973,
+      "grad_norm": 1.6204224824905396,
+      "learning_rate": 4.6914449230947645e-06,
+      "loss": 0.2109,
+      "mean_token_accuracy": 0.9321287989616394,
+      "num_tokens": 53298201.0,
+      "step": 6565
+    },
+    {
+      "entropy": 0.28626594245433806,
+      "epoch": 3.7076749435665914,
+      "grad_norm": 1.7525380849838257,
+      "learning_rate": 4.690988035685701e-06,
+      "loss": 0.1939,
+      "mean_token_accuracy": 0.9376110672950745,
+      "num_tokens": 53338694.0,
+      "step": 6570
+    },
+    {
+      "entropy": 0.29510667324066164,
+      "epoch": 3.7104966139954856,
+      "grad_norm": 1.6574718952178955,
+      "learning_rate": 4.690530840638083e-06,
+      "loss": 0.1897,
+      "mean_token_accuracy": 0.940097987651825,
+      "num_tokens": 53379266.0,
+      "step": 6575
+    },
+    {
+      "entropy": 0.3167356073856354,
+      "epoch": 3.7133182844243793,
+      "grad_norm": 2.295748472213745,
+      "learning_rate": 4.690073338041728e-06,
+      "loss": 0.2168,
+      "mean_token_accuracy": 0.9299868464469909,
+      "num_tokens": 53419886.0,
+      "step": 6580
+    },
+    {
+      "entropy": 0.27687845230102537,
+      "epoch": 3.716139954853273,
+      "grad_norm": 1.9063183069229126,
+      "learning_rate": 4.689615527986514e-06,
+      "loss": 0.1954,
+      "mean_token_accuracy": 0.9360902070999145,
+      "num_tokens": 53460645.0,
+      "step": 6585
+    },
+    {
+      "entropy": 0.27884105145931243,
+      "epoch": 3.718961625282167,
+      "grad_norm": 1.709240436553955,
+      "learning_rate": 4.689157410562374e-06,
+      "loss": 0.1907,
+      "mean_token_accuracy": 0.939925491809845,
+      "num_tokens": 53501408.0,
+      "step": 6590
+    },
+    {
+      "entropy": 0.3242746412754059,
+      "epoch": 3.721783295711061,
+      "grad_norm": 1.7791012525558472,
+      "learning_rate": 4.688698985859309e-06,
+      "loss": 0.2124,
+      "mean_token_accuracy": 0.9305900692939758,
+      "num_tokens": 53542237.0,
+      "step": 6595
+    },
+    {
+      "entropy": 0.3170479595661163,
+      "epoch": 3.724604966139955,
+      "grad_norm": 1.7075973749160767,
+      "learning_rate": 4.688240253967374e-06,
+      "loss": 0.2162,
+      "mean_token_accuracy": 0.9304636359214783,
+      "num_tokens": 53582388.0,
+      "step": 6600
+    },
+    {
+      "entropy": 0.3019756257534027,
+      "epoch": 3.7274266365688487,
+      "grad_norm": 1.7577334642410278,
+      "learning_rate": 4.6877812149766875e-06,
+      "loss": 0.1985,
+      "mean_token_accuracy": 0.9347709894180298,
+      "num_tokens": 53623102.0,
+      "step": 6605
+    },
+    {
+      "entropy": 0.33137462139129636,
+      "epoch": 3.730248306997743,
+      "grad_norm": 1.6792027950286865,
+      "learning_rate": 4.687321868977429e-06,
+      "loss": 0.241,
+      "mean_token_accuracy": 0.9231707096099854,
+      "num_tokens": 53663401.0,
+      "step": 6610
+    },
+    {
+      "entropy": 0.33162534832954405,
+      "epoch": 3.7330699774266365,
+      "grad_norm": 1.8961528539657593,
+      "learning_rate": 4.686862216059836e-06,
+      "loss": 0.2269,
+      "mean_token_accuracy": 0.9269180774688721,
+      "num_tokens": 53703912.0,
+      "step": 6615
+    },
+    {
+      "entropy": 0.32055285573005676,
+      "epoch": 3.7358916478555306,
+      "grad_norm": 1.9672921895980835,
+      "learning_rate": 4.686402256314208e-06,
+      "loss": 0.219,
+      "mean_token_accuracy": 0.9281070351600647,
+      "num_tokens": 53744733.0,
+      "step": 6620
+    },
+    {
+      "entropy": 0.32325056195259094,
+      "epoch": 3.7387133182844243,
+      "grad_norm": 2.0314273834228516,
+      "learning_rate": 4.685941989830903e-06,
+      "loss": 0.2246,
+      "mean_token_accuracy": 0.9259540200233459,
+      "num_tokens": 53784519.0,
+      "step": 6625
+    },
+    {
+      "entropy": 0.3410928726196289,
+      "epoch": 3.741534988713318,
+      "grad_norm": 1.9816386699676514,
+      "learning_rate": 4.685481416700342e-06,
+      "loss": 0.2428,
+      "mean_token_accuracy": 0.9229821920394897,
+      "num_tokens": 53825215.0,
+      "step": 6630
+    },
+    {
+      "entropy": 0.3254290819168091,
+      "epoch": 3.744356659142212,
+      "grad_norm": 2.0321855545043945,
+      "learning_rate": 4.685020537013004e-06,
+      "loss": 0.2097,
+      "mean_token_accuracy": 0.932454526424408,
+      "num_tokens": 53865574.0,
+      "step": 6635
+    },
+    {
+      "entropy": 0.29887913465499877,
+      "epoch": 3.7471783295711063,
+      "grad_norm": 1.6626564264297485,
+      "learning_rate": 4.684559350859428e-06,
+      "loss": 0.206,
+      "mean_token_accuracy": 0.9335589647293091,
+      "num_tokens": 53906263.0,
+      "step": 6640
+    },
+    {
+      "entropy": 0.32658823728561404,
+      "epoch": 3.75,
+      "grad_norm": 1.9820867776870728,
+      "learning_rate": 4.684097858330215e-06,
+      "loss": 0.2207,
+      "mean_token_accuracy": 0.9296117424964905,
+      "num_tokens": 53946706.0,
+      "step": 6645
+    },
+    {
+      "entropy": 0.29330855011940005,
+      "epoch": 3.7528216704288937,
+      "grad_norm": 1.978268027305603,
+      "learning_rate": 4.683636059516024e-06,
+      "loss": 0.1986,
+      "mean_token_accuracy": 0.9361773490905761,
+      "num_tokens": 53987483.0,
+      "step": 6650
+    },
+    {
+      "entropy": 0.3071614384651184,
+      "epoch": 3.755643340857788,
+      "grad_norm": 1.5843520164489746,
+      "learning_rate": 4.683173954507578e-06,
+      "loss": 0.2018,
+      "mean_token_accuracy": 0.935286819934845,
+      "num_tokens": 54028166.0,
+      "step": 6655
+    },
+    {
+      "entropy": 0.2947255611419678,
+      "epoch": 3.758465011286682,
+      "grad_norm": 1.8720123767852783,
+      "learning_rate": 4.682711543395656e-06,
+      "loss": 0.1897,
+      "mean_token_accuracy": 0.9393622994422912,
+      "num_tokens": 54068959.0,
+      "step": 6660
+    },
+    {
+      "entropy": 0.3333085238933563,
+      "epoch": 3.7612866817155757,
+      "grad_norm": 2.160306453704834,
+      "learning_rate": 4.6822488262710985e-06,
+      "loss": 0.2458,
+      "mean_token_accuracy": 0.9197412252426147,
+      "num_tokens": 54109806.0,
+      "step": 6665
+    },
+    {
+      "entropy": 0.3107566237449646,
+      "epoch": 3.7641083521444694,
+      "grad_norm": 1.6867446899414062,
+      "learning_rate": 4.681785803224807e-06,
+      "loss": 0.2145,
+      "mean_token_accuracy": 0.9298100113868714,
+      "num_tokens": 54150377.0,
+      "step": 6670
+    },
+    {
+      "entropy": 0.34289470911026,
+      "epoch": 3.7669300225733635,
+      "grad_norm": 2.239997625350952,
+      "learning_rate": 4.681322474347741e-06,
+      "loss": 0.243,
+      "mean_token_accuracy": 0.9235318303108215,
+      "num_tokens": 54191007.0,
+      "step": 6675
+    },
+    {
+      "entropy": 0.300843334197998,
+      "epoch": 3.769751693002257,
+      "grad_norm": 1.7303624153137207,
+      "learning_rate": 4.680858839730923e-06,
+      "loss": 0.2098,
+      "mean_token_accuracy": 0.9318808197975159,
+      "num_tokens": 54231680.0,
+      "step": 6680
+    },
+    {
+      "entropy": 0.3099523186683655,
+      "epoch": 3.7725733634311513,
+      "grad_norm": 1.8123502731323242,
+      "learning_rate": 4.680394899465435e-06,
+      "loss": 0.2032,
+      "mean_token_accuracy": 0.9328450918197632,
+      "num_tokens": 54272199.0,
+      "step": 6685
+    },
+    {
+      "entropy": 0.3188793003559113,
+      "epoch": 3.775395033860045,
+      "grad_norm": 1.985037922859192,
+      "learning_rate": 4.679930653642415e-06,
+      "loss": 0.2159,
+      "mean_token_accuracy": 0.9308086633682251,
+      "num_tokens": 54312840.0,
+      "step": 6690
+    },
+    {
+      "entropy": 0.29359256029129027,
+      "epoch": 3.778216704288939,
+      "grad_norm": 1.7202750444412231,
+      "learning_rate": 4.679466102353068e-06,
+      "loss": 0.2156,
+      "mean_token_accuracy": 0.9294952034950257,
+      "num_tokens": 54353596.0,
+      "step": 6695
+    },
+    {
+      "entropy": 0.32029845714569094,
+      "epoch": 3.781038374717833,
+      "grad_norm": 2.217895746231079,
+      "learning_rate": 4.679001245688651e-06,
+      "loss": 0.2197,
+      "mean_token_accuracy": 0.9296623706817627,
+      "num_tokens": 54393283.0,
+      "step": 6700
+    },
+    {
+      "entropy": 0.3079906702041626,
+      "epoch": 3.783860045146727,
+      "grad_norm": 1.7621604204177856,
+      "learning_rate": 4.678536083740488e-06,
+      "loss": 0.2123,
+      "mean_token_accuracy": 0.934044873714447,
+      "num_tokens": 54433802.0,
+      "step": 6705
+    },
+    {
+      "entropy": 0.3148071825504303,
+      "epoch": 3.7866817155756207,
+      "grad_norm": 1.961150884628296,
+      "learning_rate": 4.67807061659996e-06,
+      "loss": 0.2128,
+      "mean_token_accuracy": 0.9301833033561706,
+      "num_tokens": 54474550.0,
+      "step": 6710
+    },
+    {
+      "entropy": 0.31343921422958376,
+      "epoch": 3.7895033860045144,
+      "grad_norm": 2.0900840759277344,
+      "learning_rate": 4.677604844358507e-06,
+      "loss": 0.2165,
+      "mean_token_accuracy": 0.9303097009658814,
+      "num_tokens": 54515386.0,
+      "step": 6715
+    },
+    {
+      "entropy": 0.32203570008277893,
+      "epoch": 3.7923250564334086,
+      "grad_norm": 1.9342727661132812,
+      "learning_rate": 4.677138767107631e-06,
+      "loss": 0.2161,
+      "mean_token_accuracy": 0.9303983330726624,
+      "num_tokens": 54556126.0,
+      "step": 6720
+    },
+    {
+      "entropy": 0.28948697447776794,
+      "epoch": 3.7951467268623027,
+      "grad_norm": 1.6458464860916138,
+      "learning_rate": 4.676672384938891e-06,
+      "loss": 0.1951,
+      "mean_token_accuracy": 0.9385475873947143,
+      "num_tokens": 54597036.0,
+      "step": 6725
+    },
+    {
+      "entropy": 0.3142231285572052,
+      "epoch": 3.7979683972911964,
+      "grad_norm": 1.8955562114715576,
+      "learning_rate": 4.676205697943911e-06,
+      "loss": 0.2115,
+      "mean_token_accuracy": 0.9330004334449769,
+      "num_tokens": 54637646.0,
+      "step": 6730
+    },
+    {
+      "entropy": 0.3165326237678528,
+      "epoch": 3.80079006772009,
+      "grad_norm": 1.7128779888153076,
+      "learning_rate": 4.675738706214369e-06,
+      "loss": 0.2125,
+      "mean_token_accuracy": 0.9308281064033508,
+      "num_tokens": 54678230.0,
+      "step": 6735
+    },
+    {
+      "entropy": 0.3183960378170013,
+      "epoch": 3.8036117381489842,
+      "grad_norm": 2.020563840866089,
+      "learning_rate": 4.6752714098420065e-06,
+      "loss": 0.225,
+      "mean_token_accuracy": 0.9275296330451965,
+      "num_tokens": 54718922.0,
+      "step": 6740
+    },
+    {
+      "entropy": 0.3038501560688019,
+      "epoch": 3.8064334085778784,
+      "grad_norm": 2.007378339767456,
+      "learning_rate": 4.674803808918624e-06,
+      "loss": 0.1919,
+      "mean_token_accuracy": 0.9365889430046082,
+      "num_tokens": 54759550.0,
+      "step": 6745
+    },
+    {
+      "entropy": 0.3290569126605988,
+      "epoch": 3.809255079006772,
+      "grad_norm": 1.8866428136825562,
+      "learning_rate": 4.674335903536083e-06,
+      "loss": 0.2219,
+      "mean_token_accuracy": 0.9293844699859619,
+      "num_tokens": 54800015.0,
+      "step": 6750
+    },
+    {
+      "entropy": 0.3428149461746216,
+      "epoch": 3.8120767494356658,
+      "grad_norm": 2.020414113998413,
+      "learning_rate": 4.673867693786301e-06,
+      "loss": 0.2348,
+      "mean_token_accuracy": 0.924261474609375,
+      "num_tokens": 54840616.0,
+      "step": 6755
+    },
+    {
+      "entropy": 0.32770676612854005,
+      "epoch": 3.81489841986456,
+      "grad_norm": 1.6898592710494995,
+      "learning_rate": 4.6733991797612595e-06,
+      "loss": 0.2039,
+      "mean_token_accuracy": 0.9363934993743896,
+      "num_tokens": 54881081.0,
+      "step": 6760
+    },
+    {
+      "entropy": 0.3508101344108582,
+      "epoch": 3.8177200902934536,
+      "grad_norm": 1.967923879623413,
+      "learning_rate": 4.672930361552998e-06,
+      "loss": 0.2206,
+      "mean_token_accuracy": 0.9289634466171265,
+      "num_tokens": 54921258.0,
+      "step": 6765
+    },
+    {
+      "entropy": 0.29331698417663576,
+      "epoch": 3.8205417607223477,
+      "grad_norm": 1.8218579292297363,
+      "learning_rate": 4.672461239253616e-06,
+      "loss": 0.1829,
+      "mean_token_accuracy": 0.9402494430541992,
+      "num_tokens": 54961736.0,
+      "step": 6770
+    },
+    {
+      "entropy": 0.31439730525016785,
+      "epoch": 3.8233634311512414,
+      "grad_norm": 2.268512487411499,
+      "learning_rate": 4.671991812955273e-06,
+      "loss": 0.2126,
+      "mean_token_accuracy": 0.9317742347717285,
+      "num_tokens": 55002382.0,
+      "step": 6775
+    },
+    {
+      "entropy": 0.3293932378292084,
+      "epoch": 3.8261851015801356,
+      "grad_norm": 2.075479507446289,
+      "learning_rate": 4.671522082750186e-06,
+      "loss": 0.2336,
+      "mean_token_accuracy": 0.9250609993934631,
+      "num_tokens": 55043115.0,
+      "step": 6780
+    },
+    {
+      "entropy": 0.32756795883178713,
+      "epoch": 3.8290067720090293,
+      "grad_norm": 1.9896371364593506,
+      "learning_rate": 4.671052048730635e-06,
+      "loss": 0.2114,
+      "mean_token_accuracy": 0.9321682929992676,
+      "num_tokens": 55082894.0,
+      "step": 6785
+    },
+    {
+      "entropy": 0.3386588513851166,
+      "epoch": 3.8318284424379234,
+      "grad_norm": 1.902786135673523,
+      "learning_rate": 4.670581710988958e-06,
+      "loss": 0.2271,
+      "mean_token_accuracy": 0.9268991231918335,
+      "num_tokens": 55123408.0,
+      "step": 6790
+    },
+    {
+      "entropy": 0.3413122892379761,
+      "epoch": 3.834650112866817,
+      "grad_norm": 2.145611047744751,
+      "learning_rate": 4.6701110696175546e-06,
+      "loss": 0.2295,
+      "mean_token_accuracy": 0.9283112645149231,
+      "num_tokens": 55163941.0,
+      "step": 6795
+    },
+    {
+      "entropy": 0.293692809343338,
+      "epoch": 3.837471783295711,
+      "grad_norm": 2.1431288719177246,
+      "learning_rate": 4.669640124708879e-06,
+      "loss": 0.2009,
+      "mean_token_accuracy": 0.934659731388092,
+      "num_tokens": 55204732.0,
+      "step": 6800
+    },
+    {
+      "entropy": 0.322798889875412,
+      "epoch": 3.840293453724605,
+      "grad_norm": 1.9868714809417725,
+      "learning_rate": 4.66916887635545e-06,
+      "loss": 0.2327,
+      "mean_token_accuracy": 0.9276728153228759,
+      "num_tokens": 55245323.0,
+      "step": 6805
+    },
+    {
+      "entropy": 0.3417269468307495,
+      "epoch": 3.843115124153499,
+      "grad_norm": 1.8813223838806152,
+      "learning_rate": 4.668697324649845e-06,
+      "loss": 0.2308,
+      "mean_token_accuracy": 0.9265347599983216,
+      "num_tokens": 55285830.0,
+      "step": 6810
+    },
+    {
+      "entropy": 0.32689463496208193,
+      "epoch": 3.845936794582393,
+      "grad_norm": 1.8665236234664917,
+      "learning_rate": 4.6682254696847e-06,
+      "loss": 0.2186,
+      "mean_token_accuracy": 0.928902006149292,
+      "num_tokens": 55326576.0,
+      "step": 6815
+    },
+    {
+      "entropy": 0.30429264307022097,
+      "epoch": 3.8487584650112865,
+      "grad_norm": 1.9757881164550781,
+      "learning_rate": 4.667753311552711e-06,
+      "loss": 0.2066,
+      "mean_token_accuracy": 0.9331792950630188,
+      "num_tokens": 55367147.0,
+      "step": 6820
+    },
+    {
+      "entropy": 0.3437939524650574,
+      "epoch": 3.8515801354401806,
+      "grad_norm": 1.7982680797576904,
+      "learning_rate": 4.667280850346634e-06,
+      "loss": 0.236,
+      "mean_token_accuracy": 0.9242946028709411,
+      "num_tokens": 55407720.0,
+      "step": 6825
+    },
+    {
+      "entropy": 0.3199367105960846,
+      "epoch": 3.8544018058690743,
+      "grad_norm": 1.9674409627914429,
+      "learning_rate": 4.666808086159283e-06,
+      "loss": 0.2236,
+      "mean_token_accuracy": 0.9290154814720154,
+      "num_tokens": 55448550.0,
+      "step": 6830
+    },
+    {
+      "entropy": 0.2929447114467621,
+      "epoch": 3.8572234762979685,
+      "grad_norm": 2.2248289585113525,
+      "learning_rate": 4.666335019083532e-06,
+      "loss": 0.1942,
+      "mean_token_accuracy": 0.9364341259002685,
+      "num_tokens": 55489200.0,
+      "step": 6835
+    },
+    {
+      "entropy": 0.29515990018844607,
+      "epoch": 3.860045146726862,
+      "grad_norm": 1.5697543621063232,
+      "learning_rate": 4.665861649212316e-06,
+      "loss": 0.1987,
+      "mean_token_accuracy": 0.936658239364624,
+      "num_tokens": 55530008.0,
+      "step": 6840
+    },
+    {
+      "entropy": 0.3109833776950836,
+      "epoch": 3.8628668171557563,
+      "grad_norm": 1.9397079944610596,
+      "learning_rate": 4.6653879766386305e-06,
+      "loss": 0.2049,
+      "mean_token_accuracy": 0.9348199844360352,
+      "num_tokens": 55570437.0,
+      "step": 6845
+    },
+    {
+      "entropy": 0.3064566671848297,
+      "epoch": 3.86568848758465,
+      "grad_norm": 1.5913760662078857,
+      "learning_rate": 4.664914001455526e-06,
+      "loss": 0.2074,
+      "mean_token_accuracy": 0.9345534205436706,
+      "num_tokens": 55611132.0,
+      "step": 6850
+    },
+    {
+      "entropy": 0.31966778039932253,
+      "epoch": 3.868510158013544,
+      "grad_norm": 1.8581488132476807,
+      "learning_rate": 4.664439723756116e-06,
+      "loss": 0.2171,
+      "mean_token_accuracy": 0.9305494904518128,
+      "num_tokens": 55651714.0,
+      "step": 6855
+    },
+    {
+      "entropy": 0.31408268213272095,
+      "epoch": 3.871331828442438,
+      "grad_norm": 1.9941738843917847,
+      "learning_rate": 4.6639651436335705e-06,
+      "loss": 0.2101,
+      "mean_token_accuracy": 0.933535099029541,
+      "num_tokens": 55692277.0,
+      "step": 6860
+    },
+    {
+      "entropy": 0.3109330773353577,
+      "epoch": 3.874153498871332,
+      "grad_norm": 2.0422751903533936,
+      "learning_rate": 4.663490261181124e-06,
+      "loss": 0.206,
+      "mean_token_accuracy": 0.9325674891471862,
+      "num_tokens": 55732896.0,
+      "step": 6865
+    },
+    {
+      "entropy": 0.29195868968963623,
+      "epoch": 3.8769751693002257,
+      "grad_norm": 1.6397924423217773,
+      "learning_rate": 4.663015076492065e-06,
+      "loss": 0.2098,
+      "mean_token_accuracy": 0.9333557486534119,
+      "num_tokens": 55773560.0,
+      "step": 6870
+    },
+    {
+      "entropy": 0.3036874562501907,
+      "epoch": 3.87979683972912,
+      "grad_norm": 2.033968448638916,
+      "learning_rate": 4.662539589659746e-06,
+      "loss": 0.2065,
+      "mean_token_accuracy": 0.9338034749031067,
+      "num_tokens": 55814207.0,
+      "step": 6875
+    },
+    {
+      "entropy": 0.33089557886123655,
+      "epoch": 3.8826185101580135,
+      "grad_norm": 1.942462682723999,
+      "learning_rate": 4.6620638007775735e-06,
+      "loss": 0.223,
+      "mean_token_accuracy": 0.9286533951759338,
+      "num_tokens": 55854785.0,
+      "step": 6880
+    },
+    {
+      "entropy": 0.3029870331287384,
+      "epoch": 3.885440180586907,
+      "grad_norm": 1.7418941259384155,
+      "learning_rate": 4.661587709939017e-06,
+      "loss": 0.2038,
+      "mean_token_accuracy": 0.9331433176994324,
+      "num_tokens": 55895399.0,
+      "step": 6885
+    },
+    {
+      "entropy": 0.3070501685142517,
+      "epoch": 3.8882618510158014,
+      "grad_norm": 1.9692811965942383,
+      "learning_rate": 4.661111317237606e-06,
+      "loss": 0.1969,
+      "mean_token_accuracy": 0.9379673600196838,
+      "num_tokens": 55935772.0,
+      "step": 6890
+    },
+    {
+      "entropy": 0.3398116111755371,
+      "epoch": 3.8910835214446955,
+      "grad_norm": 1.9820867776870728,
+      "learning_rate": 4.660634622766926e-06,
+      "loss": 0.2501,
+      "mean_token_accuracy": 0.9237083911895752,
+      "num_tokens": 55976410.0,
+      "step": 6895
+    },
+    {
+      "entropy": 0.32836961150169375,
+      "epoch": 3.893905191873589,
+      "grad_norm": 2.400958299636841,
+      "learning_rate": 4.660157626620625e-06,
+      "loss": 0.2248,
+      "mean_token_accuracy": 0.9286513090133667,
+      "num_tokens": 56016725.0,
+      "step": 6900
+    },
+    {
+      "entropy": 0.31524630188941954,
+      "epoch": 3.896726862302483,
+      "grad_norm": 1.8288160562515259,
+      "learning_rate": 4.65968032889241e-06,
+      "loss": 0.1952,
+      "mean_token_accuracy": 0.9361767411231995,
+      "num_tokens": 56057415.0,
+      "step": 6905
+    },
+    {
+      "entropy": 0.31340600848197936,
+      "epoch": 3.899548532731377,
+      "grad_norm": 1.9468982219696045,
+      "learning_rate": 4.6592027296760435e-06,
+      "loss": 0.2189,
+      "mean_token_accuracy": 0.9306774020195008,
+      "num_tokens": 56098039.0,
+      "step": 6910
+    },
+    {
+      "entropy": 0.2994066894054413,
+      "epoch": 3.9023702031602707,
+      "grad_norm": 1.8991196155548096,
+      "learning_rate": 4.658724829065352e-06,
+      "loss": 0.2006,
+      "mean_token_accuracy": 0.9361102461814881,
+      "num_tokens": 56138690.0,
+      "step": 6915
+    },
+    {
+      "entropy": 0.3208695352077484,
+      "epoch": 3.905191873589165,
+      "grad_norm": 2.3649332523345947,
+      "learning_rate": 4.658246627154219e-06,
+      "loss": 0.2117,
+      "mean_token_accuracy": 0.9312704563140869,
+      "num_tokens": 56179455.0,
+      "step": 6920
+    },
+    {
+      "entropy": 0.2794735461473465,
+      "epoch": 3.9080135440180586,
+      "grad_norm": 2.0509235858917236,
+      "learning_rate": 4.6577681240365856e-06,
+      "loss": 0.1958,
+      "mean_token_accuracy": 0.9353933930397034,
+      "num_tokens": 56219807.0,
+      "step": 6925
+    },
+    {
+      "entropy": 0.32446990013122556,
+      "epoch": 3.9108352144469527,
+      "grad_norm": 1.699033498764038,
+      "learning_rate": 4.657289319806456e-06,
+      "loss": 0.2198,
+      "mean_token_accuracy": 0.929268729686737,
+      "num_tokens": 56260537.0,
+      "step": 6930
+    },
+    {
+      "entropy": 0.29834595918655393,
+      "epoch": 3.9136568848758464,
+      "grad_norm": 1.6812763214111328,
+      "learning_rate": 4.656810214557889e-06,
+      "loss": 0.1882,
+      "mean_token_accuracy": 0.937093997001648,
+      "num_tokens": 56301037.0,
+      "step": 6935
+    },
+    {
+      "entropy": 0.2948793530464172,
+      "epoch": 3.9164785553047405,
+      "grad_norm": 1.752263069152832,
+      "learning_rate": 4.6563308083850075e-06,
+      "loss": 0.1935,
+      "mean_token_accuracy": 0.9361865758895874,
+      "num_tokens": 56341578.0,
+      "step": 6940
+    },
+    {
+      "entropy": 0.3016524910926819,
+      "epoch": 3.9193002257336342,
+      "grad_norm": 1.9265809059143066,
+      "learning_rate": 4.655851101381988e-06,
+      "loss": 0.1969,
+      "mean_token_accuracy": 0.9359880805015564,
+      "num_tokens": 56382334.0,
+      "step": 6945
+    },
+    {
+      "entropy": 0.31857663989067075,
+      "epoch": 3.9221218961625284,
+      "grad_norm": 2.062640905380249,
+      "learning_rate": 4.655371093643073e-06,
+      "loss": 0.2319,
+      "mean_token_accuracy": 0.9254809617996216,
+      "num_tokens": 56422744.0,
+      "step": 6950
+    },
+    {
+      "entropy": 0.3018131792545319,
+      "epoch": 3.924943566591422,
+      "grad_norm": 1.7309370040893555,
+      "learning_rate": 4.6548907852625565e-06,
+      "loss": 0.2042,
+      "mean_token_accuracy": 0.9332278847694397,
+      "num_tokens": 56463597.0,
+      "step": 6955
+    },
+    {
+      "entropy": 0.31196231842041017,
+      "epoch": 3.927765237020316,
+      "grad_norm": 1.6862263679504395,
+      "learning_rate": 4.654410176334796e-06,
+      "loss": 0.2161,
+      "mean_token_accuracy": 0.9320029973983764,
+      "num_tokens": 56504227.0,
+      "step": 6960
+    },
+    {
+      "entropy": 0.3238642394542694,
+      "epoch": 3.93058690744921,
+      "grad_norm": 1.998358130455017,
+      "learning_rate": 4.653929266954208e-06,
+      "loss": 0.2084,
+      "mean_token_accuracy": 0.9316397666931152,
+      "num_tokens": 56544904.0,
+      "step": 6965
+    },
+    {
+      "entropy": 0.34933393597602846,
+      "epoch": 3.9334085778781036,
+      "grad_norm": 2.1447250843048096,
+      "learning_rate": 4.653448057215267e-06,
+      "loss": 0.2264,
+      "mean_token_accuracy": 0.9266282439231872,
+      "num_tokens": 56585427.0,
+      "step": 6970
+    },
+    {
+      "entropy": 0.34060198068618774,
+      "epoch": 3.9362302483069977,
+      "grad_norm": 2.133936882019043,
+      "learning_rate": 4.652966547212506e-06,
+      "loss": 0.2445,
+      "mean_token_accuracy": 0.9236889362335206,
+      "num_tokens": 56626066.0,
+      "step": 6975
+    },
+    {
+      "entropy": 0.3082031667232513,
+      "epoch": 3.939051918735892,
+      "grad_norm": 1.9675047397613525,
+      "learning_rate": 4.652484737040518e-06,
+      "loss": 0.2007,
+      "mean_token_accuracy": 0.9365509033203125,
+      "num_tokens": 56666827.0,
+      "step": 6980
+    },
+    {
+      "entropy": 0.3459632158279419,
+      "epoch": 3.9418735891647856,
+      "grad_norm": 1.927329421043396,
+      "learning_rate": 4.652002626793956e-06,
+      "loss": 0.2278,
+      "mean_token_accuracy": 0.9272386431694031,
+      "num_tokens": 56707612.0,
+      "step": 6985
+    },
+    {
+      "entropy": 0.3056627333164215,
+      "epoch": 3.9446952595936793,
+      "grad_norm": 1.6812715530395508,
+      "learning_rate": 4.651520216567528e-06,
+      "loss": 0.1971,
+      "mean_token_accuracy": 0.9381381869316101,
+      "num_tokens": 56748245.0,
+      "step": 6990
+    },
+    {
+      "entropy": 0.2965980887413025,
+      "epoch": 3.9475169300225734,
+      "grad_norm": 1.688714861869812,
+      "learning_rate": 4.651037506456006e-06,
+      "loss": 0.2001,
+      "mean_token_accuracy": 0.9357024908065796,
+      "num_tokens": 56788909.0,
+      "step": 6995
+    },
+    {
+      "entropy": 0.32579890489578245,
+      "epoch": 3.950338600451467,
+      "grad_norm": 2.0904133319854736,
+      "learning_rate": 4.650554496554217e-06,
+      "loss": 0.2383,
+      "mean_token_accuracy": 0.9262525558471679,
+      "num_tokens": 56829431.0,
+      "step": 7000
+    },
+    {
+      "epoch": 3.950338600451467,
+      "eval_entropy": 0.3211408257484436,
+      "eval_loss": 0.14595237374305725,
+      "eval_mean_token_accuracy": 0.9543886780738831,
+      "eval_num_tokens": 56829431.0,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 24.384,
+      "eval_steps_per_second": 6.096,
+      "step": 7000
+    },
+    {
+      "entropy": 0.32015385031700133,
+      "epoch": 3.9531602708803613,
+      "grad_norm": 1.8100392818450928,
+      "learning_rate": 4.650071186957049e-06,
+      "loss": 0.2236,
+      "mean_token_accuracy": 0.9274610161781311,
+      "num_tokens": 56870084.0,
+      "step": 7005
+    },
+    {
+      "entropy": 0.31498663425445556,
+      "epoch": 3.955981941309255,
+      "grad_norm": 1.7717326879501343,
+      "learning_rate": 4.6495875777594485e-06,
+      "loss": 0.2223,
+      "mean_token_accuracy": 0.927357268333435,
+      "num_tokens": 56910874.0,
+      "step": 7010
+    },
+    {
+      "entropy": 0.2908495903015137,
+      "epoch": 3.958803611738149,
+      "grad_norm": 2.0041091442108154,
+      "learning_rate": 4.64910366905642e-06,
+      "loss": 0.1944,
+      "mean_token_accuracy": 0.9367148995399475,
+      "num_tokens": 56951468.0,
+      "step": 7015
+    },
+    {
+      "entropy": 0.3460938811302185,
+      "epoch": 3.961625282167043,
+      "grad_norm": 2.049760580062866,
+      "learning_rate": 4.648619460943027e-06,
+      "loss": 0.2347,
+      "mean_token_accuracy": 0.9227741837501526,
+      "num_tokens": 56992209.0,
+      "step": 7020
+    },
+    {
+      "entropy": 0.31545761227607727,
+      "epoch": 3.964446952595937,
+      "grad_norm": 2.2006888389587402,
+      "learning_rate": 4.6481349535143934e-06,
+      "loss": 0.21,
+      "mean_token_accuracy": 0.9324324250221252,
+      "num_tokens": 57033074.0,
+      "step": 7025
+    },
+    {
+      "entropy": 0.2913439154624939,
+      "epoch": 3.9672686230248306,
+      "grad_norm": 1.7497133016586304,
+      "learning_rate": 4.6476501468657e-06,
+      "loss": 0.1987,
+      "mean_token_accuracy": 0.934621024131775,
+      "num_tokens": 57073763.0,
+      "step": 7030
+    },
+    {
+      "entropy": 0.3054608583450317,
+      "epoch": 3.9700902934537243,
+      "grad_norm": 1.3627029657363892,
+      "learning_rate": 4.647165041092187e-06,
+      "loss": 0.1965,
+      "mean_token_accuracy": 0.9352095484733581,
+      "num_tokens": 57114140.0,
+      "step": 7035
+    },
+    {
+      "entropy": 0.31089224219322203,
+      "epoch": 3.9729119638826185,
+      "grad_norm": 1.926986575126648,
+      "learning_rate": 4.646679636289154e-06,
+      "loss": 0.2129,
+      "mean_token_accuracy": 0.9315391659736634,
+      "num_tokens": 57154688.0,
+      "step": 7040
+    },
+    {
+      "entropy": 0.29714728593826295,
+      "epoch": 3.9757336343115126,
+      "grad_norm": 2.036738872528076,
+      "learning_rate": 4.646193932551959e-06,
+      "loss": 0.2094,
+      "mean_token_accuracy": 0.9334526896476746,
+      "num_tokens": 57195301.0,
+      "step": 7045
+    },
+    {
+      "entropy": 0.3167958378791809,
+      "epoch": 3.9785553047404063,
+      "grad_norm": 2.016421318054199,
+      "learning_rate": 4.645707929976018e-06,
+      "loss": 0.2166,
+      "mean_token_accuracy": 0.9300056099891663,
+      "num_tokens": 57235979.0,
+      "step": 7050
+    },
+    {
+      "entropy": 0.2984152317047119,
+      "epoch": 3.9813769751693,
+      "grad_norm": 2.002079486846924,
+      "learning_rate": 4.645221628656806e-06,
+      "loss": 0.1954,
+      "mean_token_accuracy": 0.9366149425506591,
+      "num_tokens": 57276713.0,
+      "step": 7055
+    },
+    {
+      "entropy": 0.32194399237632754,
+      "epoch": 3.984198645598194,
+      "grad_norm": 1.7218655347824097,
+      "learning_rate": 4.644735028689858e-06,
+      "loss": 0.2147,
+      "mean_token_accuracy": 0.9305589437484741,
+      "num_tokens": 57317326.0,
+      "step": 7060
+    },
+    {
+      "entropy": 0.317461222410202,
+      "epoch": 3.9870203160270883,
+      "grad_norm": 2.030597686767578,
+      "learning_rate": 4.644248130170766e-06,
+      "loss": 0.225,
+      "mean_token_accuracy": 0.927132534980774,
+      "num_tokens": 57358188.0,
+      "step": 7065
+    },
+    {
+      "entropy": 0.30844367146492,
+      "epoch": 3.989841986455982,
+      "grad_norm": 1.770613193511963,
+      "learning_rate": 4.643760933195182e-06,
+      "loss": 0.2062,
+      "mean_token_accuracy": 0.9343749284744263,
+      "num_tokens": 57398939.0,
+      "step": 7070
+    },
+    {
+      "entropy": 0.2940232276916504,
+      "epoch": 3.9926636568848757,
+      "grad_norm": 1.9694218635559082,
+      "learning_rate": 4.643273437858814e-06,
+      "loss": 0.1933,
+      "mean_token_accuracy": 0.9365314722061158,
+      "num_tokens": 57439332.0,
+      "step": 7075
+    },
+    {
+      "entropy": 0.30736419558525085,
+      "epoch": 3.99548532731377,
+      "grad_norm": 1.9869581460952759,
+      "learning_rate": 4.642785644257432e-06,
+      "loss": 0.2046,
+      "mean_token_accuracy": 0.9342207312583923,
+      "num_tokens": 57479943.0,
+      "step": 7080
+    },
+    {
+      "entropy": 0.3101332366466522,
+      "epoch": 3.9983069977426635,
+      "grad_norm": 1.9268759489059448,
+      "learning_rate": 4.6422975524868635e-06,
+      "loss": 0.2151,
+      "mean_token_accuracy": 0.9306729078292847,
+      "num_tokens": 57520661.0,
+      "step": 7085
+    },
+    {
+      "entropy": 0.3069190442562103,
+      "epoch": 4.001128668171558,
+      "grad_norm": 1.7287434339523315,
+      "learning_rate": 4.641809162642993e-06,
+      "loss": 0.1752,
+      "mean_token_accuracy": 0.9477459907531738,
+      "num_tokens": 57555350.0,
+      "step": 7090
+    },
+    {
+      "entropy": 0.2743695855140686,
+      "epoch": 4.003950338600451,
+      "grad_norm": 1.4140337705612183,
+      "learning_rate": 4.641320474821765e-06,
+      "loss": 0.128,
+      "mean_token_accuracy": 0.9617153882980347,
+      "num_tokens": 57596053.0,
+      "step": 7095
+    },
+    {
+      "entropy": 0.2616827428340912,
+      "epoch": 4.006772009029345,
+      "grad_norm": 2.273259162902832,
+      "learning_rate": 4.640831489119184e-06,
+      "loss": 0.1496,
+      "mean_token_accuracy": 0.9549853324890136,
+      "num_tokens": 57636820.0,
+      "step": 7100
+    },
+    {
+      "entropy": 0.21969364285469056,
+      "epoch": 4.00959367945824,
+      "grad_norm": 1.9657589197158813,
+      "learning_rate": 4.640342205631309e-06,
+      "loss": 0.1234,
+      "mean_token_accuracy": 0.961889123916626,
+      "num_tokens": 57677102.0,
+      "step": 7105
+    },
+    {
+      "entropy": 0.2414553850889206,
+      "epoch": 4.012415349887133,
+      "grad_norm": 2.4210760593414307,
+      "learning_rate": 4.639852624454261e-06,
+      "loss": 0.1491,
+      "mean_token_accuracy": 0.9532399654388428,
+      "num_tokens": 57717694.0,
+      "step": 7110
+    },
+    {
+      "entropy": 0.22773428559303283,
+      "epoch": 4.015237020316027,
+      "grad_norm": 2.0775444507598877,
+      "learning_rate": 4.639362745684219e-06,
+      "loss": 0.1188,
+      "mean_token_accuracy": 0.9613750100135803,
+      "num_tokens": 57758420.0,
+      "step": 7115
+    },
+    {
+      "entropy": 0.24161579608917236,
+      "epoch": 4.018058690744921,
+      "grad_norm": 1.9448866844177246,
+      "learning_rate": 4.638872569417417e-06,
+      "loss": 0.1306,
+      "mean_token_accuracy": 0.9591588497161865,
+      "num_tokens": 57799103.0,
+      "step": 7120
+    },
+    {
+      "entropy": 0.2467570424079895,
+      "epoch": 4.020880361173815,
+      "grad_norm": 1.8447470664978027,
+      "learning_rate": 4.638382095750152e-06,
+      "loss": 0.1285,
+      "mean_token_accuracy": 0.9602084875106811,
+      "num_tokens": 57839297.0,
+      "step": 7125
+    },
+    {
+      "entropy": 0.24987946152687074,
+      "epoch": 4.023702031602709,
+      "grad_norm": 2.2855517864227295,
+      "learning_rate": 4.6378913247787786e-06,
+      "loss": 0.1235,
+      "mean_token_accuracy": 0.9607668399810791,
+      "num_tokens": 57879861.0,
+      "step": 7130
+    },
+    {
+      "entropy": 0.23636153638362883,
+      "epoch": 4.026523702031603,
+      "grad_norm": 1.8914059400558472,
+      "learning_rate": 4.637400256599707e-06,
+      "loss": 0.133,
+      "mean_token_accuracy": 0.9598736882209777,
+      "num_tokens": 57920707.0,
+      "step": 7135
+    },
+    {
+      "entropy": 0.2302469491958618,
+      "epoch": 4.029345372460496,
+      "grad_norm": 2.0456087589263916,
+      "learning_rate": 4.636908891309408e-06,
+      "loss": 0.1263,
+      "mean_token_accuracy": 0.9601135969161987,
+      "num_tokens": 57961297.0,
+      "step": 7140
+    },
+    {
+      "entropy": 0.23698365390300752,
+      "epoch": 4.03216704288939,
+      "grad_norm": 1.853820562362671,
+      "learning_rate": 4.636417229004412e-06,
+      "loss": 0.1325,
+      "mean_token_accuracy": 0.9583729267120361,
+      "num_tokens": 58001984.0,
+      "step": 7145
+    },
+    {
+      "entropy": 0.22845596373081206,
+      "epoch": 4.034988713318285,
+      "grad_norm": 1.8284083604812622,
+      "learning_rate": 4.635925269781305e-06,
+      "loss": 0.1264,
+      "mean_token_accuracy": 0.9601821541786194,
+      "num_tokens": 58042688.0,
+      "step": 7150
+    },
+    {
+      "entropy": 0.21759492456912993,
+      "epoch": 4.037810383747178,
+      "grad_norm": 2.047950267791748,
+      "learning_rate": 4.6354330137367305e-06,
+      "loss": 0.1157,
+      "mean_token_accuracy": 0.963828158378601,
+      "num_tokens": 58083388.0,
+      "step": 7155
+    },
+    {
+      "entropy": 0.22680849432945252,
+      "epoch": 4.040632054176072,
+      "grad_norm": 1.9068692922592163,
+      "learning_rate": 4.634940460967396e-06,
+      "loss": 0.1217,
+      "mean_token_accuracy": 0.9618602633476258,
+      "num_tokens": 58124041.0,
+      "step": 7160
+    },
+    {
+      "entropy": 0.22470956444740295,
+      "epoch": 4.043453724604966,
+      "grad_norm": 1.6333588361740112,
+      "learning_rate": 4.634447611570061e-06,
+      "loss": 0.1182,
+      "mean_token_accuracy": 0.9632346987724304,
+      "num_tokens": 58164726.0,
+      "step": 7165
+    },
+    {
+      "entropy": 0.2346551775932312,
+      "epoch": 4.04627539503386,
+      "grad_norm": 1.8143978118896484,
+      "learning_rate": 4.633954465641546e-06,
+      "loss": 0.1325,
+      "mean_token_accuracy": 0.9576727986335755,
+      "num_tokens": 58205286.0,
+      "step": 7170
+    },
+    {
+      "entropy": 0.23285722136497497,
+      "epoch": 4.049097065462754,
+      "grad_norm": 1.8555634021759033,
+      "learning_rate": 4.633461023278731e-06,
+      "loss": 0.1157,
+      "mean_token_accuracy": 0.9634591937065125,
+      "num_tokens": 58246019.0,
+      "step": 7175
+    },
+    {
+      "entropy": 0.23194559514522553,
+      "epoch": 4.051918735891648,
+      "grad_norm": 1.8216824531555176,
+      "learning_rate": 4.632967284578551e-06,
+      "loss": 0.1168,
+      "mean_token_accuracy": 0.9622029185295105,
+      "num_tokens": 58286809.0,
+      "step": 7180
+    },
+    {
+      "entropy": 0.25160720348358157,
+      "epoch": 4.0547404063205414,
+      "grad_norm": 1.9061291217803955,
+      "learning_rate": 4.632473249638003e-06,
+      "loss": 0.1255,
+      "mean_token_accuracy": 0.9632068753242493,
+      "num_tokens": 58327226.0,
+      "step": 7185
+    },
+    {
+      "entropy": 0.22852468192577363,
+      "epoch": 4.057562076749436,
+      "grad_norm": 1.9731613397598267,
+      "learning_rate": 4.631978918554139e-06,
+      "loss": 0.1235,
+      "mean_token_accuracy": 0.9616599082946777,
+      "num_tokens": 58367956.0,
+      "step": 7190
+    },
+    {
+      "entropy": 0.2459454208612442,
+      "epoch": 4.06038374717833,
+      "grad_norm": 1.9121125936508179,
+      "learning_rate": 4.631484291424069e-06,
+      "loss": 0.1461,
+      "mean_token_accuracy": 0.9540547609329224,
+      "num_tokens": 58408674.0,
+      "step": 7195
+    },
+    {
+      "entropy": 0.24432432651519775,
+      "epoch": 4.063205417607223,
+      "grad_norm": 1.7734456062316895,
+      "learning_rate": 4.630989368344966e-06,
+      "loss": 0.1334,
+      "mean_token_accuracy": 0.9592110633850097,
+      "num_tokens": 58448721.0,
+      "step": 7200
+    },
+    {
+      "entropy": 0.24515534937381744,
+      "epoch": 4.066027088036117,
+      "grad_norm": 2.12031888961792,
+      "learning_rate": 4.630494149414054e-06,
+      "loss": 0.1274,
+      "mean_token_accuracy": 0.9601586699485779,
+      "num_tokens": 58489397.0,
+      "step": 7205
+    },
+    {
+      "entropy": 0.23616698682308196,
+      "epoch": 4.068848758465012,
+      "grad_norm": 1.635472059249878,
+      "learning_rate": 4.629998634728622e-06,
+      "loss": 0.1233,
+      "mean_token_accuracy": 0.9608691930770874,
+      "num_tokens": 58529908.0,
+      "step": 7210
+    },
+    {
+      "entropy": 0.24083788096904754,
+      "epoch": 4.071670428893905,
+      "grad_norm": 1.929811716079712,
+      "learning_rate": 4.629502824386013e-06,
+      "loss": 0.1248,
+      "mean_token_accuracy": 0.9608209013938904,
+      "num_tokens": 58570278.0,
+      "step": 7215
+    },
+    {
+      "entropy": 0.24579550325870514,
+      "epoch": 4.074492099322799,
+      "grad_norm": 2.148528814315796,
+      "learning_rate": 4.629006718483627e-06,
+      "loss": 0.1347,
+      "mean_token_accuracy": 0.9587805867195129,
+      "num_tokens": 58611040.0,
+      "step": 7220
+    },
+    {
+      "entropy": 0.23470092117786406,
+      "epoch": 4.077313769751693,
+      "grad_norm": 1.7747644186019897,
+      "learning_rate": 4.628510317118927e-06,
+      "loss": 0.1234,
+      "mean_token_accuracy": 0.961380934715271,
+      "num_tokens": 58651615.0,
+      "step": 7225
+    },
+    {
+      "entropy": 0.21981629133224487,
+      "epoch": 4.0801354401805865,
+      "grad_norm": 2.0685737133026123,
+      "learning_rate": 4.628013620389429e-06,
+      "loss": 0.1237,
+      "mean_token_accuracy": 0.960912036895752,
+      "num_tokens": 58692473.0,
+      "step": 7230
+    },
+    {
+      "entropy": 0.22954968512058258,
+      "epoch": 4.082957110609481,
+      "grad_norm": 2.0184037685394287,
+      "learning_rate": 4.62751662839271e-06,
+      "loss": 0.1328,
+      "mean_token_accuracy": 0.9574317932128906,
+      "num_tokens": 58733055.0,
+      "step": 7235
+    },
+    {
+      "entropy": 0.2246454894542694,
+      "epoch": 4.085778781038375,
+      "grad_norm": 1.9148306846618652,
+      "learning_rate": 4.627019341226404e-06,
+      "loss": 0.1156,
+      "mean_token_accuracy": 0.9631801724433899,
+      "num_tokens": 58773853.0,
+      "step": 7240
+    },
+    {
+      "entropy": 0.23507523834705352,
+      "epoch": 4.0886004514672685,
+      "grad_norm": 1.794526219367981,
+      "learning_rate": 4.626521758988204e-06,
+      "loss": 0.1167,
+      "mean_token_accuracy": 0.9632613658905029,
+      "num_tokens": 58814514.0,
+      "step": 7245
+    },
+    {
+      "entropy": 0.23969656229019165,
+      "epoch": 4.091422121896162,
+      "grad_norm": 1.9400169849395752,
+      "learning_rate": 4.626023881775858e-06,
+      "loss": 0.1284,
+      "mean_token_accuracy": 0.9598143339157105,
+      "num_tokens": 58854908.0,
+      "step": 7250
+    },
+    {
+      "entropy": 0.2598514884710312,
+      "epoch": 4.094243792325057,
+      "grad_norm": 2.0811522006988525,
+      "learning_rate": 4.625525709687176e-06,
+      "loss": 0.1422,
+      "mean_token_accuracy": 0.9565314888954163,
+      "num_tokens": 58895394.0,
+      "step": 7255
+    },
+    {
+      "entropy": 0.24920229315757753,
+      "epoch": 4.0970654627539504,
+      "grad_norm": 1.6480082273483276,
+      "learning_rate": 4.625027242820023e-06,
+      "loss": 0.1286,
+      "mean_token_accuracy": 0.9599669218063355,
+      "num_tokens": 58935939.0,
+      "step": 7260
+    },
+    {
+      "entropy": 0.23545166254043579,
+      "epoch": 4.099887133182844,
+      "grad_norm": 2.00404691696167,
+      "learning_rate": 4.6245284812723234e-06,
+      "loss": 0.1257,
+      "mean_token_accuracy": 0.9604501008987427,
+      "num_tokens": 58976341.0,
+      "step": 7265
+    },
+    {
+      "entropy": 0.23354564011096954,
+      "epoch": 4.102708803611738,
+      "grad_norm": 2.3596303462982178,
+      "learning_rate": 4.624029425142059e-06,
+      "loss": 0.1312,
+      "mean_token_accuracy": 0.9599619388580323,
+      "num_tokens": 59016822.0,
+      "step": 7270
+    },
+    {
+      "entropy": 0.24546316564083098,
+      "epoch": 4.105530474040632,
+      "grad_norm": 1.6743675470352173,
+      "learning_rate": 4.623530074527269e-06,
+      "loss": 0.1286,
+      "mean_token_accuracy": 0.9613336443901062,
+      "num_tokens": 59057357.0,
+      "step": 7275
+    },
+    {
+      "entropy": 0.23465977013111114,
+      "epoch": 4.108352144469526,
+      "grad_norm": 1.9552369117736816,
+      "learning_rate": 4.6230304295260504e-06,
+      "loss": 0.1138,
+      "mean_token_accuracy": 0.963888657093048,
+      "num_tokens": 59098128.0,
+      "step": 7280
+    },
+    {
+      "entropy": 0.22476655542850493,
+      "epoch": 4.11117381489842,
+      "grad_norm": 1.9228246212005615,
+      "learning_rate": 4.62253049023656e-06,
+      "loss": 0.132,
+      "mean_token_accuracy": 0.9577814340591431,
+      "num_tokens": 59138682.0,
+      "step": 7285
+    },
+    {
+      "entropy": 0.23238305151462554,
+      "epoch": 4.1139954853273135,
+      "grad_norm": 1.8508713245391846,
+      "learning_rate": 4.62203025675701e-06,
+      "loss": 0.1336,
+      "mean_token_accuracy": 0.9592737793922425,
+      "num_tokens": 59179330.0,
+      "step": 7290
+    },
+    {
+      "entropy": 0.21901310980319977,
+      "epoch": 4.116817155756208,
+      "grad_norm": 1.9854564666748047,
+      "learning_rate": 4.621529729185671e-06,
+      "loss": 0.1146,
+      "mean_token_accuracy": 0.9635023355484009,
+      "num_tokens": 59220102.0,
+      "step": 7295
+    },
+    {
+      "entropy": 0.23093846142292024,
+      "epoch": 4.119638826185102,
+      "grad_norm": 1.7176353931427002,
+      "learning_rate": 4.621028907620873e-06,
+      "loss": 0.129,
+      "mean_token_accuracy": 0.9593886017799378,
+      "num_tokens": 59260732.0,
+      "step": 7300
+    },
+    {
+      "entropy": 0.2379404127597809,
+      "epoch": 4.1224604966139955,
+      "grad_norm": 1.8273789882659912,
+      "learning_rate": 4.620527792161001e-06,
+      "loss": 0.1124,
+      "mean_token_accuracy": 0.9650070428848266,
+      "num_tokens": 59301380.0,
+      "step": 7305
+    },
+    {
+      "entropy": 0.22208026945590972,
+      "epoch": 4.125282167042889,
+      "grad_norm": 1.7196441888809204,
+      "learning_rate": 4.620026382904499e-06,
+      "loss": 0.1052,
+      "mean_token_accuracy": 0.9668426513671875,
+      "num_tokens": 59342144.0,
+      "step": 7310
+    },
+    {
+      "entropy": 0.23061503767967223,
+      "epoch": 4.128103837471783,
+      "grad_norm": 2.7583887577056885,
+      "learning_rate": 4.61952467994987e-06,
+      "loss": 0.1236,
+      "mean_token_accuracy": 0.9621009945869445,
+      "num_tokens": 59382762.0,
+      "step": 7315
+    },
+    {
+      "entropy": 0.22966778576374053,
+      "epoch": 4.1309255079006775,
+      "grad_norm": 2.4049155712127686,
+      "learning_rate": 4.619022683395675e-06,
+      "loss": 0.131,
+      "mean_token_accuracy": 0.9585760951042175,
+      "num_tokens": 59423356.0,
+      "step": 7320
+    },
+    {
+      "entropy": 0.23155851662158966,
+      "epoch": 4.133747178329571,
+      "grad_norm": 1.8571085929870605,
+      "learning_rate": 4.618520393340528e-06,
+      "loss": 0.1282,
+      "mean_token_accuracy": 0.9598821759223938,
+      "num_tokens": 59463990.0,
+      "step": 7325
+    },
+    {
+      "entropy": 0.2637378484010696,
+      "epoch": 4.136568848758465,
+      "grad_norm": 2.0401177406311035,
+      "learning_rate": 4.618017809883107e-06,
+      "loss": 0.1305,
+      "mean_token_accuracy": 0.9606005668640136,
+      "num_tokens": 59504742.0,
+      "step": 7330
+    },
+    {
+      "entropy": 0.2353299468755722,
+      "epoch": 4.139390519187359,
+      "grad_norm": 2.042680263519287,
+      "learning_rate": 4.617514933122142e-06,
+      "loss": 0.1304,
+      "mean_token_accuracy": 0.9596582651138306,
+      "num_tokens": 59545424.0,
+      "step": 7335
+    },
+    {
+      "entropy": 0.22721785604953765,
+      "epoch": 4.142212189616253,
+      "grad_norm": 2.117658853530884,
+      "learning_rate": 4.6170117631564246e-06,
+      "loss": 0.115,
+      "mean_token_accuracy": 0.9629755258560181,
+      "num_tokens": 59586178.0,
+      "step": 7340
+    },
+    {
+      "entropy": 0.2232923239469528,
+      "epoch": 4.145033860045147,
+      "grad_norm": 1.8017418384552002,
+      "learning_rate": 4.616508300084803e-06,
+      "loss": 0.129,
+      "mean_token_accuracy": 0.9588123679161071,
+      "num_tokens": 59626782.0,
+      "step": 7345
+    },
+    {
+      "entropy": 0.22652737498283387,
+      "epoch": 4.1478555304740405,
+      "grad_norm": 2.2765512466430664,
+      "learning_rate": 4.616004544006181e-06,
+      "loss": 0.1175,
+      "mean_token_accuracy": 0.9620057940483093,
+      "num_tokens": 59667236.0,
+      "step": 7350
+    },
+    {
+      "entropy": 0.23027748465538025,
+      "epoch": 4.150677200902934,
+      "grad_norm": 1.8993479013442993,
+      "learning_rate": 4.615500495019523e-06,
+      "loss": 0.1223,
+      "mean_token_accuracy": 0.9609902024269104,
+      "num_tokens": 59708042.0,
+      "step": 7355
+    },
+    {
+      "entropy": 0.22403572499752045,
+      "epoch": 4.153498871331829,
+      "grad_norm": 2.1379618644714355,
+      "learning_rate": 4.614996153223849e-06,
+      "loss": 0.1266,
+      "mean_token_accuracy": 0.9605488657951355,
+      "num_tokens": 59748650.0,
+      "step": 7360
+    },
+    {
+      "entropy": 0.2490081638097763,
+      "epoch": 4.1563205417607225,
+      "grad_norm": 2.036970853805542,
+      "learning_rate": 4.614491518718237e-06,
+      "loss": 0.1421,
+      "mean_token_accuracy": 0.9558643579483033,
+      "num_tokens": 59789397.0,
+      "step": 7365
+    },
+    {
+      "entropy": 0.21482129395008087,
+      "epoch": 4.159142212189616,
+      "grad_norm": 1.7927038669586182,
+      "learning_rate": 4.613986591601823e-06,
+      "loss": 0.1081,
+      "mean_token_accuracy": 0.9661535143852233,
+      "num_tokens": 59830163.0,
+      "step": 7370
+    },
+    {
+      "entropy": 0.23096846640110016,
+      "epoch": 4.16196388261851,
+      "grad_norm": 2.013101816177368,
+      "learning_rate": 4.613481371973799e-06,
+      "loss": 0.1247,
+      "mean_token_accuracy": 0.9610580563545227,
+      "num_tokens": 59870838.0,
+      "step": 7375
+    },
+    {
+      "entropy": 0.23506858050823212,
+      "epoch": 4.164785553047404,
+      "grad_norm": 2.304497480392456,
+      "learning_rate": 4.612975859933415e-06,
+      "loss": 0.1292,
+      "mean_token_accuracy": 0.9596829771995544,
+      "num_tokens": 59911676.0,
+      "step": 7380
+    },
+    {
+      "entropy": 0.2367018163204193,
+      "epoch": 4.167607223476298,
+      "grad_norm": 2.2349987030029297,
+      "learning_rate": 4.612470055579982e-06,
+      "loss": 0.1316,
+      "mean_token_accuracy": 0.9579625487327575,
+      "num_tokens": 59952218.0,
+      "step": 7385
+    },
+    {
+      "entropy": 0.22246408462524414,
+      "epoch": 4.170428893905192,
+      "grad_norm": 1.9588614702224731,
+      "learning_rate": 4.611963959012862e-06,
+      "loss": 0.1166,
+      "mean_token_accuracy": 0.9629753589630127,
+      "num_tokens": 59992290.0,
+      "step": 7390
+    },
+    {
+      "entropy": 0.22336077094078063,
+      "epoch": 4.173250564334086,
+      "grad_norm": 1.994891881942749,
+      "learning_rate": 4.611457570331479e-06,
+      "loss": 0.1155,
+      "mean_token_accuracy": 0.9629181385040283,
+      "num_tokens": 60033130.0,
+      "step": 7395
+    },
+    {
+      "entropy": 0.24217330813407897,
+      "epoch": 4.176072234762979,
+      "grad_norm": 1.8636099100112915,
+      "learning_rate": 4.610950889635313e-06,
+      "loss": 0.1315,
+      "mean_token_accuracy": 0.9586685538291931,
+      "num_tokens": 60073764.0,
+      "step": 7400
+    },
+    {
+      "entropy": 0.2141832858324051,
+      "epoch": 4.178893905191874,
+      "grad_norm": 1.7918341159820557,
+      "learning_rate": 4.6104439170239015e-06,
+      "loss": 0.1143,
+      "mean_token_accuracy": 0.9644059538841248,
+      "num_tokens": 60114234.0,
+      "step": 7405
+    },
+    {
+      "entropy": 0.2275536447763443,
+      "epoch": 4.181715575620768,
+      "grad_norm": 1.8082565069198608,
+      "learning_rate": 4.609936652596841e-06,
+      "loss": 0.121,
+      "mean_token_accuracy": 0.961764144897461,
+      "num_tokens": 60154149.0,
+      "step": 7410
+    },
+    {
+      "entropy": 0.24505477845668794,
+      "epoch": 4.184537246049661,
+      "grad_norm": 1.9884941577911377,
+      "learning_rate": 4.60942909645378e-06,
+      "loss": 0.1369,
+      "mean_token_accuracy": 0.956404197216034,
+      "num_tokens": 60194901.0,
+      "step": 7415
+    },
+    {
+      "entropy": 0.2267145186662674,
+      "epoch": 4.187358916478555,
+      "grad_norm": 2.0457382202148438,
+      "learning_rate": 4.608921248694431e-06,
+      "loss": 0.129,
+      "mean_token_accuracy": 0.9587201476097107,
+      "num_tokens": 60235805.0,
+      "step": 7420
+    },
+    {
+      "entropy": 0.24519952833652497,
+      "epoch": 4.1901805869074495,
+      "grad_norm": 2.398599147796631,
+      "learning_rate": 4.6084131094185594e-06,
+      "loss": 0.1345,
+      "mean_token_accuracy": 0.9566538453102111,
+      "num_tokens": 60276550.0,
+      "step": 7425
+    },
+    {
+      "entropy": 0.23022598624229432,
+      "epoch": 4.193002257336343,
+      "grad_norm": 1.9493690729141235,
+      "learning_rate": 4.607904678725989e-06,
+      "loss": 0.1237,
+      "mean_token_accuracy": 0.9610730051994324,
+      "num_tokens": 60316464.0,
+      "step": 7430
+    },
+    {
+      "entropy": 0.22809267342090606,
+      "epoch": 4.195823927765237,
+      "grad_norm": 2.157726287841797,
+      "learning_rate": 4.607395956716603e-06,
+      "loss": 0.12,
+      "mean_token_accuracy": 0.9622081279754638,
+      "num_tokens": 60356857.0,
+      "step": 7435
+    },
+    {
+      "entropy": 0.2448444426059723,
+      "epoch": 4.198645598194131,
+      "grad_norm": 2.016291379928589,
+      "learning_rate": 4.606886943490338e-06,
+      "loss": 0.1329,
+      "mean_token_accuracy": 0.9577858328819275,
+      "num_tokens": 60397631.0,
+      "step": 7440
+    },
+    {
+      "entropy": 0.2451613575220108,
+      "epoch": 4.201467268623025,
+      "grad_norm": 2.0515503883361816,
+      "learning_rate": 4.60637763914719e-06,
+      "loss": 0.1414,
+      "mean_token_accuracy": 0.9543289184570313,
+      "num_tokens": 60437991.0,
+      "step": 7445
+    },
+    {
+      "entropy": 0.2392250567674637,
+      "epoch": 4.204288939051919,
+      "grad_norm": 1.8992726802825928,
+      "learning_rate": 4.605868043787213e-06,
+      "loss": 0.1336,
+      "mean_token_accuracy": 0.9581851363182068,
+      "num_tokens": 60478732.0,
+      "step": 7450
+    },
+    {
+      "entropy": 0.23091658651828767,
+      "epoch": 4.207110609480813,
+      "grad_norm": 1.9759114980697632,
+      "learning_rate": 4.605358157510516e-06,
+      "loss": 0.1108,
+      "mean_token_accuracy": 0.9652905344963074,
+      "num_tokens": 60518543.0,
+      "step": 7455
+    },
+    {
+      "entropy": 0.23277661204338074,
+      "epoch": 4.209932279909706,
+      "grad_norm": 1.8714420795440674,
+      "learning_rate": 4.6048479804172666e-06,
+      "loss": 0.1099,
+      "mean_token_accuracy": 0.966291892528534,
+      "num_tokens": 60559260.0,
+      "step": 7460
+    },
+    {
+      "entropy": 0.23967285752296447,
+      "epoch": 4.2127539503386,
+      "grad_norm": 2.056525468826294,
+      "learning_rate": 4.604337512607689e-06,
+      "loss": 0.1228,
+      "mean_token_accuracy": 0.9611961007118225,
+      "num_tokens": 60600020.0,
+      "step": 7465
+    },
+    {
+      "entropy": 0.24567246735095977,
+      "epoch": 4.215575620767495,
+      "grad_norm": 2.1545474529266357,
+      "learning_rate": 4.603826754182065e-06,
+      "loss": 0.1493,
+      "mean_token_accuracy": 0.9537813544273377,
+      "num_tokens": 60640575.0,
+      "step": 7470
+    },
+    {
+      "entropy": 0.2454807221889496,
+      "epoch": 4.218397291196388,
+      "grad_norm": 2.1680426597595215,
+      "learning_rate": 4.603315705240732e-06,
+      "loss": 0.1382,
+      "mean_token_accuracy": 0.956198763847351,
+      "num_tokens": 60681128.0,
+      "step": 7475
+    },
+    {
+      "entropy": 0.24590539932250977,
+      "epoch": 4.221218961625282,
+      "grad_norm": 1.8926993608474731,
+      "learning_rate": 4.602804365884088e-06,
+      "loss": 0.1257,
+      "mean_token_accuracy": 0.960571038722992,
+      "num_tokens": 60721973.0,
+      "step": 7480
+    },
+    {
+      "entropy": 0.23549313545227052,
+      "epoch": 4.224040632054176,
+      "grad_norm": 2.156057119369507,
+      "learning_rate": 4.602292736212583e-06,
+      "loss": 0.1329,
+      "mean_token_accuracy": 0.9587239861488343,
+      "num_tokens": 60762628.0,
+      "step": 7485
+    },
+    {
+      "entropy": 0.2402738958597183,
+      "epoch": 4.22686230248307,
+      "grad_norm": 1.9733823537826538,
+      "learning_rate": 4.601780816326728e-06,
+      "loss": 0.1245,
+      "mean_token_accuracy": 0.9602890253067017,
+      "num_tokens": 60803434.0,
+      "step": 7490
+    },
+    {
+      "entropy": 0.243062624335289,
+      "epoch": 4.229683972911964,
+      "grad_norm": 2.058126211166382,
+      "learning_rate": 4.60126860632709e-06,
+      "loss": 0.1329,
+      "mean_token_accuracy": 0.9585610747337341,
+      "num_tokens": 60844279.0,
+      "step": 7495
+    },
+    {
+      "entropy": 0.23409413993358613,
+      "epoch": 4.232505643340858,
+      "grad_norm": 1.9692258834838867,
+      "learning_rate": 4.600756106314292e-06,
+      "loss": 0.1167,
+      "mean_token_accuracy": 0.9634551763534546,
+      "num_tokens": 60884801.0,
+      "step": 7500
+    },
+    {
+      "epoch": 4.232505643340858,
+      "eval_entropy": 0.2763104736804962,
+      "eval_loss": 0.11991167068481445,
+      "eval_mean_token_accuracy": 0.9616711139678955,
+      "eval_num_tokens": 60884801.0,
+      "eval_runtime": 0.1639,
+      "eval_samples_per_second": 24.403,
+      "eval_steps_per_second": 6.101,
+      "step": 7500
+    },
+    {
+      "entropy": 0.24236758947372436,
+      "epoch": 4.235327313769751,
+      "grad_norm": 2.3060202598571777,
+      "learning_rate": 4.6002433163890156e-06,
+      "loss": 0.1294,
+      "mean_token_accuracy": 0.959705400466919,
+      "num_tokens": 60925426.0,
+      "step": 7505
+    },
+    {
+      "entropy": 0.21380599737167358,
+      "epoch": 4.238148984198646,
+      "grad_norm": 2.0318894386291504,
+      "learning_rate": 4.599730236651998e-06,
+      "loss": 0.1058,
+      "mean_token_accuracy": 0.965471088886261,
+      "num_tokens": 60966196.0,
+      "step": 7510
+    },
+    {
+      "entropy": 0.22909758388996124,
+      "epoch": 4.24097065462754,
+      "grad_norm": 1.807207465171814,
+      "learning_rate": 4.5992168672040335e-06,
+      "loss": 0.1201,
+      "mean_token_accuracy": 0.9617739319801331,
+      "num_tokens": 61006851.0,
+      "step": 7515
+    },
+    {
+      "entropy": 0.2145596742630005,
+      "epoch": 4.243792325056433,
+      "grad_norm": 1.704498052597046,
+      "learning_rate": 4.598703208145974e-06,
+      "loss": 0.1079,
+      "mean_token_accuracy": 0.9660002827644348,
+      "num_tokens": 61047687.0,
+      "step": 7520
+    },
+    {
+      "entropy": 0.2247183084487915,
+      "epoch": 4.246613995485327,
+      "grad_norm": 2.1221797466278076,
+      "learning_rate": 4.598189259578727e-06,
+      "loss": 0.1279,
+      "mean_token_accuracy": 0.9582555651664734,
+      "num_tokens": 61088309.0,
+      "step": 7525
+    },
+    {
+      "entropy": 0.23647244274616241,
+      "epoch": 4.249435665914222,
+      "grad_norm": 2.2665932178497314,
+      "learning_rate": 4.597675021603259e-06,
+      "loss": 0.1294,
+      "mean_token_accuracy": 0.9585322737693787,
+      "num_tokens": 61129103.0,
+      "step": 7530
+    },
+    {
+      "entropy": 0.21924723386764527,
+      "epoch": 4.252257336343115,
+      "grad_norm": 2.0743370056152344,
+      "learning_rate": 4.597160494320592e-06,
+      "loss": 0.1204,
+      "mean_token_accuracy": 0.9610903978347778,
+      "num_tokens": 61169817.0,
+      "step": 7535
+    },
+    {
+      "entropy": 0.22164719700813293,
+      "epoch": 4.255079006772009,
+      "grad_norm": 1.8417187929153442,
+      "learning_rate": 4.596645677831804e-06,
+      "loss": 0.1236,
+      "mean_token_accuracy": 0.9602578401565551,
+      "num_tokens": 61210598.0,
+      "step": 7540
+    },
+    {
+      "entropy": 0.2332447052001953,
+      "epoch": 4.257900677200903,
+      "grad_norm": 1.8406648635864258,
+      "learning_rate": 4.596130572238031e-06,
+      "loss": 0.1174,
+      "mean_token_accuracy": 0.9627135396003723,
+      "num_tokens": 61251185.0,
+      "step": 7545
+    },
+    {
+      "entropy": 0.23616610169410707,
+      "epoch": 4.260722347629796,
+      "grad_norm": 1.8603347539901733,
+      "learning_rate": 4.595615177640466e-06,
+      "loss": 0.1263,
+      "mean_token_accuracy": 0.9601802945137023,
+      "num_tokens": 61291752.0,
+      "step": 7550
+    },
+    {
+      "entropy": 0.22554007172584534,
+      "epoch": 4.263544018058691,
+      "grad_norm": 2.3185107707977295,
+      "learning_rate": 4.59509949414036e-06,
+      "loss": 0.119,
+      "mean_token_accuracy": 0.9624182105064392,
+      "num_tokens": 61332441.0,
+      "step": 7555
+    },
+    {
+      "entropy": 0.2530226230621338,
+      "epoch": 4.266365688487585,
+      "grad_norm": 2.326624631881714,
+      "learning_rate": 4.594583521839015e-06,
+      "loss": 0.1415,
+      "mean_token_accuracy": 0.9568467020988465,
+      "num_tokens": 61373014.0,
+      "step": 7560
+    },
+    {
+      "entropy": 0.22727547585964203,
+      "epoch": 4.269187358916478,
+      "grad_norm": 2.162659168243408,
+      "learning_rate": 4.594067260837796e-06,
+      "loss": 0.1309,
+      "mean_token_accuracy": 0.9592653393745423,
+      "num_tokens": 61413745.0,
+      "step": 7565
+    },
+    {
+      "entropy": 0.24552667737007142,
+      "epoch": 4.272009029345372,
+      "grad_norm": 2.3903839588165283,
+      "learning_rate": 4.593550711238123e-06,
+      "loss": 0.1324,
+      "mean_token_accuracy": 0.9594314217567443,
+      "num_tokens": 61454234.0,
+      "step": 7570
+    },
+    {
+      "entropy": 0.23494355678558348,
+      "epoch": 4.274830699774267,
+      "grad_norm": 1.9757755994796753,
+      "learning_rate": 4.5930338731414726e-06,
+      "loss": 0.1304,
+      "mean_token_accuracy": 0.9587761878967285,
+      "num_tokens": 61494930.0,
+      "step": 7575
+    },
+    {
+      "entropy": 0.24661367535591125,
+      "epoch": 4.27765237020316,
+      "grad_norm": 2.0399813652038574,
+      "learning_rate": 4.592516746649377e-06,
+      "loss": 0.1321,
+      "mean_token_accuracy": 0.9585609912872315,
+      "num_tokens": 61535747.0,
+      "step": 7580
+    },
+    {
+      "entropy": 0.23092095851898192,
+      "epoch": 4.280474040632054,
+      "grad_norm": 2.1672401428222656,
+      "learning_rate": 4.591999331863425e-06,
+      "loss": 0.1239,
+      "mean_token_accuracy": 0.9604844927787781,
+      "num_tokens": 61576529.0,
+      "step": 7585
+    },
+    {
+      "entropy": 0.2526959002017975,
+      "epoch": 4.283295711060948,
+      "grad_norm": 2.322620153427124,
+      "learning_rate": 4.5914816288852645e-06,
+      "loss": 0.1319,
+      "mean_token_accuracy": 0.9580205917358399,
+      "num_tokens": 61617030.0,
+      "step": 7590
+    },
+    {
+      "entropy": 0.22551030814647674,
+      "epoch": 4.286117381489842,
+      "grad_norm": 1.7364836931228638,
+      "learning_rate": 4.590963637816596e-06,
+      "loss": 0.127,
+      "mean_token_accuracy": 0.960225510597229,
+      "num_tokens": 61657559.0,
+      "step": 7595
+    },
+    {
+      "entropy": 0.22876082360744476,
+      "epoch": 4.288939051918736,
+      "grad_norm": 1.735487461090088,
+      "learning_rate": 4.590445358759181e-06,
+      "loss": 0.1165,
+      "mean_token_accuracy": 0.961608099937439,
+      "num_tokens": 61697918.0,
+      "step": 7600
+    },
+    {
+      "entropy": 0.23216151893138887,
+      "epoch": 4.29176072234763,
+      "grad_norm": 2.070525646209717,
+      "learning_rate": 4.589926791814836e-06,
+      "loss": 0.1204,
+      "mean_token_accuracy": 0.9610988855361938,
+      "num_tokens": 61738812.0,
+      "step": 7605
+    },
+    {
+      "entropy": 0.23693689107894897,
+      "epoch": 4.294582392776523,
+      "grad_norm": 2.203585147857666,
+      "learning_rate": 4.589407937085431e-06,
+      "loss": 0.1247,
+      "mean_token_accuracy": 0.9603411674499511,
+      "num_tokens": 61779432.0,
+      "step": 7610
+    },
+    {
+      "entropy": 0.2458764672279358,
+      "epoch": 4.297404063205418,
+      "grad_norm": 2.2714271545410156,
+      "learning_rate": 4.5888887946728966e-06,
+      "loss": 0.1412,
+      "mean_token_accuracy": 0.9558348536491394,
+      "num_tokens": 61820028.0,
+      "step": 7615
+    },
+    {
+      "entropy": 0.23253954946994781,
+      "epoch": 4.300225733634312,
+      "grad_norm": 2.077993154525757,
+      "learning_rate": 4.588369364679217e-06,
+      "loss": 0.1219,
+      "mean_token_accuracy": 0.9627041578292846,
+      "num_tokens": 61860687.0,
+      "step": 7620
+    },
+    {
+      "entropy": 0.2426275998353958,
+      "epoch": 4.303047404063205,
+      "grad_norm": 2.0825798511505127,
+      "learning_rate": 4.587849647206437e-06,
+      "loss": 0.1395,
+      "mean_token_accuracy": 0.9566293358802795,
+      "num_tokens": 61901354.0,
+      "step": 7625
+    },
+    {
+      "entropy": 0.2425705909729004,
+      "epoch": 4.305869074492099,
+      "grad_norm": 1.9618566036224365,
+      "learning_rate": 4.587329642356654e-06,
+      "loss": 0.1382,
+      "mean_token_accuracy": 0.9561069011688232,
+      "num_tokens": 61941799.0,
+      "step": 7630
+    },
+    {
+      "entropy": 0.2432687759399414,
+      "epoch": 4.308690744920993,
+      "grad_norm": 1.9312220811843872,
+      "learning_rate": 4.586809350232022e-06,
+      "loss": 0.1262,
+      "mean_token_accuracy": 0.9600926876068115,
+      "num_tokens": 61982412.0,
+      "step": 7635
+    },
+    {
+      "entropy": 0.22498698830604552,
+      "epoch": 4.311512415349887,
+      "grad_norm": 2.0869600772857666,
+      "learning_rate": 4.586288770934753e-06,
+      "loss": 0.1247,
+      "mean_token_accuracy": 0.9600854396820069,
+      "num_tokens": 62022929.0,
+      "step": 7640
+    },
+    {
+      "entropy": 0.2627303659915924,
+      "epoch": 4.314334085778781,
+      "grad_norm": 2.0073163509368896,
+      "learning_rate": 4.585767904567115e-06,
+      "loss": 0.1315,
+      "mean_token_accuracy": 0.9589227676391602,
+      "num_tokens": 62063278.0,
+      "step": 7645
+    },
+    {
+      "entropy": 0.2304110735654831,
+      "epoch": 4.317155756207675,
+      "grad_norm": 1.8721083402633667,
+      "learning_rate": 4.585246751231433e-06,
+      "loss": 0.1155,
+      "mean_token_accuracy": 0.962857437133789,
+      "num_tokens": 62103307.0,
+      "step": 7650
+    },
+    {
+      "entropy": 0.24186222553253173,
+      "epoch": 4.3199774266365685,
+      "grad_norm": 2.2419540882110596,
+      "learning_rate": 4.584725311030085e-06,
+      "loss": 0.126,
+      "mean_token_accuracy": 0.9607646465301514,
+      "num_tokens": 62144023.0,
+      "step": 7655
+    },
+    {
+      "entropy": 0.22962496876716615,
+      "epoch": 4.322799097065463,
+      "grad_norm": 2.0205132961273193,
+      "learning_rate": 4.584203584065512e-06,
+      "loss": 0.1339,
+      "mean_token_accuracy": 0.9580964565277099,
+      "num_tokens": 62184285.0,
+      "step": 7660
+    },
+    {
+      "entropy": 0.22456653118133546,
+      "epoch": 4.325620767494357,
+      "grad_norm": 1.9401427507400513,
+      "learning_rate": 4.583681570440204e-06,
+      "loss": 0.1209,
+      "mean_token_accuracy": 0.96328786611557,
+      "num_tokens": 62224200.0,
+      "step": 7665
+    },
+    {
+      "entropy": 0.22748615145683287,
+      "epoch": 4.3284424379232505,
+      "grad_norm": 1.9831054210662842,
+      "learning_rate": 4.583159270256712e-06,
+      "loss": 0.1157,
+      "mean_token_accuracy": 0.9626774907112121,
+      "num_tokens": 62264973.0,
+      "step": 7670
+    },
+    {
+      "entropy": 0.2501515805721283,
+      "epoch": 4.331264108352144,
+      "grad_norm": 2.2783570289611816,
+      "learning_rate": 4.582636683617643e-06,
+      "loss": 0.1268,
+      "mean_token_accuracy": 0.9602941393852233,
+      "num_tokens": 62305564.0,
+      "step": 7675
+    },
+    {
+      "entropy": 0.238142928481102,
+      "epoch": 4.334085778781039,
+      "grad_norm": 2.0721852779388428,
+      "learning_rate": 4.582113810625657e-06,
+      "loss": 0.133,
+      "mean_token_accuracy": 0.958099901676178,
+      "num_tokens": 62345999.0,
+      "step": 7680
+    },
+    {
+      "entropy": 0.22630032598972322,
+      "epoch": 4.336907449209932,
+      "grad_norm": 1.8380122184753418,
+      "learning_rate": 4.581590651383473e-06,
+      "loss": 0.1168,
+      "mean_token_accuracy": 0.9639391660690307,
+      "num_tokens": 62386861.0,
+      "step": 7685
+    },
+    {
+      "entropy": 0.2205124467611313,
+      "epoch": 4.339729119638826,
+      "grad_norm": 2.063775062561035,
+      "learning_rate": 4.581067205993867e-06,
+      "loss": 0.1078,
+      "mean_token_accuracy": 0.9652288436889649,
+      "num_tokens": 62427526.0,
+      "step": 7690
+    },
+    {
+      "entropy": 0.23350533246994018,
+      "epoch": 4.34255079006772,
+      "grad_norm": 2.0544824600219727,
+      "learning_rate": 4.580543474559669e-06,
+      "loss": 0.1333,
+      "mean_token_accuracy": 0.9583484888076782,
+      "num_tokens": 62468033.0,
+      "step": 7695
+    },
+    {
+      "entropy": 0.23622917830944062,
+      "epoch": 4.345372460496614,
+      "grad_norm": 1.9537707567214966,
+      "learning_rate": 4.580019457183766e-06,
+      "loss": 0.1248,
+      "mean_token_accuracy": 0.9608685970306396,
+      "num_tokens": 62508513.0,
+      "step": 7700
+    },
+    {
+      "entropy": 0.22541263699531555,
+      "epoch": 4.348194130925508,
+      "grad_norm": 1.8316019773483276,
+      "learning_rate": 4.579495153969102e-06,
+      "loss": 0.1235,
+      "mean_token_accuracy": 0.9609684944152832,
+      "num_tokens": 62549331.0,
+      "step": 7705
+    },
+    {
+      "entropy": 0.2408158302307129,
+      "epoch": 4.351015801354402,
+      "grad_norm": 2.1585938930511475,
+      "learning_rate": 4.578970565018676e-06,
+      "loss": 0.1357,
+      "mean_token_accuracy": 0.9570087313652038,
+      "num_tokens": 62590028.0,
+      "step": 7710
+    },
+    {
+      "entropy": 0.2397814780473709,
+      "epoch": 4.3538374717832955,
+      "grad_norm": 1.9146168231964111,
+      "learning_rate": 4.578445690435542e-06,
+      "loss": 0.1205,
+      "mean_token_accuracy": 0.9615342140197753,
+      "num_tokens": 62630583.0,
+      "step": 7715
+    },
+    {
+      "entropy": 0.2495466351509094,
+      "epoch": 4.356659142212189,
+      "grad_norm": 2.213092565536499,
+      "learning_rate": 4.577920530322815e-06,
+      "loss": 0.1317,
+      "mean_token_accuracy": 0.9592637538909912,
+      "num_tokens": 62671379.0,
+      "step": 7720
+    },
+    {
+      "entropy": 0.25447321236133574,
+      "epoch": 4.359480812641084,
+      "grad_norm": 2.162130355834961,
+      "learning_rate": 4.5773950847836604e-06,
+      "loss": 0.1476,
+      "mean_token_accuracy": 0.9534005284309387,
+      "num_tokens": 62712032.0,
+      "step": 7725
+    },
+    {
+      "entropy": 0.242551389336586,
+      "epoch": 4.3623024830699775,
+      "grad_norm": 2.0905115604400635,
+      "learning_rate": 4.576869353921302e-06,
+      "loss": 0.1325,
+      "mean_token_accuracy": 0.9586278200149536,
+      "num_tokens": 62752939.0,
+      "step": 7730
+    },
+    {
+      "entropy": 0.2576675295829773,
+      "epoch": 4.365124153498871,
+      "grad_norm": 2.0513839721679688,
+      "learning_rate": 4.5763433378390205e-06,
+      "loss": 0.1467,
+      "mean_token_accuracy": 0.9542353510856628,
+      "num_tokens": 62793276.0,
+      "step": 7735
+    },
+    {
+      "entropy": 0.2226796418428421,
+      "epoch": 4.367945823927765,
+      "grad_norm": 2.049790382385254,
+      "learning_rate": 4.575817036640153e-06,
+      "loss": 0.1265,
+      "mean_token_accuracy": 0.9590314865112305,
+      "num_tokens": 62833681.0,
+      "step": 7740
+    },
+    {
+      "entropy": 0.22169795334339143,
+      "epoch": 4.3707674943566595,
+      "grad_norm": 2.1573617458343506,
+      "learning_rate": 4.575290450428088e-06,
+      "loss": 0.1141,
+      "mean_token_accuracy": 0.9643651604652405,
+      "num_tokens": 62874247.0,
+      "step": 7745
+    },
+    {
+      "entropy": 0.21064240634441375,
+      "epoch": 4.373589164785553,
+      "grad_norm": 2.0029399394989014,
+      "learning_rate": 4.574763579306276e-06,
+      "loss": 0.114,
+      "mean_token_accuracy": 0.9654483675956727,
+      "num_tokens": 62914852.0,
+      "step": 7750
+    },
+    {
+      "entropy": 0.2374986946582794,
+      "epoch": 4.376410835214447,
+      "grad_norm": 2.325894355773926,
+      "learning_rate": 4.574236423378221e-06,
+      "loss": 0.1237,
+      "mean_token_accuracy": 0.9615754604339599,
+      "num_tokens": 62955498.0,
+      "step": 7755
+    },
+    {
+      "entropy": 0.2390218436717987,
+      "epoch": 4.3792325056433405,
+      "grad_norm": 2.234550952911377,
+      "learning_rate": 4.5737089827474826e-06,
+      "loss": 0.1278,
+      "mean_token_accuracy": 0.9586256504058838,
+      "num_tokens": 62995636.0,
+      "step": 7760
+    },
+    {
+      "entropy": 0.2208180695772171,
+      "epoch": 4.382054176072235,
+      "grad_norm": 2.097153902053833,
+      "learning_rate": 4.573181257517675e-06,
+      "loss": 0.1101,
+      "mean_token_accuracy": 0.9645330548286438,
+      "num_tokens": 63036376.0,
+      "step": 7765
+    },
+    {
+      "entropy": 0.25867641568183897,
+      "epoch": 4.384875846501129,
+      "grad_norm": 1.9841035604476929,
+      "learning_rate": 4.572653247792471e-06,
+      "loss": 0.136,
+      "mean_token_accuracy": 0.9573712944984436,
+      "num_tokens": 63076910.0,
+      "step": 7770
+    },
+    {
+      "entropy": 0.22610498666763307,
+      "epoch": 4.3876975169300225,
+      "grad_norm": 1.8051813840866089,
+      "learning_rate": 4.572124953675599e-06,
+      "loss": 0.1146,
+      "mean_token_accuracy": 0.963974404335022,
+      "num_tokens": 63117280.0,
+      "step": 7775
+    },
+    {
+      "entropy": 0.24536430835723877,
+      "epoch": 4.390519187358916,
+      "grad_norm": 2.267284870147705,
+      "learning_rate": 4.571596375270843e-06,
+      "loss": 0.1263,
+      "mean_token_accuracy": 0.9603652000427246,
+      "num_tokens": 63158036.0,
+      "step": 7780
+    },
+    {
+      "entropy": 0.2449027419090271,
+      "epoch": 4.393340857787811,
+      "grad_norm": 1.866507649421692,
+      "learning_rate": 4.5710675126820394e-06,
+      "loss": 0.1414,
+      "mean_token_accuracy": 0.9538567900657654,
+      "num_tokens": 63198608.0,
+      "step": 7785
+    },
+    {
+      "entropy": 0.2529394537210464,
+      "epoch": 4.3961625282167045,
+      "grad_norm": 2.741823673248291,
+      "learning_rate": 4.570538366013085e-06,
+      "loss": 0.136,
+      "mean_token_accuracy": 0.957249391078949,
+      "num_tokens": 63239140.0,
+      "step": 7790
+    },
+    {
+      "entropy": 0.2259007692337036,
+      "epoch": 4.398984198645598,
+      "grad_norm": 1.8987911939620972,
+      "learning_rate": 4.570008935367931e-06,
+      "loss": 0.1297,
+      "mean_token_accuracy": 0.9585871219635009,
+      "num_tokens": 63279866.0,
+      "step": 7795
+    },
+    {
+      "entropy": 0.22481289207935334,
+      "epoch": 4.401805869074492,
+      "grad_norm": 2.1305699348449707,
+      "learning_rate": 4.569479220850583e-06,
+      "loss": 0.125,
+      "mean_token_accuracy": 0.9605366587638855,
+      "num_tokens": 63320412.0,
+      "step": 7800
+    },
+    {
+      "entropy": 0.23074941635131835,
+      "epoch": 4.404627539503386,
+      "grad_norm": 1.930199384689331,
+      "learning_rate": 4.568949222565105e-06,
+      "loss": 0.1218,
+      "mean_token_accuracy": 0.9614872097969055,
+      "num_tokens": 63361131.0,
+      "step": 7805
+    },
+    {
+      "entropy": 0.22710799276828766,
+      "epoch": 4.40744920993228,
+      "grad_norm": 1.8718585968017578,
+      "learning_rate": 4.568418940615616e-06,
+      "loss": 0.119,
+      "mean_token_accuracy": 0.9626579761505127,
+      "num_tokens": 63401708.0,
+      "step": 7810
+    },
+    {
+      "entropy": 0.24245786666870117,
+      "epoch": 4.410270880361174,
+      "grad_norm": 2.3666772842407227,
+      "learning_rate": 4.567888375106286e-06,
+      "loss": 0.1383,
+      "mean_token_accuracy": 0.956715726852417,
+      "num_tokens": 63442479.0,
+      "step": 7815
+    },
+    {
+      "entropy": 0.2398514539003372,
+      "epoch": 4.413092550790068,
+      "grad_norm": 2.2359414100646973,
+      "learning_rate": 4.567357526141349e-06,
+      "loss": 0.133,
+      "mean_token_accuracy": 0.9572691440582275,
+      "num_tokens": 63482841.0,
+      "step": 7820
+    },
+    {
+      "entropy": 0.23423077166080475,
+      "epoch": 4.415914221218961,
+      "grad_norm": 2.1519811153411865,
+      "learning_rate": 4.5668263938250876e-06,
+      "loss": 0.1264,
+      "mean_token_accuracy": 0.9597557783126831,
+      "num_tokens": 63523538.0,
+      "step": 7825
+    },
+    {
+      "entropy": 0.2178510457277298,
+      "epoch": 4.418735891647856,
+      "grad_norm": 1.9678175449371338,
+      "learning_rate": 4.566294978261844e-06,
+      "loss": 0.1131,
+      "mean_token_accuracy": 0.9639857292175293,
+      "num_tokens": 63564255.0,
+      "step": 7830
+    },
+    {
+      "entropy": 0.23961224555969238,
+      "epoch": 4.4215575620767495,
+      "grad_norm": 2.6398816108703613,
+      "learning_rate": 4.565763279556014e-06,
+      "loss": 0.137,
+      "mean_token_accuracy": 0.9575521588325501,
+      "num_tokens": 63604882.0,
+      "step": 7835
+    },
+    {
+      "entropy": 0.24535090029239653,
+      "epoch": 4.424379232505643,
+      "grad_norm": 2.135787010192871,
+      "learning_rate": 4.565231297812051e-06,
+      "loss": 0.127,
+      "mean_token_accuracy": 0.9597046256065369,
+      "num_tokens": 63645157.0,
+      "step": 7840
+    },
+    {
+      "entropy": 0.2281971275806427,
+      "epoch": 4.427200902934537,
+      "grad_norm": 2.0932908058166504,
+      "learning_rate": 4.564699033134462e-06,
+      "loss": 0.1229,
+      "mean_token_accuracy": 0.9616926431655883,
+      "num_tokens": 63685797.0,
+      "step": 7845
+    },
+    {
+      "entropy": 0.24232443869113923,
+      "epoch": 4.4300225733634315,
+      "grad_norm": 1.8690487146377563,
+      "learning_rate": 4.564166485627811e-06,
+      "loss": 0.1439,
+      "mean_token_accuracy": 0.9539458394050598,
+      "num_tokens": 63726385.0,
+      "step": 7850
+    },
+    {
+      "entropy": 0.2243878573179245,
+      "epoch": 4.432844243792325,
+      "grad_norm": 2.3627984523773193,
+      "learning_rate": 4.563633655396717e-06,
+      "loss": 0.1282,
+      "mean_token_accuracy": 0.9588661789894104,
+      "num_tokens": 63767053.0,
+      "step": 7855
+    },
+    {
+      "entropy": 0.25156151950359346,
+      "epoch": 4.435665914221219,
+      "grad_norm": 2.046766996383667,
+      "learning_rate": 4.563100542545854e-06,
+      "loss": 0.1355,
+      "mean_token_accuracy": 0.9556037425994873,
+      "num_tokens": 63807612.0,
+      "step": 7860
+    },
+    {
+      "entropy": 0.23665276169776917,
+      "epoch": 4.438487584650113,
+      "grad_norm": 2.13456130027771,
+      "learning_rate": 4.5625671471799535e-06,
+      "loss": 0.1412,
+      "mean_token_accuracy": 0.955809724330902,
+      "num_tokens": 63848217.0,
+      "step": 7865
+    },
+    {
+      "entropy": 0.22619423866271973,
+      "epoch": 4.441309255079007,
+      "grad_norm": 2.092322587966919,
+      "learning_rate": 4.562033469403799e-06,
+      "loss": 0.1183,
+      "mean_token_accuracy": 0.9627872586250306,
+      "num_tokens": 63888701.0,
+      "step": 7870
+    },
+    {
+      "entropy": 0.22337720394134522,
+      "epoch": 4.444130925507901,
+      "grad_norm": 1.8521798849105835,
+      "learning_rate": 4.561499509322233e-06,
+      "loss": 0.1193,
+      "mean_token_accuracy": 0.9614850997924804,
+      "num_tokens": 63929411.0,
+      "step": 7875
+    },
+    {
+      "entropy": 0.233727565407753,
+      "epoch": 4.446952595936795,
+      "grad_norm": 1.6389681100845337,
+      "learning_rate": 4.560965267040151e-06,
+      "loss": 0.1149,
+      "mean_token_accuracy": 0.9633271217346191,
+      "num_tokens": 63970272.0,
+      "step": 7880
+    },
+    {
+      "entropy": 0.21891143321990966,
+      "epoch": 4.449774266365688,
+      "grad_norm": 2.2698888778686523,
+      "learning_rate": 4.560430742662506e-06,
+      "loss": 0.1212,
+      "mean_token_accuracy": 0.9615803718566894,
+      "num_tokens": 64010993.0,
+      "step": 7885
+    },
+    {
+      "entropy": 0.257635697722435,
+      "epoch": 4.452595936794582,
+      "grad_norm": 2.019932270050049,
+      "learning_rate": 4.559895936294305e-06,
+      "loss": 0.1328,
+      "mean_token_accuracy": 0.9580089688301087,
+      "num_tokens": 64051747.0,
+      "step": 7890
+    },
+    {
+      "entropy": 0.2612759441137314,
+      "epoch": 4.455417607223477,
+      "grad_norm": 2.1498031616210938,
+      "learning_rate": 4.559360848040611e-06,
+      "loss": 0.1354,
+      "mean_token_accuracy": 0.9571520805358886,
+      "num_tokens": 64092298.0,
+      "step": 7895
+    },
+    {
+      "entropy": 0.24472321271896363,
+      "epoch": 4.45823927765237,
+      "grad_norm": 2.2443246841430664,
+      "learning_rate": 4.558825478006543e-06,
+      "loss": 0.1298,
+      "mean_token_accuracy": 0.9582807421684265,
+      "num_tokens": 64132966.0,
+      "step": 7900
+    },
+    {
+      "entropy": 0.24086927771568298,
+      "epoch": 4.461060948081264,
+      "grad_norm": 2.0586602687835693,
+      "learning_rate": 4.5582898262972715e-06,
+      "loss": 0.1147,
+      "mean_token_accuracy": 0.9626161217689514,
+      "num_tokens": 64173716.0,
+      "step": 7905
+    },
+    {
+      "entropy": 0.22729225158691407,
+      "epoch": 4.463882618510158,
+      "grad_norm": 2.192944049835205,
+      "learning_rate": 4.557753893018028e-06,
+      "loss": 0.1231,
+      "mean_token_accuracy": 0.962181007862091,
+      "num_tokens": 64214398.0,
+      "step": 7910
+    },
+    {
+      "entropy": 0.24374393820762635,
+      "epoch": 4.466704288939052,
+      "grad_norm": 2.042626142501831,
+      "learning_rate": 4.557217678274097e-06,
+      "loss": 0.1412,
+      "mean_token_accuracy": 0.9556024193763732,
+      "num_tokens": 64254854.0,
+      "step": 7915
+    },
+    {
+      "entropy": 0.23933674693107604,
+      "epoch": 4.469525959367946,
+      "grad_norm": 1.807453989982605,
+      "learning_rate": 4.556681182170816e-06,
+      "loss": 0.1334,
+      "mean_token_accuracy": 0.9586431503295898,
+      "num_tokens": 64295435.0,
+      "step": 7920
+    },
+    {
+      "entropy": 0.2443247377872467,
+      "epoch": 4.47234762979684,
+      "grad_norm": 2.1040701866149902,
+      "learning_rate": 4.55614440481358e-06,
+      "loss": 0.1288,
+      "mean_token_accuracy": 0.9598910212516785,
+      "num_tokens": 64336069.0,
+      "step": 7925
+    },
+    {
+      "entropy": 0.24531608521938325,
+      "epoch": 4.475169300225733,
+      "grad_norm": 2.0442538261413574,
+      "learning_rate": 4.555607346307841e-06,
+      "loss": 0.1359,
+      "mean_token_accuracy": 0.9561710357666016,
+      "num_tokens": 64376772.0,
+      "step": 7930
+    },
+    {
+      "entropy": 0.2342956393957138,
+      "epoch": 4.477990970654628,
+      "grad_norm": 1.9000775814056396,
+      "learning_rate": 4.555070006759102e-06,
+      "loss": 0.1351,
+      "mean_token_accuracy": 0.9563033699989318,
+      "num_tokens": 64417469.0,
+      "step": 7935
+    },
+    {
+      "entropy": 0.22305310368537903,
+      "epoch": 4.480812641083522,
+      "grad_norm": 1.8340357542037964,
+      "learning_rate": 4.554532386272925e-06,
+      "loss": 0.1233,
+      "mean_token_accuracy": 0.9609080910682678,
+      "num_tokens": 64458183.0,
+      "step": 7940
+    },
+    {
+      "entropy": 0.24366792142391205,
+      "epoch": 4.483634311512415,
+      "grad_norm": 2.0451269149780273,
+      "learning_rate": 4.5539944849549244e-06,
+      "loss": 0.1332,
+      "mean_token_accuracy": 0.9595473170280456,
+      "num_tokens": 64499084.0,
+      "step": 7945
+    },
+    {
+      "entropy": 0.2458895593881607,
+      "epoch": 4.486455981941309,
+      "grad_norm": 2.0097408294677734,
+      "learning_rate": 4.553456302910771e-06,
+      "loss": 0.1273,
+      "mean_token_accuracy": 0.9585718154907227,
+      "num_tokens": 64539911.0,
+      "step": 7950
+    },
+    {
+      "entropy": 0.26139111518859864,
+      "epoch": 4.489277652370204,
+      "grad_norm": 2.031979560852051,
+      "learning_rate": 4.552917840246191e-06,
+      "loss": 0.1372,
+      "mean_token_accuracy": 0.9563012480735779,
+      "num_tokens": 64580156.0,
+      "step": 7955
+    },
+    {
+      "entropy": 0.2303838014602661,
+      "epoch": 4.492099322799097,
+      "grad_norm": 1.742417573928833,
+      "learning_rate": 4.552379097066967e-06,
+      "loss": 0.128,
+      "mean_token_accuracy": 0.959154736995697,
+      "num_tokens": 64620865.0,
+      "step": 7960
+    },
+    {
+      "entropy": 0.24330146610736847,
+      "epoch": 4.494920993227991,
+      "grad_norm": 2.139296770095825,
+      "learning_rate": 4.551840073478934e-06,
+      "loss": 0.134,
+      "mean_token_accuracy": 0.9584646821022034,
+      "num_tokens": 64661404.0,
+      "step": 7965
+    },
+    {
+      "entropy": 0.24887127578258514,
+      "epoch": 4.497742663656885,
+      "grad_norm": 2.17922306060791,
+      "learning_rate": 4.551300769587982e-06,
+      "loss": 0.133,
+      "mean_token_accuracy": 0.9567908644676208,
+      "num_tokens": 64701858.0,
+      "step": 7970
+    },
+    {
+      "entropy": 0.2185787171125412,
+      "epoch": 4.500564334085778,
+      "grad_norm": 2.0657732486724854,
+      "learning_rate": 4.550761185500059e-06,
+      "loss": 0.1253,
+      "mean_token_accuracy": 0.9603734970092773,
+      "num_tokens": 64742554.0,
+      "step": 7975
+    },
+    {
+      "entropy": 0.2412070006132126,
+      "epoch": 4.503386004514673,
+      "grad_norm": 1.9913356304168701,
+      "learning_rate": 4.550221321321165e-06,
+      "loss": 0.1238,
+      "mean_token_accuracy": 0.959455955028534,
+      "num_tokens": 64783017.0,
+      "step": 7980
+    },
+    {
+      "entropy": 0.24967373609542848,
+      "epoch": 4.506207674943567,
+      "grad_norm": 2.3349766731262207,
+      "learning_rate": 4.549681177157358e-06,
+      "loss": 0.1318,
+      "mean_token_accuracy": 0.9588298797607422,
+      "num_tokens": 64823373.0,
+      "step": 7985
+    },
+    {
+      "entropy": 0.22077546417713165,
+      "epoch": 4.50902934537246,
+      "grad_norm": 1.7486339807510376,
+      "learning_rate": 4.549140753114748e-06,
+      "loss": 0.1058,
+      "mean_token_accuracy": 0.9653358578681945,
+      "num_tokens": 64864026.0,
+      "step": 7990
+    },
+    {
+      "entropy": 0.22933962047100068,
+      "epoch": 4.511851015801354,
+      "grad_norm": 1.8534975051879883,
+      "learning_rate": 4.548600049299502e-06,
+      "loss": 0.1226,
+      "mean_token_accuracy": 0.9605401158332825,
+      "num_tokens": 64904483.0,
+      "step": 7995
+    },
+    {
+      "entropy": 0.22776411473751068,
+      "epoch": 4.514672686230249,
+      "grad_norm": 1.8297226428985596,
+      "learning_rate": 4.548059065817841e-06,
+      "loss": 0.1186,
+      "mean_token_accuracy": 0.96176518201828,
+      "num_tokens": 64945138.0,
+      "step": 8000
+    },
+    {
+      "epoch": 4.514672686230249,
+      "eval_entropy": 0.2636869251728058,
+      "eval_loss": 0.0795026570558548,
+      "eval_mean_token_accuracy": 0.9754695296287537,
+      "eval_num_tokens": 64945138.0,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 24.429,
+      "eval_steps_per_second": 6.107,
+      "step": 8000
+    },
+    {
+      "entropy": 0.21750448644161224,
+      "epoch": 4.517494356659142,
+      "grad_norm": 2.122471570968628,
+      "learning_rate": 4.547517802776042e-06,
+      "loss": 0.1224,
+      "mean_token_accuracy": 0.9609117388725281,
+      "num_tokens": 64985873.0,
+      "step": 8005
+    },
+    {
+      "entropy": 0.22921790778636933,
+      "epoch": 4.520316027088036,
+      "grad_norm": 2.1541335582733154,
+      "learning_rate": 4.546976260280435e-06,
+      "loss": 0.1131,
+      "mean_token_accuracy": 0.9636703491210937,
+      "num_tokens": 65026415.0,
+      "step": 8010
+    },
+    {
+      "entropy": 0.23894762694835664,
+      "epoch": 4.52313769751693,
+      "grad_norm": 2.3302664756774902,
+      "learning_rate": 4.546434438437408e-06,
+      "loss": 0.1196,
+      "mean_token_accuracy": 0.9631991147994995,
+      "num_tokens": 65067004.0,
+      "step": 8015
+    },
+    {
+      "entropy": 0.26045531034469604,
+      "epoch": 4.525959367945823,
+      "grad_norm": 2.2902634143829346,
+      "learning_rate": 4.5458923373534e-06,
+      "loss": 0.1442,
+      "mean_token_accuracy": 0.9546260356903076,
+      "num_tokens": 65107609.0,
+      "step": 8020
+    },
+    {
+      "entropy": 0.23849842548370362,
+      "epoch": 4.528781038374718,
+      "grad_norm": 2.1312549114227295,
+      "learning_rate": 4.545349957134908e-06,
+      "loss": 0.1253,
+      "mean_token_accuracy": 0.9602569699287414,
+      "num_tokens": 65148321.0,
+      "step": 8025
+    },
+    {
+      "entropy": 0.242272087931633,
+      "epoch": 4.531602708803612,
+      "grad_norm": 2.1353938579559326,
+      "learning_rate": 4.544807297888482e-06,
+      "loss": 0.1329,
+      "mean_token_accuracy": 0.9582337856292724,
+      "num_tokens": 65188859.0,
+      "step": 8030
+    },
+    {
+      "entropy": 0.21073226928710936,
+      "epoch": 4.534424379232505,
+      "grad_norm": 1.7557628154754639,
+      "learning_rate": 4.544264359720728e-06,
+      "loss": 0.1058,
+      "mean_token_accuracy": 0.9665800213813782,
+      "num_tokens": 65229611.0,
+      "step": 8035
+    },
+    {
+      "entropy": 0.2263183683156967,
+      "epoch": 4.5372460496614,
+      "grad_norm": 1.9420866966247559,
+      "learning_rate": 4.543721142738306e-06,
+      "loss": 0.1327,
+      "mean_token_accuracy": 0.9569009423255921,
+      "num_tokens": 65270007.0,
+      "step": 8040
+    },
+    {
+      "entropy": 0.24126074612140655,
+      "epoch": 4.540067720090294,
+      "grad_norm": 1.8423664569854736,
+      "learning_rate": 4.543177647047931e-06,
+      "loss": 0.1261,
+      "mean_token_accuracy": 0.9594087600708008,
+      "num_tokens": 65310681.0,
+      "step": 8045
+    },
+    {
+      "entropy": 0.2388361632823944,
+      "epoch": 4.542889390519187,
+      "grad_norm": 1.9569209814071655,
+      "learning_rate": 4.542633872756374e-06,
+      "loss": 0.1303,
+      "mean_token_accuracy": 0.9593725204467773,
+      "num_tokens": 65351285.0,
+      "step": 8050
+    },
+    {
+      "entropy": 0.23481654226779938,
+      "epoch": 4.545711060948081,
+      "grad_norm": 1.949885606765747,
+      "learning_rate": 4.542089819970456e-06,
+      "loss": 0.1192,
+      "mean_token_accuracy": 0.9611479878425598,
+      "num_tokens": 65391577.0,
+      "step": 8055
+    },
+    {
+      "entropy": 0.25715220272541045,
+      "epoch": 4.548532731376975,
+      "grad_norm": 2.203080892562866,
+      "learning_rate": 4.541545488797061e-06,
+      "loss": 0.1468,
+      "mean_token_accuracy": 0.9533620238304138,
+      "num_tokens": 65432197.0,
+      "step": 8060
+    },
+    {
+      "entropy": 0.23479729294776916,
+      "epoch": 4.551354401805869,
+      "grad_norm": 1.9314894676208496,
+      "learning_rate": 4.541000879343119e-06,
+      "loss": 0.1233,
+      "mean_token_accuracy": 0.9603962182998658,
+      "num_tokens": 65473100.0,
+      "step": 8065
+    },
+    {
+      "entropy": 0.23400575220584868,
+      "epoch": 4.554176072234763,
+      "grad_norm": 1.9281415939331055,
+      "learning_rate": 4.540455991715621e-06,
+      "loss": 0.139,
+      "mean_token_accuracy": 0.9566599488258362,
+      "num_tokens": 65513690.0,
+      "step": 8070
+    },
+    {
+      "entropy": 0.22714907824993133,
+      "epoch": 4.556997742663657,
+      "grad_norm": 1.8072905540466309,
+      "learning_rate": 4.539910826021609e-06,
+      "loss": 0.1232,
+      "mean_token_accuracy": 0.9606075048446655,
+      "num_tokens": 65554560.0,
+      "step": 8075
+    },
+    {
+      "entropy": 0.23093242347240447,
+      "epoch": 4.5598194130925505,
+      "grad_norm": 1.9328153133392334,
+      "learning_rate": 4.539365382368182e-06,
+      "loss": 0.1313,
+      "mean_token_accuracy": 0.9587170481681824,
+      "num_tokens": 65595278.0,
+      "step": 8080
+    },
+    {
+      "entropy": 0.2399729460477829,
+      "epoch": 4.562641083521445,
+      "grad_norm": 2.074904203414917,
+      "learning_rate": 4.5388196608624915e-06,
+      "loss": 0.126,
+      "mean_token_accuracy": 0.9598425269126892,
+      "num_tokens": 65635798.0,
+      "step": 8085
+    },
+    {
+      "entropy": 0.22223535776138306,
+      "epoch": 4.565462753950339,
+      "grad_norm": 1.8036304712295532,
+      "learning_rate": 4.538273661611744e-06,
+      "loss": 0.128,
+      "mean_token_accuracy": 0.960320234298706,
+      "num_tokens": 65676485.0,
+      "step": 8090
+    },
+    {
+      "entropy": 0.2333895742893219,
+      "epoch": 4.568284424379232,
+      "grad_norm": 2.1625099182128906,
+      "learning_rate": 4.537727384723203e-06,
+      "loss": 0.1396,
+      "mean_token_accuracy": 0.9549428701400757,
+      "num_tokens": 65717380.0,
+      "step": 8095
+    },
+    {
+      "entropy": 0.2488487184047699,
+      "epoch": 4.571106094808126,
+      "grad_norm": 2.21354079246521,
+      "learning_rate": 4.537180830304183e-06,
+      "loss": 0.1413,
+      "mean_token_accuracy": 0.9569279074668884,
+      "num_tokens": 65757893.0,
+      "step": 8100
+    },
+    {
+      "entropy": 0.22758035659790038,
+      "epoch": 4.57392776523702,
+      "grad_norm": 2.3330676555633545,
+      "learning_rate": 4.536633998462055e-06,
+      "loss": 0.1163,
+      "mean_token_accuracy": 0.9631718873977662,
+      "num_tokens": 65798444.0,
+      "step": 8105
+    },
+    {
+      "entropy": 0.235689178109169,
+      "epoch": 4.576749435665914,
+      "grad_norm": 2.2506041526794434,
+      "learning_rate": 4.536086889304246e-06,
+      "loss": 0.134,
+      "mean_token_accuracy": 0.9576963305473327,
+      "num_tokens": 65839127.0,
+      "step": 8110
+    },
+    {
+      "entropy": 0.2473171055316925,
+      "epoch": 4.579571106094808,
+      "grad_norm": 2.3321518898010254,
+      "learning_rate": 4.535539502938233e-06,
+      "loss": 0.128,
+      "mean_token_accuracy": 0.9609445095062256,
+      "num_tokens": 65879975.0,
+      "step": 8115
+    },
+    {
+      "entropy": 0.25173512697219846,
+      "epoch": 4.582392776523702,
+      "grad_norm": 2.5491793155670166,
+      "learning_rate": 4.534991839471551e-06,
+      "loss": 0.1372,
+      "mean_token_accuracy": 0.9561426758766174,
+      "num_tokens": 65920672.0,
+      "step": 8120
+    },
+    {
+      "entropy": 0.23555795848369598,
+      "epoch": 4.585214446952596,
+      "grad_norm": 2.0215837955474854,
+      "learning_rate": 4.534443899011789e-06,
+      "loss": 0.1274,
+      "mean_token_accuracy": 0.9595631241798401,
+      "num_tokens": 65960881.0,
+      "step": 8125
+    },
+    {
+      "entropy": 0.2525820404291153,
+      "epoch": 4.58803611738149,
+      "grad_norm": 2.073427438735962,
+      "learning_rate": 4.533895681666591e-06,
+      "loss": 0.1387,
+      "mean_token_accuracy": 0.9559617161750793,
+      "num_tokens": 66001503.0,
+      "step": 8130
+    },
+    {
+      "entropy": 0.23859679996967315,
+      "epoch": 4.590857787810384,
+      "grad_norm": 1.9522126913070679,
+      "learning_rate": 4.533347187543652e-06,
+      "loss": 0.125,
+      "mean_token_accuracy": 0.9605787754058838,
+      "num_tokens": 66042200.0,
+      "step": 8135
+    },
+    {
+      "entropy": 0.23188299536705018,
+      "epoch": 4.5936794582392775,
+      "grad_norm": 1.8732774257659912,
+      "learning_rate": 4.5327984167507255e-06,
+      "loss": 0.1232,
+      "mean_token_accuracy": 0.9610357999801635,
+      "num_tokens": 66082649.0,
+      "step": 8140
+    },
+    {
+      "entropy": 0.2427914947271347,
+      "epoch": 4.596501128668171,
+      "grad_norm": 2.2590503692626953,
+      "learning_rate": 4.532249369395616e-06,
+      "loss": 0.1242,
+      "mean_token_accuracy": 0.9602537393569947,
+      "num_tokens": 66123110.0,
+      "step": 8145
+    },
+    {
+      "entropy": 0.24748845994472504,
+      "epoch": 4.599322799097066,
+      "grad_norm": 2.055881977081299,
+      "learning_rate": 4.531700045586187e-06,
+      "loss": 0.1205,
+      "mean_token_accuracy": 0.9622966647148132,
+      "num_tokens": 66163660.0,
+      "step": 8150
+    },
+    {
+      "entropy": 0.23012515604496003,
+      "epoch": 4.6021444695259595,
+      "grad_norm": 1.8205794095993042,
+      "learning_rate": 4.53115044543035e-06,
+      "loss": 0.1247,
+      "mean_token_accuracy": 0.9605541706085206,
+      "num_tokens": 66204458.0,
+      "step": 8155
+    },
+    {
+      "entropy": 0.2356725037097931,
+      "epoch": 4.604966139954853,
+      "grad_norm": 2.3245763778686523,
+      "learning_rate": 4.530600569036075e-06,
+      "loss": 0.1337,
+      "mean_token_accuracy": 0.9578644394874573,
+      "num_tokens": 66245231.0,
+      "step": 8160
+    },
+    {
+      "entropy": 0.2583352416753769,
+      "epoch": 4.607787810383747,
+      "grad_norm": 2.2044198513031006,
+      "learning_rate": 4.530050416511386e-06,
+      "loss": 0.1491,
+      "mean_token_accuracy": 0.9534392952919006,
+      "num_tokens": 66285856.0,
+      "step": 8165
+    },
+    {
+      "entropy": 0.2439739376306534,
+      "epoch": 4.610609480812641,
+      "grad_norm": 2.181988477706909,
+      "learning_rate": 4.529499987964359e-06,
+      "loss": 0.1372,
+      "mean_token_accuracy": 0.9561636447906494,
+      "num_tokens": 66326451.0,
+      "step": 8170
+    },
+    {
+      "entropy": 0.24100883007049562,
+      "epoch": 4.613431151241535,
+      "grad_norm": 2.0313637256622314,
+      "learning_rate": 4.5289492835031275e-06,
+      "loss": 0.1421,
+      "mean_token_accuracy": 0.9549835324287415,
+      "num_tokens": 66367157.0,
+      "step": 8175
+    },
+    {
+      "entropy": 0.22571427822113038,
+      "epoch": 4.616252821670429,
+      "grad_norm": 1.996917963027954,
+      "learning_rate": 4.528398303235877e-06,
+      "loss": 0.1191,
+      "mean_token_accuracy": 0.9621818661689758,
+      "num_tokens": 66407802.0,
+      "step": 8180
+    },
+    {
+      "entropy": 0.23621368408203125,
+      "epoch": 4.6190744920993225,
+      "grad_norm": 1.7542959451675415,
+      "learning_rate": 4.527847047270847e-06,
+      "loss": 0.1379,
+      "mean_token_accuracy": 0.9554514408111572,
+      "num_tokens": 66448374.0,
+      "step": 8185
+    },
+    {
+      "entropy": 0.2359769821166992,
+      "epoch": 4.621896162528216,
+      "grad_norm": 2.465397834777832,
+      "learning_rate": 4.527295515716332e-06,
+      "loss": 0.128,
+      "mean_token_accuracy": 0.9583778262138367,
+      "num_tokens": 66488251.0,
+      "step": 8190
+    },
+    {
+      "entropy": 0.22702213227748871,
+      "epoch": 4.624717832957111,
+      "grad_norm": 2.1334855556488037,
+      "learning_rate": 4.526743708680681e-06,
+      "loss": 0.1218,
+      "mean_token_accuracy": 0.9614168047904968,
+      "num_tokens": 66528989.0,
+      "step": 8195
+    },
+    {
+      "entropy": 0.2207847625017166,
+      "epoch": 4.6275395033860045,
+      "grad_norm": 2.2564754486083984,
+      "learning_rate": 4.526191626272297e-06,
+      "loss": 0.1245,
+      "mean_token_accuracy": 0.9604014396667481,
+      "num_tokens": 66569555.0,
+      "step": 8200
+    },
+    {
+      "entropy": 0.24324011504650117,
+      "epoch": 4.630361173814898,
+      "grad_norm": 1.8627599477767944,
+      "learning_rate": 4.525639268599635e-06,
+      "loss": 0.1284,
+      "mean_token_accuracy": 0.9589541673660278,
+      "num_tokens": 66610160.0,
+      "step": 8205
+    },
+    {
+      "entropy": 0.2532497227191925,
+      "epoch": 4.633182844243793,
+      "grad_norm": 2.4542059898376465,
+      "learning_rate": 4.5250866357712066e-06,
+      "loss": 0.1418,
+      "mean_token_accuracy": 0.9555447816848754,
+      "num_tokens": 66650856.0,
+      "step": 8210
+    },
+    {
+      "entropy": 0.23340786695480348,
+      "epoch": 4.6360045146726865,
+      "grad_norm": 2.2496676445007324,
+      "learning_rate": 4.524533727895577e-06,
+      "loss": 0.118,
+      "mean_token_accuracy": 0.9636736989021302,
+      "num_tokens": 66691192.0,
+      "step": 8215
+    },
+    {
+      "entropy": 0.24236354231834412,
+      "epoch": 4.63882618510158,
+      "grad_norm": 1.9971568584442139,
+      "learning_rate": 4.5239805450813646e-06,
+      "loss": 0.134,
+      "mean_token_accuracy": 0.9569501161575318,
+      "num_tokens": 66731579.0,
+      "step": 8220
+    },
+    {
+      "entropy": 0.23493029475212096,
+      "epoch": 4.641647855530474,
+      "grad_norm": 2.242898941040039,
+      "learning_rate": 4.523427087437241e-06,
+      "loss": 0.1212,
+      "mean_token_accuracy": 0.962970244884491,
+      "num_tokens": 66771885.0,
+      "step": 8225
+    },
+    {
+      "entropy": 0.22397561073303224,
+      "epoch": 4.644469525959368,
+      "grad_norm": 1.9795989990234375,
+      "learning_rate": 4.522873355071936e-06,
+      "loss": 0.1089,
+      "mean_token_accuracy": 0.9651905417442321,
+      "num_tokens": 66812487.0,
+      "step": 8230
+    },
+    {
+      "entropy": 0.21759993433952332,
+      "epoch": 4.647291196388262,
+      "grad_norm": 5.298044204711914,
+      "learning_rate": 4.5223193480942275e-06,
+      "loss": 0.1171,
+      "mean_token_accuracy": 0.9617223024368287,
+      "num_tokens": 66853201.0,
+      "step": 8235
+    },
+    {
+      "entropy": 0.24825019538402557,
+      "epoch": 4.650112866817156,
+      "grad_norm": 2.4175469875335693,
+      "learning_rate": 4.521765066612952e-06,
+      "loss": 0.1532,
+      "mean_token_accuracy": 0.9531420230865478,
+      "num_tokens": 66893997.0,
+      "step": 8240
+    },
+    {
+      "entropy": 0.23018406629562377,
+      "epoch": 4.6529345372460496,
+      "grad_norm": 1.9281044006347656,
+      "learning_rate": 4.521210510736998e-06,
+      "loss": 0.1235,
+      "mean_token_accuracy": 0.9599860429763794,
+      "num_tokens": 66934876.0,
+      "step": 8245
+    },
+    {
+      "entropy": 0.23870559334754943,
+      "epoch": 4.655756207674943,
+      "grad_norm": 1.8871232271194458,
+      "learning_rate": 4.520655680575306e-06,
+      "loss": 0.1159,
+      "mean_token_accuracy": 0.9634807467460632,
+      "num_tokens": 66975475.0,
+      "step": 8250
+    },
+    {
+      "entropy": 0.244593945145607,
+      "epoch": 4.658577878103838,
+      "grad_norm": 2.5723750591278076,
+      "learning_rate": 4.520100576236877e-06,
+      "loss": 0.1325,
+      "mean_token_accuracy": 0.9575012445449829,
+      "num_tokens": 67016014.0,
+      "step": 8255
+    },
+    {
+      "entropy": 0.23622966110706328,
+      "epoch": 4.6613995485327315,
+      "grad_norm": 2.5368049144744873,
+      "learning_rate": 4.5195451978307556e-06,
+      "loss": 0.131,
+      "mean_token_accuracy": 0.9596938133239746,
+      "num_tokens": 67056774.0,
+      "step": 8260
+    },
+    {
+      "entropy": 0.22385245859622954,
+      "epoch": 4.664221218961625,
+      "grad_norm": 2.3085649013519287,
+      "learning_rate": 4.51898954546605e-06,
+      "loss": 0.1206,
+      "mean_token_accuracy": 0.9616567969322205,
+      "num_tokens": 67096578.0,
+      "step": 8265
+    },
+    {
+      "entropy": 0.24655192196369172,
+      "epoch": 4.667042889390519,
+      "grad_norm": 2.2656362056732178,
+      "learning_rate": 4.518433619251918e-06,
+      "loss": 0.1407,
+      "mean_token_accuracy": 0.9564512014389038,
+      "num_tokens": 67137138.0,
+      "step": 8270
+    },
+    {
+      "entropy": 0.24504003524780274,
+      "epoch": 4.669864559819413,
+      "grad_norm": 1.9192149639129639,
+      "learning_rate": 4.5178774192975685e-06,
+      "loss": 0.1408,
+      "mean_token_accuracy": 0.9554814100265503,
+      "num_tokens": 67176712.0,
+      "step": 8275
+    },
+    {
+      "entropy": 0.22429890930652618,
+      "epoch": 4.672686230248307,
+      "grad_norm": 2.07617449760437,
+      "learning_rate": 4.51732094571227e-06,
+      "loss": 0.1242,
+      "mean_token_accuracy": 0.9604406833648682,
+      "num_tokens": 67217311.0,
+      "step": 8280
+    },
+    {
+      "entropy": 0.250271201133728,
+      "epoch": 4.675507900677201,
+      "grad_norm": 1.9916068315505981,
+      "learning_rate": 4.51676419860534e-06,
+      "loss": 0.1517,
+      "mean_token_accuracy": 0.9521410703659058,
+      "num_tokens": 67257957.0,
+      "step": 8285
+    },
+    {
+      "entropy": 0.243993404507637,
+      "epoch": 4.678329571106095,
+      "grad_norm": 2.026092290878296,
+      "learning_rate": 4.516207178086153e-06,
+      "loss": 0.121,
+      "mean_token_accuracy": 0.962063193321228,
+      "num_tokens": 67298719.0,
+      "step": 8290
+    },
+    {
+      "entropy": 0.23810403943061828,
+      "epoch": 4.681151241534989,
+      "grad_norm": 1.7401542663574219,
+      "learning_rate": 4.515649884264135e-06,
+      "loss": 0.1302,
+      "mean_token_accuracy": 0.9585550904273987,
+      "num_tokens": 67339410.0,
+      "step": 8295
+    },
+    {
+      "entropy": 0.24030295610427857,
+      "epoch": 4.683972911963883,
+      "grad_norm": 1.8887089490890503,
+      "learning_rate": 4.515092317248766e-06,
+      "loss": 0.1216,
+      "mean_token_accuracy": 0.9610766172409058,
+      "num_tokens": 67379808.0,
+      "step": 8300
+    },
+    {
+      "entropy": 0.2178757071495056,
+      "epoch": 4.686794582392777,
+      "grad_norm": 2.0333640575408936,
+      "learning_rate": 4.514534477149581e-06,
+      "loss": 0.1048,
+      "mean_token_accuracy": 0.9664766073226929,
+      "num_tokens": 67420340.0,
+      "step": 8305
+    },
+    {
+      "entropy": 0.23881784081459045,
+      "epoch": 4.68961625282167,
+      "grad_norm": 2.146456480026245,
+      "learning_rate": 4.513976364076167e-06,
+      "loss": 0.1213,
+      "mean_token_accuracy": 0.9607988595962524,
+      "num_tokens": 67461118.0,
+      "step": 8310
+    },
+    {
+      "entropy": 0.22699067294597625,
+      "epoch": 4.692437923250564,
+      "grad_norm": 2.0983293056488037,
+      "learning_rate": 4.513417978138166e-06,
+      "loss": 0.1382,
+      "mean_token_accuracy": 0.9564812779426575,
+      "num_tokens": 67501770.0,
+      "step": 8315
+    },
+    {
+      "entropy": 0.21569800674915313,
+      "epoch": 4.6952595936794586,
+      "grad_norm": 1.7150790691375732,
+      "learning_rate": 4.5128593194452725e-06,
+      "loss": 0.124,
+      "mean_token_accuracy": 0.9598898530006409,
+      "num_tokens": 67542371.0,
+      "step": 8320
+    },
+    {
+      "entropy": 0.2471790909767151,
+      "epoch": 4.698081264108352,
+      "grad_norm": 2.265134811401367,
+      "learning_rate": 4.5123003881072345e-06,
+      "loss": 0.1441,
+      "mean_token_accuracy": 0.9535491704940796,
+      "num_tokens": 67582960.0,
+      "step": 8325
+    },
+    {
+      "entropy": 0.22231938540935517,
+      "epoch": 4.700902934537246,
+      "grad_norm": 1.7570245265960693,
+      "learning_rate": 4.511741184233856e-06,
+      "loss": 0.1158,
+      "mean_token_accuracy": 0.9630581736564636,
+      "num_tokens": 67623636.0,
+      "step": 8330
+    },
+    {
+      "entropy": 0.24155616462230683,
+      "epoch": 4.70372460496614,
+      "grad_norm": 2.0228629112243652,
+      "learning_rate": 4.511181707934992e-06,
+      "loss": 0.1354,
+      "mean_token_accuracy": 0.9561177611351013,
+      "num_tokens": 67664179.0,
+      "step": 8335
+    },
+    {
+      "entropy": 0.23686327040195465,
+      "epoch": 4.706546275395034,
+      "grad_norm": 2.0804860591888428,
+      "learning_rate": 4.5106219593205505e-06,
+      "loss": 0.1402,
+      "mean_token_accuracy": 0.9536548256874084,
+      "num_tokens": 67704834.0,
+      "step": 8340
+    },
+    {
+      "entropy": 0.22247822284698487,
+      "epoch": 4.709367945823928,
+      "grad_norm": 2.316359519958496,
+      "learning_rate": 4.510061938500495e-06,
+      "loss": 0.1317,
+      "mean_token_accuracy": 0.9575009226799012,
+      "num_tokens": 67745324.0,
+      "step": 8345
+    },
+    {
+      "entropy": 0.25336918532848357,
+      "epoch": 4.712189616252822,
+      "grad_norm": 1.8472719192504883,
+      "learning_rate": 4.509501645584842e-06,
+      "loss": 0.1437,
+      "mean_token_accuracy": 0.9530521392822265,
+      "num_tokens": 67786074.0,
+      "step": 8350
+    },
+    {
+      "entropy": 0.23464283645153045,
+      "epoch": 4.715011286681715,
+      "grad_norm": 2.232201337814331,
+      "learning_rate": 4.508941080683661e-06,
+      "loss": 0.1304,
+      "mean_token_accuracy": 0.9586957097053528,
+      "num_tokens": 67826602.0,
+      "step": 8355
+    },
+    {
+      "entropy": 0.2284877747297287,
+      "epoch": 4.717832957110609,
+      "grad_norm": 2.0694456100463867,
+      "learning_rate": 4.508380243907074e-06,
+      "loss": 0.1335,
+      "mean_token_accuracy": 0.9572539687156677,
+      "num_tokens": 67867168.0,
+      "step": 8360
+    },
+    {
+      "entropy": 0.23733165860176086,
+      "epoch": 4.720654627539504,
+      "grad_norm": 1.836682915687561,
+      "learning_rate": 4.5078191353652575e-06,
+      "loss": 0.1427,
+      "mean_token_accuracy": 0.9549849033355713,
+      "num_tokens": 67907809.0,
+      "step": 8365
+    },
+    {
+      "entropy": 0.23421334028244017,
+      "epoch": 4.723476297968397,
+      "grad_norm": 2.212164878845215,
+      "learning_rate": 4.507257755168444e-06,
+      "loss": 0.1401,
+      "mean_token_accuracy": 0.9548245668411255,
+      "num_tokens": 67948509.0,
+      "step": 8370
+    },
+    {
+      "entropy": 0.2503542214632034,
+      "epoch": 4.726297968397291,
+      "grad_norm": 2.0967366695404053,
+      "learning_rate": 4.506696103426914e-06,
+      "loss": 0.1403,
+      "mean_token_accuracy": 0.9554270505905151,
+      "num_tokens": 67989217.0,
+      "step": 8375
+    },
+    {
+      "entropy": 0.22551278471946717,
+      "epoch": 4.729119638826186,
+      "grad_norm": 2.0582823753356934,
+      "learning_rate": 4.506134180251005e-06,
+      "loss": 0.1292,
+      "mean_token_accuracy": 0.9604162216186524,
+      "num_tokens": 68029546.0,
+      "step": 8380
+    },
+    {
+      "entropy": 0.23066943287849426,
+      "epoch": 4.731941309255079,
+      "grad_norm": 1.9239495992660522,
+      "learning_rate": 4.5055719857511065e-06,
+      "loss": 0.1443,
+      "mean_token_accuracy": 0.9540068626403808,
+      "num_tokens": 68070249.0,
+      "step": 8385
+    },
+    {
+      "entropy": 0.24158512353897094,
+      "epoch": 4.734762979683973,
+      "grad_norm": 2.0336618423461914,
+      "learning_rate": 4.505009520037662e-06,
+      "loss": 0.1339,
+      "mean_token_accuracy": 0.9556556820869446,
+      "num_tokens": 68110991.0,
+      "step": 8390
+    },
+    {
+      "entropy": 0.24518156945705413,
+      "epoch": 4.737584650112867,
+      "grad_norm": 2.0292651653289795,
+      "learning_rate": 4.504446783221168e-06,
+      "loss": 0.1303,
+      "mean_token_accuracy": 0.9584499716758728,
+      "num_tokens": 68151625.0,
+      "step": 8395
+    },
+    {
+      "entropy": 0.23519628643989562,
+      "epoch": 4.74040632054176,
+      "grad_norm": 2.051119327545166,
+      "learning_rate": 4.503883775412174e-06,
+      "loss": 0.1334,
+      "mean_token_accuracy": 0.9583625435829163,
+      "num_tokens": 68192104.0,
+      "step": 8400
+    },
+    {
+      "entropy": 0.2450340747833252,
+      "epoch": 4.743227990970655,
+      "grad_norm": 2.1468701362609863,
+      "learning_rate": 4.503320496721283e-06,
+      "loss": 0.1437,
+      "mean_token_accuracy": 0.9554325222969056,
+      "num_tokens": 68232969.0,
+      "step": 8405
+    },
+    {
+      "entropy": 0.22379915416240692,
+      "epoch": 4.746049661399549,
+      "grad_norm": 2.243983268737793,
+      "learning_rate": 4.5027569472591515e-06,
+      "loss": 0.1262,
+      "mean_token_accuracy": 0.9602833390235901,
+      "num_tokens": 68273416.0,
+      "step": 8410
+    },
+    {
+      "entropy": 0.23107292354106904,
+      "epoch": 4.748871331828442,
+      "grad_norm": 1.9481990337371826,
+      "learning_rate": 4.502193127136489e-06,
+      "loss": 0.1195,
+      "mean_token_accuracy": 0.9613822102546692,
+      "num_tokens": 68314053.0,
+      "step": 8415
+    },
+    {
+      "entropy": 0.24034847021102906,
+      "epoch": 4.751693002257336,
+      "grad_norm": 2.4841561317443848,
+      "learning_rate": 4.501629036464057e-06,
+      "loss": 0.1295,
+      "mean_token_accuracy": 0.9587523937225342,
+      "num_tokens": 68354766.0,
+      "step": 8420
+    },
+    {
+      "entropy": 0.23407194018363953,
+      "epoch": 4.754514672686231,
+      "grad_norm": 2.2087182998657227,
+      "learning_rate": 4.501064675352671e-06,
+      "loss": 0.125,
+      "mean_token_accuracy": 0.9598301172256469,
+      "num_tokens": 68395431.0,
+      "step": 8425
+    },
+    {
+      "entropy": 0.25628364980220797,
+      "epoch": 4.757336343115124,
+      "grad_norm": 2.0910379886627197,
+      "learning_rate": 4.500500043913203e-06,
+      "loss": 0.1406,
+      "mean_token_accuracy": 0.9561483979225158,
+      "num_tokens": 68436109.0,
+      "step": 8430
+    },
+    {
+      "entropy": 0.26322224736213684,
+      "epoch": 4.760158013544018,
+      "grad_norm": 2.235393762588501,
+      "learning_rate": 4.499935142256571e-06,
+      "loss": 0.1605,
+      "mean_token_accuracy": 0.9494641661643982,
+      "num_tokens": 68476762.0,
+      "step": 8435
+    },
+    {
+      "entropy": 0.24928564131259917,
+      "epoch": 4.762979683972912,
+      "grad_norm": 2.2493717670440674,
+      "learning_rate": 4.499369970493751e-06,
+      "loss": 0.1457,
+      "mean_token_accuracy": 0.9544100880622863,
+      "num_tokens": 68517508.0,
+      "step": 8440
+    },
+    {
+      "entropy": 0.23149662911891938,
+      "epoch": 4.765801354401805,
+      "grad_norm": 1.8034436702728271,
+      "learning_rate": 4.498804528735773e-06,
+      "loss": 0.1316,
+      "mean_token_accuracy": 0.958375072479248,
+      "num_tokens": 68558346.0,
+      "step": 8445
+    },
+    {
+      "entropy": 0.2590792328119278,
+      "epoch": 4.7686230248307,
+      "grad_norm": 1.9933143854141235,
+      "learning_rate": 4.498238817093717e-06,
+      "loss": 0.1211,
+      "mean_token_accuracy": 0.9613798260688782,
+      "num_tokens": 68598995.0,
+      "step": 8450
+    },
+    {
+      "entropy": 0.21191737055778503,
+      "epoch": 4.771444695259594,
+      "grad_norm": 1.8412405252456665,
+      "learning_rate": 4.497672835678716e-06,
+      "loss": 0.1165,
+      "mean_token_accuracy": 0.9624577045440674,
+      "num_tokens": 68639676.0,
+      "step": 8455
+    },
+    {
+      "entropy": 0.2465408831834793,
+      "epoch": 4.774266365688487,
+      "grad_norm": 2.141119956970215,
+      "learning_rate": 4.497106584601957e-06,
+      "loss": 0.1499,
+      "mean_token_accuracy": 0.9527180552482605,
+      "num_tokens": 68680313.0,
+      "step": 8460
+    },
+    {
+      "entropy": 0.2483223021030426,
+      "epoch": 4.777088036117382,
+      "grad_norm": 2.286459445953369,
+      "learning_rate": 4.496540063974683e-06,
+      "loss": 0.1449,
+      "mean_token_accuracy": 0.9531118035316467,
+      "num_tokens": 68721184.0,
+      "step": 8465
+    },
+    {
+      "entropy": 0.23041402101516723,
+      "epoch": 4.779909706546276,
+      "grad_norm": 1.8058642148971558,
+      "learning_rate": 4.495973273908184e-06,
+      "loss": 0.1271,
+      "mean_token_accuracy": 0.9595191001892089,
+      "num_tokens": 68761833.0,
+      "step": 8470
+    },
+    {
+      "entropy": 0.24951161742210387,
+      "epoch": 4.782731376975169,
+      "grad_norm": 2.057999849319458,
+      "learning_rate": 4.495406214513807e-06,
+      "loss": 0.1334,
+      "mean_token_accuracy": 0.9589525699615479,
+      "num_tokens": 68802489.0,
+      "step": 8475
+    },
+    {
+      "entropy": 0.2543626993894577,
+      "epoch": 4.785553047404063,
+      "grad_norm": 2.1733288764953613,
+      "learning_rate": 4.494838885902952e-06,
+      "loss": 0.1569,
+      "mean_token_accuracy": 0.9508596181869506,
+      "num_tokens": 68843038.0,
+      "step": 8480
+    },
+    {
+      "entropy": 0.22672601640224457,
+      "epoch": 4.788374717832957,
+      "grad_norm": 3.0629308223724365,
+      "learning_rate": 4.4942712881870684e-06,
+      "loss": 0.13,
+      "mean_token_accuracy": 0.9586154460906983,
+      "num_tokens": 68883645.0,
+      "step": 8485
+    },
+    {
+      "entropy": 0.24403004348278046,
+      "epoch": 4.791196388261851,
+      "grad_norm": 2.2074005603790283,
+      "learning_rate": 4.493703421477663e-06,
+      "loss": 0.1449,
+      "mean_token_accuracy": 0.9540403723716736,
+      "num_tokens": 68924239.0,
+      "step": 8490
+    },
+    {
+      "entropy": 0.2356546252965927,
+      "epoch": 4.794018058690745,
+      "grad_norm": 1.874244213104248,
+      "learning_rate": 4.493135285886293e-06,
+      "loss": 0.1392,
+      "mean_token_accuracy": 0.9564555406570434,
+      "num_tokens": 68964945.0,
+      "step": 8495
+    },
+    {
+      "entropy": 0.24922839403152466,
+      "epoch": 4.796839729119639,
+      "grad_norm": 2.270205020904541,
+      "learning_rate": 4.492566881524568e-06,
+      "loss": 0.1371,
+      "mean_token_accuracy": 0.9573128700256348,
+      "num_tokens": 69005804.0,
+      "step": 8500
+    },
+    {
+      "epoch": 4.796839729119639,
+      "eval_entropy": 0.25907859206199646,
+      "eval_loss": 0.06230160593986511,
+      "eval_mean_token_accuracy": 0.9831352829933167,
+      "eval_num_tokens": 69005804.0,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 24.392,
+      "eval_steps_per_second": 6.098,
+      "step": 8500
+    },
+    {
+      "entropy": 0.2256260484457016,
+      "epoch": 4.799661399548532,
+      "grad_norm": 2.126230001449585,
+      "learning_rate": 4.491998208504151e-06,
+      "loss": 0.1214,
+      "mean_token_accuracy": 0.9609059691429138,
+      "num_tokens": 69045948.0,
+      "step": 8505
+    },
+    {
+      "entropy": 0.2439708322286606,
+      "epoch": 4.802483069977427,
+      "grad_norm": 1.7869462966918945,
+      "learning_rate": 4.491429266936759e-06,
+      "loss": 0.1453,
+      "mean_token_accuracy": 0.9530874967575074,
+      "num_tokens": 69086575.0,
+      "step": 8510
+    },
+    {
+      "entropy": 0.23215862214565278,
+      "epoch": 4.805304740406321,
+      "grad_norm": 1.871141791343689,
+      "learning_rate": 4.490860056934158e-06,
+      "loss": 0.1164,
+      "mean_token_accuracy": 0.9630860209465026,
+      "num_tokens": 69127332.0,
+      "step": 8515
+    },
+    {
+      "entropy": 0.22940534651279448,
+      "epoch": 4.808126410835214,
+      "grad_norm": 1.7535682916641235,
+      "learning_rate": 4.490290578608173e-06,
+      "loss": 0.1133,
+      "mean_token_accuracy": 0.9624628305435181,
+      "num_tokens": 69167989.0,
+      "step": 8520
+    },
+    {
+      "entropy": 0.22979909479618071,
+      "epoch": 4.810948081264108,
+      "grad_norm": 2.0037307739257812,
+      "learning_rate": 4.489720832070676e-06,
+      "loss": 0.1249,
+      "mean_token_accuracy": 0.9602430582046508,
+      "num_tokens": 69208619.0,
+      "step": 8525
+    },
+    {
+      "entropy": 0.24260205030441284,
+      "epoch": 4.813769751693002,
+      "grad_norm": 2.209491491317749,
+      "learning_rate": 4.489150817433594e-06,
+      "loss": 0.1345,
+      "mean_token_accuracy": 0.9570428252220153,
+      "num_tokens": 69249329.0,
+      "step": 8530
+    },
+    {
+      "entropy": 0.22956629991531372,
+      "epoch": 4.816591422121896,
+      "grad_norm": 2.2682173252105713,
+      "learning_rate": 4.488580534808908e-06,
+      "loss": 0.1358,
+      "mean_token_accuracy": 0.9560045003890991,
+      "num_tokens": 69289846.0,
+      "step": 8535
+    },
+    {
+      "entropy": 0.2220306247472763,
+      "epoch": 4.81941309255079,
+      "grad_norm": 1.8845446109771729,
+      "learning_rate": 4.488009984308647e-06,
+      "loss": 0.1276,
+      "mean_token_accuracy": 0.9602849364280701,
+      "num_tokens": 69330609.0,
+      "step": 8540
+    },
+    {
+      "entropy": 0.25677892565727234,
+      "epoch": 4.822234762979684,
+      "grad_norm": 2.3458547592163086,
+      "learning_rate": 4.487439166044898e-06,
+      "loss": 0.1333,
+      "mean_token_accuracy": 0.9575768232345581,
+      "num_tokens": 69370952.0,
+      "step": 8545
+    },
+    {
+      "entropy": 0.2511105537414551,
+      "epoch": 4.825056433408578,
+      "grad_norm": 2.185471773147583,
+      "learning_rate": 4.486868080129797e-06,
+      "loss": 0.1403,
+      "mean_token_accuracy": 0.9561097264289856,
+      "num_tokens": 69411446.0,
+      "step": 8550
+    },
+    {
+      "entropy": 0.2550018668174744,
+      "epoch": 4.827878103837472,
+      "grad_norm": 2.0615882873535156,
+      "learning_rate": 4.486296726675535e-06,
+      "loss": 0.1344,
+      "mean_token_accuracy": 0.957464849948883,
+      "num_tokens": 69452055.0,
+      "step": 8555
+    },
+    {
+      "entropy": 0.2429851323366165,
+      "epoch": 4.830699774266366,
+      "grad_norm": 2.123836040496826,
+      "learning_rate": 4.485725105794354e-06,
+      "loss": 0.1285,
+      "mean_token_accuracy": 0.9590805530548095,
+      "num_tokens": 69492630.0,
+      "step": 8560
+    },
+    {
+      "entropy": 0.23657283782958985,
+      "epoch": 4.8335214446952595,
+      "grad_norm": 2.050199508666992,
+      "learning_rate": 4.48515321759855e-06,
+      "loss": 0.1305,
+      "mean_token_accuracy": 0.9578537583351135,
+      "num_tokens": 69533313.0,
+      "step": 8565
+    },
+    {
+      "entropy": 0.23308176994323732,
+      "epoch": 4.836343115124153,
+      "grad_norm": 1.9955555200576782,
+      "learning_rate": 4.4845810622004685e-06,
+      "loss": 0.1443,
+      "mean_token_accuracy": 0.9529533624649048,
+      "num_tokens": 69573940.0,
+      "step": 8570
+    },
+    {
+      "entropy": 0.23426572382450103,
+      "epoch": 4.839164785553048,
+      "grad_norm": 1.7304738759994507,
+      "learning_rate": 4.484008639712511e-06,
+      "loss": 0.1191,
+      "mean_token_accuracy": 0.9620694041252136,
+      "num_tokens": 69614517.0,
+      "step": 8575
+    },
+    {
+      "entropy": 0.2388251841068268,
+      "epoch": 4.841986455981941,
+      "grad_norm": 2.242614269256592,
+      "learning_rate": 4.48343595024713e-06,
+      "loss": 0.138,
+      "mean_token_accuracy": 0.9573504686355591,
+      "num_tokens": 69654972.0,
+      "step": 8580
+    },
+    {
+      "entropy": 0.2321117788553238,
+      "epoch": 4.844808126410835,
+      "grad_norm": 2.4217350482940674,
+      "learning_rate": 4.482862993916829e-06,
+      "loss": 0.1388,
+      "mean_token_accuracy": 0.95455002784729,
+      "num_tokens": 69695607.0,
+      "step": 8585
+    },
+    {
+      "entropy": 0.22771554291248322,
+      "epoch": 4.847629796839729,
+      "grad_norm": 1.7584290504455566,
+      "learning_rate": 4.482289770834168e-06,
+      "loss": 0.1239,
+      "mean_token_accuracy": 0.9602065324783325,
+      "num_tokens": 69736187.0,
+      "step": 8590
+    },
+    {
+      "entropy": 0.22587930858135224,
+      "epoch": 4.850451467268623,
+      "grad_norm": 2.119394302368164,
+      "learning_rate": 4.481716281111753e-06,
+      "loss": 0.1207,
+      "mean_token_accuracy": 0.9609213948249817,
+      "num_tokens": 69777050.0,
+      "step": 8595
+    },
+    {
+      "entropy": 0.23252309560775758,
+      "epoch": 4.853273137697517,
+      "grad_norm": 2.079322576522827,
+      "learning_rate": 4.481142524862249e-06,
+      "loss": 0.1269,
+      "mean_token_accuracy": 0.9575224757194519,
+      "num_tokens": 69817801.0,
+      "step": 8600
+    },
+    {
+      "entropy": 0.2291330635547638,
+      "epoch": 4.856094808126411,
+      "grad_norm": 2.186591148376465,
+      "learning_rate": 4.48056850219837e-06,
+      "loss": 0.1236,
+      "mean_token_accuracy": 0.9605337023735047,
+      "num_tokens": 69858177.0,
+      "step": 8605
+    },
+    {
+      "entropy": 0.22314091324806212,
+      "epoch": 4.8589164785553045,
+      "grad_norm": 1.9909532070159912,
+      "learning_rate": 4.479994213232882e-06,
+      "loss": 0.1238,
+      "mean_token_accuracy": 0.9604300022125244,
+      "num_tokens": 69898890.0,
+      "step": 8610
+    },
+    {
+      "entropy": 0.2294975697994232,
+      "epoch": 4.861738148984198,
+      "grad_norm": 2.1407694816589355,
+      "learning_rate": 4.479419658078606e-06,
+      "loss": 0.1272,
+      "mean_token_accuracy": 0.9603155970573425,
+      "num_tokens": 69939533.0,
+      "step": 8615
+    },
+    {
+      "entropy": 0.23732683360576629,
+      "epoch": 4.864559819413093,
+      "grad_norm": 2.215397357940674,
+      "learning_rate": 4.478844836848411e-06,
+      "loss": 0.1376,
+      "mean_token_accuracy": 0.9563262820243835,
+      "num_tokens": 69980394.0,
+      "step": 8620
+    },
+    {
+      "entropy": 0.24985795021057128,
+      "epoch": 4.8673814898419865,
+      "grad_norm": 1.9875785112380981,
+      "learning_rate": 4.478269749655222e-06,
+      "loss": 0.1414,
+      "mean_token_accuracy": 0.954399836063385,
+      "num_tokens": 70020932.0,
+      "step": 8625
+    },
+    {
+      "entropy": 0.23618102967739105,
+      "epoch": 4.87020316027088,
+      "grad_norm": 2.1521682739257812,
+      "learning_rate": 4.477694396612014e-06,
+      "loss": 0.124,
+      "mean_token_accuracy": 0.9608348488807679,
+      "num_tokens": 70061566.0,
+      "step": 8630
+    },
+    {
+      "entropy": 0.21555082201957704,
+      "epoch": 4.873024830699774,
+      "grad_norm": 1.909771203994751,
+      "learning_rate": 4.477118777831817e-06,
+      "loss": 0.1278,
+      "mean_token_accuracy": 0.9597031474113464,
+      "num_tokens": 70102195.0,
+      "step": 8635
+    },
+    {
+      "entropy": 0.24607637226581575,
+      "epoch": 4.8758465011286685,
+      "grad_norm": 2.6042673587799072,
+      "learning_rate": 4.47654289342771e-06,
+      "loss": 0.1404,
+      "mean_token_accuracy": 0.9529093623161315,
+      "num_tokens": 70142718.0,
+      "step": 8640
+    },
+    {
+      "entropy": 0.22525620758533477,
+      "epoch": 4.878668171557562,
+      "grad_norm": 2.1628029346466064,
+      "learning_rate": 4.475966743512826e-06,
+      "loss": 0.1384,
+      "mean_token_accuracy": 0.9556032299995423,
+      "num_tokens": 70183458.0,
+      "step": 8645
+    },
+    {
+      "entropy": 0.24763034284114838,
+      "epoch": 4.881489841986456,
+      "grad_norm": 2.244004726409912,
+      "learning_rate": 4.47539032820035e-06,
+      "loss": 0.1385,
+      "mean_token_accuracy": 0.9558951020240783,
+      "num_tokens": 70224283.0,
+      "step": 8650
+    },
+    {
+      "entropy": 0.23140210211277007,
+      "epoch": 4.8843115124153496,
+      "grad_norm": 2.290135145187378,
+      "learning_rate": 4.474813647603518e-06,
+      "loss": 0.1229,
+      "mean_token_accuracy": 0.9592395067214966,
+      "num_tokens": 70264983.0,
+      "step": 8655
+    },
+    {
+      "entropy": 0.2398304522037506,
+      "epoch": 4.887133182844244,
+      "grad_norm": 1.9016507863998413,
+      "learning_rate": 4.4742367018356195e-06,
+      "loss": 0.1311,
+      "mean_token_accuracy": 0.95786372423172,
+      "num_tokens": 70305649.0,
+      "step": 8660
+    },
+    {
+      "entropy": 0.23110878467559814,
+      "epoch": 4.889954853273138,
+      "grad_norm": 1.8025792837142944,
+      "learning_rate": 4.4736594910099956e-06,
+      "loss": 0.1405,
+      "mean_token_accuracy": 0.9554929614067078,
+      "num_tokens": 70345976.0,
+      "step": 8665
+    },
+    {
+      "entropy": 0.24155546426773072,
+      "epoch": 4.8927765237020315,
+      "grad_norm": 1.9929896593093872,
+      "learning_rate": 4.47308201524004e-06,
+      "loss": 0.1383,
+      "mean_token_accuracy": 0.955091404914856,
+      "num_tokens": 70386380.0,
+      "step": 8670
+    },
+    {
+      "entropy": 0.23273320198059083,
+      "epoch": 4.895598194130925,
+      "grad_norm": 1.964606523513794,
+      "learning_rate": 4.4725042746391965e-06,
+      "loss": 0.1267,
+      "mean_token_accuracy": 0.9588501572608947,
+      "num_tokens": 70427041.0,
+      "step": 8675
+    },
+    {
+      "entropy": 0.24253876507282257,
+      "epoch": 4.89841986455982,
+      "grad_norm": 2.12473464012146,
+      "learning_rate": 4.471926269320963e-06,
+      "loss": 0.1228,
+      "mean_token_accuracy": 0.9616016149520874,
+      "num_tokens": 70467666.0,
+      "step": 8680
+    },
+    {
+      "entropy": 0.23109547793865204,
+      "epoch": 4.9012415349887135,
+      "grad_norm": 2.302859306335449,
+      "learning_rate": 4.471347999398888e-06,
+      "loss": 0.1198,
+      "mean_token_accuracy": 0.9608020305633544,
+      "num_tokens": 70507748.0,
+      "step": 8685
+    },
+    {
+      "entropy": 0.21591301262378693,
+      "epoch": 4.904063205417607,
+      "grad_norm": 1.8348546028137207,
+      "learning_rate": 4.4707694649865755e-06,
+      "loss": 0.1206,
+      "mean_token_accuracy": 0.9611794590950012,
+      "num_tokens": 70548511.0,
+      "step": 8690
+    },
+    {
+      "entropy": 0.2398153394460678,
+      "epoch": 4.906884875846501,
+      "grad_norm": 1.7756059169769287,
+      "learning_rate": 4.470190666197675e-06,
+      "loss": 0.132,
+      "mean_token_accuracy": 0.9583917617797851,
+      "num_tokens": 70589210.0,
+      "step": 8695
+    },
+    {
+      "entropy": 0.24644698202610016,
+      "epoch": 4.909706546275395,
+      "grad_norm": 2.212986469268799,
+      "learning_rate": 4.469611603145895e-06,
+      "loss": 0.1325,
+      "mean_token_accuracy": 0.957687258720398,
+      "num_tokens": 70629759.0,
+      "step": 8700
+    },
+    {
+      "entropy": 0.2506899446249008,
+      "epoch": 4.912528216704289,
+      "grad_norm": 2.2517426013946533,
+      "learning_rate": 4.469032275944989e-06,
+      "loss": 0.1354,
+      "mean_token_accuracy": 0.9561644673347474,
+      "num_tokens": 70669828.0,
+      "step": 8705
+    },
+    {
+      "entropy": 0.26197082698345187,
+      "epoch": 4.915349887133183,
+      "grad_norm": 2.4998831748962402,
+      "learning_rate": 4.468452684708769e-06,
+      "loss": 0.1506,
+      "mean_token_accuracy": 0.953766405582428,
+      "num_tokens": 70710349.0,
+      "step": 8710
+    },
+    {
+      "entropy": 0.2373470038175583,
+      "epoch": 4.918171557562077,
+      "grad_norm": 2.084594249725342,
+      "learning_rate": 4.467872829551093e-06,
+      "loss": 0.1306,
+      "mean_token_accuracy": 0.9590035080909729,
+      "num_tokens": 70751075.0,
+      "step": 8715
+    },
+    {
+      "entropy": 0.23836889564990998,
+      "epoch": 4.92099322799097,
+      "grad_norm": 2.043854236602783,
+      "learning_rate": 4.467292710585876e-06,
+      "loss": 0.1263,
+      "mean_token_accuracy": 0.9599133729934692,
+      "num_tokens": 70791615.0,
+      "step": 8720
+    },
+    {
+      "entropy": 0.2358907401561737,
+      "epoch": 4.923814898419865,
+      "grad_norm": 1.9788299798965454,
+      "learning_rate": 4.466712327927082e-06,
+      "loss": 0.1422,
+      "mean_token_accuracy": 0.9546040058135986,
+      "num_tokens": 70832420.0,
+      "step": 8725
+    },
+    {
+      "entropy": 0.2312130182981491,
+      "epoch": 4.926636568848759,
+      "grad_norm": 2.282057762145996,
+      "learning_rate": 4.466131681688725e-06,
+      "loss": 0.1265,
+      "mean_token_accuracy": 0.9591808557510376,
+      "num_tokens": 70873131.0,
+      "step": 8730
+    },
+    {
+      "entropy": 0.22316355705261232,
+      "epoch": 4.929458239277652,
+      "grad_norm": 2.02055025100708,
+      "learning_rate": 4.465550771984877e-06,
+      "loss": 0.1187,
+      "mean_token_accuracy": 0.9624357938766479,
+      "num_tokens": 70913503.0,
+      "step": 8735
+    },
+    {
+      "entropy": 0.2304389774799347,
+      "epoch": 4.932279909706546,
+      "grad_norm": 2.3464314937591553,
+      "learning_rate": 4.464969598929654e-06,
+      "loss": 0.1205,
+      "mean_token_accuracy": 0.9610739946365356,
+      "num_tokens": 70954047.0,
+      "step": 8740
+    },
+    {
+      "entropy": 0.23224081397056578,
+      "epoch": 4.9351015801354405,
+      "grad_norm": 2.0249598026275635,
+      "learning_rate": 4.4643881626372305e-06,
+      "loss": 0.125,
+      "mean_token_accuracy": 0.959325659275055,
+      "num_tokens": 70994812.0,
+      "step": 8745
+    },
+    {
+      "entropy": 0.2500057280063629,
+      "epoch": 4.937923250564334,
+      "grad_norm": 2.1669914722442627,
+      "learning_rate": 4.463806463221827e-06,
+      "loss": 0.1378,
+      "mean_token_accuracy": 0.9567291259765625,
+      "num_tokens": 71035388.0,
+      "step": 8750
+    },
+    {
+      "entropy": 0.23122502863407135,
+      "epoch": 4.940744920993228,
+      "grad_norm": 2.2984166145324707,
+      "learning_rate": 4.463224500797721e-06,
+      "loss": 0.1396,
+      "mean_token_accuracy": 0.954986846446991,
+      "num_tokens": 71076225.0,
+      "step": 8755
+    },
+    {
+      "entropy": 0.22869625687599182,
+      "epoch": 4.943566591422122,
+      "grad_norm": 1.968353509902954,
+      "learning_rate": 4.462642275479236e-06,
+      "loss": 0.1187,
+      "mean_token_accuracy": 0.9607104420661926,
+      "num_tokens": 71116934.0,
+      "step": 8760
+    },
+    {
+      "entropy": 0.23740582168102264,
+      "epoch": 4.946388261851016,
+      "grad_norm": 1.954420566558838,
+      "learning_rate": 4.462059787380754e-06,
+      "loss": 0.1371,
+      "mean_token_accuracy": 0.9562184929847717,
+      "num_tokens": 71157332.0,
+      "step": 8765
+    },
+    {
+      "entropy": 0.23098691403865815,
+      "epoch": 4.94920993227991,
+      "grad_norm": 1.9516210556030273,
+      "learning_rate": 4.461477036616702e-06,
+      "loss": 0.1415,
+      "mean_token_accuracy": 0.9550253391265869,
+      "num_tokens": 71198170.0,
+      "step": 8770
+    },
+    {
+      "entropy": 0.22629740536212922,
+      "epoch": 4.952031602708804,
+      "grad_norm": 2.029778480529785,
+      "learning_rate": 4.460894023301563e-06,
+      "loss": 0.1159,
+      "mean_token_accuracy": 0.9626461505889893,
+      "num_tokens": 71238652.0,
+      "step": 8775
+    },
+    {
+      "entropy": 0.22563746571540833,
+      "epoch": 4.954853273137697,
+      "grad_norm": 2.0067238807678223,
+      "learning_rate": 4.460310747549869e-06,
+      "loss": 0.128,
+      "mean_token_accuracy": 0.9585816025733948,
+      "num_tokens": 71279449.0,
+      "step": 8780
+    },
+    {
+      "entropy": 0.264061963558197,
+      "epoch": 4.957674943566591,
+      "grad_norm": 2.42386794090271,
+      "learning_rate": 4.459727209476205e-06,
+      "loss": 0.1534,
+      "mean_token_accuracy": 0.9512446284294128,
+      "num_tokens": 71320179.0,
+      "step": 8785
+    },
+    {
+      "entropy": 0.2313307523727417,
+      "epoch": 4.960496613995486,
+      "grad_norm": 1.9760738611221313,
+      "learning_rate": 4.459143409195208e-06,
+      "loss": 0.1184,
+      "mean_token_accuracy": 0.9614473700523376,
+      "num_tokens": 71360778.0,
+      "step": 8790
+    },
+    {
+      "entropy": 0.23215781152248383,
+      "epoch": 4.963318284424379,
+      "grad_norm": 2.0897891521453857,
+      "learning_rate": 4.458559346821564e-06,
+      "loss": 0.1406,
+      "mean_token_accuracy": 0.9561804056167602,
+      "num_tokens": 71401477.0,
+      "step": 8795
+    },
+    {
+      "entropy": 0.2455916315317154,
+      "epoch": 4.966139954853273,
+      "grad_norm": 2.2576193809509277,
+      "learning_rate": 4.457975022470013e-06,
+      "loss": 0.139,
+      "mean_token_accuracy": 0.9552298545837402,
+      "num_tokens": 71442131.0,
+      "step": 8800
+    },
+    {
+      "entropy": 0.2560196042060852,
+      "epoch": 4.968961625282167,
+      "grad_norm": 2.23545503616333,
+      "learning_rate": 4.457390436255345e-06,
+      "loss": 0.1483,
+      "mean_token_accuracy": 0.952830684185028,
+      "num_tokens": 71482712.0,
+      "step": 8805
+    },
+    {
+      "entropy": 0.24611234068870544,
+      "epoch": 4.971783295711061,
+      "grad_norm": 1.8639657497406006,
+      "learning_rate": 4.456805588292404e-06,
+      "loss": 0.1302,
+      "mean_token_accuracy": 0.9585993766784668,
+      "num_tokens": 71523349.0,
+      "step": 8810
+    },
+    {
+      "entropy": 0.2496563047170639,
+      "epoch": 4.974604966139955,
+      "grad_norm": 2.166790723800659,
+      "learning_rate": 4.456220478696081e-06,
+      "loss": 0.1514,
+      "mean_token_accuracy": 0.9514712452888489,
+      "num_tokens": 71563712.0,
+      "step": 8815
+    },
+    {
+      "entropy": 0.23716452717781067,
+      "epoch": 4.977426636568849,
+      "grad_norm": 2.0670976638793945,
+      "learning_rate": 4.455635107581322e-06,
+      "loss": 0.1365,
+      "mean_token_accuracy": 0.9566124558448792,
+      "num_tokens": 71604371.0,
+      "step": 8820
+    },
+    {
+      "entropy": 0.21934570670127868,
+      "epoch": 4.980248306997742,
+      "grad_norm": 2.0929582118988037,
+      "learning_rate": 4.455049475063124e-06,
+      "loss": 0.1163,
+      "mean_token_accuracy": 0.9618055701255799,
+      "num_tokens": 71645153.0,
+      "step": 8825
+    },
+    {
+      "entropy": 0.24300400614738465,
+      "epoch": 4.983069977426637,
+      "grad_norm": 1.9957592487335205,
+      "learning_rate": 4.4544635812565335e-06,
+      "loss": 0.1322,
+      "mean_token_accuracy": 0.9577077746391296,
+      "num_tokens": 71685968.0,
+      "step": 8830
+    },
+    {
+      "entropy": 0.24642634391784668,
+      "epoch": 4.985891647855531,
+      "grad_norm": 1.812232494354248,
+      "learning_rate": 4.453877426276649e-06,
+      "loss": 0.1327,
+      "mean_token_accuracy": 0.9573553204536438,
+      "num_tokens": 71726713.0,
+      "step": 8835
+    },
+    {
+      "entropy": 0.24452163875102997,
+      "epoch": 4.988713318284424,
+      "grad_norm": 2.2379374504089355,
+      "learning_rate": 4.4532910102386215e-06,
+      "loss": 0.1368,
+      "mean_token_accuracy": 0.9563169717788697,
+      "num_tokens": 71767291.0,
+      "step": 8840
+    },
+    {
+      "entropy": 0.23888054192066194,
+      "epoch": 4.991534988713318,
+      "grad_norm": 2.2258286476135254,
+      "learning_rate": 4.452704333257653e-06,
+      "loss": 0.1338,
+      "mean_token_accuracy": 0.9571861267089844,
+      "num_tokens": 71807947.0,
+      "step": 8845
+    },
+    {
+      "entropy": 0.23455857038497924,
+      "epoch": 4.994356659142213,
+      "grad_norm": 2.2029316425323486,
+      "learning_rate": 4.452117395448995e-06,
+      "loss": 0.1321,
+      "mean_token_accuracy": 0.9573328852653503,
+      "num_tokens": 71848678.0,
+      "step": 8850
+    },
+    {
+      "entropy": 0.23345192968845369,
+      "epoch": 4.997178329571106,
+      "grad_norm": 2.2948131561279297,
+      "learning_rate": 4.451530196927952e-06,
+      "loss": 0.1239,
+      "mean_token_accuracy": 0.9603603482246399,
+      "num_tokens": 71889260.0,
+      "step": 8855
+    },
+    {
+      "entropy": 0.22622182071208954,
+      "epoch": 5.0,
+      "grad_norm": 3.6322450637817383,
+      "learning_rate": 4.45094273780988e-06,
+      "loss": 0.1305,
+      "mean_token_accuracy": 0.9572938919067383,
+      "num_tokens": 71923845.0,
+      "step": 8860
+    },
+    {
+      "entropy": 0.22580887079238893,
+      "epoch": 5.002821670428894,
+      "grad_norm": 1.672566294670105,
+      "learning_rate": 4.450355018210185e-06,
+      "loss": 0.0748,
+      "mean_token_accuracy": 0.9793183445930481,
+      "num_tokens": 71964397.0,
+      "step": 8865
+    },
+    {
+      "entropy": 0.1980500638484955,
+      "epoch": 5.005643340857787,
+      "grad_norm": 1.7135627269744873,
+      "learning_rate": 4.4497670382443235e-06,
+      "loss": 0.0595,
+      "mean_token_accuracy": 0.9829611539840698,
+      "num_tokens": 72004845.0,
+      "step": 8870
+    },
+    {
+      "entropy": 0.19046223759651185,
+      "epoch": 5.008465011286682,
+      "grad_norm": 1.7195278406143188,
+      "learning_rate": 4.449178798027806e-06,
+      "loss": 0.0699,
+      "mean_token_accuracy": 0.9804732322692871,
+      "num_tokens": 72045486.0,
+      "step": 8875
+    },
+    {
+      "entropy": 0.1900404006242752,
+      "epoch": 5.011286681715576,
+      "grad_norm": 1.8688246011734009,
+      "learning_rate": 4.4485902976761905e-06,
+      "loss": 0.0699,
+      "mean_token_accuracy": 0.9789183497428894,
+      "num_tokens": 72086189.0,
+      "step": 8880
+    },
+    {
+      "entropy": 0.17116313576698303,
+      "epoch": 5.014108352144469,
+      "grad_norm": 2.521716833114624,
+      "learning_rate": 4.44800153730509e-06,
+      "loss": 0.0676,
+      "mean_token_accuracy": 0.979660964012146,
+      "num_tokens": 72126926.0,
+      "step": 8885
+    },
+    {
+      "entropy": 0.1692497670650482,
+      "epoch": 5.016930022573363,
+      "grad_norm": 2.2940783500671387,
+      "learning_rate": 4.447412517030165e-06,
+      "loss": 0.0638,
+      "mean_token_accuracy": 0.980719518661499,
+      "num_tokens": 72167664.0,
+      "step": 8890
+    },
+    {
+      "entropy": 0.18492531478405,
+      "epoch": 5.019751693002258,
+      "grad_norm": 2.423424005508423,
+      "learning_rate": 4.446823236967129e-06,
+      "loss": 0.086,
+      "mean_token_accuracy": 0.9749364614486694,
+      "num_tokens": 72208162.0,
+      "step": 8895
+    },
+    {
+      "entropy": 0.1805371016263962,
+      "epoch": 5.022573363431151,
+      "grad_norm": 2.1918728351593018,
+      "learning_rate": 4.446233697231747e-06,
+      "loss": 0.0717,
+      "mean_token_accuracy": 0.9780835628509521,
+      "num_tokens": 72248751.0,
+      "step": 8900
+    },
+    {
+      "entropy": 0.19484019577503203,
+      "epoch": 5.025395033860045,
+      "grad_norm": 2.126352071762085,
+      "learning_rate": 4.445643897939832e-06,
+      "loss": 0.0653,
+      "mean_token_accuracy": 0.9807442188262939,
+      "num_tokens": 72289553.0,
+      "step": 8905
+    },
+    {
+      "entropy": 0.16935117840766906,
+      "epoch": 5.028216704288939,
+      "grad_norm": 1.7222833633422852,
+      "learning_rate": 4.445053839207252e-06,
+      "loss": 0.0582,
+      "mean_token_accuracy": 0.9815907716751099,
+      "num_tokens": 72330231.0,
+      "step": 8910
+    },
+    {
+      "entropy": 0.17409595847129822,
+      "epoch": 5.031038374717833,
+      "grad_norm": 1.9635859727859497,
+      "learning_rate": 4.4444635211499245e-06,
+      "loss": 0.0644,
+      "mean_token_accuracy": 0.9801131606101989,
+      "num_tokens": 72370964.0,
+      "step": 8915
+    },
+    {
+      "entropy": 0.18664040863513948,
+      "epoch": 5.033860045146727,
+      "grad_norm": 2.004497528076172,
+      "learning_rate": 4.443872943883817e-06,
+      "loss": 0.0686,
+      "mean_token_accuracy": 0.9801067471504211,
+      "num_tokens": 72411632.0,
+      "step": 8920
+    },
+    {
+      "entropy": 0.19012978672981262,
+      "epoch": 5.036681715575621,
+      "grad_norm": 2.1313867568969727,
+      "learning_rate": 4.443282107524947e-06,
+      "loss": 0.0779,
+      "mean_token_accuracy": 0.9768623113632202,
+      "num_tokens": 72452367.0,
+      "step": 8925
+    },
+    {
+      "entropy": 0.18184551000595092,
+      "epoch": 5.039503386004514,
+      "grad_norm": 2.076591730117798,
+      "learning_rate": 4.442691012189386e-06,
+      "loss": 0.0734,
+      "mean_token_accuracy": 0.9780202388763428,
+      "num_tokens": 72493043.0,
+      "step": 8930
+    },
+    {
+      "entropy": 0.17082217931747437,
+      "epoch": 5.042325056433409,
+      "grad_norm": 1.6795393228530884,
+      "learning_rate": 4.4420996579932555e-06,
+      "loss": 0.0697,
+      "mean_token_accuracy": 0.9782551288604736,
+      "num_tokens": 72533217.0,
+      "step": 8935
+    },
+    {
+      "entropy": 0.19154114723205568,
+      "epoch": 5.045146726862303,
+      "grad_norm": 1.8840192556381226,
+      "learning_rate": 4.4415080450527244e-06,
+      "loss": 0.0783,
+      "mean_token_accuracy": 0.9762571692466736,
+      "num_tokens": 72573491.0,
+      "step": 8940
+    },
+    {
+      "entropy": 0.1921801507472992,
+      "epoch": 5.047968397291196,
+      "grad_norm": 2.051151752471924,
+      "learning_rate": 4.440916173484018e-06,
+      "loss": 0.0735,
+      "mean_token_accuracy": 0.9775419950485229,
+      "num_tokens": 72614111.0,
+      "step": 8945
+    },
+    {
+      "entropy": 0.171154123544693,
+      "epoch": 5.05079006772009,
+      "grad_norm": 1.7416431903839111,
+      "learning_rate": 4.440324043403408e-06,
+      "loss": 0.0562,
+      "mean_token_accuracy": 0.982762086391449,
+      "num_tokens": 72654942.0,
+      "step": 8950
+    },
+    {
+      "entropy": 0.1801618218421936,
+      "epoch": 5.053611738148984,
+      "grad_norm": 2.1165449619293213,
+      "learning_rate": 4.439731654927218e-06,
+      "loss": 0.0651,
+      "mean_token_accuracy": 0.9803321242332459,
+      "num_tokens": 72695401.0,
+      "step": 8955
+    },
+    {
+      "entropy": 0.17029935419559478,
+      "epoch": 5.056433408577878,
+      "grad_norm": 2.3758440017700195,
+      "learning_rate": 4.439139008171824e-06,
+      "loss": 0.0691,
+      "mean_token_accuracy": 0.9787796020507813,
+      "num_tokens": 72736277.0,
+      "step": 8960
+    },
+    {
+      "entropy": 0.1681983232498169,
+      "epoch": 5.059255079006772,
+      "grad_norm": 1.9630498886108398,
+      "learning_rate": 4.43854610325365e-06,
+      "loss": 0.0617,
+      "mean_token_accuracy": 0.981269919872284,
+      "num_tokens": 72777155.0,
+      "step": 8965
+    },
+    {
+      "entropy": 0.20426134169101715,
+      "epoch": 5.062076749435666,
+      "grad_norm": 2.2527589797973633,
+      "learning_rate": 4.437952940289175e-06,
+      "loss": 0.0815,
+      "mean_token_accuracy": 0.9752603888511657,
+      "num_tokens": 72817863.0,
+      "step": 8970
+    },
+    {
+      "entropy": 0.17029196619987488,
+      "epoch": 5.0648984198645595,
+      "grad_norm": 1.9497177600860596,
+      "learning_rate": 4.437359519394923e-06,
+      "loss": 0.0664,
+      "mean_token_accuracy": 0.9784554839134216,
+      "num_tokens": 72858530.0,
+      "step": 8975
+    },
+    {
+      "entropy": 0.18146368563175203,
+      "epoch": 5.067720090293454,
+      "grad_norm": 1.7516709566116333,
+      "learning_rate": 4.436765840687473e-06,
+      "loss": 0.0697,
+      "mean_token_accuracy": 0.9783901810646057,
+      "num_tokens": 72899175.0,
+      "step": 8980
+    },
+    {
+      "entropy": 0.17964912950992584,
+      "epoch": 5.070541760722348,
+      "grad_norm": 1.487107515335083,
+      "learning_rate": 4.4361719042834525e-06,
+      "loss": 0.0595,
+      "mean_token_accuracy": 0.9815647959709167,
+      "num_tokens": 72939846.0,
+      "step": 8985
+    },
+    {
+      "entropy": 0.1833242356777191,
+      "epoch": 5.073363431151241,
+      "grad_norm": 2.1468114852905273,
+      "learning_rate": 4.435577710299542e-06,
+      "loss": 0.0569,
+      "mean_token_accuracy": 0.982574725151062,
+      "num_tokens": 72979909.0,
+      "step": 8990
+    },
+    {
+      "entropy": 0.1870110362768173,
+      "epoch": 5.076185101580135,
+      "grad_norm": 1.998921513557434,
+      "learning_rate": 4.43498325885247e-06,
+      "loss": 0.0646,
+      "mean_token_accuracy": 0.9806098222732544,
+      "num_tokens": 73020545.0,
+      "step": 8995
+    },
+    {
+      "entropy": 0.18318285942077636,
+      "epoch": 5.07900677200903,
+      "grad_norm": 1.859528660774231,
+      "learning_rate": 4.434388550059016e-06,
+      "loss": 0.0739,
+      "mean_token_accuracy": 0.9773783922195435,
+      "num_tokens": 73061009.0,
+      "step": 9000
+    },
+    {
+      "epoch": 5.07900677200903,
+      "eval_entropy": 0.23515231907367706,
+      "eval_loss": 0.055877141654491425,
+      "eval_mean_token_accuracy": 0.9827520251274109,
+      "eval_num_tokens": 73061009.0,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 24.455,
+      "eval_steps_per_second": 6.114,
+      "step": 9000
+    },
+    {
+      "entropy": 0.1605137288570404,
+      "epoch": 5.081828442437923,
+      "grad_norm": 2.0149075984954834,
+      "learning_rate": 4.433793584036011e-06,
+      "loss": 0.0591,
+      "mean_token_accuracy": 0.9814065933227539,
+      "num_tokens": 73101615.0,
+      "step": 9005
+    },
+    {
+      "entropy": 0.19753072261810303,
+      "epoch": 5.084650112866817,
+      "grad_norm": 1.9284378290176392,
+      "learning_rate": 4.433198360900337e-06,
+      "loss": 0.0738,
+      "mean_token_accuracy": 0.9784594297409057,
+      "num_tokens": 73142398.0,
+      "step": 9010
+    },
+    {
+      "entropy": 0.18846041858196258,
+      "epoch": 5.087471783295711,
+      "grad_norm": 2.1237173080444336,
+      "learning_rate": 4.432602880768925e-06,
+      "loss": 0.076,
+      "mean_token_accuracy": 0.9769237160682678,
+      "num_tokens": 73183058.0,
+      "step": 9015
+    },
+    {
+      "entropy": 0.1911979854106903,
+      "epoch": 5.090293453724605,
+      "grad_norm": 1.8477128744125366,
+      "learning_rate": 4.4320071437587554e-06,
+      "loss": 0.0704,
+      "mean_token_accuracy": 0.9788993120193481,
+      "num_tokens": 73223530.0,
+      "step": 9020
+    },
+    {
+      "entropy": 0.19110194444656373,
+      "epoch": 5.093115124153499,
+      "grad_norm": 2.072406053543091,
+      "learning_rate": 4.431411149986865e-06,
+      "loss": 0.074,
+      "mean_token_accuracy": 0.9778715133666992,
+      "num_tokens": 73264278.0,
+      "step": 9025
+    },
+    {
+      "entropy": 0.18987910449504852,
+      "epoch": 5.095936794582393,
+      "grad_norm": 1.7068977355957031,
+      "learning_rate": 4.430814899570333e-06,
+      "loss": 0.0745,
+      "mean_token_accuracy": 0.9774025797843933,
+      "num_tokens": 73304905.0,
+      "step": 9030
+    },
+    {
+      "entropy": 0.18311900794506072,
+      "epoch": 5.0987584650112865,
+      "grad_norm": 2.1360621452331543,
+      "learning_rate": 4.430218392626295e-06,
+      "loss": 0.065,
+      "mean_token_accuracy": 0.9798977851867676,
+      "num_tokens": 73345687.0,
+      "step": 9035
+    },
+    {
+      "entropy": 0.17857904732227325,
+      "epoch": 5.10158013544018,
+      "grad_norm": 1.964221477508545,
+      "learning_rate": 4.429621629271933e-06,
+      "loss": 0.0628,
+      "mean_token_accuracy": 0.980586564540863,
+      "num_tokens": 73386265.0,
+      "step": 9040
+    },
+    {
+      "entropy": 0.17305072247982026,
+      "epoch": 5.104401805869075,
+      "grad_norm": 2.037114143371582,
+      "learning_rate": 4.429024609624482e-06,
+      "loss": 0.0636,
+      "mean_token_accuracy": 0.9802641272544861,
+      "num_tokens": 73427151.0,
+      "step": 9045
+    },
+    {
+      "entropy": 0.18654557764530183,
+      "epoch": 5.1072234762979685,
+      "grad_norm": 2.1122262477874756,
+      "learning_rate": 4.428427333801228e-06,
+      "loss": 0.0626,
+      "mean_token_accuracy": 0.9806604266166687,
+      "num_tokens": 73467943.0,
+      "step": 9050
+    },
+    {
+      "entropy": 0.1888034909963608,
+      "epoch": 5.110045146726862,
+      "grad_norm": 2.032395124435425,
+      "learning_rate": 4.4278298019195044e-06,
+      "loss": 0.0677,
+      "mean_token_accuracy": 0.9790974020957947,
+      "num_tokens": 73508586.0,
+      "step": 9055
+    },
+    {
+      "entropy": 0.201117542386055,
+      "epoch": 5.112866817155756,
+      "grad_norm": 2.6121714115142822,
+      "learning_rate": 4.4272320140966965e-06,
+      "loss": 0.0789,
+      "mean_token_accuracy": 0.9769583702087402,
+      "num_tokens": 73549036.0,
+      "step": 9060
+    },
+    {
+      "entropy": 0.17489383816719056,
+      "epoch": 5.1156884875846504,
+      "grad_norm": 2.1628427505493164,
+      "learning_rate": 4.4266339704502415e-06,
+      "loss": 0.0704,
+      "mean_token_accuracy": 0.9784565687179565,
+      "num_tokens": 73589624.0,
+      "step": 9065
+    },
+    {
+      "entropy": 0.17749419510364534,
+      "epoch": 5.118510158013544,
+      "grad_norm": 2.2119405269622803,
+      "learning_rate": 4.426035671097623e-06,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.980502438545227,
+      "num_tokens": 73630265.0,
+      "step": 9070
+    },
+    {
+      "entropy": 0.17612397372722627,
+      "epoch": 5.121331828442438,
+      "grad_norm": 1.97671377658844,
+      "learning_rate": 4.425437116156377e-06,
+      "loss": 0.0741,
+      "mean_token_accuracy": 0.9769485712051391,
+      "num_tokens": 73670966.0,
+      "step": 9075
+    },
+    {
+      "entropy": 0.1710589587688446,
+      "epoch": 5.1241534988713315,
+      "grad_norm": 2.027311325073242,
+      "learning_rate": 4.424838305744091e-06,
+      "loss": 0.0636,
+      "mean_token_accuracy": 0.9812980055809021,
+      "num_tokens": 73711495.0,
+      "step": 9080
+    },
+    {
+      "entropy": 0.16620054244995117,
+      "epoch": 5.126975169300226,
+      "grad_norm": 2.079716920852661,
+      "learning_rate": 4.4242392399784015e-06,
+      "loss": 0.0599,
+      "mean_token_accuracy": 0.9820851564407349,
+      "num_tokens": 73751986.0,
+      "step": 9085
+    },
+    {
+      "entropy": 0.19693105220794677,
+      "epoch": 5.12979683972912,
+      "grad_norm": 1.8494852781295776,
+      "learning_rate": 4.423639918976994e-06,
+      "loss": 0.0841,
+      "mean_token_accuracy": 0.9751617550849915,
+      "num_tokens": 73792696.0,
+      "step": 9090
+    },
+    {
+      "entropy": 0.16846722960472107,
+      "epoch": 5.1326185101580135,
+      "grad_norm": 2.0323588848114014,
+      "learning_rate": 4.4230403428576055e-06,
+      "loss": 0.0576,
+      "mean_token_accuracy": 0.983081865310669,
+      "num_tokens": 73833479.0,
+      "step": 9095
+    },
+    {
+      "entropy": 0.18227221965789794,
+      "epoch": 5.135440180586907,
+      "grad_norm": 1.713901400566101,
+      "learning_rate": 4.4224405117380235e-06,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9784926295280456,
+      "num_tokens": 73874141.0,
+      "step": 9100
+    },
+    {
+      "entropy": 0.17101970911026002,
+      "epoch": 5.138261851015802,
+      "grad_norm": 1.8212043046951294,
+      "learning_rate": 4.421840425736084e-06,
+      "loss": 0.0694,
+      "mean_token_accuracy": 0.9785187363624572,
+      "num_tokens": 73914676.0,
+      "step": 9105
+    },
+    {
+      "entropy": 0.18568507134914397,
+      "epoch": 5.1410835214446955,
+      "grad_norm": 2.0307278633117676,
+      "learning_rate": 4.421240084969673e-06,
+      "loss": 0.0681,
+      "mean_token_accuracy": 0.9798570513725281,
+      "num_tokens": 73955243.0,
+      "step": 9110
+    },
+    {
+      "entropy": 0.18314568400382997,
+      "epoch": 5.143905191873589,
+      "grad_norm": 2.26824951171875,
+      "learning_rate": 4.42063948955673e-06,
+      "loss": 0.0819,
+      "mean_token_accuracy": 0.9749473333358765,
+      "num_tokens": 73996064.0,
+      "step": 9115
+    },
+    {
+      "entropy": 0.18159986436367034,
+      "epoch": 5.146726862302483,
+      "grad_norm": 2.368440866470337,
+      "learning_rate": 4.420038639615241e-06,
+      "loss": 0.0759,
+      "mean_token_accuracy": 0.97624671459198,
+      "num_tokens": 74036627.0,
+      "step": 9120
+    },
+    {
+      "entropy": 0.18119294345378875,
+      "epoch": 5.149548532731377,
+      "grad_norm": 1.908693552017212,
+      "learning_rate": 4.419437535263243e-06,
+      "loss": 0.067,
+      "mean_token_accuracy": 0.9788689136505127,
+      "num_tokens": 74077307.0,
+      "step": 9125
+    },
+    {
+      "entropy": 0.18659307360649108,
+      "epoch": 5.152370203160271,
+      "grad_norm": 1.8129174709320068,
+      "learning_rate": 4.418836176618823e-06,
+      "loss": 0.0727,
+      "mean_token_accuracy": 0.9778983354568481,
+      "num_tokens": 74117925.0,
+      "step": 9130
+    },
+    {
+      "entropy": 0.19027889668941497,
+      "epoch": 5.155191873589165,
+      "grad_norm": 1.8075789213180542,
+      "learning_rate": 4.418234563800117e-06,
+      "loss": 0.0733,
+      "mean_token_accuracy": 0.9768654942512512,
+      "num_tokens": 74158594.0,
+      "step": 9135
+    },
+    {
+      "entropy": 0.17420806884765624,
+      "epoch": 5.158013544018059,
+      "grad_norm": 2.1815521717071533,
+      "learning_rate": 4.417632696925314e-06,
+      "loss": 0.0643,
+      "mean_token_accuracy": 0.980584466457367,
+      "num_tokens": 74199375.0,
+      "step": 9140
+    },
+    {
+      "entropy": 0.19676790237426758,
+      "epoch": 5.160835214446952,
+      "grad_norm": 2.2495217323303223,
+      "learning_rate": 4.417030576112649e-06,
+      "loss": 0.0635,
+      "mean_token_accuracy": 0.9816282868385315,
+      "num_tokens": 74239922.0,
+      "step": 9145
+    },
+    {
+      "entropy": 0.17044782042503356,
+      "epoch": 5.163656884875847,
+      "grad_norm": 2.1400983333587646,
+      "learning_rate": 4.416428201480409e-06,
+      "loss": 0.0741,
+      "mean_token_accuracy": 0.9771974563598633,
+      "num_tokens": 74280490.0,
+      "step": 9150
+    },
+    {
+      "entropy": 0.17132185995578766,
+      "epoch": 5.1664785553047405,
+      "grad_norm": 1.5042531490325928,
+      "learning_rate": 4.415825573146931e-06,
+      "loss": 0.066,
+      "mean_token_accuracy": 0.9790427088737488,
+      "num_tokens": 74321233.0,
+      "step": 9155
+    },
+    {
+      "entropy": 0.1608135759830475,
+      "epoch": 5.169300225733634,
+      "grad_norm": 1.817196011543274,
+      "learning_rate": 4.415222691230602e-06,
+      "loss": 0.0584,
+      "mean_token_accuracy": 0.9813137888908386,
+      "num_tokens": 74361867.0,
+      "step": 9160
+    },
+    {
+      "entropy": 0.19295158088207245,
+      "epoch": 5.172121896162528,
+      "grad_norm": 2.3909380435943604,
+      "learning_rate": 4.414619555849857e-06,
+      "loss": 0.0688,
+      "mean_token_accuracy": 0.9787463188171387,
+      "num_tokens": 74402471.0,
+      "step": 9165
+    },
+    {
+      "entropy": 0.1939996987581253,
+      "epoch": 5.1749435665914225,
+      "grad_norm": 2.0213077068328857,
+      "learning_rate": 4.414016167123183e-06,
+      "loss": 0.0715,
+      "mean_token_accuracy": 0.978550124168396,
+      "num_tokens": 74442930.0,
+      "step": 9170
+    },
+    {
+      "entropy": 0.18027763068675995,
+      "epoch": 5.177765237020316,
+      "grad_norm": 1.568903923034668,
+      "learning_rate": 4.413412525169115e-06,
+      "loss": 0.0628,
+      "mean_token_accuracy": 0.9806607604026795,
+      "num_tokens": 74483369.0,
+      "step": 9175
+    },
+    {
+      "entropy": 0.17010557353496553,
+      "epoch": 5.18058690744921,
+      "grad_norm": 2.0850884914398193,
+      "learning_rate": 4.4128086301062405e-06,
+      "loss": 0.0652,
+      "mean_token_accuracy": 0.9795267939567566,
+      "num_tokens": 74523998.0,
+      "step": 9180
+    },
+    {
+      "entropy": 0.1618386447429657,
+      "epoch": 5.183408577878104,
+      "grad_norm": 2.156888484954834,
+      "learning_rate": 4.412204482053191e-06,
+      "loss": 0.0653,
+      "mean_token_accuracy": 0.9797653436660767,
+      "num_tokens": 74564676.0,
+      "step": 9185
+    },
+    {
+      "entropy": 0.1769125282764435,
+      "epoch": 5.186230248306998,
+      "grad_norm": 1.982959270477295,
+      "learning_rate": 4.411600081128655e-06,
+      "loss": 0.0692,
+      "mean_token_accuracy": 0.9781957626342773,
+      "num_tokens": 74605459.0,
+      "step": 9190
+    },
+    {
+      "entropy": 0.1831894338130951,
+      "epoch": 5.189051918735892,
+      "grad_norm": 2.3006227016448975,
+      "learning_rate": 4.410995427451365e-06,
+      "loss": 0.0659,
+      "mean_token_accuracy": 0.9792438864707946,
+      "num_tokens": 74646245.0,
+      "step": 9195
+    },
+    {
+      "entropy": 0.16881802082061767,
+      "epoch": 5.191873589164786,
+      "grad_norm": 1.7162789106369019,
+      "learning_rate": 4.410390521140107e-06,
+      "loss": 0.0598,
+      "mean_token_accuracy": 0.9816855788230896,
+      "num_tokens": 74686744.0,
+      "step": 9200
+    },
+    {
+      "entropy": 0.17817852199077605,
+      "epoch": 5.194695259593679,
+      "grad_norm": 1.815919041633606,
+      "learning_rate": 4.409785362313714e-06,
+      "loss": 0.0659,
+      "mean_token_accuracy": 0.9804519295692444,
+      "num_tokens": 74727229.0,
+      "step": 9205
+    },
+    {
+      "entropy": 0.16955990195274354,
+      "epoch": 5.197516930022573,
+      "grad_norm": 1.5127193927764893,
+      "learning_rate": 4.409179951091069e-06,
+      "loss": 0.0704,
+      "mean_token_accuracy": 0.9778647541999816,
+      "num_tokens": 74767941.0,
+      "step": 9210
+    },
+    {
+      "entropy": 0.17718103229999543,
+      "epoch": 5.200338600451468,
+      "grad_norm": 1.9324251413345337,
+      "learning_rate": 4.408574287591105e-06,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9797194123268127,
+      "num_tokens": 74808629.0,
+      "step": 9215
+    },
+    {
+      "entropy": 0.18118757009506226,
+      "epoch": 5.203160270880361,
+      "grad_norm": 1.6411625146865845,
+      "learning_rate": 4.407968371932807e-06,
+      "loss": 0.0668,
+      "mean_token_accuracy": 0.9800692081451416,
+      "num_tokens": 74849290.0,
+      "step": 9220
+    },
+    {
+      "entropy": 0.16765645444393157,
+      "epoch": 5.205981941309255,
+      "grad_norm": 2.054521322250366,
+      "learning_rate": 4.407362204235205e-06,
+      "loss": 0.0622,
+      "mean_token_accuracy": 0.9807947397232055,
+      "num_tokens": 74889988.0,
+      "step": 9225
+    },
+    {
+      "entropy": 0.18616848886013032,
+      "epoch": 5.208803611738149,
+      "grad_norm": 2.160146474838257,
+      "learning_rate": 4.40675578461738e-06,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.9792033910751343,
+      "num_tokens": 74930594.0,
+      "step": 9230
+    },
+    {
+      "entropy": 0.17592273354530336,
+      "epoch": 5.211625282167043,
+      "grad_norm": 2.01505708694458,
+      "learning_rate": 4.4061491131984655e-06,
+      "loss": 0.0708,
+      "mean_token_accuracy": 0.9776349067687988,
+      "num_tokens": 74971295.0,
+      "step": 9235
+    },
+    {
+      "entropy": 0.16667756140232087,
+      "epoch": 5.214446952595937,
+      "grad_norm": 1.9631502628326416,
+      "learning_rate": 4.405542190097641e-06,
+      "loss": 0.0639,
+      "mean_token_accuracy": 0.9816940546035766,
+      "num_tokens": 75011986.0,
+      "step": 9240
+    },
+    {
+      "entropy": 0.17029538750648499,
+      "epoch": 5.217268623024831,
+      "grad_norm": 1.7982895374298096,
+      "learning_rate": 4.404935015434138e-06,
+      "loss": 0.0641,
+      "mean_token_accuracy": 0.9802758932113648,
+      "num_tokens": 75052805.0,
+      "step": 9245
+    },
+    {
+      "entropy": 0.172264164686203,
+      "epoch": 5.220090293453724,
+      "grad_norm": 2.2566871643066406,
+      "learning_rate": 4.404327589327234e-06,
+      "loss": 0.0668,
+      "mean_token_accuracy": 0.9802647709846497,
+      "num_tokens": 75093268.0,
+      "step": 9250
+    },
+    {
+      "entropy": 0.16985282599925994,
+      "epoch": 5.222911963882619,
+      "grad_norm": 1.5910284519195557,
+      "learning_rate": 4.403719911896258e-06,
+      "loss": 0.0634,
+      "mean_token_accuracy": 0.9810275197029114,
+      "num_tokens": 75134071.0,
+      "step": 9255
+    },
+    {
+      "entropy": 0.19104610085487367,
+      "epoch": 5.225733634311513,
+      "grad_norm": 2.561847448348999,
+      "learning_rate": 4.40311198326059e-06,
+      "loss": 0.0704,
+      "mean_token_accuracy": 0.9786494374275208,
+      "num_tokens": 75174864.0,
+      "step": 9260
+    },
+    {
+      "entropy": 0.19318519532680511,
+      "epoch": 5.228555304740406,
+      "grad_norm": 1.789509654045105,
+      "learning_rate": 4.402503803539656e-06,
+      "loss": 0.0748,
+      "mean_token_accuracy": 0.9768137574195862,
+      "num_tokens": 75215612.0,
+      "step": 9265
+    },
+    {
+      "entropy": 0.1831961005926132,
+      "epoch": 5.2313769751693,
+      "grad_norm": 1.7263920307159424,
+      "learning_rate": 4.401895372852935e-06,
+      "loss": 0.0658,
+      "mean_token_accuracy": 0.9797954320907593,
+      "num_tokens": 75255813.0,
+      "step": 9270
+    },
+    {
+      "entropy": 0.16736743450164795,
+      "epoch": 5.234198645598195,
+      "grad_norm": 1.8194705247879028,
+      "learning_rate": 4.401286691319951e-06,
+      "loss": 0.0677,
+      "mean_token_accuracy": 0.9790467739105224,
+      "num_tokens": 75296532.0,
+      "step": 9275
+    },
+    {
+      "entropy": 0.17123483419418334,
+      "epoch": 5.237020316027088,
+      "grad_norm": 1.8370773792266846,
+      "learning_rate": 4.40067775906028e-06,
+      "loss": 0.0634,
+      "mean_token_accuracy": 0.9816179037094116,
+      "num_tokens": 75337252.0,
+      "step": 9280
+    },
+    {
+      "entropy": 0.18256573975086213,
+      "epoch": 5.239841986455982,
+      "grad_norm": 1.8240233659744263,
+      "learning_rate": 4.400068576193549e-06,
+      "loss": 0.0654,
+      "mean_token_accuracy": 0.9805835843086242,
+      "num_tokens": 75377833.0,
+      "step": 9285
+    },
+    {
+      "entropy": 0.19250957071781158,
+      "epoch": 5.242663656884876,
+      "grad_norm": 1.8019994497299194,
+      "learning_rate": 4.399459142839429e-06,
+      "loss": 0.0775,
+      "mean_token_accuracy": 0.9766400456428528,
+      "num_tokens": 75418751.0,
+      "step": 9290
+    },
+    {
+      "entropy": 0.204935023188591,
+      "epoch": 5.245485327313769,
+      "grad_norm": 2.1010115146636963,
+      "learning_rate": 4.398849459117645e-06,
+      "loss": 0.0687,
+      "mean_token_accuracy": 0.9798161029815674,
+      "num_tokens": 75459490.0,
+      "step": 9295
+    },
+    {
+      "entropy": 0.1841699182987213,
+      "epoch": 5.248306997742664,
+      "grad_norm": 1.935957431793213,
+      "learning_rate": 4.3982395251479705e-06,
+      "loss": 0.066,
+      "mean_token_accuracy": 0.9796889901161194,
+      "num_tokens": 75500096.0,
+      "step": 9300
+    },
+    {
+      "entropy": 0.16842667162418365,
+      "epoch": 5.251128668171558,
+      "grad_norm": 2.501762866973877,
+      "learning_rate": 4.3976293410502245e-06,
+      "loss": 0.0566,
+      "mean_token_accuracy": 0.9825867414474487,
+      "num_tokens": 75540820.0,
+      "step": 9305
+    },
+    {
+      "entropy": 0.18842605650424957,
+      "epoch": 5.253950338600451,
+      "grad_norm": 1.8708136081695557,
+      "learning_rate": 4.397018906944279e-06,
+      "loss": 0.0661,
+      "mean_token_accuracy": 0.9798398852348328,
+      "num_tokens": 75581478.0,
+      "step": 9310
+    },
+    {
+      "entropy": 0.17341751158237456,
+      "epoch": 5.256772009029345,
+      "grad_norm": 1.717519998550415,
+      "learning_rate": 4.3964082229500545e-06,
+      "loss": 0.0754,
+      "mean_token_accuracy": 0.977244210243225,
+      "num_tokens": 75622057.0,
+      "step": 9315
+    },
+    {
+      "entropy": 0.17311387956142427,
+      "epoch": 5.25959367945824,
+      "grad_norm": 2.0607640743255615,
+      "learning_rate": 4.39579728918752e-06,
+      "loss": 0.0712,
+      "mean_token_accuracy": 0.977792501449585,
+      "num_tokens": 75662674.0,
+      "step": 9320
+    },
+    {
+      "entropy": 0.17308418452739716,
+      "epoch": 5.262415349887133,
+      "grad_norm": 1.5914183855056763,
+      "learning_rate": 4.395186105776691e-06,
+      "loss": 0.0619,
+      "mean_token_accuracy": 0.9809646487236023,
+      "num_tokens": 75703277.0,
+      "step": 9325
+    },
+    {
+      "entropy": 0.15740045607089997,
+      "epoch": 5.265237020316027,
+      "grad_norm": 2.155003547668457,
+      "learning_rate": 4.394574672837637e-06,
+      "loss": 0.0599,
+      "mean_token_accuracy": 0.9820808887481689,
+      "num_tokens": 75743221.0,
+      "step": 9330
+    },
+    {
+      "entropy": 0.17217525243759155,
+      "epoch": 5.268058690744921,
+      "grad_norm": 2.087221384048462,
+      "learning_rate": 4.393962990490475e-06,
+      "loss": 0.0632,
+      "mean_token_accuracy": 0.9806132674217224,
+      "num_tokens": 75783998.0,
+      "step": 9335
+    },
+    {
+      "entropy": 0.1740059047937393,
+      "epoch": 5.270880361173815,
+      "grad_norm": 2.0336315631866455,
+      "learning_rate": 4.393351058855366e-06,
+      "loss": 0.0739,
+      "mean_token_accuracy": 0.9778852939605713,
+      "num_tokens": 75824530.0,
+      "step": 9340
+    },
+    {
+      "entropy": 0.19505321085453034,
+      "epoch": 5.273702031602709,
+      "grad_norm": 2.201986074447632,
+      "learning_rate": 4.392738878052528e-06,
+      "loss": 0.071,
+      "mean_token_accuracy": 0.9772972106933594,
+      "num_tokens": 75865197.0,
+      "step": 9345
+    },
+    {
+      "entropy": 0.17578763961791993,
+      "epoch": 5.276523702031603,
+      "grad_norm": 1.827558159828186,
+      "learning_rate": 4.392126448202223e-06,
+      "loss": 0.0617,
+      "mean_token_accuracy": 0.9813929557800293,
+      "num_tokens": 75905554.0,
+      "step": 9350
+    },
+    {
+      "entropy": 0.16351258158683776,
+      "epoch": 5.279345372460496,
+      "grad_norm": 2.258183479309082,
+      "learning_rate": 4.391513769424762e-06,
+      "loss": 0.0643,
+      "mean_token_accuracy": 0.9805266618728637,
+      "num_tokens": 75946320.0,
+      "step": 9355
+    },
+    {
+      "entropy": 0.17927370667457582,
+      "epoch": 5.282167042889391,
+      "grad_norm": 2.4332005977630615,
+      "learning_rate": 4.390900841840506e-06,
+      "loss": 0.0721,
+      "mean_token_accuracy": 0.9778699398040771,
+      "num_tokens": 75987141.0,
+      "step": 9360
+    },
+    {
+      "entropy": 0.16708399951457978,
+      "epoch": 5.284988713318285,
+      "grad_norm": 1.8501638174057007,
+      "learning_rate": 4.3902876655698666e-06,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.978826928138733,
+      "num_tokens": 76027705.0,
+      "step": 9365
+    },
+    {
+      "entropy": 0.18877400755882262,
+      "epoch": 5.287810383747178,
+      "grad_norm": 1.798592448234558,
+      "learning_rate": 4.3896742407332995e-06,
+      "loss": 0.0649,
+      "mean_token_accuracy": 0.9791201710700989,
+      "num_tokens": 76068472.0,
+      "step": 9370
+    },
+    {
+      "entropy": 0.1839359939098358,
+      "epoch": 5.290632054176072,
+      "grad_norm": 2.170732259750366,
+      "learning_rate": 4.389060567451313e-06,
+      "loss": 0.0688,
+      "mean_token_accuracy": 0.9787888526916504,
+      "num_tokens": 76109233.0,
+      "step": 9375
+    },
+    {
+      "entropy": 0.1790493279695511,
+      "epoch": 5.293453724604966,
+      "grad_norm": 1.9541733264923096,
+      "learning_rate": 4.388446645844465e-06,
+      "loss": 0.0708,
+      "mean_token_accuracy": 0.9779038906097413,
+      "num_tokens": 76149974.0,
+      "step": 9380
+    },
+    {
+      "entropy": 0.17843609154224396,
+      "epoch": 5.29627539503386,
+      "grad_norm": 1.8375680446624756,
+      "learning_rate": 4.387832476033358e-06,
+      "loss": 0.0692,
+      "mean_token_accuracy": 0.9785267472267151,
+      "num_tokens": 76190641.0,
+      "step": 9385
+    },
+    {
+      "entropy": 0.17534771859645842,
+      "epoch": 5.299097065462754,
+      "grad_norm": 2.224586009979248,
+      "learning_rate": 4.3872180581386485e-06,
+      "loss": 0.0762,
+      "mean_token_accuracy": 0.9763685941696167,
+      "num_tokens": 76231345.0,
+      "step": 9390
+    },
+    {
+      "entropy": 0.17150465846061708,
+      "epoch": 5.301918735891648,
+      "grad_norm": 2.095118999481201,
+      "learning_rate": 4.3866033922810355e-06,
+      "loss": 0.0719,
+      "mean_token_accuracy": 0.9783193111419678,
+      "num_tokens": 76272069.0,
+      "step": 9395
+    },
+    {
+      "entropy": 0.18151159584522247,
+      "epoch": 5.3047404063205414,
+      "grad_norm": 2.1895596981048584,
+      "learning_rate": 4.385988478581274e-06,
+      "loss": 0.0718,
+      "mean_token_accuracy": 0.9773975610733032,
+      "num_tokens": 76312608.0,
+      "step": 9400
+    },
+    {
+      "entropy": 0.19453785419464112,
+      "epoch": 5.307562076749436,
+      "grad_norm": 2.0922398567199707,
+      "learning_rate": 4.38537331716016e-06,
+      "loss": 0.0709,
+      "mean_token_accuracy": 0.9785321354866028,
+      "num_tokens": 76353353.0,
+      "step": 9405
+    },
+    {
+      "entropy": 0.18575883209705352,
+      "epoch": 5.31038374717833,
+      "grad_norm": 3.6228082180023193,
+      "learning_rate": 4.384757908138545e-06,
+      "loss": 0.0727,
+      "mean_token_accuracy": 0.9786053776741028,
+      "num_tokens": 76394028.0,
+      "step": 9410
+    },
+    {
+      "entropy": 0.18422328531742097,
+      "epoch": 5.313205417607223,
+      "grad_norm": 1.974879503250122,
+      "learning_rate": 4.3841422516373255e-06,
+      "loss": 0.0632,
+      "mean_token_accuracy": 0.9804863810539246,
+      "num_tokens": 76434694.0,
+      "step": 9415
+    },
+    {
+      "entropy": 0.17816999256610871,
+      "epoch": 5.316027088036117,
+      "grad_norm": 1.6818512678146362,
+      "learning_rate": 4.383526347777446e-06,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9786510825157165,
+      "num_tokens": 76475096.0,
+      "step": 9420
+    },
+    {
+      "entropy": 0.18150631487369537,
+      "epoch": 5.318848758465011,
+      "grad_norm": 1.903784990310669,
+      "learning_rate": 4.3829101966799025e-06,
+      "loss": 0.0732,
+      "mean_token_accuracy": 0.9780114173889161,
+      "num_tokens": 76515735.0,
+      "step": 9425
+    },
+    {
+      "entropy": 0.17466551959514617,
+      "epoch": 5.321670428893905,
+      "grad_norm": 1.8316190242767334,
+      "learning_rate": 4.382293798465738e-06,
+      "loss": 0.0729,
+      "mean_token_accuracy": 0.9783891916275025,
+      "num_tokens": 76556187.0,
+      "step": 9430
+    },
+    {
+      "entropy": 0.1856124997138977,
+      "epoch": 5.324492099322799,
+      "grad_norm": 2.0181853771209717,
+      "learning_rate": 4.381677153256042e-06,
+      "loss": 0.077,
+      "mean_token_accuracy": 0.9763926148414612,
+      "num_tokens": 76596840.0,
+      "step": 9435
+    },
+    {
+      "entropy": 0.16592672169208528,
+      "epoch": 5.327313769751693,
+      "grad_norm": 2.4706497192382812,
+      "learning_rate": 4.381060261171956e-06,
+      "loss": 0.0765,
+      "mean_token_accuracy": 0.9765696048736572,
+      "num_tokens": 76637424.0,
+      "step": 9440
+    },
+    {
+      "entropy": 0.17519499063491822,
+      "epoch": 5.3301354401805865,
+      "grad_norm": 1.9613145589828491,
+      "learning_rate": 4.38044312233467e-06,
+      "loss": 0.0649,
+      "mean_token_accuracy": 0.9802795290946961,
+      "num_tokens": 76678086.0,
+      "step": 9445
+    },
+    {
+      "entropy": 0.19584991335868834,
+      "epoch": 5.332957110609481,
+      "grad_norm": 1.8354933261871338,
+      "learning_rate": 4.379825736865419e-06,
+      "loss": 0.068,
+      "mean_token_accuracy": 0.9790003061294555,
+      "num_tokens": 76718348.0,
+      "step": 9450
+    },
+    {
+      "entropy": 0.17431385517120362,
+      "epoch": 5.335778781038375,
+      "grad_norm": 2.1092381477355957,
+      "learning_rate": 4.3792081048854875e-06,
+      "loss": 0.072,
+      "mean_token_accuracy": 0.9774962902069092,
+      "num_tokens": 76759148.0,
+      "step": 9455
+    },
+    {
+      "entropy": 0.1774332642555237,
+      "epoch": 5.3386004514672685,
+      "grad_norm": 2.28005313873291,
+      "learning_rate": 4.378590226516211e-06,
+      "loss": 0.0746,
+      "mean_token_accuracy": 0.9770366787910462,
+      "num_tokens": 76799597.0,
+      "step": 9460
+    },
+    {
+      "entropy": 0.17988061010837555,
+      "epoch": 5.341422121896162,
+      "grad_norm": 2.3315978050231934,
+      "learning_rate": 4.3779721018789735e-06,
+      "loss": 0.0781,
+      "mean_token_accuracy": 0.9758471369743347,
+      "num_tokens": 76840462.0,
+      "step": 9465
+    },
+    {
+      "entropy": 0.16905871629714966,
+      "epoch": 5.344243792325057,
+      "grad_norm": 2.204232931137085,
+      "learning_rate": 4.377353731095202e-06,
+      "loss": 0.0604,
+      "mean_token_accuracy": 0.9812986135482789,
+      "num_tokens": 76881153.0,
+      "step": 9470
+    },
+    {
+      "entropy": 0.17078822553157808,
+      "epoch": 5.3470654627539504,
+      "grad_norm": 1.9415490627288818,
+      "learning_rate": 4.376735114286378e-06,
+      "loss": 0.0697,
+      "mean_token_accuracy": 0.9775596261024475,
+      "num_tokens": 76922045.0,
+      "step": 9475
+    },
+    {
+      "entropy": 0.17137598097324372,
+      "epoch": 5.349887133182844,
+      "grad_norm": 1.894740343093872,
+      "learning_rate": 4.3761162515740276e-06,
+      "loss": 0.0631,
+      "mean_token_accuracy": 0.9806401252746582,
+      "num_tokens": 76962699.0,
+      "step": 9480
+    },
+    {
+      "entropy": 0.1780911296606064,
+      "epoch": 5.352708803611738,
+      "grad_norm": 1.8693207502365112,
+      "learning_rate": 4.375497143079726e-06,
+      "loss": 0.0728,
+      "mean_token_accuracy": 0.9785822868347168,
+      "num_tokens": 77003042.0,
+      "step": 9485
+    },
+    {
+      "entropy": 0.18042166531085968,
+      "epoch": 5.355530474040632,
+      "grad_norm": 1.8227202892303467,
+      "learning_rate": 4.3748777889250995e-06,
+      "loss": 0.0743,
+      "mean_token_accuracy": 0.9775307297706604,
+      "num_tokens": 77043699.0,
+      "step": 9490
+    },
+    {
+      "entropy": 0.1943429082632065,
+      "epoch": 5.358352144469526,
+      "grad_norm": 1.9916659593582153,
+      "learning_rate": 4.374258189231818e-06,
+      "loss": 0.0667,
+      "mean_token_accuracy": 0.9789568185806274,
+      "num_tokens": 77084496.0,
+      "step": 9495
+    },
+    {
+      "entropy": 0.19134209752082826,
+      "epoch": 5.36117381489842,
+      "grad_norm": 2.394291877746582,
+      "learning_rate": 4.3736383441216036e-06,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.9791797280311585,
+      "num_tokens": 77124779.0,
+      "step": 9500
+    },
+    {
+      "epoch": 5.36117381489842,
+      "eval_entropy": 0.22530123591423035,
+      "eval_loss": 0.03946574404835701,
+      "eval_mean_token_accuracy": 0.9888846278190613,
+      "eval_num_tokens": 77124779.0,
+      "eval_runtime": 0.1643,
+      "eval_samples_per_second": 24.339,
+      "eval_steps_per_second": 6.085,
+      "step": 9500
+    },
+    {
+      "entropy": 0.19695676267147064,
+      "epoch": 5.3639954853273135,
+      "grad_norm": 1.9585169553756714,
+      "learning_rate": 4.3730182537162235e-06,
+      "loss": 0.0782,
+      "mean_token_accuracy": 0.9759560108184815,
+      "num_tokens": 77165622.0,
+      "step": 9505
+    },
+    {
+      "entropy": 0.18560438454151154,
+      "epoch": 5.366817155756207,
+      "grad_norm": 2.292339563369751,
+      "learning_rate": 4.3723979181374964e-06,
+      "loss": 0.0743,
+      "mean_token_accuracy": 0.9771399736404419,
+      "num_tokens": 77206242.0,
+      "step": 9510
+    },
+    {
+      "entropy": 0.17267344892024994,
+      "epoch": 5.369638826185102,
+      "grad_norm": 2.0678932666778564,
+      "learning_rate": 4.371777337507285e-06,
+      "loss": 0.0692,
+      "mean_token_accuracy": 0.979387903213501,
+      "num_tokens": 77246858.0,
+      "step": 9515
+    },
+    {
+      "entropy": 0.18678390383720397,
+      "epoch": 5.3724604966139955,
+      "grad_norm": 2.482128858566284,
+      "learning_rate": 4.371156511947504e-06,
+      "loss": 0.0713,
+      "mean_token_accuracy": 0.9776126742362976,
+      "num_tokens": 77287179.0,
+      "step": 9520
+    },
+    {
+      "entropy": 0.16702256202697754,
+      "epoch": 5.375282167042889,
+      "grad_norm": 2.2040395736694336,
+      "learning_rate": 4.370535441580114e-06,
+      "loss": 0.0623,
+      "mean_token_accuracy": 0.981655502319336,
+      "num_tokens": 77327907.0,
+      "step": 9525
+    },
+    {
+      "entropy": 0.16512243151664735,
+      "epoch": 5.378103837471783,
+      "grad_norm": 1.9599778652191162,
+      "learning_rate": 4.369914126527125e-06,
+      "loss": 0.0643,
+      "mean_token_accuracy": 0.9801242351531982,
+      "num_tokens": 77368670.0,
+      "step": 9530
+    },
+    {
+      "entropy": 0.17791537344455718,
+      "epoch": 5.3809255079006775,
+      "grad_norm": 1.9389322996139526,
+      "learning_rate": 4.369292566910594e-06,
+      "loss": 0.0642,
+      "mean_token_accuracy": 0.9804812669754028,
+      "num_tokens": 77409415.0,
+      "step": 9535
+    },
+    {
+      "entropy": 0.18755693435668946,
+      "epoch": 5.383747178329571,
+      "grad_norm": 1.7628971338272095,
+      "learning_rate": 4.368670762852626e-06,
+      "loss": 0.0655,
+      "mean_token_accuracy": 0.9798348426818848,
+      "num_tokens": 77450190.0,
+      "step": 9540
+    },
+    {
+      "entropy": 0.1958528518676758,
+      "epoch": 5.386568848758465,
+      "grad_norm": 2.0312836170196533,
+      "learning_rate": 4.368048714475375e-06,
+      "loss": 0.0711,
+      "mean_token_accuracy": 0.9780890107154846,
+      "num_tokens": 77490984.0,
+      "step": 9545
+    },
+    {
+      "entropy": 0.1894278347492218,
+      "epoch": 5.389390519187359,
+      "grad_norm": 2.1138522624969482,
+      "learning_rate": 4.367426421901042e-06,
+      "loss": 0.0687,
+      "mean_token_accuracy": 0.9790720820426941,
+      "num_tokens": 77531707.0,
+      "step": 9550
+    },
+    {
+      "entropy": 0.18264504075050353,
+      "epoch": 5.392212189616253,
+      "grad_norm": 2.236631155014038,
+      "learning_rate": 4.366803885251879e-06,
+      "loss": 0.068,
+      "mean_token_accuracy": 0.9791576147079468,
+      "num_tokens": 77572138.0,
+      "step": 9555
+    },
+    {
+      "entropy": 0.17964340448379518,
+      "epoch": 5.395033860045147,
+      "grad_norm": 1.9369723796844482,
+      "learning_rate": 4.366181104650179e-06,
+      "loss": 0.0726,
+      "mean_token_accuracy": 0.978190791606903,
+      "num_tokens": 77612964.0,
+      "step": 9560
+    },
+    {
+      "entropy": 0.19620463848114014,
+      "epoch": 5.3978555304740405,
+      "grad_norm": 2.114414930343628,
+      "learning_rate": 4.36555808021829e-06,
+      "loss": 0.0781,
+      "mean_token_accuracy": 0.975774621963501,
+      "num_tokens": 77653620.0,
+      "step": 9565
+    },
+    {
+      "entropy": 0.18174751698970795,
+      "epoch": 5.400677200902934,
+      "grad_norm": 2.574272394180298,
+      "learning_rate": 4.364934812078606e-06,
+      "loss": 0.071,
+      "mean_token_accuracy": 0.9776658058166504,
+      "num_tokens": 77694231.0,
+      "step": 9570
+    },
+    {
+      "entropy": 0.19385533034801483,
+      "epoch": 5.403498871331829,
+      "grad_norm": 1.886120319366455,
+      "learning_rate": 4.364311300353567e-06,
+      "loss": 0.0706,
+      "mean_token_accuracy": 0.9778653621673584,
+      "num_tokens": 77734981.0,
+      "step": 9575
+    },
+    {
+      "entropy": 0.1768470048904419,
+      "epoch": 5.4063205417607225,
+      "grad_norm": 1.6945990324020386,
+      "learning_rate": 4.363687545165661e-06,
+      "loss": 0.0748,
+      "mean_token_accuracy": 0.9762227416038514,
+      "num_tokens": 77775525.0,
+      "step": 9580
+    },
+    {
+      "entropy": 0.18447025418281554,
+      "epoch": 5.409142212189616,
+      "grad_norm": 2.226775646209717,
+      "learning_rate": 4.363063546637426e-06,
+      "loss": 0.0772,
+      "mean_token_accuracy": 0.9765905261039733,
+      "num_tokens": 77816146.0,
+      "step": 9585
+    },
+    {
+      "entropy": 0.20254374146461487,
+      "epoch": 5.41196388261851,
+      "grad_norm": 1.4906258583068848,
+      "learning_rate": 4.3624393048914465e-06,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.9796987056732178,
+      "num_tokens": 77856760.0,
+      "step": 9590
+    },
+    {
+      "entropy": 0.20228754878044128,
+      "epoch": 5.414785553047404,
+      "grad_norm": 2.128209114074707,
+      "learning_rate": 4.361814820050355e-06,
+      "loss": 0.074,
+      "mean_token_accuracy": 0.9770419836044312,
+      "num_tokens": 77897182.0,
+      "step": 9595
+    },
+    {
+      "entropy": 0.18532134592533112,
+      "epoch": 5.417607223476298,
+      "grad_norm": 1.994026780128479,
+      "learning_rate": 4.36119009223683e-06,
+      "loss": 0.0698,
+      "mean_token_accuracy": 0.9782418012619019,
+      "num_tokens": 77937769.0,
+      "step": 9600
+    },
+    {
+      "entropy": 0.17942964434623718,
+      "epoch": 5.420428893905192,
+      "grad_norm": 1.7702689170837402,
+      "learning_rate": 4.3605651215736025e-06,
+      "loss": 0.0697,
+      "mean_token_accuracy": 0.9785507321357727,
+      "num_tokens": 77978608.0,
+      "step": 9605
+    },
+    {
+      "entropy": 0.18152170181274413,
+      "epoch": 5.423250564334086,
+      "grad_norm": 2.2829222679138184,
+      "learning_rate": 4.359939908183445e-06,
+      "loss": 0.0748,
+      "mean_token_accuracy": 0.9763474941253663,
+      "num_tokens": 78019201.0,
+      "step": 9610
+    },
+    {
+      "entropy": 0.16868273317813873,
+      "epoch": 5.426072234762979,
+      "grad_norm": 1.9242933988571167,
+      "learning_rate": 4.3593144521891825e-06,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9792532801628113,
+      "num_tokens": 78059853.0,
+      "step": 9615
+    },
+    {
+      "entropy": 0.17837615609169005,
+      "epoch": 5.428893905191874,
+      "grad_norm": 1.9785670042037964,
+      "learning_rate": 4.358688753713685e-06,
+      "loss": 0.063,
+      "mean_token_accuracy": 0.9807116985321045,
+      "num_tokens": 78100351.0,
+      "step": 9620
+    },
+    {
+      "entropy": 0.16789911687374115,
+      "epoch": 5.431715575620768,
+      "grad_norm": 1.8135242462158203,
+      "learning_rate": 4.358062812879873e-06,
+      "loss": 0.0667,
+      "mean_token_accuracy": 0.979233467578888,
+      "num_tokens": 78140931.0,
+      "step": 9625
+    },
+    {
+      "entropy": 0.16792088150978088,
+      "epoch": 5.434537246049661,
+      "grad_norm": 2.1166341304779053,
+      "learning_rate": 4.357436629810709e-06,
+      "loss": 0.0682,
+      "mean_token_accuracy": 0.9792353987693787,
+      "num_tokens": 78181470.0,
+      "step": 9630
+    },
+    {
+      "entropy": 0.18325534760951995,
+      "epoch": 5.437358916478555,
+      "grad_norm": 1.728598952293396,
+      "learning_rate": 4.35681020462921e-06,
+      "loss": 0.0747,
+      "mean_token_accuracy": 0.977009117603302,
+      "num_tokens": 78222347.0,
+      "step": 9635
+    },
+    {
+      "entropy": 0.18267615139484406,
+      "epoch": 5.4401805869074495,
+      "grad_norm": 1.7328113317489624,
+      "learning_rate": 4.356183537458436e-06,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9778196811676025,
+      "num_tokens": 78262982.0,
+      "step": 9640
+    },
+    {
+      "entropy": 0.180474916100502,
+      "epoch": 5.443002257336343,
+      "grad_norm": 1.5890209674835205,
+      "learning_rate": 4.3555566284214955e-06,
+      "loss": 0.0663,
+      "mean_token_accuracy": 0.9788239240646363,
+      "num_tokens": 78303575.0,
+      "step": 9645
+    },
+    {
+      "entropy": 0.20339135229587554,
+      "epoch": 5.445823927765237,
+      "grad_norm": 1.9945831298828125,
+      "learning_rate": 4.354929477641547e-06,
+      "loss": 0.0792,
+      "mean_token_accuracy": 0.9748901128768921,
+      "num_tokens": 78343987.0,
+      "step": 9650
+    },
+    {
+      "entropy": 0.18225070238113403,
+      "epoch": 5.448645598194131,
+      "grad_norm": 1.9023123979568481,
+      "learning_rate": 4.354302085241791e-06,
+      "loss": 0.072,
+      "mean_token_accuracy": 0.9779296040534973,
+      "num_tokens": 78384670.0,
+      "step": 9655
+    },
+    {
+      "entropy": 0.17491951882839202,
+      "epoch": 5.451467268623025,
+      "grad_norm": 1.9437018632888794,
+      "learning_rate": 4.353674451345481e-06,
+      "loss": 0.0649,
+      "mean_token_accuracy": 0.9799412608146667,
+      "num_tokens": 78425444.0,
+      "step": 9660
+    },
+    {
+      "entropy": 0.19293942749500276,
+      "epoch": 5.454288939051919,
+      "grad_norm": 2.0638647079467773,
+      "learning_rate": 4.353046576075915e-06,
+      "loss": 0.0706,
+      "mean_token_accuracy": 0.9783964395523072,
+      "num_tokens": 78465938.0,
+      "step": 9665
+    },
+    {
+      "entropy": 0.17496614456176757,
+      "epoch": 5.457110609480813,
+      "grad_norm": 2.2454192638397217,
+      "learning_rate": 4.35241845955644e-06,
+      "loss": 0.0733,
+      "mean_token_accuracy": 0.9771980404853821,
+      "num_tokens": 78506288.0,
+      "step": 9670
+    },
+    {
+      "entropy": 0.18418402671813966,
+      "epoch": 5.459932279909706,
+      "grad_norm": 1.9852421283721924,
+      "learning_rate": 4.3517901019104494e-06,
+      "loss": 0.0624,
+      "mean_token_accuracy": 0.9815483212471008,
+      "num_tokens": 78546897.0,
+      "step": 9675
+    },
+    {
+      "entropy": 0.17411540746688842,
+      "epoch": 5.4627539503386,
+      "grad_norm": 2.471421957015991,
+      "learning_rate": 4.351161503261384e-06,
+      "loss": 0.0593,
+      "mean_token_accuracy": 0.981855833530426,
+      "num_tokens": 78587494.0,
+      "step": 9680
+    },
+    {
+      "entropy": 0.18723950386047364,
+      "epoch": 5.465575620767495,
+      "grad_norm": 2.042299509048462,
+      "learning_rate": 4.3505326637327315e-06,
+      "loss": 0.0789,
+      "mean_token_accuracy": 0.9755418181419373,
+      "num_tokens": 78628058.0,
+      "step": 9685
+    },
+    {
+      "entropy": 0.1705529898405075,
+      "epoch": 5.468397291196388,
+      "grad_norm": 2.377067804336548,
+      "learning_rate": 4.3499035834480275e-06,
+      "loss": 0.0714,
+      "mean_token_accuracy": 0.9782310485839844,
+      "num_tokens": 78668915.0,
+      "step": 9690
+    },
+    {
+      "entropy": 0.1720373123884201,
+      "epoch": 5.471218961625282,
+      "grad_norm": 2.115959405899048,
+      "learning_rate": 4.349274262530856e-06,
+      "loss": 0.0724,
+      "mean_token_accuracy": 0.9783347487449646,
+      "num_tokens": 78709535.0,
+      "step": 9695
+    },
+    {
+      "entropy": 0.1738339751958847,
+      "epoch": 5.474040632054176,
+      "grad_norm": 1.7836189270019531,
+      "learning_rate": 4.348644701104845e-06,
+      "loss": 0.067,
+      "mean_token_accuracy": 0.9795932054519654,
+      "num_tokens": 78750224.0,
+      "step": 9700
+    },
+    {
+      "entropy": 0.19387414753437043,
+      "epoch": 5.47686230248307,
+      "grad_norm": 1.597865343093872,
+      "learning_rate": 4.348014899293675e-06,
+      "loss": 0.0782,
+      "mean_token_accuracy": 0.9758282899856567,
+      "num_tokens": 78790822.0,
+      "step": 9705
+    },
+    {
+      "entropy": 0.1638443261384964,
+      "epoch": 5.479683972911964,
+      "grad_norm": 1.7164875268936157,
+      "learning_rate": 4.3473848572210685e-06,
+      "loss": 0.0667,
+      "mean_token_accuracy": 0.9790258288383484,
+      "num_tokens": 78831384.0,
+      "step": 9710
+    },
+    {
+      "entropy": 0.17148119807243348,
+      "epoch": 5.482505643340858,
+      "grad_norm": 1.7814518213272095,
+      "learning_rate": 4.346754575010798e-06,
+      "loss": 0.055,
+      "mean_token_accuracy": 0.9819474458694458,
+      "num_tokens": 78871893.0,
+      "step": 9715
+    },
+    {
+      "entropy": 0.1933718204498291,
+      "epoch": 5.485327313769751,
+      "grad_norm": 2.326171398162842,
+      "learning_rate": 4.346124052786682e-06,
+      "loss": 0.0807,
+      "mean_token_accuracy": 0.9755822420120239,
+      "num_tokens": 78912523.0,
+      "step": 9720
+    },
+    {
+      "entropy": 0.17635531425476075,
+      "epoch": 5.488148984198646,
+      "grad_norm": 1.7191340923309326,
+      "learning_rate": 4.3454932906725875e-06,
+      "loss": 0.0705,
+      "mean_token_accuracy": 0.9771863937377929,
+      "num_tokens": 78952985.0,
+      "step": 9725
+    },
+    {
+      "entropy": 0.17937880754470825,
+      "epoch": 5.49097065462754,
+      "grad_norm": 2.046811819076538,
+      "learning_rate": 4.3448622887924265e-06,
+      "loss": 0.0618,
+      "mean_token_accuracy": 0.981409227848053,
+      "num_tokens": 78993356.0,
+      "step": 9730
+    },
+    {
+      "entropy": 0.19099608957767486,
+      "epoch": 5.493792325056433,
+      "grad_norm": 2.5957789421081543,
+      "learning_rate": 4.3442310472701615e-06,
+      "loss": 0.0765,
+      "mean_token_accuracy": 0.9757651925086975,
+      "num_tokens": 79034026.0,
+      "step": 9735
+    },
+    {
+      "entropy": 0.1835557132959366,
+      "epoch": 5.496613995485327,
+      "grad_norm": 2.360753059387207,
+      "learning_rate": 4.343599566229799e-06,
+      "loss": 0.0732,
+      "mean_token_accuracy": 0.9772834777832031,
+      "num_tokens": 79074662.0,
+      "step": 9740
+    },
+    {
+      "entropy": 0.17511331737041474,
+      "epoch": 5.499435665914222,
+      "grad_norm": 2.0439140796661377,
+      "learning_rate": 4.342967845795393e-06,
+      "loss": 0.0642,
+      "mean_token_accuracy": 0.9799216628074646,
+      "num_tokens": 79115192.0,
+      "step": 9745
+    },
+    {
+      "entropy": 0.18398658633232118,
+      "epoch": 5.502257336343115,
+      "grad_norm": 1.601022481918335,
+      "learning_rate": 4.342335886091045e-06,
+      "loss": 0.0729,
+      "mean_token_accuracy": 0.9781609892845153,
+      "num_tokens": 79155981.0,
+      "step": 9750
+    },
+    {
+      "entropy": 0.18661101460456847,
+      "epoch": 5.505079006772009,
+      "grad_norm": 2.0731983184814453,
+      "learning_rate": 4.341703687240903e-06,
+      "loss": 0.0754,
+      "mean_token_accuracy": 0.9759285807609558,
+      "num_tokens": 79196610.0,
+      "step": 9755
+    },
+    {
+      "entropy": 0.19010823965072632,
+      "epoch": 5.507900677200903,
+      "grad_norm": 1.9401448965072632,
+      "learning_rate": 4.341071249369164e-06,
+      "loss": 0.0742,
+      "mean_token_accuracy": 0.9756409406661988,
+      "num_tokens": 79237061.0,
+      "step": 9760
+    },
+    {
+      "entropy": 0.17404641211032867,
+      "epoch": 5.510722347629796,
+      "grad_norm": 2.2344675064086914,
+      "learning_rate": 4.340438572600072e-06,
+      "loss": 0.0677,
+      "mean_token_accuracy": 0.9782544851303101,
+      "num_tokens": 79277948.0,
+      "step": 9765
+    },
+    {
+      "entropy": 0.1966948091983795,
+      "epoch": 5.513544018058691,
+      "grad_norm": 2.11859130859375,
+      "learning_rate": 4.3398056570579125e-06,
+      "loss": 0.0865,
+      "mean_token_accuracy": 0.9740624785423279,
+      "num_tokens": 79317957.0,
+      "step": 9770
+    },
+    {
+      "entropy": 0.17077078223228453,
+      "epoch": 5.516365688487585,
+      "grad_norm": 1.9903873205184937,
+      "learning_rate": 4.339172502867023e-06,
+      "loss": 0.0667,
+      "mean_token_accuracy": 0.9797818660736084,
+      "num_tokens": 79358659.0,
+      "step": 9775
+    },
+    {
+      "entropy": 0.17696953415870667,
+      "epoch": 5.519187358916478,
+      "grad_norm": 2.100498676300049,
+      "learning_rate": 4.338539110151789e-06,
+      "loss": 0.0834,
+      "mean_token_accuracy": 0.974258816242218,
+      "num_tokens": 79399262.0,
+      "step": 9780
+    },
+    {
+      "entropy": 0.17644532322883605,
+      "epoch": 5.522009029345372,
+      "grad_norm": 2.0375876426696777,
+      "learning_rate": 4.337905479036639e-06,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9780573964118957,
+      "num_tokens": 79439526.0,
+      "step": 9785
+    },
+    {
+      "entropy": 0.18530842363834382,
+      "epoch": 5.524830699774267,
+      "grad_norm": 2.0761780738830566,
+      "learning_rate": 4.33727160964605e-06,
+      "loss": 0.0706,
+      "mean_token_accuracy": 0.9781399369239807,
+      "num_tokens": 79480123.0,
+      "step": 9790
+    },
+    {
+      "entropy": 0.19907425343990326,
+      "epoch": 5.52765237020316,
+      "grad_norm": 1.8196872472763062,
+      "learning_rate": 4.336637502104545e-06,
+      "loss": 0.0821,
+      "mean_token_accuracy": 0.9743606328964234,
+      "num_tokens": 79520700.0,
+      "step": 9795
+    },
+    {
+      "entropy": 0.18495319783687592,
+      "epoch": 5.530474040632054,
+      "grad_norm": 1.844511866569519,
+      "learning_rate": 4.336003156536696e-06,
+      "loss": 0.0689,
+      "mean_token_accuracy": 0.9784881234169006,
+      "num_tokens": 79561218.0,
+      "step": 9800
+    },
+    {
+      "entropy": 0.17770458459854127,
+      "epoch": 5.533295711060948,
+      "grad_norm": 2.0217959880828857,
+      "learning_rate": 4.335368573067118e-06,
+      "loss": 0.0743,
+      "mean_token_accuracy": 0.9774918794631958,
+      "num_tokens": 79601835.0,
+      "step": 9805
+    },
+    {
+      "entropy": 0.18182687163352967,
+      "epoch": 5.536117381489842,
+      "grad_norm": 2.815741777420044,
+      "learning_rate": 4.334733751820478e-06,
+      "loss": 0.0754,
+      "mean_token_accuracy": 0.9765426874160766,
+      "num_tokens": 79642074.0,
+      "step": 9810
+    },
+    {
+      "entropy": 0.17728038728237153,
+      "epoch": 5.538939051918736,
+      "grad_norm": 1.9506757259368896,
+      "learning_rate": 4.334098692921484e-06,
+      "loss": 0.0674,
+      "mean_token_accuracy": 0.9793436288833618,
+      "num_tokens": 79682666.0,
+      "step": 9815
+    },
+    {
+      "entropy": 0.1820712596178055,
+      "epoch": 5.54176072234763,
+      "grad_norm": 2.2737324237823486,
+      "learning_rate": 4.333463396494896e-06,
+      "loss": 0.0664,
+      "mean_token_accuracy": 0.9783164978027343,
+      "num_tokens": 79723398.0,
+      "step": 9820
+    },
+    {
+      "entropy": 0.1908366858959198,
+      "epoch": 5.544582392776523,
+      "grad_norm": 1.7559250593185425,
+      "learning_rate": 4.332827862665515e-06,
+      "loss": 0.0673,
+      "mean_token_accuracy": 0.9796145915985107,
+      "num_tokens": 79764063.0,
+      "step": 9825
+    },
+    {
+      "entropy": 0.18218895196914672,
+      "epoch": 5.547404063205418,
+      "grad_norm": 2.057884693145752,
+      "learning_rate": 4.332192091558195e-06,
+      "loss": 0.084,
+      "mean_token_accuracy": 0.9729039072990417,
+      "num_tokens": 79804649.0,
+      "step": 9830
+    },
+    {
+      "entropy": 0.19951559603214264,
+      "epoch": 5.550225733634312,
+      "grad_norm": 2.015651226043701,
+      "learning_rate": 4.331556083297831e-06,
+      "loss": 0.0819,
+      "mean_token_accuracy": 0.9746386528015136,
+      "num_tokens": 79845171.0,
+      "step": 9835
+    },
+    {
+      "entropy": 0.17591853737831115,
+      "epoch": 5.553047404063205,
+      "grad_norm": 2.3496806621551514,
+      "learning_rate": 4.330919838009368e-06,
+      "loss": 0.0591,
+      "mean_token_accuracy": 0.9815265893936157,
+      "num_tokens": 79885183.0,
+      "step": 9840
+    },
+    {
+      "entropy": 0.176205512881279,
+      "epoch": 5.555869074492099,
+      "grad_norm": 2.1586620807647705,
+      "learning_rate": 4.330283355817796e-06,
+      "loss": 0.0767,
+      "mean_token_accuracy": 0.9763123154640198,
+      "num_tokens": 79925909.0,
+      "step": 9845
+    },
+    {
+      "entropy": 0.17635221481323243,
+      "epoch": 5.558690744920993,
+      "grad_norm": 2.0332553386688232,
+      "learning_rate": 4.329646636848151e-06,
+      "loss": 0.0669,
+      "mean_token_accuracy": 0.9792371988296509,
+      "num_tokens": 79966534.0,
+      "step": 9850
+    },
+    {
+      "entropy": 0.18909758031368257,
+      "epoch": 5.561512415349887,
+      "grad_norm": 1.8824989795684814,
+      "learning_rate": 4.3290096812255185e-06,
+      "loss": 0.0767,
+      "mean_token_accuracy": 0.976757287979126,
+      "num_tokens": 80007224.0,
+      "step": 9855
+    },
+    {
+      "entropy": 0.18777174949645997,
+      "epoch": 5.564334085778781,
+      "grad_norm": 2.0460257530212402,
+      "learning_rate": 4.328372489075028e-06,
+      "loss": 0.0755,
+      "mean_token_accuracy": 0.9778550267219543,
+      "num_tokens": 80047897.0,
+      "step": 9860
+    },
+    {
+      "entropy": 0.18754212260246278,
+      "epoch": 5.567155756207675,
+      "grad_norm": 2.3515195846557617,
+      "learning_rate": 4.327735060521855e-06,
+      "loss": 0.0802,
+      "mean_token_accuracy": 0.9749362111091614,
+      "num_tokens": 80088428.0,
+      "step": 9865
+    },
+    {
+      "entropy": 0.17782265245914458,
+      "epoch": 5.5699774266365685,
+      "grad_norm": 1.882233738899231,
+      "learning_rate": 4.3270973956912225e-06,
+      "loss": 0.0696,
+      "mean_token_accuracy": 0.9777340531349182,
+      "num_tokens": 80129154.0,
+      "step": 9870
+    },
+    {
+      "entropy": 0.16835958659648895,
+      "epoch": 5.572799097065463,
+      "grad_norm": 1.7510329484939575,
+      "learning_rate": 4.326459494708401e-06,
+      "loss": 0.0638,
+      "mean_token_accuracy": 0.9795340061187744,
+      "num_tokens": 80169798.0,
+      "step": 9875
+    },
+    {
+      "entropy": 0.17615444958209991,
+      "epoch": 5.575620767494357,
+      "grad_norm": 1.4721903800964355,
+      "learning_rate": 4.325821357698705e-06,
+      "loss": 0.0693,
+      "mean_token_accuracy": 0.9783699035644531,
+      "num_tokens": 80210503.0,
+      "step": 9880
+    },
+    {
+      "entropy": 0.1756734073162079,
+      "epoch": 5.5784424379232505,
+      "grad_norm": 2.1826069355010986,
+      "learning_rate": 4.325182984787499e-06,
+      "loss": 0.067,
+      "mean_token_accuracy": 0.9788246631622315,
+      "num_tokens": 80251068.0,
+      "step": 9885
+    },
+    {
+      "entropy": 0.19139577448368073,
+      "epoch": 5.581264108352144,
+      "grad_norm": 2.021207094192505,
+      "learning_rate": 4.324544376100188e-06,
+      "loss": 0.0744,
+      "mean_token_accuracy": 0.9779517769813537,
+      "num_tokens": 80291849.0,
+      "step": 9890
+    },
+    {
+      "entropy": 0.17277522683143615,
+      "epoch": 5.584085778781039,
+      "grad_norm": 2.1115806102752686,
+      "learning_rate": 4.323905531762229e-06,
+      "loss": 0.0694,
+      "mean_token_accuracy": 0.9781728863716126,
+      "num_tokens": 80332416.0,
+      "step": 9895
+    },
+    {
+      "entropy": 0.17857773303985597,
+      "epoch": 5.586907449209932,
+      "grad_norm": 1.9007774591445923,
+      "learning_rate": 4.323266451899122e-06,
+      "loss": 0.0666,
+      "mean_token_accuracy": 0.9787236452102661,
+      "num_tokens": 80373154.0,
+      "step": 9900
+    },
+    {
+      "entropy": 0.18921782374382018,
+      "epoch": 5.589729119638826,
+      "grad_norm": 1.874495267868042,
+      "learning_rate": 4.322627136636415e-06,
+      "loss": 0.0715,
+      "mean_token_accuracy": 0.9781720876693726,
+      "num_tokens": 80413856.0,
+      "step": 9905
+    },
+    {
+      "entropy": 0.18341380059719087,
+      "epoch": 5.59255079006772,
+      "grad_norm": 2.1999471187591553,
+      "learning_rate": 4.321987586099702e-06,
+      "loss": 0.0603,
+      "mean_token_accuracy": 0.9813642621040344,
+      "num_tokens": 80454655.0,
+      "step": 9910
+    },
+    {
+      "entropy": 0.17606990933418273,
+      "epoch": 5.595372460496614,
+      "grad_norm": 2.1019020080566406,
+      "learning_rate": 4.32134780041462e-06,
+      "loss": 0.0651,
+      "mean_token_accuracy": 0.9796467542648315,
+      "num_tokens": 80495377.0,
+      "step": 9915
+    },
+    {
+      "entropy": 0.18233582377433777,
+      "epoch": 5.598194130925508,
+      "grad_norm": 1.8605166673660278,
+      "learning_rate": 4.320707779706859e-06,
+      "loss": 0.0701,
+      "mean_token_accuracy": 0.9781240105628968,
+      "num_tokens": 80536019.0,
+      "step": 9920
+    },
+    {
+      "entropy": 0.18825688362121581,
+      "epoch": 5.601015801354402,
+      "grad_norm": 1.838043451309204,
+      "learning_rate": 4.320067524102149e-06,
+      "loss": 0.0755,
+      "mean_token_accuracy": 0.9771787285804748,
+      "num_tokens": 80576772.0,
+      "step": 9925
+    },
+    {
+      "entropy": 0.1714703172445297,
+      "epoch": 5.6038374717832955,
+      "grad_norm": 2.1907753944396973,
+      "learning_rate": 4.319427033726268e-06,
+      "loss": 0.0685,
+      "mean_token_accuracy": 0.9775472521781922,
+      "num_tokens": 80617419.0,
+      "step": 9930
+    },
+    {
+      "entropy": 0.1989262282848358,
+      "epoch": 5.606659142212189,
+      "grad_norm": 2.381535530090332,
+      "learning_rate": 4.3187863087050405e-06,
+      "loss": 0.0764,
+      "mean_token_accuracy": 0.9764021635055542,
+      "num_tokens": 80657689.0,
+      "step": 9935
+    },
+    {
+      "entropy": 0.1726620763540268,
+      "epoch": 5.609480812641084,
+      "grad_norm": 2.249459743499756,
+      "learning_rate": 4.318145349164339e-06,
+      "loss": 0.067,
+      "mean_token_accuracy": 0.9787798404693604,
+      "num_tokens": 80698324.0,
+      "step": 9940
+    },
+    {
+      "entropy": 0.17213599681854247,
+      "epoch": 5.6123024830699775,
+      "grad_norm": 1.9886410236358643,
+      "learning_rate": 4.3175041552300775e-06,
+      "loss": 0.062,
+      "mean_token_accuracy": 0.9811339974403381,
+      "num_tokens": 80738375.0,
+      "step": 9945
+    },
+    {
+      "entropy": 0.16617416441440583,
+      "epoch": 5.615124153498871,
+      "grad_norm": 2.2124135494232178,
+      "learning_rate": 4.31686272702822e-06,
+      "loss": 0.0702,
+      "mean_token_accuracy": 0.978109622001648,
+      "num_tokens": 80778942.0,
+      "step": 9950
+    },
+    {
+      "entropy": 0.1947992593050003,
+      "epoch": 5.617945823927765,
+      "grad_norm": 2.2856040000915527,
+      "learning_rate": 4.316221064684775e-06,
+      "loss": 0.0789,
+      "mean_token_accuracy": 0.9748259305953979,
+      "num_tokens": 80819710.0,
+      "step": 9955
+    },
+    {
+      "entropy": 0.19779794812202453,
+      "epoch": 5.6207674943566595,
+      "grad_norm": 3.436311721801758,
+      "learning_rate": 4.3155791683257965e-06,
+      "loss": 0.074,
+      "mean_token_accuracy": 0.9769353747367859,
+      "num_tokens": 80860043.0,
+      "step": 9960
+    },
+    {
+      "entropy": 0.18495135605335236,
+      "epoch": 5.623589164785553,
+      "grad_norm": 2.0917840003967285,
+      "learning_rate": 4.314937038077386e-06,
+      "loss": 0.0708,
+      "mean_token_accuracy": 0.9777469873428345,
+      "num_tokens": 80900666.0,
+      "step": 9965
+    },
+    {
+      "entropy": 0.1749154359102249,
+      "epoch": 5.626410835214447,
+      "grad_norm": 1.831464409828186,
+      "learning_rate": 4.314294674065689e-06,
+      "loss": 0.0673,
+      "mean_token_accuracy": 0.9792719483375549,
+      "num_tokens": 80941260.0,
+      "step": 9970
+    },
+    {
+      "entropy": 0.17395665049552916,
+      "epoch": 5.6292325056433405,
+      "grad_norm": 2.2329459190368652,
+      "learning_rate": 4.3136520764168996e-06,
+      "loss": 0.0716,
+      "mean_token_accuracy": 0.9781460523605346,
+      "num_tokens": 80981968.0,
+      "step": 9975
+    },
+    {
+      "entropy": 0.19316057562828065,
+      "epoch": 5.632054176072235,
+      "grad_norm": 1.6789857149124146,
+      "learning_rate": 4.3130092452572545e-06,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9778452634811401,
+      "num_tokens": 81022614.0,
+      "step": 9980
+    },
+    {
+      "entropy": 0.17241209149360656,
+      "epoch": 5.634875846501129,
+      "grad_norm": 1.6188982725143433,
+      "learning_rate": 4.312366180713039e-06,
+      "loss": 0.0668,
+      "mean_token_accuracy": 0.9798883080482483,
+      "num_tokens": 81063456.0,
+      "step": 9985
+    },
+    {
+      "entropy": 0.18288934230804443,
+      "epoch": 5.6376975169300225,
+      "grad_norm": 12.512042999267578,
+      "learning_rate": 4.311722882910584e-06,
+      "loss": 0.0767,
+      "mean_token_accuracy": 0.9758876323699951,
+      "num_tokens": 81104260.0,
+      "step": 9990
+    },
+    {
+      "entropy": 0.17049558758735656,
+      "epoch": 5.640519187358916,
+      "grad_norm": 2.0652711391448975,
+      "learning_rate": 4.3110793519762625e-06,
+      "loss": 0.0753,
+      "mean_token_accuracy": 0.9753929495811462,
+      "num_tokens": 81145141.0,
+      "step": 9995
+    },
+    {
+      "entropy": 0.17534006536006927,
+      "epoch": 5.643340857787811,
+      "grad_norm": 1.9431458711624146,
+      "learning_rate": 4.3104355880365e-06,
+      "loss": 0.0686,
+      "mean_token_accuracy": 0.9783696532249451,
+      "num_tokens": 81185597.0,
+      "step": 10000
+    },
+    {
+      "epoch": 5.643340857787811,
+      "eval_entropy": 0.2248619794845581,
+      "eval_loss": 0.03424257040023804,
+      "eval_mean_token_accuracy": 0.9908010959625244,
+      "eval_num_tokens": 81185597.0,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 24.396,
+      "eval_steps_per_second": 6.099,
+      "step": 10000
+    },
+    {
+      "entropy": 0.17620908617973327,
+      "epoch": 5.6461625282167045,
+      "grad_norm": 2.0070059299468994,
+      "learning_rate": 4.3097915912177615e-06,
+      "loss": 0.0726,
+      "mean_token_accuracy": 0.9765960216522217,
+      "num_tokens": 81225965.0,
+      "step": 10005
+    },
+    {
+      "entropy": 0.19689911603927612,
+      "epoch": 5.648984198645598,
+      "grad_norm": 2.1444599628448486,
+      "learning_rate": 4.30914736164656e-06,
+      "loss": 0.0715,
+      "mean_token_accuracy": 0.9776429057121276,
+      "num_tokens": 81266144.0,
+      "step": 10010
+    },
+    {
+      "entropy": 0.1702759474515915,
+      "epoch": 5.651805869074492,
+      "grad_norm": 1.7831538915634155,
+      "learning_rate": 4.308502899449456e-06,
+      "loss": 0.0646,
+      "mean_token_accuracy": 0.9793395638465882,
+      "num_tokens": 81306666.0,
+      "step": 10015
+    },
+    {
+      "entropy": 0.19171405136585234,
+      "epoch": 5.654627539503386,
+      "grad_norm": 6.789673328399658,
+      "learning_rate": 4.307858204753054e-06,
+      "loss": 0.0827,
+      "mean_token_accuracy": 0.9753363370895386,
+      "num_tokens": 81347319.0,
+      "step": 10020
+    },
+    {
+      "entropy": 0.18764612972736358,
+      "epoch": 5.65744920993228,
+      "grad_norm": 2.124255895614624,
+      "learning_rate": 4.3072132776840035e-06,
+      "loss": 0.0692,
+      "mean_token_accuracy": 0.9783503413200378,
+      "num_tokens": 81387613.0,
+      "step": 10025
+    },
+    {
+      "entropy": 0.17401687800884247,
+      "epoch": 5.660270880361174,
+      "grad_norm": 2.335118293762207,
+      "learning_rate": 4.306568118369001e-06,
+      "loss": 0.0691,
+      "mean_token_accuracy": 0.9781079888343811,
+      "num_tokens": 81428246.0,
+      "step": 10030
+    },
+    {
+      "entropy": 0.1741603434085846,
+      "epoch": 5.663092550790068,
+      "grad_norm": 1.9242067337036133,
+      "learning_rate": 4.305922726934788e-06,
+      "loss": 0.0657,
+      "mean_token_accuracy": 0.9787984848022461,
+      "num_tokens": 81468991.0,
+      "step": 10035
+    },
+    {
+      "entropy": 0.18352454900741577,
+      "epoch": 5.665914221218961,
+      "grad_norm": 2.054337501525879,
+      "learning_rate": 4.305277103508152e-06,
+      "loss": 0.0738,
+      "mean_token_accuracy": 0.976045835018158,
+      "num_tokens": 81509669.0,
+      "step": 10040
+    },
+    {
+      "entropy": 0.19397950172424316,
+      "epoch": 5.668735891647856,
+      "grad_norm": 2.4687507152557373,
+      "learning_rate": 4.304631248215925e-06,
+      "loss": 0.0815,
+      "mean_token_accuracy": 0.9752507686614991,
+      "num_tokens": 81550264.0,
+      "step": 10045
+    },
+    {
+      "entropy": 0.19177696108818054,
+      "epoch": 5.6715575620767495,
+      "grad_norm": 2.18593168258667,
+      "learning_rate": 4.303985161184986e-06,
+      "loss": 0.081,
+      "mean_token_accuracy": 0.9757527351379395,
+      "num_tokens": 81590845.0,
+      "step": 10050
+    },
+    {
+      "entropy": 0.19264190196990966,
+      "epoch": 5.674379232505643,
+      "grad_norm": 1.6302083730697632,
+      "learning_rate": 4.3033388425422595e-06,
+      "loss": 0.0754,
+      "mean_token_accuracy": 0.9771258950233459,
+      "num_tokens": 81631530.0,
+      "step": 10055
+    },
+    {
+      "entropy": 0.18041291534900666,
+      "epoch": 5.677200902934537,
+      "grad_norm": 2.0120229721069336,
+      "learning_rate": 4.302692292414713e-06,
+      "loss": 0.0701,
+      "mean_token_accuracy": 0.9786185026168823,
+      "num_tokens": 81672307.0,
+      "step": 10060
+    },
+    {
+      "entropy": 0.18173338174819947,
+      "epoch": 5.6800225733634315,
+      "grad_norm": 1.9405615329742432,
+      "learning_rate": 4.302045510929364e-06,
+      "loss": 0.0692,
+      "mean_token_accuracy": 0.9786619186401367,
+      "num_tokens": 81712760.0,
+      "step": 10065
+    },
+    {
+      "entropy": 0.18286008834838868,
+      "epoch": 5.682844243792325,
+      "grad_norm": 2.343987464904785,
+      "learning_rate": 4.3013984982132705e-06,
+      "loss": 0.0807,
+      "mean_token_accuracy": 0.9762983083724975,
+      "num_tokens": 81753384.0,
+      "step": 10070
+    },
+    {
+      "entropy": 0.1787573516368866,
+      "epoch": 5.685665914221219,
+      "grad_norm": 2.1654746532440186,
+      "learning_rate": 4.30075125439354e-06,
+      "loss": 0.0803,
+      "mean_token_accuracy": 0.9749480605125427,
+      "num_tokens": 81793907.0,
+      "step": 10075
+    },
+    {
+      "entropy": 0.1876184046268463,
+      "epoch": 5.688487584650113,
+      "grad_norm": 2.1528499126434326,
+      "learning_rate": 4.3001037795973225e-06,
+      "loss": 0.065,
+      "mean_token_accuracy": 0.9796617031097412,
+      "num_tokens": 81834372.0,
+      "step": 10080
+    },
+    {
+      "entropy": 0.17926348149776458,
+      "epoch": 5.691309255079007,
+      "grad_norm": 2.1884028911590576,
+      "learning_rate": 4.299456073951814e-06,
+      "loss": 0.0723,
+      "mean_token_accuracy": 0.9770533204078674,
+      "num_tokens": 81875199.0,
+      "step": 10085
+    },
+    {
+      "entropy": 0.17883450984954835,
+      "epoch": 5.694130925507901,
+      "grad_norm": 1.955631971359253,
+      "learning_rate": 4.2988081375842575e-06,
+      "loss": 0.0658,
+      "mean_token_accuracy": 0.9807510495185852,
+      "num_tokens": 81915920.0,
+      "step": 10090
+    },
+    {
+      "entropy": 0.1908187061548233,
+      "epoch": 5.696952595936795,
+      "grad_norm": 2.022860288619995,
+      "learning_rate": 4.29815997062194e-06,
+      "loss": 0.0767,
+      "mean_token_accuracy": 0.9755529642105103,
+      "num_tokens": 81956148.0,
+      "step": 10095
+    },
+    {
+      "entropy": 0.1840174227952957,
+      "epoch": 5.699774266365688,
+      "grad_norm": 2.5862724781036377,
+      "learning_rate": 4.297511573192194e-06,
+      "loss": 0.0737,
+      "mean_token_accuracy": 0.9770397782325745,
+      "num_tokens": 81996746.0,
+      "step": 10100
+    },
+    {
+      "entropy": 0.18789410591125488,
+      "epoch": 5.702595936794582,
+      "grad_norm": 2.2070839405059814,
+      "learning_rate": 4.296862945422396e-06,
+      "loss": 0.0763,
+      "mean_token_accuracy": 0.9756775379180909,
+      "num_tokens": 82037275.0,
+      "step": 10105
+    },
+    {
+      "entropy": 0.1947762042284012,
+      "epoch": 5.705417607223477,
+      "grad_norm": 2.0239546298980713,
+      "learning_rate": 4.2962140874399705e-06,
+      "loss": 0.0736,
+      "mean_token_accuracy": 0.9776275634765625,
+      "num_tokens": 82077888.0,
+      "step": 10110
+    },
+    {
+      "entropy": 0.19028544127941133,
+      "epoch": 5.70823927765237,
+      "grad_norm": 2.6127588748931885,
+      "learning_rate": 4.295564999372385e-06,
+      "loss": 0.0712,
+      "mean_token_accuracy": 0.97735435962677,
+      "num_tokens": 82118455.0,
+      "step": 10115
+    },
+    {
+      "entropy": 0.18263841271400452,
+      "epoch": 5.711060948081264,
+      "grad_norm": 1.8415166139602661,
+      "learning_rate": 4.294915681347154e-06,
+      "loss": 0.0719,
+      "mean_token_accuracy": 0.9764939785003662,
+      "num_tokens": 82158890.0,
+      "step": 10120
+    },
+    {
+      "entropy": 0.18013592660427094,
+      "epoch": 5.713882618510158,
+      "grad_norm": 2.214005470275879,
+      "learning_rate": 4.294266133491834e-06,
+      "loss": 0.0654,
+      "mean_token_accuracy": 0.9796643853187561,
+      "num_tokens": 82199368.0,
+      "step": 10125
+    },
+    {
+      "entropy": 0.17933290004730223,
+      "epoch": 5.716704288939052,
+      "grad_norm": 2.40313720703125,
+      "learning_rate": 4.293616355934032e-06,
+      "loss": 0.0761,
+      "mean_token_accuracy": 0.9756105422973633,
+      "num_tokens": 82240162.0,
+      "step": 10130
+    },
+    {
+      "entropy": 0.19547135531902313,
+      "epoch": 5.719525959367946,
+      "grad_norm": 2.1128058433532715,
+      "learning_rate": 4.292966348801394e-06,
+      "loss": 0.0715,
+      "mean_token_accuracy": 0.9782087802886963,
+      "num_tokens": 82281077.0,
+      "step": 10135
+    },
+    {
+      "entropy": 0.17906437516212464,
+      "epoch": 5.72234762979684,
+      "grad_norm": 2.4157004356384277,
+      "learning_rate": 4.292316112221615e-06,
+      "loss": 0.0637,
+      "mean_token_accuracy": 0.9807783722877502,
+      "num_tokens": 82321293.0,
+      "step": 10140
+    },
+    {
+      "entropy": 0.17458663284778594,
+      "epoch": 5.725169300225733,
+      "grad_norm": 2.1524438858032227,
+      "learning_rate": 4.291665646322434e-06,
+      "loss": 0.0691,
+      "mean_token_accuracy": 0.9782337784767151,
+      "num_tokens": 82361543.0,
+      "step": 10145
+    },
+    {
+      "entropy": 0.18585495054721832,
+      "epoch": 5.727990970654628,
+      "grad_norm": 2.3089470863342285,
+      "learning_rate": 4.2910149512316345e-06,
+      "loss": 0.0743,
+      "mean_token_accuracy": 0.976749575138092,
+      "num_tokens": 82402317.0,
+      "step": 10150
+    },
+    {
+      "entropy": 0.1775616079568863,
+      "epoch": 5.730812641083522,
+      "grad_norm": 2.647921323776245,
+      "learning_rate": 4.290364027077047e-06,
+      "loss": 0.0775,
+      "mean_token_accuracy": 0.975775396823883,
+      "num_tokens": 82443034.0,
+      "step": 10155
+    },
+    {
+      "entropy": 0.16766503751277922,
+      "epoch": 5.733634311512415,
+      "grad_norm": 2.180415391921997,
+      "learning_rate": 4.2897128739865446e-06,
+      "loss": 0.0681,
+      "mean_token_accuracy": 0.9786939501762391,
+      "num_tokens": 82483796.0,
+      "step": 10160
+    },
+    {
+      "entropy": 0.20999548733234405,
+      "epoch": 5.736455981941309,
+      "grad_norm": 2.0857527256011963,
+      "learning_rate": 4.289061492088047e-06,
+      "loss": 0.086,
+      "mean_token_accuracy": 0.9734646320343018,
+      "num_tokens": 82524597.0,
+      "step": 10165
+    },
+    {
+      "entropy": 0.18432653844356536,
+      "epoch": 5.739277652370204,
+      "grad_norm": 2.0384151935577393,
+      "learning_rate": 4.2884098815095175e-06,
+      "loss": 0.0702,
+      "mean_token_accuracy": 0.977328622341156,
+      "num_tokens": 82565390.0,
+      "step": 10170
+    },
+    {
+      "entropy": 0.17555322945117952,
+      "epoch": 5.742099322799097,
+      "grad_norm": 2.352675199508667,
+      "learning_rate": 4.287758042378966e-06,
+      "loss": 0.0657,
+      "mean_token_accuracy": 0.979662299156189,
+      "num_tokens": 82605127.0,
+      "step": 10175
+    },
+    {
+      "entropy": 0.1978653311729431,
+      "epoch": 5.744920993227991,
+      "grad_norm": 2.1638011932373047,
+      "learning_rate": 4.287105974824446e-06,
+      "loss": 0.0753,
+      "mean_token_accuracy": 0.9755738496780395,
+      "num_tokens": 82645570.0,
+      "step": 10180
+    },
+    {
+      "entropy": 0.18076943159103392,
+      "epoch": 5.747742663656885,
+      "grad_norm": 2.197725534439087,
+      "learning_rate": 4.286453678974055e-06,
+      "loss": 0.0757,
+      "mean_token_accuracy": 0.9774895668029785,
+      "num_tokens": 82686162.0,
+      "step": 10185
+    },
+    {
+      "entropy": 0.1678744912147522,
+      "epoch": 5.750564334085778,
+      "grad_norm": 1.8131523132324219,
+      "learning_rate": 4.28580115495594e-06,
+      "loss": 0.0712,
+      "mean_token_accuracy": 0.9775494694709778,
+      "num_tokens": 82726804.0,
+      "step": 10190
+    },
+    {
+      "entropy": 0.1693742036819458,
+      "epoch": 5.753386004514673,
+      "grad_norm": 1.9890544414520264,
+      "learning_rate": 4.285148402898285e-06,
+      "loss": 0.0696,
+      "mean_token_accuracy": 0.9783661007881165,
+      "num_tokens": 82767575.0,
+      "step": 10195
+    },
+    {
+      "entropy": 0.16243847608566284,
+      "epoch": 5.756207674943567,
+      "grad_norm": 2.3177292346954346,
+      "learning_rate": 4.284495422929326e-06,
+      "loss": 0.063,
+      "mean_token_accuracy": 0.9801602840423584,
+      "num_tokens": 82808426.0,
+      "step": 10200
+    },
+    {
+      "entropy": 0.18572303354740144,
+      "epoch": 5.75902934537246,
+      "grad_norm": 2.198636531829834,
+      "learning_rate": 4.283842215177341e-06,
+      "loss": 0.0701,
+      "mean_token_accuracy": 0.9777385950088501,
+      "num_tokens": 82849159.0,
+      "step": 10205
+    },
+    {
+      "entropy": 0.18345086872577668,
+      "epoch": 5.761851015801354,
+      "grad_norm": 2.125354290008545,
+      "learning_rate": 4.283188779770652e-06,
+      "loss": 0.0782,
+      "mean_token_accuracy": 0.9764749526977539,
+      "num_tokens": 82889959.0,
+      "step": 10210
+    },
+    {
+      "entropy": 0.18493236899375914,
+      "epoch": 5.764672686230249,
+      "grad_norm": 2.407668352127075,
+      "learning_rate": 4.2825351168376275e-06,
+      "loss": 0.0869,
+      "mean_token_accuracy": 0.971848976612091,
+      "num_tokens": 82930471.0,
+      "step": 10215
+    },
+    {
+      "entropy": 0.18200556337833404,
+      "epoch": 5.767494356659142,
+      "grad_norm": 1.8151724338531494,
+      "learning_rate": 4.281881226506677e-06,
+      "loss": 0.0811,
+      "mean_token_accuracy": 0.974114739894867,
+      "num_tokens": 82971060.0,
+      "step": 10220
+    },
+    {
+      "entropy": 0.17582403719425202,
+      "epoch": 5.770316027088036,
+      "grad_norm": 1.7381181716918945,
+      "learning_rate": 4.28122710890626e-06,
+      "loss": 0.0811,
+      "mean_token_accuracy": 0.974997091293335,
+      "num_tokens": 83011536.0,
+      "step": 10225
+    },
+    {
+      "entropy": 0.19064405858516692,
+      "epoch": 5.77313769751693,
+      "grad_norm": 2.46132493019104,
+      "learning_rate": 4.2805727641648775e-06,
+      "loss": 0.0911,
+      "mean_token_accuracy": 0.9718187093734741,
+      "num_tokens": 83052238.0,
+      "step": 10230
+    },
+    {
+      "entropy": 0.18355910778045653,
+      "epoch": 5.775959367945823,
+      "grad_norm": 2.21754789352417,
+      "learning_rate": 4.2799181924110755e-06,
+      "loss": 0.075,
+      "mean_token_accuracy": 0.9755819082260132,
+      "num_tokens": 83092882.0,
+      "step": 10235
+    },
+    {
+      "entropy": 0.18562229871749877,
+      "epoch": 5.778781038374718,
+      "grad_norm": 2.161637306213379,
+      "learning_rate": 4.279263393773444e-06,
+      "loss": 0.0767,
+      "mean_token_accuracy": 0.9759423017501831,
+      "num_tokens": 83133727.0,
+      "step": 10240
+    },
+    {
+      "entropy": 0.1935652256011963,
+      "epoch": 5.781602708803612,
+      "grad_norm": 1.7781506776809692,
+      "learning_rate": 4.278608368380618e-06,
+      "loss": 0.0818,
+      "mean_token_accuracy": 0.9746825933456421,
+      "num_tokens": 83173500.0,
+      "step": 10245
+    },
+    {
+      "entropy": 0.19277122020721435,
+      "epoch": 5.784424379232505,
+      "grad_norm": 1.752469539642334,
+      "learning_rate": 4.27795311636128e-06,
+      "loss": 0.0694,
+      "mean_token_accuracy": 0.9785934090614319,
+      "num_tokens": 83214016.0,
+      "step": 10250
+    },
+    {
+      "entropy": 0.179434072971344,
+      "epoch": 5.7872460496614,
+      "grad_norm": 2.006405830383301,
+      "learning_rate": 4.277297637844151e-06,
+      "loss": 0.071,
+      "mean_token_accuracy": 0.9783080697059632,
+      "num_tokens": 83254826.0,
+      "step": 10255
+    },
+    {
+      "entropy": 0.17800212800502777,
+      "epoch": 5.790067720090294,
+      "grad_norm": 2.5060832500457764,
+      "learning_rate": 4.276641932958002e-06,
+      "loss": 0.074,
+      "mean_token_accuracy": 0.9775895595550537,
+      "num_tokens": 83295437.0,
+      "step": 10260
+    },
+    {
+      "entropy": 0.1740177422761917,
+      "epoch": 5.792889390519187,
+      "grad_norm": 1.8866690397262573,
+      "learning_rate": 4.275986001831645e-06,
+      "loss": 0.0737,
+      "mean_token_accuracy": 0.9761949896812439,
+      "num_tokens": 83336071.0,
+      "step": 10265
+    },
+    {
+      "entropy": 0.17147966027259826,
+      "epoch": 5.795711060948081,
+      "grad_norm": 2.06299090385437,
+      "learning_rate": 4.275329844593938e-06,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9773301362991333,
+      "num_tokens": 83376866.0,
+      "step": 10270
+    },
+    {
+      "entropy": 0.18029634952545165,
+      "epoch": 5.798532731376975,
+      "grad_norm": 2.3944549560546875,
+      "learning_rate": 4.274673461373784e-06,
+      "loss": 0.0886,
+      "mean_token_accuracy": 0.9716618418693542,
+      "num_tokens": 83417485.0,
+      "step": 10275
+    },
+    {
+      "entropy": 0.18746162354946136,
+      "epoch": 5.801354401805869,
+      "grad_norm": 2.2615456581115723,
+      "learning_rate": 4.274016852300129e-06,
+      "loss": 0.0709,
+      "mean_token_accuracy": 0.9775063872337342,
+      "num_tokens": 83458100.0,
+      "step": 10280
+    },
+    {
+      "entropy": 0.18002947568893432,
+      "epoch": 5.804176072234763,
+      "grad_norm": 1.8940982818603516,
+      "learning_rate": 4.273360017501964e-06,
+      "loss": 0.0732,
+      "mean_token_accuracy": 0.9775622367858887,
+      "num_tokens": 83498674.0,
+      "step": 10285
+    },
+    {
+      "entropy": 0.1664568930864334,
+      "epoch": 5.806997742663657,
+      "grad_norm": 2.5354745388031006,
+      "learning_rate": 4.272702957108325e-06,
+      "loss": 0.0729,
+      "mean_token_accuracy": 0.976758623123169,
+      "num_tokens": 83539467.0,
+      "step": 10290
+    },
+    {
+      "entropy": 0.18044437170028688,
+      "epoch": 5.8098194130925505,
+      "grad_norm": 1.8379076719284058,
+      "learning_rate": 4.272045671248289e-06,
+      "loss": 0.0661,
+      "mean_token_accuracy": 0.9795686602592468,
+      "num_tokens": 83580065.0,
+      "step": 10295
+    },
+    {
+      "entropy": 0.18124404847621917,
+      "epoch": 5.812641083521445,
+      "grad_norm": 2.4792568683624268,
+      "learning_rate": 4.2713881600509835e-06,
+      "loss": 0.0632,
+      "mean_token_accuracy": 0.9804369568824768,
+      "num_tokens": 83619608.0,
+      "step": 10300
+    },
+    {
+      "entropy": 0.1773180991411209,
+      "epoch": 5.815462753950339,
+      "grad_norm": 2.801755428314209,
+      "learning_rate": 4.270730423645574e-06,
+      "loss": 0.066,
+      "mean_token_accuracy": 0.9787346959114075,
+      "num_tokens": 83659117.0,
+      "step": 10305
+    },
+    {
+      "entropy": 0.17246809601783752,
+      "epoch": 5.818284424379232,
+      "grad_norm": 1.8683533668518066,
+      "learning_rate": 4.2700724621612745e-06,
+      "loss": 0.0701,
+      "mean_token_accuracy": 0.9784170150756836,
+      "num_tokens": 83699987.0,
+      "step": 10310
+    },
+    {
+      "entropy": 0.17956862449645997,
+      "epoch": 5.821106094808126,
+      "grad_norm": 1.9856154918670654,
+      "learning_rate": 4.269414275727341e-06,
+      "loss": 0.083,
+      "mean_token_accuracy": 0.9736566543579102,
+      "num_tokens": 83740362.0,
+      "step": 10315
+    },
+    {
+      "entropy": 0.17001102566719056,
+      "epoch": 5.82392776523702,
+      "grad_norm": 2.348156452178955,
+      "learning_rate": 4.2687558644730735e-06,
+      "loss": 0.0603,
+      "mean_token_accuracy": 0.9807537913322448,
+      "num_tokens": 83781025.0,
+      "step": 10320
+    },
+    {
+      "entropy": 0.17963250279426574,
+      "epoch": 5.826749435665914,
+      "grad_norm": 2.1682050228118896,
+      "learning_rate": 4.268097228527818e-06,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9780291318893433,
+      "num_tokens": 83821718.0,
+      "step": 10325
+    },
+    {
+      "entropy": 0.18213966786861419,
+      "epoch": 5.829571106094808,
+      "grad_norm": 2.4645328521728516,
+      "learning_rate": 4.267438368020964e-06,
+      "loss": 0.0777,
+      "mean_token_accuracy": 0.9760030627250671,
+      "num_tokens": 83862498.0,
+      "step": 10330
+    },
+    {
+      "entropy": 0.16510314047336577,
+      "epoch": 5.832392776523702,
+      "grad_norm": 1.8456097841262817,
+      "learning_rate": 4.2667792830819435e-06,
+      "loss": 0.0668,
+      "mean_token_accuracy": 0.9788869500160218,
+      "num_tokens": 83903238.0,
+      "step": 10335
+    },
+    {
+      "entropy": 0.17967475354671478,
+      "epoch": 5.835214446952596,
+      "grad_norm": 1.8277400732040405,
+      "learning_rate": 4.266119973840235e-06,
+      "loss": 0.0766,
+      "mean_token_accuracy": 0.975653862953186,
+      "num_tokens": 83944020.0,
+      "step": 10340
+    },
+    {
+      "entropy": 0.18739196360111238,
+      "epoch": 5.83803611738149,
+      "grad_norm": 2.124237060546875,
+      "learning_rate": 4.2654604404253585e-06,
+      "loss": 0.0773,
+      "mean_token_accuracy": 0.9757479429244995,
+      "num_tokens": 83984438.0,
+      "step": 10345
+    },
+    {
+      "entropy": 0.1835099220275879,
+      "epoch": 5.840857787810384,
+      "grad_norm": 2.48547625541687,
+      "learning_rate": 4.264800682966881e-06,
+      "loss": 0.0658,
+      "mean_token_accuracy": 0.9790345907211304,
+      "num_tokens": 84025304.0,
+      "step": 10350
+    },
+    {
+      "entropy": 0.17234547436237335,
+      "epoch": 5.8436794582392775,
+      "grad_norm": 2.024911642074585,
+      "learning_rate": 4.26414070159441e-06,
+      "loss": 0.0661,
+      "mean_token_accuracy": 0.9787455081939698,
+      "num_tokens": 84065848.0,
+      "step": 10355
+    },
+    {
+      "entropy": 0.19444843232631684,
+      "epoch": 5.846501128668171,
+      "grad_norm": 2.464071273803711,
+      "learning_rate": 4.263480496437601e-06,
+      "loss": 0.0811,
+      "mean_token_accuracy": 0.974936056137085,
+      "num_tokens": 84106680.0,
+      "step": 10360
+    },
+    {
+      "entropy": 0.173148912191391,
+      "epoch": 5.849322799097066,
+      "grad_norm": 1.9251296520233154,
+      "learning_rate": 4.26282006762615e-06,
+      "loss": 0.0701,
+      "mean_token_accuracy": 0.9775914549827576,
+      "num_tokens": 84147405.0,
+      "step": 10365
+    },
+    {
+      "entropy": 0.18452159464359283,
+      "epoch": 5.8521444695259595,
+      "grad_norm": 2.2456061840057373,
+      "learning_rate": 4.262159415289799e-06,
+      "loss": 0.0768,
+      "mean_token_accuracy": 0.9760621428489685,
+      "num_tokens": 84188045.0,
+      "step": 10370
+    },
+    {
+      "entropy": 0.18307125866413115,
+      "epoch": 5.854966139954853,
+      "grad_norm": 2.077425241470337,
+      "learning_rate": 4.261498539558333e-06,
+      "loss": 0.0692,
+      "mean_token_accuracy": 0.9785082459449768,
+      "num_tokens": 84227724.0,
+      "step": 10375
+    },
+    {
+      "entropy": 0.1773241400718689,
+      "epoch": 5.857787810383747,
+      "grad_norm": 2.191255569458008,
+      "learning_rate": 4.260837440561583e-06,
+      "loss": 0.0711,
+      "mean_token_accuracy": 0.976841950416565,
+      "num_tokens": 84268109.0,
+      "step": 10380
+    },
+    {
+      "entropy": 0.18978277146816253,
+      "epoch": 5.860609480812641,
+      "grad_norm": 2.40100359916687,
+      "learning_rate": 4.260176118429418e-06,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9782564878463745,
+      "num_tokens": 84308793.0,
+      "step": 10385
+    },
+    {
+      "entropy": 0.18334400355815889,
+      "epoch": 5.863431151241535,
+      "grad_norm": 2.2183425426483154,
+      "learning_rate": 4.259514573291757e-06,
+      "loss": 0.0792,
+      "mean_token_accuracy": 0.9754730105400086,
+      "num_tokens": 84349159.0,
+      "step": 10390
+    },
+    {
+      "entropy": 0.2000194698572159,
+      "epoch": 5.866252821670429,
+      "grad_norm": 2.203062057495117,
+      "learning_rate": 4.258852805278562e-06,
+      "loss": 0.0838,
+      "mean_token_accuracy": 0.9722742438316345,
+      "num_tokens": 84389870.0,
+      "step": 10395
+    },
+    {
+      "entropy": 0.18452554047107697,
+      "epoch": 5.8690744920993225,
+      "grad_norm": 2.1212892532348633,
+      "learning_rate": 4.258190814519834e-06,
+      "loss": 0.0774,
+      "mean_token_accuracy": 0.976058554649353,
+      "num_tokens": 84430501.0,
+      "step": 10400
+    },
+    {
+      "entropy": 0.1830653041601181,
+      "epoch": 5.871896162528216,
+      "grad_norm": 2.3475375175476074,
+      "learning_rate": 4.2575286011456255e-06,
+      "loss": 0.074,
+      "mean_token_accuracy": 0.9763685941696167,
+      "num_tokens": 84471098.0,
+      "step": 10405
+    },
+    {
+      "entropy": 0.16962153613567352,
+      "epoch": 5.874717832957111,
+      "grad_norm": 1.6830201148986816,
+      "learning_rate": 4.256866165286024e-06,
+      "loss": 0.069,
+      "mean_token_accuracy": 0.9781726360321045,
+      "num_tokens": 84511664.0,
+      "step": 10410
+    },
+    {
+      "entropy": 0.1886516332626343,
+      "epoch": 5.8775395033860045,
+      "grad_norm": 2.5439491271972656,
+      "learning_rate": 4.256203507071168e-06,
+      "loss": 0.0834,
+      "mean_token_accuracy": 0.974034309387207,
+      "num_tokens": 84552223.0,
+      "step": 10415
+    },
+    {
+      "entropy": 0.17994187772274017,
+      "epoch": 5.880361173814898,
+      "grad_norm": 2.874234437942505,
+      "learning_rate": 4.255540626631236e-06,
+      "loss": 0.0763,
+      "mean_token_accuracy": 0.9750635504722596,
+      "num_tokens": 84592774.0,
+      "step": 10420
+    },
+    {
+      "entropy": 0.19643645584583283,
+      "epoch": 5.883182844243793,
+      "grad_norm": 2.5835118293762207,
+      "learning_rate": 4.2548775240964515e-06,
+      "loss": 0.0883,
+      "mean_token_accuracy": 0.9736735224723816,
+      "num_tokens": 84633572.0,
+      "step": 10425
+    },
+    {
+      "entropy": 0.17592164278030395,
+      "epoch": 5.8860045146726865,
+      "grad_norm": 2.1674015522003174,
+      "learning_rate": 4.25421419959708e-06,
+      "loss": 0.074,
+      "mean_token_accuracy": 0.9766902089118957,
+      "num_tokens": 84674314.0,
+      "step": 10430
+    },
+    {
+      "entropy": 0.19005840122699738,
+      "epoch": 5.88882618510158,
+      "grad_norm": 1.8254783153533936,
+      "learning_rate": 4.253550653263432e-06,
+      "loss": 0.0783,
+      "mean_token_accuracy": 0.9754180431365966,
+      "num_tokens": 84714985.0,
+      "step": 10435
+    },
+    {
+      "entropy": 0.17556642293930053,
+      "epoch": 5.891647855530474,
+      "grad_norm": 1.7899775505065918,
+      "learning_rate": 4.2528868852258615e-06,
+      "loss": 0.0664,
+      "mean_token_accuracy": 0.9791262030601502,
+      "num_tokens": 84755700.0,
+      "step": 10440
+    },
+    {
+      "entropy": 0.17945761382579803,
+      "epoch": 5.894469525959368,
+      "grad_norm": 2.6520256996154785,
+      "learning_rate": 4.252222895614766e-06,
+      "loss": 0.0713,
+      "mean_token_accuracy": 0.9770576000213623,
+      "num_tokens": 84796363.0,
+      "step": 10445
+    },
+    {
+      "entropy": 0.19162935614585877,
+      "epoch": 5.897291196388262,
+      "grad_norm": 2.3933403491973877,
+      "learning_rate": 4.2515586845605864e-06,
+      "loss": 0.0795,
+      "mean_token_accuracy": 0.9750626444816589,
+      "num_tokens": 84836849.0,
+      "step": 10450
+    },
+    {
+      "entropy": 0.17797043323516845,
+      "epoch": 5.900112866817156,
+      "grad_norm": 2.0211195945739746,
+      "learning_rate": 4.250894252193806e-06,
+      "loss": 0.0713,
+      "mean_token_accuracy": 0.9775516271591187,
+      "num_tokens": 84877674.0,
+      "step": 10455
+    },
+    {
+      "entropy": 0.1765431433916092,
+      "epoch": 5.9029345372460496,
+      "grad_norm": 2.351975440979004,
+      "learning_rate": 4.250229598644954e-06,
+      "loss": 0.0678,
+      "mean_token_accuracy": 0.9786093235015869,
+      "num_tokens": 84918343.0,
+      "step": 10460
+    },
+    {
+      "entropy": 0.18122887313365937,
+      "epoch": 5.905756207674943,
+      "grad_norm": 1.8683475255966187,
+      "learning_rate": 4.249564724044602e-06,
+      "loss": 0.072,
+      "mean_token_accuracy": 0.9761911749839782,
+      "num_tokens": 84958846.0,
+      "step": 10465
+    },
+    {
+      "entropy": 0.17490617036819459,
+      "epoch": 5.908577878103838,
+      "grad_norm": 1.9673949480056763,
+      "learning_rate": 4.248899628523362e-06,
+      "loss": 0.0763,
+      "mean_token_accuracy": 0.9763616800308228,
+      "num_tokens": 84999643.0,
+      "step": 10470
+    },
+    {
+      "entropy": 0.18483209908008574,
+      "epoch": 5.9113995485327315,
+      "grad_norm": 1.887024164199829,
+      "learning_rate": 4.248234312211895e-06,
+      "loss": 0.0688,
+      "mean_token_accuracy": 0.978734838962555,
+      "num_tokens": 85040465.0,
+      "step": 10475
+    },
+    {
+      "entropy": 0.19448497295379638,
+      "epoch": 5.914221218961625,
+      "grad_norm": 2.0865771770477295,
+      "learning_rate": 4.247568775240901e-06,
+      "loss": 0.0623,
+      "mean_token_accuracy": 0.9806246161460876,
+      "num_tokens": 85081222.0,
+      "step": 10480
+    },
+    {
+      "entropy": 0.1752565920352936,
+      "epoch": 5.917042889390519,
+      "grad_norm": 2.1277925968170166,
+      "learning_rate": 4.246903017741124e-06,
+      "loss": 0.0841,
+      "mean_token_accuracy": 0.9734927892684937,
+      "num_tokens": 85121900.0,
+      "step": 10485
+    },
+    {
+      "entropy": 0.18596406877040864,
+      "epoch": 5.919864559819413,
+      "grad_norm": 2.1438140869140625,
+      "learning_rate": 4.246237039843355e-06,
+      "loss": 0.0659,
+      "mean_token_accuracy": 0.9792929530143738,
+      "num_tokens": 85162395.0,
+      "step": 10490
+    },
+    {
+      "entropy": 0.1806576669216156,
+      "epoch": 5.922686230248307,
+      "grad_norm": 2.2096385955810547,
+      "learning_rate": 4.2455708416784235e-06,
+      "loss": 0.0719,
+      "mean_token_accuracy": 0.977632212638855,
+      "num_tokens": 85203275.0,
+      "step": 10495
+    },
+    {
+      "entropy": 0.20309909880161287,
+      "epoch": 5.925507900677201,
+      "grad_norm": 1.782074213027954,
+      "learning_rate": 4.244904423377204e-06,
+      "loss": 0.0736,
+      "mean_token_accuracy": 0.9776381611824035,
+      "num_tokens": 85243951.0,
+      "step": 10500
+    },
+    {
+      "epoch": 5.925507900677201,
+      "eval_entropy": 0.2187255471944809,
+      "eval_loss": 0.030472468584775925,
+      "eval_mean_token_accuracy": 0.9911843538284302,
+      "eval_num_tokens": 85243951.0,
+      "eval_runtime": 0.1635,
+      "eval_samples_per_second": 24.46,
+      "eval_steps_per_second": 6.115,
+      "step": 10500
+    },
+    {
+      "entropy": 0.18639110922813415,
+      "epoch": 5.928329571106095,
+      "grad_norm": 2.001336097717285,
+      "learning_rate": 4.244237785070615e-06,
+      "loss": 0.0712,
+      "mean_token_accuracy": 0.9782602667808533,
+      "num_tokens": 85284411.0,
+      "step": 10505
+    },
+    {
+      "entropy": 0.18605276346206664,
+      "epoch": 5.931151241534989,
+      "grad_norm": 2.1870064735412598,
+      "learning_rate": 4.243570926889618e-06,
+      "loss": 0.0759,
+      "mean_token_accuracy": 0.9759214639663696,
+      "num_tokens": 85325253.0,
+      "step": 10510
+    },
+    {
+      "entropy": 0.17156625986099244,
+      "epoch": 5.933972911963883,
+      "grad_norm": 2.035531520843506,
+      "learning_rate": 4.242903848965217e-06,
+      "loss": 0.0656,
+      "mean_token_accuracy": 0.9796592235565186,
+      "num_tokens": 85365632.0,
+      "step": 10515
+    },
+    {
+      "entropy": 0.20702783465385438,
+      "epoch": 5.936794582392777,
+      "grad_norm": 2.6586201190948486,
+      "learning_rate": 4.242236551428459e-06,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9800093650817872,
+      "num_tokens": 85405847.0,
+      "step": 10520
+    },
+    {
+      "entropy": 0.18220070600509644,
+      "epoch": 5.93961625282167,
+      "grad_norm": 2.27356219291687,
+      "learning_rate": 4.241569034410436e-06,
+      "loss": 0.0749,
+      "mean_token_accuracy": 0.9759989261627198,
+      "num_tokens": 85446548.0,
+      "step": 10525
+    },
+    {
+      "entropy": 0.17777936458587645,
+      "epoch": 5.942437923250564,
+      "grad_norm": 1.8858665227890015,
+      "learning_rate": 4.24090129804228e-06,
+      "loss": 0.0764,
+      "mean_token_accuracy": 0.9763510704040528,
+      "num_tokens": 85487166.0,
+      "step": 10530
+    },
+    {
+      "entropy": 0.19082893133163453,
+      "epoch": 5.9452595936794586,
+      "grad_norm": 2.3828372955322266,
+      "learning_rate": 4.24023334245517e-06,
+      "loss": 0.0726,
+      "mean_token_accuracy": 0.9769440650939941,
+      "num_tokens": 85527909.0,
+      "step": 10535
+    },
+    {
+      "entropy": 0.16436271667480468,
+      "epoch": 5.948081264108352,
+      "grad_norm": 2.2857789993286133,
+      "learning_rate": 4.2395651677803244e-06,
+      "loss": 0.0691,
+      "mean_token_accuracy": 0.9776111841201782,
+      "num_tokens": 85568713.0,
+      "step": 10540
+    },
+    {
+      "entropy": 0.19756191074848176,
+      "epoch": 5.950902934537246,
+      "grad_norm": 1.9905869960784912,
+      "learning_rate": 4.238896774149007e-06,
+      "loss": 0.0671,
+      "mean_token_accuracy": 0.9784975528717041,
+      "num_tokens": 85609278.0,
+      "step": 10545
+    },
+    {
+      "entropy": 0.17361958026885987,
+      "epoch": 5.95372460496614,
+      "grad_norm": 1.9324471950531006,
+      "learning_rate": 4.2382281616925235e-06,
+      "loss": 0.0715,
+      "mean_token_accuracy": 0.9772069692611695,
+      "num_tokens": 85649844.0,
+      "step": 10550
+    },
+    {
+      "entropy": 0.17714324295520784,
+      "epoch": 5.956546275395034,
+      "grad_norm": 2.4276859760284424,
+      "learning_rate": 4.237559330542223e-06,
+      "loss": 0.0732,
+      "mean_token_accuracy": 0.9776072025299072,
+      "num_tokens": 85690372.0,
+      "step": 10555
+    },
+    {
+      "entropy": 0.18243989944458008,
+      "epoch": 5.959367945823928,
+      "grad_norm": 2.2024104595184326,
+      "learning_rate": 4.236890280829496e-06,
+      "loss": 0.0758,
+      "mean_token_accuracy": 0.9765413165092468,
+      "num_tokens": 85730984.0,
+      "step": 10560
+    },
+    {
+      "entropy": 0.18148486614227294,
+      "epoch": 5.962189616252822,
+      "grad_norm": 2.123389720916748,
+      "learning_rate": 4.236221012685781e-06,
+      "loss": 0.0675,
+      "mean_token_accuracy": 0.9791842937469483,
+      "num_tokens": 85771776.0,
+      "step": 10565
+    },
+    {
+      "entropy": 0.19152417182922363,
+      "epoch": 5.965011286681715,
+      "grad_norm": 2.2187232971191406,
+      "learning_rate": 4.235551526242552e-06,
+      "loss": 0.0727,
+      "mean_token_accuracy": 0.9772362232208252,
+      "num_tokens": 85812366.0,
+      "step": 10570
+    },
+    {
+      "entropy": 0.18775533139705658,
+      "epoch": 5.967832957110609,
+      "grad_norm": 2.338989496231079,
+      "learning_rate": 4.234881821631332e-06,
+      "loss": 0.0823,
+      "mean_token_accuracy": 0.974126398563385,
+      "num_tokens": 85851986.0,
+      "step": 10575
+    },
+    {
+      "entropy": 0.19790259897708892,
+      "epoch": 5.970654627539504,
+      "grad_norm": 1.8901110887527466,
+      "learning_rate": 4.234211898983684e-06,
+      "loss": 0.0742,
+      "mean_token_accuracy": 0.976418399810791,
+      "num_tokens": 85892394.0,
+      "step": 10580
+    },
+    {
+      "entropy": 0.18636341392993927,
+      "epoch": 5.973476297968397,
+      "grad_norm": 2.382916212081909,
+      "learning_rate": 4.233541758431213e-06,
+      "loss": 0.083,
+      "mean_token_accuracy": 0.9738220572471619,
+      "num_tokens": 85932953.0,
+      "step": 10585
+    },
+    {
+      "entropy": 0.16605794429779053,
+      "epoch": 5.976297968397291,
+      "grad_norm": 1.9576960802078247,
+      "learning_rate": 4.232871400105572e-06,
+      "loss": 0.0698,
+      "mean_token_accuracy": 0.9788704752922058,
+      "num_tokens": 85973329.0,
+      "step": 10590
+    },
+    {
+      "entropy": 0.16949787735939026,
+      "epoch": 5.979119638826186,
+      "grad_norm": 1.836525559425354,
+      "learning_rate": 4.232200824138448e-06,
+      "loss": 0.0688,
+      "mean_token_accuracy": 0.9781976103782654,
+      "num_tokens": 86013764.0,
+      "step": 10595
+    },
+    {
+      "entropy": 0.17068464457988738,
+      "epoch": 5.981941309255079,
+      "grad_norm": 1.9783167839050293,
+      "learning_rate": 4.231530030661579e-06,
+      "loss": 0.0685,
+      "mean_token_accuracy": 0.9786436915397644,
+      "num_tokens": 86054432.0,
+      "step": 10600
+    },
+    {
+      "entropy": 0.18707202672958373,
+      "epoch": 5.984762979683973,
+      "grad_norm": 1.955723524093628,
+      "learning_rate": 4.230859019806741e-06,
+      "loss": 0.06,
+      "mean_token_accuracy": 0.9814386010169983,
+      "num_tokens": 86095074.0,
+      "step": 10605
+    },
+    {
+      "entropy": 0.19427139759063722,
+      "epoch": 5.987584650112867,
+      "grad_norm": 2.150771141052246,
+      "learning_rate": 4.230187791705756e-06,
+      "loss": 0.0887,
+      "mean_token_accuracy": 0.9718084454536438,
+      "num_tokens": 86135924.0,
+      "step": 10610
+    },
+    {
+      "entropy": 0.18580053150653839,
+      "epoch": 5.99040632054176,
+      "grad_norm": 1.954380989074707,
+      "learning_rate": 4.229516346490485e-06,
+      "loss": 0.0782,
+      "mean_token_accuracy": 0.9761150360107422,
+      "num_tokens": 86176503.0,
+      "step": 10615
+    },
+    {
+      "entropy": 0.18900538682937623,
+      "epoch": 5.993227990970655,
+      "grad_norm": 2.2284088134765625,
+      "learning_rate": 4.2288446842928345e-06,
+      "loss": 0.0784,
+      "mean_token_accuracy": 0.9764190793037415,
+      "num_tokens": 86217113.0,
+      "step": 10620
+    },
+    {
+      "entropy": 0.18157367706298827,
+      "epoch": 5.996049661399549,
+      "grad_norm": 2.151667833328247,
+      "learning_rate": 4.228172805244753e-06,
+      "loss": 0.0793,
+      "mean_token_accuracy": 0.9748581528663636,
+      "num_tokens": 86257786.0,
+      "step": 10625
+    },
+    {
+      "entropy": 0.19717211425304412,
+      "epoch": 5.998871331828442,
+      "grad_norm": 2.68282413482666,
+      "learning_rate": 4.227500709478229e-06,
+      "loss": 0.0845,
+      "mean_token_accuracy": 0.9748492002487182,
+      "num_tokens": 86298375.0,
+      "step": 10630
+    },
+    {
+      "entropy": 0.16193274557590484,
+      "epoch": 6.001693002257336,
+      "grad_norm": 1.6442623138427734,
+      "learning_rate": 4.226828397125298e-06,
+      "loss": 0.0563,
+      "mean_token_accuracy": 0.9836606025695801,
+      "num_tokens": 86333142.0,
+      "step": 10635
+    },
+    {
+      "entropy": 0.15360802710056304,
+      "epoch": 6.004514672686231,
+      "grad_norm": 1.2552261352539062,
+      "learning_rate": 4.226155868318035e-06,
+      "loss": 0.0358,
+      "mean_token_accuracy": 0.9905863881111145,
+      "num_tokens": 86373661.0,
+      "step": 10640
+    },
+    {
+      "entropy": 0.15118006765842437,
+      "epoch": 6.007336343115124,
+      "grad_norm": 2.012360095977783,
+      "learning_rate": 4.225483123188559e-06,
+      "loss": 0.0349,
+      "mean_token_accuracy": 0.990486478805542,
+      "num_tokens": 86414362.0,
+      "step": 10645
+    },
+    {
+      "entropy": 0.13885512351989746,
+      "epoch": 6.010158013544018,
+      "grad_norm": 2.042280435562134,
+      "learning_rate": 4.224810161869029e-06,
+      "loss": 0.0329,
+      "mean_token_accuracy": 0.990878415107727,
+      "num_tokens": 86454954.0,
+      "step": 10650
+    },
+    {
+      "entropy": 0.1436757594347,
+      "epoch": 6.012979683972912,
+      "grad_norm": 1.7394871711730957,
+      "learning_rate": 4.224136984491651e-06,
+      "loss": 0.0358,
+      "mean_token_accuracy": 0.9893459677696228,
+      "num_tokens": 86495647.0,
+      "step": 10655
+    },
+    {
+      "entropy": 0.1538625791668892,
+      "epoch": 6.015801354401806,
+      "grad_norm": 1.6942013502120972,
+      "learning_rate": 4.22346359118867e-06,
+      "loss": 0.0335,
+      "mean_token_accuracy": 0.9908841848373413,
+      "num_tokens": 86536010.0,
+      "step": 10660
+    },
+    {
+      "entropy": 0.1632068932056427,
+      "epoch": 6.0186230248307,
+      "grad_norm": 2.3148810863494873,
+      "learning_rate": 4.222789982092373e-06,
+      "loss": 0.0313,
+      "mean_token_accuracy": 0.9914784550666809,
+      "num_tokens": 86576406.0,
+      "step": 10665
+    },
+    {
+      "entropy": 0.1600690007209778,
+      "epoch": 6.021444695259594,
+      "grad_norm": 2.032419443130493,
+      "learning_rate": 4.222116157335091e-06,
+      "loss": 0.0394,
+      "mean_token_accuracy": 0.9886528730392456,
+      "num_tokens": 86616837.0,
+      "step": 10670
+    },
+    {
+      "entropy": 0.13843523859977722,
+      "epoch": 6.024266365688487,
+      "grad_norm": 2.1894493103027344,
+      "learning_rate": 4.2214421170491975e-06,
+      "loss": 0.0281,
+      "mean_token_accuracy": 0.9916700005531311,
+      "num_tokens": 86657356.0,
+      "step": 10675
+    },
+    {
+      "entropy": 0.1451241284608841,
+      "epoch": 6.027088036117381,
+      "grad_norm": 1.5372796058654785,
+      "learning_rate": 4.220767861367108e-06,
+      "loss": 0.032,
+      "mean_token_accuracy": 0.9909778356552124,
+      "num_tokens": 86697851.0,
+      "step": 10680
+    },
+    {
+      "entropy": 0.1586732506752014,
+      "epoch": 6.029909706546276,
+      "grad_norm": 1.9109724760055542,
+      "learning_rate": 4.220093390421279e-06,
+      "loss": 0.0396,
+      "mean_token_accuracy": 0.9885246515274048,
+      "num_tokens": 86738396.0,
+      "step": 10685
+    },
+    {
+      "entropy": 0.14336285442113877,
+      "epoch": 6.032731376975169,
+      "grad_norm": 1.615535855293274,
+      "learning_rate": 4.219418704344211e-06,
+      "loss": 0.0323,
+      "mean_token_accuracy": 0.9910237193107605,
+      "num_tokens": 86779215.0,
+      "step": 10690
+    },
+    {
+      "entropy": 0.15025009214878082,
+      "epoch": 6.035553047404063,
+      "grad_norm": 2.157390594482422,
+      "learning_rate": 4.218743803268447e-06,
+      "loss": 0.0399,
+      "mean_token_accuracy": 0.9880735993385314,
+      "num_tokens": 86819698.0,
+      "step": 10695
+    },
+    {
+      "entropy": 0.15084101557731627,
+      "epoch": 6.038374717832957,
+      "grad_norm": 1.7998473644256592,
+      "learning_rate": 4.218068687326571e-06,
+      "loss": 0.0354,
+      "mean_token_accuracy": 0.9899451732635498,
+      "num_tokens": 86860397.0,
+      "step": 10700
+    },
+    {
+      "entropy": 0.14166499078273773,
+      "epoch": 6.041196388261851,
+      "grad_norm": 1.578946590423584,
+      "learning_rate": 4.217393356651208e-06,
+      "loss": 0.0367,
+      "mean_token_accuracy": 0.9898417115211486,
+      "num_tokens": 86901077.0,
+      "step": 10705
+    },
+    {
+      "entropy": 0.14319468885660172,
+      "epoch": 6.044018058690745,
+      "grad_norm": 1.6669219732284546,
+      "learning_rate": 4.216717811375028e-06,
+      "loss": 0.0345,
+      "mean_token_accuracy": 0.9898629069328309,
+      "num_tokens": 86941792.0,
+      "step": 10710
+    },
+    {
+      "entropy": 0.13536919057369232,
+      "epoch": 6.046839729119639,
+      "grad_norm": 1.7521120309829712,
+      "learning_rate": 4.216042051630743e-06,
+      "loss": 0.0326,
+      "mean_token_accuracy": 0.9901854276657105,
+      "num_tokens": 86982279.0,
+      "step": 10715
+    },
+    {
+      "entropy": 0.1333491548895836,
+      "epoch": 6.049661399548532,
+      "grad_norm": 1.9163808822631836,
+      "learning_rate": 4.215366077551105e-06,
+      "loss": 0.0341,
+      "mean_token_accuracy": 0.9900972962379455,
+      "num_tokens": 87023083.0,
+      "step": 10720
+    },
+    {
+      "entropy": 0.15708822011947632,
+      "epoch": 6.052483069977427,
+      "grad_norm": 2.149850368499756,
+      "learning_rate": 4.21468988926891e-06,
+      "loss": 0.0382,
+      "mean_token_accuracy": 0.9889817953109741,
+      "num_tokens": 87063613.0,
+      "step": 10725
+    },
+    {
+      "entropy": 0.14288501143455506,
+      "epoch": 6.055304740406321,
+      "grad_norm": 1.6220556497573853,
+      "learning_rate": 4.2140134869169934e-06,
+      "loss": 0.0397,
+      "mean_token_accuracy": 0.9880726933479309,
+      "num_tokens": 87104292.0,
+      "step": 10730
+    },
+    {
+      "entropy": 0.13960061967372894,
+      "epoch": 6.058126410835214,
+      "grad_norm": 1.8582446575164795,
+      "learning_rate": 4.213336870628236e-06,
+      "loss": 0.0337,
+      "mean_token_accuracy": 0.9908414721488953,
+      "num_tokens": 87144991.0,
+      "step": 10735
+    },
+    {
+      "entropy": 0.1556618928909302,
+      "epoch": 6.060948081264108,
+      "grad_norm": 1.7569464445114136,
+      "learning_rate": 4.21266004053556e-06,
+      "loss": 0.035,
+      "mean_token_accuracy": 0.9900591254234314,
+      "num_tokens": 87185620.0,
+      "step": 10740
+    },
+    {
+      "entropy": 0.15336497128009796,
+      "epoch": 6.063769751693002,
+      "grad_norm": 1.2546908855438232,
+      "learning_rate": 4.211982996771926e-06,
+      "loss": 0.0348,
+      "mean_token_accuracy": 0.9904587507247925,
+      "num_tokens": 87226351.0,
+      "step": 10745
+    },
+    {
+      "entropy": 0.1377393662929535,
+      "epoch": 6.066591422121896,
+      "grad_norm": 1.875976800918579,
+      "learning_rate": 4.211305739470342e-06,
+      "loss": 0.0305,
+      "mean_token_accuracy": 0.9914319157600403,
+      "num_tokens": 87267095.0,
+      "step": 10750
+    },
+    {
+      "entropy": 0.1543271839618683,
+      "epoch": 6.06941309255079,
+      "grad_norm": 1.8693065643310547,
+      "learning_rate": 4.210628268763854e-06,
+      "loss": 0.0337,
+      "mean_token_accuracy": 0.9902760028839112,
+      "num_tokens": 87307784.0,
+      "step": 10755
+    },
+    {
+      "entropy": 0.14306709170341492,
+      "epoch": 6.072234762979684,
+      "grad_norm": 2.66241717338562,
+      "learning_rate": 4.209950584785552e-06,
+      "loss": 0.0307,
+      "mean_token_accuracy": 0.9919659614562988,
+      "num_tokens": 87348340.0,
+      "step": 10760
+    },
+    {
+      "entropy": 0.14404877722263337,
+      "epoch": 6.0750564334085775,
+      "grad_norm": 1.838407278060913,
+      "learning_rate": 4.209272687668565e-06,
+      "loss": 0.0314,
+      "mean_token_accuracy": 0.9915964126586914,
+      "num_tokens": 87389066.0,
+      "step": 10765
+    },
+    {
+      "entropy": 0.15369834899902343,
+      "epoch": 6.077878103837472,
+      "grad_norm": 1.1812386512756348,
+      "learning_rate": 4.2085945775460685e-06,
+      "loss": 0.0294,
+      "mean_token_accuracy": 0.9916245102882385,
+      "num_tokens": 87429380.0,
+      "step": 10770
+    },
+    {
+      "entropy": 0.15424088537693023,
+      "epoch": 6.080699774266366,
+      "grad_norm": 1.5837703943252563,
+      "learning_rate": 4.207916254551276e-06,
+      "loss": 0.0328,
+      "mean_token_accuracy": 0.9913092613220215,
+      "num_tokens": 87470021.0,
+      "step": 10775
+    },
+    {
+      "entropy": 0.1431649297475815,
+      "epoch": 6.0835214446952595,
+      "grad_norm": 1.973551630973816,
+      "learning_rate": 4.207237718817446e-06,
+      "loss": 0.0312,
+      "mean_token_accuracy": 0.9908066749572754,
+      "num_tokens": 87510700.0,
+      "step": 10780
+    },
+    {
+      "entropy": 0.14944909512996674,
+      "epoch": 6.086343115124153,
+      "grad_norm": 1.8268595933914185,
+      "learning_rate": 4.2065589704778745e-06,
+      "loss": 0.0325,
+      "mean_token_accuracy": 0.9908172607421875,
+      "num_tokens": 87551408.0,
+      "step": 10785
+    },
+    {
+      "entropy": 0.1600016713142395,
+      "epoch": 6.089164785553048,
+      "grad_norm": 1.7882065773010254,
+      "learning_rate": 4.205880009665902e-06,
+      "loss": 0.0378,
+      "mean_token_accuracy": 0.9892760753631592,
+      "num_tokens": 87592139.0,
+      "step": 10790
+    },
+    {
+      "entropy": 0.14885053634643555,
+      "epoch": 6.091986455981941,
+      "grad_norm": 1.7751688957214355,
+      "learning_rate": 4.205200836514912e-06,
+      "loss": 0.0372,
+      "mean_token_accuracy": 0.9887823104858399,
+      "num_tokens": 87632768.0,
+      "step": 10795
+    },
+    {
+      "entropy": 0.14843485057353972,
+      "epoch": 6.094808126410835,
+      "grad_norm": 1.5204575061798096,
+      "learning_rate": 4.204521451158327e-06,
+      "loss": 0.0345,
+      "mean_token_accuracy": 0.9894916415214539,
+      "num_tokens": 87673525.0,
+      "step": 10800
+    },
+    {
+      "entropy": 0.13986097276210785,
+      "epoch": 6.097629796839729,
+      "grad_norm": 1.9457536935806274,
+      "learning_rate": 4.2038418537296126e-06,
+      "loss": 0.032,
+      "mean_token_accuracy": 0.9908498883247375,
+      "num_tokens": 87714086.0,
+      "step": 10805
+    },
+    {
+      "entropy": 0.14706320762634278,
+      "epoch": 6.100451467268623,
+      "grad_norm": 1.7336926460266113,
+      "learning_rate": 4.203162044362276e-06,
+      "loss": 0.0347,
+      "mean_token_accuracy": 0.9906623005867005,
+      "num_tokens": 87753943.0,
+      "step": 10810
+    },
+    {
+      "entropy": 0.1470971792936325,
+      "epoch": 6.103273137697517,
+      "grad_norm": 2.0381176471710205,
+      "learning_rate": 4.2024820231898655e-06,
+      "loss": 0.0328,
+      "mean_token_accuracy": 0.9907041072845459,
+      "num_tokens": 87794673.0,
+      "step": 10815
+    },
+    {
+      "entropy": 0.1387518674135208,
+      "epoch": 6.106094808126411,
+      "grad_norm": 1.6845587491989136,
+      "learning_rate": 4.201801790345971e-06,
+      "loss": 0.0372,
+      "mean_token_accuracy": 0.9895907640457153,
+      "num_tokens": 87835382.0,
+      "step": 10820
+    },
+    {
+      "entropy": 0.14503337144851686,
+      "epoch": 6.1089164785553045,
+      "grad_norm": 1.6855789422988892,
+      "learning_rate": 4.201121345964225e-06,
+      "loss": 0.0334,
+      "mean_token_accuracy": 0.9907157897949219,
+      "num_tokens": 87876038.0,
+      "step": 10825
+    },
+    {
+      "entropy": 0.13844524025917054,
+      "epoch": 6.111738148984198,
+      "grad_norm": 1.9881229400634766,
+      "learning_rate": 4.200440690178301e-06,
+      "loss": 0.0339,
+      "mean_token_accuracy": 0.9900532841682435,
+      "num_tokens": 87916705.0,
+      "step": 10830
+    },
+    {
+      "entropy": 0.15506267845630645,
+      "epoch": 6.114559819413093,
+      "grad_norm": 1.9544332027435303,
+      "learning_rate": 4.199759823121914e-06,
+      "loss": 0.0346,
+      "mean_token_accuracy": 0.9900209069252014,
+      "num_tokens": 87957566.0,
+      "step": 10835
+    },
+    {
+      "entropy": 0.16678837686777115,
+      "epoch": 6.1173814898419865,
+      "grad_norm": 1.6686840057373047,
+      "learning_rate": 4.199078744928819e-06,
+      "loss": 0.0337,
+      "mean_token_accuracy": 0.9909396409988404,
+      "num_tokens": 87997564.0,
+      "step": 10840
+    },
+    {
+      "entropy": 0.14651824533939362,
+      "epoch": 6.12020316027088,
+      "grad_norm": 1.7264286279678345,
+      "learning_rate": 4.198397455732816e-06,
+      "loss": 0.0373,
+      "mean_token_accuracy": 0.9898837804794312,
+      "num_tokens": 88037998.0,
+      "step": 10845
+    },
+    {
+      "entropy": 0.15673071444034575,
+      "epoch": 6.123024830699774,
+      "grad_norm": 1.6216492652893066,
+      "learning_rate": 4.197715955667742e-06,
+      "loss": 0.0348,
+      "mean_token_accuracy": 0.9899758458137512,
+      "num_tokens": 88078569.0,
+      "step": 10850
+    },
+    {
+      "entropy": 0.1553318828344345,
+      "epoch": 6.1258465011286685,
+      "grad_norm": 2.029632329940796,
+      "learning_rate": 4.19703424486748e-06,
+      "loss": 0.0277,
+      "mean_token_accuracy": 0.9923609852790832,
+      "num_tokens": 88118867.0,
+      "step": 10855
+    },
+    {
+      "entropy": 0.13455870598554612,
+      "epoch": 6.128668171557562,
+      "grad_norm": 1.8276398181915283,
+      "learning_rate": 4.196352323465951e-06,
+      "loss": 0.0256,
+      "mean_token_accuracy": 0.9924584746360778,
+      "num_tokens": 88159338.0,
+      "step": 10860
+    },
+    {
+      "entropy": 0.13744568973779678,
+      "epoch": 6.131489841986456,
+      "grad_norm": 2.144724130630493,
+      "learning_rate": 4.1956701915971196e-06,
+      "loss": 0.0358,
+      "mean_token_accuracy": 0.9901370286941529,
+      "num_tokens": 88199943.0,
+      "step": 10865
+    },
+    {
+      "entropy": 0.1540704548358917,
+      "epoch": 6.1343115124153496,
+      "grad_norm": 2.058373212814331,
+      "learning_rate": 4.194987849394988e-06,
+      "loss": 0.0333,
+      "mean_token_accuracy": 0.9897918820381164,
+      "num_tokens": 88240426.0,
+      "step": 10870
+    },
+    {
+      "entropy": 0.14623642861843109,
+      "epoch": 6.137133182844244,
+      "grad_norm": 2.4471840858459473,
+      "learning_rate": 4.194305296993606e-06,
+      "loss": 0.0359,
+      "mean_token_accuracy": 0.9897229671478271,
+      "num_tokens": 88281095.0,
+      "step": 10875
+    },
+    {
+      "entropy": 0.1303885832428932,
+      "epoch": 6.139954853273138,
+      "grad_norm": 1.7725268602371216,
+      "learning_rate": 4.193622534527058e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9896966218948364,
+      "num_tokens": 88321683.0,
+      "step": 10880
+    },
+    {
+      "entropy": 0.13377143442630768,
+      "epoch": 6.1427765237020315,
+      "grad_norm": 2.0175435543060303,
+      "learning_rate": 4.192939562129476e-06,
+      "loss": 0.0306,
+      "mean_token_accuracy": 0.9910183906555176,
+      "num_tokens": 88362423.0,
+      "step": 10885
+    },
+    {
+      "entropy": 0.13516291081905366,
+      "epoch": 6.145598194130925,
+      "grad_norm": 2.2618424892425537,
+      "learning_rate": 4.192256379935027e-06,
+      "loss": 0.0277,
+      "mean_token_accuracy": 0.9920958876609802,
+      "num_tokens": 88402934.0,
+      "step": 10890
+    },
+    {
+      "entropy": 0.14001532047986984,
+      "epoch": 6.14841986455982,
+      "grad_norm": 1.8463703393936157,
+      "learning_rate": 4.191572988077924e-06,
+      "loss": 0.0333,
+      "mean_token_accuracy": 0.9904056191444397,
+      "num_tokens": 88443748.0,
+      "step": 10895
+    },
+    {
+      "entropy": 0.12817184776067733,
+      "epoch": 6.1512415349887135,
+      "grad_norm": 1.4544408321380615,
+      "learning_rate": 4.190889386692418e-06,
+      "loss": 0.0263,
+      "mean_token_accuracy": 0.9931593537330627,
+      "num_tokens": 88484016.0,
+      "step": 10900
+    },
+    {
+      "entropy": 0.14389062374830247,
+      "epoch": 6.154063205417607,
+      "grad_norm": 1.9694024324417114,
+      "learning_rate": 4.190205575912804e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9898261427879333,
+      "num_tokens": 88524836.0,
+      "step": 10905
+    },
+    {
+      "entropy": 0.14219014048576356,
+      "epoch": 6.156884875846501,
+      "grad_norm": 2.0647833347320557,
+      "learning_rate": 4.189521555873416e-06,
+      "loss": 0.035,
+      "mean_token_accuracy": 0.9893984079360962,
+      "num_tokens": 88565548.0,
+      "step": 10910
+    },
+    {
+      "entropy": 0.14767142832279206,
+      "epoch": 6.159706546275395,
+      "grad_norm": 1.9720888137817383,
+      "learning_rate": 4.18883732670863e-06,
+      "loss": 0.0334,
+      "mean_token_accuracy": 0.9906517505645752,
+      "num_tokens": 88605978.0,
+      "step": 10915
+    },
+    {
+      "entropy": 0.14821238815784454,
+      "epoch": 6.162528216704289,
+      "grad_norm": 2.0277600288391113,
+      "learning_rate": 4.188152888552864e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.9893009543418885,
+      "num_tokens": 88646706.0,
+      "step": 10920
+    },
+    {
+      "entropy": 0.15523377060890198,
+      "epoch": 6.165349887133183,
+      "grad_norm": 1.6985244750976562,
+      "learning_rate": 4.1874682415405735e-06,
+      "loss": 0.0361,
+      "mean_token_accuracy": 0.9894434928894043,
+      "num_tokens": 88687198.0,
+      "step": 10925
+    },
+    {
+      "entropy": 0.15274632275104522,
+      "epoch": 6.168171557562077,
+      "grad_norm": 1.9225730895996094,
+      "learning_rate": 4.186783385806259e-06,
+      "loss": 0.0328,
+      "mean_token_accuracy": 0.9911539673805236,
+      "num_tokens": 88727828.0,
+      "step": 10930
+    },
+    {
+      "entropy": 0.13707538247108458,
+      "epoch": 6.17099322799097,
+      "grad_norm": 1.9640833139419556,
+      "learning_rate": 4.186098321484459e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9896294713020325,
+      "num_tokens": 88768488.0,
+      "step": 10935
+    },
+    {
+      "entropy": 0.14530183970928193,
+      "epoch": 6.173814898419865,
+      "grad_norm": 1.6039245128631592,
+      "learning_rate": 4.185413048709757e-06,
+      "loss": 0.0314,
+      "mean_token_accuracy": 0.9907344460487366,
+      "num_tokens": 88809035.0,
+      "step": 10940
+    },
+    {
+      "entropy": 0.14951497465372085,
+      "epoch": 6.176636568848759,
+      "grad_norm": 1.7432762384414673,
+      "learning_rate": 4.184727567616775e-06,
+      "loss": 0.0387,
+      "mean_token_accuracy": 0.9889516711235047,
+      "num_tokens": 88849740.0,
+      "step": 10945
+    },
+    {
+      "entropy": 0.14462248533964156,
+      "epoch": 6.179458239277652,
+      "grad_norm": 1.5465824604034424,
+      "learning_rate": 4.184041878340174e-06,
+      "loss": 0.0338,
+      "mean_token_accuracy": 0.990413224697113,
+      "num_tokens": 88890461.0,
+      "step": 10950
+    },
+    {
+      "entropy": 0.1421406477689743,
+      "epoch": 6.182279909706546,
+      "grad_norm": 2.111994743347168,
+      "learning_rate": 4.183355981014658e-06,
+      "loss": 0.0405,
+      "mean_token_accuracy": 0.9883662700653076,
+      "num_tokens": 88931166.0,
+      "step": 10955
+    },
+    {
+      "entropy": 0.1487989455461502,
+      "epoch": 6.1851015801354405,
+      "grad_norm": 2.039963722229004,
+      "learning_rate": 4.1826698757749715e-06,
+      "loss": 0.0361,
+      "mean_token_accuracy": 0.989516270160675,
+      "num_tokens": 88971798.0,
+      "step": 10960
+    },
+    {
+      "entropy": 0.14600317776203156,
+      "epoch": 6.187923250564334,
+      "grad_norm": 1.9209233522415161,
+      "learning_rate": 4.1819835627559e-06,
+      "loss": 0.0341,
+      "mean_token_accuracy": 0.9893284320831299,
+      "num_tokens": 89012519.0,
+      "step": 10965
+    },
+    {
+      "entropy": 0.1442936509847641,
+      "epoch": 6.190744920993228,
+      "grad_norm": 1.5550873279571533,
+      "learning_rate": 4.1812970420922725e-06,
+      "loss": 0.0373,
+      "mean_token_accuracy": 0.9887455940246582,
+      "num_tokens": 89053198.0,
+      "step": 10970
+    },
+    {
+      "entropy": 0.15094742476940154,
+      "epoch": 6.193566591422122,
+      "grad_norm": 1.8063254356384277,
+      "learning_rate": 4.180610313918952e-06,
+      "loss": 0.0401,
+      "mean_token_accuracy": 0.9881560444831848,
+      "num_tokens": 89093860.0,
+      "step": 10975
+    },
+    {
+      "entropy": 0.15021807849407195,
+      "epoch": 6.196388261851016,
+      "grad_norm": 2.1231496334075928,
+      "learning_rate": 4.1799233783708474e-06,
+      "loss": 0.0369,
+      "mean_token_accuracy": 0.9901463985443115,
+      "num_tokens": 89134320.0,
+      "step": 10980
+    },
+    {
+      "entropy": 0.1368527665734291,
+      "epoch": 6.19920993227991,
+      "grad_norm": 2.1702585220336914,
+      "learning_rate": 4.1792362355829094e-06,
+      "loss": 0.0336,
+      "mean_token_accuracy": 0.9903720498085022,
+      "num_tokens": 89175018.0,
+      "step": 10985
+    },
+    {
+      "entropy": 0.15346645712852477,
+      "epoch": 6.202031602708804,
+      "grad_norm": 1.8872740268707275,
+      "learning_rate": 4.178548885690126e-06,
+      "loss": 0.0384,
+      "mean_token_accuracy": 0.9887521862983704,
+      "num_tokens": 89215458.0,
+      "step": 10990
+    },
+    {
+      "entropy": 0.14233822673559188,
+      "epoch": 6.204853273137697,
+      "grad_norm": 1.8049365282058716,
+      "learning_rate": 4.177861328827526e-06,
+      "loss": 0.0391,
+      "mean_token_accuracy": 0.9878924250602722,
+      "num_tokens": 89255912.0,
+      "step": 10995
+    },
+    {
+      "entropy": 0.14570232629776,
+      "epoch": 6.207674943566591,
+      "grad_norm": 1.8370901346206665,
+      "learning_rate": 4.1771735651301815e-06,
+      "loss": 0.0314,
+      "mean_token_accuracy": 0.9908752083778382,
+      "num_tokens": 89296589.0,
+      "step": 11000
+    },
+    {
+      "epoch": 6.207674943566591,
+      "eval_entropy": 0.20263180136680603,
+      "eval_loss": 0.02532457746565342,
+      "eval_mean_token_accuracy": 0.9938673973083496,
+      "eval_num_tokens": 89296589.0,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 24.369,
+      "eval_steps_per_second": 6.092,
+      "step": 11000
+    },
+    {
+      "entropy": 0.14294780194759368,
+      "epoch": 6.210496613995486,
+      "grad_norm": 1.8399924039840698,
+      "learning_rate": 4.176485594733203e-06,
+      "loss": 0.037,
+      "mean_token_accuracy": 0.9889877915382386,
+      "num_tokens": 89337302.0,
+      "step": 11005
+    },
+    {
+      "entropy": 0.14533422291278839,
+      "epoch": 6.213318284424379,
+      "grad_norm": 1.7887989282608032,
+      "learning_rate": 4.175797417771744e-06,
+      "loss": 0.0325,
+      "mean_token_accuracy": 0.9902246236801148,
+      "num_tokens": 89377799.0,
+      "step": 11010
+    },
+    {
+      "entropy": 0.1417243927717209,
+      "epoch": 6.216139954853273,
+      "grad_norm": 1.6449092626571655,
+      "learning_rate": 4.175109034380994e-06,
+      "loss": 0.0311,
+      "mean_token_accuracy": 0.9912659168243408,
+      "num_tokens": 89418429.0,
+      "step": 11015
+    },
+    {
+      "entropy": 0.14027564078569413,
+      "epoch": 6.218961625282167,
+      "grad_norm": 1.5085475444793701,
+      "learning_rate": 4.1744204446961885e-06,
+      "loss": 0.031,
+      "mean_token_accuracy": 0.9910840392112732,
+      "num_tokens": 89458957.0,
+      "step": 11020
+    },
+    {
+      "entropy": 0.1335515111684799,
+      "epoch": 6.221783295711061,
+      "grad_norm": 1.927368402481079,
+      "learning_rate": 4.1737316488526005e-06,
+      "loss": 0.0369,
+      "mean_token_accuracy": 0.9889877438545227,
+      "num_tokens": 89499568.0,
+      "step": 11025
+    },
+    {
+      "entropy": 0.14035135358572007,
+      "epoch": 6.224604966139955,
+      "grad_norm": 1.654311180114746,
+      "learning_rate": 4.173042646985544e-06,
+      "loss": 0.0325,
+      "mean_token_accuracy": 0.9901265263557434,
+      "num_tokens": 89540184.0,
+      "step": 11030
+    },
+    {
+      "entropy": 0.14204625636339188,
+      "epoch": 6.227426636568849,
+      "grad_norm": 1.987782597541809,
+      "learning_rate": 4.172353439230372e-06,
+      "loss": 0.0328,
+      "mean_token_accuracy": 0.9901327252388,
+      "num_tokens": 89580911.0,
+      "step": 11035
+    },
+    {
+      "entropy": 0.15092624723911285,
+      "epoch": 6.230248306997742,
+      "grad_norm": 1.7647333145141602,
+      "learning_rate": 4.1716640257224815e-06,
+      "loss": 0.0374,
+      "mean_token_accuracy": 0.9883796095848083,
+      "num_tokens": 89620532.0,
+      "step": 11040
+    },
+    {
+      "entropy": 0.15512611269950866,
+      "epoch": 6.233069977426637,
+      "grad_norm": 1.8042088747024536,
+      "learning_rate": 4.170974406597307e-06,
+      "loss": 0.0377,
+      "mean_token_accuracy": 0.9896431684494018,
+      "num_tokens": 89661252.0,
+      "step": 11045
+    },
+    {
+      "entropy": 0.13535745441913605,
+      "epoch": 6.235891647855531,
+      "grad_norm": 2.0197017192840576,
+      "learning_rate": 4.170284581990325e-06,
+      "loss": 0.0321,
+      "mean_token_accuracy": 0.9908074617385865,
+      "num_tokens": 89702049.0,
+      "step": 11050
+    },
+    {
+      "entropy": 0.14527721405029298,
+      "epoch": 6.238713318284424,
+      "grad_norm": 2.0421767234802246,
+      "learning_rate": 4.1695945520370505e-06,
+      "loss": 0.0355,
+      "mean_token_accuracy": 0.9893774628639221,
+      "num_tokens": 89742588.0,
+      "step": 11055
+    },
+    {
+      "entropy": 0.159025439620018,
+      "epoch": 6.241534988713318,
+      "grad_norm": 2.0698342323303223,
+      "learning_rate": 4.16890431687304e-06,
+      "loss": 0.0317,
+      "mean_token_accuracy": 0.9911215782165528,
+      "num_tokens": 89783240.0,
+      "step": 11060
+    },
+    {
+      "entropy": 0.15658071339130403,
+      "epoch": 6.244356659142213,
+      "grad_norm": 1.8461730480194092,
+      "learning_rate": 4.168213876633891e-06,
+      "loss": 0.0354,
+      "mean_token_accuracy": 0.988905155658722,
+      "num_tokens": 89823822.0,
+      "step": 11065
+    },
+    {
+      "entropy": 0.13386445641517639,
+      "epoch": 6.247178329571106,
+      "grad_norm": 2.02535343170166,
+      "learning_rate": 4.167523231455241e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9896762609481812,
+      "num_tokens": 89864500.0,
+      "step": 11070
+    },
+    {
+      "entropy": 0.14109272211790086,
+      "epoch": 6.25,
+      "grad_norm": 2.1409685611724854,
+      "learning_rate": 4.166832381472766e-06,
+      "loss": 0.0341,
+      "mean_token_accuracy": 0.9898652195930481,
+      "num_tokens": 89905040.0,
+      "step": 11075
+    },
+    {
+      "entropy": 0.1469561368227005,
+      "epoch": 6.252821670428894,
+      "grad_norm": 1.8599631786346436,
+      "learning_rate": 4.166141326822184e-06,
+      "loss": 0.0307,
+      "mean_token_accuracy": 0.9907646656036377,
+      "num_tokens": 89945211.0,
+      "step": 11080
+    },
+    {
+      "entropy": 0.13446701616048812,
+      "epoch": 6.255643340857787,
+      "grad_norm": 1.5020068883895874,
+      "learning_rate": 4.165450067639254e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.9891453385353088,
+      "num_tokens": 89985851.0,
+      "step": 11085
+    },
+    {
+      "entropy": 0.14148662984371185,
+      "epoch": 6.258465011286682,
+      "grad_norm": 2.185192346572876,
+      "learning_rate": 4.164758604059772e-06,
+      "loss": 0.0323,
+      "mean_token_accuracy": 0.9905752301216125,
+      "num_tokens": 90026635.0,
+      "step": 11090
+    },
+    {
+      "entropy": 0.13741403222084045,
+      "epoch": 6.261286681715576,
+      "grad_norm": 1.7420563697814941,
+      "learning_rate": 4.164066936219577e-06,
+      "loss": 0.0321,
+      "mean_token_accuracy": 0.9903434753417969,
+      "num_tokens": 90066598.0,
+      "step": 11095
+    },
+    {
+      "entropy": 0.14443300664424896,
+      "epoch": 6.264108352144469,
+      "grad_norm": 1.9935277700424194,
+      "learning_rate": 4.163375064254549e-06,
+      "loss": 0.041,
+      "mean_token_accuracy": 0.988189947605133,
+      "num_tokens": 90107478.0,
+      "step": 11100
+    },
+    {
+      "entropy": 0.1385072499513626,
+      "epoch": 6.266930022573363,
+      "grad_norm": 2.011894702911377,
+      "learning_rate": 4.162682988300602e-06,
+      "loss": 0.0394,
+      "mean_token_accuracy": 0.9881589770317077,
+      "num_tokens": 90148095.0,
+      "step": 11105
+    },
+    {
+      "entropy": 0.15410477817058563,
+      "epoch": 6.269751693002258,
+      "grad_norm": 1.801830530166626,
+      "learning_rate": 4.1619907084937e-06,
+      "loss": 0.0365,
+      "mean_token_accuracy": 0.9895232796669007,
+      "num_tokens": 90188661.0,
+      "step": 11110
+    },
+    {
+      "entropy": 0.1497156322002411,
+      "epoch": 6.272573363431151,
+      "grad_norm": 2.2699999809265137,
+      "learning_rate": 4.161298224969836e-06,
+      "loss": 0.0365,
+      "mean_token_accuracy": 0.9894330143928528,
+      "num_tokens": 90229364.0,
+      "step": 11115
+    },
+    {
+      "entropy": 0.14462217986583709,
+      "epoch": 6.275395033860045,
+      "grad_norm": 2.064868450164795,
+      "learning_rate": 4.1606055378650516e-06,
+      "loss": 0.034,
+      "mean_token_accuracy": 0.9895681977272034,
+      "num_tokens": 90270064.0,
+      "step": 11120
+    },
+    {
+      "entropy": 0.14862312972545624,
+      "epoch": 6.278216704288939,
+      "grad_norm": 1.742047905921936,
+      "learning_rate": 4.159912647315425e-06,
+      "loss": 0.0354,
+      "mean_token_accuracy": 0.9892676591873169,
+      "num_tokens": 90310329.0,
+      "step": 11125
+    },
+    {
+      "entropy": 0.14780815839767455,
+      "epoch": 6.281038374717833,
+      "grad_norm": 1.733389139175415,
+      "learning_rate": 4.159219553457074e-06,
+      "loss": 0.0338,
+      "mean_token_accuracy": 0.9897500872612,
+      "num_tokens": 90351138.0,
+      "step": 11130
+    },
+    {
+      "entropy": 0.13568034768104553,
+      "epoch": 6.283860045146727,
+      "grad_norm": 1.8001052141189575,
+      "learning_rate": 4.158526256426158e-06,
+      "loss": 0.0333,
+      "mean_token_accuracy": 0.9904277682304382,
+      "num_tokens": 90391910.0,
+      "step": 11135
+    },
+    {
+      "entropy": 0.14421828389167785,
+      "epoch": 6.286681715575621,
+      "grad_norm": 2.015634298324585,
+      "learning_rate": 4.157832756358874e-06,
+      "loss": 0.0406,
+      "mean_token_accuracy": 0.9873695731163025,
+      "num_tokens": 90432746.0,
+      "step": 11140
+    },
+    {
+      "entropy": 0.14020881354808806,
+      "epoch": 6.289503386004514,
+      "grad_norm": 2.3647797107696533,
+      "learning_rate": 4.157139053391461e-06,
+      "loss": 0.0362,
+      "mean_token_accuracy": 0.9892059206962586,
+      "num_tokens": 90473415.0,
+      "step": 11145
+    },
+    {
+      "entropy": 0.1497344046831131,
+      "epoch": 6.292325056433409,
+      "grad_norm": 2.052783727645874,
+      "learning_rate": 4.156445147660197e-06,
+      "loss": 0.0382,
+      "mean_token_accuracy": 0.9883453965187072,
+      "num_tokens": 90514134.0,
+      "step": 11150
+    },
+    {
+      "entropy": 0.14050848484039308,
+      "epoch": 6.295146726862303,
+      "grad_norm": 1.6089463233947754,
+      "learning_rate": 4.1557510393014e-06,
+      "loss": 0.0388,
+      "mean_token_accuracy": 0.9887239336967468,
+      "num_tokens": 90554962.0,
+      "step": 11155
+    },
+    {
+      "entropy": 0.14407500624656677,
+      "epoch": 6.297968397291196,
+      "grad_norm": 2.434351682662964,
+      "learning_rate": 4.155056728451426e-06,
+      "loss": 0.0376,
+      "mean_token_accuracy": 0.9894903421401977,
+      "num_tokens": 90595436.0,
+      "step": 11160
+    },
+    {
+      "entropy": 0.13423256427049637,
+      "epoch": 6.30079006772009,
+      "grad_norm": 2.1055705547332764,
+      "learning_rate": 4.154362215246675e-06,
+      "loss": 0.0352,
+      "mean_token_accuracy": 0.9897322297096253,
+      "num_tokens": 90636082.0,
+      "step": 11165
+    },
+    {
+      "entropy": 0.1400793805718422,
+      "epoch": 6.303611738148984,
+      "grad_norm": 2.3050198554992676,
+      "learning_rate": 4.1536674998235825e-06,
+      "loss": 0.0382,
+      "mean_token_accuracy": 0.9891103863716125,
+      "num_tokens": 90676676.0,
+      "step": 11170
+    },
+    {
+      "entropy": 0.1425952732563019,
+      "epoch": 6.306433408577878,
+      "grad_norm": 2.2005064487457275,
+      "learning_rate": 4.152972582318626e-06,
+      "loss": 0.0357,
+      "mean_token_accuracy": 0.9899725079536438,
+      "num_tokens": 90717384.0,
+      "step": 11175
+    },
+    {
+      "entropy": 0.1526680827140808,
+      "epoch": 6.309255079006772,
+      "grad_norm": 1.6304471492767334,
+      "learning_rate": 4.152277462868321e-06,
+      "loss": 0.0371,
+      "mean_token_accuracy": 0.9893090009689331,
+      "num_tokens": 90757930.0,
+      "step": 11180
+    },
+    {
+      "entropy": 0.14085469841957093,
+      "epoch": 6.312076749435666,
+      "grad_norm": 1.5589280128479004,
+      "learning_rate": 4.1515821416092264e-06,
+      "loss": 0.0282,
+      "mean_token_accuracy": 0.9919806241989135,
+      "num_tokens": 90798800.0,
+      "step": 11185
+    },
+    {
+      "entropy": 0.13670676797628403,
+      "epoch": 6.3148984198645595,
+      "grad_norm": 1.8319333791732788,
+      "learning_rate": 4.150886618677936e-06,
+      "loss": 0.0367,
+      "mean_token_accuracy": 0.9886664271354675,
+      "num_tokens": 90838684.0,
+      "step": 11190
+    },
+    {
+      "entropy": 0.16441910862922668,
+      "epoch": 6.317720090293454,
+      "grad_norm": 2.1681110858917236,
+      "learning_rate": 4.150190894211087e-06,
+      "loss": 0.0416,
+      "mean_token_accuracy": 0.9879634499549865,
+      "num_tokens": 90879475.0,
+      "step": 11195
+    },
+    {
+      "entropy": 0.15294736325740815,
+      "epoch": 6.320541760722348,
+      "grad_norm": 1.9503849744796753,
+      "learning_rate": 4.1494949683453525e-06,
+      "loss": 0.033,
+      "mean_token_accuracy": 0.990665853023529,
+      "num_tokens": 90920247.0,
+      "step": 11200
+    },
+    {
+      "entropy": 0.13768713921308517,
+      "epoch": 6.323363431151241,
+      "grad_norm": 1.801827311515808,
+      "learning_rate": 4.148798841217448e-06,
+      "loss": 0.036,
+      "mean_token_accuracy": 0.9888361096382141,
+      "num_tokens": 90960908.0,
+      "step": 11205
+    },
+    {
+      "entropy": 0.1548486292362213,
+      "epoch": 6.326185101580135,
+      "grad_norm": 2.232868194580078,
+      "learning_rate": 4.148102512964129e-06,
+      "loss": 0.0361,
+      "mean_token_accuracy": 0.9894163966178894,
+      "num_tokens": 91001543.0,
+      "step": 11210
+    },
+    {
+      "entropy": 0.13173782229423522,
+      "epoch": 6.32900677200903,
+      "grad_norm": 1.801491141319275,
+      "learning_rate": 4.1474059837221884e-06,
+      "loss": 0.0347,
+      "mean_token_accuracy": 0.9903063178062439,
+      "num_tokens": 91042245.0,
+      "step": 11215
+    },
+    {
+      "entropy": 0.14220143854618073,
+      "epoch": 6.331828442437923,
+      "grad_norm": 2.110448122024536,
+      "learning_rate": 4.146709253628458e-06,
+      "loss": 0.0377,
+      "mean_token_accuracy": 0.9890905261039734,
+      "num_tokens": 91082706.0,
+      "step": 11220
+    },
+    {
+      "entropy": 0.1490049421787262,
+      "epoch": 6.334650112866817,
+      "grad_norm": 2.0357728004455566,
+      "learning_rate": 4.146012322819814e-06,
+      "loss": 0.0372,
+      "mean_token_accuracy": 0.9890268802642822,
+      "num_tokens": 91123300.0,
+      "step": 11225
+    },
+    {
+      "entropy": 0.16816553473472595,
+      "epoch": 6.337471783295711,
+      "grad_norm": 2.126213550567627,
+      "learning_rate": 4.145315191433165e-06,
+      "loss": 0.0379,
+      "mean_token_accuracy": 0.989051103591919,
+      "num_tokens": 91163864.0,
+      "step": 11230
+    },
+    {
+      "entropy": 0.15646686255931855,
+      "epoch": 6.340293453724605,
+      "grad_norm": 1.7959010601043701,
+      "learning_rate": 4.144617859605464e-06,
+      "loss": 0.0367,
+      "mean_token_accuracy": 0.9899686336517334,
+      "num_tokens": 91204489.0,
+      "step": 11235
+    },
+    {
+      "entropy": 0.14400748908519745,
+      "epoch": 6.343115124153499,
+      "grad_norm": 1.6874085664749146,
+      "learning_rate": 4.1439203274737015e-06,
+      "loss": 0.0352,
+      "mean_token_accuracy": 0.9898784995079041,
+      "num_tokens": 91245210.0,
+      "step": 11240
+    },
+    {
+      "entropy": 0.12948621511459352,
+      "epoch": 6.345936794582393,
+      "grad_norm": 1.4666175842285156,
+      "learning_rate": 4.143222595174909e-06,
+      "loss": 0.0392,
+      "mean_token_accuracy": 0.9886078119277955,
+      "num_tokens": 91285941.0,
+      "step": 11245
+    },
+    {
+      "entropy": 0.14075332880020142,
+      "epoch": 6.3487584650112865,
+      "grad_norm": 2.0497982501983643,
+      "learning_rate": 4.142524662846156e-06,
+      "loss": 0.0376,
+      "mean_token_accuracy": 0.9891217827796936,
+      "num_tokens": 91326822.0,
+      "step": 11250
+    },
+    {
+      "entropy": 0.15241701006889344,
+      "epoch": 6.35158013544018,
+      "grad_norm": 1.943670392036438,
+      "learning_rate": 4.14182653062455e-06,
+      "loss": 0.0322,
+      "mean_token_accuracy": 0.9905896425247193,
+      "num_tokens": 91367515.0,
+      "step": 11255
+    },
+    {
+      "entropy": 0.1496208757162094,
+      "epoch": 6.354401805869075,
+      "grad_norm": 2.349369525909424,
+      "learning_rate": 4.14112819864724e-06,
+      "loss": 0.0394,
+      "mean_token_accuracy": 0.9895246982574463,
+      "num_tokens": 91408287.0,
+      "step": 11260
+    },
+    {
+      "entropy": 0.12518291175365448,
+      "epoch": 6.3572234762979685,
+      "grad_norm": 1.957039475440979,
+      "learning_rate": 4.140429667051412e-06,
+      "loss": 0.034,
+      "mean_token_accuracy": 0.9896358966827392,
+      "num_tokens": 91448827.0,
+      "step": 11265
+    },
+    {
+      "entropy": 0.13367379158735276,
+      "epoch": 6.360045146726862,
+      "grad_norm": 1.9315341711044312,
+      "learning_rate": 4.139730935974295e-06,
+      "loss": 0.0331,
+      "mean_token_accuracy": 0.9905829906463623,
+      "num_tokens": 91489610.0,
+      "step": 11270
+    },
+    {
+      "entropy": 0.13425323665142058,
+      "epoch": 6.362866817155756,
+      "grad_norm": 1.8871325254440308,
+      "learning_rate": 4.1390320055531545e-06,
+      "loss": 0.0374,
+      "mean_token_accuracy": 0.9884009838104248,
+      "num_tokens": 91530511.0,
+      "step": 11275
+    },
+    {
+      "entropy": 0.1478370100259781,
+      "epoch": 6.3656884875846504,
+      "grad_norm": 1.611417531967163,
+      "learning_rate": 4.1383328759252935e-06,
+      "loss": 0.031,
+      "mean_token_accuracy": 0.9910028457641602,
+      "num_tokens": 91571131.0,
+      "step": 11280
+    },
+    {
+      "entropy": 0.14289433658123016,
+      "epoch": 6.368510158013544,
+      "grad_norm": 1.8513978719711304,
+      "learning_rate": 4.137633547228058e-06,
+      "loss": 0.0401,
+      "mean_token_accuracy": 0.9881711721420288,
+      "num_tokens": 91612000.0,
+      "step": 11285
+    },
+    {
+      "entropy": 0.1438445121049881,
+      "epoch": 6.371331828442438,
+      "grad_norm": 2.4446709156036377,
+      "learning_rate": 4.13693401959883e-06,
+      "loss": 0.0344,
+      "mean_token_accuracy": 0.9899812698364258,
+      "num_tokens": 91652645.0,
+      "step": 11290
+    },
+    {
+      "entropy": 0.1517265945672989,
+      "epoch": 6.3741534988713315,
+      "grad_norm": 1.8532661199569702,
+      "learning_rate": 4.136234293175033e-06,
+      "loss": 0.0325,
+      "mean_token_accuracy": 0.9901606917381287,
+      "num_tokens": 91693216.0,
+      "step": 11295
+    },
+    {
+      "entropy": 0.1488794654607773,
+      "epoch": 6.376975169300226,
+      "grad_norm": 1.7429970502853394,
+      "learning_rate": 4.135534368094127e-06,
+      "loss": 0.0413,
+      "mean_token_accuracy": 0.9876135468482972,
+      "num_tokens": 91733904.0,
+      "step": 11300
+    },
+    {
+      "entropy": 0.15159515142440796,
+      "epoch": 6.37979683972912,
+      "grad_norm": 1.6426184177398682,
+      "learning_rate": 4.1348342444936134e-06,
+      "loss": 0.0405,
+      "mean_token_accuracy": 0.9885798454284668,
+      "num_tokens": 91774778.0,
+      "step": 11305
+    },
+    {
+      "entropy": 0.14828484505414963,
+      "epoch": 6.3826185101580135,
+      "grad_norm": 1.2837458848953247,
+      "learning_rate": 4.134133922511032e-06,
+      "loss": 0.0389,
+      "mean_token_accuracy": 0.9885473847389221,
+      "num_tokens": 91815379.0,
+      "step": 11310
+    },
+    {
+      "entropy": 0.13970376253128053,
+      "epoch": 6.385440180586907,
+      "grad_norm": 1.9679341316223145,
+      "learning_rate": 4.133433402283958e-06,
+      "loss": 0.0327,
+      "mean_token_accuracy": 0.9905824422836303,
+      "num_tokens": 91855882.0,
+      "step": 11315
+    },
+    {
+      "entropy": 0.13457007706165314,
+      "epoch": 6.388261851015802,
+      "grad_norm": 1.7451153993606567,
+      "learning_rate": 4.132732683950013e-06,
+      "loss": 0.0351,
+      "mean_token_accuracy": 0.9886767625808716,
+      "num_tokens": 91896363.0,
+      "step": 11320
+    },
+    {
+      "entropy": 0.14434319883584976,
+      "epoch": 6.3910835214446955,
+      "grad_norm": 2.0313048362731934,
+      "learning_rate": 4.13203176764685e-06,
+      "loss": 0.0348,
+      "mean_token_accuracy": 0.9894550204277038,
+      "num_tokens": 91936957.0,
+      "step": 11325
+    },
+    {
+      "entropy": 0.14290865659713745,
+      "epoch": 6.393905191873589,
+      "grad_norm": 1.3977835178375244,
+      "learning_rate": 4.131330653512167e-06,
+      "loss": 0.0361,
+      "mean_token_accuracy": 0.9899007439613342,
+      "num_tokens": 91977634.0,
+      "step": 11330
+    },
+    {
+      "entropy": 0.14777366816997528,
+      "epoch": 6.396726862302483,
+      "grad_norm": 1.7547193765640259,
+      "learning_rate": 4.130629341683695e-06,
+      "loss": 0.0359,
+      "mean_token_accuracy": 0.9890438318252563,
+      "num_tokens": 92018299.0,
+      "step": 11335
+    },
+    {
+      "entropy": 0.16221860945224761,
+      "epoch": 6.399548532731377,
+      "grad_norm": 2.0626728534698486,
+      "learning_rate": 4.129927832299209e-06,
+      "loss": 0.0361,
+      "mean_token_accuracy": 0.989508330821991,
+      "num_tokens": 92058961.0,
+      "step": 11340
+    },
+    {
+      "entropy": 0.13226877599954606,
+      "epoch": 6.402370203160271,
+      "grad_norm": 1.7659369707107544,
+      "learning_rate": 4.129226125496519e-06,
+      "loss": 0.0363,
+      "mean_token_accuracy": 0.9892664194107056,
+      "num_tokens": 92099870.0,
+      "step": 11345
+    },
+    {
+      "entropy": 0.14557516872882842,
+      "epoch": 6.405191873589165,
+      "grad_norm": 2.086122751235962,
+      "learning_rate": 4.128524221413477e-06,
+      "loss": 0.0347,
+      "mean_token_accuracy": 0.9899450778961182,
+      "num_tokens": 92140746.0,
+      "step": 11350
+    },
+    {
+      "entropy": 0.15533352494239808,
+      "epoch": 6.408013544018059,
+      "grad_norm": 2.2414488792419434,
+      "learning_rate": 4.12782212018797e-06,
+      "loss": 0.0414,
+      "mean_token_accuracy": 0.9879919528961182,
+      "num_tokens": 92181327.0,
+      "step": 11355
+    },
+    {
+      "entropy": 0.14174574315547944,
+      "epoch": 6.410835214446952,
+      "grad_norm": 1.9369492530822754,
+      "learning_rate": 4.127119821957927e-06,
+      "loss": 0.0337,
+      "mean_token_accuracy": 0.9892768621444702,
+      "num_tokens": 92221857.0,
+      "step": 11360
+    },
+    {
+      "entropy": 0.13032906055450438,
+      "epoch": 6.413656884875847,
+      "grad_norm": 1.955140471458435,
+      "learning_rate": 4.126417326861316e-06,
+      "loss": 0.0331,
+      "mean_token_accuracy": 0.9905326247215271,
+      "num_tokens": 92262534.0,
+      "step": 11365
+    },
+    {
+      "entropy": 0.13662864565849303,
+      "epoch": 6.4164785553047405,
+      "grad_norm": 1.978060245513916,
+      "learning_rate": 4.1257146350361395e-06,
+      "loss": 0.0436,
+      "mean_token_accuracy": 0.9868420958518982,
+      "num_tokens": 92303036.0,
+      "step": 11370
+    },
+    {
+      "entropy": 0.15215873420238496,
+      "epoch": 6.419300225733634,
+      "grad_norm": 2.898237466812134,
+      "learning_rate": 4.125011746620444e-06,
+      "loss": 0.0403,
+      "mean_token_accuracy": 0.9886619448661804,
+      "num_tokens": 92342876.0,
+      "step": 11375
+    },
+    {
+      "entropy": 0.16599242389202118,
+      "epoch": 6.422121896162528,
+      "grad_norm": 1.9874509572982788,
+      "learning_rate": 4.1243086617523105e-06,
+      "loss": 0.0404,
+      "mean_token_accuracy": 0.98833726644516,
+      "num_tokens": 92383402.0,
+      "step": 11380
+    },
+    {
+      "entropy": 0.14803114980459214,
+      "epoch": 6.4249435665914225,
+      "grad_norm": 2.0754692554473877,
+      "learning_rate": 4.12360538056986e-06,
+      "loss": 0.0355,
+      "mean_token_accuracy": 0.9896466255187988,
+      "num_tokens": 92423873.0,
+      "step": 11385
+    },
+    {
+      "entropy": 0.14688670337200166,
+      "epoch": 6.427765237020316,
+      "grad_norm": 1.7059087753295898,
+      "learning_rate": 4.122901903211254e-06,
+      "loss": 0.0363,
+      "mean_token_accuracy": 0.9896903872489929,
+      "num_tokens": 92464592.0,
+      "step": 11390
+    },
+    {
+      "entropy": 0.137593112885952,
+      "epoch": 6.43058690744921,
+      "grad_norm": 1.807116985321045,
+      "learning_rate": 4.122198229814689e-06,
+      "loss": 0.0329,
+      "mean_token_accuracy": 0.9902451157569885,
+      "num_tokens": 92505357.0,
+      "step": 11395
+    },
+    {
+      "entropy": 0.1294647052884102,
+      "epoch": 6.433408577878104,
+      "grad_norm": 1.739324927330017,
+      "learning_rate": 4.121494360518401e-06,
+      "loss": 0.0359,
+      "mean_token_accuracy": 0.9892550468444824,
+      "num_tokens": 92546032.0,
+      "step": 11400
+    },
+    {
+      "entropy": 0.16134763658046722,
+      "epoch": 6.436230248306998,
+      "grad_norm": 1.8106449842453003,
+      "learning_rate": 4.120790295460668e-06,
+      "loss": 0.0378,
+      "mean_token_accuracy": 0.9890878438949585,
+      "num_tokens": 92586727.0,
+      "step": 11405
+    },
+    {
+      "entropy": 0.13998755365610122,
+      "epoch": 6.439051918735892,
+      "grad_norm": 1.5401268005371094,
+      "learning_rate": 4.1200860347798e-06,
+      "loss": 0.0369,
+      "mean_token_accuracy": 0.9888973951339721,
+      "num_tokens": 92627301.0,
+      "step": 11410
+    },
+    {
+      "entropy": 0.14669421464204788,
+      "epoch": 6.441873589164786,
+      "grad_norm": 1.9903889894485474,
+      "learning_rate": 4.119381578614153e-06,
+      "loss": 0.0388,
+      "mean_token_accuracy": 0.9879800200462341,
+      "num_tokens": 92668052.0,
+      "step": 11415
+    },
+    {
+      "entropy": 0.14707699716091155,
+      "epoch": 6.444695259593679,
+      "grad_norm": 1.9357092380523682,
+      "learning_rate": 4.118676927102115e-06,
+      "loss": 0.0356,
+      "mean_token_accuracy": 0.9894900679588318,
+      "num_tokens": 92708627.0,
+      "step": 11420
+    },
+    {
+      "entropy": 0.14949690401554108,
+      "epoch": 6.447516930022573,
+      "grad_norm": 1.9960310459136963,
+      "learning_rate": 4.117972080382115e-06,
+      "loss": 0.0378,
+      "mean_token_accuracy": 0.9885159015655518,
+      "num_tokens": 92749188.0,
+      "step": 11425
+    },
+    {
+      "entropy": 0.14854080080986024,
+      "epoch": 6.450338600451468,
+      "grad_norm": 1.9244252443313599,
+      "learning_rate": 4.117267038592621e-06,
+      "loss": 0.0391,
+      "mean_token_accuracy": 0.988319730758667,
+      "num_tokens": 92789998.0,
+      "step": 11430
+    },
+    {
+      "entropy": 0.14139661490917205,
+      "epoch": 6.453160270880361,
+      "grad_norm": 1.5770515203475952,
+      "learning_rate": 4.1165618018721385e-06,
+      "loss": 0.0379,
+      "mean_token_accuracy": 0.9883534550666809,
+      "num_tokens": 92830742.0,
+      "step": 11435
+    },
+    {
+      "entropy": 0.1313648045063019,
+      "epoch": 6.455981941309255,
+      "grad_norm": 1.4559335708618164,
+      "learning_rate": 4.115856370359211e-06,
+      "loss": 0.0352,
+      "mean_token_accuracy": 0.9895216822624207,
+      "num_tokens": 92871340.0,
+      "step": 11440
+    },
+    {
+      "entropy": 0.14764813482761383,
+      "epoch": 6.458803611738149,
+      "grad_norm": 1.990279197692871,
+      "learning_rate": 4.115150744192421e-06,
+      "loss": 0.0383,
+      "mean_token_accuracy": 0.9883626222610473,
+      "num_tokens": 92911986.0,
+      "step": 11445
+    },
+    {
+      "entropy": 0.14003989696502686,
+      "epoch": 6.461625282167043,
+      "grad_norm": 2.3407559394836426,
+      "learning_rate": 4.114444923510388e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9898525953292847,
+      "num_tokens": 92952713.0,
+      "step": 11450
+    },
+    {
+      "entropy": 0.13716760277748108,
+      "epoch": 6.464446952595937,
+      "grad_norm": 1.8746510744094849,
+      "learning_rate": 4.113738908451771e-06,
+      "loss": 0.0314,
+      "mean_token_accuracy": 0.9901943325996398,
+      "num_tokens": 92993264.0,
+      "step": 11455
+    },
+    {
+      "entropy": 0.14014555811882018,
+      "epoch": 6.467268623024831,
+      "grad_norm": 1.970062255859375,
+      "learning_rate": 4.113032699155268e-06,
+      "loss": 0.042,
+      "mean_token_accuracy": 0.9883641839027405,
+      "num_tokens": 93033836.0,
+      "step": 11460
+    },
+    {
+      "entropy": 0.15491693019866942,
+      "epoch": 6.470090293453724,
+      "grad_norm": 1.8556119203567505,
+      "learning_rate": 4.112326295759612e-06,
+      "loss": 0.0394,
+      "mean_token_accuracy": 0.9879988074302674,
+      "num_tokens": 93074450.0,
+      "step": 11465
+    },
+    {
+      "entropy": 0.1430242419242859,
+      "epoch": 6.472911963882619,
+      "grad_norm": 2.5160515308380127,
+      "learning_rate": 4.111619698403577e-06,
+      "loss": 0.0413,
+      "mean_token_accuracy": 0.9870900750160218,
+      "num_tokens": 93115187.0,
+      "step": 11470
+    },
+    {
+      "entropy": 0.1444158598780632,
+      "epoch": 6.475733634311513,
+      "grad_norm": 1.70442795753479,
+      "learning_rate": 4.110912907225974e-06,
+      "loss": 0.0362,
+      "mean_token_accuracy": 0.9890662670135498,
+      "num_tokens": 93155595.0,
+      "step": 11475
+    },
+    {
+      "entropy": 0.13908229172229766,
+      "epoch": 6.478555304740406,
+      "grad_norm": 1.6484752893447876,
+      "learning_rate": 4.110205922365652e-06,
+      "loss": 0.0341,
+      "mean_token_accuracy": 0.9901829957962036,
+      "num_tokens": 93196137.0,
+      "step": 11480
+    },
+    {
+      "entropy": 0.1475514531135559,
+      "epoch": 6.4813769751693,
+      "grad_norm": 1.759164571762085,
+      "learning_rate": 4.1094987439615e-06,
+      "loss": 0.0328,
+      "mean_token_accuracy": 0.990956437587738,
+      "num_tokens": 93236943.0,
+      "step": 11485
+    },
+    {
+      "entropy": 0.16040101349353791,
+      "epoch": 6.484198645598195,
+      "grad_norm": 2.2340142726898193,
+      "learning_rate": 4.10879137215244e-06,
+      "loss": 0.038,
+      "mean_token_accuracy": 0.9892443895339966,
+      "num_tokens": 93277309.0,
+      "step": 11490
+    },
+    {
+      "entropy": 0.1610852897167206,
+      "epoch": 6.487020316027088,
+      "grad_norm": 1.9512050151824951,
+      "learning_rate": 4.108083807077437e-06,
+      "loss": 0.0445,
+      "mean_token_accuracy": 0.9860481977462768,
+      "num_tokens": 93317646.0,
+      "step": 11495
+    },
+    {
+      "entropy": 0.15572181046009065,
+      "epoch": 6.489841986455982,
+      "grad_norm": 2.4991040229797363,
+      "learning_rate": 4.1073760488754935e-06,
+      "loss": 0.0448,
+      "mean_token_accuracy": 0.9868571162223816,
+      "num_tokens": 93358277.0,
+      "step": 11500
+    },
+    {
+      "epoch": 6.489841986455982,
+      "eval_entropy": 0.1996920257806778,
+      "eval_loss": 0.023422840982675552,
+      "eval_mean_token_accuracy": 0.9938673973083496,
+      "eval_num_tokens": 93358277.0,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 24.421,
+      "eval_steps_per_second": 6.105,
+      "step": 11500
+    },
+    {
+      "entropy": 0.1445915535092354,
+      "epoch": 6.492663656884876,
+      "grad_norm": 1.8517767190933228,
+      "learning_rate": 4.106668097685647e-06,
+      "loss": 0.0377,
+      "mean_token_accuracy": 0.9889588594436646,
+      "num_tokens": 93398334.0,
+      "step": 11505
+    },
+    {
+      "entropy": 0.15220266580581665,
+      "epoch": 6.495485327313769,
+      "grad_norm": 1.8239866495132446,
+      "learning_rate": 4.105959953646975e-06,
+      "loss": 0.0392,
+      "mean_token_accuracy": 0.9890074372291565,
+      "num_tokens": 93438747.0,
+      "step": 11510
+    },
+    {
+      "entropy": 0.150148668885231,
+      "epoch": 6.498306997742664,
+      "grad_norm": 1.4895845651626587,
+      "learning_rate": 4.105251616898592e-06,
+      "loss": 0.0376,
+      "mean_token_accuracy": 0.9892801523208619,
+      "num_tokens": 93479380.0,
+      "step": 11515
+    },
+    {
+      "entropy": 0.14549075663089753,
+      "epoch": 6.501128668171558,
+      "grad_norm": 1.9964500665664673,
+      "learning_rate": 4.104543087579652e-06,
+      "loss": 0.0315,
+      "mean_token_accuracy": 0.9905564427375794,
+      "num_tokens": 93520126.0,
+      "step": 11520
+    },
+    {
+      "entropy": 0.1546286165714264,
+      "epoch": 6.503950338600451,
+      "grad_norm": 2.1431045532226562,
+      "learning_rate": 4.103834365829346e-06,
+      "loss": 0.0387,
+      "mean_token_accuracy": 0.9886301279067993,
+      "num_tokens": 93560843.0,
+      "step": 11525
+    },
+    {
+      "entropy": 0.14803966879844666,
+      "epoch": 6.506772009029345,
+      "grad_norm": 1.7916501760482788,
+      "learning_rate": 4.1031254517869e-06,
+      "loss": 0.0302,
+      "mean_token_accuracy": 0.9909810662269593,
+      "num_tokens": 93601247.0,
+      "step": 11530
+    },
+    {
+      "entropy": 0.14564967155456543,
+      "epoch": 6.50959367945824,
+      "grad_norm": 2.3298871517181396,
+      "learning_rate": 4.102416345591583e-06,
+      "loss": 0.0387,
+      "mean_token_accuracy": 0.9888354897499084,
+      "num_tokens": 93641820.0,
+      "step": 11535
+    },
+    {
+      "entropy": 0.1511426329612732,
+      "epoch": 6.512415349887133,
+      "grad_norm": 1.7131050825119019,
+      "learning_rate": 4.101707047382697e-06,
+      "loss": 0.0292,
+      "mean_token_accuracy": 0.9912442445755005,
+      "num_tokens": 93682410.0,
+      "step": 11540
+    },
+    {
+      "entropy": 0.14388006627559663,
+      "epoch": 6.515237020316027,
+      "grad_norm": 1.829458475112915,
+      "learning_rate": 4.100997557299585e-06,
+      "loss": 0.0371,
+      "mean_token_accuracy": 0.9888779282569885,
+      "num_tokens": 93723161.0,
+      "step": 11545
+    },
+    {
+      "entropy": 0.15411291718482972,
+      "epoch": 6.518058690744921,
+      "grad_norm": 2.0121374130249023,
+      "learning_rate": 4.100287875481627e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.9880946636199951,
+      "num_tokens": 93763892.0,
+      "step": 11550
+    },
+    {
+      "entropy": 0.16307685375213624,
+      "epoch": 6.520880361173814,
+      "grad_norm": 2.0615103244781494,
+      "learning_rate": 4.099578002068238e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.9895309686660767,
+      "num_tokens": 93804514.0,
+      "step": 11555
+    },
+    {
+      "entropy": 0.14214831590652466,
+      "epoch": 6.523702031602709,
+      "grad_norm": 1.7495334148406982,
+      "learning_rate": 4.098867937198873e-06,
+      "loss": 0.0356,
+      "mean_token_accuracy": 0.989754056930542,
+      "num_tokens": 93844886.0,
+      "step": 11560
+    },
+    {
+      "entropy": 0.13692098259925842,
+      "epoch": 6.526523702031603,
+      "grad_norm": 2.2061078548431396,
+      "learning_rate": 4.098157681013027e-06,
+      "loss": 0.0405,
+      "mean_token_accuracy": 0.9879782795906067,
+      "num_tokens": 93885343.0,
+      "step": 11565
+    },
+    {
+      "entropy": 0.1372235894203186,
+      "epoch": 6.529345372460496,
+      "grad_norm": 1.7683453559875488,
+      "learning_rate": 4.097447233650226e-06,
+      "loss": 0.0356,
+      "mean_token_accuracy": 0.9889456748962402,
+      "num_tokens": 93926121.0,
+      "step": 11570
+    },
+    {
+      "entropy": 0.15331811606884002,
+      "epoch": 6.532167042889391,
+      "grad_norm": 2.0020864009857178,
+      "learning_rate": 4.0967365952500416e-06,
+      "loss": 0.0332,
+      "mean_token_accuracy": 0.9897879481315612,
+      "num_tokens": 93966780.0,
+      "step": 11575
+    },
+    {
+      "entropy": 0.154784095287323,
+      "epoch": 6.534988713318285,
+      "grad_norm": 1.7012702226638794,
+      "learning_rate": 4.096025765952076e-06,
+      "loss": 0.0342,
+      "mean_token_accuracy": 0.9907065868377686,
+      "num_tokens": 94007365.0,
+      "step": 11580
+    },
+    {
+      "entropy": 0.13629684150218963,
+      "epoch": 6.537810383747178,
+      "grad_norm": 2.087913990020752,
+      "learning_rate": 4.095314745895972e-06,
+      "loss": 0.0376,
+      "mean_token_accuracy": 0.9887621521949768,
+      "num_tokens": 94048019.0,
+      "step": 11585
+    },
+    {
+      "entropy": 0.14386317431926726,
+      "epoch": 6.540632054176072,
+      "grad_norm": 1.9246793985366821,
+      "learning_rate": 4.0946035352214106e-06,
+      "loss": 0.0308,
+      "mean_token_accuracy": 0.990863585472107,
+      "num_tokens": 94088462.0,
+      "step": 11590
+    },
+    {
+      "entropy": 0.1515270948410034,
+      "epoch": 6.543453724604966,
+      "grad_norm": 2.080392837524414,
+      "learning_rate": 4.093892134068108e-06,
+      "loss": 0.0403,
+      "mean_token_accuracy": 0.9882361173629761,
+      "num_tokens": 94128923.0,
+      "step": 11595
+    },
+    {
+      "entropy": 0.15317191779613495,
+      "epoch": 6.54627539503386,
+      "grad_norm": 1.7559170722961426,
+      "learning_rate": 4.09318054257582e-06,
+      "loss": 0.036,
+      "mean_token_accuracy": 0.9891046166419983,
+      "num_tokens": 94169548.0,
+      "step": 11600
+    },
+    {
+      "entropy": 0.1485581949353218,
+      "epoch": 6.549097065462754,
+      "grad_norm": 1.9086334705352783,
+      "learning_rate": 4.092468760884338e-06,
+      "loss": 0.0403,
+      "mean_token_accuracy": 0.9878690481185913,
+      "num_tokens": 94210344.0,
+      "step": 11605
+    },
+    {
+      "entropy": 0.14423695504665374,
+      "epoch": 6.551918735891648,
+      "grad_norm": 2.0206217765808105,
+      "learning_rate": 4.0917567891334935e-06,
+      "loss": 0.0367,
+      "mean_token_accuracy": 0.9891125440597535,
+      "num_tokens": 94250622.0,
+      "step": 11610
+    },
+    {
+      "entropy": 0.14233740866184236,
+      "epoch": 6.5547404063205414,
+      "grad_norm": 2.0237369537353516,
+      "learning_rate": 4.091044627463151e-06,
+      "loss": 0.033,
+      "mean_token_accuracy": 0.9899827241897583,
+      "num_tokens": 94291129.0,
+      "step": 11615
+    },
+    {
+      "entropy": 0.1390757292509079,
+      "epoch": 6.557562076749436,
+      "grad_norm": 2.0195369720458984,
+      "learning_rate": 4.0903322760132165e-06,
+      "loss": 0.0393,
+      "mean_token_accuracy": 0.9883224606513977,
+      "num_tokens": 94331922.0,
+      "step": 11620
+    },
+    {
+      "entropy": 0.15587489008903505,
+      "epoch": 6.56038374717833,
+      "grad_norm": 2.020552158355713,
+      "learning_rate": 4.0896197349236306e-06,
+      "loss": 0.0397,
+      "mean_token_accuracy": 0.9889486074447632,
+      "num_tokens": 94372528.0,
+      "step": 11625
+    },
+    {
+      "entropy": 0.15144974291324614,
+      "epoch": 6.563205417607223,
+      "grad_norm": 1.6519806385040283,
+      "learning_rate": 4.0889070043343725e-06,
+      "loss": 0.0409,
+      "mean_token_accuracy": 0.9875991225242615,
+      "num_tokens": 94413314.0,
+      "step": 11630
+    },
+    {
+      "entropy": 0.16432562172412873,
+      "epoch": 6.566027088036117,
+      "grad_norm": 2.0881333351135254,
+      "learning_rate": 4.088194084385459e-06,
+      "loss": 0.0379,
+      "mean_token_accuracy": 0.9884780287742615,
+      "num_tokens": 94453628.0,
+      "step": 11635
+    },
+    {
+      "entropy": 0.15171082615852355,
+      "epoch": 6.568848758465011,
+      "grad_norm": 1.8627437353134155,
+      "learning_rate": 4.08748097521694e-06,
+      "loss": 0.0423,
+      "mean_token_accuracy": 0.9872801780700684,
+      "num_tokens": 94494421.0,
+      "step": 11640
+    },
+    {
+      "entropy": 0.1384830117225647,
+      "epoch": 6.571670428893905,
+      "grad_norm": 1.6030412912368774,
+      "learning_rate": 4.0867676769689104e-06,
+      "loss": 0.0408,
+      "mean_token_accuracy": 0.988001823425293,
+      "num_tokens": 94534981.0,
+      "step": 11645
+    },
+    {
+      "entropy": 0.15062596201896666,
+      "epoch": 6.574492099322799,
+      "grad_norm": 2.3316404819488525,
+      "learning_rate": 4.086054189781495e-06,
+      "loss": 0.0375,
+      "mean_token_accuracy": 0.9890991806983948,
+      "num_tokens": 94575864.0,
+      "step": 11650
+    },
+    {
+      "entropy": 0.15638395845890046,
+      "epoch": 6.577313769751693,
+      "grad_norm": 2.025421142578125,
+      "learning_rate": 4.085340513794859e-06,
+      "loss": 0.032,
+      "mean_token_accuracy": 0.9899617671966553,
+      "num_tokens": 94616541.0,
+      "step": 11655
+    },
+    {
+      "entropy": 0.1384786307811737,
+      "epoch": 6.580135440180587,
+      "grad_norm": 2.037663459777832,
+      "learning_rate": 4.084626649149204e-06,
+      "loss": 0.0357,
+      "mean_token_accuracy": 0.9894083857536315,
+      "num_tokens": 94657102.0,
+      "step": 11660
+    },
+    {
+      "entropy": 0.14564592242240906,
+      "epoch": 6.582957110609481,
+      "grad_norm": 2.0661332607269287,
+      "learning_rate": 4.083912595984769e-06,
+      "loss": 0.0367,
+      "mean_token_accuracy": 0.9883635878562927,
+      "num_tokens": 94697865.0,
+      "step": 11665
+    },
+    {
+      "entropy": 0.15481620132923127,
+      "epoch": 6.585778781038375,
+      "grad_norm": 1.89729642868042,
+      "learning_rate": 4.083198354441831e-06,
+      "loss": 0.0391,
+      "mean_token_accuracy": 0.9885015726089478,
+      "num_tokens": 94738274.0,
+      "step": 11670
+    },
+    {
+      "entropy": 0.14785862267017363,
+      "epoch": 6.5886004514672685,
+      "grad_norm": 2.5949220657348633,
+      "learning_rate": 4.082483924660701e-06,
+      "loss": 0.0326,
+      "mean_token_accuracy": 0.9899281978607177,
+      "num_tokens": 94778892.0,
+      "step": 11675
+    },
+    {
+      "entropy": 0.1447735384106636,
+      "epoch": 6.591422121896162,
+      "grad_norm": 1.6759907007217407,
+      "learning_rate": 4.081769306781729e-06,
+      "loss": 0.0374,
+      "mean_token_accuracy": 0.9890196442604064,
+      "num_tokens": 94819675.0,
+      "step": 11680
+    },
+    {
+      "entropy": 0.14461695849895478,
+      "epoch": 6.594243792325057,
+      "grad_norm": 1.7010953426361084,
+      "learning_rate": 4.081054500945303e-06,
+      "loss": 0.0343,
+      "mean_token_accuracy": 0.9894821643829346,
+      "num_tokens": 94860190.0,
+      "step": 11685
+    },
+    {
+      "entropy": 0.14982332289218903,
+      "epoch": 6.5970654627539504,
+      "grad_norm": 2.2242343425750732,
+      "learning_rate": 4.080339507291845e-06,
+      "loss": 0.0377,
+      "mean_token_accuracy": 0.9886794209480285,
+      "num_tokens": 94900923.0,
+      "step": 11690
+    },
+    {
+      "entropy": 0.15086464136838912,
+      "epoch": 6.599887133182844,
+      "grad_norm": 2.197476863861084,
+      "learning_rate": 4.079624325961818e-06,
+      "loss": 0.0359,
+      "mean_token_accuracy": 0.988996946811676,
+      "num_tokens": 94941485.0,
+      "step": 11695
+    },
+    {
+      "entropy": 0.1545707941055298,
+      "epoch": 6.602708803611738,
+      "grad_norm": 1.8683733940124512,
+      "learning_rate": 4.0789089570957175e-06,
+      "loss": 0.0394,
+      "mean_token_accuracy": 0.9879536390304565,
+      "num_tokens": 94982052.0,
+      "step": 11700
+    },
+    {
+      "entropy": 0.13810085207223893,
+      "epoch": 6.605530474040632,
+      "grad_norm": 2.827000856399536,
+      "learning_rate": 4.078193400834078e-06,
+      "loss": 0.0388,
+      "mean_token_accuracy": 0.9887730002403259,
+      "num_tokens": 95022767.0,
+      "step": 11705
+    },
+    {
+      "entropy": 0.14364120215177537,
+      "epoch": 6.608352144469526,
+      "grad_norm": 2.075237989425659,
+      "learning_rate": 4.077477657317471e-06,
+      "loss": 0.0323,
+      "mean_token_accuracy": 0.9897770881652832,
+      "num_tokens": 95063645.0,
+      "step": 11710
+    },
+    {
+      "entropy": 0.1465618520975113,
+      "epoch": 6.61117381489842,
+      "grad_norm": 2.806278705596924,
+      "learning_rate": 4.076761726686505e-06,
+      "loss": 0.0466,
+      "mean_token_accuracy": 0.9863425970077515,
+      "num_tokens": 95104053.0,
+      "step": 11715
+    },
+    {
+      "entropy": 0.14402187317609788,
+      "epoch": 6.6139954853273135,
+      "grad_norm": 1.9875422716140747,
+      "learning_rate": 4.076045609081824e-06,
+      "loss": 0.0403,
+      "mean_token_accuracy": 0.9877901315689087,
+      "num_tokens": 95144803.0,
+      "step": 11720
+    },
+    {
+      "entropy": 0.154840087890625,
+      "epoch": 6.616817155756207,
+      "grad_norm": 1.8063068389892578,
+      "learning_rate": 4.075329304644109e-06,
+      "loss": 0.0397,
+      "mean_token_accuracy": 0.9879529476165771,
+      "num_tokens": 95185612.0,
+      "step": 11725
+    },
+    {
+      "entropy": 0.13759687691926956,
+      "epoch": 6.619638826185102,
+      "grad_norm": 1.7499263286590576,
+      "learning_rate": 4.074612813514079e-06,
+      "loss": 0.0373,
+      "mean_token_accuracy": 0.988704776763916,
+      "num_tokens": 95226271.0,
+      "step": 11730
+    },
+    {
+      "entropy": 0.1382174924015999,
+      "epoch": 6.6224604966139955,
+      "grad_norm": 2.57856822013855,
+      "learning_rate": 4.073896135832488e-06,
+      "loss": 0.0423,
+      "mean_token_accuracy": 0.987503731250763,
+      "num_tokens": 95266935.0,
+      "step": 11735
+    },
+    {
+      "entropy": 0.15132977664470673,
+      "epoch": 6.625282167042889,
+      "grad_norm": 2.365727424621582,
+      "learning_rate": 4.073179271740128e-06,
+      "loss": 0.0439,
+      "mean_token_accuracy": 0.9868868947029114,
+      "num_tokens": 95307554.0,
+      "step": 11740
+    },
+    {
+      "entropy": 0.1502486765384674,
+      "epoch": 6.628103837471784,
+      "grad_norm": 2.003241777420044,
+      "learning_rate": 4.072462221377827e-06,
+      "loss": 0.0392,
+      "mean_token_accuracy": 0.9876524209976196,
+      "num_tokens": 95348372.0,
+      "step": 11745
+    },
+    {
+      "entropy": 0.15410052090883256,
+      "epoch": 6.6309255079006775,
+      "grad_norm": 1.5356718301773071,
+      "learning_rate": 4.07174498488645e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.989504873752594,
+      "num_tokens": 95389076.0,
+      "step": 11750
+    },
+    {
+      "entropy": 0.1485922798514366,
+      "epoch": 6.633747178329571,
+      "grad_norm": 1.6456876993179321,
+      "learning_rate": 4.071027562406896e-06,
+      "loss": 0.0336,
+      "mean_token_accuracy": 0.9904733896255493,
+      "num_tokens": 95429753.0,
+      "step": 11755
+    },
+    {
+      "entropy": 0.1574637770652771,
+      "epoch": 6.636568848758465,
+      "grad_norm": 2.0578815937042236,
+      "learning_rate": 4.070309954080106e-06,
+      "loss": 0.0393,
+      "mean_token_accuracy": 0.9888066649436951,
+      "num_tokens": 95470465.0,
+      "step": 11760
+    },
+    {
+      "entropy": 0.1463371992111206,
+      "epoch": 6.639390519187359,
+      "grad_norm": 1.576945185661316,
+      "learning_rate": 4.069592160047051e-06,
+      "loss": 0.0363,
+      "mean_token_accuracy": 0.9890714764595032,
+      "num_tokens": 95510938.0,
+      "step": 11765
+    },
+    {
+      "entropy": 0.15714893043041228,
+      "epoch": 6.642212189616253,
+      "grad_norm": 1.784827709197998,
+      "learning_rate": 4.0688741804487446e-06,
+      "loss": 0.0374,
+      "mean_token_accuracy": 0.9894795179367065,
+      "num_tokens": 95551359.0,
+      "step": 11770
+    },
+    {
+      "entropy": 0.14306671023368836,
+      "epoch": 6.645033860045147,
+      "grad_norm": 1.9207254648208618,
+      "learning_rate": 4.0681560154262326e-06,
+      "loss": 0.0399,
+      "mean_token_accuracy": 0.9883310794830322,
+      "num_tokens": 95591909.0,
+      "step": 11775
+    },
+    {
+      "entropy": 0.14147693812847137,
+      "epoch": 6.6478555304740405,
+      "grad_norm": 2.073054552078247,
+      "learning_rate": 4.067437665120598e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9901525735855102,
+      "num_tokens": 95632630.0,
+      "step": 11780
+    },
+    {
+      "entropy": 0.12939851135015487,
+      "epoch": 6.650677200902934,
+      "grad_norm": 1.6735975742340088,
+      "learning_rate": 4.066719129672962e-06,
+      "loss": 0.0346,
+      "mean_token_accuracy": 0.9895193934440613,
+      "num_tokens": 95673068.0,
+      "step": 11785
+    },
+    {
+      "entropy": 0.13276245892047883,
+      "epoch": 6.653498871331829,
+      "grad_norm": 2.009474754333496,
+      "learning_rate": 4.066000409224481e-06,
+      "loss": 0.0377,
+      "mean_token_accuracy": 0.9889036655426026,
+      "num_tokens": 95713693.0,
+      "step": 11790
+    },
+    {
+      "entropy": 0.13992193043231965,
+      "epoch": 6.6563205417607225,
+      "grad_norm": 2.217224359512329,
+      "learning_rate": 4.0652815039163475e-06,
+      "loss": 0.0363,
+      "mean_token_accuracy": 0.9892218232154846,
+      "num_tokens": 95754580.0,
+      "step": 11795
+    },
+    {
+      "entropy": 0.14898284673690795,
+      "epoch": 6.659142212189616,
+      "grad_norm": 1.6819255352020264,
+      "learning_rate": 4.06456241388979e-06,
+      "loss": 0.0351,
+      "mean_token_accuracy": 0.9898673295974731,
+      "num_tokens": 95795256.0,
+      "step": 11800
+    },
+    {
+      "entropy": 0.14159813523292542,
+      "epoch": 6.66196388261851,
+      "grad_norm": 1.9627163410186768,
+      "learning_rate": 4.063843139286073e-06,
+      "loss": 0.0384,
+      "mean_token_accuracy": 0.9878716945648194,
+      "num_tokens": 95835802.0,
+      "step": 11805
+    },
+    {
+      "entropy": 0.1385001763701439,
+      "epoch": 6.664785553047404,
+      "grad_norm": 2.2941231727600098,
+      "learning_rate": 4.063123680246501e-06,
+      "loss": 0.0414,
+      "mean_token_accuracy": 0.9872376441955566,
+      "num_tokens": 95876651.0,
+      "step": 11810
+    },
+    {
+      "entropy": 0.14992179572582245,
+      "epoch": 6.667607223476298,
+      "grad_norm": 1.722368836402893,
+      "learning_rate": 4.062404036912409e-06,
+      "loss": 0.0365,
+      "mean_token_accuracy": 0.9891926527023316,
+      "num_tokens": 95917227.0,
+      "step": 11815
+    },
+    {
+      "entropy": 0.14835633486509323,
+      "epoch": 6.670428893905192,
+      "grad_norm": 2.5152008533477783,
+      "learning_rate": 4.061684209425173e-06,
+      "loss": 0.0407,
+      "mean_token_accuracy": 0.9874385356903076,
+      "num_tokens": 95957630.0,
+      "step": 11820
+    },
+    {
+      "entropy": 0.1346977487206459,
+      "epoch": 6.673250564334086,
+      "grad_norm": 2.1163482666015625,
+      "learning_rate": 4.060964197926201e-06,
+      "loss": 0.0327,
+      "mean_token_accuracy": 0.9909256815910339,
+      "num_tokens": 95998284.0,
+      "step": 11825
+    },
+    {
+      "entropy": 0.14587730467319487,
+      "epoch": 6.67607223476298,
+      "grad_norm": 2.0600922107696533,
+      "learning_rate": 4.060244002556942e-06,
+      "loss": 0.0354,
+      "mean_token_accuracy": 0.9892189502716064,
+      "num_tokens": 96039045.0,
+      "step": 11830
+    },
+    {
+      "entropy": 0.157058185338974,
+      "epoch": 6.678893905191874,
+      "grad_norm": 2.0479791164398193,
+      "learning_rate": 4.0595236234588755e-06,
+      "loss": 0.0423,
+      "mean_token_accuracy": 0.9872640252113343,
+      "num_tokens": 96079680.0,
+      "step": 11835
+    },
+    {
+      "entropy": 0.15050775408744813,
+      "epoch": 6.681715575620768,
+      "grad_norm": 2.05046010017395,
+      "learning_rate": 4.058803060773523e-06,
+      "loss": 0.041,
+      "mean_token_accuracy": 0.9878550887107849,
+      "num_tokens": 96120155.0,
+      "step": 11840
+    },
+    {
+      "entropy": 0.15183096528053283,
+      "epoch": 6.684537246049661,
+      "grad_norm": 2.1310324668884277,
+      "learning_rate": 4.058082314642438e-06,
+      "loss": 0.0417,
+      "mean_token_accuracy": 0.9870881080627442,
+      "num_tokens": 96160925.0,
+      "step": 11845
+    },
+    {
+      "entropy": 0.13959557116031646,
+      "epoch": 6.687358916478555,
+      "grad_norm": 2.0781123638153076,
+      "learning_rate": 4.05736138520721e-06,
+      "loss": 0.0383,
+      "mean_token_accuracy": 0.9885563850402832,
+      "num_tokens": 96201452.0,
+      "step": 11850
+    },
+    {
+      "entropy": 0.16393175423145295,
+      "epoch": 6.6901805869074495,
+      "grad_norm": 1.6570488214492798,
+      "learning_rate": 4.056640272609467e-06,
+      "loss": 0.0327,
+      "mean_token_accuracy": 0.9903802037239074,
+      "num_tokens": 96242166.0,
+      "step": 11855
+    },
+    {
+      "entropy": 0.1479017674922943,
+      "epoch": 6.693002257336343,
+      "grad_norm": 1.9548949003219604,
+      "learning_rate": 4.055918976990872e-06,
+      "loss": 0.0321,
+      "mean_token_accuracy": 0.9904338598251343,
+      "num_tokens": 96282855.0,
+      "step": 11860
+    },
+    {
+      "entropy": 0.1482335865497589,
+      "epoch": 6.695823927765237,
+      "grad_norm": 2.1175918579101562,
+      "learning_rate": 4.055197498493123e-06,
+      "loss": 0.0418,
+      "mean_token_accuracy": 0.9874824643135071,
+      "num_tokens": 96323775.0,
+      "step": 11865
+    },
+    {
+      "entropy": 0.14961565434932708,
+      "epoch": 6.698645598194131,
+      "grad_norm": 1.942327618598938,
+      "learning_rate": 4.054475837257953e-06,
+      "loss": 0.032,
+      "mean_token_accuracy": 0.9901739597320557,
+      "num_tokens": 96364548.0,
+      "step": 11870
+    },
+    {
+      "entropy": 0.15366184413433076,
+      "epoch": 6.701467268623025,
+      "grad_norm": 2.2145068645477295,
+      "learning_rate": 4.053753993427135e-06,
+      "loss": 0.0378,
+      "mean_token_accuracy": 0.988074266910553,
+      "num_tokens": 96404933.0,
+      "step": 11875
+    },
+    {
+      "entropy": 0.15021361410617828,
+      "epoch": 6.704288939051919,
+      "grad_norm": 1.6474920511245728,
+      "learning_rate": 4.053031967142475e-06,
+      "loss": 0.0341,
+      "mean_token_accuracy": 0.98985835313797,
+      "num_tokens": 96445402.0,
+      "step": 11880
+    },
+    {
+      "entropy": 0.16894918978214263,
+      "epoch": 6.707110609480813,
+      "grad_norm": 1.8737285137176514,
+      "learning_rate": 4.052309758545813e-06,
+      "loss": 0.0475,
+      "mean_token_accuracy": 0.9867665410041809,
+      "num_tokens": 96486119.0,
+      "step": 11885
+    },
+    {
+      "entropy": 0.15583183765411376,
+      "epoch": 6.709932279909706,
+      "grad_norm": 1.9295543432235718,
+      "learning_rate": 4.051587367779029e-06,
+      "loss": 0.0321,
+      "mean_token_accuracy": 0.9901478886604309,
+      "num_tokens": 96526926.0,
+      "step": 11890
+    },
+    {
+      "entropy": 0.15508732497692107,
+      "epoch": 6.7127539503386,
+      "grad_norm": 1.60502290725708,
+      "learning_rate": 4.050864794984036e-06,
+      "loss": 0.0316,
+      "mean_token_accuracy": 0.9912713050842286,
+      "num_tokens": 96567737.0,
+      "step": 11895
+    },
+    {
+      "entropy": 0.15691338181495668,
+      "epoch": 6.715575620767495,
+      "grad_norm": 1.7124838829040527,
+      "learning_rate": 4.050142040302784e-06,
+      "loss": 0.0368,
+      "mean_token_accuracy": 0.9893094897270203,
+      "num_tokens": 96607281.0,
+      "step": 11900
+    },
+    {
+      "entropy": 0.1330145627260208,
+      "epoch": 6.718397291196388,
+      "grad_norm": 1.8225212097167969,
+      "learning_rate": 4.049419103877258e-06,
+      "loss": 0.035,
+      "mean_token_accuracy": 0.9894989252090454,
+      "num_tokens": 96648016.0,
+      "step": 11905
+    },
+    {
+      "entropy": 0.15559065639972686,
+      "epoch": 6.721218961625282,
+      "grad_norm": 1.6894086599349976,
+      "learning_rate": 4.048695985849479e-06,
+      "loss": 0.036,
+      "mean_token_accuracy": 0.9887082815170288,
+      "num_tokens": 96688807.0,
+      "step": 11910
+    },
+    {
+      "entropy": 0.14794491827487946,
+      "epoch": 6.724040632054177,
+      "grad_norm": 1.7598350048065186,
+      "learning_rate": 4.047972686361503e-06,
+      "loss": 0.0399,
+      "mean_token_accuracy": 0.9888983845710755,
+      "num_tokens": 96729457.0,
+      "step": 11915
+    },
+    {
+      "entropy": 0.1369953766465187,
+      "epoch": 6.72686230248307,
+      "grad_norm": 1.9057563543319702,
+      "learning_rate": 4.047249205555423e-06,
+      "loss": 0.0321,
+      "mean_token_accuracy": 0.9902025699615479,
+      "num_tokens": 96769286.0,
+      "step": 11920
+    },
+    {
+      "entropy": 0.14871995151042938,
+      "epoch": 6.729683972911964,
+      "grad_norm": 2.072307586669922,
+      "learning_rate": 4.046525543573366e-06,
+      "loss": 0.0397,
+      "mean_token_accuracy": 0.9880092859268188,
+      "num_tokens": 96809834.0,
+      "step": 11925
+    },
+    {
+      "entropy": 0.13897217512130738,
+      "epoch": 6.732505643340858,
+      "grad_norm": 2.005223274230957,
+      "learning_rate": 4.045801700557497e-06,
+      "loss": 0.0378,
+      "mean_token_accuracy": 0.9891113877296448,
+      "num_tokens": 96850609.0,
+      "step": 11930
+    },
+    {
+      "entropy": 0.1497331142425537,
+      "epoch": 6.735327313769751,
+      "grad_norm": 2.2140676975250244,
+      "learning_rate": 4.045077676650014e-06,
+      "loss": 0.0385,
+      "mean_token_accuracy": 0.9883991718292237,
+      "num_tokens": 96891222.0,
+      "step": 11935
+    },
+    {
+      "entropy": 0.14488416612148286,
+      "epoch": 6.738148984198646,
+      "grad_norm": 1.802141547203064,
+      "learning_rate": 4.044353471993152e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.9881270170211792,
+      "num_tokens": 96931928.0,
+      "step": 11940
+    },
+    {
+      "entropy": 0.14439502954483033,
+      "epoch": 6.74097065462754,
+      "grad_norm": 2.055842399597168,
+      "learning_rate": 4.0436290867291806e-06,
+      "loss": 0.0436,
+      "mean_token_accuracy": 0.986508822441101,
+      "num_tokens": 96972565.0,
+      "step": 11945
+    },
+    {
+      "entropy": 0.14606373012065887,
+      "epoch": 6.743792325056433,
+      "grad_norm": 1.8258408308029175,
+      "learning_rate": 4.042904521000406e-06,
+      "loss": 0.0348,
+      "mean_token_accuracy": 0.9901708602905274,
+      "num_tokens": 97013177.0,
+      "step": 11950
+    },
+    {
+      "entropy": 0.15913362801074982,
+      "epoch": 6.746613995485327,
+      "grad_norm": 2.4801416397094727,
+      "learning_rate": 4.042179774949169e-06,
+      "loss": 0.0495,
+      "mean_token_accuracy": 0.9845641493797302,
+      "num_tokens": 97053721.0,
+      "step": 11955
+    },
+    {
+      "entropy": 0.1454220324754715,
+      "epoch": 6.749435665914222,
+      "grad_norm": 2.5556042194366455,
+      "learning_rate": 4.041454848717845e-06,
+      "loss": 0.0492,
+      "mean_token_accuracy": 0.9852368235588074,
+      "num_tokens": 97094386.0,
+      "step": 11960
+    },
+    {
+      "entropy": 0.14474825710058212,
+      "epoch": 6.752257336343115,
+      "grad_norm": 1.8604875802993774,
+      "learning_rate": 4.040729742448848e-06,
+      "loss": 0.0389,
+      "mean_token_accuracy": 0.9878225922584534,
+      "num_tokens": 97134920.0,
+      "step": 11965
+    },
+    {
+      "entropy": 0.14345545172691346,
+      "epoch": 6.755079006772009,
+      "grad_norm": 2.0168981552124023,
+      "learning_rate": 4.040004456284623e-06,
+      "loss": 0.0399,
+      "mean_token_accuracy": 0.9885003328323364,
+      "num_tokens": 97175551.0,
+      "step": 11970
+    },
+    {
+      "entropy": 0.14142475724220277,
+      "epoch": 6.757900677200903,
+      "grad_norm": 1.5360438823699951,
+      "learning_rate": 4.0392789903676545e-06,
+      "loss": 0.0319,
+      "mean_token_accuracy": 0.9909959673881531,
+      "num_tokens": 97216255.0,
+      "step": 11975
+    },
+    {
+      "entropy": 0.14621337950229646,
+      "epoch": 6.760722347629796,
+      "grad_norm": 1.5952950716018677,
+      "learning_rate": 4.03855334484046e-06,
+      "loss": 0.0362,
+      "mean_token_accuracy": 0.9889961838722229,
+      "num_tokens": 97257015.0,
+      "step": 11980
+    },
+    {
+      "entropy": 0.14792815148830413,
+      "epoch": 6.763544018058691,
+      "grad_norm": 2.06760835647583,
+      "learning_rate": 4.037827519845591e-06,
+      "loss": 0.0299,
+      "mean_token_accuracy": 0.9917079448699951,
+      "num_tokens": 97297751.0,
+      "step": 11985
+    },
+    {
+      "entropy": 0.1481049656867981,
+      "epoch": 6.766365688487585,
+      "grad_norm": 1.856473684310913,
+      "learning_rate": 4.037101515525637e-06,
+      "loss": 0.0368,
+      "mean_token_accuracy": 0.9893781185150147,
+      "num_tokens": 97338310.0,
+      "step": 11990
+    },
+    {
+      "entropy": 0.14933787584304808,
+      "epoch": 6.769187358916478,
+      "grad_norm": 1.7966928482055664,
+      "learning_rate": 4.036375332023222e-06,
+      "loss": 0.0408,
+      "mean_token_accuracy": 0.9873141407966614,
+      "num_tokens": 97378801.0,
+      "step": 11995
+    },
+    {
+      "entropy": 0.15227662324905394,
+      "epoch": 6.772009029345372,
+      "grad_norm": 2.43129301071167,
+      "learning_rate": 4.0356489694810055e-06,
+      "loss": 0.0365,
+      "mean_token_accuracy": 0.9891417384147644,
+      "num_tokens": 97419558.0,
+      "step": 12000
+    },
+    {
+      "epoch": 6.772009029345372,
+      "eval_entropy": 0.1956140100955963,
+      "eval_loss": 0.012606015428900719,
+      "eval_mean_token_accuracy": 0.9965503811836243,
+      "eval_num_tokens": 97419558.0,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 24.418,
+      "eval_steps_per_second": 6.105,
+      "step": 12000
+    },
+    {
+      "entropy": 0.15393645465373992,
+      "epoch": 6.774830699774267,
+      "grad_norm": 2.1532039642333984,
+      "learning_rate": 4.03492242804168e-06,
+      "loss": 0.0442,
+      "mean_token_accuracy": 0.9862104415893554,
+      "num_tokens": 97460095.0,
+      "step": 12005
+    },
+    {
+      "entropy": 0.1350156396627426,
+      "epoch": 6.77765237020316,
+      "grad_norm": 1.9720470905303955,
+      "learning_rate": 4.034195707847975e-06,
+      "loss": 0.0332,
+      "mean_token_accuracy": 0.9898574113845825,
+      "num_tokens": 97500756.0,
+      "step": 12010
+    },
+    {
+      "entropy": 0.1415847659111023,
+      "epoch": 6.780474040632054,
+      "grad_norm": 1.9140161275863647,
+      "learning_rate": 4.033468809042655e-06,
+      "loss": 0.0365,
+      "mean_token_accuracy": 0.9890129208564759,
+      "num_tokens": 97541530.0,
+      "step": 12015
+    },
+    {
+      "entropy": 0.14546948671340942,
+      "epoch": 6.783295711060948,
+      "grad_norm": 2.1538074016571045,
+      "learning_rate": 4.032741731768519e-06,
+      "loss": 0.0406,
+      "mean_token_accuracy": 0.986963152885437,
+      "num_tokens": 97581567.0,
+      "step": 12020
+    },
+    {
+      "entropy": 0.14426875561475755,
+      "epoch": 6.786117381489842,
+      "grad_norm": 1.9157716035842896,
+      "learning_rate": 4.032014476168403e-06,
+      "loss": 0.0355,
+      "mean_token_accuracy": 0.9900845766067505,
+      "num_tokens": 97622196.0,
+      "step": 12025
+    },
+    {
+      "entropy": 0.14931005537509917,
+      "epoch": 6.788939051918736,
+      "grad_norm": 1.9808646440505981,
+      "learning_rate": 4.031287042385174e-06,
+      "loss": 0.0398,
+      "mean_token_accuracy": 0.9882163524627685,
+      "num_tokens": 97662914.0,
+      "step": 12030
+    },
+    {
+      "entropy": 0.1658826380968094,
+      "epoch": 6.79176072234763,
+      "grad_norm": 2.1118521690368652,
+      "learning_rate": 4.030559430561738e-06,
+      "loss": 0.0371,
+      "mean_token_accuracy": 0.9887429594993591,
+      "num_tokens": 97703480.0,
+      "step": 12035
+    },
+    {
+      "entropy": 0.15551794469356536,
+      "epoch": 6.794582392776523,
+      "grad_norm": 2.5292036533355713,
+      "learning_rate": 4.029831640841035e-06,
+      "loss": 0.0463,
+      "mean_token_accuracy": 0.9871610522270202,
+      "num_tokens": 97744240.0,
+      "step": 12040
+    },
+    {
+      "entropy": 0.15738882422447203,
+      "epoch": 6.797404063205418,
+      "grad_norm": 1.7646291255950928,
+      "learning_rate": 4.029103673366037e-06,
+      "loss": 0.0386,
+      "mean_token_accuracy": 0.988528847694397,
+      "num_tokens": 97784307.0,
+      "step": 12045
+    },
+    {
+      "entropy": 0.1447429984807968,
+      "epoch": 6.800225733634312,
+      "grad_norm": 2.040116310119629,
+      "learning_rate": 4.028375528279757e-06,
+      "loss": 0.0399,
+      "mean_token_accuracy": 0.9880939960479737,
+      "num_tokens": 97825063.0,
+      "step": 12050
+    },
+    {
+      "entropy": 0.15236329436302185,
+      "epoch": 6.803047404063205,
+      "grad_norm": 1.7679275274276733,
+      "learning_rate": 4.027647205725235e-06,
+      "loss": 0.0347,
+      "mean_token_accuracy": 0.9887324929237366,
+      "num_tokens": 97865669.0,
+      "step": 12055
+    },
+    {
+      "entropy": 0.15214796960353852,
+      "epoch": 6.805869074492099,
+      "grad_norm": 1.926100730895996,
+      "learning_rate": 4.026918705845553e-06,
+      "loss": 0.0309,
+      "mean_token_accuracy": 0.9904511570930481,
+      "num_tokens": 97906513.0,
+      "step": 12060
+    },
+    {
+      "entropy": 0.14448803663253784,
+      "epoch": 6.808690744920993,
+      "grad_norm": 1.732896327972412,
+      "learning_rate": 4.026190028783824e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.9893592596054077,
+      "num_tokens": 97946867.0,
+      "step": 12065
+    },
+    {
+      "entropy": 0.15263217091560363,
+      "epoch": 6.811512415349887,
+      "grad_norm": 1.5818462371826172,
+      "learning_rate": 4.025461174683195e-06,
+      "loss": 0.038,
+      "mean_token_accuracy": 0.9896114826202392,
+      "num_tokens": 97987436.0,
+      "step": 12070
+    },
+    {
+      "entropy": 0.14199225902557372,
+      "epoch": 6.814334085778781,
+      "grad_norm": 2.4520139694213867,
+      "learning_rate": 4.024732143686854e-06,
+      "loss": 0.0343,
+      "mean_token_accuracy": 0.9893141746520996,
+      "num_tokens": 98028061.0,
+      "step": 12075
+    },
+    {
+      "entropy": 0.1499510258436203,
+      "epoch": 6.817155756207675,
+      "grad_norm": 2.507272720336914,
+      "learning_rate": 4.024002935938015e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9894481897354126,
+      "num_tokens": 98068816.0,
+      "step": 12080
+    },
+    {
+      "entropy": 0.1536535143852234,
+      "epoch": 6.8199774266365685,
+      "grad_norm": 1.787207007408142,
+      "learning_rate": 4.0232735515799325e-06,
+      "loss": 0.04,
+      "mean_token_accuracy": 0.9880955934524536,
+      "num_tokens": 98109350.0,
+      "step": 12085
+    },
+    {
+      "entropy": 0.15066614747047424,
+      "epoch": 6.822799097065463,
+      "grad_norm": 2.0756945610046387,
+      "learning_rate": 4.022543990755894e-06,
+      "loss": 0.0397,
+      "mean_token_accuracy": 0.9888906359672547,
+      "num_tokens": 98149953.0,
+      "step": 12090
+    },
+    {
+      "entropy": 0.1614619642496109,
+      "epoch": 6.825620767494357,
+      "grad_norm": 2.169306516647339,
+      "learning_rate": 4.021814253609222e-06,
+      "loss": 0.0418,
+      "mean_token_accuracy": 0.98787602186203,
+      "num_tokens": 98190613.0,
+      "step": 12095
+    },
+    {
+      "entropy": 0.13885380923748017,
+      "epoch": 6.8284424379232505,
+      "grad_norm": 1.65521240234375,
+      "learning_rate": 4.021084340283273e-06,
+      "loss": 0.0377,
+      "mean_token_accuracy": 0.9888322472572326,
+      "num_tokens": 98231228.0,
+      "step": 12100
+    },
+    {
+      "entropy": 0.13987670242786407,
+      "epoch": 6.831264108352144,
+      "grad_norm": 2.0256147384643555,
+      "learning_rate": 4.020354250921439e-06,
+      "loss": 0.0374,
+      "mean_token_accuracy": 0.9885939717292785,
+      "num_tokens": 98271713.0,
+      "step": 12105
+    },
+    {
+      "entropy": 0.14241887032985687,
+      "epoch": 6.834085778781039,
+      "grad_norm": 2.178802728652954,
+      "learning_rate": 4.0196239856671465e-06,
+      "loss": 0.0343,
+      "mean_token_accuracy": 0.9898516416549683,
+      "num_tokens": 98312133.0,
+      "step": 12110
+    },
+    {
+      "entropy": 0.15942923724651337,
+      "epoch": 6.836907449209932,
+      "grad_norm": 1.7774022817611694,
+      "learning_rate": 4.0188935446638545e-06,
+      "loss": 0.0417,
+      "mean_token_accuracy": 0.9870691418647766,
+      "num_tokens": 98352906.0,
+      "step": 12115
+    },
+    {
+      "entropy": 0.13171774595975877,
+      "epoch": 6.839729119638826,
+      "grad_norm": 1.836889624595642,
+      "learning_rate": 4.018162928055061e-06,
+      "loss": 0.04,
+      "mean_token_accuracy": 0.9872918844223022,
+      "num_tokens": 98393669.0,
+      "step": 12120
+    },
+    {
+      "entropy": 0.1482748955488205,
+      "epoch": 6.84255079006772,
+      "grad_norm": 1.7053948640823364,
+      "learning_rate": 4.017432135984293e-06,
+      "loss": 0.034,
+      "mean_token_accuracy": 0.990520977973938,
+      "num_tokens": 98434346.0,
+      "step": 12125
+    },
+    {
+      "entropy": 0.15172266364097595,
+      "epoch": 6.845372460496614,
+      "grad_norm": 1.739862084388733,
+      "learning_rate": 4.016701168595116e-06,
+      "loss": 0.0434,
+      "mean_token_accuracy": 0.9869125485420227,
+      "num_tokens": 98474870.0,
+      "step": 12130
+    },
+    {
+      "entropy": 0.15046610832214355,
+      "epoch": 6.848194130925508,
+      "grad_norm": 2.3250515460968018,
+      "learning_rate": 4.01597002603113e-06,
+      "loss": 0.0452,
+      "mean_token_accuracy": 0.9872011065483093,
+      "num_tokens": 98515557.0,
+      "step": 12135
+    },
+    {
+      "entropy": 0.14062797725200654,
+      "epoch": 6.851015801354402,
+      "grad_norm": 2.180325746536255,
+      "learning_rate": 4.015238708435965e-06,
+      "loss": 0.0368,
+      "mean_token_accuracy": 0.9888609290122986,
+      "num_tokens": 98556251.0,
+      "step": 12140
+    },
+    {
+      "entropy": 0.1476264402270317,
+      "epoch": 6.8538374717832955,
+      "grad_norm": 2.2180349826812744,
+      "learning_rate": 4.0145072159532906e-06,
+      "loss": 0.0376,
+      "mean_token_accuracy": 0.9886261582374573,
+      "num_tokens": 98597018.0,
+      "step": 12145
+    },
+    {
+      "entropy": 0.1369374930858612,
+      "epoch": 6.856659142212189,
+      "grad_norm": 1.9766490459442139,
+      "learning_rate": 4.013775548726807e-06,
+      "loss": 0.0375,
+      "mean_token_accuracy": 0.9877067446708679,
+      "num_tokens": 98637642.0,
+      "step": 12150
+    },
+    {
+      "entropy": 0.15410472452640533,
+      "epoch": 6.859480812641084,
+      "grad_norm": 2.415297269821167,
+      "learning_rate": 4.013043706900252e-06,
+      "loss": 0.0404,
+      "mean_token_accuracy": 0.9889116406440734,
+      "num_tokens": 98678395.0,
+      "step": 12155
+    },
+    {
+      "entropy": 0.16218827664852142,
+      "epoch": 6.8623024830699775,
+      "grad_norm": 1.7888035774230957,
+      "learning_rate": 4.012311690617396e-06,
+      "loss": 0.0408,
+      "mean_token_accuracy": 0.9877578258514405,
+      "num_tokens": 98718623.0,
+      "step": 12160
+    },
+    {
+      "entropy": 0.1510297805070877,
+      "epoch": 6.865124153498871,
+      "grad_norm": 2.012989044189453,
+      "learning_rate": 4.011579500022043e-06,
+      "loss": 0.0388,
+      "mean_token_accuracy": 0.9888627529144287,
+      "num_tokens": 98759483.0,
+      "step": 12165
+    },
+    {
+      "entropy": 0.15312339663505553,
+      "epoch": 6.867945823927765,
+      "grad_norm": 1.867661714553833,
+      "learning_rate": 4.010847135258031e-06,
+      "loss": 0.0349,
+      "mean_token_accuracy": 0.9899600625038147,
+      "num_tokens": 98799890.0,
+      "step": 12170
+    },
+    {
+      "entropy": 0.14693421721458436,
+      "epoch": 6.8707674943566595,
+      "grad_norm": 1.5122771263122559,
+      "learning_rate": 4.010114596469234e-06,
+      "loss": 0.0333,
+      "mean_token_accuracy": 0.9900424122810364,
+      "num_tokens": 98840808.0,
+      "step": 12175
+    },
+    {
+      "entropy": 0.17182584404945372,
+      "epoch": 6.873589164785553,
+      "grad_norm": 1.7709141969680786,
+      "learning_rate": 4.009381883799561e-06,
+      "loss": 0.0355,
+      "mean_token_accuracy": 0.989521062374115,
+      "num_tokens": 98881544.0,
+      "step": 12180
+    },
+    {
+      "entropy": 0.15575533509254455,
+      "epoch": 6.876410835214447,
+      "grad_norm": 2.0675461292266846,
+      "learning_rate": 4.00864899739295e-06,
+      "loss": 0.0394,
+      "mean_token_accuracy": 0.9877262592315674,
+      "num_tokens": 98922210.0,
+      "step": 12185
+    },
+    {
+      "entropy": 0.15429140627384186,
+      "epoch": 6.8792325056433405,
+      "grad_norm": 2.4544715881347656,
+      "learning_rate": 4.00791593739338e-06,
+      "loss": 0.0494,
+      "mean_token_accuracy": 0.9845258593559265,
+      "num_tokens": 98962824.0,
+      "step": 12190
+    },
+    {
+      "entropy": 0.15189413875341415,
+      "epoch": 6.882054176072235,
+      "grad_norm": 2.8407533168792725,
+      "learning_rate": 4.007182703944859e-06,
+      "loss": 0.0446,
+      "mean_token_accuracy": 0.9873796224594116,
+      "num_tokens": 99003619.0,
+      "step": 12195
+    },
+    {
+      "entropy": 0.15428635776042937,
+      "epoch": 6.884875846501129,
+      "grad_norm": 2.224959135055542,
+      "learning_rate": 4.006449297191432e-06,
+      "loss": 0.0365,
+      "mean_token_accuracy": 0.9887138962745666,
+      "num_tokens": 99044216.0,
+      "step": 12200
+    },
+    {
+      "entropy": 0.14300165176391602,
+      "epoch": 6.8876975169300225,
+      "grad_norm": 1.9736629724502563,
+      "learning_rate": 4.005715717277174e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.988729465007782,
+      "num_tokens": 99084439.0,
+      "step": 12205
+    },
+    {
+      "entropy": 0.13568673729896547,
+      "epoch": 6.890519187358916,
+      "grad_norm": 1.9142423868179321,
+      "learning_rate": 4.004981964346201e-06,
+      "loss": 0.0332,
+      "mean_token_accuracy": 0.990368640422821,
+      "num_tokens": 99124909.0,
+      "step": 12210
+    },
+    {
+      "entropy": 0.14406311213970185,
+      "epoch": 6.893340857787811,
+      "grad_norm": 2.4211018085479736,
+      "learning_rate": 4.004248038542656e-06,
+      "loss": 0.039,
+      "mean_token_accuracy": 0.988343071937561,
+      "num_tokens": 99165645.0,
+      "step": 12215
+    },
+    {
+      "entropy": 0.15503831803798676,
+      "epoch": 6.8961625282167045,
+      "grad_norm": 2.471714496612549,
+      "learning_rate": 4.003513940010718e-06,
+      "loss": 0.0432,
+      "mean_token_accuracy": 0.9861461281776428,
+      "num_tokens": 99206347.0,
+      "step": 12220
+    },
+    {
+      "entropy": 0.1480024516582489,
+      "epoch": 6.898984198645598,
+      "grad_norm": 2.0117077827453613,
+      "learning_rate": 4.002779668894604e-06,
+      "loss": 0.0362,
+      "mean_token_accuracy": 0.9891183733940124,
+      "num_tokens": 99246987.0,
+      "step": 12225
+    },
+    {
+      "entropy": 0.16451097428798675,
+      "epoch": 6.901805869074492,
+      "grad_norm": 1.6899313926696777,
+      "learning_rate": 4.002045225338559e-06,
+      "loss": 0.0404,
+      "mean_token_accuracy": 0.9885852932929993,
+      "num_tokens": 99287449.0,
+      "step": 12230
+    },
+    {
+      "entropy": 0.14375028908252716,
+      "epoch": 6.904627539503386,
+      "grad_norm": 1.818851351737976,
+      "learning_rate": 4.001310609486866e-06,
+      "loss": 0.0362,
+      "mean_token_accuracy": 0.9893089056015014,
+      "num_tokens": 99327979.0,
+      "step": 12235
+    },
+    {
+      "entropy": 0.15548242926597594,
+      "epoch": 6.90744920993228,
+      "grad_norm": 1.614169955253601,
+      "learning_rate": 4.000575821483839e-06,
+      "loss": 0.0383,
+      "mean_token_accuracy": 0.9892736315727234,
+      "num_tokens": 99368473.0,
+      "step": 12240
+    },
+    {
+      "entropy": 0.16284603476524354,
+      "epoch": 6.910270880361174,
+      "grad_norm": 1.6673612594604492,
+      "learning_rate": 3.9998408614738295e-06,
+      "loss": 0.039,
+      "mean_token_accuracy": 0.9885139226913452,
+      "num_tokens": 99408923.0,
+      "step": 12245
+    },
+    {
+      "entropy": 0.14805781245231628,
+      "epoch": 6.913092550790068,
+      "grad_norm": 2.2017159461975098,
+      "learning_rate": 3.999105729601218e-06,
+      "loss": 0.0378,
+      "mean_token_accuracy": 0.9887848377227784,
+      "num_tokens": 99449422.0,
+      "step": 12250
+    },
+    {
+      "entropy": 0.14871875047683716,
+      "epoch": 6.915914221218961,
+      "grad_norm": 1.9085617065429688,
+      "learning_rate": 3.9983704260104225e-06,
+      "loss": 0.0395,
+      "mean_token_accuracy": 0.9885958671569824,
+      "num_tokens": 99490171.0,
+      "step": 12255
+    },
+    {
+      "entropy": 0.1603003293275833,
+      "epoch": 6.918735891647856,
+      "grad_norm": 1.8411649465560913,
+      "learning_rate": 3.997634950845893e-06,
+      "loss": 0.0391,
+      "mean_token_accuracy": 0.9881568551063538,
+      "num_tokens": 99530789.0,
+      "step": 12260
+    },
+    {
+      "entropy": 0.14625448882579803,
+      "epoch": 6.9215575620767495,
+      "grad_norm": 1.8774826526641846,
+      "learning_rate": 3.996899304252116e-06,
+      "loss": 0.0342,
+      "mean_token_accuracy": 0.9903077602386474,
+      "num_tokens": 99571506.0,
+      "step": 12265
+    },
+    {
+      "entropy": 0.14785387217998505,
+      "epoch": 6.924379232505643,
+      "grad_norm": 1.910869836807251,
+      "learning_rate": 3.996163486373605e-06,
+      "loss": 0.0408,
+      "mean_token_accuracy": 0.9881960272789001,
+      "num_tokens": 99612260.0,
+      "step": 12270
+    },
+    {
+      "entropy": 0.15447575151920317,
+      "epoch": 6.927200902934537,
+      "grad_norm": 1.985054612159729,
+      "learning_rate": 3.9954274973549144e-06,
+      "loss": 0.0399,
+      "mean_token_accuracy": 0.9876292586326599,
+      "num_tokens": 99652726.0,
+      "step": 12275
+    },
+    {
+      "entropy": 0.15240433514118196,
+      "epoch": 6.9300225733634315,
+      "grad_norm": 1.6934657096862793,
+      "learning_rate": 3.994691337340629e-06,
+      "loss": 0.0353,
+      "mean_token_accuracy": 0.9899464964866638,
+      "num_tokens": 99693195.0,
+      "step": 12280
+    },
+    {
+      "entropy": 0.14032234251499176,
+      "epoch": 6.932844243792325,
+      "grad_norm": 2.0978596210479736,
+      "learning_rate": 3.9939550064753674e-06,
+      "loss": 0.0421,
+      "mean_token_accuracy": 0.9877179741859436,
+      "num_tokens": 99733715.0,
+      "step": 12285
+    },
+    {
+      "entropy": 0.1609266698360443,
+      "epoch": 6.935665914221219,
+      "grad_norm": 2.033520221710205,
+      "learning_rate": 3.993218504903781e-06,
+      "loss": 0.0396,
+      "mean_token_accuracy": 0.9879401564598084,
+      "num_tokens": 99774295.0,
+      "step": 12290
+    },
+    {
+      "entropy": 0.1533256560564041,
+      "epoch": 6.938487584650113,
+      "grad_norm": 2.056273937225342,
+      "learning_rate": 3.992481832770558e-06,
+      "loss": 0.0481,
+      "mean_token_accuracy": 0.9857136845588684,
+      "num_tokens": 99814847.0,
+      "step": 12295
+    },
+    {
+      "entropy": 0.14914492666721343,
+      "epoch": 6.941309255079007,
+      "grad_norm": 1.809752345085144,
+      "learning_rate": 3.991744990220415e-06,
+      "loss": 0.0403,
+      "mean_token_accuracy": 0.9875779390335083,
+      "num_tokens": 99855659.0,
+      "step": 12300
+    },
+    {
+      "entropy": 0.13904739618301393,
+      "epoch": 6.944130925507901,
+      "grad_norm": 1.739696979522705,
+      "learning_rate": 3.9910079773981055e-06,
+      "loss": 0.0357,
+      "mean_token_accuracy": 0.989600396156311,
+      "num_tokens": 99896329.0,
+      "step": 12305
+    },
+    {
+      "entropy": 0.1490867018699646,
+      "epoch": 6.946952595936795,
+      "grad_norm": 2.174006223678589,
+      "learning_rate": 3.990270794448418e-06,
+      "loss": 0.0397,
+      "mean_token_accuracy": 0.9880833983421325,
+      "num_tokens": 99936915.0,
+      "step": 12310
+    },
+    {
+      "entropy": 0.1486115574836731,
+      "epoch": 6.949774266365688,
+      "grad_norm": 2.1487553119659424,
+      "learning_rate": 3.989533441516169e-06,
+      "loss": 0.0412,
+      "mean_token_accuracy": 0.988270103931427,
+      "num_tokens": 99977400.0,
+      "step": 12315
+    },
+    {
+      "entropy": 0.15158516466617583,
+      "epoch": 6.952595936794582,
+      "grad_norm": 2.021366834640503,
+      "learning_rate": 3.9887959187462145e-06,
+      "loss": 0.0385,
+      "mean_token_accuracy": 0.988451075553894,
+      "num_tokens": 100018070.0,
+      "step": 12320
+    },
+    {
+      "entropy": 0.15477460920810698,
+      "epoch": 6.955417607223477,
+      "grad_norm": 2.3939874172210693,
+      "learning_rate": 3.988058226283438e-06,
+      "loss": 0.0459,
+      "mean_token_accuracy": 0.9860603213310242,
+      "num_tokens": 100058592.0,
+      "step": 12325
+    },
+    {
+      "entropy": 0.13858523964881897,
+      "epoch": 6.95823927765237,
+      "grad_norm": 1.722223162651062,
+      "learning_rate": 3.987320364272761e-06,
+      "loss": 0.0366,
+      "mean_token_accuracy": 0.9893313646316528,
+      "num_tokens": 100098381.0,
+      "step": 12330
+    },
+    {
+      "entropy": 0.14853320121765137,
+      "epoch": 6.961060948081264,
+      "grad_norm": 1.6259387731552124,
+      "learning_rate": 3.986582332859138e-06,
+      "loss": 0.029,
+      "mean_token_accuracy": 0.9913646578788757,
+      "num_tokens": 100138509.0,
+      "step": 12335
+    },
+    {
+      "entropy": 0.1412474274635315,
+      "epoch": 6.963882618510158,
+      "grad_norm": 1.655497431755066,
+      "learning_rate": 3.985844132187552e-06,
+      "loss": 0.0377,
+      "mean_token_accuracy": 0.9879396677017211,
+      "num_tokens": 100178940.0,
+      "step": 12340
+    },
+    {
+      "entropy": 0.14502420425415039,
+      "epoch": 6.966704288939052,
+      "grad_norm": 2.259775161743164,
+      "learning_rate": 3.985105762403024e-06,
+      "loss": 0.0385,
+      "mean_token_accuracy": 0.9885640144348145,
+      "num_tokens": 100219679.0,
+      "step": 12345
+    },
+    {
+      "entropy": 0.1412261575460434,
+      "epoch": 6.969525959367946,
+      "grad_norm": 2.183638095855713,
+      "learning_rate": 3.984367223650608e-06,
+      "loss": 0.0391,
+      "mean_token_accuracy": 0.988431978225708,
+      "num_tokens": 100260265.0,
+      "step": 12350
+    },
+    {
+      "entropy": 0.16001304388046264,
+      "epoch": 6.97234762979684,
+      "grad_norm": 1.7587566375732422,
+      "learning_rate": 3.983628516075389e-06,
+      "loss": 0.0422,
+      "mean_token_accuracy": 0.9870441198348999,
+      "num_tokens": 100301028.0,
+      "step": 12355
+    },
+    {
+      "entropy": 0.1356129452586174,
+      "epoch": 6.975169300225733,
+      "grad_norm": 1.582041621208191,
+      "learning_rate": 3.982889639822487e-06,
+      "loss": 0.0351,
+      "mean_token_accuracy": 0.9892949461936951,
+      "num_tokens": 100341780.0,
+      "step": 12360
+    },
+    {
+      "entropy": 0.15091146826744078,
+      "epoch": 6.977990970654628,
+      "grad_norm": 1.9201163053512573,
+      "learning_rate": 3.982150595037053e-06,
+      "loss": 0.0396,
+      "mean_token_accuracy": 0.9874477505683898,
+      "num_tokens": 100382504.0,
+      "step": 12365
+    },
+    {
+      "entropy": 0.14935719966888428,
+      "epoch": 6.980812641083522,
+      "grad_norm": 1.8997581005096436,
+      "learning_rate": 3.981411381864274e-06,
+      "loss": 0.0357,
+      "mean_token_accuracy": 0.9886354565620422,
+      "num_tokens": 100423339.0,
+      "step": 12370
+    },
+    {
+      "entropy": 0.15034773349761962,
+      "epoch": 6.983634311512415,
+      "grad_norm": 2.1917574405670166,
+      "learning_rate": 3.980672000449367e-06,
+      "loss": 0.0354,
+      "mean_token_accuracy": 0.9892937660217285,
+      "num_tokens": 100463746.0,
+      "step": 12375
+    },
+    {
+      "entropy": 0.15186431109905243,
+      "epoch": 6.986455981941309,
+      "grad_norm": 1.7991350889205933,
+      "learning_rate": 3.9799324509375846e-06,
+      "loss": 0.0449,
+      "mean_token_accuracy": 0.9864747166633606,
+      "num_tokens": 100504500.0,
+      "step": 12380
+    },
+    {
+      "entropy": 0.162436842918396,
+      "epoch": 6.989277652370204,
+      "grad_norm": 1.9422773122787476,
+      "learning_rate": 3.979192733474211e-06,
+      "loss": 0.0436,
+      "mean_token_accuracy": 0.9862201929092407,
+      "num_tokens": 100545119.0,
+      "step": 12385
+    },
+    {
+      "entropy": 0.15107800364494323,
+      "epoch": 6.992099322799097,
+      "grad_norm": 2.049860715866089,
+      "learning_rate": 3.978452848204563e-06,
+      "loss": 0.0403,
+      "mean_token_accuracy": 0.9873886942863465,
+      "num_tokens": 100585667.0,
+      "step": 12390
+    },
+    {
+      "entropy": 0.1658845365047455,
+      "epoch": 6.994920993227991,
+      "grad_norm": 2.350536346435547,
+      "learning_rate": 3.977712795273993e-06,
+      "loss": 0.0426,
+      "mean_token_accuracy": 0.9870129108428956,
+      "num_tokens": 100626337.0,
+      "step": 12395
+    },
+    {
+      "entropy": 0.14828409254550934,
+      "epoch": 6.997742663656885,
+      "grad_norm": 2.069523811340332,
+      "learning_rate": 3.976972574827883e-06,
+      "loss": 0.0379,
+      "mean_token_accuracy": 0.9882864952087402,
+      "num_tokens": 100666972.0,
+      "step": 12400
+    },
+    {
+      "entropy": 0.15599106550216674,
+      "epoch": 7.000564334085778,
+      "grad_norm": 1.5296047925949097,
+      "learning_rate": 3.97623218701165e-06,
+      "loss": 0.0328,
+      "mean_token_accuracy": 0.9913263559341431,
+      "num_tokens": 100701561.0,
+      "step": 12405
+    },
+    {
+      "entropy": 0.13645475655794143,
+      "epoch": 7.003386004514673,
+      "grad_norm": 1.3886555433273315,
+      "learning_rate": 3.975491631970744e-06,
+      "loss": 0.0159,
+      "mean_token_accuracy": 0.9963740348815918,
+      "num_tokens": 100742162.0,
+      "step": 12410
+    },
+    {
+      "entropy": 0.13274604827165604,
+      "epoch": 7.006207674943567,
+      "grad_norm": 1.7409141063690186,
+      "learning_rate": 3.974750909850646e-06,
+      "loss": 0.0181,
+      "mean_token_accuracy": 0.9954428911209107,
+      "num_tokens": 100782515.0,
+      "step": 12415
+    },
+    {
+      "entropy": 0.12155016213655472,
+      "epoch": 7.00902934537246,
+      "grad_norm": 1.6582471132278442,
+      "learning_rate": 3.9740100207968716e-06,
+      "loss": 0.0183,
+      "mean_token_accuracy": 0.9957684516906739,
+      "num_tokens": 100823151.0,
+      "step": 12420
+    },
+    {
+      "entropy": 0.12232804298400879,
+      "epoch": 7.011851015801354,
+      "grad_norm": 1.9185771942138672,
+      "learning_rate": 3.973268964954967e-06,
+      "loss": 0.0156,
+      "mean_token_accuracy": 0.9957268714904786,
+      "num_tokens": 100863921.0,
+      "step": 12425
+    },
+    {
+      "entropy": 0.13372556120157242,
+      "epoch": 7.014672686230249,
+      "grad_norm": 1.302709937095642,
+      "learning_rate": 3.972527742470515e-06,
+      "loss": 0.017,
+      "mean_token_accuracy": 0.9956133961677551,
+      "num_tokens": 100904652.0,
+      "step": 12430
+    },
+    {
+      "entropy": 0.1276037335395813,
+      "epoch": 7.017494356659142,
+      "grad_norm": 2.0763673782348633,
+      "learning_rate": 3.971786353489127e-06,
+      "loss": 0.0174,
+      "mean_token_accuracy": 0.9954843640327453,
+      "num_tokens": 100945350.0,
+      "step": 12435
+    },
+    {
+      "entropy": 0.1303536906838417,
+      "epoch": 7.020316027088036,
+      "grad_norm": 1.7125003337860107,
+      "learning_rate": 3.97104479815645e-06,
+      "loss": 0.0193,
+      "mean_token_accuracy": 0.9947200775146484,
+      "num_tokens": 100986018.0,
+      "step": 12440
+    },
+    {
+      "entropy": 0.13805813938379288,
+      "epoch": 7.02313769751693,
+      "grad_norm": 1.2138659954071045,
+      "learning_rate": 3.9703030766181634e-06,
+      "loss": 0.0166,
+      "mean_token_accuracy": 0.9959219574928284,
+      "num_tokens": 101026679.0,
+      "step": 12445
+    },
+    {
+      "entropy": 0.11290259510278702,
+      "epoch": 7.025959367945824,
+      "grad_norm": 1.9468364715576172,
+      "learning_rate": 3.969561189019977e-06,
+      "loss": 0.0156,
+      "mean_token_accuracy": 0.9957970261573792,
+      "num_tokens": 101067288.0,
+      "step": 12450
+    },
+    {
+      "entropy": 0.12177327871322632,
+      "epoch": 7.028781038374718,
+      "grad_norm": 1.656708836555481,
+      "learning_rate": 3.968819135507636e-06,
+      "loss": 0.0212,
+      "mean_token_accuracy": 0.994585645198822,
+      "num_tokens": 101107815.0,
+      "step": 12455
+    },
+    {
+      "entropy": 0.13150275498628616,
+      "epoch": 7.031602708803612,
+      "grad_norm": 1.2296918630599976,
+      "learning_rate": 3.968076916226914e-06,
+      "loss": 0.0143,
+      "mean_token_accuracy": 0.9961319208145142,
+      "num_tokens": 101148543.0,
+      "step": 12460
+    },
+    {
+      "entropy": 0.13319746106863023,
+      "epoch": 7.034424379232505,
+      "grad_norm": 1.9725842475891113,
+      "learning_rate": 3.967334531323624e-06,
+      "loss": 0.0166,
+      "mean_token_accuracy": 0.995919120311737,
+      "num_tokens": 101189223.0,
+      "step": 12465
+    },
+    {
+      "entropy": 0.13151731789112092,
+      "epoch": 7.0372460496614,
+      "grad_norm": 1.4561344385147095,
+      "learning_rate": 3.966591980943605e-06,
+      "loss": 0.0189,
+      "mean_token_accuracy": 0.995412039756775,
+      "num_tokens": 101229757.0,
+      "step": 12470
+    },
+    {
+      "entropy": 0.12682377845048903,
+      "epoch": 7.040067720090294,
+      "grad_norm": 1.5752171277999878,
+      "learning_rate": 3.965849265232732e-06,
+      "loss": 0.0162,
+      "mean_token_accuracy": 0.996173620223999,
+      "num_tokens": 101270029.0,
+      "step": 12475
+    },
+    {
+      "entropy": 0.1453809142112732,
+      "epoch": 7.042889390519187,
+      "grad_norm": 1.8831138610839844,
+      "learning_rate": 3.965106384336912e-06,
+      "loss": 0.0178,
+      "mean_token_accuracy": 0.995201587677002,
+      "num_tokens": 101310756.0,
+      "step": 12480
+    },
+    {
+      "entropy": 0.13473598957061766,
+      "epoch": 7.045711060948081,
+      "grad_norm": 1.9781758785247803,
+      "learning_rate": 3.964363338402083e-06,
+      "loss": 0.0181,
+      "mean_token_accuracy": 0.9956030488014221,
+      "num_tokens": 101350870.0,
+      "step": 12485
+    },
+    {
+      "entropy": 0.13227078914642335,
+      "epoch": 7.048532731376975,
+      "grad_norm": 1.273861289024353,
+      "learning_rate": 3.9636201275742175e-06,
+      "loss": 0.0178,
+      "mean_token_accuracy": 0.9954041123390198,
+      "num_tokens": 101391393.0,
+      "step": 12490
+    },
+    {
+      "entropy": 0.130999419093132,
+      "epoch": 7.051354401805869,
+      "grad_norm": 1.901928186416626,
+      "learning_rate": 3.962876751999318e-06,
+      "loss": 0.0211,
+      "mean_token_accuracy": 0.9943174481391907,
+      "num_tokens": 101431784.0,
+      "step": 12495
+    },
+    {
+      "entropy": 0.13049811720848084,
+      "epoch": 7.054176072234763,
+      "grad_norm": 1.6979538202285767,
+      "learning_rate": 3.962133211823424e-06,
+      "loss": 0.0206,
+      "mean_token_accuracy": 0.9947118163108826,
+      "num_tokens": 101472388.0,
+      "step": 12500
+    },
+    {
+      "epoch": 7.054176072234763,
+      "eval_entropy": 0.1905188113451004,
+      "eval_loss": 0.012303678318858147,
+      "eval_mean_token_accuracy": 0.9973169565200806,
+      "eval_num_tokens": 101472388.0,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 24.454,
+      "eval_steps_per_second": 6.113,
+      "step": 12500
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 35440,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 20,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.212206797834539e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}