diff --git "a/training_log.json" "b/training_log.json"
new file mode 100644--- /dev/null
+++ "b/training_log.json"
@@ -0,0 +1,3884 @@
+[
+  {
+    "loss": 1.8227,
+    "grad_norm": 1.0382107496261597,
+    "learning_rate": 9e-06,
+    "entropy": 1.3477498710155487,
+    "num_tokens": 212422.0,
+    "mean_token_accuracy": 0.6026485413312912,
+    "epoch": 0.0025823111684958036,
+    "step": 10
+  },
+  {
+    "loss": 1.7607,
+    "grad_norm": 0.5150073766708374,
+    "learning_rate": 1.9e-05,
+    "entropy": 1.4502712726593017,
+    "num_tokens": 439120.0,
+    "mean_token_accuracy": 0.6050827592611313,
+    "epoch": 0.005164622336991607,
+    "step": 20
+  },
+  {
+    "loss": 1.6637,
+    "grad_norm": 0.49609556794166565,
+    "learning_rate": 2.9e-05,
+    "entropy": 1.6171577751636506,
+    "num_tokens": 659373.0,
+    "mean_token_accuracy": 0.6135089665651321,
+    "epoch": 0.007746933505487412,
+    "step": 30
+  },
+  {
+    "loss": 1.5172,
+    "grad_norm": 0.5204024314880371,
+    "learning_rate": 3.9000000000000006e-05,
+    "entropy": 1.5392299711704254,
+    "num_tokens": 892258.0,
+    "mean_token_accuracy": 0.6360862851142883,
+    "epoch": 0.010329244673983214,
+    "step": 40
+  },
+  {
+    "loss": 1.2945,
+    "grad_norm": 0.5096025466918945,
+    "learning_rate": 4.9e-05,
+    "entropy": 1.2994968891143799,
+    "num_tokens": 1105690.0,
+    "mean_token_accuracy": 0.6796375066041946,
+    "epoch": 0.012911555842479019,
+    "step": 50
+  },
+  {
+    "loss": 1.0857,
+    "grad_norm": 0.46880456805229187,
+    "learning_rate": 5.9e-05,
+    "entropy": 1.0753329753875733,
+    "num_tokens": 1330615.0,
+    "mean_token_accuracy": 0.7262010276317596,
+    "epoch": 0.015493867010974823,
+    "step": 60
+  },
+  {
+    "loss": 1.0636,
+    "grad_norm": 0.34878015518188477,
+    "learning_rate": 6.9e-05,
+    "entropy": 1.0663362741470337,
+    "num_tokens": 1550884.0,
+    "mean_token_accuracy": 0.7337213784456253,
+    "epoch": 0.018076178179470628,
+    "step": 70
+  },
+  {
+    "loss": 0.9654,
+    "grad_norm": 0.3773444592952728,
+    "learning_rate": 7.900000000000001e-05,
+    "entropy": 0.968677031993866,
+    "num_tokens": 1771016.0,
+    "mean_token_accuracy": 0.7530804932117462,
+    "epoch": 0.02065848934796643,
+    "step": 80
+  },
+  {
+    "loss": 0.9163,
+    "grad_norm": 0.35969284176826477,
+    "learning_rate": 8.900000000000001e-05,
+    "entropy": 0.9248217672109604,
+    "num_tokens": 1989636.0,
+    "mean_token_accuracy": 0.7628412812948226,
+    "epoch": 0.023240800516462233,
+    "step": 90
+  },
+  {
+    "loss": 0.9032,
+    "grad_norm": 0.30777958035469055,
+    "learning_rate": 9.900000000000001e-05,
+    "entropy": 0.9049190044403076,
+    "num_tokens": 2216762.0,
+    "mean_token_accuracy": 0.7661656141281128,
+    "epoch": 0.025823111684958037,
+    "step": 100
+  },
+  {
+    "loss": 0.8708,
+    "grad_norm": 0.3418641686439514,
+    "learning_rate": 9.999859605811758e-05,
+    "entropy": 0.8667158752679824,
+    "num_tokens": 2449222.0,
+    "mean_token_accuracy": 0.7731464982032776,
+    "epoch": 0.028405422853453842,
+    "step": 110
+  },
+  {
+    "loss": 0.876,
+    "grad_norm": 0.3700331449508667,
+    "learning_rate": 9.999374302690578e-05,
+    "entropy": 0.881954425573349,
+    "num_tokens": 2679757.0,
+    "mean_token_accuracy": 0.771819531917572,
+    "epoch": 0.030987734021949646,
+    "step": 120
+  },
+  {
+    "loss": 0.8714,
+    "grad_norm": 0.3392849266529083,
+    "learning_rate": 9.998542391013121e-05,
+    "entropy": 0.881954836845398,
+    "num_tokens": 2899369.0,
+    "mean_token_accuracy": 0.7725747585296631,
+    "epoch": 0.03357004519044545,
+    "step": 130
+  },
+  {
+    "loss": 0.8221,
+    "grad_norm": 0.313702791929245,
+    "learning_rate": 9.997363928456144e-05,
+    "entropy": 0.8353541791439056,
+    "num_tokens": 3117811.0,
+    "mean_token_accuracy": 0.7830256283283233,
+    "epoch": 0.036152356358941255,
+    "step": 140
+  },
+  {
+    "loss": 0.8141,
+    "grad_norm": 0.34913378953933716,
+    "learning_rate": 9.995838996722914e-05,
+    "entropy": 0.8243900567293168,
+    "num_tokens": 3335852.0,
+    "mean_token_accuracy": 0.7871499985456467,
+    "epoch": 0.03873466752743705,
+    "step": 150
+  },
+  {
+    "loss": 0.9178,
+    "grad_norm": 0.3551943302154541,
+    "learning_rate": 9.99396770153754e-05,
+    "entropy": 0.9281795680522918,
+    "num_tokens": 3547153.0,
+    "mean_token_accuracy": 0.7648159861564636,
+    "epoch": 0.04131697869593286,
+    "step": 160
+  },
+  {
+    "loss": 0.8314,
+    "grad_norm": 0.3343612849712372,
+    "learning_rate": 9.991750172637635e-05,
+    "entropy": 0.8414248585700989,
+    "num_tokens": 3774587.0,
+    "mean_token_accuracy": 0.7814970403909683,
+    "epoch": 0.04389928986442866,
+    "step": 170
+  },
+  {
+    "loss": 0.8441,
+    "grad_norm": 0.38748034834861755,
+    "learning_rate": 9.989186563765343e-05,
+    "entropy": 0.8512898027896881,
+    "num_tokens": 3987027.0,
+    "mean_token_accuracy": 0.7771281927824021,
+    "epoch": 0.046481601032924466,
+    "step": 180
+  },
+  {
+    "loss": 0.8497,
+    "grad_norm": 0.3136347830295563,
+    "learning_rate": 9.986277052656658e-05,
+    "entropy": 0.8494892001152039,
+    "num_tokens": 4224162.0,
+    "mean_token_accuracy": 0.7764846205711364,
+    "epoch": 0.04906391220142027,
+    "step": 190
+  },
+  {
+    "loss": 0.8767,
+    "grad_norm": 0.2760690152645111,
+    "learning_rate": 9.983021841029111e-05,
+    "entropy": 0.8813158571720123,
+    "num_tokens": 4461540.0,
+    "mean_token_accuracy": 0.7718124866485596,
+    "epoch": 0.051646223369916075,
+    "step": 200
+  },
+  {
+    "loss": 0.8616,
+    "grad_norm": 0.3373132646083832,
+    "learning_rate": 9.979421154567785e-05,
+    "entropy": 0.8658105224370957,
+    "num_tokens": 4675517.0,
+    "mean_token_accuracy": 0.7753888577222824,
+    "epoch": 0.05422853453841188,
+    "step": 210
+  },
+  {
+    "loss": 0.8315,
+    "grad_norm": 0.3122693598270416,
+    "learning_rate": 9.975475242909667e-05,
+    "entropy": 0.8357755273580552,
+    "num_tokens": 4893926.0,
+    "mean_token_accuracy": 0.7797913908958435,
+    "epoch": 0.056810845706907684,
+    "step": 220
+  },
+  {
+    "loss": 0.777,
+    "grad_norm": 0.31195691227912903,
+    "learning_rate": 9.971184379626339e-05,
+    "entropy": 0.7857547640800476,
+    "num_tokens": 5119968.0,
+    "mean_token_accuracy": 0.7917549222707748,
+    "epoch": 0.05939315687540349,
+    "step": 230
+  },
+  {
+    "loss": 0.8304,
+    "grad_norm": 0.39890217781066895,
+    "learning_rate": 9.966548862205016e-05,
+    "entropy": 0.8422511845827103,
+    "num_tokens": 5342685.0,
+    "mean_token_accuracy": 0.7815836250782013,
+    "epoch": 0.06197546804389929,
+    "step": 240
+  },
+  {
+    "loss": 0.7829,
+    "grad_norm": 0.295213907957077,
+    "learning_rate": 9.961569012027915e-05,
+    "entropy": 0.7870919585227967,
+    "num_tokens": 5570339.0,
+    "mean_token_accuracy": 0.7925563097000122,
+    "epoch": 0.0645577792123951,
+    "step": 250
+  },
+  {
+    "loss": 0.8125,
+    "grad_norm": 0.3115305006504059,
+    "learning_rate": 9.956245174349976e-05,
+    "entropy": 0.8125756323337555,
+    "num_tokens": 5811219.0,
+    "mean_token_accuracy": 0.78687684237957,
+    "epoch": 0.0671400903808909,
+    "step": 260
+  },
+  {
+    "loss": 0.808,
+    "grad_norm": 0.3087800443172455,
+    "learning_rate": 9.950577718274927e-05,
+    "entropy": 0.8085541218519211,
+    "num_tokens": 6027751.0,
+    "mean_token_accuracy": 0.7848722189664841,
+    "epoch": 0.0697224015493867,
+    "step": 270
+  },
+  {
+    "loss": 0.793,
+    "grad_norm": 0.3529994487762451,
+    "learning_rate": 9.944567036729695e-05,
+    "entropy": 0.8006815254688263,
+    "num_tokens": 6244346.0,
+    "mean_token_accuracy": 0.78866326212883,
+    "epoch": 0.07230471271788251,
+    "step": 280
+  },
+  {
+    "loss": 0.8033,
+    "grad_norm": 0.33018946647644043,
+    "learning_rate": 9.938213546437154e-05,
+    "entropy": 0.8124665051698685,
+    "num_tokens": 6468953.0,
+    "mean_token_accuracy": 0.7864397615194321,
+    "epoch": 0.07488702388637831,
+    "step": 290
+  },
+  {
+    "loss": 0.8236,
+    "grad_norm": 0.3586885929107666,
+    "learning_rate": 9.93151768788725e-05,
+    "entropy": 0.822173684835434,
+    "num_tokens": 6683224.0,
+    "mean_token_accuracy": 0.783802455663681,
+    "epoch": 0.0774693350548741,
+    "step": 300
+  },
+  {
+    "loss": 0.803,
+    "grad_norm": 0.36473768949508667,
+    "learning_rate": 9.924479925306448e-05,
+    "entropy": 0.8035257875919342,
+    "num_tokens": 6899814.0,
+    "mean_token_accuracy": 0.7880063384771347,
+    "epoch": 0.08005164622336991,
+    "step": 310
+  },
+  {
+    "loss": 0.7717,
+    "grad_norm": 0.29985126852989197,
+    "learning_rate": 9.917100746625552e-05,
+    "entropy": 0.7767777234315872,
+    "num_tokens": 7127264.0,
+    "mean_token_accuracy": 0.7941146910190582,
+    "epoch": 0.08263395739186571,
+    "step": 320
+  },
+  {
+    "loss": 0.8628,
+    "grad_norm": 0.38821646571159363,
+    "learning_rate": 9.90938066344588e-05,
+    "entropy": 0.8686172008514405,
+    "num_tokens": 7346312.0,
+    "mean_token_accuracy": 0.7750474840402604,
+    "epoch": 0.08521626856036152,
+    "step": 330
+  },
+  {
+    "loss": 0.7993,
+    "grad_norm": 0.2704952657222748,
+    "learning_rate": 9.901320211003786e-05,
+    "entropy": 0.8089859664440155,
+    "num_tokens": 7568639.0,
+    "mean_token_accuracy": 0.7889086365699768,
+    "epoch": 0.08779857972885732,
+    "step": 340
+  },
+  {
+    "loss": 0.8385,
+    "grad_norm": 0.3369337022304535,
+    "learning_rate": 9.892919948133558e-05,
+    "entropy": 0.8445587635040284,
+    "num_tokens": 7782903.0,
+    "mean_token_accuracy": 0.7784519046545029,
+    "epoch": 0.09038089089735313,
+    "step": 350
+  },
+  {
+    "loss": 0.8203,
+    "grad_norm": 0.3692948818206787,
+    "learning_rate": 9.884180457228678e-05,
+    "entropy": 0.8273489683866501,
+    "num_tokens": 7992733.0,
+    "mean_token_accuracy": 0.7822743952274323,
+    "epoch": 0.09296320206584893,
+    "step": 360
+  },
+  {
+    "loss": 0.7603,
+    "grad_norm": 0.27002501487731934,
+    "learning_rate": 9.875102344201428e-05,
+    "entropy": 0.7642450213432312,
+    "num_tokens": 8229494.0,
+    "mean_token_accuracy": 0.7939036816358567,
+    "epoch": 0.09554551323434474,
+    "step": 370
+  },
+  {
+    "loss": 0.8318,
+    "grad_norm": 0.29771047830581665,
+    "learning_rate": 9.865686238440905e-05,
+    "entropy": 0.8396833568811417,
+    "num_tokens": 8451547.0,
+    "mean_token_accuracy": 0.7823069781064987,
+    "epoch": 0.09812782440284054,
+    "step": 380
+  },
+  {
+    "loss": 0.8068,
+    "grad_norm": 0.26213449239730835,
+    "learning_rate": 9.855932792769367e-05,
+    "entropy": 0.8070472866296768,
+    "num_tokens": 8659006.0,
+    "mean_token_accuracy": 0.7854753315448761,
+    "epoch": 0.10071013557133635,
+    "step": 390
+  },
+  {
+    "loss": 0.8116,
+    "grad_norm": 0.3281964361667633,
+    "learning_rate": 9.845842683396975e-05,
+    "entropy": 0.8121274918317795,
+    "num_tokens": 8880841.0,
+    "mean_token_accuracy": 0.785520127415657,
+    "epoch": 0.10329244673983215,
+    "step": 400
+  },
+  {
+    "loss": 0.7812,
+    "grad_norm": 0.28127405047416687,
+    "learning_rate": 9.835416609874923e-05,
+    "entropy": 0.782671543955803,
+    "num_tokens": 9105513.0,
+    "mean_token_accuracy": 0.7911499351263046,
+    "epoch": 0.10587475790832795,
+    "step": 410
+  },
+  {
+    "loss": 0.7683,
+    "grad_norm": 0.3277055323123932,
+    "learning_rate": 9.824655295046925e-05,
+    "entropy": 0.7686008930206298,
+    "num_tokens": 9328555.0,
+    "mean_token_accuracy": 0.7945983529090881,
+    "epoch": 0.10845706907682376,
+    "step": 420
+  },
+  {
+    "loss": 0.797,
+    "grad_norm": 0.339370459318161,
+    "learning_rate": 9.813559484999102e-05,
+    "entropy": 0.7977692514657975,
+    "num_tokens": 9548239.0,
+    "mean_token_accuracy": 0.7889350891113281,
+    "epoch": 0.11103938024531956,
+    "step": 430
+  },
+  {
+    "loss": 0.7888,
+    "grad_norm": 0.2581377923488617,
+    "learning_rate": 9.802129949008265e-05,
+    "entropy": 0.7947029381990433,
+    "num_tokens": 9773006.0,
+    "mean_token_accuracy": 0.7894155353307724,
+    "epoch": 0.11362169141381537,
+    "step": 440
+  },
+  {
+    "loss": 0.7939,
+    "grad_norm": 0.2780890464782715,
+    "learning_rate": 9.790367479488567e-05,
+    "entropy": 0.7900374501943588,
+    "num_tokens": 10007849.0,
+    "mean_token_accuracy": 0.7889153778553009,
+    "epoch": 0.11620400258231117,
+    "step": 450
+  },
+  {
+    "loss": 0.7843,
+    "grad_norm": 0.35706827044487,
+    "learning_rate": 9.778272891936578e-05,
+    "entropy": 0.7888166964054107,
+    "num_tokens": 10228340.0,
+    "mean_token_accuracy": 0.7894989937543869,
+    "epoch": 0.11878631375080698,
+    "step": 460
+  },
+  {
+    "loss": 0.7818,
+    "grad_norm": 0.28400421142578125,
+    "learning_rate": 9.765847024874732e-05,
+    "entropy": 0.788197135925293,
+    "num_tokens": 10449248.0,
+    "mean_token_accuracy": 0.7903656899929047,
+    "epoch": 0.12136862491930278,
+    "step": 470
+  },
+  {
+    "loss": 0.7724,
+    "grad_norm": 0.3045465648174286,
+    "learning_rate": 9.753090739793208e-05,
+    "entropy": 0.7795355767011642,
+    "num_tokens": 10693608.0,
+    "mean_token_accuracy": 0.792780527472496,
+    "epoch": 0.12395093608779859,
+    "step": 480
+  },
+  {
+    "loss": 0.8257,
+    "grad_norm": 0.2519194781780243,
+    "learning_rate": 9.740004921090188e-05,
+    "entropy": 0.8308536618947983,
+    "num_tokens": 10916236.0,
+    "mean_token_accuracy": 0.7820773988962173,
+    "epoch": 0.1265332472562944,
+    "step": 490
+  },
+  {
+    "loss": 0.89,
+    "grad_norm": 0.26520147919654846,
+    "learning_rate": 9.726590476010548e-05,
+    "entropy": 0.891560536623001,
+    "num_tokens": 11141566.0,
+    "mean_token_accuracy": 0.7719975858926773,
+    "epoch": 0.1291155584247902,
+    "step": 500
+  },
+  {
+    "loss": 0.7747,
+    "grad_norm": 0.3187003433704376,
+    "learning_rate": 9.712848334582965e-05,
+    "entropy": 0.7743502765893936,
+    "num_tokens": 11355026.0,
+    "mean_token_accuracy": 0.7932853370904922,
+    "epoch": 0.131697869593286,
+    "step": 510
+  },
+  {
+    "loss": 0.7687,
+    "grad_norm": 0.2777857780456543,
+    "learning_rate": 9.698779449555421e-05,
+    "entropy": 0.7742303222417831,
+    "num_tokens": 11568694.0,
+    "mean_token_accuracy": 0.7947982579469681,
+    "epoch": 0.1342801807617818,
+    "step": 520
+  },
+  {
+    "loss": 0.7613,
+    "grad_norm": 0.26509273052215576,
+    "learning_rate": 9.684384796329165e-05,
+    "entropy": 0.7646809101104737,
+    "num_tokens": 11801619.0,
+    "mean_token_accuracy": 0.7968462467193603,
+    "epoch": 0.1368624919302776,
+    "step": 530
+  },
+  {
+    "loss": 0.7922,
+    "grad_norm": 0.30422940850257874,
+    "learning_rate": 9.669665372891076e-05,
+    "entropy": 0.797106859087944,
+    "num_tokens": 12028866.0,
+    "mean_token_accuracy": 0.7882502853870392,
+    "epoch": 0.1394448030987734,
+    "step": 540
+  },
+  {
+    "loss": 0.7981,
+    "grad_norm": 0.3401212692260742,
+    "learning_rate": 9.654622199744485e-05,
+    "entropy": 0.802043342590332,
+    "num_tokens": 12231531.0,
+    "mean_token_accuracy": 0.788466426730156,
+    "epoch": 0.14202711426726922,
+    "step": 550
+  },
+  {
+    "loss": 0.7113,
+    "grad_norm": 0.2457897961139679,
+    "learning_rate": 9.639256319838408e-05,
+    "entropy": 0.713046595454216,
+    "num_tokens": 12467239.0,
+    "mean_token_accuracy": 0.8070330142974853,
+    "epoch": 0.14460942543576502,
+    "step": 560
+  },
+  {
+    "loss": 0.7966,
+    "grad_norm": 0.36826762557029724,
+    "learning_rate": 9.62356879849525e-05,
+    "entropy": 0.8055638283491134,
+    "num_tokens": 12691681.0,
+    "mean_token_accuracy": 0.789168867468834,
+    "epoch": 0.14719173660426083,
+    "step": 570
+  },
+  {
+    "loss": 0.77,
+    "grad_norm": 0.23838452994823456,
+    "learning_rate": 9.607560723336941e-05,
+    "entropy": 0.7768332690000535,
+    "num_tokens": 12916444.0,
+    "mean_token_accuracy": 0.795351967215538,
+    "epoch": 0.14977404777275663,
+    "step": 580
+  },
+  {
+    "loss": 0.7308,
+    "grad_norm": 0.2863340377807617,
+    "learning_rate": 9.591233204209532e-05,
+    "entropy": 0.7349940836429596,
+    "num_tokens": 13132831.0,
+    "mean_token_accuracy": 0.8030821114778519,
+    "epoch": 0.15235635894125243,
+    "step": 590
+  },
+  {
+    "loss": 0.7557,
+    "grad_norm": 0.2508617639541626,
+    "learning_rate": 9.574587373106241e-05,
+    "entropy": 0.7568652123212815,
+    "num_tokens": 13351405.0,
+    "mean_token_accuracy": 0.7971890032291412,
+    "epoch": 0.1549386701097482,
+    "step": 600
+  },
+  {
+    "loss": 0.7656,
+    "grad_norm": 0.2594195008277893,
+    "learning_rate": 9.557624384088988e-05,
+    "entropy": 0.7718629479408264,
+    "num_tokens": 13586644.0,
+    "mean_token_accuracy": 0.796920120716095,
+    "epoch": 0.15752098127824402,
+    "step": 610
+  },
+  {
+    "loss": 0.8103,
+    "grad_norm": 0.26011860370635986,
+    "learning_rate": 9.540345413208371e-05,
+    "entropy": 0.8115774095058441,
+    "num_tokens": 13807188.0,
+    "mean_token_accuracy": 0.7873918890953064,
+    "epoch": 0.16010329244673982,
+    "step": 620
+  },
+  {
+    "loss": 0.7465,
+    "grad_norm": 0.26401853561401367,
+    "learning_rate": 9.522751658422128e-05,
+    "entropy": 0.7476193904876709,
+    "num_tokens": 14037773.0,
+    "mean_token_accuracy": 0.7999627381563187,
+    "epoch": 0.16268560361523562,
+    "step": 630
+  },
+  {
+    "loss": 0.7885,
+    "grad_norm": 0.2792288064956665,
+    "learning_rate": 9.504844339512095e-05,
+    "entropy": 0.7934227615594864,
+    "num_tokens": 14250931.0,
+    "mean_token_accuracy": 0.7914494186639786,
+    "epoch": 0.16526791478373143,
+    "step": 640
+  },
+  {
+    "loss": 0.8174,
+    "grad_norm": 0.265168696641922,
+    "learning_rate": 9.486624697999625e-05,
+    "entropy": 0.8207276076078415,
+    "num_tokens": 14472590.0,
+    "mean_token_accuracy": 0.7836927026510239,
+    "epoch": 0.16785022595222723,
+    "step": 650
+  },
+  {
+    "loss": 0.7785,
+    "grad_norm": 0.30072885751724243,
+    "learning_rate": 9.468093997059518e-05,
+    "entropy": 0.7893655002117157,
+    "num_tokens": 14696942.0,
+    "mean_token_accuracy": 0.7929514676332474,
+    "epoch": 0.17043253712072304,
+    "step": 660
+  },
+  {
+    "loss": 0.803,
+    "grad_norm": 0.2837442457675934,
+    "learning_rate": 9.449253521432447e-05,
+    "entropy": 0.8055104643106461,
+    "num_tokens": 14917939.0,
+    "mean_token_accuracy": 0.7881365656852722,
+    "epoch": 0.17301484828921884,
+    "step": 670
+  },
+  {
+    "loss": 0.7755,
+    "grad_norm": 0.3586549460887909,
+    "learning_rate": 9.430104577335881e-05,
+    "entropy": 0.785092431306839,
+    "num_tokens": 15126055.0,
+    "mean_token_accuracy": 0.7926269322633743,
+    "epoch": 0.17559715945771465,
+    "step": 680
+  },
+  {
+    "loss": 0.7333,
+    "grad_norm": 0.2736169695854187,
+    "learning_rate": 9.41064849237353e-05,
+    "entropy": 0.7377870678901672,
+    "num_tokens": 15345502.0,
+    "mean_token_accuracy": 0.800521832704544,
+    "epoch": 0.17817947062621045,
+    "step": 690
+  },
+  {
+    "loss": 0.7181,
+    "grad_norm": 0.24931730329990387,
+    "learning_rate": 9.390886615443299e-05,
+    "entropy": 0.7208522915840149,
+    "num_tokens": 15576219.0,
+    "mean_token_accuracy": 0.8049475431442261,
+    "epoch": 0.18076178179470626,
+    "step": 700
+  },
+  {
+    "loss": 0.738,
+    "grad_norm": 0.28357696533203125,
+    "learning_rate": 9.370820316643768e-05,
+    "entropy": 0.735668683052063,
+    "num_tokens": 15799601.0,
+    "mean_token_accuracy": 0.8012147158384323,
+    "epoch": 0.18334409296320206,
+    "step": 710
+  },
+  {
+    "loss": 0.778,
+    "grad_norm": 0.2858926057815552,
+    "learning_rate": 9.350450987179202e-05,
+    "entropy": 0.781666374206543,
+    "num_tokens": 16018324.0,
+    "mean_token_accuracy": 0.7933406472206116,
+    "epoch": 0.18592640413169786,
+    "step": 720
+  },
+  {
+    "loss": 0.7659,
+    "grad_norm": 0.2707382142543793,
+    "learning_rate": 9.329780039263101e-05,
+    "entropy": 0.7684377580881119,
+    "num_tokens": 16249193.0,
+    "mean_token_accuracy": 0.7941454380750657,
+    "epoch": 0.18850871530019367,
+    "step": 730
+  },
+  {
+    "loss": 0.7866,
+    "grad_norm": 0.272834450006485,
+    "learning_rate": 9.308808906020287e-05,
+    "entropy": 0.7898376554250717,
+    "num_tokens": 16474694.0,
+    "mean_token_accuracy": 0.7909801512956619,
+    "epoch": 0.19109102646868947,
+    "step": 740
+  },
+  {
+    "loss": 0.7656,
+    "grad_norm": 0.2885092794895172,
+    "learning_rate": 9.287539041387547e-05,
+    "entropy": 0.7683475404977799,
+    "num_tokens": 16704204.0,
+    "mean_token_accuracy": 0.7954743087291718,
+    "epoch": 0.19367333763718528,
+    "step": 750
+  },
+  {
+    "loss": 0.7307,
+    "grad_norm": 0.3181609809398651,
+    "learning_rate": 9.265971920012837e-05,
+    "entropy": 0.735710608959198,
+    "num_tokens": 16933045.0,
+    "mean_token_accuracy": 0.8025461912155152,
+    "epoch": 0.19625564880568108,
+    "step": 760
+  },
+  {
+    "loss": 0.7112,
+    "grad_norm": 0.2486344277858734,
+    "learning_rate": 9.244109037153029e-05,
+    "entropy": 0.7170950889587402,
+    "num_tokens": 17164834.0,
+    "mean_token_accuracy": 0.8034955650568009,
+    "epoch": 0.1988379599741769,
+    "step": 770
+  },
+  {
+    "loss": 0.708,
+    "grad_norm": 0.27479955554008484,
+    "learning_rate": 9.221951908570258e-05,
+    "entropy": 0.7130055457353592,
+    "num_tokens": 17391047.0,
+    "mean_token_accuracy": 0.8064507722854615,
+    "epoch": 0.2014202711426727,
+    "step": 780
+  },
+  {
+    "loss": 0.7404,
+    "grad_norm": 0.27712559700012207,
+    "learning_rate": 9.199502070426833e-05,
+    "entropy": 0.7458436578512192,
+    "num_tokens": 17600014.0,
+    "mean_token_accuracy": 0.8010766774415969,
+    "epoch": 0.2040025823111685,
+    "step": 790
+  },
+  {
+    "loss": 0.7459,
+    "grad_norm": 0.23461352288722992,
+    "learning_rate": 9.176761079178717e-05,
+    "entropy": 0.7532986223697662,
+    "num_tokens": 17839375.0,
+    "mean_token_accuracy": 0.7996405005455017,
+    "epoch": 0.2065848934796643,
+    "step": 800
+  },
+  {
+    "loss": 0.7889,
+    "grad_norm": 0.3072088658809662,
+    "learning_rate": 9.153730511467646e-05,
+    "entropy": 0.7842026621103286,
+    "num_tokens": 18051756.0,
+    "mean_token_accuracy": 0.7949094295501709,
+    "epoch": 0.2091672046481601,
+    "step": 810
+  },
+  {
+    "loss": 0.7256,
+    "grad_norm": 0.2747524380683899,
+    "learning_rate": 9.130411964011795e-05,
+    "entropy": 0.7265078127384186,
+    "num_tokens": 18271944.0,
+    "mean_token_accuracy": 0.8043490052223206,
+    "epoch": 0.2117495158166559,
+    "step": 820
+  },
+  {
+    "loss": 0.7522,
+    "grad_norm": 0.2839372456073761,
+    "learning_rate": 9.10680705349509e-05,
+    "entropy": 0.7600590527057648,
+    "num_tokens": 18493573.0,
+    "mean_token_accuracy": 0.7979195952415467,
+    "epoch": 0.2143318269851517,
+    "step": 830
+  },
+  {
+    "loss": 0.7184,
+    "grad_norm": 0.26603153347969055,
+    "learning_rate": 9.082917416455114e-05,
+    "entropy": 0.727937662601471,
+    "num_tokens": 18732461.0,
+    "mean_token_accuracy": 0.8036612659692765,
+    "epoch": 0.21691413815364752,
+    "step": 840
+  },
+  {
+    "loss": 0.7606,
+    "grad_norm": 0.28745344281196594,
+    "learning_rate": 9.058744709169657e-05,
+    "entropy": 0.7691597014665603,
+    "num_tokens": 18948836.0,
+    "mean_token_accuracy": 0.7954544007778168,
+    "epoch": 0.21949644932214332,
+    "step": 850
+  },
+  {
+    "loss": 0.7432,
+    "grad_norm": 0.24271559715270996,
+    "learning_rate": 9.034290607541875e-05,
+    "entropy": 0.7455548018217086,
+    "num_tokens": 19175411.0,
+    "mean_token_accuracy": 0.7987944543361664,
+    "epoch": 0.22207876049063913,
+    "step": 860
+  },
+  {
+    "loss": 0.7808,
+    "grad_norm": 0.26113608479499817,
+    "learning_rate": 9.009556806984102e-05,
+    "entropy": 0.7865129858255386,
+    "num_tokens": 19401674.0,
+    "mean_token_accuracy": 0.7918182015419006,
+    "epoch": 0.22466107165913493,
+    "step": 870
+  },
+  {
+    "loss": 0.7627,
+    "grad_norm": 0.2622512876987457,
+    "learning_rate": 8.984545022300308e-05,
+    "entropy": 0.7637326002120972,
+    "num_tokens": 19616114.0,
+    "mean_token_accuracy": 0.796609079837799,
+    "epoch": 0.22724338282763074,
+    "step": 880
+  },
+  {
+    "loss": 0.7635,
+    "grad_norm": 0.28230419754981995,
+    "learning_rate": 8.95925698756721e-05,
+    "entropy": 0.7649184703826905,
+    "num_tokens": 19833283.0,
+    "mean_token_accuracy": 0.7957229435443878,
+    "epoch": 0.22982569399612654,
+    "step": 890
+  },
+  {
+    "loss": 0.7139,
+    "grad_norm": 0.2590973675251007,
+    "learning_rate": 8.93369445601405e-05,
+    "entropy": 0.7141411513090133,
+    "num_tokens": 20067846.0,
+    "mean_token_accuracy": 0.8063762694597244,
+    "epoch": 0.23240800516462234,
+    "step": 900
+  },
+  {
+    "loss": 0.7258,
+    "grad_norm": 0.29914793372154236,
+    "learning_rate": 8.907859199901036e-05,
+    "entropy": 0.7249648362398148,
+    "num_tokens": 20282958.0,
+    "mean_token_accuracy": 0.8041715890169143,
+    "epoch": 0.23499031633311815,
+    "step": 910
+  },
+  {
+    "loss": 0.7468,
+    "grad_norm": 0.33904722332954407,
+    "learning_rate": 8.881753010396483e-05,
+    "entropy": 0.7552884995937348,
+    "num_tokens": 20514217.0,
+    "mean_token_accuracy": 0.7984365493059158,
+    "epoch": 0.23757262750161395,
+    "step": 920
+  },
+  {
+    "loss": 0.7469,
+    "grad_norm": 0.2916295528411865,
+    "learning_rate": 8.855377697452613e-05,
+    "entropy": 0.7542608201503753,
+    "num_tokens": 20742976.0,
+    "mean_token_accuracy": 0.8005632907152176,
+    "epoch": 0.24015493867010976,
+    "step": 930
+  },
+  {
+    "loss": 0.7635,
+    "grad_norm": 0.2670243978500366,
+    "learning_rate": 8.828735089680089e-05,
+    "entropy": 0.7616350531578064,
+    "num_tokens": 20967064.0,
+    "mean_token_accuracy": 0.796535101532936,
+    "epoch": 0.24273724983860556,
+    "step": 940
+  },
+  {
+    "loss": 0.7721,
+    "grad_norm": 0.32767635583877563,
+    "learning_rate": 8.801827034221222e-05,
+    "entropy": 0.7816158264875412,
+    "num_tokens": 21198991.0,
+    "mean_token_accuracy": 0.7934033811092377,
+    "epoch": 0.24531956100710137,
+    "step": 950
+  },
+  {
+    "loss": 0.7324,
+    "grad_norm": 0.2466670572757721,
+    "learning_rate": 8.77465539662192e-05,
+    "entropy": 0.7298496663570404,
+    "num_tokens": 21416342.0,
+    "mean_token_accuracy": 0.8032382041215896,
+    "epoch": 0.24790187217559717,
+    "step": 960
+  },
+  {
+    "loss": 0.7203,
+    "grad_norm": 0.32663026452064514,
+    "learning_rate": 8.747222060702336e-05,
+    "entropy": 0.7248791009187698,
+    "num_tokens": 21639020.0,
+    "mean_token_accuracy": 0.8066118597984314,
+    "epoch": 0.25048418334409295,
+    "step": 970
+  },
+  {
+    "loss": 0.787,
+    "grad_norm": 0.2980765700340271,
+    "learning_rate": 8.719528928426273e-05,
+    "entropy": 0.7887016743421554,
+    "num_tokens": 21848300.0,
+    "mean_token_accuracy": 0.7913496702909469,
+    "epoch": 0.2530664945125888,
+    "step": 980
+  },
+  {
+    "loss": 0.7164,
+    "grad_norm": 0.2752622067928314,
+    "learning_rate": 8.691577919769316e-05,
+    "entropy": 0.7202609479427338,
+    "num_tokens": 22067594.0,
+    "mean_token_accuracy": 0.8054278463125228,
+    "epoch": 0.25564880568108456,
+    "step": 990
+  },
+  {
+    "loss": 0.7741,
+    "grad_norm": 0.26739493012428284,
+    "learning_rate": 8.663370972585717e-05,
+    "entropy": 0.7841310113668442,
+    "num_tokens": 22292216.0,
+    "mean_token_accuracy": 0.7926570892333984,
+    "epoch": 0.2582311168495804,
+    "step": 1000
+  },
+  {
+    "loss": 0.7975,
+    "grad_norm": 0.2629258334636688,
+    "learning_rate": 8.634910042474047e-05,
+    "entropy": 0.792700219154358,
+    "num_tokens": 22513422.0,
+    "mean_token_accuracy": 0.7921054184436798,
+    "epoch": 0.26081342801807617,
+    "step": 1010
+  },
+  {
+    "loss": 0.7759,
+    "grad_norm": 0.2839512228965759,
+    "learning_rate": 8.606197102641614e-05,
+    "entropy": 0.7746599346399308,
+    "num_tokens": 22738553.0,
+    "mean_token_accuracy": 0.7940926223993301,
+    "epoch": 0.263395739186572,
+    "step": 1020
+  },
+  {
+    "loss": 0.7525,
+    "grad_norm": 0.3013017177581787,
+    "learning_rate": 8.577234143767649e-05,
+    "entropy": 0.7538807690143585,
+    "num_tokens": 22959380.0,
+    "mean_token_accuracy": 0.7982607930898666,
+    "epoch": 0.2659780503550678,
+    "step": 1030
+  },
+  {
+    "loss": 0.7531,
+    "grad_norm": 0.24685463309288025,
+    "learning_rate": 8.548023173865313e-05,
+    "entropy": 0.75996915102005,
+    "num_tokens": 23183220.0,
+    "mean_token_accuracy": 0.7970141679048538,
+    "epoch": 0.2685603615235636,
+    "step": 1040
+  },
+  {
+    "loss": 0.7856,
+    "grad_norm": 0.2837998867034912,
+    "learning_rate": 8.518566218142454e-05,
+    "entropy": 0.7838736146688461,
+    "num_tokens": 23404006.0,
+    "mean_token_accuracy": 0.7927649199962616,
+    "epoch": 0.2711426726920594,
+    "step": 1050
+  },
+  {
+    "loss": 0.8175,
+    "grad_norm": 0.29661616683006287,
+    "learning_rate": 8.488865318861224e-05,
+    "entropy": 0.8234259188175201,
+    "num_tokens": 23628937.0,
+    "mean_token_accuracy": 0.7849996358156204,
+    "epoch": 0.2737249838605552,
+    "step": 1060
+  },
+  {
+    "loss": 0.722,
+    "grad_norm": 0.25703901052474976,
+    "learning_rate": 8.458922535196467e-05,
+    "entropy": 0.7301439672708512,
+    "num_tokens": 23865644.0,
+    "mean_token_accuracy": 0.8037035495042801,
+    "epoch": 0.276307295029051,
+    "step": 1070
+  },
+  {
+    "loss": 0.7135,
+    "grad_norm": 0.2932882010936737,
+    "learning_rate": 8.428739943092971e-05,
+    "entropy": 0.7183157622814178,
+    "num_tokens": 24084332.0,
+    "mean_token_accuracy": 0.8063235968351364,
+    "epoch": 0.2788896061975468,
+    "step": 1080
+  },
+  {
+    "loss": 0.7216,
+    "grad_norm": 0.27650290727615356,
+    "learning_rate": 8.398319635121534e-05,
+    "entropy": 0.726809737086296,
+    "num_tokens": 24302911.0,
+    "mean_token_accuracy": 0.8047617524862289,
+    "epoch": 0.2814719173660426,
+    "step": 1090
+  },
+  {
+    "loss": 0.7758,
+    "grad_norm": 0.30116310715675354,
+    "learning_rate": 8.367663720333887e-05,
+    "entropy": 0.7749881535768509,
+    "num_tokens": 24536462.0,
+    "mean_token_accuracy": 0.792735579609871,
+    "epoch": 0.28405422853453843,
+    "step": 1100
+  },
+  {
+    "loss": 0.7322,
+    "grad_norm": 0.268449991941452,
+    "learning_rate": 8.336774324116472e-05,
+    "entropy": 0.7460200548171997,
+    "num_tokens": 24746352.0,
+    "mean_token_accuracy": 0.8020590513944625,
+    "epoch": 0.2866365397030342,
+    "step": 1110
+  },
+  {
+    "loss": 0.7794,
+    "grad_norm": 0.2835862934589386,
+    "learning_rate": 8.305653588043092e-05,
+    "entropy": 0.778589203953743,
+    "num_tokens": 24970626.0,
+    "mean_token_accuracy": 0.794257864356041,
+    "epoch": 0.28921885087153004,
+    "step": 1120
+  },
+  {
+    "loss": 0.7859,
+    "grad_norm": 0.2543643116950989,
+    "learning_rate": 8.274303669726426e-05,
+    "entropy": 0.7844608157873154,
+    "num_tokens": 25202908.0,
+    "mean_token_accuracy": 0.7923130601644516,
+    "epoch": 0.2918011620400258,
+    "step": 1130
+  },
+  {
+    "loss": 0.7586,
+    "grad_norm": 0.29340291023254395,
+    "learning_rate": 8.242726742668453e-05,
+    "entropy": 0.7621672093868256,
+    "num_tokens": 25430498.0,
+    "mean_token_accuracy": 0.7958371937274933,
+    "epoch": 0.29438347320852165,
+    "step": 1140
+  },
+  {
+    "loss": 0.796,
+    "grad_norm": 0.2856576442718506,
+    "learning_rate": 8.210924996109752e-05,
+    "entropy": 0.7996839165687561,
+    "num_tokens": 25655907.0,
+    "mean_token_accuracy": 0.7891602605581284,
+    "epoch": 0.2969657843770174,
+    "step": 1150
+  },
+  {
+    "loss": 0.7431,
+    "grad_norm": 0.31065818667411804,
+    "learning_rate": 8.178900634877728e-05,
+    "entropy": 0.7477968245744705,
+    "num_tokens": 25877824.0,
+    "mean_token_accuracy": 0.8015917330980301,
+    "epoch": 0.29954809554551326,
+    "step": 1160
+  },
+  {
+    "loss": 0.7728,
+    "grad_norm": 0.24908140301704407,
+    "learning_rate": 8.146655879233744e-05,
+    "entropy": 0.7769822120666504,
+    "num_tokens": 26110819.0,
+    "mean_token_accuracy": 0.7936273038387298,
+    "epoch": 0.30213040671400904,
+    "step": 1170
+  },
+  {
+    "loss": 0.8123,
+    "grad_norm": 0.3190557658672333,
+    "learning_rate": 8.114192964719197e-05,
+    "entropy": 0.8135595470666885,
+    "num_tokens": 26324524.0,
+    "mean_token_accuracy": 0.7863456726074218,
+    "epoch": 0.30471271788250487,
+    "step": 1180
+  },
+  {
+    "loss": 0.7591,
+    "grad_norm": 0.28709912300109863,
+    "learning_rate": 8.081514142000517e-05,
+    "entropy": 0.76651451587677,
+    "num_tokens": 26543338.0,
+    "mean_token_accuracy": 0.7968069106340409,
+    "epoch": 0.30729502905100065,
+    "step": 1190
+  },
+  {
+    "loss": 0.713,
+    "grad_norm": 0.3083307445049286,
+    "learning_rate": 8.04862167671314e-05,
+    "entropy": 0.7108349561691284,
+    "num_tokens": 26768445.0,
+    "mean_token_accuracy": 0.8063337951898575,
+    "epoch": 0.3098773402194964,
+    "step": 1200
+  },
+  {
+    "loss": 0.7237,
+    "grad_norm": 0.2864898443222046,
+    "learning_rate": 8.015517849304419e-05,
+    "entropy": 0.730633094906807,
+    "num_tokens": 26986834.0,
+    "mean_token_accuracy": 0.8043313831090927,
+    "epoch": 0.31245965138799225,
+    "step": 1210
+  },
+  {
+    "loss": 0.827,
+    "grad_norm": 0.26509392261505127,
+    "learning_rate": 7.98220495487553e-05,
+    "entropy": 0.8305549472570419,
+    "num_tokens": 27212130.0,
+    "mean_token_accuracy": 0.7802114427089691,
+    "epoch": 0.31504196255648803,
+    "step": 1220
+  },
+  {
+    "loss": 0.7437,
+    "grad_norm": 0.27083709836006165,
+    "learning_rate": 7.948685303022341e-05,
+    "entropy": 0.7459201335906982,
+    "num_tokens": 27440385.0,
+    "mean_token_accuracy": 0.7997552633285523,
+    "epoch": 0.31762427372498386,
+    "step": 1230
+  },
+  {
+    "loss": 0.7803,
+    "grad_norm": 0.27397236227989197,
+    "learning_rate": 7.914961217675294e-05,
+    "entropy": 0.7889262080192566,
+    "num_tokens": 27660653.0,
+    "mean_token_accuracy": 0.7899086564779282,
+    "epoch": 0.32020658489347964,
+    "step": 1240
+  },
+  {
+    "loss": 0.6871,
+    "grad_norm": 0.3037557303905487,
+    "learning_rate": 7.881035036938287e-05,
+    "entropy": 0.6911757588386536,
+    "num_tokens": 27879066.0,
+    "mean_token_accuracy": 0.8117859929800033,
+    "epoch": 0.32278889606197547,
+    "step": 1250
+  },
+  {
+    "loss": 0.7546,
+    "grad_norm": 0.28858184814453125,
+    "learning_rate": 7.846909112926564e-05,
+    "entropy": 0.7656300634145736,
+    "num_tokens": 28099197.0,
+    "mean_token_accuracy": 0.7982858657836914,
+    "epoch": 0.32537120723047125,
+    "step": 1260
+  },
+  {
+    "loss": 0.7821,
+    "grad_norm": 0.2936239540576935,
+    "learning_rate": 7.812585811603655e-05,
+    "entropy": 0.781898695230484,
+    "num_tokens": 28318775.0,
+    "mean_token_accuracy": 0.7906658828258515,
+    "epoch": 0.3279535183989671,
+    "step": 1270
+  },
+  {
+    "loss": 0.7853,
+    "grad_norm": 0.28847548365592957,
+    "learning_rate": 7.77806751261733e-05,
+    "entropy": 0.7884863972663879,
+    "num_tokens": 28535279.0,
+    "mean_token_accuracy": 0.7927415996789933,
+    "epoch": 0.33053582956746286,
+    "step": 1280
+  },
+  {
+    "loss": 0.7315,
+    "grad_norm": 0.2842661440372467,
+    "learning_rate": 7.743356609134622e-05,
+    "entropy": 0.7407895147800445,
+    "num_tokens": 28776109.0,
+    "mean_token_accuracy": 0.8013403862714767,
+    "epoch": 0.3331181407359587,
+    "step": 1290
+  },
+  {
+    "loss": 0.7278,
+    "grad_norm": 0.2504037320613861,
+    "learning_rate": 7.708455507675913e-05,
+    "entropy": 0.7367011368274688,
+    "num_tokens": 28991068.0,
+    "mean_token_accuracy": 0.8018357723951339,
+    "epoch": 0.33570045190445447,
+    "step": 1300
+  },
+  {
+    "loss": 0.7557,
+    "grad_norm": 0.2458297312259674,
+    "learning_rate": 7.673366627948082e-05,
+    "entropy": 0.7577834516763687,
+    "num_tokens": 29210689.0,
+    "mean_token_accuracy": 0.7981645226478576,
+    "epoch": 0.3382827630729503,
+    "step": 1310
+  },
+  {
+    "loss": 0.7664,
+    "grad_norm": 0.300417959690094,
+    "learning_rate": 7.638092402676747e-05,
+    "entropy": 0.7715034753084182,
+    "num_tokens": 29429547.0,
+    "mean_token_accuracy": 0.79455007314682,
+    "epoch": 0.3408650742414461,
+    "step": 1320
+  },
+  {
+    "loss": 0.7038,
+    "grad_norm": 0.31124284863471985,
+    "learning_rate": 7.602635277437602e-05,
+    "entropy": 0.7138365268707275,
+    "num_tokens": 29655589.0,
+    "mean_token_accuracy": 0.8058148920536041,
+    "epoch": 0.3434473854099419,
+    "step": 1330
+  },
+  {
+    "loss": 0.7297,
+    "grad_norm": 0.249782532453537,
+    "learning_rate": 7.566997710486873e-05,
+    "entropy": 0.732063102722168,
+    "num_tokens": 29870517.0,
+    "mean_token_accuracy": 0.8033244669437408,
+    "epoch": 0.3460296965784377,
+    "step": 1340
+  },
+  {
+    "loss": 0.7096,
+    "grad_norm": 0.2807137966156006,
+    "learning_rate": 7.531182172590876e-05,
+    "entropy": 0.7162998780608177,
+    "num_tokens": 30106464.0,
+    "mean_token_accuracy": 0.806077653169632,
+    "epoch": 0.3486120077469335,
+    "step": 1350
+  },
+  {
+    "loss": 0.7289,
+    "grad_norm": 0.26910290122032166,
+    "learning_rate": 7.495191146854721e-05,
+    "entropy": 0.7291241198778152,
+    "num_tokens": 30327778.0,
+    "mean_token_accuracy": 0.8027620315551758,
+    "epoch": 0.3511943189154293,
+    "step": 1360
+  },
+  {
+    "loss": 0.6953,
+    "grad_norm": 0.28704628348350525,
+    "learning_rate": 7.45902712855016e-05,
+    "entropy": 0.6994198083877563,
+    "num_tokens": 30559521.0,
+    "mean_token_accuracy": 0.8095674574375152,
+    "epoch": 0.3537766300839251,
+    "step": 1370
+  },
+  {
+    "loss": 0.7287,
+    "grad_norm": 0.29035428166389465,
+    "learning_rate": 7.422692624942585e-05,
+    "entropy": 0.7311005294322968,
+    "num_tokens": 30789659.0,
+    "mean_token_accuracy": 0.8028050363063812,
+    "epoch": 0.3563589412524209,
+    "step": 1380
+  },
+  {
+    "loss": 0.7664,
+    "grad_norm": 0.3093847930431366,
+    "learning_rate": 7.386190155117199e-05,
+    "entropy": 0.7770371347665787,
+    "num_tokens": 31002624.0,
+    "mean_token_accuracy": 0.795510447025299,
+    "epoch": 0.35894125242091673,
+    "step": 1390
+  },
+  {
+    "loss": 0.686,
+    "grad_norm": 0.25263503193855286,
+    "learning_rate": 7.349522249804373e-05,
+    "entropy": 0.6919069796800613,
+    "num_tokens": 31234430.0,
+    "mean_token_accuracy": 0.8122060149908066,
+    "epoch": 0.3615235635894125,
+    "step": 1400
+  },
+  {
+    "loss": 0.7581,
+    "grad_norm": 0.3199212849140167,
+    "learning_rate": 7.312691451204178e-05,
+    "entropy": 0.7603460341691971,
+    "num_tokens": 31451877.0,
+    "mean_token_accuracy": 0.7980766743421555,
+    "epoch": 0.36410587475790834,
+    "step": 1410
+  },
+  {
+    "loss": 0.7254,
+    "grad_norm": 0.28610947728157043,
+    "learning_rate": 7.275700312810144e-05,
+    "entropy": 0.7191128998994827,
+    "num_tokens": 31674092.0,
+    "mean_token_accuracy": 0.8061201065778733,
+    "epoch": 0.3666881859264041,
+    "step": 1420
+  },
+  {
+    "loss": 0.6975,
+    "grad_norm": 0.2749674320220947,
+    "learning_rate": 7.238551399232218e-05,
+    "entropy": 0.6995274364948273,
+    "num_tokens": 31906321.0,
+    "mean_token_accuracy": 0.8110896348953247,
+    "epoch": 0.36927049709489995,
+    "step": 1430
+  },
+  {
+    "loss": 0.7399,
+    "grad_norm": 0.2573988735675812,
+    "learning_rate": 7.201247286018967e-05,
+    "entropy": 0.738545709848404,
+    "num_tokens": 32125362.0,
+    "mean_token_accuracy": 0.8022010952234269,
+    "epoch": 0.37185280826339573,
+    "step": 1440
+  },
+  {
+    "loss": 0.6956,
+    "grad_norm": 0.2564477026462555,
+    "learning_rate": 7.163790559479003e-05,
+    "entropy": 0.703942996263504,
+    "num_tokens": 32355592.0,
+    "mean_token_accuracy": 0.809988585114479,
+    "epoch": 0.37443511943189156,
+    "step": 1450
+  },
+  {
+    "loss": 0.7422,
+    "grad_norm": 0.2665322721004486,
+    "learning_rate": 7.12618381650168e-05,
+    "entropy": 0.7470666587352752,
+    "num_tokens": 32589331.0,
+    "mean_token_accuracy": 0.7986473143100739,
+    "epoch": 0.37701743060038734,
+    "step": 1460
+  },
+  {
+    "loss": 0.7288,
+    "grad_norm": 0.2737554907798767,
+    "learning_rate": 7.088429664377051e-05,
+    "entropy": 0.737991139292717,
+    "num_tokens": 32816610.0,
+    "mean_token_accuracy": 0.8015279024839401,
+    "epoch": 0.37959974176888317,
+    "step": 1470
+  },
+  {
+    "loss": 0.7404,
+    "grad_norm": 0.28172188997268677,
+    "learning_rate": 7.050530720615104e-05,
+    "entropy": 0.7469445317983627,
+    "num_tokens": 33036716.0,
+    "mean_token_accuracy": 0.8000003874301911,
+    "epoch": 0.38218205293737895,
+    "step": 1480
+  },
+  {
+    "loss": 0.7394,
+    "grad_norm": 0.28893858194351196,
+    "learning_rate": 7.012489612764284e-05,
+    "entropy": 0.7446531891822815,
+    "num_tokens": 33262921.0,
+    "mean_token_accuracy": 0.7989077031612396,
+    "epoch": 0.3847643641058748,
+    "step": 1490
+  },
+  {
+    "loss": 0.6756,
+    "grad_norm": 0.2717651426792145,
+    "learning_rate": 6.97430897822933e-05,
+    "entropy": 0.6801834493875504,
+    "num_tokens": 33489654.0,
+    "mean_token_accuracy": 0.8149780780076981,
+    "epoch": 0.38734667527437056,
+    "step": 1500
+  },
+  {
+    "loss": 0.815,
+    "grad_norm": 0.24628806114196777,
+    "learning_rate": 6.935991464088416e-05,
+    "entropy": 0.8130639553070068,
+    "num_tokens": 33715529.0,
+    "mean_token_accuracy": 0.7862419575452805,
+    "epoch": 0.3899289864428664,
+    "step": 1510
+  },
+  {
+    "loss": 0.7699,
+    "grad_norm": 0.3145904242992401,
+    "learning_rate": 6.897539726909637e-05,
+    "entropy": 0.7687238410115242,
+    "num_tokens": 33937441.0,
+    "mean_token_accuracy": 0.7945855349302292,
+    "epoch": 0.39251129761136216,
+    "step": 1520
+  },
+  {
+    "loss": 0.6814,
+    "grad_norm": 0.2653945982456207,
+    "learning_rate": 6.85895643256682e-05,
+    "entropy": 0.6895080089569092,
+    "num_tokens": 34154899.0,
+    "mean_token_accuracy": 0.8130895376205445,
+    "epoch": 0.395093608779858,
+    "step": 1530
+  },
+  {
+    "loss": 0.7194,
+    "grad_norm": 0.2962811291217804,
+    "learning_rate": 6.820244256054704e-05,
+    "entropy": 0.7286921381950379,
+    "num_tokens": 34376870.0,
+    "mean_token_accuracy": 0.8044507384300232,
+    "epoch": 0.3976759199483538,
+    "step": 1540
+  },
+  {
+    "loss": 0.7307,
+    "grad_norm": 0.2986937463283539,
+    "learning_rate": 6.781405881303474e-05,
+    "entropy": 0.7333788439631462,
+    "num_tokens": 34615762.0,
+    "mean_token_accuracy": 0.8029743075370789,
+    "epoch": 0.4002582311168496,
+    "step": 1550
+  },
+  {
+    "loss": 0.7345,
+    "grad_norm": 0.30212172865867615,
+    "learning_rate": 6.742444000992694e-05,
+    "entropy": 0.7404761523008346,
+    "num_tokens": 34841635.0,
+    "mean_token_accuracy": 0.8009481459856034,
+    "epoch": 0.4028405422853454,
+    "step": 1560
+  },
+  {
+    "loss": 0.7055,
+    "grad_norm": 0.31336578726768494,
+    "learning_rate": 6.70336131636461e-05,
+    "entropy": 0.7085718095302582,
+    "num_tokens": 35065220.0,
+    "mean_token_accuracy": 0.8091004997491836,
+    "epoch": 0.4054228534538412,
+    "step": 1570
+  },
+  {
+    "loss": 0.7711,
+    "grad_norm": 0.2713090181350708,
+    "learning_rate": 6.66416053703688e-05,
+    "entropy": 0.7731089383363724,
+    "num_tokens": 35297893.0,
+    "mean_token_accuracy": 0.7932068020105362,
+    "epoch": 0.408005164622337,
+    "step": 1580
+  },
+  {
+    "loss": 0.7632,
+    "grad_norm": 0.2571941018104553,
+    "learning_rate": 6.624844380814717e-05,
+    "entropy": 0.7549047037959099,
+    "num_tokens": 35529896.0,
+    "mean_token_accuracy": 0.7992217838764191,
+    "epoch": 0.41058747579083277,
+    "step": 1590
+  },
+  {
+    "loss": 0.779,
+    "grad_norm": 0.3086455762386322,
+    "learning_rate": 6.585415573502455e-05,
+    "entropy": 0.7779949843883515,
+    "num_tokens": 35748202.0,
+    "mean_token_accuracy": 0.795146381855011,
+    "epoch": 0.4131697869593286,
+    "step": 1600
+  },
+  {
+    "loss": 0.7009,
+    "grad_norm": 0.28316617012023926,
+    "learning_rate": 6.545876848714574e-05,
+    "entropy": 0.7077500820159912,
+    "num_tokens": 35963953.0,
+    "mean_token_accuracy": 0.8086227357387543,
+    "epoch": 0.4157520981278244,
+    "step": 1610
+  },
+  {
+    "loss": 0.6681,
+    "grad_norm": 0.2951730191707611,
+    "learning_rate": 6.506230947686172e-05,
+    "entropy": 0.6698646396398544,
+    "num_tokens": 36187006.0,
+    "mean_token_accuracy": 0.8158697307109832,
+    "epoch": 0.4183344092963202,
+    "step": 1620
+  },
+  {
+    "loss": 0.7176,
+    "grad_norm": 0.28841638565063477,
+    "learning_rate": 6.466480619082919e-05,
+    "entropy": 0.7250782191753388,
+    "num_tokens": 36397760.0,
+    "mean_token_accuracy": 0.8054184794425965,
+    "epoch": 0.420916720464816,
+    "step": 1630
+  },
+  {
+    "loss": 0.689,
+    "grad_norm": 0.2621630132198334,
+    "learning_rate": 6.426628618810491e-05,
+    "entropy": 0.7003191500902176,
+    "num_tokens": 36618227.0,
+    "mean_token_accuracy": 0.8108294665813446,
+    "epoch": 0.4234990316333118,
+    "step": 1640
+  },
+  {
+    "loss": 0.7341,
+    "grad_norm": 0.25907382369041443,
+    "learning_rate": 6.386677709823496e-05,
+    "entropy": 0.7378379344940186,
+    "num_tokens": 36829652.0,
+    "mean_token_accuracy": 0.801531919836998,
+    "epoch": 0.4260813428018076,
+    "step": 1650
+  },
+  {
+    "loss": 0.7483,
+    "grad_norm": 0.3523082137107849,
+    "learning_rate": 6.346630661933927e-05,
+    "entropy": 0.742335370182991,
+    "num_tokens": 37049020.0,
+    "mean_token_accuracy": 0.8001793205738068,
+    "epoch": 0.4286636539703034,
+    "step": 1660
+  },
+  {
+    "loss": 0.7819,
+    "grad_norm": 0.2581123113632202,
+    "learning_rate": 6.306490251619121e-05,
+    "entropy": 0.78986496925354,
+    "num_tokens": 37270090.0,
+    "mean_token_accuracy": 0.7922927856445312,
+    "epoch": 0.4312459651387992,
+    "step": 1670
+  },
+  {
+    "loss": 0.7424,
+    "grad_norm": 0.28055480122566223,
+    "learning_rate": 6.266259261829266e-05,
+    "entropy": 0.7482966244220733,
+    "num_tokens": 37500241.0,
+    "mean_token_accuracy": 0.8013508349657059,
+    "epoch": 0.43382827630729504,
+    "step": 1680
+  },
+  {
+    "loss": 0.8196,
+    "grad_norm": 0.2631305158138275,
+    "learning_rate": 6.225940481794463e-05,
+    "entropy": 0.8206893265247345,
+    "num_tokens": 37722357.0,
+    "mean_token_accuracy": 0.7845668315887451,
+    "epoch": 0.4364105874757908,
+    "step": 1690
+  },
+  {
+    "loss": 0.7689,
+    "grad_norm": 0.27216827869415283,
+    "learning_rate": 6.185536706831346e-05,
+    "entropy": 0.7689682602882385,
+    "num_tokens": 37950299.0,
+    "mean_token_accuracy": 0.7950216263532639,
+    "epoch": 0.43899289864428664,
+    "step": 1700
+  },
+  {
+    "loss": 0.7273,
+    "grad_norm": 0.30833864212036133,
+    "learning_rate": 6.145050738149277e-05,
+    "entropy": 0.7360947251319885,
+    "num_tokens": 38167912.0,
+    "mean_token_accuracy": 0.8035003364086151,
+    "epoch": 0.4415752098127824,
+    "step": 1710
+  },
+  {
+    "loss": 0.7299,
+    "grad_norm": 0.31663721799850464,
+    "learning_rate": 6.104485382656142e-05,
+    "entropy": 0.7425105035305023,
+    "num_tokens": 38390268.0,
+    "mean_token_accuracy": 0.8013977319002151,
+    "epoch": 0.44415752098127825,
+    "step": 1720
+  },
+  {
+    "loss": 0.7602,
+    "grad_norm": 0.28522592782974243,
+    "learning_rate": 6.063843452763744e-05,
+    "entropy": 0.758876496553421,
+    "num_tokens": 38611911.0,
+    "mean_token_accuracy": 0.7959836840629577,
+    "epoch": 0.44673983214977403,
+    "step": 1730
+  },
+  {
+    "loss": 0.747,
+    "grad_norm": 0.2787739932537079,
+    "learning_rate": 6.023127766192824e-05,
+    "entropy": 0.7534547060728073,
+    "num_tokens": 38830188.0,
+    "mean_token_accuracy": 0.7991040110588074,
+    "epoch": 0.44932214331826986,
+    "step": 1740
+  },
+  {
+    "loss": 0.7143,
+    "grad_norm": 0.3082408010959625,
+    "learning_rate": 5.982341145777695e-05,
+    "entropy": 0.7169277727603912,
+    "num_tokens": 39049777.0,
+    "mean_token_accuracy": 0.8069923728704452,
+    "epoch": 0.45190445448676564,
+    "step": 1750
+  },
+  {
+    "loss": 0.7273,
+    "grad_norm": 0.2924341857433319,
+    "learning_rate": 5.9414864192705424e-05,
+    "entropy": 0.7336899906396865,
+    "num_tokens": 39269635.0,
+    "mean_token_accuracy": 0.802969342470169,
+    "epoch": 0.45448676565526147,
+    "step": 1760
+  },
+  {
+    "loss": 0.7215,
+    "grad_norm": 0.27919134497642517,
+    "learning_rate": 5.900566419145377e-05,
+    "entropy": 0.7257662117481232,
+    "num_tokens": 39487714.0,
+    "mean_token_accuracy": 0.8041402757167816,
+    "epoch": 0.45706907682375725,
+    "step": 1770
+  },
+  {
+    "loss": 0.6977,
+    "grad_norm": 0.25231438875198364,
+    "learning_rate": 5.859583982401653e-05,
+    "entropy": 0.6998866051435471,
+    "num_tokens": 39714170.0,
+    "mean_token_accuracy": 0.8111774027347565,
+    "epoch": 0.4596513879922531,
+    "step": 1780
+  },
+  {
+    "loss": 0.742,
+    "grad_norm": 0.24185897409915924,
+    "learning_rate": 5.818541950367579e-05,
+    "entropy": 0.7464011371135711,
+    "num_tokens": 39927978.0,
+    "mean_token_accuracy": 0.8016654849052429,
+    "epoch": 0.46223369916074886,
+    "step": 1790
+  },
+  {
+    "loss": 0.703,
+    "grad_norm": 0.25616031885147095,
+    "learning_rate": 5.7774431685031274e-05,
+    "entropy": 0.7071524858474731,
+    "num_tokens": 40144816.0,
+    "mean_token_accuracy": 0.8098747581243515,
+    "epoch": 0.4648160103292447,
+    "step": 1800
+  },
+  {
+    "loss": 0.6964,
+    "grad_norm": 0.31371667981147766,
+    "learning_rate": 5.736290486202759e-05,
+    "entropy": 0.6939981669187546,
+    "num_tokens": 40345533.0,
+    "mean_token_accuracy": 0.8119114637374878,
+    "epoch": 0.46739832149774047,
+    "step": 1810
+  },
+  {
+    "loss": 0.8156,
+    "grad_norm": 0.3086910545825958,
+    "learning_rate": 5.6950867565978735e-05,
+    "entropy": 0.8133882731199265,
+    "num_tokens": 40554468.0,
+    "mean_token_accuracy": 0.7859878808259964,
+    "epoch": 0.4699806326662363,
+    "step": 1820
+  },
+  {
+    "loss": 0.6717,
+    "grad_norm": 0.2788376808166504,
+    "learning_rate": 5.653834836359e-05,
+    "entropy": 0.6821775257587432,
+    "num_tokens": 40783644.0,
+    "mean_token_accuracy": 0.8164048582315445,
+    "epoch": 0.4725629438347321,
+    "step": 1830
+  },
+  {
+    "loss": 0.7689,
+    "grad_norm": 0.3284967243671417,
+    "learning_rate": 5.6125375854977426e-05,
+    "entropy": 0.7694633185863495,
+    "num_tokens": 40996183.0,
+    "mean_token_accuracy": 0.7952580213546753,
+    "epoch": 0.4751452550032279,
+    "step": 1840
+  },
+  {
+    "loss": 0.7344,
+    "grad_norm": 0.2739453613758087,
+    "learning_rate": 5.571197867168493e-05,
+    "entropy": 0.7373751908540725,
+    "num_tokens": 41216379.0,
+    "mean_token_accuracy": 0.8036257177591324,
+    "epoch": 0.4777275661717237,
+    "step": 1850
+  },
+  {
+    "loss": 0.7846,
+    "grad_norm": 0.2590038478374481,
+    "learning_rate": 5.529818547469932e-05,
+    "entropy": 0.7804774641990662,
+    "num_tokens": 41444373.0,
+    "mean_token_accuracy": 0.7919269859790802,
+    "epoch": 0.4803098773402195,
+    "step": 1860
+  },
+  {
+    "loss": 0.7529,
+    "grad_norm": 0.34748631715774536,
+    "learning_rate": 5.488402495246316e-05,
+    "entropy": 0.7593414902687072,
+    "num_tokens": 41669265.0,
+    "mean_token_accuracy": 0.7967518717050552,
+    "epoch": 0.4828921885087153,
+    "step": 1870
+  },
+  {
+    "loss": 0.7025,
+    "grad_norm": 0.2958257496356964,
+    "learning_rate": 5.446952581888586e-05,
+    "entropy": 0.7056142598390579,
+    "num_tokens": 41888599.0,
+    "mean_token_accuracy": 0.8074656426906586,
+    "epoch": 0.4854744996772111,
+    "step": 1880
+  },
+  {
+    "loss": 0.6623,
+    "grad_norm": 0.32922279834747314,
+    "learning_rate": 5.4054716811352816e-05,
+    "entropy": 0.6658655419945717,
+    "num_tokens": 42103445.0,
+    "mean_token_accuracy": 0.8173845767974853,
+    "epoch": 0.4880568108457069,
+    "step": 1890
+  },
+  {
+    "loss": 0.6835,
+    "grad_norm": 0.27834922075271606,
+    "learning_rate": 5.363962668873317e-05,
+    "entropy": 0.6853355497121811,
+    "num_tokens": 42324754.0,
+    "mean_token_accuracy": 0.8134206265211106,
+    "epoch": 0.49063912201420273,
+    "step": 1900
+  },
+  {
+    "loss": 0.6927,
+    "grad_norm": 0.36218395829200745,
+    "learning_rate": 5.322428422938585e-05,
+    "entropy": 0.6967261880636215,
+    "num_tokens": 42548676.0,
+    "mean_token_accuracy": 0.8099091470241546,
+    "epoch": 0.4932214331826985,
+    "step": 1910
+  },
+  {
+    "loss": 0.7314,
+    "grad_norm": 0.26530325412750244,
+    "learning_rate": 5.2808718229164375e-05,
+    "entropy": 0.7388436585664749,
+    "num_tokens": 42780674.0,
+    "mean_token_accuracy": 0.8034113436937332,
+    "epoch": 0.49580374435119434,
+    "step": 1920
+  },
+  {
+    "loss": 0.7003,
+    "grad_norm": 0.28567594289779663,
+    "learning_rate": 5.239295749942049e-05,
+    "entropy": 0.700775408744812,
+    "num_tokens": 42993726.0,
+    "mean_token_accuracy": 0.8110381156206131,
+    "epoch": 0.4983860555196901,
+    "step": 1930
+  },
+  {
+    "loss": 0.7674,
+    "grad_norm": 0.2515343725681305,
+    "learning_rate": 5.197703086500656e-05,
+    "entropy": 0.7715499877929688,
+    "num_tokens": 43224427.0,
+    "mean_token_accuracy": 0.796505457162857,
+    "epoch": 0.5009683666881859,
+    "step": 1940
+  },
+  {
+    "loss": 0.7176,
+    "grad_norm": 0.28791186213493347,
+    "learning_rate": 5.156096716227719e-05,
+    "entropy": 0.7235445290803909,
+    "num_tokens": 43435326.0,
+    "mean_token_accuracy": 0.8063592225313186,
+    "epoch": 0.5035506778566817,
+    "step": 1950
+  },
+  {
+    "loss": 0.6719,
+    "grad_norm": 0.2640407383441925,
+    "learning_rate": 5.114479523709003e-05,
+    "entropy": 0.6716303527355194,
+    "num_tokens": 43643191.0,
+    "mean_token_accuracy": 0.8156255722045899,
+    "epoch": 0.5061329890251776,
+    "step": 1960
+  },
+  {
+    "loss": 0.7137,
+    "grad_norm": 0.2662080228328705,
+    "learning_rate": 5.072854394280578e-05,
+    "entropy": 0.7221474066376686,
+    "num_tokens": 43868551.0,
+    "mean_token_accuracy": 0.8072103232145309,
+    "epoch": 0.5087153001936734,
+    "step": 1970
+  },
+  {
+    "loss": 0.7388,
+    "grad_norm": 0.266189306974411,
+    "learning_rate": 5.031224213828788e-05,
+    "entropy": 0.7444506168365479,
+    "num_tokens": 44092695.0,
+    "mean_token_accuracy": 0.8010527819395066,
+    "epoch": 0.5112976113621691,
+    "step": 1980
+  },
+  {
+    "loss": 0.7565,
+    "grad_norm": 0.27445709705352783,
+    "learning_rate": 4.989591868590162e-05,
+    "entropy": 0.7575427994132042,
+    "num_tokens": 44308972.0,
+    "mean_token_accuracy": 0.7987733721733093,
+    "epoch": 0.513879922530665,
+    "step": 1990
+  },
+  {
+    "loss": 0.675,
+    "grad_norm": 0.24522656202316284,
+    "learning_rate": 4.9479602449513174e-05,
+    "entropy": 0.6674839437007904,
+    "num_tokens": 44539892.0,
+    "mean_token_accuracy": 0.8155701875686645,
+    "epoch": 0.5164622336991608,
+    "step": 2000
+  },
+  {
+    "loss": 0.6908,
+    "grad_norm": 0.27964839339256287,
+    "learning_rate": 4.9063322292488414e-05,
+    "entropy": 0.6955849170684815,
+    "num_tokens": 44757659.0,
+    "mean_token_accuracy": 0.8129804879426956,
+    "epoch": 0.5190445448676565,
+    "step": 2010
+  },
+  {
+    "loss": 0.714,
+    "grad_norm": 0.28458577394485474,
+    "learning_rate": 4.864710707569181e-05,
+    "entropy": 0.7280423313379287,
+    "num_tokens": 44969447.0,
+    "mean_token_accuracy": 0.8045047700405121,
+    "epoch": 0.5216268560361523,
+    "step": 2020
+  },
+  {
+    "loss": 0.7488,
+    "grad_norm": 0.3118044435977936,
+    "learning_rate": 4.8230985655485525e-05,
+    "entropy": 0.7563327640295029,
+    "num_tokens": 45191063.0,
+    "mean_token_accuracy": 0.7996702790260315,
+    "epoch": 0.5242091672046482,
+    "step": 2030
+  },
+  {
+    "loss": 0.7041,
+    "grad_norm": 0.2871478796005249,
+    "learning_rate": 4.781498688172875e-05,
+    "entropy": 0.7103944063186646,
+    "num_tokens": 45410657.0,
+    "mean_token_accuracy": 0.8094416737556458,
+    "epoch": 0.526791478373144,
+    "step": 2040
+  },
+  {
+    "loss": 0.7267,
+    "grad_norm": 0.2895406186580658,
+    "learning_rate": 4.739913959577755e-05,
+    "entropy": 0.7345693051815033,
+    "num_tokens": 45633161.0,
+    "mean_token_accuracy": 0.802816191315651,
+    "epoch": 0.5293737895416397,
+    "step": 2050
+  },
+  {
+    "loss": 0.7216,
+    "grad_norm": 0.3061612844467163,
+    "learning_rate": 4.6983472628485286e-05,
+    "entropy": 0.7228217035531997,
+    "num_tokens": 45863461.0,
+    "mean_token_accuracy": 0.8049194812774658,
+    "epoch": 0.5319561007101355,
+    "step": 2060
+  },
+  {
+    "loss": 0.7492,
+    "grad_norm": 0.28822118043899536,
+    "learning_rate": 4.6568014798203765e-05,
+    "entropy": 0.7588370621204377,
+    "num_tokens": 46086225.0,
+    "mean_token_accuracy": 0.7975336879491806,
+    "epoch": 0.5345384118786314,
+    "step": 2070
+  },
+  {
+    "loss": 0.7334,
+    "grad_norm": 0.2947237491607666,
+    "learning_rate": 4.615279490878522e-05,
+    "entropy": 0.7364494383335114,
+    "num_tokens": 46306577.0,
+    "mean_token_accuracy": 0.8028736978769302,
+    "epoch": 0.5371207230471272,
+    "step": 2080
+  },
+  {
+    "loss": 0.7402,
+    "grad_norm": 0.28399375081062317,
+    "learning_rate": 4.573784174758538e-05,
+    "entropy": 0.7495404034852982,
+    "num_tokens": 46521889.0,
+    "mean_token_accuracy": 0.8020780056715011,
+    "epoch": 0.5397030342156229,
+    "step": 2090
+  },
+  {
+    "loss": 0.7247,
+    "grad_norm": 0.3146340548992157,
+    "learning_rate": 4.532318408346757e-05,
+    "entropy": 0.732389822602272,
+    "num_tokens": 46745832.0,
+    "mean_token_accuracy": 0.8045948296785355,
+    "epoch": 0.5422853453841188,
+    "step": 2100
+  },
+  {
+    "loss": 0.7203,
+    "grad_norm": 0.33135199546813965,
+    "learning_rate": 4.4908850664808245e-05,
+    "entropy": 0.7190593853592873,
+    "num_tokens": 46968971.0,
+    "mean_token_accuracy": 0.8061521053314209,
+    "epoch": 0.5448676565526146,
+    "step": 2110
+  },
+  {
+    "loss": 0.7253,
+    "grad_norm": 0.25495028495788574,
+    "learning_rate": 4.449487021750373e-05,
+    "entropy": 0.7277108535170556,
+    "num_tokens": 47177907.0,
+    "mean_token_accuracy": 0.8039905726909637,
+    "epoch": 0.5474499677211104,
+    "step": 2120
+  },
+  {
+    "loss": 0.7326,
+    "grad_norm": 0.2718372046947479,
+    "learning_rate": 4.4081271442978745e-05,
+    "entropy": 0.7357953459024429,
+    "num_tokens": 47405943.0,
+    "mean_token_accuracy": 0.8034853279590607,
+    "epoch": 0.5500322788896062,
+    "step": 2130
+  },
+  {
+    "loss": 0.7994,
+    "grad_norm": 0.29703348875045776,
+    "learning_rate": 4.3668083016196486e-05,
+    "entropy": 0.8072989761829377,
+    "num_tokens": 47625664.0,
+    "mean_token_accuracy": 0.7880438774824142,
+    "epoch": 0.552614590058102,
+    "step": 2140
+  },
+  {
+    "loss": 0.7094,
+    "grad_norm": 0.2337184101343155,
+    "learning_rate": 4.325533358367058e-05,
+    "entropy": 0.710915943980217,
+    "num_tokens": 47847735.0,
+    "mean_token_accuracy": 0.8080082088708878,
+    "epoch": 0.5551969012265978,
+    "step": 2150
+  },
+  {
+    "loss": 0.6969,
+    "grad_norm": 0.25921082496643066,
+    "learning_rate": 4.284305176147901e-05,
+    "entropy": 0.7076527655124665,
+    "num_tokens": 48088651.0,
+    "mean_token_accuracy": 0.8098387718200684,
+    "epoch": 0.5577792123950936,
+    "step": 2160
+  },
+  {
+    "loss": 0.727,
+    "grad_norm": 0.29743558168411255,
+    "learning_rate": 4.243126613328015e-05,
+    "entropy": 0.7333346515893936,
+    "num_tokens": 48311841.0,
+    "mean_token_accuracy": 0.8033136278390884,
+    "epoch": 0.5603615235635894,
+    "step": 2170
+  },
+  {
+    "loss": 0.668,
+    "grad_norm": 0.2839380204677582,
+    "learning_rate": 4.2020005248331054e-05,
+    "entropy": 0.680330915749073,
+    "num_tokens": 48541172.0,
+    "mean_token_accuracy": 0.8162445783615112,
+    "epoch": 0.5629438347320852,
+    "step": 2180
+  },
+  {
+    "loss": 0.7175,
+    "grad_norm": 0.28205934166908264,
+    "learning_rate": 4.160929761950811e-05,
+    "entropy": 0.719845575094223,
+    "num_tokens": 48764865.0,
+    "mean_token_accuracy": 0.8060573965311051,
+    "epoch": 0.565526145900581,
+    "step": 2190
+  },
+  {
+    "loss": 0.6914,
+    "grad_norm": 0.32426223158836365,
+    "learning_rate": 4.11991717213303e-05,
+    "entropy": 0.6926346108317375,
+    "num_tokens": 48994462.0,
+    "mean_token_accuracy": 0.8123312324285508,
+    "epoch": 0.5681084570690769,
+    "step": 2200
+  },
+  {
+    "loss": 0.726,
+    "grad_norm": 0.26659542322158813,
+    "learning_rate": 4.0789655987984925e-05,
+    "entropy": 0.7318926438689232,
+    "num_tokens": 49217593.0,
+    "mean_token_accuracy": 0.8045945227146148,
+    "epoch": 0.5706907682375726,
+    "step": 2210
+  },
+  {
+    "loss": 0.7102,
+    "grad_norm": 0.28396132588386536,
+    "learning_rate": 4.0380778811356365e-05,
+    "entropy": 0.7165648102760315,
+    "num_tokens": 49451191.0,
+    "mean_token_accuracy": 0.8072537958621979,
+    "epoch": 0.5732730794060684,
+    "step": 2220
+  },
+  {
+    "loss": 0.6689,
+    "grad_norm": 0.28661495447158813,
+    "learning_rate": 3.99725685390576e-05,
+    "entropy": 0.6788897529244423,
+    "num_tokens": 49682456.0,
+    "mean_token_accuracy": 0.815646868944168,
+    "epoch": 0.5758553905745643,
+    "step": 2230
+  },
+  {
+    "loss": 0.7025,
+    "grad_norm": 0.2851995825767517,
+    "learning_rate": 3.956505347246487e-05,
+    "entropy": 0.7025539427995682,
+    "num_tokens": 49918044.0,
+    "mean_token_accuracy": 0.8090035438537597,
+    "epoch": 0.5784377017430601,
+    "step": 2240
+  },
+  {
+    "loss": 0.6964,
+    "grad_norm": 0.2995574474334717,
+    "learning_rate": 3.915826186475555e-05,
+    "entropy": 0.6998808637261391,
+    "num_tokens": 50139989.0,
+    "mean_token_accuracy": 0.8096735835075378,
+    "epoch": 0.5810200129115558,
+    "step": 2250
+  },
+  {
+    "loss": 0.724,
+    "grad_norm": 0.29106131196022034,
+    "learning_rate": 3.8752221918949306e-05,
+    "entropy": 0.7294921964406967,
+    "num_tokens": 50359939.0,
+    "mean_token_accuracy": 0.8044734835624695,
+    "epoch": 0.5836023240800516,
+    "step": 2260
+  },
+  {
+    "loss": 0.7017,
+    "grad_norm": 0.32531774044036865,
+    "learning_rate": 3.834696178595282e-05,
+    "entropy": 0.7081922739744186,
+    "num_tokens": 50586979.0,
+    "mean_token_accuracy": 0.8094498455524445,
+    "epoch": 0.5861846352485475,
+    "step": 2270
+  },
+  {
+    "loss": 0.7084,
+    "grad_norm": 0.26263847947120667,
+    "learning_rate": 3.794250956260805e-05,
+    "entropy": 0.7074251011013984,
+    "num_tokens": 50813902.0,
+    "mean_token_accuracy": 0.808785879611969,
+    "epoch": 0.5887669464170433,
+    "step": 2280
+  },
+  {
+    "loss": 0.7549,
+    "grad_norm": 0.29434698820114136,
+    "learning_rate": 3.753889328974423e-05,
+    "entropy": 0.7641531646251678,
+    "num_tokens": 51044696.0,
+    "mean_token_accuracy": 0.7972988486289978,
+    "epoch": 0.591349257585539,
+    "step": 2290
+  },
+  {
+    "loss": 0.7627,
+    "grad_norm": 0.2781457304954529,
+    "learning_rate": 3.7136140950233864e-05,
+    "entropy": 0.764645341038704,
+    "num_tokens": 51267662.0,
+    "mean_token_accuracy": 0.79625363945961,
+    "epoch": 0.5939315687540349,
+    "step": 2300
+  },
+  {
+    "loss": 0.7725,
+    "grad_norm": 0.27123042941093445,
+    "learning_rate": 3.6734280467052595e-05,
+    "entropy": 0.7800057739019394,
+    "num_tokens": 51486424.0,
+    "mean_token_accuracy": 0.7927343875169754,
+    "epoch": 0.5965138799225307,
+    "step": 2310
+  },
+  {
+    "loss": 0.7276,
+    "grad_norm": 0.3310370445251465,
+    "learning_rate": 3.633333970134334e-05,
+    "entropy": 0.7255799725651741,
+    "num_tokens": 51710712.0,
+    "mean_token_accuracy": 0.8043308049440384,
+    "epoch": 0.5990961910910265,
+    "step": 2320
+  },
+  {
+    "loss": 0.7186,
+    "grad_norm": 0.30834323167800903,
+    "learning_rate": 3.593334645048463e-05,
+    "entropy": 0.7251825213432312,
+    "num_tokens": 51929200.0,
+    "mean_token_accuracy": 0.8063902944326401,
+    "epoch": 0.6016785022595222,
+    "step": 2330
+  },
+  {
+    "loss": 0.7208,
+    "grad_norm": 0.2710685431957245,
+    "learning_rate": 3.5534328446163444e-05,
+    "entropy": 0.7191109150648117,
+    "num_tokens": 52150269.0,
+    "mean_token_accuracy": 0.8055253535509109,
+    "epoch": 0.6042608134280181,
+    "step": 2340
+  },
+  {
+    "loss": 0.6743,
+    "grad_norm": 0.29122015833854675,
+    "learning_rate": 3.5136313352452523e-05,
+    "entropy": 0.6803427249193191,
+    "num_tokens": 52373985.0,
+    "mean_token_accuracy": 0.8142094522714615,
+    "epoch": 0.6068431245965139,
+    "step": 2350
+  },
+  {
+    "loss": 0.7097,
+    "grad_norm": 0.28865063190460205,
+    "learning_rate": 3.473932876389244e-05,
+    "entropy": 0.7169732004404068,
+    "num_tokens": 52596955.0,
+    "mean_token_accuracy": 0.8064839720726014,
+    "epoch": 0.6094254357650097,
+    "step": 2360
+  },
+  {
+    "loss": 0.7808,
+    "grad_norm": 0.2779035270214081,
+    "learning_rate": 3.434340220357843e-05,
+    "entropy": 0.7823223441839218,
+    "num_tokens": 52828509.0,
+    "mean_token_accuracy": 0.7934709966182709,
+    "epoch": 0.6120077469335055,
+    "step": 2370
+  },
+  {
+    "loss": 0.7063,
+    "grad_norm": 0.30650147795677185,
+    "learning_rate": 3.3948561121252254e-05,
+    "entropy": 0.7046063154935837,
+    "num_tokens": 53056392.0,
+    "mean_token_accuracy": 0.8085776209831238,
+    "epoch": 0.6145900581020013,
+    "step": 2380
+  },
+  {
+    "loss": 0.7573,
+    "grad_norm": 0.2515726387500763,
+    "learning_rate": 3.3554832891399015e-05,
+    "entropy": 0.7701946586370468,
+    "num_tokens": 53289453.0,
+    "mean_token_accuracy": 0.7973541855812073,
+    "epoch": 0.6171723692704971,
+    "step": 2390
+  },
+  {
+    "loss": 0.6877,
+    "grad_norm": 0.2916530966758728,
+    "learning_rate": 3.3162244811349366e-05,
+    "entropy": 0.6941643834114075,
+    "num_tokens": 53511361.0,
+    "mean_token_accuracy": 0.8097677916288376,
+    "epoch": 0.6197546804389928,
+    "step": 2400
+  },
+  {
+    "loss": 0.6698,
+    "grad_norm": 0.35168713331222534,
+    "learning_rate": 3.277082409938691e-05,
+    "entropy": 0.6763507977128029,
+    "num_tokens": 53733013.0,
+    "mean_token_accuracy": 0.8161178439855575,
+    "epoch": 0.6223369916074887,
+    "step": 2410
+  },
+  {
+    "loss": 0.6756,
+    "grad_norm": 0.2850055396556854,
+    "learning_rate": 3.2380597892861175e-05,
+    "entropy": 0.676644204556942,
+    "num_tokens": 53957269.0,
+    "mean_token_accuracy": 0.8149404138326645,
+    "epoch": 0.6249193027759845,
+    "step": 2420
+  },
+  {
+    "loss": 0.7038,
+    "grad_norm": 0.30560407042503357,
+    "learning_rate": 3.1991593246306164e-05,
+    "entropy": 0.7020898759365082,
+    "num_tokens": 54171717.0,
+    "mean_token_accuracy": 0.8107654899358749,
+    "epoch": 0.6275016139444803,
+    "step": 2430
+  },
+  {
+    "loss": 0.6545,
+    "grad_norm": 0.30098938941955566,
+    "learning_rate": 3.160383712956466e-05,
+    "entropy": 0.656835712492466,
+    "num_tokens": 54398417.0,
+    "mean_token_accuracy": 0.82038534283638,
+    "epoch": 0.6300839251129761,
+    "step": 2440
+  },
+  {
+    "loss": 0.7134,
+    "grad_norm": 0.24808907508850098,
+    "learning_rate": 3.121735642591838e-05,
+    "entropy": 0.716155156493187,
+    "num_tokens": 54616605.0,
+    "mean_token_accuracy": 0.8066381633281707,
+    "epoch": 0.6326662362814719,
+    "step": 2450
+  },
+  {
+    "loss": 0.7269,
+    "grad_norm": 0.32362425327301025,
+    "learning_rate": 3.083217793022415e-05,
+    "entropy": 0.7308804154396057,
+    "num_tokens": 54835949.0,
+    "mean_token_accuracy": 0.8038643062114715,
+    "epoch": 0.6352485474499677,
+    "step": 2460
+  },
+  {
+    "loss": 0.6981,
+    "grad_norm": 0.31787142157554626,
+    "learning_rate": 3.0448328347056244e-05,
+    "entropy": 0.6972566187381745,
+    "num_tokens": 55063212.0,
+    "mean_token_accuracy": 0.8092873215675354,
+    "epoch": 0.6378308586184636,
+    "step": 2470
+  },
+  {
+    "loss": 0.7066,
+    "grad_norm": 0.3164233863353729,
+    "learning_rate": 3.0065834288854893e-05,
+    "entropy": 0.7190612822771072,
+    "num_tokens": 55288008.0,
+    "mean_token_accuracy": 0.8090256512165069,
+    "epoch": 0.6404131697869593,
+    "step": 2480
+  },
+  {
+    "loss": 0.7241,
+    "grad_norm": 0.3039920926094055,
+    "learning_rate": 2.968472227408127e-05,
+    "entropy": 0.718104162812233,
+    "num_tokens": 55501656.0,
+    "mean_token_accuracy": 0.8054259717464447,
+    "epoch": 0.6429954809554551,
+    "step": 2490
+  },
+  {
+    "loss": 0.7009,
+    "grad_norm": 0.2654428482055664,
+    "learning_rate": 2.930501872537891e-05,
+    "entropy": 0.7064283490180969,
+    "num_tokens": 55720044.0,
+    "mean_token_accuracy": 0.8091771513223648,
+    "epoch": 0.6455777921239509,
+    "step": 2500
+  },
+  {
+    "loss": 0.672,
+    "grad_norm": 0.3438260555267334,
+    "learning_rate": 2.8926749967741908e-05,
+    "entropy": 0.673621541261673,
+    "num_tokens": 55945963.0,
+    "mean_token_accuracy": 0.8174030244350433,
+    "epoch": 0.6481601032924468,
+    "step": 2510
+  },
+  {
+    "loss": 0.7376,
+    "grad_norm": 0.2591122090816498,
+    "learning_rate": 2.8549942226689673e-05,
+    "entropy": 0.7400184273719788,
+    "num_tokens": 56180380.0,
+    "mean_token_accuracy": 0.8010205596685409,
+    "epoch": 0.6507424144609425,
+    "step": 2520
+  },
+  {
+    "loss": 0.7021,
+    "grad_norm": 0.30535343289375305,
+    "learning_rate": 2.8174621626448827e-05,
+    "entropy": 0.7036702126264572,
+    "num_tokens": 56403176.0,
+    "mean_token_accuracy": 0.8093780845403671,
+    "epoch": 0.6533247256294383,
+    "step": 2530
+  },
+  {
+    "loss": 0.6855,
+    "grad_norm": 0.30807363986968994,
+    "learning_rate": 2.78008141881419e-05,
+    "entropy": 0.6923295110464096,
+    "num_tokens": 56630304.0,
+    "mean_token_accuracy": 0.8130094796419144,
+    "epoch": 0.6559070367979342,
+    "step": 2540
+  },
+  {
+    "loss": 0.6795,
+    "grad_norm": 0.2982047200202942,
+    "learning_rate": 2.7428545827983342e-05,
+    "entropy": 0.6848388686776161,
+    "num_tokens": 56866175.0,
+    "mean_token_accuracy": 0.8150498539209365,
+    "epoch": 0.65848934796643,
+    "step": 2550
+  },
+  {
+    "loss": 0.6908,
+    "grad_norm": 0.30639222264289856,
+    "learning_rate": 2.705784235548272e-05,
+    "entropy": 0.6895878106355667,
+    "num_tokens": 57092835.0,
+    "mean_token_accuracy": 0.8119323909282684,
+    "epoch": 0.6610716591349257,
+    "step": 2560
+  },
+  {
+    "loss": 0.7266,
+    "grad_norm": 0.3296997845172882,
+    "learning_rate": 2.6688729471655328e-05,
+    "entropy": 0.7319988161325455,
+    "num_tokens": 57300056.0,
+    "mean_token_accuracy": 0.8027160823345184,
+    "epoch": 0.6636539703034215,
+    "step": 2570
+  },
+  {
+    "loss": 0.6983,
+    "grad_norm": 0.2694888114929199,
+    "learning_rate": 2.632123276724031e-05,
+    "entropy": 0.7012559026479721,
+    "num_tokens": 57526144.0,
+    "mean_token_accuracy": 0.8083841979503632,
+    "epoch": 0.6662362814719174,
+    "step": 2580
+  },
+  {
+    "loss": 0.6938,
+    "grad_norm": 0.31541094183921814,
+    "learning_rate": 2.5955377720926512e-05,
+    "entropy": 0.6915117859840393,
+    "num_tokens": 57737388.0,
+    "mean_token_accuracy": 0.8134361296892166,
+    "epoch": 0.6688185926404132,
+    "step": 2590
+  },
+  {
+    "loss": 0.7445,
+    "grad_norm": 0.274829626083374,
+    "learning_rate": 2.559118969758595e-05,
+    "entropy": 0.7476611018180848,
+    "num_tokens": 57966900.0,
+    "mean_token_accuracy": 0.7994938224554062,
+    "epoch": 0.6714009038089089,
+    "step": 2600
+  },
+  {
+    "loss": 0.7141,
+    "grad_norm": 0.3201714754104614,
+    "learning_rate": 2.5228693946515302e-05,
+    "entropy": 0.7212372869253159,
+    "num_tokens": 58193368.0,
+    "mean_token_accuracy": 0.8060443252325058,
+    "epoch": 0.6739832149774048,
+    "step": 2610
+  },
+  {
+    "loss": 0.698,
+    "grad_norm": 0.29826799035072327,
+    "learning_rate": 2.486791559968536e-05,
+    "entropy": 0.7093008697032929,
+    "num_tokens": 58403598.0,
+    "mean_token_accuracy": 0.8120593637228012,
+    "epoch": 0.6765655261459006,
+    "step": 2620
+  },
+  {
+    "loss": 0.7146,
+    "grad_norm": 0.3138905465602875,
+    "learning_rate": 2.4508879669998614e-05,
+    "entropy": 0.7180470049381256,
+    "num_tokens": 58634383.0,
+    "mean_token_accuracy": 0.8049383610486984,
+    "epoch": 0.6791478373143964,
+    "step": 2630
+  },
+  {
+    "loss": 0.6845,
+    "grad_norm": 0.34878674149513245,
+    "learning_rate": 2.415161104955509e-05,
+    "entropy": 0.6881518125534057,
+    "num_tokens": 58858055.0,
+    "mean_token_accuracy": 0.8141413390636444,
+    "epoch": 0.6817301484828922,
+    "step": 2640
+  },
+  {
+    "loss": 0.745,
+    "grad_norm": 0.2723364233970642,
+    "learning_rate": 2.3796134507926594e-05,
+    "entropy": 0.7326867192983627,
+    "num_tokens": 59071166.0,
+    "mean_token_accuracy": 0.8036218643188476,
+    "epoch": 0.684312459651388,
+    "step": 2650
+  },
+  {
+    "loss": 0.6955,
+    "grad_norm": 0.2990936040878296,
+    "learning_rate": 2.3442474690439398e-05,
+    "entropy": 0.6928721308708191,
+    "num_tokens": 59279102.0,
+    "mean_token_accuracy": 0.8137792259454727,
+    "epoch": 0.6868947708198838,
+    "step": 2660
+  },
+  {
+    "loss": 0.7412,
+    "grad_norm": 0.3078380823135376,
+    "learning_rate": 2.30906561164656e-05,
+    "entropy": 0.737492860853672,
+    "num_tokens": 59493613.0,
+    "mean_token_accuracy": 0.803581839799881,
+    "epoch": 0.6894770819883796,
+    "step": 2670
+  },
+  {
+    "loss": 0.7322,
+    "grad_norm": 0.3019523024559021,
+    "learning_rate": 2.274070317772316e-05,
+    "entropy": 0.7335268735885621,
+    "num_tokens": 59705132.0,
+    "mean_token_accuracy": 0.8033424109220505,
+    "epoch": 0.6920593931568754,
+    "step": 2680
+  },
+  {
+    "loss": 0.7013,
+    "grad_norm": 0.2935568392276764,
+    "learning_rate": 2.2392640136584834e-05,
+    "entropy": 0.7151971668004989,
+    "num_tokens": 59914822.0,
+    "mean_token_accuracy": 0.8081181108951568,
+    "epoch": 0.6946417043253712,
+    "step": 2690
+  },
+  {
+    "loss": 0.6921,
+    "grad_norm": 0.3314759135246277,
+    "learning_rate": 2.204649112439604e-05,
+    "entropy": 0.6989267513155937,
+    "num_tokens": 60130448.0,
+    "mean_token_accuracy": 0.8114324450492859,
+    "epoch": 0.697224015493867,
+    "step": 2700
+  },
+  {
+    "loss": 0.7304,
+    "grad_norm": 0.2831194996833801,
+    "learning_rate": 2.1702280139801835e-05,
+    "entropy": 0.7311217278242111,
+    "num_tokens": 60356454.0,
+    "mean_token_accuracy": 0.8026905834674836,
+    "epoch": 0.6998063266623629,
+    "step": 2710
+  },
+  {
+    "loss": 0.7007,
+    "grad_norm": 0.33581551909446716,
+    "learning_rate": 2.1360031047083045e-05,
+    "entropy": 0.7018767982721329,
+    "num_tokens": 60563637.0,
+    "mean_token_accuracy": 0.8092585116624832,
+    "epoch": 0.7023886378308586,
+    "step": 2720
+  },
+  {
+    "loss": 0.724,
+    "grad_norm": 0.29153990745544434,
+    "learning_rate": 2.101976757450183e-05,
+    "entropy": 0.7301119029521942,
+    "num_tokens": 60791248.0,
+    "mean_token_accuracy": 0.8054573953151702,
+    "epoch": 0.7049709489993544,
+    "step": 2730
+  },
+  {
+    "loss": 0.7099,
+    "grad_norm": 0.3348991274833679,
+    "learning_rate": 2.0681513312656487e-05,
+    "entropy": 0.7101235866546631,
+    "num_tokens": 61012834.0,
+    "mean_token_accuracy": 0.8071693569421768,
+    "epoch": 0.7075532601678503,
+    "step": 2740
+  },
+  {
+    "loss": 0.7157,
+    "grad_norm": 0.33254438638687134,
+    "learning_rate": 2.0345291712845966e-05,
+    "entropy": 0.7233776718378067,
+    "num_tokens": 61240008.0,
+    "mean_token_accuracy": 0.804712375998497,
+    "epoch": 0.7101355713363461,
+    "step": 2750
+  },
+  {
+    "loss": 0.6452,
+    "grad_norm": 0.31784573197364807,
+    "learning_rate": 2.0011126085443978e-05,
+    "entropy": 0.6461900442838668,
+    "num_tokens": 61456216.0,
+    "mean_token_accuracy": 0.8218548208475113,
+    "epoch": 0.7127178825048418,
+    "step": 2760
+  },
+  {
+    "loss": 0.777,
+    "grad_norm": 0.3528957664966583,
+    "learning_rate": 1.9679039598282866e-05,
+    "entropy": 0.7807377189397812,
+    "num_tokens": 61671856.0,
+    "mean_token_accuracy": 0.7943974494934082,
+    "epoch": 0.7153001936733376,
+    "step": 2770
+  },
+  {
+    "loss": 0.6969,
+    "grad_norm": 0.27970144152641296,
+    "learning_rate": 1.9349055275047378e-05,
+    "entropy": 0.6933271333575248,
+    "num_tokens": 61901204.0,
+    "mean_token_accuracy": 0.8111155271530152,
+    "epoch": 0.7178825048418335,
+    "step": 2780
+  },
+  {
+    "loss": 0.7737,
+    "grad_norm": 0.2799535393714905,
+    "learning_rate": 1.9021195993678416e-05,
+    "entropy": 0.7860652267932892,
+    "num_tokens": 62115655.0,
+    "mean_token_accuracy": 0.7939832776784896,
+    "epoch": 0.7204648160103292,
+    "step": 2790
+  },
+  {
+    "loss": 0.6883,
+    "grad_norm": 0.2660212516784668,
+    "learning_rate": 1.8695484484786917e-05,
+    "entropy": 0.6971263945102691,
+    "num_tokens": 62338640.0,
+    "mean_token_accuracy": 0.8095773428678512,
+    "epoch": 0.723047127178825,
+    "step": 2800
+  },
+  {
+    "loss": 0.6817,
+    "grad_norm": 0.34406208992004395,
+    "learning_rate": 1.8371943330077918e-05,
+    "entropy": 0.6870502740144729,
+    "num_tokens": 62558349.0,
+    "mean_token_accuracy": 0.8148900002241135,
+    "epoch": 0.7256294383473209,
+    "step": 2810
+  },
+  {
+    "loss": 0.6707,
+    "grad_norm": 0.3069903552532196,
+    "learning_rate": 1.805059496078495e-05,
+    "entropy": 0.6698985725641251,
+    "num_tokens": 62784119.0,
+    "mean_token_accuracy": 0.8161512583494186,
+    "epoch": 0.7282117495158167,
+    "step": 2820
+  },
+  {
+    "loss": 0.6912,
+    "grad_norm": 0.3088005781173706,
+    "learning_rate": 1.7731461656114893e-05,
+    "entropy": 0.6937626540660858,
+    "num_tokens": 63025574.0,
+    "mean_token_accuracy": 0.8110652893781662,
+    "epoch": 0.7307940606843124,
+    "step": 2830
+  },
+  {
+    "loss": 0.7005,
+    "grad_norm": 0.2753876745700836,
+    "learning_rate": 1.7414565541703342e-05,
+    "entropy": 0.7088986724615097,
+    "num_tokens": 63241986.0,
+    "mean_token_accuracy": 0.8078723192214966,
+    "epoch": 0.7333763718528082,
+    "step": 2840
+  },
+  {
+    "loss": 0.7379,
+    "grad_norm": 0.27868911623954773,
+    "learning_rate": 1.7099928588080605e-05,
+    "entropy": 0.7435200735926628,
+    "num_tokens": 63465840.0,
+    "mean_token_accuracy": 0.8005112290382386,
+    "epoch": 0.7359586830213041,
+    "step": 2850
+  },
+  {
+    "loss": 0.6757,
+    "grad_norm": 0.31595101952552795,
+    "learning_rate": 1.6787572609148523e-05,
+    "entropy": 0.6833824813365936,
+    "num_tokens": 63695553.0,
+    "mean_token_accuracy": 0.8157312601804734,
+    "epoch": 0.7385409941897999,
+    "step": 2860
+  },
+  {
+    "loss": 0.7016,
+    "grad_norm": 0.2744700610637665,
+    "learning_rate": 1.6477519260668062e-05,
+    "entropy": 0.7068268299102783,
+    "num_tokens": 63922146.0,
+    "mean_token_accuracy": 0.8094365686178208,
+    "epoch": 0.7411233053582956,
+    "step": 2870
+  },
+  {
+    "loss": 0.7036,
+    "grad_norm": 0.29427847266197205,
+    "learning_rate": 1.6169790038757932e-05,
+    "entropy": 0.6996650636196137,
+    "num_tokens": 64140210.0,
+    "mean_token_accuracy": 0.8086274355649948,
+    "epoch": 0.7437056165267915,
+    "step": 2880
+  },
+  {
+    "loss": 0.6855,
+    "grad_norm": 0.318248987197876,
+    "learning_rate": 1.586440627840425e-05,
+    "entropy": 0.6926235496997833,
+    "num_tokens": 64367668.0,
+    "mean_token_accuracy": 0.8126532167196274,
+    "epoch": 0.7462879276952873,
+    "step": 2890
+  },
+  {
+    "loss": 0.6587,
+    "grad_norm": 0.26778993010520935,
+    "learning_rate": 1.5561389151981366e-05,
+    "entropy": 0.6617662817239761,
+    "num_tokens": 64596890.0,
+    "mean_token_accuracy": 0.8181241929531098,
+    "epoch": 0.7488702388637831,
+    "step": 2900
+  },
+  {
+    "loss": 0.7325,
+    "grad_norm": 0.27768194675445557,
+    "learning_rate": 1.5260759667783997e-05,
+    "entropy": 0.7412423819303513,
+    "num_tokens": 64806702.0,
+    "mean_token_accuracy": 0.8047612726688385,
+    "epoch": 0.7514525500322788,
+    "step": 2910
+  },
+  {
+    "loss": 0.6484,
+    "grad_norm": 0.2950485050678253,
+    "learning_rate": 1.4962538668570691e-05,
+    "entropy": 0.6552188992500305,
+    "num_tokens": 65032975.0,
+    "mean_token_accuracy": 0.8191455602645874,
+    "epoch": 0.7540348612007747,
+    "step": 2920
+  },
+  {
+    "loss": 0.7559,
+    "grad_norm": 0.305338591337204,
+    "learning_rate": 1.4666746830118804e-05,
+    "entropy": 0.7579466968774795,
+    "num_tokens": 65245187.0,
+    "mean_token_accuracy": 0.796593251824379,
+    "epoch": 0.7566171723692705,
+    "step": 2930
+  },
+  {
+    "loss": 0.7234,
+    "grad_norm": 0.29828450083732605,
+    "learning_rate": 1.4373404659791028e-05,
+    "entropy": 0.7265091866254807,
+    "num_tokens": 65478479.0,
+    "mean_token_accuracy": 0.8035373449325561,
+    "epoch": 0.7591994835377663,
+    "step": 2940
+  },
+  {
+    "loss": 0.658,
+    "grad_norm": 0.29163649678230286,
+    "learning_rate": 1.4082532495113626e-05,
+    "entropy": 0.6767950043082237,
+    "num_tokens": 65695750.0,
+    "mean_token_accuracy": 0.8175939649343491,
+    "epoch": 0.7617817947062621,
+    "step": 2950
+  },
+  {
+    "loss": 0.6907,
+    "grad_norm": 0.29603517055511475,
+    "learning_rate": 1.3794150502366404e-05,
+    "entropy": 0.6919728189706802,
+    "num_tokens": 65932520.0,
+    "mean_token_accuracy": 0.8117793321609497,
+    "epoch": 0.7643641058747579,
+    "step": 2960
+  },
+  {
+    "loss": 0.7528,
+    "grad_norm": 0.29973626136779785,
+    "learning_rate": 1.3508278675184593e-05,
+    "entropy": 0.7577356785535813,
+    "num_tokens": 66148085.0,
+    "mean_token_accuracy": 0.798463636636734,
+    "epoch": 0.7669464170432537,
+    "step": 2970
+  },
+  {
+    "loss": 0.6982,
+    "grad_norm": 0.3467027246952057,
+    "learning_rate": 1.3224936833172657e-05,
+    "entropy": 0.7075864315032959,
+    "num_tokens": 66350441.0,
+    "mean_token_accuracy": 0.8102767705917359,
+    "epoch": 0.7695287282117496,
+    "step": 2980
+  },
+  {
+    "loss": 0.7345,
+    "grad_norm": 0.3444955050945282,
+    "learning_rate": 1.2944144620530229e-05,
+    "entropy": 0.7401443183422088,
+    "num_tokens": 66574442.0,
+    "mean_token_accuracy": 0.802479374408722,
+    "epoch": 0.7721110393802453,
+    "step": 2990
+  },
+  {
+    "loss": 0.7497,
+    "grad_norm": 0.2776440382003784,
+    "learning_rate": 1.266592150469012e-05,
+    "entropy": 0.7472027570009232,
+    "num_tokens": 66801152.0,
+    "mean_token_accuracy": 0.7987334460020066,
+    "epoch": 0.7746933505487411,
+    "step": 3000
+  },
+  {
+    "loss": 0.6939,
+    "grad_norm": 0.27906349301338196,
+    "learning_rate": 1.2390286774968718e-05,
+    "entropy": 0.6968228965997696,
+    "num_tokens": 67029964.0,
+    "mean_token_accuracy": 0.8106195479631424,
+    "epoch": 0.7772756617172369,
+    "step": 3010
+  },
+  {
+    "loss": 0.69,
+    "grad_norm": 0.2762759029865265,
+    "learning_rate": 1.2117259541228549e-05,
+    "entropy": 0.6997145384550094,
+    "num_tokens": 67247542.0,
+    "mean_token_accuracy": 0.812323608994484,
+    "epoch": 0.7798579728857328,
+    "step": 3020
+  },
+  {
+    "loss": 0.7269,
+    "grad_norm": 0.3158010244369507,
+    "learning_rate": 1.1846858732553456e-05,
+    "entropy": 0.7227586776018142,
+    "num_tokens": 67461274.0,
+    "mean_token_accuracy": 0.8032592833042145,
+    "epoch": 0.7824402840542285,
+    "step": 3030
+  },
+  {
+    "loss": 0.7255,
+    "grad_norm": 0.31994199752807617,
+    "learning_rate": 1.1579103095936206e-05,
+    "entropy": 0.7367910444736481,
+    "num_tokens": 67672498.0,
+    "mean_token_accuracy": 0.8042957931756973,
+    "epoch": 0.7850225952227243,
+    "step": 3040
+  },
+  {
+    "loss": 0.7221,
+    "grad_norm": 0.3173619508743286,
+    "learning_rate": 1.1314011194978764e-05,
+    "entropy": 0.7295337229967117,
+    "num_tokens": 67896909.0,
+    "mean_token_accuracy": 0.8035631746053695,
+    "epoch": 0.7876049063912202,
+    "step": 3050
+  },
+  {
+    "loss": 0.7016,
+    "grad_norm": 0.26975125074386597,
+    "learning_rate": 1.1051601408605277e-05,
+    "entropy": 0.6950832739472389,
+    "num_tokens": 68127225.0,
+    "mean_token_accuracy": 0.8104336768388748,
+    "epoch": 0.790187217559716,
+    "step": 3060
+  },
+  {
+    "loss": 0.7297,
+    "grad_norm": 0.2895670235157013,
+    "learning_rate": 1.079189192978785e-05,
+    "entropy": 0.728104168176651,
+    "num_tokens": 68354043.0,
+    "mean_token_accuracy": 0.8038080394268036,
+    "epoch": 0.7927695287282117,
+    "step": 3070
+  },
+  {
+    "loss": 0.6808,
+    "grad_norm": 0.31210049986839294,
+    "learning_rate": 1.0534900764285206e-05,
+    "entropy": 0.6828968375921249,
+    "num_tokens": 68568009.0,
+    "mean_token_accuracy": 0.813372614979744,
+    "epoch": 0.7953518398967075,
+    "step": 3080
+  },
+  {
+    "loss": 0.7699,
+    "grad_norm": 0.3012009859085083,
+    "learning_rate": 1.0280645729394366e-05,
+    "entropy": 0.7736784100532532,
+    "num_tokens": 68780475.0,
+    "mean_token_accuracy": 0.7947412014007569,
+    "epoch": 0.7979341510652034,
+    "step": 3090
+  },
+  {
+    "loss": 0.7059,
+    "grad_norm": 0.2901022136211395,
+    "learning_rate": 1.002914445271535e-05,
+    "entropy": 0.7141468852758408,
+    "num_tokens": 68993221.0,
+    "mean_token_accuracy": 0.8091312497854233,
+    "epoch": 0.8005164622336992,
+    "step": 3100
+  },
+  {
+    "loss": 0.7216,
+    "grad_norm": 0.34709054231643677,
+    "learning_rate": 9.780414370929064e-06,
+    "entropy": 0.7236628875136375,
+    "num_tokens": 69217587.0,
+    "mean_token_accuracy": 0.806429061293602,
+    "epoch": 0.8030987734021949,
+    "step": 3110
+  },
+  {
+    "loss": 0.7232,
+    "grad_norm": 0.2870867848396301,
+    "learning_rate": 9.534472728588383e-06,
+    "entropy": 0.7272969245910644,
+    "num_tokens": 69443310.0,
+    "mean_token_accuracy": 0.8060679912567139,
+    "epoch": 0.8056810845706908,
+    "step": 3120
+  },
+  {
+    "loss": 0.7545,
+    "grad_norm": 0.25815922021865845,
+    "learning_rate": 9.291336576922616e-06,
+    "entropy": 0.7626336216926575,
+    "num_tokens": 69672892.0,
+    "mean_token_accuracy": 0.7972894191741944,
+    "epoch": 0.8082633957391866,
+    "step": 3130
+  },
+  {
+    "loss": 0.6876,
+    "grad_norm": 0.29713913798332214,
+    "learning_rate": 9.051022772655293e-06,
+    "entropy": 0.6986047714948654,
+    "num_tokens": 69890743.0,
+    "mean_token_accuracy": 0.8120619833469391,
+    "epoch": 0.8108457069076824,
+    "step": 3140
+  },
+  {
+    "loss": 0.7198,
+    "grad_norm": 0.2686369717121124,
+    "learning_rate": 8.813547976835528e-06,
+    "entropy": 0.7254200860857963,
+    "num_tokens": 70116053.0,
+    "mean_token_accuracy": 0.8052540808916092,
+    "epoch": 0.8134280180761781,
+    "step": 3150
+  },
+  {
+    "loss": 0.6943,
+    "grad_norm": 0.2916862368583679,
+    "learning_rate": 8.578928653682866e-06,
+    "entropy": 0.6961161434650421,
+    "num_tokens": 70329456.0,
+    "mean_token_accuracy": 0.8126390129327774,
+    "epoch": 0.816010329244674,
+    "step": 3160
+  },
+  {
+    "loss": 0.7393,
+    "grad_norm": 0.30221208930015564,
+    "learning_rate": 8.347181069445836e-06,
+    "entropy": 0.7420953810214996,
+    "num_tokens": 70553938.0,
+    "mean_token_accuracy": 0.7991205245256424,
+    "epoch": 0.8185926404131698,
+    "step": 3170
+  },
+  {
+    "loss": 0.7443,
+    "grad_norm": 0.30007004737854004,
+    "learning_rate": 8.118321291274184e-06,
+    "entropy": 0.7562640652060508,
+    "num_tokens": 70776813.0,
+    "mean_token_accuracy": 0.7976048231124878,
+    "epoch": 0.8211749515816655,
+    "step": 3180
+  },
+  {
+    "loss": 0.7273,
+    "grad_norm": 0.28486886620521545,
+    "learning_rate": 7.892365186104955e-06,
+    "entropy": 0.7327735960483551,
+    "num_tokens": 71001316.0,
+    "mean_token_accuracy": 0.8032251060009002,
+    "epoch": 0.8237572627501614,
+    "step": 3190
+  },
+  {
+    "loss": 0.673,
+    "grad_norm": 0.26758044958114624,
+    "learning_rate": 7.669328419562422e-06,
+    "entropy": 0.6777101784944535,
+    "num_tokens": 71225607.0,
+    "mean_token_accuracy": 0.8158749163150787,
+    "epoch": 0.8263395739186572,
+    "step": 3200
+  },
+  {
+    "loss": 0.6451,
+    "grad_norm": 0.31398946046829224,
+    "learning_rate": 7.4492264548719605e-06,
+    "entropy": 0.6516683280467988,
+    "num_tokens": 71460072.0,
+    "mean_token_accuracy": 0.8190358489751816,
+    "epoch": 0.828921885087153,
+    "step": 3210
+  },
+  {
+    "loss": 0.687,
+    "grad_norm": 0.28087955713272095,
+    "learning_rate": 7.2320745517880205e-06,
+    "entropy": 0.6877734363079071,
+    "num_tokens": 71691287.0,
+    "mean_token_accuracy": 0.8136042594909668,
+    "epoch": 0.8315041962556488,
+    "step": 3220
+  },
+  {
+    "loss": 0.766,
+    "grad_norm": 0.3055887818336487,
+    "learning_rate": 7.01788776553613e-06,
+    "entropy": 0.7661666601896286,
+    "num_tokens": 71915757.0,
+    "mean_token_accuracy": 0.7935220867395401,
+    "epoch": 0.8340865074241446,
+    "step": 3230
+  },
+  {
+    "loss": 0.6704,
+    "grad_norm": 0.26342907547950745,
+    "learning_rate": 6.806680945769112e-06,
+    "entropy": 0.6782629758119583,
+    "num_tokens": 72138609.0,
+    "mean_token_accuracy": 0.815047687292099,
+    "epoch": 0.8366688185926404,
+    "step": 3240
+  },
+  {
+    "loss": 0.6974,
+    "grad_norm": 0.30100300908088684,
+    "learning_rate": 6.598468735537566e-06,
+    "entropy": 0.7019865930080413,
+    "num_tokens": 72355904.0,
+    "mean_token_accuracy": 0.8122695237398148,
+    "epoch": 0.8392511297611362,
+    "step": 3250
+  },
+  {
+    "loss": 0.6812,
+    "grad_norm": 0.31231164932250977,
+    "learning_rate": 6.3932655702746594e-06,
+    "entropy": 0.6826063901185989,
+    "num_tokens": 72582185.0,
+    "mean_token_accuracy": 0.8142417192459106,
+    "epoch": 0.841833440929632,
+    "step": 3260
+  },
+  {
+    "loss": 0.739,
+    "grad_norm": 0.2941230535507202,
+    "learning_rate": 6.191085676795289e-06,
+    "entropy": 0.7482027873396874,
+    "num_tokens": 72786873.0,
+    "mean_token_accuracy": 0.800478795170784,
+    "epoch": 0.8444157520981278,
+    "step": 3270
+  },
+  {
+    "loss": 0.6933,
+    "grad_norm": 0.3058132529258728,
+    "learning_rate": 5.991943072309764e-06,
+    "entropy": 0.7009373277425766,
+    "num_tokens": 73007186.0,
+    "mean_token_accuracy": 0.8104920625686646,
+    "epoch": 0.8469980632666236,
+    "step": 3280
+  },
+  {
+    "loss": 0.691,
+    "grad_norm": 0.34394147992134094,
+    "learning_rate": 5.795851563451965e-06,
+    "entropy": 0.6961993753910065,
+    "num_tokens": 73221227.0,
+    "mean_token_accuracy": 0.8104782938957215,
+    "epoch": 0.8495803744351195,
+    "step": 3290
+  },
+  {
+    "loss": 0.6845,
+    "grad_norm": 0.27636125683784485,
+    "learning_rate": 5.602824745322127e-06,
+    "entropy": 0.6845636337995529,
+    "num_tokens": 73446914.0,
+    "mean_token_accuracy": 0.8126100540161133,
+    "epoch": 0.8521626856036152,
+    "step": 3300
+  },
+  {
+    "loss": 0.6588,
+    "grad_norm": 0.28646811842918396,
+    "learning_rate": 5.412876000544309e-06,
+    "entropy": 0.6653542622923851,
+    "num_tokens": 73665913.0,
+    "mean_token_accuracy": 0.8176908433437348,
+    "epoch": 0.854744996772111,
+    "step": 3310
+  },
+  {
+    "loss": 0.7334,
+    "grad_norm": 0.28045985102653503,
+    "learning_rate": 5.226018498338526e-06,
+    "entropy": 0.7391698479652404,
+    "num_tokens": 73893662.0,
+    "mean_token_accuracy": 0.8034716665744781,
+    "epoch": 0.8573273079406069,
+    "step": 3320
+  },
+  {
+    "loss": 0.6959,
+    "grad_norm": 0.27868810296058655,
+    "learning_rate": 5.042265193607765e-06,
+    "entropy": 0.7021783247590065,
+    "num_tokens": 74117461.0,
+    "mean_token_accuracy": 0.8107198417186737,
+    "epoch": 0.8599096191091027,
+    "step": 3330
+  },
+  {
+    "loss": 0.6257,
+    "grad_norm": 0.289795458316803,
+    "learning_rate": 4.861628826039799e-06,
+    "entropy": 0.6267055228352547,
+    "num_tokens": 74350458.0,
+    "mean_token_accuracy": 0.8263014882802964,
+    "epoch": 0.8624919302775984,
+    "step": 3340
+  },
+  {
+    "loss": 0.6857,
+    "grad_norm": 0.3118912875652313,
+    "learning_rate": 4.684121919223938e-06,
+    "entropy": 0.6892769366502762,
+    "num_tokens": 74570028.0,
+    "mean_token_accuracy": 0.8128671914339065,
+    "epoch": 0.8650742414460942,
+    "step": 3350
+  },
+  {
+    "loss": 0.7262,
+    "grad_norm": 0.2844097912311554,
+    "learning_rate": 4.509756779782765e-06,
+    "entropy": 0.7205952927470207,
+    "num_tokens": 74788292.0,
+    "mean_token_accuracy": 0.8064137637615204,
+    "epoch": 0.8676565526145901,
+    "step": 3360
+  },
+  {
+    "loss": 0.6939,
+    "grad_norm": 0.2834044098854065,
+    "learning_rate": 4.338545496518908e-06,
+    "entropy": 0.7010853558778762,
+    "num_tokens": 75015661.0,
+    "mean_token_accuracy": 0.8110615521669388,
+    "epoch": 0.8702388637830859,
+    "step": 3370
+  },
+  {
+    "loss": 0.6466,
+    "grad_norm": 0.31039345264434814,
+    "learning_rate": 4.17049993957695e-06,
+    "entropy": 0.6548168569803238,
+    "num_tokens": 75237714.0,
+    "mean_token_accuracy": 0.8208600997924804,
+    "epoch": 0.8728211749515816,
+    "step": 3380
+  },
+  {
+    "loss": 0.7192,
+    "grad_norm": 0.31068113446235657,
+    "learning_rate": 4.005631759620409e-06,
+    "entropy": 0.7178841561079026,
+    "num_tokens": 75466595.0,
+    "mean_token_accuracy": 0.8053244233131409,
+    "epoch": 0.8754034861200775,
+    "step": 3390
+  },
+  {
+    "loss": 0.7341,
+    "grad_norm": 0.2801123559474945,
+    "learning_rate": 3.8439523870240466e-06,
+    "entropy": 0.7397258043289184,
+    "num_tokens": 75685746.0,
+    "mean_token_accuracy": 0.8019078284502029,
+    "epoch": 0.8779857972885733,
+    "step": 3400
+  },
+  {
+    "loss": 0.749,
+    "grad_norm": 0.28393077850341797,
+    "learning_rate": 3.685473031081377e-06,
+    "entropy": 0.7531062722206116,
+    "num_tokens": 75891129.0,
+    "mean_token_accuracy": 0.799371811747551,
+    "epoch": 0.8805681084570691,
+    "step": 3410
+  },
+  {
+    "loss": 0.7069,
+    "grad_norm": 0.2816139757633209,
+    "learning_rate": 3.5302046792275057e-06,
+    "entropy": 0.7183081269264221,
+    "num_tokens": 76127939.0,
+    "mean_token_accuracy": 0.8063154727220535,
+    "epoch": 0.8831504196255648,
+    "step": 3420
+  },
+  {
+    "loss": 0.6473,
+    "grad_norm": 0.3396845757961273,
+    "learning_rate": 3.378158096277395e-06,
+    "entropy": 0.6580532863736153,
+    "num_tokens": 76356205.0,
+    "mean_token_accuracy": 0.8217125713825226,
+    "epoch": 0.8857327307940607,
+    "step": 3430
+  },
+  {
+    "loss": 0.6843,
+    "grad_norm": 0.3237053155899048,
+    "learning_rate": 3.2293438236795203e-06,
+    "entropy": 0.6855427086353302,
+    "num_tokens": 76583682.0,
+    "mean_token_accuracy": 0.814071336388588,
+    "epoch": 0.8883150419625565,
+    "step": 3440
+  },
+  {
+    "loss": 0.717,
+    "grad_norm": 0.24803553521633148,
+    "learning_rate": 3.083772178785016e-06,
+    "entropy": 0.7196818366646767,
+    "num_tokens": 76805913.0,
+    "mean_token_accuracy": 0.8079381316900254,
+    "epoch": 0.8908973531310523,
+    "step": 3450
+  },
+  {
+    "loss": 0.6765,
+    "grad_norm": 0.2812926173210144,
+    "learning_rate": 2.941453254132398e-06,
+    "entropy": 0.6827230915427208,
+    "num_tokens": 77042857.0,
+    "mean_token_accuracy": 0.8151951909065247,
+    "epoch": 0.8934796642995481,
+    "step": 3460
+  },
+  {
+    "loss": 0.7466,
+    "grad_norm": 0.25068199634552,
+    "learning_rate": 2.802396916747818e-06,
+    "entropy": 0.7477584034204483,
+    "num_tokens": 77274945.0,
+    "mean_token_accuracy": 0.7997568786144257,
+    "epoch": 0.8960619754680439,
+    "step": 3470
+  },
+  {
+    "loss": 0.7463,
+    "grad_norm": 0.29893818497657776,
+    "learning_rate": 2.666612807461e-06,
+    "entropy": 0.7611301898956299,
+    "num_tokens": 77507117.0,
+    "mean_token_accuracy": 0.7970719575881958,
+    "epoch": 0.8986442866365397,
+    "step": 3480
+  },
+  {
+    "loss": 0.734,
+    "grad_norm": 0.25104856491088867,
+    "learning_rate": 2.5341103402368116e-06,
+    "entropy": 0.7325819343328476,
+    "num_tokens": 77741888.0,
+    "mean_token_accuracy": 0.802145466208458,
+    "epoch": 0.9012265978050356,
+    "step": 3490
+  },
+  {
+    "loss": 0.7342,
+    "grad_norm": 0.2604316174983978,
+    "learning_rate": 2.404898701522612e-06,
+    "entropy": 0.7358214914798736,
+    "num_tokens": 77976461.0,
+    "mean_token_accuracy": 0.8025443255901337,
+    "epoch": 0.9038089089735313,
+    "step": 3500
+  },
+  {
+    "loss": 0.6718,
+    "grad_norm": 0.27755334973335266,
+    "learning_rate": 2.278986849611353e-06,
+    "entropy": 0.6820023268461227,
+    "num_tokens": 78199709.0,
+    "mean_token_accuracy": 0.8153169691562653,
+    "epoch": 0.9063912201420271,
+    "step": 3510
+  },
+  {
+    "loss": 0.7006,
+    "grad_norm": 0.27558284997940063,
+    "learning_rate": 2.1563835140204825e-06,
+    "entropy": 0.6964967548847198,
+    "num_tokens": 78427884.0,
+    "mean_token_accuracy": 0.8086929380893707,
+    "epoch": 0.9089735313105229,
+    "step": 3520
+  },
+  {
+    "loss": 0.6621,
+    "grad_norm": 0.29580724239349365,
+    "learning_rate": 2.037097194886728e-06,
+    "entropy": 0.6729359984397888,
+    "num_tokens": 78640676.0,
+    "mean_token_accuracy": 0.8171943426132202,
+    "epoch": 0.9115558424790188,
+    "step": 3530
+  },
+  {
+    "loss": 0.6216,
+    "grad_norm": 0.37162065505981445,
+    "learning_rate": 1.921136162376791e-06,
+    "entropy": 0.6259115815162659,
+    "num_tokens": 78854352.0,
+    "mean_token_accuracy": 0.8265722841024399,
+    "epoch": 0.9141381536475145,
+    "step": 3540
+  },
+  {
+    "loss": 0.7331,
+    "grad_norm": 0.2595006227493286,
+    "learning_rate": 1.8085084561139555e-06,
+    "entropy": 0.7356774628162384,
+    "num_tokens": 79083671.0,
+    "mean_token_accuracy": 0.8027292221784592,
+    "epoch": 0.9167204648160103,
+    "step": 3550
+  },
+  {
+    "loss": 0.7051,
+    "grad_norm": 0.2967183589935303,
+    "learning_rate": 1.6992218846207053e-06,
+    "entropy": 0.7063945472240448,
+    "num_tokens": 79299500.0,
+    "mean_token_accuracy": 0.8102407902479172,
+    "epoch": 0.9193027759845062,
+    "step": 3560
+  },
+  {
+    "loss": 0.6776,
+    "grad_norm": 0.31715014576911926,
+    "learning_rate": 1.593284024777364e-06,
+    "entropy": 0.678543546795845,
+    "num_tokens": 79516676.0,
+    "mean_token_accuracy": 0.81387257874012,
+    "epoch": 0.921885087153002,
+    "step": 3570
+  },
+  {
+    "loss": 0.6979,
+    "grad_norm": 0.2775455713272095,
+    "learning_rate": 1.4907022212967803e-06,
+    "entropy": 0.6995210900902749,
+    "num_tokens": 79753818.0,
+    "mean_token_accuracy": 0.8093364298343658,
+    "epoch": 0.9244673983214977,
+    "step": 3580
+  },
+  {
+    "loss": 0.6913,
+    "grad_norm": 0.3358924686908722,
+    "learning_rate": 1.3914835862151033e-06,
+    "entropy": 0.6959970772266388,
+    "num_tokens": 79987679.0,
+    "mean_token_accuracy": 0.8107724577188492,
+    "epoch": 0.9270497094899935,
+    "step": 3590
+  },
+  {
+    "loss": 0.721,
+    "grad_norm": 0.25691378116607666,
+    "learning_rate": 1.295634998398726e-06,
+    "entropy": 0.7186200335621834,
+    "num_tokens": 80219837.0,
+    "mean_token_accuracy": 0.8068506330251694,
+    "epoch": 0.9296320206584894,
+    "step": 3600
+  },
+  {
+    "loss": 0.7425,
+    "grad_norm": 0.291149765253067,
+    "learning_rate": 1.2031631030673618e-06,
+    "entropy": 0.7435696452856064,
+    "num_tokens": 80451637.0,
+    "mean_token_accuracy": 0.8018748521804809,
+    "epoch": 0.9322143318269851,
+    "step": 3610
+  },
+  {
+    "loss": 0.6562,
+    "grad_norm": 0.28476300835609436,
+    "learning_rate": 1.114074311333313e-06,
+    "entropy": 0.6657967358827591,
+    "num_tokens": 80672000.0,
+    "mean_token_accuracy": 0.818168243765831,
+    "epoch": 0.9347966429954809,
+    "step": 3620
+  },
+  {
+    "loss": 0.7104,
+    "grad_norm": 0.3295108377933502,
+    "learning_rate": 1.0283747997570092e-06,
+    "entropy": 0.7145670384168625,
+    "num_tokens": 80885606.0,
+    "mean_token_accuracy": 0.807201874256134,
+    "epoch": 0.9373789541639768,
+    "step": 3630
+  },
+  {
+    "loss": 0.7107,
+    "grad_norm": 0.28491243720054626,
+    "learning_rate": 9.460705099187783e-07,
+    "entropy": 0.7149154737591743,
+    "num_tokens": 81100462.0,
+    "mean_token_accuracy": 0.8080836564302445,
+    "epoch": 0.9399612653324726,
+    "step": 3640
+  },
+  {
+    "loss": 0.715,
+    "grad_norm": 0.3006097376346588,
+    "learning_rate": 8.671671480069032e-07,
+    "entropy": 0.7211218193173409,
+    "num_tokens": 81324223.0,
+    "mean_token_accuracy": 0.8062219977378845,
+    "epoch": 0.9425435765009683,
+    "step": 3650
+  },
+  {
+    "loss": 0.6908,
+    "grad_norm": 0.3136330544948578,
+    "learning_rate": 7.916701844220109e-07,
+    "entropy": 0.6916179537773133,
+    "num_tokens": 81546085.0,
+    "mean_token_accuracy": 0.8120699375867844,
+    "epoch": 0.9451258876694641,
+    "step": 3660
+  },
+  {
+    "loss": 0.7581,
+    "grad_norm": 0.30285340547561646,
+    "learning_rate": 7.195848533978311e-07,
+    "entropy": 0.7561731487512589,
+    "num_tokens": 81764441.0,
+    "mean_token_accuracy": 0.7981429368257522,
+    "epoch": 0.94770819883796,
+    "step": 3670
+  },
+  {
+    "loss": 0.7366,
+    "grad_norm": 0.26118963956832886,
+    "learning_rate": 6.509161526382701e-07,
+    "entropy": 0.7330646455287934,
+    "num_tokens": 81993927.0,
+    "mean_token_accuracy": 0.8043831527233124,
+    "epoch": 0.9502905100064558,
+    "step": 3680
+  },
+  {
+    "loss": 0.748,
+    "grad_norm": 0.27240875363349915,
+    "learning_rate": 5.856688429709434e-07,
+    "entropy": 0.7462060987949372,
+    "num_tokens": 82223112.0,
+    "mean_token_accuracy": 0.8016072034835815,
+    "epoch": 0.9528728211749515,
+    "step": 3690
+  },
+  {
+    "loss": 0.6935,
+    "grad_norm": 0.2734779715538025,
+    "learning_rate": 5.238474480170952e-07,
+    "entropy": 0.6943931892514229,
+    "num_tokens": 82455082.0,
+    "mean_token_accuracy": 0.8110814154148102,
+    "epoch": 0.9554551323434474,
+    "step": 3700
+  },
+  {
+    "loss": 0.7164,
+    "grad_norm": 0.3203067481517792,
+    "learning_rate": 4.6545625387798274e-07,
+    "entropy": 0.7151076972484589,
+    "num_tokens": 82671312.0,
+    "mean_token_accuracy": 0.8075156539678574,
+    "epoch": 0.9580374435119432,
+    "step": 3710
+  },
+  {
+    "loss": 0.6983,
+    "grad_norm": 0.2791406810283661,
+    "learning_rate": 4.104993088376974e-07,
+    "entropy": 0.7000140130519867,
+    "num_tokens": 82901834.0,
+    "mean_token_accuracy": 0.811114838719368,
+    "epoch": 0.960619754680439,
+    "step": 3720
+  },
+  {
+    "loss": 0.6859,
+    "grad_norm": 0.3430984318256378,
+    "learning_rate": 3.5898042308252224e-07,
+    "entropy": 0.6891811639070511,
+    "num_tokens": 83126676.0,
+    "mean_token_accuracy": 0.8148478597402573,
+    "epoch": 0.9632020658489348,
+    "step": 3730
+  },
+  {
+    "loss": 0.7634,
+    "grad_norm": 0.2839677035808563,
+    "learning_rate": 3.109031684367547e-07,
+    "entropy": 0.7797995641827583,
+    "num_tokens": 83330169.0,
+    "mean_token_accuracy": 0.7926018327474594,
+    "epoch": 0.9657843770174306,
+    "step": 3740
+  },
+  {
+    "loss": 0.6537,
+    "grad_norm": 0.2943638563156128,
+    "learning_rate": 2.662708781150769e-07,
+    "entropy": 0.6571737483143807,
+    "num_tokens": 83552644.0,
+    "mean_token_accuracy": 0.8210361778736115,
+    "epoch": 0.9683666881859264,
+    "step": 3750
+  },
+  {
+    "loss": 0.6883,
+    "grad_norm": 0.30458876490592957,
+    "learning_rate": 2.250866464914625e-07,
+    "entropy": 0.6957482188940048,
+    "num_tokens": 83768196.0,
+    "mean_token_accuracy": 0.8107382476329803,
+    "epoch": 0.9709489993544222,
+    "step": 3760
+  },
+  {
+    "loss": 0.6383,
+    "grad_norm": 0.3314257264137268,
+    "learning_rate": 1.8735332888463742e-07,
+    "entropy": 0.6453122898936272,
+    "num_tokens": 83990317.0,
+    "mean_token_accuracy": 0.8240663319826126,
+    "epoch": 0.973531310522918,
+    "step": 3770
+  },
+  {
+    "loss": 0.7142,
+    "grad_norm": 0.30556467175483704,
+    "learning_rate": 1.5307354136011586e-07,
+    "entropy": 0.7219583809375762,
+    "num_tokens": 84209863.0,
+    "mean_token_accuracy": 0.8086627215147019,
+    "epoch": 0.9761136216914138,
+    "step": 3780
+  },
+  {
+    "loss": 0.7143,
+    "grad_norm": 0.29667413234710693,
+    "learning_rate": 1.22249660548851e-07,
+    "entropy": 0.7240412354469299,
+    "num_tokens": 84447812.0,
+    "mean_token_accuracy": 0.8065494656562805,
+    "epoch": 0.9786959328599096,
+    "step": 3790
+  },
+  {
+    "loss": 0.7197,
+    "grad_norm": 0.2865903377532959,
+    "learning_rate": 9.48838234824223e-08,
+    "entropy": 0.7270282983779908,
+    "num_tokens": 84680310.0,
+    "mean_token_accuracy": 0.8049559354782104,
+    "epoch": 0.9812782440284055,
+    "step": 3800
+  },
+  {
+    "loss": 0.7106,
+    "grad_norm": 0.2443910390138626,
+    "learning_rate": 7.097792744492071e-08,
+    "entropy": 0.7112312227487564,
+    "num_tokens": 84909798.0,
+    "mean_token_accuracy": 0.8073030918836593,
+    "epoch": 0.9838605551969012,
+    "step": 3810
+  },
+  {
+    "loss": 0.7667,
+    "grad_norm": 0.3134123384952545,
+    "learning_rate": 5.053362984137611e-08,
+    "entropy": 0.765004500746727,
+    "num_tokens": 85131658.0,
+    "mean_token_accuracy": 0.7956638902425766,
+    "epoch": 0.986442866365397,
+    "step": 3820
+  },
+  {
+    "loss": 0.7177,
+    "grad_norm": 0.3139767646789551,
+    "learning_rate": 3.355234808284924e-08,
+    "entropy": 0.7268019631505013,
+    "num_tokens": 85351505.0,
+    "mean_token_accuracy": 0.8048588812351227,
+    "epoch": 0.9890251775338929,
+    "step": 3830
+  },
+  {
+    "loss": 0.7362,
+    "grad_norm": 0.2681005895137787,
+    "learning_rate": 2.0035259488182523e-08,
+    "entropy": 0.74400754570961,
+    "num_tokens": 85579956.0,
+    "mean_token_accuracy": 0.8011926084756851,
+    "epoch": 0.9916074887023887,
+    "step": 3840
+  },
+  {
+    "loss": 0.7149,
+    "grad_norm": 0.2973273992538452,
+    "learning_rate": 9.983301202365347e-09,
+    "entropy": 0.7227866470813751,
+    "num_tokens": 85807276.0,
+    "mean_token_accuracy": 0.8073938339948654,
+    "epoch": 0.9941897998708844,
+    "step": 3850
+  },
+  {
+    "loss": 0.7912,
+    "grad_norm": 0.3043336272239685,
+    "learning_rate": 3.3971701315638206e-09,
+    "entropy": 0.7958255261182785,
+    "num_tokens": 86031243.0,
+    "mean_token_accuracy": 0.7904368698596954,
+    "epoch": 0.9967721110393802,
+    "step": 3860
+  },
+  {
+    "loss": 0.7145,
+    "grad_norm": 0.3112486004829407,
+    "learning_rate": 2.7732289478166196e-10,
+    "entropy": 0.716927382349968,
+    "num_tokens": 86249568.0,
+    "mean_token_accuracy": 0.8073446452617645,
+    "epoch": 0.9993544222078761,
+    "step": 3870
+  },
+  {
+    "train_runtime": 49267.654,
+    "train_samples_per_second": 2.515,
+    "train_steps_per_second": 0.079,
+    "total_flos": 5.541551362900722e+19,
+    "train_loss": 0.74992161919958,
+    "entropy": 0.6862897753715516,
+    "num_tokens": 86293367.0,
+    "mean_token_accuracy": 0.8136373043060303,
+    "epoch": 1.0,
+    "step": 3873
+  }
+]
\ No newline at end of file