diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4004 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 3971,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.6497728377580643,
+      "epoch": 0.002518653779554828,
+      "grad_norm": 0.6953125,
+      "learning_rate": 4.988667841853438e-05,
+      "loss": 1.1745,
+      "mean_token_accuracy": 0.7103757172822952,
+      "num_tokens": 927987.0,
+      "step": 10
+    },
+    {
+      "entropy": 1.5930162012577056,
+      "epoch": 0.005037307559109656,
+      "grad_norm": 0.625,
+      "learning_rate": 4.9760765550239234e-05,
+      "loss": 1.1032,
+      "mean_token_accuracy": 0.7197690099477768,
+      "num_tokens": 1859988.0,
+      "step": 20
+    },
+    {
+      "entropy": 1.5709561467170716,
+      "epoch": 0.0075559613386644835,
+      "grad_norm": 0.640625,
+      "learning_rate": 4.96348526819441e-05,
+      "loss": 1.085,
+      "mean_token_accuracy": 0.7227144911885262,
+      "num_tokens": 2800856.0,
+      "step": 30
+    },
+    {
+      "entropy": 1.566073715686798,
+      "epoch": 0.010074615118219311,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.9508939813648956e-05,
+      "loss": 1.0955,
+      "mean_token_accuracy": 0.7216884844005108,
+      "num_tokens": 3739770.0,
+      "step": 40
+    },
+    {
+      "entropy": 1.570231558382511,
+      "epoch": 0.01259326889777414,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.938302694535382e-05,
+      "loss": 1.1137,
+      "mean_token_accuracy": 0.7198784925043583,
+      "num_tokens": 4662394.0,
+      "step": 50
+    },
+    {
+      "entropy": 1.5752479314804078,
+      "epoch": 0.015111922677328967,
+      "grad_norm": 0.67578125,
+      "learning_rate": 4.925711407705868e-05,
+      "loss": 1.1067,
+      "mean_token_accuracy": 0.7184384018182755,
+      "num_tokens": 5592065.0,
+      "step": 60
+    },
+    {
+      "entropy": 1.5744277507066726,
+      "epoch": 0.017630576456883797,
+      "grad_norm": 0.58984375,
+      "learning_rate": 4.913120120876354e-05,
+      "loss": 1.0994,
+      "mean_token_accuracy": 0.7210577458143235,
+      "num_tokens": 6494850.0,
+      "step": 70
+    },
+    {
+      "entropy": 1.5574209734797477,
+      "epoch": 0.020149230236438623,
+      "grad_norm": 0.63671875,
+      "learning_rate": 4.90052883404684e-05,
+      "loss": 1.0789,
+      "mean_token_accuracy": 0.7261395655572415,
+      "num_tokens": 7402054.0,
+      "step": 80
+    },
+    {
+      "entropy": 1.536399021744728,
+      "epoch": 0.022667884015993452,
+      "grad_norm": 0.64453125,
+      "learning_rate": 4.887937547217326e-05,
+      "loss": 1.0834,
+      "mean_token_accuracy": 0.7243645377457142,
+      "num_tokens": 8302725.0,
+      "step": 90
+    },
+    {
+      "entropy": 1.5392677783966064,
+      "epoch": 0.02518653779554828,
+      "grad_norm": 0.63671875,
+      "learning_rate": 4.875346260387812e-05,
+      "loss": 1.0579,
+      "mean_token_accuracy": 0.7294937998056412,
+      "num_tokens": 9218531.0,
+      "step": 100
+    },
+    {
+      "entropy": 1.5304515838623047,
+      "epoch": 0.027705191575103108,
+      "grad_norm": 0.62109375,
+      "learning_rate": 4.862754973558298e-05,
+      "loss": 1.0607,
+      "mean_token_accuracy": 0.7303975030779839,
+      "num_tokens": 10136124.0,
+      "step": 110
+    },
+    {
+      "entropy": 1.5561267986893654,
+      "epoch": 0.030223845354657934,
+      "grad_norm": 0.62109375,
+      "learning_rate": 4.8501636867287845e-05,
+      "loss": 1.0511,
+      "mean_token_accuracy": 0.7304783284664154,
+      "num_tokens": 11083595.0,
+      "step": 120
+    },
+    {
+      "entropy": 1.5530587568879128,
+      "epoch": 0.03274249913421276,
+      "grad_norm": 0.609375,
+      "learning_rate": 4.83757239989927e-05,
+      "loss": 1.1037,
+      "mean_token_accuracy": 0.7202336438000202,
+      "num_tokens": 11990008.0,
+      "step": 130
+    },
+    {
+      "entropy": 1.5585834830999374,
+      "epoch": 0.03526115291376759,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.824981113069756e-05,
+      "loss": 1.103,
+      "mean_token_accuracy": 0.7219732508063317,
+      "num_tokens": 12902256.0,
+      "step": 140
+    },
+    {
+      "entropy": 1.5430516287684442,
+      "epoch": 0.03777980669332242,
+      "grad_norm": 0.62890625,
+      "learning_rate": 4.812389826240242e-05,
+      "loss": 1.0711,
+      "mean_token_accuracy": 0.7261470817029476,
+      "num_tokens": 13855059.0,
+      "step": 150
+    },
+    {
+      "entropy": 1.5312249392271042,
+      "epoch": 0.040298460472877246,
+      "grad_norm": 0.65625,
+      "learning_rate": 4.799798539410728e-05,
+      "loss": 1.0753,
+      "mean_token_accuracy": 0.7257871173322201,
+      "num_tokens": 14779771.0,
+      "step": 160
+    },
+    {
+      "entropy": 1.5230548575520515,
+      "epoch": 0.04281711425243207,
+      "grad_norm": 0.6015625,
+      "learning_rate": 4.7872072525812137e-05,
+      "loss": 1.0393,
+      "mean_token_accuracy": 0.735567494481802,
+      "num_tokens": 15690913.0,
+      "step": 170
+    },
+    {
+      "entropy": 1.5378432273864746,
+      "epoch": 0.045335768031986905,
+      "grad_norm": 0.64453125,
+      "learning_rate": 4.7746159657517e-05,
+      "loss": 1.0876,
+      "mean_token_accuracy": 0.7245238177478314,
+      "num_tokens": 16633243.0,
+      "step": 180
+    },
+    {
+      "entropy": 1.5589430660009385,
+      "epoch": 0.04785442181154173,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.762024678922186e-05,
+      "loss": 1.0532,
+      "mean_token_accuracy": 0.7320627510547638,
+      "num_tokens": 17553351.0,
+      "step": 190
+    },
+    {
+      "entropy": 1.5157382100820542,
+      "epoch": 0.05037307559109656,
+      "grad_norm": 0.63671875,
+      "learning_rate": 4.749433392092672e-05,
+      "loss": 1.0358,
+      "mean_token_accuracy": 0.731314767897129,
+      "num_tokens": 18490557.0,
+      "step": 200
+    },
+    {
+      "entropy": 1.5197158992290496,
+      "epoch": 0.05289172937065139,
+      "grad_norm": 0.6171875,
+      "learning_rate": 4.736842105263158e-05,
+      "loss": 1.0799,
+      "mean_token_accuracy": 0.7257929719984532,
+      "num_tokens": 19433634.0,
+      "step": 210
+    },
+    {
+      "entropy": 1.5323385715484619,
+      "epoch": 0.055410383150206216,
+      "grad_norm": 0.67578125,
+      "learning_rate": 4.724250818433644e-05,
+      "loss": 1.0491,
+      "mean_token_accuracy": 0.7314196087419986,
+      "num_tokens": 20340871.0,
+      "step": 220
+    },
+    {
+      "entropy": 1.5352398797869682,
+      "epoch": 0.05792903692976104,
+      "grad_norm": 0.58984375,
+      "learning_rate": 4.71165953160413e-05,
+      "loss": 1.0442,
+      "mean_token_accuracy": 0.7328816823661327,
+      "num_tokens": 21256808.0,
+      "step": 230
+    },
+    {
+      "entropy": 1.5354523420333863,
+      "epoch": 0.06044769070931587,
+      "grad_norm": 0.6484375,
+      "learning_rate": 4.699068244774616e-05,
+      "loss": 1.073,
+      "mean_token_accuracy": 0.7256046369671821,
+      "num_tokens": 22171600.0,
+      "step": 240
+    },
+    {
+      "entropy": 1.5628203928470612,
+      "epoch": 0.0629663444888707,
+      "grad_norm": 0.58984375,
+      "learning_rate": 4.6864769579451025e-05,
+      "loss": 1.1204,
+      "mean_token_accuracy": 0.7178477592766285,
+      "num_tokens": 23116116.0,
+      "step": 250
+    },
+    {
+      "entropy": 1.5106180116534234,
+      "epoch": 0.06548499826842552,
+      "grad_norm": 0.62890625,
+      "learning_rate": 4.673885671115588e-05,
+      "loss": 1.0505,
+      "mean_token_accuracy": 0.7322093218564987,
+      "num_tokens": 24022009.0,
+      "step": 260
+    },
+    {
+      "entropy": 1.5455271020531653,
+      "epoch": 0.06800365204798035,
+      "grad_norm": 0.640625,
+      "learning_rate": 4.661294384286074e-05,
+      "loss": 1.1028,
+      "mean_token_accuracy": 0.7194493912160397,
+      "num_tokens": 24938320.0,
+      "step": 270
+    },
+    {
+      "entropy": 1.4870401427149773,
+      "epoch": 0.07052230582753519,
+      "grad_norm": 0.625,
+      "learning_rate": 4.64870309745656e-05,
+      "loss": 1.0062,
+      "mean_token_accuracy": 0.7405259512364865,
+      "num_tokens": 25872274.0,
+      "step": 280
+    },
+    {
+      "entropy": 1.5163322016596794,
+      "epoch": 0.07304095960709,
+      "grad_norm": 0.59765625,
+      "learning_rate": 4.636111810627046e-05,
+      "loss": 1.0475,
+      "mean_token_accuracy": 0.7297493070363998,
+      "num_tokens": 26787336.0,
+      "step": 290
+    },
+    {
+      "entropy": 1.5091008096933365,
+      "epoch": 0.07555961338664484,
+      "grad_norm": 0.5546875,
+      "learning_rate": 4.6235205237975324e-05,
+      "loss": 1.038,
+      "mean_token_accuracy": 0.7351058594882488,
+      "num_tokens": 27725039.0,
+      "step": 300
+    },
+    {
+      "entropy": 1.535538549721241,
+      "epoch": 0.07807826716619967,
+      "grad_norm": 0.63671875,
+      "learning_rate": 4.6109292369680185e-05,
+      "loss": 1.0576,
+      "mean_token_accuracy": 0.7299107275903225,
+      "num_tokens": 28643618.0,
+      "step": 310
+    },
+    {
+      "entropy": 1.5314571484923363,
+      "epoch": 0.08059692094575449,
+      "grad_norm": 0.640625,
+      "learning_rate": 4.5983379501385046e-05,
+      "loss": 1.0666,
+      "mean_token_accuracy": 0.7284630224108696,
+      "num_tokens": 29563205.0,
+      "step": 320
+    },
+    {
+      "entropy": 1.5281456634402275,
+      "epoch": 0.08311557472530932,
+      "grad_norm": 0.65234375,
+      "learning_rate": 4.58574666330899e-05,
+      "loss": 1.0655,
+      "mean_token_accuracy": 0.7297011129558086,
+      "num_tokens": 30480512.0,
+      "step": 330
+    },
+    {
+      "entropy": 1.5301422774791718,
+      "epoch": 0.08563422850486414,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.573155376479477e-05,
+      "loss": 1.0605,
+      "mean_token_accuracy": 0.7305384628474713,
+      "num_tokens": 31442273.0,
+      "step": 340
+    },
+    {
+      "entropy": 1.4844171449542045,
+      "epoch": 0.08815288228441898,
+      "grad_norm": 0.62109375,
+      "learning_rate": 4.560564089649962e-05,
+      "loss": 1.0442,
+      "mean_token_accuracy": 0.7323296993970871,
+      "num_tokens": 32344410.0,
+      "step": 350
+    },
+    {
+      "entropy": 1.5229773551225663,
+      "epoch": 0.09067153606397381,
+      "grad_norm": 0.6015625,
+      "learning_rate": 4.5479728028204484e-05,
+      "loss": 1.0332,
+      "mean_token_accuracy": 0.7339819706976414,
+      "num_tokens": 33250917.0,
+      "step": 360
+    },
+    {
+      "entropy": 1.5061728432774544,
+      "epoch": 0.09319018984352863,
+      "grad_norm": 0.6328125,
+      "learning_rate": 4.5353815159909345e-05,
+      "loss": 1.0196,
+      "mean_token_accuracy": 0.7394672840833664,
+      "num_tokens": 34175027.0,
+      "step": 370
+    },
+    {
+      "entropy": 1.5456583075225354,
+      "epoch": 0.09570884362308346,
+      "grad_norm": 0.640625,
+      "learning_rate": 4.5227902291614206e-05,
+      "loss": 1.0619,
+      "mean_token_accuracy": 0.7283155068755149,
+      "num_tokens": 35109724.0,
+      "step": 380
+    },
+    {
+      "entropy": 1.5447079360485076,
+      "epoch": 0.0982274974026383,
+      "grad_norm": 0.62109375,
+      "learning_rate": 4.510198942331906e-05,
+      "loss": 1.0899,
+      "mean_token_accuracy": 0.7229381762444973,
+      "num_tokens": 36062415.0,
+      "step": 390
+    },
+    {
+      "entropy": 1.533456189930439,
+      "epoch": 0.10074615118219311,
+      "grad_norm": 0.65625,
+      "learning_rate": 4.497607655502393e-05,
+      "loss": 1.0645,
+      "mean_token_accuracy": 0.7263024888932705,
+      "num_tokens": 37007526.0,
+      "step": 400
+    },
+    {
+      "entropy": 1.522585578262806,
+      "epoch": 0.10326480496174795,
+      "grad_norm": 0.6171875,
+      "learning_rate": 4.485016368672878e-05,
+      "loss": 1.0556,
+      "mean_token_accuracy": 0.7289167314767837,
+      "num_tokens": 37935164.0,
+      "step": 410
+    },
+    {
+      "entropy": 1.4904729962348937,
+      "epoch": 0.10578345874130278,
+      "grad_norm": 0.66796875,
+      "learning_rate": 4.4724250818433643e-05,
+      "loss": 1.0408,
+      "mean_token_accuracy": 0.7305458515882493,
+      "num_tokens": 38872935.0,
+      "step": 420
+    },
+    {
+      "entropy": 1.51086837798357,
+      "epoch": 0.1083021125208576,
+      "grad_norm": 0.59765625,
+      "learning_rate": 4.459833795013851e-05,
+      "loss": 1.0184,
+      "mean_token_accuracy": 0.7377710983157157,
+      "num_tokens": 39800020.0,
+      "step": 430
+    },
+    {
+      "entropy": 1.527511714398861,
+      "epoch": 0.11082076630041243,
+      "grad_norm": 0.62890625,
+      "learning_rate": 4.4472425081843366e-05,
+      "loss": 1.0671,
+      "mean_token_accuracy": 0.728612695634365,
+      "num_tokens": 40736503.0,
+      "step": 440
+    },
+    {
+      "entropy": 1.5698168337345124,
+      "epoch": 0.11333942007996725,
+      "grad_norm": 0.60546875,
+      "learning_rate": 4.434651221354823e-05,
+      "loss": 1.0474,
+      "mean_token_accuracy": 0.7324717722833156,
+      "num_tokens": 41666603.0,
+      "step": 450
+    },
+    {
+      "entropy": 1.5131051123142243,
+      "epoch": 0.11585807385952208,
+      "grad_norm": 0.625,
+      "learning_rate": 4.422059934525309e-05,
+      "loss": 1.0566,
+      "mean_token_accuracy": 0.7301515676081181,
+      "num_tokens": 42598424.0,
+      "step": 460
+    },
+    {
+      "entropy": 1.50111276358366,
+      "epoch": 0.11837672763907692,
+      "grad_norm": 0.578125,
+      "learning_rate": 4.409468647695795e-05,
+      "loss": 1.0339,
+      "mean_token_accuracy": 0.7336028017103672,
+      "num_tokens": 43559521.0,
+      "step": 470
+    },
+    {
+      "entropy": 1.4837490528821946,
+      "epoch": 0.12089538141863174,
+      "grad_norm": 0.59765625,
+      "learning_rate": 4.39687736086628e-05,
+      "loss": 0.9844,
+      "mean_token_accuracy": 0.7455279111862183,
+      "num_tokens": 44494185.0,
+      "step": 480
+    },
+    {
+      "entropy": 1.5296257749199866,
+      "epoch": 0.12341403519818657,
+      "grad_norm": 0.58203125,
+      "learning_rate": 4.384286074036767e-05,
+      "loss": 1.0266,
+      "mean_token_accuracy": 0.7367306031286717,
+      "num_tokens": 45420386.0,
+      "step": 490
+    },
+    {
+      "entropy": 1.5529645070433618,
+      "epoch": 0.1259326889777414,
+      "grad_norm": 0.609375,
+      "learning_rate": 4.3716947872072525e-05,
+      "loss": 1.0647,
+      "mean_token_accuracy": 0.7286850355565548,
+      "num_tokens": 46328398.0,
+      "step": 500
+    },
+    {
+      "entropy": 1.5489887461066245,
+      "epoch": 0.12845134275729622,
+      "grad_norm": 0.62890625,
+      "learning_rate": 4.3591035003777386e-05,
+      "loss": 1.0497,
+      "mean_token_accuracy": 0.7323333226144314,
+      "num_tokens": 47237802.0,
+      "step": 510
+    },
+    {
+      "entropy": 1.5361493453383446,
+      "epoch": 0.13096999653685104,
+      "grad_norm": 0.59375,
+      "learning_rate": 4.346512213548225e-05,
+      "loss": 1.0247,
+      "mean_token_accuracy": 0.7348957560956478,
+      "num_tokens": 48176536.0,
+      "step": 520
+    },
+    {
+      "entropy": 1.5109021827578544,
+      "epoch": 0.1334886503164059,
+      "grad_norm": 0.58203125,
+      "learning_rate": 4.333920926718711e-05,
+      "loss": 1.0204,
+      "mean_token_accuracy": 0.7367621898651123,
+      "num_tokens": 49109015.0,
+      "step": 530
+    },
+    {
+      "entropy": 1.4852696403861045,
+      "epoch": 0.1360073040959607,
+      "grad_norm": 0.5859375,
+      "learning_rate": 4.321329639889197e-05,
+      "loss": 1.0364,
+      "mean_token_accuracy": 0.7326985791325569,
+      "num_tokens": 50040438.0,
+      "step": 540
+    },
+    {
+      "entropy": 1.5346143633127212,
+      "epoch": 0.13852595787551553,
+      "grad_norm": 0.65625,
+      "learning_rate": 4.308738353059683e-05,
+      "loss": 1.0536,
+      "mean_token_accuracy": 0.7286248430609703,
+      "num_tokens": 50976370.0,
+      "step": 550
+    },
+    {
+      "entropy": 1.514106061309576,
+      "epoch": 0.14104461165507037,
+      "grad_norm": 0.69921875,
+      "learning_rate": 4.296147066230169e-05,
+      "loss": 1.041,
+      "mean_token_accuracy": 0.7335773780941963,
+      "num_tokens": 51918138.0,
+      "step": 560
+    },
+    {
+      "entropy": 1.5336113646626472,
+      "epoch": 0.1435632654346252,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.2835557794006546e-05,
+      "loss": 1.0518,
+      "mean_token_accuracy": 0.7317119717597962,
+      "num_tokens": 52817229.0,
+      "step": 570
+    },
+    {
+      "entropy": 1.5196146205067635,
+      "epoch": 0.14608191921418,
+      "grad_norm": 0.59765625,
+      "learning_rate": 4.2709644925711414e-05,
+      "loss": 1.0267,
+      "mean_token_accuracy": 0.736695246398449,
+      "num_tokens": 53760765.0,
+      "step": 580
+    },
+    {
+      "entropy": 1.5179338246583938,
+      "epoch": 0.14860057299373486,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.258373205741627e-05,
+      "loss": 1.0466,
+      "mean_token_accuracy": 0.7297343403100968,
+      "num_tokens": 54671519.0,
+      "step": 590
+    },
+    {
+      "entropy": 1.5304696060717107,
+      "epoch": 0.15111922677328968,
+      "grad_norm": 0.62890625,
+      "learning_rate": 4.245781918912113e-05,
+      "loss": 1.0458,
+      "mean_token_accuracy": 0.7315973825752735,
+      "num_tokens": 55593071.0,
+      "step": 600
+    },
+    {
+      "entropy": 1.4869036689400672,
+      "epoch": 0.1536378805528445,
+      "grad_norm": 0.6640625,
+      "learning_rate": 4.233190632082599e-05,
+      "loss": 1.0077,
+      "mean_token_accuracy": 0.7418606728315353,
+      "num_tokens": 56514341.0,
+      "step": 610
+    },
+    {
+      "entropy": 1.4934701904654504,
+      "epoch": 0.15615653433239934,
+      "grad_norm": 0.578125,
+      "learning_rate": 4.220599345253085e-05,
+      "loss": 1.0083,
+      "mean_token_accuracy": 0.7406033039093017,
+      "num_tokens": 57462853.0,
+      "step": 620
+    },
+    {
+      "entropy": 1.5260894536972045,
+      "epoch": 0.15867518811195416,
+      "grad_norm": 0.63671875,
+      "learning_rate": 4.2080080584235706e-05,
+      "loss": 1.052,
+      "mean_token_accuracy": 0.7272073470056057,
+      "num_tokens": 58377529.0,
+      "step": 630
+    },
+    {
+      "entropy": 1.548849606513977,
+      "epoch": 0.16119384189150898,
+      "grad_norm": 0.609375,
+      "learning_rate": 4.1954167715940574e-05,
+      "loss": 1.0904,
+      "mean_token_accuracy": 0.724496615678072,
+      "num_tokens": 59304012.0,
+      "step": 640
+    },
+    {
+      "entropy": 1.5230202600359917,
+      "epoch": 0.16371249567106383,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.1828254847645435e-05,
+      "loss": 1.0317,
+      "mean_token_accuracy": 0.7327270671725273,
+      "num_tokens": 60226985.0,
+      "step": 650
+    },
+    {
+      "entropy": 1.4835097789764404,
+      "epoch": 0.16623114945061865,
+      "grad_norm": 0.60546875,
+      "learning_rate": 4.170234197935029e-05,
+      "loss": 1.0267,
+      "mean_token_accuracy": 0.7380013577640057,
+      "num_tokens": 61164672.0,
+      "step": 660
+    },
+    {
+      "entropy": 1.5426082447171212,
+      "epoch": 0.16874980323017347,
+      "grad_norm": 0.56640625,
+      "learning_rate": 4.157642911105516e-05,
+      "loss": 1.0308,
+      "mean_token_accuracy": 0.7336142487823963,
+      "num_tokens": 62097480.0,
+      "step": 670
+    },
+    {
+      "entropy": 1.5340310111641884,
+      "epoch": 0.1712684570097283,
+      "grad_norm": 0.609375,
+      "learning_rate": 4.145051624276001e-05,
+      "loss": 1.072,
+      "mean_token_accuracy": 0.7251117169857025,
+      "num_tokens": 63026840.0,
+      "step": 680
+    },
+    {
+      "entropy": 1.500159528851509,
+      "epoch": 0.17378711078928313,
+      "grad_norm": 0.6328125,
+      "learning_rate": 4.132460337446487e-05,
+      "loss": 1.0297,
+      "mean_token_accuracy": 0.7330615803599357,
+      "num_tokens": 63949950.0,
+      "step": 690
+    },
+    {
+      "entropy": 1.5545173302292823,
+      "epoch": 0.17630576456883795,
+      "grad_norm": 0.57421875,
+      "learning_rate": 4.1198690506169734e-05,
+      "loss": 1.0545,
+      "mean_token_accuracy": 0.7317077338695526,
+      "num_tokens": 64876440.0,
+      "step": 700
+    },
+    {
+      "entropy": 1.5059973791241645,
+      "epoch": 0.17882441834839277,
+      "grad_norm": 0.58203125,
+      "learning_rate": 4.1072777637874595e-05,
+      "loss": 1.01,
+      "mean_token_accuracy": 0.7392976120114326,
+      "num_tokens": 65798419.0,
+      "step": 710
+    },
+    {
+      "entropy": 1.5090464890003203,
+      "epoch": 0.18134307212794762,
+      "grad_norm": 0.60546875,
+      "learning_rate": 4.094686476957945e-05,
+      "loss": 1.0592,
+      "mean_token_accuracy": 0.7292688339948654,
+      "num_tokens": 66749039.0,
+      "step": 720
+    },
+    {
+      "entropy": 1.5077268376946449,
+      "epoch": 0.18386172590750244,
+      "grad_norm": 0.66015625,
+      "learning_rate": 4.082095190128432e-05,
+      "loss": 1.0422,
+      "mean_token_accuracy": 0.7338137224316597,
+      "num_tokens": 67710358.0,
+      "step": 730
+    },
+    {
+      "entropy": 1.5227509185671806,
+      "epoch": 0.18638037968705726,
+      "grad_norm": 0.60546875,
+      "learning_rate": 4.069503903298917e-05,
+      "loss": 1.0484,
+      "mean_token_accuracy": 0.7323349170386791,
+      "num_tokens": 68624639.0,
+      "step": 740
+    },
+    {
+      "entropy": 1.4950523242354392,
+      "epoch": 0.1888990334666121,
+      "grad_norm": 0.609375,
+      "learning_rate": 4.056912616469403e-05,
+      "loss": 1.0318,
+      "mean_token_accuracy": 0.7348136067390442,
+      "num_tokens": 69559409.0,
+      "step": 750
+    },
+    {
+      "entropy": 1.4976729974150658,
+      "epoch": 0.19141768724616692,
+      "grad_norm": 0.61328125,
+      "learning_rate": 4.044321329639889e-05,
+      "loss": 1.0152,
+      "mean_token_accuracy": 0.7377664625644684,
+      "num_tokens": 70498976.0,
+      "step": 760
+    },
+    {
+      "entropy": 1.5174943268299104,
+      "epoch": 0.19393634102572174,
+      "grad_norm": 0.58984375,
+      "learning_rate": 4.0317300428103754e-05,
+      "loss": 1.0128,
+      "mean_token_accuracy": 0.7391440957784653,
+      "num_tokens": 71449612.0,
+      "step": 770
+    },
+    {
+      "entropy": 1.5164562672376634,
+      "epoch": 0.1964549948052766,
+      "grad_norm": 0.671875,
+      "learning_rate": 4.0191387559808616e-05,
+      "loss": 0.9954,
+      "mean_token_accuracy": 0.7393902830779553,
+      "num_tokens": 72373886.0,
+      "step": 780
+    },
+    {
+      "entropy": 1.4641567483544349,
+      "epoch": 0.1989736485848314,
+      "grad_norm": 0.58203125,
+      "learning_rate": 4.0065474691513477e-05,
+      "loss": 0.9909,
+      "mean_token_accuracy": 0.7428550243377685,
+      "num_tokens": 73312402.0,
+      "step": 790
+    },
+    {
+      "entropy": 1.507477380335331,
+      "epoch": 0.20149230236438623,
+      "grad_norm": 0.578125,
+      "learning_rate": 3.993956182321834e-05,
+      "loss": 1.0209,
+      "mean_token_accuracy": 0.7389532752335072,
+      "num_tokens": 74255599.0,
+      "step": 800
+    },
+    {
+      "entropy": 1.4874264925718308,
+      "epoch": 0.20401095614394107,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.981364895492319e-05,
+      "loss": 1.0178,
+      "mean_token_accuracy": 0.7369325928390026,
+      "num_tokens": 75173757.0,
+      "step": 810
+    },
+    {
+      "entropy": 1.4983343809843064,
+      "epoch": 0.2065296099234959,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.968773608662806e-05,
+      "loss": 1.0488,
+      "mean_token_accuracy": 0.7323521085083484,
+      "num_tokens": 76090812.0,
+      "step": 820
+    },
+    {
+      "entropy": 1.5155110970139503,
+      "epoch": 0.2090482637030507,
+      "grad_norm": 0.64453125,
+      "learning_rate": 3.9561823218332914e-05,
+      "loss": 1.0699,
+      "mean_token_accuracy": 0.7274161614477634,
+      "num_tokens": 77029951.0,
+      "step": 830
+    },
+    {
+      "entropy": 1.4780758187174796,
+      "epoch": 0.21156691748260556,
+      "grad_norm": 0.625,
+      "learning_rate": 3.9435910350037775e-05,
+      "loss": 1.0051,
+      "mean_token_accuracy": 0.7377786092460156,
+      "num_tokens": 77960751.0,
+      "step": 840
+    },
+    {
+      "entropy": 1.4968378961086273,
+      "epoch": 0.21408557126216038,
+      "grad_norm": 0.6015625,
+      "learning_rate": 3.9309997481742636e-05,
+      "loss": 1.0027,
+      "mean_token_accuracy": 0.739619717001915,
+      "num_tokens": 78871850.0,
+      "step": 850
+    },
+    {
+      "entropy": 1.5046116828918457,
+      "epoch": 0.2166042250417152,
+      "grad_norm": 0.59375,
+      "learning_rate": 3.91840846134475e-05,
+      "loss": 1.0261,
+      "mean_token_accuracy": 0.7376008428633213,
+      "num_tokens": 79789618.0,
+      "step": 860
+    },
+    {
+      "entropy": 1.5208716675639153,
+      "epoch": 0.21912287882127002,
+      "grad_norm": 0.609375,
+      "learning_rate": 3.905817174515236e-05,
+      "loss": 1.051,
+      "mean_token_accuracy": 0.7320379175245761,
+      "num_tokens": 80728419.0,
+      "step": 870
+    },
+    {
+      "entropy": 1.5159122616052627,
+      "epoch": 0.22164153260082486,
+      "grad_norm": 0.60546875,
+      "learning_rate": 3.893225887685722e-05,
+      "loss": 1.0243,
+      "mean_token_accuracy": 0.7333739779889583,
+      "num_tokens": 81673574.0,
+      "step": 880
+    },
+    {
+      "entropy": 1.51417468637228,
+      "epoch": 0.22416018638037968,
+      "grad_norm": 0.59765625,
+      "learning_rate": 3.880634600856208e-05,
+      "loss": 1.0413,
+      "mean_token_accuracy": 0.7333143278956413,
+      "num_tokens": 82609636.0,
+      "step": 890
+    },
+    {
+      "entropy": 1.4946632251143455,
+      "epoch": 0.2266788401599345,
+      "grad_norm": 0.56640625,
+      "learning_rate": 3.8680433140266935e-05,
+      "loss": 1.0272,
+      "mean_token_accuracy": 0.7353226900100708,
+      "num_tokens": 83566048.0,
+      "step": 900
+    },
+    {
+      "entropy": 1.4703357517719269,
+      "epoch": 0.22919749393948935,
+      "grad_norm": 0.609375,
+      "learning_rate": 3.8554520271971796e-05,
+      "loss": 1.0239,
+      "mean_token_accuracy": 0.7346428163349629,
+      "num_tokens": 84490230.0,
+      "step": 910
+    },
+    {
+      "entropy": 1.5199193447828292,
+      "epoch": 0.23171614771904417,
+      "grad_norm": 0.64453125,
+      "learning_rate": 3.842860740367666e-05,
+      "loss": 1.0443,
+      "mean_token_accuracy": 0.7323003455996513,
+      "num_tokens": 85420861.0,
+      "step": 920
+    },
+    {
+      "entropy": 1.52715914696455,
+      "epoch": 0.234234801498599,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.830269453538152e-05,
+      "loss": 1.0432,
+      "mean_token_accuracy": 0.7316192977130413,
+      "num_tokens": 86352852.0,
+      "step": 930
+    },
+    {
+      "entropy": 1.4817651391029358,
+      "epoch": 0.23675345527815383,
+      "grad_norm": 0.6328125,
+      "learning_rate": 3.817678166708637e-05,
+      "loss": 1.0019,
+      "mean_token_accuracy": 0.7411134377121925,
+      "num_tokens": 87249992.0,
+      "step": 940
+    },
+    {
+      "entropy": 1.4907526612281798,
+      "epoch": 0.23927210905770865,
+      "grad_norm": 0.59765625,
+      "learning_rate": 3.805086879879124e-05,
+      "loss": 1.0013,
+      "mean_token_accuracy": 0.7385142982006073,
+      "num_tokens": 88185598.0,
+      "step": 950
+    },
+    {
+      "entropy": 1.5223437398672104,
+      "epoch": 0.24179076283726347,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.7924955930496095e-05,
+      "loss": 1.0563,
+      "mean_token_accuracy": 0.7298605613410473,
+      "num_tokens": 89109791.0,
+      "step": 960
+    },
+    {
+      "entropy": 1.486008982360363,
+      "epoch": 0.24430941661681832,
+      "grad_norm": 0.56640625,
+      "learning_rate": 3.7799043062200956e-05,
+      "loss": 1.0103,
+      "mean_token_accuracy": 0.737246710062027,
+      "num_tokens": 90058609.0,
+      "step": 970
+    },
+    {
+      "entropy": 1.512269377708435,
+      "epoch": 0.24682807039637314,
+      "grad_norm": 0.59375,
+      "learning_rate": 3.767313019390582e-05,
+      "loss": 1.0284,
+      "mean_token_accuracy": 0.7351725302636624,
+      "num_tokens": 90982493.0,
+      "step": 980
+    },
+    {
+      "entropy": 1.5016800329089164,
+      "epoch": 0.24934672417592796,
+      "grad_norm": 0.6171875,
+      "learning_rate": 3.754721732561068e-05,
+      "loss": 1.0298,
+      "mean_token_accuracy": 0.7350243054330349,
+      "num_tokens": 91919533.0,
+      "step": 990
+    },
+    {
+      "entropy": 1.4913472577929496,
+      "epoch": 0.2518653779554828,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.742130445731554e-05,
+      "loss": 0.992,
+      "mean_token_accuracy": 0.7438635163009166,
+      "num_tokens": 92854895.0,
+      "step": 1000
+    },
+    {
+      "entropy": 1.4977505892515182,
+      "epoch": 0.2543840317350376,
+      "grad_norm": 0.58984375,
+      "learning_rate": 3.72953915890204e-05,
+      "loss": 1.0344,
+      "mean_token_accuracy": 0.7339183062314987,
+      "num_tokens": 93790271.0,
+      "step": 1010
+    },
+    {
+      "entropy": 1.5159107878804208,
+      "epoch": 0.25690268551459244,
+      "grad_norm": 0.64453125,
+      "learning_rate": 3.716947872072526e-05,
+      "loss": 1.0492,
+      "mean_token_accuracy": 0.7301998473703861,
+      "num_tokens": 94686152.0,
+      "step": 1020
+    },
+    {
+      "entropy": 1.5307456016540528,
+      "epoch": 0.2594213392941473,
+      "grad_norm": 0.60546875,
+      "learning_rate": 3.7043565852430116e-05,
+      "loss": 1.0364,
+      "mean_token_accuracy": 0.7321190066635609,
+      "num_tokens": 95608968.0,
+      "step": 1030
+    },
+    {
+      "entropy": 1.4946022421121596,
+      "epoch": 0.2619399930737021,
+      "grad_norm": 0.63671875,
+      "learning_rate": 3.6917652984134983e-05,
+      "loss": 1.0151,
+      "mean_token_accuracy": 0.7384577453136444,
+      "num_tokens": 96528931.0,
+      "step": 1040
+    },
+    {
+      "entropy": 1.4895016327500343,
+      "epoch": 0.26445864685325693,
+      "grad_norm": 0.65234375,
+      "learning_rate": 3.679174011583984e-05,
+      "loss": 1.0396,
+      "mean_token_accuracy": 0.7310051701962947,
+      "num_tokens": 97431432.0,
+      "step": 1050
+    },
+    {
+      "entropy": 1.5223950266838073,
+      "epoch": 0.2669773006328118,
+      "grad_norm": 0.6328125,
+      "learning_rate": 3.66658272475447e-05,
+      "loss": 1.0245,
+      "mean_token_accuracy": 0.7353874854743481,
+      "num_tokens": 98357299.0,
+      "step": 1060
+    },
+    {
+      "entropy": 1.502314467728138,
+      "epoch": 0.26949595441236657,
+      "grad_norm": 0.671875,
+      "learning_rate": 3.653991437924956e-05,
+      "loss": 1.0212,
+      "mean_token_accuracy": 0.7371881239116191,
+      "num_tokens": 99287298.0,
+      "step": 1070
+    },
+    {
+      "entropy": 1.487211187183857,
+      "epoch": 0.2720146081919214,
+      "grad_norm": 0.66015625,
+      "learning_rate": 3.641400151095442e-05,
+      "loss": 1.018,
+      "mean_token_accuracy": 0.7396972067654133,
+      "num_tokens": 100201770.0,
+      "step": 1080
+    },
+    {
+      "entropy": 1.4796535983681678,
+      "epoch": 0.27453326197147626,
+      "grad_norm": 0.609375,
+      "learning_rate": 3.628808864265928e-05,
+      "loss": 1.0203,
+      "mean_token_accuracy": 0.73761862590909,
+      "num_tokens": 101133473.0,
+      "step": 1090
+    },
+    {
+      "entropy": 1.5198125943541527,
+      "epoch": 0.27705191575103105,
+      "grad_norm": 0.5625,
+      "learning_rate": 3.616217577436414e-05,
+      "loss": 1.0277,
+      "mean_token_accuracy": 0.7344238288700581,
+      "num_tokens": 102079400.0,
+      "step": 1100
+    },
+    {
+      "entropy": 1.4751672700047493,
+      "epoch": 0.2795705695305859,
+      "grad_norm": 0.59765625,
+      "learning_rate": 3.6036262906069004e-05,
+      "loss": 0.992,
+      "mean_token_accuracy": 0.7427970752120018,
+      "num_tokens": 103002746.0,
+      "step": 1110
+    },
+    {
+      "entropy": 1.4785253152251243,
+      "epoch": 0.28208922331014075,
+      "grad_norm": 0.60546875,
+      "learning_rate": 3.591035003777386e-05,
+      "loss": 0.9957,
+      "mean_token_accuracy": 0.7420700334012509,
+      "num_tokens": 103918330.0,
+      "step": 1120
+    },
+    {
+      "entropy": 1.5064472571015357,
+      "epoch": 0.28460787708969554,
+      "grad_norm": 0.6015625,
+      "learning_rate": 3.5784437169478727e-05,
+      "loss": 1.0121,
+      "mean_token_accuracy": 0.7389346733689308,
+      "num_tokens": 104854940.0,
+      "step": 1130
+    },
+    {
+      "entropy": 1.5104423373937608,
+      "epoch": 0.2871265308692504,
+      "grad_norm": 0.68359375,
+      "learning_rate": 3.565852430118358e-05,
+      "loss": 1.0577,
+      "mean_token_accuracy": 0.7281268581748008,
+      "num_tokens": 105757398.0,
+      "step": 1140
+    },
+    {
+      "entropy": 1.4889473721385003,
+      "epoch": 0.28964518464880523,
+      "grad_norm": 0.58984375,
+      "learning_rate": 3.553261143288844e-05,
+      "loss": 1.0116,
+      "mean_token_accuracy": 0.7378810577094554,
+      "num_tokens": 106709345.0,
+      "step": 1150
+    },
+    {
+      "entropy": 1.4943735346198082,
+      "epoch": 0.29216383842836,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.54066985645933e-05,
+      "loss": 1.0252,
+      "mean_token_accuracy": 0.7358876734972,
+      "num_tokens": 107635433.0,
+      "step": 1160
+    },
+    {
+      "entropy": 1.502584198117256,
+      "epoch": 0.29468249220791487,
+      "grad_norm": 0.60546875,
+      "learning_rate": 3.5280785696298164e-05,
+      "loss": 1.0513,
+      "mean_token_accuracy": 0.7284682184457779,
+      "num_tokens": 108555011.0,
+      "step": 1170
+    },
+    {
+      "entropy": 1.4951962947845459,
+      "epoch": 0.2972011459874697,
+      "grad_norm": 0.5859375,
+      "learning_rate": 3.515487282800302e-05,
+      "loss": 1.0058,
+      "mean_token_accuracy": 0.7395492434501648,
+      "num_tokens": 109490480.0,
+      "step": 1180
+    },
+    {
+      "entropy": 1.4874925374984742,
+      "epoch": 0.2997197997670245,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.5028959959707886e-05,
+      "loss": 0.996,
+      "mean_token_accuracy": 0.7426116019487381,
+      "num_tokens": 110419525.0,
+      "step": 1190
+    },
+    {
+      "entropy": 1.4831220969557761,
+      "epoch": 0.30223845354657936,
+      "grad_norm": 0.5625,
+      "learning_rate": 3.490304709141274e-05,
+      "loss": 1.0094,
+      "mean_token_accuracy": 0.7389338947832584,
+      "num_tokens": 111349186.0,
+      "step": 1200
+    },
+    {
+      "entropy": 1.4941108256578446,
+      "epoch": 0.3047571073261342,
+      "grad_norm": 0.63671875,
+      "learning_rate": 3.47771342231176e-05,
+      "loss": 1.0472,
+      "mean_token_accuracy": 0.7327656015753746,
+      "num_tokens": 112282566.0,
+      "step": 1210
+    },
+    {
+      "entropy": 1.5015305012464524,
+      "epoch": 0.307275761105689,
+      "grad_norm": 0.63671875,
+      "learning_rate": 3.465122135482247e-05,
+      "loss": 1.0308,
+      "mean_token_accuracy": 0.73332554474473,
+      "num_tokens": 113202119.0,
+      "step": 1220
+    },
+    {
+      "entropy": 1.4797065630555153,
+      "epoch": 0.30979441488524384,
+      "grad_norm": 0.5703125,
+      "learning_rate": 3.4525308486527324e-05,
+      "loss": 1.037,
+      "mean_token_accuracy": 0.7331598915159703,
+      "num_tokens": 114138832.0,
+      "step": 1230
+    },
+    {
+      "entropy": 1.5040980368852614,
+      "epoch": 0.3123130686647987,
+      "grad_norm": 0.59375,
+      "learning_rate": 3.4399395618232185e-05,
+      "loss": 1.0141,
+      "mean_token_accuracy": 0.7366115286946296,
+      "num_tokens": 115039255.0,
+      "step": 1240
+    },
+    {
+      "entropy": 1.491380912065506,
+      "epoch": 0.3148317224443535,
+      "grad_norm": 0.6328125,
+      "learning_rate": 3.4273482749937046e-05,
+      "loss": 1.0256,
+      "mean_token_accuracy": 0.7349315457046032,
+      "num_tokens": 115940525.0,
+      "step": 1250
+    },
+    {
+      "entropy": 1.474766517430544,
+      "epoch": 0.3173503762239083,
+      "grad_norm": 0.6015625,
+      "learning_rate": 3.414756988164191e-05,
+      "loss": 1.0051,
+      "mean_token_accuracy": 0.7393092684447765,
+      "num_tokens": 116859333.0,
+      "step": 1260
+    },
+    {
+      "entropy": 1.5051109313964843,
+      "epoch": 0.3198690300034632,
+      "grad_norm": 0.57421875,
+      "learning_rate": 3.402165701334676e-05,
+      "loss": 1.016,
+      "mean_token_accuracy": 0.7380836345255375,
+      "num_tokens": 117776423.0,
+      "step": 1270
+    },
+    {
+      "entropy": 1.4819408521056174,
+      "epoch": 0.32238768378301796,
+      "grad_norm": 0.578125,
+      "learning_rate": 3.389574414505163e-05,
+      "loss": 0.991,
+      "mean_token_accuracy": 0.7428291216492653,
+      "num_tokens": 118716777.0,
+      "step": 1280
+    },
+    {
+      "entropy": 1.4922457024455071,
+      "epoch": 0.3249063375625728,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.3769831276756484e-05,
+      "loss": 1.0247,
+      "mean_token_accuracy": 0.7364048793911934,
+      "num_tokens": 119626650.0,
+      "step": 1290
+    },
+    {
+      "entropy": 1.4869423642754556,
+      "epoch": 0.32742499134212766,
+      "grad_norm": 0.62109375,
+      "learning_rate": 3.3643918408461345e-05,
+      "loss": 1.0159,
+      "mean_token_accuracy": 0.7364346459507942,
+      "num_tokens": 120559670.0,
+      "step": 1300
+    },
+    {
+      "entropy": 1.5164387181401253,
+      "epoch": 0.32994364512168245,
+      "grad_norm": 0.6171875,
+      "learning_rate": 3.3518005540166206e-05,
+      "loss": 1.0452,
+      "mean_token_accuracy": 0.7311400182545185,
+      "num_tokens": 121489609.0,
+      "step": 1310
+    },
+    {
+      "entropy": 1.4925027176737786,
+      "epoch": 0.3324622989012373,
+      "grad_norm": 0.58984375,
+      "learning_rate": 3.339209267187107e-05,
+      "loss": 1.0261,
+      "mean_token_accuracy": 0.7373234689235687,
+      "num_tokens": 122416454.0,
+      "step": 1320
+    },
+    {
+      "entropy": 1.507451492547989,
+      "epoch": 0.33498095268079214,
+      "grad_norm": 0.5859375,
+      "learning_rate": 3.326617980357593e-05,
+      "loss": 1.0082,
+      "mean_token_accuracy": 0.7384895570576191,
+      "num_tokens": 123342649.0,
+      "step": 1330
+    },
+    {
+      "entropy": 1.4801776722073554,
+      "epoch": 0.33749960646034693,
+      "grad_norm": 0.60546875,
+      "learning_rate": 3.314026693528079e-05,
+      "loss": 1.026,
+      "mean_token_accuracy": 0.734679427742958,
+      "num_tokens": 124277565.0,
+      "step": 1340
+    },
+    {
+      "entropy": 1.5057601898908615,
+      "epoch": 0.3400182602399018,
+      "grad_norm": 0.6328125,
+      "learning_rate": 3.301435406698565e-05,
+      "loss": 1.0532,
+      "mean_token_accuracy": 0.7300509728491307,
+      "num_tokens": 125193040.0,
+      "step": 1350
+    },
+    {
+      "entropy": 1.504169335961342,
+      "epoch": 0.3425369140194566,
+      "grad_norm": 0.6484375,
+      "learning_rate": 3.2888441198690504e-05,
+      "loss": 1.0018,
+      "mean_token_accuracy": 0.7397960677742959,
+      "num_tokens": 126119671.0,
+      "step": 1360
+    },
+    {
+      "entropy": 1.495669847726822,
+      "epoch": 0.3450555677990114,
+      "grad_norm": 0.62109375,
+      "learning_rate": 3.276252833039537e-05,
+      "loss": 1.0296,
+      "mean_token_accuracy": 0.7364421933889389,
+      "num_tokens": 127044727.0,
+      "step": 1370
+    },
+    {
+      "entropy": 1.4692699402570724,
+      "epoch": 0.34757422157856627,
+      "grad_norm": 0.58984375,
+      "learning_rate": 3.263661546210023e-05,
+      "loss": 0.972,
+      "mean_token_accuracy": 0.7507402546703815,
+      "num_tokens": 127984703.0,
+      "step": 1380
+    },
+    {
+      "entropy": 1.5226725831627845,
+      "epoch": 0.35009287535812106,
+      "grad_norm": 0.62109375,
+      "learning_rate": 3.251070259380509e-05,
+      "loss": 1.0447,
+      "mean_token_accuracy": 0.7286883890628815,
+      "num_tokens": 128914516.0,
+      "step": 1390
+    },
+    {
+      "entropy": 1.4715677112340928,
+      "epoch": 0.3526115291376759,
+      "grad_norm": 0.59765625,
+      "learning_rate": 3.238478972550995e-05,
+      "loss": 0.9847,
+      "mean_token_accuracy": 0.7421014286577702,
+      "num_tokens": 129828449.0,
+      "step": 1400
+    },
+    {
+      "entropy": 1.4714261531829833,
+      "epoch": 0.35513018291723075,
+      "grad_norm": 0.59765625,
+      "learning_rate": 3.225887685721481e-05,
+      "loss": 1.0038,
+      "mean_token_accuracy": 0.7429524183273315,
+      "num_tokens": 130786715.0,
+      "step": 1410
+    },
+    {
+      "entropy": 1.485981273651123,
+      "epoch": 0.35764883669678554,
+      "grad_norm": 0.56640625,
+      "learning_rate": 3.213296398891967e-05,
+      "loss": 1.0028,
+      "mean_token_accuracy": 0.7397522330284119,
+      "num_tokens": 131737266.0,
+      "step": 1420
+    },
+    {
+      "entropy": 1.5126312345266342,
+      "epoch": 0.3601674904763404,
+      "grad_norm": 0.58984375,
+      "learning_rate": 3.200705112062453e-05,
+      "loss": 1.0257,
+      "mean_token_accuracy": 0.7358055524528027,
+      "num_tokens": 132679586.0,
+      "step": 1430
+    },
+    {
+      "entropy": 1.5042190700769424,
+      "epoch": 0.36268614425589524,
+      "grad_norm": 0.6640625,
+      "learning_rate": 3.188113825232939e-05,
+      "loss": 1.0099,
+      "mean_token_accuracy": 0.7401847161352635,
+      "num_tokens": 133603713.0,
+      "step": 1440
+    },
+    {
+      "entropy": 1.4730105847120285,
+      "epoch": 0.36520479803545003,
+      "grad_norm": 0.62890625,
+      "learning_rate": 3.175522538403425e-05,
+      "loss": 1.0513,
+      "mean_token_accuracy": 0.7313306540250778,
+      "num_tokens": 134522232.0,
+      "step": 1450
+    },
+    {
+      "entropy": 1.4745427146553993,
+      "epoch": 0.3677234518150049,
+      "grad_norm": 0.578125,
+      "learning_rate": 3.1629312515739115e-05,
+      "loss": 1.028,
+      "mean_token_accuracy": 0.7363212339580059,
+      "num_tokens": 135466122.0,
+      "step": 1460
+    },
+    {
+      "entropy": 1.4661547526717187,
+      "epoch": 0.3702421055945597,
+      "grad_norm": 0.57421875,
+      "learning_rate": 3.150339964744397e-05,
+      "loss": 1.0065,
+      "mean_token_accuracy": 0.7393469184637069,
+      "num_tokens": 136411704.0,
+      "step": 1470
+    },
+    {
+      "entropy": 1.4818128019571304,
+      "epoch": 0.3727607593741145,
+      "grad_norm": 0.58984375,
+      "learning_rate": 3.137748677914883e-05,
+      "loss": 1.005,
+      "mean_token_accuracy": 0.7411336719989776,
+      "num_tokens": 137343925.0,
+      "step": 1480
+    },
+    {
+      "entropy": 1.5323064669966697,
+      "epoch": 0.37527941315366936,
+      "grad_norm": 0.6953125,
+      "learning_rate": 3.125157391085369e-05,
+      "loss": 1.0397,
+      "mean_token_accuracy": 0.7332744777202607,
+      "num_tokens": 138258013.0,
+      "step": 1490
+    },
+    {
+      "entropy": 1.4970730185508727,
+      "epoch": 0.3777980669332242,
+      "grad_norm": 0.59375,
+      "learning_rate": 3.112566104255855e-05,
+      "loss": 1.0267,
+      "mean_token_accuracy": 0.7344107307493687,
+      "num_tokens": 139191604.0,
+      "step": 1500
+    },
+    {
+      "entropy": 1.5285832583904266,
+      "epoch": 0.380316720712779,
+      "grad_norm": 0.62109375,
+      "learning_rate": 3.099974817426341e-05,
+      "loss": 1.0279,
+      "mean_token_accuracy": 0.7337069801986218,
+      "num_tokens": 140119182.0,
+      "step": 1510
+    },
+    {
+      "entropy": 1.4828715667128562,
+      "epoch": 0.38283537449233385,
+      "grad_norm": 0.609375,
+      "learning_rate": 3.0873835305968275e-05,
+      "loss": 0.98,
+      "mean_token_accuracy": 0.742748036235571,
+      "num_tokens": 141028056.0,
+      "step": 1520
+    },
+    {
+      "entropy": 1.4884025305509567,
+      "epoch": 0.3853540282718887,
+      "grad_norm": 0.55078125,
+      "learning_rate": 3.074792243767313e-05,
+      "loss": 1.0096,
+      "mean_token_accuracy": 0.740665141493082,
+      "num_tokens": 141973431.0,
+      "step": 1530
+    },
+    {
+      "entropy": 1.5160500556230545,
+      "epoch": 0.3878726820514435,
+      "grad_norm": 0.60546875,
+      "learning_rate": 3.062200956937799e-05,
+      "loss": 1.0196,
+      "mean_token_accuracy": 0.7365249536931515,
+      "num_tokens": 142929741.0,
+      "step": 1540
+    },
+    {
+      "entropy": 1.5132767364382744,
+      "epoch": 0.39039133583099833,
+      "grad_norm": 0.6875,
+      "learning_rate": 3.0496096701082855e-05,
+      "loss": 0.9694,
+      "mean_token_accuracy": 0.7474554844200612,
+      "num_tokens": 143858628.0,
+      "step": 1550
+    },
+    {
+      "entropy": 1.4559171989560127,
+      "epoch": 0.3929099896105532,
+      "grad_norm": 0.56640625,
+      "learning_rate": 3.0370183832787713e-05,
+      "loss": 0.991,
+      "mean_token_accuracy": 0.7405735231935978,
+      "num_tokens": 144791852.0,
+      "step": 1560
+    },
+    {
+      "entropy": 1.4635022155940534,
+      "epoch": 0.39542864339010797,
+      "grad_norm": 0.625,
+      "learning_rate": 3.024427096449257e-05,
+      "loss": 0.9773,
+      "mean_token_accuracy": 0.7441258721053601,
+      "num_tokens": 145700233.0,
+      "step": 1570
+    },
+    {
+      "entropy": 1.4716689318418503,
+      "epoch": 0.3979472971696628,
+      "grad_norm": 0.5859375,
+      "learning_rate": 3.0118358096197435e-05,
+      "loss": 0.9752,
+      "mean_token_accuracy": 0.7468914896249771,
+      "num_tokens": 146628030.0,
+      "step": 1580
+    },
+    {
+      "entropy": 1.4879262149333954,
+      "epoch": 0.40046595094921766,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.9992445227902293e-05,
+      "loss": 0.9802,
+      "mean_token_accuracy": 0.7438114926218986,
+      "num_tokens": 147555920.0,
+      "step": 1590
+    },
+    {
+      "entropy": 1.479422764480114,
+      "epoch": 0.40298460472877246,
+      "grad_norm": 0.60546875,
+      "learning_rate": 2.9866532359607154e-05,
+      "loss": 1.0069,
+      "mean_token_accuracy": 0.736028865724802,
+      "num_tokens": 148480143.0,
+      "step": 1600
+    },
+    {
+      "entropy": 1.5022207498550415,
+      "epoch": 0.4055032585083273,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.9740619491312015e-05,
+      "loss": 1.0107,
+      "mean_token_accuracy": 0.7401330538094044,
+      "num_tokens": 149430672.0,
+      "step": 1610
+    },
+    {
+      "entropy": 1.4303474977612496,
+      "epoch": 0.40802191228788215,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.9614706623016876e-05,
+      "loss": 0.98,
+      "mean_token_accuracy": 0.7455232538282871,
+      "num_tokens": 150340592.0,
+      "step": 1620
+    },
+    {
+      "entropy": 1.5223925068974495,
+      "epoch": 0.41054056606743694,
+      "grad_norm": 0.671875,
+      "learning_rate": 2.9488793754721734e-05,
+      "loss": 1.0375,
+      "mean_token_accuracy": 0.7336984358727932,
+      "num_tokens": 151262776.0,
+      "step": 1630
+    },
+    {
+      "entropy": 1.5078096300363542,
+      "epoch": 0.4130592198469918,
+      "grad_norm": 0.6015625,
+      "learning_rate": 2.9362880886426598e-05,
+      "loss": 1.0294,
+      "mean_token_accuracy": 0.7375610016286374,
+      "num_tokens": 152191345.0,
+      "step": 1640
+    },
+    {
+      "entropy": 1.4810595944523812,
+      "epoch": 0.41557787362654663,
+      "grad_norm": 0.64453125,
+      "learning_rate": 2.9236968018131456e-05,
+      "loss": 1.0057,
+      "mean_token_accuracy": 0.7403605856001377,
+      "num_tokens": 153133627.0,
+      "step": 1650
+    },
+    {
+      "entropy": 1.5020821630954742,
+      "epoch": 0.4180965274061014,
+      "grad_norm": 0.5859375,
+      "learning_rate": 2.9111055149836313e-05,
+      "loss": 1.0231,
+      "mean_token_accuracy": 0.7343570776283741,
+      "num_tokens": 154077853.0,
+      "step": 1660
+    },
+    {
+      "entropy": 1.4656946510076523,
+      "epoch": 0.4206151811856563,
+      "grad_norm": 0.609375,
+      "learning_rate": 2.898514228154117e-05,
+      "loss": 0.9825,
+      "mean_token_accuracy": 0.7439971931278706,
+      "num_tokens": 154999470.0,
+      "step": 1670
+    },
+    {
+      "entropy": 1.4865729957818985,
+      "epoch": 0.4231338349652111,
+      "grad_norm": 0.60546875,
+      "learning_rate": 2.8859229413246036e-05,
+      "loss": 0.9847,
+      "mean_token_accuracy": 0.741909223049879,
+      "num_tokens": 155919939.0,
+      "step": 1680
+    },
+    {
+      "entropy": 1.442151701450348,
+      "epoch": 0.4256524887447659,
+      "grad_norm": 0.59765625,
+      "learning_rate": 2.8733316544950893e-05,
+      "loss": 0.971,
+      "mean_token_accuracy": 0.747497969865799,
+      "num_tokens": 156857683.0,
+      "step": 1690
+    },
+    {
+      "entropy": 1.4939273685216903,
+      "epoch": 0.42817114252432076,
+      "grad_norm": 0.58203125,
+      "learning_rate": 2.8607403676655754e-05,
+      "loss": 1.027,
+      "mean_token_accuracy": 0.73572783395648,
+      "num_tokens": 157789868.0,
+      "step": 1700
+    },
+    {
+      "entropy": 1.4739609718322755,
+      "epoch": 0.4306897963038756,
+      "grad_norm": 0.62109375,
+      "learning_rate": 2.8481490808360615e-05,
+      "loss": 1.0036,
+      "mean_token_accuracy": 0.7423903658986092,
+      "num_tokens": 158671912.0,
+      "step": 1710
+    },
+    {
+      "entropy": 1.5179324716329574,
+      "epoch": 0.4332084500834304,
+      "grad_norm": 0.609375,
+      "learning_rate": 2.8355577940065477e-05,
+      "loss": 1.0389,
+      "mean_token_accuracy": 0.7320630177855492,
+      "num_tokens": 159601676.0,
+      "step": 1720
+    },
+    {
+      "entropy": 1.500729313492775,
+      "epoch": 0.43572710386298524,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.8229665071770334e-05,
+      "loss": 1.0081,
+      "mean_token_accuracy": 0.7389924034476281,
+      "num_tokens": 160526625.0,
+      "step": 1730
+    },
+    {
+      "entropy": 1.4945330180227756,
+      "epoch": 0.43824575764254003,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.81037522034752e-05,
+      "loss": 0.9928,
+      "mean_token_accuracy": 0.7408586353063583,
+      "num_tokens": 161444586.0,
+      "step": 1740
+    },
+    {
+      "entropy": 1.4932605415582656,
+      "epoch": 0.4407644114220949,
+      "grad_norm": 0.63671875,
+      "learning_rate": 2.7977839335180056e-05,
+      "loss": 1.0089,
+      "mean_token_accuracy": 0.7384962946176529,
+      "num_tokens": 162384356.0,
+      "step": 1750
+    },
+    {
+      "entropy": 1.5264878809452056,
+      "epoch": 0.44328306520164973,
+      "grad_norm": 0.625,
+      "learning_rate": 2.7851926466884914e-05,
+      "loss": 1.003,
+      "mean_token_accuracy": 0.7381562039256095,
+      "num_tokens": 163308126.0,
+      "step": 1760
+    },
+    {
+      "entropy": 1.4538481026887893,
+      "epoch": 0.4458017189812045,
+      "grad_norm": 0.58203125,
+      "learning_rate": 2.772601359858978e-05,
+      "loss": 0.9907,
+      "mean_token_accuracy": 0.7408280216157437,
+      "num_tokens": 164233223.0,
+      "step": 1770
+    },
+    {
+      "entropy": 1.4829386696219444,
+      "epoch": 0.44832037276075937,
+      "grad_norm": 0.5625,
+      "learning_rate": 2.7600100730294636e-05,
+      "loss": 0.9865,
+      "mean_token_accuracy": 0.7451853767037392,
+      "num_tokens": 165127331.0,
+      "step": 1780
+    },
+    {
+      "entropy": 1.4769062221050262,
+      "epoch": 0.4508390265403142,
+      "grad_norm": 0.65625,
+      "learning_rate": 2.7474187861999494e-05,
+      "loss": 0.9878,
+      "mean_token_accuracy": 0.7435719795525074,
+      "num_tokens": 166030010.0,
+      "step": 1790
+    },
+    {
+      "entropy": 1.4725798398256302,
+      "epoch": 0.453357680319869,
+      "grad_norm": 0.6171875,
+      "learning_rate": 2.734827499370436e-05,
+      "loss": 0.9798,
+      "mean_token_accuracy": 0.7441974826157093,
+      "num_tokens": 166972675.0,
+      "step": 1800
+    },
+    {
+      "entropy": 1.5000967025756835,
+      "epoch": 0.45587633409942385,
+      "grad_norm": 0.6015625,
+      "learning_rate": 2.7222362125409216e-05,
+      "loss": 0.9785,
+      "mean_token_accuracy": 0.7470294989645481,
+      "num_tokens": 167918655.0,
+      "step": 1810
+    },
+    {
+      "entropy": 1.507605105638504,
+      "epoch": 0.4583949878789787,
+      "grad_norm": 0.6953125,
+      "learning_rate": 2.7096449257114077e-05,
+      "loss": 1.0254,
+      "mean_token_accuracy": 0.7376879565417767,
+      "num_tokens": 168837593.0,
+      "step": 1820
+    },
+    {
+      "entropy": 1.483200278878212,
+      "epoch": 0.4609136416585335,
+      "grad_norm": 0.609375,
+      "learning_rate": 2.697053638881894e-05,
+      "loss": 1.0047,
+      "mean_token_accuracy": 0.7400114268064499,
+      "num_tokens": 169770322.0,
+      "step": 1830
+    },
+    {
+      "entropy": 1.503974023461342,
+      "epoch": 0.46343229543808834,
+      "grad_norm": 0.62109375,
+      "learning_rate": 2.68446235205238e-05,
+      "loss": 1.0527,
+      "mean_token_accuracy": 0.7306334659457207,
+      "num_tokens": 170718172.0,
+      "step": 1840
+    },
+    {
+      "entropy": 1.4733003690838813,
+      "epoch": 0.4659509492176432,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.6718710652228657e-05,
+      "loss": 0.9902,
+      "mean_token_accuracy": 0.7422384157776832,
+      "num_tokens": 171646451.0,
+      "step": 1850
+    },
+    {
+      "entropy": 1.4795169189572335,
+      "epoch": 0.468469602997198,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.659279778393352e-05,
+      "loss": 0.994,
+      "mean_token_accuracy": 0.7433700770139694,
+      "num_tokens": 172571288.0,
+      "step": 1860
+    },
+    {
+      "entropy": 1.544270959496498,
+      "epoch": 0.4709882567767528,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.646688491563838e-05,
+      "loss": 1.0684,
+      "mean_token_accuracy": 0.7255132175981999,
+      "num_tokens": 173493482.0,
+      "step": 1870
+    },
+    {
+      "entropy": 1.4687143832445144,
+      "epoch": 0.47350691055630767,
+      "grad_norm": 0.6015625,
+      "learning_rate": 2.6340972047343237e-05,
+      "loss": 1.0264,
+      "mean_token_accuracy": 0.7358646772801876,
+      "num_tokens": 174430178.0,
+      "step": 1880
+    },
+    {
+      "entropy": 1.5258139103651047,
+      "epoch": 0.47602556433586246,
+      "grad_norm": 0.6328125,
+      "learning_rate": 2.62150591790481e-05,
+      "loss": 1.0474,
+      "mean_token_accuracy": 0.7309989266097545,
+      "num_tokens": 175370191.0,
+      "step": 1890
+    },
+    {
+      "entropy": 1.5031465500593186,
+      "epoch": 0.4785442181154173,
+      "grad_norm": 0.6640625,
+      "learning_rate": 2.608914631075296e-05,
+      "loss": 1.0244,
+      "mean_token_accuracy": 0.7371547490358352,
+      "num_tokens": 176327610.0,
+      "step": 1900
+    },
+    {
+      "entropy": 1.4879642963409423,
+      "epoch": 0.48106287189497215,
+      "grad_norm": 0.62890625,
+      "learning_rate": 2.5963233442457817e-05,
+      "loss": 0.9919,
+      "mean_token_accuracy": 0.7449640512466431,
+      "num_tokens": 177261487.0,
+      "step": 1910
+    },
+    {
+      "entropy": 1.4944231912493706,
+      "epoch": 0.48358152567452695,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.583732057416268e-05,
+      "loss": 1.0104,
+      "mean_token_accuracy": 0.7372982695698738,
+      "num_tokens": 178202468.0,
+      "step": 1920
+    },
+    {
+      "entropy": 1.5143211469054223,
+      "epoch": 0.4861001794540818,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.571140770586754e-05,
+      "loss": 1.0548,
+      "mean_token_accuracy": 0.726641795784235,
+      "num_tokens": 179133391.0,
+      "step": 1930
+    },
+    {
+      "entropy": 1.5140092059969903,
+      "epoch": 0.48861883323363664,
+      "grad_norm": 0.609375,
+      "learning_rate": 2.55854948375724e-05,
+      "loss": 0.9964,
+      "mean_token_accuracy": 0.7447598196566105,
+      "num_tokens": 180047435.0,
+      "step": 1940
+    },
+    {
+      "entropy": 1.4550499767065048,
+      "epoch": 0.49113748701319143,
+      "grad_norm": 0.56640625,
+      "learning_rate": 2.5459581969277265e-05,
+      "loss": 0.9904,
+      "mean_token_accuracy": 0.7420261397957801,
+      "num_tokens": 180983171.0,
+      "step": 1950
+    },
+    {
+      "entropy": 1.4785954818129539,
+      "epoch": 0.4936561407927463,
+      "grad_norm": 0.546875,
+      "learning_rate": 2.5333669100982122e-05,
+      "loss": 0.9914,
+      "mean_token_accuracy": 0.7422729566693306,
+      "num_tokens": 181901605.0,
+      "step": 1960
+    },
+    {
+      "entropy": 1.4713733956217765,
+      "epoch": 0.4961747945723011,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.520775623268698e-05,
+      "loss": 1.0136,
+      "mean_token_accuracy": 0.7407643228769303,
+      "num_tokens": 182808597.0,
+      "step": 1970
+    },
+    {
+      "entropy": 1.4882659435272216,
+      "epoch": 0.4986934483518559,
+      "grad_norm": 0.59765625,
+      "learning_rate": 2.5081843364391845e-05,
+      "loss": 1.0281,
+      "mean_token_accuracy": 0.733605157583952,
+      "num_tokens": 183760991.0,
+      "step": 1980
+    },
+    {
+      "entropy": 1.4794856160879135,
+      "epoch": 0.5012121021314108,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.4955930496096702e-05,
+      "loss": 0.9757,
+      "mean_token_accuracy": 0.7448467329144478,
+      "num_tokens": 184676166.0,
+      "step": 1990
+    },
+    {
+      "entropy": 1.4704470828175544,
+      "epoch": 0.5037307559109656,
+      "grad_norm": 0.5859375,
+      "learning_rate": 2.4830017627801563e-05,
+      "loss": 0.9887,
+      "mean_token_accuracy": 0.7421532794833183,
+      "num_tokens": 185591403.0,
+      "step": 2000
+    },
+    {
+      "entropy": 1.4506453573703766,
+      "epoch": 0.5062494096905205,
+      "grad_norm": 0.6015625,
+      "learning_rate": 2.470410475950642e-05,
+      "loss": 0.9482,
+      "mean_token_accuracy": 0.751684907823801,
+      "num_tokens": 186530825.0,
+      "step": 2010
+    },
+    {
+      "entropy": 1.505790999531746,
+      "epoch": 0.5087680634700752,
+      "grad_norm": 0.64453125,
+      "learning_rate": 2.4578191891211282e-05,
+      "loss": 1.01,
+      "mean_token_accuracy": 0.738251518458128,
+      "num_tokens": 187483104.0,
+      "step": 2020
+    },
+    {
+      "entropy": 1.4906439885497094,
+      "epoch": 0.51128671724963,
+      "grad_norm": 0.609375,
+      "learning_rate": 2.4452279022916143e-05,
+      "loss": 1.0081,
+      "mean_token_accuracy": 0.7367694586515426,
+      "num_tokens": 188419415.0,
+      "step": 2030
+    },
+    {
+      "entropy": 1.5162854179739953,
+      "epoch": 0.5138053710291849,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.4326366154621e-05,
+      "loss": 1.0318,
+      "mean_token_accuracy": 0.7348616622388363,
+      "num_tokens": 189364907.0,
+      "step": 2040
+    },
+    {
+      "entropy": 1.515935306251049,
+      "epoch": 0.5163240248087397,
+      "grad_norm": 0.625,
+      "learning_rate": 2.4200453286325865e-05,
+      "loss": 1.037,
+      "mean_token_accuracy": 0.7346496164798737,
+      "num_tokens": 190305289.0,
+      "step": 2050
+    },
+    {
+      "entropy": 1.5043818101286888,
+      "epoch": 0.5188426785882946,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.4074540418030726e-05,
+      "loss": 1.0202,
+      "mean_token_accuracy": 0.737002718448639,
+      "num_tokens": 191240393.0,
+      "step": 2060
+    },
+    {
+      "entropy": 1.4879194065928458,
+      "epoch": 0.5213613323678494,
+      "grad_norm": 0.625,
+      "learning_rate": 2.3948627549735584e-05,
+      "loss": 1.0149,
+      "mean_token_accuracy": 0.7377648189663887,
+      "num_tokens": 192175710.0,
+      "step": 2070
+    },
+    {
+      "entropy": 1.476494650542736,
+      "epoch": 0.5238799861474042,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.3822714681440445e-05,
+      "loss": 1.0126,
+      "mean_token_accuracy": 0.7382982887327671,
+      "num_tokens": 193089119.0,
+      "step": 2080
+    },
+    {
+      "entropy": 1.4906734734773637,
+      "epoch": 0.526398639926959,
+      "grad_norm": 0.58203125,
+      "learning_rate": 2.3696801813145306e-05,
+      "loss": 1.0228,
+      "mean_token_accuracy": 0.7358043491840363,
+      "num_tokens": 194014464.0,
+      "step": 2090
+    },
+    {
+      "entropy": 1.463935060799122,
+      "epoch": 0.5289172937065139,
+      "grad_norm": 0.546875,
+      "learning_rate": 2.3570888944850164e-05,
+      "loss": 0.979,
+      "mean_token_accuracy": 0.7440614722669124,
+      "num_tokens": 194937483.0,
+      "step": 2100
+    },
+    {
+      "entropy": 1.4706278055906297,
+      "epoch": 0.5314359474860687,
+      "grad_norm": 0.609375,
+      "learning_rate": 2.3444976076555025e-05,
+      "loss": 0.9844,
+      "mean_token_accuracy": 0.742541317641735,
+      "num_tokens": 195881861.0,
+      "step": 2110
+    },
+    {
+      "entropy": 1.510326901078224,
+      "epoch": 0.5339546012656236,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.3319063208259886e-05,
+      "loss": 1.0143,
+      "mean_token_accuracy": 0.7375903382897377,
+      "num_tokens": 196792310.0,
+      "step": 2120
+    },
+    {
+      "entropy": 1.500123569369316,
+      "epoch": 0.5364732550451784,
+      "grad_norm": 0.6171875,
+      "learning_rate": 2.3193150339964744e-05,
+      "loss": 1.0156,
+      "mean_token_accuracy": 0.737464863806963,
+      "num_tokens": 197740447.0,
+      "step": 2130
+    },
+    {
+      "entropy": 1.5234432876110078,
+      "epoch": 0.5389919088247331,
+      "grad_norm": 0.66015625,
+      "learning_rate": 2.3067237471669605e-05,
+      "loss": 1.0293,
+      "mean_token_accuracy": 0.7343058608472347,
+      "num_tokens": 198650674.0,
+      "step": 2140
+    },
+    {
+      "entropy": 1.4829162567853929,
+      "epoch": 0.541510562604288,
+      "grad_norm": 0.59765625,
+      "learning_rate": 2.2941324603374466e-05,
+      "loss": 1.0158,
+      "mean_token_accuracy": 0.7392853118479252,
+      "num_tokens": 199579745.0,
+      "step": 2150
+    },
+    {
+      "entropy": 1.4706959769129753,
+      "epoch": 0.5440292163838428,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.2815411735079327e-05,
+      "loss": 0.9895,
+      "mean_token_accuracy": 0.7438252337276936,
+      "num_tokens": 200490996.0,
+      "step": 2160
+    },
+    {
+      "entropy": 1.528316856920719,
+      "epoch": 0.5465478701633977,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.2689498866784188e-05,
+      "loss": 1.0492,
+      "mean_token_accuracy": 0.7321145072579384,
+      "num_tokens": 201428466.0,
+      "step": 2170
+    },
+    {
+      "entropy": 1.4623350918292999,
+      "epoch": 0.5490665239429525,
+      "grad_norm": 0.59375,
+      "learning_rate": 2.256358599848905e-05,
+      "loss": 0.981,
+      "mean_token_accuracy": 0.744949608296156,
+      "num_tokens": 202354085.0,
+      "step": 2180
+    },
+    {
+      "entropy": 1.5169390365481377,
+      "epoch": 0.5515851777225074,
+      "grad_norm": 0.6484375,
+      "learning_rate": 2.2437673130193907e-05,
+      "loss": 1.0167,
+      "mean_token_accuracy": 0.737620497494936,
+      "num_tokens": 203296048.0,
+      "step": 2190
+    },
+    {
+      "entropy": 1.4908597692847252,
+      "epoch": 0.5541038315020621,
+      "grad_norm": 0.60546875,
+      "learning_rate": 2.2311760261898768e-05,
+      "loss": 0.9884,
+      "mean_token_accuracy": 0.7446716167032719,
+      "num_tokens": 204251239.0,
+      "step": 2200
+    },
+    {
+      "entropy": 1.4890633031725884,
+      "epoch": 0.556622485281617,
+      "grad_norm": 0.6171875,
+      "learning_rate": 2.218584739360363e-05,
+      "loss": 0.9989,
+      "mean_token_accuracy": 0.7413519226014614,
+      "num_tokens": 205151325.0,
+      "step": 2210
+    },
+    {
+      "entropy": 1.519281443953514,
+      "epoch": 0.5591411390611718,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.2059934525308487e-05,
+      "loss": 1.021,
+      "mean_token_accuracy": 0.7366731703281403,
+      "num_tokens": 206096182.0,
+      "step": 2220
+    },
+    {
+      "entropy": 1.4802996426820756,
+      "epoch": 0.5616597928407266,
+      "grad_norm": 0.5859375,
+      "learning_rate": 2.1934021657013348e-05,
+      "loss": 0.9981,
+      "mean_token_accuracy": 0.738096322119236,
+      "num_tokens": 207029179.0,
+      "step": 2230
+    },
+    {
+      "entropy": 1.493895760178566,
+      "epoch": 0.5641784466202815,
+      "grad_norm": 0.6015625,
+      "learning_rate": 2.1808108788718206e-05,
+      "loss": 1.0104,
+      "mean_token_accuracy": 0.7363863155245781,
+      "num_tokens": 207960947.0,
+      "step": 2240
+    },
+    {
+      "entropy": 1.472677892446518,
+      "epoch": 0.5666971003998363,
+      "grad_norm": 0.62109375,
+      "learning_rate": 2.1682195920423067e-05,
+      "loss": 0.9692,
+      "mean_token_accuracy": 0.7454905331134796,
+      "num_tokens": 208889854.0,
+      "step": 2250
+    },
+    {
+      "entropy": 1.5076108410954476,
+      "epoch": 0.5692157541793911,
+      "grad_norm": 0.60546875,
+      "learning_rate": 2.1556283052127928e-05,
+      "loss": 1.0289,
+      "mean_token_accuracy": 0.7352958224713803,
+      "num_tokens": 209818677.0,
+      "step": 2260
+    },
+    {
+      "entropy": 1.4948209643363952,
+      "epoch": 0.5717344079589459,
+      "grad_norm": 0.56640625,
+      "learning_rate": 2.143037018383279e-05,
+      "loss": 0.9812,
+      "mean_token_accuracy": 0.7438007473945618,
+      "num_tokens": 210768812.0,
+      "step": 2270
+    },
+    {
+      "entropy": 1.4773472845554352,
+      "epoch": 0.5742530617385008,
+      "grad_norm": 0.8125,
+      "learning_rate": 2.130445731553765e-05,
+      "loss": 1.0183,
+      "mean_token_accuracy": 0.7372843034565448,
+      "num_tokens": 211681636.0,
+      "step": 2280
+    },
+    {
+      "entropy": 1.4644770205020905,
+      "epoch": 0.5767717155180556,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.117854444724251e-05,
+      "loss": 0.9989,
+      "mean_token_accuracy": 0.7396629564464092,
+      "num_tokens": 212615595.0,
+      "step": 2290
+    },
+    {
+      "entropy": 1.5075991481542588,
+      "epoch": 0.5792903692976105,
+      "grad_norm": 0.640625,
+      "learning_rate": 2.105263157894737e-05,
+      "loss": 1.0286,
+      "mean_token_accuracy": 0.7359593294560909,
+      "num_tokens": 213552263.0,
+      "step": 2300
+    },
+    {
+      "entropy": 1.4733971416950227,
+      "epoch": 0.5818090230771653,
+      "grad_norm": 0.6015625,
+      "learning_rate": 2.092671871065223e-05,
+      "loss": 1.0125,
+      "mean_token_accuracy": 0.7387680150568485,
+      "num_tokens": 214486865.0,
+      "step": 2310
+    },
+    {
+      "entropy": 1.4752314537763596,
+      "epoch": 0.58432767685672,
+      "grad_norm": 0.6171875,
+      "learning_rate": 2.080080584235709e-05,
+      "loss": 0.9759,
+      "mean_token_accuracy": 0.7432409018278122,
+      "num_tokens": 215442702.0,
+      "step": 2320
+    },
+    {
+      "entropy": 1.4748615235090257,
+      "epoch": 0.5868463306362749,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.067489297406195e-05,
+      "loss": 0.9963,
+      "mean_token_accuracy": 0.740365145355463,
+      "num_tokens": 216371631.0,
+      "step": 2330
+    },
+    {
+      "entropy": 1.4902510225772858,
+      "epoch": 0.5893649844158297,
+      "grad_norm": 0.6015625,
+      "learning_rate": 2.054898010576681e-05,
+      "loss": 0.9753,
+      "mean_token_accuracy": 0.7458043657243252,
+      "num_tokens": 217312756.0,
+      "step": 2340
+    },
+    {
+      "entropy": 1.4833617210388184,
+      "epoch": 0.5918836381953846,
+      "grad_norm": 0.58203125,
+      "learning_rate": 2.042306723747167e-05,
+      "loss": 1.0325,
+      "mean_token_accuracy": 0.7355702064931393,
+      "num_tokens": 218231302.0,
+      "step": 2350
+    },
+    {
+      "entropy": 1.510320021212101,
+      "epoch": 0.5944022919749394,
+      "grad_norm": 0.62890625,
+      "learning_rate": 2.029715436917653e-05,
+      "loss": 0.9958,
+      "mean_token_accuracy": 0.7420489639043808,
+      "num_tokens": 219167712.0,
+      "step": 2360
+    },
+    {
+      "entropy": 1.4711307182908058,
+      "epoch": 0.5969209457544942,
+      "grad_norm": 0.58203125,
+      "learning_rate": 2.017124150088139e-05,
+      "loss": 0.9686,
+      "mean_token_accuracy": 0.7474707916378975,
+      "num_tokens": 220082656.0,
+      "step": 2370
+    },
+    {
+      "entropy": 1.501902323961258,
+      "epoch": 0.599439599534049,
+      "grad_norm": 0.578125,
+      "learning_rate": 2.004532863258625e-05,
+      "loss": 1.0207,
+      "mean_token_accuracy": 0.7335725836455822,
+      "num_tokens": 220988673.0,
+      "step": 2380
+    },
+    {
+      "entropy": 1.4819249659776688,
+      "epoch": 0.6019582533136039,
+      "grad_norm": 0.63671875,
+      "learning_rate": 1.9919415764291112e-05,
+      "loss": 0.9895,
+      "mean_token_accuracy": 0.7437810882925987,
+      "num_tokens": 221890469.0,
+      "step": 2390
+    },
+    {
+      "entropy": 1.4631519243121147,
+      "epoch": 0.6044769070931587,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.9793502895995973e-05,
+      "loss": 0.9768,
+      "mean_token_accuracy": 0.7473115906119346,
+      "num_tokens": 222833708.0,
+      "step": 2400
+    },
+    {
+      "entropy": 1.4829539999365806,
+      "epoch": 0.6069955608727136,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.9667590027700834e-05,
+      "loss": 1.0029,
+      "mean_token_accuracy": 0.7405181847512722,
+      "num_tokens": 223767218.0,
+      "step": 2410
+    },
+    {
+      "entropy": 1.4947340697050095,
+      "epoch": 0.6095142146522684,
+      "grad_norm": 0.63671875,
+      "learning_rate": 1.9541677159405692e-05,
+      "loss": 1.0127,
+      "mean_token_accuracy": 0.7391134299337864,
+      "num_tokens": 224688061.0,
+      "step": 2420
+    },
+    {
+      "entropy": 1.450956329703331,
+      "epoch": 0.6120328684318231,
+      "grad_norm": 0.640625,
+      "learning_rate": 1.9415764291110553e-05,
+      "loss": 0.9603,
+      "mean_token_accuracy": 0.7479838944971562,
+      "num_tokens": 225619117.0,
+      "step": 2430
+    },
+    {
+      "entropy": 1.4841125264763833,
+      "epoch": 0.614551522211378,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.9289851422815414e-05,
+      "loss": 0.9819,
+      "mean_token_accuracy": 0.7439415737986564,
+      "num_tokens": 226543813.0,
+      "step": 2440
+    },
+    {
+      "entropy": 1.4869384542107582,
+      "epoch": 0.6170701759909328,
+      "grad_norm": 0.6484375,
+      "learning_rate": 1.916393855452027e-05,
+      "loss": 1.0238,
+      "mean_token_accuracy": 0.7360433027148247,
+      "num_tokens": 227466179.0,
+      "step": 2450
+    },
+    {
+      "entropy": 1.4867294758558274,
+      "epoch": 0.6195888297704877,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.9038025686225133e-05,
+      "loss": 1.0198,
+      "mean_token_accuracy": 0.7386736191809178,
+      "num_tokens": 228391779.0,
+      "step": 2460
+    },
+    {
+      "entropy": 1.4967486664652825,
+      "epoch": 0.6221074835500425,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.8912112817929994e-05,
+      "loss": 0.9842,
+      "mean_token_accuracy": 0.7447447247803212,
+      "num_tokens": 229318854.0,
+      "step": 2470
+    },
+    {
+      "entropy": 1.4630463495850563,
+      "epoch": 0.6246261373295974,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.878619994963485e-05,
+      "loss": 0.9822,
+      "mean_token_accuracy": 0.7429227210581303,
+      "num_tokens": 230243272.0,
+      "step": 2480
+    },
+    {
+      "entropy": 1.5132867440581321,
+      "epoch": 0.6271447911091521,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.8660287081339713e-05,
+      "loss": 1.0411,
+      "mean_token_accuracy": 0.7323724761605263,
+      "num_tokens": 231165164.0,
+      "step": 2490
+    },
+    {
+      "entropy": 1.505117255449295,
+      "epoch": 0.629663444888707,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.8534374213044574e-05,
+      "loss": 1.0083,
+      "mean_token_accuracy": 0.7365673772990704,
+      "num_tokens": 232090344.0,
+      "step": 2500
+    },
+    {
+      "entropy": 1.5145672991871835,
+      "epoch": 0.6321820986682618,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.8408461344749435e-05,
+      "loss": 1.0037,
+      "mean_token_accuracy": 0.7394612200558186,
+      "num_tokens": 233012123.0,
+      "step": 2510
+    },
+    {
+      "entropy": 1.4879171073436737,
+      "epoch": 0.6347007524478167,
+      "grad_norm": 0.62109375,
+      "learning_rate": 1.8282548476454296e-05,
+      "loss": 1.0255,
+      "mean_token_accuracy": 0.734065717458725,
+      "num_tokens": 233944623.0,
+      "step": 2520
+    },
+    {
+      "entropy": 1.5197929657995701,
+      "epoch": 0.6372194062273715,
+      "grad_norm": 0.5703125,
+      "learning_rate": 1.8156635608159157e-05,
+      "loss": 1.0018,
+      "mean_token_accuracy": 0.7394672058522701,
+      "num_tokens": 234854307.0,
+      "step": 2530
+    },
+    {
+      "entropy": 1.4701609089970589,
+      "epoch": 0.6397380600069263,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.8030722739864015e-05,
+      "loss": 0.9855,
+      "mean_token_accuracy": 0.7438199289143086,
+      "num_tokens": 235800851.0,
+      "step": 2540
+    },
+    {
+      "entropy": 1.4832564637064933,
+      "epoch": 0.6422567137864811,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.7904809871568876e-05,
+      "loss": 1.011,
+      "mean_token_accuracy": 0.73984669521451,
+      "num_tokens": 236747858.0,
+      "step": 2550
+    },
+    {
+      "entropy": 1.4942113682627678,
+      "epoch": 0.6447753675660359,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.7778897003273737e-05,
+      "loss": 1.0095,
+      "mean_token_accuracy": 0.737560261040926,
+      "num_tokens": 237690926.0,
+      "step": 2560
+    },
+    {
+      "entropy": 1.4950286373496056,
+      "epoch": 0.6472940213455908,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.7652984134978595e-05,
+      "loss": 1.0272,
+      "mean_token_accuracy": 0.7348176456987858,
+      "num_tokens": 238601585.0,
+      "step": 2570
+    },
+    {
+      "entropy": 1.507275079190731,
+      "epoch": 0.6498126751251456,
+      "grad_norm": 0.62109375,
+      "learning_rate": 1.7527071266683456e-05,
+      "loss": 0.9862,
+      "mean_token_accuracy": 0.7441342391073704,
+      "num_tokens": 239561971.0,
+      "step": 2580
+    },
+    {
+      "entropy": 1.4744239822030067,
+      "epoch": 0.6523313289047005,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.7401158398388313e-05,
+      "loss": 0.9914,
+      "mean_token_accuracy": 0.7422404564917088,
+      "num_tokens": 240510814.0,
+      "step": 2590
+    },
+    {
+      "entropy": 1.4942139282822609,
+      "epoch": 0.6548499826842553,
+      "grad_norm": 0.62890625,
+      "learning_rate": 1.7275245530093174e-05,
+      "loss": 0.9984,
+      "mean_token_accuracy": 0.7415653012692929,
+      "num_tokens": 241429091.0,
+      "step": 2600
+    },
+    {
+      "entropy": 1.5210790306329727,
+      "epoch": 0.65736863646381,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.7149332661798036e-05,
+      "loss": 1.0669,
+      "mean_token_accuracy": 0.7277764029800892,
+      "num_tokens": 242326342.0,
+      "step": 2610
+    },
+    {
+      "entropy": 1.501256914436817,
+      "epoch": 0.6598872902433649,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.7023419793502897e-05,
+      "loss": 1.0371,
+      "mean_token_accuracy": 0.7327685542404652,
+      "num_tokens": 243262918.0,
+      "step": 2620
+    },
+    {
+      "entropy": 1.4907999977469444,
+      "epoch": 0.6624059440229197,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.6897506925207758e-05,
+      "loss": 1.0074,
+      "mean_token_accuracy": 0.7394888132810593,
+      "num_tokens": 244186200.0,
+      "step": 2630
+    },
+    {
+      "entropy": 1.5089690536260605,
+      "epoch": 0.6649245978024746,
+      "grad_norm": 0.640625,
+      "learning_rate": 1.677159405691262e-05,
+      "loss": 1.0026,
+      "mean_token_accuracy": 0.7411562100052833,
+      "num_tokens": 245110985.0,
+      "step": 2640
+    },
+    {
+      "entropy": 1.4826441787183284,
+      "epoch": 0.6674432515820294,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.6645681188617477e-05,
+      "loss": 1.0147,
+      "mean_token_accuracy": 0.7376951925456524,
+      "num_tokens": 246023602.0,
+      "step": 2650
+    },
+    {
+      "entropy": 1.5016205877065658,
+      "epoch": 0.6699619053615843,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.6519768320322338e-05,
+      "loss": 1.0179,
+      "mean_token_accuracy": 0.7369609415531159,
+      "num_tokens": 246986524.0,
+      "step": 2660
+    },
+    {
+      "entropy": 1.4834182366728783,
+      "epoch": 0.672480559141139,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.63938554520272e-05,
+      "loss": 0.9773,
+      "mean_token_accuracy": 0.7454558834433556,
+      "num_tokens": 247926005.0,
+      "step": 2670
+    },
+    {
+      "entropy": 1.4759241729974746,
+      "epoch": 0.6749992129206939,
+      "grad_norm": 0.62109375,
+      "learning_rate": 1.6267942583732056e-05,
+      "loss": 1.0038,
+      "mean_token_accuracy": 0.7400371395051479,
+      "num_tokens": 248847171.0,
+      "step": 2680
+    },
+    {
+      "entropy": 1.5304366648197174,
+      "epoch": 0.6775178667002487,
+      "grad_norm": 0.63671875,
+      "learning_rate": 1.6142029715436917e-05,
+      "loss": 1.0356,
+      "mean_token_accuracy": 0.7320890232920647,
+      "num_tokens": 249786960.0,
+      "step": 2690
+    },
+    {
+      "entropy": 1.4903511464595796,
+      "epoch": 0.6800365204798036,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.601611684714178e-05,
+      "loss": 0.9906,
+      "mean_token_accuracy": 0.7420216508209705,
+      "num_tokens": 250698887.0,
+      "step": 2700
+    },
+    {
+      "entropy": 1.5175952777266501,
+      "epoch": 0.6825551742593584,
+      "grad_norm": 0.625,
+      "learning_rate": 1.5890203978846636e-05,
+      "loss": 0.9865,
+      "mean_token_accuracy": 0.7433042749762535,
+      "num_tokens": 251619967.0,
+      "step": 2710
+    },
+    {
+      "entropy": 1.4962933480739593,
+      "epoch": 0.6850738280389131,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.5764291110551497e-05,
+      "loss": 1.0039,
+      "mean_token_accuracy": 0.7438869446516037,
+      "num_tokens": 252542015.0,
+      "step": 2720
+    },
+    {
+      "entropy": 1.475045308470726,
+      "epoch": 0.687592481818468,
+      "grad_norm": 0.62109375,
+      "learning_rate": 1.5638378242256362e-05,
+      "loss": 0.9653,
+      "mean_token_accuracy": 0.7465457111597061,
+      "num_tokens": 253479410.0,
+      "step": 2730
+    },
+    {
+      "entropy": 1.4475198343396187,
+      "epoch": 0.6901111355980228,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.551246537396122e-05,
+      "loss": 0.9883,
+      "mean_token_accuracy": 0.7442029610276222,
+      "num_tokens": 254421008.0,
+      "step": 2740
+    },
+    {
+      "entropy": 1.4826817378401755,
+      "epoch": 0.6926297893775777,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.538655250566608e-05,
+      "loss": 1.0162,
+      "mean_token_accuracy": 0.7372715182602405,
+      "num_tokens": 255356938.0,
+      "step": 2750
+    },
+    {
+      "entropy": 1.4607293367385865,
+      "epoch": 0.6951484431571325,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.5260639637370942e-05,
+      "loss": 0.982,
+      "mean_token_accuracy": 0.7464394047856331,
+      "num_tokens": 256276194.0,
+      "step": 2760
+    },
+    {
+      "entropy": 1.5108062788844108,
+      "epoch": 0.6976670969366874,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.51347267690758e-05,
+      "loss": 1.0184,
+      "mean_token_accuracy": 0.736438725143671,
+      "num_tokens": 257222479.0,
+      "step": 2770
+    },
+    {
+      "entropy": 1.4691065698862076,
+      "epoch": 0.7001857507162421,
+      "grad_norm": 0.6328125,
+      "learning_rate": 1.500881390078066e-05,
+      "loss": 1.0069,
+      "mean_token_accuracy": 0.7403544157743454,
+      "num_tokens": 258141986.0,
+      "step": 2780
+    },
+    {
+      "entropy": 1.4614216104149818,
+      "epoch": 0.702704404495797,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.4882901032485522e-05,
+      "loss": 0.9821,
+      "mean_token_accuracy": 0.7435299441218376,
+      "num_tokens": 259054793.0,
+      "step": 2790
+    },
+    {
+      "entropy": 1.4462628573179246,
+      "epoch": 0.7052230582753518,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.475698816419038e-05,
+      "loss": 0.9687,
+      "mean_token_accuracy": 0.7498079493641854,
+      "num_tokens": 260012807.0,
+      "step": 2800
+    },
+    {
+      "entropy": 1.4848258540034294,
+      "epoch": 0.7077417120549067,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.4631075295895242e-05,
+      "loss": 0.989,
+      "mean_token_accuracy": 0.7408308543264865,
+      "num_tokens": 260948752.0,
+      "step": 2810
+    },
+    {
+      "entropy": 1.4752742052078247,
+      "epoch": 0.7102603658344615,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.4505162427600103e-05,
+      "loss": 0.9764,
+      "mean_token_accuracy": 0.7448857210576534,
+      "num_tokens": 261830676.0,
+      "step": 2820
+    },
+    {
+      "entropy": 1.4709856286644936,
+      "epoch": 0.7127790196140164,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.4379249559304961e-05,
+      "loss": 0.9866,
+      "mean_token_accuracy": 0.7452083975076675,
+      "num_tokens": 262789665.0,
+      "step": 2830
+    },
+    {
+      "entropy": 1.506988750398159,
+      "epoch": 0.7152976733935711,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.4253336691009822e-05,
+      "loss": 0.9845,
+      "mean_token_accuracy": 0.742633256316185,
+      "num_tokens": 263719914.0,
+      "step": 2840
+    },
+    {
+      "entropy": 1.4864262834191322,
+      "epoch": 0.7178163271731259,
+      "grad_norm": 0.6328125,
+      "learning_rate": 1.4127423822714683e-05,
+      "loss": 1.0123,
+      "mean_token_accuracy": 0.735482431203127,
+      "num_tokens": 264631489.0,
+      "step": 2850
+    },
+    {
+      "entropy": 1.4856854885816575,
+      "epoch": 0.7203349809526808,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.4001510954419542e-05,
+      "loss": 1.0006,
+      "mean_token_accuracy": 0.7381072089076042,
+      "num_tokens": 265559953.0,
+      "step": 2860
+    },
+    {
+      "entropy": 1.5200473442673683,
+      "epoch": 0.7228536347322356,
+      "grad_norm": 0.65625,
+      "learning_rate": 1.3875598086124404e-05,
+      "loss": 1.0176,
+      "mean_token_accuracy": 0.7391909696161747,
+      "num_tokens": 266506028.0,
+      "step": 2870
+    },
+    {
+      "entropy": 1.5127366796135902,
+      "epoch": 0.7253722885117905,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.3749685217829265e-05,
+      "loss": 1.0437,
+      "mean_token_accuracy": 0.7318307563662529,
+      "num_tokens": 267439853.0,
+      "step": 2880
+    },
+    {
+      "entropy": 1.4693277314305306,
+      "epoch": 0.7278909422913453,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.3623772349534122e-05,
+      "loss": 0.9981,
+      "mean_token_accuracy": 0.7403404414653778,
+      "num_tokens": 268367954.0,
+      "step": 2890
+    },
+    {
+      "entropy": 1.4881853014230728,
+      "epoch": 0.7304095960709001,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.3497859481238983e-05,
+      "loss": 1.0046,
+      "mean_token_accuracy": 0.7405321776866913,
+      "num_tokens": 269276694.0,
+      "step": 2900
+    },
+    {
+      "entropy": 1.5084335424005986,
+      "epoch": 0.7329282498504549,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.3371946612943845e-05,
+      "loss": 1.0238,
+      "mean_token_accuracy": 0.7338868103921413,
+      "num_tokens": 270192213.0,
+      "step": 2910
+    },
+    {
+      "entropy": 1.4800263881683349,
+      "epoch": 0.7354469036300098,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.3246033744648704e-05,
+      "loss": 0.982,
+      "mean_token_accuracy": 0.7456442162394523,
+      "num_tokens": 271149665.0,
+      "step": 2920
+    },
+    {
+      "entropy": 1.5040636241436005,
+      "epoch": 0.7379655574095646,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.3120120876353565e-05,
+      "loss": 1.0298,
+      "mean_token_accuracy": 0.733680484443903,
+      "num_tokens": 272066248.0,
+      "step": 2930
+    },
+    {
+      "entropy": 1.4629945561289788,
+      "epoch": 0.7404842111891194,
+      "grad_norm": 0.6328125,
+      "learning_rate": 1.2994208008058426e-05,
+      "loss": 0.9934,
+      "mean_token_accuracy": 0.7426059558987618,
+      "num_tokens": 272970451.0,
+      "step": 2940
+    },
+    {
+      "entropy": 1.4538513720035553,
+      "epoch": 0.7430028649686743,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.2868295139763284e-05,
+      "loss": 1.0009,
+      "mean_token_accuracy": 0.7420231208205224,
+      "num_tokens": 273900781.0,
+      "step": 2950
+    },
+    {
+      "entropy": 1.4652035534381866,
+      "epoch": 0.745521518748229,
+      "grad_norm": 0.5625,
+      "learning_rate": 1.2742382271468145e-05,
+      "loss": 1.0068,
+      "mean_token_accuracy": 0.7384018860757351,
+      "num_tokens": 274840032.0,
+      "step": 2960
+    },
+    {
+      "entropy": 1.515961553156376,
+      "epoch": 0.7480401725277839,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.2616469403173004e-05,
+      "loss": 1.005,
+      "mean_token_accuracy": 0.7410760201513767,
+      "num_tokens": 275780300.0,
+      "step": 2970
+    },
+    {
+      "entropy": 1.5057893320918083,
+      "epoch": 0.7505588263073387,
+      "grad_norm": 0.59375,
+      "learning_rate": 1.2490556534877865e-05,
+      "loss": 1.0121,
+      "mean_token_accuracy": 0.7373140566051006,
+      "num_tokens": 276711260.0,
+      "step": 2980
+    },
+    {
+      "entropy": 1.4707092121243477,
+      "epoch": 0.7530774800868936,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.2364643666582726e-05,
+      "loss": 0.9748,
+      "mean_token_accuracy": 0.7442017652094364,
+      "num_tokens": 277641984.0,
+      "step": 2990
+    },
+    {
+      "entropy": 1.5000385448336602,
+      "epoch": 0.7555961338664484,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.2238730798287586e-05,
+      "loss": 1.0032,
+      "mean_token_accuracy": 0.7411497220396995,
+      "num_tokens": 278587148.0,
+      "step": 3000
+    },
+    {
+      "entropy": 1.5032398775219917,
+      "epoch": 0.7581147876460033,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.2112817929992445e-05,
+      "loss": 0.984,
+      "mean_token_accuracy": 0.7436385117471218,
+      "num_tokens": 279492864.0,
+      "step": 3010
+    },
+    {
+      "entropy": 1.5076435655355453,
+      "epoch": 0.760633441425558,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.1986905061697306e-05,
+      "loss": 1.0129,
+      "mean_token_accuracy": 0.7388051725924015,
+      "num_tokens": 280409665.0,
+      "step": 3020
+    },
+    {
+      "entropy": 1.4777460366487503,
+      "epoch": 0.7631520952051128,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.1860992193402166e-05,
+      "loss": 0.987,
+      "mean_token_accuracy": 0.7431692533195019,
+      "num_tokens": 281347495.0,
+      "step": 3030
+    },
+    {
+      "entropy": 1.4848545402288438,
+      "epoch": 0.7656707489846677,
+      "grad_norm": 0.59765625,
+      "learning_rate": 1.1735079325107027e-05,
+      "loss": 1.012,
+      "mean_token_accuracy": 0.7397673189640045,
+      "num_tokens": 282259845.0,
+      "step": 3040
+    },
+    {
+      "entropy": 1.4779446840286254,
+      "epoch": 0.7681894027642225,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.1609166456811888e-05,
+      "loss": 0.9848,
+      "mean_token_accuracy": 0.7452454976737499,
+      "num_tokens": 283198308.0,
+      "step": 3050
+    },
+    {
+      "entropy": 1.4629323229193687,
+      "epoch": 0.7707080565437774,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.1483253588516747e-05,
+      "loss": 1.0083,
+      "mean_token_accuracy": 0.7391570150852204,
+      "num_tokens": 284134979.0,
+      "step": 3060
+    },
+    {
+      "entropy": 1.490471364557743,
+      "epoch": 0.7732267103233322,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.1357340720221607e-05,
+      "loss": 1.0078,
+      "mean_token_accuracy": 0.7395447090268135,
+      "num_tokens": 285088130.0,
+      "step": 3070
+    },
+    {
+      "entropy": 1.4384993091225624,
+      "epoch": 0.775745364102887,
+      "grad_norm": 0.58203125,
+      "learning_rate": 1.1231427851926466e-05,
+      "loss": 0.972,
+      "mean_token_accuracy": 0.7457866981625557,
+      "num_tokens": 286034044.0,
+      "step": 3080
+    },
+    {
+      "entropy": 1.474022053182125,
+      "epoch": 0.7782640178824418,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.1105514983631327e-05,
+      "loss": 1.0109,
+      "mean_token_accuracy": 0.7370796717703343,
+      "num_tokens": 286953119.0,
+      "step": 3090
+    },
+    {
+      "entropy": 1.4849003747105598,
+      "epoch": 0.7807826716619967,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.0979602115336188e-05,
+      "loss": 0.9955,
+      "mean_token_accuracy": 0.7422776162624359,
+      "num_tokens": 287893431.0,
+      "step": 3100
+    },
+    {
+      "entropy": 1.4983975902199744,
+      "epoch": 0.7833013254415515,
+      "grad_norm": 0.6328125,
+      "learning_rate": 1.0853689247041048e-05,
+      "loss": 0.9983,
+      "mean_token_accuracy": 0.7403753191232681,
+      "num_tokens": 288840355.0,
+      "step": 3110
+    },
+    {
+      "entropy": 1.5098316431045533,
+      "epoch": 0.7858199792211064,
+      "grad_norm": 0.64453125,
+      "learning_rate": 1.0727776378745909e-05,
+      "loss": 1.0409,
+      "mean_token_accuracy": 0.7355413317680359,
+      "num_tokens": 289761229.0,
+      "step": 3120
+    },
+    {
+      "entropy": 1.5024679169058799,
+      "epoch": 0.7883386330006611,
+      "grad_norm": 0.609375,
+      "learning_rate": 1.0601863510450768e-05,
+      "loss": 1.0068,
+      "mean_token_accuracy": 0.7369636178016663,
+      "num_tokens": 290690390.0,
+      "step": 3130
+    },
+    {
+      "entropy": 1.522695505619049,
+      "epoch": 0.7908572867802159,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.0475950642155628e-05,
+      "loss": 1.0523,
+      "mean_token_accuracy": 0.7308142200112343,
+      "num_tokens": 291636996.0,
+      "step": 3140
+    },
+    {
+      "entropy": 1.512753178179264,
+      "epoch": 0.7933759405597708,
+      "grad_norm": 0.55859375,
+      "learning_rate": 1.035003777386049e-05,
+      "loss": 1.0027,
+      "mean_token_accuracy": 0.7391382239758968,
+      "num_tokens": 292575769.0,
+      "step": 3150
+    },
+    {
+      "entropy": 1.49990826100111,
+      "epoch": 0.7958945943393256,
+      "grad_norm": 0.6171875,
+      "learning_rate": 1.022412490556535e-05,
+      "loss": 1.0215,
+      "mean_token_accuracy": 0.7357432976365089,
+      "num_tokens": 293492388.0,
+      "step": 3160
+    },
+    {
+      "entropy": 1.5039694808423518,
+      "epoch": 0.7984132481188805,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.0098212037270209e-05,
+      "loss": 1.0502,
+      "mean_token_accuracy": 0.7306272312998772,
+      "num_tokens": 294428696.0,
+      "step": 3170
+    },
+    {
+      "entropy": 1.5101190388202668,
+      "epoch": 0.8009319018984353,
+      "grad_norm": 0.6015625,
+      "learning_rate": 9.97229916897507e-06,
+      "loss": 1.0399,
+      "mean_token_accuracy": 0.7336853705346584,
+      "num_tokens": 295346544.0,
+      "step": 3180
+    },
+    {
+      "entropy": 1.50984176248312,
+      "epoch": 0.8034505556779901,
+      "grad_norm": 0.578125,
+      "learning_rate": 9.84638630067993e-06,
+      "loss": 1.0043,
+      "mean_token_accuracy": 0.7394494645297527,
+      "num_tokens": 296248020.0,
+      "step": 3190
+    },
+    {
+      "entropy": 1.474437852203846,
+      "epoch": 0.8059692094575449,
+      "grad_norm": 0.5625,
+      "learning_rate": 9.72047343238479e-06,
+      "loss": 0.9938,
+      "mean_token_accuracy": 0.7405206322669983,
+      "num_tokens": 297188131.0,
+      "step": 3200
+    },
+    {
+      "entropy": 1.498723168671131,
+      "epoch": 0.8084878632370998,
+      "grad_norm": 0.61328125,
+      "learning_rate": 9.594560564089652e-06,
+      "loss": 1.0374,
+      "mean_token_accuracy": 0.7325737461447716,
+      "num_tokens": 298109292.0,
+      "step": 3210
+    },
+    {
+      "entropy": 1.4560261756181716,
+      "epoch": 0.8110065170166546,
+      "grad_norm": 0.578125,
+      "learning_rate": 9.468647695794511e-06,
+      "loss": 0.9786,
+      "mean_token_accuracy": 0.7439539857208729,
+      "num_tokens": 299051156.0,
+      "step": 3220
+    },
+    {
+      "entropy": 1.491245013475418,
+      "epoch": 0.8135251707962095,
+      "grad_norm": 0.5859375,
+      "learning_rate": 9.34273482749937e-06,
+      "loss": 0.9968,
+      "mean_token_accuracy": 0.7427712038159371,
+      "num_tokens": 299999619.0,
+      "step": 3230
+    },
+    {
+      "entropy": 1.4857505142688752,
+      "epoch": 0.8160438245757643,
+      "grad_norm": 0.57421875,
+      "learning_rate": 9.216821959204232e-06,
+      "loss": 1.0178,
+      "mean_token_accuracy": 0.732306931167841,
+      "num_tokens": 300908853.0,
+      "step": 3240
+    },
+    {
+      "entropy": 1.4923428311944007,
+      "epoch": 0.818562478355319,
+      "grad_norm": 0.62890625,
+      "learning_rate": 9.090909090909091e-06,
+      "loss": 1.0103,
+      "mean_token_accuracy": 0.7387787699699402,
+      "num_tokens": 301830035.0,
+      "step": 3250
+    },
+    {
+      "entropy": 1.4485341638326645,
+      "epoch": 0.8210811321348739,
+      "grad_norm": 0.59375,
+      "learning_rate": 8.964996222613952e-06,
+      "loss": 0.949,
+      "mean_token_accuracy": 0.7519924528896809,
+      "num_tokens": 302774780.0,
+      "step": 3260
+    },
+    {
+      "entropy": 1.482700689136982,
+      "epoch": 0.8235997859144287,
+      "grad_norm": 0.6015625,
+      "learning_rate": 8.839083354318812e-06,
+      "loss": 1.0185,
+      "mean_token_accuracy": 0.7383209764957428,
+      "num_tokens": 303698822.0,
+      "step": 3270
+    },
+    {
+      "entropy": 1.5057204052805901,
+      "epoch": 0.8261184396939836,
+      "grad_norm": 0.57421875,
+      "learning_rate": 8.713170486023673e-06,
+      "loss": 1.023,
+      "mean_token_accuracy": 0.7377226896584034,
+      "num_tokens": 304620345.0,
+      "step": 3280
+    },
+    {
+      "entropy": 1.4904442220926284,
+      "epoch": 0.8286370934735384,
+      "grad_norm": 0.5390625,
+      "learning_rate": 8.587257617728532e-06,
+      "loss": 1.0087,
+      "mean_token_accuracy": 0.7416534572839737,
+      "num_tokens": 305551710.0,
+      "step": 3290
+    },
+    {
+      "entropy": 1.5239005982875824,
+      "epoch": 0.8311557472530933,
+      "grad_norm": 0.62890625,
+      "learning_rate": 8.461344749433391e-06,
+      "loss": 1.059,
+      "mean_token_accuracy": 0.7288159102201461,
+      "num_tokens": 306466206.0,
+      "step": 3300
+    },
+    {
+      "entropy": 1.513123245537281,
+      "epoch": 0.833674401032648,
+      "grad_norm": 0.625,
+      "learning_rate": 8.335431881138252e-06,
+      "loss": 1.0271,
+      "mean_token_accuracy": 0.7379589311778545,
+      "num_tokens": 307399178.0,
+      "step": 3310
+    },
+    {
+      "entropy": 1.4970497369766236,
+      "epoch": 0.8361930548122029,
+      "grad_norm": 0.59765625,
+      "learning_rate": 8.209519012843114e-06,
+      "loss": 0.9927,
+      "mean_token_accuracy": 0.7436629004776478,
+      "num_tokens": 308340686.0,
+      "step": 3320
+    },
+    {
+      "entropy": 1.4847823336720467,
+      "epoch": 0.8387117085917577,
+      "grad_norm": 0.59765625,
+      "learning_rate": 8.083606144547973e-06,
+      "loss": 1.0018,
+      "mean_token_accuracy": 0.7416875958442688,
+      "num_tokens": 309290930.0,
+      "step": 3330
+    },
+    {
+      "entropy": 1.4939886629581451,
+      "epoch": 0.8412303623713125,
+      "grad_norm": 0.66015625,
+      "learning_rate": 7.957693276252834e-06,
+      "loss": 1.0149,
+      "mean_token_accuracy": 0.736718226224184,
+      "num_tokens": 310228345.0,
+      "step": 3340
+    },
+    {
+      "entropy": 1.5165553405880927,
+      "epoch": 0.8437490161508674,
+      "grad_norm": 0.61328125,
+      "learning_rate": 7.831780407957693e-06,
+      "loss": 1.0194,
+      "mean_token_accuracy": 0.735667246580124,
+      "num_tokens": 311147031.0,
+      "step": 3350
+    },
+    {
+      "entropy": 1.5101538181304932,
+      "epoch": 0.8462676699304222,
+      "grad_norm": 0.5625,
+      "learning_rate": 7.705867539662553e-06,
+      "loss": 1.0143,
+      "mean_token_accuracy": 0.7377640530467033,
+      "num_tokens": 312071813.0,
+      "step": 3360
+    },
+    {
+      "entropy": 1.4858613476157188,
+      "epoch": 0.848786323709977,
+      "grad_norm": 0.5859375,
+      "learning_rate": 7.579954671367415e-06,
+      "loss": 1.0042,
+      "mean_token_accuracy": 0.7420022577047348,
+      "num_tokens": 313019044.0,
+      "step": 3370
+    },
+    {
+      "entropy": 1.4877135768532752,
+      "epoch": 0.8513049774895318,
+      "grad_norm": 0.66796875,
+      "learning_rate": 7.454041803072274e-06,
+      "loss": 1.0116,
+      "mean_token_accuracy": 0.7370616056025028,
+      "num_tokens": 313924648.0,
+      "step": 3380
+    },
+    {
+      "entropy": 1.4523219719529152,
+      "epoch": 0.8538236312690867,
+      "grad_norm": 0.5703125,
+      "learning_rate": 7.3281289347771344e-06,
+      "loss": 0.972,
+      "mean_token_accuracy": 0.7463509783148765,
+      "num_tokens": 314876822.0,
+      "step": 3390
+    },
+    {
+      "entropy": 1.5021595641970635,
+      "epoch": 0.8563422850486415,
+      "grad_norm": 0.671875,
+      "learning_rate": 7.2022160664819955e-06,
+      "loss": 1.0164,
+      "mean_token_accuracy": 0.7378572486341,
+      "num_tokens": 315830857.0,
+      "step": 3400
+    },
+    {
+      "entropy": 1.4737745821475983,
+      "epoch": 0.8588609388281964,
+      "grad_norm": 0.578125,
+      "learning_rate": 7.076303198186855e-06,
+      "loss": 0.992,
+      "mean_token_accuracy": 0.7421701058745385,
+      "num_tokens": 316765928.0,
+      "step": 3410
+    },
+    {
+      "entropy": 1.5143599852919578,
+      "epoch": 0.8613795926077512,
+      "grad_norm": 0.6015625,
+      "learning_rate": 6.950390329891715e-06,
+      "loss": 1.0368,
+      "mean_token_accuracy": 0.7320093587040901,
+      "num_tokens": 317700949.0,
+      "step": 3420
+    },
+    {
+      "entropy": 1.5028961822390556,
+      "epoch": 0.863898246387306,
+      "grad_norm": 0.6328125,
+      "learning_rate": 6.8244774615965745e-06,
+      "loss": 1.0503,
+      "mean_token_accuracy": 0.7298037901520729,
+      "num_tokens": 318642426.0,
+      "step": 3430
+    },
+    {
+      "entropy": 1.487327341735363,
+      "epoch": 0.8664169001668608,
+      "grad_norm": 0.6171875,
+      "learning_rate": 6.698564593301436e-06,
+      "loss": 0.9981,
+      "mean_token_accuracy": 0.7411175973713398,
+      "num_tokens": 319554341.0,
+      "step": 3440
+    },
+    {
+      "entropy": 1.4858398377895354,
+      "epoch": 0.8689355539464156,
+      "grad_norm": 0.671875,
+      "learning_rate": 6.572651725006296e-06,
+      "loss": 0.9707,
+      "mean_token_accuracy": 0.7463601998984813,
+      "num_tokens": 320469718.0,
+      "step": 3450
+    },
+    {
+      "entropy": 1.4977013349533081,
+      "epoch": 0.8714542077259705,
+      "grad_norm": 0.63671875,
+      "learning_rate": 6.446738856711155e-06,
+      "loss": 1.0111,
+      "mean_token_accuracy": 0.7389899030327797,
+      "num_tokens": 321408669.0,
+      "step": 3460
+    },
+    {
+      "entropy": 1.4885832622647286,
+      "epoch": 0.8739728615055253,
+      "grad_norm": 0.69140625,
+      "learning_rate": 6.320825988416016e-06,
+      "loss": 1.002,
+      "mean_token_accuracy": 0.7399603985249996,
+      "num_tokens": 322319368.0,
+      "step": 3470
+    },
+    {
+      "entropy": 1.5308593392372132,
+      "epoch": 0.8764915152850801,
+      "grad_norm": 0.59765625,
+      "learning_rate": 6.194913120120877e-06,
+      "loss": 1.0263,
+      "mean_token_accuracy": 0.7362756490707397,
+      "num_tokens": 323249158.0,
+      "step": 3480
+    },
+    {
+      "entropy": 1.489845983684063,
+      "epoch": 0.8790101690646349,
+      "grad_norm": 0.55859375,
+      "learning_rate": 6.069000251825737e-06,
+      "loss": 0.962,
+      "mean_token_accuracy": 0.7470691427588463,
+      "num_tokens": 324177591.0,
+      "step": 3490
+    },
+    {
+      "entropy": 1.4811480522155762,
+      "epoch": 0.8815288228441898,
+      "grad_norm": 0.6171875,
+      "learning_rate": 5.943087383530597e-06,
+      "loss": 0.9931,
+      "mean_token_accuracy": 0.74410410374403,
+      "num_tokens": 325086173.0,
+      "step": 3500
+    },
+    {
+      "entropy": 1.517330525815487,
+      "epoch": 0.8840474766237446,
+      "grad_norm": 0.625,
+      "learning_rate": 5.817174515235457e-06,
+      "loss": 1.0269,
+      "mean_token_accuracy": 0.7340725965797901,
+      "num_tokens": 325998476.0,
+      "step": 3510
+    },
+    {
+      "entropy": 1.4777013659477234,
+      "epoch": 0.8865661304032995,
+      "grad_norm": 0.61328125,
+      "learning_rate": 5.6912616469403176e-06,
+      "loss": 1.0367,
+      "mean_token_accuracy": 0.7329237192869187,
+      "num_tokens": 326888629.0,
+      "step": 3520
+    },
+    {
+      "entropy": 1.488082292675972,
+      "epoch": 0.8890847841828543,
+      "grad_norm": 0.5625,
+      "learning_rate": 5.565348778645178e-06,
+      "loss": 0.9912,
+      "mean_token_accuracy": 0.742741683870554,
+      "num_tokens": 327829681.0,
+      "step": 3530
+    },
+    {
+      "entropy": 1.4924616023898125,
+      "epoch": 0.891603437962409,
+      "grad_norm": 0.59765625,
+      "learning_rate": 5.439435910350038e-06,
+      "loss": 0.9738,
+      "mean_token_accuracy": 0.7450707465410232,
+      "num_tokens": 328777620.0,
+      "step": 3540
+    },
+    {
+      "entropy": 1.4866561606526374,
+      "epoch": 0.8941220917419639,
+      "grad_norm": 0.6015625,
+      "learning_rate": 5.313523042054898e-06,
+      "loss": 1.0161,
+      "mean_token_accuracy": 0.7406411737203598,
+      "num_tokens": 329712464.0,
+      "step": 3550
+    },
+    {
+      "entropy": 1.4689026042819022,
+      "epoch": 0.8966407455215187,
+      "grad_norm": 0.59375,
+      "learning_rate": 5.1876101737597585e-06,
+      "loss": 0.9691,
+      "mean_token_accuracy": 0.7475285783410073,
+      "num_tokens": 330641995.0,
+      "step": 3560
+    },
+    {
+      "entropy": 1.5136161372065544,
+      "epoch": 0.8991593993010736,
+      "grad_norm": 0.58984375,
+      "learning_rate": 5.061697305464619e-06,
+      "loss": 1.0013,
+      "mean_token_accuracy": 0.7394486844539643,
+      "num_tokens": 331568134.0,
+      "step": 3570
+    },
+    {
+      "entropy": 1.491369791328907,
+      "epoch": 0.9016780530806284,
+      "grad_norm": 0.55859375,
+      "learning_rate": 4.935784437169479e-06,
+      "loss": 1.0122,
+      "mean_token_accuracy": 0.735909091681242,
+      "num_tokens": 332501528.0,
+      "step": 3580
+    },
+    {
+      "entropy": 1.4946796283125878,
+      "epoch": 0.9041967068601833,
+      "grad_norm": 0.62890625,
+      "learning_rate": 4.809871568874339e-06,
+      "loss": 0.9903,
+      "mean_token_accuracy": 0.7426324136555195,
+      "num_tokens": 333409554.0,
+      "step": 3590
+    },
+    {
+      "entropy": 1.4981262043118477,
+      "epoch": 0.906715360639738,
+      "grad_norm": 0.609375,
+      "learning_rate": 4.6839587005791995e-06,
+      "loss": 1.0273,
+      "mean_token_accuracy": 0.7351961940526962,
+      "num_tokens": 334331813.0,
+      "step": 3600
+    },
+    {
+      "entropy": 1.4707158148288726,
+      "epoch": 0.9092340144192929,
+      "grad_norm": 0.59375,
+      "learning_rate": 4.55804583228406e-06,
+      "loss": 1.0127,
+      "mean_token_accuracy": 0.7383458167314529,
+      "num_tokens": 335268161.0,
+      "step": 3610
+    },
+    {
+      "entropy": 1.529105232656002,
+      "epoch": 0.9117526681988477,
+      "grad_norm": 0.63671875,
+      "learning_rate": 4.43213296398892e-06,
+      "loss": 1.0746,
+      "mean_token_accuracy": 0.7247563876211643,
+      "num_tokens": 336166223.0,
+      "step": 3620
+    },
+    {
+      "entropy": 1.4632112085819244,
+      "epoch": 0.9142713219784026,
+      "grad_norm": 0.640625,
+      "learning_rate": 4.30622009569378e-06,
+      "loss": 0.9751,
+      "mean_token_accuracy": 0.7464818298816681,
+      "num_tokens": 337114488.0,
+      "step": 3630
+    },
+    {
+      "entropy": 1.4965253099799156,
+      "epoch": 0.9167899757579574,
+      "grad_norm": 0.59765625,
+      "learning_rate": 4.1803072273986405e-06,
+      "loss": 1.0182,
+      "mean_token_accuracy": 0.7366202019155026,
+      "num_tokens": 338046633.0,
+      "step": 3640
+    },
+    {
+      "entropy": 1.4636427968740464,
+      "epoch": 0.9193086295375122,
+      "grad_norm": 0.640625,
+      "learning_rate": 4.054394359103501e-06,
+      "loss": 0.9868,
+      "mean_token_accuracy": 0.7432655967772007,
+      "num_tokens": 338968280.0,
+      "step": 3650
+    },
+    {
+      "entropy": 1.4739444583654404,
+      "epoch": 0.921827283317067,
+      "grad_norm": 0.5625,
+      "learning_rate": 3.928481490808361e-06,
+      "loss": 0.9648,
+      "mean_token_accuracy": 0.7487529583275319,
+      "num_tokens": 339893347.0,
+      "step": 3660
+    },
+    {
+      "entropy": 1.415738572180271,
+      "epoch": 0.9243459370966218,
+      "grad_norm": 0.61328125,
+      "learning_rate": 3.8025686225132208e-06,
+      "loss": 0.9282,
+      "mean_token_accuracy": 0.756267712265253,
+      "num_tokens": 340816571.0,
+      "step": 3670
+    },
+    {
+      "entropy": 1.4868112698197364,
+      "epoch": 0.9268645908761767,
+      "grad_norm": 0.609375,
+      "learning_rate": 3.6766557542180814e-06,
+      "loss": 1.0149,
+      "mean_token_accuracy": 0.7356903217732906,
+      "num_tokens": 341727649.0,
+      "step": 3680
+    },
+    {
+      "entropy": 1.489026002585888,
+      "epoch": 0.9293832446557315,
+      "grad_norm": 0.59375,
+      "learning_rate": 3.5507428859229413e-06,
+      "loss": 0.9725,
+      "mean_token_accuracy": 0.7466869972646236,
+      "num_tokens": 342670871.0,
+      "step": 3690
+    },
+    {
+      "entropy": 1.4483363971114158,
+      "epoch": 0.9319018984352864,
+      "grad_norm": 0.54296875,
+      "learning_rate": 3.4248300176278015e-06,
+      "loss": 0.9528,
+      "mean_token_accuracy": 0.7525399282574654,
+      "num_tokens": 343575810.0,
+      "step": 3700
+    },
+    {
+      "entropy": 1.4972086012363435,
+      "epoch": 0.9344205522148412,
+      "grad_norm": 0.59765625,
+      "learning_rate": 3.298917149332662e-06,
+      "loss": 0.9881,
+      "mean_token_accuracy": 0.7425376117229462,
+      "num_tokens": 344505435.0,
+      "step": 3710
+    },
+    {
+      "entropy": 1.4751506879925729,
+      "epoch": 0.936939205994396,
+      "grad_norm": 0.56640625,
+      "learning_rate": 3.173004281037522e-06,
+      "loss": 0.9995,
+      "mean_token_accuracy": 0.7415733940899372,
+      "num_tokens": 345454166.0,
+      "step": 3720
+    },
+    {
+      "entropy": 1.4912448197603225,
+      "epoch": 0.9394578597739508,
+      "grad_norm": 0.58984375,
+      "learning_rate": 3.0470914127423827e-06,
+      "loss": 1.0192,
+      "mean_token_accuracy": 0.7367800146341323,
+      "num_tokens": 346361862.0,
+      "step": 3730
+    },
+    {
+      "entropy": 1.5106546625494957,
+      "epoch": 0.9419765135535056,
+      "grad_norm": 0.62890625,
+      "learning_rate": 2.9211785444472425e-06,
+      "loss": 1.045,
+      "mean_token_accuracy": 0.7318918511271477,
+      "num_tokens": 347306780.0,
+      "step": 3740
+    },
+    {
+      "entropy": 1.4863861680030823,
+      "epoch": 0.9444951673330605,
+      "grad_norm": 0.625,
+      "learning_rate": 2.7952656761521027e-06,
+      "loss": 0.9934,
+      "mean_token_accuracy": 0.7438393741846084,
+      "num_tokens": 348230042.0,
+      "step": 3750
+    },
+    {
+      "entropy": 1.4912439316511155,
+      "epoch": 0.9470138211126153,
+      "grad_norm": 0.62109375,
+      "learning_rate": 2.6693528078569634e-06,
+      "loss": 0.9985,
+      "mean_token_accuracy": 0.7412358790636062,
+      "num_tokens": 349130792.0,
+      "step": 3760
+    },
+    {
+      "entropy": 1.5144632264971734,
+      "epoch": 0.9495324748921702,
+      "grad_norm": 0.66015625,
+      "learning_rate": 2.543439939561823e-06,
+      "loss": 1.0468,
+      "mean_token_accuracy": 0.7318057663738727,
+      "num_tokens": 350050566.0,
+      "step": 3770
+    },
+    {
+      "entropy": 1.4877790048718453,
+      "epoch": 0.9520511286717249,
+      "grad_norm": 0.5546875,
+      "learning_rate": 2.4175270712666834e-06,
+      "loss": 1.0022,
+      "mean_token_accuracy": 0.7390140548348427,
+      "num_tokens": 350980304.0,
+      "step": 3780
+    },
+    {
+      "entropy": 1.4984122559428215,
+      "epoch": 0.9545697824512798,
+      "grad_norm": 0.62890625,
+      "learning_rate": 2.2916142029715437e-06,
+      "loss": 0.9951,
+      "mean_token_accuracy": 0.7424575209617614,
+      "num_tokens": 351911007.0,
+      "step": 3790
+    },
+    {
+      "entropy": 1.495207953453064,
+      "epoch": 0.9570884362308346,
+      "grad_norm": 0.61328125,
+      "learning_rate": 2.1657013346764043e-06,
+      "loss": 1.015,
+      "mean_token_accuracy": 0.7366439893841743,
+      "num_tokens": 352846519.0,
+      "step": 3800
+    },
+    {
+      "entropy": 1.481141744554043,
+      "epoch": 0.9596070900103895,
+      "grad_norm": 0.5703125,
+      "learning_rate": 2.039788466381264e-06,
+      "loss": 1.0275,
+      "mean_token_accuracy": 0.7345097243785859,
+      "num_tokens": 353782132.0,
+      "step": 3810
+    },
+    {
+      "entropy": 1.4692917168140411,
+      "epoch": 0.9621257437899443,
+      "grad_norm": 0.62890625,
+      "learning_rate": 1.9138755980861244e-06,
+      "loss": 0.9799,
+      "mean_token_accuracy": 0.744342253357172,
+      "num_tokens": 354668645.0,
+      "step": 3820
+    },
+    {
+      "entropy": 1.5071855306625366,
+      "epoch": 0.9646443975694992,
+      "grad_norm": 0.56640625,
+      "learning_rate": 1.7879627297909846e-06,
+      "loss": 1.0593,
+      "mean_token_accuracy": 0.7290410205721856,
+      "num_tokens": 355600887.0,
+      "step": 3830
+    },
+    {
+      "entropy": 1.5132853239774704,
+      "epoch": 0.9671630513490539,
+      "grad_norm": 0.61328125,
+      "learning_rate": 1.662049861495845e-06,
+      "loss": 1.0435,
+      "mean_token_accuracy": 0.7298006169497967,
+      "num_tokens": 356522841.0,
+      "step": 3840
+    },
+    {
+      "entropy": 1.4672839909791946,
+      "epoch": 0.9696817051286087,
+      "grad_norm": 0.60546875,
+      "learning_rate": 1.5361369932007051e-06,
+      "loss": 0.9753,
+      "mean_token_accuracy": 0.7466705776751041,
+      "num_tokens": 357452055.0,
+      "step": 3850
+    },
+    {
+      "entropy": 1.4822666108608247,
+      "epoch": 0.9722003589081636,
+      "grad_norm": 0.6328125,
+      "learning_rate": 1.4102241249055656e-06,
+      "loss": 1.0259,
+      "mean_token_accuracy": 0.7341479398310184,
+      "num_tokens": 358377891.0,
+      "step": 3860
+    },
+    {
+      "entropy": 1.4937451481819153,
+      "epoch": 0.9747190126877184,
+      "grad_norm": 0.578125,
+      "learning_rate": 1.2843112566104256e-06,
+      "loss": 0.9759,
+      "mean_token_accuracy": 0.7456471361219883,
+      "num_tokens": 359300390.0,
+      "step": 3870
+    },
+    {
+      "entropy": 1.5173174068331718,
+      "epoch": 0.9772376664672733,
+      "grad_norm": 0.6015625,
+      "learning_rate": 1.1583983883152859e-06,
+      "loss": 1.0301,
+      "mean_token_accuracy": 0.7358597233891487,
+      "num_tokens": 360238533.0,
+      "step": 3880
+    },
+    {
+      "entropy": 1.4928388759493827,
+      "epoch": 0.979756320246828,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.032485520020146e-06,
+      "loss": 1.0117,
+      "mean_token_accuracy": 0.7398792788386345,
+      "num_tokens": 361150526.0,
+      "step": 3890
+    },
+    {
+      "entropy": 1.5063897222280502,
+      "epoch": 0.9822749740263829,
+      "grad_norm": 0.62109375,
+      "learning_rate": 9.065726517250064e-07,
+      "loss": 1.017,
+      "mean_token_accuracy": 0.7376690439879894,
+      "num_tokens": 362100458.0,
+      "step": 3900
+    },
+    {
+      "entropy": 1.4783270865678788,
+      "epoch": 0.9847936278059377,
+      "grad_norm": 0.6328125,
+      "learning_rate": 7.806597834298666e-07,
+      "loss": 1.0081,
+      "mean_token_accuracy": 0.7380148410797119,
+      "num_tokens": 363014425.0,
+      "step": 3910
+    },
+    {
+      "entropy": 1.5196350157260894,
+      "epoch": 0.9873122815854926,
+      "grad_norm": 0.6328125,
+      "learning_rate": 6.547469151347267e-07,
+      "loss": 1.02,
+      "mean_token_accuracy": 0.7349734269082546,
+      "num_tokens": 363929023.0,
+      "step": 3920
+    },
+    {
+      "entropy": 1.4695521786808967,
+      "epoch": 0.9898309353650474,
+      "grad_norm": 0.59375,
+      "learning_rate": 5.28834046839587e-07,
+      "loss": 0.9597,
+      "mean_token_accuracy": 0.7482846342027187,
+      "num_tokens": 364898379.0,
+      "step": 3930
+    },
+    {
+      "entropy": 1.4966929897665977,
+      "epoch": 0.9923495891446023,
+      "grad_norm": 0.6328125,
+      "learning_rate": 4.029211785444472e-07,
+      "loss": 1.0137,
+      "mean_token_accuracy": 0.7382561720907688,
+      "num_tokens": 365874642.0,
+      "step": 3940
+    },
+    {
+      "entropy": 1.485151782631874,
+      "epoch": 0.994868242924157,
+      "grad_norm": 0.58984375,
+      "learning_rate": 2.770083102493075e-07,
+      "loss": 0.9854,
+      "mean_token_accuracy": 0.7435578644275666,
+      "num_tokens": 366766965.0,
+      "step": 3950
+    },
+    {
+      "entropy": 1.4845439344644547,
+      "epoch": 0.9973868967037118,
+      "grad_norm": 0.58984375,
+      "learning_rate": 1.5109544195416771e-07,
+      "loss": 0.9723,
+      "mean_token_accuracy": 0.7472767226397992,
+      "num_tokens": 367711087.0,
+      "step": 3960
+    },
+    {
+      "entropy": 1.5110925331711769,
+      "epoch": 0.9999055504832667,
+      "grad_norm": 0.6328125,
+      "learning_rate": 2.518257365902795e-08,
+      "loss": 1.0354,
+      "mean_token_accuracy": 0.7324677541851997,
+      "num_tokens": 368669225.0,
+      "step": 3970
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 3971,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.986803437820707e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}