diff --git "a/checkpoint-3000/trainer_state.json" "b/checkpoint-3000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-3000/trainer_state.json"
@@ -0,0 +1,6034 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.15430908108942212,
+  "eval_steps": 500,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 10.7425612449646,
+      "epoch": 0.0002571818018157035,
+      "grad_norm": 5.65625,
+      "learning_rate": 2e-06,
+      "loss": 10.7388,
+      "mean_token_accuracy": 0.0,
+      "num_tokens": 9030.0,
+      "step": 5
+    },
+    {
+      "entropy": 10.742591857910156,
+      "epoch": 0.000514363603631407,
+      "grad_norm": 5.25,
+      "learning_rate": 4.5e-06,
+      "loss": 10.7331,
+      "mean_token_accuracy": 0.0,
+      "num_tokens": 18084.0,
+      "step": 10
+    },
+    {
+      "entropy": 10.742617130279541,
+      "epoch": 0.0007715454054471106,
+      "grad_norm": 5.1875,
+      "learning_rate": 7e-06,
+      "loss": 10.6991,
+      "mean_token_accuracy": 0.0,
+      "num_tokens": 26309.0,
+      "step": 15
+    },
+    {
+      "entropy": 10.742614936828613,
+      "epoch": 0.001028727207262814,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.5e-06,
+      "loss": 10.661,
+      "mean_token_accuracy": 0.0013652202673256398,
+      "num_tokens": 35204.0,
+      "step": 20
+    },
+    {
+      "entropy": 10.74255428314209,
+      "epoch": 0.0012859090090785177,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.2e-05,
+      "loss": 10.5708,
+      "mean_token_accuracy": 0.009783844836056233,
+      "num_tokens": 44540.0,
+      "step": 25
+    },
+    {
+      "entropy": 10.74219331741333,
+      "epoch": 0.0015430908108942211,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.4500000000000002e-05,
+      "loss": 10.4335,
+      "mean_token_accuracy": 0.035427753999829295,
+      "num_tokens": 53118.0,
+      "step": 30
+    },
+    {
+      "entropy": 10.740796661376953,
+      "epoch": 0.0018002726127099246,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 10.3091,
+      "mean_token_accuracy": 0.04362985044717789,
+      "num_tokens": 61753.0,
+      "step": 35
+    },
+    {
+      "entropy": 10.736623382568359,
+      "epoch": 0.002057454414525628,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.95e-05,
+      "loss": 10.1821,
+      "mean_token_accuracy": 0.044929017499089244,
+      "num_tokens": 71102.0,
+      "step": 40
+    },
+    {
+      "entropy": 10.729742622375488,
+      "epoch": 0.0023146362163413317,
+      "grad_norm": 2.3125,
+      "learning_rate": 2.2e-05,
+      "loss": 10.0932,
+      "mean_token_accuracy": 0.04235424343496561,
+      "num_tokens": 80094.0,
+      "step": 45
+    },
+    {
+      "entropy": 10.723683643341065,
+      "epoch": 0.0025718180181570354,
+      "grad_norm": 2.109375,
+      "learning_rate": 2.4500000000000003e-05,
+      "loss": 9.9838,
+      "mean_token_accuracy": 0.04485577754676342,
+      "num_tokens": 88770.0,
+      "step": 50
+    },
+    {
+      "entropy": 10.720681381225585,
+      "epoch": 0.0028289998199727386,
+      "grad_norm": 2.046875,
+      "learning_rate": 2.7e-05,
+      "loss": 9.944,
+      "mean_token_accuracy": 0.043965942412614825,
+      "num_tokens": 97608.0,
+      "step": 55
+    },
+    {
+      "entropy": 10.717811107635498,
+      "epoch": 0.0030861816217884423,
+      "grad_norm": 1.875,
+      "learning_rate": 2.95e-05,
+      "loss": 9.8684,
+      "mean_token_accuracy": 0.04390094466507435,
+      "num_tokens": 106204.0,
+      "step": 60
+    },
+    {
+      "entropy": 10.713484477996825,
+      "epoch": 0.003343363423604146,
+      "grad_norm": 1.90625,
+      "learning_rate": 3.2e-05,
+      "loss": 9.841,
+      "mean_token_accuracy": 0.04340735748410225,
+      "num_tokens": 115241.0,
+      "step": 65
+    },
+    {
+      "entropy": 10.708874225616455,
+      "epoch": 0.003600545225419849,
+      "grad_norm": 1.984375,
+      "learning_rate": 3.4500000000000005e-05,
+      "loss": 9.7416,
+      "mean_token_accuracy": 0.04423734098672867,
+      "num_tokens": 123734.0,
+      "step": 70
+    },
+    {
+      "entropy": 10.703791618347168,
+      "epoch": 0.003857727027235553,
+      "grad_norm": 1.8984375,
+      "learning_rate": 3.7e-05,
+      "loss": 9.7488,
+      "mean_token_accuracy": 0.04389217719435692,
+      "num_tokens": 132701.0,
+      "step": 75
+    },
+    {
+      "entropy": 10.697965526580811,
+      "epoch": 0.004114908829051256,
+      "grad_norm": 1.921875,
+      "learning_rate": 3.95e-05,
+      "loss": 9.6173,
+      "mean_token_accuracy": 0.04561489634215832,
+      "num_tokens": 141064.0,
+      "step": 80
+    },
+    {
+      "entropy": 10.6871018409729,
+      "epoch": 0.00437209063086696,
+      "grad_norm": 1.96875,
+      "learning_rate": 4.2000000000000004e-05,
+      "loss": 9.5084,
+      "mean_token_accuracy": 0.05098861046135426,
+      "num_tokens": 149340.0,
+      "step": 85
+    },
+    {
+      "entropy": 10.67187786102295,
+      "epoch": 0.004629272432682663,
+      "grad_norm": 1.828125,
+      "learning_rate": 4.45e-05,
+      "loss": 9.4949,
+      "mean_token_accuracy": 0.05393466129899025,
+      "num_tokens": 158312.0,
+      "step": 90
+    },
+    {
+      "entropy": 10.659277820587159,
+      "epoch": 0.004886454234498367,
+      "grad_norm": 1.640625,
+      "learning_rate": 4.7000000000000004e-05,
+      "loss": 9.445,
+      "mean_token_accuracy": 0.0573005523532629,
+      "num_tokens": 167203.0,
+      "step": 95
+    },
+    {
+      "entropy": 10.643789291381836,
+      "epoch": 0.005143636036314071,
+      "grad_norm": 1.8046875,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 9.2598,
+      "mean_token_accuracy": 0.0629499789327383,
+      "num_tokens": 175929.0,
+      "step": 100
+    },
+    {
+      "entropy": 10.604194068908692,
+      "epoch": 0.0054008178381297735,
+      "grad_norm": 1.7734375,
+      "learning_rate": 5.2e-05,
+      "loss": 9.2354,
+      "mean_token_accuracy": 0.05866244621574879,
+      "num_tokens": 185619.0,
+      "step": 105
+    },
+    {
+      "entropy": 10.549863052368163,
+      "epoch": 0.005657999639945477,
+      "grad_norm": 1.6875,
+      "learning_rate": 5.45e-05,
+      "loss": 9.1507,
+      "mean_token_accuracy": 0.05989648550748825,
+      "num_tokens": 195005.0,
+      "step": 110
+    },
+    {
+      "entropy": 10.489420223236085,
+      "epoch": 0.005915181441761181,
+      "grad_norm": 1.59375,
+      "learning_rate": 5.7e-05,
+      "loss": 8.9373,
+      "mean_token_accuracy": 0.0647052500396967,
+      "num_tokens": 203392.0,
+      "step": 115
+    },
+    {
+      "entropy": 10.426068687438965,
+      "epoch": 0.0061723632435768845,
+      "grad_norm": 1.578125,
+      "learning_rate": 5.9499999999999996e-05,
+      "loss": 8.8217,
+      "mean_token_accuracy": 0.060671578347682956,
+      "num_tokens": 211689.0,
+      "step": 120
+    },
+    {
+      "entropy": 10.359361457824708,
+      "epoch": 0.006429545045392588,
+      "grad_norm": 1.4296875,
+      "learning_rate": 6.2e-05,
+      "loss": 8.756,
+      "mean_token_accuracy": 0.06377223357558251,
+      "num_tokens": 220798.0,
+      "step": 125
+    },
+    {
+      "entropy": 10.30004062652588,
+      "epoch": 0.006686726847208292,
+      "grad_norm": 1.4453125,
+      "learning_rate": 6.450000000000001e-05,
+      "loss": 8.6622,
+      "mean_token_accuracy": 0.06106479801237583,
+      "num_tokens": 229652.0,
+      "step": 130
+    },
+    {
+      "entropy": 10.221406269073487,
+      "epoch": 0.006943908649023995,
+      "grad_norm": 1.453125,
+      "learning_rate": 6.7e-05,
+      "loss": 8.4588,
+      "mean_token_accuracy": 0.06428722217679024,
+      "num_tokens": 238320.0,
+      "step": 135
+    },
+    {
+      "entropy": 10.105729293823241,
+      "epoch": 0.007201090450839698,
+      "grad_norm": 1.5,
+      "learning_rate": 6.950000000000001e-05,
+      "loss": 8.431,
+      "mean_token_accuracy": 0.06200639493763447,
+      "num_tokens": 247033.0,
+      "step": 140
+    },
+    {
+      "entropy": 9.967057228088379,
+      "epoch": 0.007458272252655402,
+      "grad_norm": 1.203125,
+      "learning_rate": 7.2e-05,
+      "loss": 8.3441,
+      "mean_token_accuracy": 0.06121456101536751,
+      "num_tokens": 256666.0,
+      "step": 145
+    },
+    {
+      "entropy": 9.81930513381958,
+      "epoch": 0.007715454054471106,
+      "grad_norm": 1.15625,
+      "learning_rate": 7.45e-05,
+      "loss": 8.1686,
+      "mean_token_accuracy": 0.06682926267385483,
+      "num_tokens": 265689.0,
+      "step": 150
+    },
+    {
+      "entropy": 9.67783327102661,
+      "epoch": 0.00797263585628681,
+      "grad_norm": 0.9921875,
+      "learning_rate": 7.7e-05,
+      "loss": 8.1019,
+      "mean_token_accuracy": 0.06226279996335506,
+      "num_tokens": 274276.0,
+      "step": 155
+    },
+    {
+      "entropy": 9.495966053009033,
+      "epoch": 0.008229817658102512,
+      "grad_norm": 1.0703125,
+      "learning_rate": 7.950000000000001e-05,
+      "loss": 7.9704,
+      "mean_token_accuracy": 0.06119140088558197,
+      "num_tokens": 283825.0,
+      "step": 160
+    },
+    {
+      "entropy": 9.309716987609864,
+      "epoch": 0.008486999459918217,
+      "grad_norm": 0.96484375,
+      "learning_rate": 8.2e-05,
+      "loss": 7.8795,
+      "mean_token_accuracy": 0.06309534795582294,
+      "num_tokens": 292223.0,
+      "step": 165
+    },
+    {
+      "entropy": 9.042353534698487,
+      "epoch": 0.00874418126173392,
+      "grad_norm": 0.8046875,
+      "learning_rate": 8.450000000000001e-05,
+      "loss": 7.8385,
+      "mean_token_accuracy": 0.06506023034453393,
+      "num_tokens": 300889.0,
+      "step": 170
+    },
+    {
+      "entropy": 8.90816011428833,
+      "epoch": 0.009001363063549624,
+      "grad_norm": 1.0078125,
+      "learning_rate": 8.7e-05,
+      "loss": 7.7842,
+      "mean_token_accuracy": 0.0640747856348753,
+      "num_tokens": 309729.0,
+      "step": 175
+    },
+    {
+      "entropy": 8.687435150146484,
+      "epoch": 0.009258544865365327,
+      "grad_norm": 0.8125,
+      "learning_rate": 8.95e-05,
+      "loss": 7.7211,
+      "mean_token_accuracy": 0.060701503232121465,
+      "num_tokens": 318611.0,
+      "step": 180
+    },
+    {
+      "entropy": 8.534039783477784,
+      "epoch": 0.00951572666718103,
+      "grad_norm": 0.8828125,
+      "learning_rate": 9.2e-05,
+      "loss": 7.6722,
+      "mean_token_accuracy": 0.06427375338971615,
+      "num_tokens": 326948.0,
+      "step": 185
+    },
+    {
+      "entropy": 8.478750228881836,
+      "epoch": 0.009772908468996734,
+      "grad_norm": 0.81640625,
+      "learning_rate": 9.45e-05,
+      "loss": 7.6806,
+      "mean_token_accuracy": 0.06457847729325294,
+      "num_tokens": 336249.0,
+      "step": 190
+    },
+    {
+      "entropy": 8.388401317596436,
+      "epoch": 0.010030090270812437,
+      "grad_norm": 0.7265625,
+      "learning_rate": 9.7e-05,
+      "loss": 7.6163,
+      "mean_token_accuracy": 0.06882799975574017,
+      "num_tokens": 345182.0,
+      "step": 195
+    },
+    {
+      "entropy": 8.318638229370118,
+      "epoch": 0.010287272072628141,
+      "grad_norm": 0.79296875,
+      "learning_rate": 9.95e-05,
+      "loss": 7.632,
+      "mean_token_accuracy": 0.06467234715819359,
+      "num_tokens": 353642.0,
+      "step": 200
+    },
+    {
+      "entropy": 8.238949489593505,
+      "epoch": 0.010544453874443844,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.000102,
+      "loss": 7.6098,
+      "mean_token_accuracy": 0.06644487231969834,
+      "num_tokens": 362269.0,
+      "step": 205
+    },
+    {
+      "entropy": 8.207266712188721,
+      "epoch": 0.010801635676259547,
+      "grad_norm": 0.92578125,
+      "learning_rate": 0.00010449999999999999,
+      "loss": 7.5609,
+      "mean_token_accuracy": 0.06878665201365948,
+      "num_tokens": 371024.0,
+      "step": 210
+    },
+    {
+      "entropy": 8.225072002410888,
+      "epoch": 0.011058817478075252,
+      "grad_norm": 0.8671875,
+      "learning_rate": 0.000107,
+      "loss": 7.5979,
+      "mean_token_accuracy": 0.06695215404033661,
+      "num_tokens": 379876.0,
+      "step": 215
+    },
+    {
+      "entropy": 8.130556297302245,
+      "epoch": 0.011315999279890954,
+      "grad_norm": 0.8671875,
+      "learning_rate": 0.0001095,
+      "loss": 7.6058,
+      "mean_token_accuracy": 0.06531222015619279,
+      "num_tokens": 388907.0,
+      "step": 220
+    },
+    {
+      "entropy": 8.187319278717041,
+      "epoch": 0.011573181081706659,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.000112,
+      "loss": 7.5483,
+      "mean_token_accuracy": 0.0684064768254757,
+      "num_tokens": 397629.0,
+      "step": 225
+    },
+    {
+      "entropy": 8.132099628448486,
+      "epoch": 0.011830362883522362,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.0001145,
+      "loss": 7.5356,
+      "mean_token_accuracy": 0.06274137385189534,
+      "num_tokens": 406919.0,
+      "step": 230
+    },
+    {
+      "entropy": 8.15430727005005,
+      "epoch": 0.012087544685338066,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.00011700000000000001,
+      "loss": 7.5962,
+      "mean_token_accuracy": 0.068911187723279,
+      "num_tokens": 415471.0,
+      "step": 235
+    },
+    {
+      "entropy": 8.099409580230713,
+      "epoch": 0.012344726487153769,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00011949999999999999,
+      "loss": 7.5735,
+      "mean_token_accuracy": 0.0665768850594759,
+      "num_tokens": 424532.0,
+      "step": 240
+    },
+    {
+      "entropy": 8.095283269882202,
+      "epoch": 0.012601908288969472,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.000122,
+      "loss": 7.5317,
+      "mean_token_accuracy": 0.06533802710473538,
+      "num_tokens": 433457.0,
+      "step": 245
+    },
+    {
+      "entropy": 8.048843669891358,
+      "epoch": 0.012859090090785176,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.0001245,
+      "loss": 7.5112,
+      "mean_token_accuracy": 0.06810796111822129,
+      "num_tokens": 442765.0,
+      "step": 250
+    },
+    {
+      "entropy": 8.033838605880737,
+      "epoch": 0.01311627189260088,
+      "grad_norm": 0.890625,
+      "learning_rate": 0.000127,
+      "loss": 7.5382,
+      "mean_token_accuracy": 0.06802100911736489,
+      "num_tokens": 451853.0,
+      "step": 255
+    },
+    {
+      "entropy": 8.045299959182739,
+      "epoch": 0.013373453694416584,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0001295,
+      "loss": 7.4256,
+      "mean_token_accuracy": 0.07066994272172451,
+      "num_tokens": 460210.0,
+      "step": 260
+    },
+    {
+      "entropy": 7.9866725444793705,
+      "epoch": 0.013630635496232286,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.000132,
+      "loss": 7.4002,
+      "mean_token_accuracy": 0.08060714676976204,
+      "num_tokens": 468673.0,
+      "step": 265
+    },
+    {
+      "entropy": 8.092905473709106,
+      "epoch": 0.01388781729804799,
+      "grad_norm": 0.94140625,
+      "learning_rate": 0.00013450000000000002,
+      "loss": 7.4123,
+      "mean_token_accuracy": 0.07691804766654968,
+      "num_tokens": 477194.0,
+      "step": 270
+    },
+    {
+      "entropy": 7.961445760726929,
+      "epoch": 0.014144999099863694,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00013700000000000002,
+      "loss": 7.381,
+      "mean_token_accuracy": 0.07125308774411679,
+      "num_tokens": 486869.0,
+      "step": 275
+    },
+    {
+      "entropy": 7.966795444488525,
+      "epoch": 0.014402180901679397,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0001395,
+      "loss": 7.3379,
+      "mean_token_accuracy": 0.07665478959679603,
+      "num_tokens": 495731.0,
+      "step": 280
+    },
+    {
+      "entropy": 7.958342027664185,
+      "epoch": 0.014659362703495101,
+      "grad_norm": 0.94140625,
+      "learning_rate": 0.00014199999999999998,
+      "loss": 7.3933,
+      "mean_token_accuracy": 0.07646878808736801,
+      "num_tokens": 504071.0,
+      "step": 285
+    },
+    {
+      "entropy": 7.9638275623321535,
+      "epoch": 0.014916544505310804,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0001445,
+      "loss": 7.4247,
+      "mean_token_accuracy": 0.07719123288989067,
+      "num_tokens": 512927.0,
+      "step": 290
+    },
+    {
+      "entropy": 7.965741872787476,
+      "epoch": 0.015173726307126508,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.000147,
+      "loss": 7.2639,
+      "mean_token_accuracy": 0.07586801275610924,
+      "num_tokens": 521405.0,
+      "step": 295
+    },
+    {
+      "entropy": 7.823179912567139,
+      "epoch": 0.015430908108942211,
+      "grad_norm": 0.8984375,
+      "learning_rate": 0.0001495,
+      "loss": 7.3976,
+      "mean_token_accuracy": 0.07964370101690292,
+      "num_tokens": 530473.0,
+      "step": 300
+    },
+    {
+      "entropy": 7.85285005569458,
+      "epoch": 0.015688089910757916,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.000152,
+      "loss": 7.3457,
+      "mean_token_accuracy": 0.07974019274115562,
+      "num_tokens": 540058.0,
+      "step": 305
+    },
+    {
+      "entropy": 7.974370861053467,
+      "epoch": 0.01594527171257362,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.00015450000000000001,
+      "loss": 7.407,
+      "mean_token_accuracy": 0.07686455249786377,
+      "num_tokens": 548492.0,
+      "step": 310
+    },
+    {
+      "entropy": 7.900070238113403,
+      "epoch": 0.01620245351438932,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.000157,
+      "loss": 7.313,
+      "mean_token_accuracy": 0.07973849661648273,
+      "num_tokens": 556788.0,
+      "step": 315
+    },
+    {
+      "entropy": 7.812545537948608,
+      "epoch": 0.016459635316205024,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0001595,
+      "loss": 7.2565,
+      "mean_token_accuracy": 0.07418388202786445,
+      "num_tokens": 566122.0,
+      "step": 320
+    },
+    {
+      "entropy": 7.9200437545776365,
+      "epoch": 0.01671681711802073,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000162,
+      "loss": 7.2952,
+      "mean_token_accuracy": 0.07806292176246643,
+      "num_tokens": 575574.0,
+      "step": 325
+    },
+    {
+      "entropy": 7.8590294361114506,
+      "epoch": 0.016973998919836433,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00016450000000000001,
+      "loss": 7.3445,
+      "mean_token_accuracy": 0.07793254256248475,
+      "num_tokens": 584552.0,
+      "step": 330
+    },
+    {
+      "entropy": 7.8465721130371096,
+      "epoch": 0.017231180721652136,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00016700000000000002,
+      "loss": 7.3252,
+      "mean_token_accuracy": 0.07768969610333443,
+      "num_tokens": 593572.0,
+      "step": 335
+    },
+    {
+      "entropy": 7.80656795501709,
+      "epoch": 0.01748836252346784,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00016950000000000003,
+      "loss": 7.2293,
+      "mean_token_accuracy": 0.07511881664395333,
+      "num_tokens": 603205.0,
+      "step": 340
+    },
+    {
+      "entropy": 7.84014139175415,
+      "epoch": 0.01774554432528354,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 7.2884,
+      "mean_token_accuracy": 0.07499681413173676,
+      "num_tokens": 612268.0,
+      "step": 345
+    },
+    {
+      "entropy": 7.784244871139526,
+      "epoch": 0.018002726127099248,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00017449999999999999,
+      "loss": 7.2643,
+      "mean_token_accuracy": 0.07596472017467022,
+      "num_tokens": 621213.0,
+      "step": 350
+    },
+    {
+      "entropy": 7.803834342956543,
+      "epoch": 0.01825990792891495,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.000177,
+      "loss": 7.3069,
+      "mean_token_accuracy": 0.08319237679243088,
+      "num_tokens": 630231.0,
+      "step": 355
+    },
+    {
+      "entropy": 7.815704870223999,
+      "epoch": 0.018517089730730654,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0001795,
+      "loss": 7.2149,
+      "mean_token_accuracy": 0.0825786791741848,
+      "num_tokens": 639217.0,
+      "step": 360
+    },
+    {
+      "entropy": 7.837387752532959,
+      "epoch": 0.018774271532546356,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000182,
+      "loss": 7.2476,
+      "mean_token_accuracy": 0.07730407565832138,
+      "num_tokens": 648233.0,
+      "step": 365
+    },
+    {
+      "entropy": 7.710758781433105,
+      "epoch": 0.01903145333436206,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.0001845,
+      "loss": 7.2316,
+      "mean_token_accuracy": 0.08014441356062889,
+      "num_tokens": 657817.0,
+      "step": 370
+    },
+    {
+      "entropy": 7.747053003311157,
+      "epoch": 0.019288635136177765,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000187,
+      "loss": 7.1701,
+      "mean_token_accuracy": 0.08062060177326202,
+      "num_tokens": 666335.0,
+      "step": 375
+    },
+    {
+      "entropy": 7.7258199691772464,
+      "epoch": 0.019545816937993468,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0001895,
+      "loss": 7.0682,
+      "mean_token_accuracy": 0.08676183447241784,
+      "num_tokens": 674766.0,
+      "step": 380
+    },
+    {
+      "entropy": 7.720611429214477,
+      "epoch": 0.01980299873980917,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.000192,
+      "loss": 7.1426,
+      "mean_token_accuracy": 0.0820883870124817,
+      "num_tokens": 683125.0,
+      "step": 385
+    },
+    {
+      "entropy": 7.673658752441407,
+      "epoch": 0.020060180541624874,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.0001945,
+      "loss": 7.1867,
+      "mean_token_accuracy": 0.08598564639687538,
+      "num_tokens": 691561.0,
+      "step": 390
+    },
+    {
+      "entropy": 7.715641736984253,
+      "epoch": 0.020317362343440577,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00019700000000000002,
+      "loss": 7.1201,
+      "mean_token_accuracy": 0.08909595757722855,
+      "num_tokens": 701015.0,
+      "step": 395
+    },
+    {
+      "entropy": 7.7091387748718265,
+      "epoch": 0.020574544145256283,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00019950000000000002,
+      "loss": 7.1455,
+      "mean_token_accuracy": 0.0916555643081665,
+      "num_tokens": 709940.0,
+      "step": 400
+    },
+    {
+      "entropy": 7.6161031246185305,
+      "epoch": 0.020831725947071986,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.000202,
+      "loss": 7.0992,
+      "mean_token_accuracy": 0.08667092099785804,
+      "num_tokens": 719304.0,
+      "step": 405
+    },
+    {
+      "entropy": 7.635785293579102,
+      "epoch": 0.02108890774888769,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00020449999999999998,
+      "loss": 7.1057,
+      "mean_token_accuracy": 0.08669973537325859,
+      "num_tokens": 728489.0,
+      "step": 410
+    },
+    {
+      "entropy": 7.667731189727784,
+      "epoch": 0.02134608955070339,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000207,
+      "loss": 7.1789,
+      "mean_token_accuracy": 0.08465257212519646,
+      "num_tokens": 737769.0,
+      "step": 415
+    },
+    {
+      "entropy": 7.64384446144104,
+      "epoch": 0.021603271352519094,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0002095,
+      "loss": 7.0319,
+      "mean_token_accuracy": 0.08635761365294456,
+      "num_tokens": 746540.0,
+      "step": 420
+    },
+    {
+      "entropy": 7.658045482635498,
+      "epoch": 0.0218604531543348,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000212,
+      "loss": 7.216,
+      "mean_token_accuracy": 0.08415560573339462,
+      "num_tokens": 756118.0,
+      "step": 425
+    },
+    {
+      "entropy": 7.623081970214844,
+      "epoch": 0.022117634956150503,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.0002145,
+      "loss": 7.0563,
+      "mean_token_accuracy": 0.09152830839157104,
+      "num_tokens": 764091.0,
+      "step": 430
+    },
+    {
+      "entropy": 7.5571778297424315,
+      "epoch": 0.022374816757966206,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00021700000000000002,
+      "loss": 7.0872,
+      "mean_token_accuracy": 0.08672986775636674,
+      "num_tokens": 773220.0,
+      "step": 435
+    },
+    {
+      "entropy": 7.645488929748535,
+      "epoch": 0.02263199855978191,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0002195,
+      "loss": 7.1345,
+      "mean_token_accuracy": 0.08709268644452095,
+      "num_tokens": 781937.0,
+      "step": 440
+    },
+    {
+      "entropy": 7.57753210067749,
+      "epoch": 0.022889180361597615,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000222,
+      "loss": 7.0728,
+      "mean_token_accuracy": 0.08992104157805443,
+      "num_tokens": 791685.0,
+      "step": 445
+    },
+    {
+      "entropy": 7.5337488651275635,
+      "epoch": 0.023146362163413318,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0002245,
+      "loss": 6.9439,
+      "mean_token_accuracy": 0.08900289312005043,
+      "num_tokens": 800759.0,
+      "step": 450
+    },
+    {
+      "entropy": 7.569718551635742,
+      "epoch": 0.02340354396522902,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00022700000000000002,
+      "loss": 7.0617,
+      "mean_token_accuracy": 0.08903328478336334,
+      "num_tokens": 809508.0,
+      "step": 455
+    },
+    {
+      "entropy": 7.4335291385650635,
+      "epoch": 0.023660725767044723,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00022950000000000002,
+      "loss": 6.9777,
+      "mean_token_accuracy": 0.08975348770618438,
+      "num_tokens": 819262.0,
+      "step": 460
+    },
+    {
+      "entropy": 7.5286821842193605,
+      "epoch": 0.023917907568860426,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 7.0061,
+      "mean_token_accuracy": 0.09333336725831032,
+      "num_tokens": 827902.0,
+      "step": 465
+    },
+    {
+      "entropy": 7.5438226699829105,
+      "epoch": 0.024175089370676132,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00023449999999999998,
+      "loss": 7.071,
+      "mean_token_accuracy": 0.08738465905189514,
+      "num_tokens": 836678.0,
+      "step": 470
+    },
+    {
+      "entropy": 7.578305292129516,
+      "epoch": 0.024432271172491835,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000237,
+      "loss": 7.0481,
+      "mean_token_accuracy": 0.08780269175767899,
+      "num_tokens": 845604.0,
+      "step": 475
+    },
+    {
+      "entropy": 7.495909070968628,
+      "epoch": 0.024689452974307538,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0002395,
+      "loss": 6.9237,
+      "mean_token_accuracy": 0.09766526669263839,
+      "num_tokens": 854530.0,
+      "step": 480
+    },
+    {
+      "entropy": 7.472972011566162,
+      "epoch": 0.02494663477612324,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.000242,
+      "loss": 6.9801,
+      "mean_token_accuracy": 0.08929606899619102,
+      "num_tokens": 863038.0,
+      "step": 485
+    },
+    {
+      "entropy": 7.4176201820373535,
+      "epoch": 0.025203816577938944,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0002445,
+      "loss": 6.9488,
+      "mean_token_accuracy": 0.09348267316818237,
+      "num_tokens": 871691.0,
+      "step": 490
+    },
+    {
+      "entropy": 7.523787260055542,
+      "epoch": 0.02546099837975465,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000247,
+      "loss": 6.9412,
+      "mean_token_accuracy": 0.09506363123655319,
+      "num_tokens": 881026.0,
+      "step": 495
+    },
+    {
+      "entropy": 7.421153450012207,
+      "epoch": 0.025718180181570353,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.0002495,
+      "loss": 6.9932,
+      "mean_token_accuracy": 0.0955754928290844,
+      "num_tokens": 889829.0,
+      "step": 500
+    },
+    {
+      "entropy": 7.390109062194824,
+      "epoch": 0.025975361983386056,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.000252,
+      "loss": 7.0112,
+      "mean_token_accuracy": 0.0914897546172142,
+      "num_tokens": 898992.0,
+      "step": 505
+    },
+    {
+      "entropy": 7.555384731292724,
+      "epoch": 0.02623254378520176,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0002545,
+      "loss": 7.0507,
+      "mean_token_accuracy": 0.08820229396224022,
+      "num_tokens": 908237.0,
+      "step": 510
+    },
+    {
+      "entropy": 7.433199691772461,
+      "epoch": 0.02648972558701746,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000257,
+      "loss": 6.8716,
+      "mean_token_accuracy": 0.09959734976291656,
+      "num_tokens": 916715.0,
+      "step": 515
+    },
+    {
+      "entropy": 7.360012674331665,
+      "epoch": 0.026746907388833167,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0002595,
+      "loss": 6.9746,
+      "mean_token_accuracy": 0.0938393734395504,
+      "num_tokens": 925834.0,
+      "step": 520
+    },
+    {
+      "entropy": 7.439189624786377,
+      "epoch": 0.02700408919064887,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.000262,
+      "loss": 6.9216,
+      "mean_token_accuracy": 0.09481378644704819,
+      "num_tokens": 935725.0,
+      "step": 525
+    },
+    {
+      "entropy": 7.516362428665161,
+      "epoch": 0.027261270992464573,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.00026450000000000003,
+      "loss": 7.0258,
+      "mean_token_accuracy": 0.09422887191176414,
+      "num_tokens": 944788.0,
+      "step": 530
+    },
+    {
+      "entropy": 7.29243836402893,
+      "epoch": 0.027518452794280276,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00026700000000000004,
+      "loss": 6.8317,
+      "mean_token_accuracy": 0.09764747545123101,
+      "num_tokens": 953561.0,
+      "step": 535
+    },
+    {
+      "entropy": 7.504419994354248,
+      "epoch": 0.02777563459609598,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00026950000000000005,
+      "loss": 6.9259,
+      "mean_token_accuracy": 0.09662783890962601,
+      "num_tokens": 961942.0,
+      "step": 540
+    },
+    {
+      "entropy": 7.3940901279449465,
+      "epoch": 0.028032816397911685,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 6.8816,
+      "mean_token_accuracy": 0.09403559938073158,
+      "num_tokens": 970699.0,
+      "step": 545
+    },
+    {
+      "entropy": 7.347795295715332,
+      "epoch": 0.028289998199727388,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0002745,
+      "loss": 6.9946,
+      "mean_token_accuracy": 0.09083672165870667,
+      "num_tokens": 980352.0,
+      "step": 550
+    },
+    {
+      "entropy": 7.459839010238648,
+      "epoch": 0.02854718000154309,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000277,
+      "loss": 6.8925,
+      "mean_token_accuracy": 0.09576373025774956,
+      "num_tokens": 988771.0,
+      "step": 555
+    },
+    {
+      "entropy": 7.296186876296997,
+      "epoch": 0.028804361803358793,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.0002795,
+      "loss": 6.7979,
+      "mean_token_accuracy": 0.09578228294849396,
+      "num_tokens": 997564.0,
+      "step": 560
+    },
+    {
+      "entropy": 7.341630935668945,
+      "epoch": 0.0290615436051745,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 6.8745,
+      "mean_token_accuracy": 0.09284131526947022,
+      "num_tokens": 1005945.0,
+      "step": 565
+    },
+    {
+      "entropy": 7.347255086898803,
+      "epoch": 0.029318725406990202,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0002845,
+      "loss": 6.9394,
+      "mean_token_accuracy": 0.09283768609166146,
+      "num_tokens": 1015012.0,
+      "step": 570
+    },
+    {
+      "entropy": 7.4236492156982425,
+      "epoch": 0.029575907208805905,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000287,
+      "loss": 6.9536,
+      "mean_token_accuracy": 0.09391803517937661,
+      "num_tokens": 1024240.0,
+      "step": 575
+    },
+    {
+      "entropy": 7.367508506774902,
+      "epoch": 0.029833089010621608,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.0002895,
+      "loss": 6.8857,
+      "mean_token_accuracy": 0.09755372777581214,
+      "num_tokens": 1033277.0,
+      "step": 580
+    },
+    {
+      "entropy": 7.279513311386109,
+      "epoch": 0.03009027081243731,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.000292,
+      "loss": 6.8268,
+      "mean_token_accuracy": 0.09293166324496269,
+      "num_tokens": 1042312.0,
+      "step": 585
+    },
+    {
+      "entropy": 7.353345108032227,
+      "epoch": 0.030347452614253017,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0002945,
+      "loss": 6.8653,
+      "mean_token_accuracy": 0.0949218288064003,
+      "num_tokens": 1052756.0,
+      "step": 590
+    },
+    {
+      "entropy": 7.368993330001831,
+      "epoch": 0.03060463441606872,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.000297,
+      "loss": 6.8974,
+      "mean_token_accuracy": 0.10047846585512162,
+      "num_tokens": 1061563.0,
+      "step": 595
+    },
+    {
+      "entropy": 7.286983442306519,
+      "epoch": 0.030861816217884423,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.0002995,
+      "loss": 6.7907,
+      "mean_token_accuracy": 0.10389479920268059,
+      "num_tokens": 1070177.0,
+      "step": 600
+    },
+    {
+      "entropy": 7.252815437316895,
+      "epoch": 0.031118998019700125,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.000302,
+      "loss": 6.8292,
+      "mean_token_accuracy": 0.09795110002160072,
+      "num_tokens": 1079662.0,
+      "step": 605
+    },
+    {
+      "entropy": 7.261862182617188,
+      "epoch": 0.03137617982151583,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003045,
+      "loss": 6.7916,
+      "mean_token_accuracy": 0.0980616994202137,
+      "num_tokens": 1088719.0,
+      "step": 610
+    },
+    {
+      "entropy": 7.310345602035523,
+      "epoch": 0.03163336162333153,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000307,
+      "loss": 6.7727,
+      "mean_token_accuracy": 0.10127385258674622,
+      "num_tokens": 1096663.0,
+      "step": 615
+    },
+    {
+      "entropy": 7.34635648727417,
+      "epoch": 0.03189054342514724,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0003095,
+      "loss": 6.8751,
+      "mean_token_accuracy": 0.0975107342004776,
+      "num_tokens": 1105484.0,
+      "step": 620
+    },
+    {
+      "entropy": 7.2787559032440186,
+      "epoch": 0.032147725226962943,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.000312,
+      "loss": 6.8409,
+      "mean_token_accuracy": 0.096168252825737,
+      "num_tokens": 1114781.0,
+      "step": 625
+    },
+    {
+      "entropy": 7.194593715667724,
+      "epoch": 0.03240490702877864,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003145,
+      "loss": 6.8495,
+      "mean_token_accuracy": 0.1023260697722435,
+      "num_tokens": 1124075.0,
+      "step": 630
+    },
+    {
+      "entropy": 7.284073257446289,
+      "epoch": 0.03266208883059435,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.000317,
+      "loss": 6.7928,
+      "mean_token_accuracy": 0.09392334222793579,
+      "num_tokens": 1132863.0,
+      "step": 635
+    },
+    {
+      "entropy": 7.23821120262146,
+      "epoch": 0.03291927063241005,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0003195,
+      "loss": 6.8477,
+      "mean_token_accuracy": 0.10305669158697128,
+      "num_tokens": 1141324.0,
+      "step": 640
+    },
+    {
+      "entropy": 7.2563272476196286,
+      "epoch": 0.033176452434225755,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.000322,
+      "loss": 6.716,
+      "mean_token_accuracy": 0.10566487908363342,
+      "num_tokens": 1149406.0,
+      "step": 645
+    },
+    {
+      "entropy": 7.335630702972412,
+      "epoch": 0.03343363423604146,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00032450000000000003,
+      "loss": 6.8656,
+      "mean_token_accuracy": 0.09870733916759492,
+      "num_tokens": 1158488.0,
+      "step": 650
+    },
+    {
+      "entropy": 7.255077314376831,
+      "epoch": 0.03369081603785716,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.00032700000000000003,
+      "loss": 6.8075,
+      "mean_token_accuracy": 0.10482278540730476,
+      "num_tokens": 1166756.0,
+      "step": 655
+    },
+    {
+      "entropy": 7.2207207679748535,
+      "epoch": 0.03394799783967287,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00032950000000000004,
+      "loss": 6.7814,
+      "mean_token_accuracy": 0.10137899518013001,
+      "num_tokens": 1175930.0,
+      "step": 660
+    },
+    {
+      "entropy": 7.146165227890014,
+      "epoch": 0.034205179641488566,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 6.7864,
+      "mean_token_accuracy": 0.1040356308221817,
+      "num_tokens": 1184348.0,
+      "step": 665
+    },
+    {
+      "entropy": 7.224062728881836,
+      "epoch": 0.03446236144330427,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00033450000000000005,
+      "loss": 6.7721,
+      "mean_token_accuracy": 0.10236084014177323,
+      "num_tokens": 1192800.0,
+      "step": 670
+    },
+    {
+      "entropy": 7.279936122894287,
+      "epoch": 0.03471954324511998,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000337,
+      "loss": 6.8795,
+      "mean_token_accuracy": 0.09708840399980545,
+      "num_tokens": 1201446.0,
+      "step": 675
+    },
+    {
+      "entropy": 7.065876436233521,
+      "epoch": 0.03497672504693568,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.0003395,
+      "loss": 6.6507,
+      "mean_token_accuracy": 0.1062047965824604,
+      "num_tokens": 1210088.0,
+      "step": 680
+    },
+    {
+      "entropy": 7.077348375320435,
+      "epoch": 0.035233906848751384,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000342,
+      "loss": 6.5917,
+      "mean_token_accuracy": 0.11215948089957237,
+      "num_tokens": 1219425.0,
+      "step": 685
+    },
+    {
+      "entropy": 7.214649200439453,
+      "epoch": 0.03549108865056708,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00034449999999999997,
+      "loss": 6.7802,
+      "mean_token_accuracy": 0.09808368533849716,
+      "num_tokens": 1228332.0,
+      "step": 690
+    },
+    {
+      "entropy": 7.219751453399658,
+      "epoch": 0.03574827045238279,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.000347,
+      "loss": 6.8329,
+      "mean_token_accuracy": 0.09870714694261551,
+      "num_tokens": 1237953.0,
+      "step": 695
+    },
+    {
+      "entropy": 7.24316897392273,
+      "epoch": 0.036005452254198496,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.0003495,
+      "loss": 6.8479,
+      "mean_token_accuracy": 0.09783646762371064,
+      "num_tokens": 1247387.0,
+      "step": 700
+    },
+    {
+      "entropy": 7.2066491603851315,
+      "epoch": 0.036262634056014195,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.000352,
+      "loss": 6.7471,
+      "mean_token_accuracy": 0.10154265984892845,
+      "num_tokens": 1256215.0,
+      "step": 705
+    },
+    {
+      "entropy": 7.0408368587493895,
+      "epoch": 0.0365198158578299,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.0003545,
+      "loss": 6.6315,
+      "mean_token_accuracy": 0.10381938368082047,
+      "num_tokens": 1265963.0,
+      "step": 710
+    },
+    {
+      "entropy": 7.164453172683716,
+      "epoch": 0.0367769976596456,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.000357,
+      "loss": 6.6903,
+      "mean_token_accuracy": 0.1119053155183792,
+      "num_tokens": 1275024.0,
+      "step": 715
+    },
+    {
+      "entropy": 7.111461877822876,
+      "epoch": 0.03703417946146131,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0003595,
+      "loss": 6.7586,
+      "mean_token_accuracy": 0.10169535055756569,
+      "num_tokens": 1283981.0,
+      "step": 720
+    },
+    {
+      "entropy": 7.1354100704193115,
+      "epoch": 0.03729136126327701,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000362,
+      "loss": 6.62,
+      "mean_token_accuracy": 0.10666845366358757,
+      "num_tokens": 1292834.0,
+      "step": 725
+    },
+    {
+      "entropy": 7.122345590591431,
+      "epoch": 0.03754854306509271,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0003645,
+      "loss": 6.7978,
+      "mean_token_accuracy": 0.1032055139541626,
+      "num_tokens": 1301429.0,
+      "step": 730
+    },
+    {
+      "entropy": 7.0510656356811525,
+      "epoch": 0.03780572486690842,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000367,
+      "loss": 6.7053,
+      "mean_token_accuracy": 0.10353485718369485,
+      "num_tokens": 1310088.0,
+      "step": 735
+    },
+    {
+      "entropy": 7.182754850387573,
+      "epoch": 0.03806290666872412,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0003695,
+      "loss": 6.7723,
+      "mean_token_accuracy": 0.10199789181351662,
+      "num_tokens": 1318249.0,
+      "step": 740
+    },
+    {
+      "entropy": 7.0828611850738525,
+      "epoch": 0.038320088470539825,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.000372,
+      "loss": 6.6795,
+      "mean_token_accuracy": 0.10390995517373085,
+      "num_tokens": 1327075.0,
+      "step": 745
+    },
+    {
+      "entropy": 7.114889192581177,
+      "epoch": 0.03857727027235553,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0003745,
+      "loss": 6.7128,
+      "mean_token_accuracy": 0.1041706345975399,
+      "num_tokens": 1335995.0,
+      "step": 750
+    },
+    {
+      "entropy": 7.040337038040161,
+      "epoch": 0.03883445207417123,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.000377,
+      "loss": 6.6914,
+      "mean_token_accuracy": 0.10317371264100075,
+      "num_tokens": 1344789.0,
+      "step": 755
+    },
+    {
+      "entropy": 7.120604038238525,
+      "epoch": 0.039091633875986936,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0003795,
+      "loss": 6.6955,
+      "mean_token_accuracy": 0.10629725456237793,
+      "num_tokens": 1352940.0,
+      "step": 760
+    },
+    {
+      "entropy": 7.052408075332641,
+      "epoch": 0.039348815677802636,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.000382,
+      "loss": 6.637,
+      "mean_token_accuracy": 0.10945691764354706,
+      "num_tokens": 1361803.0,
+      "step": 765
+    },
+    {
+      "entropy": 7.069132375717163,
+      "epoch": 0.03960599747961834,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0003845,
+      "loss": 6.7294,
+      "mean_token_accuracy": 0.10019565001130104,
+      "num_tokens": 1371647.0,
+      "step": 770
+    },
+    {
+      "entropy": 7.164604377746582,
+      "epoch": 0.03986317928143405,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00038700000000000003,
+      "loss": 6.7523,
+      "mean_token_accuracy": 0.10618292987346649,
+      "num_tokens": 1380447.0,
+      "step": 775
+    },
+    {
+      "entropy": 7.114570999145508,
+      "epoch": 0.04012036108324975,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00038950000000000003,
+      "loss": 6.7284,
+      "mean_token_accuracy": 0.09737010598182679,
+      "num_tokens": 1388902.0,
+      "step": 780
+    },
+    {
+      "entropy": 7.0223623752594,
+      "epoch": 0.040377542885065454,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 6.5808,
+      "mean_token_accuracy": 0.1079009048640728,
+      "num_tokens": 1398178.0,
+      "step": 785
+    },
+    {
+      "entropy": 7.056776905059815,
+      "epoch": 0.04063472468688115,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00039450000000000005,
+      "loss": 6.6511,
+      "mean_token_accuracy": 0.11126778572797776,
+      "num_tokens": 1406042.0,
+      "step": 790
+    },
+    {
+      "entropy": 7.084379482269287,
+      "epoch": 0.04089190648869686,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00039700000000000005,
+      "loss": 6.6815,
+      "mean_token_accuracy": 0.10524625033140182,
+      "num_tokens": 1415562.0,
+      "step": 795
+    },
+    {
+      "entropy": 7.051506662368775,
+      "epoch": 0.041149088290512566,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0003995,
+      "loss": 6.6008,
+      "mean_token_accuracy": 0.10510810911655426,
+      "num_tokens": 1424782.0,
+      "step": 800
+    },
+    {
+      "entropy": 6.966880178451538,
+      "epoch": 0.041406270092328265,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000402,
+      "loss": 6.6177,
+      "mean_token_accuracy": 0.10489397794008255,
+      "num_tokens": 1433737.0,
+      "step": 805
+    },
+    {
+      "entropy": 7.0082066535949705,
+      "epoch": 0.04166345189414397,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004045,
+      "loss": 6.6886,
+      "mean_token_accuracy": 0.10373986288905143,
+      "num_tokens": 1442948.0,
+      "step": 810
+    },
+    {
+      "entropy": 7.030193662643432,
+      "epoch": 0.04192063369595967,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00040699999999999997,
+      "loss": 6.617,
+      "mean_token_accuracy": 0.1067310206592083,
+      "num_tokens": 1452333.0,
+      "step": 815
+    },
+    {
+      "entropy": 6.962583208084107,
+      "epoch": 0.04217781549777538,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004095,
+      "loss": 6.6201,
+      "mean_token_accuracy": 0.10676174983382225,
+      "num_tokens": 1461160.0,
+      "step": 820
+    },
+    {
+      "entropy": 6.962631416320801,
+      "epoch": 0.04243499729959108,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000412,
+      "loss": 6.6449,
+      "mean_token_accuracy": 0.10844494923949241,
+      "num_tokens": 1470248.0,
+      "step": 825
+    },
+    {
+      "entropy": 7.0632740497589115,
+      "epoch": 0.04269217910140678,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004145,
+      "loss": 6.6382,
+      "mean_token_accuracy": 0.11155002191662788,
+      "num_tokens": 1478992.0,
+      "step": 830
+    },
+    {
+      "entropy": 7.077846574783325,
+      "epoch": 0.04294936090322249,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.000417,
+      "loss": 6.6672,
+      "mean_token_accuracy": 0.1010983519256115,
+      "num_tokens": 1488341.0,
+      "step": 835
+    },
+    {
+      "entropy": 6.8585468292236325,
+      "epoch": 0.04320654270503819,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0004195,
+      "loss": 6.5807,
+      "mean_token_accuracy": 0.10849210470914841,
+      "num_tokens": 1496986.0,
+      "step": 840
+    },
+    {
+      "entropy": 7.0004956245422365,
+      "epoch": 0.043463724506853894,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000422,
+      "loss": 6.5236,
+      "mean_token_accuracy": 0.11627273112535477,
+      "num_tokens": 1505455.0,
+      "step": 845
+    },
+    {
+      "entropy": 7.062360429763794,
+      "epoch": 0.0437209063086696,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004245,
+      "loss": 6.6677,
+      "mean_token_accuracy": 0.11094246879220009,
+      "num_tokens": 1515240.0,
+      "step": 850
+    },
+    {
+      "entropy": 6.843272590637207,
+      "epoch": 0.0439780881104853,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000427,
+      "loss": 6.4627,
+      "mean_token_accuracy": 0.11083339005708695,
+      "num_tokens": 1523234.0,
+      "step": 855
+    },
+    {
+      "entropy": 6.870317792892456,
+      "epoch": 0.044235269912301006,
+      "grad_norm": 1.25,
+      "learning_rate": 0.0004295,
+      "loss": 6.5185,
+      "mean_token_accuracy": 0.10689858049154281,
+      "num_tokens": 1531836.0,
+      "step": 860
+    },
+    {
+      "entropy": 6.954329633712769,
+      "epoch": 0.04449245171411671,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000432,
+      "loss": 6.5619,
+      "mean_token_accuracy": 0.1082220919430256,
+      "num_tokens": 1540611.0,
+      "step": 865
+    },
+    {
+      "entropy": 6.9597697257995605,
+      "epoch": 0.04474963351593241,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0004345,
+      "loss": 6.5767,
+      "mean_token_accuracy": 0.10847441479563713,
+      "num_tokens": 1549652.0,
+      "step": 870
+    },
+    {
+      "entropy": 6.933281469345093,
+      "epoch": 0.04500681531774812,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.000437,
+      "loss": 6.654,
+      "mean_token_accuracy": 0.10323531776666642,
+      "num_tokens": 1559040.0,
+      "step": 875
+    },
+    {
+      "entropy": 6.976513767242432,
+      "epoch": 0.04526399711956382,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004395,
+      "loss": 6.5506,
+      "mean_token_accuracy": 0.10878390446305275,
+      "num_tokens": 1567745.0,
+      "step": 880
+    },
+    {
+      "entropy": 6.939139747619629,
+      "epoch": 0.045521178921379524,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.000442,
+      "loss": 6.5311,
+      "mean_token_accuracy": 0.11226821094751357,
+      "num_tokens": 1576077.0,
+      "step": 885
+    },
+    {
+      "entropy": 6.885157585144043,
+      "epoch": 0.04577836072319523,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.0004445,
+      "loss": 6.597,
+      "mean_token_accuracy": 0.10303749218583107,
+      "num_tokens": 1584641.0,
+      "step": 890
+    },
+    {
+      "entropy": 6.9989728927612305,
+      "epoch": 0.04603554252501093,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000447,
+      "loss": 6.5356,
+      "mean_token_accuracy": 0.10896688029170036,
+      "num_tokens": 1594131.0,
+      "step": 895
+    },
+    {
+      "entropy": 6.9140125751495365,
+      "epoch": 0.046292724326826636,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00044950000000000003,
+      "loss": 6.5217,
+      "mean_token_accuracy": 0.10893905013799668,
+      "num_tokens": 1603099.0,
+      "step": 900
+    },
+    {
+      "entropy": 6.941550970077515,
+      "epoch": 0.046549906128642335,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 6.5503,
+      "mean_token_accuracy": 0.11116070300340652,
+      "num_tokens": 1611453.0,
+      "step": 905
+    },
+    {
+      "entropy": 6.77474455833435,
+      "epoch": 0.04680708793045804,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00045450000000000004,
+      "loss": 6.5397,
+      "mean_token_accuracy": 0.10877310186624527,
+      "num_tokens": 1619577.0,
+      "step": 910
+    },
+    {
+      "entropy": 7.0068763256072994,
+      "epoch": 0.04706426973227375,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00045700000000000005,
+      "loss": 6.6618,
+      "mean_token_accuracy": 0.10706020519137383,
+      "num_tokens": 1629026.0,
+      "step": 915
+    },
+    {
+      "entropy": 6.97544412612915,
+      "epoch": 0.04732145153408945,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00045950000000000006,
+      "loss": 6.5855,
+      "mean_token_accuracy": 0.11112762615084648,
+      "num_tokens": 1638556.0,
+      "step": 920
+    },
+    {
+      "entropy": 6.83156042098999,
+      "epoch": 0.04757863333590515,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.000462,
+      "loss": 6.4674,
+      "mean_token_accuracy": 0.11065400391817093,
+      "num_tokens": 1647756.0,
+      "step": 925
+    },
+    {
+      "entropy": 6.925829887390137,
+      "epoch": 0.04783581513772085,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004645,
+      "loss": 6.5459,
+      "mean_token_accuracy": 0.10694767236709594,
+      "num_tokens": 1656614.0,
+      "step": 930
+    },
+    {
+      "entropy": 6.851103353500366,
+      "epoch": 0.04809299693953656,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.000467,
+      "loss": 6.5566,
+      "mean_token_accuracy": 0.11177869439125061,
+      "num_tokens": 1666213.0,
+      "step": 935
+    },
+    {
+      "entropy": 6.823415803909302,
+      "epoch": 0.048350178741352265,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004695,
+      "loss": 6.5108,
+      "mean_token_accuracy": 0.11019086241722106,
+      "num_tokens": 1675435.0,
+      "step": 940
+    },
+    {
+      "entropy": 6.8377485275268555,
+      "epoch": 0.048607360543167964,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000472,
+      "loss": 6.4806,
+      "mean_token_accuracy": 0.11481974124908448,
+      "num_tokens": 1684268.0,
+      "step": 945
+    },
+    {
+      "entropy": 6.837112474441528,
+      "epoch": 0.04886454234498367,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004745,
+      "loss": 6.566,
+      "mean_token_accuracy": 0.10879503637552261,
+      "num_tokens": 1692665.0,
+      "step": 950
+    },
+    {
+      "entropy": 7.00029149055481,
+      "epoch": 0.04912172414679937,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000477,
+      "loss": 6.6023,
+      "mean_token_accuracy": 0.11165886968374253,
+      "num_tokens": 1701321.0,
+      "step": 955
+    },
+    {
+      "entropy": 6.796492385864258,
+      "epoch": 0.049378905948615076,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004795,
+      "loss": 6.5413,
+      "mean_token_accuracy": 0.10716139823198319,
+      "num_tokens": 1710476.0,
+      "step": 960
+    },
+    {
+      "entropy": 6.805513620376587,
+      "epoch": 0.04963608775043078,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000482,
+      "loss": 6.4891,
+      "mean_token_accuracy": 0.10693883448839188,
+      "num_tokens": 1719949.0,
+      "step": 965
+    },
+    {
+      "entropy": 6.972143602371216,
+      "epoch": 0.04989326955224648,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004845,
+      "loss": 6.6165,
+      "mean_token_accuracy": 0.10792995691299438,
+      "num_tokens": 1729362.0,
+      "step": 970
+    },
+    {
+      "entropy": 6.834133863449097,
+      "epoch": 0.05015045135406219,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.000487,
+      "loss": 6.5197,
+      "mean_token_accuracy": 0.10525891557335854,
+      "num_tokens": 1739001.0,
+      "step": 975
+    },
+    {
+      "entropy": 6.7445004940032955,
+      "epoch": 0.05040763315587789,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004895,
+      "loss": 6.4256,
+      "mean_token_accuracy": 0.11702816039323807,
+      "num_tokens": 1748249.0,
+      "step": 980
+    },
+    {
+      "entropy": 6.755212068557739,
+      "epoch": 0.050664814957693594,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.000492,
+      "loss": 6.4386,
+      "mean_token_accuracy": 0.1141409620642662,
+      "num_tokens": 1757070.0,
+      "step": 985
+    },
+    {
+      "entropy": 6.799167490005493,
+      "epoch": 0.0509219967595093,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004945,
+      "loss": 6.4032,
+      "mean_token_accuracy": 0.11792294010519981,
+      "num_tokens": 1765737.0,
+      "step": 990
+    },
+    {
+      "entropy": 6.891513156890869,
+      "epoch": 0.051179178561325,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000497,
+      "loss": 6.5878,
+      "mean_token_accuracy": 0.11050201207399368,
+      "num_tokens": 1774942.0,
+      "step": 995
+    },
+    {
+      "entropy": 6.857363891601563,
+      "epoch": 0.051436360363140705,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004995,
+      "loss": 6.5936,
+      "mean_token_accuracy": 0.10368210449814796,
+      "num_tokens": 1784256.0,
+      "step": 1000
+    },
+    {
+      "entropy": 6.849530935287476,
+      "epoch": 0.051693542164956405,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.000499998026082006,
+      "loss": 6.5383,
+      "mean_token_accuracy": 0.10572032853960991,
+      "num_tokens": 1792926.0,
+      "step": 1005
+    },
+    {
+      "entropy": 6.72293267250061,
+      "epoch": 0.05195072396677211,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004999900070995136,
+      "loss": 6.4893,
+      "mean_token_accuracy": 0.11575818508863449,
+      "num_tokens": 1802070.0,
+      "step": 1010
+    },
+    {
+      "entropy": 6.841842365264893,
+      "epoch": 0.05220790576858782,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004999758199023239,
+      "loss": 6.5344,
+      "mean_token_accuracy": 0.10665113478899002,
+      "num_tokens": 1811507.0,
+      "step": 1015
+    },
+    {
+      "entropy": 6.851677942276001,
+      "epoch": 0.05246508757040352,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.0004999554648793858,
+      "loss": 6.4522,
+      "mean_token_accuracy": 0.11116268336772919,
+      "num_tokens": 1820311.0,
+      "step": 1020
+    },
+    {
+      "entropy": 6.710117959976197,
+      "epoch": 0.05272226937221922,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0004999289425887425,
+      "loss": 6.4099,
+      "mean_token_accuracy": 0.1138461858034134,
+      "num_tokens": 1828948.0,
+      "step": 1025
+    },
+    {
+      "entropy": 6.834085750579834,
+      "epoch": 0.05297945117403492,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004998962537575161,
+      "loss": 6.5724,
+      "mean_token_accuracy": 0.1098745584487915,
+      "num_tokens": 1838350.0,
+      "step": 1030
+    },
+    {
+      "entropy": 6.842246675491333,
+      "epoch": 0.05323663297585063,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004998573992818874,
+      "loss": 6.4004,
+      "mean_token_accuracy": 0.12005882859230041,
+      "num_tokens": 1847162.0,
+      "step": 1035
+    },
+    {
+      "entropy": 6.665293884277344,
+      "epoch": 0.053493814777666335,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004998123802270715,
+      "loss": 6.4849,
+      "mean_token_accuracy": 0.11263752803206444,
+      "num_tokens": 1856275.0,
+      "step": 1040
+    },
+    {
+      "entropy": 6.729813432693481,
+      "epoch": 0.053750996579482034,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004997611978272886,
+      "loss": 6.3855,
+      "mean_token_accuracy": 0.11651406660676003,
+      "num_tokens": 1864663.0,
+      "step": 1045
+    },
+    {
+      "entropy": 6.733455944061279,
+      "epoch": 0.05400817838129774,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004997038534857298,
+      "loss": 6.424,
+      "mean_token_accuracy": 0.1177938535809517,
+      "num_tokens": 1874622.0,
+      "step": 1050
+    },
+    {
+      "entropy": 6.7023192882537845,
+      "epoch": 0.05426536018311344,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004996403487745194,
+      "loss": 6.3922,
+      "mean_token_accuracy": 0.11079680770635605,
+      "num_tokens": 1884257.0,
+      "step": 1055
+    },
+    {
+      "entropy": 6.801939916610718,
+      "epoch": 0.054522541984929146,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.000499570685434671,
+      "loss": 6.4462,
+      "mean_token_accuracy": 0.11080927178263664,
+      "num_tokens": 1894336.0,
+      "step": 1060
+    },
+    {
+      "entropy": 6.734097909927368,
+      "epoch": 0.05477972378674485,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0004994948653760405,
+      "loss": 6.4129,
+      "mean_token_accuracy": 0.11459591910243035,
+      "num_tokens": 1903305.0,
+      "step": 1065
+    },
+    {
+      "entropy": 6.686365461349487,
+      "epoch": 0.05503690558856055,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0004994128906772729,
+      "loss": 6.4414,
+      "mean_token_accuracy": 0.11344051584601403,
+      "num_tokens": 1911992.0,
+      "step": 1070
+    },
+    {
+      "entropy": 6.707837152481079,
+      "epoch": 0.05529408739037626,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000499324763585746,
+      "loss": 6.3462,
+      "mean_token_accuracy": 0.12238444536924362,
+      "num_tokens": 1920826.0,
+      "step": 1075
+    },
+    {
+      "entropy": 6.851500606536865,
+      "epoch": 0.05555126919219196,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004992304865175085,
+      "loss": 6.5005,
+      "mean_token_accuracy": 0.11342206448316575,
+      "num_tokens": 1930181.0,
+      "step": 1080
+    },
+    {
+      "entropy": 6.683105993270874,
+      "epoch": 0.05580845099400766,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0004991300620572138,
+      "loss": 6.3627,
+      "mean_token_accuracy": 0.11612597554922104,
+      "num_tokens": 1938865.0,
+      "step": 1085
+    },
+    {
+      "entropy": 6.6627490520477295,
+      "epoch": 0.05606563279582337,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004990234929580494,
+      "loss": 6.3018,
+      "mean_token_accuracy": 0.11800351217389107,
+      "num_tokens": 1947979.0,
+      "step": 1090
+    },
+    {
+      "entropy": 6.705393505096436,
+      "epoch": 0.05632281459763907,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0004989107821416609,
+      "loss": 6.4207,
+      "mean_token_accuracy": 0.11706310212612152,
+      "num_tokens": 1957012.0,
+      "step": 1095
+    },
+    {
+      "entropy": 6.785688734054565,
+      "epoch": 0.056579996399454775,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004987919326980723,
+      "loss": 6.5063,
+      "mean_token_accuracy": 0.11153422519564629,
+      "num_tokens": 1966613.0,
+      "step": 1100
+    },
+    {
+      "entropy": 6.662994146347046,
+      "epoch": 0.05683717820127048,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004986669478856011,
+      "loss": 6.4567,
+      "mean_token_accuracy": 0.11049317792057992,
+      "num_tokens": 1975878.0,
+      "step": 1105
+    },
+    {
+      "entropy": 6.703376007080078,
+      "epoch": 0.05709436000308618,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004985358311307688,
+      "loss": 6.3858,
+      "mean_token_accuracy": 0.12022560909390449,
+      "num_tokens": 1984828.0,
+      "step": 1110
+    },
+    {
+      "entropy": 6.687681436538696,
+      "epoch": 0.05735154180490189,
+      "grad_norm": 1.25,
+      "learning_rate": 0.0004983985860282081,
+      "loss": 6.3282,
+      "mean_token_accuracy": 0.11830979585647583,
+      "num_tokens": 1992706.0,
+      "step": 1115
+    },
+    {
+      "entropy": 6.691852474212647,
+      "epoch": 0.057608723606717586,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004982552163405623,
+      "loss": 6.3836,
+      "mean_token_accuracy": 0.11681668907403946,
+      "num_tokens": 2001461.0,
+      "step": 1120
+    },
+    {
+      "entropy": 6.5880796909332275,
+      "epoch": 0.05786590540853329,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004981057259983839,
+      "loss": 6.3809,
+      "mean_token_accuracy": 0.11749306172132493,
+      "num_tokens": 2011029.0,
+      "step": 1125
+    },
+    {
+      "entropy": 6.7658202171325685,
+      "epoch": 0.058123087210349,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004979501191000262,
+      "loss": 6.4011,
+      "mean_token_accuracy": 0.11484191119670868,
+      "num_tokens": 2019712.0,
+      "step": 1130
+    },
+    {
+      "entropy": 6.552847003936767,
+      "epoch": 0.0583802690121647,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004977883999115311,
+      "loss": 6.2373,
+      "mean_token_accuracy": 0.12176965028047562,
+      "num_tokens": 2028411.0,
+      "step": 1135
+    },
+    {
+      "entropy": 6.681560611724853,
+      "epoch": 0.058637450813980405,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004976205728665113,
+      "loss": 6.395,
+      "mean_token_accuracy": 0.11524997726082802,
+      "num_tokens": 2037874.0,
+      "step": 1140
+    },
+    {
+      "entropy": 6.68039722442627,
+      "epoch": 0.058894632615796104,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0004974466425660307,
+      "loss": 6.3971,
+      "mean_token_accuracy": 0.11318053975701332,
+      "num_tokens": 2046752.0,
+      "step": 1145
+    },
+    {
+      "entropy": 6.605433797836303,
+      "epoch": 0.05915181441761181,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004972666137784759,
+      "loss": 6.3035,
+      "mean_token_accuracy": 0.12526411563158035,
+      "num_tokens": 2054893.0,
+      "step": 1150
+    },
+    {
+      "entropy": 6.69100604057312,
+      "epoch": 0.059408996219427516,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004970804914394271,
+      "loss": 6.2801,
+      "mean_token_accuracy": 0.12134814634919167,
+      "num_tokens": 2062794.0,
+      "step": 1155
+    },
+    {
+      "entropy": 6.600447416305542,
+      "epoch": 0.059666178021243216,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0004968882806515225,
+      "loss": 6.4006,
+      "mean_token_accuracy": 0.11898823902010917,
+      "num_tokens": 2072225.0,
+      "step": 1160
+    },
+    {
+      "entropy": 6.584507894515991,
+      "epoch": 0.05992335982305892,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.0004966899866843177,
+      "loss": 6.3768,
+      "mean_token_accuracy": 0.11224907487630845,
+      "num_tokens": 2080521.0,
+      "step": 1165
+    },
+    {
+      "entropy": 6.6145092964172365,
+      "epoch": 0.06018054162487462,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000496485614974142,
+      "loss": 6.2801,
+      "mean_token_accuracy": 0.1179177112877369,
+      "num_tokens": 2090633.0,
+      "step": 1170
+    },
+    {
+      "entropy": 6.679648780822754,
+      "epoch": 0.06043772342669033,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0004962751711239492,
+      "loss": 6.4617,
+      "mean_token_accuracy": 0.1087147168815136,
+      "num_tokens": 2100284.0,
+      "step": 1175
+    },
+    {
+      "entropy": 6.599804067611695,
+      "epoch": 0.060694905228506034,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004960586609031636,
+      "loss": 6.2978,
+      "mean_token_accuracy": 0.11903873831033707,
+      "num_tokens": 2109612.0,
+      "step": 1180
+    },
+    {
+      "entropy": 6.658626890182495,
+      "epoch": 0.06095208703032173,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004958360902475224,
+      "loss": 6.3076,
+      "mean_token_accuracy": 0.1250632882118225,
+      "num_tokens": 2117817.0,
+      "step": 1185
+    },
+    {
+      "entropy": 6.6112429141998295,
+      "epoch": 0.06120926883213744,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004956074652589125,
+      "loss": 6.4007,
+      "mean_token_accuracy": 0.1222650684416294,
+      "num_tokens": 2126283.0,
+      "step": 1190
+    },
+    {
+      "entropy": 6.64881386756897,
+      "epoch": 0.06146645063395314,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004953727922052035,
+      "loss": 6.3017,
+      "mean_token_accuracy": 0.11719283163547516,
+      "num_tokens": 2134361.0,
+      "step": 1195
+    },
+    {
+      "entropy": 6.705007314682007,
+      "epoch": 0.061723632435768845,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004951320775200756,
+      "loss": 6.389,
+      "mean_token_accuracy": 0.11199460625648498,
+      "num_tokens": 2143304.0,
+      "step": 1200
+    },
+    {
+      "entropy": 6.519802904129028,
+      "epoch": 0.06198081423758455,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0004948853278028436,
+      "loss": 6.2467,
+      "mean_token_accuracy": 0.11709044873714447,
+      "num_tokens": 2151650.0,
+      "step": 1205
+    },
+    {
+      "entropy": 6.573320007324218,
+      "epoch": 0.06223799603940025,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004946325498182755,
+      "loss": 6.3113,
+      "mean_token_accuracy": 0.11818666979670525,
+      "num_tokens": 2161139.0,
+      "step": 1210
+    },
+    {
+      "entropy": 6.78554859161377,
+      "epoch": 0.06249517784121596,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004943737504964076,
+      "loss": 6.4733,
+      "mean_token_accuracy": 0.11637833341956139,
+      "num_tokens": 2170606.0,
+      "step": 1215
+    },
+    {
+      "entropy": 6.549322175979614,
+      "epoch": 0.06275235964303166,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000494108936932354,
+      "loss": 6.3382,
+      "mean_token_accuracy": 0.11747744753956794,
+      "num_tokens": 2180284.0,
+      "step": 1220
+    },
+    {
+      "entropy": 6.6795018196105955,
+      "epoch": 0.06300954144484737,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004938381163861124,
+      "loss": 6.3362,
+      "mean_token_accuracy": 0.12103840410709381,
+      "num_tokens": 2189130.0,
+      "step": 1225
+    },
+    {
+      "entropy": 6.5853265762329105,
+      "epoch": 0.06326672324666306,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004935612962823645,
+      "loss": 6.3564,
+      "mean_token_accuracy": 0.11555552780628205,
+      "num_tokens": 2198083.0,
+      "step": 1230
+    },
+    {
+      "entropy": 6.717031812667846,
+      "epoch": 0.06352390504847877,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004932784842102739,
+      "loss": 6.3463,
+      "mean_token_accuracy": 0.12118180245161056,
+      "num_tokens": 2206470.0,
+      "step": 1235
+    },
+    {
+      "entropy": 6.5396645069122314,
+      "epoch": 0.06378108685029447,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004929896879232758,
+      "loss": 6.3004,
+      "mean_token_accuracy": 0.12020415365695954,
+      "num_tokens": 2215241.0,
+      "step": 1240
+    },
+    {
+      "entropy": 6.611561393737793,
+      "epoch": 0.06403826865211018,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004926949153388668,
+      "loss": 6.3158,
+      "mean_token_accuracy": 0.12042762041091919,
+      "num_tokens": 2224047.0,
+      "step": 1245
+    },
+    {
+      "entropy": 6.717500877380371,
+      "epoch": 0.06429545045392589,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004923941745383859,
+      "loss": 6.4761,
+      "mean_token_accuracy": 0.11400299072265625,
+      "num_tokens": 2234030.0,
+      "step": 1250
+    },
+    {
+      "entropy": 6.551206016540528,
+      "epoch": 0.06455263225574158,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.000492087473766794,
+      "loss": 6.2506,
+      "mean_token_accuracy": 0.12165699452161789,
+      "num_tokens": 2243123.0,
+      "step": 1255
+    },
+    {
+      "entropy": 6.559505128860474,
+      "epoch": 0.06480981405755729,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000491774821432448,
+      "loss": 6.2329,
+      "mean_token_accuracy": 0.12427677810192109,
+      "num_tokens": 2251778.0,
+      "step": 1260
+    },
+    {
+      "entropy": 6.605167245864868,
+      "epoch": 0.06506699585937299,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004914562261068693,
+      "loss": 6.2228,
+      "mean_token_accuracy": 0.12148385569453239,
+      "num_tokens": 2261552.0,
+      "step": 1265
+    },
+    {
+      "entropy": 6.4509941101074215,
+      "epoch": 0.0653241776611887,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004911316965245098,
+      "loss": 6.2464,
+      "mean_token_accuracy": 0.12392940372228622,
+      "num_tokens": 2270543.0,
+      "step": 1270
+    },
+    {
+      "entropy": 6.684006547927856,
+      "epoch": 0.0655813594630044,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.000490801241582512,
+      "loss": 6.2811,
+      "mean_token_accuracy": 0.12095867022871971,
+      "num_tokens": 2278856.0,
+      "step": 1275
+    },
+    {
+      "entropy": 6.489159202575683,
+      "epoch": 0.0658385412648201,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.000490464870340465,
+      "loss": 6.2578,
+      "mean_token_accuracy": 0.12322199791669845,
+      "num_tokens": 2287208.0,
+      "step": 1280
+    },
+    {
+      "entropy": 6.573577213287353,
+      "epoch": 0.0660957230666358,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004901225920201563,
+      "loss": 6.3295,
+      "mean_token_accuracy": 0.11680645272135734,
+      "num_tokens": 2296947.0,
+      "step": 1285
+    },
+    {
+      "entropy": 6.5740087032318115,
+      "epoch": 0.06635290486845151,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000489774416005319,
+      "loss": 6.2783,
+      "mean_token_accuracy": 0.12419264763593674,
+      "num_tokens": 2305853.0,
+      "step": 1290
+    },
+    {
+      "entropy": 6.506151580810547,
+      "epoch": 0.06661008667026722,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004894203518413742,
+      "loss": 6.2311,
+      "mean_token_accuracy": 0.11923173069953918,
+      "num_tokens": 2315344.0,
+      "step": 1295
+    },
+    {
+      "entropy": 6.499106740951538,
+      "epoch": 0.06686726847208292,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004890604092351701,
+      "loss": 6.2209,
+      "mean_token_accuracy": 0.1210881769657135,
+      "num_tokens": 2324007.0,
+      "step": 1300
+    },
+    {
+      "entropy": 6.517620944976807,
+      "epoch": 0.06712445027389861,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000488694598054715,
+      "loss": 6.2672,
+      "mean_token_accuracy": 0.11856792867183685,
+      "num_tokens": 2332613.0,
+      "step": 1305
+    },
+    {
+      "entropy": 6.554975318908691,
+      "epoch": 0.06738163207571432,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004883229283289071,
+      "loss": 6.2676,
+      "mean_token_accuracy": 0.12289687469601632,
+      "num_tokens": 2342019.0,
+      "step": 1310
+    },
+    {
+      "entropy": 6.47722110748291,
+      "epoch": 0.06763881387753003,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00048794541024725993,
+      "loss": 6.2182,
+      "mean_token_accuracy": 0.12205516248941421,
+      "num_tokens": 2350715.0,
+      "step": 1315
+    },
+    {
+      "entropy": 6.656716966629029,
+      "epoch": 0.06789599567934573,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004875620541596221,
+      "loss": 6.3923,
+      "mean_token_accuracy": 0.12098478376865388,
+      "num_tokens": 2359699.0,
+      "step": 1320
+    },
+    {
+      "entropy": 6.4792803764343265,
+      "epoch": 0.06815317748116144,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00048717287057589454,
+      "loss": 6.1595,
+      "mean_token_accuracy": 0.12839691638946532,
+      "num_tokens": 2368782.0,
+      "step": 1325
+    },
+    {
+      "entropy": 6.383249664306641,
+      "epoch": 0.06841035928297713,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004867778701657417,
+      "loss": 6.0096,
+      "mean_token_accuracy": 0.13362448140978814,
+      "num_tokens": 2377235.0,
+      "step": 1330
+    },
+    {
+      "entropy": 6.409832763671875,
+      "epoch": 0.06866754108479284,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00048637706375829955,
+      "loss": 6.2013,
+      "mean_token_accuracy": 0.12747161388397216,
+      "num_tokens": 2386265.0,
+      "step": 1335
+    },
+    {
+      "entropy": 6.527412176132202,
+      "epoch": 0.06892472288660854,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000485970462341878,
+      "loss": 6.1131,
+      "mean_token_accuracy": 0.1332041934132576,
+      "num_tokens": 2394953.0,
+      "step": 1340
+    },
+    {
+      "entropy": 6.503418636322022,
+      "epoch": 0.06918190468842425,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00048555807706366044,
+      "loss": 6.3003,
+      "mean_token_accuracy": 0.11948163360357285,
+      "num_tokens": 2403792.0,
+      "step": 1345
+    },
+    {
+      "entropy": 6.4240118026733395,
+      "epoch": 0.06943908649023996,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00048513991922939756,
+      "loss": 6.1478,
+      "mean_token_accuracy": 0.1259373866021633,
+      "num_tokens": 2412252.0,
+      "step": 1350
+    },
+    {
+      "entropy": 6.6844557762146,
+      "epoch": 0.06969626829205565,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00048471600030309744,
+      "loss": 6.2847,
+      "mean_token_accuracy": 0.1220081552863121,
+      "num_tokens": 2420364.0,
+      "step": 1355
+    },
+    {
+      "entropy": 6.442555618286133,
+      "epoch": 0.06995345009387136,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00048428633190671186,
+      "loss": 6.1597,
+      "mean_token_accuracy": 0.12498654872179031,
+      "num_tokens": 2429333.0,
+      "step": 1360
+    },
+    {
+      "entropy": 6.515508556365967,
+      "epoch": 0.07021063189568706,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004838509258198167,
+      "loss": 6.2702,
+      "mean_token_accuracy": 0.12291633188724518,
+      "num_tokens": 2437439.0,
+      "step": 1365
+    },
+    {
+      "entropy": 6.4559262752532955,
+      "epoch": 0.07046781369750277,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00048340979397929,
+      "loss": 6.2392,
+      "mean_token_accuracy": 0.12877265959978104,
+      "num_tokens": 2446084.0,
+      "step": 1370
+    },
+    {
+      "entropy": 6.574571943283081,
+      "epoch": 0.07072499549931847,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00048296294847898386,
+      "loss": 6.2038,
+      "mean_token_accuracy": 0.11378729492425918,
+      "num_tokens": 2455613.0,
+      "step": 1375
+    },
+    {
+      "entropy": 6.547432279586792,
+      "epoch": 0.07098217730113417,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0004825104015693934,
+      "loss": 6.336,
+      "mean_token_accuracy": 0.12563420534133912,
+      "num_tokens": 2464123.0,
+      "step": 1380
+    },
+    {
+      "entropy": 6.496583318710327,
+      "epoch": 0.07123935910294987,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004820521656573208,
+      "loss": 6.2003,
+      "mean_token_accuracy": 0.12900396287441254,
+      "num_tokens": 2472972.0,
+      "step": 1385
+    },
+    {
+      "entropy": 6.55055980682373,
+      "epoch": 0.07149654090476558,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00048158825330553505,
+      "loss": 6.3127,
+      "mean_token_accuracy": 0.12014695033431053,
+      "num_tokens": 2482017.0,
+      "step": 1390
+    },
+    {
+      "entropy": 6.498235702514648,
+      "epoch": 0.07175372270658129,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00048111867723242763,
+      "loss": 6.1993,
+      "mean_token_accuracy": 0.12744636759161948,
+      "num_tokens": 2490771.0,
+      "step": 1395
+    },
+    {
+      "entropy": 6.492023563385009,
+      "epoch": 0.07201090450839699,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0004806434503116637,
+      "loss": 6.2633,
+      "mean_token_accuracy": 0.12012107148766518,
+      "num_tokens": 2501047.0,
+      "step": 1400
+    },
+    {
+      "entropy": 6.459356260299683,
+      "epoch": 0.07226808631021268,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004801625855718296,
+      "loss": 6.2797,
+      "mean_token_accuracy": 0.12193729802966118,
+      "num_tokens": 2509689.0,
+      "step": 1405
+    },
+    {
+      "entropy": 6.523634672164917,
+      "epoch": 0.07252526811202839,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00047967609619607477,
+      "loss": 6.2162,
+      "mean_token_accuracy": 0.1272864632308483,
+      "num_tokens": 2519067.0,
+      "step": 1410
+    },
+    {
+      "entropy": 6.3656079292297365,
+      "epoch": 0.0727824499138441,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004791839955217513,
+      "loss": 6.1233,
+      "mean_token_accuracy": 0.12799235731363295,
+      "num_tokens": 2527921.0,
+      "step": 1415
+    },
+    {
+      "entropy": 6.555266809463501,
+      "epoch": 0.0730396317156598,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00047868629704004786,
+      "loss": 6.2048,
+      "mean_token_accuracy": 0.12755087167024612,
+      "num_tokens": 2536424.0,
+      "step": 1420
+    },
+    {
+      "entropy": 6.4298278331756595,
+      "epoch": 0.07329681351747551,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00047818301439561965,
+      "loss": 6.1967,
+      "mean_token_accuracy": 0.12071703523397445,
+      "num_tokens": 2546029.0,
+      "step": 1425
+    },
+    {
+      "entropy": 6.526437473297119,
+      "epoch": 0.0735539953192912,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00047767416138621454,
+      "loss": 6.2481,
+      "mean_token_accuracy": 0.12410413399338723,
+      "num_tokens": 2554861.0,
+      "step": 1430
+    },
+    {
+      "entropy": 6.460501337051392,
+      "epoch": 0.07381117712110691,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000477159751962295,
+      "loss": 6.1015,
+      "mean_token_accuracy": 0.12634221836924553,
+      "num_tokens": 2563405.0,
+      "step": 1435
+    },
+    {
+      "entropy": 6.288591575622559,
+      "epoch": 0.07406835892292261,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00047663980022665507,
+      "loss": 6.0832,
+      "mean_token_accuracy": 0.1300654336810112,
+      "num_tokens": 2571409.0,
+      "step": 1440
+    },
+    {
+      "entropy": 6.524644994735718,
+      "epoch": 0.07432554072473832,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00047611432043403437,
+      "loss": 6.323,
+      "mean_token_accuracy": 0.11622402295470238,
+      "num_tokens": 2581233.0,
+      "step": 1445
+    },
+    {
+      "entropy": 6.429956722259521,
+      "epoch": 0.07458272252655403,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004755833269907267,
+      "loss": 6.273,
+      "mean_token_accuracy": 0.11982614472508431,
+      "num_tokens": 2591528.0,
+      "step": 1450
+    },
+    {
+      "entropy": 6.653250694274902,
+      "epoch": 0.07483990432836972,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004750468344541857,
+      "loss": 6.2093,
+      "mean_token_accuracy": 0.1260430246591568,
+      "num_tokens": 2600033.0,
+      "step": 1455
+    },
+    {
+      "entropy": 6.354333925247192,
+      "epoch": 0.07509708613018543,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00047450485753262525,
+      "loss": 6.1861,
+      "mean_token_accuracy": 0.12549542263150215,
+      "num_tokens": 2608648.0,
+      "step": 1460
+    },
+    {
+      "entropy": 6.368370294570923,
+      "epoch": 0.07535426793200113,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00047395741108461633,
+      "loss": 6.0852,
+      "mean_token_accuracy": 0.13009310886263847,
+      "num_tokens": 2617727.0,
+      "step": 1465
+    },
+    {
+      "entropy": 6.389603805541992,
+      "epoch": 0.07561144973381684,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00047340451011867985,
+      "loss": 6.0666,
+      "mean_token_accuracy": 0.13752511888742447,
+      "num_tokens": 2625463.0,
+      "step": 1470
+    },
+    {
+      "entropy": 6.439476728439331,
+      "epoch": 0.07586863153563254,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00047284616979287515,
+      "loss": 6.1768,
+      "mean_token_accuracy": 0.12700055465102195,
+      "num_tokens": 2635050.0,
+      "step": 1475
+    },
+    {
+      "entropy": 6.452608394622803,
+      "epoch": 0.07612581333744824,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00047228240541438433,
+      "loss": 6.2373,
+      "mean_token_accuracy": 0.1270735152065754,
+      "num_tokens": 2645129.0,
+      "step": 1480
+    },
+    {
+      "entropy": 6.402850103378296,
+      "epoch": 0.07638299513926394,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00047171323243909257,
+      "loss": 6.1132,
+      "mean_token_accuracy": 0.12463590428233147,
+      "num_tokens": 2653966.0,
+      "step": 1485
+    },
+    {
+      "entropy": 6.413847303390503,
+      "epoch": 0.07664017694107965,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00047113866647116457,
+      "loss": 6.1521,
+      "mean_token_accuracy": 0.1268475502729416,
+      "num_tokens": 2663219.0,
+      "step": 1490
+    },
+    {
+      "entropy": 6.427061939239502,
+      "epoch": 0.07689735874289536,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004705587232626164,
+      "loss": 6.0832,
+      "mean_token_accuracy": 0.1297985441982746,
+      "num_tokens": 2671468.0,
+      "step": 1495
+    },
+    {
+      "entropy": 6.292563104629517,
+      "epoch": 0.07715454054471106,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00046997341871288424,
+      "loss": 6.036,
+      "mean_token_accuracy": 0.1342359036207199,
+      "num_tokens": 2680380.0,
+      "step": 1500
+    },
+    {
+      "entropy": 6.436664009094239,
+      "epoch": 0.07741172234652675,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004693827688683879,
+      "loss": 6.1687,
+      "mean_token_accuracy": 0.12308031022548675,
+      "num_tokens": 2689976.0,
+      "step": 1505
+    },
+    {
+      "entropy": 6.508772373199463,
+      "epoch": 0.07766890414834246,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004687867899220914,
+      "loss": 6.1793,
+      "mean_token_accuracy": 0.12510153502225876,
+      "num_tokens": 2698780.0,
+      "step": 1510
+    },
+    {
+      "entropy": 6.226078796386719,
+      "epoch": 0.07792608595015817,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00046818549821305846,
+      "loss": 6.03,
+      "mean_token_accuracy": 0.13029744401574134,
+      "num_tokens": 2706887.0,
+      "step": 1515
+    },
+    {
+      "entropy": 6.41196641921997,
+      "epoch": 0.07818326775197387,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00046757891022600494,
+      "loss": 6.0603,
+      "mean_token_accuracy": 0.1270065888762474,
+      "num_tokens": 2715484.0,
+      "step": 1520
+    },
+    {
+      "entropy": 6.334155130386352,
+      "epoch": 0.07844044955378958,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004669670425908471,
+      "loss": 6.1788,
+      "mean_token_accuracy": 0.12362223640084266,
+      "num_tokens": 2724272.0,
+      "step": 1525
+    },
+    {
+      "entropy": 6.397065687179565,
+      "epoch": 0.07869763135560527,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004663499120822451,
+      "loss": 6.0992,
+      "mean_token_accuracy": 0.13246706426143645,
+      "num_tokens": 2733379.0,
+      "step": 1530
+    },
+    {
+      "entropy": 6.365872526168824,
+      "epoch": 0.07895481315742098,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004657275356191437,
+      "loss": 6.0708,
+      "mean_token_accuracy": 0.13027092143893243,
+      "num_tokens": 2742584.0,
+      "step": 1535
+    },
+    {
+      "entropy": 6.467748212814331,
+      "epoch": 0.07921199495923668,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00046509993026430804,
+      "loss": 6.2481,
+      "mean_token_accuracy": 0.12418549954891205,
+      "num_tokens": 2751491.0,
+      "step": 1540
+    },
+    {
+      "entropy": 6.414959955215454,
+      "epoch": 0.07946917676105239,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004644671132238558,
+      "loss": 6.0621,
+      "mean_token_accuracy": 0.1291202872991562,
+      "num_tokens": 2760320.0,
+      "step": 1545
+    },
+    {
+      "entropy": 6.405585384368896,
+      "epoch": 0.0797263585628681,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00046382910184678585,
+      "loss": 6.2345,
+      "mean_token_accuracy": 0.11954343914985657,
+      "num_tokens": 2770247.0,
+      "step": 1550
+    },
+    {
+      "entropy": 6.431417655944824,
+      "epoch": 0.07998354036468379,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004631859136245025,
+      "loss": 6.0881,
+      "mean_token_accuracy": 0.13392823338508605,
+      "num_tokens": 2779969.0,
+      "step": 1555
+    },
+    {
+      "entropy": 6.324232196807861,
+      "epoch": 0.0802407221664995,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004625375661903357,
+      "loss": 6.1817,
+      "mean_token_accuracy": 0.13248204365372657,
+      "num_tokens": 2788584.0,
+      "step": 1560
+    },
+    {
+      "entropy": 6.307961273193359,
+      "epoch": 0.0804979039683152,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00046188407731905787,
+      "loss": 6.0618,
+      "mean_token_accuracy": 0.13163200318813323,
+      "num_tokens": 2796982.0,
+      "step": 1565
+    },
+    {
+      "entropy": 6.487869310379028,
+      "epoch": 0.08075508577013091,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00046122546492639643,
+      "loss": 6.2135,
+      "mean_token_accuracy": 0.12365017086267471,
+      "num_tokens": 2806788.0,
+      "step": 1570
+    },
+    {
+      "entropy": 6.321759557723999,
+      "epoch": 0.08101226757194661,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000460561747068543,
+      "loss": 6.0832,
+      "mean_token_accuracy": 0.12752610817551613,
+      "num_tokens": 2816086.0,
+      "step": 1575
+    },
+    {
+      "entropy": 6.285551452636719,
+      "epoch": 0.0812694493737623,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004598929419416578,
+      "loss": 6.084,
+      "mean_token_accuracy": 0.12898513302206993,
+      "num_tokens": 2825623.0,
+      "step": 1580
+    },
+    {
+      "entropy": 6.406826734542847,
+      "epoch": 0.08152663117557801,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00045921906788137123,
+      "loss": 6.229,
+      "mean_token_accuracy": 0.1251101866364479,
+      "num_tokens": 2835241.0,
+      "step": 1585
+    },
+    {
+      "entropy": 6.380240058898925,
+      "epoch": 0.08178381297739372,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00045854014336228115,
+      "loss": 6.0826,
+      "mean_token_accuracy": 0.1333278901875019,
+      "num_tokens": 2843428.0,
+      "step": 1590
+    },
+    {
+      "entropy": 6.348821783065796,
+      "epoch": 0.08204099477920943,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00045785618699744615,
+      "loss": 6.0808,
+      "mean_token_accuracy": 0.1321525380015373,
+      "num_tokens": 2852662.0,
+      "step": 1595
+    },
+    {
+      "entropy": 6.395787191390991,
+      "epoch": 0.08229817658102513,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00045716721753787543,
+      "loss": 6.0908,
+      "mean_token_accuracy": 0.13208872005343436,
+      "num_tokens": 2861094.0,
+      "step": 1600
+    },
+    {
+      "entropy": 6.31074743270874,
+      "epoch": 0.08255535838284082,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004564732538720148,
+      "loss": 5.9817,
+      "mean_token_accuracy": 0.1272851899266243,
+      "num_tokens": 2870214.0,
+      "step": 1605
+    },
+    {
+      "entropy": 6.283258056640625,
+      "epoch": 0.08281254018465653,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00045577431502522877,
+      "loss": 6.1137,
+      "mean_token_accuracy": 0.1351046971976757,
+      "num_tokens": 2878876.0,
+      "step": 1610
+    },
+    {
+      "entropy": 6.478934669494629,
+      "epoch": 0.08306972198647224,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004550704201592787,
+      "loss": 6.1111,
+      "mean_token_accuracy": 0.12919269502162933,
+      "num_tokens": 2888111.0,
+      "step": 1615
+    },
+    {
+      "entropy": 6.340307664871216,
+      "epoch": 0.08332690378828794,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004543615885717981,
+      "loss": 6.1395,
+      "mean_token_accuracy": 0.13248469904065133,
+      "num_tokens": 2897589.0,
+      "step": 1620
+    },
+    {
+      "entropy": 6.388657903671264,
+      "epoch": 0.08358408559010365,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00045364783969576296,
+      "loss": 6.0822,
+      "mean_token_accuracy": 0.1332689106464386,
+      "num_tokens": 2906803.0,
+      "step": 1625
+    },
+    {
+      "entropy": 6.2963072776794435,
+      "epoch": 0.08384126739191934,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004529291930989592,
+      "loss": 5.9716,
+      "mean_token_accuracy": 0.13755856528878213,
+      "num_tokens": 2916903.0,
+      "step": 1630
+    },
+    {
+      "entropy": 6.342654514312744,
+      "epoch": 0.08409844919373505,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004522056684834464,
+      "loss": 6.0949,
+      "mean_token_accuracy": 0.13052690103650094,
+      "num_tokens": 2925952.0,
+      "step": 1635
+    },
+    {
+      "entropy": 6.325797176361084,
+      "epoch": 0.08435563099555075,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004514772856850173,
+      "loss": 6.0416,
+      "mean_token_accuracy": 0.12649609819054602,
+      "num_tokens": 2934538.0,
+      "step": 1640
+    },
+    {
+      "entropy": 6.3612961769104,
+      "epoch": 0.08461281279736646,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0004507440646726542,
+      "loss": 6.1471,
+      "mean_token_accuracy": 0.132496327906847,
+      "num_tokens": 2943553.0,
+      "step": 1645
+    },
+    {
+      "entropy": 6.292884588241577,
+      "epoch": 0.08486999459918217,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004500060255479818,
+      "loss": 6.031,
+      "mean_token_accuracy": 0.1341935083270073,
+      "num_tokens": 2951854.0,
+      "step": 1650
+    },
+    {
+      "entropy": 6.448796558380127,
+      "epoch": 0.08512717640099786,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004492631885447151,
+      "loss": 6.1977,
+      "mean_token_accuracy": 0.12419203296303749,
+      "num_tokens": 2961684.0,
+      "step": 1655
+    },
+    {
+      "entropy": 6.379677057266235,
+      "epoch": 0.08538435820281357,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00044851557402810616,
+      "loss": 6.0526,
+      "mean_token_accuracy": 0.12697965502738953,
+      "num_tokens": 2970984.0,
+      "step": 1660
+    },
+    {
+      "entropy": 6.304157495498657,
+      "epoch": 0.08564154000462927,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00044776320249438444,
+      "loss": 6.0435,
+      "mean_token_accuracy": 0.12820193246006967,
+      "num_tokens": 2979319.0,
+      "step": 1665
+    },
+    {
+      "entropy": 6.231897449493408,
+      "epoch": 0.08589872180644498,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00044700609457019565,
+      "loss": 5.9506,
+      "mean_token_accuracy": 0.1356399580836296,
+      "num_tokens": 2987780.0,
+      "step": 1670
+    },
+    {
+      "entropy": 6.239076566696167,
+      "epoch": 0.08615590360826068,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.0004462442710120359,
+      "loss": 6.0657,
+      "mean_token_accuracy": 0.12994374260306357,
+      "num_tokens": 2996890.0,
+      "step": 1675
+    },
+    {
+      "entropy": 6.530197525024414,
+      "epoch": 0.08641308541007638,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.000445477752705683,
+      "loss": 6.1939,
+      "mean_token_accuracy": 0.12460953146219253,
+      "num_tokens": 3005736.0,
+      "step": 1680
+    },
+    {
+      "entropy": 6.189916801452637,
+      "epoch": 0.08667026721189208,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00044470656066562336,
+      "loss": 5.9827,
+      "mean_token_accuracy": 0.13694657981395722,
+      "num_tokens": 3014167.0,
+      "step": 1685
+    },
+    {
+      "entropy": 6.1757958889007565,
+      "epoch": 0.08692744901370779,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004439307160344765,
+      "loss": 5.9809,
+      "mean_token_accuracy": 0.140267164260149,
+      "num_tokens": 3023018.0,
+      "step": 1690
+    },
+    {
+      "entropy": 6.499619054794311,
+      "epoch": 0.0871846308155235,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.00044315024008241473,
+      "loss": 6.1587,
+      "mean_token_accuracy": 0.12271130084991455,
+      "num_tokens": 3032659.0,
+      "step": 1695
+    },
+    {
+      "entropy": 6.313583087921143,
+      "epoch": 0.0874418126173392,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0004423651542065806,
+      "loss": 5.9887,
+      "mean_token_accuracy": 0.13750137612223626,
+      "num_tokens": 3040882.0,
+      "step": 1700
+    },
+    {
+      "entropy": 6.182634162902832,
+      "epoch": 0.0876989944191549,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00044157547993050006,
+      "loss": 5.9244,
+      "mean_token_accuracy": 0.13822244331240655,
+      "num_tokens": 3050242.0,
+      "step": 1705
+    },
+    {
+      "entropy": 6.321706867218017,
+      "epoch": 0.0879561762209706,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00044078123890349227,
+      "loss": 6.0384,
+      "mean_token_accuracy": 0.12941291332244872,
+      "num_tokens": 3059928.0,
+      "step": 1710
+    },
+    {
+      "entropy": 6.401517152786255,
+      "epoch": 0.0882133580227863,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00043998245290007606,
+      "loss": 6.2233,
+      "mean_token_accuracy": 0.12163035720586776,
+      "num_tokens": 3069787.0,
+      "step": 1715
+    },
+    {
+      "entropy": 6.403897905349732,
+      "epoch": 0.08847053982460201,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00043917914381937323,
+      "loss": 6.1021,
+      "mean_token_accuracy": 0.12949455678462982,
+      "num_tokens": 3079171.0,
+      "step": 1720
+    },
+    {
+      "entropy": 6.267082691192627,
+      "epoch": 0.08872772162641772,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00043837133368450815,
+      "loss": 6.0348,
+      "mean_token_accuracy": 0.13504951894283296,
+      "num_tokens": 3088371.0,
+      "step": 1725
+    },
+    {
+      "entropy": 6.264813995361328,
+      "epoch": 0.08898490342823343,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004375590446420037,
+      "loss": 6.0367,
+      "mean_token_accuracy": 0.13473087251186372,
+      "num_tokens": 3098489.0,
+      "step": 1730
+    },
+    {
+      "entropy": 6.363276672363281,
+      "epoch": 0.08924208523004912,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0004367422989611743,
+      "loss": 6.1169,
+      "mean_token_accuracy": 0.12966250479221345,
+      "num_tokens": 3107161.0,
+      "step": 1735
+    },
+    {
+      "entropy": 6.385274219512939,
+      "epoch": 0.08949926703186482,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004359211190335153,
+      "loss": 6.0351,
+      "mean_token_accuracy": 0.13608441054821013,
+      "num_tokens": 3115435.0,
+      "step": 1740
+    },
+    {
+      "entropy": 6.3607177734375,
+      "epoch": 0.08975644883368053,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00043509552737208923,
+      "loss": 6.1034,
+      "mean_token_accuracy": 0.13773850798606874,
+      "num_tokens": 3124543.0,
+      "step": 1745
+    },
+    {
+      "entropy": 6.311847782135009,
+      "epoch": 0.09001363063549624,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.00043426554661090853,
+      "loss": 5.9807,
+      "mean_token_accuracy": 0.12751075103878975,
+      "num_tokens": 3133856.0,
+      "step": 1750
+    },
+    {
+      "entropy": 6.20699553489685,
+      "epoch": 0.09027081243731194,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00043343119950431516,
+      "loss": 6.0226,
+      "mean_token_accuracy": 0.13111255168914795,
+      "num_tokens": 3143643.0,
+      "step": 1755
+    },
+    {
+      "entropy": 6.388123416900635,
+      "epoch": 0.09052799423912763,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00043259250892635644,
+      "loss": 6.0525,
+      "mean_token_accuracy": 0.13792511597275733,
+      "num_tokens": 3152084.0,
+      "step": 1760
+    },
+    {
+      "entropy": 6.389042806625366,
+      "epoch": 0.09078517604094334,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004317494978701582,
+      "loss": 6.0266,
+      "mean_token_accuracy": 0.13580404818058014,
+      "num_tokens": 3160091.0,
+      "step": 1765
+    },
+    {
+      "entropy": 6.2101809024810795,
+      "epoch": 0.09104235784275905,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004309021894472943,
+      "loss": 6.014,
+      "mean_token_accuracy": 0.12847821563482284,
+      "num_tokens": 3169569.0,
+      "step": 1770
+    },
+    {
+      "entropy": 6.378039360046387,
+      "epoch": 0.09129953964457475,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004300506068871534,
+      "loss": 6.1253,
+      "mean_token_accuracy": 0.12418064251542091,
+      "num_tokens": 3179974.0,
+      "step": 1775
+    },
+    {
+      "entropy": 6.311042928695679,
+      "epoch": 0.09155672144639046,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00042919477353630135,
+      "loss": 6.0914,
+      "mean_token_accuracy": 0.13122028410434722,
+      "num_tokens": 3189489.0,
+      "step": 1780
+    },
+    {
+      "entropy": 6.401360130310058,
+      "epoch": 0.09181390324820615,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.000428334712857842,
+      "loss": 6.1483,
+      "mean_token_accuracy": 0.1267679288983345,
+      "num_tokens": 3198335.0,
+      "step": 1785
+    },
+    {
+      "entropy": 6.154367589950562,
+      "epoch": 0.09207108505002186,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00042747044843077304,
+      "loss": 5.8831,
+      "mean_token_accuracy": 0.1401224859058857,
+      "num_tokens": 3206400.0,
+      "step": 1790
+    },
+    {
+      "entropy": 6.391241455078125,
+      "epoch": 0.09232826685183756,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00042660200394934047,
+      "loss": 5.9885,
+      "mean_token_accuracy": 0.1355439156293869,
+      "num_tokens": 3214480.0,
+      "step": 1795
+    },
+    {
+      "entropy": 6.293077850341797,
+      "epoch": 0.09258544865365327,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00042572940322238844,
+      "loss": 6.0104,
+      "mean_token_accuracy": 0.13572105020284653,
+      "num_tokens": 3222386.0,
+      "step": 1800
+    },
+    {
+      "entropy": 6.25052399635315,
+      "epoch": 0.09284263045546898,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00042485267017270664,
+      "loss": 5.9602,
+      "mean_token_accuracy": 0.13466379642486573,
+      "num_tokens": 3231203.0,
+      "step": 1805
+    },
+    {
+      "entropy": 6.233933353424073,
+      "epoch": 0.09309981225728467,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004239718288363745,
+      "loss": 6.0089,
+      "mean_token_accuracy": 0.1300691097974777,
+      "num_tokens": 3240693.0,
+      "step": 1810
+    },
+    {
+      "entropy": 6.2747550964355465,
+      "epoch": 0.09335699405910038,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004230869033621023,
+      "loss": 6.0764,
+      "mean_token_accuracy": 0.13697424307465553,
+      "num_tokens": 3249518.0,
+      "step": 1815
+    },
+    {
+      "entropy": 6.3998106002807615,
+      "epoch": 0.09361417586091608,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004221979180105688,
+      "loss": 6.0622,
+      "mean_token_accuracy": 0.1327414371073246,
+      "num_tokens": 3258712.0,
+      "step": 1820
+    },
+    {
+      "entropy": 6.279940986633301,
+      "epoch": 0.09387135766273179,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00042130489715375645,
+      "loss": 5.9342,
+      "mean_token_accuracy": 0.140862063318491,
+      "num_tokens": 3266797.0,
+      "step": 1825
+    },
+    {
+      "entropy": 6.2281848907470705,
+      "epoch": 0.0941285394645475,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00042040786527428335,
+      "loss": 6.0584,
+      "mean_token_accuracy": 0.13048559799790382,
+      "num_tokens": 3275811.0,
+      "step": 1830
+    },
+    {
+      "entropy": 6.248003005981445,
+      "epoch": 0.09438572126636319,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004195068469647315,
+      "loss": 5.9652,
+      "mean_token_accuracy": 0.13584049865603448,
+      "num_tokens": 3284813.0,
+      "step": 1835
+    },
+    {
+      "entropy": 6.284520101547241,
+      "epoch": 0.0946429030681789,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00041860186692697297,
+      "loss": 6.0291,
+      "mean_token_accuracy": 0.1311030738055706,
+      "num_tokens": 3293509.0,
+      "step": 1840
+    },
+    {
+      "entropy": 6.319161224365234,
+      "epoch": 0.0949000848699946,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00041769294997149264,
+      "loss": 5.9418,
+      "mean_token_accuracy": 0.1394168458878994,
+      "num_tokens": 3302010.0,
+      "step": 1845
+    },
+    {
+      "entropy": 6.28194146156311,
+      "epoch": 0.0951572666718103,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.0004167801210167081,
+      "loss": 6.0737,
+      "mean_token_accuracy": 0.13149803429841994,
+      "num_tokens": 3312307.0,
+      "step": 1850
+    },
+    {
+      "entropy": 6.298348474502563,
+      "epoch": 0.09541444847362601,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004158634050882861,
+      "loss": 6.0619,
+      "mean_token_accuracy": 0.12632458582520484,
+      "num_tokens": 3321859.0,
+      "step": 1855
+    },
+    {
+      "entropy": 6.352298450469971,
+      "epoch": 0.0956716302754417,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.0004149428273184569,
+      "loss": 6.0642,
+      "mean_token_accuracy": 0.13412729203701018,
+      "num_tokens": 3331727.0,
+      "step": 1860
+    },
+    {
+      "entropy": 6.263995742797851,
+      "epoch": 0.09592881207725741,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004140184129453253,
+      "loss": 5.9321,
+      "mean_token_accuracy": 0.13026233464479448,
+      "num_tokens": 3340449.0,
+      "step": 1865
+    },
+    {
+      "entropy": 6.30117506980896,
+      "epoch": 0.09618599387907312,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.000413090187312178,
+      "loss": 5.9886,
+      "mean_token_accuracy": 0.13512412831187248,
+      "num_tokens": 3349374.0,
+      "step": 1870
+    },
+    {
+      "entropy": 6.266588401794434,
+      "epoch": 0.09644317568088882,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0004121581758667898,
+      "loss": 6.0195,
+      "mean_token_accuracy": 0.13124436065554618,
+      "num_tokens": 3358792.0,
+      "step": 1875
+    },
+    {
+      "entropy": 6.184923267364502,
+      "epoch": 0.09670035748270453,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.00041122240416072533,
+      "loss": 5.9184,
+      "mean_token_accuracy": 0.13883327171206475,
+      "num_tokens": 3368066.0,
+      "step": 1880
+    },
+    {
+      "entropy": 6.24156346321106,
+      "epoch": 0.09695753928452022,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004102828978486385,
+      "loss": 5.954,
+      "mean_token_accuracy": 0.13969675824046135,
+      "num_tokens": 3376872.0,
+      "step": 1885
+    },
+    {
+      "entropy": 6.241713762283325,
+      "epoch": 0.09721472108633593,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004093396826875695,
+      "loss": 5.9471,
+      "mean_token_accuracy": 0.13266203328967094,
+      "num_tokens": 3385851.0,
+      "step": 1890
+    },
+    {
+      "entropy": 6.231948804855347,
+      "epoch": 0.09747190288815163,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00040839278453623837,
+      "loss": 6.0056,
+      "mean_token_accuracy": 0.1346826121211052,
+      "num_tokens": 3395263.0,
+      "step": 1895
+    },
+    {
+      "entropy": 6.228415203094483,
+      "epoch": 0.09772908468996734,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004074422293543363,
+      "loss": 6.0445,
+      "mean_token_accuracy": 0.13088179305195807,
+      "num_tokens": 3404814.0,
+      "step": 1900
+    },
+    {
+      "entropy": 6.329216241836548,
+      "epoch": 0.09798626649178305,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004064880432018137,
+      "loss": 6.1074,
+      "mean_token_accuracy": 0.1315300315618515,
+      "num_tokens": 3413822.0,
+      "step": 1905
+    },
+    {
+      "entropy": 6.301573801040649,
+      "epoch": 0.09824344829359874,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00040553025223816615,
+      "loss": 6.0124,
+      "mean_token_accuracy": 0.13628942891955376,
+      "num_tokens": 3422318.0,
+      "step": 1910
+    },
+    {
+      "entropy": 6.206828165054321,
+      "epoch": 0.09850063009541445,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00040456888272171653,
+      "loss": 5.9329,
+      "mean_token_accuracy": 0.13764095306396484,
+      "num_tokens": 3431066.0,
+      "step": 1915
+    },
+    {
+      "entropy": 6.251052141189575,
+      "epoch": 0.09875781189723015,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00040360396100889577,
+      "loss": 5.9485,
+      "mean_token_accuracy": 0.1366151362657547,
+      "num_tokens": 3439822.0,
+      "step": 1920
+    },
+    {
+      "entropy": 6.248995113372803,
+      "epoch": 0.09901499369904586,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004026355135535202,
+      "loss": 6.0449,
+      "mean_token_accuracy": 0.13020109459757806,
+      "num_tokens": 3449430.0,
+      "step": 1925
+    },
+    {
+      "entropy": 6.266049480438232,
+      "epoch": 0.09927217550086156,
+      "grad_norm": 1.125,
+      "learning_rate": 0.000401663566906066,
+      "loss": 5.9646,
+      "mean_token_accuracy": 0.13827199935913087,
+      "num_tokens": 3457519.0,
+      "step": 1930
+    },
+    {
+      "entropy": 6.185720539093017,
+      "epoch": 0.09952935730267726,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00040068814771294134,
+      "loss": 5.8726,
+      "mean_token_accuracy": 0.13870462104678155,
+      "num_tokens": 3466328.0,
+      "step": 1935
+    },
+    {
+      "entropy": 6.133827590942383,
+      "epoch": 0.09978653910449296,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0003997092827157562,
+      "loss": 5.8982,
+      "mean_token_accuracy": 0.1401980772614479,
+      "num_tokens": 3475821.0,
+      "step": 1940
+    },
+    {
+      "entropy": 6.2277813911437985,
+      "epoch": 0.10004372090630867,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000398726998750589,
+      "loss": 5.9881,
+      "mean_token_accuracy": 0.1336817152798176,
+      "num_tokens": 3484902.0,
+      "step": 1945
+    },
+    {
+      "entropy": 6.296745872497558,
+      "epoch": 0.10030090270812438,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00039774132274725076,
+      "loss": 6.0906,
+      "mean_token_accuracy": 0.12511072233319281,
+      "num_tokens": 3494951.0,
+      "step": 1950
+    },
+    {
+      "entropy": 6.261813402175903,
+      "epoch": 0.10055808450994008,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00039675228172854707,
+      "loss": 5.9523,
+      "mean_token_accuracy": 0.14045180454850198,
+      "num_tokens": 3503688.0,
+      "step": 1955
+    },
+    {
+      "entropy": 6.224357604980469,
+      "epoch": 0.10081526631175577,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003957599028095371,
+      "loss": 5.9971,
+      "mean_token_accuracy": 0.13452794998884202,
+      "num_tokens": 3512337.0,
+      "step": 1960
+    },
+    {
+      "entropy": 6.310317754745483,
+      "epoch": 0.10107244811357148,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00039476421319679017,
+      "loss": 6.0837,
+      "mean_token_accuracy": 0.1328693710267544,
+      "num_tokens": 3521370.0,
+      "step": 1965
+    },
+    {
+      "entropy": 6.193576526641846,
+      "epoch": 0.10132962991538719,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00039376524018764,
+      "loss": 5.8877,
+      "mean_token_accuracy": 0.14604497849941253,
+      "num_tokens": 3530225.0,
+      "step": 1970
+    },
+    {
+      "entropy": 6.205514764785766,
+      "epoch": 0.1015868117172029,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00039276301116943616,
+      "loss": 5.9687,
+      "mean_token_accuracy": 0.14247402623295785,
+      "num_tokens": 3538843.0,
+      "step": 1975
+    },
+    {
+      "entropy": 6.286650371551514,
+      "epoch": 0.1018439935190186,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0003917575536187936,
+      "loss": 5.9858,
+      "mean_token_accuracy": 0.1346484936773777,
+      "num_tokens": 3547833.0,
+      "step": 1980
+    },
+    {
+      "entropy": 6.177884578704834,
+      "epoch": 0.10210117532083429,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00039074889510083894,
+      "loss": 5.9072,
+      "mean_token_accuracy": 0.1447339817881584,
+      "num_tokens": 3556021.0,
+      "step": 1985
+    },
+    {
+      "entropy": 6.251054954528809,
+      "epoch": 0.10235835712265,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00038973706326845495,
+      "loss": 5.8984,
+      "mean_token_accuracy": 0.13767489194869995,
+      "num_tokens": 3564483.0,
+      "step": 1990
+    },
+    {
+      "entropy": 6.208160734176635,
+      "epoch": 0.1026155389244657,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0003887220858615225,
+      "loss": 5.9995,
+      "mean_token_accuracy": 0.12966525331139564,
+      "num_tokens": 3574761.0,
+      "step": 1995
+    },
+    {
+      "entropy": 6.26411509513855,
+      "epoch": 0.10287272072628141,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0003877039907061597,
+      "loss": 6.0126,
+      "mean_token_accuracy": 0.13828732147812844,
+      "num_tokens": 3583401.0,
+      "step": 2000
+    },
+    {
+      "entropy": 6.262833213806152,
+      "epoch": 0.10312990252809712,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0003866828057139598,
+      "loss": 5.8656,
+      "mean_token_accuracy": 0.14422237426042556,
+      "num_tokens": 3592534.0,
+      "step": 2005
+    },
+    {
+      "entropy": 6.177701854705811,
+      "epoch": 0.10338708432991281,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00038565855888122503,
+      "loss": 5.9444,
+      "mean_token_accuracy": 0.1334696665406227,
+      "num_tokens": 3601578.0,
+      "step": 2010
+    },
+    {
+      "entropy": 6.221395444869995,
+      "epoch": 0.10364426613172852,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00038463127828819975,
+      "loss": 5.9308,
+      "mean_token_accuracy": 0.1387752816081047,
+      "num_tokens": 3610069.0,
+      "step": 2015
+    },
+    {
+      "entropy": 6.283290386199951,
+      "epoch": 0.10390144793354422,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00038360099209830043,
+      "loss": 5.8238,
+      "mean_token_accuracy": 0.1489806979894638,
+      "num_tokens": 3618959.0,
+      "step": 2020
+    },
+    {
+      "entropy": 6.1479912281036375,
+      "epoch": 0.10415862973535993,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0003825677285573433,
+      "loss": 5.9255,
+      "mean_token_accuracy": 0.1377922661602497,
+      "num_tokens": 3627675.0,
+      "step": 2025
+    },
+    {
+      "entropy": 6.19081678390503,
+      "epoch": 0.10441581153717563,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00038153151599277027,
+      "loss": 5.9258,
+      "mean_token_accuracy": 0.1381417006254196,
+      "num_tokens": 3636183.0,
+      "step": 2030
+    },
+    {
+      "entropy": 6.213774585723877,
+      "epoch": 0.10467299333899133,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003804923828128723,
+      "loss": 5.902,
+      "mean_token_accuracy": 0.14034309834241868,
+      "num_tokens": 3644523.0,
+      "step": 2035
+    },
+    {
+      "entropy": 6.226733255386352,
+      "epoch": 0.10493017514080703,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0003794503575060104,
+      "loss": 6.0011,
+      "mean_token_accuracy": 0.13573786467313767,
+      "num_tokens": 3653199.0,
+      "step": 2040
+    },
+    {
+      "entropy": 6.145754480361939,
+      "epoch": 0.10518735694262274,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00037840546863983484,
+      "loss": 5.8755,
+      "mean_token_accuracy": 0.1409320466220379,
+      "num_tokens": 3661401.0,
+      "step": 2045
+    },
+    {
+      "entropy": 6.191087007522583,
+      "epoch": 0.10544453874443845,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0003773577448605015,
+      "loss": 5.9555,
+      "mean_token_accuracy": 0.14488928020000458,
+      "num_tokens": 3670665.0,
+      "step": 2050
+    },
+    {
+      "entropy": 6.2919593334198,
+      "epoch": 0.10570172054625415,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003763072148918872,
+      "loss": 6.0581,
+      "mean_token_accuracy": 0.132058534771204,
+      "num_tokens": 3679737.0,
+      "step": 2055
+    },
+    {
+      "entropy": 6.227366399765015,
+      "epoch": 0.10595890234806984,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0003752539075348017,
+      "loss": 5.8594,
+      "mean_token_accuracy": 0.13849994465708731,
+      "num_tokens": 3688427.0,
+      "step": 2060
+    },
+    {
+      "entropy": 6.180734157562256,
+      "epoch": 0.10621608414988555,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00037419785166619817,
+      "loss": 5.9415,
+      "mean_token_accuracy": 0.13965051546692847,
+      "num_tokens": 3697234.0,
+      "step": 2065
+    },
+    {
+      "entropy": 6.145251703262329,
+      "epoch": 0.10647326595170126,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.0003731390762383818,
+      "loss": 5.8466,
+      "mean_token_accuracy": 0.14111215248703957,
+      "num_tokens": 3706639.0,
+      "step": 2070
+    },
+    {
+      "entropy": 6.1391058444976805,
+      "epoch": 0.10673044775351696,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.0003720776102782158,
+      "loss": 5.8297,
+      "mean_token_accuracy": 0.14451524391770362,
+      "num_tokens": 3715282.0,
+      "step": 2075
+    },
+    {
+      "entropy": 6.170806932449341,
+      "epoch": 0.10698762955533267,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00037101348288632555,
+      "loss": 5.8653,
+      "mean_token_accuracy": 0.13939128443598747,
+      "num_tokens": 3724262.0,
+      "step": 2080
+    },
+    {
+      "entropy": 6.220163202285766,
+      "epoch": 0.10724481135714836,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0003699467232363012,
+      "loss": 5.8956,
+      "mean_token_accuracy": 0.14401227980852127,
+      "num_tokens": 3732687.0,
+      "step": 2085
+    },
+    {
+      "entropy": 6.141675901412964,
+      "epoch": 0.10750199315896407,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0003688773605738973,
+      "loss": 5.8874,
+      "mean_token_accuracy": 0.14037907868623734,
+      "num_tokens": 3741952.0,
+      "step": 2090
+    },
+    {
+      "entropy": 6.175952386856079,
+      "epoch": 0.10775917496077977,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00036780542421623134,
+      "loss": 5.8905,
+      "mean_token_accuracy": 0.1411571592092514,
+      "num_tokens": 3750629.0,
+      "step": 2095
+    },
+    {
+      "entropy": 6.2291477680206295,
+      "epoch": 0.10801635676259548,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0003667309435509802,
+      "loss": 5.8878,
+      "mean_token_accuracy": 0.13913479447364807,
+      "num_tokens": 3758741.0,
+      "step": 2100
+    },
+    {
+      "entropy": 6.117889499664306,
+      "epoch": 0.10827353856441119,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003656539480355741,
+      "loss": 5.9091,
+      "mean_token_accuracy": 0.137098728120327,
+      "num_tokens": 3766947.0,
+      "step": 2105
+    },
+    {
+      "entropy": 6.259656620025635,
+      "epoch": 0.10853072036622688,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0003645744671963891,
+      "loss": 5.9503,
+      "mean_token_accuracy": 0.14083122685551644,
+      "num_tokens": 3775437.0,
+      "step": 2110
+    },
+    {
+      "entropy": 6.182394504547119,
+      "epoch": 0.10878790216804259,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003634925306279376,
+      "loss": 5.882,
+      "mean_token_accuracy": 0.13808514177799225,
+      "num_tokens": 3784479.0,
+      "step": 2115
+    },
+    {
+      "entropy": 6.155785608291626,
+      "epoch": 0.10904508396985829,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0003624081679920574,
+      "loss": 5.8596,
+      "mean_token_accuracy": 0.14330154284834862,
+      "num_tokens": 3793116.0,
+      "step": 2120
+    },
+    {
+      "entropy": 6.155031490325928,
+      "epoch": 0.109302265771674,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0003613214090170977,
+      "loss": 5.912,
+      "mean_token_accuracy": 0.1452380448579788,
+      "num_tokens": 3802132.0,
+      "step": 2125
+    },
+    {
+      "entropy": 6.152184724807739,
+      "epoch": 0.1095594475734897,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.0003602322834971048,
+      "loss": 5.9048,
+      "mean_token_accuracy": 0.1368635557591915,
+      "num_tokens": 3811227.0,
+      "step": 2130
+    },
+    {
+      "entropy": 6.172056388854981,
+      "epoch": 0.1098166293753054,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003591408212910051,
+      "loss": 5.7761,
+      "mean_token_accuracy": 0.1443565383553505,
+      "num_tokens": 3819390.0,
+      "step": 2135
+    },
+    {
+      "entropy": 6.0653129577636715,
+      "epoch": 0.1100738111771211,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0003580470523217863,
+      "loss": 5.9103,
+      "mean_token_accuracy": 0.13197219669818877,
+      "num_tokens": 3828294.0,
+      "step": 2140
+    },
+    {
+      "entropy": 6.209930038452148,
+      "epoch": 0.11033099297893681,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0003569510065756771,
+      "loss": 5.9416,
+      "mean_token_accuracy": 0.14223602637648583,
+      "num_tokens": 3836449.0,
+      "step": 2145
+    },
+    {
+      "entropy": 6.269498205184936,
+      "epoch": 0.11058817478075252,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.0003558527141013254,
+      "loss": 6.004,
+      "mean_token_accuracy": 0.12444452196359634,
+      "num_tokens": 3846045.0,
+      "step": 2150
+    },
+    {
+      "entropy": 6.1805016040802006,
+      "epoch": 0.11084535658256822,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0003547522050089742,
+      "loss": 5.9299,
+      "mean_token_accuracy": 0.13361917287111283,
+      "num_tokens": 3855279.0,
+      "step": 2155
+    },
+    {
+      "entropy": 6.155153894424439,
+      "epoch": 0.11110253838438391,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00035364950946963606,
+      "loss": 5.7911,
+      "mean_token_accuracy": 0.14686265140771865,
+      "num_tokens": 3864328.0,
+      "step": 2160
+    },
+    {
+      "entropy": 6.148785877227783,
+      "epoch": 0.11135972018619962,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.0003525446577142663,
+      "loss": 5.8913,
+      "mean_token_accuracy": 0.13261692225933075,
+      "num_tokens": 3873601.0,
+      "step": 2165
+    },
+    {
+      "entropy": 6.128220462799073,
+      "epoch": 0.11161690198801533,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00035143768003293395,
+      "loss": 5.7752,
+      "mean_token_accuracy": 0.14257467091083526,
+      "num_tokens": 3882600.0,
+      "step": 2170
+    },
+    {
+      "entropy": 6.230699729919434,
+      "epoch": 0.11187408378983103,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0003503286067739913,
+      "loss": 5.9735,
+      "mean_token_accuracy": 0.13481164574623108,
+      "num_tokens": 3891380.0,
+      "step": 2175
+    },
+    {
+      "entropy": 6.201015424728394,
+      "epoch": 0.11213126559164674,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00034921746834324193,
+      "loss": 5.9691,
+      "mean_token_accuracy": 0.13804642111063004,
+      "num_tokens": 3900760.0,
+      "step": 2180
+    },
+    {
+      "entropy": 6.157192897796631,
+      "epoch": 0.11238844739346243,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003481042952031072,
+      "loss": 5.8215,
+      "mean_token_accuracy": 0.13898649886250497,
+      "num_tokens": 3909992.0,
+      "step": 2185
+    },
+    {
+      "entropy": 6.084528255462646,
+      "epoch": 0.11264562919527814,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0003469891178717911,
+      "loss": 5.8355,
+      "mean_token_accuracy": 0.14062371253967285,
+      "num_tokens": 3919103.0,
+      "step": 2190
+    },
+    {
+      "entropy": 6.174018383026123,
+      "epoch": 0.11290281099709384,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003458719669224436,
+      "loss": 5.8548,
+      "mean_token_accuracy": 0.13782422468066216,
+      "num_tokens": 3927645.0,
+      "step": 2195
+    },
+    {
+      "entropy": 6.1444765567779545,
+      "epoch": 0.11315999279890955,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0003447528729823221,
+      "loss": 5.9277,
+      "mean_token_accuracy": 0.138858176022768,
+      "num_tokens": 3936292.0,
+      "step": 2200
+    },
+    {
+      "entropy": 6.249077701568604,
+      "epoch": 0.11341717460072526,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0003436318667319525,
+      "loss": 5.8706,
+      "mean_token_accuracy": 0.1403427816927433,
+      "num_tokens": 3944799.0,
+      "step": 2205
+    },
+    {
+      "entropy": 6.100788831710815,
+      "epoch": 0.11367435640254096,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00034250897890428716,
+      "loss": 5.815,
+      "mean_token_accuracy": 0.1452298142015934,
+      "num_tokens": 3953904.0,
+      "step": 2210
+    },
+    {
+      "entropy": 6.122757530212402,
+      "epoch": 0.11393153820435666,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003413842402838633,
+      "loss": 5.8773,
+      "mean_token_accuracy": 0.14187178686261176,
+      "num_tokens": 3962687.0,
+      "step": 2215
+    },
+    {
+      "entropy": 6.139140462875366,
+      "epoch": 0.11418872000617236,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00034025768170595834,
+      "loss": 5.7781,
+      "mean_token_accuracy": 0.14838410168886185,
+      "num_tokens": 3971611.0,
+      "step": 2220
+    },
+    {
+      "entropy": 6.095836210250854,
+      "epoch": 0.11444590180798807,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0003391293340557446,
+      "loss": 5.8363,
+      "mean_token_accuracy": 0.14200367480516435,
+      "num_tokens": 3980742.0,
+      "step": 2225
+    },
+    {
+      "entropy": 6.170583009719849,
+      "epoch": 0.11470308360980377,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0003379992282674431,
+      "loss": 5.863,
+      "mean_token_accuracy": 0.14318625777959823,
+      "num_tokens": 3989289.0,
+      "step": 2230
+    },
+    {
+      "entropy": 6.051944446563721,
+      "epoch": 0.11496026541161948,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0003368673953234749,
+      "loss": 5.7969,
+      "mean_token_accuracy": 0.14933174550533296,
+      "num_tokens": 3997320.0,
+      "step": 2235
+    },
+    {
+      "entropy": 6.068231821060181,
+      "epoch": 0.11521744721343517,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00033573386625361176,
+      "loss": 5.8942,
+      "mean_token_accuracy": 0.13985462710261345,
+      "num_tokens": 4006521.0,
+      "step": 2240
+    },
+    {
+      "entropy": 6.175307178497315,
+      "epoch": 0.11547462901525088,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00033459867213412567,
+      "loss": 5.7411,
+      "mean_token_accuracy": 0.15053482055664064,
+      "num_tokens": 4015686.0,
+      "step": 2245
+    },
+    {
+      "entropy": 6.166437149047852,
+      "epoch": 0.11573181081706659,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000333461844086937,
+      "loss": 5.8909,
+      "mean_token_accuracy": 0.13881933465600013,
+      "num_tokens": 4024038.0,
+      "step": 2250
+    },
+    {
+      "entropy": 6.187738752365112,
+      "epoch": 0.11598899261888229,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00033232341327876097,
+      "loss": 5.9012,
+      "mean_token_accuracy": 0.14501129165291787,
+      "num_tokens": 4032138.0,
+      "step": 2255
+    },
+    {
+      "entropy": 6.085036993026733,
+      "epoch": 0.116246174420698,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0003311834109202531,
+      "loss": 5.8445,
+      "mean_token_accuracy": 0.14039366841316223,
+      "num_tokens": 4040460.0,
+      "step": 2260
+    },
+    {
+      "entropy": 6.152256155014038,
+      "epoch": 0.11650335622251369,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00033004186826515416,
+      "loss": 5.8079,
+      "mean_token_accuracy": 0.14548811465501785,
+      "num_tokens": 4049151.0,
+      "step": 2265
+    },
+    {
+      "entropy": 6.053188514709473,
+      "epoch": 0.1167605380243294,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0003288988166094324,
+      "loss": 5.8077,
+      "mean_token_accuracy": 0.14649804309010506,
+      "num_tokens": 4057642.0,
+      "step": 2270
+    },
+    {
+      "entropy": 6.170608472824097,
+      "epoch": 0.1170177198261451,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00032775428729042656,
+      "loss": 5.9238,
+      "mean_token_accuracy": 0.13787836730480194,
+      "num_tokens": 4066997.0,
+      "step": 2275
+    },
+    {
+      "entropy": 6.14677357673645,
+      "epoch": 0.11727490162796081,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.000326608311685986,
+      "loss": 5.8088,
+      "mean_token_accuracy": 0.1431390479207039,
+      "num_tokens": 4076233.0,
+      "step": 2280
+    },
+    {
+      "entropy": 6.0628547191619875,
+      "epoch": 0.11753208342977652,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0003254609212136108,
+      "loss": 5.768,
+      "mean_token_accuracy": 0.151872681081295,
+      "num_tokens": 4084379.0,
+      "step": 2285
+    },
+    {
+      "entropy": 6.142666673660278,
+      "epoch": 0.11778926523159221,
+      "grad_norm": 0.9765625,
+      "learning_rate": 0.00032431214732959036,
+      "loss": 5.8573,
+      "mean_token_accuracy": 0.14202747270464897,
+      "num_tokens": 4094135.0,
+      "step": 2290
+    },
+    {
+      "entropy": 6.119070816040039,
+      "epoch": 0.11804644703340791,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000323162021528141,
+      "loss": 5.8174,
+      "mean_token_accuracy": 0.14889189898967742,
+      "num_tokens": 4103125.0,
+      "step": 2295
+    },
+    {
+      "entropy": 6.128068208694458,
+      "epoch": 0.11830362883522362,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00032201057534054264,
+      "loss": 5.7736,
+      "mean_token_accuracy": 0.1462913542985916,
+      "num_tokens": 4110852.0,
+      "step": 2300
+    },
+    {
+      "entropy": 6.033051919937134,
+      "epoch": 0.11856081063703933,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00032085784033427414,
+      "loss": 5.8141,
+      "mean_token_accuracy": 0.14134329929947853,
+      "num_tokens": 4119911.0,
+      "step": 2305
+    },
+    {
+      "entropy": 6.131434345245362,
+      "epoch": 0.11881799243885503,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0003197038481121478,
+      "loss": 5.7666,
+      "mean_token_accuracy": 0.14859718158841134,
+      "num_tokens": 4128215.0,
+      "step": 2310
+    },
+    {
+      "entropy": 6.168793773651123,
+      "epoch": 0.11907517424067073,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0003185486303114436,
+      "loss": 5.8695,
+      "mean_token_accuracy": 0.13778275102376938,
+      "num_tokens": 4137547.0,
+      "step": 2315
+    },
+    {
+      "entropy": 6.040676689147949,
+      "epoch": 0.11933235604248643,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0003173922186030409,
+      "loss": 5.8115,
+      "mean_token_accuracy": 0.14293806701898576,
+      "num_tokens": 4146532.0,
+      "step": 2320
+    },
+    {
+      "entropy": 6.0469255447387695,
+      "epoch": 0.11958953784430214,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000316234644690551,
+      "loss": 5.7047,
+      "mean_token_accuracy": 0.14709992557764054,
+      "num_tokens": 4154665.0,
+      "step": 2325
+    },
+    {
+      "entropy": 6.059593057632446,
+      "epoch": 0.11984671964611784,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0003150759403094473,
+      "loss": 5.7577,
+      "mean_token_accuracy": 0.14673785269260406,
+      "num_tokens": 4164077.0,
+      "step": 2330
+    },
+    {
+      "entropy": 6.11279263496399,
+      "epoch": 0.12010390144793355,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00031391613722619587,
+      "loss": 5.846,
+      "mean_token_accuracy": 0.1490771532058716,
+      "num_tokens": 4172768.0,
+      "step": 2335
+    },
+    {
+      "entropy": 6.129426527023315,
+      "epoch": 0.12036108324974924,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.000312755267237384,
+      "loss": 5.8131,
+      "mean_token_accuracy": 0.1410787008702755,
+      "num_tokens": 4183164.0,
+      "step": 2340
+    },
+    {
+      "entropy": 6.10960693359375,
+      "epoch": 0.12061826505156495,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003115933621688488,
+      "loss": 5.8081,
+      "mean_token_accuracy": 0.14593723863363267,
+      "num_tokens": 4192396.0,
+      "step": 2345
+    },
+    {
+      "entropy": 6.070228147506714,
+      "epoch": 0.12087544685338066,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00031043045387480487,
+      "loss": 5.8152,
+      "mean_token_accuracy": 0.15205326676368713,
+      "num_tokens": 4200895.0,
+      "step": 2350
+    },
+    {
+      "entropy": 6.134779262542724,
+      "epoch": 0.12113262865519636,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0003092665742369703,
+      "loss": 5.8362,
+      "mean_token_accuracy": 0.14449836313724518,
+      "num_tokens": 4209448.0,
+      "step": 2355
+    },
+    {
+      "entropy": 6.061989641189575,
+      "epoch": 0.12138981045701207,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00030810175516369343,
+      "loss": 5.8572,
+      "mean_token_accuracy": 0.14029947221279143,
+      "num_tokens": 4218861.0,
+      "step": 2360
+    },
+    {
+      "entropy": 6.108126449584961,
+      "epoch": 0.12164699225882776,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0003069360285890775,
+      "loss": 5.7775,
+      "mean_token_accuracy": 0.14581253603100777,
+      "num_tokens": 4228123.0,
+      "step": 2365
+    },
+    {
+      "entropy": 6.158972692489624,
+      "epoch": 0.12190417406064347,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00030576942647210547,
+      "loss": 5.8698,
+      "mean_token_accuracy": 0.1389954224228859,
+      "num_tokens": 4237430.0,
+      "step": 2370
+    },
+    {
+      "entropy": 6.108787727355957,
+      "epoch": 0.12216135586245917,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00030460198079576355,
+      "loss": 5.7897,
+      "mean_token_accuracy": 0.14067094400525093,
+      "num_tokens": 4246106.0,
+      "step": 2375
+    },
+    {
+      "entropy": 6.181547927856445,
+      "epoch": 0.12241853766427488,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0003034337235661648,
+      "loss": 5.8474,
+      "mean_token_accuracy": 0.15081686154007912,
+      "num_tokens": 4255087.0,
+      "step": 2380
+    },
+    {
+      "entropy": 6.1383575916290285,
+      "epoch": 0.12267571946609059,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0003022646868116714,
+      "loss": 5.8509,
+      "mean_token_accuracy": 0.14444693699479103,
+      "num_tokens": 4263413.0,
+      "step": 2385
+    },
+    {
+      "entropy": 6.136625051498413,
+      "epoch": 0.12293290126790628,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.0003010949025820163,
+      "loss": 5.8514,
+      "mean_token_accuracy": 0.1409892752766609,
+      "num_tokens": 4272595.0,
+      "step": 2390
+    },
+    {
+      "entropy": 6.147713708877563,
+      "epoch": 0.12319008306972198,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0002999244029474252,
+      "loss": 5.791,
+      "mean_token_accuracy": 0.14426918625831603,
+      "num_tokens": 4281576.0,
+      "step": 2395
+    },
+    {
+      "entropy": 6.0478291511535645,
+      "epoch": 0.12344726487153769,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00029875321999773684,
+      "loss": 5.6831,
+      "mean_token_accuracy": 0.1507292740046978,
+      "num_tokens": 4289698.0,
+      "step": 2400
+    },
+    {
+      "entropy": 6.066676187515259,
+      "epoch": 0.1237044466733534,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00029758138584152333,
+      "loss": 5.8092,
+      "mean_token_accuracy": 0.13970909789204597,
+      "num_tokens": 4299010.0,
+      "step": 2405
+    },
+    {
+      "entropy": 6.06015453338623,
+      "epoch": 0.1239616284751691,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0002964089326052102,
+      "loss": 5.7582,
+      "mean_token_accuracy": 0.1479499265551567,
+      "num_tokens": 4307870.0,
+      "step": 2410
+    },
+    {
+      "entropy": 6.090758895874023,
+      "epoch": 0.1242188102769848,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0002952358924321949,
+      "loss": 5.8444,
+      "mean_token_accuracy": 0.1424449399113655,
+      "num_tokens": 4316599.0,
+      "step": 2415
+    },
+    {
+      "entropy": 6.045522451400757,
+      "epoch": 0.1244759920788005,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00029406229748196657,
+      "loss": 5.6963,
+      "mean_token_accuracy": 0.15066781789064407,
+      "num_tokens": 4326065.0,
+      "step": 2420
+    },
+    {
+      "entropy": 5.994427728652954,
+      "epoch": 0.12473317388061621,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0002928881799292235,
+      "loss": 5.7029,
+      "mean_token_accuracy": 0.14825093299150466,
+      "num_tokens": 4335225.0,
+      "step": 2425
+    },
+    {
+      "entropy": 6.015228986740112,
+      "epoch": 0.12499035568243191,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00029171357196299154,
+      "loss": 5.6892,
+      "mean_token_accuracy": 0.14861446171998977,
+      "num_tokens": 4344634.0,
+      "step": 2430
+    },
+    {
+      "entropy": 6.074378967285156,
+      "epoch": 0.1252475374842476,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0002905385057857414,
+      "loss": 5.6841,
+      "mean_token_accuracy": 0.15366790145635606,
+      "num_tokens": 4352815.0,
+      "step": 2435
+    },
+    {
+      "entropy": 5.976829957962036,
+      "epoch": 0.12550471928606333,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0002893630136125058,
+      "loss": 5.6631,
+      "mean_token_accuracy": 0.1512479767203331,
+      "num_tokens": 4361217.0,
+      "step": 2440
+    },
+    {
+      "entropy": 6.115370368957519,
+      "epoch": 0.12576190108787902,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0002881871276699967,
+      "loss": 5.8411,
+      "mean_token_accuracy": 0.14890666753053666,
+      "num_tokens": 4369963.0,
+      "step": 2445
+    },
+    {
+      "entropy": 6.115329885482788,
+      "epoch": 0.12601908288969474,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.00028701088019572114,
+      "loss": 5.9072,
+      "mean_token_accuracy": 0.13943761885166167,
+      "num_tokens": 4379146.0,
+      "step": 2450
+    },
+    {
+      "entropy": 6.071151351928711,
+      "epoch": 0.12627626469151043,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0002858343034370977,
+      "loss": 5.7355,
+      "mean_token_accuracy": 0.14560586810112,
+      "num_tokens": 4388240.0,
+      "step": 2455
+    },
+    {
+      "entropy": 6.015738677978516,
+      "epoch": 0.12653344649332612,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00028465742965057267,
+      "loss": 5.6636,
+      "mean_token_accuracy": 0.15043488889932632,
+      "num_tokens": 4396460.0,
+      "step": 2460
+    },
+    {
+      "entropy": 6.124741458892823,
+      "epoch": 0.12679062829514184,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00028348029110073533,
+      "loss": 5.8333,
+      "mean_token_accuracy": 0.14380111843347548,
+      "num_tokens": 4404559.0,
+      "step": 2465
+    },
+    {
+      "entropy": 6.0465250492095945,
+      "epoch": 0.12704781009695754,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00028230292005943365,
+      "loss": 5.7417,
+      "mean_token_accuracy": 0.14473437517881393,
+      "num_tokens": 4413795.0,
+      "step": 2470
+    },
+    {
+      "entropy": 6.124128723144532,
+      "epoch": 0.12730499189877326,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00028112534880488945,
+      "loss": 5.8881,
+      "mean_token_accuracy": 0.14171595126390457,
+      "num_tokens": 4422724.0,
+      "step": 2475
+    },
+    {
+      "entropy": 6.177429819107056,
+      "epoch": 0.12756217370058895,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0002799476096208137,
+      "loss": 5.7924,
+      "mean_token_accuracy": 0.14749524891376495,
+      "num_tokens": 4432454.0,
+      "step": 2480
+    },
+    {
+      "entropy": 5.974297523498535,
+      "epoch": 0.12781935550240464,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00027876973479552087,
+      "loss": 5.7018,
+      "mean_token_accuracy": 0.15010210499167442,
+      "num_tokens": 4441004.0,
+      "step": 2485
+    },
+    {
+      "entropy": 6.044048643112182,
+      "epoch": 0.12807653730422036,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.00027759175662104424,
+      "loss": 5.7713,
+      "mean_token_accuracy": 0.1527440369129181,
+      "num_tokens": 4450077.0,
+      "step": 2490
+    },
+    {
+      "entropy": 6.111342573165894,
+      "epoch": 0.12833371910603605,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0002764137073922508,
+      "loss": 5.7829,
+      "mean_token_accuracy": 0.14307816997170447,
+      "num_tokens": 4458821.0,
+      "step": 2495
+    },
+    {
+      "entropy": 6.062967014312744,
+      "epoch": 0.12859090090785177,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00027523561940595505,
+      "loss": 5.6736,
+      "mean_token_accuracy": 0.1584312215447426,
+      "num_tokens": 4466806.0,
+      "step": 2500
+    },
+    {
+      "entropy": 6.015162229537964,
+      "epoch": 0.12884808270966747,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0002740575249600342,
+      "loss": 5.747,
+      "mean_token_accuracy": 0.14772430807352066,
+      "num_tokens": 4475025.0,
+      "step": 2505
+    },
+    {
+      "entropy": 6.172916507720947,
+      "epoch": 0.12910526451148316,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00027287945635254263,
+      "loss": 5.7744,
+      "mean_token_accuracy": 0.1500643014907837,
+      "num_tokens": 4483790.0,
+      "step": 2510
+    },
+    {
+      "entropy": 6.092405891418457,
+      "epoch": 0.12936244631329888,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00027170144588082635,
+      "loss": 5.762,
+      "mean_token_accuracy": 0.15136242061853408,
+      "num_tokens": 4493367.0,
+      "step": 2515
+    },
+    {
+      "entropy": 6.146335029602051,
+      "epoch": 0.12961962811511457,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00027052352584063763,
+      "loss": 5.8209,
+      "mean_token_accuracy": 0.14450475126504897,
+      "num_tokens": 4502043.0,
+      "step": 2520
+    },
+    {
+      "entropy": 6.095106935501098,
+      "epoch": 0.1298768099169303,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00026934572852524907,
+      "loss": 5.7306,
+      "mean_token_accuracy": 0.14429289028048514,
+      "num_tokens": 4510459.0,
+      "step": 2525
+    },
+    {
+      "entropy": 6.059567832946778,
+      "epoch": 0.13013399171874598,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00026816808622456937,
+      "loss": 5.8032,
+      "mean_token_accuracy": 0.14666858240962027,
+      "num_tokens": 4519462.0,
+      "step": 2530
+    },
+    {
+      "entropy": 6.07892336845398,
+      "epoch": 0.13039117352056168,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0002669906312242569,
+      "loss": 5.8342,
+      "mean_token_accuracy": 0.14680682569742204,
+      "num_tokens": 4528651.0,
+      "step": 2535
+    },
+    {
+      "entropy": 6.166332626342774,
+      "epoch": 0.1306483553223774,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00026581339580483525,
+      "loss": 5.7735,
+      "mean_token_accuracy": 0.14615342170000076,
+      "num_tokens": 4537415.0,
+      "step": 2540
+    },
+    {
+      "entropy": 6.106481647491455,
+      "epoch": 0.1309055371241931,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.0002646364122408082,
+      "loss": 5.8251,
+      "mean_token_accuracy": 0.14407150372862815,
+      "num_tokens": 4547175.0,
+      "step": 2545
+    },
+    {
+      "entropy": 6.118671560287476,
+      "epoch": 0.1311627189260088,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0002634597127997749,
+      "loss": 5.8591,
+      "mean_token_accuracy": 0.1475327268242836,
+      "num_tokens": 4556028.0,
+      "step": 2550
+    },
+    {
+      "entropy": 6.080974769592285,
+      "epoch": 0.1314199007278245,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0002622833297415445,
+      "loss": 5.8042,
+      "mean_token_accuracy": 0.14328806847333908,
+      "num_tokens": 4564769.0,
+      "step": 2555
+    },
+    {
+      "entropy": 6.13230185508728,
+      "epoch": 0.1316770825296402,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0002611072953172531,
+      "loss": 5.7484,
+      "mean_token_accuracy": 0.14350520372390746,
+      "num_tokens": 4573654.0,
+      "step": 2560
+    },
+    {
+      "entropy": 6.029941511154175,
+      "epoch": 0.1319342643314559,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00025993164176847845,
+      "loss": 5.7879,
+      "mean_token_accuracy": 0.1487342558801174,
+      "num_tokens": 4583916.0,
+      "step": 2565
+    },
+    {
+      "entropy": 5.982626152038574,
+      "epoch": 0.1321914461332716,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0002587564013263564,
+      "loss": 5.6726,
+      "mean_token_accuracy": 0.15445442348718644,
+      "num_tokens": 4592281.0,
+      "step": 2570
+    },
+    {
+      "entropy": 6.086978912353516,
+      "epoch": 0.13244862793508733,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0002575816062106974,
+      "loss": 5.8636,
+      "mean_token_accuracy": 0.1401931382715702,
+      "num_tokens": 4601838.0,
+      "step": 2575
+    },
+    {
+      "entropy": 6.107604694366455,
+      "epoch": 0.13270580973690302,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00025640728862910293,
+      "loss": 5.7927,
+      "mean_token_accuracy": 0.14364690631628035,
+      "num_tokens": 4610533.0,
+      "step": 2580
+    },
+    {
+      "entropy": 6.074134016036988,
+      "epoch": 0.1329629915387187,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00025523348077608285,
+      "loss": 5.7005,
+      "mean_token_accuracy": 0.15348859429359435,
+      "num_tokens": 4618325.0,
+      "step": 2585
+    },
+    {
+      "entropy": 6.045175695419312,
+      "epoch": 0.13322017334053443,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00025406021483217225,
+      "loss": 5.8188,
+      "mean_token_accuracy": 0.14634289145469664,
+      "num_tokens": 4627594.0,
+      "step": 2590
+    },
+    {
+      "entropy": 6.118519401550293,
+      "epoch": 0.13347735514235012,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00025288752296304963,
+      "loss": 5.7381,
+      "mean_token_accuracy": 0.1461601436138153,
+      "num_tokens": 4636701.0,
+      "step": 2595
+    },
+    {
+      "entropy": 6.042619895935059,
+      "epoch": 0.13373453694416584,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.000251715437318655,
+      "loss": 5.73,
+      "mean_token_accuracy": 0.15269380956888198,
+      "num_tokens": 4645351.0,
+      "step": 2600
+    },
+    {
+      "entropy": 6.1090494155883786,
+      "epoch": 0.13399171874598154,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0002505439900323084,
+      "loss": 5.8194,
+      "mean_token_accuracy": 0.14812767803668975,
+      "num_tokens": 4654550.0,
+      "step": 2605
+    },
+    {
+      "entropy": 6.076578140258789,
+      "epoch": 0.13424890054779723,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00024937321321982894,
+      "loss": 5.7404,
+      "mean_token_accuracy": 0.14685189351439476,
+      "num_tokens": 4663603.0,
+      "step": 2610
+    },
+    {
+      "entropy": 6.114797639846802,
+      "epoch": 0.13450608234961295,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00024820313897865433,
+      "loss": 5.8668,
+      "mean_token_accuracy": 0.14367469623684884,
+      "num_tokens": 4672140.0,
+      "step": 2615
+    },
+    {
+      "entropy": 6.078299140930175,
+      "epoch": 0.13476326415142864,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00024703379938696105,
+      "loss": 5.8506,
+      "mean_token_accuracy": 0.1418767161667347,
+      "num_tokens": 4681550.0,
+      "step": 2620
+    },
+    {
+      "entropy": 6.064907884597778,
+      "epoch": 0.13502044595324436,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00024586522650278447,
+      "loss": 5.7066,
+      "mean_token_accuracy": 0.1540905401110649,
+      "num_tokens": 4689805.0,
+      "step": 2625
+    },
+    {
+      "entropy": 6.06515941619873,
+      "epoch": 0.13527762775506005,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00024469745236314064,
+      "loss": 5.6682,
+      "mean_token_accuracy": 0.15574965476989747,
+      "num_tokens": 4698201.0,
+      "step": 2630
+    },
+    {
+      "entropy": 6.10120153427124,
+      "epoch": 0.13553480955687575,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00024353050898314767,
+      "loss": 5.759,
+      "mean_token_accuracy": 0.14834545105695723,
+      "num_tokens": 4706625.0,
+      "step": 2635
+    },
+    {
+      "entropy": 6.046721982955932,
+      "epoch": 0.13579199135869147,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00024236442835514743,
+      "loss": 5.6776,
+      "mean_token_accuracy": 0.15349911004304886,
+      "num_tokens": 4715487.0,
+      "step": 2640
+    },
+    {
+      "entropy": 6.128892707824707,
+      "epoch": 0.13604917316050716,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00024119924244782965,
+      "loss": 5.8295,
+      "mean_token_accuracy": 0.13918910250067712,
+      "num_tokens": 4724851.0,
+      "step": 2645
+    },
+    {
+      "entropy": 6.108962345123291,
+      "epoch": 0.13630635496232288,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00024003498320535462,
+      "loss": 5.8328,
+      "mean_token_accuracy": 0.14703123271465302,
+      "num_tokens": 4732863.0,
+      "step": 2650
+    },
+    {
+      "entropy": 6.194251680374146,
+      "epoch": 0.13656353676413857,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00023887168254647727,
+      "loss": 5.9352,
+      "mean_token_accuracy": 0.13815246075391768,
+      "num_tokens": 4743023.0,
+      "step": 2655
+    },
+    {
+      "entropy": 6.134444332122802,
+      "epoch": 0.13682071856595426,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00023770937236367308,
+      "loss": 5.8393,
+      "mean_token_accuracy": 0.14146638661623,
+      "num_tokens": 4752009.0,
+      "step": 2660
+    },
+    {
+      "entropy": 6.196498537063599,
+      "epoch": 0.13707790036776998,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.00023654808452226278,
+      "loss": 5.7816,
+      "mean_token_accuracy": 0.14370832592248917,
+      "num_tokens": 4761731.0,
+      "step": 2665
+    },
+    {
+      "entropy": 6.083220052719116,
+      "epoch": 0.13733508216958568,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00023538785085953912,
+      "loss": 5.7217,
+      "mean_token_accuracy": 0.14778869450092316,
+      "num_tokens": 4769991.0,
+      "step": 2670
+    },
+    {
+      "entropy": 6.005928564071655,
+      "epoch": 0.1375922639714014,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00023422870318389404,
+      "loss": 5.6433,
+      "mean_token_accuracy": 0.14906827211380005,
+      "num_tokens": 4778481.0,
+      "step": 2675
+    },
+    {
+      "entropy": 6.045457029342652,
+      "epoch": 0.1378494457732171,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0002330706732739468,
+      "loss": 5.7499,
+      "mean_token_accuracy": 0.14508197754621505,
+      "num_tokens": 4787974.0,
+      "step": 2680
+    },
+    {
+      "entropy": 6.096870279312133,
+      "epoch": 0.13810662757503278,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00023191379287767211,
+      "loss": 5.7255,
+      "mean_token_accuracy": 0.15906407088041305,
+      "num_tokens": 4796296.0,
+      "step": 2685
+    },
+    {
+      "entropy": 5.97481255531311,
+      "epoch": 0.1383638093768485,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0002307580937115305,
+      "loss": 5.6343,
+      "mean_token_accuracy": 0.15172418504953383,
+      "num_tokens": 4805338.0,
+      "step": 2690
+    },
+    {
+      "entropy": 6.05340428352356,
+      "epoch": 0.1386209911786642,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00022960360745959846,
+      "loss": 5.7981,
+      "mean_token_accuracy": 0.14350073263049126,
+      "num_tokens": 4814859.0,
+      "step": 2695
+    },
+    {
+      "entropy": 6.117996311187744,
+      "epoch": 0.1388781729804799,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00022845036577269972,
+      "loss": 5.7159,
+      "mean_token_accuracy": 0.1534138098359108,
+      "num_tokens": 4822932.0,
+      "step": 2700
+    },
+    {
+      "entropy": 6.12151985168457,
+      "epoch": 0.1391353547822956,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00022729840026753777,
+      "loss": 5.871,
+      "mean_token_accuracy": 0.1421603888273239,
+      "num_tokens": 4831771.0,
+      "step": 2705
+    },
+    {
+      "entropy": 6.000537586212158,
+      "epoch": 0.1393925365841113,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0002261477425258287,
+      "loss": 5.6885,
+      "mean_token_accuracy": 0.15709090158343314,
+      "num_tokens": 4840470.0,
+      "step": 2710
+    },
+    {
+      "entropy": 6.1092977046966555,
+      "epoch": 0.13964971838592702,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0002249984240934358,
+      "loss": 5.8418,
+      "mean_token_accuracy": 0.14331340342760085,
+      "num_tokens": 4848989.0,
+      "step": 2715
+    },
+    {
+      "entropy": 6.073454236984253,
+      "epoch": 0.1399069001877427,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00022385047647950464,
+      "loss": 5.6271,
+      "mean_token_accuracy": 0.15826509073376654,
+      "num_tokens": 4858595.0,
+      "step": 2720
+    },
+    {
+      "entropy": 6.021537399291992,
+      "epoch": 0.14016408198955843,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0002227039311555986,
+      "loss": 5.7031,
+      "mean_token_accuracy": 0.14996479600667953,
+      "num_tokens": 4867744.0,
+      "step": 2725
+    },
+    {
+      "entropy": 6.065329742431641,
+      "epoch": 0.14042126379137412,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0002215588195548372,
+      "loss": 5.7416,
+      "mean_token_accuracy": 0.14959547519683838,
+      "num_tokens": 4876648.0,
+      "step": 2730
+    },
+    {
+      "entropy": 6.054684972763061,
+      "epoch": 0.14067844559318982,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00022041517307103337,
+      "loss": 5.676,
+      "mean_token_accuracy": 0.152857705950737,
+      "num_tokens": 4884794.0,
+      "step": 2735
+    },
+    {
+      "entropy": 6.127657175064087,
+      "epoch": 0.14093562739500554,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0002192730230578331,
+      "loss": 5.8685,
+      "mean_token_accuracy": 0.14753987938165664,
+      "num_tokens": 4894425.0,
+      "step": 2740
+    },
+    {
+      "entropy": 6.0618555545806885,
+      "epoch": 0.14119280919682123,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0002181324008278559,
+      "loss": 5.742,
+      "mean_token_accuracy": 0.15118976533412934,
+      "num_tokens": 4903135.0,
+      "step": 2745
+    },
+    {
+      "entropy": 6.038413095474243,
+      "epoch": 0.14144999099863695,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00021699333765183655,
+      "loss": 5.6718,
+      "mean_token_accuracy": 0.15931579023599624,
+      "num_tokens": 4911777.0,
+      "step": 2750
+    },
+    {
+      "entropy": 6.064218330383301,
+      "epoch": 0.14170717280045264,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0002158558647577673,
+      "loss": 5.7467,
+      "mean_token_accuracy": 0.15042026042938234,
+      "num_tokens": 4920350.0,
+      "step": 2755
+    },
+    {
+      "entropy": 6.01921124458313,
+      "epoch": 0.14196435460226833,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00021472001333004215,
+      "loss": 5.7043,
+      "mean_token_accuracy": 0.15539543628692626,
+      "num_tokens": 4928917.0,
+      "step": 2760
+    },
+    {
+      "entropy": 6.121454429626465,
+      "epoch": 0.14222153640408405,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00021358581450860186,
+      "loss": 5.877,
+      "mean_token_accuracy": 0.14617596417665482,
+      "num_tokens": 4937317.0,
+      "step": 2765
+    },
+    {
+      "entropy": 6.176073312759399,
+      "epoch": 0.14247871820589975,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0002124532993880799,
+      "loss": 5.8984,
+      "mean_token_accuracy": 0.14659789949655533,
+      "num_tokens": 4946679.0,
+      "step": 2770
+    },
+    {
+      "entropy": 6.103835725784302,
+      "epoch": 0.14273590000771547,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00021132249901695044,
+      "loss": 5.7354,
+      "mean_token_accuracy": 0.14755650311708451,
+      "num_tokens": 4955617.0,
+      "step": 2775
+    },
+    {
+      "entropy": 6.038599872589112,
+      "epoch": 0.14299308180953116,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00021019344439667705,
+      "loss": 5.6085,
+      "mean_token_accuracy": 0.15284328907728195,
+      "num_tokens": 4964376.0,
+      "step": 2780
+    },
+    {
+      "entropy": 6.036679029464722,
+      "epoch": 0.14325026361134685,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00020906616648086213,
+      "loss": 5.8006,
+      "mean_token_accuracy": 0.14090735837817192,
+      "num_tokens": 4974308.0,
+      "step": 2785
+    },
+    {
+      "entropy": 6.100146913528443,
+      "epoch": 0.14350744541316257,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00020794069617439942,
+      "loss": 5.7213,
+      "mean_token_accuracy": 0.14757935106754302,
+      "num_tokens": 4983922.0,
+      "step": 2790
+    },
+    {
+      "entropy": 6.111583518981933,
+      "epoch": 0.14376462721497826,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00020681706433262593,
+      "loss": 5.7919,
+      "mean_token_accuracy": 0.14720299541950227,
+      "num_tokens": 4992417.0,
+      "step": 2795
+    },
+    {
+      "entropy": 6.074492073059082,
+      "epoch": 0.14402180901679398,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00020569530176047602,
+      "loss": 5.6844,
+      "mean_token_accuracy": 0.15296519845724105,
+      "num_tokens": 5000782.0,
+      "step": 2800
+    },
+    {
+      "entropy": 6.0138218879699705,
+      "epoch": 0.14427899081860968,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0002045754392116374,
+      "loss": 5.6534,
+      "mean_token_accuracy": 0.1558382123708725,
+      "num_tokens": 5009515.0,
+      "step": 2805
+    },
+    {
+      "entropy": 6.1086820602417,
+      "epoch": 0.14453617262042537,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00020345750738770757,
+      "loss": 5.78,
+      "mean_token_accuracy": 0.1543252319097519,
+      "num_tokens": 5019326.0,
+      "step": 2810
+    },
+    {
+      "entropy": 6.052635383605957,
+      "epoch": 0.1447933544222411,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00020234153693735214,
+      "loss": 5.6338,
+      "mean_token_accuracy": 0.153486018627882,
+      "num_tokens": 5028355.0,
+      "step": 2815
+    },
+    {
+      "entropy": 6.0126776695251465,
+      "epoch": 0.14505053622405678,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0002012275584554647,
+      "loss": 5.7085,
+      "mean_token_accuracy": 0.14389130845665932,
+      "num_tokens": 5037558.0,
+      "step": 2820
+    },
+    {
+      "entropy": 6.078545379638672,
+      "epoch": 0.1453077180258725,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00020011560248232803,
+      "loss": 5.7932,
+      "mean_token_accuracy": 0.1469465374946594,
+      "num_tokens": 5046154.0,
+      "step": 2825
+    },
+    {
+      "entropy": 6.114815616607666,
+      "epoch": 0.1455648998276882,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00019900569950277692,
+      "loss": 5.7511,
+      "mean_token_accuracy": 0.14542998522520065,
+      "num_tokens": 5054483.0,
+      "step": 2830
+    },
+    {
+      "entropy": 6.104913187026978,
+      "epoch": 0.14582208162950389,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00019789787994536228,
+      "loss": 5.7582,
+      "mean_token_accuracy": 0.15545508861541749,
+      "num_tokens": 5063131.0,
+      "step": 2835
+    },
+    {
+      "entropy": 6.084230899810791,
+      "epoch": 0.1460792634313196,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00019679217418151667,
+      "loss": 5.8028,
+      "mean_token_accuracy": 0.1484840542078018,
+      "num_tokens": 5072165.0,
+      "step": 2840
+    },
+    {
+      "entropy": 6.020099258422851,
+      "epoch": 0.1463364452331353,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00019568861252472236,
+      "loss": 5.6214,
+      "mean_token_accuracy": 0.15922571569681168,
+      "num_tokens": 5080203.0,
+      "step": 2845
+    },
+    {
+      "entropy": 6.0180675983428955,
+      "epoch": 0.14659362703495102,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00019458722522967952,
+      "loss": 5.6505,
+      "mean_token_accuracy": 0.15302405059337615,
+      "num_tokens": 5088434.0,
+      "step": 2850
+    },
+    {
+      "entropy": 6.0362659931182865,
+      "epoch": 0.1468508088367667,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00019348804249147723,
+      "loss": 5.7373,
+      "mean_token_accuracy": 0.15490390062332154,
+      "num_tokens": 5096528.0,
+      "step": 2855
+    },
+    {
+      "entropy": 6.023699522018433,
+      "epoch": 0.1471079906385824,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0001923910944447655,
+      "loss": 5.7418,
+      "mean_token_accuracy": 0.15109253525733948,
+      "num_tokens": 5105684.0,
+      "step": 2860
+    },
+    {
+      "entropy": 6.057016038894654,
+      "epoch": 0.14736517244039812,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00019129641116292928,
+      "loss": 5.6806,
+      "mean_token_accuracy": 0.15375835597515106,
+      "num_tokens": 5113761.0,
+      "step": 2865
+    },
+    {
+      "entropy": 6.084755897521973,
+      "epoch": 0.14762235424221382,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00019020402265726343,
+      "loss": 5.7504,
+      "mean_token_accuracy": 0.15073051154613495,
+      "num_tokens": 5122186.0,
+      "step": 2870
+    },
+    {
+      "entropy": 6.134022569656372,
+      "epoch": 0.14787953604402954,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0001891139588761509,
+      "loss": 5.8892,
+      "mean_token_accuracy": 0.14310402274131775,
+      "num_tokens": 5132312.0,
+      "step": 2875
+    },
+    {
+      "entropy": 6.000723934173584,
+      "epoch": 0.14813671784584523,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00018802624970424076,
+      "loss": 5.619,
+      "mean_token_accuracy": 0.15236654356122017,
+      "num_tokens": 5140811.0,
+      "step": 2880
+    },
+    {
+      "entropy": 6.063713455200196,
+      "epoch": 0.14839389964766092,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00018694092496162945,
+      "loss": 5.6289,
+      "mean_token_accuracy": 0.15638696998357773,
+      "num_tokens": 5148382.0,
+      "step": 2885
+    },
+    {
+      "entropy": 6.060241270065307,
+      "epoch": 0.14865108144947664,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00018585801440304306,
+      "loss": 5.6513,
+      "mean_token_accuracy": 0.16285014897584915,
+      "num_tokens": 5157551.0,
+      "step": 2890
+    },
+    {
+      "entropy": 6.027216815948487,
+      "epoch": 0.14890826325129233,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00018477754771702165,
+      "loss": 5.6163,
+      "mean_token_accuracy": 0.15568110346794128,
+      "num_tokens": 5165423.0,
+      "step": 2895
+    },
+    {
+      "entropy": 6.013796472549439,
+      "epoch": 0.14916544505310805,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00018369955452510506,
+      "loss": 5.7244,
+      "mean_token_accuracy": 0.15133277401328088,
+      "num_tokens": 5174290.0,
+      "step": 2900
+    },
+    {
+      "entropy": 6.072081470489502,
+      "epoch": 0.14942262685492375,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0001826240643810212,
+      "loss": 5.73,
+      "mean_token_accuracy": 0.1497554212808609,
+      "num_tokens": 5183516.0,
+      "step": 2905
+    },
+    {
+      "entropy": 6.041428518295288,
+      "epoch": 0.14967980865673944,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0001815511067698758,
+      "loss": 5.6703,
+      "mean_token_accuracy": 0.151530821621418,
+      "num_tokens": 5192140.0,
+      "step": 2910
+    },
+    {
+      "entropy": 6.060357141494751,
+      "epoch": 0.14993699045855516,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0001804807111073436,
+      "loss": 5.7481,
+      "mean_token_accuracy": 0.15596057176589967,
+      "num_tokens": 5201062.0,
+      "step": 2915
+    },
+    {
+      "entropy": 6.018105745315552,
+      "epoch": 0.15019417226037085,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0001794129067388625,
+      "loss": 5.5261,
+      "mean_token_accuracy": 0.16392314583063125,
+      "num_tokens": 5209850.0,
+      "step": 2920
+    },
+    {
+      "entropy": 5.980852556228638,
+      "epoch": 0.15045135406218657,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00017834772293882868,
+      "loss": 5.6941,
+      "mean_token_accuracy": 0.15812414288520812,
+      "num_tokens": 5218309.0,
+      "step": 2925
+    },
+    {
+      "entropy": 6.012538290023803,
+      "epoch": 0.15070853586400226,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000177285188909794,
+      "loss": 5.7583,
+      "mean_token_accuracy": 0.15520099699497222,
+      "num_tokens": 5226475.0,
+      "step": 2930
+    },
+    {
+      "entropy": 6.05089168548584,
+      "epoch": 0.15096571766581796,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0001762253337816656,
+      "loss": 5.5913,
+      "mean_token_accuracy": 0.16246795505285264,
+      "num_tokens": 5235665.0,
+      "step": 2935
+    },
+    {
+      "entropy": 6.068499660491943,
+      "epoch": 0.15122289946763368,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00017516818661090738,
+      "loss": 5.6451,
+      "mean_token_accuracy": 0.15650007724761963,
+      "num_tokens": 5245129.0,
+      "step": 2940
+    },
+    {
+      "entropy": 5.932179880142212,
+      "epoch": 0.15148008126944937,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0001741137763797428,
+      "loss": 5.6156,
+      "mean_token_accuracy": 0.15633189380168916,
+      "num_tokens": 5253374.0,
+      "step": 2945
+    },
+    {
+      "entropy": 6.057919502258301,
+      "epoch": 0.1517372630712651,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00017306213199536115,
+      "loss": 5.7014,
+      "mean_token_accuracy": 0.15033402144908906,
+      "num_tokens": 5262186.0,
+      "step": 2950
+    },
+    {
+      "entropy": 6.089097738265991,
+      "epoch": 0.15199444487308078,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0001720132822891243,
+      "loss": 5.7236,
+      "mean_token_accuracy": 0.15335725396871566,
+      "num_tokens": 5270708.0,
+      "step": 2955
+    },
+    {
+      "entropy": 6.062357234954834,
+      "epoch": 0.15225162667489647,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0001709672560157769,
+      "loss": 5.6618,
+      "mean_token_accuracy": 0.1679921880364418,
+      "num_tokens": 5279170.0,
+      "step": 2960
+    },
+    {
+      "entropy": 5.953275728225708,
+      "epoch": 0.1525088084767122,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00016992408185265758,
+      "loss": 5.6085,
+      "mean_token_accuracy": 0.16066092550754546,
+      "num_tokens": 5287278.0,
+      "step": 2965
+    },
+    {
+      "entropy": 6.021520376205444,
+      "epoch": 0.15276599027852789,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00016888378839891298,
+      "loss": 5.7349,
+      "mean_token_accuracy": 0.15320995151996614,
+      "num_tokens": 5296333.0,
+      "step": 2970
+    },
+    {
+      "entropy": 6.035226774215698,
+      "epoch": 0.1530231720803436,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0001678464041747137,
+      "loss": 5.7335,
+      "mean_token_accuracy": 0.1488783285021782,
+      "num_tokens": 5306121.0,
+      "step": 2975
+    },
+    {
+      "entropy": 6.089285039901734,
+      "epoch": 0.1532803538821593,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00016681195762047223,
+      "loss": 5.8083,
+      "mean_token_accuracy": 0.14437277913093566,
+      "num_tokens": 5315569.0,
+      "step": 2980
+    },
+    {
+      "entropy": 6.052507734298706,
+      "epoch": 0.153537535683975,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00016578047709606337,
+      "loss": 5.6236,
+      "mean_token_accuracy": 0.15959543138742446,
+      "num_tokens": 5325185.0,
+      "step": 2985
+    },
+    {
+      "entropy": 6.066595077514648,
+      "epoch": 0.1537947174857907,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00016475199088004678,
+      "loss": 5.7174,
+      "mean_token_accuracy": 0.1542446307837963,
+      "num_tokens": 5334798.0,
+      "step": 2990
+    },
+    {
+      "entropy": 6.041294384002685,
+      "epoch": 0.1540518992876064,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00016372652716889163,
+      "loss": 5.6552,
+      "mean_token_accuracy": 0.15341484695672988,
+      "num_tokens": 5343994.0,
+      "step": 2995
+    },
+    {
+      "entropy": 5.968040609359742,
+      "epoch": 0.15430908108942212,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0001627041140762035,
+      "loss": 5.6459,
+      "mean_token_accuracy": 0.155483378469944,
+      "num_tokens": 5352357.0,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 4000,
+  "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": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1069056693043200.0,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}