diff --git "a/checkpoint-2000/trainer_state.json" "b/checkpoint-2000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-2000/trainer_state.json"
@@ -0,0 +1,4034 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.10287272072628141,
+  "eval_steps": 500,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 10.742102813720702,
+      "epoch": 0.0002571818018157035,
+      "grad_norm": 5.21875,
+      "learning_rate": 2e-06,
+      "loss": 10.7614,
+      "mean_token_accuracy": 8.620689623057842e-05,
+      "num_tokens": 9573.0,
+      "step": 5
+    },
+    {
+      "entropy": 10.742151069641114,
+      "epoch": 0.000514363603631407,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.5e-06,
+      "loss": 10.7583,
+      "mean_token_accuracy": 0.0,
+      "num_tokens": 18933.0,
+      "step": 10
+    },
+    {
+      "entropy": 10.742130088806153,
+      "epoch": 0.0007715454054471106,
+      "grad_norm": 5.21875,
+      "learning_rate": 7e-06,
+      "loss": 10.7448,
+      "mean_token_accuracy": 0.0,
+      "num_tokens": 29028.0,
+      "step": 15
+    },
+    {
+      "entropy": 10.742134284973144,
+      "epoch": 0.001028727207262814,
+      "grad_norm": 4.9375,
+      "learning_rate": 9.5e-06,
+      "loss": 10.6881,
+      "mean_token_accuracy": 0.00012970168609172106,
+      "num_tokens": 37987.0,
+      "step": 20
+    },
+    {
+      "entropy": 10.742049884796142,
+      "epoch": 0.0012859090090785177,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.2e-05,
+      "loss": 10.5852,
+      "mean_token_accuracy": 0.0029032987426035107,
+      "num_tokens": 47163.0,
+      "step": 25
+    },
+    {
+      "entropy": 10.741710948944093,
+      "epoch": 0.0015430908108942211,
+      "grad_norm": 4.25,
+      "learning_rate": 1.4500000000000002e-05,
+      "loss": 10.4809,
+      "mean_token_accuracy": 0.02479298785328865,
+      "num_tokens": 56096.0,
+      "step": 30
+    },
+    {
+      "entropy": 10.740671920776368,
+      "epoch": 0.0018002726127099246,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 10.3641,
+      "mean_token_accuracy": 0.03960379287600517,
+      "num_tokens": 65823.0,
+      "step": 35
+    },
+    {
+      "entropy": 10.737626838684083,
+      "epoch": 0.002057454414525628,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.95e-05,
+      "loss": 10.2147,
+      "mean_token_accuracy": 0.04078627172857523,
+      "num_tokens": 75005.0,
+      "step": 40
+    },
+    {
+      "entropy": 10.73199462890625,
+      "epoch": 0.0023146362163413317,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.2e-05,
+      "loss": 10.1015,
+      "mean_token_accuracy": 0.04431910365819931,
+      "num_tokens": 83340.0,
+      "step": 45
+    },
+    {
+      "entropy": 10.72521743774414,
+      "epoch": 0.0025718180181570354,
+      "grad_norm": 2.140625,
+      "learning_rate": 2.4500000000000003e-05,
+      "loss": 10.0212,
+      "mean_token_accuracy": 0.049253365769982335,
+      "num_tokens": 92004.0,
+      "step": 50
+    },
+    {
+      "entropy": 10.719530010223389,
+      "epoch": 0.0028289998199727386,
+      "grad_norm": 2.078125,
+      "learning_rate": 2.7e-05,
+      "loss": 9.9673,
+      "mean_token_accuracy": 0.04715999849140644,
+      "num_tokens": 101196.0,
+      "step": 55
+    },
+    {
+      "entropy": 10.713758850097657,
+      "epoch": 0.0030861816217884423,
+      "grad_norm": 2.078125,
+      "learning_rate": 2.95e-05,
+      "loss": 9.8887,
+      "mean_token_accuracy": 0.04930844120681286,
+      "num_tokens": 109931.0,
+      "step": 60
+    },
+    {
+      "entropy": 10.709882545471192,
+      "epoch": 0.003343363423604146,
+      "grad_norm": 2.0625,
+      "learning_rate": 3.2e-05,
+      "loss": 9.8563,
+      "mean_token_accuracy": 0.046213630586862564,
+      "num_tokens": 118450.0,
+      "step": 65
+    },
+    {
+      "entropy": 10.70816535949707,
+      "epoch": 0.003600545225419849,
+      "grad_norm": 1.96875,
+      "learning_rate": 3.4500000000000005e-05,
+      "loss": 9.7843,
+      "mean_token_accuracy": 0.04638329595327377,
+      "num_tokens": 127311.0,
+      "step": 70
+    },
+    {
+      "entropy": 10.705178928375243,
+      "epoch": 0.003857727027235553,
+      "grad_norm": 1.875,
+      "learning_rate": 3.7e-05,
+      "loss": 9.7153,
+      "mean_token_accuracy": 0.05029761902987957,
+      "num_tokens": 136529.0,
+      "step": 75
+    },
+    {
+      "entropy": 10.696865940093994,
+      "epoch": 0.004114908829051256,
+      "grad_norm": 1.828125,
+      "learning_rate": 3.95e-05,
+      "loss": 9.6593,
+      "mean_token_accuracy": 0.050337107852101326,
+      "num_tokens": 145754.0,
+      "step": 80
+    },
+    {
+      "entropy": 10.685219669342041,
+      "epoch": 0.00437209063086696,
+      "grad_norm": 1.8046875,
+      "learning_rate": 4.2000000000000004e-05,
+      "loss": 9.6217,
+      "mean_token_accuracy": 0.04981501027941704,
+      "num_tokens": 155229.0,
+      "step": 85
+    },
+    {
+      "entropy": 10.669653797149659,
+      "epoch": 0.004629272432682663,
+      "grad_norm": 1.8984375,
+      "learning_rate": 4.45e-05,
+      "loss": 9.4894,
+      "mean_token_accuracy": 0.06370054408907891,
+      "num_tokens": 163766.0,
+      "step": 90
+    },
+    {
+      "entropy": 10.649745845794678,
+      "epoch": 0.004886454234498367,
+      "grad_norm": 1.7265625,
+      "learning_rate": 4.7000000000000004e-05,
+      "loss": 9.4341,
+      "mean_token_accuracy": 0.05732584930956364,
+      "num_tokens": 172309.0,
+      "step": 95
+    },
+    {
+      "entropy": 10.618613624572754,
+      "epoch": 0.005143636036314071,
+      "grad_norm": 1.7421875,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 9.2855,
+      "mean_token_accuracy": 0.06327233165502548,
+      "num_tokens": 180242.0,
+      "step": 100
+    },
+    {
+      "entropy": 10.569914817810059,
+      "epoch": 0.0054008178381297735,
+      "grad_norm": 1.75,
+      "learning_rate": 5.2e-05,
+      "loss": 9.2324,
+      "mean_token_accuracy": 0.060527439787983896,
+      "num_tokens": 189044.0,
+      "step": 105
+    },
+    {
+      "entropy": 10.520080757141113,
+      "epoch": 0.005657999639945477,
+      "grad_norm": 1.5390625,
+      "learning_rate": 5.45e-05,
+      "loss": 9.1122,
+      "mean_token_accuracy": 0.06187737137079239,
+      "num_tokens": 197779.0,
+      "step": 110
+    },
+    {
+      "entropy": 10.468755054473878,
+      "epoch": 0.005915181441761181,
+      "grad_norm": 1.5859375,
+      "learning_rate": 5.7e-05,
+      "loss": 8.9561,
+      "mean_token_accuracy": 0.06379008032381535,
+      "num_tokens": 206025.0,
+      "step": 115
+    },
+    {
+      "entropy": 10.407035160064698,
+      "epoch": 0.0061723632435768845,
+      "grad_norm": 1.453125,
+      "learning_rate": 5.9499999999999996e-05,
+      "loss": 8.9663,
+      "mean_token_accuracy": 0.05617102049291134,
+      "num_tokens": 216620.0,
+      "step": 120
+    },
+    {
+      "entropy": 10.3653733253479,
+      "epoch": 0.006429545045392588,
+      "grad_norm": 1.375,
+      "learning_rate": 6.2e-05,
+      "loss": 8.7968,
+      "mean_token_accuracy": 0.059108608216047284,
+      "num_tokens": 225654.0,
+      "step": 125
+    },
+    {
+      "entropy": 10.301421070098877,
+      "epoch": 0.006686726847208292,
+      "grad_norm": 1.4609375,
+      "learning_rate": 6.450000000000001e-05,
+      "loss": 8.6036,
+      "mean_token_accuracy": 0.06164416745305061,
+      "num_tokens": 233967.0,
+      "step": 130
+    },
+    {
+      "entropy": 10.185540676116943,
+      "epoch": 0.006943908649023995,
+      "grad_norm": 1.4140625,
+      "learning_rate": 6.7e-05,
+      "loss": 8.5469,
+      "mean_token_accuracy": 0.06072752773761749,
+      "num_tokens": 241911.0,
+      "step": 135
+    },
+    {
+      "entropy": 10.0550705909729,
+      "epoch": 0.007201090450839698,
+      "grad_norm": 1.25,
+      "learning_rate": 6.950000000000001e-05,
+      "loss": 8.4323,
+      "mean_token_accuracy": 0.06439580023288727,
+      "num_tokens": 250771.0,
+      "step": 140
+    },
+    {
+      "entropy": 9.967238235473634,
+      "epoch": 0.007458272252655402,
+      "grad_norm": 1.1171875,
+      "learning_rate": 7.2e-05,
+      "loss": 8.3391,
+      "mean_token_accuracy": 0.06209088861942291,
+      "num_tokens": 260036.0,
+      "step": 145
+    },
+    {
+      "entropy": 9.832792282104492,
+      "epoch": 0.007715454054471106,
+      "grad_norm": 1.1875,
+      "learning_rate": 7.45e-05,
+      "loss": 8.2272,
+      "mean_token_accuracy": 0.06231529638171196,
+      "num_tokens": 268614.0,
+      "step": 150
+    },
+    {
+      "entropy": 9.642397022247314,
+      "epoch": 0.00797263585628681,
+      "grad_norm": 1.0546875,
+      "learning_rate": 7.7e-05,
+      "loss": 8.085,
+      "mean_token_accuracy": 0.06268828995525837,
+      "num_tokens": 278205.0,
+      "step": 155
+    },
+    {
+      "entropy": 9.471720123291016,
+      "epoch": 0.008229817658102512,
+      "grad_norm": 1.109375,
+      "learning_rate": 7.950000000000001e-05,
+      "loss": 8.0285,
+      "mean_token_accuracy": 0.054571619257330894,
+      "num_tokens": 287369.0,
+      "step": 160
+    },
+    {
+      "entropy": 9.277729606628418,
+      "epoch": 0.008486999459918217,
+      "grad_norm": 0.9765625,
+      "learning_rate": 8.2e-05,
+      "loss": 7.8389,
+      "mean_token_accuracy": 0.0658249743282795,
+      "num_tokens": 295469.0,
+      "step": 165
+    },
+    {
+      "entropy": 9.065653324127197,
+      "epoch": 0.00874418126173392,
+      "grad_norm": 0.875,
+      "learning_rate": 8.450000000000001e-05,
+      "loss": 7.8802,
+      "mean_token_accuracy": 0.05946259275078773,
+      "num_tokens": 304108.0,
+      "step": 170
+    },
+    {
+      "entropy": 8.9242995262146,
+      "epoch": 0.009001363063549624,
+      "grad_norm": 0.83203125,
+      "learning_rate": 8.7e-05,
+      "loss": 7.7549,
+      "mean_token_accuracy": 0.06102342195808887,
+      "num_tokens": 312797.0,
+      "step": 175
+    },
+    {
+      "entropy": 8.724368000030518,
+      "epoch": 0.009258544865365327,
+      "grad_norm": 0.89453125,
+      "learning_rate": 8.95e-05,
+      "loss": 7.726,
+      "mean_token_accuracy": 0.05993276461958885,
+      "num_tokens": 321947.0,
+      "step": 180
+    },
+    {
+      "entropy": 8.550490474700927,
+      "epoch": 0.00951572666718103,
+      "grad_norm": 0.859375,
+      "learning_rate": 9.2e-05,
+      "loss": 7.7268,
+      "mean_token_accuracy": 0.06107203662395477,
+      "num_tokens": 330881.0,
+      "step": 185
+    },
+    {
+      "entropy": 8.48734483718872,
+      "epoch": 0.009772908468996734,
+      "grad_norm": 0.8125,
+      "learning_rate": 9.45e-05,
+      "loss": 7.6738,
+      "mean_token_accuracy": 0.06392239518463612,
+      "num_tokens": 339615.0,
+      "step": 190
+    },
+    {
+      "entropy": 8.382297706604003,
+      "epoch": 0.010030090270812437,
+      "grad_norm": 0.9765625,
+      "learning_rate": 9.7e-05,
+      "loss": 7.5319,
+      "mean_token_accuracy": 0.0668424092233181,
+      "num_tokens": 347445.0,
+      "step": 195
+    },
+    {
+      "entropy": 8.23064022064209,
+      "epoch": 0.010287272072628141,
+      "grad_norm": 0.81640625,
+      "learning_rate": 9.95e-05,
+      "loss": 7.6527,
+      "mean_token_accuracy": 0.0642503272742033,
+      "num_tokens": 356466.0,
+      "step": 200
+    },
+    {
+      "entropy": 8.293260860443116,
+      "epoch": 0.010544453874443844,
+      "grad_norm": 0.890625,
+      "learning_rate": 0.000102,
+      "loss": 7.6586,
+      "mean_token_accuracy": 0.06924829706549644,
+      "num_tokens": 364582.0,
+      "step": 205
+    },
+    {
+      "entropy": 8.233483600616456,
+      "epoch": 0.010801635676259547,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.00010449999999999999,
+      "loss": 7.5999,
+      "mean_token_accuracy": 0.06699038669466972,
+      "num_tokens": 373816.0,
+      "step": 210
+    },
+    {
+      "entropy": 8.192768096923828,
+      "epoch": 0.011058817478075252,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.000107,
+      "loss": 7.6079,
+      "mean_token_accuracy": 0.06721281111240388,
+      "num_tokens": 382663.0,
+      "step": 215
+    },
+    {
+      "entropy": 8.15766954421997,
+      "epoch": 0.011315999279890954,
+      "grad_norm": 0.875,
+      "learning_rate": 0.0001095,
+      "loss": 7.5084,
+      "mean_token_accuracy": 0.06859233565628528,
+      "num_tokens": 391146.0,
+      "step": 220
+    },
+    {
+      "entropy": 8.103716373443604,
+      "epoch": 0.011573181081706659,
+      "grad_norm": 0.8125,
+      "learning_rate": 0.000112,
+      "loss": 7.5845,
+      "mean_token_accuracy": 0.06810487173497677,
+      "num_tokens": 400693.0,
+      "step": 225
+    },
+    {
+      "entropy": 8.105987453460694,
+      "epoch": 0.011830362883522362,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0001145,
+      "loss": 7.539,
+      "mean_token_accuracy": 0.06601319462060928,
+      "num_tokens": 409706.0,
+      "step": 230
+    },
+    {
+      "entropy": 8.056668090820313,
+      "epoch": 0.012087544685338066,
+      "grad_norm": 0.9140625,
+      "learning_rate": 0.00011700000000000001,
+      "loss": 7.4696,
+      "mean_token_accuracy": 0.06857900470495223,
+      "num_tokens": 418717.0,
+      "step": 235
+    },
+    {
+      "entropy": 8.077108192443848,
+      "epoch": 0.012344726487153769,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00011949999999999999,
+      "loss": 7.5335,
+      "mean_token_accuracy": 0.06780378967523575,
+      "num_tokens": 428630.0,
+      "step": 240
+    },
+    {
+      "entropy": 8.065593910217284,
+      "epoch": 0.012601908288969472,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.000122,
+      "loss": 7.559,
+      "mean_token_accuracy": 0.06397245228290557,
+      "num_tokens": 438762.0,
+      "step": 245
+    },
+    {
+      "entropy": 8.016668844223023,
+      "epoch": 0.012859090090785176,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0001245,
+      "loss": 7.4803,
+      "mean_token_accuracy": 0.06500645838677883,
+      "num_tokens": 447419.0,
+      "step": 250
+    },
+    {
+      "entropy": 8.066792869567871,
+      "epoch": 0.01311627189260088,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.000127,
+      "loss": 7.4872,
+      "mean_token_accuracy": 0.07560092583298683,
+      "num_tokens": 455818.0,
+      "step": 255
+    },
+    {
+      "entropy": 8.00608172416687,
+      "epoch": 0.013373453694416584,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0001295,
+      "loss": 7.4407,
+      "mean_token_accuracy": 0.0703579068183899,
+      "num_tokens": 464411.0,
+      "step": 260
+    },
+    {
+      "entropy": 8.006528282165528,
+      "epoch": 0.013630635496232286,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.000132,
+      "loss": 7.4791,
+      "mean_token_accuracy": 0.07310711704194546,
+      "num_tokens": 473181.0,
+      "step": 265
+    },
+    {
+      "entropy": 7.954403877258301,
+      "epoch": 0.01388781729804799,
+      "grad_norm": 0.890625,
+      "learning_rate": 0.00013450000000000002,
+      "loss": 7.3802,
+      "mean_token_accuracy": 0.07561531476676464,
+      "num_tokens": 481776.0,
+      "step": 270
+    },
+    {
+      "entropy": 7.960335302352905,
+      "epoch": 0.014144999099863694,
+      "grad_norm": 0.91796875,
+      "learning_rate": 0.00013700000000000002,
+      "loss": 7.3739,
+      "mean_token_accuracy": 0.07225163355469703,
+      "num_tokens": 490482.0,
+      "step": 275
+    },
+    {
+      "entropy": 7.908543682098388,
+      "epoch": 0.014402180901679397,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0001395,
+      "loss": 7.3373,
+      "mean_token_accuracy": 0.08000247925519943,
+      "num_tokens": 498394.0,
+      "step": 280
+    },
+    {
+      "entropy": 7.912304162979126,
+      "epoch": 0.014659362703495101,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00014199999999999998,
+      "loss": 7.3847,
+      "mean_token_accuracy": 0.07234559133648873,
+      "num_tokens": 507194.0,
+      "step": 285
+    },
+    {
+      "entropy": 7.905929660797119,
+      "epoch": 0.014916544505310804,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0001445,
+      "loss": 7.3824,
+      "mean_token_accuracy": 0.07954801470041276,
+      "num_tokens": 515538.0,
+      "step": 290
+    },
+    {
+      "entropy": 7.906702852249145,
+      "epoch": 0.015173726307126508,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.000147,
+      "loss": 7.4258,
+      "mean_token_accuracy": 0.0705490980297327,
+      "num_tokens": 524544.0,
+      "step": 295
+    },
+    {
+      "entropy": 7.956319189071655,
+      "epoch": 0.015430908108942211,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.0001495,
+      "loss": 7.3581,
+      "mean_token_accuracy": 0.0792141430079937,
+      "num_tokens": 533096.0,
+      "step": 300
+    },
+    {
+      "entropy": 7.908858060836792,
+      "epoch": 0.015688089910757916,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.000152,
+      "loss": 7.2829,
+      "mean_token_accuracy": 0.07599962502717972,
+      "num_tokens": 542254.0,
+      "step": 305
+    },
+    {
+      "entropy": 7.929060173034668,
+      "epoch": 0.01594527171257362,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.00015450000000000001,
+      "loss": 7.3804,
+      "mean_token_accuracy": 0.07816362082958221,
+      "num_tokens": 550985.0,
+      "step": 310
+    },
+    {
+      "entropy": 7.865027332305909,
+      "epoch": 0.01620245351438932,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.000157,
+      "loss": 7.3625,
+      "mean_token_accuracy": 0.07863678447902203,
+      "num_tokens": 560224.0,
+      "step": 315
+    },
+    {
+      "entropy": 7.856458759307861,
+      "epoch": 0.016459635316205024,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0001595,
+      "loss": 7.2099,
+      "mean_token_accuracy": 0.0831373617053032,
+      "num_tokens": 569268.0,
+      "step": 320
+    },
+    {
+      "entropy": 7.8261981964111325,
+      "epoch": 0.01671681711802073,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000162,
+      "loss": 7.305,
+      "mean_token_accuracy": 0.07629284039139747,
+      "num_tokens": 577876.0,
+      "step": 325
+    },
+    {
+      "entropy": 7.825508499145508,
+      "epoch": 0.016973998919836433,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.00016450000000000001,
+      "loss": 7.1964,
+      "mean_token_accuracy": 0.08476228266954422,
+      "num_tokens": 586622.0,
+      "step": 330
+    },
+    {
+      "entropy": 7.87431845664978,
+      "epoch": 0.017231180721652136,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.00016700000000000002,
+      "loss": 7.3968,
+      "mean_token_accuracy": 0.07733883932232857,
+      "num_tokens": 596544.0,
+      "step": 335
+    },
+    {
+      "entropy": 7.786734390258789,
+      "epoch": 0.01748836252346784,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00016950000000000003,
+      "loss": 7.2342,
+      "mean_token_accuracy": 0.07911348566412926,
+      "num_tokens": 605349.0,
+      "step": 340
+    },
+    {
+      "entropy": 7.809148788452148,
+      "epoch": 0.01774554432528354,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 7.2457,
+      "mean_token_accuracy": 0.07915201708674431,
+      "num_tokens": 614266.0,
+      "step": 345
+    },
+    {
+      "entropy": 7.771593856811523,
+      "epoch": 0.018002726127099248,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00017449999999999999,
+      "loss": 7.2829,
+      "mean_token_accuracy": 0.07773459404706955,
+      "num_tokens": 622534.0,
+      "step": 350
+    },
+    {
+      "entropy": 7.822271633148193,
+      "epoch": 0.01825990792891495,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000177,
+      "loss": 7.2365,
+      "mean_token_accuracy": 0.08085938617587089,
+      "num_tokens": 631838.0,
+      "step": 355
+    },
+    {
+      "entropy": 7.698997259140015,
+      "epoch": 0.018517089730730654,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0001795,
+      "loss": 7.2539,
+      "mean_token_accuracy": 0.08351444229483604,
+      "num_tokens": 641386.0,
+      "step": 360
+    },
+    {
+      "entropy": 7.703382873535157,
+      "epoch": 0.018774271532546356,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.000182,
+      "loss": 7.2384,
+      "mean_token_accuracy": 0.0824521966278553,
+      "num_tokens": 650465.0,
+      "step": 365
+    },
+    {
+      "entropy": 7.835447978973389,
+      "epoch": 0.01903145333436206,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0001845,
+      "loss": 7.1401,
+      "mean_token_accuracy": 0.08636756986379623,
+      "num_tokens": 659509.0,
+      "step": 370
+    },
+    {
+      "entropy": 7.586853075027466,
+      "epoch": 0.019288635136177765,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000187,
+      "loss": 7.1678,
+      "mean_token_accuracy": 0.08730523958802223,
+      "num_tokens": 668137.0,
+      "step": 375
+    },
+    {
+      "entropy": 7.862004041671753,
+      "epoch": 0.019545816937993468,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0001895,
+      "loss": 7.2499,
+      "mean_token_accuracy": 0.0831152357161045,
+      "num_tokens": 677784.0,
+      "step": 380
+    },
+    {
+      "entropy": 7.637575674057007,
+      "epoch": 0.01980299873980917,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000192,
+      "loss": 7.1701,
+      "mean_token_accuracy": 0.08491914793848991,
+      "num_tokens": 686088.0,
+      "step": 385
+    },
+    {
+      "entropy": 7.752926158905029,
+      "epoch": 0.020060180541624874,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0001945,
+      "loss": 7.3326,
+      "mean_token_accuracy": 0.08027631081640721,
+      "num_tokens": 695131.0,
+      "step": 390
+    },
+    {
+      "entropy": 7.793869876861573,
+      "epoch": 0.020317362343440577,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00019700000000000002,
+      "loss": 7.1152,
+      "mean_token_accuracy": 0.08675874844193458,
+      "num_tokens": 703976.0,
+      "step": 395
+    },
+    {
+      "entropy": 7.674239635467529,
+      "epoch": 0.020574544145256283,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00019950000000000002,
+      "loss": 7.1607,
+      "mean_token_accuracy": 0.08738962039351464,
+      "num_tokens": 712080.0,
+      "step": 400
+    },
+    {
+      "entropy": 7.662798118591309,
+      "epoch": 0.020831725947071986,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000202,
+      "loss": 7.2073,
+      "mean_token_accuracy": 0.08529370948672295,
+      "num_tokens": 721151.0,
+      "step": 405
+    },
+    {
+      "entropy": 7.703192710876465,
+      "epoch": 0.02108890774888769,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00020449999999999998,
+      "loss": 7.2059,
+      "mean_token_accuracy": 0.08231882154941558,
+      "num_tokens": 730957.0,
+      "step": 410
+    },
+    {
+      "entropy": 7.6682192325592045,
+      "epoch": 0.02134608955070339,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.000207,
+      "loss": 7.115,
+      "mean_token_accuracy": 0.09367254376411438,
+      "num_tokens": 740054.0,
+      "step": 415
+    },
+    {
+      "entropy": 7.710998058319092,
+      "epoch": 0.021603271352519094,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0002095,
+      "loss": 7.1138,
+      "mean_token_accuracy": 0.0901214174926281,
+      "num_tokens": 748576.0,
+      "step": 420
+    },
+    {
+      "entropy": 7.615434312820435,
+      "epoch": 0.0218604531543348,
+      "grad_norm": 0.91015625,
+      "learning_rate": 0.000212,
+      "loss": 7.1257,
+      "mean_token_accuracy": 0.08843043595552444,
+      "num_tokens": 758078.0,
+      "step": 425
+    },
+    {
+      "entropy": 7.6622356414794925,
+      "epoch": 0.022117634956150503,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0002145,
+      "loss": 7.1976,
+      "mean_token_accuracy": 0.08554179668426513,
+      "num_tokens": 767041.0,
+      "step": 430
+    },
+    {
+      "entropy": 7.6705575466156,
+      "epoch": 0.022374816757966206,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00021700000000000002,
+      "loss": 7.192,
+      "mean_token_accuracy": 0.08901726603507995,
+      "num_tokens": 775675.0,
+      "step": 435
+    },
+    {
+      "entropy": 7.627669334411621,
+      "epoch": 0.02263199855978191,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0002195,
+      "loss": 7.0133,
+      "mean_token_accuracy": 0.09209411665797233,
+      "num_tokens": 784105.0,
+      "step": 440
+    },
+    {
+      "entropy": 7.484321117401123,
+      "epoch": 0.022889180361597615,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000222,
+      "loss": 6.9712,
+      "mean_token_accuracy": 0.09117011949419976,
+      "num_tokens": 792861.0,
+      "step": 445
+    },
+    {
+      "entropy": 7.594349336624146,
+      "epoch": 0.023146362163413318,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0002245,
+      "loss": 7.0314,
+      "mean_token_accuracy": 0.08824475556612014,
+      "num_tokens": 801642.0,
+      "step": 450
+    },
+    {
+      "entropy": 7.497442388534546,
+      "epoch": 0.02340354396522902,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00022700000000000002,
+      "loss": 6.9939,
+      "mean_token_accuracy": 0.08806813433766365,
+      "num_tokens": 811105.0,
+      "step": 455
+    },
+    {
+      "entropy": 7.556538200378418,
+      "epoch": 0.023660725767044723,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00022950000000000002,
+      "loss": 7.0195,
+      "mean_token_accuracy": 0.09214109480381012,
+      "num_tokens": 819989.0,
+      "step": 460
+    },
+    {
+      "entropy": 7.516587781906128,
+      "epoch": 0.023917907568860426,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 7.0303,
+      "mean_token_accuracy": 0.0912808708846569,
+      "num_tokens": 829532.0,
+      "step": 465
+    },
+    {
+      "entropy": 7.524700498580932,
+      "epoch": 0.024175089370676132,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.00023449999999999998,
+      "loss": 6.9313,
+      "mean_token_accuracy": 0.09493043571710587,
+      "num_tokens": 838071.0,
+      "step": 470
+    },
+    {
+      "entropy": 7.6139366149902346,
+      "epoch": 0.024432271172491835,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000237,
+      "loss": 7.1399,
+      "mean_token_accuracy": 0.0876034751534462,
+      "num_tokens": 848094.0,
+      "step": 475
+    },
+    {
+      "entropy": 7.520268821716309,
+      "epoch": 0.024689452974307538,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0002395,
+      "loss": 7.0461,
+      "mean_token_accuracy": 0.09003157168626785,
+      "num_tokens": 856585.0,
+      "step": 480
+    },
+    {
+      "entropy": 7.4084981918334964,
+      "epoch": 0.02494663477612324,
+      "grad_norm": 1.0,
+      "learning_rate": 0.000242,
+      "loss": 7.0064,
+      "mean_token_accuracy": 0.09524489566683769,
+      "num_tokens": 865866.0,
+      "step": 485
+    },
+    {
+      "entropy": 7.565576791763306,
+      "epoch": 0.025203816577938944,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.0002445,
+      "loss": 7.0857,
+      "mean_token_accuracy": 0.09243087470531464,
+      "num_tokens": 874963.0,
+      "step": 490
+    },
+    {
+      "entropy": 7.522153425216675,
+      "epoch": 0.02546099837975465,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000247,
+      "loss": 7.0253,
+      "mean_token_accuracy": 0.09747636616230011,
+      "num_tokens": 883670.0,
+      "step": 495
+    },
+    {
+      "entropy": 7.411280775070191,
+      "epoch": 0.025718180181570353,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0002495,
+      "loss": 6.9456,
+      "mean_token_accuracy": 0.09364836364984512,
+      "num_tokens": 892322.0,
+      "step": 500
+    },
+    {
+      "entropy": 7.502190637588501,
+      "epoch": 0.025975361983386056,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.000252,
+      "loss": 7.0,
+      "mean_token_accuracy": 0.09010454788804054,
+      "num_tokens": 901450.0,
+      "step": 505
+    },
+    {
+      "entropy": 7.450657367706299,
+      "epoch": 0.02623254378520176,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0002545,
+      "loss": 6.9258,
+      "mean_token_accuracy": 0.0917640596628189,
+      "num_tokens": 910499.0,
+      "step": 510
+    },
+    {
+      "entropy": 7.4553405284881595,
+      "epoch": 0.02648972558701746,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.000257,
+      "loss": 7.042,
+      "mean_token_accuracy": 0.08597794771194459,
+      "num_tokens": 920223.0,
+      "step": 515
+    },
+    {
+      "entropy": 7.50772066116333,
+      "epoch": 0.026746907388833167,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.0002595,
+      "loss": 7.014,
+      "mean_token_accuracy": 0.08768303692340851,
+      "num_tokens": 929414.0,
+      "step": 520
+    },
+    {
+      "entropy": 7.4105640888214115,
+      "epoch": 0.02700408919064887,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.000262,
+      "loss": 6.98,
+      "mean_token_accuracy": 0.08690723404288292,
+      "num_tokens": 938059.0,
+      "step": 525
+    },
+    {
+      "entropy": 7.362936401367188,
+      "epoch": 0.027261270992464573,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.00026450000000000003,
+      "loss": 7.0251,
+      "mean_token_accuracy": 0.08964922651648521,
+      "num_tokens": 946868.0,
+      "step": 530
+    },
+    {
+      "entropy": 7.468600177764893,
+      "epoch": 0.027518452794280276,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00026700000000000004,
+      "loss": 6.867,
+      "mean_token_accuracy": 0.09712589457631111,
+      "num_tokens": 954997.0,
+      "step": 535
+    },
+    {
+      "entropy": 7.428448724746704,
+      "epoch": 0.02777563459609598,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00026950000000000005,
+      "loss": 6.9816,
+      "mean_token_accuracy": 0.09053401798009872,
+      "num_tokens": 964488.0,
+      "step": 540
+    },
+    {
+      "entropy": 7.317347574234009,
+      "epoch": 0.028032816397911685,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 6.8074,
+      "mean_token_accuracy": 0.10062685832381249,
+      "num_tokens": 972901.0,
+      "step": 545
+    },
+    {
+      "entropy": 7.45094141960144,
+      "epoch": 0.028289998199727388,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0002745,
+      "loss": 7.0093,
+      "mean_token_accuracy": 0.09124675542116165,
+      "num_tokens": 981922.0,
+      "step": 550
+    },
+    {
+      "entropy": 7.441890525817871,
+      "epoch": 0.02854718000154309,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000277,
+      "loss": 6.9774,
+      "mean_token_accuracy": 0.09369566291570663,
+      "num_tokens": 990419.0,
+      "step": 555
+    },
+    {
+      "entropy": 7.304813051223755,
+      "epoch": 0.028804361803358793,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0002795,
+      "loss": 6.8424,
+      "mean_token_accuracy": 0.09960516914725304,
+      "num_tokens": 998236.0,
+      "step": 560
+    },
+    {
+      "entropy": 7.3862227439880375,
+      "epoch": 0.0290615436051745,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 6.8905,
+      "mean_token_accuracy": 0.09915700182318687,
+      "num_tokens": 1008258.0,
+      "step": 565
+    },
+    {
+      "entropy": 7.355383062362671,
+      "epoch": 0.029318725406990202,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0002845,
+      "loss": 6.8729,
+      "mean_token_accuracy": 0.09238902553915977,
+      "num_tokens": 1017029.0,
+      "step": 570
+    },
+    {
+      "entropy": 7.315412712097168,
+      "epoch": 0.029575907208805905,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.000287,
+      "loss": 6.9067,
+      "mean_token_accuracy": 0.09177370145916938,
+      "num_tokens": 1026497.0,
+      "step": 575
+    },
+    {
+      "entropy": 7.337841367721557,
+      "epoch": 0.029833089010621608,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0002895,
+      "loss": 6.8092,
+      "mean_token_accuracy": 0.10336045101284981,
+      "num_tokens": 1034728.0,
+      "step": 580
+    },
+    {
+      "entropy": 7.256781911849975,
+      "epoch": 0.03009027081243731,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000292,
+      "loss": 6.8645,
+      "mean_token_accuracy": 0.09807764887809753,
+      "num_tokens": 1043556.0,
+      "step": 585
+    },
+    {
+      "entropy": 7.315465259552002,
+      "epoch": 0.030347452614253017,
+      "grad_norm": 1.25,
+      "learning_rate": 0.0002945,
+      "loss": 6.7914,
+      "mean_token_accuracy": 0.10462642684578896,
+      "num_tokens": 1053449.0,
+      "step": 590
+    },
+    {
+      "entropy": 7.294723844528198,
+      "epoch": 0.03060463441606872,
+      "grad_norm": 1.25,
+      "learning_rate": 0.000297,
+      "loss": 6.8701,
+      "mean_token_accuracy": 0.09261535704135895,
+      "num_tokens": 1062691.0,
+      "step": 595
+    },
+    {
+      "entropy": 7.328425025939941,
+      "epoch": 0.030861816217884423,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0002995,
+      "loss": 6.8732,
+      "mean_token_accuracy": 0.09672143906354905,
+      "num_tokens": 1072194.0,
+      "step": 600
+    },
+    {
+      "entropy": 7.280484533309936,
+      "epoch": 0.031118998019700125,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000302,
+      "loss": 6.7832,
+      "mean_token_accuracy": 0.09872276112437248,
+      "num_tokens": 1080781.0,
+      "step": 605
+    },
+    {
+      "entropy": 7.2788474559783936,
+      "epoch": 0.03137617982151583,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0003045,
+      "loss": 6.9191,
+      "mean_token_accuracy": 0.09485599771142006,
+      "num_tokens": 1090048.0,
+      "step": 610
+    },
+    {
+      "entropy": 7.34011402130127,
+      "epoch": 0.03163336162333153,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000307,
+      "loss": 6.9368,
+      "mean_token_accuracy": 0.09523111060261727,
+      "num_tokens": 1100121.0,
+      "step": 615
+    },
+    {
+      "entropy": 7.326197099685669,
+      "epoch": 0.03189054342514724,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003095,
+      "loss": 6.8036,
+      "mean_token_accuracy": 0.09568134695291519,
+      "num_tokens": 1108569.0,
+      "step": 620
+    },
+    {
+      "entropy": 7.213407182693482,
+      "epoch": 0.032147725226962943,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.000312,
+      "loss": 6.776,
+      "mean_token_accuracy": 0.09914244189858437,
+      "num_tokens": 1118247.0,
+      "step": 625
+    },
+    {
+      "entropy": 7.210325479507446,
+      "epoch": 0.03240490702877864,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0003145,
+      "loss": 6.7999,
+      "mean_token_accuracy": 0.09887302219867707,
+      "num_tokens": 1126951.0,
+      "step": 630
+    },
+    {
+      "entropy": 7.209123611450195,
+      "epoch": 0.03266208883059435,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000317,
+      "loss": 6.7651,
+      "mean_token_accuracy": 0.10773184821009636,
+      "num_tokens": 1135390.0,
+      "step": 635
+    },
+    {
+      "entropy": 7.2418413162231445,
+      "epoch": 0.03291927063241005,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0003195,
+      "loss": 6.8057,
+      "mean_token_accuracy": 0.09772193506360054,
+      "num_tokens": 1143911.0,
+      "step": 640
+    },
+    {
+      "entropy": 7.288219833374024,
+      "epoch": 0.033176452434225755,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.000322,
+      "loss": 6.7833,
+      "mean_token_accuracy": 0.10312128961086273,
+      "num_tokens": 1152957.0,
+      "step": 645
+    },
+    {
+      "entropy": 7.144729995727539,
+      "epoch": 0.03343363423604146,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.00032450000000000003,
+      "loss": 6.7073,
+      "mean_token_accuracy": 0.10490344762802124,
+      "num_tokens": 1161911.0,
+      "step": 650
+    },
+    {
+      "entropy": 7.214617156982422,
+      "epoch": 0.03369081603785716,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00032700000000000003,
+      "loss": 6.7621,
+      "mean_token_accuracy": 0.10395861864089966,
+      "num_tokens": 1170932.0,
+      "step": 655
+    },
+    {
+      "entropy": 7.226040458679199,
+      "epoch": 0.03394799783967287,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00032950000000000004,
+      "loss": 6.8979,
+      "mean_token_accuracy": 0.09301602616906166,
+      "num_tokens": 1180417.0,
+      "step": 660
+    },
+    {
+      "entropy": 7.148654270172119,
+      "epoch": 0.034205179641488566,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 6.7629,
+      "mean_token_accuracy": 0.09860587194561958,
+      "num_tokens": 1189153.0,
+      "step": 665
+    },
+    {
+      "entropy": 7.214554643630981,
+      "epoch": 0.03446236144330427,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00033450000000000005,
+      "loss": 6.7476,
+      "mean_token_accuracy": 0.1043582171201706,
+      "num_tokens": 1197903.0,
+      "step": 670
+    },
+    {
+      "entropy": 7.1135680198669435,
+      "epoch": 0.03471954324511998,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000337,
+      "loss": 6.7435,
+      "mean_token_accuracy": 0.10275317579507828,
+      "num_tokens": 1207310.0,
+      "step": 675
+    },
+    {
+      "entropy": 7.197217321395874,
+      "epoch": 0.03497672504693568,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.0003395,
+      "loss": 6.7117,
+      "mean_token_accuracy": 0.10364129021763802,
+      "num_tokens": 1215577.0,
+      "step": 680
+    },
+    {
+      "entropy": 7.160997724533081,
+      "epoch": 0.035233906848751384,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.000342,
+      "loss": 6.801,
+      "mean_token_accuracy": 0.10323198661208152,
+      "num_tokens": 1223906.0,
+      "step": 685
+    },
+    {
+      "entropy": 7.1171308040618895,
+      "epoch": 0.03549108865056708,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00034449999999999997,
+      "loss": 6.7434,
+      "mean_token_accuracy": 0.10217863768339157,
+      "num_tokens": 1232689.0,
+      "step": 690
+    },
+    {
+      "entropy": 7.243052577972412,
+      "epoch": 0.03574827045238279,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000347,
+      "loss": 6.777,
+      "mean_token_accuracy": 0.10172305777668952,
+      "num_tokens": 1241079.0,
+      "step": 695
+    },
+    {
+      "entropy": 7.0958874225616455,
+      "epoch": 0.036005452254198496,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0003495,
+      "loss": 6.6994,
+      "mean_token_accuracy": 0.10428043976426124,
+      "num_tokens": 1249540.0,
+      "step": 700
+    },
+    {
+      "entropy": 7.236631059646607,
+      "epoch": 0.036262634056014195,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000352,
+      "loss": 6.767,
+      "mean_token_accuracy": 0.10423633679747582,
+      "num_tokens": 1257823.0,
+      "step": 705
+    },
+    {
+      "entropy": 7.162586879730225,
+      "epoch": 0.0365198158578299,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0003545,
+      "loss": 6.7752,
+      "mean_token_accuracy": 0.10328136906027793,
+      "num_tokens": 1267242.0,
+      "step": 710
+    },
+    {
+      "entropy": 7.095451688766479,
+      "epoch": 0.0367769976596456,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.000357,
+      "loss": 6.6616,
+      "mean_token_accuracy": 0.10539852529764175,
+      "num_tokens": 1275942.0,
+      "step": 715
+    },
+    {
+      "entropy": 7.087832736968994,
+      "epoch": 0.03703417946146131,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0003595,
+      "loss": 6.716,
+      "mean_token_accuracy": 0.10562722086906433,
+      "num_tokens": 1284840.0,
+      "step": 720
+    },
+    {
+      "entropy": 7.123103380203247,
+      "epoch": 0.03729136126327701,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000362,
+      "loss": 6.6864,
+      "mean_token_accuracy": 0.10260284543037415,
+      "num_tokens": 1293592.0,
+      "step": 725
+    },
+    {
+      "entropy": 7.171104860305786,
+      "epoch": 0.03754854306509271,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0003645,
+      "loss": 6.8006,
+      "mean_token_accuracy": 0.10065393671393394,
+      "num_tokens": 1301903.0,
+      "step": 730
+    },
+    {
+      "entropy": 7.110880374908447,
+      "epoch": 0.03780572486690842,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.000367,
+      "loss": 6.6262,
+      "mean_token_accuracy": 0.105719206482172,
+      "num_tokens": 1310230.0,
+      "step": 735
+    },
+    {
+      "entropy": 7.178726625442505,
+      "epoch": 0.03806290666872412,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0003695,
+      "loss": 6.7676,
+      "mean_token_accuracy": 0.10490798503160477,
+      "num_tokens": 1320029.0,
+      "step": 740
+    },
+    {
+      "entropy": 7.001174449920654,
+      "epoch": 0.038320088470539825,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.000372,
+      "loss": 6.6133,
+      "mean_token_accuracy": 0.10703234896063804,
+      "num_tokens": 1329155.0,
+      "step": 745
+    },
+    {
+      "entropy": 7.125448083877563,
+      "epoch": 0.03857727027235553,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0003745,
+      "loss": 6.8029,
+      "mean_token_accuracy": 0.1019430361688137,
+      "num_tokens": 1337958.0,
+      "step": 750
+    },
+    {
+      "entropy": 7.109465599060059,
+      "epoch": 0.03883445207417123,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.000377,
+      "loss": 6.7375,
+      "mean_token_accuracy": 0.09966553151607513,
+      "num_tokens": 1347105.0,
+      "step": 755
+    },
+    {
+      "entropy": 7.161317777633667,
+      "epoch": 0.039091633875986936,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0003795,
+      "loss": 6.6791,
+      "mean_token_accuracy": 0.10060147866606713,
+      "num_tokens": 1356161.0,
+      "step": 760
+    },
+    {
+      "entropy": 6.9970433712005615,
+      "epoch": 0.039348815677802636,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000382,
+      "loss": 6.6891,
+      "mean_token_accuracy": 0.10932381451129913,
+      "num_tokens": 1365291.0,
+      "step": 765
+    },
+    {
+      "entropy": 7.0630707263946535,
+      "epoch": 0.03960599747961834,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0003845,
+      "loss": 6.6481,
+      "mean_token_accuracy": 0.1059026412665844,
+      "num_tokens": 1374057.0,
+      "step": 770
+    },
+    {
+      "entropy": 7.118219041824341,
+      "epoch": 0.03986317928143405,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00038700000000000003,
+      "loss": 6.6895,
+      "mean_token_accuracy": 0.09991606175899506,
+      "num_tokens": 1383204.0,
+      "step": 775
+    },
+    {
+      "entropy": 7.0793726444244385,
+      "epoch": 0.04012036108324975,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00038950000000000003,
+      "loss": 6.7114,
+      "mean_token_accuracy": 0.1018392451107502,
+      "num_tokens": 1392303.0,
+      "step": 780
+    },
+    {
+      "entropy": 7.041213703155518,
+      "epoch": 0.040377542885065454,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 6.6789,
+      "mean_token_accuracy": 0.10870900377631187,
+      "num_tokens": 1401955.0,
+      "step": 785
+    },
+    {
+      "entropy": 7.127471685409546,
+      "epoch": 0.04063472468688115,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00039450000000000005,
+      "loss": 6.6934,
+      "mean_token_accuracy": 0.10036104544997215,
+      "num_tokens": 1410467.0,
+      "step": 790
+    },
+    {
+      "entropy": 6.975720405578613,
+      "epoch": 0.04089190648869686,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00039700000000000005,
+      "loss": 6.6748,
+      "mean_token_accuracy": 0.09731148779392243,
+      "num_tokens": 1420337.0,
+      "step": 795
+    },
+    {
+      "entropy": 7.059706926345825,
+      "epoch": 0.041149088290512566,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0003995,
+      "loss": 6.6711,
+      "mean_token_accuracy": 0.10144198536872864,
+      "num_tokens": 1429453.0,
+      "step": 800
+    },
+    {
+      "entropy": 7.094644546508789,
+      "epoch": 0.041406270092328265,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000402,
+      "loss": 6.6208,
+      "mean_token_accuracy": 0.10604123547673225,
+      "num_tokens": 1438417.0,
+      "step": 805
+    },
+    {
+      "entropy": 6.986465740203857,
+      "epoch": 0.04166345189414397,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.0004045,
+      "loss": 6.6667,
+      "mean_token_accuracy": 0.10473878160119057,
+      "num_tokens": 1446480.0,
+      "step": 810
+    },
+    {
+      "entropy": 6.946633005142212,
+      "epoch": 0.04192063369595967,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00040699999999999997,
+      "loss": 6.6306,
+      "mean_token_accuracy": 0.10496868267655372,
+      "num_tokens": 1455102.0,
+      "step": 815
+    },
+    {
+      "entropy": 7.14404935836792,
+      "epoch": 0.04217781549777538,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004095,
+      "loss": 6.6538,
+      "mean_token_accuracy": 0.10508396551012993,
+      "num_tokens": 1463839.0,
+      "step": 820
+    },
+    {
+      "entropy": 7.024435377120971,
+      "epoch": 0.04243499729959108,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000412,
+      "loss": 6.6836,
+      "mean_token_accuracy": 0.10788170769810676,
+      "num_tokens": 1472806.0,
+      "step": 825
+    },
+    {
+      "entropy": 6.979895305633545,
+      "epoch": 0.04269217910140678,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004145,
+      "loss": 6.6016,
+      "mean_token_accuracy": 0.10319365486502648,
+      "num_tokens": 1481707.0,
+      "step": 830
+    },
+    {
+      "entropy": 6.948284530639649,
+      "epoch": 0.04294936090322249,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.000417,
+      "loss": 6.6068,
+      "mean_token_accuracy": 0.10952205657958984,
+      "num_tokens": 1491057.0,
+      "step": 835
+    },
+    {
+      "entropy": 6.963069152832031,
+      "epoch": 0.04320654270503819,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004195,
+      "loss": 6.5666,
+      "mean_token_accuracy": 0.10177317038178443,
+      "num_tokens": 1500331.0,
+      "step": 840
+    },
+    {
+      "entropy": 6.964898347854614,
+      "epoch": 0.043463724506853894,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000422,
+      "loss": 6.6376,
+      "mean_token_accuracy": 0.10519776120781898,
+      "num_tokens": 1509051.0,
+      "step": 845
+    },
+    {
+      "entropy": 7.032815408706665,
+      "epoch": 0.0437209063086696,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.0004245,
+      "loss": 6.5931,
+      "mean_token_accuracy": 0.10714114159345627,
+      "num_tokens": 1517082.0,
+      "step": 850
+    },
+    {
+      "entropy": 6.869009304046631,
+      "epoch": 0.0439780881104853,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.000427,
+      "loss": 6.5338,
+      "mean_token_accuracy": 0.11201286613941193,
+      "num_tokens": 1525498.0,
+      "step": 855
+    },
+    {
+      "entropy": 6.827907609939575,
+      "epoch": 0.044235269912301006,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004295,
+      "loss": 6.4978,
+      "mean_token_accuracy": 0.11273056790232658,
+      "num_tokens": 1534476.0,
+      "step": 860
+    },
+    {
+      "entropy": 6.958985805511475,
+      "epoch": 0.04449245171411671,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000432,
+      "loss": 6.5896,
+      "mean_token_accuracy": 0.10771357268095016,
+      "num_tokens": 1542905.0,
+      "step": 865
+    },
+    {
+      "entropy": 6.942825794219971,
+      "epoch": 0.04474963351593241,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004345,
+      "loss": 6.583,
+      "mean_token_accuracy": 0.10546278730034828,
+      "num_tokens": 1551763.0,
+      "step": 870
+    },
+    {
+      "entropy": 6.906496191024781,
+      "epoch": 0.04500681531774812,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000437,
+      "loss": 6.5085,
+      "mean_token_accuracy": 0.11341241300106049,
+      "num_tokens": 1560520.0,
+      "step": 875
+    },
+    {
+      "entropy": 6.904255485534668,
+      "epoch": 0.04526399711956382,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004395,
+      "loss": 6.4871,
+      "mean_token_accuracy": 0.11152468100190163,
+      "num_tokens": 1569401.0,
+      "step": 880
+    },
+    {
+      "entropy": 6.922533941268921,
+      "epoch": 0.045521178921379524,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000442,
+      "loss": 6.5564,
+      "mean_token_accuracy": 0.10919757708907127,
+      "num_tokens": 1578440.0,
+      "step": 885
+    },
+    {
+      "entropy": 6.789845991134643,
+      "epoch": 0.04577836072319523,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0004445,
+      "loss": 6.4462,
+      "mean_token_accuracy": 0.11144870221614837,
+      "num_tokens": 1587551.0,
+      "step": 890
+    },
+    {
+      "entropy": 6.944276809692383,
+      "epoch": 0.04603554252501093,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.000447,
+      "loss": 6.6154,
+      "mean_token_accuracy": 0.10818377807736397,
+      "num_tokens": 1596314.0,
+      "step": 895
+    },
+    {
+      "entropy": 6.900253582000732,
+      "epoch": 0.046292724326826636,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00044950000000000003,
+      "loss": 6.4889,
+      "mean_token_accuracy": 0.1112966388463974,
+      "num_tokens": 1605025.0,
+      "step": 900
+    },
+    {
+      "entropy": 6.905750894546509,
+      "epoch": 0.046549906128642335,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 6.5348,
+      "mean_token_accuracy": 0.10826415568590164,
+      "num_tokens": 1613295.0,
+      "step": 905
+    },
+    {
+      "entropy": 6.862349224090576,
+      "epoch": 0.04680708793045804,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.00045450000000000004,
+      "loss": 6.5131,
+      "mean_token_accuracy": 0.1080511450767517,
+      "num_tokens": 1621865.0,
+      "step": 910
+    },
+    {
+      "entropy": 6.87268099784851,
+      "epoch": 0.04706426973227375,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00045700000000000005,
+      "loss": 6.517,
+      "mean_token_accuracy": 0.11334388554096222,
+      "num_tokens": 1631036.0,
+      "step": 915
+    },
+    {
+      "entropy": 6.960796642303467,
+      "epoch": 0.04732145153408945,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00045950000000000006,
+      "loss": 6.6083,
+      "mean_token_accuracy": 0.10900305211544037,
+      "num_tokens": 1640311.0,
+      "step": 920
+    },
+    {
+      "entropy": 6.877505731582642,
+      "epoch": 0.04757863333590515,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.000462,
+      "loss": 6.5707,
+      "mean_token_accuracy": 0.11122839301824569,
+      "num_tokens": 1648651.0,
+      "step": 925
+    },
+    {
+      "entropy": 6.814219570159912,
+      "epoch": 0.04783581513772085,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.0004645,
+      "loss": 6.5091,
+      "mean_token_accuracy": 0.11311314180493355,
+      "num_tokens": 1657493.0,
+      "step": 930
+    },
+    {
+      "entropy": 6.919810724258423,
+      "epoch": 0.04809299693953656,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.000467,
+      "loss": 6.5672,
+      "mean_token_accuracy": 0.10320627093315124,
+      "num_tokens": 1666909.0,
+      "step": 935
+    },
+    {
+      "entropy": 6.835229587554932,
+      "epoch": 0.048350178741352265,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004695,
+      "loss": 6.5857,
+      "mean_token_accuracy": 0.10411025136709214,
+      "num_tokens": 1676071.0,
+      "step": 940
+    },
+    {
+      "entropy": 6.990052223205566,
+      "epoch": 0.048607360543167964,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.000472,
+      "loss": 6.6305,
+      "mean_token_accuracy": 0.10452140420675278,
+      "num_tokens": 1685240.0,
+      "step": 945
+    },
+    {
+      "entropy": 6.916949987411499,
+      "epoch": 0.04886454234498367,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004745,
+      "loss": 6.5346,
+      "mean_token_accuracy": 0.117493636906147,
+      "num_tokens": 1694441.0,
+      "step": 950
+    },
+    {
+      "entropy": 6.721475076675415,
+      "epoch": 0.04912172414679937,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.000477,
+      "loss": 6.4578,
+      "mean_token_accuracy": 0.1129125975072384,
+      "num_tokens": 1702925.0,
+      "step": 955
+    },
+    {
+      "entropy": 6.895248603820801,
+      "epoch": 0.049378905948615076,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004795,
+      "loss": 6.4835,
+      "mean_token_accuracy": 0.10699952915310859,
+      "num_tokens": 1711708.0,
+      "step": 960
+    },
+    {
+      "entropy": 6.799575614929199,
+      "epoch": 0.04963608775043078,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.000482,
+      "loss": 6.4901,
+      "mean_token_accuracy": 0.10597067773342132,
+      "num_tokens": 1720543.0,
+      "step": 965
+    },
+    {
+      "entropy": 6.862289619445801,
+      "epoch": 0.04989326955224648,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0004845,
+      "loss": 6.4988,
+      "mean_token_accuracy": 0.11304299309849739,
+      "num_tokens": 1729776.0,
+      "step": 970
+    },
+    {
+      "entropy": 6.8804943561553955,
+      "epoch": 0.05015045135406219,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.000487,
+      "loss": 6.4822,
+      "mean_token_accuracy": 0.11637242883443832,
+      "num_tokens": 1738249.0,
+      "step": 975
+    },
+    {
+      "entropy": 6.828695917129517,
+      "epoch": 0.05040763315587789,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0004895,
+      "loss": 6.5063,
+      "mean_token_accuracy": 0.11184739843010902,
+      "num_tokens": 1747262.0,
+      "step": 980
+    },
+    {
+      "entropy": 6.891150426864624,
+      "epoch": 0.050664814957693594,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.000492,
+      "loss": 6.5072,
+      "mean_token_accuracy": 0.11530979126691818,
+      "num_tokens": 1755320.0,
+      "step": 985
+    },
+    {
+      "entropy": 6.7496922492980955,
+      "epoch": 0.0509219967595093,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004945,
+      "loss": 6.5038,
+      "mean_token_accuracy": 0.11211839094758033,
+      "num_tokens": 1763798.0,
+      "step": 990
+    },
+    {
+      "entropy": 6.788990116119384,
+      "epoch": 0.051179178561325,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.000497,
+      "loss": 6.4671,
+      "mean_token_accuracy": 0.11327328234910965,
+      "num_tokens": 1773862.0,
+      "step": 995
+    },
+    {
+      "entropy": 6.837374210357666,
+      "epoch": 0.051436360363140705,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004995,
+      "loss": 6.5248,
+      "mean_token_accuracy": 0.11146914437413216,
+      "num_tokens": 1783552.0,
+      "step": 1000
+    },
+    {
+      "entropy": 6.746343898773193,
+      "epoch": 0.051693542164956405,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000499998026082006,
+      "loss": 6.426,
+      "mean_token_accuracy": 0.11493942067027092,
+      "num_tokens": 1792245.0,
+      "step": 1005
+    },
+    {
+      "entropy": 6.774056434631348,
+      "epoch": 0.05195072396677211,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004999900070995136,
+      "loss": 6.5069,
+      "mean_token_accuracy": 0.11212562620639802,
+      "num_tokens": 1800813.0,
+      "step": 1010
+    },
+    {
+      "entropy": 6.852898216247558,
+      "epoch": 0.05220790576858782,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004999758199023239,
+      "loss": 6.5207,
+      "mean_token_accuracy": 0.11174971088767052,
+      "num_tokens": 1809671.0,
+      "step": 1015
+    },
+    {
+      "entropy": 6.741954612731933,
+      "epoch": 0.05246508757040352,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004999554648793858,
+      "loss": 6.4463,
+      "mean_token_accuracy": 0.11097860932350159,
+      "num_tokens": 1817968.0,
+      "step": 1020
+    },
+    {
+      "entropy": 6.815918350219727,
+      "epoch": 0.05272226937221922,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004999289425887425,
+      "loss": 6.529,
+      "mean_token_accuracy": 0.10783923268318177,
+      "num_tokens": 1826302.0,
+      "step": 1025
+    },
+    {
+      "entropy": 6.708252668380737,
+      "epoch": 0.05297945117403492,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004998962537575161,
+      "loss": 6.4088,
+      "mean_token_accuracy": 0.11643101274967194,
+      "num_tokens": 1836008.0,
+      "step": 1030
+    },
+    {
+      "entropy": 6.737126874923706,
+      "epoch": 0.05323663297585063,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0004998573992818874,
+      "loss": 6.4282,
+      "mean_token_accuracy": 0.11566728353500366,
+      "num_tokens": 1844730.0,
+      "step": 1035
+    },
+    {
+      "entropy": 6.724181604385376,
+      "epoch": 0.053493814777666335,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004998123802270715,
+      "loss": 6.402,
+      "mean_token_accuracy": 0.11093823388218879,
+      "num_tokens": 1853186.0,
+      "step": 1040
+    },
+    {
+      "entropy": 6.778470802307129,
+      "epoch": 0.053750996579482034,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.0004997611978272886,
+      "loss": 6.376,
+      "mean_token_accuracy": 0.1124820165336132,
+      "num_tokens": 1861579.0,
+      "step": 1045
+    },
+    {
+      "entropy": 6.778669786453247,
+      "epoch": 0.05400817838129774,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004997038534857298,
+      "loss": 6.5295,
+      "mean_token_accuracy": 0.1051288291811943,
+      "num_tokens": 1871948.0,
+      "step": 1050
+    },
+    {
+      "entropy": 6.779291105270386,
+      "epoch": 0.05426536018311344,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004996403487745194,
+      "loss": 6.4291,
+      "mean_token_accuracy": 0.11674032881855964,
+      "num_tokens": 1880142.0,
+      "step": 1055
+    },
+    {
+      "entropy": 6.749887180328369,
+      "epoch": 0.054522541984929146,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.000499570685434671,
+      "loss": 6.4415,
+      "mean_token_accuracy": 0.10835661590099335,
+      "num_tokens": 1888929.0,
+      "step": 1060
+    },
+    {
+      "entropy": 6.763294887542725,
+      "epoch": 0.05477972378674485,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0004994948653760405,
+      "loss": 6.4847,
+      "mean_token_accuracy": 0.10644253864884376,
+      "num_tokens": 1897982.0,
+      "step": 1065
+    },
+    {
+      "entropy": 6.66923336982727,
+      "epoch": 0.05503690558856055,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004994128906772729,
+      "loss": 6.3776,
+      "mean_token_accuracy": 0.11694736555218696,
+      "num_tokens": 1906583.0,
+      "step": 1070
+    },
+    {
+      "entropy": 6.752908086776733,
+      "epoch": 0.05529408739037626,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000499324763585746,
+      "loss": 6.4433,
+      "mean_token_accuracy": 0.11435103714466095,
+      "num_tokens": 1915457.0,
+      "step": 1075
+    },
+    {
+      "entropy": 6.66996021270752,
+      "epoch": 0.05555126919219196,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004992304865175085,
+      "loss": 6.3785,
+      "mean_token_accuracy": 0.12019199803471566,
+      "num_tokens": 1923802.0,
+      "step": 1080
+    },
+    {
+      "entropy": 6.735610485076904,
+      "epoch": 0.05580845099400766,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004991300620572138,
+      "loss": 6.3917,
+      "mean_token_accuracy": 0.11757480353116989,
+      "num_tokens": 1933097.0,
+      "step": 1085
+    },
+    {
+      "entropy": 6.72240138053894,
+      "epoch": 0.05606563279582337,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004990234929580494,
+      "loss": 6.3646,
+      "mean_token_accuracy": 0.11228772699832916,
+      "num_tokens": 1942240.0,
+      "step": 1090
+    },
+    {
+      "entropy": 6.5773022174835205,
+      "epoch": 0.05632281459763907,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004989107821416609,
+      "loss": 6.3542,
+      "mean_token_accuracy": 0.11520460173487664,
+      "num_tokens": 1950501.0,
+      "step": 1095
+    },
+    {
+      "entropy": 6.835616827011108,
+      "epoch": 0.056579996399454775,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.0004987919326980723,
+      "loss": 6.4121,
+      "mean_token_accuracy": 0.11566629931330681,
+      "num_tokens": 1958621.0,
+      "step": 1100
+    },
+    {
+      "entropy": 6.723128461837769,
+      "epoch": 0.05683717820127048,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004986669478856011,
+      "loss": 6.5071,
+      "mean_token_accuracy": 0.11123462915420532,
+      "num_tokens": 1966836.0,
+      "step": 1105
+    },
+    {
+      "entropy": 6.7716064453125,
+      "epoch": 0.05709436000308618,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004985358311307688,
+      "loss": 6.4827,
+      "mean_token_accuracy": 0.11127110049128533,
+      "num_tokens": 1976046.0,
+      "step": 1110
+    },
+    {
+      "entropy": 6.752973651885986,
+      "epoch": 0.05735154180490189,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004983985860282081,
+      "loss": 6.4645,
+      "mean_token_accuracy": 0.11258459836244583,
+      "num_tokens": 1984074.0,
+      "step": 1115
+    },
+    {
+      "entropy": 6.730236387252807,
+      "epoch": 0.057608723606717586,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004982552163405623,
+      "loss": 6.539,
+      "mean_token_accuracy": 0.11384548544883728,
+      "num_tokens": 1992898.0,
+      "step": 1120
+    },
+    {
+      "entropy": 6.832723760604859,
+      "epoch": 0.05786590540853329,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004981057259983839,
+      "loss": 6.4669,
+      "mean_token_accuracy": 0.11225469931960105,
+      "num_tokens": 2002390.0,
+      "step": 1125
+    },
+    {
+      "entropy": 6.657161140441895,
+      "epoch": 0.058123087210349,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004979501191000262,
+      "loss": 6.3538,
+      "mean_token_accuracy": 0.1199690505862236,
+      "num_tokens": 2011499.0,
+      "step": 1130
+    },
+    {
+      "entropy": 6.6502784252166744,
+      "epoch": 0.0583802690121647,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004977883999115311,
+      "loss": 6.4782,
+      "mean_token_accuracy": 0.11357165277004241,
+      "num_tokens": 2021336.0,
+      "step": 1135
+    },
+    {
+      "entropy": 6.731290435791015,
+      "epoch": 0.058637450813980405,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004976205728665113,
+      "loss": 6.3638,
+      "mean_token_accuracy": 0.11489428207278252,
+      "num_tokens": 2029706.0,
+      "step": 1140
+    },
+    {
+      "entropy": 6.830005741119384,
+      "epoch": 0.058894632615796104,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0004974466425660307,
+      "loss": 6.5844,
+      "mean_token_accuracy": 0.10661658570170403,
+      "num_tokens": 2038915.0,
+      "step": 1145
+    },
+    {
+      "entropy": 6.701237726211548,
+      "epoch": 0.05915181441761181,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004972666137784759,
+      "loss": 6.418,
+      "mean_token_accuracy": 0.11196302473545075,
+      "num_tokens": 2047661.0,
+      "step": 1150
+    },
+    {
+      "entropy": 6.637448024749756,
+      "epoch": 0.059408996219427516,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004970804914394271,
+      "loss": 6.4053,
+      "mean_token_accuracy": 0.11128832325339318,
+      "num_tokens": 2056768.0,
+      "step": 1155
+    },
+    {
+      "entropy": 6.7576159000396725,
+      "epoch": 0.059666178021243216,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004968882806515225,
+      "loss": 6.4509,
+      "mean_token_accuracy": 0.11101424172520638,
+      "num_tokens": 2066862.0,
+      "step": 1160
+    },
+    {
+      "entropy": 6.65157413482666,
+      "epoch": 0.05992335982305892,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004966899866843177,
+      "loss": 6.2771,
+      "mean_token_accuracy": 0.11530844420194626,
+      "num_tokens": 2075794.0,
+      "step": 1165
+    },
+    {
+      "entropy": 6.693396663665771,
+      "epoch": 0.06018054162487462,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000496485614974142,
+      "loss": 6.4464,
+      "mean_token_accuracy": 0.11457487121224404,
+      "num_tokens": 2084806.0,
+      "step": 1170
+    },
+    {
+      "entropy": 6.578125905990601,
+      "epoch": 0.06043772342669033,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0004962751711239492,
+      "loss": 6.3015,
+      "mean_token_accuracy": 0.11791201233863831,
+      "num_tokens": 2093884.0,
+      "step": 1175
+    },
+    {
+      "entropy": 6.601912546157837,
+      "epoch": 0.060694905228506034,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004960586609031636,
+      "loss": 6.2546,
+      "mean_token_accuracy": 0.12269692420959473,
+      "num_tokens": 2103716.0,
+      "step": 1180
+    },
+    {
+      "entropy": 6.633868169784546,
+      "epoch": 0.06095208703032173,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004958360902475224,
+      "loss": 6.3092,
+      "mean_token_accuracy": 0.11640048250555993,
+      "num_tokens": 2113627.0,
+      "step": 1185
+    },
+    {
+      "entropy": 6.588856792449951,
+      "epoch": 0.06120926883213744,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.0004956074652589125,
+      "loss": 6.2842,
+      "mean_token_accuracy": 0.1199034109711647,
+      "num_tokens": 2121455.0,
+      "step": 1190
+    },
+    {
+      "entropy": 6.636424875259399,
+      "epoch": 0.06146645063395314,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004953727922052035,
+      "loss": 6.2933,
+      "mean_token_accuracy": 0.11959701478481292,
+      "num_tokens": 2130364.0,
+      "step": 1195
+    },
+    {
+      "entropy": 6.704743337631226,
+      "epoch": 0.061723632435768845,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.0004951320775200756,
+      "loss": 6.4765,
+      "mean_token_accuracy": 0.11397643610835076,
+      "num_tokens": 2138893.0,
+      "step": 1200
+    },
+    {
+      "entropy": 6.7864457130432125,
+      "epoch": 0.06198081423758455,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004948853278028436,
+      "loss": 6.4775,
+      "mean_token_accuracy": 0.11080277115106582,
+      "num_tokens": 2148441.0,
+      "step": 1205
+    },
+    {
+      "entropy": 6.467393255233764,
+      "epoch": 0.06223799603940025,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004946325498182755,
+      "loss": 6.3488,
+      "mean_token_accuracy": 0.11968965232372283,
+      "num_tokens": 2157566.0,
+      "step": 1210
+    },
+    {
+      "entropy": 6.594685506820679,
+      "epoch": 0.06249517784121596,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004943737504964076,
+      "loss": 6.2783,
+      "mean_token_accuracy": 0.12351583763957023,
+      "num_tokens": 2166652.0,
+      "step": 1215
+    },
+    {
+      "entropy": 6.595351934432983,
+      "epoch": 0.06275235964303166,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.000494108936932354,
+      "loss": 6.3271,
+      "mean_token_accuracy": 0.11962940394878388,
+      "num_tokens": 2175786.0,
+      "step": 1220
+    },
+    {
+      "entropy": 6.599175357818604,
+      "epoch": 0.06300954144484737,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004938381163861124,
+      "loss": 6.3318,
+      "mean_token_accuracy": 0.11382425799965859,
+      "num_tokens": 2184282.0,
+      "step": 1225
+    },
+    {
+      "entropy": 6.65804877281189,
+      "epoch": 0.06326672324666306,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004935612962823645,
+      "loss": 6.2634,
+      "mean_token_accuracy": 0.1293702930212021,
+      "num_tokens": 2192588.0,
+      "step": 1230
+    },
+    {
+      "entropy": 6.570264053344727,
+      "epoch": 0.06352390504847877,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004932784842102739,
+      "loss": 6.224,
+      "mean_token_accuracy": 0.1252226300537586,
+      "num_tokens": 2201247.0,
+      "step": 1235
+    },
+    {
+      "entropy": 6.581180286407471,
+      "epoch": 0.06378108685029447,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004929896879232758,
+      "loss": 6.4,
+      "mean_token_accuracy": 0.11070948839187622,
+      "num_tokens": 2210096.0,
+      "step": 1240
+    },
+    {
+      "entropy": 6.548767185211181,
+      "epoch": 0.06403826865211018,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004926949153388668,
+      "loss": 6.2372,
+      "mean_token_accuracy": 0.12190242633223533,
+      "num_tokens": 2219221.0,
+      "step": 1245
+    },
+    {
+      "entropy": 6.613304138183594,
+      "epoch": 0.06429545045392589,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004923941745383859,
+      "loss": 6.3914,
+      "mean_token_accuracy": 0.11916265785694122,
+      "num_tokens": 2229094.0,
+      "step": 1250
+    },
+    {
+      "entropy": 6.622950220108033,
+      "epoch": 0.06455263225574158,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000492087473766794,
+      "loss": 6.2814,
+      "mean_token_accuracy": 0.12494359165430069,
+      "num_tokens": 2236964.0,
+      "step": 1255
+    },
+    {
+      "entropy": 6.639262056350708,
+      "epoch": 0.06480981405755729,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.000491774821432448,
+      "loss": 6.2964,
+      "mean_token_accuracy": 0.11379757001996041,
+      "num_tokens": 2247044.0,
+      "step": 1260
+    },
+    {
+      "entropy": 6.578648328781128,
+      "epoch": 0.06506699585937299,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004914562261068693,
+      "loss": 6.3464,
+      "mean_token_accuracy": 0.11853989958763123,
+      "num_tokens": 2256106.0,
+      "step": 1265
+    },
+    {
+      "entropy": 6.5819896221160885,
+      "epoch": 0.0653241776611887,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0004911316965245098,
+      "loss": 6.333,
+      "mean_token_accuracy": 0.1119878426194191,
+      "num_tokens": 2265432.0,
+      "step": 1270
+    },
+    {
+      "entropy": 6.685585260391235,
+      "epoch": 0.0655813594630044,
+      "grad_norm": 1.125,
+      "learning_rate": 0.000490801241582512,
+      "loss": 6.3091,
+      "mean_token_accuracy": 0.11969617679715157,
+      "num_tokens": 2273882.0,
+      "step": 1275
+    },
+    {
+      "entropy": 6.507035255432129,
+      "epoch": 0.0658385412648201,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000490464870340465,
+      "loss": 6.2311,
+      "mean_token_accuracy": 0.12235765084624291,
+      "num_tokens": 2282754.0,
+      "step": 1280
+    },
+    {
+      "entropy": 6.6120068550109865,
+      "epoch": 0.0660957230666358,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004901225920201563,
+      "loss": 6.2385,
+      "mean_token_accuracy": 0.11968952193856239,
+      "num_tokens": 2291637.0,
+      "step": 1285
+    },
+    {
+      "entropy": 6.5052814960479735,
+      "epoch": 0.06635290486845151,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000489774416005319,
+      "loss": 6.2752,
+      "mean_token_accuracy": 0.1284623309969902,
+      "num_tokens": 2300279.0,
+      "step": 1290
+    },
+    {
+      "entropy": 6.605392360687256,
+      "epoch": 0.06661008667026722,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.0004894203518413742,
+      "loss": 6.3063,
+      "mean_token_accuracy": 0.11685983091592789,
+      "num_tokens": 2309413.0,
+      "step": 1295
+    },
+    {
+      "entropy": 6.453005313873291,
+      "epoch": 0.06686726847208292,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004890604092351701,
+      "loss": 6.2339,
+      "mean_token_accuracy": 0.12537124752998352,
+      "num_tokens": 2318284.0,
+      "step": 1300
+    },
+    {
+      "entropy": 6.663999080657959,
+      "epoch": 0.06712445027389861,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000488694598054715,
+      "loss": 6.4158,
+      "mean_token_accuracy": 0.11596757918596268,
+      "num_tokens": 2327483.0,
+      "step": 1305
+    },
+    {
+      "entropy": 6.604641580581665,
+      "epoch": 0.06738163207571432,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004883229283289071,
+      "loss": 6.3298,
+      "mean_token_accuracy": 0.11663304418325424,
+      "num_tokens": 2336096.0,
+      "step": 1310
+    },
+    {
+      "entropy": 6.647005414962768,
+      "epoch": 0.06763881387753003,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00048794541024725993,
+      "loss": 6.2797,
+      "mean_token_accuracy": 0.11775869503617287,
+      "num_tokens": 2345393.0,
+      "step": 1315
+    },
+    {
+      "entropy": 6.554132890701294,
+      "epoch": 0.06789599567934573,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004875620541596221,
+      "loss": 6.2501,
+      "mean_token_accuracy": 0.12467899695038795,
+      "num_tokens": 2353824.0,
+      "step": 1320
+    },
+    {
+      "entropy": 6.465411710739136,
+      "epoch": 0.06815317748116144,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00048717287057589454,
+      "loss": 6.2802,
+      "mean_token_accuracy": 0.12414587661623955,
+      "num_tokens": 2361992.0,
+      "step": 1325
+    },
+    {
+      "entropy": 6.607723951339722,
+      "epoch": 0.06841035928297713,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0004867778701657417,
+      "loss": 6.1856,
+      "mean_token_accuracy": 0.12355614453554153,
+      "num_tokens": 2370849.0,
+      "step": 1330
+    },
+    {
+      "entropy": 6.542122459411621,
+      "epoch": 0.06866754108479284,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00048637706375829955,
+      "loss": 6.2951,
+      "mean_token_accuracy": 0.125213223695755,
+      "num_tokens": 2379460.0,
+      "step": 1335
+    },
+    {
+      "entropy": 6.524771022796631,
+      "epoch": 0.06892472288660854,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000485970462341878,
+      "loss": 6.1841,
+      "mean_token_accuracy": 0.12624969705939293,
+      "num_tokens": 2388236.0,
+      "step": 1340
+    },
+    {
+      "entropy": 6.408080244064331,
+      "epoch": 0.06918190468842425,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00048555807706366044,
+      "loss": 6.2347,
+      "mean_token_accuracy": 0.12203069031238556,
+      "num_tokens": 2397598.0,
+      "step": 1345
+    },
+    {
+      "entropy": 6.627772331237793,
+      "epoch": 0.06943908649023996,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00048513991922939756,
+      "loss": 6.2066,
+      "mean_token_accuracy": 0.12295120730996131,
+      "num_tokens": 2406097.0,
+      "step": 1350
+    },
+    {
+      "entropy": 6.502293634414673,
+      "epoch": 0.06969626829205565,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00048471600030309744,
+      "loss": 6.3098,
+      "mean_token_accuracy": 0.11426047384738922,
+      "num_tokens": 2414590.0,
+      "step": 1355
+    },
+    {
+      "entropy": 6.440834903717041,
+      "epoch": 0.06995345009387136,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00048428633190671186,
+      "loss": 6.1354,
+      "mean_token_accuracy": 0.1235704205930233,
+      "num_tokens": 2424100.0,
+      "step": 1360
+    },
+    {
+      "entropy": 6.563741016387939,
+      "epoch": 0.07021063189568706,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004838509258198167,
+      "loss": 6.2091,
+      "mean_token_accuracy": 0.121873689442873,
+      "num_tokens": 2432751.0,
+      "step": 1365
+    },
+    {
+      "entropy": 6.543551826477051,
+      "epoch": 0.07046781369750277,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00048340979397929,
+      "loss": 6.1861,
+      "mean_token_accuracy": 0.12034041285514832,
+      "num_tokens": 2442054.0,
+      "step": 1370
+    },
+    {
+      "entropy": 6.375408792495728,
+      "epoch": 0.07072499549931847,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00048296294847898386,
+      "loss": 6.1099,
+      "mean_token_accuracy": 0.12799823582172393,
+      "num_tokens": 2450711.0,
+      "step": 1375
+    },
+    {
+      "entropy": 6.405136442184448,
+      "epoch": 0.07098217730113417,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004825104015693934,
+      "loss": 6.1895,
+      "mean_token_accuracy": 0.12983948215842248,
+      "num_tokens": 2460203.0,
+      "step": 1380
+    },
+    {
+      "entropy": 6.558651971817016,
+      "epoch": 0.07123935910294987,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004820521656573208,
+      "loss": 6.2482,
+      "mean_token_accuracy": 0.12185781002044678,
+      "num_tokens": 2469656.0,
+      "step": 1385
+    },
+    {
+      "entropy": 6.4783261775970455,
+      "epoch": 0.07149654090476558,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00048158825330553505,
+      "loss": 6.1986,
+      "mean_token_accuracy": 0.12839986979961396,
+      "num_tokens": 2478359.0,
+      "step": 1390
+    },
+    {
+      "entropy": 6.440516662597656,
+      "epoch": 0.07175372270658129,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00048111867723242763,
+      "loss": 6.1319,
+      "mean_token_accuracy": 0.12275411933660507,
+      "num_tokens": 2487330.0,
+      "step": 1395
+    },
+    {
+      "entropy": 6.464969682693481,
+      "epoch": 0.07201090450839699,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004806434503116637,
+      "loss": 6.1196,
+      "mean_token_accuracy": 0.12776034623384475,
+      "num_tokens": 2495445.0,
+      "step": 1400
+    },
+    {
+      "entropy": 6.352128219604492,
+      "epoch": 0.07226808631021268,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004801625855718296,
+      "loss": 6.1221,
+      "mean_token_accuracy": 0.129915539175272,
+      "num_tokens": 2504061.0,
+      "step": 1405
+    },
+    {
+      "entropy": 6.479479789733887,
+      "epoch": 0.07252526811202839,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00047967609619607477,
+      "loss": 6.1634,
+      "mean_token_accuracy": 0.12523439154028893,
+      "num_tokens": 2513044.0,
+      "step": 1410
+    },
+    {
+      "entropy": 6.456380176544189,
+      "epoch": 0.0727824499138441,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004791839955217513,
+      "loss": 6.2452,
+      "mean_token_accuracy": 0.12148747369647026,
+      "num_tokens": 2522027.0,
+      "step": 1415
+    },
+    {
+      "entropy": 6.455992269515991,
+      "epoch": 0.0730396317156598,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00047868629704004786,
+      "loss": 6.1438,
+      "mean_token_accuracy": 0.12664651721715928,
+      "num_tokens": 2530888.0,
+      "step": 1420
+    },
+    {
+      "entropy": 6.309210348129272,
+      "epoch": 0.07329681351747551,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00047818301439561965,
+      "loss": 6.1205,
+      "mean_token_accuracy": 0.13230895176529883,
+      "num_tokens": 2540210.0,
+      "step": 1425
+    },
+    {
+      "entropy": 6.500010251998901,
+      "epoch": 0.0735539953192912,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00047767416138621454,
+      "loss": 6.1576,
+      "mean_token_accuracy": 0.13035310953855514,
+      "num_tokens": 2547943.0,
+      "step": 1430
+    },
+    {
+      "entropy": 6.4128905773162845,
+      "epoch": 0.07381117712110691,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.000477159751962295,
+      "loss": 6.1758,
+      "mean_token_accuracy": 0.12096773236989974,
+      "num_tokens": 2557174.0,
+      "step": 1435
+    },
+    {
+      "entropy": 6.467635154724121,
+      "epoch": 0.07406835892292261,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00047663980022665507,
+      "loss": 6.1847,
+      "mean_token_accuracy": 0.12423764616250992,
+      "num_tokens": 2566139.0,
+      "step": 1440
+    },
+    {
+      "entropy": 6.443244886398316,
+      "epoch": 0.07432554072473832,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00047611432043403437,
+      "loss": 6.1833,
+      "mean_token_accuracy": 0.12400569245219231,
+      "num_tokens": 2574701.0,
+      "step": 1445
+    },
+    {
+      "entropy": 6.48380708694458,
+      "epoch": 0.07458272252655403,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004755833269907267,
+      "loss": 6.1683,
+      "mean_token_accuracy": 0.13298066854476928,
+      "num_tokens": 2583065.0,
+      "step": 1450
+    },
+    {
+      "entropy": 6.337066030502319,
+      "epoch": 0.07483990432836972,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004750468344541857,
+      "loss": 6.0415,
+      "mean_token_accuracy": 0.13096242472529412,
+      "num_tokens": 2592619.0,
+      "step": 1455
+    },
+    {
+      "entropy": 6.387078046798706,
+      "epoch": 0.07509708613018543,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00047450485753262525,
+      "loss": 6.1234,
+      "mean_token_accuracy": 0.12428422570228577,
+      "num_tokens": 2601434.0,
+      "step": 1460
+    },
+    {
+      "entropy": 6.404777574539184,
+      "epoch": 0.07535426793200113,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00047395741108461633,
+      "loss": 6.247,
+      "mean_token_accuracy": 0.11993053182959557,
+      "num_tokens": 2610925.0,
+      "step": 1465
+    },
+    {
+      "entropy": 6.445322704315186,
+      "epoch": 0.07561144973381684,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00047340451011867985,
+      "loss": 6.2036,
+      "mean_token_accuracy": 0.1239316701889038,
+      "num_tokens": 2620275.0,
+      "step": 1470
+    },
+    {
+      "entropy": 6.519756078720093,
+      "epoch": 0.07586863153563254,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00047284616979287515,
+      "loss": 6.2099,
+      "mean_token_accuracy": 0.1233117938041687,
+      "num_tokens": 2629826.0,
+      "step": 1475
+    },
+    {
+      "entropy": 6.344198274612427,
+      "epoch": 0.07612581333744824,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00047228240541438433,
+      "loss": 5.9902,
+      "mean_token_accuracy": 0.13270971104502677,
+      "num_tokens": 2638680.0,
+      "step": 1480
+    },
+    {
+      "entropy": 6.357509803771973,
+      "epoch": 0.07638299513926394,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00047171323243909257,
+      "loss": 6.1679,
+      "mean_token_accuracy": 0.12576281726360322,
+      "num_tokens": 2647739.0,
+      "step": 1485
+    },
+    {
+      "entropy": 6.3423261642456055,
+      "epoch": 0.07664017694107965,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00047113866647116457,
+      "loss": 6.0689,
+      "mean_token_accuracy": 0.12693566605448722,
+      "num_tokens": 2656137.0,
+      "step": 1490
+    },
+    {
+      "entropy": 6.399060344696045,
+      "epoch": 0.07689735874289536,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004705587232626164,
+      "loss": 6.1868,
+      "mean_token_accuracy": 0.1253664128482342,
+      "num_tokens": 2665173.0,
+      "step": 1495
+    },
+    {
+      "entropy": 6.466083621978759,
+      "epoch": 0.07715454054471106,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00046997341871288424,
+      "loss": 6.1583,
+      "mean_token_accuracy": 0.13294297680258751,
+      "num_tokens": 2673574.0,
+      "step": 1500
+    },
+    {
+      "entropy": 6.281017255783081,
+      "epoch": 0.07741172234652675,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004693827688683879,
+      "loss": 6.0623,
+      "mean_token_accuracy": 0.1325296178460121,
+      "num_tokens": 2682550.0,
+      "step": 1505
+    },
+    {
+      "entropy": 6.448086404800415,
+      "epoch": 0.07766890414834246,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004687867899220914,
+      "loss": 6.1484,
+      "mean_token_accuracy": 0.1286767065525055,
+      "num_tokens": 2691473.0,
+      "step": 1510
+    },
+    {
+      "entropy": 6.355803918838501,
+      "epoch": 0.07792608595015817,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00046818549821305846,
+      "loss": 6.1298,
+      "mean_token_accuracy": 0.12500087693333625,
+      "num_tokens": 2700351.0,
+      "step": 1515
+    },
+    {
+      "entropy": 6.375854969024658,
+      "epoch": 0.07818326775197387,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00046757891022600494,
+      "loss": 6.134,
+      "mean_token_accuracy": 0.13383759930729866,
+      "num_tokens": 2709286.0,
+      "step": 1520
+    },
+    {
+      "entropy": 6.397713947296142,
+      "epoch": 0.07844044955378958,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004669670425908471,
+      "loss": 6.1319,
+      "mean_token_accuracy": 0.1305990234017372,
+      "num_tokens": 2718003.0,
+      "step": 1525
+    },
+    {
+      "entropy": 6.5097356796264645,
+      "epoch": 0.07869763135560527,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004663499120822451,
+      "loss": 6.1401,
+      "mean_token_accuracy": 0.13283091709017752,
+      "num_tokens": 2726488.0,
+      "step": 1530
+    },
+    {
+      "entropy": 6.442542362213135,
+      "epoch": 0.07895481315742098,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004657275356191437,
+      "loss": 6.2381,
+      "mean_token_accuracy": 0.12060219943523406,
+      "num_tokens": 2735183.0,
+      "step": 1535
+    },
+    {
+      "entropy": 6.431625938415527,
+      "epoch": 0.07921199495923668,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00046509993026430804,
+      "loss": 6.1846,
+      "mean_token_accuracy": 0.13017381876707076,
+      "num_tokens": 2744730.0,
+      "step": 1540
+    },
+    {
+      "entropy": 6.391227769851684,
+      "epoch": 0.07946917676105239,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004644671132238558,
+      "loss": 6.078,
+      "mean_token_accuracy": 0.13144444525241852,
+      "num_tokens": 2753148.0,
+      "step": 1545
+    },
+    {
+      "entropy": 6.3710565090179445,
+      "epoch": 0.0797263585628681,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00046382910184678585,
+      "loss": 6.1304,
+      "mean_token_accuracy": 0.12514668405056,
+      "num_tokens": 2762394.0,
+      "step": 1550
+    },
+    {
+      "entropy": 6.383265018463135,
+      "epoch": 0.07998354036468379,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004631859136245025,
+      "loss": 6.0604,
+      "mean_token_accuracy": 0.13881182968616484,
+      "num_tokens": 2770778.0,
+      "step": 1555
+    },
+    {
+      "entropy": 6.303916597366333,
+      "epoch": 0.0802407221664995,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004625375661903357,
+      "loss": 6.0143,
+      "mean_token_accuracy": 0.1306192882359028,
+      "num_tokens": 2779530.0,
+      "step": 1560
+    },
+    {
+      "entropy": 6.338508129119873,
+      "epoch": 0.0804979039683152,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00046188407731905787,
+      "loss": 6.0973,
+      "mean_token_accuracy": 0.12936475947499276,
+      "num_tokens": 2788587.0,
+      "step": 1565
+    },
+    {
+      "entropy": 6.286006164550781,
+      "epoch": 0.08075508577013091,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00046122546492639643,
+      "loss": 6.0359,
+      "mean_token_accuracy": 0.13100264444947243,
+      "num_tokens": 2797759.0,
+      "step": 1570
+    },
+    {
+      "entropy": 6.315873813629151,
+      "epoch": 0.08101226757194661,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.000460561747068543,
+      "loss": 5.9986,
+      "mean_token_accuracy": 0.13158408626914025,
+      "num_tokens": 2806030.0,
+      "step": 1575
+    },
+    {
+      "entropy": 6.338774108886719,
+      "epoch": 0.0812694493737623,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004598929419416578,
+      "loss": 6.0144,
+      "mean_token_accuracy": 0.13145786076784133,
+      "num_tokens": 2815051.0,
+      "step": 1580
+    },
+    {
+      "entropy": 6.2580437660217285,
+      "epoch": 0.08152663117557801,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00045921906788137123,
+      "loss": 6.0245,
+      "mean_token_accuracy": 0.1368385747075081,
+      "num_tokens": 2824019.0,
+      "step": 1585
+    },
+    {
+      "entropy": 6.473824882507325,
+      "epoch": 0.08178381297739372,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00045854014336228115,
+      "loss": 6.1763,
+      "mean_token_accuracy": 0.12589069753885268,
+      "num_tokens": 2832494.0,
+      "step": 1590
+    },
+    {
+      "entropy": 6.306181764602661,
+      "epoch": 0.08204099477920943,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00045785618699744615,
+      "loss": 6.0168,
+      "mean_token_accuracy": 0.13415324538946152,
+      "num_tokens": 2841314.0,
+      "step": 1595
+    },
+    {
+      "entropy": 6.2222880840301515,
+      "epoch": 0.08229817658102513,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00045716721753787543,
+      "loss": 5.9476,
+      "mean_token_accuracy": 0.13951284289360047,
+      "num_tokens": 2849635.0,
+      "step": 1600
+    },
+    {
+      "entropy": 6.297337532043457,
+      "epoch": 0.08255535838284082,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.0004564732538720148,
+      "loss": 6.094,
+      "mean_token_accuracy": 0.12806924805045128,
+      "num_tokens": 2859306.0,
+      "step": 1605
+    },
+    {
+      "entropy": 6.361379623413086,
+      "epoch": 0.08281254018465653,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00045577431502522877,
+      "loss": 6.1886,
+      "mean_token_accuracy": 0.1268698565661907,
+      "num_tokens": 2868316.0,
+      "step": 1610
+    },
+    {
+      "entropy": 6.415389966964722,
+      "epoch": 0.08306972198647224,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004550704201592787,
+      "loss": 6.1611,
+      "mean_token_accuracy": 0.12649422660470008,
+      "num_tokens": 2877508.0,
+      "step": 1615
+    },
+    {
+      "entropy": 6.425594854354858,
+      "epoch": 0.08332690378828794,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0004543615885717981,
+      "loss": 6.1123,
+      "mean_token_accuracy": 0.1241836428642273,
+      "num_tokens": 2886927.0,
+      "step": 1620
+    },
+    {
+      "entropy": 6.364439010620117,
+      "epoch": 0.08358408559010365,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00045364783969576296,
+      "loss": 6.1434,
+      "mean_token_accuracy": 0.12308017164468765,
+      "num_tokens": 2896484.0,
+      "step": 1625
+    },
+    {
+      "entropy": 6.411374950408936,
+      "epoch": 0.08384126739191934,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004529291930989592,
+      "loss": 6.0181,
+      "mean_token_accuracy": 0.13011238351464272,
+      "num_tokens": 2905298.0,
+      "step": 1630
+    },
+    {
+      "entropy": 6.292522811889649,
+      "epoch": 0.08409844919373505,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004522056684834464,
+      "loss": 6.0992,
+      "mean_token_accuracy": 0.1334770753979683,
+      "num_tokens": 2913811.0,
+      "step": 1635
+    },
+    {
+      "entropy": 6.370844841003418,
+      "epoch": 0.08435563099555075,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004514772856850173,
+      "loss": 6.0902,
+      "mean_token_accuracy": 0.128910331428051,
+      "num_tokens": 2922531.0,
+      "step": 1640
+    },
+    {
+      "entropy": 6.311898422241211,
+      "epoch": 0.08461281279736646,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004507440646726542,
+      "loss": 6.0567,
+      "mean_token_accuracy": 0.13458428382873536,
+      "num_tokens": 2931178.0,
+      "step": 1645
+    },
+    {
+      "entropy": 6.309841632843018,
+      "epoch": 0.08486999459918217,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004500060255479818,
+      "loss": 6.0937,
+      "mean_token_accuracy": 0.13025279939174653,
+      "num_tokens": 2940419.0,
+      "step": 1650
+    },
+    {
+      "entropy": 6.36889328956604,
+      "epoch": 0.08512717640099786,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004492631885447151,
+      "loss": 5.9964,
+      "mean_token_accuracy": 0.1294368840754032,
+      "num_tokens": 2949119.0,
+      "step": 1655
+    },
+    {
+      "entropy": 6.307538890838623,
+      "epoch": 0.08538435820281357,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00044851557402810616,
+      "loss": 6.1656,
+      "mean_token_accuracy": 0.12390446290373802,
+      "num_tokens": 2957775.0,
+      "step": 1660
+    },
+    {
+      "entropy": 6.376241397857666,
+      "epoch": 0.08564154000462927,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00044776320249438444,
+      "loss": 6.0929,
+      "mean_token_accuracy": 0.1286988750100136,
+      "num_tokens": 2967644.0,
+      "step": 1665
+    },
+    {
+      "entropy": 6.352219152450561,
+      "epoch": 0.08589872180644498,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00044700609457019565,
+      "loss": 6.0603,
+      "mean_token_accuracy": 0.13165615424513816,
+      "num_tokens": 2975608.0,
+      "step": 1670
+    },
+    {
+      "entropy": 6.289653968811035,
+      "epoch": 0.08615590360826068,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004462442710120359,
+      "loss": 6.0552,
+      "mean_token_accuracy": 0.1328391693532467,
+      "num_tokens": 2984138.0,
+      "step": 1675
+    },
+    {
+      "entropy": 6.410796213150024,
+      "epoch": 0.08641308541007638,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000445477752705683,
+      "loss": 6.1551,
+      "mean_token_accuracy": 0.12801553606986998,
+      "num_tokens": 2992770.0,
+      "step": 1680
+    },
+    {
+      "entropy": 6.379126834869385,
+      "epoch": 0.08667026721189208,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00044470656066562336,
+      "loss": 5.9911,
+      "mean_token_accuracy": 0.1325019121170044,
+      "num_tokens": 3001685.0,
+      "step": 1685
+    },
+    {
+      "entropy": 6.266774129867554,
+      "epoch": 0.08692744901370779,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004439307160344765,
+      "loss": 6.0246,
+      "mean_token_accuracy": 0.13351587057113648,
+      "num_tokens": 3010905.0,
+      "step": 1690
+    },
+    {
+      "entropy": 6.312497520446778,
+      "epoch": 0.0871846308155235,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00044315024008241473,
+      "loss": 6.0822,
+      "mean_token_accuracy": 0.12986138314008713,
+      "num_tokens": 3020257.0,
+      "step": 1695
+    },
+    {
+      "entropy": 6.329559803009033,
+      "epoch": 0.0874418126173392,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004423651542065806,
+      "loss": 6.086,
+      "mean_token_accuracy": 0.1287504680454731,
+      "num_tokens": 3029397.0,
+      "step": 1700
+    },
+    {
+      "entropy": 6.40656681060791,
+      "epoch": 0.0876989944191549,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00044157547993050006,
+      "loss": 6.0878,
+      "mean_token_accuracy": 0.12193093374371529,
+      "num_tokens": 3038822.0,
+      "step": 1705
+    },
+    {
+      "entropy": 6.286016368865967,
+      "epoch": 0.0879561762209706,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00044078123890349227,
+      "loss": 5.9675,
+      "mean_token_accuracy": 0.13392331525683404,
+      "num_tokens": 3047435.0,
+      "step": 1710
+    },
+    {
+      "entropy": 6.301493215560913,
+      "epoch": 0.0882133580227863,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00043998245290007606,
+      "loss": 6.1045,
+      "mean_token_accuracy": 0.13060200214385986,
+      "num_tokens": 3056011.0,
+      "step": 1715
+    },
+    {
+      "entropy": 6.378498220443726,
+      "epoch": 0.08847053982460201,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00043917914381937323,
+      "loss": 6.0839,
+      "mean_token_accuracy": 0.1344021290540695,
+      "num_tokens": 3064182.0,
+      "step": 1720
+    },
+    {
+      "entropy": 6.304371786117554,
+      "epoch": 0.08872772162641772,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00043837133368450815,
+      "loss": 6.0605,
+      "mean_token_accuracy": 0.13042551502585412,
+      "num_tokens": 3072669.0,
+      "step": 1725
+    },
+    {
+      "entropy": 6.319208002090454,
+      "epoch": 0.08898490342823343,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004375590446420037,
+      "loss": 6.0262,
+      "mean_token_accuracy": 0.13751724287867545,
+      "num_tokens": 3081024.0,
+      "step": 1730
+    },
+    {
+      "entropy": 6.245740699768066,
+      "epoch": 0.08924208523004912,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004367422989611743,
+      "loss": 5.9442,
+      "mean_token_accuracy": 0.13434263542294503,
+      "num_tokens": 3089747.0,
+      "step": 1735
+    },
+    {
+      "entropy": 6.31815619468689,
+      "epoch": 0.08949926703186482,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.0004359211190335153,
+      "loss": 6.0425,
+      "mean_token_accuracy": 0.13445181101560594,
+      "num_tokens": 3099011.0,
+      "step": 1740
+    },
+    {
+      "entropy": 6.286892604827881,
+      "epoch": 0.08975644883368053,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00043509552737208923,
+      "loss": 5.9967,
+      "mean_token_accuracy": 0.13267636224627494,
+      "num_tokens": 3107645.0,
+      "step": 1745
+    },
+    {
+      "entropy": 6.278048324584961,
+      "epoch": 0.09001363063549624,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00043426554661090853,
+      "loss": 6.0557,
+      "mean_token_accuracy": 0.1316031478345394,
+      "num_tokens": 3116292.0,
+      "step": 1750
+    },
+    {
+      "entropy": 6.359972858428955,
+      "epoch": 0.09027081243731194,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00043343119950431516,
+      "loss": 6.0723,
+      "mean_token_accuracy": 0.13226044401526452,
+      "num_tokens": 3126356.0,
+      "step": 1755
+    },
+    {
+      "entropy": 6.250633955001831,
+      "epoch": 0.09052799423912763,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00043259250892635644,
+      "loss": 6.0472,
+      "mean_token_accuracy": 0.12974994629621506,
+      "num_tokens": 3135685.0,
+      "step": 1760
+    },
+    {
+      "entropy": 6.314090394973755,
+      "epoch": 0.09078517604094334,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004317494978701582,
+      "loss": 6.0026,
+      "mean_token_accuracy": 0.12911238446831702,
+      "num_tokens": 3144788.0,
+      "step": 1765
+    },
+    {
+      "entropy": 6.288527154922486,
+      "epoch": 0.09104235784275905,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004309021894472943,
+      "loss": 6.0151,
+      "mean_token_accuracy": 0.1408878244459629,
+      "num_tokens": 3153273.0,
+      "step": 1770
+    },
+    {
+      "entropy": 6.267010736465454,
+      "epoch": 0.09129953964457475,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004300506068871534,
+      "loss": 6.0249,
+      "mean_token_accuracy": 0.1345413491129875,
+      "num_tokens": 3161806.0,
+      "step": 1775
+    },
+    {
+      "entropy": 6.174133968353272,
+      "epoch": 0.09155672144639046,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00042919477353630135,
+      "loss": 5.8371,
+      "mean_token_accuracy": 0.14690690338611603,
+      "num_tokens": 3170401.0,
+      "step": 1780
+    },
+    {
+      "entropy": 6.2911989212036135,
+      "epoch": 0.09181390324820615,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.000428334712857842,
+      "loss": 5.9583,
+      "mean_token_accuracy": 0.13602055236697197,
+      "num_tokens": 3179742.0,
+      "step": 1785
+    },
+    {
+      "entropy": 6.309875202178955,
+      "epoch": 0.09207108505002186,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00042747044843077304,
+      "loss": 6.1082,
+      "mean_token_accuracy": 0.13044340386986733,
+      "num_tokens": 3188898.0,
+      "step": 1790
+    },
+    {
+      "entropy": 6.30495662689209,
+      "epoch": 0.09232826685183756,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00042660200394934047,
+      "loss": 6.0906,
+      "mean_token_accuracy": 0.12872936055064202,
+      "num_tokens": 3197860.0,
+      "step": 1795
+    },
+    {
+      "entropy": 6.351209688186645,
+      "epoch": 0.09258544865365327,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00042572940322238844,
+      "loss": 6.0633,
+      "mean_token_accuracy": 0.133078370988369,
+      "num_tokens": 3206779.0,
+      "step": 1800
+    },
+    {
+      "entropy": 6.252921724319458,
+      "epoch": 0.09284263045546898,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.00042485267017270664,
+      "loss": 5.9835,
+      "mean_token_accuracy": 0.1363011360168457,
+      "num_tokens": 3215051.0,
+      "step": 1805
+    },
+    {
+      "entropy": 6.277395391464234,
+      "epoch": 0.09309981225728467,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004239718288363745,
+      "loss": 5.9666,
+      "mean_token_accuracy": 0.1378926604986191,
+      "num_tokens": 3223977.0,
+      "step": 1810
+    },
+    {
+      "entropy": 6.305054616928101,
+      "epoch": 0.09335699405910038,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004230869033621023,
+      "loss": 5.9981,
+      "mean_token_accuracy": 0.1367839142680168,
+      "num_tokens": 3232714.0,
+      "step": 1815
+    },
+    {
+      "entropy": 6.307405042648315,
+      "epoch": 0.09361417586091608,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004221979180105688,
+      "loss": 5.9666,
+      "mean_token_accuracy": 0.13633009940385818,
+      "num_tokens": 3241405.0,
+      "step": 1820
+    },
+    {
+      "entropy": 6.269251775741577,
+      "epoch": 0.09387135766273179,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00042130489715375645,
+      "loss": 6.0106,
+      "mean_token_accuracy": 0.13501947596669198,
+      "num_tokens": 3249846.0,
+      "step": 1825
+    },
+    {
+      "entropy": 6.296307373046875,
+      "epoch": 0.0941285394645475,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00042040786527428335,
+      "loss": 5.9188,
+      "mean_token_accuracy": 0.13773071020841599,
+      "num_tokens": 3258231.0,
+      "step": 1830
+    },
+    {
+      "entropy": 6.150211763381958,
+      "epoch": 0.09438572126636319,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004195068469647315,
+      "loss": 5.9593,
+      "mean_token_accuracy": 0.1389425627887249,
+      "num_tokens": 3267333.0,
+      "step": 1835
+    },
+    {
+      "entropy": 6.298181486129761,
+      "epoch": 0.0946429030681789,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00041860186692697297,
+      "loss": 6.1112,
+      "mean_token_accuracy": 0.12236591503024101,
+      "num_tokens": 3276575.0,
+      "step": 1840
+    },
+    {
+      "entropy": 6.347346067428589,
+      "epoch": 0.0949000848699946,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00041769294997149264,
+      "loss": 6.1014,
+      "mean_token_accuracy": 0.12974566370248794,
+      "num_tokens": 3284771.0,
+      "step": 1845
+    },
+    {
+      "entropy": 6.261059379577636,
+      "epoch": 0.0951572666718103,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.0004167801210167081,
+      "loss": 6.0138,
+      "mean_token_accuracy": 0.13305832222104072,
+      "num_tokens": 3294091.0,
+      "step": 1850
+    },
+    {
+      "entropy": 6.248409795761108,
+      "epoch": 0.09541444847362601,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004158634050882861,
+      "loss": 5.8796,
+      "mean_token_accuracy": 0.13779047578573228,
+      "num_tokens": 3302572.0,
+      "step": 1855
+    },
+    {
+      "entropy": 6.177290248870849,
+      "epoch": 0.0956716302754417,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004149428273184569,
+      "loss": 5.9015,
+      "mean_token_accuracy": 0.14134511575102807,
+      "num_tokens": 3310709.0,
+      "step": 1860
+    },
+    {
+      "entropy": 6.239241313934326,
+      "epoch": 0.09592881207725741,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004140184129453253,
+      "loss": 5.9881,
+      "mean_token_accuracy": 0.13656110763549806,
+      "num_tokens": 3319959.0,
+      "step": 1865
+    },
+    {
+      "entropy": 6.237763357162476,
+      "epoch": 0.09618599387907312,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.000413090187312178,
+      "loss": 5.9338,
+      "mean_token_accuracy": 0.1348011650145054,
+      "num_tokens": 3328475.0,
+      "step": 1870
+    },
+    {
+      "entropy": 6.189648580551148,
+      "epoch": 0.09644317568088882,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.0004121581758667898,
+      "loss": 6.0114,
+      "mean_token_accuracy": 0.1303848333656788,
+      "num_tokens": 3337785.0,
+      "step": 1875
+    },
+    {
+      "entropy": 6.246728134155274,
+      "epoch": 0.09670035748270453,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.00041122240416072533,
+      "loss": 5.9514,
+      "mean_token_accuracy": 0.13871703594923018,
+      "num_tokens": 3347102.0,
+      "step": 1880
+    },
+    {
+      "entropy": 6.295105886459351,
+      "epoch": 0.09695753928452022,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004102828978486385,
+      "loss": 5.9854,
+      "mean_token_accuracy": 0.13081275448203086,
+      "num_tokens": 3356787.0,
+      "step": 1885
+    },
+    {
+      "entropy": 6.240402698516846,
+      "epoch": 0.09721472108633593,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004093396826875695,
+      "loss": 6.058,
+      "mean_token_accuracy": 0.1320380300283432,
+      "num_tokens": 3365497.0,
+      "step": 1890
+    },
+    {
+      "entropy": 6.231711196899414,
+      "epoch": 0.09747190288815163,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00040839278453623837,
+      "loss": 6.0093,
+      "mean_token_accuracy": 0.13135660141706468,
+      "num_tokens": 3373622.0,
+      "step": 1895
+    },
+    {
+      "entropy": 6.235841464996338,
+      "epoch": 0.09772908468996734,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0004074422293543363,
+      "loss": 5.9967,
+      "mean_token_accuracy": 0.13504694923758506,
+      "num_tokens": 3382563.0,
+      "step": 1900
+    },
+    {
+      "entropy": 6.295864820480347,
+      "epoch": 0.09798626649178305,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004064880432018137,
+      "loss": 5.8712,
+      "mean_token_accuracy": 0.14425868391990662,
+      "num_tokens": 3390686.0,
+      "step": 1905
+    },
+    {
+      "entropy": 6.186135339736938,
+      "epoch": 0.09824344829359874,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00040553025223816615,
+      "loss": 5.9351,
+      "mean_token_accuracy": 0.13617005720734596,
+      "num_tokens": 3399616.0,
+      "step": 1910
+    },
+    {
+      "entropy": 6.16251573562622,
+      "epoch": 0.09850063009541445,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00040456888272171653,
+      "loss": 5.834,
+      "mean_token_accuracy": 0.1381289020180702,
+      "num_tokens": 3407572.0,
+      "step": 1915
+    },
+    {
+      "entropy": 6.25083360671997,
+      "epoch": 0.09875781189723015,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00040360396100889577,
+      "loss": 5.9848,
+      "mean_token_accuracy": 0.13736421391367912,
+      "num_tokens": 3416174.0,
+      "step": 1920
+    },
+    {
+      "entropy": 6.121949052810669,
+      "epoch": 0.09901499369904586,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.0004026355135535202,
+      "loss": 5.844,
+      "mean_token_accuracy": 0.1408976547420025,
+      "num_tokens": 3425115.0,
+      "step": 1925
+    },
+    {
+      "entropy": 6.129916381835938,
+      "epoch": 0.09927217550086156,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.000401663566906066,
+      "loss": 5.8997,
+      "mean_token_accuracy": 0.14465468153357505,
+      "num_tokens": 3433589.0,
+      "step": 1930
+    },
+    {
+      "entropy": 6.251618909835815,
+      "epoch": 0.09952935730267726,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00040068814771294134,
+      "loss": 5.9948,
+      "mean_token_accuracy": 0.14055624082684517,
+      "num_tokens": 3443205.0,
+      "step": 1935
+    },
+    {
+      "entropy": 6.300161504745484,
+      "epoch": 0.09978653910449296,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0003997092827157562,
+      "loss": 5.9464,
+      "mean_token_accuracy": 0.136979528516531,
+      "num_tokens": 3451738.0,
+      "step": 1940
+    },
+    {
+      "entropy": 6.163929224014282,
+      "epoch": 0.10004372090630867,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000398726998750589,
+      "loss": 5.9042,
+      "mean_token_accuracy": 0.1414563961327076,
+      "num_tokens": 3461201.0,
+      "step": 1945
+    },
+    {
+      "entropy": 6.146855640411377,
+      "epoch": 0.10030090270812438,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00039774132274725076,
+      "loss": 5.9047,
+      "mean_token_accuracy": 0.1325928770005703,
+      "num_tokens": 3470788.0,
+      "step": 1950
+    },
+    {
+      "entropy": 6.3281749248504635,
+      "epoch": 0.10055808450994008,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00039675228172854707,
+      "loss": 6.0299,
+      "mean_token_accuracy": 0.13255303725600243,
+      "num_tokens": 3480143.0,
+      "step": 1955
+    },
+    {
+      "entropy": 6.215788984298706,
+      "epoch": 0.10081526631175577,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0003957599028095371,
+      "loss": 5.9038,
+      "mean_token_accuracy": 0.14231238663196563,
+      "num_tokens": 3488217.0,
+      "step": 1960
+    },
+    {
+      "entropy": 6.2849045753479,
+      "epoch": 0.10107244811357148,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00039476421319679017,
+      "loss": 6.0992,
+      "mean_token_accuracy": 0.1307053431868553,
+      "num_tokens": 3496960.0,
+      "step": 1965
+    },
+    {
+      "entropy": 6.365100383758545,
+      "epoch": 0.10132962991538719,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00039376524018764,
+      "loss": 6.0058,
+      "mean_token_accuracy": 0.1358587734401226,
+      "num_tokens": 3506314.0,
+      "step": 1970
+    },
+    {
+      "entropy": 6.257384538650513,
+      "epoch": 0.1015868117172029,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00039276301116943616,
+      "loss": 5.9707,
+      "mean_token_accuracy": 0.14363735169172287,
+      "num_tokens": 3514425.0,
+      "step": 1975
+    },
+    {
+      "entropy": 6.19381799697876,
+      "epoch": 0.1018439935190186,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0003917575536187936,
+      "loss": 5.9653,
+      "mean_token_accuracy": 0.1379447251558304,
+      "num_tokens": 3524153.0,
+      "step": 1980
+    },
+    {
+      "entropy": 6.209758520126343,
+      "epoch": 0.10210117532083429,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00039074889510083894,
+      "loss": 5.8675,
+      "mean_token_accuracy": 0.14806670546531678,
+      "num_tokens": 3532252.0,
+      "step": 1985
+    },
+    {
+      "entropy": 6.207709884643554,
+      "epoch": 0.10235835712265,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00038973706326845495,
+      "loss": 5.9073,
+      "mean_token_accuracy": 0.13528766930103303,
+      "num_tokens": 3540689.0,
+      "step": 1990
+    },
+    {
+      "entropy": 6.209512042999267,
+      "epoch": 0.1026155389244657,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003887220858615225,
+      "loss": 5.9125,
+      "mean_token_accuracy": 0.14427549242973328,
+      "num_tokens": 3549110.0,
+      "step": 1995
+    },
+    {
+      "entropy": 6.194009447097779,
+      "epoch": 0.10287272072628141,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0003877039907061597,
+      "loss": 5.8742,
+      "mean_token_accuracy": 0.14361105114221573,
+      "num_tokens": 3558252.0,
+      "step": 2000
+    }
+  ],
+  "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": 706424604917760.0,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}