diff --git "a/checkpoint-2500/trainer_state.json" "b/checkpoint-2500/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-2500/trainer_state.json"
@@ -0,0 +1,5034 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.0676818950930626,
+  "eval_steps": 500,
+  "global_step": 2500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 10.69194974899292,
+      "epoch": 0.00013536379018612522,
+      "grad_norm": 12.5,
+      "learning_rate": 2e-06,
+      "loss": 10.8777,
+      "mean_token_accuracy": 5.577245028689504e-05,
+      "num_tokens": 15054.0,
+      "step": 5
+    },
+    {
+      "entropy": 10.691948223114014,
+      "epoch": 0.00027072758037225043,
+      "grad_norm": 12.0625,
+      "learning_rate": 4.5e-06,
+      "loss": 10.8153,
+      "mean_token_accuracy": 0.00044745010090991856,
+      "num_tokens": 28250.0,
+      "step": 10
+    },
+    {
+      "entropy": 10.691227531433105,
+      "epoch": 0.0004060913705583756,
+      "grad_norm": 9.1875,
+      "learning_rate": 7e-06,
+      "loss": 10.5817,
+      "mean_token_accuracy": 0.0174422716954723,
+      "num_tokens": 41527.0,
+      "step": 15
+    },
+    {
+      "entropy": 10.685559177398682,
+      "epoch": 0.0005414551607445009,
+      "grad_norm": 5.9375,
+      "learning_rate": 9.5e-06,
+      "loss": 10.2793,
+      "mean_token_accuracy": 0.034390256740152836,
+      "num_tokens": 54285.0,
+      "step": 20
+    },
+    {
+      "entropy": 10.661216259002686,
+      "epoch": 0.0006768189509306261,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.2e-05,
+      "loss": 10.0155,
+      "mean_token_accuracy": 0.031836986914277075,
+      "num_tokens": 69833.0,
+      "step": 25
+    },
+    {
+      "entropy": 10.635576629638672,
+      "epoch": 0.0008121827411167512,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.4500000000000002e-05,
+      "loss": 9.8988,
+      "mean_token_accuracy": 0.033391524106264114,
+      "num_tokens": 84065.0,
+      "step": 30
+    },
+    {
+      "entropy": 10.61604061126709,
+      "epoch": 0.0009475465313028764,
+      "grad_norm": 2.375,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 9.7824,
+      "mean_token_accuracy": 0.035429037548601626,
+      "num_tokens": 97347.0,
+      "step": 35
+    },
+    {
+      "entropy": 10.616745281219483,
+      "epoch": 0.0010829103214890017,
+      "grad_norm": 2.5,
+      "learning_rate": 1.95e-05,
+      "loss": 9.722,
+      "mean_token_accuracy": 0.034273325465619564,
+      "num_tokens": 110883.0,
+      "step": 40
+    },
+    {
+      "entropy": 10.620021438598632,
+      "epoch": 0.001218274111675127,
+      "grad_norm": 2.1875,
+      "learning_rate": 2.2e-05,
+      "loss": 9.7299,
+      "mean_token_accuracy": 0.031383275240659717,
+      "num_tokens": 125407.0,
+      "step": 45
+    },
+    {
+      "entropy": 10.61033935546875,
+      "epoch": 0.0013536379018612521,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.4500000000000003e-05,
+      "loss": 9.6781,
+      "mean_token_accuracy": 0.03162609338760376,
+      "num_tokens": 138221.0,
+      "step": 50
+    },
+    {
+      "entropy": 10.609558010101319,
+      "epoch": 0.0014890016920473773,
+      "grad_norm": 2.21875,
+      "learning_rate": 2.7e-05,
+      "loss": 9.6075,
+      "mean_token_accuracy": 0.037006295286118984,
+      "num_tokens": 150716.0,
+      "step": 55
+    },
+    {
+      "entropy": 10.592256355285645,
+      "epoch": 0.0016243654822335025,
+      "grad_norm": 2.15625,
+      "learning_rate": 2.95e-05,
+      "loss": 9.5363,
+      "mean_token_accuracy": 0.033432438410818575,
+      "num_tokens": 167084.0,
+      "step": 60
+    },
+    {
+      "entropy": 10.596331214904785,
+      "epoch": 0.0017597292724196277,
+      "grad_norm": 2.125,
+      "learning_rate": 3.2e-05,
+      "loss": 9.5357,
+      "mean_token_accuracy": 0.028936176747083663,
+      "num_tokens": 181029.0,
+      "step": 65
+    },
+    {
+      "entropy": 10.597533226013184,
+      "epoch": 0.0018950930626057529,
+      "grad_norm": 2.09375,
+      "learning_rate": 3.4500000000000005e-05,
+      "loss": 9.3621,
+      "mean_token_accuracy": 0.03304864317178726,
+      "num_tokens": 194629.0,
+      "step": 70
+    },
+    {
+      "entropy": 10.554742336273193,
+      "epoch": 0.0020304568527918783,
+      "grad_norm": 2.0625,
+      "learning_rate": 3.7e-05,
+      "loss": 9.2827,
+      "mean_token_accuracy": 0.029723377712070943,
+      "num_tokens": 207367.0,
+      "step": 75
+    },
+    {
+      "entropy": 10.533198165893555,
+      "epoch": 0.0021658206429780035,
+      "grad_norm": 1.9453125,
+      "learning_rate": 3.95e-05,
+      "loss": 9.1578,
+      "mean_token_accuracy": 0.029196519777178763,
+      "num_tokens": 221208.0,
+      "step": 80
+    },
+    {
+      "entropy": 10.477063274383545,
+      "epoch": 0.0023011844331641287,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.2000000000000004e-05,
+      "loss": 9.0349,
+      "mean_token_accuracy": 0.03920750468969345,
+      "num_tokens": 233176.0,
+      "step": 85
+    },
+    {
+      "entropy": 10.465512084960938,
+      "epoch": 0.002436548223350254,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.45e-05,
+      "loss": 8.9479,
+      "mean_token_accuracy": 0.037744084000587465,
+      "num_tokens": 247745.0,
+      "step": 90
+    },
+    {
+      "entropy": 10.410159492492676,
+      "epoch": 0.002571912013536379,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.7000000000000004e-05,
+      "loss": 8.8433,
+      "mean_token_accuracy": 0.037843816168606284,
+      "num_tokens": 262129.0,
+      "step": 95
+    },
+    {
+      "entropy": 10.357386016845703,
+      "epoch": 0.0027072758037225042,
+      "grad_norm": 2.03125,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 8.7074,
+      "mean_token_accuracy": 0.042482448555529115,
+      "num_tokens": 275500.0,
+      "step": 100
+    },
+    {
+      "entropy": 10.285986518859863,
+      "epoch": 0.0028426395939086294,
+      "grad_norm": 1.875,
+      "learning_rate": 5.2e-05,
+      "loss": 8.6332,
+      "mean_token_accuracy": 0.04186442382633686,
+      "num_tokens": 290437.0,
+      "step": 105
+    },
+    {
+      "entropy": 10.227664184570312,
+      "epoch": 0.0029780033840947546,
+      "grad_norm": 1.90625,
+      "learning_rate": 5.45e-05,
+      "loss": 8.4353,
+      "mean_token_accuracy": 0.0504638671875,
+      "num_tokens": 304615.0,
+      "step": 110
+    },
+    {
+      "entropy": 10.08832664489746,
+      "epoch": 0.00311336717428088,
+      "grad_norm": 1.8671875,
+      "learning_rate": 5.7e-05,
+      "loss": 8.263,
+      "mean_token_accuracy": 0.05822908580303192,
+      "num_tokens": 318560.0,
+      "step": 115
+    },
+    {
+      "entropy": 9.972474098205566,
+      "epoch": 0.003248730964467005,
+      "grad_norm": 2.03125,
+      "learning_rate": 5.9499999999999996e-05,
+      "loss": 8.1931,
+      "mean_token_accuracy": 0.05676005966961384,
+      "num_tokens": 332748.0,
+      "step": 120
+    },
+    {
+      "entropy": 9.847494888305665,
+      "epoch": 0.00338409475465313,
+      "grad_norm": 2.125,
+      "learning_rate": 6.2e-05,
+      "loss": 7.9947,
+      "mean_token_accuracy": 0.06220810636878014,
+      "num_tokens": 347127.0,
+      "step": 125
+    },
+    {
+      "entropy": 9.678038311004638,
+      "epoch": 0.0035194585448392554,
+      "grad_norm": 1.6171875,
+      "learning_rate": 6.450000000000001e-05,
+      "loss": 7.9391,
+      "mean_token_accuracy": 0.058373385295271876,
+      "num_tokens": 359407.0,
+      "step": 130
+    },
+    {
+      "entropy": 9.540247344970703,
+      "epoch": 0.0036548223350253805,
+      "grad_norm": 1.4375,
+      "learning_rate": 6.7e-05,
+      "loss": 7.7636,
+      "mean_token_accuracy": 0.061562583222985266,
+      "num_tokens": 375021.0,
+      "step": 135
+    },
+    {
+      "entropy": 9.30407886505127,
+      "epoch": 0.0037901861252115057,
+      "grad_norm": 1.484375,
+      "learning_rate": 6.950000000000001e-05,
+      "loss": 7.5962,
+      "mean_token_accuracy": 0.070381173864007,
+      "num_tokens": 389496.0,
+      "step": 140
+    },
+    {
+      "entropy": 9.132857418060302,
+      "epoch": 0.003925549915397631,
+      "grad_norm": 1.5703125,
+      "learning_rate": 7.2e-05,
+      "loss": 7.6112,
+      "mean_token_accuracy": 0.06686735302209854,
+      "num_tokens": 402465.0,
+      "step": 145
+    },
+    {
+      "entropy": 8.873841571807862,
+      "epoch": 0.0040609137055837565,
+      "grad_norm": 1.515625,
+      "learning_rate": 7.45e-05,
+      "loss": 7.502,
+      "mean_token_accuracy": 0.0693404596298933,
+      "num_tokens": 415980.0,
+      "step": 150
+    },
+    {
+      "entropy": 8.732354259490966,
+      "epoch": 0.004196277495769882,
+      "grad_norm": 1.3515625,
+      "learning_rate": 7.7e-05,
+      "loss": 7.4634,
+      "mean_token_accuracy": 0.07001067139208317,
+      "num_tokens": 428623.0,
+      "step": 155
+    },
+    {
+      "entropy": 8.544918155670166,
+      "epoch": 0.004331641285956007,
+      "grad_norm": 1.5859375,
+      "learning_rate": 7.950000000000001e-05,
+      "loss": 7.4298,
+      "mean_token_accuracy": 0.06541897803544998,
+      "num_tokens": 441999.0,
+      "step": 160
+    },
+    {
+      "entropy": 8.378799533843994,
+      "epoch": 0.004467005076142132,
+      "grad_norm": 1.5,
+      "learning_rate": 8.2e-05,
+      "loss": 7.3005,
+      "mean_token_accuracy": 0.07170183844864368,
+      "num_tokens": 455098.0,
+      "step": 165
+    },
+    {
+      "entropy": 8.280582523345947,
+      "epoch": 0.004602368866328257,
+      "grad_norm": 1.9375,
+      "learning_rate": 8.450000000000001e-05,
+      "loss": 7.2898,
+      "mean_token_accuracy": 0.07008971199393273,
+      "num_tokens": 468506.0,
+      "step": 170
+    },
+    {
+      "entropy": 8.127180671691894,
+      "epoch": 0.0047377326565143825,
+      "grad_norm": 1.421875,
+      "learning_rate": 8.7e-05,
+      "loss": 7.3215,
+      "mean_token_accuracy": 0.06721526682376862,
+      "num_tokens": 482795.0,
+      "step": 175
+    },
+    {
+      "entropy": 8.071395254135131,
+      "epoch": 0.004873096446700508,
+      "grad_norm": 1.7109375,
+      "learning_rate": 8.95e-05,
+      "loss": 7.2161,
+      "mean_token_accuracy": 0.07408471405506134,
+      "num_tokens": 497102.0,
+      "step": 180
+    },
+    {
+      "entropy": 8.015200138092041,
+      "epoch": 0.005008460236886633,
+      "grad_norm": 1.3203125,
+      "learning_rate": 9.2e-05,
+      "loss": 7.236,
+      "mean_token_accuracy": 0.07463390305638314,
+      "num_tokens": 511777.0,
+      "step": 185
+    },
+    {
+      "entropy": 8.013217878341674,
+      "epoch": 0.005143824027072758,
+      "grad_norm": 1.5234375,
+      "learning_rate": 9.45e-05,
+      "loss": 7.2509,
+      "mean_token_accuracy": 0.07550472915172576,
+      "num_tokens": 525855.0,
+      "step": 190
+    },
+    {
+      "entropy": 7.909638738632202,
+      "epoch": 0.005279187817258883,
+      "grad_norm": 1.84375,
+      "learning_rate": 9.7e-05,
+      "loss": 7.1878,
+      "mean_token_accuracy": 0.0701899241656065,
+      "num_tokens": 541364.0,
+      "step": 195
+    },
+    {
+      "entropy": 7.90043454170227,
+      "epoch": 0.0054145516074450084,
+      "grad_norm": 1.546875,
+      "learning_rate": 9.95e-05,
+      "loss": 7.1873,
+      "mean_token_accuracy": 0.0761424608528614,
+      "num_tokens": 555489.0,
+      "step": 200
+    },
+    {
+      "entropy": 7.926745128631592,
+      "epoch": 0.005549915397631134,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.000102,
+      "loss": 7.2169,
+      "mean_token_accuracy": 0.07424422949552537,
+      "num_tokens": 569207.0,
+      "step": 205
+    },
+    {
+      "entropy": 7.8149010181427006,
+      "epoch": 0.005685279187817259,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00010449999999999999,
+      "loss": 7.1699,
+      "mean_token_accuracy": 0.07592894844710826,
+      "num_tokens": 583689.0,
+      "step": 210
+    },
+    {
+      "entropy": 7.787826442718506,
+      "epoch": 0.005820642978003384,
+      "grad_norm": 1.5625,
+      "learning_rate": 0.000107,
+      "loss": 7.1253,
+      "mean_token_accuracy": 0.0805364765226841,
+      "num_tokens": 597729.0,
+      "step": 215
+    },
+    {
+      "entropy": 7.7665050506591795,
+      "epoch": 0.005956006768189509,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.0001095,
+      "loss": 7.1149,
+      "mean_token_accuracy": 0.08125128298997879,
+      "num_tokens": 610354.0,
+      "step": 220
+    },
+    {
+      "entropy": 7.738793563842774,
+      "epoch": 0.006091370558375634,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.000112,
+      "loss": 6.9965,
+      "mean_token_accuracy": 0.08436518236994743,
+      "num_tokens": 625591.0,
+      "step": 225
+    },
+    {
+      "entropy": 7.677218818664551,
+      "epoch": 0.00622673434856176,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0001145,
+      "loss": 7.1185,
+      "mean_token_accuracy": 0.08086443841457366,
+      "num_tokens": 640936.0,
+      "step": 230
+    },
+    {
+      "entropy": 7.736022233963013,
+      "epoch": 0.006362098138747885,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.00011700000000000001,
+      "loss": 7.0841,
+      "mean_token_accuracy": 0.0765869714319706,
+      "num_tokens": 655949.0,
+      "step": 235
+    },
+    {
+      "entropy": 7.609257936477661,
+      "epoch": 0.00649746192893401,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.00011949999999999999,
+      "loss": 6.948,
+      "mean_token_accuracy": 0.08549271896481514,
+      "num_tokens": 669357.0,
+      "step": 240
+    },
+    {
+      "entropy": 7.508241319656372,
+      "epoch": 0.006632825719120135,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.000122,
+      "loss": 6.8378,
+      "mean_token_accuracy": 0.08860795795917512,
+      "num_tokens": 683906.0,
+      "step": 245
+    },
+    {
+      "entropy": 7.5739704132080075,
+      "epoch": 0.00676818950930626,
+      "grad_norm": 1.3828125,
+      "learning_rate": 0.0001245,
+      "loss": 6.9511,
+      "mean_token_accuracy": 0.0863402545452118,
+      "num_tokens": 700467.0,
+      "step": 250
+    },
+    {
+      "entropy": 7.5351649761199955,
+      "epoch": 0.0069035532994923855,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.000127,
+      "loss": 6.9582,
+      "mean_token_accuracy": 0.09148906618356704,
+      "num_tokens": 711779.0,
+      "step": 255
+    },
+    {
+      "entropy": 7.530111169815063,
+      "epoch": 0.007038917089678511,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.0001295,
+      "loss": 6.9595,
+      "mean_token_accuracy": 0.08548436611890793,
+      "num_tokens": 725455.0,
+      "step": 260
+    },
+    {
+      "entropy": 7.483809947967529,
+      "epoch": 0.007174280879864636,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.000132,
+      "loss": 6.8334,
+      "mean_token_accuracy": 0.08580193966627121,
+      "num_tokens": 740428.0,
+      "step": 265
+    },
+    {
+      "entropy": 7.460272359848022,
+      "epoch": 0.007309644670050761,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00013450000000000002,
+      "loss": 6.8215,
+      "mean_token_accuracy": 0.09169955775141717,
+      "num_tokens": 755327.0,
+      "step": 270
+    },
+    {
+      "entropy": 7.53057975769043,
+      "epoch": 0.007445008460236886,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.00013700000000000002,
+      "loss": 6.9262,
+      "mean_token_accuracy": 0.09086679890751839,
+      "num_tokens": 768335.0,
+      "step": 275
+    },
+    {
+      "entropy": 7.381830310821533,
+      "epoch": 0.0075803722504230115,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.0001395,
+      "loss": 6.9661,
+      "mean_token_accuracy": 0.08472498953342437,
+      "num_tokens": 781500.0,
+      "step": 280
+    },
+    {
+      "entropy": 7.446859312057495,
+      "epoch": 0.007715736040609137,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00014199999999999998,
+      "loss": 6.8651,
+      "mean_token_accuracy": 0.09278287589550019,
+      "num_tokens": 794859.0,
+      "step": 285
+    },
+    {
+      "entropy": 7.3814451694488525,
+      "epoch": 0.007851099830795263,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.0001445,
+      "loss": 6.8369,
+      "mean_token_accuracy": 0.09373516142368317,
+      "num_tokens": 808653.0,
+      "step": 290
+    },
+    {
+      "entropy": 7.418990850448608,
+      "epoch": 0.007986463620981387,
+      "grad_norm": 1.3828125,
+      "learning_rate": 0.000147,
+      "loss": 6.8866,
+      "mean_token_accuracy": 0.08324785232543945,
+      "num_tokens": 824876.0,
+      "step": 295
+    },
+    {
+      "entropy": 7.358400630950928,
+      "epoch": 0.008121827411167513,
+      "grad_norm": 1.25,
+      "learning_rate": 0.0001495,
+      "loss": 6.8195,
+      "mean_token_accuracy": 0.09909781068563461,
+      "num_tokens": 838065.0,
+      "step": 300
+    },
+    {
+      "entropy": 7.24797306060791,
+      "epoch": 0.008257191201353637,
+      "grad_norm": 1.375,
+      "learning_rate": 0.000152,
+      "loss": 6.7376,
+      "mean_token_accuracy": 0.09417691752314568,
+      "num_tokens": 850591.0,
+      "step": 305
+    },
+    {
+      "entropy": 7.293615865707397,
+      "epoch": 0.008392554991539763,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00015450000000000001,
+      "loss": 6.8553,
+      "mean_token_accuracy": 0.09107527285814285,
+      "num_tokens": 865254.0,
+      "step": 310
+    },
+    {
+      "entropy": 7.388335704803467,
+      "epoch": 0.008527918781725888,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.000157,
+      "loss": 6.8402,
+      "mean_token_accuracy": 0.08597931787371635,
+      "num_tokens": 879791.0,
+      "step": 315
+    },
+    {
+      "entropy": 7.291468143463135,
+      "epoch": 0.008663282571912014,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.0001595,
+      "loss": 6.8454,
+      "mean_token_accuracy": 0.09367485567927361,
+      "num_tokens": 891708.0,
+      "step": 320
+    },
+    {
+      "entropy": 7.272171115875244,
+      "epoch": 0.008798646362098138,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000162,
+      "loss": 6.7682,
+      "mean_token_accuracy": 0.09533382132649422,
+      "num_tokens": 905454.0,
+      "step": 325
+    },
+    {
+      "entropy": 7.248970413208008,
+      "epoch": 0.008934010152284264,
+      "grad_norm": 1.359375,
+      "learning_rate": 0.00016450000000000001,
+      "loss": 6.7542,
+      "mean_token_accuracy": 0.09669646397233009,
+      "num_tokens": 919080.0,
+      "step": 330
+    },
+    {
+      "entropy": 7.180841302871704,
+      "epoch": 0.009069373942470389,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00016700000000000002,
+      "loss": 6.6389,
+      "mean_token_accuracy": 0.09361326694488525,
+      "num_tokens": 932145.0,
+      "step": 335
+    },
+    {
+      "entropy": 7.175373554229736,
+      "epoch": 0.009204737732656515,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00016950000000000003,
+      "loss": 6.7554,
+      "mean_token_accuracy": 0.09223678484559059,
+      "num_tokens": 946502.0,
+      "step": 340
+    },
+    {
+      "entropy": 7.258382320404053,
+      "epoch": 0.009340101522842639,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 6.7061,
+      "mean_token_accuracy": 0.09237500727176666,
+      "num_tokens": 959414.0,
+      "step": 345
+    },
+    {
+      "entropy": 7.242776870727539,
+      "epoch": 0.009475465313028765,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00017449999999999999,
+      "loss": 6.7714,
+      "mean_token_accuracy": 0.09881808012723922,
+      "num_tokens": 972868.0,
+      "step": 350
+    },
+    {
+      "entropy": 7.061698722839355,
+      "epoch": 0.00961082910321489,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.000177,
+      "loss": 6.6902,
+      "mean_token_accuracy": 0.09196390882134438,
+      "num_tokens": 987469.0,
+      "step": 355
+    },
+    {
+      "entropy": 7.158710718154907,
+      "epoch": 0.009746192893401015,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0001795,
+      "loss": 6.693,
+      "mean_token_accuracy": 0.09573714062571526,
+      "num_tokens": 1001797.0,
+      "step": 360
+    },
+    {
+      "entropy": 7.105884790420532,
+      "epoch": 0.00988155668358714,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.000182,
+      "loss": 6.6582,
+      "mean_token_accuracy": 0.09512505084276199,
+      "num_tokens": 1016034.0,
+      "step": 365
+    },
+    {
+      "entropy": 7.184845685958862,
+      "epoch": 0.010016920473773266,
+      "grad_norm": 2.125,
+      "learning_rate": 0.0001845,
+      "loss": 6.7164,
+      "mean_token_accuracy": 0.09947735443711281,
+      "num_tokens": 1030566.0,
+      "step": 370
+    },
+    {
+      "entropy": 7.0832836627960205,
+      "epoch": 0.01015228426395939,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.000187,
+      "loss": 6.763,
+      "mean_token_accuracy": 0.09124055355787278,
+      "num_tokens": 1044735.0,
+      "step": 375
+    },
+    {
+      "entropy": 7.09069504737854,
+      "epoch": 0.010287648054145516,
+      "grad_norm": 1.359375,
+      "learning_rate": 0.0001895,
+      "loss": 6.7385,
+      "mean_token_accuracy": 0.0939220942556858,
+      "num_tokens": 1059567.0,
+      "step": 380
+    },
+    {
+      "entropy": 7.149837684631348,
+      "epoch": 0.01042301184433164,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.000192,
+      "loss": 6.6309,
+      "mean_token_accuracy": 0.1013464093208313,
+      "num_tokens": 1074612.0,
+      "step": 385
+    },
+    {
+      "entropy": 7.109938526153565,
+      "epoch": 0.010558375634517767,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.0001945,
+      "loss": 6.6941,
+      "mean_token_accuracy": 0.09540026187896729,
+      "num_tokens": 1087092.0,
+      "step": 390
+    },
+    {
+      "entropy": 6.9963634490966795,
+      "epoch": 0.010693739424703893,
+      "grad_norm": 1.484375,
+      "learning_rate": 0.00019700000000000002,
+      "loss": 6.6652,
+      "mean_token_accuracy": 0.09614972546696662,
+      "num_tokens": 1101966.0,
+      "step": 395
+    },
+    {
+      "entropy": 7.059049797058106,
+      "epoch": 0.010829103214890017,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00019950000000000002,
+      "loss": 6.6127,
+      "mean_token_accuracy": 0.10141809433698654,
+      "num_tokens": 1116287.0,
+      "step": 400
+    },
+    {
+      "entropy": 7.021635675430298,
+      "epoch": 0.010964467005076143,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.000202,
+      "loss": 6.5773,
+      "mean_token_accuracy": 0.10086113512516022,
+      "num_tokens": 1132775.0,
+      "step": 405
+    },
+    {
+      "entropy": 6.978186845779419,
+      "epoch": 0.011099830795262267,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00020449999999999998,
+      "loss": 6.6211,
+      "mean_token_accuracy": 0.09556624442338943,
+      "num_tokens": 1149372.0,
+      "step": 410
+    },
+    {
+      "entropy": 7.039114618301392,
+      "epoch": 0.011235194585448393,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.000207,
+      "loss": 6.7039,
+      "mean_token_accuracy": 0.09624231159687043,
+      "num_tokens": 1163772.0,
+      "step": 415
+    },
+    {
+      "entropy": 6.965072917938232,
+      "epoch": 0.011370558375634518,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0002095,
+      "loss": 6.5853,
+      "mean_token_accuracy": 0.09794198945164681,
+      "num_tokens": 1176056.0,
+      "step": 420
+    },
+    {
+      "entropy": 7.030379724502564,
+      "epoch": 0.011505922165820644,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.000212,
+      "loss": 6.5365,
+      "mean_token_accuracy": 0.10276012122631073,
+      "num_tokens": 1189759.0,
+      "step": 425
+    },
+    {
+      "entropy": 7.0084771633148195,
+      "epoch": 0.011641285956006768,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0002145,
+      "loss": 6.591,
+      "mean_token_accuracy": 0.10040640980005264,
+      "num_tokens": 1203142.0,
+      "step": 430
+    },
+    {
+      "entropy": 6.9573570728302006,
+      "epoch": 0.011776649746192894,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00021700000000000002,
+      "loss": 6.5819,
+      "mean_token_accuracy": 0.10083244740962982,
+      "num_tokens": 1217924.0,
+      "step": 435
+    },
+    {
+      "entropy": 7.0246354103088375,
+      "epoch": 0.011912013536379018,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.0002195,
+      "loss": 6.6666,
+      "mean_token_accuracy": 0.10064776018261909,
+      "num_tokens": 1231713.0,
+      "step": 440
+    },
+    {
+      "entropy": 6.966695070266724,
+      "epoch": 0.012047377326565144,
+      "grad_norm": 1.6015625,
+      "learning_rate": 0.000222,
+      "loss": 6.56,
+      "mean_token_accuracy": 0.0989975780248642,
+      "num_tokens": 1245538.0,
+      "step": 445
+    },
+    {
+      "entropy": 6.837519836425781,
+      "epoch": 0.012182741116751269,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0002245,
+      "loss": 6.574,
+      "mean_token_accuracy": 0.10176394879817963,
+      "num_tokens": 1262574.0,
+      "step": 450
+    },
+    {
+      "entropy": 6.953103399276733,
+      "epoch": 0.012318104906937395,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.00022700000000000002,
+      "loss": 6.4812,
+      "mean_token_accuracy": 0.10641162097454071,
+      "num_tokens": 1276727.0,
+      "step": 455
+    },
+    {
+      "entropy": 6.89174976348877,
+      "epoch": 0.01245346869712352,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00022950000000000002,
+      "loss": 6.4861,
+      "mean_token_accuracy": 0.10931641608476639,
+      "num_tokens": 1291029.0,
+      "step": 460
+    },
+    {
+      "entropy": 6.775276374816895,
+      "epoch": 0.012588832487309645,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 6.4553,
+      "mean_token_accuracy": 0.1023554228246212,
+      "num_tokens": 1305720.0,
+      "step": 465
+    },
+    {
+      "entropy": 6.879015350341797,
+      "epoch": 0.01272419627749577,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00023449999999999998,
+      "loss": 6.4732,
+      "mean_token_accuracy": 0.10401393547654152,
+      "num_tokens": 1321681.0,
+      "step": 470
+    },
+    {
+      "entropy": 6.755443096160889,
+      "epoch": 0.012859560067681896,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.000237,
+      "loss": 6.4114,
+      "mean_token_accuracy": 0.11262697726488113,
+      "num_tokens": 1334960.0,
+      "step": 475
+    },
+    {
+      "entropy": 6.846341848373413,
+      "epoch": 0.01299492385786802,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.0002395,
+      "loss": 6.58,
+      "mean_token_accuracy": 0.10540890470147132,
+      "num_tokens": 1349479.0,
+      "step": 480
+    },
+    {
+      "entropy": 6.880573081970215,
+      "epoch": 0.013130287648054146,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.000242,
+      "loss": 6.4669,
+      "mean_token_accuracy": 0.10141590088605881,
+      "num_tokens": 1363240.0,
+      "step": 485
+    },
+    {
+      "entropy": 6.843642616271973,
+      "epoch": 0.01326565143824027,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0002445,
+      "loss": 6.5119,
+      "mean_token_accuracy": 0.09850463718175888,
+      "num_tokens": 1375568.0,
+      "step": 490
+    },
+    {
+      "entropy": 6.877911949157715,
+      "epoch": 0.013401015228426396,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000247,
+      "loss": 6.5538,
+      "mean_token_accuracy": 0.10284689888358116,
+      "num_tokens": 1389099.0,
+      "step": 495
+    },
+    {
+      "entropy": 6.875770854949951,
+      "epoch": 0.01353637901861252,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.0002495,
+      "loss": 6.5636,
+      "mean_token_accuracy": 0.09843793138861656,
+      "num_tokens": 1402107.0,
+      "step": 500
+    },
+    {
+      "entropy": 6.843701934814453,
+      "epoch": 0.013671742808798647,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.000252,
+      "loss": 6.537,
+      "mean_token_accuracy": 0.10457578748464584,
+      "num_tokens": 1417644.0,
+      "step": 505
+    },
+    {
+      "entropy": 6.870151567459106,
+      "epoch": 0.013807106598984771,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0002545,
+      "loss": 6.5303,
+      "mean_token_accuracy": 0.09835479706525803,
+      "num_tokens": 1433103.0,
+      "step": 510
+    },
+    {
+      "entropy": 6.636357831954956,
+      "epoch": 0.013942470389170897,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.000257,
+      "loss": 6.3019,
+      "mean_token_accuracy": 0.10845862105488777,
+      "num_tokens": 1445449.0,
+      "step": 515
+    },
+    {
+      "entropy": 6.858165597915649,
+      "epoch": 0.014077834179357021,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.0002595,
+      "loss": 6.4757,
+      "mean_token_accuracy": 0.11344253048300743,
+      "num_tokens": 1458077.0,
+      "step": 520
+    },
+    {
+      "entropy": 6.699763393402099,
+      "epoch": 0.014213197969543147,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.000262,
+      "loss": 6.4336,
+      "mean_token_accuracy": 0.10050448402762413,
+      "num_tokens": 1471773.0,
+      "step": 525
+    },
+    {
+      "entropy": 6.785126543045044,
+      "epoch": 0.014348561759729272,
+      "grad_norm": 1.375,
+      "learning_rate": 0.00026450000000000003,
+      "loss": 6.4653,
+      "mean_token_accuracy": 0.1011991024017334,
+      "num_tokens": 1486826.0,
+      "step": 530
+    },
+    {
+      "entropy": 6.653678703308105,
+      "epoch": 0.014483925549915398,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00026700000000000004,
+      "loss": 6.392,
+      "mean_token_accuracy": 0.1090740442276001,
+      "num_tokens": 1501027.0,
+      "step": 535
+    },
+    {
+      "entropy": 6.773468685150147,
+      "epoch": 0.014619289340101522,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00026950000000000005,
+      "loss": 6.3659,
+      "mean_token_accuracy": 0.10750545933842659,
+      "num_tokens": 1512547.0,
+      "step": 540
+    },
+    {
+      "entropy": 6.705225276947021,
+      "epoch": 0.014754653130287648,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 6.3975,
+      "mean_token_accuracy": 0.11041399985551834,
+      "num_tokens": 1525485.0,
+      "step": 545
+    },
+    {
+      "entropy": 6.779332685470581,
+      "epoch": 0.014890016920473773,
+      "grad_norm": 1.375,
+      "learning_rate": 0.0002745,
+      "loss": 6.3915,
+      "mean_token_accuracy": 0.11499902307987213,
+      "num_tokens": 1538824.0,
+      "step": 550
+    },
+    {
+      "entropy": 6.653468084335327,
+      "epoch": 0.015025380710659899,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.000277,
+      "loss": 6.4215,
+      "mean_token_accuracy": 0.10652948096394539,
+      "num_tokens": 1553913.0,
+      "step": 555
+    },
+    {
+      "entropy": 6.714767408370972,
+      "epoch": 0.015160744500846023,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.0002795,
+      "loss": 6.4351,
+      "mean_token_accuracy": 0.10829391852021217,
+      "num_tokens": 1567865.0,
+      "step": 560
+    },
+    {
+      "entropy": 6.673421001434326,
+      "epoch": 0.015296108291032149,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 6.393,
+      "mean_token_accuracy": 0.10847804918885232,
+      "num_tokens": 1581334.0,
+      "step": 565
+    },
+    {
+      "entropy": 6.721217823028565,
+      "epoch": 0.015431472081218273,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.0002845,
+      "loss": 6.3429,
+      "mean_token_accuracy": 0.1096364676952362,
+      "num_tokens": 1595181.0,
+      "step": 570
+    },
+    {
+      "entropy": 6.660063457489014,
+      "epoch": 0.0155668358714044,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.000287,
+      "loss": 6.3849,
+      "mean_token_accuracy": 0.11193926930427552,
+      "num_tokens": 1609117.0,
+      "step": 575
+    },
+    {
+      "entropy": 6.630877065658569,
+      "epoch": 0.015702199661590525,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0002895,
+      "loss": 6.3853,
+      "mean_token_accuracy": 0.10854767486453057,
+      "num_tokens": 1623200.0,
+      "step": 580
+    },
+    {
+      "entropy": 6.59787392616272,
+      "epoch": 0.01583756345177665,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.000292,
+      "loss": 6.2754,
+      "mean_token_accuracy": 0.11524961069226265,
+      "num_tokens": 1637372.0,
+      "step": 585
+    },
+    {
+      "entropy": 6.6487486362457275,
+      "epoch": 0.015972927241962774,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.0002945,
+      "loss": 6.292,
+      "mean_token_accuracy": 0.10983927696943283,
+      "num_tokens": 1651853.0,
+      "step": 590
+    },
+    {
+      "entropy": 6.589022636413574,
+      "epoch": 0.016108291032148902,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.000297,
+      "loss": 6.3127,
+      "mean_token_accuracy": 0.11347366273403167,
+      "num_tokens": 1664889.0,
+      "step": 595
+    },
+    {
+      "entropy": 6.664232492446899,
+      "epoch": 0.016243654822335026,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.0002995,
+      "loss": 6.3829,
+      "mean_token_accuracy": 0.10112749561667442,
+      "num_tokens": 1679357.0,
+      "step": 600
+    },
+    {
+      "entropy": 6.555701160430909,
+      "epoch": 0.01637901861252115,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.000302,
+      "loss": 6.2406,
+      "mean_token_accuracy": 0.11562057062983513,
+      "num_tokens": 1693179.0,
+      "step": 605
+    },
+    {
+      "entropy": 6.557049798965454,
+      "epoch": 0.016514382402707275,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.0003045,
+      "loss": 6.2925,
+      "mean_token_accuracy": 0.1144544430077076,
+      "num_tokens": 1706554.0,
+      "step": 610
+    },
+    {
+      "entropy": 6.619402074813843,
+      "epoch": 0.016649746192893403,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.000307,
+      "loss": 6.39,
+      "mean_token_accuracy": 0.1108802892267704,
+      "num_tokens": 1720010.0,
+      "step": 615
+    },
+    {
+      "entropy": 6.585338306427002,
+      "epoch": 0.016785109983079527,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.0003095,
+      "loss": 6.3556,
+      "mean_token_accuracy": 0.10875253900885581,
+      "num_tokens": 1733041.0,
+      "step": 620
+    },
+    {
+      "entropy": 6.681243371963501,
+      "epoch": 0.01692047377326565,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.000312,
+      "loss": 6.3521,
+      "mean_token_accuracy": 0.10595379248261452,
+      "num_tokens": 1747324.0,
+      "step": 625
+    },
+    {
+      "entropy": 6.517426633834839,
+      "epoch": 0.017055837563451776,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.0003145,
+      "loss": 6.2758,
+      "mean_token_accuracy": 0.11371921375393867,
+      "num_tokens": 1760321.0,
+      "step": 630
+    },
+    {
+      "entropy": 6.659466218948364,
+      "epoch": 0.017191201353637903,
+      "grad_norm": 1.25,
+      "learning_rate": 0.000317,
+      "loss": 6.4047,
+      "mean_token_accuracy": 0.10822949036955834,
+      "num_tokens": 1775552.0,
+      "step": 635
+    },
+    {
+      "entropy": 6.524487066268921,
+      "epoch": 0.017326565143824028,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0003195,
+      "loss": 6.3124,
+      "mean_token_accuracy": 0.11699947118759155,
+      "num_tokens": 1788698.0,
+      "step": 640
+    },
+    {
+      "entropy": 6.539701795578003,
+      "epoch": 0.017461928934010152,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000322,
+      "loss": 6.3669,
+      "mean_token_accuracy": 0.1119937926530838,
+      "num_tokens": 1803775.0,
+      "step": 645
+    },
+    {
+      "entropy": 6.564684867858887,
+      "epoch": 0.017597292724196276,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00032450000000000003,
+      "loss": 6.3122,
+      "mean_token_accuracy": 0.1049906738102436,
+      "num_tokens": 1817010.0,
+      "step": 650
+    },
+    {
+      "entropy": 6.499929475784302,
+      "epoch": 0.017732656514382404,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.00032700000000000003,
+      "loss": 6.2747,
+      "mean_token_accuracy": 0.10665653720498085,
+      "num_tokens": 1831788.0,
+      "step": 655
+    },
+    {
+      "entropy": 6.599995756149292,
+      "epoch": 0.01786802030456853,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.00032950000000000004,
+      "loss": 6.2993,
+      "mean_token_accuracy": 0.10768605694174767,
+      "num_tokens": 1846624.0,
+      "step": 660
+    },
+    {
+      "entropy": 6.432188606262207,
+      "epoch": 0.018003384094754653,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 6.2798,
+      "mean_token_accuracy": 0.1103129394352436,
+      "num_tokens": 1863033.0,
+      "step": 665
+    },
+    {
+      "entropy": 6.561399269104004,
+      "epoch": 0.018138747884940777,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00033450000000000005,
+      "loss": 6.3231,
+      "mean_token_accuracy": 0.109639373421669,
+      "num_tokens": 1877067.0,
+      "step": 670
+    },
+    {
+      "entropy": 6.520267629623413,
+      "epoch": 0.018274111675126905,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000337,
+      "loss": 6.1885,
+      "mean_token_accuracy": 0.1139417678117752,
+      "num_tokens": 1890950.0,
+      "step": 675
+    },
+    {
+      "entropy": 6.526244306564331,
+      "epoch": 0.01840947546531303,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0003395,
+      "loss": 6.3383,
+      "mean_token_accuracy": 0.111404038220644,
+      "num_tokens": 1904091.0,
+      "step": 680
+    },
+    {
+      "entropy": 6.561566352844238,
+      "epoch": 0.018544839255499154,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.000342,
+      "loss": 6.2958,
+      "mean_token_accuracy": 0.10963592603802681,
+      "num_tokens": 1917469.0,
+      "step": 685
+    },
+    {
+      "entropy": 6.326075553894043,
+      "epoch": 0.018680203045685278,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00034449999999999997,
+      "loss": 6.1969,
+      "mean_token_accuracy": 0.11997934058308601,
+      "num_tokens": 1932512.0,
+      "step": 690
+    },
+    {
+      "entropy": 6.496285247802734,
+      "epoch": 0.018815566835871406,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.000347,
+      "loss": 6.2037,
+      "mean_token_accuracy": 0.11709487736225128,
+      "num_tokens": 1946953.0,
+      "step": 695
+    },
+    {
+      "entropy": 6.45741024017334,
+      "epoch": 0.01895093062605753,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0003495,
+      "loss": 6.2194,
+      "mean_token_accuracy": 0.11923639252781867,
+      "num_tokens": 1959479.0,
+      "step": 700
+    },
+    {
+      "entropy": 6.558106708526611,
+      "epoch": 0.019086294416243654,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.000352,
+      "loss": 6.2801,
+      "mean_token_accuracy": 0.10900771766901016,
+      "num_tokens": 1973683.0,
+      "step": 705
+    },
+    {
+      "entropy": 6.336573648452759,
+      "epoch": 0.01922165820642978,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.0003545,
+      "loss": 6.2242,
+      "mean_token_accuracy": 0.11210474893450736,
+      "num_tokens": 1988255.0,
+      "step": 710
+    },
+    {
+      "entropy": 6.4287889957427975,
+      "epoch": 0.019357021996615906,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000357,
+      "loss": 6.2052,
+      "mean_token_accuracy": 0.11728888675570488,
+      "num_tokens": 2003306.0,
+      "step": 715
+    },
+    {
+      "entropy": 6.432671117782593,
+      "epoch": 0.01949238578680203,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0003595,
+      "loss": 6.2291,
+      "mean_token_accuracy": 0.11440798714756965,
+      "num_tokens": 2015810.0,
+      "step": 720
+    },
+    {
+      "entropy": 6.465086364746094,
+      "epoch": 0.019627749576988155,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000362,
+      "loss": 6.2413,
+      "mean_token_accuracy": 0.11951256170868874,
+      "num_tokens": 2029477.0,
+      "step": 725
+    },
+    {
+      "entropy": 6.478774738311768,
+      "epoch": 0.01976311336717428,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0003645,
+      "loss": 6.2714,
+      "mean_token_accuracy": 0.1068502850830555,
+      "num_tokens": 2044779.0,
+      "step": 730
+    },
+    {
+      "entropy": 6.394287252426148,
+      "epoch": 0.019898477157360407,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.000367,
+      "loss": 6.2824,
+      "mean_token_accuracy": 0.11770006865262986,
+      "num_tokens": 2059640.0,
+      "step": 735
+    },
+    {
+      "entropy": 6.4458049774169925,
+      "epoch": 0.02003384094754653,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0003695,
+      "loss": 6.2743,
+      "mean_token_accuracy": 0.11187923103570938,
+      "num_tokens": 2073378.0,
+      "step": 740
+    },
+    {
+      "entropy": 6.440818643569946,
+      "epoch": 0.020169204737732656,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.000372,
+      "loss": 6.1675,
+      "mean_token_accuracy": 0.11438429951667786,
+      "num_tokens": 2087484.0,
+      "step": 745
+    },
+    {
+      "entropy": 6.520938920974731,
+      "epoch": 0.02030456852791878,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0003745,
+      "loss": 6.3903,
+      "mean_token_accuracy": 0.10714041441679001,
+      "num_tokens": 2103186.0,
+      "step": 750
+    },
+    {
+      "entropy": 6.335501766204834,
+      "epoch": 0.020439932318104908,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.000377,
+      "loss": 6.1203,
+      "mean_token_accuracy": 0.1230811521410942,
+      "num_tokens": 2117578.0,
+      "step": 755
+    },
+    {
+      "entropy": 6.347048139572143,
+      "epoch": 0.020575296108291032,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0003795,
+      "loss": 6.2012,
+      "mean_token_accuracy": 0.11709161922335624,
+      "num_tokens": 2133742.0,
+      "step": 760
+    },
+    {
+      "entropy": 6.374935579299927,
+      "epoch": 0.020710659898477157,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.000382,
+      "loss": 6.2038,
+      "mean_token_accuracy": 0.11305479407310486,
+      "num_tokens": 2147709.0,
+      "step": 765
+    },
+    {
+      "entropy": 6.456587076187134,
+      "epoch": 0.02084602368866328,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0003845,
+      "loss": 6.1872,
+      "mean_token_accuracy": 0.11732911616563797,
+      "num_tokens": 2161898.0,
+      "step": 770
+    },
+    {
+      "entropy": 6.387871885299683,
+      "epoch": 0.02098138747884941,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00038700000000000003,
+      "loss": 6.2099,
+      "mean_token_accuracy": 0.11243800446391106,
+      "num_tokens": 2177720.0,
+      "step": 775
+    },
+    {
+      "entropy": 6.285234069824218,
+      "epoch": 0.021116751269035533,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00038950000000000003,
+      "loss": 6.0929,
+      "mean_token_accuracy": 0.12101836651563644,
+      "num_tokens": 2191675.0,
+      "step": 780
+    },
+    {
+      "entropy": 6.226907634735108,
+      "epoch": 0.021252115059221657,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 6.0661,
+      "mean_token_accuracy": 0.12149770706892013,
+      "num_tokens": 2205366.0,
+      "step": 785
+    },
+    {
+      "entropy": 6.322827768325806,
+      "epoch": 0.021387478849407785,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00039450000000000005,
+      "loss": 6.165,
+      "mean_token_accuracy": 0.11490083187818527,
+      "num_tokens": 2219062.0,
+      "step": 790
+    },
+    {
+      "entropy": 6.354215955734253,
+      "epoch": 0.02152284263959391,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00039700000000000005,
+      "loss": 6.1218,
+      "mean_token_accuracy": 0.12039542272686958,
+      "num_tokens": 2232700.0,
+      "step": 795
+    },
+    {
+      "entropy": 6.188680410385132,
+      "epoch": 0.021658206429780034,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0003995,
+      "loss": 6.0871,
+      "mean_token_accuracy": 0.12226910442113877,
+      "num_tokens": 2246944.0,
+      "step": 800
+    },
+    {
+      "entropy": 6.37126350402832,
+      "epoch": 0.021793570219966158,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.000402,
+      "loss": 6.1353,
+      "mean_token_accuracy": 0.12089133262634277,
+      "num_tokens": 2260988.0,
+      "step": 805
+    },
+    {
+      "entropy": 6.415036964416504,
+      "epoch": 0.021928934010152286,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.0004045,
+      "loss": 6.208,
+      "mean_token_accuracy": 0.1159418523311615,
+      "num_tokens": 2275403.0,
+      "step": 810
+    },
+    {
+      "entropy": 6.222449016571045,
+      "epoch": 0.02206429780033841,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00040699999999999997,
+      "loss": 6.1736,
+      "mean_token_accuracy": 0.11669653207063675,
+      "num_tokens": 2290861.0,
+      "step": 815
+    },
+    {
+      "entropy": 6.282603788375854,
+      "epoch": 0.022199661590524535,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0004095,
+      "loss": 6.1153,
+      "mean_token_accuracy": 0.12357514351606369,
+      "num_tokens": 2304171.0,
+      "step": 820
+    },
+    {
+      "entropy": 6.458443546295166,
+      "epoch": 0.02233502538071066,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000412,
+      "loss": 6.2533,
+      "mean_token_accuracy": 0.11721441000699998,
+      "num_tokens": 2318805.0,
+      "step": 825
+    },
+    {
+      "entropy": 6.250155591964722,
+      "epoch": 0.022470389170896787,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004145,
+      "loss": 6.1643,
+      "mean_token_accuracy": 0.12117116153240204,
+      "num_tokens": 2332894.0,
+      "step": 830
+    },
+    {
+      "entropy": 6.294081735610962,
+      "epoch": 0.02260575296108291,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.000417,
+      "loss": 6.1301,
+      "mean_token_accuracy": 0.11678448617458344,
+      "num_tokens": 2346718.0,
+      "step": 835
+    },
+    {
+      "entropy": 6.230988359451294,
+      "epoch": 0.022741116751269035,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0004195,
+      "loss": 6.1202,
+      "mean_token_accuracy": 0.11509295254945755,
+      "num_tokens": 2359480.0,
+      "step": 840
+    },
+    {
+      "entropy": 6.359671497344971,
+      "epoch": 0.02287648054145516,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.000422,
+      "loss": 6.2173,
+      "mean_token_accuracy": 0.11525185033679008,
+      "num_tokens": 2373095.0,
+      "step": 845
+    },
+    {
+      "entropy": 6.199200820922852,
+      "epoch": 0.023011844331641287,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.0004245,
+      "loss": 6.0346,
+      "mean_token_accuracy": 0.12598701417446137,
+      "num_tokens": 2386224.0,
+      "step": 850
+    },
+    {
+      "entropy": 6.216959381103516,
+      "epoch": 0.02314720812182741,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.000427,
+      "loss": 6.1023,
+      "mean_token_accuracy": 0.12083085179328919,
+      "num_tokens": 2402813.0,
+      "step": 855
+    },
+    {
+      "entropy": 6.328297233581543,
+      "epoch": 0.023282571912013536,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004295,
+      "loss": 6.097,
+      "mean_token_accuracy": 0.12417078986763955,
+      "num_tokens": 2415423.0,
+      "step": 860
+    },
+    {
+      "entropy": 6.259360122680664,
+      "epoch": 0.02341793570219966,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.000432,
+      "loss": 6.1033,
+      "mean_token_accuracy": 0.11193115115165711,
+      "num_tokens": 2427845.0,
+      "step": 865
+    },
+    {
+      "entropy": 6.251777505874633,
+      "epoch": 0.023553299492385788,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004345,
+      "loss": 6.0958,
+      "mean_token_accuracy": 0.11671509072184563,
+      "num_tokens": 2442498.0,
+      "step": 870
+    },
+    {
+      "entropy": 6.154549264907837,
+      "epoch": 0.023688663282571912,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.000437,
+      "loss": 6.094,
+      "mean_token_accuracy": 0.11765043288469315,
+      "num_tokens": 2455483.0,
+      "step": 875
+    },
+    {
+      "entropy": 6.1775672912597654,
+      "epoch": 0.023824027072758037,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.0004395,
+      "loss": 6.0294,
+      "mean_token_accuracy": 0.125057340413332,
+      "num_tokens": 2469133.0,
+      "step": 880
+    },
+    {
+      "entropy": 6.341091442108154,
+      "epoch": 0.02395939086294416,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.000442,
+      "loss": 6.12,
+      "mean_token_accuracy": 0.11728312373161316,
+      "num_tokens": 2484060.0,
+      "step": 885
+    },
+    {
+      "entropy": 6.137259387969971,
+      "epoch": 0.02409475465313029,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004445,
+      "loss": 5.9998,
+      "mean_token_accuracy": 0.1305142655968666,
+      "num_tokens": 2497132.0,
+      "step": 890
+    },
+    {
+      "entropy": 6.238704776763916,
+      "epoch": 0.024230118443316413,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.000447,
+      "loss": 6.1717,
+      "mean_token_accuracy": 0.11579567641019821,
+      "num_tokens": 2511285.0,
+      "step": 895
+    },
+    {
+      "entropy": 6.297119235992431,
+      "epoch": 0.024365482233502538,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00044950000000000003,
+      "loss": 6.1921,
+      "mean_token_accuracy": 0.12004831880331039,
+      "num_tokens": 2524660.0,
+      "step": 900
+    },
+    {
+      "entropy": 6.38506498336792,
+      "epoch": 0.024500846023688662,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 6.153,
+      "mean_token_accuracy": 0.12265655994415284,
+      "num_tokens": 2536944.0,
+      "step": 905
+    },
+    {
+      "entropy": 6.106860065460205,
+      "epoch": 0.02463620981387479,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00045450000000000004,
+      "loss": 6.113,
+      "mean_token_accuracy": 0.1159572497010231,
+      "num_tokens": 2552167.0,
+      "step": 910
+    },
+    {
+      "entropy": 6.307010126113892,
+      "epoch": 0.024771573604060914,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00045700000000000005,
+      "loss": 6.1229,
+      "mean_token_accuracy": 0.1155533254146576,
+      "num_tokens": 2565263.0,
+      "step": 915
+    },
+    {
+      "entropy": 6.212118196487427,
+      "epoch": 0.02490693739424704,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00045950000000000006,
+      "loss": 6.0779,
+      "mean_token_accuracy": 0.12355258241295815,
+      "num_tokens": 2579209.0,
+      "step": 920
+    },
+    {
+      "entropy": 6.278075122833252,
+      "epoch": 0.025042301184433163,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.000462,
+      "loss": 6.127,
+      "mean_token_accuracy": 0.12122250273823738,
+      "num_tokens": 2594162.0,
+      "step": 925
+    },
+    {
+      "entropy": 6.144100475311279,
+      "epoch": 0.02517766497461929,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004645,
+      "loss": 6.1096,
+      "mean_token_accuracy": 0.11865596175193786,
+      "num_tokens": 2608748.0,
+      "step": 930
+    },
+    {
+      "entropy": 6.206922101974487,
+      "epoch": 0.025313028764805415,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.000467,
+      "loss": 6.0455,
+      "mean_token_accuracy": 0.11540281772613525,
+      "num_tokens": 2625671.0,
+      "step": 935
+    },
+    {
+      "entropy": 6.185919237136841,
+      "epoch": 0.02544839255499154,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.0004695,
+      "loss": 6.0258,
+      "mean_token_accuracy": 0.13033671155571938,
+      "num_tokens": 2640156.0,
+      "step": 940
+    },
+    {
+      "entropy": 6.142274999618531,
+      "epoch": 0.025583756345177663,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000472,
+      "loss": 6.0462,
+      "mean_token_accuracy": 0.12545183822512626,
+      "num_tokens": 2655609.0,
+      "step": 945
+    },
+    {
+      "entropy": 6.160119819641113,
+      "epoch": 0.02571912013536379,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004745,
+      "loss": 6.0096,
+      "mean_token_accuracy": 0.12306903898715973,
+      "num_tokens": 2668473.0,
+      "step": 950
+    },
+    {
+      "entropy": 6.120021915435791,
+      "epoch": 0.025854483925549916,
+      "grad_norm": 1.125,
+      "learning_rate": 0.000477,
+      "loss": 6.0944,
+      "mean_token_accuracy": 0.1181985966861248,
+      "num_tokens": 2681997.0,
+      "step": 955
+    },
+    {
+      "entropy": 6.164200687408448,
+      "epoch": 0.02598984771573604,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0004795,
+      "loss": 6.0733,
+      "mean_token_accuracy": 0.12185675352811813,
+      "num_tokens": 2697502.0,
+      "step": 960
+    },
+    {
+      "entropy": 6.162236452102661,
+      "epoch": 0.026125211505922164,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.000482,
+      "loss": 6.024,
+      "mean_token_accuracy": 0.12753638476133347,
+      "num_tokens": 2710780.0,
+      "step": 965
+    },
+    {
+      "entropy": 6.187419700622558,
+      "epoch": 0.026260575296108292,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004845,
+      "loss": 6.0861,
+      "mean_token_accuracy": 0.12425143718719482,
+      "num_tokens": 2725643.0,
+      "step": 970
+    },
+    {
+      "entropy": 6.1629087924957275,
+      "epoch": 0.026395939086294416,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.000487,
+      "loss": 6.0179,
+      "mean_token_accuracy": 0.12270609810948371,
+      "num_tokens": 2738610.0,
+      "step": 975
+    },
+    {
+      "entropy": 6.142755031585693,
+      "epoch": 0.02653130287648054,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004895,
+      "loss": 6.0373,
+      "mean_token_accuracy": 0.1201431192457676,
+      "num_tokens": 2752765.0,
+      "step": 980
+    },
+    {
+      "entropy": 6.1001872539520265,
+      "epoch": 0.02666666666666667,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000492,
+      "loss": 5.9777,
+      "mean_token_accuracy": 0.12657394856214524,
+      "num_tokens": 2765355.0,
+      "step": 985
+    },
+    {
+      "entropy": 6.141935205459594,
+      "epoch": 0.026802030456852793,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0004945,
+      "loss": 6.0975,
+      "mean_token_accuracy": 0.12246641293168067,
+      "num_tokens": 2781318.0,
+      "step": 990
+    },
+    {
+      "entropy": 6.18066816329956,
+      "epoch": 0.026937394247038917,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.000497,
+      "loss": 6.0927,
+      "mean_token_accuracy": 0.11901046261191368,
+      "num_tokens": 2794329.0,
+      "step": 995
+    },
+    {
+      "entropy": 6.097706031799317,
+      "epoch": 0.02707275803722504,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0004995,
+      "loss": 6.0651,
+      "mean_token_accuracy": 0.12499085217714309,
+      "num_tokens": 2807743.0,
+      "step": 1000
+    },
+    {
+      "entropy": 6.1797788619995115,
+      "epoch": 0.02720812182741117,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000499998026082006,
+      "loss": 6.0732,
+      "mean_token_accuracy": 0.12034357115626335,
+      "num_tokens": 2821700.0,
+      "step": 1005
+    },
+    {
+      "entropy": 6.11210823059082,
+      "epoch": 0.027343485617597293,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0004999900070995136,
+      "loss": 6.1228,
+      "mean_token_accuracy": 0.12435722276568413,
+      "num_tokens": 2836110.0,
+      "step": 1010
+    },
+    {
+      "entropy": 6.177678155899048,
+      "epoch": 0.027478849407783418,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.0004999758199023239,
+      "loss": 6.0535,
+      "mean_token_accuracy": 0.12844400405883788,
+      "num_tokens": 2850084.0,
+      "step": 1015
+    },
+    {
+      "entropy": 6.247022819519043,
+      "epoch": 0.027614213197969542,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0004999554648793858,
+      "loss": 6.1175,
+      "mean_token_accuracy": 0.12100091725587844,
+      "num_tokens": 2864286.0,
+      "step": 1020
+    },
+    {
+      "entropy": 6.0084270477294925,
+      "epoch": 0.02774957698815567,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0004999289425887425,
+      "loss": 5.9948,
+      "mean_token_accuracy": 0.12859246730804444,
+      "num_tokens": 2878742.0,
+      "step": 1025
+    },
+    {
+      "entropy": 6.1349043369293215,
+      "epoch": 0.027884940778341794,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004998962537575161,
+      "loss": 5.9836,
+      "mean_token_accuracy": 0.1273471213877201,
+      "num_tokens": 2891855.0,
+      "step": 1030
+    },
+    {
+      "entropy": 5.931607675552368,
+      "epoch": 0.02802030456852792,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004998573992818874,
+      "loss": 5.9324,
+      "mean_token_accuracy": 0.1273692585527897,
+      "num_tokens": 2905629.0,
+      "step": 1035
+    },
+    {
+      "entropy": 6.158669471740723,
+      "epoch": 0.028155668358714043,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004998123802270715,
+      "loss": 6.0355,
+      "mean_token_accuracy": 0.12390008270740509,
+      "num_tokens": 2918358.0,
+      "step": 1040
+    },
+    {
+      "entropy": 6.0347072124481205,
+      "epoch": 0.02829103214890017,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.0004997611978272886,
+      "loss": 5.9851,
+      "mean_token_accuracy": 0.12076426520943642,
+      "num_tokens": 2931814.0,
+      "step": 1045
+    },
+    {
+      "entropy": 6.025428819656372,
+      "epoch": 0.028426395939086295,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.0004997038534857298,
+      "loss": 5.9487,
+      "mean_token_accuracy": 0.12540491446852683,
+      "num_tokens": 2945457.0,
+      "step": 1050
+    },
+    {
+      "entropy": 6.2262214660644535,
+      "epoch": 0.02856175972927242,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004996403487745194,
+      "loss": 6.0524,
+      "mean_token_accuracy": 0.12540149614214896,
+      "num_tokens": 2960790.0,
+      "step": 1055
+    },
+    {
+      "entropy": 6.016364479064942,
+      "epoch": 0.028697123519458544,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.000499570685434671,
+      "loss": 5.9462,
+      "mean_token_accuracy": 0.12737049981951715,
+      "num_tokens": 2972851.0,
+      "step": 1060
+    },
+    {
+      "entropy": 6.061508703231811,
+      "epoch": 0.02883248730964467,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004994948653760405,
+      "loss": 6.0193,
+      "mean_token_accuracy": 0.12827756106853486,
+      "num_tokens": 2986167.0,
+      "step": 1065
+    },
+    {
+      "entropy": 6.138202476501465,
+      "epoch": 0.028967851099830796,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004994128906772729,
+      "loss": 6.0747,
+      "mean_token_accuracy": 0.11862326189875602,
+      "num_tokens": 3001051.0,
+      "step": 1070
+    },
+    {
+      "entropy": 6.000337219238281,
+      "epoch": 0.02910321489001692,
+      "grad_norm": 1.0,
+      "learning_rate": 0.000499324763585746,
+      "loss": 5.8905,
+      "mean_token_accuracy": 0.1284564033150673,
+      "num_tokens": 3014393.0,
+      "step": 1075
+    },
+    {
+      "entropy": 5.964756202697754,
+      "epoch": 0.029238578680203044,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0004992304865175085,
+      "loss": 5.9813,
+      "mean_token_accuracy": 0.12596782818436622,
+      "num_tokens": 3029072.0,
+      "step": 1080
+    },
+    {
+      "entropy": 6.101922702789307,
+      "epoch": 0.029373942470389172,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004991300620572138,
+      "loss": 5.9816,
+      "mean_token_accuracy": 0.12776771411299706,
+      "num_tokens": 3041844.0,
+      "step": 1085
+    },
+    {
+      "entropy": 6.014832973480225,
+      "epoch": 0.029509306260575296,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0004990234929580494,
+      "loss": 6.0048,
+      "mean_token_accuracy": 0.12292343974113465,
+      "num_tokens": 3056570.0,
+      "step": 1090
+    },
+    {
+      "entropy": 6.118423938751221,
+      "epoch": 0.02964467005076142,
+      "grad_norm": 0.90234375,
+      "learning_rate": 0.0004989107821416609,
+      "loss": 6.0251,
+      "mean_token_accuracy": 0.12079276815056801,
+      "num_tokens": 3070407.0,
+      "step": 1095
+    },
+    {
+      "entropy": 6.080052423477173,
+      "epoch": 0.029780033840947545,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0004987919326980723,
+      "loss": 5.9603,
+      "mean_token_accuracy": 0.1254470780491829,
+      "num_tokens": 3083845.0,
+      "step": 1100
+    },
+    {
+      "entropy": 6.030987882614136,
+      "epoch": 0.029915397631133673,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.0004986669478856011,
+      "loss": 6.0241,
+      "mean_token_accuracy": 0.12584121003746987,
+      "num_tokens": 3097637.0,
+      "step": 1105
+    },
+    {
+      "entropy": 6.138483095169067,
+      "epoch": 0.030050761421319797,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004985358311307688,
+      "loss": 6.0386,
+      "mean_token_accuracy": 0.12023792713880539,
+      "num_tokens": 3112788.0,
+      "step": 1110
+    },
+    {
+      "entropy": 6.025177240371704,
+      "epoch": 0.03018612521150592,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.0004983985860282081,
+      "loss": 6.0024,
+      "mean_token_accuracy": 0.12313967794179917,
+      "num_tokens": 3125927.0,
+      "step": 1115
+    },
+    {
+      "entropy": 6.122660446166992,
+      "epoch": 0.030321489001692046,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004982552163405623,
+      "loss": 6.0601,
+      "mean_token_accuracy": 0.12063769325613975,
+      "num_tokens": 3141294.0,
+      "step": 1120
+    },
+    {
+      "entropy": 5.91901388168335,
+      "epoch": 0.030456852791878174,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0004981057259983839,
+      "loss": 5.8293,
+      "mean_token_accuracy": 0.1289824903011322,
+      "num_tokens": 3154662.0,
+      "step": 1125
+    },
+    {
+      "entropy": 6.029631614685059,
+      "epoch": 0.030592216582064298,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004979501191000262,
+      "loss": 5.9944,
+      "mean_token_accuracy": 0.1223167359828949,
+      "num_tokens": 3169356.0,
+      "step": 1130
+    },
+    {
+      "entropy": 6.0573914527893065,
+      "epoch": 0.030727580372250422,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004977883999115311,
+      "loss": 5.9411,
+      "mean_token_accuracy": 0.12958179041743279,
+      "num_tokens": 3183432.0,
+      "step": 1135
+    },
+    {
+      "entropy": 5.989719247817993,
+      "epoch": 0.030862944162436547,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.0004976205728665113,
+      "loss": 5.96,
+      "mean_token_accuracy": 0.13433775156736374,
+      "num_tokens": 3197440.0,
+      "step": 1140
+    },
+    {
+      "entropy": 5.956574392318726,
+      "epoch": 0.030998307952622674,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.0004974466425660307,
+      "loss": 5.9362,
+      "mean_token_accuracy": 0.12498937547206879,
+      "num_tokens": 3213348.0,
+      "step": 1145
+    },
+    {
+      "entropy": 6.026212120056153,
+      "epoch": 0.0311336717428088,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.0004972666137784759,
+      "loss": 5.9483,
+      "mean_token_accuracy": 0.1215012364089489,
+      "num_tokens": 3226835.0,
+      "step": 1150
+    },
+    {
+      "entropy": 5.991872882843017,
+      "epoch": 0.03126903553299493,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0004970804914394271,
+      "loss": 5.9253,
+      "mean_token_accuracy": 0.1271298736333847,
+      "num_tokens": 3240815.0,
+      "step": 1155
+    },
+    {
+      "entropy": 5.99332070350647,
+      "epoch": 0.03140439932318105,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004968882806515225,
+      "loss": 5.9465,
+      "mean_token_accuracy": 0.13174123987555503,
+      "num_tokens": 3254567.0,
+      "step": 1160
+    },
+    {
+      "entropy": 5.977180242538452,
+      "epoch": 0.031539763113367175,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004966899866843177,
+      "loss": 5.871,
+      "mean_token_accuracy": 0.13577421456575395,
+      "num_tokens": 3268802.0,
+      "step": 1165
+    },
+    {
+      "entropy": 6.024600458145142,
+      "epoch": 0.0316751269035533,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000496485614974142,
+      "loss": 5.9607,
+      "mean_token_accuracy": 0.12890158146619796,
+      "num_tokens": 3284215.0,
+      "step": 1170
+    },
+    {
+      "entropy": 5.917850065231323,
+      "epoch": 0.031810490693739424,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004962751711239492,
+      "loss": 5.9356,
+      "mean_token_accuracy": 0.12962086424231528,
+      "num_tokens": 3300179.0,
+      "step": 1175
+    },
+    {
+      "entropy": 6.076095533370972,
+      "epoch": 0.03194585448392555,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004960586609031636,
+      "loss": 5.9531,
+      "mean_token_accuracy": 0.12908913493156432,
+      "num_tokens": 3314282.0,
+      "step": 1180
+    },
+    {
+      "entropy": 5.97105712890625,
+      "epoch": 0.03208121827411167,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004958360902475224,
+      "loss": 6.0269,
+      "mean_token_accuracy": 0.12777508646249772,
+      "num_tokens": 3328343.0,
+      "step": 1185
+    },
+    {
+      "entropy": 6.013218975067138,
+      "epoch": 0.032216582064297804,
+      "grad_norm": 0.88671875,
+      "learning_rate": 0.0004956074652589125,
+      "loss": 5.8409,
+      "mean_token_accuracy": 0.13359294980764388,
+      "num_tokens": 3342751.0,
+      "step": 1190
+    },
+    {
+      "entropy": 5.94380931854248,
+      "epoch": 0.03235194585448393,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.0004953727922052035,
+      "loss": 5.8721,
+      "mean_token_accuracy": 0.1346684418618679,
+      "num_tokens": 3357406.0,
+      "step": 1195
+    },
+    {
+      "entropy": 6.028680276870728,
+      "epoch": 0.03248730964467005,
+      "grad_norm": 0.92578125,
+      "learning_rate": 0.0004951320775200756,
+      "loss": 5.9773,
+      "mean_token_accuracy": 0.12906075417995452,
+      "num_tokens": 3372090.0,
+      "step": 1200
+    },
+    {
+      "entropy": 6.016203355789185,
+      "epoch": 0.03262267343485618,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004948853278028436,
+      "loss": 5.9553,
+      "mean_token_accuracy": 0.1359055757522583,
+      "num_tokens": 3384007.0,
+      "step": 1205
+    },
+    {
+      "entropy": 5.911950778961182,
+      "epoch": 0.0327580372250423,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004946325498182755,
+      "loss": 5.9622,
+      "mean_token_accuracy": 0.12463960498571396,
+      "num_tokens": 3398653.0,
+      "step": 1210
+    },
+    {
+      "entropy": 6.000537252426147,
+      "epoch": 0.032893401015228425,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0004943737504964076,
+      "loss": 5.8577,
+      "mean_token_accuracy": 0.13106542453169823,
+      "num_tokens": 3412892.0,
+      "step": 1215
+    },
+    {
+      "entropy": 5.9030508518219,
+      "epoch": 0.03302876480541455,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.000494108936932354,
+      "loss": 5.7772,
+      "mean_token_accuracy": 0.14335963129997253,
+      "num_tokens": 3427162.0,
+      "step": 1220
+    },
+    {
+      "entropy": 5.881168079376221,
+      "epoch": 0.033164128595600674,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004938381163861124,
+      "loss": 5.8865,
+      "mean_token_accuracy": 0.1323596365749836,
+      "num_tokens": 3440402.0,
+      "step": 1225
+    },
+    {
+      "entropy": 5.966072845458984,
+      "epoch": 0.033299492385786805,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0004935612962823645,
+      "loss": 5.9626,
+      "mean_token_accuracy": 0.12719930335879326,
+      "num_tokens": 3453989.0,
+      "step": 1230
+    },
+    {
+      "entropy": 6.010539197921753,
+      "epoch": 0.03343485617597293,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0004932784842102739,
+      "loss": 5.9214,
+      "mean_token_accuracy": 0.12816225066781045,
+      "num_tokens": 3468962.0,
+      "step": 1235
+    },
+    {
+      "entropy": 5.930458641052246,
+      "epoch": 0.033570219966159054,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.0004929896879232758,
+      "loss": 5.7972,
+      "mean_token_accuracy": 0.1380744606256485,
+      "num_tokens": 3481385.0,
+      "step": 1240
+    },
+    {
+      "entropy": 5.931494283676147,
+      "epoch": 0.03370558375634518,
+      "grad_norm": 0.92578125,
+      "learning_rate": 0.0004926949153388668,
+      "loss": 5.8538,
+      "mean_token_accuracy": 0.13487453013658524,
+      "num_tokens": 3495904.0,
+      "step": 1245
+    },
+    {
+      "entropy": 5.931593751907348,
+      "epoch": 0.0338409475465313,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.0004923941745383859,
+      "loss": 5.8606,
+      "mean_token_accuracy": 0.12867458090186118,
+      "num_tokens": 3511946.0,
+      "step": 1250
+    },
+    {
+      "entropy": 5.901215839385986,
+      "epoch": 0.03397631133671743,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.000492087473766794,
+      "loss": 5.8359,
+      "mean_token_accuracy": 0.13469822704792023,
+      "num_tokens": 3524470.0,
+      "step": 1255
+    },
+    {
+      "entropy": 5.945947980880737,
+      "epoch": 0.03411167512690355,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.000491774821432448,
+      "loss": 5.9056,
+      "mean_token_accuracy": 0.13596510142087936,
+      "num_tokens": 3537444.0,
+      "step": 1260
+    },
+    {
+      "entropy": 6.070478630065918,
+      "epoch": 0.034247038917089676,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004914562261068693,
+      "loss": 6.0304,
+      "mean_token_accuracy": 0.12992693334817887,
+      "num_tokens": 3551171.0,
+      "step": 1265
+    },
+    {
+      "entropy": 5.8572030544281,
+      "epoch": 0.03438240270727581,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.0004911316965245098,
+      "loss": 5.8561,
+      "mean_token_accuracy": 0.13198562785983087,
+      "num_tokens": 3565608.0,
+      "step": 1270
+    },
+    {
+      "entropy": 5.899595928192139,
+      "epoch": 0.03451776649746193,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.000490801241582512,
+      "loss": 5.8328,
+      "mean_token_accuracy": 0.1360912725329399,
+      "num_tokens": 3580423.0,
+      "step": 1275
+    },
+    {
+      "entropy": 5.88170690536499,
+      "epoch": 0.034653130287648055,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000490464870340465,
+      "loss": 5.8666,
+      "mean_token_accuracy": 0.1351574994623661,
+      "num_tokens": 3594068.0,
+      "step": 1280
+    },
+    {
+      "entropy": 5.838206243515015,
+      "epoch": 0.03478849407783418,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.0004901225920201563,
+      "loss": 5.8293,
+      "mean_token_accuracy": 0.13220815882086753,
+      "num_tokens": 3607978.0,
+      "step": 1285
+    },
+    {
+      "entropy": 5.945906925201416,
+      "epoch": 0.034923857868020304,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.000489774416005319,
+      "loss": 5.9087,
+      "mean_token_accuracy": 0.1275140456855297,
+      "num_tokens": 3621091.0,
+      "step": 1290
+    },
+    {
+      "entropy": 5.873530435562134,
+      "epoch": 0.03505922165820643,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.0004894203518413742,
+      "loss": 5.8965,
+      "mean_token_accuracy": 0.13213202804327012,
+      "num_tokens": 3634615.0,
+      "step": 1295
+    },
+    {
+      "entropy": 5.950916433334351,
+      "epoch": 0.03519458544839255,
+      "grad_norm": 0.9296875,
+      "learning_rate": 0.0004890604092351701,
+      "loss": 5.8203,
+      "mean_token_accuracy": 0.1343979686498642,
+      "num_tokens": 3648952.0,
+      "step": 1300
+    },
+    {
+      "entropy": 5.9373682022094725,
+      "epoch": 0.03532994923857868,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000488694598054715,
+      "loss": 5.9016,
+      "mean_token_accuracy": 0.1337265945971012,
+      "num_tokens": 3662631.0,
+      "step": 1305
+    },
+    {
+      "entropy": 5.8529746532440186,
+      "epoch": 0.03546531302876481,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0004883229283289071,
+      "loss": 5.8703,
+      "mean_token_accuracy": 0.13501393496990205,
+      "num_tokens": 3677762.0,
+      "step": 1310
+    },
+    {
+      "entropy": 5.979293251037598,
+      "epoch": 0.03560067681895093,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.00048794541024725993,
+      "loss": 5.8974,
+      "mean_token_accuracy": 0.12559727132320403,
+      "num_tokens": 3692657.0,
+      "step": 1315
+    },
+    {
+      "entropy": 5.79110279083252,
+      "epoch": 0.03573604060913706,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0004875620541596221,
+      "loss": 5.7665,
+      "mean_token_accuracy": 0.13931511640548705,
+      "num_tokens": 3706615.0,
+      "step": 1320
+    },
+    {
+      "entropy": 5.879328918457031,
+      "epoch": 0.03587140439932318,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.00048717287057589454,
+      "loss": 5.7665,
+      "mean_token_accuracy": 0.13615642935037614,
+      "num_tokens": 3720054.0,
+      "step": 1325
+    },
+    {
+      "entropy": 5.7798095703125,
+      "epoch": 0.036006768189509306,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0004867778701657417,
+      "loss": 5.8004,
+      "mean_token_accuracy": 0.13352179080247878,
+      "num_tokens": 3732578.0,
+      "step": 1330
+    },
+    {
+      "entropy": 5.896574926376343,
+      "epoch": 0.03614213197969543,
+      "grad_norm": 0.8671875,
+      "learning_rate": 0.00048637706375829955,
+      "loss": 5.8389,
+      "mean_token_accuracy": 0.13618992641568184,
+      "num_tokens": 3747745.0,
+      "step": 1335
+    },
+    {
+      "entropy": 5.925070810317993,
+      "epoch": 0.036277495769881554,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.000485970462341878,
+      "loss": 5.8093,
+      "mean_token_accuracy": 0.1370240181684494,
+      "num_tokens": 3760715.0,
+      "step": 1340
+    },
+    {
+      "entropy": 5.853664588928223,
+      "epoch": 0.03641285956006768,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00048555807706366044,
+      "loss": 5.8601,
+      "mean_token_accuracy": 0.13756127581000327,
+      "num_tokens": 3773670.0,
+      "step": 1345
+    },
+    {
+      "entropy": 5.945131015777588,
+      "epoch": 0.03654822335025381,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00048513991922939756,
+      "loss": 5.7907,
+      "mean_token_accuracy": 0.14311188608407974,
+      "num_tokens": 3786738.0,
+      "step": 1350
+    },
+    {
+      "entropy": 5.859251689910889,
+      "epoch": 0.036683587140439934,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00048471600030309744,
+      "loss": 5.7949,
+      "mean_token_accuracy": 0.14019912704825402,
+      "num_tokens": 3800536.0,
+      "step": 1355
+    },
+    {
+      "entropy": 5.906093645095825,
+      "epoch": 0.03681895093062606,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00048428633190671186,
+      "loss": 5.917,
+      "mean_token_accuracy": 0.13591600805521012,
+      "num_tokens": 3813241.0,
+      "step": 1360
+    },
+    {
+      "entropy": 5.841483879089355,
+      "epoch": 0.03695431472081218,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0004838509258198167,
+      "loss": 5.8021,
+      "mean_token_accuracy": 0.14016170799732208,
+      "num_tokens": 3827311.0,
+      "step": 1365
+    },
+    {
+      "entropy": 5.877211189270019,
+      "epoch": 0.03708967851099831,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00048340979397929,
+      "loss": 5.867,
+      "mean_token_accuracy": 0.13220226317644118,
+      "num_tokens": 3841034.0,
+      "step": 1370
+    },
+    {
+      "entropy": 5.83872618675232,
+      "epoch": 0.03722504230118443,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00048296294847898386,
+      "loss": 5.6653,
+      "mean_token_accuracy": 0.13773941323161126,
+      "num_tokens": 3853559.0,
+      "step": 1375
+    },
+    {
+      "entropy": 5.855471277236939,
+      "epoch": 0.037360406091370556,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.0004825104015693934,
+      "loss": 5.864,
+      "mean_token_accuracy": 0.12592799887061118,
+      "num_tokens": 3868327.0,
+      "step": 1380
+    },
+    {
+      "entropy": 5.8454270362854,
+      "epoch": 0.03749576988155669,
+      "grad_norm": 0.9296875,
+      "learning_rate": 0.0004820521656573208,
+      "loss": 5.8182,
+      "mean_token_accuracy": 0.13926773220300676,
+      "num_tokens": 3881981.0,
+      "step": 1385
+    },
+    {
+      "entropy": 5.975482940673828,
+      "epoch": 0.03763113367174281,
+      "grad_norm": 0.92578125,
+      "learning_rate": 0.00048158825330553505,
+      "loss": 5.8992,
+      "mean_token_accuracy": 0.13271776139736174,
+      "num_tokens": 3895911.0,
+      "step": 1390
+    },
+    {
+      "entropy": 5.823576545715332,
+      "epoch": 0.037766497461928936,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.00048111867723242763,
+      "loss": 5.797,
+      "mean_token_accuracy": 0.1352679081261158,
+      "num_tokens": 3910179.0,
+      "step": 1395
+    },
+    {
+      "entropy": 5.880249977111816,
+      "epoch": 0.03790186125211506,
+      "grad_norm": 0.85546875,
+      "learning_rate": 0.0004806434503116637,
+      "loss": 5.8109,
+      "mean_token_accuracy": 0.13052432835102082,
+      "num_tokens": 3925053.0,
+      "step": 1400
+    },
+    {
+      "entropy": 5.788560056686402,
+      "epoch": 0.038037225042301184,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004801625855718296,
+      "loss": 5.8081,
+      "mean_token_accuracy": 0.13492835760116578,
+      "num_tokens": 3938544.0,
+      "step": 1405
+    },
+    {
+      "entropy": 5.931741094589233,
+      "epoch": 0.03817258883248731,
+      "grad_norm": 0.91796875,
+      "learning_rate": 0.00047967609619607477,
+      "loss": 5.9446,
+      "mean_token_accuracy": 0.12690417915582658,
+      "num_tokens": 3953088.0,
+      "step": 1410
+    },
+    {
+      "entropy": 5.938955354690552,
+      "epoch": 0.03830795262267343,
+      "grad_norm": 0.84765625,
+      "learning_rate": 0.0004791839955217513,
+      "loss": 5.9114,
+      "mean_token_accuracy": 0.13125399351119996,
+      "num_tokens": 3970136.0,
+      "step": 1415
+    },
+    {
+      "entropy": 5.8262474060058596,
+      "epoch": 0.03844331641285956,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00047868629704004786,
+      "loss": 5.7126,
+      "mean_token_accuracy": 0.14028473943471909,
+      "num_tokens": 3984170.0,
+      "step": 1420
+    },
+    {
+      "entropy": 5.777353525161743,
+      "epoch": 0.03857868020304569,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.00047818301439561965,
+      "loss": 5.7676,
+      "mean_token_accuracy": 0.14551205635070802,
+      "num_tokens": 3998030.0,
+      "step": 1425
+    },
+    {
+      "entropy": 5.755202388763427,
+      "epoch": 0.03871404399323181,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00047767416138621454,
+      "loss": 5.6717,
+      "mean_token_accuracy": 0.13998118191957473,
+      "num_tokens": 4013158.0,
+      "step": 1430
+    },
+    {
+      "entropy": 5.910610723495483,
+      "epoch": 0.03884940778341794,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.000477159751962295,
+      "loss": 5.8218,
+      "mean_token_accuracy": 0.13360931798815728,
+      "num_tokens": 4028147.0,
+      "step": 1435
+    },
+    {
+      "entropy": 5.76572093963623,
+      "epoch": 0.03898477157360406,
+      "grad_norm": 0.9765625,
+      "learning_rate": 0.00047663980022665507,
+      "loss": 5.6998,
+      "mean_token_accuracy": 0.14273363947868348,
+      "num_tokens": 4043039.0,
+      "step": 1440
+    },
+    {
+      "entropy": 5.883444166183471,
+      "epoch": 0.039120135363790186,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00047611432043403437,
+      "loss": 5.8944,
+      "mean_token_accuracy": 0.13942254111170768,
+      "num_tokens": 4056414.0,
+      "step": 1445
+    },
+    {
+      "entropy": 5.877360773086548,
+      "epoch": 0.03925549915397631,
+      "grad_norm": 0.92578125,
+      "learning_rate": 0.0004755833269907267,
+      "loss": 5.8143,
+      "mean_token_accuracy": 0.1393973112106323,
+      "num_tokens": 4071776.0,
+      "step": 1450
+    },
+    {
+      "entropy": 5.851318216323852,
+      "epoch": 0.039390862944162434,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.0004750468344541857,
+      "loss": 5.7161,
+      "mean_token_accuracy": 0.13444317877292633,
+      "num_tokens": 4085216.0,
+      "step": 1455
+    },
+    {
+      "entropy": 5.749738121032715,
+      "epoch": 0.03952622673434856,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00047450485753262525,
+      "loss": 5.8215,
+      "mean_token_accuracy": 0.14175896048545839,
+      "num_tokens": 4098523.0,
+      "step": 1460
+    },
+    {
+      "entropy": 5.883940315246582,
+      "epoch": 0.03966159052453469,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.00047395741108461633,
+      "loss": 5.7887,
+      "mean_token_accuracy": 0.13734084963798524,
+      "num_tokens": 4112383.0,
+      "step": 1465
+    },
+    {
+      "entropy": 5.746815729141235,
+      "epoch": 0.039796954314720814,
+      "grad_norm": 0.9765625,
+      "learning_rate": 0.00047340451011867985,
+      "loss": 5.8546,
+      "mean_token_accuracy": 0.1313462048768997,
+      "num_tokens": 4125524.0,
+      "step": 1470
+    },
+    {
+      "entropy": 5.8694652080535885,
+      "epoch": 0.03993231810490694,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.00047284616979287515,
+      "loss": 5.7026,
+      "mean_token_accuracy": 0.14162963032722473,
+      "num_tokens": 4138946.0,
+      "step": 1475
+    },
+    {
+      "entropy": 5.773919725418091,
+      "epoch": 0.04006768189509306,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00047228240541438433,
+      "loss": 5.7453,
+      "mean_token_accuracy": 0.14252977669239045,
+      "num_tokens": 4152066.0,
+      "step": 1480
+    },
+    {
+      "entropy": 5.838653135299682,
+      "epoch": 0.04020304568527919,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.00047171323243909257,
+      "loss": 5.7556,
+      "mean_token_accuracy": 0.1399710148572922,
+      "num_tokens": 4166171.0,
+      "step": 1485
+    },
+    {
+      "entropy": 5.763364934921265,
+      "epoch": 0.04033840947546531,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00047113866647116457,
+      "loss": 5.7782,
+      "mean_token_accuracy": 0.14722683578729628,
+      "num_tokens": 4178809.0,
+      "step": 1490
+    },
+    {
+      "entropy": 5.734166765213013,
+      "epoch": 0.040473773265651436,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004705587232626164,
+      "loss": 5.7097,
+      "mean_token_accuracy": 0.1362159803509712,
+      "num_tokens": 4192211.0,
+      "step": 1495
+    },
+    {
+      "entropy": 5.872725915908814,
+      "epoch": 0.04060913705583756,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.00046997341871288424,
+      "loss": 5.7196,
+      "mean_token_accuracy": 0.14548214226961137,
+      "num_tokens": 4206310.0,
+      "step": 1500
+    },
+    {
+      "entropy": 5.856945276260376,
+      "epoch": 0.04074450084602369,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0004693827688683879,
+      "loss": 5.8553,
+      "mean_token_accuracy": 0.14121681377291678,
+      "num_tokens": 4219479.0,
+      "step": 1505
+    },
+    {
+      "entropy": 5.895747375488281,
+      "epoch": 0.040879864636209816,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.0004687867899220914,
+      "loss": 5.8104,
+      "mean_token_accuracy": 0.1397186741232872,
+      "num_tokens": 4234129.0,
+      "step": 1510
+    },
+    {
+      "entropy": 5.804001092910767,
+      "epoch": 0.04101522842639594,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00046818549821305846,
+      "loss": 5.7749,
+      "mean_token_accuracy": 0.13784123361110687,
+      "num_tokens": 4247785.0,
+      "step": 1515
+    },
+    {
+      "entropy": 5.728389310836792,
+      "epoch": 0.041150592216582065,
+      "grad_norm": 0.91015625,
+      "learning_rate": 0.00046757891022600494,
+      "loss": 5.6397,
+      "mean_token_accuracy": 0.14299456030130386,
+      "num_tokens": 4262983.0,
+      "step": 1520
+    },
+    {
+      "entropy": 5.9353930950164795,
+      "epoch": 0.04128595600676819,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.0004669670425908471,
+      "loss": 5.9331,
+      "mean_token_accuracy": 0.1264545239508152,
+      "num_tokens": 4278373.0,
+      "step": 1525
+    },
+    {
+      "entropy": 5.817500162124634,
+      "epoch": 0.04142131979695431,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.0004663499120822451,
+      "loss": 5.6499,
+      "mean_token_accuracy": 0.14248120337724685,
+      "num_tokens": 4293058.0,
+      "step": 1530
+    },
+    {
+      "entropy": 5.697978162765503,
+      "epoch": 0.04155668358714044,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.0004657275356191437,
+      "loss": 5.7497,
+      "mean_token_accuracy": 0.14827215373516084,
+      "num_tokens": 4305107.0,
+      "step": 1535
+    },
+    {
+      "entropy": 5.7474280834198,
+      "epoch": 0.04169204737732656,
+      "grad_norm": 0.875,
+      "learning_rate": 0.00046509993026430804,
+      "loss": 5.7442,
+      "mean_token_accuracy": 0.14127848967909812,
+      "num_tokens": 4320791.0,
+      "step": 1540
+    },
+    {
+      "entropy": 5.863645458221436,
+      "epoch": 0.04182741116751269,
+      "grad_norm": 0.859375,
+      "learning_rate": 0.0004644671132238558,
+      "loss": 5.7167,
+      "mean_token_accuracy": 0.13956723287701606,
+      "num_tokens": 4336989.0,
+      "step": 1545
+    },
+    {
+      "entropy": 5.8170019626617435,
+      "epoch": 0.04196277495769882,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00046382910184678585,
+      "loss": 5.6923,
+      "mean_token_accuracy": 0.1433095797896385,
+      "num_tokens": 4350286.0,
+      "step": 1550
+    },
+    {
+      "entropy": 5.611237478256226,
+      "epoch": 0.04209813874788494,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0004631859136245025,
+      "loss": 5.6368,
+      "mean_token_accuracy": 0.14433954954147338,
+      "num_tokens": 4366115.0,
+      "step": 1555
+    },
+    {
+      "entropy": 5.8419585704803465,
+      "epoch": 0.042233502538071066,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0004625375661903357,
+      "loss": 5.8694,
+      "mean_token_accuracy": 0.13394705951213837,
+      "num_tokens": 4381538.0,
+      "step": 1560
+    },
+    {
+      "entropy": 5.869414138793945,
+      "epoch": 0.04236886632825719,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00046188407731905787,
+      "loss": 5.8672,
+      "mean_token_accuracy": 0.13035987243056296,
+      "num_tokens": 4395460.0,
+      "step": 1565
+    },
+    {
+      "entropy": 5.894603872299195,
+      "epoch": 0.042504230118443315,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.00046122546492639643,
+      "loss": 5.6597,
+      "mean_token_accuracy": 0.14258980005979538,
+      "num_tokens": 4411841.0,
+      "step": 1570
+    },
+    {
+      "entropy": 5.751853179931641,
+      "epoch": 0.04263959390862944,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.000460561747068543,
+      "loss": 5.7583,
+      "mean_token_accuracy": 0.138499815762043,
+      "num_tokens": 4425555.0,
+      "step": 1575
+    },
+    {
+      "entropy": 5.86604118347168,
+      "epoch": 0.04277495769881557,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.0004598929419416578,
+      "loss": 5.6867,
+      "mean_token_accuracy": 0.14240610525012015,
+      "num_tokens": 4437822.0,
+      "step": 1580
+    },
+    {
+      "entropy": 5.737030029296875,
+      "epoch": 0.042910321489001695,
+      "grad_norm": 0.890625,
+      "learning_rate": 0.00045921906788137123,
+      "loss": 5.7327,
+      "mean_token_accuracy": 0.14339831471443176,
+      "num_tokens": 4452974.0,
+      "step": 1585
+    },
+    {
+      "entropy": 5.74243483543396,
+      "epoch": 0.04304568527918782,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.00045854014336228115,
+      "loss": 5.6716,
+      "mean_token_accuracy": 0.1469262346625328,
+      "num_tokens": 4465057.0,
+      "step": 1590
+    },
+    {
+      "entropy": 5.721040344238281,
+      "epoch": 0.04318104906937394,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00045785618699744615,
+      "loss": 5.6387,
+      "mean_token_accuracy": 0.14491022154688835,
+      "num_tokens": 4479131.0,
+      "step": 1595
+    },
+    {
+      "entropy": 5.748002433776856,
+      "epoch": 0.04331641285956007,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00045716721753787543,
+      "loss": 5.7572,
+      "mean_token_accuracy": 0.14127787053585053,
+      "num_tokens": 4495131.0,
+      "step": 1600
+    },
+    {
+      "entropy": 5.711660146713257,
+      "epoch": 0.04345177664974619,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.0004564732538720148,
+      "loss": 5.7059,
+      "mean_token_accuracy": 0.14483382552862167,
+      "num_tokens": 4508267.0,
+      "step": 1605
+    },
+    {
+      "entropy": 5.848050451278686,
+      "epoch": 0.043587140439932316,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00045577431502522877,
+      "loss": 5.7053,
+      "mean_token_accuracy": 0.14630670472979546,
+      "num_tokens": 4522298.0,
+      "step": 1610
+    },
+    {
+      "entropy": 5.725667428970337,
+      "epoch": 0.04372250423011844,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0004550704201592787,
+      "loss": 5.7321,
+      "mean_token_accuracy": 0.1426356054842472,
+      "num_tokens": 4536509.0,
+      "step": 1615
+    },
+    {
+      "entropy": 5.708344316482544,
+      "epoch": 0.04385786802030457,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0004543615885717981,
+      "loss": 5.5904,
+      "mean_token_accuracy": 0.1510692059993744,
+      "num_tokens": 4548164.0,
+      "step": 1620
+    },
+    {
+      "entropy": 5.699061918258667,
+      "epoch": 0.043993231810490696,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.00045364783969576296,
+      "loss": 5.6169,
+      "mean_token_accuracy": 0.14389338046312333,
+      "num_tokens": 4562588.0,
+      "step": 1625
+    },
+    {
+      "entropy": 5.745128536224366,
+      "epoch": 0.04412859560067682,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.0004529291930989592,
+      "loss": 5.7788,
+      "mean_token_accuracy": 0.14362882375717162,
+      "num_tokens": 4575607.0,
+      "step": 1630
+    },
+    {
+      "entropy": 5.842718124389648,
+      "epoch": 0.044263959390862945,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.0004522056684834464,
+      "loss": 5.6764,
+      "mean_token_accuracy": 0.1434960126876831,
+      "num_tokens": 4590102.0,
+      "step": 1635
+    },
+    {
+      "entropy": 5.678610563278198,
+      "epoch": 0.04439932318104907,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.0004514772856850173,
+      "loss": 5.7023,
+      "mean_token_accuracy": 0.14198361933231354,
+      "num_tokens": 4604192.0,
+      "step": 1640
+    },
+    {
+      "entropy": 5.788250541687011,
+      "epoch": 0.04453468697123519,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0004507440646726542,
+      "loss": 5.6749,
+      "mean_token_accuracy": 0.14550164267420768,
+      "num_tokens": 4617643.0,
+      "step": 1645
+    },
+    {
+      "entropy": 5.704812479019165,
+      "epoch": 0.04467005076142132,
+      "grad_norm": 0.890625,
+      "learning_rate": 0.0004500060255479818,
+      "loss": 5.7478,
+      "mean_token_accuracy": 0.139235407859087,
+      "num_tokens": 4633818.0,
+      "step": 1650
+    },
+    {
+      "entropy": 5.7535816669464115,
+      "epoch": 0.04480541455160744,
+      "grad_norm": 0.82421875,
+      "learning_rate": 0.0004492631885447151,
+      "loss": 5.7359,
+      "mean_token_accuracy": 0.13843510597944259,
+      "num_tokens": 4648527.0,
+      "step": 1655
+    },
+    {
+      "entropy": 5.807849359512329,
+      "epoch": 0.04494077834179357,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00044851557402810616,
+      "loss": 5.6926,
+      "mean_token_accuracy": 0.1437424451112747,
+      "num_tokens": 4661545.0,
+      "step": 1660
+    },
+    {
+      "entropy": 5.759947490692139,
+      "epoch": 0.0450761421319797,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00044776320249438444,
+      "loss": 5.7466,
+      "mean_token_accuracy": 0.14138651117682458,
+      "num_tokens": 4675255.0,
+      "step": 1665
+    },
+    {
+      "entropy": 5.719445466995239,
+      "epoch": 0.04521150592216582,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.00044700609457019565,
+      "loss": 5.5664,
+      "mean_token_accuracy": 0.15053462088108063,
+      "num_tokens": 4688863.0,
+      "step": 1670
+    },
+    {
+      "entropy": 5.777239894866943,
+      "epoch": 0.045346869712351946,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0004462442710120359,
+      "loss": 5.73,
+      "mean_token_accuracy": 0.14415158182382584,
+      "num_tokens": 4702336.0,
+      "step": 1675
+    },
+    {
+      "entropy": 5.72534589767456,
+      "epoch": 0.04548223350253807,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000445477752705683,
+      "loss": 5.5325,
+      "mean_token_accuracy": 0.14911517053842543,
+      "num_tokens": 4716433.0,
+      "step": 1680
+    },
+    {
+      "entropy": 5.71527886390686,
+      "epoch": 0.045617597292724195,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.00044470656066562336,
+      "loss": 5.7004,
+      "mean_token_accuracy": 0.14622605815529824,
+      "num_tokens": 4730458.0,
+      "step": 1685
+    },
+    {
+      "entropy": 5.669246053695678,
+      "epoch": 0.04575296108291032,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004439307160344765,
+      "loss": 5.6398,
+      "mean_token_accuracy": 0.14573194235563278,
+      "num_tokens": 4744262.0,
+      "step": 1690
+    },
+    {
+      "entropy": 5.70993275642395,
+      "epoch": 0.045888324873096444,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00044315024008241473,
+      "loss": 5.6693,
+      "mean_token_accuracy": 0.14355338960886002,
+      "num_tokens": 4758673.0,
+      "step": 1695
+    },
+    {
+      "entropy": 5.847352313995361,
+      "epoch": 0.046023688663282575,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.0004423651542065806,
+      "loss": 5.6926,
+      "mean_token_accuracy": 0.14469362199306487,
+      "num_tokens": 4772220.0,
+      "step": 1700
+    },
+    {
+      "entropy": 5.645872116088867,
+      "epoch": 0.0461590524534687,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00044157547993050006,
+      "loss": 5.5701,
+      "mean_token_accuracy": 0.1518155261874199,
+      "num_tokens": 4785273.0,
+      "step": 1705
+    },
+    {
+      "entropy": 5.716040754318238,
+      "epoch": 0.04629441624365482,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.00044078123890349227,
+      "loss": 5.6269,
+      "mean_token_accuracy": 0.1508127622306347,
+      "num_tokens": 4799530.0,
+      "step": 1710
+    },
+    {
+      "entropy": 5.730141544342041,
+      "epoch": 0.04642978003384095,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00043998245290007606,
+      "loss": 5.6882,
+      "mean_token_accuracy": 0.141463815420866,
+      "num_tokens": 4813985.0,
+      "step": 1715
+    },
+    {
+      "entropy": 5.686599445343018,
+      "epoch": 0.04656514382402707,
+      "grad_norm": 0.94140625,
+      "learning_rate": 0.00043917914381937323,
+      "loss": 5.6193,
+      "mean_token_accuracy": 0.14998584240674973,
+      "num_tokens": 4826131.0,
+      "step": 1720
+    },
+    {
+      "entropy": 5.733201026916504,
+      "epoch": 0.046700507614213196,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.00043837133368450815,
+      "loss": 5.6939,
+      "mean_token_accuracy": 0.1447956845164299,
+      "num_tokens": 4840854.0,
+      "step": 1725
+    },
+    {
+      "entropy": 5.695008659362793,
+      "epoch": 0.04683587140439932,
+      "grad_norm": 0.9765625,
+      "learning_rate": 0.0004375590446420037,
+      "loss": 5.6338,
+      "mean_token_accuracy": 0.146064093708992,
+      "num_tokens": 4855690.0,
+      "step": 1730
+    },
+    {
+      "entropy": 5.772643423080444,
+      "epoch": 0.046971235194585445,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.0004367422989611743,
+      "loss": 5.6561,
+      "mean_token_accuracy": 0.15210988670587539,
+      "num_tokens": 4868909.0,
+      "step": 1735
+    },
+    {
+      "entropy": 5.729992532730103,
+      "epoch": 0.047106598984771576,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004359211190335153,
+      "loss": 5.6925,
+      "mean_token_accuracy": 0.1468656025826931,
+      "num_tokens": 4883021.0,
+      "step": 1740
+    },
+    {
+      "entropy": 5.736793279647827,
+      "epoch": 0.0472419627749577,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00043509552737208923,
+      "loss": 5.6105,
+      "mean_token_accuracy": 0.14630509316921234,
+      "num_tokens": 4897056.0,
+      "step": 1745
+    },
+    {
+      "entropy": 5.618852758407593,
+      "epoch": 0.047377326565143825,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00043426554661090853,
+      "loss": 5.6044,
+      "mean_token_accuracy": 0.14477996677160263,
+      "num_tokens": 4910514.0,
+      "step": 1750
+    },
+    {
+      "entropy": 5.719070625305176,
+      "epoch": 0.04751269035532995,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00043343119950431516,
+      "loss": 5.6632,
+      "mean_token_accuracy": 0.14679592475295067,
+      "num_tokens": 4923888.0,
+      "step": 1755
+    },
+    {
+      "entropy": 5.694540119171142,
+      "epoch": 0.047648054145516074,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.00043259250892635644,
+      "loss": 5.7051,
+      "mean_token_accuracy": 0.1482004776597023,
+      "num_tokens": 4939282.0,
+      "step": 1760
+    },
+    {
+      "entropy": 5.804504442214966,
+      "epoch": 0.0477834179357022,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.0004317494978701582,
+      "loss": 5.6336,
+      "mean_token_accuracy": 0.14055619165301322,
+      "num_tokens": 4954903.0,
+      "step": 1765
+    },
+    {
+      "entropy": 5.703070640563965,
+      "epoch": 0.04791878172588832,
+      "grad_norm": 0.9296875,
+      "learning_rate": 0.0004309021894472943,
+      "loss": 5.6719,
+      "mean_token_accuracy": 0.15038193613290787,
+      "num_tokens": 4969840.0,
+      "step": 1770
+    },
+    {
+      "entropy": 5.734935092926025,
+      "epoch": 0.048054145516074454,
+      "grad_norm": 0.89453125,
+      "learning_rate": 0.0004300506068871534,
+      "loss": 5.6401,
+      "mean_token_accuracy": 0.14663783758878707,
+      "num_tokens": 4984608.0,
+      "step": 1775
+    },
+    {
+      "entropy": 5.63493800163269,
+      "epoch": 0.04818950930626058,
+      "grad_norm": 0.8125,
+      "learning_rate": 0.00042919477353630135,
+      "loss": 5.5685,
+      "mean_token_accuracy": 0.1455909326672554,
+      "num_tokens": 5000369.0,
+      "step": 1780
+    },
+    {
+      "entropy": 5.715186595916748,
+      "epoch": 0.0483248730964467,
+      "grad_norm": 0.91015625,
+      "learning_rate": 0.000428334712857842,
+      "loss": 5.5859,
+      "mean_token_accuracy": 0.14593576341867448,
+      "num_tokens": 5014858.0,
+      "step": 1785
+    },
+    {
+      "entropy": 5.569622945785523,
+      "epoch": 0.048460236886632826,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.00042747044843077304,
+      "loss": 5.6116,
+      "mean_token_accuracy": 0.1430406704545021,
+      "num_tokens": 5029339.0,
+      "step": 1790
+    },
+    {
+      "entropy": 5.745399045944214,
+      "epoch": 0.04859560067681895,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00042660200394934047,
+      "loss": 5.6182,
+      "mean_token_accuracy": 0.14943531677126884,
+      "num_tokens": 5043065.0,
+      "step": 1795
+    },
+    {
+      "entropy": 5.723861646652222,
+      "epoch": 0.048730964467005075,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.00042572940322238844,
+      "loss": 5.6695,
+      "mean_token_accuracy": 0.14552638530731202,
+      "num_tokens": 5057186.0,
+      "step": 1800
+    },
+    {
+      "entropy": 5.697584199905395,
+      "epoch": 0.0488663282571912,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00042485267017270664,
+      "loss": 5.6677,
+      "mean_token_accuracy": 0.14745727330446243,
+      "num_tokens": 5070656.0,
+      "step": 1805
+    },
+    {
+      "entropy": 5.696168994903564,
+      "epoch": 0.049001692047377324,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.0004239718288363745,
+      "loss": 5.5746,
+      "mean_token_accuracy": 0.15220395475625992,
+      "num_tokens": 5084586.0,
+      "step": 1810
+    },
+    {
+      "entropy": 5.629924297332764,
+      "epoch": 0.049137055837563455,
+      "grad_norm": 0.89453125,
+      "learning_rate": 0.0004230869033621023,
+      "loss": 5.5757,
+      "mean_token_accuracy": 0.14787699431180953,
+      "num_tokens": 5098994.0,
+      "step": 1815
+    },
+    {
+      "entropy": 5.748041343688965,
+      "epoch": 0.04927241962774958,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0004221979180105688,
+      "loss": 5.6422,
+      "mean_token_accuracy": 0.14642422050237655,
+      "num_tokens": 5112063.0,
+      "step": 1820
+    },
+    {
+      "entropy": 5.589199495315552,
+      "epoch": 0.049407783417935704,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00042130489715375645,
+      "loss": 5.5928,
+      "mean_token_accuracy": 0.14088452607393265,
+      "num_tokens": 5125114.0,
+      "step": 1825
+    },
+    {
+      "entropy": 5.657587385177612,
+      "epoch": 0.04954314720812183,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00042040786527428335,
+      "loss": 5.5843,
+      "mean_token_accuracy": 0.1488766610622406,
+      "num_tokens": 5137439.0,
+      "step": 1830
+    },
+    {
+      "entropy": 5.620833206176758,
+      "epoch": 0.04967851099830795,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0004195068469647315,
+      "loss": 5.5275,
+      "mean_token_accuracy": 0.15381399989128114,
+      "num_tokens": 5151456.0,
+      "step": 1835
+    },
+    {
+      "entropy": 5.56364073753357,
+      "epoch": 0.04981387478849408,
+      "grad_norm": 0.921875,
+      "learning_rate": 0.00041860186692697297,
+      "loss": 5.5334,
+      "mean_token_accuracy": 0.1509672984480858,
+      "num_tokens": 5166494.0,
+      "step": 1840
+    },
+    {
+      "entropy": 5.724608564376831,
+      "epoch": 0.0499492385786802,
+      "grad_norm": 0.921875,
+      "learning_rate": 0.00041769294997149264,
+      "loss": 5.6809,
+      "mean_token_accuracy": 0.1437281422317028,
+      "num_tokens": 5182077.0,
+      "step": 1845
+    },
+    {
+      "entropy": 5.814406871795654,
+      "epoch": 0.050084602368866325,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0004167801210167081,
+      "loss": 5.7406,
+      "mean_token_accuracy": 0.14046868160367013,
+      "num_tokens": 5195702.0,
+      "step": 1850
+    },
+    {
+      "entropy": 5.645143127441406,
+      "epoch": 0.05021996615905246,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0004158634050882861,
+      "loss": 5.5753,
+      "mean_token_accuracy": 0.149927581846714,
+      "num_tokens": 5209855.0,
+      "step": 1855
+    },
+    {
+      "entropy": 5.817540025711059,
+      "epoch": 0.05035532994923858,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0004149428273184569,
+      "loss": 5.6542,
+      "mean_token_accuracy": 0.14699074774980544,
+      "num_tokens": 5224585.0,
+      "step": 1860
+    },
+    {
+      "entropy": 5.497271156311035,
+      "epoch": 0.050490693739424705,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.0004140184129453253,
+      "loss": 5.4861,
+      "mean_token_accuracy": 0.1623205564916134,
+      "num_tokens": 5238681.0,
+      "step": 1865
+    },
+    {
+      "entropy": 5.642212295532227,
+      "epoch": 0.05062605752961083,
+      "grad_norm": 0.87109375,
+      "learning_rate": 0.000413090187312178,
+      "loss": 5.6379,
+      "mean_token_accuracy": 0.14622695744037628,
+      "num_tokens": 5253380.0,
+      "step": 1870
+    },
+    {
+      "entropy": 5.752071571350098,
+      "epoch": 0.050761421319796954,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0004121581758667898,
+      "loss": 5.5888,
+      "mean_token_accuracy": 0.14917353242635728,
+      "num_tokens": 5266227.0,
+      "step": 1875
+    },
+    {
+      "entropy": 5.602586889266968,
+      "epoch": 0.05089678510998308,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00041122240416072533,
+      "loss": 5.5358,
+      "mean_token_accuracy": 0.14778123795986176,
+      "num_tokens": 5280582.0,
+      "step": 1880
+    },
+    {
+      "entropy": 5.582650089263916,
+      "epoch": 0.0510321489001692,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0004102828978486385,
+      "loss": 5.6277,
+      "mean_token_accuracy": 0.14625633805990218,
+      "num_tokens": 5293588.0,
+      "step": 1885
+    },
+    {
+      "entropy": 5.778658056259156,
+      "epoch": 0.05116751269035533,
+      "grad_norm": 0.921875,
+      "learning_rate": 0.0004093396826875695,
+      "loss": 5.6307,
+      "mean_token_accuracy": 0.15253614038228988,
+      "num_tokens": 5306660.0,
+      "step": 1890
+    },
+    {
+      "entropy": 5.693345212936402,
+      "epoch": 0.05130287648054146,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00040839278453623837,
+      "loss": 5.6564,
+      "mean_token_accuracy": 0.14705868065357208,
+      "num_tokens": 5321213.0,
+      "step": 1895
+    },
+    {
+      "entropy": 5.5740069389343265,
+      "epoch": 0.05143824027072758,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0004074422293543363,
+      "loss": 5.4629,
+      "mean_token_accuracy": 0.1561117112636566,
+      "num_tokens": 5333388.0,
+      "step": 1900
+    },
+    {
+      "entropy": 5.731689023971557,
+      "epoch": 0.05157360406091371,
+      "grad_norm": 0.91796875,
+      "learning_rate": 0.0004064880432018137,
+      "loss": 5.6677,
+      "mean_token_accuracy": 0.1454417496919632,
+      "num_tokens": 5347818.0,
+      "step": 1905
+    },
+    {
+      "entropy": 5.6437784194946286,
+      "epoch": 0.05170896785109983,
+      "grad_norm": 0.84375,
+      "learning_rate": 0.00040553025223816615,
+      "loss": 5.5623,
+      "mean_token_accuracy": 0.15110951960086821,
+      "num_tokens": 5361113.0,
+      "step": 1910
+    },
+    {
+      "entropy": 5.6192018508911135,
+      "epoch": 0.051844331641285955,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00040456888272171653,
+      "loss": 5.514,
+      "mean_token_accuracy": 0.15319421291351318,
+      "num_tokens": 5374588.0,
+      "step": 1915
+    },
+    {
+      "entropy": 5.594708824157715,
+      "epoch": 0.05197969543147208,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00040360396100889577,
+      "loss": 5.4534,
+      "mean_token_accuracy": 0.1579025626182556,
+      "num_tokens": 5387947.0,
+      "step": 1920
+    },
+    {
+      "entropy": 5.686883783340454,
+      "epoch": 0.052115059221658204,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0004026355135535202,
+      "loss": 5.5768,
+      "mean_token_accuracy": 0.14642110466957092,
+      "num_tokens": 5401973.0,
+      "step": 1925
+    },
+    {
+      "entropy": 5.65032434463501,
+      "epoch": 0.05225042301184433,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.000401663566906066,
+      "loss": 5.6805,
+      "mean_token_accuracy": 0.14293227344751358,
+      "num_tokens": 5414320.0,
+      "step": 1930
+    },
+    {
+      "entropy": 5.6758674621582035,
+      "epoch": 0.05238578680203046,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.00040068814771294134,
+      "loss": 5.5703,
+      "mean_token_accuracy": 0.15153372585773467,
+      "num_tokens": 5428566.0,
+      "step": 1935
+    },
+    {
+      "entropy": 5.69721827507019,
+      "epoch": 0.052521150592216584,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0003997092827157562,
+      "loss": 5.5449,
+      "mean_token_accuracy": 0.15088408589363098,
+      "num_tokens": 5442237.0,
+      "step": 1940
+    },
+    {
+      "entropy": 5.631793069839477,
+      "epoch": 0.05265651438240271,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.000398726998750589,
+      "loss": 5.555,
+      "mean_token_accuracy": 0.14634450227022172,
+      "num_tokens": 5456123.0,
+      "step": 1945
+    },
+    {
+      "entropy": 5.550398159027099,
+      "epoch": 0.05279187817258883,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00039774132274725076,
+      "loss": 5.4601,
+      "mean_token_accuracy": 0.15713558942079545,
+      "num_tokens": 5469958.0,
+      "step": 1950
+    },
+    {
+      "entropy": 5.612749671936035,
+      "epoch": 0.05292724196277496,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00039675228172854707,
+      "loss": 5.5706,
+      "mean_token_accuracy": 0.1515437036752701,
+      "num_tokens": 5483537.0,
+      "step": 1955
+    },
+    {
+      "entropy": 5.541254329681396,
+      "epoch": 0.05306260575296108,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0003957599028095371,
+      "loss": 5.52,
+      "mean_token_accuracy": 0.15656681656837462,
+      "num_tokens": 5496113.0,
+      "step": 1960
+    },
+    {
+      "entropy": 5.681048393249512,
+      "epoch": 0.053197969543147205,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00039476421319679017,
+      "loss": 5.5984,
+      "mean_token_accuracy": 0.14592843279242515,
+      "num_tokens": 5510774.0,
+      "step": 1965
+    },
+    {
+      "entropy": 5.631022214889526,
+      "epoch": 0.05333333333333334,
+      "grad_norm": 0.90234375,
+      "learning_rate": 0.00039376524018764,
+      "loss": 5.5633,
+      "mean_token_accuracy": 0.15251765251159669,
+      "num_tokens": 5524804.0,
+      "step": 1970
+    },
+    {
+      "entropy": 5.578512382507324,
+      "epoch": 0.05346869712351946,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00039276301116943616,
+      "loss": 5.5762,
+      "mean_token_accuracy": 0.14840420782566072,
+      "num_tokens": 5538080.0,
+      "step": 1975
+    },
+    {
+      "entropy": 5.642090654373169,
+      "epoch": 0.053604060913705585,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0003917575536187936,
+      "loss": 5.4796,
+      "mean_token_accuracy": 0.15499964654445647,
+      "num_tokens": 5551163.0,
+      "step": 1980
+    },
+    {
+      "entropy": 5.6105047225952145,
+      "epoch": 0.05373942470389171,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00039074889510083894,
+      "loss": 5.5321,
+      "mean_token_accuracy": 0.14916370213031768,
+      "num_tokens": 5563033.0,
+      "step": 1985
+    },
+    {
+      "entropy": 5.6050855159759525,
+      "epoch": 0.053874788494077834,
+      "grad_norm": 0.8359375,
+      "learning_rate": 0.00038973706326845495,
+      "loss": 5.5097,
+      "mean_token_accuracy": 0.14798914641141891,
+      "num_tokens": 5578003.0,
+      "step": 1990
+    },
+    {
+      "entropy": 5.528244829177856,
+      "epoch": 0.05401015228426396,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0003887220858615225,
+      "loss": 5.4304,
+      "mean_token_accuracy": 0.1600724771618843,
+      "num_tokens": 5590105.0,
+      "step": 1995
+    },
+    {
+      "entropy": 5.6911087989807125,
+      "epoch": 0.05414551607445008,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0003877039907061597,
+      "loss": 5.6708,
+      "mean_token_accuracy": 0.14893444925546645,
+      "num_tokens": 5604410.0,
+      "step": 2000
+    },
+    {
+      "entropy": 5.698815584182739,
+      "epoch": 0.05428087986463621,
+      "grad_norm": 0.8828125,
+      "learning_rate": 0.0003866828057139598,
+      "loss": 5.5792,
+      "mean_token_accuracy": 0.15375161916017532,
+      "num_tokens": 5619735.0,
+      "step": 2005
+    },
+    {
+      "entropy": 5.719092321395874,
+      "epoch": 0.05441624365482234,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00038565855888122503,
+      "loss": 5.6744,
+      "mean_token_accuracy": 0.14042194560170174,
+      "num_tokens": 5634119.0,
+      "step": 2010
+    },
+    {
+      "entropy": 5.584566402435303,
+      "epoch": 0.05455160744500846,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00038463127828819975,
+      "loss": 5.4887,
+      "mean_token_accuracy": 0.1561438649892807,
+      "num_tokens": 5648254.0,
+      "step": 2015
+    },
+    {
+      "entropy": 5.588643312454224,
+      "epoch": 0.05468697123519459,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.00038360099209830043,
+      "loss": 5.5368,
+      "mean_token_accuracy": 0.1553065925836563,
+      "num_tokens": 5662954.0,
+      "step": 2020
+    },
+    {
+      "entropy": 5.600089883804321,
+      "epoch": 0.05482233502538071,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0003825677285573433,
+      "loss": 5.4919,
+      "mean_token_accuracy": 0.15998355746269227,
+      "num_tokens": 5677230.0,
+      "step": 2025
+    },
+    {
+      "entropy": 5.567089080810547,
+      "epoch": 0.054957698815566836,
+      "grad_norm": 0.91015625,
+      "learning_rate": 0.00038153151599277027,
+      "loss": 5.5599,
+      "mean_token_accuracy": 0.1493852972984314,
+      "num_tokens": 5691909.0,
+      "step": 2030
+    },
+    {
+      "entropy": 5.523836088180542,
+      "epoch": 0.05509306260575296,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0003804923828128723,
+      "loss": 5.4634,
+      "mean_token_accuracy": 0.15521894842386247,
+      "num_tokens": 5704644.0,
+      "step": 2035
+    },
+    {
+      "entropy": 5.595105218887329,
+      "epoch": 0.055228426395939084,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.0003794503575060104,
+      "loss": 5.4873,
+      "mean_token_accuracy": 0.15124136358499526,
+      "num_tokens": 5719314.0,
+      "step": 2040
+    },
+    {
+      "entropy": 5.615251255035401,
+      "epoch": 0.05536379018612521,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00037840546863983484,
+      "loss": 5.6091,
+      "mean_token_accuracy": 0.15352071672677994,
+      "num_tokens": 5733464.0,
+      "step": 2045
+    },
+    {
+      "entropy": 5.596045446395874,
+      "epoch": 0.05549915397631134,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0003773577448605015,
+      "loss": 5.4564,
+      "mean_token_accuracy": 0.1583792507648468,
+      "num_tokens": 5747934.0,
+      "step": 2050
+    },
+    {
+      "entropy": 5.621100521087646,
+      "epoch": 0.055634517766497464,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0003763072148918872,
+      "loss": 5.595,
+      "mean_token_accuracy": 0.14875475019216539,
+      "num_tokens": 5763152.0,
+      "step": 2055
+    },
+    {
+      "entropy": 5.664612340927124,
+      "epoch": 0.05576988155668359,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.0003752539075348017,
+      "loss": 5.621,
+      "mean_token_accuracy": 0.1492762856185436,
+      "num_tokens": 5777125.0,
+      "step": 2060
+    },
+    {
+      "entropy": 5.5557661056518555,
+      "epoch": 0.05590524534686971,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00037419785166619817,
+      "loss": 5.4585,
+      "mean_token_accuracy": 0.15448152869939805,
+      "num_tokens": 5790844.0,
+      "step": 2065
+    },
+    {
+      "entropy": 5.711333417892456,
+      "epoch": 0.05604060913705584,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0003731390762383818,
+      "loss": 5.6297,
+      "mean_token_accuracy": 0.15116460025310516,
+      "num_tokens": 5804419.0,
+      "step": 2070
+    },
+    {
+      "entropy": 5.697661542892456,
+      "epoch": 0.05617597292724196,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.0003720776102782158,
+      "loss": 5.5238,
+      "mean_token_accuracy": 0.15059797763824462,
+      "num_tokens": 5818710.0,
+      "step": 2075
+    },
+    {
+      "entropy": 5.481573247909546,
+      "epoch": 0.056311336717428086,
+      "grad_norm": 0.90234375,
+      "learning_rate": 0.00037101348288632555,
+      "loss": 5.4792,
+      "mean_token_accuracy": 0.15036241486668586,
+      "num_tokens": 5832986.0,
+      "step": 2080
+    },
+    {
+      "entropy": 5.6798604965209964,
+      "epoch": 0.05644670050761421,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.0003699467232363012,
+      "loss": 5.6005,
+      "mean_token_accuracy": 0.14628931134939194,
+      "num_tokens": 5847273.0,
+      "step": 2085
+    },
+    {
+      "entropy": 5.626612424850464,
+      "epoch": 0.05658206429780034,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.0003688773605738973,
+      "loss": 5.4952,
+      "mean_token_accuracy": 0.15603184401988984,
+      "num_tokens": 5860585.0,
+      "step": 2090
+    },
+    {
+      "entropy": 5.655035018920898,
+      "epoch": 0.056717428087986466,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00036780542421623134,
+      "loss": 5.6352,
+      "mean_token_accuracy": 0.15595826059579848,
+      "num_tokens": 5874305.0,
+      "step": 2095
+    },
+    {
+      "entropy": 5.631872653961182,
+      "epoch": 0.05685279187817259,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.0003667309435509802,
+      "loss": 5.5025,
+      "mean_token_accuracy": 0.15525192618370057,
+      "num_tokens": 5888975.0,
+      "step": 2100
+    },
+    {
+      "entropy": 5.599838733673096,
+      "epoch": 0.056988155668358714,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0003656539480355741,
+      "loss": 5.4398,
+      "mean_token_accuracy": 0.15944063067436218,
+      "num_tokens": 5902132.0,
+      "step": 2105
+    },
+    {
+      "entropy": 5.597221994400025,
+      "epoch": 0.05712351945854484,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.0003645744671963891,
+      "loss": 5.4945,
+      "mean_token_accuracy": 0.15443069338798524,
+      "num_tokens": 5916030.0,
+      "step": 2110
+    },
+    {
+      "entropy": 5.598980569839478,
+      "epoch": 0.05725888324873096,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003634925306279376,
+      "loss": 5.4403,
+      "mean_token_accuracy": 0.15974231958389282,
+      "num_tokens": 5928290.0,
+      "step": 2115
+    },
+    {
+      "entropy": 5.579548406600952,
+      "epoch": 0.05739424703891709,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0003624081679920574,
+      "loss": 5.5052,
+      "mean_token_accuracy": 0.1567055270075798,
+      "num_tokens": 5940463.0,
+      "step": 2120
+    },
+    {
+      "entropy": 5.538625049591064,
+      "epoch": 0.05752961082910321,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0003613214090170977,
+      "loss": 5.4353,
+      "mean_token_accuracy": 0.1593435138463974,
+      "num_tokens": 5954114.0,
+      "step": 2125
+    },
+    {
+      "entropy": 5.529782009124756,
+      "epoch": 0.05766497461928934,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0003602322834971048,
+      "loss": 5.4826,
+      "mean_token_accuracy": 0.16051046177744865,
+      "num_tokens": 5968004.0,
+      "step": 2130
+    },
+    {
+      "entropy": 5.56360445022583,
+      "epoch": 0.05780033840947547,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.0003591408212910051,
+      "loss": 5.4107,
+      "mean_token_accuracy": 0.15503834784030915,
+      "num_tokens": 5981749.0,
+      "step": 2135
+    },
+    {
+      "entropy": 5.674638128280639,
+      "epoch": 0.05793570219966159,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0003580470523217863,
+      "loss": 5.6603,
+      "mean_token_accuracy": 0.1459052950143814,
+      "num_tokens": 5995872.0,
+      "step": 2140
+    },
+    {
+      "entropy": 5.516640329360962,
+      "epoch": 0.058071065989847716,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0003569510065756771,
+      "loss": 5.4613,
+      "mean_token_accuracy": 0.159725384414196,
+      "num_tokens": 6010511.0,
+      "step": 2145
+    },
+    {
+      "entropy": 5.602073812484742,
+      "epoch": 0.05820642978003384,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.0003558527141013254,
+      "loss": 5.5,
+      "mean_token_accuracy": 0.15657427310943603,
+      "num_tokens": 6024122.0,
+      "step": 2150
+    },
+    {
+      "entropy": 5.603561973571777,
+      "epoch": 0.058341793570219964,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.0003547522050089742,
+      "loss": 5.4806,
+      "mean_token_accuracy": 0.16240517124533654,
+      "num_tokens": 6038779.0,
+      "step": 2155
+    },
+    {
+      "entropy": 5.605594062805176,
+      "epoch": 0.05847715736040609,
+      "grad_norm": 0.9140625,
+      "learning_rate": 0.00035364950946963606,
+      "loss": 5.4582,
+      "mean_token_accuracy": 0.155762480199337,
+      "num_tokens": 6052030.0,
+      "step": 2160
+    },
+    {
+      "entropy": 5.573745393753052,
+      "epoch": 0.05861252115059222,
+      "grad_norm": 0.921875,
+      "learning_rate": 0.0003525446577142663,
+      "loss": 5.5071,
+      "mean_token_accuracy": 0.1532558962702751,
+      "num_tokens": 6066580.0,
+      "step": 2165
+    },
+    {
+      "entropy": 5.605597829818725,
+      "epoch": 0.058747884940778344,
+      "grad_norm": 0.921875,
+      "learning_rate": 0.00035143768003293395,
+      "loss": 5.5391,
+      "mean_token_accuracy": 0.15521732419729234,
+      "num_tokens": 6081902.0,
+      "step": 2170
+    },
+    {
+      "entropy": 5.6139661312103275,
+      "epoch": 0.05888324873096447,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.0003503286067739913,
+      "loss": 5.5386,
+      "mean_token_accuracy": 0.1476532280445099,
+      "num_tokens": 6096694.0,
+      "step": 2175
+    },
+    {
+      "entropy": 5.541563320159912,
+      "epoch": 0.05901861252115059,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.00034921746834324193,
+      "loss": 5.4467,
+      "mean_token_accuracy": 0.1517432898283005,
+      "num_tokens": 6112894.0,
+      "step": 2180
+    },
+    {
+      "entropy": 5.5798078060150145,
+      "epoch": 0.05915397631133672,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0003481042952031072,
+      "loss": 5.499,
+      "mean_token_accuracy": 0.16076818704605103,
+      "num_tokens": 6125753.0,
+      "step": 2185
+    },
+    {
+      "entropy": 5.762272024154663,
+      "epoch": 0.05928934010152284,
+      "grad_norm": 0.91796875,
+      "learning_rate": 0.0003469891178717911,
+      "loss": 5.6864,
+      "mean_token_accuracy": 0.14467110186815263,
+      "num_tokens": 6139487.0,
+      "step": 2190
+    },
+    {
+      "entropy": 5.568824195861817,
+      "epoch": 0.059424703891708966,
+      "grad_norm": 0.84765625,
+      "learning_rate": 0.0003458719669224436,
+      "loss": 5.4139,
+      "mean_token_accuracy": 0.15835953801870345,
+      "num_tokens": 6155195.0,
+      "step": 2195
+    },
+    {
+      "entropy": 5.551574659347534,
+      "epoch": 0.05956006768189509,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0003447528729823221,
+      "loss": 5.5192,
+      "mean_token_accuracy": 0.15438710078597068,
+      "num_tokens": 6168836.0,
+      "step": 2200
+    },
+    {
+      "entropy": 5.640334033966065,
+      "epoch": 0.05969543147208122,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0003436318667319525,
+      "loss": 5.522,
+      "mean_token_accuracy": 0.16132964789867402,
+      "num_tokens": 6182513.0,
+      "step": 2205
+    },
+    {
+      "entropy": 5.578224229812622,
+      "epoch": 0.059830795262267346,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00034250897890428716,
+      "loss": 5.4526,
+      "mean_token_accuracy": 0.15591600388288498,
+      "num_tokens": 6196641.0,
+      "step": 2210
+    },
+    {
+      "entropy": 5.471917009353637,
+      "epoch": 0.05996615905245347,
+      "grad_norm": 0.96484375,
+      "learning_rate": 0.0003413842402838633,
+      "loss": 5.3832,
+      "mean_token_accuracy": 0.16338237076997758,
+      "num_tokens": 6210446.0,
+      "step": 2215
+    },
+    {
+      "entropy": 5.558266496658325,
+      "epoch": 0.060101522842639594,
+      "grad_norm": 1.0,
+      "learning_rate": 0.00034025768170595834,
+      "loss": 5.4801,
+      "mean_token_accuracy": 0.15260896682739258,
+      "num_tokens": 6223168.0,
+      "step": 2220
+    },
+    {
+      "entropy": 5.594317674636841,
+      "epoch": 0.06023688663282572,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.0003391293340557446,
+      "loss": 5.5317,
+      "mean_token_accuracy": 0.15804025083780288,
+      "num_tokens": 6236136.0,
+      "step": 2225
+    },
+    {
+      "entropy": 5.661670207977295,
+      "epoch": 0.06037225042301184,
+      "grad_norm": 0.92578125,
+      "learning_rate": 0.0003379992282674431,
+      "loss": 5.6567,
+      "mean_token_accuracy": 0.14315905719995498,
+      "num_tokens": 6250260.0,
+      "step": 2230
+    },
+    {
+      "entropy": 5.584939670562744,
+      "epoch": 0.06050761421319797,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0003368673953234749,
+      "loss": 5.4497,
+      "mean_token_accuracy": 0.1615369811654091,
+      "num_tokens": 6264510.0,
+      "step": 2235
+    },
+    {
+      "entropy": 5.551972198486328,
+      "epoch": 0.06064297800338409,
+      "grad_norm": 0.83203125,
+      "learning_rate": 0.00033573386625361176,
+      "loss": 5.3992,
+      "mean_token_accuracy": 0.15868130326271057,
+      "num_tokens": 6280784.0,
+      "step": 2240
+    },
+    {
+      "entropy": 5.652569103240967,
+      "epoch": 0.06077834179357022,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00033459867213412567,
+      "loss": 5.5289,
+      "mean_token_accuracy": 0.15628519803285598,
+      "num_tokens": 6294819.0,
+      "step": 2245
+    },
+    {
+      "entropy": 5.400842475891113,
+      "epoch": 0.06091370558375635,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.000333461844086937,
+      "loss": 5.3094,
+      "mean_token_accuracy": 0.16555650681257247,
+      "num_tokens": 6309654.0,
+      "step": 2250
+    },
+    {
+      "entropy": 5.477272796630859,
+      "epoch": 0.06104906937394247,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00033232341327876097,
+      "loss": 5.4673,
+      "mean_token_accuracy": 0.16471943408250808,
+      "num_tokens": 6322804.0,
+      "step": 2255
+    },
+    {
+      "entropy": 5.658356094360352,
+      "epoch": 0.061184433164128596,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0003311834109202531,
+      "loss": 5.5506,
+      "mean_token_accuracy": 0.1548818677663803,
+      "num_tokens": 6336513.0,
+      "step": 2260
+    },
+    {
+      "entropy": 5.56692852973938,
+      "epoch": 0.06131979695431472,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00033004186826515416,
+      "loss": 5.5298,
+      "mean_token_accuracy": 0.15601305216550826,
+      "num_tokens": 6349883.0,
+      "step": 2265
+    },
+    {
+      "entropy": 5.621473360061645,
+      "epoch": 0.061455160744500845,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0003288988166094324,
+      "loss": 5.5471,
+      "mean_token_accuracy": 0.15290187150239945,
+      "num_tokens": 6363163.0,
+      "step": 2270
+    },
+    {
+      "entropy": 5.6147716522216795,
+      "epoch": 0.06159052453468697,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00032775428729042656,
+      "loss": 5.4855,
+      "mean_token_accuracy": 0.15710740610957147,
+      "num_tokens": 6377689.0,
+      "step": 2275
+    },
+    {
+      "entropy": 5.543179178237915,
+      "epoch": 0.06172588832487309,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.000326608311685986,
+      "loss": 5.3916,
+      "mean_token_accuracy": 0.15376305133104323,
+      "num_tokens": 6391921.0,
+      "step": 2280
+    },
+    {
+      "entropy": 5.585430002212524,
+      "epoch": 0.061861252115059225,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0003254609212136108,
+      "loss": 5.5179,
+      "mean_token_accuracy": 0.1545792117714882,
+      "num_tokens": 6406731.0,
+      "step": 2285
+    },
+    {
+      "entropy": 5.570159673690796,
+      "epoch": 0.06199661590524535,
+      "grad_norm": 0.91796875,
+      "learning_rate": 0.00032431214732959036,
+      "loss": 5.5019,
+      "mean_token_accuracy": 0.15775589048862457,
+      "num_tokens": 6420403.0,
+      "step": 2290
+    },
+    {
+      "entropy": 5.589111185073852,
+      "epoch": 0.06213197969543147,
+      "grad_norm": 0.92578125,
+      "learning_rate": 0.000323162021528141,
+      "loss": 5.496,
+      "mean_token_accuracy": 0.153490050137043,
+      "num_tokens": 6434852.0,
+      "step": 2295
+    },
+    {
+      "entropy": 5.513012456893921,
+      "epoch": 0.0622673434856176,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00032201057534054264,
+      "loss": 5.4013,
+      "mean_token_accuracy": 0.16139222010970117,
+      "num_tokens": 6448481.0,
+      "step": 2300
+    },
+    {
+      "entropy": 5.525697946548462,
+      "epoch": 0.06240270727580372,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00032085784033427414,
+      "loss": 5.4119,
+      "mean_token_accuracy": 0.15614687353372575,
+      "num_tokens": 6462149.0,
+      "step": 2305
+    },
+    {
+      "entropy": 5.631346750259399,
+      "epoch": 0.06253807106598985,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0003197038481121478,
+      "loss": 5.57,
+      "mean_token_accuracy": 0.15901135355234147,
+      "num_tokens": 6475847.0,
+      "step": 2310
+    },
+    {
+      "entropy": 5.519312572479248,
+      "epoch": 0.06267343485617598,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0003185486303114436,
+      "loss": 5.4104,
+      "mean_token_accuracy": 0.15896950513124466,
+      "num_tokens": 6488881.0,
+      "step": 2315
+    },
+    {
+      "entropy": 5.459926891326904,
+      "epoch": 0.0628087986463621,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0003173922186030409,
+      "loss": 5.3378,
+      "mean_token_accuracy": 0.16899482160806656,
+      "num_tokens": 6502019.0,
+      "step": 2320
+    },
+    {
+      "entropy": 5.500802850723266,
+      "epoch": 0.06294416243654823,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.000316234644690551,
+      "loss": 5.4286,
+      "mean_token_accuracy": 0.1598815768957138,
+      "num_tokens": 6516025.0,
+      "step": 2325
+    },
+    {
+      "entropy": 5.611294984817505,
+      "epoch": 0.06307952622673435,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.0003150759403094473,
+      "loss": 5.5137,
+      "mean_token_accuracy": 0.15525443553924562,
+      "num_tokens": 6532164.0,
+      "step": 2330
+    },
+    {
+      "entropy": 5.551161003112793,
+      "epoch": 0.06321489001692047,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00031391613722619587,
+      "loss": 5.3971,
+      "mean_token_accuracy": 0.15851665437221527,
+      "num_tokens": 6545729.0,
+      "step": 2335
+    },
+    {
+      "entropy": 5.541462421417236,
+      "epoch": 0.0633502538071066,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.000312755267237384,
+      "loss": 5.3707,
+      "mean_token_accuracy": 0.15949714332818984,
+      "num_tokens": 6559398.0,
+      "step": 2340
+    },
+    {
+      "entropy": 5.5119524002075195,
+      "epoch": 0.06348561759729272,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.0003115933621688488,
+      "loss": 5.481,
+      "mean_token_accuracy": 0.15978508144617082,
+      "num_tokens": 6575516.0,
+      "step": 2345
+    },
+    {
+      "entropy": 5.600857067108154,
+      "epoch": 0.06362098138747885,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00031043045387480487,
+      "loss": 5.4257,
+      "mean_token_accuracy": 0.15539521276950835,
+      "num_tokens": 6587894.0,
+      "step": 2350
+    },
+    {
+      "entropy": 5.5432634353637695,
+      "epoch": 0.06375634517766497,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.0003092665742369703,
+      "loss": 5.4473,
+      "mean_token_accuracy": 0.16352478563785552,
+      "num_tokens": 6603355.0,
+      "step": 2355
+    },
+    {
+      "entropy": 5.5917826175689695,
+      "epoch": 0.0638917089678511,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00030810175516369343,
+      "loss": 5.5569,
+      "mean_token_accuracy": 0.152334725856781,
+      "num_tokens": 6618677.0,
+      "step": 2360
+    },
+    {
+      "entropy": 5.666406059265137,
+      "epoch": 0.06402707275803722,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0003069360285890775,
+      "loss": 5.5159,
+      "mean_token_accuracy": 0.15209025740623475,
+      "num_tokens": 6635415.0,
+      "step": 2365
+    },
+    {
+      "entropy": 5.668917846679688,
+      "epoch": 0.06416243654822334,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00030576942647210547,
+      "loss": 5.6671,
+      "mean_token_accuracy": 0.15406512916088105,
+      "num_tokens": 6650605.0,
+      "step": 2370
+    },
+    {
+      "entropy": 5.528806543350219,
+      "epoch": 0.06429780033840947,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.00030460198079576355,
+      "loss": 5.3962,
+      "mean_token_accuracy": 0.16389574855566025,
+      "num_tokens": 6664430.0,
+      "step": 2375
+    },
+    {
+      "entropy": 5.590758180618286,
+      "epoch": 0.06443316412859561,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0003034337235661648,
+      "loss": 5.3776,
+      "mean_token_accuracy": 0.16676643639802932,
+      "num_tokens": 6677763.0,
+      "step": 2380
+    },
+    {
+      "entropy": 5.4963232517242435,
+      "epoch": 0.06456852791878173,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.0003022646868116714,
+      "loss": 5.3369,
+      "mean_token_accuracy": 0.1656413570046425,
+      "num_tokens": 6691694.0,
+      "step": 2385
+    },
+    {
+      "entropy": 5.495297574996949,
+      "epoch": 0.06470389170896786,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0003010949025820163,
+      "loss": 5.4282,
+      "mean_token_accuracy": 0.16844454854726792,
+      "num_tokens": 6707451.0,
+      "step": 2390
+    },
+    {
+      "entropy": 5.507102966308594,
+      "epoch": 0.06483925549915398,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0002999244029474252,
+      "loss": 5.3812,
+      "mean_token_accuracy": 0.16256919205188752,
+      "num_tokens": 6721556.0,
+      "step": 2395
+    },
+    {
+      "entropy": 5.582904243469239,
+      "epoch": 0.0649746192893401,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00029875321999773684,
+      "loss": 5.4365,
+      "mean_token_accuracy": 0.1632665753364563,
+      "num_tokens": 6734588.0,
+      "step": 2400
+    },
+    {
+      "entropy": 5.4878991603851315,
+      "epoch": 0.06510998307952623,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.00029758138584152333,
+      "loss": 5.4356,
+      "mean_token_accuracy": 0.16168006360530854,
+      "num_tokens": 6747712.0,
+      "step": 2405
+    },
+    {
+      "entropy": 5.454759693145752,
+      "epoch": 0.06524534686971235,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0002964089326052102,
+      "loss": 5.3916,
+      "mean_token_accuracy": 0.16523440480232238,
+      "num_tokens": 6761410.0,
+      "step": 2410
+    },
+    {
+      "entropy": 5.61132550239563,
+      "epoch": 0.06538071065989848,
+      "grad_norm": 0.91796875,
+      "learning_rate": 0.0002952358924321949,
+      "loss": 5.518,
+      "mean_token_accuracy": 0.1567309945821762,
+      "num_tokens": 6776003.0,
+      "step": 2415
+    },
+    {
+      "entropy": 5.593502044677734,
+      "epoch": 0.0655160744500846,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.00029406229748196657,
+      "loss": 5.4497,
+      "mean_token_accuracy": 0.16475784331560134,
+      "num_tokens": 6790709.0,
+      "step": 2420
+    },
+    {
+      "entropy": 5.519558477401733,
+      "epoch": 0.06565143824027073,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.0002928881799292235,
+      "loss": 5.4502,
+      "mean_token_accuracy": 0.15292711034417153,
+      "num_tokens": 6804682.0,
+      "step": 2425
+    },
+    {
+      "entropy": 5.496097755432129,
+      "epoch": 0.06578680203045685,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00029171357196299154,
+      "loss": 5.3023,
+      "mean_token_accuracy": 0.1671382039785385,
+      "num_tokens": 6816787.0,
+      "step": 2430
+    },
+    {
+      "entropy": 5.5248020648956295,
+      "epoch": 0.06592216582064298,
+      "grad_norm": 0.8984375,
+      "learning_rate": 0.0002905385057857414,
+      "loss": 5.3747,
+      "mean_token_accuracy": 0.16119658052921296,
+      "num_tokens": 6832773.0,
+      "step": 2435
+    },
+    {
+      "entropy": 5.525621557235718,
+      "epoch": 0.0660575296108291,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.0002893630136125058,
+      "loss": 5.4688,
+      "mean_token_accuracy": 0.15827736407518386,
+      "num_tokens": 6847640.0,
+      "step": 2440
+    },
+    {
+      "entropy": 5.430527544021606,
+      "epoch": 0.06619289340101522,
+      "grad_norm": 0.97265625,
+      "learning_rate": 0.0002881871276699967,
+      "loss": 5.3552,
+      "mean_token_accuracy": 0.16867494881153106,
+      "num_tokens": 6860824.0,
+      "step": 2445
+    },
+    {
+      "entropy": 5.561198711395264,
+      "epoch": 0.06632825719120135,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.00028701088019572114,
+      "loss": 5.4154,
+      "mean_token_accuracy": 0.16522417813539506,
+      "num_tokens": 6875275.0,
+      "step": 2450
+    },
+    {
+      "entropy": 5.504469585418701,
+      "epoch": 0.06646362098138747,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.0002858343034370977,
+      "loss": 5.4315,
+      "mean_token_accuracy": 0.16247233897447586,
+      "num_tokens": 6889204.0,
+      "step": 2455
+    },
+    {
+      "entropy": 5.62698335647583,
+      "epoch": 0.06659898477157361,
+      "grad_norm": 0.87109375,
+      "learning_rate": 0.00028465742965057267,
+      "loss": 5.4637,
+      "mean_token_accuracy": 0.15535432547330857,
+      "num_tokens": 6904921.0,
+      "step": 2460
+    },
+    {
+      "entropy": 5.487014389038086,
+      "epoch": 0.06673434856175973,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.00028348029110073533,
+      "loss": 5.4044,
+      "mean_token_accuracy": 0.16557617038488387,
+      "num_tokens": 6918655.0,
+      "step": 2465
+    },
+    {
+      "entropy": 5.577527332305908,
+      "epoch": 0.06686971235194586,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00028230292005943365,
+      "loss": 5.4711,
+      "mean_token_accuracy": 0.16226180791854858,
+      "num_tokens": 6932828.0,
+      "step": 2470
+    },
+    {
+      "entropy": 5.516256475448609,
+      "epoch": 0.06700507614213198,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.00028112534880488945,
+      "loss": 5.3411,
+      "mean_token_accuracy": 0.16999999433755875,
+      "num_tokens": 6945109.0,
+      "step": 2475
+    },
+    {
+      "entropy": 5.428887033462525,
+      "epoch": 0.06714043993231811,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0002799476096208137,
+      "loss": 5.3381,
+      "mean_token_accuracy": 0.16663481295108795,
+      "num_tokens": 6958474.0,
+      "step": 2480
+    },
+    {
+      "entropy": 5.509528636932373,
+      "epoch": 0.06727580372250423,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00027876973479552087,
+      "loss": 5.4112,
+      "mean_token_accuracy": 0.16581467241048814,
+      "num_tokens": 6971847.0,
+      "step": 2485
+    },
+    {
+      "entropy": 5.571677732467651,
+      "epoch": 0.06741116751269036,
+      "grad_norm": 0.91796875,
+      "learning_rate": 0.00027759175662104424,
+      "loss": 5.3789,
+      "mean_token_accuracy": 0.16470448225736617,
+      "num_tokens": 6985637.0,
+      "step": 2490
+    },
+    {
+      "entropy": 5.432705068588257,
+      "epoch": 0.06754653130287648,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0002764137073922508,
+      "loss": 5.3402,
+      "mean_token_accuracy": 0.17244728654623032,
+      "num_tokens": 6999074.0,
+      "step": 2495
+    },
+    {
+      "entropy": 5.488432264328003,
+      "epoch": 0.0676818950930626,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00027523561940595505,
+      "loss": 5.2963,
+      "mean_token_accuracy": 0.17337176203727722,
+      "num_tokens": 7012184.0,
+      "step": 2500
+    }
+  ],
+  "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": 1.2391276529664e+16,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}