diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,7021 @@
+{
+  "best_global_step": 3400,
+  "best_metric": 0.06878264993429184,
+  "best_model_checkpoint": "local/checkpoints/belief-sft-qwen25-3b-v1/checkpoint-3400",
+  "epoch": 1.1643905989126246,
+  "eval_steps": 200,
+  "global_step": 3400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 0.9062692061066627,
+      "epoch": 0.0017124020720065071,
+      "grad_norm": 23.75,
+      "learning_rate": 1.8264840182648401e-07,
+      "loss": 0.3624,
+      "mean_token_accuracy": 0.9272528678178787,
+      "num_tokens": 95074.0,
+      "step": 5
+    },
+    {
+      "entropy": 0.9045031666755676,
+      "epoch": 0.0034248041440130142,
+      "grad_norm": 18.625,
+      "learning_rate": 4.1095890410958903e-07,
+      "loss": 0.4053,
+      "mean_token_accuracy": 0.9153304010629654,
+      "num_tokens": 191118.0,
+      "step": 10
+    },
+    {
+      "entropy": 0.9171398520469666,
+      "epoch": 0.005137206216019522,
+      "grad_norm": 18.75,
+      "learning_rate": 6.39269406392694e-07,
+      "loss": 0.3463,
+      "mean_token_accuracy": 0.9308812618255615,
+      "num_tokens": 285855.0,
+      "step": 15
+    },
+    {
+      "entropy": 0.9197553902864456,
+      "epoch": 0.0068496082880260285,
+      "grad_norm": 20.5,
+      "learning_rate": 8.675799086757991e-07,
+      "loss": 0.3903,
+      "mean_token_accuracy": 0.9208075940608978,
+      "num_tokens": 380927.0,
+      "step": 20
+    },
+    {
+      "entropy": 0.9234911277890205,
+      "epoch": 0.008562010360032535,
+      "grad_norm": 19.75,
+      "learning_rate": 1.095890410958904e-06,
+      "loss": 0.3015,
+      "mean_token_accuracy": 0.9363403037190438,
+      "num_tokens": 476032.0,
+      "step": 25
+    },
+    {
+      "entropy": 0.9328053191304206,
+      "epoch": 0.010274412432039044,
+      "grad_norm": 17.5,
+      "learning_rate": 1.3242009132420092e-06,
+      "loss": 0.3091,
+      "mean_token_accuracy": 0.9269585222005844,
+      "num_tokens": 571950.0,
+      "step": 30
+    },
+    {
+      "entropy": 0.9477812826633454,
+      "epoch": 0.01198681450404555,
+      "grad_norm": 14.4375,
+      "learning_rate": 1.5525114155251142e-06,
+      "loss": 0.2599,
+      "mean_token_accuracy": 0.9340660348534584,
+      "num_tokens": 666916.0,
+      "step": 35
+    },
+    {
+      "entropy": 0.9567943319678307,
+      "epoch": 0.013699216576052057,
+      "grad_norm": 5.125,
+      "learning_rate": 1.7808219178082193e-06,
+      "loss": 0.2449,
+      "mean_token_accuracy": 0.9327369257807732,
+      "num_tokens": 762379.0,
+      "step": 40
+    },
+    {
+      "entropy": 0.9490906789898872,
+      "epoch": 0.015411618648058564,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.0091324200913244e-06,
+      "loss": 0.1546,
+      "mean_token_accuracy": 0.9585488125681877,
+      "num_tokens": 858381.0,
+      "step": 45
+    },
+    {
+      "entropy": 0.9891362741589547,
+      "epoch": 0.01712402072006507,
+      "grad_norm": 3.875,
+      "learning_rate": 2.2374429223744292e-06,
+      "loss": 0.2014,
+      "mean_token_accuracy": 0.946811343729496,
+      "num_tokens": 954332.0,
+      "step": 50
+    },
+    {
+      "entropy": 1.0156420469284058,
+      "epoch": 0.018836422792071577,
+      "grad_norm": 3.375,
+      "learning_rate": 2.4657534246575345e-06,
+      "loss": 0.1856,
+      "mean_token_accuracy": 0.9508478149771691,
+      "num_tokens": 1050610.0,
+      "step": 55
+    },
+    {
+      "entropy": 1.0485677316784858,
+      "epoch": 0.020548824864078087,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.6940639269406394e-06,
+      "loss": 0.185,
+      "mean_token_accuracy": 0.9523112565279007,
+      "num_tokens": 1146057.0,
+      "step": 60
+    },
+    {
+      "entropy": 1.0706076920032501,
+      "epoch": 0.022261226936084594,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.9223744292237442e-06,
+      "loss": 0.1839,
+      "mean_token_accuracy": 0.9491461858153343,
+      "num_tokens": 1241736.0,
+      "step": 65
+    },
+    {
+      "entropy": 1.105829471349716,
+      "epoch": 0.0239736290080911,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.1506849315068495e-06,
+      "loss": 0.2111,
+      "mean_token_accuracy": 0.940838199853897,
+      "num_tokens": 1336341.0,
+      "step": 70
+    },
+    {
+      "entropy": 1.0772303640842438,
+      "epoch": 0.025686031080097607,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.3789954337899544e-06,
+      "loss": 0.1148,
+      "mean_token_accuracy": 0.9684880748391151,
+      "num_tokens": 1433244.0,
+      "step": 75
+    },
+    {
+      "entropy": 1.1036281421780587,
+      "epoch": 0.027398433152104114,
+      "grad_norm": 2.109375,
+      "learning_rate": 3.6073059360730597e-06,
+      "loss": 0.1933,
+      "mean_token_accuracy": 0.9462000370025635,
+      "num_tokens": 1529034.0,
+      "step": 80
+    },
+    {
+      "entropy": 1.1001263737678528,
+      "epoch": 0.02911083522411062,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.8356164383561645e-06,
+      "loss": 0.1368,
+      "mean_token_accuracy": 0.9612271264195442,
+      "num_tokens": 1625159.0,
+      "step": 85
+    },
+    {
+      "entropy": 1.1086553126573562,
+      "epoch": 0.030823237296117127,
+      "grad_norm": 2.0,
+      "learning_rate": 4.063926940639269e-06,
+      "loss": 0.1608,
+      "mean_token_accuracy": 0.9532164826989173,
+      "num_tokens": 1720210.0,
+      "step": 90
+    },
+    {
+      "entropy": 1.1092362314462663,
+      "epoch": 0.032535639368123634,
+      "grad_norm": 1.859375,
+      "learning_rate": 4.292237442922374e-06,
+      "loss": 0.1366,
+      "mean_token_accuracy": 0.9635654240846634,
+      "num_tokens": 1816227.0,
+      "step": 95
+    },
+    {
+      "entropy": 1.099499014019966,
+      "epoch": 0.03424804144013014,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.52054794520548e-06,
+      "loss": 0.1566,
+      "mean_token_accuracy": 0.9522104755043983,
+      "num_tokens": 1912001.0,
+      "step": 100
+    },
+    {
+      "entropy": 1.091651801764965,
+      "epoch": 0.03596044351213665,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.748858447488585e-06,
+      "loss": 0.1476,
+      "mean_token_accuracy": 0.9571668624877929,
+      "num_tokens": 2009919.0,
+      "step": 105
+    },
+    {
+      "entropy": 1.0983191147446631,
+      "epoch": 0.037672845584143154,
+      "grad_norm": 1.9765625,
+      "learning_rate": 4.97716894977169e-06,
+      "loss": 0.1467,
+      "mean_token_accuracy": 0.9581803992390633,
+      "num_tokens": 2105818.0,
+      "step": 110
+    },
+    {
+      "entropy": 1.0989631101489068,
+      "epoch": 0.03938524765614966,
+      "grad_norm": 2.28125,
+      "learning_rate": 5.2054794520547945e-06,
+      "loss": 0.1498,
+      "mean_token_accuracy": 0.956357330083847,
+      "num_tokens": 2200642.0,
+      "step": 115
+    },
+    {
+      "entropy": 1.0744837760925292,
+      "epoch": 0.041097649728156174,
+      "grad_norm": 2.046875,
+      "learning_rate": 5.4337899543379e-06,
+      "loss": 0.1278,
+      "mean_token_accuracy": 0.9602996557950974,
+      "num_tokens": 2297903.0,
+      "step": 120
+    },
+    {
+      "entropy": 1.0978670805692672,
+      "epoch": 0.04281005180016268,
+      "grad_norm": 2.28125,
+      "learning_rate": 5.662100456621005e-06,
+      "loss": 0.1491,
+      "mean_token_accuracy": 0.9575389578938485,
+      "num_tokens": 2394082.0,
+      "step": 125
+    },
+    {
+      "entropy": 1.0745210871100426,
+      "epoch": 0.04452245387216919,
+      "grad_norm": 1.7109375,
+      "learning_rate": 5.89041095890411e-06,
+      "loss": 0.1064,
+      "mean_token_accuracy": 0.9694158479571342,
+      "num_tokens": 2490541.0,
+      "step": 130
+    },
+    {
+      "entropy": 1.087765595316887,
+      "epoch": 0.046234855944175694,
+      "grad_norm": 1.7890625,
+      "learning_rate": 6.118721461187215e-06,
+      "loss": 0.1288,
+      "mean_token_accuracy": 0.9622066810727119,
+      "num_tokens": 2586861.0,
+      "step": 135
+    },
+    {
+      "entropy": 1.0718995586037636,
+      "epoch": 0.0479472580161822,
+      "grad_norm": 1.9765625,
+      "learning_rate": 6.3470319634703205e-06,
+      "loss": 0.1253,
+      "mean_token_accuracy": 0.962417995929718,
+      "num_tokens": 2682340.0,
+      "step": 140
+    },
+    {
+      "entropy": 1.0599022552371025,
+      "epoch": 0.04965966008818871,
+      "grad_norm": 2.28125,
+      "learning_rate": 6.5753424657534245e-06,
+      "loss": 0.125,
+      "mean_token_accuracy": 0.9616312265396119,
+      "num_tokens": 2780443.0,
+      "step": 145
+    },
+    {
+      "entropy": 1.070662696659565,
+      "epoch": 0.051372062160195214,
+      "grad_norm": 2.109375,
+      "learning_rate": 6.80365296803653e-06,
+      "loss": 0.1508,
+      "mean_token_accuracy": 0.9566436842083931,
+      "num_tokens": 2877567.0,
+      "step": 150
+    },
+    {
+      "entropy": 1.0815016210079194,
+      "epoch": 0.05308446423220172,
+      "grad_norm": 2.03125,
+      "learning_rate": 7.031963470319635e-06,
+      "loss": 0.1324,
+      "mean_token_accuracy": 0.9586755976080894,
+      "num_tokens": 2972826.0,
+      "step": 155
+    },
+    {
+      "entropy": 1.0920762062072753,
+      "epoch": 0.05479686630420823,
+      "grad_norm": 2.109375,
+      "learning_rate": 7.260273972602741e-06,
+      "loss": 0.1616,
+      "mean_token_accuracy": 0.9534945413470268,
+      "num_tokens": 3070233.0,
+      "step": 160
+    },
+    {
+      "entropy": 1.0888029769062997,
+      "epoch": 0.056509268376214734,
+      "grad_norm": 2.359375,
+      "learning_rate": 7.488584474885845e-06,
+      "loss": 0.1146,
+      "mean_token_accuracy": 0.9652556926012039,
+      "num_tokens": 3165086.0,
+      "step": 165
+    },
+    {
+      "entropy": 1.077832542359829,
+      "epoch": 0.05822167044822124,
+      "grad_norm": 1.4765625,
+      "learning_rate": 7.71689497716895e-06,
+      "loss": 0.1051,
+      "mean_token_accuracy": 0.9666829258203506,
+      "num_tokens": 3260237.0,
+      "step": 170
+    },
+    {
+      "entropy": 1.0615464583039285,
+      "epoch": 0.05993407252022775,
+      "grad_norm": 1.8046875,
+      "learning_rate": 7.945205479452055e-06,
+      "loss": 0.1493,
+      "mean_token_accuracy": 0.9574898958206177,
+      "num_tokens": 3355865.0,
+      "step": 175
+    },
+    {
+      "entropy": 1.0586259469389916,
+      "epoch": 0.061646474592234254,
+      "grad_norm": 2.0625,
+      "learning_rate": 8.17351598173516e-06,
+      "loss": 0.1285,
+      "mean_token_accuracy": 0.9624573588371277,
+      "num_tokens": 3452124.0,
+      "step": 180
+    },
+    {
+      "entropy": 1.0579442173242568,
+      "epoch": 0.06335887666424077,
+      "grad_norm": 2.421875,
+      "learning_rate": 8.401826484018264e-06,
+      "loss": 0.1327,
+      "mean_token_accuracy": 0.9598859220743179,
+      "num_tokens": 3546736.0,
+      "step": 185
+    },
+    {
+      "entropy": 1.0489359453320504,
+      "epoch": 0.06507127873624727,
+      "grad_norm": 1.59375,
+      "learning_rate": 8.63013698630137e-06,
+      "loss": 0.1192,
+      "mean_token_accuracy": 0.9659060955047607,
+      "num_tokens": 3642877.0,
+      "step": 190
+    },
+    {
+      "entropy": 1.0695990741252899,
+      "epoch": 0.06678368080825378,
+      "grad_norm": 1.6875,
+      "learning_rate": 8.858447488584476e-06,
+      "loss": 0.1305,
+      "mean_token_accuracy": 0.9588893681764603,
+      "num_tokens": 3738361.0,
+      "step": 195
+    },
+    {
+      "entropy": 1.0633935362100602,
+      "epoch": 0.06849608288026028,
+      "grad_norm": 1.609375,
+      "learning_rate": 9.086757990867581e-06,
+      "loss": 0.1313,
+      "mean_token_accuracy": 0.9605677902698517,
+      "num_tokens": 3834262.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.06849608288026028,
+      "eval_entropy": 1.0730068477631718,
+      "eval_loss": 0.12394363433122635,
+      "eval_mean_token_accuracy": 0.9631483021142556,
+      "eval_num_tokens": 3834262.0,
+      "eval_runtime": 343.5784,
+      "eval_samples_per_second": 15.109,
+      "eval_steps_per_second": 15.109,
+      "step": 200
+    },
+    {
+      "entropy": 1.0728650525212289,
+      "epoch": 0.0702084849522668,
+      "grad_norm": 1.453125,
+      "learning_rate": 9.315068493150685e-06,
+      "loss": 0.1143,
+      "mean_token_accuracy": 0.9654545396566391,
+      "num_tokens": 3929861.0,
+      "step": 205
+    },
+    {
+      "entropy": 1.0676488667726516,
+      "epoch": 0.0719208870242733,
+      "grad_norm": 2.265625,
+      "learning_rate": 9.543378995433791e-06,
+      "loss": 0.1057,
+      "mean_token_accuracy": 0.9684385940432548,
+      "num_tokens": 4025272.0,
+      "step": 210
+    },
+    {
+      "entropy": 1.045828601717949,
+      "epoch": 0.07363328909627981,
+      "grad_norm": 1.9765625,
+      "learning_rate": 9.771689497716895e-06,
+      "loss": 0.1155,
+      "mean_token_accuracy": 0.9644280582666397,
+      "num_tokens": 4122183.0,
+      "step": 215
+    },
+    {
+      "entropy": 1.0757814526557923,
+      "epoch": 0.07534569116828631,
+      "grad_norm": 1.859375,
+      "learning_rate": 1e-05,
+      "loss": 0.1135,
+      "mean_token_accuracy": 0.9650613278150558,
+      "num_tokens": 4217774.0,
+      "step": 220
+    },
+    {
+      "entropy": 1.0543020576238633,
+      "epoch": 0.07705809324029282,
+      "grad_norm": 1.8671875,
+      "learning_rate": 1.0228310502283105e-05,
+      "loss": 0.1043,
+      "mean_token_accuracy": 0.9682487115263939,
+      "num_tokens": 4314260.0,
+      "step": 225
+    },
+    {
+      "entropy": 1.0522808894515037,
+      "epoch": 0.07877049531229932,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.0456621004566212e-05,
+      "loss": 0.1145,
+      "mean_token_accuracy": 0.9673304915428161,
+      "num_tokens": 4409217.0,
+      "step": 230
+    },
+    {
+      "entropy": 1.0580141142010688,
+      "epoch": 0.08048289738430583,
+      "grad_norm": 2.03125,
+      "learning_rate": 1.0684931506849316e-05,
+      "loss": 0.1208,
+      "mean_token_accuracy": 0.9653781488537788,
+      "num_tokens": 4504189.0,
+      "step": 235
+    },
+    {
+      "entropy": 1.0517845913767814,
+      "epoch": 0.08219529945631235,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.091324200913242e-05,
+      "loss": 0.1045,
+      "mean_token_accuracy": 0.9695414170622826,
+      "num_tokens": 4602030.0,
+      "step": 240
+    },
+    {
+      "entropy": 1.0771506279706955,
+      "epoch": 0.08390770152831885,
+      "grad_norm": 1.6484375,
+      "learning_rate": 1.1141552511415526e-05,
+      "loss": 0.1211,
+      "mean_token_accuracy": 0.9629720881581306,
+      "num_tokens": 4698720.0,
+      "step": 245
+    },
+    {
+      "entropy": 1.060033243894577,
+      "epoch": 0.08562010360032536,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.1369863013698632e-05,
+      "loss": 0.118,
+      "mean_token_accuracy": 0.9626638069748878,
+      "num_tokens": 4794093.0,
+      "step": 250
+    },
+    {
+      "entropy": 1.0580409437417984,
+      "epoch": 0.08733250567233186,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.1598173515981737e-05,
+      "loss": 0.1113,
+      "mean_token_accuracy": 0.9661624982953072,
+      "num_tokens": 4890592.0,
+      "step": 255
+    },
+    {
+      "entropy": 1.0918870344758034,
+      "epoch": 0.08904490774433838,
+      "grad_norm": 1.6640625,
+      "learning_rate": 1.1826484018264841e-05,
+      "loss": 0.1218,
+      "mean_token_accuracy": 0.9616506293416023,
+      "num_tokens": 4985570.0,
+      "step": 260
+    },
+    {
+      "entropy": 1.1070211112499238,
+      "epoch": 0.09075730981634487,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.2054794520547945e-05,
+      "loss": 0.1068,
+      "mean_token_accuracy": 0.9692709237337113,
+      "num_tokens": 5081557.0,
+      "step": 265
+    },
+    {
+      "entropy": 1.1139537230134011,
+      "epoch": 0.09246971188835139,
+      "grad_norm": 1.9921875,
+      "learning_rate": 1.2283105022831053e-05,
+      "loss": 0.0959,
+      "mean_token_accuracy": 0.9693335697054863,
+      "num_tokens": 5177210.0,
+      "step": 270
+    },
+    {
+      "entropy": 1.1202767953276633,
+      "epoch": 0.09418211396035789,
+      "grad_norm": 2.0,
+      "learning_rate": 1.2511415525114157e-05,
+      "loss": 0.1145,
+      "mean_token_accuracy": 0.9640319466590881,
+      "num_tokens": 5271205.0,
+      "step": 275
+    },
+    {
+      "entropy": 1.087700729072094,
+      "epoch": 0.0958945160323644,
+      "grad_norm": 2.03125,
+      "learning_rate": 1.273972602739726e-05,
+      "loss": 0.1125,
+      "mean_token_accuracy": 0.9662732973694801,
+      "num_tokens": 5367915.0,
+      "step": 280
+    },
+    {
+      "entropy": 1.1220319345593452,
+      "epoch": 0.0976069181043709,
+      "grad_norm": 1.921875,
+      "learning_rate": 1.2968036529680365e-05,
+      "loss": 0.1334,
+      "mean_token_accuracy": 0.959275484085083,
+      "num_tokens": 5462804.0,
+      "step": 285
+    },
+    {
+      "entropy": 1.1254572600126267,
+      "epoch": 0.09931932017637742,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.3196347031963472e-05,
+      "loss": 0.1106,
+      "mean_token_accuracy": 0.9642360165715218,
+      "num_tokens": 5558010.0,
+      "step": 290
+    },
+    {
+      "entropy": 1.1263577818870545,
+      "epoch": 0.10103172224838392,
+      "grad_norm": 1.890625,
+      "learning_rate": 1.3424657534246576e-05,
+      "loss": 0.1167,
+      "mean_token_accuracy": 0.9638412594795227,
+      "num_tokens": 5655146.0,
+      "step": 295
+    },
+    {
+      "entropy": 1.1474478513002395,
+      "epoch": 0.10274412432039043,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.3652968036529682e-05,
+      "loss": 0.1071,
+      "mean_token_accuracy": 0.9681661799550056,
+      "num_tokens": 5750776.0,
+      "step": 300
+    },
+    {
+      "entropy": 1.1563830554485321,
+      "epoch": 0.10445652639239693,
+      "grad_norm": 1.796875,
+      "learning_rate": 1.3881278538812786e-05,
+      "loss": 0.1046,
+      "mean_token_accuracy": 0.9668073713779449,
+      "num_tokens": 5846538.0,
+      "step": 305
+    },
+    {
+      "entropy": 1.158979158103466,
+      "epoch": 0.10616892846440344,
+      "grad_norm": 1.8671875,
+      "learning_rate": 1.4109589041095892e-05,
+      "loss": 0.0867,
+      "mean_token_accuracy": 0.9745153188705444,
+      "num_tokens": 5943710.0,
+      "step": 310
+    },
+    {
+      "entropy": 1.1690414190292358,
+      "epoch": 0.10788133053640996,
+      "grad_norm": 2.0,
+      "learning_rate": 1.4337899543378997e-05,
+      "loss": 0.1042,
+      "mean_token_accuracy": 0.9689471706748009,
+      "num_tokens": 6038625.0,
+      "step": 315
+    },
+    {
+      "entropy": 1.1702216923236848,
+      "epoch": 0.10959373260841646,
+      "grad_norm": 2.28125,
+      "learning_rate": 1.4566210045662101e-05,
+      "loss": 0.1068,
+      "mean_token_accuracy": 0.9637582913041115,
+      "num_tokens": 6132955.0,
+      "step": 320
+    },
+    {
+      "entropy": 1.1256336629390717,
+      "epoch": 0.11130613468042297,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.4794520547945205e-05,
+      "loss": 0.0899,
+      "mean_token_accuracy": 0.9711817696690559,
+      "num_tokens": 6228776.0,
+      "step": 325
+    },
+    {
+      "entropy": 1.1827411025762558,
+      "epoch": 0.11301853675242947,
+      "grad_norm": 1.796875,
+      "learning_rate": 1.5022831050228313e-05,
+      "loss": 0.0978,
+      "mean_token_accuracy": 0.9692988291382789,
+      "num_tokens": 6323051.0,
+      "step": 330
+    },
+    {
+      "entropy": 1.1549245923757554,
+      "epoch": 0.11473093882443598,
+      "grad_norm": 1.7734375,
+      "learning_rate": 1.5251141552511417e-05,
+      "loss": 0.0923,
+      "mean_token_accuracy": 0.9706217840313911,
+      "num_tokens": 6418545.0,
+      "step": 335
+    },
+    {
+      "entropy": 1.1360580712556838,
+      "epoch": 0.11644334089644248,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.5479452054794523e-05,
+      "loss": 0.0909,
+      "mean_token_accuracy": 0.9723489090800286,
+      "num_tokens": 6514653.0,
+      "step": 340
+    },
+    {
+      "entropy": 1.1471509724855422,
+      "epoch": 0.118155742968449,
+      "grad_norm": 1.703125,
+      "learning_rate": 1.5707762557077627e-05,
+      "loss": 0.1191,
+      "mean_token_accuracy": 0.9618223562836647,
+      "num_tokens": 6610026.0,
+      "step": 345
+    },
+    {
+      "entropy": 1.1042737737298012,
+      "epoch": 0.1198681450404555,
+      "grad_norm": 2.328125,
+      "learning_rate": 1.5936073059360734e-05,
+      "loss": 0.085,
+      "mean_token_accuracy": 0.9750692829489708,
+      "num_tokens": 6705202.0,
+      "step": 350
+    },
+    {
+      "entropy": 1.1440322995185852,
+      "epoch": 0.12158054711246201,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.6164383561643838e-05,
+      "loss": 0.1037,
+      "mean_token_accuracy": 0.9680525943636894,
+      "num_tokens": 6800938.0,
+      "step": 355
+    },
+    {
+      "entropy": 1.2064876496791839,
+      "epoch": 0.12329294918446851,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.6392694063926942e-05,
+      "loss": 0.0904,
+      "mean_token_accuracy": 0.9723571017384529,
+      "num_tokens": 6896919.0,
+      "step": 360
+    },
+    {
+      "entropy": 1.1749372839927674,
+      "epoch": 0.12500535125647502,
+      "grad_norm": 2.171875,
+      "learning_rate": 1.6621004566210046e-05,
+      "loss": 0.0985,
+      "mean_token_accuracy": 0.9676112428307533,
+      "num_tokens": 6992718.0,
+      "step": 365
+    },
+    {
+      "entropy": 1.1613423347473144,
+      "epoch": 0.12671775332848154,
+      "grad_norm": 1.6328125,
+      "learning_rate": 1.6849315068493153e-05,
+      "loss": 0.106,
+      "mean_token_accuracy": 0.9668835282325745,
+      "num_tokens": 7089013.0,
+      "step": 370
+    },
+    {
+      "entropy": 1.1506674513220787,
+      "epoch": 0.12843015540048802,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.7077625570776257e-05,
+      "loss": 0.1162,
+      "mean_token_accuracy": 0.9608615890145302,
+      "num_tokens": 7183768.0,
+      "step": 375
+    },
+    {
+      "entropy": 1.136763733625412,
+      "epoch": 0.13014255747249454,
+      "grad_norm": 1.3046875,
+      "learning_rate": 1.730593607305936e-05,
+      "loss": 0.1008,
+      "mean_token_accuracy": 0.969757254421711,
+      "num_tokens": 7279011.0,
+      "step": 380
+    },
+    {
+      "entropy": 1.139556035399437,
+      "epoch": 0.13185495954450105,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.7534246575342465e-05,
+      "loss": 0.0772,
+      "mean_token_accuracy": 0.9765607297420502,
+      "num_tokens": 7375455.0,
+      "step": 385
+    },
+    {
+      "entropy": 1.1232876256108284,
+      "epoch": 0.13356736161650756,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.7762557077625573e-05,
+      "loss": 0.0986,
+      "mean_token_accuracy": 0.9692042410373688,
+      "num_tokens": 7471860.0,
+      "step": 390
+    },
+    {
+      "entropy": 1.1268639057874679,
+      "epoch": 0.13527976368851408,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.7990867579908677e-05,
+      "loss": 0.0999,
+      "mean_token_accuracy": 0.9696505635976791,
+      "num_tokens": 7568173.0,
+      "step": 395
+    },
+    {
+      "entropy": 1.1772323668003082,
+      "epoch": 0.13699216576052056,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.821917808219178e-05,
+      "loss": 0.095,
+      "mean_token_accuracy": 0.9713818535208703,
+      "num_tokens": 7664220.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.13699216576052056,
+      "eval_entropy": 1.1686793487440803,
+      "eval_loss": 0.09909827262163162,
+      "eval_mean_token_accuracy": 0.968796361069118,
+      "eval_num_tokens": 7664220.0,
+      "eval_runtime": 344.0747,
+      "eval_samples_per_second": 15.087,
+      "eval_steps_per_second": 15.087,
+      "step": 400
+    },
+    {
+      "entropy": 1.148973396420479,
+      "epoch": 0.13870456783252708,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.8447488584474885e-05,
+      "loss": 0.1047,
+      "mean_token_accuracy": 0.9654049828648568,
+      "num_tokens": 7759804.0,
+      "step": 405
+    },
+    {
+      "entropy": 1.1473380595445632,
+      "epoch": 0.1404169699045336,
+      "grad_norm": 1.890625,
+      "learning_rate": 1.8675799086757992e-05,
+      "loss": 0.1012,
+      "mean_token_accuracy": 0.9676817387342453,
+      "num_tokens": 7857072.0,
+      "step": 410
+    },
+    {
+      "entropy": 1.1635912582278252,
+      "epoch": 0.1421293719765401,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.8904109589041096e-05,
+      "loss": 0.0903,
+      "mean_token_accuracy": 0.9715405374765396,
+      "num_tokens": 7952930.0,
+      "step": 415
+    },
+    {
+      "entropy": 1.1438232004642486,
+      "epoch": 0.1438417740485466,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.91324200913242e-05,
+      "loss": 0.0865,
+      "mean_token_accuracy": 0.9731761604547501,
+      "num_tokens": 8049054.0,
+      "step": 420
+    },
+    {
+      "entropy": 1.1792391508817672,
+      "epoch": 0.1455541761205531,
+      "grad_norm": 1.328125,
+      "learning_rate": 1.9360730593607308e-05,
+      "loss": 0.1092,
+      "mean_token_accuracy": 0.9646677434444427,
+      "num_tokens": 8143158.0,
+      "step": 425
+    },
+    {
+      "entropy": 1.1537054866552352,
+      "epoch": 0.14726657819255962,
+      "grad_norm": 1.8125,
+      "learning_rate": 1.9589041095890412e-05,
+      "loss": 0.0948,
+      "mean_token_accuracy": 0.966642065346241,
+      "num_tokens": 8238274.0,
+      "step": 430
+    },
+    {
+      "entropy": 1.1381796672940254,
+      "epoch": 0.14897898026456613,
+      "grad_norm": 1.7109375,
+      "learning_rate": 1.981735159817352e-05,
+      "loss": 0.1097,
+      "mean_token_accuracy": 0.9673643127083779,
+      "num_tokens": 8334829.0,
+      "step": 435
+    },
+    {
+      "entropy": 1.171370232105255,
+      "epoch": 0.15069138233657262,
+      "grad_norm": 1.6953125,
+      "learning_rate": 1.9998587770088972e-05,
+      "loss": 0.1055,
+      "mean_token_accuracy": 0.9659725993871688,
+      "num_tokens": 8431326.0,
+      "step": 440
+    },
+    {
+      "entropy": 1.178799495100975,
+      "epoch": 0.15240378440857913,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.9991526620533824e-05,
+      "loss": 0.0931,
+      "mean_token_accuracy": 0.9694060906767845,
+      "num_tokens": 8525674.0,
+      "step": 445
+    },
+    {
+      "entropy": 1.1232061132788658,
+      "epoch": 0.15411618648058564,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.9984465470978676e-05,
+      "loss": 0.0784,
+      "mean_token_accuracy": 0.9741345658898354,
+      "num_tokens": 8622874.0,
+      "step": 450
+    },
+    {
+      "entropy": 1.098988600075245,
+      "epoch": 0.15582858855259216,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.997740432142353e-05,
+      "loss": 0.0816,
+      "mean_token_accuracy": 0.9723601341247559,
+      "num_tokens": 8719645.0,
+      "step": 455
+    },
+    {
+      "entropy": 1.1582636207342147,
+      "epoch": 0.15754099062459864,
+      "grad_norm": 1.5703125,
+      "learning_rate": 1.9970343171868384e-05,
+      "loss": 0.0876,
+      "mean_token_accuracy": 0.971201179921627,
+      "num_tokens": 8813685.0,
+      "step": 460
+    },
+    {
+      "entropy": 1.2026127398014068,
+      "epoch": 0.15925339269660516,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.9963282022313236e-05,
+      "loss": 0.0785,
+      "mean_token_accuracy": 0.9762142851948739,
+      "num_tokens": 8909361.0,
+      "step": 465
+    },
+    {
+      "entropy": 1.179990318417549,
+      "epoch": 0.16096579476861167,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.995622087275809e-05,
+      "loss": 0.0994,
+      "mean_token_accuracy": 0.9681125238537789,
+      "num_tokens": 9005271.0,
+      "step": 470
+    },
+    {
+      "entropy": 1.162953832745552,
+      "epoch": 0.16267819684061818,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.994915972320294e-05,
+      "loss": 0.0902,
+      "mean_token_accuracy": 0.9697920635342598,
+      "num_tokens": 9100052.0,
+      "step": 475
+    },
+    {
+      "entropy": 1.0724239751696587,
+      "epoch": 0.1643905989126247,
+      "grad_norm": 1.265625,
+      "learning_rate": 1.9942098573647793e-05,
+      "loss": 0.09,
+      "mean_token_accuracy": 0.9717666566371918,
+      "num_tokens": 9197437.0,
+      "step": 480
+    },
+    {
+      "entropy": 1.0595092862844466,
+      "epoch": 0.16610300098463118,
+      "grad_norm": 1.3125,
+      "learning_rate": 1.9935037424092645e-05,
+      "loss": 0.0894,
+      "mean_token_accuracy": 0.9705368980765343,
+      "num_tokens": 9292794.0,
+      "step": 485
+    },
+    {
+      "entropy": 1.1322496384382248,
+      "epoch": 0.1678154030566377,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.9927976274537497e-05,
+      "loss": 0.0755,
+      "mean_token_accuracy": 0.9746683239936829,
+      "num_tokens": 9388472.0,
+      "step": 490
+    },
+    {
+      "entropy": 1.168193370103836,
+      "epoch": 0.1695278051286442,
+      "grad_norm": 1.328125,
+      "learning_rate": 1.992091512498235e-05,
+      "loss": 0.0974,
+      "mean_token_accuracy": 0.9690256848931312,
+      "num_tokens": 9484945.0,
+      "step": 495
+    },
+    {
+      "entropy": 1.1688465356826783,
+      "epoch": 0.17124020720065072,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.99138539754272e-05,
+      "loss": 0.0946,
+      "mean_token_accuracy": 0.9699531555175781,
+      "num_tokens": 9580746.0,
+      "step": 500
+    },
+    {
+      "entropy": 1.122713667154312,
+      "epoch": 0.1729526092726572,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.9906792825872054e-05,
+      "loss": 0.0836,
+      "mean_token_accuracy": 0.9738435730338096,
+      "num_tokens": 9676251.0,
+      "step": 505
+    },
+    {
+      "entropy": 1.088064730167389,
+      "epoch": 0.17466501134466372,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9899731676316906e-05,
+      "loss": 0.0923,
+      "mean_token_accuracy": 0.9695730015635491,
+      "num_tokens": 9770972.0,
+      "step": 510
+    },
+    {
+      "entropy": 1.100546683371067,
+      "epoch": 0.17637741341667024,
+      "grad_norm": 2.109375,
+      "learning_rate": 1.9892670526761758e-05,
+      "loss": 0.089,
+      "mean_token_accuracy": 0.9697953119874001,
+      "num_tokens": 9868115.0,
+      "step": 515
+    },
+    {
+      "entropy": 1.1451151102781296,
+      "epoch": 0.17808981548867675,
+      "grad_norm": 1.6328125,
+      "learning_rate": 1.988560937720661e-05,
+      "loss": 0.0874,
+      "mean_token_accuracy": 0.9729209512472152,
+      "num_tokens": 9963643.0,
+      "step": 520
+    },
+    {
+      "entropy": 1.1592493295669555,
+      "epoch": 0.17980221756068324,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.9878548227651462e-05,
+      "loss": 0.0986,
+      "mean_token_accuracy": 0.9683105036616325,
+      "num_tokens": 10057510.0,
+      "step": 525
+    },
+    {
+      "entropy": 1.1292186185717583,
+      "epoch": 0.18151461963268975,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.9871487078096318e-05,
+      "loss": 0.0842,
+      "mean_token_accuracy": 0.9726386845111847,
+      "num_tokens": 10152827.0,
+      "step": 530
+    },
+    {
+      "entropy": 1.0931081146001815,
+      "epoch": 0.18322702170469626,
+      "grad_norm": 1.921875,
+      "learning_rate": 1.986442592854117e-05,
+      "loss": 0.112,
+      "mean_token_accuracy": 0.9615588992834091,
+      "num_tokens": 10247075.0,
+      "step": 535
+    },
+    {
+      "entropy": 1.0847404524683952,
+      "epoch": 0.18493942377670278,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.9857364778986022e-05,
+      "loss": 0.0937,
+      "mean_token_accuracy": 0.9710121795535087,
+      "num_tokens": 10342138.0,
+      "step": 540
+    },
+    {
+      "entropy": 1.1010636031627654,
+      "epoch": 0.1866518258487093,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.9850303629430874e-05,
+      "loss": 0.0801,
+      "mean_token_accuracy": 0.9743027776479721,
+      "num_tokens": 10437650.0,
+      "step": 545
+    },
+    {
+      "entropy": 1.155510675907135,
+      "epoch": 0.18836422792071578,
+      "grad_norm": 1.296875,
+      "learning_rate": 1.9843242479875727e-05,
+      "loss": 0.0896,
+      "mean_token_accuracy": 0.9708870887756348,
+      "num_tokens": 10534660.0,
+      "step": 550
+    },
+    {
+      "entropy": 1.267996037006378,
+      "epoch": 0.1900766299927223,
+      "grad_norm": 1.5234375,
+      "learning_rate": 1.983618133032058e-05,
+      "loss": 0.1095,
+      "mean_token_accuracy": 0.9658202782273293,
+      "num_tokens": 10630076.0,
+      "step": 555
+    },
+    {
+      "entropy": 1.2236505270004272,
+      "epoch": 0.1917890320647288,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.982912018076543e-05,
+      "loss": 0.0743,
+      "mean_token_accuracy": 0.976417088508606,
+      "num_tokens": 10727315.0,
+      "step": 560
+    },
+    {
+      "entropy": 1.1612312823534012,
+      "epoch": 0.19350143413673532,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9822059031210283e-05,
+      "loss": 0.0838,
+      "mean_token_accuracy": 0.9727051377296447,
+      "num_tokens": 10823469.0,
+      "step": 565
+    },
+    {
+      "entropy": 1.1151698872447013,
+      "epoch": 0.1952138362087418,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.9814997881655135e-05,
+      "loss": 0.0797,
+      "mean_token_accuracy": 0.9749635219573974,
+      "num_tokens": 10919960.0,
+      "step": 570
+    },
+    {
+      "entropy": 1.1441911906003952,
+      "epoch": 0.19692623828074832,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1.9807936732099988e-05,
+      "loss": 0.0884,
+      "mean_token_accuracy": 0.9716255396604538,
+      "num_tokens": 11015791.0,
+      "step": 575
+    },
+    {
+      "entropy": 1.1635893791913987,
+      "epoch": 0.19863864035275483,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.980087558254484e-05,
+      "loss": 0.0706,
+      "mean_token_accuracy": 0.9784829944372178,
+      "num_tokens": 11112082.0,
+      "step": 580
+    },
+    {
+      "entropy": 1.2036704629659654,
+      "epoch": 0.20035104242476134,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.9793814432989692e-05,
+      "loss": 0.1041,
+      "mean_token_accuracy": 0.9667044058442116,
+      "num_tokens": 11207132.0,
+      "step": 585
+    },
+    {
+      "entropy": 1.2154219895601273,
+      "epoch": 0.20206344449676783,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.9786753283434544e-05,
+      "loss": 0.0915,
+      "mean_token_accuracy": 0.9702539920806885,
+      "num_tokens": 11302991.0,
+      "step": 590
+    },
+    {
+      "entropy": 1.1827192723751068,
+      "epoch": 0.20377584656877434,
+      "grad_norm": 1.8125,
+      "learning_rate": 1.9779692133879396e-05,
+      "loss": 0.0802,
+      "mean_token_accuracy": 0.9724412202835083,
+      "num_tokens": 11398820.0,
+      "step": 595
+    },
+    {
+      "entropy": 1.1439438670873643,
+      "epoch": 0.20548824864078086,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.9772630984324252e-05,
+      "loss": 0.0882,
+      "mean_token_accuracy": 0.9716157883405685,
+      "num_tokens": 11494342.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.20548824864078086,
+      "eval_entropy": 1.1280909870153584,
+      "eval_loss": 0.08968911319971085,
+      "eval_mean_token_accuracy": 0.9710176917910002,
+      "eval_num_tokens": 11494342.0,
+      "eval_runtime": 344.7673,
+      "eval_samples_per_second": 15.057,
+      "eval_steps_per_second": 15.057,
+      "step": 600
+    },
+    {
+      "entropy": 1.135306817293167,
+      "epoch": 0.20720065071278737,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9765569834769104e-05,
+      "loss": 0.0969,
+      "mean_token_accuracy": 0.9689493924379349,
+      "num_tokens": 11590982.0,
+      "step": 605
+    },
+    {
+      "entropy": 1.1494647443294526,
+      "epoch": 0.20891305278479386,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.9758508685213956e-05,
+      "loss": 0.082,
+      "mean_token_accuracy": 0.9725746884942055,
+      "num_tokens": 11685670.0,
+      "step": 610
+    },
+    {
+      "entropy": 1.1234963625669478,
+      "epoch": 0.21062545485680037,
+      "grad_norm": 1.7265625,
+      "learning_rate": 1.975144753565881e-05,
+      "loss": 0.0934,
+      "mean_token_accuracy": 0.9686412587761879,
+      "num_tokens": 11780904.0,
+      "step": 615
+    },
+    {
+      "entropy": 1.1179186254739761,
+      "epoch": 0.21233785692880688,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.974438638610366e-05,
+      "loss": 0.0993,
+      "mean_token_accuracy": 0.967847254872322,
+      "num_tokens": 11877613.0,
+      "step": 620
+    },
+    {
+      "entropy": 1.1603994369506836,
+      "epoch": 0.2140502590008134,
+      "grad_norm": 1.078125,
+      "learning_rate": 1.9737325236548513e-05,
+      "loss": 0.0915,
+      "mean_token_accuracy": 0.9702446043491364,
+      "num_tokens": 11973440.0,
+      "step": 625
+    },
+    {
+      "entropy": 1.1549611002206803,
+      "epoch": 0.2157626610728199,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.9730264086993365e-05,
+      "loss": 0.0948,
+      "mean_token_accuracy": 0.9691407144069671,
+      "num_tokens": 12070847.0,
+      "step": 630
+    },
+    {
+      "entropy": 1.1574709385633468,
+      "epoch": 0.2174750631448264,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.9723202937438217e-05,
+      "loss": 0.0986,
+      "mean_token_accuracy": 0.9684703022241592,
+      "num_tokens": 12165926.0,
+      "step": 635
+    },
+    {
+      "entropy": 1.1355842426419258,
+      "epoch": 0.2191874652168329,
+      "grad_norm": 1.6953125,
+      "learning_rate": 1.971614178788307e-05,
+      "loss": 0.0702,
+      "mean_token_accuracy": 0.9763172134757042,
+      "num_tokens": 12262182.0,
+      "step": 640
+    },
+    {
+      "entropy": 1.1412763714790344,
+      "epoch": 0.22089986728883942,
+      "grad_norm": 1.25,
+      "learning_rate": 1.970908063832792e-05,
+      "loss": 0.078,
+      "mean_token_accuracy": 0.9748011395335198,
+      "num_tokens": 12357992.0,
+      "step": 645
+    },
+    {
+      "entropy": 1.1249642685055732,
+      "epoch": 0.22261226936084594,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.9702019488772774e-05,
+      "loss": 0.0885,
+      "mean_token_accuracy": 0.9710410177707672,
+      "num_tokens": 12454012.0,
+      "step": 650
+    },
+    {
+      "entropy": 1.1118563190102577,
+      "epoch": 0.22432467143285242,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.9694958339217626e-05,
+      "loss": 0.0889,
+      "mean_token_accuracy": 0.9718759000301361,
+      "num_tokens": 12549174.0,
+      "step": 655
+    },
+    {
+      "entropy": 1.1437419265508653,
+      "epoch": 0.22603707350485894,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.9687897189662478e-05,
+      "loss": 0.0872,
+      "mean_token_accuracy": 0.9693790361285209,
+      "num_tokens": 12645263.0,
+      "step": 660
+    },
+    {
+      "entropy": 1.1714959651231767,
+      "epoch": 0.22774947557686545,
+      "grad_norm": 0.84375,
+      "learning_rate": 1.968083604010733e-05,
+      "loss": 0.0758,
+      "mean_token_accuracy": 0.9737798690795898,
+      "num_tokens": 12740666.0,
+      "step": 665
+    },
+    {
+      "entropy": 1.1677046597003937,
+      "epoch": 0.22946187764887196,
+      "grad_norm": 1.2890625,
+      "learning_rate": 1.9673774890552186e-05,
+      "loss": 0.0872,
+      "mean_token_accuracy": 0.9732278332114219,
+      "num_tokens": 12836292.0,
+      "step": 670
+    },
+    {
+      "entropy": 1.138964867591858,
+      "epoch": 0.23117427972087845,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.9666713740997038e-05,
+      "loss": 0.0823,
+      "mean_token_accuracy": 0.9749679490923882,
+      "num_tokens": 12931270.0,
+      "step": 675
+    },
+    {
+      "entropy": 1.1000571802258492,
+      "epoch": 0.23288668179288496,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.965965259144189e-05,
+      "loss": 0.0823,
+      "mean_token_accuracy": 0.9703440621495247,
+      "num_tokens": 13026422.0,
+      "step": 680
+    },
+    {
+      "entropy": 1.0885263308882713,
+      "epoch": 0.23459908386489148,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.9652591441886742e-05,
+      "loss": 0.1056,
+      "mean_token_accuracy": 0.9662298679351806,
+      "num_tokens": 13121571.0,
+      "step": 685
+    },
+    {
+      "entropy": 1.1132431641221046,
+      "epoch": 0.236311485936898,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.9645530292331594e-05,
+      "loss": 0.0835,
+      "mean_token_accuracy": 0.9716737136244774,
+      "num_tokens": 13217964.0,
+      "step": 690
+    },
+    {
+      "entropy": 1.1341713339090347,
+      "epoch": 0.23802388800890448,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.9638469142776446e-05,
+      "loss": 0.0885,
+      "mean_token_accuracy": 0.9692564889788627,
+      "num_tokens": 13313525.0,
+      "step": 695
+    },
+    {
+      "entropy": 1.1248771339654922,
+      "epoch": 0.239736290080911,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.96314079932213e-05,
+      "loss": 0.0867,
+      "mean_token_accuracy": 0.9700839623808861,
+      "num_tokens": 13407346.0,
+      "step": 700
+    },
+    {
+      "entropy": 1.1088782325387,
+      "epoch": 0.2414486921529175,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.962434684366615e-05,
+      "loss": 0.1042,
+      "mean_token_accuracy": 0.9658521458506584,
+      "num_tokens": 13501784.0,
+      "step": 705
+    },
+    {
+      "entropy": 1.0953863456845283,
+      "epoch": 0.24316109422492402,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.9617285694111003e-05,
+      "loss": 0.075,
+      "mean_token_accuracy": 0.9736921682953834,
+      "num_tokens": 13597646.0,
+      "step": 710
+    },
+    {
+      "entropy": 1.0996337860822678,
+      "epoch": 0.24487349629693053,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.9610224544555855e-05,
+      "loss": 0.0775,
+      "mean_token_accuracy": 0.9732078939676285,
+      "num_tokens": 13693988.0,
+      "step": 715
+    },
+    {
+      "entropy": 1.1099782168865204,
+      "epoch": 0.24658589836893702,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.9603163395000707e-05,
+      "loss": 0.1032,
+      "mean_token_accuracy": 0.9639217257499695,
+      "num_tokens": 13791115.0,
+      "step": 720
+    },
+    {
+      "entropy": 1.1337740927934647,
+      "epoch": 0.24829830044094353,
+      "grad_norm": 1.125,
+      "learning_rate": 1.959610224544556e-05,
+      "loss": 0.0647,
+      "mean_token_accuracy": 0.9774993389844895,
+      "num_tokens": 13885912.0,
+      "step": 725
+    },
+    {
+      "entropy": 1.1596150428056717,
+      "epoch": 0.25001070251295004,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.9589041095890412e-05,
+      "loss": 0.0916,
+      "mean_token_accuracy": 0.9706126675009727,
+      "num_tokens": 13981889.0,
+      "step": 730
+    },
+    {
+      "entropy": 1.161018443107605,
+      "epoch": 0.25172310458495656,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.9581979946335264e-05,
+      "loss": 0.088,
+      "mean_token_accuracy": 0.9711202338337899,
+      "num_tokens": 14077836.0,
+      "step": 735
+    },
+    {
+      "entropy": 1.1635111689567565,
+      "epoch": 0.25343550665696307,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.957491879678012e-05,
+      "loss": 0.0825,
+      "mean_token_accuracy": 0.9722855061292648,
+      "num_tokens": 14176376.0,
+      "step": 740
+    },
+    {
+      "entropy": 1.1742104530334472,
+      "epoch": 0.2551479087289696,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.956785764722497e-05,
+      "loss": 0.0708,
+      "mean_token_accuracy": 0.9766950204968452,
+      "num_tokens": 14270408.0,
+      "step": 745
+    },
+    {
+      "entropy": 1.123642298579216,
+      "epoch": 0.25686031080097604,
+      "grad_norm": 0.99609375,
+      "learning_rate": 1.9560796497669824e-05,
+      "loss": 0.0759,
+      "mean_token_accuracy": 0.9750209525227547,
+      "num_tokens": 14365163.0,
+      "step": 750
+    },
+    {
+      "entropy": 1.0996820986270905,
+      "epoch": 0.25857271287298256,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.9553735348114676e-05,
+      "loss": 0.0865,
+      "mean_token_accuracy": 0.9717126682400703,
+      "num_tokens": 14462014.0,
+      "step": 755
+    },
+    {
+      "entropy": 1.0929965630173684,
+      "epoch": 0.26028511494498907,
+      "grad_norm": 1.25,
+      "learning_rate": 1.9546674198559528e-05,
+      "loss": 0.1059,
+      "mean_token_accuracy": 0.9640148550271987,
+      "num_tokens": 14557382.0,
+      "step": 760
+    },
+    {
+      "entropy": 1.0437636464834212,
+      "epoch": 0.2619975170169956,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.953961304900438e-05,
+      "loss": 0.0698,
+      "mean_token_accuracy": 0.9776692152023315,
+      "num_tokens": 14655228.0,
+      "step": 765
+    },
+    {
+      "entropy": 1.0500921428203582,
+      "epoch": 0.2637099190890021,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.9532551899449232e-05,
+      "loss": 0.0942,
+      "mean_token_accuracy": 0.9697719410061836,
+      "num_tokens": 14750671.0,
+      "step": 770
+    },
+    {
+      "entropy": 1.0289709255099297,
+      "epoch": 0.2654223211610086,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9525490749894085e-05,
+      "loss": 0.0734,
+      "mean_token_accuracy": 0.9765213608741761,
+      "num_tokens": 14847422.0,
+      "step": 775
+    },
+    {
+      "entropy": 1.0425951570272445,
+      "epoch": 0.2671347232330151,
+      "grad_norm": 1.296875,
+      "learning_rate": 1.9518429600338937e-05,
+      "loss": 0.0761,
+      "mean_token_accuracy": 0.9735964551568032,
+      "num_tokens": 14944284.0,
+      "step": 780
+    },
+    {
+      "entropy": 1.0833984687924385,
+      "epoch": 0.26884712530502164,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.951136845078379e-05,
+      "loss": 0.0944,
+      "mean_token_accuracy": 0.9690458863973618,
+      "num_tokens": 15040626.0,
+      "step": 785
+    },
+    {
+      "entropy": 1.116835503280163,
+      "epoch": 0.27055952737702815,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.950430730122864e-05,
+      "loss": 0.0752,
+      "mean_token_accuracy": 0.9745122998952865,
+      "num_tokens": 15136331.0,
+      "step": 790
+    },
+    {
+      "entropy": 1.1012332648038865,
+      "epoch": 0.2722719294490346,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.9497246151673493e-05,
+      "loss": 0.0756,
+      "mean_token_accuracy": 0.9743741288781166,
+      "num_tokens": 15232579.0,
+      "step": 795
+    },
+    {
+      "entropy": 1.1151490032672882,
+      "epoch": 0.2739843315210411,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.9490185002118346e-05,
+      "loss": 0.0901,
+      "mean_token_accuracy": 0.9711347505450248,
+      "num_tokens": 15328221.0,
+      "step": 800
+    },
+    {
+      "epoch": 0.2739843315210411,
+      "eval_entropy": 1.1329338754837832,
+      "eval_loss": 0.08416474610567093,
+      "eval_mean_token_accuracy": 0.9723265243619192,
+      "eval_num_tokens": 15328221.0,
+      "eval_runtime": 342.6765,
+      "eval_samples_per_second": 15.148,
+      "eval_steps_per_second": 15.148,
+      "step": 800
+    },
+    {
+      "entropy": 1.1421309560537338,
+      "epoch": 0.27569673359304764,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.9483123852563198e-05,
+      "loss": 0.1002,
+      "mean_token_accuracy": 0.9669936135411262,
+      "num_tokens": 15424262.0,
+      "step": 805
+    },
+    {
+      "entropy": 1.1184192776679993,
+      "epoch": 0.27740913566505415,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.9476062703008053e-05,
+      "loss": 0.0778,
+      "mean_token_accuracy": 0.9732958853244782,
+      "num_tokens": 15520651.0,
+      "step": 810
+    },
+    {
+      "entropy": 1.1349014431238174,
+      "epoch": 0.27912153773706067,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.9469001553452905e-05,
+      "loss": 0.0959,
+      "mean_token_accuracy": 0.9689885884523392,
+      "num_tokens": 15616857.0,
+      "step": 815
+    },
+    {
+      "entropy": 1.0879802733659745,
+      "epoch": 0.2808339398090672,
+      "grad_norm": 0.9765625,
+      "learning_rate": 1.9461940403897758e-05,
+      "loss": 0.0805,
+      "mean_token_accuracy": 0.9737834885716439,
+      "num_tokens": 15712455.0,
+      "step": 820
+    },
+    {
+      "entropy": 1.0658569231629371,
+      "epoch": 0.2825463418810737,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.945487925434261e-05,
+      "loss": 0.0865,
+      "mean_token_accuracy": 0.9749824166297912,
+      "num_tokens": 15807961.0,
+      "step": 825
+    },
+    {
+      "entropy": 1.0840911537408828,
+      "epoch": 0.2842587439530802,
+      "grad_norm": 1.1953125,
+      "learning_rate": 1.9447818104787462e-05,
+      "loss": 0.08,
+      "mean_token_accuracy": 0.973600634932518,
+      "num_tokens": 15903352.0,
+      "step": 830
+    },
+    {
+      "entropy": 1.1055504277348518,
+      "epoch": 0.28597114602508666,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.9440756955232314e-05,
+      "loss": 0.0803,
+      "mean_token_accuracy": 0.9731408044695854,
+      "num_tokens": 15999315.0,
+      "step": 835
+    },
+    {
+      "entropy": 1.133864377439022,
+      "epoch": 0.2876835480970932,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.9433695805677166e-05,
+      "loss": 0.0806,
+      "mean_token_accuracy": 0.9727028325200081,
+      "num_tokens": 16094373.0,
+      "step": 840
+    },
+    {
+      "entropy": 1.132483857870102,
+      "epoch": 0.2893959501690997,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.942663465612202e-05,
+      "loss": 0.092,
+      "mean_token_accuracy": 0.9698078826069831,
+      "num_tokens": 16190419.0,
+      "step": 845
+    },
+    {
+      "entropy": 1.1321695446968079,
+      "epoch": 0.2911083522411062,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.941957350656687e-05,
+      "loss": 0.0891,
+      "mean_token_accuracy": 0.9687074229121209,
+      "num_tokens": 16286761.0,
+      "step": 850
+    },
+    {
+      "entropy": 1.1051161468029023,
+      "epoch": 0.2928207543131127,
+      "grad_norm": 1.8984375,
+      "learning_rate": 1.9412512357011723e-05,
+      "loss": 0.0802,
+      "mean_token_accuracy": 0.9729267492890358,
+      "num_tokens": 16384242.0,
+      "step": 855
+    },
+    {
+      "entropy": 1.0977824166417123,
+      "epoch": 0.29453315638511923,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.9405451207456575e-05,
+      "loss": 0.0729,
+      "mean_token_accuracy": 0.9771168872714042,
+      "num_tokens": 16479877.0,
+      "step": 860
+    },
+    {
+      "entropy": 1.1167832091450691,
+      "epoch": 0.29624555845712575,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.9398390057901427e-05,
+      "loss": 0.0823,
+      "mean_token_accuracy": 0.9717945232987404,
+      "num_tokens": 16577026.0,
+      "step": 865
+    },
+    {
+      "entropy": 1.1650871604681015,
+      "epoch": 0.29795796052913226,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.939132890834628e-05,
+      "loss": 0.0952,
+      "mean_token_accuracy": 0.967523755133152,
+      "num_tokens": 16672193.0,
+      "step": 870
+    },
+    {
+      "entropy": 1.1446998178958894,
+      "epoch": 0.2996703626011388,
+      "grad_norm": 1.5234375,
+      "learning_rate": 1.938426775879113e-05,
+      "loss": 0.0867,
+      "mean_token_accuracy": 0.9706576213240623,
+      "num_tokens": 16768103.0,
+      "step": 875
+    },
+    {
+      "entropy": 1.1523787319660186,
+      "epoch": 0.30138276467314523,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.9377206609235987e-05,
+      "loss": 0.0861,
+      "mean_token_accuracy": 0.9712656766176224,
+      "num_tokens": 16863858.0,
+      "step": 880
+    },
+    {
+      "entropy": 1.1596208423376084,
+      "epoch": 0.30309516674515175,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.937014545968084e-05,
+      "loss": 0.0876,
+      "mean_token_accuracy": 0.9702959269285202,
+      "num_tokens": 16959453.0,
+      "step": 885
+    },
+    {
+      "entropy": 1.1840025901794433,
+      "epoch": 0.30480756881715826,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.936308431012569e-05,
+      "loss": 0.0792,
+      "mean_token_accuracy": 0.9746014758944511,
+      "num_tokens": 17054032.0,
+      "step": 890
+    },
+    {
+      "entropy": 1.1704025954008102,
+      "epoch": 0.3065199708891648,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.9356023160570544e-05,
+      "loss": 0.0944,
+      "mean_token_accuracy": 0.968149097263813,
+      "num_tokens": 17149702.0,
+      "step": 895
+    },
+    {
+      "entropy": 1.166870293021202,
+      "epoch": 0.3082323729611713,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.9348962011015396e-05,
+      "loss": 0.0854,
+      "mean_token_accuracy": 0.9683062836527825,
+      "num_tokens": 17246104.0,
+      "step": 900
+    },
+    {
+      "entropy": 1.1797128587961196,
+      "epoch": 0.3099447750331778,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.9341900861460248e-05,
+      "loss": 0.082,
+      "mean_token_accuracy": 0.9712266325950623,
+      "num_tokens": 17343307.0,
+      "step": 905
+    },
+    {
+      "entropy": 1.191875371336937,
+      "epoch": 0.3116571771051843,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.93348397119051e-05,
+      "loss": 0.0908,
+      "mean_token_accuracy": 0.9702668696641922,
+      "num_tokens": 17439118.0,
+      "step": 910
+    },
+    {
+      "entropy": 1.178999388217926,
+      "epoch": 0.3133695791771908,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.9327778562349952e-05,
+      "loss": 0.0726,
+      "mean_token_accuracy": 0.9764754861593247,
+      "num_tokens": 17535956.0,
+      "step": 915
+    },
+    {
+      "entropy": 1.2161524027585984,
+      "epoch": 0.3150819812491973,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.9320717412794804e-05,
+      "loss": 0.0751,
+      "mean_token_accuracy": 0.9758885353803635,
+      "num_tokens": 17630101.0,
+      "step": 920
+    },
+    {
+      "entropy": 1.1891431301832198,
+      "epoch": 0.3167943833212038,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.9313656263239657e-05,
+      "loss": 0.0689,
+      "mean_token_accuracy": 0.977847607433796,
+      "num_tokens": 17723904.0,
+      "step": 925
+    },
+    {
+      "entropy": 1.1786861628293992,
+      "epoch": 0.3185067853932103,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.930659511368451e-05,
+      "loss": 0.1093,
+      "mean_token_accuracy": 0.9639420002698899,
+      "num_tokens": 17820104.0,
+      "step": 930
+    },
+    {
+      "entropy": 1.1687486171722412,
+      "epoch": 0.3202191874652168,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.929953396412936e-05,
+      "loss": 0.0933,
+      "mean_token_accuracy": 0.9705073401331902,
+      "num_tokens": 17917624.0,
+      "step": 935
+    },
+    {
+      "entropy": 1.216786414384842,
+      "epoch": 0.32193158953722334,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.9292472814574213e-05,
+      "loss": 0.0778,
+      "mean_token_accuracy": 0.9745463475584983,
+      "num_tokens": 18014476.0,
+      "step": 940
+    },
+    {
+      "entropy": 1.2187707602977753,
+      "epoch": 0.32364399160922985,
+      "grad_norm": 0.93359375,
+      "learning_rate": 1.9285411665019065e-05,
+      "loss": 0.0795,
+      "mean_token_accuracy": 0.9739805281162262,
+      "num_tokens": 18111245.0,
+      "step": 945
+    },
+    {
+      "entropy": 1.204078447818756,
+      "epoch": 0.32535639368123637,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.927835051546392e-05,
+      "loss": 0.0901,
+      "mean_token_accuracy": 0.969705119729042,
+      "num_tokens": 18207009.0,
+      "step": 950
+    },
+    {
+      "entropy": 1.1895687460899353,
+      "epoch": 0.3270687957532429,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.9271289365908773e-05,
+      "loss": 0.0721,
+      "mean_token_accuracy": 0.9744245424866677,
+      "num_tokens": 18303021.0,
+      "step": 955
+    },
+    {
+      "entropy": 1.1670367866754532,
+      "epoch": 0.3287811978252494,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.9264228216353625e-05,
+      "loss": 0.0766,
+      "mean_token_accuracy": 0.974539938569069,
+      "num_tokens": 18401085.0,
+      "step": 960
+    },
+    {
+      "entropy": 1.1691705465316773,
+      "epoch": 0.33049359989725585,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.9257167066798477e-05,
+      "loss": 0.0905,
+      "mean_token_accuracy": 0.9697417616844177,
+      "num_tokens": 18495961.0,
+      "step": 965
+    },
+    {
+      "entropy": 1.197907094657421,
+      "epoch": 0.33220600196926237,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.925010591724333e-05,
+      "loss": 0.0698,
+      "mean_token_accuracy": 0.9757021367549896,
+      "num_tokens": 18591333.0,
+      "step": 970
+    },
+    {
+      "entropy": 1.2224839985370637,
+      "epoch": 0.3339184040412689,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.9243044767688182e-05,
+      "loss": 0.0923,
+      "mean_token_accuracy": 0.9703406438231468,
+      "num_tokens": 18686307.0,
+      "step": 975
+    },
+    {
+      "entropy": 1.1912748873233796,
+      "epoch": 0.3356308061132754,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.9235983618133034e-05,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9767110109329223,
+      "num_tokens": 18781938.0,
+      "step": 980
+    },
+    {
+      "entropy": 1.1654059916734696,
+      "epoch": 0.3373432081852819,
+      "grad_norm": 1.3984375,
+      "learning_rate": 1.9228922468577886e-05,
+      "loss": 0.0776,
+      "mean_token_accuracy": 0.9733897715806961,
+      "num_tokens": 18877859.0,
+      "step": 985
+    },
+    {
+      "entropy": 1.1751596838235856,
+      "epoch": 0.3390556102572884,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.9221861319022738e-05,
+      "loss": 0.1021,
+      "mean_token_accuracy": 0.9660407736897468,
+      "num_tokens": 18973276.0,
+      "step": 990
+    },
+    {
+      "entropy": 1.2088831305503844,
+      "epoch": 0.34076801232929493,
+      "grad_norm": 1.78125,
+      "learning_rate": 1.921480016946759e-05,
+      "loss": 0.0946,
+      "mean_token_accuracy": 0.9672083750367164,
+      "num_tokens": 19067271.0,
+      "step": 995
+    },
+    {
+      "entropy": 1.2160977214574813,
+      "epoch": 0.34248041440130145,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.9207739019912443e-05,
+      "loss": 0.0874,
+      "mean_token_accuracy": 0.9694693282246589,
+      "num_tokens": 19161798.0,
+      "step": 1000
+    },
+    {
+      "epoch": 0.34248041440130145,
+      "eval_entropy": 1.184665808430115,
+      "eval_loss": 0.08057019859552383,
+      "eval_mean_token_accuracy": 0.9731951729338698,
+      "eval_num_tokens": 19161798.0,
+      "eval_runtime": 343.6988,
+      "eval_samples_per_second": 15.103,
+      "eval_steps_per_second": 15.103,
+      "step": 1000
+    },
+    {
+      "entropy": 1.1762341767549516,
+      "epoch": 0.3441928164733079,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.9200677870357295e-05,
+      "loss": 0.0827,
+      "mean_token_accuracy": 0.9712936699390411,
+      "num_tokens": 19258020.0,
+      "step": 1005
+    },
+    {
+      "entropy": 1.176533767580986,
+      "epoch": 0.3459052185453144,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.9193616720802147e-05,
+      "loss": 0.072,
+      "mean_token_accuracy": 0.9765922427177429,
+      "num_tokens": 19352852.0,
+      "step": 1010
+    },
+    {
+      "entropy": 1.1692177772521972,
+      "epoch": 0.34761762061732093,
+      "grad_norm": 1.125,
+      "learning_rate": 1.9186555571247003e-05,
+      "loss": 0.0903,
+      "mean_token_accuracy": 0.9706610023975373,
+      "num_tokens": 19448321.0,
+      "step": 1015
+    },
+    {
+      "entropy": 1.185525080561638,
+      "epoch": 0.34933002268932745,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.9179494421691855e-05,
+      "loss": 0.0914,
+      "mean_token_accuracy": 0.9682233199477196,
+      "num_tokens": 19544813.0,
+      "step": 1020
+    },
+    {
+      "entropy": 1.1823008358478546,
+      "epoch": 0.35104242476133396,
+      "grad_norm": 1.3984375,
+      "learning_rate": 1.9172433272136707e-05,
+      "loss": 0.0824,
+      "mean_token_accuracy": 0.9712761744856835,
+      "num_tokens": 19638802.0,
+      "step": 1025
+    },
+    {
+      "entropy": 1.181131473183632,
+      "epoch": 0.3527548268333405,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.916537212258156e-05,
+      "loss": 0.0824,
+      "mean_token_accuracy": 0.973716153204441,
+      "num_tokens": 19734540.0,
+      "step": 1030
+    },
+    {
+      "entropy": 1.1901806712150573,
+      "epoch": 0.354467228905347,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.915831097302641e-05,
+      "loss": 0.0807,
+      "mean_token_accuracy": 0.9732174187898636,
+      "num_tokens": 19828641.0,
+      "step": 1035
+    },
+    {
+      "entropy": 1.1445277482271194,
+      "epoch": 0.3561796309773535,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.9151249823471263e-05,
+      "loss": 0.0789,
+      "mean_token_accuracy": 0.9736713021993637,
+      "num_tokens": 19924642.0,
+      "step": 1040
+    },
+    {
+      "entropy": 1.134951826930046,
+      "epoch": 0.35789203304936,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.9144188673916116e-05,
+      "loss": 0.0759,
+      "mean_token_accuracy": 0.9738373890519142,
+      "num_tokens": 20019895.0,
+      "step": 1045
+    },
+    {
+      "entropy": 1.1359785586595534,
+      "epoch": 0.3596044351213665,
+      "grad_norm": 1.1015625,
+      "learning_rate": 1.9137127524360968e-05,
+      "loss": 0.0794,
+      "mean_token_accuracy": 0.9729860201478004,
+      "num_tokens": 20115100.0,
+      "step": 1050
+    },
+    {
+      "entropy": 1.139900442957878,
+      "epoch": 0.361316837193373,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.913006637480582e-05,
+      "loss": 0.0711,
+      "mean_token_accuracy": 0.9760159462690353,
+      "num_tokens": 20209626.0,
+      "step": 1055
+    },
+    {
+      "entropy": 1.1350515216588974,
+      "epoch": 0.3630292392653795,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.9123005225250672e-05,
+      "loss": 0.0903,
+      "mean_token_accuracy": 0.9710188835859299,
+      "num_tokens": 20304525.0,
+      "step": 1060
+    },
+    {
+      "entropy": 1.1176694065332413,
+      "epoch": 0.364741641337386,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.9115944075695524e-05,
+      "loss": 0.0841,
+      "mean_token_accuracy": 0.9716746240854264,
+      "num_tokens": 20398498.0,
+      "step": 1065
+    },
+    {
+      "entropy": 1.1096349090337754,
+      "epoch": 0.3664540434093925,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.9108882926140376e-05,
+      "loss": 0.0858,
+      "mean_token_accuracy": 0.9728362917900085,
+      "num_tokens": 20495470.0,
+      "step": 1070
+    },
+    {
+      "entropy": 1.1560020476579667,
+      "epoch": 0.36816644548139904,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.910182177658523e-05,
+      "loss": 0.0811,
+      "mean_token_accuracy": 0.9735475718975067,
+      "num_tokens": 20591476.0,
+      "step": 1075
+    },
+    {
+      "entropy": 1.136641663312912,
+      "epoch": 0.36987884755340555,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.909476062703008e-05,
+      "loss": 0.0703,
+      "mean_token_accuracy": 0.9762310981750488,
+      "num_tokens": 20687762.0,
+      "step": 1080
+    },
+    {
+      "entropy": 1.1244812905788422,
+      "epoch": 0.37159124962541207,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.9087699477474936e-05,
+      "loss": 0.0883,
+      "mean_token_accuracy": 0.9711073860526085,
+      "num_tokens": 20783330.0,
+      "step": 1085
+    },
+    {
+      "entropy": 1.1223245054483413,
+      "epoch": 0.3733036516974186,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.908063832791979e-05,
+      "loss": 0.0731,
+      "mean_token_accuracy": 0.9754893377423286,
+      "num_tokens": 20879017.0,
+      "step": 1090
+    },
+    {
+      "entropy": 1.122566458582878,
+      "epoch": 0.37501605376942504,
+      "grad_norm": 1.0,
+      "learning_rate": 1.907357717836464e-05,
+      "loss": 0.0871,
+      "mean_token_accuracy": 0.9713185086846352,
+      "num_tokens": 20976461.0,
+      "step": 1095
+    },
+    {
+      "entropy": 1.1225494295358658,
+      "epoch": 0.37672845584143155,
+      "grad_norm": 0.9296875,
+      "learning_rate": 1.9066516028809493e-05,
+      "loss": 0.0703,
+      "mean_token_accuracy": 0.9770632356405258,
+      "num_tokens": 21072346.0,
+      "step": 1100
+    },
+    {
+      "entropy": 1.130875328183174,
+      "epoch": 0.37844085791343807,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.9059454879254345e-05,
+      "loss": 0.0852,
+      "mean_token_accuracy": 0.9714759200811386,
+      "num_tokens": 21166383.0,
+      "step": 1105
+    },
+    {
+      "entropy": 1.1106991320848465,
+      "epoch": 0.3801532599854446,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.9052393729699197e-05,
+      "loss": 0.0847,
+      "mean_token_accuracy": 0.9714381963014602,
+      "num_tokens": 21261335.0,
+      "step": 1110
+    },
+    {
+      "entropy": 1.1340815603733063,
+      "epoch": 0.3818656620574511,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.904533258014405e-05,
+      "loss": 0.0863,
+      "mean_token_accuracy": 0.9708552584052086,
+      "num_tokens": 21356040.0,
+      "step": 1115
+    },
+    {
+      "entropy": 1.1361641854047775,
+      "epoch": 0.3835780641294576,
+      "grad_norm": 1.296875,
+      "learning_rate": 1.90382714305889e-05,
+      "loss": 0.08,
+      "mean_token_accuracy": 0.9730976328253746,
+      "num_tokens": 21451169.0,
+      "step": 1120
+    },
+    {
+      "entropy": 1.1605969429016114,
+      "epoch": 0.3852904662014641,
+      "grad_norm": 1.375,
+      "learning_rate": 1.9031210281033754e-05,
+      "loss": 0.079,
+      "mean_token_accuracy": 0.9726199865341186,
+      "num_tokens": 21547984.0,
+      "step": 1125
+    },
+    {
+      "entropy": 1.1953546226024627,
+      "epoch": 0.38700286827347063,
+      "grad_norm": 0.890625,
+      "learning_rate": 1.9024149131478606e-05,
+      "loss": 0.0723,
+      "mean_token_accuracy": 0.9743477463722229,
+      "num_tokens": 21643581.0,
+      "step": 1130
+    },
+    {
+      "entropy": 1.202703031897545,
+      "epoch": 0.3887152703454771,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.9017087981923458e-05,
+      "loss": 0.0883,
+      "mean_token_accuracy": 0.9702430635690689,
+      "num_tokens": 21739881.0,
+      "step": 1135
+    },
+    {
+      "entropy": 1.1869259536266328,
+      "epoch": 0.3904276724174836,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.901002683236831e-05,
+      "loss": 0.0721,
+      "mean_token_accuracy": 0.9763333141803742,
+      "num_tokens": 21837269.0,
+      "step": 1140
+    },
+    {
+      "entropy": 1.2013211756944657,
+      "epoch": 0.3921400744894901,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.9002965682813162e-05,
+      "loss": 0.0828,
+      "mean_token_accuracy": 0.9735760897397995,
+      "num_tokens": 21932370.0,
+      "step": 1145
+    },
+    {
+      "entropy": 1.1329059958457948,
+      "epoch": 0.39385247656149663,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.8995904533258015e-05,
+      "loss": 0.0583,
+      "mean_token_accuracy": 0.9816904410719871,
+      "num_tokens": 22030080.0,
+      "step": 1150
+    },
+    {
+      "entropy": 1.1252098083496094,
+      "epoch": 0.39556487863350315,
+      "grad_norm": 1.3125,
+      "learning_rate": 1.898884338370287e-05,
+      "loss": 0.0813,
+      "mean_token_accuracy": 0.9703188315033913,
+      "num_tokens": 22125289.0,
+      "step": 1155
+    },
+    {
+      "entropy": 1.1263560235500336,
+      "epoch": 0.39727728070550966,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.8981782234147722e-05,
+      "loss": 0.0821,
+      "mean_token_accuracy": 0.9734075844287873,
+      "num_tokens": 22220042.0,
+      "step": 1160
+    },
+    {
+      "entropy": 1.1657070100307465,
+      "epoch": 0.3989896827775162,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.8974721084592575e-05,
+      "loss": 0.0773,
+      "mean_token_accuracy": 0.9741544544696807,
+      "num_tokens": 22315619.0,
+      "step": 1165
+    },
+    {
+      "entropy": 1.1746282905340195,
+      "epoch": 0.4007020848495227,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.8967659935037427e-05,
+      "loss": 0.086,
+      "mean_token_accuracy": 0.9719067171216011,
+      "num_tokens": 22411621.0,
+      "step": 1170
+    },
+    {
+      "entropy": 1.146896904706955,
+      "epoch": 0.4024144869215292,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.896059878548228e-05,
+      "loss": 0.0714,
+      "mean_token_accuracy": 0.9758814170956611,
+      "num_tokens": 22507999.0,
+      "step": 1175
+    },
+    {
+      "entropy": 1.134201779961586,
+      "epoch": 0.40412688899353566,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.895353763592713e-05,
+      "loss": 0.0802,
+      "mean_token_accuracy": 0.9721911549568176,
+      "num_tokens": 22602128.0,
+      "step": 1180
+    },
+    {
+      "entropy": 1.1231216818094254,
+      "epoch": 0.4058392910655422,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.8946476486371983e-05,
+      "loss": 0.0834,
+      "mean_token_accuracy": 0.9721213683485985,
+      "num_tokens": 22698359.0,
+      "step": 1185
+    },
+    {
+      "entropy": 1.1180975079536437,
+      "epoch": 0.4075516931375487,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.8939415336816835e-05,
+      "loss": 0.0682,
+      "mean_token_accuracy": 0.9775525942444802,
+      "num_tokens": 22794691.0,
+      "step": 1190
+    },
+    {
+      "entropy": 1.1064689725637435,
+      "epoch": 0.4092640952095552,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.8932354187261688e-05,
+      "loss": 0.0721,
+      "mean_token_accuracy": 0.9743425354361535,
+      "num_tokens": 22890528.0,
+      "step": 1195
+    },
+    {
+      "entropy": 1.0751424506306648,
+      "epoch": 0.4109764972815617,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.892529303770654e-05,
+      "loss": 0.0875,
+      "mean_token_accuracy": 0.9724462166428566,
+      "num_tokens": 22986341.0,
+      "step": 1200
+    },
+    {
+      "epoch": 0.4109764972815617,
+      "eval_entropy": 1.0812944073916813,
+      "eval_loss": 0.07848075777292252,
+      "eval_mean_token_accuracy": 0.9738113799322045,
+      "eval_num_tokens": 22986341.0,
+      "eval_runtime": 344.4528,
+      "eval_samples_per_second": 15.07,
+      "eval_steps_per_second": 15.07,
+      "step": 1200
+    },
+    {
+      "entropy": 1.0729010492563247,
+      "epoch": 0.41268889935356823,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.8918231888151392e-05,
+      "loss": 0.0742,
+      "mean_token_accuracy": 0.9755285680294037,
+      "num_tokens": 23082988.0,
+      "step": 1205
+    },
+    {
+      "entropy": 1.1028600260615349,
+      "epoch": 0.41440130142557474,
+      "grad_norm": 1.2890625,
+      "learning_rate": 1.8911170738596244e-05,
+      "loss": 0.0886,
+      "mean_token_accuracy": 0.9688351228833199,
+      "num_tokens": 23177562.0,
+      "step": 1210
+    },
+    {
+      "entropy": 1.1172732263803482,
+      "epoch": 0.41611370349758126,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.8904109589041096e-05,
+      "loss": 0.0775,
+      "mean_token_accuracy": 0.9736831769347191,
+      "num_tokens": 23272355.0,
+      "step": 1215
+    },
+    {
+      "entropy": 1.1159149900078773,
+      "epoch": 0.4178261055695877,
+      "grad_norm": 0.953125,
+      "learning_rate": 1.889704843948595e-05,
+      "loss": 0.0782,
+      "mean_token_accuracy": 0.974421463906765,
+      "num_tokens": 23369391.0,
+      "step": 1220
+    },
+    {
+      "entropy": 1.0732599392533302,
+      "epoch": 0.4195385076415942,
+      "grad_norm": 1.375,
+      "learning_rate": 1.8889987289930804e-05,
+      "loss": 0.0793,
+      "mean_token_accuracy": 0.9726202577352524,
+      "num_tokens": 23464327.0,
+      "step": 1225
+    },
+    {
+      "entropy": 1.0523262411355971,
+      "epoch": 0.42125090971360074,
+      "grad_norm": 0.92578125,
+      "learning_rate": 1.8882926140375656e-05,
+      "loss": 0.0732,
+      "mean_token_accuracy": 0.9761245265603066,
+      "num_tokens": 23560315.0,
+      "step": 1230
+    },
+    {
+      "entropy": 1.0740356594324112,
+      "epoch": 0.42296331178560725,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.887586499082051e-05,
+      "loss": 0.0679,
+      "mean_token_accuracy": 0.9772997796535492,
+      "num_tokens": 23656475.0,
+      "step": 1235
+    },
+    {
+      "entropy": 1.0834780499339103,
+      "epoch": 0.42467571385761377,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.886880384126536e-05,
+      "loss": 0.0637,
+      "mean_token_accuracy": 0.9777759954333305,
+      "num_tokens": 23753007.0,
+      "step": 1240
+    },
+    {
+      "entropy": 1.1189991503953933,
+      "epoch": 0.4263881159296203,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.8861742691710213e-05,
+      "loss": 0.0768,
+      "mean_token_accuracy": 0.9747056648135185,
+      "num_tokens": 23846466.0,
+      "step": 1245
+    },
+    {
+      "entropy": 1.0957374036312104,
+      "epoch": 0.4281005180016268,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.8854681542155065e-05,
+      "loss": 0.064,
+      "mean_token_accuracy": 0.9799482583999634,
+      "num_tokens": 23943373.0,
+      "step": 1250
+    },
+    {
+      "entropy": 1.115578231215477,
+      "epoch": 0.4298129200736333,
+      "grad_norm": 0.9765625,
+      "learning_rate": 1.8847620392599917e-05,
+      "loss": 0.0777,
+      "mean_token_accuracy": 0.9736348867416382,
+      "num_tokens": 24038741.0,
+      "step": 1255
+    },
+    {
+      "entropy": 1.0859564393758774,
+      "epoch": 0.4315253221456398,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.884055924304477e-05,
+      "loss": 0.0673,
+      "mean_token_accuracy": 0.9761564552783966,
+      "num_tokens": 24134677.0,
+      "step": 1260
+    },
+    {
+      "entropy": 1.0911764219403266,
+      "epoch": 0.4332377242176463,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.883349809348962e-05,
+      "loss": 0.0675,
+      "mean_token_accuracy": 0.9767653375864029,
+      "num_tokens": 24230410.0,
+      "step": 1265
+    },
+    {
+      "entropy": 1.0949071258306504,
+      "epoch": 0.4349501262896528,
+      "grad_norm": 1.1953125,
+      "learning_rate": 1.8826436943934474e-05,
+      "loss": 0.0742,
+      "mean_token_accuracy": 0.9746360689401626,
+      "num_tokens": 24327089.0,
+      "step": 1270
+    },
+    {
+      "entropy": 1.1227848380804062,
+      "epoch": 0.4366625283616593,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.8819375794379326e-05,
+      "loss": 0.0714,
+      "mean_token_accuracy": 0.9749162957072258,
+      "num_tokens": 24422324.0,
+      "step": 1275
+    },
+    {
+      "entropy": 1.1244820475578308,
+      "epoch": 0.4383749304336658,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.8812314644824178e-05,
+      "loss": 0.0775,
+      "mean_token_accuracy": 0.9728047639131546,
+      "num_tokens": 24517608.0,
+      "step": 1280
+    },
+    {
+      "entropy": 1.1480940580368042,
+      "epoch": 0.44008733250567234,
+      "grad_norm": 1.5703125,
+      "learning_rate": 1.880525349526903e-05,
+      "loss": 0.0806,
+      "mean_token_accuracy": 0.9733371615409852,
+      "num_tokens": 24611844.0,
+      "step": 1285
+    },
+    {
+      "entropy": 1.1213059842586517,
+      "epoch": 0.44179973457767885,
+      "grad_norm": 0.8046875,
+      "learning_rate": 1.8798192345713882e-05,
+      "loss": 0.0641,
+      "mean_token_accuracy": 0.978345374763012,
+      "num_tokens": 24707385.0,
+      "step": 1290
+    },
+    {
+      "entropy": 1.1082172825932504,
+      "epoch": 0.44351213664968536,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.8791131196158738e-05,
+      "loss": 0.0654,
+      "mean_token_accuracy": 0.9781587600708008,
+      "num_tokens": 24804482.0,
+      "step": 1295
+    },
+    {
+      "entropy": 1.1288210570812225,
+      "epoch": 0.4452245387216919,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.878407004660359e-05,
+      "loss": 0.0907,
+      "mean_token_accuracy": 0.9689996346831322,
+      "num_tokens": 24898992.0,
+      "step": 1300
+    },
+    {
+      "entropy": 1.1171072974801064,
+      "epoch": 0.44693694079369833,
+      "grad_norm": 0.953125,
+      "learning_rate": 1.8777008897048442e-05,
+      "loss": 0.0747,
+      "mean_token_accuracy": 0.9750395983457565,
+      "num_tokens": 24994998.0,
+      "step": 1305
+    },
+    {
+      "entropy": 1.1184355363249778,
+      "epoch": 0.44864934286570485,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.8769947747493294e-05,
+      "loss": 0.0795,
+      "mean_token_accuracy": 0.9729250952601433,
+      "num_tokens": 25088978.0,
+      "step": 1310
+    },
+    {
+      "entropy": 1.0859876736998557,
+      "epoch": 0.45036174493771136,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.8762886597938147e-05,
+      "loss": 0.0842,
+      "mean_token_accuracy": 0.9722599759697914,
+      "num_tokens": 25186519.0,
+      "step": 1315
+    },
+    {
+      "entropy": 1.0826068818569183,
+      "epoch": 0.4520741470097179,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.8755825448383e-05,
+      "loss": 0.0784,
+      "mean_token_accuracy": 0.9736651584506035,
+      "num_tokens": 25283303.0,
+      "step": 1320
+    },
+    {
+      "entropy": 1.1052915394306182,
+      "epoch": 0.4537865490817244,
+      "grad_norm": 0.9375,
+      "learning_rate": 1.874876429882785e-05,
+      "loss": 0.0693,
+      "mean_token_accuracy": 0.9749015867710114,
+      "num_tokens": 25377545.0,
+      "step": 1325
+    },
+    {
+      "entropy": 1.1142669394612312,
+      "epoch": 0.4554989511537309,
+      "grad_norm": 0.9140625,
+      "learning_rate": 1.8741703149272703e-05,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9768253564834595,
+      "num_tokens": 25471575.0,
+      "step": 1330
+    },
+    {
+      "entropy": 1.1039756879210472,
+      "epoch": 0.4572113532257374,
+      "grad_norm": 1.328125,
+      "learning_rate": 1.8734641999717555e-05,
+      "loss": 0.095,
+      "mean_token_accuracy": 0.9675980389118195,
+      "num_tokens": 25569541.0,
+      "step": 1335
+    },
+    {
+      "entropy": 1.1077512711286546,
+      "epoch": 0.45892375529774393,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.8727580850162407e-05,
+      "loss": 0.0804,
+      "mean_token_accuracy": 0.9747905045747757,
+      "num_tokens": 25664334.0,
+      "step": 1340
+    },
+    {
+      "entropy": 1.1091766357421875,
+      "epoch": 0.46063615736975044,
+      "grad_norm": 1.171875,
+      "learning_rate": 1.872051970060726e-05,
+      "loss": 0.0792,
+      "mean_token_accuracy": 0.9721957504749298,
+      "num_tokens": 25761088.0,
+      "step": 1345
+    },
+    {
+      "entropy": 1.1294567346572877,
+      "epoch": 0.4623485594417569,
+      "grad_norm": 1.6796875,
+      "learning_rate": 1.8713458551052112e-05,
+      "loss": 0.0814,
+      "mean_token_accuracy": 0.97224922478199,
+      "num_tokens": 25856690.0,
+      "step": 1350
+    },
+    {
+      "entropy": 1.152076357603073,
+      "epoch": 0.4640609615137634,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.8706397401496964e-05,
+      "loss": 0.0893,
+      "mean_token_accuracy": 0.9689715698361396,
+      "num_tokens": 25951668.0,
+      "step": 1355
+    },
+    {
+      "entropy": 1.1267032206058503,
+      "epoch": 0.46577336358576993,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.8699336251941816e-05,
+      "loss": 0.0769,
+      "mean_token_accuracy": 0.9732311427593231,
+      "num_tokens": 26047774.0,
+      "step": 1360
+    },
+    {
+      "entropy": 1.1367964625358582,
+      "epoch": 0.46748576565777644,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.869227510238667e-05,
+      "loss": 0.0806,
+      "mean_token_accuracy": 0.9723409309983253,
+      "num_tokens": 26141057.0,
+      "step": 1365
+    },
+    {
+      "entropy": 1.1020300745964051,
+      "epoch": 0.46919816772978296,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.8685213952831524e-05,
+      "loss": 0.0809,
+      "mean_token_accuracy": 0.9728222474455833,
+      "num_tokens": 26235989.0,
+      "step": 1370
+    },
+    {
+      "entropy": 1.105322277545929,
+      "epoch": 0.47091056980178947,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.8678152803276376e-05,
+      "loss": 0.0871,
+      "mean_token_accuracy": 0.9722603857517242,
+      "num_tokens": 26332587.0,
+      "step": 1375
+    },
+    {
+      "entropy": 1.120736826956272,
+      "epoch": 0.472622971873796,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.8671091653721228e-05,
+      "loss": 0.08,
+      "mean_token_accuracy": 0.9735232293605804,
+      "num_tokens": 26429198.0,
+      "step": 1380
+    },
+    {
+      "entropy": 1.1338796734809875,
+      "epoch": 0.4743353739458025,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.866403050416608e-05,
+      "loss": 0.0662,
+      "mean_token_accuracy": 0.977371446788311,
+      "num_tokens": 26526685.0,
+      "step": 1385
+    },
+    {
+      "entropy": 1.12310651242733,
+      "epoch": 0.47604777601780895,
+      "grad_norm": 0.96484375,
+      "learning_rate": 1.8656969354610933e-05,
+      "loss": 0.0898,
+      "mean_token_accuracy": 0.968597300350666,
+      "num_tokens": 26621660.0,
+      "step": 1390
+    },
+    {
+      "entropy": 1.1365027964115142,
+      "epoch": 0.47776017808981547,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.8649908205055785e-05,
+      "loss": 0.0969,
+      "mean_token_accuracy": 0.9676954999566079,
+      "num_tokens": 26715957.0,
+      "step": 1395
+    },
+    {
+      "entropy": 1.1175832107663155,
+      "epoch": 0.479472580161822,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.8642847055500637e-05,
+      "loss": 0.0725,
+      "mean_token_accuracy": 0.9743657141923905,
+      "num_tokens": 26812726.0,
+      "step": 1400
+    },
+    {
+      "epoch": 0.479472580161822,
+      "eval_entropy": 1.1091446698320506,
+      "eval_loss": 0.07670548558235168,
+      "eval_mean_token_accuracy": 0.9741229142005493,
+      "eval_num_tokens": 26812726.0,
+      "eval_runtime": 343.5433,
+      "eval_samples_per_second": 15.11,
+      "eval_steps_per_second": 15.11,
+      "step": 1400
+    },
+    {
+      "entropy": 1.1173082008957862,
+      "epoch": 0.4811849822338285,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.863578590594549e-05,
+      "loss": 0.0694,
+      "mean_token_accuracy": 0.9754531055688858,
+      "num_tokens": 26907455.0,
+      "step": 1405
+    },
+    {
+      "entropy": 1.1009103521704673,
+      "epoch": 0.482897384305835,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.862872475639034e-05,
+      "loss": 0.0961,
+      "mean_token_accuracy": 0.9665542811155319,
+      "num_tokens": 27002282.0,
+      "step": 1410
+    },
+    {
+      "entropy": 1.0813133031129838,
+      "epoch": 0.4846097863778415,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.8621663606835193e-05,
+      "loss": 0.0776,
+      "mean_token_accuracy": 0.9734799340367317,
+      "num_tokens": 27097740.0,
+      "step": 1415
+    },
+    {
+      "entropy": 1.1056796163320541,
+      "epoch": 0.48632218844984804,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.8614602457280046e-05,
+      "loss": 0.0701,
+      "mean_token_accuracy": 0.9753427192568779,
+      "num_tokens": 27192342.0,
+      "step": 1420
+    },
+    {
+      "entropy": 1.0902955189347268,
+      "epoch": 0.48803459052185455,
+      "grad_norm": 1.328125,
+      "learning_rate": 1.8607541307724898e-05,
+      "loss": 0.0824,
+      "mean_token_accuracy": 0.971779927611351,
+      "num_tokens": 27286470.0,
+      "step": 1425
+    },
+    {
+      "entropy": 1.0736634850502014,
+      "epoch": 0.48974699259386106,
+      "grad_norm": 1.3359375,
+      "learning_rate": 1.860048015816975e-05,
+      "loss": 0.0688,
+      "mean_token_accuracy": 0.9764139145612717,
+      "num_tokens": 27381777.0,
+      "step": 1430
+    },
+    {
+      "entropy": 1.0835579365491868,
+      "epoch": 0.4914593946658675,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.8593419008614605e-05,
+      "loss": 0.0749,
+      "mean_token_accuracy": 0.9745497271418572,
+      "num_tokens": 27478230.0,
+      "step": 1435
+    },
+    {
+      "entropy": 1.0621713683009149,
+      "epoch": 0.49317179673787404,
+      "grad_norm": 0.94140625,
+      "learning_rate": 1.8586357859059458e-05,
+      "loss": 0.0758,
+      "mean_token_accuracy": 0.9750664070248604,
+      "num_tokens": 27574710.0,
+      "step": 1440
+    },
+    {
+      "entropy": 1.0537893816828727,
+      "epoch": 0.49488419880988055,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.857929670950431e-05,
+      "loss": 0.0732,
+      "mean_token_accuracy": 0.974408708512783,
+      "num_tokens": 27671297.0,
+      "step": 1445
+    },
+    {
+      "entropy": 1.0498380094766617,
+      "epoch": 0.49659660088188706,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.8572235559949162e-05,
+      "loss": 0.0717,
+      "mean_token_accuracy": 0.9738784611225129,
+      "num_tokens": 27767609.0,
+      "step": 1450
+    },
+    {
+      "entropy": 1.0688601344823838,
+      "epoch": 0.4983090029538936,
+      "grad_norm": 1.0,
+      "learning_rate": 1.8565174410394014e-05,
+      "loss": 0.0764,
+      "mean_token_accuracy": 0.9731876507401467,
+      "num_tokens": 27865163.0,
+      "step": 1455
+    },
+    {
+      "entropy": 1.093703928589821,
+      "epoch": 0.5000214050259001,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.8558113260838866e-05,
+      "loss": 0.0766,
+      "mean_token_accuracy": 0.9748765379190445,
+      "num_tokens": 27961199.0,
+      "step": 1460
+    },
+    {
+      "entropy": 1.1209586292505265,
+      "epoch": 0.5017338070979066,
+      "grad_norm": 0.94140625,
+      "learning_rate": 1.855105211128372e-05,
+      "loss": 0.076,
+      "mean_token_accuracy": 0.9739302903413772,
+      "num_tokens": 28056637.0,
+      "step": 1465
+    },
+    {
+      "entropy": 1.1187533110380172,
+      "epoch": 0.5034462091699131,
+      "grad_norm": 1.125,
+      "learning_rate": 1.854399096172857e-05,
+      "loss": 0.0888,
+      "mean_token_accuracy": 0.9700263679027558,
+      "num_tokens": 28153784.0,
+      "step": 1470
+    },
+    {
+      "entropy": 1.1052709579467774,
+      "epoch": 0.5051586112419196,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.8536929812173423e-05,
+      "loss": 0.0878,
+      "mean_token_accuracy": 0.9698923602700233,
+      "num_tokens": 28248434.0,
+      "step": 1475
+    },
+    {
+      "entropy": 1.073799839615822,
+      "epoch": 0.5068710133139261,
+      "grad_norm": 1.0,
+      "learning_rate": 1.8529868662618275e-05,
+      "loss": 0.0695,
+      "mean_token_accuracy": 0.9760001584887504,
+      "num_tokens": 28347592.0,
+      "step": 1480
+    },
+    {
+      "entropy": 1.0971616566181184,
+      "epoch": 0.5085834153859327,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.8522807513063127e-05,
+      "loss": 0.0845,
+      "mean_token_accuracy": 0.9726158648729324,
+      "num_tokens": 28442474.0,
+      "step": 1485
+    },
+    {
+      "entropy": 1.0925533041357993,
+      "epoch": 0.5102958174579392,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.851574636350798e-05,
+      "loss": 0.0866,
+      "mean_token_accuracy": 0.9709009021520615,
+      "num_tokens": 28538777.0,
+      "step": 1490
+    },
+    {
+      "entropy": 1.0947575479745866,
+      "epoch": 0.5120082195299457,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.850868521395283e-05,
+      "loss": 0.0702,
+      "mean_token_accuracy": 0.9763423159718514,
+      "num_tokens": 28635186.0,
+      "step": 1495
+    },
+    {
+      "entropy": 1.1196099147200584,
+      "epoch": 0.5137206216019521,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.8501624064397684e-05,
+      "loss": 0.076,
+      "mean_token_accuracy": 0.9735683783888817,
+      "num_tokens": 28730973.0,
+      "step": 1500
+    },
+    {
+      "entropy": 1.112065863609314,
+      "epoch": 0.5154330236739586,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.849456291484254e-05,
+      "loss": 0.0977,
+      "mean_token_accuracy": 0.9675616696476936,
+      "num_tokens": 28826544.0,
+      "step": 1505
+    },
+    {
+      "entropy": 1.099971030652523,
+      "epoch": 0.5171454257459651,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.848750176528739e-05,
+      "loss": 0.0904,
+      "mean_token_accuracy": 0.9687278628349304,
+      "num_tokens": 28922516.0,
+      "step": 1510
+    },
+    {
+      "entropy": 1.0816262930631637,
+      "epoch": 0.5188578278179716,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.8480440615732244e-05,
+      "loss": 0.0824,
+      "mean_token_accuracy": 0.9728308230638504,
+      "num_tokens": 29018290.0,
+      "step": 1515
+    },
+    {
+      "entropy": 1.0814829096198082,
+      "epoch": 0.5205702298899781,
+      "grad_norm": 1.265625,
+      "learning_rate": 1.8473379466177096e-05,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9751347601413727,
+      "num_tokens": 29114830.0,
+      "step": 1520
+    },
+    {
+      "entropy": 1.1055495515465736,
+      "epoch": 0.5222826319619847,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.8466318316621948e-05,
+      "loss": 0.0787,
+      "mean_token_accuracy": 0.9734732165932656,
+      "num_tokens": 29210613.0,
+      "step": 1525
+    },
+    {
+      "entropy": 1.1160762310028076,
+      "epoch": 0.5239950340339912,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.84592571670668e-05,
+      "loss": 0.0851,
+      "mean_token_accuracy": 0.9684530064463616,
+      "num_tokens": 29305506.0,
+      "step": 1530
+    },
+    {
+      "entropy": 1.128273183107376,
+      "epoch": 0.5257074361059977,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.8452196017511652e-05,
+      "loss": 0.08,
+      "mean_token_accuracy": 0.9709644392132759,
+      "num_tokens": 29400275.0,
+      "step": 1535
+    },
+    {
+      "entropy": 1.1217272847890853,
+      "epoch": 0.5274198381780042,
+      "grad_norm": 1.1015625,
+      "learning_rate": 1.8445134867956505e-05,
+      "loss": 0.0779,
+      "mean_token_accuracy": 0.9738597184419632,
+      "num_tokens": 29495597.0,
+      "step": 1540
+    },
+    {
+      "entropy": 1.1326696157455445,
+      "epoch": 0.5291322402500107,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.8438073718401357e-05,
+      "loss": 0.0712,
+      "mean_token_accuracy": 0.9752731010317802,
+      "num_tokens": 29590665.0,
+      "step": 1545
+    },
+    {
+      "entropy": 1.1165046751499177,
+      "epoch": 0.5308446423220172,
+      "grad_norm": 0.8984375,
+      "learning_rate": 1.843101256884621e-05,
+      "loss": 0.0655,
+      "mean_token_accuracy": 0.9761840745806694,
+      "num_tokens": 29688035.0,
+      "step": 1550
+    },
+    {
+      "entropy": 1.1177081316709518,
+      "epoch": 0.5325570443940237,
+      "grad_norm": 1.0,
+      "learning_rate": 1.842395141929106e-05,
+      "loss": 0.0585,
+      "mean_token_accuracy": 0.9791857406497002,
+      "num_tokens": 29782471.0,
+      "step": 1555
+    },
+    {
+      "entropy": 1.099551397562027,
+      "epoch": 0.5342694464660303,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.8416890269735913e-05,
+      "loss": 0.0739,
+      "mean_token_accuracy": 0.9745470687747002,
+      "num_tokens": 29876290.0,
+      "step": 1560
+    },
+    {
+      "entropy": 1.08235345184803,
+      "epoch": 0.5359818485380368,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.8409829120180765e-05,
+      "loss": 0.0772,
+      "mean_token_accuracy": 0.9738533869385719,
+      "num_tokens": 29971080.0,
+      "step": 1565
+    },
+    {
+      "entropy": 1.0917438238859176,
+      "epoch": 0.5376942506100433,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.8402767970625618e-05,
+      "loss": 0.0752,
+      "mean_token_accuracy": 0.9734141230583191,
+      "num_tokens": 30067971.0,
+      "step": 1570
+    },
+    {
+      "entropy": 1.1280545637011528,
+      "epoch": 0.5394066526820498,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.8395706821070473e-05,
+      "loss": 0.0791,
+      "mean_token_accuracy": 0.9722218483686447,
+      "num_tokens": 30163407.0,
+      "step": 1575
+    },
+    {
+      "entropy": 1.1376834154129027,
+      "epoch": 0.5411190547540563,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.8388645671515325e-05,
+      "loss": 0.0844,
+      "mean_token_accuracy": 0.9709107607603074,
+      "num_tokens": 30258139.0,
+      "step": 1580
+    },
+    {
+      "entropy": 1.125060684978962,
+      "epoch": 0.5428314568260627,
+      "grad_norm": 1.125,
+      "learning_rate": 1.8381584521960177e-05,
+      "loss": 0.0767,
+      "mean_token_accuracy": 0.974628047645092,
+      "num_tokens": 30355266.0,
+      "step": 1585
+    },
+    {
+      "entropy": 1.1403662621974946,
+      "epoch": 0.5445438588980692,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.837452337240503e-05,
+      "loss": 0.0702,
+      "mean_token_accuracy": 0.976131497323513,
+      "num_tokens": 30449150.0,
+      "step": 1590
+    },
+    {
+      "entropy": 1.1283534318208694,
+      "epoch": 0.5462562609700757,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.8367462222849882e-05,
+      "loss": 0.0723,
+      "mean_token_accuracy": 0.9750758990645408,
+      "num_tokens": 30545831.0,
+      "step": 1595
+    },
+    {
+      "entropy": 1.1351796254515647,
+      "epoch": 0.5479686630420822,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.8360401073294734e-05,
+      "loss": 0.0739,
+      "mean_token_accuracy": 0.9743103638291359,
+      "num_tokens": 30643392.0,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5479686630420822,
+      "eval_entropy": 1.1286422604197277,
+      "eval_loss": 0.07486434280872345,
+      "eval_mean_token_accuracy": 0.9743929062360882,
+      "eval_num_tokens": 30643392.0,
+      "eval_runtime": 343.9142,
+      "eval_samples_per_second": 15.094,
+      "eval_steps_per_second": 15.094,
+      "step": 1600
+    },
+    {
+      "entropy": 1.1078941345214843,
+      "epoch": 0.5496810651140888,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.8353339923739586e-05,
+      "loss": 0.0584,
+      "mean_token_accuracy": 0.9796475604176521,
+      "num_tokens": 30738742.0,
+      "step": 1605
+    },
+    {
+      "entropy": 1.1059675514698029,
+      "epoch": 0.5513934671860953,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.834627877418444e-05,
+      "loss": 0.0833,
+      "mean_token_accuracy": 0.9719104275107384,
+      "num_tokens": 30835712.0,
+      "step": 1610
+    },
+    {
+      "entropy": 1.094590501487255,
+      "epoch": 0.5531058692581018,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.833921762462929e-05,
+      "loss": 0.0661,
+      "mean_token_accuracy": 0.97759860008955,
+      "num_tokens": 30932938.0,
+      "step": 1615
+    },
+    {
+      "entropy": 1.0952149495482444,
+      "epoch": 0.5548182713301083,
+      "grad_norm": 1.1953125,
+      "learning_rate": 1.8332156475074143e-05,
+      "loss": 0.073,
+      "mean_token_accuracy": 0.9743284121155739,
+      "num_tokens": 31029358.0,
+      "step": 1620
+    },
+    {
+      "entropy": 1.0952559739351273,
+      "epoch": 0.5565306734021148,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.8325095325518995e-05,
+      "loss": 0.0734,
+      "mean_token_accuracy": 0.9750207558274269,
+      "num_tokens": 31124669.0,
+      "step": 1625
+    },
+    {
+      "entropy": 1.065773655474186,
+      "epoch": 0.5582430754741213,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.8318034175963847e-05,
+      "loss": 0.0738,
+      "mean_token_accuracy": 0.9743185415863991,
+      "num_tokens": 31220346.0,
+      "step": 1630
+    },
+    {
+      "entropy": 1.0924474328756333,
+      "epoch": 0.5599554775461278,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.83109730264087e-05,
+      "loss": 0.0849,
+      "mean_token_accuracy": 0.9700090736150742,
+      "num_tokens": 31315005.0,
+      "step": 1635
+    },
+    {
+      "entropy": 1.0915605559945107,
+      "epoch": 0.5616678796181344,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.830391187685355e-05,
+      "loss": 0.0744,
+      "mean_token_accuracy": 0.9750060677528382,
+      "num_tokens": 31410551.0,
+      "step": 1640
+    },
+    {
+      "entropy": 1.0771109610795975,
+      "epoch": 0.5633802816901409,
+      "grad_norm": 0.99609375,
+      "learning_rate": 1.8296850727298407e-05,
+      "loss": 0.0707,
+      "mean_token_accuracy": 0.9761269629001618,
+      "num_tokens": 31508633.0,
+      "step": 1645
+    },
+    {
+      "entropy": 1.1171609967947007,
+      "epoch": 0.5650926837621474,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.828978957774326e-05,
+      "loss": 0.0648,
+      "mean_token_accuracy": 0.9781188502907753,
+      "num_tokens": 31605184.0,
+      "step": 1650
+    },
+    {
+      "entropy": 1.1474918186664582,
+      "epoch": 0.5668050858341539,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.828272842818811e-05,
+      "loss": 0.091,
+      "mean_token_accuracy": 0.9693744152784347,
+      "num_tokens": 31701845.0,
+      "step": 1655
+    },
+    {
+      "entropy": 1.1507633730769158,
+      "epoch": 0.5685174879061604,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.8275667278632963e-05,
+      "loss": 0.0938,
+      "mean_token_accuracy": 0.9682975247502327,
+      "num_tokens": 31797362.0,
+      "step": 1660
+    },
+    {
+      "entropy": 1.1760642051696777,
+      "epoch": 0.5702298899781669,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.8268606129077816e-05,
+      "loss": 0.0798,
+      "mean_token_accuracy": 0.9729333654046058,
+      "num_tokens": 31892426.0,
+      "step": 1665
+    },
+    {
+      "entropy": 1.1229776382446288,
+      "epoch": 0.5719422920501733,
+      "grad_norm": 0.96875,
+      "learning_rate": 1.8261544979522668e-05,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9753845900297164,
+      "num_tokens": 31989986.0,
+      "step": 1670
+    },
+    {
+      "entropy": 1.1255135804414749,
+      "epoch": 0.5736546941221798,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.825448382996752e-05,
+      "loss": 0.0687,
+      "mean_token_accuracy": 0.9756256759166717,
+      "num_tokens": 32086075.0,
+      "step": 1675
+    },
+    {
+      "entropy": 1.0982681214809418,
+      "epoch": 0.5753670961941864,
+      "grad_norm": 0.828125,
+      "learning_rate": 1.8247422680412372e-05,
+      "loss": 0.0591,
+      "mean_token_accuracy": 0.979189184308052,
+      "num_tokens": 32181987.0,
+      "step": 1680
+    },
+    {
+      "entropy": 1.0815881341695786,
+      "epoch": 0.5770794982661929,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.8240361530857224e-05,
+      "loss": 0.0625,
+      "mean_token_accuracy": 0.9776791915297508,
+      "num_tokens": 32276951.0,
+      "step": 1685
+    },
+    {
+      "entropy": 1.0643230020999908,
+      "epoch": 0.5787919003381994,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.8233300381302077e-05,
+      "loss": 0.0781,
+      "mean_token_accuracy": 0.9719484120607376,
+      "num_tokens": 32373805.0,
+      "step": 1690
+    },
+    {
+      "entropy": 1.0802049085497856,
+      "epoch": 0.5805043024102059,
+      "grad_norm": 1.25,
+      "learning_rate": 1.822623923174693e-05,
+      "loss": 0.0594,
+      "mean_token_accuracy": 0.9803182989358902,
+      "num_tokens": 32468960.0,
+      "step": 1695
+    },
+    {
+      "entropy": 1.057201734185219,
+      "epoch": 0.5822167044822124,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.821917808219178e-05,
+      "loss": 0.0789,
+      "mean_token_accuracy": 0.9730697616934776,
+      "num_tokens": 32566865.0,
+      "step": 1700
+    },
+    {
+      "entropy": 1.0866804271936417,
+      "epoch": 0.5839291065542189,
+      "grad_norm": 1.1015625,
+      "learning_rate": 1.8212116932636633e-05,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.9764628037810326,
+      "num_tokens": 32662398.0,
+      "step": 1705
+    },
+    {
+      "entropy": 1.1012579217553138,
+      "epoch": 0.5856415086262254,
+      "grad_norm": 1.25,
+      "learning_rate": 1.8205055783081485e-05,
+      "loss": 0.0797,
+      "mean_token_accuracy": 0.9717117264866829,
+      "num_tokens": 32758010.0,
+      "step": 1710
+    },
+    {
+      "entropy": 1.092042450606823,
+      "epoch": 0.587353910698232,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.819799463352634e-05,
+      "loss": 0.0649,
+      "mean_token_accuracy": 0.9781408607959747,
+      "num_tokens": 32852960.0,
+      "step": 1715
+    },
+    {
+      "entropy": 1.0619764447212219,
+      "epoch": 0.5890663127702385,
+      "grad_norm": 0.8671875,
+      "learning_rate": 1.8190933483971193e-05,
+      "loss": 0.0679,
+      "mean_token_accuracy": 0.975428955256939,
+      "num_tokens": 32950249.0,
+      "step": 1720
+    },
+    {
+      "entropy": 1.0530066207051276,
+      "epoch": 0.590778714842245,
+      "grad_norm": 0.8515625,
+      "learning_rate": 1.8183872334416045e-05,
+      "loss": 0.0691,
+      "mean_token_accuracy": 0.9765373587608337,
+      "num_tokens": 33045951.0,
+      "step": 1725
+    },
+    {
+      "entropy": 1.0625995561480521,
+      "epoch": 0.5924911169142515,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.8176811184860897e-05,
+      "loss": 0.0828,
+      "mean_token_accuracy": 0.972608408331871,
+      "num_tokens": 33141621.0,
+      "step": 1730
+    },
+    {
+      "entropy": 1.0762339487671853,
+      "epoch": 0.594203518986258,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.816975003530575e-05,
+      "loss": 0.0865,
+      "mean_token_accuracy": 0.970147967338562,
+      "num_tokens": 33236913.0,
+      "step": 1735
+    },
+    {
+      "entropy": 1.072178229689598,
+      "epoch": 0.5959159210582645,
+      "grad_norm": 1.171875,
+      "learning_rate": 1.81626888857506e-05,
+      "loss": 0.0669,
+      "mean_token_accuracy": 0.9769659787416458,
+      "num_tokens": 33332173.0,
+      "step": 1740
+    },
+    {
+      "entropy": 1.0843237221240998,
+      "epoch": 0.597628323130271,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.8155627736195454e-05,
+      "loss": 0.0849,
+      "mean_token_accuracy": 0.9713902458548546,
+      "num_tokens": 33427806.0,
+      "step": 1745
+    },
+    {
+      "entropy": 1.0805180773139,
+      "epoch": 0.5993407252022775,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.8148566586640306e-05,
+      "loss": 0.0674,
+      "mean_token_accuracy": 0.9748094469308853,
+      "num_tokens": 33524332.0,
+      "step": 1750
+    },
+    {
+      "entropy": 1.0669876322150231,
+      "epoch": 0.601053127274284,
+      "grad_norm": 0.90234375,
+      "learning_rate": 1.8141505437085158e-05,
+      "loss": 0.0729,
+      "mean_token_accuracy": 0.9747355252504348,
+      "num_tokens": 33621199.0,
+      "step": 1755
+    },
+    {
+      "entropy": 1.0862119302153588,
+      "epoch": 0.6027655293462905,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.813444428753001e-05,
+      "loss": 0.0706,
+      "mean_token_accuracy": 0.9761605769395828,
+      "num_tokens": 33717423.0,
+      "step": 1760
+    },
+    {
+      "entropy": 1.0796998351812364,
+      "epoch": 0.604477931418297,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.8127383137974862e-05,
+      "loss": 0.0602,
+      "mean_token_accuracy": 0.978901955485344,
+      "num_tokens": 33812150.0,
+      "step": 1765
+    },
+    {
+      "entropy": 1.0544181913137436,
+      "epoch": 0.6061903334903035,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.8120321988419715e-05,
+      "loss": 0.0647,
+      "mean_token_accuracy": 0.9788004845380783,
+      "num_tokens": 33907966.0,
+      "step": 1770
+    },
+    {
+      "entropy": 1.0548089489340782,
+      "epoch": 0.60790273556231,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.8113260838864567e-05,
+      "loss": 0.0591,
+      "mean_token_accuracy": 0.9789862528443336,
+      "num_tokens": 34004605.0,
+      "step": 1775
+    },
+    {
+      "entropy": 1.084161141514778,
+      "epoch": 0.6096151376343165,
+      "grad_norm": 1.0625,
+      "learning_rate": 1.810619968930942e-05,
+      "loss": 0.0791,
+      "mean_token_accuracy": 0.9731057688593865,
+      "num_tokens": 34100565.0,
+      "step": 1780
+    },
+    {
+      "entropy": 1.0813875496387482,
+      "epoch": 0.611327539706323,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.8099138539754275e-05,
+      "loss": 0.0613,
+      "mean_token_accuracy": 0.9782551780343056,
+      "num_tokens": 34194891.0,
+      "step": 1785
+    },
+    {
+      "entropy": 1.0782890811562538,
+      "epoch": 0.6130399417783295,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.8092077390199127e-05,
+      "loss": 0.0822,
+      "mean_token_accuracy": 0.9718855217099189,
+      "num_tokens": 34291416.0,
+      "step": 1790
+    },
+    {
+      "entropy": 1.0406158477067948,
+      "epoch": 0.6147523438503361,
+      "grad_norm": 0.95703125,
+      "learning_rate": 1.808501624064398e-05,
+      "loss": 0.0635,
+      "mean_token_accuracy": 0.9779545396566391,
+      "num_tokens": 34388459.0,
+      "step": 1795
+    },
+    {
+      "entropy": 1.0450820043683051,
+      "epoch": 0.6164647459223426,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.807795509108883e-05,
+      "loss": 0.0792,
+      "mean_token_accuracy": 0.9734042763710022,
+      "num_tokens": 34483640.0,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6164647459223426,
+      "eval_entropy": 1.0251484084418687,
+      "eval_loss": 0.07354158908128738,
+      "eval_mean_token_accuracy": 0.9749061561370259,
+      "eval_num_tokens": 34483640.0,
+      "eval_runtime": 341.6354,
+      "eval_samples_per_second": 15.195,
+      "eval_steps_per_second": 15.195,
+      "step": 1800
+    },
+    {
+      "entropy": 1.0292415514588356,
+      "epoch": 0.6181771479943491,
+      "grad_norm": 1.0625,
+      "learning_rate": 1.8070893941533683e-05,
+      "loss": 0.0747,
+      "mean_token_accuracy": 0.9743593811988831,
+      "num_tokens": 34580800.0,
+      "step": 1805
+    },
+    {
+      "entropy": 1.0573086097836495,
+      "epoch": 0.6198895500663556,
+      "grad_norm": 0.86328125,
+      "learning_rate": 1.8063832791978535e-05,
+      "loss": 0.0617,
+      "mean_token_accuracy": 0.9797847896814347,
+      "num_tokens": 34677291.0,
+      "step": 1810
+    },
+    {
+      "entropy": 1.060031969845295,
+      "epoch": 0.6216019521383621,
+      "grad_norm": 1.2890625,
+      "learning_rate": 1.8056771642423388e-05,
+      "loss": 0.0673,
+      "mean_token_accuracy": 0.9779677763581276,
+      "num_tokens": 34773835.0,
+      "step": 1815
+    },
+    {
+      "entropy": 1.0707685634493829,
+      "epoch": 0.6233143542103686,
+      "grad_norm": 0.94921875,
+      "learning_rate": 1.804971049286824e-05,
+      "loss": 0.0748,
+      "mean_token_accuracy": 0.9731908693909646,
+      "num_tokens": 34871431.0,
+      "step": 1820
+    },
+    {
+      "entropy": 1.056957259774208,
+      "epoch": 0.6250267562823751,
+      "grad_norm": 1.125,
+      "learning_rate": 1.8042649343313092e-05,
+      "loss": 0.0749,
+      "mean_token_accuracy": 0.9742481961846352,
+      "num_tokens": 34967562.0,
+      "step": 1825
+    },
+    {
+      "entropy": 1.041993832588196,
+      "epoch": 0.6267391583543817,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.8035588193757944e-05,
+      "loss": 0.0692,
+      "mean_token_accuracy": 0.9756530299782753,
+      "num_tokens": 35064952.0,
+      "step": 1830
+    },
+    {
+      "entropy": 1.0403671085834503,
+      "epoch": 0.6284515604263882,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.8028527044202796e-05,
+      "loss": 0.0873,
+      "mean_token_accuracy": 0.9713320881128311,
+      "num_tokens": 35160442.0,
+      "step": 1835
+    },
+    {
+      "entropy": 1.0346712574362755,
+      "epoch": 0.6301639624983946,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.802146589464765e-05,
+      "loss": 0.0626,
+      "mean_token_accuracy": 0.9778290823101997,
+      "num_tokens": 35255546.0,
+      "step": 1840
+    },
+    {
+      "entropy": 1.0228869020938873,
+      "epoch": 0.6318763645704011,
+      "grad_norm": 1.328125,
+      "learning_rate": 1.80144047450925e-05,
+      "loss": 0.0615,
+      "mean_token_accuracy": 0.9791113749146462,
+      "num_tokens": 35352822.0,
+      "step": 1845
+    },
+    {
+      "entropy": 1.0281955137848855,
+      "epoch": 0.6335887666424076,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.8007343595537353e-05,
+      "loss": 0.0819,
+      "mean_token_accuracy": 0.9717352703213692,
+      "num_tokens": 35449769.0,
+      "step": 1850
+    },
+    {
+      "entropy": 1.0498301699757575,
+      "epoch": 0.6353011687144141,
+      "grad_norm": 1.1953125,
+      "learning_rate": 1.800028244598221e-05,
+      "loss": 0.0716,
+      "mean_token_accuracy": 0.9757066443562508,
+      "num_tokens": 35546271.0,
+      "step": 1855
+    },
+    {
+      "entropy": 1.0463885709643364,
+      "epoch": 0.6370135707864206,
+      "grad_norm": 0.96484375,
+      "learning_rate": 1.799322129642706e-05,
+      "loss": 0.0663,
+      "mean_token_accuracy": 0.9780217185616493,
+      "num_tokens": 35644183.0,
+      "step": 1860
+    },
+    {
+      "entropy": 1.0546012654900552,
+      "epoch": 0.6387259728584271,
+      "grad_norm": 1.078125,
+      "learning_rate": 1.7986160146871913e-05,
+      "loss": 0.0713,
+      "mean_token_accuracy": 0.9775891959667206,
+      "num_tokens": 35740578.0,
+      "step": 1865
+    },
+    {
+      "entropy": 1.0453653767704965,
+      "epoch": 0.6404383749304337,
+      "grad_norm": 0.84375,
+      "learning_rate": 1.7979098997316765e-05,
+      "loss": 0.0625,
+      "mean_token_accuracy": 0.9783225864171982,
+      "num_tokens": 35837143.0,
+      "step": 1870
+    },
+    {
+      "entropy": 1.0281398952007295,
+      "epoch": 0.6421507770024402,
+      "grad_norm": 0.984375,
+      "learning_rate": 1.7972037847761617e-05,
+      "loss": 0.0645,
+      "mean_token_accuracy": 0.9764010831713676,
+      "num_tokens": 35933446.0,
+      "step": 1875
+    },
+    {
+      "entropy": 1.0106928437948226,
+      "epoch": 0.6438631790744467,
+      "grad_norm": 1.3359375,
+      "learning_rate": 1.796497669820647e-05,
+      "loss": 0.0832,
+      "mean_token_accuracy": 0.9733222305774689,
+      "num_tokens": 36030394.0,
+      "step": 1880
+    },
+    {
+      "entropy": 1.0186400935053825,
+      "epoch": 0.6455755811464532,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.795791554865132e-05,
+      "loss": 0.0694,
+      "mean_token_accuracy": 0.9761451289057732,
+      "num_tokens": 36126049.0,
+      "step": 1885
+    },
+    {
+      "entropy": 1.0065297693014146,
+      "epoch": 0.6472879832184597,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.7950854399096174e-05,
+      "loss": 0.0634,
+      "mean_token_accuracy": 0.977371571958065,
+      "num_tokens": 36220337.0,
+      "step": 1890
+    },
+    {
+      "entropy": 0.9792183041572571,
+      "epoch": 0.6490003852904662,
+      "grad_norm": 1.0,
+      "learning_rate": 1.7943793249541026e-05,
+      "loss": 0.0597,
+      "mean_token_accuracy": 0.9776044756174087,
+      "num_tokens": 36315432.0,
+      "step": 1895
+    },
+    {
+      "entropy": 0.9598794534802437,
+      "epoch": 0.6507127873624727,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.7936732099985878e-05,
+      "loss": 0.0632,
+      "mean_token_accuracy": 0.9762929394841194,
+      "num_tokens": 36411889.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.9933922991156579,
+      "epoch": 0.6524251894344792,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.792967095043073e-05,
+      "loss": 0.075,
+      "mean_token_accuracy": 0.9754076436161995,
+      "num_tokens": 36505347.0,
+      "step": 1905
+    },
+    {
+      "entropy": 0.9948342055082321,
+      "epoch": 0.6541375915064858,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.7922609800875582e-05,
+      "loss": 0.062,
+      "mean_token_accuracy": 0.9795832648873329,
+      "num_tokens": 36601609.0,
+      "step": 1910
+    },
+    {
+      "entropy": 1.023300763964653,
+      "epoch": 0.6558499935784923,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.7915548651320434e-05,
+      "loss": 0.0782,
+      "mean_token_accuracy": 0.9718105450272561,
+      "num_tokens": 36696884.0,
+      "step": 1915
+    },
+    {
+      "entropy": 1.0569759473204612,
+      "epoch": 0.6575623956504988,
+      "grad_norm": 0.8359375,
+      "learning_rate": 1.7908487501765287e-05,
+      "loss": 0.0657,
+      "mean_token_accuracy": 0.9762900114059448,
+      "num_tokens": 36792437.0,
+      "step": 1920
+    },
+    {
+      "entropy": 1.0571024104952813,
+      "epoch": 0.6592747977225052,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.7901426352210142e-05,
+      "loss": 0.0758,
+      "mean_token_accuracy": 0.9737093850970269,
+      "num_tokens": 36887385.0,
+      "step": 1925
+    },
+    {
+      "entropy": 1.0637590497732163,
+      "epoch": 0.6609871997945117,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.7894365202654994e-05,
+      "loss": 0.0764,
+      "mean_token_accuracy": 0.9738867059350014,
+      "num_tokens": 36983013.0,
+      "step": 1930
+    },
+    {
+      "entropy": 1.0555319100618363,
+      "epoch": 0.6626996018665182,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.7887304053099847e-05,
+      "loss": 0.0679,
+      "mean_token_accuracy": 0.9779734149575233,
+      "num_tokens": 37080695.0,
+      "step": 1935
+    },
+    {
+      "entropy": 1.0645917877554893,
+      "epoch": 0.6644120039385247,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.78802429035447e-05,
+      "loss": 0.0654,
+      "mean_token_accuracy": 0.9791171714663506,
+      "num_tokens": 37177183.0,
+      "step": 1940
+    },
+    {
+      "entropy": 1.0873429045081138,
+      "epoch": 0.6661244060105312,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.787318175398955e-05,
+      "loss": 0.0799,
+      "mean_token_accuracy": 0.9705887973308563,
+      "num_tokens": 37273106.0,
+      "step": 1945
+    },
+    {
+      "entropy": 1.0863715693354608,
+      "epoch": 0.6678368080825378,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.7866120604434403e-05,
+      "loss": 0.0696,
+      "mean_token_accuracy": 0.9750098019838334,
+      "num_tokens": 37367702.0,
+      "step": 1950
+    },
+    {
+      "entropy": 1.0704414397478104,
+      "epoch": 0.6695492101545443,
+      "grad_norm": 0.91015625,
+      "learning_rate": 1.7859059454879255e-05,
+      "loss": 0.0628,
+      "mean_token_accuracy": 0.9776540219783783,
+      "num_tokens": 37463877.0,
+      "step": 1955
+    },
+    {
+      "entropy": 1.0778814405202866,
+      "epoch": 0.6712616122265508,
+      "grad_norm": 1.078125,
+      "learning_rate": 1.7851998305324107e-05,
+      "loss": 0.0668,
+      "mean_token_accuracy": 0.976428185403347,
+      "num_tokens": 37559306.0,
+      "step": 1960
+    },
+    {
+      "entropy": 1.0520638614892959,
+      "epoch": 0.6729740142985573,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.784493715576896e-05,
+      "loss": 0.0631,
+      "mean_token_accuracy": 0.9773733288049697,
+      "num_tokens": 37655853.0,
+      "step": 1965
+    },
+    {
+      "entropy": 1.0375919058918952,
+      "epoch": 0.6746864163705638,
+      "grad_norm": 0.8125,
+      "learning_rate": 1.7837876006213812e-05,
+      "loss": 0.0612,
+      "mean_token_accuracy": 0.978010119497776,
+      "num_tokens": 37752446.0,
+      "step": 1970
+    },
+    {
+      "entropy": 1.0213906228542329,
+      "epoch": 0.6763988184425703,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.7830814856658664e-05,
+      "loss": 0.0727,
+      "mean_token_accuracy": 0.9736576318740845,
+      "num_tokens": 37849320.0,
+      "step": 1975
+    },
+    {
+      "entropy": 1.0170167818665505,
+      "epoch": 0.6781112205145768,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.7823753707103516e-05,
+      "loss": 0.0764,
+      "mean_token_accuracy": 0.972841563820839,
+      "num_tokens": 37944960.0,
+      "step": 1980
+    },
+    {
+      "entropy": 1.011356957256794,
+      "epoch": 0.6798236225865834,
+      "grad_norm": 0.86328125,
+      "learning_rate": 1.781669255754837e-05,
+      "loss": 0.0685,
+      "mean_token_accuracy": 0.9748612508177757,
+      "num_tokens": 38039123.0,
+      "step": 1985
+    },
+    {
+      "entropy": 1.0182710096240044,
+      "epoch": 0.6815360246585899,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.780963140799322e-05,
+      "loss": 0.0765,
+      "mean_token_accuracy": 0.9738985195755958,
+      "num_tokens": 38134854.0,
+      "step": 1990
+    },
+    {
+      "entropy": 1.0277080595493318,
+      "epoch": 0.6832484267305964,
+      "grad_norm": 0.78125,
+      "learning_rate": 1.7802570258438076e-05,
+      "loss": 0.0817,
+      "mean_token_accuracy": 0.9715014040470124,
+      "num_tokens": 38232336.0,
+      "step": 1995
+    },
+    {
+      "entropy": 1.0434579849243164,
+      "epoch": 0.6849608288026029,
+      "grad_norm": 0.87890625,
+      "learning_rate": 1.7795509108882928e-05,
+      "loss": 0.0706,
+      "mean_token_accuracy": 0.9753948897123337,
+      "num_tokens": 38327467.0,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6849608288026029,
+      "eval_entropy": 1.0512260044041315,
+      "eval_loss": 0.07304668426513672,
+      "eval_mean_token_accuracy": 0.974862716717427,
+      "eval_num_tokens": 38327467.0,
+      "eval_runtime": 342.9674,
+      "eval_samples_per_second": 15.136,
+      "eval_steps_per_second": 15.136,
+      "step": 2000
+    },
+    {
+      "entropy": 1.052195043861866,
+      "epoch": 0.6866732308746094,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.778844795932778e-05,
+      "loss": 0.0682,
+      "mean_token_accuracy": 0.9773333162069321,
+      "num_tokens": 38421803.0,
+      "step": 2005
+    },
+    {
+      "entropy": 1.0468317985534668,
+      "epoch": 0.6883856329466158,
+      "grad_norm": 1.078125,
+      "learning_rate": 1.7781386809772633e-05,
+      "loss": 0.0763,
+      "mean_token_accuracy": 0.9746113270521164,
+      "num_tokens": 38517841.0,
+      "step": 2010
+    },
+    {
+      "entropy": 1.0659818336367608,
+      "epoch": 0.6900980350186223,
+      "grad_norm": 0.8671875,
+      "learning_rate": 1.7774325660217485e-05,
+      "loss": 0.0709,
+      "mean_token_accuracy": 0.9763822495937348,
+      "num_tokens": 38612565.0,
+      "step": 2015
+    },
+    {
+      "entropy": 1.0623872816562652,
+      "epoch": 0.6918104370906288,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.7767264510662337e-05,
+      "loss": 0.0851,
+      "mean_token_accuracy": 0.9718340992927551,
+      "num_tokens": 38707505.0,
+      "step": 2020
+    },
+    {
+      "entropy": 1.0433957114815713,
+      "epoch": 0.6935228391626354,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.776020336110719e-05,
+      "loss": 0.0774,
+      "mean_token_accuracy": 0.9720458060503006,
+      "num_tokens": 38805875.0,
+      "step": 2025
+    },
+    {
+      "entropy": 1.0362237706780433,
+      "epoch": 0.6952352412346419,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.775314221155204e-05,
+      "loss": 0.0618,
+      "mean_token_accuracy": 0.9778729319572449,
+      "num_tokens": 38902387.0,
+      "step": 2030
+    },
+    {
+      "entropy": 1.031554362177849,
+      "epoch": 0.6969476433066484,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.7746081061996893e-05,
+      "loss": 0.0703,
+      "mean_token_accuracy": 0.9737389236688614,
+      "num_tokens": 38997735.0,
+      "step": 2035
+    },
+    {
+      "entropy": 0.9809193640947342,
+      "epoch": 0.6986600453786549,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.7739019912441746e-05,
+      "loss": 0.0674,
+      "mean_token_accuracy": 0.9763004139065743,
+      "num_tokens": 39096421.0,
+      "step": 2040
+    },
+    {
+      "entropy": 0.9762739643454552,
+      "epoch": 0.7003724474506614,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.7731958762886598e-05,
+      "loss": 0.0809,
+      "mean_token_accuracy": 0.9694921225309372,
+      "num_tokens": 39190258.0,
+      "step": 2045
+    },
+    {
+      "entropy": 0.9988986641168595,
+      "epoch": 0.7020848495226679,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.772489761333145e-05,
+      "loss": 0.0751,
+      "mean_token_accuracy": 0.9740806385874748,
+      "num_tokens": 39286204.0,
+      "step": 2050
+    },
+    {
+      "entropy": 1.0018026500940322,
+      "epoch": 0.7037972515946744,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.7717836463776302e-05,
+      "loss": 0.0594,
+      "mean_token_accuracy": 0.9806592509150505,
+      "num_tokens": 39380963.0,
+      "step": 2055
+    },
+    {
+      "entropy": 1.0104161515831946,
+      "epoch": 0.705509653666681,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.7710775314221154e-05,
+      "loss": 0.06,
+      "mean_token_accuracy": 0.9781637147068978,
+      "num_tokens": 39476538.0,
+      "step": 2060
+    },
+    {
+      "entropy": 0.9990322396159173,
+      "epoch": 0.7072220557386875,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.770371416466601e-05,
+      "loss": 0.0748,
+      "mean_token_accuracy": 0.9751879677176476,
+      "num_tokens": 39572679.0,
+      "step": 2065
+    },
+    {
+      "entropy": 1.0114371925592422,
+      "epoch": 0.708934457810694,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.7696653015110862e-05,
+      "loss": 0.0718,
+      "mean_token_accuracy": 0.9758079364895821,
+      "num_tokens": 39669064.0,
+      "step": 2070
+    },
+    {
+      "entropy": 1.0140913486480714,
+      "epoch": 0.7106468598827005,
+      "grad_norm": 1.1953125,
+      "learning_rate": 1.7689591865555714e-05,
+      "loss": 0.0616,
+      "mean_token_accuracy": 0.9792987465858459,
+      "num_tokens": 39765572.0,
+      "step": 2075
+    },
+    {
+      "entropy": 1.0250459358096122,
+      "epoch": 0.712359261954707,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.7682530716000566e-05,
+      "loss": 0.0587,
+      "mean_token_accuracy": 0.9775450319051743,
+      "num_tokens": 39863159.0,
+      "step": 2080
+    },
+    {
+      "entropy": 1.0532150954008102,
+      "epoch": 0.7140716640267135,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.767546956644542e-05,
+      "loss": 0.083,
+      "mean_token_accuracy": 0.970898550748825,
+      "num_tokens": 39957804.0,
+      "step": 2085
+    },
+    {
+      "entropy": 1.0311722368001939,
+      "epoch": 0.71578406609872,
+      "grad_norm": 1.1953125,
+      "learning_rate": 1.766840841689027e-05,
+      "loss": 0.0492,
+      "mean_token_accuracy": 0.9832075580954551,
+      "num_tokens": 40052573.0,
+      "step": 2090
+    },
+    {
+      "entropy": 1.0157158136367799,
+      "epoch": 0.7174964681707264,
+      "grad_norm": 0.96875,
+      "learning_rate": 1.7661347267335123e-05,
+      "loss": 0.0678,
+      "mean_token_accuracy": 0.9764246463775634,
+      "num_tokens": 40148232.0,
+      "step": 2095
+    },
+    {
+      "entropy": 1.010478426516056,
+      "epoch": 0.719208870242733,
+      "grad_norm": 0.93359375,
+      "learning_rate": 1.7654286117779975e-05,
+      "loss": 0.0671,
+      "mean_token_accuracy": 0.9778443798422813,
+      "num_tokens": 40243563.0,
+      "step": 2100
+    },
+    {
+      "entropy": 1.002381108701229,
+      "epoch": 0.7209212723147395,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.7647224968224827e-05,
+      "loss": 0.0673,
+      "mean_token_accuracy": 0.9749844714999198,
+      "num_tokens": 40339635.0,
+      "step": 2105
+    },
+    {
+      "entropy": 1.0059390485286712,
+      "epoch": 0.722633674386746,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.764016381866968e-05,
+      "loss": 0.0709,
+      "mean_token_accuracy": 0.9754815220832824,
+      "num_tokens": 40436076.0,
+      "step": 2110
+    },
+    {
+      "entropy": 0.9956211879849434,
+      "epoch": 0.7243460764587525,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.763310266911453e-05,
+      "loss": 0.0541,
+      "mean_token_accuracy": 0.981034567952156,
+      "num_tokens": 40532212.0,
+      "step": 2115
+    },
+    {
+      "entropy": 1.000788713991642,
+      "epoch": 0.726058478530759,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.7626041519559384e-05,
+      "loss": 0.0741,
+      "mean_token_accuracy": 0.9746564134955407,
+      "num_tokens": 40626923.0,
+      "step": 2120
+    },
+    {
+      "entropy": 0.9876314714550972,
+      "epoch": 0.7277708806027655,
+      "grad_norm": 1.0625,
+      "learning_rate": 1.7618980370004236e-05,
+      "loss": 0.0716,
+      "mean_token_accuracy": 0.974675789475441,
+      "num_tokens": 40722779.0,
+      "step": 2125
+    },
+    {
+      "entropy": 0.9961372017860413,
+      "epoch": 0.729483282674772,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.7611919220449088e-05,
+      "loss": 0.0765,
+      "mean_token_accuracy": 0.9724399492144584,
+      "num_tokens": 40819002.0,
+      "step": 2130
+    },
+    {
+      "entropy": 1.002322755753994,
+      "epoch": 0.7311956847467785,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.7604858070893944e-05,
+      "loss": 0.0679,
+      "mean_token_accuracy": 0.977602270245552,
+      "num_tokens": 40915494.0,
+      "step": 2135
+    },
+    {
+      "entropy": 1.0378064289689064,
+      "epoch": 0.732908086818785,
+      "grad_norm": 0.6171875,
+      "learning_rate": 1.7597796921338796e-05,
+      "loss": 0.08,
+      "mean_token_accuracy": 0.9735590949654579,
+      "num_tokens": 41011681.0,
+      "step": 2140
+    },
+    {
+      "entropy": 1.0606622621417046,
+      "epoch": 0.7346204888907916,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.7590735771783648e-05,
+      "loss": 0.0764,
+      "mean_token_accuracy": 0.9734551355242729,
+      "num_tokens": 41107429.0,
+      "step": 2145
+    },
+    {
+      "entropy": 1.0482859820127488,
+      "epoch": 0.7363328909627981,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.75836746222285e-05,
+      "loss": 0.0633,
+      "mean_token_accuracy": 0.9766931667923927,
+      "num_tokens": 41202802.0,
+      "step": 2150
+    },
+    {
+      "entropy": 1.044474896788597,
+      "epoch": 0.7380452930348046,
+      "grad_norm": 1.3125,
+      "learning_rate": 1.7576613472673352e-05,
+      "loss": 0.0775,
+      "mean_token_accuracy": 0.9722099617123604,
+      "num_tokens": 41299346.0,
+      "step": 2155
+    },
+    {
+      "entropy": 1.0398955419659615,
+      "epoch": 0.7397576951068111,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.7569552323118205e-05,
+      "loss": 0.0702,
+      "mean_token_accuracy": 0.9754580482840538,
+      "num_tokens": 41394294.0,
+      "step": 2160
+    },
+    {
+      "entropy": 1.027321819961071,
+      "epoch": 0.7414700971788176,
+      "grad_norm": 0.8828125,
+      "learning_rate": 1.7562491173563057e-05,
+      "loss": 0.066,
+      "mean_token_accuracy": 0.9773415267467499,
+      "num_tokens": 41491754.0,
+      "step": 2165
+    },
+    {
+      "entropy": 1.0326685786247254,
+      "epoch": 0.7431824992508241,
+      "grad_norm": 0.828125,
+      "learning_rate": 1.755543002400791e-05,
+      "loss": 0.059,
+      "mean_token_accuracy": 0.978815944492817,
+      "num_tokens": 41585799.0,
+      "step": 2170
+    },
+    {
+      "entropy": 1.0269860818982124,
+      "epoch": 0.7448949013228306,
+      "grad_norm": 0.97265625,
+      "learning_rate": 1.754836887445276e-05,
+      "loss": 0.0815,
+      "mean_token_accuracy": 0.9711510434746742,
+      "num_tokens": 41681350.0,
+      "step": 2175
+    },
+    {
+      "entropy": 1.008481813967228,
+      "epoch": 0.7466073033948372,
+      "grad_norm": 0.734375,
+      "learning_rate": 1.7541307724897613e-05,
+      "loss": 0.0499,
+      "mean_token_accuracy": 0.9836753293871879,
+      "num_tokens": 41779208.0,
+      "step": 2180
+    },
+    {
+      "entropy": 1.0394549012184142,
+      "epoch": 0.7483197054668436,
+      "grad_norm": 1.5234375,
+      "learning_rate": 1.7534246575342465e-05,
+      "loss": 0.0751,
+      "mean_token_accuracy": 0.974792493879795,
+      "num_tokens": 41875544.0,
+      "step": 2185
+    },
+    {
+      "entropy": 1.0370924472808838,
+      "epoch": 0.7500321075388501,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.7527185425787318e-05,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.9779522895812989,
+      "num_tokens": 41972127.0,
+      "step": 2190
+    },
+    {
+      "entropy": 1.0175135120749474,
+      "epoch": 0.7517445096108566,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.752012427623217e-05,
+      "loss": 0.0647,
+      "mean_token_accuracy": 0.9783763736486435,
+      "num_tokens": 42068575.0,
+      "step": 2195
+    },
+    {
+      "entropy": 1.027130389213562,
+      "epoch": 0.7534569116828631,
+      "grad_norm": 0.953125,
+      "learning_rate": 1.7513063126677022e-05,
+      "loss": 0.0589,
+      "mean_token_accuracy": 0.9790372952818871,
+      "num_tokens": 42165170.0,
+      "step": 2200
+    },
+    {
+      "epoch": 0.7534569116828631,
+      "eval_entropy": 1.0479857029195605,
+      "eval_loss": 0.07274039834737778,
+      "eval_mean_token_accuracy": 0.9750881667105914,
+      "eval_num_tokens": 42165170.0,
+      "eval_runtime": 342.0986,
+      "eval_samples_per_second": 15.174,
+      "eval_steps_per_second": 15.174,
+      "step": 2200
+    },
+    {
+      "entropy": 1.049907286465168,
+      "epoch": 0.7551693137548696,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.7506001977121878e-05,
+      "loss": 0.0777,
+      "mean_token_accuracy": 0.9732275575399398,
+      "num_tokens": 42260030.0,
+      "step": 2205
+    },
+    {
+      "entropy": 1.068640261888504,
+      "epoch": 0.7568817158268761,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.749894082756673e-05,
+      "loss": 0.0788,
+      "mean_token_accuracy": 0.9729743257164956,
+      "num_tokens": 42356295.0,
+      "step": 2210
+    },
+    {
+      "entropy": 1.0697579666972161,
+      "epoch": 0.7585941178988826,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.7491879678011582e-05,
+      "loss": 0.0862,
+      "mean_token_accuracy": 0.970036880671978,
+      "num_tokens": 42452124.0,
+      "step": 2215
+    },
+    {
+      "entropy": 1.0525717750191688,
+      "epoch": 0.7603065199708892,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.7484818528456434e-05,
+      "loss": 0.0735,
+      "mean_token_accuracy": 0.9742584452033043,
+      "num_tokens": 42548362.0,
+      "step": 2220
+    },
+    {
+      "entropy": 1.0521694049239159,
+      "epoch": 0.7620189220428957,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.7477757378901286e-05,
+      "loss": 0.0727,
+      "mean_token_accuracy": 0.9752233937382698,
+      "num_tokens": 42643227.0,
+      "step": 2225
+    },
+    {
+      "entropy": 1.0245699137449265,
+      "epoch": 0.7637313241149022,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.747069622934614e-05,
+      "loss": 0.0673,
+      "mean_token_accuracy": 0.9771186649799347,
+      "num_tokens": 42740731.0,
+      "step": 2230
+    },
+    {
+      "entropy": 1.0340376988053321,
+      "epoch": 0.7654437261869087,
+      "grad_norm": 1.2890625,
+      "learning_rate": 1.746363507979099e-05,
+      "loss": 0.0772,
+      "mean_token_accuracy": 0.9736419811844825,
+      "num_tokens": 42835316.0,
+      "step": 2235
+    },
+    {
+      "entropy": 1.0107177391648292,
+      "epoch": 0.7671561282589152,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.7456573930235843e-05,
+      "loss": 0.0665,
+      "mean_token_accuracy": 0.9768707409501076,
+      "num_tokens": 42931658.0,
+      "step": 2240
+    },
+    {
+      "entropy": 1.0158133059740067,
+      "epoch": 0.7688685303309217,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.7449512780680695e-05,
+      "loss": 0.0636,
+      "mean_token_accuracy": 0.9763818144798279,
+      "num_tokens": 43030400.0,
+      "step": 2245
+    },
+    {
+      "entropy": 1.0594315737485887,
+      "epoch": 0.7705809324029282,
+      "grad_norm": 1.328125,
+      "learning_rate": 1.7442451631125547e-05,
+      "loss": 0.0893,
+      "mean_token_accuracy": 0.9690458342432976,
+      "num_tokens": 43126487.0,
+      "step": 2250
+    },
+    {
+      "entropy": 1.0822157844901086,
+      "epoch": 0.7722933344749348,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.74353904815704e-05,
+      "loss": 0.0843,
+      "mean_token_accuracy": 0.97084331959486,
+      "num_tokens": 43221995.0,
+      "step": 2255
+    },
+    {
+      "entropy": 1.108020193874836,
+      "epoch": 0.7740057365469413,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.742832933201525e-05,
+      "loss": 0.0813,
+      "mean_token_accuracy": 0.9706909582018852,
+      "num_tokens": 43316125.0,
+      "step": 2260
+    },
+    {
+      "entropy": 1.0929044395685197,
+      "epoch": 0.7757181386189478,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.7421268182460104e-05,
+      "loss": 0.0746,
+      "mean_token_accuracy": 0.9729498103260994,
+      "num_tokens": 43410800.0,
+      "step": 2265
+    },
+    {
+      "entropy": 1.0594649076461793,
+      "epoch": 0.7774305406909542,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.7414207032904956e-05,
+      "loss": 0.0703,
+      "mean_token_accuracy": 0.9744070082902908,
+      "num_tokens": 43506610.0,
+      "step": 2270
+    },
+    {
+      "entropy": 1.0071976557374,
+      "epoch": 0.7791429427629607,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.740714588334981e-05,
+      "loss": 0.0659,
+      "mean_token_accuracy": 0.9776173815131187,
+      "num_tokens": 43605142.0,
+      "step": 2275
+    },
+    {
+      "entropy": 1.01753339022398,
+      "epoch": 0.7808553448349672,
+      "grad_norm": 1.1015625,
+      "learning_rate": 1.7400084733794664e-05,
+      "loss": 0.066,
+      "mean_token_accuracy": 0.9779619738459587,
+      "num_tokens": 43700194.0,
+      "step": 2280
+    },
+    {
+      "entropy": 1.0066962376236916,
+      "epoch": 0.7825677469069737,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.7393023584239516e-05,
+      "loss": 0.0672,
+      "mean_token_accuracy": 0.9750949293375015,
+      "num_tokens": 43796959.0,
+      "step": 2285
+    },
+    {
+      "entropy": 1.0055259376764298,
+      "epoch": 0.7842801489789802,
+      "grad_norm": 0.9765625,
+      "learning_rate": 1.7385962434684368e-05,
+      "loss": 0.0733,
+      "mean_token_accuracy": 0.9744447767734528,
+      "num_tokens": 43893523.0,
+      "step": 2290
+    },
+    {
+      "entropy": 0.9918966501951217,
+      "epoch": 0.7859925510509868,
+      "grad_norm": 0.92578125,
+      "learning_rate": 1.737890128512922e-05,
+      "loss": 0.054,
+      "mean_token_accuracy": 0.982974311709404,
+      "num_tokens": 43988874.0,
+      "step": 2295
+    },
+    {
+      "entropy": 1.018340316414833,
+      "epoch": 0.7877049531229933,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.7371840135574072e-05,
+      "loss": 0.0825,
+      "mean_token_accuracy": 0.9712464198470115,
+      "num_tokens": 44083420.0,
+      "step": 2300
+    },
+    {
+      "entropy": 1.014200821518898,
+      "epoch": 0.7894173551949998,
+      "grad_norm": 1.25,
+      "learning_rate": 1.7364778986018924e-05,
+      "loss": 0.0665,
+      "mean_token_accuracy": 0.975614894926548,
+      "num_tokens": 44179106.0,
+      "step": 2305
+    },
+    {
+      "entropy": 1.0206679552793503,
+      "epoch": 0.7911297572670063,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.7357717836463777e-05,
+      "loss": 0.072,
+      "mean_token_accuracy": 0.9756124913692474,
+      "num_tokens": 44275761.0,
+      "step": 2310
+    },
+    {
+      "entropy": 1.029144886136055,
+      "epoch": 0.7928421593390128,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.7350656686908632e-05,
+      "loss": 0.0693,
+      "mean_token_accuracy": 0.977510643005371,
+      "num_tokens": 44372693.0,
+      "step": 2315
+    },
+    {
+      "entropy": 1.0367668330669404,
+      "epoch": 0.7945545614110193,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.7343595537353484e-05,
+      "loss": 0.0698,
+      "mean_token_accuracy": 0.9765833675861358,
+      "num_tokens": 44468586.0,
+      "step": 2320
+    },
+    {
+      "entropy": 1.0696975022554398,
+      "epoch": 0.7962669634830258,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.7336534387798336e-05,
+      "loss": 0.0721,
+      "mean_token_accuracy": 0.9733602672815322,
+      "num_tokens": 44563560.0,
+      "step": 2325
+    },
+    {
+      "entropy": 1.0519574165344239,
+      "epoch": 0.7979793655550323,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.732947323824319e-05,
+      "loss": 0.072,
+      "mean_token_accuracy": 0.975217056274414,
+      "num_tokens": 44658506.0,
+      "step": 2330
+    },
+    {
+      "entropy": 0.9864769175648689,
+      "epoch": 0.7996917676270389,
+      "grad_norm": 0.90234375,
+      "learning_rate": 1.732241208868804e-05,
+      "loss": 0.0541,
+      "mean_token_accuracy": 0.9797987848520279,
+      "num_tokens": 44756473.0,
+      "step": 2335
+    },
+    {
+      "entropy": 0.9787307888269424,
+      "epoch": 0.8014041696990454,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.7315350939132893e-05,
+      "loss": 0.066,
+      "mean_token_accuracy": 0.9785678714513779,
+      "num_tokens": 44853971.0,
+      "step": 2340
+    },
+    {
+      "entropy": 0.9830088958144187,
+      "epoch": 0.8031165717710519,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.7308289789577745e-05,
+      "loss": 0.086,
+      "mean_token_accuracy": 0.97075584679842,
+      "num_tokens": 44951011.0,
+      "step": 2345
+    },
+    {
+      "entropy": 1.0064200356602668,
+      "epoch": 0.8048289738430584,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.7301228640022597e-05,
+      "loss": 0.0731,
+      "mean_token_accuracy": 0.972475990653038,
+      "num_tokens": 45046605.0,
+      "step": 2350
+    },
+    {
+      "entropy": 1.0114048093557357,
+      "epoch": 0.8065413759150648,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.729416749046745e-05,
+      "loss": 0.0761,
+      "mean_token_accuracy": 0.971944771707058,
+      "num_tokens": 45142842.0,
+      "step": 2355
+    },
+    {
+      "entropy": 1.0145201429724693,
+      "epoch": 0.8082537779870713,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.72871063409123e-05,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9744179874658585,
+      "num_tokens": 45237455.0,
+      "step": 2360
+    },
+    {
+      "entropy": 0.9942709267139435,
+      "epoch": 0.8099661800590778,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.7280045191357154e-05,
+      "loss": 0.0645,
+      "mean_token_accuracy": 0.9777603447437286,
+      "num_tokens": 45333811.0,
+      "step": 2365
+    },
+    {
+      "entropy": 1.0171207472682,
+      "epoch": 0.8116785821310843,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.727298404180201e-05,
+      "loss": 0.0642,
+      "mean_token_accuracy": 0.9767574414610862,
+      "num_tokens": 45428647.0,
+      "step": 2370
+    },
+    {
+      "entropy": 1.0124354764819146,
+      "epoch": 0.8133909842030909,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.726592289224686e-05,
+      "loss": 0.0777,
+      "mean_token_accuracy": 0.9743141904473305,
+      "num_tokens": 45525017.0,
+      "step": 2375
+    },
+    {
+      "entropy": 1.030722188949585,
+      "epoch": 0.8151033862750974,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.7258861742691714e-05,
+      "loss": 0.0797,
+      "mean_token_accuracy": 0.9719723522663116,
+      "num_tokens": 45620247.0,
+      "step": 2380
+    },
+    {
+      "entropy": 1.0066359415650368,
+      "epoch": 0.8168157883471039,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.7251800593136566e-05,
+      "loss": 0.0524,
+      "mean_token_accuracy": 0.9814678773283958,
+      "num_tokens": 45716028.0,
+      "step": 2385
+    },
+    {
+      "entropy": 1.0095541879534722,
+      "epoch": 0.8185281904191104,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.7244739443581418e-05,
+      "loss": 0.0763,
+      "mean_token_accuracy": 0.9737412840127945,
+      "num_tokens": 45811731.0,
+      "step": 2390
+    },
+    {
+      "entropy": 1.0151244834065438,
+      "epoch": 0.8202405924911169,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.723767829402627e-05,
+      "loss": 0.081,
+      "mean_token_accuracy": 0.9724341601133346,
+      "num_tokens": 45905973.0,
+      "step": 2395
+    },
+    {
+      "entropy": 1.0098653748631476,
+      "epoch": 0.8219529945631234,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.7230617144471122e-05,
+      "loss": 0.0708,
+      "mean_token_accuracy": 0.9755938917398452,
+      "num_tokens": 46001551.0,
+      "step": 2400
+    },
+    {
+      "epoch": 0.8219529945631234,
+      "eval_entropy": 1.0198914397775587,
+      "eval_loss": 0.07129265367984772,
+      "eval_mean_token_accuracy": 0.9753302702943246,
+      "eval_num_tokens": 46001551.0,
+      "eval_runtime": 343.1526,
+      "eval_samples_per_second": 15.127,
+      "eval_steps_per_second": 15.127,
+      "step": 2400
+    },
+    {
+      "entropy": 1.02613263130188,
+      "epoch": 0.8236653966351299,
+      "grad_norm": 0.9296875,
+      "learning_rate": 1.7223555994915975e-05,
+      "loss": 0.0716,
+      "mean_token_accuracy": 0.9752387389540672,
+      "num_tokens": 46097300.0,
+      "step": 2405
+    },
+    {
+      "entropy": 1.0430584996938705,
+      "epoch": 0.8253777987071365,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.7216494845360827e-05,
+      "loss": 0.0867,
+      "mean_token_accuracy": 0.9701804518699646,
+      "num_tokens": 46194057.0,
+      "step": 2410
+    },
+    {
+      "entropy": 1.0283590614795686,
+      "epoch": 0.827090200779143,
+      "grad_norm": 0.9375,
+      "learning_rate": 1.720943369580568e-05,
+      "loss": 0.0619,
+      "mean_token_accuracy": 0.9786317780613899,
+      "num_tokens": 46290426.0,
+      "step": 2415
+    },
+    {
+      "entropy": 1.0015892446041108,
+      "epoch": 0.8288026028511495,
+      "grad_norm": 1.0,
+      "learning_rate": 1.720237254625053e-05,
+      "loss": 0.0715,
+      "mean_token_accuracy": 0.9746826022863389,
+      "num_tokens": 46386426.0,
+      "step": 2420
+    },
+    {
+      "entropy": 1.0031271770596504,
+      "epoch": 0.830515004923156,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.7195311396695383e-05,
+      "loss": 0.0595,
+      "mean_token_accuracy": 0.9782519474625587,
+      "num_tokens": 46482184.0,
+      "step": 2425
+    },
+    {
+      "entropy": 1.010385772585869,
+      "epoch": 0.8322274069951625,
+      "grad_norm": 1.0,
+      "learning_rate": 1.7188250247140235e-05,
+      "loss": 0.0689,
+      "mean_token_accuracy": 0.9767611041665077,
+      "num_tokens": 46579032.0,
+      "step": 2430
+    },
+    {
+      "entropy": 1.0388280734419824,
+      "epoch": 0.833939809067169,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.7181189097585088e-05,
+      "loss": 0.0695,
+      "mean_token_accuracy": 0.9758057102560997,
+      "num_tokens": 46673505.0,
+      "step": 2435
+    },
+    {
+      "entropy": 1.0433961763978004,
+      "epoch": 0.8356522111391754,
+      "grad_norm": 1.1015625,
+      "learning_rate": 1.7174127948029943e-05,
+      "loss": 0.0762,
+      "mean_token_accuracy": 0.9739331617951393,
+      "num_tokens": 46770016.0,
+      "step": 2440
+    },
+    {
+      "entropy": 1.0399362683296203,
+      "epoch": 0.8373646132111819,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.7167066798474795e-05,
+      "loss": 0.0785,
+      "mean_token_accuracy": 0.9725722745060921,
+      "num_tokens": 46866987.0,
+      "step": 2445
+    },
+    {
+      "entropy": 1.0495048731565475,
+      "epoch": 0.8390770152831885,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.7160005648919648e-05,
+      "loss": 0.0708,
+      "mean_token_accuracy": 0.9755064710974694,
+      "num_tokens": 46964689.0,
+      "step": 2450
+    },
+    {
+      "entropy": 1.041603434085846,
+      "epoch": 0.840789417355195,
+      "grad_norm": 1.078125,
+      "learning_rate": 1.71529444993645e-05,
+      "loss": 0.0821,
+      "mean_token_accuracy": 0.9712515100836754,
+      "num_tokens": 47061389.0,
+      "step": 2455
+    },
+    {
+      "entropy": 1.0497812896966934,
+      "epoch": 0.8425018194272015,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.7145883349809352e-05,
+      "loss": 0.083,
+      "mean_token_accuracy": 0.9728558778762817,
+      "num_tokens": 47155893.0,
+      "step": 2460
+    },
+    {
+      "entropy": 1.0246960267424583,
+      "epoch": 0.844214221499208,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.7138822200254204e-05,
+      "loss": 0.0761,
+      "mean_token_accuracy": 0.9728069946169853,
+      "num_tokens": 47252784.0,
+      "step": 2465
+    },
+    {
+      "entropy": 1.0220705047249794,
+      "epoch": 0.8459266235712145,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.7131761050699056e-05,
+      "loss": 0.0719,
+      "mean_token_accuracy": 0.9752938598394394,
+      "num_tokens": 47348747.0,
+      "step": 2470
+    },
+    {
+      "entropy": 1.0130750119686127,
+      "epoch": 0.847639025643221,
+      "grad_norm": 1.1328125,
+      "learning_rate": 1.712469990114391e-05,
+      "loss": 0.0546,
+      "mean_token_accuracy": 0.9807667568325996,
+      "num_tokens": 47445858.0,
+      "step": 2475
+    },
+    {
+      "entropy": 1.017496681213379,
+      "epoch": 0.8493514277152275,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.711763875158876e-05,
+      "loss": 0.0808,
+      "mean_token_accuracy": 0.9720327258110046,
+      "num_tokens": 47541005.0,
+      "step": 2480
+    },
+    {
+      "entropy": 1.0123622953891753,
+      "epoch": 0.851063829787234,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.7110577602033613e-05,
+      "loss": 0.0647,
+      "mean_token_accuracy": 0.9786107212305069,
+      "num_tokens": 47638217.0,
+      "step": 2485
+    },
+    {
+      "entropy": 0.9845497235655785,
+      "epoch": 0.8527762318592406,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.7103516452478465e-05,
+      "loss": 0.0642,
+      "mean_token_accuracy": 0.9764533504843712,
+      "num_tokens": 47734699.0,
+      "step": 2490
+    },
+    {
+      "entropy": 0.9638260096311569,
+      "epoch": 0.8544886339312471,
+      "grad_norm": 0.99609375,
+      "learning_rate": 1.7096455302923317e-05,
+      "loss": 0.0582,
+      "mean_token_accuracy": 0.9796235024929046,
+      "num_tokens": 47831330.0,
+      "step": 2495
+    },
+    {
+      "entropy": 0.9459871113300323,
+      "epoch": 0.8562010360032536,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.708939415336817e-05,
+      "loss": 0.05,
+      "mean_token_accuracy": 0.9809640198945999,
+      "num_tokens": 47927988.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.9511372432112694,
+      "epoch": 0.8579134380752601,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.708233300381302e-05,
+      "loss": 0.0734,
+      "mean_token_accuracy": 0.9751604914665222,
+      "num_tokens": 48025492.0,
+      "step": 2505
+    },
+    {
+      "entropy": 0.9695461690425873,
+      "epoch": 0.8596258401472666,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.7075271854257877e-05,
+      "loss": 0.082,
+      "mean_token_accuracy": 0.9733020886778831,
+      "num_tokens": 48121097.0,
+      "step": 2510
+    },
+    {
+      "entropy": 0.9868211299180984,
+      "epoch": 0.8613382422192731,
+      "grad_norm": 1.0546875,
+      "learning_rate": 1.706821070470273e-05,
+      "loss": 0.0781,
+      "mean_token_accuracy": 0.9722427085042,
+      "num_tokens": 48216606.0,
+      "step": 2515
+    },
+    {
+      "entropy": 1.0152612164616586,
+      "epoch": 0.8630506442912796,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.706114955514758e-05,
+      "loss": 0.0636,
+      "mean_token_accuracy": 0.9768360331654549,
+      "num_tokens": 48311035.0,
+      "step": 2520
+    },
+    {
+      "entropy": 1.020312374830246,
+      "epoch": 0.864763046363286,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.7054088405592434e-05,
+      "loss": 0.0725,
+      "mean_token_accuracy": 0.9769116640090942,
+      "num_tokens": 48407026.0,
+      "step": 2525
+    },
+    {
+      "entropy": 1.0195435583591461,
+      "epoch": 0.8664754484352926,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.7047027256037286e-05,
+      "loss": 0.065,
+      "mean_token_accuracy": 0.9760793089866638,
+      "num_tokens": 48502604.0,
+      "step": 2530
+    },
+    {
+      "entropy": 1.025297474861145,
+      "epoch": 0.8681878505072991,
+      "grad_norm": 0.921875,
+      "learning_rate": 1.7039966106482138e-05,
+      "loss": 0.078,
+      "mean_token_accuracy": 0.9716022863984108,
+      "num_tokens": 48599138.0,
+      "step": 2535
+    },
+    {
+      "entropy": 1.032269251346588,
+      "epoch": 0.8699002525793056,
+      "grad_norm": 1.296875,
+      "learning_rate": 1.703290495692699e-05,
+      "loss": 0.0735,
+      "mean_token_accuracy": 0.9734241843223572,
+      "num_tokens": 48693107.0,
+      "step": 2540
+    },
+    {
+      "entropy": 0.9952888503670693,
+      "epoch": 0.8716126546513121,
+      "grad_norm": 0.74609375,
+      "learning_rate": 1.7025843807371842e-05,
+      "loss": 0.0553,
+      "mean_token_accuracy": 0.9810656040906907,
+      "num_tokens": 48792074.0,
+      "step": 2545
+    },
+    {
+      "entropy": 1.0151646718382836,
+      "epoch": 0.8733250567233186,
+      "grad_norm": 1.0625,
+      "learning_rate": 1.7018782657816694e-05,
+      "loss": 0.073,
+      "mean_token_accuracy": 0.9738075092434884,
+      "num_tokens": 48886925.0,
+      "step": 2550
+    },
+    {
+      "entropy": 1.007457859814167,
+      "epoch": 0.8750374587953251,
+      "grad_norm": 1.203125,
+      "learning_rate": 1.7011721508261547e-05,
+      "loss": 0.0789,
+      "mean_token_accuracy": 0.9715763330459595,
+      "num_tokens": 48979664.0,
+      "step": 2555
+    },
+    {
+      "entropy": 0.9924128890037537,
+      "epoch": 0.8767498608673316,
+      "grad_norm": 0.859375,
+      "learning_rate": 1.70046603587064e-05,
+      "loss": 0.0638,
+      "mean_token_accuracy": 0.9770465239882469,
+      "num_tokens": 49076087.0,
+      "step": 2560
+    },
+    {
+      "entropy": 1.00923193693161,
+      "epoch": 0.8784622629393382,
+      "grad_norm": 1.3125,
+      "learning_rate": 1.699759920915125e-05,
+      "loss": 0.0626,
+      "mean_token_accuracy": 0.9779347315430641,
+      "num_tokens": 49171087.0,
+      "step": 2565
+    },
+    {
+      "entropy": 1.0142344176769256,
+      "epoch": 0.8801746650113447,
+      "grad_norm": 0.921875,
+      "learning_rate": 1.6990538059596103e-05,
+      "loss": 0.0768,
+      "mean_token_accuracy": 0.9728077635169029,
+      "num_tokens": 49267306.0,
+      "step": 2570
+    },
+    {
+      "entropy": 1.0244693771004676,
+      "epoch": 0.8818870670833512,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.6983476910040955e-05,
+      "loss": 0.0626,
+      "mean_token_accuracy": 0.9775442227721214,
+      "num_tokens": 49362515.0,
+      "step": 2575
+    },
+    {
+      "entropy": 1.001212653517723,
+      "epoch": 0.8835994691553577,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.697641576048581e-05,
+      "loss": 0.0716,
+      "mean_token_accuracy": 0.9777348548173904,
+      "num_tokens": 49459724.0,
+      "step": 2580
+    },
+    {
+      "entropy": 1.0120882764458656,
+      "epoch": 0.8853118712273642,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.6969354610930663e-05,
+      "loss": 0.0628,
+      "mean_token_accuracy": 0.9770969331264496,
+      "num_tokens": 49555001.0,
+      "step": 2585
+    },
+    {
+      "entropy": 1.0079264789819717,
+      "epoch": 0.8870242732993707,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.6962293461375515e-05,
+      "loss": 0.0731,
+      "mean_token_accuracy": 0.9721629783511162,
+      "num_tokens": 49650849.0,
+      "step": 2590
+    },
+    {
+      "entropy": 1.0020647004246712,
+      "epoch": 0.8887366753713772,
+      "grad_norm": 0.90625,
+      "learning_rate": 1.6955232311820367e-05,
+      "loss": 0.0662,
+      "mean_token_accuracy": 0.9765786185860634,
+      "num_tokens": 49745808.0,
+      "step": 2595
+    },
+    {
+      "entropy": 1.0287819892168044,
+      "epoch": 0.8904490774433838,
+      "grad_norm": 0.953125,
+      "learning_rate": 1.694817116226522e-05,
+      "loss": 0.0771,
+      "mean_token_accuracy": 0.9737220153212547,
+      "num_tokens": 49839490.0,
+      "step": 2600
+    },
+    {
+      "epoch": 0.8904490774433838,
+      "eval_entropy": 1.0099742332535473,
+      "eval_loss": 0.07065261900424957,
+      "eval_mean_token_accuracy": 0.9756279491638774,
+      "eval_num_tokens": 49839490.0,
+      "eval_runtime": 342.8046,
+      "eval_samples_per_second": 15.143,
+      "eval_steps_per_second": 15.143,
+      "step": 2600
+    },
+    {
+      "entropy": 1.0079515293240546,
+      "epoch": 0.8921614795153903,
+      "grad_norm": 0.8515625,
+      "learning_rate": 1.6941110012710072e-05,
+      "loss": 0.0661,
+      "mean_token_accuracy": 0.9761309459805488,
+      "num_tokens": 49936257.0,
+      "step": 2605
+    },
+    {
+      "entropy": 0.9924996554851532,
+      "epoch": 0.8938738815873967,
+      "grad_norm": 1.125,
+      "learning_rate": 1.6934048863154924e-05,
+      "loss": 0.0645,
+      "mean_token_accuracy": 0.9769900470972062,
+      "num_tokens": 50033249.0,
+      "step": 2610
+    },
+    {
+      "entropy": 1.005980095267296,
+      "epoch": 0.8955862836594032,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.6926987713599776e-05,
+      "loss": 0.0774,
+      "mean_token_accuracy": 0.9721055164933204,
+      "num_tokens": 50128073.0,
+      "step": 2615
+    },
+    {
+      "entropy": 0.9839522838592529,
+      "epoch": 0.8972986857314097,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.6919926564044628e-05,
+      "loss": 0.0596,
+      "mean_token_accuracy": 0.9800780564546585,
+      "num_tokens": 50224373.0,
+      "step": 2620
+    },
+    {
+      "entropy": 0.9890779614448547,
+      "epoch": 0.8990110878034162,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.691286541448948e-05,
+      "loss": 0.0857,
+      "mean_token_accuracy": 0.9688116520643234,
+      "num_tokens": 50321477.0,
+      "step": 2625
+    },
+    {
+      "entropy": 0.9660723656415939,
+      "epoch": 0.9007234898754227,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.6905804264934333e-05,
+      "loss": 0.0556,
+      "mean_token_accuracy": 0.9800270527601243,
+      "num_tokens": 50416898.0,
+      "step": 2630
+    },
+    {
+      "entropy": 0.9696930199861526,
+      "epoch": 0.9024358919474292,
+      "grad_norm": 1.125,
+      "learning_rate": 1.6898743115379185e-05,
+      "loss": 0.0902,
+      "mean_token_accuracy": 0.9687868222594261,
+      "num_tokens": 50512924.0,
+      "step": 2635
+    },
+    {
+      "entropy": 0.9480961576104164,
+      "epoch": 0.9041482940194357,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.6891681965824037e-05,
+      "loss": 0.0621,
+      "mean_token_accuracy": 0.9769589707255364,
+      "num_tokens": 50609393.0,
+      "step": 2640
+    },
+    {
+      "entropy": 0.9467855021357536,
+      "epoch": 0.9058606960914423,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.688462081626889e-05,
+      "loss": 0.0697,
+      "mean_token_accuracy": 0.9777101933956146,
+      "num_tokens": 50704611.0,
+      "step": 2645
+    },
+    {
+      "entropy": 0.963311468064785,
+      "epoch": 0.9075730981634488,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.6877559666713745e-05,
+      "loss": 0.0728,
+      "mean_token_accuracy": 0.9736167207360268,
+      "num_tokens": 50799331.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.9705994367599488,
+      "epoch": 0.9092855002354553,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.6870498517158597e-05,
+      "loss": 0.0696,
+      "mean_token_accuracy": 0.9770171776413917,
+      "num_tokens": 50896320.0,
+      "step": 2655
+    },
+    {
+      "entropy": 1.0018626376986504,
+      "epoch": 0.9109979023074618,
+      "grad_norm": 1.3359375,
+      "learning_rate": 1.686343736760345e-05,
+      "loss": 0.0682,
+      "mean_token_accuracy": 0.9750515669584274,
+      "num_tokens": 50991083.0,
+      "step": 2660
+    },
+    {
+      "entropy": 0.9972934260964393,
+      "epoch": 0.9127103043794683,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.68563762180483e-05,
+      "loss": 0.0816,
+      "mean_token_accuracy": 0.9727978155016899,
+      "num_tokens": 51086938.0,
+      "step": 2665
+    },
+    {
+      "entropy": 0.9934064865112304,
+      "epoch": 0.9144227064514748,
+      "grad_norm": 0.8046875,
+      "learning_rate": 1.6849315068493153e-05,
+      "loss": 0.048,
+      "mean_token_accuracy": 0.983874998986721,
+      "num_tokens": 51184939.0,
+      "step": 2670
+    },
+    {
+      "entropy": 1.0016051083803177,
+      "epoch": 0.9161351085234813,
+      "grad_norm": 1.125,
+      "learning_rate": 1.6842253918938006e-05,
+      "loss": 0.0705,
+      "mean_token_accuracy": 0.9754848092794418,
+      "num_tokens": 51280919.0,
+      "step": 2675
+    },
+    {
+      "entropy": 0.9878206342458725,
+      "epoch": 0.9178475105954879,
+      "grad_norm": 1.1015625,
+      "learning_rate": 1.6835192769382858e-05,
+      "loss": 0.0687,
+      "mean_token_accuracy": 0.976571673154831,
+      "num_tokens": 51375535.0,
+      "step": 2680
+    },
+    {
+      "entropy": 0.975702029466629,
+      "epoch": 0.9195599126674944,
+      "grad_norm": 0.91796875,
+      "learning_rate": 1.682813161982771e-05,
+      "loss": 0.0701,
+      "mean_token_accuracy": 0.9743976220488548,
+      "num_tokens": 51471696.0,
+      "step": 2685
+    },
+    {
+      "entropy": 0.9879275470972061,
+      "epoch": 0.9212723147395009,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.6821070470272562e-05,
+      "loss": 0.0765,
+      "mean_token_accuracy": 0.9727325603365898,
+      "num_tokens": 51568434.0,
+      "step": 2690
+    },
+    {
+      "entropy": 0.9859485492110253,
+      "epoch": 0.9229847168115073,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.6814009320717414e-05,
+      "loss": 0.0646,
+      "mean_token_accuracy": 0.9771854683756829,
+      "num_tokens": 51664822.0,
+      "step": 2695
+    },
+    {
+      "entropy": 0.9673167899250984,
+      "epoch": 0.9246971188835138,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.6806948171162266e-05,
+      "loss": 0.073,
+      "mean_token_accuracy": 0.9765182673931122,
+      "num_tokens": 51761270.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.9643731415271759,
+      "epoch": 0.9264095209555203,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.679988702160712e-05,
+      "loss": 0.073,
+      "mean_token_accuracy": 0.9760948166251182,
+      "num_tokens": 51857201.0,
+      "step": 2705
+    },
+    {
+      "entropy": 0.9687802597880364,
+      "epoch": 0.9281219230275268,
+      "grad_norm": 0.77734375,
+      "learning_rate": 1.679282587205197e-05,
+      "loss": 0.0697,
+      "mean_token_accuracy": 0.9760758623480796,
+      "num_tokens": 51954590.0,
+      "step": 2710
+    },
+    {
+      "entropy": 0.9934337750077248,
+      "epoch": 0.9298343250995333,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.6785764722496823e-05,
+      "loss": 0.0843,
+      "mean_token_accuracy": 0.972426763176918,
+      "num_tokens": 52051243.0,
+      "step": 2715
+    },
+    {
+      "entropy": 0.9970568343997002,
+      "epoch": 0.9315467271715399,
+      "grad_norm": 0.96484375,
+      "learning_rate": 1.677870357294168e-05,
+      "loss": 0.0743,
+      "mean_token_accuracy": 0.9749300390481949,
+      "num_tokens": 52147494.0,
+      "step": 2720
+    },
+    {
+      "entropy": 0.9861614435911179,
+      "epoch": 0.9332591292435464,
+      "grad_norm": 0.96484375,
+      "learning_rate": 1.677164242338653e-05,
+      "loss": 0.0669,
+      "mean_token_accuracy": 0.9751275151968002,
+      "num_tokens": 52243542.0,
+      "step": 2725
+    },
+    {
+      "entropy": 0.9868069842457772,
+      "epoch": 0.9349715313155529,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.6764581273831383e-05,
+      "loss": 0.0734,
+      "mean_token_accuracy": 0.9750597506761551,
+      "num_tokens": 52338332.0,
+      "step": 2730
+    },
+    {
+      "entropy": 0.9374115392565727,
+      "epoch": 0.9366839333875594,
+      "grad_norm": 0.94921875,
+      "learning_rate": 1.6757520124276235e-05,
+      "loss": 0.0564,
+      "mean_token_accuracy": 0.9803335517644882,
+      "num_tokens": 52435934.0,
+      "step": 2735
+    },
+    {
+      "entropy": 0.9441143423318863,
+      "epoch": 0.9383963354595659,
+      "grad_norm": 1.2890625,
+      "learning_rate": 1.6750458974721087e-05,
+      "loss": 0.0722,
+      "mean_token_accuracy": 0.9737128004431724,
+      "num_tokens": 52533690.0,
+      "step": 2740
+    },
+    {
+      "entropy": 0.950364151597023,
+      "epoch": 0.9401087375315724,
+      "grad_norm": 1.0625,
+      "learning_rate": 1.674339782516594e-05,
+      "loss": 0.0738,
+      "mean_token_accuracy": 0.973547138273716,
+      "num_tokens": 52630075.0,
+      "step": 2745
+    },
+    {
+      "entropy": 0.953400057554245,
+      "epoch": 0.9418211396035789,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.673633667561079e-05,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9747496351599694,
+      "num_tokens": 52725993.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.9978614062070846,
+      "epoch": 0.9435335416755855,
+      "grad_norm": 0.8984375,
+      "learning_rate": 1.6729275526055644e-05,
+      "loss": 0.0785,
+      "mean_token_accuracy": 0.9707878053188324,
+      "num_tokens": 52821301.0,
+      "step": 2755
+    },
+    {
+      "entropy": 0.9885793298482894,
+      "epoch": 0.945245943747592,
+      "grad_norm": 1.078125,
+      "learning_rate": 1.6722214376500496e-05,
+      "loss": 0.0673,
+      "mean_token_accuracy": 0.9776830092072487,
+      "num_tokens": 52918107.0,
+      "step": 2760
+    },
+    {
+      "entropy": 0.9933202862739563,
+      "epoch": 0.9469583458195985,
+      "grad_norm": 1.078125,
+      "learning_rate": 1.6715153226945348e-05,
+      "loss": 0.0663,
+      "mean_token_accuracy": 0.9772309347987175,
+      "num_tokens": 53014405.0,
+      "step": 2765
+    },
+    {
+      "entropy": 0.963502575457096,
+      "epoch": 0.948670747891605,
+      "grad_norm": 1.3046875,
+      "learning_rate": 1.67080920773902e-05,
+      "loss": 0.0685,
+      "mean_token_accuracy": 0.9766884252429009,
+      "num_tokens": 53111410.0,
+      "step": 2770
+    },
+    {
+      "entropy": 0.9509256482124329,
+      "epoch": 0.9503831499636115,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.6701030927835052e-05,
+      "loss": 0.0781,
+      "mean_token_accuracy": 0.9729023471474647,
+      "num_tokens": 53205549.0,
+      "step": 2775
+    },
+    {
+      "entropy": 0.9464489609003067,
+      "epoch": 0.9520955520356179,
+      "grad_norm": 0.8203125,
+      "learning_rate": 1.6693969778279905e-05,
+      "loss": 0.0603,
+      "mean_token_accuracy": 0.979080531001091,
+      "num_tokens": 53302318.0,
+      "step": 2780
+    },
+    {
+      "entropy": 0.9415407299995422,
+      "epoch": 0.9538079541076244,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.6686908628724757e-05,
+      "loss": 0.0607,
+      "mean_token_accuracy": 0.9790382295846939,
+      "num_tokens": 53396766.0,
+      "step": 2785
+    },
+    {
+      "entropy": 0.9295911982655525,
+      "epoch": 0.9555203561796309,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.6679847479169612e-05,
+      "loss": 0.0709,
+      "mean_token_accuracy": 0.9747038275003433,
+      "num_tokens": 53493300.0,
+      "step": 2790
+    },
+    {
+      "entropy": 0.9324138760566711,
+      "epoch": 0.9572327582516375,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.6672786329614465e-05,
+      "loss": 0.0941,
+      "mean_token_accuracy": 0.9678083568811416,
+      "num_tokens": 53588322.0,
+      "step": 2795
+    },
+    {
+      "entropy": 0.9405335456132888,
+      "epoch": 0.958945160323644,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.6665725180059317e-05,
+      "loss": 0.0705,
+      "mean_token_accuracy": 0.9749884650111198,
+      "num_tokens": 53683343.0,
+      "step": 2800
+    },
+    {
+      "epoch": 0.958945160323644,
+      "eval_entropy": 0.9480948406041093,
+      "eval_loss": 0.07012418657541275,
+      "eval_mean_token_accuracy": 0.9758322776868104,
+      "eval_num_tokens": 53683343.0,
+      "eval_runtime": 342.17,
+      "eval_samples_per_second": 15.171,
+      "eval_steps_per_second": 15.171,
+      "step": 2800
+    },
+    {
+      "entropy": 0.9476415947079658,
+      "epoch": 0.9606575623956505,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.665866403050417e-05,
+      "loss": 0.0669,
+      "mean_token_accuracy": 0.976867787539959,
+      "num_tokens": 53780021.0,
+      "step": 2805
+    },
+    {
+      "entropy": 0.9423150017857551,
+      "epoch": 0.962369964467657,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.665160288094902e-05,
+      "loss": 0.0715,
+      "mean_token_accuracy": 0.9760167255997658,
+      "num_tokens": 53876300.0,
+      "step": 2810
+    },
+    {
+      "entropy": 0.9487619116902352,
+      "epoch": 0.9640823665396635,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.6644541731393873e-05,
+      "loss": 0.0653,
+      "mean_token_accuracy": 0.9764960244297981,
+      "num_tokens": 53971307.0,
+      "step": 2815
+    },
+    {
+      "entropy": 0.9514382466673851,
+      "epoch": 0.96579476861167,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.6637480581838725e-05,
+      "loss": 0.063,
+      "mean_token_accuracy": 0.9773913502693177,
+      "num_tokens": 54066342.0,
+      "step": 2820
+    },
+    {
+      "entropy": 0.9581455826759339,
+      "epoch": 0.9675071706836765,
+      "grad_norm": 0.85546875,
+      "learning_rate": 1.6630419432283578e-05,
+      "loss": 0.0703,
+      "mean_token_accuracy": 0.974689881503582,
+      "num_tokens": 54162967.0,
+      "step": 2825
+    },
+    {
+      "entropy": 0.9725223809480668,
+      "epoch": 0.969219572755683,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.662335828272843e-05,
+      "loss": 0.0762,
+      "mean_token_accuracy": 0.9736184805631638,
+      "num_tokens": 54257202.0,
+      "step": 2830
+    },
+    {
+      "entropy": 0.9718520253896713,
+      "epoch": 0.9709319748276896,
+      "grad_norm": 0.86328125,
+      "learning_rate": 1.6616297133173282e-05,
+      "loss": 0.063,
+      "mean_token_accuracy": 0.9791649013757706,
+      "num_tokens": 54352522.0,
+      "step": 2835
+    },
+    {
+      "entropy": 0.9721231788396836,
+      "epoch": 0.9726443768996961,
+      "grad_norm": 1.0859375,
+      "learning_rate": 1.6609235983618134e-05,
+      "loss": 0.0636,
+      "mean_token_accuracy": 0.976552638411522,
+      "num_tokens": 54447948.0,
+      "step": 2840
+    },
+    {
+      "entropy": 0.9618054986000061,
+      "epoch": 0.9743567789717026,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.6602174834062986e-05,
+      "loss": 0.0657,
+      "mean_token_accuracy": 0.977731990814209,
+      "num_tokens": 54542337.0,
+      "step": 2845
+    },
+    {
+      "entropy": 0.9337889388203621,
+      "epoch": 0.9760691810437091,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.659511368450784e-05,
+      "loss": 0.0642,
+      "mean_token_accuracy": 0.9778771981596946,
+      "num_tokens": 54639109.0,
+      "step": 2850
+    },
+    {
+      "entropy": 0.9521181106567382,
+      "epoch": 0.9777815831157156,
+      "grad_norm": 0.8671875,
+      "learning_rate": 1.658805253495269e-05,
+      "loss": 0.0782,
+      "mean_token_accuracy": 0.9726430088281631,
+      "num_tokens": 54734283.0,
+      "step": 2855
+    },
+    {
+      "entropy": 0.9422412633895874,
+      "epoch": 0.9794939851877221,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.6580991385397546e-05,
+      "loss": 0.0577,
+      "mean_token_accuracy": 0.9813840791583062,
+      "num_tokens": 54830058.0,
+      "step": 2860
+    },
+    {
+      "entropy": 0.9561690777540207,
+      "epoch": 0.9812063872597285,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.65739302358424e-05,
+      "loss": 0.0731,
+      "mean_token_accuracy": 0.9734148949384689,
+      "num_tokens": 54925941.0,
+      "step": 2865
+    },
+    {
+      "entropy": 0.9799683928489685,
+      "epoch": 0.982918789331735,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.656686908628725e-05,
+      "loss": 0.0787,
+      "mean_token_accuracy": 0.9713241353631019,
+      "num_tokens": 55022550.0,
+      "step": 2870
+    },
+    {
+      "entropy": 0.9695358350872993,
+      "epoch": 0.9846311914037416,
+      "grad_norm": 0.95703125,
+      "learning_rate": 1.6559807936732103e-05,
+      "loss": 0.0597,
+      "mean_token_accuracy": 0.9795848146080971,
+      "num_tokens": 55120430.0,
+      "step": 2875
+    },
+    {
+      "entropy": 1.0011437699198722,
+      "epoch": 0.9863435934757481,
+      "grad_norm": 0.96875,
+      "learning_rate": 1.6552746787176955e-05,
+      "loss": 0.0684,
+      "mean_token_accuracy": 0.9753994226455689,
+      "num_tokens": 55215406.0,
+      "step": 2880
+    },
+    {
+      "entropy": 1.0233532801270484,
+      "epoch": 0.9880559955477546,
+      "grad_norm": 1.1953125,
+      "learning_rate": 1.6545685637621807e-05,
+      "loss": 0.0829,
+      "mean_token_accuracy": 0.9695719391107559,
+      "num_tokens": 55310180.0,
+      "step": 2885
+    },
+    {
+      "entropy": 1.0195039480924606,
+      "epoch": 0.9897683976197611,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.653862448806666e-05,
+      "loss": 0.071,
+      "mean_token_accuracy": 0.9743932992219925,
+      "num_tokens": 55404325.0,
+      "step": 2890
+    },
+    {
+      "entropy": 1.0267487913370132,
+      "epoch": 0.9914807996917676,
+      "grad_norm": 1.625,
+      "learning_rate": 1.653156333851151e-05,
+      "loss": 0.0806,
+      "mean_token_accuracy": 0.9701411068439484,
+      "num_tokens": 55499762.0,
+      "step": 2895
+    },
+    {
+      "entropy": 1.0262441650032996,
+      "epoch": 0.9931932017637741,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.6524502188956364e-05,
+      "loss": 0.0639,
+      "mean_token_accuracy": 0.9763584673404694,
+      "num_tokens": 55595048.0,
+      "step": 2900
+    },
+    {
+      "entropy": 1.0100090682506562,
+      "epoch": 0.9949056038357806,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.6517441039401216e-05,
+      "loss": 0.076,
+      "mean_token_accuracy": 0.9738606035709381,
+      "num_tokens": 55690060.0,
+      "step": 2905
+    },
+    {
+      "entropy": 0.9947652608156204,
+      "epoch": 0.9966180059077872,
+      "grad_norm": 0.98046875,
+      "learning_rate": 1.6510379889846068e-05,
+      "loss": 0.0618,
+      "mean_token_accuracy": 0.9773861810564994,
+      "num_tokens": 55786280.0,
+      "step": 2910
+    },
+    {
+      "entropy": 0.9965799048542976,
+      "epoch": 0.9983304079797937,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.650331874029092e-05,
+      "loss": 0.0642,
+      "mean_token_accuracy": 0.9770623341202735,
+      "num_tokens": 55881848.0,
+      "step": 2915
+    },
+    {
+      "entropy": 0.9892813471647409,
+      "epoch": 1.0,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.6496257590735772e-05,
+      "loss": 0.0637,
+      "mean_token_accuracy": 0.9787636078321017,
+      "num_tokens": 55974635.0,
+      "step": 2920
+    },
+    {
+      "entropy": 1.0098810076713562,
+      "epoch": 1.0017124020720065,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.6489196441180624e-05,
+      "loss": 0.0741,
+      "mean_token_accuracy": 0.9739048555493355,
+      "num_tokens": 56067997.0,
+      "step": 2925
+    },
+    {
+      "entropy": 0.9655358344316483,
+      "epoch": 1.003424804144013,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.648213529162548e-05,
+      "loss": 0.0659,
+      "mean_token_accuracy": 0.9775921151041984,
+      "num_tokens": 56163159.0,
+      "step": 2930
+    },
+    {
+      "entropy": 0.9377493798732758,
+      "epoch": 1.0051372062160195,
+      "grad_norm": 0.89453125,
+      "learning_rate": 1.6475074142070332e-05,
+      "loss": 0.0584,
+      "mean_token_accuracy": 0.9798688739538193,
+      "num_tokens": 56259435.0,
+      "step": 2935
+    },
+    {
+      "entropy": 0.9051266267895699,
+      "epoch": 1.006849608288026,
+      "grad_norm": 0.71484375,
+      "learning_rate": 1.6468012992515184e-05,
+      "loss": 0.0551,
+      "mean_token_accuracy": 0.9802899986505509,
+      "num_tokens": 56355929.0,
+      "step": 2940
+    },
+    {
+      "entropy": 0.8824210211634635,
+      "epoch": 1.0085620103600326,
+      "grad_norm": 0.9765625,
+      "learning_rate": 1.6460951842960036e-05,
+      "loss": 0.0558,
+      "mean_token_accuracy": 0.9798930689692498,
+      "num_tokens": 56453200.0,
+      "step": 2945
+    },
+    {
+      "entropy": 0.8910400390625,
+      "epoch": 1.010274412432039,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.645389069340489e-05,
+      "loss": 0.0674,
+      "mean_token_accuracy": 0.9765957236289978,
+      "num_tokens": 56547090.0,
+      "step": 2950
+    },
+    {
+      "entropy": 0.8885609060525894,
+      "epoch": 1.0119868145040456,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.644682954384974e-05,
+      "loss": 0.058,
+      "mean_token_accuracy": 0.9788002461194992,
+      "num_tokens": 56642362.0,
+      "step": 2955
+    },
+    {
+      "entropy": 0.9048280924558639,
+      "epoch": 1.013699216576052,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.6439768394294593e-05,
+      "loss": 0.0716,
+      "mean_token_accuracy": 0.9741968616843224,
+      "num_tokens": 56737484.0,
+      "step": 2960
+    },
+    {
+      "entropy": 0.9081128984689713,
+      "epoch": 1.0154116186480586,
+      "grad_norm": 0.81640625,
+      "learning_rate": 1.6432707244739445e-05,
+      "loss": 0.0674,
+      "mean_token_accuracy": 0.9759755790233612,
+      "num_tokens": 56833397.0,
+      "step": 2965
+    },
+    {
+      "entropy": 0.9255183428525925,
+      "epoch": 1.0171240207200651,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.6425646095184297e-05,
+      "loss": 0.053,
+      "mean_token_accuracy": 0.9818565338850022,
+      "num_tokens": 56928501.0,
+      "step": 2970
+    },
+    {
+      "entropy": 0.9304577693343162,
+      "epoch": 1.0188364227920716,
+      "grad_norm": 1.125,
+      "learning_rate": 1.641858494562915e-05,
+      "loss": 0.0632,
+      "mean_token_accuracy": 0.9773348525166512,
+      "num_tokens": 57025240.0,
+      "step": 2975
+    },
+    {
+      "entropy": 0.9325456500053406,
+      "epoch": 1.0205488248640782,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.6411523796074002e-05,
+      "loss": 0.0641,
+      "mean_token_accuracy": 0.9783900573849678,
+      "num_tokens": 57119449.0,
+      "step": 2980
+    },
+    {
+      "entropy": 0.8931903123855591,
+      "epoch": 1.0222612269360847,
+      "grad_norm": 0.953125,
+      "learning_rate": 1.6404462646518854e-05,
+      "loss": 0.0565,
+      "mean_token_accuracy": 0.979720464348793,
+      "num_tokens": 57215738.0,
+      "step": 2985
+    },
+    {
+      "entropy": 0.8842774495482445,
+      "epoch": 1.0239736290080912,
+      "grad_norm": 0.92578125,
+      "learning_rate": 1.6397401496963706e-05,
+      "loss": 0.0508,
+      "mean_token_accuracy": 0.9802591532468796,
+      "num_tokens": 57311502.0,
+      "step": 2990
+    },
+    {
+      "entropy": 0.8743513241410256,
+      "epoch": 1.0256860310800977,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.6390340347408558e-05,
+      "loss": 0.0634,
+      "mean_token_accuracy": 0.976072047650814,
+      "num_tokens": 57406748.0,
+      "step": 2995
+    },
+    {
+      "entropy": 0.8594859436154365,
+      "epoch": 1.0273984331521042,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.6383279197853414e-05,
+      "loss": 0.0557,
+      "mean_token_accuracy": 0.9787166699767113,
+      "num_tokens": 57500843.0,
+      "step": 3000
+    },
+    {
+      "epoch": 1.0273984331521042,
+      "eval_entropy": 0.8527569949248006,
+      "eval_loss": 0.07056921720504761,
+      "eval_mean_token_accuracy": 0.9757542677339941,
+      "eval_num_tokens": 57500843.0,
+      "eval_runtime": 341.7814,
+      "eval_samples_per_second": 15.188,
+      "eval_steps_per_second": 15.188,
+      "step": 3000
+    },
+    {
+      "entropy": 0.8743501231074333,
+      "epoch": 1.0291108352241105,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.6376218048298266e-05,
+      "loss": 0.0799,
+      "mean_token_accuracy": 0.9730860769748688,
+      "num_tokens": 57594939.0,
+      "step": 3005
+    },
+    {
+      "entropy": 0.8810740992426872,
+      "epoch": 1.030823237296117,
+      "grad_norm": 1.2109375,
+      "learning_rate": 1.6369156898743118e-05,
+      "loss": 0.0495,
+      "mean_token_accuracy": 0.9817509651184082,
+      "num_tokens": 57690465.0,
+      "step": 3010
+    },
+    {
+      "entropy": 0.8944238230586052,
+      "epoch": 1.0325356393681235,
+      "grad_norm": 0.91015625,
+      "learning_rate": 1.636209574918797e-05,
+      "loss": 0.0593,
+      "mean_token_accuracy": 0.9786673188209534,
+      "num_tokens": 57787457.0,
+      "step": 3015
+    },
+    {
+      "entropy": 0.9141129747033119,
+      "epoch": 1.03424804144013,
+      "grad_norm": 0.875,
+      "learning_rate": 1.6355034599632822e-05,
+      "loss": 0.0453,
+      "mean_token_accuracy": 0.9834377780556679,
+      "num_tokens": 57883004.0,
+      "step": 3020
+    },
+    {
+      "entropy": 0.9124521747231483,
+      "epoch": 1.0359604435121366,
+      "grad_norm": 1.0703125,
+      "learning_rate": 1.6347973450077675e-05,
+      "loss": 0.0617,
+      "mean_token_accuracy": 0.9783773273229599,
+      "num_tokens": 57980536.0,
+      "step": 3025
+    },
+    {
+      "entropy": 0.9251243323087692,
+      "epoch": 1.037672845584143,
+      "grad_norm": 1.2578125,
+      "learning_rate": 1.6340912300522527e-05,
+      "loss": 0.0704,
+      "mean_token_accuracy": 0.9733127430081367,
+      "num_tokens": 58075288.0,
+      "step": 3030
+    },
+    {
+      "entropy": 0.8932579517364502,
+      "epoch": 1.0393852476561496,
+      "grad_norm": 1.1875,
+      "learning_rate": 1.633385115096738e-05,
+      "loss": 0.0639,
+      "mean_token_accuracy": 0.9764333948493004,
+      "num_tokens": 58172569.0,
+      "step": 3035
+    },
+    {
+      "entropy": 0.9052770018577576,
+      "epoch": 1.041097649728156,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.632679000141223e-05,
+      "loss": 0.0623,
+      "mean_token_accuracy": 0.9783546477556229,
+      "num_tokens": 58266912.0,
+      "step": 3040
+    },
+    {
+      "entropy": 0.8669688135385514,
+      "epoch": 1.0428100518001626,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.6319728851857083e-05,
+      "loss": 0.0602,
+      "mean_token_accuracy": 0.9792937994003296,
+      "num_tokens": 58363072.0,
+      "step": 3045
+    },
+    {
+      "entropy": 0.8911859810352325,
+      "epoch": 1.0445224538721691,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.6312667702301936e-05,
+      "loss": 0.068,
+      "mean_token_accuracy": 0.9762070998549461,
+      "num_tokens": 58459144.0,
+      "step": 3050
+    },
+    {
+      "entropy": 0.903916372358799,
+      "epoch": 1.0462348559441756,
+      "grad_norm": 0.66015625,
+      "learning_rate": 1.6305606552746788e-05,
+      "loss": 0.0603,
+      "mean_token_accuracy": 0.9786941900849342,
+      "num_tokens": 58554334.0,
+      "step": 3055
+    },
+    {
+      "entropy": 0.891601949930191,
+      "epoch": 1.0479472580161822,
+      "grad_norm": 0.97265625,
+      "learning_rate": 1.629854540319164e-05,
+      "loss": 0.0603,
+      "mean_token_accuracy": 0.9767902821302414,
+      "num_tokens": 58649808.0,
+      "step": 3060
+    },
+    {
+      "entropy": 0.8664698898792267,
+      "epoch": 1.0496596600881887,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.6291484253636492e-05,
+      "loss": 0.0629,
+      "mean_token_accuracy": 0.9775587156414985,
+      "num_tokens": 58745866.0,
+      "step": 3065
+    },
+    {
+      "entropy": 0.869505487382412,
+      "epoch": 1.0513720621601952,
+      "grad_norm": 1.625,
+      "learning_rate": 1.6284423104081348e-05,
+      "loss": 0.0652,
+      "mean_token_accuracy": 0.9769822880625725,
+      "num_tokens": 58840795.0,
+      "step": 3070
+    },
+    {
+      "entropy": 0.87770616710186,
+      "epoch": 1.0530844642322017,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.62773619545262e-05,
+      "loss": 0.0637,
+      "mean_token_accuracy": 0.9759916350245476,
+      "num_tokens": 58937683.0,
+      "step": 3075
+    },
+    {
+      "entropy": 0.879734282195568,
+      "epoch": 1.0547968663042082,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.6270300804971052e-05,
+      "loss": 0.0654,
+      "mean_token_accuracy": 0.9750653713941574,
+      "num_tokens": 59033051.0,
+      "step": 3080
+    },
+    {
+      "entropy": 0.8639825642108917,
+      "epoch": 1.0565092683762147,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.6263239655415904e-05,
+      "loss": 0.0609,
+      "mean_token_accuracy": 0.978702911734581,
+      "num_tokens": 59129974.0,
+      "step": 3085
+    },
+    {
+      "entropy": 0.8483097955584527,
+      "epoch": 1.0582216704482212,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.6256178505860756e-05,
+      "loss": 0.0675,
+      "mean_token_accuracy": 0.9770919218659401,
+      "num_tokens": 59224530.0,
+      "step": 3090
+    },
+    {
+      "entropy": 0.8549211889505386,
+      "epoch": 1.0599340725202278,
+      "grad_norm": 0.8125,
+      "learning_rate": 1.624911735630561e-05,
+      "loss": 0.0428,
+      "mean_token_accuracy": 0.9845057934522629,
+      "num_tokens": 59319662.0,
+      "step": 3095
+    },
+    {
+      "entropy": 0.8503558397293091,
+      "epoch": 1.0616464745922343,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.624205620675046e-05,
+      "loss": 0.0552,
+      "mean_token_accuracy": 0.9804422736167908,
+      "num_tokens": 59414639.0,
+      "step": 3100
+    },
+    {
+      "entropy": 0.8334200575947761,
+      "epoch": 1.0633588766642408,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.6234995057195313e-05,
+      "loss": 0.0661,
+      "mean_token_accuracy": 0.9767545774579048,
+      "num_tokens": 59509672.0,
+      "step": 3105
+    },
+    {
+      "entropy": 0.8355286598205567,
+      "epoch": 1.0650712787362473,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.6227933907640165e-05,
+      "loss": 0.0522,
+      "mean_token_accuracy": 0.980382539331913,
+      "num_tokens": 59604818.0,
+      "step": 3110
+    },
+    {
+      "entropy": 0.8298099577426911,
+      "epoch": 1.0667836808082538,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.6220872758085017e-05,
+      "loss": 0.0524,
+      "mean_token_accuracy": 0.9813072860240937,
+      "num_tokens": 59701120.0,
+      "step": 3115
+    },
+    {
+      "entropy": 0.848918391764164,
+      "epoch": 1.0684960828802603,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.621381160852987e-05,
+      "loss": 0.0574,
+      "mean_token_accuracy": 0.9779295608401298,
+      "num_tokens": 59795998.0,
+      "step": 3120
+    },
+    {
+      "entropy": 0.8719707027077674,
+      "epoch": 1.0702084849522668,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.620675045897472e-05,
+      "loss": 0.0817,
+      "mean_token_accuracy": 0.9715264976024628,
+      "num_tokens": 59890754.0,
+      "step": 3125
+    },
+    {
+      "entropy": 0.8746989250183106,
+      "epoch": 1.0719208870242734,
+      "grad_norm": 0.85546875,
+      "learning_rate": 1.6199689309419574e-05,
+      "loss": 0.0616,
+      "mean_token_accuracy": 0.978082337975502,
+      "num_tokens": 59987122.0,
+      "step": 3130
+    },
+    {
+      "entropy": 0.8864504680037498,
+      "epoch": 1.0736332890962799,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.6192628159864426e-05,
+      "loss": 0.0639,
+      "mean_token_accuracy": 0.9779197230935097,
+      "num_tokens": 60083355.0,
+      "step": 3135
+    },
+    {
+      "entropy": 0.8923596352338791,
+      "epoch": 1.0753456911682864,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.618556701030928e-05,
+      "loss": 0.0683,
+      "mean_token_accuracy": 0.9750730499625206,
+      "num_tokens": 60178207.0,
+      "step": 3140
+    },
+    {
+      "entropy": 0.87035953104496,
+      "epoch": 1.077058093240293,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.6178505860754134e-05,
+      "loss": 0.0776,
+      "mean_token_accuracy": 0.9722903192043304,
+      "num_tokens": 60274905.0,
+      "step": 3145
+    },
+    {
+      "entropy": 0.8527958109974861,
+      "epoch": 1.0787704953122994,
+      "grad_norm": 0.984375,
+      "learning_rate": 1.6171444711198986e-05,
+      "loss": 0.0624,
+      "mean_token_accuracy": 0.9777600631117821,
+      "num_tokens": 60371634.0,
+      "step": 3150
+    },
+    {
+      "entropy": 0.8534418120980263,
+      "epoch": 1.080482897384306,
+      "grad_norm": 0.8828125,
+      "learning_rate": 1.6164383561643838e-05,
+      "loss": 0.0666,
+      "mean_token_accuracy": 0.976043076813221,
+      "num_tokens": 60468234.0,
+      "step": 3155
+    },
+    {
+      "entropy": 0.8695672944188118,
+      "epoch": 1.0821952994563124,
+      "grad_norm": 1.3359375,
+      "learning_rate": 1.615732241208869e-05,
+      "loss": 0.0632,
+      "mean_token_accuracy": 0.9761998042464256,
+      "num_tokens": 60564159.0,
+      "step": 3160
+    },
+    {
+      "entropy": 0.8551686093211174,
+      "epoch": 1.083907701528319,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.6150261262533542e-05,
+      "loss": 0.0598,
+      "mean_token_accuracy": 0.978071017563343,
+      "num_tokens": 60659341.0,
+      "step": 3165
+    },
+    {
+      "entropy": 0.8612403854727745,
+      "epoch": 1.0856201036003255,
+      "grad_norm": 1.171875,
+      "learning_rate": 1.6143200112978394e-05,
+      "loss": 0.0699,
+      "mean_token_accuracy": 0.9772937417030334,
+      "num_tokens": 60755077.0,
+      "step": 3170
+    },
+    {
+      "entropy": 0.84052122682333,
+      "epoch": 1.087332505672332,
+      "grad_norm": 0.96484375,
+      "learning_rate": 1.6136138963423247e-05,
+      "loss": 0.0557,
+      "mean_token_accuracy": 0.9802494704723358,
+      "num_tokens": 60850358.0,
+      "step": 3175
+    },
+    {
+      "entropy": 0.8205159649252891,
+      "epoch": 1.0890449077443383,
+      "grad_norm": 0.9921875,
+      "learning_rate": 1.61290778138681e-05,
+      "loss": 0.0514,
+      "mean_token_accuracy": 0.9810981079936028,
+      "num_tokens": 60946555.0,
+      "step": 3180
+    },
+    {
+      "entropy": 0.8325344190001488,
+      "epoch": 1.0907573098163448,
+      "grad_norm": 0.80859375,
+      "learning_rate": 1.612201666431295e-05,
+      "loss": 0.0626,
+      "mean_token_accuracy": 0.9768680885434151,
+      "num_tokens": 61042354.0,
+      "step": 3185
+    },
+    {
+      "entropy": 0.8242442235350609,
+      "epoch": 1.0924697118883513,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.6114955514757803e-05,
+      "loss": 0.0646,
+      "mean_token_accuracy": 0.9768854349851608,
+      "num_tokens": 61137891.0,
+      "step": 3190
+    },
+    {
+      "entropy": 0.826060627400875,
+      "epoch": 1.0941821139603578,
+      "grad_norm": 0.76171875,
+      "learning_rate": 1.6107894365202655e-05,
+      "loss": 0.0568,
+      "mean_token_accuracy": 0.9801831752061844,
+      "num_tokens": 61233917.0,
+      "step": 3195
+    },
+    {
+      "entropy": 0.856308788061142,
+      "epoch": 1.0958945160323643,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.6100833215647508e-05,
+      "loss": 0.0666,
+      "mean_token_accuracy": 0.9756682381033898,
+      "num_tokens": 61329330.0,
+      "step": 3200
+    },
+    {
+      "epoch": 1.0958945160323643,
+      "eval_entropy": 0.8625875975878522,
+      "eval_loss": 0.06928043067455292,
+      "eval_mean_token_accuracy": 0.9760321049395381,
+      "eval_num_tokens": 61329330.0,
+      "eval_runtime": 343.1694,
+      "eval_samples_per_second": 15.127,
+      "eval_steps_per_second": 15.127,
+      "step": 3200
+    },
+    {
+      "entropy": 0.8722243189811707,
+      "epoch": 1.0976069181043708,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.609377206609236e-05,
+      "loss": 0.0638,
+      "mean_token_accuracy": 0.9757747337222099,
+      "num_tokens": 61425357.0,
+      "step": 3205
+    },
+    {
+      "entropy": 0.8615012854337692,
+      "epoch": 1.0993193201763773,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.6086710916537215e-05,
+      "loss": 0.0569,
+      "mean_token_accuracy": 0.9792459204792976,
+      "num_tokens": 61523543.0,
+      "step": 3210
+    },
+    {
+      "entropy": 0.8531653940677643,
+      "epoch": 1.1010317222483839,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.6079649766982067e-05,
+      "loss": 0.0623,
+      "mean_token_accuracy": 0.9775284275412559,
+      "num_tokens": 61620579.0,
+      "step": 3215
+    },
+    {
+      "entropy": 0.8647532910108566,
+      "epoch": 1.1027441243203904,
+      "grad_norm": 1.109375,
+      "learning_rate": 1.607258861742692e-05,
+      "loss": 0.0763,
+      "mean_token_accuracy": 0.973094767332077,
+      "num_tokens": 61715429.0,
+      "step": 3220
+    },
+    {
+      "entropy": 0.8397126033902168,
+      "epoch": 1.1044565263923969,
+      "grad_norm": 1.03125,
+      "learning_rate": 1.6065527467871772e-05,
+      "loss": 0.0524,
+      "mean_token_accuracy": 0.9821034133434295,
+      "num_tokens": 61811471.0,
+      "step": 3225
+    },
+    {
+      "entropy": 0.8370882570743561,
+      "epoch": 1.1061689284644034,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.6058466318316624e-05,
+      "loss": 0.0576,
+      "mean_token_accuracy": 0.9802135825157166,
+      "num_tokens": 61908540.0,
+      "step": 3230
+    },
+    {
+      "entropy": 0.8354578971862793,
+      "epoch": 1.10788133053641,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.6051405168761476e-05,
+      "loss": 0.0549,
+      "mean_token_accuracy": 0.9807366400957107,
+      "num_tokens": 62003553.0,
+      "step": 3235
+    },
+    {
+      "entropy": 0.8561637192964554,
+      "epoch": 1.1095937326084164,
+      "grad_norm": 1.3984375,
+      "learning_rate": 1.6044344019206328e-05,
+      "loss": 0.0588,
+      "mean_token_accuracy": 0.9789238944649696,
+      "num_tokens": 62098752.0,
+      "step": 3240
+    },
+    {
+      "entropy": 0.8672030672430993,
+      "epoch": 1.111306134680423,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.603728286965118e-05,
+      "loss": 0.0599,
+      "mean_token_accuracy": 0.9799745827913284,
+      "num_tokens": 62195600.0,
+      "step": 3245
+    },
+    {
+      "entropy": 0.8879306688904762,
+      "epoch": 1.1130185367524295,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.6030221720096033e-05,
+      "loss": 0.0605,
+      "mean_token_accuracy": 0.9782017156481743,
+      "num_tokens": 62290384.0,
+      "step": 3250
+    },
+    {
+      "entropy": 0.920284028351307,
+      "epoch": 1.114730938824436,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.6023160570540885e-05,
+      "loss": 0.0695,
+      "mean_token_accuracy": 0.9749491840600968,
+      "num_tokens": 62384249.0,
+      "step": 3255
+    },
+    {
+      "entropy": 0.897847993671894,
+      "epoch": 1.1164433408964425,
+      "grad_norm": 1.265625,
+      "learning_rate": 1.6016099420985737e-05,
+      "loss": 0.0589,
+      "mean_token_accuracy": 0.9784821212291718,
+      "num_tokens": 62480900.0,
+      "step": 3260
+    },
+    {
+      "entropy": 0.9038872599601746,
+      "epoch": 1.118155742968449,
+      "grad_norm": 1.3046875,
+      "learning_rate": 1.600903827143059e-05,
+      "loss": 0.076,
+      "mean_token_accuracy": 0.971885883808136,
+      "num_tokens": 62576526.0,
+      "step": 3265
+    },
+    {
+      "entropy": 0.8733076989650727,
+      "epoch": 1.1198681450404555,
+      "grad_norm": 1.25,
+      "learning_rate": 1.600197712187544e-05,
+      "loss": 0.055,
+      "mean_token_accuracy": 0.9807557061314582,
+      "num_tokens": 62672919.0,
+      "step": 3270
+    },
+    {
+      "entropy": 0.8596949741244316,
+      "epoch": 1.121580547112462,
+      "grad_norm": 1.265625,
+      "learning_rate": 1.5994915972320294e-05,
+      "loss": 0.066,
+      "mean_token_accuracy": 0.9782865032553673,
+      "num_tokens": 62768117.0,
+      "step": 3275
+    },
+    {
+      "entropy": 0.8478237375617027,
+      "epoch": 1.1232929491844685,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.598785482276515e-05,
+      "loss": 0.0666,
+      "mean_token_accuracy": 0.9767226159572602,
+      "num_tokens": 62864461.0,
+      "step": 3280
+    },
+    {
+      "entropy": 0.8456330627202988,
+      "epoch": 1.125005351256475,
+      "grad_norm": 1.046875,
+      "learning_rate": 1.598079367321e-05,
+      "loss": 0.0646,
+      "mean_token_accuracy": 0.9766177698969841,
+      "num_tokens": 62960611.0,
+      "step": 3285
+    },
+    {
+      "entropy": 0.8405693635344506,
+      "epoch": 1.1267177533284816,
+      "grad_norm": 1.171875,
+      "learning_rate": 1.5973732523654853e-05,
+      "loss": 0.064,
+      "mean_token_accuracy": 0.9772413223981857,
+      "num_tokens": 63057073.0,
+      "step": 3290
+    },
+    {
+      "entropy": 0.8662106260657311,
+      "epoch": 1.128430155400488,
+      "grad_norm": 1.296875,
+      "learning_rate": 1.5966671374099706e-05,
+      "loss": 0.0581,
+      "mean_token_accuracy": 0.9803111746907234,
+      "num_tokens": 63151959.0,
+      "step": 3295
+    },
+    {
+      "entropy": 0.8654055118560791,
+      "epoch": 1.1301425574724946,
+      "grad_norm": 1.15625,
+      "learning_rate": 1.5959610224544558e-05,
+      "loss": 0.06,
+      "mean_token_accuracy": 0.9784368440508843,
+      "num_tokens": 63247698.0,
+      "step": 3300
+    },
+    {
+      "entropy": 0.8705126538872718,
+      "epoch": 1.131854959544501,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.595254907498941e-05,
+      "loss": 0.0508,
+      "mean_token_accuracy": 0.9814905971288681,
+      "num_tokens": 63344447.0,
+      "step": 3305
+    },
+    {
+      "entropy": 0.8566119238734246,
+      "epoch": 1.1335673616165076,
+      "grad_norm": 1.296875,
+      "learning_rate": 1.5945487925434262e-05,
+      "loss": 0.0567,
+      "mean_token_accuracy": 0.9788544490933418,
+      "num_tokens": 63441001.0,
+      "step": 3310
+    },
+    {
+      "entropy": 0.8461316138505935,
+      "epoch": 1.1352797636885141,
+      "grad_norm": 1.0625,
+      "learning_rate": 1.5938426775879114e-05,
+      "loss": 0.055,
+      "mean_token_accuracy": 0.9801321387290954,
+      "num_tokens": 63539397.0,
+      "step": 3315
+    },
+    {
+      "entropy": 0.8657594919204712,
+      "epoch": 1.1369921657605206,
+      "grad_norm": 0.953125,
+      "learning_rate": 1.5931365626323966e-05,
+      "loss": 0.0699,
+      "mean_token_accuracy": 0.9751646563410759,
+      "num_tokens": 63634394.0,
+      "step": 3320
+    },
+    {
+      "entropy": 0.8594745621085167,
+      "epoch": 1.1387045678325272,
+      "grad_norm": 1.0625,
+      "learning_rate": 1.592430447676882e-05,
+      "loss": 0.0616,
+      "mean_token_accuracy": 0.978434470295906,
+      "num_tokens": 63728055.0,
+      "step": 3325
+    },
+    {
+      "entropy": 0.8570857509970665,
+      "epoch": 1.1404169699045337,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.591724332721367e-05,
+      "loss": 0.0511,
+      "mean_token_accuracy": 0.9822077453136444,
+      "num_tokens": 63823452.0,
+      "step": 3330
+    },
+    {
+      "entropy": 0.8722443252801895,
+      "epoch": 1.1421293719765402,
+      "grad_norm": 0.86328125,
+      "learning_rate": 1.5910182177658523e-05,
+      "loss": 0.0624,
+      "mean_token_accuracy": 0.9779923439025879,
+      "num_tokens": 63916800.0,
+      "step": 3335
+    },
+    {
+      "entropy": 0.8692046105861664,
+      "epoch": 1.1438417740485467,
+      "grad_norm": 1.09375,
+      "learning_rate": 1.5903121028103375e-05,
+      "loss": 0.0675,
+      "mean_token_accuracy": 0.9750535070896149,
+      "num_tokens": 64013259.0,
+      "step": 3340
+    },
+    {
+      "entropy": 0.8847835958003998,
+      "epoch": 1.145554176120553,
+      "grad_norm": 1.1484375,
+      "learning_rate": 1.5896059878548227e-05,
+      "loss": 0.0651,
+      "mean_token_accuracy": 0.9789275035262108,
+      "num_tokens": 64108854.0,
+      "step": 3345
+    },
+    {
+      "entropy": 0.918972896039486,
+      "epoch": 1.1472665781925597,
+      "grad_norm": 1.0078125,
+      "learning_rate": 1.5888998728993083e-05,
+      "loss": 0.0705,
+      "mean_token_accuracy": 0.9746803641319275,
+      "num_tokens": 64204590.0,
+      "step": 3350
+    },
+    {
+      "entropy": 0.9159505844116211,
+      "epoch": 1.148978980264566,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.5881937579437935e-05,
+      "loss": 0.0488,
+      "mean_token_accuracy": 0.9825111612677574,
+      "num_tokens": 64299790.0,
+      "step": 3355
+    },
+    {
+      "entropy": 0.9065679743885994,
+      "epoch": 1.1506913823365725,
+      "grad_norm": 1.3125,
+      "learning_rate": 1.5874876429882787e-05,
+      "loss": 0.071,
+      "mean_token_accuracy": 0.9745347738265991,
+      "num_tokens": 64395948.0,
+      "step": 3360
+    },
+    {
+      "entropy": 0.8990728467702865,
+      "epoch": 1.152403784408579,
+      "grad_norm": 0.875,
+      "learning_rate": 1.586781528032764e-05,
+      "loss": 0.0674,
+      "mean_token_accuracy": 0.9762404799461365,
+      "num_tokens": 64491887.0,
+      "step": 3365
+    },
+    {
+      "entropy": 0.8890839830040932,
+      "epoch": 1.1541161864805856,
+      "grad_norm": 0.89453125,
+      "learning_rate": 1.586075413077249e-05,
+      "loss": 0.057,
+      "mean_token_accuracy": 0.9794210523366929,
+      "num_tokens": 64589745.0,
+      "step": 3370
+    },
+    {
+      "entropy": 0.9014985293149949,
+      "epoch": 1.155828588552592,
+      "grad_norm": 1.3046875,
+      "learning_rate": 1.5853692981217344e-05,
+      "loss": 0.0787,
+      "mean_token_accuracy": 0.9704926788806916,
+      "num_tokens": 64684817.0,
+      "step": 3375
+    },
+    {
+      "entropy": 0.8692110657691956,
+      "epoch": 1.1575409906245986,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.5846631831662196e-05,
+      "loss": 0.0535,
+      "mean_token_accuracy": 0.980925039947033,
+      "num_tokens": 64781869.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.8538182318210602,
+      "epoch": 1.159253392696605,
+      "grad_norm": 1.1015625,
+      "learning_rate": 1.5839570682107048e-05,
+      "loss": 0.0474,
+      "mean_token_accuracy": 0.9825504943728447,
+      "num_tokens": 64878748.0,
+      "step": 3385
+    },
+    {
+      "entropy": 0.8495529070496559,
+      "epoch": 1.1609657947686116,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.58325095325519e-05,
+      "loss": 0.0693,
+      "mean_token_accuracy": 0.9754141867160797,
+      "num_tokens": 64973042.0,
+      "step": 3390
+    },
+    {
+      "entropy": 0.8619621366262435,
+      "epoch": 1.1626781968406181,
+      "grad_norm": 0.9453125,
+      "learning_rate": 1.5825448382996752e-05,
+      "loss": 0.0648,
+      "mean_token_accuracy": 0.9773951455950737,
+      "num_tokens": 65069539.0,
+      "step": 3395
+    },
+    {
+      "entropy": 0.8898006737232208,
+      "epoch": 1.1643905989126246,
+      "grad_norm": 1.015625,
+      "learning_rate": 1.5818387233441605e-05,
+      "loss": 0.0753,
+      "mean_token_accuracy": 0.9713006168603897,
+      "num_tokens": 65163815.0,
+      "step": 3400
+    },
+    {
+      "epoch": 1.1643905989126246,
+      "eval_entropy": 0.880468895166188,
+      "eval_loss": 0.06878264993429184,
+      "eval_mean_token_accuracy": 0.9761748176460618,
+      "eval_num_tokens": 65163815.0,
+      "eval_runtime": 343.3384,
+      "eval_samples_per_second": 15.119,
+      "eval_steps_per_second": 15.119,
+      "step": 3400
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 14600,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1414110720830259e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}