diff --git "a/checkpoint-3000/trainer_state.json" "b/checkpoint-3000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-3000/trainer_state.json"
@@ -0,0 +1,3694 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.5856,
+  "eval_steps": 100,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9365717947483063,
+      "epoch": 0.032,
+      "grad_norm": 245.0,
+      "learning_rate": 1.8e-07,
+      "loss": 6.202,
+      "mean_token_accuracy": 0.2543207791633904,
+      "num_tokens": 144208.0,
+      "step": 10
+    },
+    {
+      "entropy": 1.9173382006585598,
+      "epoch": 0.064,
+      "grad_norm": 230.0,
+      "learning_rate": 3.8e-07,
+      "loss": 6.1081,
+      "mean_token_accuracy": 0.25344993276521566,
+      "num_tokens": 285280.0,
+      "step": 20
+    },
+    {
+      "entropy": 1.9177092142403125,
+      "epoch": 0.096,
+      "grad_norm": 246.0,
+      "learning_rate": 5.800000000000001e-07,
+      "loss": 5.9977,
+      "mean_token_accuracy": 0.25974713498726487,
+      "num_tokens": 430075.0,
+      "step": 30
+    },
+    {
+      "entropy": 1.9969916075468064,
+      "epoch": 0.128,
+      "grad_norm": 218.0,
+      "learning_rate": 7.8e-07,
+      "loss": 5.7736,
+      "mean_token_accuracy": 0.25647984803654256,
+      "num_tokens": 574626.0,
+      "step": 40
+    },
+    {
+      "entropy": 2.0409375332295894,
+      "epoch": 0.16,
+      "grad_norm": 198.0,
+      "learning_rate": 9.800000000000001e-07,
+      "loss": 5.4968,
+      "mean_token_accuracy": 0.25714964754879477,
+      "num_tokens": 721064.0,
+      "step": 50
+    },
+    {
+      "entropy": 2.080457562953234,
+      "epoch": 0.192,
+      "grad_norm": 141.0,
+      "learning_rate": 1.1800000000000001e-06,
+      "loss": 4.9089,
+      "mean_token_accuracy": 0.2747303694486618,
+      "num_tokens": 865275.0,
+      "step": 60
+    },
+    {
+      "entropy": 2.1661732770502566,
+      "epoch": 0.224,
+      "grad_norm": 95.0,
+      "learning_rate": 1.3800000000000001e-06,
+      "loss": 4.2411,
+      "mean_token_accuracy": 0.3079248811118305,
+      "num_tokens": 1005142.0,
+      "step": 70
+    },
+    {
+      "entropy": 2.440109082311392,
+      "epoch": 0.256,
+      "grad_norm": 65.5,
+      "learning_rate": 1.5800000000000001e-06,
+      "loss": 3.7265,
+      "mean_token_accuracy": 0.32527065742760897,
+      "num_tokens": 1147262.0,
+      "step": 80
+    },
+    {
+      "entropy": 2.6965860120952128,
+      "epoch": 0.288,
+      "grad_norm": 43.75,
+      "learning_rate": 1.7800000000000001e-06,
+      "loss": 3.3189,
+      "mean_token_accuracy": 0.3538851010613143,
+      "num_tokens": 1283845.0,
+      "step": 90
+    },
+    {
+      "entropy": 2.7403977632522585,
+      "epoch": 0.32,
+      "grad_norm": 26.25,
+      "learning_rate": 1.98e-06,
+      "loss": 3.0218,
+      "mean_token_accuracy": 0.3812081998214126,
+      "num_tokens": 1430551.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.32,
+      "eval_coding_entropy": 2.1376109790802,
+      "eval_coding_loss": 2.211063861846924,
+      "eval_coding_mean_token_accuracy": 0.4966651816368103,
+      "eval_coding_num_tokens": 1430551.0,
+      "eval_coding_runtime": 81.7119,
+      "eval_coding_samples_per_second": 6.119,
+      "eval_coding_steps_per_second": 3.06,
+      "step": 100
+    },
+    {
+      "epoch": 0.32,
+      "eval_math_entropy": 2.7154639225006103,
+      "eval_math_loss": 2.8518309593200684,
+      "eval_math_mean_token_accuracy": 0.402160507440567,
+      "eval_math_num_tokens": 1430551.0,
+      "eval_math_runtime": 46.4902,
+      "eval_math_samples_per_second": 10.755,
+      "eval_math_steps_per_second": 5.377,
+      "step": 100
+    },
+    {
+      "entropy": 2.6445168428123,
+      "epoch": 0.352,
+      "grad_norm": 22.25,
+      "learning_rate": 2.1800000000000003e-06,
+      "loss": 2.7264,
+      "mean_token_accuracy": 0.41560478005558255,
+      "num_tokens": 1573121.0,
+      "step": 110
+    },
+    {
+      "entropy": 2.5231464847922327,
+      "epoch": 0.384,
+      "grad_norm": 17.5,
+      "learning_rate": 2.38e-06,
+      "loss": 2.597,
+      "mean_token_accuracy": 0.4375463806092739,
+      "num_tokens": 1714945.0,
+      "step": 120
+    },
+    {
+      "entropy": 2.426661425083876,
+      "epoch": 0.416,
+      "grad_norm": 18.875,
+      "learning_rate": 2.5800000000000003e-06,
+      "loss": 2.4805,
+      "mean_token_accuracy": 0.45005974899977447,
+      "num_tokens": 1851617.0,
+      "step": 130
+    },
+    {
+      "entropy": 2.3277198508381844,
+      "epoch": 0.448,
+      "grad_norm": 22.0,
+      "learning_rate": 2.7800000000000005e-06,
+      "loss": 2.3694,
+      "mean_token_accuracy": 0.4707961905747652,
+      "num_tokens": 1994075.0,
+      "step": 140
+    },
+    {
+      "entropy": 2.2954090014100075,
+      "epoch": 0.48,
+      "grad_norm": 20.875,
+      "learning_rate": 2.9800000000000003e-06,
+      "loss": 2.3273,
+      "mean_token_accuracy": 0.4766052259132266,
+      "num_tokens": 2136748.0,
+      "step": 150
+    },
+    {
+      "entropy": 2.2432560376822948,
+      "epoch": 0.512,
+      "grad_norm": 15.1875,
+      "learning_rate": 3.1800000000000005e-06,
+      "loss": 2.2433,
+      "mean_token_accuracy": 0.4955187674611807,
+      "num_tokens": 2276430.0,
+      "step": 160
+    },
+    {
+      "entropy": 2.1506235644221308,
+      "epoch": 0.544,
+      "grad_norm": 20.375,
+      "learning_rate": 3.3800000000000007e-06,
+      "loss": 2.1835,
+      "mean_token_accuracy": 0.5024478340521454,
+      "num_tokens": 2420005.0,
+      "step": 170
+    },
+    {
+      "entropy": 2.173030565679073,
+      "epoch": 0.576,
+      "grad_norm": 21.25,
+      "learning_rate": 3.58e-06,
+      "loss": 2.2005,
+      "mean_token_accuracy": 0.5029806947335601,
+      "num_tokens": 2567140.0,
+      "step": 180
+    },
+    {
+      "entropy": 2.067853280156851,
+      "epoch": 0.608,
+      "grad_norm": 15.625,
+      "learning_rate": 3.7800000000000002e-06,
+      "loss": 2.0934,
+      "mean_token_accuracy": 0.52086176071316,
+      "num_tokens": 2707769.0,
+      "step": 190
+    },
+    {
+      "entropy": 2.0607680074870585,
+      "epoch": 0.64,
+      "grad_norm": 16.125,
+      "learning_rate": 3.980000000000001e-06,
+      "loss": 2.0413,
+      "mean_token_accuracy": 0.5280526082962751,
+      "num_tokens": 2851391.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.64,
+      "eval_coding_entropy": 2.082627496242523,
+      "eval_coding_loss": 2.0564982891082764,
+      "eval_coding_mean_token_accuracy": 0.5250193012952804,
+      "eval_coding_num_tokens": 2851391.0,
+      "eval_coding_runtime": 81.149,
+      "eval_coding_samples_per_second": 6.162,
+      "eval_coding_steps_per_second": 3.081,
+      "step": 200
+    },
+    {
+      "epoch": 0.64,
+      "eval_math_entropy": 2.0227903814315797,
+      "eval_math_loss": 2.056239604949951,
+      "eval_math_mean_token_accuracy": 0.5309821690320968,
+      "eval_math_num_tokens": 2851391.0,
+      "eval_math_runtime": 46.4944,
+      "eval_math_samples_per_second": 10.754,
+      "eval_math_steps_per_second": 5.377,
+      "step": 200
+    },
+    {
+      "entropy": 2.0263434551656245,
+      "epoch": 0.672,
+      "grad_norm": 15.6875,
+      "learning_rate": 4.18e-06,
+      "loss": 2.0178,
+      "mean_token_accuracy": 0.5313251735642552,
+      "num_tokens": 2997713.0,
+      "step": 210
+    },
+    {
+      "entropy": 2.0123437851667405,
+      "epoch": 0.704,
+      "grad_norm": 17.375,
+      "learning_rate": 4.38e-06,
+      "loss": 2.0443,
+      "mean_token_accuracy": 0.5295042948797345,
+      "num_tokens": 3141249.0,
+      "step": 220
+    },
+    {
+      "entropy": 2.039450883120298,
+      "epoch": 0.736,
+      "grad_norm": 18.5,
+      "learning_rate": 4.58e-06,
+      "loss": 2.0519,
+      "mean_token_accuracy": 0.5338158709928393,
+      "num_tokens": 3279135.0,
+      "step": 230
+    },
+    {
+      "entropy": 1.9529575437307358,
+      "epoch": 0.768,
+      "grad_norm": 15.4375,
+      "learning_rate": 4.78e-06,
+      "loss": 1.9503,
+      "mean_token_accuracy": 0.5456137385219335,
+      "num_tokens": 3429580.0,
+      "step": 240
+    },
+    {
+      "entropy": 1.9847303189337253,
+      "epoch": 0.8,
+      "grad_norm": 18.875,
+      "learning_rate": 4.980000000000001e-06,
+      "loss": 1.9973,
+      "mean_token_accuracy": 0.5430868744850159,
+      "num_tokens": 3574102.0,
+      "step": 250
+    },
+    {
+      "entropy": 1.941353114694357,
+      "epoch": 0.832,
+      "grad_norm": 17.625,
+      "learning_rate": 5.18e-06,
+      "loss": 1.9468,
+      "mean_token_accuracy": 0.5475187944248319,
+      "num_tokens": 3711682.0,
+      "step": 260
+    },
+    {
+      "entropy": 1.8798863902688026,
+      "epoch": 0.864,
+      "grad_norm": 21.0,
+      "learning_rate": 5.380000000000001e-06,
+      "loss": 1.873,
+      "mean_token_accuracy": 0.5597963750362396,
+      "num_tokens": 3855113.0,
+      "step": 270
+    },
+    {
+      "entropy": 1.835599235445261,
+      "epoch": 0.896,
+      "grad_norm": 14.375,
+      "learning_rate": 5.580000000000001e-06,
+      "loss": 1.8228,
+      "mean_token_accuracy": 0.5651934513822198,
+      "num_tokens": 3995349.0,
+      "step": 280
+    },
+    {
+      "entropy": 1.8507938131690025,
+      "epoch": 0.928,
+      "grad_norm": 16.875,
+      "learning_rate": 5.78e-06,
+      "loss": 1.8394,
+      "mean_token_accuracy": 0.5661519970744848,
+      "num_tokens": 4135852.0,
+      "step": 290
+    },
+    {
+      "entropy": 1.8499364234507083,
+      "epoch": 0.96,
+      "grad_norm": 16.75,
+      "learning_rate": 5.98e-06,
+      "loss": 1.8409,
+      "mean_token_accuracy": 0.5702653925865888,
+      "num_tokens": 4274655.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.96,
+      "eval_coding_entropy": 2.037804766178131,
+      "eval_coding_loss": 2.022986650466919,
+      "eval_coding_mean_token_accuracy": 0.5313419116735458,
+      "eval_coding_num_tokens": 4274655.0,
+      "eval_coding_runtime": 81.1157,
+      "eval_coding_samples_per_second": 6.164,
+      "eval_coding_steps_per_second": 3.082,
+      "step": 300
+    },
+    {
+      "epoch": 0.96,
+      "eval_math_entropy": 1.802053134918213,
+      "eval_math_loss": 1.8271188735961914,
+      "eval_math_mean_token_accuracy": 0.5726956763267517,
+      "eval_math_num_tokens": 4274655.0,
+      "eval_math_runtime": 46.5128,
+      "eval_math_samples_per_second": 10.75,
+      "eval_math_steps_per_second": 5.375,
+      "step": 300
+    },
+    {
+      "entropy": 1.8301428437232972,
+      "epoch": 0.992,
+      "grad_norm": 17.25,
+      "learning_rate": 6.18e-06,
+      "loss": 1.8206,
+      "mean_token_accuracy": 0.569915696978569,
+      "num_tokens": 4415600.0,
+      "step": 310
+    },
+    {
+      "entropy": 1.8003562704512948,
+      "epoch": 1.0224,
+      "grad_norm": 12.75,
+      "learning_rate": 6.380000000000001e-06,
+      "loss": 1.7748,
+      "mean_token_accuracy": 0.5778473768579332,
+      "num_tokens": 4551683.0,
+      "step": 320
+    },
+    {
+      "entropy": 1.7793153546750546,
+      "epoch": 1.0544,
+      "grad_norm": 15.125,
+      "learning_rate": 6.5800000000000005e-06,
+      "loss": 1.7851,
+      "mean_token_accuracy": 0.579111011326313,
+      "num_tokens": 4693696.0,
+      "step": 330
+    },
+    {
+      "entropy": 1.7894929252564906,
+      "epoch": 1.0864,
+      "grad_norm": 14.625,
+      "learning_rate": 6.780000000000001e-06,
+      "loss": 1.7635,
+      "mean_token_accuracy": 0.5829346276819706,
+      "num_tokens": 4834396.0,
+      "step": 340
+    },
+    {
+      "entropy": 1.7216685689985751,
+      "epoch": 1.1184,
+      "grad_norm": 15.1875,
+      "learning_rate": 6.98e-06,
+      "loss": 1.7294,
+      "mean_token_accuracy": 0.5859142445027828,
+      "num_tokens": 4971857.0,
+      "step": 350
+    },
+    {
+      "entropy": 1.795823823660612,
+      "epoch": 1.1504,
+      "grad_norm": 19.0,
+      "learning_rate": 7.180000000000001e-06,
+      "loss": 1.7665,
+      "mean_token_accuracy": 0.5832353200763464,
+      "num_tokens": 5110683.0,
+      "step": 360
+    },
+    {
+      "entropy": 1.7300004258751869,
+      "epoch": 1.1824,
+      "grad_norm": 13.625,
+      "learning_rate": 7.3800000000000005e-06,
+      "loss": 1.6956,
+      "mean_token_accuracy": 0.589705759473145,
+      "num_tokens": 5252621.0,
+      "step": 370
+    },
+    {
+      "entropy": 1.6924258664250373,
+      "epoch": 1.2144,
+      "grad_norm": 18.875,
+      "learning_rate": 7.58e-06,
+      "loss": 1.687,
+      "mean_token_accuracy": 0.5896685650572181,
+      "num_tokens": 5394768.0,
+      "step": 380
+    },
+    {
+      "entropy": 1.754713100194931,
+      "epoch": 1.2464,
+      "grad_norm": 12.5,
+      "learning_rate": 7.78e-06,
+      "loss": 1.7225,
+      "mean_token_accuracy": 0.5889439214020967,
+      "num_tokens": 5533439.0,
+      "step": 390
+    },
+    {
+      "entropy": 1.704547866433859,
+      "epoch": 1.2784,
+      "grad_norm": 20.5,
+      "learning_rate": 7.980000000000002e-06,
+      "loss": 1.6608,
+      "mean_token_accuracy": 0.599461050145328,
+      "num_tokens": 5670712.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.2784,
+      "eval_coding_entropy": 1.9745532732009887,
+      "eval_coding_loss": 2.004408359527588,
+      "eval_coding_mean_token_accuracy": 0.5342881647348404,
+      "eval_coding_num_tokens": 5670712.0,
+      "eval_coding_runtime": 81.3362,
+      "eval_coding_samples_per_second": 6.147,
+      "eval_coding_steps_per_second": 3.074,
+      "step": 400
+    },
+    {
+      "epoch": 1.2784,
+      "eval_math_entropy": 1.6563503637313843,
+      "eval_math_loss": 1.6785366535186768,
+      "eval_math_mean_token_accuracy": 0.6000865260362626,
+      "eval_math_num_tokens": 5670712.0,
+      "eval_math_runtime": 46.6244,
+      "eval_math_samples_per_second": 10.724,
+      "eval_math_steps_per_second": 5.362,
+      "step": 400
+    },
+    {
+      "entropy": 1.6839238464832307,
+      "epoch": 1.3104,
+      "grad_norm": 16.625,
+      "learning_rate": 8.18e-06,
+      "loss": 1.6869,
+      "mean_token_accuracy": 0.5953959075734019,
+      "num_tokens": 5813458.0,
+      "step": 410
+    },
+    {
+      "entropy": 1.6726863503456115,
+      "epoch": 1.3424,
+      "grad_norm": 12.75,
+      "learning_rate": 8.380000000000001e-06,
+      "loss": 1.6315,
+      "mean_token_accuracy": 0.6037707706913352,
+      "num_tokens": 5951880.0,
+      "step": 420
+    },
+    {
+      "entropy": 1.6084760576486588,
+      "epoch": 1.3744,
+      "grad_norm": 14.3125,
+      "learning_rate": 8.580000000000001e-06,
+      "loss": 1.5881,
+      "mean_token_accuracy": 0.6148429989814759,
+      "num_tokens": 6098256.0,
+      "step": 430
+    },
+    {
+      "entropy": 1.6628840889781713,
+      "epoch": 1.4064,
+      "grad_norm": 15.6875,
+      "learning_rate": 8.78e-06,
+      "loss": 1.6761,
+      "mean_token_accuracy": 0.6000858852639794,
+      "num_tokens": 6242646.0,
+      "step": 440
+    },
+    {
+      "entropy": 1.6300812467932702,
+      "epoch": 1.4384000000000001,
+      "grad_norm": 21.25,
+      "learning_rate": 8.98e-06,
+      "loss": 1.6003,
+      "mean_token_accuracy": 0.6083745462819934,
+      "num_tokens": 6386025.0,
+      "step": 450
+    },
+    {
+      "entropy": 1.6391606278717519,
+      "epoch": 1.4704,
+      "grad_norm": 20.375,
+      "learning_rate": 9.180000000000002e-06,
+      "loss": 1.6273,
+      "mean_token_accuracy": 0.6097741341218352,
+      "num_tokens": 6533075.0,
+      "step": 460
+    },
+    {
+      "entropy": 1.6905927129089833,
+      "epoch": 1.5024,
+      "grad_norm": 15.375,
+      "learning_rate": 9.38e-06,
+      "loss": 1.6572,
+      "mean_token_accuracy": 0.6046760326251388,
+      "num_tokens": 6677208.0,
+      "step": 470
+    },
+    {
+      "entropy": 1.6522716350853444,
+      "epoch": 1.5344,
+      "grad_norm": 13.5,
+      "learning_rate": 9.58e-06,
+      "loss": 1.6494,
+      "mean_token_accuracy": 0.6073551412671805,
+      "num_tokens": 6814455.0,
+      "step": 480
+    },
+    {
+      "entropy": 1.6329746030271053,
+      "epoch": 1.5664,
+      "grad_norm": 13.6875,
+      "learning_rate": 9.780000000000001e-06,
+      "loss": 1.5924,
+      "mean_token_accuracy": 0.614310959354043,
+      "num_tokens": 6958149.0,
+      "step": 490
+    },
+    {
+      "entropy": 1.5567305848002433,
+      "epoch": 1.5984,
+      "grad_norm": 12.6875,
+      "learning_rate": 9.980000000000001e-06,
+      "loss": 1.5416,
+      "mean_token_accuracy": 0.6263186866417527,
+      "num_tokens": 7099528.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.5984,
+      "eval_coding_entropy": 1.9465989785194397,
+      "eval_coding_loss": 1.990442156791687,
+      "eval_coding_mean_token_accuracy": 0.5386206238269806,
+      "eval_coding_num_tokens": 7099528.0,
+      "eval_coding_runtime": 81.2403,
+      "eval_coding_samples_per_second": 6.155,
+      "eval_coding_steps_per_second": 3.077,
+      "step": 500
+    },
+    {
+      "epoch": 1.5984,
+      "eval_math_entropy": 1.5453263101577759,
+      "eval_math_loss": 1.5754268169403076,
+      "eval_math_mean_token_accuracy": 0.6199949022531509,
+      "eval_math_num_tokens": 7099528.0,
+      "eval_math_runtime": 45.9145,
+      "eval_math_samples_per_second": 10.89,
+      "eval_math_steps_per_second": 5.445,
+      "step": 500
+    },
+    {
+      "entropy": 1.5644792415201665,
+      "epoch": 1.6303999999999998,
+      "grad_norm": 15.5625,
+      "learning_rate": 1.018e-05,
+      "loss": 1.5511,
+      "mean_token_accuracy": 0.6213703533634544,
+      "num_tokens": 7241222.0,
+      "step": 510
+    },
+    {
+      "entropy": 1.6044137500226499,
+      "epoch": 1.6623999999999999,
+      "grad_norm": 14.6875,
+      "learning_rate": 1.038e-05,
+      "loss": 1.5761,
+      "mean_token_accuracy": 0.6146464478224516,
+      "num_tokens": 7387899.0,
+      "step": 520
+    },
+    {
+      "entropy": 1.5174766711890697,
+      "epoch": 1.6944,
+      "grad_norm": 14.4375,
+      "learning_rate": 1.0580000000000002e-05,
+      "loss": 1.515,
+      "mean_token_accuracy": 0.6280366148799658,
+      "num_tokens": 7535577.0,
+      "step": 530
+    },
+    {
+      "entropy": 1.5012709453701973,
+      "epoch": 1.7264,
+      "grad_norm": 18.0,
+      "learning_rate": 1.0780000000000002e-05,
+      "loss": 1.4795,
+      "mean_token_accuracy": 0.6314780458807945,
+      "num_tokens": 7679139.0,
+      "step": 540
+    },
+    {
+      "entropy": 1.557135234773159,
+      "epoch": 1.7584,
+      "grad_norm": 21.0,
+      "learning_rate": 1.0980000000000002e-05,
+      "loss": 1.5279,
+      "mean_token_accuracy": 0.6282323149964213,
+      "num_tokens": 7821116.0,
+      "step": 550
+    },
+    {
+      "entropy": 1.5534105684608221,
+      "epoch": 1.7904,
+      "grad_norm": 15.9375,
+      "learning_rate": 1.1180000000000001e-05,
+      "loss": 1.5329,
+      "mean_token_accuracy": 0.6273708928376436,
+      "num_tokens": 7955945.0,
+      "step": 560
+    },
+    {
+      "entropy": 1.5024470299482346,
+      "epoch": 1.8224,
+      "grad_norm": 16.75,
+      "learning_rate": 1.138e-05,
+      "loss": 1.5027,
+      "mean_token_accuracy": 0.6360822120681405,
+      "num_tokens": 8098517.0,
+      "step": 570
+    },
+    {
+      "entropy": 1.468005495518446,
+      "epoch": 1.8544,
+      "grad_norm": 14.75,
+      "learning_rate": 1.1580000000000001e-05,
+      "loss": 1.4433,
+      "mean_token_accuracy": 0.6402354558929801,
+      "num_tokens": 8247652.0,
+      "step": 580
+    },
+    {
+      "entropy": 1.4894821517169476,
+      "epoch": 1.8864,
+      "grad_norm": 15.125,
+      "learning_rate": 1.178e-05,
+      "loss": 1.4651,
+      "mean_token_accuracy": 0.6387512655928731,
+      "num_tokens": 8390969.0,
+      "step": 590
+    },
+    {
+      "entropy": 1.5070936933159829,
+      "epoch": 1.9184,
+      "grad_norm": 14.625,
+      "learning_rate": 1.198e-05,
+      "loss": 1.4912,
+      "mean_token_accuracy": 0.6379801956936717,
+      "num_tokens": 8532905.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.9184,
+      "eval_coding_entropy": 1.9700443706512452,
+      "eval_coding_loss": 1.9896337985992432,
+      "eval_coding_mean_token_accuracy": 0.5387515650987625,
+      "eval_coding_num_tokens": 8532905.0,
+      "eval_coding_runtime": 81.1982,
+      "eval_coding_samples_per_second": 6.158,
+      "eval_coding_steps_per_second": 3.079,
+      "step": 600
+    },
+    {
+      "epoch": 1.9184,
+      "eval_math_entropy": 1.479466641664505,
+      "eval_math_loss": 1.4867653846740723,
+      "eval_math_mean_token_accuracy": 0.6389840010404587,
+      "eval_math_num_tokens": 8532905.0,
+      "eval_math_runtime": 46.5533,
+      "eval_math_samples_per_second": 10.74,
+      "eval_math_steps_per_second": 5.37,
+      "step": 600
+    },
+    {
+      "entropy": 1.5151923824101687,
+      "epoch": 1.9504000000000001,
+      "grad_norm": 17.125,
+      "learning_rate": 1.218e-05,
+      "loss": 1.4863,
+      "mean_token_accuracy": 0.6339302582666277,
+      "num_tokens": 8676028.0,
+      "step": 610
+    },
+    {
+      "entropy": 1.468616184964776,
+      "epoch": 1.9824000000000002,
+      "grad_norm": 14.5,
+      "learning_rate": 1.2380000000000002e-05,
+      "loss": 1.4732,
+      "mean_token_accuracy": 0.6385495683178306,
+      "num_tokens": 8817323.0,
+      "step": 620
+    },
+    {
+      "entropy": 1.5188211822196056,
+      "epoch": 2.0128,
+      "grad_norm": 14.4375,
+      "learning_rate": 1.2580000000000002e-05,
+      "loss": 1.4835,
+      "mean_token_accuracy": 0.634301284426137,
+      "num_tokens": 8949173.0,
+      "step": 630
+    },
+    {
+      "entropy": 1.3755222257226705,
+      "epoch": 2.0448,
+      "grad_norm": 12.9375,
+      "learning_rate": 1.2780000000000001e-05,
+      "loss": 1.3601,
+      "mean_token_accuracy": 0.6608490329235792,
+      "num_tokens": 9085612.0,
+      "step": 640
+    },
+    {
+      "entropy": 1.4207779351621865,
+      "epoch": 2.0768,
+      "grad_norm": 13.9375,
+      "learning_rate": 1.2980000000000001e-05,
+      "loss": 1.4097,
+      "mean_token_accuracy": 0.6495957912877202,
+      "num_tokens": 9233211.0,
+      "step": 650
+    },
+    {
+      "entropy": 1.437428993731737,
+      "epoch": 2.1088,
+      "grad_norm": 17.5,
+      "learning_rate": 1.3180000000000001e-05,
+      "loss": 1.4124,
+      "mean_token_accuracy": 0.6517477501183748,
+      "num_tokens": 9376523.0,
+      "step": 660
+    },
+    {
+      "entropy": 1.3855586621910334,
+      "epoch": 2.1408,
+      "grad_norm": 14.75,
+      "learning_rate": 1.3380000000000002e-05,
+      "loss": 1.3542,
+      "mean_token_accuracy": 0.6590309431776404,
+      "num_tokens": 9523963.0,
+      "step": 670
+    },
+    {
+      "entropy": 1.4285461019724608,
+      "epoch": 2.1728,
+      "grad_norm": 16.375,
+      "learning_rate": 1.3580000000000002e-05,
+      "loss": 1.4063,
+      "mean_token_accuracy": 0.6481046233326196,
+      "num_tokens": 9670446.0,
+      "step": 680
+    },
+    {
+      "entropy": 1.389345682412386,
+      "epoch": 2.2048,
+      "grad_norm": 14.5625,
+      "learning_rate": 1.378e-05,
+      "loss": 1.368,
+      "mean_token_accuracy": 0.6584958679974079,
+      "num_tokens": 9809324.0,
+      "step": 690
+    },
+    {
+      "entropy": 1.4275502871721983,
+      "epoch": 2.2368,
+      "grad_norm": 14.875,
+      "learning_rate": 1.398e-05,
+      "loss": 1.4041,
+      "mean_token_accuracy": 0.6543006511405111,
+      "num_tokens": 9943194.0,
+      "step": 700
+    },
+    {
+      "epoch": 2.2368,
+      "eval_coding_entropy": 1.8811444640159607,
+      "eval_coding_loss": 1.999657154083252,
+      "eval_coding_mean_token_accuracy": 0.5383623929023743,
+      "eval_coding_num_tokens": 9943194.0,
+      "eval_coding_runtime": 80.9539,
+      "eval_coding_samples_per_second": 6.176,
+      "eval_coding_steps_per_second": 3.088,
+      "step": 700
+    },
+    {
+      "epoch": 2.2368,
+      "eval_math_entropy": 1.3948393285274505,
+      "eval_math_loss": 1.4332354068756104,
+      "eval_math_mean_token_accuracy": 0.6499490721225738,
+      "eval_math_num_tokens": 9943194.0,
+      "eval_math_runtime": 46.0836,
+      "eval_math_samples_per_second": 10.85,
+      "eval_math_steps_per_second": 5.425,
+      "step": 700
+    },
+    {
+      "entropy": 1.407985271140933,
+      "epoch": 2.2688,
+      "grad_norm": 16.625,
+      "learning_rate": 1.418e-05,
+      "loss": 1.3837,
+      "mean_token_accuracy": 0.6551688494160771,
+      "num_tokens": 10087633.0,
+      "step": 710
+    },
+    {
+      "entropy": 1.378972141817212,
+      "epoch": 2.3008,
+      "grad_norm": 13.375,
+      "learning_rate": 1.4380000000000001e-05,
+      "loss": 1.3544,
+      "mean_token_accuracy": 0.6556477602571249,
+      "num_tokens": 10229382.0,
+      "step": 720
+    },
+    {
+      "entropy": 1.354974314570427,
+      "epoch": 2.3327999999999998,
+      "grad_norm": 14.9375,
+      "learning_rate": 1.4580000000000001e-05,
+      "loss": 1.3344,
+      "mean_token_accuracy": 0.6645016176626086,
+      "num_tokens": 10376268.0,
+      "step": 730
+    },
+    {
+      "entropy": 1.3568570345640183,
+      "epoch": 2.3648,
+      "grad_norm": 17.75,
+      "learning_rate": 1.478e-05,
+      "loss": 1.3345,
+      "mean_token_accuracy": 0.6638497773557901,
+      "num_tokens": 10518064.0,
+      "step": 740
+    },
+    {
+      "entropy": 1.418501327931881,
+      "epoch": 2.3968,
+      "grad_norm": 15.5625,
+      "learning_rate": 1.498e-05,
+      "loss": 1.3993,
+      "mean_token_accuracy": 0.6545212138444185,
+      "num_tokens": 10654090.0,
+      "step": 750
+    },
+    {
+      "entropy": 1.3498825568705797,
+      "epoch": 2.4288,
+      "grad_norm": 14.8125,
+      "learning_rate": 1.5180000000000002e-05,
+      "loss": 1.3302,
+      "mean_token_accuracy": 0.6608284536749125,
+      "num_tokens": 10803539.0,
+      "step": 760
+    },
+    {
+      "entropy": 1.3939791657030582,
+      "epoch": 2.4608,
+      "grad_norm": 15.1875,
+      "learning_rate": 1.5380000000000002e-05,
+      "loss": 1.3721,
+      "mean_token_accuracy": 0.6571606319397688,
+      "num_tokens": 10951500.0,
+      "step": 770
+    },
+    {
+      "entropy": 1.4095406230539083,
+      "epoch": 2.4928,
+      "grad_norm": 12.6875,
+      "learning_rate": 1.5580000000000003e-05,
+      "loss": 1.3827,
+      "mean_token_accuracy": 0.6542418278753758,
+      "num_tokens": 11089421.0,
+      "step": 780
+    },
+    {
+      "entropy": 1.3628040976822375,
+      "epoch": 2.5248,
+      "grad_norm": 16.0,
+      "learning_rate": 1.578e-05,
+      "loss": 1.3434,
+      "mean_token_accuracy": 0.6625945869833231,
+      "num_tokens": 11233056.0,
+      "step": 790
+    },
+    {
+      "entropy": 1.353104681521654,
+      "epoch": 2.5568,
+      "grad_norm": 13.5,
+      "learning_rate": 1.5980000000000003e-05,
+      "loss": 1.3441,
+      "mean_token_accuracy": 0.663878696039319,
+      "num_tokens": 11382658.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.5568,
+      "eval_coding_entropy": 1.88368141412735,
+      "eval_coding_loss": 2.0113229751586914,
+      "eval_coding_mean_token_accuracy": 0.5368727495670319,
+      "eval_coding_num_tokens": 11382658.0,
+      "eval_coding_runtime": 80.7647,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 800
+    },
+    {
+      "epoch": 2.5568,
+      "eval_math_entropy": 1.3451528911590576,
+      "eval_math_loss": 1.3778893947601318,
+      "eval_math_mean_token_accuracy": 0.6607881124019623,
+      "eval_math_num_tokens": 11382658.0,
+      "eval_math_runtime": 46.1509,
+      "eval_math_samples_per_second": 10.834,
+      "eval_math_steps_per_second": 5.417,
+      "step": 800
+    },
+    {
+      "entropy": 1.3698780447244645,
+      "epoch": 2.5888,
+      "grad_norm": 14.0625,
+      "learning_rate": 1.618e-05,
+      "loss": 1.3446,
+      "mean_token_accuracy": 0.667889142036438,
+      "num_tokens": 11520248.0,
+      "step": 810
+    },
+    {
+      "entropy": 1.3439954832196235,
+      "epoch": 2.6208,
+      "grad_norm": 13.25,
+      "learning_rate": 1.638e-05,
+      "loss": 1.3411,
+      "mean_token_accuracy": 0.6649587988853455,
+      "num_tokens": 11662426.0,
+      "step": 820
+    },
+    {
+      "entropy": 1.3382667150348424,
+      "epoch": 2.6528,
+      "grad_norm": 15.125,
+      "learning_rate": 1.658e-05,
+      "loss": 1.2968,
+      "mean_token_accuracy": 0.667837655171752,
+      "num_tokens": 11801529.0,
+      "step": 830
+    },
+    {
+      "entropy": 1.3607995260506869,
+      "epoch": 2.6848,
+      "grad_norm": 12.8125,
+      "learning_rate": 1.6780000000000002e-05,
+      "loss": 1.3342,
+      "mean_token_accuracy": 0.6632724912837148,
+      "num_tokens": 11945073.0,
+      "step": 840
+    },
+    {
+      "entropy": 1.340226661413908,
+      "epoch": 2.7168,
+      "grad_norm": 18.375,
+      "learning_rate": 1.698e-05,
+      "loss": 1.3183,
+      "mean_token_accuracy": 0.6692224076017738,
+      "num_tokens": 12086334.0,
+      "step": 850
+    },
+    {
+      "entropy": 1.3257515206933022,
+      "epoch": 2.7488,
+      "grad_norm": 17.75,
+      "learning_rate": 1.718e-05,
+      "loss": 1.3204,
+      "mean_token_accuracy": 0.6661742279306054,
+      "num_tokens": 12227580.0,
+      "step": 860
+    },
+    {
+      "entropy": 1.2998232152312994,
+      "epoch": 2.7808,
+      "grad_norm": 14.0,
+      "learning_rate": 1.7380000000000003e-05,
+      "loss": 1.2841,
+      "mean_token_accuracy": 0.6723095076158643,
+      "num_tokens": 12366568.0,
+      "step": 870
+    },
+    {
+      "entropy": 1.321110926195979,
+      "epoch": 2.8128,
+      "grad_norm": 14.1875,
+      "learning_rate": 1.758e-05,
+      "loss": 1.2809,
+      "mean_token_accuracy": 0.6734412474557757,
+      "num_tokens": 12508616.0,
+      "step": 880
+    },
+    {
+      "entropy": 1.3441704627126456,
+      "epoch": 2.8448,
+      "grad_norm": 14.5625,
+      "learning_rate": 1.7780000000000003e-05,
+      "loss": 1.3264,
+      "mean_token_accuracy": 0.6696716655045748,
+      "num_tokens": 12651142.0,
+      "step": 890
+    },
+    {
+      "entropy": 1.350888105481863,
+      "epoch": 2.8768000000000002,
+      "grad_norm": 13.125,
+      "learning_rate": 1.798e-05,
+      "loss": 1.3305,
+      "mean_token_accuracy": 0.6628765376284719,
+      "num_tokens": 12790697.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.8768000000000002,
+      "eval_coding_entropy": 1.923588216304779,
+      "eval_coding_loss": 2.011129379272461,
+      "eval_coding_mean_token_accuracy": 0.5363702869415283,
+      "eval_coding_num_tokens": 12790697.0,
+      "eval_coding_runtime": 80.767,
+      "eval_coding_samples_per_second": 6.191,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 900
+    },
+    {
+      "epoch": 2.8768000000000002,
+      "eval_math_entropy": 1.3416466426849365,
+      "eval_math_loss": 1.3308217525482178,
+      "eval_math_mean_token_accuracy": 0.6704786785840988,
+      "eval_math_num_tokens": 12790697.0,
+      "eval_math_runtime": 46.0849,
+      "eval_math_samples_per_second": 10.85,
+      "eval_math_steps_per_second": 5.425,
+      "step": 900
+    },
+    {
+      "entropy": 1.3483149357140065,
+      "epoch": 2.9088000000000003,
+      "grad_norm": 11.875,
+      "learning_rate": 1.8180000000000002e-05,
+      "loss": 1.3227,
+      "mean_token_accuracy": 0.6711992308497429,
+      "num_tokens": 12931565.0,
+      "step": 910
+    },
+    {
+      "entropy": 1.2966109592467547,
+      "epoch": 2.9408,
+      "grad_norm": 13.4375,
+      "learning_rate": 1.8380000000000004e-05,
+      "loss": 1.2763,
+      "mean_token_accuracy": 0.6791517404839397,
+      "num_tokens": 13073417.0,
+      "step": 920
+    },
+    {
+      "entropy": 1.3717037633061409,
+      "epoch": 2.9728,
+      "grad_norm": 13.0,
+      "learning_rate": 1.858e-05,
+      "loss": 1.3564,
+      "mean_token_accuracy": 0.6620129646733404,
+      "num_tokens": 13220708.0,
+      "step": 930
+    },
+    {
+      "entropy": 1.3031759885580916,
+      "epoch": 3.0032,
+      "grad_norm": 12.5,
+      "learning_rate": 1.878e-05,
+      "loss": 1.2593,
+      "mean_token_accuracy": 0.6780896206435404,
+      "num_tokens": 13357917.0,
+      "step": 940
+    },
+    {
+      "entropy": 1.2132978197187185,
+      "epoch": 3.0352,
+      "grad_norm": 12.125,
+      "learning_rate": 1.898e-05,
+      "loss": 1.2061,
+      "mean_token_accuracy": 0.6939747478812933,
+      "num_tokens": 13504598.0,
+      "step": 950
+    },
+    {
+      "entropy": 1.2069394052028657,
+      "epoch": 3.0672,
+      "grad_norm": 11.625,
+      "learning_rate": 1.918e-05,
+      "loss": 1.1756,
+      "mean_token_accuracy": 0.695431136712432,
+      "num_tokens": 13654124.0,
+      "step": 960
+    },
+    {
+      "entropy": 1.2009814836084842,
+      "epoch": 3.0992,
+      "grad_norm": 11.0625,
+      "learning_rate": 1.938e-05,
+      "loss": 1.1898,
+      "mean_token_accuracy": 0.6938818197697401,
+      "num_tokens": 13797243.0,
+      "step": 970
+    },
+    {
+      "entropy": 1.2126072812825441,
+      "epoch": 3.1312,
+      "grad_norm": 12.0,
+      "learning_rate": 1.9580000000000002e-05,
+      "loss": 1.1748,
+      "mean_token_accuracy": 0.6926786806434393,
+      "num_tokens": 13944269.0,
+      "step": 980
+    },
+    {
+      "entropy": 1.2055565655231475,
+      "epoch": 3.1632,
+      "grad_norm": 10.75,
+      "learning_rate": 1.978e-05,
+      "loss": 1.1736,
+      "mean_token_accuracy": 0.6961859103292227,
+      "num_tokens": 14084426.0,
+      "step": 990
+    },
+    {
+      "entropy": 1.2099793013185263,
+      "epoch": 3.1952,
+      "grad_norm": 12.8125,
+      "learning_rate": 1.9980000000000002e-05,
+      "loss": 1.1884,
+      "mean_token_accuracy": 0.6956615813076497,
+      "num_tokens": 14225578.0,
+      "step": 1000
+    },
+    {
+      "epoch": 3.1952,
+      "eval_coding_entropy": 1.8568685698509215,
+      "eval_coding_loss": 2.0376784801483154,
+      "eval_coding_mean_token_accuracy": 0.5336039383411407,
+      "eval_coding_num_tokens": 14225578.0,
+      "eval_coding_runtime": 81.5817,
+      "eval_coding_samples_per_second": 6.129,
+      "eval_coding_steps_per_second": 3.064,
+      "step": 1000
+    },
+    {
+      "epoch": 3.1952,
+      "eval_math_entropy": 1.2256325538158417,
+      "eval_math_loss": 1.2980027198791504,
+      "eval_math_mean_token_accuracy": 0.6779666702747345,
+      "eval_math_num_tokens": 14225578.0,
+      "eval_math_runtime": 46.295,
+      "eval_math_samples_per_second": 10.8,
+      "eval_math_steps_per_second": 5.4,
+      "step": 1000
+    },
+    {
+      "entropy": 1.2310455348342657,
+      "epoch": 3.2272,
+      "grad_norm": 11.75,
+      "learning_rate": 1.9980000000000002e-05,
+      "loss": 1.2069,
+      "mean_token_accuracy": 0.6908577650785446,
+      "num_tokens": 14371313.0,
+      "step": 1010
+    },
+    {
+      "entropy": 1.2366832181811334,
+      "epoch": 3.2592,
+      "grad_norm": 9.9375,
+      "learning_rate": 1.995777777777778e-05,
+      "loss": 1.2053,
+      "mean_token_accuracy": 0.6909629859030246,
+      "num_tokens": 14510372.0,
+      "step": 1020
+    },
+    {
+      "entropy": 1.1998371604830027,
+      "epoch": 3.2912,
+      "grad_norm": 11.1875,
+      "learning_rate": 1.9935555555555557e-05,
+      "loss": 1.1928,
+      "mean_token_accuracy": 0.6971310704946518,
+      "num_tokens": 14654048.0,
+      "step": 1030
+    },
+    {
+      "entropy": 1.2052522465586661,
+      "epoch": 3.3232,
+      "grad_norm": 10.6875,
+      "learning_rate": 1.9913333333333335e-05,
+      "loss": 1.1736,
+      "mean_token_accuracy": 0.6949702069163323,
+      "num_tokens": 14799061.0,
+      "step": 1040
+    },
+    {
+      "entropy": 1.2304670121520758,
+      "epoch": 3.3552,
+      "grad_norm": 9.125,
+      "learning_rate": 1.9891111111111112e-05,
+      "loss": 1.208,
+      "mean_token_accuracy": 0.6886732917279005,
+      "num_tokens": 14948615.0,
+      "step": 1050
+    },
+    {
+      "entropy": 1.195398748293519,
+      "epoch": 3.3872,
+      "grad_norm": 10.4375,
+      "learning_rate": 1.986888888888889e-05,
+      "loss": 1.1855,
+      "mean_token_accuracy": 0.7008309479802847,
+      "num_tokens": 15091830.0,
+      "step": 1060
+    },
+    {
+      "entropy": 1.1796185825020076,
+      "epoch": 3.4192,
+      "grad_norm": 11.375,
+      "learning_rate": 1.9846666666666668e-05,
+      "loss": 1.1623,
+      "mean_token_accuracy": 0.7005136132240295,
+      "num_tokens": 15223941.0,
+      "step": 1070
+    },
+    {
+      "entropy": 1.1732951115816832,
+      "epoch": 3.4512,
+      "grad_norm": 13.8125,
+      "learning_rate": 1.9824444444444445e-05,
+      "loss": 1.1437,
+      "mean_token_accuracy": 0.7002243742346763,
+      "num_tokens": 15364119.0,
+      "step": 1080
+    },
+    {
+      "entropy": 1.1816517032682896,
+      "epoch": 3.4832,
+      "grad_norm": 11.125,
+      "learning_rate": 1.9802222222222226e-05,
+      "loss": 1.1425,
+      "mean_token_accuracy": 0.6995426103472709,
+      "num_tokens": 15508933.0,
+      "step": 1090
+    },
+    {
+      "entropy": 1.2135872263461351,
+      "epoch": 3.5152,
+      "grad_norm": 8.75,
+      "learning_rate": 1.978e-05,
+      "loss": 1.1847,
+      "mean_token_accuracy": 0.6952282149344683,
+      "num_tokens": 15654554.0,
+      "step": 1100
+    },
+    {
+      "epoch": 3.5152,
+      "eval_coding_entropy": 1.8463355026245116,
+      "eval_coding_loss": 2.0515615940093994,
+      "eval_coding_mean_token_accuracy": 0.5331288994550705,
+      "eval_coding_num_tokens": 15654554.0,
+      "eval_coding_runtime": 80.9144,
+      "eval_coding_samples_per_second": 6.179,
+      "eval_coding_steps_per_second": 3.09,
+      "step": 1100
+    },
+    {
+      "epoch": 3.5152,
+      "eval_math_entropy": 1.2076501297950744,
+      "eval_math_loss": 1.240770697593689,
+      "eval_math_mean_token_accuracy": 0.6900603187084198,
+      "eval_math_num_tokens": 15654554.0,
+      "eval_math_runtime": 46.9328,
+      "eval_math_samples_per_second": 10.654,
+      "eval_math_steps_per_second": 5.327,
+      "step": 1100
+    },
+    {
+      "entropy": 1.212555281817913,
+      "epoch": 3.5472,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.975777777777778e-05,
+      "loss": 1.1833,
+      "mean_token_accuracy": 0.6934925394132734,
+      "num_tokens": 15793364.0,
+      "step": 1110
+    },
+    {
+      "entropy": 1.207034119591117,
+      "epoch": 3.5792,
+      "grad_norm": 10.875,
+      "learning_rate": 1.9735555555555556e-05,
+      "loss": 1.1711,
+      "mean_token_accuracy": 0.697374514490366,
+      "num_tokens": 15935657.0,
+      "step": 1120
+    },
+    {
+      "entropy": 1.2222992777824402,
+      "epoch": 3.6112,
+      "grad_norm": 9.9375,
+      "learning_rate": 1.9713333333333337e-05,
+      "loss": 1.2014,
+      "mean_token_accuracy": 0.6935698345303536,
+      "num_tokens": 16077275.0,
+      "step": 1130
+    },
+    {
+      "entropy": 1.1657866805791854,
+      "epoch": 3.6432,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.969111111111111e-05,
+      "loss": 1.1546,
+      "mean_token_accuracy": 0.7033114738762378,
+      "num_tokens": 16222318.0,
+      "step": 1140
+    },
+    {
+      "entropy": 1.1633354667574167,
+      "epoch": 3.6752000000000002,
+      "grad_norm": 9.25,
+      "learning_rate": 1.9668888888888892e-05,
+      "loss": 1.1251,
+      "mean_token_accuracy": 0.7043058726936579,
+      "num_tokens": 16361855.0,
+      "step": 1150
+    },
+    {
+      "entropy": 1.1620853256434203,
+      "epoch": 3.7072000000000003,
+      "grad_norm": 9.75,
+      "learning_rate": 1.9646666666666666e-05,
+      "loss": 1.1342,
+      "mean_token_accuracy": 0.7022159334272147,
+      "num_tokens": 16500375.0,
+      "step": 1160
+    },
+    {
+      "entropy": 1.1738709833472967,
+      "epoch": 3.7392,
+      "grad_norm": 10.0625,
+      "learning_rate": 1.9624444444444447e-05,
+      "loss": 1.1561,
+      "mean_token_accuracy": 0.7008879505097866,
+      "num_tokens": 16643092.0,
+      "step": 1170
+    },
+    {
+      "entropy": 1.1810582045465707,
+      "epoch": 3.7712,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.9602222222222225e-05,
+      "loss": 1.1661,
+      "mean_token_accuracy": 0.6994068440049886,
+      "num_tokens": 16789527.0,
+      "step": 1180
+    },
+    {
+      "entropy": 1.1981235884130002,
+      "epoch": 3.8032,
+      "grad_norm": 9.25,
+      "learning_rate": 1.9580000000000002e-05,
+      "loss": 1.184,
+      "mean_token_accuracy": 0.7009957425296307,
+      "num_tokens": 16930493.0,
+      "step": 1190
+    },
+    {
+      "entropy": 1.1709371741861105,
+      "epoch": 3.8352,
+      "grad_norm": 9.25,
+      "learning_rate": 1.955777777777778e-05,
+      "loss": 1.1328,
+      "mean_token_accuracy": 0.7039620807394386,
+      "num_tokens": 17066856.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.8352,
+      "eval_coding_entropy": 1.7909702219963073,
+      "eval_coding_loss": 2.0417468547821045,
+      "eval_coding_mean_token_accuracy": 0.5346206701993942,
+      "eval_coding_num_tokens": 17066856.0,
+      "eval_coding_runtime": 81.4914,
+      "eval_coding_samples_per_second": 6.136,
+      "eval_coding_steps_per_second": 3.068,
+      "step": 1200
+    },
+    {
+      "epoch": 3.8352,
+      "eval_math_entropy": 1.1532045022249222,
+      "eval_math_loss": 1.2045187950134277,
+      "eval_math_mean_token_accuracy": 0.6980784633159638,
+      "eval_math_num_tokens": 17066856.0,
+      "eval_math_runtime": 46.8505,
+      "eval_math_samples_per_second": 10.672,
+      "eval_math_steps_per_second": 5.336,
+      "step": 1200
+    },
+    {
+      "entropy": 1.156894364207983,
+      "epoch": 3.8672,
+      "grad_norm": 9.0625,
+      "learning_rate": 1.9535555555555557e-05,
+      "loss": 1.1334,
+      "mean_token_accuracy": 0.7076525427401066,
+      "num_tokens": 17205604.0,
+      "step": 1210
+    },
+    {
+      "entropy": 1.174615489691496,
+      "epoch": 3.8992,
+      "grad_norm": 8.375,
+      "learning_rate": 1.9513333333333335e-05,
+      "loss": 1.165,
+      "mean_token_accuracy": 0.6988233763724565,
+      "num_tokens": 17348293.0,
+      "step": 1220
+    },
+    {
+      "entropy": 1.1527796313166618,
+      "epoch": 3.9312,
+      "grad_norm": 9.3125,
+      "learning_rate": 1.9491111111111113e-05,
+      "loss": 1.1257,
+      "mean_token_accuracy": 0.707377451658249,
+      "num_tokens": 17491463.0,
+      "step": 1230
+    },
+    {
+      "entropy": 1.1771091677248477,
+      "epoch": 3.9632,
+      "grad_norm": 10.0625,
+      "learning_rate": 1.946888888888889e-05,
+      "loss": 1.1595,
+      "mean_token_accuracy": 0.7032928232103586,
+      "num_tokens": 17628099.0,
+      "step": 1240
+    },
+    {
+      "entropy": 1.1742881268262864,
+      "epoch": 3.9952,
+      "grad_norm": 10.5625,
+      "learning_rate": 1.9446666666666668e-05,
+      "loss": 1.1505,
+      "mean_token_accuracy": 0.7026802890002728,
+      "num_tokens": 17768068.0,
+      "step": 1250
+    },
+    {
+      "entropy": 1.0864369343770177,
+      "epoch": 4.0256,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.9424444444444446e-05,
+      "loss": 1.0635,
+      "mean_token_accuracy": 0.7234101566044908,
+      "num_tokens": 17904915.0,
+      "step": 1260
+    },
+    {
+      "entropy": 1.0292031805962325,
+      "epoch": 4.0576,
+      "grad_norm": 10.375,
+      "learning_rate": 1.9402222222222223e-05,
+      "loss": 0.9885,
+      "mean_token_accuracy": 0.7318673226982355,
+      "num_tokens": 18050621.0,
+      "step": 1270
+    },
+    {
+      "entropy": 1.0440782070159913,
+      "epoch": 4.0896,
+      "grad_norm": 10.75,
+      "learning_rate": 1.938e-05,
+      "loss": 1.0119,
+      "mean_token_accuracy": 0.7287190053611994,
+      "num_tokens": 18195139.0,
+      "step": 1280
+    },
+    {
+      "entropy": 1.0936193302273751,
+      "epoch": 4.1216,
+      "grad_norm": 9.5,
+      "learning_rate": 1.935777777777778e-05,
+      "loss": 1.0587,
+      "mean_token_accuracy": 0.7225902825593948,
+      "num_tokens": 18335483.0,
+      "step": 1290
+    },
+    {
+      "entropy": 1.0716597214341164,
+      "epoch": 4.1536,
+      "grad_norm": 8.0,
+      "learning_rate": 1.9335555555555556e-05,
+      "loss": 1.0521,
+      "mean_token_accuracy": 0.7205776769667864,
+      "num_tokens": 18471464.0,
+      "step": 1300
+    },
+    {
+      "epoch": 4.1536,
+      "eval_coding_entropy": 1.666171245574951,
+      "eval_coding_loss": 2.089735984802246,
+      "eval_coding_mean_token_accuracy": 0.5322686601877212,
+      "eval_coding_num_tokens": 18471464.0,
+      "eval_coding_runtime": 81.1644,
+      "eval_coding_samples_per_second": 6.16,
+      "eval_coding_steps_per_second": 3.08,
+      "step": 1300
+    },
+    {
+      "epoch": 4.1536,
+      "eval_math_entropy": 1.060988509297371,
+      "eval_math_loss": 1.184299111366272,
+      "eval_math_mean_token_accuracy": 0.7041985726356507,
+      "eval_math_num_tokens": 18471464.0,
+      "eval_math_runtime": 46.7403,
+      "eval_math_samples_per_second": 10.697,
+      "eval_math_steps_per_second": 5.349,
+      "step": 1300
+    },
+    {
+      "entropy": 1.042165393382311,
+      "epoch": 4.1856,
+      "grad_norm": 8.375,
+      "learning_rate": 1.9313333333333334e-05,
+      "loss": 1.0117,
+      "mean_token_accuracy": 0.7323273476213217,
+      "num_tokens": 18613144.0,
+      "step": 1310
+    },
+    {
+      "entropy": 1.0549738246947526,
+      "epoch": 4.2176,
+      "grad_norm": 12.0625,
+      "learning_rate": 1.9291111111111115e-05,
+      "loss": 1.0356,
+      "mean_token_accuracy": 0.7262615174055099,
+      "num_tokens": 18757354.0,
+      "step": 1320
+    },
+    {
+      "entropy": 0.9945293262600898,
+      "epoch": 4.2496,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.926888888888889e-05,
+      "loss": 0.9893,
+      "mean_token_accuracy": 0.737324832752347,
+      "num_tokens": 18901730.0,
+      "step": 1330
+    },
+    {
+      "entropy": 1.0042558293789625,
+      "epoch": 4.2816,
+      "grad_norm": 8.1875,
+      "learning_rate": 1.924666666666667e-05,
+      "loss": 0.976,
+      "mean_token_accuracy": 0.7383205603808165,
+      "num_tokens": 19044725.0,
+      "step": 1340
+    },
+    {
+      "entropy": 1.0393657265231013,
+      "epoch": 4.3136,
+      "grad_norm": 8.625,
+      "learning_rate": 1.9224444444444444e-05,
+      "loss": 1.0086,
+      "mean_token_accuracy": 0.7331446453928947,
+      "num_tokens": 19193912.0,
+      "step": 1350
+    },
+    {
+      "entropy": 1.0386454924941062,
+      "epoch": 4.3456,
+      "grad_norm": 9.0,
+      "learning_rate": 1.9202222222222225e-05,
+      "loss": 1.0083,
+      "mean_token_accuracy": 0.7298929043114185,
+      "num_tokens": 19336463.0,
+      "step": 1360
+    },
+    {
+      "entropy": 1.0017262276262044,
+      "epoch": 4.3776,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.918e-05,
+      "loss": 0.979,
+      "mean_token_accuracy": 0.7369209341704845,
+      "num_tokens": 19482758.0,
+      "step": 1370
+    },
+    {
+      "entropy": 1.0357928197830915,
+      "epoch": 4.4096,
+      "grad_norm": 11.25,
+      "learning_rate": 1.915777777777778e-05,
+      "loss": 1.0003,
+      "mean_token_accuracy": 0.7308139387518168,
+      "num_tokens": 19624149.0,
+      "step": 1380
+    },
+    {
+      "entropy": 1.0128012642264366,
+      "epoch": 4.4416,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.9135555555555555e-05,
+      "loss": 0.994,
+      "mean_token_accuracy": 0.7347184054553508,
+      "num_tokens": 19762641.0,
+      "step": 1390
+    },
+    {
+      "entropy": 1.0030905462801456,
+      "epoch": 4.4736,
+      "grad_norm": 7.96875,
+      "learning_rate": 1.9113333333333336e-05,
+      "loss": 0.9852,
+      "mean_token_accuracy": 0.7376307789236307,
+      "num_tokens": 19906714.0,
+      "step": 1400
+    },
+    {
+      "epoch": 4.4736,
+      "eval_coding_entropy": 1.6414423942565919,
+      "eval_coding_loss": 2.086019277572632,
+      "eval_coding_mean_token_accuracy": 0.533721309542656,
+      "eval_coding_num_tokens": 19906714.0,
+      "eval_coding_runtime": 81.4357,
+      "eval_coding_samples_per_second": 6.14,
+      "eval_coding_steps_per_second": 3.07,
+      "step": 1400
+    },
+    {
+      "epoch": 4.4736,
+      "eval_math_entropy": 1.023936753153801,
+      "eval_math_loss": 1.1772490739822388,
+      "eval_math_mean_token_accuracy": 0.707464236497879,
+      "eval_math_num_tokens": 19906714.0,
+      "eval_math_runtime": 46.7296,
+      "eval_math_samples_per_second": 10.7,
+      "eval_math_steps_per_second": 5.35,
+      "step": 1400
+    },
+    {
+      "entropy": 1.0678872428834438,
+      "epoch": 4.5056,
+      "grad_norm": 8.5,
+      "learning_rate": 1.9091111111111113e-05,
+      "loss": 1.0421,
+      "mean_token_accuracy": 0.7248407501727343,
+      "num_tokens": 20042824.0,
+      "step": 1410
+    },
+    {
+      "entropy": 1.0317748140543699,
+      "epoch": 4.5376,
+      "grad_norm": 7.28125,
+      "learning_rate": 1.906888888888889e-05,
+      "loss": 1.0118,
+      "mean_token_accuracy": 0.7290458243340254,
+      "num_tokens": 20184198.0,
+      "step": 1420
+    },
+    {
+      "entropy": 1.0234299018979072,
+      "epoch": 4.5696,
+      "grad_norm": 8.1875,
+      "learning_rate": 1.904666666666667e-05,
+      "loss": 0.9974,
+      "mean_token_accuracy": 0.734089670330286,
+      "num_tokens": 20326769.0,
+      "step": 1430
+    },
+    {
+      "entropy": 1.0432322915643453,
+      "epoch": 4.6016,
+      "grad_norm": 10.4375,
+      "learning_rate": 1.9024444444444446e-05,
+      "loss": 1.0359,
+      "mean_token_accuracy": 0.7288251578807831,
+      "num_tokens": 20463248.0,
+      "step": 1440
+    },
+    {
+      "entropy": 0.9960782427340746,
+      "epoch": 4.6336,
+      "grad_norm": 7.75,
+      "learning_rate": 1.9002222222222224e-05,
+      "loss": 0.9673,
+      "mean_token_accuracy": 0.7419588100165129,
+      "num_tokens": 20606156.0,
+      "step": 1450
+    },
+    {
+      "entropy": 1.0214475937187673,
+      "epoch": 4.6655999999999995,
+      "grad_norm": 9.75,
+      "learning_rate": 1.898e-05,
+      "loss": 1.0073,
+      "mean_token_accuracy": 0.7355681076645851,
+      "num_tokens": 20742663.0,
+      "step": 1460
+    },
+    {
+      "entropy": 1.042779802531004,
+      "epoch": 4.6975999999999996,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.895777777777778e-05,
+      "loss": 1.015,
+      "mean_token_accuracy": 0.7308492448180914,
+      "num_tokens": 20888422.0,
+      "step": 1470
+    },
+    {
+      "entropy": 0.9938499752432108,
+      "epoch": 4.7296,
+      "grad_norm": 8.4375,
+      "learning_rate": 1.8935555555555556e-05,
+      "loss": 0.9681,
+      "mean_token_accuracy": 0.7414834558963775,
+      "num_tokens": 21032183.0,
+      "step": 1480
+    },
+    {
+      "entropy": 1.016091812774539,
+      "epoch": 4.7616,
+      "grad_norm": 10.3125,
+      "learning_rate": 1.8913333333333334e-05,
+      "loss": 0.9861,
+      "mean_token_accuracy": 0.7363665714859963,
+      "num_tokens": 21171363.0,
+      "step": 1490
+    },
+    {
+      "entropy": 1.0135044915601612,
+      "epoch": 4.7936,
+      "grad_norm": 7.5625,
+      "learning_rate": 1.8891111111111115e-05,
+      "loss": 0.9953,
+      "mean_token_accuracy": 0.7363950561732053,
+      "num_tokens": 21316124.0,
+      "step": 1500
+    },
+    {
+      "epoch": 4.7936,
+      "eval_coding_entropy": 1.6775614714622498,
+      "eval_coding_loss": 2.1003665924072266,
+      "eval_coding_mean_token_accuracy": 0.5311660279035568,
+      "eval_coding_num_tokens": 21316124.0,
+      "eval_coding_runtime": 82.243,
+      "eval_coding_samples_per_second": 6.08,
+      "eval_coding_steps_per_second": 3.04,
+      "step": 1500
+    },
+    {
+      "epoch": 4.7936,
+      "eval_math_entropy": 1.0237463138103484,
+      "eval_math_loss": 1.1396057605743408,
+      "eval_math_mean_token_accuracy": 0.715588841676712,
+      "eval_math_num_tokens": 21316124.0,
+      "eval_math_runtime": 47.1978,
+      "eval_math_samples_per_second": 10.594,
+      "eval_math_steps_per_second": 5.297,
+      "step": 1500
+    },
+    {
+      "entropy": 1.0302676513791085,
+      "epoch": 4.8256,
+      "grad_norm": 7.5625,
+      "learning_rate": 1.886888888888889e-05,
+      "loss": 1.0034,
+      "mean_token_accuracy": 0.7323874134570361,
+      "num_tokens": 21456811.0,
+      "step": 1510
+    },
+    {
+      "entropy": 1.0087147317826748,
+      "epoch": 4.8576,
+      "grad_norm": 10.375,
+      "learning_rate": 1.884666666666667e-05,
+      "loss": 0.9966,
+      "mean_token_accuracy": 0.7373051200062036,
+      "num_tokens": 21598174.0,
+      "step": 1520
+    },
+    {
+      "entropy": 1.0385818116366863,
+      "epoch": 4.8896,
+      "grad_norm": 8.75,
+      "learning_rate": 1.8824444444444445e-05,
+      "loss": 1.0234,
+      "mean_token_accuracy": 0.7324004743248225,
+      "num_tokens": 21749526.0,
+      "step": 1530
+    },
+    {
+      "entropy": 1.0246109385043383,
+      "epoch": 4.9216,
+      "grad_norm": 8.125,
+      "learning_rate": 1.8802222222222226e-05,
+      "loss": 0.9977,
+      "mean_token_accuracy": 0.7329407773911953,
+      "num_tokens": 21885800.0,
+      "step": 1540
+    },
+    {
+      "entropy": 1.005864929407835,
+      "epoch": 4.9536,
+      "grad_norm": 9.4375,
+      "learning_rate": 1.878e-05,
+      "loss": 0.9762,
+      "mean_token_accuracy": 0.7406043078750372,
+      "num_tokens": 22028858.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.9871604397892952,
+      "epoch": 4.9856,
+      "grad_norm": 7.84375,
+      "learning_rate": 1.875777777777778e-05,
+      "loss": 0.9605,
+      "mean_token_accuracy": 0.7413943439722062,
+      "num_tokens": 22175575.0,
+      "step": 1560
+    },
+    {
+      "entropy": 0.9552774119533991,
+      "epoch": 5.016,
+      "grad_norm": 7.96875,
+      "learning_rate": 1.873555555555556e-05,
+      "loss": 0.9271,
+      "mean_token_accuracy": 0.7519231669996914,
+      "num_tokens": 22313061.0,
+      "step": 1570
+    },
+    {
+      "entropy": 0.908423101902008,
+      "epoch": 5.048,
+      "grad_norm": 6.65625,
+      "learning_rate": 1.8713333333333336e-05,
+      "loss": 0.8928,
+      "mean_token_accuracy": 0.7576735183596611,
+      "num_tokens": 22457889.0,
+      "step": 1580
+    },
+    {
+      "entropy": 0.9394706435501575,
+      "epoch": 5.08,
+      "grad_norm": 7.5,
+      "learning_rate": 1.8691111111111114e-05,
+      "loss": 0.9083,
+      "mean_token_accuracy": 0.7530260667204857,
+      "num_tokens": 22594618.0,
+      "step": 1590
+    },
+    {
+      "entropy": 0.9052924428135156,
+      "epoch": 5.112,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.866888888888889e-05,
+      "loss": 0.876,
+      "mean_token_accuracy": 0.7579276110976935,
+      "num_tokens": 22735034.0,
+      "step": 1600
+    },
+    {
+      "epoch": 5.112,
+      "eval_coding_entropy": 1.5108720426559448,
+      "eval_coding_loss": 2.168074131011963,
+      "eval_coding_mean_token_accuracy": 0.5293694581985474,
+      "eval_coding_num_tokens": 22735034.0,
+      "eval_coding_runtime": 80.6791,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.099,
+      "step": 1600
+    },
+    {
+      "epoch": 5.112,
+      "eval_math_entropy": 0.9210035672187805,
+      "eval_math_loss": 1.1561440229415894,
+      "eval_math_mean_token_accuracy": 0.7158858454227448,
+      "eval_math_num_tokens": 22735034.0,
+      "eval_math_runtime": 46.0157,
+      "eval_math_samples_per_second": 10.866,
+      "eval_math_steps_per_second": 5.433,
+      "step": 1600
+    },
+    {
+      "entropy": 0.8932699026539922,
+      "epoch": 5.144,
+      "grad_norm": 6.5,
+      "learning_rate": 1.864666666666667e-05,
+      "loss": 0.8714,
+      "mean_token_accuracy": 0.762346514314413,
+      "num_tokens": 22875863.0,
+      "step": 1610
+    },
+    {
+      "entropy": 0.8730874558910727,
+      "epoch": 5.176,
+      "grad_norm": 7.3125,
+      "learning_rate": 1.8624444444444446e-05,
+      "loss": 0.8248,
+      "mean_token_accuracy": 0.767206709459424,
+      "num_tokens": 23019827.0,
+      "step": 1620
+    },
+    {
+      "entropy": 0.9459875440225005,
+      "epoch": 5.208,
+      "grad_norm": 7.625,
+      "learning_rate": 1.8602222222222224e-05,
+      "loss": 0.9073,
+      "mean_token_accuracy": 0.7505842503160238,
+      "num_tokens": 23160295.0,
+      "step": 1630
+    },
+    {
+      "entropy": 0.8937128365039826,
+      "epoch": 5.24,
+      "grad_norm": 6.375,
+      "learning_rate": 1.858e-05,
+      "loss": 0.87,
+      "mean_token_accuracy": 0.7625281598418951,
+      "num_tokens": 23302273.0,
+      "step": 1640
+    },
+    {
+      "entropy": 0.8976007211953402,
+      "epoch": 5.272,
+      "grad_norm": 6.75,
+      "learning_rate": 1.855777777777778e-05,
+      "loss": 0.8753,
+      "mean_token_accuracy": 0.7627121699973941,
+      "num_tokens": 23449683.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.8997082270681858,
+      "epoch": 5.304,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.8535555555555557e-05,
+      "loss": 0.888,
+      "mean_token_accuracy": 0.7588404752314091,
+      "num_tokens": 23588004.0,
+      "step": 1660
+    },
+    {
+      "entropy": 0.9175663651898504,
+      "epoch": 5.336,
+      "grad_norm": 7.5,
+      "learning_rate": 1.8513333333333335e-05,
+      "loss": 0.8712,
+      "mean_token_accuracy": 0.7588019911199808,
+      "num_tokens": 23735695.0,
+      "step": 1670
+    },
+    {
+      "entropy": 0.8917087573558092,
+      "epoch": 5.368,
+      "grad_norm": 6.75,
+      "learning_rate": 1.8491111111111112e-05,
+      "loss": 0.872,
+      "mean_token_accuracy": 0.7614140216261148,
+      "num_tokens": 23878624.0,
+      "step": 1680
+    },
+    {
+      "entropy": 0.9097595578059554,
+      "epoch": 5.4,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.846888888888889e-05,
+      "loss": 0.8898,
+      "mean_token_accuracy": 0.7600292358547449,
+      "num_tokens": 24021664.0,
+      "step": 1690
+    },
+    {
+      "entropy": 0.8914705088362098,
+      "epoch": 5.432,
+      "grad_norm": 8.75,
+      "learning_rate": 1.8446666666666667e-05,
+      "loss": 0.8629,
+      "mean_token_accuracy": 0.7618724539875984,
+      "num_tokens": 24162453.0,
+      "step": 1700
+    },
+    {
+      "epoch": 5.432,
+      "eval_coding_entropy": 1.5077661447525024,
+      "eval_coding_loss": 2.1740472316741943,
+      "eval_coding_mean_token_accuracy": 0.5293574998378754,
+      "eval_coding_num_tokens": 24162453.0,
+      "eval_coding_runtime": 81.2553,
+      "eval_coding_samples_per_second": 6.153,
+      "eval_coding_steps_per_second": 3.077,
+      "step": 1700
+    },
+    {
+      "epoch": 5.432,
+      "eval_math_entropy": 0.907505418419838,
+      "eval_math_loss": 1.1403419971466064,
+      "eval_math_mean_token_accuracy": 0.7196878392696381,
+      "eval_math_num_tokens": 24162453.0,
+      "eval_math_runtime": 46.3752,
+      "eval_math_samples_per_second": 10.782,
+      "eval_math_steps_per_second": 5.391,
+      "step": 1700
+    },
+    {
+      "entropy": 0.8794696152210235,
+      "epoch": 5.464,
+      "grad_norm": 7.8125,
+      "learning_rate": 1.842444444444445e-05,
+      "loss": 0.8491,
+      "mean_token_accuracy": 0.7637434259057045,
+      "num_tokens": 24302751.0,
+      "step": 1710
+    },
+    {
+      "entropy": 0.8770243087783456,
+      "epoch": 5.496,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.8402222222222223e-05,
+      "loss": 0.8569,
+      "mean_token_accuracy": 0.7662156637758016,
+      "num_tokens": 24447316.0,
+      "step": 1720
+    },
+    {
+      "entropy": 0.9151918144896627,
+      "epoch": 5.5280000000000005,
+      "grad_norm": 7.9375,
+      "learning_rate": 1.8380000000000004e-05,
+      "loss": 0.8896,
+      "mean_token_accuracy": 0.7570474561303854,
+      "num_tokens": 24588812.0,
+      "step": 1730
+    },
+    {
+      "entropy": 0.9098712190985679,
+      "epoch": 5.5600000000000005,
+      "grad_norm": 6.96875,
+      "learning_rate": 1.8357777777777778e-05,
+      "loss": 0.8752,
+      "mean_token_accuracy": 0.7589137274771929,
+      "num_tokens": 24729042.0,
+      "step": 1740
+    },
+    {
+      "entropy": 0.9058109503239393,
+      "epoch": 5.592,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.833555555555556e-05,
+      "loss": 0.8724,
+      "mean_token_accuracy": 0.7591247230768203,
+      "num_tokens": 24861573.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.8967058792710304,
+      "epoch": 5.624,
+      "grad_norm": 7.65625,
+      "learning_rate": 1.8313333333333333e-05,
+      "loss": 0.8772,
+      "mean_token_accuracy": 0.7598790679126978,
+      "num_tokens": 25009913.0,
+      "step": 1760
+    },
+    {
+      "entropy": 0.8895885275676847,
+      "epoch": 5.656,
+      "grad_norm": 7.125,
+      "learning_rate": 1.8291111111111114e-05,
+      "loss": 0.8706,
+      "mean_token_accuracy": 0.7640030801296234,
+      "num_tokens": 25148190.0,
+      "step": 1770
+    },
+    {
+      "entropy": 0.9247543184086681,
+      "epoch": 5.688,
+      "grad_norm": 7.75,
+      "learning_rate": 1.8268888888888888e-05,
+      "loss": 0.899,
+      "mean_token_accuracy": 0.7559223342686892,
+      "num_tokens": 25291745.0,
+      "step": 1780
+    },
+    {
+      "entropy": 0.9217993386089802,
+      "epoch": 5.72,
+      "grad_norm": 7.8125,
+      "learning_rate": 1.824666666666667e-05,
+      "loss": 0.895,
+      "mean_token_accuracy": 0.7565487738698721,
+      "num_tokens": 25434441.0,
+      "step": 1790
+    },
+    {
+      "entropy": 0.8966394849121571,
+      "epoch": 5.752,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.8224444444444447e-05,
+      "loss": 0.8758,
+      "mean_token_accuracy": 0.7607072979211807,
+      "num_tokens": 25573574.0,
+      "step": 1800
+    },
+    {
+      "epoch": 5.752,
+      "eval_coding_entropy": 1.5287488384246826,
+      "eval_coding_loss": 2.1693997383117676,
+      "eval_coding_mean_token_accuracy": 0.5300438596010209,
+      "eval_coding_num_tokens": 25573574.0,
+      "eval_coding_runtime": 81.0569,
+      "eval_coding_samples_per_second": 6.169,
+      "eval_coding_steps_per_second": 3.084,
+      "step": 1800
+    },
+    {
+      "epoch": 5.752,
+      "eval_math_entropy": 0.9324754639863968,
+      "eval_math_loss": 1.1225923299789429,
+      "eval_math_mean_token_accuracy": 0.7223285484313965,
+      "eval_math_num_tokens": 25573574.0,
+      "eval_math_runtime": 46.1182,
+      "eval_math_samples_per_second": 10.842,
+      "eval_math_steps_per_second": 5.421,
+      "step": 1800
+    },
+    {
+      "entropy": 0.8990588784217834,
+      "epoch": 5.784,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.8202222222222225e-05,
+      "loss": 0.8611,
+      "mean_token_accuracy": 0.7620632901787758,
+      "num_tokens": 25720633.0,
+      "step": 1810
+    },
+    {
+      "entropy": 0.9017492324113846,
+      "epoch": 5.816,
+      "grad_norm": 7.0,
+      "learning_rate": 1.8180000000000002e-05,
+      "loss": 0.8795,
+      "mean_token_accuracy": 0.7607263036072254,
+      "num_tokens": 25860386.0,
+      "step": 1820
+    },
+    {
+      "entropy": 0.8764655807986855,
+      "epoch": 5.848,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.815777777777778e-05,
+      "loss": 0.8454,
+      "mean_token_accuracy": 0.7668800290673972,
+      "num_tokens": 26003831.0,
+      "step": 1830
+    },
+    {
+      "entropy": 0.9092245759442449,
+      "epoch": 5.88,
+      "grad_norm": 7.28125,
+      "learning_rate": 1.8135555555555557e-05,
+      "loss": 0.9039,
+      "mean_token_accuracy": 0.7570795770734549,
+      "num_tokens": 26141832.0,
+      "step": 1840
+    },
+    {
+      "entropy": 0.9221448406577111,
+      "epoch": 5.912,
+      "grad_norm": 7.6875,
+      "learning_rate": 1.8113333333333335e-05,
+      "loss": 0.889,
+      "mean_token_accuracy": 0.7561331510543823,
+      "num_tokens": 26285027.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.8948194550350308,
+      "epoch": 5.944,
+      "grad_norm": 8.6875,
+      "learning_rate": 1.8091111111111113e-05,
+      "loss": 0.8728,
+      "mean_token_accuracy": 0.760323528200388,
+      "num_tokens": 26432812.0,
+      "step": 1860
+    },
+    {
+      "entropy": 0.8912792162969708,
+      "epoch": 5.976,
+      "grad_norm": 7.4375,
+      "learning_rate": 1.806888888888889e-05,
+      "loss": 0.8637,
+      "mean_token_accuracy": 0.7622829657047987,
+      "num_tokens": 26576198.0,
+      "step": 1870
+    },
+    {
+      "entropy": 0.8932076808261243,
+      "epoch": 6.0064,
+      "grad_norm": 6.59375,
+      "learning_rate": 1.8046666666666668e-05,
+      "loss": 0.8527,
+      "mean_token_accuracy": 0.7719342869363333,
+      "num_tokens": 26714418.0,
+      "step": 1880
+    },
+    {
+      "entropy": 0.7547721007838846,
+      "epoch": 6.0384,
+      "grad_norm": 7.78125,
+      "learning_rate": 1.8024444444444445e-05,
+      "loss": 0.7466,
+      "mean_token_accuracy": 0.7900280602276325,
+      "num_tokens": 26854471.0,
+      "step": 1890
+    },
+    {
+      "entropy": 0.8012739174067974,
+      "epoch": 6.0704,
+      "grad_norm": 6.9375,
+      "learning_rate": 1.8002222222222223e-05,
+      "loss": 0.7694,
+      "mean_token_accuracy": 0.7837659023702145,
+      "num_tokens": 26997548.0,
+      "step": 1900
+    },
+    {
+      "epoch": 6.0704,
+      "eval_coding_entropy": 1.4375424737930298,
+      "eval_coding_loss": 2.2412030696868896,
+      "eval_coding_mean_token_accuracy": 0.5258033001422882,
+      "eval_coding_num_tokens": 26997548.0,
+      "eval_coding_runtime": 80.7826,
+      "eval_coding_samples_per_second": 6.189,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 1900
+    },
+    {
+      "epoch": 6.0704,
+      "eval_math_entropy": 0.8488073872327805,
+      "eval_math_loss": 1.1494157314300537,
+      "eval_math_mean_token_accuracy": 0.7230218532085418,
+      "eval_math_num_tokens": 26997548.0,
+      "eval_math_runtime": 46.1166,
+      "eval_math_samples_per_second": 10.842,
+      "eval_math_steps_per_second": 5.421,
+      "step": 1900
+    },
+    {
+      "entropy": 0.7635609801858664,
+      "epoch": 6.1024,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.798e-05,
+      "loss": 0.7384,
+      "mean_token_accuracy": 0.7916049119085073,
+      "num_tokens": 27142854.0,
+      "step": 1910
+    },
+    {
+      "entropy": 0.756542045250535,
+      "epoch": 6.1344,
+      "grad_norm": 7.84375,
+      "learning_rate": 1.7957777777777778e-05,
+      "loss": 0.7285,
+      "mean_token_accuracy": 0.7928035575896502,
+      "num_tokens": 27287562.0,
+      "step": 1920
+    },
+    {
+      "entropy": 0.7951679166406393,
+      "epoch": 6.1664,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.7935555555555556e-05,
+      "loss": 0.7664,
+      "mean_token_accuracy": 0.7882673852145672,
+      "num_tokens": 27424818.0,
+      "step": 1930
+    },
+    {
+      "entropy": 0.7970327842980623,
+      "epoch": 6.1984,
+      "grad_norm": 7.75,
+      "learning_rate": 1.7913333333333337e-05,
+      "loss": 0.7689,
+      "mean_token_accuracy": 0.7828843638300895,
+      "num_tokens": 27562593.0,
+      "step": 1940
+    },
+    {
+      "entropy": 0.7904769157990813,
+      "epoch": 6.2304,
+      "grad_norm": 6.84375,
+      "learning_rate": 1.789111111111111e-05,
+      "loss": 0.7573,
+      "mean_token_accuracy": 0.7857714165002108,
+      "num_tokens": 27704822.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.7835376903414726,
+      "epoch": 6.2624,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.7868888888888892e-05,
+      "loss": 0.7641,
+      "mean_token_accuracy": 0.7862238857895136,
+      "num_tokens": 27850462.0,
+      "step": 1960
+    },
+    {
+      "entropy": 0.7813543854281306,
+      "epoch": 6.2943999999999996,
+      "grad_norm": 6.09375,
+      "learning_rate": 1.7846666666666666e-05,
+      "loss": 0.7619,
+      "mean_token_accuracy": 0.7848350748419761,
+      "num_tokens": 27995867.0,
+      "step": 1970
+    },
+    {
+      "entropy": 0.7751486446708441,
+      "epoch": 6.3264,
+      "grad_norm": 6.53125,
+      "learning_rate": 1.7824444444444447e-05,
+      "loss": 0.7454,
+      "mean_token_accuracy": 0.7895105637609958,
+      "num_tokens": 28141531.0,
+      "step": 1980
+    },
+    {
+      "entropy": 0.8145087849348783,
+      "epoch": 6.3584,
+      "grad_norm": 7.1875,
+      "learning_rate": 1.780222222222222e-05,
+      "loss": 0.7932,
+      "mean_token_accuracy": 0.7784016776829958,
+      "num_tokens": 28283236.0,
+      "step": 1990
+    },
+    {
+      "entropy": 0.7936712250113487,
+      "epoch": 6.3904,
+      "grad_norm": 7.5625,
+      "learning_rate": 1.7780000000000003e-05,
+      "loss": 0.7684,
+      "mean_token_accuracy": 0.7840494114905596,
+      "num_tokens": 28424091.0,
+      "step": 2000
+    },
+    {
+      "epoch": 6.3904,
+      "eval_coding_entropy": 1.3973456344604491,
+      "eval_coding_loss": 2.2944459915161133,
+      "eval_coding_mean_token_accuracy": 0.523113137125969,
+      "eval_coding_num_tokens": 28424091.0,
+      "eval_coding_runtime": 80.7861,
+      "eval_coding_samples_per_second": 6.189,
+      "eval_coding_steps_per_second": 3.095,
+      "step": 2000
+    },
+    {
+      "epoch": 6.3904,
+      "eval_math_entropy": 0.8243015093803405,
+      "eval_math_loss": 1.1564151048660278,
+      "eval_math_mean_token_accuracy": 0.7230146683454514,
+      "eval_math_num_tokens": 28424091.0,
+      "eval_math_runtime": 45.6525,
+      "eval_math_samples_per_second": 10.952,
+      "eval_math_steps_per_second": 5.476,
+      "step": 2000
+    },
+    {
+      "entropy": 0.8037007624283433,
+      "epoch": 6.4224,
+      "grad_norm": 5.875,
+      "learning_rate": 1.7757777777777777e-05,
+      "loss": 0.7719,
+      "mean_token_accuracy": 0.7816891677677631,
+      "num_tokens": 28567440.0,
+      "step": 2010
+    },
+    {
+      "entropy": 0.834957654774189,
+      "epoch": 6.4544,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.7735555555555558e-05,
+      "loss": 0.7928,
+      "mean_token_accuracy": 0.7771519389003515,
+      "num_tokens": 28710513.0,
+      "step": 2020
+    },
+    {
+      "entropy": 0.8089170817285776,
+      "epoch": 6.4864,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.7713333333333335e-05,
+      "loss": 0.7809,
+      "mean_token_accuracy": 0.7798932507634163,
+      "num_tokens": 28845867.0,
+      "step": 2030
+    },
+    {
+      "entropy": 0.7893792005255819,
+      "epoch": 6.5184,
+      "grad_norm": 7.0,
+      "learning_rate": 1.7691111111111113e-05,
+      "loss": 0.772,
+      "mean_token_accuracy": 0.7853275798261166,
+      "num_tokens": 28986087.0,
+      "step": 2040
+    },
+    {
+      "entropy": 0.817673640884459,
+      "epoch": 6.5504,
+      "grad_norm": 7.25,
+      "learning_rate": 1.766888888888889e-05,
+      "loss": 0.7786,
+      "mean_token_accuracy": 0.7811060819774867,
+      "num_tokens": 29127986.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.7762261744588613,
+      "epoch": 6.5824,
+      "grad_norm": 6.75,
+      "learning_rate": 1.7646666666666668e-05,
+      "loss": 0.7558,
+      "mean_token_accuracy": 0.7873165447264909,
+      "num_tokens": 29274419.0,
+      "step": 2060
+    },
+    {
+      "entropy": 0.834563871473074,
+      "epoch": 6.6144,
+      "grad_norm": 7.0625,
+      "learning_rate": 1.7624444444444446e-05,
+      "loss": 0.7983,
+      "mean_token_accuracy": 0.775597108900547,
+      "num_tokens": 29416469.0,
+      "step": 2070
+    },
+    {
+      "entropy": 0.7912350505590439,
+      "epoch": 6.6464,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.7602222222222223e-05,
+      "loss": 0.7725,
+      "mean_token_accuracy": 0.7829791463911533,
+      "num_tokens": 29552171.0,
+      "step": 2080
+    },
+    {
+      "entropy": 0.7903627971187234,
+      "epoch": 6.6784,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.758e-05,
+      "loss": 0.7663,
+      "mean_token_accuracy": 0.7871049847453833,
+      "num_tokens": 29697358.0,
+      "step": 2090
+    },
+    {
+      "entropy": 0.7845524778589607,
+      "epoch": 6.7104,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.755777777777778e-05,
+      "loss": 0.7538,
+      "mean_token_accuracy": 0.7856491521000862,
+      "num_tokens": 29837721.0,
+      "step": 2100
+    },
+    {
+      "epoch": 6.7104,
+      "eval_coding_entropy": 1.4155221619606018,
+      "eval_coding_loss": 2.2857377529144287,
+      "eval_coding_mean_token_accuracy": 0.5220260412693024,
+      "eval_coding_num_tokens": 29837721.0,
+      "eval_coding_runtime": 80.7321,
+      "eval_coding_samples_per_second": 6.193,
+      "eval_coding_steps_per_second": 3.097,
+      "step": 2100
+    },
+    {
+      "epoch": 6.7104,
+      "eval_math_entropy": 0.8305150240659713,
+      "eval_math_loss": 1.140716791152954,
+      "eval_math_mean_token_accuracy": 0.7266455957889557,
+      "eval_math_num_tokens": 29837721.0,
+      "eval_math_runtime": 46.0673,
+      "eval_math_samples_per_second": 10.854,
+      "eval_math_steps_per_second": 5.427,
+      "step": 2100
+    },
+    {
+      "entropy": 0.789828123897314,
+      "epoch": 6.7424,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.7535555555555556e-05,
+      "loss": 0.7616,
+      "mean_token_accuracy": 0.7844318665564061,
+      "num_tokens": 29975475.0,
+      "step": 2110
+    },
+    {
+      "entropy": 0.8087119059637189,
+      "epoch": 6.7744,
+      "grad_norm": 6.5,
+      "learning_rate": 1.7513333333333334e-05,
+      "loss": 0.7842,
+      "mean_token_accuracy": 0.7817314140498638,
+      "num_tokens": 30122898.0,
+      "step": 2120
+    },
+    {
+      "entropy": 0.8047438763082028,
+      "epoch": 6.8064,
+      "grad_norm": 7.6875,
+      "learning_rate": 1.749111111111111e-05,
+      "loss": 0.7719,
+      "mean_token_accuracy": 0.7834227979183197,
+      "num_tokens": 30267350.0,
+      "step": 2130
+    },
+    {
+      "entropy": 0.7722480308264494,
+      "epoch": 6.8384,
+      "grad_norm": 5.875,
+      "learning_rate": 1.746888888888889e-05,
+      "loss": 0.7482,
+      "mean_token_accuracy": 0.7906263262033463,
+      "num_tokens": 30410425.0,
+      "step": 2140
+    },
+    {
+      "entropy": 0.7863922705873847,
+      "epoch": 6.8704,
+      "grad_norm": 5.75,
+      "learning_rate": 1.7446666666666667e-05,
+      "loss": 0.7573,
+      "mean_token_accuracy": 0.7874143712222577,
+      "num_tokens": 30556104.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.7923967076465488,
+      "epoch": 6.9024,
+      "grad_norm": 6.875,
+      "learning_rate": 1.7424444444444444e-05,
+      "loss": 0.7673,
+      "mean_token_accuracy": 0.7842209670692682,
+      "num_tokens": 30696633.0,
+      "step": 2160
+    },
+    {
+      "entropy": 0.7635320017114282,
+      "epoch": 6.9344,
+      "grad_norm": 7.8125,
+      "learning_rate": 1.7402222222222222e-05,
+      "loss": 0.7348,
+      "mean_token_accuracy": 0.7918086532503367,
+      "num_tokens": 30846050.0,
+      "step": 2170
+    },
+    {
+      "entropy": 0.8394313918426632,
+      "epoch": 6.9664,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.7380000000000003e-05,
+      "loss": 0.8089,
+      "mean_token_accuracy": 0.7738638300448656,
+      "num_tokens": 30985566.0,
+      "step": 2180
+    },
+    {
+      "entropy": 0.7745752094313503,
+      "epoch": 6.9984,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.735777777777778e-05,
+      "loss": 0.7555,
+      "mean_token_accuracy": 0.7885649967938662,
+      "num_tokens": 31126781.0,
+      "step": 2190
+    },
+    {
+      "entropy": 0.6780358164718276,
+      "epoch": 7.0288,
+      "grad_norm": 6.34375,
+      "learning_rate": 1.7335555555555558e-05,
+      "loss": 0.6546,
+      "mean_token_accuracy": 0.8142786818115335,
+      "num_tokens": 31264201.0,
+      "step": 2200
+    },
+    {
+      "epoch": 7.0288,
+      "eval_coding_entropy": 1.304100867509842,
+      "eval_coding_loss": 2.3947367668151855,
+      "eval_coding_mean_token_accuracy": 0.5186028212308884,
+      "eval_coding_num_tokens": 31264201.0,
+      "eval_coding_runtime": 80.6872,
+      "eval_coding_samples_per_second": 6.197,
+      "eval_coding_steps_per_second": 3.098,
+      "step": 2200
+    },
+    {
+      "epoch": 7.0288,
+      "eval_math_entropy": 0.7535352064371109,
+      "eval_math_loss": 1.187160849571228,
+      "eval_math_mean_token_accuracy": 0.7245078366994858,
+      "eval_math_num_tokens": 31264201.0,
+      "eval_math_runtime": 45.9956,
+      "eval_math_samples_per_second": 10.871,
+      "eval_math_steps_per_second": 5.435,
+      "step": 2200
+    },
+    {
+      "entropy": 0.6964505778625607,
+      "epoch": 7.0608,
+      "grad_norm": 7.15625,
+      "learning_rate": 1.7313333333333336e-05,
+      "loss": 0.6526,
+      "mean_token_accuracy": 0.8109941691160202,
+      "num_tokens": 31407669.0,
+      "step": 2210
+    },
+    {
+      "entropy": 0.6898874808102846,
+      "epoch": 7.0928,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.7291111111111113e-05,
+      "loss": 0.6482,
+      "mean_token_accuracy": 0.8115406323224306,
+      "num_tokens": 31545593.0,
+      "step": 2220
+    },
+    {
+      "entropy": 0.6895740699023009,
+      "epoch": 7.1248,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.726888888888889e-05,
+      "loss": 0.6556,
+      "mean_token_accuracy": 0.8106831427663564,
+      "num_tokens": 31682971.0,
+      "step": 2230
+    },
+    {
+      "entropy": 0.6958585649728775,
+      "epoch": 7.1568,
+      "grad_norm": 7.0625,
+      "learning_rate": 1.724666666666667e-05,
+      "loss": 0.6613,
+      "mean_token_accuracy": 0.8089716974645853,
+      "num_tokens": 31823741.0,
+      "step": 2240
+    },
+    {
+      "entropy": 0.6874750999733805,
+      "epoch": 7.1888,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.7224444444444446e-05,
+      "loss": 0.6497,
+      "mean_token_accuracy": 0.8098180022090673,
+      "num_tokens": 31961567.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.7004628121852875,
+      "epoch": 7.2208,
+      "grad_norm": 6.59375,
+      "learning_rate": 1.7202222222222224e-05,
+      "loss": 0.6678,
+      "mean_token_accuracy": 0.806743647903204,
+      "num_tokens": 32099978.0,
+      "step": 2260
+    },
+    {
+      "entropy": 0.6884488120675087,
+      "epoch": 7.2528,
+      "grad_norm": 5.90625,
+      "learning_rate": 1.718e-05,
+      "loss": 0.6613,
+      "mean_token_accuracy": 0.8100233267992735,
+      "num_tokens": 32241652.0,
+      "step": 2270
+    },
+    {
+      "entropy": 0.677618951164186,
+      "epoch": 7.2848,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.715777777777778e-05,
+      "loss": 0.6566,
+      "mean_token_accuracy": 0.8116524960845709,
+      "num_tokens": 32384902.0,
+      "step": 2280
+    },
+    {
+      "entropy": 0.7019629061222077,
+      "epoch": 7.3168,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.7135555555555557e-05,
+      "loss": 0.6562,
+      "mean_token_accuracy": 0.807775004953146,
+      "num_tokens": 32528866.0,
+      "step": 2290
+    },
+    {
+      "entropy": 0.7041911154985427,
+      "epoch": 7.3488,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.7113333333333334e-05,
+      "loss": 0.6684,
+      "mean_token_accuracy": 0.8070357060059905,
+      "num_tokens": 32676568.0,
+      "step": 2300
+    },
+    {
+      "epoch": 7.3488,
+      "eval_coding_entropy": 1.2615632174015046,
+      "eval_coding_loss": 2.4397706985473633,
+      "eval_coding_mean_token_accuracy": 0.5178579885959625,
+      "eval_coding_num_tokens": 32676568.0,
+      "eval_coding_runtime": 80.7921,
+      "eval_coding_samples_per_second": 6.189,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 2300
+    },
+    {
+      "epoch": 7.3488,
+      "eval_math_entropy": 0.7428458639383316,
+      "eval_math_loss": 1.1929874420166016,
+      "eval_math_mean_token_accuracy": 0.7246040363311768,
+      "eval_math_num_tokens": 32676568.0,
+      "eval_math_runtime": 46.0974,
+      "eval_math_samples_per_second": 10.847,
+      "eval_math_steps_per_second": 5.423,
+      "step": 2300
+    },
+    {
+      "entropy": 0.6917074115946888,
+      "epoch": 7.3808,
+      "grad_norm": 5.5,
+      "learning_rate": 1.7091111111111112e-05,
+      "loss": 0.6567,
+      "mean_token_accuracy": 0.8070734158158303,
+      "num_tokens": 32815334.0,
+      "step": 2310
+    },
+    {
+      "entropy": 0.701219811849296,
+      "epoch": 7.4128,
+      "grad_norm": 6.78125,
+      "learning_rate": 1.706888888888889e-05,
+      "loss": 0.6619,
+      "mean_token_accuracy": 0.8089694399386644,
+      "num_tokens": 32953795.0,
+      "step": 2320
+    },
+    {
+      "entropy": 0.7057253098115325,
+      "epoch": 7.4448,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.704666666666667e-05,
+      "loss": 0.6766,
+      "mean_token_accuracy": 0.8057920385152102,
+      "num_tokens": 33095143.0,
+      "step": 2330
+    },
+    {
+      "entropy": 0.7029675403609872,
+      "epoch": 7.4768,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.7024444444444445e-05,
+      "loss": 0.6654,
+      "mean_token_accuracy": 0.8057617463171483,
+      "num_tokens": 33231603.0,
+      "step": 2340
+    },
+    {
+      "entropy": 0.7067077737301588,
+      "epoch": 7.5088,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.7002222222222226e-05,
+      "loss": 0.6633,
+      "mean_token_accuracy": 0.8062474392354488,
+      "num_tokens": 33373326.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.6964337192475796,
+      "epoch": 7.5408,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.698e-05,
+      "loss": 0.6706,
+      "mean_token_accuracy": 0.8072092153131962,
+      "num_tokens": 33518601.0,
+      "step": 2360
+    },
+    {
+      "entropy": 0.7400129863992333,
+      "epoch": 7.5728,
+      "grad_norm": 7.125,
+      "learning_rate": 1.695777777777778e-05,
+      "loss": 0.7088,
+      "mean_token_accuracy": 0.7985821563750506,
+      "num_tokens": 33658816.0,
+      "step": 2370
+    },
+    {
+      "entropy": 0.716340858861804,
+      "epoch": 7.6048,
+      "grad_norm": 7.4375,
+      "learning_rate": 1.6935555555555555e-05,
+      "loss": 0.6818,
+      "mean_token_accuracy": 0.8039816755801439,
+      "num_tokens": 33801497.0,
+      "step": 2380
+    },
+    {
+      "entropy": 0.6917989738285542,
+      "epoch": 7.6368,
+      "grad_norm": 6.125,
+      "learning_rate": 1.6913333333333336e-05,
+      "loss": 0.6613,
+      "mean_token_accuracy": 0.8100811116397381,
+      "num_tokens": 33953968.0,
+      "step": 2390
+    },
+    {
+      "entropy": 0.7106034072116018,
+      "epoch": 7.6688,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.689111111111111e-05,
+      "loss": 0.6834,
+      "mean_token_accuracy": 0.8061708252876997,
+      "num_tokens": 34098980.0,
+      "step": 2400
+    },
+    {
+      "epoch": 7.6688,
+      "eval_coding_entropy": 1.275503532409668,
+      "eval_coding_loss": 2.4567227363586426,
+      "eval_coding_mean_token_accuracy": 0.5148999409675599,
+      "eval_coding_num_tokens": 34098980.0,
+      "eval_coding_runtime": 80.8055,
+      "eval_coding_samples_per_second": 6.188,
+      "eval_coding_steps_per_second": 3.094,
+      "step": 2400
+    },
+    {
+      "epoch": 7.6688,
+      "eval_math_entropy": 0.7483759069442749,
+      "eval_math_loss": 1.1908166408538818,
+      "eval_math_mean_token_accuracy": 0.725921373128891,
+      "eval_math_num_tokens": 34098980.0,
+      "eval_math_runtime": 46.1081,
+      "eval_math_samples_per_second": 10.844,
+      "eval_math_steps_per_second": 5.422,
+      "step": 2400
+    },
+    {
+      "entropy": 0.6861953908577562,
+      "epoch": 7.7008,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.686888888888889e-05,
+      "loss": 0.658,
+      "mean_token_accuracy": 0.8090121451765299,
+      "num_tokens": 34244356.0,
+      "step": 2410
+    },
+    {
+      "entropy": 0.6957589897327126,
+      "epoch": 7.7328,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.684666666666667e-05,
+      "loss": 0.6682,
+      "mean_token_accuracy": 0.8056382369250059,
+      "num_tokens": 34386627.0,
+      "step": 2420
+    },
+    {
+      "entropy": 0.6806132102385163,
+      "epoch": 7.7648,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.6824444444444447e-05,
+      "loss": 0.6505,
+      "mean_token_accuracy": 0.8116094917058945,
+      "num_tokens": 34529410.0,
+      "step": 2430
+    },
+    {
+      "entropy": 0.7054837485775352,
+      "epoch": 7.7968,
+      "grad_norm": 5.90625,
+      "learning_rate": 1.6802222222222224e-05,
+      "loss": 0.6771,
+      "mean_token_accuracy": 0.8048388287425041,
+      "num_tokens": 34672533.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.723476623184979,
+      "epoch": 7.8288,
+      "grad_norm": 6.59375,
+      "learning_rate": 1.6780000000000002e-05,
+      "loss": 0.6912,
+      "mean_token_accuracy": 0.7992072824388743,
+      "num_tokens": 34816086.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.7052285142242909,
+      "epoch": 7.8608,
+      "grad_norm": 5.75,
+      "learning_rate": 1.675777777777778e-05,
+      "loss": 0.6724,
+      "mean_token_accuracy": 0.8062153965234756,
+      "num_tokens": 34957664.0,
+      "step": 2460
+    },
+    {
+      "entropy": 0.689465108141303,
+      "epoch": 7.8928,
+      "grad_norm": 6.65625,
+      "learning_rate": 1.6735555555555557e-05,
+      "loss": 0.6624,
+      "mean_token_accuracy": 0.8078870676457882,
+      "num_tokens": 35103008.0,
+      "step": 2470
+    },
+    {
+      "entropy": 0.7182905808091163,
+      "epoch": 7.9248,
+      "grad_norm": 7.5625,
+      "learning_rate": 1.6713333333333335e-05,
+      "loss": 0.6914,
+      "mean_token_accuracy": 0.8023614939302206,
+      "num_tokens": 35246605.0,
+      "step": 2480
+    },
+    {
+      "entropy": 0.6776346262544394,
+      "epoch": 7.9568,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.6691111111111112e-05,
+      "loss": 0.6477,
+      "mean_token_accuracy": 0.8146071847528219,
+      "num_tokens": 35390831.0,
+      "step": 2490
+    },
+    {
+      "entropy": 0.7015655115246773,
+      "epoch": 7.9888,
+      "grad_norm": 6.5,
+      "learning_rate": 1.666888888888889e-05,
+      "loss": 0.6571,
+      "mean_token_accuracy": 0.8058168698102236,
+      "num_tokens": 35532109.0,
+      "step": 2500
+    },
+    {
+      "epoch": 7.9888,
+      "eval_coding_entropy": 1.2702913012504577,
+      "eval_coding_loss": 2.4451162815093994,
+      "eval_coding_mean_token_accuracy": 0.5166184566020966,
+      "eval_coding_num_tokens": 35532109.0,
+      "eval_coding_runtime": 80.8534,
+      "eval_coding_samples_per_second": 6.184,
+      "eval_coding_steps_per_second": 3.092,
+      "step": 2500
+    },
+    {
+      "epoch": 7.9888,
+      "eval_math_entropy": 0.7409075227975845,
+      "eval_math_loss": 1.1822519302368164,
+      "eval_math_mean_token_accuracy": 0.7277101860046387,
+      "eval_math_num_tokens": 35532109.0,
+      "eval_math_runtime": 45.659,
+      "eval_math_samples_per_second": 10.951,
+      "eval_math_steps_per_second": 5.475,
+      "step": 2500
+    },
+    {
+      "entropy": 0.6560120561013096,
+      "epoch": 8.0192,
+      "grad_norm": 6.5,
+      "learning_rate": 1.6646666666666668e-05,
+      "loss": 0.6135,
+      "mean_token_accuracy": 0.8244276019303423,
+      "num_tokens": 35665123.0,
+      "step": 2510
+    },
+    {
+      "entropy": 0.6025455558672548,
+      "epoch": 8.0512,
+      "grad_norm": 6.03125,
+      "learning_rate": 1.6624444444444445e-05,
+      "loss": 0.5613,
+      "mean_token_accuracy": 0.8370782844722271,
+      "num_tokens": 35807960.0,
+      "step": 2520
+    },
+    {
+      "entropy": 0.59874182138592,
+      "epoch": 8.0832,
+      "grad_norm": 6.78125,
+      "learning_rate": 1.6602222222222223e-05,
+      "loss": 0.5531,
+      "mean_token_accuracy": 0.8351179607212543,
+      "num_tokens": 35951873.0,
+      "step": 2530
+    },
+    {
+      "entropy": 0.5861171767115593,
+      "epoch": 8.1152,
+      "grad_norm": 6.59375,
+      "learning_rate": 1.658e-05,
+      "loss": 0.5421,
+      "mean_token_accuracy": 0.8375212635844946,
+      "num_tokens": 36094800.0,
+      "step": 2540
+    },
+    {
+      "entropy": 0.5902227221988141,
+      "epoch": 8.1472,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.6557777777777778e-05,
+      "loss": 0.5441,
+      "mean_token_accuracy": 0.837360543012619,
+      "num_tokens": 36237253.0,
+      "step": 2550
+    },
+    {
+      "entropy": 0.6122677305713295,
+      "epoch": 8.1792,
+      "grad_norm": 5.75,
+      "learning_rate": 1.6535555555555556e-05,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.831546526774764,
+      "num_tokens": 36382754.0,
+      "step": 2560
+    },
+    {
+      "entropy": 0.5965196141973138,
+      "epoch": 8.2112,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.6513333333333333e-05,
+      "loss": 0.5472,
+      "mean_token_accuracy": 0.8380160730332136,
+      "num_tokens": 36524612.0,
+      "step": 2570
+    },
+    {
+      "entropy": 0.6061937671154738,
+      "epoch": 8.2432,
+      "grad_norm": 7.125,
+      "learning_rate": 1.6491111111111114e-05,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8334988377988338,
+      "num_tokens": 36665472.0,
+      "step": 2580
+    },
+    {
+      "entropy": 0.5859136013314128,
+      "epoch": 8.2752,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.646888888888889e-05,
+      "loss": 0.547,
+      "mean_token_accuracy": 0.8391021471470594,
+      "num_tokens": 36816358.0,
+      "step": 2590
+    },
+    {
+      "entropy": 0.6046942584216595,
+      "epoch": 8.3072,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.644666666666667e-05,
+      "loss": 0.5638,
+      "mean_token_accuracy": 0.8341225437819958,
+      "num_tokens": 36957057.0,
+      "step": 2600
+    },
+    {
+      "epoch": 8.3072,
+      "eval_coding_entropy": 1.190221048116684,
+      "eval_coding_loss": 2.6115522384643555,
+      "eval_coding_mean_token_accuracy": 0.5068803449869156,
+      "eval_coding_num_tokens": 36957057.0,
+      "eval_coding_runtime": 80.5416,
+      "eval_coding_samples_per_second": 6.208,
+      "eval_coding_steps_per_second": 3.104,
+      "step": 2600
+    },
+    {
+      "epoch": 8.3072,
+      "eval_math_entropy": 0.6728351511955262,
+      "eval_math_loss": 1.2562575340270996,
+      "eval_math_mean_token_accuracy": 0.7232639222145081,
+      "eval_math_num_tokens": 36957057.0,
+      "eval_math_runtime": 45.9495,
+      "eval_math_samples_per_second": 10.882,
+      "eval_math_steps_per_second": 5.441,
+      "step": 2600
+    },
+    {
+      "entropy": 0.6147088201716542,
+      "epoch": 8.3392,
+      "grad_norm": 6.75,
+      "learning_rate": 1.6424444444444444e-05,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8301718145608902,
+      "num_tokens": 37100317.0,
+      "step": 2610
+    },
+    {
+      "entropy": 0.6111037125810981,
+      "epoch": 8.3712,
+      "grad_norm": 6.5,
+      "learning_rate": 1.6402222222222225e-05,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8321197848767042,
+      "num_tokens": 37238699.0,
+      "step": 2620
+    },
+    {
+      "entropy": 0.6080829676240682,
+      "epoch": 8.4032,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.638e-05,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.831758837401867,
+      "num_tokens": 37380463.0,
+      "step": 2630
+    },
+    {
+      "entropy": 0.5950629852712155,
+      "epoch": 8.4352,
+      "grad_norm": 7.03125,
+      "learning_rate": 1.635777777777778e-05,
+      "loss": 0.5566,
+      "mean_token_accuracy": 0.8358054164797067,
+      "num_tokens": 37523481.0,
+      "step": 2640
+    },
+    {
+      "entropy": 0.6291573172435164,
+      "epoch": 8.4672,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.6335555555555558e-05,
+      "loss": 0.5907,
+      "mean_token_accuracy": 0.8268824059516191,
+      "num_tokens": 37664219.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.6096789442002774,
+      "epoch": 8.4992,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.6313333333333335e-05,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8319921869784593,
+      "num_tokens": 37804845.0,
+      "step": 2660
+    },
+    {
+      "entropy": 0.6148664908483624,
+      "epoch": 8.5312,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.6291111111111113e-05,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8318193539977073,
+      "num_tokens": 37946895.0,
+      "step": 2670
+    },
+    {
+      "entropy": 0.6144907113164664,
+      "epoch": 8.5632,
+      "grad_norm": 6.25,
+      "learning_rate": 1.626888888888889e-05,
+      "loss": 0.5703,
+      "mean_token_accuracy": 0.8314372502267361,
+      "num_tokens": 38092590.0,
+      "step": 2680
+    },
+    {
+      "entropy": 0.6098557639867067,
+      "epoch": 8.5952,
+      "grad_norm": 6.375,
+      "learning_rate": 1.6246666666666668e-05,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8304973334074021,
+      "num_tokens": 38240389.0,
+      "step": 2690
+    },
+    {
+      "entropy": 0.6045034935697913,
+      "epoch": 8.6272,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.6224444444444446e-05,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8335471376776695,
+      "num_tokens": 38387161.0,
+      "step": 2700
+    },
+    {
+      "epoch": 8.6272,
+      "eval_coding_entropy": 1.2006142148971557,
+      "eval_coding_loss": 2.568366765975952,
+      "eval_coding_mean_token_accuracy": 0.5114180357456207,
+      "eval_coding_num_tokens": 38387161.0,
+      "eval_coding_runtime": 80.161,
+      "eval_coding_samples_per_second": 6.237,
+      "eval_coding_steps_per_second": 3.119,
+      "step": 2700
+    },
+    {
+      "epoch": 8.6272,
+      "eval_math_entropy": 0.6876461794376373,
+      "eval_math_loss": 1.2407033443450928,
+      "eval_math_mean_token_accuracy": 0.7249098746776581,
+      "eval_math_num_tokens": 38387161.0,
+      "eval_math_runtime": 45.7722,
+      "eval_math_samples_per_second": 10.924,
+      "eval_math_steps_per_second": 5.462,
+      "step": 2700
+    },
+    {
+      "entropy": 0.6427660673856735,
+      "epoch": 8.6592,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.6202222222222223e-05,
+      "loss": 0.6024,
+      "mean_token_accuracy": 0.8231842331588268,
+      "num_tokens": 38531282.0,
+      "step": 2710
+    },
+    {
+      "entropy": 0.6149282867088914,
+      "epoch": 8.6912,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.618e-05,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.832622691988945,
+      "num_tokens": 38675669.0,
+      "step": 2720
+    },
+    {
+      "entropy": 0.5920048102736473,
+      "epoch": 8.7232,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.615777777777778e-05,
+      "loss": 0.5556,
+      "mean_token_accuracy": 0.8333576548844576,
+      "num_tokens": 38812594.0,
+      "step": 2730
+    },
+    {
+      "entropy": 0.6234932013787329,
+      "epoch": 8.7552,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.6135555555555556e-05,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8268820513039827,
+      "num_tokens": 38956131.0,
+      "step": 2740
+    },
+    {
+      "entropy": 0.5987358037382364,
+      "epoch": 8.7872,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.6113333333333334e-05,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8351681742817163,
+      "num_tokens": 39091841.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.6189692422747612,
+      "epoch": 8.8192,
+      "grad_norm": 5.90625,
+      "learning_rate": 1.609111111111111e-05,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8277230367064476,
+      "num_tokens": 39231403.0,
+      "step": 2760
+    },
+    {
+      "entropy": 0.612622588314116,
+      "epoch": 8.8512,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.606888888888889e-05,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8315247103571892,
+      "num_tokens": 39369492.0,
+      "step": 2770
+    },
+    {
+      "entropy": 0.6111666314303875,
+      "epoch": 8.8832,
+      "grad_norm": 6.75,
+      "learning_rate": 1.6046666666666667e-05,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8307797733694315,
+      "num_tokens": 39509140.0,
+      "step": 2780
+    },
+    {
+      "entropy": 0.6081953683868051,
+      "epoch": 8.9152,
+      "grad_norm": 7.1875,
+      "learning_rate": 1.6024444444444444e-05,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8316793866455555,
+      "num_tokens": 39651013.0,
+      "step": 2790
+    },
+    {
+      "entropy": 0.6254503468051553,
+      "epoch": 8.9472,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.6002222222222222e-05,
+      "loss": 0.5835,
+      "mean_token_accuracy": 0.8256080824881792,
+      "num_tokens": 39792941.0,
+      "step": 2800
+    },
+    {
+      "epoch": 8.9472,
+      "eval_coding_entropy": 1.182766483783722,
+      "eval_coding_loss": 2.6028387546539307,
+      "eval_coding_mean_token_accuracy": 0.5101276164054871,
+      "eval_coding_num_tokens": 39792941.0,
+      "eval_coding_runtime": 80.5635,
+      "eval_coding_samples_per_second": 6.206,
+      "eval_coding_steps_per_second": 3.103,
+      "step": 2800
+    },
+    {
+      "epoch": 8.9472,
+      "eval_math_entropy": 0.6721872200369835,
+      "eval_math_loss": 1.250475287437439,
+      "eval_math_mean_token_accuracy": 0.7252113993167877,
+      "eval_math_num_tokens": 39792941.0,
+      "eval_math_runtime": 46.1028,
+      "eval_math_samples_per_second": 10.845,
+      "eval_math_steps_per_second": 5.423,
+      "step": 2800
+    },
+    {
+      "entropy": 0.6324062248691916,
+      "epoch": 8.9792,
+      "grad_norm": 6.125,
+      "learning_rate": 1.5980000000000003e-05,
+      "loss": 0.5989,
+      "mean_token_accuracy": 0.826288477703929,
+      "num_tokens": 39939134.0,
+      "step": 2810
+    },
+    {
+      "entropy": 0.6012408115754002,
+      "epoch": 9.0096,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.5957777777777777e-05,
+      "loss": 0.5458,
+      "mean_token_accuracy": 0.8398351175220389,
+      "num_tokens": 40074891.0,
+      "step": 2820
+    },
+    {
+      "entropy": 0.5318261187523603,
+      "epoch": 9.0416,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.5935555555555558e-05,
+      "loss": 0.4762,
+      "mean_token_accuracy": 0.8588335525244475,
+      "num_tokens": 40222571.0,
+      "step": 2830
+    },
+    {
+      "entropy": 0.5274458668194711,
+      "epoch": 9.0736,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.5913333333333332e-05,
+      "loss": 0.4708,
+      "mean_token_accuracy": 0.8612128391861915,
+      "num_tokens": 40367991.0,
+      "step": 2840
+    },
+    {
+      "entropy": 0.5246741400100291,
+      "epoch": 9.1056,
+      "grad_norm": 6.25,
+      "learning_rate": 1.5891111111111113e-05,
+      "loss": 0.4579,
+      "mean_token_accuracy": 0.8627629213035106,
+      "num_tokens": 40509352.0,
+      "step": 2850
+    },
+    {
+      "entropy": 0.5131453761830926,
+      "epoch": 9.1376,
+      "grad_norm": 9.0,
+      "learning_rate": 1.5868888888888888e-05,
+      "loss": 0.4494,
+      "mean_token_accuracy": 0.8649103097617626,
+      "num_tokens": 40651112.0,
+      "step": 2860
+    },
+    {
+      "entropy": 0.5197440345771611,
+      "epoch": 9.1696,
+      "grad_norm": 6.15625,
+      "learning_rate": 1.584666666666667e-05,
+      "loss": 0.4622,
+      "mean_token_accuracy": 0.8633057914674283,
+      "num_tokens": 40787831.0,
+      "step": 2870
+    },
+    {
+      "entropy": 0.5328541550785303,
+      "epoch": 9.2016,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.5824444444444446e-05,
+      "loss": 0.4764,
+      "mean_token_accuracy": 0.8565490908920765,
+      "num_tokens": 40932702.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.5416258215904236,
+      "epoch": 9.2336,
+      "grad_norm": 6.78125,
+      "learning_rate": 1.5802222222222224e-05,
+      "loss": 0.4798,
+      "mean_token_accuracy": 0.8568971626460552,
+      "num_tokens": 41080043.0,
+      "step": 2890
+    },
+    {
+      "entropy": 0.5248058075085282,
+      "epoch": 9.2656,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.578e-05,
+      "loss": 0.4775,
+      "mean_token_accuracy": 0.8570398628711701,
+      "num_tokens": 41220738.0,
+      "step": 2900
+    },
+    {
+      "epoch": 9.2656,
+      "eval_coding_entropy": 1.1136542909145355,
+      "eval_coding_loss": 2.7673168182373047,
+      "eval_coding_mean_token_accuracy": 0.5006222115755081,
+      "eval_coding_num_tokens": 41220738.0,
+      "eval_coding_runtime": 80.8421,
+      "eval_coding_samples_per_second": 6.185,
+      "eval_coding_steps_per_second": 3.092,
+      "step": 2900
+    },
+    {
+      "epoch": 9.2656,
+      "eval_math_entropy": 0.6322178390026093,
+      "eval_math_loss": 1.325918197631836,
+      "eval_math_mean_token_accuracy": 0.7193420985937119,
+      "eval_math_num_tokens": 41220738.0,
+      "eval_math_runtime": 46.1351,
+      "eval_math_samples_per_second": 10.838,
+      "eval_math_steps_per_second": 5.419,
+      "step": 2900
+    },
+    {
+      "entropy": 0.5244877910241484,
+      "epoch": 9.2976,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.575777777777778e-05,
+      "loss": 0.4692,
+      "mean_token_accuracy": 0.8596760775893927,
+      "num_tokens": 41364745.0,
+      "step": 2910
+    },
+    {
+      "entropy": 0.5472216626629234,
+      "epoch": 9.3296,
+      "grad_norm": 6.5625,
+      "learning_rate": 1.5735555555555557e-05,
+      "loss": 0.4817,
+      "mean_token_accuracy": 0.8550189323723316,
+      "num_tokens": 41506028.0,
+      "step": 2920
+    },
+    {
+      "entropy": 0.5349522266536951,
+      "epoch": 9.3616,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.5713333333333334e-05,
+      "loss": 0.4839,
+      "mean_token_accuracy": 0.8562944039702416,
+      "num_tokens": 41651227.0,
+      "step": 2930
+    },
+    {
+      "entropy": 0.5615787353366614,
+      "epoch": 9.3936,
+      "grad_norm": 6.78125,
+      "learning_rate": 1.5691111111111112e-05,
+      "loss": 0.5062,
+      "mean_token_accuracy": 0.8496946014463902,
+      "num_tokens": 41793383.0,
+      "step": 2940
+    },
+    {
+      "entropy": 0.5343781635165215,
+      "epoch": 9.4256,
+      "grad_norm": 6.75,
+      "learning_rate": 1.5668888888888893e-05,
+      "loss": 0.4748,
+      "mean_token_accuracy": 0.8585585359483957,
+      "num_tokens": 41931388.0,
+      "step": 2950
+    },
+    {
+      "entropy": 0.5460910867899657,
+      "epoch": 9.4576,
+      "grad_norm": 6.21875,
+      "learning_rate": 1.5646666666666667e-05,
+      "loss": 0.4908,
+      "mean_token_accuracy": 0.8529999420046807,
+      "num_tokens": 42072732.0,
+      "step": 2960
+    },
+    {
+      "entropy": 0.5411690116859973,
+      "epoch": 9.4896,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.5624444444444448e-05,
+      "loss": 0.491,
+      "mean_token_accuracy": 0.8546163454651833,
+      "num_tokens": 42216880.0,
+      "step": 2970
+    },
+    {
+      "entropy": 0.5196895817294717,
+      "epoch": 9.5216,
+      "grad_norm": 6.03125,
+      "learning_rate": 1.5602222222222222e-05,
+      "loss": 0.4608,
+      "mean_token_accuracy": 0.8605905815958976,
+      "num_tokens": 42363129.0,
+      "step": 2980
+    },
+    {
+      "entropy": 0.5527703028172255,
+      "epoch": 9.5536,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.5580000000000003e-05,
+      "loss": 0.4997,
+      "mean_token_accuracy": 0.8506911654025316,
+      "num_tokens": 42501873.0,
+      "step": 2990
+    },
+    {
+      "entropy": 0.5321281863376498,
+      "epoch": 9.5856,
+      "grad_norm": 7.3125,
+      "learning_rate": 1.5557777777777778e-05,
+      "loss": 0.4759,
+      "mean_token_accuracy": 0.8594888877123594,
+      "num_tokens": 42642989.0,
+      "step": 3000
+    },
+    {
+      "epoch": 9.5856,
+      "eval_coding_entropy": 1.0940111632347107,
+      "eval_coding_loss": 2.7935373783111572,
+      "eval_coding_mean_token_accuracy": 0.49977762436866763,
+      "eval_coding_num_tokens": 42642989.0,
+      "eval_coding_runtime": 81.0858,
+      "eval_coding_samples_per_second": 6.166,
+      "eval_coding_steps_per_second": 3.083,
+      "step": 3000
+    },
+    {
+      "epoch": 9.5856,
+      "eval_math_entropy": 0.6124070265889168,
+      "eval_math_loss": 1.3337898254394531,
+      "eval_math_mean_token_accuracy": 0.7198901507854462,
+      "eval_math_num_tokens": 42642989.0,
+      "eval_math_runtime": 45.8245,
+      "eval_math_samples_per_second": 10.911,
+      "eval_math_steps_per_second": 5.456,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 10000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 32,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.702990516896309e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}