diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,39734 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 10.0,
+  "eval_steps": 220,
+  "global_step": 4390,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.002280501710376283,
+      "grad_norm": 44.5,
+      "learning_rate": 0.0,
+      "loss": 2.4621,
+      "mean_token_accuracy": 0.659612387418747,
+      "num_tokens": 107751.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.004561003420752566,
+      "grad_norm": 42.25,
+      "learning_rate": 2.272727272727273e-08,
+      "loss": 2.5474,
+      "mean_token_accuracy": 0.6517423838376999,
+      "num_tokens": 214469.0,
+      "step": 2
+    },
+    {
+      "epoch": 0.0068415051311288486,
+      "grad_norm": 48.75,
+      "learning_rate": 4.545454545454546e-08,
+      "loss": 2.5963,
+      "mean_token_accuracy": 0.6451424211263657,
+      "num_tokens": 321493.0,
+      "step": 3
+    },
+    {
+      "epoch": 0.009122006841505131,
+      "grad_norm": 52.75,
+      "learning_rate": 6.818181818181819e-08,
+      "loss": 2.5346,
+      "mean_token_accuracy": 0.656927764415741,
+      "num_tokens": 428374.0,
+      "step": 4
+    },
+    {
+      "epoch": 0.011402508551881414,
+      "grad_norm": 33.75,
+      "learning_rate": 9.090909090909091e-08,
+      "loss": 2.5273,
+      "mean_token_accuracy": 0.6588353216648102,
+      "num_tokens": 535756.0,
+      "step": 5
+    },
+    {
+      "epoch": 0.013683010262257697,
+      "grad_norm": 32.0,
+      "learning_rate": 1.1363636363636364e-07,
+      "loss": 2.4468,
+      "mean_token_accuracy": 0.6671261787414551,
+      "num_tokens": 642560.0,
+      "step": 6
+    },
+    {
+      "epoch": 0.01596351197263398,
+      "grad_norm": 33.5,
+      "learning_rate": 1.3636363636363637e-07,
+      "loss": 2.5426,
+      "mean_token_accuracy": 0.6598037928342819,
+      "num_tokens": 749804.0,
+      "step": 7
+    },
+    {
+      "epoch": 0.018244013683010263,
+      "grad_norm": 33.25,
+      "learning_rate": 1.590909090909091e-07,
+      "loss": 2.5504,
+      "mean_token_accuracy": 0.6524623930454254,
+      "num_tokens": 857462.0,
+      "step": 8
+    },
+    {
+      "epoch": 0.020524515393386546,
+      "grad_norm": 39.25,
+      "learning_rate": 1.8181818181818183e-07,
+      "loss": 2.4795,
+      "mean_token_accuracy": 0.6614948958158493,
+      "num_tokens": 964652.0,
+      "step": 9
+    },
+    {
+      "epoch": 0.02280501710376283,
+      "grad_norm": 32.25,
+      "learning_rate": 2.0454545454545456e-07,
+      "loss": 2.4681,
+      "mean_token_accuracy": 0.6571934372186661,
+      "num_tokens": 1072114.0,
+      "step": 10
+    },
+    {
+      "epoch": 0.02508551881413911,
+      "grad_norm": 56.75,
+      "learning_rate": 2.2727272727272729e-07,
+      "loss": 2.5162,
+      "mean_token_accuracy": 0.6555981487035751,
+      "num_tokens": 1179487.0,
+      "step": 11
+    },
+    {
+      "epoch": 0.027366020524515394,
+      "grad_norm": 50.0,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 2.5639,
+      "mean_token_accuracy": 0.6516416519880295,
+      "num_tokens": 1286278.0,
+      "step": 12
+    },
+    {
+      "epoch": 0.029646522234891677,
+      "grad_norm": 36.5,
+      "learning_rate": 2.7272727272727274e-07,
+      "loss": 2.4399,
+      "mean_token_accuracy": 0.669710099697113,
+      "num_tokens": 1393298.0,
+      "step": 13
+    },
+    {
+      "epoch": 0.03192702394526796,
+      "grad_norm": 39.0,
+      "learning_rate": 2.954545454545455e-07,
+      "loss": 2.4725,
+      "mean_token_accuracy": 0.6617565155029297,
+      "num_tokens": 1500702.0,
+      "step": 14
+    },
+    {
+      "epoch": 0.03420752565564424,
+      "grad_norm": 35.5,
+      "learning_rate": 3.181818181818182e-07,
+      "loss": 2.4827,
+      "mean_token_accuracy": 0.656602531671524,
+      "num_tokens": 1607486.0,
+      "step": 15
+    },
+    {
+      "epoch": 0.036488027366020526,
+      "grad_norm": 47.0,
+      "learning_rate": 3.409090909090909e-07,
+      "loss": 2.5839,
+      "mean_token_accuracy": 0.6526448428630829,
+      "num_tokens": 1714872.0,
+      "step": 16
+    },
+    {
+      "epoch": 0.03876852907639681,
+      "grad_norm": 35.25,
+      "learning_rate": 3.6363636363636366e-07,
+      "loss": 2.467,
+      "mean_token_accuracy": 0.6649379134178162,
+      "num_tokens": 1822351.0,
+      "step": 17
+    },
+    {
+      "epoch": 0.04104903078677309,
+      "grad_norm": 40.75,
+      "learning_rate": 3.8636363636363636e-07,
+      "loss": 2.5003,
+      "mean_token_accuracy": 0.6564287096261978,
+      "num_tokens": 1930114.0,
+      "step": 18
+    },
+    {
+      "epoch": 0.043329532497149374,
+      "grad_norm": 33.5,
+      "learning_rate": 4.090909090909091e-07,
+      "loss": 2.5395,
+      "mean_token_accuracy": 0.6506040990352631,
+      "num_tokens": 2038041.0,
+      "step": 19
+    },
+    {
+      "epoch": 0.04561003420752566,
+      "grad_norm": 32.75,
+      "learning_rate": 4.3181818181818187e-07,
+      "loss": 2.5359,
+      "mean_token_accuracy": 0.6554941684007645,
+      "num_tokens": 2144846.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04789053591790194,
+      "grad_norm": 31.125,
+      "learning_rate": 4.5454545454545457e-07,
+      "loss": 2.422,
+      "mean_token_accuracy": 0.6640460044145584,
+      "num_tokens": 2252295.0,
+      "step": 21
+    },
+    {
+      "epoch": 0.05017103762827822,
+      "grad_norm": 37.5,
+      "learning_rate": 4.772727272727274e-07,
+      "loss": 2.5004,
+      "mean_token_accuracy": 0.659651443362236,
+      "num_tokens": 2359439.0,
+      "step": 22
+    },
+    {
+      "epoch": 0.052451539338654506,
+      "grad_norm": 33.5,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 2.5621,
+      "mean_token_accuracy": 0.6528295874595642,
+      "num_tokens": 2466337.0,
+      "step": 23
+    },
+    {
+      "epoch": 0.05473204104903079,
+      "grad_norm": 33.0,
+      "learning_rate": 5.227272727272728e-07,
+      "loss": 2.4917,
+      "mean_token_accuracy": 0.661816731095314,
+      "num_tokens": 2572770.0,
+      "step": 24
+    },
+    {
+      "epoch": 0.05701254275940707,
+      "grad_norm": 45.0,
+      "learning_rate": 5.454545454545455e-07,
+      "loss": 2.402,
+      "mean_token_accuracy": 0.663006380200386,
+      "num_tokens": 2680167.0,
+      "step": 25
+    },
+    {
+      "epoch": 0.059293044469783354,
+      "grad_norm": 44.5,
+      "learning_rate": 5.681818181818182e-07,
+      "loss": 2.4679,
+      "mean_token_accuracy": 0.6532121747732162,
+      "num_tokens": 2787861.0,
+      "step": 26
+    },
+    {
+      "epoch": 0.06157354618015964,
+      "grad_norm": 34.0,
+      "learning_rate": 5.90909090909091e-07,
+      "loss": 2.5954,
+      "mean_token_accuracy": 0.64468814432621,
+      "num_tokens": 2894486.0,
+      "step": 27
+    },
+    {
+      "epoch": 0.06385404789053592,
+      "grad_norm": 56.5,
+      "learning_rate": 6.136363636363637e-07,
+      "loss": 2.5431,
+      "mean_token_accuracy": 0.6518365442752838,
+      "num_tokens": 3001693.0,
+      "step": 28
+    },
+    {
+      "epoch": 0.0661345496009122,
+      "grad_norm": 37.75,
+      "learning_rate": 6.363636363636364e-07,
+      "loss": 2.5421,
+      "mean_token_accuracy": 0.6533436477184296,
+      "num_tokens": 3108771.0,
+      "step": 29
+    },
+    {
+      "epoch": 0.06841505131128849,
+      "grad_norm": 31.5,
+      "learning_rate": 6.590909090909091e-07,
+      "loss": 2.4357,
+      "mean_token_accuracy": 0.6666090935468674,
+      "num_tokens": 3215312.0,
+      "step": 30
+    },
+    {
+      "epoch": 0.07069555302166476,
+      "grad_norm": 31.5,
+      "learning_rate": 6.818181818181818e-07,
+      "loss": 2.4502,
+      "mean_token_accuracy": 0.6624381989240646,
+      "num_tokens": 3322496.0,
+      "step": 31
+    },
+    {
+      "epoch": 0.07297605473204105,
+      "grad_norm": 32.25,
+      "learning_rate": 7.045454545454545e-07,
+      "loss": 2.5082,
+      "mean_token_accuracy": 0.6581026613712311,
+      "num_tokens": 3429869.0,
+      "step": 32
+    },
+    {
+      "epoch": 0.07525655644241733,
+      "grad_norm": 32.5,
+      "learning_rate": 7.272727272727273e-07,
+      "loss": 2.4792,
+      "mean_token_accuracy": 0.6572518944740295,
+      "num_tokens": 3536755.0,
+      "step": 33
+    },
+    {
+      "epoch": 0.07753705815279362,
+      "grad_norm": 31.75,
+      "learning_rate": 7.5e-07,
+      "loss": 2.4916,
+      "mean_token_accuracy": 0.6520580649375916,
+      "num_tokens": 3644347.0,
+      "step": 34
+    },
+    {
+      "epoch": 0.07981755986316989,
+      "grad_norm": 36.0,
+      "learning_rate": 7.727272727272727e-07,
+      "loss": 2.4185,
+      "mean_token_accuracy": 0.6675815433263779,
+      "num_tokens": 3751037.0,
+      "step": 35
+    },
+    {
+      "epoch": 0.08209806157354618,
+      "grad_norm": 32.75,
+      "learning_rate": 7.954545454545455e-07,
+      "loss": 2.4929,
+      "mean_token_accuracy": 0.6536126732826233,
+      "num_tokens": 3857928.0,
+      "step": 36
+    },
+    {
+      "epoch": 0.08437856328392246,
+      "grad_norm": 31.125,
+      "learning_rate": 8.181818181818182e-07,
+      "loss": 2.401,
+      "mean_token_accuracy": 0.6677189022302628,
+      "num_tokens": 3965628.0,
+      "step": 37
+    },
+    {
+      "epoch": 0.08665906499429875,
+      "grad_norm": 31.875,
+      "learning_rate": 8.409090909090909e-07,
+      "loss": 2.447,
+      "mean_token_accuracy": 0.6582602560520172,
+      "num_tokens": 4072519.0,
+      "step": 38
+    },
+    {
+      "epoch": 0.08893956670467502,
+      "grad_norm": 31.375,
+      "learning_rate": 8.636363636363637e-07,
+      "loss": 2.4231,
+      "mean_token_accuracy": 0.6686751246452332,
+      "num_tokens": 4180022.0,
+      "step": 39
+    },
+    {
+      "epoch": 0.09122006841505131,
+      "grad_norm": 47.25,
+      "learning_rate": 8.863636363636364e-07,
+      "loss": 2.4306,
+      "mean_token_accuracy": 0.6607790142297745,
+      "num_tokens": 4287045.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09350057012542759,
+      "grad_norm": 33.0,
+      "learning_rate": 9.090909090909091e-07,
+      "loss": 2.326,
+      "mean_token_accuracy": 0.6676892191171646,
+      "num_tokens": 4394995.0,
+      "step": 41
+    },
+    {
+      "epoch": 0.09578107183580388,
+      "grad_norm": 32.75,
+      "learning_rate": 9.31818181818182e-07,
+      "loss": 2.4869,
+      "mean_token_accuracy": 0.6459519267082214,
+      "num_tokens": 4501329.0,
+      "step": 42
+    },
+    {
+      "epoch": 0.09806157354618016,
+      "grad_norm": 31.875,
+      "learning_rate": 9.545454545454548e-07,
+      "loss": 2.5084,
+      "mean_token_accuracy": 0.6592301279306412,
+      "num_tokens": 4608718.0,
+      "step": 43
+    },
+    {
+      "epoch": 0.10034207525655645,
+      "grad_norm": 31.375,
+      "learning_rate": 9.772727272727275e-07,
+      "loss": 2.4805,
+      "mean_token_accuracy": 0.6500378400087357,
+      "num_tokens": 4715658.0,
+      "step": 44
+    },
+    {
+      "epoch": 0.10262257696693272,
+      "grad_norm": 30.75,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 2.4318,
+      "mean_token_accuracy": 0.6553644686937332,
+      "num_tokens": 4822368.0,
+      "step": 45
+    },
+    {
+      "epoch": 0.10490307867730901,
+      "grad_norm": 31.625,
+      "learning_rate": 1.0227272727272729e-06,
+      "loss": 2.4468,
+      "mean_token_accuracy": 0.6600785255432129,
+      "num_tokens": 4929696.0,
+      "step": 46
+    },
+    {
+      "epoch": 0.10718358038768529,
+      "grad_norm": 30.25,
+      "learning_rate": 1.0454545454545456e-06,
+      "loss": 2.3474,
+      "mean_token_accuracy": 0.6583467125892639,
+      "num_tokens": 5037621.0,
+      "step": 47
+    },
+    {
+      "epoch": 0.10946408209806158,
+      "grad_norm": 30.375,
+      "learning_rate": 1.0681818181818183e-06,
+      "loss": 2.3681,
+      "mean_token_accuracy": 0.6633005887269974,
+      "num_tokens": 5144969.0,
+      "step": 48
+    },
+    {
+      "epoch": 0.11174458380843785,
+      "grad_norm": 30.25,
+      "learning_rate": 1.090909090909091e-06,
+      "loss": 2.4083,
+      "mean_token_accuracy": 0.6564080715179443,
+      "num_tokens": 5252296.0,
+      "step": 49
+    },
+    {
+      "epoch": 0.11402508551881414,
+      "grad_norm": 31.125,
+      "learning_rate": 1.1136363636363637e-06,
+      "loss": 2.3943,
+      "mean_token_accuracy": 0.6526119261980057,
+      "num_tokens": 5358939.0,
+      "step": 50
+    },
+    {
+      "epoch": 0.11630558722919042,
+      "grad_norm": 31.125,
+      "learning_rate": 1.1363636363636364e-06,
+      "loss": 2.4018,
+      "mean_token_accuracy": 0.6534301191568375,
+      "num_tokens": 5465936.0,
+      "step": 51
+    },
+    {
+      "epoch": 0.11858608893956671,
+      "grad_norm": 30.125,
+      "learning_rate": 1.159090909090909e-06,
+      "loss": 2.3053,
+      "mean_token_accuracy": 0.6658475399017334,
+      "num_tokens": 5572802.0,
+      "step": 52
+    },
+    {
+      "epoch": 0.12086659064994298,
+      "grad_norm": 28.625,
+      "learning_rate": 1.181818181818182e-06,
+      "loss": 2.2423,
+      "mean_token_accuracy": 0.6766009479761124,
+      "num_tokens": 5680067.0,
+      "step": 53
+    },
+    {
+      "epoch": 0.12314709236031927,
+      "grad_norm": 29.125,
+      "learning_rate": 1.2045454545454547e-06,
+      "loss": 2.3047,
+      "mean_token_accuracy": 0.6613978892564774,
+      "num_tokens": 5787260.0,
+      "step": 54
+    },
+    {
+      "epoch": 0.12542759407069556,
+      "grad_norm": 28.875,
+      "learning_rate": 1.2272727272727274e-06,
+      "loss": 2.2284,
+      "mean_token_accuracy": 0.6748173385858536,
+      "num_tokens": 5894097.0,
+      "step": 55
+    },
+    {
+      "epoch": 0.12770809578107184,
+      "grad_norm": 28.5,
+      "learning_rate": 1.25e-06,
+      "loss": 2.3062,
+      "mean_token_accuracy": 0.665241539478302,
+      "num_tokens": 6001210.0,
+      "step": 56
+    },
+    {
+      "epoch": 0.12998859749144812,
+      "grad_norm": 26.75,
+      "learning_rate": 1.2727272727272728e-06,
+      "loss": 2.1486,
+      "mean_token_accuracy": 0.6805437654256821,
+      "num_tokens": 6109107.0,
+      "step": 57
+    },
+    {
+      "epoch": 0.1322690992018244,
+      "grad_norm": 28.25,
+      "learning_rate": 1.2954545454545455e-06,
+      "loss": 2.2641,
+      "mean_token_accuracy": 0.6705231070518494,
+      "num_tokens": 6216222.0,
+      "step": 58
+    },
+    {
+      "epoch": 0.1345496009122007,
+      "grad_norm": 28.75,
+      "learning_rate": 1.3181818181818182e-06,
+      "loss": 2.2772,
+      "mean_token_accuracy": 0.6666675806045532,
+      "num_tokens": 6323452.0,
+      "step": 59
+    },
+    {
+      "epoch": 0.13683010262257697,
+      "grad_norm": 29.625,
+      "learning_rate": 1.3409090909090911e-06,
+      "loss": 2.231,
+      "mean_token_accuracy": 0.6695638597011566,
+      "num_tokens": 6430912.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.13911060433295325,
+      "grad_norm": 27.375,
+      "learning_rate": 1.3636363636363636e-06,
+      "loss": 2.2394,
+      "mean_token_accuracy": 0.6609823554754257,
+      "num_tokens": 6538019.0,
+      "step": 61
+    },
+    {
+      "epoch": 0.14139110604332952,
+      "grad_norm": 27.75,
+      "learning_rate": 1.3863636363636365e-06,
+      "loss": 2.2672,
+      "mean_token_accuracy": 0.6677799224853516,
+      "num_tokens": 6644863.0,
+      "step": 62
+    },
+    {
+      "epoch": 0.14367160775370583,
+      "grad_norm": 27.5,
+      "learning_rate": 1.409090909090909e-06,
+      "loss": 2.2019,
+      "mean_token_accuracy": 0.66669762134552,
+      "num_tokens": 6752333.0,
+      "step": 63
+    },
+    {
+      "epoch": 0.1459521094640821,
+      "grad_norm": 27.0,
+      "learning_rate": 1.431818181818182e-06,
+      "loss": 2.2077,
+      "mean_token_accuracy": 0.6657529324293137,
+      "num_tokens": 6859405.0,
+      "step": 64
+    },
+    {
+      "epoch": 0.14823261117445838,
+      "grad_norm": 37.75,
+      "learning_rate": 1.4545454545454546e-06,
+      "loss": 2.1799,
+      "mean_token_accuracy": 0.6656498908996582,
+      "num_tokens": 6966421.0,
+      "step": 65
+    },
+    {
+      "epoch": 0.15051311288483465,
+      "grad_norm": 27.25,
+      "learning_rate": 1.4772727272727275e-06,
+      "loss": 2.2099,
+      "mean_token_accuracy": 0.6709634065628052,
+      "num_tokens": 7073157.0,
+      "step": 66
+    },
+    {
+      "epoch": 0.15279361459521096,
+      "grad_norm": 29.5,
+      "learning_rate": 1.5e-06,
+      "loss": 2.1091,
+      "mean_token_accuracy": 0.6776281297206879,
+      "num_tokens": 7180043.0,
+      "step": 67
+    },
+    {
+      "epoch": 0.15507411630558723,
+      "grad_norm": 27.25,
+      "learning_rate": 1.522727272727273e-06,
+      "loss": 2.1265,
+      "mean_token_accuracy": 0.6751858294010162,
+      "num_tokens": 7287133.0,
+      "step": 68
+    },
+    {
+      "epoch": 0.1573546180159635,
+      "grad_norm": 25.875,
+      "learning_rate": 1.5454545454545454e-06,
+      "loss": 2.1089,
+      "mean_token_accuracy": 0.6743377894163132,
+      "num_tokens": 7394475.0,
+      "step": 69
+    },
+    {
+      "epoch": 0.15963511972633979,
+      "grad_norm": 29.375,
+      "learning_rate": 1.5681818181818184e-06,
+      "loss": 2.0552,
+      "mean_token_accuracy": 0.6845020353794098,
+      "num_tokens": 7502908.0,
+      "step": 70
+    },
+    {
+      "epoch": 0.1619156214367161,
+      "grad_norm": 26.25,
+      "learning_rate": 1.590909090909091e-06,
+      "loss": 2.1523,
+      "mean_token_accuracy": 0.6694881021976471,
+      "num_tokens": 7609855.0,
+      "step": 71
+    },
+    {
+      "epoch": 0.16419612314709237,
+      "grad_norm": 26.125,
+      "learning_rate": 1.613636363636364e-06,
+      "loss": 2.168,
+      "mean_token_accuracy": 0.6710262894630432,
+      "num_tokens": 7716660.0,
+      "step": 72
+    },
+    {
+      "epoch": 0.16647662485746864,
+      "grad_norm": 25.875,
+      "learning_rate": 1.6363636363636365e-06,
+      "loss": 2.0526,
+      "mean_token_accuracy": 0.6768545061349869,
+      "num_tokens": 7823567.0,
+      "step": 73
+    },
+    {
+      "epoch": 0.16875712656784492,
+      "grad_norm": 25.375,
+      "learning_rate": 1.6590909090909094e-06,
+      "loss": 2.0024,
+      "mean_token_accuracy": 0.6849700808525085,
+      "num_tokens": 7930741.0,
+      "step": 74
+    },
+    {
+      "epoch": 0.17103762827822122,
+      "grad_norm": 24.375,
+      "learning_rate": 1.6818181818181819e-06,
+      "loss": 2.0285,
+      "mean_token_accuracy": 0.6854856461286545,
+      "num_tokens": 8037391.0,
+      "step": 75
+    },
+    {
+      "epoch": 0.1733181299885975,
+      "grad_norm": 24.125,
+      "learning_rate": 1.7045454545454546e-06,
+      "loss": 1.9599,
+      "mean_token_accuracy": 0.6963415741920471,
+      "num_tokens": 8144926.0,
+      "step": 76
+    },
+    {
+      "epoch": 0.17559863169897377,
+      "grad_norm": 25.375,
+      "learning_rate": 1.7272727272727275e-06,
+      "loss": 1.9961,
+      "mean_token_accuracy": 0.6796167641878128,
+      "num_tokens": 8251774.0,
+      "step": 77
+    },
+    {
+      "epoch": 0.17787913340935005,
+      "grad_norm": 23.25,
+      "learning_rate": 1.75e-06,
+      "loss": 1.8617,
+      "mean_token_accuracy": 0.7025072127580643,
+      "num_tokens": 8358808.0,
+      "step": 78
+    },
+    {
+      "epoch": 0.18015963511972635,
+      "grad_norm": 23.625,
+      "learning_rate": 1.7727272727272729e-06,
+      "loss": 1.8932,
+      "mean_token_accuracy": 0.6979697048664093,
+      "num_tokens": 8466050.0,
+      "step": 79
+    },
+    {
+      "epoch": 0.18244013683010263,
+      "grad_norm": 24.25,
+      "learning_rate": 1.7954545454545456e-06,
+      "loss": 1.9363,
+      "mean_token_accuracy": 0.6897751837968826,
+      "num_tokens": 8572892.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.1847206385404789,
+      "grad_norm": 23.375,
+      "learning_rate": 1.8181818181818183e-06,
+      "loss": 1.9175,
+      "mean_token_accuracy": 0.6948112696409225,
+      "num_tokens": 8679995.0,
+      "step": 81
+    },
+    {
+      "epoch": 0.18700114025085518,
+      "grad_norm": 22.625,
+      "learning_rate": 1.840909090909091e-06,
+      "loss": 1.825,
+      "mean_token_accuracy": 0.7063078880310059,
+      "num_tokens": 8787388.0,
+      "step": 82
+    },
+    {
+      "epoch": 0.18928164196123148,
+      "grad_norm": 21.875,
+      "learning_rate": 1.863636363636364e-06,
+      "loss": 1.8128,
+      "mean_token_accuracy": 0.7035099864006042,
+      "num_tokens": 8894141.0,
+      "step": 83
+    },
+    {
+      "epoch": 0.19156214367160776,
+      "grad_norm": 22.375,
+      "learning_rate": 1.8863636363636364e-06,
+      "loss": 1.8433,
+      "mean_token_accuracy": 0.6927594691514969,
+      "num_tokens": 9000732.0,
+      "step": 84
+    },
+    {
+      "epoch": 0.19384264538198404,
+      "grad_norm": 21.5,
+      "learning_rate": 1.9090909090909095e-06,
+      "loss": 1.7329,
+      "mean_token_accuracy": 0.7171757072210312,
+      "num_tokens": 9107855.0,
+      "step": 85
+    },
+    {
+      "epoch": 0.1961231470923603,
+      "grad_norm": 21.5,
+      "learning_rate": 1.931818181818182e-06,
+      "loss": 1.7188,
+      "mean_token_accuracy": 0.7101240158081055,
+      "num_tokens": 9215544.0,
+      "step": 86
+    },
+    {
+      "epoch": 0.19840364880273662,
+      "grad_norm": 21.625,
+      "learning_rate": 1.954545454545455e-06,
+      "loss": 1.7445,
+      "mean_token_accuracy": 0.7046354413032532,
+      "num_tokens": 9322163.0,
+      "step": 87
+    },
+    {
+      "epoch": 0.2006841505131129,
+      "grad_norm": 31.75,
+      "learning_rate": 1.977272727272727e-06,
+      "loss": 1.6605,
+      "mean_token_accuracy": 0.7182554006576538,
+      "num_tokens": 9428959.0,
+      "step": 88
+    },
+    {
+      "epoch": 0.20296465222348917,
+      "grad_norm": 20.625,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.6955,
+      "mean_token_accuracy": 0.7105437815189362,
+      "num_tokens": 9535594.0,
+      "step": 89
+    },
+    {
+      "epoch": 0.20524515393386544,
+      "grad_norm": 20.5,
+      "learning_rate": 2.022727272727273e-06,
+      "loss": 1.6534,
+      "mean_token_accuracy": 0.7167372107505798,
+      "num_tokens": 9642716.0,
+      "step": 90
+    },
+    {
+      "epoch": 0.20752565564424175,
+      "grad_norm": 20.125,
+      "learning_rate": 2.0454545454545457e-06,
+      "loss": 1.6508,
+      "mean_token_accuracy": 0.7186834812164307,
+      "num_tokens": 9749032.0,
+      "step": 91
+    },
+    {
+      "epoch": 0.20980615735461802,
+      "grad_norm": 19.625,
+      "learning_rate": 2.0681818181818184e-06,
+      "loss": 1.6197,
+      "mean_token_accuracy": 0.7244506776332855,
+      "num_tokens": 9855340.0,
+      "step": 92
+    },
+    {
+      "epoch": 0.2120866590649943,
+      "grad_norm": 19.5,
+      "learning_rate": 2.090909090909091e-06,
+      "loss": 1.5969,
+      "mean_token_accuracy": 0.7231418937444687,
+      "num_tokens": 9962265.0,
+      "step": 93
+    },
+    {
+      "epoch": 0.21436716077537057,
+      "grad_norm": 18.5,
+      "learning_rate": 2.113636363636364e-06,
+      "loss": 1.5364,
+      "mean_token_accuracy": 0.7311695367097855,
+      "num_tokens": 10069156.0,
+      "step": 94
+    },
+    {
+      "epoch": 0.21664766248574688,
+      "grad_norm": 17.625,
+      "learning_rate": 2.1363636363636365e-06,
+      "loss": 1.4815,
+      "mean_token_accuracy": 0.7454914003610611,
+      "num_tokens": 10176688.0,
+      "step": 95
+    },
+    {
+      "epoch": 0.21892816419612315,
+      "grad_norm": 17.625,
+      "learning_rate": 2.1590909090909092e-06,
+      "loss": 1.4978,
+      "mean_token_accuracy": 0.7258649468421936,
+      "num_tokens": 10283306.0,
+      "step": 96
+    },
+    {
+      "epoch": 0.22120866590649943,
+      "grad_norm": 17.5,
+      "learning_rate": 2.181818181818182e-06,
+      "loss": 1.4677,
+      "mean_token_accuracy": 0.7376868277788162,
+      "num_tokens": 10390084.0,
+      "step": 97
+    },
+    {
+      "epoch": 0.2234891676168757,
+      "grad_norm": 17.125,
+      "learning_rate": 2.2045454545454547e-06,
+      "loss": 1.4221,
+      "mean_token_accuracy": 0.7373760044574738,
+      "num_tokens": 10497627.0,
+      "step": 98
+    },
+    {
+      "epoch": 0.22576966932725198,
+      "grad_norm": 23.875,
+      "learning_rate": 2.2272727272727274e-06,
+      "loss": 1.3852,
+      "mean_token_accuracy": 0.7427248060703278,
+      "num_tokens": 10604483.0,
+      "step": 99
+    },
+    {
+      "epoch": 0.22805017103762829,
+      "grad_norm": 18.0,
+      "learning_rate": 2.25e-06,
+      "loss": 1.3928,
+      "mean_token_accuracy": 0.7454716116189957,
+      "num_tokens": 10711923.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.23033067274800456,
+      "grad_norm": 19.375,
+      "learning_rate": 2.2727272727272728e-06,
+      "loss": 1.3648,
+      "mean_token_accuracy": 0.749587893486023,
+      "num_tokens": 10818737.0,
+      "step": 101
+    },
+    {
+      "epoch": 0.23261117445838084,
+      "grad_norm": 21.5,
+      "learning_rate": 2.295454545454546e-06,
+      "loss": 1.3399,
+      "mean_token_accuracy": 0.7486944496631622,
+      "num_tokens": 10925328.0,
+      "step": 102
+    },
+    {
+      "epoch": 0.2348916761687571,
+      "grad_norm": 14.9375,
+      "learning_rate": 2.318181818181818e-06,
+      "loss": 1.2484,
+      "mean_token_accuracy": 0.7670646905899048,
+      "num_tokens": 11032717.0,
+      "step": 103
+    },
+    {
+      "epoch": 0.23717217787913342,
+      "grad_norm": 14.9375,
+      "learning_rate": 2.3409090909090913e-06,
+      "loss": 1.2661,
+      "mean_token_accuracy": 0.7583093047142029,
+      "num_tokens": 11139345.0,
+      "step": 104
+    },
+    {
+      "epoch": 0.2394526795895097,
+      "grad_norm": 14.375,
+      "learning_rate": 2.363636363636364e-06,
+      "loss": 1.2064,
+      "mean_token_accuracy": 0.7690076977014542,
+      "num_tokens": 11245962.0,
+      "step": 105
+    },
+    {
+      "epoch": 0.24173318129988597,
+      "grad_norm": 13.9375,
+      "learning_rate": 2.3863636363636367e-06,
+      "loss": 1.2029,
+      "mean_token_accuracy": 0.7617898732423782,
+      "num_tokens": 11352972.0,
+      "step": 106
+    },
+    {
+      "epoch": 0.24401368301026224,
+      "grad_norm": 13.75,
+      "learning_rate": 2.4090909090909094e-06,
+      "loss": 1.153,
+      "mean_token_accuracy": 0.7770701050758362,
+      "num_tokens": 11460132.0,
+      "step": 107
+    },
+    {
+      "epoch": 0.24629418472063855,
+      "grad_norm": 13.8125,
+      "learning_rate": 2.431818181818182e-06,
+      "loss": 1.1218,
+      "mean_token_accuracy": 0.7785277962684631,
+      "num_tokens": 11567105.0,
+      "step": 108
+    },
+    {
+      "epoch": 0.24857468643101482,
+      "grad_norm": 12.5625,
+      "learning_rate": 2.454545454545455e-06,
+      "loss": 1.0815,
+      "mean_token_accuracy": 0.7852117866277695,
+      "num_tokens": 11674498.0,
+      "step": 109
+    },
+    {
+      "epoch": 0.2508551881413911,
+      "grad_norm": 12.625,
+      "learning_rate": 2.4772727272727275e-06,
+      "loss": 1.0581,
+      "mean_token_accuracy": 0.7985803633928299,
+      "num_tokens": 11780696.0,
+      "step": 110
+    },
+    {
+      "epoch": 0.2531356898517674,
+      "grad_norm": 12.0,
+      "learning_rate": 2.5e-06,
+      "loss": 1.0621,
+      "mean_token_accuracy": 0.7939664274454117,
+      "num_tokens": 11887719.0,
+      "step": 111
+    },
+    {
+      "epoch": 0.2554161915621437,
+      "grad_norm": 10.875,
+      "learning_rate": 2.522727272727273e-06,
+      "loss": 0.9893,
+      "mean_token_accuracy": 0.812217116355896,
+      "num_tokens": 11994518.0,
+      "step": 112
+    },
+    {
+      "epoch": 0.2576966932725199,
+      "grad_norm": 10.4375,
+      "learning_rate": 2.5454545454545456e-06,
+      "loss": 1.009,
+      "mean_token_accuracy": 0.8142502456903458,
+      "num_tokens": 12101625.0,
+      "step": 113
+    },
+    {
+      "epoch": 0.25997719498289623,
+      "grad_norm": 9.5625,
+      "learning_rate": 2.5681818181818187e-06,
+      "loss": 1.0013,
+      "mean_token_accuracy": 0.8154999315738678,
+      "num_tokens": 12208599.0,
+      "step": 114
+    },
+    {
+      "epoch": 0.26225769669327254,
+      "grad_norm": 9.0625,
+      "learning_rate": 2.590909090909091e-06,
+      "loss": 0.9603,
+      "mean_token_accuracy": 0.8178321123123169,
+      "num_tokens": 12315919.0,
+      "step": 115
+    },
+    {
+      "epoch": 0.2645381984036488,
+      "grad_norm": 8.25,
+      "learning_rate": 2.6136363636363637e-06,
+      "loss": 0.9488,
+      "mean_token_accuracy": 0.8180135637521744,
+      "num_tokens": 12422670.0,
+      "step": 116
+    },
+    {
+      "epoch": 0.2668187001140251,
+      "grad_norm": 7.625,
+      "learning_rate": 2.6363636363636364e-06,
+      "loss": 0.9013,
+      "mean_token_accuracy": 0.8282536566257477,
+      "num_tokens": 12529416.0,
+      "step": 117
+    },
+    {
+      "epoch": 0.2690992018244014,
+      "grad_norm": 13.1875,
+      "learning_rate": 2.6590909090909095e-06,
+      "loss": 0.8886,
+      "mean_token_accuracy": 0.8260611891746521,
+      "num_tokens": 12636798.0,
+      "step": 118
+    },
+    {
+      "epoch": 0.27137970353477764,
+      "grad_norm": 6.9375,
+      "learning_rate": 2.6818181818181822e-06,
+      "loss": 0.896,
+      "mean_token_accuracy": 0.8283491432666779,
+      "num_tokens": 12743702.0,
+      "step": 119
+    },
+    {
+      "epoch": 0.27366020524515394,
+      "grad_norm": 6.6875,
+      "learning_rate": 2.7045454545454545e-06,
+      "loss": 0.9264,
+      "mean_token_accuracy": 0.8232417404651642,
+      "num_tokens": 12851550.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.2759407069555302,
+      "grad_norm": 6.0625,
+      "learning_rate": 2.7272727272727272e-06,
+      "loss": 0.8567,
+      "mean_token_accuracy": 0.8310881853103638,
+      "num_tokens": 12958606.0,
+      "step": 121
+    },
+    {
+      "epoch": 0.2782212086659065,
+      "grad_norm": 5.21875,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.8334,
+      "mean_token_accuracy": 0.8362232446670532,
+      "num_tokens": 13065521.0,
+      "step": 122
+    },
+    {
+      "epoch": 0.2805017103762828,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.772727272727273e-06,
+      "loss": 0.8364,
+      "mean_token_accuracy": 0.8403038382530212,
+      "num_tokens": 13172067.0,
+      "step": 123
+    },
+    {
+      "epoch": 0.28278221208665905,
+      "grad_norm": 4.75,
+      "learning_rate": 2.7954545454545458e-06,
+      "loss": 0.8005,
+      "mean_token_accuracy": 0.8477163910865784,
+      "num_tokens": 13279427.0,
+      "step": 124
+    },
+    {
+      "epoch": 0.28506271379703535,
+      "grad_norm": 4.78125,
+      "learning_rate": 2.818181818181818e-06,
+      "loss": 0.8036,
+      "mean_token_accuracy": 0.8398231416940689,
+      "num_tokens": 13386332.0,
+      "step": 125
+    },
+    {
+      "epoch": 0.28734321550741165,
+      "grad_norm": 4.625,
+      "learning_rate": 2.8409090909090916e-06,
+      "loss": 0.7985,
+      "mean_token_accuracy": 0.8417368233203888,
+      "num_tokens": 13493485.0,
+      "step": 126
+    },
+    {
+      "epoch": 0.2896237172177879,
+      "grad_norm": 4.40625,
+      "learning_rate": 2.863636363636364e-06,
+      "loss": 0.8215,
+      "mean_token_accuracy": 0.8364972919225693,
+      "num_tokens": 13600526.0,
+      "step": 127
+    },
+    {
+      "epoch": 0.2919042189281642,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.8863636363636366e-06,
+      "loss": 0.7888,
+      "mean_token_accuracy": 0.8411660343408585,
+      "num_tokens": 13707335.0,
+      "step": 128
+    },
+    {
+      "epoch": 0.29418472063854045,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.9090909090909093e-06,
+      "loss": 0.7955,
+      "mean_token_accuracy": 0.840514287352562,
+      "num_tokens": 13814123.0,
+      "step": 129
+    },
+    {
+      "epoch": 0.29646522234891676,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.931818181818182e-06,
+      "loss": 0.7764,
+      "mean_token_accuracy": 0.8397897183895111,
+      "num_tokens": 13921444.0,
+      "step": 130
+    },
+    {
+      "epoch": 0.29874572405929306,
+      "grad_norm": 3.84375,
+      "learning_rate": 2.954545454545455e-06,
+      "loss": 0.7721,
+      "mean_token_accuracy": 0.8480268269777298,
+      "num_tokens": 14028739.0,
+      "step": 131
+    },
+    {
+      "epoch": 0.3010262257696693,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.9772727272727274e-06,
+      "loss": 0.7483,
+      "mean_token_accuracy": 0.8492945581674576,
+      "num_tokens": 14136015.0,
+      "step": 132
+    },
+    {
+      "epoch": 0.3033067274800456,
+      "grad_norm": 4.46875,
+      "learning_rate": 3e-06,
+      "loss": 0.7765,
+      "mean_token_accuracy": 0.8442358523607254,
+      "num_tokens": 14243016.0,
+      "step": 133
+    },
+    {
+      "epoch": 0.3055872291904219,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.0227272727272728e-06,
+      "loss": 0.7711,
+      "mean_token_accuracy": 0.8423920571804047,
+      "num_tokens": 14350319.0,
+      "step": 134
+    },
+    {
+      "epoch": 0.30786773090079816,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.045454545454546e-06,
+      "loss": 0.767,
+      "mean_token_accuracy": 0.845877543091774,
+      "num_tokens": 14457699.0,
+      "step": 135
+    },
+    {
+      "epoch": 0.31014823261117447,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.0681818181818186e-06,
+      "loss": 0.7406,
+      "mean_token_accuracy": 0.8514658808708191,
+      "num_tokens": 14565228.0,
+      "step": 136
+    },
+    {
+      "epoch": 0.3124287343215507,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.090909090909091e-06,
+      "loss": 0.7493,
+      "mean_token_accuracy": 0.8524320423603058,
+      "num_tokens": 14671853.0,
+      "step": 137
+    },
+    {
+      "epoch": 0.314709236031927,
+      "grad_norm": 3.625,
+      "learning_rate": 3.1136363636363636e-06,
+      "loss": 0.7529,
+      "mean_token_accuracy": 0.8469192087650299,
+      "num_tokens": 14778717.0,
+      "step": 138
+    },
+    {
+      "epoch": 0.3169897377423033,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.1363636363636367e-06,
+      "loss": 0.751,
+      "mean_token_accuracy": 0.847698837518692,
+      "num_tokens": 14885625.0,
+      "step": 139
+    },
+    {
+      "epoch": 0.31927023945267957,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.1590909090909094e-06,
+      "loss": 0.7293,
+      "mean_token_accuracy": 0.8514856845140457,
+      "num_tokens": 14992824.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.3215507411630559,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.181818181818182e-06,
+      "loss": 0.7383,
+      "mean_token_accuracy": 0.8546594232320786,
+      "num_tokens": 15100514.0,
+      "step": 141
+    },
+    {
+      "epoch": 0.3238312428734322,
+      "grad_norm": 6.03125,
+      "learning_rate": 3.204545454545455e-06,
+      "loss": 0.7278,
+      "mean_token_accuracy": 0.8488231599330902,
+      "num_tokens": 15208020.0,
+      "step": 142
+    },
+    {
+      "epoch": 0.3261117445838084,
+      "grad_norm": 4.875,
+      "learning_rate": 3.227272727272728e-06,
+      "loss": 0.7396,
+      "mean_token_accuracy": 0.8457832187414169,
+      "num_tokens": 15314965.0,
+      "step": 143
+    },
+    {
+      "epoch": 0.32839224629418473,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.7361,
+      "mean_token_accuracy": 0.8558014631271362,
+      "num_tokens": 15421585.0,
+      "step": 144
+    },
+    {
+      "epoch": 0.330672748004561,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.272727272727273e-06,
+      "loss": 0.7336,
+      "mean_token_accuracy": 0.8494018763303757,
+      "num_tokens": 15528412.0,
+      "step": 145
+    },
+    {
+      "epoch": 0.3329532497149373,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.2954545454545456e-06,
+      "loss": 0.7171,
+      "mean_token_accuracy": 0.8540607988834381,
+      "num_tokens": 15635689.0,
+      "step": 146
+    },
+    {
+      "epoch": 0.3352337514253136,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.3181818181818188e-06,
+      "loss": 0.6972,
+      "mean_token_accuracy": 0.856657013297081,
+      "num_tokens": 15742801.0,
+      "step": 147
+    },
+    {
+      "epoch": 0.33751425313568983,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.3409090909090915e-06,
+      "loss": 0.7182,
+      "mean_token_accuracy": 0.8510304987430573,
+      "num_tokens": 15850125.0,
+      "step": 148
+    },
+    {
+      "epoch": 0.33979475484606614,
+      "grad_norm": 3.625,
+      "learning_rate": 3.3636363636363637e-06,
+      "loss": 0.7362,
+      "mean_token_accuracy": 0.84682796895504,
+      "num_tokens": 15956807.0,
+      "step": 149
+    },
+    {
+      "epoch": 0.34207525655644244,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.3863636363636364e-06,
+      "loss": 0.7022,
+      "mean_token_accuracy": 0.85152368247509,
+      "num_tokens": 16063797.0,
+      "step": 150
+    },
+    {
+      "epoch": 0.3443557582668187,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.409090909090909e-06,
+      "loss": 0.6747,
+      "mean_token_accuracy": 0.8609962910413742,
+      "num_tokens": 16170407.0,
+      "step": 151
+    },
+    {
+      "epoch": 0.346636259977195,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.4318181818181823e-06,
+      "loss": 0.7054,
+      "mean_token_accuracy": 0.8567307740449905,
+      "num_tokens": 16276735.0,
+      "step": 152
+    },
+    {
+      "epoch": 0.34891676168757124,
+      "grad_norm": 3.375,
+      "learning_rate": 3.454545454545455e-06,
+      "loss": 0.6954,
+      "mean_token_accuracy": 0.8543938845396042,
+      "num_tokens": 16383665.0,
+      "step": 153
+    },
+    {
+      "epoch": 0.35119726339794755,
+      "grad_norm": 3.984375,
+      "learning_rate": 3.4772727272727277e-06,
+      "loss": 0.7353,
+      "mean_token_accuracy": 0.8494099676609039,
+      "num_tokens": 16490654.0,
+      "step": 154
+    },
+    {
+      "epoch": 0.35347776510832385,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.5e-06,
+      "loss": 0.6977,
+      "mean_token_accuracy": 0.8554576933383942,
+      "num_tokens": 16597928.0,
+      "step": 155
+    },
+    {
+      "epoch": 0.3557582668187001,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.522727272727273e-06,
+      "loss": 0.666,
+      "mean_token_accuracy": 0.8644505739212036,
+      "num_tokens": 16705070.0,
+      "step": 156
+    },
+    {
+      "epoch": 0.3580387685290764,
+      "grad_norm": 4.6875,
+      "learning_rate": 3.5454545454545458e-06,
+      "loss": 0.7023,
+      "mean_token_accuracy": 0.8530746251344681,
+      "num_tokens": 16812219.0,
+      "step": 157
+    },
+    {
+      "epoch": 0.3603192702394527,
+      "grad_norm": 3.125,
+      "learning_rate": 3.5681818181818185e-06,
+      "loss": 0.6717,
+      "mean_token_accuracy": 0.8598745912313461,
+      "num_tokens": 16919765.0,
+      "step": 158
+    },
+    {
+      "epoch": 0.36259977194982895,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.590909090909091e-06,
+      "loss": 0.7021,
+      "mean_token_accuracy": 0.8543113619089127,
+      "num_tokens": 17026571.0,
+      "step": 159
+    },
+    {
+      "epoch": 0.36488027366020526,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.6136363636363643e-06,
+      "loss": 0.7072,
+      "mean_token_accuracy": 0.8556502610445023,
+      "num_tokens": 17133437.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.3671607753705815,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.6363636363636366e-06,
+      "loss": 0.686,
+      "mean_token_accuracy": 0.859677642583847,
+      "num_tokens": 17240745.0,
+      "step": 161
+    },
+    {
+      "epoch": 0.3694412770809578,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.6590909090909093e-06,
+      "loss": 0.6827,
+      "mean_token_accuracy": 0.8606048673391342,
+      "num_tokens": 17347700.0,
+      "step": 162
+    },
+    {
+      "epoch": 0.3717217787913341,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.681818181818182e-06,
+      "loss": 0.6758,
+      "mean_token_accuracy": 0.8614284843206406,
+      "num_tokens": 17455182.0,
+      "step": 163
+    },
+    {
+      "epoch": 0.37400228050171036,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.704545454545455e-06,
+      "loss": 0.6935,
+      "mean_token_accuracy": 0.8534423410892487,
+      "num_tokens": 17562182.0,
+      "step": 164
+    },
+    {
+      "epoch": 0.37628278221208666,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.727272727272728e-06,
+      "loss": 0.7017,
+      "mean_token_accuracy": 0.855252206325531,
+      "num_tokens": 17668986.0,
+      "step": 165
+    },
+    {
+      "epoch": 0.37856328392246297,
+      "grad_norm": 2.625,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.6615,
+      "mean_token_accuracy": 0.8632187992334366,
+      "num_tokens": 17776052.0,
+      "step": 166
+    },
+    {
+      "epoch": 0.3808437856328392,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.772727272727273e-06,
+      "loss": 0.6868,
+      "mean_token_accuracy": 0.854138657450676,
+      "num_tokens": 17883630.0,
+      "step": 167
+    },
+    {
+      "epoch": 0.3831242873432155,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.7954545454545455e-06,
+      "loss": 0.6734,
+      "mean_token_accuracy": 0.8595366477966309,
+      "num_tokens": 17990666.0,
+      "step": 168
+    },
+    {
+      "epoch": 0.38540478905359177,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.818181818181819e-06,
+      "loss": 0.6898,
+      "mean_token_accuracy": 0.8539352118968964,
+      "num_tokens": 18097695.0,
+      "step": 169
+    },
+    {
+      "epoch": 0.38768529076396807,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.840909090909091e-06,
+      "loss": 0.691,
+      "mean_token_accuracy": 0.8536562919616699,
+      "num_tokens": 18204503.0,
+      "step": 170
+    },
+    {
+      "epoch": 0.3899657924743444,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.863636363636364e-06,
+      "loss": 0.6677,
+      "mean_token_accuracy": 0.8611757457256317,
+      "num_tokens": 18311196.0,
+      "step": 171
+    },
+    {
+      "epoch": 0.3922462941847206,
+      "grad_norm": 3.125,
+      "learning_rate": 3.886363636363637e-06,
+      "loss": 0.7125,
+      "mean_token_accuracy": 0.8503219485282898,
+      "num_tokens": 18417786.0,
+      "step": 172
+    },
+    {
+      "epoch": 0.3945267958950969,
+      "grad_norm": 2.0625,
+      "learning_rate": 3.90909090909091e-06,
+      "loss": 0.6665,
+      "mean_token_accuracy": 0.8595822900533676,
+      "num_tokens": 18525225.0,
+      "step": 173
+    },
+    {
+      "epoch": 0.39680729760547323,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.931818181818182e-06,
+      "loss": 0.6908,
+      "mean_token_accuracy": 0.8568900525569916,
+      "num_tokens": 18632231.0,
+      "step": 174
+    },
+    {
+      "epoch": 0.3990877993158495,
+      "grad_norm": 2.3125,
+      "learning_rate": 3.954545454545454e-06,
+      "loss": 0.6955,
+      "mean_token_accuracy": 0.8539032638072968,
+      "num_tokens": 18739610.0,
+      "step": 175
+    },
+    {
+      "epoch": 0.4013683010262258,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.9772727272727275e-06,
+      "loss": 0.6726,
+      "mean_token_accuracy": 0.8602974861860275,
+      "num_tokens": 18846794.0,
+      "step": 176
+    },
+    {
+      "epoch": 0.40364880273660203,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.6678,
+      "mean_token_accuracy": 0.8603723049163818,
+      "num_tokens": 18953693.0,
+      "step": 177
+    },
+    {
+      "epoch": 0.40592930444697833,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.022727272727273e-06,
+      "loss": 0.6585,
+      "mean_token_accuracy": 0.8616438657045364,
+      "num_tokens": 19060250.0,
+      "step": 178
+    },
+    {
+      "epoch": 0.40820980615735464,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.045454545454546e-06,
+      "loss": 0.6679,
+      "mean_token_accuracy": 0.8587404191493988,
+      "num_tokens": 19167003.0,
+      "step": 179
+    },
+    {
+      "epoch": 0.4104903078677309,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.068181818181818e-06,
+      "loss": 0.665,
+      "mean_token_accuracy": 0.8621865957975388,
+      "num_tokens": 19274555.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.4127708095781072,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.0909090909090915e-06,
+      "loss": 0.6714,
+      "mean_token_accuracy": 0.8582023978233337,
+      "num_tokens": 19381948.0,
+      "step": 181
+    },
+    {
+      "epoch": 0.4150513112884835,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.113636363636364e-06,
+      "loss": 0.6432,
+      "mean_token_accuracy": 0.8636541664600372,
+      "num_tokens": 19489531.0,
+      "step": 182
+    },
+    {
+      "epoch": 0.41733181299885974,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.136363636363637e-06,
+      "loss": 0.6553,
+      "mean_token_accuracy": 0.8614503294229507,
+      "num_tokens": 19596788.0,
+      "step": 183
+    },
+    {
+      "epoch": 0.41961231470923605,
+      "grad_norm": 2.375,
+      "learning_rate": 4.159090909090909e-06,
+      "loss": 0.6557,
+      "mean_token_accuracy": 0.865259125828743,
+      "num_tokens": 19703381.0,
+      "step": 184
+    },
+    {
+      "epoch": 0.4218928164196123,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.181818181818182e-06,
+      "loss": 0.675,
+      "mean_token_accuracy": 0.8583291471004486,
+      "num_tokens": 19810512.0,
+      "step": 185
+    },
+    {
+      "epoch": 0.4241733181299886,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.204545454545455e-06,
+      "loss": 0.6707,
+      "mean_token_accuracy": 0.8560606837272644,
+      "num_tokens": 19917996.0,
+      "step": 186
+    },
+    {
+      "epoch": 0.4264538198403649,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.227272727272728e-06,
+      "loss": 0.6663,
+      "mean_token_accuracy": 0.8619653731584549,
+      "num_tokens": 20025261.0,
+      "step": 187
+    },
+    {
+      "epoch": 0.42873432155074115,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.25e-06,
+      "loss": 0.6772,
+      "mean_token_accuracy": 0.8561718165874481,
+      "num_tokens": 20131720.0,
+      "step": 188
+    },
+    {
+      "epoch": 0.43101482326111745,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.272727272727273e-06,
+      "loss": 0.6727,
+      "mean_token_accuracy": 0.8626906722784042,
+      "num_tokens": 20238797.0,
+      "step": 189
+    },
+    {
+      "epoch": 0.43329532497149376,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.295454545454546e-06,
+      "loss": 0.6861,
+      "mean_token_accuracy": 0.8580966591835022,
+      "num_tokens": 20345319.0,
+      "step": 190
+    },
+    {
+      "epoch": 0.43557582668187,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.3181818181818185e-06,
+      "loss": 0.6749,
+      "mean_token_accuracy": 0.8572024405002594,
+      "num_tokens": 20452421.0,
+      "step": 191
+    },
+    {
+      "epoch": 0.4378563283922463,
+      "grad_norm": 4.0,
+      "learning_rate": 4.340909090909091e-06,
+      "loss": 0.6529,
+      "mean_token_accuracy": 0.862261489033699,
+      "num_tokens": 20559771.0,
+      "step": 192
+    },
+    {
+      "epoch": 0.44013683010262256,
+      "grad_norm": 2.75,
+      "learning_rate": 4.363636363636364e-06,
+      "loss": 0.6361,
+      "mean_token_accuracy": 0.8651402294635773,
+      "num_tokens": 20667009.0,
+      "step": 193
+    },
+    {
+      "epoch": 0.44241733181299886,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.386363636363637e-06,
+      "loss": 0.6849,
+      "mean_token_accuracy": 0.8569408059120178,
+      "num_tokens": 20774389.0,
+      "step": 194
+    },
+    {
+      "epoch": 0.44469783352337516,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.409090909090909e-06,
+      "loss": 0.667,
+      "mean_token_accuracy": 0.8601485341787338,
+      "num_tokens": 20881703.0,
+      "step": 195
+    },
+    {
+      "epoch": 0.4469783352337514,
+      "grad_norm": 3.875,
+      "learning_rate": 4.4318181818181824e-06,
+      "loss": 0.656,
+      "mean_token_accuracy": 0.8644906729459763,
+      "num_tokens": 20988710.0,
+      "step": 196
+    },
+    {
+      "epoch": 0.4492588369441277,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.454545454545455e-06,
+      "loss": 0.6577,
+      "mean_token_accuracy": 0.8626122176647186,
+      "num_tokens": 21095340.0,
+      "step": 197
+    },
+    {
+      "epoch": 0.45153933865450396,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.477272727272728e-06,
+      "loss": 0.6547,
+      "mean_token_accuracy": 0.8651628345251083,
+      "num_tokens": 21202493.0,
+      "step": 198
+    },
+    {
+      "epoch": 0.45381984036488027,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.5e-06,
+      "loss": 0.6607,
+      "mean_token_accuracy": 0.8622200191020966,
+      "num_tokens": 21309591.0,
+      "step": 199
+    },
+    {
+      "epoch": 0.45610034207525657,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.522727272727273e-06,
+      "loss": 0.6405,
+      "mean_token_accuracy": 0.8680934458971024,
+      "num_tokens": 21417007.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4583808437856328,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.5454545454545455e-06,
+      "loss": 0.6589,
+      "mean_token_accuracy": 0.8650902211666107,
+      "num_tokens": 21524653.0,
+      "step": 201
+    },
+    {
+      "epoch": 0.4606613454960091,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.568181818181819e-06,
+      "loss": 0.6672,
+      "mean_token_accuracy": 0.8640954792499542,
+      "num_tokens": 21631985.0,
+      "step": 202
+    },
+    {
+      "epoch": 0.4629418472063854,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.590909090909092e-06,
+      "loss": 0.6749,
+      "mean_token_accuracy": 0.8592006415128708,
+      "num_tokens": 21738904.0,
+      "step": 203
+    },
+    {
+      "epoch": 0.4652223489167617,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.613636363636364e-06,
+      "loss": 0.6362,
+      "mean_token_accuracy": 0.8670714944601059,
+      "num_tokens": 21846587.0,
+      "step": 204
+    },
+    {
+      "epoch": 0.467502850627138,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.636363636363636e-06,
+      "loss": 0.6724,
+      "mean_token_accuracy": 0.861112967133522,
+      "num_tokens": 21953786.0,
+      "step": 205
+    },
+    {
+      "epoch": 0.4697833523375142,
+      "grad_norm": 4.125,
+      "learning_rate": 4.6590909090909095e-06,
+      "loss": 0.6757,
+      "mean_token_accuracy": 0.860434964299202,
+      "num_tokens": 22061456.0,
+      "step": 206
+    },
+    {
+      "epoch": 0.47206385404789053,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.681818181818183e-06,
+      "loss": 0.6586,
+      "mean_token_accuracy": 0.8619724959135056,
+      "num_tokens": 22169315.0,
+      "step": 207
+    },
+    {
+      "epoch": 0.47434435575826683,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.704545454545455e-06,
+      "loss": 0.6591,
+      "mean_token_accuracy": 0.8640032708644867,
+      "num_tokens": 22276762.0,
+      "step": 208
+    },
+    {
+      "epoch": 0.4766248574686431,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.727272727272728e-06,
+      "loss": 0.6554,
+      "mean_token_accuracy": 0.8638423681259155,
+      "num_tokens": 22383321.0,
+      "step": 209
+    },
+    {
+      "epoch": 0.4789053591790194,
+      "grad_norm": 2.03125,
+      "learning_rate": 4.75e-06,
+      "loss": 0.6595,
+      "mean_token_accuracy": 0.8663541227579117,
+      "num_tokens": 22490163.0,
+      "step": 210
+    },
+    {
+      "epoch": 0.4811858608893957,
+      "grad_norm": 1.8984375,
+      "learning_rate": 4.772727272727273e-06,
+      "loss": 0.6458,
+      "mean_token_accuracy": 0.8662288784980774,
+      "num_tokens": 22596745.0,
+      "step": 211
+    },
+    {
+      "epoch": 0.48346636259977194,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.795454545454546e-06,
+      "loss": 0.6634,
+      "mean_token_accuracy": 0.8609149158000946,
+      "num_tokens": 22704241.0,
+      "step": 212
+    },
+    {
+      "epoch": 0.48574686431014824,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.818181818181819e-06,
+      "loss": 0.6559,
+      "mean_token_accuracy": 0.8637297302484512,
+      "num_tokens": 22810936.0,
+      "step": 213
+    },
+    {
+      "epoch": 0.4880273660205245,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.840909090909091e-06,
+      "loss": 0.6616,
+      "mean_token_accuracy": 0.8652000278234482,
+      "num_tokens": 22917635.0,
+      "step": 214
+    },
+    {
+      "epoch": 0.4903078677309008,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.863636363636364e-06,
+      "loss": 0.6662,
+      "mean_token_accuracy": 0.8612215965986252,
+      "num_tokens": 23024940.0,
+      "step": 215
+    },
+    {
+      "epoch": 0.4925883694412771,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.8863636363636365e-06,
+      "loss": 0.6785,
+      "mean_token_accuracy": 0.8561449646949768,
+      "num_tokens": 23132260.0,
+      "step": 216
+    },
+    {
+      "epoch": 0.49486887115165334,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.90909090909091e-06,
+      "loss": 0.6614,
+      "mean_token_accuracy": 0.8678651452064514,
+      "num_tokens": 23239311.0,
+      "step": 217
+    },
+    {
+      "epoch": 0.49714937286202965,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.931818181818182e-06,
+      "loss": 0.6485,
+      "mean_token_accuracy": 0.8589539378881454,
+      "num_tokens": 23346340.0,
+      "step": 218
+    },
+    {
+      "epoch": 0.49942987457240595,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.954545454545455e-06,
+      "loss": 0.6416,
+      "mean_token_accuracy": 0.8672493249177933,
+      "num_tokens": 23454269.0,
+      "step": 219
+    },
+    {
+      "epoch": 0.5017103762827823,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.977272727272728e-06,
+      "loss": 0.6669,
+      "mean_token_accuracy": 0.8612665235996246,
+      "num_tokens": 23560711.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5017103762827823,
+      "eval_loss": 0.6530081033706665,
+      "eval_mean_token_accuracy": 0.8651766348701013,
+      "eval_num_tokens": 23560711.0,
+      "eval_runtime": 58.5845,
+      "eval_samples_per_second": 143.126,
+      "eval_steps_per_second": 4.489,
+      "step": 220
+    },
+    {
+      "epoch": 0.5039908779931584,
+      "grad_norm": 2.296875,
+      "learning_rate": 5e-06,
+      "loss": 0.62,
+      "mean_token_accuracy": 0.8713521808385849,
+      "num_tokens": 23668025.0,
+      "step": 221
+    },
+    {
+      "epoch": 0.5062713797035348,
+      "grad_norm": 1.890625,
+      "learning_rate": 4.999999290524132e-06,
+      "loss": 0.6252,
+      "mean_token_accuracy": 0.8714602589607239,
+      "num_tokens": 23775006.0,
+      "step": 222
+    },
+    {
+      "epoch": 0.508551881413911,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.999997162096932e-06,
+      "loss": 0.6736,
+      "mean_token_accuracy": 0.8643509894609451,
+      "num_tokens": 23882355.0,
+      "step": 223
+    },
+    {
+      "epoch": 0.5108323831242874,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.999993614719606e-06,
+      "loss": 0.6412,
+      "mean_token_accuracy": 0.8682572692632675,
+      "num_tokens": 23989600.0,
+      "step": 224
+    },
+    {
+      "epoch": 0.5131128848346637,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.999988648394169e-06,
+      "loss": 0.6096,
+      "mean_token_accuracy": 0.8725505918264389,
+      "num_tokens": 24097243.0,
+      "step": 225
+    },
+    {
+      "epoch": 0.5153933865450399,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.99998226312344e-06,
+      "loss": 0.6531,
+      "mean_token_accuracy": 0.8634347319602966,
+      "num_tokens": 24203953.0,
+      "step": 226
+    },
+    {
+      "epoch": 0.5176738882554162,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.999974458911041e-06,
+      "loss": 0.6314,
+      "mean_token_accuracy": 0.8699511885643005,
+      "num_tokens": 24310863.0,
+      "step": 227
+    },
+    {
+      "epoch": 0.5199543899657925,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.999965235761404e-06,
+      "loss": 0.6486,
+      "mean_token_accuracy": 0.8652396500110626,
+      "num_tokens": 24417755.0,
+      "step": 228
+    },
+    {
+      "epoch": 0.5222348916761688,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.999954593679762e-06,
+      "loss": 0.659,
+      "mean_token_accuracy": 0.864411011338234,
+      "num_tokens": 24524785.0,
+      "step": 229
+    },
+    {
+      "epoch": 0.5245153933865451,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.999942532672157e-06,
+      "loss": 0.688,
+      "mean_token_accuracy": 0.854931503534317,
+      "num_tokens": 24631291.0,
+      "step": 230
+    },
+    {
+      "epoch": 0.5267958950969214,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.999929052745434e-06,
+      "loss": 0.6541,
+      "mean_token_accuracy": 0.8636079728603363,
+      "num_tokens": 24738568.0,
+      "step": 231
+    },
+    {
+      "epoch": 0.5290763968072976,
+      "grad_norm": 3.953125,
+      "learning_rate": 4.999914153907243e-06,
+      "loss": 0.6538,
+      "mean_token_accuracy": 0.8641977459192276,
+      "num_tokens": 24845646.0,
+      "step": 232
+    },
+    {
+      "epoch": 0.5313568985176739,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.999897836166041e-06,
+      "loss": 0.6473,
+      "mean_token_accuracy": 0.8661595731973648,
+      "num_tokens": 24952336.0,
+      "step": 233
+    },
+    {
+      "epoch": 0.5336374002280502,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.999880099531089e-06,
+      "loss": 0.6576,
+      "mean_token_accuracy": 0.8643601685762405,
+      "num_tokens": 25058478.0,
+      "step": 234
+    },
+    {
+      "epoch": 0.5359179019384265,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.999860944012455e-06,
+      "loss": 0.6463,
+      "mean_token_accuracy": 0.8663338273763657,
+      "num_tokens": 25165512.0,
+      "step": 235
+    },
+    {
+      "epoch": 0.5381984036488028,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.999840369621011e-06,
+      "loss": 0.6453,
+      "mean_token_accuracy": 0.8652057200670242,
+      "num_tokens": 25272671.0,
+      "step": 236
+    },
+    {
+      "epoch": 0.540478905359179,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.999818376368435e-06,
+      "loss": 0.625,
+      "mean_token_accuracy": 0.8729881942272186,
+      "num_tokens": 25379431.0,
+      "step": 237
+    },
+    {
+      "epoch": 0.5427594070695553,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.999794964267208e-06,
+      "loss": 0.6275,
+      "mean_token_accuracy": 0.8688799887895584,
+      "num_tokens": 25486822.0,
+      "step": 238
+    },
+    {
+      "epoch": 0.5450399087799316,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.9997701333306215e-06,
+      "loss": 0.6471,
+      "mean_token_accuracy": 0.8656931668519974,
+      "num_tokens": 25594363.0,
+      "step": 239
+    },
+    {
+      "epoch": 0.5473204104903079,
+      "grad_norm": 5.34375,
+      "learning_rate": 4.999743883572766e-06,
+      "loss": 0.6471,
+      "mean_token_accuracy": 0.8659048229455948,
+      "num_tokens": 25701619.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5496009122006842,
+      "grad_norm": 4.125,
+      "learning_rate": 4.999716215008542e-06,
+      "loss": 0.6573,
+      "mean_token_accuracy": 0.866451695561409,
+      "num_tokens": 25808261.0,
+      "step": 241
+    },
+    {
+      "epoch": 0.5518814139110604,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.999687127653654e-06,
+      "loss": 0.6614,
+      "mean_token_accuracy": 0.8627986013889313,
+      "num_tokens": 25915402.0,
+      "step": 242
+    },
+    {
+      "epoch": 0.5541619156214367,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.99965662152461e-06,
+      "loss": 0.6355,
+      "mean_token_accuracy": 0.8666711002588272,
+      "num_tokens": 26023048.0,
+      "step": 243
+    },
+    {
+      "epoch": 0.556442417331813,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.999624696638725e-06,
+      "loss": 0.6311,
+      "mean_token_accuracy": 0.869884267449379,
+      "num_tokens": 26130137.0,
+      "step": 244
+    },
+    {
+      "epoch": 0.5587229190421893,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.999591353014119e-06,
+      "loss": 0.6488,
+      "mean_token_accuracy": 0.8630170971155167,
+      "num_tokens": 26237107.0,
+      "step": 245
+    },
+    {
+      "epoch": 0.5610034207525656,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.999556590669718e-06,
+      "loss": 0.6274,
+      "mean_token_accuracy": 0.8699584752321243,
+      "num_tokens": 26344577.0,
+      "step": 246
+    },
+    {
+      "epoch": 0.5632839224629419,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.999520409625253e-06,
+      "loss": 0.6507,
+      "mean_token_accuracy": 0.8654352575540543,
+      "num_tokens": 26451583.0,
+      "step": 247
+    },
+    {
+      "epoch": 0.5655644241733181,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.999482809901257e-06,
+      "loss": 0.6765,
+      "mean_token_accuracy": 0.8594978898763657,
+      "num_tokens": 26558350.0,
+      "step": 248
+    },
+    {
+      "epoch": 0.5678449258836944,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.999443791519074e-06,
+      "loss": 0.6401,
+      "mean_token_accuracy": 0.8702575117349625,
+      "num_tokens": 26665344.0,
+      "step": 249
+    },
+    {
+      "epoch": 0.5701254275940707,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.999403354500847e-06,
+      "loss": 0.6382,
+      "mean_token_accuracy": 0.8670401573181152,
+      "num_tokens": 26772368.0,
+      "step": 250
+    },
+    {
+      "epoch": 0.572405929304447,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.99936149886953e-06,
+      "loss": 0.6368,
+      "mean_token_accuracy": 0.8696161508560181,
+      "num_tokens": 26879565.0,
+      "step": 251
+    },
+    {
+      "epoch": 0.5746864310148233,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.999318224648878e-06,
+      "loss": 0.6335,
+      "mean_token_accuracy": 0.8677998781204224,
+      "num_tokens": 26986810.0,
+      "step": 252
+    },
+    {
+      "epoch": 0.5769669327251995,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.999273531863453e-06,
+      "loss": 0.6261,
+      "mean_token_accuracy": 0.8713762909173965,
+      "num_tokens": 27094179.0,
+      "step": 253
+    },
+    {
+      "epoch": 0.5792474344355758,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.999227420538622e-06,
+      "loss": 0.6654,
+      "mean_token_accuracy": 0.8648567199707031,
+      "num_tokens": 27201502.0,
+      "step": 254
+    },
+    {
+      "epoch": 0.5815279361459521,
+      "grad_norm": 2.25,
+      "learning_rate": 4.999179890700555e-06,
+      "loss": 0.663,
+      "mean_token_accuracy": 0.8646685779094696,
+      "num_tokens": 27308101.0,
+      "step": 255
+    },
+    {
+      "epoch": 0.5838084378563284,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.999130942376232e-06,
+      "loss": 0.6307,
+      "mean_token_accuracy": 0.8694456964731216,
+      "num_tokens": 27414923.0,
+      "step": 256
+    },
+    {
+      "epoch": 0.5860889395667047,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.999080575593433e-06,
+      "loss": 0.6417,
+      "mean_token_accuracy": 0.8683536648750305,
+      "num_tokens": 27521506.0,
+      "step": 257
+    },
+    {
+      "epoch": 0.5883694412770809,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.999028790380746e-06,
+      "loss": 0.6632,
+      "mean_token_accuracy": 0.8628116995096207,
+      "num_tokens": 27628384.0,
+      "step": 258
+    },
+    {
+      "epoch": 0.5906499429874572,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.9989755867675635e-06,
+      "loss": 0.638,
+      "mean_token_accuracy": 0.8686887472867966,
+      "num_tokens": 27735457.0,
+      "step": 259
+    },
+    {
+      "epoch": 0.5929304446978335,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.998920964784082e-06,
+      "loss": 0.6394,
+      "mean_token_accuracy": 0.868531346321106,
+      "num_tokens": 27842440.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.5952109464082098,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.998864924461305e-06,
+      "loss": 0.6169,
+      "mean_token_accuracy": 0.8760952204465866,
+      "num_tokens": 27949150.0,
+      "step": 261
+    },
+    {
+      "epoch": 0.5974914481185861,
+      "grad_norm": 2.75,
+      "learning_rate": 4.998807465831039e-06,
+      "loss": 0.6455,
+      "mean_token_accuracy": 0.8662375062704086,
+      "num_tokens": 28056218.0,
+      "step": 262
+    },
+    {
+      "epoch": 0.5997719498289624,
+      "grad_norm": 3.0,
+      "learning_rate": 4.998748588925897e-06,
+      "loss": 0.6565,
+      "mean_token_accuracy": 0.8636495620012283,
+      "num_tokens": 28163194.0,
+      "step": 263
+    },
+    {
+      "epoch": 0.6020524515393386,
+      "grad_norm": 2.25,
+      "learning_rate": 4.998688293779297e-06,
+      "loss": 0.621,
+      "mean_token_accuracy": 0.8721827417612076,
+      "num_tokens": 28270611.0,
+      "step": 264
+    },
+    {
+      "epoch": 0.6043329532497149,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.998626580425459e-06,
+      "loss": 0.6558,
+      "mean_token_accuracy": 0.8628446161746979,
+      "num_tokens": 28377681.0,
+      "step": 265
+    },
+    {
+      "epoch": 0.6066134549600912,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.998563448899413e-06,
+      "loss": 0.6314,
+      "mean_token_accuracy": 0.8687665909528732,
+      "num_tokens": 28484960.0,
+      "step": 266
+    },
+    {
+      "epoch": 0.6088939566704675,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.998498899236989e-06,
+      "loss": 0.6571,
+      "mean_token_accuracy": 0.8647327572107315,
+      "num_tokens": 28592263.0,
+      "step": 267
+    },
+    {
+      "epoch": 0.6111744583808438,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.998432931474825e-06,
+      "loss": 0.6395,
+      "mean_token_accuracy": 0.8670562505722046,
+      "num_tokens": 28699389.0,
+      "step": 268
+    },
+    {
+      "epoch": 0.61345496009122,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.998365545650365e-06,
+      "loss": 0.6289,
+      "mean_token_accuracy": 0.8697399348020554,
+      "num_tokens": 28807311.0,
+      "step": 269
+    },
+    {
+      "epoch": 0.6157354618015963,
+      "grad_norm": 6.0,
+      "learning_rate": 4.998296741801852e-06,
+      "loss": 0.657,
+      "mean_token_accuracy": 0.8628436028957367,
+      "num_tokens": 28913892.0,
+      "step": 270
+    },
+    {
+      "epoch": 0.6180159635119726,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.998226519968341e-06,
+      "loss": 0.6215,
+      "mean_token_accuracy": 0.8716912418603897,
+      "num_tokens": 29020775.0,
+      "step": 271
+    },
+    {
+      "epoch": 0.6202964652223489,
+      "grad_norm": 4.0,
+      "learning_rate": 4.998154880189688e-06,
+      "loss": 0.6409,
+      "mean_token_accuracy": 0.8669329136610031,
+      "num_tokens": 29127842.0,
+      "step": 272
+    },
+    {
+      "epoch": 0.6225769669327252,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.998081822506552e-06,
+      "loss": 0.643,
+      "mean_token_accuracy": 0.8676059246063232,
+      "num_tokens": 29235156.0,
+      "step": 273
+    },
+    {
+      "epoch": 0.6248574686431014,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.998007346960402e-06,
+      "loss": 0.6382,
+      "mean_token_accuracy": 0.8684564977884293,
+      "num_tokens": 29342694.0,
+      "step": 274
+    },
+    {
+      "epoch": 0.6271379703534777,
+      "grad_norm": 2.375,
+      "learning_rate": 4.997931453593507e-06,
+      "loss": 0.6342,
+      "mean_token_accuracy": 0.8661504536867142,
+      "num_tokens": 29449934.0,
+      "step": 275
+    },
+    {
+      "epoch": 0.629418472063854,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.997854142448944e-06,
+      "loss": 0.632,
+      "mean_token_accuracy": 0.8716614693403244,
+      "num_tokens": 29557007.0,
+      "step": 276
+    },
+    {
+      "epoch": 0.6316989737742303,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.997775413570593e-06,
+      "loss": 0.6217,
+      "mean_token_accuracy": 0.8713774085044861,
+      "num_tokens": 29664526.0,
+      "step": 277
+    },
+    {
+      "epoch": 0.6339794754846066,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.997695267003139e-06,
+      "loss": 0.6231,
+      "mean_token_accuracy": 0.8719380050897598,
+      "num_tokens": 29771396.0,
+      "step": 278
+    },
+    {
+      "epoch": 0.636259977194983,
+      "grad_norm": 2.5,
+      "learning_rate": 4.99761370279207e-06,
+      "loss": 0.6335,
+      "mean_token_accuracy": 0.8700685054063797,
+      "num_tokens": 29877736.0,
+      "step": 279
+    },
+    {
+      "epoch": 0.6385404789053591,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.997530720983682e-06,
+      "loss": 0.6331,
+      "mean_token_accuracy": 0.8674991726875305,
+      "num_tokens": 29985140.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6408209806157354,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.9974463216250735e-06,
+      "loss": 0.6618,
+      "mean_token_accuracy": 0.8660477548837662,
+      "num_tokens": 30092099.0,
+      "step": 281
+    },
+    {
+      "epoch": 0.6431014823261118,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.997360504764148e-06,
+      "loss": 0.6333,
+      "mean_token_accuracy": 0.8705120533704758,
+      "num_tokens": 30198927.0,
+      "step": 282
+    },
+    {
+      "epoch": 0.645381984036488,
+      "grad_norm": 2.25,
+      "learning_rate": 4.997273270449614e-06,
+      "loss": 0.6165,
+      "mean_token_accuracy": 0.8727722465991974,
+      "num_tokens": 30306090.0,
+      "step": 283
+    },
+    {
+      "epoch": 0.6476624857468644,
+      "grad_norm": 2.875,
+      "learning_rate": 4.997184618730983e-06,
+      "loss": 0.648,
+      "mean_token_accuracy": 0.8658933788537979,
+      "num_tokens": 30413636.0,
+      "step": 284
+    },
+    {
+      "epoch": 0.6499429874572406,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.997094549658572e-06,
+      "loss": 0.6261,
+      "mean_token_accuracy": 0.8701870143413544,
+      "num_tokens": 30521647.0,
+      "step": 285
+    },
+    {
+      "epoch": 0.6522234891676169,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.997003063283503e-06,
+      "loss": 0.6482,
+      "mean_token_accuracy": 0.8652370274066925,
+      "num_tokens": 30628370.0,
+      "step": 286
+    },
+    {
+      "epoch": 0.6545039908779932,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.996910159657703e-06,
+      "loss": 0.6343,
+      "mean_token_accuracy": 0.8690200746059418,
+      "num_tokens": 30735820.0,
+      "step": 287
+    },
+    {
+      "epoch": 0.6567844925883695,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.996815838833899e-06,
+      "loss": 0.6356,
+      "mean_token_accuracy": 0.8680445849895477,
+      "num_tokens": 30843566.0,
+      "step": 288
+    },
+    {
+      "epoch": 0.6590649942987458,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.99672010086563e-06,
+      "loss": 0.6028,
+      "mean_token_accuracy": 0.8737114071846008,
+      "num_tokens": 30951346.0,
+      "step": 289
+    },
+    {
+      "epoch": 0.661345496009122,
+      "grad_norm": 3.125,
+      "learning_rate": 4.996622945807231e-06,
+      "loss": 0.6135,
+      "mean_token_accuracy": 0.8741404414176941,
+      "num_tokens": 31058663.0,
+      "step": 290
+    },
+    {
+      "epoch": 0.6636259977194983,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.996524373713848e-06,
+      "loss": 0.6218,
+      "mean_token_accuracy": 0.8712268769741058,
+      "num_tokens": 31166704.0,
+      "step": 291
+    },
+    {
+      "epoch": 0.6659064994298746,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.996424384641428e-06,
+      "loss": 0.6119,
+      "mean_token_accuracy": 0.874211773276329,
+      "num_tokens": 31273610.0,
+      "step": 292
+    },
+    {
+      "epoch": 0.6681870011402509,
+      "grad_norm": 6.875,
+      "learning_rate": 4.996322978646722e-06,
+      "loss": 0.6088,
+      "mean_token_accuracy": 0.8734241724014282,
+      "num_tokens": 31380662.0,
+      "step": 293
+    },
+    {
+      "epoch": 0.6704675028506272,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.996220155787287e-06,
+      "loss": 0.6526,
+      "mean_token_accuracy": 0.8677922487258911,
+      "num_tokens": 31487775.0,
+      "step": 294
+    },
+    {
+      "epoch": 0.6727480045610034,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.996115916121483e-06,
+      "loss": 0.6319,
+      "mean_token_accuracy": 0.8704153895378113,
+      "num_tokens": 31595052.0,
+      "step": 295
+    },
+    {
+      "epoch": 0.6750285062713797,
+      "grad_norm": 5.28125,
+      "learning_rate": 4.996010259708475e-06,
+      "loss": 0.6412,
+      "mean_token_accuracy": 0.8663373440504074,
+      "num_tokens": 31702051.0,
+      "step": 296
+    },
+    {
+      "epoch": 0.677309007981756,
+      "grad_norm": 2.75,
+      "learning_rate": 4.99590318660823e-06,
+      "loss": 0.6066,
+      "mean_token_accuracy": 0.87331423163414,
+      "num_tokens": 31809437.0,
+      "step": 297
+    },
+    {
+      "epoch": 0.6795895096921323,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.9957946968815215e-06,
+      "loss": 0.6374,
+      "mean_token_accuracy": 0.8677855134010315,
+      "num_tokens": 31916405.0,
+      "step": 298
+    },
+    {
+      "epoch": 0.6818700114025086,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.995684790589927e-06,
+      "loss": 0.6389,
+      "mean_token_accuracy": 0.8677714616060257,
+      "num_tokens": 32023457.0,
+      "step": 299
+    },
+    {
+      "epoch": 0.6841505131128849,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.995573467795825e-06,
+      "loss": 0.6213,
+      "mean_token_accuracy": 0.8719353079795837,
+      "num_tokens": 32130881.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.6864310148232611,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.995460728562403e-06,
+      "loss": 0.6392,
+      "mean_token_accuracy": 0.8694345206022263,
+      "num_tokens": 32237937.0,
+      "step": 301
+    },
+    {
+      "epoch": 0.6887115165336374,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.9953465729536475e-06,
+      "loss": 0.6415,
+      "mean_token_accuracy": 0.8696666061878204,
+      "num_tokens": 32345103.0,
+      "step": 302
+    },
+    {
+      "epoch": 0.6909920182440137,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.995231001034352e-06,
+      "loss": 0.6148,
+      "mean_token_accuracy": 0.868248924612999,
+      "num_tokens": 32451683.0,
+      "step": 303
+    },
+    {
+      "epoch": 0.69327251995439,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.995114012870112e-06,
+      "loss": 0.6406,
+      "mean_token_accuracy": 0.8688762336969376,
+      "num_tokens": 32558767.0,
+      "step": 304
+    },
+    {
+      "epoch": 0.6955530216647663,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.99499560852733e-06,
+      "loss": 0.629,
+      "mean_token_accuracy": 0.8722851425409317,
+      "num_tokens": 32665998.0,
+      "step": 305
+    },
+    {
+      "epoch": 0.6978335233751425,
+      "grad_norm": 5.0,
+      "learning_rate": 4.994875788073207e-06,
+      "loss": 0.6373,
+      "mean_token_accuracy": 0.8682472556829453,
+      "num_tokens": 32773798.0,
+      "step": 306
+    },
+    {
+      "epoch": 0.7001140250855188,
+      "grad_norm": 3.5,
+      "learning_rate": 4.994754551575752e-06,
+      "loss": 0.65,
+      "mean_token_accuracy": 0.8666471391916275,
+      "num_tokens": 32880403.0,
+      "step": 307
+    },
+    {
+      "epoch": 0.7023945267958951,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.994631899103777e-06,
+      "loss": 0.6419,
+      "mean_token_accuracy": 0.867155522108078,
+      "num_tokens": 32987552.0,
+      "step": 308
+    },
+    {
+      "epoch": 0.7046750285062714,
+      "grad_norm": 2.375,
+      "learning_rate": 4.9945078307268974e-06,
+      "loss": 0.6452,
+      "mean_token_accuracy": 0.8653433471918106,
+      "num_tokens": 33094230.0,
+      "step": 309
+    },
+    {
+      "epoch": 0.7069555302166477,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.994382346515531e-06,
+      "loss": 0.6171,
+      "mean_token_accuracy": 0.8708781599998474,
+      "num_tokens": 33201782.0,
+      "step": 310
+    },
+    {
+      "epoch": 0.7092360319270239,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.9942554465409e-06,
+      "loss": 0.6404,
+      "mean_token_accuracy": 0.8667383790016174,
+      "num_tokens": 33308197.0,
+      "step": 311
+    },
+    {
+      "epoch": 0.7115165336374002,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.994127130875032e-06,
+      "loss": 0.6098,
+      "mean_token_accuracy": 0.8729591369628906,
+      "num_tokens": 33415348.0,
+      "step": 312
+    },
+    {
+      "epoch": 0.7137970353477765,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.993997399590755e-06,
+      "loss": 0.6182,
+      "mean_token_accuracy": 0.8707993477582932,
+      "num_tokens": 33522331.0,
+      "step": 313
+    },
+    {
+      "epoch": 0.7160775370581528,
+      "grad_norm": 4.25,
+      "learning_rate": 4.993866252761702e-06,
+      "loss": 0.6517,
+      "mean_token_accuracy": 0.8653054535388947,
+      "num_tokens": 33629329.0,
+      "step": 314
+    },
+    {
+      "epoch": 0.7183580387685291,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.993733690462311e-06,
+      "loss": 0.6286,
+      "mean_token_accuracy": 0.8681600391864777,
+      "num_tokens": 33736412.0,
+      "step": 315
+    },
+    {
+      "epoch": 0.7206385404789054,
+      "grad_norm": 1.953125,
+      "learning_rate": 4.99359971276782e-06,
+      "loss": 0.6184,
+      "mean_token_accuracy": 0.8719700872898102,
+      "num_tokens": 33843463.0,
+      "step": 316
+    },
+    {
+      "epoch": 0.7229190421892816,
+      "grad_norm": 1.8203125,
+      "learning_rate": 4.993464319754273e-06,
+      "loss": 0.6154,
+      "mean_token_accuracy": 0.8726183176040649,
+      "num_tokens": 33950234.0,
+      "step": 317
+    },
+    {
+      "epoch": 0.7251995438996579,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.993327511498516e-06,
+      "loss": 0.625,
+      "mean_token_accuracy": 0.8715388774871826,
+      "num_tokens": 34057495.0,
+      "step": 318
+    },
+    {
+      "epoch": 0.7274800456100342,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.9931892880782e-06,
+      "loss": 0.6312,
+      "mean_token_accuracy": 0.8697730153799057,
+      "num_tokens": 34164527.0,
+      "step": 319
+    },
+    {
+      "epoch": 0.7297605473204105,
+      "grad_norm": 2.75,
+      "learning_rate": 4.993049649571775e-06,
+      "loss": 0.6444,
+      "mean_token_accuracy": 0.8663023114204407,
+      "num_tokens": 34271517.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7320410490307868,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.992908596058501e-06,
+      "loss": 0.6272,
+      "mean_token_accuracy": 0.8702614009380341,
+      "num_tokens": 34378878.0,
+      "step": 321
+    },
+    {
+      "epoch": 0.734321550741163,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.992766127618434e-06,
+      "loss": 0.6261,
+      "mean_token_accuracy": 0.8703981339931488,
+      "num_tokens": 34485783.0,
+      "step": 322
+    },
+    {
+      "epoch": 0.7366020524515393,
+      "grad_norm": 2.5,
+      "learning_rate": 4.992622244332439e-06,
+      "loss": 0.6452,
+      "mean_token_accuracy": 0.8647294193506241,
+      "num_tokens": 34592416.0,
+      "step": 323
+    },
+    {
+      "epoch": 0.7388825541619156,
+      "grad_norm": 1.953125,
+      "learning_rate": 4.992476946282179e-06,
+      "loss": 0.6313,
+      "mean_token_accuracy": 0.8711864650249481,
+      "num_tokens": 34699693.0,
+      "step": 324
+    },
+    {
+      "epoch": 0.7411630558722919,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.992330233550124e-06,
+      "loss": 0.6155,
+      "mean_token_accuracy": 0.8703418523073196,
+      "num_tokens": 34806837.0,
+      "step": 325
+    },
+    {
+      "epoch": 0.7434435575826682,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.9921821062195445e-06,
+      "loss": 0.6482,
+      "mean_token_accuracy": 0.8661866039037704,
+      "num_tokens": 34913354.0,
+      "step": 326
+    },
+    {
+      "epoch": 0.7457240592930444,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.9920325643745145e-06,
+      "loss": 0.6217,
+      "mean_token_accuracy": 0.8691374510526657,
+      "num_tokens": 35020432.0,
+      "step": 327
+    },
+    {
+      "epoch": 0.7480045610034207,
+      "grad_norm": 6.3125,
+      "learning_rate": 4.991881608099912e-06,
+      "loss": 0.6274,
+      "mean_token_accuracy": 0.8629052639007568,
+      "num_tokens": 35127877.0,
+      "step": 328
+    },
+    {
+      "epoch": 0.750285062713797,
+      "grad_norm": 4.5,
+      "learning_rate": 4.991729237481417e-06,
+      "loss": 0.6361,
+      "mean_token_accuracy": 0.8679073601961136,
+      "num_tokens": 35234847.0,
+      "step": 329
+    },
+    {
+      "epoch": 0.7525655644241733,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.991575452605511e-06,
+      "loss": 0.6226,
+      "mean_token_accuracy": 0.8713287711143494,
+      "num_tokens": 35342056.0,
+      "step": 330
+    },
+    {
+      "epoch": 0.7548460661345496,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.9914202535594795e-06,
+      "loss": 0.6452,
+      "mean_token_accuracy": 0.8670705258846283,
+      "num_tokens": 35448907.0,
+      "step": 331
+    },
+    {
+      "epoch": 0.7571265678449259,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.991263640431411e-06,
+      "loss": 0.6152,
+      "mean_token_accuracy": 0.8704476803541183,
+      "num_tokens": 35556170.0,
+      "step": 332
+    },
+    {
+      "epoch": 0.7594070695553021,
+      "grad_norm": 5.5,
+      "learning_rate": 4.9911056133101965e-06,
+      "loss": 0.6195,
+      "mean_token_accuracy": 0.8712238371372223,
+      "num_tokens": 35662971.0,
+      "step": 333
+    },
+    {
+      "epoch": 0.7616875712656784,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.990946172285528e-06,
+      "loss": 0.6431,
+      "mean_token_accuracy": 0.8710373938083649,
+      "num_tokens": 35769814.0,
+      "step": 334
+    },
+    {
+      "epoch": 0.7639680729760547,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.990785317447901e-06,
+      "loss": 0.6003,
+      "mean_token_accuracy": 0.8746102750301361,
+      "num_tokens": 35877019.0,
+      "step": 335
+    },
+    {
+      "epoch": 0.766248574686431,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.990623048888615e-06,
+      "loss": 0.627,
+      "mean_token_accuracy": 0.8735091537237167,
+      "num_tokens": 35983784.0,
+      "step": 336
+    },
+    {
+      "epoch": 0.7685290763968073,
+      "grad_norm": 3.9375,
+      "learning_rate": 4.9904593666997704e-06,
+      "loss": 0.6629,
+      "mean_token_accuracy": 0.8641745299100876,
+      "num_tokens": 36090599.0,
+      "step": 337
+    },
+    {
+      "epoch": 0.7708095781071835,
+      "grad_norm": 3.875,
+      "learning_rate": 4.990294270974268e-06,
+      "loss": 0.6328,
+      "mean_token_accuracy": 0.8708094358444214,
+      "num_tokens": 36197751.0,
+      "step": 338
+    },
+    {
+      "epoch": 0.7730900798175598,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.990127761805816e-06,
+      "loss": 0.6345,
+      "mean_token_accuracy": 0.8706918656826019,
+      "num_tokens": 36304049.0,
+      "step": 339
+    },
+    {
+      "epoch": 0.7753705815279361,
+      "grad_norm": 1.9375,
+      "learning_rate": 4.989959839288919e-06,
+      "loss": 0.6264,
+      "mean_token_accuracy": 0.8699767887592316,
+      "num_tokens": 36411843.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.7776510832383124,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.989790503518888e-06,
+      "loss": 0.6192,
+      "mean_token_accuracy": 0.8728293031454086,
+      "num_tokens": 36518890.0,
+      "step": 341
+    },
+    {
+      "epoch": 0.7799315849486887,
+      "grad_norm": 5.875,
+      "learning_rate": 4.9896197545918345e-06,
+      "loss": 0.6203,
+      "mean_token_accuracy": 0.8697379231452942,
+      "num_tokens": 36626348.0,
+      "step": 342
+    },
+    {
+      "epoch": 0.7822120866590649,
+      "grad_norm": 1.8984375,
+      "learning_rate": 4.989447592604673e-06,
+      "loss": 0.6028,
+      "mean_token_accuracy": 0.8770763874053955,
+      "num_tokens": 36733926.0,
+      "step": 343
+    },
+    {
+      "epoch": 0.7844925883694412,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.989274017655117e-06,
+      "loss": 0.6,
+      "mean_token_accuracy": 0.8766891658306122,
+      "num_tokens": 36841051.0,
+      "step": 344
+    },
+    {
+      "epoch": 0.7867730900798175,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.989099029841687e-06,
+      "loss": 0.6305,
+      "mean_token_accuracy": 0.8703635185956955,
+      "num_tokens": 36947674.0,
+      "step": 345
+    },
+    {
+      "epoch": 0.7890535917901939,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.988922629263701e-06,
+      "loss": 0.6234,
+      "mean_token_accuracy": 0.87026646733284,
+      "num_tokens": 37054412.0,
+      "step": 346
+    },
+    {
+      "epoch": 0.7913340935005702,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.988744816021283e-06,
+      "loss": 0.64,
+      "mean_token_accuracy": 0.8698112666606903,
+      "num_tokens": 37162022.0,
+      "step": 347
+    },
+    {
+      "epoch": 0.7936145952109465,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.988565590215352e-06,
+      "loss": 0.6163,
+      "mean_token_accuracy": 0.8709748089313507,
+      "num_tokens": 37269040.0,
+      "step": 348
+    },
+    {
+      "epoch": 0.7958950969213227,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.9883849519476364e-06,
+      "loss": 0.6075,
+      "mean_token_accuracy": 0.8744599372148514,
+      "num_tokens": 37376445.0,
+      "step": 349
+    },
+    {
+      "epoch": 0.798175598631699,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.988202901320663e-06,
+      "loss": 0.6218,
+      "mean_token_accuracy": 0.8723867684602737,
+      "num_tokens": 37483781.0,
+      "step": 350
+    },
+    {
+      "epoch": 0.8004561003420753,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.988019438437759e-06,
+      "loss": 0.6543,
+      "mean_token_accuracy": 0.8646156787872314,
+      "num_tokens": 37590388.0,
+      "step": 351
+    },
+    {
+      "epoch": 0.8027366020524516,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.987834563403055e-06,
+      "loss": 0.6162,
+      "mean_token_accuracy": 0.8731140941381454,
+      "num_tokens": 37698197.0,
+      "step": 352
+    },
+    {
+      "epoch": 0.8050171037628279,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.987648276321482e-06,
+      "loss": 0.658,
+      "mean_token_accuracy": 0.8623346835374832,
+      "num_tokens": 37804721.0,
+      "step": 353
+    },
+    {
+      "epoch": 0.8072976054732041,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.987460577298774e-06,
+      "loss": 0.6171,
+      "mean_token_accuracy": 0.8700351715087891,
+      "num_tokens": 37912089.0,
+      "step": 354
+    },
+    {
+      "epoch": 0.8095781071835804,
+      "grad_norm": 5.53125,
+      "learning_rate": 4.9872714664414635e-06,
+      "loss": 0.6288,
+      "mean_token_accuracy": 0.8672404289245605,
+      "num_tokens": 38018353.0,
+      "step": 355
+    },
+    {
+      "epoch": 0.8118586088939567,
+      "grad_norm": 5.0,
+      "learning_rate": 4.987080943856887e-06,
+      "loss": 0.6396,
+      "mean_token_accuracy": 0.8685039430856705,
+      "num_tokens": 38124920.0,
+      "step": 356
+    },
+    {
+      "epoch": 0.814139110604333,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.986889009653183e-06,
+      "loss": 0.6103,
+      "mean_token_accuracy": 0.8747462034225464,
+      "num_tokens": 38232424.0,
+      "step": 357
+    },
+    {
+      "epoch": 0.8164196123147093,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.986695663939288e-06,
+      "loss": 0.6204,
+      "mean_token_accuracy": 0.8709569126367569,
+      "num_tokens": 38339566.0,
+      "step": 358
+    },
+    {
+      "epoch": 0.8187001140250855,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.986500906824942e-06,
+      "loss": 0.6181,
+      "mean_token_accuracy": 0.8750324100255966,
+      "num_tokens": 38446417.0,
+      "step": 359
+    },
+    {
+      "epoch": 0.8209806157354618,
+      "grad_norm": 5.5,
+      "learning_rate": 4.986304738420684e-06,
+      "loss": 0.6115,
+      "mean_token_accuracy": 0.8724595308303833,
+      "num_tokens": 38553490.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8232611174458381,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.9861071588378565e-06,
+      "loss": 0.644,
+      "mean_token_accuracy": 0.8690180629491806,
+      "num_tokens": 38660341.0,
+      "step": 361
+    },
+    {
+      "epoch": 0.8255416191562144,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.985908168188602e-06,
+      "loss": 0.6163,
+      "mean_token_accuracy": 0.8723935782909393,
+      "num_tokens": 38767171.0,
+      "step": 362
+    },
+    {
+      "epoch": 0.8278221208665907,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.985707766585865e-06,
+      "loss": 0.6186,
+      "mean_token_accuracy": 0.8713973164558411,
+      "num_tokens": 38873913.0,
+      "step": 363
+    },
+    {
+      "epoch": 0.830102622576967,
+      "grad_norm": 2.75,
+      "learning_rate": 4.985505954143387e-06,
+      "loss": 0.6212,
+      "mean_token_accuracy": 0.8741555064916611,
+      "num_tokens": 38981023.0,
+      "step": 364
+    },
+    {
+      "epoch": 0.8323831242873432,
+      "grad_norm": 1.8359375,
+      "learning_rate": 4.985302730975713e-06,
+      "loss": 0.6252,
+      "mean_token_accuracy": 0.8745421171188354,
+      "num_tokens": 39088181.0,
+      "step": 365
+    },
+    {
+      "epoch": 0.8346636259977195,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.9850980971981914e-06,
+      "loss": 0.6419,
+      "mean_token_accuracy": 0.870025172829628,
+      "num_tokens": 39195055.0,
+      "step": 366
+    },
+    {
+      "epoch": 0.8369441277080958,
+      "grad_norm": 3.375,
+      "learning_rate": 4.984892052926965e-06,
+      "loss": 0.6414,
+      "mean_token_accuracy": 0.8648830950260162,
+      "num_tokens": 39302023.0,
+      "step": 367
+    },
+    {
+      "epoch": 0.8392246294184721,
+      "grad_norm": 1.9609375,
+      "learning_rate": 4.984684598278982e-06,
+      "loss": 0.6332,
+      "mean_token_accuracy": 0.8697508275508881,
+      "num_tokens": 39408616.0,
+      "step": 368
+    },
+    {
+      "epoch": 0.8415051311288484,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.984475733371991e-06,
+      "loss": 0.649,
+      "mean_token_accuracy": 0.8658408671617508,
+      "num_tokens": 39515948.0,
+      "step": 369
+    },
+    {
+      "epoch": 0.8437856328392246,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.984265458324538e-06,
+      "loss": 0.6415,
+      "mean_token_accuracy": 0.8648363202810287,
+      "num_tokens": 39622628.0,
+      "step": 370
+    },
+    {
+      "epoch": 0.8460661345496009,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.984053773255971e-06,
+      "loss": 0.6158,
+      "mean_token_accuracy": 0.8734622299671173,
+      "num_tokens": 39730107.0,
+      "step": 371
+    },
+    {
+      "epoch": 0.8483466362599772,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.9838406782864394e-06,
+      "loss": 0.6134,
+      "mean_token_accuracy": 0.8723733127117157,
+      "num_tokens": 39837288.0,
+      "step": 372
+    },
+    {
+      "epoch": 0.8506271379703535,
+      "grad_norm": 1.9296875,
+      "learning_rate": 4.983626173536891e-06,
+      "loss": 0.6114,
+      "mean_token_accuracy": 0.8771011680364609,
+      "num_tokens": 39944365.0,
+      "step": 373
+    },
+    {
+      "epoch": 0.8529076396807298,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.983410259129075e-06,
+      "loss": 0.6515,
+      "mean_token_accuracy": 0.8640440553426743,
+      "num_tokens": 40051377.0,
+      "step": 374
+    },
+    {
+      "epoch": 0.855188141391106,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.983192935185539e-06,
+      "loss": 0.6192,
+      "mean_token_accuracy": 0.8728392422199249,
+      "num_tokens": 40158071.0,
+      "step": 375
+    },
+    {
+      "epoch": 0.8574686431014823,
+      "grad_norm": 2.03125,
+      "learning_rate": 4.9829742018296335e-06,
+      "loss": 0.6264,
+      "mean_token_accuracy": 0.8718508183956146,
+      "num_tokens": 40265163.0,
+      "step": 376
+    },
+    {
+      "epoch": 0.8597491448118586,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.9827540591855064e-06,
+      "loss": 0.6263,
+      "mean_token_accuracy": 0.8694735020399094,
+      "num_tokens": 40372053.0,
+      "step": 377
+    },
+    {
+      "epoch": 0.8620296465222349,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.9825325073781075e-06,
+      "loss": 0.6302,
+      "mean_token_accuracy": 0.8715538680553436,
+      "num_tokens": 40479008.0,
+      "step": 378
+    },
+    {
+      "epoch": 0.8643101482326112,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.982309546533184e-06,
+      "loss": 0.6379,
+      "mean_token_accuracy": 0.866651251912117,
+      "num_tokens": 40585719.0,
+      "step": 379
+    },
+    {
+      "epoch": 0.8665906499429875,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.982085176777285e-06,
+      "loss": 0.6138,
+      "mean_token_accuracy": 0.8743686825037003,
+      "num_tokens": 40692870.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.8688711516533637,
+      "grad_norm": 4.75,
+      "learning_rate": 4.981859398237758e-06,
+      "loss": 0.6326,
+      "mean_token_accuracy": 0.8692844212055206,
+      "num_tokens": 40800059.0,
+      "step": 381
+    },
+    {
+      "epoch": 0.87115165336374,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.9816322110427505e-06,
+      "loss": 0.6292,
+      "mean_token_accuracy": 0.8697078227996826,
+      "num_tokens": 40907542.0,
+      "step": 382
+    },
+    {
+      "epoch": 0.8734321550741163,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.98140361532121e-06,
+      "loss": 0.6362,
+      "mean_token_accuracy": 0.871175691485405,
+      "num_tokens": 41014351.0,
+      "step": 383
+    },
+    {
+      "epoch": 0.8757126567844926,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.981173611202883e-06,
+      "loss": 0.6454,
+      "mean_token_accuracy": 0.8670124560594559,
+      "num_tokens": 41121154.0,
+      "step": 384
+    },
+    {
+      "epoch": 0.8779931584948689,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.980942198818315e-06,
+      "loss": 0.63,
+      "mean_token_accuracy": 0.8680609911680222,
+      "num_tokens": 41228488.0,
+      "step": 385
+    },
+    {
+      "epoch": 0.8802736602052451,
+      "grad_norm": 4.96875,
+      "learning_rate": 4.980709378298851e-06,
+      "loss": 0.6381,
+      "mean_token_accuracy": 0.8670986741781235,
+      "num_tokens": 41335487.0,
+      "step": 386
+    },
+    {
+      "epoch": 0.8825541619156214,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.980475149776636e-06,
+      "loss": 0.6267,
+      "mean_token_accuracy": 0.8722383230924606,
+      "num_tokens": 41442292.0,
+      "step": 387
+    },
+    {
+      "epoch": 0.8848346636259977,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.980239513384614e-06,
+      "loss": 0.6291,
+      "mean_token_accuracy": 0.8699522018432617,
+      "num_tokens": 41549340.0,
+      "step": 388
+    },
+    {
+      "epoch": 0.887115165336374,
+      "grad_norm": 7.5,
+      "learning_rate": 4.980002469256527e-06,
+      "loss": 0.6157,
+      "mean_token_accuracy": 0.8692153990268707,
+      "num_tokens": 41656335.0,
+      "step": 389
+    },
+    {
+      "epoch": 0.8893956670467503,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.979764017526916e-06,
+      "loss": 0.6327,
+      "mean_token_accuracy": 0.8701380044221878,
+      "num_tokens": 41763377.0,
+      "step": 390
+    },
+    {
+      "epoch": 0.8916761687571265,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.979524158331123e-06,
+      "loss": 0.635,
+      "mean_token_accuracy": 0.8717087209224701,
+      "num_tokens": 41870195.0,
+      "step": 391
+    },
+    {
+      "epoch": 0.8939566704675028,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.979282891805287e-06,
+      "loss": 0.6182,
+      "mean_token_accuracy": 0.8729406297206879,
+      "num_tokens": 41977586.0,
+      "step": 392
+    },
+    {
+      "epoch": 0.8962371721778791,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.979040218086345e-06,
+      "loss": 0.6338,
+      "mean_token_accuracy": 0.8731931746006012,
+      "num_tokens": 42084265.0,
+      "step": 393
+    },
+    {
+      "epoch": 0.8985176738882554,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.978796137312036e-06,
+      "loss": 0.6323,
+      "mean_token_accuracy": 0.8713531494140625,
+      "num_tokens": 42191215.0,
+      "step": 394
+    },
+    {
+      "epoch": 0.9007981755986317,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.978550649620894e-06,
+      "loss": 0.6286,
+      "mean_token_accuracy": 0.868993267416954,
+      "num_tokens": 42298587.0,
+      "step": 395
+    },
+    {
+      "epoch": 0.9030786773090079,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.978303755152254e-06,
+      "loss": 0.6027,
+      "mean_token_accuracy": 0.8749285191297531,
+      "num_tokens": 42406175.0,
+      "step": 396
+    },
+    {
+      "epoch": 0.9053591790193842,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.978055454046247e-06,
+      "loss": 0.6109,
+      "mean_token_accuracy": 0.8721934705972672,
+      "num_tokens": 42513138.0,
+      "step": 397
+    },
+    {
+      "epoch": 0.9076396807297605,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.977805746443807e-06,
+      "loss": 0.6093,
+      "mean_token_accuracy": 0.87620410323143,
+      "num_tokens": 42620098.0,
+      "step": 398
+    },
+    {
+      "epoch": 0.9099201824401368,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.9775546324866596e-06,
+      "loss": 0.5789,
+      "mean_token_accuracy": 0.8835956156253815,
+      "num_tokens": 42727074.0,
+      "step": 399
+    },
+    {
+      "epoch": 0.9122006841505131,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.977302112317334e-06,
+      "loss": 0.6503,
+      "mean_token_accuracy": 0.8638210296630859,
+      "num_tokens": 42834015.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9144811858608894,
+      "grad_norm": 5.78125,
+      "learning_rate": 4.977048186079155e-06,
+      "loss": 0.6314,
+      "mean_token_accuracy": 0.8687157332897186,
+      "num_tokens": 42940807.0,
+      "step": 401
+    },
+    {
+      "epoch": 0.9167616875712656,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.976792853916248e-06,
+      "loss": 0.6368,
+      "mean_token_accuracy": 0.8687217086553574,
+      "num_tokens": 43047379.0,
+      "step": 402
+    },
+    {
+      "epoch": 0.9190421892816419,
+      "grad_norm": 4.625,
+      "learning_rate": 4.9765361159735335e-06,
+      "loss": 0.6258,
+      "mean_token_accuracy": 0.8750667423009872,
+      "num_tokens": 43154795.0,
+      "step": 403
+    },
+    {
+      "epoch": 0.9213226909920182,
+      "grad_norm": 6.25,
+      "learning_rate": 4.97627797239673e-06,
+      "loss": 0.6312,
+      "mean_token_accuracy": 0.8691485226154327,
+      "num_tokens": 43261640.0,
+      "step": 404
+    },
+    {
+      "epoch": 0.9236031927023945,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.976018423332357e-06,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.875212773680687,
+      "num_tokens": 43368543.0,
+      "step": 405
+    },
+    {
+      "epoch": 0.9258836944127709,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.975757468927727e-06,
+      "loss": 0.6041,
+      "mean_token_accuracy": 0.8764694184064865,
+      "num_tokens": 43475828.0,
+      "step": 406
+    },
+    {
+      "epoch": 0.928164196123147,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.975495109330954e-06,
+      "loss": 0.6317,
+      "mean_token_accuracy": 0.8676637560129166,
+      "num_tokens": 43582569.0,
+      "step": 407
+    },
+    {
+      "epoch": 0.9304446978335233,
+      "grad_norm": 8.4375,
+      "learning_rate": 4.97523134469095e-06,
+      "loss": 0.6399,
+      "mean_token_accuracy": 0.8702896982431412,
+      "num_tokens": 43689655.0,
+      "step": 408
+    },
+    {
+      "epoch": 0.9327251995438997,
+      "grad_norm": 2.875,
+      "learning_rate": 4.97496617515742e-06,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.8774754852056503,
+      "num_tokens": 43797009.0,
+      "step": 409
+    },
+    {
+      "epoch": 0.935005701254276,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.974699600880869e-06,
+      "loss": 0.6182,
+      "mean_token_accuracy": 0.8728543817996979,
+      "num_tokens": 43904821.0,
+      "step": 410
+    },
+    {
+      "epoch": 0.9372862029646523,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.974431622012601e-06,
+      "loss": 0.6565,
+      "mean_token_accuracy": 0.866395503282547,
+      "num_tokens": 44011564.0,
+      "step": 411
+    },
+    {
+      "epoch": 0.9395667046750285,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.974162238704716e-06,
+      "loss": 0.594,
+      "mean_token_accuracy": 0.8790825754404068,
+      "num_tokens": 44118685.0,
+      "step": 412
+    },
+    {
+      "epoch": 0.9418472063854048,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.973891451110109e-06,
+      "loss": 0.6196,
+      "mean_token_accuracy": 0.8710049241781235,
+      "num_tokens": 44225866.0,
+      "step": 413
+    },
+    {
+      "epoch": 0.9441277080957811,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.973619259382475e-06,
+      "loss": 0.6301,
+      "mean_token_accuracy": 0.8691826015710831,
+      "num_tokens": 44333305.0,
+      "step": 414
+    },
+    {
+      "epoch": 0.9464082098061574,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.973345663676305e-06,
+      "loss": 0.6057,
+      "mean_token_accuracy": 0.8764727264642715,
+      "num_tokens": 44439926.0,
+      "step": 415
+    },
+    {
+      "epoch": 0.9486887115165337,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.973070664146885e-06,
+      "loss": 0.6067,
+      "mean_token_accuracy": 0.871231347322464,
+      "num_tokens": 44547096.0,
+      "step": 416
+    },
+    {
+      "epoch": 0.95096921322691,
+      "grad_norm": 4.25,
+      "learning_rate": 4.972794260950301e-06,
+      "loss": 0.6118,
+      "mean_token_accuracy": 0.8720909953117371,
+      "num_tokens": 44654153.0,
+      "step": 417
+    },
+    {
+      "epoch": 0.9532497149372862,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.972516454243433e-06,
+      "loss": 0.6346,
+      "mean_token_accuracy": 0.8684164136648178,
+      "num_tokens": 44761805.0,
+      "step": 418
+    },
+    {
+      "epoch": 0.9555302166476625,
+      "grad_norm": 6.25,
+      "learning_rate": 4.972237244183961e-06,
+      "loss": 0.6235,
+      "mean_token_accuracy": 0.8743065893650055,
+      "num_tokens": 44869180.0,
+      "step": 419
+    },
+    {
+      "epoch": 0.9578107183580388,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.971956630930356e-06,
+      "loss": 0.6178,
+      "mean_token_accuracy": 0.8736863434314728,
+      "num_tokens": 44976595.0,
+      "step": 420
+    },
+    {
+      "epoch": 0.9600912200684151,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.971674614641891e-06,
+      "loss": 0.618,
+      "mean_token_accuracy": 0.8725884109735489,
+      "num_tokens": 45083791.0,
+      "step": 421
+    },
+    {
+      "epoch": 0.9623717217787914,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.971391195478632e-06,
+      "loss": 0.6255,
+      "mean_token_accuracy": 0.875871405005455,
+      "num_tokens": 45190655.0,
+      "step": 422
+    },
+    {
+      "epoch": 0.9646522234891676,
+      "grad_norm": 4.0,
+      "learning_rate": 4.971106373601443e-06,
+      "loss": 0.6342,
+      "mean_token_accuracy": 0.8689168095588684,
+      "num_tokens": 45297243.0,
+      "step": 423
+    },
+    {
+      "epoch": 0.9669327251995439,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.9708201491719825e-06,
+      "loss": 0.623,
+      "mean_token_accuracy": 0.8707065731287003,
+      "num_tokens": 45404013.0,
+      "step": 424
+    },
+    {
+      "epoch": 0.9692132269099202,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.9705325223527055e-06,
+      "loss": 0.6136,
+      "mean_token_accuracy": 0.8755741715431213,
+      "num_tokens": 45511141.0,
+      "step": 425
+    },
+    {
+      "epoch": 0.9714937286202965,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.970243493306865e-06,
+      "loss": 0.6068,
+      "mean_token_accuracy": 0.8742416203022003,
+      "num_tokens": 45618184.0,
+      "step": 426
+    },
+    {
+      "epoch": 0.9737742303306728,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.969953062198508e-06,
+      "loss": 0.6242,
+      "mean_token_accuracy": 0.8716663718223572,
+      "num_tokens": 45725266.0,
+      "step": 427
+    },
+    {
+      "epoch": 0.976054732041049,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.969661229192477e-06,
+      "loss": 0.6127,
+      "mean_token_accuracy": 0.8720242828130722,
+      "num_tokens": 45832262.0,
+      "step": 428
+    },
+    {
+      "epoch": 0.9783352337514253,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.969367994454412e-06,
+      "loss": 0.5931,
+      "mean_token_accuracy": 0.8772017359733582,
+      "num_tokens": 45939904.0,
+      "step": 429
+    },
+    {
+      "epoch": 0.9806157354618016,
+      "grad_norm": 1.9609375,
+      "learning_rate": 4.9690733581507445e-06,
+      "loss": 0.6109,
+      "mean_token_accuracy": 0.8746745586395264,
+      "num_tokens": 46046844.0,
+      "step": 430
+    },
+    {
+      "epoch": 0.9828962371721779,
+      "grad_norm": 1.84375,
+      "learning_rate": 4.968777320448707e-06,
+      "loss": 0.622,
+      "mean_token_accuracy": 0.8727656751871109,
+      "num_tokens": 46154121.0,
+      "step": 431
+    },
+    {
+      "epoch": 0.9851767388825542,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.9684798815163235e-06,
+      "loss": 0.6052,
+      "mean_token_accuracy": 0.8737929463386536,
+      "num_tokens": 46260970.0,
+      "step": 432
+    },
+    {
+      "epoch": 0.9874572405929305,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.968181041522416e-06,
+      "loss": 0.6273,
+      "mean_token_accuracy": 0.8705646842718124,
+      "num_tokens": 46368205.0,
+      "step": 433
+    },
+    {
+      "epoch": 0.9897377423033067,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.967880800636599e-06,
+      "loss": 0.6285,
+      "mean_token_accuracy": 0.8682654201984406,
+      "num_tokens": 46475197.0,
+      "step": 434
+    },
+    {
+      "epoch": 0.992018244013683,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.967579159029284e-06,
+      "loss": 0.6114,
+      "mean_token_accuracy": 0.8732243329286575,
+      "num_tokens": 46582041.0,
+      "step": 435
+    },
+    {
+      "epoch": 0.9942987457240593,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.9672761168716766e-06,
+      "loss": 0.6324,
+      "mean_token_accuracy": 0.8705407828092575,
+      "num_tokens": 46688998.0,
+      "step": 436
+    },
+    {
+      "epoch": 0.9965792474344356,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.966971674335778e-06,
+      "loss": 0.6035,
+      "mean_token_accuracy": 0.8775283247232437,
+      "num_tokens": 46795830.0,
+      "step": 437
+    },
+    {
+      "epoch": 0.9988597491448119,
+      "grad_norm": 2.125,
+      "learning_rate": 4.966665831594383e-06,
+      "loss": 0.6028,
+      "mean_token_accuracy": 0.8748981207609177,
+      "num_tokens": 46903068.0,
+      "step": 438
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 7.34375,
+      "learning_rate": 4.966358588821084e-06,
+      "loss": 0.6496,
+      "mean_token_accuracy": 0.8653521537780762,
+      "num_tokens": 46942232.0,
+      "step": 439
+    },
+    {
+      "epoch": 1.0022805017103762,
+      "grad_norm": 2.0,
+      "learning_rate": 4.966049946190265e-06,
+      "loss": 0.624,
+      "mean_token_accuracy": 0.8737205117940903,
+      "num_tokens": 47049296.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0022805017103762,
+      "eval_loss": 0.6221891045570374,
+      "eval_mean_token_accuracy": 0.8722115945453426,
+      "eval_num_tokens": 47049296.0,
+      "eval_runtime": 58.6597,
+      "eval_samples_per_second": 142.943,
+      "eval_steps_per_second": 4.483,
+      "step": 440
+    },
+    {
+      "epoch": 1.0045610034207526,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.9657399038771045e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8777357935905457,
+      "num_tokens": 47156210.0,
+      "step": 441
+    },
+    {
+      "epoch": 1.0068415051311288,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.965428462057578e-06,
+      "loss": 0.6234,
+      "mean_token_accuracy": 0.8698296397924423,
+      "num_tokens": 47263285.0,
+      "step": 442
+    },
+    {
+      "epoch": 1.0091220068415052,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.965115620908453e-06,
+      "loss": 0.5926,
+      "mean_token_accuracy": 0.8760122805833817,
+      "num_tokens": 47370395.0,
+      "step": 443
+    },
+    {
+      "epoch": 1.0114025085518814,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.964801380607293e-06,
+      "loss": 0.6211,
+      "mean_token_accuracy": 0.8716463297605515,
+      "num_tokens": 47477476.0,
+      "step": 444
+    },
+    {
+      "epoch": 1.0136830102622576,
+      "grad_norm": 3.25,
+      "learning_rate": 4.964485741332453e-06,
+      "loss": 0.6086,
+      "mean_token_accuracy": 0.8730912059545517,
+      "num_tokens": 47585481.0,
+      "step": 445
+    },
+    {
+      "epoch": 1.015963511972634,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.964168703263086e-06,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8801993578672409,
+      "num_tokens": 47693090.0,
+      "step": 446
+    },
+    {
+      "epoch": 1.0182440136830102,
+      "grad_norm": 5.8125,
+      "learning_rate": 4.963850266579136e-06,
+      "loss": 0.6235,
+      "mean_token_accuracy": 0.8687308132648468,
+      "num_tokens": 47800542.0,
+      "step": 447
+    },
+    {
+      "epoch": 1.0205245153933866,
+      "grad_norm": 5.0,
+      "learning_rate": 4.963530431461341e-06,
+      "loss": 0.6048,
+      "mean_token_accuracy": 0.8741424828767776,
+      "num_tokens": 47907379.0,
+      "step": 448
+    },
+    {
+      "epoch": 1.0228050171037628,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.963209198091232e-06,
+      "loss": 0.6295,
+      "mean_token_accuracy": 0.8697365671396255,
+      "num_tokens": 48014994.0,
+      "step": 449
+    },
+    {
+      "epoch": 1.025085518814139,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.962886566651138e-06,
+      "loss": 0.6177,
+      "mean_token_accuracy": 0.873694971203804,
+      "num_tokens": 48122545.0,
+      "step": 450
+    },
+    {
+      "epoch": 1.0273660205245154,
+      "grad_norm": 8.5625,
+      "learning_rate": 4.962562537324176e-06,
+      "loss": 0.6066,
+      "mean_token_accuracy": 0.8789347857236862,
+      "num_tokens": 48230071.0,
+      "step": 451
+    },
+    {
+      "epoch": 1.0296465222348916,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.96223711029426e-06,
+      "loss": 0.641,
+      "mean_token_accuracy": 0.8675385862588882,
+      "num_tokens": 48336830.0,
+      "step": 452
+    },
+    {
+      "epoch": 1.031927023945268,
+      "grad_norm": 5.90625,
+      "learning_rate": 4.961910285746094e-06,
+      "loss": 0.5954,
+      "mean_token_accuracy": 0.8765672594308853,
+      "num_tokens": 48443698.0,
+      "step": 453
+    },
+    {
+      "epoch": 1.0342075256556442,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.9615820638651805e-06,
+      "loss": 0.642,
+      "mean_token_accuracy": 0.8699254840612411,
+      "num_tokens": 48550340.0,
+      "step": 454
+    },
+    {
+      "epoch": 1.0364880273660204,
+      "grad_norm": 5.71875,
+      "learning_rate": 4.961252444837809e-06,
+      "loss": 0.6359,
+      "mean_token_accuracy": 0.8699039667844772,
+      "num_tokens": 48657031.0,
+      "step": 455
+    },
+    {
+      "epoch": 1.0387685290763968,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.960921428851066e-06,
+      "loss": 0.6169,
+      "mean_token_accuracy": 0.8748660534620285,
+      "num_tokens": 48763597.0,
+      "step": 456
+    },
+    {
+      "epoch": 1.041049030786773,
+      "grad_norm": 5.0,
+      "learning_rate": 4.960589016092832e-06,
+      "loss": 0.6144,
+      "mean_token_accuracy": 0.8746305704116821,
+      "num_tokens": 48871423.0,
+      "step": 457
+    },
+    {
+      "epoch": 1.0433295324971494,
+      "grad_norm": 6.84375,
+      "learning_rate": 4.960255206751774e-06,
+      "loss": 0.627,
+      "mean_token_accuracy": 0.8704584836959839,
+      "num_tokens": 48978350.0,
+      "step": 458
+    },
+    {
+      "epoch": 1.0456100342075256,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.959920001017358e-06,
+      "loss": 0.6198,
+      "mean_token_accuracy": 0.8744053989648819,
+      "num_tokens": 49085534.0,
+      "step": 459
+    },
+    {
+      "epoch": 1.047890535917902,
+      "grad_norm": 4.59375,
+      "learning_rate": 4.95958339907984e-06,
+      "loss": 0.6185,
+      "mean_token_accuracy": 0.8731739073991776,
+      "num_tokens": 49192856.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.0501710376282782,
+      "grad_norm": 9.1875,
+      "learning_rate": 4.959245401130269e-06,
+      "loss": 0.6314,
+      "mean_token_accuracy": 0.8677381575107574,
+      "num_tokens": 49299373.0,
+      "step": 461
+    },
+    {
+      "epoch": 1.0524515393386544,
+      "grad_norm": 6.5625,
+      "learning_rate": 4.958906007360487e-06,
+      "loss": 0.5992,
+      "mean_token_accuracy": 0.876073032617569,
+      "num_tokens": 49406428.0,
+      "step": 462
+    },
+    {
+      "epoch": 1.0547320410490308,
+      "grad_norm": 3.96875,
+      "learning_rate": 4.958565217963125e-06,
+      "loss": 0.6201,
+      "mean_token_accuracy": 0.8712072819471359,
+      "num_tokens": 49513674.0,
+      "step": 463
+    },
+    {
+      "epoch": 1.057012542759407,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.95822303313161e-06,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.8719818741083145,
+      "num_tokens": 49620690.0,
+      "step": 464
+    },
+    {
+      "epoch": 1.0592930444697835,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.957879453060159e-06,
+      "loss": 0.607,
+      "mean_token_accuracy": 0.874313622713089,
+      "num_tokens": 49728319.0,
+      "step": 465
+    },
+    {
+      "epoch": 1.0615735461801596,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.957534477943782e-06,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8772328495979309,
+      "num_tokens": 49835825.0,
+      "step": 466
+    },
+    {
+      "epoch": 1.0638540478905358,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.957188107978279e-06,
+      "loss": 0.6051,
+      "mean_token_accuracy": 0.8736355155706406,
+      "num_tokens": 49943303.0,
+      "step": 467
+    },
+    {
+      "epoch": 1.0661345496009123,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.956840343360245e-06,
+      "loss": 0.6138,
+      "mean_token_accuracy": 0.8750191628932953,
+      "num_tokens": 50050828.0,
+      "step": 468
+    },
+    {
+      "epoch": 1.0684150513112884,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.956491184287062e-06,
+      "loss": 0.6051,
+      "mean_token_accuracy": 0.8774718195199966,
+      "num_tokens": 50158185.0,
+      "step": 469
+    },
+    {
+      "epoch": 1.0706955530216649,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.9561406309569084e-06,
+      "loss": 0.6381,
+      "mean_token_accuracy": 0.8694217354059219,
+      "num_tokens": 50265385.0,
+      "step": 470
+    },
+    {
+      "epoch": 1.072976054732041,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.955788683568749e-06,
+      "loss": 0.6035,
+      "mean_token_accuracy": 0.8717086911201477,
+      "num_tokens": 50372238.0,
+      "step": 471
+    },
+    {
+      "epoch": 1.0752565564424172,
+      "grad_norm": 3.625,
+      "learning_rate": 4.955435342322345e-06,
+      "loss": 0.5958,
+      "mean_token_accuracy": 0.8751126825809479,
+      "num_tokens": 50479115.0,
+      "step": 472
+    },
+    {
+      "epoch": 1.0775370581527937,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.955080607418244e-06,
+      "loss": 0.6061,
+      "mean_token_accuracy": 0.8753761649131775,
+      "num_tokens": 50586257.0,
+      "step": 473
+    },
+    {
+      "epoch": 1.0798175598631699,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.954724479057788e-06,
+      "loss": 0.6195,
+      "mean_token_accuracy": 0.8744408041238785,
+      "num_tokens": 50692536.0,
+      "step": 474
+    },
+    {
+      "epoch": 1.0820980615735463,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.954366957443107e-06,
+      "loss": 0.6102,
+      "mean_token_accuracy": 0.8744495362043381,
+      "num_tokens": 50799101.0,
+      "step": 475
+    },
+    {
+      "epoch": 1.0843785632839225,
+      "grad_norm": 4.96875,
+      "learning_rate": 4.954008042777125e-06,
+      "loss": 0.6226,
+      "mean_token_accuracy": 0.871866300702095,
+      "num_tokens": 50906128.0,
+      "step": 476
+    },
+    {
+      "epoch": 1.0866590649942987,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.953647735263555e-06,
+      "loss": 0.5962,
+      "mean_token_accuracy": 0.8779451102018356,
+      "num_tokens": 51013166.0,
+      "step": 477
+    },
+    {
+      "epoch": 1.088939566704675,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.953286035106898e-06,
+      "loss": 0.6202,
+      "mean_token_accuracy": 0.8745466768741608,
+      "num_tokens": 51120460.0,
+      "step": 478
+    },
+    {
+      "epoch": 1.0912200684150513,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.952922942512452e-06,
+      "loss": 0.6224,
+      "mean_token_accuracy": 0.8701803684234619,
+      "num_tokens": 51227376.0,
+      "step": 479
+    },
+    {
+      "epoch": 1.0935005701254277,
+      "grad_norm": 1.8515625,
+      "learning_rate": 4.9525584576862985e-06,
+      "loss": 0.623,
+      "mean_token_accuracy": 0.8713761270046234,
+      "num_tokens": 51334284.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.0957810718358039,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.952192580835313e-06,
+      "loss": 0.6286,
+      "mean_token_accuracy": 0.8694994300603867,
+      "num_tokens": 51441531.0,
+      "step": 481
+    },
+    {
+      "epoch": 1.09806157354618,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.9518253121671595e-06,
+      "loss": 0.6201,
+      "mean_token_accuracy": 0.8716080784797668,
+      "num_tokens": 51549055.0,
+      "step": 482
+    },
+    {
+      "epoch": 1.1003420752565565,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.951456651890294e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.881346806883812,
+      "num_tokens": 51656798.0,
+      "step": 483
+    },
+    {
+      "epoch": 1.1026225769669327,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.951086600213959e-06,
+      "loss": 0.6183,
+      "mean_token_accuracy": 0.8716641664505005,
+      "num_tokens": 51763790.0,
+      "step": 484
+    },
+    {
+      "epoch": 1.104903078677309,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.950715157348191e-06,
+      "loss": 0.6196,
+      "mean_token_accuracy": 0.8753807693719864,
+      "num_tokens": 51870653.0,
+      "step": 485
+    },
+    {
+      "epoch": 1.1071835803876853,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.950342323503812e-06,
+      "loss": 0.6146,
+      "mean_token_accuracy": 0.870696559548378,
+      "num_tokens": 51977455.0,
+      "step": 486
+    },
+    {
+      "epoch": 1.1094640820980617,
+      "grad_norm": 4.625,
+      "learning_rate": 4.949968098892436e-06,
+      "loss": 0.642,
+      "mean_token_accuracy": 0.8685554414987564,
+      "num_tokens": 52083822.0,
+      "step": 487
+    },
+    {
+      "epoch": 1.1117445838084379,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.949592483726465e-06,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8732410371303558,
+      "num_tokens": 52191177.0,
+      "step": 488
+    },
+    {
+      "epoch": 1.114025085518814,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.949215478219092e-06,
+      "loss": 0.5947,
+      "mean_token_accuracy": 0.8758023232221603,
+      "num_tokens": 52297640.0,
+      "step": 489
+    },
+    {
+      "epoch": 1.1163055872291905,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.948837082584298e-06,
+      "loss": 0.6011,
+      "mean_token_accuracy": 0.8783663511276245,
+      "num_tokens": 52404875.0,
+      "step": 490
+    },
+    {
+      "epoch": 1.1185860889395667,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.9484572970368516e-06,
+      "loss": 0.6263,
+      "mean_token_accuracy": 0.8689621537923813,
+      "num_tokens": 52511463.0,
+      "step": 491
+    },
+    {
+      "epoch": 1.120866590649943,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.948076121792313e-06,
+      "loss": 0.6115,
+      "mean_token_accuracy": 0.8712314665317535,
+      "num_tokens": 52617733.0,
+      "step": 492
+    },
+    {
+      "epoch": 1.1231470923603193,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.9476935570670294e-06,
+      "loss": 0.6101,
+      "mean_token_accuracy": 0.8728629648685455,
+      "num_tokens": 52725104.0,
+      "step": 493
+    },
+    {
+      "epoch": 1.1254275940706955,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.947309603078138e-06,
+      "loss": 0.646,
+      "mean_token_accuracy": 0.8683103322982788,
+      "num_tokens": 52832460.0,
+      "step": 494
+    },
+    {
+      "epoch": 1.127708095781072,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.946924260043563e-06,
+      "loss": 0.6341,
+      "mean_token_accuracy": 0.8687001317739487,
+      "num_tokens": 52939314.0,
+      "step": 495
+    },
+    {
+      "epoch": 1.129988597491448,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.946537528182017e-06,
+      "loss": 0.6301,
+      "mean_token_accuracy": 0.8715783804655075,
+      "num_tokens": 53045997.0,
+      "step": 496
+    },
+    {
+      "epoch": 1.1322690992018245,
+      "grad_norm": 2.5,
+      "learning_rate": 4.946149407713002e-06,
+      "loss": 0.6211,
+      "mean_token_accuracy": 0.8704388439655304,
+      "num_tokens": 53153160.0,
+      "step": 497
+    },
+    {
+      "epoch": 1.1345496009122007,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.945759898856809e-06,
+      "loss": 0.6162,
+      "mean_token_accuracy": 0.8732296079397202,
+      "num_tokens": 53260206.0,
+      "step": 498
+    },
+    {
+      "epoch": 1.1368301026225769,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.9453690018345144e-06,
+      "loss": 0.6392,
+      "mean_token_accuracy": 0.8706572353839874,
+      "num_tokens": 53366657.0,
+      "step": 499
+    },
+    {
+      "epoch": 1.1391106043329533,
+      "grad_norm": 3.25,
+      "learning_rate": 4.944976716867984e-06,
+      "loss": 0.635,
+      "mean_token_accuracy": 0.8699849843978882,
+      "num_tokens": 53473137.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.1413911060433295,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.944583044179871e-06,
+      "loss": 0.6214,
+      "mean_token_accuracy": 0.875198557972908,
+      "num_tokens": 53580438.0,
+      "step": 501
+    },
+    {
+      "epoch": 1.143671607753706,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.944187983993617e-06,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8774997144937515,
+      "num_tokens": 53687875.0,
+      "step": 502
+    },
+    {
+      "epoch": 1.145952109464082,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.94379153653345e-06,
+      "loss": 0.6153,
+      "mean_token_accuracy": 0.8719532340764999,
+      "num_tokens": 53794260.0,
+      "step": 503
+    },
+    {
+      "epoch": 1.1482326111744583,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.9433937020243854e-06,
+      "loss": 0.6351,
+      "mean_token_accuracy": 0.869365006685257,
+      "num_tokens": 53901213.0,
+      "step": 504
+    },
+    {
+      "epoch": 1.1505131128848347,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.942994480692228e-06,
+      "loss": 0.6199,
+      "mean_token_accuracy": 0.8735174834728241,
+      "num_tokens": 54007865.0,
+      "step": 505
+    },
+    {
+      "epoch": 1.152793614595211,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.942593872763566e-06,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8790851831436157,
+      "num_tokens": 54114696.0,
+      "step": 506
+    },
+    {
+      "epoch": 1.1550741163055873,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.9421918784657795e-06,
+      "loss": 0.6112,
+      "mean_token_accuracy": 0.8767211586236954,
+      "num_tokens": 54221690.0,
+      "step": 507
+    },
+    {
+      "epoch": 1.1573546180159635,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.94178849802703e-06,
+      "loss": 0.6044,
+      "mean_token_accuracy": 0.8764821738004684,
+      "num_tokens": 54328424.0,
+      "step": 508
+    },
+    {
+      "epoch": 1.1596351197263397,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.9413837316762705e-06,
+      "loss": 0.6264,
+      "mean_token_accuracy": 0.8703786879777908,
+      "num_tokens": 54435061.0,
+      "step": 509
+    },
+    {
+      "epoch": 1.1619156214367161,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.940977579643237e-06,
+      "loss": 0.6096,
+      "mean_token_accuracy": 0.8734538704156876,
+      "num_tokens": 54541917.0,
+      "step": 510
+    },
+    {
+      "epoch": 1.1641961231470923,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.940570042158454e-06,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8750556707382202,
+      "num_tokens": 54649030.0,
+      "step": 511
+    },
+    {
+      "epoch": 1.1664766248574687,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.940161119453232e-06,
+      "loss": 0.6084,
+      "mean_token_accuracy": 0.8727173060178757,
+      "num_tokens": 54756245.0,
+      "step": 512
+    },
+    {
+      "epoch": 1.168757126567845,
+      "grad_norm": 4.53125,
+      "learning_rate": 4.939750811759668e-06,
+      "loss": 0.6083,
+      "mean_token_accuracy": 0.8715848624706268,
+      "num_tokens": 54863315.0,
+      "step": 513
+    },
+    {
+      "epoch": 1.171037628278221,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.939339119310645e-06,
+      "loss": 0.621,
+      "mean_token_accuracy": 0.8712294846773148,
+      "num_tokens": 54970714.0,
+      "step": 514
+    },
+    {
+      "epoch": 1.1733181299885975,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.93892604233983e-06,
+      "loss": 0.6305,
+      "mean_token_accuracy": 0.8707058429718018,
+      "num_tokens": 55077726.0,
+      "step": 515
+    },
+    {
+      "epoch": 1.1755986316989737,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.93851158108168e-06,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8780573010444641,
+      "num_tokens": 55185327.0,
+      "step": 516
+    },
+    {
+      "epoch": 1.1778791334093501,
+      "grad_norm": 3.953125,
+      "learning_rate": 4.938095735771433e-06,
+      "loss": 0.6184,
+      "mean_token_accuracy": 0.8706268519163132,
+      "num_tokens": 55292419.0,
+      "step": 517
+    },
+    {
+      "epoch": 1.1801596351197263,
+      "grad_norm": 1.8984375,
+      "learning_rate": 4.937678506645116e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.873799741268158,
+      "num_tokens": 55399684.0,
+      "step": 518
+    },
+    {
+      "epoch": 1.1824401368301025,
+      "grad_norm": 5.84375,
+      "learning_rate": 4.937259893939539e-06,
+      "loss": 0.6202,
+      "mean_token_accuracy": 0.8741802275180817,
+      "num_tokens": 55507140.0,
+      "step": 519
+    },
+    {
+      "epoch": 1.184720638540479,
+      "grad_norm": 4.78125,
+      "learning_rate": 4.9368398978923e-06,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8741926103830338,
+      "num_tokens": 55614345.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.1870011402508551,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.93641851874178e-06,
+      "loss": 0.631,
+      "mean_token_accuracy": 0.8710474520921707,
+      "num_tokens": 55721149.0,
+      "step": 521
+    },
+    {
+      "epoch": 1.1892816419612315,
+      "grad_norm": 4.5,
+      "learning_rate": 4.935995756727146e-06,
+      "loss": 0.6192,
+      "mean_token_accuracy": 0.8732537627220154,
+      "num_tokens": 55828356.0,
+      "step": 522
+    },
+    {
+      "epoch": 1.1915621436716077,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.935571612088349e-06,
+      "loss": 0.6065,
+      "mean_token_accuracy": 0.875878319144249,
+      "num_tokens": 55935753.0,
+      "step": 523
+    },
+    {
+      "epoch": 1.193842645381984,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.935146085066125e-06,
+      "loss": 0.6175,
+      "mean_token_accuracy": 0.874293178319931,
+      "num_tokens": 56042443.0,
+      "step": 524
+    },
+    {
+      "epoch": 1.1961231470923603,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.934719175901996e-06,
+      "loss": 0.6016,
+      "mean_token_accuracy": 0.8778766244649887,
+      "num_tokens": 56149378.0,
+      "step": 525
+    },
+    {
+      "epoch": 1.1984036488027365,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.934290884838266e-06,
+      "loss": 0.6168,
+      "mean_token_accuracy": 0.8713551312685013,
+      "num_tokens": 56256623.0,
+      "step": 526
+    },
+    {
+      "epoch": 1.200684150513113,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.933861212118027e-06,
+      "loss": 0.6319,
+      "mean_token_accuracy": 0.8705524206161499,
+      "num_tokens": 56363653.0,
+      "step": 527
+    },
+    {
+      "epoch": 1.2029646522234891,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.933430157985151e-06,
+      "loss": 0.6279,
+      "mean_token_accuracy": 0.8724389970302582,
+      "num_tokens": 56470381.0,
+      "step": 528
+    },
+    {
+      "epoch": 1.2052451539338653,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.932997722684296e-06,
+      "loss": 0.6157,
+      "mean_token_accuracy": 0.8716868907213211,
+      "num_tokens": 56577262.0,
+      "step": 529
+    },
+    {
+      "epoch": 1.2075256556442417,
+      "grad_norm": 4.625,
+      "learning_rate": 4.932563906460905e-06,
+      "loss": 0.6121,
+      "mean_token_accuracy": 0.8729743212461472,
+      "num_tokens": 56683964.0,
+      "step": 530
+    },
+    {
+      "epoch": 1.209806157354618,
+      "grad_norm": 7.375,
+      "learning_rate": 4.932128709561202e-06,
+      "loss": 0.6031,
+      "mean_token_accuracy": 0.8723554909229279,
+      "num_tokens": 56791540.0,
+      "step": 531
+    },
+    {
+      "epoch": 1.2120866590649944,
+      "grad_norm": 6.375,
+      "learning_rate": 4.931692132232198e-06,
+      "loss": 0.601,
+      "mean_token_accuracy": 0.8752623051404953,
+      "num_tokens": 56898876.0,
+      "step": 532
+    },
+    {
+      "epoch": 1.2143671607753705,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.931254174721687e-06,
+      "loss": 0.6311,
+      "mean_token_accuracy": 0.867269441485405,
+      "num_tokens": 57005808.0,
+      "step": 533
+    },
+    {
+      "epoch": 1.216647662485747,
+      "grad_norm": 2.125,
+      "learning_rate": 4.930814837278242e-06,
+      "loss": 0.5913,
+      "mean_token_accuracy": 0.8760450780391693,
+      "num_tokens": 57113633.0,
+      "step": 534
+    },
+    {
+      "epoch": 1.2189281641961232,
+      "grad_norm": 5.71875,
+      "learning_rate": 4.930374120151225e-06,
+      "loss": 0.6388,
+      "mean_token_accuracy": 0.8680548220872879,
+      "num_tokens": 57220723.0,
+      "step": 535
+    },
+    {
+      "epoch": 1.2212086659064993,
+      "grad_norm": 5.71875,
+      "learning_rate": 4.929932023590776e-06,
+      "loss": 0.6436,
+      "mean_token_accuracy": 0.867982491850853,
+      "num_tokens": 57327185.0,
+      "step": 536
+    },
+    {
+      "epoch": 1.2234891676168758,
+      "grad_norm": 7.25,
+      "learning_rate": 4.929488547847823e-06,
+      "loss": 0.6068,
+      "mean_token_accuracy": 0.8724008947610855,
+      "num_tokens": 57433998.0,
+      "step": 537
+    },
+    {
+      "epoch": 1.225769669327252,
+      "grad_norm": 5.0,
+      "learning_rate": 4.9290436931740735e-06,
+      "loss": 0.6194,
+      "mean_token_accuracy": 0.8717511296272278,
+      "num_tokens": 57541338.0,
+      "step": 538
+    },
+    {
+      "epoch": 1.2280501710376284,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.928597459822018e-06,
+      "loss": 0.6031,
+      "mean_token_accuracy": 0.879209354519844,
+      "num_tokens": 57648210.0,
+      "step": 539
+    },
+    {
+      "epoch": 1.2303306727480046,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.928149848044931e-06,
+      "loss": 0.6069,
+      "mean_token_accuracy": 0.8770927637815475,
+      "num_tokens": 57755187.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.2326111744583808,
+      "grad_norm": 4.375,
+      "learning_rate": 4.9277008580968665e-06,
+      "loss": 0.5956,
+      "mean_token_accuracy": 0.8778523355722427,
+      "num_tokens": 57862318.0,
+      "step": 541
+    },
+    {
+      "epoch": 1.2348916761687572,
+      "grad_norm": 5.03125,
+      "learning_rate": 4.927250490232664e-06,
+      "loss": 0.6139,
+      "mean_token_accuracy": 0.8774797171354294,
+      "num_tokens": 57969230.0,
+      "step": 542
+    },
+    {
+      "epoch": 1.2371721778791334,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.926798744707943e-06,
+      "loss": 0.5933,
+      "mean_token_accuracy": 0.8759450018405914,
+      "num_tokens": 58076744.0,
+      "step": 543
+    },
+    {
+      "epoch": 1.2394526795895098,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.926345621779106e-06,
+      "loss": 0.6241,
+      "mean_token_accuracy": 0.8749475181102753,
+      "num_tokens": 58184125.0,
+      "step": 544
+    },
+    {
+      "epoch": 1.241733181299886,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.9258911217033355e-06,
+      "loss": 0.6161,
+      "mean_token_accuracy": 0.8739427775144577,
+      "num_tokens": 58290980.0,
+      "step": 545
+    },
+    {
+      "epoch": 1.2440136830102622,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.925435244738599e-06,
+      "loss": 0.6038,
+      "mean_token_accuracy": 0.8763206750154495,
+      "num_tokens": 58397982.0,
+      "step": 546
+    },
+    {
+      "epoch": 1.2462941847206386,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.924977991143642e-06,
+      "loss": 0.5985,
+      "mean_token_accuracy": 0.8783524632453918,
+      "num_tokens": 58505934.0,
+      "step": 547
+    },
+    {
+      "epoch": 1.2485746864310148,
+      "grad_norm": 2.5,
+      "learning_rate": 4.924519361177993e-06,
+      "loss": 0.5877,
+      "mean_token_accuracy": 0.8753952831029892,
+      "num_tokens": 58613394.0,
+      "step": 548
+    },
+    {
+      "epoch": 1.2508551881413912,
+      "grad_norm": 1.8984375,
+      "learning_rate": 4.9240593551019625e-06,
+      "loss": 0.6099,
+      "mean_token_accuracy": 0.8755134344100952,
+      "num_tokens": 58721077.0,
+      "step": 549
+    },
+    {
+      "epoch": 1.2531356898517674,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.92359797317664e-06,
+      "loss": 0.6153,
+      "mean_token_accuracy": 0.8757786601781845,
+      "num_tokens": 58828151.0,
+      "step": 550
+    },
+    {
+      "epoch": 1.2554161915621438,
+      "grad_norm": 2.375,
+      "learning_rate": 4.923135215663897e-06,
+      "loss": 0.6188,
+      "mean_token_accuracy": 0.8740972727537155,
+      "num_tokens": 58935083.0,
+      "step": 551
+    },
+    {
+      "epoch": 1.25769669327252,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.922671082826386e-06,
+      "loss": 0.6087,
+      "mean_token_accuracy": 0.8730472475290298,
+      "num_tokens": 59041838.0,
+      "step": 552
+    },
+    {
+      "epoch": 1.2599771949828962,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.92220557492754e-06,
+      "loss": 0.6085,
+      "mean_token_accuracy": 0.8769043833017349,
+      "num_tokens": 59149008.0,
+      "step": 553
+    },
+    {
+      "epoch": 1.2622576966932726,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.921738692231572e-06,
+      "loss": 0.6053,
+      "mean_token_accuracy": 0.8772878795862198,
+      "num_tokens": 59255903.0,
+      "step": 554
+    },
+    {
+      "epoch": 1.2645381984036488,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.9212704350034764e-06,
+      "loss": 0.6136,
+      "mean_token_accuracy": 0.8748277425765991,
+      "num_tokens": 59362809.0,
+      "step": 555
+    },
+    {
+      "epoch": 1.2668187001140252,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.920800803509026e-06,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.87485072016716,
+      "num_tokens": 59469454.0,
+      "step": 556
+    },
+    {
+      "epoch": 1.2690992018244014,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.920329798014775e-06,
+      "loss": 0.6206,
+      "mean_token_accuracy": 0.8719239979982376,
+      "num_tokens": 59576709.0,
+      "step": 557
+    },
+    {
+      "epoch": 1.2713797035347776,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.919857418788056e-06,
+      "loss": 0.6285,
+      "mean_token_accuracy": 0.8737199753522873,
+      "num_tokens": 59683387.0,
+      "step": 558
+    },
+    {
+      "epoch": 1.273660205245154,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.919383666096985e-06,
+      "loss": 0.6085,
+      "mean_token_accuracy": 0.8742072433233261,
+      "num_tokens": 59790354.0,
+      "step": 559
+    },
+    {
+      "epoch": 1.2759407069555302,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.918908540210452e-06,
+      "loss": 0.6398,
+      "mean_token_accuracy": 0.869797870516777,
+      "num_tokens": 59897756.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.2782212086659066,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.91843204139813e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.88178950548172,
+      "num_tokens": 60004961.0,
+      "step": 561
+    },
+    {
+      "epoch": 1.2805017103762828,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.917954169930472e-06,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8791621029376984,
+      "num_tokens": 60112141.0,
+      "step": 562
+    },
+    {
+      "epoch": 1.282782212086659,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.917474926078707e-06,
+      "loss": 0.6084,
+      "mean_token_accuracy": 0.8750548511743546,
+      "num_tokens": 60220073.0,
+      "step": 563
+    },
+    {
+      "epoch": 1.2850627137970354,
+      "grad_norm": 2.75,
+      "learning_rate": 4.916994310114845e-06,
+      "loss": 0.5995,
+      "mean_token_accuracy": 0.8793503940105438,
+      "num_tokens": 60327794.0,
+      "step": 564
+    },
+    {
+      "epoch": 1.2873432155074116,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.916512322311675e-06,
+      "loss": 0.6275,
+      "mean_token_accuracy": 0.86960369348526,
+      "num_tokens": 60434871.0,
+      "step": 565
+    },
+    {
+      "epoch": 1.289623717217788,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.916028962942763e-06,
+      "loss": 0.5944,
+      "mean_token_accuracy": 0.8765391111373901,
+      "num_tokens": 60542243.0,
+      "step": 566
+    },
+    {
+      "epoch": 1.2919042189281642,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.915544232282455e-06,
+      "loss": 0.6144,
+      "mean_token_accuracy": 0.8739331364631653,
+      "num_tokens": 60649576.0,
+      "step": 567
+    },
+    {
+      "epoch": 1.2941847206385404,
+      "grad_norm": 1.8828125,
+      "learning_rate": 4.915058130605874e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8797845989465714,
+      "num_tokens": 60756480.0,
+      "step": 568
+    },
+    {
+      "epoch": 1.2964652223489168,
+      "grad_norm": 1.9609375,
+      "learning_rate": 4.9145706581889235e-06,
+      "loss": 0.6264,
+      "mean_token_accuracy": 0.8693940490484238,
+      "num_tokens": 60863068.0,
+      "step": 569
+    },
+    {
+      "epoch": 1.298745724059293,
+      "grad_norm": 5.1875,
+      "learning_rate": 4.914081815308283e-06,
+      "loss": 0.5955,
+      "mean_token_accuracy": 0.8792444169521332,
+      "num_tokens": 60969966.0,
+      "step": 570
+    },
+    {
+      "epoch": 1.3010262257696694,
+      "grad_norm": 2.75,
+      "learning_rate": 4.913591602241409e-06,
+      "loss": 0.6145,
+      "mean_token_accuracy": 0.8721539825201035,
+      "num_tokens": 61076972.0,
+      "step": 571
+    },
+    {
+      "epoch": 1.3033067274800456,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.9131000192665365e-06,
+      "loss": 0.6187,
+      "mean_token_accuracy": 0.8705037534236908,
+      "num_tokens": 61183878.0,
+      "step": 572
+    },
+    {
+      "epoch": 1.3055872291904218,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.9126070666626815e-06,
+      "loss": 0.6065,
+      "mean_token_accuracy": 0.8774714469909668,
+      "num_tokens": 61290915.0,
+      "step": 573
+    },
+    {
+      "epoch": 1.3078677309007982,
+      "grad_norm": 6.53125,
+      "learning_rate": 4.912112744709632e-06,
+      "loss": 0.599,
+      "mean_token_accuracy": 0.876824364066124,
+      "num_tokens": 61397686.0,
+      "step": 574
+    },
+    {
+      "epoch": 1.3101482326111744,
+      "grad_norm": 4.25,
+      "learning_rate": 4.911617053687957e-06,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8789774626493454,
+      "num_tokens": 61504802.0,
+      "step": 575
+    },
+    {
+      "epoch": 1.3124287343215508,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.911119993878999e-06,
+      "loss": 0.615,
+      "mean_token_accuracy": 0.8741600215435028,
+      "num_tokens": 61611778.0,
+      "step": 576
+    },
+    {
+      "epoch": 1.314709236031927,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.910621565564882e-06,
+      "loss": 0.6029,
+      "mean_token_accuracy": 0.8803430199623108,
+      "num_tokens": 61719136.0,
+      "step": 577
+    },
+    {
+      "epoch": 1.3169897377423032,
+      "grad_norm": 4.375,
+      "learning_rate": 4.910121769028503e-06,
+      "loss": 0.5989,
+      "mean_token_accuracy": 0.8754008263349533,
+      "num_tokens": 61826166.0,
+      "step": 578
+    },
+    {
+      "epoch": 1.3192702394526796,
+      "grad_norm": 8.625,
+      "learning_rate": 4.909620604553537e-06,
+      "loss": 0.6315,
+      "mean_token_accuracy": 0.8695260733366013,
+      "num_tokens": 61933001.0,
+      "step": 579
+    },
+    {
+      "epoch": 1.3215507411630558,
+      "grad_norm": 7.34375,
+      "learning_rate": 4.909118072424436e-06,
+      "loss": 0.5984,
+      "mean_token_accuracy": 0.880437433719635,
+      "num_tokens": 62039905.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3238312428734322,
+      "grad_norm": 6.21875,
+      "learning_rate": 4.908614172926426e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.8764496594667435,
+      "num_tokens": 62147016.0,
+      "step": 581
+    },
+    {
+      "epoch": 1.3261117445838084,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.908108906345512e-06,
+      "loss": 0.6047,
+      "mean_token_accuracy": 0.8747891783714294,
+      "num_tokens": 62254504.0,
+      "step": 582
+    },
+    {
+      "epoch": 1.3283922462941846,
+      "grad_norm": 3.75,
+      "learning_rate": 4.907602272968473e-06,
+      "loss": 0.6253,
+      "mean_token_accuracy": 0.8759507387876511,
+      "num_tokens": 62361146.0,
+      "step": 583
+    },
+    {
+      "epoch": 1.330672748004561,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.907094273082865e-06,
+      "loss": 0.6254,
+      "mean_token_accuracy": 0.8714919984340668,
+      "num_tokens": 62467825.0,
+      "step": 584
+    },
+    {
+      "epoch": 1.3329532497149372,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.906584906977018e-06,
+      "loss": 0.606,
+      "mean_token_accuracy": 0.8751325905323029,
+      "num_tokens": 62575437.0,
+      "step": 585
+    },
+    {
+      "epoch": 1.3352337514253136,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.906074174940038e-06,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.8743661940097809,
+      "num_tokens": 62682439.0,
+      "step": 586
+    },
+    {
+      "epoch": 1.3375142531356898,
+      "grad_norm": 2.125,
+      "learning_rate": 4.905562077261808e-06,
+      "loss": 0.6095,
+      "mean_token_accuracy": 0.8771595358848572,
+      "num_tokens": 62789139.0,
+      "step": 587
+    },
+    {
+      "epoch": 1.339794754846066,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.905048614232984e-06,
+      "loss": 0.62,
+      "mean_token_accuracy": 0.8709544539451599,
+      "num_tokens": 62896172.0,
+      "step": 588
+    },
+    {
+      "epoch": 1.3420752565564424,
+      "grad_norm": 1.9765625,
+      "learning_rate": 4.904533786144998e-06,
+      "loss": 0.6211,
+      "mean_token_accuracy": 0.8748406618833542,
+      "num_tokens": 63003536.0,
+      "step": 589
+    },
+    {
+      "epoch": 1.3443557582668186,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.904017593290056e-06,
+      "loss": 0.6059,
+      "mean_token_accuracy": 0.8730844259262085,
+      "num_tokens": 63110204.0,
+      "step": 590
+    },
+    {
+      "epoch": 1.346636259977195,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.903500035961139e-06,
+      "loss": 0.5986,
+      "mean_token_accuracy": 0.8759681731462479,
+      "num_tokens": 63217782.0,
+      "step": 591
+    },
+    {
+      "epoch": 1.3489167616875712,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.902981114452005e-06,
+      "loss": 0.6124,
+      "mean_token_accuracy": 0.8746581524610519,
+      "num_tokens": 63324751.0,
+      "step": 592
+    },
+    {
+      "epoch": 1.3511972633979474,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.90246082905718e-06,
+      "loss": 0.617,
+      "mean_token_accuracy": 0.8720570355653763,
+      "num_tokens": 63431709.0,
+      "step": 593
+    },
+    {
+      "epoch": 1.3534777651083238,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.90193918007197e-06,
+      "loss": 0.6156,
+      "mean_token_accuracy": 0.8742910176515579,
+      "num_tokens": 63538728.0,
+      "step": 594
+    },
+    {
+      "epoch": 1.3557582668187,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.901416167792452e-06,
+      "loss": 0.6044,
+      "mean_token_accuracy": 0.875694528222084,
+      "num_tokens": 63645888.0,
+      "step": 595
+    },
+    {
+      "epoch": 1.3580387685290765,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.9008917925154795e-06,
+      "loss": 0.595,
+      "mean_token_accuracy": 0.8773985803127289,
+      "num_tokens": 63753292.0,
+      "step": 596
+    },
+    {
+      "epoch": 1.3603192702394526,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.900366054538675e-06,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8744669109582901,
+      "num_tokens": 63859832.0,
+      "step": 597
+    },
+    {
+      "epoch": 1.3625997719498288,
+      "grad_norm": 9.0625,
+      "learning_rate": 4.8998389541604405e-06,
+      "loss": 0.6178,
+      "mean_token_accuracy": 0.8743558824062347,
+      "num_tokens": 63966869.0,
+      "step": 598
+    },
+    {
+      "epoch": 1.3648802736602053,
+      "grad_norm": 7.9375,
+      "learning_rate": 4.899310491679945e-06,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.8777134567499161,
+      "num_tokens": 64074013.0,
+      "step": 599
+    },
+    {
+      "epoch": 1.3671607753705814,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.898780667397136e-06,
+      "loss": 0.6098,
+      "mean_token_accuracy": 0.8756420910358429,
+      "num_tokens": 64181228.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.3694412770809579,
+      "grad_norm": 4.9375,
+      "learning_rate": 4.89824948161273e-06,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8778956681489944,
+      "num_tokens": 64288189.0,
+      "step": 601
+    },
+    {
+      "epoch": 1.371721778791334,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.8977169346282184e-06,
+      "loss": 0.6101,
+      "mean_token_accuracy": 0.8754762411117554,
+      "num_tokens": 64395116.0,
+      "step": 602
+    },
+    {
+      "epoch": 1.3740022805017102,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.8971830267458645e-06,
+      "loss": 0.6076,
+      "mean_token_accuracy": 0.8785329908132553,
+      "num_tokens": 64502388.0,
+      "step": 603
+    },
+    {
+      "epoch": 1.3762827822120867,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.896647758268703e-06,
+      "loss": 0.6121,
+      "mean_token_accuracy": 0.8748378753662109,
+      "num_tokens": 64609541.0,
+      "step": 604
+    },
+    {
+      "epoch": 1.378563283922463,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.8961111295005444e-06,
+      "loss": 0.6057,
+      "mean_token_accuracy": 0.8767747431993484,
+      "num_tokens": 64716199.0,
+      "step": 605
+    },
+    {
+      "epoch": 1.3808437856328393,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.895573140745967e-06,
+      "loss": 0.6154,
+      "mean_token_accuracy": 0.8689918667078018,
+      "num_tokens": 64823123.0,
+      "step": 606
+    },
+    {
+      "epoch": 1.3831242873432155,
+      "grad_norm": 1.9296875,
+      "learning_rate": 4.895033792310323e-06,
+      "loss": 0.5859,
+      "mean_token_accuracy": 0.879554808139801,
+      "num_tokens": 64930304.0,
+      "step": 607
+    },
+    {
+      "epoch": 1.3854047890535917,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.894493084499736e-06,
+      "loss": 0.6176,
+      "mean_token_accuracy": 0.8728460669517517,
+      "num_tokens": 65037508.0,
+      "step": 608
+    },
+    {
+      "epoch": 1.387685290763968,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.893951017621103e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8793998062610626,
+      "num_tokens": 65145188.0,
+      "step": 609
+    },
+    {
+      "epoch": 1.3899657924743445,
+      "grad_norm": 6.59375,
+      "learning_rate": 4.893407591982088e-06,
+      "loss": 0.6289,
+      "mean_token_accuracy": 0.8728020936250687,
+      "num_tokens": 65251638.0,
+      "step": 610
+    },
+    {
+      "epoch": 1.3922462941847207,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.892862807891131e-06,
+      "loss": 0.6187,
+      "mean_token_accuracy": 0.8728552013635635,
+      "num_tokens": 65358821.0,
+      "step": 611
+    },
+    {
+      "epoch": 1.3945267958950969,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.89231666565744e-06,
+      "loss": 0.6356,
+      "mean_token_accuracy": 0.8699388653039932,
+      "num_tokens": 65466008.0,
+      "step": 612
+    },
+    {
+      "epoch": 1.3968072976054733,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.891769165590995e-06,
+      "loss": 0.6067,
+      "mean_token_accuracy": 0.8760533332824707,
+      "num_tokens": 65573422.0,
+      "step": 613
+    },
+    {
+      "epoch": 1.3990877993158495,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.891220308002547e-06,
+      "loss": 0.6115,
+      "mean_token_accuracy": 0.8774368613958359,
+      "num_tokens": 65679987.0,
+      "step": 614
+    },
+    {
+      "epoch": 1.401368301026226,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.890670093203617e-06,
+      "loss": 0.6178,
+      "mean_token_accuracy": 0.8716036677360535,
+      "num_tokens": 65786496.0,
+      "step": 615
+    },
+    {
+      "epoch": 1.403648802736602,
+      "grad_norm": 5.25,
+      "learning_rate": 4.890118521506494e-06,
+      "loss": 0.6349,
+      "mean_token_accuracy": 0.8668168038129807,
+      "num_tokens": 65893146.0,
+      "step": 616
+    },
+    {
+      "epoch": 1.4059293044469783,
+      "grad_norm": 3.25,
+      "learning_rate": 4.889565593224242e-06,
+      "loss": 0.6086,
+      "mean_token_accuracy": 0.8759677559137344,
+      "num_tokens": 65999804.0,
+      "step": 617
+    },
+    {
+      "epoch": 1.4082098061573547,
+      "grad_norm": 7.4375,
+      "learning_rate": 4.889011308670693e-06,
+      "loss": 0.601,
+      "mean_token_accuracy": 0.8765671104192734,
+      "num_tokens": 66106566.0,
+      "step": 618
+    },
+    {
+      "epoch": 1.4104903078677309,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.8884556681604445e-06,
+      "loss": 0.6154,
+      "mean_token_accuracy": 0.8747722804546356,
+      "num_tokens": 66213934.0,
+      "step": 619
+    },
+    {
+      "epoch": 1.4127708095781073,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.8878986720088715e-06,
+      "loss": 0.6039,
+      "mean_token_accuracy": 0.878739207983017,
+      "num_tokens": 66320771.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4150513112884835,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.8873403205321115e-06,
+      "loss": 0.6327,
+      "mean_token_accuracy": 0.8718791306018829,
+      "num_tokens": 66428168.0,
+      "step": 621
+    },
+    {
+      "epoch": 1.4173318129988597,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.886780614047075e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.877121314406395,
+      "num_tokens": 66535417.0,
+      "step": 622
+    },
+    {
+      "epoch": 1.419612314709236,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.886219552871441e-06,
+      "loss": 0.6183,
+      "mean_token_accuracy": 0.8733110427856445,
+      "num_tokens": 66642970.0,
+      "step": 623
+    },
+    {
+      "epoch": 1.4218928164196123,
+      "grad_norm": 6.0,
+      "learning_rate": 4.885657137323656e-06,
+      "loss": 0.6033,
+      "mean_token_accuracy": 0.8745451271533966,
+      "num_tokens": 66750313.0,
+      "step": 624
+    },
+    {
+      "epoch": 1.4241733181299887,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.885093367722937e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8829841762781143,
+      "num_tokens": 66857610.0,
+      "step": 625
+    },
+    {
+      "epoch": 1.426453819840365,
+      "grad_norm": 6.15625,
+      "learning_rate": 4.884528244389269e-06,
+      "loss": 0.6207,
+      "mean_token_accuracy": 0.8738205432891846,
+      "num_tokens": 66964325.0,
+      "step": 626
+    },
+    {
+      "epoch": 1.428734321550741,
+      "grad_norm": 1.9921875,
+      "learning_rate": 4.883961767643404e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8796354234218597,
+      "num_tokens": 67071449.0,
+      "step": 627
+    },
+    {
+      "epoch": 1.4310148232611175,
+      "grad_norm": 4.125,
+      "learning_rate": 4.883393937806864e-06,
+      "loss": 0.5991,
+      "mean_token_accuracy": 0.8746996223926544,
+      "num_tokens": 67178773.0,
+      "step": 628
+    },
+    {
+      "epoch": 1.4332953249714937,
+      "grad_norm": 6.21875,
+      "learning_rate": 4.882824755201938e-06,
+      "loss": 0.5954,
+      "mean_token_accuracy": 0.8757723271846771,
+      "num_tokens": 67286302.0,
+      "step": 629
+    },
+    {
+      "epoch": 1.4355758266818701,
+      "grad_norm": 8.4375,
+      "learning_rate": 4.8822542201516835e-06,
+      "loss": 0.6061,
+      "mean_token_accuracy": 0.8711700737476349,
+      "num_tokens": 67393466.0,
+      "step": 630
+    },
+    {
+      "epoch": 1.4378563283922463,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.881682332979925e-06,
+      "loss": 0.6216,
+      "mean_token_accuracy": 0.869665339589119,
+      "num_tokens": 67501293.0,
+      "step": 631
+    },
+    {
+      "epoch": 1.4401368301026225,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.881109094011254e-06,
+      "loss": 0.5956,
+      "mean_token_accuracy": 0.8787660300731659,
+      "num_tokens": 67609069.0,
+      "step": 632
+    },
+    {
+      "epoch": 1.442417331812999,
+      "grad_norm": 5.5,
+      "learning_rate": 4.88053450357103e-06,
+      "loss": 0.6089,
+      "mean_token_accuracy": 0.8752091825008392,
+      "num_tokens": 67716379.0,
+      "step": 633
+    },
+    {
+      "epoch": 1.444697833523375,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.87995856198538e-06,
+      "loss": 0.6164,
+      "mean_token_accuracy": 0.8721603453159332,
+      "num_tokens": 67823022.0,
+      "step": 634
+    },
+    {
+      "epoch": 1.4469783352337515,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.879381269581197e-06,
+      "loss": 0.6158,
+      "mean_token_accuracy": 0.8731086701154709,
+      "num_tokens": 67929811.0,
+      "step": 635
+    },
+    {
+      "epoch": 1.4492588369441277,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.878802626686141e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8828457891941071,
+      "num_tokens": 68036844.0,
+      "step": 636
+    },
+    {
+      "epoch": 1.451539338654504,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.8782226336286395e-06,
+      "loss": 0.6151,
+      "mean_token_accuracy": 0.8716375380754471,
+      "num_tokens": 68144018.0,
+      "step": 637
+    },
+    {
+      "epoch": 1.4538198403648803,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.8776412907378845e-06,
+      "loss": 0.5829,
+      "mean_token_accuracy": 0.8821705430746078,
+      "num_tokens": 68251176.0,
+      "step": 638
+    },
+    {
+      "epoch": 1.4561003420752565,
+      "grad_norm": 2.375,
+      "learning_rate": 4.877058598343835e-06,
+      "loss": 0.6087,
+      "mean_token_accuracy": 0.8753101974725723,
+      "num_tokens": 68358878.0,
+      "step": 639
+    },
+    {
+      "epoch": 1.458380843785633,
+      "grad_norm": 1.8515625,
+      "learning_rate": 4.876474556777216e-06,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.8794443160295486,
+      "num_tokens": 68466299.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.4606613454960091,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.8758891663695165e-06,
+      "loss": 0.602,
+      "mean_token_accuracy": 0.8779085278511047,
+      "num_tokens": 68573383.0,
+      "step": 641
+    },
+    {
+      "epoch": 1.4629418472063853,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.875302427452996e-06,
+      "loss": 0.6235,
+      "mean_token_accuracy": 0.8709381222724915,
+      "num_tokens": 68680054.0,
+      "step": 642
+    },
+    {
+      "epoch": 1.4652223489167617,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.874714340360674e-06,
+      "loss": 0.605,
+      "mean_token_accuracy": 0.8749048262834549,
+      "num_tokens": 68787643.0,
+      "step": 643
+    },
+    {
+      "epoch": 1.467502850627138,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.874124905426339e-06,
+      "loss": 0.6054,
+      "mean_token_accuracy": 0.8763098567724228,
+      "num_tokens": 68894888.0,
+      "step": 644
+    },
+    {
+      "epoch": 1.4697833523375143,
+      "grad_norm": 3.0,
+      "learning_rate": 4.873534122984541e-06,
+      "loss": 0.611,
+      "mean_token_accuracy": 0.8762774467468262,
+      "num_tokens": 69001443.0,
+      "step": 645
+    },
+    {
+      "epoch": 1.4720638540478905,
+      "grad_norm": 2.125,
+      "learning_rate": 4.872941993370598e-06,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.882948562502861,
+      "num_tokens": 69108474.0,
+      "step": 646
+    },
+    {
+      "epoch": 1.4743443557582667,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.872348516920591e-06,
+      "loss": 0.6272,
+      "mean_token_accuracy": 0.8692348450422287,
+      "num_tokens": 69215286.0,
+      "step": 647
+    },
+    {
+      "epoch": 1.4766248574686431,
+      "grad_norm": 1.890625,
+      "learning_rate": 4.8717536939713665e-06,
+      "loss": 0.5848,
+      "mean_token_accuracy": 0.8783621490001678,
+      "num_tokens": 69322791.0,
+      "step": 648
+    },
+    {
+      "epoch": 1.4789053591790193,
+      "grad_norm": 4.75,
+      "learning_rate": 4.871157524860533e-06,
+      "loss": 0.603,
+      "mean_token_accuracy": 0.8760869204998016,
+      "num_tokens": 69429973.0,
+      "step": 649
+    },
+    {
+      "epoch": 1.4811858608893957,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.870560009926465e-06,
+      "loss": 0.6125,
+      "mean_token_accuracy": 0.8765671104192734,
+      "num_tokens": 69537546.0,
+      "step": 650
+    },
+    {
+      "epoch": 1.483466362599772,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.869961149508301e-06,
+      "loss": 0.6213,
+      "mean_token_accuracy": 0.8740588128566742,
+      "num_tokens": 69644078.0,
+      "step": 651
+    },
+    {
+      "epoch": 1.4857468643101481,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.869360943945943e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8758237510919571,
+      "num_tokens": 69750942.0,
+      "step": 652
+    },
+    {
+      "epoch": 1.4880273660205245,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.868759393580054e-06,
+      "loss": 0.6004,
+      "mean_token_accuracy": 0.8765598982572556,
+      "num_tokens": 69857823.0,
+      "step": 653
+    },
+    {
+      "epoch": 1.4903078677309007,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.868156498752066e-06,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8746924102306366,
+      "num_tokens": 69964959.0,
+      "step": 654
+    },
+    {
+      "epoch": 1.4925883694412772,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.8675522598041675e-06,
+      "loss": 0.6092,
+      "mean_token_accuracy": 0.8732979595661163,
+      "num_tokens": 70071686.0,
+      "step": 655
+    },
+    {
+      "epoch": 1.4948688711516533,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.866946677079314e-06,
+      "loss": 0.5976,
+      "mean_token_accuracy": 0.875095933675766,
+      "num_tokens": 70179037.0,
+      "step": 656
+    },
+    {
+      "epoch": 1.4971493728620295,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.866339750921222e-06,
+      "loss": 0.6294,
+      "mean_token_accuracy": 0.8688063323497772,
+      "num_tokens": 70285939.0,
+      "step": 657
+    },
+    {
+      "epoch": 1.499429874572406,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.86573148167437e-06,
+      "loss": 0.6153,
+      "mean_token_accuracy": 0.8754535466432571,
+      "num_tokens": 70393254.0,
+      "step": 658
+    },
+    {
+      "epoch": 1.5017103762827824,
+      "grad_norm": 5.34375,
+      "learning_rate": 4.865121869684003e-06,
+      "loss": 0.5896,
+      "mean_token_accuracy": 0.8764495700597763,
+      "num_tokens": 70500856.0,
+      "step": 659
+    },
+    {
+      "epoch": 1.5039908779931586,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.864510915296122e-06,
+      "loss": 0.597,
+      "mean_token_accuracy": 0.8765180259943008,
+      "num_tokens": 70608662.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5039908779931586,
+      "eval_loss": 0.612367570400238,
+      "eval_mean_token_accuracy": 0.8747105800153638,
+      "eval_num_tokens": 70608662.0,
+      "eval_runtime": 58.5955,
+      "eval_samples_per_second": 143.1,
+      "eval_steps_per_second": 4.488,
+      "step": 660
+    },
+    {
+      "epoch": 1.5062713797035348,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.8638986188574955e-06,
+      "loss": 0.5952,
+      "mean_token_accuracy": 0.87701815366745,
+      "num_tokens": 70715623.0,
+      "step": 661
+    },
+    {
+      "epoch": 1.508551881413911,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.863284980715649e-06,
+      "loss": 0.6125,
+      "mean_token_accuracy": 0.8763703256845474,
+      "num_tokens": 70822494.0,
+      "step": 662
+    },
+    {
+      "epoch": 1.5108323831242874,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.8626700012188724e-06,
+      "loss": 0.6303,
+      "mean_token_accuracy": 0.8716311007738113,
+      "num_tokens": 70929521.0,
+      "step": 663
+    },
+    {
+      "epoch": 1.5131128848346638,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.8620536807162164e-06,
+      "loss": 0.6084,
+      "mean_token_accuracy": 0.8747840225696564,
+      "num_tokens": 71036264.0,
+      "step": 664
+    },
+    {
+      "epoch": 1.51539338654504,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.861436019557492e-06,
+      "loss": 0.6227,
+      "mean_token_accuracy": 0.8713407516479492,
+      "num_tokens": 71142930.0,
+      "step": 665
+    },
+    {
+      "epoch": 1.5176738882554162,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.8608170180932725e-06,
+      "loss": 0.6033,
+      "mean_token_accuracy": 0.8776234537363052,
+      "num_tokens": 71249846.0,
+      "step": 666
+    },
+    {
+      "epoch": 1.5199543899657924,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.860196676674891e-06,
+      "loss": 0.5998,
+      "mean_token_accuracy": 0.8785290718078613,
+      "num_tokens": 71356611.0,
+      "step": 667
+    },
+    {
+      "epoch": 1.5222348916761688,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.8595749956544414e-06,
+      "loss": 0.6036,
+      "mean_token_accuracy": 0.8775981813669205,
+      "num_tokens": 71463513.0,
+      "step": 668
+    },
+    {
+      "epoch": 1.5245153933865452,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.858951975384777e-06,
+      "loss": 0.6166,
+      "mean_token_accuracy": 0.874157264828682,
+      "num_tokens": 71570255.0,
+      "step": 669
+    },
+    {
+      "epoch": 1.5267958950969214,
+      "grad_norm": 5.875,
+      "learning_rate": 4.858327616219513e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8791629523038864,
+      "num_tokens": 71677310.0,
+      "step": 670
+    },
+    {
+      "epoch": 1.5290763968072976,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.857701918513023e-06,
+      "loss": 0.6137,
+      "mean_token_accuracy": 0.8735476732254028,
+      "num_tokens": 71783915.0,
+      "step": 671
+    },
+    {
+      "epoch": 1.5313568985176738,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.857074882620442e-06,
+      "loss": 0.6026,
+      "mean_token_accuracy": 0.8756736516952515,
+      "num_tokens": 71891495.0,
+      "step": 672
+    },
+    {
+      "epoch": 1.5336374002280502,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.856446508897662e-06,
+      "loss": 0.5949,
+      "mean_token_accuracy": 0.8775335848331451,
+      "num_tokens": 71999208.0,
+      "step": 673
+    },
+    {
+      "epoch": 1.5359179019384266,
+      "grad_norm": 8.1875,
+      "learning_rate": 4.8558167977013365e-06,
+      "loss": 0.6152,
+      "mean_token_accuracy": 0.8727478533983231,
+      "num_tokens": 72106904.0,
+      "step": 674
+    },
+    {
+      "epoch": 1.5381984036488028,
+      "grad_norm": 6.59375,
+      "learning_rate": 4.8551857493888775e-06,
+      "loss": 0.5947,
+      "mean_token_accuracy": 0.8790593892335892,
+      "num_tokens": 72213551.0,
+      "step": 675
+    },
+    {
+      "epoch": 1.540478905359179,
+      "grad_norm": 6.1875,
+      "learning_rate": 4.854553364318456e-06,
+      "loss": 0.6048,
+      "mean_token_accuracy": 0.8734369874000549,
+      "num_tokens": 72320910.0,
+      "step": 676
+    },
+    {
+      "epoch": 1.5427594070695552,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.8539196428490016e-06,
+      "loss": 0.6387,
+      "mean_token_accuracy": 0.8696837574243546,
+      "num_tokens": 72427539.0,
+      "step": 677
+    },
+    {
+      "epoch": 1.5450399087799316,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.8532845853402015e-06,
+      "loss": 0.622,
+      "mean_token_accuracy": 0.8762244433164597,
+      "num_tokens": 72534311.0,
+      "step": 678
+    },
+    {
+      "epoch": 1.547320410490308,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.8526481921525035e-06,
+      "loss": 0.615,
+      "mean_token_accuracy": 0.872207522392273,
+      "num_tokens": 72641361.0,
+      "step": 679
+    },
+    {
+      "epoch": 1.5496009122006842,
+      "grad_norm": 3.625,
+      "learning_rate": 4.85201046364711e-06,
+      "loss": 0.6082,
+      "mean_token_accuracy": 0.8779794573783875,
+      "num_tokens": 72748703.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.5518814139110604,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.851371400185986e-06,
+      "loss": 0.6091,
+      "mean_token_accuracy": 0.8768866658210754,
+      "num_tokens": 72855762.0,
+      "step": 681
+    },
+    {
+      "epoch": 1.5541619156214366,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.85073100213185e-06,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.8712277412414551,
+      "num_tokens": 72962645.0,
+      "step": 682
+    },
+    {
+      "epoch": 1.556442417331813,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.8500892698481784e-06,
+      "loss": 0.6018,
+      "mean_token_accuracy": 0.8731422871351242,
+      "num_tokens": 73069760.0,
+      "step": 683
+    },
+    {
+      "epoch": 1.5587229190421894,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.849446203699209e-06,
+      "loss": 0.6125,
+      "mean_token_accuracy": 0.8730466067790985,
+      "num_tokens": 73176865.0,
+      "step": 684
+    },
+    {
+      "epoch": 1.5610034207525656,
+      "grad_norm": 2.375,
+      "learning_rate": 4.848801804049932e-06,
+      "loss": 0.6069,
+      "mean_token_accuracy": 0.8776687532663345,
+      "num_tokens": 73284011.0,
+      "step": 685
+    },
+    {
+      "epoch": 1.5632839224629418,
+      "grad_norm": 3.9375,
+      "learning_rate": 4.848156071266095e-06,
+      "loss": 0.5939,
+      "mean_token_accuracy": 0.8770481199026108,
+      "num_tokens": 73390880.0,
+      "step": 686
+    },
+    {
+      "epoch": 1.565564424173318,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.847509005714207e-06,
+      "loss": 0.6151,
+      "mean_token_accuracy": 0.8751052618026733,
+      "num_tokens": 73497575.0,
+      "step": 687
+    },
+    {
+      "epoch": 1.5678449258836944,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.846860607761527e-06,
+      "loss": 0.6022,
+      "mean_token_accuracy": 0.8736310601234436,
+      "num_tokens": 73604375.0,
+      "step": 688
+    },
+    {
+      "epoch": 1.5701254275940708,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.8462108777760734e-06,
+      "loss": 0.6045,
+      "mean_token_accuracy": 0.873234361410141,
+      "num_tokens": 73712100.0,
+      "step": 689
+    },
+    {
+      "epoch": 1.572405929304447,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.845559816126622e-06,
+      "loss": 0.6198,
+      "mean_token_accuracy": 0.8704589009284973,
+      "num_tokens": 73819126.0,
+      "step": 690
+    },
+    {
+      "epoch": 1.5746864310148232,
+      "grad_norm": 4.25,
+      "learning_rate": 4.844907423182699e-06,
+      "loss": 0.6048,
+      "mean_token_accuracy": 0.8772129714488983,
+      "num_tokens": 73925905.0,
+      "step": 691
+    },
+    {
+      "epoch": 1.5769669327251994,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.844253699314596e-06,
+      "loss": 0.5999,
+      "mean_token_accuracy": 0.875739112496376,
+      "num_tokens": 74033596.0,
+      "step": 692
+    },
+    {
+      "epoch": 1.5792474344355758,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.843598644893349e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.8757044672966003,
+      "num_tokens": 74140480.0,
+      "step": 693
+    },
+    {
+      "epoch": 1.5815279361459522,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.842942260290757e-06,
+      "loss": 0.6247,
+      "mean_token_accuracy": 0.8707884252071381,
+      "num_tokens": 74247691.0,
+      "step": 694
+    },
+    {
+      "epoch": 1.5838084378563284,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.84228454587937e-06,
+      "loss": 0.6162,
+      "mean_token_accuracy": 0.8712183386087418,
+      "num_tokens": 74354568.0,
+      "step": 695
+    },
+    {
+      "epoch": 1.5860889395667046,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.841625502032495e-06,
+      "loss": 0.5966,
+      "mean_token_accuracy": 0.8787146657705307,
+      "num_tokens": 74461346.0,
+      "step": 696
+    },
+    {
+      "epoch": 1.5883694412770808,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.84096512912419e-06,
+      "loss": 0.6141,
+      "mean_token_accuracy": 0.8755437433719635,
+      "num_tokens": 74567912.0,
+      "step": 697
+    },
+    {
+      "epoch": 1.5906499429874572,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.8403034275292735e-06,
+      "loss": 0.6005,
+      "mean_token_accuracy": 0.8786727637052536,
+      "num_tokens": 74674714.0,
+      "step": 698
+    },
+    {
+      "epoch": 1.5929304446978336,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.839640397623312e-06,
+      "loss": 0.604,
+      "mean_token_accuracy": 0.8754930794239044,
+      "num_tokens": 74781656.0,
+      "step": 699
+    },
+    {
+      "epoch": 1.5952109464082098,
+      "grad_norm": 4.125,
+      "learning_rate": 4.83897603978263e-06,
+      "loss": 0.6078,
+      "mean_token_accuracy": 0.8756787329912186,
+      "num_tokens": 74888714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.597491448118586,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.838310354384304e-06,
+      "loss": 0.6248,
+      "mean_token_accuracy": 0.8733028322458267,
+      "num_tokens": 74995535.0,
+      "step": 701
+    },
+    {
+      "epoch": 1.5997719498289624,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.8376433418061615e-06,
+      "loss": 0.6095,
+      "mean_token_accuracy": 0.8739898204803467,
+      "num_tokens": 75102558.0,
+      "step": 702
+    },
+    {
+      "epoch": 1.6020524515393386,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.8369750024267904e-06,
+      "loss": 0.6103,
+      "mean_token_accuracy": 0.8737581223249435,
+      "num_tokens": 75209920.0,
+      "step": 703
+    },
+    {
+      "epoch": 1.604332953249715,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.836305336625523e-06,
+      "loss": 0.6218,
+      "mean_token_accuracy": 0.8702377676963806,
+      "num_tokens": 75316855.0,
+      "step": 704
+    },
+    {
+      "epoch": 1.6066134549600912,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.835634344782453e-06,
+      "loss": 0.6076,
+      "mean_token_accuracy": 0.8709095418453217,
+      "num_tokens": 75423929.0,
+      "step": 705
+    },
+    {
+      "epoch": 1.6088939566704674,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.834962027278418e-06,
+      "loss": 0.6074,
+      "mean_token_accuracy": 0.873084306716919,
+      "num_tokens": 75531031.0,
+      "step": 706
+    },
+    {
+      "epoch": 1.6111744583808438,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.834288384495015e-06,
+      "loss": 0.6219,
+      "mean_token_accuracy": 0.8735803365707397,
+      "num_tokens": 75637818.0,
+      "step": 707
+    },
+    {
+      "epoch": 1.61345496009122,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.833613416814591e-06,
+      "loss": 0.5976,
+      "mean_token_accuracy": 0.8781051784753799,
+      "num_tokens": 75744879.0,
+      "step": 708
+    },
+    {
+      "epoch": 1.6157354618015964,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.832937124620243e-06,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8801741451025009,
+      "num_tokens": 75852406.0,
+      "step": 709
+    },
+    {
+      "epoch": 1.6180159635119726,
+      "grad_norm": 5.84375,
+      "learning_rate": 4.832259508295822e-06,
+      "loss": 0.5971,
+      "mean_token_accuracy": 0.8746786862611771,
+      "num_tokens": 75959416.0,
+      "step": 710
+    },
+    {
+      "epoch": 1.6202964652223488,
+      "grad_norm": 4.125,
+      "learning_rate": 4.831580568225931e-06,
+      "loss": 0.6059,
+      "mean_token_accuracy": 0.8775349110364914,
+      "num_tokens": 76067055.0,
+      "step": 711
+    },
+    {
+      "epoch": 1.6225769669327252,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.830900304795921e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8796486407518387,
+      "num_tokens": 76174493.0,
+      "step": 712
+    },
+    {
+      "epoch": 1.6248574686431014,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.8302187183918996e-06,
+      "loss": 0.6092,
+      "mean_token_accuracy": 0.8721607774496078,
+      "num_tokens": 76281621.0,
+      "step": 713
+    },
+    {
+      "epoch": 1.6271379703534778,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.8295358094007184e-06,
+      "loss": 0.6131,
+      "mean_token_accuracy": 0.8758230358362198,
+      "num_tokens": 76387782.0,
+      "step": 714
+    },
+    {
+      "epoch": 1.629418472063854,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.828851578209986e-06,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.8804755955934525,
+      "num_tokens": 76495248.0,
+      "step": 715
+    },
+    {
+      "epoch": 1.6316989737742302,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.828166025208059e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8788855522871017,
+      "num_tokens": 76602322.0,
+      "step": 716
+    },
+    {
+      "epoch": 1.6339794754846066,
+      "grad_norm": 3.5,
+      "learning_rate": 4.8274791507840416e-06,
+      "loss": 0.6142,
+      "mean_token_accuracy": 0.876452699303627,
+      "num_tokens": 76709429.0,
+      "step": 717
+    },
+    {
+      "epoch": 1.636259977194983,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.826790955327793e-06,
+      "loss": 0.6182,
+      "mean_token_accuracy": 0.8712098151445389,
+      "num_tokens": 76817050.0,
+      "step": 718
+    },
+    {
+      "epoch": 1.6385404789053593,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.826101439229918e-06,
+      "loss": 0.6144,
+      "mean_token_accuracy": 0.8716437220573425,
+      "num_tokens": 76923670.0,
+      "step": 719
+    },
+    {
+      "epoch": 1.6408209806157354,
+      "grad_norm": 5.0,
+      "learning_rate": 4.825410602881774e-06,
+      "loss": 0.6147,
+      "mean_token_accuracy": 0.8725506961345673,
+      "num_tokens": 77030550.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.6431014823261116,
+      "grad_norm": 4.25,
+      "learning_rate": 4.824718446675465e-06,
+      "loss": 0.6036,
+      "mean_token_accuracy": 0.8758846968412399,
+      "num_tokens": 77137310.0,
+      "step": 721
+    },
+    {
+      "epoch": 1.645381984036488,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.8240249710038455e-06,
+      "loss": 0.5956,
+      "mean_token_accuracy": 0.8754921108484268,
+      "num_tokens": 77244391.0,
+      "step": 722
+    },
+    {
+      "epoch": 1.6476624857468645,
+      "grad_norm": 5.40625,
+      "learning_rate": 4.82333017626052e-06,
+      "loss": 0.609,
+      "mean_token_accuracy": 0.8759109824895859,
+      "num_tokens": 77350980.0,
+      "step": 723
+    },
+    {
+      "epoch": 1.6499429874572407,
+      "grad_norm": 7.5,
+      "learning_rate": 4.82263406283984e-06,
+      "loss": 0.6217,
+      "mean_token_accuracy": 0.8713529407978058,
+      "num_tokens": 77457863.0,
+      "step": 724
+    },
+    {
+      "epoch": 1.6522234891676169,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.821936631136907e-06,
+      "loss": 0.612,
+      "mean_token_accuracy": 0.8779774755239487,
+      "num_tokens": 77564956.0,
+      "step": 725
+    },
+    {
+      "epoch": 1.654503990877993,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.821237881547567e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8784075975418091,
+      "num_tokens": 77672980.0,
+      "step": 726
+    },
+    {
+      "epoch": 1.6567844925883695,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.82053781446842e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8801317662000656,
+      "num_tokens": 77779982.0,
+      "step": 727
+    },
+    {
+      "epoch": 1.6590649942987459,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.819836430296809e-06,
+      "loss": 0.6133,
+      "mean_token_accuracy": 0.8747014552354813,
+      "num_tokens": 77886824.0,
+      "step": 728
+    },
+    {
+      "epoch": 1.661345496009122,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.819133729430826e-06,
+      "loss": 0.606,
+      "mean_token_accuracy": 0.874286100268364,
+      "num_tokens": 77993947.0,
+      "step": 729
+    },
+    {
+      "epoch": 1.6636259977194983,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.818429712269312e-06,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.8773009479045868,
+      "num_tokens": 78101464.0,
+      "step": 730
+    },
+    {
+      "epoch": 1.6659064994298745,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.8177243792118515e-06,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8781311810016632,
+      "num_tokens": 78208412.0,
+      "step": 731
+    },
+    {
+      "epoch": 1.6681870011402509,
+      "grad_norm": 5.03125,
+      "learning_rate": 4.8170177306587785e-06,
+      "loss": 0.5947,
+      "mean_token_accuracy": 0.8763359040021896,
+      "num_tokens": 78316677.0,
+      "step": 732
+    },
+    {
+      "epoch": 1.6704675028506273,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.8163097670111735e-06,
+      "loss": 0.6005,
+      "mean_token_accuracy": 0.8768177330493927,
+      "num_tokens": 78424519.0,
+      "step": 733
+    },
+    {
+      "epoch": 1.6727480045610035,
+      "grad_norm": 5.1875,
+      "learning_rate": 4.815600488670863e-06,
+      "loss": 0.6127,
+      "mean_token_accuracy": 0.8756203055381775,
+      "num_tokens": 78531596.0,
+      "step": 734
+    },
+    {
+      "epoch": 1.6750285062713797,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.81488989604042e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8781074434518814,
+      "num_tokens": 78638185.0,
+      "step": 735
+    },
+    {
+      "epoch": 1.6773090079817559,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.814177989523162e-06,
+      "loss": 0.5993,
+      "mean_token_accuracy": 0.8762544244527817,
+      "num_tokens": 78745981.0,
+      "step": 736
+    },
+    {
+      "epoch": 1.6795895096921323,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.813464769523154e-06,
+      "loss": 0.6209,
+      "mean_token_accuracy": 0.8740936666727066,
+      "num_tokens": 78853266.0,
+      "step": 737
+    },
+    {
+      "epoch": 1.6818700114025087,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.812750236445206e-06,
+      "loss": 0.6025,
+      "mean_token_accuracy": 0.8746587634086609,
+      "num_tokens": 78960887.0,
+      "step": 738
+    },
+    {
+      "epoch": 1.6841505131128849,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.812034390694874e-06,
+      "loss": 0.6203,
+      "mean_token_accuracy": 0.871962770819664,
+      "num_tokens": 79068181.0,
+      "step": 739
+    },
+    {
+      "epoch": 1.686431014823261,
+      "grad_norm": 2.625,
+      "learning_rate": 4.811317232678456e-06,
+      "loss": 0.5949,
+      "mean_token_accuracy": 0.8781395107507706,
+      "num_tokens": 79175196.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.6887115165336373,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.810598762803e-06,
+      "loss": 0.6016,
+      "mean_token_accuracy": 0.8768087178468704,
+      "num_tokens": 79282678.0,
+      "step": 741
+    },
+    {
+      "epoch": 1.6909920182440137,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.809878981476293e-06,
+      "loss": 0.6145,
+      "mean_token_accuracy": 0.8780290633440018,
+      "num_tokens": 79389475.0,
+      "step": 742
+    },
+    {
+      "epoch": 1.69327251995439,
+      "grad_norm": 3.5,
+      "learning_rate": 4.80915788910687e-06,
+      "loss": 0.6114,
+      "mean_token_accuracy": 0.8754473924636841,
+      "num_tokens": 79496094.0,
+      "step": 743
+    },
+    {
+      "epoch": 1.6955530216647663,
+      "grad_norm": 4.125,
+      "learning_rate": 4.80843548610401e-06,
+      "loss": 0.5918,
+      "mean_token_accuracy": 0.8804915100336075,
+      "num_tokens": 79603317.0,
+      "step": 744
+    },
+    {
+      "epoch": 1.6978335233751425,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.807711772877733e-06,
+      "loss": 0.6007,
+      "mean_token_accuracy": 0.875912070274353,
+      "num_tokens": 79709931.0,
+      "step": 745
+    },
+    {
+      "epoch": 1.7001140250855187,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.8069867498388066e-06,
+      "loss": 0.6093,
+      "mean_token_accuracy": 0.8721041977405548,
+      "num_tokens": 79816476.0,
+      "step": 746
+    },
+    {
+      "epoch": 1.702394526795895,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.806260417398739e-06,
+      "loss": 0.6006,
+      "mean_token_accuracy": 0.8775873631238937,
+      "num_tokens": 79923080.0,
+      "step": 747
+    },
+    {
+      "epoch": 1.7046750285062715,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.805532775969783e-06,
+      "loss": 0.607,
+      "mean_token_accuracy": 0.8796124309301376,
+      "num_tokens": 80030054.0,
+      "step": 748
+    },
+    {
+      "epoch": 1.7069555302166477,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.804803825964933e-06,
+      "loss": 0.6043,
+      "mean_token_accuracy": 0.8757105469703674,
+      "num_tokens": 80137000.0,
+      "step": 749
+    },
+    {
+      "epoch": 1.709236031927024,
+      "grad_norm": 4.9375,
+      "learning_rate": 4.804073567797928e-06,
+      "loss": 0.6095,
+      "mean_token_accuracy": 0.874399796128273,
+      "num_tokens": 80244175.0,
+      "step": 750
+    },
+    {
+      "epoch": 1.7115165336374,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.803342001883247e-06,
+      "loss": 0.6234,
+      "mean_token_accuracy": 0.8730274736881256,
+      "num_tokens": 80351450.0,
+      "step": 751
+    },
+    {
+      "epoch": 1.7137970353477765,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.802609128636113e-06,
+      "loss": 0.609,
+      "mean_token_accuracy": 0.8778874278068542,
+      "num_tokens": 80458377.0,
+      "step": 752
+    },
+    {
+      "epoch": 1.716077537058153,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.801874948472492e-06,
+      "loss": 0.6195,
+      "mean_token_accuracy": 0.8746256977319717,
+      "num_tokens": 80565320.0,
+      "step": 753
+    },
+    {
+      "epoch": 1.718358038768529,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.801139461809089e-06,
+      "loss": 0.5944,
+      "mean_token_accuracy": 0.8773505538702011,
+      "num_tokens": 80672618.0,
+      "step": 754
+    },
+    {
+      "epoch": 1.7206385404789053,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.800402669063353e-06,
+      "loss": 0.6376,
+      "mean_token_accuracy": 0.8712969422340393,
+      "num_tokens": 80779049.0,
+      "step": 755
+    },
+    {
+      "epoch": 1.7229190421892815,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.799664570653473e-06,
+      "loss": 0.6028,
+      "mean_token_accuracy": 0.8769599795341492,
+      "num_tokens": 80886174.0,
+      "step": 756
+    },
+    {
+      "epoch": 1.725199543899658,
+      "grad_norm": 4.53125,
+      "learning_rate": 4.79892516699838e-06,
+      "loss": 0.5913,
+      "mean_token_accuracy": 0.8780887126922607,
+      "num_tokens": 80993329.0,
+      "step": 757
+    },
+    {
+      "epoch": 1.7274800456100343,
+      "grad_norm": 6.96875,
+      "learning_rate": 4.798184458517745e-06,
+      "loss": 0.5998,
+      "mean_token_accuracy": 0.8770759999752045,
+      "num_tokens": 81101103.0,
+      "step": 758
+    },
+    {
+      "epoch": 1.7297605473204105,
+      "grad_norm": 6.6875,
+      "learning_rate": 4.797442445631978e-06,
+      "loss": 0.5998,
+      "mean_token_accuracy": 0.8771584331989288,
+      "num_tokens": 81208123.0,
+      "step": 759
+    },
+    {
+      "epoch": 1.7320410490307867,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.7966991287622335e-06,
+      "loss": 0.6297,
+      "mean_token_accuracy": 0.8718820959329605,
+      "num_tokens": 81315003.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.734321550741163,
+      "grad_norm": 8.0625,
+      "learning_rate": 4.795954508330403e-06,
+      "loss": 0.6039,
+      "mean_token_accuracy": 0.8779590427875519,
+      "num_tokens": 81421672.0,
+      "step": 761
+    },
+    {
+      "epoch": 1.7366020524515393,
+      "grad_norm": 11.4375,
+      "learning_rate": 4.795208584759119e-06,
+      "loss": 0.6015,
+      "mean_token_accuracy": 0.8754216283559799,
+      "num_tokens": 81528963.0,
+      "step": 762
+    },
+    {
+      "epoch": 1.7388825541619157,
+      "grad_norm": 7.625,
+      "learning_rate": 4.794461358471753e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8796775341033936,
+      "num_tokens": 81636332.0,
+      "step": 763
+    },
+    {
+      "epoch": 1.741163055872292,
+      "grad_norm": 9.5625,
+      "learning_rate": 4.7937128298924155e-06,
+      "loss": 0.6187,
+      "mean_token_accuracy": 0.8761927634477615,
+      "num_tokens": 81743814.0,
+      "step": 764
+    },
+    {
+      "epoch": 1.7434435575826681,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.7929629994459584e-06,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8852293342351913,
+      "num_tokens": 81851249.0,
+      "step": 765
+    },
+    {
+      "epoch": 1.7457240592930443,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.792211867557969e-06,
+      "loss": 0.5976,
+      "mean_token_accuracy": 0.8791010826826096,
+      "num_tokens": 81958684.0,
+      "step": 766
+    },
+    {
+      "epoch": 1.7480045610034207,
+      "grad_norm": 5.625,
+      "learning_rate": 4.7914594346547774e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8784136772155762,
+      "num_tokens": 82065522.0,
+      "step": 767
+    },
+    {
+      "epoch": 1.7502850627137971,
+      "grad_norm": 5.75,
+      "learning_rate": 4.790705701163449e-06,
+      "loss": 0.6138,
+      "mean_token_accuracy": 0.8732837587594986,
+      "num_tokens": 82172365.0,
+      "step": 768
+    },
+    {
+      "epoch": 1.7525655644241733,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.789950667511789e-06,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.87830650806427,
+      "num_tokens": 82279622.0,
+      "step": 769
+    },
+    {
+      "epoch": 1.7548460661345495,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.789194334128338e-06,
+      "loss": 0.6031,
+      "mean_token_accuracy": 0.877634271979332,
+      "num_tokens": 82386626.0,
+      "step": 770
+    },
+    {
+      "epoch": 1.757126567844926,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.788436701442378e-06,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.8743524998426437,
+      "num_tokens": 82494070.0,
+      "step": 771
+    },
+    {
+      "epoch": 1.7594070695553021,
+      "grad_norm": 4.9375,
+      "learning_rate": 4.787677769883926e-06,
+      "loss": 0.6018,
+      "mean_token_accuracy": 0.8753460049629211,
+      "num_tokens": 82601259.0,
+      "step": 772
+    },
+    {
+      "epoch": 1.7616875712656785,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.786917539883738e-06,
+      "loss": 0.5842,
+      "mean_token_accuracy": 0.8806093335151672,
+      "num_tokens": 82708412.0,
+      "step": 773
+    },
+    {
+      "epoch": 1.7639680729760547,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.786156011873304e-06,
+      "loss": 0.6122,
+      "mean_token_accuracy": 0.8751035928726196,
+      "num_tokens": 82815320.0,
+      "step": 774
+    },
+    {
+      "epoch": 1.766248574686431,
+      "grad_norm": 2.875,
+      "learning_rate": 4.785393186284854e-06,
+      "loss": 0.5978,
+      "mean_token_accuracy": 0.8772329390048981,
+      "num_tokens": 82922009.0,
+      "step": 775
+    },
+    {
+      "epoch": 1.7685290763968073,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.784629063551354e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8803988993167877,
+      "num_tokens": 83029138.0,
+      "step": 776
+    },
+    {
+      "epoch": 1.7708095781071835,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.783863644106502e-06,
+      "loss": 0.5954,
+      "mean_token_accuracy": 0.8786198645830154,
+      "num_tokens": 83136422.0,
+      "step": 777
+    },
+    {
+      "epoch": 1.77309007981756,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.783096928384739e-06,
+      "loss": 0.6054,
+      "mean_token_accuracy": 0.8743276447057724,
+      "num_tokens": 83243905.0,
+      "step": 778
+    },
+    {
+      "epoch": 1.7753705815279361,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.782328916821235e-06,
+      "loss": 0.6018,
+      "mean_token_accuracy": 0.8728191703557968,
+      "num_tokens": 83350681.0,
+      "step": 779
+    },
+    {
+      "epoch": 1.7776510832383123,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.7815596098519004e-06,
+      "loss": 0.6099,
+      "mean_token_accuracy": 0.873179629445076,
+      "num_tokens": 83457754.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.7799315849486887,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.780789007913379e-06,
+      "loss": 0.6103,
+      "mean_token_accuracy": 0.8735631853342056,
+      "num_tokens": 83565053.0,
+      "step": 781
+    },
+    {
+      "epoch": 1.782212086659065,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.780017111443048e-06,
+      "loss": 0.5854,
+      "mean_token_accuracy": 0.8767279386520386,
+      "num_tokens": 83672036.0,
+      "step": 782
+    },
+    {
+      "epoch": 1.7844925883694414,
+      "grad_norm": 4.25,
+      "learning_rate": 4.779243920879023e-06,
+      "loss": 0.6206,
+      "mean_token_accuracy": 0.8717406392097473,
+      "num_tokens": 83778589.0,
+      "step": 783
+    },
+    {
+      "epoch": 1.7867730900798175,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.77846943666015e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8808600008487701,
+      "num_tokens": 83886086.0,
+      "step": 784
+    },
+    {
+      "epoch": 1.7890535917901937,
+      "grad_norm": 2.5,
+      "learning_rate": 4.777693659226013e-06,
+      "loss": 0.612,
+      "mean_token_accuracy": 0.8717308193445206,
+      "num_tokens": 83993388.0,
+      "step": 785
+    },
+    {
+      "epoch": 1.7913340935005702,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.776916589016928e-06,
+      "loss": 0.612,
+      "mean_token_accuracy": 0.874013751745224,
+      "num_tokens": 84100238.0,
+      "step": 786
+    },
+    {
+      "epoch": 1.7936145952109466,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.776138226473944e-06,
+      "loss": 0.6069,
+      "mean_token_accuracy": 0.8767507672309875,
+      "num_tokens": 84206817.0,
+      "step": 787
+    },
+    {
+      "epoch": 1.7958950969213228,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.775358572038845e-06,
+      "loss": 0.6163,
+      "mean_token_accuracy": 0.8742737025022507,
+      "num_tokens": 84314065.0,
+      "step": 788
+    },
+    {
+      "epoch": 1.798175598631699,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.774577626154148e-06,
+      "loss": 0.6074,
+      "mean_token_accuracy": 0.873917818069458,
+      "num_tokens": 84421307.0,
+      "step": 789
+    },
+    {
+      "epoch": 1.8004561003420751,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.773795389263104e-06,
+      "loss": 0.598,
+      "mean_token_accuracy": 0.8776136040687561,
+      "num_tokens": 84528605.0,
+      "step": 790
+    },
+    {
+      "epoch": 1.8027366020524516,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.773011861809694e-06,
+      "loss": 0.6154,
+      "mean_token_accuracy": 0.8759381324052811,
+      "num_tokens": 84635657.0,
+      "step": 791
+    },
+    {
+      "epoch": 1.805017103762828,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.772227044238632e-06,
+      "loss": 0.6173,
+      "mean_token_accuracy": 0.8727094084024429,
+      "num_tokens": 84742929.0,
+      "step": 792
+    },
+    {
+      "epoch": 1.8072976054732042,
+      "grad_norm": 3.375,
+      "learning_rate": 4.771440936995367e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.876227617263794,
+      "num_tokens": 84849828.0,
+      "step": 793
+    },
+    {
+      "epoch": 1.8095781071835804,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.770653540526079e-06,
+      "loss": 0.6116,
+      "mean_token_accuracy": 0.8727654963731766,
+      "num_tokens": 84956976.0,
+      "step": 794
+    },
+    {
+      "epoch": 1.8118586088939566,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.7698648552776785e-06,
+      "loss": 0.6202,
+      "mean_token_accuracy": 0.8722630441188812,
+      "num_tokens": 85063714.0,
+      "step": 795
+    },
+    {
+      "epoch": 1.814139110604333,
+      "grad_norm": 6.9375,
+      "learning_rate": 4.769074881697806e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.883360892534256,
+      "num_tokens": 85171018.0,
+      "step": 796
+    },
+    {
+      "epoch": 1.8164196123147094,
+      "grad_norm": 8.1875,
+      "learning_rate": 4.768283620234838e-06,
+      "loss": 0.5899,
+      "mean_token_accuracy": 0.8781389743089676,
+      "num_tokens": 85278037.0,
+      "step": 797
+    },
+    {
+      "epoch": 1.8187001140250856,
+      "grad_norm": 7.28125,
+      "learning_rate": 4.767491071337877e-06,
+      "loss": 0.6095,
+      "mean_token_accuracy": 0.8743828237056732,
+      "num_tokens": 85385317.0,
+      "step": 798
+    },
+    {
+      "epoch": 1.8209806157354618,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.766697235456761e-06,
+      "loss": 0.5999,
+      "mean_token_accuracy": 0.8782155364751816,
+      "num_tokens": 85493351.0,
+      "step": 799
+    },
+    {
+      "epoch": 1.823261117445838,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.765902113042053e-06,
+      "loss": 0.6179,
+      "mean_token_accuracy": 0.8723445981740952,
+      "num_tokens": 85599811.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.8255416191562144,
+      "grad_norm": 4.875,
+      "learning_rate": 4.765105704545052e-06,
+      "loss": 0.5962,
+      "mean_token_accuracy": 0.8751823008060455,
+      "num_tokens": 85706831.0,
+      "step": 801
+    },
+    {
+      "epoch": 1.8278221208665908,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.7643080104177815e-06,
+      "loss": 0.611,
+      "mean_token_accuracy": 0.8715473413467407,
+      "num_tokens": 85813856.0,
+      "step": 802
+    },
+    {
+      "epoch": 1.830102622576967,
+      "grad_norm": 4.125,
+      "learning_rate": 4.763509031113e-06,
+      "loss": 0.6133,
+      "mean_token_accuracy": 0.8734237253665924,
+      "num_tokens": 85921277.0,
+      "step": 803
+    },
+    {
+      "epoch": 1.8323831242873432,
+      "grad_norm": 2.5,
+      "learning_rate": 4.7627087670841894e-06,
+      "loss": 0.6146,
+      "mean_token_accuracy": 0.8760107606649399,
+      "num_tokens": 86028244.0,
+      "step": 804
+    },
+    {
+      "epoch": 1.8346636259977194,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.761907218785566e-06,
+      "loss": 0.6267,
+      "mean_token_accuracy": 0.8721684813499451,
+      "num_tokens": 86134885.0,
+      "step": 805
+    },
+    {
+      "epoch": 1.8369441277080958,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.761104386672074e-06,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.8805957436561584,
+      "num_tokens": 86242405.0,
+      "step": 806
+    },
+    {
+      "epoch": 1.8392246294184722,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.760300271199384e-06,
+      "loss": 0.6013,
+      "mean_token_accuracy": 0.875573918223381,
+      "num_tokens": 86349613.0,
+      "step": 807
+    },
+    {
+      "epoch": 1.8415051311288484,
+      "grad_norm": 2.375,
+      "learning_rate": 4.759494872823896e-06,
+      "loss": 0.6257,
+      "mean_token_accuracy": 0.8726819604635239,
+      "num_tokens": 86456544.0,
+      "step": 808
+    },
+    {
+      "epoch": 1.8437856328392246,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.758688192002741e-06,
+      "loss": 0.6027,
+      "mean_token_accuracy": 0.8799534738063812,
+      "num_tokens": 86563859.0,
+      "step": 809
+    },
+    {
+      "epoch": 1.8460661345496008,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.757880229193773e-06,
+      "loss": 0.599,
+      "mean_token_accuracy": 0.8771905303001404,
+      "num_tokens": 86670639.0,
+      "step": 810
+    },
+    {
+      "epoch": 1.8483466362599772,
+      "grad_norm": 2.125,
+      "learning_rate": 4.757070984855577e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8817077875137329,
+      "num_tokens": 86777613.0,
+      "step": 811
+    },
+    {
+      "epoch": 1.8506271379703536,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.756260459447465e-06,
+      "loss": 0.6173,
+      "mean_token_accuracy": 0.8729428201913834,
+      "num_tokens": 86884393.0,
+      "step": 812
+    },
+    {
+      "epoch": 1.8529076396807298,
+      "grad_norm": 3.0,
+      "learning_rate": 4.755448653429475e-06,
+      "loss": 0.6015,
+      "mean_token_accuracy": 0.877841129899025,
+      "num_tokens": 86991781.0,
+      "step": 813
+    },
+    {
+      "epoch": 1.855188141391106,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.754635567262372e-06,
+      "loss": 0.5992,
+      "mean_token_accuracy": 0.8776097446680069,
+      "num_tokens": 87098943.0,
+      "step": 814
+    },
+    {
+      "epoch": 1.8574686431014822,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.753821201407648e-06,
+      "loss": 0.603,
+      "mean_token_accuracy": 0.8773338198661804,
+      "num_tokens": 87205996.0,
+      "step": 815
+    },
+    {
+      "epoch": 1.8597491448118586,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.7530055563275225e-06,
+      "loss": 0.6032,
+      "mean_token_accuracy": 0.8762088716030121,
+      "num_tokens": 87313545.0,
+      "step": 816
+    },
+    {
+      "epoch": 1.862029646522235,
+      "grad_norm": 2.75,
+      "learning_rate": 4.7521886324849385e-06,
+      "loss": 0.626,
+      "mean_token_accuracy": 0.8743972480297089,
+      "num_tokens": 87420224.0,
+      "step": 817
+    },
+    {
+      "epoch": 1.8643101482326112,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.751370430343568e-06,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.8746133893728256,
+      "num_tokens": 87527454.0,
+      "step": 818
+    },
+    {
+      "epoch": 1.8665906499429874,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.750550950367805e-06,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8807315528392792,
+      "num_tokens": 87634497.0,
+      "step": 819
+    },
+    {
+      "epoch": 1.8688711516533636,
+      "grad_norm": 2.25,
+      "learning_rate": 4.749730193022771e-06,
+      "loss": 0.6099,
+      "mean_token_accuracy": 0.8758885711431503,
+      "num_tokens": 87741318.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.87115165336374,
+      "grad_norm": 5.78125,
+      "learning_rate": 4.748908158774312e-06,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8823642432689667,
+      "num_tokens": 87848637.0,
+      "step": 821
+    },
+    {
+      "epoch": 1.8734321550741164,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.748084848089e-06,
+      "loss": 0.602,
+      "mean_token_accuracy": 0.876085102558136,
+      "num_tokens": 87955464.0,
+      "step": 822
+    },
+    {
+      "epoch": 1.8757126567844926,
+      "grad_norm": 2.0,
+      "learning_rate": 4.747260261434128e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8766376227140427,
+      "num_tokens": 88062827.0,
+      "step": 823
+    },
+    {
+      "epoch": 1.8779931584948688,
+      "grad_norm": 4.375,
+      "learning_rate": 4.7464343992777175e-06,
+      "loss": 0.6083,
+      "mean_token_accuracy": 0.8759739398956299,
+      "num_tokens": 88169682.0,
+      "step": 824
+    },
+    {
+      "epoch": 1.880273660205245,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.74560726208851e-06,
+      "loss": 0.6146,
+      "mean_token_accuracy": 0.8760395795106888,
+      "num_tokens": 88276642.0,
+      "step": 825
+    },
+    {
+      "epoch": 1.8825541619156214,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.744778850335974e-06,
+      "loss": 0.6127,
+      "mean_token_accuracy": 0.8747592866420746,
+      "num_tokens": 88383703.0,
+      "step": 826
+    },
+    {
+      "epoch": 1.8848346636259978,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.7439491644903e-06,
+      "loss": 0.6065,
+      "mean_token_accuracy": 0.8750014901161194,
+      "num_tokens": 88490924.0,
+      "step": 827
+    },
+    {
+      "epoch": 1.887115165336374,
+      "grad_norm": 5.5,
+      "learning_rate": 4.743118205022402e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8755096942186356,
+      "num_tokens": 88597962.0,
+      "step": 828
+    },
+    {
+      "epoch": 1.8893956670467502,
+      "grad_norm": 5.25,
+      "learning_rate": 4.742285972403915e-06,
+      "loss": 0.6042,
+      "mean_token_accuracy": 0.8763994574546814,
+      "num_tokens": 88705068.0,
+      "step": 829
+    },
+    {
+      "epoch": 1.8916761687571264,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.7414524671071995e-06,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8812403380870819,
+      "num_tokens": 88812440.0,
+      "step": 830
+    },
+    {
+      "epoch": 1.8939566704675028,
+      "grad_norm": 4.9375,
+      "learning_rate": 4.7406176896053356e-06,
+      "loss": 0.6033,
+      "mean_token_accuracy": 0.8769374042749405,
+      "num_tokens": 88919411.0,
+      "step": 831
+    },
+    {
+      "epoch": 1.8962371721778792,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.739781640372129e-06,
+      "loss": 0.6026,
+      "mean_token_accuracy": 0.8767253011465073,
+      "num_tokens": 89026619.0,
+      "step": 832
+    },
+    {
+      "epoch": 1.8985176738882554,
+      "grad_norm": 3.984375,
+      "learning_rate": 4.7389443198821035e-06,
+      "loss": 0.6013,
+      "mean_token_accuracy": 0.8763202428817749,
+      "num_tokens": 89133774.0,
+      "step": 833
+    },
+    {
+      "epoch": 1.9007981755986316,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.738105728610507e-06,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8780044764280319,
+      "num_tokens": 89240785.0,
+      "step": 834
+    },
+    {
+      "epoch": 1.9030786773090078,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.737265867033307e-06,
+      "loss": 0.6104,
+      "mean_token_accuracy": 0.876687616109848,
+      "num_tokens": 89347651.0,
+      "step": 835
+    },
+    {
+      "epoch": 1.9053591790193842,
+      "grad_norm": 6.125,
+      "learning_rate": 4.736424735627193e-06,
+      "loss": 0.5907,
+      "mean_token_accuracy": 0.8766518086194992,
+      "num_tokens": 89454086.0,
+      "step": 836
+    },
+    {
+      "epoch": 1.9076396807297606,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.735582334869575e-06,
+      "loss": 0.5983,
+      "mean_token_accuracy": 0.8772305101156235,
+      "num_tokens": 89561179.0,
+      "step": 837
+    },
+    {
+      "epoch": 1.9099201824401368,
+      "grad_norm": 2.875,
+      "learning_rate": 4.734738665238583e-06,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8851803094148636,
+      "num_tokens": 89668499.0,
+      "step": 838
+    },
+    {
+      "epoch": 1.912200684150513,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.733893727213068e-06,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8792501837015152,
+      "num_tokens": 89775580.0,
+      "step": 839
+    },
+    {
+      "epoch": 1.9144811858608894,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.7330475212726e-06,
+      "loss": 0.5953,
+      "mean_token_accuracy": 0.8790338486433029,
+      "num_tokens": 89883322.0,
+      "step": 840
+    },
+    {
+      "epoch": 1.9167616875712656,
+      "grad_norm": 4.78125,
+      "learning_rate": 4.73220004789747e-06,
+      "loss": 0.6213,
+      "mean_token_accuracy": 0.8717529475688934,
+      "num_tokens": 89989995.0,
+      "step": 841
+    },
+    {
+      "epoch": 1.919042189281642,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.7313513075686875e-06,
+      "loss": 0.6019,
+      "mean_token_accuracy": 0.8774217069149017,
+      "num_tokens": 90097055.0,
+      "step": 842
+    },
+    {
+      "epoch": 1.9213226909920182,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.73050130076798e-06,
+      "loss": 0.5949,
+      "mean_token_accuracy": 0.8793853372335434,
+      "num_tokens": 90204483.0,
+      "step": 843
+    },
+    {
+      "epoch": 1.9236031927023944,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.729650027977797e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8808252811431885,
+      "num_tokens": 90311830.0,
+      "step": 844
+    },
+    {
+      "epoch": 1.9258836944127709,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.728797489681302e-06,
+      "loss": 0.5869,
+      "mean_token_accuracy": 0.8766701221466064,
+      "num_tokens": 90418653.0,
+      "step": 845
+    },
+    {
+      "epoch": 1.928164196123147,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.7279436863623805e-06,
+      "loss": 0.5924,
+      "mean_token_accuracy": 0.880815863609314,
+      "num_tokens": 90526355.0,
+      "step": 846
+    },
+    {
+      "epoch": 1.9304446978335235,
+      "grad_norm": 2.625,
+      "learning_rate": 4.7270886185056355e-06,
+      "loss": 0.6098,
+      "mean_token_accuracy": 0.8776162266731262,
+      "num_tokens": 90633735.0,
+      "step": 847
+    },
+    {
+      "epoch": 1.9327251995438997,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.726232286596385e-06,
+      "loss": 0.6064,
+      "mean_token_accuracy": 0.8745153993368149,
+      "num_tokens": 90740939.0,
+      "step": 848
+    },
+    {
+      "epoch": 1.9350057012542758,
+      "grad_norm": 3.625,
+      "learning_rate": 4.725374691120669e-06,
+      "loss": 0.6194,
+      "mean_token_accuracy": 0.8727088868618011,
+      "num_tokens": 90847747.0,
+      "step": 849
+    },
+    {
+      "epoch": 1.9372862029646523,
+      "grad_norm": 4.5,
+      "learning_rate": 4.7245158325652396e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.878357782959938,
+      "num_tokens": 90954915.0,
+      "step": 850
+    },
+    {
+      "epoch": 1.9395667046750285,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.7236557114175705e-06,
+      "loss": 0.6183,
+      "mean_token_accuracy": 0.8722603023052216,
+      "num_tokens": 91061821.0,
+      "step": 851
+    },
+    {
+      "epoch": 1.9418472063854049,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.722794328165849e-06,
+      "loss": 0.6068,
+      "mean_token_accuracy": 0.8738327622413635,
+      "num_tokens": 91168909.0,
+      "step": 852
+    },
+    {
+      "epoch": 1.944127708095781,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.721931683298979e-06,
+      "loss": 0.597,
+      "mean_token_accuracy": 0.8767188042402267,
+      "num_tokens": 91275666.0,
+      "step": 853
+    },
+    {
+      "epoch": 1.9464082098061573,
+      "grad_norm": 6.75,
+      "learning_rate": 4.721067777306582e-06,
+      "loss": 0.6291,
+      "mean_token_accuracy": 0.8720938414335251,
+      "num_tokens": 91382425.0,
+      "step": 854
+    },
+    {
+      "epoch": 1.9486887115165337,
+      "grad_norm": 6.9375,
+      "learning_rate": 4.7202026106789935e-06,
+      "loss": 0.6055,
+      "mean_token_accuracy": 0.87422114610672,
+      "num_tokens": 91489317.0,
+      "step": 855
+    },
+    {
+      "epoch": 1.95096921322691,
+      "grad_norm": 4.875,
+      "learning_rate": 4.719336183907266e-06,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8765899688005447,
+      "num_tokens": 91596598.0,
+      "step": 856
+    },
+    {
+      "epoch": 1.9532497149372863,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.718468497483166e-06,
+      "loss": 0.5942,
+      "mean_token_accuracy": 0.8780688494443893,
+      "num_tokens": 91703876.0,
+      "step": 857
+    },
+    {
+      "epoch": 1.9555302166476625,
+      "grad_norm": 3.5,
+      "learning_rate": 4.717599551899177e-06,
+      "loss": 0.5909,
+      "mean_token_accuracy": 0.8795433938503265,
+      "num_tokens": 91810908.0,
+      "step": 858
+    },
+    {
+      "epoch": 1.9578107183580387,
+      "grad_norm": 6.5,
+      "learning_rate": 4.716729347648494e-06,
+      "loss": 0.6067,
+      "mean_token_accuracy": 0.8784368187189102,
+      "num_tokens": 91918098.0,
+      "step": 859
+    },
+    {
+      "epoch": 1.960091220068415,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.71585788522503e-06,
+      "loss": 0.6027,
+      "mean_token_accuracy": 0.8751647025346756,
+      "num_tokens": 92025449.0,
+      "step": 860
+    },
+    {
+      "epoch": 1.9623717217787915,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.7149851651234085e-06,
+      "loss": 0.6008,
+      "mean_token_accuracy": 0.8768515288829803,
+      "num_tokens": 92132495.0,
+      "step": 861
+    },
+    {
+      "epoch": 1.9646522234891677,
+      "grad_norm": 6.25,
+      "learning_rate": 4.714111187838969e-06,
+      "loss": 0.6026,
+      "mean_token_accuracy": 0.8760241121053696,
+      "num_tokens": 92240414.0,
+      "step": 862
+    },
+    {
+      "epoch": 1.9669327251995439,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.713235953867764e-06,
+      "loss": 0.6127,
+      "mean_token_accuracy": 0.8732226341962814,
+      "num_tokens": 92346560.0,
+      "step": 863
+    },
+    {
+      "epoch": 1.96921322690992,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.712359463706561e-06,
+      "loss": 0.6059,
+      "mean_token_accuracy": 0.8731984049081802,
+      "num_tokens": 92453339.0,
+      "step": 864
+    },
+    {
+      "epoch": 1.9714937286202965,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.711481717852837e-06,
+      "loss": 0.6093,
+      "mean_token_accuracy": 0.8733879327774048,
+      "num_tokens": 92560826.0,
+      "step": 865
+    },
+    {
+      "epoch": 1.973774230330673,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.710602716804784e-06,
+      "loss": 0.614,
+      "mean_token_accuracy": 0.8723070323467255,
+      "num_tokens": 92667923.0,
+      "step": 866
+    },
+    {
+      "epoch": 1.976054732041049,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.709722461061307e-06,
+      "loss": 0.6102,
+      "mean_token_accuracy": 0.8759769946336746,
+      "num_tokens": 92774822.0,
+      "step": 867
+    },
+    {
+      "epoch": 1.9783352337514253,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.70884095112202e-06,
+      "loss": 0.5922,
+      "mean_token_accuracy": 0.8767600506544113,
+      "num_tokens": 92881978.0,
+      "step": 868
+    },
+    {
+      "epoch": 1.9806157354618015,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.707958187487254e-06,
+      "loss": 0.6102,
+      "mean_token_accuracy": 0.8756605833768845,
+      "num_tokens": 92989352.0,
+      "step": 869
+    },
+    {
+      "epoch": 1.982896237172178,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.707074170658046e-06,
+      "loss": 0.6259,
+      "mean_token_accuracy": 0.8736278414726257,
+      "num_tokens": 93095509.0,
+      "step": 870
+    },
+    {
+      "epoch": 1.9851767388825543,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.706188901136148e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8808690756559372,
+      "num_tokens": 93203007.0,
+      "step": 871
+    },
+    {
+      "epoch": 1.9874572405929305,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.705302379424023e-06,
+      "loss": 0.6058,
+      "mean_token_accuracy": 0.8731043189764023,
+      "num_tokens": 93310443.0,
+      "step": 872
+    },
+    {
+      "epoch": 1.9897377423033067,
+      "grad_norm": 4.25,
+      "learning_rate": 4.704414606024842e-06,
+      "loss": 0.6241,
+      "mean_token_accuracy": 0.8732631206512451,
+      "num_tokens": 93416826.0,
+      "step": 873
+    },
+    {
+      "epoch": 1.9920182440136829,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.703525581442488e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8822022676467896,
+      "num_tokens": 93523746.0,
+      "step": 874
+    },
+    {
+      "epoch": 1.9942987457240593,
+      "grad_norm": 3.5,
+      "learning_rate": 4.702635306181554e-06,
+      "loss": 0.5902,
+      "mean_token_accuracy": 0.8761956989765167,
+      "num_tokens": 93630669.0,
+      "step": 875
+    },
+    {
+      "epoch": 1.9965792474344357,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.701743780747345e-06,
+      "loss": 0.6131,
+      "mean_token_accuracy": 0.8748619705438614,
+      "num_tokens": 93738175.0,
+      "step": 876
+    },
+    {
+      "epoch": 1.998859749144812,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.700851005645872e-06,
+      "loss": 0.6047,
+      "mean_token_accuracy": 0.8752595335245132,
+      "num_tokens": 93845319.0,
+      "step": 877
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 5.40625,
+      "learning_rate": 4.699956981383857e-06,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8865090310573578,
+      "num_tokens": 93884464.0,
+      "step": 878
+    },
+    {
+      "epoch": 2.002280501710376,
+      "grad_norm": 2.625,
+      "learning_rate": 4.699061708468732e-06,
+      "loss": 0.604,
+      "mean_token_accuracy": 0.8751733750104904,
+      "num_tokens": 93991225.0,
+      "step": 879
+    },
+    {
+      "epoch": 2.0045610034207524,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.698165187408635e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8786071985960007,
+      "num_tokens": 94098859.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.0045610034207524,
+      "eval_loss": 0.6056435108184814,
+      "eval_mean_token_accuracy": 0.8761785948231193,
+      "eval_num_tokens": 94098859.0,
+      "eval_runtime": 58.5991,
+      "eval_samples_per_second": 143.091,
+      "eval_steps_per_second": 4.488,
+      "step": 880
+    },
+    {
+      "epoch": 2.006841505131129,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.697267418712415e-06,
+      "loss": 0.5955,
+      "mean_token_accuracy": 0.8779609203338623,
+      "num_tokens": 94205931.0,
+      "step": 881
+    },
+    {
+      "epoch": 2.009122006841505,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.6963684028896285e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.877967357635498,
+      "num_tokens": 94313152.0,
+      "step": 882
+    },
+    {
+      "epoch": 2.0114025085518814,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.695468140450539e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.8791947513818741,
+      "num_tokens": 94420319.0,
+      "step": 883
+    },
+    {
+      "epoch": 2.0136830102622576,
+      "grad_norm": 4.875,
+      "learning_rate": 4.6945666319061166e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8798933327198029,
+      "num_tokens": 94527420.0,
+      "step": 884
+    },
+    {
+      "epoch": 2.015963511972634,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.6936638777680435e-06,
+      "loss": 0.5986,
+      "mean_token_accuracy": 0.8755367547273636,
+      "num_tokens": 94634228.0,
+      "step": 885
+    },
+    {
+      "epoch": 2.0182440136830104,
+      "grad_norm": 1.9140625,
+      "learning_rate": 4.6927598785487026e-06,
+      "loss": 0.6088,
+      "mean_token_accuracy": 0.873708963394165,
+      "num_tokens": 94740813.0,
+      "step": 886
+    },
+    {
+      "epoch": 2.0205245153933866,
+      "grad_norm": 3.25,
+      "learning_rate": 4.691854634761188e-06,
+      "loss": 0.5952,
+      "mean_token_accuracy": 0.8793173581361771,
+      "num_tokens": 94848566.0,
+      "step": 887
+    },
+    {
+      "epoch": 2.022805017103763,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.690948146919299e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8790719360113144,
+      "num_tokens": 94955659.0,
+      "step": 888
+    },
+    {
+      "epoch": 2.025085518814139,
+      "grad_norm": 4.53125,
+      "learning_rate": 4.690040415537538e-06,
+      "loss": 0.5906,
+      "mean_token_accuracy": 0.8785732984542847,
+      "num_tokens": 95062760.0,
+      "step": 889
+    },
+    {
+      "epoch": 2.027366020524515,
+      "grad_norm": 3.5,
+      "learning_rate": 4.689131441131119e-06,
+      "loss": 0.6134,
+      "mean_token_accuracy": 0.8747037500143051,
+      "num_tokens": 95170435.0,
+      "step": 890
+    },
+    {
+      "epoch": 2.029646522234892,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.6882212242159555e-06,
+      "loss": 0.5945,
+      "mean_token_accuracy": 0.8810647279024124,
+      "num_tokens": 95277421.0,
+      "step": 891
+    },
+    {
+      "epoch": 2.031927023945268,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.687309765308671e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.881627082824707,
+      "num_tokens": 95384868.0,
+      "step": 892
+    },
+    {
+      "epoch": 2.034207525655644,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.6863970649265914e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8783137798309326,
+      "num_tokens": 95492292.0,
+      "step": 893
+    },
+    {
+      "epoch": 2.0364880273660204,
+      "grad_norm": 5.78125,
+      "learning_rate": 4.685483123587748e-06,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8808389455080032,
+      "num_tokens": 95599389.0,
+      "step": 894
+    },
+    {
+      "epoch": 2.0387685290763966,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.684567941810876e-06,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.8855313509702682,
+      "num_tokens": 95706662.0,
+      "step": 895
+    },
+    {
+      "epoch": 2.0410490307867732,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.683651520115414e-06,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.8826097548007965,
+      "num_tokens": 95814265.0,
+      "step": 896
+    },
+    {
+      "epoch": 2.0433295324971494,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.682733859021508e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8777095377445221,
+      "num_tokens": 95921432.0,
+      "step": 897
+    },
+    {
+      "epoch": 2.0456100342075256,
+      "grad_norm": 5.71875,
+      "learning_rate": 4.681814959050002e-06,
+      "loss": 0.6054,
+      "mean_token_accuracy": 0.8748744130134583,
+      "num_tokens": 96029029.0,
+      "step": 898
+    },
+    {
+      "epoch": 2.047890535917902,
+      "grad_norm": 5.65625,
+      "learning_rate": 4.680894820722446e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8802212625741959,
+      "num_tokens": 96136531.0,
+      "step": 899
+    },
+    {
+      "epoch": 2.050171037628278,
+      "grad_norm": 6.875,
+      "learning_rate": 4.679973444561095e-06,
+      "loss": 0.6056,
+      "mean_token_accuracy": 0.873874619603157,
+      "num_tokens": 96243395.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.0524515393386547,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.679050831088902e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.8777193129062653,
+      "num_tokens": 96350083.0,
+      "step": 901
+    },
+    {
+      "epoch": 2.054732041049031,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.678126980829525e-06,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8807910680770874,
+      "num_tokens": 96457480.0,
+      "step": 902
+    },
+    {
+      "epoch": 2.057012542759407,
+      "grad_norm": 5.0,
+      "learning_rate": 4.677201894307325e-06,
+      "loss": 0.6114,
+      "mean_token_accuracy": 0.872836172580719,
+      "num_tokens": 96564282.0,
+      "step": 903
+    },
+    {
+      "epoch": 2.0592930444697832,
+      "grad_norm": 4.75,
+      "learning_rate": 4.676275572047362e-06,
+      "loss": 0.6123,
+      "mean_token_accuracy": 0.871756300330162,
+      "num_tokens": 96671356.0,
+      "step": 904
+    },
+    {
+      "epoch": 2.0615735461801594,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.675348014575399e-06,
+      "loss": 0.6142,
+      "mean_token_accuracy": 0.8757220953702927,
+      "num_tokens": 96778413.0,
+      "step": 905
+    },
+    {
+      "epoch": 2.063854047890536,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.674419222417899e-06,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.8768678158521652,
+      "num_tokens": 96885296.0,
+      "step": 906
+    },
+    {
+      "epoch": 2.0661345496009123,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.673489196102028e-06,
+      "loss": 0.6045,
+      "mean_token_accuracy": 0.874673992395401,
+      "num_tokens": 96992919.0,
+      "step": 907
+    },
+    {
+      "epoch": 2.0684150513112884,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.67255793615565e-06,
+      "loss": 0.6077,
+      "mean_token_accuracy": 0.8760568499565125,
+      "num_tokens": 97099795.0,
+      "step": 908
+    },
+    {
+      "epoch": 2.0706955530216646,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.67162544310733e-06,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.8787839561700821,
+      "num_tokens": 97206999.0,
+      "step": 909
+    },
+    {
+      "epoch": 2.072976054732041,
+      "grad_norm": 7.03125,
+      "learning_rate": 4.670691717486333e-06,
+      "loss": 0.5961,
+      "mean_token_accuracy": 0.8767089992761612,
+      "num_tokens": 97314451.0,
+      "step": 910
+    },
+    {
+      "epoch": 2.0752565564424175,
+      "grad_norm": 5.84375,
+      "learning_rate": 4.669756759822625e-06,
+      "loss": 0.6157,
+      "mean_token_accuracy": 0.8748338222503662,
+      "num_tokens": 97421703.0,
+      "step": 911
+    },
+    {
+      "epoch": 2.0775370581527937,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.668820570646868e-06,
+      "loss": 0.5903,
+      "mean_token_accuracy": 0.8794323354959488,
+      "num_tokens": 97528420.0,
+      "step": 912
+    },
+    {
+      "epoch": 2.07981755986317,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.667883150490427e-06,
+      "loss": 0.6027,
+      "mean_token_accuracy": 0.8755746185779572,
+      "num_tokens": 97634954.0,
+      "step": 913
+    },
+    {
+      "epoch": 2.082098061573546,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.666944499885361e-06,
+      "loss": 0.6083,
+      "mean_token_accuracy": 0.8736224919557571,
+      "num_tokens": 97742085.0,
+      "step": 914
+    },
+    {
+      "epoch": 2.0843785632839227,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.6660046193644315e-06,
+      "loss": 0.6008,
+      "mean_token_accuracy": 0.8764249086380005,
+      "num_tokens": 97849138.0,
+      "step": 915
+    },
+    {
+      "epoch": 2.086659064994299,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.665063509461098e-06,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.8795299530029297,
+      "num_tokens": 97956499.0,
+      "step": 916
+    },
+    {
+      "epoch": 2.088939566704675,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.664121170709512e-06,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8816528022289276,
+      "num_tokens": 98063712.0,
+      "step": 917
+    },
+    {
+      "epoch": 2.0912200684150513,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.663177603644532e-06,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.8774126917123795,
+      "num_tokens": 98170731.0,
+      "step": 918
+    },
+    {
+      "epoch": 2.0935005701254275,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.662232808801704e-06,
+      "loss": 0.6308,
+      "mean_token_accuracy": 0.8704886138439178,
+      "num_tokens": 98277880.0,
+      "step": 919
+    },
+    {
+      "epoch": 2.095781071835804,
+      "grad_norm": 4.0,
+      "learning_rate": 4.661286786717278e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8788218349218369,
+      "num_tokens": 98385335.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.0980615735461803,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.660339537928198e-06,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8791757225990295,
+      "num_tokens": 98492150.0,
+      "step": 921
+    },
+    {
+      "epoch": 2.1003420752565565,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.659391062972102e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8784543722867966,
+      "num_tokens": 98599554.0,
+      "step": 922
+    },
+    {
+      "epoch": 2.1026225769669327,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.658441362387328e-06,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.8808614313602448,
+      "num_tokens": 98706337.0,
+      "step": 923
+    },
+    {
+      "epoch": 2.104903078677309,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.657490436712907e-06,
+      "loss": 0.62,
+      "mean_token_accuracy": 0.8705125451087952,
+      "num_tokens": 98812884.0,
+      "step": 924
+    },
+    {
+      "epoch": 2.1071835803876855,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.6565382864885665e-06,
+      "loss": 0.5909,
+      "mean_token_accuracy": 0.8783458173274994,
+      "num_tokens": 98920205.0,
+      "step": 925
+    },
+    {
+      "epoch": 2.1094640820980617,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.655584912254727e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.8786965757608414,
+      "num_tokens": 99027392.0,
+      "step": 926
+    },
+    {
+      "epoch": 2.111744583808438,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.654630314552508e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8778886198997498,
+      "num_tokens": 99134550.0,
+      "step": 927
+    },
+    {
+      "epoch": 2.114025085518814,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.653674493923718e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8823033571243286,
+      "num_tokens": 99241929.0,
+      "step": 928
+    },
+    {
+      "epoch": 2.1163055872291903,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.652717450910864e-06,
+      "loss": 0.6113,
+      "mean_token_accuracy": 0.8728591203689575,
+      "num_tokens": 99349158.0,
+      "step": 929
+    },
+    {
+      "epoch": 2.118586088939567,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.651759186057144e-06,
+      "loss": 0.5993,
+      "mean_token_accuracy": 0.87718665599823,
+      "num_tokens": 99456113.0,
+      "step": 930
+    },
+    {
+      "epoch": 2.120866590649943,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.650799699906452e-06,
+      "loss": 0.6052,
+      "mean_token_accuracy": 0.8780839145183563,
+      "num_tokens": 99563254.0,
+      "step": 931
+    },
+    {
+      "epoch": 2.1231470923603193,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.649838993003373e-06,
+      "loss": 0.5993,
+      "mean_token_accuracy": 0.8814170211553574,
+      "num_tokens": 99670632.0,
+      "step": 932
+    },
+    {
+      "epoch": 2.1254275940706955,
+      "grad_norm": 4.625,
+      "learning_rate": 4.648877065893186e-06,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8802737295627594,
+      "num_tokens": 99777773.0,
+      "step": 933
+    },
+    {
+      "epoch": 2.1277080957810717,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.647913919121861e-06,
+      "loss": 0.5973,
+      "mean_token_accuracy": 0.8730404675006866,
+      "num_tokens": 99885179.0,
+      "step": 934
+    },
+    {
+      "epoch": 2.1299885974914483,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.646949553236064e-06,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.8850271999835968,
+      "num_tokens": 99992436.0,
+      "step": 935
+    },
+    {
+      "epoch": 2.1322690992018245,
+      "grad_norm": 4.75,
+      "learning_rate": 4.645983968783148e-06,
+      "loss": 0.5917,
+      "mean_token_accuracy": 0.8796833902597427,
+      "num_tokens": 100099422.0,
+      "step": 936
+    },
+    {
+      "epoch": 2.1345496009122007,
+      "grad_norm": 5.9375,
+      "learning_rate": 4.645017166311163e-06,
+      "loss": 0.596,
+      "mean_token_accuracy": 0.8761509507894516,
+      "num_tokens": 100207243.0,
+      "step": 937
+    },
+    {
+      "epoch": 2.136830102622577,
+      "grad_norm": 2.875,
+      "learning_rate": 4.644049146368844e-06,
+      "loss": 0.6185,
+      "mean_token_accuracy": 0.8749994188547134,
+      "num_tokens": 100313796.0,
+      "step": 938
+    },
+    {
+      "epoch": 2.139110604332953,
+      "grad_norm": 1.9609375,
+      "learning_rate": 4.643079909505622e-06,
+      "loss": 0.5775,
+      "mean_token_accuracy": 0.8829146772623062,
+      "num_tokens": 100420992.0,
+      "step": 939
+    },
+    {
+      "epoch": 2.1413911060433297,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.642109456271618e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8793570548295975,
+      "num_tokens": 100528158.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.143671607753706,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.64113778721764e-06,
+      "loss": 0.6007,
+      "mean_token_accuracy": 0.8770193010568619,
+      "num_tokens": 100635032.0,
+      "step": 941
+    },
+    {
+      "epoch": 2.145952109464082,
+      "grad_norm": 5.4375,
+      "learning_rate": 4.640164902895192e-06,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8772297501564026,
+      "num_tokens": 100742739.0,
+      "step": 942
+    },
+    {
+      "epoch": 2.1482326111744583,
+      "grad_norm": 2.5,
+      "learning_rate": 4.6391908038564615e-06,
+      "loss": 0.6109,
+      "mean_token_accuracy": 0.8756915777921677,
+      "num_tokens": 100849851.0,
+      "step": 943
+    },
+    {
+      "epoch": 2.1505131128848345,
+      "grad_norm": 6.5,
+      "learning_rate": 4.6382154906543295e-06,
+      "loss": 0.5989,
+      "mean_token_accuracy": 0.8732953071594238,
+      "num_tokens": 100957153.0,
+      "step": 944
+    },
+    {
+      "epoch": 2.152793614595211,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.637238963842365e-06,
+      "loss": 0.6082,
+      "mean_token_accuracy": 0.8774750828742981,
+      "num_tokens": 101063587.0,
+      "step": 945
+    },
+    {
+      "epoch": 2.1550741163055873,
+      "grad_norm": 3.96875,
+      "learning_rate": 4.636261223974826e-06,
+      "loss": 0.5955,
+      "mean_token_accuracy": 0.8747821748256683,
+      "num_tokens": 101170370.0,
+      "step": 946
+    },
+    {
+      "epoch": 2.1573546180159635,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.635282271606658e-06,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8776073157787323,
+      "num_tokens": 101277426.0,
+      "step": 947
+    },
+    {
+      "epoch": 2.1596351197263397,
+      "grad_norm": 2.5,
+      "learning_rate": 4.634302107293497e-06,
+      "loss": 0.5965,
+      "mean_token_accuracy": 0.876614436507225,
+      "num_tokens": 101383939.0,
+      "step": 948
+    },
+    {
+      "epoch": 2.161915621436716,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.633320731591663e-06,
+      "loss": 0.605,
+      "mean_token_accuracy": 0.8757250159978867,
+      "num_tokens": 101491379.0,
+      "step": 949
+    },
+    {
+      "epoch": 2.1641961231470925,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.632338145058167e-06,
+      "loss": 0.6244,
+      "mean_token_accuracy": 0.8718846142292023,
+      "num_tokens": 101598608.0,
+      "step": 950
+    },
+    {
+      "epoch": 2.1664766248574687,
+      "grad_norm": 1.96875,
+      "learning_rate": 4.631354348250706e-06,
+      "loss": 0.5809,
+      "mean_token_accuracy": 0.8830228000879288,
+      "num_tokens": 101706020.0,
+      "step": 951
+    },
+    {
+      "epoch": 2.168757126567845,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.630369341727665e-06,
+      "loss": 0.6116,
+      "mean_token_accuracy": 0.872904360294342,
+      "num_tokens": 101812698.0,
+      "step": 952
+    },
+    {
+      "epoch": 2.171037628278221,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.629383126048114e-06,
+      "loss": 0.608,
+      "mean_token_accuracy": 0.8747462183237076,
+      "num_tokens": 101919836.0,
+      "step": 953
+    },
+    {
+      "epoch": 2.1733181299885973,
+      "grad_norm": 4.875,
+      "learning_rate": 4.6283957017718105e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8803711831569672,
+      "num_tokens": 102027244.0,
+      "step": 954
+    },
+    {
+      "epoch": 2.175598631698974,
+      "grad_norm": 1.96875,
+      "learning_rate": 4.627407069459196e-06,
+      "loss": 0.5976,
+      "mean_token_accuracy": 0.8786141574382782,
+      "num_tokens": 102134050.0,
+      "step": 955
+    },
+    {
+      "epoch": 2.17787913340935,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.626417229671401e-06,
+      "loss": 0.586,
+      "mean_token_accuracy": 0.8817842751741409,
+      "num_tokens": 102241872.0,
+      "step": 956
+    },
+    {
+      "epoch": 2.1801596351197263,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.625426182970237e-06,
+      "loss": 0.6055,
+      "mean_token_accuracy": 0.8747357130050659,
+      "num_tokens": 102348674.0,
+      "step": 957
+    },
+    {
+      "epoch": 2.1824401368301025,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.6244339299182065e-06,
+      "loss": 0.6021,
+      "mean_token_accuracy": 0.8756296783685684,
+      "num_tokens": 102455493.0,
+      "step": 958
+    },
+    {
+      "epoch": 2.1847206385404787,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.62344047107849e-06,
+      "loss": 0.5923,
+      "mean_token_accuracy": 0.8777966946363449,
+      "num_tokens": 102562617.0,
+      "step": 959
+    },
+    {
+      "epoch": 2.1870011402508553,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.622445807014956e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8824858963489532,
+      "num_tokens": 102670281.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.1892816419612315,
+      "grad_norm": 3.125,
+      "learning_rate": 4.621449938292159e-06,
+      "loss": 0.603,
+      "mean_token_accuracy": 0.8756995797157288,
+      "num_tokens": 102778053.0,
+      "step": 961
+    },
+    {
+      "epoch": 2.1915621436716077,
+      "grad_norm": 4.46875,
+      "learning_rate": 4.620452865475331e-06,
+      "loss": 0.5975,
+      "mean_token_accuracy": 0.8776930570602417,
+      "num_tokens": 102885553.0,
+      "step": 962
+    },
+    {
+      "epoch": 2.193842645381984,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.6194545891303955e-06,
+      "loss": 0.6116,
+      "mean_token_accuracy": 0.8760574162006378,
+      "num_tokens": 102992589.0,
+      "step": 963
+    },
+    {
+      "epoch": 2.19612314709236,
+      "grad_norm": 6.25,
+      "learning_rate": 4.618455109823952e-06,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.87519970536232,
+      "num_tokens": 103099395.0,
+      "step": 964
+    },
+    {
+      "epoch": 2.1984036488027368,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.617454428123287e-06,
+      "loss": 0.6091,
+      "mean_token_accuracy": 0.8743910491466522,
+      "num_tokens": 103205910.0,
+      "step": 965
+    },
+    {
+      "epoch": 2.200684150513113,
+      "grad_norm": 5.21875,
+      "learning_rate": 4.616452544596367e-06,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.8808601945638657,
+      "num_tokens": 103313312.0,
+      "step": 966
+    },
+    {
+      "epoch": 2.202964652223489,
+      "grad_norm": 2.25,
+      "learning_rate": 4.615449459811843e-06,
+      "loss": 0.587,
+      "mean_token_accuracy": 0.882127583026886,
+      "num_tokens": 103420313.0,
+      "step": 967
+    },
+    {
+      "epoch": 2.2052451539338653,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.614445174339045e-06,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8822465240955353,
+      "num_tokens": 103527336.0,
+      "step": 968
+    },
+    {
+      "epoch": 2.2075256556442415,
+      "grad_norm": 6.46875,
+      "learning_rate": 4.613439688747988e-06,
+      "loss": 0.5882,
+      "mean_token_accuracy": 0.8785429149866104,
+      "num_tokens": 103634750.0,
+      "step": 969
+    },
+    {
+      "epoch": 2.209806157354618,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.612433003609365e-06,
+      "loss": 0.6146,
+      "mean_token_accuracy": 0.8751912713050842,
+      "num_tokens": 103741472.0,
+      "step": 970
+    },
+    {
+      "epoch": 2.2120866590649944,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.611425119494552e-06,
+      "loss": 0.6132,
+      "mean_token_accuracy": 0.8733502626419067,
+      "num_tokens": 103848855.0,
+      "step": 971
+    },
+    {
+      "epoch": 2.2143671607753705,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.6104160369756025e-06,
+      "loss": 0.5832,
+      "mean_token_accuracy": 0.8768988847732544,
+      "num_tokens": 103956579.0,
+      "step": 972
+    },
+    {
+      "epoch": 2.2166476624857467,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.609405756625254e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8784401416778564,
+      "num_tokens": 104063842.0,
+      "step": 973
+    },
+    {
+      "epoch": 2.2189281641961234,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.608394279016921e-06,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8794825226068497,
+      "num_tokens": 104171052.0,
+      "step": 974
+    },
+    {
+      "epoch": 2.2212086659064996,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.6073816047247e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8819704353809357,
+      "num_tokens": 104278607.0,
+      "step": 975
+    },
+    {
+      "epoch": 2.2234891676168758,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.606367734323365e-06,
+      "loss": 0.5971,
+      "mean_token_accuracy": 0.8778064996004105,
+      "num_tokens": 104385870.0,
+      "step": 976
+    },
+    {
+      "epoch": 2.225769669327252,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.605352668388369e-06,
+      "loss": 0.6163,
+      "mean_token_accuracy": 0.8713217675685883,
+      "num_tokens": 104493565.0,
+      "step": 977
+    },
+    {
+      "epoch": 2.228050171037628,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.6043364074958435e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8773491680622101,
+      "num_tokens": 104600945.0,
+      "step": 978
+    },
+    {
+      "epoch": 2.2303306727480043,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.6033189522226e-06,
+      "loss": 0.6037,
+      "mean_token_accuracy": 0.8754960596561432,
+      "num_tokens": 104707727.0,
+      "step": 979
+    },
+    {
+      "epoch": 2.232611174458381,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.602300303146123e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8821861445903778,
+      "num_tokens": 104815334.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.234891676168757,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.601280460844583e-06,
+      "loss": 0.5935,
+      "mean_token_accuracy": 0.8821869492530823,
+      "num_tokens": 104922027.0,
+      "step": 981
+    },
+    {
+      "epoch": 2.2371721778791334,
+      "grad_norm": 3.125,
+      "learning_rate": 4.6002594258968185e-06,
+      "loss": 0.5679,
+      "mean_token_accuracy": 0.8820158392190933,
+      "num_tokens": 105028843.0,
+      "step": 982
+    },
+    {
+      "epoch": 2.2394526795895096,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.599237198882351e-06,
+      "loss": 0.6006,
+      "mean_token_accuracy": 0.8787632882595062,
+      "num_tokens": 105135688.0,
+      "step": 983
+    },
+    {
+      "epoch": 2.241733181299886,
+      "grad_norm": 2.875,
+      "learning_rate": 4.598213780381377e-06,
+      "loss": 0.5986,
+      "mean_token_accuracy": 0.8801819086074829,
+      "num_tokens": 105242376.0,
+      "step": 984
+    },
+    {
+      "epoch": 2.2440136830102624,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.59718917097477e-06,
+      "loss": 0.5932,
+      "mean_token_accuracy": 0.8789917528629303,
+      "num_tokens": 105349325.0,
+      "step": 985
+    },
+    {
+      "epoch": 2.2462941847206386,
+      "grad_norm": 5.8125,
+      "learning_rate": 4.596163371244076e-06,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8811417818069458,
+      "num_tokens": 105456213.0,
+      "step": 986
+    },
+    {
+      "epoch": 2.2485746864310148,
+      "grad_norm": 7.59375,
+      "learning_rate": 4.595136381771521e-06,
+      "loss": 0.5982,
+      "mean_token_accuracy": 0.8762593269348145,
+      "num_tokens": 105562679.0,
+      "step": 987
+    },
+    {
+      "epoch": 2.250855188141391,
+      "grad_norm": 7.8125,
+      "learning_rate": 4.594108203140004e-06,
+      "loss": 0.6039,
+      "mean_token_accuracy": 0.8737095445394516,
+      "num_tokens": 105670038.0,
+      "step": 988
+    },
+    {
+      "epoch": 2.253135689851767,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.593078835933099e-06,
+      "loss": 0.5871,
+      "mean_token_accuracy": 0.8798780590295792,
+      "num_tokens": 105777357.0,
+      "step": 989
+    },
+    {
+      "epoch": 2.255416191562144,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.592048280735055e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.879479706287384,
+      "num_tokens": 105884657.0,
+      "step": 990
+    },
+    {
+      "epoch": 2.25769669327252,
+      "grad_norm": 5.71875,
+      "learning_rate": 4.591016538130796e-06,
+      "loss": 0.5854,
+      "mean_token_accuracy": 0.8783192932605743,
+      "num_tokens": 105991145.0,
+      "step": 991
+    },
+    {
+      "epoch": 2.259977194982896,
+      "grad_norm": 5.59375,
+      "learning_rate": 4.589983608705918e-06,
+      "loss": 0.6045,
+      "mean_token_accuracy": 0.8753475993871689,
+      "num_tokens": 106097808.0,
+      "step": 992
+    },
+    {
+      "epoch": 2.2622576966932724,
+      "grad_norm": 5.78125,
+      "learning_rate": 4.588949493046693e-06,
+      "loss": 0.628,
+      "mean_token_accuracy": 0.8707805275917053,
+      "num_tokens": 106204975.0,
+      "step": 993
+    },
+    {
+      "epoch": 2.264538198403649,
+      "grad_norm": 4.375,
+      "learning_rate": 4.587914191740064e-06,
+      "loss": 0.6265,
+      "mean_token_accuracy": 0.8718110471963882,
+      "num_tokens": 106311506.0,
+      "step": 994
+    },
+    {
+      "epoch": 2.266818700114025,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.586877705373648e-06,
+      "loss": 0.5866,
+      "mean_token_accuracy": 0.8814664483070374,
+      "num_tokens": 106418583.0,
+      "step": 995
+    },
+    {
+      "epoch": 2.2690992018244014,
+      "grad_norm": 3.625,
+      "learning_rate": 4.585840034535736e-06,
+      "loss": 0.5829,
+      "mean_token_accuracy": 0.881786435842514,
+      "num_tokens": 106526615.0,
+      "step": 996
+    },
+    {
+      "epoch": 2.2713797035347776,
+      "grad_norm": 7.40625,
+      "learning_rate": 4.584801179815289e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.878081277012825,
+      "num_tokens": 106633800.0,
+      "step": 997
+    },
+    {
+      "epoch": 2.2736602052451538,
+      "grad_norm": 7.59375,
+      "learning_rate": 4.583761141801941e-06,
+      "loss": 0.6034,
+      "mean_token_accuracy": 0.8735693991184235,
+      "num_tokens": 106740741.0,
+      "step": 998
+    },
+    {
+      "epoch": 2.27594070695553,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.5827199210859975e-06,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8779504746198654,
+      "num_tokens": 106847997.0,
+      "step": 999
+    },
+    {
+      "epoch": 2.2782212086659066,
+      "grad_norm": 5.4375,
+      "learning_rate": 4.581677518258435e-06,
+      "loss": 0.6154,
+      "mean_token_accuracy": 0.8708099573850632,
+      "num_tokens": 106955093.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.280501710376283,
+      "grad_norm": 5.25,
+      "learning_rate": 4.580633933910901e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.881976768374443,
+      "num_tokens": 107062229.0,
+      "step": 1001
+    },
+    {
+      "epoch": 2.282782212086659,
+      "grad_norm": 5.5625,
+      "learning_rate": 4.579589168635715e-06,
+      "loss": 0.6241,
+      "mean_token_accuracy": 0.8718415945768356,
+      "num_tokens": 107168673.0,
+      "step": 1002
+    },
+    {
+      "epoch": 2.285062713797035,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.578543223025865e-06,
+      "loss": 0.6294,
+      "mean_token_accuracy": 0.8693426996469498,
+      "num_tokens": 107275384.0,
+      "step": 1003
+    },
+    {
+      "epoch": 2.287343215507412,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.577496097675009e-06,
+      "loss": 0.5871,
+      "mean_token_accuracy": 0.8758938610553741,
+      "num_tokens": 107382155.0,
+      "step": 1004
+    },
+    {
+      "epoch": 2.289623717217788,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.576447793177476e-06,
+      "loss": 0.6103,
+      "mean_token_accuracy": 0.8743392378091812,
+      "num_tokens": 107488845.0,
+      "step": 1005
+    },
+    {
+      "epoch": 2.291904218928164,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.575398310128263e-06,
+      "loss": 0.5773,
+      "mean_token_accuracy": 0.8819352090358734,
+      "num_tokens": 107596244.0,
+      "step": 1006
+    },
+    {
+      "epoch": 2.2941847206385404,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.574347649123036e-06,
+      "loss": 0.6106,
+      "mean_token_accuracy": 0.8723960518836975,
+      "num_tokens": 107703259.0,
+      "step": 1007
+    },
+    {
+      "epoch": 2.2964652223489166,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.57329581075813e-06,
+      "loss": 0.626,
+      "mean_token_accuracy": 0.8719481080770493,
+      "num_tokens": 107810343.0,
+      "step": 1008
+    },
+    {
+      "epoch": 2.2987457240592932,
+      "grad_norm": 5.1875,
+      "learning_rate": 4.572242795630549e-06,
+      "loss": 0.5856,
+      "mean_token_accuracy": 0.8785750865936279,
+      "num_tokens": 107917748.0,
+      "step": 1009
+    },
+    {
+      "epoch": 2.3010262257696694,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.571188604337963e-06,
+      "loss": 0.5905,
+      "mean_token_accuracy": 0.8774587363004684,
+      "num_tokens": 108025056.0,
+      "step": 1010
+    },
+    {
+      "epoch": 2.3033067274800456,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.570133237478711e-06,
+      "loss": 0.6028,
+      "mean_token_accuracy": 0.8772729784250259,
+      "num_tokens": 108131874.0,
+      "step": 1011
+    },
+    {
+      "epoch": 2.305587229190422,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.5690766956517985e-06,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.883285716176033,
+      "num_tokens": 108238676.0,
+      "step": 1012
+    },
+    {
+      "epoch": 2.307867730900798,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.568018979456899e-06,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.87696273624897,
+      "num_tokens": 108345907.0,
+      "step": 1013
+    },
+    {
+      "epoch": 2.3101482326111746,
+      "grad_norm": 5.46875,
+      "learning_rate": 4.566960089494351e-06,
+      "loss": 0.6027,
+      "mean_token_accuracy": 0.8753390312194824,
+      "num_tokens": 108453246.0,
+      "step": 1014
+    },
+    {
+      "epoch": 2.312428734321551,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.5659000263651615e-06,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8795862644910812,
+      "num_tokens": 108560344.0,
+      "step": 1015
+    },
+    {
+      "epoch": 2.314709236031927,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.564838790671e-06,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.8801321387290955,
+      "num_tokens": 108667748.0,
+      "step": 1016
+    },
+    {
+      "epoch": 2.316989737742303,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.5637763830142046e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8832177221775055,
+      "num_tokens": 108775087.0,
+      "step": 1017
+    },
+    {
+      "epoch": 2.3192702394526794,
+      "grad_norm": 4.96875,
+      "learning_rate": 4.562712803997776e-06,
+      "loss": 0.6119,
+      "mean_token_accuracy": 0.8724011331796646,
+      "num_tokens": 108882040.0,
+      "step": 1018
+    },
+    {
+      "epoch": 2.321550741163056,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.5616480542253825e-06,
+      "loss": 0.5942,
+      "mean_token_accuracy": 0.8780981302261353,
+      "num_tokens": 108988642.0,
+      "step": 1019
+    },
+    {
+      "epoch": 2.3238312428734322,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.5605821343013555e-06,
+      "loss": 0.61,
+      "mean_token_accuracy": 0.8755066245794296,
+      "num_tokens": 109096022.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.3261117445838084,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.55951504483069e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.878739207983017,
+      "num_tokens": 109203309.0,
+      "step": 1021
+    },
+    {
+      "epoch": 2.3283922462941846,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.558446786419045e-06,
+      "loss": 0.599,
+      "mean_token_accuracy": 0.8794536143541336,
+      "num_tokens": 109309812.0,
+      "step": 1022
+    },
+    {
+      "epoch": 2.330672748004561,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.557377359672745e-06,
+      "loss": 0.5817,
+      "mean_token_accuracy": 0.8791648596525192,
+      "num_tokens": 109416293.0,
+      "step": 1023
+    },
+    {
+      "epoch": 2.3329532497149374,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.556306765198775e-06,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.8814702332019806,
+      "num_tokens": 109523500.0,
+      "step": 1024
+    },
+    {
+      "epoch": 2.3352337514253136,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.555235003604782e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.8794472515583038,
+      "num_tokens": 109630312.0,
+      "step": 1025
+    },
+    {
+      "epoch": 2.33751425313569,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.55416207549908e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8786284774541855,
+      "num_tokens": 109737301.0,
+      "step": 1026
+    },
+    {
+      "epoch": 2.339794754846066,
+      "grad_norm": 4.625,
+      "learning_rate": 4.5530879814906404e-06,
+      "loss": 0.6159,
+      "mean_token_accuracy": 0.8746591061353683,
+      "num_tokens": 109844193.0,
+      "step": 1027
+    },
+    {
+      "epoch": 2.342075256556442,
+      "grad_norm": 2.5,
+      "learning_rate": 4.5520127221891e-06,
+      "loss": 0.6225,
+      "mean_token_accuracy": 0.8710612952709198,
+      "num_tokens": 109951493.0,
+      "step": 1028
+    },
+    {
+      "epoch": 2.344355758266819,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.5509362982047525e-06,
+      "loss": 0.5982,
+      "mean_token_accuracy": 0.8757833987474442,
+      "num_tokens": 110058293.0,
+      "step": 1029
+    },
+    {
+      "epoch": 2.346636259977195,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.549858710148558e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.876603439450264,
+      "num_tokens": 110165755.0,
+      "step": 1030
+    },
+    {
+      "epoch": 2.3489167616875712,
+      "grad_norm": 4.125,
+      "learning_rate": 4.548779958632134e-06,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.8827737718820572,
+      "num_tokens": 110273020.0,
+      "step": 1031
+    },
+    {
+      "epoch": 2.3511972633979474,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.5477000442677575e-06,
+      "loss": 0.6005,
+      "mean_token_accuracy": 0.8793087154626846,
+      "num_tokens": 110379923.0,
+      "step": 1032
+    },
+    {
+      "epoch": 2.353477765108324,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.546618967668369e-06,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8808025866746902,
+      "num_tokens": 110487050.0,
+      "step": 1033
+    },
+    {
+      "epoch": 2.3557582668187003,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.545536729447566e-06,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8820274025201797,
+      "num_tokens": 110594620.0,
+      "step": 1034
+    },
+    {
+      "epoch": 2.3580387685290765,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.544453330219606e-06,
+      "loss": 0.5961,
+      "mean_token_accuracy": 0.8768385797739029,
+      "num_tokens": 110701829.0,
+      "step": 1035
+    },
+    {
+      "epoch": 2.3603192702394526,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.543368770599406e-06,
+      "loss": 0.5974,
+      "mean_token_accuracy": 0.8801583051681519,
+      "num_tokens": 110809252.0,
+      "step": 1036
+    },
+    {
+      "epoch": 2.362599771949829,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.542283051202539e-06,
+      "loss": 0.5896,
+      "mean_token_accuracy": 0.8764929324388504,
+      "num_tokens": 110916093.0,
+      "step": 1037
+    },
+    {
+      "epoch": 2.364880273660205,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.541196172645242e-06,
+      "loss": 0.5974,
+      "mean_token_accuracy": 0.8785767704248428,
+      "num_tokens": 111023473.0,
+      "step": 1038
+    },
+    {
+      "epoch": 2.3671607753705817,
+      "grad_norm": 6.9375,
+      "learning_rate": 4.540108135544403e-06,
+      "loss": 0.5914,
+      "mean_token_accuracy": 0.8778847008943558,
+      "num_tokens": 111130747.0,
+      "step": 1039
+    },
+    {
+      "epoch": 2.369441277080958,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.5390189405175725e-06,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8778575360774994,
+      "num_tokens": 111237498.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.371721778791334,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.537928588182955e-06,
+      "loss": 0.6031,
+      "mean_token_accuracy": 0.8711429536342621,
+      "num_tokens": 111344543.0,
+      "step": 1041
+    },
+    {
+      "epoch": 2.3740022805017102,
+      "grad_norm": 3.125,
+      "learning_rate": 4.536837079159416e-06,
+      "loss": 0.5712,
+      "mean_token_accuracy": 0.8806983381509781,
+      "num_tokens": 111451314.0,
+      "step": 1042
+    },
+    {
+      "epoch": 2.376282782212087,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.535744414066473e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8755781650543213,
+      "num_tokens": 111558317.0,
+      "step": 1043
+    },
+    {
+      "epoch": 2.378563283922463,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.534650593524302e-06,
+      "loss": 0.6024,
+      "mean_token_accuracy": 0.874798059463501,
+      "num_tokens": 111664986.0,
+      "step": 1044
+    },
+    {
+      "epoch": 2.3808437856328393,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.533555618153735e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8804291188716888,
+      "num_tokens": 111772027.0,
+      "step": 1045
+    },
+    {
+      "epoch": 2.3831242873432155,
+      "grad_norm": 2.5,
+      "learning_rate": 4.532459488576258e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8797705769538879,
+      "num_tokens": 111878671.0,
+      "step": 1046
+    },
+    {
+      "epoch": 2.3854047890535917,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.531362205414013e-06,
+      "loss": 0.59,
+      "mean_token_accuracy": 0.8787082433700562,
+      "num_tokens": 111985253.0,
+      "step": 1047
+    },
+    {
+      "epoch": 2.387685290763968,
+      "grad_norm": 3.5,
+      "learning_rate": 4.530263769289798e-06,
+      "loss": 0.5874,
+      "mean_token_accuracy": 0.8787430375814438,
+      "num_tokens": 112092830.0,
+      "step": 1048
+    },
+    {
+      "epoch": 2.3899657924743445,
+      "grad_norm": 4.625,
+      "learning_rate": 4.529164180827063e-06,
+      "loss": 0.6002,
+      "mean_token_accuracy": 0.8749395608901978,
+      "num_tokens": 112199272.0,
+      "step": 1049
+    },
+    {
+      "epoch": 2.3922462941847207,
+      "grad_norm": 5.875,
+      "learning_rate": 4.528063440649913e-06,
+      "loss": 0.5932,
+      "mean_token_accuracy": 0.8786779493093491,
+      "num_tokens": 112306480.0,
+      "step": 1050
+    },
+    {
+      "epoch": 2.394526795895097,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.526961549383109e-06,
+      "loss": 0.5999,
+      "mean_token_accuracy": 0.8784288913011551,
+      "num_tokens": 112413379.0,
+      "step": 1051
+    },
+    {
+      "epoch": 2.396807297605473,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.52585850765206e-06,
+      "loss": 0.6035,
+      "mean_token_accuracy": 0.8735271543264389,
+      "num_tokens": 112520374.0,
+      "step": 1052
+    },
+    {
+      "epoch": 2.3990877993158497,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.524754316082833e-06,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8810184299945831,
+      "num_tokens": 112627321.0,
+      "step": 1053
+    },
+    {
+      "epoch": 2.401368301026226,
+      "grad_norm": 6.4375,
+      "learning_rate": 4.5236489753021465e-06,
+      "loss": 0.6087,
+      "mean_token_accuracy": 0.8768182098865509,
+      "num_tokens": 112734136.0,
+      "step": 1054
+    },
+    {
+      "epoch": 2.403648802736602,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.522542485937369e-06,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8769848495721817,
+      "num_tokens": 112841135.0,
+      "step": 1055
+    },
+    {
+      "epoch": 2.4059293044469783,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.521434848616523e-06,
+      "loss": 0.5983,
+      "mean_token_accuracy": 0.8787764012813568,
+      "num_tokens": 112948501.0,
+      "step": 1056
+    },
+    {
+      "epoch": 2.4082098061573545,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.520326063968283e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8776495456695557,
+      "num_tokens": 113055115.0,
+      "step": 1057
+    },
+    {
+      "epoch": 2.4104903078677307,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.5192161326219716e-06,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8828926682472229,
+      "num_tokens": 113162261.0,
+      "step": 1058
+    },
+    {
+      "epoch": 2.4127708095781073,
+      "grad_norm": 2.03125,
+      "learning_rate": 4.5181050552075665e-06,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.8777969181537628,
+      "num_tokens": 113269282.0,
+      "step": 1059
+    },
+    {
+      "epoch": 2.4150513112884835,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.516992832355694e-06,
+      "loss": 0.5973,
+      "mean_token_accuracy": 0.8747198283672333,
+      "num_tokens": 113375649.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.4173318129988597,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.515879464697629e-06,
+      "loss": 0.5993,
+      "mean_token_accuracy": 0.8800251632928848,
+      "num_tokens": 113483268.0,
+      "step": 1061
+    },
+    {
+      "epoch": 2.419612314709236,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.514764952865297e-06,
+      "loss": 0.6132,
+      "mean_token_accuracy": 0.8789681494235992,
+      "num_tokens": 113590044.0,
+      "step": 1062
+    },
+    {
+      "epoch": 2.4218928164196125,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.513649297491275e-06,
+      "loss": 0.6064,
+      "mean_token_accuracy": 0.8741051852703094,
+      "num_tokens": 113696761.0,
+      "step": 1063
+    },
+    {
+      "epoch": 2.4241733181299887,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.512532499208787e-06,
+      "loss": 0.606,
+      "mean_token_accuracy": 0.8762391060590744,
+      "num_tokens": 113804097.0,
+      "step": 1064
+    },
+    {
+      "epoch": 2.426453819840365,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.511414558651706e-06,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8823606818914413,
+      "num_tokens": 113911914.0,
+      "step": 1065
+    },
+    {
+      "epoch": 2.428734321550741,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.5102954764545525e-06,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8803216964006424,
+      "num_tokens": 114018818.0,
+      "step": 1066
+    },
+    {
+      "epoch": 2.4310148232611173,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.509175253252497e-06,
+      "loss": 0.6118,
+      "mean_token_accuracy": 0.8786596357822418,
+      "num_tokens": 114125625.0,
+      "step": 1067
+    },
+    {
+      "epoch": 2.433295324971494,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.508053889681357e-06,
+      "loss": 0.5957,
+      "mean_token_accuracy": 0.87969671189785,
+      "num_tokens": 114232527.0,
+      "step": 1068
+    },
+    {
+      "epoch": 2.43557582668187,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.5069313863775956e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8781027346849442,
+      "num_tokens": 114339323.0,
+      "step": 1069
+    },
+    {
+      "epoch": 2.4378563283922463,
+      "grad_norm": 2.625,
+      "learning_rate": 4.505807743978325e-06,
+      "loss": 0.5874,
+      "mean_token_accuracy": 0.877943754196167,
+      "num_tokens": 114446811.0,
+      "step": 1070
+    },
+    {
+      "epoch": 2.4401368301026225,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.5046829631213014e-06,
+      "loss": 0.613,
+      "mean_token_accuracy": 0.8732447922229767,
+      "num_tokens": 114553671.0,
+      "step": 1071
+    },
+    {
+      "epoch": 2.4424173318129987,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.503557044444931e-06,
+      "loss": 0.5986,
+      "mean_token_accuracy": 0.8797748982906342,
+      "num_tokens": 114660235.0,
+      "step": 1072
+    },
+    {
+      "epoch": 2.4446978335233753,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.502429988588263e-06,
+      "loss": 0.6077,
+      "mean_token_accuracy": 0.8732910007238388,
+      "num_tokens": 114767907.0,
+      "step": 1073
+    },
+    {
+      "epoch": 2.4469783352337515,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.50130179619099e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8783971816301346,
+      "num_tokens": 114874874.0,
+      "step": 1074
+    },
+    {
+      "epoch": 2.4492588369441277,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.500172467893455e-06,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.8795461803674698,
+      "num_tokens": 114981618.0,
+      "step": 1075
+    },
+    {
+      "epoch": 2.451539338654504,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.499042004336642e-06,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.8823413103818893,
+      "num_tokens": 115088543.0,
+      "step": 1076
+    },
+    {
+      "epoch": 2.45381984036488,
+      "grad_norm": 6.90625,
+      "learning_rate": 4.497910406162182e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.8751686066389084,
+      "num_tokens": 115195316.0,
+      "step": 1077
+    },
+    {
+      "epoch": 2.4561003420752567,
+      "grad_norm": 6.28125,
+      "learning_rate": 4.496777674012345e-06,
+      "loss": 0.6081,
+      "mean_token_accuracy": 0.8768873512744904,
+      "num_tokens": 115302063.0,
+      "step": 1078
+    },
+    {
+      "epoch": 2.458380843785633,
+      "grad_norm": 2.125,
+      "learning_rate": 4.495643808530049e-06,
+      "loss": 0.6031,
+      "mean_token_accuracy": 0.8746660053730011,
+      "num_tokens": 115409028.0,
+      "step": 1079
+    },
+    {
+      "epoch": 2.460661345496009,
+      "grad_norm": 4.375,
+      "learning_rate": 4.494508810358855e-06,
+      "loss": 0.5939,
+      "mean_token_accuracy": 0.8768034875392914,
+      "num_tokens": 115516132.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.4629418472063853,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.4933726801429665e-06,
+      "loss": 0.6003,
+      "mean_token_accuracy": 0.8732063323259354,
+      "num_tokens": 115623543.0,
+      "step": 1081
+    },
+    {
+      "epoch": 2.4652223489167615,
+      "grad_norm": 6.0,
+      "learning_rate": 4.4922354185272275e-06,
+      "loss": 0.6137,
+      "mean_token_accuracy": 0.875907376408577,
+      "num_tokens": 115730261.0,
+      "step": 1082
+    },
+    {
+      "epoch": 2.467502850627138,
+      "grad_norm": 5.90625,
+      "learning_rate": 4.491097026157127e-06,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.8774657398462296,
+      "num_tokens": 115837186.0,
+      "step": 1083
+    },
+    {
+      "epoch": 2.4697833523375143,
+      "grad_norm": 3.375,
+      "learning_rate": 4.489957503678794e-06,
+      "loss": 0.5833,
+      "mean_token_accuracy": 0.8786139786243439,
+      "num_tokens": 115944585.0,
+      "step": 1084
+    },
+    {
+      "epoch": 2.4720638540478905,
+      "grad_norm": 6.3125,
+      "learning_rate": 4.488816851738999e-06,
+      "loss": 0.5921,
+      "mean_token_accuracy": 0.881994903087616,
+      "num_tokens": 116051127.0,
+      "step": 1085
+    },
+    {
+      "epoch": 2.4743443557582667,
+      "grad_norm": 4.0,
+      "learning_rate": 4.487675070985156e-06,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.879449263215065,
+      "num_tokens": 116158228.0,
+      "step": 1086
+    },
+    {
+      "epoch": 2.476624857468643,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.4865321620653144e-06,
+      "loss": 0.5908,
+      "mean_token_accuracy": 0.8789650350809097,
+      "num_tokens": 116265116.0,
+      "step": 1087
+    },
+    {
+      "epoch": 2.4789053591790196,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.485388125628171e-06,
+      "loss": 0.5956,
+      "mean_token_accuracy": 0.8748547732830048,
+      "num_tokens": 116372159.0,
+      "step": 1088
+    },
+    {
+      "epoch": 2.4811858608893957,
+      "grad_norm": 5.40625,
+      "learning_rate": 4.484242962323056e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8844872564077377,
+      "num_tokens": 116478884.0,
+      "step": 1089
+    },
+    {
+      "epoch": 2.483466362599772,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.483096672799942e-06,
+      "loss": 0.597,
+      "mean_token_accuracy": 0.8777136653661728,
+      "num_tokens": 116585905.0,
+      "step": 1090
+    },
+    {
+      "epoch": 2.485746864310148,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.481949257709442e-06,
+      "loss": 0.5792,
+      "mean_token_accuracy": 0.8820765465497971,
+      "num_tokens": 116693091.0,
+      "step": 1091
+    },
+    {
+      "epoch": 2.4880273660205243,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.480800717702807e-06,
+      "loss": 0.5935,
+      "mean_token_accuracy": 0.8752633780241013,
+      "num_tokens": 116801048.0,
+      "step": 1092
+    },
+    {
+      "epoch": 2.490307867730901,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.479651053431926e-06,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8796508759260178,
+      "num_tokens": 116907982.0,
+      "step": 1093
+    },
+    {
+      "epoch": 2.492588369441277,
+      "grad_norm": 2.875,
+      "learning_rate": 4.4785002655493246e-06,
+      "loss": 0.5905,
+      "mean_token_accuracy": 0.8795785456895828,
+      "num_tokens": 117015153.0,
+      "step": 1094
+    },
+    {
+      "epoch": 2.4948688711516533,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.477348354708169e-06,
+      "loss": 0.5995,
+      "mean_token_accuracy": 0.8773641586303711,
+      "num_tokens": 117122420.0,
+      "step": 1095
+    },
+    {
+      "epoch": 2.4971493728620295,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.476195321562262e-06,
+      "loss": 0.5962,
+      "mean_token_accuracy": 0.8779642879962921,
+      "num_tokens": 117229161.0,
+      "step": 1096
+    },
+    {
+      "epoch": 2.4994298745724057,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.475041166766042e-06,
+      "loss": 0.6134,
+      "mean_token_accuracy": 0.8731478750705719,
+      "num_tokens": 117336152.0,
+      "step": 1097
+    },
+    {
+      "epoch": 2.5017103762827824,
+      "grad_norm": 4.875,
+      "learning_rate": 4.473885890974586e-06,
+      "loss": 0.614,
+      "mean_token_accuracy": 0.8745481222867966,
+      "num_tokens": 117443787.0,
+      "step": 1098
+    },
+    {
+      "epoch": 2.5039908779931586,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.472729494843605e-06,
+      "loss": 0.5917,
+      "mean_token_accuracy": 0.8793596476316452,
+      "num_tokens": 117550788.0,
+      "step": 1099
+    },
+    {
+      "epoch": 2.5062713797035348,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.471571979029448e-06,
+      "loss": 0.5808,
+      "mean_token_accuracy": 0.8826311677694321,
+      "num_tokens": 117658107.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.5062713797035348,
+      "eval_loss": 0.6002530455589294,
+      "eval_mean_token_accuracy": 0.8773398306433239,
+      "eval_num_tokens": 117658107.0,
+      "eval_runtime": 58.5432,
+      "eval_samples_per_second": 143.228,
+      "eval_steps_per_second": 4.492,
+      "step": 1100
+    },
+    {
+      "epoch": 2.508551881413911,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.470413344189098e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8801444619894028,
+      "num_tokens": 117764556.0,
+      "step": 1101
+    },
+    {
+      "epoch": 2.5108323831242876,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.469253590980175e-06,
+      "loss": 0.5925,
+      "mean_token_accuracy": 0.8782180845737457,
+      "num_tokens": 117871482.0,
+      "step": 1102
+    },
+    {
+      "epoch": 2.5131128848346638,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.46809272006093e-06,
+      "loss": 0.5886,
+      "mean_token_accuracy": 0.8766061216592789,
+      "num_tokens": 117978366.0,
+      "step": 1103
+    },
+    {
+      "epoch": 2.51539338654504,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.466930732090254e-06,
+      "loss": 0.5901,
+      "mean_token_accuracy": 0.8791538327932358,
+      "num_tokens": 118085055.0,
+      "step": 1104
+    },
+    {
+      "epoch": 2.517673888255416,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.465767627727668e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8798196315765381,
+      "num_tokens": 118192713.0,
+      "step": 1105
+    },
+    {
+      "epoch": 2.5199543899657924,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.464603407633326e-06,
+      "loss": 0.6121,
+      "mean_token_accuracy": 0.8753893822431564,
+      "num_tokens": 118299578.0,
+      "step": 1106
+    },
+    {
+      "epoch": 2.5222348916761685,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.463438072468018e-06,
+      "loss": 0.5943,
+      "mean_token_accuracy": 0.8764231652021408,
+      "num_tokens": 118406017.0,
+      "step": 1107
+    },
+    {
+      "epoch": 2.524515393386545,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.462271622893166e-06,
+      "loss": 0.6068,
+      "mean_token_accuracy": 0.8774004876613617,
+      "num_tokens": 118512926.0,
+      "step": 1108
+    },
+    {
+      "epoch": 2.5267958950969214,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.461104059570825e-06,
+      "loss": 0.6248,
+      "mean_token_accuracy": 0.8694035857915878,
+      "num_tokens": 118619939.0,
+      "step": 1109
+    },
+    {
+      "epoch": 2.5290763968072976,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.4599353831636785e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8803573846817017,
+      "num_tokens": 118727362.0,
+      "step": 1110
+    },
+    {
+      "epoch": 2.5313568985176738,
+      "grad_norm": 6.84375,
+      "learning_rate": 4.458765594335048e-06,
+      "loss": 0.5913,
+      "mean_token_accuracy": 0.877105325460434,
+      "num_tokens": 118834306.0,
+      "step": 1111
+    },
+    {
+      "epoch": 2.5336374002280504,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.457594693748881e-06,
+      "loss": 0.6192,
+      "mean_token_accuracy": 0.8765625059604645,
+      "num_tokens": 118941239.0,
+      "step": 1112
+    },
+    {
+      "epoch": 2.5359179019384266,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.456422682069758e-06,
+      "loss": 0.5766,
+      "mean_token_accuracy": 0.8844843655824661,
+      "num_tokens": 119048172.0,
+      "step": 1113
+    },
+    {
+      "epoch": 2.538198403648803,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.455249559962892e-06,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.880544051527977,
+      "num_tokens": 119155079.0,
+      "step": 1114
+    },
+    {
+      "epoch": 2.540478905359179,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.454075328094123e-06,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8793693780899048,
+      "num_tokens": 119262757.0,
+      "step": 1115
+    },
+    {
+      "epoch": 2.542759407069555,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.452899987129922e-06,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8787418156862259,
+      "num_tokens": 119370566.0,
+      "step": 1116
+    },
+    {
+      "epoch": 2.5450399087799314,
+      "grad_norm": 2.625,
+      "learning_rate": 4.4517235377373915e-06,
+      "loss": 0.6146,
+      "mean_token_accuracy": 0.8720010370016098,
+      "num_tokens": 119477698.0,
+      "step": 1117
+    },
+    {
+      "epoch": 2.547320410490308,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.45054598058426e-06,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8797213733196259,
+      "num_tokens": 119584695.0,
+      "step": 1118
+    },
+    {
+      "epoch": 2.549600912200684,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.449367316338887e-06,
+      "loss": 0.5789,
+      "mean_token_accuracy": 0.8773613125085831,
+      "num_tokens": 119691823.0,
+      "step": 1119
+    },
+    {
+      "epoch": 2.5518814139110604,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.448187545670258e-06,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.881779134273529,
+      "num_tokens": 119799172.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.5541619156214366,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.44700666924799e-06,
+      "loss": 0.6025,
+      "mean_token_accuracy": 0.874766156077385,
+      "num_tokens": 119905784.0,
+      "step": 1121
+    },
+    {
+      "epoch": 2.556442417331813,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.4458246877423254e-06,
+      "loss": 0.6032,
+      "mean_token_accuracy": 0.8753187209367752,
+      "num_tokens": 120012658.0,
+      "step": 1122
+    },
+    {
+      "epoch": 2.5587229190421894,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.444641601824134e-06,
+      "loss": 0.5871,
+      "mean_token_accuracy": 0.8781961053609848,
+      "num_tokens": 120119347.0,
+      "step": 1123
+    },
+    {
+      "epoch": 2.5610034207525656,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.443457412164911e-06,
+      "loss": 0.5931,
+      "mean_token_accuracy": 0.8751012235879898,
+      "num_tokens": 120226332.0,
+      "step": 1124
+    },
+    {
+      "epoch": 2.563283922462942,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.442272119436781e-06,
+      "loss": 0.5954,
+      "mean_token_accuracy": 0.8773366808891296,
+      "num_tokens": 120334032.0,
+      "step": 1125
+    },
+    {
+      "epoch": 2.565564424173318,
+      "grad_norm": 3.5,
+      "learning_rate": 4.441085724312494e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8794488459825516,
+      "num_tokens": 120441226.0,
+      "step": 1126
+    },
+    {
+      "epoch": 2.567844925883694,
+      "grad_norm": 2.875,
+      "learning_rate": 4.4398982274654235e-06,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8812543451786041,
+      "num_tokens": 120549055.0,
+      "step": 1127
+    },
+    {
+      "epoch": 2.570125427594071,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.43870962956957e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8838596194982529,
+      "num_tokens": 120656376.0,
+      "step": 1128
+    },
+    {
+      "epoch": 2.572405929304447,
+      "grad_norm": 3.125,
+      "learning_rate": 4.437519931299559e-06,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.878927618265152,
+      "num_tokens": 120764475.0,
+      "step": 1129
+    },
+    {
+      "epoch": 2.574686431014823,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.43632913333064e-06,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8791298568248749,
+      "num_tokens": 120871237.0,
+      "step": 1130
+    },
+    {
+      "epoch": 2.5769669327251994,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.435137236338688e-06,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8821601420640945,
+      "num_tokens": 120978615.0,
+      "step": 1131
+    },
+    {
+      "epoch": 2.579247434435576,
+      "grad_norm": 6.46875,
+      "learning_rate": 4.433944241000199e-06,
+      "loss": 0.6004,
+      "mean_token_accuracy": 0.8753758072853088,
+      "num_tokens": 121085644.0,
+      "step": 1132
+    },
+    {
+      "epoch": 2.581527936145952,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.4327501479922955e-06,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8809983879327774,
+      "num_tokens": 121193058.0,
+      "step": 1133
+    },
+    {
+      "epoch": 2.5838084378563284,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.431554957992722e-06,
+      "loss": 0.6168,
+      "mean_token_accuracy": 0.8755798637866974,
+      "num_tokens": 121299745.0,
+      "step": 1134
+    },
+    {
+      "epoch": 2.5860889395667046,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.430358671679843e-06,
+      "loss": 0.6168,
+      "mean_token_accuracy": 0.8684369623661041,
+      "num_tokens": 121406377.0,
+      "step": 1135
+    },
+    {
+      "epoch": 2.588369441277081,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.42916128973265e-06,
+      "loss": 0.6126,
+      "mean_token_accuracy": 0.8750972300767899,
+      "num_tokens": 121513300.0,
+      "step": 1136
+    },
+    {
+      "epoch": 2.590649942987457,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.427962812830753e-06,
+      "loss": 0.6178,
+      "mean_token_accuracy": 0.8743036985397339,
+      "num_tokens": 121620489.0,
+      "step": 1137
+    },
+    {
+      "epoch": 2.5929304446978336,
+      "grad_norm": 4.25,
+      "learning_rate": 4.426763241654383e-06,
+      "loss": 0.6034,
+      "mean_token_accuracy": 0.8752316683530807,
+      "num_tokens": 121727667.0,
+      "step": 1138
+    },
+    {
+      "epoch": 2.59521094640821,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.425562576884396e-06,
+      "loss": 0.5877,
+      "mean_token_accuracy": 0.8789777606725693,
+      "num_tokens": 121834628.0,
+      "step": 1139
+    },
+    {
+      "epoch": 2.597491448118586,
+      "grad_norm": 4.125,
+      "learning_rate": 4.424360819202264e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.8804343789815903,
+      "num_tokens": 121941613.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.5997719498289626,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.423157969290081e-06,
+      "loss": 0.6015,
+      "mean_token_accuracy": 0.8748785555362701,
+      "num_tokens": 122048337.0,
+      "step": 1141
+    },
+    {
+      "epoch": 2.602052451539339,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.421954027830565e-06,
+      "loss": 0.5984,
+      "mean_token_accuracy": 0.8761484026908875,
+      "num_tokens": 122155743.0,
+      "step": 1142
+    },
+    {
+      "epoch": 2.604332953249715,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.4207489955070465e-06,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8797977864742279,
+      "num_tokens": 122262415.0,
+      "step": 1143
+    },
+    {
+      "epoch": 2.6066134549600912,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.419542873003479e-06,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.879091739654541,
+      "num_tokens": 122369829.0,
+      "step": 1144
+    },
+    {
+      "epoch": 2.6088939566704674,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.418335661004436e-06,
+      "loss": 0.6004,
+      "mean_token_accuracy": 0.8749701827764511,
+      "num_tokens": 122476808.0,
+      "step": 1145
+    },
+    {
+      "epoch": 2.6111744583808436,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.417127360195107e-06,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8770206123590469,
+      "num_tokens": 122583941.0,
+      "step": 1146
+    },
+    {
+      "epoch": 2.61345496009122,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.415917971261299e-06,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8799891471862793,
+      "num_tokens": 122691451.0,
+      "step": 1147
+    },
+    {
+      "epoch": 2.6157354618015964,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.414707494889439e-06,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8810625970363617,
+      "num_tokens": 122798569.0,
+      "step": 1148
+    },
+    {
+      "epoch": 2.6180159635119726,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.413495931766571e-06,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.8791490197181702,
+      "num_tokens": 122906142.0,
+      "step": 1149
+    },
+    {
+      "epoch": 2.620296465222349,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.412283282580352e-06,
+      "loss": 0.5818,
+      "mean_token_accuracy": 0.880540743470192,
+      "num_tokens": 123012808.0,
+      "step": 1150
+    },
+    {
+      "epoch": 2.6225769669327255,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.41106954801906e-06,
+      "loss": 0.5901,
+      "mean_token_accuracy": 0.8770016133785248,
+      "num_tokens": 123119979.0,
+      "step": 1151
+    },
+    {
+      "epoch": 2.6248574686431017,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.409854728771588e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8810955137014389,
+      "num_tokens": 123227161.0,
+      "step": 1152
+    },
+    {
+      "epoch": 2.627137970353478,
+      "grad_norm": 4.375,
+      "learning_rate": 4.4086388255274425e-06,
+      "loss": 0.5973,
+      "mean_token_accuracy": 0.8776374161243439,
+      "num_tokens": 123333866.0,
+      "step": 1153
+    },
+    {
+      "epoch": 2.629418472063854,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.407421838976747e-06,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8836002796888351,
+      "num_tokens": 123441095.0,
+      "step": 1154
+    },
+    {
+      "epoch": 2.6316989737742302,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.40620376981024e-06,
+      "loss": 0.5776,
+      "mean_token_accuracy": 0.8822665065526962,
+      "num_tokens": 123548255.0,
+      "step": 1155
+    },
+    {
+      "epoch": 2.6339794754846064,
+      "grad_norm": 3.875,
+      "learning_rate": 4.404984618719275e-06,
+      "loss": 0.6157,
+      "mean_token_accuracy": 0.8759646117687225,
+      "num_tokens": 123654882.0,
+      "step": 1156
+    },
+    {
+      "epoch": 2.636259977194983,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.403764386395817e-06,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8818103224039078,
+      "num_tokens": 123762537.0,
+      "step": 1157
+    },
+    {
+      "epoch": 2.6385404789053593,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.402543073532446e-06,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8845062106847763,
+      "num_tokens": 123870045.0,
+      "step": 1158
+    },
+    {
+      "epoch": 2.6408209806157354,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.401320680822357e-06,
+      "loss": 0.5626,
+      "mean_token_accuracy": 0.8820901811122894,
+      "num_tokens": 123977756.0,
+      "step": 1159
+    },
+    {
+      "epoch": 2.6431014823261116,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.400097208959357e-06,
+      "loss": 0.5636,
+      "mean_token_accuracy": 0.8831377625465393,
+      "num_tokens": 124085720.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.6453819840364883,
+      "grad_norm": 5.375,
+      "learning_rate": 4.398872658637863e-06,
+      "loss": 0.5896,
+      "mean_token_accuracy": 0.8773757070302963,
+      "num_tokens": 124192659.0,
+      "step": 1161
+    },
+    {
+      "epoch": 2.6476624857468645,
+      "grad_norm": 4.9375,
+      "learning_rate": 4.397647030552907e-06,
+      "loss": 0.6128,
+      "mean_token_accuracy": 0.8716171085834503,
+      "num_tokens": 124299363.0,
+      "step": 1162
+    },
+    {
+      "epoch": 2.6499429874572407,
+      "grad_norm": 4.75,
+      "learning_rate": 4.396420325400132e-06,
+      "loss": 0.5984,
+      "mean_token_accuracy": 0.8754719495773315,
+      "num_tokens": 124406432.0,
+      "step": 1163
+    },
+    {
+      "epoch": 2.652223489167617,
+      "grad_norm": 8.6875,
+      "learning_rate": 4.3951925438757936e-06,
+      "loss": 0.6075,
+      "mean_token_accuracy": 0.8721831738948822,
+      "num_tokens": 124513942.0,
+      "step": 1164
+    },
+    {
+      "epoch": 2.654503990877993,
+      "grad_norm": 5.6875,
+      "learning_rate": 4.3939636866767535e-06,
+      "loss": 0.5874,
+      "mean_token_accuracy": 0.8772239238023758,
+      "num_tokens": 124621397.0,
+      "step": 1165
+    },
+    {
+      "epoch": 2.6567844925883692,
+      "grad_norm": 9.8125,
+      "learning_rate": 4.39273375450049e-06,
+      "loss": 0.6082,
+      "mean_token_accuracy": 0.8729386478662491,
+      "num_tokens": 124728132.0,
+      "step": 1166
+    },
+    {
+      "epoch": 2.659064994298746,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.391502748045088e-06,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.8794106990098953,
+      "num_tokens": 124834818.0,
+      "step": 1167
+    },
+    {
+      "epoch": 2.661345496009122,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.390270668009244e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8794341683387756,
+      "num_tokens": 124941733.0,
+      "step": 1168
+    },
+    {
+      "epoch": 2.6636259977194983,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.38903751509226e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8788381963968277,
+      "num_tokens": 125049080.0,
+      "step": 1169
+    },
+    {
+      "epoch": 2.6659064994298745,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.3878032899940534e-06,
+      "loss": 0.601,
+      "mean_token_accuracy": 0.8787420690059662,
+      "num_tokens": 125156160.0,
+      "step": 1170
+    },
+    {
+      "epoch": 2.668187001140251,
+      "grad_norm": 6.1875,
+      "learning_rate": 4.386567993415144e-06,
+      "loss": 0.5948,
+      "mean_token_accuracy": 0.8779895603656769,
+      "num_tokens": 125263507.0,
+      "step": 1171
+    },
+    {
+      "epoch": 2.6704675028506273,
+      "grad_norm": 5.375,
+      "learning_rate": 4.3853316260566635e-06,
+      "loss": 0.5665,
+      "mean_token_accuracy": 0.8852428048849106,
+      "num_tokens": 125370490.0,
+      "step": 1172
+    },
+    {
+      "epoch": 2.6727480045610035,
+      "grad_norm": 5.03125,
+      "learning_rate": 4.384094188620349e-06,
+      "loss": 0.6148,
+      "mean_token_accuracy": 0.8747402876615524,
+      "num_tokens": 125477079.0,
+      "step": 1173
+    },
+    {
+      "epoch": 2.6750285062713797,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.3828556818085485e-06,
+      "loss": 0.5914,
+      "mean_token_accuracy": 0.8805000483989716,
+      "num_tokens": 125584049.0,
+      "step": 1174
+    },
+    {
+      "epoch": 2.677309007981756,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.3816161063242115e-06,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8765310496091843,
+      "num_tokens": 125690629.0,
+      "step": 1175
+    },
+    {
+      "epoch": 2.679589509692132,
+      "grad_norm": 2.625,
+      "learning_rate": 4.3803754628708995e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8820624649524689,
+      "num_tokens": 125798232.0,
+      "step": 1176
+    },
+    {
+      "epoch": 2.6818700114025087,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.379133752152776e-06,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8863883018493652,
+      "num_tokens": 125905642.0,
+      "step": 1177
+    },
+    {
+      "epoch": 2.684150513112885,
+      "grad_norm": 2.25,
+      "learning_rate": 4.377890974874614e-06,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.8794945180416107,
+      "num_tokens": 126013558.0,
+      "step": 1178
+    },
+    {
+      "epoch": 2.686431014823261,
+      "grad_norm": 2.125,
+      "learning_rate": 4.376647131741787e-06,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8828789293766022,
+      "num_tokens": 126120692.0,
+      "step": 1179
+    },
+    {
+      "epoch": 2.6887115165336373,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.375402223460279e-06,
+      "loss": 0.6034,
+      "mean_token_accuracy": 0.8755092322826385,
+      "num_tokens": 126227571.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.690992018244014,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.3741562507366754e-06,
+      "loss": 0.6015,
+      "mean_token_accuracy": 0.8771520704030991,
+      "num_tokens": 126334666.0,
+      "step": 1181
+    },
+    {
+      "epoch": 2.69327251995439,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.3729092142781655e-06,
+      "loss": 0.5598,
+      "mean_token_accuracy": 0.8833112865686417,
+      "num_tokens": 126441830.0,
+      "step": 1182
+    },
+    {
+      "epoch": 2.6955530216647663,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.3716611147925435e-06,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8818268924951553,
+      "num_tokens": 126548549.0,
+      "step": 1183
+    },
+    {
+      "epoch": 2.6978335233751425,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.370411952988207e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8798163086175919,
+      "num_tokens": 126655776.0,
+      "step": 1184
+    },
+    {
+      "epoch": 2.7001140250855187,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.369161729574155e-06,
+      "loss": 0.6053,
+      "mean_token_accuracy": 0.8758595883846283,
+      "num_tokens": 126762585.0,
+      "step": 1185
+    },
+    {
+      "epoch": 2.702394526795895,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.367910445259991e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8808387517929077,
+      "num_tokens": 126869208.0,
+      "step": 1186
+    },
+    {
+      "epoch": 2.7046750285062715,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.36665810075592e-06,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.877717912197113,
+      "num_tokens": 126976031.0,
+      "step": 1187
+    },
+    {
+      "epoch": 2.7069555302166477,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.365404696772748e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8818590492010117,
+      "num_tokens": 127083381.0,
+      "step": 1188
+    },
+    {
+      "epoch": 2.709236031927024,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.364150234021883e-06,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8820794522762299,
+      "num_tokens": 127190792.0,
+      "step": 1189
+    },
+    {
+      "epoch": 2.7115165336374,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.362894713215334e-06,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8834023624658585,
+      "num_tokens": 127298871.0,
+      "step": 1190
+    },
+    {
+      "epoch": 2.7137970353477767,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.361638135065711e-06,
+      "loss": 0.557,
+      "mean_token_accuracy": 0.8848878443241119,
+      "num_tokens": 127407231.0,
+      "step": 1191
+    },
+    {
+      "epoch": 2.716077537058153,
+      "grad_norm": 3.25,
+      "learning_rate": 4.360380500286222e-06,
+      "loss": 0.5808,
+      "mean_token_accuracy": 0.8773325383663177,
+      "num_tokens": 127514509.0,
+      "step": 1192
+    },
+    {
+      "epoch": 2.718358038768529,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.359121809590678e-06,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8788514882326126,
+      "num_tokens": 127621769.0,
+      "step": 1193
+    },
+    {
+      "epoch": 2.7206385404789053,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.357862063693486e-06,
+      "loss": 0.6107,
+      "mean_token_accuracy": 0.8736522942781448,
+      "num_tokens": 127728235.0,
+      "step": 1194
+    },
+    {
+      "epoch": 2.7229190421892815,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.356601263309654e-06,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8788127601146698,
+      "num_tokens": 127834915.0,
+      "step": 1195
+    },
+    {
+      "epoch": 2.7251995438996577,
+      "grad_norm": 2.5,
+      "learning_rate": 4.355339409154788e-06,
+      "loss": 0.5965,
+      "mean_token_accuracy": 0.8762216120958328,
+      "num_tokens": 127942200.0,
+      "step": 1196
+    },
+    {
+      "epoch": 2.7274800456100343,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.354076501945093e-06,
+      "loss": 0.6054,
+      "mean_token_accuracy": 0.8778090626001358,
+      "num_tokens": 128049307.0,
+      "step": 1197
+    },
+    {
+      "epoch": 2.7297605473204105,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.352812542397369e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8793000727891922,
+      "num_tokens": 128155836.0,
+      "step": 1198
+    },
+    {
+      "epoch": 2.7320410490307867,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.351547531229016e-06,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.8827884644269943,
+      "num_tokens": 128262857.0,
+      "step": 1199
+    },
+    {
+      "epoch": 2.734321550741163,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.350281469158029e-06,
+      "loss": 0.61,
+      "mean_token_accuracy": 0.8760698437690735,
+      "num_tokens": 128369643.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.7366020524515395,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.3490143569030025e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.8797038048505783,
+      "num_tokens": 128476571.0,
+      "step": 1201
+    },
+    {
+      "epoch": 2.7388825541619157,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.347746195183123e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8761547356843948,
+      "num_tokens": 128583850.0,
+      "step": 1202
+    },
+    {
+      "epoch": 2.741163055872292,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.346476984718176e-06,
+      "loss": 0.6078,
+      "mean_token_accuracy": 0.8726416677236557,
+      "num_tokens": 128691263.0,
+      "step": 1203
+    },
+    {
+      "epoch": 2.743443557582668,
+      "grad_norm": 5.59375,
+      "learning_rate": 4.345206726228538e-06,
+      "loss": 0.5859,
+      "mean_token_accuracy": 0.8812550455331802,
+      "num_tokens": 128798421.0,
+      "step": 1204
+    },
+    {
+      "epoch": 2.7457240592930443,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.343935420435187e-06,
+      "loss": 0.5942,
+      "mean_token_accuracy": 0.8780511021614075,
+      "num_tokens": 128905472.0,
+      "step": 1205
+    },
+    {
+      "epoch": 2.7480045610034205,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.34266306805969e-06,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8798847049474716,
+      "num_tokens": 129012064.0,
+      "step": 1206
+    },
+    {
+      "epoch": 2.750285062713797,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.341389669824209e-06,
+      "loss": 0.5915,
+      "mean_token_accuracy": 0.8787447810173035,
+      "num_tokens": 129118812.0,
+      "step": 1207
+    },
+    {
+      "epoch": 2.7525655644241733,
+      "grad_norm": 4.0,
+      "learning_rate": 4.340115226451501e-06,
+      "loss": 0.602,
+      "mean_token_accuracy": 0.8773921728134155,
+      "num_tokens": 129226022.0,
+      "step": 1208
+    },
+    {
+      "epoch": 2.7548460661345495,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.338839738664915e-06,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8822150230407715,
+      "num_tokens": 129333172.0,
+      "step": 1209
+    },
+    {
+      "epoch": 2.757126567844926,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.3375632071883935e-06,
+      "loss": 0.5966,
+      "mean_token_accuracy": 0.8766026943922043,
+      "num_tokens": 129440070.0,
+      "step": 1210
+    },
+    {
+      "epoch": 2.7594070695553023,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.336285632746472e-06,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.8804485648870468,
+      "num_tokens": 129547114.0,
+      "step": 1211
+    },
+    {
+      "epoch": 2.7616875712656785,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.3350070160642754e-06,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8801703006029129,
+      "num_tokens": 129653799.0,
+      "step": 1212
+    },
+    {
+      "epoch": 2.7639680729760547,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.333727357867523e-06,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8849901556968689,
+      "num_tokens": 129761360.0,
+      "step": 1213
+    },
+    {
+      "epoch": 2.766248574686431,
+      "grad_norm": 2.5,
+      "learning_rate": 4.3324466588825235e-06,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8838348835706711,
+      "num_tokens": 129868322.0,
+      "step": 1214
+    },
+    {
+      "epoch": 2.768529076396807,
+      "grad_norm": 2.125,
+      "learning_rate": 4.331164919836177e-06,
+      "loss": 0.5913,
+      "mean_token_accuracy": 0.8776644319295883,
+      "num_tokens": 129975643.0,
+      "step": 1215
+    },
+    {
+      "epoch": 2.7708095781071833,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.329882141455974e-06,
+      "loss": 0.5871,
+      "mean_token_accuracy": 0.8794203400611877,
+      "num_tokens": 130082157.0,
+      "step": 1216
+    },
+    {
+      "epoch": 2.77309007981756,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.3285983244699955e-06,
+      "loss": 0.5854,
+      "mean_token_accuracy": 0.8823015242815018,
+      "num_tokens": 130189446.0,
+      "step": 1217
+    },
+    {
+      "epoch": 2.775370581527936,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.327313469606911e-06,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8795596957206726,
+      "num_tokens": 130296713.0,
+      "step": 1218
+    },
+    {
+      "epoch": 2.7776510832383123,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.326027577595977e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8805474638938904,
+      "num_tokens": 130403572.0,
+      "step": 1219
+    },
+    {
+      "epoch": 2.779931584948689,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.324740649167044e-06,
+      "loss": 0.5829,
+      "mean_token_accuracy": 0.8781325221061707,
+      "num_tokens": 130510616.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.782212086659065,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.323452685050545e-06,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8842640519142151,
+      "num_tokens": 130618088.0,
+      "step": 1221
+    },
+    {
+      "epoch": 2.7844925883694414,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.3221636859775075e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8810494989156723,
+      "num_tokens": 130725134.0,
+      "step": 1222
+    },
+    {
+      "epoch": 2.7867730900798175,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.320873652679538e-06,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.8809178620576859,
+      "num_tokens": 130832079.0,
+      "step": 1223
+    },
+    {
+      "epoch": 2.7890535917901937,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.319582585888838e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8811828643083572,
+      "num_tokens": 130939236.0,
+      "step": 1224
+    },
+    {
+      "epoch": 2.79133409350057,
+      "grad_norm": 2.25,
+      "learning_rate": 4.31829048633819e-06,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8782161772251129,
+      "num_tokens": 131046662.0,
+      "step": 1225
+    },
+    {
+      "epoch": 2.7936145952109466,
+      "grad_norm": 2.625,
+      "learning_rate": 4.316997354760965e-06,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8795737326145172,
+      "num_tokens": 131153949.0,
+      "step": 1226
+    },
+    {
+      "epoch": 2.7958950969213228,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.3157031918911204e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.8794270306825638,
+      "num_tokens": 131260517.0,
+      "step": 1227
+    },
+    {
+      "epoch": 2.798175598631699,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.314407998463198e-06,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.881917729973793,
+      "num_tokens": 131367523.0,
+      "step": 1228
+    },
+    {
+      "epoch": 2.800456100342075,
+      "grad_norm": 6.40625,
+      "learning_rate": 4.3131117752123235e-06,
+      "loss": 0.598,
+      "mean_token_accuracy": 0.8779633790254593,
+      "num_tokens": 131474262.0,
+      "step": 1229
+    },
+    {
+      "epoch": 2.802736602052452,
+      "grad_norm": 3.5,
+      "learning_rate": 4.311814522874209e-06,
+      "loss": 0.5936,
+      "mean_token_accuracy": 0.8795278668403625,
+      "num_tokens": 131581107.0,
+      "step": 1230
+    },
+    {
+      "epoch": 2.805017103762828,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.3105162421851494e-06,
+      "loss": 0.6044,
+      "mean_token_accuracy": 0.8744796365499496,
+      "num_tokens": 131687930.0,
+      "step": 1231
+    },
+    {
+      "epoch": 2.807297605473204,
+      "grad_norm": 5.8125,
+      "learning_rate": 4.309216933882025e-06,
+      "loss": 0.5833,
+      "mean_token_accuracy": 0.8764165937900543,
+      "num_tokens": 131794765.0,
+      "step": 1232
+    },
+    {
+      "epoch": 2.8095781071835804,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.307916598702296e-06,
+      "loss": 0.5676,
+      "mean_token_accuracy": 0.883644163608551,
+      "num_tokens": 131902317.0,
+      "step": 1233
+    },
+    {
+      "epoch": 2.8118586088939566,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.3066152373840105e-06,
+      "loss": 0.6001,
+      "mean_token_accuracy": 0.8781489878892899,
+      "num_tokens": 132009851.0,
+      "step": 1234
+    },
+    {
+      "epoch": 2.8141391106043327,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.305312850665794e-06,
+      "loss": 0.5967,
+      "mean_token_accuracy": 0.8778474628925323,
+      "num_tokens": 132116939.0,
+      "step": 1235
+    },
+    {
+      "epoch": 2.8164196123147094,
+      "grad_norm": 4.0,
+      "learning_rate": 4.304009439286855e-06,
+      "loss": 0.5897,
+      "mean_token_accuracy": 0.8780102729797363,
+      "num_tokens": 132223406.0,
+      "step": 1236
+    },
+    {
+      "epoch": 2.8187001140250856,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.3027050039869865e-06,
+      "loss": 0.5879,
+      "mean_token_accuracy": 0.8773478418588638,
+      "num_tokens": 132330700.0,
+      "step": 1237
+    },
+    {
+      "epoch": 2.8209806157354618,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.301399545506561e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8794009536504745,
+      "num_tokens": 132437921.0,
+      "step": 1238
+    },
+    {
+      "epoch": 2.823261117445838,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.3000930645865305e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8809588998556137,
+      "num_tokens": 132545056.0,
+      "step": 1239
+    },
+    {
+      "epoch": 2.8255416191562146,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.298785561968428e-06,
+      "loss": 0.5899,
+      "mean_token_accuracy": 0.8774296343326569,
+      "num_tokens": 132651620.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.827822120866591,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.297477038394368e-06,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.8836159557104111,
+      "num_tokens": 132758652.0,
+      "step": 1241
+    },
+    {
+      "epoch": 2.830102622576967,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.296167494607043e-06,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.8778567016124725,
+      "num_tokens": 132866236.0,
+      "step": 1242
+    },
+    {
+      "epoch": 2.832383124287343,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.294856931349724e-06,
+      "loss": 0.6096,
+      "mean_token_accuracy": 0.8745770305395126,
+      "num_tokens": 132973334.0,
+      "step": 1243
+    },
+    {
+      "epoch": 2.8346636259977194,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.293545349366262e-06,
+      "loss": 0.5915,
+      "mean_token_accuracy": 0.8771042227745056,
+      "num_tokens": 133079721.0,
+      "step": 1244
+    },
+    {
+      "epoch": 2.8369441277080956,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.292232749401085e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8794742226600647,
+      "num_tokens": 133186577.0,
+      "step": 1245
+    },
+    {
+      "epoch": 2.839224629418472,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.2909191321992e-06,
+      "loss": 0.6038,
+      "mean_token_accuracy": 0.8782872408628464,
+      "num_tokens": 133293871.0,
+      "step": 1246
+    },
+    {
+      "epoch": 2.8415051311288484,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.2896044985061915e-06,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.8800533413887024,
+      "num_tokens": 133400863.0,
+      "step": 1247
+    },
+    {
+      "epoch": 2.8437856328392246,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.288288849068218e-06,
+      "loss": 0.5854,
+      "mean_token_accuracy": 0.8809787184000015,
+      "num_tokens": 133507932.0,
+      "step": 1248
+    },
+    {
+      "epoch": 2.846066134549601,
+      "grad_norm": 3.25,
+      "learning_rate": 4.286972184632019e-06,
+      "loss": 0.6134,
+      "mean_token_accuracy": 0.8738250583410263,
+      "num_tokens": 133615287.0,
+      "step": 1249
+    },
+    {
+      "epoch": 2.8483466362599774,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.285654505944906e-06,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8837039470672607,
+      "num_tokens": 133722203.0,
+      "step": 1250
+    },
+    {
+      "epoch": 2.8506271379703536,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.28433581375477e-06,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.8851412385702133,
+      "num_tokens": 133829983.0,
+      "step": 1251
+    },
+    {
+      "epoch": 2.85290763968073,
+      "grad_norm": 3.5,
+      "learning_rate": 4.283016108810073e-06,
+      "loss": 0.6047,
+      "mean_token_accuracy": 0.8773223906755447,
+      "num_tokens": 133937207.0,
+      "step": 1252
+    },
+    {
+      "epoch": 2.855188141391106,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.281695391859854e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8886352777481079,
+      "num_tokens": 134044589.0,
+      "step": 1253
+    },
+    {
+      "epoch": 2.857468643101482,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.28037366365373e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.8755118399858475,
+      "num_tokens": 134151607.0,
+      "step": 1254
+    },
+    {
+      "epoch": 2.8597491448118584,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.279050924941885e-06,
+      "loss": 0.5854,
+      "mean_token_accuracy": 0.8779601603746414,
+      "num_tokens": 134258813.0,
+      "step": 1255
+    },
+    {
+      "epoch": 2.862029646522235,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.2777271764750805e-06,
+      "loss": 0.5679,
+      "mean_token_accuracy": 0.8823637515306473,
+      "num_tokens": 134366131.0,
+      "step": 1256
+    },
+    {
+      "epoch": 2.864310148232611,
+      "grad_norm": 2.125,
+      "learning_rate": 4.276402419004652e-06,
+      "loss": 0.5843,
+      "mean_token_accuracy": 0.8769189864397049,
+      "num_tokens": 134472934.0,
+      "step": 1257
+    },
+    {
+      "epoch": 2.8665906499429874,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.275076653282504e-06,
+      "loss": 0.5915,
+      "mean_token_accuracy": 0.8746793568134308,
+      "num_tokens": 134579651.0,
+      "step": 1258
+    },
+    {
+      "epoch": 2.8688711516533636,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.273749880061118e-06,
+      "loss": 0.599,
+      "mean_token_accuracy": 0.8750852793455124,
+      "num_tokens": 134686349.0,
+      "step": 1259
+    },
+    {
+      "epoch": 2.8711516533637402,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.272422100093542e-06,
+      "loss": 0.6092,
+      "mean_token_accuracy": 0.8773278295993805,
+      "num_tokens": 134793245.0,
+      "step": 1260
+    },
+    {
+      "epoch": 2.8734321550741164,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.271093314133401e-06,
+      "loss": 0.6051,
+      "mean_token_accuracy": 0.8744064420461655,
+      "num_tokens": 134900346.0,
+      "step": 1261
+    },
+    {
+      "epoch": 2.8757126567844926,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.269763522934888e-06,
+      "loss": 0.5923,
+      "mean_token_accuracy": 0.8789143562316895,
+      "num_tokens": 135007561.0,
+      "step": 1262
+    },
+    {
+      "epoch": 2.877993158494869,
+      "grad_norm": 7.625,
+      "learning_rate": 4.268432727252765e-06,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.8737368434667587,
+      "num_tokens": 135115241.0,
+      "step": 1263
+    },
+    {
+      "epoch": 2.880273660205245,
+      "grad_norm": 5.4375,
+      "learning_rate": 4.2671009278423665e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8808005452156067,
+      "num_tokens": 135221997.0,
+      "step": 1264
+    },
+    {
+      "epoch": 2.882554161915621,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.265768125459597e-06,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.879405215382576,
+      "num_tokens": 135329159.0,
+      "step": 1265
+    },
+    {
+      "epoch": 2.884834663625998,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.264434320860929e-06,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.8776859790086746,
+      "num_tokens": 135436122.0,
+      "step": 1266
+    },
+    {
+      "epoch": 2.887115165336374,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.2630995148034044e-06,
+      "loss": 0.6147,
+      "mean_token_accuracy": 0.8750355541706085,
+      "num_tokens": 135542672.0,
+      "step": 1267
+    },
+    {
+      "epoch": 2.88939566704675,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.261763708044633e-06,
+      "loss": 0.5957,
+      "mean_token_accuracy": 0.8799539804458618,
+      "num_tokens": 135649649.0,
+      "step": 1268
+    },
+    {
+      "epoch": 2.8916761687571264,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.2604269013427925e-06,
+      "loss": 0.5871,
+      "mean_token_accuracy": 0.8781967163085938,
+      "num_tokens": 135756953.0,
+      "step": 1269
+    },
+    {
+      "epoch": 2.893956670467503,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.25908909545663e-06,
+      "loss": 0.6056,
+      "mean_token_accuracy": 0.8765884786844254,
+      "num_tokens": 135863953.0,
+      "step": 1270
+    },
+    {
+      "epoch": 2.8962371721778792,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.257750291145457e-06,
+      "loss": 0.611,
+      "mean_token_accuracy": 0.8713682293891907,
+      "num_tokens": 135970439.0,
+      "step": 1271
+    },
+    {
+      "epoch": 2.8985176738882554,
+      "grad_norm": 2.125,
+      "learning_rate": 4.256410489169154e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8807791769504547,
+      "num_tokens": 136077645.0,
+      "step": 1272
+    },
+    {
+      "epoch": 2.9007981755986316,
+      "grad_norm": 5.125,
+      "learning_rate": 4.255069690288166e-06,
+      "loss": 0.6053,
+      "mean_token_accuracy": 0.8755863457918167,
+      "num_tokens": 136184852.0,
+      "step": 1273
+    },
+    {
+      "epoch": 2.903078677309008,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.253727895263504e-06,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8805266171693802,
+      "num_tokens": 136292297.0,
+      "step": 1274
+    },
+    {
+      "epoch": 2.905359179019384,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.252385104856746e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8784385919570923,
+      "num_tokens": 136399187.0,
+      "step": 1275
+    },
+    {
+      "epoch": 2.9076396807297606,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.251041319830034e-06,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8847327828407288,
+      "num_tokens": 136506050.0,
+      "step": 1276
+    },
+    {
+      "epoch": 2.909920182440137,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.249696540946074e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8823814243078232,
+      "num_tokens": 136613948.0,
+      "step": 1277
+    },
+    {
+      "epoch": 2.912200684150513,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.248350768968136e-06,
+      "loss": 0.5926,
+      "mean_token_accuracy": 0.8782215118408203,
+      "num_tokens": 136720344.0,
+      "step": 1278
+    },
+    {
+      "epoch": 2.9144811858608897,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.247004004660055e-06,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8780831098556519,
+      "num_tokens": 136827167.0,
+      "step": 1279
+    },
+    {
+      "epoch": 2.916761687571266,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.245656248786228e-06,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.880301833152771,
+      "num_tokens": 136933946.0,
+      "step": 1280
+    },
+    {
+      "epoch": 2.919042189281642,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.2443075021116166e-06,
+      "loss": 0.5953,
+      "mean_token_accuracy": 0.8767362087965012,
+      "num_tokens": 137040549.0,
+      "step": 1281
+    },
+    {
+      "epoch": 2.9213226909920182,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.242957765401741e-06,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8834853321313858,
+      "num_tokens": 137147698.0,
+      "step": 1282
+    },
+    {
+      "epoch": 2.9236031927023944,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.241607039422687e-06,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8839240074157715,
+      "num_tokens": 137254610.0,
+      "step": 1283
+    },
+    {
+      "epoch": 2.9258836944127706,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.2402553249411e-06,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.8824329972267151,
+      "num_tokens": 137361723.0,
+      "step": 1284
+    },
+    {
+      "epoch": 2.928164196123147,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.238902622724188e-06,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8801659643650055,
+      "num_tokens": 137469233.0,
+      "step": 1285
+    },
+    {
+      "epoch": 2.9304446978335235,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.237548933539718e-06,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.8831808865070343,
+      "num_tokens": 137575939.0,
+      "step": 1286
+    },
+    {
+      "epoch": 2.9327251995438997,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.236194258156019e-06,
+      "loss": 0.5652,
+      "mean_token_accuracy": 0.8819428384304047,
+      "num_tokens": 137683411.0,
+      "step": 1287
+    },
+    {
+      "epoch": 2.935005701254276,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.234838597341977e-06,
+      "loss": 0.5775,
+      "mean_token_accuracy": 0.8816139101982117,
+      "num_tokens": 137790631.0,
+      "step": 1288
+    },
+    {
+      "epoch": 2.9372862029646525,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.233481951867039e-06,
+      "loss": 0.6042,
+      "mean_token_accuracy": 0.8764592856168747,
+      "num_tokens": 137897306.0,
+      "step": 1289
+    },
+    {
+      "epoch": 2.9395667046750287,
+      "grad_norm": 2.375,
+      "learning_rate": 4.232124322501212e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8771606087684631,
+      "num_tokens": 138004247.0,
+      "step": 1290
+    },
+    {
+      "epoch": 2.941847206385405,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.230765710015058e-06,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8792049288749695,
+      "num_tokens": 138110738.0,
+      "step": 1291
+    },
+    {
+      "epoch": 2.944127708095781,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.229406115179703e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8832272589206696,
+      "num_tokens": 138218085.0,
+      "step": 1292
+    },
+    {
+      "epoch": 2.9464082098061573,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.228045538766823e-06,
+      "loss": 0.5946,
+      "mean_token_accuracy": 0.8802038431167603,
+      "num_tokens": 138324630.0,
+      "step": 1293
+    },
+    {
+      "epoch": 2.9486887115165334,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.226683981548656e-06,
+      "loss": 0.6021,
+      "mean_token_accuracy": 0.8761596828699112,
+      "num_tokens": 138431811.0,
+      "step": 1294
+    },
+    {
+      "epoch": 2.95096921322691,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.2253214442979975e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8778193593025208,
+      "num_tokens": 138539059.0,
+      "step": 1295
+    },
+    {
+      "epoch": 2.9532497149372863,
+      "grad_norm": 6.34375,
+      "learning_rate": 4.223957927788195e-06,
+      "loss": 0.6276,
+      "mean_token_accuracy": 0.8704663217067719,
+      "num_tokens": 138646166.0,
+      "step": 1296
+    },
+    {
+      "epoch": 2.9555302166476625,
+      "grad_norm": 7.0625,
+      "learning_rate": 4.222593432793155e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.879715234041214,
+      "num_tokens": 138753154.0,
+      "step": 1297
+    },
+    {
+      "epoch": 2.9578107183580387,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.2212279600873385e-06,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8822585344314575,
+      "num_tokens": 138860601.0,
+      "step": 1298
+    },
+    {
+      "epoch": 2.9600912200684153,
+      "grad_norm": 6.03125,
+      "learning_rate": 4.219861510445762e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8831232637166977,
+      "num_tokens": 138967531.0,
+      "step": 1299
+    },
+    {
+      "epoch": 2.9623717217787915,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.2184940846439946e-06,
+      "loss": 0.6037,
+      "mean_token_accuracy": 0.8782211989164352,
+      "num_tokens": 139074913.0,
+      "step": 1300
+    },
+    {
+      "epoch": 2.9646522234891677,
+      "grad_norm": 9.3125,
+      "learning_rate": 4.217125683458162e-06,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8801289498806,
+      "num_tokens": 139182021.0,
+      "step": 1301
+    },
+    {
+      "epoch": 2.966932725199544,
+      "grad_norm": 11.3125,
+      "learning_rate": 4.215756307664941e-06,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.8780060112476349,
+      "num_tokens": 139289235.0,
+      "step": 1302
+    },
+    {
+      "epoch": 2.96921322690992,
+      "grad_norm": 13.75,
+      "learning_rate": 4.214385958041565e-06,
+      "loss": 0.6209,
+      "mean_token_accuracy": 0.8715900182723999,
+      "num_tokens": 139396474.0,
+      "step": 1303
+    },
+    {
+      "epoch": 2.9714937286202963,
+      "grad_norm": 9.8125,
+      "learning_rate": 4.213014635365816e-06,
+      "loss": 0.614,
+      "mean_token_accuracy": 0.8730522990226746,
+      "num_tokens": 139502964.0,
+      "step": 1304
+    },
+    {
+      "epoch": 2.973774230330673,
+      "grad_norm": 7.125,
+      "learning_rate": 4.2116423404160316e-06,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8835666477680206,
+      "num_tokens": 139610473.0,
+      "step": 1305
+    },
+    {
+      "epoch": 2.976054732041049,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.210269073971098e-06,
+      "loss": 0.5923,
+      "mean_token_accuracy": 0.8774100840091705,
+      "num_tokens": 139718229.0,
+      "step": 1306
+    },
+    {
+      "epoch": 2.9783352337514253,
+      "grad_norm": 4.0,
+      "learning_rate": 4.208894836810457e-06,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.8828433156013489,
+      "num_tokens": 139825370.0,
+      "step": 1307
+    },
+    {
+      "epoch": 2.9806157354618015,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.207519629714099e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8787718713283539,
+      "num_tokens": 139932155.0,
+      "step": 1308
+    },
+    {
+      "epoch": 2.982896237172178,
+      "grad_norm": 6.28125,
+      "learning_rate": 4.206143453462562e-06,
+      "loss": 0.5967,
+      "mean_token_accuracy": 0.8779617547988892,
+      "num_tokens": 140039280.0,
+      "step": 1309
+    },
+    {
+      "epoch": 2.9851767388825543,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.204766308836941e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8793751299381256,
+      "num_tokens": 140146757.0,
+      "step": 1310
+    },
+    {
+      "epoch": 2.9874572405929305,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.203388196618874e-06,
+      "loss": 0.6091,
+      "mean_token_accuracy": 0.8730627149343491,
+      "num_tokens": 140253574.0,
+      "step": 1311
+    },
+    {
+      "epoch": 2.9897377423033067,
+      "grad_norm": 5.25,
+      "learning_rate": 4.202009117590552e-06,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8779786825180054,
+      "num_tokens": 140360312.0,
+      "step": 1312
+    },
+    {
+      "epoch": 2.992018244013683,
+      "grad_norm": 5.1875,
+      "learning_rate": 4.200629072534713e-06,
+      "loss": 0.5836,
+      "mean_token_accuracy": 0.8807083070278168,
+      "num_tokens": 140467414.0,
+      "step": 1313
+    },
+    {
+      "epoch": 2.994298745724059,
+      "grad_norm": 6.75,
+      "learning_rate": 4.1992480622346455e-06,
+      "loss": 0.5945,
+      "mean_token_accuracy": 0.8760360926389694,
+      "num_tokens": 140574002.0,
+      "step": 1314
+    },
+    {
+      "epoch": 2.9965792474344357,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.197866087474181e-06,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8772165775299072,
+      "num_tokens": 140680798.0,
+      "step": 1315
+    },
+    {
+      "epoch": 2.998859749144812,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.196483149037707e-06,
+      "loss": 0.5989,
+      "mean_token_accuracy": 0.8783533871173859,
+      "num_tokens": 140787588.0,
+      "step": 1316
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 7.1875,
+      "learning_rate": 4.195099247710147e-06,
+      "loss": 0.624,
+      "mean_token_accuracy": 0.8783180117607117,
+      "num_tokens": 140826696.0,
+      "step": 1317
+    },
+    {
+      "epoch": 3.002280501710376,
+      "grad_norm": 5.28125,
+      "learning_rate": 4.1937143842769805e-06,
+      "loss": 0.5899,
+      "mean_token_accuracy": 0.8779798746109009,
+      "num_tokens": 140933415.0,
+      "step": 1318
+    },
+    {
+      "epoch": 3.0045610034207524,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.192328559524227e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8831545263528824,
+      "num_tokens": 141041011.0,
+      "step": 1319
+    },
+    {
+      "epoch": 3.006841505131129,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.190941774238454e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8806227445602417,
+      "num_tokens": 141148541.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.006841505131129,
+      "eval_loss": 0.596246063709259,
+      "eval_mean_token_accuracy": 0.8780989914339304,
+      "eval_num_tokens": 141148541.0,
+      "eval_runtime": 58.6547,
+      "eval_samples_per_second": 142.955,
+      "eval_steps_per_second": 4.484,
+      "step": 1320
+    },
+    {
+      "epoch": 3.009122006841505,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.1895540292067765e-06,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8786788433790207,
+      "num_tokens": 141255428.0,
+      "step": 1321
+    },
+    {
+      "epoch": 3.0114025085518814,
+      "grad_norm": 5.5625,
+      "learning_rate": 4.18816532521685e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8818920701742172,
+      "num_tokens": 141362931.0,
+      "step": 1322
+    },
+    {
+      "epoch": 3.0136830102622576,
+      "grad_norm": 8.25,
+      "learning_rate": 4.1867756630568755e-06,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.8816585689783096,
+      "num_tokens": 141470783.0,
+      "step": 1323
+    },
+    {
+      "epoch": 3.015963511972634,
+      "grad_norm": 5.65625,
+      "learning_rate": 4.1853850435156e-06,
+      "loss": 0.5817,
+      "mean_token_accuracy": 0.8797392249107361,
+      "num_tokens": 141578103.0,
+      "step": 1324
+    },
+    {
+      "epoch": 3.0182440136830104,
+      "grad_norm": 4.125,
+      "learning_rate": 4.18399346738231e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.8776365965604782,
+      "num_tokens": 141684405.0,
+      "step": 1325
+    },
+    {
+      "epoch": 3.0205245153933866,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.18260093544684e-06,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8821319788694382,
+      "num_tokens": 141790905.0,
+      "step": 1326
+    },
+    {
+      "epoch": 3.022805017103763,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.181207448499562e-06,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8799286335706711,
+      "num_tokens": 141898196.0,
+      "step": 1327
+    },
+    {
+      "epoch": 3.025085518814139,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.179813007331394e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8794847130775452,
+      "num_tokens": 142005139.0,
+      "step": 1328
+    },
+    {
+      "epoch": 3.027366020524515,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.178417612733792e-06,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8797455281019211,
+      "num_tokens": 142111855.0,
+      "step": 1329
+    },
+    {
+      "epoch": 3.029646522234892,
+      "grad_norm": 6.53125,
+      "learning_rate": 4.177021265498757e-06,
+      "loss": 0.5931,
+      "mean_token_accuracy": 0.879165843129158,
+      "num_tokens": 142218523.0,
+      "step": 1330
+    },
+    {
+      "epoch": 3.031927023945268,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.1756239664188275e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8838147222995758,
+      "num_tokens": 142325909.0,
+      "step": 1331
+    },
+    {
+      "epoch": 3.034207525655644,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.1742257162870835e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8792033344507217,
+      "num_tokens": 142433088.0,
+      "step": 1332
+    },
+    {
+      "epoch": 3.0364880273660204,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.172826515897146e-06,
+      "loss": 0.6002,
+      "mean_token_accuracy": 0.8756905943155289,
+      "num_tokens": 142540035.0,
+      "step": 1333
+    },
+    {
+      "epoch": 3.0387685290763966,
+      "grad_norm": 4.59375,
+      "learning_rate": 4.171426366043172e-06,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8791181296110153,
+      "num_tokens": 142647392.0,
+      "step": 1334
+    },
+    {
+      "epoch": 3.0410490307867732,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.170025267519862e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8790037631988525,
+      "num_tokens": 142754270.0,
+      "step": 1335
+    },
+    {
+      "epoch": 3.0433295324971494,
+      "grad_norm": 5.28125,
+      "learning_rate": 4.168623221122451e-06,
+      "loss": 0.5575,
+      "mean_token_accuracy": 0.8844562470912933,
+      "num_tokens": 142861663.0,
+      "step": 1336
+    },
+    {
+      "epoch": 3.0456100342075256,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.167220227646713e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8789408653974533,
+      "num_tokens": 142968943.0,
+      "step": 1337
+    },
+    {
+      "epoch": 3.047890535917902,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.165816287888962e-06,
+      "loss": 0.5843,
+      "mean_token_accuracy": 0.879507377743721,
+      "num_tokens": 143076504.0,
+      "step": 1338
+    },
+    {
+      "epoch": 3.050171037628278,
+      "grad_norm": 2.5,
+      "learning_rate": 4.164411402646045e-06,
+      "loss": 0.5905,
+      "mean_token_accuracy": 0.8781361430883408,
+      "num_tokens": 143183423.0,
+      "step": 1339
+    },
+    {
+      "epoch": 3.0524515393386547,
+      "grad_norm": 5.46875,
+      "learning_rate": 4.163005572715348e-06,
+      "loss": 0.5969,
+      "mean_token_accuracy": 0.879531055688858,
+      "num_tokens": 143290366.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.054732041049031,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.161598798894795e-06,
+      "loss": 0.595,
+      "mean_token_accuracy": 0.8771509379148483,
+      "num_tokens": 143397090.0,
+      "step": 1341
+    },
+    {
+      "epoch": 3.057012542759407,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.160191081982841e-06,
+      "loss": 0.6039,
+      "mean_token_accuracy": 0.8763537108898163,
+      "num_tokens": 143504171.0,
+      "step": 1342
+    },
+    {
+      "epoch": 3.0592930444697832,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.15878242277848e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8811188638210297,
+      "num_tokens": 143612068.0,
+      "step": 1343
+    },
+    {
+      "epoch": 3.0615735461801594,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.157372822081241e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.8822892308235168,
+      "num_tokens": 143719411.0,
+      "step": 1344
+    },
+    {
+      "epoch": 3.063854047890536,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.155962280691184e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8773934096097946,
+      "num_tokens": 143826292.0,
+      "step": 1345
+    },
+    {
+      "epoch": 3.0661345496009123,
+      "grad_norm": 6.125,
+      "learning_rate": 4.154550799408906e-06,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8825222551822662,
+      "num_tokens": 143933142.0,
+      "step": 1346
+    },
+    {
+      "epoch": 3.0684150513112884,
+      "grad_norm": 4.125,
+      "learning_rate": 4.153138379035537e-06,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8797077685594559,
+      "num_tokens": 144040202.0,
+      "step": 1347
+    },
+    {
+      "epoch": 3.0706955530216646,
+      "grad_norm": 4.25,
+      "learning_rate": 4.1517250203727395e-06,
+      "loss": 0.5917,
+      "mean_token_accuracy": 0.8776738494634628,
+      "num_tokens": 144147431.0,
+      "step": 1348
+    },
+    {
+      "epoch": 3.072976054732041,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.150310724222708e-06,
+      "loss": 0.5766,
+      "mean_token_accuracy": 0.8820817023515701,
+      "num_tokens": 144254552.0,
+      "step": 1349
+    },
+    {
+      "epoch": 3.0752565564424175,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.14889549138817e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8843452036380768,
+      "num_tokens": 144362233.0,
+      "step": 1350
+    },
+    {
+      "epoch": 3.0775370581527937,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.147479322672383e-06,
+      "loss": 0.5932,
+      "mean_token_accuracy": 0.8773269057273865,
+      "num_tokens": 144469328.0,
+      "step": 1351
+    },
+    {
+      "epoch": 3.07981755986317,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.14606221887914e-06,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.882498249411583,
+      "num_tokens": 144576504.0,
+      "step": 1352
+    },
+    {
+      "epoch": 3.082098061573546,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.144644180812759e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8806255161762238,
+      "num_tokens": 144683928.0,
+      "step": 1353
+    },
+    {
+      "epoch": 3.0843785632839227,
+      "grad_norm": 2.875,
+      "learning_rate": 4.143225209278093e-06,
+      "loss": 0.6075,
+      "mean_token_accuracy": 0.8764054775238037,
+      "num_tokens": 144791032.0,
+      "step": 1354
+    },
+    {
+      "epoch": 3.086659064994299,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.141805305080521e-06,
+      "loss": 0.6094,
+      "mean_token_accuracy": 0.8736419975757599,
+      "num_tokens": 144897700.0,
+      "step": 1355
+    },
+    {
+      "epoch": 3.088939566704675,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.1403844690259544e-06,
+      "loss": 0.6007,
+      "mean_token_accuracy": 0.8770473003387451,
+      "num_tokens": 145004603.0,
+      "step": 1356
+    },
+    {
+      "epoch": 3.0912200684150513,
+      "grad_norm": 2.03125,
+      "learning_rate": 4.138962701920831e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8825189918279648,
+      "num_tokens": 145112050.0,
+      "step": 1357
+    },
+    {
+      "epoch": 3.0935005701254275,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.13754000457212e-06,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8764393627643585,
+      "num_tokens": 145218872.0,
+      "step": 1358
+    },
+    {
+      "epoch": 3.095781071835804,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.136116377787317e-06,
+      "loss": 0.6087,
+      "mean_token_accuracy": 0.8764399290084839,
+      "num_tokens": 145325751.0,
+      "step": 1359
+    },
+    {
+      "epoch": 3.0980615735461803,
+      "grad_norm": 2.5,
+      "learning_rate": 4.134691822374445e-06,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8835193067789078,
+      "num_tokens": 145432654.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.1003420752565565,
+      "grad_norm": 3.0,
+      "learning_rate": 4.1332663391420515e-06,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8790077865123749,
+      "num_tokens": 145539487.0,
+      "step": 1361
+    },
+    {
+      "epoch": 3.1026225769669327,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.131839928899217e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.880421444773674,
+      "num_tokens": 145647251.0,
+      "step": 1362
+    },
+    {
+      "epoch": 3.104903078677309,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.130412592455542e-06,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8812432438135147,
+      "num_tokens": 145754411.0,
+      "step": 1363
+    },
+    {
+      "epoch": 3.1071835803876855,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.128984330621157e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8790383189916611,
+      "num_tokens": 145861509.0,
+      "step": 1364
+    },
+    {
+      "epoch": 3.1094640820980617,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.127555144206713e-06,
+      "loss": 0.5979,
+      "mean_token_accuracy": 0.8779343664646149,
+      "num_tokens": 145968759.0,
+      "step": 1365
+    },
+    {
+      "epoch": 3.111744583808438,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.126125034023392e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.8722660690546036,
+      "num_tokens": 146075699.0,
+      "step": 1366
+    },
+    {
+      "epoch": 3.114025085518814,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.124694000882894e-06,
+      "loss": 0.6043,
+      "mean_token_accuracy": 0.8754329532384872,
+      "num_tokens": 146183526.0,
+      "step": 1367
+    },
+    {
+      "epoch": 3.1163055872291903,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.123262045597447e-06,
+      "loss": 0.601,
+      "mean_token_accuracy": 0.8771016895771027,
+      "num_tokens": 146290729.0,
+      "step": 1368
+    },
+    {
+      "epoch": 3.118586088939567,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.121829168979802e-06,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8831167221069336,
+      "num_tokens": 146397777.0,
+      "step": 1369
+    },
+    {
+      "epoch": 3.120866590649943,
+      "grad_norm": 6.4375,
+      "learning_rate": 4.120395371843231e-06,
+      "loss": 0.5964,
+      "mean_token_accuracy": 0.8783126771450043,
+      "num_tokens": 146504599.0,
+      "step": 1370
+    },
+    {
+      "epoch": 3.1231470923603193,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.11896065500153e-06,
+      "loss": 0.5848,
+      "mean_token_accuracy": 0.8771827965974808,
+      "num_tokens": 146611271.0,
+      "step": 1371
+    },
+    {
+      "epoch": 3.1254275940706955,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.117525019269016e-06,
+      "loss": 0.5836,
+      "mean_token_accuracy": 0.8794733434915543,
+      "num_tokens": 146718261.0,
+      "step": 1372
+    },
+    {
+      "epoch": 3.1277080957810717,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.116088465460529e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8812500983476639,
+      "num_tokens": 146825123.0,
+      "step": 1373
+    },
+    {
+      "epoch": 3.1299885974914483,
+      "grad_norm": 3.25,
+      "learning_rate": 4.114650994391428e-06,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8858306407928467,
+      "num_tokens": 146932562.0,
+      "step": 1374
+    },
+    {
+      "epoch": 3.1322690992018245,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.113212606877596e-06,
+      "loss": 0.5975,
+      "mean_token_accuracy": 0.8762167394161224,
+      "num_tokens": 147039235.0,
+      "step": 1375
+    },
+    {
+      "epoch": 3.1345496009122007,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.111773303735432e-06,
+      "loss": 0.6084,
+      "mean_token_accuracy": 0.8757900893688202,
+      "num_tokens": 147146449.0,
+      "step": 1376
+    },
+    {
+      "epoch": 3.136830102622577,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.110333085781857e-06,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8781153112649918,
+      "num_tokens": 147253224.0,
+      "step": 1377
+    },
+    {
+      "epoch": 3.139110604332953,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.108891953834312e-06,
+      "loss": 0.6022,
+      "mean_token_accuracy": 0.8766528069972992,
+      "num_tokens": 147359809.0,
+      "step": 1378
+    },
+    {
+      "epoch": 3.1413911060433297,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.107449908710753e-06,
+      "loss": 0.5898,
+      "mean_token_accuracy": 0.8773281574249268,
+      "num_tokens": 147467375.0,
+      "step": 1379
+    },
+    {
+      "epoch": 3.143671607753706,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.106006951229661e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.8796637058258057,
+      "num_tokens": 147574319.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.145952109464082,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.104563082210028e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8800371438264847,
+      "num_tokens": 147681603.0,
+      "step": 1381
+    },
+    {
+      "epoch": 3.1482326111744583,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.1031183024713665e-06,
+      "loss": 0.5898,
+      "mean_token_accuracy": 0.8773266673088074,
+      "num_tokens": 147788610.0,
+      "step": 1382
+    },
+    {
+      "epoch": 3.1505131128848345,
+      "grad_norm": 4.5,
+      "learning_rate": 4.101672612833706e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8835958689451218,
+      "num_tokens": 147895321.0,
+      "step": 1383
+    },
+    {
+      "epoch": 3.152793614595211,
+      "grad_norm": 4.75,
+      "learning_rate": 4.100226014117592e-06,
+      "loss": 0.6047,
+      "mean_token_accuracy": 0.8747086226940155,
+      "num_tokens": 148002099.0,
+      "step": 1384
+    },
+    {
+      "epoch": 3.1550741163055873,
+      "grad_norm": 4.0,
+      "learning_rate": 4.098778507144086e-06,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.883410856127739,
+      "num_tokens": 148109199.0,
+      "step": 1385
+    },
+    {
+      "epoch": 3.1573546180159635,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.097330092734765e-06,
+      "loss": 0.5878,
+      "mean_token_accuracy": 0.8802912831306458,
+      "num_tokens": 148215632.0,
+      "step": 1386
+    },
+    {
+      "epoch": 3.1596351197263397,
+      "grad_norm": 6.375,
+      "learning_rate": 4.09588077171172e-06,
+      "loss": 0.5637,
+      "mean_token_accuracy": 0.8851134181022644,
+      "num_tokens": 148323133.0,
+      "step": 1387
+    },
+    {
+      "epoch": 3.161915621436716,
+      "grad_norm": 6.21875,
+      "learning_rate": 4.094430544897559e-06,
+      "loss": 0.5913,
+      "mean_token_accuracy": 0.8810424953699112,
+      "num_tokens": 148429831.0,
+      "step": 1388
+    },
+    {
+      "epoch": 3.1641961231470925,
+      "grad_norm": 6.0625,
+      "learning_rate": 4.092979413115404e-06,
+      "loss": 0.586,
+      "mean_token_accuracy": 0.8824973404407501,
+      "num_tokens": 148536848.0,
+      "step": 1389
+    },
+    {
+      "epoch": 3.1664766248574687,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.091527377188886e-06,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8795595914125443,
+      "num_tokens": 148644319.0,
+      "step": 1390
+    },
+    {
+      "epoch": 3.168757126567845,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.090074437942155e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.879493236541748,
+      "num_tokens": 148751953.0,
+      "step": 1391
+    },
+    {
+      "epoch": 3.171037628278221,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.088620596199872e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8819226622581482,
+      "num_tokens": 148859244.0,
+      "step": 1392
+    },
+    {
+      "epoch": 3.1733181299885973,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.087165852787206e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8783839493989944,
+      "num_tokens": 148966518.0,
+      "step": 1393
+    },
+    {
+      "epoch": 3.175598631698974,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.085710208529844e-06,
+      "loss": 0.5879,
+      "mean_token_accuracy": 0.8783328980207443,
+      "num_tokens": 149074306.0,
+      "step": 1394
+    },
+    {
+      "epoch": 3.17787913340935,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.084253664253981e-06,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8816855847835541,
+      "num_tokens": 149181408.0,
+      "step": 1395
+    },
+    {
+      "epoch": 3.1801596351197263,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.082796220786324e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8770945519208908,
+      "num_tokens": 149288874.0,
+      "step": 1396
+    },
+    {
+      "epoch": 3.1824401368301025,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.081337878954088e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8788999617099762,
+      "num_tokens": 149395972.0,
+      "step": 1397
+    },
+    {
+      "epoch": 3.1847206385404787,
+      "grad_norm": 2.25,
+      "learning_rate": 4.079878639585002e-06,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.884684830904007,
+      "num_tokens": 149503225.0,
+      "step": 1398
+    },
+    {
+      "epoch": 3.1870011402508553,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.0784185035072996e-06,
+      "loss": 0.593,
+      "mean_token_accuracy": 0.8796551078557968,
+      "num_tokens": 149609665.0,
+      "step": 1399
+    },
+    {
+      "epoch": 3.1892816419612315,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.076957471549728e-06,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8772178441286087,
+      "num_tokens": 149716324.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.1915621436716077,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.0754955445415405e-06,
+      "loss": 0.5978,
+      "mean_token_accuracy": 0.8765067309141159,
+      "num_tokens": 149823176.0,
+      "step": 1401
+    },
+    {
+      "epoch": 3.193842645381984,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.074032723312497e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8809285014867783,
+      "num_tokens": 149930057.0,
+      "step": 1402
+    },
+    {
+      "epoch": 3.19612314709236,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.072569008692868e-06,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8809255510568619,
+      "num_tokens": 150036837.0,
+      "step": 1403
+    },
+    {
+      "epoch": 3.1984036488027368,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.071104401513429e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8793366551399231,
+      "num_tokens": 150144003.0,
+      "step": 1404
+    },
+    {
+      "epoch": 3.200684150513113,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.069638902605464e-06,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.8811514675617218,
+      "num_tokens": 150252100.0,
+      "step": 1405
+    },
+    {
+      "epoch": 3.202964652223489,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.06817251280076e-06,
+      "loss": 0.6003,
+      "mean_token_accuracy": 0.8807378858327866,
+      "num_tokens": 150359077.0,
+      "step": 1406
+    },
+    {
+      "epoch": 3.2052451539338653,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.0667052329316125e-06,
+      "loss": 0.5956,
+      "mean_token_accuracy": 0.8794204890727997,
+      "num_tokens": 150466473.0,
+      "step": 1407
+    },
+    {
+      "epoch": 3.2075256556442415,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.0652370638308215e-06,
+      "loss": 0.603,
+      "mean_token_accuracy": 0.8787625581026077,
+      "num_tokens": 150572929.0,
+      "step": 1408
+    },
+    {
+      "epoch": 3.209806157354618,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.063768006331691e-06,
+      "loss": 0.598,
+      "mean_token_accuracy": 0.874964028596878,
+      "num_tokens": 150679963.0,
+      "step": 1409
+    },
+    {
+      "epoch": 3.2120866590649944,
+      "grad_norm": 6.4375,
+      "learning_rate": 4.06229806126803e-06,
+      "loss": 0.5632,
+      "mean_token_accuracy": 0.882831260561943,
+      "num_tokens": 150787240.0,
+      "step": 1410
+    },
+    {
+      "epoch": 3.2143671607753705,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.06082722947415e-06,
+      "loss": 0.5531,
+      "mean_token_accuracy": 0.8842821419239044,
+      "num_tokens": 150894724.0,
+      "step": 1411
+    },
+    {
+      "epoch": 3.2166476624857467,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.059355511784868e-06,
+      "loss": 0.5953,
+      "mean_token_accuracy": 0.8783928453922272,
+      "num_tokens": 151001455.0,
+      "step": 1412
+    },
+    {
+      "epoch": 3.2189281641961234,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.057882909035503e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8786003589630127,
+      "num_tokens": 151108702.0,
+      "step": 1413
+    },
+    {
+      "epoch": 3.2212086659064996,
+      "grad_norm": 4.75,
+      "learning_rate": 4.0564094220618735e-06,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8806807845830917,
+      "num_tokens": 151215645.0,
+      "step": 1414
+    },
+    {
+      "epoch": 3.2234891676168758,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.054935051700305e-06,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8828134685754776,
+      "num_tokens": 151322626.0,
+      "step": 1415
+    },
+    {
+      "epoch": 3.225769669327252,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.053459798787619e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.8766085356473923,
+      "num_tokens": 151429877.0,
+      "step": 1416
+    },
+    {
+      "epoch": 3.228050171037628,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.0519836641611425e-06,
+      "loss": 0.5851,
+      "mean_token_accuracy": 0.8795545697212219,
+      "num_tokens": 151536994.0,
+      "step": 1417
+    },
+    {
+      "epoch": 3.2303306727480043,
+      "grad_norm": 5.6875,
+      "learning_rate": 4.050506648658701e-06,
+      "loss": 0.5921,
+      "mean_token_accuracy": 0.8756987452507019,
+      "num_tokens": 151643935.0,
+      "step": 1418
+    },
+    {
+      "epoch": 3.232611174458381,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.049028753118619e-06,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.8784631341695786,
+      "num_tokens": 151751200.0,
+      "step": 1419
+    },
+    {
+      "epoch": 3.234891676168757,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.047549978379721e-06,
+      "loss": 0.6032,
+      "mean_token_accuracy": 0.8733428716659546,
+      "num_tokens": 151858059.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.2371721778791334,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.046070325281333e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8788390904664993,
+      "num_tokens": 151965037.0,
+      "step": 1421
+    },
+    {
+      "epoch": 3.2394526795895096,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.044589794663275e-06,
+      "loss": 0.6012,
+      "mean_token_accuracy": 0.8751859962940216,
+      "num_tokens": 152072452.0,
+      "step": 1422
+    },
+    {
+      "epoch": 3.241733181299886,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.04310838736587e-06,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8823445439338684,
+      "num_tokens": 152180256.0,
+      "step": 1423
+    },
+    {
+      "epoch": 3.2440136830102624,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.041626104229937e-06,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8832966238260269,
+      "num_tokens": 152287148.0,
+      "step": 1424
+    },
+    {
+      "epoch": 3.2462941847206386,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.0401429460967864e-06,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8766773343086243,
+      "num_tokens": 152393325.0,
+      "step": 1425
+    },
+    {
+      "epoch": 3.2485746864310148,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.038658913808235e-06,
+      "loss": 0.595,
+      "mean_token_accuracy": 0.877457782626152,
+      "num_tokens": 152500677.0,
+      "step": 1426
+    },
+    {
+      "epoch": 3.250855188141391,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.037174008206589e-06,
+      "loss": 0.5955,
+      "mean_token_accuracy": 0.8779526948928833,
+      "num_tokens": 152606926.0,
+      "step": 1427
+    },
+    {
+      "epoch": 3.253135689851767,
+      "grad_norm": 2.875,
+      "learning_rate": 4.035688230134651e-06,
+      "loss": 0.5891,
+      "mean_token_accuracy": 0.8811050057411194,
+      "num_tokens": 152713975.0,
+      "step": 1428
+    },
+    {
+      "epoch": 3.255416191562144,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.034201580435723e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.8812618553638458,
+      "num_tokens": 152820865.0,
+      "step": 1429
+    },
+    {
+      "epoch": 3.25769669327252,
+      "grad_norm": 3.0,
+      "learning_rate": 4.0327140599535954e-06,
+      "loss": 0.6084,
+      "mean_token_accuracy": 0.8711613863706589,
+      "num_tokens": 152927618.0,
+      "step": 1430
+    },
+    {
+      "epoch": 3.259977194982896,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.031225669532558e-06,
+      "loss": 0.5991,
+      "mean_token_accuracy": 0.8751797676086426,
+      "num_tokens": 153034150.0,
+      "step": 1431
+    },
+    {
+      "epoch": 3.2622576966932724,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.029736410017392e-06,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8789128512144089,
+      "num_tokens": 153141189.0,
+      "step": 1432
+    },
+    {
+      "epoch": 3.264538198403649,
+      "grad_norm": 4.5,
+      "learning_rate": 4.028246282253373e-06,
+      "loss": 0.5939,
+      "mean_token_accuracy": 0.8788965493440628,
+      "num_tokens": 153247856.0,
+      "step": 1433
+    },
+    {
+      "epoch": 3.266818700114025,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.026755287086267e-06,
+      "loss": 0.5809,
+      "mean_token_accuracy": 0.8801927268505096,
+      "num_tokens": 153355189.0,
+      "step": 1434
+    },
+    {
+      "epoch": 3.2690992018244014,
+      "grad_norm": 6.34375,
+      "learning_rate": 4.025263425362335e-06,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8794397115707397,
+      "num_tokens": 153461890.0,
+      "step": 1435
+    },
+    {
+      "epoch": 3.2713797035347776,
+      "grad_norm": 3.953125,
+      "learning_rate": 4.0237706979283306e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8837473094463348,
+      "num_tokens": 153568765.0,
+      "step": 1436
+    },
+    {
+      "epoch": 3.2736602052451538,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.022277105631495e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.882627084851265,
+      "num_tokens": 153675854.0,
+      "step": 1437
+    },
+    {
+      "epoch": 3.27594070695553,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.020782649319563e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8796237707138062,
+      "num_tokens": 153783094.0,
+      "step": 1438
+    },
+    {
+      "epoch": 3.2782212086659066,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.019287329840759e-06,
+      "loss": 0.5933,
+      "mean_token_accuracy": 0.8813715130090714,
+      "num_tokens": 153889764.0,
+      "step": 1439
+    },
+    {
+      "epoch": 3.280501710376283,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.017791148043797e-06,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.882407933473587,
+      "num_tokens": 153996810.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.282782212086659,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.016294104777883e-06,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.8812000006437302,
+      "num_tokens": 154103786.0,
+      "step": 1441
+    },
+    {
+      "epoch": 3.285062713797035,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.0147962008927065e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8824433386325836,
+      "num_tokens": 154211857.0,
+      "step": 1442
+    },
+    {
+      "epoch": 3.287343215507412,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.013297437238452e-06,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8849389404058456,
+      "num_tokens": 154318483.0,
+      "step": 1443
+    },
+    {
+      "epoch": 3.289623717217788,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.011797814665787e-06,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8829954713582993,
+      "num_tokens": 154425544.0,
+      "step": 1444
+    },
+    {
+      "epoch": 3.291904218928164,
+      "grad_norm": 2.375,
+      "learning_rate": 4.010297334025869e-06,
+      "loss": 0.5487,
+      "mean_token_accuracy": 0.8852455317974091,
+      "num_tokens": 154532774.0,
+      "step": 1445
+    },
+    {
+      "epoch": 3.2941847206385404,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.008795996170341e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8847774416208267,
+      "num_tokens": 154639810.0,
+      "step": 1446
+    },
+    {
+      "epoch": 3.2964652223489166,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.0072938019513345e-06,
+      "loss": 0.5896,
+      "mean_token_accuracy": 0.8774778991937637,
+      "num_tokens": 154747419.0,
+      "step": 1447
+    },
+    {
+      "epoch": 3.2987457240592932,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.0057907522214646e-06,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8795448988676071,
+      "num_tokens": 154855354.0,
+      "step": 1448
+    },
+    {
+      "epoch": 3.3010262257696694,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.004286847833835e-06,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8828533291816711,
+      "num_tokens": 154962700.0,
+      "step": 1449
+    },
+    {
+      "epoch": 3.3033067274800456,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.002782089642031e-06,
+      "loss": 0.5988,
+      "mean_token_accuracy": 0.8755003958940506,
+      "num_tokens": 155069571.0,
+      "step": 1450
+    },
+    {
+      "epoch": 3.305587229190422,
+      "grad_norm": 2.375,
+      "learning_rate": 4.001276478500127e-06,
+      "loss": 0.5454,
+      "mean_token_accuracy": 0.8879126906394958,
+      "num_tokens": 155177593.0,
+      "step": 1451
+    },
+    {
+      "epoch": 3.307867730900798,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.9997700152626755e-06,
+      "loss": 0.5674,
+      "mean_token_accuracy": 0.8830482810735703,
+      "num_tokens": 155284329.0,
+      "step": 1452
+    },
+    {
+      "epoch": 3.3101482326111746,
+      "grad_norm": 8.375,
+      "learning_rate": 3.9982627007847186e-06,
+      "loss": 0.5848,
+      "mean_token_accuracy": 0.8783923089504242,
+      "num_tokens": 155392445.0,
+      "step": 1453
+    },
+    {
+      "epoch": 3.312428734321551,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.996754535921777e-06,
+      "loss": 0.5655,
+      "mean_token_accuracy": 0.8857136368751526,
+      "num_tokens": 155499630.0,
+      "step": 1454
+    },
+    {
+      "epoch": 3.314709236031927,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.995245521529857e-06,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.8815815448760986,
+      "num_tokens": 155606583.0,
+      "step": 1455
+    },
+    {
+      "epoch": 3.316989737742303,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.993735658465446e-06,
+      "loss": 0.5964,
+      "mean_token_accuracy": 0.880055382847786,
+      "num_tokens": 155712881.0,
+      "step": 1456
+    },
+    {
+      "epoch": 3.3192702394526794,
+      "grad_norm": 4.375,
+      "learning_rate": 3.992224947585513e-06,
+      "loss": 0.6071,
+      "mean_token_accuracy": 0.8751248270273209,
+      "num_tokens": 155820166.0,
+      "step": 1457
+    },
+    {
+      "epoch": 3.321550741163056,
+      "grad_norm": 8.3125,
+      "learning_rate": 3.990713389747508e-06,
+      "loss": 0.5909,
+      "mean_token_accuracy": 0.8792251795530319,
+      "num_tokens": 155927043.0,
+      "step": 1458
+    },
+    {
+      "epoch": 3.3238312428734322,
+      "grad_norm": 6.21875,
+      "learning_rate": 3.989200985809362e-06,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8834533840417862,
+      "num_tokens": 156034044.0,
+      "step": 1459
+    },
+    {
+      "epoch": 3.3261117445838084,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.987687736629487e-06,
+      "loss": 0.5925,
+      "mean_token_accuracy": 0.8784044981002808,
+      "num_tokens": 156140733.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.3283922462941846,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.986173643066774e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.8763966262340546,
+      "num_tokens": 156247552.0,
+      "step": 1461
+    },
+    {
+      "epoch": 3.330672748004561,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.984658705980593e-06,
+      "loss": 0.6053,
+      "mean_token_accuracy": 0.8726420104503632,
+      "num_tokens": 156355329.0,
+      "step": 1462
+    },
+    {
+      "epoch": 3.3329532497149374,
+      "grad_norm": 5.34375,
+      "learning_rate": 3.983142926230792e-06,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.880921483039856,
+      "num_tokens": 156462676.0,
+      "step": 1463
+    },
+    {
+      "epoch": 3.3352337514253136,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.981626304677701e-06,
+      "loss": 0.5803,
+      "mean_token_accuracy": 0.8802833706140518,
+      "num_tokens": 156570120.0,
+      "step": 1464
+    },
+    {
+      "epoch": 3.33751425313569,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.980108842182121e-06,
+      "loss": 0.605,
+      "mean_token_accuracy": 0.8745223730802536,
+      "num_tokens": 156677584.0,
+      "step": 1465
+    },
+    {
+      "epoch": 3.339794754846066,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.978590539605338e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8785550892353058,
+      "num_tokens": 156785538.0,
+      "step": 1466
+    },
+    {
+      "epoch": 3.342075256556442,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.97707139780911e-06,
+      "loss": 0.5614,
+      "mean_token_accuracy": 0.8855056762695312,
+      "num_tokens": 156893407.0,
+      "step": 1467
+    },
+    {
+      "epoch": 3.344355758266819,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.975551417655673e-06,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8805434554815292,
+      "num_tokens": 157000135.0,
+      "step": 1468
+    },
+    {
+      "epoch": 3.346636259977195,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.974030600007737e-06,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.880170151591301,
+      "num_tokens": 157106985.0,
+      "step": 1469
+    },
+    {
+      "epoch": 3.3489167616875712,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.97250894572849e-06,
+      "loss": 0.5935,
+      "mean_token_accuracy": 0.8791963756084442,
+      "num_tokens": 157213948.0,
+      "step": 1470
+    },
+    {
+      "epoch": 3.3511972633979474,
+      "grad_norm": 5.0625,
+      "learning_rate": 3.970986455681593e-06,
+      "loss": 0.5577,
+      "mean_token_accuracy": 0.8822743445634842,
+      "num_tokens": 157321436.0,
+      "step": 1471
+    },
+    {
+      "epoch": 3.353477765108324,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.969463130731183e-06,
+      "loss": 0.6327,
+      "mean_token_accuracy": 0.8710167407989502,
+      "num_tokens": 157427856.0,
+      "step": 1472
+    },
+    {
+      "epoch": 3.3557582668187003,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.967938971741869e-06,
+      "loss": 0.597,
+      "mean_token_accuracy": 0.880034327507019,
+      "num_tokens": 157535511.0,
+      "step": 1473
+    },
+    {
+      "epoch": 3.3580387685290765,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.966413979578734e-06,
+      "loss": 0.5949,
+      "mean_token_accuracy": 0.8794209510087967,
+      "num_tokens": 157642677.0,
+      "step": 1474
+    },
+    {
+      "epoch": 3.3603192702394526,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.964888155107335e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8807871788740158,
+      "num_tokens": 157749664.0,
+      "step": 1475
+    },
+    {
+      "epoch": 3.362599771949829,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.963361499193699e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.8778406232595444,
+      "num_tokens": 157857160.0,
+      "step": 1476
+    },
+    {
+      "epoch": 3.364880273660205,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.9618340127043274e-06,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.881255105137825,
+      "num_tokens": 157964040.0,
+      "step": 1477
+    },
+    {
+      "epoch": 3.3671607753705817,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.960305696506192e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8831611126661301,
+      "num_tokens": 158071025.0,
+      "step": 1478
+    },
+    {
+      "epoch": 3.369441277080958,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.958776551466737e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8826898485422134,
+      "num_tokens": 158177813.0,
+      "step": 1479
+    },
+    {
+      "epoch": 3.371721778791334,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.957246578453873e-06,
+      "loss": 0.6079,
+      "mean_token_accuracy": 0.8766767531633377,
+      "num_tokens": 158284056.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.3740022805017102,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.955715778335984e-06,
+      "loss": 0.5648,
+      "mean_token_accuracy": 0.882177472114563,
+      "num_tokens": 158390740.0,
+      "step": 1481
+    },
+    {
+      "epoch": 3.376282782212087,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.954184151981924e-06,
+      "loss": 0.5948,
+      "mean_token_accuracy": 0.8798738569021225,
+      "num_tokens": 158497577.0,
+      "step": 1482
+    },
+    {
+      "epoch": 3.378563283922463,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.952651700261012e-06,
+      "loss": 0.5969,
+      "mean_token_accuracy": 0.8804901242256165,
+      "num_tokens": 158604558.0,
+      "step": 1483
+    },
+    {
+      "epoch": 3.3808437856328393,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.95111842404304e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8816994279623032,
+      "num_tokens": 158712603.0,
+      "step": 1484
+    },
+    {
+      "epoch": 3.3831242873432155,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.949584324198266e-06,
+      "loss": 0.5931,
+      "mean_token_accuracy": 0.8765831738710403,
+      "num_tokens": 158819917.0,
+      "step": 1485
+    },
+    {
+      "epoch": 3.3854047890535917,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.948049401597414e-06,
+      "loss": 0.576,
+      "mean_token_accuracy": 0.881806880235672,
+      "num_tokens": 158927113.0,
+      "step": 1486
+    },
+    {
+      "epoch": 3.387685290763968,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.946513657111678e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.881575807929039,
+      "num_tokens": 159034584.0,
+      "step": 1487
+    },
+    {
+      "epoch": 3.3899657924743445,
+      "grad_norm": 2.40625,
+      "learning_rate": 3.944977091612716e-06,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.880961686372757,
+      "num_tokens": 159141695.0,
+      "step": 1488
+    },
+    {
+      "epoch": 3.3922462941847207,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.943439705972654e-06,
+      "loss": 0.604,
+      "mean_token_accuracy": 0.8772746175527573,
+      "num_tokens": 159248763.0,
+      "step": 1489
+    },
+    {
+      "epoch": 3.394526795895097,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.94190150106408e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8807346522808075,
+      "num_tokens": 159355766.0,
+      "step": 1490
+    },
+    {
+      "epoch": 3.396807297605473,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.9403624777600526e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8799644559621811,
+      "num_tokens": 159462660.0,
+      "step": 1491
+    },
+    {
+      "epoch": 3.3990877993158497,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.938822636934089e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.8790238797664642,
+      "num_tokens": 159569562.0,
+      "step": 1492
+    },
+    {
+      "epoch": 3.401368301026226,
+      "grad_norm": 7.1875,
+      "learning_rate": 3.937281979460175e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8814513683319092,
+      "num_tokens": 159676589.0,
+      "step": 1493
+    },
+    {
+      "epoch": 3.403648802736602,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.9357405062127565e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8827214390039444,
+      "num_tokens": 159783562.0,
+      "step": 1494
+    },
+    {
+      "epoch": 3.4059293044469783,
+      "grad_norm": 2.875,
+      "learning_rate": 3.934198218066745e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.882139042019844,
+      "num_tokens": 159891070.0,
+      "step": 1495
+    },
+    {
+      "epoch": 3.4082098061573545,
+      "grad_norm": 6.34375,
+      "learning_rate": 3.932655115897513e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8820941895246506,
+      "num_tokens": 159997890.0,
+      "step": 1496
+    },
+    {
+      "epoch": 3.4104903078677307,
+      "grad_norm": 4.0,
+      "learning_rate": 3.9311112005808955e-06,
+      "loss": 0.6003,
+      "mean_token_accuracy": 0.8780702501535416,
+      "num_tokens": 160104922.0,
+      "step": 1497
+    },
+    {
+      "epoch": 3.4127708095781073,
+      "grad_norm": 6.15625,
+      "learning_rate": 3.92956647299319e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8826514631509781,
+      "num_tokens": 160211984.0,
+      "step": 1498
+    },
+    {
+      "epoch": 3.4150513112884835,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.928020934011153e-06,
+      "loss": 0.6038,
+      "mean_token_accuracy": 0.876240611076355,
+      "num_tokens": 160319031.0,
+      "step": 1499
+    },
+    {
+      "epoch": 3.4173318129988597,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.926474584512002e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.8836843520402908,
+      "num_tokens": 160425892.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.419612314709236,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.924927425373417e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8801800161600113,
+      "num_tokens": 160533225.0,
+      "step": 1501
+    },
+    {
+      "epoch": 3.4218928164196125,
+      "grad_norm": 7.0625,
+      "learning_rate": 3.9233794574735345e-06,
+      "loss": 0.5941,
+      "mean_token_accuracy": 0.8788256347179413,
+      "num_tokens": 160640613.0,
+      "step": 1502
+    },
+    {
+      "epoch": 3.4241733181299887,
+      "grad_norm": 5.8125,
+      "learning_rate": 3.921830681690951e-06,
+      "loss": 0.6065,
+      "mean_token_accuracy": 0.8749442100524902,
+      "num_tokens": 160746763.0,
+      "step": 1503
+    },
+    {
+      "epoch": 3.426453819840365,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.920281098904722e-06,
+      "loss": 0.5889,
+      "mean_token_accuracy": 0.8785606771707535,
+      "num_tokens": 160854121.0,
+      "step": 1504
+    },
+    {
+      "epoch": 3.428734321550741,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.918730709994361e-06,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8799080401659012,
+      "num_tokens": 160961324.0,
+      "step": 1505
+    },
+    {
+      "epoch": 3.4310148232611173,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.91717951583984e-06,
+      "loss": 0.5808,
+      "mean_token_accuracy": 0.8835075944662094,
+      "num_tokens": 161068109.0,
+      "step": 1506
+    },
+    {
+      "epoch": 3.433295324971494,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.915627517321584e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8793487548828125,
+      "num_tokens": 161175558.0,
+      "step": 1507
+    },
+    {
+      "epoch": 3.43557582668187,
+      "grad_norm": 5.03125,
+      "learning_rate": 3.914074715320479e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8839425444602966,
+      "num_tokens": 161282715.0,
+      "step": 1508
+    },
+    {
+      "epoch": 3.4378563283922463,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.912521110717866e-06,
+      "loss": 0.5878,
+      "mean_token_accuracy": 0.879220187664032,
+      "num_tokens": 161390083.0,
+      "step": 1509
+    },
+    {
+      "epoch": 3.4401368301026225,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.9109667043955405e-06,
+      "loss": 0.5901,
+      "mean_token_accuracy": 0.8790208101272583,
+      "num_tokens": 161497044.0,
+      "step": 1510
+    },
+    {
+      "epoch": 3.4424173318129987,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.909411497235752e-06,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8846622407436371,
+      "num_tokens": 161603896.0,
+      "step": 1511
+    },
+    {
+      "epoch": 3.4446978335233753,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.907855490121208e-06,
+      "loss": 0.5878,
+      "mean_token_accuracy": 0.8800790905952454,
+      "num_tokens": 161710931.0,
+      "step": 1512
+    },
+    {
+      "epoch": 3.4469783352337515,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.906298683935068e-06,
+      "loss": 0.5958,
+      "mean_token_accuracy": 0.8813421279191971,
+      "num_tokens": 161817975.0,
+      "step": 1513
+    },
+    {
+      "epoch": 3.4492588369441277,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.904741079560944e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8818980902433395,
+      "num_tokens": 161925410.0,
+      "step": 1514
+    },
+    {
+      "epoch": 3.451539338654504,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.903182677882904e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8834614157676697,
+      "num_tokens": 162032741.0,
+      "step": 1515
+    },
+    {
+      "epoch": 3.45381984036488,
+      "grad_norm": 3.875,
+      "learning_rate": 3.901623479785465e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8787531405687332,
+      "num_tokens": 162139480.0,
+      "step": 1516
+    },
+    {
+      "epoch": 3.4561003420752567,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.900063486153598e-06,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8862394094467163,
+      "num_tokens": 162246597.0,
+      "step": 1517
+    },
+    {
+      "epoch": 3.458380843785633,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.898502697872725e-06,
+      "loss": 0.5941,
+      "mean_token_accuracy": 0.8790966123342514,
+      "num_tokens": 162353845.0,
+      "step": 1518
+    },
+    {
+      "epoch": 3.460661345496009,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.896941115828721e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8806100636720657,
+      "num_tokens": 162460754.0,
+      "step": 1519
+    },
+    {
+      "epoch": 3.4629418472063853,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.895378740907908e-06,
+      "loss": 0.6153,
+      "mean_token_accuracy": 0.8727654367685318,
+      "num_tokens": 162567222.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.4652223489167615,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.89381557399706e-06,
+      "loss": 0.5973,
+      "mean_token_accuracy": 0.8774588257074356,
+      "num_tokens": 162674330.0,
+      "step": 1521
+    },
+    {
+      "epoch": 3.467502850627138,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.892251615983401e-06,
+      "loss": 0.587,
+      "mean_token_accuracy": 0.8811975121498108,
+      "num_tokens": 162781710.0,
+      "step": 1522
+    },
+    {
+      "epoch": 3.4697833523375143,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.890686867754604e-06,
+      "loss": 0.6,
+      "mean_token_accuracy": 0.8789321780204773,
+      "num_tokens": 162888560.0,
+      "step": 1523
+    },
+    {
+      "epoch": 3.4720638540478905,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.889121330198788e-06,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8869301974773407,
+      "num_tokens": 162995694.0,
+      "step": 1524
+    },
+    {
+      "epoch": 3.4743443557582667,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.887555004204524e-06,
+      "loss": 0.5906,
+      "mean_token_accuracy": 0.8777419775724411,
+      "num_tokens": 163103255.0,
+      "step": 1525
+    },
+    {
+      "epoch": 3.476624857468643,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.885987890660828e-06,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.8802089840173721,
+      "num_tokens": 163210670.0,
+      "step": 1526
+    },
+    {
+      "epoch": 3.4789053591790196,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.884419990457161e-06,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8770112693309784,
+      "num_tokens": 163317663.0,
+      "step": 1527
+    },
+    {
+      "epoch": 3.4811858608893957,
+      "grad_norm": 3.0,
+      "learning_rate": 3.882851304483436e-06,
+      "loss": 0.599,
+      "mean_token_accuracy": 0.8784664124250412,
+      "num_tokens": 163424638.0,
+      "step": 1528
+    },
+    {
+      "epoch": 3.483466362599772,
+      "grad_norm": 2.75,
+      "learning_rate": 3.881281833630007e-06,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.8907039016485214,
+      "num_tokens": 163532458.0,
+      "step": 1529
+    },
+    {
+      "epoch": 3.485746864310148,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.879711578787676e-06,
+      "loss": 0.5905,
+      "mean_token_accuracy": 0.8753378838300705,
+      "num_tokens": 163639976.0,
+      "step": 1530
+    },
+    {
+      "epoch": 3.4880273660205243,
+      "grad_norm": 2.625,
+      "learning_rate": 3.87814054084769e-06,
+      "loss": 0.5656,
+      "mean_token_accuracy": 0.8845688998699188,
+      "num_tokens": 163747328.0,
+      "step": 1531
+    },
+    {
+      "epoch": 3.490307867730901,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.8765687207017375e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8783632963895798,
+      "num_tokens": 163854482.0,
+      "step": 1532
+    },
+    {
+      "epoch": 3.492588369441277,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.874996119241956e-06,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8804601579904556,
+      "num_tokens": 163961669.0,
+      "step": 1533
+    },
+    {
+      "epoch": 3.4948688711516533,
+      "grad_norm": 3.125,
+      "learning_rate": 3.873422737360922e-06,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8807873427867889,
+      "num_tokens": 164068614.0,
+      "step": 1534
+    },
+    {
+      "epoch": 3.4971493728620295,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.871848575951658e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8857271820306778,
+      "num_tokens": 164175640.0,
+      "step": 1535
+    },
+    {
+      "epoch": 3.4994298745724057,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.8702736359076265e-06,
+      "loss": 0.5652,
+      "mean_token_accuracy": 0.8821373879909515,
+      "num_tokens": 164283025.0,
+      "step": 1536
+    },
+    {
+      "epoch": 3.5017103762827824,
+      "grad_norm": 2.140625,
+      "learning_rate": 3.868697918122733e-06,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.8805082738399506,
+      "num_tokens": 164389945.0,
+      "step": 1537
+    },
+    {
+      "epoch": 3.5039908779931586,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.867121423491325e-06,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8843076974153519,
+      "num_tokens": 164497358.0,
+      "step": 1538
+    },
+    {
+      "epoch": 3.5062713797035348,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.86554415290819e-06,
+      "loss": 0.5839,
+      "mean_token_accuracy": 0.8772299587726593,
+      "num_tokens": 164604547.0,
+      "step": 1539
+    },
+    {
+      "epoch": 3.508551881413911,
+      "grad_norm": 3.0,
+      "learning_rate": 3.8639661072685575e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8764625191688538,
+      "num_tokens": 164711483.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.508551881413911,
+      "eval_loss": 0.5934817790985107,
+      "eval_mean_token_accuracy": 0.8786315757058872,
+      "eval_num_tokens": 164711483.0,
+      "eval_runtime": 58.6311,
+      "eval_samples_per_second": 143.013,
+      "eval_steps_per_second": 4.486,
+      "step": 1540
+    },
+    {
+      "epoch": 3.5108323831242876,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.862387287468095e-06,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.884019061923027,
+      "num_tokens": 164818766.0,
+      "step": 1541
+    },
+    {
+      "epoch": 3.5131128848346638,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.860807694402909e-06,
+      "loss": 0.6091,
+      "mean_token_accuracy": 0.872483566403389,
+      "num_tokens": 164925458.0,
+      "step": 1542
+    },
+    {
+      "epoch": 3.51539338654504,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.859227328969547e-06,
+      "loss": 0.5638,
+      "mean_token_accuracy": 0.8828099370002747,
+      "num_tokens": 165033022.0,
+      "step": 1543
+    },
+    {
+      "epoch": 3.517673888255416,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.857646192064995e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.885229229927063,
+      "num_tokens": 165140269.0,
+      "step": 1544
+    },
+    {
+      "epoch": 3.5199543899657924,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.856064284586674e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.878631979227066,
+      "num_tokens": 165247172.0,
+      "step": 1545
+    },
+    {
+      "epoch": 3.5222348916761685,
+      "grad_norm": 4.84375,
+      "learning_rate": 3.854481607432445e-06,
+      "loss": 0.5869,
+      "mean_token_accuracy": 0.881737545132637,
+      "num_tokens": 165353974.0,
+      "step": 1546
+    },
+    {
+      "epoch": 3.524515393386545,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.852898161500605e-06,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8790680915117264,
+      "num_tokens": 165461254.0,
+      "step": 1547
+    },
+    {
+      "epoch": 3.5267958950969214,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.851313947689888e-06,
+      "loss": 0.6022,
+      "mean_token_accuracy": 0.8754421025514603,
+      "num_tokens": 165568029.0,
+      "step": 1548
+    },
+    {
+      "epoch": 3.5290763968072976,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.849728966899462e-06,
+      "loss": 0.5803,
+      "mean_token_accuracy": 0.8801588118076324,
+      "num_tokens": 165675255.0,
+      "step": 1549
+    },
+    {
+      "epoch": 3.5313568985176738,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.848143220028931e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8785509765148163,
+      "num_tokens": 165781381.0,
+      "step": 1550
+    },
+    {
+      "epoch": 3.5336374002280504,
+      "grad_norm": 2.203125,
+      "learning_rate": 3.846556707978337e-06,
+      "loss": 0.556,
+      "mean_token_accuracy": 0.8853710889816284,
+      "num_tokens": 165888356.0,
+      "step": 1551
+    },
+    {
+      "epoch": 3.5359179019384266,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.844969431648151e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8783566206693649,
+      "num_tokens": 165995028.0,
+      "step": 1552
+    },
+    {
+      "epoch": 3.538198403648803,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.843381391939281e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.8758636116981506,
+      "num_tokens": 166102173.0,
+      "step": 1553
+    },
+    {
+      "epoch": 3.540478905359179,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.841792589753067e-06,
+      "loss": 0.5566,
+      "mean_token_accuracy": 0.8812205046415329,
+      "num_tokens": 166209381.0,
+      "step": 1554
+    },
+    {
+      "epoch": 3.542759407069555,
+      "grad_norm": 5.65625,
+      "learning_rate": 3.840203025991285e-06,
+      "loss": 0.5516,
+      "mean_token_accuracy": 0.8851082473993301,
+      "num_tokens": 166316578.0,
+      "step": 1555
+    },
+    {
+      "epoch": 3.5450399087799314,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.838612701556138e-06,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.879833921790123,
+      "num_tokens": 166423915.0,
+      "step": 1556
+    },
+    {
+      "epoch": 3.547320410490308,
+      "grad_norm": 3.125,
+      "learning_rate": 3.837021617350266e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8793807029724121,
+      "num_tokens": 166530907.0,
+      "step": 1557
+    },
+    {
+      "epoch": 3.549600912200684,
+      "grad_norm": 2.375,
+      "learning_rate": 3.8354297742767345e-06,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8835282325744629,
+      "num_tokens": 166638263.0,
+      "step": 1558
+    },
+    {
+      "epoch": 3.5518814139110604,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.833837173239044e-06,
+      "loss": 0.5898,
+      "mean_token_accuracy": 0.8798348754644394,
+      "num_tokens": 166745468.0,
+      "step": 1559
+    },
+    {
+      "epoch": 3.5541619156214366,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.832243815141126e-06,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8829097151756287,
+      "num_tokens": 166852138.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.556442417331813,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.830649700887339e-06,
+      "loss": 0.6001,
+      "mean_token_accuracy": 0.8771270960569382,
+      "num_tokens": 166958904.0,
+      "step": 1561
+    },
+    {
+      "epoch": 3.5587229190421894,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.829054831382471e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8783409595489502,
+      "num_tokens": 167066230.0,
+      "step": 1562
+    },
+    {
+      "epoch": 3.5610034207525656,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.827459207531739e-06,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8838016241788864,
+      "num_tokens": 167173536.0,
+      "step": 1563
+    },
+    {
+      "epoch": 3.563283922462942,
+      "grad_norm": 2.625,
+      "learning_rate": 3.825862830240787e-06,
+      "loss": 0.6079,
+      "mean_token_accuracy": 0.8744715452194214,
+      "num_tokens": 167280709.0,
+      "step": 1564
+    },
+    {
+      "epoch": 3.565564424173318,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.82426570041569e-06,
+      "loss": 0.5893,
+      "mean_token_accuracy": 0.8770562261343002,
+      "num_tokens": 167387081.0,
+      "step": 1565
+    },
+    {
+      "epoch": 3.567844925883694,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.822667818962948e-06,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.8852861523628235,
+      "num_tokens": 167494025.0,
+      "step": 1566
+    },
+    {
+      "epoch": 3.570125427594071,
+      "grad_norm": 2.5,
+      "learning_rate": 3.821069186789486e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8804926127195358,
+      "num_tokens": 167600832.0,
+      "step": 1567
+    },
+    {
+      "epoch": 3.572405929304447,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.819469804802659e-06,
+      "loss": 0.5898,
+      "mean_token_accuracy": 0.8771549314260483,
+      "num_tokens": 167707990.0,
+      "step": 1568
+    },
+    {
+      "epoch": 3.574686431014823,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.8178696739102435e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8794214427471161,
+      "num_tokens": 167815132.0,
+      "step": 1569
+    },
+    {
+      "epoch": 3.5769669327251994,
+      "grad_norm": 5.65625,
+      "learning_rate": 3.816268795020443e-06,
+      "loss": 0.5548,
+      "mean_token_accuracy": 0.8854867666959763,
+      "num_tokens": 167922544.0,
+      "step": 1570
+    },
+    {
+      "epoch": 3.579247434435576,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.814667169041887e-06,
+      "loss": 0.6035,
+      "mean_token_accuracy": 0.8798968493938446,
+      "num_tokens": 168029140.0,
+      "step": 1571
+    },
+    {
+      "epoch": 3.581527936145952,
+      "grad_norm": 5.09375,
+      "learning_rate": 3.8130647968836254e-06,
+      "loss": 0.6082,
+      "mean_token_accuracy": 0.8785337060689926,
+      "num_tokens": 168136139.0,
+      "step": 1572
+    },
+    {
+      "epoch": 3.5838084378563284,
+      "grad_norm": 2.34375,
+      "learning_rate": 3.811461679455136e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8779959082603455,
+      "num_tokens": 168242818.0,
+      "step": 1573
+    },
+    {
+      "epoch": 3.5860889395667046,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.809857817666316e-06,
+      "loss": 0.5924,
+      "mean_token_accuracy": 0.8785370439291,
+      "num_tokens": 168350600.0,
+      "step": 1574
+    },
+    {
+      "epoch": 3.588369441277081,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.808253212427486e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.879870593547821,
+      "num_tokens": 168457735.0,
+      "step": 1575
+    },
+    {
+      "epoch": 3.590649942987457,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.8066478646493898e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8814902901649475,
+      "num_tokens": 168564725.0,
+      "step": 1576
+    },
+    {
+      "epoch": 3.5929304446978336,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.805041775243191e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.8772566318511963,
+      "num_tokens": 168671913.0,
+      "step": 1577
+    },
+    {
+      "epoch": 3.59521094640821,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.803434945120475e-06,
+      "loss": 0.5593,
+      "mean_token_accuracy": 0.8856948614120483,
+      "num_tokens": 168779191.0,
+      "step": 1578
+    },
+    {
+      "epoch": 3.597491448118586,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.801827375193249e-06,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8798210918903351,
+      "num_tokens": 168886143.0,
+      "step": 1579
+    },
+    {
+      "epoch": 3.5997719498289626,
+      "grad_norm": 6.9375,
+      "learning_rate": 3.8002190663739362e-06,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8814046531915665,
+      "num_tokens": 168993134.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.602052451539339,
+      "grad_norm": 8.6875,
+      "learning_rate": 3.798610019575384e-06,
+      "loss": 0.5846,
+      "mean_token_accuracy": 0.8799954205751419,
+      "num_tokens": 169099760.0,
+      "step": 1581
+    },
+    {
+      "epoch": 3.604332953249715,
+      "grad_norm": 6.90625,
+      "learning_rate": 3.7970002357108554e-06,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8843145817518234,
+      "num_tokens": 169206545.0,
+      "step": 1582
+    },
+    {
+      "epoch": 3.6066134549600912,
+      "grad_norm": 3.625,
+      "learning_rate": 3.7953897156940323e-06,
+      "loss": 0.5927,
+      "mean_token_accuracy": 0.8787747472524643,
+      "num_tokens": 169313229.0,
+      "step": 1583
+    },
+    {
+      "epoch": 3.6088939566704674,
+      "grad_norm": 2.75,
+      "learning_rate": 3.793778460439015e-06,
+      "loss": 0.5878,
+      "mean_token_accuracy": 0.8789673447608948,
+      "num_tokens": 169420044.0,
+      "step": 1584
+    },
+    {
+      "epoch": 3.6111744583808436,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.792166470860321e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8790552765130997,
+      "num_tokens": 169526883.0,
+      "step": 1585
+    },
+    {
+      "epoch": 3.61345496009122,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.790553747872885e-06,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8849265277385712,
+      "num_tokens": 169633699.0,
+      "step": 1586
+    },
+    {
+      "epoch": 3.6157354618015964,
+      "grad_norm": 4.96875,
+      "learning_rate": 3.788940292392056e-06,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8801223486661911,
+      "num_tokens": 169741250.0,
+      "step": 1587
+    },
+    {
+      "epoch": 3.6180159635119726,
+      "grad_norm": 5.59375,
+      "learning_rate": 3.787326105333601e-06,
+      "loss": 0.5694,
+      "mean_token_accuracy": 0.8809481114149094,
+      "num_tokens": 169848862.0,
+      "step": 1588
+    },
+    {
+      "epoch": 3.620296465222349,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.7857111876137017e-06,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8804655224084854,
+      "num_tokens": 169955712.0,
+      "step": 1589
+    },
+    {
+      "epoch": 3.6225769669327255,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.784095540148954e-06,
+      "loss": 0.5981,
+      "mean_token_accuracy": 0.8755791783332825,
+      "num_tokens": 170062534.0,
+      "step": 1590
+    },
+    {
+      "epoch": 3.6248574686431017,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.7824791638563674e-06,
+      "loss": 0.5882,
+      "mean_token_accuracy": 0.8761962950229645,
+      "num_tokens": 170169349.0,
+      "step": 1591
+    },
+    {
+      "epoch": 3.627137970353478,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.7808620596533675e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8823762983083725,
+      "num_tokens": 170276611.0,
+      "step": 1592
+    },
+    {
+      "epoch": 3.629418472063854,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.77924422845779e-06,
+      "loss": 0.5998,
+      "mean_token_accuracy": 0.873461440205574,
+      "num_tokens": 170383192.0,
+      "step": 1593
+    },
+    {
+      "epoch": 3.6316989737742302,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.7776256711878856e-06,
+      "loss": 0.5914,
+      "mean_token_accuracy": 0.8810142874717712,
+      "num_tokens": 170490101.0,
+      "step": 1594
+    },
+    {
+      "epoch": 3.6339794754846064,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.7760063887623155e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8790360242128372,
+      "num_tokens": 170596909.0,
+      "step": 1595
+    },
+    {
+      "epoch": 3.636259977194983,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.7743863821001538e-06,
+      "loss": 0.5571,
+      "mean_token_accuracy": 0.886503741145134,
+      "num_tokens": 170704691.0,
+      "step": 1596
+    },
+    {
+      "epoch": 3.6385404789053593,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.7727656521208843e-06,
+      "loss": 0.5991,
+      "mean_token_accuracy": 0.8780782669782639,
+      "num_tokens": 170811447.0,
+      "step": 1597
+    },
+    {
+      "epoch": 3.6408209806157354,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.771144199744402e-06,
+      "loss": 0.5773,
+      "mean_token_accuracy": 0.8789367228746414,
+      "num_tokens": 170918012.0,
+      "step": 1598
+    },
+    {
+      "epoch": 3.6431014823261116,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.7695220258910124e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8771436512470245,
+      "num_tokens": 171024919.0,
+      "step": 1599
+    },
+    {
+      "epoch": 3.6453819840364883,
+      "grad_norm": 3.625,
+      "learning_rate": 3.7678991314814305e-06,
+      "loss": 0.6,
+      "mean_token_accuracy": 0.8739635199308395,
+      "num_tokens": 171131530.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.6476624857468645,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.766275517436779e-06,
+      "loss": 0.5897,
+      "mean_token_accuracy": 0.8780031353235245,
+      "num_tokens": 171238533.0,
+      "step": 1601
+    },
+    {
+      "epoch": 3.6499429874572407,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.7646511846785904e-06,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8816352784633636,
+      "num_tokens": 171345881.0,
+      "step": 1602
+    },
+    {
+      "epoch": 3.652223489167617,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.7630261341288044e-06,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.883026197552681,
+      "num_tokens": 171452807.0,
+      "step": 1603
+    },
+    {
+      "epoch": 3.654503990877993,
+      "grad_norm": 4.0,
+      "learning_rate": 3.7614003667097674e-06,
+      "loss": 0.6065,
+      "mean_token_accuracy": 0.8770434111356735,
+      "num_tokens": 171559794.0,
+      "step": 1604
+    },
+    {
+      "epoch": 3.6567844925883692,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.759773883344236e-06,
+      "loss": 0.6215,
+      "mean_token_accuracy": 0.8752636462450027,
+      "num_tokens": 171666284.0,
+      "step": 1605
+    },
+    {
+      "epoch": 3.659064994298746,
+      "grad_norm": 5.625,
+      "learning_rate": 3.7581466849553685e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.884151503443718,
+      "num_tokens": 171774291.0,
+      "step": 1606
+    },
+    {
+      "epoch": 3.661345496009122,
+      "grad_norm": 2.5,
+      "learning_rate": 3.7565187724667324e-06,
+      "loss": 0.562,
+      "mean_token_accuracy": 0.8847729563713074,
+      "num_tokens": 171881709.0,
+      "step": 1607
+    },
+    {
+      "epoch": 3.6636259977194983,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.7548901468022993e-06,
+      "loss": 0.5917,
+      "mean_token_accuracy": 0.8790787905454636,
+      "num_tokens": 171988700.0,
+      "step": 1608
+    },
+    {
+      "epoch": 3.6659064994298745,
+      "grad_norm": 3.5,
+      "learning_rate": 3.7532608088864444e-06,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.876745343208313,
+      "num_tokens": 172095530.0,
+      "step": 1609
+    },
+    {
+      "epoch": 3.668187001140251,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.75163075964395e-06,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8810919225215912,
+      "num_tokens": 172202580.0,
+      "step": 1610
+    },
+    {
+      "epoch": 3.6704675028506273,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8769372403621674,
+      "num_tokens": 172310027.0,
+      "step": 1611
+    },
+    {
+      "epoch": 3.6727480045610035,
+      "grad_norm": 2.265625,
+      "learning_rate": 3.748368530880183e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8828428983688354,
+      "num_tokens": 172416427.0,
+      "step": 1612
+    },
+    {
+      "epoch": 3.6750285062713797,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.7467363532104874e-06,
+      "loss": 0.5595,
+      "mean_token_accuracy": 0.8848859220743179,
+      "num_tokens": 172523861.0,
+      "step": 1613
+    },
+    {
+      "epoch": 3.677309007981756,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.7451034679173082e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8770375549793243,
+      "num_tokens": 172630804.0,
+      "step": 1614
+    },
+    {
+      "epoch": 3.679589509692132,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.7434698759274366e-06,
+      "loss": 0.5891,
+      "mean_token_accuracy": 0.8780438005924225,
+      "num_tokens": 172738455.0,
+      "step": 1615
+    },
+    {
+      "epoch": 3.6818700114025087,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.741835578168071e-06,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.8857946693897247,
+      "num_tokens": 172845911.0,
+      "step": 1616
+    },
+    {
+      "epoch": 3.684150513112885,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.740200575566806e-06,
+      "loss": 0.599,
+      "mean_token_accuracy": 0.8763966113328934,
+      "num_tokens": 172952684.0,
+      "step": 1617
+    },
+    {
+      "epoch": 3.686431014823261,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.7385648690516364e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.879105344414711,
+      "num_tokens": 173059713.0,
+      "step": 1618
+    },
+    {
+      "epoch": 3.6887115165336373,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.7369284595509587e-06,
+      "loss": 0.5858,
+      "mean_token_accuracy": 0.873798742890358,
+      "num_tokens": 173167102.0,
+      "step": 1619
+    },
+    {
+      "epoch": 3.690992018244014,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.7352913479935672e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8798558115959167,
+      "num_tokens": 173274635.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.69327251995439,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.7336535353086546e-06,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.8765471577644348,
+      "num_tokens": 173382044.0,
+      "step": 1621
+    },
+    {
+      "epoch": 3.6955530216647663,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.7320150224258124e-06,
+      "loss": 0.5818,
+      "mean_token_accuracy": 0.8792853504419327,
+      "num_tokens": 173489148.0,
+      "step": 1622
+    },
+    {
+      "epoch": 3.6978335233751425,
+      "grad_norm": 3.25,
+      "learning_rate": 3.7303758102750274e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8817913979291916,
+      "num_tokens": 173596404.0,
+      "step": 1623
+    },
+    {
+      "epoch": 3.7001140250855187,
+      "grad_norm": 2.75,
+      "learning_rate": 3.7287358997866872e-06,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8805738687515259,
+      "num_tokens": 173703327.0,
+      "step": 1624
+    },
+    {
+      "epoch": 3.702394526795895,
+      "grad_norm": 3.25,
+      "learning_rate": 3.7270952918915715e-06,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8786198794841766,
+      "num_tokens": 173809938.0,
+      "step": 1625
+    },
+    {
+      "epoch": 3.7046750285062715,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.7254539875208577e-06,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8781810998916626,
+      "num_tokens": 173916969.0,
+      "step": 1626
+    },
+    {
+      "epoch": 3.7069555302166477,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.7238119876061196e-06,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8811867833137512,
+      "num_tokens": 174023879.0,
+      "step": 1627
+    },
+    {
+      "epoch": 3.709236031927024,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.7221692930793234e-06,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8845594227313995,
+      "num_tokens": 174130854.0,
+      "step": 1628
+    },
+    {
+      "epoch": 3.7115165336374,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.7205259048728316e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8832680881023407,
+      "num_tokens": 174238556.0,
+      "step": 1629
+    },
+    {
+      "epoch": 3.7137970353477767,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.718881823919399e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8850577622652054,
+      "num_tokens": 174346096.0,
+      "step": 1630
+    },
+    {
+      "epoch": 3.716077537058153,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.717237051152175e-06,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8809388726949692,
+      "num_tokens": 174453264.0,
+      "step": 1631
+    },
+    {
+      "epoch": 3.718358038768529,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.7155915875047005e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8833019137382507,
+      "num_tokens": 174560352.0,
+      "step": 1632
+    },
+    {
+      "epoch": 3.7206385404789053,
+      "grad_norm": 4.0,
+      "learning_rate": 3.7139454339109082e-06,
+      "loss": 0.59,
+      "mean_token_accuracy": 0.8777986019849777,
+      "num_tokens": 174667183.0,
+      "step": 1633
+    },
+    {
+      "epoch": 3.7229190421892815,
+      "grad_norm": 5.375,
+      "learning_rate": 3.7122985913051242e-06,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.8856906592845917,
+      "num_tokens": 174774451.0,
+      "step": 1634
+    },
+    {
+      "epoch": 3.7251995438996577,
+      "grad_norm": 2.203125,
+      "learning_rate": 3.710651060622064e-06,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8867538422346115,
+      "num_tokens": 174881357.0,
+      "step": 1635
+    },
+    {
+      "epoch": 3.7274800456100343,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.7090028427968343e-06,
+      "loss": 0.6073,
+      "mean_token_accuracy": 0.8716320544481277,
+      "num_tokens": 174987742.0,
+      "step": 1636
+    },
+    {
+      "epoch": 3.7297605473204105,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.7073539387649316e-06,
+      "loss": 0.5933,
+      "mean_token_accuracy": 0.8786651045084,
+      "num_tokens": 175094511.0,
+      "step": 1637
+    },
+    {
+      "epoch": 3.7320410490307867,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.7057043494622423e-06,
+      "loss": 0.6106,
+      "mean_token_accuracy": 0.8771418780088425,
+      "num_tokens": 175201546.0,
+      "step": 1638
+    },
+    {
+      "epoch": 3.734321550741163,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.704054075825042e-06,
+      "loss": 0.5927,
+      "mean_token_accuracy": 0.8767105042934418,
+      "num_tokens": 175308332.0,
+      "step": 1639
+    },
+    {
+      "epoch": 3.7366020524515395,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.702403118789992e-06,
+      "loss": 0.5588,
+      "mean_token_accuracy": 0.8864465802907944,
+      "num_tokens": 175415556.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.7388825541619157,
+      "grad_norm": 5.78125,
+      "learning_rate": 3.7007514792941462e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8830393701791763,
+      "num_tokens": 175522759.0,
+      "step": 1641
+    },
+    {
+      "epoch": 3.741163055872292,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.6990991582749414e-06,
+      "loss": 0.5712,
+      "mean_token_accuracy": 0.8837911188602448,
+      "num_tokens": 175630033.0,
+      "step": 1642
+    },
+    {
+      "epoch": 3.743443557582668,
+      "grad_norm": 3.125,
+      "learning_rate": 3.6974461566702048e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8770407438278198,
+      "num_tokens": 175737297.0,
+      "step": 1643
+    },
+    {
+      "epoch": 3.7457240592930443,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.695792475418146e-06,
+      "loss": 0.5899,
+      "mean_token_accuracy": 0.8814648389816284,
+      "num_tokens": 175844521.0,
+      "step": 1644
+    },
+    {
+      "epoch": 3.7480045610034205,
+      "grad_norm": 6.34375,
+      "learning_rate": 3.6941381154573646e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.878902330994606,
+      "num_tokens": 175951433.0,
+      "step": 1645
+    },
+    {
+      "epoch": 3.750285062713797,
+      "grad_norm": 4.84375,
+      "learning_rate": 3.692483077726843e-06,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.884130209684372,
+      "num_tokens": 176058602.0,
+      "step": 1646
+    },
+    {
+      "epoch": 3.7525655644241733,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.6908273631659475e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8837015181779861,
+      "num_tokens": 176165024.0,
+      "step": 1647
+    },
+    {
+      "epoch": 3.7548460661345495,
+      "grad_norm": 4.5,
+      "learning_rate": 3.689170972714431e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8826144188642502,
+      "num_tokens": 176272341.0,
+      "step": 1648
+    },
+    {
+      "epoch": 3.757126567844926,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.6875139073124277e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8824542462825775,
+      "num_tokens": 176379289.0,
+      "step": 1649
+    },
+    {
+      "epoch": 3.7594070695553023,
+      "grad_norm": 4.375,
+      "learning_rate": 3.6858561679004567e-06,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.8805209249258041,
+      "num_tokens": 176486406.0,
+      "step": 1650
+    },
+    {
+      "epoch": 3.7616875712656785,
+      "grad_norm": 5.21875,
+      "learning_rate": 3.684197755419419e-06,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8811310976743698,
+      "num_tokens": 176593647.0,
+      "step": 1651
+    },
+    {
+      "epoch": 3.7639680729760547,
+      "grad_norm": 6.21875,
+      "learning_rate": 3.6825386708105963e-06,
+      "loss": 0.5575,
+      "mean_token_accuracy": 0.8876552581787109,
+      "num_tokens": 176700583.0,
+      "step": 1652
+    },
+    {
+      "epoch": 3.766248574686431,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.6808789150156545e-06,
+      "loss": 0.5908,
+      "mean_token_accuracy": 0.8768218755722046,
+      "num_tokens": 176807605.0,
+      "step": 1653
+    },
+    {
+      "epoch": 3.768529076396807,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.679218488976638e-06,
+      "loss": 0.5967,
+      "mean_token_accuracy": 0.8734856247901917,
+      "num_tokens": 176914867.0,
+      "step": 1654
+    },
+    {
+      "epoch": 3.7708095781071833,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.677557393635973e-06,
+      "loss": 0.5803,
+      "mean_token_accuracy": 0.8815526366233826,
+      "num_tokens": 177021989.0,
+      "step": 1655
+    },
+    {
+      "epoch": 3.77309007981756,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.6758956299364643e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8808587789535522,
+      "num_tokens": 177128640.0,
+      "step": 1656
+    },
+    {
+      "epoch": 3.775370581527936,
+      "grad_norm": 2.375,
+      "learning_rate": 3.674233198821299e-06,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8834094703197479,
+      "num_tokens": 177236418.0,
+      "step": 1657
+    },
+    {
+      "epoch": 3.7776510832383123,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.6725701012340387e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8827496618032455,
+      "num_tokens": 177343371.0,
+      "step": 1658
+    },
+    {
+      "epoch": 3.779931584948689,
+      "grad_norm": 2.25,
+      "learning_rate": 3.6709063381186267e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8798905909061432,
+      "num_tokens": 177450862.0,
+      "step": 1659
+    },
+    {
+      "epoch": 3.782212086659065,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.6692419104193823e-06,
+      "loss": 0.6005,
+      "mean_token_accuracy": 0.8762509971857071,
+      "num_tokens": 177557890.0,
+      "step": 1660
+    },
+    {
+      "epoch": 3.7844925883694414,
+      "grad_norm": 3.0,
+      "learning_rate": 3.6675768190810023e-06,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8809296637773514,
+      "num_tokens": 177664846.0,
+      "step": 1661
+    },
+    {
+      "epoch": 3.7867730900798175,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.665911065048561e-06,
+      "loss": 0.5714,
+      "mean_token_accuracy": 0.8829169422388077,
+      "num_tokens": 177772527.0,
+      "step": 1662
+    },
+    {
+      "epoch": 3.7890535917901937,
+      "grad_norm": 5.25,
+      "learning_rate": 3.6642446492675075e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8831394165754318,
+      "num_tokens": 177879427.0,
+      "step": 1663
+    },
+    {
+      "epoch": 3.79133409350057,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.6625775726836677e-06,
+      "loss": 0.5886,
+      "mean_token_accuracy": 0.8788637816905975,
+      "num_tokens": 177986497.0,
+      "step": 1664
+    },
+    {
+      "epoch": 3.7936145952109466,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.6609098362432425e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8837129175662994,
+      "num_tokens": 178093286.0,
+      "step": 1665
+    },
+    {
+      "epoch": 3.7958950969213228,
+      "grad_norm": 4.0,
+      "learning_rate": 3.659241440892806e-06,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8812308311462402,
+      "num_tokens": 178200373.0,
+      "step": 1666
+    },
+    {
+      "epoch": 3.798175598631699,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.6575723875793085e-06,
+      "loss": 0.5917,
+      "mean_token_accuracy": 0.8776303678750992,
+      "num_tokens": 178307338.0,
+      "step": 1667
+    },
+    {
+      "epoch": 3.800456100342075,
+      "grad_norm": 6.3125,
+      "learning_rate": 3.655902677250071e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8835581988096237,
+      "num_tokens": 178414549.0,
+      "step": 1668
+    },
+    {
+      "epoch": 3.802736602052452,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.6542323108527896e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8782322406768799,
+      "num_tokens": 178521662.0,
+      "step": 1669
+    },
+    {
+      "epoch": 3.805017103762828,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.652561289335532e-06,
+      "loss": 0.5973,
+      "mean_token_accuracy": 0.8794067353010178,
+      "num_tokens": 178627910.0,
+      "step": 1670
+    },
+    {
+      "epoch": 3.807297605473204,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.6508896136467376e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8825301080942154,
+      "num_tokens": 178735254.0,
+      "step": 1671
+    },
+    {
+      "epoch": 3.8095781071835804,
+      "grad_norm": 3.0,
+      "learning_rate": 3.649217284735217e-06,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8833526968955994,
+      "num_tokens": 178842561.0,
+      "step": 1672
+    },
+    {
+      "epoch": 3.8118586088939566,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.6475443035501522e-06,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.8792631328105927,
+      "num_tokens": 178949352.0,
+      "step": 1673
+    },
+    {
+      "epoch": 3.8141391106043327,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.645870671041095e-06,
+      "loss": 0.5765,
+      "mean_token_accuracy": 0.8842526823282242,
+      "num_tokens": 179056304.0,
+      "step": 1674
+    },
+    {
+      "epoch": 3.8164196123147094,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.6441963881579668e-06,
+      "loss": 0.586,
+      "mean_token_accuracy": 0.8808672279119492,
+      "num_tokens": 179163897.0,
+      "step": 1675
+    },
+    {
+      "epoch": 3.8187001140250856,
+      "grad_norm": 4.75,
+      "learning_rate": 3.642521455851058e-06,
+      "loss": 0.5614,
+      "mean_token_accuracy": 0.8823509216308594,
+      "num_tokens": 179270806.0,
+      "step": 1676
+    },
+    {
+      "epoch": 3.8209806157354618,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.6408458750710284e-06,
+      "loss": 0.5578,
+      "mean_token_accuracy": 0.8836124241352081,
+      "num_tokens": 179378081.0,
+      "step": 1677
+    },
+    {
+      "epoch": 3.823261117445838,
+      "grad_norm": 4.625,
+      "learning_rate": 3.639169646768905e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8754040449857712,
+      "num_tokens": 179485558.0,
+      "step": 1678
+    },
+    {
+      "epoch": 3.8255416191562146,
+      "grad_norm": 5.21875,
+      "learning_rate": 3.637492771896082e-06,
+      "loss": 0.5817,
+      "mean_token_accuracy": 0.8810373991727829,
+      "num_tokens": 179592323.0,
+      "step": 1679
+    },
+    {
+      "epoch": 3.827822120866591,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.6358152514043226e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8800250738859177,
+      "num_tokens": 179699420.0,
+      "step": 1680
+    },
+    {
+      "epoch": 3.830102622576967,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.634137086245754e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8808536231517792,
+      "num_tokens": 179806666.0,
+      "step": 1681
+    },
+    {
+      "epoch": 3.832383124287343,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.6324582773728712e-06,
+      "loss": 0.5973,
+      "mean_token_accuracy": 0.8750377893447876,
+      "num_tokens": 179914021.0,
+      "step": 1682
+    },
+    {
+      "epoch": 3.8346636259977194,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.6307788257385325e-06,
+      "loss": 0.5859,
+      "mean_token_accuracy": 0.8790106326341629,
+      "num_tokens": 180020983.0,
+      "step": 1683
+    },
+    {
+      "epoch": 3.8369441277080956,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.6290987322959624e-06,
+      "loss": 0.5493,
+      "mean_token_accuracy": 0.8865607380867004,
+      "num_tokens": 180128755.0,
+      "step": 1684
+    },
+    {
+      "epoch": 3.839224629418472,
+      "grad_norm": 2.375,
+      "learning_rate": 3.6274179979987507e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.881371021270752,
+      "num_tokens": 180235525.0,
+      "step": 1685
+    },
+    {
+      "epoch": 3.8415051311288484,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.625736623800849e-06,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.878914400935173,
+      "num_tokens": 180342761.0,
+      "step": 1686
+    },
+    {
+      "epoch": 3.8437856328392246,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.624054610656572e-06,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.879774883389473,
+      "num_tokens": 180450260.0,
+      "step": 1687
+    },
+    {
+      "epoch": 3.846066134549601,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.622371959520599e-06,
+      "loss": 0.5879,
+      "mean_token_accuracy": 0.881365180015564,
+      "num_tokens": 180557292.0,
+      "step": 1688
+    },
+    {
+      "epoch": 3.8483466362599774,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.6206886713479705e-06,
+      "loss": 0.6012,
+      "mean_token_accuracy": 0.8755958080291748,
+      "num_tokens": 180664350.0,
+      "step": 1689
+    },
+    {
+      "epoch": 3.8506271379703536,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.6190047470940875e-06,
+      "loss": 0.5677,
+      "mean_token_accuracy": 0.8807881772518158,
+      "num_tokens": 180771157.0,
+      "step": 1690
+    },
+    {
+      "epoch": 3.85290763968073,
+      "grad_norm": 2.875,
+      "learning_rate": 3.6173201877147134e-06,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.8763711750507355,
+      "num_tokens": 180877774.0,
+      "step": 1691
+    },
+    {
+      "epoch": 3.855188141391106,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.6156349941659717e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8838952630758286,
+      "num_tokens": 180985055.0,
+      "step": 1692
+    },
+    {
+      "epoch": 3.857468643101482,
+      "grad_norm": 4.53125,
+      "learning_rate": 3.613949167404345e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8797314018011093,
+      "num_tokens": 181091651.0,
+      "step": 1693
+    },
+    {
+      "epoch": 3.8597491448118584,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.6122627083866773e-06,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8829943537712097,
+      "num_tokens": 181198250.0,
+      "step": 1694
+    },
+    {
+      "epoch": 3.862029646522235,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.610575618070169e-06,
+      "loss": 0.559,
+      "mean_token_accuracy": 0.8832627683877945,
+      "num_tokens": 181305182.0,
+      "step": 1695
+    },
+    {
+      "epoch": 3.864310148232611,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.6088878974123796e-06,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8848606944084167,
+      "num_tokens": 181412694.0,
+      "step": 1696
+    },
+    {
+      "epoch": 3.8665906499429874,
+      "grad_norm": 5.125,
+      "learning_rate": 3.6071995473712284e-06,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.8789099603891373,
+      "num_tokens": 181519464.0,
+      "step": 1697
+    },
+    {
+      "epoch": 3.8688711516533636,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.605510568904989e-06,
+      "loss": 0.6049,
+      "mean_token_accuracy": 0.8752625435590744,
+      "num_tokens": 181626561.0,
+      "step": 1698
+    },
+    {
+      "epoch": 3.8711516533637402,
+      "grad_norm": 4.90625,
+      "learning_rate": 3.6038209629722936e-06,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8840775638818741,
+      "num_tokens": 181733816.0,
+      "step": 1699
+    },
+    {
+      "epoch": 3.8734321550741164,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.6021307305321295e-06,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8823692500591278,
+      "num_tokens": 181841269.0,
+      "step": 1700
+    },
+    {
+      "epoch": 3.8757126567844926,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.6004398725438406e-06,
+      "loss": 0.5469,
+      "mean_token_accuracy": 0.8876504898071289,
+      "num_tokens": 181948467.0,
+      "step": 1701
+    },
+    {
+      "epoch": 3.877993158494869,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.5987483899671245e-06,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.8821342885494232,
+      "num_tokens": 182055789.0,
+      "step": 1702
+    },
+    {
+      "epoch": 3.880273660205245,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.597056283762034e-06,
+      "loss": 0.5856,
+      "mean_token_accuracy": 0.8786395937204361,
+      "num_tokens": 182163625.0,
+      "step": 1703
+    },
+    {
+      "epoch": 3.882554161915621,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.5953635548889777e-06,
+      "loss": 0.5699,
+      "mean_token_accuracy": 0.8824369460344315,
+      "num_tokens": 182270282.0,
+      "step": 1704
+    },
+    {
+      "epoch": 3.884834663625998,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.5936702043087134e-06,
+      "loss": 0.6037,
+      "mean_token_accuracy": 0.8777669966220856,
+      "num_tokens": 182376966.0,
+      "step": 1705
+    },
+    {
+      "epoch": 3.887115165336374,
+      "grad_norm": 2.75,
+      "learning_rate": 3.5919762329823556e-06,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8837083727121353,
+      "num_tokens": 182484405.0,
+      "step": 1706
+    },
+    {
+      "epoch": 3.88939566704675,
+      "grad_norm": 4.125,
+      "learning_rate": 3.5902816418713694e-06,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8821840286254883,
+      "num_tokens": 182591485.0,
+      "step": 1707
+    },
+    {
+      "epoch": 3.8916761687571264,
+      "grad_norm": 8.375,
+      "learning_rate": 3.5885864319375717e-06,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8777542114257812,
+      "num_tokens": 182699148.0,
+      "step": 1708
+    },
+    {
+      "epoch": 3.893956670467503,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.5868906041431313e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8843749016523361,
+      "num_tokens": 182805780.0,
+      "step": 1709
+    },
+    {
+      "epoch": 3.8962371721778792,
+      "grad_norm": 4.125,
+      "learning_rate": 3.5851941594505674e-06,
+      "loss": 0.5906,
+      "mean_token_accuracy": 0.8774453997612,
+      "num_tokens": 182912492.0,
+      "step": 1710
+    },
+    {
+      "epoch": 3.8985176738882554,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.5834970988227484e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8813342452049255,
+      "num_tokens": 183019581.0,
+      "step": 1711
+    },
+    {
+      "epoch": 3.9007981755986316,
+      "grad_norm": 5.125,
+      "learning_rate": 3.581799423222895e-06,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8827546834945679,
+      "num_tokens": 183126811.0,
+      "step": 1712
+    },
+    {
+      "epoch": 3.903078677309008,
+      "grad_norm": 8.5,
+      "learning_rate": 3.580101133614573e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.879145547747612,
+      "num_tokens": 183234837.0,
+      "step": 1713
+    },
+    {
+      "epoch": 3.905359179019384,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.5784022309617006e-06,
+      "loss": 0.5614,
+      "mean_token_accuracy": 0.8833577632904053,
+      "num_tokens": 183342206.0,
+      "step": 1714
+    },
+    {
+      "epoch": 3.9076396807297606,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.57670271622854e-06,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8815594464540482,
+      "num_tokens": 183448841.0,
+      "step": 1715
+    },
+    {
+      "epoch": 3.909920182440137,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.5750025903797053e-06,
+      "loss": 0.5777,
+      "mean_token_accuracy": 0.8809472620487213,
+      "num_tokens": 183555198.0,
+      "step": 1716
+    },
+    {
+      "epoch": 3.912200684150513,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.5733018543801534e-06,
+      "loss": 0.5877,
+      "mean_token_accuracy": 0.8786370903253555,
+      "num_tokens": 183661758.0,
+      "step": 1717
+    },
+    {
+      "epoch": 3.9144811858608897,
+      "grad_norm": 6.25,
+      "learning_rate": 3.5716005091951906e-06,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8804649114608765,
+      "num_tokens": 183768385.0,
+      "step": 1718
+    },
+    {
+      "epoch": 3.916761687571266,
+      "grad_norm": 5.34375,
+      "learning_rate": 3.569898555790466e-06,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8821226358413696,
+      "num_tokens": 183874915.0,
+      "step": 1719
+    },
+    {
+      "epoch": 3.919042189281642,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.5681959951319766e-06,
+      "loss": 0.6009,
+      "mean_token_accuracy": 0.8761325925588608,
+      "num_tokens": 183982390.0,
+      "step": 1720
+    },
+    {
+      "epoch": 3.9213226909920182,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.566492828186063e-06,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8820738047361374,
+      "num_tokens": 184089244.0,
+      "step": 1721
+    },
+    {
+      "epoch": 3.9236031927023944,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.564789055919409e-06,
+      "loss": 0.5631,
+      "mean_token_accuracy": 0.8825008720159531,
+      "num_tokens": 184196515.0,
+      "step": 1722
+    },
+    {
+      "epoch": 3.9258836944127706,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.5630846792990435e-06,
+      "loss": 0.6072,
+      "mean_token_accuracy": 0.8774504512548447,
+      "num_tokens": 184303445.0,
+      "step": 1723
+    },
+    {
+      "epoch": 3.928164196123147,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.5613796992923382e-06,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8825356364250183,
+      "num_tokens": 184410680.0,
+      "step": 1724
+    },
+    {
+      "epoch": 3.9304446978335235,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.559674116867006e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.881066232919693,
+      "num_tokens": 184517883.0,
+      "step": 1725
+    },
+    {
+      "epoch": 3.9327251995438997,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.5579679329911025e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.8830063939094543,
+      "num_tokens": 184625069.0,
+      "step": 1726
+    },
+    {
+      "epoch": 3.935005701254276,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.556261148633026e-06,
+      "loss": 0.576,
+      "mean_token_accuracy": 0.8830278366804123,
+      "num_tokens": 184732358.0,
+      "step": 1727
+    },
+    {
+      "epoch": 3.9372862029646525,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.5545537647615125e-06,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8799594938755035,
+      "num_tokens": 184839208.0,
+      "step": 1728
+    },
+    {
+      "epoch": 3.9395667046750287,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.552845782345642e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8789944648742676,
+      "num_tokens": 184945933.0,
+      "step": 1729
+    },
+    {
+      "epoch": 3.941847206385405,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.551137202354831e-06,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.8809881657361984,
+      "num_tokens": 185052752.0,
+      "step": 1730
+    },
+    {
+      "epoch": 3.944127708095781,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.5494280257588367e-06,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8846791535615921,
+      "num_tokens": 185159970.0,
+      "step": 1731
+    },
+    {
+      "epoch": 3.9464082098061573,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.547718253527755e-06,
+      "loss": 0.5646,
+      "mean_token_accuracy": 0.8843375146389008,
+      "num_tokens": 185267250.0,
+      "step": 1732
+    },
+    {
+      "epoch": 3.9486887115165334,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.546007886632019e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8792973756790161,
+      "num_tokens": 185373715.0,
+      "step": 1733
+    },
+    {
+      "epoch": 3.95096921322691,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.5442969260424022e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8815480768680573,
+      "num_tokens": 185480906.0,
+      "step": 1734
+    },
+    {
+      "epoch": 3.9532497149372863,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.5425853727300095e-06,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.8790358453989029,
+      "num_tokens": 185588351.0,
+      "step": 1735
+    },
+    {
+      "epoch": 3.9555302166476625,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.5408732276662882e-06,
+      "loss": 0.5974,
+      "mean_token_accuracy": 0.8829907327890396,
+      "num_tokens": 185695047.0,
+      "step": 1736
+    },
+    {
+      "epoch": 3.9578107183580387,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.5391604918230173e-06,
+      "loss": 0.582,
+      "mean_token_accuracy": 0.878580242395401,
+      "num_tokens": 185802784.0,
+      "step": 1737
+    },
+    {
+      "epoch": 3.9600912200684153,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.537447166172313e-06,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.879580944776535,
+      "num_tokens": 185909720.0,
+      "step": 1738
+    },
+    {
+      "epoch": 3.9623717217787915,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.5357332516866256e-06,
+      "loss": 0.5551,
+      "mean_token_accuracy": 0.8843010365962982,
+      "num_tokens": 186016443.0,
+      "step": 1739
+    },
+    {
+      "epoch": 3.9646522234891677,
+      "grad_norm": 3.375,
+      "learning_rate": 3.534018749338741e-06,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8813291490077972,
+      "num_tokens": 186123915.0,
+      "step": 1740
+    },
+    {
+      "epoch": 3.966932725199544,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.532303660101776e-06,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8835762590169907,
+      "num_tokens": 186230831.0,
+      "step": 1741
+    },
+    {
+      "epoch": 3.96921322690992,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.530587984949183e-06,
+      "loss": 0.6007,
+      "mean_token_accuracy": 0.8778104931116104,
+      "num_tokens": 186337578.0,
+      "step": 1742
+    },
+    {
+      "epoch": 3.9714937286202963,
+      "grad_norm": 3.125,
+      "learning_rate": 3.5288717248547453e-06,
+      "loss": 0.5634,
+      "mean_token_accuracy": 0.8861571699380875,
+      "num_tokens": 186445605.0,
+      "step": 1743
+    },
+    {
+      "epoch": 3.973774230330673,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.5271548807925803e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8777986913919449,
+      "num_tokens": 186552754.0,
+      "step": 1744
+    },
+    {
+      "epoch": 3.976054732041049,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.525437453737136e-06,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.8793213814496994,
+      "num_tokens": 186659906.0,
+      "step": 1745
+    },
+    {
+      "epoch": 3.9783352337514253,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.5237194446631883e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8833697587251663,
+      "num_tokens": 186766193.0,
+      "step": 1746
+    },
+    {
+      "epoch": 3.9806157354618015,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.522000854545849e-06,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8789832890033722,
+      "num_tokens": 186872982.0,
+      "step": 1747
+    },
+    {
+      "epoch": 3.982896237172178,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.520281684360554e-06,
+      "loss": 0.5893,
+      "mean_token_accuracy": 0.8788421899080276,
+      "num_tokens": 186979970.0,
+      "step": 1748
+    },
+    {
+      "epoch": 3.9851767388825543,
+      "grad_norm": 2.875,
+      "learning_rate": 3.5185619350830725e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8799102157354355,
+      "num_tokens": 187086646.0,
+      "step": 1749
+    },
+    {
+      "epoch": 3.9874572405929305,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.516841607689501e-06,
+      "loss": 0.5596,
+      "mean_token_accuracy": 0.8887099474668503,
+      "num_tokens": 187194165.0,
+      "step": 1750
+    },
+    {
+      "epoch": 3.9897377423033067,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.515120703156264e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8788872063159943,
+      "num_tokens": 187300450.0,
+      "step": 1751
+    },
+    {
+      "epoch": 3.992018244013683,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.5133992224601126e-06,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.8765320926904678,
+      "num_tokens": 187407316.0,
+      "step": 1752
+    },
+    {
+      "epoch": 3.994298745724059,
+      "grad_norm": 4.46875,
+      "learning_rate": 3.511677166578128e-06,
+      "loss": 0.5517,
+      "mean_token_accuracy": 0.8845993727445602,
+      "num_tokens": 187515138.0,
+      "step": 1753
+    },
+    {
+      "epoch": 3.9965792474344357,
+      "grad_norm": 2.328125,
+      "learning_rate": 3.509954536487714e-06,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8815711736679077,
+      "num_tokens": 187622052.0,
+      "step": 1754
+    },
+    {
+      "epoch": 3.998859749144812,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.5082313331666035e-06,
+      "loss": 0.5521,
+      "mean_token_accuracy": 0.8849412500858307,
+      "num_tokens": 187729710.0,
+      "step": 1755
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 4.90625,
+      "learning_rate": 3.506507557592853e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.879196435213089,
+      "num_tokens": 187768928.0,
+      "step": 1756
+    },
+    {
+      "epoch": 4.002280501710376,
+      "grad_norm": 3.0,
+      "learning_rate": 3.5047832107448437e-06,
+      "loss": 0.5866,
+      "mean_token_accuracy": 0.8787702172994614,
+      "num_tokens": 187875429.0,
+      "step": 1757
+    },
+    {
+      "epoch": 4.004561003420752,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.503058293601283e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.8796392232179642,
+      "num_tokens": 187982859.0,
+      "step": 1758
+    },
+    {
+      "epoch": 4.006841505131129,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.5013328071411995e-06,
+      "loss": 0.5651,
+      "mean_token_accuracy": 0.8837246298789978,
+      "num_tokens": 188090247.0,
+      "step": 1759
+    },
+    {
+      "epoch": 4.009122006841505,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.499606752343945e-06,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8794354498386383,
+      "num_tokens": 188198071.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.009122006841505,
+      "eval_loss": 0.5914322733879089,
+      "eval_mean_token_accuracy": 0.8791430533612182,
+      "eval_num_tokens": 188198071.0,
+      "eval_runtime": 58.7059,
+      "eval_samples_per_second": 142.831,
+      "eval_steps_per_second": 4.48,
+      "step": 1760
+    },
+    {
+      "epoch": 4.011402508551882,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.4978801301891972e-06,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8818523436784744,
+      "num_tokens": 188305205.0,
+      "step": 1761
+    },
+    {
+      "epoch": 4.013683010262258,
+      "grad_norm": 3.625,
+      "learning_rate": 3.496152941656952e-06,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8844558596611023,
+      "num_tokens": 188411935.0,
+      "step": 1762
+    },
+    {
+      "epoch": 4.015963511972634,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.494425187727528e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8814276456832886,
+      "num_tokens": 188519090.0,
+      "step": 1763
+    },
+    {
+      "epoch": 4.01824401368301,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.4926968693815667e-06,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8852830976247787,
+      "num_tokens": 188626017.0,
+      "step": 1764
+    },
+    {
+      "epoch": 4.020524515393387,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.4909679876000256e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8862181156873703,
+      "num_tokens": 188733308.0,
+      "step": 1765
+    },
+    {
+      "epoch": 4.022805017103763,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.4892385433641875e-06,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8816091120243073,
+      "num_tokens": 188840326.0,
+      "step": 1766
+    },
+    {
+      "epoch": 4.025085518814139,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.4875085376556493e-06,
+      "loss": 0.5696,
+      "mean_token_accuracy": 0.881555825471878,
+      "num_tokens": 188947739.0,
+      "step": 1767
+    },
+    {
+      "epoch": 4.027366020524515,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.4857779714563305e-06,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8749964982271194,
+      "num_tokens": 189054532.0,
+      "step": 1768
+    },
+    {
+      "epoch": 4.029646522234891,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.4840468457484654e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8818257004022598,
+      "num_tokens": 189161911.0,
+      "step": 1769
+    },
+    {
+      "epoch": 4.031927023945268,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.4823151615146093e-06,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8840532153844833,
+      "num_tokens": 189269010.0,
+      "step": 1770
+    },
+    {
+      "epoch": 4.034207525655645,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.480582919737631e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8821255415678024,
+      "num_tokens": 189375921.0,
+      "step": 1771
+    },
+    {
+      "epoch": 4.036488027366021,
+      "grad_norm": 4.875,
+      "learning_rate": 3.478850121400719e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8794375658035278,
+      "num_tokens": 189483390.0,
+      "step": 1772
+    },
+    {
+      "epoch": 4.038768529076397,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.477116767487375e-06,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8811006546020508,
+      "num_tokens": 189590557.0,
+      "step": 1773
+    },
+    {
+      "epoch": 4.041049030786773,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.475382858981418e-06,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8820205479860306,
+      "num_tokens": 189697618.0,
+      "step": 1774
+    },
+    {
+      "epoch": 4.043329532497149,
+      "grad_norm": 7.40625,
+      "learning_rate": 3.473648396866981e-06,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8759682923555374,
+      "num_tokens": 189804472.0,
+      "step": 1775
+    },
+    {
+      "epoch": 4.045610034207526,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.4719133821285108e-06,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8793429881334305,
+      "num_tokens": 189911753.0,
+      "step": 1776
+    },
+    {
+      "epoch": 4.047890535917902,
+      "grad_norm": 4.78125,
+      "learning_rate": 3.470177815750769e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8850299268960953,
+      "num_tokens": 190019669.0,
+      "step": 1777
+    },
+    {
+      "epoch": 4.050171037628278,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.4684416987188273e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8823755383491516,
+      "num_tokens": 190127195.0,
+      "step": 1778
+    },
+    {
+      "epoch": 4.052451539338654,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.4667050320180755e-06,
+      "loss": 0.5832,
+      "mean_token_accuracy": 0.8784758597612381,
+      "num_tokens": 190234158.0,
+      "step": 1779
+    },
+    {
+      "epoch": 4.05473204104903,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.4649678166342104e-06,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8827002048492432,
+      "num_tokens": 190341000.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.0570125427594075,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.4632300535532415e-06,
+      "loss": 0.6008,
+      "mean_token_accuracy": 0.8756804317235947,
+      "num_tokens": 190448392.0,
+      "step": 1781
+    },
+    {
+      "epoch": 4.059293044469784,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.46149174376149e-06,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8806837797164917,
+      "num_tokens": 190555060.0,
+      "step": 1782
+    },
+    {
+      "epoch": 4.06157354618016,
+      "grad_norm": 5.09375,
+      "learning_rate": 3.459752888245587e-06,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8827811032533646,
+      "num_tokens": 190662364.0,
+      "step": 1783
+    },
+    {
+      "epoch": 4.063854047890536,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.4580134879924732e-06,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.877520278096199,
+      "num_tokens": 190769329.0,
+      "step": 1784
+    },
+    {
+      "epoch": 4.066134549600912,
+      "grad_norm": 3.0,
+      "learning_rate": 3.4562735439894e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.88214111328125,
+      "num_tokens": 190876871.0,
+      "step": 1785
+    },
+    {
+      "epoch": 4.068415051311288,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.4545330572239234e-06,
+      "loss": 0.6067,
+      "mean_token_accuracy": 0.8759456723928452,
+      "num_tokens": 190983692.0,
+      "step": 1786
+    },
+    {
+      "epoch": 4.070695553021665,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.452792028683912e-06,
+      "loss": 0.59,
+      "mean_token_accuracy": 0.8788794428110123,
+      "num_tokens": 191090686.0,
+      "step": 1787
+    },
+    {
+      "epoch": 4.072976054732041,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.4510504593575396e-06,
+      "loss": 0.5766,
+      "mean_token_accuracy": 0.8841027617454529,
+      "num_tokens": 191197679.0,
+      "step": 1788
+    },
+    {
+      "epoch": 4.075256556442417,
+      "grad_norm": 3.96875,
+      "learning_rate": 3.449308350233287e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8803805410861969,
+      "num_tokens": 191305286.0,
+      "step": 1789
+    },
+    {
+      "epoch": 4.077537058152793,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.447565702299942e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8774025738239288,
+      "num_tokens": 191412336.0,
+      "step": 1790
+    },
+    {
+      "epoch": 4.07981755986317,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.445822516546598e-06,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.881226509809494,
+      "num_tokens": 191519740.0,
+      "step": 1791
+    },
+    {
+      "epoch": 4.0820980615735465,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.444078793962653e-06,
+      "loss": 0.5699,
+      "mean_token_accuracy": 0.8817672431468964,
+      "num_tokens": 191626597.0,
+      "step": 1792
+    },
+    {
+      "epoch": 4.084378563283923,
+      "grad_norm": 2.328125,
+      "learning_rate": 3.4423345355378114e-06,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8822430968284607,
+      "num_tokens": 191733873.0,
+      "step": 1793
+    },
+    {
+      "epoch": 4.086659064994299,
+      "grad_norm": 4.625,
+      "learning_rate": 3.440589742262079e-06,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8803035467863083,
+      "num_tokens": 191840789.0,
+      "step": 1794
+    },
+    {
+      "epoch": 4.088939566704675,
+      "grad_norm": 5.25,
+      "learning_rate": 3.438844415125768e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8835125714540482,
+      "num_tokens": 191947897.0,
+      "step": 1795
+    },
+    {
+      "epoch": 4.091220068415051,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.437098555119493e-06,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8822659403085709,
+      "num_tokens": 192055072.0,
+      "step": 1796
+    },
+    {
+      "epoch": 4.0935005701254275,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.4353521632341686e-06,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8791303038597107,
+      "num_tokens": 192162000.0,
+      "step": 1797
+    },
+    {
+      "epoch": 4.095781071835804,
+      "grad_norm": 7.75,
+      "learning_rate": 3.4336052404610138e-06,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8811719119548798,
+      "num_tokens": 192269139.0,
+      "step": 1798
+    },
+    {
+      "epoch": 4.09806157354618,
+      "grad_norm": 8.6875,
+      "learning_rate": 3.431857787791549e-06,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.8792611956596375,
+      "num_tokens": 192376137.0,
+      "step": 1799
+    },
+    {
+      "epoch": 4.100342075256556,
+      "grad_norm": 6.96875,
+      "learning_rate": 3.4301098062175936e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.8786915689706802,
+      "num_tokens": 192483573.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.102622576966933,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.4283612967312692e-06,
+      "loss": 0.6022,
+      "mean_token_accuracy": 0.8749962151050568,
+      "num_tokens": 192590098.0,
+      "step": 1801
+    },
+    {
+      "epoch": 4.104903078677309,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.426612260324996e-06,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8811412006616592,
+      "num_tokens": 192696798.0,
+      "step": 1802
+    },
+    {
+      "epoch": 4.1071835803876855,
+      "grad_norm": 4.84375,
+      "learning_rate": 3.424862697991491e-06,
+      "loss": 0.5495,
+      "mean_token_accuracy": 0.8869053721427917,
+      "num_tokens": 192804017.0,
+      "step": 1803
+    },
+    {
+      "epoch": 4.109464082098062,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.4231126107237754e-06,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8787698745727539,
+      "num_tokens": 192912495.0,
+      "step": 1804
+    },
+    {
+      "epoch": 4.111744583808438,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.4213619995151628e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.8815167546272278,
+      "num_tokens": 193019325.0,
+      "step": 1805
+    },
+    {
+      "epoch": 4.114025085518814,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.4196108653592662e-06,
+      "loss": 0.5571,
+      "mean_token_accuracy": 0.8847940266132355,
+      "num_tokens": 193126418.0,
+      "step": 1806
+    },
+    {
+      "epoch": 4.11630558722919,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.417859209249997e-06,
+      "loss": 0.5618,
+      "mean_token_accuracy": 0.8841525167226791,
+      "num_tokens": 193233113.0,
+      "step": 1807
+    },
+    {
+      "epoch": 4.1185860889395665,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.4161070321815605e-06,
+      "loss": 0.5985,
+      "mean_token_accuracy": 0.8764221370220184,
+      "num_tokens": 193339988.0,
+      "step": 1808
+    },
+    {
+      "epoch": 4.120866590649943,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.4143543351484585e-06,
+      "loss": 0.586,
+      "mean_token_accuracy": 0.8799712508916855,
+      "num_tokens": 193446945.0,
+      "step": 1809
+    },
+    {
+      "epoch": 4.123147092360319,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.4126011191454877e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8829108774662018,
+      "num_tokens": 193553550.0,
+      "step": 1810
+    },
+    {
+      "epoch": 4.125427594070696,
+      "grad_norm": 5.15625,
+      "learning_rate": 3.4108473851677408e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8795969039201736,
+      "num_tokens": 193660308.0,
+      "step": 1811
+    },
+    {
+      "epoch": 4.127708095781072,
+      "grad_norm": 3.375,
+      "learning_rate": 3.4090931342106024e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8809874951839447,
+      "num_tokens": 193768000.0,
+      "step": 1812
+    },
+    {
+      "epoch": 4.129988597491448,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.4073383672697524e-06,
+      "loss": 0.5897,
+      "mean_token_accuracy": 0.8771644234657288,
+      "num_tokens": 193875084.0,
+      "step": 1813
+    },
+    {
+      "epoch": 4.1322690992018245,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.4055830853411616e-06,
+      "loss": 0.6109,
+      "mean_token_accuracy": 0.8745006024837494,
+      "num_tokens": 193981792.0,
+      "step": 1814
+    },
+    {
+      "epoch": 4.134549600912201,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.4038272894210945e-06,
+      "loss": 0.5889,
+      "mean_token_accuracy": 0.8784460872411728,
+      "num_tokens": 194088451.0,
+      "step": 1815
+    },
+    {
+      "epoch": 4.136830102622577,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.4020709805061066e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8836447149515152,
+      "num_tokens": 194195882.0,
+      "step": 1816
+    },
+    {
+      "epoch": 4.139110604332953,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.4003141595930456e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8817628026008606,
+      "num_tokens": 194302902.0,
+      "step": 1817
+    },
+    {
+      "epoch": 4.141391106043329,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.3985568276790487e-06,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8798153549432755,
+      "num_tokens": 194410857.0,
+      "step": 1818
+    },
+    {
+      "epoch": 4.1436716077537055,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.3967989857615434e-06,
+      "loss": 0.589,
+      "mean_token_accuracy": 0.8804911673069,
+      "num_tokens": 194517991.0,
+      "step": 1819
+    },
+    {
+      "epoch": 4.145952109464082,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.3950406348382483e-06,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.8826321363449097,
+      "num_tokens": 194624664.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.148232611174459,
+      "grad_norm": 4.375,
+      "learning_rate": 3.3932817759071666e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.8814938217401505,
+      "num_tokens": 194731631.0,
+      "step": 1821
+    },
+    {
+      "epoch": 4.150513112884835,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.3915224099665962e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8795495629310608,
+      "num_tokens": 194838909.0,
+      "step": 1822
+    },
+    {
+      "epoch": 4.152793614595211,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.389762538015116e-06,
+      "loss": 0.5996,
+      "mean_token_accuracy": 0.8770654946565628,
+      "num_tokens": 194945773.0,
+      "step": 1823
+    },
+    {
+      "epoch": 4.155074116305587,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.388002161051598e-06,
+      "loss": 0.5975,
+      "mean_token_accuracy": 0.8804232776165009,
+      "num_tokens": 195052653.0,
+      "step": 1824
+    },
+    {
+      "epoch": 4.1573546180159635,
+      "grad_norm": 4.625,
+      "learning_rate": 3.3862412800751963e-06,
+      "loss": 0.595,
+      "mean_token_accuracy": 0.8783848881721497,
+      "num_tokens": 195159206.0,
+      "step": 1825
+    },
+    {
+      "epoch": 4.15963511972634,
+      "grad_norm": 5.15625,
+      "learning_rate": 3.3844798960853533e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.879721149802208,
+      "num_tokens": 195266220.0,
+      "step": 1826
+    },
+    {
+      "epoch": 4.161915621436716,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.382718010081797e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8808034509420395,
+      "num_tokens": 195373432.0,
+      "step": 1827
+    },
+    {
+      "epoch": 4.164196123147092,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.38095562306454e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8810047507286072,
+      "num_tokens": 195480280.0,
+      "step": 1828
+    },
+    {
+      "epoch": 4.166476624857468,
+      "grad_norm": 2.28125,
+      "learning_rate": 3.3791927360338785e-06,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8826696872711182,
+      "num_tokens": 195587648.0,
+      "step": 1829
+    },
+    {
+      "epoch": 4.168757126567845,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.3774293499903934e-06,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.8804333359003067,
+      "num_tokens": 195694201.0,
+      "step": 1830
+    },
+    {
+      "epoch": 4.1710376282782216,
+      "grad_norm": 4.53125,
+      "learning_rate": 3.3756654659349487e-06,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.879314973950386,
+      "num_tokens": 195800840.0,
+      "step": 1831
+    },
+    {
+      "epoch": 4.173318129988598,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.373901084868691e-06,
+      "loss": 0.5692,
+      "mean_token_accuracy": 0.8824407607316971,
+      "num_tokens": 195908050.0,
+      "step": 1832
+    },
+    {
+      "epoch": 4.175598631698974,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.372136207793049e-06,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8842916339635849,
+      "num_tokens": 196014977.0,
+      "step": 1833
+    },
+    {
+      "epoch": 4.17787913340935,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.3703708357097333e-06,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8835365027189255,
+      "num_tokens": 196122033.0,
+      "step": 1834
+    },
+    {
+      "epoch": 4.180159635119726,
+      "grad_norm": 5.25,
+      "learning_rate": 3.3686049696207336e-06,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8810452073812485,
+      "num_tokens": 196228948.0,
+      "step": 1835
+    },
+    {
+      "epoch": 4.1824401368301025,
+      "grad_norm": 4.90625,
+      "learning_rate": 3.3668386105283226e-06,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.882371187210083,
+      "num_tokens": 196335775.0,
+      "step": 1836
+    },
+    {
+      "epoch": 4.184720638540479,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.365071759435051e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8814999759197235,
+      "num_tokens": 196442967.0,
+      "step": 1837
+    },
+    {
+      "epoch": 4.187001140250855,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.363304417343749e-06,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8800509870052338,
+      "num_tokens": 196550345.0,
+      "step": 1838
+    },
+    {
+      "epoch": 4.189281641961231,
+      "grad_norm": 2.625,
+      "learning_rate": 3.3615365852575276e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8801786154508591,
+      "num_tokens": 196658198.0,
+      "step": 1839
+    },
+    {
+      "epoch": 4.191562143671608,
+      "grad_norm": 5.03125,
+      "learning_rate": 3.359768264179772e-06,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8778839707374573,
+      "num_tokens": 196765684.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.193842645381984,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.357999455114148e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8802204430103302,
+      "num_tokens": 196872724.0,
+      "step": 1841
+    },
+    {
+      "epoch": 4.196123147092361,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.356230159064599e-06,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.8841566443443298,
+      "num_tokens": 196979553.0,
+      "step": 1842
+    },
+    {
+      "epoch": 4.198403648802737,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.3544603770353407e-06,
+      "loss": 0.5583,
+      "mean_token_accuracy": 0.8825344890356064,
+      "num_tokens": 197086673.0,
+      "step": 1843
+    },
+    {
+      "epoch": 4.200684150513113,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.352690110030869e-06,
+      "loss": 0.5991,
+      "mean_token_accuracy": 0.8824966847896576,
+      "num_tokens": 197193889.0,
+      "step": 1844
+    },
+    {
+      "epoch": 4.202964652223489,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.350919359055953e-06,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8796708136796951,
+      "num_tokens": 197301220.0,
+      "step": 1845
+    },
+    {
+      "epoch": 4.205245153933865,
+      "grad_norm": 3.0,
+      "learning_rate": 3.3491481251156355e-06,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.8849449008703232,
+      "num_tokens": 197408502.0,
+      "step": 1846
+    },
+    {
+      "epoch": 4.2075256556442415,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.347376409215236e-06,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8793286383152008,
+      "num_tokens": 197515859.0,
+      "step": 1847
+    },
+    {
+      "epoch": 4.209806157354618,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.345604212360346e-06,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8793947696685791,
+      "num_tokens": 197622609.0,
+      "step": 1848
+    },
+    {
+      "epoch": 4.212086659064994,
+      "grad_norm": 4.5,
+      "learning_rate": 3.3438315355568295e-06,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.8782702833414078,
+      "num_tokens": 197729199.0,
+      "step": 1849
+    },
+    {
+      "epoch": 4.214367160775371,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.3420583798108253e-06,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8802850246429443,
+      "num_tokens": 197835966.0,
+      "step": 1850
+    },
+    {
+      "epoch": 4.216647662485747,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.34028474612874e-06,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8822648972272873,
+      "num_tokens": 197942829.0,
+      "step": 1851
+    },
+    {
+      "epoch": 4.218928164196123,
+      "grad_norm": 5.375,
+      "learning_rate": 3.338510635517256e-06,
+      "loss": 0.6037,
+      "mean_token_accuracy": 0.8766555935144424,
+      "num_tokens": 198050288.0,
+      "step": 1852
+    },
+    {
+      "epoch": 4.2212086659065,
+      "grad_norm": 6.46875,
+      "learning_rate": 3.3367360489833236e-06,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8814980983734131,
+      "num_tokens": 198158005.0,
+      "step": 1853
+    },
+    {
+      "epoch": 4.223489167616876,
+      "grad_norm": 8.0625,
+      "learning_rate": 3.3349609875341626e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.882731556892395,
+      "num_tokens": 198265125.0,
+      "step": 1854
+    },
+    {
+      "epoch": 4.225769669327252,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.3331854521772656e-06,
+      "loss": 0.558,
+      "mean_token_accuracy": 0.8883755952119827,
+      "num_tokens": 198371867.0,
+      "step": 1855
+    },
+    {
+      "epoch": 4.228050171037628,
+      "grad_norm": 2.375,
+      "learning_rate": 3.3314094439203903e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8845670074224472,
+      "num_tokens": 198478602.0,
+      "step": 1856
+    },
+    {
+      "epoch": 4.230330672748004,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.3296329637715662e-06,
+      "loss": 0.5775,
+      "mean_token_accuracy": 0.8805161267518997,
+      "num_tokens": 198585016.0,
+      "step": 1857
+    },
+    {
+      "epoch": 4.2326111744583805,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.3278560127390892e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8802484422922134,
+      "num_tokens": 198692273.0,
+      "step": 1858
+    },
+    {
+      "epoch": 4.234891676168757,
+      "grad_norm": 4.6875,
+      "learning_rate": 3.32607859183152e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8806602954864502,
+      "num_tokens": 198798563.0,
+      "step": 1859
+    },
+    {
+      "epoch": 4.237172177879134,
+      "grad_norm": 2.34375,
+      "learning_rate": 3.3243007020576917e-06,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.8770420551300049,
+      "num_tokens": 198905412.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.23945267958951,
+      "grad_norm": 6.125,
+      "learning_rate": 3.322522344426698e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.883165031671524,
+      "num_tokens": 199012136.0,
+      "step": 1861
+    },
+    {
+      "epoch": 4.241733181299886,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.320743519947901e-06,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.880427822470665,
+      "num_tokens": 199119481.0,
+      "step": 1862
+    },
+    {
+      "epoch": 4.244013683010262,
+      "grad_norm": 6.875,
+      "learning_rate": 3.318964229630927e-06,
+      "loss": 0.5902,
+      "mean_token_accuracy": 0.8777157664299011,
+      "num_tokens": 199226442.0,
+      "step": 1863
+    },
+    {
+      "epoch": 4.246294184720639,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.3171844744856675e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8849682509899139,
+      "num_tokens": 199333370.0,
+      "step": 1864
+    },
+    {
+      "epoch": 4.248574686431015,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.3154042555222758e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8795311897993088,
+      "num_tokens": 199440819.0,
+      "step": 1865
+    },
+    {
+      "epoch": 4.250855188141391,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.3136235737511715e-06,
+      "loss": 0.589,
+      "mean_token_accuracy": 0.8744902908802032,
+      "num_tokens": 199547464.0,
+      "step": 1866
+    },
+    {
+      "epoch": 4.253135689851767,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.3118424301830343e-06,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8773903250694275,
+      "num_tokens": 199654391.0,
+      "step": 1867
+    },
+    {
+      "epoch": 4.255416191562143,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.310060825828807e-06,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8800706118345261,
+      "num_tokens": 199761255.0,
+      "step": 1868
+    },
+    {
+      "epoch": 4.2576966932725195,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.3082787616996938e-06,
+      "loss": 0.5681,
+      "mean_token_accuracy": 0.8814914971590042,
+      "num_tokens": 199868966.0,
+      "step": 1869
+    },
+    {
+      "epoch": 4.259977194982897,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.3064962388071586e-06,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8803159892559052,
+      "num_tokens": 199975974.0,
+      "step": 1870
+    },
+    {
+      "epoch": 4.262257696693273,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.3047132581629297e-06,
+      "loss": 0.5617,
+      "mean_token_accuracy": 0.8839969784021378,
+      "num_tokens": 200083683.0,
+      "step": 1871
+    },
+    {
+      "epoch": 4.264538198403649,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.3029298207789907e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8819585740566254,
+      "num_tokens": 200191486.0,
+      "step": 1872
+    },
+    {
+      "epoch": 4.266818700114025,
+      "grad_norm": 6.03125,
+      "learning_rate": 3.301145927667586e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8768231570720673,
+      "num_tokens": 200298545.0,
+      "step": 1873
+    },
+    {
+      "epoch": 4.269099201824401,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.2993615798412204e-06,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.8820153176784515,
+      "num_tokens": 200405570.0,
+      "step": 1874
+    },
+    {
+      "epoch": 4.271379703534778,
+      "grad_norm": 4.9375,
+      "learning_rate": 3.297576778312654e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8819970637559891,
+      "num_tokens": 200512849.0,
+      "step": 1875
+    },
+    {
+      "epoch": 4.273660205245154,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.295791524094906e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.878609761595726,
+      "num_tokens": 200620132.0,
+      "step": 1876
+    },
+    {
+      "epoch": 4.27594070695553,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.294005818201252e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8842368423938751,
+      "num_tokens": 200726719.0,
+      "step": 1877
+    },
+    {
+      "epoch": 4.278221208665906,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.2922196616452253e-06,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.8771747201681137,
+      "num_tokens": 200833547.0,
+      "step": 1878
+    },
+    {
+      "epoch": 4.280501710376283,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.2904330554406126e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.880841001868248,
+      "num_tokens": 200941003.0,
+      "step": 1879
+    },
+    {
+      "epoch": 4.282782212086659,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.288646000601457e-06,
+      "loss": 0.5993,
+      "mean_token_accuracy": 0.8747004866600037,
+      "num_tokens": 201048114.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.285062713797036,
+      "grad_norm": 5.0625,
+      "learning_rate": 3.286858498142057e-06,
+      "loss": 0.5821,
+      "mean_token_accuracy": 0.8779664039611816,
+      "num_tokens": 201155149.0,
+      "step": 1881
+    },
+    {
+      "epoch": 4.287343215507412,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.285070549076965e-06,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.877086952328682,
+      "num_tokens": 201262291.0,
+      "step": 1882
+    },
+    {
+      "epoch": 4.289623717217788,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.283282154420985e-06,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8825719207525253,
+      "num_tokens": 201369699.0,
+      "step": 1883
+    },
+    {
+      "epoch": 4.291904218928164,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.2814933151891766e-06,
+      "loss": 0.5674,
+      "mean_token_accuracy": 0.881410762667656,
+      "num_tokens": 201476718.0,
+      "step": 1884
+    },
+    {
+      "epoch": 4.29418472063854,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.2797040323968493e-06,
+      "loss": 0.5776,
+      "mean_token_accuracy": 0.8789144903421402,
+      "num_tokens": 201583789.0,
+      "step": 1885
+    },
+    {
+      "epoch": 4.296465222348917,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.277914307059566e-06,
+      "loss": 0.5567,
+      "mean_token_accuracy": 0.886613667011261,
+      "num_tokens": 201690953.0,
+      "step": 1886
+    },
+    {
+      "epoch": 4.298745724059293,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.276124140193141e-06,
+      "loss": 0.5685,
+      "mean_token_accuracy": 0.882378563284874,
+      "num_tokens": 201798723.0,
+      "step": 1887
+    },
+    {
+      "epoch": 4.301026225769669,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.274333532813637e-06,
+      "loss": 0.6009,
+      "mean_token_accuracy": 0.8767693191766739,
+      "num_tokens": 201905583.0,
+      "step": 1888
+    },
+    {
+      "epoch": 4.303306727480045,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.272542485937369e-06,
+      "loss": 0.5607,
+      "mean_token_accuracy": 0.89011649787426,
+      "num_tokens": 202012658.0,
+      "step": 1889
+    },
+    {
+      "epoch": 4.305587229190422,
+      "grad_norm": 2.5,
+      "learning_rate": 3.2707510005809005e-06,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8810762017965317,
+      "num_tokens": 202120313.0,
+      "step": 1890
+    },
+    {
+      "epoch": 4.307867730900798,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.2689590777610443e-06,
+      "loss": 0.5833,
+      "mean_token_accuracy": 0.8805640786886215,
+      "num_tokens": 202227525.0,
+      "step": 1891
+    },
+    {
+      "epoch": 4.310148232611175,
+      "grad_norm": 6.125,
+      "learning_rate": 3.267166718494861e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8817348927259445,
+      "num_tokens": 202334247.0,
+      "step": 1892
+    },
+    {
+      "epoch": 4.312428734321551,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.265373923799658e-06,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8808832913637161,
+      "num_tokens": 202441565.0,
+      "step": 1893
+    },
+    {
+      "epoch": 4.314709236031927,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.263580694692992e-06,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.882627323269844,
+      "num_tokens": 202549008.0,
+      "step": 1894
+    },
+    {
+      "epoch": 4.316989737742303,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.261787032192666e-06,
+      "loss": 0.5842,
+      "mean_token_accuracy": 0.8810908943414688,
+      "num_tokens": 202655622.0,
+      "step": 1895
+    },
+    {
+      "epoch": 4.319270239452679,
+      "grad_norm": 4.625,
+      "learning_rate": 3.259992937316727e-06,
+      "loss": 0.5903,
+      "mean_token_accuracy": 0.882026731967926,
+      "num_tokens": 202761692.0,
+      "step": 1896
+    },
+    {
+      "epoch": 4.321550741163056,
+      "grad_norm": 5.1875,
+      "learning_rate": 3.258198411083469e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8775666505098343,
+      "num_tokens": 202868233.0,
+      "step": 1897
+    },
+    {
+      "epoch": 4.323831242873432,
+      "grad_norm": 8.4375,
+      "learning_rate": 3.2564034545114308e-06,
+      "loss": 0.5866,
+      "mean_token_accuracy": 0.8776109367609024,
+      "num_tokens": 202975704.0,
+      "step": 1898
+    },
+    {
+      "epoch": 4.326111744583809,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.2546080686193947e-06,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8826450109481812,
+      "num_tokens": 203082943.0,
+      "step": 1899
+    },
+    {
+      "epoch": 4.328392246294185,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.2528122544263873e-06,
+      "loss": 0.5614,
+      "mean_token_accuracy": 0.8836265951395035,
+      "num_tokens": 203189879.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.330672748004561,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.251016012951678e-06,
+      "loss": 0.5576,
+      "mean_token_accuracy": 0.8853756338357925,
+      "num_tokens": 203296924.0,
+      "step": 1901
+    },
+    {
+      "epoch": 4.3329532497149374,
+      "grad_norm": 8.4375,
+      "learning_rate": 3.2492193452147774e-06,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.8769858479499817,
+      "num_tokens": 203403322.0,
+      "step": 1902
+    },
+    {
+      "epoch": 4.335233751425314,
+      "grad_norm": 5.5,
+      "learning_rate": 3.247422252235442e-06,
+      "loss": 0.5485,
+      "mean_token_accuracy": 0.8848748654127121,
+      "num_tokens": 203510621.0,
+      "step": 1903
+    },
+    {
+      "epoch": 4.33751425313569,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.245624735033665e-06,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8844299167394638,
+      "num_tokens": 203617576.0,
+      "step": 1904
+    },
+    {
+      "epoch": 4.339794754846066,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.2438267946296836e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8842675089836121,
+      "num_tokens": 203724726.0,
+      "step": 1905
+    },
+    {
+      "epoch": 4.342075256556442,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.242028432043974e-06,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8774498254060745,
+      "num_tokens": 203832260.0,
+      "step": 1906
+    },
+    {
+      "epoch": 4.344355758266818,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.2402296482972513e-06,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8836204558610916,
+      "num_tokens": 203940076.0,
+      "step": 1907
+    },
+    {
+      "epoch": 4.346636259977195,
+      "grad_norm": 7.75,
+      "learning_rate": 3.238430444410471e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8820386677980423,
+      "num_tokens": 204047654.0,
+      "step": 1908
+    },
+    {
+      "epoch": 4.348916761687571,
+      "grad_norm": 5.09375,
+      "learning_rate": 3.2366308214048262e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8806447833776474,
+      "num_tokens": 204154623.0,
+      "step": 1909
+    },
+    {
+      "epoch": 4.351197263397948,
+      "grad_norm": 5.03125,
+      "learning_rate": 3.2348307803017493e-06,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.886213019490242,
+      "num_tokens": 204261536.0,
+      "step": 1910
+    },
+    {
+      "epoch": 4.353477765108324,
+      "grad_norm": 4.6875,
+      "learning_rate": 3.2330303221229078e-06,
+      "loss": 0.5915,
+      "mean_token_accuracy": 0.8766489624977112,
+      "num_tokens": 204368307.0,
+      "step": 1911
+    },
+    {
+      "epoch": 4.3557582668187,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.231229447890206e-06,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8825319856405258,
+      "num_tokens": 204475273.0,
+      "step": 1912
+    },
+    {
+      "epoch": 4.3580387685290765,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.229428158625787e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8858949691057205,
+      "num_tokens": 204582592.0,
+      "step": 1913
+    },
+    {
+      "epoch": 4.360319270239453,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.2276264553520275e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8835407346487045,
+      "num_tokens": 204689248.0,
+      "step": 1914
+    },
+    {
+      "epoch": 4.362599771949829,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.2258243390915397e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8794312477111816,
+      "num_tokens": 204796588.0,
+      "step": 1915
+    },
+    {
+      "epoch": 4.364880273660205,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.2240218108671683e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8816234618425369,
+      "num_tokens": 204903575.0,
+      "step": 1916
+    },
+    {
+      "epoch": 4.367160775370581,
+      "grad_norm": 4.625,
+      "learning_rate": 3.2222188717019965e-06,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.8821093142032623,
+      "num_tokens": 205010359.0,
+      "step": 1917
+    },
+    {
+      "epoch": 4.369441277080957,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.220415522619335e-06,
+      "loss": 0.5984,
+      "mean_token_accuracy": 0.8760864734649658,
+      "num_tokens": 205117682.0,
+      "step": 1918
+    },
+    {
+      "epoch": 4.3717217787913345,
+      "grad_norm": 2.34375,
+      "learning_rate": 3.218611764642732e-06,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8813271969556808,
+      "num_tokens": 205224758.0,
+      "step": 1919
+    },
+    {
+      "epoch": 4.374002280501711,
+      "grad_norm": 3.375,
+      "learning_rate": 3.2168075987959633e-06,
+      "loss": 0.6073,
+      "mean_token_accuracy": 0.8754995763301849,
+      "num_tokens": 205331474.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.376282782212087,
+      "grad_norm": 4.0,
+      "learning_rate": 3.2150030261030414e-06,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8807307332754135,
+      "num_tokens": 205438379.0,
+      "step": 1921
+    },
+    {
+      "epoch": 4.378563283922463,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.2131980475882053e-06,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.8842781186103821,
+      "num_tokens": 205545207.0,
+      "step": 1922
+    },
+    {
+      "epoch": 4.380843785632839,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.2113926642759256e-06,
+      "loss": 0.605,
+      "mean_token_accuracy": 0.8773124665021896,
+      "num_tokens": 205651526.0,
+      "step": 1923
+    },
+    {
+      "epoch": 4.3831242873432155,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.2095868771909037e-06,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8833547234535217,
+      "num_tokens": 205758217.0,
+      "step": 1924
+    },
+    {
+      "epoch": 4.385404789053592,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.2077806873580696e-06,
+      "loss": 0.5551,
+      "mean_token_accuracy": 0.8848164230585098,
+      "num_tokens": 205865218.0,
+      "step": 1925
+    },
+    {
+      "epoch": 4.387685290763968,
+      "grad_norm": 5.25,
+      "learning_rate": 3.205974095802582e-06,
+      "loss": 0.5773,
+      "mean_token_accuracy": 0.8819639384746552,
+      "num_tokens": 205971573.0,
+      "step": 1926
+    },
+    {
+      "epoch": 4.389965792474344,
+      "grad_norm": 6.0,
+      "learning_rate": 3.204167103549827e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8817443251609802,
+      "num_tokens": 206078630.0,
+      "step": 1927
+    },
+    {
+      "epoch": 4.39224629418472,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.2023597116254175e-06,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8799740672111511,
+      "num_tokens": 206185346.0,
+      "step": 1928
+    },
+    {
+      "epoch": 4.394526795895097,
+      "grad_norm": 2.625,
+      "learning_rate": 3.2005519210551955e-06,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.885019987821579,
+      "num_tokens": 206292483.0,
+      "step": 1929
+    },
+    {
+      "epoch": 4.3968072976054735,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.1987437328652287e-06,
+      "loss": 0.5626,
+      "mean_token_accuracy": 0.8824569880962372,
+      "num_tokens": 206399488.0,
+      "step": 1930
+    },
+    {
+      "epoch": 4.39908779931585,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.196935148081808e-06,
+      "loss": 0.5953,
+      "mean_token_accuracy": 0.8730659335851669,
+      "num_tokens": 206506633.0,
+      "step": 1931
+    },
+    {
+      "epoch": 4.401368301026226,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.1951261677314526e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8806923627853394,
+      "num_tokens": 206614034.0,
+      "step": 1932
+    },
+    {
+      "epoch": 4.403648802736602,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.1933167928409046e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.882137656211853,
+      "num_tokens": 206721766.0,
+      "step": 1933
+    },
+    {
+      "epoch": 4.405929304446978,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.1915070244371295e-06,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8834485858678818,
+      "num_tokens": 206829226.0,
+      "step": 1934
+    },
+    {
+      "epoch": 4.4082098061573545,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.1896968635473174e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8820080310106277,
+      "num_tokens": 206936587.0,
+      "step": 1935
+    },
+    {
+      "epoch": 4.410490307867731,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.187886311198881e-06,
+      "loss": 0.5638,
+      "mean_token_accuracy": 0.8846455514431,
+      "num_tokens": 207043902.0,
+      "step": 1936
+    },
+    {
+      "epoch": 4.412770809578107,
+      "grad_norm": 4.75,
+      "learning_rate": 3.1860753684194536e-06,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8813911080360413,
+      "num_tokens": 207151288.0,
+      "step": 1937
+    },
+    {
+      "epoch": 4.415051311288483,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.1842640362368932e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.878994882106781,
+      "num_tokens": 207258999.0,
+      "step": 1938
+    },
+    {
+      "epoch": 4.41733181299886,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.182452315679276e-06,
+      "loss": 0.545,
+      "mean_token_accuracy": 0.8886394798755646,
+      "num_tokens": 207366568.0,
+      "step": 1939
+    },
+    {
+      "epoch": 4.419612314709236,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.1806402077748987e-06,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.8816164880990982,
+      "num_tokens": 207473843.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.4218928164196125,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.178827713552281e-06,
+      "loss": 0.5808,
+      "mean_token_accuracy": 0.8799097836017609,
+      "num_tokens": 207580995.0,
+      "step": 1941
+    },
+    {
+      "epoch": 4.424173318129989,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.177014834040158e-06,
+      "loss": 0.5652,
+      "mean_token_accuracy": 0.883696660399437,
+      "num_tokens": 207688161.0,
+      "step": 1942
+    },
+    {
+      "epoch": 4.426453819840365,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.1752015702674855e-06,
+      "loss": 0.6015,
+      "mean_token_accuracy": 0.8793037235736847,
+      "num_tokens": 207795164.0,
+      "step": 1943
+    },
+    {
+      "epoch": 4.428734321550741,
+      "grad_norm": 2.5,
+      "learning_rate": 3.173387923263437e-06,
+      "loss": 0.5574,
+      "mean_token_accuracy": 0.8862407803535461,
+      "num_tokens": 207902155.0,
+      "step": 1944
+    },
+    {
+      "epoch": 4.431014823261117,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.1715738940574032e-06,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8828154355287552,
+      "num_tokens": 208009297.0,
+      "step": 1945
+    },
+    {
+      "epoch": 4.4332953249714935,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.1697594836789924e-06,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8778323084115982,
+      "num_tokens": 208116608.0,
+      "step": 1946
+    },
+    {
+      "epoch": 4.43557582668187,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.167944693158029e-06,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8808430731296539,
+      "num_tokens": 208223390.0,
+      "step": 1947
+    },
+    {
+      "epoch": 4.437856328392247,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.166129523524553e-06,
+      "loss": 0.6007,
+      "mean_token_accuracy": 0.8762641549110413,
+      "num_tokens": 208330138.0,
+      "step": 1948
+    },
+    {
+      "epoch": 4.440136830102623,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.1643139758088194e-06,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.8787393867969513,
+      "num_tokens": 208437552.0,
+      "step": 1949
+    },
+    {
+      "epoch": 4.442417331812999,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.1624980510412984e-06,
+      "loss": 0.5987,
+      "mean_token_accuracy": 0.8784570544958115,
+      "num_tokens": 208544652.0,
+      "step": 1950
+    },
+    {
+      "epoch": 4.444697833523375,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.160681750252674e-06,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8795022666454315,
+      "num_tokens": 208651498.0,
+      "step": 1951
+    },
+    {
+      "epoch": 4.4469783352337515,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.1588650744738418e-06,
+      "loss": 0.5617,
+      "mean_token_accuracy": 0.8832292854785919,
+      "num_tokens": 208758479.0,
+      "step": 1952
+    },
+    {
+      "epoch": 4.449258836944128,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.1570480247359147e-06,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8820718824863434,
+      "num_tokens": 208865610.0,
+      "step": 1953
+    },
+    {
+      "epoch": 4.451539338654504,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.155230602070213e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.879115030169487,
+      "num_tokens": 208972259.0,
+      "step": 1954
+    },
+    {
+      "epoch": 4.45381984036488,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.153412807508271e-06,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8792910128831863,
+      "num_tokens": 209078252.0,
+      "step": 1955
+    },
+    {
+      "epoch": 4.456100342075256,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.1515946420818343e-06,
+      "loss": 0.5557,
+      "mean_token_accuracy": 0.8846887648105621,
+      "num_tokens": 209185333.0,
+      "step": 1956
+    },
+    {
+      "epoch": 4.4583808437856325,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.1497761068228585e-06,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8841688334941864,
+      "num_tokens": 209292465.0,
+      "step": 1957
+    },
+    {
+      "epoch": 4.460661345496009,
+      "grad_norm": 3.5,
+      "learning_rate": 3.1479572027635085e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8820602297782898,
+      "num_tokens": 209399257.0,
+      "step": 1958
+    },
+    {
+      "epoch": 4.462941847206386,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.1461379309361594e-06,
+      "loss": 0.5927,
+      "mean_token_accuracy": 0.8786464035511017,
+      "num_tokens": 209506400.0,
+      "step": 1959
+    },
+    {
+      "epoch": 4.465222348916762,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.144318292373395e-06,
+      "loss": 0.5487,
+      "mean_token_accuracy": 0.8865722268819809,
+      "num_tokens": 209613571.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.467502850627138,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.142498288108007e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8826761692762375,
+      "num_tokens": 209720956.0,
+      "step": 1961
+    },
+    {
+      "epoch": 4.469783352337514,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.1406779191729954e-06,
+      "loss": 0.5548,
+      "mean_token_accuracy": 0.8883082419633865,
+      "num_tokens": 209827663.0,
+      "step": 1962
+    },
+    {
+      "epoch": 4.4720638540478905,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.1388571866015645e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8810272365808487,
+      "num_tokens": 209934611.0,
+      "step": 1963
+    },
+    {
+      "epoch": 4.474344355758267,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.1370360914271286e-06,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8840647041797638,
+      "num_tokens": 210041595.0,
+      "step": 1964
+    },
+    {
+      "epoch": 4.476624857468643,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.1352146346833057e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.877694696187973,
+      "num_tokens": 210148918.0,
+      "step": 1965
+    },
+    {
+      "epoch": 4.478905359179019,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.133392817403919e-06,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8824283182621002,
+      "num_tokens": 210256121.0,
+      "step": 1966
+    },
+    {
+      "epoch": 4.481185860889395,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.131570640622998e-06,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8796381056308746,
+      "num_tokens": 210363033.0,
+      "step": 1967
+    },
+    {
+      "epoch": 4.483466362599772,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.1297481053747737e-06,
+      "loss": 0.5526,
+      "mean_token_accuracy": 0.8865970373153687,
+      "num_tokens": 210470814.0,
+      "step": 1968
+    },
+    {
+      "epoch": 4.485746864310149,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.127925212693682e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8791713118553162,
+      "num_tokens": 210577490.0,
+      "step": 1969
+    },
+    {
+      "epoch": 4.488027366020525,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.1261019636143636e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8812829405069351,
+      "num_tokens": 210684376.0,
+      "step": 1970
+    },
+    {
+      "epoch": 4.490307867730901,
+      "grad_norm": 2.484375,
+      "learning_rate": 3.124278359171657e-06,
+      "loss": 0.5712,
+      "mean_token_accuracy": 0.8833597153425217,
+      "num_tokens": 210791620.0,
+      "step": 1971
+    },
+    {
+      "epoch": 4.492588369441277,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.122454400400606e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8824751228094101,
+      "num_tokens": 210898333.0,
+      "step": 1972
+    },
+    {
+      "epoch": 4.494868871151653,
+      "grad_norm": 2.34375,
+      "learning_rate": 3.1206300883364547e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8802805244922638,
+      "num_tokens": 211005919.0,
+      "step": 1973
+    },
+    {
+      "epoch": 4.4971493728620295,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.1188054240146463e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.880774512887001,
+      "num_tokens": 211113271.0,
+      "step": 1974
+    },
+    {
+      "epoch": 4.499429874572406,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.1169804084708267e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.880780965089798,
+      "num_tokens": 211220394.0,
+      "step": 1975
+    },
+    {
+      "epoch": 4.501710376282782,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.1151550427408383e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8800267279148102,
+      "num_tokens": 211327515.0,
+      "step": 1976
+    },
+    {
+      "epoch": 4.503990877993158,
+      "grad_norm": 4.46875,
+      "learning_rate": 3.1133293278607228e-06,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.882025346159935,
+      "num_tokens": 211434652.0,
+      "step": 1977
+    },
+    {
+      "epoch": 4.506271379703534,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.1115032648667224e-06,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.8816207945346832,
+      "num_tokens": 211542180.0,
+      "step": 1978
+    },
+    {
+      "epoch": 4.508551881413911,
+      "grad_norm": 3.25,
+      "learning_rate": 3.1096768547952743e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8829948008060455,
+      "num_tokens": 211649219.0,
+      "step": 1979
+    },
+    {
+      "epoch": 4.510832383124288,
+      "grad_norm": 2.5,
+      "learning_rate": 3.1078500986830134e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8840111643075943,
+      "num_tokens": 211756443.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.510832383124288,
+      "eval_loss": 0.5894109606742859,
+      "eval_mean_token_accuracy": 0.8794237710677172,
+      "eval_num_tokens": 211756443.0,
+      "eval_runtime": 58.6388,
+      "eval_samples_per_second": 142.994,
+      "eval_steps_per_second": 4.485,
+      "step": 1980
+    },
+    {
+      "epoch": 4.513112884834664,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.1060229975667716e-06,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8810431063175201,
+      "num_tokens": 211863529.0,
+      "step": 1981
+    },
+    {
+      "epoch": 4.51539338654504,
+      "grad_norm": 5.625,
+      "learning_rate": 3.104195552483576e-06,
+      "loss": 0.5842,
+      "mean_token_accuracy": 0.8787284940481186,
+      "num_tokens": 211970403.0,
+      "step": 1982
+    },
+    {
+      "epoch": 4.517673888255416,
+      "grad_norm": 5.21875,
+      "learning_rate": 3.102367764470649e-06,
+      "loss": 0.568,
+      "mean_token_accuracy": 0.8812828063964844,
+      "num_tokens": 212077504.0,
+      "step": 1983
+    },
+    {
+      "epoch": 4.519954389965792,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.1005396345654087e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.8794949948787689,
+      "num_tokens": 212185300.0,
+      "step": 1984
+    },
+    {
+      "epoch": 4.5222348916761685,
+      "grad_norm": 4.25,
+      "learning_rate": 3.0987111638054657e-06,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8824008107185364,
+      "num_tokens": 212292097.0,
+      "step": 1985
+    },
+    {
+      "epoch": 4.524515393386545,
+      "grad_norm": 5.65625,
+      "learning_rate": 3.0968823532286246e-06,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8847775161266327,
+      "num_tokens": 212399289.0,
+      "step": 1986
+    },
+    {
+      "epoch": 4.526795895096921,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.095053203872883e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8831916600465775,
+      "num_tokens": 212506647.0,
+      "step": 1987
+    },
+    {
+      "epoch": 4.529076396807298,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.0932237167764306e-06,
+      "loss": 0.5765,
+      "mean_token_accuracy": 0.883448526263237,
+      "num_tokens": 212613620.0,
+      "step": 1988
+    },
+    {
+      "epoch": 4.531356898517674,
+      "grad_norm": 4.46875,
+      "learning_rate": 3.0913938929776493e-06,
+      "loss": 0.5899,
+      "mean_token_accuracy": 0.8791800290346146,
+      "num_tokens": 212720388.0,
+      "step": 1989
+    },
+    {
+      "epoch": 4.53363740022805,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.0895637335151117e-06,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.8825742602348328,
+      "num_tokens": 212827508.0,
+      "step": 1990
+    },
+    {
+      "epoch": 4.535917901938427,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.0877332394275806e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.878256767988205,
+      "num_tokens": 212933843.0,
+      "step": 1991
+    },
+    {
+      "epoch": 4.538198403648803,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.08590241175401e-06,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8803077787160873,
+      "num_tokens": 213040798.0,
+      "step": 1992
+    },
+    {
+      "epoch": 4.540478905359179,
+      "grad_norm": 5.1875,
+      "learning_rate": 3.0840712515335412e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8778508752584457,
+      "num_tokens": 213147454.0,
+      "step": 1993
+    },
+    {
+      "epoch": 4.542759407069555,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.0822397598055065e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8773138970136642,
+      "num_tokens": 213255022.0,
+      "step": 1994
+    },
+    {
+      "epoch": 4.545039908779931,
+      "grad_norm": 4.375,
+      "learning_rate": 3.080407937609424e-06,
+      "loss": 0.5926,
+      "mean_token_accuracy": 0.8787964135408401,
+      "num_tokens": 213362436.0,
+      "step": 1995
+    },
+    {
+      "epoch": 4.5473204104903076,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.0785757859850025e-06,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.8769951462745667,
+      "num_tokens": 213469549.0,
+      "step": 1996
+    },
+    {
+      "epoch": 4.549600912200685,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.0767433059721338e-06,
+      "loss": 0.5777,
+      "mean_token_accuracy": 0.8808469474315643,
+      "num_tokens": 213576382.0,
+      "step": 1997
+    },
+    {
+      "epoch": 4.55188141391106,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.074910498610899e-06,
+      "loss": 0.5572,
+      "mean_token_accuracy": 0.887230396270752,
+      "num_tokens": 213683677.0,
+      "step": 1998
+    },
+    {
+      "epoch": 4.554161915621437,
+      "grad_norm": 5.03125,
+      "learning_rate": 3.0730773649415647e-06,
+      "loss": 0.5931,
+      "mean_token_accuracy": 0.8774087131023407,
+      "num_tokens": 213790372.0,
+      "step": 1999
+    },
+    {
+      "epoch": 4.556442417331813,
+      "grad_norm": 6.09375,
+      "learning_rate": 3.0712439060045818e-06,
+      "loss": 0.5667,
+      "mean_token_accuracy": 0.884253740310669,
+      "num_tokens": 213897106.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.558722919042189,
+      "grad_norm": 5.25,
+      "learning_rate": 3.069410122840585e-06,
+      "loss": 0.5908,
+      "mean_token_accuracy": 0.8797920346260071,
+      "num_tokens": 214004255.0,
+      "step": 2001
+    },
+    {
+      "epoch": 4.561003420752566,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.0675760164903972e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8811387270689011,
+      "num_tokens": 214111233.0,
+      "step": 2002
+    },
+    {
+      "epoch": 4.563283922462942,
+      "grad_norm": 3.125,
+      "learning_rate": 3.065741587995019e-06,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.8835022449493408,
+      "num_tokens": 214218120.0,
+      "step": 2003
+    },
+    {
+      "epoch": 4.565564424173318,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.0639068383956373e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.880307137966156,
+      "num_tokens": 214324888.0,
+      "step": 2004
+    },
+    {
+      "epoch": 4.567844925883694,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.062071768733621e-06,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.886817216873169,
+      "num_tokens": 214432054.0,
+      "step": 2005
+    },
+    {
+      "epoch": 4.57012542759407,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.0602363800505198e-06,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.882687121629715,
+      "num_tokens": 214539517.0,
+      "step": 2006
+    },
+    {
+      "epoch": 4.572405929304447,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.0584006733880656e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.8774117529392242,
+      "num_tokens": 214646420.0,
+      "step": 2007
+    },
+    {
+      "epoch": 4.574686431014824,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.0565646497881697e-06,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8867736458778381,
+      "num_tokens": 214753267.0,
+      "step": 2008
+    },
+    {
+      "epoch": 4.5769669327252,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.0547283102929228e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8841065913438797,
+      "num_tokens": 214860275.0,
+      "step": 2009
+    },
+    {
+      "epoch": 4.579247434435576,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.0528916559445967e-06,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8801997601985931,
+      "num_tokens": 214967499.0,
+      "step": 2010
+    },
+    {
+      "epoch": 4.581527936145952,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.05105468778564e-06,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8840168863534927,
+      "num_tokens": 215075121.0,
+      "step": 2011
+    },
+    {
+      "epoch": 4.583808437856328,
+      "grad_norm": 5.3125,
+      "learning_rate": 3.049217406858681e-06,
+      "loss": 0.6082,
+      "mean_token_accuracy": 0.8759740740060806,
+      "num_tokens": 215181867.0,
+      "step": 2012
+    },
+    {
+      "epoch": 4.586088939566705,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.047379814206526e-06,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8838594704866409,
+      "num_tokens": 215288922.0,
+      "step": 2013
+    },
+    {
+      "epoch": 4.588369441277081,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.0455419108721556e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8816310912370682,
+      "num_tokens": 215395792.0,
+      "step": 2014
+    },
+    {
+      "epoch": 4.590649942987457,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.043703697898728e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8817844688892365,
+      "num_tokens": 215503184.0,
+      "step": 2015
+    },
+    {
+      "epoch": 4.592930444697833,
+      "grad_norm": 2.234375,
+      "learning_rate": 3.041865176329579e-06,
+      "loss": 0.5444,
+      "mean_token_accuracy": 0.8872413337230682,
+      "num_tokens": 215610549.0,
+      "step": 2016
+    },
+    {
+      "epoch": 4.59521094640821,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.040026347208217e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.882647380232811,
+      "num_tokens": 215718112.0,
+      "step": 2017
+    },
+    {
+      "epoch": 4.5974914481185865,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.0381872115783256e-06,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8802604079246521,
+      "num_tokens": 215825609.0,
+      "step": 2018
+    },
+    {
+      "epoch": 4.599771949828963,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.0363477704837633e-06,
+      "loss": 0.552,
+      "mean_token_accuracy": 0.8875061571598053,
+      "num_tokens": 215932611.0,
+      "step": 2019
+    },
+    {
+      "epoch": 4.602052451539339,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.034508024968561e-06,
+      "loss": 0.5714,
+      "mean_token_accuracy": 0.8848912864923477,
+      "num_tokens": 216040088.0,
+      "step": 2020
+    },
+    {
+      "epoch": 4.604332953249715,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.032667976076923e-06,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.8811133801937103,
+      "num_tokens": 216147228.0,
+      "step": 2021
+    },
+    {
+      "epoch": 4.606613454960091,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.0308276248532244e-06,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.8812815397977829,
+      "num_tokens": 216254056.0,
+      "step": 2022
+    },
+    {
+      "epoch": 4.608893956670467,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.0289869723420144e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8829955011606216,
+      "num_tokens": 216361270.0,
+      "step": 2023
+    },
+    {
+      "epoch": 4.611174458380844,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.027146019588012e-06,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.8840386420488358,
+      "num_tokens": 216468575.0,
+      "step": 2024
+    },
+    {
+      "epoch": 4.61345496009122,
+      "grad_norm": 5.53125,
+      "learning_rate": 3.025304767636105e-06,
+      "loss": 0.5558,
+      "mean_token_accuracy": 0.8865272402763367,
+      "num_tokens": 216576234.0,
+      "step": 2025
+    },
+    {
+      "epoch": 4.615735461801596,
+      "grad_norm": 5.40625,
+      "learning_rate": 3.0234632175313537e-06,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8783838450908661,
+      "num_tokens": 216683054.0,
+      "step": 2026
+    },
+    {
+      "epoch": 4.618015963511972,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.0216213703189856e-06,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8805306851863861,
+      "num_tokens": 216790089.0,
+      "step": 2027
+    },
+    {
+      "epoch": 4.620296465222349,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.019779227044398e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8831235766410828,
+      "num_tokens": 216897778.0,
+      "step": 2028
+    },
+    {
+      "epoch": 4.6225769669327255,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.0179367887531567e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8810001909732819,
+      "num_tokens": 217004442.0,
+      "step": 2029
+    },
+    {
+      "epoch": 4.624857468643102,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.016094056490993e-06,
+      "loss": 0.5503,
+      "mean_token_accuracy": 0.8854062855243683,
+      "num_tokens": 217111586.0,
+      "step": 2030
+    },
+    {
+      "epoch": 4.627137970353478,
+      "grad_norm": 2.40625,
+      "learning_rate": 3.0142510313038057e-06,
+      "loss": 0.5533,
+      "mean_token_accuracy": 0.8880758136510849,
+      "num_tokens": 217219168.0,
+      "step": 2031
+    },
+    {
+      "epoch": 4.629418472063854,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.012407714237662e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8846538215875626,
+      "num_tokens": 217326211.0,
+      "step": 2032
+    },
+    {
+      "epoch": 4.63169897377423,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.010564106338791e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8847674578428268,
+      "num_tokens": 217433707.0,
+      "step": 2033
+    },
+    {
+      "epoch": 4.633979475484606,
+      "grad_norm": 2.375,
+      "learning_rate": 3.0087202086535915e-06,
+      "loss": 0.5532,
+      "mean_token_accuracy": 0.8850553631782532,
+      "num_tokens": 217541577.0,
+      "step": 2034
+    },
+    {
+      "epoch": 4.636259977194983,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.006876022228622e-06,
+      "loss": 0.5832,
+      "mean_token_accuracy": 0.882301077246666,
+      "num_tokens": 217648764.0,
+      "step": 2035
+    },
+    {
+      "epoch": 4.638540478905359,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.0050315481106074e-06,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8841645568609238,
+      "num_tokens": 217755467.0,
+      "step": 2036
+    },
+    {
+      "epoch": 4.640820980615736,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.0031867873464372e-06,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8814872205257416,
+      "num_tokens": 217862349.0,
+      "step": 2037
+    },
+    {
+      "epoch": 4.643101482326112,
+      "grad_norm": 2.875,
+      "learning_rate": 3.00134174098316e-06,
+      "loss": 0.6096,
+      "mean_token_accuracy": 0.8733366578817368,
+      "num_tokens": 217968209.0,
+      "step": 2038
+    },
+    {
+      "epoch": 4.645381984036488,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.999496410067989e-06,
+      "loss": 0.5549,
+      "mean_token_accuracy": 0.8875339925289154,
+      "num_tokens": 218076237.0,
+      "step": 2039
+    },
+    {
+      "epoch": 4.6476624857468645,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.9976507956482996e-06,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8795648515224457,
+      "num_tokens": 218183464.0,
+      "step": 2040
+    },
+    {
+      "epoch": 4.649942987457241,
+      "grad_norm": 2.46875,
+      "learning_rate": 2.9958048987716266e-06,
+      "loss": 0.6071,
+      "mean_token_accuracy": 0.8746256977319717,
+      "num_tokens": 218290055.0,
+      "step": 2041
+    },
+    {
+      "epoch": 4.652223489167617,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.993958720485664e-06,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.8757225871086121,
+      "num_tokens": 218397051.0,
+      "step": 2042
+    },
+    {
+      "epoch": 4.654503990877993,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.9921122618382687e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8802944868803024,
+      "num_tokens": 218504413.0,
+      "step": 2043
+    },
+    {
+      "epoch": 4.656784492588369,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.9902655238774537e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8826518207788467,
+      "num_tokens": 218611326.0,
+      "step": 2044
+    },
+    {
+      "epoch": 4.659064994298745,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.988418507651392e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8791725039482117,
+      "num_tokens": 218718666.0,
+      "step": 2045
+    },
+    {
+      "epoch": 4.661345496009122,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.9865712142084145e-06,
+      "loss": 0.5578,
+      "mean_token_accuracy": 0.8884040713310242,
+      "num_tokens": 218826339.0,
+      "step": 2046
+    },
+    {
+      "epoch": 4.663625997719498,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.98472364459701e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8853495866060257,
+      "num_tokens": 218933583.0,
+      "step": 2047
+    },
+    {
+      "epoch": 4.665906499429875,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.982875799865823e-06,
+      "loss": 0.5685,
+      "mean_token_accuracy": 0.8821221739053726,
+      "num_tokens": 219040654.0,
+      "step": 2048
+    },
+    {
+      "epoch": 4.668187001140251,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.9810276810636535e-06,
+      "loss": 0.6013,
+      "mean_token_accuracy": 0.8774379193782806,
+      "num_tokens": 219147467.0,
+      "step": 2049
+    },
+    {
+      "epoch": 4.670467502850627,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.97917928923946e-06,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.879958301782608,
+      "num_tokens": 219254877.0,
+      "step": 2050
+    },
+    {
+      "epoch": 4.6727480045610035,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.977330625442352e-06,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.8850719779729843,
+      "num_tokens": 219362293.0,
+      "step": 2051
+    },
+    {
+      "epoch": 4.67502850627138,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.9754816907215963e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8815977722406387,
+      "num_tokens": 219469187.0,
+      "step": 2052
+    },
+    {
+      "epoch": 4.677309007981756,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.9736324861266125e-06,
+      "loss": 0.5835,
+      "mean_token_accuracy": 0.8775934129953384,
+      "num_tokens": 219576287.0,
+      "step": 2053
+    },
+    {
+      "epoch": 4.679589509692132,
+      "grad_norm": 5.3125,
+      "learning_rate": 2.9717830127069734e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8794510513544083,
+      "num_tokens": 219683068.0,
+      "step": 2054
+    },
+    {
+      "epoch": 4.681870011402508,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.969933271512404e-06,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8821385949850082,
+      "num_tokens": 219790016.0,
+      "step": 2055
+    },
+    {
+      "epoch": 4.684150513112884,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.9680832635927824e-06,
+      "loss": 0.5908,
+      "mean_token_accuracy": 0.8790445029735565,
+      "num_tokens": 219896728.0,
+      "step": 2056
+    },
+    {
+      "epoch": 4.6864310148232615,
+      "grad_norm": 2.75,
+      "learning_rate": 2.9662329899981375e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8832840174436569,
+      "num_tokens": 220004033.0,
+      "step": 2057
+    },
+    {
+      "epoch": 4.688711516533638,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.964382451778648e-06,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8850243538618088,
+      "num_tokens": 220111159.0,
+      "step": 2058
+    },
+    {
+      "epoch": 4.690992018244014,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.9625316499846444e-06,
+      "loss": 0.582,
+      "mean_token_accuracy": 0.8815952986478806,
+      "num_tokens": 220218001.0,
+      "step": 2059
+    },
+    {
+      "epoch": 4.69327251995439,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.9606805856666053e-06,
+      "loss": 0.6012,
+      "mean_token_accuracy": 0.8748088330030441,
+      "num_tokens": 220325346.0,
+      "step": 2060
+    },
+    {
+      "epoch": 4.695553021664766,
+      "grad_norm": 2.34375,
+      "learning_rate": 2.95882925987516e-06,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8794578313827515,
+      "num_tokens": 220432554.0,
+      "step": 2061
+    },
+    {
+      "epoch": 4.6978335233751425,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.9569776736610855e-06,
+      "loss": 0.5848,
+      "mean_token_accuracy": 0.8808020949363708,
+      "num_tokens": 220539630.0,
+      "step": 2062
+    },
+    {
+      "epoch": 4.700114025085519,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.9551258280753046e-06,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8800801783800125,
+      "num_tokens": 220646814.0,
+      "step": 2063
+    },
+    {
+      "epoch": 4.702394526795895,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.953273724168891e-06,
+      "loss": 0.568,
+      "mean_token_accuracy": 0.8822778612375259,
+      "num_tokens": 220753926.0,
+      "step": 2064
+    },
+    {
+      "epoch": 4.704675028506271,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.9514213629930614e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8794926106929779,
+      "num_tokens": 220860498.0,
+      "step": 2065
+    },
+    {
+      "epoch": 4.706955530216648,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.949568745599182e-06,
+      "loss": 0.5579,
+      "mean_token_accuracy": 0.8850435167551041,
+      "num_tokens": 220967240.0,
+      "step": 2066
+    },
+    {
+      "epoch": 4.7092360319270234,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.9477158730387615e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8826223760843277,
+      "num_tokens": 221074023.0,
+      "step": 2067
+    },
+    {
+      "epoch": 4.7115165336374005,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.945862746363455e-06,
+      "loss": 0.5591,
+      "mean_token_accuracy": 0.8868841081857681,
+      "num_tokens": 221181425.0,
+      "step": 2068
+    },
+    {
+      "epoch": 4.713797035347777,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.944009366625061e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8801351636648178,
+      "num_tokens": 221288251.0,
+      "step": 2069
+    },
+    {
+      "epoch": 4.716077537058153,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.942155734875523e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8805908411741257,
+      "num_tokens": 221395168.0,
+      "step": 2070
+    },
+    {
+      "epoch": 4.718358038768529,
+      "grad_norm": 6.375,
+      "learning_rate": 2.9403018521669256e-06,
+      "loss": 0.5966,
+      "mean_token_accuracy": 0.8779006004333496,
+      "num_tokens": 221502170.0,
+      "step": 2071
+    },
+    {
+      "epoch": 4.720638540478905,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.938447719551498e-06,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8802389800548553,
+      "num_tokens": 221608850.0,
+      "step": 2072
+    },
+    {
+      "epoch": 4.7229190421892815,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.9365933380816092e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8805797696113586,
+      "num_tokens": 221715991.0,
+      "step": 2073
+    },
+    {
+      "epoch": 4.725199543899658,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.93473870880977e-06,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8762219101190567,
+      "num_tokens": 221823317.0,
+      "step": 2074
+    },
+    {
+      "epoch": 4.727480045610034,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.932883832788633e-06,
+      "loss": 0.5591,
+      "mean_token_accuracy": 0.8835586160421371,
+      "num_tokens": 221930410.0,
+      "step": 2075
+    },
+    {
+      "epoch": 4.72976054732041,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.9310287110709895e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.881802573800087,
+      "num_tokens": 222037632.0,
+      "step": 2076
+    },
+    {
+      "epoch": 4.732041049030787,
+      "grad_norm": 3.828125,
+      "learning_rate": 2.9291733447097714e-06,
+      "loss": 0.5588,
+      "mean_token_accuracy": 0.8866459727287292,
+      "num_tokens": 222144620.0,
+      "step": 2077
+    },
+    {
+      "epoch": 4.734321550741163,
+      "grad_norm": 3.625,
+      "learning_rate": 2.927317734758047e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.879328653216362,
+      "num_tokens": 222251771.0,
+      "step": 2078
+    },
+    {
+      "epoch": 4.7366020524515395,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.925461882269027e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8794370591640472,
+      "num_tokens": 222359272.0,
+      "step": 2079
+    },
+    {
+      "epoch": 4.738882554161916,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.9236057882960567e-06,
+      "loss": 0.5677,
+      "mean_token_accuracy": 0.8789798021316528,
+      "num_tokens": 222466677.0,
+      "step": 2080
+    },
+    {
+      "epoch": 4.741163055872292,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.921749453892618e-06,
+      "loss": 0.579,
+      "mean_token_accuracy": 0.8830148726701736,
+      "num_tokens": 222573940.0,
+      "step": 2081
+    },
+    {
+      "epoch": 4.743443557582668,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.919892880112332e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8783297091722488,
+      "num_tokens": 222680851.0,
+      "step": 2082
+    },
+    {
+      "epoch": 4.745724059293044,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.9180360680089542e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8823717683553696,
+      "num_tokens": 222787611.0,
+      "step": 2083
+    },
+    {
+      "epoch": 4.7480045610034205,
+      "grad_norm": 2.75,
+      "learning_rate": 2.9161790186363746e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8819119483232498,
+      "num_tokens": 222894602.0,
+      "step": 2084
+    },
+    {
+      "epoch": 4.750285062713797,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.9143217330486186e-06,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.8828014135360718,
+      "num_tokens": 223001331.0,
+      "step": 2085
+    },
+    {
+      "epoch": 4.752565564424174,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.9124642122998453e-06,
+      "loss": 0.5765,
+      "mean_token_accuracy": 0.8834296762943268,
+      "num_tokens": 223108795.0,
+      "step": 2086
+    },
+    {
+      "epoch": 4.75484606613455,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.9106064574443477e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8811309486627579,
+      "num_tokens": 223215668.0,
+      "step": 2087
+    },
+    {
+      "epoch": 4.757126567844926,
+      "grad_norm": 6.09375,
+      "learning_rate": 2.9087484695365523e-06,
+      "loss": 0.5878,
+      "mean_token_accuracy": 0.8807103782892227,
+      "num_tokens": 223322617.0,
+      "step": 2088
+    },
+    {
+      "epoch": 4.759407069555302,
+      "grad_norm": 5.6875,
+      "learning_rate": 2.906890249631017e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8810094445943832,
+      "num_tokens": 223429512.0,
+      "step": 2089
+    },
+    {
+      "epoch": 4.7616875712656785,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.905031798782431e-06,
+      "loss": 0.5522,
+      "mean_token_accuracy": 0.8895687907934189,
+      "num_tokens": 223536691.0,
+      "step": 2090
+    },
+    {
+      "epoch": 4.763968072976055,
+      "grad_norm": 2.75,
+      "learning_rate": 2.903173118045616e-06,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8834663927555084,
+      "num_tokens": 223643826.0,
+      "step": 2091
+    },
+    {
+      "epoch": 4.766248574686431,
+      "grad_norm": 4.40625,
+      "learning_rate": 2.901314208475522e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8782705664634705,
+      "num_tokens": 223750903.0,
+      "step": 2092
+    },
+    {
+      "epoch": 4.768529076396807,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.8994550711272317e-06,
+      "loss": 0.5789,
+      "mean_token_accuracy": 0.8804999589920044,
+      "num_tokens": 223858002.0,
+      "step": 2093
+    },
+    {
+      "epoch": 4.770809578107183,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.897595707055954e-06,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8842459321022034,
+      "num_tokens": 223965056.0,
+      "step": 2094
+    },
+    {
+      "epoch": 4.7730900798175595,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.8957361173170297e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8789258599281311,
+      "num_tokens": 224071784.0,
+      "step": 2095
+    },
+    {
+      "epoch": 4.775370581527936,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.893876302965925e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8828511238098145,
+      "num_tokens": 224178796.0,
+      "step": 2096
+    },
+    {
+      "epoch": 4.777651083238313,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.8920162650582344e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8815236538648605,
+      "num_tokens": 224287157.0,
+      "step": 2097
+    },
+    {
+      "epoch": 4.779931584948689,
+      "grad_norm": 4.625,
+      "learning_rate": 2.8901560046496797e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8796006739139557,
+      "num_tokens": 224394174.0,
+      "step": 2098
+    },
+    {
+      "epoch": 4.782212086659065,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.8882955227961098e-06,
+      "loss": 0.5891,
+      "mean_token_accuracy": 0.8783185184001923,
+      "num_tokens": 224501173.0,
+      "step": 2099
+    },
+    {
+      "epoch": 4.784492588369441,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.886434820553497e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8836559951305389,
+      "num_tokens": 224608530.0,
+      "step": 2100
+    },
+    {
+      "epoch": 4.7867730900798175,
+      "grad_norm": 3.25,
+      "learning_rate": 2.884573898977941e-06,
+      "loss": 0.5632,
+      "mean_token_accuracy": 0.8841763436794281,
+      "num_tokens": 224715326.0,
+      "step": 2101
+    },
+    {
+      "epoch": 4.789053591790194,
+      "grad_norm": 2.5,
+      "learning_rate": 2.882712759125664e-06,
+      "loss": 0.5582,
+      "mean_token_accuracy": 0.8851722776889801,
+      "num_tokens": 224822759.0,
+      "step": 2102
+    },
+    {
+      "epoch": 4.79133409350057,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.8808514020530127e-06,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8818291127681732,
+      "num_tokens": 224929583.0,
+      "step": 2103
+    },
+    {
+      "epoch": 4.793614595210946,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.8789898288164595e-06,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8850146681070328,
+      "num_tokens": 225036349.0,
+      "step": 2104
+    },
+    {
+      "epoch": 4.795895096921322,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.8771280404725953e-06,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8813169300556183,
+      "num_tokens": 225142845.0,
+      "step": 2105
+    },
+    {
+      "epoch": 4.798175598631699,
+      "grad_norm": 5.875,
+      "learning_rate": 2.8752660380781367e-06,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8808184266090393,
+      "num_tokens": 225250066.0,
+      "step": 2106
+    },
+    {
+      "epoch": 4.800456100342076,
+      "grad_norm": 4.0,
+      "learning_rate": 2.8734038226899198e-06,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8856227844953537,
+      "num_tokens": 225356570.0,
+      "step": 2107
+    },
+    {
+      "epoch": 4.802736602052452,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.8715413953649012e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8814336508512497,
+      "num_tokens": 225463833.0,
+      "step": 2108
+    },
+    {
+      "epoch": 4.805017103762828,
+      "grad_norm": 4.125,
+      "learning_rate": 2.8696787571601597e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8789616823196411,
+      "num_tokens": 225570947.0,
+      "step": 2109
+    },
+    {
+      "epoch": 4.807297605473204,
+      "grad_norm": 5.4375,
+      "learning_rate": 2.8678159091328926e-06,
+      "loss": 0.5651,
+      "mean_token_accuracy": 0.8813024163246155,
+      "num_tokens": 225678130.0,
+      "step": 2110
+    },
+    {
+      "epoch": 4.80957810718358,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.865952852340417e-06,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8798904716968536,
+      "num_tokens": 225785246.0,
+      "step": 2111
+    },
+    {
+      "epoch": 4.811858608893957,
+      "grad_norm": 6.53125,
+      "learning_rate": 2.864089587840167e-06,
+      "loss": 0.5956,
+      "mean_token_accuracy": 0.8791421502828598,
+      "num_tokens": 225891598.0,
+      "step": 2112
+    },
+    {
+      "epoch": 4.814139110604333,
+      "grad_norm": 3.25,
+      "learning_rate": 2.862226116689696e-06,
+      "loss": 0.5793,
+      "mean_token_accuracy": 0.8808872699737549,
+      "num_tokens": 225998427.0,
+      "step": 2113
+    },
+    {
+      "epoch": 4.816419612314709,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.8603624399466732e-06,
+      "loss": 0.561,
+      "mean_token_accuracy": 0.8853624612092972,
+      "num_tokens": 226106239.0,
+      "step": 2114
+    },
+    {
+      "epoch": 4.818700114025085,
+      "grad_norm": 2.46875,
+      "learning_rate": 2.858498558668888e-06,
+      "loss": 0.549,
+      "mean_token_accuracy": 0.886398509144783,
+      "num_tokens": 226213512.0,
+      "step": 2115
+    },
+    {
+      "epoch": 4.820980615735461,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.856634473914242e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8814459592103958,
+      "num_tokens": 226321236.0,
+      "step": 2116
+    },
+    {
+      "epoch": 4.823261117445838,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.854770186740753e-06,
+      "loss": 0.5703,
+      "mean_token_accuracy": 0.8803770244121552,
+      "num_tokens": 226428875.0,
+      "step": 2117
+    },
+    {
+      "epoch": 4.825541619156215,
+      "grad_norm": 6.09375,
+      "learning_rate": 2.8529056982065557e-06,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8832362145185471,
+      "num_tokens": 226535810.0,
+      "step": 2118
+    },
+    {
+      "epoch": 4.827822120866591,
+      "grad_norm": 5.875,
+      "learning_rate": 2.8510410093698966e-06,
+      "loss": 0.5848,
+      "mean_token_accuracy": 0.8776679933071136,
+      "num_tokens": 226642320.0,
+      "step": 2119
+    },
+    {
+      "epoch": 4.830102622576967,
+      "grad_norm": 6.03125,
+      "learning_rate": 2.849176121289138e-06,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8798095434904099,
+      "num_tokens": 226750138.0,
+      "step": 2120
+    },
+    {
+      "epoch": 4.832383124287343,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.8473110350227536e-06,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8817552924156189,
+      "num_tokens": 226857421.0,
+      "step": 2121
+    },
+    {
+      "epoch": 4.834663625997719,
+      "grad_norm": 4.125,
+      "learning_rate": 2.845445751629331e-06,
+      "loss": 0.5896,
+      "mean_token_accuracy": 0.8786935806274414,
+      "num_tokens": 226965168.0,
+      "step": 2122
+    },
+    {
+      "epoch": 4.836944127708096,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.843580272167569e-06,
+      "loss": 0.5703,
+      "mean_token_accuracy": 0.8808483481407166,
+      "num_tokens": 227072404.0,
+      "step": 2123
+    },
+    {
+      "epoch": 4.839224629418472,
+      "grad_norm": 6.625,
+      "learning_rate": 2.8417145976962773e-06,
+      "loss": 0.5926,
+      "mean_token_accuracy": 0.8761008530855179,
+      "num_tokens": 227179076.0,
+      "step": 2124
+    },
+    {
+      "epoch": 4.841505131128848,
+      "grad_norm": 4.75,
+      "learning_rate": 2.8398487292743772e-06,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8841915279626846,
+      "num_tokens": 227285768.0,
+      "step": 2125
+    },
+    {
+      "epoch": 4.843785632839225,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.8379826679609e-06,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8802362382411957,
+      "num_tokens": 227392431.0,
+      "step": 2126
+    },
+    {
+      "epoch": 4.846066134549601,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.836116414814985e-06,
+      "loss": 0.6062,
+      "mean_token_accuracy": 0.8763725161552429,
+      "num_tokens": 227499046.0,
+      "step": 2127
+    },
+    {
+      "epoch": 4.848346636259977,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.8342499708958827e-06,
+      "loss": 0.5858,
+      "mean_token_accuracy": 0.8813872933387756,
+      "num_tokens": 227606044.0,
+      "step": 2128
+    },
+    {
+      "epoch": 4.850627137970354,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.8323833372629485e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8817006945610046,
+      "num_tokens": 227713533.0,
+      "step": 2129
+    },
+    {
+      "epoch": 4.85290763968073,
+      "grad_norm": 5.625,
+      "learning_rate": 2.8305165149756496e-06,
+      "loss": 0.5596,
+      "mean_token_accuracy": 0.8857055157423019,
+      "num_tokens": 227820658.0,
+      "step": 2130
+    },
+    {
+      "epoch": 4.855188141391106,
+      "grad_norm": 3.25,
+      "learning_rate": 2.828649505093558e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8798857480287552,
+      "num_tokens": 227927701.0,
+      "step": 2131
+    },
+    {
+      "epoch": 4.857468643101482,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.826782308676351e-06,
+      "loss": 0.5818,
+      "mean_token_accuracy": 0.8805892914533615,
+      "num_tokens": 228034374.0,
+      "step": 2132
+    },
+    {
+      "epoch": 4.859749144811858,
+      "grad_norm": 2.75,
+      "learning_rate": 2.824914926783815e-06,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8796881288290024,
+      "num_tokens": 228141770.0,
+      "step": 2133
+    },
+    {
+      "epoch": 4.862029646522235,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.82304736047584e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.8820720165967941,
+      "num_tokens": 228249025.0,
+      "step": 2134
+    },
+    {
+      "epoch": 4.864310148232612,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.821179610812419e-06,
+      "loss": 0.5874,
+      "mean_token_accuracy": 0.8788727074861526,
+      "num_tokens": 228356604.0,
+      "step": 2135
+    },
+    {
+      "epoch": 4.866590649942988,
+      "grad_norm": 6.71875,
+      "learning_rate": 2.819311678853652e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8829772472381592,
+      "num_tokens": 228464360.0,
+      "step": 2136
+    },
+    {
+      "epoch": 4.868871151653364,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.8174435656597403e-06,
+      "loss": 0.557,
+      "mean_token_accuracy": 0.8821916729211807,
+      "num_tokens": 228571499.0,
+      "step": 2137
+    },
+    {
+      "epoch": 4.87115165336374,
+      "grad_norm": 4.78125,
+      "learning_rate": 2.8155752722909896e-06,
+      "loss": 0.5777,
+      "mean_token_accuracy": 0.8816975653171539,
+      "num_tokens": 228678610.0,
+      "step": 2138
+    },
+    {
+      "epoch": 4.873432155074116,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.8137067998078073e-06,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8804251700639725,
+      "num_tokens": 228785226.0,
+      "step": 2139
+    },
+    {
+      "epoch": 4.875712656784493,
+      "grad_norm": 2.421875,
+      "learning_rate": 2.8118381492707004e-06,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8774864822626114,
+      "num_tokens": 228891949.0,
+      "step": 2140
+    },
+    {
+      "epoch": 4.877993158494869,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.8099693217402807e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8809685558080673,
+      "num_tokens": 228999525.0,
+      "step": 2141
+    },
+    {
+      "epoch": 4.880273660205245,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.808100318277258e-06,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8823101967573166,
+      "num_tokens": 229106700.0,
+      "step": 2142
+    },
+    {
+      "epoch": 4.882554161915621,
+      "grad_norm": 2.359375,
+      "learning_rate": 2.806231139942443e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8847738802433014,
+      "num_tokens": 229214479.0,
+      "step": 2143
+    },
+    {
+      "epoch": 4.884834663625997,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.8043617877967456e-06,
+      "loss": 0.5809,
+      "mean_token_accuracy": 0.8798518478870392,
+      "num_tokens": 229321261.0,
+      "step": 2144
+    },
+    {
+      "epoch": 4.887115165336374,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.8024922629011727e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.883740559220314,
+      "num_tokens": 229428653.0,
+      "step": 2145
+    },
+    {
+      "epoch": 4.889395667046751,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.800622566316831e-06,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8811280727386475,
+      "num_tokens": 229535592.0,
+      "step": 2146
+    },
+    {
+      "epoch": 4.891676168757127,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.798752699104925e-06,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8848204910755157,
+      "num_tokens": 229642187.0,
+      "step": 2147
+    },
+    {
+      "epoch": 4.893956670467503,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.7968826623267542e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8824999332427979,
+      "num_tokens": 229749302.0,
+      "step": 2148
+    },
+    {
+      "epoch": 4.896237172177879,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.7950124570437163e-06,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8834424167871475,
+      "num_tokens": 229856402.0,
+      "step": 2149
+    },
+    {
+      "epoch": 4.898517673888255,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.793142084317303e-06,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8832953125238419,
+      "num_tokens": 229963106.0,
+      "step": 2150
+    },
+    {
+      "epoch": 4.900798175598632,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.7912715452091014e-06,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8826733380556107,
+      "num_tokens": 230070472.0,
+      "step": 2151
+    },
+    {
+      "epoch": 4.903078677309008,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.789400840780795e-06,
+      "loss": 0.5928,
+      "mean_token_accuracy": 0.8755721002817154,
+      "num_tokens": 230177193.0,
+      "step": 2152
+    },
+    {
+      "epoch": 4.905359179019384,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.7875299720941577e-06,
+      "loss": 0.5809,
+      "mean_token_accuracy": 0.8802897483110428,
+      "num_tokens": 230284630.0,
+      "step": 2153
+    },
+    {
+      "epoch": 4.90763968072976,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.785658940211059e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8808315396308899,
+      "num_tokens": 230391476.0,
+      "step": 2154
+    },
+    {
+      "epoch": 4.909920182440137,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.7837877461934616e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.881686270236969,
+      "num_tokens": 230498008.0,
+      "step": 2155
+    },
+    {
+      "epoch": 4.9122006841505135,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.7819163911034175e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.882976695895195,
+      "num_tokens": 230605141.0,
+      "step": 2156
+    },
+    {
+      "epoch": 4.91448118586089,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.7800448760030724e-06,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8807835131883621,
+      "num_tokens": 230712400.0,
+      "step": 2157
+    },
+    {
+      "epoch": 4.916761687571266,
+      "grad_norm": 7.28125,
+      "learning_rate": 2.7781732019546625e-06,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8794967532157898,
+      "num_tokens": 230819368.0,
+      "step": 2158
+    },
+    {
+      "epoch": 4.919042189281642,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.776301370020513e-06,
+      "loss": 0.5858,
+      "mean_token_accuracy": 0.8819598108530045,
+      "num_tokens": 230925639.0,
+      "step": 2159
+    },
+    {
+      "epoch": 4.921322690992018,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.7744293812630412e-06,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.881621241569519,
+      "num_tokens": 231032712.0,
+      "step": 2160
+    },
+    {
+      "epoch": 4.923603192702394,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.77255723674475e-06,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8840135484933853,
+      "num_tokens": 231139944.0,
+      "step": 2161
+    },
+    {
+      "epoch": 4.925883694412771,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.770684937528233e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8798476308584213,
+      "num_tokens": 231246948.0,
+      "step": 2162
+    },
+    {
+      "epoch": 4.928164196123147,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.7688124846761716e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8791171163320541,
+      "num_tokens": 231353810.0,
+      "step": 2163
+    },
+    {
+      "epoch": 4.930444697833523,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.766939879251333e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.876966580748558,
+      "num_tokens": 231461394.0,
+      "step": 2164
+    },
+    {
+      "epoch": 4.932725199543899,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.7650671223165726e-06,
+      "loss": 0.5547,
+      "mean_token_accuracy": 0.8889574855566025,
+      "num_tokens": 231568225.0,
+      "step": 2165
+    },
+    {
+      "epoch": 4.935005701254276,
+      "grad_norm": 3.0,
+      "learning_rate": 2.7631942149348313e-06,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8834359645843506,
+      "num_tokens": 231675236.0,
+      "step": 2166
+    },
+    {
+      "epoch": 4.9372862029646525,
+      "grad_norm": 4.78125,
+      "learning_rate": 2.761321158169134e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8793386965990067,
+      "num_tokens": 231782454.0,
+      "step": 2167
+    },
+    {
+      "epoch": 4.939566704675029,
+      "grad_norm": 4.78125,
+      "learning_rate": 2.759447953082593e-06,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8823465257883072,
+      "num_tokens": 231889063.0,
+      "step": 2168
+    },
+    {
+      "epoch": 4.941847206385405,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.757574600738402e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.8793519288301468,
+      "num_tokens": 231995594.0,
+      "step": 2169
+    },
+    {
+      "epoch": 4.944127708095781,
+      "grad_norm": 4.78125,
+      "learning_rate": 2.755701102199841e-06,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8815154582262039,
+      "num_tokens": 232102587.0,
+      "step": 2170
+    },
+    {
+      "epoch": 4.946408209806157,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.7538274585302707e-06,
+      "loss": 0.5541,
+      "mean_token_accuracy": 0.8854823410511017,
+      "num_tokens": 232209760.0,
+      "step": 2171
+    },
+    {
+      "epoch": 4.9486887115165334,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.751953670793135e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8862673789262772,
+      "num_tokens": 232316636.0,
+      "step": 2172
+    },
+    {
+      "epoch": 4.95096921322691,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.7500797400519595e-06,
+      "loss": 0.5792,
+      "mean_token_accuracy": 0.8799717128276825,
+      "num_tokens": 232423639.0,
+      "step": 2173
+    },
+    {
+      "epoch": 4.953249714937286,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.7482056673703526e-06,
+      "loss": 0.589,
+      "mean_token_accuracy": 0.8792567402124405,
+      "num_tokens": 232530768.0,
+      "step": 2174
+    },
+    {
+      "epoch": 4.955530216647663,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.746331453812e-06,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8782864660024643,
+      "num_tokens": 232637783.0,
+      "step": 2175
+    },
+    {
+      "epoch": 4.957810718358039,
+      "grad_norm": 3.125,
+      "learning_rate": 2.74445710044067e-06,
+      "loss": 0.5983,
+      "mean_token_accuracy": 0.8764726370573044,
+      "num_tokens": 232745202.0,
+      "step": 2176
+    },
+    {
+      "epoch": 4.960091220068415,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.7425826083202096e-06,
+      "loss": 0.5832,
+      "mean_token_accuracy": 0.8824016451835632,
+      "num_tokens": 232852068.0,
+      "step": 2177
+    },
+    {
+      "epoch": 4.9623717217787915,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.740707978514543e-06,
+      "loss": 0.5526,
+      "mean_token_accuracy": 0.8894577920436859,
+      "num_tokens": 232959452.0,
+      "step": 2178
+    },
+    {
+      "epoch": 4.964652223489168,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.738833212087676e-06,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8816526979207993,
+      "num_tokens": 233066740.0,
+      "step": 2179
+    },
+    {
+      "epoch": 4.966932725199544,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.736958310103688e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8829785734415054,
+      "num_tokens": 233173753.0,
+      "step": 2180
+    },
+    {
+      "epoch": 4.96921322690992,
+      "grad_norm": 2.421875,
+      "learning_rate": 2.735083273626738e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8816793113946915,
+      "num_tokens": 233280384.0,
+      "step": 2181
+    },
+    {
+      "epoch": 4.971493728620296,
+      "grad_norm": 5.0,
+      "learning_rate": 2.7332081037210607e-06,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8826627433300018,
+      "num_tokens": 233388025.0,
+      "step": 2182
+    },
+    {
+      "epoch": 4.9737742303306725,
+      "grad_norm": 3.875,
+      "learning_rate": 2.7313328014509653e-06,
+      "loss": 0.5789,
+      "mean_token_accuracy": 0.8806509524583817,
+      "num_tokens": 233494917.0,
+      "step": 2183
+    },
+    {
+      "epoch": 4.976054732041049,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.729457367880838e-06,
+      "loss": 0.5431,
+      "mean_token_accuracy": 0.886359453201294,
+      "num_tokens": 233602067.0,
+      "step": 2184
+    },
+    {
+      "epoch": 4.978335233751425,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.727581804075139e-06,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8818333595991135,
+      "num_tokens": 233708533.0,
+      "step": 2185
+    },
+    {
+      "epoch": 4.980615735461802,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.7257061110984005e-06,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8846048712730408,
+      "num_tokens": 233815478.0,
+      "step": 2186
+    },
+    {
+      "epoch": 4.982896237172178,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.7238302900152327e-06,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8843528181314468,
+      "num_tokens": 233922328.0,
+      "step": 2187
+    },
+    {
+      "epoch": 4.985176738882554,
+      "grad_norm": 5.40625,
+      "learning_rate": 2.7219543418903115e-06,
+      "loss": 0.6009,
+      "mean_token_accuracy": 0.8763009756803513,
+      "num_tokens": 234028772.0,
+      "step": 2188
+    },
+    {
+      "epoch": 4.9874572405929305,
+      "grad_norm": 2.4375,
+      "learning_rate": 2.720078267788392e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8788257986307144,
+      "num_tokens": 234135491.0,
+      "step": 2189
+    },
+    {
+      "epoch": 4.989737742303307,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.718202068774296e-06,
+      "loss": 0.5923,
+      "mean_token_accuracy": 0.8749925345182419,
+      "num_tokens": 234243399.0,
+      "step": 2190
+    },
+    {
+      "epoch": 4.992018244013683,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.7163257459129184e-06,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8842321634292603,
+      "num_tokens": 234350519.0,
+      "step": 2191
+    },
+    {
+      "epoch": 4.994298745724059,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.7144493002692242e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8810799866914749,
+      "num_tokens": 234457691.0,
+      "step": 2192
+    },
+    {
+      "epoch": 4.996579247434435,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.7125727329082474e-06,
+      "loss": 0.568,
+      "mean_token_accuracy": 0.8835117220878601,
+      "num_tokens": 234564624.0,
+      "step": 2193
+    },
+    {
+      "epoch": 4.9988597491448115,
+      "grad_norm": 4.625,
+      "learning_rate": 2.7106960448950904e-06,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.8846527189016342,
+      "num_tokens": 234671664.0,
+      "step": 2194
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.7088192372949267e-06,
+      "loss": 0.5468,
+      "mean_token_accuracy": 0.893031507730484,
+      "num_tokens": 234711160.0,
+      "step": 2195
+    },
+    {
+      "epoch": 5.002280501710376,
+      "grad_norm": 2.40625,
+      "learning_rate": 2.7069423111729948e-06,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8820315301418304,
+      "num_tokens": 234818290.0,
+      "step": 2196
+    },
+    {
+      "epoch": 5.004561003420752,
+      "grad_norm": 2.625,
+      "learning_rate": 2.705065267594602e-06,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8801742941141129,
+      "num_tokens": 234925842.0,
+      "step": 2197
+    },
+    {
+      "epoch": 5.006841505131129,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.703188107625123e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8789117336273193,
+      "num_tokens": 235032605.0,
+      "step": 2198
+    },
+    {
+      "epoch": 5.009122006841505,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.701310832329996e-06,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8840703815221786,
+      "num_tokens": 235139505.0,
+      "step": 2199
+    },
+    {
+      "epoch": 5.011402508551882,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.6994334427747276e-06,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8826315701007843,
+      "num_tokens": 235246340.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.011402508551882,
+      "eval_loss": 0.5884966850280762,
+      "eval_mean_token_accuracy": 0.8795779687370184,
+      "eval_num_tokens": 235246340.0,
+      "eval_runtime": 58.673,
+      "eval_samples_per_second": 142.911,
+      "eval_steps_per_second": 4.482,
+      "step": 2200
+    },
+    {
+      "epoch": 5.013683010262258,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.6975559400248876e-06,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8832818120718002,
+      "num_tokens": 235354018.0,
+      "step": 2201
+    },
+    {
+      "epoch": 5.015963511972634,
+      "grad_norm": 3.625,
+      "learning_rate": 2.6956783251461093e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8836443275213242,
+      "num_tokens": 235461263.0,
+      "step": 2202
+    },
+    {
+      "epoch": 5.01824401368301,
+      "grad_norm": 5.09375,
+      "learning_rate": 2.6938005992040923e-06,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.88588847219944,
+      "num_tokens": 235568391.0,
+      "step": 2203
+    },
+    {
+      "epoch": 5.020524515393387,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.6919227632645963e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8812144100666046,
+      "num_tokens": 235675302.0,
+      "step": 2204
+    },
+    {
+      "epoch": 5.022805017103763,
+      "grad_norm": 5.46875,
+      "learning_rate": 2.690044818393444e-06,
+      "loss": 0.5707,
+      "mean_token_accuracy": 0.8809924572706223,
+      "num_tokens": 235781916.0,
+      "step": 2205
+    },
+    {
+      "epoch": 5.025085518814139,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.688166765656523e-06,
+      "loss": 0.5886,
+      "mean_token_accuracy": 0.8800527006387711,
+      "num_tokens": 235889095.0,
+      "step": 2206
+    },
+    {
+      "epoch": 5.027366020524515,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.686288606119778e-06,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8826051503419876,
+      "num_tokens": 235995817.0,
+      "step": 2207
+    },
+    {
+      "epoch": 5.029646522234891,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.6844103408492165e-06,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8821920156478882,
+      "num_tokens": 236102620.0,
+      "step": 2208
+    },
+    {
+      "epoch": 5.031927023945268,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.682531970910906e-06,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.882752850651741,
+      "num_tokens": 236210097.0,
+      "step": 2209
+    },
+    {
+      "epoch": 5.034207525655645,
+      "grad_norm": 3.609375,
+      "learning_rate": 2.6806534973709723e-06,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.8835909813642502,
+      "num_tokens": 236317553.0,
+      "step": 2210
+    },
+    {
+      "epoch": 5.036488027366021,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.6787749212956023e-06,
+      "loss": 0.5567,
+      "mean_token_accuracy": 0.8868236839771271,
+      "num_tokens": 236424162.0,
+      "step": 2211
+    },
+    {
+      "epoch": 5.038768529076397,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.676896243751037e-06,
+      "loss": 0.6022,
+      "mean_token_accuracy": 0.8755950033664703,
+      "num_tokens": 236531359.0,
+      "step": 2212
+    },
+    {
+      "epoch": 5.041049030786773,
+      "grad_norm": 2.1875,
+      "learning_rate": 2.6750174658035793e-06,
+      "loss": 0.559,
+      "mean_token_accuracy": 0.8848745375871658,
+      "num_tokens": 236638705.0,
+      "step": 2213
+    },
+    {
+      "epoch": 5.043329532497149,
+      "grad_norm": 4.5,
+      "learning_rate": 2.673138588519587e-06,
+      "loss": 0.6018,
+      "mean_token_accuracy": 0.8795148730278015,
+      "num_tokens": 236745788.0,
+      "step": 2214
+    },
+    {
+      "epoch": 5.045610034207526,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.671259612965475e-06,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8808691650629044,
+      "num_tokens": 236852575.0,
+      "step": 2215
+    },
+    {
+      "epoch": 5.047890535917902,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.6693805402077123e-06,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.8838084042072296,
+      "num_tokens": 236959458.0,
+      "step": 2216
+    },
+    {
+      "epoch": 5.050171037628278,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.6675013713128252e-06,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8833956867456436,
+      "num_tokens": 237067046.0,
+      "step": 2217
+    },
+    {
+      "epoch": 5.052451539338654,
+      "grad_norm": 3.0,
+      "learning_rate": 2.665622107347393e-06,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8833256661891937,
+      "num_tokens": 237174163.0,
+      "step": 2218
+    },
+    {
+      "epoch": 5.05473204104903,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.6637427493780503e-06,
+      "loss": 0.5704,
+      "mean_token_accuracy": 0.8838570863008499,
+      "num_tokens": 237280734.0,
+      "step": 2219
+    },
+    {
+      "epoch": 5.0570125427594075,
+      "grad_norm": 5.21875,
+      "learning_rate": 2.6618632984714843e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8816571533679962,
+      "num_tokens": 237388128.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.059293044469784,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.6599837556944353e-06,
+      "loss": 0.6045,
+      "mean_token_accuracy": 0.8754912316799164,
+      "num_tokens": 237494752.0,
+      "step": 2221
+    },
+    {
+      "epoch": 5.06157354618016,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.658104122113695e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8795156031847,
+      "num_tokens": 237601646.0,
+      "step": 2222
+    },
+    {
+      "epoch": 5.063854047890536,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.6562243987961066e-06,
+      "loss": 0.5483,
+      "mean_token_accuracy": 0.889316976070404,
+      "num_tokens": 237709147.0,
+      "step": 2223
+    },
+    {
+      "epoch": 5.066134549600912,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.6543445868085665e-06,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8827229887247086,
+      "num_tokens": 237816406.0,
+      "step": 2224
+    },
+    {
+      "epoch": 5.068415051311288,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.652464687218018e-06,
+      "loss": 0.5996,
+      "mean_token_accuracy": 0.8769262731075287,
+      "num_tokens": 237923355.0,
+      "step": 2225
+    },
+    {
+      "epoch": 5.070695553021665,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.6505847010914575e-06,
+      "loss": 0.5856,
+      "mean_token_accuracy": 0.8756005167961121,
+      "num_tokens": 238030151.0,
+      "step": 2226
+    },
+    {
+      "epoch": 5.072976054732041,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.6487046294959275e-06,
+      "loss": 0.5846,
+      "mean_token_accuracy": 0.8792047202587128,
+      "num_tokens": 238137145.0,
+      "step": 2227
+    },
+    {
+      "epoch": 5.075256556442417,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.64682447349852e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8841764777898788,
+      "num_tokens": 238243562.0,
+      "step": 2228
+    },
+    {
+      "epoch": 5.077537058152793,
+      "grad_norm": 3.375,
+      "learning_rate": 2.6449442341663755e-06,
+      "loss": 0.5626,
+      "mean_token_accuracy": 0.8807887881994247,
+      "num_tokens": 238350893.0,
+      "step": 2229
+    },
+    {
+      "epoch": 5.07981755986317,
+      "grad_norm": 3.5,
+      "learning_rate": 2.643063912566683e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8790174126625061,
+      "num_tokens": 238457625.0,
+      "step": 2230
+    },
+    {
+      "epoch": 5.0820980615735465,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.641183509766675e-06,
+      "loss": 0.5586,
+      "mean_token_accuracy": 0.8864348530769348,
+      "num_tokens": 238564821.0,
+      "step": 2231
+    },
+    {
+      "epoch": 5.084378563283923,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.639303026833632e-06,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8856208771467209,
+      "num_tokens": 238671917.0,
+      "step": 2232
+    },
+    {
+      "epoch": 5.086659064994299,
+      "grad_norm": 5.46875,
+      "learning_rate": 2.6374224648348815e-06,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8855437189340591,
+      "num_tokens": 238779139.0,
+      "step": 2233
+    },
+    {
+      "epoch": 5.088939566704675,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.6355418248377928e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8826903700828552,
+      "num_tokens": 238886730.0,
+      "step": 2234
+    },
+    {
+      "epoch": 5.091220068415051,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.633661107909781e-06,
+      "loss": 0.5568,
+      "mean_token_accuracy": 0.8886370956897736,
+      "num_tokens": 238993760.0,
+      "step": 2235
+    },
+    {
+      "epoch": 5.0935005701254275,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.6317803151183053e-06,
+      "loss": 0.5576,
+      "mean_token_accuracy": 0.8850623667240143,
+      "num_tokens": 239100806.0,
+      "step": 2236
+    },
+    {
+      "epoch": 5.095781071835804,
+      "grad_norm": 2.625,
+      "learning_rate": 2.629899447530866e-06,
+      "loss": 0.5615,
+      "mean_token_accuracy": 0.8851571381092072,
+      "num_tokens": 239209162.0,
+      "step": 2237
+    },
+    {
+      "epoch": 5.09806157354618,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.6280185062150084e-06,
+      "loss": 0.5565,
+      "mean_token_accuracy": 0.8866082727909088,
+      "num_tokens": 239316377.0,
+      "step": 2238
+    },
+    {
+      "epoch": 5.100342075256556,
+      "grad_norm": 3.84375,
+      "learning_rate": 2.6261374922383176e-06,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8804745078086853,
+      "num_tokens": 239423635.0,
+      "step": 2239
+    },
+    {
+      "epoch": 5.102622576966933,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.6242564066684217e-06,
+      "loss": 0.5942,
+      "mean_token_accuracy": 0.8799638897180557,
+      "num_tokens": 239530264.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.104903078677309,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.6223752505729884e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8811380565166473,
+      "num_tokens": 239637865.0,
+      "step": 2241
+    },
+    {
+      "epoch": 5.1071835803876855,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.6204940250197253e-06,
+      "loss": 0.5634,
+      "mean_token_accuracy": 0.8820418566465378,
+      "num_tokens": 239744908.0,
+      "step": 2242
+    },
+    {
+      "epoch": 5.109464082098062,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.61861273107638e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8828686624765396,
+      "num_tokens": 239852149.0,
+      "step": 2243
+    },
+    {
+      "epoch": 5.111744583808438,
+      "grad_norm": 6.84375,
+      "learning_rate": 2.6167313698107385e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.880137637257576,
+      "num_tokens": 239959160.0,
+      "step": 2244
+    },
+    {
+      "epoch": 5.114025085518814,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.6148499422906243e-06,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8834390491247177,
+      "num_tokens": 240066499.0,
+      "step": 2245
+    },
+    {
+      "epoch": 5.11630558722919,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.6129684495839013e-06,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8807038068771362,
+      "num_tokens": 240173494.0,
+      "step": 2246
+    },
+    {
+      "epoch": 5.1185860889395665,
+      "grad_norm": 4.5,
+      "learning_rate": 2.611086892758467e-06,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.883940801024437,
+      "num_tokens": 240280234.0,
+      "step": 2247
+    },
+    {
+      "epoch": 5.120866590649943,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.6092052728822564e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8823877573013306,
+      "num_tokens": 240387350.0,
+      "step": 2248
+    },
+    {
+      "epoch": 5.123147092360319,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.607323591023242e-06,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8841145634651184,
+      "num_tokens": 240494417.0,
+      "step": 2249
+    },
+    {
+      "epoch": 5.125427594070696,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.605441848249428e-06,
+      "loss": 0.5501,
+      "mean_token_accuracy": 0.8849078863859177,
+      "num_tokens": 240601696.0,
+      "step": 2250
+    },
+    {
+      "epoch": 5.127708095781072,
+      "grad_norm": 6.34375,
+      "learning_rate": 2.6035600456288573e-06,
+      "loss": 0.5773,
+      "mean_token_accuracy": 0.8804211169481277,
+      "num_tokens": 240708216.0,
+      "step": 2251
+    },
+    {
+      "epoch": 5.129988597491448,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.6016781842296044e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.881677657365799,
+      "num_tokens": 240814705.0,
+      "step": 2252
+    },
+    {
+      "epoch": 5.1322690992018245,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.599796265119777e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.881735309958458,
+      "num_tokens": 240921724.0,
+      "step": 2253
+    },
+    {
+      "epoch": 5.134549600912201,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.597914289367516e-06,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8814171552658081,
+      "num_tokens": 241029263.0,
+      "step": 2254
+    },
+    {
+      "epoch": 5.136830102622577,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.596032258040994e-06,
+      "loss": 0.5699,
+      "mean_token_accuracy": 0.8828539550304413,
+      "num_tokens": 241136198.0,
+      "step": 2255
+    },
+    {
+      "epoch": 5.139110604332953,
+      "grad_norm": 3.375,
+      "learning_rate": 2.594150172208417e-06,
+      "loss": 0.5987,
+      "mean_token_accuracy": 0.872844010591507,
+      "num_tokens": 241243108.0,
+      "step": 2256
+    },
+    {
+      "epoch": 5.141391106043329,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.59226803293802e-06,
+      "loss": 0.5773,
+      "mean_token_accuracy": 0.881134495139122,
+      "num_tokens": 241350038.0,
+      "step": 2257
+    },
+    {
+      "epoch": 5.1436716077537055,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.5903858412980688e-06,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.882486030459404,
+      "num_tokens": 241456705.0,
+      "step": 2258
+    },
+    {
+      "epoch": 5.145952109464082,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.5885035983568584e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8809113800525665,
+      "num_tokens": 241564205.0,
+      "step": 2259
+    },
+    {
+      "epoch": 5.148232611174459,
+      "grad_norm": 8.3125,
+      "learning_rate": 2.5866213051827148e-06,
+      "loss": 0.5681,
+      "mean_token_accuracy": 0.8836153000593185,
+      "num_tokens": 241671302.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.150513112884835,
+      "grad_norm": 4.96875,
+      "learning_rate": 2.5847389628439905e-06,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8828669935464859,
+      "num_tokens": 241778608.0,
+      "step": 2261
+    },
+    {
+      "epoch": 5.152793614595211,
+      "grad_norm": 6.4375,
+      "learning_rate": 2.5828565724090672e-06,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.881557047367096,
+      "num_tokens": 241886293.0,
+      "step": 2262
+    },
+    {
+      "epoch": 5.155074116305587,
+      "grad_norm": 4.25,
+      "learning_rate": 2.5809741349463526e-06,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.882814958691597,
+      "num_tokens": 241993190.0,
+      "step": 2263
+    },
+    {
+      "epoch": 5.1573546180159635,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.579091651524282e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8815194815397263,
+      "num_tokens": 242100380.0,
+      "step": 2264
+    },
+    {
+      "epoch": 5.15963511972634,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.5772091232113176e-06,
+      "loss": 0.5659,
+      "mean_token_accuracy": 0.8794299364089966,
+      "num_tokens": 242207689.0,
+      "step": 2265
+    },
+    {
+      "epoch": 5.161915621436716,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.575326551075945e-06,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8856062591075897,
+      "num_tokens": 242314652.0,
+      "step": 2266
+    },
+    {
+      "epoch": 5.164196123147092,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.5734439361866762e-06,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8838386088609695,
+      "num_tokens": 242422494.0,
+      "step": 2267
+    },
+    {
+      "epoch": 5.166476624857468,
+      "grad_norm": 5.875,
+      "learning_rate": 2.571561279612047e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8841647505760193,
+      "num_tokens": 242529583.0,
+      "step": 2268
+    },
+    {
+      "epoch": 5.168757126567845,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.5696785824206177e-06,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8775971680879593,
+      "num_tokens": 242636422.0,
+      "step": 2269
+    },
+    {
+      "epoch": 5.1710376282782216,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.5677958456809703e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8794868886470795,
+      "num_tokens": 242742862.0,
+      "step": 2270
+    },
+    {
+      "epoch": 5.173318129988598,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.5659130704617092e-06,
+      "loss": 0.579,
+      "mean_token_accuracy": 0.8852169960737228,
+      "num_tokens": 242850283.0,
+      "step": 2271
+    },
+    {
+      "epoch": 5.175598631698974,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.5640302578314614e-06,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.8777539879083633,
+      "num_tokens": 242957389.0,
+      "step": 2272
+    },
+    {
+      "epoch": 5.17787913340935,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.562147408858876e-06,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8816378861665726,
+      "num_tokens": 243064479.0,
+      "step": 2273
+    },
+    {
+      "epoch": 5.180159635119726,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.5602645246126207e-06,
+      "loss": 0.5922,
+      "mean_token_accuracy": 0.877925843000412,
+      "num_tokens": 243171634.0,
+      "step": 2274
+    },
+    {
+      "epoch": 5.1824401368301025,
+      "grad_norm": 3.125,
+      "learning_rate": 2.5583816061613847e-06,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.885333925485611,
+      "num_tokens": 243278723.0,
+      "step": 2275
+    },
+    {
+      "epoch": 5.184720638540479,
+      "grad_norm": 2.296875,
+      "learning_rate": 2.5564986545738767e-06,
+      "loss": 0.5537,
+      "mean_token_accuracy": 0.8853138238191605,
+      "num_tokens": 243386191.0,
+      "step": 2276
+    },
+    {
+      "epoch": 5.187001140250855,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.554615670918823e-06,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.8801163583993912,
+      "num_tokens": 243492478.0,
+      "step": 2277
+    },
+    {
+      "epoch": 5.189281641961231,
+      "grad_norm": 3.0,
+      "learning_rate": 2.552732656264969e-06,
+      "loss": 0.5707,
+      "mean_token_accuracy": 0.8835473358631134,
+      "num_tokens": 243599878.0,
+      "step": 2278
+    },
+    {
+      "epoch": 5.191562143671608,
+      "grad_norm": 3.84375,
+      "learning_rate": 2.5508496116810766e-06,
+      "loss": 0.6136,
+      "mean_token_accuracy": 0.8747565001249313,
+      "num_tokens": 243706837.0,
+      "step": 2279
+    },
+    {
+      "epoch": 5.193842645381984,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.548966538235927e-06,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8776230216026306,
+      "num_tokens": 243813746.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.196123147092361,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.547083436998316e-06,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8837112933397293,
+      "num_tokens": 243920433.0,
+      "step": 2281
+    },
+    {
+      "epoch": 5.198403648802737,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.5452003090370543e-06,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8835209310054779,
+      "num_tokens": 244027394.0,
+      "step": 2282
+    },
+    {
+      "epoch": 5.200684150513113,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.5433171554209694e-06,
+      "loss": 0.5585,
+      "mean_token_accuracy": 0.8822397440671921,
+      "num_tokens": 244134310.0,
+      "step": 2283
+    },
+    {
+      "epoch": 5.202964652223489,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.5414339772189045e-06,
+      "loss": 0.5558,
+      "mean_token_accuracy": 0.8861335813999176,
+      "num_tokens": 244241565.0,
+      "step": 2284
+    },
+    {
+      "epoch": 5.205245153933865,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.5395507754997135e-06,
+      "loss": 0.5562,
+      "mean_token_accuracy": 0.8825538158416748,
+      "num_tokens": 244349038.0,
+      "step": 2285
+    },
+    {
+      "epoch": 5.2075256556442415,
+      "grad_norm": 2.375,
+      "learning_rate": 2.5376675513322665e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8824332058429718,
+      "num_tokens": 244455939.0,
+      "step": 2286
+    },
+    {
+      "epoch": 5.209806157354618,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.535784305785443e-06,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8791725635528564,
+      "num_tokens": 244563202.0,
+      "step": 2287
+    },
+    {
+      "epoch": 5.212086659064994,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.5339010399281394e-06,
+      "loss": 0.5615,
+      "mean_token_accuracy": 0.8839509189128876,
+      "num_tokens": 244670656.0,
+      "step": 2288
+    },
+    {
+      "epoch": 5.214367160775371,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.53201775482926e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8803385496139526,
+      "num_tokens": 244777602.0,
+      "step": 2289
+    },
+    {
+      "epoch": 5.216647662485747,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.530134451557722e-06,
+      "loss": 0.5667,
+      "mean_token_accuracy": 0.8833064287900925,
+      "num_tokens": 244884757.0,
+      "step": 2290
+    },
+    {
+      "epoch": 5.218928164196123,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.52825113118245e-06,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8829501569271088,
+      "num_tokens": 244992076.0,
+      "step": 2291
+    },
+    {
+      "epoch": 5.2212086659065,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.5263677947723813e-06,
+      "loss": 0.5773,
+      "mean_token_accuracy": 0.8804136514663696,
+      "num_tokens": 245099111.0,
+      "step": 2292
+    },
+    {
+      "epoch": 5.223489167616876,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.5244844433964615e-06,
+      "loss": 0.5481,
+      "mean_token_accuracy": 0.8848851323127747,
+      "num_tokens": 245206635.0,
+      "step": 2293
+    },
+    {
+      "epoch": 5.225769669327252,
+      "grad_norm": 2.484375,
+      "learning_rate": 2.522601078123645e-06,
+      "loss": 0.5586,
+      "mean_token_accuracy": 0.8856381624937057,
+      "num_tokens": 245313995.0,
+      "step": 2294
+    },
+    {
+      "epoch": 5.228050171037628,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.5207177000228916e-06,
+      "loss": 0.5651,
+      "mean_token_accuracy": 0.8863477110862732,
+      "num_tokens": 245420590.0,
+      "step": 2295
+    },
+    {
+      "epoch": 5.230330672748004,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.5188343101631717e-06,
+      "loss": 0.5998,
+      "mean_token_accuracy": 0.8750515878200531,
+      "num_tokens": 245527620.0,
+      "step": 2296
+    },
+    {
+      "epoch": 5.2326111744583805,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.516950909613462e-06,
+      "loss": 0.5574,
+      "mean_token_accuracy": 0.8836745172739029,
+      "num_tokens": 245634856.0,
+      "step": 2297
+    },
+    {
+      "epoch": 5.234891676168757,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.5150674994427427e-06,
+      "loss": 0.5914,
+      "mean_token_accuracy": 0.8803335726261139,
+      "num_tokens": 245741637.0,
+      "step": 2298
+    },
+    {
+      "epoch": 5.237172177879134,
+      "grad_norm": 6.40625,
+      "learning_rate": 2.5131840807200015e-06,
+      "loss": 0.5603,
+      "mean_token_accuracy": 0.8841415345668793,
+      "num_tokens": 245850208.0,
+      "step": 2299
+    },
+    {
+      "epoch": 5.23945267958951,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.511300654514231e-06,
+      "loss": 0.6017,
+      "mean_token_accuracy": 0.8761951625347137,
+      "num_tokens": 245956827.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.241733181299886,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.5094172218944276e-06,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8835689127445221,
+      "num_tokens": 246064233.0,
+      "step": 2301
+    },
+    {
+      "epoch": 5.244013683010262,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.5075337839295903e-06,
+      "loss": 0.5553,
+      "mean_token_accuracy": 0.881672739982605,
+      "num_tokens": 246171103.0,
+      "step": 2302
+    },
+    {
+      "epoch": 5.246294184720639,
+      "grad_norm": 6.4375,
+      "learning_rate": 2.5056503416887222e-06,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8831684589385986,
+      "num_tokens": 246278163.0,
+      "step": 2303
+    },
+    {
+      "epoch": 5.248574686431015,
+      "grad_norm": 3.375,
+      "learning_rate": 2.5037668962408295e-06,
+      "loss": 0.5733,
+      "mean_token_accuracy": 0.8839241862297058,
+      "num_tokens": 246385460.0,
+      "step": 2304
+    },
+    {
+      "epoch": 5.250855188141391,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.5018834486549198e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8784928619861603,
+      "num_tokens": 246492696.0,
+      "step": 2305
+    },
+    {
+      "epoch": 5.253135689851767,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.5e-06,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.884750634431839,
+      "num_tokens": 246599773.0,
+      "step": 2306
+    },
+    {
+      "epoch": 5.255416191562143,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.4981165513450807e-06,
+      "loss": 0.5908,
+      "mean_token_accuracy": 0.8797119110822678,
+      "num_tokens": 246706234.0,
+      "step": 2307
+    },
+    {
+      "epoch": 5.2576966932725195,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.4962331037591705e-06,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8851300626993179,
+      "num_tokens": 246813408.0,
+      "step": 2308
+    },
+    {
+      "epoch": 5.259977194982897,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.494349658311279e-06,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8815591186285019,
+      "num_tokens": 246920031.0,
+      "step": 2309
+    },
+    {
+      "epoch": 5.262257696693273,
+      "grad_norm": 5.46875,
+      "learning_rate": 2.492466216070411e-06,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.8796799033880234,
+      "num_tokens": 247027919.0,
+      "step": 2310
+    },
+    {
+      "epoch": 5.264538198403649,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.4905827781055733e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8780756443738937,
+      "num_tokens": 247134526.0,
+      "step": 2311
+    },
+    {
+      "epoch": 5.266818700114025,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.4886993454857696e-06,
+      "loss": 0.5598,
+      "mean_token_accuracy": 0.8812786191701889,
+      "num_tokens": 247242308.0,
+      "step": 2312
+    },
+    {
+      "epoch": 5.269099201824401,
+      "grad_norm": 2.875,
+      "learning_rate": 2.486815919279999e-06,
+      "loss": 0.5608,
+      "mean_token_accuracy": 0.8846468031406403,
+      "num_tokens": 247349006.0,
+      "step": 2313
+    },
+    {
+      "epoch": 5.271379703534778,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.4849325005572573e-06,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8841832727193832,
+      "num_tokens": 247455756.0,
+      "step": 2314
+    },
+    {
+      "epoch": 5.273660205245154,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.483049090386539e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8810638189315796,
+      "num_tokens": 247562448.0,
+      "step": 2315
+    },
+    {
+      "epoch": 5.27594070695553,
+      "grad_norm": 5.25,
+      "learning_rate": 2.4811656898368287e-06,
+      "loss": 0.5921,
+      "mean_token_accuracy": 0.87621209025383,
+      "num_tokens": 247669493.0,
+      "step": 2316
+    },
+    {
+      "epoch": 5.278221208665906,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.4792822999771092e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8810609132051468,
+      "num_tokens": 247776380.0,
+      "step": 2317
+    },
+    {
+      "epoch": 5.280501710376283,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.477398921876356e-06,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.8778067380189896,
+      "num_tokens": 247883246.0,
+      "step": 2318
+    },
+    {
+      "epoch": 5.282782212086659,
+      "grad_norm": 4.5,
+      "learning_rate": 2.475515556603539e-06,
+      "loss": 0.5648,
+      "mean_token_accuracy": 0.882491260766983,
+      "num_tokens": 247990244.0,
+      "step": 2319
+    },
+    {
+      "epoch": 5.285062713797036,
+      "grad_norm": 5.875,
+      "learning_rate": 2.47363220522762e-06,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8826608210802078,
+      "num_tokens": 248097490.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.287343215507412,
+      "grad_norm": 4.40625,
+      "learning_rate": 2.4717488688175513e-06,
+      "loss": 0.576,
+      "mean_token_accuracy": 0.8805565237998962,
+      "num_tokens": 248204076.0,
+      "step": 2321
+    },
+    {
+      "epoch": 5.289623717217788,
+      "grad_norm": 2.75,
+      "learning_rate": 2.469865548442279e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8818432092666626,
+      "num_tokens": 248311212.0,
+      "step": 2322
+    },
+    {
+      "epoch": 5.291904218928164,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.4679822451707404e-06,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8799884766340256,
+      "num_tokens": 248418165.0,
+      "step": 2323
+    },
+    {
+      "epoch": 5.29418472063854,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.4660989600718606e-06,
+      "loss": 0.5607,
+      "mean_token_accuracy": 0.8832896500825882,
+      "num_tokens": 248525308.0,
+      "step": 2324
+    },
+    {
+      "epoch": 5.296465222348917,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.4642156942145577e-06,
+      "loss": 0.5588,
+      "mean_token_accuracy": 0.8862069100141525,
+      "num_tokens": 248632300.0,
+      "step": 2325
+    },
+    {
+      "epoch": 5.298745724059293,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.4623324486677352e-06,
+      "loss": 0.5821,
+      "mean_token_accuracy": 0.8789513111114502,
+      "num_tokens": 248739272.0,
+      "step": 2326
+    },
+    {
+      "epoch": 5.301026225769669,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.4604492245002873e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.881064772605896,
+      "num_tokens": 248846325.0,
+      "step": 2327
+    },
+    {
+      "epoch": 5.303306727480045,
+      "grad_norm": 2.453125,
+      "learning_rate": 2.4585660227810963e-06,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8820571899414062,
+      "num_tokens": 248953637.0,
+      "step": 2328
+    },
+    {
+      "epoch": 5.305587229190422,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.4566828445790306e-06,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8833505213260651,
+      "num_tokens": 249060027.0,
+      "step": 2329
+    },
+    {
+      "epoch": 5.307867730900798,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.454799690962946e-06,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8810784071683884,
+      "num_tokens": 249166827.0,
+      "step": 2330
+    },
+    {
+      "epoch": 5.310148232611175,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.4529165630016855e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8803735673427582,
+      "num_tokens": 249273474.0,
+      "step": 2331
+    },
+    {
+      "epoch": 5.312428734321551,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.4510334617640733e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8823344260454178,
+      "num_tokens": 249380314.0,
+      "step": 2332
+    },
+    {
+      "epoch": 5.314709236031927,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.4491503883189242e-06,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8847284018993378,
+      "num_tokens": 249487519.0,
+      "step": 2333
+    },
+    {
+      "epoch": 5.316989737742303,
+      "grad_norm": 4.625,
+      "learning_rate": 2.447267343735032e-06,
+      "loss": 0.5818,
+      "mean_token_accuracy": 0.8809306025505066,
+      "num_tokens": 249594226.0,
+      "step": 2334
+    },
+    {
+      "epoch": 5.319270239452679,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.4453843290811772e-06,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8837304264307022,
+      "num_tokens": 249701458.0,
+      "step": 2335
+    },
+    {
+      "epoch": 5.321550741163056,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.4435013454261246e-06,
+      "loss": 0.5898,
+      "mean_token_accuracy": 0.8787836730480194,
+      "num_tokens": 249807967.0,
+      "step": 2336
+    },
+    {
+      "epoch": 5.323831242873432,
+      "grad_norm": 3.625,
+      "learning_rate": 2.4416183938386157e-06,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8804430663585663,
+      "num_tokens": 249915439.0,
+      "step": 2337
+    },
+    {
+      "epoch": 5.326111744583809,
+      "grad_norm": 2.875,
+      "learning_rate": 2.4397354753873797e-06,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8802332282066345,
+      "num_tokens": 250022617.0,
+      "step": 2338
+    },
+    {
+      "epoch": 5.328392246294185,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.4378525911411246e-06,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8810295462608337,
+      "num_tokens": 250129347.0,
+      "step": 2339
+    },
+    {
+      "epoch": 5.330672748004561,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.435969742168539e-06,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8794845640659332,
+      "num_tokens": 250236273.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.3329532497149374,
+      "grad_norm": 3.75,
+      "learning_rate": 2.4340869295382924e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8810597956180573,
+      "num_tokens": 250343235.0,
+      "step": 2341
+    },
+    {
+      "epoch": 5.335233751425314,
+      "grad_norm": 5.25,
+      "learning_rate": 2.432204154319031e-06,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8840866684913635,
+      "num_tokens": 250450686.0,
+      "step": 2342
+    },
+    {
+      "epoch": 5.33751425313569,
+      "grad_norm": 3.0,
+      "learning_rate": 2.4303214175793827e-06,
+      "loss": 0.599,
+      "mean_token_accuracy": 0.8780745565891266,
+      "num_tokens": 250557785.0,
+      "step": 2343
+    },
+    {
+      "epoch": 5.339794754846066,
+      "grad_norm": 5.71875,
+      "learning_rate": 2.4284387203879536e-06,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8825893849134445,
+      "num_tokens": 250665664.0,
+      "step": 2344
+    },
+    {
+      "epoch": 5.342075256556442,
+      "grad_norm": 5.09375,
+      "learning_rate": 2.426556063813324e-06,
+      "loss": 0.5485,
+      "mean_token_accuracy": 0.8859640210866928,
+      "num_tokens": 250773371.0,
+      "step": 2345
+    },
+    {
+      "epoch": 5.344355758266818,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.4246734489240554e-06,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.883133664727211,
+      "num_tokens": 250880142.0,
+      "step": 2346
+    },
+    {
+      "epoch": 5.346636259977195,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.4227908767886837e-06,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8815203011035919,
+      "num_tokens": 250987935.0,
+      "step": 2347
+    },
+    {
+      "epoch": 5.348916761687571,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.420908348475719e-06,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.8813404738903046,
+      "num_tokens": 251094697.0,
+      "step": 2348
+    },
+    {
+      "epoch": 5.351197263397948,
+      "grad_norm": 4.9375,
+      "learning_rate": 2.4190258650536483e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.883138045668602,
+      "num_tokens": 251201631.0,
+      "step": 2349
+    },
+    {
+      "epoch": 5.353477765108324,
+      "grad_norm": 2.421875,
+      "learning_rate": 2.417143427590933e-06,
+      "loss": 0.5414,
+      "mean_token_accuracy": 0.8879837691783905,
+      "num_tokens": 251309130.0,
+      "step": 2350
+    },
+    {
+      "epoch": 5.3557582668187,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.4152610371560095e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.886492133140564,
+      "num_tokens": 251416268.0,
+      "step": 2351
+    },
+    {
+      "epoch": 5.3580387685290765,
+      "grad_norm": 7.125,
+      "learning_rate": 2.413378694817286e-06,
+      "loss": 0.5553,
+      "mean_token_accuracy": 0.8863291144371033,
+      "num_tokens": 251523690.0,
+      "step": 2352
+    },
+    {
+      "epoch": 5.360319270239453,
+      "grad_norm": 6.96875,
+      "learning_rate": 2.411496401643142e-06,
+      "loss": 0.59,
+      "mean_token_accuracy": 0.8787893354892731,
+      "num_tokens": 251630852.0,
+      "step": 2353
+    },
+    {
+      "epoch": 5.362599771949829,
+      "grad_norm": 4.25,
+      "learning_rate": 2.409614158701932e-06,
+      "loss": 0.5525,
+      "mean_token_accuracy": 0.888729602098465,
+      "num_tokens": 251738309.0,
+      "step": 2354
+    },
+    {
+      "epoch": 5.364880273660205,
+      "grad_norm": 5.0,
+      "learning_rate": 2.407731967061981e-06,
+      "loss": 0.559,
+      "mean_token_accuracy": 0.8860864639282227,
+      "num_tokens": 251845628.0,
+      "step": 2355
+    },
+    {
+      "epoch": 5.367160775370581,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.4058498277915835e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8805464655160904,
+      "num_tokens": 251951993.0,
+      "step": 2356
+    },
+    {
+      "epoch": 5.369441277080957,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.4039677419590064e-06,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8828859180212021,
+      "num_tokens": 252059284.0,
+      "step": 2357
+    },
+    {
+      "epoch": 5.3717217787913345,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.4020857106324853e-06,
+      "loss": 0.6042,
+      "mean_token_accuracy": 0.872930184006691,
+      "num_tokens": 252166320.0,
+      "step": 2358
+    },
+    {
+      "epoch": 5.374002280501711,
+      "grad_norm": 4.375,
+      "learning_rate": 2.4002037348802245e-06,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8799397349357605,
+      "num_tokens": 252273488.0,
+      "step": 2359
+    },
+    {
+      "epoch": 5.376282782212087,
+      "grad_norm": 6.0625,
+      "learning_rate": 2.3983218157703964e-06,
+      "loss": 0.6086,
+      "mean_token_accuracy": 0.8736053854227066,
+      "num_tokens": 252380458.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.378563283922463,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.3964399543711427e-06,
+      "loss": 0.59,
+      "mean_token_accuracy": 0.8807500749826431,
+      "num_tokens": 252487767.0,
+      "step": 2361
+    },
+    {
+      "epoch": 5.380843785632839,
+      "grad_norm": 5.84375,
+      "learning_rate": 2.394558151750572e-06,
+      "loss": 0.5971,
+      "mean_token_accuracy": 0.8786357641220093,
+      "num_tokens": 252595094.0,
+      "step": 2362
+    },
+    {
+      "epoch": 5.3831242873432155,
+      "grad_norm": 5.84375,
+      "learning_rate": 2.3926764089767594e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8844779282808304,
+      "num_tokens": 252702007.0,
+      "step": 2363
+    },
+    {
+      "epoch": 5.385404789053592,
+      "grad_norm": 6.125,
+      "learning_rate": 2.3907947271177444e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8814315497875214,
+      "num_tokens": 252808779.0,
+      "step": 2364
+    },
+    {
+      "epoch": 5.387685290763968,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.388913107241534e-06,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8800296932458878,
+      "num_tokens": 252915277.0,
+      "step": 2365
+    },
+    {
+      "epoch": 5.389965792474344,
+      "grad_norm": 3.375,
+      "learning_rate": 2.3870315504160995e-06,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8823367357254028,
+      "num_tokens": 253022601.0,
+      "step": 2366
+    },
+    {
+      "epoch": 5.39224629418472,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.3851500577093757e-06,
+      "loss": 0.5541,
+      "mean_token_accuracy": 0.8841412514448166,
+      "num_tokens": 253129731.0,
+      "step": 2367
+    },
+    {
+      "epoch": 5.394526795895097,
+      "grad_norm": 4.8125,
+      "learning_rate": 2.3832686301892628e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8834680169820786,
+      "num_tokens": 253236662.0,
+      "step": 2368
+    },
+    {
+      "epoch": 5.3968072976054735,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.381387268923621e-06,
+      "loss": 0.594,
+      "mean_token_accuracy": 0.8795076310634613,
+      "num_tokens": 253343900.0,
+      "step": 2369
+    },
+    {
+      "epoch": 5.39908779931585,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.3795059749802756e-06,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8820982426404953,
+      "num_tokens": 253450924.0,
+      "step": 2370
+    },
+    {
+      "epoch": 5.401368301026226,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.377624749427012e-06,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.884630486369133,
+      "num_tokens": 253558210.0,
+      "step": 2371
+    },
+    {
+      "epoch": 5.403648802736602,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.3757435933315787e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.878990963101387,
+      "num_tokens": 253665052.0,
+      "step": 2372
+    },
+    {
+      "epoch": 5.405929304446978,
+      "grad_norm": 5.03125,
+      "learning_rate": 2.3738625077616837e-06,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8815456032752991,
+      "num_tokens": 253772336.0,
+      "step": 2373
+    },
+    {
+      "epoch": 5.4082098061573545,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.371981493784993e-06,
+      "loss": 0.5833,
+      "mean_token_accuracy": 0.8819355368614197,
+      "num_tokens": 253878690.0,
+      "step": 2374
+    },
+    {
+      "epoch": 5.410490307867731,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.370100552469135e-06,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.8823709636926651,
+      "num_tokens": 253985411.0,
+      "step": 2375
+    },
+    {
+      "epoch": 5.412770809578107,
+      "grad_norm": 4.6875,
+      "learning_rate": 2.3682196848816955e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.881668359041214,
+      "num_tokens": 254092624.0,
+      "step": 2376
+    },
+    {
+      "epoch": 5.415051311288483,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.3663388920902198e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8824621737003326,
+      "num_tokens": 254199578.0,
+      "step": 2377
+    },
+    {
+      "epoch": 5.41733181299886,
+      "grad_norm": 4.6875,
+      "learning_rate": 2.3644581751622076e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8811022043228149,
+      "num_tokens": 254306615.0,
+      "step": 2378
+    },
+    {
+      "epoch": 5.419612314709236,
+      "grad_norm": 7.15625,
+      "learning_rate": 2.3625775351651193e-06,
+      "loss": 0.5955,
+      "mean_token_accuracy": 0.8790160417556763,
+      "num_tokens": 254413447.0,
+      "step": 2379
+    },
+    {
+      "epoch": 5.4218928164196125,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.3606969731663683e-06,
+      "loss": 0.5955,
+      "mean_token_accuracy": 0.8772003203630447,
+      "num_tokens": 254520623.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.424173318129989,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.358816490233326e-06,
+      "loss": 0.5646,
+      "mean_token_accuracy": 0.8835051357746124,
+      "num_tokens": 254627369.0,
+      "step": 2381
+    },
+    {
+      "epoch": 5.426453819840365,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.356936087433318e-06,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8843741118907928,
+      "num_tokens": 254734806.0,
+      "step": 2382
+    },
+    {
+      "epoch": 5.428734321550741,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.3550557658336245e-06,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.881999284029007,
+      "num_tokens": 254841536.0,
+      "step": 2383
+    },
+    {
+      "epoch": 5.431014823261117,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.3531755265014818e-06,
+      "loss": 0.5481,
+      "mean_token_accuracy": 0.8869676440954208,
+      "num_tokens": 254949872.0,
+      "step": 2384
+    },
+    {
+      "epoch": 5.4332953249714935,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.3512953705040737e-06,
+      "loss": 0.525,
+      "mean_token_accuracy": 0.894694447517395,
+      "num_tokens": 255057725.0,
+      "step": 2385
+    },
+    {
+      "epoch": 5.43557582668187,
+      "grad_norm": 4.90625,
+      "learning_rate": 2.3494152989085433e-06,
+      "loss": 0.5388,
+      "mean_token_accuracy": 0.888728454709053,
+      "num_tokens": 255165920.0,
+      "step": 2386
+    },
+    {
+      "epoch": 5.437856328392247,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.3475353127819827e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8810894787311554,
+      "num_tokens": 255273480.0,
+      "step": 2387
+    },
+    {
+      "epoch": 5.440136830102623,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.345655413191434e-06,
+      "loss": 0.5962,
+      "mean_token_accuracy": 0.880548432469368,
+      "num_tokens": 255380729.0,
+      "step": 2388
+    },
+    {
+      "epoch": 5.442417331812999,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.3437756012038933e-06,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.882551446557045,
+      "num_tokens": 255488672.0,
+      "step": 2389
+    },
+    {
+      "epoch": 5.444697833523375,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.341895877886306e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8806143552064896,
+      "num_tokens": 255595607.0,
+      "step": 2390
+    },
+    {
+      "epoch": 5.4469783352337515,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.3400162443055655e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8818208873271942,
+      "num_tokens": 255702745.0,
+      "step": 2391
+    },
+    {
+      "epoch": 5.449258836944128,
+      "grad_norm": 2.625,
+      "learning_rate": 2.338136701528516e-06,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8838256001472473,
+      "num_tokens": 255810202.0,
+      "step": 2392
+    },
+    {
+      "epoch": 5.451539338654504,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.33625725062195e-06,
+      "loss": 0.6016,
+      "mean_token_accuracy": 0.8813562989234924,
+      "num_tokens": 255916916.0,
+      "step": 2393
+    },
+    {
+      "epoch": 5.45381984036488,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.3343778926526074e-06,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8832430243492126,
+      "num_tokens": 256023471.0,
+      "step": 2394
+    },
+    {
+      "epoch": 5.456100342075256,
+      "grad_norm": 3.125,
+      "learning_rate": 2.332498628687176e-06,
+      "loss": 0.5833,
+      "mean_token_accuracy": 0.8809223771095276,
+      "num_tokens": 256130280.0,
+      "step": 2395
+    },
+    {
+      "epoch": 5.4583808437856325,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.330619459792289e-06,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8768937885761261,
+      "num_tokens": 256236939.0,
+      "step": 2396
+    },
+    {
+      "epoch": 5.460661345496009,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.328740387034526e-06,
+      "loss": 0.587,
+      "mean_token_accuracy": 0.8795687556266785,
+      "num_tokens": 256344069.0,
+      "step": 2397
+    },
+    {
+      "epoch": 5.462941847206386,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.326861411480414e-06,
+      "loss": 0.545,
+      "mean_token_accuracy": 0.8894121795892715,
+      "num_tokens": 256450947.0,
+      "step": 2398
+    },
+    {
+      "epoch": 5.465222348916762,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.324982534196421e-06,
+      "loss": 0.5617,
+      "mean_token_accuracy": 0.8822938203811646,
+      "num_tokens": 256558445.0,
+      "step": 2399
+    },
+    {
+      "epoch": 5.467502850627138,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.3231037562489636e-06,
+      "loss": 0.5599,
+      "mean_token_accuracy": 0.8841269761323929,
+      "num_tokens": 256665737.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.469783352337514,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.321225078704399e-06,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8798850178718567,
+      "num_tokens": 256772658.0,
+      "step": 2401
+    },
+    {
+      "epoch": 5.4720638540478905,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.319346502629028e-06,
+      "loss": 0.5642,
+      "mean_token_accuracy": 0.883221909403801,
+      "num_tokens": 256879693.0,
+      "step": 2402
+    },
+    {
+      "epoch": 5.474344355758267,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.3174680290890945e-06,
+      "loss": 0.596,
+      "mean_token_accuracy": 0.882306694984436,
+      "num_tokens": 256986979.0,
+      "step": 2403
+    },
+    {
+      "epoch": 5.476624857468643,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.315589659150784e-06,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8786382675170898,
+      "num_tokens": 257093930.0,
+      "step": 2404
+    },
+    {
+      "epoch": 5.478905359179019,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.3137113938802224e-06,
+      "loss": 0.5936,
+      "mean_token_accuracy": 0.8781066983938217,
+      "num_tokens": 257201360.0,
+      "step": 2405
+    },
+    {
+      "epoch": 5.481185860889395,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.311833234343478e-06,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8780955076217651,
+      "num_tokens": 257308294.0,
+      "step": 2406
+    },
+    {
+      "epoch": 5.483466362599772,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.3099551816065563e-06,
+      "loss": 0.6028,
+      "mean_token_accuracy": 0.8727833181619644,
+      "num_tokens": 257414698.0,
+      "step": 2407
+    },
+    {
+      "epoch": 5.485746864310149,
+      "grad_norm": 3.125,
+      "learning_rate": 2.3080772367354046e-06,
+      "loss": 0.5579,
+      "mean_token_accuracy": 0.888285905122757,
+      "num_tokens": 257521674.0,
+      "step": 2408
+    },
+    {
+      "epoch": 5.488027366020525,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.3061994007959086e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8786954879760742,
+      "num_tokens": 257628383.0,
+      "step": 2409
+    },
+    {
+      "epoch": 5.490307867730901,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.304321674853891e-06,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8847887217998505,
+      "num_tokens": 257735646.0,
+      "step": 2410
+    },
+    {
+      "epoch": 5.492588369441277,
+      "grad_norm": 4.65625,
+      "learning_rate": 2.3024440599751132e-06,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8823476582765579,
+      "num_tokens": 257843020.0,
+      "step": 2411
+    },
+    {
+      "epoch": 5.494868871151653,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.3005665572252732e-06,
+      "loss": 0.6067,
+      "mean_token_accuracy": 0.8751996457576752,
+      "num_tokens": 257949651.0,
+      "step": 2412
+    },
+    {
+      "epoch": 5.4971493728620295,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.2986891676700042e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8818869143724442,
+      "num_tokens": 258056645.0,
+      "step": 2413
+    },
+    {
+      "epoch": 5.499429874572406,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.296811892374878e-06,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8861507475376129,
+      "num_tokens": 258163786.0,
+      "step": 2414
+    },
+    {
+      "epoch": 5.501710376282782,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.294934732405398e-06,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8789652734994888,
+      "num_tokens": 258271199.0,
+      "step": 2415
+    },
+    {
+      "epoch": 5.503990877993158,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.293057688827007e-06,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.8831315487623215,
+      "num_tokens": 258378688.0,
+      "step": 2416
+    },
+    {
+      "epoch": 5.506271379703534,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.2911807627050745e-06,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8817677646875381,
+      "num_tokens": 258486085.0,
+      "step": 2417
+    },
+    {
+      "epoch": 5.508551881413911,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.2893039551049104e-06,
+      "loss": 0.5869,
+      "mean_token_accuracy": 0.8783139288425446,
+      "num_tokens": 258593269.0,
+      "step": 2418
+    },
+    {
+      "epoch": 5.510832383124288,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.2874272670917534e-06,
+      "loss": 0.5944,
+      "mean_token_accuracy": 0.8795038163661957,
+      "num_tokens": 258700704.0,
+      "step": 2419
+    },
+    {
+      "epoch": 5.513112884834664,
+      "grad_norm": 4.40625,
+      "learning_rate": 2.2855506997307766e-06,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.8858631104230881,
+      "num_tokens": 258807861.0,
+      "step": 2420
+    },
+    {
+      "epoch": 5.513112884834664,
+      "eval_loss": 0.5877215266227722,
+      "eval_mean_token_accuracy": 0.8795312627186793,
+      "eval_num_tokens": 258807861.0,
+      "eval_runtime": 58.5506,
+      "eval_samples_per_second": 143.21,
+      "eval_steps_per_second": 4.492,
+      "step": 2420
+    },
+    {
+      "epoch": 5.51539338654504,
+      "grad_norm": 5.46875,
+      "learning_rate": 2.283674254087082e-06,
+      "loss": 0.5588,
+      "mean_token_accuracy": 0.8837638050317764,
+      "num_tokens": 258914737.0,
+      "step": 2421
+    },
+    {
+      "epoch": 5.517673888255416,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.281797931225705e-06,
+      "loss": 0.6097,
+      "mean_token_accuracy": 0.8738208562135696,
+      "num_tokens": 259021178.0,
+      "step": 2422
+    },
+    {
+      "epoch": 5.519954389965792,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.279921732211609e-06,
+      "loss": 0.5914,
+      "mean_token_accuracy": 0.8749384880065918,
+      "num_tokens": 259127761.0,
+      "step": 2423
+    },
+    {
+      "epoch": 5.5222348916761685,
+      "grad_norm": 2.453125,
+      "learning_rate": 2.278045658109689e-06,
+      "loss": 0.5707,
+      "mean_token_accuracy": 0.8829948306083679,
+      "num_tokens": 259234696.0,
+      "step": 2424
+    },
+    {
+      "epoch": 5.524515393386545,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.2761697099847686e-06,
+      "loss": 0.59,
+      "mean_token_accuracy": 0.8792230039834976,
+      "num_tokens": 259341237.0,
+      "step": 2425
+    },
+    {
+      "epoch": 5.526795895096921,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.274293888901599e-06,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8843846768140793,
+      "num_tokens": 259448732.0,
+      "step": 2426
+    },
+    {
+      "epoch": 5.529076396807298,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.2724181959248627e-06,
+      "loss": 0.5692,
+      "mean_token_accuracy": 0.8816651403903961,
+      "num_tokens": 259555843.0,
+      "step": 2427
+    },
+    {
+      "epoch": 5.531356898517674,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.270542632119163e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8784616589546204,
+      "num_tokens": 259662832.0,
+      "step": 2428
+    },
+    {
+      "epoch": 5.53363740022805,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.2686671985490355e-06,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8840401917695999,
+      "num_tokens": 259770125.0,
+      "step": 2429
+    },
+    {
+      "epoch": 5.535917901938427,
+      "grad_norm": 5.78125,
+      "learning_rate": 2.26679189627894e-06,
+      "loss": 0.5922,
+      "mean_token_accuracy": 0.8798678070306778,
+      "num_tokens": 259876661.0,
+      "step": 2430
+    },
+    {
+      "epoch": 5.538198403648803,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.264916726373263e-06,
+      "loss": 0.5936,
+      "mean_token_accuracy": 0.8790270835161209,
+      "num_tokens": 259982852.0,
+      "step": 2431
+    },
+    {
+      "epoch": 5.540478905359179,
+      "grad_norm": 5.375,
+      "learning_rate": 2.263041689896313e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.881781816482544,
+      "num_tokens": 260089664.0,
+      "step": 2432
+    },
+    {
+      "epoch": 5.542759407069555,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.261166787912325e-06,
+      "loss": 0.5976,
+      "mean_token_accuracy": 0.8784068375825882,
+      "num_tokens": 260196663.0,
+      "step": 2433
+    },
+    {
+      "epoch": 5.545039908779931,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.2592920214854573e-06,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8845723420381546,
+      "num_tokens": 260303381.0,
+      "step": 2434
+    },
+    {
+      "epoch": 5.5473204104903076,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.2574173916797912e-06,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.8818473219871521,
+      "num_tokens": 260410524.0,
+      "step": 2435
+    },
+    {
+      "epoch": 5.549600912200685,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.2555428995593303e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8806977868080139,
+      "num_tokens": 260517537.0,
+      "step": 2436
+    },
+    {
+      "epoch": 5.55188141391106,
+      "grad_norm": 6.15625,
+      "learning_rate": 2.253668546188e-06,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8790431469678879,
+      "num_tokens": 260624932.0,
+      "step": 2437
+    },
+    {
+      "epoch": 5.554161915621437,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.2517943326296487e-06,
+      "loss": 0.5477,
+      "mean_token_accuracy": 0.8867908716201782,
+      "num_tokens": 260732712.0,
+      "step": 2438
+    },
+    {
+      "epoch": 5.556442417331813,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.249920259948041e-06,
+      "loss": 0.5552,
+      "mean_token_accuracy": 0.8851701766252518,
+      "num_tokens": 260840731.0,
+      "step": 2439
+    },
+    {
+      "epoch": 5.558722919042189,
+      "grad_norm": 3.828125,
+      "learning_rate": 2.2480463292068655e-06,
+      "loss": 0.5642,
+      "mean_token_accuracy": 0.8820369690656662,
+      "num_tokens": 260947477.0,
+      "step": 2440
+    },
+    {
+      "epoch": 5.561003420752566,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.24617254146973e-06,
+      "loss": 0.5957,
+      "mean_token_accuracy": 0.879251167178154,
+      "num_tokens": 261054404.0,
+      "step": 2441
+    },
+    {
+      "epoch": 5.563283922462942,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.2442988978001594e-06,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.8821232914924622,
+      "num_tokens": 261162432.0,
+      "step": 2442
+    },
+    {
+      "epoch": 5.565564424173318,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.2424253992615983e-06,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8804102689027786,
+      "num_tokens": 261269127.0,
+      "step": 2443
+    },
+    {
+      "epoch": 5.567844925883694,
+      "grad_norm": 5.65625,
+      "learning_rate": 2.2405520469174084e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8801386058330536,
+      "num_tokens": 261376055.0,
+      "step": 2444
+    },
+    {
+      "epoch": 5.57012542759407,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.238678841830867e-06,
+      "loss": 0.5766,
+      "mean_token_accuracy": 0.8823980838060379,
+      "num_tokens": 261482935.0,
+      "step": 2445
+    },
+    {
+      "epoch": 5.572405929304447,
+      "grad_norm": 5.5625,
+      "learning_rate": 2.23680578506517e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8805437237024307,
+      "num_tokens": 261589878.0,
+      "step": 2446
+    },
+    {
+      "epoch": 5.574686431014824,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.234932877683428e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.878250241279602,
+      "num_tokens": 261696700.0,
+      "step": 2447
+    },
+    {
+      "epoch": 5.5769669327252,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.233060120748667e-06,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8817458301782608,
+      "num_tokens": 261803758.0,
+      "step": 2448
+    },
+    {
+      "epoch": 5.579247434435576,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.2311875153238296e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8814815580844879,
+      "num_tokens": 261910576.0,
+      "step": 2449
+    },
+    {
+      "epoch": 5.581527936145952,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.229315062471768e-06,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8858354687690735,
+      "num_tokens": 262017627.0,
+      "step": 2450
+    },
+    {
+      "epoch": 5.583808437856328,
+      "grad_norm": 5.59375,
+      "learning_rate": 2.2274427632552507e-06,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.8767371773719788,
+      "num_tokens": 262124172.0,
+      "step": 2451
+    },
+    {
+      "epoch": 5.586088939566705,
+      "grad_norm": 5.96875,
+      "learning_rate": 2.2255706187369596e-06,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.8821327090263367,
+      "num_tokens": 262231439.0,
+      "step": 2452
+    },
+    {
+      "epoch": 5.588369441277081,
+      "grad_norm": 5.46875,
+      "learning_rate": 2.223698629979487e-06,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8792487680912018,
+      "num_tokens": 262338613.0,
+      "step": 2453
+    },
+    {
+      "epoch": 5.590649942987457,
+      "grad_norm": 5.65625,
+      "learning_rate": 2.221826798045338e-06,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8853352516889572,
+      "num_tokens": 262445246.0,
+      "step": 2454
+    },
+    {
+      "epoch": 5.592930444697833,
+      "grad_norm": 5.34375,
+      "learning_rate": 2.2199551239969284e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8816657811403275,
+      "num_tokens": 262552338.0,
+      "step": 2455
+    },
+    {
+      "epoch": 5.59521094640821,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.2180836088965833e-06,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8782707005739212,
+      "num_tokens": 262659476.0,
+      "step": 2456
+    },
+    {
+      "epoch": 5.5974914481185865,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.216212253806539e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.88326196372509,
+      "num_tokens": 262766605.0,
+      "step": 2457
+    },
+    {
+      "epoch": 5.599771949828963,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.214341059788941e-06,
+      "loss": 0.5598,
+      "mean_token_accuracy": 0.8842233419418335,
+      "num_tokens": 262873710.0,
+      "step": 2458
+    },
+    {
+      "epoch": 5.602052451539339,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.2124700279058435e-06,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8748954981565475,
+      "num_tokens": 262980715.0,
+      "step": 2459
+    },
+    {
+      "epoch": 5.604332953249715,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.2105991592192063e-06,
+      "loss": 0.5897,
+      "mean_token_accuracy": 0.8786115646362305,
+      "num_tokens": 263087540.0,
+      "step": 2460
+    },
+    {
+      "epoch": 5.606613454960091,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.208728454790899e-06,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8810366690158844,
+      "num_tokens": 263194654.0,
+      "step": 2461
+    },
+    {
+      "epoch": 5.608893956670467,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.2068579156826974e-06,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8800259828567505,
+      "num_tokens": 263302451.0,
+      "step": 2462
+    },
+    {
+      "epoch": 5.611174458380844,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.2049875429562845e-06,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8829780966043472,
+      "num_tokens": 263409487.0,
+      "step": 2463
+    },
+    {
+      "epoch": 5.61345496009122,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.203117337673246e-06,
+      "loss": 0.5514,
+      "mean_token_accuracy": 0.8871753662824631,
+      "num_tokens": 263516373.0,
+      "step": 2464
+    },
+    {
+      "epoch": 5.615735461801596,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.2012473008950756e-06,
+      "loss": 0.5481,
+      "mean_token_accuracy": 0.8859710693359375,
+      "num_tokens": 263624001.0,
+      "step": 2465
+    },
+    {
+      "epoch": 5.618015963511972,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.1993774336831696e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8810239285230637,
+      "num_tokens": 263731549.0,
+      "step": 2466
+    },
+    {
+      "epoch": 5.620296465222349,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.197507737098828e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8845034837722778,
+      "num_tokens": 263838875.0,
+      "step": 2467
+    },
+    {
+      "epoch": 5.6225769669327255,
+      "grad_norm": 4.59375,
+      "learning_rate": 2.195638212203255e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.881766065955162,
+      "num_tokens": 263945489.0,
+      "step": 2468
+    },
+    {
+      "epoch": 5.624857468643102,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.193768860057557e-06,
+      "loss": 0.5533,
+      "mean_token_accuracy": 0.8849270343780518,
+      "num_tokens": 264052839.0,
+      "step": 2469
+    },
+    {
+      "epoch": 5.627137970353478,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.191899681722743e-06,
+      "loss": 0.5856,
+      "mean_token_accuracy": 0.8784494251012802,
+      "num_tokens": 264159607.0,
+      "step": 2470
+    },
+    {
+      "epoch": 5.629418472063854,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.19003067825972e-06,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8801737427711487,
+      "num_tokens": 264266733.0,
+      "step": 2471
+    },
+    {
+      "epoch": 5.63169897377423,
+      "grad_norm": 3.125,
+      "learning_rate": 2.1881618507293004e-06,
+      "loss": 0.5579,
+      "mean_token_accuracy": 0.8845822513103485,
+      "num_tokens": 264374129.0,
+      "step": 2472
+    },
+    {
+      "epoch": 5.633979475484606,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.186293200192194e-06,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8821998536586761,
+      "num_tokens": 264481221.0,
+      "step": 2473
+    },
+    {
+      "epoch": 5.636259977194983,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.1844247277090113e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8810071498155594,
+      "num_tokens": 264588602.0,
+      "step": 2474
+    },
+    {
+      "epoch": 5.638540478905359,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.1825564343402606e-06,
+      "loss": 0.5958,
+      "mean_token_accuracy": 0.8767081648111343,
+      "num_tokens": 264695289.0,
+      "step": 2475
+    },
+    {
+      "epoch": 5.640820980615736,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.180688321146349e-06,
+      "loss": 0.5983,
+      "mean_token_accuracy": 0.8806226551532745,
+      "num_tokens": 264801458.0,
+      "step": 2476
+    },
+    {
+      "epoch": 5.643101482326112,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.1788203891875818e-06,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8851145952939987,
+      "num_tokens": 264908510.0,
+      "step": 2477
+    },
+    {
+      "epoch": 5.645381984036488,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.176952639524161e-06,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8833800852298737,
+      "num_tokens": 265015635.0,
+      "step": 2478
+    },
+    {
+      "epoch": 5.6476624857468645,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.175085073216185e-06,
+      "loss": 0.5733,
+      "mean_token_accuracy": 0.8817281723022461,
+      "num_tokens": 265122707.0,
+      "step": 2479
+    },
+    {
+      "epoch": 5.649942987457241,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.173217691323649e-06,
+      "loss": 0.5607,
+      "mean_token_accuracy": 0.8875688314437866,
+      "num_tokens": 265229807.0,
+      "step": 2480
+    },
+    {
+      "epoch": 5.652223489167617,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.1713504949064433e-06,
+      "loss": 0.5366,
+      "mean_token_accuracy": 0.8914333879947662,
+      "num_tokens": 265337470.0,
+      "step": 2481
+    },
+    {
+      "epoch": 5.654503990877993,
+      "grad_norm": 3.125,
+      "learning_rate": 2.169483485024351e-06,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.8841661512851715,
+      "num_tokens": 265444298.0,
+      "step": 2482
+    },
+    {
+      "epoch": 5.656784492588369,
+      "grad_norm": 3.125,
+      "learning_rate": 2.167616662737052e-06,
+      "loss": 0.5528,
+      "mean_token_accuracy": 0.8871481865644455,
+      "num_tokens": 265551726.0,
+      "step": 2483
+    },
+    {
+      "epoch": 5.659064994298745,
+      "grad_norm": 4.0,
+      "learning_rate": 2.1657500291041185e-06,
+      "loss": 0.5541,
+      "mean_token_accuracy": 0.8866736739873886,
+      "num_tokens": 265659350.0,
+      "step": 2484
+    },
+    {
+      "epoch": 5.661345496009122,
+      "grad_norm": 3.75,
+      "learning_rate": 2.1638835851850155e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8781305104494095,
+      "num_tokens": 265766187.0,
+      "step": 2485
+    },
+    {
+      "epoch": 5.663625997719498,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.1620173320391007e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8814955800771713,
+      "num_tokens": 265873426.0,
+      "step": 2486
+    },
+    {
+      "epoch": 5.665906499429875,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.160151270725623e-06,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8812149912118912,
+      "num_tokens": 265980661.0,
+      "step": 2487
+    },
+    {
+      "epoch": 5.668187001140251,
+      "grad_norm": 5.96875,
+      "learning_rate": 2.158285402303723e-06,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8758508861064911,
+      "num_tokens": 266087907.0,
+      "step": 2488
+    },
+    {
+      "epoch": 5.670467502850627,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.1564197278324317e-06,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8836510181427002,
+      "num_tokens": 266195252.0,
+      "step": 2489
+    },
+    {
+      "epoch": 5.6727480045610035,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.1545542483706694e-06,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8824823647737503,
+      "num_tokens": 266302369.0,
+      "step": 2490
+    },
+    {
+      "epoch": 5.67502850627138,
+      "grad_norm": 2.625,
+      "learning_rate": 2.1526889649772477e-06,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.885822594165802,
+      "num_tokens": 266409450.0,
+      "step": 2491
+    },
+    {
+      "epoch": 5.677309007981756,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.1508238787108633e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8812492936849594,
+      "num_tokens": 266516611.0,
+      "step": 2492
+    },
+    {
+      "epoch": 5.679589509692132,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.1489589906301046e-06,
+      "loss": 0.5951,
+      "mean_token_accuracy": 0.8759992122650146,
+      "num_tokens": 266623332.0,
+      "step": 2493
+    },
+    {
+      "epoch": 5.681870011402508,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.1470943017934455e-06,
+      "loss": 0.5674,
+      "mean_token_accuracy": 0.8840955346822739,
+      "num_tokens": 266730787.0,
+      "step": 2494
+    },
+    {
+      "epoch": 5.684150513112884,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.145229813259248e-06,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8826257884502411,
+      "num_tokens": 266838139.0,
+      "step": 2495
+    },
+    {
+      "epoch": 5.6864310148232615,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.143365526085759e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.88275146484375,
+      "num_tokens": 266944920.0,
+      "step": 2496
+    },
+    {
+      "epoch": 5.688711516533638,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.1415014413311126e-06,
+      "loss": 0.5877,
+      "mean_token_accuracy": 0.880971685051918,
+      "num_tokens": 267051841.0,
+      "step": 2497
+    },
+    {
+      "epoch": 5.690992018244014,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.139637560053327e-06,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8808932155370712,
+      "num_tokens": 267159338.0,
+      "step": 2498
+    },
+    {
+      "epoch": 5.69327251995439,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.137773883310305e-06,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8823635280132294,
+      "num_tokens": 267266461.0,
+      "step": 2499
+    },
+    {
+      "epoch": 5.695553021664766,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.1359104121598337e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.881933718919754,
+      "num_tokens": 267373196.0,
+      "step": 2500
+    },
+    {
+      "epoch": 5.6978335233751425,
+      "grad_norm": 4.0,
+      "learning_rate": 2.1340471476595836e-06,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.8774979561567307,
+      "num_tokens": 267480795.0,
+      "step": 2501
+    },
+    {
+      "epoch": 5.700114025085519,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.1321840908671082e-06,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.8825899064540863,
+      "num_tokens": 267587894.0,
+      "step": 2502
+    },
+    {
+      "epoch": 5.702394526795895,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.1303212428398407e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8776924163103104,
+      "num_tokens": 267694535.0,
+      "step": 2503
+    },
+    {
+      "epoch": 5.704675028506271,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.1284586046350996e-06,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8794005513191223,
+      "num_tokens": 267801843.0,
+      "step": 2504
+    },
+    {
+      "epoch": 5.706955530216648,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.126596177310081e-06,
+      "loss": 0.5546,
+      "mean_token_accuracy": 0.8848482072353363,
+      "num_tokens": 267909136.0,
+      "step": 2505
+    },
+    {
+      "epoch": 5.7092360319270234,
+      "grad_norm": 3.125,
+      "learning_rate": 2.124733961921864e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8841657936573029,
+      "num_tokens": 268016448.0,
+      "step": 2506
+    },
+    {
+      "epoch": 5.7115165336374005,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.1228719595274056e-06,
+      "loss": 0.6123,
+      "mean_token_accuracy": 0.8750414401292801,
+      "num_tokens": 268123331.0,
+      "step": 2507
+    },
+    {
+      "epoch": 5.713797035347777,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.1210101711835413e-06,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8833545595407486,
+      "num_tokens": 268230339.0,
+      "step": 2508
+    },
+    {
+      "epoch": 5.716077537058153,
+      "grad_norm": 5.53125,
+      "learning_rate": 2.1191485979469877e-06,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.880239725112915,
+      "num_tokens": 268337047.0,
+      "step": 2509
+    },
+    {
+      "epoch": 5.718358038768529,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.1172872408743374e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8804043382406235,
+      "num_tokens": 268444006.0,
+      "step": 2510
+    },
+    {
+      "epoch": 5.720638540478905,
+      "grad_norm": 4.5625,
+      "learning_rate": 2.11542610102206e-06,
+      "loss": 0.6014,
+      "mean_token_accuracy": 0.8754177987575531,
+      "num_tokens": 268550836.0,
+      "step": 2511
+    },
+    {
+      "epoch": 5.7229190421892815,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.1135651794465032e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8840764611959457,
+      "num_tokens": 268658140.0,
+      "step": 2512
+    },
+    {
+      "epoch": 5.725199543899658,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.1117044772038915e-06,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8794988542795181,
+      "num_tokens": 268764953.0,
+      "step": 2513
+    },
+    {
+      "epoch": 5.727480045610034,
+      "grad_norm": 4.40625,
+      "learning_rate": 2.1098439953503207e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8798924386501312,
+      "num_tokens": 268872046.0,
+      "step": 2514
+    },
+    {
+      "epoch": 5.72976054732041,
+      "grad_norm": 5.96875,
+      "learning_rate": 2.1079837349417664e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8788573145866394,
+      "num_tokens": 268978631.0,
+      "step": 2515
+    },
+    {
+      "epoch": 5.732041049030787,
+      "grad_norm": 4.375,
+      "learning_rate": 2.1061236970340756e-06,
+      "loss": 0.5553,
+      "mean_token_accuracy": 0.8862823694944382,
+      "num_tokens": 269085803.0,
+      "step": 2516
+    },
+    {
+      "epoch": 5.734321550741163,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.104263882682971e-06,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8813868910074234,
+      "num_tokens": 269192710.0,
+      "step": 2517
+    },
+    {
+      "epoch": 5.7366020524515395,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.1024042929440465e-06,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8835171908140182,
+      "num_tokens": 269299690.0,
+      "step": 2518
+    },
+    {
+      "epoch": 5.738882554161916,
+      "grad_norm": 3.0,
+      "learning_rate": 2.1005449288727696e-06,
+      "loss": 0.5593,
+      "mean_token_accuracy": 0.8817470818758011,
+      "num_tokens": 269406982.0,
+      "step": 2519
+    },
+    {
+      "epoch": 5.741163055872292,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.0986857915244787e-06,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8834887593984604,
+      "num_tokens": 269514394.0,
+      "step": 2520
+    },
+    {
+      "epoch": 5.743443557582668,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.096826881954385e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8815822005271912,
+      "num_tokens": 269621655.0,
+      "step": 2521
+    },
+    {
+      "epoch": 5.745724059293044,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.0949682012175693e-06,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8810020089149475,
+      "num_tokens": 269729019.0,
+      "step": 2522
+    },
+    {
+      "epoch": 5.7480045610034205,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.093109750368983e-06,
+      "loss": 0.5615,
+      "mean_token_accuracy": 0.8838683366775513,
+      "num_tokens": 269836494.0,
+      "step": 2523
+    },
+    {
+      "epoch": 5.750285062713797,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.0912515304634485e-06,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8854202926158905,
+      "num_tokens": 269943579.0,
+      "step": 2524
+    },
+    {
+      "epoch": 5.752565564424174,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.089393542555653e-06,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8842357248067856,
+      "num_tokens": 270050301.0,
+      "step": 2525
+    },
+    {
+      "epoch": 5.75484606613455,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.0875357877001556e-06,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8842559903860092,
+      "num_tokens": 270157064.0,
+      "step": 2526
+    },
+    {
+      "epoch": 5.757126567844926,
+      "grad_norm": 4.625,
+      "learning_rate": 2.085678266951382e-06,
+      "loss": 0.5951,
+      "mean_token_accuracy": 0.8778895437717438,
+      "num_tokens": 270263910.0,
+      "step": 2527
+    },
+    {
+      "epoch": 5.759407069555302,
+      "grad_norm": 4.5,
+      "learning_rate": 2.083820981363626e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8812290579080582,
+      "num_tokens": 270371067.0,
+      "step": 2528
+    },
+    {
+      "epoch": 5.7616875712656785,
+      "grad_norm": 2.5,
+      "learning_rate": 2.0819639319910466e-06,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.8800959140062332,
+      "num_tokens": 270478072.0,
+      "step": 2529
+    },
+    {
+      "epoch": 5.763968072976055,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.0801071198876684e-06,
+      "loss": 0.5703,
+      "mean_token_accuracy": 0.8809636831283569,
+      "num_tokens": 270584725.0,
+      "step": 2530
+    },
+    {
+      "epoch": 5.766248574686431,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.0782505461073822e-06,
+      "loss": 0.5835,
+      "mean_token_accuracy": 0.8816069960594177,
+      "num_tokens": 270691204.0,
+      "step": 2531
+    },
+    {
+      "epoch": 5.768529076396807,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.076394211703944e-06,
+      "loss": 0.5505,
+      "mean_token_accuracy": 0.8873765766620636,
+      "num_tokens": 270798291.0,
+      "step": 2532
+    },
+    {
+      "epoch": 5.770809578107183,
+      "grad_norm": 4.5625,
+      "learning_rate": 2.0745381177309732e-06,
+      "loss": 0.5627,
+      "mean_token_accuracy": 0.8861078023910522,
+      "num_tokens": 270905483.0,
+      "step": 2533
+    },
+    {
+      "epoch": 5.7730900798175595,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.072682265241954e-06,
+      "loss": 0.5536,
+      "mean_token_accuracy": 0.8886701166629791,
+      "num_tokens": 271012509.0,
+      "step": 2534
+    },
+    {
+      "epoch": 5.775370581527936,
+      "grad_norm": 5.21875,
+      "learning_rate": 2.0708266552902303e-06,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8834939152002335,
+      "num_tokens": 271119387.0,
+      "step": 2535
+    },
+    {
+      "epoch": 5.777651083238313,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.0689712889290114e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8821908384561539,
+      "num_tokens": 271226607.0,
+      "step": 2536
+    },
+    {
+      "epoch": 5.779931584948689,
+      "grad_norm": 3.25,
+      "learning_rate": 2.0671161672113677e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8805604577064514,
+      "num_tokens": 271333818.0,
+      "step": 2537
+    },
+    {
+      "epoch": 5.782212086659065,
+      "grad_norm": 2.375,
+      "learning_rate": 2.06526129119023e-06,
+      "loss": 0.5637,
+      "mean_token_accuracy": 0.8876101672649384,
+      "num_tokens": 271440910.0,
+      "step": 2538
+    },
+    {
+      "epoch": 5.784492588369441,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.063406661918391e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8823280781507492,
+      "num_tokens": 271548224.0,
+      "step": 2539
+    },
+    {
+      "epoch": 5.7867730900798175,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.0615522804485027e-06,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8831372708082199,
+      "num_tokens": 271655042.0,
+      "step": 2540
+    },
+    {
+      "epoch": 5.789053591790194,
+      "grad_norm": 2.5,
+      "learning_rate": 2.059698147833075e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8820475935935974,
+      "num_tokens": 271762363.0,
+      "step": 2541
+    },
+    {
+      "epoch": 5.79133409350057,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.0578442651244774e-06,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8825207501649857,
+      "num_tokens": 271869380.0,
+      "step": 2542
+    },
+    {
+      "epoch": 5.793614595210946,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.0559906333749392e-06,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8834405839443207,
+      "num_tokens": 271976735.0,
+      "step": 2543
+    },
+    {
+      "epoch": 5.795895096921322,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.054137253636545e-06,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8806865364313126,
+      "num_tokens": 272084686.0,
+      "step": 2544
+    },
+    {
+      "epoch": 5.798175598631699,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.0522841269612397e-06,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.8805492520332336,
+      "num_tokens": 272192178.0,
+      "step": 2545
+    },
+    {
+      "epoch": 5.800456100342076,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.0504312544008193e-06,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.883878156542778,
+      "num_tokens": 272298806.0,
+      "step": 2546
+    },
+    {
+      "epoch": 5.802736602052452,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.048578637006939e-06,
+      "loss": 0.5506,
+      "mean_token_accuracy": 0.8853472024202347,
+      "num_tokens": 272405957.0,
+      "step": 2547
+    },
+    {
+      "epoch": 5.805017103762828,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.04672627583111e-06,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8789206445217133,
+      "num_tokens": 272513063.0,
+      "step": 2548
+    },
+    {
+      "epoch": 5.807297605473204,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.0448741719246962e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.883039727807045,
+      "num_tokens": 272619770.0,
+      "step": 2549
+    },
+    {
+      "epoch": 5.80957810718358,
+      "grad_norm": 2.75,
+      "learning_rate": 2.043022326338916e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8768695294857025,
+      "num_tokens": 272726957.0,
+      "step": 2550
+    },
+    {
+      "epoch": 5.811858608893957,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.0411707401248406e-06,
+      "loss": 0.5602,
+      "mean_token_accuracy": 0.8837281614542007,
+      "num_tokens": 272834440.0,
+      "step": 2551
+    },
+    {
+      "epoch": 5.814139110604333,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.0393194143333956e-06,
+      "loss": 0.5583,
+      "mean_token_accuracy": 0.8858349472284317,
+      "num_tokens": 272941653.0,
+      "step": 2552
+    },
+    {
+      "epoch": 5.816419612314709,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.0374683500153564e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.884008064866066,
+      "num_tokens": 273048925.0,
+      "step": 2553
+    },
+    {
+      "epoch": 5.818700114025085,
+      "grad_norm": 5.0625,
+      "learning_rate": 2.0356175482213523e-06,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8851135820150375,
+      "num_tokens": 273155880.0,
+      "step": 2554
+    },
+    {
+      "epoch": 5.820980615735461,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.033767010001863e-06,
+      "loss": 0.5922,
+      "mean_token_accuracy": 0.8770723342895508,
+      "num_tokens": 273262101.0,
+      "step": 2555
+    },
+    {
+      "epoch": 5.823261117445838,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.0319167364072184e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8830315619707108,
+      "num_tokens": 273369542.0,
+      "step": 2556
+    },
+    {
+      "epoch": 5.825541619156215,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.0300667284875965e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8794274926185608,
+      "num_tokens": 273477240.0,
+      "step": 2557
+    },
+    {
+      "epoch": 5.827822120866591,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.0282169872930275e-06,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8839514553546906,
+      "num_tokens": 273584941.0,
+      "step": 2558
+    },
+    {
+      "epoch": 5.830102622576967,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.026367513873388e-06,
+      "loss": 0.5544,
+      "mean_token_accuracy": 0.8844399601221085,
+      "num_tokens": 273692501.0,
+      "step": 2559
+    },
+    {
+      "epoch": 5.832383124287343,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.0245183092784046e-06,
+      "loss": 0.5856,
+      "mean_token_accuracy": 0.8805771768093109,
+      "num_tokens": 273799051.0,
+      "step": 2560
+    },
+    {
+      "epoch": 5.834663625997719,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.0226693745576494e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8822827190160751,
+      "num_tokens": 273906681.0,
+      "step": 2561
+    },
+    {
+      "epoch": 5.836944127708096,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.020820710760541e-06,
+      "loss": 0.5538,
+      "mean_token_accuracy": 0.8871154636144638,
+      "num_tokens": 274014464.0,
+      "step": 2562
+    },
+    {
+      "epoch": 5.839224629418472,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.018972318936347e-06,
+      "loss": 0.5931,
+      "mean_token_accuracy": 0.8768787384033203,
+      "num_tokens": 274121211.0,
+      "step": 2563
+    },
+    {
+      "epoch": 5.841505131128848,
+      "grad_norm": 3.25,
+      "learning_rate": 2.017124200134178e-06,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8847302943468094,
+      "num_tokens": 274228462.0,
+      "step": 2564
+    },
+    {
+      "epoch": 5.843785632839225,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.01527635540299e-06,
+      "loss": 0.5809,
+      "mean_token_accuracy": 0.8782871812582016,
+      "num_tokens": 274335776.0,
+      "step": 2565
+    },
+    {
+      "epoch": 5.846066134549601,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.0134287857915864e-06,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.8803008496761322,
+      "num_tokens": 274443604.0,
+      "step": 2566
+    },
+    {
+      "epoch": 5.848346636259977,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.0115814923486093e-06,
+      "loss": 0.557,
+      "mean_token_accuracy": 0.8840547949075699,
+      "num_tokens": 274550591.0,
+      "step": 2567
+    },
+    {
+      "epoch": 5.850627137970354,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.009734476122547e-06,
+      "loss": 0.559,
+      "mean_token_accuracy": 0.8849809020757675,
+      "num_tokens": 274657799.0,
+      "step": 2568
+    },
+    {
+      "epoch": 5.85290763968073,
+      "grad_norm": 2.875,
+      "learning_rate": 2.007887738161732e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8857974410057068,
+      "num_tokens": 274765249.0,
+      "step": 2569
+    },
+    {
+      "epoch": 5.855188141391106,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.006041279514336e-06,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8802428096532822,
+      "num_tokens": 274871966.0,
+      "step": 2570
+    },
+    {
+      "epoch": 5.857468643101482,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.004195101228374e-06,
+      "loss": 0.5453,
+      "mean_token_accuracy": 0.8866171091794968,
+      "num_tokens": 274979490.0,
+      "step": 2571
+    },
+    {
+      "epoch": 5.859749144811858,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.002349204351701e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8811267167329788,
+      "num_tokens": 275086036.0,
+      "step": 2572
+    },
+    {
+      "epoch": 5.862029646522235,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.0005035899320115e-06,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8833982795476913,
+      "num_tokens": 275193239.0,
+      "step": 2573
+    },
+    {
+      "epoch": 5.864310148232612,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.998658259016841e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8820087909698486,
+      "num_tokens": 275300065.0,
+      "step": 2574
+    },
+    {
+      "epoch": 5.866590649942988,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.996813212653564e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.8738995492458344,
+      "num_tokens": 275407231.0,
+      "step": 2575
+    },
+    {
+      "epoch": 5.868871151653364,
+      "grad_norm": 3.25,
+      "learning_rate": 1.9949684518893926e-06,
+      "loss": 0.6003,
+      "mean_token_accuracy": 0.8753588199615479,
+      "num_tokens": 275514097.0,
+      "step": 2576
+    },
+    {
+      "epoch": 5.87115165336374,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.9931239777713794e-06,
+      "loss": 0.5453,
+      "mean_token_accuracy": 0.8908671289682388,
+      "num_tokens": 275621304.0,
+      "step": 2577
+    },
+    {
+      "epoch": 5.873432155074116,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.9912797913464098e-06,
+      "loss": 0.5424,
+      "mean_token_accuracy": 0.8852016925811768,
+      "num_tokens": 275728465.0,
+      "step": 2578
+    },
+    {
+      "epoch": 5.875712656784493,
+      "grad_norm": 3.125,
+      "learning_rate": 1.989435893661209e-06,
+      "loss": 0.5527,
+      "mean_token_accuracy": 0.8849275410175323,
+      "num_tokens": 275835525.0,
+      "step": 2579
+    },
+    {
+      "epoch": 5.877993158494869,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.9875922857623387e-06,
+      "loss": 0.5584,
+      "mean_token_accuracy": 0.8849629461765289,
+      "num_tokens": 275942316.0,
+      "step": 2580
+    },
+    {
+      "epoch": 5.880273660205245,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.985748968696194e-06,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.8805515915155411,
+      "num_tokens": 276048810.0,
+      "step": 2581
+    },
+    {
+      "epoch": 5.882554161915621,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.9839059435090073e-06,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8853743970394135,
+      "num_tokens": 276155430.0,
+      "step": 2582
+    },
+    {
+      "epoch": 5.884834663625997,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.9820632112468437e-06,
+      "loss": 0.5613,
+      "mean_token_accuracy": 0.8853378593921661,
+      "num_tokens": 276262832.0,
+      "step": 2583
+    },
+    {
+      "epoch": 5.887115165336374,
+      "grad_norm": 3.625,
+      "learning_rate": 1.9802207729556023e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.876113772392273,
+      "num_tokens": 276369762.0,
+      "step": 2584
+    },
+    {
+      "epoch": 5.889395667046751,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.9783786296810148e-06,
+      "loss": 0.5496,
+      "mean_token_accuracy": 0.8896660208702087,
+      "num_tokens": 276477298.0,
+      "step": 2585
+    },
+    {
+      "epoch": 5.891676168757127,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.9765367824686467e-06,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8809142112731934,
+      "num_tokens": 276583840.0,
+      "step": 2586
+    },
+    {
+      "epoch": 5.893956670467503,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.974695232363895e-06,
+      "loss": 0.5606,
+      "mean_token_accuracy": 0.8866210728883743,
+      "num_tokens": 276690816.0,
+      "step": 2587
+    },
+    {
+      "epoch": 5.896237172177879,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.9728539804119893e-06,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8767704367637634,
+      "num_tokens": 276797914.0,
+      "step": 2588
+    },
+    {
+      "epoch": 5.898517673888255,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.9710130276579864e-06,
+      "loss": 0.586,
+      "mean_token_accuracy": 0.8765504211187363,
+      "num_tokens": 276904386.0,
+      "step": 2589
+    },
+    {
+      "epoch": 5.900798175598632,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.969172375146776e-06,
+      "loss": 0.5591,
+      "mean_token_accuracy": 0.8872781246900558,
+      "num_tokens": 277011618.0,
+      "step": 2590
+    },
+    {
+      "epoch": 5.903078677309008,
+      "grad_norm": 2.875,
+      "learning_rate": 1.9673320239230783e-06,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8807976096868515,
+      "num_tokens": 277118651.0,
+      "step": 2591
+    },
+    {
+      "epoch": 5.905359179019384,
+      "grad_norm": 2.5,
+      "learning_rate": 1.9654919750314396e-06,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.8855329602956772,
+      "num_tokens": 277225439.0,
+      "step": 2592
+    },
+    {
+      "epoch": 5.90763968072976,
+      "grad_norm": 3.75,
+      "learning_rate": 1.9636522295162375e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8821413516998291,
+      "num_tokens": 277332844.0,
+      "step": 2593
+    },
+    {
+      "epoch": 5.909920182440137,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.9618127884216753e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8815776854753494,
+      "num_tokens": 277439892.0,
+      "step": 2594
+    },
+    {
+      "epoch": 5.9122006841505135,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.959973652791784e-06,
+      "loss": 0.5554,
+      "mean_token_accuracy": 0.885757714509964,
+      "num_tokens": 277547646.0,
+      "step": 2595
+    },
+    {
+      "epoch": 5.91448118586089,
+      "grad_norm": 2.625,
+      "learning_rate": 1.9581348236704217e-06,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8831838816404343,
+      "num_tokens": 277654799.0,
+      "step": 2596
+    },
+    {
+      "epoch": 5.916761687571266,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.9562963021012723e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8817932456731796,
+      "num_tokens": 277761954.0,
+      "step": 2597
+    },
+    {
+      "epoch": 5.919042189281642,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.954458089127845e-06,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8825689405202866,
+      "num_tokens": 277869215.0,
+      "step": 2598
+    },
+    {
+      "epoch": 5.921322690992018,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.952620185793475e-06,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8813809603452682,
+      "num_tokens": 277975771.0,
+      "step": 2599
+    },
+    {
+      "epoch": 5.923603192702394,
+      "grad_norm": 4.46875,
+      "learning_rate": 1.9507825931413193e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8809202462434769,
+      "num_tokens": 278082716.0,
+      "step": 2600
+    },
+    {
+      "epoch": 5.925883694412771,
+      "grad_norm": 2.5,
+      "learning_rate": 1.9489453122143605e-06,
+      "loss": 0.5458,
+      "mean_token_accuracy": 0.8862514644861221,
+      "num_tokens": 278189379.0,
+      "step": 2601
+    },
+    {
+      "epoch": 5.928164196123147,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.947108344055404e-06,
+      "loss": 0.5578,
+      "mean_token_accuracy": 0.8841505497694016,
+      "num_tokens": 278296342.0,
+      "step": 2602
+    },
+    {
+      "epoch": 5.930444697833523,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.9452716897070785e-06,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8791919499635696,
+      "num_tokens": 278403387.0,
+      "step": 2603
+    },
+    {
+      "epoch": 5.932725199543899,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.943435350211832e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8786250501871109,
+      "num_tokens": 278510323.0,
+      "step": 2604
+    },
+    {
+      "epoch": 5.935005701254276,
+      "grad_norm": 6.0,
+      "learning_rate": 1.941599326611935e-06,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8808638751506805,
+      "num_tokens": 278616886.0,
+      "step": 2605
+    },
+    {
+      "epoch": 5.9372862029646525,
+      "grad_norm": 3.0,
+      "learning_rate": 1.939763619949481e-06,
+      "loss": 0.5821,
+      "mean_token_accuracy": 0.880496084690094,
+      "num_tokens": 278724068.0,
+      "step": 2606
+    },
+    {
+      "epoch": 5.939566704675029,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.9379282312663797e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8817182034254074,
+      "num_tokens": 278830958.0,
+      "step": 2607
+    },
+    {
+      "epoch": 5.941847206385405,
+      "grad_norm": 3.125,
+      "learning_rate": 1.936093161604363e-06,
+      "loss": 0.536,
+      "mean_token_accuracy": 0.8876519352197647,
+      "num_tokens": 278938000.0,
+      "step": 2608
+    },
+    {
+      "epoch": 5.944127708095781,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.9342584120049824e-06,
+      "loss": 0.6064,
+      "mean_token_accuracy": 0.876665860414505,
+      "num_tokens": 279044994.0,
+      "step": 2609
+    },
+    {
+      "epoch": 5.946408209806157,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.9324239835096044e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8785818070173264,
+      "num_tokens": 279152387.0,
+      "step": 2610
+    },
+    {
+      "epoch": 5.9486887115165334,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.930589877159415e-06,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8864869177341461,
+      "num_tokens": 279259832.0,
+      "step": 2611
+    },
+    {
+      "epoch": 5.95096921322691,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.928756093995419e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8803494274616241,
+      "num_tokens": 279366863.0,
+      "step": 2612
+    },
+    {
+      "epoch": 5.953249714937286,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.9269226350584357e-06,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8808418363332748,
+      "num_tokens": 279473717.0,
+      "step": 2613
+    },
+    {
+      "epoch": 5.955530216647663,
+      "grad_norm": 3.375,
+      "learning_rate": 1.9250895013891015e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8824566006660461,
+      "num_tokens": 279580810.0,
+      "step": 2614
+    },
+    {
+      "epoch": 5.957810718358039,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.9232566940278675e-06,
+      "loss": 0.5545,
+      "mean_token_accuracy": 0.8856438547372818,
+      "num_tokens": 279687888.0,
+      "step": 2615
+    },
+    {
+      "epoch": 5.960091220068415,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.9214242140149987e-06,
+      "loss": 0.5858,
+      "mean_token_accuracy": 0.8792087882757187,
+      "num_tokens": 279794989.0,
+      "step": 2616
+    },
+    {
+      "epoch": 5.9623717217787915,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.9195920623905766e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8819312304258347,
+      "num_tokens": 279901750.0,
+      "step": 2617
+    },
+    {
+      "epoch": 5.964652223489168,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.9177602401944943e-06,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8836159706115723,
+      "num_tokens": 280008000.0,
+      "step": 2618
+    },
+    {
+      "epoch": 5.966932725199544,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.915928748466459e-06,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.8819707930088043,
+      "num_tokens": 280115535.0,
+      "step": 2619
+    },
+    {
+      "epoch": 5.96921322690992,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.9140975882459912e-06,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8795643150806427,
+      "num_tokens": 280222510.0,
+      "step": 2620
+    },
+    {
+      "epoch": 5.971493728620296,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.9122667605724202e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8824938386678696,
+      "num_tokens": 280329229.0,
+      "step": 2621
+    },
+    {
+      "epoch": 5.9737742303306725,
+      "grad_norm": 4.375,
+      "learning_rate": 1.910436266484889e-06,
+      "loss": 0.5898,
+      "mean_token_accuracy": 0.8779114931821823,
+      "num_tokens": 280436155.0,
+      "step": 2622
+    },
+    {
+      "epoch": 5.976054732041049,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.908606107022351e-06,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.8845522552728653,
+      "num_tokens": 280543656.0,
+      "step": 2623
+    },
+    {
+      "epoch": 5.978335233751425,
+      "grad_norm": 3.625,
+      "learning_rate": 1.9067762832235698e-06,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8839434236288071,
+      "num_tokens": 280651076.0,
+      "step": 2624
+    },
+    {
+      "epoch": 5.980615735461802,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.9049467961271184e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.881288930773735,
+      "num_tokens": 280758080.0,
+      "step": 2625
+    },
+    {
+      "epoch": 5.982896237172178,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.9031176467713763e-06,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8865546584129333,
+      "num_tokens": 280865575.0,
+      "step": 2626
+    },
+    {
+      "epoch": 5.985176738882554,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.9012888361945354e-06,
+      "loss": 0.5608,
+      "mean_token_accuracy": 0.8823821991682053,
+      "num_tokens": 280972852.0,
+      "step": 2627
+    },
+    {
+      "epoch": 5.9874572405929305,
+      "grad_norm": 4.90625,
+      "learning_rate": 1.8994603654345917e-06,
+      "loss": 0.5856,
+      "mean_token_accuracy": 0.8786684423685074,
+      "num_tokens": 281079620.0,
+      "step": 2628
+    },
+    {
+      "epoch": 5.989737742303307,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.897632235529351e-06,
+      "loss": 0.5979,
+      "mean_token_accuracy": 0.8792405128479004,
+      "num_tokens": 281186453.0,
+      "step": 2629
+    },
+    {
+      "epoch": 5.992018244013683,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.8958044475164242e-06,
+      "loss": 0.5692,
+      "mean_token_accuracy": 0.8830309361219406,
+      "num_tokens": 281293819.0,
+      "step": 2630
+    },
+    {
+      "epoch": 5.994298745724059,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.8939770024332294e-06,
+      "loss": 0.5425,
+      "mean_token_accuracy": 0.8879213184118271,
+      "num_tokens": 281400651.0,
+      "step": 2631
+    },
+    {
+      "epoch": 5.996579247434435,
+      "grad_norm": 2.875,
+      "learning_rate": 1.8921499013169876e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8807700723409653,
+      "num_tokens": 281507623.0,
+      "step": 2632
+    },
+    {
+      "epoch": 5.9988597491448115,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.8903231452047265e-06,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8816178441047668,
+      "num_tokens": 281614487.0,
+      "step": 2633
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.8884967351332778e-06,
+      "loss": 0.5877,
+      "mean_token_accuracy": 0.8762392401695251,
+      "num_tokens": 281653392.0,
+      "step": 2634
+    },
+    {
+      "epoch": 6.002280501710376,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.886670672139277e-06,
+      "loss": 0.5603,
+      "mean_token_accuracy": 0.8846362233161926,
+      "num_tokens": 281760658.0,
+      "step": 2635
+    },
+    {
+      "epoch": 6.004561003420752,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.884844957259163e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8855700343847275,
+      "num_tokens": 281867681.0,
+      "step": 2636
+    },
+    {
+      "epoch": 6.006841505131129,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.8830195915291741e-06,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8807573318481445,
+      "num_tokens": 281974677.0,
+      "step": 2637
+    },
+    {
+      "epoch": 6.009122006841505,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.8811945759853543e-06,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8797610551118851,
+      "num_tokens": 282081659.0,
+      "step": 2638
+    },
+    {
+      "epoch": 6.011402508551882,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.879369911663546e-06,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.881444662809372,
+      "num_tokens": 282188705.0,
+      "step": 2639
+    },
+    {
+      "epoch": 6.013683010262258,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.8775455995993941e-06,
+      "loss": 0.5571,
+      "mean_token_accuracy": 0.8842919617891312,
+      "num_tokens": 282295789.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.013683010262258,
+      "eval_loss": 0.5869334936141968,
+      "eval_mean_token_accuracy": 0.8798886684863739,
+      "eval_num_tokens": 282295789.0,
+      "eval_runtime": 58.6353,
+      "eval_samples_per_second": 143.003,
+      "eval_steps_per_second": 4.485,
+      "step": 2640
+    },
+    {
+      "epoch": 6.015963511972634,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.875721640828344e-06,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8848172128200531,
+      "num_tokens": 282402500.0,
+      "step": 2641
+    },
+    {
+      "epoch": 6.01824401368301,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.8738980363856376e-06,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.8776390701532364,
+      "num_tokens": 282509194.0,
+      "step": 2642
+    },
+    {
+      "epoch": 6.020524515393387,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.8720747873063184e-06,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8839791864156723,
+      "num_tokens": 282616314.0,
+      "step": 2643
+    },
+    {
+      "epoch": 6.022805017103763,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.870251894625227e-06,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8833863288164139,
+      "num_tokens": 282723232.0,
+      "step": 2644
+    },
+    {
+      "epoch": 6.025085518814139,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.8684293593770026e-06,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8824484199285507,
+      "num_tokens": 282829758.0,
+      "step": 2645
+    },
+    {
+      "epoch": 6.027366020524515,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.866607182596081e-06,
+      "loss": 0.5839,
+      "mean_token_accuracy": 0.8822183609008789,
+      "num_tokens": 282936677.0,
+      "step": 2646
+    },
+    {
+      "epoch": 6.029646522234891,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.8647853653166953e-06,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8796509057283401,
+      "num_tokens": 283043715.0,
+      "step": 2647
+    },
+    {
+      "epoch": 6.031927023945268,
+      "grad_norm": 3.0,
+      "learning_rate": 1.862963908572872e-06,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8782380074262619,
+      "num_tokens": 283150407.0,
+      "step": 2648
+    },
+    {
+      "epoch": 6.034207525655645,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.8611428133984365e-06,
+      "loss": 0.5848,
+      "mean_token_accuracy": 0.8795290291309357,
+      "num_tokens": 283257249.0,
+      "step": 2649
+    },
+    {
+      "epoch": 6.036488027366021,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.8593220808270057e-06,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.881279394030571,
+      "num_tokens": 283363971.0,
+      "step": 2650
+    },
+    {
+      "epoch": 6.038768529076397,
+      "grad_norm": 4.78125,
+      "learning_rate": 1.857501711891993e-06,
+      "loss": 0.552,
+      "mean_token_accuracy": 0.8844835609197617,
+      "num_tokens": 283471402.0,
+      "step": 2651
+    },
+    {
+      "epoch": 6.041049030786773,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.8556817076266059e-06,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.8844869136810303,
+      "num_tokens": 283578013.0,
+      "step": 2652
+    },
+    {
+      "epoch": 6.043329532497149,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.8538620690638414e-06,
+      "loss": 0.5541,
+      "mean_token_accuracy": 0.8868309855461121,
+      "num_tokens": 283685194.0,
+      "step": 2653
+    },
+    {
+      "epoch": 6.045610034207526,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.8520427972364924e-06,
+      "loss": 0.5879,
+      "mean_token_accuracy": 0.8806286752223969,
+      "num_tokens": 283792579.0,
+      "step": 2654
+    },
+    {
+      "epoch": 6.047890535917902,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.8502238931771422e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.884283035993576,
+      "num_tokens": 283899492.0,
+      "step": 2655
+    },
+    {
+      "epoch": 6.050171037628278,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.848405357918166e-06,
+      "loss": 0.596,
+      "mean_token_accuracy": 0.8780128657817841,
+      "num_tokens": 284006120.0,
+      "step": 2656
+    },
+    {
+      "epoch": 6.052451539338654,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.8465871924917295e-06,
+      "loss": 0.6124,
+      "mean_token_accuracy": 0.8742113560438156,
+      "num_tokens": 284112587.0,
+      "step": 2657
+    },
+    {
+      "epoch": 6.05473204104903,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.8447693979297882e-06,
+      "loss": 0.5548,
+      "mean_token_accuracy": 0.8869588524103165,
+      "num_tokens": 284219902.0,
+      "step": 2658
+    },
+    {
+      "epoch": 6.0570125427594075,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.8429519752640862e-06,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8819890022277832,
+      "num_tokens": 284327196.0,
+      "step": 2659
+    },
+    {
+      "epoch": 6.059293044469784,
+      "grad_norm": 2.4375,
+      "learning_rate": 1.8411349255261587e-06,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8833761811256409,
+      "num_tokens": 284434704.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.06157354618016,
+      "grad_norm": 5.125,
+      "learning_rate": 1.8393182497473271e-06,
+      "loss": 0.5674,
+      "mean_token_accuracy": 0.8822661340236664,
+      "num_tokens": 284542263.0,
+      "step": 2661
+    },
+    {
+      "epoch": 6.063854047890536,
+      "grad_norm": 3.875,
+      "learning_rate": 1.837501948958702e-06,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.8850172013044357,
+      "num_tokens": 284649568.0,
+      "step": 2662
+    },
+    {
+      "epoch": 6.066134549600912,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.8356860241911817e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8816670328378677,
+      "num_tokens": 284757479.0,
+      "step": 2663
+    },
+    {
+      "epoch": 6.068415051311288,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.833870476475448e-06,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8798322230577469,
+      "num_tokens": 284864552.0,
+      "step": 2664
+    },
+    {
+      "epoch": 6.070695553021665,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.8320553068419716e-06,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8766209781169891,
+      "num_tokens": 284971361.0,
+      "step": 2665
+    },
+    {
+      "epoch": 6.072976054732041,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.830240516321008e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8826578855514526,
+      "num_tokens": 285078270.0,
+      "step": 2666
+    },
+    {
+      "epoch": 6.075256556442417,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.8284261059425972e-06,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8815598785877228,
+      "num_tokens": 285185441.0,
+      "step": 2667
+    },
+    {
+      "epoch": 6.077537058152793,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.8266120767365642e-06,
+      "loss": 0.5596,
+      "mean_token_accuracy": 0.8829776495695114,
+      "num_tokens": 285291883.0,
+      "step": 2668
+    },
+    {
+      "epoch": 6.07981755986317,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.8247984297325156e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8801479190587997,
+      "num_tokens": 285398806.0,
+      "step": 2669
+    },
+    {
+      "epoch": 6.0820980615735465,
+      "grad_norm": 3.25,
+      "learning_rate": 1.8229851659598425e-06,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8817605227231979,
+      "num_tokens": 285505881.0,
+      "step": 2670
+    },
+    {
+      "epoch": 6.084378563283923,
+      "grad_norm": 3.125,
+      "learning_rate": 1.8211722864477197e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.8797835558652878,
+      "num_tokens": 285612626.0,
+      "step": 2671
+    },
+    {
+      "epoch": 6.086659064994299,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.819359792225101e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8806306272745132,
+      "num_tokens": 285720025.0,
+      "step": 2672
+    },
+    {
+      "epoch": 6.088939566704675,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.8175476843207245e-06,
+      "loss": 0.5537,
+      "mean_token_accuracy": 0.8841009885072708,
+      "num_tokens": 285827144.0,
+      "step": 2673
+    },
+    {
+      "epoch": 6.091220068415051,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.8157359637631078e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8842537254095078,
+      "num_tokens": 285933624.0,
+      "step": 2674
+    },
+    {
+      "epoch": 6.0935005701254275,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.813924631580547e-06,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.881016343832016,
+      "num_tokens": 286040489.0,
+      "step": 2675
+    },
+    {
+      "epoch": 6.095781071835804,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.8121136888011198e-06,
+      "loss": 0.5817,
+      "mean_token_accuracy": 0.8789900839328766,
+      "num_tokens": 286146914.0,
+      "step": 2676
+    },
+    {
+      "epoch": 6.09806157354618,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.810303136452683e-06,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8830416947603226,
+      "num_tokens": 286254159.0,
+      "step": 2677
+    },
+    {
+      "epoch": 6.100342075256556,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.8084929755628707e-06,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8868089914321899,
+      "num_tokens": 286361203.0,
+      "step": 2678
+    },
+    {
+      "epoch": 6.102622576966933,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.8066832071590967e-06,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8792434334754944,
+      "num_tokens": 286468526.0,
+      "step": 2679
+    },
+    {
+      "epoch": 6.104903078677309,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.8048738322685478e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8830481469631195,
+      "num_tokens": 286575282.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.1071835803876855,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.8030648519181926e-06,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8848658800125122,
+      "num_tokens": 286682350.0,
+      "step": 2681
+    },
+    {
+      "epoch": 6.109464082098062,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.8012562671347721e-06,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8794921636581421,
+      "num_tokens": 286789153.0,
+      "step": 2682
+    },
+    {
+      "epoch": 6.111744583808438,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.7994480789448043e-06,
+      "loss": 0.598,
+      "mean_token_accuracy": 0.877688080072403,
+      "num_tokens": 286896549.0,
+      "step": 2683
+    },
+    {
+      "epoch": 6.114025085518814,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.7976402883745836e-06,
+      "loss": 0.5455,
+      "mean_token_accuracy": 0.8872903436422348,
+      "num_tokens": 287003999.0,
+      "step": 2684
+    },
+    {
+      "epoch": 6.11630558722919,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.7958328964501749e-06,
+      "loss": 0.5613,
+      "mean_token_accuracy": 0.8816230744123459,
+      "num_tokens": 287110496.0,
+      "step": 2685
+    },
+    {
+      "epoch": 6.1185860889395665,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.7940259041974189e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8829924911260605,
+      "num_tokens": 287217513.0,
+      "step": 2686
+    },
+    {
+      "epoch": 6.120866590649943,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.7922193126419306e-06,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8832157105207443,
+      "num_tokens": 287324344.0,
+      "step": 2687
+    },
+    {
+      "epoch": 6.123147092360319,
+      "grad_norm": 4.75,
+      "learning_rate": 1.7904131228090965e-06,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.8847935795783997,
+      "num_tokens": 287431380.0,
+      "step": 2688
+    },
+    {
+      "epoch": 6.125427594070696,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.7886073357240746e-06,
+      "loss": 0.5555,
+      "mean_token_accuracy": 0.8844207972288132,
+      "num_tokens": 287538513.0,
+      "step": 2689
+    },
+    {
+      "epoch": 6.127708095781072,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.7868019524117957e-06,
+      "loss": 0.5821,
+      "mean_token_accuracy": 0.8769582509994507,
+      "num_tokens": 287645242.0,
+      "step": 2690
+    },
+    {
+      "epoch": 6.129988597491448,
+      "grad_norm": 3.375,
+      "learning_rate": 1.7849969738969592e-06,
+      "loss": 0.5993,
+      "mean_token_accuracy": 0.8752488493919373,
+      "num_tokens": 287751854.0,
+      "step": 2691
+    },
+    {
+      "epoch": 6.1322690992018245,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.783192401204037e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8854573518037796,
+      "num_tokens": 287858938.0,
+      "step": 2692
+    },
+    {
+      "epoch": 6.134549600912201,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.7813882353572692e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8819147795438766,
+      "num_tokens": 287965788.0,
+      "step": 2693
+    },
+    {
+      "epoch": 6.136830102622577,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.7795844773806653e-06,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.884269118309021,
+      "num_tokens": 288073071.0,
+      "step": 2694
+    },
+    {
+      "epoch": 6.139110604332953,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.7777811282980047e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8846654891967773,
+      "num_tokens": 288180362.0,
+      "step": 2695
+    },
+    {
+      "epoch": 6.141391106043329,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.7759781891328321e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8841256648302078,
+      "num_tokens": 288287566.0,
+      "step": 2696
+    },
+    {
+      "epoch": 6.1436716077537055,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.7741756609084616e-06,
+      "loss": 0.5633,
+      "mean_token_accuracy": 0.8848319351673126,
+      "num_tokens": 288395039.0,
+      "step": 2697
+    },
+    {
+      "epoch": 6.145952109464082,
+      "grad_norm": 3.875,
+      "learning_rate": 1.772373544647973e-06,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8844243884086609,
+      "num_tokens": 288502651.0,
+      "step": 2698
+    },
+    {
+      "epoch": 6.148232611174459,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.770571841374213e-06,
+      "loss": 0.5655,
+      "mean_token_accuracy": 0.882807731628418,
+      "num_tokens": 288609620.0,
+      "step": 2699
+    },
+    {
+      "epoch": 6.150513112884835,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.7687705521097954e-06,
+      "loss": 0.5494,
+      "mean_token_accuracy": 0.8844881951808929,
+      "num_tokens": 288716986.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.152793614595211,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.766969677877094e-06,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8841705173254013,
+      "num_tokens": 288824522.0,
+      "step": 2701
+    },
+    {
+      "epoch": 6.155074116305587,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.7651692196982517e-06,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.8823985755443573,
+      "num_tokens": 288931504.0,
+      "step": 2702
+    },
+    {
+      "epoch": 6.1573546180159635,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.7633691785951746e-06,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8812411278486252,
+      "num_tokens": 289038321.0,
+      "step": 2703
+    },
+    {
+      "epoch": 6.15963511972634,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.7615695555895296e-06,
+      "loss": 0.5512,
+      "mean_token_accuracy": 0.8865274041891098,
+      "num_tokens": 289145591.0,
+      "step": 2704
+    },
+    {
+      "epoch": 6.161915621436716,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.7597703517027491e-06,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8811384439468384,
+      "num_tokens": 289252364.0,
+      "step": 2705
+    },
+    {
+      "epoch": 6.164196123147092,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.7579715679560273e-06,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8845842778682709,
+      "num_tokens": 289359378.0,
+      "step": 2706
+    },
+    {
+      "epoch": 6.166476624857468,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.7561732053703174e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8825568854808807,
+      "num_tokens": 289466858.0,
+      "step": 2707
+    },
+    {
+      "epoch": 6.168757126567845,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.7543752649663354e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8844540864229202,
+      "num_tokens": 289574570.0,
+      "step": 2708
+    },
+    {
+      "epoch": 6.1710376282782216,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.7525777477645586e-06,
+      "loss": 0.554,
+      "mean_token_accuracy": 0.8892973810434341,
+      "num_tokens": 289682543.0,
+      "step": 2709
+    },
+    {
+      "epoch": 6.173318129988598,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.7507806547852224e-06,
+      "loss": 0.5571,
+      "mean_token_accuracy": 0.885601207613945,
+      "num_tokens": 289789759.0,
+      "step": 2710
+    },
+    {
+      "epoch": 6.175598631698974,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.7489839870483236e-06,
+      "loss": 0.5518,
+      "mean_token_accuracy": 0.885807454586029,
+      "num_tokens": 289897075.0,
+      "step": 2711
+    },
+    {
+      "epoch": 6.17787913340935,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.7471877455736136e-06,
+      "loss": 0.5733,
+      "mean_token_accuracy": 0.8810791075229645,
+      "num_tokens": 290004392.0,
+      "step": 2712
+    },
+    {
+      "epoch": 6.180159635119726,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.7453919313806057e-06,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.880391463637352,
+      "num_tokens": 290111469.0,
+      "step": 2713
+    },
+    {
+      "epoch": 6.1824401368301025,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.7435965454885699e-06,
+      "loss": 0.5589,
+      "mean_token_accuracy": 0.8832577913999557,
+      "num_tokens": 290218539.0,
+      "step": 2714
+    },
+    {
+      "epoch": 6.184720638540479,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.7418015889165312e-06,
+      "loss": 0.533,
+      "mean_token_accuracy": 0.8896812498569489,
+      "num_tokens": 290326424.0,
+      "step": 2715
+    },
+    {
+      "epoch": 6.187001140250855,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.7400070626832732e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8802052289247513,
+      "num_tokens": 290433540.0,
+      "step": 2716
+    },
+    {
+      "epoch": 6.189281641961231,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.7382129678073351e-06,
+      "loss": 0.5833,
+      "mean_token_accuracy": 0.878139853477478,
+      "num_tokens": 290540729.0,
+      "step": 2717
+    },
+    {
+      "epoch": 6.191562143671608,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.7364193053070082e-06,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.8827469646930695,
+      "num_tokens": 290647100.0,
+      "step": 2718
+    },
+    {
+      "epoch": 6.193842645381984,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.7346260762003428e-06,
+      "loss": 0.5656,
+      "mean_token_accuracy": 0.8845665007829666,
+      "num_tokens": 290753980.0,
+      "step": 2719
+    },
+    {
+      "epoch": 6.196123147092361,
+      "grad_norm": 5.40625,
+      "learning_rate": 1.7328332815051403e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8774483948945999,
+      "num_tokens": 290861484.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.198403648802737,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.7310409222389563e-06,
+      "loss": 0.5454,
+      "mean_token_accuracy": 0.8843204379081726,
+      "num_tokens": 290968606.0,
+      "step": 2721
+    },
+    {
+      "epoch": 6.200684150513113,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.7292489994191005e-06,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8810103088617325,
+      "num_tokens": 291075885.0,
+      "step": 2722
+    },
+    {
+      "epoch": 6.202964652223489,
+      "grad_norm": 2.75,
+      "learning_rate": 1.7274575140626318e-06,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8839705884456635,
+      "num_tokens": 291183303.0,
+      "step": 2723
+    },
+    {
+      "epoch": 6.205245153933865,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.7256664671863634e-06,
+      "loss": 0.5398,
+      "mean_token_accuracy": 0.8867494761943817,
+      "num_tokens": 291290745.0,
+      "step": 2724
+    },
+    {
+      "epoch": 6.2075256556442415,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.72387585980686e-06,
+      "loss": 0.6035,
+      "mean_token_accuracy": 0.8760453164577484,
+      "num_tokens": 291397663.0,
+      "step": 2725
+    },
+    {
+      "epoch": 6.209806157354618,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.7220856929404342e-06,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.8818619549274445,
+      "num_tokens": 291504707.0,
+      "step": 2726
+    },
+    {
+      "epoch": 6.212086659064994,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.720295967603152e-06,
+      "loss": 0.5569,
+      "mean_token_accuracy": 0.8842518627643585,
+      "num_tokens": 291611837.0,
+      "step": 2727
+    },
+    {
+      "epoch": 6.214367160775371,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.7185066848108244e-06,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8826566338539124,
+      "num_tokens": 291718969.0,
+      "step": 2728
+    },
+    {
+      "epoch": 6.216647662485747,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.7167178455790157e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8848382532596588,
+      "num_tokens": 291826377.0,
+      "step": 2729
+    },
+    {
+      "epoch": 6.218928164196123,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.7149294509230357e-06,
+      "loss": 0.5986,
+      "mean_token_accuracy": 0.8756033033132553,
+      "num_tokens": 291933188.0,
+      "step": 2730
+    },
+    {
+      "epoch": 6.2212086659065,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.713141501857943e-06,
+      "loss": 0.5674,
+      "mean_token_accuracy": 0.8862540572881699,
+      "num_tokens": 292040480.0,
+      "step": 2731
+    },
+    {
+      "epoch": 6.223489167616876,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.7113539993985431e-06,
+      "loss": 0.5827,
+      "mean_token_accuracy": 0.8784683644771576,
+      "num_tokens": 292147117.0,
+      "step": 2732
+    },
+    {
+      "epoch": 6.225769669327252,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.7095669445593887e-06,
+      "loss": 0.55,
+      "mean_token_accuracy": 0.8856497257947922,
+      "num_tokens": 292254295.0,
+      "step": 2733
+    },
+    {
+      "epoch": 6.228050171037628,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.707780338354776e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8842949420213699,
+      "num_tokens": 292361287.0,
+      "step": 2734
+    },
+    {
+      "epoch": 6.230330672748004,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.7059941817987485e-06,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8810069411993027,
+      "num_tokens": 292468510.0,
+      "step": 2735
+    },
+    {
+      "epoch": 6.2326111744583805,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.7042084759050948e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8827442526817322,
+      "num_tokens": 292575697.0,
+      "step": 2736
+    },
+    {
+      "epoch": 6.234891676168757,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.7024232216873465e-06,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8848889619112015,
+      "num_tokens": 292682969.0,
+      "step": 2737
+    },
+    {
+      "epoch": 6.237172177879134,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.7006384201587809e-06,
+      "loss": 0.5584,
+      "mean_token_accuracy": 0.884590744972229,
+      "num_tokens": 292790103.0,
+      "step": 2738
+    },
+    {
+      "epoch": 6.23945267958951,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.6988540723324145e-06,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.881315678358078,
+      "num_tokens": 292897356.0,
+      "step": 2739
+    },
+    {
+      "epoch": 6.241733181299886,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.6970701792210101e-06,
+      "loss": 0.5632,
+      "mean_token_accuracy": 0.8851087540388107,
+      "num_tokens": 293004918.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.244013683010262,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.6952867418370707e-06,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8804485946893692,
+      "num_tokens": 293112328.0,
+      "step": 2741
+    },
+    {
+      "epoch": 6.246294184720639,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.6935037611928412e-06,
+      "loss": 0.5993,
+      "mean_token_accuracy": 0.8765440136194229,
+      "num_tokens": 293218993.0,
+      "step": 2742
+    },
+    {
+      "epoch": 6.248574686431015,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.691721238300308e-06,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.882225289940834,
+      "num_tokens": 293325757.0,
+      "step": 2743
+    },
+    {
+      "epoch": 6.250855188141391,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.689939174171194e-06,
+      "loss": 0.5923,
+      "mean_token_accuracy": 0.876752108335495,
+      "num_tokens": 293432687.0,
+      "step": 2744
+    },
+    {
+      "epoch": 6.253135689851767,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.6881575698169662e-06,
+      "loss": 0.5656,
+      "mean_token_accuracy": 0.8828807026147842,
+      "num_tokens": 293540213.0,
+      "step": 2745
+    },
+    {
+      "epoch": 6.255416191562143,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.6863764262488292e-06,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.8817939162254333,
+      "num_tokens": 293647497.0,
+      "step": 2746
+    },
+    {
+      "epoch": 6.2576966932725195,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.6845957444777244e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8802742511034012,
+      "num_tokens": 293754862.0,
+      "step": 2747
+    },
+    {
+      "epoch": 6.259977194982897,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.6828155255143331e-06,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.882528692483902,
+      "num_tokens": 293861700.0,
+      "step": 2748
+    },
+    {
+      "epoch": 6.262257696693273,
+      "grad_norm": 6.125,
+      "learning_rate": 1.6810357703690739e-06,
+      "loss": 0.5875,
+      "mean_token_accuracy": 0.8790639489889145,
+      "num_tokens": 293969027.0,
+      "step": 2749
+    },
+    {
+      "epoch": 6.264538198403649,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.6792564800521e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8794273883104324,
+      "num_tokens": 294075857.0,
+      "step": 2750
+    },
+    {
+      "epoch": 6.266818700114025,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.677477655573303e-06,
+      "loss": 0.5915,
+      "mean_token_accuracy": 0.8795068562030792,
+      "num_tokens": 294182659.0,
+      "step": 2751
+    },
+    {
+      "epoch": 6.269099201824401,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.675699297942309e-06,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8815376460552216,
+      "num_tokens": 294289557.0,
+      "step": 2752
+    },
+    {
+      "epoch": 6.271379703534778,
+      "grad_norm": 3.0,
+      "learning_rate": 1.6739214081684799e-06,
+      "loss": 0.5931,
+      "mean_token_accuracy": 0.876123234629631,
+      "num_tokens": 294396897.0,
+      "step": 2753
+    },
+    {
+      "epoch": 6.273660205245154,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.6721439872609125e-06,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.8840298056602478,
+      "num_tokens": 294504720.0,
+      "step": 2754
+    },
+    {
+      "epoch": 6.27594070695553,
+      "grad_norm": 6.96875,
+      "learning_rate": 1.6703670362284346e-06,
+      "loss": 0.5652,
+      "mean_token_accuracy": 0.8847046941518784,
+      "num_tokens": 294611688.0,
+      "step": 2755
+    },
+    {
+      "epoch": 6.278221208665906,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.6685905560796101e-06,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.878638818860054,
+      "num_tokens": 294718635.0,
+      "step": 2756
+    },
+    {
+      "epoch": 6.280501710376283,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.6668145478227354e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8829765170812607,
+      "num_tokens": 294825331.0,
+      "step": 2757
+    },
+    {
+      "epoch": 6.282782212086659,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.6650390124658378e-06,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.878277525305748,
+      "num_tokens": 294932357.0,
+      "step": 2758
+    },
+    {
+      "epoch": 6.285062713797036,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.663263951016678e-06,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8813124597072601,
+      "num_tokens": 295039173.0,
+      "step": 2759
+    },
+    {
+      "epoch": 6.287343215507412,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.661489364482745e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8825227618217468,
+      "num_tokens": 295146182.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.289623717217788,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.6597152538712608e-06,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.8797194510698318,
+      "num_tokens": 295253659.0,
+      "step": 2761
+    },
+    {
+      "epoch": 6.291904218928164,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.6579416201891757e-06,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8829602301120758,
+      "num_tokens": 295361105.0,
+      "step": 2762
+    },
+    {
+      "epoch": 6.29418472063854,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.6561684644431709e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8760716319084167,
+      "num_tokens": 295467879.0,
+      "step": 2763
+    },
+    {
+      "epoch": 6.296465222348917,
+      "grad_norm": 4.25,
+      "learning_rate": 1.6543957876396544e-06,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8809865862131119,
+      "num_tokens": 295575043.0,
+      "step": 2764
+    },
+    {
+      "epoch": 6.298745724059293,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.6526235907847649e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8808043301105499,
+      "num_tokens": 295682338.0,
+      "step": 2765
+    },
+    {
+      "epoch": 6.301026225769669,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.6508518748843651e-06,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8840913772583008,
+      "num_tokens": 295789718.0,
+      "step": 2766
+    },
+    {
+      "epoch": 6.303306727480045,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.649080640944048e-06,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8826287239789963,
+      "num_tokens": 295897773.0,
+      "step": 2767
+    },
+    {
+      "epoch": 6.305587229190422,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.6473098899691313e-06,
+      "loss": 0.5909,
+      "mean_token_accuracy": 0.8791597187519073,
+      "num_tokens": 296004423.0,
+      "step": 2768
+    },
+    {
+      "epoch": 6.307867730900798,
+      "grad_norm": 4.0,
+      "learning_rate": 1.6455396229646595e-06,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8822521567344666,
+      "num_tokens": 296111701.0,
+      "step": 2769
+    },
+    {
+      "epoch": 6.310148232611175,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.6437698409354025e-06,
+      "loss": 0.5532,
+      "mean_token_accuracy": 0.887429729104042,
+      "num_tokens": 296218964.0,
+      "step": 2770
+    },
+    {
+      "epoch": 6.312428734321551,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.6420005448858522e-06,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8810682147741318,
+      "num_tokens": 296325719.0,
+      "step": 2771
+    },
+    {
+      "epoch": 6.314709236031927,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.6402317358202286e-06,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8821781575679779,
+      "num_tokens": 296432464.0,
+      "step": 2772
+    },
+    {
+      "epoch": 6.316989737742303,
+      "grad_norm": 4.125,
+      "learning_rate": 1.6384634147424732e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8829574584960938,
+      "num_tokens": 296539951.0,
+      "step": 2773
+    },
+    {
+      "epoch": 6.319270239452679,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.636695582656251e-06,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8789397776126862,
+      "num_tokens": 296646664.0,
+      "step": 2774
+    },
+    {
+      "epoch": 6.321550741163056,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.6349282405649506e-06,
+      "loss": 0.5393,
+      "mean_token_accuracy": 0.889549657702446,
+      "num_tokens": 296754082.0,
+      "step": 2775
+    },
+    {
+      "epoch": 6.323831242873432,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.6331613894716787e-06,
+      "loss": 0.5546,
+      "mean_token_accuracy": 0.8843847513198853,
+      "num_tokens": 296861010.0,
+      "step": 2776
+    },
+    {
+      "epoch": 6.326111744583809,
+      "grad_norm": 6.21875,
+      "learning_rate": 1.6313950303792672e-06,
+      "loss": 0.5679,
+      "mean_token_accuracy": 0.8804360330104828,
+      "num_tokens": 296968168.0,
+      "step": 2777
+    },
+    {
+      "epoch": 6.328392246294185,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.6296291642902673e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8830510526895523,
+      "num_tokens": 297075136.0,
+      "step": 2778
+    },
+    {
+      "epoch": 6.330672748004561,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.6278637922069512e-06,
+      "loss": 0.5924,
+      "mean_token_accuracy": 0.8757496923208237,
+      "num_tokens": 297182280.0,
+      "step": 2779
+    },
+    {
+      "epoch": 6.3329532497149374,
+      "grad_norm": 4.125,
+      "learning_rate": 1.6260989151313091e-06,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8829326927661896,
+      "num_tokens": 297288826.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.335233751425314,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.6243345340650523e-06,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8820019513368607,
+      "num_tokens": 297395720.0,
+      "step": 2781
+    },
+    {
+      "epoch": 6.33751425313569,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.6225706500096079e-06,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8875852972269058,
+      "num_tokens": 297502615.0,
+      "step": 2782
+    },
+    {
+      "epoch": 6.339794754846066,
+      "grad_norm": 2.875,
+      "learning_rate": 1.6208072639661226e-06,
+      "loss": 0.5507,
+      "mean_token_accuracy": 0.8859220743179321,
+      "num_tokens": 297609799.0,
+      "step": 2783
+    },
+    {
+      "epoch": 6.342075256556442,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.6190443769354608e-06,
+      "loss": 0.5932,
+      "mean_token_accuracy": 0.8765744715929031,
+      "num_tokens": 297716620.0,
+      "step": 2784
+    },
+    {
+      "epoch": 6.344355758266818,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.6172819899182036e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8797289431095123,
+      "num_tokens": 297823497.0,
+      "step": 2785
+    },
+    {
+      "epoch": 6.346636259977195,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.6155201039146478e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.876724436879158,
+      "num_tokens": 297930683.0,
+      "step": 2786
+    },
+    {
+      "epoch": 6.348916761687571,
+      "grad_norm": 3.75,
+      "learning_rate": 1.613758719924805e-06,
+      "loss": 0.589,
+      "mean_token_accuracy": 0.8765078336000443,
+      "num_tokens": 298037433.0,
+      "step": 2787
+    },
+    {
+      "epoch": 6.351197263397948,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.611997838948403e-06,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8840204477310181,
+      "num_tokens": 298145428.0,
+      "step": 2788
+    },
+    {
+      "epoch": 6.353477765108324,
+      "grad_norm": 4.71875,
+      "learning_rate": 1.6102374619848845e-06,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8840190768241882,
+      "num_tokens": 298252510.0,
+      "step": 2789
+    },
+    {
+      "epoch": 6.3557582668187,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.6084775900334046e-06,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8814188688993454,
+      "num_tokens": 298359365.0,
+      "step": 2790
+    },
+    {
+      "epoch": 6.3580387685290765,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6067182240928332e-06,
+      "loss": 0.5766,
+      "mean_token_accuracy": 0.8799934387207031,
+      "num_tokens": 298466492.0,
+      "step": 2791
+    },
+    {
+      "epoch": 6.360319270239453,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.6049593651617534e-06,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8819076120853424,
+      "num_tokens": 298573579.0,
+      "step": 2792
+    },
+    {
+      "epoch": 6.362599771949829,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.6032010142384572e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.880363255739212,
+      "num_tokens": 298680688.0,
+      "step": 2793
+    },
+    {
+      "epoch": 6.364880273660205,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.6014431723209522e-06,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.8782119452953339,
+      "num_tokens": 298787550.0,
+      "step": 2794
+    },
+    {
+      "epoch": 6.367160775370581,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.599685840406955e-06,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8798055797815323,
+      "num_tokens": 298895010.0,
+      "step": 2795
+    },
+    {
+      "epoch": 6.369441277080957,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.5979290194938938e-06,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8793617337942123,
+      "num_tokens": 299001708.0,
+      "step": 2796
+    },
+    {
+      "epoch": 6.3717217787913345,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.5961727105789072e-06,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8799160867929459,
+      "num_tokens": 299108766.0,
+      "step": 2797
+    },
+    {
+      "epoch": 6.374002280501711,
+      "grad_norm": 3.375,
+      "learning_rate": 1.5944169146588395e-06,
+      "loss": 0.5493,
+      "mean_token_accuracy": 0.8847594708204269,
+      "num_tokens": 299215718.0,
+      "step": 2798
+    },
+    {
+      "epoch": 6.376282782212087,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.5926616327302482e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8823742717504501,
+      "num_tokens": 299322716.0,
+      "step": 2799
+    },
+    {
+      "epoch": 6.378563283922463,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.5909068657893978e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8821296989917755,
+      "num_tokens": 299429926.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.380843785632839,
+      "grad_norm": 2.75,
+      "learning_rate": 1.5891526148322594e-06,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8831614851951599,
+      "num_tokens": 299537185.0,
+      "step": 2801
+    },
+    {
+      "epoch": 6.3831242873432155,
+      "grad_norm": 3.125,
+      "learning_rate": 1.5873988808545127e-06,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8860654383897781,
+      "num_tokens": 299644399.0,
+      "step": 2802
+    },
+    {
+      "epoch": 6.385404789053592,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.5856456648515425e-06,
+      "loss": 0.5397,
+      "mean_token_accuracy": 0.8893062770366669,
+      "num_tokens": 299751766.0,
+      "step": 2803
+    },
+    {
+      "epoch": 6.387685290763968,
+      "grad_norm": 2.5,
+      "learning_rate": 1.5838929678184405e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.880397379398346,
+      "num_tokens": 299857907.0,
+      "step": 2804
+    },
+    {
+      "epoch": 6.389965792474344,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.5821407907500036e-06,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8866594880819321,
+      "num_tokens": 299965673.0,
+      "step": 2805
+    },
+    {
+      "epoch": 6.39224629418472,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.5803891346407342e-06,
+      "loss": 0.5633,
+      "mean_token_accuracy": 0.8857071846723557,
+      "num_tokens": 300072879.0,
+      "step": 2806
+    },
+    {
+      "epoch": 6.394526795895097,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.5786380004848379e-06,
+      "loss": 0.582,
+      "mean_token_accuracy": 0.8817501962184906,
+      "num_tokens": 300180006.0,
+      "step": 2807
+    },
+    {
+      "epoch": 6.3968072976054735,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.576887389276226e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8835788518190384,
+      "num_tokens": 300287280.0,
+      "step": 2808
+    },
+    {
+      "epoch": 6.39908779931585,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.5751373020085093e-06,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8786880671977997,
+      "num_tokens": 300394445.0,
+      "step": 2809
+    },
+    {
+      "epoch": 6.401368301026226,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.5733877396750051e-06,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8815035223960876,
+      "num_tokens": 300501045.0,
+      "step": 2810
+    },
+    {
+      "epoch": 6.403648802736602,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.5716387032687314e-06,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.883103683590889,
+      "num_tokens": 300607844.0,
+      "step": 2811
+    },
+    {
+      "epoch": 6.405929304446978,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.5698901937824066e-06,
+      "loss": 0.5944,
+      "mean_token_accuracy": 0.8773173838853836,
+      "num_tokens": 300714942.0,
+      "step": 2812
+    },
+    {
+      "epoch": 6.4082098061573545,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.5681422122084522e-06,
+      "loss": 0.5497,
+      "mean_token_accuracy": 0.8895971029996872,
+      "num_tokens": 300822920.0,
+      "step": 2813
+    },
+    {
+      "epoch": 6.410490307867731,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.5663947595389873e-06,
+      "loss": 0.5817,
+      "mean_token_accuracy": 0.8785582333803177,
+      "num_tokens": 300930457.0,
+      "step": 2814
+    },
+    {
+      "epoch": 6.412770809578107,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.5646478367658325e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8812803626060486,
+      "num_tokens": 301037070.0,
+      "step": 2815
+    },
+    {
+      "epoch": 6.415051311288483,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.562901444880508e-06,
+      "loss": 0.5524,
+      "mean_token_accuracy": 0.8858920186758041,
+      "num_tokens": 301144185.0,
+      "step": 2816
+    },
+    {
+      "epoch": 6.41733181299886,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.5611555848742318e-06,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.8812452852725983,
+      "num_tokens": 301251360.0,
+      "step": 2817
+    },
+    {
+      "epoch": 6.419612314709236,
+      "grad_norm": 4.25,
+      "learning_rate": 1.5594102577379216e-06,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8858301192522049,
+      "num_tokens": 301358185.0,
+      "step": 2818
+    },
+    {
+      "epoch": 6.4218928164196125,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.5576654644621897e-06,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8778669685125351,
+      "num_tokens": 301464922.0,
+      "step": 2819
+    },
+    {
+      "epoch": 6.424173318129989,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.5559212060373474e-06,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8837179839611053,
+      "num_tokens": 301571945.0,
+      "step": 2820
+    },
+    {
+      "epoch": 6.426453819840365,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.5541774834534024e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.8790989071130753,
+      "num_tokens": 301678995.0,
+      "step": 2821
+    },
+    {
+      "epoch": 6.428734321550741,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.5524342977000587e-06,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.880190521478653,
+      "num_tokens": 301786689.0,
+      "step": 2822
+    },
+    {
+      "epoch": 6.431014823261117,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.5506916497667134e-06,
+      "loss": 0.5599,
+      "mean_token_accuracy": 0.8833551555871964,
+      "num_tokens": 301894003.0,
+      "step": 2823
+    },
+    {
+      "epoch": 6.4332953249714935,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.5489495406424618e-06,
+      "loss": 0.5376,
+      "mean_token_accuracy": 0.8886000365018845,
+      "num_tokens": 302001595.0,
+      "step": 2824
+    },
+    {
+      "epoch": 6.43557582668187,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.5472079713160892e-06,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8830095380544662,
+      "num_tokens": 302109787.0,
+      "step": 2825
+    },
+    {
+      "epoch": 6.437856328392247,
+      "grad_norm": 4.0,
+      "learning_rate": 1.5454669427760774e-06,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8822568953037262,
+      "num_tokens": 302216211.0,
+      "step": 2826
+    },
+    {
+      "epoch": 6.440136830102623,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.5437264560106014e-06,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8839522004127502,
+      "num_tokens": 302323292.0,
+      "step": 2827
+    },
+    {
+      "epoch": 6.442417331812999,
+      "grad_norm": 2.75,
+      "learning_rate": 1.5419865120075267e-06,
+      "loss": 0.5576,
+      "mean_token_accuracy": 0.8829583674669266,
+      "num_tokens": 302430317.0,
+      "step": 2828
+    },
+    {
+      "epoch": 6.444697833523375,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.5402471117544143e-06,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8828336149454117,
+      "num_tokens": 302537141.0,
+      "step": 2829
+    },
+    {
+      "epoch": 6.4469783352337515,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.5385082562385112e-06,
+      "loss": 0.5505,
+      "mean_token_accuracy": 0.8884585201740265,
+      "num_tokens": 302644739.0,
+      "step": 2830
+    },
+    {
+      "epoch": 6.449258836944128,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.5367699464467596e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8810509741306305,
+      "num_tokens": 302751800.0,
+      "step": 2831
+    },
+    {
+      "epoch": 6.451539338654504,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.5350321833657904e-06,
+      "loss": 0.6045,
+      "mean_token_accuracy": 0.8751820474863052,
+      "num_tokens": 302858532.0,
+      "step": 2832
+    },
+    {
+      "epoch": 6.45381984036488,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.5332949679819251e-06,
+      "loss": 0.5479,
+      "mean_token_accuracy": 0.8875158727169037,
+      "num_tokens": 302965381.0,
+      "step": 2833
+    },
+    {
+      "epoch": 6.456100342075256,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.531558301281173e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8815342336893082,
+      "num_tokens": 303072747.0,
+      "step": 2834
+    },
+    {
+      "epoch": 6.4583808437856325,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.5298221842492328e-06,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8833811283111572,
+      "num_tokens": 303179347.0,
+      "step": 2835
+    },
+    {
+      "epoch": 6.460661345496009,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.5280866178714898e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8801722079515457,
+      "num_tokens": 303287273.0,
+      "step": 2836
+    },
+    {
+      "epoch": 6.462941847206386,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.5263516031330195e-06,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.8790634274482727,
+      "num_tokens": 303394414.0,
+      "step": 2837
+    },
+    {
+      "epoch": 6.465222348916762,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.524617141018582e-06,
+      "loss": 0.5903,
+      "mean_token_accuracy": 0.8738131821155548,
+      "num_tokens": 303501157.0,
+      "step": 2838
+    },
+    {
+      "epoch": 6.467502850627138,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.5228832325126248e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.881572037935257,
+      "num_tokens": 303608294.0,
+      "step": 2839
+    },
+    {
+      "epoch": 6.469783352337514,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.5211498785992818e-06,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8818154633045197,
+      "num_tokens": 303715138.0,
+      "step": 2840
+    },
+    {
+      "epoch": 6.4720638540478905,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.5194170802623692e-06,
+      "loss": 0.5776,
+      "mean_token_accuracy": 0.8801160901784897,
+      "num_tokens": 303822023.0,
+      "step": 2841
+    },
+    {
+      "epoch": 6.474344355758267,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.5176848384853913e-06,
+      "loss": 0.5854,
+      "mean_token_accuracy": 0.8776775300502777,
+      "num_tokens": 303929007.0,
+      "step": 2842
+    },
+    {
+      "epoch": 6.476624857468643,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.515953154251535e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8842798173427582,
+      "num_tokens": 304035636.0,
+      "step": 2843
+    },
+    {
+      "epoch": 6.478905359179019,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.5142220285436701e-06,
+      "loss": 0.5627,
+      "mean_token_accuracy": 0.8824753314256668,
+      "num_tokens": 304142349.0,
+      "step": 2844
+    },
+    {
+      "epoch": 6.481185860889395,
+      "grad_norm": 4.0,
+      "learning_rate": 1.512491462344351e-06,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8803005963563919,
+      "num_tokens": 304249360.0,
+      "step": 2845
+    },
+    {
+      "epoch": 6.483466362599772,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.5107614566358136e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8824707418680191,
+      "num_tokens": 304357003.0,
+      "step": 2846
+    },
+    {
+      "epoch": 6.485746864310149,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.5090320123999746e-06,
+      "loss": 0.5918,
+      "mean_token_accuracy": 0.8782989233732224,
+      "num_tokens": 304463367.0,
+      "step": 2847
+    },
+    {
+      "epoch": 6.488027366020525,
+      "grad_norm": 4.90625,
+      "learning_rate": 1.5073031306184343e-06,
+      "loss": 0.5776,
+      "mean_token_accuracy": 0.882141649723053,
+      "num_tokens": 304570072.0,
+      "step": 2848
+    },
+    {
+      "epoch": 6.490307867730901,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.5055748122724722e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8819586336612701,
+      "num_tokens": 304677214.0,
+      "step": 2849
+    },
+    {
+      "epoch": 6.492588369441277,
+      "grad_norm": 2.4375,
+      "learning_rate": 1.5038470583430485e-06,
+      "loss": 0.5446,
+      "mean_token_accuracy": 0.8865179270505905,
+      "num_tokens": 304784053.0,
+      "step": 2850
+    },
+    {
+      "epoch": 6.494868871151653,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.5021198698108038e-06,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.8798740953207016,
+      "num_tokens": 304891175.0,
+      "step": 2851
+    },
+    {
+      "epoch": 6.4971493728620295,
+      "grad_norm": 6.3125,
+      "learning_rate": 1.5003932476560554e-06,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8817462623119354,
+      "num_tokens": 304997679.0,
+      "step": 2852
+    },
+    {
+      "epoch": 6.499429874572406,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.4986671928588016e-06,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8805342614650726,
+      "num_tokens": 305104659.0,
+      "step": 2853
+    },
+    {
+      "epoch": 6.501710376282782,
+      "grad_norm": 7.0,
+      "learning_rate": 1.496941706398718e-06,
+      "loss": 0.5651,
+      "mean_token_accuracy": 0.8827610611915588,
+      "num_tokens": 305211801.0,
+      "step": 2854
+    },
+    {
+      "epoch": 6.503990877993158,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.495216789255156e-06,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.8812805563211441,
+      "num_tokens": 305317978.0,
+      "step": 2855
+    },
+    {
+      "epoch": 6.506271379703534,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.4934924424071479e-06,
+      "loss": 0.5954,
+      "mean_token_accuracy": 0.8763656914234161,
+      "num_tokens": 305424656.0,
+      "step": 2856
+    },
+    {
+      "epoch": 6.508551881413911,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.4917686668333975e-06,
+      "loss": 0.5881,
+      "mean_token_accuracy": 0.881349191069603,
+      "num_tokens": 305531468.0,
+      "step": 2857
+    },
+    {
+      "epoch": 6.510832383124288,
+      "grad_norm": 6.21875,
+      "learning_rate": 1.4900454635122866e-06,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8755508661270142,
+      "num_tokens": 305638201.0,
+      "step": 2858
+    },
+    {
+      "epoch": 6.513112884834664,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.4883228334218727e-06,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8790022134780884,
+      "num_tokens": 305745038.0,
+      "step": 2859
+    },
+    {
+      "epoch": 6.51539338654504,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.4866007775398874e-06,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8798090815544128,
+      "num_tokens": 305852570.0,
+      "step": 2860
+    },
+    {
+      "epoch": 6.51539338654504,
+      "eval_loss": 0.5866165161132812,
+      "eval_mean_token_accuracy": 0.8798247088497583,
+      "eval_num_tokens": 305852570.0,
+      "eval_runtime": 58.6358,
+      "eval_samples_per_second": 143.001,
+      "eval_steps_per_second": 4.485,
+      "step": 2860
+    },
+    {
+      "epoch": 6.517673888255416,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.4848792968437376e-06,
+      "loss": 0.54,
+      "mean_token_accuracy": 0.888011172413826,
+      "num_tokens": 305960317.0,
+      "step": 2861
+    },
+    {
+      "epoch": 6.519954389965792,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.4831583923105e-06,
+      "loss": 0.5677,
+      "mean_token_accuracy": 0.8860371559858322,
+      "num_tokens": 306067692.0,
+      "step": 2862
+    },
+    {
+      "epoch": 6.5222348916761685,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.481438064916928e-06,
+      "loss": 0.5586,
+      "mean_token_accuracy": 0.8848079293966293,
+      "num_tokens": 306174375.0,
+      "step": 2863
+    },
+    {
+      "epoch": 6.524515393386545,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.4797183156394462e-06,
+      "loss": 0.5544,
+      "mean_token_accuracy": 0.8868988156318665,
+      "num_tokens": 306282289.0,
+      "step": 2864
+    },
+    {
+      "epoch": 6.526795895096921,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.477999145454152e-06,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8846388012170792,
+      "num_tokens": 306389414.0,
+      "step": 2865
+    },
+    {
+      "epoch": 6.529076396807298,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.4762805553368115e-06,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.8790937066078186,
+      "num_tokens": 306496182.0,
+      "step": 2866
+    },
+    {
+      "epoch": 6.531356898517674,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.4745625462628654e-06,
+      "loss": 0.5926,
+      "mean_token_accuracy": 0.8804367631673813,
+      "num_tokens": 306603764.0,
+      "step": 2867
+    },
+    {
+      "epoch": 6.53363740022805,
+      "grad_norm": 4.0,
+      "learning_rate": 1.47284511920742e-06,
+      "loss": 0.5766,
+      "mean_token_accuracy": 0.8812845051288605,
+      "num_tokens": 306710535.0,
+      "step": 2868
+    },
+    {
+      "epoch": 6.535917901938427,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.4711282751452549e-06,
+      "loss": 0.6152,
+      "mean_token_accuracy": 0.8755820095539093,
+      "num_tokens": 306817409.0,
+      "step": 2869
+    },
+    {
+      "epoch": 6.538198403648803,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.4694120150508179e-06,
+      "loss": 0.5614,
+      "mean_token_accuracy": 0.8860864490270615,
+      "num_tokens": 306924651.0,
+      "step": 2870
+    },
+    {
+      "epoch": 6.540478905359179,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.4676963398982248e-06,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8829345405101776,
+      "num_tokens": 307032470.0,
+      "step": 2871
+    },
+    {
+      "epoch": 6.542759407069555,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.4659812506612608e-06,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8877181857824326,
+      "num_tokens": 307139315.0,
+      "step": 2872
+    },
+    {
+      "epoch": 6.545039908779931,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.4642667483133753e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8820486813783646,
+      "num_tokens": 307245855.0,
+      "step": 2873
+    },
+    {
+      "epoch": 6.5473204104903076,
+      "grad_norm": 4.0,
+      "learning_rate": 1.4625528338276879e-06,
+      "loss": 0.5888,
+      "mean_token_accuracy": 0.8795684576034546,
+      "num_tokens": 307352674.0,
+      "step": 2874
+    },
+    {
+      "epoch": 6.549600912200685,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.4608395081769833e-06,
+      "loss": 0.551,
+      "mean_token_accuracy": 0.885527640581131,
+      "num_tokens": 307460336.0,
+      "step": 2875
+    },
+    {
+      "epoch": 6.55188141391106,
+      "grad_norm": 2.875,
+      "learning_rate": 1.4591267723337122e-06,
+      "loss": 0.5699,
+      "mean_token_accuracy": 0.8819777965545654,
+      "num_tokens": 307567742.0,
+      "step": 2876
+    },
+    {
+      "epoch": 6.554161915621437,
+      "grad_norm": 2.75,
+      "learning_rate": 1.4574146272699914e-06,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8834614157676697,
+      "num_tokens": 307674838.0,
+      "step": 2877
+    },
+    {
+      "epoch": 6.556442417331813,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.4557030739575988e-06,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8842470943927765,
+      "num_tokens": 307782038.0,
+      "step": 2878
+    },
+    {
+      "epoch": 6.558722919042189,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.4539921133679808e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8773439824581146,
+      "num_tokens": 307889575.0,
+      "step": 2879
+    },
+    {
+      "epoch": 6.561003420752566,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.4522817464722453e-06,
+      "loss": 0.5685,
+      "mean_token_accuracy": 0.8828644007444382,
+      "num_tokens": 307996562.0,
+      "step": 2880
+    },
+    {
+      "epoch": 6.563283922462942,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.4505719742411644e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8790508657693863,
+      "num_tokens": 308103487.0,
+      "step": 2881
+    },
+    {
+      "epoch": 6.565564424173318,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.44886279764517e-06,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.881091371178627,
+      "num_tokens": 308210017.0,
+      "step": 2882
+    },
+    {
+      "epoch": 6.567844925883694,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.4471542176543587e-06,
+      "loss": 0.5541,
+      "mean_token_accuracy": 0.8873138725757599,
+      "num_tokens": 308316619.0,
+      "step": 2883
+    },
+    {
+      "epoch": 6.57012542759407,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.4454462352384885e-06,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8838976472616196,
+      "num_tokens": 308424126.0,
+      "step": 2884
+    },
+    {
+      "epoch": 6.572405929304447,
+      "grad_norm": 2.40625,
+      "learning_rate": 1.4437388513669754e-06,
+      "loss": 0.5733,
+      "mean_token_accuracy": 0.880536213517189,
+      "num_tokens": 308531132.0,
+      "step": 2885
+    },
+    {
+      "epoch": 6.574686431014824,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.4420320670088977e-06,
+      "loss": 0.5665,
+      "mean_token_accuracy": 0.8818255960941315,
+      "num_tokens": 308639423.0,
+      "step": 2886
+    },
+    {
+      "epoch": 6.5769669327252,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.4403258831329947e-06,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.8803568929433823,
+      "num_tokens": 308746298.0,
+      "step": 2887
+    },
+    {
+      "epoch": 6.579247434435576,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.4386203007076632e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8798245489597321,
+      "num_tokens": 308853686.0,
+      "step": 2888
+    },
+    {
+      "epoch": 6.581527936145952,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.4369153207009573e-06,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8845240324735641,
+      "num_tokens": 308960699.0,
+      "step": 2889
+    },
+    {
+      "epoch": 6.583808437856328,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.4352109440805917e-06,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.880035325884819,
+      "num_tokens": 309067894.0,
+      "step": 2890
+    },
+    {
+      "epoch": 6.586088939566705,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.4335071718139379e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8802482336759567,
+      "num_tokens": 309175232.0,
+      "step": 2891
+    },
+    {
+      "epoch": 6.588369441277081,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.4318040048680238e-06,
+      "loss": 0.5359,
+      "mean_token_accuracy": 0.8893679231405258,
+      "num_tokens": 309282678.0,
+      "step": 2892
+    },
+    {
+      "epoch": 6.590649942987457,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.430101444209535e-06,
+      "loss": 0.5545,
+      "mean_token_accuracy": 0.8879834562540054,
+      "num_tokens": 309390130.0,
+      "step": 2893
+    },
+    {
+      "epoch": 6.592930444697833,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.4283994908048107e-06,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8846594393253326,
+      "num_tokens": 309497117.0,
+      "step": 2894
+    },
+    {
+      "epoch": 6.59521094640821,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.426698145619847e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8807022422552109,
+      "num_tokens": 309604240.0,
+      "step": 2895
+    },
+    {
+      "epoch": 6.5974914481185865,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.424997409620295e-06,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.881415531039238,
+      "num_tokens": 309712035.0,
+      "step": 2896
+    },
+    {
+      "epoch": 6.599771949828963,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.4232972837714598e-06,
+      "loss": 0.5659,
+      "mean_token_accuracy": 0.8861576318740845,
+      "num_tokens": 309818871.0,
+      "step": 2897
+    },
+    {
+      "epoch": 6.602052451539339,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.4215977690382998e-06,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8870658576488495,
+      "num_tokens": 309926653.0,
+      "step": 2898
+    },
+    {
+      "epoch": 6.604332953249715,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.4198988663854276e-06,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8792040795087814,
+      "num_tokens": 310033580.0,
+      "step": 2899
+    },
+    {
+      "epoch": 6.606613454960091,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.4182005767771057e-06,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8829011768102646,
+      "num_tokens": 310141344.0,
+      "step": 2900
+    },
+    {
+      "epoch": 6.608893956670467,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.4165029011772513e-06,
+      "loss": 0.5765,
+      "mean_token_accuracy": 0.8812812268733978,
+      "num_tokens": 310247976.0,
+      "step": 2901
+    },
+    {
+      "epoch": 6.611174458380844,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.4148058405494328e-06,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8837906569242477,
+      "num_tokens": 310354966.0,
+      "step": 2902
+    },
+    {
+      "epoch": 6.61345496009122,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.4131093958568695e-06,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8823096752166748,
+      "num_tokens": 310462145.0,
+      "step": 2903
+    },
+    {
+      "epoch": 6.615735461801596,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.4114135680624291e-06,
+      "loss": 0.5919,
+      "mean_token_accuracy": 0.8795100599527359,
+      "num_tokens": 310568947.0,
+      "step": 2904
+    },
+    {
+      "epoch": 6.618015963511972,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.4097183581286322e-06,
+      "loss": 0.5776,
+      "mean_token_accuracy": 0.8784335255622864,
+      "num_tokens": 310675494.0,
+      "step": 2905
+    },
+    {
+      "epoch": 6.620296465222349,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.4080237670176456e-06,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.8832641243934631,
+      "num_tokens": 310782897.0,
+      "step": 2906
+    },
+    {
+      "epoch": 6.6225769669327255,
+      "grad_norm": 3.125,
+      "learning_rate": 1.4063297956912875e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8814590126276016,
+      "num_tokens": 310890161.0,
+      "step": 2907
+    },
+    {
+      "epoch": 6.624857468643102,
+      "grad_norm": 2.5,
+      "learning_rate": 1.4046364451110234e-06,
+      "loss": 0.5637,
+      "mean_token_accuracy": 0.884671688079834,
+      "num_tokens": 310996832.0,
+      "step": 2908
+    },
+    {
+      "epoch": 6.627137970353478,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.4029437162379666e-06,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.8783999532461166,
+      "num_tokens": 311103426.0,
+      "step": 2909
+    },
+    {
+      "epoch": 6.629418472063854,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.4012516100328766e-06,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8824923038482666,
+      "num_tokens": 311210237.0,
+      "step": 2910
+    },
+    {
+      "epoch": 6.63169897377423,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.3995601274561605e-06,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8831194937229156,
+      "num_tokens": 311317987.0,
+      "step": 2911
+    },
+    {
+      "epoch": 6.633979475484606,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.3978692694678711e-06,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.8810025453567505,
+      "num_tokens": 311424839.0,
+      "step": 2912
+    },
+    {
+      "epoch": 6.636259977194983,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.3961790370277068e-06,
+      "loss": 0.5655,
+      "mean_token_accuracy": 0.8853645473718643,
+      "num_tokens": 311531661.0,
+      "step": 2913
+    },
+    {
+      "epoch": 6.638540478905359,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.3944894310950113e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8820922076702118,
+      "num_tokens": 311639163.0,
+      "step": 2914
+    },
+    {
+      "epoch": 6.640820980615736,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.3928004526287729e-06,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8839680552482605,
+      "num_tokens": 311746011.0,
+      "step": 2915
+    },
+    {
+      "epoch": 6.643101482326112,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.3911121025876212e-06,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8828011155128479,
+      "num_tokens": 311852925.0,
+      "step": 2916
+    },
+    {
+      "epoch": 6.645381984036488,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.389424381929832e-06,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.8824039697647095,
+      "num_tokens": 311960025.0,
+      "step": 2917
+    },
+    {
+      "epoch": 6.6476624857468645,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.3877372916133234e-06,
+      "loss": 0.5964,
+      "mean_token_accuracy": 0.8741250783205032,
+      "num_tokens": 312066896.0,
+      "step": 2918
+    },
+    {
+      "epoch": 6.649942987457241,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.3860508325956549e-06,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.8746428936719894,
+      "num_tokens": 312173856.0,
+      "step": 2919
+    },
+    {
+      "epoch": 6.652223489167617,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.3843650058340291e-06,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8829982429742813,
+      "num_tokens": 312280996.0,
+      "step": 2920
+    },
+    {
+      "epoch": 6.654503990877993,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.382679812285287e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8866865336894989,
+      "num_tokens": 312388440.0,
+      "step": 2921
+    },
+    {
+      "epoch": 6.656784492588369,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.3809952529059127e-06,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8881336748600006,
+      "num_tokens": 312495421.0,
+      "step": 2922
+    },
+    {
+      "epoch": 6.659064994298745,
+      "grad_norm": 4.0,
+      "learning_rate": 1.3793113286520293e-06,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.884793609380722,
+      "num_tokens": 312602917.0,
+      "step": 2923
+    },
+    {
+      "epoch": 6.661345496009122,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.3776280404794016e-06,
+      "loss": 0.5775,
+      "mean_token_accuracy": 0.8788406997919083,
+      "num_tokens": 312710035.0,
+      "step": 2924
+    },
+    {
+      "epoch": 6.663625997719498,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.3759453893434285e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8826200515031815,
+      "num_tokens": 312816809.0,
+      "step": 2925
+    },
+    {
+      "epoch": 6.665906499429875,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.3742633761991519e-06,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8815308213233948,
+      "num_tokens": 312923988.0,
+      "step": 2926
+    },
+    {
+      "epoch": 6.668187001140251,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.3725820020012506e-06,
+      "loss": 0.5486,
+      "mean_token_accuracy": 0.8900953233242035,
+      "num_tokens": 313031446.0,
+      "step": 2927
+    },
+    {
+      "epoch": 6.670467502850627,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.3709012677040385e-06,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.882040336728096,
+      "num_tokens": 313138636.0,
+      "step": 2928
+    },
+    {
+      "epoch": 6.6727480045610035,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.3692211742614686e-06,
+      "loss": 0.5656,
+      "mean_token_accuracy": 0.8804366439580917,
+      "num_tokens": 313245263.0,
+      "step": 2929
+    },
+    {
+      "epoch": 6.67502850627138,
+      "grad_norm": 4.25,
+      "learning_rate": 1.3675417226271298e-06,
+      "loss": 0.5615,
+      "mean_token_accuracy": 0.8823003172874451,
+      "num_tokens": 313353027.0,
+      "step": 2930
+    },
+    {
+      "epoch": 6.677309007981756,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.365862913754247e-06,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8799808919429779,
+      "num_tokens": 313460650.0,
+      "step": 2931
+    },
+    {
+      "epoch": 6.679589509692132,
+      "grad_norm": 5.90625,
+      "learning_rate": 1.3641847485956782e-06,
+      "loss": 0.5846,
+      "mean_token_accuracy": 0.8758883625268936,
+      "num_tokens": 313567543.0,
+      "step": 2932
+    },
+    {
+      "epoch": 6.681870011402508,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.362507228103918e-06,
+      "loss": 0.5508,
+      "mean_token_accuracy": 0.8851123601198196,
+      "num_tokens": 313674854.0,
+      "step": 2933
+    },
+    {
+      "epoch": 6.684150513112884,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.3608303532310956e-06,
+      "loss": 0.5605,
+      "mean_token_accuracy": 0.884984016418457,
+      "num_tokens": 313781754.0,
+      "step": 2934
+    },
+    {
+      "epoch": 6.6864310148232615,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.3591541249289718e-06,
+      "loss": 0.5382,
+      "mean_token_accuracy": 0.8888740241527557,
+      "num_tokens": 313889331.0,
+      "step": 2935
+    },
+    {
+      "epoch": 6.688711516533638,
+      "grad_norm": 2.75,
+      "learning_rate": 1.357478544148943e-06,
+      "loss": 0.5517,
+      "mean_token_accuracy": 0.8864340782165527,
+      "num_tokens": 313997028.0,
+      "step": 2936
+    },
+    {
+      "epoch": 6.690992018244014,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.3558036118420343e-06,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8856192529201508,
+      "num_tokens": 314104052.0,
+      "step": 2937
+    },
+    {
+      "epoch": 6.69327251995439,
+      "grad_norm": 3.25,
+      "learning_rate": 1.3541293289589058e-06,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.882482260465622,
+      "num_tokens": 314211575.0,
+      "step": 2938
+    },
+    {
+      "epoch": 6.695553021664766,
+      "grad_norm": 4.125,
+      "learning_rate": 1.3524556964498482e-06,
+      "loss": 0.5554,
+      "mean_token_accuracy": 0.88676318526268,
+      "num_tokens": 314318790.0,
+      "step": 2939
+    },
+    {
+      "epoch": 6.6978335233751425,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.3507827152647835e-06,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8853241056203842,
+      "num_tokens": 314425965.0,
+      "step": 2940
+    },
+    {
+      "epoch": 6.700114025085519,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.3491103863532626e-06,
+      "loss": 0.5577,
+      "mean_token_accuracy": 0.8847237229347229,
+      "num_tokens": 314533161.0,
+      "step": 2941
+    },
+    {
+      "epoch": 6.702394526795895,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.3474387106644688e-06,
+      "loss": 0.5874,
+      "mean_token_accuracy": 0.8763073831796646,
+      "num_tokens": 314640305.0,
+      "step": 2942
+    },
+    {
+      "epoch": 6.704675028506271,
+      "grad_norm": 4.5,
+      "learning_rate": 1.345767689147211e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8827333003282547,
+      "num_tokens": 314747484.0,
+      "step": 2943
+    },
+    {
+      "epoch": 6.706955530216648,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.3440973227499293e-06,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.8817901015281677,
+      "num_tokens": 314854356.0,
+      "step": 2944
+    },
+    {
+      "epoch": 6.7092360319270234,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.3424276124206917e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8832406252622604,
+      "num_tokens": 314961203.0,
+      "step": 2945
+    },
+    {
+      "epoch": 6.7115165336374005,
+      "grad_norm": 3.0,
+      "learning_rate": 1.3407585591071944e-06,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8804884552955627,
+      "num_tokens": 315067663.0,
+      "step": 2946
+    },
+    {
+      "epoch": 6.713797035347777,
+      "grad_norm": 6.0,
+      "learning_rate": 1.3390901637567579e-06,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8794064670801163,
+      "num_tokens": 315175060.0,
+      "step": 2947
+    },
+    {
+      "epoch": 6.716077537058153,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.3374224273163334e-06,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8783332705497742,
+      "num_tokens": 315281889.0,
+      "step": 2948
+    },
+    {
+      "epoch": 6.718358038768529,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.3357553507324938e-06,
+      "loss": 0.5681,
+      "mean_token_accuracy": 0.8839272111654282,
+      "num_tokens": 315388559.0,
+      "step": 2949
+    },
+    {
+      "epoch": 6.720638540478905,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.3340889349514403e-06,
+      "loss": 0.5636,
+      "mean_token_accuracy": 0.8809895366430283,
+      "num_tokens": 315495344.0,
+      "step": 2950
+    },
+    {
+      "epoch": 6.7229190421892815,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.3324231809189985e-06,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8813484013080597,
+      "num_tokens": 315602116.0,
+      "step": 2951
+    },
+    {
+      "epoch": 6.725199543899658,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.3307580895806194e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8836003094911575,
+      "num_tokens": 315709172.0,
+      "step": 2952
+    },
+    {
+      "epoch": 6.727480045610034,
+      "grad_norm": 3.375,
+      "learning_rate": 1.3290936618813747e-06,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8845668733119965,
+      "num_tokens": 315816102.0,
+      "step": 2953
+    },
+    {
+      "epoch": 6.72976054732041,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.327429898765962e-06,
+      "loss": 0.5712,
+      "mean_token_accuracy": 0.8821906894445419,
+      "num_tokens": 315923130.0,
+      "step": 2954
+    },
+    {
+      "epoch": 6.732041049030787,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.3257668011787018e-06,
+      "loss": 0.5899,
+      "mean_token_accuracy": 0.8804189115762711,
+      "num_tokens": 316030172.0,
+      "step": 2955
+    },
+    {
+      "epoch": 6.734321550741163,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.3241043700635352e-06,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.8796321153640747,
+      "num_tokens": 316136895.0,
+      "step": 2956
+    },
+    {
+      "epoch": 6.7366020524515395,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.3224426063640272e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8832730054855347,
+      "num_tokens": 316243931.0,
+      "step": 2957
+    },
+    {
+      "epoch": 6.738882554161916,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.320781511023363e-06,
+      "loss": 0.5692,
+      "mean_token_accuracy": 0.883409783244133,
+      "num_tokens": 316351625.0,
+      "step": 2958
+    },
+    {
+      "epoch": 6.741163055872292,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.3191210849843461e-06,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8800751566886902,
+      "num_tokens": 316459647.0,
+      "step": 2959
+    },
+    {
+      "epoch": 6.743443557582668,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.3174613291894039e-06,
+      "loss": 0.5953,
+      "mean_token_accuracy": 0.8763148337602615,
+      "num_tokens": 316566718.0,
+      "step": 2960
+    },
+    {
+      "epoch": 6.745724059293044,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.3158022445805816e-06,
+      "loss": 0.5492,
+      "mean_token_accuracy": 0.8837677389383316,
+      "num_tokens": 316673847.0,
+      "step": 2961
+    },
+    {
+      "epoch": 6.7480045610034205,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.3141438320995433e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8806140273809433,
+      "num_tokens": 316780635.0,
+      "step": 2962
+    },
+    {
+      "epoch": 6.750285062713797,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.3124860926875732e-06,
+      "loss": 0.5505,
+      "mean_token_accuracy": 0.8896244168281555,
+      "num_tokens": 316887938.0,
+      "step": 2963
+    },
+    {
+      "epoch": 6.752565564424174,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.3108290272855697e-06,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.8813838809728622,
+      "num_tokens": 316995335.0,
+      "step": 2964
+    },
+    {
+      "epoch": 6.75484606613455,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.309172636834053e-06,
+      "loss": 0.5642,
+      "mean_token_accuracy": 0.8819121271371841,
+      "num_tokens": 317102837.0,
+      "step": 2965
+    },
+    {
+      "epoch": 6.757126567844926,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.3075169222731573e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8826228231191635,
+      "num_tokens": 317210216.0,
+      "step": 2966
+    },
+    {
+      "epoch": 6.759407069555302,
+      "grad_norm": 3.5,
+      "learning_rate": 1.305861884542636e-06,
+      "loss": 0.5793,
+      "mean_token_accuracy": 0.8787815272808075,
+      "num_tokens": 317317069.0,
+      "step": 2967
+    },
+    {
+      "epoch": 6.7616875712656785,
+      "grad_norm": 4.875,
+      "learning_rate": 1.3042075245818542e-06,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8827643990516663,
+      "num_tokens": 317424086.0,
+      "step": 2968
+    },
+    {
+      "epoch": 6.763968072976055,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.3025538433297957e-06,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8830547332763672,
+      "num_tokens": 317531744.0,
+      "step": 2969
+    },
+    {
+      "epoch": 6.766248574686431,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.3009008417250597e-06,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.883535161614418,
+      "num_tokens": 317638896.0,
+      "step": 2970
+    },
+    {
+      "epoch": 6.768529076396807,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.2992485207058548e-06,
+      "loss": 0.5958,
+      "mean_token_accuracy": 0.8761639446020126,
+      "num_tokens": 317745636.0,
+      "step": 2971
+    },
+    {
+      "epoch": 6.770809578107183,
+      "grad_norm": 2.625,
+      "learning_rate": 1.2975968812100081e-06,
+      "loss": 0.5413,
+      "mean_token_accuracy": 0.8862400203943253,
+      "num_tokens": 317852822.0,
+      "step": 2972
+    },
+    {
+      "epoch": 6.7730900798175595,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.295945924174959e-06,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8786645531654358,
+      "num_tokens": 317959386.0,
+      "step": 2973
+    },
+    {
+      "epoch": 6.775370581527936,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.2942956505377585e-06,
+      "loss": 0.565,
+      "mean_token_accuracy": 0.8845228105783463,
+      "num_tokens": 318066201.0,
+      "step": 2974
+    },
+    {
+      "epoch": 6.777651083238313,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.2926460612350688e-06,
+      "loss": 0.5577,
+      "mean_token_accuracy": 0.883774995803833,
+      "num_tokens": 318173322.0,
+      "step": 2975
+    },
+    {
+      "epoch": 6.779931584948689,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.2909971572031663e-06,
+      "loss": 0.558,
+      "mean_token_accuracy": 0.8879963010549545,
+      "num_tokens": 318281003.0,
+      "step": 2976
+    },
+    {
+      "epoch": 6.782212086659065,
+      "grad_norm": 3.375,
+      "learning_rate": 1.2893489393779362e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8839980363845825,
+      "num_tokens": 318387942.0,
+      "step": 2977
+    },
+    {
+      "epoch": 6.784492588369441,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.2877014086948762e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8807279914617538,
+      "num_tokens": 318494877.0,
+      "step": 2978
+    },
+    {
+      "epoch": 6.7867730900798175,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.2860545660890928e-06,
+      "loss": 0.5537,
+      "mean_token_accuracy": 0.8867141306400299,
+      "num_tokens": 318601668.0,
+      "step": 2979
+    },
+    {
+      "epoch": 6.789053591790194,
+      "grad_norm": 2.34375,
+      "learning_rate": 1.2844084124953006e-06,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8846204280853271,
+      "num_tokens": 318708916.0,
+      "step": 2980
+    },
+    {
+      "epoch": 6.79133409350057,
+      "grad_norm": 5.25,
+      "learning_rate": 1.2827629488478254e-06,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8803286552429199,
+      "num_tokens": 318816351.0,
+      "step": 2981
+    },
+    {
+      "epoch": 6.793614595210946,
+      "grad_norm": 3.25,
+      "learning_rate": 1.2811181760806013e-06,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8851221948862076,
+      "num_tokens": 318923078.0,
+      "step": 2982
+    },
+    {
+      "epoch": 6.795895096921322,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.2794740951271686e-06,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8797405809164047,
+      "num_tokens": 319029672.0,
+      "step": 2983
+    },
+    {
+      "epoch": 6.798175598631699,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.2778307069206764e-06,
+      "loss": 0.5599,
+      "mean_token_accuracy": 0.8823461681604385,
+      "num_tokens": 319136574.0,
+      "step": 2984
+    },
+    {
+      "epoch": 6.800456100342076,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.2761880123938814e-06,
+      "loss": 0.5618,
+      "mean_token_accuracy": 0.8868078589439392,
+      "num_tokens": 319244374.0,
+      "step": 2985
+    },
+    {
+      "epoch": 6.802736602052452,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.2745460124791425e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8795672506093979,
+      "num_tokens": 319352045.0,
+      "step": 2986
+    },
+    {
+      "epoch": 6.805017103762828,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.272904708108429e-06,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8827357441186905,
+      "num_tokens": 319458393.0,
+      "step": 2987
+    },
+    {
+      "epoch": 6.807297605473204,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.2712641002133128e-06,
+      "loss": 0.6055,
+      "mean_token_accuracy": 0.8749971687793732,
+      "num_tokens": 319564852.0,
+      "step": 2988
+    },
+    {
+      "epoch": 6.80957810718358,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.2696241897249728e-06,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8824810981750488,
+      "num_tokens": 319671606.0,
+      "step": 2989
+    },
+    {
+      "epoch": 6.811858608893957,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.2679849775741884e-06,
+      "loss": 0.5606,
+      "mean_token_accuracy": 0.8874455839395523,
+      "num_tokens": 319778774.0,
+      "step": 2990
+    },
+    {
+      "epoch": 6.814139110604333,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.266346464691346e-06,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8811825066804886,
+      "num_tokens": 319885624.0,
+      "step": 2991
+    },
+    {
+      "epoch": 6.816419612314709,
+      "grad_norm": 3.25,
+      "learning_rate": 1.2647086520064343e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.880624532699585,
+      "num_tokens": 319992897.0,
+      "step": 2992
+    },
+    {
+      "epoch": 6.818700114025085,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.2630715404490424e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8810169398784637,
+      "num_tokens": 320099665.0,
+      "step": 2993
+    },
+    {
+      "epoch": 6.820980615735461,
+      "grad_norm": 2.625,
+      "learning_rate": 1.2614351309483646e-06,
+      "loss": 0.5474,
+      "mean_token_accuracy": 0.8881228417158127,
+      "num_tokens": 320207542.0,
+      "step": 2994
+    },
+    {
+      "epoch": 6.823261117445838,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.259799424433196e-06,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8811034262180328,
+      "num_tokens": 320314111.0,
+      "step": 2995
+    },
+    {
+      "epoch": 6.825541619156215,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.25816442183193e-06,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8809540122747421,
+      "num_tokens": 320421117.0,
+      "step": 2996
+    },
+    {
+      "epoch": 6.827822120866591,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.2565301240725636e-06,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8778086453676224,
+      "num_tokens": 320527962.0,
+      "step": 2997
+    },
+    {
+      "epoch": 6.830102622576967,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.2548965320826928e-06,
+      "loss": 0.5667,
+      "mean_token_accuracy": 0.883897066116333,
+      "num_tokens": 320634881.0,
+      "step": 2998
+    },
+    {
+      "epoch": 6.832383124287343,
+      "grad_norm": 3.0,
+      "learning_rate": 1.2532636467895126e-06,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8828509300947189,
+      "num_tokens": 320741780.0,
+      "step": 2999
+    },
+    {
+      "epoch": 6.834663625997719,
+      "grad_norm": 3.375,
+      "learning_rate": 1.2516314691198172e-06,
+      "loss": 0.5606,
+      "mean_token_accuracy": 0.8836653828620911,
+      "num_tokens": 320849047.0,
+      "step": 3000
+    },
+    {
+      "epoch": 6.836944127708096,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.2500000000000007e-06,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8793622553348541,
+      "num_tokens": 320955644.0,
+      "step": 3001
+    },
+    {
+      "epoch": 6.839224629418472,
+      "grad_norm": 3.75,
+      "learning_rate": 1.2483692403560507e-06,
+      "loss": 0.5484,
+      "mean_token_accuracy": 0.8884700238704681,
+      "num_tokens": 321063849.0,
+      "step": 3002
+    },
+    {
+      "epoch": 6.841505131128848,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.2467391911135562e-06,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8834027796983719,
+      "num_tokens": 321170372.0,
+      "step": 3003
+    },
+    {
+      "epoch": 6.843785632839225,
+      "grad_norm": 4.0,
+      "learning_rate": 1.2451098531977015e-06,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.8847081363201141,
+      "num_tokens": 321277858.0,
+      "step": 3004
+    },
+    {
+      "epoch": 6.846066134549601,
+      "grad_norm": 2.875,
+      "learning_rate": 1.2434812275332678e-06,
+      "loss": 0.536,
+      "mean_token_accuracy": 0.8904493749141693,
+      "num_tokens": 321384888.0,
+      "step": 3005
+    },
+    {
+      "epoch": 6.848346636259977,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.2418533150446324e-06,
+      "loss": 0.5543,
+      "mean_token_accuracy": 0.8845214545726776,
+      "num_tokens": 321491996.0,
+      "step": 3006
+    },
+    {
+      "epoch": 6.850627137970354,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.2402261166557647e-06,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.8830210715532303,
+      "num_tokens": 321598410.0,
+      "step": 3007
+    },
+    {
+      "epoch": 6.85290763968073,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.2385996332902326e-06,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8822718113660812,
+      "num_tokens": 321705299.0,
+      "step": 3008
+    },
+    {
+      "epoch": 6.855188141391106,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.236973865871196e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8847811222076416,
+      "num_tokens": 321811999.0,
+      "step": 3009
+    },
+    {
+      "epoch": 6.857468643101482,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.2353488153214096e-06,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8817588984966278,
+      "num_tokens": 321919613.0,
+      "step": 3010
+    },
+    {
+      "epoch": 6.859749144811858,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.2337244825632217e-06,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8837725669145584,
+      "num_tokens": 322026837.0,
+      "step": 3011
+    },
+    {
+      "epoch": 6.862029646522235,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.2321008685185699e-06,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8756814897060394,
+      "num_tokens": 322132737.0,
+      "step": 3012
+    },
+    {
+      "epoch": 6.864310148232612,
+      "grad_norm": 2.46875,
+      "learning_rate": 1.2304779741089884e-06,
+      "loss": 0.559,
+      "mean_token_accuracy": 0.8843032121658325,
+      "num_tokens": 322239975.0,
+      "step": 3013
+    },
+    {
+      "epoch": 6.866590649942988,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.228855800255599e-06,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8782118856906891,
+      "num_tokens": 322346564.0,
+      "step": 3014
+    },
+    {
+      "epoch": 6.868871151653364,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.2272343478791165e-06,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.8836709409952164,
+      "num_tokens": 322453680.0,
+      "step": 3015
+    },
+    {
+      "epoch": 6.87115165336374,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.2256136178998468e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8812815845012665,
+      "num_tokens": 322561481.0,
+      "step": 3016
+    },
+    {
+      "epoch": 6.873432155074116,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.2239936112376858e-06,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8811852186918259,
+      "num_tokens": 322668877.0,
+      "step": 3017
+    },
+    {
+      "epoch": 6.875712656784493,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.2223743288121155e-06,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8837666660547256,
+      "num_tokens": 322775613.0,
+      "step": 3018
+    },
+    {
+      "epoch": 6.877993158494869,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.2207557715422106e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8822371959686279,
+      "num_tokens": 322883149.0,
+      "step": 3019
+    },
+    {
+      "epoch": 6.880273660205245,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.219137940346633e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.880626305937767,
+      "num_tokens": 322989979.0,
+      "step": 3020
+    },
+    {
+      "epoch": 6.882554161915621,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.2175208361436328e-06,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8829739540815353,
+      "num_tokens": 323096950.0,
+      "step": 3021
+    },
+    {
+      "epoch": 6.884834663625997,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.2159044598510473e-06,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8852163255214691,
+      "num_tokens": 323204095.0,
+      "step": 3022
+    },
+    {
+      "epoch": 6.887115165336374,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.2142888123862992e-06,
+      "loss": 0.5789,
+      "mean_token_accuracy": 0.8801557719707489,
+      "num_tokens": 323311535.0,
+      "step": 3023
+    },
+    {
+      "epoch": 6.889395667046751,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.2126738946663996e-06,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8806950151920319,
+      "num_tokens": 323418174.0,
+      "step": 3024
+    },
+    {
+      "epoch": 6.891676168757127,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.2110597076079448e-06,
+      "loss": 0.5536,
+      "mean_token_accuracy": 0.8857726603746414,
+      "num_tokens": 323524933.0,
+      "step": 3025
+    },
+    {
+      "epoch": 6.893956670467503,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.2094462521271156e-06,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8852896243333817,
+      "num_tokens": 323631823.0,
+      "step": 3026
+    },
+    {
+      "epoch": 6.896237172177879,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.2078335291396798e-06,
+      "loss": 0.5733,
+      "mean_token_accuracy": 0.8801656812429428,
+      "num_tokens": 323738919.0,
+      "step": 3027
+    },
+    {
+      "epoch": 6.898517673888255,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.2062215395609856e-06,
+      "loss": 0.5623,
+      "mean_token_accuracy": 0.8838664293289185,
+      "num_tokens": 323845761.0,
+      "step": 3028
+    },
+    {
+      "epoch": 6.900798175598632,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.2046102843059681e-06,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8809877783060074,
+      "num_tokens": 323952934.0,
+      "step": 3029
+    },
+    {
+      "epoch": 6.903078677309008,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.202999764289145e-06,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8807922154664993,
+      "num_tokens": 324060199.0,
+      "step": 3030
+    },
+    {
+      "epoch": 6.905359179019384,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.201389980424616e-06,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8789155036211014,
+      "num_tokens": 324166935.0,
+      "step": 3031
+    },
+    {
+      "epoch": 6.90763968072976,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.1997809336260644e-06,
+      "loss": 0.5569,
+      "mean_token_accuracy": 0.8868138492107391,
+      "num_tokens": 324273307.0,
+      "step": 3032
+    },
+    {
+      "epoch": 6.909920182440137,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.1981726248067521e-06,
+      "loss": 0.5478,
+      "mean_token_accuracy": 0.8851444870233536,
+      "num_tokens": 324380120.0,
+      "step": 3033
+    },
+    {
+      "epoch": 6.9122006841505135,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.1965650548795251e-06,
+      "loss": 0.5511,
+      "mean_token_accuracy": 0.8851585388183594,
+      "num_tokens": 324487235.0,
+      "step": 3034
+    },
+    {
+      "epoch": 6.91448118586089,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.1949582247568107e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8827362954616547,
+      "num_tokens": 324594111.0,
+      "step": 3035
+    },
+    {
+      "epoch": 6.916761687571266,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.1933521353506117e-06,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8829712569713593,
+      "num_tokens": 324701818.0,
+      "step": 3036
+    },
+    {
+      "epoch": 6.919042189281642,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.1917467875725148e-06,
+      "loss": 0.5528,
+      "mean_token_accuracy": 0.8853205442428589,
+      "num_tokens": 324808789.0,
+      "step": 3037
+    },
+    {
+      "epoch": 6.921322690992018,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.1901421823336856e-06,
+      "loss": 0.5456,
+      "mean_token_accuracy": 0.885327011346817,
+      "num_tokens": 324916062.0,
+      "step": 3038
+    },
+    {
+      "epoch": 6.923603192702394,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.188538320544865e-06,
+      "loss": 0.5632,
+      "mean_token_accuracy": 0.880491703748703,
+      "num_tokens": 325023122.0,
+      "step": 3039
+    },
+    {
+      "epoch": 6.925883694412771,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.1869352031163746e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.880371481180191,
+      "num_tokens": 325130007.0,
+      "step": 3040
+    },
+    {
+      "epoch": 6.928164196123147,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.1853328309581139e-06,
+      "loss": 0.5676,
+      "mean_token_accuracy": 0.884147435426712,
+      "num_tokens": 325236627.0,
+      "step": 3041
+    },
+    {
+      "epoch": 6.930444697833523,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.183731204979557e-06,
+      "loss": 0.5679,
+      "mean_token_accuracy": 0.8821654170751572,
+      "num_tokens": 325344292.0,
+      "step": 3042
+    },
+    {
+      "epoch": 6.932725199543899,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.182130326089758e-06,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8848479390144348,
+      "num_tokens": 325451752.0,
+      "step": 3043
+    },
+    {
+      "epoch": 6.935005701254276,
+      "grad_norm": 4.71875,
+      "learning_rate": 1.1805301951973423e-06,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8822412639856339,
+      "num_tokens": 325558775.0,
+      "step": 3044
+    },
+    {
+      "epoch": 6.9372862029646525,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.1789308132105145e-06,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8826945275068283,
+      "num_tokens": 325665451.0,
+      "step": 3045
+    },
+    {
+      "epoch": 6.939566704675029,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.1773321810370527e-06,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8854436427354813,
+      "num_tokens": 325772180.0,
+      "step": 3046
+    },
+    {
+      "epoch": 6.941847206385405,
+      "grad_norm": 2.515625,
+      "learning_rate": 1.1757342995843103e-06,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8795358538627625,
+      "num_tokens": 325879003.0,
+      "step": 3047
+    },
+    {
+      "epoch": 6.944127708095781,
+      "grad_norm": 3.125,
+      "learning_rate": 1.1741371697592134e-06,
+      "loss": 0.5642,
+      "mean_token_accuracy": 0.8838000446557999,
+      "num_tokens": 325986026.0,
+      "step": 3048
+    },
+    {
+      "epoch": 6.946408209806157,
+      "grad_norm": 3.625,
+      "learning_rate": 1.1725407924682628e-06,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8830106258392334,
+      "num_tokens": 326093663.0,
+      "step": 3049
+    },
+    {
+      "epoch": 6.9486887115165334,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.17094516861753e-06,
+      "loss": 0.5393,
+      "mean_token_accuracy": 0.8868276029825211,
+      "num_tokens": 326201499.0,
+      "step": 3050
+    },
+    {
+      "epoch": 6.95096921322691,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.1693502991126609e-06,
+      "loss": 0.5703,
+      "mean_token_accuracy": 0.8848345726728439,
+      "num_tokens": 326308543.0,
+      "step": 3051
+    },
+    {
+      "epoch": 6.953249714937286,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.1677561848588734e-06,
+      "loss": 0.5626,
+      "mean_token_accuracy": 0.8820251226425171,
+      "num_tokens": 326415635.0,
+      "step": 3052
+    },
+    {
+      "epoch": 6.955530216647663,
+      "grad_norm": 3.375,
+      "learning_rate": 1.166162826760955e-06,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8849399089813232,
+      "num_tokens": 326523279.0,
+      "step": 3053
+    },
+    {
+      "epoch": 6.957810718358039,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.1645702257232663e-06,
+      "loss": 0.5785,
+      "mean_token_accuracy": 0.8775086104869843,
+      "num_tokens": 326630109.0,
+      "step": 3054
+    },
+    {
+      "epoch": 6.960091220068415,
+      "grad_norm": 3.0,
+      "learning_rate": 1.1629783826497351e-06,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8805986046791077,
+      "num_tokens": 326737084.0,
+      "step": 3055
+    },
+    {
+      "epoch": 6.9623717217787915,
+      "grad_norm": 2.625,
+      "learning_rate": 1.161387298443863e-06,
+      "loss": 0.5631,
+      "mean_token_accuracy": 0.8876520395278931,
+      "num_tokens": 326844156.0,
+      "step": 3056
+    },
+    {
+      "epoch": 6.964652223489168,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.1597969740087159e-06,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8821894228458405,
+      "num_tokens": 326951005.0,
+      "step": 3057
+    },
+    {
+      "epoch": 6.966932725199544,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.1582074102469332e-06,
+      "loss": 0.5554,
+      "mean_token_accuracy": 0.8848588168621063,
+      "num_tokens": 327058614.0,
+      "step": 3058
+    },
+    {
+      "epoch": 6.96921322690992,
+      "grad_norm": 3.0,
+      "learning_rate": 1.1566186080607198e-06,
+      "loss": 0.5495,
+      "mean_token_accuracy": 0.8877767473459244,
+      "num_tokens": 327166015.0,
+      "step": 3059
+    },
+    {
+      "epoch": 6.971493728620296,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.1550305683518506e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8813445121049881,
+      "num_tokens": 327273077.0,
+      "step": 3060
+    },
+    {
+      "epoch": 6.9737742303306725,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.1534432920216643e-06,
+      "loss": 0.5984,
+      "mean_token_accuracy": 0.8741457164287567,
+      "num_tokens": 327379556.0,
+      "step": 3061
+    },
+    {
+      "epoch": 6.976054732041049,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.151856779971069e-06,
+      "loss": 0.5636,
+      "mean_token_accuracy": 0.8825311958789825,
+      "num_tokens": 327486300.0,
+      "step": 3062
+    },
+    {
+      "epoch": 6.978335233751425,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.1502710331005384e-06,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.8804716914892197,
+      "num_tokens": 327593244.0,
+      "step": 3063
+    },
+    {
+      "epoch": 6.980615735461802,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.148686052310112e-06,
+      "loss": 0.596,
+      "mean_token_accuracy": 0.875918909907341,
+      "num_tokens": 327699724.0,
+      "step": 3064
+    },
+    {
+      "epoch": 6.982896237172178,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.147101838499395e-06,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8763432502746582,
+      "num_tokens": 327806813.0,
+      "step": 3065
+    },
+    {
+      "epoch": 6.985176738882554,
+      "grad_norm": 2.875,
+      "learning_rate": 1.145518392567555e-06,
+      "loss": 0.5514,
+      "mean_token_accuracy": 0.8882738053798676,
+      "num_tokens": 327913705.0,
+      "step": 3066
+    },
+    {
+      "epoch": 6.9874572405929305,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.1439357154133263e-06,
+      "loss": 0.5451,
+      "mean_token_accuracy": 0.8874611854553223,
+      "num_tokens": 328021868.0,
+      "step": 3067
+    },
+    {
+      "epoch": 6.989737742303307,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.1423538079350053e-06,
+      "loss": 0.5511,
+      "mean_token_accuracy": 0.8833891451358795,
+      "num_tokens": 328129123.0,
+      "step": 3068
+    },
+    {
+      "epoch": 6.992018244013683,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.1407726710304525e-06,
+      "loss": 0.6043,
+      "mean_token_accuracy": 0.8766579329967499,
+      "num_tokens": 328235771.0,
+      "step": 3069
+    },
+    {
+      "epoch": 6.994298745724059,
+      "grad_norm": 4.90625,
+      "learning_rate": 1.139192305597092e-06,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8802380710840225,
+      "num_tokens": 328342486.0,
+      "step": 3070
+    },
+    {
+      "epoch": 6.996579247434435,
+      "grad_norm": 3.625,
+      "learning_rate": 1.1376127125319065e-06,
+      "loss": 0.5908,
+      "mean_token_accuracy": 0.8793238997459412,
+      "num_tokens": 328449792.0,
+      "step": 3071
+    },
+    {
+      "epoch": 6.9988597491448115,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.1360338927314432e-06,
+      "loss": 0.5572,
+      "mean_token_accuracy": 0.8853235244750977,
+      "num_tokens": 328556382.0,
+      "step": 3072
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 5.625,
+      "learning_rate": 1.1344558470918098e-06,
+      "loss": 0.6197,
+      "mean_token_accuracy": 0.8733022511005402,
+      "num_tokens": 328595624.0,
+      "step": 3073
+    },
+    {
+      "epoch": 7.002280501710376,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.1328785765086752e-06,
+      "loss": 0.5623,
+      "mean_token_accuracy": 0.8809218257665634,
+      "num_tokens": 328702654.0,
+      "step": 3074
+    },
+    {
+      "epoch": 7.004561003420752,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.131302081877268e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8839420080184937,
+      "num_tokens": 328809480.0,
+      "step": 3075
+    },
+    {
+      "epoch": 7.006841505131129,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.1297263640923745e-06,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8816477954387665,
+      "num_tokens": 328916070.0,
+      "step": 3076
+    },
+    {
+      "epoch": 7.009122006841505,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.1281514240483427e-06,
+      "loss": 0.5495,
+      "mean_token_accuracy": 0.8879576325416565,
+      "num_tokens": 329022615.0,
+      "step": 3077
+    },
+    {
+      "epoch": 7.011402508551882,
+      "grad_norm": 3.875,
+      "learning_rate": 1.1265772626390786e-06,
+      "loss": 0.5521,
+      "mean_token_accuracy": 0.888176754117012,
+      "num_tokens": 329129873.0,
+      "step": 3078
+    },
+    {
+      "epoch": 7.013683010262258,
+      "grad_norm": 3.0,
+      "learning_rate": 1.1250038807580449e-06,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8815352469682693,
+      "num_tokens": 329236596.0,
+      "step": 3079
+    },
+    {
+      "epoch": 7.015963511972634,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.1234312792982627e-06,
+      "loss": 0.5557,
+      "mean_token_accuracy": 0.8854427635669708,
+      "num_tokens": 329344039.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.015963511972634,
+      "eval_loss": 0.5865316987037659,
+      "eval_mean_token_accuracy": 0.8799761234580791,
+      "eval_num_tokens": 329344039.0,
+      "eval_runtime": 58.6794,
+      "eval_samples_per_second": 142.895,
+      "eval_steps_per_second": 4.482,
+      "step": 3080
+    },
+    {
+      "epoch": 7.01824401368301,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.1218594591523118e-06,
+      "loss": 0.5583,
+      "mean_token_accuracy": 0.8825812339782715,
+      "num_tokens": 329451249.0,
+      "step": 3081
+    },
+    {
+      "epoch": 7.020524515393387,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.120288421212325e-06,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8850286900997162,
+      "num_tokens": 329558258.0,
+      "step": 3082
+    },
+    {
+      "epoch": 7.022805017103763,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.1187181663699935e-06,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8803520053625107,
+      "num_tokens": 329665648.0,
+      "step": 3083
+    },
+    {
+      "epoch": 7.025085518814139,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.1171486955165645e-06,
+      "loss": 0.5544,
+      "mean_token_accuracy": 0.886691614985466,
+      "num_tokens": 329772942.0,
+      "step": 3084
+    },
+    {
+      "epoch": 7.027366020524515,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.115580009542839e-06,
+      "loss": 0.5551,
+      "mean_token_accuracy": 0.8843899071216583,
+      "num_tokens": 329880278.0,
+      "step": 3085
+    },
+    {
+      "epoch": 7.029646522234891,
+      "grad_norm": 2.875,
+      "learning_rate": 1.1140121093391736e-06,
+      "loss": 0.5499,
+      "mean_token_accuracy": 0.8839717209339142,
+      "num_tokens": 329987716.0,
+      "step": 3086
+    },
+    {
+      "epoch": 7.031927023945268,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.1124449957954764e-06,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8822407573461533,
+      "num_tokens": 330094855.0,
+      "step": 3087
+    },
+    {
+      "epoch": 7.034207525655645,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.110878669801212e-06,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.8854365050792694,
+      "num_tokens": 330202300.0,
+      "step": 3088
+    },
+    {
+      "epoch": 7.036488027366021,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.1093131322453966e-06,
+      "loss": 0.562,
+      "mean_token_accuracy": 0.8828160762786865,
+      "num_tokens": 330309184.0,
+      "step": 3089
+    },
+    {
+      "epoch": 7.038768529076397,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.1077483840165986e-06,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8839673697948456,
+      "num_tokens": 330415860.0,
+      "step": 3090
+    },
+    {
+      "epoch": 7.041049030786773,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.10618442600294e-06,
+      "loss": 0.5521,
+      "mean_token_accuracy": 0.8864448517560959,
+      "num_tokens": 330522700.0,
+      "step": 3091
+    },
+    {
+      "epoch": 7.043329532497149,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.1046212590920931e-06,
+      "loss": 0.55,
+      "mean_token_accuracy": 0.8861220180988312,
+      "num_tokens": 330630445.0,
+      "step": 3092
+    },
+    {
+      "epoch": 7.045610034207526,
+      "grad_norm": 3.25,
+      "learning_rate": 1.10305888417128e-06,
+      "loss": 0.5674,
+      "mean_token_accuracy": 0.8770045042037964,
+      "num_tokens": 330737176.0,
+      "step": 3093
+    },
+    {
+      "epoch": 7.047890535917902,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.101497302127275e-06,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.880455732345581,
+      "num_tokens": 330843523.0,
+      "step": 3094
+    },
+    {
+      "epoch": 7.050171037628278,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.0999365138464024e-06,
+      "loss": 0.5503,
+      "mean_token_accuracy": 0.8857882469892502,
+      "num_tokens": 330950912.0,
+      "step": 3095
+    },
+    {
+      "epoch": 7.052451539338654,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.0983765202145351e-06,
+      "loss": 0.5694,
+      "mean_token_accuracy": 0.8827014863491058,
+      "num_tokens": 331057394.0,
+      "step": 3096
+    },
+    {
+      "epoch": 7.05473204104903,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.0968173221170966e-06,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.88177290558815,
+      "num_tokens": 331164519.0,
+      "step": 3097
+    },
+    {
+      "epoch": 7.0570125427594075,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.0952589204390557e-06,
+      "loss": 0.5574,
+      "mean_token_accuracy": 0.8831107765436172,
+      "num_tokens": 331272101.0,
+      "step": 3098
+    },
+    {
+      "epoch": 7.059293044469784,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.0937013160649328e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8797809779644012,
+      "num_tokens": 331379251.0,
+      "step": 3099
+    },
+    {
+      "epoch": 7.06157354618016,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.0921445098787923e-06,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8834618926048279,
+      "num_tokens": 331486795.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.063854047890536,
+      "grad_norm": 3.375,
+      "learning_rate": 1.0905885027642484e-06,
+      "loss": 0.5577,
+      "mean_token_accuracy": 0.882593035697937,
+      "num_tokens": 331593690.0,
+      "step": 3101
+    },
+    {
+      "epoch": 7.066134549600912,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.0890332956044614e-06,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8840242773294449,
+      "num_tokens": 331700378.0,
+      "step": 3102
+    },
+    {
+      "epoch": 7.068415051311288,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.0874788892821354e-06,
+      "loss": 0.5603,
+      "mean_token_accuracy": 0.886967346072197,
+      "num_tokens": 331807432.0,
+      "step": 3103
+    },
+    {
+      "epoch": 7.070695553021665,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.0859252846795215e-06,
+      "loss": 0.5571,
+      "mean_token_accuracy": 0.8852237462997437,
+      "num_tokens": 331914837.0,
+      "step": 3104
+    },
+    {
+      "epoch": 7.072976054732041,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.0843724826784165e-06,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8843671381473541,
+      "num_tokens": 332021366.0,
+      "step": 3105
+    },
+    {
+      "epoch": 7.075256556442417,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.0828204841601608e-06,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8801092207431793,
+      "num_tokens": 332128357.0,
+      "step": 3106
+    },
+    {
+      "epoch": 7.077537058152793,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.0812692900056384e-06,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8785477876663208,
+      "num_tokens": 332235371.0,
+      "step": 3107
+    },
+    {
+      "epoch": 7.07981755986317,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.0797189010952784e-06,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8777376860380173,
+      "num_tokens": 332342307.0,
+      "step": 3108
+    },
+    {
+      "epoch": 7.0820980615735465,
+      "grad_norm": 4.125,
+      "learning_rate": 1.0781693183090495e-06,
+      "loss": 0.5818,
+      "mean_token_accuracy": 0.8822515457868576,
+      "num_tokens": 332449426.0,
+      "step": 3109
+    },
+    {
+      "epoch": 7.084378563283923,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.076620542526466e-06,
+      "loss": 0.5588,
+      "mean_token_accuracy": 0.8819447606801987,
+      "num_tokens": 332556668.0,
+      "step": 3110
+    },
+    {
+      "epoch": 7.086659064994299,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.0750725746265832e-06,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8780965954065323,
+      "num_tokens": 332664662.0,
+      "step": 3111
+    },
+    {
+      "epoch": 7.088939566704675,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.0735254154879979e-06,
+      "loss": 0.5859,
+      "mean_token_accuracy": 0.880235344171524,
+      "num_tokens": 332771413.0,
+      "step": 3112
+    },
+    {
+      "epoch": 7.091220068415051,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.0719790659888481e-06,
+      "loss": 0.5952,
+      "mean_token_accuracy": 0.8757160604000092,
+      "num_tokens": 332878585.0,
+      "step": 3113
+    },
+    {
+      "epoch": 7.0935005701254275,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.070433527006811e-06,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8773680776357651,
+      "num_tokens": 332985770.0,
+      "step": 3114
+    },
+    {
+      "epoch": 7.095781071835804,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.0688887994191049e-06,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8812025189399719,
+      "num_tokens": 333093197.0,
+      "step": 3115
+    },
+    {
+      "epoch": 7.09806157354618,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.0673448841024875e-06,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8841598778963089,
+      "num_tokens": 333200787.0,
+      "step": 3116
+    },
+    {
+      "epoch": 7.100342075256556,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.0658017819332556e-06,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8818695098161697,
+      "num_tokens": 333308356.0,
+      "step": 3117
+    },
+    {
+      "epoch": 7.102622576966933,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.064259493787244e-06,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8795827180147171,
+      "num_tokens": 333415586.0,
+      "step": 3118
+    },
+    {
+      "epoch": 7.104903078677309,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.0627180205398263e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8810782432556152,
+      "num_tokens": 333522537.0,
+      "step": 3119
+    },
+    {
+      "epoch": 7.1071835803876855,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.0611773630659117e-06,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8795880377292633,
+      "num_tokens": 333629886.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.109464082098062,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.0596375222399491e-06,
+      "loss": 0.5703,
+      "mean_token_accuracy": 0.8833544999361038,
+      "num_tokens": 333737026.0,
+      "step": 3121
+    },
+    {
+      "epoch": 7.111744583808438,
+      "grad_norm": 2.875,
+      "learning_rate": 1.0580984989359205e-06,
+      "loss": 0.5599,
+      "mean_token_accuracy": 0.8837230205535889,
+      "num_tokens": 333843902.0,
+      "step": 3122
+    },
+    {
+      "epoch": 7.114025085518814,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.0565602940273472e-06,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8842257410287857,
+      "num_tokens": 333951056.0,
+      "step": 3123
+    },
+    {
+      "epoch": 7.11630558722919,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.055022908387285e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8821539282798767,
+      "num_tokens": 334058094.0,
+      "step": 3124
+    },
+    {
+      "epoch": 7.1185860889395665,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.053486342888323e-06,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8864398002624512,
+      "num_tokens": 334165433.0,
+      "step": 3125
+    },
+    {
+      "epoch": 7.120866590649943,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.0519505984025865e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8731893301010132,
+      "num_tokens": 334272058.0,
+      "step": 3126
+    },
+    {
+      "epoch": 7.123147092360319,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.050415675801735e-06,
+      "loss": 0.5417,
+      "mean_token_accuracy": 0.8889250755310059,
+      "num_tokens": 334379380.0,
+      "step": 3127
+    },
+    {
+      "epoch": 7.125427594070696,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.0488815759569605e-06,
+      "loss": 0.5574,
+      "mean_token_accuracy": 0.8852666765451431,
+      "num_tokens": 334486391.0,
+      "step": 3128
+    },
+    {
+      "epoch": 7.127708095781072,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.0473482997389891e-06,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8808193802833557,
+      "num_tokens": 334593174.0,
+      "step": 3129
+    },
+    {
+      "epoch": 7.129988597491448,
+      "grad_norm": 3.25,
+      "learning_rate": 1.0458158480180777e-06,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8826655298471451,
+      "num_tokens": 334700093.0,
+      "step": 3130
+    },
+    {
+      "epoch": 7.1322690992018245,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.0442842216640168e-06,
+      "loss": 0.5637,
+      "mean_token_accuracy": 0.8832122981548309,
+      "num_tokens": 334807273.0,
+      "step": 3131
+    },
+    {
+      "epoch": 7.134549600912201,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.042753421546128e-06,
+      "loss": 0.5607,
+      "mean_token_accuracy": 0.883466050028801,
+      "num_tokens": 334914093.0,
+      "step": 3132
+    },
+    {
+      "epoch": 7.136830102622577,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.0412234485332636e-06,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8840949237346649,
+      "num_tokens": 335021072.0,
+      "step": 3133
+    },
+    {
+      "epoch": 7.139110604332953,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.0396943034938077e-06,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8802383244037628,
+      "num_tokens": 335127732.0,
+      "step": 3134
+    },
+    {
+      "epoch": 7.141391106043329,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.0381659872956732e-06,
+      "loss": 0.5877,
+      "mean_token_accuracy": 0.8789878785610199,
+      "num_tokens": 335234354.0,
+      "step": 3135
+    },
+    {
+      "epoch": 7.1436716077537055,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.0366385008063015e-06,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.8824616819620132,
+      "num_tokens": 335341184.0,
+      "step": 3136
+    },
+    {
+      "epoch": 7.145952109464082,
+      "grad_norm": 2.625,
+      "learning_rate": 1.0351118448926658e-06,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8839305341243744,
+      "num_tokens": 335448536.0,
+      "step": 3137
+    },
+    {
+      "epoch": 7.148232611174459,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.0335860204212662e-06,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.880255714058876,
+      "num_tokens": 335555399.0,
+      "step": 3138
+    },
+    {
+      "epoch": 7.150513112884835,
+      "grad_norm": 2.875,
+      "learning_rate": 1.0320610282581309e-06,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8826555460691452,
+      "num_tokens": 335662374.0,
+      "step": 3139
+    },
+    {
+      "epoch": 7.152793614595211,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.0305368692688175e-06,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8849854469299316,
+      "num_tokens": 335769670.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.155074116305587,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.029013544318407e-06,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8815399259328842,
+      "num_tokens": 335876896.0,
+      "step": 3141
+    },
+    {
+      "epoch": 7.1573546180159635,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.0274910542715103e-06,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8821052312850952,
+      "num_tokens": 335983657.0,
+      "step": 3142
+    },
+    {
+      "epoch": 7.15963511972634,
+      "grad_norm": 4.625,
+      "learning_rate": 1.025969399992264e-06,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8829544335603714,
+      "num_tokens": 336090948.0,
+      "step": 3143
+    },
+    {
+      "epoch": 7.161915621436716,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.0244485823443281e-06,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8804685175418854,
+      "num_tokens": 336197903.0,
+      "step": 3144
+    },
+    {
+      "epoch": 7.164196123147092,
+      "grad_norm": 2.421875,
+      "learning_rate": 1.0229286021908913e-06,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8843429386615753,
+      "num_tokens": 336305488.0,
+      "step": 3145
+    },
+    {
+      "epoch": 7.166476624857468,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.021409460394663e-06,
+      "loss": 0.5326,
+      "mean_token_accuracy": 0.8926713168621063,
+      "num_tokens": 336413922.0,
+      "step": 3146
+    },
+    {
+      "epoch": 7.168757126567845,
+      "grad_norm": 2.875,
+      "learning_rate": 1.0198911578178797e-06,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8805201500654221,
+      "num_tokens": 336521353.0,
+      "step": 3147
+    },
+    {
+      "epoch": 7.1710376282782216,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.0183736953223005e-06,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8848622888326645,
+      "num_tokens": 336628098.0,
+      "step": 3148
+    },
+    {
+      "epoch": 7.173318129988598,
+      "grad_norm": 4.5,
+      "learning_rate": 1.0168570737692082e-06,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8816571831703186,
+      "num_tokens": 336735354.0,
+      "step": 3149
+    },
+    {
+      "epoch": 7.175598631698974,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.0153412940194073e-06,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8790102005004883,
+      "num_tokens": 336843105.0,
+      "step": 3150
+    },
+    {
+      "epoch": 7.17787913340935,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.0138263569332268e-06,
+      "loss": 0.5677,
+      "mean_token_accuracy": 0.8861614763736725,
+      "num_tokens": 336950109.0,
+      "step": 3151
+    },
+    {
+      "epoch": 7.180159635119726,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.0123122633705131e-06,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.8783406764268875,
+      "num_tokens": 337056741.0,
+      "step": 3152
+    },
+    {
+      "epoch": 7.1824401368301025,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.0107990141906378e-06,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8800586462020874,
+      "num_tokens": 337163436.0,
+      "step": 3153
+    },
+    {
+      "epoch": 7.184720638540479,
+      "grad_norm": 3.0,
+      "learning_rate": 1.0092866102524922e-06,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8837520331144333,
+      "num_tokens": 337270521.0,
+      "step": 3154
+    },
+    {
+      "epoch": 7.187001140250855,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.0077750524144871e-06,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.887342780828476,
+      "num_tokens": 337377963.0,
+      "step": 3155
+    },
+    {
+      "epoch": 7.189281641961231,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.0062643415345546e-06,
+      "loss": 0.5532,
+      "mean_token_accuracy": 0.8867213129997253,
+      "num_tokens": 337484753.0,
+      "step": 3156
+    },
+    {
+      "epoch": 7.191562143671608,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.0047544784701435e-06,
+      "loss": 0.5591,
+      "mean_token_accuracy": 0.8835946470499039,
+      "num_tokens": 337592359.0,
+      "step": 3157
+    },
+    {
+      "epoch": 7.193842645381984,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.0032454640782232e-06,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8831405937671661,
+      "num_tokens": 337698920.0,
+      "step": 3158
+    },
+    {
+      "epoch": 7.196123147092361,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.0017372992152819e-06,
+      "loss": 0.5524,
+      "mean_token_accuracy": 0.8846133053302765,
+      "num_tokens": 337806568.0,
+      "step": 3159
+    },
+    {
+      "epoch": 7.198403648802737,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.0002299847373243e-06,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.88174769282341,
+      "num_tokens": 337913427.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.200684150513113,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.987235214998741e-07,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8783168196678162,
+      "num_tokens": 338020074.0,
+      "step": 3161
+    },
+    {
+      "epoch": 7.202964652223489,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.972179103579687e-07,
+      "loss": 0.5932,
+      "mean_token_accuracy": 0.8791584521532059,
+      "num_tokens": 338126865.0,
+      "step": 3162
+    },
+    {
+      "epoch": 7.205245153933865,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.957131521661655e-07,
+      "loss": 0.561,
+      "mean_token_accuracy": 0.8836409598588943,
+      "num_tokens": 338234022.0,
+      "step": 3163
+    },
+    {
+      "epoch": 7.2075256556442415,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.942092477785365e-07,
+      "loss": 0.5694,
+      "mean_token_accuracy": 0.8865296542644501,
+      "num_tokens": 338341091.0,
+      "step": 3164
+    },
+    {
+      "epoch": 7.209806157354618,
+      "grad_norm": 2.875,
+      "learning_rate": 9.927061980486668e-07,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.883445993065834,
+      "num_tokens": 338447802.0,
+      "step": 3165
+    },
+    {
+      "epoch": 7.212086659064994,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.9120400382966e-07,
+      "loss": 0.5991,
+      "mean_token_accuracy": 0.8793905973434448,
+      "num_tokens": 338554661.0,
+      "step": 3166
+    },
+    {
+      "epoch": 7.214367160775371,
+      "grad_norm": 5.5625,
+      "learning_rate": 9.897026659741328e-07,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8824618011713028,
+      "num_tokens": 338661638.0,
+      "step": 3167
+    },
+    {
+      "epoch": 7.216647662485747,
+      "grad_norm": 6.0,
+      "learning_rate": 9.882021853342143e-07,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.882818415760994,
+      "num_tokens": 338768332.0,
+      "step": 3168
+    },
+    {
+      "epoch": 7.218928164196123,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.867025627615493e-07,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8801371902227402,
+      "num_tokens": 338875605.0,
+      "step": 3169
+    },
+    {
+      "epoch": 7.2212086659065,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.852037991072941e-07,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8800628185272217,
+      "num_tokens": 338982521.0,
+      "step": 3170
+    },
+    {
+      "epoch": 7.223489167616876,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.837058952221182e-07,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8799264430999756,
+      "num_tokens": 339089755.0,
+      "step": 3171
+    },
+    {
+      "epoch": 7.225769669327252,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.822088519562038e-07,
+      "loss": 0.5651,
+      "mean_token_accuracy": 0.8843246251344681,
+      "num_tokens": 339196888.0,
+      "step": 3172
+    },
+    {
+      "epoch": 7.228050171037628,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.80712670159242e-07,
+      "loss": 0.5808,
+      "mean_token_accuracy": 0.8821865916252136,
+      "num_tokens": 339303828.0,
+      "step": 3173
+    },
+    {
+      "epoch": 7.230330672748004,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.792173506804378e-07,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.880447655916214,
+      "num_tokens": 339410331.0,
+      "step": 3174
+    },
+    {
+      "epoch": 7.2326111744583805,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.777228943685055e-07,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8847004473209381,
+      "num_tokens": 339517573.0,
+      "step": 3175
+    },
+    {
+      "epoch": 7.234891676168757,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.762293020716696e-07,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8809026032686234,
+      "num_tokens": 339624146.0,
+      "step": 3176
+    },
+    {
+      "epoch": 7.237172177879134,
+      "grad_norm": 2.75,
+      "learning_rate": 9.74736574637665e-07,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8855401873588562,
+      "num_tokens": 339731001.0,
+      "step": 3177
+    },
+    {
+      "epoch": 7.23945267958951,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.732447129137337e-07,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8829121142625809,
+      "num_tokens": 339838319.0,
+      "step": 3178
+    },
+    {
+      "epoch": 7.241733181299886,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.717537177466279e-07,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8830237686634064,
+      "num_tokens": 339945278.0,
+      "step": 3179
+    },
+    {
+      "epoch": 7.244013683010262,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.702635899826082e-07,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.8862001597881317,
+      "num_tokens": 340052303.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.246294184720639,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.687743304674421e-07,
+      "loss": 0.5634,
+      "mean_token_accuracy": 0.8842123001813889,
+      "num_tokens": 340158489.0,
+      "step": 3181
+    },
+    {
+      "epoch": 7.248574686431015,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.672859400464046e-07,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8775025904178619,
+      "num_tokens": 340265222.0,
+      "step": 3182
+    },
+    {
+      "epoch": 7.250855188141391,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.657984195642783e-07,
+      "loss": 0.5997,
+      "mean_token_accuracy": 0.8756692260503769,
+      "num_tokens": 340371418.0,
+      "step": 3183
+    },
+    {
+      "epoch": 7.253135689851767,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.64311769865349e-07,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8807047456502914,
+      "num_tokens": 340478369.0,
+      "step": 3184
+    },
+    {
+      "epoch": 7.255416191562143,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.628259917934118e-07,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8846138417720795,
+      "num_tokens": 340585428.0,
+      "step": 3185
+    },
+    {
+      "epoch": 7.2576966932725195,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.613410861917661e-07,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8808754086494446,
+      "num_tokens": 340692420.0,
+      "step": 3186
+    },
+    {
+      "epoch": 7.259977194982897,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.59857053903214e-07,
+      "loss": 0.5659,
+      "mean_token_accuracy": 0.8816579431295395,
+      "num_tokens": 340799529.0,
+      "step": 3187
+    },
+    {
+      "epoch": 7.262257696693273,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.583738957700653e-07,
+      "loss": 0.5858,
+      "mean_token_accuracy": 0.8804793953895569,
+      "num_tokens": 340906789.0,
+      "step": 3188
+    },
+    {
+      "epoch": 7.264538198403649,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.568916126341305e-07,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8775817602872849,
+      "num_tokens": 341013458.0,
+      "step": 3189
+    },
+    {
+      "epoch": 7.266818700114025,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.554102053367253e-07,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8787823021411896,
+      "num_tokens": 341120727.0,
+      "step": 3190
+    },
+    {
+      "epoch": 7.269099201824401,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.53929674718668e-07,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.8785655200481415,
+      "num_tokens": 341227639.0,
+      "step": 3191
+    },
+    {
+      "epoch": 7.271379703534778,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.524500216202795e-07,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8779048174619675,
+      "num_tokens": 341334736.0,
+      "step": 3192
+    },
+    {
+      "epoch": 7.273660205245154,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.50971246881382e-07,
+      "loss": 0.5946,
+      "mean_token_accuracy": 0.8770471662282944,
+      "num_tokens": 341441533.0,
+      "step": 3193
+    },
+    {
+      "epoch": 7.27594070695553,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.494933513413007e-07,
+      "loss": 0.5982,
+      "mean_token_accuracy": 0.8785821348428726,
+      "num_tokens": 341547940.0,
+      "step": 3194
+    },
+    {
+      "epoch": 7.278221208665906,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.480163358388584e-07,
+      "loss": 0.5544,
+      "mean_token_accuracy": 0.8853187263011932,
+      "num_tokens": 341655509.0,
+      "step": 3195
+    },
+    {
+      "epoch": 7.280501710376283,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.465402012123818e-07,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.8840262442827225,
+      "num_tokens": 341762697.0,
+      "step": 3196
+    },
+    {
+      "epoch": 7.282782212086659,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.45064948299696e-07,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8797417432069778,
+      "num_tokens": 341869635.0,
+      "step": 3197
+    },
+    {
+      "epoch": 7.285062713797036,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.435905779381265e-07,
+      "loss": 0.6004,
+      "mean_token_accuracy": 0.8786374479532242,
+      "num_tokens": 341975850.0,
+      "step": 3198
+    },
+    {
+      "epoch": 7.287343215507412,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.421170909644983e-07,
+      "loss": 0.5555,
+      "mean_token_accuracy": 0.8853399753570557,
+      "num_tokens": 342083686.0,
+      "step": 3199
+    },
+    {
+      "epoch": 7.289623717217788,
+      "grad_norm": 3.375,
+      "learning_rate": 9.406444882151322e-07,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8812949508428574,
+      "num_tokens": 342191163.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.291904218928164,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.391727705258502e-07,
+      "loss": 0.5411,
+      "mean_token_accuracy": 0.8899759203195572,
+      "num_tokens": 342298933.0,
+      "step": 3201
+    },
+    {
+      "epoch": 7.29418472063854,
+      "grad_norm": 5.71875,
+      "learning_rate": 9.377019387319705e-07,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8839350640773773,
+      "num_tokens": 342405767.0,
+      "step": 3202
+    },
+    {
+      "epoch": 7.296465222348917,
+      "grad_norm": 3.0,
+      "learning_rate": 9.362319936683092e-07,
+      "loss": 0.5962,
+      "mean_token_accuracy": 0.8760559111833572,
+      "num_tokens": 342512638.0,
+      "step": 3203
+    },
+    {
+      "epoch": 7.298745724059293,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.347629361691795e-07,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8815828859806061,
+      "num_tokens": 342619888.0,
+      "step": 3204
+    },
+    {
+      "epoch": 7.301026225769669,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.332947670683882e-07,
+      "loss": 0.5465,
+      "mean_token_accuracy": 0.8862589299678802,
+      "num_tokens": 342726859.0,
+      "step": 3205
+    },
+    {
+      "epoch": 7.303306727480045,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.318274871992408e-07,
+      "loss": 0.5435,
+      "mean_token_accuracy": 0.8866942375898361,
+      "num_tokens": 342834871.0,
+      "step": 3206
+    },
+    {
+      "epoch": 7.305587229190422,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.303610973945376e-07,
+      "loss": 0.5552,
+      "mean_token_accuracy": 0.8848265260457993,
+      "num_tokens": 342942485.0,
+      "step": 3207
+    },
+    {
+      "epoch": 7.307867730900798,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.288955984865717e-07,
+      "loss": 0.5443,
+      "mean_token_accuracy": 0.8879365026950836,
+      "num_tokens": 343049386.0,
+      "step": 3208
+    },
+    {
+      "epoch": 7.310148232611175,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.274309913071328e-07,
+      "loss": 0.5866,
+      "mean_token_accuracy": 0.8782593309879303,
+      "num_tokens": 343155935.0,
+      "step": 3209
+    },
+    {
+      "epoch": 7.312428734321551,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.259672766875044e-07,
+      "loss": 0.5835,
+      "mean_token_accuracy": 0.8784520477056503,
+      "num_tokens": 343262869.0,
+      "step": 3210
+    },
+    {
+      "epoch": 7.314709236031927,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.245044554584609e-07,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.884000301361084,
+      "num_tokens": 343369889.0,
+      "step": 3211
+    },
+    {
+      "epoch": 7.316989737742303,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.230425284502725e-07,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8830729424953461,
+      "num_tokens": 343476997.0,
+      "step": 3212
+    },
+    {
+      "epoch": 7.319270239452679,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.215814964927005e-07,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8840525597333908,
+      "num_tokens": 343584267.0,
+      "step": 3213
+    },
+    {
+      "epoch": 7.321550741163056,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.201213604149989e-07,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8782940655946732,
+      "num_tokens": 343690596.0,
+      "step": 3214
+    },
+    {
+      "epoch": 7.323831242873432,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.186621210459129e-07,
+      "loss": 0.6009,
+      "mean_token_accuracy": 0.877897322177887,
+      "num_tokens": 343797695.0,
+      "step": 3215
+    },
+    {
+      "epoch": 7.326111744583809,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.172037792136773e-07,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8841624855995178,
+      "num_tokens": 343905377.0,
+      "step": 3216
+    },
+    {
+      "epoch": 7.328392246294185,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.157463357460194e-07,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8875628858804703,
+      "num_tokens": 344012519.0,
+      "step": 3217
+    },
+    {
+      "epoch": 7.330672748004561,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.142897914701565e-07,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8802744299173355,
+      "num_tokens": 344120062.0,
+      "step": 3218
+    },
+    {
+      "epoch": 7.3329532497149374,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.128341472127944e-07,
+      "loss": 0.562,
+      "mean_token_accuracy": 0.8852846622467041,
+      "num_tokens": 344227315.0,
+      "step": 3219
+    },
+    {
+      "epoch": 7.335233751425314,
+      "grad_norm": 2.5,
+      "learning_rate": 9.113794038001298e-07,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8799241334199905,
+      "num_tokens": 344334679.0,
+      "step": 3220
+    },
+    {
+      "epoch": 7.33751425313569,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.099255620578451e-07,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8857553601264954,
+      "num_tokens": 344441934.0,
+      "step": 3221
+    },
+    {
+      "epoch": 7.339794754846066,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.084726228111141e-07,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8808804005384445,
+      "num_tokens": 344548474.0,
+      "step": 3222
+    },
+    {
+      "epoch": 7.342075256556442,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.070205868845966e-07,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8860443085432053,
+      "num_tokens": 344655190.0,
+      "step": 3223
+    },
+    {
+      "epoch": 7.344355758266818,
+      "grad_norm": 5.09375,
+      "learning_rate": 9.055694551024402e-07,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8791934102773666,
+      "num_tokens": 344762745.0,
+      "step": 3224
+    },
+    {
+      "epoch": 7.346636259977195,
+      "grad_norm": 2.625,
+      "learning_rate": 9.041192282882796e-07,
+      "loss": 0.5971,
+      "mean_token_accuracy": 0.8765173703432083,
+      "num_tokens": 344869812.0,
+      "step": 3225
+    },
+    {
+      "epoch": 7.348916761687571,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.026699072652361e-07,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8783349841833115,
+      "num_tokens": 344976836.0,
+      "step": 3226
+    },
+    {
+      "epoch": 7.351197263397948,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.012214928559149e-07,
+      "loss": 0.551,
+      "mean_token_accuracy": 0.8875472396612167,
+      "num_tokens": 345084950.0,
+      "step": 3227
+    },
+    {
+      "epoch": 7.353477765108324,
+      "grad_norm": 6.34375,
+      "learning_rate": 8.997739858824083e-07,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.8791827261447906,
+      "num_tokens": 345191976.0,
+      "step": 3228
+    },
+    {
+      "epoch": 7.3557582668187,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.983273871662951e-07,
+      "loss": 0.587,
+      "mean_token_accuracy": 0.8798834979534149,
+      "num_tokens": 345299539.0,
+      "step": 3229
+    },
+    {
+      "epoch": 7.3580387685290765,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.968816975286346e-07,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8791932165622711,
+      "num_tokens": 345406386.0,
+      "step": 3230
+    },
+    {
+      "epoch": 7.360319270239453,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.954369177899727e-07,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8776541501283646,
+      "num_tokens": 345513832.0,
+      "step": 3231
+    },
+    {
+      "epoch": 7.362599771949829,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.939930487703402e-07,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8819593489170074,
+      "num_tokens": 345621238.0,
+      "step": 3232
+    },
+    {
+      "epoch": 7.364880273660205,
+      "grad_norm": 4.75,
+      "learning_rate": 8.925500912892471e-07,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.878967210650444,
+      "num_tokens": 345727718.0,
+      "step": 3233
+    },
+    {
+      "epoch": 7.367160775370581,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.911080461656893e-07,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.885176420211792,
+      "num_tokens": 345834217.0,
+      "step": 3234
+    },
+    {
+      "epoch": 7.369441277080957,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.896669142181436e-07,
+      "loss": 0.5766,
+      "mean_token_accuracy": 0.8792836219072342,
+      "num_tokens": 345941843.0,
+      "step": 3235
+    },
+    {
+      "epoch": 7.3717217787913345,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.882266962645695e-07,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8821652680635452,
+      "num_tokens": 346049168.0,
+      "step": 3236
+    },
+    {
+      "epoch": 7.374002280501711,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.867873931224053e-07,
+      "loss": 0.5566,
+      "mean_token_accuracy": 0.8830807954072952,
+      "num_tokens": 346156521.0,
+      "step": 3237
+    },
+    {
+      "epoch": 7.376282782212087,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.853490056085723e-07,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8781331777572632,
+      "num_tokens": 346263445.0,
+      "step": 3238
+    },
+    {
+      "epoch": 7.378563283922463,
+      "grad_norm": 2.625,
+      "learning_rate": 8.839115345394716e-07,
+      "loss": 0.5608,
+      "mean_token_accuracy": 0.8834449648857117,
+      "num_tokens": 346370432.0,
+      "step": 3239
+    },
+    {
+      "epoch": 7.380843785632839,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.824749807309846e-07,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8865119814872742,
+      "num_tokens": 346477909.0,
+      "step": 3240
+    },
+    {
+      "epoch": 7.3831242873432155,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.810393449984706e-07,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8794442266225815,
+      "num_tokens": 346584653.0,
+      "step": 3241
+    },
+    {
+      "epoch": 7.385404789053592,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.7960462815677e-07,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8795085549354553,
+      "num_tokens": 346691178.0,
+      "step": 3242
+    },
+    {
+      "epoch": 7.387685290763968,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.781708310201989e-07,
+      "loss": 0.5395,
+      "mean_token_accuracy": 0.8901244848966599,
+      "num_tokens": 346798766.0,
+      "step": 3243
+    },
+    {
+      "epoch": 7.389965792474344,
+      "grad_norm": 3.75,
+      "learning_rate": 8.767379544025531e-07,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8783616870641708,
+      "num_tokens": 346905556.0,
+      "step": 3244
+    },
+    {
+      "epoch": 7.39224629418472,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.753059991171065e-07,
+      "loss": 0.6058,
+      "mean_token_accuracy": 0.8745017796754837,
+      "num_tokens": 347012108.0,
+      "step": 3245
+    },
+    {
+      "epoch": 7.394526795895097,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.738749659766085e-07,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8834935277700424,
+      "num_tokens": 347119108.0,
+      "step": 3246
+    },
+    {
+      "epoch": 7.3968072976054735,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.724448557932874e-07,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8808663189411163,
+      "num_tokens": 347226523.0,
+      "step": 3247
+    },
+    {
+      "epoch": 7.39908779931585,
+      "grad_norm": 3.9375,
+      "learning_rate": 8.71015669378844e-07,
+      "loss": 0.5613,
+      "mean_token_accuracy": 0.883145734667778,
+      "num_tokens": 347333636.0,
+      "step": 3248
+    },
+    {
+      "epoch": 7.401368301026226,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.69587407544458e-07,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8786642998456955,
+      "num_tokens": 347440255.0,
+      "step": 3249
+    },
+    {
+      "epoch": 7.403648802736602,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.681600711007832e-07,
+      "loss": 0.5679,
+      "mean_token_accuracy": 0.8843037188053131,
+      "num_tokens": 347546867.0,
+      "step": 3250
+    },
+    {
+      "epoch": 7.405929304446978,
+      "grad_norm": 2.75,
+      "learning_rate": 8.667336608579488e-07,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8831322491168976,
+      "num_tokens": 347653956.0,
+      "step": 3251
+    },
+    {
+      "epoch": 7.4082098061573545,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.653081776255562e-07,
+      "loss": 0.557,
+      "mean_token_accuracy": 0.8858849257230759,
+      "num_tokens": 347761532.0,
+      "step": 3252
+    },
+    {
+      "epoch": 7.410490307867731,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.638836222126839e-07,
+      "loss": 0.5495,
+      "mean_token_accuracy": 0.8869747668504715,
+      "num_tokens": 347869055.0,
+      "step": 3253
+    },
+    {
+      "epoch": 7.412770809578107,
+      "grad_norm": 5.0625,
+      "learning_rate": 8.624599954278803e-07,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8779660612344742,
+      "num_tokens": 347975665.0,
+      "step": 3254
+    },
+    {
+      "epoch": 7.415051311288483,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.610372980791695e-07,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8830466419458389,
+      "num_tokens": 348082812.0,
+      "step": 3255
+    },
+    {
+      "epoch": 7.41733181299886,
+      "grad_norm": 6.40625,
+      "learning_rate": 8.59615530974047e-07,
+      "loss": 0.5588,
+      "mean_token_accuracy": 0.8859402984380722,
+      "num_tokens": 348190346.0,
+      "step": 3256
+    },
+    {
+      "epoch": 7.419612314709236,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.581946949194802e-07,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.881903350353241,
+      "num_tokens": 348297177.0,
+      "step": 3257
+    },
+    {
+      "epoch": 7.4218928164196125,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.56774790721909e-07,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8854805678129196,
+      "num_tokens": 348404580.0,
+      "step": 3258
+    },
+    {
+      "epoch": 7.424173318129989,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.553558191872422e-07,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8778659850358963,
+      "num_tokens": 348511494.0,
+      "step": 3259
+    },
+    {
+      "epoch": 7.426453819840365,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.539377811208613e-07,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8794593662023544,
+      "num_tokens": 348618636.0,
+      "step": 3260
+    },
+    {
+      "epoch": 7.428734321550741,
+      "grad_norm": 4.15625,
+      "learning_rate": 8.525206773276173e-07,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.883561447262764,
+      "num_tokens": 348725346.0,
+      "step": 3261
+    },
+    {
+      "epoch": 7.431014823261117,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.511045086118311e-07,
+      "loss": 0.5403,
+      "mean_token_accuracy": 0.8908537030220032,
+      "num_tokens": 348832898.0,
+      "step": 3262
+    },
+    {
+      "epoch": 7.4332953249714935,
+      "grad_norm": 2.875,
+      "learning_rate": 8.496892757772934e-07,
+      "loss": 0.5582,
+      "mean_token_accuracy": 0.8817539364099503,
+      "num_tokens": 348940485.0,
+      "step": 3263
+    },
+    {
+      "epoch": 7.43557582668187,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.482749796272613e-07,
+      "loss": 0.586,
+      "mean_token_accuracy": 0.8792066723108292,
+      "num_tokens": 349047156.0,
+      "step": 3264
+    },
+    {
+      "epoch": 7.437856328392247,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.468616209644634e-07,
+      "loss": 0.5426,
+      "mean_token_accuracy": 0.8881279230117798,
+      "num_tokens": 349153893.0,
+      "step": 3265
+    },
+    {
+      "epoch": 7.440136830102623,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.454492005910942e-07,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8840852081775665,
+      "num_tokens": 349260811.0,
+      "step": 3266
+    },
+    {
+      "epoch": 7.442417331812999,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.440377193088162e-07,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8842138350009918,
+      "num_tokens": 349368153.0,
+      "step": 3267
+    },
+    {
+      "epoch": 7.444697833523375,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.426271779187592e-07,
+      "loss": 0.5885,
+      "mean_token_accuracy": 0.8814673125743866,
+      "num_tokens": 349475704.0,
+      "step": 3268
+    },
+    {
+      "epoch": 7.4469783352337515,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.4121757722152e-07,
+      "loss": 0.554,
+      "mean_token_accuracy": 0.8854426443576813,
+      "num_tokens": 349582741.0,
+      "step": 3269
+    },
+    {
+      "epoch": 7.449258836944128,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.398089180171592e-07,
+      "loss": 0.5512,
+      "mean_token_accuracy": 0.8872993588447571,
+      "num_tokens": 349689828.0,
+      "step": 3270
+    },
+    {
+      "epoch": 7.451539338654504,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.384012011052053e-07,
+      "loss": 0.541,
+      "mean_token_accuracy": 0.8934241831302643,
+      "num_tokens": 349797473.0,
+      "step": 3271
+    },
+    {
+      "epoch": 7.45381984036488,
+      "grad_norm": 4.6875,
+      "learning_rate": 8.369944272846522e-07,
+      "loss": 0.5388,
+      "mean_token_accuracy": 0.8873519897460938,
+      "num_tokens": 349904634.0,
+      "step": 3272
+    },
+    {
+      "epoch": 7.456100342075256,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.355885973539557e-07,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8826733976602554,
+      "num_tokens": 350011607.0,
+      "step": 3273
+    },
+    {
+      "epoch": 7.4583808437856325,
+      "grad_norm": 3.0,
+      "learning_rate": 8.341837121110386e-07,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8846445381641388,
+      "num_tokens": 350119034.0,
+      "step": 3274
+    },
+    {
+      "epoch": 7.460661345496009,
+      "grad_norm": 2.5625,
+      "learning_rate": 8.327797723532874e-07,
+      "loss": 0.5438,
+      "mean_token_accuracy": 0.8912030011415482,
+      "num_tokens": 350226528.0,
+      "step": 3275
+    },
+    {
+      "epoch": 7.462941847206386,
+      "grad_norm": 2.75,
+      "learning_rate": 8.313767788775498e-07,
+      "loss": 0.596,
+      "mean_token_accuracy": 0.8776167631149292,
+      "num_tokens": 350333295.0,
+      "step": 3276
+    },
+    {
+      "epoch": 7.465222348916762,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.299747324801385e-07,
+      "loss": 0.5985,
+      "mean_token_accuracy": 0.8792443871498108,
+      "num_tokens": 350440043.0,
+      "step": 3277
+    },
+    {
+      "epoch": 7.467502850627138,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.285736339568279e-07,
+      "loss": 0.5906,
+      "mean_token_accuracy": 0.8789392858743668,
+      "num_tokens": 350546875.0,
+      "step": 3278
+    },
+    {
+      "epoch": 7.469783352337514,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.271734841028553e-07,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8837863206863403,
+      "num_tokens": 350653887.0,
+      "step": 3279
+    },
+    {
+      "epoch": 7.4720638540478905,
+      "grad_norm": 2.625,
+      "learning_rate": 8.25774283712917e-07,
+      "loss": 0.5674,
+      "mean_token_accuracy": 0.8839046061038971,
+      "num_tokens": 350761695.0,
+      "step": 3280
+    },
+    {
+      "epoch": 7.474344355758267,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.243760335811734e-07,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8785386085510254,
+      "num_tokens": 350868270.0,
+      "step": 3281
+    },
+    {
+      "epoch": 7.476624857468643,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.229787345012439e-07,
+      "loss": 0.5577,
+      "mean_token_accuracy": 0.8850255161523819,
+      "num_tokens": 350975311.0,
+      "step": 3282
+    },
+    {
+      "epoch": 7.478905359179019,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.215823872662084e-07,
+      "loss": 0.5602,
+      "mean_token_accuracy": 0.8832758516073227,
+      "num_tokens": 351082788.0,
+      "step": 3283
+    },
+    {
+      "epoch": 7.481185860889395,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.201869926686068e-07,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8810236304998398,
+      "num_tokens": 351190001.0,
+      "step": 3284
+    },
+    {
+      "epoch": 7.483466362599772,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.187925515004391e-07,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8823070824146271,
+      "num_tokens": 351296897.0,
+      "step": 3285
+    },
+    {
+      "epoch": 7.485746864310149,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.173990645531612e-07,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8813482820987701,
+      "num_tokens": 351404175.0,
+      "step": 3286
+    },
+    {
+      "epoch": 7.488027366020525,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.160065326176905e-07,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8828651309013367,
+      "num_tokens": 351510970.0,
+      "step": 3287
+    },
+    {
+      "epoch": 7.490307867730901,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.14614956484401e-07,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.8815842270851135,
+      "num_tokens": 351617878.0,
+      "step": 3288
+    },
+    {
+      "epoch": 7.492588369441277,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.132243369431248e-07,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8866011053323746,
+      "num_tokens": 351725881.0,
+      "step": 3289
+    },
+    {
+      "epoch": 7.494868871151653,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.11834674783151e-07,
+      "loss": 0.5889,
+      "mean_token_accuracy": 0.8792490214109421,
+      "num_tokens": 351832802.0,
+      "step": 3290
+    },
+    {
+      "epoch": 7.4971493728620295,
+      "grad_norm": 2.75,
+      "learning_rate": 8.104459707932238e-07,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8836107850074768,
+      "num_tokens": 351939776.0,
+      "step": 3291
+    },
+    {
+      "epoch": 7.499429874572406,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.090582257615456e-07,
+      "loss": 0.5699,
+      "mean_token_accuracy": 0.8809670358896255,
+      "num_tokens": 352046740.0,
+      "step": 3292
+    },
+    {
+      "epoch": 7.501710376282782,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.076714404757735e-07,
+      "loss": 0.5593,
+      "mean_token_accuracy": 0.8861764520406723,
+      "num_tokens": 352153749.0,
+      "step": 3293
+    },
+    {
+      "epoch": 7.503990877993158,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.062856157230209e-07,
+      "loss": 0.5891,
+      "mean_token_accuracy": 0.8781421184539795,
+      "num_tokens": 352260556.0,
+      "step": 3294
+    },
+    {
+      "epoch": 7.506271379703534,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.049007522898536e-07,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8831081688404083,
+      "num_tokens": 352367485.0,
+      "step": 3295
+    },
+    {
+      "epoch": 7.508551881413911,
+      "grad_norm": 4.46875,
+      "learning_rate": 8.035168509622948e-07,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8832761347293854,
+      "num_tokens": 352473992.0,
+      "step": 3296
+    },
+    {
+      "epoch": 7.510832383124288,
+      "grad_norm": 3.375,
+      "learning_rate": 8.02133912525819e-07,
+      "loss": 0.5534,
+      "mean_token_accuracy": 0.8868482112884521,
+      "num_tokens": 352582216.0,
+      "step": 3297
+    },
+    {
+      "epoch": 7.513112884834664,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.007519377653558e-07,
+      "loss": 0.5803,
+      "mean_token_accuracy": 0.8818371295928955,
+      "num_tokens": 352688608.0,
+      "step": 3298
+    },
+    {
+      "epoch": 7.51539338654504,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.993709274652872e-07,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8833436369895935,
+      "num_tokens": 352794995.0,
+      "step": 3299
+    },
+    {
+      "epoch": 7.517673888255416,
+      "grad_norm": 2.625,
+      "learning_rate": 7.979908824094484e-07,
+      "loss": 0.5448,
+      "mean_token_accuracy": 0.8875368237495422,
+      "num_tokens": 352902959.0,
+      "step": 3300
+    },
+    {
+      "epoch": 7.517673888255416,
+      "eval_loss": 0.5862932801246643,
+      "eval_mean_token_accuracy": 0.8799256974753318,
+      "eval_num_tokens": 352902959.0,
+      "eval_runtime": 58.6491,
+      "eval_samples_per_second": 142.969,
+      "eval_steps_per_second": 4.484,
+      "step": 3300
+    },
+    {
+      "epoch": 7.519954389965792,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.966118033811271e-07,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8830474466085434,
+      "num_tokens": 353010544.0,
+      "step": 3301
+    },
+    {
+      "epoch": 7.5222348916761685,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.952336911630604e-07,
+      "loss": 0.5655,
+      "mean_token_accuracy": 0.8814462274312973,
+      "num_tokens": 353117940.0,
+      "step": 3302
+    },
+    {
+      "epoch": 7.524515393386545,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.938565465374384e-07,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8806511014699936,
+      "num_tokens": 353225388.0,
+      "step": 3303
+    },
+    {
+      "epoch": 7.526795895096921,
+      "grad_norm": 5.21875,
+      "learning_rate": 7.924803702859024e-07,
+      "loss": 0.5648,
+      "mean_token_accuracy": 0.8838271647691727,
+      "num_tokens": 353332497.0,
+      "step": 3304
+    },
+    {
+      "epoch": 7.529076396807298,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.911051631895433e-07,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8843842297792435,
+      "num_tokens": 353439584.0,
+      "step": 3305
+    },
+    {
+      "epoch": 7.531356898517674,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.897309260289027e-07,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.882561057806015,
+      "num_tokens": 353546370.0,
+      "step": 3306
+    },
+    {
+      "epoch": 7.53363740022805,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.883576595839698e-07,
+      "loss": 0.5814,
+      "mean_token_accuracy": 0.879010796546936,
+      "num_tokens": 353653493.0,
+      "step": 3307
+    },
+    {
+      "epoch": 7.535917901938427,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.869853646341849e-07,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8826321810483932,
+      "num_tokens": 353759627.0,
+      "step": 3308
+    },
+    {
+      "epoch": 7.538198403648803,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.856140419584357e-07,
+      "loss": 0.5969,
+      "mean_token_accuracy": 0.8757929354906082,
+      "num_tokens": 353866548.0,
+      "step": 3309
+    },
+    {
+      "epoch": 7.540478905359179,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.842436923350591e-07,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8822405785322189,
+      "num_tokens": 353973896.0,
+      "step": 3310
+    },
+    {
+      "epoch": 7.542759407069555,
+      "grad_norm": 4.25,
+      "learning_rate": 7.828743165418393e-07,
+      "loss": 0.5866,
+      "mean_token_accuracy": 0.8774521350860596,
+      "num_tokens": 354080599.0,
+      "step": 3311
+    },
+    {
+      "epoch": 7.545039908779931,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.815059153560065e-07,
+      "loss": 0.5634,
+      "mean_token_accuracy": 0.8832540661096573,
+      "num_tokens": 354188075.0,
+      "step": 3312
+    },
+    {
+      "epoch": 7.5473204104903076,
+      "grad_norm": 4.5,
+      "learning_rate": 7.801384895542391e-07,
+      "loss": 0.5556,
+      "mean_token_accuracy": 0.8853226155042648,
+      "num_tokens": 354295337.0,
+      "step": 3313
+    },
+    {
+      "epoch": 7.549600912200685,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.78772039912662e-07,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8821288347244263,
+      "num_tokens": 354402522.0,
+      "step": 3314
+    },
+    {
+      "epoch": 7.55188141391106,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.774065672068463e-07,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8814186155796051,
+      "num_tokens": 354509530.0,
+      "step": 3315
+    },
+    {
+      "epoch": 7.554161915621437,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.760420722118059e-07,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8813648819923401,
+      "num_tokens": 354616615.0,
+      "step": 3316
+    },
+    {
+      "epoch": 7.556442417331813,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.746785557020034e-07,
+      "loss": 0.564,
+      "mean_token_accuracy": 0.8848537355661392,
+      "num_tokens": 354724411.0,
+      "step": 3317
+    },
+    {
+      "epoch": 7.558722919042189,
+      "grad_norm": 4.5,
+      "learning_rate": 7.733160184513447e-07,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8836092352867126,
+      "num_tokens": 354831136.0,
+      "step": 3318
+    },
+    {
+      "epoch": 7.561003420752566,
+      "grad_norm": 4.25,
+      "learning_rate": 7.719544612331781e-07,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8814618289470673,
+      "num_tokens": 354938130.0,
+      "step": 3319
+    },
+    {
+      "epoch": 7.563283922462942,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.705938848202985e-07,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8781514316797256,
+      "num_tokens": 355044678.0,
+      "step": 3320
+    },
+    {
+      "epoch": 7.565564424173318,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.692342899849419e-07,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.883265346288681,
+      "num_tokens": 355152110.0,
+      "step": 3321
+    },
+    {
+      "epoch": 7.567844925883694,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.678756774987897e-07,
+      "loss": 0.5704,
+      "mean_token_accuracy": 0.8851431161165237,
+      "num_tokens": 355259414.0,
+      "step": 3322
+    },
+    {
+      "epoch": 7.57012542759407,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.665180481329621e-07,
+      "loss": 0.5549,
+      "mean_token_accuracy": 0.8860851675271988,
+      "num_tokens": 355366168.0,
+      "step": 3323
+    },
+    {
+      "epoch": 7.572405929304447,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.651614026580243e-07,
+      "loss": 0.5704,
+      "mean_token_accuracy": 0.8835692703723907,
+      "num_tokens": 355472724.0,
+      "step": 3324
+    },
+    {
+      "epoch": 7.574686431014824,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.638057418439818e-07,
+      "loss": 0.5896,
+      "mean_token_accuracy": 0.8791484832763672,
+      "num_tokens": 355579877.0,
+      "step": 3325
+    },
+    {
+      "epoch": 7.5769669327252,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.624510664602819e-07,
+      "loss": 0.5607,
+      "mean_token_accuracy": 0.8831581175327301,
+      "num_tokens": 355686885.0,
+      "step": 3326
+    },
+    {
+      "epoch": 7.579247434435576,
+      "grad_norm": 3.625,
+      "learning_rate": 7.610973772758118e-07,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8818655163049698,
+      "num_tokens": 355793461.0,
+      "step": 3327
+    },
+    {
+      "epoch": 7.581527936145952,
+      "grad_norm": 3.0,
+      "learning_rate": 7.597446750589005e-07,
+      "loss": 0.579,
+      "mean_token_accuracy": 0.8854977488517761,
+      "num_tokens": 355900140.0,
+      "step": 3328
+    },
+    {
+      "epoch": 7.583808437856328,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.583929605773138e-07,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8835340738296509,
+      "num_tokens": 356007033.0,
+      "step": 3329
+    },
+    {
+      "epoch": 7.586088939566705,
+      "grad_norm": 2.53125,
+      "learning_rate": 7.570422345982598e-07,
+      "loss": 0.558,
+      "mean_token_accuracy": 0.8873008489608765,
+      "num_tokens": 356114209.0,
+      "step": 3330
+    },
+    {
+      "epoch": 7.588369441277081,
+      "grad_norm": 3.75,
+      "learning_rate": 7.556924978883843e-07,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8829249292612076,
+      "num_tokens": 356220979.0,
+      "step": 3331
+    },
+    {
+      "epoch": 7.590649942987457,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.543437512137717e-07,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8828289657831192,
+      "num_tokens": 356327873.0,
+      "step": 3332
+    },
+    {
+      "epoch": 7.592930444697833,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.529959953399455e-07,
+      "loss": 0.5503,
+      "mean_token_accuracy": 0.8852836340665817,
+      "num_tokens": 356435610.0,
+      "step": 3333
+    },
+    {
+      "epoch": 7.59521094640821,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.516492310318643e-07,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8795439153909683,
+      "num_tokens": 356542182.0,
+      "step": 3334
+    },
+    {
+      "epoch": 7.5974914481185865,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.503034590539266e-07,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.88029345870018,
+      "num_tokens": 356648911.0,
+      "step": 3335
+    },
+    {
+      "epoch": 7.599771949828963,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.489586801699661e-07,
+      "loss": 0.5624,
+      "mean_token_accuracy": 0.8867314010858536,
+      "num_tokens": 356755662.0,
+      "step": 3336
+    },
+    {
+      "epoch": 7.602052451539339,
+      "grad_norm": 2.265625,
+      "learning_rate": 7.476148951432543e-07,
+      "loss": 0.5522,
+      "mean_token_accuracy": 0.8867203295230865,
+      "num_tokens": 356862489.0,
+      "step": 3337
+    },
+    {
+      "epoch": 7.604332953249715,
+      "grad_norm": 3.375,
+      "learning_rate": 7.462721047364965e-07,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8798790127038956,
+      "num_tokens": 356969949.0,
+      "step": 3338
+    },
+    {
+      "epoch": 7.606613454960091,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.449303097118355e-07,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8800079971551895,
+      "num_tokens": 357077056.0,
+      "step": 3339
+    },
+    {
+      "epoch": 7.608893956670467,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.435895108308472e-07,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8835026770830154,
+      "num_tokens": 357184699.0,
+      "step": 3340
+    },
+    {
+      "epoch": 7.611174458380844,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.422497088545436e-07,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8809027224779129,
+      "num_tokens": 357291871.0,
+      "step": 3341
+    },
+    {
+      "epoch": 7.61345496009122,
+      "grad_norm": 2.578125,
+      "learning_rate": 7.409109045433704e-07,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8787789940834045,
+      "num_tokens": 357399018.0,
+      "step": 3342
+    },
+    {
+      "epoch": 7.615735461801596,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.395730986572075e-07,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.8798029869794846,
+      "num_tokens": 357505735.0,
+      "step": 3343
+    },
+    {
+      "epoch": 7.618015963511972,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.382362919553682e-07,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8806295096874237,
+      "num_tokens": 357613067.0,
+      "step": 3344
+    },
+    {
+      "epoch": 7.620296465222349,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.369004851965966e-07,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8788366466760635,
+      "num_tokens": 357720663.0,
+      "step": 3345
+    },
+    {
+      "epoch": 7.6225769669327255,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.355656791390717e-07,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8791816085577011,
+      "num_tokens": 357828239.0,
+      "step": 3346
+    },
+    {
+      "epoch": 7.624857468643102,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.342318745404034e-07,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8827292621135712,
+      "num_tokens": 357935648.0,
+      "step": 3347
+    },
+    {
+      "epoch": 7.627137970353478,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.32899072157634e-07,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.8853467255830765,
+      "num_tokens": 358042684.0,
+      "step": 3348
+    },
+    {
+      "epoch": 7.629418472063854,
+      "grad_norm": 4.96875,
+      "learning_rate": 7.315672727472365e-07,
+      "loss": 0.5439,
+      "mean_token_accuracy": 0.8852219432592392,
+      "num_tokens": 358150191.0,
+      "step": 3349
+    },
+    {
+      "epoch": 7.63169897377423,
+      "grad_norm": 3.125,
+      "learning_rate": 7.302364770651132e-07,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.888086274266243,
+      "num_tokens": 358257157.0,
+      "step": 3350
+    },
+    {
+      "epoch": 7.633979475484606,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.289066858665991e-07,
+      "loss": 0.5475,
+      "mean_token_accuracy": 0.8860320299863815,
+      "num_tokens": 358364536.0,
+      "step": 3351
+    },
+    {
+      "epoch": 7.636259977194983,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.275778999064578e-07,
+      "loss": 0.5927,
+      "mean_token_accuracy": 0.8757172375917435,
+      "num_tokens": 358471817.0,
+      "step": 3352
+    },
+    {
+      "epoch": 7.638540478905359,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.262501199388827e-07,
+      "loss": 0.5552,
+      "mean_token_accuracy": 0.8869472742080688,
+      "num_tokens": 358578963.0,
+      "step": 3353
+    },
+    {
+      "epoch": 7.640820980615736,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.249233467174965e-07,
+      "loss": 0.5882,
+      "mean_token_accuracy": 0.8762264847755432,
+      "num_tokens": 358685361.0,
+      "step": 3354
+    },
+    {
+      "epoch": 7.643101482326112,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.235975809953491e-07,
+      "loss": 0.55,
+      "mean_token_accuracy": 0.8874974101781845,
+      "num_tokens": 358792677.0,
+      "step": 3355
+    },
+    {
+      "epoch": 7.645381984036488,
+      "grad_norm": 4.125,
+      "learning_rate": 7.222728235249196e-07,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8794031143188477,
+      "num_tokens": 358899865.0,
+      "step": 3356
+    },
+    {
+      "epoch": 7.6476624857468645,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.209490750581152e-07,
+      "loss": 0.553,
+      "mean_token_accuracy": 0.8865524530410767,
+      "num_tokens": 359006856.0,
+      "step": 3357
+    },
+    {
+      "epoch": 7.649942987457241,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.196263363462699e-07,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.88320592045784,
+      "num_tokens": 359113787.0,
+      "step": 3358
+    },
+    {
+      "epoch": 7.652223489167617,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.183046081401454e-07,
+      "loss": 0.565,
+      "mean_token_accuracy": 0.8821183294057846,
+      "num_tokens": 359220641.0,
+      "step": 3359
+    },
+    {
+      "epoch": 7.654503990877993,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.169838911899276e-07,
+      "loss": 0.5597,
+      "mean_token_accuracy": 0.8826924860477448,
+      "num_tokens": 359327949.0,
+      "step": 3360
+    },
+    {
+      "epoch": 7.656784492588369,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.156641862452316e-07,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8792467415332794,
+      "num_tokens": 359434539.0,
+      "step": 3361
+    },
+    {
+      "epoch": 7.659064994298745,
+      "grad_norm": 5.59375,
+      "learning_rate": 7.143454940550948e-07,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8802933245897293,
+      "num_tokens": 359541747.0,
+      "step": 3362
+    },
+    {
+      "epoch": 7.661345496009122,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.13027815367982e-07,
+      "loss": 0.5676,
+      "mean_token_accuracy": 0.8832086026668549,
+      "num_tokens": 359648549.0,
+      "step": 3363
+    },
+    {
+      "epoch": 7.663625997719498,
+      "grad_norm": 4.5,
+      "learning_rate": 7.117111509317823e-07,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8798936158418655,
+      "num_tokens": 359755583.0,
+      "step": 3364
+    },
+    {
+      "epoch": 7.665906499429875,
+      "grad_norm": 5.09375,
+      "learning_rate": 7.103955014938099e-07,
+      "loss": 0.5509,
+      "mean_token_accuracy": 0.885413408279419,
+      "num_tokens": 359862616.0,
+      "step": 3365
+    },
+    {
+      "epoch": 7.668187001140251,
+      "grad_norm": 3.125,
+      "learning_rate": 7.090808678008005e-07,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8743928670883179,
+      "num_tokens": 359969068.0,
+      "step": 3366
+    },
+    {
+      "epoch": 7.670467502850627,
+      "grad_norm": 4.375,
+      "learning_rate": 7.077672505989155e-07,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8788380324840546,
+      "num_tokens": 360076499.0,
+      "step": 3367
+    },
+    {
+      "epoch": 7.6727480045610035,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.064546506337386e-07,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8831783980131149,
+      "num_tokens": 360183723.0,
+      "step": 3368
+    },
+    {
+      "epoch": 7.67502850627138,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.051430686502764e-07,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8796802908182144,
+      "num_tokens": 360291600.0,
+      "step": 3369
+    },
+    {
+      "epoch": 7.677309007981756,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.038325053929582e-07,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.8816768825054169,
+      "num_tokens": 360398736.0,
+      "step": 3370
+    },
+    {
+      "epoch": 7.679589509692132,
+      "grad_norm": 4.3125,
+      "learning_rate": 7.025229616056326e-07,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.8814764469861984,
+      "num_tokens": 360505355.0,
+      "step": 3371
+    },
+    {
+      "epoch": 7.681870011402508,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.012144380315724e-07,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8859875947237015,
+      "num_tokens": 360612219.0,
+      "step": 3372
+    },
+    {
+      "epoch": 7.684150513112884,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.999069354134703e-07,
+      "loss": 0.5471,
+      "mean_token_accuracy": 0.8861010521650314,
+      "num_tokens": 360720165.0,
+      "step": 3373
+    },
+    {
+      "epoch": 7.6864310148232615,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.986004544934394e-07,
+      "loss": 0.588,
+      "mean_token_accuracy": 0.8770526051521301,
+      "num_tokens": 360827091.0,
+      "step": 3374
+    },
+    {
+      "epoch": 7.688711516533638,
+      "grad_norm": 3.0,
+      "learning_rate": 6.972949960130135e-07,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8841615915298462,
+      "num_tokens": 360934168.0,
+      "step": 3375
+    },
+    {
+      "epoch": 7.690992018244014,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.959905607131457e-07,
+      "loss": 0.5652,
+      "mean_token_accuracy": 0.8876559138298035,
+      "num_tokens": 361040859.0,
+      "step": 3376
+    },
+    {
+      "epoch": 7.69327251995439,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.946871493342072e-07,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8838866651058197,
+      "num_tokens": 361148098.0,
+      "step": 3377
+    },
+    {
+      "epoch": 7.695553021664766,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.933847626159898e-07,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8839807361364365,
+      "num_tokens": 361255309.0,
+      "step": 3378
+    },
+    {
+      "epoch": 7.6978335233751425,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.920834012977032e-07,
+      "loss": 0.5935,
+      "mean_token_accuracy": 0.878353163599968,
+      "num_tokens": 361362125.0,
+      "step": 3379
+    },
+    {
+      "epoch": 7.700114025085519,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.907830661179757e-07,
+      "loss": 0.555,
+      "mean_token_accuracy": 0.8843338489532471,
+      "num_tokens": 361468866.0,
+      "step": 3380
+    },
+    {
+      "epoch": 7.702394526795895,
+      "grad_norm": 2.5,
+      "learning_rate": 6.894837578148505e-07,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8871625512838364,
+      "num_tokens": 361576767.0,
+      "step": 3381
+    },
+    {
+      "epoch": 7.704675028506271,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.881854771257912e-07,
+      "loss": 0.5575,
+      "mean_token_accuracy": 0.8831998556852341,
+      "num_tokens": 361683691.0,
+      "step": 3382
+    },
+    {
+      "epoch": 7.706955530216648,
+      "grad_norm": 3.125,
+      "learning_rate": 6.868882247876776e-07,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8782538622617722,
+      "num_tokens": 361790388.0,
+      "step": 3383
+    },
+    {
+      "epoch": 7.7092360319270234,
+      "grad_norm": 3.75,
+      "learning_rate": 6.855920015368032e-07,
+      "loss": 0.5964,
+      "mean_token_accuracy": 0.8798200190067291,
+      "num_tokens": 361897277.0,
+      "step": 3384
+    },
+    {
+      "epoch": 7.7115165336374005,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.8429680810888e-07,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.882276862859726,
+      "num_tokens": 362003897.0,
+      "step": 3385
+    },
+    {
+      "epoch": 7.713797035347777,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.830026452390354e-07,
+      "loss": 0.5599,
+      "mean_token_accuracy": 0.8820279538631439,
+      "num_tokens": 362111309.0,
+      "step": 3386
+    },
+    {
+      "epoch": 7.716077537058153,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.817095136618113e-07,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8817304819822311,
+      "num_tokens": 362218467.0,
+      "step": 3387
+    },
+    {
+      "epoch": 7.718358038768529,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.804174141111631e-07,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8817280679941177,
+      "num_tokens": 362326122.0,
+      "step": 3388
+    },
+    {
+      "epoch": 7.720638540478905,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.791263473204624e-07,
+      "loss": 0.586,
+      "mean_token_accuracy": 0.878828227519989,
+      "num_tokens": 362433157.0,
+      "step": 3389
+    },
+    {
+      "epoch": 7.7229190421892815,
+      "grad_norm": 2.625,
+      "learning_rate": 6.778363140224933e-07,
+      "loss": 0.5615,
+      "mean_token_accuracy": 0.8845448940992355,
+      "num_tokens": 362540668.0,
+      "step": 3390
+    },
+    {
+      "epoch": 7.725199543899658,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.765473149494545e-07,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.884161964058876,
+      "num_tokens": 362648198.0,
+      "step": 3391
+    },
+    {
+      "epoch": 7.727480045610034,
+      "grad_norm": 3.75,
+      "learning_rate": 6.752593508329572e-07,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8821451961994171,
+      "num_tokens": 362755339.0,
+      "step": 3392
+    },
+    {
+      "epoch": 7.72976054732041,
+      "grad_norm": 2.625,
+      "learning_rate": 6.739724224040236e-07,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8820271492004395,
+      "num_tokens": 362862783.0,
+      "step": 3393
+    },
+    {
+      "epoch": 7.732041049030787,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.726865303930905e-07,
+      "loss": 0.5809,
+      "mean_token_accuracy": 0.8791635781526566,
+      "num_tokens": 362969748.0,
+      "step": 3394
+    },
+    {
+      "epoch": 7.734321550741163,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.714016755300048e-07,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8794443905353546,
+      "num_tokens": 363076599.0,
+      "step": 3395
+    },
+    {
+      "epoch": 7.7366020524515395,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.701178585440257e-07,
+      "loss": 0.5509,
+      "mean_token_accuracy": 0.8836392611265182,
+      "num_tokens": 363184034.0,
+      "step": 3396
+    },
+    {
+      "epoch": 7.738882554161916,
+      "grad_norm": 2.75,
+      "learning_rate": 6.688350801638235e-07,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8802366703748703,
+      "num_tokens": 363290837.0,
+      "step": 3397
+    },
+    {
+      "epoch": 7.741163055872292,
+      "grad_norm": 4.84375,
+      "learning_rate": 6.67553341117477e-07,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8817736506462097,
+      "num_tokens": 363398189.0,
+      "step": 3398
+    },
+    {
+      "epoch": 7.743443557582668,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.662726421324775e-07,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8845831155776978,
+      "num_tokens": 363505198.0,
+      "step": 3399
+    },
+    {
+      "epoch": 7.745724059293044,
+      "grad_norm": 4.84375,
+      "learning_rate": 6.649929839357247e-07,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.8841813802719116,
+      "num_tokens": 363612622.0,
+      "step": 3400
+    },
+    {
+      "epoch": 7.7480045610034205,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.637143672535282e-07,
+      "loss": 0.561,
+      "mean_token_accuracy": 0.8850229233503342,
+      "num_tokens": 363719127.0,
+      "step": 3401
+    },
+    {
+      "epoch": 7.750285062713797,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.624367928116066e-07,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8808367401361465,
+      "num_tokens": 363826026.0,
+      "step": 3402
+    },
+    {
+      "epoch": 7.752565564424174,
+      "grad_norm": 4.65625,
+      "learning_rate": 6.611602613350854e-07,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8835535049438477,
+      "num_tokens": 363933444.0,
+      "step": 3403
+    },
+    {
+      "epoch": 7.75484606613455,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.598847735485001e-07,
+      "loss": 0.5656,
+      "mean_token_accuracy": 0.8848689496517181,
+      "num_tokens": 364040465.0,
+      "step": 3404
+    },
+    {
+      "epoch": 7.757126567844926,
+      "grad_norm": 3.921875,
+      "learning_rate": 6.586103301757918e-07,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.880999892950058,
+      "num_tokens": 364147442.0,
+      "step": 3405
+    },
+    {
+      "epoch": 7.759407069555302,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.573369319403108e-07,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.883205309510231,
+      "num_tokens": 364255015.0,
+      "step": 3406
+    },
+    {
+      "epoch": 7.7616875712656785,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.560645795648132e-07,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8813123106956482,
+      "num_tokens": 364362027.0,
+      "step": 3407
+    },
+    {
+      "epoch": 7.763968072976055,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.547932737714624e-07,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8851943165063858,
+      "num_tokens": 364469642.0,
+      "step": 3408
+    },
+    {
+      "epoch": 7.766248574686431,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.535230152818256e-07,
+      "loss": 0.5576,
+      "mean_token_accuracy": 0.8846663683652878,
+      "num_tokens": 364576740.0,
+      "step": 3409
+    },
+    {
+      "epoch": 7.768529076396807,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.522538048168777e-07,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8818811625242233,
+      "num_tokens": 364683980.0,
+      "step": 3410
+    },
+    {
+      "epoch": 7.770809578107183,
+      "grad_norm": 4.71875,
+      "learning_rate": 6.509856430969982e-07,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.8817369639873505,
+      "num_tokens": 364790604.0,
+      "step": 3411
+    },
+    {
+      "epoch": 7.7730900798175595,
+      "grad_norm": 4.71875,
+      "learning_rate": 6.49718530841971e-07,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.876135990023613,
+      "num_tokens": 364897881.0,
+      "step": 3412
+    },
+    {
+      "epoch": 7.775370581527936,
+      "grad_norm": 4.375,
+      "learning_rate": 6.484524687709853e-07,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8819546699523926,
+      "num_tokens": 365004584.0,
+      "step": 3413
+    },
+    {
+      "epoch": 7.777651083238313,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.471874576026321e-07,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8798770606517792,
+      "num_tokens": 365111502.0,
+      "step": 3414
+    },
+    {
+      "epoch": 7.779931584948689,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.459234980549081e-07,
+      "loss": 0.5942,
+      "mean_token_accuracy": 0.8755921274423599,
+      "num_tokens": 365217932.0,
+      "step": 3415
+    },
+    {
+      "epoch": 7.782212086659065,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.446605908452122e-07,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8808364421129227,
+      "num_tokens": 365324803.0,
+      "step": 3416
+    },
+    {
+      "epoch": 7.784492588369441,
+      "grad_norm": 3.0,
+      "learning_rate": 6.433987366903461e-07,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8788796812295914,
+      "num_tokens": 365432029.0,
+      "step": 3417
+    },
+    {
+      "epoch": 7.7867730900798175,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.421379363065142e-07,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8842990696430206,
+      "num_tokens": 365539725.0,
+      "step": 3418
+    },
+    {
+      "epoch": 7.789053591790194,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.408781904093228e-07,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8797939419746399,
+      "num_tokens": 365646379.0,
+      "step": 3419
+    },
+    {
+      "epoch": 7.79133409350057,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.39619499713778e-07,
+      "loss": 0.5456,
+      "mean_token_accuracy": 0.8892961293458939,
+      "num_tokens": 365753657.0,
+      "step": 3420
+    },
+    {
+      "epoch": 7.793614595210946,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.383618649342894e-07,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8814832419157028,
+      "num_tokens": 365859878.0,
+      "step": 3421
+    },
+    {
+      "epoch": 7.795895096921322,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.371052867846658e-07,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8824429661035538,
+      "num_tokens": 365966802.0,
+      "step": 3422
+    },
+    {
+      "epoch": 7.798175598631699,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.358497659781177e-07,
+      "loss": 0.5513,
+      "mean_token_accuracy": 0.8872740119695663,
+      "num_tokens": 366073841.0,
+      "step": 3423
+    },
+    {
+      "epoch": 7.800456100342076,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.345953032272525e-07,
+      "loss": 0.5757,
+      "mean_token_accuracy": 0.8818953335285187,
+      "num_tokens": 366180970.0,
+      "step": 3424
+    },
+    {
+      "epoch": 7.802736602052452,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.333418992440804e-07,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8802995383739471,
+      "num_tokens": 366287962.0,
+      "step": 3425
+    },
+    {
+      "epoch": 7.805017103762828,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.3208955474001e-07,
+      "loss": 0.5699,
+      "mean_token_accuracy": 0.8855053037405014,
+      "num_tokens": 366395208.0,
+      "step": 3426
+    },
+    {
+      "epoch": 7.807297605473204,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.308382704258459e-07,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8811968117952347,
+      "num_tokens": 366502174.0,
+      "step": 3427
+    },
+    {
+      "epoch": 7.80957810718358,
+      "grad_norm": 5.4375,
+      "learning_rate": 6.29588047011794e-07,
+      "loss": 0.5882,
+      "mean_token_accuracy": 0.8795420229434967,
+      "num_tokens": 366609589.0,
+      "step": 3428
+    },
+    {
+      "epoch": 7.811858608893957,
+      "grad_norm": 2.375,
+      "learning_rate": 6.283388852074576e-07,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.8850322365760803,
+      "num_tokens": 366716534.0,
+      "step": 3429
+    },
+    {
+      "epoch": 7.814139110604333,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.270907857218356e-07,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.882201686501503,
+      "num_tokens": 366823931.0,
+      "step": 3430
+    },
+    {
+      "epoch": 7.816419612314709,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.258437492633254e-07,
+      "loss": 0.552,
+      "mean_token_accuracy": 0.887025773525238,
+      "num_tokens": 366930916.0,
+      "step": 3431
+    },
+    {
+      "epoch": 7.818700114025085,
+      "grad_norm": 2.484375,
+      "learning_rate": 6.245977765397216e-07,
+      "loss": 0.5631,
+      "mean_token_accuracy": 0.8866972625255585,
+      "num_tokens": 367038203.0,
+      "step": 3432
+    },
+    {
+      "epoch": 7.820980615735461,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.233528682582132e-07,
+      "loss": 0.5618,
+      "mean_token_accuracy": 0.8818273693323135,
+      "num_tokens": 367145120.0,
+      "step": 3433
+    },
+    {
+      "epoch": 7.823261117445838,
+      "grad_norm": 3.796875,
+      "learning_rate": 6.221090251253872e-07,
+      "loss": 0.5571,
+      "mean_token_accuracy": 0.8845161199569702,
+      "num_tokens": 367252872.0,
+      "step": 3434
+    },
+    {
+      "epoch": 7.825541619156215,
+      "grad_norm": 2.625,
+      "learning_rate": 6.208662478472249e-07,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.8831181675195694,
+      "num_tokens": 367359928.0,
+      "step": 3435
+    },
+    {
+      "epoch": 7.827822120866591,
+      "grad_norm": 2.53125,
+      "learning_rate": 6.196245371291015e-07,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8835492432117462,
+      "num_tokens": 367467008.0,
+      "step": 3436
+    },
+    {
+      "epoch": 7.830102622576967,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.183838936757891e-07,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8792263269424438,
+      "num_tokens": 367574251.0,
+      "step": 3437
+    },
+    {
+      "epoch": 7.832383124287343,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.171443181914524e-07,
+      "loss": 0.5484,
+      "mean_token_accuracy": 0.8878850936889648,
+      "num_tokens": 367682131.0,
+      "step": 3438
+    },
+    {
+      "epoch": 7.834663625997719,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.159058113796507e-07,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8801481425762177,
+      "num_tokens": 367789432.0,
+      "step": 3439
+    },
+    {
+      "epoch": 7.836944127708096,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.146683739433374e-07,
+      "loss": 0.5901,
+      "mean_token_accuracy": 0.8798154145479202,
+      "num_tokens": 367896299.0,
+      "step": 3440
+    },
+    {
+      "epoch": 7.839224629418472,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.134320065848564e-07,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8814557641744614,
+      "num_tokens": 368003358.0,
+      "step": 3441
+    },
+    {
+      "epoch": 7.841505131128848,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.121967100059473e-07,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.8781363666057587,
+      "num_tokens": 368110282.0,
+      "step": 3442
+    },
+    {
+      "epoch": 7.843785632839225,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.109624849077397e-07,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8851889967918396,
+      "num_tokens": 368217316.0,
+      "step": 3443
+    },
+    {
+      "epoch": 7.846066134549601,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.097293319907566e-07,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8824188113212585,
+      "num_tokens": 368324078.0,
+      "step": 3444
+    },
+    {
+      "epoch": 7.848346636259977,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.084972519549123e-07,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8789727538824081,
+      "num_tokens": 368430695.0,
+      "step": 3445
+    },
+    {
+      "epoch": 7.850627137970354,
+      "grad_norm": 4.78125,
+      "learning_rate": 6.072662454995101e-07,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.8788609057664871,
+      "num_tokens": 368537559.0,
+      "step": 3446
+    },
+    {
+      "epoch": 7.85290763968073,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.060363133232472e-07,
+      "loss": 0.5585,
+      "mean_token_accuracy": 0.8846858441829681,
+      "num_tokens": 368645105.0,
+      "step": 3447
+    },
+    {
+      "epoch": 7.855188141391106,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.048074561242076e-07,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8781778067350388,
+      "num_tokens": 368752515.0,
+      "step": 3448
+    },
+    {
+      "epoch": 7.857468643101482,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.035796745998679e-07,
+      "loss": 0.5448,
+      "mean_token_accuracy": 0.8889462649822235,
+      "num_tokens": 368859963.0,
+      "step": 3449
+    },
+    {
+      "epoch": 7.859749144811858,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.023529694470931e-07,
+      "loss": 0.568,
+      "mean_token_accuracy": 0.8810675591230392,
+      "num_tokens": 368966665.0,
+      "step": 3450
+    },
+    {
+      "epoch": 7.862029646522235,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.01127341362138e-07,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8824652433395386,
+      "num_tokens": 369073545.0,
+      "step": 3451
+    },
+    {
+      "epoch": 7.864310148232612,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.999027910406441e-07,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.8798021376132965,
+      "num_tokens": 369180613.0,
+      "step": 3452
+    },
+    {
+      "epoch": 7.866590649942988,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.98679319177643e-07,
+      "loss": 0.5666,
+      "mean_token_accuracy": 0.8812180906534195,
+      "num_tokens": 369287568.0,
+      "step": 3453
+    },
+    {
+      "epoch": 7.868871151653364,
+      "grad_norm": 3.125,
+      "learning_rate": 5.974569264675542e-07,
+      "loss": 0.5585,
+      "mean_token_accuracy": 0.8855846971273422,
+      "num_tokens": 369394613.0,
+      "step": 3454
+    },
+    {
+      "epoch": 7.87115165336374,
+      "grad_norm": 3.25,
+      "learning_rate": 5.962356136041835e-07,
+      "loss": 0.5601,
+      "mean_token_accuracy": 0.8867956697940826,
+      "num_tokens": 369501322.0,
+      "step": 3455
+    },
+    {
+      "epoch": 7.873432155074116,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.95015381280726e-07,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8840297311544418,
+      "num_tokens": 369608680.0,
+      "step": 3456
+    },
+    {
+      "epoch": 7.875712656784493,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.937962301897604e-07,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.8828248530626297,
+      "num_tokens": 369716370.0,
+      "step": 3457
+    },
+    {
+      "epoch": 7.877993158494869,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.925781610232534e-07,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.8825751543045044,
+      "num_tokens": 369823697.0,
+      "step": 3458
+    },
+    {
+      "epoch": 7.880273660205245,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.913611744725584e-07,
+      "loss": 0.5753,
+      "mean_token_accuracy": 0.8807247132062912,
+      "num_tokens": 369931399.0,
+      "step": 3459
+    },
+    {
+      "epoch": 7.882554161915621,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.901452712284128e-07,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.881399855017662,
+      "num_tokens": 370038799.0,
+      "step": 3460
+    },
+    {
+      "epoch": 7.884834663625997,
+      "grad_norm": 5.0,
+      "learning_rate": 5.889304519809402e-07,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8788302540779114,
+      "num_tokens": 370146637.0,
+      "step": 3461
+    },
+    {
+      "epoch": 7.887115165336374,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.877167174196491e-07,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8876172602176666,
+      "num_tokens": 370253754.0,
+      "step": 3462
+    },
+    {
+      "epoch": 7.889395667046751,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.865040682334303e-07,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.8796623796224594,
+      "num_tokens": 370360287.0,
+      "step": 3463
+    },
+    {
+      "epoch": 7.891676168757127,
+      "grad_norm": 3.875,
+      "learning_rate": 5.852925051105609e-07,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8807369768619537,
+      "num_tokens": 370467034.0,
+      "step": 3464
+    },
+    {
+      "epoch": 7.893956670467503,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.840820287387009e-07,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8808783739805222,
+      "num_tokens": 370573928.0,
+      "step": 3465
+    },
+    {
+      "epoch": 7.896237172177879,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.828726398048939e-07,
+      "loss": 0.5845,
+      "mean_token_accuracy": 0.8822729289531708,
+      "num_tokens": 370680253.0,
+      "step": 3466
+    },
+    {
+      "epoch": 7.898517673888255,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.816643389955642e-07,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8849064260721207,
+      "num_tokens": 370786997.0,
+      "step": 3467
+    },
+    {
+      "epoch": 7.900798175598632,
+      "grad_norm": 3.5,
+      "learning_rate": 5.804571269965206e-07,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8811779320240021,
+      "num_tokens": 370893845.0,
+      "step": 3468
+    },
+    {
+      "epoch": 7.903078677309008,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.792510044929545e-07,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8799442201852798,
+      "num_tokens": 371000695.0,
+      "step": 3469
+    },
+    {
+      "epoch": 7.905359179019384,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.780459721694359e-07,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.8812065422534943,
+      "num_tokens": 371107589.0,
+      "step": 3470
+    },
+    {
+      "epoch": 7.90763968072976,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.768420307099188e-07,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8848077207803726,
+      "num_tokens": 371214705.0,
+      "step": 3471
+    },
+    {
+      "epoch": 7.909920182440137,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.756391807977377e-07,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8829392790794373,
+      "num_tokens": 371321954.0,
+      "step": 3472
+    },
+    {
+      "epoch": 7.9122006841505135,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.744374231156056e-07,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.87844018638134,
+      "num_tokens": 371428415.0,
+      "step": 3473
+    },
+    {
+      "epoch": 7.91448118586089,
+      "grad_norm": 4.96875,
+      "learning_rate": 5.732367583456177e-07,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.879816859960556,
+      "num_tokens": 371535322.0,
+      "step": 3474
+    },
+    {
+      "epoch": 7.916761687571266,
+      "grad_norm": 2.625,
+      "learning_rate": 5.720371871692484e-07,
+      "loss": 0.5599,
+      "mean_token_accuracy": 0.8838074952363968,
+      "num_tokens": 371642326.0,
+      "step": 3475
+    },
+    {
+      "epoch": 7.919042189281642,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.708387102673507e-07,
+      "loss": 0.5495,
+      "mean_token_accuracy": 0.8860156238079071,
+      "num_tokens": 371749387.0,
+      "step": 3476
+    },
+    {
+      "epoch": 7.921322690992018,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.696413283201571e-07,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8818910270929337,
+      "num_tokens": 371856387.0,
+      "step": 3477
+    },
+    {
+      "epoch": 7.923603192702394,
+      "grad_norm": 2.75,
+      "learning_rate": 5.684450420072792e-07,
+      "loss": 0.5595,
+      "mean_token_accuracy": 0.8838623762130737,
+      "num_tokens": 371963635.0,
+      "step": 3478
+    },
+    {
+      "epoch": 7.925883694412771,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.67249852007705e-07,
+      "loss": 0.5367,
+      "mean_token_accuracy": 0.8921998292207718,
+      "num_tokens": 372071179.0,
+      "step": 3479
+    },
+    {
+      "epoch": 7.928164196123147,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.660557589998014e-07,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.88170425593853,
+      "num_tokens": 372178998.0,
+      "step": 3480
+    },
+    {
+      "epoch": 7.930444697833523,
+      "grad_norm": 4.84375,
+      "learning_rate": 5.648627636613127e-07,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.8786900341510773,
+      "num_tokens": 372286187.0,
+      "step": 3481
+    },
+    {
+      "epoch": 7.932725199543899,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.636708666693599e-07,
+      "loss": 0.5475,
+      "mean_token_accuracy": 0.887066513299942,
+      "num_tokens": 372393703.0,
+      "step": 3482
+    },
+    {
+      "epoch": 7.935005701254276,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.62480068700442e-07,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.879653736948967,
+      "num_tokens": 372500635.0,
+      "step": 3483
+    },
+    {
+      "epoch": 7.9372862029646525,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.612903704304309e-07,
+      "loss": 0.5343,
+      "mean_token_accuracy": 0.8889681696891785,
+      "num_tokens": 372608260.0,
+      "step": 3484
+    },
+    {
+      "epoch": 7.939566704675029,
+      "grad_norm": 3.984375,
+      "learning_rate": 5.601017725345772e-07,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.8840508311986923,
+      "num_tokens": 372715730.0,
+      "step": 3485
+    },
+    {
+      "epoch": 7.941847206385405,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.589142756875065e-07,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.8822902888059616,
+      "num_tokens": 372822393.0,
+      "step": 3486
+    },
+    {
+      "epoch": 7.944127708095781,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.577278805632186e-07,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8777175396680832,
+      "num_tokens": 372929636.0,
+      "step": 3487
+    },
+    {
+      "epoch": 7.946408209806157,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.565425878350895e-07,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8786440938711166,
+      "num_tokens": 373036503.0,
+      "step": 3488
+    },
+    {
+      "epoch": 7.9486887115165334,
+      "grad_norm": 2.34375,
+      "learning_rate": 5.553583981758668e-07,
+      "loss": 0.5385,
+      "mean_token_accuracy": 0.8891723304986954,
+      "num_tokens": 373143771.0,
+      "step": 3489
+    },
+    {
+      "epoch": 7.95096921322691,
+      "grad_norm": 2.625,
+      "learning_rate": 5.541753122576746e-07,
+      "loss": 0.5589,
+      "mean_token_accuracy": 0.8819065690040588,
+      "num_tokens": 373250701.0,
+      "step": 3490
+    },
+    {
+      "epoch": 7.953249714937286,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.529933307520102e-07,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8800751566886902,
+      "num_tokens": 373357804.0,
+      "step": 3491
+    },
+    {
+      "epoch": 7.955530216647663,
+      "grad_norm": 2.75,
+      "learning_rate": 5.518124543297423e-07,
+      "loss": 0.5565,
+      "mean_token_accuracy": 0.8874002993106842,
+      "num_tokens": 373464625.0,
+      "step": 3492
+    },
+    {
+      "epoch": 7.957810718358039,
+      "grad_norm": 4.09375,
+      "learning_rate": 5.506326836611139e-07,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8810268938541412,
+      "num_tokens": 373572124.0,
+      "step": 3493
+    },
+    {
+      "epoch": 7.960091220068415,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.494540194157411e-07,
+      "loss": 0.5765,
+      "mean_token_accuracy": 0.8803278654813766,
+      "num_tokens": 373678778.0,
+      "step": 3494
+    },
+    {
+      "epoch": 7.9623717217787915,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.482764622626094e-07,
+      "loss": 0.554,
+      "mean_token_accuracy": 0.8871753364801407,
+      "num_tokens": 373785455.0,
+      "step": 3495
+    },
+    {
+      "epoch": 7.964652223489168,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.471000128700784e-07,
+      "loss": 0.5617,
+      "mean_token_accuracy": 0.8817069977521896,
+      "num_tokens": 373892856.0,
+      "step": 3496
+    },
+    {
+      "epoch": 7.966932725199544,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.459246719058778e-07,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.8817310035228729,
+      "num_tokens": 374000078.0,
+      "step": 3497
+    },
+    {
+      "epoch": 7.96921322690992,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.447504400371084e-07,
+      "loss": 0.5618,
+      "mean_token_accuracy": 0.8855457454919815,
+      "num_tokens": 374107031.0,
+      "step": 3498
+    },
+    {
+      "epoch": 7.971493728620296,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.435773179302426e-07,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8799401372671127,
+      "num_tokens": 374213605.0,
+      "step": 3499
+    },
+    {
+      "epoch": 7.9737742303306725,
+      "grad_norm": 2.75,
+      "learning_rate": 5.4240530625112e-07,
+      "loss": 0.5501,
+      "mean_token_accuracy": 0.885622963309288,
+      "num_tokens": 374321608.0,
+      "step": 3500
+    },
+    {
+      "epoch": 7.976054732041049,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.412344056649527e-07,
+      "loss": 0.5896,
+      "mean_token_accuracy": 0.8806732445955276,
+      "num_tokens": 374427692.0,
+      "step": 3501
+    },
+    {
+      "epoch": 7.978335233751425,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.400646168363216e-07,
+      "loss": 0.5627,
+      "mean_token_accuracy": 0.885017529129982,
+      "num_tokens": 374534923.0,
+      "step": 3502
+    },
+    {
+      "epoch": 7.980615735461802,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.388959404291757e-07,
+      "loss": 0.5489,
+      "mean_token_accuracy": 0.8873688131570816,
+      "num_tokens": 374641673.0,
+      "step": 3503
+    },
+    {
+      "epoch": 7.982896237172178,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.377283771068342e-07,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8803865015506744,
+      "num_tokens": 374748674.0,
+      "step": 3504
+    },
+    {
+      "epoch": 7.985176738882554,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.365619275319823e-07,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.879953607916832,
+      "num_tokens": 374855343.0,
+      "step": 3505
+    },
+    {
+      "epoch": 7.9874572405929305,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.353965923666743e-07,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8788940012454987,
+      "num_tokens": 374962436.0,
+      "step": 3506
+    },
+    {
+      "epoch": 7.989737742303307,
+      "grad_norm": 2.75,
+      "learning_rate": 5.342323722723324e-07,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8827593922615051,
+      "num_tokens": 375068957.0,
+      "step": 3507
+    },
+    {
+      "epoch": 7.992018244013683,
+      "grad_norm": 3.125,
+      "learning_rate": 5.330692679097457e-07,
+      "loss": 0.5441,
+      "mean_token_accuracy": 0.8856361508369446,
+      "num_tokens": 375176504.0,
+      "step": 3508
+    },
+    {
+      "epoch": 7.994298745724059,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.319072799390693e-07,
+      "loss": 0.5954,
+      "mean_token_accuracy": 0.8784863501787186,
+      "num_tokens": 375283610.0,
+      "step": 3509
+    },
+    {
+      "epoch": 7.996579247434435,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.307464090198258e-07,
+      "loss": 0.5606,
+      "mean_token_accuracy": 0.8854129165410995,
+      "num_tokens": 375390831.0,
+      "step": 3510
+    },
+    {
+      "epoch": 7.9988597491448115,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.295866558109023e-07,
+      "loss": 0.5461,
+      "mean_token_accuracy": 0.8877543658018112,
+      "num_tokens": 375498660.0,
+      "step": 3511
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 5.6875,
+      "learning_rate": 5.284280209705531e-07,
+      "loss": 0.5913,
+      "mean_token_accuracy": 0.8680516481399536,
+      "num_tokens": 375537856.0,
+      "step": 3512
+    },
+    {
+      "epoch": 8.002280501710377,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.272705051563959e-07,
+      "loss": 0.5911,
+      "mean_token_accuracy": 0.8796858638525009,
+      "num_tokens": 375645015.0,
+      "step": 3513
+    },
+    {
+      "epoch": 8.004561003420752,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.261141090254149e-07,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.880715399980545,
+      "num_tokens": 375752133.0,
+      "step": 3514
+    },
+    {
+      "epoch": 8.00684150513113,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.249588332339589e-07,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.8854578584432602,
+      "num_tokens": 375859485.0,
+      "step": 3515
+    },
+    {
+      "epoch": 8.009122006841505,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.238046784377388e-07,
+      "loss": 0.5818,
+      "mean_token_accuracy": 0.8783604055643082,
+      "num_tokens": 375965833.0,
+      "step": 3516
+    },
+    {
+      "epoch": 8.011402508551882,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.226516452918315e-07,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8814006298780441,
+      "num_tokens": 376072854.0,
+      "step": 3517
+    },
+    {
+      "epoch": 8.013683010262257,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.214997344506758e-07,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.884096547961235,
+      "num_tokens": 376179522.0,
+      "step": 3518
+    },
+    {
+      "epoch": 8.015963511972634,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.203489465680747e-07,
+      "loss": 0.5921,
+      "mean_token_accuracy": 0.8821551650762558,
+      "num_tokens": 376286201.0,
+      "step": 3519
+    },
+    {
+      "epoch": 8.01824401368301,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.19199282297193e-07,
+      "loss": 0.5474,
+      "mean_token_accuracy": 0.8864544034004211,
+      "num_tokens": 376394233.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.01824401368301,
+      "eval_loss": 0.5862906575202942,
+      "eval_mean_token_accuracy": 0.8799242350085154,
+      "eval_num_tokens": 376394233.0,
+      "eval_runtime": 58.678,
+      "eval_samples_per_second": 142.899,
+      "eval_steps_per_second": 4.482,
+      "step": 3520
+    },
+    {
+      "epoch": 8.020524515393387,
+      "grad_norm": 2.34375,
+      "learning_rate": 5.180507422905585e-07,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8834462761878967,
+      "num_tokens": 376501043.0,
+      "step": 3521
+    },
+    {
+      "epoch": 8.022805017103764,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.169033272000587e-07,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8795521408319473,
+      "num_tokens": 376607841.0,
+      "step": 3522
+    },
+    {
+      "epoch": 8.025085518814139,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.157570376769452e-07,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8829406499862671,
+      "num_tokens": 376714615.0,
+      "step": 3523
+    },
+    {
+      "epoch": 8.027366020524516,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.146118743718301e-07,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.885052427649498,
+      "num_tokens": 376822883.0,
+      "step": 3524
+    },
+    {
+      "epoch": 8.029646522234891,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.134678379346856e-07,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8828334510326385,
+      "num_tokens": 376929625.0,
+      "step": 3525
+    },
+    {
+      "epoch": 8.031927023945268,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.123249290148452e-07,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8789701759815216,
+      "num_tokens": 377036782.0,
+      "step": 3526
+    },
+    {
+      "epoch": 8.034207525655644,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.111831482610011e-07,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8839947730302811,
+      "num_tokens": 377143972.0,
+      "step": 3527
+    },
+    {
+      "epoch": 8.03648802736602,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.100424963212064e-07,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8813960254192352,
+      "num_tokens": 377250375.0,
+      "step": 3528
+    },
+    {
+      "epoch": 8.038768529076396,
+      "grad_norm": 2.46875,
+      "learning_rate": 5.089029738428733e-07,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8840717822313309,
+      "num_tokens": 377357434.0,
+      "step": 3529
+    },
+    {
+      "epoch": 8.041049030786773,
+      "grad_norm": 3.90625,
+      "learning_rate": 5.077645814727725e-07,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8795655369758606,
+      "num_tokens": 377464466.0,
+      "step": 3530
+    },
+    {
+      "epoch": 8.043329532497149,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.066273198570343e-07,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8808377087116241,
+      "num_tokens": 377570549.0,
+      "step": 3531
+    },
+    {
+      "epoch": 8.045610034207526,
+      "grad_norm": 4.09375,
+      "learning_rate": 5.054911896411452e-07,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8835866749286652,
+      "num_tokens": 377678050.0,
+      "step": 3532
+    },
+    {
+      "epoch": 8.047890535917903,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.043561914699513e-07,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8832181841135025,
+      "num_tokens": 377785329.0,
+      "step": 3533
+    },
+    {
+      "epoch": 8.050171037628278,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.032223259876565e-07,
+      "loss": 0.5881,
+      "mean_token_accuracy": 0.8773612827062607,
+      "num_tokens": 377892051.0,
+      "step": 3534
+    },
+    {
+      "epoch": 8.052451539338655,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.020895938378195e-07,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8818919360637665,
+      "num_tokens": 377999467.0,
+      "step": 3535
+    },
+    {
+      "epoch": 8.05473204104903,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.009579956633578e-07,
+      "loss": 0.5568,
+      "mean_token_accuracy": 0.884328693151474,
+      "num_tokens": 378106965.0,
+      "step": 3536
+    },
+    {
+      "epoch": 8.057012542759407,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.998275321065454e-07,
+      "loss": 0.5704,
+      "mean_token_accuracy": 0.8801892846822739,
+      "num_tokens": 378214261.0,
+      "step": 3537
+    },
+    {
+      "epoch": 8.059293044469783,
+      "grad_norm": 3.5,
+      "learning_rate": 4.986982038090104e-07,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8822603821754456,
+      "num_tokens": 378320449.0,
+      "step": 3538
+    },
+    {
+      "epoch": 8.06157354618016,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.975700114117385e-07,
+      "loss": 0.5418,
+      "mean_token_accuracy": 0.8873136639595032,
+      "num_tokens": 378427513.0,
+      "step": 3539
+    },
+    {
+      "epoch": 8.063854047890535,
+      "grad_norm": 2.875,
+      "learning_rate": 4.964429555550693e-07,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.8837717622518539,
+      "num_tokens": 378534328.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.066134549600912,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.953170368786985e-07,
+      "loss": 0.5401,
+      "mean_token_accuracy": 0.8866355121135712,
+      "num_tokens": 378641467.0,
+      "step": 3541
+    },
+    {
+      "epoch": 8.06841505131129,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.941922560216764e-07,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8841693699359894,
+      "num_tokens": 378748281.0,
+      "step": 3542
+    },
+    {
+      "epoch": 8.070695553021665,
+      "grad_norm": 3.125,
+      "learning_rate": 4.930686136224056e-07,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8798070102930069,
+      "num_tokens": 378854984.0,
+      "step": 3543
+    },
+    {
+      "epoch": 8.072976054732042,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.91946110318644e-07,
+      "loss": 0.5871,
+      "mean_token_accuracy": 0.8762340992689133,
+      "num_tokens": 378962257.0,
+      "step": 3544
+    },
+    {
+      "epoch": 8.075256556442417,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.908247467475036e-07,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8846005648374557,
+      "num_tokens": 379069977.0,
+      "step": 3545
+    },
+    {
+      "epoch": 8.077537058152794,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.897045235454481e-07,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.882202997803688,
+      "num_tokens": 379177064.0,
+      "step": 3546
+    },
+    {
+      "epoch": 8.07981755986317,
+      "grad_norm": 5.0,
+      "learning_rate": 4.885854413482955e-07,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8858351409435272,
+      "num_tokens": 379284331.0,
+      "step": 3547
+    },
+    {
+      "epoch": 8.082098061573546,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.874675007912138e-07,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.8848386406898499,
+      "num_tokens": 379391861.0,
+      "step": 3548
+    },
+    {
+      "epoch": 8.084378563283922,
+      "grad_norm": 4.96875,
+      "learning_rate": 4.863507025087255e-07,
+      "loss": 0.5525,
+      "mean_token_accuracy": 0.8858330547809601,
+      "num_tokens": 379499126.0,
+      "step": 3549
+    },
+    {
+      "epoch": 8.086659064994299,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.852350471347031e-07,
+      "loss": 0.5846,
+      "mean_token_accuracy": 0.8791882544755936,
+      "num_tokens": 379605713.0,
+      "step": 3550
+    },
+    {
+      "epoch": 8.088939566704674,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.841205353023715e-07,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.881306603550911,
+      "num_tokens": 379712727.0,
+      "step": 3551
+    },
+    {
+      "epoch": 8.091220068415051,
+      "grad_norm": 2.875,
+      "learning_rate": 4.83007167644306e-07,
+      "loss": 0.5438,
+      "mean_token_accuracy": 0.8896623253822327,
+      "num_tokens": 379819603.0,
+      "step": 3552
+    },
+    {
+      "epoch": 8.093500570125428,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.818949447924334e-07,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.881472259759903,
+      "num_tokens": 379926447.0,
+      "step": 3553
+    },
+    {
+      "epoch": 8.095781071835804,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.807838673780282e-07,
+      "loss": 0.5636,
+      "mean_token_accuracy": 0.8816197216510773,
+      "num_tokens": 380033718.0,
+      "step": 3554
+    },
+    {
+      "epoch": 8.09806157354618,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.796739360317181e-07,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.8843680173158646,
+      "num_tokens": 380140392.0,
+      "step": 3555
+    },
+    {
+      "epoch": 8.100342075256556,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.785651513834774e-07,
+      "loss": 0.5553,
+      "mean_token_accuracy": 0.8834531456232071,
+      "num_tokens": 380247239.0,
+      "step": 3556
+    },
+    {
+      "epoch": 8.102622576966933,
+      "grad_norm": 4.78125,
+      "learning_rate": 4.774575140626317e-07,
+      "loss": 0.5562,
+      "mean_token_accuracy": 0.8883015066385269,
+      "num_tokens": 380354451.0,
+      "step": 3557
+    },
+    {
+      "epoch": 8.104903078677308,
+      "grad_norm": 4.125,
+      "learning_rate": 4.763510246978548e-07,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8845552057027817,
+      "num_tokens": 380462073.0,
+      "step": 3558
+    },
+    {
+      "epoch": 8.107183580387685,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.7524568391716736e-07,
+      "loss": 0.5809,
+      "mean_token_accuracy": 0.8840623497962952,
+      "num_tokens": 380569477.0,
+      "step": 3559
+    },
+    {
+      "epoch": 8.10946408209806,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.7414149234794064e-07,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8819562494754791,
+      "num_tokens": 380676862.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.111744583808438,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.7303845061689197e-07,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8820130825042725,
+      "num_tokens": 380783938.0,
+      "step": 3561
+    },
+    {
+      "epoch": 8.114025085518815,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.719365593500866e-07,
+      "loss": 0.5504,
+      "mean_token_accuracy": 0.8882433474063873,
+      "num_tokens": 380891146.0,
+      "step": 3562
+    },
+    {
+      "epoch": 8.11630558722919,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.7083581917293784e-07,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.8783487379550934,
+      "num_tokens": 380998117.0,
+      "step": 3563
+    },
+    {
+      "epoch": 8.118586088939567,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.6973623071020267e-07,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8855432868003845,
+      "num_tokens": 381104807.0,
+      "step": 3564
+    },
+    {
+      "epoch": 8.120866590649943,
+      "grad_norm": 2.75,
+      "learning_rate": 4.686377945859874e-07,
+      "loss": 0.5585,
+      "mean_token_accuracy": 0.8837299197912216,
+      "num_tokens": 381212129.0,
+      "step": 3565
+    },
+    {
+      "epoch": 8.12314709236032,
+      "grad_norm": 4.375,
+      "learning_rate": 4.6754051142374275e-07,
+      "loss": 0.5605,
+      "mean_token_accuracy": 0.8824782967567444,
+      "num_tokens": 381319310.0,
+      "step": 3566
+    },
+    {
+      "epoch": 8.125427594070695,
+      "grad_norm": 3.375,
+      "learning_rate": 4.664443818462658e-07,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8837754726409912,
+      "num_tokens": 381426819.0,
+      "step": 3567
+    },
+    {
+      "epoch": 8.127708095781072,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.653494064756983e-07,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.882844015955925,
+      "num_tokens": 381533803.0,
+      "step": 3568
+    },
+    {
+      "epoch": 8.129988597491447,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.6425558593352796e-07,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8839477002620697,
+      "num_tokens": 381640832.0,
+      "step": 3569
+    },
+    {
+      "epoch": 8.132269099201825,
+      "grad_norm": 4.125,
+      "learning_rate": 4.631629208405847e-07,
+      "loss": 0.5459,
+      "mean_token_accuracy": 0.8868978321552277,
+      "num_tokens": 381748214.0,
+      "step": 3570
+    },
+    {
+      "epoch": 8.134549600912202,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.620714118170452e-07,
+      "loss": 0.5508,
+      "mean_token_accuracy": 0.8868376761674881,
+      "num_tokens": 381855821.0,
+      "step": 3571
+    },
+    {
+      "epoch": 8.136830102622577,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.609810594824282e-07,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.883231908082962,
+      "num_tokens": 381962902.0,
+      "step": 3572
+    },
+    {
+      "epoch": 8.139110604332954,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.598918644555975e-07,
+      "loss": 0.5536,
+      "mean_token_accuracy": 0.8844647407531738,
+      "num_tokens": 382070254.0,
+      "step": 3573
+    },
+    {
+      "epoch": 8.14139110604333,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.58803827354759e-07,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8791768401861191,
+      "num_tokens": 382177004.0,
+      "step": 3574
+    },
+    {
+      "epoch": 8.143671607753706,
+      "grad_norm": 3.625,
+      "learning_rate": 4.5771694879746087e-07,
+      "loss": 0.5942,
+      "mean_token_accuracy": 0.874402716755867,
+      "num_tokens": 382283679.0,
+      "step": 3575
+    },
+    {
+      "epoch": 8.145952109464082,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.566312294005948e-07,
+      "loss": 0.5765,
+      "mean_token_accuracy": 0.8818236142396927,
+      "num_tokens": 382390417.0,
+      "step": 3576
+    },
+    {
+      "epoch": 8.148232611174459,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.5554666978039455e-07,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8824535757303238,
+      "num_tokens": 382497498.0,
+      "step": 3577
+    },
+    {
+      "epoch": 8.150513112884834,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.544632705524343e-07,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8810164928436279,
+      "num_tokens": 382604536.0,
+      "step": 3578
+    },
+    {
+      "epoch": 8.152793614595211,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.5338103233163175e-07,
+      "loss": 0.5817,
+      "mean_token_accuracy": 0.8840733021497726,
+      "num_tokens": 382711132.0,
+      "step": 3579
+    },
+    {
+      "epoch": 8.155074116305586,
+      "grad_norm": 2.375,
+      "learning_rate": 4.522999557322433e-07,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8804437518119812,
+      "num_tokens": 382817671.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.157354618015964,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.512200413678672e-07,
+      "loss": 0.562,
+      "mean_token_accuracy": 0.8838673382997513,
+      "num_tokens": 382924893.0,
+      "step": 3581
+    },
+    {
+      "epoch": 8.15963511972634,
+      "grad_norm": 3.0,
+      "learning_rate": 4.501412898514426e-07,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8765487670898438,
+      "num_tokens": 383031251.0,
+      "step": 3582
+    },
+    {
+      "epoch": 8.161915621436716,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.490637017952479e-07,
+      "loss": 0.5583,
+      "mean_token_accuracy": 0.8857912421226501,
+      "num_tokens": 383138617.0,
+      "step": 3583
+    },
+    {
+      "epoch": 8.164196123147093,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.4798727781090096e-07,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.883015900850296,
+      "num_tokens": 383245038.0,
+      "step": 3584
+    },
+    {
+      "epoch": 8.166476624857468,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.4691201850936034e-07,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8790012896060944,
+      "num_tokens": 383351713.0,
+      "step": 3585
+    },
+    {
+      "epoch": 8.168757126567845,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.458379245009209e-07,
+      "loss": 0.5598,
+      "mean_token_accuracy": 0.8837389200925827,
+      "num_tokens": 383458654.0,
+      "step": 3586
+    },
+    {
+      "epoch": 8.17103762827822,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.447649963952183e-07,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.8778487145900726,
+      "num_tokens": 383565763.0,
+      "step": 3587
+    },
+    {
+      "epoch": 8.173318129988598,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.43693234801226e-07,
+      "loss": 0.5907,
+      "mean_token_accuracy": 0.8766829818487167,
+      "num_tokens": 383672820.0,
+      "step": 3588
+    },
+    {
+      "epoch": 8.175598631698973,
+      "grad_norm": 5.4375,
+      "learning_rate": 4.4262264032725517e-07,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8824921548366547,
+      "num_tokens": 383780010.0,
+      "step": 3589
+    },
+    {
+      "epoch": 8.17787913340935,
+      "grad_norm": 3.125,
+      "learning_rate": 4.41553213580955e-07,
+      "loss": 0.5849,
+      "mean_token_accuracy": 0.8807243406772614,
+      "num_tokens": 383886905.0,
+      "step": 3590
+    },
+    {
+      "epoch": 8.180159635119727,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.404849551693102e-07,
+      "loss": 0.5775,
+      "mean_token_accuracy": 0.8819932341575623,
+      "num_tokens": 383994254.0,
+      "step": 3591
+    },
+    {
+      "epoch": 8.182440136830103,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.394178656986448e-07,
+      "loss": 0.5586,
+      "mean_token_accuracy": 0.881662055850029,
+      "num_tokens": 384101925.0,
+      "step": 3592
+    },
+    {
+      "epoch": 8.18472063854048,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.383519457746174e-07,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8843112587928772,
+      "num_tokens": 384209051.0,
+      "step": 3593
+    },
+    {
+      "epoch": 8.187001140250855,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.3728719600222374e-07,
+      "loss": 0.5566,
+      "mean_token_accuracy": 0.8843548893928528,
+      "num_tokens": 384316445.0,
+      "step": 3594
+    },
+    {
+      "epoch": 8.189281641961232,
+      "grad_norm": 2.875,
+      "learning_rate": 4.3622361698579586e-07,
+      "loss": 0.5965,
+      "mean_token_accuracy": 0.8787325769662857,
+      "num_tokens": 384422950.0,
+      "step": 3595
+    },
+    {
+      "epoch": 8.191562143671607,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.351612093290006e-07,
+      "loss": 0.5696,
+      "mean_token_accuracy": 0.8826261758804321,
+      "num_tokens": 384530108.0,
+      "step": 3596
+    },
+    {
+      "epoch": 8.193842645381984,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.340999736348389e-07,
+      "loss": 0.5549,
+      "mean_token_accuracy": 0.8850667327642441,
+      "num_tokens": 384637297.0,
+      "step": 3597
+    },
+    {
+      "epoch": 8.19612314709236,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.3303991050564877e-07,
+      "loss": 0.5562,
+      "mean_token_accuracy": 0.8847819566726685,
+      "num_tokens": 384744448.0,
+      "step": 3598
+    },
+    {
+      "epoch": 8.198403648802737,
+      "grad_norm": 2.875,
+      "learning_rate": 4.3198102054310157e-07,
+      "loss": 0.5707,
+      "mean_token_accuracy": 0.8816685527563095,
+      "num_tokens": 384851534.0,
+      "step": 3599
+    },
+    {
+      "epoch": 8.200684150513112,
+      "grad_norm": 4.25,
+      "learning_rate": 4.30923304348202e-07,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8833976686000824,
+      "num_tokens": 384958686.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.20296465222349,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.2986676252129047e-07,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8791931867599487,
+      "num_tokens": 385065535.0,
+      "step": 3601
+    },
+    {
+      "epoch": 8.205245153933866,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.288113956620382e-07,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8821399062871933,
+      "num_tokens": 385172528.0,
+      "step": 3602
+    },
+    {
+      "epoch": 8.207525655644242,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.2775720436945225e-07,
+      "loss": 0.5577,
+      "mean_token_accuracy": 0.8869880139827728,
+      "num_tokens": 385279822.0,
+      "step": 3603
+    },
+    {
+      "epoch": 8.209806157354619,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.267041892418705e-07,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8820626139640808,
+      "num_tokens": 385387057.0,
+      "step": 3604
+    },
+    {
+      "epoch": 8.212086659064994,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.256523508769647e-07,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8827876895666122,
+      "num_tokens": 385494227.0,
+      "step": 3605
+    },
+    {
+      "epoch": 8.214367160775371,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.246016898717381e-07,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8795175403356552,
+      "num_tokens": 385600675.0,
+      "step": 3606
+    },
+    {
+      "epoch": 8.216647662485746,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.235522068225248e-07,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8818391412496567,
+      "num_tokens": 385707591.0,
+      "step": 3607
+    },
+    {
+      "epoch": 8.218928164196123,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.225039023249916e-07,
+      "loss": 0.5563,
+      "mean_token_accuracy": 0.8863593190908432,
+      "num_tokens": 385814554.0,
+      "step": 3608
+    },
+    {
+      "epoch": 8.221208665906499,
+      "grad_norm": 3.0,
+      "learning_rate": 4.2145677697413566e-07,
+      "loss": 0.5972,
+      "mean_token_accuracy": 0.8759454041719437,
+      "num_tokens": 385921328.0,
+      "step": 3609
+    },
+    {
+      "epoch": 8.223489167616876,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.204108313642852e-07,
+      "loss": 0.5589,
+      "mean_token_accuracy": 0.8848637193441391,
+      "num_tokens": 386028022.0,
+      "step": 3610
+    },
+    {
+      "epoch": 8.225769669327253,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.1936606608909887e-07,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.883062332868576,
+      "num_tokens": 386135191.0,
+      "step": 3611
+    },
+    {
+      "epoch": 8.228050171037628,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.1832248174156597e-07,
+      "loss": 0.5584,
+      "mean_token_accuracy": 0.885178342461586,
+      "num_tokens": 386242119.0,
+      "step": 3612
+    },
+    {
+      "epoch": 8.230330672748005,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.1728007891400356e-07,
+      "loss": 0.5714,
+      "mean_token_accuracy": 0.8812027722597122,
+      "num_tokens": 386349161.0,
+      "step": 3613
+    },
+    {
+      "epoch": 8.23261117445838,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.1623885819805977e-07,
+      "loss": 0.5699,
+      "mean_token_accuracy": 0.8800331354141235,
+      "num_tokens": 386456594.0,
+      "step": 3614
+    },
+    {
+      "epoch": 8.234891676168758,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.151988201847112e-07,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.884452149271965,
+      "num_tokens": 386563773.0,
+      "step": 3615
+    },
+    {
+      "epoch": 8.237172177879133,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.141599654642642e-07,
+      "loss": 0.5577,
+      "mean_token_accuracy": 0.8853688091039658,
+      "num_tokens": 386670901.0,
+      "step": 3616
+    },
+    {
+      "epoch": 8.23945267958951,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.1312229462635243e-07,
+      "loss": 0.5623,
+      "mean_token_accuracy": 0.8857151865959167,
+      "num_tokens": 386777914.0,
+      "step": 3617
+    },
+    {
+      "epoch": 8.241733181299885,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.1208580825993686e-07,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8840577751398087,
+      "num_tokens": 386884846.0,
+      "step": 3618
+    },
+    {
+      "epoch": 8.244013683010262,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.1105050695330774e-07,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8825150728225708,
+      "num_tokens": 386992013.0,
+      "step": 3619
+    },
+    {
+      "epoch": 8.246294184720638,
+      "grad_norm": 3.953125,
+      "learning_rate": 4.100163912940827e-07,
+      "loss": 0.5946,
+      "mean_token_accuracy": 0.8755682855844498,
+      "num_tokens": 387098665.0,
+      "step": 3620
+    },
+    {
+      "epoch": 8.248574686431015,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.0898346186920484e-07,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8821796476840973,
+      "num_tokens": 387205278.0,
+      "step": 3621
+    },
+    {
+      "epoch": 8.250855188141392,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.0795171926494543e-07,
+      "loss": 0.5953,
+      "mean_token_accuracy": 0.8782723695039749,
+      "num_tokens": 387312537.0,
+      "step": 3622
+    },
+    {
+      "epoch": 8.253135689851767,
+      "grad_norm": 3.25,
+      "learning_rate": 4.0692116406690214e-07,
+      "loss": 0.5792,
+      "mean_token_accuracy": 0.8817024827003479,
+      "num_tokens": 387419037.0,
+      "step": 3623
+    },
+    {
+      "epoch": 8.255416191562144,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.058917968599968e-07,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8763107359409332,
+      "num_tokens": 387525971.0,
+      "step": 3624
+    },
+    {
+      "epoch": 8.25769669327252,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.048636182284796e-07,
+      "loss": 0.5607,
+      "mean_token_accuracy": 0.8834527432918549,
+      "num_tokens": 387633084.0,
+      "step": 3625
+    },
+    {
+      "epoch": 8.259977194982897,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.038366287559245e-07,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8804467916488647,
+      "num_tokens": 387740264.0,
+      "step": 3626
+    },
+    {
+      "epoch": 8.262257696693272,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.0281082902523055e-07,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8822118043899536,
+      "num_tokens": 387847552.0,
+      "step": 3627
+    },
+    {
+      "epoch": 8.264538198403649,
+      "grad_norm": 2.875,
+      "learning_rate": 4.0178621961862315e-07,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8843962848186493,
+      "num_tokens": 387954393.0,
+      "step": 3628
+    },
+    {
+      "epoch": 8.266818700114024,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.0076280111764927e-07,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8815756440162659,
+      "num_tokens": 388061243.0,
+      "step": 3629
+    },
+    {
+      "epoch": 8.269099201824401,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.997405741031821e-07,
+      "loss": 0.5575,
+      "mean_token_accuracy": 0.885826587677002,
+      "num_tokens": 388168371.0,
+      "step": 3630
+    },
+    {
+      "epoch": 8.271379703534778,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.98719539155418e-07,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8817395716905594,
+      "num_tokens": 388276038.0,
+      "step": 3631
+    },
+    {
+      "epoch": 8.273660205245154,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.9769969685387684e-07,
+      "loss": 0.5584,
+      "mean_token_accuracy": 0.8838264048099518,
+      "num_tokens": 388382766.0,
+      "step": 3632
+    },
+    {
+      "epoch": 8.27594070695553,
+      "grad_norm": 3.625,
+      "learning_rate": 3.966810477774016e-07,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.8818796724081039,
+      "num_tokens": 388489939.0,
+      "step": 3633
+    },
+    {
+      "epoch": 8.278221208665906,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.9566359250415686e-07,
+      "loss": 0.5851,
+      "mean_token_accuracy": 0.8771698474884033,
+      "num_tokens": 388596579.0,
+      "step": 3634
+    },
+    {
+      "epoch": 8.280501710376283,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.9464733161163144e-07,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.881603792309761,
+      "num_tokens": 388703184.0,
+      "step": 3635
+    },
+    {
+      "epoch": 8.282782212086659,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.9363226567663503e-07,
+      "loss": 0.5835,
+      "mean_token_accuracy": 0.8799589574337006,
+      "num_tokens": 388809921.0,
+      "step": 3636
+    },
+    {
+      "epoch": 8.285062713797036,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.926183952752999e-07,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8831854462623596,
+      "num_tokens": 388916867.0,
+      "step": 3637
+    },
+    {
+      "epoch": 8.287343215507411,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.9160572098307923e-07,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8826831877231598,
+      "num_tokens": 389024041.0,
+      "step": 3638
+    },
+    {
+      "epoch": 8.289623717217788,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.90594243374747e-07,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8816553503274918,
+      "num_tokens": 389131535.0,
+      "step": 3639
+    },
+    {
+      "epoch": 8.291904218928163,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.895839630243983e-07,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8834093660116196,
+      "num_tokens": 389238175.0,
+      "step": 3640
+    },
+    {
+      "epoch": 8.29418472063854,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.8857488050544903e-07,
+      "loss": 0.5675,
+      "mean_token_accuracy": 0.8817841410636902,
+      "num_tokens": 389344785.0,
+      "step": 3641
+    },
+    {
+      "epoch": 8.296465222348917,
+      "grad_norm": 2.625,
+      "learning_rate": 3.875669963906356e-07,
+      "loss": 0.5535,
+      "mean_token_accuracy": 0.8869169652462006,
+      "num_tokens": 389452290.0,
+      "step": 3642
+    },
+    {
+      "epoch": 8.298745724059293,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.865603112520125e-07,
+      "loss": 0.5573,
+      "mean_token_accuracy": 0.8858313113451004,
+      "num_tokens": 389559422.0,
+      "step": 3643
+    },
+    {
+      "epoch": 8.30102622576967,
+      "grad_norm": 3.125,
+      "learning_rate": 3.855548256609556e-07,
+      "loss": 0.5437,
+      "mean_token_accuracy": 0.8861667513847351,
+      "num_tokens": 389667410.0,
+      "step": 3644
+    },
+    {
+      "epoch": 8.303306727480045,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.8455054018815803e-07,
+      "loss": 0.5598,
+      "mean_token_accuracy": 0.8834993839263916,
+      "num_tokens": 389774294.0,
+      "step": 3645
+    },
+    {
+      "epoch": 8.305587229190422,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.8354745540363364e-07,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8835310786962509,
+      "num_tokens": 389881962.0,
+      "step": 3646
+    },
+    {
+      "epoch": 8.307867730900798,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.8254557187671374e-07,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.881848931312561,
+      "num_tokens": 389989039.0,
+      "step": 3647
+    },
+    {
+      "epoch": 8.310148232611175,
+      "grad_norm": 3.0,
+      "learning_rate": 3.815448901760485e-07,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8844071328639984,
+      "num_tokens": 390096310.0,
+      "step": 3648
+    },
+    {
+      "epoch": 8.31242873432155,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.805454108696055e-07,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8806750923395157,
+      "num_tokens": 390202884.0,
+      "step": 3649
+    },
+    {
+      "epoch": 8.314709236031927,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.7954713452466927e-07,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8855821341276169,
+      "num_tokens": 390310246.0,
+      "step": 3650
+    },
+    {
+      "epoch": 8.316989737742304,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.785500617078425e-07,
+      "loss": 0.5854,
+      "mean_token_accuracy": 0.8808927685022354,
+      "num_tokens": 390416919.0,
+      "step": 3651
+    },
+    {
+      "epoch": 8.31927023945268,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.775541929850443e-07,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8799949437379837,
+      "num_tokens": 390524810.0,
+      "step": 3652
+    },
+    {
+      "epoch": 8.321550741163056,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.76559528921511e-07,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8805143088102341,
+      "num_tokens": 390631927.0,
+      "step": 3653
+    },
+    {
+      "epoch": 8.323831242873432,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.7556607008179454e-07,
+      "loss": 0.557,
+      "mean_token_accuracy": 0.885184645652771,
+      "num_tokens": 390738890.0,
+      "step": 3654
+    },
+    {
+      "epoch": 8.326111744583809,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.745738170297633e-07,
+      "loss": 0.5676,
+      "mean_token_accuracy": 0.8800319731235504,
+      "num_tokens": 390845227.0,
+      "step": 3655
+    },
+    {
+      "epoch": 8.328392246294184,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.7358277032860016e-07,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8837781846523285,
+      "num_tokens": 390951990.0,
+      "step": 3656
+    },
+    {
+      "epoch": 8.330672748004561,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.7259293054080435e-07,
+      "loss": 0.594,
+      "mean_token_accuracy": 0.8773994743824005,
+      "num_tokens": 391059266.0,
+      "step": 3657
+    },
+    {
+      "epoch": 8.332953249714937,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.7160429822819003e-07,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8825381994247437,
+      "num_tokens": 391166339.0,
+      "step": 3658
+    },
+    {
+      "epoch": 8.335233751425314,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.706168739518859e-07,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.881363645195961,
+      "num_tokens": 391273827.0,
+      "step": 3659
+    },
+    {
+      "epoch": 8.33751425313569,
+      "grad_norm": 4.375,
+      "learning_rate": 3.6963065827233524e-07,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8808765411376953,
+      "num_tokens": 391381343.0,
+      "step": 3660
+    },
+    {
+      "epoch": 8.339794754846066,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.6864565174929393e-07,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8825783133506775,
+      "num_tokens": 391488599.0,
+      "step": 3661
+    },
+    {
+      "epoch": 8.342075256556443,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.676618549418334e-07,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8835724592208862,
+      "num_tokens": 391596245.0,
+      "step": 3662
+    },
+    {
+      "epoch": 8.344355758266818,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.666792684083381e-07,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8827132731676102,
+      "num_tokens": 391702960.0,
+      "step": 3663
+    },
+    {
+      "epoch": 8.346636259977195,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.656978927065041e-07,
+      "loss": 0.5646,
+      "mean_token_accuracy": 0.8864104300737381,
+      "num_tokens": 391811197.0,
+      "step": 3664
+    },
+    {
+      "epoch": 8.34891676168757,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.64717728393342e-07,
+      "loss": 0.602,
+      "mean_token_accuracy": 0.8749131262302399,
+      "num_tokens": 391918416.0,
+      "step": 3665
+    },
+    {
+      "epoch": 8.351197263397948,
+      "grad_norm": 2.875,
+      "learning_rate": 3.6373877602517457e-07,
+      "loss": 0.5567,
+      "mean_token_accuracy": 0.884322926402092,
+      "num_tokens": 392025371.0,
+      "step": 3666
+    },
+    {
+      "epoch": 8.353477765108323,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.627610361576353e-07,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.8826847523450851,
+      "num_tokens": 392132210.0,
+      "step": 3667
+    },
+    {
+      "epoch": 8.3557582668187,
+      "grad_norm": 3.0,
+      "learning_rate": 3.6178450934567065e-07,
+      "loss": 0.5909,
+      "mean_token_accuracy": 0.8771587163209915,
+      "num_tokens": 392239261.0,
+      "step": 3668
+    },
+    {
+      "epoch": 8.358038768529076,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.6080919614353895e-07,
+      "loss": 0.5765,
+      "mean_token_accuracy": 0.8804507404565811,
+      "num_tokens": 392345489.0,
+      "step": 3669
+    },
+    {
+      "epoch": 8.360319270239453,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.598350971048087e-07,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8826466053724289,
+      "num_tokens": 392452753.0,
+      "step": 3670
+    },
+    {
+      "epoch": 8.36259977194983,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.5886221278236045e-07,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8848947435617447,
+      "num_tokens": 392559803.0,
+      "step": 3671
+    },
+    {
+      "epoch": 8.364880273660205,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.578905437283833e-07,
+      "loss": 0.5829,
+      "mean_token_accuracy": 0.8808190226554871,
+      "num_tokens": 392666034.0,
+      "step": 3672
+    },
+    {
+      "epoch": 8.367160775370582,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.569200904943784e-07,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.8789727836847305,
+      "num_tokens": 392772892.0,
+      "step": 3673
+    },
+    {
+      "epoch": 8.369441277080957,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.559508536311568e-07,
+      "loss": 0.5528,
+      "mean_token_accuracy": 0.8866614103317261,
+      "num_tokens": 392880120.0,
+      "step": 3674
+    },
+    {
+      "epoch": 8.371721778791335,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.549828336888378e-07,
+      "loss": 0.5579,
+      "mean_token_accuracy": 0.8827106803655624,
+      "num_tokens": 392987873.0,
+      "step": 3675
+    },
+    {
+      "epoch": 8.37400228050171,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.5401603121685197e-07,
+      "loss": 0.5466,
+      "mean_token_accuracy": 0.8857832849025726,
+      "num_tokens": 393095077.0,
+      "step": 3676
+    },
+    {
+      "epoch": 8.376282782212087,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.5305044676393645e-07,
+      "loss": 0.5893,
+      "mean_token_accuracy": 0.8792509883642197,
+      "num_tokens": 393202410.0,
+      "step": 3677
+    },
+    {
+      "epoch": 8.378563283922462,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.5208608087813874e-07,
+      "loss": 0.5906,
+      "mean_token_accuracy": 0.8767274022102356,
+      "num_tokens": 393309342.0,
+      "step": 3678
+    },
+    {
+      "epoch": 8.38084378563284,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.5112293410681455e-07,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.881255492568016,
+      "num_tokens": 393416921.0,
+      "step": 3679
+    },
+    {
+      "epoch": 8.383124287343216,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.501610069966271e-07,
+      "loss": 0.5514,
+      "mean_token_accuracy": 0.8843955397605896,
+      "num_tokens": 393524843.0,
+      "step": 3680
+    },
+    {
+      "epoch": 8.385404789053592,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.492003000935487e-07,
+      "loss": 0.5547,
+      "mean_token_accuracy": 0.8885585069656372,
+      "num_tokens": 393632101.0,
+      "step": 3681
+    },
+    {
+      "epoch": 8.387685290763969,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.482408139428564e-07,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8814975172281265,
+      "num_tokens": 393739142.0,
+      "step": 3682
+    },
+    {
+      "epoch": 8.389965792474344,
+      "grad_norm": 3.0,
+      "learning_rate": 3.4728254908913683e-07,
+      "loss": 0.5596,
+      "mean_token_accuracy": 0.8834152221679688,
+      "num_tokens": 393846302.0,
+      "step": 3683
+    },
+    {
+      "epoch": 8.392246294184721,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.463255060762827e-07,
+      "loss": 0.5646,
+      "mean_token_accuracy": 0.8830453753471375,
+      "num_tokens": 393952665.0,
+      "step": 3684
+    },
+    {
+      "epoch": 8.394526795895096,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.4536968544749333e-07,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8798201084136963,
+      "num_tokens": 394059602.0,
+      "step": 3685
+    },
+    {
+      "epoch": 8.396807297605474,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.4441508774527345e-07,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.8854030817747116,
+      "num_tokens": 394166384.0,
+      "step": 3686
+    },
+    {
+      "epoch": 8.399087799315849,
+      "grad_norm": 3.875,
+      "learning_rate": 3.434617135114349e-07,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8825076222419739,
+      "num_tokens": 394273389.0,
+      "step": 3687
+    },
+    {
+      "epoch": 8.401368301026226,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.425095632870937e-07,
+      "loss": 0.594,
+      "mean_token_accuracy": 0.8796204030513763,
+      "num_tokens": 394380226.0,
+      "step": 3688
+    },
+    {
+      "epoch": 8.403648802736601,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.4155863761267256e-07,
+      "loss": 0.5841,
+      "mean_token_accuracy": 0.8805309385061264,
+      "num_tokens": 394486821.0,
+      "step": 3689
+    },
+    {
+      "epoch": 8.405929304446978,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.406089370278981e-07,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8786486685276031,
+      "num_tokens": 394593768.0,
+      "step": 3690
+    },
+    {
+      "epoch": 8.408209806157355,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.396604620718025e-07,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.8826608210802078,
+      "num_tokens": 394700911.0,
+      "step": 3691
+    },
+    {
+      "epoch": 8.41049030786773,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.387132132827223e-07,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8810716569423676,
+      "num_tokens": 394807801.0,
+      "step": 3692
+    },
+    {
+      "epoch": 8.412770809578108,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.377671911982963e-07,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8786879032850266,
+      "num_tokens": 394914319.0,
+      "step": 3693
+    },
+    {
+      "epoch": 8.415051311288483,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.3682239635546927e-07,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8792416155338287,
+      "num_tokens": 395021338.0,
+      "step": 3694
+    },
+    {
+      "epoch": 8.41733181299886,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.35878829290488e-07,
+      "loss": 0.5768,
+      "mean_token_accuracy": 0.8806908279657364,
+      "num_tokens": 395128126.0,
+      "step": 3695
+    },
+    {
+      "epoch": 8.419612314709235,
+      "grad_norm": 4.625,
+      "learning_rate": 3.3493649053890325e-07,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8870299756526947,
+      "num_tokens": 395234907.0,
+      "step": 3696
+    },
+    {
+      "epoch": 8.421892816419613,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.339953806355692e-07,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8846122622489929,
+      "num_tokens": 395342739.0,
+      "step": 3697
+    },
+    {
+      "epoch": 8.424173318129988,
+      "grad_norm": 5.5625,
+      "learning_rate": 3.330555001146399e-07,
+      "loss": 0.5829,
+      "mean_token_accuracy": 0.8779343664646149,
+      "num_tokens": 395449718.0,
+      "step": 3698
+    },
+    {
+      "epoch": 8.426453819840365,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.3211684950957416e-07,
+      "loss": 0.5808,
+      "mean_token_accuracy": 0.8803275525569916,
+      "num_tokens": 395556795.0,
+      "step": 3699
+    },
+    {
+      "epoch": 8.428734321550742,
+      "grad_norm": 3.0,
+      "learning_rate": 3.311794293531323e-07,
+      "loss": 0.5608,
+      "mean_token_accuracy": 0.8883070945739746,
+      "num_tokens": 395664009.0,
+      "step": 3700
+    },
+    {
+      "epoch": 8.431014823261117,
+      "grad_norm": 4.125,
+      "learning_rate": 3.3024324017737555e-07,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8797996789216995,
+      "num_tokens": 395771593.0,
+      "step": 3701
+    },
+    {
+      "epoch": 8.433295324971494,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.2930828251366703e-07,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8806504160165787,
+      "num_tokens": 395878663.0,
+      "step": 3702
+    },
+    {
+      "epoch": 8.43557582668187,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.283745568926708e-07,
+      "loss": 0.5562,
+      "mean_token_accuracy": 0.8880557268857956,
+      "num_tokens": 395986031.0,
+      "step": 3703
+    },
+    {
+      "epoch": 8.437856328392247,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.274420638443507e-07,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8808362036943436,
+      "num_tokens": 396093478.0,
+      "step": 3704
+    },
+    {
+      "epoch": 8.440136830102622,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.2651080389797253e-07,
+      "loss": 0.5438,
+      "mean_token_accuracy": 0.8885146230459213,
+      "num_tokens": 396200577.0,
+      "step": 3705
+    },
+    {
+      "epoch": 8.442417331813,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.255807775821015e-07,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8813702762126923,
+      "num_tokens": 396307631.0,
+      "step": 3706
+    },
+    {
+      "epoch": 8.444697833523374,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.246519854246022e-07,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.882218137383461,
+      "num_tokens": 396415007.0,
+      "step": 3707
+    },
+    {
+      "epoch": 8.446978335233752,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.2372442795263885e-07,
+      "loss": 0.5712,
+      "mean_token_accuracy": 0.8808966130018234,
+      "num_tokens": 396521935.0,
+      "step": 3708
+    },
+    {
+      "epoch": 8.449258836944129,
+      "grad_norm": 3.75,
+      "learning_rate": 3.227981056926763e-07,
+      "loss": 0.597,
+      "mean_token_accuracy": 0.8786474019289017,
+      "num_tokens": 396629351.0,
+      "step": 3709
+    },
+    {
+      "epoch": 8.451539338654504,
+      "grad_norm": 3.25,
+      "learning_rate": 3.218730191704758e-07,
+      "loss": 0.5879,
+      "mean_token_accuracy": 0.8776755332946777,
+      "num_tokens": 396736310.0,
+      "step": 3710
+    },
+    {
+      "epoch": 8.453819840364881,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.209491689110994e-07,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.8844281584024429,
+      "num_tokens": 396843391.0,
+      "step": 3711
+    },
+    {
+      "epoch": 8.456100342075256,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.2002655543890646e-07,
+      "loss": 0.5692,
+      "mean_token_accuracy": 0.8826111257076263,
+      "num_tokens": 396950647.0,
+      "step": 3712
+    },
+    {
+      "epoch": 8.458380843785633,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.1910517927755516e-07,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.879383772611618,
+      "num_tokens": 397058121.0,
+      "step": 3713
+    },
+    {
+      "epoch": 8.460661345496009,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.181850409499995e-07,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8790426999330521,
+      "num_tokens": 397165268.0,
+      "step": 3714
+    },
+    {
+      "epoch": 8.462941847206386,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.1726614097849326e-07,
+      "loss": 0.5596,
+      "mean_token_accuracy": 0.8887834399938583,
+      "num_tokens": 397272308.0,
+      "step": 3715
+    },
+    {
+      "epoch": 8.465222348916761,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.163484798845862e-07,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8812167048454285,
+      "num_tokens": 397379017.0,
+      "step": 3716
+    },
+    {
+      "epoch": 8.467502850627138,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.1543205818912484e-07,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8864733725786209,
+      "num_tokens": 397485914.0,
+      "step": 3717
+    },
+    {
+      "epoch": 8.469783352337513,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.145168764122525e-07,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8842067122459412,
+      "num_tokens": 397592552.0,
+      "step": 3718
+    },
+    {
+      "epoch": 8.47206385404789,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.1360293507340934e-07,
+      "loss": 0.5665,
+      "mean_token_accuracy": 0.8827557861804962,
+      "num_tokens": 397699338.0,
+      "step": 3719
+    },
+    {
+      "epoch": 8.474344355758268,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.1269023469132937e-07,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.8828287869691849,
+      "num_tokens": 397806861.0,
+      "step": 3720
+    },
+    {
+      "epoch": 8.476624857468643,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.117787757840449e-07,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.8807165771722794,
+      "num_tokens": 397913908.0,
+      "step": 3721
+    },
+    {
+      "epoch": 8.47890535917902,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.10868558868882e-07,
+      "loss": 0.5543,
+      "mean_token_accuracy": 0.8827250152826309,
+      "num_tokens": 398021549.0,
+      "step": 3722
+    },
+    {
+      "epoch": 8.481185860889395,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.0995958446246197e-07,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.881324291229248,
+      "num_tokens": 398128775.0,
+      "step": 3723
+    },
+    {
+      "epoch": 8.483466362599772,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.090518530807021e-07,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.8850361853837967,
+      "num_tokens": 398235464.0,
+      "step": 3724
+    },
+    {
+      "epoch": 8.485746864310148,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.0814536523881224e-07,
+      "loss": 0.5474,
+      "mean_token_accuracy": 0.8856307417154312,
+      "num_tokens": 398342478.0,
+      "step": 3725
+    },
+    {
+      "epoch": 8.488027366020525,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.072401214512974e-07,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8800984472036362,
+      "num_tokens": 398449103.0,
+      "step": 3726
+    },
+    {
+      "epoch": 8.4903078677309,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.063361222319569e-07,
+      "loss": 0.5692,
+      "mean_token_accuracy": 0.8843390345573425,
+      "num_tokens": 398556171.0,
+      "step": 3727
+    },
+    {
+      "epoch": 8.492588369441277,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.054333680938837e-07,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.8837459683418274,
+      "num_tokens": 398663128.0,
+      "step": 3728
+    },
+    {
+      "epoch": 8.494868871151652,
+      "grad_norm": 4.25,
+      "learning_rate": 3.045318595494623e-07,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8831415474414825,
+      "num_tokens": 398770220.0,
+      "step": 3729
+    },
+    {
+      "epoch": 8.49714937286203,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.036315971103723e-07,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8805525749921799,
+      "num_tokens": 398876750.0,
+      "step": 3730
+    },
+    {
+      "epoch": 8.499429874572407,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.0273258128758585e-07,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.8824233710765839,
+      "num_tokens": 398983846.0,
+      "step": 3731
+    },
+    {
+      "epoch": 8.501710376282782,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.018348125913659e-07,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8839138001203537,
+      "num_tokens": 399090705.0,
+      "step": 3732
+    },
+    {
+      "epoch": 8.503990877993159,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.009382915312689e-07,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8818164467811584,
+      "num_tokens": 399197937.0,
+      "step": 3733
+    },
+    {
+      "epoch": 8.506271379703534,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.000430186161432e-07,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8860601335763931,
+      "num_tokens": 399304810.0,
+      "step": 3734
+    },
+    {
+      "epoch": 8.508551881413911,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.991489943541287e-07,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8790812194347382,
+      "num_tokens": 399411951.0,
+      "step": 3735
+    },
+    {
+      "epoch": 8.510832383124287,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.982562192526556e-07,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.881022498011589,
+      "num_tokens": 399519175.0,
+      "step": 3736
+    },
+    {
+      "epoch": 8.513112884834664,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.97364693818446e-07,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8782191127538681,
+      "num_tokens": 399626295.0,
+      "step": 3737
+    },
+    {
+      "epoch": 8.515393386545039,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.9647441855751274e-07,
+      "loss": 0.5584,
+      "mean_token_accuracy": 0.8843040764331818,
+      "num_tokens": 399733595.0,
+      "step": 3738
+    },
+    {
+      "epoch": 8.517673888255416,
+      "grad_norm": 4.59375,
+      "learning_rate": 2.9558539397515905e-07,
+      "loss": 0.5624,
+      "mean_token_accuracy": 0.884150817990303,
+      "num_tokens": 399840730.0,
+      "step": 3739
+    },
+    {
+      "epoch": 8.519954389965793,
+      "grad_norm": 3.25,
+      "learning_rate": 2.94697620575978e-07,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8824001997709274,
+      "num_tokens": 399947355.0,
+      "step": 3740
+    },
+    {
+      "epoch": 8.519954389965793,
+      "eval_loss": 0.586450457572937,
+      "eval_mean_token_accuracy": 0.8799830790708274,
+      "eval_num_tokens": 399947355.0,
+      "eval_runtime": 58.61,
+      "eval_samples_per_second": 143.064,
+      "eval_steps_per_second": 4.487,
+      "step": 3740
+    },
+    {
+      "epoch": 8.522234891676169,
+      "grad_norm": 3.0,
+      "learning_rate": 2.938110988638521e-07,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8857548832893372,
+      "num_tokens": 400053792.0,
+      "step": 3741
+    },
+    {
+      "epoch": 8.524515393386546,
+      "grad_norm": 2.625,
+      "learning_rate": 2.9292582934195427e-07,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8823639154434204,
+      "num_tokens": 400160716.0,
+      "step": 3742
+    },
+    {
+      "epoch": 8.526795895096921,
+      "grad_norm": 4.75,
+      "learning_rate": 2.9204181251274665e-07,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8819014728069305,
+      "num_tokens": 400267286.0,
+      "step": 3743
+    },
+    {
+      "epoch": 8.529076396807298,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.9115904887798005e-07,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8814072459936142,
+      "num_tokens": 400374052.0,
+      "step": 3744
+    },
+    {
+      "epoch": 8.531356898517673,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.9027753893869387e-07,
+      "loss": 0.5511,
+      "mean_token_accuracy": 0.8873879760503769,
+      "num_tokens": 400481323.0,
+      "step": 3745
+    },
+    {
+      "epoch": 8.53363740022805,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.893972831952166e-07,
+      "loss": 0.5714,
+      "mean_token_accuracy": 0.8847419172525406,
+      "num_tokens": 400588604.0,
+      "step": 3746
+    },
+    {
+      "epoch": 8.535917901938426,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.8851828214716383e-07,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8783596158027649,
+      "num_tokens": 400695471.0,
+      "step": 3747
+    },
+    {
+      "epoch": 8.538198403648803,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.876405362934395e-07,
+      "loss": 0.5835,
+      "mean_token_accuracy": 0.8776167929172516,
+      "num_tokens": 400802607.0,
+      "step": 3748
+    },
+    {
+      "epoch": 8.54047890535918,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.8676404613223573e-07,
+      "loss": 0.5807,
+      "mean_token_accuracy": 0.8812108039855957,
+      "num_tokens": 400909340.0,
+      "step": 3749
+    },
+    {
+      "epoch": 8.542759407069555,
+      "grad_norm": 4.0,
+      "learning_rate": 2.858888121610315e-07,
+      "loss": 0.5536,
+      "mean_token_accuracy": 0.887622594833374,
+      "num_tokens": 401015859.0,
+      "step": 3750
+    },
+    {
+      "epoch": 8.545039908779932,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.8501483487659217e-07,
+      "loss": 0.5961,
+      "mean_token_accuracy": 0.8768871873617172,
+      "num_tokens": 401122411.0,
+      "step": 3751
+    },
+    {
+      "epoch": 8.547320410490308,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.841421147749709e-07,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8786147236824036,
+      "num_tokens": 401229660.0,
+      "step": 3752
+    },
+    {
+      "epoch": 8.549600912200685,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.832706523515061e-07,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8837973475456238,
+      "num_tokens": 401336325.0,
+      "step": 3753
+    },
+    {
+      "epoch": 8.55188141391106,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.824004481008233e-07,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8784725517034531,
+      "num_tokens": 401444403.0,
+      "step": 3754
+    },
+    {
+      "epoch": 8.554161915621437,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.815315025168339e-07,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.880395233631134,
+      "num_tokens": 401551536.0,
+      "step": 3755
+    },
+    {
+      "epoch": 8.556442417331812,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.8066381609273497e-07,
+      "loss": 0.5666,
+      "mean_token_accuracy": 0.8854324221611023,
+      "num_tokens": 401658365.0,
+      "step": 3756
+    },
+    {
+      "epoch": 8.55872291904219,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.7979738932100734e-07,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8865665048360825,
+      "num_tokens": 401765495.0,
+      "step": 3757
+    },
+    {
+      "epoch": 8.561003420752566,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.7893222269341906e-07,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8840383142232895,
+      "num_tokens": 401872546.0,
+      "step": 3758
+    },
+    {
+      "epoch": 8.563283922462942,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.7806831670102176e-07,
+      "loss": 0.5856,
+      "mean_token_accuracy": 0.8813675791025162,
+      "num_tokens": 401979397.0,
+      "step": 3759
+    },
+    {
+      "epoch": 8.565564424173319,
+      "grad_norm": 2.875,
+      "learning_rate": 2.7720567183415175e-07,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8806767165660858,
+      "num_tokens": 402086207.0,
+      "step": 3760
+    },
+    {
+      "epoch": 8.567844925883694,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.7634428858242995e-07,
+      "loss": 0.5909,
+      "mean_token_accuracy": 0.8802158385515213,
+      "num_tokens": 402192757.0,
+      "step": 3761
+    },
+    {
+      "epoch": 8.570125427594071,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.754841674347608e-07,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8806091845035553,
+      "num_tokens": 402299546.0,
+      "step": 3762
+    },
+    {
+      "epoch": 8.572405929304447,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.7462530887933216e-07,
+      "loss": 0.559,
+      "mean_token_accuracy": 0.8859734088182449,
+      "num_tokens": 402406468.0,
+      "step": 3763
+    },
+    {
+      "epoch": 8.574686431014824,
+      "grad_norm": 3.375,
+      "learning_rate": 2.737677134036154e-07,
+      "loss": 0.5665,
+      "mean_token_accuracy": 0.8834273219108582,
+      "num_tokens": 402513385.0,
+      "step": 3764
+    },
+    {
+      "epoch": 8.576966932725199,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.729113814943654e-07,
+      "loss": 0.5879,
+      "mean_token_accuracy": 0.8773778825998306,
+      "num_tokens": 402620374.0,
+      "step": 3765
+    },
+    {
+      "epoch": 8.579247434435576,
+      "grad_norm": 3.25,
+      "learning_rate": 2.7205631363761976e-07,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8817746788263321,
+      "num_tokens": 402727727.0,
+      "step": 3766
+    },
+    {
+      "epoch": 8.581527936145951,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.7120251031869884e-07,
+      "loss": 0.5487,
+      "mean_token_accuracy": 0.886865645647049,
+      "num_tokens": 402834747.0,
+      "step": 3767
+    },
+    {
+      "epoch": 8.583808437856328,
+      "grad_norm": 3.0,
+      "learning_rate": 2.7034997202220384e-07,
+      "loss": 0.5794,
+      "mean_token_accuracy": 0.8793386965990067,
+      "num_tokens": 402941219.0,
+      "step": 3768
+    },
+    {
+      "epoch": 8.586088939566705,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.6949869923202e-07,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8822144120931625,
+      "num_tokens": 403048515.0,
+      "step": 3769
+    },
+    {
+      "epoch": 8.58836944127708,
+      "grad_norm": 3.0,
+      "learning_rate": 2.686486924313128e-07,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8802978694438934,
+      "num_tokens": 403156281.0,
+      "step": 3770
+    },
+    {
+      "epoch": 8.590649942987458,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.6779995210253015e-07,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8790141940116882,
+      "num_tokens": 403263439.0,
+      "step": 3771
+    },
+    {
+      "epoch": 8.592930444697833,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.6695247872740027e-07,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8814330548048019,
+      "num_tokens": 403370460.0,
+      "step": 3772
+    },
+    {
+      "epoch": 8.59521094640821,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.6610627278693265e-07,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8820819407701492,
+      "num_tokens": 403477203.0,
+      "step": 3773
+    },
+    {
+      "epoch": 8.597491448118586,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.6526133476141804e-07,
+      "loss": 0.5748,
+      "mean_token_accuracy": 0.8831133842468262,
+      "num_tokens": 403583814.0,
+      "step": 3774
+    },
+    {
+      "epoch": 8.599771949828963,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.644176651304259e-07,
+      "loss": 0.5623,
+      "mean_token_accuracy": 0.8832801431417465,
+      "num_tokens": 403690937.0,
+      "step": 3775
+    },
+    {
+      "epoch": 8.602052451539338,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.6357526437280764e-07,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8839538246393204,
+      "num_tokens": 403798148.0,
+      "step": 3776
+    },
+    {
+      "epoch": 8.604332953249715,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.6273413296669353e-07,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8829332441091537,
+      "num_tokens": 403904875.0,
+      "step": 3777
+    },
+    {
+      "epoch": 8.60661345496009,
+      "grad_norm": 2.75,
+      "learning_rate": 2.618942713894937e-07,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8815491497516632,
+      "num_tokens": 404011913.0,
+      "step": 3778
+    },
+    {
+      "epoch": 8.608893956670467,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.610556801178968e-07,
+      "loss": 0.5873,
+      "mean_token_accuracy": 0.8805450052022934,
+      "num_tokens": 404119324.0,
+      "step": 3779
+    },
+    {
+      "epoch": 8.611174458380844,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.602183596278715e-07,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.889171689748764,
+      "num_tokens": 404226267.0,
+      "step": 3780
+    },
+    {
+      "epoch": 8.61345496009122,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.5938231039466436e-07,
+      "loss": 0.5535,
+      "mean_token_accuracy": 0.8866444677114487,
+      "num_tokens": 404333323.0,
+      "step": 3781
+    },
+    {
+      "epoch": 8.615735461801597,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.585475328928011e-07,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8843964338302612,
+      "num_tokens": 404440123.0,
+      "step": 3782
+    },
+    {
+      "epoch": 8.618015963511972,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.577140275960857e-07,
+      "loss": 0.5852,
+      "mean_token_accuracy": 0.882148802280426,
+      "num_tokens": 404546230.0,
+      "step": 3783
+    },
+    {
+      "epoch": 8.62029646522235,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.5688179497759895e-07,
+      "loss": 0.5542,
+      "mean_token_accuracy": 0.8827191293239594,
+      "num_tokens": 404653034.0,
+      "step": 3784
+    },
+    {
+      "epoch": 8.622576966932725,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.560508355097002e-07,
+      "loss": 0.5588,
+      "mean_token_accuracy": 0.8882575631141663,
+      "num_tokens": 404760581.0,
+      "step": 3785
+    },
+    {
+      "epoch": 8.624857468643102,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.552211496640261e-07,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.8854646235704422,
+      "num_tokens": 404868271.0,
+      "step": 3786
+    },
+    {
+      "epoch": 8.627137970353477,
+      "grad_norm": 2.375,
+      "learning_rate": 2.543927379114902e-07,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8811837881803513,
+      "num_tokens": 404974915.0,
+      "step": 3787
+    },
+    {
+      "epoch": 8.629418472063854,
+      "grad_norm": 2.875,
+      "learning_rate": 2.5356560072228335e-07,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8805698156356812,
+      "num_tokens": 405081754.0,
+      "step": 3788
+    },
+    {
+      "epoch": 8.631698973774231,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.5273973856587283e-07,
+      "loss": 0.579,
+      "mean_token_accuracy": 0.879248857498169,
+      "num_tokens": 405188743.0,
+      "step": 3789
+    },
+    {
+      "epoch": 8.633979475484606,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.5191515191100107e-07,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8826890587806702,
+      "num_tokens": 405296046.0,
+      "step": 3790
+    },
+    {
+      "epoch": 8.636259977194984,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.5109184122568797e-07,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.88176129758358,
+      "num_tokens": 405402833.0,
+      "step": 3791
+    },
+    {
+      "epoch": 8.638540478905359,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.502698069772294e-07,
+      "loss": 0.5776,
+      "mean_token_accuracy": 0.8807597756385803,
+      "num_tokens": 405509745.0,
+      "step": 3792
+    },
+    {
+      "epoch": 8.640820980615736,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.494490496321958e-07,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8785258382558823,
+      "num_tokens": 405616827.0,
+      "step": 3793
+    },
+    {
+      "epoch": 8.643101482326111,
+      "grad_norm": 2.375,
+      "learning_rate": 2.4862956965643253e-07,
+      "loss": 0.582,
+      "mean_token_accuracy": 0.8794488459825516,
+      "num_tokens": 405723490.0,
+      "step": 3794
+    },
+    {
+      "epoch": 8.645381984036488,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.4781136751506176e-07,
+      "loss": 0.5634,
+      "mean_token_accuracy": 0.8818620294332504,
+      "num_tokens": 405830225.0,
+      "step": 3795
+    },
+    {
+      "epoch": 8.647662485746864,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.4699444367247834e-07,
+      "loss": 0.5658,
+      "mean_token_accuracy": 0.886525347828865,
+      "num_tokens": 405937278.0,
+      "step": 3796
+    },
+    {
+      "epoch": 8.64994298745724,
+      "grad_norm": 2.625,
+      "learning_rate": 2.461787985923525e-07,
+      "loss": 0.5698,
+      "mean_token_accuracy": 0.8777887523174286,
+      "num_tokens": 406044118.0,
+      "step": 3797
+    },
+    {
+      "epoch": 8.652223489167618,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.4536443273762864e-07,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8797616511583328,
+      "num_tokens": 406151330.0,
+      "step": 3798
+    },
+    {
+      "epoch": 8.654503990877993,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.4455134657052626e-07,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8845881819725037,
+      "num_tokens": 406258516.0,
+      "step": 3799
+    },
+    {
+      "epoch": 8.65678449258837,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.437395405525356e-07,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8821278810501099,
+      "num_tokens": 406364734.0,
+      "step": 3800
+    },
+    {
+      "epoch": 8.659064994298745,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.429290151444233e-07,
+      "loss": 0.5531,
+      "mean_token_accuracy": 0.8842455595731735,
+      "num_tokens": 406471849.0,
+      "step": 3801
+    },
+    {
+      "epoch": 8.661345496009123,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.421197708062273e-07,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8807191848754883,
+      "num_tokens": 406579175.0,
+      "step": 3802
+    },
+    {
+      "epoch": 8.663625997719498,
+      "grad_norm": 4.59375,
+      "learning_rate": 2.413118079972593e-07,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8790168464183807,
+      "num_tokens": 406686082.0,
+      "step": 3803
+    },
+    {
+      "epoch": 8.665906499429875,
+      "grad_norm": 4.0,
+      "learning_rate": 2.405051271761036e-07,
+      "loss": 0.5646,
+      "mean_token_accuracy": 0.8817109614610672,
+      "num_tokens": 406793281.0,
+      "step": 3804
+    },
+    {
+      "epoch": 8.66818700114025,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.396997288006167e-07,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.8831426650285721,
+      "num_tokens": 406900397.0,
+      "step": 3805
+    },
+    {
+      "epoch": 8.670467502850627,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.388956133279266e-07,
+      "loss": 0.5637,
+      "mean_token_accuracy": 0.885161817073822,
+      "num_tokens": 407007825.0,
+      "step": 3806
+    },
+    {
+      "epoch": 8.672748004561003,
+      "grad_norm": 2.875,
+      "learning_rate": 2.3809278121443403e-07,
+      "loss": 0.5742,
+      "mean_token_accuracy": 0.8806300610303879,
+      "num_tokens": 407115154.0,
+      "step": 3807
+    },
+    {
+      "epoch": 8.67502850627138,
+      "grad_norm": 4.625,
+      "learning_rate": 2.3729123291581112e-07,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.8829029500484467,
+      "num_tokens": 407222499.0,
+      "step": 3808
+    },
+    {
+      "epoch": 8.677309007981757,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.3649096888700095e-07,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8817735910415649,
+      "num_tokens": 407329789.0,
+      "step": 3809
+    },
+    {
+      "epoch": 8.679589509692132,
+      "grad_norm": 2.421875,
+      "learning_rate": 2.356919895822188e-07,
+      "loss": 0.5606,
+      "mean_token_accuracy": 0.885416716337204,
+      "num_tokens": 407436906.0,
+      "step": 3810
+    },
+    {
+      "epoch": 8.68187001140251,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.3489429545494851e-07,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8833657205104828,
+      "num_tokens": 407544131.0,
+      "step": 3811
+    },
+    {
+      "epoch": 8.684150513112884,
+      "grad_norm": 2.625,
+      "learning_rate": 2.3409788695794688e-07,
+      "loss": 0.5696,
+      "mean_token_accuracy": 0.8843205720186234,
+      "num_tokens": 407652008.0,
+      "step": 3812
+    },
+    {
+      "epoch": 8.686431014823262,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.3330276454323926e-07,
+      "loss": 0.5669,
+      "mean_token_accuracy": 0.8810292482376099,
+      "num_tokens": 407759542.0,
+      "step": 3813
+    },
+    {
+      "epoch": 8.688711516533637,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.3250892866212294e-07,
+      "loss": 0.5823,
+      "mean_token_accuracy": 0.8779249489307404,
+      "num_tokens": 407865934.0,
+      "step": 3814
+    },
+    {
+      "epoch": 8.690992018244014,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.3171637976516253e-07,
+      "loss": 0.5513,
+      "mean_token_accuracy": 0.8835948258638382,
+      "num_tokens": 407973474.0,
+      "step": 3815
+    },
+    {
+      "epoch": 8.69327251995439,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.3092511830219405e-07,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8819569945335388,
+      "num_tokens": 408080757.0,
+      "step": 3816
+    },
+    {
+      "epoch": 8.695553021664766,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.3013514472232295e-07,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8784191906452179,
+      "num_tokens": 408188778.0,
+      "step": 3817
+    },
+    {
+      "epoch": 8.697833523375142,
+      "grad_norm": 2.421875,
+      "learning_rate": 2.293464594739214e-07,
+      "loss": 0.5593,
+      "mean_token_accuracy": 0.8858485221862793,
+      "num_tokens": 408295942.0,
+      "step": 3818
+    },
+    {
+      "epoch": 8.700114025085519,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.2855906300463305e-07,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.8777198940515518,
+      "num_tokens": 408402995.0,
+      "step": 3819
+    },
+    {
+      "epoch": 8.702394526795896,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.2777295576136865e-07,
+      "loss": 0.5557,
+      "mean_token_accuracy": 0.888022854924202,
+      "num_tokens": 408510620.0,
+      "step": 3820
+    },
+    {
+      "epoch": 8.704675028506271,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.2698813819030802e-07,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8824697285890579,
+      "num_tokens": 408617724.0,
+      "step": 3821
+    },
+    {
+      "epoch": 8.706955530216648,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.2620461073689732e-07,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8824862539768219,
+      "num_tokens": 408724694.0,
+      "step": 3822
+    },
+    {
+      "epoch": 8.709236031927023,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.254223738458522e-07,
+      "loss": 0.5593,
+      "mean_token_accuracy": 0.8855371475219727,
+      "num_tokens": 408832529.0,
+      "step": 3823
+    },
+    {
+      "epoch": 8.7115165336374,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.2464142796115557e-07,
+      "loss": 0.553,
+      "mean_token_accuracy": 0.8828513324260712,
+      "num_tokens": 408939774.0,
+      "step": 3824
+    },
+    {
+      "epoch": 8.713797035347776,
+      "grad_norm": 3.125,
+      "learning_rate": 2.2386177352605677e-07,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8790740817785263,
+      "num_tokens": 409046308.0,
+      "step": 3825
+    },
+    {
+      "epoch": 8.716077537058153,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.2308341098307318e-07,
+      "loss": 0.554,
+      "mean_token_accuracy": 0.8837715536355972,
+      "num_tokens": 409153352.0,
+      "step": 3826
+    },
+    {
+      "epoch": 8.718358038768528,
+      "grad_norm": 3.0,
+      "learning_rate": 2.2230634077398755e-07,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8824053555727005,
+      "num_tokens": 409260531.0,
+      "step": 3827
+    },
+    {
+      "epoch": 8.720638540478905,
+      "grad_norm": 3.0,
+      "learning_rate": 2.2153056333985014e-07,
+      "loss": 0.5613,
+      "mean_token_accuracy": 0.8850871324539185,
+      "num_tokens": 409367568.0,
+      "step": 3828
+    },
+    {
+      "epoch": 8.722919042189282,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.2075607912097758e-07,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8864067494869232,
+      "num_tokens": 409474914.0,
+      "step": 3829
+    },
+    {
+      "epoch": 8.725199543899658,
+      "grad_norm": 5.375,
+      "learning_rate": 2.1998288855695189e-07,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8827249854803085,
+      "num_tokens": 409581981.0,
+      "step": 3830
+    },
+    {
+      "epoch": 8.727480045610035,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.1921099208662173e-07,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8793773353099823,
+      "num_tokens": 409689221.0,
+      "step": 3831
+    },
+    {
+      "epoch": 8.72976054732041,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.184403901480997e-07,
+      "loss": 0.5921,
+      "mean_token_accuracy": 0.8778216391801834,
+      "num_tokens": 409796146.0,
+      "step": 3832
+    },
+    {
+      "epoch": 8.732041049030787,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.176710831787651e-07,
+      "loss": 0.565,
+      "mean_token_accuracy": 0.8842185884714127,
+      "num_tokens": 409902678.0,
+      "step": 3833
+    },
+    {
+      "epoch": 8.734321550741162,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.1690307161526148e-07,
+      "loss": 0.5482,
+      "mean_token_accuracy": 0.8871366381645203,
+      "num_tokens": 410010272.0,
+      "step": 3834
+    },
+    {
+      "epoch": 8.73660205245154,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.1613635589349756e-07,
+      "loss": 0.5492,
+      "mean_token_accuracy": 0.885500431060791,
+      "num_tokens": 410117832.0,
+      "step": 3835
+    },
+    {
+      "epoch": 8.738882554161915,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.153709364486467e-07,
+      "loss": 0.5601,
+      "mean_token_accuracy": 0.8852385729551315,
+      "num_tokens": 410225406.0,
+      "step": 3836
+    },
+    {
+      "epoch": 8.741163055872292,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.1460681371514552e-07,
+      "loss": 0.5574,
+      "mean_token_accuracy": 0.8837675005197525,
+      "num_tokens": 410332633.0,
+      "step": 3837
+    },
+    {
+      "epoch": 8.743443557582669,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.13843988126696e-07,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.8787273913621902,
+      "num_tokens": 410439603.0,
+      "step": 3838
+    },
+    {
+      "epoch": 8.745724059293044,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.130824601162626e-07,
+      "loss": 0.5633,
+      "mean_token_accuracy": 0.8858155012130737,
+      "num_tokens": 410546706.0,
+      "step": 3839
+    },
+    {
+      "epoch": 8.748004561003421,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.1232223011607406e-07,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8737548291683197,
+      "num_tokens": 410653837.0,
+      "step": 3840
+    },
+    {
+      "epoch": 8.750285062713797,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.1156329855762243e-07,
+      "loss": 0.5517,
+      "mean_token_accuracy": 0.8844601809978485,
+      "num_tokens": 410761316.0,
+      "step": 3841
+    },
+    {
+      "epoch": 8.752565564424174,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.1080566587166286e-07,
+      "loss": 0.5618,
+      "mean_token_accuracy": 0.8839665204286575,
+      "num_tokens": 410868322.0,
+      "step": 3842
+    },
+    {
+      "epoch": 8.754846066134549,
+      "grad_norm": 3.25,
+      "learning_rate": 2.1004933248821247e-07,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8804609328508377,
+      "num_tokens": 410976264.0,
+      "step": 3843
+    },
+    {
+      "epoch": 8.757126567844926,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.0929429883655151e-07,
+      "loss": 0.5793,
+      "mean_token_accuracy": 0.8776623606681824,
+      "num_tokens": 411083822.0,
+      "step": 3844
+    },
+    {
+      "epoch": 8.759407069555301,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.08540565345223e-07,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8814112395048141,
+      "num_tokens": 411191400.0,
+      "step": 3845
+    },
+    {
+      "epoch": 8.761687571265679,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.0778813244203111e-07,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8791114687919617,
+      "num_tokens": 411298004.0,
+      "step": 3846
+    },
+    {
+      "epoch": 8.763968072976056,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.0703700055404285e-07,
+      "loss": 0.5826,
+      "mean_token_accuracy": 0.879044771194458,
+      "num_tokens": 411404179.0,
+      "step": 3847
+    },
+    {
+      "epoch": 8.766248574686431,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.0628717010758526e-07,
+      "loss": 0.5598,
+      "mean_token_accuracy": 0.8879903703927994,
+      "num_tokens": 411511695.0,
+      "step": 3848
+    },
+    {
+      "epoch": 8.768529076396808,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.0553864152824815e-07,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.8788136094808578,
+      "num_tokens": 411619621.0,
+      "step": 3849
+    },
+    {
+      "epoch": 8.770809578107183,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.0479141524088169e-07,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.883419394493103,
+      "num_tokens": 411726749.0,
+      "step": 3850
+    },
+    {
+      "epoch": 8.77309007981756,
+      "grad_norm": 3.0,
+      "learning_rate": 2.040454916695972e-07,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8799424916505814,
+      "num_tokens": 411833865.0,
+      "step": 3851
+    },
+    {
+      "epoch": 8.775370581527936,
+      "grad_norm": 2.46875,
+      "learning_rate": 2.0330087123776655e-07,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.884945884346962,
+      "num_tokens": 411941568.0,
+      "step": 3852
+    },
+    {
+      "epoch": 8.777651083238313,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.0255755436802248e-07,
+      "loss": 0.5659,
+      "mean_token_accuracy": 0.884887769818306,
+      "num_tokens": 412048897.0,
+      "step": 3853
+    },
+    {
+      "epoch": 8.779931584948688,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.0181554148225618e-07,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8816755414009094,
+      "num_tokens": 412155508.0,
+      "step": 3854
+    },
+    {
+      "epoch": 8.782212086659065,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.0107483300162018e-07,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.885670393705368,
+      "num_tokens": 412262951.0,
+      "step": 3855
+    },
+    {
+      "epoch": 8.78449258836944,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.0033542934652679e-07,
+      "loss": 0.5682,
+      "mean_token_accuracy": 0.8826614916324615,
+      "num_tokens": 412370424.0,
+      "step": 3856
+    },
+    {
+      "epoch": 8.786773090079818,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.9959733093664696e-07,
+      "loss": 0.5513,
+      "mean_token_accuracy": 0.8853038400411606,
+      "num_tokens": 412478059.0,
+      "step": 3857
+    },
+    {
+      "epoch": 8.789053591790195,
+      "grad_norm": 2.875,
+      "learning_rate": 1.9886053819091116e-07,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8780199140310287,
+      "num_tokens": 412584823.0,
+      "step": 3858
+    },
+    {
+      "epoch": 8.79133409350057,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.981250515275085e-07,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8827866017818451,
+      "num_tokens": 412692334.0,
+      "step": 3859
+    },
+    {
+      "epoch": 8.793614595210947,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.973908713638878e-07,
+      "loss": 0.5509,
+      "mean_token_accuracy": 0.8867141306400299,
+      "num_tokens": 412799411.0,
+      "step": 3860
+    },
+    {
+      "epoch": 8.795895096921322,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.9665799811675407e-07,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.8858144879341125,
+      "num_tokens": 412906629.0,
+      "step": 3861
+    },
+    {
+      "epoch": 8.7981755986317,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.959264322020732e-07,
+      "loss": 0.5846,
+      "mean_token_accuracy": 0.8803358823060989,
+      "num_tokens": 413013873.0,
+      "step": 3862
+    },
+    {
+      "epoch": 8.800456100342075,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.9519617403506747e-07,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8833809643983841,
+      "num_tokens": 413120858.0,
+      "step": 3863
+    },
+    {
+      "epoch": 8.802736602052452,
+      "grad_norm": 3.0,
+      "learning_rate": 1.9446722403021757e-07,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8800954222679138,
+      "num_tokens": 413227696.0,
+      "step": 3864
+    },
+    {
+      "epoch": 8.805017103762827,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.9373958260126113e-07,
+      "loss": 0.5579,
+      "mean_token_accuracy": 0.8825733810663223,
+      "num_tokens": 413334739.0,
+      "step": 3865
+    },
+    {
+      "epoch": 8.807297605473204,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.9301325016119338e-07,
+      "loss": 0.5645,
+      "mean_token_accuracy": 0.8862461596727371,
+      "num_tokens": 413441973.0,
+      "step": 3866
+    },
+    {
+      "epoch": 8.80957810718358,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.9228822712226675e-07,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.880901426076889,
+      "num_tokens": 413549047.0,
+      "step": 3867
+    },
+    {
+      "epoch": 8.811858608893957,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.915645138959904e-07,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8816043883562088,
+      "num_tokens": 413656651.0,
+      "step": 3868
+    },
+    {
+      "epoch": 8.814139110604334,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.908421108931302e-07,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8817258775234222,
+      "num_tokens": 413763796.0,
+      "step": 3869
+    },
+    {
+      "epoch": 8.816419612314709,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.9012101852370763e-07,
+      "loss": 0.5479,
+      "mean_token_accuracy": 0.8846326470375061,
+      "num_tokens": 413871019.0,
+      "step": 3870
+    },
+    {
+      "epoch": 8.818700114025086,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.894012371970008e-07,
+      "loss": 0.556,
+      "mean_token_accuracy": 0.8854041546583176,
+      "num_tokens": 413978295.0,
+      "step": 3871
+    },
+    {
+      "epoch": 8.820980615735461,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.8868276732154384e-07,
+      "loss": 0.5792,
+      "mean_token_accuracy": 0.8823095858097076,
+      "num_tokens": 414085389.0,
+      "step": 3872
+    },
+    {
+      "epoch": 8.823261117445838,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.879656093051266e-07,
+      "loss": 0.5627,
+      "mean_token_accuracy": 0.8850487917661667,
+      "num_tokens": 414192622.0,
+      "step": 3873
+    },
+    {
+      "epoch": 8.825541619156214,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.872497635547943e-07,
+      "loss": 0.5427,
+      "mean_token_accuracy": 0.8869512677192688,
+      "num_tokens": 414299632.0,
+      "step": 3874
+    },
+    {
+      "epoch": 8.82782212086659,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.8653523047684642e-07,
+      "loss": 0.5858,
+      "mean_token_accuracy": 0.882158100605011,
+      "num_tokens": 414406824.0,
+      "step": 3875
+    },
+    {
+      "epoch": 8.830102622576966,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.858220104768385e-07,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8781119138002396,
+      "num_tokens": 414513673.0,
+      "step": 3876
+    },
+    {
+      "epoch": 8.832383124287343,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.8511010395958067e-07,
+      "loss": 0.5843,
+      "mean_token_accuracy": 0.8815079480409622,
+      "num_tokens": 414621071.0,
+      "step": 3877
+    },
+    {
+      "epoch": 8.83466362599772,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.843995113291372e-07,
+      "loss": 0.544,
+      "mean_token_accuracy": 0.8865767568349838,
+      "num_tokens": 414728284.0,
+      "step": 3878
+    },
+    {
+      "epoch": 8.836944127708096,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.836902329888268e-07,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8805812895298004,
+      "num_tokens": 414835107.0,
+      "step": 3879
+    },
+    {
+      "epoch": 8.839224629418473,
+      "grad_norm": 3.875,
+      "learning_rate": 1.829822693412217e-07,
+      "loss": 0.5468,
+      "mean_token_accuracy": 0.8886751085519791,
+      "num_tokens": 414942335.0,
+      "step": 3880
+    },
+    {
+      "epoch": 8.841505131128848,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.8227562078814903e-07,
+      "loss": 0.5498,
+      "mean_token_accuracy": 0.8845852613449097,
+      "num_tokens": 415050100.0,
+      "step": 3881
+    },
+    {
+      "epoch": 8.843785632839225,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.815702877306888e-07,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8805902749300003,
+      "num_tokens": 415157356.0,
+      "step": 3882
+    },
+    {
+      "epoch": 8.8460661345496,
+      "grad_norm": 5.375,
+      "learning_rate": 1.8086627056917382e-07,
+      "loss": 0.5927,
+      "mean_token_accuracy": 0.878188282251358,
+      "num_tokens": 415264053.0,
+      "step": 3883
+    },
+    {
+      "epoch": 8.848346636259977,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.8016356970319116e-07,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.883014589548111,
+      "num_tokens": 415371061.0,
+      "step": 3884
+    },
+    {
+      "epoch": 8.850627137970353,
+      "grad_norm": 4.5625,
+      "learning_rate": 1.7946218553158062e-07,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8840430676937103,
+      "num_tokens": 415478709.0,
+      "step": 3885
+    },
+    {
+      "epoch": 8.85290763968073,
+      "grad_norm": 3.375,
+      "learning_rate": 1.7876211845243325e-07,
+      "loss": 0.5506,
+      "mean_token_accuracy": 0.8877336978912354,
+      "num_tokens": 415586877.0,
+      "step": 3886
+    },
+    {
+      "epoch": 8.855188141391107,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.780633688630942e-07,
+      "loss": 0.5946,
+      "mean_token_accuracy": 0.8771243989467621,
+      "num_tokens": 415693262.0,
+      "step": 3887
+    },
+    {
+      "epoch": 8.857468643101482,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.773659371601605e-07,
+      "loss": 0.537,
+      "mean_token_accuracy": 0.8894240111112595,
+      "num_tokens": 415800803.0,
+      "step": 3888
+    },
+    {
+      "epoch": 8.85974914481186,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.7666982373948038e-07,
+      "loss": 0.5917,
+      "mean_token_accuracy": 0.8793479204177856,
+      "num_tokens": 415907854.0,
+      "step": 3889
+    },
+    {
+      "epoch": 8.862029646522235,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.7597502899615538e-07,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.8833720535039902,
+      "num_tokens": 416015169.0,
+      "step": 3890
+    },
+    {
+      "epoch": 8.864310148232612,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.752815533245364e-07,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8827664852142334,
+      "num_tokens": 416122300.0,
+      "step": 3891
+    },
+    {
+      "epoch": 8.866590649942987,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.745893971182272e-07,
+      "loss": 0.6053,
+      "mean_token_accuracy": 0.8742794394493103,
+      "num_tokens": 416228785.0,
+      "step": 3892
+    },
+    {
+      "epoch": 8.868871151653364,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.7389856077008245e-07,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8836385905742645,
+      "num_tokens": 416336321.0,
+      "step": 3893
+    },
+    {
+      "epoch": 8.87115165336374,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.7320904467220762e-07,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8823292106389999,
+      "num_tokens": 416443558.0,
+      "step": 3894
+    },
+    {
+      "epoch": 8.873432155074116,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.725208492159583e-07,
+      "loss": 0.5829,
+      "mean_token_accuracy": 0.8806789815425873,
+      "num_tokens": 416550695.0,
+      "step": 3895
+    },
+    {
+      "epoch": 8.875712656784494,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.7183397479194175e-07,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8829580396413803,
+      "num_tokens": 416657754.0,
+      "step": 3896
+    },
+    {
+      "epoch": 8.877993158494869,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.711484217900139e-07,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8825534284114838,
+      "num_tokens": 416764992.0,
+      "step": 3897
+    },
+    {
+      "epoch": 8.880273660205246,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.7046419059928154e-07,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8812769651412964,
+      "num_tokens": 416872041.0,
+      "step": 3898
+    },
+    {
+      "epoch": 8.882554161915621,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.6978128160810098e-07,
+      "loss": 0.5447,
+      "mean_token_accuracy": 0.8902477920055389,
+      "num_tokens": 416979187.0,
+      "step": 3899
+    },
+    {
+      "epoch": 8.884834663625998,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.6909969520407854e-07,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8812951445579529,
+      "num_tokens": 417086032.0,
+      "step": 3900
+    },
+    {
+      "epoch": 8.887115165336374,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.6841943177406976e-07,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8830399364233017,
+      "num_tokens": 417192794.0,
+      "step": 3901
+    },
+    {
+      "epoch": 8.88939566704675,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.6774049170417806e-07,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8833406120538712,
+      "num_tokens": 417300018.0,
+      "step": 3902
+    },
+    {
+      "epoch": 8.891676168757126,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.6706287537975763e-07,
+      "loss": 0.5555,
+      "mean_token_accuracy": 0.8872157335281372,
+      "num_tokens": 417407045.0,
+      "step": 3903
+    },
+    {
+      "epoch": 8.893956670467503,
+      "grad_norm": 3.375,
+      "learning_rate": 1.6638658318540973e-07,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8826020061969757,
+      "num_tokens": 417514079.0,
+      "step": 3904
+    },
+    {
+      "epoch": 8.896237172177878,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.657116155049851e-07,
+      "loss": 0.5573,
+      "mean_token_accuracy": 0.8848617970943451,
+      "num_tokens": 417621403.0,
+      "step": 3905
+    },
+    {
+      "epoch": 8.898517673888255,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.6503797272158284e-07,
+      "loss": 0.571,
+      "mean_token_accuracy": 0.8829448074102402,
+      "num_tokens": 417728575.0,
+      "step": 3906
+    },
+    {
+      "epoch": 8.900798175598633,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.643656552175485e-07,
+      "loss": 0.5632,
+      "mean_token_accuracy": 0.8809054046869278,
+      "num_tokens": 417835651.0,
+      "step": 3907
+    },
+    {
+      "epoch": 8.903078677309008,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.6369466337447708e-07,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8784161955118179,
+      "num_tokens": 417943079.0,
+      "step": 3908
+    },
+    {
+      "epoch": 8.905359179019385,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.6302499757321066e-07,
+      "loss": 0.5859,
+      "mean_token_accuracy": 0.8787552714347839,
+      "num_tokens": 418050175.0,
+      "step": 3909
+    },
+    {
+      "epoch": 8.90763968072976,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.623566581938385e-07,
+      "loss": 0.5655,
+      "mean_token_accuracy": 0.8856528103351593,
+      "num_tokens": 418157102.0,
+      "step": 3910
+    },
+    {
+      "epoch": 8.909920182440137,
+      "grad_norm": 3.5,
+      "learning_rate": 1.6168964561569716e-07,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.880318284034729,
+      "num_tokens": 418264196.0,
+      "step": 3911
+    },
+    {
+      "epoch": 8.912200684150513,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.6102396021737077e-07,
+      "loss": 0.5442,
+      "mean_token_accuracy": 0.886517733335495,
+      "num_tokens": 418371298.0,
+      "step": 3912
+    },
+    {
+      "epoch": 8.91448118586089,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.6035960237668818e-07,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8844355195760727,
+      "num_tokens": 418478409.0,
+      "step": 3913
+    },
+    {
+      "epoch": 8.916761687571265,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.5969657247072695e-07,
+      "loss": 0.5548,
+      "mean_token_accuracy": 0.8860626816749573,
+      "num_tokens": 418585134.0,
+      "step": 3914
+    },
+    {
+      "epoch": 8.919042189281642,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.5903487087580994e-07,
+      "loss": 0.5435,
+      "mean_token_accuracy": 0.8894283324480057,
+      "num_tokens": 418691838.0,
+      "step": 3915
+    },
+    {
+      "epoch": 8.921322690992017,
+      "grad_norm": 3.0,
+      "learning_rate": 1.5837449796750588e-07,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.8798937052488327,
+      "num_tokens": 418798999.0,
+      "step": 3916
+    },
+    {
+      "epoch": 8.923603192702394,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.577154541206305e-07,
+      "loss": 0.5511,
+      "mean_token_accuracy": 0.8867987394332886,
+      "num_tokens": 418906023.0,
+      "step": 3917
+    },
+    {
+      "epoch": 8.925883694412772,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.5705773970924349e-07,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8823074996471405,
+      "num_tokens": 419013460.0,
+      "step": 3918
+    },
+    {
+      "epoch": 8.928164196123147,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.5640135510665094e-07,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.8819030374288559,
+      "num_tokens": 419120119.0,
+      "step": 3919
+    },
+    {
+      "epoch": 8.930444697833524,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.5574630068540458e-07,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.8858517855405807,
+      "num_tokens": 419227218.0,
+      "step": 3920
+    },
+    {
+      "epoch": 8.9327251995439,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.5509257681730034e-07,
+      "loss": 0.5511,
+      "mean_token_accuracy": 0.8868541866540909,
+      "num_tokens": 419334292.0,
+      "step": 3921
+    },
+    {
+      "epoch": 8.935005701254276,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.5444018387337946e-07,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8833621442317963,
+      "num_tokens": 419441056.0,
+      "step": 3922
+    },
+    {
+      "epoch": 8.937286202964652,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.537891222239271e-07,
+      "loss": 0.5803,
+      "mean_token_accuracy": 0.8798842430114746,
+      "num_tokens": 419548438.0,
+      "step": 3923
+    },
+    {
+      "epoch": 8.939566704675029,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.5313939223847384e-07,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8819588273763657,
+      "num_tokens": 419655399.0,
+      "step": 3924
+    },
+    {
+      "epoch": 8.941847206385404,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.5249099428579383e-07,
+      "loss": 0.5513,
+      "mean_token_accuracy": 0.8892662078142166,
+      "num_tokens": 419763644.0,
+      "step": 3925
+    },
+    {
+      "epoch": 8.944127708095781,
+      "grad_norm": 2.875,
+      "learning_rate": 1.5184392873390463e-07,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8778876066207886,
+      "num_tokens": 419870546.0,
+      "step": 3926
+    },
+    {
+      "epoch": 8.946408209806158,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.5119819595006857e-07,
+      "loss": 0.5773,
+      "mean_token_accuracy": 0.8791385740041733,
+      "num_tokens": 419977393.0,
+      "step": 3927
+    },
+    {
+      "epoch": 8.948688711516533,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.5055379630079163e-07,
+      "loss": 0.5893,
+      "mean_token_accuracy": 0.8777313679456711,
+      "num_tokens": 420084260.0,
+      "step": 3928
+    },
+    {
+      "epoch": 8.95096921322691,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.4991073015182184e-07,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8823111951351166,
+      "num_tokens": 420191156.0,
+      "step": 3929
+    },
+    {
+      "epoch": 8.953249714937286,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.4926899786815107e-07,
+      "loss": 0.5715,
+      "mean_token_accuracy": 0.883946105837822,
+      "num_tokens": 420298497.0,
+      "step": 3930
+    },
+    {
+      "epoch": 8.955530216647663,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.4862859981401468e-07,
+      "loss": 0.583,
+      "mean_token_accuracy": 0.882349282503128,
+      "num_tokens": 420405474.0,
+      "step": 3931
+    },
+    {
+      "epoch": 8.957810718358038,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.4798953635288994e-07,
+      "loss": 0.5589,
+      "mean_token_accuracy": 0.8836222589015961,
+      "num_tokens": 420513022.0,
+      "step": 3932
+    },
+    {
+      "epoch": 8.960091220068415,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.4735180784749754e-07,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.880601704120636,
+      "num_tokens": 420620084.0,
+      "step": 3933
+    },
+    {
+      "epoch": 8.96237172177879,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.4671541465979877e-07,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8797517418861389,
+      "num_tokens": 420726931.0,
+      "step": 3934
+    },
+    {
+      "epoch": 8.964652223489168,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.460803571509989e-07,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8833180367946625,
+      "num_tokens": 420834187.0,
+      "step": 3935
+    },
+    {
+      "epoch": 8.966932725199545,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.4544663568154427e-07,
+      "loss": 0.5821,
+      "mean_token_accuracy": 0.8844758570194244,
+      "num_tokens": 420940822.0,
+      "step": 3936
+    },
+    {
+      "epoch": 8.96921322690992,
+      "grad_norm": 3.125,
+      "learning_rate": 1.448142506111225e-07,
+      "loss": 0.568,
+      "mean_token_accuracy": 0.8841772228479385,
+      "num_tokens": 421048221.0,
+      "step": 3937
+    },
+    {
+      "epoch": 8.971493728620297,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.441832022986636e-07,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8786625862121582,
+      "num_tokens": 421154952.0,
+      "step": 3938
+    },
+    {
+      "epoch": 8.973774230330672,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.4355349110233868e-07,
+      "loss": 0.5533,
+      "mean_token_accuracy": 0.8868290036916733,
+      "num_tokens": 421261851.0,
+      "step": 3939
+    },
+    {
+      "epoch": 8.97605473204105,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.42925117379559e-07,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.8779198080301285,
+      "num_tokens": 421368450.0,
+      "step": 3940
+    },
+    {
+      "epoch": 8.978335233751425,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.4229808148697732e-07,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8809890896081924,
+      "num_tokens": 421475775.0,
+      "step": 3941
+    },
+    {
+      "epoch": 8.980615735461802,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.416723837804876e-07,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8843148350715637,
+      "num_tokens": 421582954.0,
+      "step": 3942
+    },
+    {
+      "epoch": 8.982896237172177,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.410480246152235e-07,
+      "loss": 0.567,
+      "mean_token_accuracy": 0.8814869672060013,
+      "num_tokens": 421690190.0,
+      "step": 3943
+    },
+    {
+      "epoch": 8.985176738882554,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.4042500434555961e-07,
+      "loss": 0.5812,
+      "mean_token_accuracy": 0.8802603036165237,
+      "num_tokens": 421797293.0,
+      "step": 3944
+    },
+    {
+      "epoch": 8.987457240592931,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.398033233251095e-07,
+      "loss": 0.5761,
+      "mean_token_accuracy": 0.8830568790435791,
+      "num_tokens": 421904481.0,
+      "step": 3945
+    },
+    {
+      "epoch": 8.989737742303307,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.3918298190672806e-07,
+      "loss": 0.5455,
+      "mean_token_accuracy": 0.8889565318822861,
+      "num_tokens": 422011408.0,
+      "step": 3946
+    },
+    {
+      "epoch": 8.992018244013684,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.3856398044250846e-07,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.8796057999134064,
+      "num_tokens": 422118456.0,
+      "step": 3947
+    },
+    {
+      "epoch": 8.994298745724059,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.3794631928378434e-07,
+      "loss": 0.5586,
+      "mean_token_accuracy": 0.8840108662843704,
+      "num_tokens": 422225963.0,
+      "step": 3948
+    },
+    {
+      "epoch": 8.996579247434436,
+      "grad_norm": 2.515625,
+      "learning_rate": 1.3732999878112856e-07,
+      "loss": 0.5605,
+      "mean_token_accuracy": 0.8850581645965576,
+      "num_tokens": 422333254.0,
+      "step": 3949
+    },
+    {
+      "epoch": 8.998859749144811,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.3671501928435193e-07,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8808369934558868,
+      "num_tokens": 422440753.0,
+      "step": 3950
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 9.0,
+      "learning_rate": 1.361013811425052e-07,
+      "loss": 0.5721,
+      "mean_token_accuracy": 0.8868737518787384,
+      "num_tokens": 422480088.0,
+      "step": 3951
+    },
+    {
+      "epoch": 9.002280501710377,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.3548908470387783e-07,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8844682276248932,
+      "num_tokens": 422587127.0,
+      "step": 3952
+    },
+    {
+      "epoch": 9.004561003420752,
+      "grad_norm": 2.421875,
+      "learning_rate": 1.348781303159974e-07,
+      "loss": 0.5231,
+      "mean_token_accuracy": 0.8936333805322647,
+      "num_tokens": 422694591.0,
+      "step": 3953
+    },
+    {
+      "epoch": 9.00684150513113,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.3426851832562982e-07,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.8864496648311615,
+      "num_tokens": 422802195.0,
+      "step": 3954
+    },
+    {
+      "epoch": 9.009122006841505,
+      "grad_norm": 2.5,
+      "learning_rate": 1.3366024907877917e-07,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.8846766799688339,
+      "num_tokens": 422909976.0,
+      "step": 3955
+    },
+    {
+      "epoch": 9.011402508551882,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.3305332292068706e-07,
+      "loss": 0.5821,
+      "mean_token_accuracy": 0.8805938214063644,
+      "num_tokens": 423016740.0,
+      "step": 3956
+    },
+    {
+      "epoch": 9.013683010262257,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.3244774019583296e-07,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8852843195199966,
+      "num_tokens": 423123757.0,
+      "step": 3957
+    },
+    {
+      "epoch": 9.015963511972634,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.318435012479341e-07,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8831017911434174,
+      "num_tokens": 423229906.0,
+      "step": 3958
+    },
+    {
+      "epoch": 9.01824401368301,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.3124060641994507e-07,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.88319331407547,
+      "num_tokens": 423337912.0,
+      "step": 3959
+    },
+    {
+      "epoch": 9.020524515393387,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.306390560540577e-07,
+      "loss": 0.5514,
+      "mean_token_accuracy": 0.8833757936954498,
+      "num_tokens": 423445320.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.020524515393387,
+      "eval_loss": 0.5863176584243774,
+      "eval_mean_token_accuracy": 0.8799216114976107,
+      "eval_num_tokens": 423445320.0,
+      "eval_runtime": 58.6917,
+      "eval_samples_per_second": 142.865,
+      "eval_steps_per_second": 4.481,
+      "step": 3960
+    },
+    {
+      "epoch": 9.022805017103764,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.300388504916991e-07,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8808793723583221,
+      "num_tokens": 423551972.0,
+      "step": 3961
+    },
+    {
+      "epoch": 9.025085518814139,
+      "grad_norm": 3.375,
+      "learning_rate": 1.2943999007353518e-07,
+      "loss": 0.5633,
+      "mean_token_accuracy": 0.8862560987472534,
+      "num_tokens": 423658903.0,
+      "step": 3962
+    },
+    {
+      "epoch": 9.027366020524516,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.2884247513946761e-07,
+      "loss": 0.5818,
+      "mean_token_accuracy": 0.8841415643692017,
+      "num_tokens": 423765920.0,
+      "step": 3963
+    },
+    {
+      "epoch": 9.029646522234891,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.2824630602863402e-07,
+      "loss": 0.5812,
+      "mean_token_accuracy": 0.8800481855869293,
+      "num_tokens": 423872709.0,
+      "step": 3964
+    },
+    {
+      "epoch": 9.031927023945268,
+      "grad_norm": 2.75,
+      "learning_rate": 1.2765148307940927e-07,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8826144337654114,
+      "num_tokens": 423979876.0,
+      "step": 3965
+    },
+    {
+      "epoch": 9.034207525655644,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.270580066294022e-07,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8837986141443253,
+      "num_tokens": 424086963.0,
+      "step": 3966
+    },
+    {
+      "epoch": 9.03648802736602,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.264658770154592e-07,
+      "loss": 0.5676,
+      "mean_token_accuracy": 0.8831162303686142,
+      "num_tokens": 424193595.0,
+      "step": 3967
+    },
+    {
+      "epoch": 9.038768529076396,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.258750945736617e-07,
+      "loss": 0.5552,
+      "mean_token_accuracy": 0.8876445442438126,
+      "num_tokens": 424300328.0,
+      "step": 3968
+    },
+    {
+      "epoch": 9.041049030786773,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.252856596393262e-07,
+      "loss": 0.5579,
+      "mean_token_accuracy": 0.8856015801429749,
+      "num_tokens": 424407500.0,
+      "step": 3969
+    },
+    {
+      "epoch": 9.043329532497149,
+      "grad_norm": 2.75,
+      "learning_rate": 1.2469757254700454e-07,
+      "loss": 0.5624,
+      "mean_token_accuracy": 0.8857114911079407,
+      "num_tokens": 424515230.0,
+      "step": 3970
+    },
+    {
+      "epoch": 9.045610034207526,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.2411083363048386e-07,
+      "loss": 0.5901,
+      "mean_token_accuracy": 0.878177598118782,
+      "num_tokens": 424622085.0,
+      "step": 3971
+    },
+    {
+      "epoch": 9.047890535917903,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.2352544322278558e-07,
+      "loss": 0.5995,
+      "mean_token_accuracy": 0.8785828799009323,
+      "num_tokens": 424728586.0,
+      "step": 3972
+    },
+    {
+      "epoch": 9.050171037628278,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.2294140165616613e-07,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.883171334862709,
+      "num_tokens": 424835689.0,
+      "step": 3973
+    },
+    {
+      "epoch": 9.052451539338655,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.223587092621162e-07,
+      "loss": 0.5783,
+      "mean_token_accuracy": 0.880466416478157,
+      "num_tokens": 424942400.0,
+      "step": 3974
+    },
+    {
+      "epoch": 9.05473204104903,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.2177736637136063e-07,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8840536028146744,
+      "num_tokens": 425049454.0,
+      "step": 3975
+    },
+    {
+      "epoch": 9.057012542759407,
+      "grad_norm": 4.125,
+      "learning_rate": 1.2119737331385885e-07,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8803068101406097,
+      "num_tokens": 425156346.0,
+      "step": 3976
+    },
+    {
+      "epoch": 9.059293044469783,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.2061873041880335e-07,
+      "loss": 0.5798,
+      "mean_token_accuracy": 0.8801010698080063,
+      "num_tokens": 425262768.0,
+      "step": 3977
+    },
+    {
+      "epoch": 9.06157354618016,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.200414380146206e-07,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.882665142416954,
+      "num_tokens": 425370275.0,
+      "step": 3978
+    },
+    {
+      "epoch": 9.063854047890535,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.1946549642897043e-07,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8818689733743668,
+      "num_tokens": 425477099.0,
+      "step": 3979
+    },
+    {
+      "epoch": 9.066134549600912,
+      "grad_norm": 3.125,
+      "learning_rate": 1.1889090598874692e-07,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8815028220415115,
+      "num_tokens": 425584752.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.06841505131129,
+      "grad_norm": 4.25,
+      "learning_rate": 1.1831766702007613e-07,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8803604692220688,
+      "num_tokens": 425691333.0,
+      "step": 3981
+    },
+    {
+      "epoch": 9.070695553021665,
+      "grad_norm": 3.0,
+      "learning_rate": 1.1774577984831725e-07,
+      "loss": 0.5801,
+      "mean_token_accuracy": 0.8821363896131516,
+      "num_tokens": 425798165.0,
+      "step": 3982
+    },
+    {
+      "epoch": 9.072976054732042,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.1717524479806231e-07,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8794075697660446,
+      "num_tokens": 425904837.0,
+      "step": 3983
+    },
+    {
+      "epoch": 9.075256556442417,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.1660606219313642e-07,
+      "loss": 0.5743,
+      "mean_token_accuracy": 0.8835705667734146,
+      "num_tokens": 426012137.0,
+      "step": 3984
+    },
+    {
+      "epoch": 9.077537058152794,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.1603823235659644e-07,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.8850951492786407,
+      "num_tokens": 426119123.0,
+      "step": 3985
+    },
+    {
+      "epoch": 9.07981755986317,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.1547175561073154e-07,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8812219202518463,
+      "num_tokens": 426226278.0,
+      "step": 3986
+    },
+    {
+      "epoch": 9.082098061573546,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.1490663227706311e-07,
+      "loss": 0.5866,
+      "mean_token_accuracy": 0.88066066801548,
+      "num_tokens": 426333015.0,
+      "step": 3987
+    },
+    {
+      "epoch": 9.084378563283922,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.1434286267634432e-07,
+      "loss": 0.5532,
+      "mean_token_accuracy": 0.8882465213537216,
+      "num_tokens": 426440373.0,
+      "step": 3988
+    },
+    {
+      "epoch": 9.086659064994299,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.1378044712855946e-07,
+      "loss": 0.6018,
+      "mean_token_accuracy": 0.8752111494541168,
+      "num_tokens": 426546913.0,
+      "step": 3989
+    },
+    {
+      "epoch": 9.088939566704674,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.1321938595292542e-07,
+      "loss": 0.568,
+      "mean_token_accuracy": 0.8806653469800949,
+      "num_tokens": 426654053.0,
+      "step": 3990
+    },
+    {
+      "epoch": 9.091220068415051,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.1265967946788913e-07,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8836180865764618,
+      "num_tokens": 426761077.0,
+      "step": 3991
+    },
+    {
+      "epoch": 9.093500570125428,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.1210132799112954e-07,
+      "loss": 0.5426,
+      "mean_token_accuracy": 0.8866991996765137,
+      "num_tokens": 426868400.0,
+      "step": 3992
+    },
+    {
+      "epoch": 9.095781071835804,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.1154433183955593e-07,
+      "loss": 0.5666,
+      "mean_token_accuracy": 0.880575105547905,
+      "num_tokens": 426975681.0,
+      "step": 3993
+    },
+    {
+      "epoch": 9.09806157354618,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.1098869132930846e-07,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.880921483039856,
+      "num_tokens": 427082443.0,
+      "step": 3994
+    },
+    {
+      "epoch": 9.100342075256556,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.1043440677575818e-07,
+      "loss": 0.5642,
+      "mean_token_accuracy": 0.88336580991745,
+      "num_tokens": 427189040.0,
+      "step": 3995
+    },
+    {
+      "epoch": 9.102622576966933,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.0988147849350623e-07,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8802250772714615,
+      "num_tokens": 427295944.0,
+      "step": 3996
+    },
+    {
+      "epoch": 9.104903078677308,
+      "grad_norm": 2.46875,
+      "learning_rate": 1.0932990679638406e-07,
+      "loss": 0.5563,
+      "mean_token_accuracy": 0.8847872316837311,
+      "num_tokens": 427402191.0,
+      "step": 3997
+    },
+    {
+      "epoch": 9.107183580387685,
+      "grad_norm": 3.5,
+      "learning_rate": 1.0877969199745347e-07,
+      "loss": 0.5764,
+      "mean_token_accuracy": 0.8834807723760605,
+      "num_tokens": 427509134.0,
+      "step": 3998
+    },
+    {
+      "epoch": 9.10946408209806,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.0823083440900523e-07,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.8807051777839661,
+      "num_tokens": 427616092.0,
+      "step": 3999
+    },
+    {
+      "epoch": 9.111744583808438,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.0768333434256039e-07,
+      "loss": 0.587,
+      "mean_token_accuracy": 0.8833329081535339,
+      "num_tokens": 427723041.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.114025085518815,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.071371921088693e-07,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8830773681402206,
+      "num_tokens": 427830278.0,
+      "step": 4001
+    },
+    {
+      "epoch": 9.11630558722919,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.0659240801791204e-07,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8835585117340088,
+      "num_tokens": 427937452.0,
+      "step": 4002
+    },
+    {
+      "epoch": 9.118586088939567,
+      "grad_norm": 3.125,
+      "learning_rate": 1.0604898237889794e-07,
+      "loss": 0.5665,
+      "mean_token_accuracy": 0.8840138912200928,
+      "num_tokens": 428044476.0,
+      "step": 4003
+    },
+    {
+      "epoch": 9.120866590649943,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.0550691550026415e-07,
+      "loss": 0.5759,
+      "mean_token_accuracy": 0.8799526989459991,
+      "num_tokens": 428151794.0,
+      "step": 4004
+    },
+    {
+      "epoch": 9.12314709236032,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.0496620768967736e-07,
+      "loss": 0.5874,
+      "mean_token_accuracy": 0.8784212470054626,
+      "num_tokens": 428258124.0,
+      "step": 4005
+    },
+    {
+      "epoch": 9.125427594070695,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.0442685925403346e-07,
+      "loss": 0.5583,
+      "mean_token_accuracy": 0.8846355378627777,
+      "num_tokens": 428365889.0,
+      "step": 4006
+    },
+    {
+      "epoch": 9.127708095781072,
+      "grad_norm": 4.125,
+      "learning_rate": 1.0388887049945589e-07,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8810495138168335,
+      "num_tokens": 428472825.0,
+      "step": 4007
+    },
+    {
+      "epoch": 9.129988597491447,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.0335224173129683e-07,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8812502026557922,
+      "num_tokens": 428580741.0,
+      "step": 4008
+    },
+    {
+      "epoch": 9.132269099201825,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.0281697325413593e-07,
+      "loss": 0.5508,
+      "mean_token_accuracy": 0.8845276087522507,
+      "num_tokens": 428687964.0,
+      "step": 4009
+    },
+    {
+      "epoch": 9.134549600912202,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.0228306537178185e-07,
+      "loss": 0.5828,
+      "mean_token_accuracy": 0.8836958706378937,
+      "num_tokens": 428794803.0,
+      "step": 4010
+    },
+    {
+      "epoch": 9.136830102622577,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.0175051838727023e-07,
+      "loss": 0.5516,
+      "mean_token_accuracy": 0.8883339315652847,
+      "num_tokens": 428901676.0,
+      "step": 4011
+    },
+    {
+      "epoch": 9.139110604332954,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.0121933260286432e-07,
+      "loss": 0.5539,
+      "mean_token_accuracy": 0.8851373344659805,
+      "num_tokens": 429009258.0,
+      "step": 4012
+    },
+    {
+      "epoch": 9.14139110604333,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.0068950832005487e-07,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8837276846170425,
+      "num_tokens": 429116708.0,
+      "step": 4013
+    },
+    {
+      "epoch": 9.143671607753706,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.0016104583956021e-07,
+      "loss": 0.5865,
+      "mean_token_accuracy": 0.8805951774120331,
+      "num_tokens": 429223799.0,
+      "step": 4014
+    },
+    {
+      "epoch": 9.145952109464082,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.963394546132488e-08,
+      "loss": 0.5714,
+      "mean_token_accuracy": 0.8829237967729568,
+      "num_tokens": 429330842.0,
+      "step": 4015
+    },
+    {
+      "epoch": 9.148232611174459,
+      "grad_norm": 2.625,
+      "learning_rate": 9.91082074845215e-08,
+      "loss": 0.5576,
+      "mean_token_accuracy": 0.8858160525560379,
+      "num_tokens": 429438548.0,
+      "step": 4016
+    },
+    {
+      "epoch": 9.150513112884834,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.85838322075483e-08,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8821554481983185,
+      "num_tokens": 429545699.0,
+      "step": 4017
+    },
+    {
+      "epoch": 9.152793614595211,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.806081992803084e-08,
+      "loss": 0.5884,
+      "mean_token_accuracy": 0.8780340850353241,
+      "num_tokens": 429653689.0,
+      "step": 4018
+    },
+    {
+      "epoch": 9.155074116305586,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.753917094282112e-08,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8829029351472855,
+      "num_tokens": 429759979.0,
+      "step": 4019
+    },
+    {
+      "epoch": 9.157354618015964,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.701888554799643e-08,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.883645236492157,
+      "num_tokens": 429867491.0,
+      "step": 4020
+    },
+    {
+      "epoch": 9.15963511972634,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.649996403886086e-08,
+      "loss": 0.5653,
+      "mean_token_accuracy": 0.8832023441791534,
+      "num_tokens": 429974470.0,
+      "step": 4021
+    },
+    {
+      "epoch": 9.161915621436716,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.598240670994435e-08,
+      "loss": 0.5901,
+      "mean_token_accuracy": 0.8796355575323105,
+      "num_tokens": 430081389.0,
+      "step": 4022
+    },
+    {
+      "epoch": 9.164196123147093,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.546621385500249e-08,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8781373351812363,
+      "num_tokens": 430188309.0,
+      "step": 4023
+    },
+    {
+      "epoch": 9.166476624857468,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.495138576701673e-08,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.883067861199379,
+      "num_tokens": 430295602.0,
+      "step": 4024
+    },
+    {
+      "epoch": 9.168757126567845,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.443792273819252e-08,
+      "loss": 0.5593,
+      "mean_token_accuracy": 0.8858152031898499,
+      "num_tokens": 430402252.0,
+      "step": 4025
+    },
+    {
+      "epoch": 9.17103762827822,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.392582505996256e-08,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8808012902736664,
+      "num_tokens": 430508615.0,
+      "step": 4026
+    },
+    {
+      "epoch": 9.173318129988598,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.341509302298295e-08,
+      "loss": 0.5636,
+      "mean_token_accuracy": 0.8854473978281021,
+      "num_tokens": 430615447.0,
+      "step": 4027
+    },
+    {
+      "epoch": 9.175598631698973,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.290572691713573e-08,
+      "loss": 0.5635,
+      "mean_token_accuracy": 0.885080024600029,
+      "num_tokens": 430722374.0,
+      "step": 4028
+    },
+    {
+      "epoch": 9.17787913340935,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.23977270315271e-08,
+      "loss": 0.5819,
+      "mean_token_accuracy": 0.8824837505817413,
+      "num_tokens": 430829818.0,
+      "step": 4029
+    },
+    {
+      "epoch": 9.180159635119727,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.18910936544884e-08,
+      "loss": 0.548,
+      "mean_token_accuracy": 0.8877168446779251,
+      "num_tokens": 430936968.0,
+      "step": 4030
+    },
+    {
+      "epoch": 9.182440136830103,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.138582707357429e-08,
+      "loss": 0.5572,
+      "mean_token_accuracy": 0.8826006203889847,
+      "num_tokens": 431043889.0,
+      "step": 4031
+    },
+    {
+      "epoch": 9.18472063854048,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.088192757556457e-08,
+      "loss": 0.576,
+      "mean_token_accuracy": 0.8780194073915482,
+      "num_tokens": 431150431.0,
+      "step": 4032
+    },
+    {
+      "epoch": 9.187001140250855,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.037939544646324e-08,
+      "loss": 0.551,
+      "mean_token_accuracy": 0.8834515959024429,
+      "num_tokens": 431257816.0,
+      "step": 4033
+    },
+    {
+      "epoch": 9.189281641961232,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.987823097149739e-08,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8835292160511017,
+      "num_tokens": 431364325.0,
+      "step": 4034
+    },
+    {
+      "epoch": 9.191562143671607,
+      "grad_norm": 3.0,
+      "learning_rate": 8.93784344351184e-08,
+      "loss": 0.5672,
+      "mean_token_accuracy": 0.8845600485801697,
+      "num_tokens": 431471086.0,
+      "step": 4035
+    },
+    {
+      "epoch": 9.193842645381984,
+      "grad_norm": 3.0,
+      "learning_rate": 8.888000612100128e-08,
+      "loss": 0.5694,
+      "mean_token_accuracy": 0.8847863525152206,
+      "num_tokens": 431578776.0,
+      "step": 4036
+    },
+    {
+      "epoch": 9.19612314709236,
+      "grad_norm": 2.875,
+      "learning_rate": 8.838294631204391e-08,
+      "loss": 0.5679,
+      "mean_token_accuracy": 0.8858127593994141,
+      "num_tokens": 431685845.0,
+      "step": 4037
+    },
+    {
+      "epoch": 9.198403648802737,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.788725529036812e-08,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.885520726442337,
+      "num_tokens": 431793922.0,
+      "step": 4038
+    },
+    {
+      "epoch": 9.200684150513112,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.739293333731886e-08,
+      "loss": 0.5813,
+      "mean_token_accuracy": 0.8812271952629089,
+      "num_tokens": 431900233.0,
+      "step": 4039
+    },
+    {
+      "epoch": 9.20296465222349,
+      "grad_norm": 2.75,
+      "learning_rate": 8.689998073346361e-08,
+      "loss": 0.5602,
+      "mean_token_accuracy": 0.8870201855897903,
+      "num_tokens": 432007117.0,
+      "step": 4040
+    },
+    {
+      "epoch": 9.205245153933866,
+      "grad_norm": 3.25,
+      "learning_rate": 8.640839775859222e-08,
+      "loss": 0.5591,
+      "mean_token_accuracy": 0.8841793239116669,
+      "num_tokens": 432114160.0,
+      "step": 4041
+    },
+    {
+      "epoch": 9.207525655644242,
+      "grad_norm": 2.75,
+      "learning_rate": 8.591818469171815e-08,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8814292848110199,
+      "num_tokens": 432221294.0,
+      "step": 4042
+    },
+    {
+      "epoch": 9.209806157354619,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.542934181107687e-08,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8811886161565781,
+      "num_tokens": 432328223.0,
+      "step": 4043
+    },
+    {
+      "epoch": 9.212086659064994,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.494186939412591e-08,
+      "loss": 0.5692,
+      "mean_token_accuracy": 0.8848684132099152,
+      "num_tokens": 432435810.0,
+      "step": 4044
+    },
+    {
+      "epoch": 9.214367160775371,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.44557677175456e-08,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8800568580627441,
+      "num_tokens": 432542846.0,
+      "step": 4045
+    },
+    {
+      "epoch": 9.216647662485746,
+      "grad_norm": 2.5625,
+      "learning_rate": 8.397103705723774e-08,
+      "loss": 0.5613,
+      "mean_token_accuracy": 0.8849961012601852,
+      "num_tokens": 432649987.0,
+      "step": 4046
+    },
+    {
+      "epoch": 9.218928164196123,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.348767768832561e-08,
+      "loss": 0.566,
+      "mean_token_accuracy": 0.8844825327396393,
+      "num_tokens": 432756682.0,
+      "step": 4047
+    },
+    {
+      "epoch": 9.221208665906499,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.300568988515529e-08,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.8798407763242722,
+      "num_tokens": 432863598.0,
+      "step": 4048
+    },
+    {
+      "epoch": 9.223489167616876,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.25250739212935e-08,
+      "loss": 0.5694,
+      "mean_token_accuracy": 0.8815117180347443,
+      "num_tokens": 432970619.0,
+      "step": 4049
+    },
+    {
+      "epoch": 9.225769669327253,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.204583006952843e-08,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.88057541847229,
+      "num_tokens": 433077375.0,
+      "step": 4050
+    },
+    {
+      "epoch": 9.228050171037628,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.156795860187028e-08,
+      "loss": 0.5483,
+      "mean_token_accuracy": 0.8875467032194138,
+      "num_tokens": 433184582.0,
+      "step": 4051
+    },
+    {
+      "epoch": 9.230330672748005,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.109145978954874e-08,
+      "loss": 0.582,
+      "mean_token_accuracy": 0.8819152861833572,
+      "num_tokens": 433291314.0,
+      "step": 4052
+    },
+    {
+      "epoch": 9.23261117445838,
+      "grad_norm": 2.75,
+      "learning_rate": 8.061633390301582e-08,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8808294236660004,
+      "num_tokens": 433398910.0,
+      "step": 4053
+    },
+    {
+      "epoch": 9.234891676168758,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.014258121194385e-08,
+      "loss": 0.5633,
+      "mean_token_accuracy": 0.885577842593193,
+      "num_tokens": 433505751.0,
+      "step": 4054
+    },
+    {
+      "epoch": 9.237172177879133,
+      "grad_norm": 3.984375,
+      "learning_rate": 7.967020198522579e-08,
+      "loss": 0.5536,
+      "mean_token_accuracy": 0.885611966252327,
+      "num_tokens": 433613189.0,
+      "step": 4055
+    },
+    {
+      "epoch": 9.23945267958951,
+      "grad_norm": 2.3125,
+      "learning_rate": 7.91991964909744e-08,
+      "loss": 0.5444,
+      "mean_token_accuracy": 0.8886383771896362,
+      "num_tokens": 433720725.0,
+      "step": 4056
+    },
+    {
+      "epoch": 9.241733181299885,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.872956499652418e-08,
+      "loss": 0.5946,
+      "mean_token_accuracy": 0.8744540363550186,
+      "num_tokens": 433827682.0,
+      "step": 4057
+    },
+    {
+      "epoch": 9.244013683010262,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.826130776842828e-08,
+      "loss": 0.5958,
+      "mean_token_accuracy": 0.8749952912330627,
+      "num_tokens": 433934460.0,
+      "step": 4058
+    },
+    {
+      "epoch": 9.246294184720638,
+      "grad_norm": 3.375,
+      "learning_rate": 7.779442507246021e-08,
+      "loss": 0.5542,
+      "mean_token_accuracy": 0.8850563615560532,
+      "num_tokens": 434041663.0,
+      "step": 4059
+    },
+    {
+      "epoch": 9.248574686431015,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.73289171736144e-08,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.882745549082756,
+      "num_tokens": 434147970.0,
+      "step": 4060
+    },
+    {
+      "epoch": 9.250855188141392,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.686478433610339e-08,
+      "loss": 0.5937,
+      "mean_token_accuracy": 0.8776877820491791,
+      "num_tokens": 434255682.0,
+      "step": 4061
+    },
+    {
+      "epoch": 9.253135689851767,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.64020268233609e-08,
+      "loss": 0.5453,
+      "mean_token_accuracy": 0.8867078274488449,
+      "num_tokens": 434362871.0,
+      "step": 4062
+    },
+    {
+      "epoch": 9.255416191562144,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.594064489803821e-08,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8832486718893051,
+      "num_tokens": 434470335.0,
+      "step": 4063
+    },
+    {
+      "epoch": 9.25769669327252,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.548063882200724e-08,
+      "loss": 0.5691,
+      "mean_token_accuracy": 0.8802042752504349,
+      "num_tokens": 434577136.0,
+      "step": 4064
+    },
+    {
+      "epoch": 9.259977194982897,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.502200885635858e-08,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8827760517597198,
+      "num_tokens": 434684066.0,
+      "step": 4065
+    },
+    {
+      "epoch": 9.262257696693272,
+      "grad_norm": 5.09375,
+      "learning_rate": 7.45647552614015e-08,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.8822989910840988,
+      "num_tokens": 434791074.0,
+      "step": 4066
+    },
+    {
+      "epoch": 9.264538198403649,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.410887829666479e-08,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8809485137462616,
+      "num_tokens": 434897922.0,
+      "step": 4067
+    },
+    {
+      "epoch": 9.266818700114024,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.365437822089482e-08,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.8806920945644379,
+      "num_tokens": 435005339.0,
+      "step": 4068
+    },
+    {
+      "epoch": 9.269099201824401,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.320125529205746e-08,
+      "loss": 0.5583,
+      "mean_token_accuracy": 0.8838251531124115,
+      "num_tokens": 435112853.0,
+      "step": 4069
+    },
+    {
+      "epoch": 9.271379703534778,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.274950976733642e-08,
+      "loss": 0.5781,
+      "mean_token_accuracy": 0.8822825849056244,
+      "num_tokens": 435220066.0,
+      "step": 4070
+    },
+    {
+      "epoch": 9.273660205245154,
+      "grad_norm": 2.875,
+      "learning_rate": 7.22991419031338e-08,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8838555067777634,
+      "num_tokens": 435327142.0,
+      "step": 4071
+    },
+    {
+      "epoch": 9.27594070695553,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.185015195506961e-08,
+      "loss": 0.5666,
+      "mean_token_accuracy": 0.8847849667072296,
+      "num_tokens": 435433996.0,
+      "step": 4072
+    },
+    {
+      "epoch": 9.278221208665906,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.140254017798221e-08,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8813326507806778,
+      "num_tokens": 435541275.0,
+      "step": 4073
+    },
+    {
+      "epoch": 9.280501710376283,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.095630682592669e-08,
+      "loss": 0.5792,
+      "mean_token_accuracy": 0.8786799758672714,
+      "num_tokens": 435648369.0,
+      "step": 4074
+    },
+    {
+      "epoch": 9.282782212086659,
+      "grad_norm": 2.578125,
+      "learning_rate": 7.051145215217715e-08,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8824814110994339,
+      "num_tokens": 435755124.0,
+      "step": 4075
+    },
+    {
+      "epoch": 9.285062713797036,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.006797640922436e-08,
+      "loss": 0.5746,
+      "mean_token_accuracy": 0.8815957754850388,
+      "num_tokens": 435861525.0,
+      "step": 4076
+    },
+    {
+      "epoch": 9.287343215507411,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.962587984877617e-08,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.8820008486509323,
+      "num_tokens": 435968074.0,
+      "step": 4077
+    },
+    {
+      "epoch": 9.289623717217788,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.918516272175879e-08,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8813347816467285,
+      "num_tokens": 436075305.0,
+      "step": 4078
+    },
+    {
+      "epoch": 9.291904218928163,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.874582527831409e-08,
+      "loss": 0.5631,
+      "mean_token_accuracy": 0.8813544809818268,
+      "num_tokens": 436183200.0,
+      "step": 4079
+    },
+    {
+      "epoch": 9.29418472063854,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.830786776780174e-08,
+      "loss": 0.607,
+      "mean_token_accuracy": 0.8745275288820267,
+      "num_tokens": 436289903.0,
+      "step": 4080
+    },
+    {
+      "epoch": 9.296465222348917,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.78712904387982e-08,
+      "loss": 0.5851,
+      "mean_token_accuracy": 0.8781695067882538,
+      "num_tokens": 436396884.0,
+      "step": 4081
+    },
+    {
+      "epoch": 9.298745724059293,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.74360935390958e-08,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.8871065676212311,
+      "num_tokens": 436504033.0,
+      "step": 4082
+    },
+    {
+      "epoch": 9.30102622576967,
+      "grad_norm": 7.09375,
+      "learning_rate": 6.700227731570475e-08,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8803330808877945,
+      "num_tokens": 436610843.0,
+      "step": 4083
+    },
+    {
+      "epoch": 9.303306727480045,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.656984201485001e-08,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8825246691703796,
+      "num_tokens": 436717813.0,
+      "step": 4084
+    },
+    {
+      "epoch": 9.305587229190422,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.613878788197359e-08,
+      "loss": 0.5887,
+      "mean_token_accuracy": 0.8762739151716232,
+      "num_tokens": 436824381.0,
+      "step": 4085
+    },
+    {
+      "epoch": 9.307867730900798,
+      "grad_norm": 2.46875,
+      "learning_rate": 6.570911516173368e-08,
+      "loss": 0.5568,
+      "mean_token_accuracy": 0.8874360471963882,
+      "num_tokens": 436931568.0,
+      "step": 4086
+    },
+    {
+      "epoch": 9.310148232611175,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.528082409800434e-08,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8826966434717178,
+      "num_tokens": 437038163.0,
+      "step": 4087
+    },
+    {
+      "epoch": 9.31242873432155,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.485391493387505e-08,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.883079007267952,
+      "num_tokens": 437145250.0,
+      "step": 4088
+    },
+    {
+      "epoch": 9.314709236031927,
+      "grad_norm": 3.25,
+      "learning_rate": 6.442838791165168e-08,
+      "loss": 0.5927,
+      "mean_token_accuracy": 0.8812314420938492,
+      "num_tokens": 437251824.0,
+      "step": 4089
+    },
+    {
+      "epoch": 9.316989737742304,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.400424327285437e-08,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8754829615354538,
+      "num_tokens": 437358958.0,
+      "step": 4090
+    },
+    {
+      "epoch": 9.31927023945268,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.358148125822e-08,
+      "loss": 0.5529,
+      "mean_token_accuracy": 0.8846077919006348,
+      "num_tokens": 437466352.0,
+      "step": 4091
+    },
+    {
+      "epoch": 9.321550741163056,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.316010210769997e-08,
+      "loss": 0.5987,
+      "mean_token_accuracy": 0.8769486397504807,
+      "num_tokens": 437573435.0,
+      "step": 4092
+    },
+    {
+      "epoch": 9.323831242873432,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.274010606046071e-08,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8839032351970673,
+      "num_tokens": 437680600.0,
+      "step": 4093
+    },
+    {
+      "epoch": 9.326111744583809,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.232149335488463e-08,
+      "loss": 0.572,
+      "mean_token_accuracy": 0.8846787661314011,
+      "num_tokens": 437787269.0,
+      "step": 4094
+    },
+    {
+      "epoch": 9.328392246294184,
+      "grad_norm": 3.84375,
+      "learning_rate": 6.190426422856749e-08,
+      "loss": 0.5708,
+      "mean_token_accuracy": 0.8827256411314011,
+      "num_tokens": 437893943.0,
+      "step": 4095
+    },
+    {
+      "epoch": 9.330672748004561,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.148841891832069e-08,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.8799492716789246,
+      "num_tokens": 438000911.0,
+      "step": 4096
+    },
+    {
+      "epoch": 9.332953249714937,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.107395766016988e-08,
+      "loss": 0.5613,
+      "mean_token_accuracy": 0.8834062963724136,
+      "num_tokens": 438107441.0,
+      "step": 4097
+    },
+    {
+      "epoch": 9.335233751425314,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.066088068935577e-08,
+      "loss": 0.5703,
+      "mean_token_accuracy": 0.8836114853620529,
+      "num_tokens": 438214716.0,
+      "step": 4098
+    },
+    {
+      "epoch": 9.33751425313569,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.024918824033221e-08,
+      "loss": 0.5945,
+      "mean_token_accuracy": 0.878153920173645,
+      "num_tokens": 438320941.0,
+      "step": 4099
+    },
+    {
+      "epoch": 9.339794754846066,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.983888054676867e-08,
+      "loss": 0.5657,
+      "mean_token_accuracy": 0.8838547617197037,
+      "num_tokens": 438428175.0,
+      "step": 4100
+    },
+    {
+      "epoch": 9.342075256556443,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.9429957841546926e-08,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8802156448364258,
+      "num_tokens": 438534915.0,
+      "step": 4101
+    },
+    {
+      "epoch": 9.344355758266818,
+      "grad_norm": 5.0625,
+      "learning_rate": 5.902242035676409e-08,
+      "loss": 0.563,
+      "mean_token_accuracy": 0.8827090561389923,
+      "num_tokens": 438641893.0,
+      "step": 4102
+    },
+    {
+      "epoch": 9.346636259977195,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.8616268323730685e-08,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8797716945409775,
+      "num_tokens": 438749068.0,
+      "step": 4103
+    },
+    {
+      "epoch": 9.34891676168757,
+      "grad_norm": 5.15625,
+      "learning_rate": 5.821150197297038e-08,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.878428652882576,
+      "num_tokens": 438855850.0,
+      "step": 4104
+    },
+    {
+      "epoch": 9.351197263397948,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.780812153422161e-08,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8853023201227188,
+      "num_tokens": 438962709.0,
+      "step": 4105
+    },
+    {
+      "epoch": 9.353477765108323,
+      "grad_norm": 4.53125,
+      "learning_rate": 5.7406127236434016e-08,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.8817013502120972,
+      "num_tokens": 439069382.0,
+      "step": 4106
+    },
+    {
+      "epoch": 9.3557582668187,
+      "grad_norm": 4.15625,
+      "learning_rate": 5.700551930777287e-08,
+      "loss": 0.5525,
+      "mean_token_accuracy": 0.8876905292272568,
+      "num_tokens": 439176711.0,
+      "step": 4107
+    },
+    {
+      "epoch": 9.358038768529076,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.66062979756149e-08,
+      "loss": 0.568,
+      "mean_token_accuracy": 0.8845077753067017,
+      "num_tokens": 439284100.0,
+      "step": 4108
+    },
+    {
+      "epoch": 9.360319270239453,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.620846346655079e-08,
+      "loss": 0.5581,
+      "mean_token_accuracy": 0.8825329095125198,
+      "num_tokens": 439391460.0,
+      "step": 4109
+    },
+    {
+      "epoch": 9.36259977194983,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.5812016006383805e-08,
+      "loss": 0.5723,
+      "mean_token_accuracy": 0.8812670260667801,
+      "num_tokens": 439498912.0,
+      "step": 4110
+    },
+    {
+      "epoch": 9.364880273660205,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.5416955820129515e-08,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8810815811157227,
+      "num_tokens": 439606318.0,
+      "step": 4111
+    },
+    {
+      "epoch": 9.367160775370582,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.50232831320166e-08,
+      "loss": 0.5562,
+      "mean_token_accuracy": 0.8831993341445923,
+      "num_tokens": 439712921.0,
+      "step": 4112
+    },
+    {
+      "epoch": 9.369441277080957,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.463099816548578e-08,
+      "loss": 0.5788,
+      "mean_token_accuracy": 0.8801111429929733,
+      "num_tokens": 439819707.0,
+      "step": 4113
+    },
+    {
+      "epoch": 9.371721778791335,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.424010114319117e-08,
+      "loss": 0.5637,
+      "mean_token_accuracy": 0.8822034150362015,
+      "num_tokens": 439926343.0,
+      "step": 4114
+    },
+    {
+      "epoch": 9.37400228050171,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.385059228699779e-08,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8807101249694824,
+      "num_tokens": 440033050.0,
+      "step": 4115
+    },
+    {
+      "epoch": 9.376282782212087,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.346247181798325e-08,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.8771592527627945,
+      "num_tokens": 440140455.0,
+      "step": 4116
+    },
+    {
+      "epoch": 9.378563283922462,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.307573995643772e-08,
+      "loss": 0.5551,
+      "mean_token_accuracy": 0.8888624608516693,
+      "num_tokens": 440247555.0,
+      "step": 4117
+    },
+    {
+      "epoch": 9.38084378563284,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.2690396921862284e-08,
+      "loss": 0.5652,
+      "mean_token_accuracy": 0.8854438215494156,
+      "num_tokens": 440354832.0,
+      "step": 4118
+    },
+    {
+      "epoch": 9.383124287343216,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.230644293297088e-08,
+      "loss": 0.5617,
+      "mean_token_accuracy": 0.8868270665407181,
+      "num_tokens": 440462224.0,
+      "step": 4119
+    },
+    {
+      "epoch": 9.385404789053592,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.192387820768752e-08,
+      "loss": 0.5614,
+      "mean_token_accuracy": 0.8861428201198578,
+      "num_tokens": 440569106.0,
+      "step": 4120
+    },
+    {
+      "epoch": 9.387685290763969,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.154270296314878e-08,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8851838260889053,
+      "num_tokens": 440676323.0,
+      "step": 4121
+    },
+    {
+      "epoch": 9.389965792474344,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.116291741570301e-08,
+      "loss": 0.5632,
+      "mean_token_accuracy": 0.8838883936405182,
+      "num_tokens": 440784286.0,
+      "step": 4122
+    },
+    {
+      "epoch": 9.392246294184721,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.078452178090831e-08,
+      "loss": 0.561,
+      "mean_token_accuracy": 0.8820521384477615,
+      "num_tokens": 440891324.0,
+      "step": 4123
+    },
+    {
+      "epoch": 9.394526795895096,
+      "grad_norm": 2.5625,
+      "learning_rate": 5.040751627353513e-08,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8817588239908218,
+      "num_tokens": 440998383.0,
+      "step": 4124
+    },
+    {
+      "epoch": 9.396807297605474,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.003190110756451e-08,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8854601830244064,
+      "num_tokens": 441105621.0,
+      "step": 4125
+    },
+    {
+      "epoch": 9.399087799315849,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.965767649618869e-08,
+      "loss": 0.5863,
+      "mean_token_accuracy": 0.8789174407720566,
+      "num_tokens": 441212053.0,
+      "step": 4126
+    },
+    {
+      "epoch": 9.401368301026226,
+      "grad_norm": 2.625,
+      "learning_rate": 4.928484265180972e-08,
+      "loss": 0.5602,
+      "mean_token_accuracy": 0.8848675191402435,
+      "num_tokens": 441318500.0,
+      "step": 4127
+    },
+    {
+      "epoch": 9.403648802736601,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.8913399786041097e-08,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8787374198436737,
+      "num_tokens": 441425061.0,
+      "step": 4128
+    },
+    {
+      "epoch": 9.405929304446978,
+      "grad_norm": 2.875,
+      "learning_rate": 4.854334810970668e-08,
+      "loss": 0.5563,
+      "mean_token_accuracy": 0.8844193369150162,
+      "num_tokens": 441531921.0,
+      "step": 4129
+    },
+    {
+      "epoch": 9.408209806157355,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.817468783284096e-08,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8798136711120605,
+      "num_tokens": 441639273.0,
+      "step": 4130
+    },
+    {
+      "epoch": 9.41049030786773,
+      "grad_norm": 2.625,
+      "learning_rate": 4.7807419164687673e-08,
+      "loss": 0.5543,
+      "mean_token_accuracy": 0.8829266577959061,
+      "num_tokens": 441745920.0,
+      "step": 4131
+    },
+    {
+      "epoch": 9.412770809578108,
+      "grad_norm": 6.03125,
+      "learning_rate": 4.7441542313702293e-08,
+      "loss": 0.5991,
+      "mean_token_accuracy": 0.8739556968212128,
+      "num_tokens": 441852952.0,
+      "step": 4132
+    },
+    {
+      "epoch": 9.415051311288483,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.707705748754898e-08,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.8853324800729752,
+      "num_tokens": 441960257.0,
+      "step": 4133
+    },
+    {
+      "epoch": 9.41733181299886,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.671396489310198e-08,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8818874061107635,
+      "num_tokens": 442067515.0,
+      "step": 4134
+    },
+    {
+      "epoch": 9.419612314709235,
+      "grad_norm": 4.375,
+      "learning_rate": 4.635226473644616e-08,
+      "loss": 0.5711,
+      "mean_token_accuracy": 0.8827540129423141,
+      "num_tokens": 442174146.0,
+      "step": 4135
+    },
+    {
+      "epoch": 9.421892816419613,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.599195722287536e-08,
+      "loss": 0.5512,
+      "mean_token_accuracy": 0.8860030323266983,
+      "num_tokens": 442281613.0,
+      "step": 4136
+    },
+    {
+      "epoch": 9.424173318129988,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.5633042556893493e-08,
+      "loss": 0.5853,
+      "mean_token_accuracy": 0.8759856522083282,
+      "num_tokens": 442388874.0,
+      "step": 4137
+    },
+    {
+      "epoch": 9.426453819840365,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.527552094221288e-08,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.8808791786432266,
+      "num_tokens": 442495992.0,
+      "step": 4138
+    },
+    {
+      "epoch": 9.428734321550742,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.4919392581756204e-08,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8829675912857056,
+      "num_tokens": 442603136.0,
+      "step": 4139
+    },
+    {
+      "epoch": 9.431014823261117,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.456465767765539e-08,
+      "loss": 0.5803,
+      "mean_token_accuracy": 0.8808294236660004,
+      "num_tokens": 442710669.0,
+      "step": 4140
+    },
+    {
+      "epoch": 9.433295324971494,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.421131643125104e-08,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8814819753170013,
+      "num_tokens": 442817853.0,
+      "step": 4141
+    },
+    {
+      "epoch": 9.43557582668187,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.3859369043092183e-08,
+      "loss": 0.558,
+      "mean_token_accuracy": 0.8864561766386032,
+      "num_tokens": 442925142.0,
+      "step": 4142
+    },
+    {
+      "epoch": 9.437856328392247,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.350881571293819e-08,
+      "loss": 0.5963,
+      "mean_token_accuracy": 0.8762660771608353,
+      "num_tokens": 443032724.0,
+      "step": 4143
+    },
+    {
+      "epoch": 9.440136830102622,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.315965663975602e-08,
+      "loss": 0.5754,
+      "mean_token_accuracy": 0.8806591182947159,
+      "num_tokens": 443140265.0,
+      "step": 4144
+    },
+    {
+      "epoch": 9.442417331813,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.281189202172131e-08,
+      "loss": 0.5769,
+      "mean_token_accuracy": 0.8811807930469513,
+      "num_tokens": 443247055.0,
+      "step": 4145
+    },
+    {
+      "epoch": 9.444697833523374,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.246552205621896e-08,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.8876753896474838,
+      "num_tokens": 443353977.0,
+      "step": 4146
+    },
+    {
+      "epoch": 9.446978335233752,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.212054693984169e-08,
+      "loss": 0.5843,
+      "mean_token_accuracy": 0.8812119662761688,
+      "num_tokens": 443460573.0,
+      "step": 4147
+    },
+    {
+      "epoch": 9.449258836944129,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.177696686839094e-08,
+      "loss": 0.593,
+      "mean_token_accuracy": 0.8792661875486374,
+      "num_tokens": 443567499.0,
+      "step": 4148
+    },
+    {
+      "epoch": 9.451539338654504,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.143478203687573e-08,
+      "loss": 0.5428,
+      "mean_token_accuracy": 0.8875003904104233,
+      "num_tokens": 443674500.0,
+      "step": 4149
+    },
+    {
+      "epoch": 9.453819840364881,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.1093992639514026e-08,
+      "loss": 0.5456,
+      "mean_token_accuracy": 0.8857799172401428,
+      "num_tokens": 443781405.0,
+      "step": 4150
+    },
+    {
+      "epoch": 9.456100342075256,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.0754598869730824e-08,
+      "loss": 0.5592,
+      "mean_token_accuracy": 0.8854545056819916,
+      "num_tokens": 443888679.0,
+      "step": 4151
+    },
+    {
+      "epoch": 9.458380843785633,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.041660092015981e-08,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8850967884063721,
+      "num_tokens": 443996366.0,
+      "step": 4152
+    },
+    {
+      "epoch": 9.460661345496009,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.007999898264225e-08,
+      "loss": 0.596,
+      "mean_token_accuracy": 0.8776397556066513,
+      "num_tokens": 444104029.0,
+      "step": 4153
+    },
+    {
+      "epoch": 9.462941847206386,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.9744793248226446e-08,
+      "loss": 0.5659,
+      "mean_token_accuracy": 0.8829027712345123,
+      "num_tokens": 444211032.0,
+      "step": 4154
+    },
+    {
+      "epoch": 9.465222348916761,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.9410983907169076e-08,
+      "loss": 0.5758,
+      "mean_token_accuracy": 0.8814502954483032,
+      "num_tokens": 444317977.0,
+      "step": 4155
+    },
+    {
+      "epoch": 9.467502850627138,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.90785711489336e-08,
+      "loss": 0.594,
+      "mean_token_accuracy": 0.8752471357584,
+      "num_tokens": 444424810.0,
+      "step": 4156
+    },
+    {
+      "epoch": 9.469783352337513,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.874755516219103e-08,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8843945115804672,
+      "num_tokens": 444532283.0,
+      "step": 4157
+    },
+    {
+      "epoch": 9.47206385404789,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.8417936134820255e-08,
+      "loss": 0.5483,
+      "mean_token_accuracy": 0.8853261023759842,
+      "num_tokens": 444639261.0,
+      "step": 4158
+    },
+    {
+      "epoch": 9.474344355758268,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.808971425390606e-08,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8793308734893799,
+      "num_tokens": 444745716.0,
+      "step": 4159
+    },
+    {
+      "epoch": 9.476624857468643,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.7762889705740824e-08,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8815562427043915,
+      "num_tokens": 444852792.0,
+      "step": 4160
+    },
+    {
+      "epoch": 9.47890535917902,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.743746267582421e-08,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.8823586702346802,
+      "num_tokens": 444960631.0,
+      "step": 4161
+    },
+    {
+      "epoch": 9.481185860889395,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.711343334886236e-08,
+      "loss": 0.5929,
+      "mean_token_accuracy": 0.8763528019189835,
+      "num_tokens": 445067005.0,
+      "step": 4162
+    },
+    {
+      "epoch": 9.483466362599772,
+      "grad_norm": 5.28125,
+      "learning_rate": 3.679080190876788e-08,
+      "loss": 0.558,
+      "mean_token_accuracy": 0.8835884630680084,
+      "num_tokens": 445174337.0,
+      "step": 4163
+    },
+    {
+      "epoch": 9.485746864310148,
+      "grad_norm": 2.875,
+      "learning_rate": 3.646956853865985e-08,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8791737705469131,
+      "num_tokens": 445281162.0,
+      "step": 4164
+    },
+    {
+      "epoch": 9.488027366020525,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.614973342086464e-08,
+      "loss": 0.5696,
+      "mean_token_accuracy": 0.8816744834184647,
+      "num_tokens": 445388038.0,
+      "step": 4165
+    },
+    {
+      "epoch": 9.4903078677309,
+      "grad_norm": 3.75,
+      "learning_rate": 3.583129673691427e-08,
+      "loss": 0.5796,
+      "mean_token_accuracy": 0.8827396035194397,
+      "num_tokens": 445494974.0,
+      "step": 4166
+    },
+    {
+      "epoch": 9.492588369441277,
+      "grad_norm": 3.96875,
+      "learning_rate": 3.551425866754693e-08,
+      "loss": 0.5716,
+      "mean_token_accuracy": 0.8809922337532043,
+      "num_tokens": 445602110.0,
+      "step": 4167
+    },
+    {
+      "epoch": 9.494868871151652,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.519861939270786e-08,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8822459131479263,
+      "num_tokens": 445709330.0,
+      "step": 4168
+    },
+    {
+      "epoch": 9.49714937286203,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.4884379091547905e-08,
+      "loss": 0.59,
+      "mean_token_accuracy": 0.8781626224517822,
+      "num_tokens": 445816251.0,
+      "step": 4169
+    },
+    {
+      "epoch": 9.499429874572407,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.457153794242302e-08,
+      "loss": 0.581,
+      "mean_token_accuracy": 0.8822884410619736,
+      "num_tokens": 445923058.0,
+      "step": 4170
+    },
+    {
+      "epoch": 9.501710376282782,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.4260096122896435e-08,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8818854689598083,
+      "num_tokens": 446030612.0,
+      "step": 4171
+    },
+    {
+      "epoch": 9.503990877993159,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.3950053809736204e-08,
+      "loss": 0.5733,
+      "mean_token_accuracy": 0.8821684867143631,
+      "num_tokens": 446137524.0,
+      "step": 4172
+    },
+    {
+      "epoch": 9.506271379703534,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.364141117891656e-08,
+      "loss": 0.5467,
+      "mean_token_accuracy": 0.8861701488494873,
+      "num_tokens": 446245159.0,
+      "step": 4173
+    },
+    {
+      "epoch": 9.508551881413911,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.333416840561709e-08,
+      "loss": 0.5643,
+      "mean_token_accuracy": 0.8825655877590179,
+      "num_tokens": 446351813.0,
+      "step": 4174
+    },
+    {
+      "epoch": 9.510832383124287,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.302832566422276e-08,
+      "loss": 0.5637,
+      "mean_token_accuracy": 0.8839205503463745,
+      "num_tokens": 446459491.0,
+      "step": 4175
+    },
+    {
+      "epoch": 9.513112884834664,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.272388312832414e-08,
+      "loss": 0.5609,
+      "mean_token_accuracy": 0.88397616147995,
+      "num_tokens": 446566325.0,
+      "step": 4176
+    },
+    {
+      "epoch": 9.515393386545039,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.242084097071663e-08,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8808989524841309,
+      "num_tokens": 446672682.0,
+      "step": 4177
+    },
+    {
+      "epoch": 9.517673888255416,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.211919936340152e-08,
+      "loss": 0.5851,
+      "mean_token_accuracy": 0.8773170709609985,
+      "num_tokens": 446779847.0,
+      "step": 4178
+    },
+    {
+      "epoch": 9.519954389965793,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.1818958477584375e-08,
+      "loss": 0.562,
+      "mean_token_accuracy": 0.8823383450508118,
+      "num_tokens": 446886733.0,
+      "step": 4179
+    },
+    {
+      "epoch": 9.522234891676169,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.152011848367664e-08,
+      "loss": 0.5527,
+      "mean_token_accuracy": 0.8837304264307022,
+      "num_tokens": 446994279.0,
+      "step": 4180
+    },
+    {
+      "epoch": 9.522234891676169,
+      "eval_loss": 0.5863840579986572,
+      "eval_mean_token_accuracy": 0.8800425250720615,
+      "eval_num_tokens": 446994279.0,
+      "eval_runtime": 58.647,
+      "eval_samples_per_second": 142.974,
+      "eval_steps_per_second": 4.484,
+      "step": 4180
+    },
+    {
+      "epoch": 9.524515393386546,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.1222679551293486e-08,
+      "loss": 0.5806,
+      "mean_token_accuracy": 0.878794476389885,
+      "num_tokens": 447100998.0,
+      "step": 4181
+    },
+    {
+      "epoch": 9.526795895096921,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.0926641849255976e-08,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8829492926597595,
+      "num_tokens": 447208281.0,
+      "step": 4182
+    },
+    {
+      "epoch": 9.529076396807298,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.063200554558915e-08,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8793532252311707,
+      "num_tokens": 447315119.0,
+      "step": 4183
+    },
+    {
+      "epoch": 9.531356898517673,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.033877080752312e-08,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8861254155635834,
+      "num_tokens": 447422365.0,
+      "step": 4184
+    },
+    {
+      "epoch": 9.53363740022805,
+      "grad_norm": 2.875,
+      "learning_rate": 3.0046937801491983e-08,
+      "loss": 0.5595,
+      "mean_token_accuracy": 0.8834296315908432,
+      "num_tokens": 447529666.0,
+      "step": 4185
+    },
+    {
+      "epoch": 9.535917901938426,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.97565066931349e-08,
+      "loss": 0.5897,
+      "mean_token_accuracy": 0.8798715174198151,
+      "num_tokens": 447636941.0,
+      "step": 4186
+    },
+    {
+      "epoch": 9.538198403648803,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.9467477647294464e-08,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.8795375227928162,
+      "num_tokens": 447744498.0,
+      "step": 4187
+    },
+    {
+      "epoch": 9.54047890535918,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.917985082801833e-08,
+      "loss": 0.5647,
+      "mean_token_accuracy": 0.8826041370630264,
+      "num_tokens": 447852156.0,
+      "step": 4188
+    },
+    {
+      "epoch": 9.542759407069555,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.8893626398557583e-08,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8823677599430084,
+      "num_tokens": 447958991.0,
+      "step": 4189
+    },
+    {
+      "epoch": 9.545039908779932,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.8608804521368382e-08,
+      "loss": 0.5586,
+      "mean_token_accuracy": 0.8871615529060364,
+      "num_tokens": 448066219.0,
+      "step": 4190
+    },
+    {
+      "epoch": 9.547320410490308,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.832538535810947e-08,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8768136501312256,
+      "num_tokens": 448173466.0,
+      "step": 4191
+    },
+    {
+      "epoch": 9.549600912200685,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.804336906964439e-08,
+      "loss": 0.5775,
+      "mean_token_accuracy": 0.8800145536661148,
+      "num_tokens": 448280207.0,
+      "step": 4192
+    },
+    {
+      "epoch": 9.55188141391106,
+      "grad_norm": 4.375,
+      "learning_rate": 2.7762755816039823e-08,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8806869983673096,
+      "num_tokens": 448387602.0,
+      "step": 4193
+    },
+    {
+      "epoch": 9.554161915621437,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.74835457565667e-08,
+      "loss": 0.5452,
+      "mean_token_accuracy": 0.8861921578645706,
+      "num_tokens": 448495042.0,
+      "step": 4194
+    },
+    {
+      "epoch": 9.556442417331812,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.7205739049699365e-08,
+      "loss": 0.5467,
+      "mean_token_accuracy": 0.8862817734479904,
+      "num_tokens": 448603055.0,
+      "step": 4195
+    },
+    {
+      "epoch": 9.55872291904219,
+      "grad_norm": 3.75,
+      "learning_rate": 2.6929335853115302e-08,
+      "loss": 0.5855,
+      "mean_token_accuracy": 0.878138080239296,
+      "num_tokens": 448710105.0,
+      "step": 4196
+    },
+    {
+      "epoch": 9.561003420752566,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.6654336323695963e-08,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8829863220453262,
+      "num_tokens": 448817588.0,
+      "step": 4197
+    },
+    {
+      "epoch": 9.563283922462942,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.63807406175251e-08,
+      "loss": 0.5701,
+      "mean_token_accuracy": 0.8825019598007202,
+      "num_tokens": 448924426.0,
+      "step": 4198
+    },
+    {
+      "epoch": 9.565564424173319,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.6108548889891005e-08,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8802784979343414,
+      "num_tokens": 449032265.0,
+      "step": 4199
+    },
+    {
+      "epoch": 9.567844925883694,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.5837761295284258e-08,
+      "loss": 0.5914,
+      "mean_token_accuracy": 0.878703162074089,
+      "num_tokens": 449139772.0,
+      "step": 4200
+    },
+    {
+      "epoch": 9.570125427594071,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.5568377987398862e-08,
+      "loss": 0.5538,
+      "mean_token_accuracy": 0.8867183774709702,
+      "num_tokens": 449247187.0,
+      "step": 4201
+    },
+    {
+      "epoch": 9.572405929304447,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.5300399119131124e-08,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8858496695756912,
+      "num_tokens": 449354632.0,
+      "step": 4202
+    },
+    {
+      "epoch": 9.574686431014824,
+      "grad_norm": 3.375,
+      "learning_rate": 2.5033824842581046e-08,
+      "loss": 0.5552,
+      "mean_token_accuracy": 0.8847160637378693,
+      "num_tokens": 449461642.0,
+      "step": 4203
+    },
+    {
+      "epoch": 9.576966932725199,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.476865530905065e-08,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8837704807519913,
+      "num_tokens": 449567936.0,
+      "step": 4204
+    },
+    {
+      "epoch": 9.579247434435576,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.4504890669045654e-08,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8814296871423721,
+      "num_tokens": 449674620.0,
+      "step": 4205
+    },
+    {
+      "epoch": 9.581527936145951,
+      "grad_norm": 3.0,
+      "learning_rate": 2.4242531072273255e-08,
+      "loss": 0.5677,
+      "mean_token_accuracy": 0.8843848407268524,
+      "num_tokens": 449782103.0,
+      "step": 4206
+    },
+    {
+      "epoch": 9.583808437856328,
+      "grad_norm": 2.484375,
+      "learning_rate": 2.398157666764378e-08,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.883353129029274,
+      "num_tokens": 449888924.0,
+      "step": 4207
+    },
+    {
+      "epoch": 9.586088939566705,
+      "grad_norm": 2.875,
+      "learning_rate": 2.3722027603270415e-08,
+      "loss": 0.5702,
+      "mean_token_accuracy": 0.8861001282930374,
+      "num_tokens": 449996145.0,
+      "step": 4208
+    },
+    {
+      "epoch": 9.58836944127708,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.3463884026467265e-08,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.8813659995794296,
+      "num_tokens": 450103843.0,
+      "step": 4209
+    },
+    {
+      "epoch": 9.590649942987458,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.320714608375241e-08,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8853833377361298,
+      "num_tokens": 450210552.0,
+      "step": 4210
+    },
+    {
+      "epoch": 9.592930444697833,
+      "grad_norm": 4.96875,
+      "learning_rate": 2.295181392084511e-08,
+      "loss": 0.5994,
+      "mean_token_accuracy": 0.877122089266777,
+      "num_tokens": 450318286.0,
+      "step": 4211
+    },
+    {
+      "epoch": 9.59521094640821,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.269788768266695e-08,
+      "loss": 0.5508,
+      "mean_token_accuracy": 0.8865731358528137,
+      "num_tokens": 450425852.0,
+      "step": 4212
+    },
+    {
+      "epoch": 9.597491448118586,
+      "grad_norm": 3.0,
+      "learning_rate": 2.2445367513341533e-08,
+      "loss": 0.5923,
+      "mean_token_accuracy": 0.8766883164644241,
+      "num_tokens": 450532753.0,
+      "step": 4213
+    },
+    {
+      "epoch": 9.599771949828963,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.21942535561942e-08,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8800808787345886,
+      "num_tokens": 450639888.0,
+      "step": 4214
+    },
+    {
+      "epoch": 9.602052451539338,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.1944545953752894e-08,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8837506324052811,
+      "num_tokens": 450746949.0,
+      "step": 4215
+    },
+    {
+      "epoch": 9.604332953249715,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.1696244847746737e-08,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8841430097818375,
+      "num_tokens": 450853966.0,
+      "step": 4216
+    },
+    {
+      "epoch": 9.60661345496009,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.1449350379106336e-08,
+      "loss": 0.5536,
+      "mean_token_accuracy": 0.8856358528137207,
+      "num_tokens": 450961525.0,
+      "step": 4217
+    },
+    {
+      "epoch": 9.608893956670467,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.1203862687964595e-08,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8838754594326019,
+      "num_tokens": 451068191.0,
+      "step": 4218
+    },
+    {
+      "epoch": 9.611174458380844,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.0959781913655053e-08,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.8790152668952942,
+      "num_tokens": 451175232.0,
+      "step": 4219
+    },
+    {
+      "epoch": 9.61345496009122,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.0717108194713566e-08,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8844882100820541,
+      "num_tokens": 451282602.0,
+      "step": 4220
+    },
+    {
+      "epoch": 9.615735461801597,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.0475841668877172e-08,
+      "loss": 0.5797,
+      "mean_token_accuracy": 0.8819600045681,
+      "num_tokens": 451389683.0,
+      "step": 4221
+    },
+    {
+      "epoch": 9.618015963511972,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.0235982473084115e-08,
+      "loss": 0.5573,
+      "mean_token_accuracy": 0.8830082267522812,
+      "num_tokens": 451496555.0,
+      "step": 4222
+    },
+    {
+      "epoch": 9.62029646522235,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.9997530743473548e-08,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.8831875026226044,
+      "num_tokens": 451603138.0,
+      "step": 4223
+    },
+    {
+      "epoch": 9.622576966932725,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.9760486615386376e-08,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8828965127468109,
+      "num_tokens": 451709899.0,
+      "step": 4224
+    },
+    {
+      "epoch": 9.624857468643102,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.9524850223363868e-08,
+      "loss": 0.5722,
+      "mean_token_accuracy": 0.8803802877664566,
+      "num_tokens": 451817093.0,
+      "step": 4225
+    },
+    {
+      "epoch": 9.627137970353477,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.9290621701149315e-08,
+      "loss": 0.5728,
+      "mean_token_accuracy": 0.883967936038971,
+      "num_tokens": 451924424.0,
+      "step": 4226
+    },
+    {
+      "epoch": 9.629418472063854,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.905780118168582e-08,
+      "loss": 0.5787,
+      "mean_token_accuracy": 0.8782493621110916,
+      "num_tokens": 452031788.0,
+      "step": 4227
+    },
+    {
+      "epoch": 9.631698973774231,
+      "grad_norm": 3.625,
+      "learning_rate": 1.882638879711768e-08,
+      "loss": 0.5915,
+      "mean_token_accuracy": 0.8770202100276947,
+      "num_tokens": 452139174.0,
+      "step": 4228
+    },
+    {
+      "epoch": 9.633979475484606,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.859638467879038e-08,
+      "loss": 0.5595,
+      "mean_token_accuracy": 0.8836124539375305,
+      "num_tokens": 452246865.0,
+      "step": 4229
+    },
+    {
+      "epoch": 9.636259977194984,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.8367788957250054e-08,
+      "loss": 0.56,
+      "mean_token_accuracy": 0.8824215233325958,
+      "num_tokens": 452354042.0,
+      "step": 4230
+    },
+    {
+      "epoch": 9.638540478905359,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.8140601762242916e-08,
+      "loss": 0.5793,
+      "mean_token_accuracy": 0.8815779983997345,
+      "num_tokens": 452461206.0,
+      "step": 4231
+    },
+    {
+      "epoch": 9.640820980615736,
+      "grad_norm": 3.625,
+      "learning_rate": 1.7914823222715817e-08,
+      "loss": 0.5824,
+      "mean_token_accuracy": 0.8769263476133347,
+      "num_tokens": 452567894.0,
+      "step": 4232
+    },
+    {
+      "epoch": 9.643101482326111,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.7690453466816805e-08,
+      "loss": 0.5695,
+      "mean_token_accuracy": 0.8843720555305481,
+      "num_tokens": 452675400.0,
+      "step": 4233
+    },
+    {
+      "epoch": 9.645381984036488,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.7467492621893457e-08,
+      "loss": 0.5697,
+      "mean_token_accuracy": 0.8844276815652847,
+      "num_tokens": 452782870.0,
+      "step": 4234
+    },
+    {
+      "epoch": 9.647662485746864,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.724594081449399e-08,
+      "loss": 0.5705,
+      "mean_token_accuracy": 0.8851383030414581,
+      "num_tokens": 452889771.0,
+      "step": 4235
+    },
+    {
+      "epoch": 9.64994298745724,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.702579817036726e-08,
+      "loss": 0.5686,
+      "mean_token_accuracy": 0.8865519165992737,
+      "num_tokens": 452997568.0,
+      "step": 4236
+    },
+    {
+      "epoch": 9.652223489167618,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.680706481446165e-08,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8826638162136078,
+      "num_tokens": 453104638.0,
+      "step": 4237
+    },
+    {
+      "epoch": 9.654503990877993,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.6589740870926186e-08,
+      "loss": 0.5501,
+      "mean_token_accuracy": 0.8905289322137833,
+      "num_tokens": 453211884.0,
+      "step": 4238
+    },
+    {
+      "epoch": 9.65678449258837,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.6373826463109976e-08,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8786414712667465,
+      "num_tokens": 453318930.0,
+      "step": 4239
+    },
+    {
+      "epoch": 9.659064994298745,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.6159321713561382e-08,
+      "loss": 0.5901,
+      "mean_token_accuracy": 0.8797651678323746,
+      "num_tokens": 453426505.0,
+      "step": 4240
+    },
+    {
+      "epoch": 9.661345496009123,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.5946226744029402e-08,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8815406709909439,
+      "num_tokens": 453533705.0,
+      "step": 4241
+    },
+    {
+      "epoch": 9.663625997719498,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.5734541675462567e-08,
+      "loss": 0.556,
+      "mean_token_accuracy": 0.8848689198493958,
+      "num_tokens": 453640428.0,
+      "step": 4242
+    },
+    {
+      "epoch": 9.665906499429875,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.5524266628009212e-08,
+      "loss": 0.5868,
+      "mean_token_accuracy": 0.8785993754863739,
+      "num_tokens": 453747745.0,
+      "step": 4243
+    },
+    {
+      "epoch": 9.66818700114025,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.5315401721017752e-08,
+      "loss": 0.5512,
+      "mean_token_accuracy": 0.8862568438053131,
+      "num_tokens": 453854721.0,
+      "step": 4244
+    },
+    {
+      "epoch": 9.670467502850627,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.5107947073035312e-08,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8809828609228134,
+      "num_tokens": 453962100.0,
+      "step": 4245
+    },
+    {
+      "epoch": 9.672748004561003,
+      "grad_norm": 4.5,
+      "learning_rate": 1.4901902801809642e-08,
+      "loss": 0.5666,
+      "mean_token_accuracy": 0.8845654428005219,
+      "num_tokens": 454069089.0,
+      "step": 4246
+    },
+    {
+      "epoch": 9.67502850627138,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.4697269024287198e-08,
+      "loss": 0.5652,
+      "mean_token_accuracy": 0.8847362250089645,
+      "num_tokens": 454176352.0,
+      "step": 4247
+    },
+    {
+      "epoch": 9.677309007981757,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.4494045856613959e-08,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8813314586877823,
+      "num_tokens": 454282952.0,
+      "step": 4248
+    },
+    {
+      "epoch": 9.679589509692132,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.4292233414135992e-08,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.8810845017433167,
+      "num_tokens": 454390017.0,
+      "step": 4249
+    },
+    {
+      "epoch": 9.68187001140251,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.4091831811397782e-08,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8813609778881073,
+      "num_tokens": 454496557.0,
+      "step": 4250
+    },
+    {
+      "epoch": 9.684150513112884,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.38928411621439e-08,
+      "loss": 0.576,
+      "mean_token_accuracy": 0.8821087181568146,
+      "num_tokens": 454603755.0,
+      "step": 4251
+    },
+    {
+      "epoch": 9.686431014823262,
+      "grad_norm": 5.5,
+      "learning_rate": 1.3695261579316776e-08,
+      "loss": 0.5867,
+      "mean_token_accuracy": 0.8810128718614578,
+      "num_tokens": 454710416.0,
+      "step": 4252
+    },
+    {
+      "epoch": 9.688711516533637,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.3499093175059208e-08,
+      "loss": 0.5492,
+      "mean_token_accuracy": 0.8873603790998459,
+      "num_tokens": 454817321.0,
+      "step": 4253
+    },
+    {
+      "epoch": 9.690992018244014,
+      "grad_norm": 2.625,
+      "learning_rate": 1.3304336060712685e-08,
+      "loss": 0.5604,
+      "mean_token_accuracy": 0.8854569047689438,
+      "num_tokens": 454924489.0,
+      "step": 4254
+    },
+    {
+      "epoch": 9.69327251995439,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.3110990346817676e-08,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8837967067956924,
+      "num_tokens": 455031171.0,
+      "step": 4255
+    },
+    {
+      "epoch": 9.695553021664766,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.2919056143113062e-08,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.8802383691072464,
+      "num_tokens": 455138285.0,
+      "step": 4256
+    },
+    {
+      "epoch": 9.697833523375142,
+      "grad_norm": 2.625,
+      "learning_rate": 1.2728533558537259e-08,
+      "loss": 0.5509,
+      "mean_token_accuracy": 0.8880133926868439,
+      "num_tokens": 455245413.0,
+      "step": 4257
+    },
+    {
+      "epoch": 9.700114025085519,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.2539422701227099e-08,
+      "loss": 0.5777,
+      "mean_token_accuracy": 0.8805587142705917,
+      "num_tokens": 455352632.0,
+      "step": 4258
+    },
+    {
+      "epoch": 9.702394526795896,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.235172367851839e-08,
+      "loss": 0.5548,
+      "mean_token_accuracy": 0.8849736303091049,
+      "num_tokens": 455460206.0,
+      "step": 4259
+    },
+    {
+      "epoch": 9.704675028506271,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.2165436596945634e-08,
+      "loss": 0.5487,
+      "mean_token_accuracy": 0.8850021660327911,
+      "num_tokens": 455567328.0,
+      "step": 4260
+    },
+    {
+      "epoch": 9.706955530216648,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.19805615622412e-08,
+      "loss": 0.5687,
+      "mean_token_accuracy": 0.8875507861375809,
+      "num_tokens": 455674739.0,
+      "step": 4261
+    },
+    {
+      "epoch": 9.709236031927023,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.179709867933726e-08,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8814339190721512,
+      "num_tokens": 455782280.0,
+      "step": 4262
+    },
+    {
+      "epoch": 9.7115165336374,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.1615048052363298e-08,
+      "loss": 0.5736,
+      "mean_token_accuracy": 0.8808081150054932,
+      "num_tokens": 455888628.0,
+      "step": 4263
+    },
+    {
+      "epoch": 9.713797035347776,
+      "grad_norm": 3.25,
+      "learning_rate": 1.1434409784648049e-08,
+      "loss": 0.5585,
+      "mean_token_accuracy": 0.8849108666181564,
+      "num_tokens": 455996087.0,
+      "step": 4264
+    },
+    {
+      "epoch": 9.716077537058153,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.125518397871811e-08,
+      "loss": 0.5839,
+      "mean_token_accuracy": 0.880156397819519,
+      "num_tokens": 456103096.0,
+      "step": 4265
+    },
+    {
+      "epoch": 9.718358038768528,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.1077370736298498e-08,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8832675367593765,
+      "num_tokens": 456209824.0,
+      "step": 4266
+    },
+    {
+      "epoch": 9.720638540478905,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.090097015831293e-08,
+      "loss": 0.5706,
+      "mean_token_accuracy": 0.8845285028219223,
+      "num_tokens": 456317318.0,
+      "step": 4267
+    },
+    {
+      "epoch": 9.722919042189282,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.0725982344882701e-08,
+      "loss": 0.5922,
+      "mean_token_accuracy": 0.8807027041912079,
+      "num_tokens": 456424411.0,
+      "step": 4268
+    },
+    {
+      "epoch": 9.725199543899658,
+      "grad_norm": 3.0,
+      "learning_rate": 1.0552407395327813e-08,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8796821981668472,
+      "num_tokens": 456531485.0,
+      "step": 4269
+    },
+    {
+      "epoch": 9.727480045610035,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.0380245408165846e-08,
+      "loss": 0.5654,
+      "mean_token_accuracy": 0.8859843015670776,
+      "num_tokens": 456638619.0,
+      "step": 4270
+    },
+    {
+      "epoch": 9.72976054732041,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.0209496481112247e-08,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.8768668621778488,
+      "num_tokens": 456745727.0,
+      "step": 4271
+    },
+    {
+      "epoch": 9.732041049030787,
+      "grad_norm": 3.125,
+      "learning_rate": 1.0040160711081437e-08,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8817434310913086,
+      "num_tokens": 456852873.0,
+      "step": 4272
+    },
+    {
+      "epoch": 9.734321550741162,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.87223819418487e-09,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.8846335262060165,
+      "num_tokens": 456960828.0,
+      "step": 4273
+    },
+    {
+      "epoch": 9.73660205245154,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.705729025732135e-09,
+      "loss": 0.5523,
+      "mean_token_accuracy": 0.8860846012830734,
+      "num_tokens": 457067866.0,
+      "step": 4274
+    },
+    {
+      "epoch": 9.738882554161915,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.540633300230418e-09,
+      "loss": 0.579,
+      "mean_token_accuracy": 0.8812527358531952,
+      "num_tokens": 457174777.0,
+      "step": 4275
+    },
+    {
+      "epoch": 9.741163055872292,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.376951111385313e-09,
+      "loss": 0.5482,
+      "mean_token_accuracy": 0.8858974128961563,
+      "num_tokens": 457282370.0,
+      "step": 4276
+    },
+    {
+      "epoch": 9.743443557582669,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.214682552099175e-09,
+      "loss": 0.5803,
+      "mean_token_accuracy": 0.8799700736999512,
+      "num_tokens": 457389651.0,
+      "step": 4277
+    },
+    {
+      "epoch": 9.745724059293044,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.053827714472773e-09,
+      "loss": 0.5549,
+      "mean_token_accuracy": 0.8852715194225311,
+      "num_tokens": 457496887.0,
+      "step": 4278
+    },
+    {
+      "epoch": 9.748004561003421,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.894386689804469e-09,
+      "loss": 0.5617,
+      "mean_token_accuracy": 0.883760392665863,
+      "num_tokens": 457603521.0,
+      "step": 4279
+    },
+    {
+      "epoch": 9.750285062713797,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.73635956858937e-09,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8828184455633163,
+      "num_tokens": 457710089.0,
+      "step": 4280
+    },
+    {
+      "epoch": 9.752565564424174,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.579746440520731e-09,
+      "loss": 0.5782,
+      "mean_token_accuracy": 0.8836691379547119,
+      "num_tokens": 457817982.0,
+      "step": 4281
+    },
+    {
+      "epoch": 9.754846066134549,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.424547394489668e-09,
+      "loss": 0.5624,
+      "mean_token_accuracy": 0.883584663271904,
+      "num_tokens": 457925306.0,
+      "step": 4282
+    },
+    {
+      "epoch": 9.757126567844926,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.270762518583498e-09,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8828044384717941,
+      "num_tokens": 458032364.0,
+      "step": 4283
+    },
+    {
+      "epoch": 9.759407069555301,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.118391900087952e-09,
+      "loss": 0.5401,
+      "mean_token_accuracy": 0.8894132673740387,
+      "num_tokens": 458139150.0,
+      "step": 4284
+    },
+    {
+      "epoch": 9.761687571265679,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.967435625485242e-09,
+      "loss": 0.5614,
+      "mean_token_accuracy": 0.8839370012283325,
+      "num_tokens": 458246236.0,
+      "step": 4285
+    },
+    {
+      "epoch": 9.763968072976056,
+      "grad_norm": 4.0625,
+      "learning_rate": 7.81789378045572e-09,
+      "loss": 0.555,
+      "mean_token_accuracy": 0.8877554684877396,
+      "num_tokens": 458353277.0,
+      "step": 4286
+    },
+    {
+      "epoch": 9.766248574686431,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.669766449876493e-09,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8832697570323944,
+      "num_tokens": 458459980.0,
+      "step": 4287
+    },
+    {
+      "epoch": 9.768529076396808,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.523053717821138e-09,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8807340711355209,
+      "num_tokens": 458566519.0,
+      "step": 4288
+    },
+    {
+      "epoch": 9.770809578107183,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.377755667561659e-09,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8821778148412704,
+      "num_tokens": 458673827.0,
+      "step": 4289
+    },
+    {
+      "epoch": 9.77309007981756,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.233872381565976e-09,
+      "loss": 0.5793,
+      "mean_token_accuracy": 0.8782579749822617,
+      "num_tokens": 458780746.0,
+      "step": 4290
+    },
+    {
+      "epoch": 9.775370581527936,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.091403941499597e-09,
+      "loss": 0.556,
+      "mean_token_accuracy": 0.885596290230751,
+      "num_tokens": 458887534.0,
+      "step": 4291
+    },
+    {
+      "epoch": 9.777651083238313,
+      "grad_norm": 3.0,
+      "learning_rate": 6.950350428225061e-09,
+      "loss": 0.5504,
+      "mean_token_accuracy": 0.8886802643537521,
+      "num_tokens": 458994895.0,
+      "step": 4292
+    },
+    {
+      "epoch": 9.779931584948688,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.810711921801105e-09,
+      "loss": 0.5644,
+      "mean_token_accuracy": 0.8828777819871902,
+      "num_tokens": 459102180.0,
+      "step": 4293
+    },
+    {
+      "epoch": 9.782212086659065,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.672488501484608e-09,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8822075426578522,
+      "num_tokens": 459209744.0,
+      "step": 4294
+    },
+    {
+      "epoch": 9.78449258836944,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.535680245727816e-09,
+      "loss": 0.556,
+      "mean_token_accuracy": 0.8854386210441589,
+      "num_tokens": 459317224.0,
+      "step": 4295
+    },
+    {
+      "epoch": 9.786773090079818,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.400287232180558e-09,
+      "loss": 0.5752,
+      "mean_token_accuracy": 0.8836027830839157,
+      "num_tokens": 459424913.0,
+      "step": 4296
+    },
+    {
+      "epoch": 9.789053591790195,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.266309537689696e-09,
+      "loss": 0.5751,
+      "mean_token_accuracy": 0.8846608698368073,
+      "num_tokens": 459531883.0,
+      "step": 4297
+    },
+    {
+      "epoch": 9.79133409350057,
+      "grad_norm": 4.5,
+      "learning_rate": 6.133747238298016e-09,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8778994530439377,
+      "num_tokens": 459638404.0,
+      "step": 4298
+    },
+    {
+      "epoch": 9.793614595210947,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.002600409245607e-09,
+      "loss": 0.5522,
+      "mean_token_accuracy": 0.8856344819068909,
+      "num_tokens": 459746677.0,
+      "step": 4299
+    },
+    {
+      "epoch": 9.795895096921322,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.872869124968761e-09,
+      "loss": 0.5939,
+      "mean_token_accuracy": 0.87481589615345,
+      "num_tokens": 459853457.0,
+      "step": 4300
+    },
+    {
+      "epoch": 9.7981755986317,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.7445534591002435e-09,
+      "loss": 0.5763,
+      "mean_token_accuracy": 0.8805107921361923,
+      "num_tokens": 459960464.0,
+      "step": 4301
+    },
+    {
+      "epoch": 9.800456100342075,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.617653484469576e-09,
+      "loss": 0.5489,
+      "mean_token_accuracy": 0.885638564825058,
+      "num_tokens": 460067123.0,
+      "step": 4302
+    },
+    {
+      "epoch": 9.802736602052452,
+      "grad_norm": 6.28125,
+      "learning_rate": 5.492169273103309e-09,
+      "loss": 0.5632,
+      "mean_token_accuracy": 0.884444460272789,
+      "num_tokens": 460173895.0,
+      "step": 4303
+    },
+    {
+      "epoch": 9.805017103762827,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.368100896223083e-09,
+      "loss": 0.5775,
+      "mean_token_accuracy": 0.8823024779558182,
+      "num_tokens": 460280558.0,
+      "step": 4304
+    },
+    {
+      "epoch": 9.807297605473204,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.245448424248123e-09,
+      "loss": 0.5707,
+      "mean_token_accuracy": 0.885224923491478,
+      "num_tokens": 460387236.0,
+      "step": 4305
+    },
+    {
+      "epoch": 9.80957810718358,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.124211926793577e-09,
+      "loss": 0.5659,
+      "mean_token_accuracy": 0.8848960250616074,
+      "num_tokens": 460493654.0,
+      "step": 4306
+    },
+    {
+      "epoch": 9.811858608893957,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.004391472670788e-09,
+      "loss": 0.5544,
+      "mean_token_accuracy": 0.8836629986763,
+      "num_tokens": 460600975.0,
+      "step": 4307
+    },
+    {
+      "epoch": 9.814139110604334,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.885987129887859e-09,
+      "loss": 0.5956,
+      "mean_token_accuracy": 0.8799479305744171,
+      "num_tokens": 460707729.0,
+      "step": 4308
+    },
+    {
+      "epoch": 9.816419612314709,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.768998965648253e-09,
+      "loss": 0.591,
+      "mean_token_accuracy": 0.8796339482069016,
+      "num_tokens": 460814870.0,
+      "step": 4309
+    },
+    {
+      "epoch": 9.818700114025086,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.653427046352743e-09,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.8791462630033493,
+      "num_tokens": 460921892.0,
+      "step": 4310
+    },
+    {
+      "epoch": 9.820980615735461,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.53927143759747e-09,
+      "loss": 0.5772,
+      "mean_token_accuracy": 0.8801241964101791,
+      "num_tokens": 461029101.0,
+      "step": 4311
+    },
+    {
+      "epoch": 9.823261117445838,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.426532204175049e-09,
+      "loss": 0.5596,
+      "mean_token_accuracy": 0.8861949443817139,
+      "num_tokens": 461136681.0,
+      "step": 4312
+    },
+    {
+      "epoch": 9.825541619156214,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.3152094100740175e-09,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8844578862190247,
+      "num_tokens": 461244074.0,
+      "step": 4313
+    },
+    {
+      "epoch": 9.82782212086659,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.205303118479109e-09,
+      "loss": 0.5693,
+      "mean_token_accuracy": 0.8827894330024719,
+      "num_tokens": 461351000.0,
+      "step": 4314
+    },
+    {
+      "epoch": 9.830102622576966,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.096813391770982e-09,
+      "loss": 0.5606,
+      "mean_token_accuracy": 0.8869377076625824,
+      "num_tokens": 461458568.0,
+      "step": 4315
+    },
+    {
+      "epoch": 9.832383124287343,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.989740291526212e-09,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.8829131424427032,
+      "num_tokens": 461565463.0,
+      "step": 4316
+    },
+    {
+      "epoch": 9.83466362599772,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.884083878517575e-09,
+      "loss": 0.5668,
+      "mean_token_accuracy": 0.8803281784057617,
+      "num_tokens": 461672563.0,
+      "step": 4317
+    },
+    {
+      "epoch": 9.836944127708096,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.779844212713213e-09,
+      "loss": 0.5542,
+      "mean_token_accuracy": 0.8861279636621475,
+      "num_tokens": 461779451.0,
+      "step": 4318
+    },
+    {
+      "epoch": 9.839224629418473,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.6770213532782985e-09,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8874040246009827,
+      "num_tokens": 461886457.0,
+      "step": 4319
+    },
+    {
+      "epoch": 9.841505131128848,
+      "grad_norm": 3.5,
+      "learning_rate": 3.5756153585725374e-09,
+      "loss": 0.5524,
+      "mean_token_accuracy": 0.8833809942007065,
+      "num_tokens": 461993587.0,
+      "step": 4320
+    },
+    {
+      "epoch": 9.843785632839225,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.475626286152112e-09,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8821329474449158,
+      "num_tokens": 462100785.0,
+      "step": 4321
+    },
+    {
+      "epoch": 9.8460661345496,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.3770541927691247e-09,
+      "loss": 0.5641,
+      "mean_token_accuracy": 0.8800762295722961,
+      "num_tokens": 462207714.0,
+      "step": 4322
+    },
+    {
+      "epoch": 9.848346636259977,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.2798991343707676e-09,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8831606060266495,
+      "num_tokens": 462314885.0,
+      "step": 4323
+    },
+    {
+      "epoch": 9.850627137970353,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.1841611661007077e-09,
+      "loss": 0.5608,
+      "mean_token_accuracy": 0.8826912045478821,
+      "num_tokens": 462422224.0,
+      "step": 4324
+    },
+    {
+      "epoch": 9.85290763968073,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.089840342297701e-09,
+      "loss": 0.5771,
+      "mean_token_accuracy": 0.8812664598226547,
+      "num_tokens": 462529435.0,
+      "step": 4325
+    },
+    {
+      "epoch": 9.855188141391107,
+      "grad_norm": 3.75,
+      "learning_rate": 2.9969367164969787e-09,
+      "loss": 0.5616,
+      "mean_token_accuracy": 0.8857505470514297,
+      "num_tokens": 462636769.0,
+      "step": 4326
+    },
+    {
+      "epoch": 9.857468643101482,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.905450341428029e-09,
+      "loss": 0.5718,
+      "mean_token_accuracy": 0.8818775713443756,
+      "num_tokens": 462743978.0,
+      "step": 4327
+    },
+    {
+      "epoch": 9.85974914481186,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.8153812690173697e-09,
+      "loss": 0.5767,
+      "mean_token_accuracy": 0.8814905136823654,
+      "num_tokens": 462850538.0,
+      "step": 4328
+    },
+    {
+      "epoch": 9.862029646522235,
+      "grad_norm": 2.625,
+      "learning_rate": 2.726729550386331e-09,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8831851780414581,
+      "num_tokens": 462957437.0,
+      "step": 4329
+    },
+    {
+      "epoch": 9.864310148232612,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.6394952358518854e-09,
+      "loss": 0.5804,
+      "mean_token_accuracy": 0.8794573098421097,
+      "num_tokens": 463063901.0,
+      "step": 4330
+    },
+    {
+      "epoch": 9.866590649942987,
+      "grad_norm": 2.875,
+      "learning_rate": 2.553678374926649e-09,
+      "loss": 0.5617,
+      "mean_token_accuracy": 0.8824156671762466,
+      "num_tokens": 463171455.0,
+      "step": 4331
+    },
+    {
+      "epoch": 9.868871151653364,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.4692790163183268e-09,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8811941742897034,
+      "num_tokens": 463278442.0,
+      "step": 4332
+    },
+    {
+      "epoch": 9.87115165336374,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.3862972079305435e-09,
+      "loss": 0.5559,
+      "mean_token_accuracy": 0.8832034319639206,
+      "num_tokens": 463385401.0,
+      "step": 4333
+    },
+    {
+      "epoch": 9.873432155074116,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.3047329968620137e-09,
+      "loss": 0.5627,
+      "mean_token_accuracy": 0.8862831592559814,
+      "num_tokens": 463492423.0,
+      "step": 4334
+    },
+    {
+      "epoch": 9.875712656784494,
+      "grad_norm": 3.0,
+      "learning_rate": 2.2245864294073715e-09,
+      "loss": 0.5696,
+      "mean_token_accuracy": 0.8842545598745346,
+      "num_tokens": 463599395.0,
+      "step": 4335
+    },
+    {
+      "epoch": 9.877993158494869,
+      "grad_norm": 2.875,
+      "learning_rate": 2.145857551056063e-09,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8817007839679718,
+      "num_tokens": 463707615.0,
+      "step": 4336
+    },
+    {
+      "epoch": 9.880273660205246,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.0685464064928996e-09,
+      "loss": 0.5538,
+      "mean_token_accuracy": 0.8855222314596176,
+      "num_tokens": 463814300.0,
+      "step": 4337
+    },
+    {
+      "epoch": 9.882554161915621,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.992653039598613e-09,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8810195475816727,
+      "num_tokens": 463921566.0,
+      "step": 4338
+    },
+    {
+      "epoch": 9.884834663625998,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.91817749344847e-09,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8760579079389572,
+      "num_tokens": 464028254.0,
+      "step": 4339
+    },
+    {
+      "epoch": 9.887115165336374,
+      "grad_norm": 2.5,
+      "learning_rate": 1.8451198103133783e-09,
+      "loss": 0.5684,
+      "mean_token_accuracy": 0.8798963725566864,
+      "num_tokens": 464135330.0,
+      "step": 4340
+    },
+    {
+      "epoch": 9.88939566704675,
+      "grad_norm": 4.78125,
+      "learning_rate": 1.7734800316596135e-09,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.8792663961648941,
+      "num_tokens": 464242517.0,
+      "step": 4341
+    },
+    {
+      "epoch": 9.891676168757126,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.703258198148261e-09,
+      "loss": 0.5634,
+      "mean_token_accuracy": 0.884123831987381,
+      "num_tokens": 464349209.0,
+      "step": 4342
+    },
+    {
+      "epoch": 9.893956670467503,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.6344543496360499e-09,
+      "loss": 0.5629,
+      "mean_token_accuracy": 0.8826566338539124,
+      "num_tokens": 464456526.0,
+      "step": 4343
+    },
+    {
+      "epoch": 9.896237172177878,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.567068525175075e-09,
+      "loss": 0.5671,
+      "mean_token_accuracy": 0.887358620762825,
+      "num_tokens": 464563578.0,
+      "step": 4344
+    },
+    {
+      "epoch": 9.898517673888255,
+      "grad_norm": 3.625,
+      "learning_rate": 1.5011007630114093e-09,
+      "loss": 0.5738,
+      "mean_token_accuracy": 0.8806861042976379,
+      "num_tokens": 464670481.0,
+      "step": 4345
+    },
+    {
+      "epoch": 9.900798175598633,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.4365511005878796e-09,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8842573165893555,
+      "num_tokens": 464778377.0,
+      "step": 4346
+    },
+    {
+      "epoch": 9.903078677309008,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.3734195745412904e-09,
+      "loss": 0.5739,
+      "mean_token_accuracy": 0.8824024051427841,
+      "num_tokens": 464885701.0,
+      "step": 4347
+    },
+    {
+      "epoch": 9.905359179019385,
+      "grad_norm": 3.625,
+      "learning_rate": 1.3117062207038123e-09,
+      "loss": 0.577,
+      "mean_token_accuracy": 0.880613699555397,
+      "num_tokens": 464992795.0,
+      "step": 4348
+    },
+    {
+      "epoch": 9.90763968072976,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.2514110741029816e-09,
+      "loss": 0.5741,
+      "mean_token_accuracy": 0.8834188729524612,
+      "num_tokens": 465099950.0,
+      "step": 4349
+    },
+    {
+      "epoch": 9.909920182440137,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.1925341689608682e-09,
+      "loss": 0.5755,
+      "mean_token_accuracy": 0.8817101418972015,
+      "num_tokens": 465207553.0,
+      "step": 4350
+    },
+    {
+      "epoch": 9.912200684150513,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.135075538695185e-09,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.8769432902336121,
+      "num_tokens": 465314057.0,
+      "step": 4351
+    },
+    {
+      "epoch": 9.91448118586089,
+      "grad_norm": 3.875,
+      "learning_rate": 1.0790352159179007e-09,
+      "loss": 0.5812,
+      "mean_token_accuracy": 0.8812268972396851,
+      "num_tokens": 465420891.0,
+      "step": 4352
+    },
+    {
+      "epoch": 9.916761687571265,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.024413232436905e-09,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.878851979970932,
+      "num_tokens": 465527864.0,
+      "step": 4353
+    },
+    {
+      "epoch": 9.919042189281642,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.71209619254343e-10,
+      "loss": 0.5747,
+      "mean_token_accuracy": 0.8795945793390274,
+      "num_tokens": 465634366.0,
+      "step": 4354
+    },
+    {
+      "epoch": 9.921322690992017,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.194244065674484e-10,
+      "loss": 0.5829,
+      "mean_token_accuracy": 0.8763747960329056,
+      "num_tokens": 465741337.0,
+      "step": 4355
+    },
+    {
+      "epoch": 9.923603192702394,
+      "grad_norm": 3.0,
+      "learning_rate": 8.690576237688208e-10,
+      "loss": 0.573,
+      "mean_token_accuracy": 0.8846324980258942,
+      "num_tokens": 465848967.0,
+      "step": 4356
+    },
+    {
+      "epoch": 9.925883694412772,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.201092994453153e-10,
+      "loss": 0.5799,
+      "mean_token_accuracy": 0.879865899682045,
+      "num_tokens": 465955628.0,
+      "step": 4357
+    },
+    {
+      "epoch": 9.928164196123147,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.725794613791527e-10,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8813536763191223,
+      "num_tokens": 466062877.0,
+      "step": 4358
+    },
+    {
+      "epoch": 9.930444697833524,
+      "grad_norm": 3.75,
+      "learning_rate": 7.264681365476422e-10,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.8843436688184738,
+      "num_tokens": 466170580.0,
+      "step": 4359
+    },
+    {
+      "epoch": 9.9327251995439,
+      "grad_norm": 3.125,
+      "learning_rate": 6.817753511226266e-10,
+      "loss": 0.5662,
+      "mean_token_accuracy": 0.8817441612482071,
+      "num_tokens": 466277744.0,
+      "step": 4360
+    },
+    {
+      "epoch": 9.935005701254276,
+      "grad_norm": 2.375,
+      "learning_rate": 6.385011304704814e-10,
+      "loss": 0.5443,
+      "mean_token_accuracy": 0.8887474536895752,
+      "num_tokens": 466384776.0,
+      "step": 4361
+    },
+    {
+      "epoch": 9.937286202964652,
+      "grad_norm": 2.875,
+      "learning_rate": 5.96645499152948e-10,
+      "loss": 0.5935,
+      "mean_token_accuracy": 0.8781139552593231,
+      "num_tokens": 466491782.0,
+      "step": 4362
+    },
+    {
+      "epoch": 9.939566704675029,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.562084809268564e-10,
+      "loss": 0.5561,
+      "mean_token_accuracy": 0.8866285234689713,
+      "num_tokens": 466598743.0,
+      "step": 4363
+    },
+    {
+      "epoch": 9.941847206385404,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.171900987430146e-10,
+      "loss": 0.5625,
+      "mean_token_accuracy": 0.8828070014715195,
+      "num_tokens": 466705736.0,
+      "step": 4364
+    },
+    {
+      "epoch": 9.944127708095781,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.795903747475961e-10,
+      "loss": 0.5786,
+      "mean_token_accuracy": 0.8803279101848602,
+      "num_tokens": 466812755.0,
+      "step": 4365
+    },
+    {
+      "epoch": 9.946408209806158,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.434093302815856e-10,
+      "loss": 0.5639,
+      "mean_token_accuracy": 0.8847870975732803,
+      "num_tokens": 466919779.0,
+      "step": 4366
+    },
+    {
+      "epoch": 9.948688711516533,
+      "grad_norm": 2.75,
+      "learning_rate": 4.0864698588077844e-10,
+      "loss": 0.5774,
+      "mean_token_accuracy": 0.883985847234726,
+      "num_tokens": 467026369.0,
+      "step": 4367
+    },
+    {
+      "epoch": 9.95096921322691,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.7530336127550306e-10,
+      "loss": 0.6001,
+      "mean_token_accuracy": 0.8794097006320953,
+      "num_tokens": 467133105.0,
+      "step": 4368
+    },
+    {
+      "epoch": 9.953249714937286,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.4337847539089866e-10,
+      "loss": 0.5515,
+      "mean_token_accuracy": 0.8836051672697067,
+      "num_tokens": 467240349.0,
+      "step": 4369
+    },
+    {
+      "epoch": 9.955530216647663,
+      "grad_norm": 3.125,
+      "learning_rate": 3.1287234634663766e-10,
+      "loss": 0.5822,
+      "mean_token_accuracy": 0.8793405294418335,
+      "num_tokens": 467347282.0,
+      "step": 4370
+    },
+    {
+      "epoch": 9.957810718358038,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.8378499145803593e-10,
+      "loss": 0.5576,
+      "mean_token_accuracy": 0.8886418640613556,
+      "num_tokens": 467454418.0,
+      "step": 4371
+    },
+    {
+      "epoch": 9.960091220068415,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.5611642723410987e-10,
+      "loss": 0.5857,
+      "mean_token_accuracy": 0.8817699551582336,
+      "num_tokens": 467561394.0,
+      "step": 4372
+    },
+    {
+      "epoch": 9.96237172177879,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.2986666937896418e-10,
+      "loss": 0.5689,
+      "mean_token_accuracy": 0.8804190456867218,
+      "num_tokens": 467668306.0,
+      "step": 4373
+    },
+    {
+      "epoch": 9.964652223489168,
+      "grad_norm": 2.875,
+      "learning_rate": 2.050357327917918e-10,
+      "loss": 0.5541,
+      "mean_token_accuracy": 0.8814870566129684,
+      "num_tokens": 467775667.0,
+      "step": 4374
+    },
+    {
+      "epoch": 9.966932725199545,
+      "grad_norm": 2.515625,
+      "learning_rate": 1.816236315657638e-10,
+      "loss": 0.5719,
+      "mean_token_accuracy": 0.8799229115247726,
+      "num_tokens": 467883505.0,
+      "step": 4375
+    },
+    {
+      "epoch": 9.96921322690992,
+      "grad_norm": 2.875,
+      "learning_rate": 1.5963037898913957e-10,
+      "loss": 0.5626,
+      "mean_token_accuracy": 0.8831915408372879,
+      "num_tokens": 467990987.0,
+      "step": 4376
+    },
+    {
+      "epoch": 9.971493728620297,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.3905598754526684e-10,
+      "loss": 0.5844,
+      "mean_token_accuracy": 0.8801503032445908,
+      "num_tokens": 468098922.0,
+      "step": 4377
+    },
+    {
+      "epoch": 9.973774230330672,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.1990046891147133e-10,
+      "loss": 0.5554,
+      "mean_token_accuracy": 0.8852853924036026,
+      "num_tokens": 468206794.0,
+      "step": 4378
+    },
+    {
+      "epoch": 9.97605473204105,
+      "grad_norm": 2.625,
+      "learning_rate": 1.021638339598896e-10,
+      "loss": 0.5712,
+      "mean_token_accuracy": 0.8826959729194641,
+      "num_tokens": 468313770.0,
+      "step": 4379
+    },
+    {
+      "epoch": 9.978335233751425,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.584609275802402e-11,
+      "loss": 0.5628,
+      "mean_token_accuracy": 0.8853756189346313,
+      "num_tokens": 468420750.0,
+      "step": 4380
+    },
+    {
+      "epoch": 9.980615735461802,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.094725456707752e-11,
+      "loss": 0.5398,
+      "mean_token_accuracy": 0.8900392800569534,
+      "num_tokens": 468527796.0,
+      "step": 4381
+    },
+    {
+      "epoch": 9.982896237172177,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.746732784361886e-11,
+      "loss": 0.5661,
+      "mean_token_accuracy": 0.8818590939044952,
+      "num_tokens": 468634501.0,
+      "step": 4382
+    },
+    {
+      "epoch": 9.985176738882554,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.540632023819491e-11,
+      "loss": 0.5673,
+      "mean_token_accuracy": 0.8829274624586105,
+      "num_tokens": 468741744.0,
+      "step": 4383
+    },
+    {
+      "epoch": 9.987457240592931,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.47642385967184e-11,
+      "loss": 0.5709,
+      "mean_token_accuracy": 0.8815714865922928,
+      "num_tokens": 468848526.0,
+      "step": 4384
+    },
+    {
+      "epoch": 9.989737742303307,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.5541088959357697e-11,
+      "loss": 0.5437,
+      "mean_token_accuracy": 0.8854826241731644,
+      "num_tokens": 468955714.0,
+      "step": 4385
+    },
+    {
+      "epoch": 9.992018244013684,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.773687656109191e-11,
+      "loss": 0.5839,
+      "mean_token_accuracy": 0.8776491433382034,
+      "num_tokens": 469062395.0,
+      "step": 4386
+    },
+    {
+      "epoch": 9.994298745724059,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.1351605831433354e-11,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8828050196170807,
+      "num_tokens": 469169221.0,
+      "step": 4387
+    },
+    {
+      "epoch": 9.996579247434436,
+      "grad_norm": 4.84375,
+      "learning_rate": 6.385280394149984e-12,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.8813241869211197,
+      "num_tokens": 469276076.0,
+      "step": 4388
+    },
+    {
+      "epoch": 9.998859749144811,
+      "grad_norm": 4.5625,
+      "learning_rate": 2.8379030686531696e-12,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8789683431386948,
+      "num_tokens": 469383376.0,
+      "step": 4389
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 4.90625,
+      "learning_rate": 7.094758677772539e-13,
+      "loss": 0.578,
+      "mean_token_accuracy": 0.8766284584999084,
+      "num_tokens": 469422320.0,
+      "step": 4390
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 4390,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.252059023052636e+18,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}