diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,15297 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.0,
+  "eval_steps": 500,
+  "global_step": 1908,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0031446540880503146,
+      "grad_norm": 1.9843024015426636,
+      "learning_rate": 4.1666666666666667e-07,
+      "loss": 0.9991,
+      "mean_token_accuracy": 0.7260034680366516,
+      "step": 1
+    },
+    {
+      "epoch": 0.006289308176100629,
+      "grad_norm": 1.9983922243118286,
+      "learning_rate": 8.333333333333333e-07,
+      "loss": 0.9796,
+      "mean_token_accuracy": 0.7287212610244751,
+      "step": 2
+    },
+    {
+      "epoch": 0.009433962264150943,
+      "grad_norm": 2.008734941482544,
+      "learning_rate": 1.25e-06,
+      "loss": 0.9978,
+      "mean_token_accuracy": 0.7293687462806702,
+      "step": 3
+    },
+    {
+      "epoch": 0.012578616352201259,
+      "grad_norm": 1.9662363529205322,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 0.995,
+      "mean_token_accuracy": 0.7301269769668579,
+      "step": 4
+    },
+    {
+      "epoch": 0.015723270440251572,
+      "grad_norm": 1.9593746662139893,
+      "learning_rate": 2.0833333333333334e-06,
+      "loss": 1.0211,
+      "mean_token_accuracy": 0.7250465154647827,
+      "step": 5
+    },
+    {
+      "epoch": 0.018867924528301886,
+      "grad_norm": 1.769201636314392,
+      "learning_rate": 2.5e-06,
+      "loss": 0.9955,
+      "mean_token_accuracy": 0.7267187833786011,
+      "step": 6
+    },
+    {
+      "epoch": 0.0220125786163522,
+      "grad_norm": 1.6486475467681885,
+      "learning_rate": 2.916666666666667e-06,
+      "loss": 0.9717,
+      "mean_token_accuracy": 0.7299320101737976,
+      "step": 7
+    },
+    {
+      "epoch": 0.025157232704402517,
+      "grad_norm": 1.1724086999893188,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.9318,
+      "mean_token_accuracy": 0.7348803877830505,
+      "step": 8
+    },
+    {
+      "epoch": 0.02830188679245283,
+      "grad_norm": 1.0883439779281616,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.9175,
+      "mean_token_accuracy": 0.7360632419586182,
+      "step": 9
+    },
+    {
+      "epoch": 0.031446540880503145,
+      "grad_norm": 1.0831983089447021,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 0.9115,
+      "mean_token_accuracy": 0.7367658019065857,
+      "step": 10
+    },
+    {
+      "epoch": 0.03459119496855346,
+      "grad_norm": 0.9518892765045166,
+      "learning_rate": 4.583333333333333e-06,
+      "loss": 0.891,
+      "mean_token_accuracy": 0.7373512387275696,
+      "step": 11
+    },
+    {
+      "epoch": 0.03773584905660377,
+      "grad_norm": 0.9544086456298828,
+      "learning_rate": 5e-06,
+      "loss": 0.904,
+      "mean_token_accuracy": 0.7343592047691345,
+      "step": 12
+    },
+    {
+      "epoch": 0.040880503144654086,
+      "grad_norm": 0.8863177299499512,
+      "learning_rate": 5.416666666666667e-06,
+      "loss": 0.8712,
+      "mean_token_accuracy": 0.7427428364753723,
+      "step": 13
+    },
+    {
+      "epoch": 0.0440251572327044,
+      "grad_norm": 0.803634524345398,
+      "learning_rate": 5.833333333333334e-06,
+      "loss": 0.8434,
+      "mean_token_accuracy": 0.7462121844291687,
+      "step": 14
+    },
+    {
+      "epoch": 0.04716981132075472,
+      "grad_norm": 0.9318463802337646,
+      "learning_rate": 6.25e-06,
+      "loss": 0.8115,
+      "mean_token_accuracy": 0.7500566244125366,
+      "step": 15
+    },
+    {
+      "epoch": 0.050314465408805034,
+      "grad_norm": 0.888688325881958,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.8099,
+      "mean_token_accuracy": 0.7527286410331726,
+      "step": 16
+    },
+    {
+      "epoch": 0.05345911949685535,
+      "grad_norm": 0.6901258230209351,
+      "learning_rate": 7.083333333333335e-06,
+      "loss": 0.7951,
+      "mean_token_accuracy": 0.7522338032722473,
+      "step": 17
+    },
+    {
+      "epoch": 0.05660377358490566,
+      "grad_norm": 0.5095185041427612,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.7702,
+      "mean_token_accuracy": 0.7567635774612427,
+      "step": 18
+    },
+    {
+      "epoch": 0.059748427672955975,
+      "grad_norm": 0.41788381338119507,
+      "learning_rate": 7.916666666666667e-06,
+      "loss": 0.7482,
+      "mean_token_accuracy": 0.7630857825279236,
+      "step": 19
+    },
+    {
+      "epoch": 0.06289308176100629,
+      "grad_norm": 0.49949783086776733,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.7309,
+      "mean_token_accuracy": 0.7646651864051819,
+      "step": 20
+    },
+    {
+      "epoch": 0.0660377358490566,
+      "grad_norm": 0.4411962032318115,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 0.7133,
+      "mean_token_accuracy": 0.7669786810874939,
+      "step": 21
+    },
+    {
+      "epoch": 0.06918238993710692,
+      "grad_norm": 0.381130188703537,
+      "learning_rate": 9.166666666666666e-06,
+      "loss": 0.7034,
+      "mean_token_accuracy": 0.7698013782501221,
+      "step": 22
+    },
+    {
+      "epoch": 0.07232704402515723,
+      "grad_norm": 0.32305705547332764,
+      "learning_rate": 9.583333333333335e-06,
+      "loss": 0.6969,
+      "mean_token_accuracy": 0.7718903422355652,
+      "step": 23
+    },
+    {
+      "epoch": 0.07547169811320754,
+      "grad_norm": 0.3746313154697418,
+      "learning_rate": 1e-05,
+      "loss": 0.6906,
+      "mean_token_accuracy": 0.7724348306655884,
+      "step": 24
+    },
+    {
+      "epoch": 0.07861635220125786,
+      "grad_norm": 0.378999263048172,
+      "learning_rate": 1.0416666666666668e-05,
+      "loss": 0.6826,
+      "mean_token_accuracy": 0.7719895243644714,
+      "step": 25
+    },
+    {
+      "epoch": 0.08176100628930817,
+      "grad_norm": 0.35218483209609985,
+      "learning_rate": 1.0833333333333334e-05,
+      "loss": 0.675,
+      "mean_token_accuracy": 0.7760418057441711,
+      "step": 26
+    },
+    {
+      "epoch": 0.08490566037735849,
+      "grad_norm": 0.3063074052333832,
+      "learning_rate": 1.125e-05,
+      "loss": 0.6707,
+      "mean_token_accuracy": 0.7769221067428589,
+      "step": 27
+    },
+    {
+      "epoch": 0.0880503144654088,
+      "grad_norm": 0.3150438070297241,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 0.6542,
+      "mean_token_accuracy": 0.7808550000190735,
+      "step": 28
+    },
+    {
+      "epoch": 0.09119496855345911,
+      "grad_norm": 0.2937377393245697,
+      "learning_rate": 1.2083333333333333e-05,
+      "loss": 0.6654,
+      "mean_token_accuracy": 0.7757168412208557,
+      "step": 29
+    },
+    {
+      "epoch": 0.09433962264150944,
+      "grad_norm": 0.2754563093185425,
+      "learning_rate": 1.25e-05,
+      "loss": 0.6553,
+      "mean_token_accuracy": 0.7805117964744568,
+      "step": 30
+    },
+    {
+      "epoch": 0.09748427672955975,
+      "grad_norm": 0.2745518386363983,
+      "learning_rate": 1.2916666666666668e-05,
+      "loss": 0.6383,
+      "mean_token_accuracy": 0.7834630012512207,
+      "step": 31
+    },
+    {
+      "epoch": 0.10062893081761007,
+      "grad_norm": 0.2507205307483673,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.6382,
+      "mean_token_accuracy": 0.7827038764953613,
+      "step": 32
+    },
+    {
+      "epoch": 0.10377358490566038,
+      "grad_norm": 0.26597172021865845,
+      "learning_rate": 1.375e-05,
+      "loss": 0.644,
+      "mean_token_accuracy": 0.7815650105476379,
+      "step": 33
+    },
+    {
+      "epoch": 0.1069182389937107,
+      "grad_norm": 0.2784893810749054,
+      "learning_rate": 1.416666666666667e-05,
+      "loss": 0.6175,
+      "mean_token_accuracy": 0.7878676056861877,
+      "step": 34
+    },
+    {
+      "epoch": 0.11006289308176101,
+      "grad_norm": 0.2109018713235855,
+      "learning_rate": 1.4583333333333333e-05,
+      "loss": 0.6174,
+      "mean_token_accuracy": 0.7876963019371033,
+      "step": 35
+    },
+    {
+      "epoch": 0.11320754716981132,
+      "grad_norm": 0.2487485110759735,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.6112,
+      "mean_token_accuracy": 0.7883297204971313,
+      "step": 36
+    },
+    {
+      "epoch": 0.11635220125786164,
+      "grad_norm": 0.21007870137691498,
+      "learning_rate": 1.5416666666666668e-05,
+      "loss": 0.6243,
+      "mean_token_accuracy": 0.7878017425537109,
+      "step": 37
+    },
+    {
+      "epoch": 0.11949685534591195,
+      "grad_norm": 0.19941003620624542,
+      "learning_rate": 1.5833333333333333e-05,
+      "loss": 0.6225,
+      "mean_token_accuracy": 0.7866714000701904,
+      "step": 38
+    },
+    {
+      "epoch": 0.12264150943396226,
+      "grad_norm": 0.22181154787540436,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 0.6172,
+      "mean_token_accuracy": 0.7870483994483948,
+      "step": 39
+    },
+    {
+      "epoch": 0.12578616352201258,
+      "grad_norm": 0.21963359415531158,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.6145,
+      "mean_token_accuracy": 0.7886669039726257,
+      "step": 40
+    },
+    {
+      "epoch": 0.1289308176100629,
+      "grad_norm": 0.19179226458072662,
+      "learning_rate": 1.7083333333333333e-05,
+      "loss": 0.6039,
+      "mean_token_accuracy": 0.7911710739135742,
+      "step": 41
+    },
+    {
+      "epoch": 0.1320754716981132,
+      "grad_norm": 0.16477134823799133,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 0.6086,
+      "mean_token_accuracy": 0.7912865281105042,
+      "step": 42
+    },
+    {
+      "epoch": 0.13522012578616352,
+      "grad_norm": 0.22841010987758636,
+      "learning_rate": 1.7916666666666667e-05,
+      "loss": 0.6055,
+      "mean_token_accuracy": 0.789507269859314,
+      "step": 43
+    },
+    {
+      "epoch": 0.13836477987421383,
+      "grad_norm": 0.1824677586555481,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 0.6083,
+      "mean_token_accuracy": 0.7901920080184937,
+      "step": 44
+    },
+    {
+      "epoch": 0.14150943396226415,
+      "grad_norm": 0.2477484941482544,
+      "learning_rate": 1.8750000000000002e-05,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.7944508790969849,
+      "step": 45
+    },
+    {
+      "epoch": 0.14465408805031446,
+      "grad_norm": 0.21475744247436523,
+      "learning_rate": 1.916666666666667e-05,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.793271005153656,
+      "step": 46
+    },
+    {
+      "epoch": 0.14779874213836477,
+      "grad_norm": 0.23494695127010345,
+      "learning_rate": 1.9583333333333333e-05,
+      "loss": 0.6035,
+      "mean_token_accuracy": 0.790216326713562,
+      "step": 47
+    },
+    {
+      "epoch": 0.1509433962264151,
+      "grad_norm": 0.19835183024406433,
+      "learning_rate": 2e-05,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.7936276197433472,
+      "step": 48
+    },
+    {
+      "epoch": 0.1540880503144654,
+      "grad_norm": 0.20389200747013092,
+      "learning_rate": 2.0416666666666667e-05,
+      "loss": 0.5883,
+      "mean_token_accuracy": 0.7947663068771362,
+      "step": 49
+    },
+    {
+      "epoch": 0.15723270440251572,
+      "grad_norm": 0.25244179368019104,
+      "learning_rate": 2.0833333333333336e-05,
+      "loss": 0.593,
+      "mean_token_accuracy": 0.7952125072479248,
+      "step": 50
+    },
+    {
+      "epoch": 0.16037735849056603,
+      "grad_norm": 0.18739324808120728,
+      "learning_rate": 2.125e-05,
+      "loss": 0.5892,
+      "mean_token_accuracy": 0.7951387763023376,
+      "step": 51
+    },
+    {
+      "epoch": 0.16352201257861634,
+      "grad_norm": 0.2114628702402115,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 0.5838,
+      "mean_token_accuracy": 0.7957105040550232,
+      "step": 52
+    },
+    {
+      "epoch": 0.16666666666666666,
+      "grad_norm": 0.2226199507713318,
+      "learning_rate": 2.2083333333333336e-05,
+      "loss": 0.5943,
+      "mean_token_accuracy": 0.7944356799125671,
+      "step": 53
+    },
+    {
+      "epoch": 0.16981132075471697,
+      "grad_norm": 0.19234661757946014,
+      "learning_rate": 2.25e-05,
+      "loss": 0.5899,
+      "mean_token_accuracy": 0.7960485219955444,
+      "step": 54
+    },
+    {
+      "epoch": 0.17295597484276728,
+      "grad_norm": 0.25666317343711853,
+      "learning_rate": 2.2916666666666667e-05,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.7957990169525146,
+      "step": 55
+    },
+    {
+      "epoch": 0.1761006289308176,
+      "grad_norm": 0.2783811390399933,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 0.5946,
+      "mean_token_accuracy": 0.7940381169319153,
+      "step": 56
+    },
+    {
+      "epoch": 0.1792452830188679,
+      "grad_norm": 0.20457403361797333,
+      "learning_rate": 2.375e-05,
+      "loss": 0.5816,
+      "mean_token_accuracy": 0.7978816628456116,
+      "step": 57
+    },
+    {
+      "epoch": 0.18238993710691823,
+      "grad_norm": 0.3623892068862915,
+      "learning_rate": 2.4166666666666667e-05,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.7961991429328918,
+      "step": 58
+    },
+    {
+      "epoch": 0.18553459119496854,
+      "grad_norm": 0.24155688285827637,
+      "learning_rate": 2.4583333333333336e-05,
+      "loss": 0.5683,
+      "mean_token_accuracy": 0.7976905703544617,
+      "step": 59
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 0.377756804227829,
+      "learning_rate": 2.5e-05,
+      "loss": 0.5762,
+      "mean_token_accuracy": 0.8002050518989563,
+      "step": 60
+    },
+    {
+      "epoch": 0.1918238993710692,
+      "grad_norm": 0.3166685104370117,
+      "learning_rate": 2.5416666666666667e-05,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.7962762713432312,
+      "step": 61
+    },
+    {
+      "epoch": 0.1949685534591195,
+      "grad_norm": 0.29503804445266724,
+      "learning_rate": 2.5833333333333336e-05,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.7990941405296326,
+      "step": 62
+    },
+    {
+      "epoch": 0.19811320754716982,
+      "grad_norm": 0.3305257558822632,
+      "learning_rate": 2.625e-05,
+      "loss": 0.5846,
+      "mean_token_accuracy": 0.7966292500495911,
+      "step": 63
+    },
+    {
+      "epoch": 0.20125786163522014,
+      "grad_norm": 0.24742531776428223,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 0.5732,
+      "mean_token_accuracy": 0.7980184555053711,
+      "step": 64
+    },
+    {
+      "epoch": 0.20440251572327045,
+      "grad_norm": 0.2809283137321472,
+      "learning_rate": 2.7083333333333335e-05,
+      "loss": 0.5749,
+      "mean_token_accuracy": 0.7999182939529419,
+      "step": 65
+    },
+    {
+      "epoch": 0.20754716981132076,
+      "grad_norm": 0.1994399130344391,
+      "learning_rate": 2.75e-05,
+      "loss": 0.5731,
+      "mean_token_accuracy": 0.7985623478889465,
+      "step": 66
+    },
+    {
+      "epoch": 0.21069182389937108,
+      "grad_norm": 0.28836458921432495,
+      "learning_rate": 2.7916666666666666e-05,
+      "loss": 0.5596,
+      "mean_token_accuracy": 0.8041678071022034,
+      "step": 67
+    },
+    {
+      "epoch": 0.2138364779874214,
+      "grad_norm": 0.25330910086631775,
+      "learning_rate": 2.833333333333334e-05,
+      "loss": 0.5756,
+      "mean_token_accuracy": 0.7999730110168457,
+      "step": 68
+    },
+    {
+      "epoch": 0.2169811320754717,
+      "grad_norm": 0.6575194001197815,
+      "learning_rate": 2.875e-05,
+      "loss": 0.5811,
+      "mean_token_accuracy": 0.7950687408447266,
+      "step": 69
+    },
+    {
+      "epoch": 0.22012578616352202,
+      "grad_norm": 0.318123996257782,
+      "learning_rate": 2.9166666666666666e-05,
+      "loss": 0.5735,
+      "mean_token_accuracy": 0.7979779243469238,
+      "step": 70
+    },
+    {
+      "epoch": 0.22327044025157233,
+      "grad_norm": 0.21200031042099,
+      "learning_rate": 2.958333333333334e-05,
+      "loss": 0.5595,
+      "mean_token_accuracy": 0.800320029258728,
+      "step": 71
+    },
+    {
+      "epoch": 0.22641509433962265,
+      "grad_norm": 0.3333161771297455,
+      "learning_rate": 3.0000000000000004e-05,
+      "loss": 0.5587,
+      "mean_token_accuracy": 0.8015144467353821,
+      "step": 72
+    },
+    {
+      "epoch": 0.22955974842767296,
+      "grad_norm": 0.2528008818626404,
+      "learning_rate": 3.0416666666666666e-05,
+      "loss": 0.5552,
+      "mean_token_accuracy": 0.8024976253509521,
+      "step": 73
+    },
+    {
+      "epoch": 0.23270440251572327,
+      "grad_norm": 0.29139530658721924,
+      "learning_rate": 3.0833333333333335e-05,
+      "loss": 0.5737,
+      "mean_token_accuracy": 0.7972533106803894,
+      "step": 74
+    },
+    {
+      "epoch": 0.2358490566037736,
+      "grad_norm": 0.31739315390586853,
+      "learning_rate": 3.125e-05,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.7987572550773621,
+      "step": 75
+    },
+    {
+      "epoch": 0.2389937106918239,
+      "grad_norm": 0.20459042489528656,
+      "learning_rate": 3.1666666666666666e-05,
+      "loss": 0.5678,
+      "mean_token_accuracy": 0.7998490929603577,
+      "step": 76
+    },
+    {
+      "epoch": 0.24213836477987422,
+      "grad_norm": 0.3210413455963135,
+      "learning_rate": 3.208333333333334e-05,
+      "loss": 0.5679,
+      "mean_token_accuracy": 0.8000871539115906,
+      "step": 77
+    },
+    {
+      "epoch": 0.24528301886792453,
+      "grad_norm": 0.2642245292663574,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8012592792510986,
+      "step": 78
+    },
+    {
+      "epoch": 0.24842767295597484,
+      "grad_norm": 0.3026284873485565,
+      "learning_rate": 3.291666666666667e-05,
+      "loss": 0.5601,
+      "mean_token_accuracy": 0.802608847618103,
+      "step": 79
+    },
+    {
+      "epoch": 0.25157232704402516,
+      "grad_norm": 0.2567735016345978,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.5488,
+      "mean_token_accuracy": 0.8038848638534546,
+      "step": 80
+    },
+    {
+      "epoch": 0.25471698113207547,
+      "grad_norm": 0.26682281494140625,
+      "learning_rate": 3.375e-05,
+      "loss": 0.5601,
+      "mean_token_accuracy": 0.801401674747467,
+      "step": 81
+    },
+    {
+      "epoch": 0.2578616352201258,
+      "grad_norm": 0.30776286125183105,
+      "learning_rate": 3.4166666666666666e-05,
+      "loss": 0.5619,
+      "mean_token_accuracy": 0.802801251411438,
+      "step": 82
+    },
+    {
+      "epoch": 0.2610062893081761,
+      "grad_norm": 0.2427256852388382,
+      "learning_rate": 3.458333333333334e-05,
+      "loss": 0.5563,
+      "mean_token_accuracy": 0.80390864610672,
+      "step": 83
+    },
+    {
+      "epoch": 0.2641509433962264,
+      "grad_norm": 0.2617015838623047,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 0.5626,
+      "mean_token_accuracy": 0.8013002276420593,
+      "step": 84
+    },
+    {
+      "epoch": 0.2672955974842767,
+      "grad_norm": 0.34007972478866577,
+      "learning_rate": 3.541666666666667e-05,
+      "loss": 0.5618,
+      "mean_token_accuracy": 0.8023524880409241,
+      "step": 85
+    },
+    {
+      "epoch": 0.27044025157232704,
+      "grad_norm": 0.22972999513149261,
+      "learning_rate": 3.5833333333333335e-05,
+      "loss": 0.5605,
+      "mean_token_accuracy": 0.8025376796722412,
+      "step": 86
+    },
+    {
+      "epoch": 0.27358490566037735,
+      "grad_norm": 0.37038788199424744,
+      "learning_rate": 3.625e-05,
+      "loss": 0.561,
+      "mean_token_accuracy": 0.8009118437767029,
+      "step": 87
+    },
+    {
+      "epoch": 0.27672955974842767,
+      "grad_norm": 0.23523087799549103,
+      "learning_rate": 3.6666666666666666e-05,
+      "loss": 0.5531,
+      "mean_token_accuracy": 0.8030558228492737,
+      "step": 88
+    },
+    {
+      "epoch": 0.279874213836478,
+      "grad_norm": 0.3976694941520691,
+      "learning_rate": 3.708333333333334e-05,
+      "loss": 0.5502,
+      "mean_token_accuracy": 0.8050774335861206,
+      "step": 89
+    },
+    {
+      "epoch": 0.2830188679245283,
+      "grad_norm": 0.2722809314727783,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 0.5506,
+      "mean_token_accuracy": 0.8029981255531311,
+      "step": 90
+    },
+    {
+      "epoch": 0.2861635220125786,
+      "grad_norm": 0.4583202004432678,
+      "learning_rate": 3.791666666666667e-05,
+      "loss": 0.5542,
+      "mean_token_accuracy": 0.8037380576133728,
+      "step": 91
+    },
+    {
+      "epoch": 0.2893081761006289,
+      "grad_norm": 0.4009253978729248,
+      "learning_rate": 3.833333333333334e-05,
+      "loss": 0.5612,
+      "mean_token_accuracy": 0.7998917698860168,
+      "step": 92
+    },
+    {
+      "epoch": 0.29245283018867924,
+      "grad_norm": 0.320212721824646,
+      "learning_rate": 3.875e-05,
+      "loss": 0.5554,
+      "mean_token_accuracy": 0.8016734719276428,
+      "step": 93
+    },
+    {
+      "epoch": 0.29559748427672955,
+      "grad_norm": 0.3593039810657501,
+      "learning_rate": 3.9166666666666665e-05,
+      "loss": 0.5599,
+      "mean_token_accuracy": 0.8014413714408875,
+      "step": 94
+    },
+    {
+      "epoch": 0.29874213836477986,
+      "grad_norm": 0.2468472719192505,
+      "learning_rate": 3.958333333333334e-05,
+      "loss": 0.5611,
+      "mean_token_accuracy": 0.8008609414100647,
+      "step": 95
+    },
+    {
+      "epoch": 0.3018867924528302,
+      "grad_norm": 0.4157026410102844,
+      "learning_rate": 4e-05,
+      "loss": 0.5603,
+      "mean_token_accuracy": 0.801600456237793,
+      "step": 96
+    },
+    {
+      "epoch": 0.3050314465408805,
+      "grad_norm": 0.2701054513454437,
+      "learning_rate": 3.9999990660718234e-05,
+      "loss": 0.5551,
+      "mean_token_accuracy": 0.8028652667999268,
+      "step": 97
+    },
+    {
+      "epoch": 0.3081761006289308,
+      "grad_norm": 0.38107001781463623,
+      "learning_rate": 3.999996264288261e-05,
+      "loss": 0.5554,
+      "mean_token_accuracy": 0.8040093183517456,
+      "step": 98
+    },
+    {
+      "epoch": 0.3113207547169811,
+      "grad_norm": 0.2702213525772095,
+      "learning_rate": 3.999991594652221e-05,
+      "loss": 0.5479,
+      "mean_token_accuracy": 0.8046407699584961,
+      "step": 99
+    },
+    {
+      "epoch": 0.31446540880503143,
+      "grad_norm": 0.274862676858902,
+      "learning_rate": 3.999985057168549e-05,
+      "loss": 0.5433,
+      "mean_token_accuracy": 0.8047059774398804,
+      "step": 100
+    },
+    {
+      "epoch": 0.31761006289308175,
+      "grad_norm": 0.2879992723464966,
+      "learning_rate": 3.999976651844029e-05,
+      "loss": 0.5565,
+      "mean_token_accuracy": 0.8013796210289001,
+      "step": 101
+    },
+    {
+      "epoch": 0.32075471698113206,
+      "grad_norm": 0.3317185342311859,
+      "learning_rate": 3.9999663786873816e-05,
+      "loss": 0.5556,
+      "mean_token_accuracy": 0.8010111451148987,
+      "step": 102
+    },
+    {
+      "epoch": 0.3238993710691824,
+      "grad_norm": 0.2893899083137512,
+      "learning_rate": 3.99995423770927e-05,
+      "loss": 0.5392,
+      "mean_token_accuracy": 0.8065787553787231,
+      "step": 103
+    },
+    {
+      "epoch": 0.3270440251572327,
+      "grad_norm": 0.27327823638916016,
+      "learning_rate": 3.999940228922291e-05,
+      "loss": 0.5418,
+      "mean_token_accuracy": 0.8071911334991455,
+      "step": 104
+    },
+    {
+      "epoch": 0.330188679245283,
+      "grad_norm": 0.301395058631897,
+      "learning_rate": 3.9999243523409826e-05,
+      "loss": 0.5467,
+      "mean_token_accuracy": 0.8041232228279114,
+      "step": 105
+    },
+    {
+      "epoch": 0.3333333333333333,
+      "grad_norm": 0.2282310575246811,
+      "learning_rate": 3.999906607981819e-05,
+      "loss": 0.5353,
+      "mean_token_accuracy": 0.8070929050445557,
+      "step": 106
+    },
+    {
+      "epoch": 0.33647798742138363,
+      "grad_norm": 0.27638840675354004,
+      "learning_rate": 3.999886995863214e-05,
+      "loss": 0.5448,
+      "mean_token_accuracy": 0.8050588369369507,
+      "step": 107
+    },
+    {
+      "epoch": 0.33962264150943394,
+      "grad_norm": 0.257600873708725,
+      "learning_rate": 3.999865516005519e-05,
+      "loss": 0.5502,
+      "mean_token_accuracy": 0.8030370473861694,
+      "step": 108
+    },
+    {
+      "epoch": 0.34276729559748426,
+      "grad_norm": 0.23493166267871857,
+      "learning_rate": 3.999842168431023e-05,
+      "loss": 0.5461,
+      "mean_token_accuracy": 0.8061801195144653,
+      "step": 109
+    },
+    {
+      "epoch": 0.34591194968553457,
+      "grad_norm": 0.33576369285583496,
+      "learning_rate": 3.9998169531639546e-05,
+      "loss": 0.5515,
+      "mean_token_accuracy": 0.8048064708709717,
+      "step": 110
+    },
+    {
+      "epoch": 0.3490566037735849,
+      "grad_norm": 0.28176188468933105,
+      "learning_rate": 3.999789870230479e-05,
+      "loss": 0.551,
+      "mean_token_accuracy": 0.8034514784812927,
+      "step": 111
+    },
+    {
+      "epoch": 0.3522012578616352,
+      "grad_norm": 0.27955150604248047,
+      "learning_rate": 3.9997609196587005e-05,
+      "loss": 0.5475,
+      "mean_token_accuracy": 0.803336501121521,
+      "step": 112
+    },
+    {
+      "epoch": 0.3553459119496855,
+      "grad_norm": 0.2750742435455322,
+      "learning_rate": 3.999730101478661e-05,
+      "loss": 0.5487,
+      "mean_token_accuracy": 0.8029544353485107,
+      "step": 113
+    },
+    {
+      "epoch": 0.3584905660377358,
+      "grad_norm": 0.27893856167793274,
+      "learning_rate": 3.9996974157223395e-05,
+      "loss": 0.5481,
+      "mean_token_accuracy": 0.8042543530464172,
+      "step": 114
+    },
+    {
+      "epoch": 0.36163522012578614,
+      "grad_norm": 0.3115741014480591,
+      "learning_rate": 3.9996628624236556e-05,
+      "loss": 0.5463,
+      "mean_token_accuracy": 0.8052036762237549,
+      "step": 115
+    },
+    {
+      "epoch": 0.36477987421383645,
+      "grad_norm": 0.21932333707809448,
+      "learning_rate": 3.999626441618464e-05,
+      "loss": 0.5362,
+      "mean_token_accuracy": 0.8065166473388672,
+      "step": 116
+    },
+    {
+      "epoch": 0.36792452830188677,
+      "grad_norm": 0.3026050329208374,
+      "learning_rate": 3.999588153344559e-05,
+      "loss": 0.5307,
+      "mean_token_accuracy": 0.8093136548995972,
+      "step": 117
+    },
+    {
+      "epoch": 0.3710691823899371,
+      "grad_norm": 0.3253287076950073,
+      "learning_rate": 3.9995479976416725e-05,
+      "loss": 0.5503,
+      "mean_token_accuracy": 0.8045260310173035,
+      "step": 118
+    },
+    {
+      "epoch": 0.3742138364779874,
+      "grad_norm": 0.30211326479911804,
+      "learning_rate": 3.999505974551473e-05,
+      "loss": 0.5492,
+      "mean_token_accuracy": 0.8044459223747253,
+      "step": 119
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.2566331923007965,
+      "learning_rate": 3.9994620841175694e-05,
+      "loss": 0.5533,
+      "mean_token_accuracy": 0.8031808733940125,
+      "step": 120
+    },
+    {
+      "epoch": 0.3805031446540881,
+      "grad_norm": 0.32716792821884155,
+      "learning_rate": 3.999416326385505e-05,
+      "loss": 0.5421,
+      "mean_token_accuracy": 0.8031218647956848,
+      "step": 121
+    },
+    {
+      "epoch": 0.3836477987421384,
+      "grad_norm": 0.2625165581703186,
+      "learning_rate": 3.999368701402763e-05,
+      "loss": 0.5465,
+      "mean_token_accuracy": 0.8032469749450684,
+      "step": 122
+    },
+    {
+      "epoch": 0.3867924528301887,
+      "grad_norm": 0.334791362285614,
+      "learning_rate": 3.9993192092187644e-05,
+      "loss": 0.5469,
+      "mean_token_accuracy": 0.8045857548713684,
+      "step": 123
+    },
+    {
+      "epoch": 0.389937106918239,
+      "grad_norm": 0.28799203038215637,
+      "learning_rate": 3.9992678498848664e-05,
+      "loss": 0.544,
+      "mean_token_accuracy": 0.8061292171478271,
+      "step": 124
+    },
+    {
+      "epoch": 0.39308176100628933,
+      "grad_norm": 0.3053019642829895,
+      "learning_rate": 3.9992146234543645e-05,
+      "loss": 0.5359,
+      "mean_token_accuracy": 0.807460606098175,
+      "step": 125
+    },
+    {
+      "epoch": 0.39622641509433965,
+      "grad_norm": 0.26873913407325745,
+      "learning_rate": 3.999159529982493e-05,
+      "loss": 0.5514,
+      "mean_token_accuracy": 0.8021828532218933,
+      "step": 126
+    },
+    {
+      "epoch": 0.39937106918238996,
+      "grad_norm": 0.2870161533355713,
+      "learning_rate": 3.9991025695264205e-05,
+      "loss": 0.539,
+      "mean_token_accuracy": 0.8052948713302612,
+      "step": 127
+    },
+    {
+      "epoch": 0.4025157232704403,
+      "grad_norm": 0.2600720226764679,
+      "learning_rate": 3.9990437421452556e-05,
+      "loss": 0.5475,
+      "mean_token_accuracy": 0.8043171763420105,
+      "step": 128
+    },
+    {
+      "epoch": 0.4056603773584906,
+      "grad_norm": 0.31923291087150574,
+      "learning_rate": 3.9989830479000435e-05,
+      "loss": 0.536,
+      "mean_token_accuracy": 0.8079201579093933,
+      "step": 129
+    },
+    {
+      "epoch": 0.4088050314465409,
+      "grad_norm": 0.2548767924308777,
+      "learning_rate": 3.9989204868537654e-05,
+      "loss": 0.5412,
+      "mean_token_accuracy": 0.8052085041999817,
+      "step": 130
+    },
+    {
+      "epoch": 0.4119496855345912,
+      "grad_norm": 0.2763267755508423,
+      "learning_rate": 3.998856059071342e-05,
+      "loss": 0.5434,
+      "mean_token_accuracy": 0.8055262565612793,
+      "step": 131
+    },
+    {
+      "epoch": 0.41509433962264153,
+      "grad_norm": 0.2880299985408783,
+      "learning_rate": 3.99878976461963e-05,
+      "loss": 0.5349,
+      "mean_token_accuracy": 0.8085739016532898,
+      "step": 132
+    },
+    {
+      "epoch": 0.41823899371069184,
+      "grad_norm": 0.2466239184141159,
+      "learning_rate": 3.998721603567422e-05,
+      "loss": 0.5464,
+      "mean_token_accuracy": 0.8047433495521545,
+      "step": 133
+    },
+    {
+      "epoch": 0.42138364779874216,
+      "grad_norm": 0.3075404763221741,
+      "learning_rate": 3.9986515759854495e-05,
+      "loss": 0.5366,
+      "mean_token_accuracy": 0.8063505291938782,
+      "step": 134
+    },
+    {
+      "epoch": 0.42452830188679247,
+      "grad_norm": 0.2623771131038666,
+      "learning_rate": 3.9985796819463795e-05,
+      "loss": 0.5354,
+      "mean_token_accuracy": 0.8071600198745728,
+      "step": 135
+    },
+    {
+      "epoch": 0.4276729559748428,
+      "grad_norm": 0.2850979268550873,
+      "learning_rate": 3.998505921524816e-05,
+      "loss": 0.5383,
+      "mean_token_accuracy": 0.8051438331604004,
+      "step": 136
+    },
+    {
+      "epoch": 0.4308176100628931,
+      "grad_norm": 0.21707168221473694,
+      "learning_rate": 3.9984302947973005e-05,
+      "loss": 0.5326,
+      "mean_token_accuracy": 0.8092407584190369,
+      "step": 137
+    },
+    {
+      "epoch": 0.4339622641509434,
+      "grad_norm": 0.2864557206630707,
+      "learning_rate": 3.9983528018423106e-05,
+      "loss": 0.5376,
+      "mean_token_accuracy": 0.8068653345108032,
+      "step": 138
+    },
+    {
+      "epoch": 0.4371069182389937,
+      "grad_norm": 0.227940633893013,
+      "learning_rate": 3.998273442740261e-05,
+      "loss": 0.5355,
+      "mean_token_accuracy": 0.8075131177902222,
+      "step": 139
+    },
+    {
+      "epoch": 0.44025157232704404,
+      "grad_norm": 0.2534600496292114,
+      "learning_rate": 3.9981922175735014e-05,
+      "loss": 0.536,
+      "mean_token_accuracy": 0.8077684640884399,
+      "step": 140
+    },
+    {
+      "epoch": 0.44339622641509435,
+      "grad_norm": 0.3271535038948059,
+      "learning_rate": 3.9981091264263205e-05,
+      "loss": 0.5227,
+      "mean_token_accuracy": 0.8108183145523071,
+      "step": 141
+    },
+    {
+      "epoch": 0.44654088050314467,
+      "grad_norm": 0.2163369208574295,
+      "learning_rate": 3.998024169384941e-05,
+      "loss": 0.5402,
+      "mean_token_accuracy": 0.8060819506645203,
+      "step": 142
+    },
+    {
+      "epoch": 0.449685534591195,
+      "grad_norm": 0.4553479850292206,
+      "learning_rate": 3.997937346537522e-05,
+      "loss": 0.5296,
+      "mean_token_accuracy": 0.8097839951515198,
+      "step": 143
+    },
+    {
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.31248709559440613,
+      "learning_rate": 3.9978486579741596e-05,
+      "loss": 0.5374,
+      "mean_token_accuracy": 0.8072243928909302,
+      "step": 144
+    },
+    {
+      "epoch": 0.4559748427672956,
+      "grad_norm": 0.5377129316329956,
+      "learning_rate": 3.9977581037868874e-05,
+      "loss": 0.5358,
+      "mean_token_accuracy": 0.8072550296783447,
+      "step": 145
+    },
+    {
+      "epoch": 0.4591194968553459,
+      "grad_norm": 0.443276584148407,
+      "learning_rate": 3.997665684069671e-05,
+      "loss": 0.5346,
+      "mean_token_accuracy": 0.8065415024757385,
+      "step": 146
+    },
+    {
+      "epoch": 0.46226415094339623,
+      "grad_norm": 0.4334987699985504,
+      "learning_rate": 3.997571398918415e-05,
+      "loss": 0.5322,
+      "mean_token_accuracy": 0.8077476620674133,
+      "step": 147
+    },
+    {
+      "epoch": 0.46540880503144655,
+      "grad_norm": 0.4879336655139923,
+      "learning_rate": 3.99747524843096e-05,
+      "loss": 0.5402,
+      "mean_token_accuracy": 0.8054540157318115,
+      "step": 148
+    },
+    {
+      "epoch": 0.46855345911949686,
+      "grad_norm": 0.3232817053794861,
+      "learning_rate": 3.9973772327070805e-05,
+      "loss": 0.5281,
+      "mean_token_accuracy": 0.8070737719535828,
+      "step": 149
+    },
+    {
+      "epoch": 0.4716981132075472,
+      "grad_norm": 0.373993843793869,
+      "learning_rate": 3.997277351848486e-05,
+      "loss": 0.542,
+      "mean_token_accuracy": 0.8036733865737915,
+      "step": 150
+    },
+    {
+      "epoch": 0.4748427672955975,
+      "grad_norm": 0.2840160131454468,
+      "learning_rate": 3.997175605958825e-05,
+      "loss": 0.525,
+      "mean_token_accuracy": 0.8079817891120911,
+      "step": 151
+    },
+    {
+      "epoch": 0.4779874213836478,
+      "grad_norm": 0.34674006700515747,
+      "learning_rate": 3.997071995143676e-05,
+      "loss": 0.5432,
+      "mean_token_accuracy": 0.8057137727737427,
+      "step": 152
+    },
+    {
+      "epoch": 0.4811320754716981,
+      "grad_norm": 0.2753613591194153,
+      "learning_rate": 3.9969665195105585e-05,
+      "loss": 0.5401,
+      "mean_token_accuracy": 0.8051686882972717,
+      "step": 153
+    },
+    {
+      "epoch": 0.48427672955974843,
+      "grad_norm": 0.338226854801178,
+      "learning_rate": 3.996859179168923e-05,
+      "loss": 0.5343,
+      "mean_token_accuracy": 0.8090754747390747,
+      "step": 154
+    },
+    {
+      "epoch": 0.48742138364779874,
+      "grad_norm": 0.32556068897247314,
+      "learning_rate": 3.996749974230157e-05,
+      "loss": 0.5375,
+      "mean_token_accuracy": 0.8059588670730591,
+      "step": 155
+    },
+    {
+      "epoch": 0.49056603773584906,
+      "grad_norm": 0.2867760956287384,
+      "learning_rate": 3.9966389048075815e-05,
+      "loss": 0.5243,
+      "mean_token_accuracy": 0.8106968998908997,
+      "step": 156
+    },
+    {
+      "epoch": 0.4937106918238994,
+      "grad_norm": 0.33583030104637146,
+      "learning_rate": 3.996525971016453e-05,
+      "loss": 0.5333,
+      "mean_token_accuracy": 0.8073939681053162,
+      "step": 157
+    },
+    {
+      "epoch": 0.4968553459119497,
+      "grad_norm": 0.25750523805618286,
+      "learning_rate": 3.9964111729739636e-05,
+      "loss": 0.5366,
+      "mean_token_accuracy": 0.8069300651550293,
+      "step": 158
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.3350702226161957,
+      "learning_rate": 3.9962945107992384e-05,
+      "loss": 0.5297,
+      "mean_token_accuracy": 0.8082327246665955,
+      "step": 159
+    },
+    {
+      "epoch": 0.5031446540880503,
+      "grad_norm": 0.25913313031196594,
+      "learning_rate": 3.996175984613337e-05,
+      "loss": 0.5346,
+      "mean_token_accuracy": 0.8069829940795898,
+      "step": 160
+    },
+    {
+      "epoch": 0.5062893081761006,
+      "grad_norm": 0.26537439227104187,
+      "learning_rate": 3.996055594539255e-05,
+      "loss": 0.5227,
+      "mean_token_accuracy": 0.8109566569328308,
+      "step": 161
+    },
+    {
+      "epoch": 0.5094339622641509,
+      "grad_norm": 0.29593321681022644,
+      "learning_rate": 3.99593334070192e-05,
+      "loss": 0.523,
+      "mean_token_accuracy": 0.8093637228012085,
+      "step": 162
+    },
+    {
+      "epoch": 0.5125786163522013,
+      "grad_norm": 0.2575869858264923,
+      "learning_rate": 3.995809223228195e-05,
+      "loss": 0.5406,
+      "mean_token_accuracy": 0.8029673099517822,
+      "step": 163
+    },
+    {
+      "epoch": 0.5157232704402516,
+      "grad_norm": 0.301028847694397,
+      "learning_rate": 3.995683242246876e-05,
+      "loss": 0.5331,
+      "mean_token_accuracy": 0.8078096508979797,
+      "step": 164
+    },
+    {
+      "epoch": 0.5188679245283019,
+      "grad_norm": 0.29553452134132385,
+      "learning_rate": 3.995555397888693e-05,
+      "loss": 0.5213,
+      "mean_token_accuracy": 0.8104479312896729,
+      "step": 165
+    },
+    {
+      "epoch": 0.5220125786163522,
+      "grad_norm": 0.30347883701324463,
+      "learning_rate": 3.995425690286311e-05,
+      "loss": 0.5386,
+      "mean_token_accuracy": 0.8064728379249573,
+      "step": 166
+    },
+    {
+      "epoch": 0.5251572327044025,
+      "grad_norm": 0.3146721422672272,
+      "learning_rate": 3.995294119574326e-05,
+      "loss": 0.5259,
+      "mean_token_accuracy": 0.8080021739006042,
+      "step": 167
+    },
+    {
+      "epoch": 0.5283018867924528,
+      "grad_norm": 0.26118120551109314,
+      "learning_rate": 3.99516068588927e-05,
+      "loss": 0.5187,
+      "mean_token_accuracy": 0.812569260597229,
+      "step": 168
+    },
+    {
+      "epoch": 0.5314465408805031,
+      "grad_norm": 0.3231758177280426,
+      "learning_rate": 3.995025389369606e-05,
+      "loss": 0.5239,
+      "mean_token_accuracy": 0.8090735673904419,
+      "step": 169
+    },
+    {
+      "epoch": 0.5345911949685535,
+      "grad_norm": 0.21853643655776978,
+      "learning_rate": 3.9948882301557306e-05,
+      "loss": 0.5162,
+      "mean_token_accuracy": 0.8130766153335571,
+      "step": 170
+    },
+    {
+      "epoch": 0.5377358490566038,
+      "grad_norm": 0.2971287667751312,
+      "learning_rate": 3.9947492083899746e-05,
+      "loss": 0.5249,
+      "mean_token_accuracy": 0.8092015385627747,
+      "step": 171
+    },
+    {
+      "epoch": 0.5408805031446541,
+      "grad_norm": 0.2766264081001282,
+      "learning_rate": 3.9946083242166e-05,
+      "loss": 0.5202,
+      "mean_token_accuracy": 0.8101739287376404,
+      "step": 172
+    },
+    {
+      "epoch": 0.5440251572327044,
+      "grad_norm": 0.22849610447883606,
+      "learning_rate": 3.9944655777818024e-05,
+      "loss": 0.5173,
+      "mean_token_accuracy": 0.8131073117256165,
+      "step": 173
+    },
+    {
+      "epoch": 0.5471698113207547,
+      "grad_norm": 0.3846280872821808,
+      "learning_rate": 3.994320969233709e-05,
+      "loss": 0.5295,
+      "mean_token_accuracy": 0.807365357875824,
+      "step": 174
+    },
+    {
+      "epoch": 0.550314465408805,
+      "grad_norm": 0.2353799194097519,
+      "learning_rate": 3.9941744987223796e-05,
+      "loss": 0.5241,
+      "mean_token_accuracy": 0.810299813747406,
+      "step": 175
+    },
+    {
+      "epoch": 0.5534591194968553,
+      "grad_norm": 0.36173367500305176,
+      "learning_rate": 3.994026166399808e-05,
+      "loss": 0.531,
+      "mean_token_accuracy": 0.8072293996810913,
+      "step": 176
+    },
+    {
+      "epoch": 0.5566037735849056,
+      "grad_norm": 0.25161418318748474,
+      "learning_rate": 3.993875972419916e-05,
+      "loss": 0.5245,
+      "mean_token_accuracy": 0.8084467053413391,
+      "step": 177
+    },
+    {
+      "epoch": 0.559748427672956,
+      "grad_norm": 0.411323219537735,
+      "learning_rate": 3.993723916938561e-05,
+      "loss": 0.5297,
+      "mean_token_accuracy": 0.8063352704048157,
+      "step": 178
+    },
+    {
+      "epoch": 0.5628930817610063,
+      "grad_norm": 0.28566470742225647,
+      "learning_rate": 3.9935700001135307e-05,
+      "loss": 0.5246,
+      "mean_token_accuracy": 0.8076804280281067,
+      "step": 179
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.3744773864746094,
+      "learning_rate": 3.9934142221045434e-05,
+      "loss": 0.5219,
+      "mean_token_accuracy": 0.8113803863525391,
+      "step": 180
+    },
+    {
+      "epoch": 0.5691823899371069,
+      "grad_norm": 0.3109516501426697,
+      "learning_rate": 3.9932565830732505e-05,
+      "loss": 0.5152,
+      "mean_token_accuracy": 0.8134464025497437,
+      "step": 181
+    },
+    {
+      "epoch": 0.5723270440251572,
+      "grad_norm": 0.40750232338905334,
+      "learning_rate": 3.993097083183233e-05,
+      "loss": 0.5119,
+      "mean_token_accuracy": 0.8139127492904663,
+      "step": 182
+    },
+    {
+      "epoch": 0.5754716981132075,
+      "grad_norm": 0.3474627733230591,
+      "learning_rate": 3.9929357226000045e-05,
+      "loss": 0.538,
+      "mean_token_accuracy": 0.8060604333877563,
+      "step": 183
+    },
+    {
+      "epoch": 0.5786163522012578,
+      "grad_norm": 0.39326900243759155,
+      "learning_rate": 3.992772501491008e-05,
+      "loss": 0.5209,
+      "mean_token_accuracy": 0.8119919300079346,
+      "step": 184
+    },
+    {
+      "epoch": 0.5817610062893082,
+      "grad_norm": 0.390663743019104,
+      "learning_rate": 3.992607420025618e-05,
+      "loss": 0.5165,
+      "mean_token_accuracy": 0.8117203712463379,
+      "step": 185
+    },
+    {
+      "epoch": 0.5849056603773585,
+      "grad_norm": 0.2819702923297882,
+      "learning_rate": 3.9924404783751385e-05,
+      "loss": 0.5287,
+      "mean_token_accuracy": 0.8078507781028748,
+      "step": 186
+    },
+    {
+      "epoch": 0.5880503144654088,
+      "grad_norm": 0.3405737280845642,
+      "learning_rate": 3.992271676712805e-05,
+      "loss": 0.5264,
+      "mean_token_accuracy": 0.8081623315811157,
+      "step": 187
+    },
+    {
+      "epoch": 0.5911949685534591,
+      "grad_norm": 0.24368081986904144,
+      "learning_rate": 3.9921010152137824e-05,
+      "loss": 0.5298,
+      "mean_token_accuracy": 0.8077815771102905,
+      "step": 188
+    },
+    {
+      "epoch": 0.5943396226415094,
+      "grad_norm": 0.2949293553829193,
+      "learning_rate": 3.991928494055166e-05,
+      "loss": 0.5169,
+      "mean_token_accuracy": 0.8117429614067078,
+      "step": 189
+    },
+    {
+      "epoch": 0.5974842767295597,
+      "grad_norm": 0.2298242747783661,
+      "learning_rate": 3.991754113415981e-05,
+      "loss": 0.5209,
+      "mean_token_accuracy": 0.8099380731582642,
+      "step": 190
+    },
+    {
+      "epoch": 0.60062893081761,
+      "grad_norm": 0.2921850383281708,
+      "learning_rate": 3.9915778734771816e-05,
+      "loss": 0.5249,
+      "mean_token_accuracy": 0.8096524477005005,
+      "step": 191
+    },
+    {
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.19392161071300507,
+      "learning_rate": 3.991399774421651e-05,
+      "loss": 0.5114,
+      "mean_token_accuracy": 0.8131522536277771,
+      "step": 192
+    },
+    {
+      "epoch": 0.6069182389937107,
+      "grad_norm": 0.3399420976638794,
+      "learning_rate": 3.991219816434204e-05,
+      "loss": 0.5156,
+      "mean_token_accuracy": 0.8106433749198914,
+      "step": 193
+    },
+    {
+      "epoch": 0.610062893081761,
+      "grad_norm": 0.2603597939014435,
+      "learning_rate": 3.99103799970158e-05,
+      "loss": 0.5152,
+      "mean_token_accuracy": 0.8112037777900696,
+      "step": 194
+    },
+    {
+      "epoch": 0.6132075471698113,
+      "grad_norm": 0.2951433062553406,
+      "learning_rate": 3.990854324412453e-05,
+      "loss": 0.5243,
+      "mean_token_accuracy": 0.8115384578704834,
+      "step": 195
+    },
+    {
+      "epoch": 0.6163522012578616,
+      "grad_norm": 0.2732630968093872,
+      "learning_rate": 3.9906687907574186e-05,
+      "loss": 0.525,
+      "mean_token_accuracy": 0.8100634813308716,
+      "step": 196
+    },
+    {
+      "epoch": 0.6194968553459119,
+      "grad_norm": 0.24679310619831085,
+      "learning_rate": 3.9904813989290084e-05,
+      "loss": 0.5178,
+      "mean_token_accuracy": 0.8117086291313171,
+      "step": 197
+    },
+    {
+      "epoch": 0.6226415094339622,
+      "grad_norm": 0.2945667505264282,
+      "learning_rate": 3.990292149121675e-05,
+      "loss": 0.5088,
+      "mean_token_accuracy": 0.8141837120056152,
+      "step": 198
+    },
+    {
+      "epoch": 0.6257861635220126,
+      "grad_norm": 0.24672532081604004,
+      "learning_rate": 3.9901010415318066e-05,
+      "loss": 0.5146,
+      "mean_token_accuracy": 0.8105906844139099,
+      "step": 199
+    },
+    {
+      "epoch": 0.6289308176100629,
+      "grad_norm": 0.32529643177986145,
+      "learning_rate": 3.9899080763577126e-05,
+      "loss": 0.5253,
+      "mean_token_accuracy": 0.8105737566947937,
+      "step": 200
+    },
+    {
+      "epoch": 0.6320754716981132,
+      "grad_norm": 0.2720421552658081,
+      "learning_rate": 3.9897132537996326e-05,
+      "loss": 0.5182,
+      "mean_token_accuracy": 0.8114479780197144,
+      "step": 201
+    },
+    {
+      "epoch": 0.6352201257861635,
+      "grad_norm": 0.2893914580345154,
+      "learning_rate": 3.9895165740597336e-05,
+      "loss": 0.5259,
+      "mean_token_accuracy": 0.8097965121269226,
+      "step": 202
+    },
+    {
+      "epoch": 0.6383647798742138,
+      "grad_norm": 0.2724241018295288,
+      "learning_rate": 3.989318037342111e-05,
+      "loss": 0.5149,
+      "mean_token_accuracy": 0.8123500347137451,
+      "step": 203
+    },
+    {
+      "epoch": 0.6415094339622641,
+      "grad_norm": 0.2764907479286194,
+      "learning_rate": 3.989117643852785e-05,
+      "loss": 0.5153,
+      "mean_token_accuracy": 0.8120966553688049,
+      "step": 204
+    },
+    {
+      "epoch": 0.6446540880503144,
+      "grad_norm": 0.24415266513824463,
+      "learning_rate": 3.988915393799703e-05,
+      "loss": 0.5111,
+      "mean_token_accuracy": 0.8132697343826294,
+      "step": 205
+    },
+    {
+      "epoch": 0.6477987421383647,
+      "grad_norm": 0.24454502761363983,
+      "learning_rate": 3.988711287392741e-05,
+      "loss": 0.5063,
+      "mean_token_accuracy": 0.8129962682723999,
+      "step": 206
+    },
+    {
+      "epoch": 0.6509433962264151,
+      "grad_norm": 0.25321948528289795,
+      "learning_rate": 3.9885053248436986e-05,
+      "loss": 0.5117,
+      "mean_token_accuracy": 0.8123859763145447,
+      "step": 207
+    },
+    {
+      "epoch": 0.6540880503144654,
+      "grad_norm": 0.21559487283229828,
+      "learning_rate": 3.9882975063663026e-05,
+      "loss": 0.5242,
+      "mean_token_accuracy": 0.8082608580589294,
+      "step": 208
+    },
+    {
+      "epoch": 0.6572327044025157,
+      "grad_norm": 0.2384122759103775,
+      "learning_rate": 3.9880878321762066e-05,
+      "loss": 0.5156,
+      "mean_token_accuracy": 0.8102776408195496,
+      "step": 209
+    },
+    {
+      "epoch": 0.660377358490566,
+      "grad_norm": 0.24567940831184387,
+      "learning_rate": 3.9878763024909884e-05,
+      "loss": 0.5135,
+      "mean_token_accuracy": 0.8105263113975525,
+      "step": 210
+    },
+    {
+      "epoch": 0.6635220125786163,
+      "grad_norm": 0.2328757643699646,
+      "learning_rate": 3.987662917530153e-05,
+      "loss": 0.5117,
+      "mean_token_accuracy": 0.812843918800354,
+      "step": 211
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "grad_norm": 0.22722886502742767,
+      "learning_rate": 3.987447677515128e-05,
+      "loss": 0.5118,
+      "mean_token_accuracy": 0.8124209642410278,
+      "step": 212
+    },
+    {
+      "epoch": 0.6698113207547169,
+      "grad_norm": 0.24911589920520782,
+      "learning_rate": 3.9872305826692686e-05,
+      "loss": 0.5164,
+      "mean_token_accuracy": 0.809599757194519,
+      "step": 213
+    },
+    {
+      "epoch": 0.6729559748427673,
+      "grad_norm": 0.21305599808692932,
+      "learning_rate": 3.987011633217853e-05,
+      "loss": 0.5184,
+      "mean_token_accuracy": 0.8106415271759033,
+      "step": 214
+    },
+    {
+      "epoch": 0.6761006289308176,
+      "grad_norm": 0.2465692013502121,
+      "learning_rate": 3.986790829388086e-05,
+      "loss": 0.5182,
+      "mean_token_accuracy": 0.8113357424736023,
+      "step": 215
+    },
+    {
+      "epoch": 0.6792452830188679,
+      "grad_norm": 0.23801495134830475,
+      "learning_rate": 3.986568171409094e-05,
+      "loss": 0.5161,
+      "mean_token_accuracy": 0.8109778761863708,
+      "step": 216
+    },
+    {
+      "epoch": 0.6823899371069182,
+      "grad_norm": 0.22071239352226257,
+      "learning_rate": 3.9863436595119284e-05,
+      "loss": 0.5156,
+      "mean_token_accuracy": 0.8117619156837463,
+      "step": 217
+    },
+    {
+      "epoch": 0.6855345911949685,
+      "grad_norm": 0.21508371829986572,
+      "learning_rate": 3.986117293929566e-05,
+      "loss": 0.4984,
+      "mean_token_accuracy": 0.8161824345588684,
+      "step": 218
+    },
+    {
+      "epoch": 0.6886792452830188,
+      "grad_norm": 0.21123456954956055,
+      "learning_rate": 3.985889074896904e-05,
+      "loss": 0.5068,
+      "mean_token_accuracy": 0.8125125169754028,
+      "step": 219
+    },
+    {
+      "epoch": 0.6918238993710691,
+      "grad_norm": 0.2076350450515747,
+      "learning_rate": 3.985659002650767e-05,
+      "loss": 0.5173,
+      "mean_token_accuracy": 0.8120942115783691,
+      "step": 220
+    },
+    {
+      "epoch": 0.6949685534591195,
+      "grad_norm": 0.20349815487861633,
+      "learning_rate": 3.9854270774299e-05,
+      "loss": 0.5088,
+      "mean_token_accuracy": 0.8142285943031311,
+      "step": 221
+    },
+    {
+      "epoch": 0.6981132075471698,
+      "grad_norm": 0.2199343591928482,
+      "learning_rate": 3.9851932994749704e-05,
+      "loss": 0.5158,
+      "mean_token_accuracy": 0.811479926109314,
+      "step": 222
+    },
+    {
+      "epoch": 0.7012578616352201,
+      "grad_norm": 0.2165432721376419,
+      "learning_rate": 3.98495766902857e-05,
+      "loss": 0.5092,
+      "mean_token_accuracy": 0.8122998476028442,
+      "step": 223
+    },
+    {
+      "epoch": 0.7044025157232704,
+      "grad_norm": 0.21602460741996765,
+      "learning_rate": 3.984720186335211e-05,
+      "loss": 0.5192,
+      "mean_token_accuracy": 0.811272382736206,
+      "step": 224
+    },
+    {
+      "epoch": 0.7075471698113207,
+      "grad_norm": 0.2049703598022461,
+      "learning_rate": 3.9844808516413305e-05,
+      "loss": 0.5076,
+      "mean_token_accuracy": 0.8121902346611023,
+      "step": 225
+    },
+    {
+      "epoch": 0.710691823899371,
+      "grad_norm": 0.22164130210876465,
+      "learning_rate": 3.9842396651952836e-05,
+      "loss": 0.5131,
+      "mean_token_accuracy": 0.8126029968261719,
+      "step": 226
+    },
+    {
+      "epoch": 0.7138364779874213,
+      "grad_norm": 0.19871044158935547,
+      "learning_rate": 3.98399662724735e-05,
+      "loss": 0.5163,
+      "mean_token_accuracy": 0.8120744228363037,
+      "step": 227
+    },
+    {
+      "epoch": 0.7169811320754716,
+      "grad_norm": 0.25331369042396545,
+      "learning_rate": 3.98375173804973e-05,
+      "loss": 0.512,
+      "mean_token_accuracy": 0.811780571937561,
+      "step": 228
+    },
+    {
+      "epoch": 0.720125786163522,
+      "grad_norm": 0.21906249225139618,
+      "learning_rate": 3.983504997856544e-05,
+      "loss": 0.5169,
+      "mean_token_accuracy": 0.8110173344612122,
+      "step": 229
+    },
+    {
+      "epoch": 0.7232704402515723,
+      "grad_norm": 0.22297775745391846,
+      "learning_rate": 3.983256406923835e-05,
+      "loss": 0.5185,
+      "mean_token_accuracy": 0.8114131689071655,
+      "step": 230
+    },
+    {
+      "epoch": 0.7264150943396226,
+      "grad_norm": 0.19183681905269623,
+      "learning_rate": 3.9830059655095625e-05,
+      "loss": 0.516,
+      "mean_token_accuracy": 0.8132800459861755,
+      "step": 231
+    },
+    {
+      "epoch": 0.7295597484276729,
+      "grad_norm": 0.24273158609867096,
+      "learning_rate": 3.9827536738736115e-05,
+      "loss": 0.5179,
+      "mean_token_accuracy": 0.8097060322761536,
+      "step": 232
+    },
+    {
+      "epoch": 0.7327044025157232,
+      "grad_norm": 0.21596857905387878,
+      "learning_rate": 3.982499532277785e-05,
+      "loss": 0.491,
+      "mean_token_accuracy": 0.8167231678962708,
+      "step": 233
+    },
+    {
+      "epoch": 0.7358490566037735,
+      "grad_norm": 0.23281732201576233,
+      "learning_rate": 3.982243540985803e-05,
+      "loss": 0.5211,
+      "mean_token_accuracy": 0.8096528649330139,
+      "step": 234
+    },
+    {
+      "epoch": 0.7389937106918238,
+      "grad_norm": 0.1850733757019043,
+      "learning_rate": 3.9819857002633084e-05,
+      "loss": 0.5119,
+      "mean_token_accuracy": 0.812339723110199,
+      "step": 235
+    },
+    {
+      "epoch": 0.7421383647798742,
+      "grad_norm": 0.2426994889974594,
+      "learning_rate": 3.981726010377862e-05,
+      "loss": 0.5155,
+      "mean_token_accuracy": 0.8118655681610107,
+      "step": 236
+    },
+    {
+      "epoch": 0.7452830188679245,
+      "grad_norm": 0.20813558995723724,
+      "learning_rate": 3.981464471598943e-05,
+      "loss": 0.5147,
+      "mean_token_accuracy": 0.8122422099113464,
+      "step": 237
+    },
+    {
+      "epoch": 0.7484276729559748,
+      "grad_norm": 0.25706005096435547,
+      "learning_rate": 3.98120108419795e-05,
+      "loss": 0.5029,
+      "mean_token_accuracy": 0.814775824546814,
+      "step": 238
+    },
+    {
+      "epoch": 0.7515723270440252,
+      "grad_norm": 0.22012631595134735,
+      "learning_rate": 3.9809358484482e-05,
+      "loss": 0.5086,
+      "mean_token_accuracy": 0.8135779500007629,
+      "step": 239
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.2591609060764313,
+      "learning_rate": 3.980668764624927e-05,
+      "loss": 0.5166,
+      "mean_token_accuracy": 0.811547577381134,
+      "step": 240
+    },
+    {
+      "epoch": 0.7578616352201258,
+      "grad_norm": 0.2425551414489746,
+      "learning_rate": 3.9803998330052834e-05,
+      "loss": 0.5106,
+      "mean_token_accuracy": 0.8126974105834961,
+      "step": 241
+    },
+    {
+      "epoch": 0.7610062893081762,
+      "grad_norm": 0.19677698612213135,
+      "learning_rate": 3.980129053868339e-05,
+      "loss": 0.5025,
+      "mean_token_accuracy": 0.8132906556129456,
+      "step": 242
+    },
+    {
+      "epoch": 0.7641509433962265,
+      "grad_norm": 0.24536176025867462,
+      "learning_rate": 3.9798564274950815e-05,
+      "loss": 0.4983,
+      "mean_token_accuracy": 0.8168281316757202,
+      "step": 243
+    },
+    {
+      "epoch": 0.7672955974842768,
+      "grad_norm": 0.23609694838523865,
+      "learning_rate": 3.979581954168414e-05,
+      "loss": 0.4942,
+      "mean_token_accuracy": 0.8176018595695496,
+      "step": 244
+    },
+    {
+      "epoch": 0.7704402515723271,
+      "grad_norm": 0.2168692797422409,
+      "learning_rate": 3.9793056341731556e-05,
+      "loss": 0.5168,
+      "mean_token_accuracy": 0.8108464479446411,
+      "step": 245
+    },
+    {
+      "epoch": 0.7735849056603774,
+      "grad_norm": 0.21630840003490448,
+      "learning_rate": 3.979027467796046e-05,
+      "loss": 0.501,
+      "mean_token_accuracy": 0.8146473169326782,
+      "step": 246
+    },
+    {
+      "epoch": 0.7767295597484277,
+      "grad_norm": 0.20055299997329712,
+      "learning_rate": 3.978747455325736e-05,
+      "loss": 0.5082,
+      "mean_token_accuracy": 0.8135594129562378,
+      "step": 247
+    },
+    {
+      "epoch": 0.779874213836478,
+      "grad_norm": 0.282925546169281,
+      "learning_rate": 3.9784655970527934e-05,
+      "loss": 0.5144,
+      "mean_token_accuracy": 0.8131219744682312,
+      "step": 248
+    },
+    {
+      "epoch": 0.7830188679245284,
+      "grad_norm": 0.21999655663967133,
+      "learning_rate": 3.978181893269703e-05,
+      "loss": 0.5046,
+      "mean_token_accuracy": 0.8133974671363831,
+      "step": 249
+    },
+    {
+      "epoch": 0.7861635220125787,
+      "grad_norm": 0.2439885139465332,
+      "learning_rate": 3.977896344270864e-05,
+      "loss": 0.5052,
+      "mean_token_accuracy": 0.8127160668373108,
+      "step": 250
+    },
+    {
+      "epoch": 0.789308176100629,
+      "grad_norm": 0.2688886225223541,
+      "learning_rate": 3.9776089503525895e-05,
+      "loss": 0.5109,
+      "mean_token_accuracy": 0.81061190366745,
+      "step": 251
+    },
+    {
+      "epoch": 0.7924528301886793,
+      "grad_norm": 0.2625892162322998,
+      "learning_rate": 3.977319711813107e-05,
+      "loss": 0.5172,
+      "mean_token_accuracy": 0.8132520914077759,
+      "step": 252
+    },
+    {
+      "epoch": 0.7955974842767296,
+      "grad_norm": 0.1982739269733429,
+      "learning_rate": 3.97702862895256e-05,
+      "loss": 0.499,
+      "mean_token_accuracy": 0.8147685527801514,
+      "step": 253
+    },
+    {
+      "epoch": 0.7987421383647799,
+      "grad_norm": 0.2820964455604553,
+      "learning_rate": 3.976735702073003e-05,
+      "loss": 0.5134,
+      "mean_token_accuracy": 0.8140535354614258,
+      "step": 254
+    },
+    {
+      "epoch": 0.8018867924528302,
+      "grad_norm": 0.19811579585075378,
+      "learning_rate": 3.9764409314784074e-05,
+      "loss": 0.5176,
+      "mean_token_accuracy": 0.8116709589958191,
+      "step": 255
+    },
+    {
+      "epoch": 0.8050314465408805,
+      "grad_norm": 0.26263049244880676,
+      "learning_rate": 3.9761443174746556e-05,
+      "loss": 0.5162,
+      "mean_token_accuracy": 0.8113158345222473,
+      "step": 256
+    },
+    {
+      "epoch": 0.8081761006289309,
+      "grad_norm": 0.24067486822605133,
+      "learning_rate": 3.975845860369542e-05,
+      "loss": 0.5055,
+      "mean_token_accuracy": 0.814015805721283,
+      "step": 257
+    },
+    {
+      "epoch": 0.8113207547169812,
+      "grad_norm": 0.19591133296489716,
+      "learning_rate": 3.975545560472776e-05,
+      "loss": 0.5074,
+      "mean_token_accuracy": 0.8133794069290161,
+      "step": 258
+    },
+    {
+      "epoch": 0.8144654088050315,
+      "grad_norm": 0.24649196863174438,
+      "learning_rate": 3.975243418095978e-05,
+      "loss": 0.5141,
+      "mean_token_accuracy": 0.8121534585952759,
+      "step": 259
+    },
+    {
+      "epoch": 0.8176100628930818,
+      "grad_norm": 0.20178139209747314,
+      "learning_rate": 3.9749394335526806e-05,
+      "loss": 0.5016,
+      "mean_token_accuracy": 0.8149169087409973,
+      "step": 260
+    },
+    {
+      "epoch": 0.8207547169811321,
+      "grad_norm": 0.2396860122680664,
+      "learning_rate": 3.9746336071583284e-05,
+      "loss": 0.5112,
+      "mean_token_accuracy": 0.8116641640663147,
+      "step": 261
+    },
+    {
+      "epoch": 0.8238993710691824,
+      "grad_norm": 0.2169322371482849,
+      "learning_rate": 3.9743259392302765e-05,
+      "loss": 0.5033,
+      "mean_token_accuracy": 0.8144574761390686,
+      "step": 262
+    },
+    {
+      "epoch": 0.8270440251572327,
+      "grad_norm": 0.27799248695373535,
+      "learning_rate": 3.9740164300877905e-05,
+      "loss": 0.5134,
+      "mean_token_accuracy": 0.8135626316070557,
+      "step": 263
+    },
+    {
+      "epoch": 0.8301886792452831,
+      "grad_norm": 0.2029547095298767,
+      "learning_rate": 3.9737050800520484e-05,
+      "loss": 0.4998,
+      "mean_token_accuracy": 0.8158847689628601,
+      "step": 264
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "grad_norm": 0.23742541670799255,
+      "learning_rate": 3.973391889446137e-05,
+      "loss": 0.5002,
+      "mean_token_accuracy": 0.8158310651779175,
+      "step": 265
+    },
+    {
+      "epoch": 0.8364779874213837,
+      "grad_norm": 0.23540346324443817,
+      "learning_rate": 3.973076858595054e-05,
+      "loss": 0.5075,
+      "mean_token_accuracy": 0.8140506148338318,
+      "step": 266
+    },
+    {
+      "epoch": 0.839622641509434,
+      "grad_norm": 0.21212562918663025,
+      "learning_rate": 3.972759987825706e-05,
+      "loss": 0.4964,
+      "mean_token_accuracy": 0.817358672618866,
+      "step": 267
+    },
+    {
+      "epoch": 0.8427672955974843,
+      "grad_norm": 0.2811741828918457,
+      "learning_rate": 3.972441277466909e-05,
+      "loss": 0.5045,
+      "mean_token_accuracy": 0.8151137828826904,
+      "step": 268
+    },
+    {
+      "epoch": 0.8459119496855346,
+      "grad_norm": 0.22324548661708832,
+      "learning_rate": 3.972120727849388e-05,
+      "loss": 0.5184,
+      "mean_token_accuracy": 0.8102690577507019,
+      "step": 269
+    },
+    {
+      "epoch": 0.8490566037735849,
+      "grad_norm": 0.2826754152774811,
+      "learning_rate": 3.9717983393057774e-05,
+      "loss": 0.5033,
+      "mean_token_accuracy": 0.8149948120117188,
+      "step": 270
+    },
+    {
+      "epoch": 0.8522012578616353,
+      "grad_norm": 0.20113104581832886,
+      "learning_rate": 3.971474112170618e-05,
+      "loss": 0.4963,
+      "mean_token_accuracy": 0.8169792890548706,
+      "step": 271
+    },
+    {
+      "epoch": 0.8553459119496856,
+      "grad_norm": 0.31522729992866516,
+      "learning_rate": 3.971148046780361e-05,
+      "loss": 0.5043,
+      "mean_token_accuracy": 0.8139733672142029,
+      "step": 272
+    },
+    {
+      "epoch": 0.8584905660377359,
+      "grad_norm": 0.24440452456474304,
+      "learning_rate": 3.970820143473363e-05,
+      "loss": 0.506,
+      "mean_token_accuracy": 0.8141810297966003,
+      "step": 273
+    },
+    {
+      "epoch": 0.8616352201257862,
+      "grad_norm": 0.34243300557136536,
+      "learning_rate": 3.970490402589889e-05,
+      "loss": 0.5009,
+      "mean_token_accuracy": 0.8145892024040222,
+      "step": 274
+    },
+    {
+      "epoch": 0.8647798742138365,
+      "grad_norm": 0.3184676468372345,
+      "learning_rate": 3.970158824472109e-05,
+      "loss": 0.5019,
+      "mean_token_accuracy": 0.8156532049179077,
+      "step": 275
+    },
+    {
+      "epoch": 0.8679245283018868,
+      "grad_norm": 0.2528303861618042,
+      "learning_rate": 3.969825409464103e-05,
+      "loss": 0.5099,
+      "mean_token_accuracy": 0.8132652044296265,
+      "step": 276
+    },
+    {
+      "epoch": 0.8710691823899371,
+      "grad_norm": 0.2652378976345062,
+      "learning_rate": 3.969490157911854e-05,
+      "loss": 0.4965,
+      "mean_token_accuracy": 0.8170227408409119,
+      "step": 277
+    },
+    {
+      "epoch": 0.8742138364779874,
+      "grad_norm": 0.2514651119709015,
+      "learning_rate": 3.9691530701632525e-05,
+      "loss": 0.4947,
+      "mean_token_accuracy": 0.8181447386741638,
+      "step": 278
+    },
+    {
+      "epoch": 0.8773584905660378,
+      "grad_norm": 0.24551190435886383,
+      "learning_rate": 3.968814146568093e-05,
+      "loss": 0.5039,
+      "mean_token_accuracy": 0.8154646158218384,
+      "step": 279
+    },
+    {
+      "epoch": 0.8805031446540881,
+      "grad_norm": 0.250418096780777,
+      "learning_rate": 3.9684733874780764e-05,
+      "loss": 0.5057,
+      "mean_token_accuracy": 0.8137528896331787,
+      "step": 280
+    },
+    {
+      "epoch": 0.8836477987421384,
+      "grad_norm": 0.24162472784519196,
+      "learning_rate": 3.9681307932468066e-05,
+      "loss": 0.499,
+      "mean_token_accuracy": 0.8132847547531128,
+      "step": 281
+    },
+    {
+      "epoch": 0.8867924528301887,
+      "grad_norm": 0.25049126148223877,
+      "learning_rate": 3.967786364229794e-05,
+      "loss": 0.5016,
+      "mean_token_accuracy": 0.8135824203491211,
+      "step": 282
+    },
+    {
+      "epoch": 0.889937106918239,
+      "grad_norm": 0.21691341698169708,
+      "learning_rate": 3.9674401007844525e-05,
+      "loss": 0.4909,
+      "mean_token_accuracy": 0.8152883052825928,
+      "step": 283
+    },
+    {
+      "epoch": 0.8930817610062893,
+      "grad_norm": 0.22291503846645355,
+      "learning_rate": 3.967092003270098e-05,
+      "loss": 0.4959,
+      "mean_token_accuracy": 0.8131790161132812,
+      "step": 284
+    },
+    {
+      "epoch": 0.8962264150943396,
+      "grad_norm": 0.21946825087070465,
+      "learning_rate": 3.96674207204795e-05,
+      "loss": 0.4943,
+      "mean_token_accuracy": 0.8152416348457336,
+      "step": 285
+    },
+    {
+      "epoch": 0.89937106918239,
+      "grad_norm": 0.21576789021492004,
+      "learning_rate": 3.966390307481133e-05,
+      "loss": 0.4963,
+      "mean_token_accuracy": 0.8158524036407471,
+      "step": 286
+    },
+    {
+      "epoch": 0.9025157232704403,
+      "grad_norm": 0.18200084567070007,
+      "learning_rate": 3.966036709934671e-05,
+      "loss": 0.4983,
+      "mean_token_accuracy": 0.8171305656433105,
+      "step": 287
+    },
+    {
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.20305195450782776,
+      "learning_rate": 3.9656812797754924e-05,
+      "loss": 0.5004,
+      "mean_token_accuracy": 0.8154840469360352,
+      "step": 288
+    },
+    {
+      "epoch": 0.9088050314465409,
+      "grad_norm": 0.17034946382045746,
+      "learning_rate": 3.965324017372426e-05,
+      "loss": 0.4975,
+      "mean_token_accuracy": 0.8167921304702759,
+      "step": 289
+    },
+    {
+      "epoch": 0.9119496855345912,
+      "grad_norm": 0.2208290696144104,
+      "learning_rate": 3.964964923096202e-05,
+      "loss": 0.5022,
+      "mean_token_accuracy": 0.815450131893158,
+      "step": 290
+    },
+    {
+      "epoch": 0.9150943396226415,
+      "grad_norm": 0.18164092302322388,
+      "learning_rate": 3.964603997319452e-05,
+      "loss": 0.4964,
+      "mean_token_accuracy": 0.8163471221923828,
+      "step": 291
+    },
+    {
+      "epoch": 0.9182389937106918,
+      "grad_norm": 0.18271009624004364,
+      "learning_rate": 3.964241240416708e-05,
+      "loss": 0.489,
+      "mean_token_accuracy": 0.8168913125991821,
+      "step": 292
+    },
+    {
+      "epoch": 0.9213836477987422,
+      "grad_norm": 0.19366014003753662,
+      "learning_rate": 3.963876652764402e-05,
+      "loss": 0.5002,
+      "mean_token_accuracy": 0.815049946308136,
+      "step": 293
+    },
+    {
+      "epoch": 0.9245283018867925,
+      "grad_norm": 0.21568985283374786,
+      "learning_rate": 3.963510234740866e-05,
+      "loss": 0.5057,
+      "mean_token_accuracy": 0.8131378293037415,
+      "step": 294
+    },
+    {
+      "epoch": 0.9276729559748428,
+      "grad_norm": 0.16388444602489471,
+      "learning_rate": 3.963141986726332e-05,
+      "loss": 0.4991,
+      "mean_token_accuracy": 0.8145990371704102,
+      "step": 295
+    },
+    {
+      "epoch": 0.9308176100628931,
+      "grad_norm": 0.21059076488018036,
+      "learning_rate": 3.962771909102928e-05,
+      "loss": 0.5068,
+      "mean_token_accuracy": 0.8122072815895081,
+      "step": 296
+    },
+    {
+      "epoch": 0.9339622641509434,
+      "grad_norm": 0.20338748395442963,
+      "learning_rate": 3.962400002254685e-05,
+      "loss": 0.5021,
+      "mean_token_accuracy": 0.8148384690284729,
+      "step": 297
+    },
+    {
+      "epoch": 0.9371069182389937,
+      "grad_norm": 0.2134150117635727,
+      "learning_rate": 3.962026266567529e-05,
+      "loss": 0.5077,
+      "mean_token_accuracy": 0.8135893940925598,
+      "step": 298
+    },
+    {
+      "epoch": 0.940251572327044,
+      "grad_norm": 0.2366155982017517,
+      "learning_rate": 3.961650702429285e-05,
+      "loss": 0.5028,
+      "mean_token_accuracy": 0.8151034712791443,
+      "step": 299
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.19872386753559113,
+      "learning_rate": 3.9612733102296757e-05,
+      "loss": 0.49,
+      "mean_token_accuracy": 0.818320095539093,
+      "step": 300
+    },
+    {
+      "epoch": 0.9465408805031447,
+      "grad_norm": 0.24557925760746002,
+      "learning_rate": 3.960894090360319e-05,
+      "loss": 0.4978,
+      "mean_token_accuracy": 0.8179511427879333,
+      "step": 301
+    },
+    {
+      "epoch": 0.949685534591195,
+      "grad_norm": 0.18634836375713348,
+      "learning_rate": 3.960513043214731e-05,
+      "loss": 0.508,
+      "mean_token_accuracy": 0.8148860335350037,
+      "step": 302
+    },
+    {
+      "epoch": 0.9528301886792453,
+      "grad_norm": 0.2479529231786728,
+      "learning_rate": 3.960130169188324e-05,
+      "loss": 0.5061,
+      "mean_token_accuracy": 0.8134167194366455,
+      "step": 303
+    },
+    {
+      "epoch": 0.9559748427672956,
+      "grad_norm": 0.2314770370721817,
+      "learning_rate": 3.959745468678407e-05,
+      "loss": 0.506,
+      "mean_token_accuracy": 0.8139018416404724,
+      "step": 304
+    },
+    {
+      "epoch": 0.9591194968553459,
+      "grad_norm": 0.18175768852233887,
+      "learning_rate": 3.959358942084179e-05,
+      "loss": 0.5025,
+      "mean_token_accuracy": 0.814060389995575,
+      "step": 305
+    },
+    {
+      "epoch": 0.9622641509433962,
+      "grad_norm": 0.26930001378059387,
+      "learning_rate": 3.958970589806741e-05,
+      "loss": 0.4899,
+      "mean_token_accuracy": 0.8199323415756226,
+      "step": 306
+    },
+    {
+      "epoch": 0.9654088050314465,
+      "grad_norm": 0.18011023104190826,
+      "learning_rate": 3.958580412249085e-05,
+      "loss": 0.4998,
+      "mean_token_accuracy": 0.8157727122306824,
+      "step": 307
+    },
+    {
+      "epoch": 0.9685534591194969,
+      "grad_norm": 0.20771653950214386,
+      "learning_rate": 3.9581884098160965e-05,
+      "loss": 0.4946,
+      "mean_token_accuracy": 0.8169547915458679,
+      "step": 308
+    },
+    {
+      "epoch": 0.9716981132075472,
+      "grad_norm": 0.20119163393974304,
+      "learning_rate": 3.9577945829145565e-05,
+      "loss": 0.493,
+      "mean_token_accuracy": 0.816156268119812,
+      "step": 309
+    },
+    {
+      "epoch": 0.9748427672955975,
+      "grad_norm": 0.1962110996246338,
+      "learning_rate": 3.9573989319531376e-05,
+      "loss": 0.4838,
+      "mean_token_accuracy": 0.8181387782096863,
+      "step": 310
+    },
+    {
+      "epoch": 0.9779874213836478,
+      "grad_norm": 0.19966855645179749,
+      "learning_rate": 3.957001457342407e-05,
+      "loss": 0.494,
+      "mean_token_accuracy": 0.8177944421768188,
+      "step": 311
+    },
+    {
+      "epoch": 0.9811320754716981,
+      "grad_norm": 0.16196459531784058,
+      "learning_rate": 3.956602159494822e-05,
+      "loss": 0.4999,
+      "mean_token_accuracy": 0.8170042037963867,
+      "step": 312
+    },
+    {
+      "epoch": 0.9842767295597484,
+      "grad_norm": 0.2277836948633194,
+      "learning_rate": 3.956201038824734e-05,
+      "loss": 0.5079,
+      "mean_token_accuracy": 0.8125689625740051,
+      "step": 313
+    },
+    {
+      "epoch": 0.9874213836477987,
+      "grad_norm": 0.15822745859622955,
+      "learning_rate": 3.9557980957483857e-05,
+      "loss": 0.4944,
+      "mean_token_accuracy": 0.8152970671653748,
+      "step": 314
+    },
+    {
+      "epoch": 0.9905660377358491,
+      "grad_norm": 0.23701541125774384,
+      "learning_rate": 3.955393330683909e-05,
+      "loss": 0.4949,
+      "mean_token_accuracy": 0.8175249695777893,
+      "step": 315
+    },
+    {
+      "epoch": 0.9937106918238994,
+      "grad_norm": 0.17451459169387817,
+      "learning_rate": 3.954986744051329e-05,
+      "loss": 0.4903,
+      "mean_token_accuracy": 0.8196095824241638,
+      "step": 316
+    },
+    {
+      "epoch": 0.9968553459119497,
+      "grad_norm": 0.20685537159442902,
+      "learning_rate": 3.954578336272559e-05,
+      "loss": 0.4872,
+      "mean_token_accuracy": 0.8189926147460938,
+      "step": 317
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.24290406703948975,
+      "learning_rate": 3.954168107771402e-05,
+      "loss": 0.4891,
+      "mean_token_accuracy": 0.8178141713142395,
+      "step": 318
+    },
+    {
+      "epoch": 1.0031446540880504,
+      "grad_norm": 0.17915529012680054,
+      "learning_rate": 3.953756058973554e-05,
+      "loss": 0.4626,
+      "mean_token_accuracy": 0.8239535689353943,
+      "step": 319
+    },
+    {
+      "epoch": 1.0062893081761006,
+      "grad_norm": 0.24146446585655212,
+      "learning_rate": 3.9533421903065946e-05,
+      "loss": 0.4629,
+      "mean_token_accuracy": 0.8243134617805481,
+      "step": 320
+    },
+    {
+      "epoch": 1.009433962264151,
+      "grad_norm": 0.24396361410617828,
+      "learning_rate": 3.9529265021999965e-05,
+      "loss": 0.4522,
+      "mean_token_accuracy": 0.8263998031616211,
+      "step": 321
+    },
+    {
+      "epoch": 1.0125786163522013,
+      "grad_norm": 0.21227137744426727,
+      "learning_rate": 3.952508995085117e-05,
+      "loss": 0.4574,
+      "mean_token_accuracy": 0.8258650898933411,
+      "step": 322
+    },
+    {
+      "epoch": 1.0157232704402517,
+      "grad_norm": 0.26239192485809326,
+      "learning_rate": 3.952089669395203e-05,
+      "loss": 0.4477,
+      "mean_token_accuracy": 0.826561689376831,
+      "step": 323
+    },
+    {
+      "epoch": 1.0188679245283019,
+      "grad_norm": 0.21801580488681793,
+      "learning_rate": 3.951668525565387e-05,
+      "loss": 0.4621,
+      "mean_token_accuracy": 0.8246572017669678,
+      "step": 324
+    },
+    {
+      "epoch": 1.0220125786163523,
+      "grad_norm": 0.24545596539974213,
+      "learning_rate": 3.95124556403269e-05,
+      "loss": 0.442,
+      "mean_token_accuracy": 0.8272949457168579,
+      "step": 325
+    },
+    {
+      "epoch": 1.0251572327044025,
+      "grad_norm": 0.21632397174835205,
+      "learning_rate": 3.950820785236018e-05,
+      "loss": 0.4616,
+      "mean_token_accuracy": 0.8258722424507141,
+      "step": 326
+    },
+    {
+      "epoch": 1.028301886792453,
+      "grad_norm": 0.25987663865089417,
+      "learning_rate": 3.950394189616164e-05,
+      "loss": 0.4433,
+      "mean_token_accuracy": 0.8266246318817139,
+      "step": 327
+    },
+    {
+      "epoch": 1.0314465408805031,
+      "grad_norm": 0.21071231365203857,
+      "learning_rate": 3.949965777615804e-05,
+      "loss": 0.4587,
+      "mean_token_accuracy": 0.8251431584358215,
+      "step": 328
+    },
+    {
+      "epoch": 1.0345911949685536,
+      "grad_norm": 0.25732994079589844,
+      "learning_rate": 3.949535549679502e-05,
+      "loss": 0.4542,
+      "mean_token_accuracy": 0.825411856174469,
+      "step": 329
+    },
+    {
+      "epoch": 1.0377358490566038,
+      "grad_norm": 0.2170933187007904,
+      "learning_rate": 3.9491035062537026e-05,
+      "loss": 0.4488,
+      "mean_token_accuracy": 0.8288849592208862,
+      "step": 330
+    },
+    {
+      "epoch": 1.0408805031446542,
+      "grad_norm": 0.23197437822818756,
+      "learning_rate": 3.948669647786738e-05,
+      "loss": 0.4475,
+      "mean_token_accuracy": 0.8281545042991638,
+      "step": 331
+    },
+    {
+      "epoch": 1.0440251572327044,
+      "grad_norm": 0.2003556787967682,
+      "learning_rate": 3.948233974728821e-05,
+      "loss": 0.4581,
+      "mean_token_accuracy": 0.8240156173706055,
+      "step": 332
+    },
+    {
+      "epoch": 1.0471698113207548,
+      "grad_norm": 0.24654971063137054,
+      "learning_rate": 3.94779648753205e-05,
+      "loss": 0.4593,
+      "mean_token_accuracy": 0.8252534866333008,
+      "step": 333
+    },
+    {
+      "epoch": 1.050314465408805,
+      "grad_norm": 0.1977737993001938,
+      "learning_rate": 3.947357186650403e-05,
+      "loss": 0.4569,
+      "mean_token_accuracy": 0.826411247253418,
+      "step": 334
+    },
+    {
+      "epoch": 1.0534591194968554,
+      "grad_norm": 0.1900518238544464,
+      "learning_rate": 3.9469160725397426e-05,
+      "loss": 0.4448,
+      "mean_token_accuracy": 0.8280865550041199,
+      "step": 335
+    },
+    {
+      "epoch": 1.0566037735849056,
+      "grad_norm": 0.20230652391910553,
+      "learning_rate": 3.946473145657813e-05,
+      "loss": 0.461,
+      "mean_token_accuracy": 0.8244823813438416,
+      "step": 336
+    },
+    {
+      "epoch": 1.059748427672956,
+      "grad_norm": 0.21481932699680328,
+      "learning_rate": 3.946028406464236e-05,
+      "loss": 0.4429,
+      "mean_token_accuracy": 0.8295682668685913,
+      "step": 337
+    },
+    {
+      "epoch": 1.0628930817610063,
+      "grad_norm": 0.20926858484745026,
+      "learning_rate": 3.9455818554205186e-05,
+      "loss": 0.4472,
+      "mean_token_accuracy": 0.8260708451271057,
+      "step": 338
+    },
+    {
+      "epoch": 1.0660377358490567,
+      "grad_norm": 0.19324615597724915,
+      "learning_rate": 3.945133492990045e-05,
+      "loss": 0.4695,
+      "mean_token_accuracy": 0.8230428099632263,
+      "step": 339
+    },
+    {
+      "epoch": 1.069182389937107,
+      "grad_norm": 0.22138728201389313,
+      "learning_rate": 3.944683319638081e-05,
+      "loss": 0.4436,
+      "mean_token_accuracy": 0.8289992213249207,
+      "step": 340
+    },
+    {
+      "epoch": 1.0723270440251573,
+      "grad_norm": 0.18856871128082275,
+      "learning_rate": 3.9442313358317694e-05,
+      "loss": 0.4393,
+      "mean_token_accuracy": 0.8271635174751282,
+      "step": 341
+    },
+    {
+      "epoch": 1.0754716981132075,
+      "grad_norm": 0.19933389127254486,
+      "learning_rate": 3.943777542040134e-05,
+      "loss": 0.4654,
+      "mean_token_accuracy": 0.8237259387969971,
+      "step": 342
+    },
+    {
+      "epoch": 1.078616352201258,
+      "grad_norm": 0.24634775519371033,
+      "learning_rate": 3.943321938734074e-05,
+      "loss": 0.4401,
+      "mean_token_accuracy": 0.829694926738739,
+      "step": 343
+    },
+    {
+      "epoch": 1.0817610062893082,
+      "grad_norm": 0.19513890147209167,
+      "learning_rate": 3.942864526386369e-05,
+      "loss": 0.4558,
+      "mean_token_accuracy": 0.825875997543335,
+      "step": 344
+    },
+    {
+      "epoch": 1.0849056603773586,
+      "grad_norm": 0.23661689460277557,
+      "learning_rate": 3.9424053054716755e-05,
+      "loss": 0.4544,
+      "mean_token_accuracy": 0.8247132301330566,
+      "step": 345
+    },
+    {
+      "epoch": 1.0880503144654088,
+      "grad_norm": 0.23907917737960815,
+      "learning_rate": 3.941944276466526e-05,
+      "loss": 0.464,
+      "mean_token_accuracy": 0.8243969082832336,
+      "step": 346
+    },
+    {
+      "epoch": 1.0911949685534592,
+      "grad_norm": 0.21444515883922577,
+      "learning_rate": 3.941481439849328e-05,
+      "loss": 0.4479,
+      "mean_token_accuracy": 0.8274782299995422,
+      "step": 347
+    },
+    {
+      "epoch": 1.0943396226415094,
+      "grad_norm": 0.23370276391506195,
+      "learning_rate": 3.941016796100368e-05,
+      "loss": 0.4586,
+      "mean_token_accuracy": 0.8257689476013184,
+      "step": 348
+    },
+    {
+      "epoch": 1.0974842767295598,
+      "grad_norm": 0.24292296171188354,
+      "learning_rate": 3.9405503457018045e-05,
+      "loss": 0.4514,
+      "mean_token_accuracy": 0.826076865196228,
+      "step": 349
+    },
+    {
+      "epoch": 1.10062893081761,
+      "grad_norm": 0.2037438154220581,
+      "learning_rate": 3.940082089137673e-05,
+      "loss": 0.4601,
+      "mean_token_accuracy": 0.8253549933433533,
+      "step": 350
+    },
+    {
+      "epoch": 1.1037735849056605,
+      "grad_norm": 0.2162904292345047,
+      "learning_rate": 3.939612026893881e-05,
+      "loss": 0.4383,
+      "mean_token_accuracy": 0.8296216726303101,
+      "step": 351
+    },
+    {
+      "epoch": 1.1069182389937107,
+      "grad_norm": 0.254822313785553,
+      "learning_rate": 3.939140159458213e-05,
+      "loss": 0.4668,
+      "mean_token_accuracy": 0.8235269784927368,
+      "step": 352
+    },
+    {
+      "epoch": 1.110062893081761,
+      "grad_norm": 0.1879785656929016,
+      "learning_rate": 3.938666487320323e-05,
+      "loss": 0.4435,
+      "mean_token_accuracy": 0.8316510915756226,
+      "step": 353
+    },
+    {
+      "epoch": 1.1132075471698113,
+      "grad_norm": 0.19271206855773926,
+      "learning_rate": 3.9381910109717415e-05,
+      "loss": 0.4598,
+      "mean_token_accuracy": 0.823648989200592,
+      "step": 354
+    },
+    {
+      "epoch": 1.1163522012578617,
+      "grad_norm": 0.22165369987487793,
+      "learning_rate": 3.937713730905868e-05,
+      "loss": 0.4611,
+      "mean_token_accuracy": 0.8243672847747803,
+      "step": 355
+    },
+    {
+      "epoch": 1.119496855345912,
+      "grad_norm": 0.17029280960559845,
+      "learning_rate": 3.937234647617975e-05,
+      "loss": 0.4536,
+      "mean_token_accuracy": 0.8255760073661804,
+      "step": 356
+    },
+    {
+      "epoch": 1.1226415094339623,
+      "grad_norm": 0.21115252375602722,
+      "learning_rate": 3.936753761605208e-05,
+      "loss": 0.4593,
+      "mean_token_accuracy": 0.8264986276626587,
+      "step": 357
+    },
+    {
+      "epoch": 1.1257861635220126,
+      "grad_norm": 0.19516626000404358,
+      "learning_rate": 3.936271073366579e-05,
+      "loss": 0.4509,
+      "mean_token_accuracy": 0.8269539475440979,
+      "step": 358
+    },
+    {
+      "epoch": 1.128930817610063,
+      "grad_norm": 0.1787671595811844,
+      "learning_rate": 3.935786583402975e-05,
+      "loss": 0.4511,
+      "mean_token_accuracy": 0.8269217014312744,
+      "step": 359
+    },
+    {
+      "epoch": 1.1320754716981132,
+      "grad_norm": 0.19312171638011932,
+      "learning_rate": 3.935300292217148e-05,
+      "loss": 0.4382,
+      "mean_token_accuracy": 0.8278734087944031,
+      "step": 360
+    },
+    {
+      "epoch": 1.1352201257861636,
+      "grad_norm": 0.1753867268562317,
+      "learning_rate": 3.9348122003137224e-05,
+      "loss": 0.4494,
+      "mean_token_accuracy": 0.8283815979957581,
+      "step": 361
+    },
+    {
+      "epoch": 1.1383647798742138,
+      "grad_norm": 0.17436008155345917,
+      "learning_rate": 3.9343223081991904e-05,
+      "loss": 0.4375,
+      "mean_token_accuracy": 0.8322175145149231,
+      "step": 362
+    },
+    {
+      "epoch": 1.1415094339622642,
+      "grad_norm": 0.17123448848724365,
+      "learning_rate": 3.933830616381912e-05,
+      "loss": 0.4603,
+      "mean_token_accuracy": 0.8241879940032959,
+      "step": 363
+    },
+    {
+      "epoch": 1.1446540880503144,
+      "grad_norm": 0.17863644659519196,
+      "learning_rate": 3.933337125372115e-05,
+      "loss": 0.443,
+      "mean_token_accuracy": 0.8289136290550232,
+      "step": 364
+    },
+    {
+      "epoch": 1.1477987421383649,
+      "grad_norm": 0.17425286769866943,
+      "learning_rate": 3.932841835681893e-05,
+      "loss": 0.4463,
+      "mean_token_accuracy": 0.8275046348571777,
+      "step": 365
+    },
+    {
+      "epoch": 1.150943396226415,
+      "grad_norm": 0.18034741282463074,
+      "learning_rate": 3.932344747825208e-05,
+      "loss": 0.4535,
+      "mean_token_accuracy": 0.8258575201034546,
+      "step": 366
+    },
+    {
+      "epoch": 1.1540880503144655,
+      "grad_norm": 0.166525736451149,
+      "learning_rate": 3.931845862317887e-05,
+      "loss": 0.4585,
+      "mean_token_accuracy": 0.8256432414054871,
+      "step": 367
+    },
+    {
+      "epoch": 1.1572327044025157,
+      "grad_norm": 0.1865280121564865,
+      "learning_rate": 3.931345179677622e-05,
+      "loss": 0.4384,
+      "mean_token_accuracy": 0.8310089111328125,
+      "step": 368
+    },
+    {
+      "epoch": 1.1603773584905661,
+      "grad_norm": 0.15174388885498047,
+      "learning_rate": 3.930842700423971e-05,
+      "loss": 0.4501,
+      "mean_token_accuracy": 0.8273698091506958,
+      "step": 369
+    },
+    {
+      "epoch": 1.1635220125786163,
+      "grad_norm": 0.19328919053077698,
+      "learning_rate": 3.9303384250783555e-05,
+      "loss": 0.4473,
+      "mean_token_accuracy": 0.8262863755226135,
+      "step": 370
+    },
+    {
+      "epoch": 1.1666666666666667,
+      "grad_norm": 0.19845743477344513,
+      "learning_rate": 3.929832354164061e-05,
+      "loss": 0.4517,
+      "mean_token_accuracy": 0.8262023329734802,
+      "step": 371
+    },
+    {
+      "epoch": 1.169811320754717,
+      "grad_norm": 0.18804894387722015,
+      "learning_rate": 3.929324488206235e-05,
+      "loss": 0.4416,
+      "mean_token_accuracy": 0.8311719298362732,
+      "step": 372
+    },
+    {
+      "epoch": 1.1729559748427674,
+      "grad_norm": 0.2016197293996811,
+      "learning_rate": 3.928814827731892e-05,
+      "loss": 0.4496,
+      "mean_token_accuracy": 0.8284710049629211,
+      "step": 373
+    },
+    {
+      "epoch": 1.1761006289308176,
+      "grad_norm": 0.1766403168439865,
+      "learning_rate": 3.928303373269903e-05,
+      "loss": 0.4613,
+      "mean_token_accuracy": 0.826279878616333,
+      "step": 374
+    },
+    {
+      "epoch": 1.179245283018868,
+      "grad_norm": 0.19094939529895782,
+      "learning_rate": 3.927790125351004e-05,
+      "loss": 0.446,
+      "mean_token_accuracy": 0.8275184035301208,
+      "step": 375
+    },
+    {
+      "epoch": 1.1823899371069182,
+      "grad_norm": 0.18953315913677216,
+      "learning_rate": 3.927275084507791e-05,
+      "loss": 0.4448,
+      "mean_token_accuracy": 0.8282947540283203,
+      "step": 376
+    },
+    {
+      "epoch": 1.1855345911949686,
+      "grad_norm": 0.19597028195858002,
+      "learning_rate": 3.9267582512747214e-05,
+      "loss": 0.4535,
+      "mean_token_accuracy": 0.825032114982605,
+      "step": 377
+    },
+    {
+      "epoch": 1.1886792452830188,
+      "grad_norm": 0.23007996380329132,
+      "learning_rate": 3.926239626188112e-05,
+      "loss": 0.4504,
+      "mean_token_accuracy": 0.8239325284957886,
+      "step": 378
+    },
+    {
+      "epoch": 1.1918238993710693,
+      "grad_norm": 0.15653976798057556,
+      "learning_rate": 3.925719209786139e-05,
+      "loss": 0.4599,
+      "mean_token_accuracy": 0.8246617317199707,
+      "step": 379
+    },
+    {
+      "epoch": 1.1949685534591195,
+      "grad_norm": 0.21206559240818024,
+      "learning_rate": 3.925197002608837e-05,
+      "loss": 0.4627,
+      "mean_token_accuracy": 0.8227834701538086,
+      "step": 380
+    },
+    {
+      "epoch": 1.1981132075471699,
+      "grad_norm": 0.15478071570396423,
+      "learning_rate": 3.924673005198099e-05,
+      "loss": 0.4491,
+      "mean_token_accuracy": 0.8279451727867126,
+      "step": 381
+    },
+    {
+      "epoch": 1.20125786163522,
+      "grad_norm": 0.19152425229549408,
+      "learning_rate": 3.924147218097678e-05,
+      "loss": 0.4454,
+      "mean_token_accuracy": 0.8282057046890259,
+      "step": 382
+    },
+    {
+      "epoch": 1.2044025157232705,
+      "grad_norm": 0.18045583367347717,
+      "learning_rate": 3.923619641853179e-05,
+      "loss": 0.4457,
+      "mean_token_accuracy": 0.8302314877510071,
+      "step": 383
+    },
+    {
+      "epoch": 1.2075471698113207,
+      "grad_norm": 0.17766040563583374,
+      "learning_rate": 3.9230902770120705e-05,
+      "loss": 0.4428,
+      "mean_token_accuracy": 0.8286284804344177,
+      "step": 384
+    },
+    {
+      "epoch": 1.2106918238993711,
+      "grad_norm": 0.21487727761268616,
+      "learning_rate": 3.9225591241236706e-05,
+      "loss": 0.4427,
+      "mean_token_accuracy": 0.829619824886322,
+      "step": 385
+    },
+    {
+      "epoch": 1.2138364779874213,
+      "grad_norm": 0.1838655024766922,
+      "learning_rate": 3.922026183739156e-05,
+      "loss": 0.4597,
+      "mean_token_accuracy": 0.823946475982666,
+      "step": 386
+    },
+    {
+      "epoch": 1.2169811320754718,
+      "grad_norm": 0.19127146899700165,
+      "learning_rate": 3.921491456411559e-05,
+      "loss": 0.45,
+      "mean_token_accuracy": 0.8261735439300537,
+      "step": 387
+    },
+    {
+      "epoch": 1.220125786163522,
+      "grad_norm": 0.22842063009738922,
+      "learning_rate": 3.920954942695764e-05,
+      "loss": 0.4478,
+      "mean_token_accuracy": 0.8252214789390564,
+      "step": 388
+    },
+    {
+      "epoch": 1.2232704402515724,
+      "grad_norm": 0.15844713151454926,
+      "learning_rate": 3.9204166431485116e-05,
+      "loss": 0.4488,
+      "mean_token_accuracy": 0.8267233967781067,
+      "step": 389
+    },
+    {
+      "epoch": 1.2264150943396226,
+      "grad_norm": 0.19568133354187012,
+      "learning_rate": 3.9198765583283915e-05,
+      "loss": 0.4478,
+      "mean_token_accuracy": 0.8291775584220886,
+      "step": 390
+    },
+    {
+      "epoch": 1.229559748427673,
+      "grad_norm": 0.1693885773420334,
+      "learning_rate": 3.919334688795851e-05,
+      "loss": 0.4669,
+      "mean_token_accuracy": 0.8229762315750122,
+      "step": 391
+    },
+    {
+      "epoch": 1.2327044025157232,
+      "grad_norm": 0.18789900839328766,
+      "learning_rate": 3.918791035113187e-05,
+      "loss": 0.4546,
+      "mean_token_accuracy": 0.8253647685050964,
+      "step": 392
+    },
+    {
+      "epoch": 1.2358490566037736,
+      "grad_norm": 0.16432693600654602,
+      "learning_rate": 3.918245597844546e-05,
+      "loss": 0.4563,
+      "mean_token_accuracy": 0.8253912329673767,
+      "step": 393
+    },
+    {
+      "epoch": 1.2389937106918238,
+      "grad_norm": 0.1900831013917923,
+      "learning_rate": 3.9176983775559285e-05,
+      "loss": 0.4435,
+      "mean_token_accuracy": 0.8277944326400757,
+      "step": 394
+    },
+    {
+      "epoch": 1.2421383647798743,
+      "grad_norm": 0.17624549567699432,
+      "learning_rate": 3.9171493748151836e-05,
+      "loss": 0.4594,
+      "mean_token_accuracy": 0.8259245753288269,
+      "step": 395
+    },
+    {
+      "epoch": 1.2452830188679245,
+      "grad_norm": 0.1721552014350891,
+      "learning_rate": 3.91659859019201e-05,
+      "loss": 0.464,
+      "mean_token_accuracy": 0.8246681094169617,
+      "step": 396
+    },
+    {
+      "epoch": 1.248427672955975,
+      "grad_norm": 0.17541447281837463,
+      "learning_rate": 3.916046024257957e-05,
+      "loss": 0.4527,
+      "mean_token_accuracy": 0.8268889784812927,
+      "step": 397
+    },
+    {
+      "epoch": 1.251572327044025,
+      "grad_norm": 0.17931130528450012,
+      "learning_rate": 3.91549167758642e-05,
+      "loss": 0.4416,
+      "mean_token_accuracy": 0.8300866484642029,
+      "step": 398
+    },
+    {
+      "epoch": 1.2547169811320755,
+      "grad_norm": 0.17346803843975067,
+      "learning_rate": 3.914935550752643e-05,
+      "loss": 0.4532,
+      "mean_token_accuracy": 0.825555145740509,
+      "step": 399
+    },
+    {
+      "epoch": 1.2578616352201257,
+      "grad_norm": 0.17272527515888214,
+      "learning_rate": 3.91437764433372e-05,
+      "loss": 0.4552,
+      "mean_token_accuracy": 0.8265715837478638,
+      "step": 400
+    },
+    {
+      "epoch": 1.2610062893081762,
+      "grad_norm": 0.18587706983089447,
+      "learning_rate": 3.913817958908587e-05,
+      "loss": 0.4512,
+      "mean_token_accuracy": 0.8276833295822144,
+      "step": 401
+    },
+    {
+      "epoch": 1.2641509433962264,
+      "grad_norm": 0.1864849030971527,
+      "learning_rate": 3.9132564950580286e-05,
+      "loss": 0.4439,
+      "mean_token_accuracy": 0.8282519578933716,
+      "step": 402
+    },
+    {
+      "epoch": 1.2672955974842768,
+      "grad_norm": 0.19797006249427795,
+      "learning_rate": 3.9126932533646756e-05,
+      "loss": 0.4452,
+      "mean_token_accuracy": 0.8268986344337463,
+      "step": 403
+    },
+    {
+      "epoch": 1.270440251572327,
+      "grad_norm": 0.2109120786190033,
+      "learning_rate": 3.912128234413002e-05,
+      "loss": 0.4516,
+      "mean_token_accuracy": 0.8287456631660461,
+      "step": 404
+    },
+    {
+      "epoch": 1.2735849056603774,
+      "grad_norm": 0.20502309501171112,
+      "learning_rate": 3.9115614387893284e-05,
+      "loss": 0.4585,
+      "mean_token_accuracy": 0.8249683380126953,
+      "step": 405
+    },
+    {
+      "epoch": 1.2767295597484276,
+      "grad_norm": 0.195379376411438,
+      "learning_rate": 3.910992867081815e-05,
+      "loss": 0.4478,
+      "mean_token_accuracy": 0.8271308541297913,
+      "step": 406
+    },
+    {
+      "epoch": 1.279874213836478,
+      "grad_norm": 0.19706301391124725,
+      "learning_rate": 3.9104225198804697e-05,
+      "loss": 0.4538,
+      "mean_token_accuracy": 0.8262473344802856,
+      "step": 407
+    },
+    {
+      "epoch": 1.2830188679245282,
+      "grad_norm": 0.22555312514305115,
+      "learning_rate": 3.90985039777714e-05,
+      "loss": 0.4541,
+      "mean_token_accuracy": 0.8253132104873657,
+      "step": 408
+    },
+    {
+      "epoch": 1.2861635220125787,
+      "grad_norm": 0.1870194524526596,
+      "learning_rate": 3.909276501365515e-05,
+      "loss": 0.4496,
+      "mean_token_accuracy": 0.8271132111549377,
+      "step": 409
+    },
+    {
+      "epoch": 1.2893081761006289,
+      "grad_norm": 0.20130424201488495,
+      "learning_rate": 3.9087008312411266e-05,
+      "loss": 0.4508,
+      "mean_token_accuracy": 0.8274027705192566,
+      "step": 410
+    },
+    {
+      "epoch": 1.2924528301886793,
+      "grad_norm": 0.19089150428771973,
+      "learning_rate": 3.908123388001347e-05,
+      "loss": 0.4437,
+      "mean_token_accuracy": 0.8283129930496216,
+      "step": 411
+    },
+    {
+      "epoch": 1.2955974842767295,
+      "grad_norm": 0.18513773381710052,
+      "learning_rate": 3.907544172245386e-05,
+      "loss": 0.4512,
+      "mean_token_accuracy": 0.8256914615631104,
+      "step": 412
+    },
+    {
+      "epoch": 1.29874213836478,
+      "grad_norm": 0.2461722195148468,
+      "learning_rate": 3.906963184574297e-05,
+      "loss": 0.4398,
+      "mean_token_accuracy": 0.8291003108024597,
+      "step": 413
+    },
+    {
+      "epoch": 1.3018867924528301,
+      "grad_norm": 0.16066083312034607,
+      "learning_rate": 3.906380425590969e-05,
+      "loss": 0.4622,
+      "mean_token_accuracy": 0.8245102167129517,
+      "step": 414
+    },
+    {
+      "epoch": 1.3050314465408805,
+      "grad_norm": 0.23007777333259583,
+      "learning_rate": 3.905795895900129e-05,
+      "loss": 0.4508,
+      "mean_token_accuracy": 0.8270963430404663,
+      "step": 415
+    },
+    {
+      "epoch": 1.3081761006289307,
+      "grad_norm": 0.1997261345386505,
+      "learning_rate": 3.905209596108342e-05,
+      "loss": 0.4428,
+      "mean_token_accuracy": 0.8287739753723145,
+      "step": 416
+    },
+    {
+      "epoch": 1.3113207547169812,
+      "grad_norm": 0.19289380311965942,
+      "learning_rate": 3.904621526824011e-05,
+      "loss": 0.4494,
+      "mean_token_accuracy": 0.8278332948684692,
+      "step": 417
+    },
+    {
+      "epoch": 1.3144654088050314,
+      "grad_norm": 0.20662692189216614,
+      "learning_rate": 3.904031688657375e-05,
+      "loss": 0.4481,
+      "mean_token_accuracy": 0.828018844127655,
+      "step": 418
+    },
+    {
+      "epoch": 1.3176100628930818,
+      "grad_norm": 0.17998790740966797,
+      "learning_rate": 3.903440082220506e-05,
+      "loss": 0.4627,
+      "mean_token_accuracy": 0.8233857154846191,
+      "step": 419
+    },
+    {
+      "epoch": 1.320754716981132,
+      "grad_norm": 0.18025898933410645,
+      "learning_rate": 3.902846708127315e-05,
+      "loss": 0.4435,
+      "mean_token_accuracy": 0.8296754956245422,
+      "step": 420
+    },
+    {
+      "epoch": 1.3238993710691824,
+      "grad_norm": 0.1748049110174179,
+      "learning_rate": 3.902251566993543e-05,
+      "loss": 0.4504,
+      "mean_token_accuracy": 0.8265761137008667,
+      "step": 421
+    },
+    {
+      "epoch": 1.3270440251572326,
+      "grad_norm": 0.1689056158065796,
+      "learning_rate": 3.901654659436768e-05,
+      "loss": 0.4475,
+      "mean_token_accuracy": 0.8279913067817688,
+      "step": 422
+    },
+    {
+      "epoch": 1.330188679245283,
+      "grad_norm": 0.18498939275741577,
+      "learning_rate": 3.901055986076399e-05,
+      "loss": 0.4612,
+      "mean_token_accuracy": 0.8243342638015747,
+      "step": 423
+    },
+    {
+      "epoch": 1.3333333333333333,
+      "grad_norm": 0.20584383606910706,
+      "learning_rate": 3.900455547533679e-05,
+      "loss": 0.4561,
+      "mean_token_accuracy": 0.8269695043563843,
+      "step": 424
+    },
+    {
+      "epoch": 1.3364779874213837,
+      "grad_norm": 0.19053617119789124,
+      "learning_rate": 3.899853344431681e-05,
+      "loss": 0.4503,
+      "mean_token_accuracy": 0.8270792961120605,
+      "step": 425
+    },
+    {
+      "epoch": 1.3396226415094339,
+      "grad_norm": 0.23964080214500427,
+      "learning_rate": 3.8992493773953103e-05,
+      "loss": 0.444,
+      "mean_token_accuracy": 0.8272242546081543,
+      "step": 426
+    },
+    {
+      "epoch": 1.3427672955974843,
+      "grad_norm": 0.17868532240390778,
+      "learning_rate": 3.898643647051303e-05,
+      "loss": 0.4478,
+      "mean_token_accuracy": 0.8284726738929749,
+      "step": 427
+    },
+    {
+      "epoch": 1.3459119496855345,
+      "grad_norm": 0.2097913771867752,
+      "learning_rate": 3.8980361540282226e-05,
+      "loss": 0.449,
+      "mean_token_accuracy": 0.8262530565261841,
+      "step": 428
+    },
+    {
+      "epoch": 1.349056603773585,
+      "grad_norm": 0.1811225712299347,
+      "learning_rate": 3.8974268989564655e-05,
+      "loss": 0.4539,
+      "mean_token_accuracy": 0.8266646265983582,
+      "step": 429
+    },
+    {
+      "epoch": 1.3522012578616351,
+      "grad_norm": 0.1976926028728485,
+      "learning_rate": 3.896815882468253e-05,
+      "loss": 0.4537,
+      "mean_token_accuracy": 0.8261896967887878,
+      "step": 430
+    },
+    {
+      "epoch": 1.3553459119496856,
+      "grad_norm": 0.17476551234722137,
+      "learning_rate": 3.8962031051976356e-05,
+      "loss": 0.45,
+      "mean_token_accuracy": 0.8286691904067993,
+      "step": 431
+    },
+    {
+      "epoch": 1.3584905660377358,
+      "grad_norm": 0.1752653270959854,
+      "learning_rate": 3.895588567780491e-05,
+      "loss": 0.4391,
+      "mean_token_accuracy": 0.830578088760376,
+      "step": 432
+    },
+    {
+      "epoch": 1.3616352201257862,
+      "grad_norm": 0.22124166786670685,
+      "learning_rate": 3.894972270854525e-05,
+      "loss": 0.4442,
+      "mean_token_accuracy": 0.8274651169776917,
+      "step": 433
+    },
+    {
+      "epoch": 1.3647798742138364,
+      "grad_norm": 0.16567222774028778,
+      "learning_rate": 3.894354215059265e-05,
+      "loss": 0.4586,
+      "mean_token_accuracy": 0.8255594968795776,
+      "step": 434
+    },
+    {
+      "epoch": 1.3679245283018868,
+      "grad_norm": 0.2491973489522934,
+      "learning_rate": 3.893734401036069e-05,
+      "loss": 0.4567,
+      "mean_token_accuracy": 0.825359046459198,
+      "step": 435
+    },
+    {
+      "epoch": 1.371069182389937,
+      "grad_norm": 0.16659407317638397,
+      "learning_rate": 3.8931128294281154e-05,
+      "loss": 0.4487,
+      "mean_token_accuracy": 0.8270349502563477,
+      "step": 436
+    },
+    {
+      "epoch": 1.3742138364779874,
+      "grad_norm": 0.21300457417964935,
+      "learning_rate": 3.892489500880408e-05,
+      "loss": 0.4584,
+      "mean_token_accuracy": 0.8247126936912537,
+      "step": 437
+    },
+    {
+      "epoch": 1.3773584905660377,
+      "grad_norm": 0.22345586121082306,
+      "learning_rate": 3.891864416039772e-05,
+      "loss": 0.4463,
+      "mean_token_accuracy": 0.8268424868583679,
+      "step": 438
+    },
+    {
+      "epoch": 1.380503144654088,
+      "grad_norm": 0.18328611552715302,
+      "learning_rate": 3.89123757555486e-05,
+      "loss": 0.441,
+      "mean_token_accuracy": 0.8294317126274109,
+      "step": 439
+    },
+    {
+      "epoch": 1.3836477987421385,
+      "grad_norm": 0.21173825860023499,
+      "learning_rate": 3.89060898007614e-05,
+      "loss": 0.4532,
+      "mean_token_accuracy": 0.826548159122467,
+      "step": 440
+    },
+    {
+      "epoch": 1.3867924528301887,
+      "grad_norm": 0.14746391773223877,
+      "learning_rate": 3.889978630255907e-05,
+      "loss": 0.4555,
+      "mean_token_accuracy": 0.826274037361145,
+      "step": 441
+    },
+    {
+      "epoch": 1.389937106918239,
+      "grad_norm": 0.2072405368089676,
+      "learning_rate": 3.8893465267482705e-05,
+      "loss": 0.4512,
+      "mean_token_accuracy": 0.8269615173339844,
+      "step": 442
+    },
+    {
+      "epoch": 1.3930817610062893,
+      "grad_norm": 0.18459253013134003,
+      "learning_rate": 3.8887126702091665e-05,
+      "loss": 0.4499,
+      "mean_token_accuracy": 0.8274534940719604,
+      "step": 443
+    },
+    {
+      "epoch": 1.3962264150943398,
+      "grad_norm": 0.1969769448041916,
+      "learning_rate": 3.8880770612963436e-05,
+      "loss": 0.4515,
+      "mean_token_accuracy": 0.8273787498474121,
+      "step": 444
+    },
+    {
+      "epoch": 1.39937106918239,
+      "grad_norm": 0.1519605964422226,
+      "learning_rate": 3.887439700669373e-05,
+      "loss": 0.4438,
+      "mean_token_accuracy": 0.8282424211502075,
+      "step": 445
+    },
+    {
+      "epoch": 1.4025157232704402,
+      "grad_norm": 0.2099246382713318,
+      "learning_rate": 3.8868005889896434e-05,
+      "loss": 0.4606,
+      "mean_token_accuracy": 0.8256029486656189,
+      "step": 446
+    },
+    {
+      "epoch": 1.4056603773584906,
+      "grad_norm": 0.16457189619541168,
+      "learning_rate": 3.886159726920359e-05,
+      "loss": 0.4593,
+      "mean_token_accuracy": 0.8245981335639954,
+      "step": 447
+    },
+    {
+      "epoch": 1.408805031446541,
+      "grad_norm": 0.2020101696252823,
+      "learning_rate": 3.8855171151265415e-05,
+      "loss": 0.4628,
+      "mean_token_accuracy": 0.8243246674537659,
+      "step": 448
+    },
+    {
+      "epoch": 1.4119496855345912,
+      "grad_norm": 0.1827785223722458,
+      "learning_rate": 3.884872754275027e-05,
+      "loss": 0.462,
+      "mean_token_accuracy": 0.8254520297050476,
+      "step": 449
+    },
+    {
+      "epoch": 1.4150943396226414,
+      "grad_norm": 0.1720219850540161,
+      "learning_rate": 3.8842266450344676e-05,
+      "loss": 0.4429,
+      "mean_token_accuracy": 0.8271845579147339,
+      "step": 450
+    },
+    {
+      "epoch": 1.4182389937106918,
+      "grad_norm": 0.1863139569759369,
+      "learning_rate": 3.8835787880753305e-05,
+      "loss": 0.4505,
+      "mean_token_accuracy": 0.8274635076522827,
+      "step": 451
+    },
+    {
+      "epoch": 1.4213836477987423,
+      "grad_norm": 0.16976246237754822,
+      "learning_rate": 3.882929184069894e-05,
+      "loss": 0.4666,
+      "mean_token_accuracy": 0.8241984844207764,
+      "step": 452
+    },
+    {
+      "epoch": 1.4245283018867925,
+      "grad_norm": 0.1592787802219391,
+      "learning_rate": 3.882277833692253e-05,
+      "loss": 0.4547,
+      "mean_token_accuracy": 0.8265584111213684,
+      "step": 453
+    },
+    {
+      "epoch": 1.4276729559748427,
+      "grad_norm": 0.18003453314304352,
+      "learning_rate": 3.8816247376183105e-05,
+      "loss": 0.456,
+      "mean_token_accuracy": 0.8237389922142029,
+      "step": 454
+    },
+    {
+      "epoch": 1.430817610062893,
+      "grad_norm": 0.15752705931663513,
+      "learning_rate": 3.880969896525784e-05,
+      "loss": 0.4535,
+      "mean_token_accuracy": 0.8278123140335083,
+      "step": 455
+    },
+    {
+      "epoch": 1.4339622641509435,
+      "grad_norm": 0.15204060077667236,
+      "learning_rate": 3.8803133110942e-05,
+      "loss": 0.4504,
+      "mean_token_accuracy": 0.8270664215087891,
+      "step": 456
+    },
+    {
+      "epoch": 1.4371069182389937,
+      "grad_norm": 0.1571756899356842,
+      "learning_rate": 3.879654982004897e-05,
+      "loss": 0.4558,
+      "mean_token_accuracy": 0.8265605568885803,
+      "step": 457
+    },
+    {
+      "epoch": 1.440251572327044,
+      "grad_norm": 0.1584654152393341,
+      "learning_rate": 3.8789949099410206e-05,
+      "loss": 0.4518,
+      "mean_token_accuracy": 0.8270092606544495,
+      "step": 458
+    },
+    {
+      "epoch": 1.4433962264150944,
+      "grad_norm": 0.15152078866958618,
+      "learning_rate": 3.878333095587527e-05,
+      "loss": 0.4461,
+      "mean_token_accuracy": 0.8276041150093079,
+      "step": 459
+    },
+    {
+      "epoch": 1.4465408805031448,
+      "grad_norm": 0.15868431329727173,
+      "learning_rate": 3.87766953963118e-05,
+      "loss": 0.4438,
+      "mean_token_accuracy": 0.8289068937301636,
+      "step": 460
+    },
+    {
+      "epoch": 1.449685534591195,
+      "grad_norm": 0.14925162494182587,
+      "learning_rate": 3.8770042427605486e-05,
+      "loss": 0.4513,
+      "mean_token_accuracy": 0.8267138600349426,
+      "step": 461
+    },
+    {
+      "epoch": 1.4528301886792452,
+      "grad_norm": 0.15520182251930237,
+      "learning_rate": 3.876337205666011e-05,
+      "loss": 0.4627,
+      "mean_token_accuracy": 0.825510561466217,
+      "step": 462
+    },
+    {
+      "epoch": 1.4559748427672956,
+      "grad_norm": 0.15460386872291565,
+      "learning_rate": 3.875668429039751e-05,
+      "loss": 0.4524,
+      "mean_token_accuracy": 0.8280324935913086,
+      "step": 463
+    },
+    {
+      "epoch": 1.459119496855346,
+      "grad_norm": 0.1757625788450241,
+      "learning_rate": 3.8749979135757564e-05,
+      "loss": 0.4476,
+      "mean_token_accuracy": 0.8284898996353149,
+      "step": 464
+    },
+    {
+      "epoch": 1.4622641509433962,
+      "grad_norm": 0.17886289954185486,
+      "learning_rate": 3.874325659969819e-05,
+      "loss": 0.4525,
+      "mean_token_accuracy": 0.8280431628227234,
+      "step": 465
+    },
+    {
+      "epoch": 1.4654088050314464,
+      "grad_norm": 0.17093369364738464,
+      "learning_rate": 3.873651668919535e-05,
+      "loss": 0.461,
+      "mean_token_accuracy": 0.8263329267501831,
+      "step": 466
+    },
+    {
+      "epoch": 1.4685534591194969,
+      "grad_norm": 0.1534358263015747,
+      "learning_rate": 3.872975941124305e-05,
+      "loss": 0.4662,
+      "mean_token_accuracy": 0.8242008090019226,
+      "step": 467
+    },
+    {
+      "epoch": 1.4716981132075473,
+      "grad_norm": 0.16316460072994232,
+      "learning_rate": 3.8722984772853276e-05,
+      "loss": 0.4401,
+      "mean_token_accuracy": 0.8306291103363037,
+      "step": 468
+    },
+    {
+      "epoch": 1.4748427672955975,
+      "grad_norm": 0.14488621056079865,
+      "learning_rate": 3.8716192781056086e-05,
+      "loss": 0.4529,
+      "mean_token_accuracy": 0.8279455900192261,
+      "step": 469
+    },
+    {
+      "epoch": 1.4779874213836477,
+      "grad_norm": 0.15546061098575592,
+      "learning_rate": 3.870938344289951e-05,
+      "loss": 0.4524,
+      "mean_token_accuracy": 0.8272173404693604,
+      "step": 470
+    },
+    {
+      "epoch": 1.4811320754716981,
+      "grad_norm": 0.14502915740013123,
+      "learning_rate": 3.8702556765449564e-05,
+      "loss": 0.4365,
+      "mean_token_accuracy": 0.8309008479118347,
+      "step": 471
+    },
+    {
+      "epoch": 1.4842767295597485,
+      "grad_norm": 0.14981509745121002,
+      "learning_rate": 3.8695712755790296e-05,
+      "loss": 0.4553,
+      "mean_token_accuracy": 0.8265926241874695,
+      "step": 472
+    },
+    {
+      "epoch": 1.4874213836477987,
+      "grad_norm": 0.1419067680835724,
+      "learning_rate": 3.8688851421023724e-05,
+      "loss": 0.4467,
+      "mean_token_accuracy": 0.8291380405426025,
+      "step": 473
+    },
+    {
+      "epoch": 1.490566037735849,
+      "grad_norm": 0.14329911768436432,
+      "learning_rate": 3.868197276826983e-05,
+      "loss": 0.4404,
+      "mean_token_accuracy": 0.8298222422599792,
+      "step": 474
+    },
+    {
+      "epoch": 1.4937106918238994,
+      "grad_norm": 0.14086921513080597,
+      "learning_rate": 3.8675076804666574e-05,
+      "loss": 0.4517,
+      "mean_token_accuracy": 0.8272147178649902,
+      "step": 475
+    },
+    {
+      "epoch": 1.4968553459119498,
+      "grad_norm": 0.13164976239204407,
+      "learning_rate": 3.86681635373699e-05,
+      "loss": 0.4543,
+      "mean_token_accuracy": 0.8258587121963501,
+      "step": 476
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 0.14340752363204956,
+      "learning_rate": 3.866123297355368e-05,
+      "loss": 0.4547,
+      "mean_token_accuracy": 0.8274818062782288,
+      "step": 477
+    },
+    {
+      "epoch": 1.5031446540880502,
+      "grad_norm": 0.15563100576400757,
+      "learning_rate": 3.865428512040975e-05,
+      "loss": 0.4546,
+      "mean_token_accuracy": 0.8245983123779297,
+      "step": 478
+    },
+    {
+      "epoch": 1.5062893081761006,
+      "grad_norm": 0.15186692774295807,
+      "learning_rate": 3.864731998514788e-05,
+      "loss": 0.4471,
+      "mean_token_accuracy": 0.8284518718719482,
+      "step": 479
+    },
+    {
+      "epoch": 1.509433962264151,
+      "grad_norm": 0.1551363319158554,
+      "learning_rate": 3.864033757499578e-05,
+      "loss": 0.4506,
+      "mean_token_accuracy": 0.8288803696632385,
+      "step": 480
+    },
+    {
+      "epoch": 1.5125786163522013,
+      "grad_norm": 0.19744457304477692,
+      "learning_rate": 3.863333789719908e-05,
+      "loss": 0.4459,
+      "mean_token_accuracy": 0.8297019600868225,
+      "step": 481
+    },
+    {
+      "epoch": 1.5157232704402515,
+      "grad_norm": 0.1606452614068985,
+      "learning_rate": 3.8626320959021336e-05,
+      "loss": 0.4574,
+      "mean_token_accuracy": 0.8262441754341125,
+      "step": 482
+    },
+    {
+      "epoch": 1.5188679245283019,
+      "grad_norm": 0.14951243996620178,
+      "learning_rate": 3.8619286767744e-05,
+      "loss": 0.4503,
+      "mean_token_accuracy": 0.8295007348060608,
+      "step": 483
+    },
+    {
+      "epoch": 1.5220125786163523,
+      "grad_norm": 0.1706402450799942,
+      "learning_rate": 3.8612235330666455e-05,
+      "loss": 0.4522,
+      "mean_token_accuracy": 0.8276010751724243,
+      "step": 484
+    },
+    {
+      "epoch": 1.5251572327044025,
+      "grad_norm": 0.18848121166229248,
+      "learning_rate": 3.860516665510595e-05,
+      "loss": 0.4478,
+      "mean_token_accuracy": 0.8272507190704346,
+      "step": 485
+    },
+    {
+      "epoch": 1.5283018867924527,
+      "grad_norm": 0.18110370635986328,
+      "learning_rate": 3.859808074839764e-05,
+      "loss": 0.4633,
+      "mean_token_accuracy": 0.8253755569458008,
+      "step": 486
+    },
+    {
+      "epoch": 1.5314465408805031,
+      "grad_norm": 0.15408791601657867,
+      "learning_rate": 3.859097761789455e-05,
+      "loss": 0.4608,
+      "mean_token_accuracy": 0.8244016766548157,
+      "step": 487
+    },
+    {
+      "epoch": 1.5345911949685536,
+      "grad_norm": 0.209492027759552,
+      "learning_rate": 3.858385727096759e-05,
+      "loss": 0.4536,
+      "mean_token_accuracy": 0.8256571292877197,
+      "step": 488
+    },
+    {
+      "epoch": 1.5377358490566038,
+      "grad_norm": 0.13831724226474762,
+      "learning_rate": 3.8576719715005534e-05,
+      "loss": 0.4471,
+      "mean_token_accuracy": 0.8289450407028198,
+      "step": 489
+    },
+    {
+      "epoch": 1.540880503144654,
+      "grad_norm": 0.1807355135679245,
+      "learning_rate": 3.856956495741501e-05,
+      "loss": 0.457,
+      "mean_token_accuracy": 0.8249189853668213,
+      "step": 490
+    },
+    {
+      "epoch": 1.5440251572327044,
+      "grad_norm": 0.14947283267974854,
+      "learning_rate": 3.856239300562047e-05,
+      "loss": 0.4517,
+      "mean_token_accuracy": 0.827312171459198,
+      "step": 491
+    },
+    {
+      "epoch": 1.5471698113207548,
+      "grad_norm": 0.1528450846672058,
+      "learning_rate": 3.855520386706427e-05,
+      "loss": 0.4469,
+      "mean_token_accuracy": 0.8281989097595215,
+      "step": 492
+    },
+    {
+      "epoch": 1.550314465408805,
+      "grad_norm": 0.1903686374425888,
+      "learning_rate": 3.854799754920654e-05,
+      "loss": 0.4517,
+      "mean_token_accuracy": 0.8267768025398254,
+      "step": 493
+    },
+    {
+      "epoch": 1.5534591194968552,
+      "grad_norm": 0.14876671135425568,
+      "learning_rate": 3.854077405952527e-05,
+      "loss": 0.4471,
+      "mean_token_accuracy": 0.8272234797477722,
+      "step": 494
+    },
+    {
+      "epoch": 1.5566037735849056,
+      "grad_norm": 0.15654216706752777,
+      "learning_rate": 3.853353340551626e-05,
+      "loss": 0.4426,
+      "mean_token_accuracy": 0.8292421698570251,
+      "step": 495
+    },
+    {
+      "epoch": 1.559748427672956,
+      "grad_norm": 0.15252597630023956,
+      "learning_rate": 3.852627559469313e-05,
+      "loss": 0.4466,
+      "mean_token_accuracy": 0.828421413898468,
+      "step": 496
+    },
+    {
+      "epoch": 1.5628930817610063,
+      "grad_norm": 0.15407709777355194,
+      "learning_rate": 3.8519000634587274e-05,
+      "loss": 0.4522,
+      "mean_token_accuracy": 0.8288334608078003,
+      "step": 497
+    },
+    {
+      "epoch": 1.5660377358490565,
+      "grad_norm": 0.15139752626419067,
+      "learning_rate": 3.851170853274793e-05,
+      "loss": 0.4591,
+      "mean_token_accuracy": 0.8242106437683105,
+      "step": 498
+    },
+    {
+      "epoch": 1.569182389937107,
+      "grad_norm": 0.16278305649757385,
+      "learning_rate": 3.8504399296742076e-05,
+      "loss": 0.4536,
+      "mean_token_accuracy": 0.8283582329750061,
+      "step": 499
+    },
+    {
+      "epoch": 1.5723270440251573,
+      "grad_norm": 0.21534794569015503,
+      "learning_rate": 3.84970729341545e-05,
+      "loss": 0.4464,
+      "mean_token_accuracy": 0.8293619155883789,
+      "step": 500
+    },
+    {
+      "epoch": 1.5754716981132075,
+      "grad_norm": 0.16945937275886536,
+      "learning_rate": 3.848972945258776e-05,
+      "loss": 0.4366,
+      "mean_token_accuracy": 0.8301461338996887,
+      "step": 501
+    },
+    {
+      "epoch": 1.5786163522012577,
+      "grad_norm": 0.19320879876613617,
+      "learning_rate": 3.8482368859662156e-05,
+      "loss": 0.4546,
+      "mean_token_accuracy": 0.8255838751792908,
+      "step": 502
+    },
+    {
+      "epoch": 1.5817610062893082,
+      "grad_norm": 0.16316726803779602,
+      "learning_rate": 3.847499116301577e-05,
+      "loss": 0.4535,
+      "mean_token_accuracy": 0.8292089104652405,
+      "step": 503
+    },
+    {
+      "epoch": 1.5849056603773586,
+      "grad_norm": 0.20331525802612305,
+      "learning_rate": 3.846759637030443e-05,
+      "loss": 0.4455,
+      "mean_token_accuracy": 0.8304234743118286,
+      "step": 504
+    },
+    {
+      "epoch": 1.5880503144654088,
+      "grad_norm": 0.14327698945999146,
+      "learning_rate": 3.846018448920168e-05,
+      "loss": 0.4388,
+      "mean_token_accuracy": 0.8320124745368958,
+      "step": 505
+    },
+    {
+      "epoch": 1.591194968553459,
+      "grad_norm": 0.1536184698343277,
+      "learning_rate": 3.845275552739883e-05,
+      "loss": 0.4611,
+      "mean_token_accuracy": 0.8233290910720825,
+      "step": 506
+    },
+    {
+      "epoch": 1.5943396226415094,
+      "grad_norm": 0.17109455168247223,
+      "learning_rate": 3.844530949260489e-05,
+      "loss": 0.4581,
+      "mean_token_accuracy": 0.8267108798027039,
+      "step": 507
+    },
+    {
+      "epoch": 1.5974842767295598,
+      "grad_norm": 0.1569495052099228,
+      "learning_rate": 3.8437846392546603e-05,
+      "loss": 0.454,
+      "mean_token_accuracy": 0.8257220387458801,
+      "step": 508
+    },
+    {
+      "epoch": 1.60062893081761,
+      "grad_norm": 0.15163108706474304,
+      "learning_rate": 3.84303662349684e-05,
+      "loss": 0.4459,
+      "mean_token_accuracy": 0.8274716734886169,
+      "step": 509
+    },
+    {
+      "epoch": 1.6037735849056602,
+      "grad_norm": 0.15992477536201477,
+      "learning_rate": 3.842286902763245e-05,
+      "loss": 0.4506,
+      "mean_token_accuracy": 0.8289050459861755,
+      "step": 510
+    },
+    {
+      "epoch": 1.6069182389937107,
+      "grad_norm": 0.16461218893527985,
+      "learning_rate": 3.841535477831855e-05,
+      "loss": 0.4523,
+      "mean_token_accuracy": 0.8275653123855591,
+      "step": 511
+    },
+    {
+      "epoch": 1.610062893081761,
+      "grad_norm": 0.15126198530197144,
+      "learning_rate": 3.840782349482426e-05,
+      "loss": 0.4442,
+      "mean_token_accuracy": 0.8301685452461243,
+      "step": 512
+    },
+    {
+      "epoch": 1.6132075471698113,
+      "grad_norm": 0.19573630392551422,
+      "learning_rate": 3.840027518496475e-05,
+      "loss": 0.4575,
+      "mean_token_accuracy": 0.8253160715103149,
+      "step": 513
+    },
+    {
+      "epoch": 1.6163522012578615,
+      "grad_norm": 0.16733072698116302,
+      "learning_rate": 3.8392709856572904e-05,
+      "loss": 0.4481,
+      "mean_token_accuracy": 0.8281794190406799,
+      "step": 514
+    },
+    {
+      "epoch": 1.619496855345912,
+      "grad_norm": 0.1560571938753128,
+      "learning_rate": 3.838512751749924e-05,
+      "loss": 0.4472,
+      "mean_token_accuracy": 0.8281599879264832,
+      "step": 515
+    },
+    {
+      "epoch": 1.6226415094339623,
+      "grad_norm": 0.17163477838039398,
+      "learning_rate": 3.837752817561194e-05,
+      "loss": 0.45,
+      "mean_token_accuracy": 0.8275222778320312,
+      "step": 516
+    },
+    {
+      "epoch": 1.6257861635220126,
+      "grad_norm": 0.16268257796764374,
+      "learning_rate": 3.8369911838796816e-05,
+      "loss": 0.4483,
+      "mean_token_accuracy": 0.8267813920974731,
+      "step": 517
+    },
+    {
+      "epoch": 1.6289308176100628,
+      "grad_norm": 0.16738560795783997,
+      "learning_rate": 3.8362278514957336e-05,
+      "loss": 0.4375,
+      "mean_token_accuracy": 0.8297300934791565,
+      "step": 518
+    },
+    {
+      "epoch": 1.6320754716981132,
+      "grad_norm": 0.16072408854961395,
+      "learning_rate": 3.8354628212014587e-05,
+      "loss": 0.4417,
+      "mean_token_accuracy": 0.8280481100082397,
+      "step": 519
+    },
+    {
+      "epoch": 1.6352201257861636,
+      "grad_norm": 0.17322075366973877,
+      "learning_rate": 3.8346960937907264e-05,
+      "loss": 0.4569,
+      "mean_token_accuracy": 0.8253117799758911,
+      "step": 520
+    },
+    {
+      "epoch": 1.6383647798742138,
+      "grad_norm": 0.1540379822254181,
+      "learning_rate": 3.833927670059168e-05,
+      "loss": 0.4352,
+      "mean_token_accuracy": 0.8304820656776428,
+      "step": 521
+    },
+    {
+      "epoch": 1.641509433962264,
+      "grad_norm": 0.20093458890914917,
+      "learning_rate": 3.833157550804176e-05,
+      "loss": 0.4502,
+      "mean_token_accuracy": 0.8276907801628113,
+      "step": 522
+    },
+    {
+      "epoch": 1.6446540880503144,
+      "grad_norm": 0.17706024646759033,
+      "learning_rate": 3.8323857368249014e-05,
+      "loss": 0.4472,
+      "mean_token_accuracy": 0.8285207748413086,
+      "step": 523
+    },
+    {
+      "epoch": 1.6477987421383649,
+      "grad_norm": 0.2173389196395874,
+      "learning_rate": 3.8316122289222535e-05,
+      "loss": 0.4482,
+      "mean_token_accuracy": 0.8268563747406006,
+      "step": 524
+    },
+    {
+      "epoch": 1.650943396226415,
+      "grad_norm": 0.14905473589897156,
+      "learning_rate": 3.8308370278989e-05,
+      "loss": 0.445,
+      "mean_token_accuracy": 0.8302363157272339,
+      "step": 525
+    },
+    {
+      "epoch": 1.6540880503144653,
+      "grad_norm": 0.1895207017660141,
+      "learning_rate": 3.8300601345592675e-05,
+      "loss": 0.455,
+      "mean_token_accuracy": 0.8297334909439087,
+      "step": 526
+    },
+    {
+      "epoch": 1.6572327044025157,
+      "grad_norm": 0.190410777926445,
+      "learning_rate": 3.829281549709533e-05,
+      "loss": 0.4456,
+      "mean_token_accuracy": 0.8273657560348511,
+      "step": 527
+    },
+    {
+      "epoch": 1.6603773584905661,
+      "grad_norm": 0.15508361160755157,
+      "learning_rate": 3.828501274157635e-05,
+      "loss": 0.4444,
+      "mean_token_accuracy": 0.8293198347091675,
+      "step": 528
+    },
+    {
+      "epoch": 1.6635220125786163,
+      "grad_norm": 0.17064864933490753,
+      "learning_rate": 3.8277193087132634e-05,
+      "loss": 0.4534,
+      "mean_token_accuracy": 0.8277645707130432,
+      "step": 529
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "grad_norm": 0.18299253284931183,
+      "learning_rate": 3.826935654187861e-05,
+      "loss": 0.4529,
+      "mean_token_accuracy": 0.8259576559066772,
+      "step": 530
+    },
+    {
+      "epoch": 1.669811320754717,
+      "grad_norm": 0.14436650276184082,
+      "learning_rate": 3.826150311394627e-05,
+      "loss": 0.4424,
+      "mean_token_accuracy": 0.8301630020141602,
+      "step": 531
+    },
+    {
+      "epoch": 1.6729559748427674,
+      "grad_norm": 0.16234935820102692,
+      "learning_rate": 3.825363281148507e-05,
+      "loss": 0.4414,
+      "mean_token_accuracy": 0.8282803297042847,
+      "step": 532
+    },
+    {
+      "epoch": 1.6761006289308176,
+      "grad_norm": 0.1424507051706314,
+      "learning_rate": 3.8245745642662025e-05,
+      "loss": 0.45,
+      "mean_token_accuracy": 0.8285747170448303,
+      "step": 533
+    },
+    {
+      "epoch": 1.6792452830188678,
+      "grad_norm": 0.17243961989879608,
+      "learning_rate": 3.8237841615661636e-05,
+      "loss": 0.4519,
+      "mean_token_accuracy": 0.8274285197257996,
+      "step": 534
+    },
+    {
+      "epoch": 1.6823899371069182,
+      "grad_norm": 0.15060077607631683,
+      "learning_rate": 3.8229920738685886e-05,
+      "loss": 0.4485,
+      "mean_token_accuracy": 0.827806293964386,
+      "step": 535
+    },
+    {
+      "epoch": 1.6855345911949686,
+      "grad_norm": 0.16962994635105133,
+      "learning_rate": 3.8221983019954254e-05,
+      "loss": 0.4526,
+      "mean_token_accuracy": 0.8264437317848206,
+      "step": 536
+    },
+    {
+      "epoch": 1.6886792452830188,
+      "grad_norm": 0.16717560589313507,
+      "learning_rate": 3.82140284677037e-05,
+      "loss": 0.4453,
+      "mean_token_accuracy": 0.8285322785377502,
+      "step": 537
+    },
+    {
+      "epoch": 1.691823899371069,
+      "grad_norm": 0.15772481262683868,
+      "learning_rate": 3.820605709018865e-05,
+      "loss": 0.4506,
+      "mean_token_accuracy": 0.8271482586860657,
+      "step": 538
+    },
+    {
+      "epoch": 1.6949685534591195,
+      "grad_norm": 0.17806661128997803,
+      "learning_rate": 3.819806889568098e-05,
+      "loss": 0.4482,
+      "mean_token_accuracy": 0.8278267979621887,
+      "step": 539
+    },
+    {
+      "epoch": 1.6981132075471699,
+      "grad_norm": 0.18837103247642517,
+      "learning_rate": 3.819006389247002e-05,
+      "loss": 0.4491,
+      "mean_token_accuracy": 0.8294506669044495,
+      "step": 540
+    },
+    {
+      "epoch": 1.70125786163522,
+      "grad_norm": 0.18263404071331024,
+      "learning_rate": 3.8182042088862555e-05,
+      "loss": 0.4549,
+      "mean_token_accuracy": 0.8264862895011902,
+      "step": 541
+    },
+    {
+      "epoch": 1.7044025157232703,
+      "grad_norm": 0.17201568186283112,
+      "learning_rate": 3.8174003493182784e-05,
+      "loss": 0.4479,
+      "mean_token_accuracy": 0.826499879360199,
+      "step": 542
+    },
+    {
+      "epoch": 1.7075471698113207,
+      "grad_norm": 0.2106884866952896,
+      "learning_rate": 3.816594811377235e-05,
+      "loss": 0.448,
+      "mean_token_accuracy": 0.8292680382728577,
+      "step": 543
+    },
+    {
+      "epoch": 1.7106918238993711,
+      "grad_norm": 0.17805208265781403,
+      "learning_rate": 3.81578759589903e-05,
+      "loss": 0.451,
+      "mean_token_accuracy": 0.8258050680160522,
+      "step": 544
+    },
+    {
+      "epoch": 1.7138364779874213,
+      "grad_norm": 0.20148631930351257,
+      "learning_rate": 3.814978703721309e-05,
+      "loss": 0.4491,
+      "mean_token_accuracy": 0.8281079530715942,
+      "step": 545
+    },
+    {
+      "epoch": 1.7169811320754715,
+      "grad_norm": 0.2121734768152237,
+      "learning_rate": 3.8141681356834587e-05,
+      "loss": 0.4503,
+      "mean_token_accuracy": 0.8274951577186584,
+      "step": 546
+    },
+    {
+      "epoch": 1.720125786163522,
+      "grad_norm": 0.16927511990070343,
+      "learning_rate": 3.813355892626603e-05,
+      "loss": 0.4358,
+      "mean_token_accuracy": 0.8312507271766663,
+      "step": 547
+    },
+    {
+      "epoch": 1.7232704402515724,
+      "grad_norm": 0.2374507337808609,
+      "learning_rate": 3.8125419753936055e-05,
+      "loss": 0.4433,
+      "mean_token_accuracy": 0.8288567662239075,
+      "step": 548
+    },
+    {
+      "epoch": 1.7264150943396226,
+      "grad_norm": 0.14739219844341278,
+      "learning_rate": 3.8117263848290656e-05,
+      "loss": 0.4432,
+      "mean_token_accuracy": 0.8277235627174377,
+      "step": 549
+    },
+    {
+      "epoch": 1.7295597484276728,
+      "grad_norm": 0.22958379983901978,
+      "learning_rate": 3.810909121779321e-05,
+      "loss": 0.4522,
+      "mean_token_accuracy": 0.8266805410385132,
+      "step": 550
+    },
+    {
+      "epoch": 1.7327044025157232,
+      "grad_norm": 0.1478036642074585,
+      "learning_rate": 3.810090187092443e-05,
+      "loss": 0.4362,
+      "mean_token_accuracy": 0.8309704065322876,
+      "step": 551
+    },
+    {
+      "epoch": 1.7358490566037736,
+      "grad_norm": 0.2060670554637909,
+      "learning_rate": 3.809269581618239e-05,
+      "loss": 0.4463,
+      "mean_token_accuracy": 0.8310072422027588,
+      "step": 552
+    },
+    {
+      "epoch": 1.7389937106918238,
+      "grad_norm": 0.157624289393425,
+      "learning_rate": 3.80844730620825e-05,
+      "loss": 0.4489,
+      "mean_token_accuracy": 0.82773357629776,
+      "step": 553
+    },
+    {
+      "epoch": 1.742138364779874,
+      "grad_norm": 0.18612465262413025,
+      "learning_rate": 3.8076233617157486e-05,
+      "loss": 0.4506,
+      "mean_token_accuracy": 0.8275358080863953,
+      "step": 554
+    },
+    {
+      "epoch": 1.7452830188679245,
+      "grad_norm": 0.15709009766578674,
+      "learning_rate": 3.806797748995741e-05,
+      "loss": 0.452,
+      "mean_token_accuracy": 0.829462468624115,
+      "step": 555
+    },
+    {
+      "epoch": 1.748427672955975,
+      "grad_norm": 0.17328409850597382,
+      "learning_rate": 3.805970468904964e-05,
+      "loss": 0.4439,
+      "mean_token_accuracy": 0.8290309906005859,
+      "step": 556
+    },
+    {
+      "epoch": 1.751572327044025,
+      "grad_norm": 0.17372116446495056,
+      "learning_rate": 3.805141522301884e-05,
+      "loss": 0.4524,
+      "mean_token_accuracy": 0.8271963000297546,
+      "step": 557
+    },
+    {
+      "epoch": 1.7547169811320755,
+      "grad_norm": 0.161929190158844,
+      "learning_rate": 3.804310910046697e-05,
+      "loss": 0.4399,
+      "mean_token_accuracy": 0.8296563029289246,
+      "step": 558
+    },
+    {
+      "epoch": 1.757861635220126,
+      "grad_norm": 0.1848653256893158,
+      "learning_rate": 3.803478633001328e-05,
+      "loss": 0.4576,
+      "mean_token_accuracy": 0.8277556896209717,
+      "step": 559
+    },
+    {
+      "epoch": 1.7610062893081762,
+      "grad_norm": 0.1677953600883484,
+      "learning_rate": 3.8026446920294295e-05,
+      "loss": 0.4554,
+      "mean_token_accuracy": 0.8266874551773071,
+      "step": 560
+    },
+    {
+      "epoch": 1.7641509433962264,
+      "grad_norm": 0.15664350986480713,
+      "learning_rate": 3.80180908799638e-05,
+      "loss": 0.4499,
+      "mean_token_accuracy": 0.8271031379699707,
+      "step": 561
+    },
+    {
+      "epoch": 1.7672955974842768,
+      "grad_norm": 0.1716042309999466,
+      "learning_rate": 3.800971821769284e-05,
+      "loss": 0.4477,
+      "mean_token_accuracy": 0.8271938562393188,
+      "step": 562
+    },
+    {
+      "epoch": 1.7704402515723272,
+      "grad_norm": 0.15244047343730927,
+      "learning_rate": 3.800132894216971e-05,
+      "loss": 0.4373,
+      "mean_token_accuracy": 0.8322244882583618,
+      "step": 563
+    },
+    {
+      "epoch": 1.7735849056603774,
+      "grad_norm": 0.1561097949743271,
+      "learning_rate": 3.799292306209995e-05,
+      "loss": 0.4342,
+      "mean_token_accuracy": 0.8296236395835876,
+      "step": 564
+    },
+    {
+      "epoch": 1.7767295597484276,
+      "grad_norm": 0.13890111446380615,
+      "learning_rate": 3.7984500586206324e-05,
+      "loss": 0.4552,
+      "mean_token_accuracy": 0.825447678565979,
+      "step": 565
+    },
+    {
+      "epoch": 1.779874213836478,
+      "grad_norm": 0.16650985181331635,
+      "learning_rate": 3.79760615232288e-05,
+      "loss": 0.4474,
+      "mean_token_accuracy": 0.8286757469177246,
+      "step": 566
+    },
+    {
+      "epoch": 1.7830188679245285,
+      "grad_norm": 0.14276254177093506,
+      "learning_rate": 3.7967605881924605e-05,
+      "loss": 0.4608,
+      "mean_token_accuracy": 0.8258764147758484,
+      "step": 567
+    },
+    {
+      "epoch": 1.7861635220125787,
+      "grad_norm": 0.13229092955589294,
+      "learning_rate": 3.795913367106812e-05,
+      "loss": 0.4489,
+      "mean_token_accuracy": 0.8294503688812256,
+      "step": 568
+    },
+    {
+      "epoch": 1.7893081761006289,
+      "grad_norm": 0.15833866596221924,
+      "learning_rate": 3.795064489945095e-05,
+      "loss": 0.4442,
+      "mean_token_accuracy": 0.829393744468689,
+      "step": 569
+    },
+    {
+      "epoch": 1.7924528301886793,
+      "grad_norm": 0.14272266626358032,
+      "learning_rate": 3.7942139575881875e-05,
+      "loss": 0.4488,
+      "mean_token_accuracy": 0.8297043442726135,
+      "step": 570
+    },
+    {
+      "epoch": 1.7955974842767297,
+      "grad_norm": 0.15338532626628876,
+      "learning_rate": 3.7933617709186845e-05,
+      "loss": 0.4637,
+      "mean_token_accuracy": 0.8241843581199646,
+      "step": 571
+    },
+    {
+      "epoch": 1.79874213836478,
+      "grad_norm": 0.16218815743923187,
+      "learning_rate": 3.7925079308209e-05,
+      "loss": 0.4581,
+      "mean_token_accuracy": 0.8253941535949707,
+      "step": 572
+    },
+    {
+      "epoch": 1.8018867924528301,
+      "grad_norm": 0.1723911464214325,
+      "learning_rate": 3.7916524381808606e-05,
+      "loss": 0.4485,
+      "mean_token_accuracy": 0.8298059701919556,
+      "step": 573
+    },
+    {
+      "epoch": 1.8050314465408805,
+      "grad_norm": 0.1315835565328598,
+      "learning_rate": 3.7907952938863095e-05,
+      "loss": 0.4409,
+      "mean_token_accuracy": 0.8307161331176758,
+      "step": 574
+    },
+    {
+      "epoch": 1.808176100628931,
+      "grad_norm": 0.1815989911556244,
+      "learning_rate": 3.7899364988267045e-05,
+      "loss": 0.4533,
+      "mean_token_accuracy": 0.8261933326721191,
+      "step": 575
+    },
+    {
+      "epoch": 1.8113207547169812,
+      "grad_norm": 0.13026003539562225,
+      "learning_rate": 3.789076053893214e-05,
+      "loss": 0.4425,
+      "mean_token_accuracy": 0.8303501009941101,
+      "step": 576
+    },
+    {
+      "epoch": 1.8144654088050314,
+      "grad_norm": 0.15672063827514648,
+      "learning_rate": 3.788213959978722e-05,
+      "loss": 0.4492,
+      "mean_token_accuracy": 0.8289505839347839,
+      "step": 577
+    },
+    {
+      "epoch": 1.8176100628930818,
+      "grad_norm": 0.1376017928123474,
+      "learning_rate": 3.7873502179778204e-05,
+      "loss": 0.4531,
+      "mean_token_accuracy": 0.8275585174560547,
+      "step": 578
+    },
+    {
+      "epoch": 1.8207547169811322,
+      "grad_norm": 0.16440635919570923,
+      "learning_rate": 3.786484828786812e-05,
+      "loss": 0.4357,
+      "mean_token_accuracy": 0.8302841782569885,
+      "step": 579
+    },
+    {
+      "epoch": 1.8238993710691824,
+      "grad_norm": 0.15636101365089417,
+      "learning_rate": 3.78561779330371e-05,
+      "loss": 0.4612,
+      "mean_token_accuracy": 0.8249790072441101,
+      "step": 580
+    },
+    {
+      "epoch": 1.8270440251572326,
+      "grad_norm": 0.14285589754581451,
+      "learning_rate": 3.7847491124282354e-05,
+      "loss": 0.4456,
+      "mean_token_accuracy": 0.828575611114502,
+      "step": 581
+    },
+    {
+      "epoch": 1.830188679245283,
+      "grad_norm": 0.16128715872764587,
+      "learning_rate": 3.783878787061817e-05,
+      "loss": 0.4499,
+      "mean_token_accuracy": 0.8283258676528931,
+      "step": 582
+    },
+    {
+      "epoch": 1.8333333333333335,
+      "grad_norm": 0.14683718979358673,
+      "learning_rate": 3.783006818107589e-05,
+      "loss": 0.4493,
+      "mean_token_accuracy": 0.8294415473937988,
+      "step": 583
+    },
+    {
+      "epoch": 1.8364779874213837,
+      "grad_norm": 0.17366930842399597,
+      "learning_rate": 3.782133206470392e-05,
+      "loss": 0.4569,
+      "mean_token_accuracy": 0.8255544900894165,
+      "step": 584
+    },
+    {
+      "epoch": 1.8396226415094339,
+      "grad_norm": 0.18530210852622986,
+      "learning_rate": 3.781257953056771e-05,
+      "loss": 0.4469,
+      "mean_token_accuracy": 0.8293200731277466,
+      "step": 585
+    },
+    {
+      "epoch": 1.8427672955974843,
+      "grad_norm": 0.1758362352848053,
+      "learning_rate": 3.780381058774975e-05,
+      "loss": 0.4444,
+      "mean_token_accuracy": 0.8289125561714172,
+      "step": 586
+    },
+    {
+      "epoch": 1.8459119496855347,
+      "grad_norm": 0.17143899202346802,
+      "learning_rate": 3.779502524534955e-05,
+      "loss": 0.4437,
+      "mean_token_accuracy": 0.8296340703964233,
+      "step": 587
+    },
+    {
+      "epoch": 1.849056603773585,
+      "grad_norm": 0.15893438458442688,
+      "learning_rate": 3.778622351248365e-05,
+      "loss": 0.461,
+      "mean_token_accuracy": 0.8271942734718323,
+      "step": 588
+    },
+    {
+      "epoch": 1.8522012578616351,
+      "grad_norm": 0.15634989738464355,
+      "learning_rate": 3.777740539828558e-05,
+      "loss": 0.4646,
+      "mean_token_accuracy": 0.8230259418487549,
+      "step": 589
+    },
+    {
+      "epoch": 1.8553459119496856,
+      "grad_norm": 0.15147466957569122,
+      "learning_rate": 3.776857091190588e-05,
+      "loss": 0.4628,
+      "mean_token_accuracy": 0.8259610533714294,
+      "step": 590
+    },
+    {
+      "epoch": 1.858490566037736,
+      "grad_norm": 0.1858360767364502,
+      "learning_rate": 3.775972006251209e-05,
+      "loss": 0.4608,
+      "mean_token_accuracy": 0.8260266184806824,
+      "step": 591
+    },
+    {
+      "epoch": 1.8616352201257862,
+      "grad_norm": 0.1573314070701599,
+      "learning_rate": 3.77508528592887e-05,
+      "loss": 0.4638,
+      "mean_token_accuracy": 0.8259332180023193,
+      "step": 592
+    },
+    {
+      "epoch": 1.8647798742138364,
+      "grad_norm": 0.1616523116827011,
+      "learning_rate": 3.7741969311437203e-05,
+      "loss": 0.4354,
+      "mean_token_accuracy": 0.8312195539474487,
+      "step": 593
+    },
+    {
+      "epoch": 1.8679245283018868,
+      "grad_norm": 0.161224365234375,
+      "learning_rate": 3.7733069428176044e-05,
+      "loss": 0.4606,
+      "mean_token_accuracy": 0.8260555267333984,
+      "step": 594
+    },
+    {
+      "epoch": 1.8710691823899372,
+      "grad_norm": 0.14999954402446747,
+      "learning_rate": 3.772415321874059e-05,
+      "loss": 0.4526,
+      "mean_token_accuracy": 0.829216480255127,
+      "step": 595
+    },
+    {
+      "epoch": 1.8742138364779874,
+      "grad_norm": 0.13627400994300842,
+      "learning_rate": 3.7715220692383206e-05,
+      "loss": 0.4451,
+      "mean_token_accuracy": 0.8285694718360901,
+      "step": 596
+    },
+    {
+      "epoch": 1.8773584905660377,
+      "grad_norm": 0.1767715960741043,
+      "learning_rate": 3.770627185837313e-05,
+      "loss": 0.4477,
+      "mean_token_accuracy": 0.8299650549888611,
+      "step": 597
+    },
+    {
+      "epoch": 1.880503144654088,
+      "grad_norm": 0.17853030562400818,
+      "learning_rate": 3.769730672599655e-05,
+      "loss": 0.4455,
+      "mean_token_accuracy": 0.8283320665359497,
+      "step": 598
+    },
+    {
+      "epoch": 1.8836477987421385,
+      "grad_norm": 0.1425485908985138,
+      "learning_rate": 3.768832530455658e-05,
+      "loss": 0.4537,
+      "mean_token_accuracy": 0.8282998204231262,
+      "step": 599
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.14554312825202942,
+      "learning_rate": 3.7679327603373224e-05,
+      "loss": 0.4365,
+      "mean_token_accuracy": 0.830973744392395,
+      "step": 600
+    },
+    {
+      "epoch": 1.889937106918239,
+      "grad_norm": 0.14549976587295532,
+      "learning_rate": 3.767031363178336e-05,
+      "loss": 0.4476,
+      "mean_token_accuracy": 0.8286392092704773,
+      "step": 601
+    },
+    {
+      "epoch": 1.8930817610062893,
+      "grad_norm": 0.1506306231021881,
+      "learning_rate": 3.766128339914079e-05,
+      "loss": 0.4344,
+      "mean_token_accuracy": 0.8309949636459351,
+      "step": 602
+    },
+    {
+      "epoch": 1.8962264150943398,
+      "grad_norm": 0.18175862729549408,
+      "learning_rate": 3.765223691481615e-05,
+      "loss": 0.4414,
+      "mean_token_accuracy": 0.8330844044685364,
+      "step": 603
+    },
+    {
+      "epoch": 1.89937106918239,
+      "grad_norm": 0.14774861931800842,
+      "learning_rate": 3.764317418819696e-05,
+      "loss": 0.447,
+      "mean_token_accuracy": 0.8281767964363098,
+      "step": 604
+    },
+    {
+      "epoch": 1.9025157232704402,
+      "grad_norm": 0.2164774239063263,
+      "learning_rate": 3.7634095228687606e-05,
+      "loss": 0.449,
+      "mean_token_accuracy": 0.827027440071106,
+      "step": 605
+    },
+    {
+      "epoch": 1.9056603773584906,
+      "grad_norm": 0.1532975137233734,
+      "learning_rate": 3.762500004570929e-05,
+      "loss": 0.439,
+      "mean_token_accuracy": 0.8315529823303223,
+      "step": 606
+    },
+    {
+      "epoch": 1.908805031446541,
+      "grad_norm": 0.17712901532649994,
+      "learning_rate": 3.761588864870009e-05,
+      "loss": 0.4586,
+      "mean_token_accuracy": 0.8272659182548523,
+      "step": 607
+    },
+    {
+      "epoch": 1.9119496855345912,
+      "grad_norm": 0.1497354805469513,
+      "learning_rate": 3.760676104711485e-05,
+      "loss": 0.4487,
+      "mean_token_accuracy": 0.8286811709403992,
+      "step": 608
+    },
+    {
+      "epoch": 1.9150943396226414,
+      "grad_norm": 0.16662414371967316,
+      "learning_rate": 3.759761725042529e-05,
+      "loss": 0.4597,
+      "mean_token_accuracy": 0.8276802897453308,
+      "step": 609
+    },
+    {
+      "epoch": 1.9182389937106918,
+      "grad_norm": 0.17103828489780426,
+      "learning_rate": 3.75884572681199e-05,
+      "loss": 0.4465,
+      "mean_token_accuracy": 0.8278771638870239,
+      "step": 610
+    },
+    {
+      "epoch": 1.9213836477987423,
+      "grad_norm": 0.18412946164608002,
+      "learning_rate": 3.7579281109703976e-05,
+      "loss": 0.4396,
+      "mean_token_accuracy": 0.8283287882804871,
+      "step": 611
+    },
+    {
+      "epoch": 1.9245283018867925,
+      "grad_norm": 0.21794135868549347,
+      "learning_rate": 3.757008878469959e-05,
+      "loss": 0.4531,
+      "mean_token_accuracy": 0.8257644772529602,
+      "step": 612
+    },
+    {
+      "epoch": 1.9276729559748427,
+      "grad_norm": 0.16859892010688782,
+      "learning_rate": 3.756088030264562e-05,
+      "loss": 0.4477,
+      "mean_token_accuracy": 0.8286134600639343,
+      "step": 613
+    },
+    {
+      "epoch": 1.930817610062893,
+      "grad_norm": 0.16001489758491516,
+      "learning_rate": 3.7551655673097664e-05,
+      "loss": 0.4477,
+      "mean_token_accuracy": 0.8258501887321472,
+      "step": 614
+    },
+    {
+      "epoch": 1.9339622641509435,
+      "grad_norm": 0.16016210615634918,
+      "learning_rate": 3.7542414905628125e-05,
+      "loss": 0.4464,
+      "mean_token_accuracy": 0.8278330564498901,
+      "step": 615
+    },
+    {
+      "epoch": 1.9371069182389937,
+      "grad_norm": 0.16711488366127014,
+      "learning_rate": 3.753315800982611e-05,
+      "loss": 0.4529,
+      "mean_token_accuracy": 0.8275772333145142,
+      "step": 616
+    },
+    {
+      "epoch": 1.940251572327044,
+      "grad_norm": 0.14468665421009064,
+      "learning_rate": 3.75238849952975e-05,
+      "loss": 0.4498,
+      "mean_token_accuracy": 0.8276384472846985,
+      "step": 617
+    },
+    {
+      "epoch": 1.9433962264150944,
+      "grad_norm": 0.16502094268798828,
+      "learning_rate": 3.751459587166486e-05,
+      "loss": 0.4418,
+      "mean_token_accuracy": 0.8288312554359436,
+      "step": 618
+    },
+    {
+      "epoch": 1.9465408805031448,
+      "grad_norm": 0.15773820877075195,
+      "learning_rate": 3.750529064856752e-05,
+      "loss": 0.4467,
+      "mean_token_accuracy": 0.8294534683227539,
+      "step": 619
+    },
+    {
+      "epoch": 1.949685534591195,
+      "grad_norm": 0.15135358273983002,
+      "learning_rate": 3.749596933566146e-05,
+      "loss": 0.4547,
+      "mean_token_accuracy": 0.826626181602478,
+      "step": 620
+    },
+    {
+      "epoch": 1.9528301886792452,
+      "grad_norm": 0.14081326127052307,
+      "learning_rate": 3.748663194261942e-05,
+      "loss": 0.4491,
+      "mean_token_accuracy": 0.8284742832183838,
+      "step": 621
+    },
+    {
+      "epoch": 1.9559748427672956,
+      "grad_norm": 0.14700694382190704,
+      "learning_rate": 3.7477278479130774e-05,
+      "loss": 0.4505,
+      "mean_token_accuracy": 0.828678548336029,
+      "step": 622
+    },
+    {
+      "epoch": 1.959119496855346,
+      "grad_norm": 0.14334739744663239,
+      "learning_rate": 3.7467908954901607e-05,
+      "loss": 0.4515,
+      "mean_token_accuracy": 0.8276963829994202,
+      "step": 623
+    },
+    {
+      "epoch": 1.9622641509433962,
+      "grad_norm": 0.13543659448623657,
+      "learning_rate": 3.745852337965463e-05,
+      "loss": 0.4372,
+      "mean_token_accuracy": 0.8290265202522278,
+      "step": 624
+    },
+    {
+      "epoch": 1.9654088050314464,
+      "grad_norm": 0.13767287135124207,
+      "learning_rate": 3.744912176312926e-05,
+      "loss": 0.4536,
+      "mean_token_accuracy": 0.829569935798645,
+      "step": 625
+    },
+    {
+      "epoch": 1.9685534591194969,
+      "grad_norm": 0.14214669167995453,
+      "learning_rate": 3.743970411508153e-05,
+      "loss": 0.4492,
+      "mean_token_accuracy": 0.8299821019172668,
+      "step": 626
+    },
+    {
+      "epoch": 1.9716981132075473,
+      "grad_norm": 0.1446889191865921,
+      "learning_rate": 3.7430270445284113e-05,
+      "loss": 0.4507,
+      "mean_token_accuracy": 0.8302611112594604,
+      "step": 627
+    },
+    {
+      "epoch": 1.9748427672955975,
+      "grad_norm": 0.1604827344417572,
+      "learning_rate": 3.74208207635263e-05,
+      "loss": 0.4503,
+      "mean_token_accuracy": 0.8277794122695923,
+      "step": 628
+    },
+    {
+      "epoch": 1.9779874213836477,
+      "grad_norm": 0.15117383003234863,
+      "learning_rate": 3.741135507961402e-05,
+      "loss": 0.4459,
+      "mean_token_accuracy": 0.8283612132072449,
+      "step": 629
+    },
+    {
+      "epoch": 1.9811320754716981,
+      "grad_norm": 0.152221217751503,
+      "learning_rate": 3.740187340336978e-05,
+      "loss": 0.4501,
+      "mean_token_accuracy": 0.8274828791618347,
+      "step": 630
+    },
+    {
+      "epoch": 1.9842767295597485,
+      "grad_norm": 0.15416257083415985,
+      "learning_rate": 3.7392375744632705e-05,
+      "loss": 0.4314,
+      "mean_token_accuracy": 0.8337308764457703,
+      "step": 631
+    },
+    {
+      "epoch": 1.9874213836477987,
+      "grad_norm": 0.15964354574680328,
+      "learning_rate": 3.73828621132585e-05,
+      "loss": 0.4477,
+      "mean_token_accuracy": 0.8281906843185425,
+      "step": 632
+    },
+    {
+      "epoch": 1.990566037735849,
+      "grad_norm": 0.176462784409523,
+      "learning_rate": 3.737333251911943e-05,
+      "loss": 0.458,
+      "mean_token_accuracy": 0.8277231454849243,
+      "step": 633
+    },
+    {
+      "epoch": 1.9937106918238994,
+      "grad_norm": 0.14236986637115479,
+      "learning_rate": 3.736378697210435e-05,
+      "loss": 0.4433,
+      "mean_token_accuracy": 0.830397367477417,
+      "step": 634
+    },
+    {
+      "epoch": 1.9968553459119498,
+      "grad_norm": 0.20474541187286377,
+      "learning_rate": 3.735422548211865e-05,
+      "loss": 0.4443,
+      "mean_token_accuracy": 0.8304517269134521,
+      "step": 635
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.1430058777332306,
+      "learning_rate": 3.7344648059084254e-05,
+      "loss": 0.4409,
+      "mean_token_accuracy": 0.8248613476753235,
+      "step": 636
+    },
+    {
+      "epoch": 2.00314465408805,
+      "grad_norm": 0.2259976714849472,
+      "learning_rate": 3.733505471293965e-05,
+      "loss": 0.3875,
+      "mean_token_accuracy": 0.8429651260375977,
+      "step": 637
+    },
+    {
+      "epoch": 2.006289308176101,
+      "grad_norm": 0.7415010929107666,
+      "learning_rate": 3.732544545363983e-05,
+      "loss": 0.396,
+      "mean_token_accuracy": 0.8418735861778259,
+      "step": 638
+    },
+    {
+      "epoch": 2.009433962264151,
+      "grad_norm": 0.20141229033470154,
+      "learning_rate": 3.73158202911563e-05,
+      "loss": 0.3975,
+      "mean_token_accuracy": 0.8419850468635559,
+      "step": 639
+    },
+    {
+      "epoch": 2.0125786163522013,
+      "grad_norm": 0.29176196455955505,
+      "learning_rate": 3.730617923547708e-05,
+      "loss": 0.38,
+      "mean_token_accuracy": 0.8438968062400818,
+      "step": 640
+    },
+    {
+      "epoch": 2.0157232704402515,
+      "grad_norm": 0.19412219524383545,
+      "learning_rate": 3.7296522296606663e-05,
+      "loss": 0.3857,
+      "mean_token_accuracy": 0.843908965587616,
+      "step": 641
+    },
+    {
+      "epoch": 2.018867924528302,
+      "grad_norm": 0.33550146222114563,
+      "learning_rate": 3.7286849484566046e-05,
+      "loss": 0.3931,
+      "mean_token_accuracy": 0.8432913422584534,
+      "step": 642
+    },
+    {
+      "epoch": 2.0220125786163523,
+      "grad_norm": 0.2596316635608673,
+      "learning_rate": 3.727716080939268e-05,
+      "loss": 0.3962,
+      "mean_token_accuracy": 0.841339111328125,
+      "step": 643
+    },
+    {
+      "epoch": 2.0251572327044025,
+      "grad_norm": 0.2917231321334839,
+      "learning_rate": 3.726745628114048e-05,
+      "loss": 0.3957,
+      "mean_token_accuracy": 0.8408085703849792,
+      "step": 644
+    },
+    {
+      "epoch": 2.0283018867924527,
+      "grad_norm": 0.22424006462097168,
+      "learning_rate": 3.725773590987982e-05,
+      "loss": 0.3824,
+      "mean_token_accuracy": 0.8433759808540344,
+      "step": 645
+    },
+    {
+      "epoch": 2.0314465408805034,
+      "grad_norm": 0.31289535760879517,
+      "learning_rate": 3.7247999705697515e-05,
+      "loss": 0.3867,
+      "mean_token_accuracy": 0.8449568152427673,
+      "step": 646
+    },
+    {
+      "epoch": 2.0345911949685536,
+      "grad_norm": 0.28442052006721497,
+      "learning_rate": 3.723824767869679e-05,
+      "loss": 0.4006,
+      "mean_token_accuracy": 0.8402043581008911,
+      "step": 647
+    },
+    {
+      "epoch": 2.0377358490566038,
+      "grad_norm": 0.282670795917511,
+      "learning_rate": 3.722847983899732e-05,
+      "loss": 0.388,
+      "mean_token_accuracy": 0.8409184217453003,
+      "step": 648
+    },
+    {
+      "epoch": 2.040880503144654,
+      "grad_norm": 0.26842135190963745,
+      "learning_rate": 3.7218696196735165e-05,
+      "loss": 0.3722,
+      "mean_token_accuracy": 0.8447781801223755,
+      "step": 649
+    },
+    {
+      "epoch": 2.0440251572327046,
+      "grad_norm": 0.22568707168102264,
+      "learning_rate": 3.720889676206279e-05,
+      "loss": 0.3783,
+      "mean_token_accuracy": 0.8424438834190369,
+      "step": 650
+    },
+    {
+      "epoch": 2.047169811320755,
+      "grad_norm": 0.23483210802078247,
+      "learning_rate": 3.719908154514904e-05,
+      "loss": 0.4051,
+      "mean_token_accuracy": 0.8413841128349304,
+      "step": 651
+    },
+    {
+      "epoch": 2.050314465408805,
+      "grad_norm": 0.19200192391872406,
+      "learning_rate": 3.7189250556179156e-05,
+      "loss": 0.391,
+      "mean_token_accuracy": 0.8443261384963989,
+      "step": 652
+    },
+    {
+      "epoch": 2.0534591194968552,
+      "grad_norm": 0.2197944074869156,
+      "learning_rate": 3.717940380535474e-05,
+      "loss": 0.3733,
+      "mean_token_accuracy": 0.8451419472694397,
+      "step": 653
+    },
+    {
+      "epoch": 2.056603773584906,
+      "grad_norm": 0.15395909547805786,
+      "learning_rate": 3.716954130289374e-05,
+      "loss": 0.396,
+      "mean_token_accuracy": 0.8405272364616394,
+      "step": 654
+    },
+    {
+      "epoch": 2.059748427672956,
+      "grad_norm": 0.2504603862762451,
+      "learning_rate": 3.7159663059030446e-05,
+      "loss": 0.3819,
+      "mean_token_accuracy": 0.8421579599380493,
+      "step": 655
+    },
+    {
+      "epoch": 2.0628930817610063,
+      "grad_norm": 0.16238710284233093,
+      "learning_rate": 3.7149769084015514e-05,
+      "loss": 0.3874,
+      "mean_token_accuracy": 0.8433874249458313,
+      "step": 656
+    },
+    {
+      "epoch": 2.0660377358490565,
+      "grad_norm": 0.25545817613601685,
+      "learning_rate": 3.713985938811588e-05,
+      "loss": 0.3916,
+      "mean_token_accuracy": 0.841235876083374,
+      "step": 657
+    },
+    {
+      "epoch": 2.069182389937107,
+      "grad_norm": 0.2000405341386795,
+      "learning_rate": 3.7129933981614816e-05,
+      "loss": 0.3717,
+      "mean_token_accuracy": 0.8454606533050537,
+      "step": 658
+    },
+    {
+      "epoch": 2.0723270440251573,
+      "grad_norm": 0.2362697422504425,
+      "learning_rate": 3.711999287481191e-05,
+      "loss": 0.3962,
+      "mean_token_accuracy": 0.840667724609375,
+      "step": 659
+    },
+    {
+      "epoch": 2.0754716981132075,
+      "grad_norm": 0.21723434329032898,
+      "learning_rate": 3.7110036078023024e-05,
+      "loss": 0.3906,
+      "mean_token_accuracy": 0.8431429266929626,
+      "step": 660
+    },
+    {
+      "epoch": 2.0786163522012577,
+      "grad_norm": 0.17892718315124512,
+      "learning_rate": 3.71000636015803e-05,
+      "loss": 0.3848,
+      "mean_token_accuracy": 0.8448073863983154,
+      "step": 661
+    },
+    {
+      "epoch": 2.0817610062893084,
+      "grad_norm": 0.18332098424434662,
+      "learning_rate": 3.709007545583217e-05,
+      "loss": 0.393,
+      "mean_token_accuracy": 0.8400642275810242,
+      "step": 662
+    },
+    {
+      "epoch": 2.0849056603773586,
+      "grad_norm": 0.1824181079864502,
+      "learning_rate": 3.708007165114329e-05,
+      "loss": 0.398,
+      "mean_token_accuracy": 0.8406155705451965,
+      "step": 663
+    },
+    {
+      "epoch": 2.088050314465409,
+      "grad_norm": 0.18380111455917358,
+      "learning_rate": 3.70700521978946e-05,
+      "loss": 0.3918,
+      "mean_token_accuracy": 0.8429200649261475,
+      "step": 664
+    },
+    {
+      "epoch": 2.091194968553459,
+      "grad_norm": 0.1803388148546219,
+      "learning_rate": 3.706001710648327e-05,
+      "loss": 0.4006,
+      "mean_token_accuracy": 0.8405814170837402,
+      "step": 665
+    },
+    {
+      "epoch": 2.0943396226415096,
+      "grad_norm": 0.17774419486522675,
+      "learning_rate": 3.7049966387322694e-05,
+      "loss": 0.3817,
+      "mean_token_accuracy": 0.842662513256073,
+      "step": 666
+    },
+    {
+      "epoch": 2.09748427672956,
+      "grad_norm": 0.19682133197784424,
+      "learning_rate": 3.703990005084248e-05,
+      "loss": 0.3837,
+      "mean_token_accuracy": 0.8443858027458191,
+      "step": 667
+    },
+    {
+      "epoch": 2.10062893081761,
+      "grad_norm": 0.19912737607955933,
+      "learning_rate": 3.702981810748844e-05,
+      "loss": 0.3818,
+      "mean_token_accuracy": 0.8457018136978149,
+      "step": 668
+    },
+    {
+      "epoch": 2.1037735849056602,
+      "grad_norm": 0.17018529772758484,
+      "learning_rate": 3.7019720567722595e-05,
+      "loss": 0.375,
+      "mean_token_accuracy": 0.8442913293838501,
+      "step": 669
+    },
+    {
+      "epoch": 2.106918238993711,
+      "grad_norm": 0.18541787564754486,
+      "learning_rate": 3.700960744202313e-05,
+      "loss": 0.3678,
+      "mean_token_accuracy": 0.8443611264228821,
+      "step": 670
+    },
+    {
+      "epoch": 2.110062893081761,
+      "grad_norm": 0.15378877520561218,
+      "learning_rate": 3.6999478740884435e-05,
+      "loss": 0.3852,
+      "mean_token_accuracy": 0.8438538312911987,
+      "step": 671
+    },
+    {
+      "epoch": 2.1132075471698113,
+      "grad_norm": 0.1910746991634369,
+      "learning_rate": 3.6989334474817026e-05,
+      "loss": 0.3812,
+      "mean_token_accuracy": 0.8418902158737183,
+      "step": 672
+    },
+    {
+      "epoch": 2.1163522012578615,
+      "grad_norm": 0.15623098611831665,
+      "learning_rate": 3.697917465434759e-05,
+      "loss": 0.3934,
+      "mean_token_accuracy": 0.8424828052520752,
+      "step": 673
+    },
+    {
+      "epoch": 2.119496855345912,
+      "grad_norm": 0.17167270183563232,
+      "learning_rate": 3.6968999290018953e-05,
+      "loss": 0.3905,
+      "mean_token_accuracy": 0.8438811898231506,
+      "step": 674
+    },
+    {
+      "epoch": 2.1226415094339623,
+      "grad_norm": 0.1417340338230133,
+      "learning_rate": 3.695880839239007e-05,
+      "loss": 0.3953,
+      "mean_token_accuracy": 0.8393605947494507,
+      "step": 675
+    },
+    {
+      "epoch": 2.1257861635220126,
+      "grad_norm": 0.19678044319152832,
+      "learning_rate": 3.694860197203602e-05,
+      "loss": 0.3838,
+      "mean_token_accuracy": 0.8427770137786865,
+      "step": 676
+    },
+    {
+      "epoch": 2.1289308176100628,
+      "grad_norm": 0.16106431186199188,
+      "learning_rate": 3.693838003954798e-05,
+      "loss": 0.397,
+      "mean_token_accuracy": 0.8402181267738342,
+      "step": 677
+    },
+    {
+      "epoch": 2.1320754716981134,
+      "grad_norm": 0.15794852375984192,
+      "learning_rate": 3.692814260553323e-05,
+      "loss": 0.3899,
+      "mean_token_accuracy": 0.8398270606994629,
+      "step": 678
+    },
+    {
+      "epoch": 2.1352201257861636,
+      "grad_norm": 0.1485368013381958,
+      "learning_rate": 3.691788968061512e-05,
+      "loss": 0.3834,
+      "mean_token_accuracy": 0.8433291912078857,
+      "step": 679
+    },
+    {
+      "epoch": 2.138364779874214,
+      "grad_norm": 0.16451844573020935,
+      "learning_rate": 3.690762127543312e-05,
+      "loss": 0.388,
+      "mean_token_accuracy": 0.8428193926811218,
+      "step": 680
+    },
+    {
+      "epoch": 2.141509433962264,
+      "grad_norm": 0.1383827030658722,
+      "learning_rate": 3.6897337400642706e-05,
+      "loss": 0.3863,
+      "mean_token_accuracy": 0.844539225101471,
+      "step": 681
+    },
+    {
+      "epoch": 2.1446540880503147,
+      "grad_norm": 0.16487407684326172,
+      "learning_rate": 3.688703806691545e-05,
+      "loss": 0.3806,
+      "mean_token_accuracy": 0.8434612154960632,
+      "step": 682
+    },
+    {
+      "epoch": 2.147798742138365,
+      "grad_norm": 0.16573506593704224,
+      "learning_rate": 3.6876723284938944e-05,
+      "loss": 0.3753,
+      "mean_token_accuracy": 0.8452996611595154,
+      "step": 683
+    },
+    {
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.16341538727283478,
+      "learning_rate": 3.686639306541681e-05,
+      "loss": 0.3854,
+      "mean_token_accuracy": 0.8425849676132202,
+      "step": 684
+    },
+    {
+      "epoch": 2.1540880503144653,
+      "grad_norm": 0.17833516001701355,
+      "learning_rate": 3.685604741906871e-05,
+      "loss": 0.3863,
+      "mean_token_accuracy": 0.8424407243728638,
+      "step": 685
+    },
+    {
+      "epoch": 2.157232704402516,
+      "grad_norm": 0.15904675424098969,
+      "learning_rate": 3.684568635663029e-05,
+      "loss": 0.3786,
+      "mean_token_accuracy": 0.8449070453643799,
+      "step": 686
+    },
+    {
+      "epoch": 2.160377358490566,
+      "grad_norm": 0.14655998349189758,
+      "learning_rate": 3.683530988885321e-05,
+      "loss": 0.3935,
+      "mean_token_accuracy": 0.8412083387374878,
+      "step": 687
+    },
+    {
+      "epoch": 2.1635220125786163,
+      "grad_norm": 0.16055339574813843,
+      "learning_rate": 3.6824918026505094e-05,
+      "loss": 0.3858,
+      "mean_token_accuracy": 0.8450068831443787,
+      "step": 688
+    },
+    {
+      "epoch": 2.1666666666666665,
+      "grad_norm": 0.1480380743741989,
+      "learning_rate": 3.681451078036957e-05,
+      "loss": 0.3874,
+      "mean_token_accuracy": 0.8429787158966064,
+      "step": 689
+    },
+    {
+      "epoch": 2.169811320754717,
+      "grad_norm": 0.1642938256263733,
+      "learning_rate": 3.680408816124622e-05,
+      "loss": 0.4035,
+      "mean_token_accuracy": 0.8393082022666931,
+      "step": 690
+    },
+    {
+      "epoch": 2.1729559748427674,
+      "grad_norm": 0.15204781293869019,
+      "learning_rate": 3.6793650179950574e-05,
+      "loss": 0.369,
+      "mean_token_accuracy": 0.8482940793037415,
+      "step": 691
+    },
+    {
+      "epoch": 2.1761006289308176,
+      "grad_norm": 0.18579435348510742,
+      "learning_rate": 3.67831968473141e-05,
+      "loss": 0.3856,
+      "mean_token_accuracy": 0.8452662229537964,
+      "step": 692
+    },
+    {
+      "epoch": 2.1792452830188678,
+      "grad_norm": 0.16767072677612305,
+      "learning_rate": 3.6772728174184185e-05,
+      "loss": 0.3887,
+      "mean_token_accuracy": 0.8428248167037964,
+      "step": 693
+    },
+    {
+      "epoch": 2.1823899371069184,
+      "grad_norm": 0.1605173796415329,
+      "learning_rate": 3.676224417142417e-05,
+      "loss": 0.3735,
+      "mean_token_accuracy": 0.8427917957305908,
+      "step": 694
+    },
+    {
+      "epoch": 2.1855345911949686,
+      "grad_norm": 0.15939317643642426,
+      "learning_rate": 3.675174484991328e-05,
+      "loss": 0.3782,
+      "mean_token_accuracy": 0.8429858088493347,
+      "step": 695
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 0.17312584817409515,
+      "learning_rate": 3.674123022054664e-05,
+      "loss": 0.3846,
+      "mean_token_accuracy": 0.843677818775177,
+      "step": 696
+    },
+    {
+      "epoch": 2.191823899371069,
+      "grad_norm": 0.14854241907596588,
+      "learning_rate": 3.6730700294235255e-05,
+      "loss": 0.401,
+      "mean_token_accuracy": 0.8407949209213257,
+      "step": 697
+    },
+    {
+      "epoch": 2.1949685534591197,
+      "grad_norm": 0.15429075062274933,
+      "learning_rate": 3.6720155081906004e-05,
+      "loss": 0.383,
+      "mean_token_accuracy": 0.8440051078796387,
+      "step": 698
+    },
+    {
+      "epoch": 2.19811320754717,
+      "grad_norm": 0.19384504854679108,
+      "learning_rate": 3.670959459450164e-05,
+      "loss": 0.3876,
+      "mean_token_accuracy": 0.8426343202590942,
+      "step": 699
+    },
+    {
+      "epoch": 2.20125786163522,
+      "grad_norm": 0.17279024422168732,
+      "learning_rate": 3.669901884298077e-05,
+      "loss": 0.3933,
+      "mean_token_accuracy": 0.8408414721488953,
+      "step": 700
+    },
+    {
+      "epoch": 2.2044025157232703,
+      "grad_norm": 0.14874425530433655,
+      "learning_rate": 3.6688427838317805e-05,
+      "loss": 0.3851,
+      "mean_token_accuracy": 0.843673586845398,
+      "step": 701
+    },
+    {
+      "epoch": 2.207547169811321,
+      "grad_norm": 0.1604325771331787,
+      "learning_rate": 3.667782159150302e-05,
+      "loss": 0.3934,
+      "mean_token_accuracy": 0.840538740158081,
+      "step": 702
+    },
+    {
+      "epoch": 2.210691823899371,
+      "grad_norm": 0.1530817151069641,
+      "learning_rate": 3.666720011354251e-05,
+      "loss": 0.3947,
+      "mean_token_accuracy": 0.8400828838348389,
+      "step": 703
+    },
+    {
+      "epoch": 2.2138364779874213,
+      "grad_norm": 0.14883075654506683,
+      "learning_rate": 3.6656563415458144e-05,
+      "loss": 0.3965,
+      "mean_token_accuracy": 0.8402729034423828,
+      "step": 704
+    },
+    {
+      "epoch": 2.2169811320754715,
+      "grad_norm": 0.13267891108989716,
+      "learning_rate": 3.66459115082876e-05,
+      "loss": 0.3857,
+      "mean_token_accuracy": 0.8434920310974121,
+      "step": 705
+    },
+    {
+      "epoch": 2.220125786163522,
+      "grad_norm": 0.17050376534461975,
+      "learning_rate": 3.663524440308436e-05,
+      "loss": 0.3995,
+      "mean_token_accuracy": 0.8391640186309814,
+      "step": 706
+    },
+    {
+      "epoch": 2.2232704402515724,
+      "grad_norm": 0.16033872961997986,
+      "learning_rate": 3.6624562110917634e-05,
+      "loss": 0.3791,
+      "mean_token_accuracy": 0.8431466817855835,
+      "step": 707
+    },
+    {
+      "epoch": 2.2264150943396226,
+      "grad_norm": 0.16110140085220337,
+      "learning_rate": 3.6613864642872433e-05,
+      "loss": 0.3815,
+      "mean_token_accuracy": 0.84393709897995,
+      "step": 708
+    },
+    {
+      "epoch": 2.229559748427673,
+      "grad_norm": 0.1578216254711151,
+      "learning_rate": 3.660315201004949e-05,
+      "loss": 0.3901,
+      "mean_token_accuracy": 0.8419655561447144,
+      "step": 709
+    },
+    {
+      "epoch": 2.2327044025157234,
+      "grad_norm": 0.15616242587566376,
+      "learning_rate": 3.659242422356528e-05,
+      "loss": 0.3755,
+      "mean_token_accuracy": 0.847722589969635,
+      "step": 710
+    },
+    {
+      "epoch": 2.2358490566037736,
+      "grad_norm": 0.16292689740657806,
+      "learning_rate": 3.658168129455201e-05,
+      "loss": 0.3778,
+      "mean_token_accuracy": 0.8432061076164246,
+      "step": 711
+    },
+    {
+      "epoch": 2.238993710691824,
+      "grad_norm": 0.14753198623657227,
+      "learning_rate": 3.657092323415759e-05,
+      "loss": 0.3908,
+      "mean_token_accuracy": 0.8414838314056396,
+      "step": 712
+    },
+    {
+      "epoch": 2.242138364779874,
+      "grad_norm": 0.14525368809700012,
+      "learning_rate": 3.656015005354565e-05,
+      "loss": 0.3864,
+      "mean_token_accuracy": 0.8425192832946777,
+      "step": 713
+    },
+    {
+      "epoch": 2.2452830188679247,
+      "grad_norm": 0.1643352061510086,
+      "learning_rate": 3.654936176389548e-05,
+      "loss": 0.3937,
+      "mean_token_accuracy": 0.8406062722206116,
+      "step": 714
+    },
+    {
+      "epoch": 2.248427672955975,
+      "grad_norm": 0.1485619693994522,
+      "learning_rate": 3.653855837640208e-05,
+      "loss": 0.3949,
+      "mean_token_accuracy": 0.8385857939720154,
+      "step": 715
+    },
+    {
+      "epoch": 2.251572327044025,
+      "grad_norm": 0.17640046775341034,
+      "learning_rate": 3.65277399022761e-05,
+      "loss": 0.3931,
+      "mean_token_accuracy": 0.84043288230896,
+      "step": 716
+    },
+    {
+      "epoch": 2.2547169811320753,
+      "grad_norm": 0.13140366971492767,
+      "learning_rate": 3.651690635274385e-05,
+      "loss": 0.382,
+      "mean_token_accuracy": 0.8446291089057922,
+      "step": 717
+    },
+    {
+      "epoch": 2.257861635220126,
+      "grad_norm": 0.1530969738960266,
+      "learning_rate": 3.650605773904728e-05,
+      "loss": 0.373,
+      "mean_token_accuracy": 0.8450080156326294,
+      "step": 718
+    },
+    {
+      "epoch": 2.261006289308176,
+      "grad_norm": 0.13690687716007233,
+      "learning_rate": 3.649519407244397e-05,
+      "loss": 0.3972,
+      "mean_token_accuracy": 0.8407341241836548,
+      "step": 719
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "grad_norm": 0.1574098914861679,
+      "learning_rate": 3.648431536420713e-05,
+      "loss": 0.3798,
+      "mean_token_accuracy": 0.8445830345153809,
+      "step": 720
+    },
+    {
+      "epoch": 2.2672955974842766,
+      "grad_norm": 0.1320675164461136,
+      "learning_rate": 3.6473421625625575e-05,
+      "loss": 0.3939,
+      "mean_token_accuracy": 0.8414525389671326,
+      "step": 721
+    },
+    {
+      "epoch": 2.270440251572327,
+      "grad_norm": 0.16731448471546173,
+      "learning_rate": 3.646251286800371e-05,
+      "loss": 0.3884,
+      "mean_token_accuracy": 0.8414167165756226,
+      "step": 722
+    },
+    {
+      "epoch": 2.2735849056603774,
+      "grad_norm": 0.15549425780773163,
+      "learning_rate": 3.645158910266154e-05,
+      "loss": 0.3905,
+      "mean_token_accuracy": 0.8417775630950928,
+      "step": 723
+    },
+    {
+      "epoch": 2.2767295597484276,
+      "grad_norm": 0.14840513467788696,
+      "learning_rate": 3.6440650340934625e-05,
+      "loss": 0.4033,
+      "mean_token_accuracy": 0.8384340405464172,
+      "step": 724
+    },
+    {
+      "epoch": 2.279874213836478,
+      "grad_norm": 0.1481187343597412,
+      "learning_rate": 3.64296965941741e-05,
+      "loss": 0.3943,
+      "mean_token_accuracy": 0.8407248854637146,
+      "step": 725
+    },
+    {
+      "epoch": 2.2830188679245285,
+      "grad_norm": 0.16942381858825684,
+      "learning_rate": 3.641872787374664e-05,
+      "loss": 0.3954,
+      "mean_token_accuracy": 0.841485857963562,
+      "step": 726
+    },
+    {
+      "epoch": 2.2861635220125787,
+      "grad_norm": 0.1288381665945053,
+      "learning_rate": 3.640774419103448e-05,
+      "loss": 0.3898,
+      "mean_token_accuracy": 0.8435741662979126,
+      "step": 727
+    },
+    {
+      "epoch": 2.289308176100629,
+      "grad_norm": 0.1845909059047699,
+      "learning_rate": 3.6396745557435344e-05,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.839992105960846,
+      "step": 728
+    },
+    {
+      "epoch": 2.292452830188679,
+      "grad_norm": 0.13671965897083282,
+      "learning_rate": 3.6385731984362505e-05,
+      "loss": 0.3952,
+      "mean_token_accuracy": 0.8399907946586609,
+      "step": 729
+    },
+    {
+      "epoch": 2.2955974842767297,
+      "grad_norm": 0.18472205102443695,
+      "learning_rate": 3.637470348324473e-05,
+      "loss": 0.3913,
+      "mean_token_accuracy": 0.8435697555541992,
+      "step": 730
+    },
+    {
+      "epoch": 2.29874213836478,
+      "grad_norm": 0.136347234249115,
+      "learning_rate": 3.6363660065526255e-05,
+      "loss": 0.3984,
+      "mean_token_accuracy": 0.8397014737129211,
+      "step": 731
+    },
+    {
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.17823082208633423,
+      "learning_rate": 3.635260174266682e-05,
+      "loss": 0.3909,
+      "mean_token_accuracy": 0.8430346846580505,
+      "step": 732
+    },
+    {
+      "epoch": 2.3050314465408803,
+      "grad_norm": 0.143354594707489,
+      "learning_rate": 3.634152852614163e-05,
+      "loss": 0.3936,
+      "mean_token_accuracy": 0.8386049270629883,
+      "step": 733
+    },
+    {
+      "epoch": 2.308176100628931,
+      "grad_norm": 0.14964474737644196,
+      "learning_rate": 3.633044042744134e-05,
+      "loss": 0.3879,
+      "mean_token_accuracy": 0.8436481952667236,
+      "step": 734
+    },
+    {
+      "epoch": 2.311320754716981,
+      "grad_norm": 0.14824624359607697,
+      "learning_rate": 3.631933745807204e-05,
+      "loss": 0.4019,
+      "mean_token_accuracy": 0.8401293754577637,
+      "step": 735
+    },
+    {
+      "epoch": 2.3144654088050314,
+      "grad_norm": 0.14450529217720032,
+      "learning_rate": 3.6308219629555264e-05,
+      "loss": 0.4031,
+      "mean_token_accuracy": 0.8403022289276123,
+      "step": 736
+    },
+    {
+      "epoch": 2.3176100628930816,
+      "grad_norm": 0.16056504845619202,
+      "learning_rate": 3.629708695342795e-05,
+      "loss": 0.3936,
+      "mean_token_accuracy": 0.8434531688690186,
+      "step": 737
+    },
+    {
+      "epoch": 2.3207547169811322,
+      "grad_norm": 0.1542007476091385,
+      "learning_rate": 3.628593944124247e-05,
+      "loss": 0.3994,
+      "mean_token_accuracy": 0.8412461280822754,
+      "step": 738
+    },
+    {
+      "epoch": 2.3238993710691824,
+      "grad_norm": 0.15933917462825775,
+      "learning_rate": 3.627477710456657e-05,
+      "loss": 0.3968,
+      "mean_token_accuracy": 0.8403854966163635,
+      "step": 739
+    },
+    {
+      "epoch": 2.3270440251572326,
+      "grad_norm": 0.15049493312835693,
+      "learning_rate": 3.626359995498337e-05,
+      "loss": 0.3845,
+      "mean_token_accuracy": 0.8435242176055908,
+      "step": 740
+    },
+    {
+      "epoch": 2.330188679245283,
+      "grad_norm": 0.15058250725269318,
+      "learning_rate": 3.625240800409139e-05,
+      "loss": 0.3976,
+      "mean_token_accuracy": 0.8395102024078369,
+      "step": 741
+    },
+    {
+      "epoch": 2.3333333333333335,
+      "grad_norm": 0.14439113438129425,
+      "learning_rate": 3.624120126350449e-05,
+      "loss": 0.3873,
+      "mean_token_accuracy": 0.8440976738929749,
+      "step": 742
+    },
+    {
+      "epoch": 2.3364779874213837,
+      "grad_norm": 0.14690877497196198,
+      "learning_rate": 3.6229979744851886e-05,
+      "loss": 0.3769,
+      "mean_token_accuracy": 0.8444263935089111,
+      "step": 743
+    },
+    {
+      "epoch": 2.339622641509434,
+      "grad_norm": 0.17405840754508972,
+      "learning_rate": 3.621874345977813e-05,
+      "loss": 0.4031,
+      "mean_token_accuracy": 0.8406171798706055,
+      "step": 744
+    },
+    {
+      "epoch": 2.342767295597484,
+      "grad_norm": 0.15319286286830902,
+      "learning_rate": 3.6207492419943065e-05,
+      "loss": 0.3899,
+      "mean_token_accuracy": 0.8424206972122192,
+      "step": 745
+    },
+    {
+      "epoch": 2.3459119496855347,
+      "grad_norm": 0.16251526772975922,
+      "learning_rate": 3.61962266370219e-05,
+      "loss": 0.3898,
+      "mean_token_accuracy": 0.8415105938911438,
+      "step": 746
+    },
+    {
+      "epoch": 2.349056603773585,
+      "grad_norm": 0.1394706815481186,
+      "learning_rate": 3.618494612270511e-05,
+      "loss": 0.3919,
+      "mean_token_accuracy": 0.8406856656074524,
+      "step": 747
+    },
+    {
+      "epoch": 2.352201257861635,
+      "grad_norm": 0.15003950893878937,
+      "learning_rate": 3.6173650888698456e-05,
+      "loss": 0.3984,
+      "mean_token_accuracy": 0.8404019474983215,
+      "step": 748
+    },
+    {
+      "epoch": 2.3553459119496853,
+      "grad_norm": 0.1499018669128418,
+      "learning_rate": 3.616234094672298e-05,
+      "loss": 0.3947,
+      "mean_token_accuracy": 0.8411343693733215,
+      "step": 749
+    },
+    {
+      "epoch": 2.358490566037736,
+      "grad_norm": 0.15295803546905518,
+      "learning_rate": 3.615101630851499e-05,
+      "loss": 0.3919,
+      "mean_token_accuracy": 0.8424092531204224,
+      "step": 750
+    },
+    {
+      "epoch": 2.361635220125786,
+      "grad_norm": 0.15068519115447998,
+      "learning_rate": 3.6139676985826035e-05,
+      "loss": 0.3825,
+      "mean_token_accuracy": 0.845098078250885,
+      "step": 751
+    },
+    {
+      "epoch": 2.3647798742138364,
+      "grad_norm": 0.17812252044677734,
+      "learning_rate": 3.6128322990422924e-05,
+      "loss": 0.3781,
+      "mean_token_accuracy": 0.8452845811843872,
+      "step": 752
+    },
+    {
+      "epoch": 2.3679245283018866,
+      "grad_norm": 0.1486268937587738,
+      "learning_rate": 3.6116954334087644e-05,
+      "loss": 0.3904,
+      "mean_token_accuracy": 0.8416242003440857,
+      "step": 753
+    },
+    {
+      "epoch": 2.3710691823899372,
+      "grad_norm": 0.20920655131340027,
+      "learning_rate": 3.6105571028617445e-05,
+      "loss": 0.3882,
+      "mean_token_accuracy": 0.8425738215446472,
+      "step": 754
+    },
+    {
+      "epoch": 2.3742138364779874,
+      "grad_norm": 0.14427687227725983,
+      "learning_rate": 3.609417308582477e-05,
+      "loss": 0.3859,
+      "mean_token_accuracy": 0.841761589050293,
+      "step": 755
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 0.1912706047296524,
+      "learning_rate": 3.608276051753722e-05,
+      "loss": 0.3996,
+      "mean_token_accuracy": 0.8410661816596985,
+      "step": 756
+    },
+    {
+      "epoch": 2.380503144654088,
+      "grad_norm": 0.15891054272651672,
+      "learning_rate": 3.607133333559761e-05,
+      "loss": 0.3871,
+      "mean_token_accuracy": 0.8439091444015503,
+      "step": 757
+    },
+    {
+      "epoch": 2.3836477987421385,
+      "grad_norm": 0.18664152920246124,
+      "learning_rate": 3.605989155186389e-05,
+      "loss": 0.3946,
+      "mean_token_accuracy": 0.8415076732635498,
+      "step": 758
+    },
+    {
+      "epoch": 2.3867924528301887,
+      "grad_norm": 0.19155050814151764,
+      "learning_rate": 3.6048435178209194e-05,
+      "loss": 0.3951,
+      "mean_token_accuracy": 0.8402332067489624,
+      "step": 759
+    },
+    {
+      "epoch": 2.389937106918239,
+      "grad_norm": 0.14341680705547333,
+      "learning_rate": 3.603696422652176e-05,
+      "loss": 0.3858,
+      "mean_token_accuracy": 0.8424145579338074,
+      "step": 760
+    },
+    {
+      "epoch": 2.3930817610062896,
+      "grad_norm": 0.1860472857952118,
+      "learning_rate": 3.602547870870498e-05,
+      "loss": 0.3829,
+      "mean_token_accuracy": 0.8446792960166931,
+      "step": 761
+    },
+    {
+      "epoch": 2.3962264150943398,
+      "grad_norm": 0.16511857509613037,
+      "learning_rate": 3.6013978636677354e-05,
+      "loss": 0.3883,
+      "mean_token_accuracy": 0.8434211611747742,
+      "step": 762
+    },
+    {
+      "epoch": 2.39937106918239,
+      "grad_norm": 0.1752832978963852,
+      "learning_rate": 3.600246402237248e-05,
+      "loss": 0.3923,
+      "mean_token_accuracy": 0.840692400932312,
+      "step": 763
+    },
+    {
+      "epoch": 2.40251572327044,
+      "grad_norm": 0.1690099984407425,
+      "learning_rate": 3.5990934877739045e-05,
+      "loss": 0.3969,
+      "mean_token_accuracy": 0.8391318917274475,
+      "step": 764
+    },
+    {
+      "epoch": 2.4056603773584904,
+      "grad_norm": 0.13980698585510254,
+      "learning_rate": 3.597939121474082e-05,
+      "loss": 0.3978,
+      "mean_token_accuracy": 0.8390938639640808,
+      "step": 765
+    },
+    {
+      "epoch": 2.408805031446541,
+      "grad_norm": 0.1642495095729828,
+      "learning_rate": 3.5967833045356664e-05,
+      "loss": 0.3863,
+      "mean_token_accuracy": 0.8447932600975037,
+      "step": 766
+    },
+    {
+      "epoch": 2.411949685534591,
+      "grad_norm": 0.1437988579273224,
+      "learning_rate": 3.595626038158043e-05,
+      "loss": 0.4029,
+      "mean_token_accuracy": 0.8407385945320129,
+      "step": 767
+    },
+    {
+      "epoch": 2.4150943396226414,
+      "grad_norm": 0.1538083404302597,
+      "learning_rate": 3.594467323542107e-05,
+      "loss": 0.3918,
+      "mean_token_accuracy": 0.8416243195533752,
+      "step": 768
+    },
+    {
+      "epoch": 2.418238993710692,
+      "grad_norm": 0.16916736960411072,
+      "learning_rate": 3.593307161890254e-05,
+      "loss": 0.3855,
+      "mean_token_accuracy": 0.8438270092010498,
+      "step": 769
+    },
+    {
+      "epoch": 2.4213836477987423,
+      "grad_norm": 0.15521636605262756,
+      "learning_rate": 3.592145554406381e-05,
+      "loss": 0.4015,
+      "mean_token_accuracy": 0.8377931118011475,
+      "step": 770
+    },
+    {
+      "epoch": 2.4245283018867925,
+      "grad_norm": 0.15567384660243988,
+      "learning_rate": 3.5909825022958857e-05,
+      "loss": 0.3918,
+      "mean_token_accuracy": 0.841164767742157,
+      "step": 771
+    },
+    {
+      "epoch": 2.4276729559748427,
+      "grad_norm": 0.15030360221862793,
+      "learning_rate": 3.5898180067656655e-05,
+      "loss": 0.3968,
+      "mean_token_accuracy": 0.8410426378250122,
+      "step": 772
+    },
+    {
+      "epoch": 2.430817610062893,
+      "grad_norm": 0.14496302604675293,
+      "learning_rate": 3.5886520690241136e-05,
+      "loss": 0.3939,
+      "mean_token_accuracy": 0.8415027260780334,
+      "step": 773
+    },
+    {
+      "epoch": 2.4339622641509435,
+      "grad_norm": 0.16270099580287933,
+      "learning_rate": 3.587484690281123e-05,
+      "loss": 0.392,
+      "mean_token_accuracy": 0.8409454822540283,
+      "step": 774
+    },
+    {
+      "epoch": 2.4371069182389937,
+      "grad_norm": 0.1545172780752182,
+      "learning_rate": 3.5863158717480794e-05,
+      "loss": 0.4027,
+      "mean_token_accuracy": 0.8391482830047607,
+      "step": 775
+    },
+    {
+      "epoch": 2.440251572327044,
+      "grad_norm": 0.17777201533317566,
+      "learning_rate": 3.585145614637864e-05,
+      "loss": 0.3875,
+      "mean_token_accuracy": 0.845243513584137,
+      "step": 776
+    },
+    {
+      "epoch": 2.4433962264150946,
+      "grad_norm": 0.13722971081733704,
+      "learning_rate": 3.583973920164849e-05,
+      "loss": 0.3907,
+      "mean_token_accuracy": 0.8422651290893555,
+      "step": 777
+    },
+    {
+      "epoch": 2.4465408805031448,
+      "grad_norm": 0.16192865371704102,
+      "learning_rate": 3.5828007895449e-05,
+      "loss": 0.3844,
+      "mean_token_accuracy": 0.8450747728347778,
+      "step": 778
+    },
+    {
+      "epoch": 2.449685534591195,
+      "grad_norm": 0.15951241552829742,
+      "learning_rate": 3.581626223995372e-05,
+      "loss": 0.3847,
+      "mean_token_accuracy": 0.8426473736763,
+      "step": 779
+    },
+    {
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.1600176841020584,
+      "learning_rate": 3.580450224735111e-05,
+      "loss": 0.3868,
+      "mean_token_accuracy": 0.8432039618492126,
+      "step": 780
+    },
+    {
+      "epoch": 2.4559748427672954,
+      "grad_norm": 0.1509629637002945,
+      "learning_rate": 3.579272792984447e-05,
+      "loss": 0.3837,
+      "mean_token_accuracy": 0.843054473400116,
+      "step": 781
+    },
+    {
+      "epoch": 2.459119496855346,
+      "grad_norm": 0.15704569220542908,
+      "learning_rate": 3.5780939299651995e-05,
+      "loss": 0.4006,
+      "mean_token_accuracy": 0.8408331274986267,
+      "step": 782
+    },
+    {
+      "epoch": 2.4622641509433962,
+      "grad_norm": 0.1845240294933319,
+      "learning_rate": 3.576913636900672e-05,
+      "loss": 0.4115,
+      "mean_token_accuracy": 0.8372202515602112,
+      "step": 783
+    },
+    {
+      "epoch": 2.4654088050314464,
+      "grad_norm": 0.13821645081043243,
+      "learning_rate": 3.5757319150156515e-05,
+      "loss": 0.3848,
+      "mean_token_accuracy": 0.840967059135437,
+      "step": 784
+    },
+    {
+      "epoch": 2.468553459119497,
+      "grad_norm": 0.18777751922607422,
+      "learning_rate": 3.57454876553641e-05,
+      "loss": 0.3991,
+      "mean_token_accuracy": 0.8414704203605652,
+      "step": 785
+    },
+    {
+      "epoch": 2.4716981132075473,
+      "grad_norm": 0.14253944158554077,
+      "learning_rate": 3.573364189690699e-05,
+      "loss": 0.401,
+      "mean_token_accuracy": 0.8404497504234314,
+      "step": 786
+    },
+    {
+      "epoch": 2.4748427672955975,
+      "grad_norm": 0.17456071078777313,
+      "learning_rate": 3.572178188707749e-05,
+      "loss": 0.3815,
+      "mean_token_accuracy": 0.8429886102676392,
+      "step": 787
+    },
+    {
+      "epoch": 2.4779874213836477,
+      "grad_norm": 0.14869949221611023,
+      "learning_rate": 3.5709907638182725e-05,
+      "loss": 0.4001,
+      "mean_token_accuracy": 0.8371372818946838,
+      "step": 788
+    },
+    {
+      "epoch": 2.481132075471698,
+      "grad_norm": 0.1542949378490448,
+      "learning_rate": 3.569801916254457e-05,
+      "loss": 0.3738,
+      "mean_token_accuracy": 0.8442022800445557,
+      "step": 789
+    },
+    {
+      "epoch": 2.4842767295597485,
+      "grad_norm": 0.1838357299566269,
+      "learning_rate": 3.5686116472499665e-05,
+      "loss": 0.3944,
+      "mean_token_accuracy": 0.8391475677490234,
+      "step": 790
+    },
+    {
+      "epoch": 2.4874213836477987,
+      "grad_norm": 0.1430555135011673,
+      "learning_rate": 3.5674199580399425e-05,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8409162163734436,
+      "step": 791
+    },
+    {
+      "epoch": 2.490566037735849,
+      "grad_norm": 0.17979463934898376,
+      "learning_rate": 3.566226849860997e-05,
+      "loss": 0.3928,
+      "mean_token_accuracy": 0.8403165936470032,
+      "step": 792
+    },
+    {
+      "epoch": 2.4937106918238996,
+      "grad_norm": 0.137798473238945,
+      "learning_rate": 3.5650323239512175e-05,
+      "loss": 0.4061,
+      "mean_token_accuracy": 0.8378630876541138,
+      "step": 793
+    },
+    {
+      "epoch": 2.49685534591195,
+      "grad_norm": 0.1714894026517868,
+      "learning_rate": 3.563836381550159e-05,
+      "loss": 0.394,
+      "mean_token_accuracy": 0.8412221074104309,
+      "step": 794
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 0.15192954242229462,
+      "learning_rate": 3.5626390238988504e-05,
+      "loss": 0.4004,
+      "mean_token_accuracy": 0.8397027254104614,
+      "step": 795
+    },
+    {
+      "epoch": 2.50314465408805,
+      "grad_norm": 0.16279102861881256,
+      "learning_rate": 3.561440252239787e-05,
+      "loss": 0.3799,
+      "mean_token_accuracy": 0.8437523245811462,
+      "step": 796
+    },
+    {
+      "epoch": 2.5062893081761004,
+      "grad_norm": 0.15632732212543488,
+      "learning_rate": 3.56024006781693e-05,
+      "loss": 0.3938,
+      "mean_token_accuracy": 0.8418835997581482,
+      "step": 797
+    },
+    {
+      "epoch": 2.509433962264151,
+      "grad_norm": 0.1601596623659134,
+      "learning_rate": 3.559038471875711e-05,
+      "loss": 0.3757,
+      "mean_token_accuracy": 0.8448925614356995,
+      "step": 798
+    },
+    {
+      "epoch": 2.5125786163522013,
+      "grad_norm": 0.15498454868793488,
+      "learning_rate": 3.557835465663021e-05,
+      "loss": 0.3966,
+      "mean_token_accuracy": 0.83966064453125,
+      "step": 799
+    },
+    {
+      "epoch": 2.5157232704402515,
+      "grad_norm": 0.15797755122184753,
+      "learning_rate": 3.5566310504272194e-05,
+      "loss": 0.3916,
+      "mean_token_accuracy": 0.8428415656089783,
+      "step": 800
+    },
+    {
+      "epoch": 2.518867924528302,
+      "grad_norm": 0.17878195643424988,
+      "learning_rate": 3.5554252274181246e-05,
+      "loss": 0.3922,
+      "mean_token_accuracy": 0.8414608240127563,
+      "step": 801
+    },
+    {
+      "epoch": 2.5220125786163523,
+      "grad_norm": 0.1354987770318985,
+      "learning_rate": 3.554217997887016e-05,
+      "loss": 0.3874,
+      "mean_token_accuracy": 0.8432824611663818,
+      "step": 802
+    },
+    {
+      "epoch": 2.5251572327044025,
+      "grad_norm": 0.17350763082504272,
+      "learning_rate": 3.553009363086634e-05,
+      "loss": 0.399,
+      "mean_token_accuracy": 0.838498055934906,
+      "step": 803
+    },
+    {
+      "epoch": 2.5283018867924527,
+      "grad_norm": 0.1341233104467392,
+      "learning_rate": 3.551799324271176e-05,
+      "loss": 0.4094,
+      "mean_token_accuracy": 0.8359872698783875,
+      "step": 804
+    },
+    {
+      "epoch": 2.531446540880503,
+      "grad_norm": 0.18266914784908295,
+      "learning_rate": 3.550587882696297e-05,
+      "loss": 0.3996,
+      "mean_token_accuracy": 0.8383893966674805,
+      "step": 805
+    },
+    {
+      "epoch": 2.5345911949685536,
+      "grad_norm": 0.14570754766464233,
+      "learning_rate": 3.549375039619109e-05,
+      "loss": 0.393,
+      "mean_token_accuracy": 0.8403996229171753,
+      "step": 806
+    },
+    {
+      "epoch": 2.5377358490566038,
+      "grad_norm": 0.1772826611995697,
+      "learning_rate": 3.5481607962981744e-05,
+      "loss": 0.4042,
+      "mean_token_accuracy": 0.8374120593070984,
+      "step": 807
+    },
+    {
+      "epoch": 2.540880503144654,
+      "grad_norm": 0.13402216136455536,
+      "learning_rate": 3.546945153993512e-05,
+      "loss": 0.3969,
+      "mean_token_accuracy": 0.8391050100326538,
+      "step": 808
+    },
+    {
+      "epoch": 2.5440251572327046,
+      "grad_norm": 0.15814465284347534,
+      "learning_rate": 3.5457281139665906e-05,
+      "loss": 0.3778,
+      "mean_token_accuracy": 0.8457577228546143,
+      "step": 809
+    },
+    {
+      "epoch": 2.547169811320755,
+      "grad_norm": 0.13784445822238922,
+      "learning_rate": 3.544509677480332e-05,
+      "loss": 0.4,
+      "mean_token_accuracy": 0.8372910022735596,
+      "step": 810
+    },
+    {
+      "epoch": 2.550314465408805,
+      "grad_norm": 0.15171018242835999,
+      "learning_rate": 3.543289845799104e-05,
+      "loss": 0.3939,
+      "mean_token_accuracy": 0.8409681916236877,
+      "step": 811
+    },
+    {
+      "epoch": 2.5534591194968552,
+      "grad_norm": 0.14729034900665283,
+      "learning_rate": 3.542068620188723e-05,
+      "loss": 0.3979,
+      "mean_token_accuracy": 0.8405581116676331,
+      "step": 812
+    },
+    {
+      "epoch": 2.5566037735849054,
+      "grad_norm": 0.16989730298519135,
+      "learning_rate": 3.540846001916454e-05,
+      "loss": 0.3912,
+      "mean_token_accuracy": 0.8416076898574829,
+      "step": 813
+    },
+    {
+      "epoch": 2.559748427672956,
+      "grad_norm": 0.16093933582305908,
+      "learning_rate": 3.5396219922510044e-05,
+      "loss": 0.4019,
+      "mean_token_accuracy": 0.8399268984794617,
+      "step": 814
+    },
+    {
+      "epoch": 2.5628930817610063,
+      "grad_norm": 0.1573500782251358,
+      "learning_rate": 3.538396592462527e-05,
+      "loss": 0.3914,
+      "mean_token_accuracy": 0.8417745232582092,
+      "step": 815
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 0.1340954303741455,
+      "learning_rate": 3.5371698038226156e-05,
+      "loss": 0.397,
+      "mean_token_accuracy": 0.8433490991592407,
+      "step": 816
+    },
+    {
+      "epoch": 2.569182389937107,
+      "grad_norm": 0.14694054424762726,
+      "learning_rate": 3.535941627604307e-05,
+      "loss": 0.3905,
+      "mean_token_accuracy": 0.8411130309104919,
+      "step": 817
+    },
+    {
+      "epoch": 2.5723270440251573,
+      "grad_norm": 0.1424230933189392,
+      "learning_rate": 3.534712065082078e-05,
+      "loss": 0.3997,
+      "mean_token_accuracy": 0.8407908082008362,
+      "step": 818
+    },
+    {
+      "epoch": 2.5754716981132075,
+      "grad_norm": 0.14012737572193146,
+      "learning_rate": 3.533481117531842e-05,
+      "loss": 0.3798,
+      "mean_token_accuracy": 0.8445196151733398,
+      "step": 819
+    },
+    {
+      "epoch": 2.5786163522012577,
+      "grad_norm": 0.14719845354557037,
+      "learning_rate": 3.5322487862309516e-05,
+      "loss": 0.4068,
+      "mean_token_accuracy": 0.8381505608558655,
+      "step": 820
+    },
+    {
+      "epoch": 2.581761006289308,
+      "grad_norm": 0.14293119311332703,
+      "learning_rate": 3.531015072458193e-05,
+      "loss": 0.387,
+      "mean_token_accuracy": 0.8414595723152161,
+      "step": 821
+    },
+    {
+      "epoch": 2.5849056603773586,
+      "grad_norm": 0.15876775979995728,
+      "learning_rate": 3.5297799774937904e-05,
+      "loss": 0.3968,
+      "mean_token_accuracy": 0.8399572968482971,
+      "step": 822
+    },
+    {
+      "epoch": 2.588050314465409,
+      "grad_norm": 0.1583719551563263,
+      "learning_rate": 3.528543502619398e-05,
+      "loss": 0.3837,
+      "mean_token_accuracy": 0.8397992253303528,
+      "step": 823
+    },
+    {
+      "epoch": 2.591194968553459,
+      "grad_norm": 0.16118523478507996,
+      "learning_rate": 3.527305649118104e-05,
+      "loss": 0.3875,
+      "mean_token_accuracy": 0.8445963263511658,
+      "step": 824
+    },
+    {
+      "epoch": 2.5943396226415096,
+      "grad_norm": 0.1711784452199936,
+      "learning_rate": 3.526066418274427e-05,
+      "loss": 0.4029,
+      "mean_token_accuracy": 0.8400377631187439,
+      "step": 825
+    },
+    {
+      "epoch": 2.59748427672956,
+      "grad_norm": 0.14684435725212097,
+      "learning_rate": 3.5248258113743126e-05,
+      "loss": 0.3746,
+      "mean_token_accuracy": 0.8455667495727539,
+      "step": 826
+    },
+    {
+      "epoch": 2.60062893081761,
+      "grad_norm": 0.16698506474494934,
+      "learning_rate": 3.523583829705137e-05,
+      "loss": 0.3929,
+      "mean_token_accuracy": 0.8404534459114075,
+      "step": 827
+    },
+    {
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.15083381533622742,
+      "learning_rate": 3.5223404745557014e-05,
+      "loss": 0.3964,
+      "mean_token_accuracy": 0.8417296409606934,
+      "step": 828
+    },
+    {
+      "epoch": 2.6069182389937104,
+      "grad_norm": 0.15367716550827026,
+      "learning_rate": 3.521095747216235e-05,
+      "loss": 0.3921,
+      "mean_token_accuracy": 0.839659571647644,
+      "step": 829
+    },
+    {
+      "epoch": 2.610062893081761,
+      "grad_norm": 0.16619056463241577,
+      "learning_rate": 3.519849648978387e-05,
+      "loss": 0.3885,
+      "mean_token_accuracy": 0.840570330619812,
+      "step": 830
+    },
+    {
+      "epoch": 2.6132075471698113,
+      "grad_norm": 0.13816650211811066,
+      "learning_rate": 3.51860218113523e-05,
+      "loss": 0.4104,
+      "mean_token_accuracy": 0.8387232422828674,
+      "step": 831
+    },
+    {
+      "epoch": 2.6163522012578615,
+      "grad_norm": 0.15900948643684387,
+      "learning_rate": 3.5173533449812627e-05,
+      "loss": 0.3917,
+      "mean_token_accuracy": 0.8438796401023865,
+      "step": 832
+    },
+    {
+      "epoch": 2.619496855345912,
+      "grad_norm": 0.14279142022132874,
+      "learning_rate": 3.516103141812396e-05,
+      "loss": 0.3886,
+      "mean_token_accuracy": 0.8426015973091125,
+      "step": 833
+    },
+    {
+      "epoch": 2.6226415094339623,
+      "grad_norm": 0.1456068903207779,
+      "learning_rate": 3.5148515729259655e-05,
+      "loss": 0.4118,
+      "mean_token_accuracy": 0.8380022644996643,
+      "step": 834
+    },
+    {
+      "epoch": 2.6257861635220126,
+      "grad_norm": 0.17821946740150452,
+      "learning_rate": 3.51359863962072e-05,
+      "loss": 0.3919,
+      "mean_token_accuracy": 0.8446857929229736,
+      "step": 835
+    },
+    {
+      "epoch": 2.6289308176100628,
+      "grad_norm": 0.13951314985752106,
+      "learning_rate": 3.512344343196827e-05,
+      "loss": 0.3948,
+      "mean_token_accuracy": 0.841708779335022,
+      "step": 836
+    },
+    {
+      "epoch": 2.632075471698113,
+      "grad_norm": 0.1878071278333664,
+      "learning_rate": 3.511088684955868e-05,
+      "loss": 0.3835,
+      "mean_token_accuracy": 0.843018114566803,
+      "step": 837
+    },
+    {
+      "epoch": 2.6352201257861636,
+      "grad_norm": 0.14683803915977478,
+      "learning_rate": 3.509831666200835e-05,
+      "loss": 0.3926,
+      "mean_token_accuracy": 0.8428792357444763,
+      "step": 838
+    },
+    {
+      "epoch": 2.638364779874214,
+      "grad_norm": 0.19433486461639404,
+      "learning_rate": 3.508573288236135e-05,
+      "loss": 0.3948,
+      "mean_token_accuracy": 0.8401256799697876,
+      "step": 839
+    },
+    {
+      "epoch": 2.641509433962264,
+      "grad_norm": 0.12940803170204163,
+      "learning_rate": 3.5073135523675853e-05,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8402566313743591,
+      "step": 840
+    },
+    {
+      "epoch": 2.6446540880503147,
+      "grad_norm": 0.16293668746948242,
+      "learning_rate": 3.506052459902409e-05,
+      "loss": 0.3869,
+      "mean_token_accuracy": 0.841245174407959,
+      "step": 841
+    },
+    {
+      "epoch": 2.647798742138365,
+      "grad_norm": 0.153008833527565,
+      "learning_rate": 3.504790012149241e-05,
+      "loss": 0.3911,
+      "mean_token_accuracy": 0.840588390827179,
+      "step": 842
+    },
+    {
+      "epoch": 2.650943396226415,
+      "grad_norm": 0.14682938158512115,
+      "learning_rate": 3.503526210418119e-05,
+      "loss": 0.3967,
+      "mean_token_accuracy": 0.8388072848320007,
+      "step": 843
+    },
+    {
+      "epoch": 2.6540880503144653,
+      "grad_norm": 0.15707483887672424,
+      "learning_rate": 3.50226105602049e-05,
+      "loss": 0.3982,
+      "mean_token_accuracy": 0.8398618102073669,
+      "step": 844
+    },
+    {
+      "epoch": 2.6572327044025155,
+      "grad_norm": 0.15881039202213287,
+      "learning_rate": 3.5009945502692e-05,
+      "loss": 0.3933,
+      "mean_token_accuracy": 0.841876745223999,
+      "step": 845
+    },
+    {
+      "epoch": 2.660377358490566,
+      "grad_norm": 0.1584368199110031,
+      "learning_rate": 3.4997266944785e-05,
+      "loss": 0.4016,
+      "mean_token_accuracy": 0.8401806950569153,
+      "step": 846
+    },
+    {
+      "epoch": 2.6635220125786163,
+      "grad_norm": 0.16472209990024567,
+      "learning_rate": 3.4984574899640415e-05,
+      "loss": 0.4028,
+      "mean_token_accuracy": 0.8396847248077393,
+      "step": 847
+    },
+    {
+      "epoch": 2.6666666666666665,
+      "grad_norm": 0.15707650780677795,
+      "learning_rate": 3.497186938042876e-05,
+      "loss": 0.4055,
+      "mean_token_accuracy": 0.8379293084144592,
+      "step": 848
+    },
+    {
+      "epoch": 2.669811320754717,
+      "grad_norm": 0.1491553783416748,
+      "learning_rate": 3.4959150400334516e-05,
+      "loss": 0.3789,
+      "mean_token_accuracy": 0.8430595397949219,
+      "step": 849
+    },
+    {
+      "epoch": 2.6729559748427674,
+      "grad_norm": 0.1792389154434204,
+      "learning_rate": 3.494641797255616e-05,
+      "loss": 0.389,
+      "mean_token_accuracy": 0.8386589884757996,
+      "step": 850
+    },
+    {
+      "epoch": 2.6761006289308176,
+      "grad_norm": 0.15633267164230347,
+      "learning_rate": 3.493367211030609e-05,
+      "loss": 0.3971,
+      "mean_token_accuracy": 0.8409385681152344,
+      "step": 851
+    },
+    {
+      "epoch": 2.6792452830188678,
+      "grad_norm": 0.18488091230392456,
+      "learning_rate": 3.492091282681066e-05,
+      "loss": 0.3912,
+      "mean_token_accuracy": 0.8436446785926819,
+      "step": 852
+    },
+    {
+      "epoch": 2.682389937106918,
+      "grad_norm": 0.17898158729076385,
+      "learning_rate": 3.490814013531017e-05,
+      "loss": 0.3944,
+      "mean_token_accuracy": 0.8414387702941895,
+      "step": 853
+    },
+    {
+      "epoch": 2.6855345911949686,
+      "grad_norm": 0.1576376110315323,
+      "learning_rate": 3.489535404905879e-05,
+      "loss": 0.4077,
+      "mean_token_accuracy": 0.8369157314300537,
+      "step": 854
+    },
+    {
+      "epoch": 2.688679245283019,
+      "grad_norm": 0.19907242059707642,
+      "learning_rate": 3.4882554581324646e-05,
+      "loss": 0.3961,
+      "mean_token_accuracy": 0.8421619534492493,
+      "step": 855
+    },
+    {
+      "epoch": 2.691823899371069,
+      "grad_norm": 0.14955250918865204,
+      "learning_rate": 3.48697417453897e-05,
+      "loss": 0.4018,
+      "mean_token_accuracy": 0.8398721814155579,
+      "step": 856
+    },
+    {
+      "epoch": 2.6949685534591197,
+      "grad_norm": 0.1245059221982956,
+      "learning_rate": 3.485691555454982e-05,
+      "loss": 0.4034,
+      "mean_token_accuracy": 0.8386282324790955,
+      "step": 857
+    },
+    {
+      "epoch": 2.69811320754717,
+      "grad_norm": 0.16106462478637695,
+      "learning_rate": 3.48440760221147e-05,
+      "loss": 0.3885,
+      "mean_token_accuracy": 0.8441129922866821,
+      "step": 858
+    },
+    {
+      "epoch": 2.70125786163522,
+      "grad_norm": 0.1262756884098053,
+      "learning_rate": 3.4831223161407904e-05,
+      "loss": 0.4013,
+      "mean_token_accuracy": 0.8384125828742981,
+      "step": 859
+    },
+    {
+      "epoch": 2.7044025157232703,
+      "grad_norm": 0.16710178554058075,
+      "learning_rate": 3.4818356985766825e-05,
+      "loss": 0.3906,
+      "mean_token_accuracy": 0.8418945670127869,
+      "step": 860
+    },
+    {
+      "epoch": 2.7075471698113205,
+      "grad_norm": 0.14853724837303162,
+      "learning_rate": 3.4805477508542665e-05,
+      "loss": 0.3895,
+      "mean_token_accuracy": 0.8417914509773254,
+      "step": 861
+    },
+    {
+      "epoch": 2.710691823899371,
+      "grad_norm": 0.1702132523059845,
+      "learning_rate": 3.4792584743100425e-05,
+      "loss": 0.4086,
+      "mean_token_accuracy": 0.8393948674201965,
+      "step": 862
+    },
+    {
+      "epoch": 2.7138364779874213,
+      "grad_norm": 0.15771856904029846,
+      "learning_rate": 3.4779678702818915e-05,
+      "loss": 0.403,
+      "mean_token_accuracy": 0.8413280844688416,
+      "step": 863
+    },
+    {
+      "epoch": 2.7169811320754715,
+      "grad_norm": 0.1526855230331421,
+      "learning_rate": 3.4766759401090695e-05,
+      "loss": 0.395,
+      "mean_token_accuracy": 0.8401949405670166,
+      "step": 864
+    },
+    {
+      "epoch": 2.720125786163522,
+      "grad_norm": 0.15966515243053436,
+      "learning_rate": 3.47538268513221e-05,
+      "loss": 0.3947,
+      "mean_token_accuracy": 0.8407111763954163,
+      "step": 865
+    },
+    {
+      "epoch": 2.7232704402515724,
+      "grad_norm": 0.15073886513710022,
+      "learning_rate": 3.474088106693323e-05,
+      "loss": 0.3969,
+      "mean_token_accuracy": 0.8384842276573181,
+      "step": 866
+    },
+    {
+      "epoch": 2.7264150943396226,
+      "grad_norm": 0.15170443058013916,
+      "learning_rate": 3.472792206135786e-05,
+      "loss": 0.3938,
+      "mean_token_accuracy": 0.8373732566833496,
+      "step": 867
+    },
+    {
+      "epoch": 2.729559748427673,
+      "grad_norm": 0.1322529911994934,
+      "learning_rate": 3.471494984804355e-05,
+      "loss": 0.4004,
+      "mean_token_accuracy": 0.8394856452941895,
+      "step": 868
+    },
+    {
+      "epoch": 2.732704402515723,
+      "grad_norm": 0.14727897942066193,
+      "learning_rate": 3.4701964440451545e-05,
+      "loss": 0.3904,
+      "mean_token_accuracy": 0.8422083258628845,
+      "step": 869
+    },
+    {
+      "epoch": 2.7358490566037736,
+      "grad_norm": 0.14118586480617523,
+      "learning_rate": 3.468896585205676e-05,
+      "loss": 0.3974,
+      "mean_token_accuracy": 0.8416357040405273,
+      "step": 870
+    },
+    {
+      "epoch": 2.738993710691824,
+      "grad_norm": 0.1370769441127777,
+      "learning_rate": 3.467595409634781e-05,
+      "loss": 0.3972,
+      "mean_token_accuracy": 0.8423879146575928,
+      "step": 871
+    },
+    {
+      "epoch": 2.742138364779874,
+      "grad_norm": 0.12729352712631226,
+      "learning_rate": 3.466292918682696e-05,
+      "loss": 0.3891,
+      "mean_token_accuracy": 0.8427572250366211,
+      "step": 872
+    },
+    {
+      "epoch": 2.7452830188679247,
+      "grad_norm": 0.1347734034061432,
+      "learning_rate": 3.464989113701016e-05,
+      "loss": 0.3784,
+      "mean_token_accuracy": 0.8443403244018555,
+      "step": 873
+    },
+    {
+      "epoch": 2.748427672955975,
+      "grad_norm": 0.1276370882987976,
+      "learning_rate": 3.4636839960426935e-05,
+      "loss": 0.3969,
+      "mean_token_accuracy": 0.8401352763175964,
+      "step": 874
+    },
+    {
+      "epoch": 2.751572327044025,
+      "grad_norm": 0.12423428148031235,
+      "learning_rate": 3.462377567062048e-05,
+      "loss": 0.3887,
+      "mean_token_accuracy": 0.8414220213890076,
+      "step": 875
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.15068255364894867,
+      "learning_rate": 3.4610698281147574e-05,
+      "loss": 0.4004,
+      "mean_token_accuracy": 0.8396066427230835,
+      "step": 876
+    },
+    {
+      "epoch": 2.757861635220126,
+      "grad_norm": 0.13070818781852722,
+      "learning_rate": 3.4597607805578614e-05,
+      "loss": 0.3903,
+      "mean_token_accuracy": 0.8402306437492371,
+      "step": 877
+    },
+    {
+      "epoch": 2.761006289308176,
+      "grad_norm": 0.14346636831760406,
+      "learning_rate": 3.458450425749754e-05,
+      "loss": 0.3924,
+      "mean_token_accuracy": 0.8397766351699829,
+      "step": 878
+    },
+    {
+      "epoch": 2.7641509433962264,
+      "grad_norm": 0.14210395514965057,
+      "learning_rate": 3.4571387650501886e-05,
+      "loss": 0.3989,
+      "mean_token_accuracy": 0.8378952145576477,
+      "step": 879
+    },
+    {
+      "epoch": 2.767295597484277,
+      "grad_norm": 0.1414806991815567,
+      "learning_rate": 3.455825799820274e-05,
+      "loss": 0.3868,
+      "mean_token_accuracy": 0.8398988246917725,
+      "step": 880
+    },
+    {
+      "epoch": 2.770440251572327,
+      "grad_norm": 0.1717570275068283,
+      "learning_rate": 3.45451153142247e-05,
+      "loss": 0.4038,
+      "mean_token_accuracy": 0.8383566737174988,
+      "step": 881
+    },
+    {
+      "epoch": 2.7735849056603774,
+      "grad_norm": 0.13245093822479248,
+      "learning_rate": 3.45319596122059e-05,
+      "loss": 0.3847,
+      "mean_token_accuracy": 0.842934787273407,
+      "step": 882
+    },
+    {
+      "epoch": 2.7767295597484276,
+      "grad_norm": 0.1757514327764511,
+      "learning_rate": 3.4518790905798004e-05,
+      "loss": 0.3936,
+      "mean_token_accuracy": 0.8390152454376221,
+      "step": 883
+    },
+    {
+      "epoch": 2.779874213836478,
+      "grad_norm": 0.12597371637821198,
+      "learning_rate": 3.450560920866613e-05,
+      "loss": 0.4083,
+      "mean_token_accuracy": 0.8398998379707336,
+      "step": 884
+    },
+    {
+      "epoch": 2.7830188679245285,
+      "grad_norm": 0.18164481222629547,
+      "learning_rate": 3.4492414534488917e-05,
+      "loss": 0.3949,
+      "mean_token_accuracy": 0.8426663875579834,
+      "step": 885
+    },
+    {
+      "epoch": 2.7861635220125787,
+      "grad_norm": 0.12726567685604095,
+      "learning_rate": 3.4479206896958434e-05,
+      "loss": 0.4044,
+      "mean_token_accuracy": 0.8387595415115356,
+      "step": 886
+    },
+    {
+      "epoch": 2.789308176100629,
+      "grad_norm": 0.1545555740594864,
+      "learning_rate": 3.446598630978024e-05,
+      "loss": 0.3938,
+      "mean_token_accuracy": 0.8420332670211792,
+      "step": 887
+    },
+    {
+      "epoch": 2.7924528301886795,
+      "grad_norm": 0.12999430298805237,
+      "learning_rate": 3.445275278667329e-05,
+      "loss": 0.3945,
+      "mean_token_accuracy": 0.8392206430435181,
+      "step": 888
+    },
+    {
+      "epoch": 2.7955974842767297,
+      "grad_norm": 0.1740381270647049,
+      "learning_rate": 3.443950634137e-05,
+      "loss": 0.3868,
+      "mean_token_accuracy": 0.841921865940094,
+      "step": 889
+    },
+    {
+      "epoch": 2.79874213836478,
+      "grad_norm": 0.1445263922214508,
+      "learning_rate": 3.4426246987616176e-05,
+      "loss": 0.3815,
+      "mean_token_accuracy": 0.844272792339325,
+      "step": 890
+    },
+    {
+      "epoch": 2.80188679245283,
+      "grad_norm": 0.16784286499023438,
+      "learning_rate": 3.4412974739171026e-05,
+      "loss": 0.393,
+      "mean_token_accuracy": 0.841655969619751,
+      "step": 891
+    },
+    {
+      "epoch": 2.8050314465408803,
+      "grad_norm": 0.16633418202400208,
+      "learning_rate": 3.439968960980712e-05,
+      "loss": 0.3883,
+      "mean_token_accuracy": 0.8415431380271912,
+      "step": 892
+    },
+    {
+      "epoch": 2.808176100628931,
+      "grad_norm": 0.14095844328403473,
+      "learning_rate": 3.438639161331042e-05,
+      "loss": 0.3899,
+      "mean_token_accuracy": 0.8424274325370789,
+      "step": 893
+    },
+    {
+      "epoch": 2.811320754716981,
+      "grad_norm": 0.1560608446598053,
+      "learning_rate": 3.437308076348024e-05,
+      "loss": 0.4066,
+      "mean_token_accuracy": 0.839648962020874,
+      "step": 894
+    },
+    {
+      "epoch": 2.8144654088050314,
+      "grad_norm": 0.13541994988918304,
+      "learning_rate": 3.4359757074129205e-05,
+      "loss": 0.3911,
+      "mean_token_accuracy": 0.842206597328186,
+      "step": 895
+    },
+    {
+      "epoch": 2.817610062893082,
+      "grad_norm": 0.1437758207321167,
+      "learning_rate": 3.4346420559083285e-05,
+      "loss": 0.4054,
+      "mean_token_accuracy": 0.8397225141525269,
+      "step": 896
+    },
+    {
+      "epoch": 2.8207547169811322,
+      "grad_norm": 0.11642755568027496,
+      "learning_rate": 3.433307123218176e-05,
+      "loss": 0.3999,
+      "mean_token_accuracy": 0.8443455696105957,
+      "step": 897
+    },
+    {
+      "epoch": 2.8238993710691824,
+      "grad_norm": 0.13762716948986053,
+      "learning_rate": 3.43197091072772e-05,
+      "loss": 0.3982,
+      "mean_token_accuracy": 0.8394805192947388,
+      "step": 898
+    },
+    {
+      "epoch": 2.8270440251572326,
+      "grad_norm": 0.1345590353012085,
+      "learning_rate": 3.430633419823545e-05,
+      "loss": 0.4009,
+      "mean_token_accuracy": 0.8399420976638794,
+      "step": 899
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.13669990003108978,
+      "learning_rate": 3.429294651893563e-05,
+      "loss": 0.3908,
+      "mean_token_accuracy": 0.8423181176185608,
+      "step": 900
+    },
+    {
+      "epoch": 2.8333333333333335,
+      "grad_norm": 0.14826296269893646,
+      "learning_rate": 3.4279546083270097e-05,
+      "loss": 0.3985,
+      "mean_token_accuracy": 0.8397724628448486,
+      "step": 901
+    },
+    {
+      "epoch": 2.8364779874213837,
+      "grad_norm": 0.14717832207679749,
+      "learning_rate": 3.426613290514447e-05,
+      "loss": 0.3775,
+      "mean_token_accuracy": 0.8437859416007996,
+      "step": 902
+    },
+    {
+      "epoch": 2.839622641509434,
+      "grad_norm": 0.1572881042957306,
+      "learning_rate": 3.4252706998477574e-05,
+      "loss": 0.3908,
+      "mean_token_accuracy": 0.8429072499275208,
+      "step": 903
+    },
+    {
+      "epoch": 2.8427672955974845,
+      "grad_norm": 0.1492244154214859,
+      "learning_rate": 3.423926837720144e-05,
+      "loss": 0.4069,
+      "mean_token_accuracy": 0.8388838768005371,
+      "step": 904
+    },
+    {
+      "epoch": 2.8459119496855347,
+      "grad_norm": 0.15537552535533905,
+      "learning_rate": 3.42258170552613e-05,
+      "loss": 0.3888,
+      "mean_token_accuracy": 0.8405810594558716,
+      "step": 905
+    },
+    {
+      "epoch": 2.849056603773585,
+      "grad_norm": 0.1573428362607956,
+      "learning_rate": 3.421235304661555e-05,
+      "loss": 0.3889,
+      "mean_token_accuracy": 0.843213677406311,
+      "step": 906
+    },
+    {
+      "epoch": 2.852201257861635,
+      "grad_norm": 0.15431787073612213,
+      "learning_rate": 3.41988763652358e-05,
+      "loss": 0.3969,
+      "mean_token_accuracy": 0.8401937484741211,
+      "step": 907
+    },
+    {
+      "epoch": 2.8553459119496853,
+      "grad_norm": 0.16958792507648468,
+      "learning_rate": 3.4185387025106745e-05,
+      "loss": 0.3966,
+      "mean_token_accuracy": 0.8413075804710388,
+      "step": 908
+    },
+    {
+      "epoch": 2.858490566037736,
+      "grad_norm": 0.15407972037792206,
+      "learning_rate": 3.417188504022625e-05,
+      "loss": 0.3944,
+      "mean_token_accuracy": 0.8398054242134094,
+      "step": 909
+    },
+    {
+      "epoch": 2.861635220125786,
+      "grad_norm": 0.16856738924980164,
+      "learning_rate": 3.415837042460531e-05,
+      "loss": 0.3862,
+      "mean_token_accuracy": 0.8428761959075928,
+      "step": 910
+    },
+    {
+      "epoch": 2.8647798742138364,
+      "grad_norm": 0.16327671706676483,
+      "learning_rate": 3.414484319226799e-05,
+      "loss": 0.4078,
+      "mean_token_accuracy": 0.8394198417663574,
+      "step": 911
+    },
+    {
+      "epoch": 2.867924528301887,
+      "grad_norm": 0.13859853148460388,
+      "learning_rate": 3.4131303357251505e-05,
+      "loss": 0.3979,
+      "mean_token_accuracy": 0.839618980884552,
+      "step": 912
+    },
+    {
+      "epoch": 2.8710691823899372,
+      "grad_norm": 0.16593888401985168,
+      "learning_rate": 3.411775093360609e-05,
+      "loss": 0.3959,
+      "mean_token_accuracy": 0.8423987030982971,
+      "step": 913
+    },
+    {
+      "epoch": 2.8742138364779874,
+      "grad_norm": 0.12062633782625198,
+      "learning_rate": 3.4104185935395075e-05,
+      "loss": 0.391,
+      "mean_token_accuracy": 0.8417162895202637,
+      "step": 914
+    },
+    {
+      "epoch": 2.8773584905660377,
+      "grad_norm": 0.14128541946411133,
+      "learning_rate": 3.409060837669483e-05,
+      "loss": 0.4021,
+      "mean_token_accuracy": 0.840218186378479,
+      "step": 915
+    },
+    {
+      "epoch": 2.880503144654088,
+      "grad_norm": 0.12717147171497345,
+      "learning_rate": 3.407701827159476e-05,
+      "loss": 0.4044,
+      "mean_token_accuracy": 0.8357147574424744,
+      "step": 916
+    },
+    {
+      "epoch": 2.8836477987421385,
+      "grad_norm": 0.14301805198192596,
+      "learning_rate": 3.406341563419729e-05,
+      "loss": 0.3944,
+      "mean_token_accuracy": 0.8393138647079468,
+      "step": 917
+    },
+    {
+      "epoch": 2.8867924528301887,
+      "grad_norm": 0.12959828972816467,
+      "learning_rate": 3.4049800478617844e-05,
+      "loss": 0.3893,
+      "mean_token_accuracy": 0.8411592841148376,
+      "step": 918
+    },
+    {
+      "epoch": 2.889937106918239,
+      "grad_norm": 0.1419060230255127,
+      "learning_rate": 3.4036172818984855e-05,
+      "loss": 0.3923,
+      "mean_token_accuracy": 0.8420356512069702,
+      "step": 919
+    },
+    {
+      "epoch": 2.8930817610062896,
+      "grad_norm": 0.14367401599884033,
+      "learning_rate": 3.4022532669439706e-05,
+      "loss": 0.406,
+      "mean_token_accuracy": 0.8380276560783386,
+      "step": 920
+    },
+    {
+      "epoch": 2.8962264150943398,
+      "grad_norm": 0.13751500844955444,
+      "learning_rate": 3.4008880044136763e-05,
+      "loss": 0.3834,
+      "mean_token_accuracy": 0.8410583734512329,
+      "step": 921
+    },
+    {
+      "epoch": 2.89937106918239,
+      "grad_norm": 0.13970538973808289,
+      "learning_rate": 3.399521495724332e-05,
+      "loss": 0.3923,
+      "mean_token_accuracy": 0.842393696308136,
+      "step": 922
+    },
+    {
+      "epoch": 2.90251572327044,
+      "grad_norm": 0.1298050582408905,
+      "learning_rate": 3.39815374229396e-05,
+      "loss": 0.3951,
+      "mean_token_accuracy": 0.8414964079856873,
+      "step": 923
+    },
+    {
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.12179820239543915,
+      "learning_rate": 3.396784745541877e-05,
+      "loss": 0.3891,
+      "mean_token_accuracy": 0.8448134064674377,
+      "step": 924
+    },
+    {
+      "epoch": 2.908805031446541,
+      "grad_norm": 0.12799564003944397,
+      "learning_rate": 3.3954145068886876e-05,
+      "loss": 0.4021,
+      "mean_token_accuracy": 0.8391764163970947,
+      "step": 925
+    },
+    {
+      "epoch": 2.911949685534591,
+      "grad_norm": 0.1219654306769371,
+      "learning_rate": 3.394043027756286e-05,
+      "loss": 0.4057,
+      "mean_token_accuracy": 0.8397082686424255,
+      "step": 926
+    },
+    {
+      "epoch": 2.9150943396226414,
+      "grad_norm": 0.1398463398218155,
+      "learning_rate": 3.392670309567852e-05,
+      "loss": 0.3973,
+      "mean_token_accuracy": 0.8409194350242615,
+      "step": 927
+    },
+    {
+      "epoch": 2.918238993710692,
+      "grad_norm": 0.13440991938114166,
+      "learning_rate": 3.391296353747854e-05,
+      "loss": 0.3866,
+      "mean_token_accuracy": 0.8404433131217957,
+      "step": 928
+    },
+    {
+      "epoch": 2.9213836477987423,
+      "grad_norm": 0.13937310874462128,
+      "learning_rate": 3.389921161722042e-05,
+      "loss": 0.3968,
+      "mean_token_accuracy": 0.8404849171638489,
+      "step": 929
+    },
+    {
+      "epoch": 2.9245283018867925,
+      "grad_norm": 0.15991294384002686,
+      "learning_rate": 3.3885447349174514e-05,
+      "loss": 0.3983,
+      "mean_token_accuracy": 0.83951735496521,
+      "step": 930
+    },
+    {
+      "epoch": 2.9276729559748427,
+      "grad_norm": 0.14287906885147095,
+      "learning_rate": 3.387167074762396e-05,
+      "loss": 0.4209,
+      "mean_token_accuracy": 0.8352724313735962,
+      "step": 931
+    },
+    {
+      "epoch": 2.930817610062893,
+      "grad_norm": 0.1415240466594696,
+      "learning_rate": 3.3857881826864715e-05,
+      "loss": 0.3867,
+      "mean_token_accuracy": 0.8398596048355103,
+      "step": 932
+    },
+    {
+      "epoch": 2.9339622641509435,
+      "grad_norm": 0.14125128090381622,
+      "learning_rate": 3.3844080601205516e-05,
+      "loss": 0.4074,
+      "mean_token_accuracy": 0.8381497263908386,
+      "step": 933
+    },
+    {
+      "epoch": 2.9371069182389937,
+      "grad_norm": 0.13332779705524445,
+      "learning_rate": 3.383026708496787e-05,
+      "loss": 0.4023,
+      "mean_token_accuracy": 0.8399627804756165,
+      "step": 934
+    },
+    {
+      "epoch": 2.940251572327044,
+      "grad_norm": 0.13416799902915955,
+      "learning_rate": 3.381644129248602e-05,
+      "loss": 0.3902,
+      "mean_token_accuracy": 0.8443140983581543,
+      "step": 935
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 0.14479246735572815,
+      "learning_rate": 3.380260323810698e-05,
+      "loss": 0.3884,
+      "mean_token_accuracy": 0.8421072959899902,
+      "step": 936
+    },
+    {
+      "epoch": 2.9465408805031448,
+      "grad_norm": 0.1284419447183609,
+      "learning_rate": 3.378875293619047e-05,
+      "loss": 0.392,
+      "mean_token_accuracy": 0.8417386412620544,
+      "step": 937
+    },
+    {
+      "epoch": 2.949685534591195,
+      "grad_norm": 0.13157133758068085,
+      "learning_rate": 3.3774890401108915e-05,
+      "loss": 0.4076,
+      "mean_token_accuracy": 0.8378016352653503,
+      "step": 938
+    },
+    {
+      "epoch": 2.952830188679245,
+      "grad_norm": 0.13864070177078247,
+      "learning_rate": 3.376101564724743e-05,
+      "loss": 0.3901,
+      "mean_token_accuracy": 0.8419355750083923,
+      "step": 939
+    },
+    {
+      "epoch": 2.9559748427672954,
+      "grad_norm": 0.14290863275527954,
+      "learning_rate": 3.374712868900384e-05,
+      "loss": 0.3963,
+      "mean_token_accuracy": 0.8409423828125,
+      "step": 940
+    },
+    {
+      "epoch": 2.959119496855346,
+      "grad_norm": 0.12841610610485077,
+      "learning_rate": 3.373322954078859e-05,
+      "loss": 0.3969,
+      "mean_token_accuracy": 0.8391101956367493,
+      "step": 941
+    },
+    {
+      "epoch": 2.9622641509433962,
+      "grad_norm": 0.14718987047672272,
+      "learning_rate": 3.371931821702481e-05,
+      "loss": 0.4037,
+      "mean_token_accuracy": 0.8362467288970947,
+      "step": 942
+    },
+    {
+      "epoch": 2.9654088050314464,
+      "grad_norm": 0.1477939933538437,
+      "learning_rate": 3.3705394732148264e-05,
+      "loss": 0.4049,
+      "mean_token_accuracy": 0.8374539613723755,
+      "step": 943
+    },
+    {
+      "epoch": 2.968553459119497,
+      "grad_norm": 0.13969185948371887,
+      "learning_rate": 3.369145910060731e-05,
+      "loss": 0.3864,
+      "mean_token_accuracy": 0.8428540825843811,
+      "step": 944
+    },
+    {
+      "epoch": 2.9716981132075473,
+      "grad_norm": 0.15239839255809784,
+      "learning_rate": 3.3677511336862924e-05,
+      "loss": 0.4012,
+      "mean_token_accuracy": 0.8415625095367432,
+      "step": 945
+    },
+    {
+      "epoch": 2.9748427672955975,
+      "grad_norm": 0.1359701305627823,
+      "learning_rate": 3.366355145538867e-05,
+      "loss": 0.4019,
+      "mean_token_accuracy": 0.8378898501396179,
+      "step": 946
+    },
+    {
+      "epoch": 2.9779874213836477,
+      "grad_norm": 0.17559880018234253,
+      "learning_rate": 3.3649579470670696e-05,
+      "loss": 0.3981,
+      "mean_token_accuracy": 0.8389723300933838,
+      "step": 947
+    },
+    {
+      "epoch": 2.981132075471698,
+      "grad_norm": 0.12837964296340942,
+      "learning_rate": 3.3635595397207705e-05,
+      "loss": 0.4067,
+      "mean_token_accuracy": 0.8383262157440186,
+      "step": 948
+    },
+    {
+      "epoch": 2.9842767295597485,
+      "grad_norm": 0.16035127639770508,
+      "learning_rate": 3.3621599249510936e-05,
+      "loss": 0.4018,
+      "mean_token_accuracy": 0.8396176695823669,
+      "step": 949
+    },
+    {
+      "epoch": 2.9874213836477987,
+      "grad_norm": 0.13061131536960602,
+      "learning_rate": 3.360759104210417e-05,
+      "loss": 0.4034,
+      "mean_token_accuracy": 0.839031994342804,
+      "step": 950
+    },
+    {
+      "epoch": 2.990566037735849,
+      "grad_norm": 0.15194053947925568,
+      "learning_rate": 3.359357078952368e-05,
+      "loss": 0.385,
+      "mean_token_accuracy": 0.842530369758606,
+      "step": 951
+    },
+    {
+      "epoch": 2.9937106918238996,
+      "grad_norm": 0.1333847939968109,
+      "learning_rate": 3.3579538506318264e-05,
+      "loss": 0.3998,
+      "mean_token_accuracy": 0.8397628664970398,
+      "step": 952
+    },
+    {
+      "epoch": 2.99685534591195,
+      "grad_norm": 0.1507115513086319,
+      "learning_rate": 3.3565494207049194e-05,
+      "loss": 0.4102,
+      "mean_token_accuracy": 0.8390610218048096,
+      "step": 953
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.12280502915382385,
+      "learning_rate": 3.3551437906290207e-05,
+      "loss": 0.3837,
+      "mean_token_accuracy": 0.8393963575363159,
+      "step": 954
+    },
+    {
+      "epoch": 3.00314465408805,
+      "grad_norm": 0.2119593769311905,
+      "learning_rate": 3.353736961862751e-05,
+      "loss": 0.3276,
+      "mean_token_accuracy": 0.8589532375335693,
+      "step": 955
+    },
+    {
+      "epoch": 3.006289308176101,
+      "grad_norm": 0.5749790668487549,
+      "learning_rate": 3.352328935865972e-05,
+      "loss": 0.3309,
+      "mean_token_accuracy": 0.8566039800643921,
+      "step": 956
+    },
+    {
+      "epoch": 3.009433962264151,
+      "grad_norm": 0.30272579193115234,
+      "learning_rate": 3.350919714099791e-05,
+      "loss": 0.3372,
+      "mean_token_accuracy": 0.8549745082855225,
+      "step": 957
+    },
+    {
+      "epoch": 3.0125786163522013,
+      "grad_norm": 0.46458861231803894,
+      "learning_rate": 3.3495092980265526e-05,
+      "loss": 0.323,
+      "mean_token_accuracy": 0.8595535755157471,
+      "step": 958
+    },
+    {
+      "epoch": 3.0157232704402515,
+      "grad_norm": 0.29047465324401855,
+      "learning_rate": 3.348097689109844e-05,
+      "loss": 0.329,
+      "mean_token_accuracy": 0.859404444694519,
+      "step": 959
+    },
+    {
+      "epoch": 3.018867924528302,
+      "grad_norm": 0.1861647367477417,
+      "learning_rate": 3.346684888814489e-05,
+      "loss": 0.3285,
+      "mean_token_accuracy": 0.8579146862030029,
+      "step": 960
+    },
+    {
+      "epoch": 3.0220125786163523,
+      "grad_norm": 0.23538491129875183,
+      "learning_rate": 3.345270898606548e-05,
+      "loss": 0.3214,
+      "mean_token_accuracy": 0.8577497005462646,
+      "step": 961
+    },
+    {
+      "epoch": 3.0251572327044025,
+      "grad_norm": 0.24326147139072418,
+      "learning_rate": 3.3438557199533143e-05,
+      "loss": 0.3172,
+      "mean_token_accuracy": 0.8618540167808533,
+      "step": 962
+    },
+    {
+      "epoch": 3.0283018867924527,
+      "grad_norm": 0.19439604878425598,
+      "learning_rate": 3.342439354323317e-05,
+      "loss": 0.3343,
+      "mean_token_accuracy": 0.8569464087486267,
+      "step": 963
+    },
+    {
+      "epoch": 3.0314465408805034,
+      "grad_norm": 0.2556777596473694,
+      "learning_rate": 3.341021803186316e-05,
+      "loss": 0.3229,
+      "mean_token_accuracy": 0.859006404876709,
+      "step": 964
+    },
+    {
+      "epoch": 3.0345911949685536,
+      "grad_norm": 0.20313474535942078,
+      "learning_rate": 3.3396030680133e-05,
+      "loss": 0.319,
+      "mean_token_accuracy": 0.8617485165596008,
+      "step": 965
+    },
+    {
+      "epoch": 3.0377358490566038,
+      "grad_norm": 0.251817524433136,
+      "learning_rate": 3.3381831502764894e-05,
+      "loss": 0.3193,
+      "mean_token_accuracy": 0.863706111907959,
+      "step": 966
+    },
+    {
+      "epoch": 3.040880503144654,
+      "grad_norm": 0.20017312467098236,
+      "learning_rate": 3.3367620514493284e-05,
+      "loss": 0.3238,
+      "mean_token_accuracy": 0.8608281016349792,
+      "step": 967
+    },
+    {
+      "epoch": 3.0440251572327046,
+      "grad_norm": 0.27038446068763733,
+      "learning_rate": 3.335339773006489e-05,
+      "loss": 0.3151,
+      "mean_token_accuracy": 0.8628440499305725,
+      "step": 968
+    },
+    {
+      "epoch": 3.047169811320755,
+      "grad_norm": 0.2303992211818695,
+      "learning_rate": 3.3339163164238665e-05,
+      "loss": 0.3103,
+      "mean_token_accuracy": 0.8667078018188477,
+      "step": 969
+    },
+    {
+      "epoch": 3.050314465408805,
+      "grad_norm": 0.23521584272384644,
+      "learning_rate": 3.33249168317858e-05,
+      "loss": 0.3129,
+      "mean_token_accuracy": 0.8642385005950928,
+      "step": 970
+    },
+    {
+      "epoch": 3.0534591194968552,
+      "grad_norm": 0.20210982859134674,
+      "learning_rate": 3.331065874748967e-05,
+      "loss": 0.3125,
+      "mean_token_accuracy": 0.8629567623138428,
+      "step": 971
+    },
+    {
+      "epoch": 3.056603773584906,
+      "grad_norm": 0.22222645580768585,
+      "learning_rate": 3.329638892614587e-05,
+      "loss": 0.3041,
+      "mean_token_accuracy": 0.8669219017028809,
+      "step": 972
+    },
+    {
+      "epoch": 3.059748427672956,
+      "grad_norm": 0.1965298056602478,
+      "learning_rate": 3.328210738256215e-05,
+      "loss": 0.3232,
+      "mean_token_accuracy": 0.8628347516059875,
+      "step": 973
+    },
+    {
+      "epoch": 3.0628930817610063,
+      "grad_norm": 0.19745439291000366,
+      "learning_rate": 3.326781413155846e-05,
+      "loss": 0.3271,
+      "mean_token_accuracy": 0.8626418113708496,
+      "step": 974
+    },
+    {
+      "epoch": 3.0660377358490565,
+      "grad_norm": 0.16124515235424042,
+      "learning_rate": 3.325350918796686e-05,
+      "loss": 0.3434,
+      "mean_token_accuracy": 0.8579691648483276,
+      "step": 975
+    },
+    {
+      "epoch": 3.069182389937107,
+      "grad_norm": 0.1931421458721161,
+      "learning_rate": 3.323919256663157e-05,
+      "loss": 0.3245,
+      "mean_token_accuracy": 0.8611186146736145,
+      "step": 976
+    },
+    {
+      "epoch": 3.0723270440251573,
+      "grad_norm": 0.18141953647136688,
+      "learning_rate": 3.3224864282408925e-05,
+      "loss": 0.3094,
+      "mean_token_accuracy": 0.8668796420097351,
+      "step": 977
+    },
+    {
+      "epoch": 3.0754716981132075,
+      "grad_norm": 0.17107489705085754,
+      "learning_rate": 3.3210524350167346e-05,
+      "loss": 0.3047,
+      "mean_token_accuracy": 0.8638178110122681,
+      "step": 978
+    },
+    {
+      "epoch": 3.0786163522012577,
+      "grad_norm": 0.18757247924804688,
+      "learning_rate": 3.319617278478735e-05,
+      "loss": 0.3279,
+      "mean_token_accuracy": 0.8593670725822449,
+      "step": 979
+    },
+    {
+      "epoch": 3.0817610062893084,
+      "grad_norm": 0.15939109027385712,
+      "learning_rate": 3.3181809601161535e-05,
+      "loss": 0.3341,
+      "mean_token_accuracy": 0.8586524128913879,
+      "step": 980
+    },
+    {
+      "epoch": 3.0849056603773586,
+      "grad_norm": 0.17841064929962158,
+      "learning_rate": 3.316743481419454e-05,
+      "loss": 0.3279,
+      "mean_token_accuracy": 0.8618712425231934,
+      "step": 981
+    },
+    {
+      "epoch": 3.088050314465409,
+      "grad_norm": 0.15791453421115875,
+      "learning_rate": 3.3153048438803064e-05,
+      "loss": 0.3183,
+      "mean_token_accuracy": 0.8625402450561523,
+      "step": 982
+    },
+    {
+      "epoch": 3.091194968553459,
+      "grad_norm": 0.16996054351329803,
+      "learning_rate": 3.3138650489915805e-05,
+      "loss": 0.3459,
+      "mean_token_accuracy": 0.8569095134735107,
+      "step": 983
+    },
+    {
+      "epoch": 3.0943396226415096,
+      "grad_norm": 0.14582346379756927,
+      "learning_rate": 3.3124240982473495e-05,
+      "loss": 0.3199,
+      "mean_token_accuracy": 0.8625748157501221,
+      "step": 984
+    },
+    {
+      "epoch": 3.09748427672956,
+      "grad_norm": 0.16500362753868103,
+      "learning_rate": 3.3109819931428855e-05,
+      "loss": 0.3309,
+      "mean_token_accuracy": 0.8585406541824341,
+      "step": 985
+    },
+    {
+      "epoch": 3.10062893081761,
+      "grad_norm": 0.1490785777568817,
+      "learning_rate": 3.309538735174657e-05,
+      "loss": 0.3192,
+      "mean_token_accuracy": 0.8605947494506836,
+      "step": 986
+    },
+    {
+      "epoch": 3.1037735849056602,
+      "grad_norm": 0.1524612009525299,
+      "learning_rate": 3.30809432584033e-05,
+      "loss": 0.3223,
+      "mean_token_accuracy": 0.8592230677604675,
+      "step": 987
+    },
+    {
+      "epoch": 3.106918238993711,
+      "grad_norm": 0.1406700164079666,
+      "learning_rate": 3.306648766638765e-05,
+      "loss": 0.3349,
+      "mean_token_accuracy": 0.8577442169189453,
+      "step": 988
+    },
+    {
+      "epoch": 3.110062893081761,
+      "grad_norm": 0.15463581681251526,
+      "learning_rate": 3.3052020590700174e-05,
+      "loss": 0.3305,
+      "mean_token_accuracy": 0.8602052927017212,
+      "step": 989
+    },
+    {
+      "epoch": 3.1132075471698113,
+      "grad_norm": 0.1442749798297882,
+      "learning_rate": 3.30375420463533e-05,
+      "loss": 0.3205,
+      "mean_token_accuracy": 0.8626676201820374,
+      "step": 990
+    },
+    {
+      "epoch": 3.1163522012578615,
+      "grad_norm": 0.14940737187862396,
+      "learning_rate": 3.302305204837141e-05,
+      "loss": 0.3072,
+      "mean_token_accuracy": 0.8642657995223999,
+      "step": 991
+    },
+    {
+      "epoch": 3.119496855345912,
+      "grad_norm": 0.1608681082725525,
+      "learning_rate": 3.300855061179074e-05,
+      "loss": 0.3054,
+      "mean_token_accuracy": 0.8632607460021973,
+      "step": 992
+    },
+    {
+      "epoch": 3.1226415094339623,
+      "grad_norm": 0.1494629681110382,
+      "learning_rate": 3.2994037751659386e-05,
+      "loss": 0.3224,
+      "mean_token_accuracy": 0.8599094152450562,
+      "step": 993
+    },
+    {
+      "epoch": 3.1257861635220126,
+      "grad_norm": 0.1440831571817398,
+      "learning_rate": 3.297951348303733e-05,
+      "loss": 0.3369,
+      "mean_token_accuracy": 0.8589987754821777,
+      "step": 994
+    },
+    {
+      "epoch": 3.1289308176100628,
+      "grad_norm": 0.16431039571762085,
+      "learning_rate": 3.296497782099637e-05,
+      "loss": 0.3264,
+      "mean_token_accuracy": 0.8605884313583374,
+      "step": 995
+    },
+    {
+      "epoch": 3.1320754716981134,
+      "grad_norm": 0.14332625269889832,
+      "learning_rate": 3.2950430780620134e-05,
+      "loss": 0.3078,
+      "mean_token_accuracy": 0.8648683428764343,
+      "step": 996
+    },
+    {
+      "epoch": 3.1352201257861636,
+      "grad_norm": 0.1832369714975357,
+      "learning_rate": 3.293587237700405e-05,
+      "loss": 0.3178,
+      "mean_token_accuracy": 0.8615559339523315,
+      "step": 997
+    },
+    {
+      "epoch": 3.138364779874214,
+      "grad_norm": 0.14069117605686188,
+      "learning_rate": 3.292130262525536e-05,
+      "loss": 0.3123,
+      "mean_token_accuracy": 0.8634822368621826,
+      "step": 998
+    },
+    {
+      "epoch": 3.141509433962264,
+      "grad_norm": 0.16523166000843048,
+      "learning_rate": 3.290672154049306e-05,
+      "loss": 0.3357,
+      "mean_token_accuracy": 0.8579123616218567,
+      "step": 999
+    },
+    {
+      "epoch": 3.1446540880503147,
+      "grad_norm": 0.14075875282287598,
+      "learning_rate": 3.2892129137847886e-05,
+      "loss": 0.3249,
+      "mean_token_accuracy": 0.8633124828338623,
+      "step": 1000
+    },
+    {
+      "epoch": 3.147798742138365,
+      "grad_norm": 0.14860154688358307,
+      "learning_rate": 3.287752543246238e-05,
+      "loss": 0.3179,
+      "mean_token_accuracy": 0.8615362048149109,
+      "step": 1001
+    },
+    {
+      "epoch": 3.150943396226415,
+      "grad_norm": 0.15627436339855194,
+      "learning_rate": 3.2862910439490764e-05,
+      "loss": 0.3236,
+      "mean_token_accuracy": 0.8606275320053101,
+      "step": 1002
+    },
+    {
+      "epoch": 3.1540880503144653,
+      "grad_norm": 0.13712136447429657,
+      "learning_rate": 3.284828417409899e-05,
+      "loss": 0.3133,
+      "mean_token_accuracy": 0.8660383820533752,
+      "step": 1003
+    },
+    {
+      "epoch": 3.157232704402516,
+      "grad_norm": 0.16311588883399963,
+      "learning_rate": 3.283364665146469e-05,
+      "loss": 0.3095,
+      "mean_token_accuracy": 0.8647195100784302,
+      "step": 1004
+    },
+    {
+      "epoch": 3.160377358490566,
+      "grad_norm": 0.1587604582309723,
+      "learning_rate": 3.2818997886777214e-05,
+      "loss": 0.3333,
+      "mean_token_accuracy": 0.8577799201011658,
+      "step": 1005
+    },
+    {
+      "epoch": 3.1635220125786163,
+      "grad_norm": 0.14171631634235382,
+      "learning_rate": 3.280433789523753e-05,
+      "loss": 0.3386,
+      "mean_token_accuracy": 0.8587863445281982,
+      "step": 1006
+    },
+    {
+      "epoch": 3.1666666666666665,
+      "grad_norm": 0.14268644154071808,
+      "learning_rate": 3.2789666692058304e-05,
+      "loss": 0.3312,
+      "mean_token_accuracy": 0.8618650436401367,
+      "step": 1007
+    },
+    {
+      "epoch": 3.169811320754717,
+      "grad_norm": 0.1492924988269806,
+      "learning_rate": 3.277498429246381e-05,
+      "loss": 0.3347,
+      "mean_token_accuracy": 0.8577045798301697,
+      "step": 1008
+    },
+    {
+      "epoch": 3.1729559748427674,
+      "grad_norm": 0.13884948194026947,
+      "learning_rate": 3.276029071168993e-05,
+      "loss": 0.3257,
+      "mean_token_accuracy": 0.8589639067649841,
+      "step": 1009
+    },
+    {
+      "epoch": 3.1761006289308176,
+      "grad_norm": 0.17117413878440857,
+      "learning_rate": 3.2745585964984175e-05,
+      "loss": 0.3281,
+      "mean_token_accuracy": 0.8598911166191101,
+      "step": 1010
+    },
+    {
+      "epoch": 3.1792452830188678,
+      "grad_norm": 0.15213529765605927,
+      "learning_rate": 3.273087006760563e-05,
+      "loss": 0.3272,
+      "mean_token_accuracy": 0.8612444996833801,
+      "step": 1011
+    },
+    {
+      "epoch": 3.1823899371069184,
+      "grad_norm": 0.13560143113136292,
+      "learning_rate": 3.271614303482494e-05,
+      "loss": 0.3371,
+      "mean_token_accuracy": 0.8586222529411316,
+      "step": 1012
+    },
+    {
+      "epoch": 3.1855345911949686,
+      "grad_norm": 0.15628263354301453,
+      "learning_rate": 3.270140488192434e-05,
+      "loss": 0.3292,
+      "mean_token_accuracy": 0.8595730662345886,
+      "step": 1013
+    },
+    {
+      "epoch": 3.188679245283019,
+      "grad_norm": 0.14781375229358673,
+      "learning_rate": 3.268665562419756e-05,
+      "loss": 0.3286,
+      "mean_token_accuracy": 0.8589670658111572,
+      "step": 1014
+    },
+    {
+      "epoch": 3.191823899371069,
+      "grad_norm": 0.15790531039237976,
+      "learning_rate": 3.267189527694989e-05,
+      "loss": 0.3274,
+      "mean_token_accuracy": 0.8619564175605774,
+      "step": 1015
+    },
+    {
+      "epoch": 3.1949685534591197,
+      "grad_norm": 0.14986778795719147,
+      "learning_rate": 3.26571238554981e-05,
+      "loss": 0.3204,
+      "mean_token_accuracy": 0.8628863096237183,
+      "step": 1016
+    },
+    {
+      "epoch": 3.19811320754717,
+      "grad_norm": 0.16138505935668945,
+      "learning_rate": 3.264234137517047e-05,
+      "loss": 0.3368,
+      "mean_token_accuracy": 0.8585647344589233,
+      "step": 1017
+    },
+    {
+      "epoch": 3.20125786163522,
+      "grad_norm": 0.1323172003030777,
+      "learning_rate": 3.262754785130676e-05,
+      "loss": 0.3126,
+      "mean_token_accuracy": 0.8637034893035889,
+      "step": 1018
+    },
+    {
+      "epoch": 3.2044025157232703,
+      "grad_norm": 0.15144748985767365,
+      "learning_rate": 3.261274329925817e-05,
+      "loss": 0.3142,
+      "mean_token_accuracy": 0.863036036491394,
+      "step": 1019
+    },
+    {
+      "epoch": 3.207547169811321,
+      "grad_norm": 0.15019363164901733,
+      "learning_rate": 3.259792773438734e-05,
+      "loss": 0.3355,
+      "mean_token_accuracy": 0.8558260202407837,
+      "step": 1020
+    },
+    {
+      "epoch": 3.210691823899371,
+      "grad_norm": 0.14695622026920319,
+      "learning_rate": 3.2583101172068366e-05,
+      "loss": 0.3308,
+      "mean_token_accuracy": 0.8598349690437317,
+      "step": 1021
+    },
+    {
+      "epoch": 3.2138364779874213,
+      "grad_norm": 0.15687057375907898,
+      "learning_rate": 3.256826362768675e-05,
+      "loss": 0.3112,
+      "mean_token_accuracy": 0.8649905323982239,
+      "step": 1022
+    },
+    {
+      "epoch": 3.2169811320754715,
+      "grad_norm": 0.15487264096736908,
+      "learning_rate": 3.255341511663937e-05,
+      "loss": 0.3235,
+      "mean_token_accuracy": 0.8581811785697937,
+      "step": 1023
+    },
+    {
+      "epoch": 3.220125786163522,
+      "grad_norm": 0.15453389286994934,
+      "learning_rate": 3.253855565433449e-05,
+      "loss": 0.3157,
+      "mean_token_accuracy": 0.860891580581665,
+      "step": 1024
+    },
+    {
+      "epoch": 3.2232704402515724,
+      "grad_norm": 0.1621147245168686,
+      "learning_rate": 3.2523685256191756e-05,
+      "loss": 0.3167,
+      "mean_token_accuracy": 0.8601864576339722,
+      "step": 1025
+    },
+    {
+      "epoch": 3.2264150943396226,
+      "grad_norm": 0.14345519244670868,
+      "learning_rate": 3.250880393764215e-05,
+      "loss": 0.314,
+      "mean_token_accuracy": 0.861411988735199,
+      "step": 1026
+    },
+    {
+      "epoch": 3.229559748427673,
+      "grad_norm": 0.1536128968000412,
+      "learning_rate": 3.249391171412798e-05,
+      "loss": 0.3488,
+      "mean_token_accuracy": 0.8550587296485901,
+      "step": 1027
+    },
+    {
+      "epoch": 3.2327044025157234,
+      "grad_norm": 0.15043888986110687,
+      "learning_rate": 3.247900860110288e-05,
+      "loss": 0.3356,
+      "mean_token_accuracy": 0.8564497232437134,
+      "step": 1028
+    },
+    {
+      "epoch": 3.2358490566037736,
+      "grad_norm": 0.1707436740398407,
+      "learning_rate": 3.246409461403178e-05,
+      "loss": 0.3344,
+      "mean_token_accuracy": 0.8586674928665161,
+      "step": 1029
+    },
+    {
+      "epoch": 3.238993710691824,
+      "grad_norm": 0.152404323220253,
+      "learning_rate": 3.244916976839089e-05,
+      "loss": 0.3287,
+      "mean_token_accuracy": 0.8581715822219849,
+      "step": 1030
+    },
+    {
+      "epoch": 3.242138364779874,
+      "grad_norm": 0.1631406843662262,
+      "learning_rate": 3.243423407966769e-05,
+      "loss": 0.3234,
+      "mean_token_accuracy": 0.8623092770576477,
+      "step": 1031
+    },
+    {
+      "epoch": 3.2452830188679247,
+      "grad_norm": 0.15967321395874023,
+      "learning_rate": 3.241928756336093e-05,
+      "loss": 0.3226,
+      "mean_token_accuracy": 0.8575816750526428,
+      "step": 1032
+    },
+    {
+      "epoch": 3.248427672955975,
+      "grad_norm": 0.14411385357379913,
+      "learning_rate": 3.240433023498056e-05,
+      "loss": 0.3309,
+      "mean_token_accuracy": 0.8565492630004883,
+      "step": 1033
+    },
+    {
+      "epoch": 3.251572327044025,
+      "grad_norm": 0.16305826604366302,
+      "learning_rate": 3.238936211004779e-05,
+      "loss": 0.3238,
+      "mean_token_accuracy": 0.8587154150009155,
+      "step": 1034
+    },
+    {
+      "epoch": 3.2547169811320753,
+      "grad_norm": 0.14919938147068024,
+      "learning_rate": 3.237438320409499e-05,
+      "loss": 0.3415,
+      "mean_token_accuracy": 0.8566460609436035,
+      "step": 1035
+    },
+    {
+      "epoch": 3.257861635220126,
+      "grad_norm": 0.15982939302921295,
+      "learning_rate": 3.2359393532665744e-05,
+      "loss": 0.3393,
+      "mean_token_accuracy": 0.8567012548446655,
+      "step": 1036
+    },
+    {
+      "epoch": 3.261006289308176,
+      "grad_norm": 0.15051038563251495,
+      "learning_rate": 3.234439311131483e-05,
+      "loss": 0.3311,
+      "mean_token_accuracy": 0.8606399297714233,
+      "step": 1037
+    },
+    {
+      "epoch": 3.2641509433962264,
+      "grad_norm": 0.15478473901748657,
+      "learning_rate": 3.232938195560812e-05,
+      "loss": 0.3029,
+      "mean_token_accuracy": 0.8636201620101929,
+      "step": 1038
+    },
+    {
+      "epoch": 3.2672955974842766,
+      "grad_norm": 0.16665510833263397,
+      "learning_rate": 3.231436008112268e-05,
+      "loss": 0.3252,
+      "mean_token_accuracy": 0.8577843904495239,
+      "step": 1039
+    },
+    {
+      "epoch": 3.270440251572327,
+      "grad_norm": 0.14956386387348175,
+      "learning_rate": 3.2299327503446675e-05,
+      "loss": 0.3244,
+      "mean_token_accuracy": 0.8613575100898743,
+      "step": 1040
+    },
+    {
+      "epoch": 3.2735849056603774,
+      "grad_norm": 0.15692093968391418,
+      "learning_rate": 3.2284284238179385e-05,
+      "loss": 0.3207,
+      "mean_token_accuracy": 0.8585643172264099,
+      "step": 1041
+    },
+    {
+      "epoch": 3.2767295597484276,
+      "grad_norm": 0.14993791282176971,
+      "learning_rate": 3.226923030093117e-05,
+      "loss": 0.3244,
+      "mean_token_accuracy": 0.8617245554924011,
+      "step": 1042
+    },
+    {
+      "epoch": 3.279874213836478,
+      "grad_norm": 0.14964789152145386,
+      "learning_rate": 3.225416570732346e-05,
+      "loss": 0.3282,
+      "mean_token_accuracy": 0.8592199087142944,
+      "step": 1043
+    },
+    {
+      "epoch": 3.2830188679245285,
+      "grad_norm": 0.1455584019422531,
+      "learning_rate": 3.2239090472988775e-05,
+      "loss": 0.323,
+      "mean_token_accuracy": 0.8582662343978882,
+      "step": 1044
+    },
+    {
+      "epoch": 3.2861635220125787,
+      "grad_norm": 0.14278201758861542,
+      "learning_rate": 3.222400461357064e-05,
+      "loss": 0.3016,
+      "mean_token_accuracy": 0.8661797046661377,
+      "step": 1045
+    },
+    {
+      "epoch": 3.289308176100629,
+      "grad_norm": 0.13362888991832733,
+      "learning_rate": 3.220890814472363e-05,
+      "loss": 0.3325,
+      "mean_token_accuracy": 0.8578916788101196,
+      "step": 1046
+    },
+    {
+      "epoch": 3.292452830188679,
+      "grad_norm": 0.14459533989429474,
+      "learning_rate": 3.219380108211331e-05,
+      "loss": 0.3237,
+      "mean_token_accuracy": 0.8623485565185547,
+      "step": 1047
+    },
+    {
+      "epoch": 3.2955974842767297,
+      "grad_norm": 0.13775929808616638,
+      "learning_rate": 3.217868344141625e-05,
+      "loss": 0.3279,
+      "mean_token_accuracy": 0.8596442341804504,
+      "step": 1048
+    },
+    {
+      "epoch": 3.29874213836478,
+      "grad_norm": 0.12983611226081848,
+      "learning_rate": 3.2163555238320006e-05,
+      "loss": 0.3241,
+      "mean_token_accuracy": 0.8623194694519043,
+      "step": 1049
+    },
+    {
+      "epoch": 3.30188679245283,
+      "grad_norm": 0.13327816128730774,
+      "learning_rate": 3.214841648852308e-05,
+      "loss": 0.3283,
+      "mean_token_accuracy": 0.8586874604225159,
+      "step": 1050
+    },
+    {
+      "epoch": 3.3050314465408803,
+      "grad_norm": 0.14740483462810516,
+      "learning_rate": 3.21332672077349e-05,
+      "loss": 0.3289,
+      "mean_token_accuracy": 0.8617115616798401,
+      "step": 1051
+    },
+    {
+      "epoch": 3.308176100628931,
+      "grad_norm": 0.1283697932958603,
+      "learning_rate": 3.211810741167588e-05,
+      "loss": 0.3175,
+      "mean_token_accuracy": 0.8588584065437317,
+      "step": 1052
+    },
+    {
+      "epoch": 3.311320754716981,
+      "grad_norm": 0.15197114646434784,
+      "learning_rate": 3.210293711607729e-05,
+      "loss": 0.3347,
+      "mean_token_accuracy": 0.8559669852256775,
+      "step": 1053
+    },
+    {
+      "epoch": 3.3144654088050314,
+      "grad_norm": 0.13700467348098755,
+      "learning_rate": 3.2087756336681306e-05,
+      "loss": 0.325,
+      "mean_token_accuracy": 0.8605079054832458,
+      "step": 1054
+    },
+    {
+      "epoch": 3.3176100628930816,
+      "grad_norm": 0.13200706243515015,
+      "learning_rate": 3.207256508924101e-05,
+      "loss": 0.3317,
+      "mean_token_accuracy": 0.8585512638092041,
+      "step": 1055
+    },
+    {
+      "epoch": 3.3207547169811322,
+      "grad_norm": 0.1488047093153,
+      "learning_rate": 3.2057363389520326e-05,
+      "loss": 0.3381,
+      "mean_token_accuracy": 0.8565067052841187,
+      "step": 1056
+    },
+    {
+      "epoch": 3.3238993710691824,
+      "grad_norm": 0.1520201414823532,
+      "learning_rate": 3.204215125329402e-05,
+      "loss": 0.3342,
+      "mean_token_accuracy": 0.8572053909301758,
+      "step": 1057
+    },
+    {
+      "epoch": 3.3270440251572326,
+      "grad_norm": 0.17158624529838562,
+      "learning_rate": 3.20269286963477e-05,
+      "loss": 0.3316,
+      "mean_token_accuracy": 0.858790934085846,
+      "step": 1058
+    },
+    {
+      "epoch": 3.330188679245283,
+      "grad_norm": 0.14773574471473694,
+      "learning_rate": 3.2011695734477776e-05,
+      "loss": 0.3207,
+      "mean_token_accuracy": 0.8638202548027039,
+      "step": 1059
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "grad_norm": 0.17465537786483765,
+      "learning_rate": 3.199645238349146e-05,
+      "loss": 0.321,
+      "mean_token_accuracy": 0.8613391518592834,
+      "step": 1060
+    },
+    {
+      "epoch": 3.3364779874213837,
+      "grad_norm": 0.13760405778884888,
+      "learning_rate": 3.198119865920677e-05,
+      "loss": 0.3217,
+      "mean_token_accuracy": 0.8583642244338989,
+      "step": 1061
+    },
+    {
+      "epoch": 3.339622641509434,
+      "grad_norm": 0.15301382541656494,
+      "learning_rate": 3.196593457745243e-05,
+      "loss": 0.3197,
+      "mean_token_accuracy": 0.8593172430992126,
+      "step": 1062
+    },
+    {
+      "epoch": 3.342767295597484,
+      "grad_norm": 0.1372508555650711,
+      "learning_rate": 3.195066015406797e-05,
+      "loss": 0.3318,
+      "mean_token_accuracy": 0.8575015664100647,
+      "step": 1063
+    },
+    {
+      "epoch": 3.3459119496855347,
+      "grad_norm": 0.14857687056064606,
+      "learning_rate": 3.193537540490363e-05,
+      "loss": 0.3316,
+      "mean_token_accuracy": 0.8603220582008362,
+      "step": 1064
+    },
+    {
+      "epoch": 3.349056603773585,
+      "grad_norm": 0.12988995015621185,
+      "learning_rate": 3.192008034582034e-05,
+      "loss": 0.3467,
+      "mean_token_accuracy": 0.8531395792961121,
+      "step": 1065
+    },
+    {
+      "epoch": 3.352201257861635,
+      "grad_norm": 0.15077495574951172,
+      "learning_rate": 3.190477499268978e-05,
+      "loss": 0.3352,
+      "mean_token_accuracy": 0.8567934632301331,
+      "step": 1066
+    },
+    {
+      "epoch": 3.3553459119496853,
+      "grad_norm": 0.14799143373966217,
+      "learning_rate": 3.1889459361394266e-05,
+      "loss": 0.2899,
+      "mean_token_accuracy": 0.8670781850814819,
+      "step": 1067
+    },
+    {
+      "epoch": 3.358490566037736,
+      "grad_norm": 0.1631443202495575,
+      "learning_rate": 3.1874133467826804e-05,
+      "loss": 0.3139,
+      "mean_token_accuracy": 0.8630610108375549,
+      "step": 1068
+    },
+    {
+      "epoch": 3.361635220125786,
+      "grad_norm": 0.1495402604341507,
+      "learning_rate": 3.1858797327891034e-05,
+      "loss": 0.3071,
+      "mean_token_accuracy": 0.8628964424133301,
+      "step": 1069
+    },
+    {
+      "epoch": 3.3647798742138364,
+      "grad_norm": 0.1613866537809372,
+      "learning_rate": 3.1843450957501254e-05,
+      "loss": 0.3187,
+      "mean_token_accuracy": 0.8615692257881165,
+      "step": 1070
+    },
+    {
+      "epoch": 3.3679245283018866,
+      "grad_norm": 0.15920579433441162,
+      "learning_rate": 3.182809437258235e-05,
+      "loss": 0.3149,
+      "mean_token_accuracy": 0.862976610660553,
+      "step": 1071
+    },
+    {
+      "epoch": 3.3710691823899372,
+      "grad_norm": 0.15265892446041107,
+      "learning_rate": 3.181272758906982e-05,
+      "loss": 0.3171,
+      "mean_token_accuracy": 0.8599149584770203,
+      "step": 1072
+    },
+    {
+      "epoch": 3.3742138364779874,
+      "grad_norm": 0.15581786632537842,
+      "learning_rate": 3.179735062290974e-05,
+      "loss": 0.3255,
+      "mean_token_accuracy": 0.8590546250343323,
+      "step": 1073
+    },
+    {
+      "epoch": 3.3773584905660377,
+      "grad_norm": 0.13360513746738434,
+      "learning_rate": 3.178196349005877e-05,
+      "loss": 0.3146,
+      "mean_token_accuracy": 0.8623788952827454,
+      "step": 1074
+    },
+    {
+      "epoch": 3.380503144654088,
+      "grad_norm": 0.1339465230703354,
+      "learning_rate": 3.17665662064841e-05,
+      "loss": 0.3545,
+      "mean_token_accuracy": 0.852672278881073,
+      "step": 1075
+    },
+    {
+      "epoch": 3.3836477987421385,
+      "grad_norm": 0.1387154757976532,
+      "learning_rate": 3.175115878816346e-05,
+      "loss": 0.3286,
+      "mean_token_accuracy": 0.8580860495567322,
+      "step": 1076
+    },
+    {
+      "epoch": 3.3867924528301887,
+      "grad_norm": 0.14403823018074036,
+      "learning_rate": 3.173574125108508e-05,
+      "loss": 0.3441,
+      "mean_token_accuracy": 0.8584855794906616,
+      "step": 1077
+    },
+    {
+      "epoch": 3.389937106918239,
+      "grad_norm": 0.15723450481891632,
+      "learning_rate": 3.172031361124774e-05,
+      "loss": 0.3252,
+      "mean_token_accuracy": 0.8592239022254944,
+      "step": 1078
+    },
+    {
+      "epoch": 3.3930817610062896,
+      "grad_norm": 0.13823583722114563,
+      "learning_rate": 3.1704875884660645e-05,
+      "loss": 0.3227,
+      "mean_token_accuracy": 0.860748291015625,
+      "step": 1079
+    },
+    {
+      "epoch": 3.3962264150943398,
+      "grad_norm": 0.17623189091682434,
+      "learning_rate": 3.16894280873435e-05,
+      "loss": 0.3334,
+      "mean_token_accuracy": 0.8578466773033142,
+      "step": 1080
+    },
+    {
+      "epoch": 3.39937106918239,
+      "grad_norm": 0.15371590852737427,
+      "learning_rate": 3.1673970235326454e-05,
+      "loss": 0.3171,
+      "mean_token_accuracy": 0.8577912449836731,
+      "step": 1081
+    },
+    {
+      "epoch": 3.40251572327044,
+      "grad_norm": 0.16374284029006958,
+      "learning_rate": 3.165850234465009e-05,
+      "loss": 0.3333,
+      "mean_token_accuracy": 0.8590589761734009,
+      "step": 1082
+    },
+    {
+      "epoch": 3.4056603773584904,
+      "grad_norm": 0.1451641023159027,
+      "learning_rate": 3.16430244313654e-05,
+      "loss": 0.3259,
+      "mean_token_accuracy": 0.8558355569839478,
+      "step": 1083
+    },
+    {
+      "epoch": 3.408805031446541,
+      "grad_norm": 0.1551799774169922,
+      "learning_rate": 3.1627536511533795e-05,
+      "loss": 0.3339,
+      "mean_token_accuracy": 0.8557202219963074,
+      "step": 1084
+    },
+    {
+      "epoch": 3.411949685534591,
+      "grad_norm": 0.12902113795280457,
+      "learning_rate": 3.1612038601227054e-05,
+      "loss": 0.3279,
+      "mean_token_accuracy": 0.8582240343093872,
+      "step": 1085
+    },
+    {
+      "epoch": 3.4150943396226414,
+      "grad_norm": 0.15876270830631256,
+      "learning_rate": 3.159653071652732e-05,
+      "loss": 0.3307,
+      "mean_token_accuracy": 0.8565781712532043,
+      "step": 1086
+    },
+    {
+      "epoch": 3.418238993710692,
+      "grad_norm": 0.12267336994409561,
+      "learning_rate": 3.1581012873527095e-05,
+      "loss": 0.327,
+      "mean_token_accuracy": 0.8610835671424866,
+      "step": 1087
+    },
+    {
+      "epoch": 3.4213836477987423,
+      "grad_norm": 0.15954187512397766,
+      "learning_rate": 3.156548508832922e-05,
+      "loss": 0.3304,
+      "mean_token_accuracy": 0.8586665391921997,
+      "step": 1088
+    },
+    {
+      "epoch": 3.4245283018867925,
+      "grad_norm": 0.1352037936449051,
+      "learning_rate": 3.154994737704684e-05,
+      "loss": 0.3317,
+      "mean_token_accuracy": 0.8586193919181824,
+      "step": 1089
+    },
+    {
+      "epoch": 3.4276729559748427,
+      "grad_norm": 0.15149874985218048,
+      "learning_rate": 3.153439975580341e-05,
+      "loss": 0.3375,
+      "mean_token_accuracy": 0.854519248008728,
+      "step": 1090
+    },
+    {
+      "epoch": 3.430817610062893,
+      "grad_norm": 0.13168424367904663,
+      "learning_rate": 3.151884224073267e-05,
+      "loss": 0.332,
+      "mean_token_accuracy": 0.8577566742897034,
+      "step": 1091
+    },
+    {
+      "epoch": 3.4339622641509435,
+      "grad_norm": 0.13555015623569489,
+      "learning_rate": 3.150327484797861e-05,
+      "loss": 0.3236,
+      "mean_token_accuracy": 0.8604238033294678,
+      "step": 1092
+    },
+    {
+      "epoch": 3.4371069182389937,
+      "grad_norm": 0.132098987698555,
+      "learning_rate": 3.1487697593695495e-05,
+      "loss": 0.3232,
+      "mean_token_accuracy": 0.8571335077285767,
+      "step": 1093
+    },
+    {
+      "epoch": 3.440251572327044,
+      "grad_norm": 0.12883129715919495,
+      "learning_rate": 3.147211049404779e-05,
+      "loss": 0.3271,
+      "mean_token_accuracy": 0.8585851192474365,
+      "step": 1094
+    },
+    {
+      "epoch": 3.4433962264150946,
+      "grad_norm": 0.12580397725105286,
+      "learning_rate": 3.145651356521022e-05,
+      "loss": 0.3376,
+      "mean_token_accuracy": 0.8587803840637207,
+      "step": 1095
+    },
+    {
+      "epoch": 3.4465408805031448,
+      "grad_norm": 0.13099335134029388,
+      "learning_rate": 3.1440906823367676e-05,
+      "loss": 0.3191,
+      "mean_token_accuracy": 0.8609188795089722,
+      "step": 1096
+    },
+    {
+      "epoch": 3.449685534591195,
+      "grad_norm": 0.1349143236875534,
+      "learning_rate": 3.142529028471525e-05,
+      "loss": 0.3198,
+      "mean_token_accuracy": 0.8595296144485474,
+      "step": 1097
+    },
+    {
+      "epoch": 3.452830188679245,
+      "grad_norm": 0.14235574007034302,
+      "learning_rate": 3.140966396545817e-05,
+      "loss": 0.3402,
+      "mean_token_accuracy": 0.854846179485321,
+      "step": 1098
+    },
+    {
+      "epoch": 3.4559748427672954,
+      "grad_norm": 0.12889491021633148,
+      "learning_rate": 3.139402788181186e-05,
+      "loss": 0.3474,
+      "mean_token_accuracy": 0.8538563847541809,
+      "step": 1099
+    },
+    {
+      "epoch": 3.459119496855346,
+      "grad_norm": 0.15164689719676971,
+      "learning_rate": 3.137838205000185e-05,
+      "loss": 0.3369,
+      "mean_token_accuracy": 0.8573994040489197,
+      "step": 1100
+    },
+    {
+      "epoch": 3.4622641509433962,
+      "grad_norm": 0.1371922492980957,
+      "learning_rate": 3.136272648626377e-05,
+      "loss": 0.3334,
+      "mean_token_accuracy": 0.8574711084365845,
+      "step": 1101
+    },
+    {
+      "epoch": 3.4654088050314464,
+      "grad_norm": 0.1531357765197754,
+      "learning_rate": 3.1347061206843376e-05,
+      "loss": 0.329,
+      "mean_token_accuracy": 0.8560079336166382,
+      "step": 1102
+    },
+    {
+      "epoch": 3.468553459119497,
+      "grad_norm": 0.15599913895130157,
+      "learning_rate": 3.133138622799651e-05,
+      "loss": 0.3245,
+      "mean_token_accuracy": 0.8592775464057922,
+      "step": 1103
+    },
+    {
+      "epoch": 3.4716981132075473,
+      "grad_norm": 0.13518376648426056,
+      "learning_rate": 3.131570156598905e-05,
+      "loss": 0.325,
+      "mean_token_accuracy": 0.8589106798171997,
+      "step": 1104
+    },
+    {
+      "epoch": 3.4748427672955975,
+      "grad_norm": 0.15266206860542297,
+      "learning_rate": 3.1300007237096944e-05,
+      "loss": 0.3346,
+      "mean_token_accuracy": 0.857874870300293,
+      "step": 1105
+    },
+    {
+      "epoch": 3.4779874213836477,
+      "grad_norm": 0.1420784741640091,
+      "learning_rate": 3.128430325760616e-05,
+      "loss": 0.3296,
+      "mean_token_accuracy": 0.8598121404647827,
+      "step": 1106
+    },
+    {
+      "epoch": 3.481132075471698,
+      "grad_norm": 0.15022684633731842,
+      "learning_rate": 3.126858964381269e-05,
+      "loss": 0.3211,
+      "mean_token_accuracy": 0.8595227003097534,
+      "step": 1107
+    },
+    {
+      "epoch": 3.4842767295597485,
+      "grad_norm": 0.15471023321151733,
+      "learning_rate": 3.125286641202252e-05,
+      "loss": 0.3259,
+      "mean_token_accuracy": 0.8603160381317139,
+      "step": 1108
+    },
+    {
+      "epoch": 3.4874213836477987,
+      "grad_norm": 0.1372515708208084,
+      "learning_rate": 3.123713357855163e-05,
+      "loss": 0.3256,
+      "mean_token_accuracy": 0.8592529296875,
+      "step": 1109
+    },
+    {
+      "epoch": 3.490566037735849,
+      "grad_norm": 0.14333589375019073,
+      "learning_rate": 3.122139115972591e-05,
+      "loss": 0.3216,
+      "mean_token_accuracy": 0.8592953681945801,
+      "step": 1110
+    },
+    {
+      "epoch": 3.4937106918238996,
+      "grad_norm": 0.13496816158294678,
+      "learning_rate": 3.120563917188127e-05,
+      "loss": 0.3319,
+      "mean_token_accuracy": 0.8571884632110596,
+      "step": 1111
+    },
+    {
+      "epoch": 3.49685534591195,
+      "grad_norm": 0.13933950662612915,
+      "learning_rate": 3.1189877631363514e-05,
+      "loss": 0.3262,
+      "mean_token_accuracy": 0.8581831455230713,
+      "step": 1112
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 0.13966146111488342,
+      "learning_rate": 3.117410655452835e-05,
+      "loss": 0.3428,
+      "mean_token_accuracy": 0.8562673926353455,
+      "step": 1113
+    },
+    {
+      "epoch": 3.50314465408805,
+      "grad_norm": 0.13303886353969574,
+      "learning_rate": 3.115832595774139e-05,
+      "loss": 0.3212,
+      "mean_token_accuracy": 0.8587968349456787,
+      "step": 1114
+    },
+    {
+      "epoch": 3.5062893081761004,
+      "grad_norm": 0.14135101437568665,
+      "learning_rate": 3.114253585737813e-05,
+      "loss": 0.3278,
+      "mean_token_accuracy": 0.8600554466247559,
+      "step": 1115
+    },
+    {
+      "epoch": 3.509433962264151,
+      "grad_norm": 0.1329166740179062,
+      "learning_rate": 3.112673626982394e-05,
+      "loss": 0.3426,
+      "mean_token_accuracy": 0.853922963142395,
+      "step": 1116
+    },
+    {
+      "epoch": 3.5125786163522013,
+      "grad_norm": 0.1364695280790329,
+      "learning_rate": 3.1110927211474e-05,
+      "loss": 0.343,
+      "mean_token_accuracy": 0.8539784550666809,
+      "step": 1117
+    },
+    {
+      "epoch": 3.5157232704402515,
+      "grad_norm": 0.1294035166501999,
+      "learning_rate": 3.109510869873335e-05,
+      "loss": 0.3456,
+      "mean_token_accuracy": 0.8523107767105103,
+      "step": 1118
+    },
+    {
+      "epoch": 3.518867924528302,
+      "grad_norm": 0.1287209391593933,
+      "learning_rate": 3.107928074801682e-05,
+      "loss": 0.3331,
+      "mean_token_accuracy": 0.8596871495246887,
+      "step": 1119
+    },
+    {
+      "epoch": 3.5220125786163523,
+      "grad_norm": 0.12785553932189941,
+      "learning_rate": 3.106344337574904e-05,
+      "loss": 0.3308,
+      "mean_token_accuracy": 0.8588385581970215,
+      "step": 1120
+    },
+    {
+      "epoch": 3.5251572327044025,
+      "grad_norm": 0.14281029999256134,
+      "learning_rate": 3.1047596598364436e-05,
+      "loss": 0.3362,
+      "mean_token_accuracy": 0.8574225902557373,
+      "step": 1121
+    },
+    {
+      "epoch": 3.5283018867924527,
+      "grad_norm": 0.13183027505874634,
+      "learning_rate": 3.1031740432307164e-05,
+      "loss": 0.3127,
+      "mean_token_accuracy": 0.8617753386497498,
+      "step": 1122
+    },
+    {
+      "epoch": 3.531446540880503,
+      "grad_norm": 0.13874752819538116,
+      "learning_rate": 3.1015874894031144e-05,
+      "loss": 0.3334,
+      "mean_token_accuracy": 0.8580701351165771,
+      "step": 1123
+    },
+    {
+      "epoch": 3.5345911949685536,
+      "grad_norm": 0.12882445752620697,
+      "learning_rate": 3.1e-05,
+      "loss": 0.3373,
+      "mean_token_accuracy": 0.8580555319786072,
+      "step": 1124
+    },
+    {
+      "epoch": 3.5377358490566038,
+      "grad_norm": 0.14017626643180847,
+      "learning_rate": 3.0984115766687096e-05,
+      "loss": 0.3346,
+      "mean_token_accuracy": 0.8570591807365417,
+      "step": 1125
+    },
+    {
+      "epoch": 3.540880503144654,
+      "grad_norm": 0.12685616314411163,
+      "learning_rate": 3.0968222210575446e-05,
+      "loss": 0.3221,
+      "mean_token_accuracy": 0.8617547154426575,
+      "step": 1126
+    },
+    {
+      "epoch": 3.5440251572327046,
+      "grad_norm": 0.1380213052034378,
+      "learning_rate": 3.095231934815777e-05,
+      "loss": 0.3239,
+      "mean_token_accuracy": 0.8619950413703918,
+      "step": 1127
+    },
+    {
+      "epoch": 3.547169811320755,
+      "grad_norm": 0.136666402220726,
+      "learning_rate": 3.0936407195936444e-05,
+      "loss": 0.329,
+      "mean_token_accuracy": 0.8563064336776733,
+      "step": 1128
+    },
+    {
+      "epoch": 3.550314465408805,
+      "grad_norm": 0.1525544971227646,
+      "learning_rate": 3.092048577042347e-05,
+      "loss": 0.3268,
+      "mean_token_accuracy": 0.8579536080360413,
+      "step": 1129
+    },
+    {
+      "epoch": 3.5534591194968552,
+      "grad_norm": 0.14374618232250214,
+      "learning_rate": 3.090455508814047e-05,
+      "loss": 0.3191,
+      "mean_token_accuracy": 0.861701250076294,
+      "step": 1130
+    },
+    {
+      "epoch": 3.5566037735849054,
+      "grad_norm": 0.1475425511598587,
+      "learning_rate": 3.088861516561871e-05,
+      "loss": 0.3168,
+      "mean_token_accuracy": 0.8611730933189392,
+      "step": 1131
+    },
+    {
+      "epoch": 3.559748427672956,
+      "grad_norm": 0.1540241688489914,
+      "learning_rate": 3.087266601939898e-05,
+      "loss": 0.3125,
+      "mean_token_accuracy": 0.8643029928207397,
+      "step": 1132
+    },
+    {
+      "epoch": 3.5628930817610063,
+      "grad_norm": 0.14442327618598938,
+      "learning_rate": 3.0856707666031694e-05,
+      "loss": 0.3327,
+      "mean_token_accuracy": 0.8581700921058655,
+      "step": 1133
+    },
+    {
+      "epoch": 3.5660377358490565,
+      "grad_norm": 0.15339228510856628,
+      "learning_rate": 3.084074012207681e-05,
+      "loss": 0.3034,
+      "mean_token_accuracy": 0.8633737564086914,
+      "step": 1134
+    },
+    {
+      "epoch": 3.569182389937107,
+      "grad_norm": 0.14664144814014435,
+      "learning_rate": 3.0824763404103795e-05,
+      "loss": 0.3392,
+      "mean_token_accuracy": 0.8558675646781921,
+      "step": 1135
+    },
+    {
+      "epoch": 3.5723270440251573,
+      "grad_norm": 0.16255716979503632,
+      "learning_rate": 3.080877752869168e-05,
+      "loss": 0.3353,
+      "mean_token_accuracy": 0.8562329411506653,
+      "step": 1136
+    },
+    {
+      "epoch": 3.5754716981132075,
+      "grad_norm": 0.1473674178123474,
+      "learning_rate": 3.079278251242896e-05,
+      "loss": 0.336,
+      "mean_token_accuracy": 0.8567706942558289,
+      "step": 1137
+    },
+    {
+      "epoch": 3.5786163522012577,
+      "grad_norm": 0.1516093611717224,
+      "learning_rate": 3.0776778371913634e-05,
+      "loss": 0.305,
+      "mean_token_accuracy": 0.8626529574394226,
+      "step": 1138
+    },
+    {
+      "epoch": 3.581761006289308,
+      "grad_norm": 0.1458694338798523,
+      "learning_rate": 3.076076512375317e-05,
+      "loss": 0.3435,
+      "mean_token_accuracy": 0.8568443059921265,
+      "step": 1139
+    },
+    {
+      "epoch": 3.5849056603773586,
+      "grad_norm": 0.14711114764213562,
+      "learning_rate": 3.0744742784564476e-05,
+      "loss": 0.3441,
+      "mean_token_accuracy": 0.8580071926116943,
+      "step": 1140
+    },
+    {
+      "epoch": 3.588050314465409,
+      "grad_norm": 0.1530224233865738,
+      "learning_rate": 3.0728711370973915e-05,
+      "loss": 0.3328,
+      "mean_token_accuracy": 0.8571670055389404,
+      "step": 1141
+    },
+    {
+      "epoch": 3.591194968553459,
+      "grad_norm": 0.1374947428703308,
+      "learning_rate": 3.071267089961724e-05,
+      "loss": 0.3153,
+      "mean_token_accuracy": 0.8622951507568359,
+      "step": 1142
+    },
+    {
+      "epoch": 3.5943396226415096,
+      "grad_norm": 0.15225058794021606,
+      "learning_rate": 3.069662138713962e-05,
+      "loss": 0.3218,
+      "mean_token_accuracy": 0.8582719564437866,
+      "step": 1143
+    },
+    {
+      "epoch": 3.59748427672956,
+      "grad_norm": 0.13568712770938873,
+      "learning_rate": 3.0680562850195594e-05,
+      "loss": 0.3311,
+      "mean_token_accuracy": 0.8576697707176208,
+      "step": 1144
+    },
+    {
+      "epoch": 3.60062893081761,
+      "grad_norm": 0.16287925839424133,
+      "learning_rate": 3.0664495305449084e-05,
+      "loss": 0.3225,
+      "mean_token_accuracy": 0.8594067692756653,
+      "step": 1145
+    },
+    {
+      "epoch": 3.6037735849056602,
+      "grad_norm": 0.14078298211097717,
+      "learning_rate": 3.064841876957335e-05,
+      "loss": 0.3139,
+      "mean_token_accuracy": 0.8635094165802002,
+      "step": 1146
+    },
+    {
+      "epoch": 3.6069182389937104,
+      "grad_norm": 0.1636054962873459,
+      "learning_rate": 3.0632333259250966e-05,
+      "loss": 0.3261,
+      "mean_token_accuracy": 0.8589130640029907,
+      "step": 1147
+    },
+    {
+      "epoch": 3.610062893081761,
+      "grad_norm": 0.1595996916294098,
+      "learning_rate": 3.0616238791173846e-05,
+      "loss": 0.34,
+      "mean_token_accuracy": 0.8539798855781555,
+      "step": 1148
+    },
+    {
+      "epoch": 3.6132075471698113,
+      "grad_norm": 0.13862352073192596,
+      "learning_rate": 3.060013538204319e-05,
+      "loss": 0.3185,
+      "mean_token_accuracy": 0.8629903197288513,
+      "step": 1149
+    },
+    {
+      "epoch": 3.6163522012578615,
+      "grad_norm": 0.14423629641532898,
+      "learning_rate": 3.058402304856946e-05,
+      "loss": 0.332,
+      "mean_token_accuracy": 0.8597278594970703,
+      "step": 1150
+    },
+    {
+      "epoch": 3.619496855345912,
+      "grad_norm": 0.1510828286409378,
+      "learning_rate": 3.0567901807472394e-05,
+      "loss": 0.3137,
+      "mean_token_accuracy": 0.8636487126350403,
+      "step": 1151
+    },
+    {
+      "epoch": 3.6226415094339623,
+      "grad_norm": 0.14903636276721954,
+      "learning_rate": 3.055177167548098e-05,
+      "loss": 0.3312,
+      "mean_token_accuracy": 0.8585656881332397,
+      "step": 1152
+    },
+    {
+      "epoch": 3.6257861635220126,
+      "grad_norm": 0.1466004103422165,
+      "learning_rate": 3.053563266933343e-05,
+      "loss": 0.3295,
+      "mean_token_accuracy": 0.8588059544563293,
+      "step": 1153
+    },
+    {
+      "epoch": 3.6289308176100628,
+      "grad_norm": 0.14094513654708862,
+      "learning_rate": 3.051948480577714e-05,
+      "loss": 0.3323,
+      "mean_token_accuracy": 0.8572471737861633,
+      "step": 1154
+    },
+    {
+      "epoch": 3.632075471698113,
+      "grad_norm": 0.13098470866680145,
+      "learning_rate": 3.0503328101568713e-05,
+      "loss": 0.336,
+      "mean_token_accuracy": 0.8568713665008545,
+      "step": 1155
+    },
+    {
+      "epoch": 3.6352201257861636,
+      "grad_norm": 0.1380501538515091,
+      "learning_rate": 3.0487162573473944e-05,
+      "loss": 0.3424,
+      "mean_token_accuracy": 0.8545725345611572,
+      "step": 1156
+    },
+    {
+      "epoch": 3.638364779874214,
+      "grad_norm": 0.13335196673870087,
+      "learning_rate": 3.047098823826776e-05,
+      "loss": 0.3258,
+      "mean_token_accuracy": 0.8585770130157471,
+      "step": 1157
+    },
+    {
+      "epoch": 3.641509433962264,
+      "grad_norm": 0.14133410155773163,
+      "learning_rate": 3.0454805112734243e-05,
+      "loss": 0.3433,
+      "mean_token_accuracy": 0.8555521965026855,
+      "step": 1158
+    },
+    {
+      "epoch": 3.6446540880503147,
+      "grad_norm": 0.1298840194940567,
+      "learning_rate": 3.043861321366658e-05,
+      "loss": 0.3353,
+      "mean_token_accuracy": 0.859320342540741,
+      "step": 1159
+    },
+    {
+      "epoch": 3.647798742138365,
+      "grad_norm": 0.1415969878435135,
+      "learning_rate": 3.042241255786708e-05,
+      "loss": 0.3397,
+      "mean_token_accuracy": 0.8569203615188599,
+      "step": 1160
+    },
+    {
+      "epoch": 3.650943396226415,
+      "grad_norm": 0.1243518739938736,
+      "learning_rate": 3.040620316214713e-05,
+      "loss": 0.3227,
+      "mean_token_accuracy": 0.86089026927948,
+      "step": 1161
+    },
+    {
+      "epoch": 3.6540880503144653,
+      "grad_norm": 0.16862568259239197,
+      "learning_rate": 3.0389985043327183e-05,
+      "loss": 0.35,
+      "mean_token_accuracy": 0.8550198674201965,
+      "step": 1162
+    },
+    {
+      "epoch": 3.6572327044025155,
+      "grad_norm": 0.1371821016073227,
+      "learning_rate": 3.0373758218236747e-05,
+      "loss": 0.3112,
+      "mean_token_accuracy": 0.8630196452140808,
+      "step": 1163
+    },
+    {
+      "epoch": 3.660377358490566,
+      "grad_norm": 0.14441686868667603,
+      "learning_rate": 3.0357522703714365e-05,
+      "loss": 0.3322,
+      "mean_token_accuracy": 0.8588864207267761,
+      "step": 1164
+    },
+    {
+      "epoch": 3.6635220125786163,
+      "grad_norm": 0.14246846735477448,
+      "learning_rate": 3.03412785166076e-05,
+      "loss": 0.3236,
+      "mean_token_accuracy": 0.8609263300895691,
+      "step": 1165
+    },
+    {
+      "epoch": 3.6666666666666665,
+      "grad_norm": 0.14187541604042053,
+      "learning_rate": 3.032502567377302e-05,
+      "loss": 0.3287,
+      "mean_token_accuracy": 0.8563394546508789,
+      "step": 1166
+    },
+    {
+      "epoch": 3.669811320754717,
+      "grad_norm": 0.1399676501750946,
+      "learning_rate": 3.030876419207616e-05,
+      "loss": 0.3193,
+      "mean_token_accuracy": 0.8611453771591187,
+      "step": 1167
+    },
+    {
+      "epoch": 3.6729559748427674,
+      "grad_norm": 0.15003567934036255,
+      "learning_rate": 3.0292494088391522e-05,
+      "loss": 0.3436,
+      "mean_token_accuracy": 0.8553731441497803,
+      "step": 1168
+    },
+    {
+      "epoch": 3.6761006289308176,
+      "grad_norm": 0.1415679007768631,
+      "learning_rate": 3.027621537960257e-05,
+      "loss": 0.3343,
+      "mean_token_accuracy": 0.8586793541908264,
+      "step": 1169
+    },
+    {
+      "epoch": 3.6792452830188678,
+      "grad_norm": 0.13909806311130524,
+      "learning_rate": 3.0259928082601675e-05,
+      "loss": 0.3342,
+      "mean_token_accuracy": 0.8584513664245605,
+      "step": 1170
+    },
+    {
+      "epoch": 3.682389937106918,
+      "grad_norm": 0.1442382037639618,
+      "learning_rate": 3.0243632214290137e-05,
+      "loss": 0.3401,
+      "mean_token_accuracy": 0.854636549949646,
+      "step": 1171
+    },
+    {
+      "epoch": 3.6855345911949686,
+      "grad_norm": 0.1362895667552948,
+      "learning_rate": 3.022732779157815e-05,
+      "loss": 0.3486,
+      "mean_token_accuracy": 0.8539217710494995,
+      "step": 1172
+    },
+    {
+      "epoch": 3.688679245283019,
+      "grad_norm": 0.14030933380126953,
+      "learning_rate": 3.0211014831384773e-05,
+      "loss": 0.3305,
+      "mean_token_accuracy": 0.8588977456092834,
+      "step": 1173
+    },
+    {
+      "epoch": 3.691823899371069,
+      "grad_norm": 0.14149655401706696,
+      "learning_rate": 3.0194693350637943e-05,
+      "loss": 0.331,
+      "mean_token_accuracy": 0.8578828573226929,
+      "step": 1174
+    },
+    {
+      "epoch": 3.6949685534591197,
+      "grad_norm": 0.1377638727426529,
+      "learning_rate": 3.017836336627441e-05,
+      "loss": 0.3356,
+      "mean_token_accuracy": 0.854766845703125,
+      "step": 1175
+    },
+    {
+      "epoch": 3.69811320754717,
+      "grad_norm": 0.1417425572872162,
+      "learning_rate": 3.0162024895239787e-05,
+      "loss": 0.3372,
+      "mean_token_accuracy": 0.8555588126182556,
+      "step": 1176
+    },
+    {
+      "epoch": 3.70125786163522,
+      "grad_norm": 0.14174726605415344,
+      "learning_rate": 3.0145677954488457e-05,
+      "loss": 0.3338,
+      "mean_token_accuracy": 0.856699526309967,
+      "step": 1177
+    },
+    {
+      "epoch": 3.7044025157232703,
+      "grad_norm": 0.13998551666736603,
+      "learning_rate": 3.0129322560983616e-05,
+      "loss": 0.326,
+      "mean_token_accuracy": 0.8607649803161621,
+      "step": 1178
+    },
+    {
+      "epoch": 3.7075471698113205,
+      "grad_norm": 0.1254308521747589,
+      "learning_rate": 3.0112958731697214e-05,
+      "loss": 0.3512,
+      "mean_token_accuracy": 0.8531991243362427,
+      "step": 1179
+    },
+    {
+      "epoch": 3.710691823899371,
+      "grad_norm": 0.1437944918870926,
+      "learning_rate": 3.0096586483609966e-05,
+      "loss": 0.3354,
+      "mean_token_accuracy": 0.8561474680900574,
+      "step": 1180
+    },
+    {
+      "epoch": 3.7138364779874213,
+      "grad_norm": 0.1321929395198822,
+      "learning_rate": 3.0080205833711325e-05,
+      "loss": 0.3284,
+      "mean_token_accuracy": 0.8593660593032837,
+      "step": 1181
+    },
+    {
+      "epoch": 3.7169811320754715,
+      "grad_norm": 0.15401658415794373,
+      "learning_rate": 3.0063816798999456e-05,
+      "loss": 0.3314,
+      "mean_token_accuracy": 0.8591458201408386,
+      "step": 1182
+    },
+    {
+      "epoch": 3.720125786163522,
+      "grad_norm": 0.14057710766792297,
+      "learning_rate": 3.0047419396481227e-05,
+      "loss": 0.3475,
+      "mean_token_accuracy": 0.8535549640655518,
+      "step": 1183
+    },
+    {
+      "epoch": 3.7232704402515724,
+      "grad_norm": 0.13213631510734558,
+      "learning_rate": 3.0031013643172185e-05,
+      "loss": 0.3159,
+      "mean_token_accuracy": 0.8608952164649963,
+      "step": 1184
+    },
+    {
+      "epoch": 3.7264150943396226,
+      "grad_norm": 0.1514212042093277,
+      "learning_rate": 3.0014599556096557e-05,
+      "loss": 0.3384,
+      "mean_token_accuracy": 0.8577061295509338,
+      "step": 1185
+    },
+    {
+      "epoch": 3.729559748427673,
+      "grad_norm": 0.14160504937171936,
+      "learning_rate": 2.9998177152287204e-05,
+      "loss": 0.3297,
+      "mean_token_accuracy": 0.8586458563804626,
+      "step": 1186
+    },
+    {
+      "epoch": 3.732704402515723,
+      "grad_norm": 0.14224717020988464,
+      "learning_rate": 2.9981746448785615e-05,
+      "loss": 0.3203,
+      "mean_token_accuracy": 0.8611958622932434,
+      "step": 1187
+    },
+    {
+      "epoch": 3.7358490566037736,
+      "grad_norm": 0.12940874695777893,
+      "learning_rate": 2.9965307462641906e-05,
+      "loss": 0.3384,
+      "mean_token_accuracy": 0.8546769618988037,
+      "step": 1188
+    },
+    {
+      "epoch": 3.738993710691824,
+      "grad_norm": 0.1318395733833313,
+      "learning_rate": 2.9948860210914773e-05,
+      "loss": 0.3339,
+      "mean_token_accuracy": 0.8579802513122559,
+      "step": 1189
+    },
+    {
+      "epoch": 3.742138364779874,
+      "grad_norm": 0.14659515023231506,
+      "learning_rate": 2.9932404710671504e-05,
+      "loss": 0.3371,
+      "mean_token_accuracy": 0.8568556308746338,
+      "step": 1190
+    },
+    {
+      "epoch": 3.7452830188679247,
+      "grad_norm": 0.1241716668009758,
+      "learning_rate": 2.9915940978987926e-05,
+      "loss": 0.3199,
+      "mean_token_accuracy": 0.8621662855148315,
+      "step": 1191
+    },
+    {
+      "epoch": 3.748427672955975,
+      "grad_norm": 0.14975996315479279,
+      "learning_rate": 2.989946903294843e-05,
+      "loss": 0.3447,
+      "mean_token_accuracy": 0.8569864630699158,
+      "step": 1192
+    },
+    {
+      "epoch": 3.751572327044025,
+      "grad_norm": 0.12985654175281525,
+      "learning_rate": 2.9882988889645913e-05,
+      "loss": 0.3268,
+      "mean_token_accuracy": 0.8582627773284912,
+      "step": 1193
+    },
+    {
+      "epoch": 3.7547169811320753,
+      "grad_norm": 0.14152680337429047,
+      "learning_rate": 2.9866500566181802e-05,
+      "loss": 0.3121,
+      "mean_token_accuracy": 0.8632294535636902,
+      "step": 1194
+    },
+    {
+      "epoch": 3.757861635220126,
+      "grad_norm": 0.14638929069042206,
+      "learning_rate": 2.985000407966598e-05,
+      "loss": 0.3317,
+      "mean_token_accuracy": 0.8586151003837585,
+      "step": 1195
+    },
+    {
+      "epoch": 3.761006289308176,
+      "grad_norm": 0.1430143564939499,
+      "learning_rate": 2.9833499447216824e-05,
+      "loss": 0.3387,
+      "mean_token_accuracy": 0.8545314073562622,
+      "step": 1196
+    },
+    {
+      "epoch": 3.7641509433962264,
+      "grad_norm": 0.14943763613700867,
+      "learning_rate": 2.9816986685961156e-05,
+      "loss": 0.3322,
+      "mean_token_accuracy": 0.858390748500824,
+      "step": 1197
+    },
+    {
+      "epoch": 3.767295597484277,
+      "grad_norm": 0.14059577882289886,
+      "learning_rate": 2.9800465813034242e-05,
+      "loss": 0.3294,
+      "mean_token_accuracy": 0.8604702949523926,
+      "step": 1198
+    },
+    {
+      "epoch": 3.770440251572327,
+      "grad_norm": 0.15087758004665375,
+      "learning_rate": 2.9783936845579747e-05,
+      "loss": 0.3381,
+      "mean_token_accuracy": 0.8561389446258545,
+      "step": 1199
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "grad_norm": 0.14814791083335876,
+      "learning_rate": 2.9767399800749755e-05,
+      "loss": 0.3311,
+      "mean_token_accuracy": 0.8588146567344666,
+      "step": 1200
+    },
+    {
+      "epoch": 3.7767295597484276,
+      "grad_norm": 0.13886821269989014,
+      "learning_rate": 2.9750854695704715e-05,
+      "loss": 0.332,
+      "mean_token_accuracy": 0.8596969842910767,
+      "step": 1201
+    },
+    {
+      "epoch": 3.779874213836478,
+      "grad_norm": 0.15242241322994232,
+      "learning_rate": 2.9734301547613465e-05,
+      "loss": 0.3277,
+      "mean_token_accuracy": 0.8580199480056763,
+      "step": 1202
+    },
+    {
+      "epoch": 3.7830188679245285,
+      "grad_norm": 0.13634704053401947,
+      "learning_rate": 2.971774037365316e-05,
+      "loss": 0.3357,
+      "mean_token_accuracy": 0.8593721985816956,
+      "step": 1203
+    },
+    {
+      "epoch": 3.7861635220125787,
+      "grad_norm": 0.14197935163974762,
+      "learning_rate": 2.9701171191009304e-05,
+      "loss": 0.3378,
+      "mean_token_accuracy": 0.8571075797080994,
+      "step": 1204
+    },
+    {
+      "epoch": 3.789308176100629,
+      "grad_norm": 0.1357307732105255,
+      "learning_rate": 2.96845940168757e-05,
+      "loss": 0.3152,
+      "mean_token_accuracy": 0.8626152276992798,
+      "step": 1205
+    },
+    {
+      "epoch": 3.7924528301886795,
+      "grad_norm": 0.1317468136548996,
+      "learning_rate": 2.966800886845445e-05,
+      "loss": 0.3392,
+      "mean_token_accuracy": 0.8555001616477966,
+      "step": 1206
+    },
+    {
+      "epoch": 3.7955974842767297,
+      "grad_norm": 0.14835762977600098,
+      "learning_rate": 2.9651415762955925e-05,
+      "loss": 0.344,
+      "mean_token_accuracy": 0.8575615882873535,
+      "step": 1207
+    },
+    {
+      "epoch": 3.79874213836478,
+      "grad_norm": 0.14677360653877258,
+      "learning_rate": 2.9634814717598762e-05,
+      "loss": 0.3227,
+      "mean_token_accuracy": 0.8616207838058472,
+      "step": 1208
+    },
+    {
+      "epoch": 3.80188679245283,
+      "grad_norm": 0.13944582641124725,
+      "learning_rate": 2.961820574960982e-05,
+      "loss": 0.3467,
+      "mean_token_accuracy": 0.8558523058891296,
+      "step": 1209
+    },
+    {
+      "epoch": 3.8050314465408803,
+      "grad_norm": 0.1433691531419754,
+      "learning_rate": 2.9601588876224215e-05,
+      "loss": 0.3417,
+      "mean_token_accuracy": 0.8568807244300842,
+      "step": 1210
+    },
+    {
+      "epoch": 3.808176100628931,
+      "grad_norm": 0.14389808475971222,
+      "learning_rate": 2.958496411468522e-05,
+      "loss": 0.3118,
+      "mean_token_accuracy": 0.8629230260848999,
+      "step": 1211
+    },
+    {
+      "epoch": 3.811320754716981,
+      "grad_norm": 0.14503751695156097,
+      "learning_rate": 2.956833148224433e-05,
+      "loss": 0.3369,
+      "mean_token_accuracy": 0.8592314720153809,
+      "step": 1212
+    },
+    {
+      "epoch": 3.8144654088050314,
+      "grad_norm": 0.14210133254528046,
+      "learning_rate": 2.955169099616119e-05,
+      "loss": 0.3301,
+      "mean_token_accuracy": 0.8592085242271423,
+      "step": 1213
+    },
+    {
+      "epoch": 3.817610062893082,
+      "grad_norm": 0.12544848024845123,
+      "learning_rate": 2.95350426737036e-05,
+      "loss": 0.3211,
+      "mean_token_accuracy": 0.8602308034896851,
+      "step": 1214
+    },
+    {
+      "epoch": 3.8207547169811322,
+      "grad_norm": 0.15146970748901367,
+      "learning_rate": 2.9518386532147487e-05,
+      "loss": 0.3319,
+      "mean_token_accuracy": 0.857859194278717,
+      "step": 1215
+    },
+    {
+      "epoch": 3.8238993710691824,
+      "grad_norm": 0.13156718015670776,
+      "learning_rate": 2.95017225887769e-05,
+      "loss": 0.3218,
+      "mean_token_accuracy": 0.8625040650367737,
+      "step": 1216
+    },
+    {
+      "epoch": 3.8270440251572326,
+      "grad_norm": 0.14363285899162292,
+      "learning_rate": 2.948505086088397e-05,
+      "loss": 0.3457,
+      "mean_token_accuracy": 0.8564530611038208,
+      "step": 1217
+    },
+    {
+      "epoch": 3.830188679245283,
+      "grad_norm": 0.12839969992637634,
+      "learning_rate": 2.9468371365768926e-05,
+      "loss": 0.3099,
+      "mean_token_accuracy": 0.8643601536750793,
+      "step": 1218
+    },
+    {
+      "epoch": 3.8333333333333335,
+      "grad_norm": 0.12974990904331207,
+      "learning_rate": 2.9451684120740038e-05,
+      "loss": 0.3274,
+      "mean_token_accuracy": 0.8582882285118103,
+      "step": 1219
+    },
+    {
+      "epoch": 3.8364779874213837,
+      "grad_norm": 0.14015363156795502,
+      "learning_rate": 2.943498914311364e-05,
+      "loss": 0.3368,
+      "mean_token_accuracy": 0.8556147217750549,
+      "step": 1220
+    },
+    {
+      "epoch": 3.839622641509434,
+      "grad_norm": 0.13860003650188446,
+      "learning_rate": 2.941828645021406e-05,
+      "loss": 0.3375,
+      "mean_token_accuracy": 0.8543303608894348,
+      "step": 1221
+    },
+    {
+      "epoch": 3.8427672955974845,
+      "grad_norm": 0.14258892834186554,
+      "learning_rate": 2.9401576059373656e-05,
+      "loss": 0.3259,
+      "mean_token_accuracy": 0.8594518303871155,
+      "step": 1222
+    },
+    {
+      "epoch": 3.8459119496855347,
+      "grad_norm": 0.13030867278575897,
+      "learning_rate": 2.9384857987932768e-05,
+      "loss": 0.3162,
+      "mean_token_accuracy": 0.8623016476631165,
+      "step": 1223
+    },
+    {
+      "epoch": 3.849056603773585,
+      "grad_norm": 0.13517674803733826,
+      "learning_rate": 2.9368132253239702e-05,
+      "loss": 0.3254,
+      "mean_token_accuracy": 0.8589973449707031,
+      "step": 1224
+    },
+    {
+      "epoch": 3.852201257861635,
+      "grad_norm": 0.12886177003383636,
+      "learning_rate": 2.935139887265072e-05,
+      "loss": 0.3354,
+      "mean_token_accuracy": 0.8530575037002563,
+      "step": 1225
+    },
+    {
+      "epoch": 3.8553459119496853,
+      "grad_norm": 0.14796893298625946,
+      "learning_rate": 2.9334657863530016e-05,
+      "loss": 0.3325,
+      "mean_token_accuracy": 0.8576770424842834,
+      "step": 1226
+    },
+    {
+      "epoch": 3.858490566037736,
+      "grad_norm": 0.1297404170036316,
+      "learning_rate": 2.9317909243249706e-05,
+      "loss": 0.3436,
+      "mean_token_accuracy": 0.8564417958259583,
+      "step": 1227
+    },
+    {
+      "epoch": 3.861635220125786,
+      "grad_norm": 0.1457653045654297,
+      "learning_rate": 2.9301153029189794e-05,
+      "loss": 0.3387,
+      "mean_token_accuracy": 0.8541455268859863,
+      "step": 1228
+    },
+    {
+      "epoch": 3.8647798742138364,
+      "grad_norm": 0.135748028755188,
+      "learning_rate": 2.928438923873816e-05,
+      "loss": 0.3194,
+      "mean_token_accuracy": 0.8600180745124817,
+      "step": 1229
+    },
+    {
+      "epoch": 3.867924528301887,
+      "grad_norm": 0.14161410927772522,
+      "learning_rate": 2.926761788929058e-05,
+      "loss": 0.3369,
+      "mean_token_accuracy": 0.8556387424468994,
+      "step": 1230
+    },
+    {
+      "epoch": 3.8710691823899372,
+      "grad_norm": 0.13314643502235413,
+      "learning_rate": 2.9250838998250638e-05,
+      "loss": 0.3317,
+      "mean_token_accuracy": 0.8579115867614746,
+      "step": 1231
+    },
+    {
+      "epoch": 3.8742138364779874,
+      "grad_norm": 0.14150525629520416,
+      "learning_rate": 2.9234052583029746e-05,
+      "loss": 0.3222,
+      "mean_token_accuracy": 0.8584760427474976,
+      "step": 1232
+    },
+    {
+      "epoch": 3.8773584905660377,
+      "grad_norm": 0.13233336806297302,
+      "learning_rate": 2.9217258661047142e-05,
+      "loss": 0.3517,
+      "mean_token_accuracy": 0.852990448474884,
+      "step": 1233
+    },
+    {
+      "epoch": 3.880503144654088,
+      "grad_norm": 0.15107811987400055,
+      "learning_rate": 2.920045724972985e-05,
+      "loss": 0.3305,
+      "mean_token_accuracy": 0.8559724688529968,
+      "step": 1234
+    },
+    {
+      "epoch": 3.8836477987421385,
+      "grad_norm": 0.16093167662620544,
+      "learning_rate": 2.9183648366512648e-05,
+      "loss": 0.3315,
+      "mean_token_accuracy": 0.8561708927154541,
+      "step": 1235
+    },
+    {
+      "epoch": 3.8867924528301887,
+      "grad_norm": 0.14102764427661896,
+      "learning_rate": 2.9166832028838085e-05,
+      "loss": 0.332,
+      "mean_token_accuracy": 0.856883704662323,
+      "step": 1236
+    },
+    {
+      "epoch": 3.889937106918239,
+      "grad_norm": 0.15067261457443237,
+      "learning_rate": 2.915000825415644e-05,
+      "loss": 0.3338,
+      "mean_token_accuracy": 0.858379065990448,
+      "step": 1237
+    },
+    {
+      "epoch": 3.8930817610062896,
+      "grad_norm": 0.1495073437690735,
+      "learning_rate": 2.9133177059925715e-05,
+      "loss": 0.3218,
+      "mean_token_accuracy": 0.8599880933761597,
+      "step": 1238
+    },
+    {
+      "epoch": 3.8962264150943398,
+      "grad_norm": 0.14388205111026764,
+      "learning_rate": 2.9116338463611596e-05,
+      "loss": 0.348,
+      "mean_token_accuracy": 0.8538846373558044,
+      "step": 1239
+    },
+    {
+      "epoch": 3.89937106918239,
+      "grad_norm": 0.1329399049282074,
+      "learning_rate": 2.9099492482687478e-05,
+      "loss": 0.3267,
+      "mean_token_accuracy": 0.8591561317443848,
+      "step": 1240
+    },
+    {
+      "epoch": 3.90251572327044,
+      "grad_norm": 0.14387187361717224,
+      "learning_rate": 2.9082639134634378e-05,
+      "loss": 0.3285,
+      "mean_token_accuracy": 0.8577826023101807,
+      "step": 1241
+    },
+    {
+      "epoch": 3.9056603773584904,
+      "grad_norm": 0.1352883279323578,
+      "learning_rate": 2.9065778436941003e-05,
+      "loss": 0.3314,
+      "mean_token_accuracy": 0.8583676218986511,
+      "step": 1242
+    },
+    {
+      "epoch": 3.908805031446541,
+      "grad_norm": 0.1337868720293045,
+      "learning_rate": 2.904891040710365e-05,
+      "loss": 0.3267,
+      "mean_token_accuracy": 0.857197105884552,
+      "step": 1243
+    },
+    {
+      "epoch": 3.911949685534591,
+      "grad_norm": 0.13829989731311798,
+      "learning_rate": 2.903203506262624e-05,
+      "loss": 0.3357,
+      "mean_token_accuracy": 0.85539710521698,
+      "step": 1244
+    },
+    {
+      "epoch": 3.9150943396226414,
+      "grad_norm": 0.1476396769285202,
+      "learning_rate": 2.9015152421020296e-05,
+      "loss": 0.3344,
+      "mean_token_accuracy": 0.857477068901062,
+      "step": 1245
+    },
+    {
+      "epoch": 3.918238993710692,
+      "grad_norm": 0.14620055258274078,
+      "learning_rate": 2.899826249980489e-05,
+      "loss": 0.3423,
+      "mean_token_accuracy": 0.8560461401939392,
+      "step": 1246
+    },
+    {
+      "epoch": 3.9213836477987423,
+      "grad_norm": 0.12422426789999008,
+      "learning_rate": 2.898136531650666e-05,
+      "loss": 0.3546,
+      "mean_token_accuracy": 0.8542535901069641,
+      "step": 1247
+    },
+    {
+      "epoch": 3.9245283018867925,
+      "grad_norm": 0.13504689931869507,
+      "learning_rate": 2.8964460888659786e-05,
+      "loss": 0.3391,
+      "mean_token_accuracy": 0.858195424079895,
+      "step": 1248
+    },
+    {
+      "epoch": 3.9276729559748427,
+      "grad_norm": 0.13244280219078064,
+      "learning_rate": 2.8947549233805953e-05,
+      "loss": 0.3121,
+      "mean_token_accuracy": 0.8605561852455139,
+      "step": 1249
+    },
+    {
+      "epoch": 3.930817610062893,
+      "grad_norm": 0.13584861159324646,
+      "learning_rate": 2.893063036949435e-05,
+      "loss": 0.3304,
+      "mean_token_accuracy": 0.8578067421913147,
+      "step": 1250
+    },
+    {
+      "epoch": 3.9339622641509435,
+      "grad_norm": 0.14337222278118134,
+      "learning_rate": 2.891370431328165e-05,
+      "loss": 0.3347,
+      "mean_token_accuracy": 0.8570818305015564,
+      "step": 1251
+    },
+    {
+      "epoch": 3.9371069182389937,
+      "grad_norm": 0.1409681886434555,
+      "learning_rate": 2.8896771082731986e-05,
+      "loss": 0.3374,
+      "mean_token_accuracy": 0.8561692833900452,
+      "step": 1252
+    },
+    {
+      "epoch": 3.940251572327044,
+      "grad_norm": 0.13311010599136353,
+      "learning_rate": 2.8879830695416933e-05,
+      "loss": 0.3386,
+      "mean_token_accuracy": 0.8557576537132263,
+      "step": 1253
+    },
+    {
+      "epoch": 3.9433962264150946,
+      "grad_norm": 0.13869957625865936,
+      "learning_rate": 2.8862883168915508e-05,
+      "loss": 0.3236,
+      "mean_token_accuracy": 0.8604661822319031,
+      "step": 1254
+    },
+    {
+      "epoch": 3.9465408805031448,
+      "grad_norm": 0.15289968252182007,
+      "learning_rate": 2.884592852081412e-05,
+      "loss": 0.3256,
+      "mean_token_accuracy": 0.8590101003646851,
+      "step": 1255
+    },
+    {
+      "epoch": 3.949685534591195,
+      "grad_norm": 0.12691056728363037,
+      "learning_rate": 2.882896676870657e-05,
+      "loss": 0.327,
+      "mean_token_accuracy": 0.8594552874565125,
+      "step": 1256
+    },
+    {
+      "epoch": 3.952830188679245,
+      "grad_norm": 0.13348102569580078,
+      "learning_rate": 2.8811997930194032e-05,
+      "loss": 0.3335,
+      "mean_token_accuracy": 0.8576066493988037,
+      "step": 1257
+    },
+    {
+      "epoch": 3.9559748427672954,
+      "grad_norm": 0.13760992884635925,
+      "learning_rate": 2.8795022022885043e-05,
+      "loss": 0.3247,
+      "mean_token_accuracy": 0.8583594560623169,
+      "step": 1258
+    },
+    {
+      "epoch": 3.959119496855346,
+      "grad_norm": 0.12815621495246887,
+      "learning_rate": 2.8778039064395464e-05,
+      "loss": 0.332,
+      "mean_token_accuracy": 0.8571073412895203,
+      "step": 1259
+    },
+    {
+      "epoch": 3.9622641509433962,
+      "grad_norm": 0.14194735884666443,
+      "learning_rate": 2.8761049072348478e-05,
+      "loss": 0.3321,
+      "mean_token_accuracy": 0.8575592637062073,
+      "step": 1260
+    },
+    {
+      "epoch": 3.9654088050314464,
+      "grad_norm": 0.12652727961540222,
+      "learning_rate": 2.874405206437455e-05,
+      "loss": 0.3126,
+      "mean_token_accuracy": 0.8608962297439575,
+      "step": 1261
+    },
+    {
+      "epoch": 3.968553459119497,
+      "grad_norm": 0.13028152287006378,
+      "learning_rate": 2.8727048058111467e-05,
+      "loss": 0.3395,
+      "mean_token_accuracy": 0.8562440872192383,
+      "step": 1262
+    },
+    {
+      "epoch": 3.9716981132075473,
+      "grad_norm": 0.1258905678987503,
+      "learning_rate": 2.8710037071204235e-05,
+      "loss": 0.3402,
+      "mean_token_accuracy": 0.8553774356842041,
+      "step": 1263
+    },
+    {
+      "epoch": 3.9748427672955975,
+      "grad_norm": 0.1402544230222702,
+      "learning_rate": 2.8693019121305123e-05,
+      "loss": 0.3403,
+      "mean_token_accuracy": 0.8573205471038818,
+      "step": 1264
+    },
+    {
+      "epoch": 3.9779874213836477,
+      "grad_norm": 0.13947366178035736,
+      "learning_rate": 2.867599422607363e-05,
+      "loss": 0.3292,
+      "mean_token_accuracy": 0.8590210676193237,
+      "step": 1265
+    },
+    {
+      "epoch": 3.981132075471698,
+      "grad_norm": 0.13215205073356628,
+      "learning_rate": 2.865896240317645e-05,
+      "loss": 0.3266,
+      "mean_token_accuracy": 0.8595618605613708,
+      "step": 1266
+    },
+    {
+      "epoch": 3.9842767295597485,
+      "grad_norm": 0.1323062777519226,
+      "learning_rate": 2.8641923670287465e-05,
+      "loss": 0.3276,
+      "mean_token_accuracy": 0.8571467995643616,
+      "step": 1267
+    },
+    {
+      "epoch": 3.9874213836477987,
+      "grad_norm": 0.12528589367866516,
+      "learning_rate": 2.8624878045087744e-05,
+      "loss": 0.3238,
+      "mean_token_accuracy": 0.8595758080482483,
+      "step": 1268
+    },
+    {
+      "epoch": 3.990566037735849,
+      "grad_norm": 0.14454223215579987,
+      "learning_rate": 2.8607825545265492e-05,
+      "loss": 0.3193,
+      "mean_token_accuracy": 0.8578279614448547,
+      "step": 1269
+    },
+    {
+      "epoch": 3.9937106918238996,
+      "grad_norm": 0.13367754220962524,
+      "learning_rate": 2.8590766188516047e-05,
+      "loss": 0.324,
+      "mean_token_accuracy": 0.8594267964363098,
+      "step": 1270
+    },
+    {
+      "epoch": 3.99685534591195,
+      "grad_norm": 0.14187419414520264,
+      "learning_rate": 2.8573699992541892e-05,
+      "loss": 0.3318,
+      "mean_token_accuracy": 0.8586729168891907,
+      "step": 1271
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.1359594166278839,
+      "learning_rate": 2.8556626975052563e-05,
+      "loss": 0.3187,
+      "mean_token_accuracy": 0.8581359386444092,
+      "step": 1272
+    },
+    {
+      "epoch": 4.00314465408805,
+      "grad_norm": 0.21873971819877625,
+      "learning_rate": 2.85395471537647e-05,
+      "loss": 0.249,
+      "mean_token_accuracy": 0.883579432964325,
+      "step": 1273
+    },
+    {
+      "epoch": 4.0062893081761,
+      "grad_norm": 0.35204988718032837,
+      "learning_rate": 2.8522460546402002e-05,
+      "loss": 0.2546,
+      "mean_token_accuracy": 0.8850920796394348,
+      "step": 1274
+    },
+    {
+      "epoch": 4.009433962264151,
+      "grad_norm": 0.26803094148635864,
+      "learning_rate": 2.8505367170695207e-05,
+      "loss": 0.2628,
+      "mean_token_accuracy": 0.8832299709320068,
+      "step": 1275
+    },
+    {
+      "epoch": 4.012578616352202,
+      "grad_norm": 0.3539370894432068,
+      "learning_rate": 2.8488267044382073e-05,
+      "loss": 0.2587,
+      "mean_token_accuracy": 0.886466920375824,
+      "step": 1276
+    },
+    {
+      "epoch": 4.015723270440252,
+      "grad_norm": 0.2860783636569977,
+      "learning_rate": 2.847116018520737e-05,
+      "loss": 0.2652,
+      "mean_token_accuracy": 0.8802006840705872,
+      "step": 1277
+    },
+    {
+      "epoch": 4.018867924528302,
+      "grad_norm": 0.19191104173660278,
+      "learning_rate": 2.8454046610922847e-05,
+      "loss": 0.2587,
+      "mean_token_accuracy": 0.884539008140564,
+      "step": 1278
+    },
+    {
+      "epoch": 4.022012578616352,
+      "grad_norm": 0.2441309690475464,
+      "learning_rate": 2.843692633928723e-05,
+      "loss": 0.2266,
+      "mean_token_accuracy": 0.8915911912918091,
+      "step": 1279
+    },
+    {
+      "epoch": 4.0251572327044025,
+      "grad_norm": 0.17421981692314148,
+      "learning_rate": 2.8419799388066182e-05,
+      "loss": 0.2455,
+      "mean_token_accuracy": 0.8853268623352051,
+      "step": 1280
+    },
+    {
+      "epoch": 4.028301886792453,
+      "grad_norm": 0.20515094697475433,
+      "learning_rate": 2.840266577503232e-05,
+      "loss": 0.2521,
+      "mean_token_accuracy": 0.886737585067749,
+      "step": 1281
+    },
+    {
+      "epoch": 4.031446540880503,
+      "grad_norm": 0.1697227954864502,
+      "learning_rate": 2.8385525517965143e-05,
+      "loss": 0.2502,
+      "mean_token_accuracy": 0.8833692669868469,
+      "step": 1282
+    },
+    {
+      "epoch": 4.034591194968553,
+      "grad_norm": 0.19101938605308533,
+      "learning_rate": 2.836837863465107e-05,
+      "loss": 0.2615,
+      "mean_token_accuracy": 0.8790598511695862,
+      "step": 1283
+    },
+    {
+      "epoch": 4.037735849056604,
+      "grad_norm": 0.16805845499038696,
+      "learning_rate": 2.8351225142883395e-05,
+      "loss": 0.2596,
+      "mean_token_accuracy": 0.8821934461593628,
+      "step": 1284
+    },
+    {
+      "epoch": 4.040880503144654,
+      "grad_norm": 0.18171757459640503,
+      "learning_rate": 2.8334065060462246e-05,
+      "loss": 0.2338,
+      "mean_token_accuracy": 0.8911201357841492,
+      "step": 1285
+    },
+    {
+      "epoch": 4.044025157232705,
+      "grad_norm": 0.17769482731819153,
+      "learning_rate": 2.8316898405194617e-05,
+      "loss": 0.2514,
+      "mean_token_accuracy": 0.8844525814056396,
+      "step": 1286
+    },
+    {
+      "epoch": 4.047169811320755,
+      "grad_norm": 0.16998951137065887,
+      "learning_rate": 2.829972519489431e-05,
+      "loss": 0.24,
+      "mean_token_accuracy": 0.8891401290893555,
+      "step": 1287
+    },
+    {
+      "epoch": 4.050314465408805,
+      "grad_norm": 0.1702141910791397,
+      "learning_rate": 2.828254544738192e-05,
+      "loss": 0.2489,
+      "mean_token_accuracy": 0.8864291310310364,
+      "step": 1288
+    },
+    {
+      "epoch": 4.053459119496855,
+      "grad_norm": 0.16874520480632782,
+      "learning_rate": 2.826535918048484e-05,
+      "loss": 0.2406,
+      "mean_token_accuracy": 0.8878934383392334,
+      "step": 1289
+    },
+    {
+      "epoch": 4.056603773584905,
+      "grad_norm": 0.15455903112888336,
+      "learning_rate": 2.8248166412037238e-05,
+      "loss": 0.2479,
+      "mean_token_accuracy": 0.8852613568305969,
+      "step": 1290
+    },
+    {
+      "epoch": 4.059748427672956,
+      "grad_norm": 0.16695955395698547,
+      "learning_rate": 2.8230967159879997e-05,
+      "loss": 0.23,
+      "mean_token_accuracy": 0.8919079303741455,
+      "step": 1291
+    },
+    {
+      "epoch": 4.062893081761007,
+      "grad_norm": 0.16394896805286407,
+      "learning_rate": 2.8213761441860765e-05,
+      "loss": 0.2713,
+      "mean_token_accuracy": 0.8791160583496094,
+      "step": 1292
+    },
+    {
+      "epoch": 4.066037735849057,
+      "grad_norm": 0.15975283086299896,
+      "learning_rate": 2.819654927583386e-05,
+      "loss": 0.2384,
+      "mean_token_accuracy": 0.8890205025672913,
+      "step": 1293
+    },
+    {
+      "epoch": 4.069182389937107,
+      "grad_norm": 0.17008937895298004,
+      "learning_rate": 2.817933067966033e-05,
+      "loss": 0.2517,
+      "mean_token_accuracy": 0.8825728893280029,
+      "step": 1294
+    },
+    {
+      "epoch": 4.072327044025157,
+      "grad_norm": 0.1542217880487442,
+      "learning_rate": 2.816210567120787e-05,
+      "loss": 0.2735,
+      "mean_token_accuracy": 0.879520833492279,
+      "step": 1295
+    },
+    {
+      "epoch": 4.0754716981132075,
+      "grad_norm": 0.17298361659049988,
+      "learning_rate": 2.8144874268350827e-05,
+      "loss": 0.2556,
+      "mean_token_accuracy": 0.8827891945838928,
+      "step": 1296
+    },
+    {
+      "epoch": 4.078616352201258,
+      "grad_norm": 0.15569692850112915,
+      "learning_rate": 2.8127636488970203e-05,
+      "loss": 0.2269,
+      "mean_token_accuracy": 0.8923097252845764,
+      "step": 1297
+    },
+    {
+      "epoch": 4.081761006289308,
+      "grad_norm": 0.1688852161169052,
+      "learning_rate": 2.8110392350953606e-05,
+      "loss": 0.2302,
+      "mean_token_accuracy": 0.8880395293235779,
+      "step": 1298
+    },
+    {
+      "epoch": 4.084905660377358,
+      "grad_norm": 0.1484537571668625,
+      "learning_rate": 2.8093141872195246e-05,
+      "loss": 0.2613,
+      "mean_token_accuracy": 0.8827516436576843,
+      "step": 1299
+    },
+    {
+      "epoch": 4.088050314465409,
+      "grad_norm": 0.16927222907543182,
+      "learning_rate": 2.8075885070595906e-05,
+      "loss": 0.2536,
+      "mean_token_accuracy": 0.8848833441734314,
+      "step": 1300
+    },
+    {
+      "epoch": 4.091194968553459,
+      "grad_norm": 0.14585481584072113,
+      "learning_rate": 2.805862196406293e-05,
+      "loss": 0.2578,
+      "mean_token_accuracy": 0.8819864988327026,
+      "step": 1301
+    },
+    {
+      "epoch": 4.09433962264151,
+      "grad_norm": 0.1680736243724823,
+      "learning_rate": 2.8041352570510216e-05,
+      "loss": 0.2529,
+      "mean_token_accuracy": 0.8855777382850647,
+      "step": 1302
+    },
+    {
+      "epoch": 4.09748427672956,
+      "grad_norm": 0.14265228807926178,
+      "learning_rate": 2.802407690785818e-05,
+      "loss": 0.2368,
+      "mean_token_accuracy": 0.8872923851013184,
+      "step": 1303
+    },
+    {
+      "epoch": 4.10062893081761,
+      "grad_norm": 0.15464213490486145,
+      "learning_rate": 2.8006794994033728e-05,
+      "loss": 0.2616,
+      "mean_token_accuracy": 0.8812229037284851,
+      "step": 1304
+    },
+    {
+      "epoch": 4.10377358490566,
+      "grad_norm": 0.14933174848556519,
+      "learning_rate": 2.7989506846970284e-05,
+      "loss": 0.2466,
+      "mean_token_accuracy": 0.8838765621185303,
+      "step": 1305
+    },
+    {
+      "epoch": 4.1069182389937104,
+      "grad_norm": 0.1553312987089157,
+      "learning_rate": 2.7972212484607708e-05,
+      "loss": 0.2505,
+      "mean_token_accuracy": 0.8844928741455078,
+      "step": 1306
+    },
+    {
+      "epoch": 4.110062893081761,
+      "grad_norm": 0.15248145163059235,
+      "learning_rate": 2.7954911924892344e-05,
+      "loss": 0.2452,
+      "mean_token_accuracy": 0.8878949880599976,
+      "step": 1307
+    },
+    {
+      "epoch": 4.113207547169812,
+      "grad_norm": 0.14316493272781372,
+      "learning_rate": 2.793760518577693e-05,
+      "loss": 0.2567,
+      "mean_token_accuracy": 0.8830364942550659,
+      "step": 1308
+    },
+    {
+      "epoch": 4.116352201257862,
+      "grad_norm": 0.15453389286994934,
+      "learning_rate": 2.792029228522064e-05,
+      "loss": 0.2372,
+      "mean_token_accuracy": 0.8906042575836182,
+      "step": 1309
+    },
+    {
+      "epoch": 4.119496855345912,
+      "grad_norm": 0.14568613469600677,
+      "learning_rate": 2.7902973241189037e-05,
+      "loss": 0.254,
+      "mean_token_accuracy": 0.8863977789878845,
+      "step": 1310
+    },
+    {
+      "epoch": 4.122641509433962,
+      "grad_norm": 0.14949272572994232,
+      "learning_rate": 2.788564807165406e-05,
+      "loss": 0.2367,
+      "mean_token_accuracy": 0.8867618441581726,
+      "step": 1311
+    },
+    {
+      "epoch": 4.1257861635220126,
+      "grad_norm": 0.15481707453727722,
+      "learning_rate": 2.7868316794594e-05,
+      "loss": 0.2428,
+      "mean_token_accuracy": 0.8883107900619507,
+      "step": 1312
+    },
+    {
+      "epoch": 4.128930817610063,
+      "grad_norm": 0.145402193069458,
+      "learning_rate": 2.7850979427993484e-05,
+      "loss": 0.2408,
+      "mean_token_accuracy": 0.887713611125946,
+      "step": 1313
+    },
+    {
+      "epoch": 4.132075471698113,
+      "grad_norm": 0.14378874003887177,
+      "learning_rate": 2.7833635989843474e-05,
+      "loss": 0.2601,
+      "mean_token_accuracy": 0.8857198357582092,
+      "step": 1314
+    },
+    {
+      "epoch": 4.135220125786163,
+      "grad_norm": 0.15104271471500397,
+      "learning_rate": 2.78162864981412e-05,
+      "loss": 0.252,
+      "mean_token_accuracy": 0.8861721754074097,
+      "step": 1315
+    },
+    {
+      "epoch": 4.138364779874214,
+      "grad_norm": 0.13934184610843658,
+      "learning_rate": 2.7798930970890216e-05,
+      "loss": 0.2509,
+      "mean_token_accuracy": 0.8866226077079773,
+      "step": 1316
+    },
+    {
+      "epoch": 4.1415094339622645,
+      "grad_norm": 0.14720742404460907,
+      "learning_rate": 2.7781569426100304e-05,
+      "loss": 0.2464,
+      "mean_token_accuracy": 0.8861123919487,
+      "step": 1317
+    },
+    {
+      "epoch": 4.144654088050315,
+      "grad_norm": 0.1393972486257553,
+      "learning_rate": 2.7764201881787512e-05,
+      "loss": 0.2655,
+      "mean_token_accuracy": 0.8817521929740906,
+      "step": 1318
+    },
+    {
+      "epoch": 4.147798742138365,
+      "grad_norm": 0.1448722630739212,
+      "learning_rate": 2.7746828355974104e-05,
+      "loss": 0.2364,
+      "mean_token_accuracy": 0.8872301578521729,
+      "step": 1319
+    },
+    {
+      "epoch": 4.150943396226415,
+      "grad_norm": 0.13423529267311096,
+      "learning_rate": 2.7729448866688544e-05,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.8843175768852234,
+      "step": 1320
+    },
+    {
+      "epoch": 4.154088050314465,
+      "grad_norm": 0.13578058779239655,
+      "learning_rate": 2.771206343196551e-05,
+      "loss": 0.2603,
+      "mean_token_accuracy": 0.8812825083732605,
+      "step": 1321
+    },
+    {
+      "epoch": 4.1572327044025155,
+      "grad_norm": 0.1407482773065567,
+      "learning_rate": 2.7694672069845813e-05,
+      "loss": 0.2423,
+      "mean_token_accuracy": 0.8843265771865845,
+      "step": 1322
+    },
+    {
+      "epoch": 4.160377358490566,
+      "grad_norm": 0.14086057245731354,
+      "learning_rate": 2.767727479837644e-05,
+      "loss": 0.2347,
+      "mean_token_accuracy": 0.8889259099960327,
+      "step": 1323
+    },
+    {
+      "epoch": 4.163522012578617,
+      "grad_norm": 0.14522001147270203,
+      "learning_rate": 2.7659871635610515e-05,
+      "loss": 0.2498,
+      "mean_token_accuracy": 0.8857369422912598,
+      "step": 1324
+    },
+    {
+      "epoch": 4.166666666666667,
+      "grad_norm": 0.14944982528686523,
+      "learning_rate": 2.7642462599607252e-05,
+      "loss": 0.2333,
+      "mean_token_accuracy": 0.8869674205780029,
+      "step": 1325
+    },
+    {
+      "epoch": 4.169811320754717,
+      "grad_norm": 0.1487564593553543,
+      "learning_rate": 2.7625047708431976e-05,
+      "loss": 0.2352,
+      "mean_token_accuracy": 0.8852526545524597,
+      "step": 1326
+    },
+    {
+      "epoch": 4.172955974842767,
+      "grad_norm": 0.1437779664993286,
+      "learning_rate": 2.7607626980156092e-05,
+      "loss": 0.262,
+      "mean_token_accuracy": 0.8827253580093384,
+      "step": 1327
+    },
+    {
+      "epoch": 4.176100628930818,
+      "grad_norm": 0.1432497501373291,
+      "learning_rate": 2.7590200432857047e-05,
+      "loss": 0.2567,
+      "mean_token_accuracy": 0.8826262354850769,
+      "step": 1328
+    },
+    {
+      "epoch": 4.179245283018868,
+      "grad_norm": 0.14890795946121216,
+      "learning_rate": 2.7572768084618334e-05,
+      "loss": 0.2688,
+      "mean_token_accuracy": 0.8793274760246277,
+      "step": 1329
+    },
+    {
+      "epoch": 4.182389937106918,
+      "grad_norm": 0.14550916850566864,
+      "learning_rate": 2.755532995352947e-05,
+      "loss": 0.233,
+      "mean_token_accuracy": 0.8903444409370422,
+      "step": 1330
+    },
+    {
+      "epoch": 4.185534591194968,
+      "grad_norm": 0.18591845035552979,
+      "learning_rate": 2.753788605768596e-05,
+      "loss": 0.2596,
+      "mean_token_accuracy": 0.8813364505767822,
+      "step": 1331
+    },
+    {
+      "epoch": 4.188679245283019,
+      "grad_norm": 0.13862967491149902,
+      "learning_rate": 2.7520436415189306e-05,
+      "loss": 0.2622,
+      "mean_token_accuracy": 0.8814187049865723,
+      "step": 1332
+    },
+    {
+      "epoch": 4.1918238993710695,
+      "grad_norm": 0.16407757997512817,
+      "learning_rate": 2.7502981044146963e-05,
+      "loss": 0.2579,
+      "mean_token_accuracy": 0.8837925791740417,
+      "step": 1333
+    },
+    {
+      "epoch": 4.19496855345912,
+      "grad_norm": 0.14519527554512024,
+      "learning_rate": 2.7485519962672337e-05,
+      "loss": 0.2695,
+      "mean_token_accuracy": 0.8818954825401306,
+      "step": 1334
+    },
+    {
+      "epoch": 4.19811320754717,
+      "grad_norm": 0.14345420897006989,
+      "learning_rate": 2.746805318888476e-05,
+      "loss": 0.2494,
+      "mean_token_accuracy": 0.8830834627151489,
+      "step": 1335
+    },
+    {
+      "epoch": 4.20125786163522,
+      "grad_norm": 0.1446036845445633,
+      "learning_rate": 2.7450580740909463e-05,
+      "loss": 0.2498,
+      "mean_token_accuracy": 0.8848795890808105,
+      "step": 1336
+    },
+    {
+      "epoch": 4.20440251572327,
+      "grad_norm": 0.145370215177536,
+      "learning_rate": 2.743310263687757e-05,
+      "loss": 0.268,
+      "mean_token_accuracy": 0.8806397914886475,
+      "step": 1337
+    },
+    {
+      "epoch": 4.2075471698113205,
+      "grad_norm": 0.13711602985858917,
+      "learning_rate": 2.7415618894926072e-05,
+      "loss": 0.23,
+      "mean_token_accuracy": 0.8897081613540649,
+      "step": 1338
+    },
+    {
+      "epoch": 4.210691823899371,
+      "grad_norm": 0.15641474723815918,
+      "learning_rate": 2.739812953319782e-05,
+      "loss": 0.2406,
+      "mean_token_accuracy": 0.8866468071937561,
+      "step": 1339
+    },
+    {
+      "epoch": 4.213836477987422,
+      "grad_norm": 0.1482914835214615,
+      "learning_rate": 2.738063456984148e-05,
+      "loss": 0.2472,
+      "mean_token_accuracy": 0.885341227054596,
+      "step": 1340
+    },
+    {
+      "epoch": 4.216981132075472,
+      "grad_norm": 0.14941436052322388,
+      "learning_rate": 2.736313402301155e-05,
+      "loss": 0.2558,
+      "mean_token_accuracy": 0.8828262090682983,
+      "step": 1341
+    },
+    {
+      "epoch": 4.220125786163522,
+      "grad_norm": 0.14204715192317963,
+      "learning_rate": 2.73456279108683e-05,
+      "loss": 0.2591,
+      "mean_token_accuracy": 0.8831762671470642,
+      "step": 1342
+    },
+    {
+      "epoch": 4.223270440251572,
+      "grad_norm": 0.14757885038852692,
+      "learning_rate": 2.73281162515778e-05,
+      "loss": 0.2582,
+      "mean_token_accuracy": 0.8833774328231812,
+      "step": 1343
+    },
+    {
+      "epoch": 4.226415094339623,
+      "grad_norm": 0.14690524339675903,
+      "learning_rate": 2.7310599063311857e-05,
+      "loss": 0.258,
+      "mean_token_accuracy": 0.8827973008155823,
+      "step": 1344
+    },
+    {
+      "epoch": 4.229559748427673,
+      "grad_norm": 0.14617246389389038,
+      "learning_rate": 2.7293076364248015e-05,
+      "loss": 0.2412,
+      "mean_token_accuracy": 0.886563241481781,
+      "step": 1345
+    },
+    {
+      "epoch": 4.232704402515723,
+      "grad_norm": 0.13870076835155487,
+      "learning_rate": 2.7275548172569557e-05,
+      "loss": 0.2321,
+      "mean_token_accuracy": 0.8864620923995972,
+      "step": 1346
+    },
+    {
+      "epoch": 4.235849056603773,
+      "grad_norm": 0.15120095014572144,
+      "learning_rate": 2.7258014506465442e-05,
+      "loss": 0.2448,
+      "mean_token_accuracy": 0.8861029744148254,
+      "step": 1347
+    },
+    {
+      "epoch": 4.238993710691824,
+      "grad_norm": 0.15668298304080963,
+      "learning_rate": 2.724047538413033e-05,
+      "loss": 0.2577,
+      "mean_token_accuracy": 0.8828524947166443,
+      "step": 1348
+    },
+    {
+      "epoch": 4.2421383647798745,
+      "grad_norm": 0.13791413605213165,
+      "learning_rate": 2.7222930823764517e-05,
+      "loss": 0.2473,
+      "mean_token_accuracy": 0.8867437243461609,
+      "step": 1349
+    },
+    {
+      "epoch": 4.245283018867925,
+      "grad_norm": 0.15054844319820404,
+      "learning_rate": 2.7205380843573967e-05,
+      "loss": 0.2549,
+      "mean_token_accuracy": 0.8830904364585876,
+      "step": 1350
+    },
+    {
+      "epoch": 4.248427672955975,
+      "grad_norm": 0.14027497172355652,
+      "learning_rate": 2.718782546177026e-05,
+      "loss": 0.2551,
+      "mean_token_accuracy": 0.8864468336105347,
+      "step": 1351
+    },
+    {
+      "epoch": 4.251572327044025,
+      "grad_norm": 0.15007853507995605,
+      "learning_rate": 2.7170264696570586e-05,
+      "loss": 0.2437,
+      "mean_token_accuracy": 0.8871759176254272,
+      "step": 1352
+    },
+    {
+      "epoch": 4.254716981132075,
+      "grad_norm": 0.1608761101961136,
+      "learning_rate": 2.7152698566197703e-05,
+      "loss": 0.2557,
+      "mean_token_accuracy": 0.8846498131752014,
+      "step": 1353
+    },
+    {
+      "epoch": 4.2578616352201255,
+      "grad_norm": 0.13730347156524658,
+      "learning_rate": 2.7135127088879947e-05,
+      "loss": 0.2541,
+      "mean_token_accuracy": 0.8821074366569519,
+      "step": 1354
+    },
+    {
+      "epoch": 4.261006289308176,
+      "grad_norm": 0.14933264255523682,
+      "learning_rate": 2.7117550282851222e-05,
+      "loss": 0.2569,
+      "mean_token_accuracy": 0.8827639818191528,
+      "step": 1355
+    },
+    {
+      "epoch": 4.264150943396227,
+      "grad_norm": 0.15967227518558502,
+      "learning_rate": 2.709996816635093e-05,
+      "loss": 0.2457,
+      "mean_token_accuracy": 0.8853583335876465,
+      "step": 1356
+    },
+    {
+      "epoch": 4.267295597484277,
+      "grad_norm": 0.130757138133049,
+      "learning_rate": 2.7082380757623997e-05,
+      "loss": 0.2624,
+      "mean_token_accuracy": 0.8816043734550476,
+      "step": 1357
+    },
+    {
+      "epoch": 4.270440251572327,
+      "grad_norm": 0.17749525606632233,
+      "learning_rate": 2.7064788074920853e-05,
+      "loss": 0.2297,
+      "mean_token_accuracy": 0.8903690576553345,
+      "step": 1358
+    },
+    {
+      "epoch": 4.273584905660377,
+      "grad_norm": 0.14349985122680664,
+      "learning_rate": 2.7047190136497374e-05,
+      "loss": 0.2644,
+      "mean_token_accuracy": 0.8817373514175415,
+      "step": 1359
+    },
+    {
+      "epoch": 4.276729559748428,
+      "grad_norm": 0.15383297204971313,
+      "learning_rate": 2.702958696061492e-05,
+      "loss": 0.2536,
+      "mean_token_accuracy": 0.8856379985809326,
+      "step": 1360
+    },
+    {
+      "epoch": 4.279874213836478,
+      "grad_norm": 0.1427260786294937,
+      "learning_rate": 2.7011978565540258e-05,
+      "loss": 0.2431,
+      "mean_token_accuracy": 0.8847337365150452,
+      "step": 1361
+    },
+    {
+      "epoch": 4.283018867924528,
+      "grad_norm": 0.15544083714485168,
+      "learning_rate": 2.6994364969545596e-05,
+      "loss": 0.2447,
+      "mean_token_accuracy": 0.8849520087242126,
+      "step": 1362
+    },
+    {
+      "epoch": 4.286163522012578,
+      "grad_norm": 0.15651412308216095,
+      "learning_rate": 2.697674619090852e-05,
+      "loss": 0.2366,
+      "mean_token_accuracy": 0.8874123692512512,
+      "step": 1363
+    },
+    {
+      "epoch": 4.289308176100629,
+      "grad_norm": 0.15618132054805756,
+      "learning_rate": 2.6959122247911996e-05,
+      "loss": 0.255,
+      "mean_token_accuracy": 0.8844594359397888,
+      "step": 1364
+    },
+    {
+      "epoch": 4.2924528301886795,
+      "grad_norm": 0.14532870054244995,
+      "learning_rate": 2.694149315884436e-05,
+      "loss": 0.2429,
+      "mean_token_accuracy": 0.8856728076934814,
+      "step": 1365
+    },
+    {
+      "epoch": 4.29559748427673,
+      "grad_norm": 0.15094998478889465,
+      "learning_rate": 2.692385894199929e-05,
+      "loss": 0.257,
+      "mean_token_accuracy": 0.8823837637901306,
+      "step": 1366
+    },
+    {
+      "epoch": 4.29874213836478,
+      "grad_norm": 0.1429101526737213,
+      "learning_rate": 2.6906219615675756e-05,
+      "loss": 0.246,
+      "mean_token_accuracy": 0.8854039311408997,
+      "step": 1367
+    },
+    {
+      "epoch": 4.30188679245283,
+      "grad_norm": 0.1493634283542633,
+      "learning_rate": 2.6888575198178073e-05,
+      "loss": 0.2523,
+      "mean_token_accuracy": 0.8816664814949036,
+      "step": 1368
+    },
+    {
+      "epoch": 4.30503144654088,
+      "grad_norm": 0.1594749540090561,
+      "learning_rate": 2.6870925707815807e-05,
+      "loss": 0.2433,
+      "mean_token_accuracy": 0.8858900666236877,
+      "step": 1369
+    },
+    {
+      "epoch": 4.3081761006289305,
+      "grad_norm": 0.16246452927589417,
+      "learning_rate": 2.6853271162903792e-05,
+      "loss": 0.28,
+      "mean_token_accuracy": 0.8772223591804504,
+      "step": 1370
+    },
+    {
+      "epoch": 4.311320754716981,
+      "grad_norm": 0.15938788652420044,
+      "learning_rate": 2.6835611581762135e-05,
+      "loss": 0.2522,
+      "mean_token_accuracy": 0.8834171295166016,
+      "step": 1371
+    },
+    {
+      "epoch": 4.314465408805032,
+      "grad_norm": 0.1860852986574173,
+      "learning_rate": 2.6817946982716125e-05,
+      "loss": 0.2239,
+      "mean_token_accuracy": 0.8932704925537109,
+      "step": 1372
+    },
+    {
+      "epoch": 4.317610062893082,
+      "grad_norm": 0.1534428745508194,
+      "learning_rate": 2.6800277384096287e-05,
+      "loss": 0.2503,
+      "mean_token_accuracy": 0.8848916292190552,
+      "step": 1373
+    },
+    {
+      "epoch": 4.320754716981132,
+      "grad_norm": 0.19438529014587402,
+      "learning_rate": 2.6782602804238328e-05,
+      "loss": 0.2566,
+      "mean_token_accuracy": 0.8817089200019836,
+      "step": 1374
+    },
+    {
+      "epoch": 4.323899371069182,
+      "grad_norm": 0.1576819121837616,
+      "learning_rate": 2.6764923261483135e-05,
+      "loss": 0.252,
+      "mean_token_accuracy": 0.8846593499183655,
+      "step": 1375
+    },
+    {
+      "epoch": 4.327044025157233,
+      "grad_norm": 0.18220695853233337,
+      "learning_rate": 2.6747238774176717e-05,
+      "loss": 0.2499,
+      "mean_token_accuracy": 0.8850845098495483,
+      "step": 1376
+    },
+    {
+      "epoch": 4.330188679245283,
+      "grad_norm": 0.15748025476932526,
+      "learning_rate": 2.6729549360670244e-05,
+      "loss": 0.2523,
+      "mean_token_accuracy": 0.8865378499031067,
+      "step": 1377
+    },
+    {
+      "epoch": 4.333333333333333,
+      "grad_norm": 0.15679998695850372,
+      "learning_rate": 2.6711855039319972e-05,
+      "loss": 0.2525,
+      "mean_token_accuracy": 0.8859933614730835,
+      "step": 1378
+    },
+    {
+      "epoch": 4.336477987421383,
+      "grad_norm": 0.1435849666595459,
+      "learning_rate": 2.6694155828487272e-05,
+      "loss": 0.2553,
+      "mean_token_accuracy": 0.882286787033081,
+      "step": 1379
+    },
+    {
+      "epoch": 4.339622641509434,
+      "grad_norm": 0.14357006549835205,
+      "learning_rate": 2.6676451746538577e-05,
+      "loss": 0.2519,
+      "mean_token_accuracy": 0.8833568692207336,
+      "step": 1380
+    },
+    {
+      "epoch": 4.3427672955974845,
+      "grad_norm": 0.14986519515514374,
+      "learning_rate": 2.6658742811845377e-05,
+      "loss": 0.2494,
+      "mean_token_accuracy": 0.8818812370300293,
+      "step": 1381
+    },
+    {
+      "epoch": 4.345911949685535,
+      "grad_norm": 0.14738526940345764,
+      "learning_rate": 2.6641029042784197e-05,
+      "loss": 0.2504,
+      "mean_token_accuracy": 0.8836377859115601,
+      "step": 1382
+    },
+    {
+      "epoch": 4.349056603773585,
+      "grad_norm": 0.15075050294399261,
+      "learning_rate": 2.6623310457736575e-05,
+      "loss": 0.2638,
+      "mean_token_accuracy": 0.8816152811050415,
+      "step": 1383
+    },
+    {
+      "epoch": 4.352201257861635,
+      "grad_norm": 0.14811988174915314,
+      "learning_rate": 2.660558707508906e-05,
+      "loss": 0.2639,
+      "mean_token_accuracy": 0.8767325282096863,
+      "step": 1384
+    },
+    {
+      "epoch": 4.355345911949685,
+      "grad_norm": 0.13753147423267365,
+      "learning_rate": 2.6587858913233168e-05,
+      "loss": 0.2412,
+      "mean_token_accuracy": 0.8852753639221191,
+      "step": 1385
+    },
+    {
+      "epoch": 4.3584905660377355,
+      "grad_norm": 0.1575927436351776,
+      "learning_rate": 2.657012599056536e-05,
+      "loss": 0.252,
+      "mean_token_accuracy": 0.8817105293273926,
+      "step": 1386
+    },
+    {
+      "epoch": 4.361635220125786,
+      "grad_norm": 0.14534930884838104,
+      "learning_rate": 2.6552388325487078e-05,
+      "loss": 0.2562,
+      "mean_token_accuracy": 0.8825299143791199,
+      "step": 1387
+    },
+    {
+      "epoch": 4.364779874213837,
+      "grad_norm": 0.1528378576040268,
+      "learning_rate": 2.6534645936404655e-05,
+      "loss": 0.2606,
+      "mean_token_accuracy": 0.8781949877738953,
+      "step": 1388
+    },
+    {
+      "epoch": 4.367924528301887,
+      "grad_norm": 0.14020651578903198,
+      "learning_rate": 2.6516898841729323e-05,
+      "loss": 0.257,
+      "mean_token_accuracy": 0.8830549716949463,
+      "step": 1389
+    },
+    {
+      "epoch": 4.371069182389937,
+      "grad_norm": 0.13903018832206726,
+      "learning_rate": 2.6499147059877213e-05,
+      "loss": 0.2522,
+      "mean_token_accuracy": 0.8827590346336365,
+      "step": 1390
+    },
+    {
+      "epoch": 4.3742138364779874,
+      "grad_norm": 0.14772510528564453,
+      "learning_rate": 2.6481390609269308e-05,
+      "loss": 0.2625,
+      "mean_token_accuracy": 0.8802635073661804,
+      "step": 1391
+    },
+    {
+      "epoch": 4.377358490566038,
+      "grad_norm": 0.1616673320531845,
+      "learning_rate": 2.646362950833145e-05,
+      "loss": 0.2425,
+      "mean_token_accuracy": 0.8878819942474365,
+      "step": 1392
+    },
+    {
+      "epoch": 4.380503144654088,
+      "grad_norm": 0.13991199433803558,
+      "learning_rate": 2.6445863775494282e-05,
+      "loss": 0.2522,
+      "mean_token_accuracy": 0.8848397135734558,
+      "step": 1393
+    },
+    {
+      "epoch": 4.383647798742138,
+      "grad_norm": 0.1424325704574585,
+      "learning_rate": 2.6428093429193288e-05,
+      "loss": 0.2689,
+      "mean_token_accuracy": 0.8799664974212646,
+      "step": 1394
+    },
+    {
+      "epoch": 4.386792452830189,
+      "grad_norm": 0.1396665871143341,
+      "learning_rate": 2.6410318487868707e-05,
+      "loss": 0.2548,
+      "mean_token_accuracy": 0.8830381631851196,
+      "step": 1395
+    },
+    {
+      "epoch": 4.389937106918239,
+      "grad_norm": 0.14341147243976593,
+      "learning_rate": 2.6392538969965565e-05,
+      "loss": 0.243,
+      "mean_token_accuracy": 0.8879396319389343,
+      "step": 1396
+    },
+    {
+      "epoch": 4.3930817610062896,
+      "grad_norm": 0.1465020775794983,
+      "learning_rate": 2.637475489393363e-05,
+      "loss": 0.2391,
+      "mean_token_accuracy": 0.8867096900939941,
+      "step": 1397
+    },
+    {
+      "epoch": 4.39622641509434,
+      "grad_norm": 0.1537989377975464,
+      "learning_rate": 2.63569662782274e-05,
+      "loss": 0.2578,
+      "mean_token_accuracy": 0.8815436959266663,
+      "step": 1398
+    },
+    {
+      "epoch": 4.39937106918239,
+      "grad_norm": 0.1487364023923874,
+      "learning_rate": 2.6339173141306095e-05,
+      "loss": 0.2679,
+      "mean_token_accuracy": 0.8777388334274292,
+      "step": 1399
+    },
+    {
+      "epoch": 4.40251572327044,
+      "grad_norm": 0.15459056198596954,
+      "learning_rate": 2.6321375501633603e-05,
+      "loss": 0.2686,
+      "mean_token_accuracy": 0.8784300088882446,
+      "step": 1400
+    },
+    {
+      "epoch": 4.40566037735849,
+      "grad_norm": 0.139274001121521,
+      "learning_rate": 2.6303573377678513e-05,
+      "loss": 0.2573,
+      "mean_token_accuracy": 0.8845201134681702,
+      "step": 1401
+    },
+    {
+      "epoch": 4.408805031446541,
+      "grad_norm": 0.15442243218421936,
+      "learning_rate": 2.6285766787914034e-05,
+      "loss": 0.2497,
+      "mean_token_accuracy": 0.8847933411598206,
+      "step": 1402
+    },
+    {
+      "epoch": 4.411949685534591,
+      "grad_norm": 0.1377149522304535,
+      "learning_rate": 2.626795575081804e-05,
+      "loss": 0.2274,
+      "mean_token_accuracy": 0.8895265460014343,
+      "step": 1403
+    },
+    {
+      "epoch": 4.415094339622642,
+      "grad_norm": 0.14790353178977966,
+      "learning_rate": 2.6250140284873017e-05,
+      "loss": 0.2685,
+      "mean_token_accuracy": 0.8800943493843079,
+      "step": 1404
+    },
+    {
+      "epoch": 4.418238993710692,
+      "grad_norm": 0.13157331943511963,
+      "learning_rate": 2.6232320408566025e-05,
+      "loss": 0.2427,
+      "mean_token_accuracy": 0.8863032460212708,
+      "step": 1405
+    },
+    {
+      "epoch": 4.421383647798742,
+      "grad_norm": 0.14972056448459625,
+      "learning_rate": 2.6214496140388718e-05,
+      "loss": 0.2452,
+      "mean_token_accuracy": 0.8844586610794067,
+      "step": 1406
+    },
+    {
+      "epoch": 4.4245283018867925,
+      "grad_norm": 0.14638881385326385,
+      "learning_rate": 2.6196667498837302e-05,
+      "loss": 0.2597,
+      "mean_token_accuracy": 0.880418598651886,
+      "step": 1407
+    },
+    {
+      "epoch": 4.427672955974843,
+      "grad_norm": 0.1468118280172348,
+      "learning_rate": 2.617883450241252e-05,
+      "loss": 0.2518,
+      "mean_token_accuracy": 0.8848738670349121,
+      "step": 1408
+    },
+    {
+      "epoch": 4.430817610062893,
+      "grad_norm": 0.14775048196315765,
+      "learning_rate": 2.616099716961964e-05,
+      "loss": 0.2653,
+      "mean_token_accuracy": 0.8803516030311584,
+      "step": 1409
+    },
+    {
+      "epoch": 4.433962264150943,
+      "grad_norm": 0.14332270622253418,
+      "learning_rate": 2.6143155518968428e-05,
+      "loss": 0.2487,
+      "mean_token_accuracy": 0.8877217173576355,
+      "step": 1410
+    },
+    {
+      "epoch": 4.437106918238994,
+      "grad_norm": 0.14214986562728882,
+      "learning_rate": 2.612530956897312e-05,
+      "loss": 0.2705,
+      "mean_token_accuracy": 0.8784423470497131,
+      "step": 1411
+    },
+    {
+      "epoch": 4.440251572327044,
+      "grad_norm": 0.1403186023235321,
+      "learning_rate": 2.6107459338152427e-05,
+      "loss": 0.2665,
+      "mean_token_accuracy": 0.8789308071136475,
+      "step": 1412
+    },
+    {
+      "epoch": 4.443396226415095,
+      "grad_norm": 0.138310045003891,
+      "learning_rate": 2.60896048450295e-05,
+      "loss": 0.2499,
+      "mean_token_accuracy": 0.8843627572059631,
+      "step": 1413
+    },
+    {
+      "epoch": 4.446540880503145,
+      "grad_norm": 0.1412084996700287,
+      "learning_rate": 2.607174610813191e-05,
+      "loss": 0.2594,
+      "mean_token_accuracy": 0.8839447498321533,
+      "step": 1414
+    },
+    {
+      "epoch": 4.449685534591195,
+      "grad_norm": 0.14437635242938995,
+      "learning_rate": 2.605388314599163e-05,
+      "loss": 0.241,
+      "mean_token_accuracy": 0.8860413432121277,
+      "step": 1415
+    },
+    {
+      "epoch": 4.452830188679245,
+      "grad_norm": 0.14431603252887726,
+      "learning_rate": 2.6036015977145018e-05,
+      "loss": 0.256,
+      "mean_token_accuracy": 0.8797090649604797,
+      "step": 1416
+    },
+    {
+      "epoch": 4.455974842767295,
+      "grad_norm": 0.1461351066827774,
+      "learning_rate": 2.6018144620132805e-05,
+      "loss": 0.2438,
+      "mean_token_accuracy": 0.8827074766159058,
+      "step": 1417
+    },
+    {
+      "epoch": 4.459119496855346,
+      "grad_norm": 0.14454837143421173,
+      "learning_rate": 2.6000269093500055e-05,
+      "loss": 0.2541,
+      "mean_token_accuracy": 0.8819749355316162,
+      "step": 1418
+    },
+    {
+      "epoch": 4.462264150943396,
+      "grad_norm": 0.15199217200279236,
+      "learning_rate": 2.598238941579617e-05,
+      "loss": 0.2622,
+      "mean_token_accuracy": 0.8807674050331116,
+      "step": 1419
+    },
+    {
+      "epoch": 4.465408805031447,
+      "grad_norm": 0.1487475484609604,
+      "learning_rate": 2.596450560557485e-05,
+      "loss": 0.2574,
+      "mean_token_accuracy": 0.8806383013725281,
+      "step": 1420
+    },
+    {
+      "epoch": 4.468553459119497,
+      "grad_norm": 0.14922156929969788,
+      "learning_rate": 2.59466176813941e-05,
+      "loss": 0.2574,
+      "mean_token_accuracy": 0.8796212077140808,
+      "step": 1421
+    },
+    {
+      "epoch": 4.471698113207547,
+      "grad_norm": 0.13734014332294464,
+      "learning_rate": 2.5928725661816162e-05,
+      "loss": 0.2412,
+      "mean_token_accuracy": 0.8878635168075562,
+      "step": 1422
+    },
+    {
+      "epoch": 4.4748427672955975,
+      "grad_norm": 0.1519346535205841,
+      "learning_rate": 2.5910829565407573e-05,
+      "loss": 0.2607,
+      "mean_token_accuracy": 0.8802406191825867,
+      "step": 1423
+    },
+    {
+      "epoch": 4.477987421383648,
+      "grad_norm": 0.13957345485687256,
+      "learning_rate": 2.5892929410739066e-05,
+      "loss": 0.2437,
+      "mean_token_accuracy": 0.8885535001754761,
+      "step": 1424
+    },
+    {
+      "epoch": 4.481132075471698,
+      "grad_norm": 0.14830255508422852,
+      "learning_rate": 2.587502521638559e-05,
+      "loss": 0.2575,
+      "mean_token_accuracy": 0.8828555345535278,
+      "step": 1425
+    },
+    {
+      "epoch": 4.484276729559748,
+      "grad_norm": 0.13677102327346802,
+      "learning_rate": 2.5857117000926298e-05,
+      "loss": 0.2501,
+      "mean_token_accuracy": 0.8842593431472778,
+      "step": 1426
+    },
+    {
+      "epoch": 4.487421383647799,
+      "grad_norm": 0.1367623507976532,
+      "learning_rate": 2.5839204782944506e-05,
+      "loss": 0.2689,
+      "mean_token_accuracy": 0.8797082304954529,
+      "step": 1427
+    },
+    {
+      "epoch": 4.490566037735849,
+      "grad_norm": 0.1288725733757019,
+      "learning_rate": 2.582128858102769e-05,
+      "loss": 0.2567,
+      "mean_token_accuracy": 0.8836309909820557,
+      "step": 1428
+    },
+    {
+      "epoch": 4.4937106918239,
+      "grad_norm": 0.13104671239852905,
+      "learning_rate": 2.5803368413767443e-05,
+      "loss": 0.2503,
+      "mean_token_accuracy": 0.8835127949714661,
+      "step": 1429
+    },
+    {
+      "epoch": 4.49685534591195,
+      "grad_norm": 0.1426146924495697,
+      "learning_rate": 2.5785444299759504e-05,
+      "loss": 0.2556,
+      "mean_token_accuracy": 0.8816784620285034,
+      "step": 1430
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 0.13342611491680145,
+      "learning_rate": 2.576751625760368e-05,
+      "loss": 0.2584,
+      "mean_token_accuracy": 0.879668116569519,
+      "step": 1431
+    },
+    {
+      "epoch": 4.50314465408805,
+      "grad_norm": 0.15487535297870636,
+      "learning_rate": 2.5749584305903866e-05,
+      "loss": 0.2569,
+      "mean_token_accuracy": 0.8806970119476318,
+      "step": 1432
+    },
+    {
+      "epoch": 4.5062893081761,
+      "grad_norm": 0.14090071618556976,
+      "learning_rate": 2.5731648463268015e-05,
+      "loss": 0.2415,
+      "mean_token_accuracy": 0.8850799798965454,
+      "step": 1433
+    },
+    {
+      "epoch": 4.509433962264151,
+      "grad_norm": 0.14992307126522064,
+      "learning_rate": 2.571370874830811e-05,
+      "loss": 0.2522,
+      "mean_token_accuracy": 0.882178008556366,
+      "step": 1434
+    },
+    {
+      "epoch": 4.512578616352201,
+      "grad_norm": 0.13169358670711517,
+      "learning_rate": 2.569576517964016e-05,
+      "loss": 0.2544,
+      "mean_token_accuracy": 0.8801778554916382,
+      "step": 1435
+    },
+    {
+      "epoch": 4.515723270440252,
+      "grad_norm": 0.16984504461288452,
+      "learning_rate": 2.567781777588416e-05,
+      "loss": 0.2474,
+      "mean_token_accuracy": 0.8849585056304932,
+      "step": 1436
+    },
+    {
+      "epoch": 4.518867924528302,
+      "grad_norm": 0.1379225105047226,
+      "learning_rate": 2.565986655566411e-05,
+      "loss": 0.2451,
+      "mean_token_accuracy": 0.88438481092453,
+      "step": 1437
+    },
+    {
+      "epoch": 4.522012578616352,
+      "grad_norm": 0.18996167182922363,
+      "learning_rate": 2.5641911537607952e-05,
+      "loss": 0.2607,
+      "mean_token_accuracy": 0.8825079798698425,
+      "step": 1438
+    },
+    {
+      "epoch": 4.5251572327044025,
+      "grad_norm": 0.13947978615760803,
+      "learning_rate": 2.562395274034756e-05,
+      "loss": 0.2568,
+      "mean_token_accuracy": 0.8816097378730774,
+      "step": 1439
+    },
+    {
+      "epoch": 4.528301886792453,
+      "grad_norm": 0.20490330457687378,
+      "learning_rate": 2.5605990182518752e-05,
+      "loss": 0.2699,
+      "mean_token_accuracy": 0.8782700896263123,
+      "step": 1440
+    },
+    {
+      "epoch": 4.531446540880503,
+      "grad_norm": 0.1351514607667923,
+      "learning_rate": 2.5588023882761233e-05,
+      "loss": 0.2508,
+      "mean_token_accuracy": 0.8827007412910461,
+      "step": 1441
+    },
+    {
+      "epoch": 4.534591194968553,
+      "grad_norm": 0.18768662214279175,
+      "learning_rate": 2.5570053859718606e-05,
+      "loss": 0.2583,
+      "mean_token_accuracy": 0.8811473846435547,
+      "step": 1442
+    },
+    {
+      "epoch": 4.537735849056604,
+      "grad_norm": 0.14161916077136993,
+      "learning_rate": 2.5552080132038312e-05,
+      "loss": 0.239,
+      "mean_token_accuracy": 0.8842949867248535,
+      "step": 1443
+    },
+    {
+      "epoch": 4.540880503144654,
+      "grad_norm": 0.1600307822227478,
+      "learning_rate": 2.5534102718371664e-05,
+      "loss": 0.2711,
+      "mean_token_accuracy": 0.8794844746589661,
+      "step": 1444
+    },
+    {
+      "epoch": 4.544025157232705,
+      "grad_norm": 0.1379203200340271,
+      "learning_rate": 2.5516121637373782e-05,
+      "loss": 0.2464,
+      "mean_token_accuracy": 0.8868380188941956,
+      "step": 1445
+    },
+    {
+      "epoch": 4.547169811320755,
+      "grad_norm": 0.15503491461277008,
+      "learning_rate": 2.5498136907703594e-05,
+      "loss": 0.2582,
+      "mean_token_accuracy": 0.8819460868835449,
+      "step": 1446
+    },
+    {
+      "epoch": 4.550314465408805,
+      "grad_norm": 0.13813835382461548,
+      "learning_rate": 2.5480148548023823e-05,
+      "loss": 0.2668,
+      "mean_token_accuracy": 0.878455638885498,
+      "step": 1447
+    },
+    {
+      "epoch": 4.553459119496855,
+      "grad_norm": 0.1430351883172989,
+      "learning_rate": 2.5462156577000952e-05,
+      "loss": 0.2432,
+      "mean_token_accuracy": 0.8864967226982117,
+      "step": 1448
+    },
+    {
+      "epoch": 4.556603773584905,
+      "grad_norm": 0.14724938571453094,
+      "learning_rate": 2.5444161013305217e-05,
+      "loss": 0.2618,
+      "mean_token_accuracy": 0.8809149861335754,
+      "step": 1449
+    },
+    {
+      "epoch": 4.559748427672956,
+      "grad_norm": 0.132796511054039,
+      "learning_rate": 2.5426161875610572e-05,
+      "loss": 0.2667,
+      "mean_token_accuracy": 0.8805984854698181,
+      "step": 1450
+    },
+    {
+      "epoch": 4.562893081761006,
+      "grad_norm": 0.1508307307958603,
+      "learning_rate": 2.540815918259469e-05,
+      "loss": 0.2494,
+      "mean_token_accuracy": 0.8844383955001831,
+      "step": 1451
+    },
+    {
+      "epoch": 4.566037735849057,
+      "grad_norm": 0.13710395991802216,
+      "learning_rate": 2.539015295293893e-05,
+      "loss": 0.2698,
+      "mean_token_accuracy": 0.8794138431549072,
+      "step": 1452
+    },
+    {
+      "epoch": 4.569182389937107,
+      "grad_norm": 0.13251298666000366,
+      "learning_rate": 2.5372143205328306e-05,
+      "loss": 0.2413,
+      "mean_token_accuracy": 0.8852861523628235,
+      "step": 1453
+    },
+    {
+      "epoch": 4.572327044025157,
+      "grad_norm": 0.1437736451625824,
+      "learning_rate": 2.5354129958451513e-05,
+      "loss": 0.2661,
+      "mean_token_accuracy": 0.8799407482147217,
+      "step": 1454
+    },
+    {
+      "epoch": 4.5754716981132075,
+      "grad_norm": 0.1383388191461563,
+      "learning_rate": 2.5336113231000862e-05,
+      "loss": 0.2517,
+      "mean_token_accuracy": 0.8848426938056946,
+      "step": 1455
+    },
+    {
+      "epoch": 4.578616352201258,
+      "grad_norm": 0.14111852645874023,
+      "learning_rate": 2.5318093041672257e-05,
+      "loss": 0.2618,
+      "mean_token_accuracy": 0.8823397159576416,
+      "step": 1456
+    },
+    {
+      "epoch": 4.581761006289308,
+      "grad_norm": 0.14448735117912292,
+      "learning_rate": 2.5300069409165232e-05,
+      "loss": 0.2371,
+      "mean_token_accuracy": 0.8877670764923096,
+      "step": 1457
+    },
+    {
+      "epoch": 4.584905660377358,
+      "grad_norm": 0.13009530305862427,
+      "learning_rate": 2.5282042352182867e-05,
+      "loss": 0.2415,
+      "mean_token_accuracy": 0.8879152536392212,
+      "step": 1458
+    },
+    {
+      "epoch": 4.588050314465409,
+      "grad_norm": 0.13649174571037292,
+      "learning_rate": 2.5264011889431795e-05,
+      "loss": 0.2632,
+      "mean_token_accuracy": 0.8810548186302185,
+      "step": 1459
+    },
+    {
+      "epoch": 4.591194968553459,
+      "grad_norm": 0.12930653989315033,
+      "learning_rate": 2.5245978039622197e-05,
+      "loss": 0.2498,
+      "mean_token_accuracy": 0.8841792345046997,
+      "step": 1460
+    },
+    {
+      "epoch": 4.59433962264151,
+      "grad_norm": 0.14395128190517426,
+      "learning_rate": 2.522794082146776e-05,
+      "loss": 0.2493,
+      "mean_token_accuracy": 0.8833113312721252,
+      "step": 1461
+    },
+    {
+      "epoch": 4.59748427672956,
+      "grad_norm": 0.13741520047187805,
+      "learning_rate": 2.5209900253685674e-05,
+      "loss": 0.2596,
+      "mean_token_accuracy": 0.8786746859550476,
+      "step": 1462
+    },
+    {
+      "epoch": 4.60062893081761,
+      "grad_norm": 0.14270655810832977,
+      "learning_rate": 2.5191856354996595e-05,
+      "loss": 0.2429,
+      "mean_token_accuracy": 0.8827742338180542,
+      "step": 1463
+    },
+    {
+      "epoch": 4.60377358490566,
+      "grad_norm": 0.13880446553230286,
+      "learning_rate": 2.5173809144124635e-05,
+      "loss": 0.2521,
+      "mean_token_accuracy": 0.8830817341804504,
+      "step": 1464
+    },
+    {
+      "epoch": 4.6069182389937104,
+      "grad_norm": 0.14457201957702637,
+      "learning_rate": 2.5155758639797357e-05,
+      "loss": 0.2489,
+      "mean_token_accuracy": 0.8850651383399963,
+      "step": 1465
+    },
+    {
+      "epoch": 4.610062893081761,
+      "grad_norm": 0.13805538415908813,
+      "learning_rate": 2.513770486074574e-05,
+      "loss": 0.2488,
+      "mean_token_accuracy": 0.883724570274353,
+      "step": 1466
+    },
+    {
+      "epoch": 4.613207547169811,
+      "grad_norm": 0.1541612446308136,
+      "learning_rate": 2.5119647825704134e-05,
+      "loss": 0.2526,
+      "mean_token_accuracy": 0.8837642073631287,
+      "step": 1467
+    },
+    {
+      "epoch": 4.616352201257862,
+      "grad_norm": 0.1382627934217453,
+      "learning_rate": 2.51015875534103e-05,
+      "loss": 0.2639,
+      "mean_token_accuracy": 0.8827504515647888,
+      "step": 1468
+    },
+    {
+      "epoch": 4.619496855345912,
+      "grad_norm": 0.15054289996623993,
+      "learning_rate": 2.5083524062605344e-05,
+      "loss": 0.2509,
+      "mean_token_accuracy": 0.8836148977279663,
+      "step": 1469
+    },
+    {
+      "epoch": 4.622641509433962,
+      "grad_norm": 0.14112670719623566,
+      "learning_rate": 2.506545737203371e-05,
+      "loss": 0.2523,
+      "mean_token_accuracy": 0.8837305903434753,
+      "step": 1470
+    },
+    {
+      "epoch": 4.6257861635220126,
+      "grad_norm": 0.14425958693027496,
+      "learning_rate": 2.5047387500443177e-05,
+      "loss": 0.247,
+      "mean_token_accuracy": 0.8845484852790833,
+      "step": 1471
+    },
+    {
+      "epoch": 4.628930817610063,
+      "grad_norm": 0.14548322558403015,
+      "learning_rate": 2.5029314466584798e-05,
+      "loss": 0.2467,
+      "mean_token_accuracy": 0.883867084980011,
+      "step": 1472
+    },
+    {
+      "epoch": 4.632075471698113,
+      "grad_norm": 0.14163750410079956,
+      "learning_rate": 2.5011238289212948e-05,
+      "loss": 0.2776,
+      "mean_token_accuracy": 0.8781344294548035,
+      "step": 1473
+    },
+    {
+      "epoch": 4.635220125786163,
+      "grad_norm": 0.14586465060710907,
+      "learning_rate": 2.499315898708521e-05,
+      "loss": 0.2493,
+      "mean_token_accuracy": 0.8815778493881226,
+      "step": 1474
+    },
+    {
+      "epoch": 4.638364779874214,
+      "grad_norm": 0.14209404587745667,
+      "learning_rate": 2.4975076578962454e-05,
+      "loss": 0.2525,
+      "mean_token_accuracy": 0.8831738233566284,
+      "step": 1475
+    },
+    {
+      "epoch": 4.6415094339622645,
+      "grad_norm": 0.14334115386009216,
+      "learning_rate": 2.4956991083608766e-05,
+      "loss": 0.2669,
+      "mean_token_accuracy": 0.882244348526001,
+      "step": 1476
+    },
+    {
+      "epoch": 4.644654088050315,
+      "grad_norm": 0.1406649500131607,
+      "learning_rate": 2.493890251979141e-05,
+      "loss": 0.2515,
+      "mean_token_accuracy": 0.8821365833282471,
+      "step": 1477
+    },
+    {
+      "epoch": 4.647798742138365,
+      "grad_norm": 0.1404079645872116,
+      "learning_rate": 2.4920810906280873e-05,
+      "loss": 0.2533,
+      "mean_token_accuracy": 0.8844720125198364,
+      "step": 1478
+    },
+    {
+      "epoch": 4.650943396226415,
+      "grad_norm": 0.13785113394260406,
+      "learning_rate": 2.4902716261850764e-05,
+      "loss": 0.2638,
+      "mean_token_accuracy": 0.8819215893745422,
+      "step": 1479
+    },
+    {
+      "epoch": 4.654088050314465,
+      "grad_norm": 0.15292143821716309,
+      "learning_rate": 2.4884618605277874e-05,
+      "loss": 0.2659,
+      "mean_token_accuracy": 0.8811072111129761,
+      "step": 1480
+    },
+    {
+      "epoch": 4.6572327044025155,
+      "grad_norm": 0.13847500085830688,
+      "learning_rate": 2.4866517955342094e-05,
+      "loss": 0.2564,
+      "mean_token_accuracy": 0.8817123770713806,
+      "step": 1481
+    },
+    {
+      "epoch": 4.660377358490566,
+      "grad_norm": 0.1566178798675537,
+      "learning_rate": 2.4848414330826443e-05,
+      "loss": 0.2505,
+      "mean_token_accuracy": 0.8840153813362122,
+      "step": 1482
+    },
+    {
+      "epoch": 4.663522012578616,
+      "grad_norm": 0.1307821422815323,
+      "learning_rate": 2.4830307750517017e-05,
+      "loss": 0.268,
+      "mean_token_accuracy": 0.8796277046203613,
+      "step": 1483
+    },
+    {
+      "epoch": 4.666666666666667,
+      "grad_norm": 0.152281254529953,
+      "learning_rate": 2.481219823320296e-05,
+      "loss": 0.254,
+      "mean_token_accuracy": 0.8816496133804321,
+      "step": 1484
+    },
+    {
+      "epoch": 4.669811320754717,
+      "grad_norm": 0.14067430794239044,
+      "learning_rate": 2.479408579767649e-05,
+      "loss": 0.2512,
+      "mean_token_accuracy": 0.8842344284057617,
+      "step": 1485
+    },
+    {
+      "epoch": 4.672955974842767,
+      "grad_norm": 0.13918046653270721,
+      "learning_rate": 2.4775970462732858e-05,
+      "loss": 0.2603,
+      "mean_token_accuracy": 0.8810861110687256,
+      "step": 1486
+    },
+    {
+      "epoch": 4.676100628930818,
+      "grad_norm": 0.1361314058303833,
+      "learning_rate": 2.4757852247170293e-05,
+      "loss": 0.256,
+      "mean_token_accuracy": 0.8834766149520874,
+      "step": 1487
+    },
+    {
+      "epoch": 4.679245283018868,
+      "grad_norm": 0.1420655995607376,
+      "learning_rate": 2.4739731169790036e-05,
+      "loss": 0.2482,
+      "mean_token_accuracy": 0.8854100108146667,
+      "step": 1488
+    },
+    {
+      "epoch": 4.682389937106918,
+      "grad_norm": 0.14817163348197937,
+      "learning_rate": 2.4721607249396295e-05,
+      "loss": 0.2581,
+      "mean_token_accuracy": 0.8823645710945129,
+      "step": 1489
+    },
+    {
+      "epoch": 4.685534591194968,
+      "grad_norm": 0.14173953235149384,
+      "learning_rate": 2.4703480504796227e-05,
+      "loss": 0.2423,
+      "mean_token_accuracy": 0.8879494071006775,
+      "step": 1490
+    },
+    {
+      "epoch": 4.688679245283019,
+      "grad_norm": 0.142574280500412,
+      "learning_rate": 2.4685350954799908e-05,
+      "loss": 0.2595,
+      "mean_token_accuracy": 0.8818832635879517,
+      "step": 1491
+    },
+    {
+      "epoch": 4.6918238993710695,
+      "grad_norm": 0.14576321840286255,
+      "learning_rate": 2.466721861822034e-05,
+      "loss": 0.256,
+      "mean_token_accuracy": 0.8829867243766785,
+      "step": 1492
+    },
+    {
+      "epoch": 4.69496855345912,
+      "grad_norm": 0.13140003383159637,
+      "learning_rate": 2.4649083513873423e-05,
+      "loss": 0.251,
+      "mean_token_accuracy": 0.8816654086112976,
+      "step": 1493
+    },
+    {
+      "epoch": 4.69811320754717,
+      "grad_norm": 0.1674601286649704,
+      "learning_rate": 2.4630945660577907e-05,
+      "loss": 0.2488,
+      "mean_token_accuracy": 0.885489284992218,
+      "step": 1494
+    },
+    {
+      "epoch": 4.70125786163522,
+      "grad_norm": 0.14407427608966827,
+      "learning_rate": 2.4612805077155404e-05,
+      "loss": 0.2642,
+      "mean_token_accuracy": 0.8817337155342102,
+      "step": 1495
+    },
+    {
+      "epoch": 4.70440251572327,
+      "grad_norm": 0.14941011369228363,
+      "learning_rate": 2.4594661782430368e-05,
+      "loss": 0.2667,
+      "mean_token_accuracy": 0.8806317448616028,
+      "step": 1496
+    },
+    {
+      "epoch": 4.7075471698113205,
+      "grad_norm": 0.1350492388010025,
+      "learning_rate": 2.4576515795230057e-05,
+      "loss": 0.265,
+      "mean_token_accuracy": 0.876676619052887,
+      "step": 1497
+    },
+    {
+      "epoch": 4.710691823899371,
+      "grad_norm": 0.14951270818710327,
+      "learning_rate": 2.4558367134384516e-05,
+      "loss": 0.2466,
+      "mean_token_accuracy": 0.8872973918914795,
+      "step": 1498
+    },
+    {
+      "epoch": 4.713836477987421,
+      "grad_norm": 0.14224649965763092,
+      "learning_rate": 2.4540215818726587e-05,
+      "loss": 0.2582,
+      "mean_token_accuracy": 0.8833028078079224,
+      "step": 1499
+    },
+    {
+      "epoch": 4.716981132075472,
+      "grad_norm": 0.15279057621955872,
+      "learning_rate": 2.452206186709185e-05,
+      "loss": 0.2644,
+      "mean_token_accuracy": 0.8810228705406189,
+      "step": 1500
+    },
+    {
+      "epoch": 4.720125786163522,
+      "grad_norm": 0.1417638510465622,
+      "learning_rate": 2.4503905298318612e-05,
+      "loss": 0.2682,
+      "mean_token_accuracy": 0.8790037631988525,
+      "step": 1501
+    },
+    {
+      "epoch": 4.723270440251572,
+      "grad_norm": 0.13526369631290436,
+      "learning_rate": 2.448574613124793e-05,
+      "loss": 0.2681,
+      "mean_token_accuracy": 0.8802682161331177,
+      "step": 1502
+    },
+    {
+      "epoch": 4.726415094339623,
+      "grad_norm": 0.1354237049818039,
+      "learning_rate": 2.4467584384723512e-05,
+      "loss": 0.2644,
+      "mean_token_accuracy": 0.8789896965026855,
+      "step": 1503
+    },
+    {
+      "epoch": 4.729559748427673,
+      "grad_norm": 0.13121561706066132,
+      "learning_rate": 2.444942007759178e-05,
+      "loss": 0.2537,
+      "mean_token_accuracy": 0.8831503987312317,
+      "step": 1504
+    },
+    {
+      "epoch": 4.732704402515723,
+      "grad_norm": 0.13547173142433167,
+      "learning_rate": 2.4431253228701795e-05,
+      "loss": 0.2565,
+      "mean_token_accuracy": 0.87940514087677,
+      "step": 1505
+    },
+    {
+      "epoch": 4.735849056603773,
+      "grad_norm": 0.13463617861270905,
+      "learning_rate": 2.4413083856905257e-05,
+      "loss": 0.2424,
+      "mean_token_accuracy": 0.886320173740387,
+      "step": 1506
+    },
+    {
+      "epoch": 4.738993710691824,
+      "grad_norm": 0.12674209475517273,
+      "learning_rate": 2.4394911981056493e-05,
+      "loss": 0.2573,
+      "mean_token_accuracy": 0.880353569984436,
+      "step": 1507
+    },
+    {
+      "epoch": 4.7421383647798745,
+      "grad_norm": 0.14087824523448944,
+      "learning_rate": 2.437673762001241e-05,
+      "loss": 0.2591,
+      "mean_token_accuracy": 0.8803715705871582,
+      "step": 1508
+    },
+    {
+      "epoch": 4.745283018867925,
+      "grad_norm": 0.13621707260608673,
+      "learning_rate": 2.4358560792632515e-05,
+      "loss": 0.2791,
+      "mean_token_accuracy": 0.8760728240013123,
+      "step": 1509
+    },
+    {
+      "epoch": 4.748427672955975,
+      "grad_norm": 0.12566280364990234,
+      "learning_rate": 2.4340381517778867e-05,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.882553219795227,
+      "step": 1510
+    },
+    {
+      "epoch": 4.751572327044025,
+      "grad_norm": 0.1349441558122635,
+      "learning_rate": 2.432219981431605e-05,
+      "loss": 0.2484,
+      "mean_token_accuracy": 0.8867226839065552,
+      "step": 1511
+    },
+    {
+      "epoch": 4.754716981132075,
+      "grad_norm": 0.12752796709537506,
+      "learning_rate": 2.4304015701111197e-05,
+      "loss": 0.2615,
+      "mean_token_accuracy": 0.8807107210159302,
+      "step": 1512
+    },
+    {
+      "epoch": 4.7578616352201255,
+      "grad_norm": 0.12869000434875488,
+      "learning_rate": 2.428582919703391e-05,
+      "loss": 0.2587,
+      "mean_token_accuracy": 0.8817474842071533,
+      "step": 1513
+    },
+    {
+      "epoch": 4.761006289308176,
+      "grad_norm": 0.12598003447055817,
+      "learning_rate": 2.4267640320956302e-05,
+      "loss": 0.2672,
+      "mean_token_accuracy": 0.8799456357955933,
+      "step": 1514
+    },
+    {
+      "epoch": 4.764150943396227,
+      "grad_norm": 0.125724658370018,
+      "learning_rate": 2.4249449091752922e-05,
+      "loss": 0.2374,
+      "mean_token_accuracy": 0.8862662315368652,
+      "step": 1515
+    },
+    {
+      "epoch": 4.767295597484277,
+      "grad_norm": 0.121979720890522,
+      "learning_rate": 2.4231255528300778e-05,
+      "loss": 0.2584,
+      "mean_token_accuracy": 0.8822363615036011,
+      "step": 1516
+    },
+    {
+      "epoch": 4.770440251572327,
+      "grad_norm": 0.12360676378011703,
+      "learning_rate": 2.421305964947929e-05,
+      "loss": 0.2559,
+      "mean_token_accuracy": 0.8823013305664062,
+      "step": 1517
+    },
+    {
+      "epoch": 4.773584905660377,
+      "grad_norm": 0.13420429825782776,
+      "learning_rate": 2.419486147417028e-05,
+      "loss": 0.2326,
+      "mean_token_accuracy": 0.8917833566665649,
+      "step": 1518
+    },
+    {
+      "epoch": 4.776729559748428,
+      "grad_norm": 0.12840516865253448,
+      "learning_rate": 2.417666102125797e-05,
+      "loss": 0.2455,
+      "mean_token_accuracy": 0.8831151127815247,
+      "step": 1519
+    },
+    {
+      "epoch": 4.779874213836478,
+      "grad_norm": 0.14521858096122742,
+      "learning_rate": 2.415845830962892e-05,
+      "loss": 0.2555,
+      "mean_token_accuracy": 0.8834573030471802,
+      "step": 1520
+    },
+    {
+      "epoch": 4.783018867924528,
+      "grad_norm": 0.13063116371631622,
+      "learning_rate": 2.4140253358172064e-05,
+      "loss": 0.2442,
+      "mean_token_accuracy": 0.8842610120773315,
+      "step": 1521
+    },
+    {
+      "epoch": 4.786163522012579,
+      "grad_norm": 0.13243721425533295,
+      "learning_rate": 2.4122046185778628e-05,
+      "loss": 0.2705,
+      "mean_token_accuracy": 0.8768439888954163,
+      "step": 1522
+    },
+    {
+      "epoch": 4.789308176100629,
+      "grad_norm": 0.14029352366924286,
+      "learning_rate": 2.4103836811342167e-05,
+      "loss": 0.2562,
+      "mean_token_accuracy": 0.8825986385345459,
+      "step": 1523
+    },
+    {
+      "epoch": 4.7924528301886795,
+      "grad_norm": 0.1397697478532791,
+      "learning_rate": 2.4085625253758505e-05,
+      "loss": 0.2495,
+      "mean_token_accuracy": 0.8802828788757324,
+      "step": 1524
+    },
+    {
+      "epoch": 4.79559748427673,
+      "grad_norm": 0.13132429122924805,
+      "learning_rate": 2.4067411531925736e-05,
+      "loss": 0.2662,
+      "mean_token_accuracy": 0.8824609518051147,
+      "step": 1525
+    },
+    {
+      "epoch": 4.79874213836478,
+      "grad_norm": 0.12849049270153046,
+      "learning_rate": 2.404919566474422e-05,
+      "loss": 0.24,
+      "mean_token_accuracy": 0.887162446975708,
+      "step": 1526
+    },
+    {
+      "epoch": 4.80188679245283,
+      "grad_norm": 0.1436852514743805,
+      "learning_rate": 2.40309776711165e-05,
+      "loss": 0.255,
+      "mean_token_accuracy": 0.8830965757369995,
+      "step": 1527
+    },
+    {
+      "epoch": 4.80503144654088,
+      "grad_norm": 0.12779057025909424,
+      "learning_rate": 2.401275756994737e-05,
+      "loss": 0.2673,
+      "mean_token_accuracy": 0.8801604509353638,
+      "step": 1528
+    },
+    {
+      "epoch": 4.8081761006289305,
+      "grad_norm": 0.13711079955101013,
+      "learning_rate": 2.399453538014379e-05,
+      "loss": 0.2391,
+      "mean_token_accuracy": 0.8846153020858765,
+      "step": 1529
+    },
+    {
+      "epoch": 4.811320754716981,
+      "grad_norm": 0.12312142550945282,
+      "learning_rate": 2.397631112061488e-05,
+      "loss": 0.2518,
+      "mean_token_accuracy": 0.8821859955787659,
+      "step": 1530
+    },
+    {
+      "epoch": 4.814465408805032,
+      "grad_norm": 0.13619542121887207,
+      "learning_rate": 2.3958084810271927e-05,
+      "loss": 0.2531,
+      "mean_token_accuracy": 0.8836720585823059,
+      "step": 1531
+    },
+    {
+      "epoch": 4.817610062893082,
+      "grad_norm": 0.12221843004226685,
+      "learning_rate": 2.393985646802833e-05,
+      "loss": 0.2481,
+      "mean_token_accuracy": 0.8841218948364258,
+      "step": 1532
+    },
+    {
+      "epoch": 4.820754716981132,
+      "grad_norm": 0.12619447708129883,
+      "learning_rate": 2.392162611279961e-05,
+      "loss": 0.2502,
+      "mean_token_accuracy": 0.8820463418960571,
+      "step": 1533
+    },
+    {
+      "epoch": 4.823899371069182,
+      "grad_norm": 0.12062551081180573,
+      "learning_rate": 2.3903393763503355e-05,
+      "loss": 0.2734,
+      "mean_token_accuracy": 0.8766216039657593,
+      "step": 1534
+    },
+    {
+      "epoch": 4.827044025157233,
+      "grad_norm": 0.12718960642814636,
+      "learning_rate": 2.388515943905925e-05,
+      "loss": 0.2495,
+      "mean_token_accuracy": 0.8845596313476562,
+      "step": 1535
+    },
+    {
+      "epoch": 4.830188679245283,
+      "grad_norm": 0.12057467550039291,
+      "learning_rate": 2.3866923158389018e-05,
+      "loss": 0.2325,
+      "mean_token_accuracy": 0.8902129530906677,
+      "step": 1536
+    },
+    {
+      "epoch": 4.833333333333333,
+      "grad_norm": 0.13578729331493378,
+      "learning_rate": 2.3848684940416402e-05,
+      "loss": 0.2785,
+      "mean_token_accuracy": 0.8787068128585815,
+      "step": 1537
+    },
+    {
+      "epoch": 4.836477987421384,
+      "grad_norm": 0.12010408192873001,
+      "learning_rate": 2.383044480406717e-05,
+      "loss": 0.2825,
+      "mean_token_accuracy": 0.8767572641372681,
+      "step": 1538
+    },
+    {
+      "epoch": 4.839622641509434,
+      "grad_norm": 0.12753932178020477,
+      "learning_rate": 2.381220276826907e-05,
+      "loss": 0.2458,
+      "mean_token_accuracy": 0.8807628750801086,
+      "step": 1539
+    },
+    {
+      "epoch": 4.8427672955974845,
+      "grad_norm": 0.12194069474935532,
+      "learning_rate": 2.3793958851951828e-05,
+      "loss": 0.2588,
+      "mean_token_accuracy": 0.8817402124404907,
+      "step": 1540
+    },
+    {
+      "epoch": 4.845911949685535,
+      "grad_norm": 0.1362626850605011,
+      "learning_rate": 2.377571307404712e-05,
+      "loss": 0.2514,
+      "mean_token_accuracy": 0.8836599588394165,
+      "step": 1541
+    },
+    {
+      "epoch": 4.849056603773585,
+      "grad_norm": 0.129854217171669,
+      "learning_rate": 2.3757465453488557e-05,
+      "loss": 0.262,
+      "mean_token_accuracy": 0.8794586062431335,
+      "step": 1542
+    },
+    {
+      "epoch": 4.852201257861635,
+      "grad_norm": 0.12912380695343018,
+      "learning_rate": 2.3739216009211644e-05,
+      "loss": 0.2602,
+      "mean_token_accuracy": 0.8814219832420349,
+      "step": 1543
+    },
+    {
+      "epoch": 4.855345911949685,
+      "grad_norm": 0.12774620950222015,
+      "learning_rate": 2.37209647601538e-05,
+      "loss": 0.2629,
+      "mean_token_accuracy": 0.881087064743042,
+      "step": 1544
+    },
+    {
+      "epoch": 4.8584905660377355,
+      "grad_norm": 0.1410323977470398,
+      "learning_rate": 2.370271172525431e-05,
+      "loss": 0.2469,
+      "mean_token_accuracy": 0.8836389780044556,
+      "step": 1545
+    },
+    {
+      "epoch": 4.861635220125786,
+      "grad_norm": 0.12685927748680115,
+      "learning_rate": 2.3684456923454316e-05,
+      "loss": 0.2493,
+      "mean_token_accuracy": 0.8849629759788513,
+      "step": 1546
+    },
+    {
+      "epoch": 4.864779874213837,
+      "grad_norm": 0.1375240981578827,
+      "learning_rate": 2.3666200373696785e-05,
+      "loss": 0.2671,
+      "mean_token_accuracy": 0.881800651550293,
+      "step": 1547
+    },
+    {
+      "epoch": 4.867924528301887,
+      "grad_norm": 0.128312885761261,
+      "learning_rate": 2.36479420949265e-05,
+      "loss": 0.2664,
+      "mean_token_accuracy": 0.8808472752571106,
+      "step": 1548
+    },
+    {
+      "epoch": 4.871069182389937,
+      "grad_norm": 0.13644251227378845,
+      "learning_rate": 2.3629682106090036e-05,
+      "loss": 0.2468,
+      "mean_token_accuracy": 0.883743405342102,
+      "step": 1549
+    },
+    {
+      "epoch": 4.8742138364779874,
+      "grad_norm": 0.143430694937706,
+      "learning_rate": 2.361142042613576e-05,
+      "loss": 0.2463,
+      "mean_token_accuracy": 0.8866894245147705,
+      "step": 1550
+    },
+    {
+      "epoch": 4.877358490566038,
+      "grad_norm": 0.12638427317142487,
+      "learning_rate": 2.3593157074013768e-05,
+      "loss": 0.2547,
+      "mean_token_accuracy": 0.881319522857666,
+      "step": 1551
+    },
+    {
+      "epoch": 4.880503144654088,
+      "grad_norm": 0.1527438908815384,
+      "learning_rate": 2.35748920686759e-05,
+      "loss": 0.2457,
+      "mean_token_accuracy": 0.8852053880691528,
+      "step": 1552
+    },
+    {
+      "epoch": 4.883647798742138,
+      "grad_norm": 0.12867127358913422,
+      "learning_rate": 2.355662542907573e-05,
+      "loss": 0.2653,
+      "mean_token_accuracy": 0.8813451528549194,
+      "step": 1553
+    },
+    {
+      "epoch": 4.886792452830189,
+      "grad_norm": 0.14862971007823944,
+      "learning_rate": 2.3538357174168497e-05,
+      "loss": 0.2605,
+      "mean_token_accuracy": 0.8801792860031128,
+      "step": 1554
+    },
+    {
+      "epoch": 4.889937106918239,
+      "grad_norm": 0.13223330676555634,
+      "learning_rate": 2.352008732291115e-05,
+      "loss": 0.2311,
+      "mean_token_accuracy": 0.8890978097915649,
+      "step": 1555
+    },
+    {
+      "epoch": 4.8930817610062896,
+      "grad_norm": 0.13772611320018768,
+      "learning_rate": 2.3501815894262265e-05,
+      "loss": 0.2641,
+      "mean_token_accuracy": 0.8789441585540771,
+      "step": 1556
+    },
+    {
+      "epoch": 4.89622641509434,
+      "grad_norm": 0.1223578229546547,
+      "learning_rate": 2.3483542907182066e-05,
+      "loss": 0.2585,
+      "mean_token_accuracy": 0.8826157450675964,
+      "step": 1557
+    },
+    {
+      "epoch": 4.89937106918239,
+      "grad_norm": 0.13819842040538788,
+      "learning_rate": 2.3465268380632394e-05,
+      "loss": 0.2657,
+      "mean_token_accuracy": 0.880345344543457,
+      "step": 1558
+    },
+    {
+      "epoch": 4.90251572327044,
+      "grad_norm": 0.12548887729644775,
+      "learning_rate": 2.344699233357669e-05,
+      "loss": 0.2589,
+      "mean_token_accuracy": 0.8830808997154236,
+      "step": 1559
+    },
+    {
+      "epoch": 4.90566037735849,
+      "grad_norm": 0.13984771072864532,
+      "learning_rate": 2.342871478497998e-05,
+      "loss": 0.2667,
+      "mean_token_accuracy": 0.8818701505661011,
+      "step": 1560
+    },
+    {
+      "epoch": 4.908805031446541,
+      "grad_norm": 0.12641951441764832,
+      "learning_rate": 2.341043575380883e-05,
+      "loss": 0.2563,
+      "mean_token_accuracy": 0.883787214756012,
+      "step": 1561
+    },
+    {
+      "epoch": 4.911949685534591,
+      "grad_norm": 0.14930589497089386,
+      "learning_rate": 2.339215525903135e-05,
+      "loss": 0.2669,
+      "mean_token_accuracy": 0.8794329762458801,
+      "step": 1562
+    },
+    {
+      "epoch": 4.915094339622642,
+      "grad_norm": 0.1343175768852234,
+      "learning_rate": 2.337387331961718e-05,
+      "loss": 0.2506,
+      "mean_token_accuracy": 0.8854690790176392,
+      "step": 1563
+    },
+    {
+      "epoch": 4.918238993710692,
+      "grad_norm": 0.13801869750022888,
+      "learning_rate": 2.3355589954537448e-05,
+      "loss": 0.2556,
+      "mean_token_accuracy": 0.8809491395950317,
+      "step": 1564
+    },
+    {
+      "epoch": 4.921383647798742,
+      "grad_norm": 0.13780134916305542,
+      "learning_rate": 2.3337305182764768e-05,
+      "loss": 0.2551,
+      "mean_token_accuracy": 0.8815582394599915,
+      "step": 1565
+    },
+    {
+      "epoch": 4.9245283018867925,
+      "grad_norm": 0.12479320913553238,
+      "learning_rate": 2.3319019023273203e-05,
+      "loss": 0.2442,
+      "mean_token_accuracy": 0.8860586881637573,
+      "step": 1566
+    },
+    {
+      "epoch": 4.927672955974843,
+      "grad_norm": 0.14212225377559662,
+      "learning_rate": 2.3300731495038272e-05,
+      "loss": 0.2428,
+      "mean_token_accuracy": 0.8860717415809631,
+      "step": 1567
+    },
+    {
+      "epoch": 4.930817610062893,
+      "grad_norm": 0.12582793831825256,
+      "learning_rate": 2.3282442617036906e-05,
+      "loss": 0.2483,
+      "mean_token_accuracy": 0.8846839070320129,
+      "step": 1568
+    },
+    {
+      "epoch": 4.933962264150943,
+      "grad_norm": 0.1322171986103058,
+      "learning_rate": 2.3264152408247433e-05,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.8855623602867126,
+      "step": 1569
+    },
+    {
+      "epoch": 4.937106918238994,
+      "grad_norm": 0.1252465844154358,
+      "learning_rate": 2.3245860887649564e-05,
+      "loss": 0.2588,
+      "mean_token_accuracy": 0.8804665207862854,
+      "step": 1570
+    },
+    {
+      "epoch": 4.940251572327044,
+      "grad_norm": 0.12859833240509033,
+      "learning_rate": 2.3227568074224376e-05,
+      "loss": 0.2635,
+      "mean_token_accuracy": 0.8824833035469055,
+      "step": 1571
+    },
+    {
+      "epoch": 4.943396226415095,
+      "grad_norm": 0.12550607323646545,
+      "learning_rate": 2.3209273986954287e-05,
+      "loss": 0.2491,
+      "mean_token_accuracy": 0.886185884475708,
+      "step": 1572
+    },
+    {
+      "epoch": 4.946540880503145,
+      "grad_norm": 0.1253989338874817,
+      "learning_rate": 2.319097864482302e-05,
+      "loss": 0.2512,
+      "mean_token_accuracy": 0.8835926651954651,
+      "step": 1573
+    },
+    {
+      "epoch": 4.949685534591195,
+      "grad_norm": 0.12692295014858246,
+      "learning_rate": 2.3172682066815636e-05,
+      "loss": 0.2585,
+      "mean_token_accuracy": 0.880564272403717,
+      "step": 1574
+    },
+    {
+      "epoch": 4.952830188679245,
+      "grad_norm": 0.1225670799612999,
+      "learning_rate": 2.3154384271918443e-05,
+      "loss": 0.2451,
+      "mean_token_accuracy": 0.8834196329116821,
+      "step": 1575
+    },
+    {
+      "epoch": 4.955974842767295,
+      "grad_norm": 0.11881422251462936,
+      "learning_rate": 2.3136085279119033e-05,
+      "loss": 0.2651,
+      "mean_token_accuracy": 0.879045307636261,
+      "step": 1576
+    },
+    {
+      "epoch": 4.959119496855346,
+      "grad_norm": 0.12771472334861755,
+      "learning_rate": 2.3117785107406224e-05,
+      "loss": 0.2527,
+      "mean_token_accuracy": 0.883535623550415,
+      "step": 1577
+    },
+    {
+      "epoch": 4.962264150943396,
+      "grad_norm": 0.12338591367006302,
+      "learning_rate": 2.3099483775770078e-05,
+      "loss": 0.2542,
+      "mean_token_accuracy": 0.8821091055870056,
+      "step": 1578
+    },
+    {
+      "epoch": 4.965408805031447,
+      "grad_norm": 0.12770599126815796,
+      "learning_rate": 2.308118130320184e-05,
+      "loss": 0.2622,
+      "mean_token_accuracy": 0.8797197937965393,
+      "step": 1579
+    },
+    {
+      "epoch": 4.968553459119497,
+      "grad_norm": 0.12757545709609985,
+      "learning_rate": 2.3062877708693957e-05,
+      "loss": 0.2512,
+      "mean_token_accuracy": 0.8837404847145081,
+      "step": 1580
+    },
+    {
+      "epoch": 4.971698113207547,
+      "grad_norm": 0.13121959567070007,
+      "learning_rate": 2.3044573011240025e-05,
+      "loss": 0.2521,
+      "mean_token_accuracy": 0.8823564052581787,
+      "step": 1581
+    },
+    {
+      "epoch": 4.9748427672955975,
+      "grad_norm": 0.12287531793117523,
+      "learning_rate": 2.30262672298348e-05,
+      "loss": 0.2531,
+      "mean_token_accuracy": 0.8842577338218689,
+      "step": 1582
+    },
+    {
+      "epoch": 4.977987421383648,
+      "grad_norm": 0.13713645935058594,
+      "learning_rate": 2.3007960383474148e-05,
+      "loss": 0.2495,
+      "mean_token_accuracy": 0.8849415183067322,
+      "step": 1583
+    },
+    {
+      "epoch": 4.981132075471698,
+      "grad_norm": 0.1428612619638443,
+      "learning_rate": 2.2989652491155043e-05,
+      "loss": 0.2563,
+      "mean_token_accuracy": 0.8826706409454346,
+      "step": 1584
+    },
+    {
+      "epoch": 4.984276729559748,
+      "grad_norm": 0.12580153346061707,
+      "learning_rate": 2.297134357187556e-05,
+      "loss": 0.2639,
+      "mean_token_accuracy": 0.878861665725708,
+      "step": 1585
+    },
+    {
+      "epoch": 4.987421383647799,
+      "grad_norm": 0.13792036473751068,
+      "learning_rate": 2.2953033644634813e-05,
+      "loss": 0.2754,
+      "mean_token_accuracy": 0.8782982230186462,
+      "step": 1586
+    },
+    {
+      "epoch": 4.990566037735849,
+      "grad_norm": 0.13837353885173798,
+      "learning_rate": 2.293472272843299e-05,
+      "loss": 0.2526,
+      "mean_token_accuracy": 0.8833487033843994,
+      "step": 1587
+    },
+    {
+      "epoch": 4.9937106918239,
+      "grad_norm": 0.14238163828849792,
+      "learning_rate": 2.2916410842271274e-05,
+      "loss": 0.2435,
+      "mean_token_accuracy": 0.8903558254241943,
+      "step": 1588
+    },
+    {
+      "epoch": 4.99685534591195,
+      "grad_norm": 0.14869339764118195,
+      "learning_rate": 2.2898098005151893e-05,
+      "loss": 0.2322,
+      "mean_token_accuracy": 0.8898493647575378,
+      "step": 1589
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.14632923901081085,
+      "learning_rate": 2.2879784236078023e-05,
+      "loss": 0.2493,
+      "mean_token_accuracy": 0.8759793639183044,
+      "step": 1590
+    },
+    {
+      "epoch": 5.00314465408805,
+      "grad_norm": 0.22233015298843384,
+      "learning_rate": 2.286146955405384e-05,
+      "loss": 0.1916,
+      "mean_token_accuracy": 0.9095132946968079,
+      "step": 1591
+    },
+    {
+      "epoch": 5.0062893081761,
+      "grad_norm": 0.25025674700737,
+      "learning_rate": 2.2843153978084445e-05,
+      "loss": 0.1819,
+      "mean_token_accuracy": 0.9113137125968933,
+      "step": 1592
+    },
+    {
+      "epoch": 5.009433962264151,
+      "grad_norm": 0.2684067189693451,
+      "learning_rate": 2.282483752717587e-05,
+      "loss": 0.1972,
+      "mean_token_accuracy": 0.9079714417457581,
+      "step": 1593
+    },
+    {
+      "epoch": 5.012578616352202,
+      "grad_norm": 0.24029596149921417,
+      "learning_rate": 2.2806520220335066e-05,
+      "loss": 0.1678,
+      "mean_token_accuracy": 0.9172501564025879,
+      "step": 1594
+    },
+    {
+      "epoch": 5.015723270440252,
+      "grad_norm": 0.2321806699037552,
+      "learning_rate": 2.2788202076569866e-05,
+      "loss": 0.1724,
+      "mean_token_accuracy": 0.9155396223068237,
+      "step": 1595
+    },
+    {
+      "epoch": 5.018867924528302,
+      "grad_norm": 0.18021827936172485,
+      "learning_rate": 2.2769883114888964e-05,
+      "loss": 0.1703,
+      "mean_token_accuracy": 0.9156048893928528,
+      "step": 1596
+    },
+    {
+      "epoch": 5.022012578616352,
+      "grad_norm": 0.23715201020240784,
+      "learning_rate": 2.2751563354301915e-05,
+      "loss": 0.1776,
+      "mean_token_accuracy": 0.9144327044487,
+      "step": 1597
+    },
+    {
+      "epoch": 5.0251572327044025,
+      "grad_norm": 0.16609273850917816,
+      "learning_rate": 2.273324281381909e-05,
+      "loss": 0.1682,
+      "mean_token_accuracy": 0.9172971844673157,
+      "step": 1598
+    },
+    {
+      "epoch": 5.028301886792453,
+      "grad_norm": 0.1903139054775238,
+      "learning_rate": 2.271492151245169e-05,
+      "loss": 0.1598,
+      "mean_token_accuracy": 0.917824387550354,
+      "step": 1599
+    },
+    {
+      "epoch": 5.031446540880503,
+      "grad_norm": 0.17430105805397034,
+      "learning_rate": 2.269659946921168e-05,
+      "loss": 0.1585,
+      "mean_token_accuracy": 0.9176533818244934,
+      "step": 1600
+    },
+    {
+      "epoch": 5.034591194968553,
+      "grad_norm": 0.18532267212867737,
+      "learning_rate": 2.2678276703111814e-05,
+      "loss": 0.1731,
+      "mean_token_accuracy": 0.9151170253753662,
+      "step": 1601
+    },
+    {
+      "epoch": 5.037735849056604,
+      "grad_norm": 0.15971753001213074,
+      "learning_rate": 2.2659953233165583e-05,
+      "loss": 0.1707,
+      "mean_token_accuracy": 0.915444552898407,
+      "step": 1602
+    },
+    {
+      "epoch": 5.040880503144654,
+      "grad_norm": 0.1557256132364273,
+      "learning_rate": 2.2641629078387224e-05,
+      "loss": 0.1737,
+      "mean_token_accuracy": 0.9138659238815308,
+      "step": 1603
+    },
+    {
+      "epoch": 5.044025157232705,
+      "grad_norm": 0.16258005797863007,
+      "learning_rate": 2.2623304257791667e-05,
+      "loss": 0.1739,
+      "mean_token_accuracy": 0.91518634557724,
+      "step": 1604
+    },
+    {
+      "epoch": 5.047169811320755,
+      "grad_norm": 0.149249866604805,
+      "learning_rate": 2.260497879039455e-05,
+      "loss": 0.188,
+      "mean_token_accuracy": 0.9087100625038147,
+      "step": 1605
+    },
+    {
+      "epoch": 5.050314465408805,
+      "grad_norm": 0.17389704287052155,
+      "learning_rate": 2.2586652695212158e-05,
+      "loss": 0.1846,
+      "mean_token_accuracy": 0.9124392867088318,
+      "step": 1606
+    },
+    {
+      "epoch": 5.053459119496855,
+      "grad_norm": 0.1585531234741211,
+      "learning_rate": 2.2568325991261456e-05,
+      "loss": 0.1816,
+      "mean_token_accuracy": 0.9147888422012329,
+      "step": 1607
+    },
+    {
+      "epoch": 5.056603773584905,
+      "grad_norm": 0.14732636511325836,
+      "learning_rate": 2.254999869756002e-05,
+      "loss": 0.1642,
+      "mean_token_accuracy": 0.9162658452987671,
+      "step": 1608
+    },
+    {
+      "epoch": 5.059748427672956,
+      "grad_norm": 0.14769305288791656,
+      "learning_rate": 2.2531670833126056e-05,
+      "loss": 0.1838,
+      "mean_token_accuracy": 0.913840651512146,
+      "step": 1609
+    },
+    {
+      "epoch": 5.062893081761007,
+      "grad_norm": 0.1692722588777542,
+      "learning_rate": 2.2513342416978347e-05,
+      "loss": 0.1761,
+      "mean_token_accuracy": 0.9122536778450012,
+      "step": 1610
+    },
+    {
+      "epoch": 5.066037735849057,
+      "grad_norm": 0.14742763340473175,
+      "learning_rate": 2.2495013468136248e-05,
+      "loss": 0.1708,
+      "mean_token_accuracy": 0.9144207239151001,
+      "step": 1611
+    },
+    {
+      "epoch": 5.069182389937107,
+      "grad_norm": 0.14860284328460693,
+      "learning_rate": 2.2476684005619675e-05,
+      "loss": 0.1656,
+      "mean_token_accuracy": 0.9158710241317749,
+      "step": 1612
+    },
+    {
+      "epoch": 5.072327044025157,
+      "grad_norm": 0.14176800847053528,
+      "learning_rate": 2.2458354048449074e-05,
+      "loss": 0.1819,
+      "mean_token_accuracy": 0.9101191759109497,
+      "step": 1613
+    },
+    {
+      "epoch": 5.0754716981132075,
+      "grad_norm": 0.14405949413776398,
+      "learning_rate": 2.2440023615645404e-05,
+      "loss": 0.181,
+      "mean_token_accuracy": 0.9110443592071533,
+      "step": 1614
+    },
+    {
+      "epoch": 5.078616352201258,
+      "grad_norm": 0.13683363795280457,
+      "learning_rate": 2.242169272623012e-05,
+      "loss": 0.1714,
+      "mean_token_accuracy": 0.9149702191352844,
+      "step": 1615
+    },
+    {
+      "epoch": 5.081761006289308,
+      "grad_norm": 0.13986143469810486,
+      "learning_rate": 2.240336139922515e-05,
+      "loss": 0.1745,
+      "mean_token_accuracy": 0.9130762219429016,
+      "step": 1616
+    },
+    {
+      "epoch": 5.084905660377358,
+      "grad_norm": 0.1361878365278244,
+      "learning_rate": 2.2385029653652868e-05,
+      "loss": 0.1701,
+      "mean_token_accuracy": 0.9145857691764832,
+      "step": 1617
+    },
+    {
+      "epoch": 5.088050314465409,
+      "grad_norm": 0.1427004188299179,
+      "learning_rate": 2.23666975085361e-05,
+      "loss": 0.1744,
+      "mean_token_accuracy": 0.9139806032180786,
+      "step": 1618
+    },
+    {
+      "epoch": 5.091194968553459,
+      "grad_norm": 0.1479874551296234,
+      "learning_rate": 2.2348364982898075e-05,
+      "loss": 0.1791,
+      "mean_token_accuracy": 0.9146347045898438,
+      "step": 1619
+    },
+    {
+      "epoch": 5.09433962264151,
+      "grad_norm": 0.13859923183918,
+      "learning_rate": 2.2330032095762406e-05,
+      "loss": 0.1758,
+      "mean_token_accuracy": 0.9162372350692749,
+      "step": 1620
+    },
+    {
+      "epoch": 5.09748427672956,
+      "grad_norm": 0.14057232439517975,
+      "learning_rate": 2.2311698866153103e-05,
+      "loss": 0.1773,
+      "mean_token_accuracy": 0.9151803255081177,
+      "step": 1621
+    },
+    {
+      "epoch": 5.10062893081761,
+      "grad_norm": 0.1407633125782013,
+      "learning_rate": 2.2293365313094514e-05,
+      "loss": 0.1623,
+      "mean_token_accuracy": 0.9161638021469116,
+      "step": 1622
+    },
+    {
+      "epoch": 5.10377358490566,
+      "grad_norm": 0.14112257957458496,
+      "learning_rate": 2.2275031455611332e-05,
+      "loss": 0.1625,
+      "mean_token_accuracy": 0.9182310104370117,
+      "step": 1623
+    },
+    {
+      "epoch": 5.1069182389937104,
+      "grad_norm": 0.13362768292427063,
+      "learning_rate": 2.225669731272857e-05,
+      "loss": 0.1804,
+      "mean_token_accuracy": 0.9142118096351624,
+      "step": 1624
+    },
+    {
+      "epoch": 5.110062893081761,
+      "grad_norm": 0.1374814510345459,
+      "learning_rate": 2.2238362903471525e-05,
+      "loss": 0.1758,
+      "mean_token_accuracy": 0.9115723371505737,
+      "step": 1625
+    },
+    {
+      "epoch": 5.113207547169812,
+      "grad_norm": 0.13537247478961945,
+      "learning_rate": 2.222002824686578e-05,
+      "loss": 0.1618,
+      "mean_token_accuracy": 0.9196138978004456,
+      "step": 1626
+    },
+    {
+      "epoch": 5.116352201257862,
+      "grad_norm": 0.13499002158641815,
+      "learning_rate": 2.2201693361937164e-05,
+      "loss": 0.1802,
+      "mean_token_accuracy": 0.9118757843971252,
+      "step": 1627
+    },
+    {
+      "epoch": 5.119496855345912,
+      "grad_norm": 0.1436554193496704,
+      "learning_rate": 2.218335826771176e-05,
+      "loss": 0.1892,
+      "mean_token_accuracy": 0.913590133190155,
+      "step": 1628
+    },
+    {
+      "epoch": 5.122641509433962,
+      "grad_norm": 0.12786825001239777,
+      "learning_rate": 2.216502298321585e-05,
+      "loss": 0.1811,
+      "mean_token_accuracy": 0.9122364521026611,
+      "step": 1629
+    },
+    {
+      "epoch": 5.1257861635220126,
+      "grad_norm": 0.1332731992006302,
+      "learning_rate": 2.2146687527475924e-05,
+      "loss": 0.1716,
+      "mean_token_accuracy": 0.9132739901542664,
+      "step": 1630
+    },
+    {
+      "epoch": 5.128930817610063,
+      "grad_norm": 0.12857694923877716,
+      "learning_rate": 2.2128351919518655e-05,
+      "loss": 0.1814,
+      "mean_token_accuracy": 0.9113665819168091,
+      "step": 1631
+    },
+    {
+      "epoch": 5.132075471698113,
+      "grad_norm": 0.13038919866085052,
+      "learning_rate": 2.211001617837085e-05,
+      "loss": 0.1664,
+      "mean_token_accuracy": 0.916091799736023,
+      "step": 1632
+    },
+    {
+      "epoch": 5.135220125786163,
+      "grad_norm": 0.1276601105928421,
+      "learning_rate": 2.2091680323059487e-05,
+      "loss": 0.1762,
+      "mean_token_accuracy": 0.9136810898780823,
+      "step": 1633
+    },
+    {
+      "epoch": 5.138364779874214,
+      "grad_norm": 0.1319180130958557,
+      "learning_rate": 2.2073344372611628e-05,
+      "loss": 0.1737,
+      "mean_token_accuracy": 0.9124643802642822,
+      "step": 1634
+    },
+    {
+      "epoch": 5.1415094339622645,
+      "grad_norm": 0.13005971908569336,
+      "learning_rate": 2.205500834605447e-05,
+      "loss": 0.163,
+      "mean_token_accuracy": 0.9175032377243042,
+      "step": 1635
+    },
+    {
+      "epoch": 5.144654088050315,
+      "grad_norm": 0.12882187962532043,
+      "learning_rate": 2.2036672262415265e-05,
+      "loss": 0.1598,
+      "mean_token_accuracy": 0.9175060391426086,
+      "step": 1636
+    },
+    {
+      "epoch": 5.147798742138365,
+      "grad_norm": 0.1321483552455902,
+      "learning_rate": 2.201833614072132e-05,
+      "loss": 0.1992,
+      "mean_token_accuracy": 0.9085779190063477,
+      "step": 1637
+    },
+    {
+      "epoch": 5.150943396226415,
+      "grad_norm": 0.12345718592405319,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.1769,
+      "mean_token_accuracy": 0.9143081307411194,
+      "step": 1638
+    },
+    {
+      "epoch": 5.154088050314465,
+      "grad_norm": 0.13488464057445526,
+      "learning_rate": 2.1981663859278684e-05,
+      "loss": 0.1526,
+      "mean_token_accuracy": 0.9188219308853149,
+      "step": 1639
+    },
+    {
+      "epoch": 5.1572327044025155,
+      "grad_norm": 0.129329651594162,
+      "learning_rate": 2.1963327737584745e-05,
+      "loss": 0.1589,
+      "mean_token_accuracy": 0.9180588126182556,
+      "step": 1640
+    },
+    {
+      "epoch": 5.160377358490566,
+      "grad_norm": 0.13226263225078583,
+      "learning_rate": 2.194499165394554e-05,
+      "loss": 0.1732,
+      "mean_token_accuracy": 0.9152401685714722,
+      "step": 1641
+    },
+    {
+      "epoch": 5.163522012578617,
+      "grad_norm": 0.1298309713602066,
+      "learning_rate": 2.1926655627388378e-05,
+      "loss": 0.1664,
+      "mean_token_accuracy": 0.9183374047279358,
+      "step": 1642
+    },
+    {
+      "epoch": 5.166666666666667,
+      "grad_norm": 0.13375401496887207,
+      "learning_rate": 2.1908319676940522e-05,
+      "loss": 0.164,
+      "mean_token_accuracy": 0.9196304082870483,
+      "step": 1643
+    },
+    {
+      "epoch": 5.169811320754717,
+      "grad_norm": 0.13043470680713654,
+      "learning_rate": 2.1889983821629154e-05,
+      "loss": 0.1738,
+      "mean_token_accuracy": 0.9157758951187134,
+      "step": 1644
+    },
+    {
+      "epoch": 5.172955974842767,
+      "grad_norm": 0.13828632235527039,
+      "learning_rate": 2.1871648080481354e-05,
+      "loss": 0.1681,
+      "mean_token_accuracy": 0.9176316261291504,
+      "step": 1645
+    },
+    {
+      "epoch": 5.176100628930818,
+      "grad_norm": 0.12848587334156036,
+      "learning_rate": 2.185331247252408e-05,
+      "loss": 0.1735,
+      "mean_token_accuracy": 0.9150087237358093,
+      "step": 1646
+    },
+    {
+      "epoch": 5.179245283018868,
+      "grad_norm": 0.13007590174674988,
+      "learning_rate": 2.1834977016784155e-05,
+      "loss": 0.2002,
+      "mean_token_accuracy": 0.9054135084152222,
+      "step": 1647
+    },
+    {
+      "epoch": 5.182389937106918,
+      "grad_norm": 0.1344444900751114,
+      "learning_rate": 2.181664173228825e-05,
+      "loss": 0.1761,
+      "mean_token_accuracy": 0.9127033948898315,
+      "step": 1648
+    },
+    {
+      "epoch": 5.185534591194968,
+      "grad_norm": 0.1295919120311737,
+      "learning_rate": 2.1798306638062842e-05,
+      "loss": 0.1714,
+      "mean_token_accuracy": 0.9170975685119629,
+      "step": 1649
+    },
+    {
+      "epoch": 5.188679245283019,
+      "grad_norm": 0.1406605988740921,
+      "learning_rate": 2.177997175313423e-05,
+      "loss": 0.1525,
+      "mean_token_accuracy": 0.9223353266716003,
+      "step": 1650
+    },
+    {
+      "epoch": 5.1918238993710695,
+      "grad_norm": 0.127178356051445,
+      "learning_rate": 2.1761637096528477e-05,
+      "loss": 0.1552,
+      "mean_token_accuracy": 0.9182380437850952,
+      "step": 1651
+    },
+    {
+      "epoch": 5.19496855345912,
+      "grad_norm": 0.13574561476707458,
+      "learning_rate": 2.174330268727143e-05,
+      "loss": 0.1718,
+      "mean_token_accuracy": 0.9154348969459534,
+      "step": 1652
+    },
+    {
+      "epoch": 5.19811320754717,
+      "grad_norm": 0.1314130425453186,
+      "learning_rate": 2.172496854438867e-05,
+      "loss": 0.1687,
+      "mean_token_accuracy": 0.9156013131141663,
+      "step": 1653
+    },
+    {
+      "epoch": 5.20125786163522,
+      "grad_norm": 0.13452252745628357,
+      "learning_rate": 2.1706634686905495e-05,
+      "loss": 0.161,
+      "mean_token_accuracy": 0.9191825985908508,
+      "step": 1654
+    },
+    {
+      "epoch": 5.20440251572327,
+      "grad_norm": 0.13440141081809998,
+      "learning_rate": 2.1688301133846906e-05,
+      "loss": 0.1809,
+      "mean_token_accuracy": 0.9128972887992859,
+      "step": 1655
+    },
+    {
+      "epoch": 5.2075471698113205,
+      "grad_norm": 0.12809255719184875,
+      "learning_rate": 2.1669967904237603e-05,
+      "loss": 0.1729,
+      "mean_token_accuracy": 0.9155882000923157,
+      "step": 1656
+    },
+    {
+      "epoch": 5.210691823899371,
+      "grad_norm": 0.1266845464706421,
+      "learning_rate": 2.1651635017101934e-05,
+      "loss": 0.1741,
+      "mean_token_accuracy": 0.9155396223068237,
+      "step": 1657
+    },
+    {
+      "epoch": 5.213836477987422,
+      "grad_norm": 0.1275676041841507,
+      "learning_rate": 2.1633302491463905e-05,
+      "loss": 0.1685,
+      "mean_token_accuracy": 0.916077196598053,
+      "step": 1658
+    },
+    {
+      "epoch": 5.216981132075472,
+      "grad_norm": 0.13584747910499573,
+      "learning_rate": 2.1614970346347137e-05,
+      "loss": 0.1769,
+      "mean_token_accuracy": 0.9124895334243774,
+      "step": 1659
+    },
+    {
+      "epoch": 5.220125786163522,
+      "grad_norm": 0.12740573287010193,
+      "learning_rate": 2.1596638600774855e-05,
+      "loss": 0.1623,
+      "mean_token_accuracy": 0.9163357615470886,
+      "step": 1660
+    },
+    {
+      "epoch": 5.223270440251572,
+      "grad_norm": 0.12712137401103973,
+      "learning_rate": 2.1578307273769887e-05,
+      "loss": 0.1706,
+      "mean_token_accuracy": 0.9162405133247375,
+      "step": 1661
+    },
+    {
+      "epoch": 5.226415094339623,
+      "grad_norm": 0.12990765273571014,
+      "learning_rate": 2.1559976384354605e-05,
+      "loss": 0.1547,
+      "mean_token_accuracy": 0.9218593835830688,
+      "step": 1662
+    },
+    {
+      "epoch": 5.229559748427673,
+      "grad_norm": 0.13632164895534515,
+      "learning_rate": 2.1541645951550935e-05,
+      "loss": 0.1727,
+      "mean_token_accuracy": 0.9177563190460205,
+      "step": 1663
+    },
+    {
+      "epoch": 5.232704402515723,
+      "grad_norm": 0.1234545111656189,
+      "learning_rate": 2.1523315994380338e-05,
+      "loss": 0.1638,
+      "mean_token_accuracy": 0.9165735840797424,
+      "step": 1664
+    },
+    {
+      "epoch": 5.235849056603773,
+      "grad_norm": 0.14210264384746552,
+      "learning_rate": 2.1504986531863765e-05,
+      "loss": 0.1689,
+      "mean_token_accuracy": 0.9141770005226135,
+      "step": 1665
+    },
+    {
+      "epoch": 5.238993710691824,
+      "grad_norm": 0.13037016987800598,
+      "learning_rate": 2.148665758302167e-05,
+      "loss": 0.1708,
+      "mean_token_accuracy": 0.915972113609314,
+      "step": 1666
+    },
+    {
+      "epoch": 5.2421383647798745,
+      "grad_norm": 0.1427024006843567,
+      "learning_rate": 2.1468329166873953e-05,
+      "loss": 0.1919,
+      "mean_token_accuracy": 0.9098140597343445,
+      "step": 1667
+    },
+    {
+      "epoch": 5.245283018867925,
+      "grad_norm": 0.12425141036510468,
+      "learning_rate": 2.1450001302439984e-05,
+      "loss": 0.1646,
+      "mean_token_accuracy": 0.9146043658256531,
+      "step": 1668
+    },
+    {
+      "epoch": 5.248427672955975,
+      "grad_norm": 0.13127084076404572,
+      "learning_rate": 2.1431674008738553e-05,
+      "loss": 0.1644,
+      "mean_token_accuracy": 0.9161396026611328,
+      "step": 1669
+    },
+    {
+      "epoch": 5.251572327044025,
+      "grad_norm": 0.12927378714084625,
+      "learning_rate": 2.1413347304787854e-05,
+      "loss": 0.1653,
+      "mean_token_accuracy": 0.9174731373786926,
+      "step": 1670
+    },
+    {
+      "epoch": 5.254716981132075,
+      "grad_norm": 0.13044291734695435,
+      "learning_rate": 2.1395021209605464e-05,
+      "loss": 0.1757,
+      "mean_token_accuracy": 0.9150776863098145,
+      "step": 1671
+    },
+    {
+      "epoch": 5.2578616352201255,
+      "grad_norm": 0.1301099807024002,
+      "learning_rate": 2.1376695742208345e-05,
+      "loss": 0.1895,
+      "mean_token_accuracy": 0.9120878577232361,
+      "step": 1672
+    },
+    {
+      "epoch": 5.261006289308176,
+      "grad_norm": 0.135678231716156,
+      "learning_rate": 2.135837092161279e-05,
+      "loss": 0.1811,
+      "mean_token_accuracy": 0.9135366082191467,
+      "step": 1673
+    },
+    {
+      "epoch": 5.264150943396227,
+      "grad_norm": 0.12597790360450745,
+      "learning_rate": 2.1340046766834423e-05,
+      "loss": 0.183,
+      "mean_token_accuracy": 0.9111697673797607,
+      "step": 1674
+    },
+    {
+      "epoch": 5.267295597484277,
+      "grad_norm": 0.1251760572195053,
+      "learning_rate": 2.1321723296888198e-05,
+      "loss": 0.1835,
+      "mean_token_accuracy": 0.9082761406898499,
+      "step": 1675
+    },
+    {
+      "epoch": 5.270440251572327,
+      "grad_norm": 0.12519238889217377,
+      "learning_rate": 2.130340053078833e-05,
+      "loss": 0.166,
+      "mean_token_accuracy": 0.9158846735954285,
+      "step": 1676
+    },
+    {
+      "epoch": 5.273584905660377,
+      "grad_norm": 0.1259630024433136,
+      "learning_rate": 2.128507848754832e-05,
+      "loss": 0.1748,
+      "mean_token_accuracy": 0.9130755662918091,
+      "step": 1677
+    },
+    {
+      "epoch": 5.276729559748428,
+      "grad_norm": 0.13136066496372223,
+      "learning_rate": 2.1266757186180915e-05,
+      "loss": 0.1796,
+      "mean_token_accuracy": 0.9120888113975525,
+      "step": 1678
+    },
+    {
+      "epoch": 5.279874213836478,
+      "grad_norm": 0.1244376003742218,
+      "learning_rate": 2.1248436645698094e-05,
+      "loss": 0.1751,
+      "mean_token_accuracy": 0.9147348999977112,
+      "step": 1679
+    },
+    {
+      "epoch": 5.283018867924528,
+      "grad_norm": 0.1334747076034546,
+      "learning_rate": 2.1230116885111048e-05,
+      "loss": 0.1773,
+      "mean_token_accuracy": 0.9149150252342224,
+      "step": 1680
+    },
+    {
+      "epoch": 5.286163522012578,
+      "grad_norm": 0.12724976241588593,
+      "learning_rate": 2.1211797923430146e-05,
+      "loss": 0.1748,
+      "mean_token_accuracy": 0.9149224162101746,
+      "step": 1681
+    },
+    {
+      "epoch": 5.289308176100629,
+      "grad_norm": 0.13602736592292786,
+      "learning_rate": 2.119347977966494e-05,
+      "loss": 0.1758,
+      "mean_token_accuracy": 0.9136221408843994,
+      "step": 1682
+    },
+    {
+      "epoch": 5.2924528301886795,
+      "grad_norm": 0.12472439557313919,
+      "learning_rate": 2.117516247282414e-05,
+      "loss": 0.1674,
+      "mean_token_accuracy": 0.9162774682044983,
+      "step": 1683
+    },
+    {
+      "epoch": 5.29559748427673,
+      "grad_norm": 0.1277409791946411,
+      "learning_rate": 2.1156846021915568e-05,
+      "loss": 0.1737,
+      "mean_token_accuracy": 0.9161177277565002,
+      "step": 1684
+    },
+    {
+      "epoch": 5.29874213836478,
+      "grad_norm": 0.13467463850975037,
+      "learning_rate": 2.1138530445946167e-05,
+      "loss": 0.1769,
+      "mean_token_accuracy": 0.9126297235488892,
+      "step": 1685
+    },
+    {
+      "epoch": 5.30188679245283,
+      "grad_norm": 0.13067930936813354,
+      "learning_rate": 2.1120215763921982e-05,
+      "loss": 0.1792,
+      "mean_token_accuracy": 0.9156489968299866,
+      "step": 1686
+    },
+    {
+      "epoch": 5.30503144654088,
+      "grad_norm": 0.13587665557861328,
+      "learning_rate": 2.1101901994848113e-05,
+      "loss": 0.1763,
+      "mean_token_accuracy": 0.9159653186798096,
+      "step": 1687
+    },
+    {
+      "epoch": 5.3081761006289305,
+      "grad_norm": 0.13601058721542358,
+      "learning_rate": 2.1083589157728728e-05,
+      "loss": 0.1758,
+      "mean_token_accuracy": 0.9139232635498047,
+      "step": 1688
+    },
+    {
+      "epoch": 5.311320754716981,
+      "grad_norm": 0.13745374977588654,
+      "learning_rate": 2.1065277271567017e-05,
+      "loss": 0.1679,
+      "mean_token_accuracy": 0.9164733290672302,
+      "step": 1689
+    },
+    {
+      "epoch": 5.314465408805032,
+      "grad_norm": 0.1407531201839447,
+      "learning_rate": 2.104696635536519e-05,
+      "loss": 0.1876,
+      "mean_token_accuracy": 0.9122368693351746,
+      "step": 1690
+    },
+    {
+      "epoch": 5.317610062893082,
+      "grad_norm": 0.1359671652317047,
+      "learning_rate": 2.1028656428124442e-05,
+      "loss": 0.1859,
+      "mean_token_accuracy": 0.9110690355300903,
+      "step": 1691
+    },
+    {
+      "epoch": 5.320754716981132,
+      "grad_norm": 0.13148832321166992,
+      "learning_rate": 2.101034750884496e-05,
+      "loss": 0.1611,
+      "mean_token_accuracy": 0.9165538549423218,
+      "step": 1692
+    },
+    {
+      "epoch": 5.323899371069182,
+      "grad_norm": 0.13991323113441467,
+      "learning_rate": 2.0992039616525858e-05,
+      "loss": 0.1793,
+      "mean_token_accuracy": 0.9136778712272644,
+      "step": 1693
+    },
+    {
+      "epoch": 5.327044025157233,
+      "grad_norm": 0.14232312142848969,
+      "learning_rate": 2.09737327701652e-05,
+      "loss": 0.1744,
+      "mean_token_accuracy": 0.9157080054283142,
+      "step": 1694
+    },
+    {
+      "epoch": 5.330188679245283,
+      "grad_norm": 0.12538862228393555,
+      "learning_rate": 2.0955426988759978e-05,
+      "loss": 0.1733,
+      "mean_token_accuracy": 0.9145515561103821,
+      "step": 1695
+    },
+    {
+      "epoch": 5.333333333333333,
+      "grad_norm": 0.14959652721881866,
+      "learning_rate": 2.093712229130605e-05,
+      "loss": 0.1776,
+      "mean_token_accuracy": 0.9126619100570679,
+      "step": 1696
+    },
+    {
+      "epoch": 5.336477987421383,
+      "grad_norm": 0.13184374570846558,
+      "learning_rate": 2.0918818696798162e-05,
+      "loss": 0.1756,
+      "mean_token_accuracy": 0.9145973920822144,
+      "step": 1697
+    },
+    {
+      "epoch": 5.339622641509434,
+      "grad_norm": 0.15026512742042542,
+      "learning_rate": 2.0900516224229924e-05,
+      "loss": 0.1735,
+      "mean_token_accuracy": 0.9127100706100464,
+      "step": 1698
+    },
+    {
+      "epoch": 5.3427672955974845,
+      "grad_norm": 0.12787233293056488,
+      "learning_rate": 2.088221489259378e-05,
+      "loss": 0.165,
+      "mean_token_accuracy": 0.9178786873817444,
+      "step": 1699
+    },
+    {
+      "epoch": 5.345911949685535,
+      "grad_norm": 0.14001305401325226,
+      "learning_rate": 2.086391472088097e-05,
+      "loss": 0.1707,
+      "mean_token_accuracy": 0.916415810585022,
+      "step": 1700
+    },
+    {
+      "epoch": 5.349056603773585,
+      "grad_norm": 0.13622647523880005,
+      "learning_rate": 2.084561572808156e-05,
+      "loss": 0.1764,
+      "mean_token_accuracy": 0.9140340685844421,
+      "step": 1701
+    },
+    {
+      "epoch": 5.352201257861635,
+      "grad_norm": 0.13659904897212982,
+      "learning_rate": 2.0827317933184366e-05,
+      "loss": 0.1679,
+      "mean_token_accuracy": 0.9176750779151917,
+      "step": 1702
+    },
+    {
+      "epoch": 5.355345911949685,
+      "grad_norm": 0.1333339959383011,
+      "learning_rate": 2.0809021355176982e-05,
+      "loss": 0.1776,
+      "mean_token_accuracy": 0.9108352065086365,
+      "step": 1703
+    },
+    {
+      "epoch": 5.3584905660377355,
+      "grad_norm": 0.1358334869146347,
+      "learning_rate": 2.0790726013045722e-05,
+      "loss": 0.1773,
+      "mean_token_accuracy": 0.9096439480781555,
+      "step": 1704
+    },
+    {
+      "epoch": 5.361635220125786,
+      "grad_norm": 0.13096484541893005,
+      "learning_rate": 2.077243192577563e-05,
+      "loss": 0.175,
+      "mean_token_accuracy": 0.915550172328949,
+      "step": 1705
+    },
+    {
+      "epoch": 5.364779874213837,
+      "grad_norm": 0.1310902237892151,
+      "learning_rate": 2.0754139112350442e-05,
+      "loss": 0.1847,
+      "mean_token_accuracy": 0.9114246368408203,
+      "step": 1706
+    },
+    {
+      "epoch": 5.367924528301887,
+      "grad_norm": 0.12905707955360413,
+      "learning_rate": 2.0735847591752573e-05,
+      "loss": 0.1707,
+      "mean_token_accuracy": 0.9200414419174194,
+      "step": 1707
+    },
+    {
+      "epoch": 5.371069182389937,
+      "grad_norm": 0.14065292477607727,
+      "learning_rate": 2.07175573829631e-05,
+      "loss": 0.1855,
+      "mean_token_accuracy": 0.9103058576583862,
+      "step": 1708
+    },
+    {
+      "epoch": 5.3742138364779874,
+      "grad_norm": 0.1275883913040161,
+      "learning_rate": 2.069926850496173e-05,
+      "loss": 0.1907,
+      "mean_token_accuracy": 0.9100257158279419,
+      "step": 1709
+    },
+    {
+      "epoch": 5.377358490566038,
+      "grad_norm": 0.12739497423171997,
+      "learning_rate": 2.0680980976726803e-05,
+      "loss": 0.1741,
+      "mean_token_accuracy": 0.9135763049125671,
+      "step": 1710
+    },
+    {
+      "epoch": 5.380503144654088,
+      "grad_norm": 0.13199791312217712,
+      "learning_rate": 2.066269481723524e-05,
+      "loss": 0.1763,
+      "mean_token_accuracy": 0.9162415862083435,
+      "step": 1711
+    },
+    {
+      "epoch": 5.383647798742138,
+      "grad_norm": 0.13175837695598602,
+      "learning_rate": 2.0644410045462558e-05,
+      "loss": 0.171,
+      "mean_token_accuracy": 0.9147182703018188,
+      "step": 1712
+    },
+    {
+      "epoch": 5.386792452830189,
+      "grad_norm": 0.12402436137199402,
+      "learning_rate": 2.0626126680382827e-05,
+      "loss": 0.1772,
+      "mean_token_accuracy": 0.9118449091911316,
+      "step": 1713
+    },
+    {
+      "epoch": 5.389937106918239,
+      "grad_norm": 0.13606113195419312,
+      "learning_rate": 2.060784474096866e-05,
+      "loss": 0.1791,
+      "mean_token_accuracy": 0.9129552841186523,
+      "step": 1714
+    },
+    {
+      "epoch": 5.3930817610062896,
+      "grad_norm": 0.12390584498643875,
+      "learning_rate": 2.0589564246191175e-05,
+      "loss": 0.1739,
+      "mean_token_accuracy": 0.911918044090271,
+      "step": 1715
+    },
+    {
+      "epoch": 5.39622641509434,
+      "grad_norm": 0.13724228739738464,
+      "learning_rate": 2.057128521502002e-05,
+      "loss": 0.1911,
+      "mean_token_accuracy": 0.9111776947975159,
+      "step": 1716
+    },
+    {
+      "epoch": 5.39937106918239,
+      "grad_norm": 0.1254095882177353,
+      "learning_rate": 2.055300766642331e-05,
+      "loss": 0.1768,
+      "mean_token_accuracy": 0.9121692180633545,
+      "step": 1717
+    },
+    {
+      "epoch": 5.40251572327044,
+      "grad_norm": 0.12512388825416565,
+      "learning_rate": 2.0534731619367608e-05,
+      "loss": 0.1692,
+      "mean_token_accuracy": 0.9146745800971985,
+      "step": 1718
+    },
+    {
+      "epoch": 5.40566037735849,
+      "grad_norm": 0.131209135055542,
+      "learning_rate": 2.0516457092817946e-05,
+      "loss": 0.1871,
+      "mean_token_accuracy": 0.9084495306015015,
+      "step": 1719
+    },
+    {
+      "epoch": 5.408805031446541,
+      "grad_norm": 0.12044340372085571,
+      "learning_rate": 2.0498184105737744e-05,
+      "loss": 0.185,
+      "mean_token_accuracy": 0.9076829552650452,
+      "step": 1720
+    },
+    {
+      "epoch": 5.411949685534591,
+      "grad_norm": 0.13348786532878876,
+      "learning_rate": 2.047991267708886e-05,
+      "loss": 0.1724,
+      "mean_token_accuracy": 0.9145122170448303,
+      "step": 1721
+    },
+    {
+      "epoch": 5.415094339622642,
+      "grad_norm": 0.132951557636261,
+      "learning_rate": 2.0461642825831505e-05,
+      "loss": 0.1833,
+      "mean_token_accuracy": 0.9118179678916931,
+      "step": 1722
+    },
+    {
+      "epoch": 5.418238993710692,
+      "grad_norm": 0.1334230601787567,
+      "learning_rate": 2.0443374570924276e-05,
+      "loss": 0.1851,
+      "mean_token_accuracy": 0.9120337963104248,
+      "step": 1723
+    },
+    {
+      "epoch": 5.421383647798742,
+      "grad_norm": 0.12848825752735138,
+      "learning_rate": 2.04251079313241e-05,
+      "loss": 0.1833,
+      "mean_token_accuracy": 0.9081471562385559,
+      "step": 1724
+    },
+    {
+      "epoch": 5.4245283018867925,
+      "grad_norm": 0.13386452198028564,
+      "learning_rate": 2.0406842925986244e-05,
+      "loss": 0.1887,
+      "mean_token_accuracy": 0.9072440266609192,
+      "step": 1725
+    },
+    {
+      "epoch": 5.427672955974843,
+      "grad_norm": 0.12583881616592407,
+      "learning_rate": 2.0388579573864248e-05,
+      "loss": 0.1695,
+      "mean_token_accuracy": 0.9142518639564514,
+      "step": 1726
+    },
+    {
+      "epoch": 5.430817610062893,
+      "grad_norm": 0.12993434071540833,
+      "learning_rate": 2.0370317893909967e-05,
+      "loss": 0.181,
+      "mean_token_accuracy": 0.9098360538482666,
+      "step": 1727
+    },
+    {
+      "epoch": 5.433962264150943,
+      "grad_norm": 0.13069041073322296,
+      "learning_rate": 2.035205790507351e-05,
+      "loss": 0.1693,
+      "mean_token_accuracy": 0.9145693778991699,
+      "step": 1728
+    },
+    {
+      "epoch": 5.437106918238994,
+      "grad_norm": 0.12848743796348572,
+      "learning_rate": 2.0333799626303224e-05,
+      "loss": 0.1598,
+      "mean_token_accuracy": 0.9185672998428345,
+      "step": 1729
+    },
+    {
+      "epoch": 5.440251572327044,
+      "grad_norm": 0.12107231467962265,
+      "learning_rate": 2.031554307654569e-05,
+      "loss": 0.1802,
+      "mean_token_accuracy": 0.9108401536941528,
+      "step": 1730
+    },
+    {
+      "epoch": 5.443396226415095,
+      "grad_norm": 0.1265760362148285,
+      "learning_rate": 2.0297288274745694e-05,
+      "loss": 0.1724,
+      "mean_token_accuracy": 0.9128392338752747,
+      "step": 1731
+    },
+    {
+      "epoch": 5.446540880503145,
+      "grad_norm": 0.1304648518562317,
+      "learning_rate": 2.0279035239846204e-05,
+      "loss": 0.1863,
+      "mean_token_accuracy": 0.9115615487098694,
+      "step": 1732
+    },
+    {
+      "epoch": 5.449685534591195,
+      "grad_norm": 0.12654311954975128,
+      "learning_rate": 2.0260783990788365e-05,
+      "loss": 0.1839,
+      "mean_token_accuracy": 0.9107187986373901,
+      "step": 1733
+    },
+    {
+      "epoch": 5.452830188679245,
+      "grad_norm": 0.12844638526439667,
+      "learning_rate": 2.0242534546511456e-05,
+      "loss": 0.1736,
+      "mean_token_accuracy": 0.9131293296813965,
+      "step": 1734
+    },
+    {
+      "epoch": 5.455974842767295,
+      "grad_norm": 0.12657323479652405,
+      "learning_rate": 2.0224286925952885e-05,
+      "loss": 0.1663,
+      "mean_token_accuracy": 0.9194426536560059,
+      "step": 1735
+    },
+    {
+      "epoch": 5.459119496855346,
+      "grad_norm": 0.12937375903129578,
+      "learning_rate": 2.0206041148048177e-05,
+      "loss": 0.1948,
+      "mean_token_accuracy": 0.9075583219528198,
+      "step": 1736
+    },
+    {
+      "epoch": 5.462264150943396,
+      "grad_norm": 0.12683670222759247,
+      "learning_rate": 2.0187797231730933e-05,
+      "loss": 0.1623,
+      "mean_token_accuracy": 0.9198793172836304,
+      "step": 1737
+    },
+    {
+      "epoch": 5.465408805031447,
+      "grad_norm": 0.12588712573051453,
+      "learning_rate": 2.016955519593284e-05,
+      "loss": 0.1909,
+      "mean_token_accuracy": 0.9105217456817627,
+      "step": 1738
+    },
+    {
+      "epoch": 5.468553459119497,
+      "grad_norm": 0.12617500126361847,
+      "learning_rate": 2.0151315059583603e-05,
+      "loss": 0.1721,
+      "mean_token_accuracy": 0.915179431438446,
+      "step": 1739
+    },
+    {
+      "epoch": 5.471698113207547,
+      "grad_norm": 0.12191683799028397,
+      "learning_rate": 2.0133076841610987e-05,
+      "loss": 0.1552,
+      "mean_token_accuracy": 0.9211398959159851,
+      "step": 1740
+    },
+    {
+      "epoch": 5.4748427672955975,
+      "grad_norm": 0.12133314460515976,
+      "learning_rate": 2.011484056094075e-05,
+      "loss": 0.1857,
+      "mean_token_accuracy": 0.9101731777191162,
+      "step": 1741
+    },
+    {
+      "epoch": 5.477987421383648,
+      "grad_norm": 0.12632842361927032,
+      "learning_rate": 2.009660623649665e-05,
+      "loss": 0.1752,
+      "mean_token_accuracy": 0.9160427451133728,
+      "step": 1742
+    },
+    {
+      "epoch": 5.481132075471698,
+      "grad_norm": 0.12612132728099823,
+      "learning_rate": 2.0078373887200402e-05,
+      "loss": 0.1484,
+      "mean_token_accuracy": 0.9197033047676086,
+      "step": 1743
+    },
+    {
+      "epoch": 5.484276729559748,
+      "grad_norm": 0.1312422752380371,
+      "learning_rate": 2.0060143531971676e-05,
+      "loss": 0.1624,
+      "mean_token_accuracy": 0.9184711575508118,
+      "step": 1744
+    },
+    {
+      "epoch": 5.487421383647799,
+      "grad_norm": 0.12046711146831512,
+      "learning_rate": 2.0041915189728082e-05,
+      "loss": 0.1493,
+      "mean_token_accuracy": 0.9204055070877075,
+      "step": 1745
+    },
+    {
+      "epoch": 5.490566037735849,
+      "grad_norm": 0.13251107931137085,
+      "learning_rate": 2.0023688879385123e-05,
+      "loss": 0.1806,
+      "mean_token_accuracy": 0.9120727181434631,
+      "step": 1746
+    },
+    {
+      "epoch": 5.4937106918239,
+      "grad_norm": 0.11958485841751099,
+      "learning_rate": 2.000546461985622e-05,
+      "loss": 0.1747,
+      "mean_token_accuracy": 0.9143047332763672,
+      "step": 1747
+    },
+    {
+      "epoch": 5.49685534591195,
+      "grad_norm": 0.12881681323051453,
+      "learning_rate": 1.998724243005264e-05,
+      "loss": 0.1825,
+      "mean_token_accuracy": 0.9094586968421936,
+      "step": 1748
+    },
+    {
+      "epoch": 5.5,
+      "grad_norm": 0.1296965628862381,
+      "learning_rate": 1.99690223288835e-05,
+      "loss": 0.1849,
+      "mean_token_accuracy": 0.9092406630516052,
+      "step": 1749
+    },
+    {
+      "epoch": 5.50314465408805,
+      "grad_norm": 0.1288004368543625,
+      "learning_rate": 1.995080433525579e-05,
+      "loss": 0.1914,
+      "mean_token_accuracy": 0.9089462757110596,
+      "step": 1750
+    },
+    {
+      "epoch": 5.5062893081761,
+      "grad_norm": 0.11885450780391693,
+      "learning_rate": 1.9932588468074266e-05,
+      "loss": 0.1653,
+      "mean_token_accuracy": 0.9152488708496094,
+      "step": 1751
+    },
+    {
+      "epoch": 5.509433962264151,
+      "grad_norm": 0.12399091571569443,
+      "learning_rate": 1.9914374746241504e-05,
+      "loss": 0.1836,
+      "mean_token_accuracy": 0.9103871583938599,
+      "step": 1752
+    },
+    {
+      "epoch": 5.512578616352201,
+      "grad_norm": 0.12655200064182281,
+      "learning_rate": 1.9896163188657846e-05,
+      "loss": 0.1718,
+      "mean_token_accuracy": 0.9169356822967529,
+      "step": 1753
+    },
+    {
+      "epoch": 5.515723270440252,
+      "grad_norm": 0.13177834451198578,
+      "learning_rate": 1.9877953814221378e-05,
+      "loss": 0.182,
+      "mean_token_accuracy": 0.9119299650192261,
+      "step": 1754
+    },
+    {
+      "epoch": 5.518867924528302,
+      "grad_norm": 0.12567467987537384,
+      "learning_rate": 1.9859746641827945e-05,
+      "loss": 0.1893,
+      "mean_token_accuracy": 0.9095203876495361,
+      "step": 1755
+    },
+    {
+      "epoch": 5.522012578616352,
+      "grad_norm": 0.12787534296512604,
+      "learning_rate": 1.984154169037108e-05,
+      "loss": 0.1861,
+      "mean_token_accuracy": 0.9093389511108398,
+      "step": 1756
+    },
+    {
+      "epoch": 5.5251572327044025,
+      "grad_norm": 0.12269425392150879,
+      "learning_rate": 1.9823338978742038e-05,
+      "loss": 0.1841,
+      "mean_token_accuracy": 0.9123227000236511,
+      "step": 1757
+    },
+    {
+      "epoch": 5.528301886792453,
+      "grad_norm": 0.12528327107429504,
+      "learning_rate": 1.9805138525829724e-05,
+      "loss": 0.176,
+      "mean_token_accuracy": 0.9133109450340271,
+      "step": 1758
+    },
+    {
+      "epoch": 5.531446540880503,
+      "grad_norm": 0.1250915825366974,
+      "learning_rate": 1.978694035052072e-05,
+      "loss": 0.1739,
+      "mean_token_accuracy": 0.9134659171104431,
+      "step": 1759
+    },
+    {
+      "epoch": 5.534591194968553,
+      "grad_norm": 0.1279488205909729,
+      "learning_rate": 1.9768744471699234e-05,
+      "loss": 0.1851,
+      "mean_token_accuracy": 0.9121567606925964,
+      "step": 1760
+    },
+    {
+      "epoch": 5.537735849056604,
+      "grad_norm": 0.12545357644557953,
+      "learning_rate": 1.9750550908247087e-05,
+      "loss": 0.1719,
+      "mean_token_accuracy": 0.9140828251838684,
+      "step": 1761
+    },
+    {
+      "epoch": 5.540880503144654,
+      "grad_norm": 0.12257158756256104,
+      "learning_rate": 1.973235967904371e-05,
+      "loss": 0.166,
+      "mean_token_accuracy": 0.9187101125717163,
+      "step": 1762
+    },
+    {
+      "epoch": 5.544025157232705,
+      "grad_norm": 0.12817668914794922,
+      "learning_rate": 1.97141708029661e-05,
+      "loss": 0.165,
+      "mean_token_accuracy": 0.9164320230484009,
+      "step": 1763
+    },
+    {
+      "epoch": 5.547169811320755,
+      "grad_norm": 0.11913950741291046,
+      "learning_rate": 1.9695984298888815e-05,
+      "loss": 0.1715,
+      "mean_token_accuracy": 0.9155896902084351,
+      "step": 1764
+    },
+    {
+      "epoch": 5.550314465408805,
+      "grad_norm": 0.12410813570022583,
+      "learning_rate": 1.9677800185683957e-05,
+      "loss": 0.1698,
+      "mean_token_accuracy": 0.9146424531936646,
+      "step": 1765
+    },
+    {
+      "epoch": 5.553459119496855,
+      "grad_norm": 0.12645314633846283,
+      "learning_rate": 1.9659618482221142e-05,
+      "loss": 0.1757,
+      "mean_token_accuracy": 0.9114490747451782,
+      "step": 1766
+    },
+    {
+      "epoch": 5.556603773584905,
+      "grad_norm": 0.1207004114985466,
+      "learning_rate": 1.964143920736749e-05,
+      "loss": 0.1799,
+      "mean_token_accuracy": 0.9112210273742676,
+      "step": 1767
+    },
+    {
+      "epoch": 5.559748427672956,
+      "grad_norm": 0.12781473994255066,
+      "learning_rate": 1.96232623799876e-05,
+      "loss": 0.1891,
+      "mean_token_accuracy": 0.9095780253410339,
+      "step": 1768
+    },
+    {
+      "epoch": 5.562893081761006,
+      "grad_norm": 0.12544502317905426,
+      "learning_rate": 1.960508801894352e-05,
+      "loss": 0.168,
+      "mean_token_accuracy": 0.9177098274230957,
+      "step": 1769
+    },
+    {
+      "epoch": 5.566037735849057,
+      "grad_norm": 0.13463300466537476,
+      "learning_rate": 1.958691614309475e-05,
+      "loss": 0.1853,
+      "mean_token_accuracy": 0.9120203256607056,
+      "step": 1770
+    },
+    {
+      "epoch": 5.569182389937107,
+      "grad_norm": 0.12383488565683365,
+      "learning_rate": 1.9568746771298214e-05,
+      "loss": 0.1787,
+      "mean_token_accuracy": 0.9111815094947815,
+      "step": 1771
+    },
+    {
+      "epoch": 5.572327044025157,
+      "grad_norm": 0.13381510972976685,
+      "learning_rate": 1.9550579922408232e-05,
+      "loss": 0.1758,
+      "mean_token_accuracy": 0.9136260747909546,
+      "step": 1772
+    },
+    {
+      "epoch": 5.5754716981132075,
+      "grad_norm": 0.11905031651258469,
+      "learning_rate": 1.9532415615276497e-05,
+      "loss": 0.1768,
+      "mean_token_accuracy": 0.9129524827003479,
+      "step": 1773
+    },
+    {
+      "epoch": 5.578616352201258,
+      "grad_norm": 0.12877804040908813,
+      "learning_rate": 1.951425386875208e-05,
+      "loss": 0.1911,
+      "mean_token_accuracy": 0.9062658548355103,
+      "step": 1774
+    },
+    {
+      "epoch": 5.581761006289308,
+      "grad_norm": 0.13063427805900574,
+      "learning_rate": 1.949609470168139e-05,
+      "loss": 0.1784,
+      "mean_token_accuracy": 0.9149458408355713,
+      "step": 1775
+    },
+    {
+      "epoch": 5.584905660377358,
+      "grad_norm": 0.1270115077495575,
+      "learning_rate": 1.947793813290816e-05,
+      "loss": 0.1811,
+      "mean_token_accuracy": 0.911460816860199,
+      "step": 1776
+    },
+    {
+      "epoch": 5.588050314465409,
+      "grad_norm": 0.12923474609851837,
+      "learning_rate": 1.945978418127342e-05,
+      "loss": 0.1786,
+      "mean_token_accuracy": 0.9101095199584961,
+      "step": 1777
+    },
+    {
+      "epoch": 5.591194968553459,
+      "grad_norm": 0.12055451422929764,
+      "learning_rate": 1.944163286561549e-05,
+      "loss": 0.1749,
+      "mean_token_accuracy": 0.9124361872673035,
+      "step": 1778
+    },
+    {
+      "epoch": 5.59433962264151,
+      "grad_norm": 0.13345621526241302,
+      "learning_rate": 1.9423484204769955e-05,
+      "loss": 0.1647,
+      "mean_token_accuracy": 0.9183674454689026,
+      "step": 1779
+    },
+    {
+      "epoch": 5.59748427672956,
+      "grad_norm": 0.1298670917749405,
+      "learning_rate": 1.940533821756964e-05,
+      "loss": 0.172,
+      "mean_token_accuracy": 0.9127525091171265,
+      "step": 1780
+    },
+    {
+      "epoch": 5.60062893081761,
+      "grad_norm": 0.12359672784805298,
+      "learning_rate": 1.93871949228446e-05,
+      "loss": 0.182,
+      "mean_token_accuracy": 0.9089277386665344,
+      "step": 1781
+    },
+    {
+      "epoch": 5.60377358490566,
+      "grad_norm": 0.12858720123767853,
+      "learning_rate": 1.9369054339422102e-05,
+      "loss": 0.1941,
+      "mean_token_accuracy": 0.9098080992698669,
+      "step": 1782
+    },
+    {
+      "epoch": 5.6069182389937104,
+      "grad_norm": 0.12454716861248016,
+      "learning_rate": 1.935091648612658e-05,
+      "loss": 0.1633,
+      "mean_token_accuracy": 0.9196724891662598,
+      "step": 1783
+    },
+    {
+      "epoch": 5.610062893081761,
+      "grad_norm": 0.1321592628955841,
+      "learning_rate": 1.9332781381779657e-05,
+      "loss": 0.1621,
+      "mean_token_accuracy": 0.9194785952568054,
+      "step": 1784
+    },
+    {
+      "epoch": 5.613207547169811,
+      "grad_norm": 0.12804855406284332,
+      "learning_rate": 1.9314649045200098e-05,
+      "loss": 0.187,
+      "mean_token_accuracy": 0.912610650062561,
+      "step": 1785
+    },
+    {
+      "epoch": 5.616352201257862,
+      "grad_norm": 0.12372619658708572,
+      "learning_rate": 1.9296519495203778e-05,
+      "loss": 0.176,
+      "mean_token_accuracy": 0.9116061925888062,
+      "step": 1786
+    },
+    {
+      "epoch": 5.619496855345912,
+      "grad_norm": 0.12314292788505554,
+      "learning_rate": 1.9278392750603704e-05,
+      "loss": 0.184,
+      "mean_token_accuracy": 0.9096059203147888,
+      "step": 1787
+    },
+    {
+      "epoch": 5.622641509433962,
+      "grad_norm": 0.12165670096874237,
+      "learning_rate": 1.9260268830209963e-05,
+      "loss": 0.168,
+      "mean_token_accuracy": 0.9138624668121338,
+      "step": 1788
+    },
+    {
+      "epoch": 5.6257861635220126,
+      "grad_norm": 0.1263875663280487,
+      "learning_rate": 1.924214775282971e-05,
+      "loss": 0.1991,
+      "mean_token_accuracy": 0.906342089176178,
+      "step": 1789
+    },
+    {
+      "epoch": 5.628930817610063,
+      "grad_norm": 0.12722733616828918,
+      "learning_rate": 1.9224029537267147e-05,
+      "loss": 0.178,
+      "mean_token_accuracy": 0.9143258929252625,
+      "step": 1790
+    },
+    {
+      "epoch": 5.632075471698113,
+      "grad_norm": 0.12726683914661407,
+      "learning_rate": 1.9205914202323506e-05,
+      "loss": 0.1733,
+      "mean_token_accuracy": 0.9135202765464783,
+      "step": 1791
+    },
+    {
+      "epoch": 5.635220125786163,
+      "grad_norm": 0.13158226013183594,
+      "learning_rate": 1.9187801766797042e-05,
+      "loss": 0.1799,
+      "mean_token_accuracy": 0.9105492830276489,
+      "step": 1792
+    },
+    {
+      "epoch": 5.638364779874214,
+      "grad_norm": 0.1274365484714508,
+      "learning_rate": 1.9169692249482992e-05,
+      "loss": 0.1552,
+      "mean_token_accuracy": 0.9198361039161682,
+      "step": 1793
+    },
+    {
+      "epoch": 5.6415094339622645,
+      "grad_norm": 0.12876805663108826,
+      "learning_rate": 1.915158566917356e-05,
+      "loss": 0.1762,
+      "mean_token_accuracy": 0.9127059578895569,
+      "step": 1794
+    },
+    {
+      "epoch": 5.644654088050315,
+      "grad_norm": 0.12105433642864227,
+      "learning_rate": 1.9133482044657904e-05,
+      "loss": 0.17,
+      "mean_token_accuracy": 0.9126316905021667,
+      "step": 1795
+    },
+    {
+      "epoch": 5.647798742138365,
+      "grad_norm": 0.12377595901489258,
+      "learning_rate": 1.911538139472213e-05,
+      "loss": 0.1673,
+      "mean_token_accuracy": 0.9142662882804871,
+      "step": 1796
+    },
+    {
+      "epoch": 5.650943396226415,
+      "grad_norm": 0.12388768792152405,
+      "learning_rate": 1.9097283738149238e-05,
+      "loss": 0.1675,
+      "mean_token_accuracy": 0.9142235517501831,
+      "step": 1797
+    },
+    {
+      "epoch": 5.654088050314465,
+      "grad_norm": 0.12548349797725677,
+      "learning_rate": 1.9079189093719136e-05,
+      "loss": 0.1847,
+      "mean_token_accuracy": 0.9101769328117371,
+      "step": 1798
+    },
+    {
+      "epoch": 5.6572327044025155,
+      "grad_norm": 0.12843972444534302,
+      "learning_rate": 1.906109748020859e-05,
+      "loss": 0.1691,
+      "mean_token_accuracy": 0.9158428311347961,
+      "step": 1799
+    },
+    {
+      "epoch": 5.660377358490566,
+      "grad_norm": 0.12968730926513672,
+      "learning_rate": 1.904300891639124e-05,
+      "loss": 0.1724,
+      "mean_token_accuracy": 0.9115962386131287,
+      "step": 1800
+    },
+    {
+      "epoch": 5.663522012578616,
+      "grad_norm": 0.13056190311908722,
+      "learning_rate": 1.9024923421037548e-05,
+      "loss": 0.1788,
+      "mean_token_accuracy": 0.9096761345863342,
+      "step": 1801
+    },
+    {
+      "epoch": 5.666666666666667,
+      "grad_norm": 0.13141171634197235,
+      "learning_rate": 1.9006841012914797e-05,
+      "loss": 0.1787,
+      "mean_token_accuracy": 0.9131348729133606,
+      "step": 1802
+    },
+    {
+      "epoch": 5.669811320754717,
+      "grad_norm": 0.14636099338531494,
+      "learning_rate": 1.8988761710787064e-05,
+      "loss": 0.1755,
+      "mean_token_accuracy": 0.910908043384552,
+      "step": 1803
+    },
+    {
+      "epoch": 5.672955974842767,
+      "grad_norm": 0.13519425690174103,
+      "learning_rate": 1.8970685533415197e-05,
+      "loss": 0.1797,
+      "mean_token_accuracy": 0.9129673838615417,
+      "step": 1804
+    },
+    {
+      "epoch": 5.676100628930818,
+      "grad_norm": 0.1360684335231781,
+      "learning_rate": 1.8952612499556825e-05,
+      "loss": 0.1927,
+      "mean_token_accuracy": 0.9077761769294739,
+      "step": 1805
+    },
+    {
+      "epoch": 5.679245283018868,
+      "grad_norm": 0.13671882450580597,
+      "learning_rate": 1.893454262796629e-05,
+      "loss": 0.1778,
+      "mean_token_accuracy": 0.9140453338623047,
+      "step": 1806
+    },
+    {
+      "epoch": 5.682389937106918,
+      "grad_norm": 0.13327039778232574,
+      "learning_rate": 1.8916475937394662e-05,
+      "loss": 0.1719,
+      "mean_token_accuracy": 0.9152608513832092,
+      "step": 1807
+    },
+    {
+      "epoch": 5.685534591194968,
+      "grad_norm": 0.14007729291915894,
+      "learning_rate": 1.8898412446589707e-05,
+      "loss": 0.1847,
+      "mean_token_accuracy": 0.9102155566215515,
+      "step": 1808
+    },
+    {
+      "epoch": 5.688679245283019,
+      "grad_norm": 0.12568777799606323,
+      "learning_rate": 1.8880352174295872e-05,
+      "loss": 0.1674,
+      "mean_token_accuracy": 0.9137527346611023,
+      "step": 1809
+    },
+    {
+      "epoch": 5.6918238993710695,
+      "grad_norm": 0.12971539795398712,
+      "learning_rate": 1.886229513925427e-05,
+      "loss": 0.1766,
+      "mean_token_accuracy": 0.9112458229064941,
+      "step": 1810
+    },
+    {
+      "epoch": 5.69496855345912,
+      "grad_norm": 0.12138007581233978,
+      "learning_rate": 1.8844241360202645e-05,
+      "loss": 0.1816,
+      "mean_token_accuracy": 0.9118508696556091,
+      "step": 1811
+    },
+    {
+      "epoch": 5.69811320754717,
+      "grad_norm": 0.13702593743801117,
+      "learning_rate": 1.8826190855875367e-05,
+      "loss": 0.1659,
+      "mean_token_accuracy": 0.9177764058113098,
+      "step": 1812
+    },
+    {
+      "epoch": 5.70125786163522,
+      "grad_norm": 0.12681394815444946,
+      "learning_rate": 1.8808143645003417e-05,
+      "loss": 0.1936,
+      "mean_token_accuracy": 0.907345712184906,
+      "step": 1813
+    },
+    {
+      "epoch": 5.70440251572327,
+      "grad_norm": 0.12529529631137848,
+      "learning_rate": 1.8790099746314335e-05,
+      "loss": 0.1643,
+      "mean_token_accuracy": 0.9167233109474182,
+      "step": 1814
+    },
+    {
+      "epoch": 5.7075471698113205,
+      "grad_norm": 0.12051253020763397,
+      "learning_rate": 1.877205917853225e-05,
+      "loss": 0.1693,
+      "mean_token_accuracy": 0.9154717326164246,
+      "step": 1815
+    },
+    {
+      "epoch": 5.710691823899371,
+      "grad_norm": 0.1212969496846199,
+      "learning_rate": 1.8754021960377812e-05,
+      "loss": 0.1735,
+      "mean_token_accuracy": 0.9160333275794983,
+      "step": 1816
+    },
+    {
+      "epoch": 5.713836477987421,
+      "grad_norm": 0.12194069474935532,
+      "learning_rate": 1.8735988110568214e-05,
+      "loss": 0.1646,
+      "mean_token_accuracy": 0.9180842041969299,
+      "step": 1817
+    },
+    {
+      "epoch": 5.716981132075472,
+      "grad_norm": 0.12107089906930923,
+      "learning_rate": 1.8717957647817142e-05,
+      "loss": 0.1635,
+      "mean_token_accuracy": 0.9168097972869873,
+      "step": 1818
+    },
+    {
+      "epoch": 5.720125786163522,
+      "grad_norm": 0.1317150741815567,
+      "learning_rate": 1.8699930590834774e-05,
+      "loss": 0.1687,
+      "mean_token_accuracy": 0.9154680967330933,
+      "step": 1819
+    },
+    {
+      "epoch": 5.723270440251572,
+      "grad_norm": 0.12018724530935287,
+      "learning_rate": 1.8681906958327748e-05,
+      "loss": 0.1571,
+      "mean_token_accuracy": 0.9167401194572449,
+      "step": 1820
+    },
+    {
+      "epoch": 5.726415094339623,
+      "grad_norm": 0.12499828636646271,
+      "learning_rate": 1.8663886768999144e-05,
+      "loss": 0.1754,
+      "mean_token_accuracy": 0.9121738076210022,
+      "step": 1821
+    },
+    {
+      "epoch": 5.729559748427673,
+      "grad_norm": 0.1199810653924942,
+      "learning_rate": 1.864587004154849e-05,
+      "loss": 0.158,
+      "mean_token_accuracy": 0.9177903532981873,
+      "step": 1822
+    },
+    {
+      "epoch": 5.732704402515723,
+      "grad_norm": 0.12288343161344528,
+      "learning_rate": 1.86278567946717e-05,
+      "loss": 0.18,
+      "mean_token_accuracy": 0.911756694316864,
+      "step": 1823
+    },
+    {
+      "epoch": 5.735849056603773,
+      "grad_norm": 0.11919686943292618,
+      "learning_rate": 1.860984704706108e-05,
+      "loss": 0.1709,
+      "mean_token_accuracy": 0.9164305925369263,
+      "step": 1824
+    },
+    {
+      "epoch": 5.738993710691824,
+      "grad_norm": 0.12749914824962616,
+      "learning_rate": 1.8591840817405317e-05,
+      "loss": 0.1828,
+      "mean_token_accuracy": 0.9121153950691223,
+      "step": 1825
+    },
+    {
+      "epoch": 5.7421383647798745,
+      "grad_norm": 0.11674331873655319,
+      "learning_rate": 1.8573838124389433e-05,
+      "loss": 0.1784,
+      "mean_token_accuracy": 0.9080796837806702,
+      "step": 1826
+    },
+    {
+      "epoch": 5.745283018867925,
+      "grad_norm": 0.12508632242679596,
+      "learning_rate": 1.855583898669479e-05,
+      "loss": 0.1792,
+      "mean_token_accuracy": 0.9131340980529785,
+      "step": 1827
+    },
+    {
+      "epoch": 5.748427672955975,
+      "grad_norm": 0.12435121834278107,
+      "learning_rate": 1.8537843422999057e-05,
+      "loss": 0.171,
+      "mean_token_accuracy": 0.9142288565635681,
+      "step": 1828
+    },
+    {
+      "epoch": 5.751572327044025,
+      "grad_norm": 0.12231216579675674,
+      "learning_rate": 1.8519851451976182e-05,
+      "loss": 0.1597,
+      "mean_token_accuracy": 0.9187856316566467,
+      "step": 1829
+    },
+    {
+      "epoch": 5.754716981132075,
+      "grad_norm": 0.1260603815317154,
+      "learning_rate": 1.8501863092296415e-05,
+      "loss": 0.18,
+      "mean_token_accuracy": 0.9122654795646667,
+      "step": 1830
+    },
+    {
+      "epoch": 5.7578616352201255,
+      "grad_norm": 0.12207242846488953,
+      "learning_rate": 1.8483878362626227e-05,
+      "loss": 0.1543,
+      "mean_token_accuracy": 0.9220340251922607,
+      "step": 1831
+    },
+    {
+      "epoch": 5.761006289308176,
+      "grad_norm": 0.12331138551235199,
+      "learning_rate": 1.8465897281628345e-05,
+      "loss": 0.1837,
+      "mean_token_accuracy": 0.9105554223060608,
+      "step": 1832
+    },
+    {
+      "epoch": 5.764150943396227,
+      "grad_norm": 0.1287182718515396,
+      "learning_rate": 1.8447919867961697e-05,
+      "loss": 0.157,
+      "mean_token_accuracy": 0.9220572710037231,
+      "step": 1833
+    },
+    {
+      "epoch": 5.767295597484277,
+      "grad_norm": 0.1238393783569336,
+      "learning_rate": 1.84299461402814e-05,
+      "loss": 0.1662,
+      "mean_token_accuracy": 0.915483295917511,
+      "step": 1834
+    },
+    {
+      "epoch": 5.770440251572327,
+      "grad_norm": 0.12729449570178986,
+      "learning_rate": 1.8411976117238772e-05,
+      "loss": 0.1559,
+      "mean_token_accuracy": 0.9172618985176086,
+      "step": 1835
+    },
+    {
+      "epoch": 5.773584905660377,
+      "grad_norm": 0.1212867870926857,
+      "learning_rate": 1.8394009817481257e-05,
+      "loss": 0.183,
+      "mean_token_accuracy": 0.9098618030548096,
+      "step": 1836
+    },
+    {
+      "epoch": 5.776729559748428,
+      "grad_norm": 0.1259167343378067,
+      "learning_rate": 1.8376047259652448e-05,
+      "loss": 0.1708,
+      "mean_token_accuracy": 0.9136357307434082,
+      "step": 1837
+    },
+    {
+      "epoch": 5.779874213836478,
+      "grad_norm": 0.12603263556957245,
+      "learning_rate": 1.8358088462392057e-05,
+      "loss": 0.1713,
+      "mean_token_accuracy": 0.9129566550254822,
+      "step": 1838
+    },
+    {
+      "epoch": 5.783018867924528,
+      "grad_norm": 0.12520869076251984,
+      "learning_rate": 1.8340133444335894e-05,
+      "loss": 0.1777,
+      "mean_token_accuracy": 0.9123414158821106,
+      "step": 1839
+    },
+    {
+      "epoch": 5.786163522012579,
+      "grad_norm": 0.1280491203069687,
+      "learning_rate": 1.832218222411584e-05,
+      "loss": 0.161,
+      "mean_token_accuracy": 0.9205901026725769,
+      "step": 1840
+    },
+    {
+      "epoch": 5.789308176100629,
+      "grad_norm": 0.12988995015621185,
+      "learning_rate": 1.8304234820359852e-05,
+      "loss": 0.1622,
+      "mean_token_accuracy": 0.9184454679489136,
+      "step": 1841
+    },
+    {
+      "epoch": 5.7924528301886795,
+      "grad_norm": 0.12439626455307007,
+      "learning_rate": 1.8286291251691897e-05,
+      "loss": 0.1705,
+      "mean_token_accuracy": 0.9161950945854187,
+      "step": 1842
+    },
+    {
+      "epoch": 5.79559748427673,
+      "grad_norm": 0.13772748410701752,
+      "learning_rate": 1.826835153673199e-05,
+      "loss": 0.1752,
+      "mean_token_accuracy": 0.9136615991592407,
+      "step": 1843
+    },
+    {
+      "epoch": 5.79874213836478,
+      "grad_norm": 0.1208910346031189,
+      "learning_rate": 1.825041569409614e-05,
+      "loss": 0.1845,
+      "mean_token_accuracy": 0.9119904637336731,
+      "step": 1844
+    },
+    {
+      "epoch": 5.80188679245283,
+      "grad_norm": 0.1381557136774063,
+      "learning_rate": 1.8232483742396327e-05,
+      "loss": 0.1734,
+      "mean_token_accuracy": 0.9145846962928772,
+      "step": 1845
+    },
+    {
+      "epoch": 5.80503144654088,
+      "grad_norm": 0.13252446055412292,
+      "learning_rate": 1.8214555700240498e-05,
+      "loss": 0.1712,
+      "mean_token_accuracy": 0.9147006869316101,
+      "step": 1846
+    },
+    {
+      "epoch": 5.8081761006289305,
+      "grad_norm": 0.1245056763291359,
+      "learning_rate": 1.819663158623256e-05,
+      "loss": 0.1695,
+      "mean_token_accuracy": 0.9146434664726257,
+      "step": 1847
+    },
+    {
+      "epoch": 5.811320754716981,
+      "grad_norm": 0.1266990602016449,
+      "learning_rate": 1.817871141897232e-05,
+      "loss": 0.1776,
+      "mean_token_accuracy": 0.9120575785636902,
+      "step": 1848
+    },
+    {
+      "epoch": 5.814465408805032,
+      "grad_norm": 0.12160525470972061,
+      "learning_rate": 1.8160795217055496e-05,
+      "loss": 0.1824,
+      "mean_token_accuracy": 0.9122020602226257,
+      "step": 1849
+    },
+    {
+      "epoch": 5.817610062893082,
+      "grad_norm": 0.127620130777359,
+      "learning_rate": 1.8142882999073704e-05,
+      "loss": 0.1719,
+      "mean_token_accuracy": 0.915385901927948,
+      "step": 1850
+    },
+    {
+      "epoch": 5.820754716981132,
+      "grad_norm": 0.11782664060592651,
+      "learning_rate": 1.8124974783614414e-05,
+      "loss": 0.1701,
+      "mean_token_accuracy": 0.9131607413291931,
+      "step": 1851
+    },
+    {
+      "epoch": 5.823899371069182,
+      "grad_norm": 0.13343071937561035,
+      "learning_rate": 1.8107070589260943e-05,
+      "loss": 0.1533,
+      "mean_token_accuracy": 0.9223355650901794,
+      "step": 1852
+    },
+    {
+      "epoch": 5.827044025157233,
+      "grad_norm": 0.11941219866275787,
+      "learning_rate": 1.808917043459243e-05,
+      "loss": 0.1745,
+      "mean_token_accuracy": 0.9113269448280334,
+      "step": 1853
+    },
+    {
+      "epoch": 5.830188679245283,
+      "grad_norm": 0.11961886286735535,
+      "learning_rate": 1.807127433818384e-05,
+      "loss": 0.1745,
+      "mean_token_accuracy": 0.9124999642372131,
+      "step": 1854
+    },
+    {
+      "epoch": 5.833333333333333,
+      "grad_norm": 0.12838539481163025,
+      "learning_rate": 1.8053382318605907e-05,
+      "loss": 0.1952,
+      "mean_token_accuracy": 0.9072304964065552,
+      "step": 1855
+    },
+    {
+      "epoch": 5.836477987421384,
+      "grad_norm": 0.11784368008375168,
+      "learning_rate": 1.8035494394425158e-05,
+      "loss": 0.1655,
+      "mean_token_accuracy": 0.9166972041130066,
+      "step": 1856
+    },
+    {
+      "epoch": 5.839622641509434,
+      "grad_norm": 0.12901045382022858,
+      "learning_rate": 1.801761058420384e-05,
+      "loss": 0.1855,
+      "mean_token_accuracy": 0.9092808365821838,
+      "step": 1857
+    },
+    {
+      "epoch": 5.8427672955974845,
+      "grad_norm": 0.11686600744724274,
+      "learning_rate": 1.799973090649995e-05,
+      "loss": 0.1754,
+      "mean_token_accuracy": 0.9145554900169373,
+      "step": 1858
+    },
+    {
+      "epoch": 5.845911949685535,
+      "grad_norm": 0.12752406299114227,
+      "learning_rate": 1.7981855379867204e-05,
+      "loss": 0.1793,
+      "mean_token_accuracy": 0.913017213344574,
+      "step": 1859
+    },
+    {
+      "epoch": 5.849056603773585,
+      "grad_norm": 0.12086496502161026,
+      "learning_rate": 1.796398402285499e-05,
+      "loss": 0.1873,
+      "mean_token_accuracy": 0.9081063866615295,
+      "step": 1860
+    },
+    {
+      "epoch": 5.852201257861635,
+      "grad_norm": 0.1289493292570114,
+      "learning_rate": 1.7946116854008383e-05,
+      "loss": 0.1715,
+      "mean_token_accuracy": 0.9141929149627686,
+      "step": 1861
+    },
+    {
+      "epoch": 5.855345911949685,
+      "grad_norm": 0.12896886467933655,
+      "learning_rate": 1.7928253891868103e-05,
+      "loss": 0.174,
+      "mean_token_accuracy": 0.9146431684494019,
+      "step": 1862
+    },
+    {
+      "epoch": 5.8584905660377355,
+      "grad_norm": 0.1219131350517273,
+      "learning_rate": 1.7910395154970505e-05,
+      "loss": 0.1889,
+      "mean_token_accuracy": 0.9074649214744568,
+      "step": 1863
+    },
+    {
+      "epoch": 5.861635220125786,
+      "grad_norm": 0.13309486210346222,
+      "learning_rate": 1.789254066184758e-05,
+      "loss": 0.1715,
+      "mean_token_accuracy": 0.9138280749320984,
+      "step": 1864
+    },
+    {
+      "epoch": 5.864779874213837,
+      "grad_norm": 0.11922544986009598,
+      "learning_rate": 1.7874690431026887e-05,
+      "loss": 0.1702,
+      "mean_token_accuracy": 0.9120256304740906,
+      "step": 1865
+    },
+    {
+      "epoch": 5.867924528301887,
+      "grad_norm": 0.12653562426567078,
+      "learning_rate": 1.7856844481031585e-05,
+      "loss": 0.1836,
+      "mean_token_accuracy": 0.9097647666931152,
+      "step": 1866
+    },
+    {
+      "epoch": 5.871069182389937,
+      "grad_norm": 0.12282025814056396,
+      "learning_rate": 1.7839002830380366e-05,
+      "loss": 0.1778,
+      "mean_token_accuracy": 0.9133232831954956,
+      "step": 1867
+    },
+    {
+      "epoch": 5.8742138364779874,
+      "grad_norm": 0.12195201963186264,
+      "learning_rate": 1.7821165497587488e-05,
+      "loss": 0.1622,
+      "mean_token_accuracy": 0.9166160821914673,
+      "step": 1868
+    },
+    {
+      "epoch": 5.877358490566038,
+      "grad_norm": 0.12466923892498016,
+      "learning_rate": 1.780333250116271e-05,
+      "loss": 0.1641,
+      "mean_token_accuracy": 0.9177669286727905,
+      "step": 1869
+    },
+    {
+      "epoch": 5.880503144654088,
+      "grad_norm": 0.13061249256134033,
+      "learning_rate": 1.778550385961129e-05,
+      "loss": 0.1789,
+      "mean_token_accuracy": 0.913748562335968,
+      "step": 1870
+    },
+    {
+      "epoch": 5.883647798742138,
+      "grad_norm": 0.12507785856723785,
+      "learning_rate": 1.7767679591433984e-05,
+      "loss": 0.1836,
+      "mean_token_accuracy": 0.9113100171089172,
+      "step": 1871
+    },
+    {
+      "epoch": 5.886792452830189,
+      "grad_norm": 0.13062918186187744,
+      "learning_rate": 1.774985971512699e-05,
+      "loss": 0.1872,
+      "mean_token_accuracy": 0.9101560115814209,
+      "step": 1872
+    },
+    {
+      "epoch": 5.889937106918239,
+      "grad_norm": 0.12669727206230164,
+      "learning_rate": 1.773204424918196e-05,
+      "loss": 0.1863,
+      "mean_token_accuracy": 0.9074638485908508,
+      "step": 1873
+    },
+    {
+      "epoch": 5.8930817610062896,
+      "grad_norm": 0.13299503922462463,
+      "learning_rate": 1.771423321208597e-05,
+      "loss": 0.1741,
+      "mean_token_accuracy": 0.9125397205352783,
+      "step": 1874
+    },
+    {
+      "epoch": 5.89622641509434,
+      "grad_norm": 0.13146135210990906,
+      "learning_rate": 1.7696426622321503e-05,
+      "loss": 0.1784,
+      "mean_token_accuracy": 0.9100609421730042,
+      "step": 1875
+    },
+    {
+      "epoch": 5.89937106918239,
+      "grad_norm": 0.13344912230968475,
+      "learning_rate": 1.7678624498366406e-05,
+      "loss": 0.1886,
+      "mean_token_accuracy": 0.9081234931945801,
+      "step": 1876
+    },
+    {
+      "epoch": 5.90251572327044,
+      "grad_norm": 0.13202603161334991,
+      "learning_rate": 1.7660826858693914e-05,
+      "loss": 0.1778,
+      "mean_token_accuracy": 0.9113733768463135,
+      "step": 1877
+    },
+    {
+      "epoch": 5.90566037735849,
+      "grad_norm": 0.12540169060230255,
+      "learning_rate": 1.7643033721772608e-05,
+      "loss": 0.1765,
+      "mean_token_accuracy": 0.912645697593689,
+      "step": 1878
+    },
+    {
+      "epoch": 5.908805031446541,
+      "grad_norm": 0.13628825545310974,
+      "learning_rate": 1.7625245106066372e-05,
+      "loss": 0.1614,
+      "mean_token_accuracy": 0.917786717414856,
+      "step": 1879
+    },
+    {
+      "epoch": 5.911949685534591,
+      "grad_norm": 0.12427868694067001,
+      "learning_rate": 1.7607461030034437e-05,
+      "loss": 0.1746,
+      "mean_token_accuracy": 0.9142725467681885,
+      "step": 1880
+    },
+    {
+      "epoch": 5.915094339622642,
+      "grad_norm": 0.13920536637306213,
+      "learning_rate": 1.7589681512131295e-05,
+      "loss": 0.1803,
+      "mean_token_accuracy": 0.9117159247398376,
+      "step": 1881
+    },
+    {
+      "epoch": 5.918238993710692,
+      "grad_norm": 0.13377030193805695,
+      "learning_rate": 1.7571906570806718e-05,
+      "loss": 0.184,
+      "mean_token_accuracy": 0.9084151983261108,
+      "step": 1882
+    },
+    {
+      "epoch": 5.921383647798742,
+      "grad_norm": 0.12509870529174805,
+      "learning_rate": 1.755413622450572e-05,
+      "loss": 0.1859,
+      "mean_token_accuracy": 0.907892107963562,
+      "step": 1883
+    },
+    {
+      "epoch": 5.9245283018867925,
+      "grad_norm": 0.13371077179908752,
+      "learning_rate": 1.7536370491668556e-05,
+      "loss": 0.175,
+      "mean_token_accuracy": 0.9129257798194885,
+      "step": 1884
+    },
+    {
+      "epoch": 5.927672955974843,
+      "grad_norm": 0.12641875445842743,
+      "learning_rate": 1.7518609390730694e-05,
+      "loss": 0.1706,
+      "mean_token_accuracy": 0.9163611531257629,
+      "step": 1885
+    },
+    {
+      "epoch": 5.930817610062893,
+      "grad_norm": 0.1306959092617035,
+      "learning_rate": 1.7500852940122796e-05,
+      "loss": 0.1617,
+      "mean_token_accuracy": 0.9185019731521606,
+      "step": 1886
+    },
+    {
+      "epoch": 5.933962264150943,
+      "grad_norm": 0.1335182934999466,
+      "learning_rate": 1.7483101158270683e-05,
+      "loss": 0.185,
+      "mean_token_accuracy": 0.9097692966461182,
+      "step": 1887
+    },
+    {
+      "epoch": 5.937106918238994,
+      "grad_norm": 0.12307420372962952,
+      "learning_rate": 1.7465354063595354e-05,
+      "loss": 0.1744,
+      "mean_token_accuracy": 0.9118009209632874,
+      "step": 1888
+    },
+    {
+      "epoch": 5.940251572327044,
+      "grad_norm": 0.13397091627120972,
+      "learning_rate": 1.744761167451292e-05,
+      "loss": 0.1726,
+      "mean_token_accuracy": 0.9162708520889282,
+      "step": 1889
+    },
+    {
+      "epoch": 5.943396226415095,
+      "grad_norm": 0.1265205293893814,
+      "learning_rate": 1.7429874009434642e-05,
+      "loss": 0.1685,
+      "mean_token_accuracy": 0.9149485230445862,
+      "step": 1890
+    },
+    {
+      "epoch": 5.946540880503145,
+      "grad_norm": 0.12802155315876007,
+      "learning_rate": 1.741214108676684e-05,
+      "loss": 0.1845,
+      "mean_token_accuracy": 0.9122422337532043,
+      "step": 1891
+    },
+    {
+      "epoch": 5.949685534591195,
+      "grad_norm": 0.12957172095775604,
+      "learning_rate": 1.7394412924910946e-05,
+      "loss": 0.1953,
+      "mean_token_accuracy": 0.9067781567573547,
+      "step": 1892
+    },
+    {
+      "epoch": 5.952830188679245,
+      "grad_norm": 0.1264718621969223,
+      "learning_rate": 1.7376689542263424e-05,
+      "loss": 0.1666,
+      "mean_token_accuracy": 0.9175181984901428,
+      "step": 1893
+    },
+    {
+      "epoch": 5.955974842767295,
+      "grad_norm": 0.12541256844997406,
+      "learning_rate": 1.7358970957215805e-05,
+      "loss": 0.1789,
+      "mean_token_accuracy": 0.9094677567481995,
+      "step": 1894
+    },
+    {
+      "epoch": 5.959119496855346,
+      "grad_norm": 0.12028246372938156,
+      "learning_rate": 1.7341257188154625e-05,
+      "loss": 0.1975,
+      "mean_token_accuracy": 0.9086244106292725,
+      "step": 1895
+    },
+    {
+      "epoch": 5.962264150943396,
+      "grad_norm": 0.13436433672904968,
+      "learning_rate": 1.7323548253461425e-05,
+      "loss": 0.1808,
+      "mean_token_accuracy": 0.9111320972442627,
+      "step": 1896
+    },
+    {
+      "epoch": 5.965408805031447,
+      "grad_norm": 0.11996529251337051,
+      "learning_rate": 1.730584417151273e-05,
+      "loss": 0.1758,
+      "mean_token_accuracy": 0.9111454486846924,
+      "step": 1897
+    },
+    {
+      "epoch": 5.968553459119497,
+      "grad_norm": 0.12315444648265839,
+      "learning_rate": 1.728814496068003e-05,
+      "loss": 0.1925,
+      "mean_token_accuracy": 0.9099141955375671,
+      "step": 1898
+    },
+    {
+      "epoch": 5.971698113207547,
+      "grad_norm": 0.12306390702724457,
+      "learning_rate": 1.7270450639329762e-05,
+      "loss": 0.1768,
+      "mean_token_accuracy": 0.9125403761863708,
+      "step": 1899
+    },
+    {
+      "epoch": 5.9748427672955975,
+      "grad_norm": 0.1273583173751831,
+      "learning_rate": 1.725276122582329e-05,
+      "loss": 0.1891,
+      "mean_token_accuracy": 0.9100738167762756,
+      "step": 1900
+    },
+    {
+      "epoch": 5.977987421383648,
+      "grad_norm": 0.12477988004684448,
+      "learning_rate": 1.723507673851687e-05,
+      "loss": 0.1697,
+      "mean_token_accuracy": 0.9162830710411072,
+      "step": 1901
+    },
+    {
+      "epoch": 5.981132075471698,
+      "grad_norm": 0.1281619518995285,
+      "learning_rate": 1.721739719576167e-05,
+      "loss": 0.1838,
+      "mean_token_accuracy": 0.9113216996192932,
+      "step": 1902
+    },
+    {
+      "epoch": 5.984276729559748,
+      "grad_norm": 0.12437441945075989,
+      "learning_rate": 1.7199722615903722e-05,
+      "loss": 0.1768,
+      "mean_token_accuracy": 0.9131054878234863,
+      "step": 1903
+    },
+    {
+      "epoch": 5.987421383647799,
+      "grad_norm": 0.12587052583694458,
+      "learning_rate": 1.7182053017283884e-05,
+      "loss": 0.1914,
+      "mean_token_accuracy": 0.9063124656677246,
+      "step": 1904
+    },
+    {
+      "epoch": 5.990566037735849,
+      "grad_norm": 0.1270790845155716,
+      "learning_rate": 1.7164388418237878e-05,
+      "loss": 0.1814,
+      "mean_token_accuracy": 0.9116953015327454,
+      "step": 1905
+    },
+    {
+      "epoch": 5.9937106918239,
+      "grad_norm": 0.11693906038999557,
+      "learning_rate": 1.714672883709621e-05,
+      "loss": 0.1712,
+      "mean_token_accuracy": 0.9138330817222595,
+      "step": 1906
+    },
+    {
+      "epoch": 5.99685534591195,
+      "grad_norm": 0.12291273474693298,
+      "learning_rate": 1.71290742921842e-05,
+      "loss": 0.1838,
+      "mean_token_accuracy": 0.9125950932502747,
+      "step": 1907
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 0.12398134917020798,
+      "learning_rate": 1.7111424801821933e-05,
+      "loss": 0.1596,
+      "mean_token_accuracy": 0.9150497913360596,
+      "step": 1908
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 3180,
+  "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": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.92346119564768e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}