diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4120 @@
+{
+  "best_global_step": 45850,
+  "best_metric": 0.05883299,
+  "best_model_checkpoint": "/mnt/bn/ocr-generation-lf/zhuhanshen/wuxc/ms-swift/IT-SFT-MERGE-1128/v2-20251128-180713/checkpoint-45850",
+  "epoch": 2.0,
+  "eval_steps": 1000,
+  "global_step": 45850,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 4.362050163576881e-05,
+      "grad_norm": 36.4930419921875,
+      "learning_rate": 2.1805494984736157e-09,
+      "loss": 1.106152057647705,
+      "step": 1,
+      "token_acc": 0.768033946251768
+    },
+    {
+      "epoch": 0.004362050163576881,
+      "grad_norm": 12.102117538452148,
+      "learning_rate": 2.1805494984736155e-07,
+      "loss": 0.8755137125651041,
+      "step": 100,
+      "token_acc": 0.8027415733279217
+    },
+    {
+      "epoch": 0.008724100327153763,
+      "grad_norm": 6.443205833435059,
+      "learning_rate": 4.361098996947231e-07,
+      "loss": 0.4035614776611328,
+      "step": 200,
+      "token_acc": 0.8832118196196428
+    },
+    {
+      "epoch": 0.013086150490730643,
+      "grad_norm": 9.192578315734863,
+      "learning_rate": 6.541648495420847e-07,
+      "loss": 0.319589729309082,
+      "step": 300,
+      "token_acc": 0.9065540095345141
+    },
+    {
+      "epoch": 0.017448200654307525,
+      "grad_norm": 5.373167037963867,
+      "learning_rate": 8.722197993894462e-07,
+      "loss": 0.27497081756591796,
+      "step": 400,
+      "token_acc": 0.9191371328484643
+    },
+    {
+      "epoch": 0.021810250817884406,
+      "grad_norm": 10.11259651184082,
+      "learning_rate": 1.0902747492368077e-06,
+      "loss": 0.25123407363891603,
+      "step": 500,
+      "token_acc": 0.925168467355569
+    },
+    {
+      "epoch": 0.026172300981461286,
+      "grad_norm": 10.016520500183105,
+      "learning_rate": 1.3083296990841693e-06,
+      "loss": 0.23664604187011717,
+      "step": 600,
+      "token_acc": 0.9300838425465398
+    },
+    {
+      "epoch": 0.030534351145038167,
+      "grad_norm": 6.1170501708984375,
+      "learning_rate": 1.526384648931531e-06,
+      "loss": 0.22281688690185547,
+      "step": 700,
+      "token_acc": 0.9335510798908284
+    },
+    {
+      "epoch": 0.03489640130861505,
+      "grad_norm": 5.652917385101318,
+      "learning_rate": 1.7444395987788924e-06,
+      "loss": 0.20703411102294922,
+      "step": 800,
+      "token_acc": 0.938460532766792
+    },
+    {
+      "epoch": 0.03925845147219193,
+      "grad_norm": 4.53310489654541,
+      "learning_rate": 1.9624945486262538e-06,
+      "loss": 0.19653003692626952,
+      "step": 900,
+      "token_acc": 0.9402567018683997
+    },
+    {
+      "epoch": 0.04362050163576881,
+      "grad_norm": 5.1060662269592285,
+      "learning_rate": 2.1805494984736154e-06,
+      "loss": 0.1954585075378418,
+      "step": 1000,
+      "token_acc": 0.9415839373883811
+    },
+    {
+      "epoch": 0.04362050163576881,
+      "eval_loss": 0.16520032286643982,
+      "eval_runtime": 250.1581,
+      "eval_samples_per_second": 59.243,
+      "eval_steps_per_second": 0.927,
+      "eval_token_acc": 0.9401546990968435,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04798255179934569,
+      "grad_norm": 6.413122653961182,
+      "learning_rate": 2.398604448320977e-06,
+      "loss": 0.18896270751953126,
+      "step": 1100,
+      "token_acc": 0.9428930293702289
+    },
+    {
+      "epoch": 0.05234460196292257,
+      "grad_norm": 5.541325569152832,
+      "learning_rate": 2.6166593981683387e-06,
+      "loss": 0.18908491134643554,
+      "step": 1200,
+      "token_acc": 0.9417954190899537
+    },
+    {
+      "epoch": 0.05670665212649945,
+      "grad_norm": 7.229689121246338,
+      "learning_rate": 2.8347143480157003e-06,
+      "loss": 0.1891629409790039,
+      "step": 1300,
+      "token_acc": 0.9430973872957417
+    },
+    {
+      "epoch": 0.061068702290076333,
+      "grad_norm": 8.033756256103516,
+      "learning_rate": 3.052769297863062e-06,
+      "loss": 0.1889020347595215,
+      "step": 1400,
+      "token_acc": 0.9424450611947768
+    },
+    {
+      "epoch": 0.06543075245365322,
+      "grad_norm": 4.568666458129883,
+      "learning_rate": 3.2708242477104235e-06,
+      "loss": 0.18495147705078124,
+      "step": 1500,
+      "token_acc": 0.9448793426926312
+    },
+    {
+      "epoch": 0.0697928026172301,
+      "grad_norm": 5.559028148651123,
+      "learning_rate": 3.4888791975577847e-06,
+      "loss": 0.1811268424987793,
+      "step": 1600,
+      "token_acc": 0.9450138440087491
+    },
+    {
+      "epoch": 0.07415485278080698,
+      "grad_norm": 6.583218574523926,
+      "learning_rate": 3.7069341474051464e-06,
+      "loss": 0.18984716415405273,
+      "step": 1700,
+      "token_acc": 0.9421092101583751
+    },
+    {
+      "epoch": 0.07851690294438386,
+      "grad_norm": 9.186613082885742,
+      "learning_rate": 3.9249890972525076e-06,
+      "loss": 0.17964080810546876,
+      "step": 1800,
+      "token_acc": 0.9452867846742794
+    },
+    {
+      "epoch": 0.08287895310796074,
+      "grad_norm": 5.135467052459717,
+      "learning_rate": 4.14304404709987e-06,
+      "loss": 0.1825981330871582,
+      "step": 1900,
+      "token_acc": 0.944733164609133
+    },
+    {
+      "epoch": 0.08724100327153762,
+      "grad_norm": 10.182524681091309,
+      "learning_rate": 4.361098996947231e-06,
+      "loss": 0.177908935546875,
+      "step": 2000,
+      "token_acc": 0.9461056495645801
+    },
+    {
+      "epoch": 0.08724100327153762,
+      "eval_loss": 0.1509065330028534,
+      "eval_runtime": 255.4608,
+      "eval_samples_per_second": 58.013,
+      "eval_steps_per_second": 0.908,
+      "eval_token_acc": 0.9452293660019673,
+      "step": 2000
+    },
+    {
+      "epoch": 0.0916030534351145,
+      "grad_norm": 5.487904071807861,
+      "learning_rate": 4.579153946794593e-06,
+      "loss": 0.17657140731811524,
+      "step": 2100,
+      "token_acc": 0.9470306750061767
+    },
+    {
+      "epoch": 0.09596510359869138,
+      "grad_norm": 6.613377571105957,
+      "learning_rate": 4.797208896641954e-06,
+      "loss": 0.18289306640625,
+      "step": 2200,
+      "token_acc": 0.9451059860400784
+    },
+    {
+      "epoch": 0.10032715376226826,
+      "grad_norm": 11.444597244262695,
+      "learning_rate": 4.999999681367619e-06,
+      "loss": 0.18539167404174806,
+      "step": 2300,
+      "token_acc": 0.9451543586613773
+    },
+    {
+      "epoch": 0.10468920392584515,
+      "grad_norm": 6.6478681564331055,
+      "learning_rate": 4.999925550936637e-06,
+      "loss": 0.17520687103271484,
+      "step": 2400,
+      "token_acc": 0.9466563107809991
+    },
+    {
+      "epoch": 0.10905125408942203,
+      "grad_norm": 5.2834882736206055,
+      "learning_rate": 4.999721370906455e-06,
+      "loss": 0.18468694686889647,
+      "step": 2500,
+      "token_acc": 0.944860093749592
+    },
+    {
+      "epoch": 0.1134133042529989,
+      "grad_norm": 5.122148036956787,
+      "learning_rate": 4.9993871518988e-06,
+      "loss": 0.17573898315429687,
+      "step": 2600,
+      "token_acc": 0.9476367310518581
+    },
+    {
+      "epoch": 0.11777535441657579,
+      "grad_norm": 5.123068809509277,
+      "learning_rate": 4.99892291130021e-06,
+      "loss": 0.1807489013671875,
+      "step": 2700,
+      "token_acc": 0.9463905801777397
+    },
+    {
+      "epoch": 0.12213740458015267,
+      "grad_norm": 5.782346248626709,
+      "learning_rate": 4.998328673261126e-06,
+      "loss": 0.17619085311889648,
+      "step": 2800,
+      "token_acc": 0.9469496873533967
+    },
+    {
+      "epoch": 0.12649945474372956,
+      "grad_norm": 3.942211151123047,
+      "learning_rate": 4.9976044686946355e-06,
+      "loss": 0.17367206573486327,
+      "step": 2900,
+      "token_acc": 0.9483415648386821
+    },
+    {
+      "epoch": 0.13086150490730644,
+      "grad_norm": 4.226406097412109,
+      "learning_rate": 4.996750335274866e-06,
+      "loss": 0.17105751037597655,
+      "step": 3000,
+      "token_acc": 0.948902939369314
+    },
+    {
+      "epoch": 0.13086150490730644,
+      "eval_loss": 0.1425495445728302,
+      "eval_runtime": 252.4401,
+      "eval_samples_per_second": 58.707,
+      "eval_steps_per_second": 0.919,
+      "eval_token_acc": 0.9480601135652329,
+      "step": 3000
+    },
+    {
+      "epoch": 0.13522355507088332,
+      "grad_norm": 5.636692047119141,
+      "learning_rate": 4.9957663174350245e-06,
+      "loss": 0.17473587036132812,
+      "step": 3100,
+      "token_acc": 0.9479665881897636
+    },
+    {
+      "epoch": 0.1395856052344602,
+      "grad_norm": 6.189432144165039,
+      "learning_rate": 4.9946524663650856e-06,
+      "loss": 0.17271400451660157,
+      "step": 3200,
+      "token_acc": 0.948410186178116
+    },
+    {
+      "epoch": 0.14394765539803708,
+      "grad_norm": 13.567493438720703,
+      "learning_rate": 4.993408840009129e-06,
+      "loss": 0.16095369338989257,
+      "step": 3300,
+      "token_acc": 0.951935940949682
+    },
+    {
+      "epoch": 0.14830970556161396,
+      "grad_norm": 7.645431041717529,
+      "learning_rate": 4.992035503062324e-06,
+      "loss": 0.17597726821899415,
+      "step": 3400,
+      "token_acc": 0.9478462417576773
+    },
+    {
+      "epoch": 0.15267175572519084,
+      "grad_norm": 6.895127296447754,
+      "learning_rate": 4.990532526967568e-06,
+      "loss": 0.16777828216552734,
+      "step": 3500,
+      "token_acc": 0.9506341819509079
+    },
+    {
+      "epoch": 0.15703380588876772,
+      "grad_norm": 6.079124927520752,
+      "learning_rate": 4.988899989911762e-06,
+      "loss": 0.16899423599243163,
+      "step": 3600,
+      "token_acc": 0.9485503365232598
+    },
+    {
+      "epoch": 0.1613958560523446,
+      "grad_norm": 4.00687837600708,
+      "learning_rate": 4.987137976821754e-06,
+      "loss": 0.15569159507751465,
+      "step": 3700,
+      "token_acc": 0.9528974769328038
+    },
+    {
+      "epoch": 0.16575790621592149,
+      "grad_norm": 12.302860260009766,
+      "learning_rate": 4.98524657935991e-06,
+      "loss": 0.17022018432617186,
+      "step": 3800,
+      "token_acc": 0.9489793956484122
+    },
+    {
+      "epoch": 0.17011995637949837,
+      "grad_norm": 3.2564656734466553,
+      "learning_rate": 4.983225895919354e-06,
+      "loss": 0.16914573669433594,
+      "step": 3900,
+      "token_acc": 0.9495082261518536
+    },
+    {
+      "epoch": 0.17448200654307525,
+      "grad_norm": 6.390711784362793,
+      "learning_rate": 4.981076031618844e-06,
+      "loss": 0.16581764221191406,
+      "step": 4000,
+      "token_acc": 0.9505939681455173
+    },
+    {
+      "epoch": 0.17448200654307525,
+      "eval_loss": 0.13437031209468842,
+      "eval_runtime": 249.7749,
+      "eval_samples_per_second": 59.333,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9513994455870518,
+      "step": 4000
+    },
+    {
+      "epoch": 0.17884405670665213,
+      "grad_norm": 5.330230712890625,
+      "learning_rate": 4.9787970982973085e-06,
+      "loss": 0.16594120025634765,
+      "step": 4100,
+      "token_acc": 0.9504188464871457
+    },
+    {
+      "epoch": 0.183206106870229,
+      "grad_norm": 4.424008369445801,
+      "learning_rate": 4.976389214508022e-06,
+      "loss": 0.16885875701904296,
+      "step": 4200,
+      "token_acc": 0.9493527907248485
+    },
+    {
+      "epoch": 0.1875681570338059,
+      "grad_norm": 7.039103031158447,
+      "learning_rate": 4.973852505512445e-06,
+      "loss": 0.16290786743164062,
+      "step": 4300,
+      "token_acc": 0.9510095676938338
+    },
+    {
+      "epoch": 0.19193020719738277,
+      "grad_norm": 4.440979480743408,
+      "learning_rate": 4.971187103273701e-06,
+      "loss": 0.1564283561706543,
+      "step": 4400,
+      "token_acc": 0.9533942774165869
+    },
+    {
+      "epoch": 0.19629225736095965,
+      "grad_norm": 3.2784507274627686,
+      "learning_rate": 4.968393146449718e-06,
+      "loss": 0.16346294403076173,
+      "step": 4500,
+      "token_acc": 0.9512642397065876
+    },
+    {
+      "epoch": 0.20065430752453653,
+      "grad_norm": 4.944409370422363,
+      "learning_rate": 4.9654707803860095e-06,
+      "loss": 0.15641886711120606,
+      "step": 4600,
+      "token_acc": 0.9534258320991503
+    },
+    {
+      "epoch": 0.2050163576881134,
+      "grad_norm": 4.601531505584717,
+      "learning_rate": 4.9624201571081164e-06,
+      "loss": 0.16025123596191407,
+      "step": 4700,
+      "token_acc": 0.9523697102048043
+    },
+    {
+      "epoch": 0.2093784078516903,
+      "grad_norm": 6.826204299926758,
+      "learning_rate": 4.9592414353137e-06,
+      "loss": 0.15847814559936524,
+      "step": 4800,
+      "token_acc": 0.9529489381993481
+    },
+    {
+      "epoch": 0.21374045801526717,
+      "grad_norm": 4.184462070465088,
+      "learning_rate": 4.955934780364281e-06,
+      "loss": 0.16379379272460937,
+      "step": 4900,
+      "token_acc": 0.9510652183746131
+    },
+    {
+      "epoch": 0.21810250817884405,
+      "grad_norm": 6.352680206298828,
+      "learning_rate": 4.952500364276644e-06,
+      "loss": 0.16284820556640625,
+      "step": 5000,
+      "token_acc": 0.9512094566302702
+    },
+    {
+      "epoch": 0.21810250817884405,
+      "eval_loss": 0.12947490811347961,
+      "eval_runtime": 249.5303,
+      "eval_samples_per_second": 59.392,
+      "eval_steps_per_second": 0.93,
+      "eval_token_acc": 0.9529140212823035,
+      "step": 5000
+    },
+    {
+      "epoch": 0.22246455834242093,
+      "grad_norm": 6.26838493347168,
+      "learning_rate": 4.948938365713883e-06,
+      "loss": 0.16305082321166992,
+      "step": 5100,
+      "token_acc": 0.9515196791570008
+    },
+    {
+      "epoch": 0.2268266085059978,
+      "grad_norm": 5.207289218902588,
+      "learning_rate": 4.9452489699761095e-06,
+      "loss": 0.15841794967651368,
+      "step": 5200,
+      "token_acc": 0.9528224991360029
+    },
+    {
+      "epoch": 0.2311886586695747,
+      "grad_norm": 5.445582389831543,
+      "learning_rate": 4.941432368990816e-06,
+      "loss": 0.1492086696624756,
+      "step": 5300,
+      "token_acc": 0.9551668315495998
+    },
+    {
+      "epoch": 0.23555070883315157,
+      "grad_norm": 6.060046672821045,
+      "learning_rate": 4.9374887613028845e-06,
+      "loss": 0.16065792083740235,
+      "step": 5400,
+      "token_acc": 0.9520704754902909
+    },
+    {
+      "epoch": 0.23991275899672845,
+      "grad_norm": 6.347756862640381,
+      "learning_rate": 4.933418352064265e-06,
+      "loss": 0.15583457946777343,
+      "step": 5500,
+      "token_acc": 0.9537081698366554
+    },
+    {
+      "epoch": 0.24427480916030533,
+      "grad_norm": 5.751275062561035,
+      "learning_rate": 4.929221353023299e-06,
+      "loss": 0.17144599914550782,
+      "step": 5600,
+      "token_acc": 0.9501105779801089
+    },
+    {
+      "epoch": 0.24863685932388221,
+      "grad_norm": 6.227544784545898,
+      "learning_rate": 4.924897982513706e-06,
+      "loss": 0.16434305191040038,
+      "step": 5700,
+      "token_acc": 0.9513792820279583
+    },
+    {
+      "epoch": 0.2529989094874591,
+      "grad_norm": 22.748313903808594,
+      "learning_rate": 4.920448465443224e-06,
+      "loss": 0.15164663314819335,
+      "step": 5800,
+      "token_acc": 0.9544478258894679
+    },
+    {
+      "epoch": 0.257360959651036,
+      "grad_norm": 6.24751615524292,
+      "learning_rate": 4.91587303328191e-06,
+      "loss": 0.15351497650146484,
+      "step": 5900,
+      "token_acc": 0.9543150580279145
+    },
+    {
+      "epoch": 0.2617230098146129,
+      "grad_norm": 18.683191299438477,
+      "learning_rate": 4.911171924050102e-06,
+      "loss": 0.15083285331726073,
+      "step": 6000,
+      "token_acc": 0.9558162853385074
+    },
+    {
+      "epoch": 0.2617230098146129,
+      "eval_loss": 0.12582732737064362,
+      "eval_runtime": 249.7653,
+      "eval_samples_per_second": 59.336,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9549986586783511,
+      "step": 6000
+    },
+    {
+      "epoch": 0.26608505997818976,
+      "grad_norm": 4.456823348999023,
+      "learning_rate": 4.906345382306029e-06,
+      "loss": 0.15498119354248047,
+      "step": 6100,
+      "token_acc": 0.9538203027177121
+    },
+    {
+      "epoch": 0.27044711014176664,
+      "grad_norm": 5.5751261711120605,
+      "learning_rate": 4.901393659133097e-06,
+      "loss": 0.160562744140625,
+      "step": 6200,
+      "token_acc": 0.95239975292447
+    },
+    {
+      "epoch": 0.2748091603053435,
+      "grad_norm": 8.7870512008667,
+      "learning_rate": 4.896317012126823e-06,
+      "loss": 0.1592543315887451,
+      "step": 6300,
+      "token_acc": 0.9525488439137464
+    },
+    {
+      "epoch": 0.2791712104689204,
+      "grad_norm": 6.550958156585693,
+      "learning_rate": 4.891115705381435e-06,
+      "loss": 0.15966859817504883,
+      "step": 6400,
+      "token_acc": 0.9526111128964875
+    },
+    {
+      "epoch": 0.2835332606324973,
+      "grad_norm": 4.731610298156738,
+      "learning_rate": 4.885790009476132e-06,
+      "loss": 0.15472820281982422,
+      "step": 6500,
+      "token_acc": 0.9540492592976276
+    },
+    {
+      "epoch": 0.28789531079607417,
+      "grad_norm": 3.9044625759124756,
+      "learning_rate": 4.880340201461015e-06,
+      "loss": 0.15400704383850097,
+      "step": 6600,
+      "token_acc": 0.9542899599431451
+    },
+    {
+      "epoch": 0.29225736095965105,
+      "grad_norm": 4.739759922027588,
+      "learning_rate": 4.874766564842662e-06,
+      "loss": 0.1476905059814453,
+      "step": 6700,
+      "token_acc": 0.9566351143445723
+    },
+    {
+      "epoch": 0.2966194111232279,
+      "grad_norm": 4.103377342224121,
+      "learning_rate": 4.869069389569394e-06,
+      "loss": 0.14938350677490234,
+      "step": 6800,
+      "token_acc": 0.9559722494927678
+    },
+    {
+      "epoch": 0.3009814612868048,
+      "grad_norm": 5.790430545806885,
+      "learning_rate": 4.863248972016179e-06,
+      "loss": 0.15306180953979492,
+      "step": 6900,
+      "token_acc": 0.9547221557428477
+    },
+    {
+      "epoch": 0.3053435114503817,
+      "grad_norm": 4.019115924835205,
+      "learning_rate": 4.857305614969224e-06,
+      "loss": 0.1495196056365967,
+      "step": 7000,
+      "token_acc": 0.9557983880238603
+    },
+    {
+      "epoch": 0.3053435114503817,
+      "eval_loss": 0.12187354266643524,
+      "eval_runtime": 248.7288,
+      "eval_samples_per_second": 59.583,
+      "eval_steps_per_second": 0.933,
+      "eval_token_acc": 0.9559152284717876,
+      "step": 7000
+    },
+    {
+      "epoch": 0.30970556161395857,
+      "grad_norm": 4.442187786102295,
+      "learning_rate": 4.851239627610216e-06,
+      "loss": 0.14913738250732422,
+      "step": 7100,
+      "token_acc": 0.9559660106753005
+    },
+    {
+      "epoch": 0.31406761177753545,
+      "grad_norm": 5.366152286529541,
+      "learning_rate": 4.8450513255002415e-06,
+      "loss": 0.14996286392211913,
+      "step": 7200,
+      "token_acc": 0.9556774135003047
+    },
+    {
+      "epoch": 0.31842966194111233,
+      "grad_norm": 4.185603618621826,
+      "learning_rate": 4.838741030563372e-06,
+      "loss": 0.15599340438842774,
+      "step": 7300,
+      "token_acc": 0.953399045204792
+    },
+    {
+      "epoch": 0.3227917121046892,
+      "grad_norm": 5.234697341918945,
+      "learning_rate": 4.832309071069914e-06,
+      "loss": 0.15486830711364746,
+      "step": 7400,
+      "token_acc": 0.9538847084824248
+    },
+    {
+      "epoch": 0.3271537622682661,
+      "grad_norm": 11.299803733825684,
+      "learning_rate": 4.825755781619333e-06,
+      "loss": 0.15180639266967774,
+      "step": 7500,
+      "token_acc": 0.9557206634731319
+    },
+    {
+      "epoch": 0.33151581243184297,
+      "grad_norm": 5.087841510772705,
+      "learning_rate": 4.819081503122847e-06,
+      "loss": 0.15074671745300294,
+      "step": 7600,
+      "token_acc": 0.9549014645668269
+    },
+    {
+      "epoch": 0.33587786259541985,
+      "grad_norm": 4.772754192352295,
+      "learning_rate": 4.812286582785697e-06,
+      "loss": 0.15997478485107422,
+      "step": 7700,
+      "token_acc": 0.9523766333888324
+    },
+    {
+      "epoch": 0.34023991275899673,
+      "grad_norm": 5.332608222961426,
+      "learning_rate": 4.805371374089071e-06,
+      "loss": 0.15052309036254882,
+      "step": 7800,
+      "token_acc": 0.9560667154238967
+    },
+    {
+      "epoch": 0.3446019629225736,
+      "grad_norm": 6.446188926696777,
+      "learning_rate": 4.798336236771733e-06,
+      "loss": 0.14415891647338866,
+      "step": 7900,
+      "token_acc": 0.957361806745217
+    },
+    {
+      "epoch": 0.3489640130861505,
+      "grad_norm": 5.3451032638549805,
+      "learning_rate": 4.791181536811295e-06,
+      "loss": 0.14976163864135741,
+      "step": 8000,
+      "token_acc": 0.9560646367214662
+    },
+    {
+      "epoch": 0.3489640130861505,
+      "eval_loss": 0.11995680630207062,
+      "eval_runtime": 247.9086,
+      "eval_samples_per_second": 59.78,
+      "eval_steps_per_second": 0.936,
+      "eval_token_acc": 0.9568038540642047,
+      "step": 8000
+    },
+    {
+      "epoch": 0.3533260632497274,
+      "grad_norm": 5.629948616027832,
+      "learning_rate": 4.783907646405187e-06,
+      "loss": 0.1462325096130371,
+      "step": 8100,
+      "token_acc": 0.9567767130904812
+    },
+    {
+      "epoch": 0.35768811341330425,
+      "grad_norm": 3.4954240322113037,
+      "learning_rate": 4.7765149439512904e-06,
+      "loss": 0.1433693313598633,
+      "step": 8200,
+      "token_acc": 0.9574897617811616
+    },
+    {
+      "epoch": 0.36205016357688113,
+      "grad_norm": 8.395429611206055,
+      "learning_rate": 4.769003814028254e-06,
+      "loss": 0.14990657806396485,
+      "step": 8300,
+      "token_acc": 0.955732380643066
+    },
+    {
+      "epoch": 0.366412213740458,
+      "grad_norm": 5.691331386566162,
+      "learning_rate": 4.761374647375488e-06,
+      "loss": 0.14862375259399413,
+      "step": 8400,
+      "token_acc": 0.9561096803907705
+    },
+    {
+      "epoch": 0.3707742639040349,
+      "grad_norm": 7.700660705566406,
+      "learning_rate": 4.753627840872838e-06,
+      "loss": 0.1447600269317627,
+      "step": 8500,
+      "token_acc": 0.9575172594313386
+    },
+    {
+      "epoch": 0.3751363140676118,
+      "grad_norm": 5.046755313873291,
+      "learning_rate": 4.745763797519937e-06,
+      "loss": 0.14034244537353516,
+      "step": 8600,
+      "token_acc": 0.9585937652755961
+    },
+    {
+      "epoch": 0.37949836423118866,
+      "grad_norm": 6.037604808807373,
+      "learning_rate": 4.737782926415242e-06,
+      "loss": 0.14538028717041016,
+      "step": 8700,
+      "token_acc": 0.956606432091631
+    },
+    {
+      "epoch": 0.38386041439476554,
+      "grad_norm": 5.605445384979248,
+      "learning_rate": 4.729685642734753e-06,
+      "loss": 0.14044106483459473,
+      "step": 8800,
+      "token_acc": 0.9584157391967529
+    },
+    {
+      "epoch": 0.3882224645583424,
+      "grad_norm": 3.2041120529174805,
+      "learning_rate": 4.721472367710412e-06,
+      "loss": 0.1447988510131836,
+      "step": 8900,
+      "token_acc": 0.9571702297990734
+    },
+    {
+      "epoch": 0.3925845147219193,
+      "grad_norm": 4.4470343589782715,
+      "learning_rate": 4.713143528608194e-06,
+      "loss": 0.14285174369812012,
+      "step": 9000,
+      "token_acc": 0.9576638578941464
+    },
+    {
+      "epoch": 0.3925845147219193,
+      "eval_loss": 0.11620920896530151,
+      "eval_runtime": 250.0276,
+      "eval_samples_per_second": 59.273,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.9578210229813109,
+      "step": 9000
+    },
+    {
+      "epoch": 0.3969465648854962,
+      "grad_norm": 5.5821356773376465,
+      "learning_rate": 4.704699558705876e-06,
+      "loss": 0.1477054214477539,
+      "step": 9100,
+      "token_acc": 0.9566980840543882
+    },
+    {
+      "epoch": 0.40130861504907306,
+      "grad_norm": 9.140554428100586,
+      "learning_rate": 4.6961408972705e-06,
+      "loss": 0.15634747505187988,
+      "step": 9200,
+      "token_acc": 0.9540431258291597
+    },
+    {
+      "epoch": 0.40567066521264994,
+      "grad_norm": 2.6849653720855713,
+      "learning_rate": 4.687467989535522e-06,
+      "loss": 0.13982583045959474,
+      "step": 9300,
+      "token_acc": 0.9585451193466928
+    },
+    {
+      "epoch": 0.4100327153762268,
+      "grad_norm": 4.507665634155273,
+      "learning_rate": 4.678681286677644e-06,
+      "loss": 0.14462839126586913,
+      "step": 9400,
+      "token_acc": 0.9575857813842605
+    },
+    {
+      "epoch": 0.4143947655398037,
+      "grad_norm": 4.900688648223877,
+      "learning_rate": 4.669781245793356e-06,
+      "loss": 0.14218985557556152,
+      "step": 9500,
+      "token_acc": 0.9583531835447112
+    },
+    {
+      "epoch": 0.4187568157033806,
+      "grad_norm": 3.7620060443878174,
+      "learning_rate": 4.6607683298751435e-06,
+      "loss": 0.13646729469299315,
+      "step": 9600,
+      "token_acc": 0.9592932889999409
+    },
+    {
+      "epoch": 0.42311886586695746,
+      "grad_norm": 5.526325225830078,
+      "learning_rate": 4.651643007787412e-06,
+      "loss": 0.13613853454589844,
+      "step": 9700,
+      "token_acc": 0.9601200068126138
+    },
+    {
+      "epoch": 0.42748091603053434,
+      "grad_norm": 6.252800464630127,
+      "learning_rate": 4.642405754242089e-06,
+      "loss": 0.14448049545288086,
+      "step": 9800,
+      "token_acc": 0.9579820567152068
+    },
+    {
+      "epoch": 0.4318429661941112,
+      "grad_norm": 5.828685283660889,
+      "learning_rate": 4.633057049773932e-06,
+      "loss": 0.14130407333374023,
+      "step": 9900,
+      "token_acc": 0.9588267632506549
+    },
+    {
+      "epoch": 0.4362050163576881,
+      "grad_norm": 6.362551689147949,
+      "learning_rate": 4.623597380715534e-06,
+      "loss": 0.1415507125854492,
+      "step": 10000,
+      "token_acc": 0.9588815789473685
+    },
+    {
+      "epoch": 0.4362050163576881,
+      "eval_loss": 0.11561030149459839,
+      "eval_runtime": 244.6104,
+      "eval_samples_per_second": 60.586,
+      "eval_steps_per_second": 0.948,
+      "eval_token_acc": 0.9586286103907717,
+      "step": 10000
+    },
+    {
+      "epoch": 0.440567066521265,
+      "grad_norm": 4.7389912605285645,
+      "learning_rate": 4.614027239172017e-06,
+      "loss": 0.13512346267700195,
+      "step": 10100,
+      "token_acc": 0.9598309226603262
+    },
+    {
+      "epoch": 0.44492911668484186,
+      "grad_norm": 6.846667289733887,
+      "learning_rate": 4.604347122995434e-06,
+      "loss": 0.1509090805053711,
+      "step": 10200,
+      "token_acc": 0.9573597748543753
+    },
+    {
+      "epoch": 0.44929116684841874,
+      "grad_norm": 6.515282154083252,
+      "learning_rate": 4.594557535758874e-06,
+      "loss": 0.1417671775817871,
+      "step": 10300,
+      "token_acc": 0.9583131932015624
+    },
+    {
+      "epoch": 0.4536532170119956,
+      "grad_norm": 4.678475379943848,
+      "learning_rate": 4.584658986730261e-06,
+      "loss": 0.13684335708618164,
+      "step": 10400,
+      "token_acc": 0.9602577662866768
+    },
+    {
+      "epoch": 0.4580152671755725,
+      "grad_norm": 7.365673542022705,
+      "learning_rate": 4.574651990845864e-06,
+      "loss": 0.13511247634887696,
+      "step": 10500,
+      "token_acc": 0.9606173846680176
+    },
+    {
+      "epoch": 0.4623773173391494,
+      "grad_norm": 8.739450454711914,
+      "learning_rate": 4.564537068683507e-06,
+      "loss": 0.1433129596710205,
+      "step": 10600,
+      "token_acc": 0.9589074287510477
+    },
+    {
+      "epoch": 0.46673936750272627,
+      "grad_norm": 5.756815433502197,
+      "learning_rate": 4.554314746435487e-06,
+      "loss": 0.13850313186645508,
+      "step": 10700,
+      "token_acc": 0.9598506612832259
+    },
+    {
+      "epoch": 0.47110141766630315,
+      "grad_norm": 5.801908493041992,
+      "learning_rate": 4.543985555881208e-06,
+      "loss": 0.1446135425567627,
+      "step": 10800,
+      "token_acc": 0.9583247382806457
+    },
+    {
+      "epoch": 0.47546346782988,
+      "grad_norm": 5.760519504547119,
+      "learning_rate": 4.533550034359506e-06,
+      "loss": 0.14331352233886718,
+      "step": 10900,
+      "token_acc": 0.9582483464207748
+    },
+    {
+      "epoch": 0.4798255179934569,
+      "grad_norm": 4.116934776306152,
+      "learning_rate": 4.523008724740705e-06,
+      "loss": 0.14439311981201172,
+      "step": 11000,
+      "token_acc": 0.9575995368881456
+    },
+    {
+      "epoch": 0.4798255179934569,
+      "eval_loss": 0.11400242149829865,
+      "eval_runtime": 249.7471,
+      "eval_samples_per_second": 59.34,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9592769158544219,
+      "step": 11000
+    },
+    {
+      "epoch": 0.4841875681570338,
+      "grad_norm": 3.357717990875244,
+      "learning_rate": 4.512362175398371e-06,
+      "loss": 0.14085225105285645,
+      "step": 11100,
+      "token_acc": 0.9586402774839236
+    },
+    {
+      "epoch": 0.48854961832061067,
+      "grad_norm": 6.080904483795166,
+      "learning_rate": 4.501610940180789e-06,
+      "loss": 0.1399165916442871,
+      "step": 11200,
+      "token_acc": 0.9586773087105651
+    },
+    {
+      "epoch": 0.49291166848418755,
+      "grad_norm": 5.0486273765563965,
+      "learning_rate": 4.490755578382145e-06,
+      "loss": 0.14417312622070313,
+      "step": 11300,
+      "token_acc": 0.9579395177290202
+    },
+    {
+      "epoch": 0.49727371864776443,
+      "grad_norm": 7.045108318328857,
+      "learning_rate": 4.479796654713438e-06,
+      "loss": 0.1347987174987793,
+      "step": 11400,
+      "token_acc": 0.9606777638720471
+    },
+    {
+      "epoch": 0.5016357688113413,
+      "grad_norm": 4.105020523071289,
+      "learning_rate": 4.468734739273095e-06,
+      "loss": 0.13610049247741698,
+      "step": 11500,
+      "token_acc": 0.9600762304805243
+    },
+    {
+      "epoch": 0.5059978189749182,
+      "grad_norm": 4.2766947746276855,
+      "learning_rate": 4.457570407517324e-06,
+      "loss": 0.13926528930664062,
+      "step": 11600,
+      "token_acc": 0.9589045543988622
+    },
+    {
+      "epoch": 0.5103598691384951,
+      "grad_norm": 17.53934097290039,
+      "learning_rate": 4.446304240230167e-06,
+      "loss": 0.13595869064331054,
+      "step": 11700,
+      "token_acc": 0.9598472992528664
+    },
+    {
+      "epoch": 0.514721919302072,
+      "grad_norm": 5.563167095184326,
+      "learning_rate": 4.434936823493293e-06,
+      "loss": 0.13069713592529297,
+      "step": 11800,
+      "token_acc": 0.9616947223112704
+    },
+    {
+      "epoch": 0.5190839694656488,
+      "grad_norm": 8.064498901367188,
+      "learning_rate": 4.4234687486555084e-06,
+      "loss": 0.13252490997314453,
+      "step": 11900,
+      "token_acc": 0.961364667811125
+    },
+    {
+      "epoch": 0.5234460196292258,
+      "grad_norm": 4.162057399749756,
+      "learning_rate": 4.4119006123019945e-06,
+      "loss": 0.14607376098632813,
+      "step": 12000,
+      "token_acc": 0.957758365222321
+    },
+    {
+      "epoch": 0.5234460196292258,
+      "eval_loss": 0.11156473308801651,
+      "eval_runtime": 250.7613,
+      "eval_samples_per_second": 59.1,
+      "eval_steps_per_second": 0.925,
+      "eval_token_acc": 0.9600286148618439,
+      "step": 12000
+    },
+    {
+      "epoch": 0.5278080697928026,
+      "grad_norm": 4.963797092437744,
+      "learning_rate": 4.400233016223271e-06,
+      "loss": 0.13640110015869142,
+      "step": 12100,
+      "token_acc": 0.9606004612884625
+    },
+    {
+      "epoch": 0.5321701199563795,
+      "grad_norm": 5.048738956451416,
+      "learning_rate": 4.38846656738389e-06,
+      "loss": 0.15161012649536132,
+      "step": 12200,
+      "token_acc": 0.9557634182813146
+    },
+    {
+      "epoch": 0.5365321701199564,
+      "grad_norm": 7.779881000518799,
+      "learning_rate": 4.3766018778908615e-06,
+      "loss": 0.1338326835632324,
+      "step": 12300,
+      "token_acc": 0.9601769911504425
+    },
+    {
+      "epoch": 0.5408942202835333,
+      "grad_norm": 4.614696502685547,
+      "learning_rate": 4.364639564961812e-06,
+      "loss": 0.12876497268676756,
+      "step": 12400,
+      "token_acc": 0.9624142336006742
+    },
+    {
+      "epoch": 0.5452562704471101,
+      "grad_norm": 4.806646823883057,
+      "learning_rate": 4.352580250892875e-06,
+      "loss": 0.134474515914917,
+      "step": 12500,
+      "token_acc": 0.9602631174026773
+    },
+    {
+      "epoch": 0.549618320610687,
+      "grad_norm": 4.253026008605957,
+      "learning_rate": 4.340424563026315e-06,
+      "loss": 0.14678755760192871,
+      "step": 12600,
+      "token_acc": 0.957807593859406
+    },
+    {
+      "epoch": 0.5539803707742639,
+      "grad_norm": 6.865671157836914,
+      "learning_rate": 4.328173133717899e-06,
+      "loss": 0.13899526596069336,
+      "step": 12700,
+      "token_acc": 0.9594970899385832
+    },
+    {
+      "epoch": 0.5583424209378408,
+      "grad_norm": 4.684392929077148,
+      "learning_rate": 4.315826600303994e-06,
+      "loss": 0.13189617156982422,
+      "step": 12800,
+      "token_acc": 0.9610336745212099
+    },
+    {
+      "epoch": 0.5627044711014176,
+      "grad_norm": 4.648969650268555,
+      "learning_rate": 4.303385605068417e-06,
+      "loss": 0.13608951568603517,
+      "step": 12900,
+      "token_acc": 0.9604944752442288
+    },
+    {
+      "epoch": 0.5670665212649946,
+      "grad_norm": 5.1123833656311035,
+      "learning_rate": 4.2908507952090174e-06,
+      "loss": 0.13563767433166504,
+      "step": 13000,
+      "token_acc": 0.960596330573991
+    },
+    {
+      "epoch": 0.5670665212649946,
+      "eval_loss": 0.1084199920296669,
+      "eval_runtime": 254.881,
+      "eval_samples_per_second": 58.145,
+      "eval_steps_per_second": 0.91,
+      "eval_token_acc": 0.9608250245908969,
+      "step": 13000
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 4.262386322021484,
+      "learning_rate": 4.278222822804015e-06,
+      "loss": 0.13114137649536134,
+      "step": 13100,
+      "token_acc": 0.962146988771691
+    },
+    {
+      "epoch": 0.5757906215921483,
+      "grad_norm": 8.251741409301758,
+      "learning_rate": 4.265502344778071e-06,
+      "loss": 0.13535155296325685,
+      "step": 13200,
+      "token_acc": 0.9604900042755529
+    },
+    {
+      "epoch": 0.5801526717557252,
+      "grad_norm": 5.114648342132568,
+      "learning_rate": 4.252690022868119e-06,
+      "loss": 0.13799442291259767,
+      "step": 13300,
+      "token_acc": 0.9601804553738738
+    },
+    {
+      "epoch": 0.5845147219193021,
+      "grad_norm": 4.916436195373535,
+      "learning_rate": 4.239786523588941e-06,
+      "loss": 0.13074012756347655,
+      "step": 13400,
+      "token_acc": 0.961597333022013
+    },
+    {
+      "epoch": 0.5888767720828789,
+      "grad_norm": 5.750058174133301,
+      "learning_rate": 4.22679251819849e-06,
+      "loss": 0.12944320678710938,
+      "step": 13500,
+      "token_acc": 0.9623505189348078
+    },
+    {
+      "epoch": 0.5932388222464559,
+      "grad_norm": 4.390042781829834,
+      "learning_rate": 4.2137086826629735e-06,
+      "loss": 0.12797122001647948,
+      "step": 13600,
+      "token_acc": 0.9626673415487246
+    },
+    {
+      "epoch": 0.5976008724100327,
+      "grad_norm": 5.039231777191162,
+      "learning_rate": 4.200535697621687e-06,
+      "loss": 0.1344309616088867,
+      "step": 13700,
+      "token_acc": 0.9587961989254508
+    },
+    {
+      "epoch": 0.6019629225736096,
+      "grad_norm": 4.506139278411865,
+      "learning_rate": 4.187274248351607e-06,
+      "loss": 0.12951894760131835,
+      "step": 13800,
+      "token_acc": 0.9618089667601382
+    },
+    {
+      "epoch": 0.6063249727371864,
+      "grad_norm": 4.769443035125732,
+      "learning_rate": 4.173925024731744e-06,
+      "loss": 0.13222533226013183,
+      "step": 13900,
+      "token_acc": 0.961495694448036
+    },
+    {
+      "epoch": 0.6106870229007634,
+      "grad_norm": 5.142453193664551,
+      "learning_rate": 4.1604887212072515e-06,
+      "loss": 0.1336323833465576,
+      "step": 14000,
+      "token_acc": 0.9605494391895384
+    },
+    {
+      "epoch": 0.6106870229007634,
+      "eval_loss": 0.10780877619981766,
+      "eval_runtime": 254.9746,
+      "eval_samples_per_second": 58.123,
+      "eval_steps_per_second": 0.91,
+      "eval_token_acc": 0.9610849056603774,
+      "step": 14000
+    },
+    {
+      "epoch": 0.6150490730643402,
+      "grad_norm": 4.960557460784912,
+      "learning_rate": 4.146966036753298e-06,
+      "loss": 0.12569026947021483,
+      "step": 14100,
+      "token_acc": 0.9631472014680823
+    },
+    {
+      "epoch": 0.6194111232279171,
+      "grad_norm": 6.882476806640625,
+      "learning_rate": 4.133357674838711e-06,
+      "loss": 0.1261822509765625,
+      "step": 14200,
+      "token_acc": 0.962310356794416
+    },
+    {
+      "epoch": 0.623773173391494,
+      "grad_norm": 4.472857475280762,
+      "learning_rate": 4.119664343389379e-06,
+      "loss": 0.14244023323059082,
+      "step": 14300,
+      "token_acc": 0.9589147138040676
+    },
+    {
+      "epoch": 0.6281352235550709,
+      "grad_norm": 4.520143508911133,
+      "learning_rate": 4.105886754751419e-06,
+      "loss": 0.14673561096191406,
+      "step": 14400,
+      "token_acc": 0.9569507419620775
+    },
+    {
+      "epoch": 0.6324972737186477,
+      "grad_norm": 5.614750385284424,
+      "learning_rate": 4.092025625654129e-06,
+      "loss": 0.13182031631469726,
+      "step": 14500,
+      "token_acc": 0.9618331926192794
+    },
+    {
+      "epoch": 0.6368593238822247,
+      "grad_norm": 5.235232830047607,
+      "learning_rate": 4.078081677172695e-06,
+      "loss": 0.13167083740234375,
+      "step": 14600,
+      "token_acc": 0.9610090699262053
+    },
+    {
+      "epoch": 0.6412213740458015,
+      "grad_norm": 6.400154113769531,
+      "learning_rate": 4.064055634690684e-06,
+      "loss": 0.1283753204345703,
+      "step": 14700,
+      "token_acc": 0.9634285038096678
+    },
+    {
+      "epoch": 0.6455834242093784,
+      "grad_norm": 4.708637237548828,
+      "learning_rate": 4.049948227862305e-06,
+      "loss": 0.12938753128051758,
+      "step": 14800,
+      "token_acc": 0.9620657866347606
+    },
+    {
+      "epoch": 0.6499454743729552,
+      "grad_norm": 3.1443099975585938,
+      "learning_rate": 4.0357601905744545e-06,
+      "loss": 0.1246177864074707,
+      "step": 14900,
+      "token_acc": 0.9637490695177302
+    },
+    {
+      "epoch": 0.6543075245365322,
+      "grad_norm": 4.834894180297852,
+      "learning_rate": 4.021492260908538e-06,
+      "loss": 0.13603127479553223,
+      "step": 15000,
+      "token_acc": 0.9604635702322964
+    },
+    {
+      "epoch": 0.6543075245365322,
+      "eval_loss": 0.10531982779502869,
+      "eval_runtime": 249.9605,
+      "eval_samples_per_second": 59.289,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.9620853080568721,
+      "step": 15000
+    },
+    {
+      "epoch": 0.658669574700109,
+      "grad_norm": 6.40881872177124,
+      "learning_rate": 4.007145181102073e-06,
+      "loss": 0.13621252059936523,
+      "step": 15100,
+      "token_acc": 0.9602403567077255
+    },
+    {
+      "epoch": 0.6630316248636859,
+      "grad_norm": 4.037698268890381,
+      "learning_rate": 3.992719697510077e-06,
+      "loss": 0.12186765670776367,
+      "step": 15200,
+      "token_acc": 0.9640692471775663
+    },
+    {
+      "epoch": 0.6673936750272628,
+      "grad_norm": 3.7594399452209473,
+      "learning_rate": 3.97821656056624e-06,
+      "loss": 0.12971938133239747,
+      "step": 15300,
+      "token_acc": 0.9622328323773512
+    },
+    {
+      "epoch": 0.6717557251908397,
+      "grad_norm": 5.0803422927856445,
+      "learning_rate": 3.963636524743891e-06,
+      "loss": 0.1299113655090332,
+      "step": 15400,
+      "token_acc": 0.9613922835754047
+    },
+    {
+      "epoch": 0.6761177753544165,
+      "grad_norm": 10.828523635864258,
+      "learning_rate": 3.948980348516742e-06,
+      "loss": 0.13089241981506347,
+      "step": 15500,
+      "token_acc": 0.9614092769880411
+    },
+    {
+      "epoch": 0.6804798255179935,
+      "grad_norm": 4.439146518707275,
+      "learning_rate": 3.934248794319435e-06,
+      "loss": 0.12397289276123047,
+      "step": 15600,
+      "token_acc": 0.963767595283974
+    },
+    {
+      "epoch": 0.6848418756815703,
+      "grad_norm": 4.211355209350586,
+      "learning_rate": 3.9194426285078794e-06,
+      "loss": 0.12579789161682128,
+      "step": 15700,
+      "token_acc": 0.9628140306004733
+    },
+    {
+      "epoch": 0.6892039258451472,
+      "grad_norm": 6.285272598266602,
+      "learning_rate": 3.904562621319385e-06,
+      "loss": 0.12318273544311524,
+      "step": 15800,
+      "token_acc": 0.9640143499860016
+    },
+    {
+      "epoch": 0.693565976008724,
+      "grad_norm": 5.098615646362305,
+      "learning_rate": 3.889609546832592e-06,
+      "loss": 0.12975415229797363,
+      "step": 15900,
+      "token_acc": 0.9626485457720456
+    },
+    {
+      "epoch": 0.697928026172301,
+      "grad_norm": 4.703002452850342,
+      "learning_rate": 3.874584182927203e-06,
+      "loss": 0.1291140651702881,
+      "step": 16000,
+      "token_acc": 0.9626422307463804
+    },
+    {
+      "epoch": 0.697928026172301,
+      "eval_loss": 0.1023755595088005,
+      "eval_runtime": 249.8751,
+      "eval_samples_per_second": 59.31,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.9632030760976482,
+      "step": 16000
+    },
+    {
+      "epoch": 0.7022900763358778,
+      "grad_norm": 5.157517910003662,
+      "learning_rate": 3.8594873112435155e-06,
+      "loss": 0.12747785568237305,
+      "step": 16100,
+      "token_acc": 0.9628675375791396
+    },
+    {
+      "epoch": 0.7066521264994547,
+      "grad_norm": 4.166143894195557,
+      "learning_rate": 3.844319717141764e-06,
+      "loss": 0.12231754302978516,
+      "step": 16200,
+      "token_acc": 0.9642421806110275
+    },
+    {
+      "epoch": 0.7110141766630316,
+      "grad_norm": 5.694368839263916,
+      "learning_rate": 3.829082189661256e-06,
+      "loss": 0.12773503303527833,
+      "step": 16300,
+      "token_acc": 0.9625522173794959
+    },
+    {
+      "epoch": 0.7153762268266085,
+      "grad_norm": 18.599973678588867,
+      "learning_rate": 3.8137755214793358e-06,
+      "loss": 0.12870462417602538,
+      "step": 16400,
+      "token_acc": 0.963150284289799
+    },
+    {
+      "epoch": 0.7197382769901853,
+      "grad_norm": 4.071221828460693,
+      "learning_rate": 3.798400508870142e-06,
+      "loss": 0.12437307357788085,
+      "step": 16500,
+      "token_acc": 0.9635778517683454
+    },
+    {
+      "epoch": 0.7241003271537623,
+      "grad_norm": 3.744788885116577,
+      "learning_rate": 3.7829579516631824e-06,
+      "loss": 0.1270986557006836,
+      "step": 16600,
+      "token_acc": 0.9625778535370091
+    },
+    {
+      "epoch": 0.7284623773173392,
+      "grad_norm": 4.1468186378479,
+      "learning_rate": 3.7674486532017314e-06,
+      "loss": 0.12220178604125977,
+      "step": 16700,
+      "token_acc": 0.9635083061616383
+    },
+    {
+      "epoch": 0.732824427480916,
+      "grad_norm": 4.590456008911133,
+      "learning_rate": 3.7518734203010353e-06,
+      "loss": 0.12810008049011232,
+      "step": 16800,
+      "token_acc": 0.9624704568475184
+    },
+    {
+      "epoch": 0.737186477644493,
+      "grad_norm": 4.199416160583496,
+      "learning_rate": 3.7362330632063403e-06,
+      "loss": 0.125726900100708,
+      "step": 16900,
+      "token_acc": 0.9630880153604012
+    },
+    {
+      "epoch": 0.7415485278080698,
+      "grad_norm": 7.344795227050781,
+      "learning_rate": 3.7205283955507455e-06,
+      "loss": 0.13037595748901368,
+      "step": 17000,
+      "token_acc": 0.9619103020492814
+    },
+    {
+      "epoch": 0.7415485278080698,
+      "eval_loss": 0.09986680746078491,
+      "eval_runtime": 249.7928,
+      "eval_samples_per_second": 59.329,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9643292273987302,
+      "step": 17000
+    },
+    {
+      "epoch": 0.7459105779716467,
+      "grad_norm": 4.345545768737793,
+      "learning_rate": 3.704760234312874e-06,
+      "loss": 0.12362366676330566,
+      "step": 17100,
+      "token_acc": 0.9646065349863416
+    },
+    {
+      "epoch": 0.7502726281352236,
+      "grad_norm": 4.425832271575928,
+      "learning_rate": 3.6889293997743723e-06,
+      "loss": 0.11739083290100098,
+      "step": 17200,
+      "token_acc": 0.965255087601945
+    },
+    {
+      "epoch": 0.7546346782988005,
+      "grad_norm": 5.167355537414551,
+      "learning_rate": 3.6730367154772414e-06,
+      "loss": 0.12582969665527344,
+      "step": 17300,
+      "token_acc": 0.9627481582817909
+    },
+    {
+      "epoch": 0.7589967284623773,
+      "grad_norm": 3.649265766143799,
+      "learning_rate": 3.65708300818099e-06,
+      "loss": 0.13049283981323243,
+      "step": 17400,
+      "token_acc": 0.9616916513063766
+    },
+    {
+      "epoch": 0.7633587786259542,
+      "grad_norm": 5.919361114501953,
+      "learning_rate": 3.6410691078196285e-06,
+      "loss": 0.12486766815185547,
+      "step": 17500,
+      "token_acc": 0.9645121544266756
+    },
+    {
+      "epoch": 0.7677208287895311,
+      "grad_norm": 3.0380961894989014,
+      "learning_rate": 3.6249958474584954e-06,
+      "loss": 0.12427781105041504,
+      "step": 17600,
+      "token_acc": 0.9628813407305854
+    },
+    {
+      "epoch": 0.772082878953108,
+      "grad_norm": 6.409759521484375,
+      "learning_rate": 3.6088640632509196e-06,
+      "loss": 0.12428503036499024,
+      "step": 17700,
+      "token_acc": 0.9637545993458708
+    },
+    {
+      "epoch": 0.7764449291166848,
+      "grad_norm": 6.19067907333374,
+      "learning_rate": 3.5926745943947185e-06,
+      "loss": 0.12289785385131836,
+      "step": 17800,
+      "token_acc": 0.963990416541882
+    },
+    {
+      "epoch": 0.7808069792802618,
+      "grad_norm": 3.6801064014434814,
+      "learning_rate": 3.5764282830885467e-06,
+      "loss": 0.1264265727996826,
+      "step": 17900,
+      "token_acc": 0.9629325823828039
+    },
+    {
+      "epoch": 0.7851690294438386,
+      "grad_norm": 10.158914566040039,
+      "learning_rate": 3.5601259744880833e-06,
+      "loss": 0.12125222206115722,
+      "step": 18000,
+      "token_acc": 0.9642156023136009
+    },
+    {
+      "epoch": 0.7851690294438386,
+      "eval_loss": 0.0974874198436737,
+      "eval_runtime": 249.4923,
+      "eval_samples_per_second": 59.401,
+      "eval_steps_per_second": 0.93,
+      "eval_token_acc": 0.9646729410712689,
+      "step": 18000
+    },
+    {
+      "epoch": 0.7895310796074155,
+      "grad_norm": 7.951091289520264,
+      "learning_rate": 3.543768516662063e-06,
+      "loss": 0.12031344413757324,
+      "step": 18100,
+      "token_acc": 0.9641105894627021
+    },
+    {
+      "epoch": 0.7938931297709924,
+      "grad_norm": 4.855859756469727,
+      "learning_rate": 3.527356760548159e-06,
+      "loss": 0.11818990707397461,
+      "step": 18200,
+      "token_acc": 0.9647326801868844
+    },
+    {
+      "epoch": 0.7982551799345693,
+      "grad_norm": 8.738151550292969,
+      "learning_rate": 3.51089155990872e-06,
+      "loss": 0.1165078067779541,
+      "step": 18300,
+      "token_acc": 0.9654456415279138
+    },
+    {
+      "epoch": 0.8026172300981461,
+      "grad_norm": 5.64874792098999,
+      "learning_rate": 3.494373771286349e-06,
+      "loss": 0.120614013671875,
+      "step": 18400,
+      "token_acc": 0.9641536709864776
+    },
+    {
+      "epoch": 0.806979280261723,
+      "grad_norm": 4.42671537399292,
+      "learning_rate": 3.477804253959352e-06,
+      "loss": 0.1219012451171875,
+      "step": 18500,
+      "token_acc": 0.9641036100346726
+    },
+    {
+      "epoch": 0.8113413304252999,
+      "grad_norm": 4.415754318237305,
+      "learning_rate": 3.4611838698970335e-06,
+      "loss": 0.11962775230407714,
+      "step": 18600,
+      "token_acc": 0.9655911722473921
+    },
+    {
+      "epoch": 0.8157033805888768,
+      "grad_norm": 6.6899027824401855,
+      "learning_rate": 3.4445134837148553e-06,
+      "loss": 0.11862802505493164,
+      "step": 18700,
+      "token_acc": 0.9649808710911543
+    },
+    {
+      "epoch": 0.8200654307524536,
+      "grad_norm": 3.721449375152588,
+      "learning_rate": 3.427793962629459e-06,
+      "loss": 0.1211440086364746,
+      "step": 18800,
+      "token_acc": 0.9638717768052799
+    },
+    {
+      "epoch": 0.8244274809160306,
+      "grad_norm": 4.877586841583252,
+      "learning_rate": 3.4110261764135525e-06,
+      "loss": 0.11775184631347656,
+      "step": 18900,
+      "token_acc": 0.9653165583354041
+    },
+    {
+      "epoch": 0.8287895310796074,
+      "grad_norm": 4.989749908447266,
+      "learning_rate": 3.3942109973506636e-06,
+      "loss": 0.1220939540863037,
+      "step": 19000,
+      "token_acc": 0.9634719545856987
+    },
+    {
+      "epoch": 0.8287895310796074,
+      "eval_loss": 0.09567935019731522,
+      "eval_runtime": 249.6723,
+      "eval_samples_per_second": 59.358,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9653603684163462,
+      "step": 19000
+    },
+    {
+      "epoch": 0.8331515812431843,
+      "grad_norm": 7.156066417694092,
+      "learning_rate": 3.377349300189761e-06,
+      "loss": 0.12005048751831054,
+      "step": 19100,
+      "token_acc": 0.9646030979629654
+    },
+    {
+      "epoch": 0.8375136314067612,
+      "grad_norm": 5.540706157684326,
+      "learning_rate": 3.360441962099748e-06,
+      "loss": 0.11568243980407715,
+      "step": 19200,
+      "token_acc": 0.9662971175166297
+    },
+    {
+      "epoch": 0.8418756815703381,
+      "grad_norm": 4.6166510581970215,
+      "learning_rate": 3.343489862623836e-06,
+      "loss": 0.11525155067443847,
+      "step": 19300,
+      "token_acc": 0.9660013046314416
+    },
+    {
+      "epoch": 0.8462377317339149,
+      "grad_norm": 3.9313056468963623,
+      "learning_rate": 3.326493883633783e-06,
+      "loss": 0.11743115425109864,
+      "step": 19400,
+      "token_acc": 0.9656000417601921
+    },
+    {
+      "epoch": 0.8505997818974919,
+      "grad_norm": 4.582152366638184,
+      "learning_rate": 3.3094549092840195e-06,
+      "loss": 0.12221628189086914,
+      "step": 19500,
+      "token_acc": 0.9639358598680389
+    },
+    {
+      "epoch": 0.8549618320610687,
+      "grad_norm": 10.774201393127441,
+      "learning_rate": 3.2923738259656585e-06,
+      "loss": 0.11821553230285645,
+      "step": 19600,
+      "token_acc": 0.9650054941503458
+    },
+    {
+      "epoch": 0.8593238822246456,
+      "grad_norm": 4.74281644821167,
+      "learning_rate": 3.2752515222603766e-06,
+      "loss": 0.11718135833740234,
+      "step": 19700,
+      "token_acc": 0.9661495155373397
+    },
+    {
+      "epoch": 0.8636859323882224,
+      "grad_norm": 5.704650402069092,
+      "learning_rate": 3.2580888888941908e-06,
+      "loss": 0.11361128807067872,
+      "step": 19800,
+      "token_acc": 0.96557928802589
+    },
+    {
+      "epoch": 0.8680479825517994,
+      "grad_norm": 2.6005077362060547,
+      "learning_rate": 3.2408868186911285e-06,
+      "loss": 0.11735478401184082,
+      "step": 19900,
+      "token_acc": 0.9663002389186196
+    },
+    {
+      "epoch": 0.8724100327153762,
+      "grad_norm": 7.039544582366943,
+      "learning_rate": 3.2236462065267715e-06,
+      "loss": 0.1120327091217041,
+      "step": 20000,
+      "token_acc": 0.9665029780149579
+    },
+    {
+      "epoch": 0.8724100327153762,
+      "eval_loss": 0.09346429258584976,
+      "eval_runtime": 254.7297,
+      "eval_samples_per_second": 58.179,
+      "eval_steps_per_second": 0.911,
+      "eval_token_acc": 0.9658605696145935,
+      "step": 20000
+    },
+    {
+      "epoch": 0.8767720828789531,
+      "grad_norm": 4.584099769592285,
+      "learning_rate": 3.206367949281708e-06,
+      "loss": 0.11660341262817382,
+      "step": 20100,
+      "token_acc": 0.965891313165129
+    },
+    {
+      "epoch": 0.88113413304253,
+      "grad_norm": 3.811115026473999,
+      "learning_rate": 3.1890529457948794e-06,
+      "loss": 0.11867828369140625,
+      "step": 20200,
+      "token_acc": 0.9647862570539326
+    },
+    {
+      "epoch": 0.8854961832061069,
+      "grad_norm": 3.324962854385376,
+      "learning_rate": 3.171702096816814e-06,
+      "loss": 0.1159024715423584,
+      "step": 20300,
+      "token_acc": 0.9654955434176213
+    },
+    {
+      "epoch": 0.8898582333696837,
+      "grad_norm": 9.917572021484375,
+      "learning_rate": 3.1543163049627727e-06,
+      "loss": 0.10901348114013672,
+      "step": 20400,
+      "token_acc": 0.9676555921987456
+    },
+    {
+      "epoch": 0.8942202835332607,
+      "grad_norm": 5.332941055297852,
+      "learning_rate": 3.136896474665796e-06,
+      "loss": 0.11584772109985352,
+      "step": 20500,
+      "token_acc": 0.9658386091204165
+    },
+    {
+      "epoch": 0.8985823336968375,
+      "grad_norm": 4.178663730621338,
+      "learning_rate": 3.1194435121296477e-06,
+      "loss": 0.11288185119628906,
+      "step": 20600,
+      "token_acc": 0.9666555365689182
+    },
+    {
+      "epoch": 0.9029443838604144,
+      "grad_norm": 5.721475124359131,
+      "learning_rate": 3.10195832528168e-06,
+      "loss": 0.11232204437255859,
+      "step": 20700,
+      "token_acc": 0.9669557900187714
+    },
+    {
+      "epoch": 0.9073064340239912,
+      "grad_norm": 2.847933053970337,
+      "learning_rate": 3.0844418237255962e-06,
+      "loss": 0.11065910339355468,
+      "step": 20800,
+      "token_acc": 0.9669379220157056
+    },
+    {
+      "epoch": 0.9116684841875682,
+      "grad_norm": 7.752315521240234,
+      "learning_rate": 3.0668949186941357e-06,
+      "loss": 0.11427606582641602,
+      "step": 20900,
+      "token_acc": 0.9659426171427099
+    },
+    {
+      "epoch": 0.916030534351145,
+      "grad_norm": 4.635708332061768,
+      "learning_rate": 3.049318523001669e-06,
+      "loss": 0.11044464111328126,
+      "step": 21000,
+      "token_acc": 0.9672973482754359
+    },
+    {
+      "epoch": 0.916030534351145,
+      "eval_loss": 0.09045585989952087,
+      "eval_runtime": 254.415,
+      "eval_samples_per_second": 58.251,
+      "eval_steps_per_second": 0.912,
+      "eval_token_acc": 0.966824644549763,
+      "step": 21000
+    },
+    {
+      "epoch": 0.920392584514722,
+      "grad_norm": 4.291543483734131,
+      "learning_rate": 3.0317135509967095e-06,
+      "loss": 0.11297191619873047,
+      "step": 21100,
+      "token_acc": 0.9673089666463106
+    },
+    {
+      "epoch": 0.9247546346782988,
+      "grad_norm": 4.2850189208984375,
+      "learning_rate": 3.014080918514353e-06,
+      "loss": 0.11317073822021484,
+      "step": 21200,
+      "token_acc": 0.9658741385799385
+    },
+    {
+      "epoch": 0.9291166848418757,
+      "grad_norm": 3.6628763675689697,
+      "learning_rate": 2.99642154282863e-06,
+      "loss": 0.10480199813842773,
+      "step": 21300,
+      "token_acc": 0.9688065128551662
+    },
+    {
+      "epoch": 0.9334787350054525,
+      "grad_norm": 6.277792930603027,
+      "learning_rate": 2.97873634260479e-06,
+      "loss": 0.1182522201538086,
+      "step": 21400,
+      "token_acc": 0.9656542191019528
+    },
+    {
+      "epoch": 0.9378407851690295,
+      "grad_norm": 3.970945358276367,
+      "learning_rate": 2.961026237851511e-06,
+      "loss": 0.10972850799560546,
+      "step": 21500,
+      "token_acc": 0.9674839091324113
+    },
+    {
+      "epoch": 0.9422028353326063,
+      "grad_norm": 5.1818366050720215,
+      "learning_rate": 2.94329214987304e-06,
+      "loss": 0.10691708564758301,
+      "step": 21600,
+      "token_acc": 0.9680576689964666
+    },
+    {
+      "epoch": 0.9465648854961832,
+      "grad_norm": 6.335598945617676,
+      "learning_rate": 2.925535001221262e-06,
+      "loss": 0.1139387321472168,
+      "step": 21700,
+      "token_acc": 0.966435651801084
+    },
+    {
+      "epoch": 0.95092693565976,
+      "grad_norm": 4.5589599609375,
+      "learning_rate": 2.9077557156477124e-06,
+      "loss": 0.11017286300659179,
+      "step": 21800,
+      "token_acc": 0.967085074906488
+    },
+    {
+      "epoch": 0.955288985823337,
+      "grad_norm": 5.797848701477051,
+      "learning_rate": 2.889955218055521e-06,
+      "loss": 0.10730340957641601,
+      "step": 21900,
+      "token_acc": 0.9681335161263331
+    },
+    {
+      "epoch": 0.9596510359869138,
+      "grad_norm": 4.161041736602783,
+      "learning_rate": 2.8721344344512934e-06,
+      "loss": 0.11035999298095703,
+      "step": 22000,
+      "token_acc": 0.9674398288808526
+    },
+    {
+      "epoch": 0.9596510359869138,
+      "eval_loss": 0.08762744069099426,
+      "eval_runtime": 249.5745,
+      "eval_samples_per_second": 59.381,
+      "eval_steps_per_second": 0.93,
+      "eval_token_acc": 0.9679787400518645,
+      "step": 22000
+    },
+    {
+      "epoch": 0.9640130861504908,
+      "grad_norm": 3.1247737407684326,
+      "learning_rate": 2.854294291896944e-06,
+      "loss": 0.1087168025970459,
+      "step": 22100,
+      "token_acc": 0.9678269463050401
+    },
+    {
+      "epoch": 0.9683751363140676,
+      "grad_norm": 6.124391078948975,
+      "learning_rate": 2.8364357184614668e-06,
+      "loss": 0.11305822372436523,
+      "step": 22200,
+      "token_acc": 0.9667042707493957
+    },
+    {
+      "epoch": 0.9727371864776445,
+      "grad_norm": 8.205916404724121,
+      "learning_rate": 2.8185596431726527e-06,
+      "loss": 0.11026342391967774,
+      "step": 22300,
+      "token_acc": 0.9671849023275145
+    },
+    {
+      "epoch": 0.9770992366412213,
+      "grad_norm": 3.7859928607940674,
+      "learning_rate": 2.8006669959687668e-06,
+      "loss": 0.10871770858764648,
+      "step": 22400,
+      "token_acc": 0.9682362974681484
+    },
+    {
+      "epoch": 0.9814612868047983,
+      "grad_norm": 7.330371856689453,
+      "learning_rate": 2.7827587076501673e-06,
+      "loss": 0.11897387504577636,
+      "step": 22500,
+      "token_acc": 0.9650655582069638
+    },
+    {
+      "epoch": 0.9858233369683751,
+      "grad_norm": 6.279266834259033,
+      "learning_rate": 2.764835709830884e-06,
+      "loss": 0.10861226081848145,
+      "step": 22600,
+      "token_acc": 0.9683295584335133
+    },
+    {
+      "epoch": 0.990185387131952,
+      "grad_norm": 3.766507148742676,
+      "learning_rate": 2.7468989348901555e-06,
+      "loss": 0.10774114608764648,
+      "step": 22700,
+      "token_acc": 0.9677088507163268
+    },
+    {
+      "epoch": 0.9945474372955289,
+      "grad_norm": 4.713306903839111,
+      "learning_rate": 2.728949315923927e-06,
+      "loss": 0.09772121429443359,
+      "step": 22800,
+      "token_acc": 0.970906443111303
+    },
+    {
+      "epoch": 0.9989094874591058,
+      "grad_norm": 3.5282657146453857,
+      "learning_rate": 2.7109877866963063e-06,
+      "loss": 0.10501280784606934,
+      "step": 22900,
+      "token_acc": 0.968406404054063
+    },
+    {
+      "epoch": 1.0032715376226826,
+      "grad_norm": 6.511656284332275,
+      "learning_rate": 2.6930152815909894e-06,
+      "loss": 0.08737098693847656,
+      "step": 23000,
+      "token_acc": 0.9734185907901215
+    },
+    {
+      "epoch": 1.0032715376226826,
+      "eval_loss": 0.0876716896891594,
+      "eval_runtime": 249.8103,
+      "eval_samples_per_second": 59.325,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9688981042654028,
+      "step": 23000
+    },
+    {
+      "epoch": 1.0076335877862594,
+      "grad_norm": 4.062010288238525,
+      "learning_rate": 2.6750327355626536e-06,
+      "loss": 0.07467103481292725,
+      "step": 23100,
+      "token_acc": 0.9765526602582927
+    },
+    {
+      "epoch": 1.0119956379498365,
+      "grad_norm": 7.563171863555908,
+      "learning_rate": 2.657041084088318e-06,
+      "loss": 0.07544662952423095,
+      "step": 23200,
+      "token_acc": 0.9759494669313387
+    },
+    {
+      "epoch": 1.0163576881134133,
+      "grad_norm": 4.794340133666992,
+      "learning_rate": 2.6390412631186802e-06,
+      "loss": 0.07672032833099365,
+      "step": 23300,
+      "token_acc": 0.9753995963996285
+    },
+    {
+      "epoch": 1.0207197382769901,
+      "grad_norm": 2.7201950550079346,
+      "learning_rate": 2.6210342090294266e-06,
+      "loss": 0.08015254974365234,
+      "step": 23400,
+      "token_acc": 0.9749328386849175
+    },
+    {
+      "epoch": 1.025081788440567,
+      "grad_norm": 2.8014349937438965,
+      "learning_rate": 2.603020858572521e-06,
+      "loss": 0.0747481918334961,
+      "step": 23500,
+      "token_acc": 0.9764472630518457
+    },
+    {
+      "epoch": 1.029443838604144,
+      "grad_norm": 4.592669486999512,
+      "learning_rate": 2.5850021488274694e-06,
+      "loss": 0.07323034763336182,
+      "step": 23600,
+      "token_acc": 0.9767670199622862
+    },
+    {
+      "epoch": 1.0338058887677208,
+      "grad_norm": 28.001813888549805,
+      "learning_rate": 2.566979017152581e-06,
+      "loss": 0.07620226860046386,
+      "step": 23700,
+      "token_acc": 0.9755634028892456
+    },
+    {
+      "epoch": 1.0381679389312977,
+      "grad_norm": 3.804150342941284,
+      "learning_rate": 2.5489524011361962e-06,
+      "loss": 0.07047548294067382,
+      "step": 23800,
+      "token_acc": 0.9773679250222524
+    },
+    {
+      "epoch": 1.0425299890948745,
+      "grad_norm": 4.701569080352783,
+      "learning_rate": 2.5309232385479155e-06,
+      "loss": 0.07270308017730713,
+      "step": 23900,
+      "token_acc": 0.977279508101026
+    },
+    {
+      "epoch": 1.0468920392584515,
+      "grad_norm": 3.494173288345337,
+      "learning_rate": 2.5128924672898183e-06,
+      "loss": 0.07390243530273438,
+      "step": 24000,
+      "token_acc": 0.9771897454554904
+    },
+    {
+      "epoch": 1.0468920392584515,
+      "eval_loss": 0.08581704646348953,
+      "eval_runtime": 256.0699,
+      "eval_samples_per_second": 57.875,
+      "eval_steps_per_second": 0.906,
+      "eval_token_acc": 0.9695715595099705,
+      "step": 24000
+    },
+    {
+      "epoch": 1.0512540894220284,
+      "grad_norm": 8.801218032836914,
+      "learning_rate": 2.4948610253476664e-06,
+      "loss": 0.0676074504852295,
+      "step": 24100,
+      "token_acc": 0.9789494252124923
+    },
+    {
+      "epoch": 1.0556161395856052,
+      "grad_norm": 19.8681583404541,
+      "learning_rate": 2.4768298507421133e-06,
+      "loss": 0.06861214637756348,
+      "step": 24200,
+      "token_acc": 0.9784243812299924
+    },
+    {
+      "epoch": 1.059978189749182,
+      "grad_norm": 4.816915512084961,
+      "learning_rate": 2.458799881479905e-06,
+      "loss": 0.06901247501373291,
+      "step": 24300,
+      "token_acc": 0.9777164475725784
+    },
+    {
+      "epoch": 1.064340239912759,
+      "grad_norm": 4.382708549499512,
+      "learning_rate": 2.4407720555050827e-06,
+      "loss": 0.07236534595489502,
+      "step": 24400,
+      "token_acc": 0.9769710653737226
+    },
+    {
+      "epoch": 1.0687022900763359,
+      "grad_norm": 4.277945041656494,
+      "learning_rate": 2.4227473106501922e-06,
+      "loss": 0.07162453174591064,
+      "step": 24500,
+      "token_acc": 0.9774640927444261
+    },
+    {
+      "epoch": 1.0730643402399127,
+      "grad_norm": 2.9444453716278076,
+      "learning_rate": 2.404726584587495e-06,
+      "loss": 0.07076330184936523,
+      "step": 24600,
+      "token_acc": 0.9775714884256834
+    },
+    {
+      "epoch": 1.0774263904034895,
+      "grad_norm": 4.69867467880249,
+      "learning_rate": 2.386710814780189e-06,
+      "loss": 0.07056962966918945,
+      "step": 24700,
+      "token_acc": 0.9776056401564706
+    },
+    {
+      "epoch": 1.0817884405670666,
+      "grad_norm": 4.424916744232178,
+      "learning_rate": 2.3687009384336403e-06,
+      "loss": 0.0715576934814453,
+      "step": 24800,
+      "token_acc": 0.977195460013599
+    },
+    {
+      "epoch": 1.0861504907306434,
+      "grad_norm": 2.6273691654205322,
+      "learning_rate": 2.3506978924466305e-06,
+      "loss": 0.0693655014038086,
+      "step": 24900,
+      "token_acc": 0.9780883749365871
+    },
+    {
+      "epoch": 1.0905125408942202,
+      "grad_norm": 3.6074914932250977,
+      "learning_rate": 2.332702613362614e-06,
+      "loss": 0.0702785873413086,
+      "step": 25000,
+      "token_acc": 0.9770827224448572
+    },
+    {
+      "epoch": 1.0905125408942202,
+      "eval_loss": 0.08393809199333191,
+      "eval_runtime": 250.0137,
+      "eval_samples_per_second": 59.277,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.9700745551283197,
+      "step": 25000
+    },
+    {
+      "epoch": 1.0948745910577973,
+      "grad_norm": 4.213324546813965,
+      "learning_rate": 2.3147160373210036e-06,
+      "loss": 0.07728311538696289,
+      "step": 25100,
+      "token_acc": 0.9752774455718015
+    },
+    {
+      "epoch": 1.099236641221374,
+      "grad_norm": 4.420457363128662,
+      "learning_rate": 2.2967391000084647e-06,
+      "loss": 0.07622474670410156,
+      "step": 25200,
+      "token_acc": 0.9757396866456227
+    },
+    {
+      "epoch": 1.103598691384951,
+      "grad_norm": 4.92677640914917,
+      "learning_rate": 2.2787727366102435e-06,
+      "loss": 0.07702836990356446,
+      "step": 25300,
+      "token_acc": 0.9761410922020535
+    },
+    {
+      "epoch": 1.1079607415485277,
+      "grad_norm": 5.670831203460693,
+      "learning_rate": 2.2608178817615165e-06,
+      "loss": 0.07056490898132324,
+      "step": 25400,
+      "token_acc": 0.9777433009634354
+    },
+    {
+      "epoch": 1.1123227917121046,
+      "grad_norm": 7.146980285644531,
+      "learning_rate": 2.2428754694987715e-06,
+      "loss": 0.0676431655883789,
+      "step": 25500,
+      "token_acc": 0.978816706848244
+    },
+    {
+      "epoch": 1.1166848418756816,
+      "grad_norm": 2.4701433181762695,
+      "learning_rate": 2.224946433211212e-06,
+      "loss": 0.07373625755310059,
+      "step": 25600,
+      "token_acc": 0.9769120234037206
+    },
+    {
+      "epoch": 1.1210468920392584,
+      "grad_norm": 5.163082122802734,
+      "learning_rate": 2.207031705592207e-06,
+      "loss": 0.0715975284576416,
+      "step": 25700,
+      "token_acc": 0.9775948493313291
+    },
+    {
+      "epoch": 1.1254089422028353,
+      "grad_norm": 3.2471859455108643,
+      "learning_rate": 2.189132218590769e-06,
+      "loss": 0.07088047981262208,
+      "step": 25800,
+      "token_acc": 0.9779005524861878
+    },
+    {
+      "epoch": 1.1297709923664123,
+      "grad_norm": 4.81692361831665,
+      "learning_rate": 2.1712489033630706e-06,
+      "loss": 0.06931174755096435,
+      "step": 25900,
+      "token_acc": 0.9776110790536642
+    },
+    {
+      "epoch": 1.1341330425299891,
+      "grad_norm": 6.65491247177124,
+      "learning_rate": 2.153382690224007e-06,
+      "loss": 0.07666608810424805,
+      "step": 26000,
+      "token_acc": 0.9760280797392595
+    },
+    {
+      "epoch": 1.1341330425299891,
+      "eval_loss": 0.08234628289937973,
+      "eval_runtime": 250.0101,
+      "eval_samples_per_second": 59.278,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.9705160735044264,
+      "step": 26000
+    },
+    {
+      "epoch": 1.138495092693566,
+      "grad_norm": 4.733185768127441,
+      "learning_rate": 2.1355345085988014e-06,
+      "loss": 0.0709274673461914,
+      "step": 26100,
+      "token_acc": 0.9779928213603264
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 5.480454444885254,
+      "learning_rate": 2.117705286974649e-06,
+      "loss": 0.07390942573547363,
+      "step": 26200,
+      "token_acc": 0.9774670075669534
+    },
+    {
+      "epoch": 1.1472191930207196,
+      "grad_norm": 4.35108757019043,
+      "learning_rate": 2.0998959528524202e-06,
+      "loss": 0.06665233612060546,
+      "step": 26300,
+      "token_acc": 0.9792105141620537
+    },
+    {
+      "epoch": 1.1515812431842967,
+      "grad_norm": 5.042696952819824,
+      "learning_rate": 2.082107432698413e-06,
+      "loss": 0.06301040172576905,
+      "step": 26400,
+      "token_acc": 0.9799591431893273
+    },
+    {
+      "epoch": 1.1559432933478735,
+      "grad_norm": 2.658958911895752,
+      "learning_rate": 2.064340651896149e-06,
+      "loss": 0.07521211624145507,
+      "step": 26500,
+      "token_acc": 0.9763495912473549
+    },
+    {
+      "epoch": 1.1603053435114503,
+      "grad_norm": 4.5513153076171875,
+      "learning_rate": 2.0465965346982427e-06,
+      "loss": 0.06404022216796874,
+      "step": 26600,
+      "token_acc": 0.9796154142384021
+    },
+    {
+      "epoch": 1.1646673936750274,
+      "grad_norm": 6.858781814575195,
+      "learning_rate": 2.028876004178315e-06,
+      "loss": 0.06282252311706543,
+      "step": 26700,
+      "token_acc": 0.9801569262999783
+    },
+    {
+      "epoch": 1.1690294438386042,
+      "grad_norm": 4.299166679382324,
+      "learning_rate": 2.011179982182973e-06,
+      "loss": 0.06631457328796386,
+      "step": 26800,
+      "token_acc": 0.9796262194883046
+    },
+    {
+      "epoch": 1.173391494002181,
+      "grad_norm": 3.9742937088012695,
+      "learning_rate": 1.9935093892838608e-06,
+      "loss": 0.0690745735168457,
+      "step": 26900,
+      "token_acc": 0.978327122747711
+    },
+    {
+      "epoch": 1.1777535441657578,
+      "grad_norm": 2.4530298709869385,
+      "learning_rate": 1.9758651447297606e-06,
+      "loss": 0.06880934715270996,
+      "step": 27000,
+      "token_acc": 0.978106382841394
+    },
+    {
+      "epoch": 1.1777535441657578,
+      "eval_loss": 0.08033985644578934,
+      "eval_runtime": 244.5702,
+      "eval_samples_per_second": 60.596,
+      "eval_steps_per_second": 0.949,
+      "eval_token_acc": 0.9716534024859161,
+      "step": 27000
+    },
+    {
+      "epoch": 1.1821155943293349,
+      "grad_norm": 3.7700579166412354,
+      "learning_rate": 1.9582481663987795e-06,
+      "loss": 0.06770928382873535,
+      "step": 27100,
+      "token_acc": 0.9788943939453049
+    },
+    {
+      "epoch": 1.1864776444929117,
+      "grad_norm": 5.555079460144043,
+      "learning_rate": 1.9406593707506e-06,
+      "loss": 0.06606301307678222,
+      "step": 27200,
+      "token_acc": 0.9796218185617092
+    },
+    {
+      "epoch": 1.1908396946564885,
+      "grad_norm": 3.66662859916687,
+      "learning_rate": 1.923099672778798e-06,
+      "loss": 0.06830612182617188,
+      "step": 27300,
+      "token_acc": 0.9784953863415075
+    },
+    {
+      "epoch": 1.1952017448200654,
+      "grad_norm": 2.253833532333374,
+      "learning_rate": 1.9055699859632502e-06,
+      "loss": 0.06353707313537597,
+      "step": 27400,
+      "token_acc": 0.9802849547555753
+    },
+    {
+      "epoch": 1.1995637949836424,
+      "grad_norm": 5.8428449630737305,
+      "learning_rate": 1.8880712222226127e-06,
+      "loss": 0.06870734214782714,
+      "step": 27500,
+      "token_acc": 0.9787229893127389
+    },
+    {
+      "epoch": 1.2039258451472192,
+      "grad_norm": 2.712040424346924,
+      "learning_rate": 1.8706042918668755e-06,
+      "loss": 0.06457172870635987,
+      "step": 27600,
+      "token_acc": 0.9792487660543853
+    },
+    {
+      "epoch": 1.208287895310796,
+      "grad_norm": 4.38408899307251,
+      "learning_rate": 1.8531701035500172e-06,
+      "loss": 0.07690935134887696,
+      "step": 27700,
+      "token_acc": 0.9740492962199486
+    },
+    {
+      "epoch": 1.2126499454743729,
+      "grad_norm": 2.473857879638672,
+      "learning_rate": 1.835769564222728e-06,
+      "loss": 0.06370730876922608,
+      "step": 27800,
+      "token_acc": 0.979825486251706
+    },
+    {
+      "epoch": 1.21701199563795,
+      "grad_norm": 3.47666072845459,
+      "learning_rate": 1.8184035790852278e-06,
+      "loss": 0.07317141532897949,
+      "step": 27900,
+      "token_acc": 0.9769554903352452
+    },
+    {
+      "epoch": 1.2213740458015268,
+      "grad_norm": 3.806673765182495,
+      "learning_rate": 1.801073051540185e-06,
+      "loss": 0.06659706592559815,
+      "step": 28000,
+      "token_acc": 0.9792695661086874
+    },
+    {
+      "epoch": 1.2213740458015268,
+      "eval_loss": 0.07924070209264755,
+      "eval_runtime": 249.7936,
+      "eval_samples_per_second": 59.329,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.972083743181615,
+      "step": 28000
+    },
+    {
+      "epoch": 1.2257360959651036,
+      "grad_norm": 2.7492287158966064,
+      "learning_rate": 1.783778883145712e-06,
+      "loss": 0.06997788429260254,
+      "step": 28100,
+      "token_acc": 0.9779032340469014
+    },
+    {
+      "epoch": 1.2300981461286804,
+      "grad_norm": 5.293479919433594,
+      "learning_rate": 1.7665219735684658e-06,
+      "loss": 0.07057114601135255,
+      "step": 28200,
+      "token_acc": 0.9777492239580683
+    },
+    {
+      "epoch": 1.2344601962922575,
+      "grad_norm": 4.3521599769592285,
+      "learning_rate": 1.7493032205368513e-06,
+      "loss": 0.0686982250213623,
+      "step": 28300,
+      "token_acc": 0.978551340703205
+    },
+    {
+      "epoch": 1.2388222464558343,
+      "grad_norm": 5.842819690704346,
+      "learning_rate": 1.7321235197943168e-06,
+      "loss": 0.0651836109161377,
+      "step": 28400,
+      "token_acc": 0.979650824442289
+    },
+    {
+      "epoch": 1.243184296619411,
+      "grad_norm": 3.996281623840332,
+      "learning_rate": 1.7149837650527523e-06,
+      "loss": 0.06602362632751464,
+      "step": 28500,
+      "token_acc": 0.9792255874942151
+    },
+    {
+      "epoch": 1.247546346782988,
+      "grad_norm": 2.58259654045105,
+      "learning_rate": 1.697884847946006e-06,
+      "loss": 0.062000880241394045,
+      "step": 28600,
+      "token_acc": 0.9799171574655362
+    },
+    {
+      "epoch": 1.2519083969465647,
+      "grad_norm": 4.274127960205078,
+      "learning_rate": 1.6808276579834943e-06,
+      "loss": 0.06488137245178223,
+      "step": 28700,
+      "token_acc": 0.9792955370379838
+    },
+    {
+      "epoch": 1.2562704471101418,
+      "grad_norm": 4.309569358825684,
+      "learning_rate": 1.6638130825039275e-06,
+      "loss": 0.0708838415145874,
+      "step": 28800,
+      "token_acc": 0.9784192156078426
+    },
+    {
+      "epoch": 1.2606324972737186,
+      "grad_norm": 3.1259348392486572,
+      "learning_rate": 1.6468420066291536e-06,
+      "loss": 0.061966466903686526,
+      "step": 28900,
+      "token_acc": 0.9806353930430375
+    },
+    {
+      "epoch": 1.2649945474372954,
+      "grad_norm": 3.426515817642212,
+      "learning_rate": 1.6299153132181106e-06,
+      "loss": 0.0645901918411255,
+      "step": 29000,
+      "token_acc": 0.9794154619736015
+    },
+    {
+      "epoch": 1.2649945474372954,
+      "eval_loss": 0.07756964862346649,
+      "eval_runtime": 254.7344,
+      "eval_samples_per_second": 58.178,
+      "eval_steps_per_second": 0.911,
+      "eval_token_acc": 0.9729472189931145,
+      "step": 29000
+    },
+    {
+      "epoch": 1.2693565976008725,
+      "grad_norm": 6.645819187164307,
+      "learning_rate": 1.6130338828208968e-06,
+      "loss": 0.06351682662963867,
+      "step": 29100,
+      "token_acc": 0.9802265678215497
+    },
+    {
+      "epoch": 1.2737186477644493,
+      "grad_norm": 4.421413421630859,
+      "learning_rate": 1.5961985936329694e-06,
+      "loss": 0.06718688488006591,
+      "step": 29200,
+      "token_acc": 0.9788990531214893
+    },
+    {
+      "epoch": 1.2780806979280261,
+      "grad_norm": 3.6267154216766357,
+      "learning_rate": 1.5794103214494528e-06,
+      "loss": 0.06186320304870605,
+      "step": 29300,
+      "token_acc": 0.9807249300449015
+    },
+    {
+      "epoch": 1.282442748091603,
+      "grad_norm": 3.70088791847229,
+      "learning_rate": 1.5626699396195821e-06,
+      "loss": 0.06981894493103027,
+      "step": 29400,
+      "token_acc": 0.9779153963414634
+    },
+    {
+      "epoch": 1.28680479825518,
+      "grad_norm": 5.71915340423584,
+      "learning_rate": 1.5459783190012707e-06,
+      "loss": 0.0649720811843872,
+      "step": 29500,
+      "token_acc": 0.9794114456024354
+    },
+    {
+      "epoch": 1.2911668484187568,
+      "grad_norm": 3.4748384952545166,
+      "learning_rate": 1.5293363279158055e-06,
+      "loss": 0.06922075271606445,
+      "step": 29600,
+      "token_acc": 0.9789335302041239
+    },
+    {
+      "epoch": 1.2955288985823337,
+      "grad_norm": 2.7103006839752197,
+      "learning_rate": 1.5127448321026756e-06,
+      "loss": 0.06484994888305665,
+      "step": 29700,
+      "token_acc": 0.979845880260818
+    },
+    {
+      "epoch": 1.2998909487459107,
+      "grad_norm": 7.105724334716797,
+      "learning_rate": 1.4962046946745369e-06,
+      "loss": 0.06801267147064209,
+      "step": 29800,
+      "token_acc": 0.9781458581259378
+    },
+    {
+      "epoch": 1.3042529989094875,
+      "grad_norm": 4.009284019470215,
+      "learning_rate": 1.4797167760723102e-06,
+      "loss": 0.06112378120422363,
+      "step": 29900,
+      "token_acc": 0.9805348371839185
+    },
+    {
+      "epoch": 1.3086150490730644,
+      "grad_norm": 3.753431558609009,
+      "learning_rate": 1.4632819340204208e-06,
+      "loss": 0.06059444427490234,
+      "step": 30000,
+      "token_acc": 0.9809932238964539
+    },
+    {
+      "epoch": 1.3086150490730644,
+      "eval_loss": 0.0756845697760582,
+      "eval_runtime": 254.1828,
+      "eval_samples_per_second": 58.304,
+      "eval_steps_per_second": 0.913,
+      "eval_token_acc": 0.9734082983099347,
+      "step": 30000
+    },
+    {
+      "epoch": 1.3129770992366412,
+      "grad_norm": 2.8951034545898438,
+      "learning_rate": 1.4469010234821765e-06,
+      "loss": 0.06441830635070801,
+      "step": 30100,
+      "token_acc": 0.9801274671845959
+    },
+    {
+      "epoch": 1.317339149400218,
+      "grad_norm": 4.482010364532471,
+      "learning_rate": 1.4305748966152978e-06,
+      "loss": 0.06479699134826661,
+      "step": 30200,
+      "token_acc": 0.9794524119947848
+    },
+    {
+      "epoch": 1.321701199563795,
+      "grad_norm": 2.2887697219848633,
+      "learning_rate": 1.414304402727574e-06,
+      "loss": 0.0626700496673584,
+      "step": 30300,
+      "token_acc": 0.9796751852488993
+    },
+    {
+      "epoch": 1.3260632497273719,
+      "grad_norm": 2.8628365993499756,
+      "learning_rate": 1.3980903882326985e-06,
+      "loss": 0.06823697090148925,
+      "step": 30400,
+      "token_acc": 0.9791249697232385
+    },
+    {
+      "epoch": 1.3304252998909487,
+      "grad_norm": 3.2624971866607666,
+      "learning_rate": 1.381933696606223e-06,
+      "loss": 0.06213871479034424,
+      "step": 30500,
+      "token_acc": 0.9808548009367681
+    },
+    {
+      "epoch": 1.3347873500545258,
+      "grad_norm": 3.104295492172241,
+      "learning_rate": 1.3658351683416838e-06,
+      "loss": 0.05973214626312256,
+      "step": 30600,
+      "token_acc": 0.9814683008691223
+    },
+    {
+      "epoch": 1.3391494002181026,
+      "grad_norm": 6.34499454498291,
+      "learning_rate": 1.3497956409068824e-06,
+      "loss": 0.06145791053771973,
+      "step": 30700,
+      "token_acc": 0.9810582677574679
+    },
+    {
+      "epoch": 1.3435114503816794,
+      "grad_norm": 4.992619514465332,
+      "learning_rate": 1.333815948700311e-06,
+      "loss": 0.06483103752136231,
+      "step": 30800,
+      "token_acc": 0.9799949403545644
+    },
+    {
+      "epoch": 1.3478735005452562,
+      "grad_norm": 6.377118110656738,
+      "learning_rate": 1.3178969230077515e-06,
+      "loss": 0.06162656784057617,
+      "step": 30900,
+      "token_acc": 0.9806395873967511
+    },
+    {
+      "epoch": 1.352235550708833,
+      "grad_norm": 3.695734739303589,
+      "learning_rate": 1.302039391959031e-06,
+      "loss": 0.06112282276153565,
+      "step": 31000,
+      "token_acc": 0.9815428766118042
+    },
+    {
+      "epoch": 1.352235550708833,
+      "eval_loss": 0.07410430908203125,
+      "eval_runtime": 249.9929,
+      "eval_samples_per_second": 59.282,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.97419353035858,
+      "step": 31000
+    },
+    {
+      "epoch": 1.35659760087241,
+      "grad_norm": 3.6432061195373535,
+      "learning_rate": 1.2862441804849376e-06,
+      "loss": 0.06405759811401367,
+      "step": 31100,
+      "token_acc": 0.980182532705931
+    },
+    {
+      "epoch": 1.360959651035987,
+      "grad_norm": 3.5472171306610107,
+      "learning_rate": 1.270512110274309e-06,
+      "loss": 0.05868791580200195,
+      "step": 31200,
+      "token_acc": 0.9816291636211757
+    },
+    {
+      "epoch": 1.3653217011995638,
+      "grad_norm": 4.953434944152832,
+      "learning_rate": 1.2548439997312883e-06,
+      "loss": 0.05954179763793945,
+      "step": 31300,
+      "token_acc": 0.9819224968854501
+    },
+    {
+      "epoch": 1.3696837513631408,
+      "grad_norm": 4.584559917449951,
+      "learning_rate": 1.2392406639327454e-06,
+      "loss": 0.05943380355834961,
+      "step": 31400,
+      "token_acc": 0.9810650276994911
+    },
+    {
+      "epoch": 1.3740458015267176,
+      "grad_norm": 5.9618425369262695,
+      "learning_rate": 1.223702914585881e-06,
+      "loss": 0.062167482376098634,
+      "step": 31500,
+      "token_acc": 0.9804242388827208
+    },
+    {
+      "epoch": 1.3784078516902945,
+      "grad_norm": 3.6155030727386475,
+      "learning_rate": 1.2082315599859954e-06,
+      "loss": 0.06322122097015381,
+      "step": 31600,
+      "token_acc": 0.9801517767413096
+    },
+    {
+      "epoch": 1.3827699018538713,
+      "grad_norm": 2.845133066177368,
+      "learning_rate": 1.1928274049744406e-06,
+      "loss": 0.05848141193389893,
+      "step": 31700,
+      "token_acc": 0.9813240790872739
+    },
+    {
+      "epoch": 1.387131952017448,
+      "grad_norm": 3.7700917720794678,
+      "learning_rate": 1.177491250896757e-06,
+      "loss": 0.058195791244506835,
+      "step": 31800,
+      "token_acc": 0.9819997313392738
+    },
+    {
+      "epoch": 1.3914940021810251,
+      "grad_norm": 3.283059597015381,
+      "learning_rate": 1.1622238955609785e-06,
+      "loss": 0.06348293304443359,
+      "step": 31900,
+      "token_acc": 0.9804331695834613
+    },
+    {
+      "epoch": 1.395856052344602,
+      "grad_norm": 4.16409969329834,
+      "learning_rate": 1.1470261331961324e-06,
+      "loss": 0.06061270713806152,
+      "step": 32000,
+      "token_acc": 0.9811250377820794
+    },
+    {
+      "epoch": 1.395856052344602,
+      "eval_loss": 0.07234535366296768,
+      "eval_runtime": 249.7003,
+      "eval_samples_per_second": 59.351,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9748586023428418,
+      "step": 32000
+    },
+    {
+      "epoch": 1.4002181025081788,
+      "grad_norm": 5.3219170570373535,
+      "learning_rate": 1.1318987544109269e-06,
+      "loss": 0.06023798942565918,
+      "step": 32100,
+      "token_acc": 0.9810668274906329
+    },
+    {
+      "epoch": 1.4045801526717558,
+      "grad_norm": 4.018277645111084,
+      "learning_rate": 1.116842546152615e-06,
+      "loss": 0.060137758255004885,
+      "step": 32200,
+      "token_acc": 0.9817034659235192
+    },
+    {
+      "epoch": 1.4089422028353327,
+      "grad_norm": 3.68279767036438,
+      "learning_rate": 1.1018582916660605e-06,
+      "loss": 0.05815181255340576,
+      "step": 32300,
+      "token_acc": 0.9821691863093951
+    },
+    {
+      "epoch": 1.4133042529989095,
+      "grad_norm": 3.751368761062622,
+      "learning_rate": 1.0869467704529942e-06,
+      "loss": 0.05569493293762207,
+      "step": 32400,
+      "token_acc": 0.9825266311370547
+    },
+    {
+      "epoch": 1.4176663031624863,
+      "grad_norm": 4.616822719573975,
+      "learning_rate": 1.072108758231459e-06,
+      "loss": 0.05629618167877197,
+      "step": 32500,
+      "token_acc": 0.9823124687557555
+    },
+    {
+      "epoch": 1.4220283533260631,
+      "grad_norm": 3.3701086044311523,
+      "learning_rate": 1.0573450268954577e-06,
+      "loss": 0.05684140205383301,
+      "step": 32600,
+      "token_acc": 0.9825464295056239
+    },
+    {
+      "epoch": 1.4263904034896402,
+      "grad_norm": 2.300793409347534,
+      "learning_rate": 1.042656344474801e-06,
+      "loss": 0.057010469436645506,
+      "step": 32700,
+      "token_acc": 0.9822899964004058
+    },
+    {
+      "epoch": 1.430752453653217,
+      "grad_norm": 4.348355770111084,
+      "learning_rate": 1.0280434750951474e-06,
+      "loss": 0.061251497268676756,
+      "step": 32800,
+      "token_acc": 0.9808127634339351
+    },
+    {
+      "epoch": 1.4351145038167938,
+      "grad_norm": 3.6465044021606445,
+      "learning_rate": 1.0135071789382566e-06,
+      "loss": 0.05879298210144043,
+      "step": 32900,
+      "token_acc": 0.9820062890640164
+    },
+    {
+      "epoch": 1.4394765539803709,
+      "grad_norm": 4.484274864196777,
+      "learning_rate": 9.990482122024458e-07,
+      "loss": 0.055512037277221676,
+      "step": 33000,
+      "token_acc": 0.9827889984324192
+    },
+    {
+      "epoch": 1.4394765539803709,
+      "eval_loss": 0.07081891596317291,
+      "eval_runtime": 249.4133,
+      "eval_samples_per_second": 59.419,
+      "eval_steps_per_second": 0.93,
+      "eval_token_acc": 0.9751492220334437,
+      "step": 33000
+    },
+    {
+      "epoch": 1.4438386041439477,
+      "grad_norm": 3.2281529903411865,
+      "learning_rate": 9.84667327063245e-07,
+      "loss": 0.05788352012634277,
+      "step": 33100,
+      "token_acc": 0.9818554954720177
+    },
+    {
+      "epoch": 1.4482006543075245,
+      "grad_norm": 5.004483222961426,
+      "learning_rate": 9.70365271634271e-07,
+      "loss": 0.0612343692779541,
+      "step": 33200,
+      "token_acc": 0.9810140790866233
+    },
+    {
+      "epoch": 1.4525627044711014,
+      "grad_norm": 3.850135087966919,
+      "learning_rate": 9.561427899283125e-07,
+      "loss": 0.05688544273376465,
+      "step": 33300,
+      "token_acc": 0.9823226364637407
+    },
+    {
+      "epoch": 1.4569247546346782,
+      "grad_norm": 3.7914345264434814,
+      "learning_rate": 9.420006218186198e-07,
+      "loss": 0.057846450805664064,
+      "step": 33400,
+      "token_acc": 0.9820036576349888
+    },
+    {
+      "epoch": 1.4612868047982552,
+      "grad_norm": 2.471034526824951,
+      "learning_rate": 9.27939503000419e-07,
+      "loss": 0.05780147552490234,
+      "step": 33500,
+      "token_acc": 0.9819559772443386
+    },
+    {
+      "epoch": 1.465648854961832,
+      "grad_norm": 4.725645542144775,
+      "learning_rate": 9.139601649526416e-07,
+      "loss": 0.059658441543579105,
+      "step": 33600,
+      "token_acc": 0.9816973978417965
+    },
+    {
+      "epoch": 1.4700109051254089,
+      "grad_norm": 3.542151927947998,
+      "learning_rate": 9.000633348998669e-07,
+      "loss": 0.05712653636932373,
+      "step": 33700,
+      "token_acc": 0.9819699628907641
+    },
+    {
+      "epoch": 1.474372955288986,
+      "grad_norm": 4.53084135055542,
+      "learning_rate": 8.862497357744945e-07,
+      "loss": 0.05675581455230713,
+      "step": 33800,
+      "token_acc": 0.9820234465526059
+    },
+    {
+      "epoch": 1.4787350054525628,
+      "grad_norm": 4.158478736877441,
+      "learning_rate": 8.725200861791378e-07,
+      "loss": 0.05835529804229736,
+      "step": 33900,
+      "token_acc": 0.9820663243996941
+    },
+    {
+      "epoch": 1.4830970556161396,
+      "grad_norm": 5.382080078125,
+      "learning_rate": 8.58875100349236e-07,
+      "loss": 0.05506091117858887,
+      "step": 34000,
+      "token_acc": 0.9825372133740343
+    },
+    {
+      "epoch": 1.4830970556161396,
+      "eval_loss": 0.06877533346414566,
+      "eval_runtime": 254.8617,
+      "eval_samples_per_second": 58.149,
+      "eval_steps_per_second": 0.91,
+      "eval_token_acc": 0.9760518197263703,
+      "step": 34000
+    },
+    {
+      "epoch": 1.4874591057797164,
+      "grad_norm": 4.1885528564453125,
+      "learning_rate": 8.453154881159023e-07,
+      "loss": 0.06418917655944824,
+      "step": 34100,
+      "token_acc": 0.9805802465917681
+    },
+    {
+      "epoch": 1.4918211559432932,
+      "grad_norm": 3.279444932937622,
+      "learning_rate": 8.318419548689993e-07,
+      "loss": 0.05651405334472656,
+      "step": 34200,
+      "token_acc": 0.9823017340079185
+    },
+    {
+      "epoch": 1.4961832061068703,
+      "grad_norm": 3.4269838333129883,
+      "learning_rate": 8.184552015204383e-07,
+      "loss": 0.05106703758239746,
+      "step": 34300,
+      "token_acc": 0.9836486664038891
+    },
+    {
+      "epoch": 1.500545256270447,
+      "grad_norm": 3.865947723388672,
+      "learning_rate": 8.051559244677199e-07,
+      "loss": 0.0522431755065918,
+      "step": 34400,
+      "token_acc": 0.9838395317031918
+    },
+    {
+      "epoch": 1.5049073064340242,
+      "grad_norm": 6.236225605010986,
+      "learning_rate": 7.919448155577089e-07,
+      "loss": 0.05624629020690918,
+      "step": 34500,
+      "token_acc": 0.9829597809300092
+    },
+    {
+      "epoch": 1.509269356597601,
+      "grad_norm": 6.7572431564331055,
+      "learning_rate": 7.788225620506384e-07,
+      "loss": 0.054970383644104004,
+      "step": 34600,
+      "token_acc": 0.9824237722392088
+    },
+    {
+      "epoch": 1.5136314067611778,
+      "grad_norm": 4.074737071990967,
+      "learning_rate": 7.657898465843599e-07,
+      "loss": 0.054400978088378904,
+      "step": 34700,
+      "token_acc": 0.9830726460195386
+    },
+    {
+      "epoch": 1.5179934569247546,
+      "grad_norm": 4.249269485473633,
+      "learning_rate": 7.528473471388342e-07,
+      "loss": 0.05802037239074707,
+      "step": 34800,
+      "token_acc": 0.9820589628377946
+    },
+    {
+      "epoch": 1.5223555070883314,
+      "grad_norm": 4.001960277557373,
+      "learning_rate": 7.399957370008568e-07,
+      "loss": 0.0512929630279541,
+      "step": 34900,
+      "token_acc": 0.9846466805286435
+    },
+    {
+      "epoch": 1.5267175572519083,
+      "grad_norm": 4.0334153175354,
+      "learning_rate": 7.272356847290357e-07,
+      "loss": 0.054765453338623045,
+      "step": 35000,
+      "token_acc": 0.9834754901641374
+    },
+    {
+      "epoch": 1.5267175572519083,
+      "eval_loss": 0.06686375290155411,
+      "eval_runtime": 249.712,
+      "eval_samples_per_second": 59.348,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9764514218009479,
+      "step": 35000
+    },
+    {
+      "epoch": 1.5310796074154853,
+      "grad_norm": 3.1412196159362793,
+      "learning_rate": 7.145678541190132e-07,
+      "loss": 0.05724092483520508,
+      "step": 35100,
+      "token_acc": 0.982305148733112
+    },
+    {
+      "epoch": 1.5354416575790621,
+      "grad_norm": 4.317452430725098,
+      "learning_rate": 7.019929041689308e-07,
+      "loss": 0.056807117462158205,
+      "step": 35200,
+      "token_acc": 0.9833986233324803
+    },
+    {
+      "epoch": 1.5398037077426392,
+      "grad_norm": 2.5507566928863525,
+      "learning_rate": 6.895114890451487e-07,
+      "loss": 0.05457932472229004,
+      "step": 35300,
+      "token_acc": 0.9831009055880454
+    },
+    {
+      "epoch": 1.544165757906216,
+      "grad_norm": 3.851468563079834,
+      "learning_rate": 6.771242580482188e-07,
+      "loss": 0.055332069396972654,
+      "step": 35400,
+      "token_acc": 0.9831133615205685
+    },
+    {
+      "epoch": 1.5485278080697928,
+      "grad_norm": 5.143461227416992,
+      "learning_rate": 6.648318555791e-07,
+      "loss": 0.05229470252990723,
+      "step": 35500,
+      "token_acc": 0.9841282328645885
+    },
+    {
+      "epoch": 1.5528898582333697,
+      "grad_norm": 3.2583236694335938,
+      "learning_rate": 6.526349211056409e-07,
+      "loss": 0.05265366554260254,
+      "step": 35600,
+      "token_acc": 0.9836261310683998
+    },
+    {
+      "epoch": 1.5572519083969465,
+      "grad_norm": 3.030510187149048,
+      "learning_rate": 6.405340891293143e-07,
+      "loss": 0.05120317459106445,
+      "step": 35700,
+      "token_acc": 0.9843944043261996
+    },
+    {
+      "epoch": 1.5616139585605233,
+      "grad_norm": 3.0830142498016357,
+      "learning_rate": 6.285299891522048e-07,
+      "loss": 0.059448165893554686,
+      "step": 35800,
+      "token_acc": 0.9815313972588839
+    },
+    {
+      "epoch": 1.5659760087241004,
+      "grad_norm": 3.755048990249634,
+      "learning_rate": 6.166232456442645e-07,
+      "loss": 0.05295629024505615,
+      "step": 35900,
+      "token_acc": 0.9834888356480129
+    },
+    {
+      "epoch": 1.5703380588876772,
+      "grad_norm": 4.288339138031006,
+      "learning_rate": 6.048144780108289e-07,
+      "loss": 0.05355148792266846,
+      "step": 36000,
+      "token_acc": 0.9836604652353047
+    },
+    {
+      "epoch": 1.5703380588876772,
+      "eval_loss": 0.06514008343219757,
+      "eval_runtime": 249.5354,
+      "eval_samples_per_second": 59.39,
+      "eval_steps_per_second": 0.93,
+      "eval_token_acc": 0.9770801663238845,
+      "step": 36000
+    },
+    {
+      "epoch": 1.5747001090512542,
+      "grad_norm": 7.691909313201904,
+      "learning_rate": 5.931043005603907e-07,
+      "loss": 0.053992023468017576,
+      "step": 36100,
+      "token_acc": 0.9830635714701752
+    },
+    {
+      "epoch": 1.579062159214831,
+      "grad_norm": 3.0412437915802,
+      "learning_rate": 5.814933224726432e-07,
+      "loss": 0.052196903228759764,
+      "step": 36200,
+      "token_acc": 0.9839295311250639
+    },
+    {
+      "epoch": 1.5834242093784079,
+      "grad_norm": 4.929081916809082,
+      "learning_rate": 5.699821477667947e-07,
+      "loss": 0.05482297897338867,
+      "step": 36300,
+      "token_acc": 0.9832719951365237
+    },
+    {
+      "epoch": 1.5877862595419847,
+      "grad_norm": 4.209242820739746,
+      "learning_rate": 5.585713752701394e-07,
+      "loss": 0.05226144790649414,
+      "step": 36400,
+      "token_acc": 0.9841242428312833
+    },
+    {
+      "epoch": 1.5921483097055615,
+      "grad_norm": 2.3831539154052734,
+      "learning_rate": 5.472615985869104e-07,
+      "loss": 0.0528890323638916,
+      "step": 36500,
+      "token_acc": 0.9840980187695516
+    },
+    {
+      "epoch": 1.5965103598691384,
+      "grad_norm": 3.2346951961517334,
+      "learning_rate": 5.360534060673994e-07,
+      "loss": 0.05202345371246338,
+      "step": 36600,
+      "token_acc": 0.9840068698790669
+    },
+    {
+      "epoch": 1.6008724100327154,
+      "grad_norm": 2.5527503490448,
+      "learning_rate": 5.249473807773472e-07,
+      "loss": 0.05285400390625,
+      "step": 36700,
+      "token_acc": 0.9839257650748936
+    },
+    {
+      "epoch": 1.6052344601962922,
+      "grad_norm": 1.54398775100708,
+      "learning_rate": 5.139441004676138e-07,
+      "loss": 0.05471937656402588,
+      "step": 36800,
+      "token_acc": 0.9833169259040494
+    },
+    {
+      "epoch": 1.6095965103598693,
+      "grad_norm": 3.731952667236328,
+      "learning_rate": 5.030441375441239e-07,
+      "loss": 0.05202892303466797,
+      "step": 36900,
+      "token_acc": 0.9836451150625598
+    },
+    {
+      "epoch": 1.613958560523446,
+      "grad_norm": 5.551258563995361,
+      "learning_rate": 4.922480590380865e-07,
+      "loss": 0.054112329483032226,
+      "step": 37000,
+      "token_acc": 0.9834300729468978
+    },
+    {
+      "epoch": 1.613958560523446,
+      "eval_loss": 0.06393582373857498,
+      "eval_runtime": 249.8278,
+      "eval_samples_per_second": 59.321,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9774182911562193,
+      "step": 37000
+    },
+    {
+      "epoch": 1.618320610687023,
+      "grad_norm": 3.7928690910339355,
+      "learning_rate": 4.815564265764994e-07,
+      "loss": 0.05264884948730469,
+      "step": 37100,
+      "token_acc": 0.9840745348053698
+    },
+    {
+      "epoch": 1.6226826608505998,
+      "grad_norm": 5.02518367767334,
+      "learning_rate": 4.709697963529333e-07,
+      "loss": 0.056769208908081056,
+      "step": 37200,
+      "token_acc": 0.9824920285478644
+    },
+    {
+      "epoch": 1.6270447110141766,
+      "grad_norm": 3.828500747680664,
+      "learning_rate": 4.604887190985957e-07,
+      "loss": 0.055812931060791014,
+      "step": 37300,
+      "token_acc": 0.9835446977359371
+    },
+    {
+      "epoch": 1.6314067611777534,
+      "grad_norm": 2.125394821166992,
+      "learning_rate": 4.5011374005368147e-07,
+      "loss": 0.051757588386535644,
+      "step": 37400,
+      "token_acc": 0.9842062000939408
+    },
+    {
+      "epoch": 1.6357688113413305,
+      "grad_norm": 2.8596160411834717,
+      "learning_rate": 4.3984539893900976e-07,
+      "loss": 0.0530195140838623,
+      "step": 37500,
+      "token_acc": 0.9834978568645278
+    },
+    {
+      "epoch": 1.6401308615049073,
+      "grad_norm": 3.184962034225464,
+      "learning_rate": 4.296842299279469e-07,
+      "loss": 0.05626538276672363,
+      "step": 37600,
+      "token_acc": 0.9829366662234632
+    },
+    {
+      "epoch": 1.6444929116684843,
+      "grad_norm": 2.6302707195281982,
+      "learning_rate": 4.196307616186185e-07,
+      "loss": 0.053105955123901365,
+      "step": 37700,
+      "token_acc": 0.9834460881105752
+    },
+    {
+      "epoch": 1.6488549618320612,
+      "grad_norm": 3.290652275085449,
+      "learning_rate": 4.0968551700640654e-07,
+      "loss": 0.05122368812561035,
+      "step": 37800,
+      "token_acc": 0.9839382289698312
+    },
+    {
+      "epoch": 1.653217011995638,
+      "grad_norm": 6.050076484680176,
+      "learning_rate": 3.998490134567498e-07,
+      "loss": 0.04652623176574707,
+      "step": 37900,
+      "token_acc": 0.9856113210284411
+    },
+    {
+      "epoch": 1.6575790621592148,
+      "grad_norm": 4.384199142456055,
+      "learning_rate": 3.90121762678225e-07,
+      "loss": 0.05324809551239014,
+      "step": 38000,
+      "token_acc": 0.9837580092385636
+    },
+    {
+      "epoch": 1.6575790621592148,
+      "eval_loss": 0.06297693401575089,
+      "eval_runtime": 249.685,
+      "eval_samples_per_second": 59.355,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9780330635786462,
+      "step": 38000
+    },
+    {
+      "epoch": 1.6619411123227916,
+      "grad_norm": 4.2886433601379395,
+      "learning_rate": 3.8050427069592526e-07,
+      "loss": 0.04979161262512207,
+      "step": 38100,
+      "token_acc": 0.9845926138438266
+    },
+    {
+      "epoch": 1.6663031624863684,
+      "grad_norm": 3.2638423442840576,
+      "learning_rate": 3.7099703782514236e-07,
+      "loss": 0.05182206630706787,
+      "step": 38200,
+      "token_acc": 0.9839714186037497
+    },
+    {
+      "epoch": 1.6706652126499455,
+      "grad_norm": 3.262000799179077,
+      "learning_rate": 3.6160055864533526e-07,
+      "loss": 0.05675747871398926,
+      "step": 38300,
+      "token_acc": 0.9826982452279391
+    },
+    {
+      "epoch": 1.6750272628135223,
+      "grad_norm": 3.224867820739746,
+      "learning_rate": 3.5231532197439884e-07,
+      "loss": 0.0510973072052002,
+      "step": 38400,
+      "token_acc": 0.9846604806212441
+    },
+    {
+      "epoch": 1.6793893129770994,
+      "grad_norm": 7.291615009307861,
+      "learning_rate": 3.4314181084324215e-07,
+      "loss": 0.04996752738952637,
+      "step": 38500,
+      "token_acc": 0.9848244576449638
+    },
+    {
+      "epoch": 1.6837513631406762,
+      "grad_norm": 3.04451060295105,
+      "learning_rate": 3.340805024706559e-07,
+      "loss": 0.0533557939529419,
+      "step": 38600,
+      "token_acc": 0.9838469152943294
+    },
+    {
+      "epoch": 1.688113413304253,
+      "grad_norm": 3.626009702682495,
+      "learning_rate": 3.251318682384838e-07,
+      "loss": 0.05408650398254394,
+      "step": 38700,
+      "token_acc": 0.983619685023991
+    },
+    {
+      "epoch": 1.6924754634678298,
+      "grad_norm": 7.859470367431641,
+      "learning_rate": 3.1629637366710877e-07,
+      "loss": 0.051053762435913086,
+      "step": 38800,
+      "token_acc": 0.9844172007416103
+    },
+    {
+      "epoch": 1.6968375136314067,
+      "grad_norm": 1.7803997993469238,
+      "learning_rate": 3.075744783912304e-07,
+      "loss": 0.0526139497756958,
+      "step": 38900,
+      "token_acc": 0.9835836512403247
+    },
+    {
+      "epoch": 1.7011995637949835,
+      "grad_norm": 2.8114235401153564,
+      "learning_rate": 2.989666361359525e-07,
+      "loss": 0.05366787433624268,
+      "step": 39000,
+      "token_acc": 0.9837218445616883
+    },
+    {
+      "epoch": 1.7011995637949835,
+      "eval_loss": 0.061601582914590836,
+      "eval_runtime": 255.8728,
+      "eval_samples_per_second": 57.919,
+      "eval_steps_per_second": 0.907,
+      "eval_token_acc": 0.9786226862201556,
+      "step": 39000
+    },
+    {
+      "epoch": 1.7055616139585605,
+      "grad_norm": 10.10936450958252,
+      "learning_rate": 2.9047329469318515e-07,
+      "loss": 0.05378860473632813,
+      "step": 39100,
+      "token_acc": 0.9834365733325614
+    },
+    {
+      "epoch": 1.7099236641221374,
+      "grad_norm": 9.03071403503418,
+      "learning_rate": 2.8209489589834713e-07,
+      "loss": 0.050548648834228514,
+      "step": 39200,
+      "token_acc": 0.9842391860108396
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 2.1598613262176514,
+      "learning_rate": 2.7383187560737825e-07,
+      "loss": 0.05034954071044922,
+      "step": 39300,
+      "token_acc": 0.9845266314810692
+    },
+    {
+      "epoch": 1.7186477644492912,
+      "grad_norm": 4.462788105010986,
+      "learning_rate": 2.656846636740712e-07,
+      "loss": 0.051555233001708986,
+      "step": 39400,
+      "token_acc": 0.9836437894791176
+    },
+    {
+      "epoch": 1.723009814612868,
+      "grad_norm": 3.6997790336608887,
+      "learning_rate": 2.576536839277069e-07,
+      "loss": 0.05385732650756836,
+      "step": 39500,
+      "token_acc": 0.9834032885363317
+    },
+    {
+      "epoch": 1.7273718647764449,
+      "grad_norm": 4.76052713394165,
+      "learning_rate": 2.497393541510046e-07,
+      "loss": 0.05027077674865723,
+      "step": 39600,
+      "token_acc": 0.9847386493704693
+    },
+    {
+      "epoch": 1.7317339149400217,
+      "grad_norm": 5.727189064025879,
+      "learning_rate": 2.4194208605839096e-07,
+      "loss": 0.04907273292541504,
+      "step": 39700,
+      "token_acc": 0.9850905973429936
+    },
+    {
+      "epoch": 1.7360959651035985,
+      "grad_norm": 3.36122989654541,
+      "learning_rate": 2.342622852745824e-07,
+      "loss": 0.04743415832519531,
+      "step": 39800,
+      "token_acc": 0.9850597151289282
+    },
+    {
+      "epoch": 1.7404580152671756,
+      "grad_norm": 3.0961196422576904,
+      "learning_rate": 2.2670035131348135e-07,
+      "loss": 0.04388811111450195,
+      "step": 39900,
+      "token_acc": 0.9863217267251401
+    },
+    {
+      "epoch": 1.7448200654307524,
+      "grad_norm": 2.886570692062378,
+      "learning_rate": 2.1925667755739427e-07,
+      "loss": 0.050310606956481936,
+      "step": 40000,
+      "token_acc": 0.9843371271999157
+    },
+    {
+      "epoch": 1.7448200654307524,
+      "eval_loss": 0.060765430331230164,
+      "eval_runtime": 250.6299,
+      "eval_samples_per_second": 59.131,
+      "eval_steps_per_second": 0.926,
+      "eval_token_acc": 0.978932866851471,
+      "step": 40000
+    },
+    {
+      "epoch": 1.7491821155943295,
+      "grad_norm": 3.1743409633636475,
+      "learning_rate": 2.119316512365696e-07,
+      "loss": 0.04878216743469238,
+      "step": 40100,
+      "token_acc": 0.9853699903545512
+    },
+    {
+      "epoch": 1.7535441657579063,
+      "grad_norm": 2.7623131275177,
+      "learning_rate": 2.0472565340904887e-07,
+      "loss": 0.05069635391235352,
+      "step": 40200,
+      "token_acc": 0.9844571383958778
+    },
+    {
+      "epoch": 1.757906215921483,
+      "grad_norm": 4.026466369628906,
+      "learning_rate": 1.9763905894084723e-07,
+      "loss": 0.052156386375427244,
+      "step": 40300,
+      "token_acc": 0.9838101347017318
+    },
+    {
+      "epoch": 1.76226826608506,
+      "grad_norm": 2.6961493492126465,
+      "learning_rate": 1.9067223648645188e-07,
+      "loss": 0.04677755355834961,
+      "step": 40400,
+      "token_acc": 0.9858562442207259
+    },
+    {
+      "epoch": 1.7666303162486368,
+      "grad_norm": 3.7131080627441406,
+      "learning_rate": 1.8382554846964218e-07,
+      "loss": 0.049273710250854495,
+      "step": 40500,
+      "token_acc": 0.985027988651177
+    },
+    {
+      "epoch": 1.7709923664122136,
+      "grad_norm": 1.813375473022461,
+      "learning_rate": 1.7709935106463792e-07,
+      "loss": 0.05270167350769043,
+      "step": 40600,
+      "token_acc": 0.9837776105069173
+    },
+    {
+      "epoch": 1.7753544165757906,
+      "grad_norm": 4.872157096862793,
+      "learning_rate": 1.7049399417757018e-07,
+      "loss": 0.04908137321472168,
+      "step": 40700,
+      "token_acc": 0.9845669230221923
+    },
+    {
+      "epoch": 1.7797164667393675,
+      "grad_norm": 3.0371429920196533,
+      "learning_rate": 1.6400982142827787e-07,
+      "loss": 0.05342394828796387,
+      "step": 40800,
+      "token_acc": 0.9832475559923447
+    },
+    {
+      "epoch": 1.7840785169029445,
+      "grad_norm": 2.185772657394409,
+      "learning_rate": 1.576471701324331e-07,
+      "loss": 0.04955248832702637,
+      "step": 40900,
+      "token_acc": 0.9850666460511651
+    },
+    {
+      "epoch": 1.7884405670665213,
+      "grad_norm": 3.0958948135375977,
+      "learning_rate": 1.514063712839925e-07,
+      "loss": 0.04495782852172851,
+      "step": 41000,
+      "token_acc": 0.9864484417348619
+    },
+    {
+      "epoch": 1.7884405670665213,
+      "eval_loss": 0.060102950781583786,
+      "eval_runtime": 250.0116,
+      "eval_samples_per_second": 59.277,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.9792290753822767,
+      "step": 41000
+    },
+    {
+      "epoch": 1.7928026172300982,
+      "grad_norm": 2.581108331680298,
+      "learning_rate": 1.4528774953798092e-07,
+      "loss": 0.049694914817810056,
+      "step": 41100,
+      "token_acc": 0.9852027649171952
+    },
+    {
+      "epoch": 1.797164667393675,
+      "grad_norm": 4.794500827789307,
+      "learning_rate": 1.3929162319359868e-07,
+      "loss": 0.052889041900634766,
+      "step": 41200,
+      "token_acc": 0.9838971151391371
+    },
+    {
+      "epoch": 1.8015267175572518,
+      "grad_norm": 3.068047523498535,
+      "learning_rate": 1.3341830417766572e-07,
+      "loss": 0.04843362331390381,
+      "step": 41300,
+      "token_acc": 0.9855183429921717
+    },
+    {
+      "epoch": 1.8058887677208286,
+      "grad_norm": 3.2265470027923584,
+      "learning_rate": 1.2766809802839536e-07,
+      "loss": 0.05105172157287598,
+      "step": 41400,
+      "token_acc": 0.9843682767213199
+    },
+    {
+      "epoch": 1.8102508178844057,
+      "grad_norm": 2.618612051010132,
+      "learning_rate": 1.2204130387949735e-07,
+      "loss": 0.04559971809387207,
+      "step": 41500,
+      "token_acc": 0.9859196891191709
+    },
+    {
+      "epoch": 1.8146128680479825,
+      "grad_norm": 2.534839153289795,
+      "learning_rate": 1.1653821444461755e-07,
+      "loss": 0.048215513229370115,
+      "step": 41600,
+      "token_acc": 0.9848556340531701
+    },
+    {
+      "epoch": 1.8189749182115595,
+      "grad_norm": 2.000978708267212,
+      "learning_rate": 1.1115911600211194e-07,
+      "loss": 0.04903024673461914,
+      "step": 41700,
+      "token_acc": 0.9847122828227678
+    },
+    {
+      "epoch": 1.8233369683751364,
+      "grad_norm": 3.0504517555236816,
+      "learning_rate": 1.0590428838015255e-07,
+      "loss": 0.05070067405700684,
+      "step": 41800,
+      "token_acc": 0.9843092181728956
+    },
+    {
+      "epoch": 1.8276990185387132,
+      "grad_norm": 3.5902857780456543,
+      "learning_rate": 1.007740049421696e-07,
+      "loss": 0.04867252349853515,
+      "step": 41900,
+      "token_acc": 0.985157748180839
+    },
+    {
+      "epoch": 1.83206106870229,
+      "grad_norm": 3.775128126144409,
+      "learning_rate": 9.576853257263374e-08,
+      "loss": 0.04645414352416992,
+      "step": 42000,
+      "token_acc": 0.9858936035130417
+    },
+    {
+      "epoch": 1.83206106870229,
+      "eval_loss": 0.05955477058887482,
+      "eval_runtime": 253.7866,
+      "eval_samples_per_second": 58.396,
+      "eval_steps_per_second": 0.914,
+      "eval_token_acc": 0.9794414513100241,
+      "step": 42000
+    },
+    {
+      "epoch": 1.8364231188658668,
+      "grad_norm": 4.869281768798828,
+      "learning_rate": 9.088813166316879e-08,
+      "loss": 0.04662482261657715,
+      "step": 42100,
+      "token_acc": 0.985914570659612
+    },
+    {
+      "epoch": 1.840785169029444,
+      "grad_norm": 2.82537579536438,
+      "learning_rate": 8.613305609900841e-08,
+      "loss": 0.04755397796630859,
+      "step": 42200,
+      "token_acc": 0.9857374328685857
+    },
+    {
+      "epoch": 1.8451472191930207,
+      "grad_norm": 3.2263779640197754,
+      "learning_rate": 8.150355324578807e-08,
+      "loss": 0.04893510818481445,
+      "step": 42300,
+      "token_acc": 0.9853583504779954
+    },
+    {
+      "epoch": 1.8495092693565978,
+      "grad_norm": 3.73339581489563,
+      "learning_rate": 7.699986393667535e-08,
+      "loss": 0.04965659618377685,
+      "step": 42400,
+      "token_acc": 0.9852581619444843
+    },
+    {
+      "epoch": 1.8538713195201746,
+      "grad_norm": 5.45037317276001,
+      "learning_rate": 7.262222245984329e-08,
+      "loss": 0.05175324916839599,
+      "step": 42500,
+      "token_acc": 0.9840829382972983
+    },
+    {
+      "epoch": 1.8582333696837514,
+      "grad_norm": 2.7033138275146484,
+      "learning_rate": 6.83708565462815e-08,
+      "loss": 0.05096693515777588,
+      "step": 42600,
+      "token_acc": 0.9843310384171561
+    },
+    {
+      "epoch": 1.8625954198473282,
+      "grad_norm": 2.480644464492798,
+      "learning_rate": 6.424598735794929e-08,
+      "loss": 0.05035747051239014,
+      "step": 42700,
+      "token_acc": 0.9847126296208089
+    },
+    {
+      "epoch": 1.866957470010905,
+      "grad_norm": 6.001003265380859,
+      "learning_rate": 6.024782947627039e-08,
+      "loss": 0.05004034996032715,
+      "step": 42800,
+      "token_acc": 0.984943015304461
+    },
+    {
+      "epoch": 1.8713195201744819,
+      "grad_norm": 2.7746217250823975,
+      "learning_rate": 5.6376590890971085e-08,
+      "loss": 0.05255108833312988,
+      "step": 42900,
+      "token_acc": 0.9837951328137426
+    },
+    {
+      "epoch": 1.875681570338059,
+      "grad_norm": 3.266484022140503,
+      "learning_rate": 5.2632472989259384e-08,
+      "loss": 0.0456111478805542,
+      "step": 43000,
+      "token_acc": 0.9861970334118388
+    },
+    {
+      "epoch": 1.875681570338059,
+      "eval_loss": 0.05910694971680641,
+      "eval_runtime": 249.7817,
+      "eval_samples_per_second": 59.332,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9795839667352231,
+      "step": 43000
+    },
+    {
+      "epoch": 1.8800436205016358,
+      "grad_norm": 2.7523422241210938,
+      "learning_rate": 4.901567054534761e-08,
+      "loss": 0.046754164695739744,
+      "step": 43100,
+      "token_acc": 0.985937347833144
+    },
+    {
+      "epoch": 1.8844056706652128,
+      "grad_norm": 4.6909003257751465,
+      "learning_rate": 4.552637171032326e-08,
+      "loss": 0.04618264198303223,
+      "step": 43200,
+      "token_acc": 0.9854022284965875
+    },
+    {
+      "epoch": 1.8887677208287896,
+      "grad_norm": 8.2206392288208,
+      "learning_rate": 4.216475800235736e-08,
+      "loss": 0.0508054256439209,
+      "step": 43300,
+      "token_acc": 0.9843883895493057
+    },
+    {
+      "epoch": 1.8931297709923665,
+      "grad_norm": 3.2640438079833984,
+      "learning_rate": 3.8931004297263765e-08,
+      "loss": 0.04786904811859131,
+      "step": 43400,
+      "token_acc": 0.985393875230365
+    },
+    {
+      "epoch": 1.8974918211559433,
+      "grad_norm": 1.9893529415130615,
+      "learning_rate": 3.5825278819401344e-08,
+      "loss": 0.047113409042358396,
+      "step": 43500,
+      "token_acc": 0.9858144453728849
+    },
+    {
+      "epoch": 1.90185387131952,
+      "grad_norm": 3.4239087104797363,
+      "learning_rate": 3.284774313292327e-08,
+      "loss": 0.04778111457824707,
+      "step": 43600,
+      "token_acc": 0.9853008490693761
+    },
+    {
+      "epoch": 1.906215921483097,
+      "grad_norm": 7.985106945037842,
+      "learning_rate": 2.999855213337094e-08,
+      "loss": 0.047200527191162106,
+      "step": 43700,
+      "token_acc": 0.9858358995324792
+    },
+    {
+      "epoch": 1.910577971646674,
+      "grad_norm": 2.5651211738586426,
+      "learning_rate": 2.7277854039617356e-08,
+      "loss": 0.047189741134643554,
+      "step": 43800,
+      "token_acc": 0.9852690961552147
+    },
+    {
+      "epoch": 1.9149400218102508,
+      "grad_norm": 4.144567489624023,
+      "learning_rate": 2.4685790386156093e-08,
+      "loss": 0.048973798751831055,
+      "step": 43900,
+      "token_acc": 0.9847606160190734
+    },
+    {
+      "epoch": 1.9193020719738279,
+      "grad_norm": 2.7335407733917236,
+      "learning_rate": 2.222249601573828e-08,
+      "loss": 0.045456886291503906,
+      "step": 44000,
+      "token_acc": 0.986095739252033
+    },
+    {
+      "epoch": 1.9193020719738279,
+      "eval_loss": 0.05891130864620209,
+      "eval_runtime": 249.6153,
+      "eval_samples_per_second": 59.371,
+      "eval_steps_per_second": 0.929,
+      "eval_token_acc": 0.9796035276759367,
+      "step": 44000
+    },
+    {
+      "epoch": 1.9236641221374047,
+      "grad_norm": 2.71073317527771,
+      "learning_rate": 1.9888099072358502e-08,
+      "loss": 0.04772358894348144,
+      "step": 44100,
+      "token_acc": 0.9852039956985347
+    },
+    {
+      "epoch": 1.9280261723009815,
+      "grad_norm": 5.084836006164551,
+      "learning_rate": 1.76827209945879e-08,
+      "loss": 0.049145450592041014,
+      "step": 44200,
+      "token_acc": 0.985264099224324
+    },
+    {
+      "epoch": 1.9323882224645583,
+      "grad_norm": 2.6878726482391357,
+      "learning_rate": 1.560647650925645e-08,
+      "loss": 0.04851144790649414,
+      "step": 44300,
+      "token_acc": 0.9848727188638159
+    },
+    {
+      "epoch": 1.9367502726281351,
+      "grad_norm": 3.3305675983428955,
+      "learning_rate": 1.3659473625486641e-08,
+      "loss": 0.05027200222015381,
+      "step": 44400,
+      "token_acc": 0.9853740048258687
+    },
+    {
+      "epoch": 1.941112322791712,
+      "grad_norm": 3.962172508239746,
+      "learning_rate": 1.1841813629072108e-08,
+      "loss": 0.04904890060424805,
+      "step": 44500,
+      "token_acc": 0.984241933059261
+    },
+    {
+      "epoch": 1.945474372955289,
+      "grad_norm": 6.480011463165283,
+      "learning_rate": 1.0153591077210479e-08,
+      "loss": 0.044611949920654294,
+      "step": 44600,
+      "token_acc": 0.986387756168967
+    },
+    {
+      "epoch": 1.9498364231188658,
+      "grad_norm": 3.123936414718628,
+      "learning_rate": 8.594893793583692e-09,
+      "loss": 0.04712478637695312,
+      "step": 44700,
+      "token_acc": 0.9857043601373849
+    },
+    {
+      "epoch": 1.954198473282443,
+      "grad_norm": 3.129568099975586,
+      "learning_rate": 7.165802863789151e-09,
+      "loss": 0.047671728134155274,
+      "step": 44800,
+      "token_acc": 0.9856617238607475
+    },
+    {
+      "epoch": 1.9585605234460197,
+      "grad_norm": 4.54371452331543,
+      "learning_rate": 5.866392631121709e-09,
+      "loss": 0.04312769412994385,
+      "step": 44900,
+      "token_acc": 0.9869960142783764
+    },
+    {
+      "epoch": 1.9629225736095965,
+      "grad_norm": 6.765581130981445,
+      "learning_rate": 4.6967306927067634e-09,
+      "loss": 0.04882259845733643,
+      "step": 45000,
+      "token_acc": 0.9849781376518219
+    },
+    {
+      "epoch": 1.9629225736095965,
+      "eval_loss": 0.05884711444377899,
+      "eval_runtime": 254.1874,
+      "eval_samples_per_second": 58.303,
+      "eval_steps_per_second": 0.913,
+      "eval_token_acc": 0.9796482383975678,
+      "step": 45000
+    },
+    {
+      "epoch": 1.9672846237731734,
+      "grad_norm": 3.8652217388153076,
+      "learning_rate": 3.6568778959825137e-09,
+      "loss": 0.04858424186706543,
+      "step": 45100,
+      "token_acc": 0.9854189208567443
+    },
+    {
+      "epoch": 1.9716466739367502,
+      "grad_norm": 4.594508171081543,
+      "learning_rate": 2.746888335535547e-09,
+      "loss": 0.045993332862854,
+      "step": 45200,
+      "token_acc": 0.985808285313814
+    },
+    {
+      "epoch": 1.976008724100327,
+      "grad_norm": 6.799201965332031,
+      "learning_rate": 1.96680935028698e-09,
+      "loss": 0.04716442108154297,
+      "step": 45300,
+      "token_acc": 0.9854687663492728
+    },
+    {
+      "epoch": 1.980370774263904,
+      "grad_norm": 3.836419105529785,
+      "learning_rate": 1.316681521028873e-09,
+      "loss": 0.04726198196411133,
+      "step": 45400,
+      "token_acc": 0.9852152973770747
+    },
+    {
+      "epoch": 1.984732824427481,
+      "grad_norm": 6.797372341156006,
+      "learning_rate": 7.965386683139731e-10,
+      "loss": 0.050753302574157715,
+      "step": 45500,
+      "token_acc": 0.984968922406266
+    },
+    {
+      "epoch": 1.989094874591058,
+      "grad_norm": 3.1903374195098877,
+      "learning_rate": 4.0640785069545606e-10,
+      "loss": 0.04919458389282227,
+      "step": 45600,
+      "token_acc": 0.9844628013569153
+    },
+    {
+      "epoch": 1.9934569247546348,
+      "grad_norm": 2.859959363937378,
+      "learning_rate": 1.4630936332082902e-10,
+      "loss": 0.04802582263946533,
+      "step": 45700,
+      "token_acc": 0.9855202858400435
+    },
+    {
+      "epoch": 1.9978189749182116,
+      "grad_norm": 5.337630748748779,
+      "learning_rate": 1.6256736874442624e-11,
+      "loss": 0.04464107036590576,
+      "step": 45800,
+      "token_acc": 0.9865233937765838
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.05883299186825752,
+      "eval_runtime": 250.142,
+      "eval_samples_per_second": 59.246,
+      "eval_steps_per_second": 0.927,
+      "eval_token_acc": 0.9796258830367522,
+      "step": 45850
+    }
+  ],
+  "logging_steps": 100,
+  "max_steps": 45850,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.039735006810985e+23,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}