diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5140 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5000436033836225,
+  "eval_steps": 100,
+  "global_step": 5734,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0008720676724513823,
+      "grad_norm": 3.2646965980529785,
+      "learning_rate": 3.4843205574912896e-07,
+      "loss": 2.1031,
+      "mean_token_accuracy": 0.5400317996740341,
+      "step": 10
+    },
+    {
+      "epoch": 0.0017441353449027645,
+      "grad_norm": 3.1658802032470703,
+      "learning_rate": 6.968641114982579e-07,
+      "loss": 2.088,
+      "mean_token_accuracy": 0.5307852253317833,
+      "step": 20
+    },
+    {
+      "epoch": 0.002616203017354147,
+      "grad_norm": 3.1891283988952637,
+      "learning_rate": 1.045296167247387e-06,
+      "loss": 2.1143,
+      "mean_token_accuracy": 0.5352128148078918,
+      "step": 30
+    },
+    {
+      "epoch": 0.003488270689805529,
+      "grad_norm": 3.1211650371551514,
+      "learning_rate": 1.3937282229965158e-06,
+      "loss": 2.1139,
+      "mean_token_accuracy": 0.5289628133177757,
+      "step": 40
+    },
+    {
+      "epoch": 0.004360338362256911,
+      "grad_norm": 3.2871005535125732,
+      "learning_rate": 1.742160278745645e-06,
+      "loss": 1.9739,
+      "mean_token_accuracy": 0.5467098891735077,
+      "step": 50
+    },
+    {
+      "epoch": 0.005232406034708294,
+      "grad_norm": 3.0558788776397705,
+      "learning_rate": 2.090592334494774e-06,
+      "loss": 1.9814,
+      "mean_token_accuracy": 0.5516878694295884,
+      "step": 60
+    },
+    {
+      "epoch": 0.0061044737071596755,
+      "grad_norm": 3.216984748840332,
+      "learning_rate": 2.4390243902439027e-06,
+      "loss": 2.0823,
+      "mean_token_accuracy": 0.5344422727823257,
+      "step": 70
+    },
+    {
+      "epoch": 0.006976541379611058,
+      "grad_norm": 3.1541659832000732,
+      "learning_rate": 2.7874564459930316e-06,
+      "loss": 1.9995,
+      "mean_token_accuracy": 0.545340022444725,
+      "step": 80
+    },
+    {
+      "epoch": 0.00784860905206244,
+      "grad_norm": 3.530167818069458,
+      "learning_rate": 3.13588850174216e-06,
+      "loss": 2.1172,
+      "mean_token_accuracy": 0.5265533238649368,
+      "step": 90
+    },
+    {
+      "epoch": 0.008720676724513822,
+      "grad_norm": 3.2280466556549072,
+      "learning_rate": 3.48432055749129e-06,
+      "loss": 1.9648,
+      "mean_token_accuracy": 0.5563111662864685,
+      "step": 100
+    },
+    {
+      "epoch": 0.008720676724513822,
+      "eval_loss": 1.9722827672958374,
+      "eval_mean_token_accuracy": 0.5575271305170926,
+      "eval_runtime": 2.9374,
+      "eval_samples_per_second": 373.458,
+      "eval_steps_per_second": 93.62,
+      "step": 100
+    },
+    {
+      "epoch": 0.009592744396965204,
+      "grad_norm": 3.1962242126464844,
+      "learning_rate": 3.832752613240418e-06,
+      "loss": 1.9852,
+      "mean_token_accuracy": 0.5532901138067245,
+      "step": 110
+    },
+    {
+      "epoch": 0.010464812069416587,
+      "grad_norm": 3.323593854904175,
+      "learning_rate": 4.181184668989548e-06,
+      "loss": 2.0632,
+      "mean_token_accuracy": 0.5446183919906616,
+      "step": 120
+    },
+    {
+      "epoch": 0.01133687974186797,
+      "grad_norm": 3.4790284633636475,
+      "learning_rate": 4.529616724738676e-06,
+      "loss": 2.0602,
+      "mean_token_accuracy": 0.5417196691036225,
+      "step": 130
+    },
+    {
+      "epoch": 0.012208947414319351,
+      "grad_norm": 3.2306854724884033,
+      "learning_rate": 4.8780487804878055e-06,
+      "loss": 1.9934,
+      "mean_token_accuracy": 0.5517245590686798,
+      "step": 140
+    },
+    {
+      "epoch": 0.013081015086770733,
+      "grad_norm": 3.3311405181884766,
+      "learning_rate": 5.226480836236935e-06,
+      "loss": 1.9607,
+      "mean_token_accuracy": 0.5559931486845017,
+      "step": 150
+    },
+    {
+      "epoch": 0.013953082759222116,
+      "grad_norm": 3.2877893447875977,
+      "learning_rate": 5.574912891986063e-06,
+      "loss": 1.953,
+      "mean_token_accuracy": 0.5566658526659012,
+      "step": 160
+    },
+    {
+      "epoch": 0.014825150431673498,
+      "grad_norm": 3.261805772781372,
+      "learning_rate": 5.923344947735193e-06,
+      "loss": 2.0449,
+      "mean_token_accuracy": 0.5460127189755439,
+      "step": 170
+    },
+    {
+      "epoch": 0.01569721810412488,
+      "grad_norm": 3.374148368835449,
+      "learning_rate": 6.27177700348432e-06,
+      "loss": 1.952,
+      "mean_token_accuracy": 0.5566658496856689,
+      "step": 180
+    },
+    {
+      "epoch": 0.01656928577657626,
+      "grad_norm": 3.4749622344970703,
+      "learning_rate": 6.62020905923345e-06,
+      "loss": 2.0616,
+      "mean_token_accuracy": 0.5360689774155617,
+      "step": 190
+    },
+    {
+      "epoch": 0.017441353449027645,
+      "grad_norm": 3.3622708320617676,
+      "learning_rate": 6.96864111498258e-06,
+      "loss": 1.9372,
+      "mean_token_accuracy": 0.5612402096390724,
+      "step": 200
+    },
+    {
+      "epoch": 0.017441353449027645,
+      "eval_loss": 1.9787964820861816,
+      "eval_mean_token_accuracy": 0.5555221512100913,
+      "eval_runtime": 2.9304,
+      "eval_samples_per_second": 374.354,
+      "eval_steps_per_second": 93.844,
+      "step": 200
+    },
+    {
+      "epoch": 0.018313421121479028,
+      "grad_norm": 3.2971785068511963,
+      "learning_rate": 7.317073170731707e-06,
+      "loss": 2.0375,
+      "mean_token_accuracy": 0.5421110570430756,
+      "step": 210
+    },
+    {
+      "epoch": 0.019185488793930408,
+      "grad_norm": 3.0966248512268066,
+      "learning_rate": 7.665505226480837e-06,
+      "loss": 1.9448,
+      "mean_token_accuracy": 0.5616560697555542,
+      "step": 220
+    },
+    {
+      "epoch": 0.02005755646638179,
+      "grad_norm": 3.1848320960998535,
+      "learning_rate": 8.013937282229966e-06,
+      "loss": 1.9626,
+      "mean_token_accuracy": 0.5553693771362305,
+      "step": 230
+    },
+    {
+      "epoch": 0.020929624138833175,
+      "grad_norm": 3.2643938064575195,
+      "learning_rate": 8.362369337979095e-06,
+      "loss": 2.0678,
+      "mean_token_accuracy": 0.53862524330616,
+      "step": 240
+    },
+    {
+      "epoch": 0.021801691811284555,
+      "grad_norm": 3.1305971145629883,
+      "learning_rate": 8.710801393728223e-06,
+      "loss": 1.9376,
+      "mean_token_accuracy": 0.5613747581839561,
+      "step": 250
+    },
+    {
+      "epoch": 0.02267375948373594,
+      "grad_norm": 3.349454164505005,
+      "learning_rate": 9.059233449477352e-06,
+      "loss": 1.9744,
+      "mean_token_accuracy": 0.5525440335273742,
+      "step": 260
+    },
+    {
+      "epoch": 0.023545827156187322,
+      "grad_norm": 3.5459022521972656,
+      "learning_rate": 9.407665505226482e-06,
+      "loss": 1.9617,
+      "mean_token_accuracy": 0.5578889518976211,
+      "step": 270
+    },
+    {
+      "epoch": 0.024417894828638702,
+      "grad_norm": 3.436187744140625,
+      "learning_rate": 9.756097560975611e-06,
+      "loss": 2.0593,
+      "mean_token_accuracy": 0.5424902111291885,
+      "step": 280
+    },
+    {
+      "epoch": 0.025289962501090085,
+      "grad_norm": 3.153066396713257,
+      "learning_rate": 1.0104529616724739e-05,
+      "loss": 1.9667,
+      "mean_token_accuracy": 0.5562377721071243,
+      "step": 290
+    },
+    {
+      "epoch": 0.026162030173541465,
+      "grad_norm": 3.3692901134490967,
+      "learning_rate": 1.045296167247387e-05,
+      "loss": 1.9894,
+      "mean_token_accuracy": 0.5470768094062806,
+      "step": 300
+    },
+    {
+      "epoch": 0.026162030173541465,
+      "eval_loss": 1.9900883436203003,
+      "eval_mean_token_accuracy": 0.5534993772073226,
+      "eval_runtime": 2.9352,
+      "eval_samples_per_second": 373.742,
+      "eval_steps_per_second": 93.691,
+      "step": 300
+    },
+    {
+      "epoch": 0.02703409784599285,
+      "grad_norm": 3.403388023376465,
+      "learning_rate": 1.0801393728222997e-05,
+      "loss": 2.0423,
+      "mean_token_accuracy": 0.5436888471245765,
+      "step": 310
+    },
+    {
+      "epoch": 0.027906165518444232,
+      "grad_norm": 3.2832868099212646,
+      "learning_rate": 1.1149825783972127e-05,
+      "loss": 2.0218,
+      "mean_token_accuracy": 0.5370352253317833,
+      "step": 320
+    },
+    {
+      "epoch": 0.028778233190895612,
+      "grad_norm": 3.327082633972168,
+      "learning_rate": 1.1498257839721256e-05,
+      "loss": 1.9284,
+      "mean_token_accuracy": 0.5605552881956101,
+      "step": 330
+    },
+    {
+      "epoch": 0.029650300863346996,
+      "grad_norm": 3.3699028491973877,
+      "learning_rate": 1.1846689895470385e-05,
+      "loss": 1.9838,
+      "mean_token_accuracy": 0.5560909986495972,
+      "step": 340
+    },
+    {
+      "epoch": 0.03052236853579838,
+      "grad_norm": 3.3363900184631348,
+      "learning_rate": 1.2195121951219513e-05,
+      "loss": 1.9309,
+      "mean_token_accuracy": 0.561166824400425,
+      "step": 350
+    },
+    {
+      "epoch": 0.03139443620824976,
+      "grad_norm": 3.691817283630371,
+      "learning_rate": 1.254355400696864e-05,
+      "loss": 2.0536,
+      "mean_token_accuracy": 0.5458170250058174,
+      "step": 360
+    },
+    {
+      "epoch": 0.03226650388070114,
+      "grad_norm": 3.43636417388916,
+      "learning_rate": 1.2891986062717772e-05,
+      "loss": 1.9653,
+      "mean_token_accuracy": 0.5548434451222419,
+      "step": 370
+    },
+    {
+      "epoch": 0.03313857155315252,
+      "grad_norm": 3.4029386043548584,
+      "learning_rate": 1.32404181184669e-05,
+      "loss": 1.9544,
+      "mean_token_accuracy": 0.5557118415832519,
+      "step": 380
+    },
+    {
+      "epoch": 0.034010639225603906,
+      "grad_norm": 3.443406820297241,
+      "learning_rate": 1.3588850174216028e-05,
+      "loss": 1.9551,
+      "mean_token_accuracy": 0.5582558631896972,
+      "step": 390
+    },
+    {
+      "epoch": 0.03488270689805529,
+      "grad_norm": 3.3380706310272217,
+      "learning_rate": 1.393728222996516e-05,
+      "loss": 1.9104,
+      "mean_token_accuracy": 0.567172209918499,
+      "step": 400
+    },
+    {
+      "epoch": 0.03488270689805529,
+      "eval_loss": 2.0009982585906982,
+      "eval_mean_token_accuracy": 0.553236078674143,
+      "eval_runtime": 2.9319,
+      "eval_samples_per_second": 374.158,
+      "eval_steps_per_second": 93.795,
+      "step": 400
+    },
+    {
+      "epoch": 0.03575477457050667,
+      "grad_norm": 3.519862413406372,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 1.9613,
+      "mean_token_accuracy": 0.5559686899185181,
+      "step": 410
+    },
+    {
+      "epoch": 0.036626842242958056,
+      "grad_norm": 3.5910491943359375,
+      "learning_rate": 1.4634146341463415e-05,
+      "loss": 1.9113,
+      "mean_token_accuracy": 0.566267117857933,
+      "step": 420
+    },
+    {
+      "epoch": 0.03749890991540943,
+      "grad_norm": 3.2554426193237305,
+      "learning_rate": 1.4982578397212544e-05,
+      "loss": 1.9542,
+      "mean_token_accuracy": 0.5568737626075745,
+      "step": 430
+    },
+    {
+      "epoch": 0.038370977587860816,
+      "grad_norm": 3.5809357166290283,
+      "learning_rate": 1.5331010452961673e-05,
+      "loss": 1.9338,
+      "mean_token_accuracy": 0.5618273049592972,
+      "step": 440
+    },
+    {
+      "epoch": 0.0392430452603122,
+      "grad_norm": 3.343474864959717,
+      "learning_rate": 1.5679442508710803e-05,
+      "loss": 1.9458,
+      "mean_token_accuracy": 0.560836598277092,
+      "step": 450
+    },
+    {
+      "epoch": 0.04011511293276358,
+      "grad_norm": 3.550192356109619,
+      "learning_rate": 1.6027874564459932e-05,
+      "loss": 1.9829,
+      "mean_token_accuracy": 0.5522627234458923,
+      "step": 460
+    },
+    {
+      "epoch": 0.040987180605214966,
+      "grad_norm": 3.4415013790130615,
+      "learning_rate": 1.637630662020906e-05,
+      "loss": 1.9704,
+      "mean_token_accuracy": 0.5508561640977859,
+      "step": 470
+    },
+    {
+      "epoch": 0.04185924827766635,
+      "grad_norm": 3.3616232872009277,
+      "learning_rate": 1.672473867595819e-05,
+      "loss": 1.9243,
+      "mean_token_accuracy": 0.5677959948778153,
+      "step": 480
+    },
+    {
+      "epoch": 0.042731315950117726,
+      "grad_norm": 3.4711287021636963,
+      "learning_rate": 1.7073170731707317e-05,
+      "loss": 1.9122,
+      "mean_token_accuracy": 0.564334636926651,
+      "step": 490
+    },
+    {
+      "epoch": 0.04360338362256911,
+      "grad_norm": 3.3743157386779785,
+      "learning_rate": 1.7421602787456446e-05,
+      "loss": 1.9858,
+      "mean_token_accuracy": 0.5544642880558968,
+      "step": 500
+    },
+    {
+      "epoch": 0.04360338362256911,
+      "eval_loss": 2.015192747116089,
+      "eval_mean_token_accuracy": 0.551407224373384,
+      "eval_runtime": 2.9236,
+      "eval_samples_per_second": 375.226,
+      "eval_steps_per_second": 94.063,
+      "step": 500
+    },
+    {
+      "epoch": 0.04447545129502049,
+      "grad_norm": 3.3399884700775146,
+      "learning_rate": 1.7770034843205575e-05,
+      "loss": 1.9659,
+      "mean_token_accuracy": 0.5558830812573433,
+      "step": 510
+    },
+    {
+      "epoch": 0.04534751896747188,
+      "grad_norm": 3.305018663406372,
+      "learning_rate": 1.8118466898954705e-05,
+      "loss": 1.9253,
+      "mean_token_accuracy": 0.562866935133934,
+      "step": 520
+    },
+    {
+      "epoch": 0.04621958663992326,
+      "grad_norm": 3.5674850940704346,
+      "learning_rate": 1.8466898954703834e-05,
+      "loss": 1.9291,
+      "mean_token_accuracy": 0.5584637999534607,
+      "step": 530
+    },
+    {
+      "epoch": 0.047091654312374644,
+      "grad_norm": 3.5518832206726074,
+      "learning_rate": 1.8815331010452963e-05,
+      "loss": 1.9729,
+      "mean_token_accuracy": 0.5560909956693649,
+      "step": 540
+    },
+    {
+      "epoch": 0.04796372198482602,
+      "grad_norm": 3.5625863075256348,
+      "learning_rate": 1.9163763066202093e-05,
+      "loss": 1.8626,
+      "mean_token_accuracy": 0.5706457912921905,
+      "step": 550
+    },
+    {
+      "epoch": 0.048835789657277404,
+      "grad_norm": 3.486783504486084,
+      "learning_rate": 1.9512195121951222e-05,
+      "loss": 1.9581,
+      "mean_token_accuracy": 0.5598336637020112,
+      "step": 560
+    },
+    {
+      "epoch": 0.04970785732972879,
+      "grad_norm": 3.670612096786499,
+      "learning_rate": 1.9860627177700348e-05,
+      "loss": 1.9866,
+      "mean_token_accuracy": 0.5558096915483475,
+      "step": 570
+    },
+    {
+      "epoch": 0.05057992500218017,
+      "grad_norm": 3.5085062980651855,
+      "learning_rate": 1.9999933277491715e-05,
+      "loss": 1.9583,
+      "mean_token_accuracy": 0.5615704476833343,
+      "step": 580
+    },
+    {
+      "epoch": 0.051451992674631554,
+      "grad_norm": 3.531385898590088,
+      "learning_rate": 1.999952553205438e-05,
+      "loss": 1.935,
+      "mean_token_accuracy": 0.5650807231664657,
+      "step": 590
+    },
+    {
+      "epoch": 0.05232406034708293,
+      "grad_norm": 3.6724016666412354,
+      "learning_rate": 1.999874712433585e-05,
+      "loss": 1.8664,
+      "mean_token_accuracy": 0.574425145983696,
+      "step": 600
+    },
+    {
+      "epoch": 0.05232406034708293,
+      "eval_loss": 2.024757146835327,
+      "eval_mean_token_accuracy": 0.5481853763623671,
+      "eval_runtime": 2.9292,
+      "eval_samples_per_second": 374.51,
+      "eval_steps_per_second": 93.883,
+      "step": 600
+    },
+    {
+      "epoch": 0.053196128019534314,
+      "grad_norm": 3.3257014751434326,
+      "learning_rate": 1.999759808319013e-05,
+      "loss": 1.883,
+      "mean_token_accuracy": 0.5699364066123962,
+      "step": 610
+    },
+    {
+      "epoch": 0.0540681956919857,
+      "grad_norm": 3.4826889038085938,
+      "learning_rate": 1.9996078451209863e-05,
+      "loss": 1.9776,
+      "mean_token_accuracy": 0.5568737760186195,
+      "step": 620
+    },
+    {
+      "epoch": 0.05494026336443708,
+      "grad_norm": 3.429898500442505,
+      "learning_rate": 1.999418828472475e-05,
+      "loss": 1.9489,
+      "mean_token_accuracy": 0.5581457942724228,
+      "step": 630
+    },
+    {
+      "epoch": 0.055812331036888464,
+      "grad_norm": 3.3332772254943848,
+      "learning_rate": 1.9991927653799458e-05,
+      "loss": 1.9224,
+      "mean_token_accuracy": 0.5676369845867157,
+      "step": 640
+    },
+    {
+      "epoch": 0.05668439870933985,
+      "grad_norm": 3.5195119380950928,
+      "learning_rate": 1.998929664223102e-05,
+      "loss": 1.937,
+      "mean_token_accuracy": 0.5581335574388504,
+      "step": 650
+    },
+    {
+      "epoch": 0.057556466381791224,
+      "grad_norm": 3.63681960105896,
+      "learning_rate": 1.9986295347545738e-05,
+      "loss": 1.9182,
+      "mean_token_accuracy": 0.5633806258440017,
+      "step": 660
+    },
+    {
+      "epoch": 0.05842853405424261,
+      "grad_norm": 3.607254981994629,
+      "learning_rate": 1.998292388099557e-05,
+      "loss": 2.0223,
+      "mean_token_accuracy": 0.5514799490571022,
+      "step": 670
+    },
+    {
+      "epoch": 0.05930060172669399,
+      "grad_norm": 3.578639268875122,
+      "learning_rate": 1.9979182367553994e-05,
+      "loss": 1.9407,
+      "mean_token_accuracy": 0.5622064501047135,
+      "step": 680
+    },
+    {
+      "epoch": 0.060172669399145375,
+      "grad_norm": 3.4681434631347656,
+      "learning_rate": 1.997507094591137e-05,
+      "loss": 1.9334,
+      "mean_token_accuracy": 0.5626467734575271,
+      "step": 690
+    },
+    {
+      "epoch": 0.06104473707159676,
+      "grad_norm": 3.5321435928344727,
+      "learning_rate": 1.9970589768469833e-05,
+      "loss": 1.9544,
+      "mean_token_accuracy": 0.5644447147846222,
+      "step": 700
+    },
+    {
+      "epoch": 0.06104473707159676,
+      "eval_loss": 2.0260138511657715,
+      "eval_mean_token_accuracy": 0.5491638492454182,
+      "eval_runtime": 2.9248,
+      "eval_samples_per_second": 375.071,
+      "eval_steps_per_second": 94.024,
+      "step": 700
+    },
+    {
+      "epoch": 0.06191680474404814,
+      "grad_norm": 3.4142160415649414,
+      "learning_rate": 1.996573900133761e-05,
+      "loss": 1.9082,
+      "mean_token_accuracy": 0.5613502949476242,
+      "step": 710
+    },
+    {
+      "epoch": 0.06278887241649952,
+      "grad_norm": 3.385344982147217,
+      "learning_rate": 1.996051882432286e-05,
+      "loss": 1.8899,
+      "mean_token_accuracy": 0.569496089220047,
+      "step": 720
+    },
+    {
+      "epoch": 0.0636609400889509,
+      "grad_norm": 3.330402374267578,
+      "learning_rate": 1.995492943092705e-05,
+      "loss": 1.8893,
+      "mean_token_accuracy": 0.5647015661001206,
+      "step": 730
+    },
+    {
+      "epoch": 0.06453300776140228,
+      "grad_norm": 3.3513271808624268,
+      "learning_rate": 1.9948971028337737e-05,
+      "loss": 1.9085,
+      "mean_token_accuracy": 0.5631115525960922,
+      "step": 740
+    },
+    {
+      "epoch": 0.06540507543385367,
+      "grad_norm": 3.5575687885284424,
+      "learning_rate": 1.9942643837420904e-05,
+      "loss": 1.9366,
+      "mean_token_accuracy": 0.5645303368568421,
+      "step": 750
+    },
+    {
+      "epoch": 0.06627714310630504,
+      "grad_norm": 3.6231048107147217,
+      "learning_rate": 1.9935948092712792e-05,
+      "loss": 1.8756,
+      "mean_token_accuracy": 0.5724682003259659,
+      "step": 760
+    },
+    {
+      "epoch": 0.06714921077875644,
+      "grad_norm": 3.6453053951263428,
+      "learning_rate": 1.992888404241117e-05,
+      "loss": 1.9674,
+      "mean_token_accuracy": 0.5565068453550339,
+      "step": 770
+    },
+    {
+      "epoch": 0.06802127845120781,
+      "grad_norm": 3.6801161766052246,
+      "learning_rate": 1.992145194836616e-05,
+      "loss": 1.9418,
+      "mean_token_accuracy": 0.5615092873573303,
+      "step": 780
+    },
+    {
+      "epoch": 0.0688933461236592,
+      "grad_norm": 3.532811403274536,
+      "learning_rate": 1.9913652086070535e-05,
+      "loss": 1.9404,
+      "mean_token_accuracy": 0.5594911932945251,
+      "step": 790
+    },
+    {
+      "epoch": 0.06976541379611058,
+      "grad_norm": 3.7176992893218994,
+      "learning_rate": 1.9905484744649484e-05,
+      "loss": 1.9642,
+      "mean_token_accuracy": 0.5557485312223435,
+      "step": 800
+    },
+    {
+      "epoch": 0.06976541379611058,
+      "eval_loss": 2.0320777893066406,
+      "eval_mean_token_accuracy": 0.5476303155855698,
+      "eval_runtime": 2.9272,
+      "eval_samples_per_second": 374.759,
+      "eval_steps_per_second": 93.946,
+      "step": 800
+    },
+    {
+      "epoch": 0.07063748146856196,
+      "grad_norm": 3.7104685306549072,
+      "learning_rate": 1.989695022684991e-05,
+      "loss": 1.9098,
+      "mean_token_accuracy": 0.5611301362514496,
+      "step": 810
+    },
+    {
+      "epoch": 0.07150954914101335,
+      "grad_norm": 3.5141100883483887,
+      "learning_rate": 1.988804884902921e-05,
+      "loss": 1.9205,
+      "mean_token_accuracy": 0.570523488521576,
+      "step": 820
+    },
+    {
+      "epoch": 0.07238161681346472,
+      "grad_norm": 3.603508949279785,
+      "learning_rate": 1.9878780941143538e-05,
+      "loss": 1.9079,
+      "mean_token_accuracy": 0.5614481419324875,
+      "step": 830
+    },
+    {
+      "epoch": 0.07325368448591611,
+      "grad_norm": 3.485046625137329,
+      "learning_rate": 1.9869146846735576e-05,
+      "loss": 1.8945,
+      "mean_token_accuracy": 0.5684931486845016,
+      "step": 840
+    },
+    {
+      "epoch": 0.07412575215836749,
+      "grad_norm": 3.6356348991394043,
+      "learning_rate": 1.985914692292182e-05,
+      "loss": 1.9382,
+      "mean_token_accuracy": 0.5595890402793884,
+      "step": 850
+    },
+    {
+      "epoch": 0.07499781983081887,
+      "grad_norm": 3.4124903678894043,
+      "learning_rate": 1.9848781540379312e-05,
+      "loss": 1.9227,
+      "mean_token_accuracy": 0.5668664410710335,
+      "step": 860
+    },
+    {
+      "epoch": 0.07586988750327026,
+      "grad_norm": 3.586756706237793,
+      "learning_rate": 1.983805108333191e-05,
+      "loss": 1.9989,
+      "mean_token_accuracy": 0.5550880625844001,
+      "step": 870
+    },
+    {
+      "epoch": 0.07674195517572163,
+      "grad_norm": 3.569127321243286,
+      "learning_rate": 1.9826955949536062e-05,
+      "loss": 1.9012,
+      "mean_token_accuracy": 0.5628669291734696,
+      "step": 880
+    },
+    {
+      "epoch": 0.07761402284817302,
+      "grad_norm": 3.446479558944702,
+      "learning_rate": 1.9815496550266036e-05,
+      "loss": 1.9369,
+      "mean_token_accuracy": 0.5592465758323669,
+      "step": 890
+    },
+    {
+      "epoch": 0.0784860905206244,
+      "grad_norm": 3.549825668334961,
+      "learning_rate": 1.98036733102987e-05,
+      "loss": 1.9849,
+      "mean_token_accuracy": 0.5525601238012314,
+      "step": 900
+    },
+    {
+      "epoch": 0.0784860905206244,
+      "eval_loss": 2.0255126953125,
+      "eval_mean_token_accuracy": 0.5478313466635617,
+      "eval_runtime": 2.9176,
+      "eval_samples_per_second": 375.992,
+      "eval_steps_per_second": 94.255,
+      "step": 900
+    },
+    {
+      "epoch": 0.07935815819307579,
+      "grad_norm": 3.392361879348755,
+      "learning_rate": 1.979148666789775e-05,
+      "loss": 1.9978,
+      "mean_token_accuracy": 0.5582069501280784,
+      "step": 910
+    },
+    {
+      "epoch": 0.08023022586552717,
+      "grad_norm": 3.431718349456787,
+      "learning_rate": 1.9778937074797494e-05,
+      "loss": 1.9129,
+      "mean_token_accuracy": 0.5620898872613906,
+      "step": 920
+    },
+    {
+      "epoch": 0.08110229353797854,
+      "grad_norm": 3.3219411373138428,
+      "learning_rate": 1.976602499618608e-05,
+      "loss": 1.9226,
+      "mean_token_accuracy": 0.5610689789056778,
+      "step": 930
+    },
+    {
+      "epoch": 0.08197436121042993,
+      "grad_norm": 3.687722682952881,
+      "learning_rate": 1.9752750910688278e-05,
+      "loss": 1.9181,
+      "mean_token_accuracy": 0.5723459005355835,
+      "step": 940
+    },
+    {
+      "epoch": 0.08284642888288131,
+      "grad_norm": 3.7822089195251465,
+      "learning_rate": 1.9739115310347698e-05,
+      "loss": 1.8826,
+      "mean_token_accuracy": 0.5668175160884857,
+      "step": 950
+    },
+    {
+      "epoch": 0.0837184965553327,
+      "grad_norm": 3.4353253841400146,
+      "learning_rate": 1.972511870060861e-05,
+      "loss": 1.9572,
+      "mean_token_accuracy": 0.5530821949243545,
+      "step": 960
+    },
+    {
+      "epoch": 0.08459056422778408,
+      "grad_norm": 3.5201773643493652,
+      "learning_rate": 1.9710761600297147e-05,
+      "loss": 1.9518,
+      "mean_token_accuracy": 0.560371819138527,
+      "step": 970
+    },
+    {
+      "epoch": 0.08546263190023545,
+      "grad_norm": 3.762462854385376,
+      "learning_rate": 1.9696044541602126e-05,
+      "loss": 1.9894,
+      "mean_token_accuracy": 0.5503913968801498,
+      "step": 980
+    },
+    {
+      "epoch": 0.08633469957268684,
+      "grad_norm": 3.530463933944702,
+      "learning_rate": 1.968096807005528e-05,
+      "loss": 1.9197,
+      "mean_token_accuracy": 0.5655455082654953,
+      "step": 990
+    },
+    {
+      "epoch": 0.08720676724513822,
+      "grad_norm": 3.4875705242156982,
+      "learning_rate": 1.966553274451106e-05,
+      "loss": 1.8823,
+      "mean_token_accuracy": 0.5705479502677917,
+      "step": 1000
+    },
+    {
+      "epoch": 0.08720676724513822,
+      "eval_loss": 2.026561737060547,
+      "eval_mean_token_accuracy": 0.5482689892161976,
+      "eval_runtime": 2.9226,
+      "eval_samples_per_second": 375.355,
+      "eval_steps_per_second": 94.095,
+      "step": 1000
+    },
+    {
+      "epoch": 0.08807883491758961,
+      "grad_norm": 3.5463833808898926,
+      "learning_rate": 1.964973913712591e-05,
+      "loss": 1.909,
+      "mean_token_accuracy": 0.5641022473573685,
+      "step": 1010
+    },
+    {
+      "epoch": 0.08895090259004099,
+      "grad_norm": 3.6753158569335938,
+      "learning_rate": 1.9633587833337064e-05,
+      "loss": 1.8993,
+      "mean_token_accuracy": 0.561472600698471,
+      "step": 1020
+    },
+    {
+      "epoch": 0.08982297026249236,
+      "grad_norm": 3.825209140777588,
+      "learning_rate": 1.961707943184083e-05,
+      "loss": 1.8815,
+      "mean_token_accuracy": 0.5647871851921081,
+      "step": 1030
+    },
+    {
+      "epoch": 0.09069503793494375,
+      "grad_norm": 3.664457082748413,
+      "learning_rate": 1.9600214544570432e-05,
+      "loss": 1.8971,
+      "mean_token_accuracy": 0.567881602048874,
+      "step": 1040
+    },
+    {
+      "epoch": 0.09156710560739513,
+      "grad_norm": 3.6888506412506104,
+      "learning_rate": 1.958299379667328e-05,
+      "loss": 1.9481,
+      "mean_token_accuracy": 0.5514554798603057,
+      "step": 1050
+    },
+    {
+      "epoch": 0.09243917327984652,
+      "grad_norm": 3.4865269660949707,
+      "learning_rate": 1.9565417826487835e-05,
+      "loss": 1.8536,
+      "mean_token_accuracy": 0.5786692768335342,
+      "step": 1060
+    },
+    {
+      "epoch": 0.0933112409522979,
+      "grad_norm": 3.71282696723938,
+      "learning_rate": 1.9547487285519922e-05,
+      "loss": 1.9755,
+      "mean_token_accuracy": 0.5581870883703232,
+      "step": 1070
+    },
+    {
+      "epoch": 0.09418330862474929,
+      "grad_norm": 3.48748779296875,
+      "learning_rate": 1.952920283841861e-05,
+      "loss": 1.9328,
+      "mean_token_accuracy": 0.5592832684516906,
+      "step": 1080
+    },
+    {
+      "epoch": 0.09505537629720066,
+      "grad_norm": 3.607158899307251,
+      "learning_rate": 1.9510565162951538e-05,
+      "loss": 1.9077,
+      "mean_token_accuracy": 0.562634539604187,
+      "step": 1090
+    },
+    {
+      "epoch": 0.09592744396965204,
+      "grad_norm": 3.3436968326568604,
+      "learning_rate": 1.9491574949979814e-05,
+      "loss": 1.9423,
+      "mean_token_accuracy": 0.5633072450757026,
+      "step": 1100
+    },
+    {
+      "epoch": 0.09592744396965204,
+      "eval_loss": 2.025158643722534,
+      "eval_mean_token_accuracy": 0.5492634755914861,
+      "eval_runtime": 2.9247,
+      "eval_samples_per_second": 375.082,
+      "eval_steps_per_second": 94.027,
+      "step": 1100
+    },
+    {
+      "epoch": 0.09679951164210343,
+      "grad_norm": 3.4683494567871094,
+      "learning_rate": 1.9472232903432406e-05,
+      "loss": 1.8574,
+      "mean_token_accuracy": 0.5820694714784622,
+      "step": 1110
+    },
+    {
+      "epoch": 0.09767157931455481,
+      "grad_norm": 3.5293450355529785,
+      "learning_rate": 1.945253974028004e-05,
+      "loss": 1.9546,
+      "mean_token_accuracy": 0.55911203622818,
+      "step": 1120
+    },
+    {
+      "epoch": 0.0985436469870062,
+      "grad_norm": 3.480376958847046,
+      "learning_rate": 1.9432496190508633e-05,
+      "loss": 1.8918,
+      "mean_token_accuracy": 0.5657289654016495,
+      "step": 1130
+    },
+    {
+      "epoch": 0.09941571465945757,
+      "grad_norm": 3.470508098602295,
+      "learning_rate": 1.941210299709222e-05,
+      "loss": 1.9403,
+      "mean_token_accuracy": 0.5624266117811203,
+      "step": 1140
+    },
+    {
+      "epoch": 0.10028778233190895,
+      "grad_norm": 3.520244598388672,
+      "learning_rate": 1.9391360915965426e-05,
+      "loss": 1.9163,
+      "mean_token_accuracy": 0.5676247596740722,
+      "step": 1150
+    },
+    {
+      "epoch": 0.10115985000436034,
+      "grad_norm": 3.808081865310669,
+      "learning_rate": 1.9370270715995447e-05,
+      "loss": 1.9633,
+      "mean_token_accuracy": 0.5575954034924507,
+      "step": 1160
+    },
+    {
+      "epoch": 0.10203191767681172,
+      "grad_norm": 3.405604124069214,
+      "learning_rate": 1.934883317895354e-05,
+      "loss": 1.8977,
+      "mean_token_accuracy": 0.5709148794412613,
+      "step": 1170
+    },
+    {
+      "epoch": 0.10290398534926311,
+      "grad_norm": 3.729879856109619,
+      "learning_rate": 1.932704909948604e-05,
+      "loss": 1.9385,
+      "mean_token_accuracy": 0.5575220078229904,
+      "step": 1180
+    },
+    {
+      "epoch": 0.10377605302171448,
+      "grad_norm": 3.677964925765991,
+      "learning_rate": 1.930491928508492e-05,
+      "loss": 1.8937,
+      "mean_token_accuracy": 0.5677103698253632,
+      "step": 1190
+    },
+    {
+      "epoch": 0.10464812069416586,
+      "grad_norm": 3.529926300048828,
+      "learning_rate": 1.9282444556057855e-05,
+      "loss": 1.885,
+      "mean_token_accuracy": 0.5668664395809173,
+      "step": 1200
+    },
+    {
+      "epoch": 0.10464812069416586,
+      "eval_loss": 2.027597427368164,
+      "eval_mean_token_accuracy": 0.5492154421589591,
+      "eval_runtime": 2.9241,
+      "eval_samples_per_second": 375.163,
+      "eval_steps_per_second": 94.047,
+      "step": 1200
+    },
+    {
+      "epoch": 0.10552018836661725,
+      "grad_norm": 3.4195098876953125,
+      "learning_rate": 1.9259625745497803e-05,
+      "loss": 1.8349,
+      "mean_token_accuracy": 0.576467701792717,
+      "step": 1210
+    },
+    {
+      "epoch": 0.10639225603906863,
+      "grad_norm": 3.41968035697937,
+      "learning_rate": 1.9236463699252136e-05,
+      "loss": 1.8824,
+      "mean_token_accuracy": 0.5667318910360336,
+      "step": 1220
+    },
+    {
+      "epoch": 0.10726432371152002,
+      "grad_norm": 3.790433645248413,
+      "learning_rate": 1.921295927589127e-05,
+      "loss": 1.8666,
+      "mean_token_accuracy": 0.5674779891967774,
+      "step": 1230
+    },
+    {
+      "epoch": 0.1081363913839714,
+      "grad_norm": 3.5514166355133057,
+      "learning_rate": 1.9189113346676878e-05,
+      "loss": 1.9275,
+      "mean_token_accuracy": 0.5630014717578888,
+      "step": 1240
+    },
+    {
+      "epoch": 0.10900845905642279,
+      "grad_norm": 3.62327241897583,
+      "learning_rate": 1.916492679552954e-05,
+      "loss": 1.8653,
+      "mean_token_accuracy": 0.5693982392549515,
+      "step": 1250
+    },
+    {
+      "epoch": 0.10988052672887416,
+      "grad_norm": 3.5460891723632812,
+      "learning_rate": 1.914040051899602e-05,
+      "loss": 1.8928,
+      "mean_token_accuracy": 0.5742661416530609,
+      "step": 1260
+    },
+    {
+      "epoch": 0.11075259440132554,
+      "grad_norm": 3.5049819946289062,
+      "learning_rate": 1.9115535426216018e-05,
+      "loss": 1.8443,
+      "mean_token_accuracy": 0.5785469651222229,
+      "step": 1270
+    },
+    {
+      "epoch": 0.11162466207377693,
+      "grad_norm": 3.552345037460327,
+      "learning_rate": 1.9090332438888458e-05,
+      "loss": 1.9099,
+      "mean_token_accuracy": 0.5679672226309777,
+      "step": 1280
+    },
+    {
+      "epoch": 0.1124967297462283,
+      "grad_norm": 3.282322406768799,
+      "learning_rate": 1.906479249123735e-05,
+      "loss": 1.8531,
+      "mean_token_accuracy": 0.5759540140628815,
+      "step": 1290
+    },
+    {
+      "epoch": 0.1133687974186797,
+      "grad_norm": 3.6338813304901123,
+      "learning_rate": 1.9038916529977136e-05,
+      "loss": 1.9086,
+      "mean_token_accuracy": 0.5693248540163041,
+      "step": 1300
+    },
+    {
+      "epoch": 0.1133687974186797,
+      "eval_loss": 2.022566556930542,
+      "eval_mean_token_accuracy": 0.5497331442616202,
+      "eval_runtime": 2.9268,
+      "eval_samples_per_second": 374.809,
+      "eval_steps_per_second": 93.958,
+      "step": 1300
+    },
+    {
+      "epoch": 0.11424086509113107,
+      "grad_norm": 3.639498472213745,
+      "learning_rate": 1.901270551427761e-05,
+      "loss": 1.9082,
+      "mean_token_accuracy": 0.5654231831431389,
+      "step": 1310
+    },
+    {
+      "epoch": 0.11511293276358245,
+      "grad_norm": 3.433962821960449,
+      "learning_rate": 1.898616041572836e-05,
+      "loss": 1.933,
+      "mean_token_accuracy": 0.5564823895692825,
+      "step": 1320
+    },
+    {
+      "epoch": 0.11598500043603384,
+      "grad_norm": 3.346561908721924,
+      "learning_rate": 1.8959282218302746e-05,
+      "loss": 1.8041,
+      "mean_token_accuracy": 0.5848214328289032,
+      "step": 1330
+    },
+    {
+      "epoch": 0.11685706810848522,
+      "grad_norm": 3.4933876991271973,
+      "learning_rate": 1.893207191832144e-05,
+      "loss": 1.9061,
+      "mean_token_accuracy": 0.56475048661232,
+      "step": 1340
+    },
+    {
+      "epoch": 0.1177291357809366,
+      "grad_norm": 3.5106682777404785,
+      "learning_rate": 1.8904530524415483e-05,
+      "loss": 1.936,
+      "mean_token_accuracy": 0.5622186839580536,
+      "step": 1350
+    },
+    {
+      "epoch": 0.11860120345338798,
+      "grad_norm": 3.614764451980591,
+      "learning_rate": 1.8876659057488905e-05,
+      "loss": 1.8617,
+      "mean_token_accuracy": 0.5762108594179154,
+      "step": 1360
+    },
+    {
+      "epoch": 0.11947327112583936,
+      "grad_norm": 3.7362213134765625,
+      "learning_rate": 1.8848458550680875e-05,
+      "loss": 1.8759,
+      "mean_token_accuracy": 0.5728473544120789,
+      "step": 1370
+    },
+    {
+      "epoch": 0.12034533879829075,
+      "grad_norm": 3.540865182876587,
+      "learning_rate": 1.8819930049327412e-05,
+      "loss": 1.8802,
+      "mean_token_accuracy": 0.5739481389522553,
+      "step": 1380
+    },
+    {
+      "epoch": 0.12121740647074213,
+      "grad_norm": 3.6314971446990967,
+      "learning_rate": 1.8791074610922624e-05,
+      "loss": 1.9243,
+      "mean_token_accuracy": 0.5682974547147751,
+      "step": 1390
+    },
+    {
+      "epoch": 0.12208947414319352,
+      "grad_norm": 3.684410333633423,
+      "learning_rate": 1.8761893305079528e-05,
+      "loss": 1.9517,
+      "mean_token_accuracy": 0.5556873708963395,
+      "step": 1400
+    },
+    {
+      "epoch": 0.12208947414319352,
+      "eval_loss": 2.020735263824463,
+      "eval_mean_token_accuracy": 0.5499928851561113,
+      "eval_runtime": 2.9246,
+      "eval_samples_per_second": 375.094,
+      "eval_steps_per_second": 94.03,
+      "step": 1400
+    },
+    {
+      "epoch": 0.12296154181564489,
+      "grad_norm": 3.7083048820495605,
+      "learning_rate": 1.873238721349038e-05,
+      "loss": 1.908,
+      "mean_token_accuracy": 0.5679794549942017,
+      "step": 1410
+    },
+    {
+      "epoch": 0.12383360948809628,
+      "grad_norm": 3.665344715118408,
+      "learning_rate": 1.8702557429886607e-05,
+      "loss": 1.9146,
+      "mean_token_accuracy": 0.5651418745517731,
+      "step": 1420
+    },
+    {
+      "epoch": 0.12470567716054766,
+      "grad_norm": 3.5886929035186768,
+      "learning_rate": 1.8672405059998228e-05,
+      "loss": 1.8477,
+      "mean_token_accuracy": 0.5771893352270127,
+      "step": 1430
+    },
+    {
+      "epoch": 0.12557774483299905,
+      "grad_norm": 3.258273124694824,
+      "learning_rate": 1.8641931221512895e-05,
+      "loss": 1.9279,
+      "mean_token_accuracy": 0.5603351265192031,
+      "step": 1440
+    },
+    {
+      "epoch": 0.12644981250545043,
+      "grad_norm": 3.558218479156494,
+      "learning_rate": 1.8611137044034454e-05,
+      "loss": 1.9204,
+      "mean_token_accuracy": 0.5681996017694473,
+      "step": 1450
+    },
+    {
+      "epoch": 0.1273218801779018,
+      "grad_norm": 3.50213360786438,
+      "learning_rate": 1.858002366904107e-05,
+      "loss": 1.9053,
+      "mean_token_accuracy": 0.5669642835855484,
+      "step": 1460
+    },
+    {
+      "epoch": 0.12819394785035318,
+      "grad_norm": 3.522390365600586,
+      "learning_rate": 1.854859224984292e-05,
+      "loss": 1.9066,
+      "mean_token_accuracy": 0.5656311169266701,
+      "step": 1470
+    },
+    {
+      "epoch": 0.12906601552280456,
+      "grad_norm": 3.6531217098236084,
+      "learning_rate": 1.851684395153944e-05,
+      "loss": 1.8943,
+      "mean_token_accuracy": 0.5665729016065597,
+      "step": 1480
+    },
+    {
+      "epoch": 0.12993808319525596,
+      "grad_norm": 3.740518569946289,
+      "learning_rate": 1.8484779950976133e-05,
+      "loss": 1.9471,
+      "mean_token_accuracy": 0.5571673199534416,
+      "step": 1490
+    },
+    {
+      "epoch": 0.13081015086770734,
+      "grad_norm": 3.618528366088867,
+      "learning_rate": 1.8452401436700954e-05,
+      "loss": 1.987,
+      "mean_token_accuracy": 0.5550269067287446,
+      "step": 1500
+    },
+    {
+      "epoch": 0.13081015086770734,
+      "eval_loss": 2.0220086574554443,
+      "eval_mean_token_accuracy": 0.5497882934050127,
+      "eval_runtime": 2.9315,
+      "eval_samples_per_second": 374.215,
+      "eval_steps_per_second": 93.81,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1316822185401587,
+      "grad_norm": 3.5748484134674072,
+      "learning_rate": 1.8419709608920243e-05,
+      "loss": 1.9666,
+      "mean_token_accuracy": 0.5587451130151748,
+      "step": 1510
+    },
+    {
+      "epoch": 0.1325542862126101,
+      "grad_norm": 3.694185256958008,
+      "learning_rate": 1.8386705679454243e-05,
+      "loss": 1.859,
+      "mean_token_accuracy": 0.5766389489173889,
+      "step": 1520
+    },
+    {
+      "epoch": 0.1334263538850615,
+      "grad_norm": 3.688524007797241,
+      "learning_rate": 1.8353390871692176e-05,
+      "loss": 1.9474,
+      "mean_token_accuracy": 0.5584271103143692,
+      "step": 1530
+    },
+    {
+      "epoch": 0.13429842155751287,
+      "grad_norm": 3.546990394592285,
+      "learning_rate": 1.8319766420546902e-05,
+      "loss": 1.8792,
+      "mean_token_accuracy": 0.568346381187439,
+      "step": 1540
+    },
+    {
+      "epoch": 0.13517048922996425,
+      "grad_norm": 3.5605790615081787,
+      "learning_rate": 1.8285833572409135e-05,
+      "loss": 1.8686,
+      "mean_token_accuracy": 0.57100048661232,
+      "step": 1550
+    },
+    {
+      "epoch": 0.13604255690241562,
+      "grad_norm": 3.7089226245880127,
+      "learning_rate": 1.8251593585101243e-05,
+      "loss": 1.8796,
+      "mean_token_accuracy": 0.5667685896158219,
+      "step": 1560
+    },
+    {
+      "epoch": 0.136914624574867,
+      "grad_norm": 3.6720075607299805,
+      "learning_rate": 1.821704772783063e-05,
+      "loss": 1.9229,
+      "mean_token_accuracy": 0.5663160502910614,
+      "step": 1570
+    },
+    {
+      "epoch": 0.1377866922473184,
+      "grad_norm": 3.6228814125061035,
+      "learning_rate": 1.818219728114267e-05,
+      "loss": 1.8947,
+      "mean_token_accuracy": 0.5709760218858719,
+      "step": 1580
+    },
+    {
+      "epoch": 0.13865875991976978,
+      "grad_norm": 3.5334060192108154,
+      "learning_rate": 1.8147043536873275e-05,
+      "loss": 1.9041,
+      "mean_token_accuracy": 0.5682607650756836,
+      "step": 1590
+    },
+    {
+      "epoch": 0.13953082759222116,
+      "grad_norm": 3.3918418884277344,
+      "learning_rate": 1.8111587798100974e-05,
+      "loss": 1.9319,
+      "mean_token_accuracy": 0.5641267120838165,
+      "step": 1600
+    },
+    {
+      "epoch": 0.13953082759222116,
+      "eval_loss": 2.0149195194244385,
+      "eval_mean_token_accuracy": 0.5498007453571666,
+      "eval_runtime": 2.926,
+      "eval_samples_per_second": 374.916,
+      "eval_steps_per_second": 93.985,
+      "step": 1600
+    },
+    {
+      "epoch": 0.14040289526467253,
+      "grad_norm": 3.6828675270080566,
+      "learning_rate": 1.807583137909862e-05,
+      "loss": 1.9258,
+      "mean_token_accuracy": 0.5654965728521347,
+      "step": 1610
+    },
+    {
+      "epoch": 0.1412749629371239,
+      "grad_norm": 3.4600281715393066,
+      "learning_rate": 1.8039775605284687e-05,
+      "loss": 1.8344,
+      "mean_token_accuracy": 0.5753791570663452,
+      "step": 1620
+    },
+    {
+      "epoch": 0.14214703060957531,
+      "grad_norm": 3.39148211479187,
+      "learning_rate": 1.800342181317413e-05,
+      "loss": 1.9394,
+      "mean_token_accuracy": 0.5610934421420097,
+      "step": 1630
+    },
+    {
+      "epoch": 0.1430190982820267,
+      "grad_norm": 3.4984686374664307,
+      "learning_rate": 1.7966771350328825e-05,
+      "loss": 1.9235,
+      "mean_token_accuracy": 0.5562989234924316,
+      "step": 1640
+    },
+    {
+      "epoch": 0.14389116595447807,
+      "grad_norm": 3.7628092765808105,
+      "learning_rate": 1.7929825575307665e-05,
+      "loss": 1.927,
+      "mean_token_accuracy": 0.5617539048194885,
+      "step": 1650
+    },
+    {
+      "epoch": 0.14476323362692944,
+      "grad_norm": 3.7283425331115723,
+      "learning_rate": 1.7892585857616144e-05,
+      "loss": 1.9028,
+      "mean_token_accuracy": 0.5693860113620758,
+      "step": 1660
+    },
+    {
+      "epoch": 0.14563530129938082,
+      "grad_norm": 3.8388149738311768,
+      "learning_rate": 1.785505357765563e-05,
+      "loss": 1.9144,
+      "mean_token_accuracy": 0.5629892379045487,
+      "step": 1670
+    },
+    {
+      "epoch": 0.14650736897183222,
+      "grad_norm": 3.470796585083008,
+      "learning_rate": 1.781723012667218e-05,
+      "loss": 1.8556,
+      "mean_token_accuracy": 0.5772015661001205,
+      "step": 1680
+    },
+    {
+      "epoch": 0.1473794366442836,
+      "grad_norm": 3.8921432495117188,
+      "learning_rate": 1.7779116906704986e-05,
+      "loss": 1.8911,
+      "mean_token_accuracy": 0.5664872765541077,
+      "step": 1690
+    },
+    {
+      "epoch": 0.14825150431673498,
+      "grad_norm": 3.590428113937378,
+      "learning_rate": 1.7740715330534383e-05,
+      "loss": 1.9019,
+      "mean_token_accuracy": 0.5726027384400367,
+      "step": 1700
+    },
+    {
+      "epoch": 0.14825150431673498,
+      "eval_loss": 2.0144217014312744,
+      "eval_mean_token_accuracy": 0.5494075776230205,
+      "eval_runtime": 2.9211,
+      "eval_samples_per_second": 375.549,
+      "eval_steps_per_second": 94.144,
+      "step": 1700
+    },
+    {
+      "epoch": 0.14912357198918635,
+      "grad_norm": 3.6635448932647705,
+      "learning_rate": 1.770202682162949e-05,
+      "loss": 1.8568,
+      "mean_token_accuracy": 0.5772504806518555,
+      "step": 1710
+    },
+    {
+      "epoch": 0.14999563966163773,
+      "grad_norm": 3.625452756881714,
+      "learning_rate": 1.7663052814095447e-05,
+      "loss": 1.8831,
+      "mean_token_accuracy": 0.5688233852386475,
+      "step": 1720
+    },
+    {
+      "epoch": 0.15086770733408914,
+      "grad_norm": 3.5212619304656982,
+      "learning_rate": 1.7623794752620255e-05,
+      "loss": 1.9596,
+      "mean_token_accuracy": 0.5570939347147942,
+      "step": 1730
+    },
+    {
+      "epoch": 0.1517397750065405,
+      "grad_norm": 3.5392465591430664,
+      "learning_rate": 1.7584254092421226e-05,
+      "loss": 1.9244,
+      "mean_token_accuracy": 0.560628667473793,
+      "step": 1740
+    },
+    {
+      "epoch": 0.1526118426789919,
+      "grad_norm": 3.5339748859405518,
+      "learning_rate": 1.754443229919103e-05,
+      "loss": 1.8832,
+      "mean_token_accuracy": 0.5658757373690605,
+      "step": 1750
+    },
+    {
+      "epoch": 0.15348391035144326,
+      "grad_norm": 3.6682333946228027,
+      "learning_rate": 1.7504330849043373e-05,
+      "loss": 1.9053,
+      "mean_token_accuracy": 0.5658635020256042,
+      "step": 1760
+    },
+    {
+      "epoch": 0.15435597802389464,
+      "grad_norm": 3.3598413467407227,
+      "learning_rate": 1.7463951228458288e-05,
+      "loss": 1.9112,
+      "mean_token_accuracy": 0.5623287558555603,
+      "step": 1770
+    },
+    {
+      "epoch": 0.15522804569634605,
+      "grad_norm": 3.4572224617004395,
+      "learning_rate": 1.7423294934227017e-05,
+      "loss": 1.9264,
+      "mean_token_accuracy": 0.5614848345518112,
+      "step": 1780
+    },
+    {
+      "epoch": 0.15610011336879742,
+      "grad_norm": 3.4597063064575195,
+      "learning_rate": 1.7382363473396543e-05,
+      "loss": 1.9578,
+      "mean_token_accuracy": 0.5607754439115524,
+      "step": 1790
+    },
+    {
+      "epoch": 0.1569721810412488,
+      "grad_norm": 3.817779541015625,
+      "learning_rate": 1.734115836321372e-05,
+      "loss": 1.9452,
+      "mean_token_accuracy": 0.5661570429801941,
+      "step": 1800
+    },
+    {
+      "epoch": 0.1569721810412488,
+      "eval_loss": 2.0112361907958984,
+      "eval_mean_token_accuracy": 0.551012275760824,
+      "eval_runtime": 2.9313,
+      "eval_samples_per_second": 374.242,
+      "eval_steps_per_second": 93.816,
+      "step": 1800
+    },
+    {
+      "epoch": 0.15784424871370017,
+      "grad_norm": 3.5462937355041504,
+      "learning_rate": 1.7299681131069026e-05,
+      "loss": 1.8223,
+      "mean_token_accuracy": 0.5791952043771744,
+      "step": 1810
+    },
+    {
+      "epoch": 0.15871631638615158,
+      "grad_norm": 3.5150744915008545,
+      "learning_rate": 1.725793331443996e-05,
+      "loss": 1.8841,
+      "mean_token_accuracy": 0.5724192708730698,
+      "step": 1820
+    },
+    {
+      "epoch": 0.15958838405860296,
+      "grad_norm": 3.523815393447876,
+      "learning_rate": 1.7215916460834048e-05,
+      "loss": 1.9568,
+      "mean_token_accuracy": 0.5602128118276596,
+      "step": 1830
+    },
+    {
+      "epoch": 0.16046045173105433,
+      "grad_norm": 3.467637062072754,
+      "learning_rate": 1.7173632127731462e-05,
+      "loss": 1.8975,
+      "mean_token_accuracy": 0.5665117412805557,
+      "step": 1840
+    },
+    {
+      "epoch": 0.1613325194035057,
+      "grad_norm": 3.6244378089904785,
+      "learning_rate": 1.7131081882527305e-05,
+      "loss": 1.8731,
+      "mean_token_accuracy": 0.5704378604888916,
+      "step": 1850
+    },
+    {
+      "epoch": 0.16220458707595709,
+      "grad_norm": 3.4595024585723877,
+      "learning_rate": 1.708826730247351e-05,
+      "loss": 1.9484,
+      "mean_token_accuracy": 0.5563967674970627,
+      "step": 1860
+    },
+    {
+      "epoch": 0.1630766547484085,
+      "grad_norm": 3.4926583766937256,
+      "learning_rate": 1.704518997462037e-05,
+      "loss": 1.9113,
+      "mean_token_accuracy": 0.5690313100814819,
+      "step": 1870
+    },
+    {
+      "epoch": 0.16394872242085987,
+      "grad_norm": 3.5621678829193115,
+      "learning_rate": 1.7001851495757708e-05,
+      "loss": 1.8423,
+      "mean_token_accuracy": 0.5772137939929962,
+      "step": 1880
+    },
+    {
+      "epoch": 0.16482079009331124,
+      "grad_norm": 3.7352819442749023,
+      "learning_rate": 1.6958253472355687e-05,
+      "loss": 1.9834,
+      "mean_token_accuracy": 0.5561643779277802,
+      "step": 1890
+    },
+    {
+      "epoch": 0.16569285776576262,
+      "grad_norm": 3.823659896850586,
+      "learning_rate": 1.6914397520505267e-05,
+      "loss": 1.9612,
+      "mean_token_accuracy": 0.5548189848661422,
+      "step": 1900
+    },
+    {
+      "epoch": 0.16569285776576262,
+      "eval_loss": 2.0061721801757812,
+      "eval_mean_token_accuracy": 0.5503504705429078,
+      "eval_runtime": 2.9323,
+      "eval_samples_per_second": 374.113,
+      "eval_steps_per_second": 93.784,
+      "step": 1900
+    },
+    {
+      "epoch": 0.166564925438214,
+      "grad_norm": 3.519951820373535,
+      "learning_rate": 1.6870285265858298e-05,
+      "loss": 1.9213,
+      "mean_token_accuracy": 0.5607876807451249,
+      "step": 1910
+    },
+    {
+      "epoch": 0.1674369931106654,
+      "grad_norm": 3.875131607055664,
+      "learning_rate": 1.6825918343567257e-05,
+      "loss": 1.9006,
+      "mean_token_accuracy": 0.5676247537136078,
+      "step": 1920
+    },
+    {
+      "epoch": 0.16830906078311678,
+      "grad_norm": 3.5626649856567383,
+      "learning_rate": 1.678129839822463e-05,
+      "loss": 1.9315,
+      "mean_token_accuracy": 0.567820456624031,
+      "step": 1930
+    },
+    {
+      "epoch": 0.16918112845556815,
+      "grad_norm": 3.5646631717681885,
+      "learning_rate": 1.673642708380198e-05,
+      "loss": 1.8352,
+      "mean_token_accuracy": 0.5709515661001205,
+      "step": 1940
+    },
+    {
+      "epoch": 0.17005319612801953,
+      "grad_norm": 3.6227667331695557,
+      "learning_rate": 1.6691306063588583e-05,
+      "loss": 1.8847,
+      "mean_token_accuracy": 0.5619618356227875,
+      "step": 1950
+    },
+    {
+      "epoch": 0.1709252638004709,
+      "grad_norm": 3.4506328105926514,
+      "learning_rate": 1.6645937010129837e-05,
+      "loss": 1.8269,
+      "mean_token_accuracy": 0.5764065533876419,
+      "step": 1960
+    },
+    {
+      "epoch": 0.1717973314729223,
+      "grad_norm": 3.461332321166992,
+      "learning_rate": 1.660032160516522e-05,
+      "loss": 1.8926,
+      "mean_token_accuracy": 0.564004409313202,
+      "step": 1970
+    },
+    {
+      "epoch": 0.1726693991453737,
+      "grad_norm": 3.537386655807495,
+      "learning_rate": 1.6554461539565953e-05,
+      "loss": 1.891,
+      "mean_token_accuracy": 0.5662915825843811,
+      "step": 1980
+    },
+    {
+      "epoch": 0.17354146681782506,
+      "grad_norm": 3.6319923400878906,
+      "learning_rate": 1.650835851327236e-05,
+      "loss": 1.8848,
+      "mean_token_accuracy": 0.5678571432828903,
+      "step": 1990
+    },
+    {
+      "epoch": 0.17441353449027644,
+      "grad_norm": 3.5577962398529053,
+      "learning_rate": 1.6462014235230805e-05,
+      "loss": 1.8614,
+      "mean_token_accuracy": 0.5675513714551925,
+      "step": 2000
+    },
+    {
+      "epoch": 0.17441353449027644,
+      "eval_loss": 2.001716375350952,
+      "eval_mean_token_accuracy": 0.5525084492293271,
+      "eval_runtime": 2.9259,
+      "eval_samples_per_second": 374.922,
+      "eval_steps_per_second": 93.987,
+      "step": 2000
+    },
+    {
+      "epoch": 0.17528560216272782,
+      "grad_norm": 3.5866129398345947,
+      "learning_rate": 1.641543042333038e-05,
+      "loss": 1.8603,
+      "mean_token_accuracy": 0.5740826815366745,
+      "step": 2010
+    },
+    {
+      "epoch": 0.17615766983517922,
+      "grad_norm": 3.563718318939209,
+      "learning_rate": 1.636860880433922e-05,
+      "loss": 2.0117,
+      "mean_token_accuracy": 0.5486912965774536,
+      "step": 2020
+    },
+    {
+      "epoch": 0.1770297375076306,
+      "grad_norm": 3.543128252029419,
+      "learning_rate": 1.632155111384047e-05,
+      "loss": 1.934,
+      "mean_token_accuracy": 0.5600782752037048,
+      "step": 2030
+    },
+    {
+      "epoch": 0.17790180518008197,
+      "grad_norm": 3.53792405128479,
+      "learning_rate": 1.6274259096168e-05,
+      "loss": 1.9506,
+      "mean_token_accuracy": 0.5586350291967392,
+      "step": 2040
+    },
+    {
+      "epoch": 0.17877387285253335,
+      "grad_norm": 3.493621826171875,
+      "learning_rate": 1.622673450434169e-05,
+      "loss": 1.8968,
+      "mean_token_accuracy": 0.5687622278928757,
+      "step": 2050
+    },
+    {
+      "epoch": 0.17964594052498473,
+      "grad_norm": 3.529428243637085,
+      "learning_rate": 1.6178979100002486e-05,
+      "loss": 1.922,
+      "mean_token_accuracy": 0.5647627174854278,
+      "step": 2060
+    },
+    {
+      "epoch": 0.18051800819743613,
+      "grad_norm": 3.6281440258026123,
+      "learning_rate": 1.6130994653347096e-05,
+      "loss": 1.9655,
+      "mean_token_accuracy": 0.5570939302444458,
+      "step": 2070
+    },
+    {
+      "epoch": 0.1813900758698875,
+      "grad_norm": 3.715604782104492,
+      "learning_rate": 1.6082782943062355e-05,
+      "loss": 1.9332,
+      "mean_token_accuracy": 0.5580357193946839,
+      "step": 2080
+    },
+    {
+      "epoch": 0.18226214354233888,
+      "grad_norm": 3.7165727615356445,
+      "learning_rate": 1.6034345756259303e-05,
+      "loss": 1.8397,
+      "mean_token_accuracy": 0.5799779832363129,
+      "step": 2090
+    },
+    {
+      "epoch": 0.18313421121479026,
+      "grad_norm": 3.582075834274292,
+      "learning_rate": 1.598568488840695e-05,
+      "loss": 1.9288,
+      "mean_token_accuracy": 0.5644447118043899,
+      "step": 2100
+    },
+    {
+      "epoch": 0.18313421121479026,
+      "eval_loss": 1.9990211725234985,
+      "eval_mean_token_accuracy": 0.5530901976065202,
+      "eval_runtime": 2.9256,
+      "eval_samples_per_second": 374.97,
+      "eval_steps_per_second": 93.999,
+      "step": 2100
+    },
+    {
+      "epoch": 0.18400627888724164,
+      "grad_norm": 3.5977823734283447,
+      "learning_rate": 1.5936802143265708e-05,
+      "loss": 1.9646,
+      "mean_token_accuracy": 0.5549779817461967,
+      "step": 2110
+    },
+    {
+      "epoch": 0.18487834655969304,
+      "grad_norm": 3.5821807384490967,
+      "learning_rate": 1.5887699332820527e-05,
+      "loss": 1.902,
+      "mean_token_accuracy": 0.5618945598602295,
+      "step": 2120
+    },
+    {
+      "epoch": 0.18575041423214442,
+      "grad_norm": 3.5157365798950195,
+      "learning_rate": 1.5838378277213745e-05,
+      "loss": 1.9042,
+      "mean_token_accuracy": 0.5667563587427139,
+      "step": 2130
+    },
+    {
+      "epoch": 0.1866224819045958,
+      "grad_norm": 3.466606378555298,
+      "learning_rate": 1.57888408046776e-05,
+      "loss": 1.9226,
+      "mean_token_accuracy": 0.5600905090570449,
+      "step": 2140
+    },
+    {
+      "epoch": 0.18749454957704717,
+      "grad_norm": 3.539996385574341,
+      "learning_rate": 1.573908875146648e-05,
+      "loss": 1.8828,
+      "mean_token_accuracy": 0.5680895298719406,
+      "step": 2150
+    },
+    {
+      "epoch": 0.18836661724949857,
+      "grad_norm": 3.545053005218506,
+      "learning_rate": 1.5689123961788834e-05,
+      "loss": 1.9182,
+      "mean_token_accuracy": 0.5615704506635666,
+      "step": 2160
+    },
+    {
+      "epoch": 0.18923868492194995,
+      "grad_norm": 3.590813636779785,
+      "learning_rate": 1.563894828773883e-05,
+      "loss": 1.8929,
+      "mean_token_accuracy": 0.569630628824234,
+      "step": 2170
+    },
+    {
+      "epoch": 0.19011075259440133,
+      "grad_norm": 3.49204683303833,
+      "learning_rate": 1.55885635892277e-05,
+      "loss": 2.0161,
+      "mean_token_accuracy": 0.5436521515250206,
+      "step": 2180
+    },
+    {
+      "epoch": 0.1909828202668527,
+      "grad_norm": 3.5304877758026123,
+      "learning_rate": 1.5537971733914784e-05,
+      "loss": 1.968,
+      "mean_token_accuracy": 0.5524339556694031,
+      "step": 2190
+    },
+    {
+      "epoch": 0.19185488793930408,
+      "grad_norm": 3.5615243911743164,
+      "learning_rate": 1.5487174597138314e-05,
+      "loss": 1.858,
+      "mean_token_accuracy": 0.5788282722234726,
+      "step": 2200
+    },
+    {
+      "epoch": 0.19185488793930408,
+      "eval_loss": 1.9950103759765625,
+      "eval_mean_token_accuracy": 0.5527432828599756,
+      "eval_runtime": 2.9254,
+      "eval_samples_per_second": 374.987,
+      "eval_steps_per_second": 94.003,
+      "step": 2200
+    },
+    {
+      "epoch": 0.19272695561175548,
+      "grad_norm": 3.4310946464538574,
+      "learning_rate": 1.543617406184589e-05,
+      "loss": 1.9119,
+      "mean_token_accuracy": 0.5654598861932755,
+      "step": 2210
+    },
+    {
+      "epoch": 0.19359902328420686,
+      "grad_norm": 3.5315332412719727,
+      "learning_rate": 1.5384972018524678e-05,
+      "loss": 1.9075,
+      "mean_token_accuracy": 0.5703400194644928,
+      "step": 2220
+    },
+    {
+      "epoch": 0.19447109095665824,
+      "grad_norm": 3.6830127239227295,
+      "learning_rate": 1.5333570365131353e-05,
+      "loss": 1.9108,
+      "mean_token_accuracy": 0.5597969651222229,
+      "step": 2230
+    },
+    {
+      "epoch": 0.19534315862910961,
+      "grad_norm": 3.7514188289642334,
+      "learning_rate": 1.5281971007021728e-05,
+      "loss": 1.8982,
+      "mean_token_accuracy": 0.5667318940162659,
+      "step": 2240
+    },
+    {
+      "epoch": 0.196215226301561,
+      "grad_norm": 3.47749662399292,
+      "learning_rate": 1.5230175856880132e-05,
+      "loss": 1.8738,
+      "mean_token_accuracy": 0.571636500954628,
+      "step": 2250
+    },
+    {
+      "epoch": 0.1970872939740124,
+      "grad_norm": 3.2910730838775635,
+      "learning_rate": 1.5178186834648509e-05,
+      "loss": 1.8697,
+      "mean_token_accuracy": 0.5772137910127639,
+      "step": 2260
+    },
+    {
+      "epoch": 0.19795936164646377,
+      "grad_norm": 3.574847936630249,
+      "learning_rate": 1.5126005867455256e-05,
+      "loss": 1.8559,
+      "mean_token_accuracy": 0.5704990208148957,
+      "step": 2270
+    },
+    {
+      "epoch": 0.19883142931891515,
+      "grad_norm": 3.706261396408081,
+      "learning_rate": 1.5073634889543778e-05,
+      "loss": 2.0243,
+      "mean_token_accuracy": 0.5446795552968979,
+      "step": 2280
+    },
+    {
+      "epoch": 0.19970349699136652,
+      "grad_norm": 3.3673837184906006,
+      "learning_rate": 1.5021075842200796e-05,
+      "loss": 1.9308,
+      "mean_token_accuracy": 0.5589652717113495,
+      "step": 2290
+    },
+    {
+      "epoch": 0.2005755646638179,
+      "grad_norm": 3.4486331939697266,
+      "learning_rate": 1.4968330673684387e-05,
+      "loss": 1.9387,
+      "mean_token_accuracy": 0.5605919793248176,
+      "step": 2300
+    },
+    {
+      "epoch": 0.2005755646638179,
+      "eval_loss": 1.990919589996338,
+      "eval_mean_token_accuracy": 0.5537448872219433,
+      "eval_runtime": 2.9235,
+      "eval_samples_per_second": 375.237,
+      "eval_steps_per_second": 94.066,
+      "step": 2300
+    },
+    {
+      "epoch": 0.2014476323362693,
+      "grad_norm": 3.601839065551758,
+      "learning_rate": 1.4915401339151769e-05,
+      "loss": 1.8633,
+      "mean_token_accuracy": 0.5703522533178329,
+      "step": 2310
+    },
+    {
+      "epoch": 0.20231970000872068,
+      "grad_norm": 3.475649118423462,
+      "learning_rate": 1.486228980058682e-05,
+      "loss": 1.9046,
+      "mean_token_accuracy": 0.5639921665191651,
+      "step": 2320
+    },
+    {
+      "epoch": 0.20319176768117206,
+      "grad_norm": 3.5041749477386475,
+      "learning_rate": 1.4808998026727348e-05,
+      "loss": 1.9579,
+      "mean_token_accuracy": 0.5542074427008629,
+      "step": 2330
+    },
+    {
+      "epoch": 0.20406383535362344,
+      "grad_norm": 3.5484228134155273,
+      "learning_rate": 1.4755527992992133e-05,
+      "loss": 1.8896,
+      "mean_token_accuracy": 0.5697162508964538,
+      "step": 2340
+    },
+    {
+      "epoch": 0.2049359030260748,
+      "grad_norm": 3.6420633792877197,
+      "learning_rate": 1.4701881681407684e-05,
+      "loss": 1.9134,
+      "mean_token_accuracy": 0.5662915885448456,
+      "step": 2350
+    },
+    {
+      "epoch": 0.20580797069852622,
+      "grad_norm": 3.5798981189727783,
+      "learning_rate": 1.464806108053477e-05,
+      "loss": 1.9019,
+      "mean_token_accuracy": 0.5640533268451691,
+      "step": 2360
+    },
+    {
+      "epoch": 0.2066800383709776,
+      "grad_norm": 3.660372734069824,
+      "learning_rate": 1.4594068185394723e-05,
+      "loss": 1.9093,
+      "mean_token_accuracy": 0.5591854214668274,
+      "step": 2370
+    },
+    {
+      "epoch": 0.20755210604342897,
+      "grad_norm": 3.6556289196014404,
+      "learning_rate": 1.4539904997395468e-05,
+      "loss": 1.9255,
+      "mean_token_accuracy": 0.5564456939697265,
+      "step": 2380
+    },
+    {
+      "epoch": 0.20842417371588035,
+      "grad_norm": 3.4871301651000977,
+      "learning_rate": 1.448557352425735e-05,
+      "loss": 1.9157,
+      "mean_token_accuracy": 0.5600660532712937,
+      "step": 2390
+    },
+    {
+      "epoch": 0.20929624138833172,
+      "grad_norm": 3.461557626724243,
+      "learning_rate": 1.44310757799387e-05,
+      "loss": 1.8434,
+      "mean_token_accuracy": 0.5769936472177506,
+      "step": 2400
+    },
+    {
+      "epoch": 0.20929624138833172,
+      "eval_loss": 1.9907763004302979,
+      "eval_mean_token_accuracy": 0.5537680130655115,
+      "eval_runtime": 2.9217,
+      "eval_samples_per_second": 375.469,
+      "eval_steps_per_second": 94.124,
+      "step": 2400
+    },
+    {
+      "epoch": 0.21016830906078313,
+      "grad_norm": 3.694026231765747,
+      "learning_rate": 1.437641378456119e-05,
+      "loss": 1.9953,
+      "mean_token_accuracy": 0.5531311199069023,
+      "step": 2410
+    },
+    {
+      "epoch": 0.2110403767332345,
+      "grad_norm": 3.566923141479492,
+      "learning_rate": 1.4321589564334946e-05,
+      "loss": 1.9027,
+      "mean_token_accuracy": 0.5658023476600647,
+      "step": 2420
+    },
+    {
+      "epoch": 0.21191244440568588,
+      "grad_norm": 3.5002241134643555,
+      "learning_rate": 1.4266605151483444e-05,
+      "loss": 1.9735,
+      "mean_token_accuracy": 0.5540606677532196,
+      "step": 2430
+    },
+    {
+      "epoch": 0.21278451207813726,
+      "grad_norm": 3.6515281200408936,
+      "learning_rate": 1.4211462584168178e-05,
+      "loss": 1.9203,
+      "mean_token_accuracy": 0.5673679009079933,
+      "step": 2440
+    },
+    {
+      "epoch": 0.21365657975058863,
+      "grad_norm": 3.421048641204834,
+      "learning_rate": 1.4156163906413113e-05,
+      "loss": 1.8613,
+      "mean_token_accuracy": 0.5732020556926727,
+      "step": 2450
+    },
+    {
+      "epoch": 0.21452864742304004,
+      "grad_norm": 3.623137950897217,
+      "learning_rate": 1.4100711168028906e-05,
+      "loss": 1.9272,
+      "mean_token_accuracy": 0.5626712322235108,
+      "step": 2460
+    },
+    {
+      "epoch": 0.2154007150954914,
+      "grad_norm": 3.645113468170166,
+      "learning_rate": 1.4045106424536938e-05,
+      "loss": 1.9162,
+      "mean_token_accuracy": 0.5643590986728668,
+      "step": 2470
+    },
+    {
+      "epoch": 0.2162727827679428,
+      "grad_norm": 3.738107204437256,
+      "learning_rate": 1.398935173709311e-05,
+      "loss": 1.9797,
+      "mean_token_accuracy": 0.5539995193481445,
+      "step": 2480
+    },
+    {
+      "epoch": 0.21714485044039417,
+      "grad_norm": 3.4691874980926514,
+      "learning_rate": 1.3933449172411446e-05,
+      "loss": 1.8819,
+      "mean_token_accuracy": 0.5657778859138489,
+      "step": 2490
+    },
+    {
+      "epoch": 0.21801691811284557,
+      "grad_norm": 3.3320319652557373,
+      "learning_rate": 1.387740080268748e-05,
+      "loss": 1.9079,
+      "mean_token_accuracy": 0.5680406093597412,
+      "step": 2500
+    },
+    {
+      "epoch": 0.21801691811284557,
+      "eval_loss": 1.9851176738739014,
+      "eval_mean_token_accuracy": 0.5547607199712233,
+      "eval_runtime": 2.9277,
+      "eval_samples_per_second": 374.703,
+      "eval_steps_per_second": 93.932,
+      "step": 2500
+    },
+    {
+      "epoch": 0.21888898578529695,
+      "grad_norm": 3.602503776550293,
+      "learning_rate": 1.3821208705521442e-05,
+      "loss": 1.9718,
+      "mean_token_accuracy": 0.5536325842142105,
+      "step": 2510
+    },
+    {
+      "epoch": 0.21976105345774832,
+      "grad_norm": 3.8319509029388428,
+      "learning_rate": 1.3764874963841255e-05,
+      "loss": 1.9679,
+      "mean_token_accuracy": 0.5532289713621139,
+      "step": 2520
+    },
+    {
+      "epoch": 0.2206331211301997,
+      "grad_norm": 3.896193027496338,
+      "learning_rate": 1.3708401665825319e-05,
+      "loss": 1.8752,
+      "mean_token_accuracy": 0.5649951100349426,
+      "step": 2530
+    },
+    {
+      "epoch": 0.22150518880265108,
+      "grad_norm": 3.625417470932007,
+      "learning_rate": 1.36517909048251e-05,
+      "loss": 1.9411,
+      "mean_token_accuracy": 0.5591364920139312,
+      "step": 2540
+    },
+    {
+      "epoch": 0.22237725647510248,
+      "grad_norm": 3.2977983951568604,
+      "learning_rate": 1.3595044779287543e-05,
+      "loss": 1.8901,
+      "mean_token_accuracy": 0.5609955936670303,
+      "step": 2550
+    },
+    {
+      "epoch": 0.22324932414755386,
+      "grad_norm": 3.7591755390167236,
+      "learning_rate": 1.3538165392677288e-05,
+      "loss": 1.946,
+      "mean_token_accuracy": 0.5591609597206115,
+      "step": 2560
+    },
+    {
+      "epoch": 0.22412139182000523,
+      "grad_norm": 3.4932961463928223,
+      "learning_rate": 1.3481154853398686e-05,
+      "loss": 1.9452,
+      "mean_token_accuracy": 0.5505503982305526,
+      "step": 2570
+    },
+    {
+      "epoch": 0.2249934594924566,
+      "grad_norm": 3.7724695205688477,
+      "learning_rate": 1.3424015274717665e-05,
+      "loss": 1.9359,
+      "mean_token_accuracy": 0.5597235798835755,
+      "step": 2580
+    },
+    {
+      "epoch": 0.225865527164908,
+      "grad_norm": 3.58357310295105,
+      "learning_rate": 1.3366748774683376e-05,
+      "loss": 1.856,
+      "mean_token_accuracy": 0.5746697694063186,
+      "step": 2590
+    },
+    {
+      "epoch": 0.2267375948373594,
+      "grad_norm": 3.539696216583252,
+      "learning_rate": 1.3309357476049686e-05,
+      "loss": 1.9073,
+      "mean_token_accuracy": 0.5640777915716171,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2267375948373594,
+      "eval_loss": 1.981577754020691,
+      "eval_mean_token_accuracy": 0.5552695239673962,
+      "eval_runtime": 2.9317,
+      "eval_samples_per_second": 374.184,
+      "eval_steps_per_second": 93.802,
+      "step": 2600
+    },
+    {
+      "epoch": 0.22760966250981077,
+      "grad_norm": 3.605401039123535,
+      "learning_rate": 1.3251843506196508e-05,
+      "loss": 1.9203,
+      "mean_token_accuracy": 0.5645058706402779,
+      "step": 2610
+    },
+    {
+      "epoch": 0.22848173018226214,
+      "grad_norm": 3.552380323410034,
+      "learning_rate": 1.3194208997050915e-05,
+      "loss": 1.8735,
+      "mean_token_accuracy": 0.575684928894043,
+      "step": 2620
+    },
+    {
+      "epoch": 0.22935379785471352,
+      "grad_norm": 3.3692588806152344,
+      "learning_rate": 1.313645608500814e-05,
+      "loss": 1.9279,
+      "mean_token_accuracy": 0.5613625228404999,
+      "step": 2630
+    },
+    {
+      "epoch": 0.2302258655271649,
+      "grad_norm": 3.645843267440796,
+      "learning_rate": 1.3078586910852364e-05,
+      "loss": 1.9269,
+      "mean_token_accuracy": 0.5609589055180549,
+      "step": 2640
+    },
+    {
+      "epoch": 0.2310979331996163,
+      "grad_norm": 3.3917734622955322,
+      "learning_rate": 1.3020603619677378e-05,
+      "loss": 1.8825,
+      "mean_token_accuracy": 0.5638820916414261,
+      "step": 2650
+    },
+    {
+      "epoch": 0.23197000087206768,
+      "grad_norm": 3.617940902709961,
+      "learning_rate": 1.296250836080706e-05,
+      "loss": 1.8613,
+      "mean_token_accuracy": 0.5684809178113938,
+      "step": 2660
+    },
+    {
+      "epoch": 0.23284206854451905,
+      "grad_norm": 3.4179887771606445,
+      "learning_rate": 1.2904303287715702e-05,
+      "loss": 1.8985,
+      "mean_token_accuracy": 0.5611668229103088,
+      "step": 2670
+    },
+    {
+      "epoch": 0.23371413621697043,
+      "grad_norm": 3.6363773345947266,
+      "learning_rate": 1.284599055794819e-05,
+      "loss": 1.9022,
+      "mean_token_accuracy": 0.5608365923166275,
+      "step": 2680
+    },
+    {
+      "epoch": 0.2345862038894218,
+      "grad_norm": 3.5546791553497314,
+      "learning_rate": 1.2787572333040022e-05,
+      "loss": 1.9519,
+      "mean_token_accuracy": 0.5555161505937576,
+      "step": 2690
+    },
+    {
+      "epoch": 0.2354582715618732,
+      "grad_norm": 3.4861562252044678,
+      "learning_rate": 1.2729050778437197e-05,
+      "loss": 1.9202,
+      "mean_token_accuracy": 0.5603228986263276,
+      "step": 2700
+    },
+    {
+      "epoch": 0.2354582715618732,
+      "eval_loss": 1.9748586416244507,
+      "eval_mean_token_accuracy": 0.5562657884034243,
+      "eval_runtime": 2.9246,
+      "eval_samples_per_second": 375.089,
+      "eval_steps_per_second": 94.029,
+      "step": 2700
+    },
+    {
+      "epoch": 0.2363303392343246,
+      "grad_norm": 3.520627021789551,
+      "learning_rate": 1.2670428063415932e-05,
+      "loss": 1.9578,
+      "mean_token_accuracy": 0.5569349259138108,
+      "step": 2710
+    },
+    {
+      "epoch": 0.23720240690677596,
+      "grad_norm": 3.7104015350341797,
+      "learning_rate": 1.2611706361002254e-05,
+      "loss": 1.8847,
+      "mean_token_accuracy": 0.5708170235157013,
+      "step": 2720
+    },
+    {
+      "epoch": 0.23807447457922734,
+      "grad_norm": 3.5505104064941406,
+      "learning_rate": 1.2552887847891462e-05,
+      "loss": 1.942,
+      "mean_token_accuracy": 0.5591854125261306,
+      "step": 2730
+    },
+    {
+      "epoch": 0.23894654225167872,
+      "grad_norm": 3.7813947200775146,
+      "learning_rate": 1.2493974704367427e-05,
+      "loss": 1.9118,
+      "mean_token_accuracy": 0.5603718221187591,
+      "step": 2740
+    },
+    {
+      "epoch": 0.23981860992413012,
+      "grad_norm": 3.5009207725524902,
+      "learning_rate": 1.2434969114221777e-05,
+      "loss": 1.9213,
+      "mean_token_accuracy": 0.5618028402328491,
+      "step": 2750
+    },
+    {
+      "epoch": 0.2406906775965815,
+      "grad_norm": 3.6487233638763428,
+      "learning_rate": 1.237587326467296e-05,
+      "loss": 1.9543,
+      "mean_token_accuracy": 0.5544520556926728,
+      "step": 2760
+    },
+    {
+      "epoch": 0.24156274526903287,
+      "grad_norm": 3.4972894191741943,
+      "learning_rate": 1.2316689346285146e-05,
+      "loss": 1.9235,
+      "mean_token_accuracy": 0.5622920677065849,
+      "step": 2770
+    },
+    {
+      "epoch": 0.24243481294148425,
+      "grad_norm": 3.3958864212036133,
+      "learning_rate": 1.2257419552887047e-05,
+      "loss": 1.9397,
+      "mean_token_accuracy": 0.5618272960186005,
+      "step": 2780
+    },
+    {
+      "epoch": 0.24330688061393563,
+      "grad_norm": 3.4992504119873047,
+      "learning_rate": 1.2198066081490585e-05,
+      "loss": 1.9026,
+      "mean_token_accuracy": 0.5659002006053925,
+      "step": 2790
+    },
+    {
+      "epoch": 0.24417894828638703,
+      "grad_norm": 3.6224024295806885,
+      "learning_rate": 1.213863113220946e-05,
+      "loss": 1.8988,
+      "mean_token_accuracy": 0.5669031322002411,
+      "step": 2800
+    },
+    {
+      "epoch": 0.24417894828638703,
+      "eval_loss": 1.970268726348877,
+      "eval_mean_token_accuracy": 0.5572407041896473,
+      "eval_runtime": 2.9281,
+      "eval_samples_per_second": 374.647,
+      "eval_steps_per_second": 93.918,
+      "step": 2800
+    },
+    {
+      "epoch": 0.2450510159588384,
+      "grad_norm": 3.330629825592041,
+      "learning_rate": 1.2079116908177592e-05,
+      "loss": 1.9452,
+      "mean_token_accuracy": 0.5591609537601471,
+      "step": 2810
+    },
+    {
+      "epoch": 0.24592308363128978,
+      "grad_norm": 3.3401975631713867,
+      "learning_rate": 1.2019525615467462e-05,
+      "loss": 1.9223,
+      "mean_token_accuracy": 0.5684809207916259,
+      "step": 2820
+    },
+    {
+      "epoch": 0.24679515130374116,
+      "grad_norm": 3.628615140914917,
+      "learning_rate": 1.1959859463008316e-05,
+      "loss": 1.9477,
+      "mean_token_accuracy": 0.5544031351804733,
+      "step": 2830
+    },
+    {
+      "epoch": 0.24766721897619257,
+      "grad_norm": 3.6269943714141846,
+      "learning_rate": 1.1900120662504315e-05,
+      "loss": 1.9112,
+      "mean_token_accuracy": 0.5584882616996765,
+      "step": 2840
+    },
+    {
+      "epoch": 0.24853928664864394,
+      "grad_norm": 3.498828887939453,
+      "learning_rate": 1.1840311428352536e-05,
+      "loss": 1.9284,
+      "mean_token_accuracy": 0.5641022458672523,
+      "step": 2850
+    },
+    {
+      "epoch": 0.24941135432109532,
+      "grad_norm": 3.47495436668396,
+      "learning_rate": 1.1780433977560879e-05,
+      "loss": 1.8903,
+      "mean_token_accuracy": 0.5707925677299499,
+      "step": 2860
+    },
+    {
+      "epoch": 0.2502834219935467,
+      "grad_norm": 3.534308910369873,
+      "learning_rate": 1.1720490529665904e-05,
+      "loss": 1.9475,
+      "mean_token_accuracy": 0.5607020556926727,
+      "step": 2870
+    },
+    {
+      "epoch": 0.2511554896659981,
+      "grad_norm": 3.3834636211395264,
+      "learning_rate": 1.1660483306650558e-05,
+      "loss": 1.8801,
+      "mean_token_accuracy": 0.5663527429103852,
+      "step": 2880
+    },
+    {
+      "epoch": 0.25202755733844945,
+      "grad_norm": 3.426891326904297,
+      "learning_rate": 1.160041453286179e-05,
+      "loss": 1.9942,
+      "mean_token_accuracy": 0.5529476553201675,
+      "step": 2890
+    },
+    {
+      "epoch": 0.25289962501090085,
+      "grad_norm": 3.628671884536743,
+      "learning_rate": 1.154028643492812e-05,
+      "loss": 1.9483,
+      "mean_token_accuracy": 0.5594422727823257,
+      "step": 2900
+    },
+    {
+      "epoch": 0.25289962501090085,
+      "eval_loss": 1.9672374725341797,
+      "eval_mean_token_accuracy": 0.5576658953319896,
+      "eval_runtime": 2.9312,
+      "eval_samples_per_second": 374.248,
+      "eval_steps_per_second": 93.818,
+      "step": 2900
+    },
+    {
+      "epoch": 0.2537716926833522,
+      "grad_norm": 3.3622570037841797,
+      "learning_rate": 1.1480101241677097e-05,
+      "loss": 1.9193,
+      "mean_token_accuracy": 0.5619985312223434,
+      "step": 2910
+    },
+    {
+      "epoch": 0.2546437603558036,
+      "grad_norm": 3.421581268310547,
+      "learning_rate": 1.1419861184052669e-05,
+      "loss": 1.9014,
+      "mean_token_accuracy": 0.5665484309196472,
+      "step": 2920
+    },
+    {
+      "epoch": 0.255515828028255,
+      "grad_norm": 3.4504427909851074,
+      "learning_rate": 1.1359568495032505e-05,
+      "loss": 1.9944,
+      "mean_token_accuracy": 0.5514065533876419,
+      "step": 2930
+    },
+    {
+      "epoch": 0.25638789570070636,
+      "grad_norm": 3.570554256439209,
+      "learning_rate": 1.1299225409545207e-05,
+      "loss": 1.873,
+      "mean_token_accuracy": 0.5731824666261673,
+      "step": 2940
+    },
+    {
+      "epoch": 0.25725996337315776,
+      "grad_norm": 3.6744372844696045,
+      "learning_rate": 1.123883416438748e-05,
+      "loss": 1.942,
+      "mean_token_accuracy": 0.5542196661233902,
+      "step": 2950
+    },
+    {
+      "epoch": 0.2581320310456091,
+      "grad_norm": 3.462813377380371,
+      "learning_rate": 1.1178396998141206e-05,
+      "loss": 1.9019,
+      "mean_token_accuracy": 0.5666707396507263,
+      "step": 2960
+    },
+    {
+      "epoch": 0.2590040987180605,
+      "grad_norm": 3.584836721420288,
+      "learning_rate": 1.1117916151090469e-05,
+      "loss": 1.908,
+      "mean_token_accuracy": 0.5664261221885681,
+      "step": 2970
+    },
+    {
+      "epoch": 0.2598761663905119,
+      "grad_norm": 3.5248260498046875,
+      "learning_rate": 1.1057393865138513e-05,
+      "loss": 1.9094,
+      "mean_token_accuracy": 0.5577177196741104,
+      "step": 2980
+    },
+    {
+      "epoch": 0.26074823406296327,
+      "grad_norm": 3.5495529174804688,
+      "learning_rate": 1.099683238372464e-05,
+      "loss": 1.9311,
+      "mean_token_accuracy": 0.5619740694761276,
+      "step": 2990
+    },
+    {
+      "epoch": 0.2616203017354147,
+      "grad_norm": 3.613942861557007,
+      "learning_rate": 1.0936233951741052e-05,
+      "loss": 1.8817,
+      "mean_token_accuracy": 0.5683586090803147,
+      "step": 3000
+    },
+    {
+      "epoch": 0.2616203017354147,
+      "eval_loss": 1.9661142826080322,
+      "eval_mean_token_accuracy": 0.5583828481760892,
+      "eval_runtime": 2.9264,
+      "eval_samples_per_second": 374.861,
+      "eval_steps_per_second": 93.971,
+      "step": 3000
+    },
+    {
+      "epoch": 0.2624923694078661,
+      "grad_norm": 3.504915475845337,
+      "learning_rate": 1.0875600815449624e-05,
+      "loss": 1.8344,
+      "mean_token_accuracy": 0.5795254379510879,
+      "step": 3010
+    },
+    {
+      "epoch": 0.2633644370803174,
+      "grad_norm": 3.4914259910583496,
+      "learning_rate": 1.081493522239866e-05,
+      "loss": 1.9344,
+      "mean_token_accuracy": 0.5580357179045677,
+      "step": 3020
+    },
+    {
+      "epoch": 0.26423650475276883,
+      "grad_norm": 3.5396957397460938,
+      "learning_rate": 1.075423942133957e-05,
+      "loss": 1.8811,
+      "mean_token_accuracy": 0.5675024509429931,
+      "step": 3030
+    },
+    {
+      "epoch": 0.2651085724252202,
+      "grad_norm": 3.6427605152130127,
+      "learning_rate": 1.0693515662143505e-05,
+      "loss": 1.9655,
+      "mean_token_accuracy": 0.5564823850989342,
+      "step": 3040
+    },
+    {
+      "epoch": 0.2659806400976716,
+      "grad_norm": 3.4619534015655518,
+      "learning_rate": 1.0632766195717979e-05,
+      "loss": 1.9342,
+      "mean_token_accuracy": 0.5576932489871979,
+      "step": 3050
+    },
+    {
+      "epoch": 0.266852707770123,
+      "grad_norm": 3.4332313537597656,
+      "learning_rate": 1.0571993273923412e-05,
+      "loss": 1.9118,
+      "mean_token_accuracy": 0.5620474547147751,
+      "step": 3060
+    },
+    {
+      "epoch": 0.26772477544257434,
+      "grad_norm": 3.5375900268554688,
+      "learning_rate": 1.0511199149489673e-05,
+      "loss": 1.8932,
+      "mean_token_accuracy": 0.5664138972759247,
+      "step": 3070
+    },
+    {
+      "epoch": 0.26859684311502574,
+      "grad_norm": 3.6080849170684814,
+      "learning_rate": 1.0450386075932571e-05,
+      "loss": 1.9305,
+      "mean_token_accuracy": 0.5603351294994354,
+      "step": 3080
+    },
+    {
+      "epoch": 0.2694689107874771,
+      "grad_norm": 3.5486888885498047,
+      "learning_rate": 1.0389556307470316e-05,
+      "loss": 1.9195,
+      "mean_token_accuracy": 0.5560909971594811,
+      "step": 3090
+    },
+    {
+      "epoch": 0.2703409784599285,
+      "grad_norm": 3.491741180419922,
+      "learning_rate": 1.0328712098939968e-05,
+      "loss": 1.9961,
+      "mean_token_accuracy": 0.5517000988125801,
+      "step": 3100
+    },
+    {
+      "epoch": 0.2703409784599285,
+      "eval_loss": 1.960378646850586,
+      "eval_mean_token_accuracy": 0.5584895927255804,
+      "eval_runtime": 2.9238,
+      "eval_samples_per_second": 375.195,
+      "eval_steps_per_second": 94.055,
+      "step": 3100
+    },
+    {
+      "epoch": 0.2712130461323799,
+      "grad_norm": 3.4540164470672607,
+      "learning_rate": 1.0267855705713854e-05,
+      "loss": 1.8456,
+      "mean_token_accuracy": 0.5719544976949692,
+      "step": 3110
+    },
+    {
+      "epoch": 0.27208511380483125,
+      "grad_norm": 3.432096481323242,
+      "learning_rate": 1.020698938361595e-05,
+      "loss": 1.9826,
+      "mean_token_accuracy": 0.5504770070314408,
+      "step": 3120
+    },
+    {
+      "epoch": 0.27295718147728265,
+      "grad_norm": 3.5643863677978516,
+      "learning_rate": 1.0146115388838293e-05,
+      "loss": 1.8509,
+      "mean_token_accuracy": 0.5684075355529785,
+      "step": 3130
+    },
+    {
+      "epoch": 0.273829249149734,
+      "grad_norm": 3.715280771255493,
+      "learning_rate": 1.0085235977857322e-05,
+      "loss": 1.8567,
+      "mean_token_accuracy": 0.5716120332479477,
+      "step": 3140
+    },
+    {
+      "epoch": 0.2747013168221854,
+      "grad_norm": 3.485016107559204,
+      "learning_rate": 1.002435340735024e-05,
+      "loss": 1.8961,
+      "mean_token_accuracy": 0.5611546009778976,
+      "step": 3150
+    },
+    {
+      "epoch": 0.2755733844946368,
+      "grad_norm": 3.383882999420166,
+      "learning_rate": 9.963469934111374e-06,
+      "loss": 1.886,
+      "mean_token_accuracy": 0.5693737775087356,
+      "step": 3160
+    },
+    {
+      "epoch": 0.27644545216708816,
+      "grad_norm": 3.463372230529785,
+      "learning_rate": 9.90258781496851e-06,
+      "loss": 1.9954,
+      "mean_token_accuracy": 0.5521037191152572,
+      "step": 3170
+    },
+    {
+      "epoch": 0.27731751983953956,
+      "grad_norm": 3.567781448364258,
+      "learning_rate": 9.841709306699245e-06,
+      "loss": 1.9702,
+      "mean_token_accuracy": 0.5553449124097825,
+      "step": 3180
+    },
+    {
+      "epoch": 0.2781895875119909,
+      "grad_norm": 3.5508370399475098,
+      "learning_rate": 9.78083666594732e-06,
+      "loss": 1.9926,
+      "mean_token_accuracy": 0.5544453486800194,
+      "step": 3190
+    },
+    {
+      "epoch": 0.2790616551844423,
+      "grad_norm": 3.380674362182617,
+      "learning_rate": 9.719972149138985e-06,
+      "loss": 1.9118,
+      "mean_token_accuracy": 0.5659735828638077,
+      "step": 3200
+    },
+    {
+      "epoch": 0.2790616551844423,
+      "eval_loss": 1.9584650993347168,
+      "eval_mean_token_accuracy": 0.5593257433717901,
+      "eval_runtime": 2.9195,
+      "eval_samples_per_second": 375.745,
+      "eval_steps_per_second": 94.193,
+      "step": 3200
+    },
+    {
+      "epoch": 0.2799337228568937,
+      "grad_norm": 3.5802674293518066,
+      "learning_rate": 9.659118012399352e-06,
+      "loss": 1.9545,
+      "mean_token_accuracy": 0.5542318969964981,
+      "step": 3210
+    },
+    {
+      "epoch": 0.28080579052934507,
+      "grad_norm": 3.5325872898101807,
+      "learning_rate": 9.598276511468763e-06,
+      "loss": 1.85,
+      "mean_token_accuracy": 0.5688845425844192,
+      "step": 3220
+    },
+    {
+      "epoch": 0.28167785820179647,
+      "grad_norm": 3.710597276687622,
+      "learning_rate": 9.537449901619174e-06,
+      "loss": 1.957,
+      "mean_token_accuracy": 0.5496086061000824,
+      "step": 3230
+    },
+    {
+      "epoch": 0.2825499258742478,
+      "grad_norm": 3.276914119720459,
+      "learning_rate": 9.476640437570562e-06,
+      "loss": 2.0204,
+      "mean_token_accuracy": 0.55030577480793,
+      "step": 3240
+    },
+    {
+      "epoch": 0.2834219935466992,
+      "grad_norm": 3.453503131866455,
+      "learning_rate": 9.415850373407342e-06,
+      "loss": 1.9207,
+      "mean_token_accuracy": 0.5621453061699867,
+      "step": 3250
+    },
+    {
+      "epoch": 0.28429406121915063,
+      "grad_norm": 3.4946365356445312,
+      "learning_rate": 9.355081962494815e-06,
+      "loss": 1.8539,
+      "mean_token_accuracy": 0.5743909686803818,
+      "step": 3260
+    },
+    {
+      "epoch": 0.285166128891602,
+      "grad_norm": 3.4860126972198486,
+      "learning_rate": 9.294337457395638e-06,
+      "loss": 1.8991,
+      "mean_token_accuracy": 0.565924659371376,
+      "step": 3270
+    },
+    {
+      "epoch": 0.2860381965640534,
+      "grad_norm": 3.5775158405303955,
+      "learning_rate": 9.233619109786332e-06,
+      "loss": 1.9027,
+      "mean_token_accuracy": 0.5682240694761276,
+      "step": 3280
+    },
+    {
+      "epoch": 0.28691026423650473,
+      "grad_norm": 3.5719025135040283,
+      "learning_rate": 9.172929170373804e-06,
+      "loss": 2.0101,
+      "mean_token_accuracy": 0.5500856161117553,
+      "step": 3290
+    },
+    {
+      "epoch": 0.28778233190895613,
+      "grad_norm": 3.4548802375793457,
+      "learning_rate": 9.112269888811934e-06,
+      "loss": 1.9406,
+      "mean_token_accuracy": 0.5581824868917465,
+      "step": 3300
+    },
+    {
+      "epoch": 0.28778233190895613,
+      "eval_loss": 1.9543726444244385,
+      "eval_mean_token_accuracy": 0.5591620698842136,
+      "eval_runtime": 2.9266,
+      "eval_samples_per_second": 374.839,
+      "eval_steps_per_second": 93.966,
+      "step": 3300
+    },
+    {
+      "epoch": 0.28865439958140754,
+      "grad_norm": 3.490755319595337,
+      "learning_rate": 9.051643513618176e-06,
+      "loss": 1.9306,
+      "mean_token_accuracy": 0.5628302425146103,
+      "step": 3310
+    },
+    {
+      "epoch": 0.2895264672538589,
+      "grad_norm": 3.5130374431610107,
+      "learning_rate": 8.99105229209021e-06,
+      "loss": 1.9957,
+      "mean_token_accuracy": 0.5466242671012879,
+      "step": 3320
+    },
+    {
+      "epoch": 0.2903985349263103,
+      "grad_norm": 3.5959911346435547,
+      "learning_rate": 8.930498470222641e-06,
+      "loss": 2.0241,
+      "mean_token_accuracy": 0.5463184952735901,
+      "step": 3330
+    },
+    {
+      "epoch": 0.29127060259876164,
+      "grad_norm": 3.674541711807251,
+      "learning_rate": 8.86998429262374e-06,
+      "loss": 1.9156,
+      "mean_token_accuracy": 0.5589408040046692,
+      "step": 3340
+    },
+    {
+      "epoch": 0.29214267027121305,
+      "grad_norm": 3.2657454013824463,
+      "learning_rate": 8.809512002432252e-06,
+      "loss": 1.8872,
+      "mean_token_accuracy": 0.5673189699649811,
+      "step": 3350
+    },
+    {
+      "epoch": 0.29301473794366445,
+      "grad_norm": 3.573329210281372,
+      "learning_rate": 8.749083841234235e-06,
+      "loss": 1.9929,
+      "mean_token_accuracy": 0.5539995074272156,
+      "step": 3360
+    },
+    {
+      "epoch": 0.2938868056161158,
+      "grad_norm": 3.3816514015197754,
+      "learning_rate": 8.688702048979974e-06,
+      "loss": 1.9497,
+      "mean_token_accuracy": 0.5583170235157013,
+      "step": 3370
+    },
+    {
+      "epoch": 0.2947588732885672,
+      "grad_norm": 3.573122501373291,
+      "learning_rate": 8.628368863900954e-06,
+      "loss": 1.9527,
+      "mean_token_accuracy": 0.555503910779953,
+      "step": 3380
+    },
+    {
+      "epoch": 0.29563094096101855,
+      "grad_norm": 3.5306241512298584,
+      "learning_rate": 8.568086522426884e-06,
+      "loss": 1.9747,
+      "mean_token_accuracy": 0.5539139002561569,
+      "step": 3390
+    },
+    {
+      "epoch": 0.29650300863346996,
+      "grad_norm": 3.3187403678894043,
+      "learning_rate": 8.507857259102814e-06,
+      "loss": 1.9246,
+      "mean_token_accuracy": 0.5613135993480682,
+      "step": 3400
+    },
+    {
+      "epoch": 0.29650300863346996,
+      "eval_loss": 1.950748324394226,
+      "eval_mean_token_accuracy": 0.5607116177949039,
+      "eval_runtime": 2.9149,
+      "eval_samples_per_second": 376.345,
+      "eval_steps_per_second": 94.344,
+      "step": 3400
+    },
+    {
+      "epoch": 0.29737507630592136,
+      "grad_norm": 3.4003677368164062,
+      "learning_rate": 8.447683306506279e-06,
+      "loss": 1.949,
+      "mean_token_accuracy": 0.5543052822351455,
+      "step": 3410
+    },
+    {
+      "epoch": 0.2982471439783727,
+      "grad_norm": 3.533676862716675,
+      "learning_rate": 8.387566895164566e-06,
+      "loss": 1.8897,
+      "mean_token_accuracy": 0.5588307231664658,
+      "step": 3420
+    },
+    {
+      "epoch": 0.2991192116508241,
+      "grad_norm": 3.4712741374969482,
+      "learning_rate": 8.327510253472023e-06,
+      "loss": 1.986,
+      "mean_token_accuracy": 0.5482020542025566,
+      "step": 3430
+    },
+    {
+      "epoch": 0.29999127932327546,
+      "grad_norm": 3.523334264755249,
+      "learning_rate": 8.267515607607458e-06,
+      "loss": 1.9021,
+      "mean_token_accuracy": 0.5668786644935608,
+      "step": 3440
+    },
+    {
+      "epoch": 0.30086334699572687,
+      "grad_norm": 3.501821279525757,
+      "learning_rate": 8.207585181451611e-06,
+      "loss": 1.9227,
+      "mean_token_accuracy": 0.5580357164144516,
+      "step": 3450
+    },
+    {
+      "epoch": 0.30173541466817827,
+      "grad_norm": 3.640744686126709,
+      "learning_rate": 8.147721196504736e-06,
+      "loss": 1.8961,
+      "mean_token_accuracy": 0.5631115436553955,
+      "step": 3460
+    },
+    {
+      "epoch": 0.3026074823406296,
+      "grad_norm": 3.525653123855591,
+      "learning_rate": 8.08792587180424e-06,
+      "loss": 2.0234,
+      "mean_token_accuracy": 0.5500489205121994,
+      "step": 3470
+    },
+    {
+      "epoch": 0.303479550013081,
+      "grad_norm": 3.3722457885742188,
+      "learning_rate": 8.028201423842437e-06,
+      "loss": 1.9279,
+      "mean_token_accuracy": 0.5553204596042634,
+      "step": 3480
+    },
+    {
+      "epoch": 0.30435161768553237,
+      "grad_norm": 3.527557611465454,
+      "learning_rate": 7.96855006648438e-06,
+      "loss": 1.8945,
+      "mean_token_accuracy": 0.5665239721536637,
+      "step": 3490
+    },
+    {
+      "epoch": 0.3052236853579838,
+      "grad_norm": 3.6900181770324707,
+      "learning_rate": 7.908974010885795e-06,
+      "loss": 2.0003,
+      "mean_token_accuracy": 0.5520670220255852,
+      "step": 3500
+    },
+    {
+      "epoch": 0.3052236853579838,
+      "eval_loss": 1.9474492073059082,
+      "eval_mean_token_accuracy": 0.5610016028447584,
+      "eval_runtime": 2.9183,
+      "eval_samples_per_second": 375.9,
+      "eval_steps_per_second": 94.232,
+      "step": 3500
+    },
+    {
+      "epoch": 0.3060957530304352,
+      "grad_norm": 3.3773767948150635,
+      "learning_rate": 7.849475465411136e-06,
+      "loss": 1.8863,
+      "mean_token_accuracy": 0.5664505988359452,
+      "step": 3510
+    },
+    {
+      "epoch": 0.30696782070288653,
+      "grad_norm": 3.6281330585479736,
+      "learning_rate": 7.790056635551704e-06,
+      "loss": 1.9699,
+      "mean_token_accuracy": 0.5561276972293854,
+      "step": 3520
+    },
+    {
+      "epoch": 0.30783988837533793,
+      "grad_norm": 3.6639270782470703,
+      "learning_rate": 7.730719723843903e-06,
+      "loss": 1.993,
+      "mean_token_accuracy": 0.5522749543190002,
+      "step": 3530
+    },
+    {
+      "epoch": 0.3087119560477893,
+      "grad_norm": 3.28005313873291,
+      "learning_rate": 7.671466929787598e-06,
+      "loss": 1.8948,
+      "mean_token_accuracy": 0.5606531292200089,
+      "step": 3540
+    },
+    {
+      "epoch": 0.3095840237202407,
+      "grad_norm": 3.7010374069213867,
+      "learning_rate": 7.61230044976458e-06,
+      "loss": 1.9899,
+      "mean_token_accuracy": 0.5536692753434181,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3104560913926921,
+      "grad_norm": 3.6204214096069336,
+      "learning_rate": 7.553222476957157e-06,
+      "loss": 1.9665,
+      "mean_token_accuracy": 0.5597969591617584,
+      "step": 3560
+    },
+    {
+      "epoch": 0.31132815906514344,
+      "grad_norm": 3.453937530517578,
+      "learning_rate": 7.494235201266849e-06,
+      "loss": 1.9408,
+      "mean_token_accuracy": 0.5594667345285416,
+      "step": 3570
+    },
+    {
+      "epoch": 0.31220022673759484,
+      "grad_norm": 3.5678353309631348,
+      "learning_rate": 7.435340809233218e-06,
+      "loss": 1.9738,
+      "mean_token_accuracy": 0.5543595537543297,
+      "step": 3580
+    },
+    {
+      "epoch": 0.3130722944100462,
+      "grad_norm": 3.684084892272949,
+      "learning_rate": 7.376541483952811e-06,
+      "loss": 2.0072,
+      "mean_token_accuracy": 0.555748525261879,
+      "step": 3590
+    },
+    {
+      "epoch": 0.3139443620824976,
+      "grad_norm": 3.332475423812866,
+      "learning_rate": 7.3178394049982485e-06,
+      "loss": 1.901,
+      "mean_token_accuracy": 0.5641022503376008,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3139443620824976,
+      "eval_loss": 1.9425283670425415,
+      "eval_mean_token_accuracy": 0.5618946792862632,
+      "eval_runtime": 2.9216,
+      "eval_samples_per_second": 375.476,
+      "eval_steps_per_second": 94.126,
+      "step": 3600
+    },
+    {
+      "epoch": 0.314816429754949,
+      "grad_norm": 3.4494006633758545,
+      "learning_rate": 7.259236748337421e-06,
+      "loss": 1.9467,
+      "mean_token_accuracy": 0.5637230962514878,
+      "step": 3610
+    },
+    {
+      "epoch": 0.31568849742740035,
+      "grad_norm": 3.4810264110565186,
+      "learning_rate": 7.20073568625284e-06,
+      "loss": 1.9812,
+      "mean_token_accuracy": 0.5533512741327286,
+      "step": 3620
+    },
+    {
+      "epoch": 0.31656056509985175,
+      "grad_norm": 3.605604648590088,
+      "learning_rate": 7.1423383872611045e-06,
+      "loss": 1.9705,
+      "mean_token_accuracy": 0.5557607665657998,
+      "step": 3630
+    },
+    {
+      "epoch": 0.31743263277230316,
+      "grad_norm": 3.363556385040283,
+      "learning_rate": 7.084047016032528e-06,
+      "loss": 1.9949,
+      "mean_token_accuracy": 0.5512720212340355,
+      "step": 3640
+    },
+    {
+      "epoch": 0.3183047004447545,
+      "grad_norm": 3.438685417175293,
+      "learning_rate": 7.025863733310894e-06,
+      "loss": 1.9139,
+      "mean_token_accuracy": 0.5619740679860115,
+      "step": 3650
+    },
+    {
+      "epoch": 0.3191767681172059,
+      "grad_norm": 3.4539272785186768,
+      "learning_rate": 6.967790695833363e-06,
+      "loss": 1.9857,
+      "mean_token_accuracy": 0.5497553870081902,
+      "step": 3660
+    },
+    {
+      "epoch": 0.32004883578965726,
+      "grad_norm": 3.3196065425872803,
+      "learning_rate": 6.909830056250527e-06,
+      "loss": 1.9075,
+      "mean_token_accuracy": 0.561442819237709,
+      "step": 3670
+    },
+    {
+      "epoch": 0.32092090346210866,
+      "grad_norm": 3.4837331771850586,
+      "learning_rate": 6.851983963046612e-06,
+      "loss": 1.9351,
+      "mean_token_accuracy": 0.5539872765541076,
+      "step": 3680
+    },
+    {
+      "epoch": 0.32179297113456007,
+      "grad_norm": 3.6272096633911133,
+      "learning_rate": 6.794254560459843e-06,
+      "loss": 1.9501,
+      "mean_token_accuracy": 0.5567759245634079,
+      "step": 3690
+    },
+    {
+      "epoch": 0.3226650388070114,
+      "grad_norm": 3.5059802532196045,
+      "learning_rate": 6.736643988402958e-06,
+      "loss": 1.9247,
+      "mean_token_accuracy": 0.5564823910593987,
+      "step": 3700
+    },
+    {
+      "epoch": 0.3226650388070114,
+      "eval_loss": 1.9405269622802734,
+      "eval_mean_token_accuracy": 0.5619106932119889,
+      "eval_runtime": 2.923,
+      "eval_samples_per_second": 375.294,
+      "eval_steps_per_second": 94.08,
+      "step": 3700
+    },
+    {
+      "epoch": 0.3235371064794628,
+      "grad_norm": 3.372619867324829,
+      "learning_rate": 6.679154382383883e-06,
+      "loss": 2.031,
+      "mean_token_accuracy": 0.547174659371376,
+      "step": 3710
+    },
+    {
+      "epoch": 0.32440917415191417,
+      "grad_norm": 3.5121965408325195,
+      "learning_rate": 6.621787873426581e-06,
+      "loss": 1.9949,
+      "mean_token_accuracy": 0.5450342565774917,
+      "step": 3720
+    },
+    {
+      "epoch": 0.3252812418243656,
+      "grad_norm": 3.432105302810669,
+      "learning_rate": 6.564546587992054e-06,
+      "loss": 1.9286,
+      "mean_token_accuracy": 0.5592955037951469,
+      "step": 3730
+    },
+    {
+      "epoch": 0.326153309496817,
+      "grad_norm": 3.5460870265960693,
+      "learning_rate": 6.507432647899519e-06,
+      "loss": 2.0162,
+      "mean_token_accuracy": 0.5443126171827316,
+      "step": 3740
+    },
+    {
+      "epoch": 0.3270253771692683,
+      "grad_norm": 3.261183261871338,
+      "learning_rate": 6.450448170247757e-06,
+      "loss": 1.9033,
+      "mean_token_accuracy": 0.5611545979976654,
+      "step": 3750
+    },
+    {
+      "epoch": 0.32789744484171973,
+      "grad_norm": 3.391395330429077,
+      "learning_rate": 6.393595267336639e-06,
+      "loss": 1.9013,
+      "mean_token_accuracy": 0.5633561611175537,
+      "step": 3760
+    },
+    {
+      "epoch": 0.3287695125141711,
+      "grad_norm": 3.458225965499878,
+      "learning_rate": 6.3368760465888226e-06,
+      "loss": 2.0028,
+      "mean_token_accuracy": 0.5453277856111527,
+      "step": 3770
+    },
+    {
+      "epoch": 0.3296415801866225,
+      "grad_norm": 3.6255977153778076,
+      "learning_rate": 6.280292610471639e-06,
+      "loss": 2.0021,
+      "mean_token_accuracy": 0.5492294445633888,
+      "step": 3780
+    },
+    {
+      "epoch": 0.3305136478590739,
+      "grad_norm": 3.650642156600952,
+      "learning_rate": 6.223847056419154e-06,
+      "loss": 1.9389,
+      "mean_token_accuracy": 0.553938353061676,
+      "step": 3790
+    },
+    {
+      "epoch": 0.33138571553152524,
+      "grad_norm": 3.583211660385132,
+      "learning_rate": 6.1675414767544285e-06,
+      "loss": 1.9802,
+      "mean_token_accuracy": 0.5481531336903572,
+      "step": 3800
+    },
+    {
+      "epoch": 0.33138571553152524,
+      "eval_loss": 1.9368196725845337,
+      "eval_mean_token_accuracy": 0.5624088236418637,
+      "eval_runtime": 2.9172,
+      "eval_samples_per_second": 376.043,
+      "eval_steps_per_second": 94.268,
+      "step": 3800
+    },
+    {
+      "epoch": 0.33225778320397664,
+      "grad_norm": 3.617384910583496,
+      "learning_rate": 6.111377958611948e-06,
+      "loss": 1.9292,
+      "mean_token_accuracy": 0.5568126231431961,
+      "step": 3810
+    },
+    {
+      "epoch": 0.333129850876428,
+      "grad_norm": 3.447025775909424,
+      "learning_rate": 6.055358583860267e-06,
+      "loss": 1.9406,
+      "mean_token_accuracy": 0.5563233852386474,
+      "step": 3820
+    },
+    {
+      "epoch": 0.3340019185488794,
+      "grad_norm": 3.4000298976898193,
+      "learning_rate": 5.99948542902483e-06,
+      "loss": 1.8873,
+      "mean_token_accuracy": 0.5638943254947663,
+      "step": 3830
+    },
+    {
+      "epoch": 0.3348739862213308,
+      "grad_norm": 3.3708372116088867,
+      "learning_rate": 5.943760565211011e-06,
+      "loss": 1.9725,
+      "mean_token_accuracy": 0.549449609220028,
+      "step": 3840
+    },
+    {
+      "epoch": 0.33574605389378215,
+      "grad_norm": 3.6719422340393066,
+      "learning_rate": 5.8881860580273285e-06,
+      "loss": 2.004,
+      "mean_token_accuracy": 0.5501467779278755,
+      "step": 3850
+    },
+    {
+      "epoch": 0.33661812156623355,
+      "grad_norm": 3.4273903369903564,
+      "learning_rate": 5.832763967508885e-06,
+      "loss": 2.0799,
+      "mean_token_accuracy": 0.5349804297089576,
+      "step": 3860
+    },
+    {
+      "epoch": 0.3374901892386849,
+      "grad_norm": 3.4551119804382324,
+      "learning_rate": 5.777496348041009e-06,
+      "loss": 1.8994,
+      "mean_token_accuracy": 0.56736790984869,
+      "step": 3870
+    },
+    {
+      "epoch": 0.3383622569111363,
+      "grad_norm": 3.470658302307129,
+      "learning_rate": 5.722385248283092e-06,
+      "loss": 1.9951,
+      "mean_token_accuracy": 0.5519936352968215,
+      "step": 3880
+    },
+    {
+      "epoch": 0.3392343245835877,
+      "grad_norm": 3.6004528999328613,
+      "learning_rate": 5.667432711092651e-06,
+      "loss": 1.967,
+      "mean_token_accuracy": 0.5478351324796676,
+      "step": 3890
+    },
+    {
+      "epoch": 0.34010639225603906,
+      "grad_norm": 3.6642818450927734,
+      "learning_rate": 5.61264077344962e-06,
+      "loss": 1.9877,
+      "mean_token_accuracy": 0.552972112596035,
+      "step": 3900
+    },
+    {
+      "epoch": 0.34010639225603906,
+      "eval_loss": 1.9334731101989746,
+      "eval_mean_token_accuracy": 0.5631951603022489,
+      "eval_runtime": 2.925,
+      "eval_samples_per_second": 375.047,
+      "eval_steps_per_second": 94.018,
+      "step": 3900
+    },
+    {
+      "epoch": 0.34097845992849046,
+      "grad_norm": 3.564228057861328,
+      "learning_rate": 5.558011466380824e-06,
+      "loss": 1.9183,
+      "mean_token_accuracy": 0.5551736742258072,
+      "step": 3910
+    },
+    {
+      "epoch": 0.3418505276009418,
+      "grad_norm": 3.4676644802093506,
+      "learning_rate": 5.5035468148846926e-06,
+      "loss": 1.939,
+      "mean_token_accuracy": 0.555406066775322,
+      "step": 3920
+    },
+    {
+      "epoch": 0.3427225952733932,
+      "grad_norm": 3.4094276428222656,
+      "learning_rate": 5.449248837856224e-06,
+      "loss": 1.9417,
+      "mean_token_accuracy": 0.5605797469615936,
+      "step": 3930
+    },
+    {
+      "epoch": 0.3435946629458446,
+      "grad_norm": 3.64096999168396,
+      "learning_rate": 5.395119548012112e-06,
+      "loss": 2.0108,
+      "mean_token_accuracy": 0.5471012726426124,
+      "step": 3940
+    },
+    {
+      "epoch": 0.34446673061829597,
+      "grad_norm": 3.3370494842529297,
+      "learning_rate": 5.34116095181616e-06,
+      "loss": 1.9591,
+      "mean_token_accuracy": 0.5542685821652412,
+      "step": 3950
+    },
+    {
+      "epoch": 0.3453387982907474,
+      "grad_norm": 3.464301347732544,
+      "learning_rate": 5.287375049404909e-06,
+      "loss": 1.9257,
+      "mean_token_accuracy": 0.5600415796041489,
+      "step": 3960
+    },
+    {
+      "epoch": 0.3462108659631987,
+      "grad_norm": 3.4731807708740234,
+      "learning_rate": 5.233763834513479e-06,
+      "loss": 1.903,
+      "mean_token_accuracy": 0.5626223087310791,
+      "step": 3970
+    },
+    {
+      "epoch": 0.3470829336356501,
+      "grad_norm": 3.3717103004455566,
+      "learning_rate": 5.180329294401685e-06,
+      "loss": 1.9075,
+      "mean_token_accuracy": 0.5634662359952927,
+      "step": 3980
+    },
+    {
+      "epoch": 0.34795500130810153,
+      "grad_norm": 3.5404505729675293,
+      "learning_rate": 5.127073409780352e-06,
+      "loss": 1.9671,
+      "mean_token_accuracy": 0.5509540200233459,
+      "step": 3990
+    },
+    {
+      "epoch": 0.3488270689805529,
+      "grad_norm": 3.4941515922546387,
+      "learning_rate": 5.0739981547379215e-06,
+      "loss": 1.9638,
+      "mean_token_accuracy": 0.5524951100349427,
+      "step": 4000
+    },
+    {
+      "epoch": 0.3488270689805529,
+      "eval_loss": 1.9310775995254517,
+      "eval_mean_token_accuracy": 0.5636452576247128,
+      "eval_runtime": 2.9211,
+      "eval_samples_per_second": 375.548,
+      "eval_steps_per_second": 94.144,
+      "step": 4000
+    },
+    {
+      "epoch": 0.3496991366530043,
+      "grad_norm": 3.378528356552124,
+      "learning_rate": 5.02110549666724e-06,
+      "loss": 1.9438,
+      "mean_token_accuracy": 0.5565190821886062,
+      "step": 4010
+    },
+    {
+      "epoch": 0.35057120432545563,
+      "grad_norm": 3.267061233520508,
+      "learning_rate": 4.968397396192675e-06,
+      "loss": 1.9108,
+      "mean_token_accuracy": 0.5627323880791664,
+      "step": 4020
+    },
+    {
+      "epoch": 0.35144327199790704,
+      "grad_norm": 3.4503090381622314,
+      "learning_rate": 4.91587580709739e-06,
+      "loss": 1.9805,
+      "mean_token_accuracy": 0.5561521515250206,
+      "step": 4030
+    },
+    {
+      "epoch": 0.35231533967035844,
+      "grad_norm": 3.533662796020508,
+      "learning_rate": 4.863542676250972e-06,
+      "loss": 1.8535,
+      "mean_token_accuracy": 0.5740704536437988,
+      "step": 4040
+    },
+    {
+      "epoch": 0.3531874073428098,
+      "grad_norm": 3.2673568725585938,
+      "learning_rate": 4.811399943537223e-06,
+      "loss": 1.9656,
+      "mean_token_accuracy": 0.5502690821886063,
+      "step": 4050
+    },
+    {
+      "epoch": 0.3540594750152612,
+      "grad_norm": 3.3562867641448975,
+      "learning_rate": 4.759449541782272e-06,
+      "loss": 1.966,
+      "mean_token_accuracy": 0.5559564545750618,
+      "step": 4060
+    },
+    {
+      "epoch": 0.35493154268771254,
+      "grad_norm": 3.2444992065429688,
+      "learning_rate": 4.707693396682936e-06,
+      "loss": 1.9658,
+      "mean_token_accuracy": 0.5579256355762482,
+      "step": 4070
+    },
+    {
+      "epoch": 0.35580361036016395,
+      "grad_norm": 3.517874002456665,
+      "learning_rate": 4.656133426735315e-06,
+      "loss": 1.9999,
+      "mean_token_accuracy": 0.5522627159953117,
+      "step": 4080
+    },
+    {
+      "epoch": 0.35667567803261535,
+      "grad_norm": 3.4915339946746826,
+      "learning_rate": 4.604771543163706e-06,
+      "loss": 1.9666,
+      "mean_token_accuracy": 0.5488992139697075,
+      "step": 4090
+    },
+    {
+      "epoch": 0.3575477457050667,
+      "grad_norm": 3.39562726020813,
+      "learning_rate": 4.5536096498497295e-06,
+      "loss": 2.0099,
+      "mean_token_accuracy": 0.5444593831896782,
+      "step": 4100
+    },
+    {
+      "epoch": 0.3575477457050667,
+      "eval_loss": 1.9278814792633057,
+      "eval_mean_token_accuracy": 0.5640989132360978,
+      "eval_runtime": 2.9224,
+      "eval_samples_per_second": 375.372,
+      "eval_steps_per_second": 94.1,
+      "step": 4100
+    },
+    {
+      "epoch": 0.3584198133775181,
+      "grad_norm": 3.524139642715454,
+      "learning_rate": 4.502649643261779e-06,
+      "loss": 1.9278,
+      "mean_token_accuracy": 0.557497552037239,
+      "step": 4110
+    },
+    {
+      "epoch": 0.35929188104996945,
+      "grad_norm": 3.4198966026306152,
+      "learning_rate": 4.451893412384707e-06,
+      "loss": 2.0413,
+      "mean_token_accuracy": 0.5419765144586564,
+      "step": 4120
+    },
+    {
+      "epoch": 0.36016394872242086,
+      "grad_norm": 3.544069766998291,
+      "learning_rate": 4.401342838649818e-06,
+      "loss": 1.9111,
+      "mean_token_accuracy": 0.567049902677536,
+      "step": 4130
+    },
+    {
+      "epoch": 0.36103601639487226,
+      "grad_norm": 3.5295345783233643,
+      "learning_rate": 4.350999795865109e-06,
+      "loss": 1.936,
+      "mean_token_accuracy": 0.5618273019790649,
+      "step": 4140
+    },
+    {
+      "epoch": 0.3619080840673236,
+      "grad_norm": 3.356889009475708,
+      "learning_rate": 4.300866150145837e-06,
+      "loss": 1.9347,
+      "mean_token_accuracy": 0.5592832654714585,
+      "step": 4150
+    },
+    {
+      "epoch": 0.362780151739775,
+      "grad_norm": 3.4614856243133545,
+      "learning_rate": 4.250943759845316e-06,
+      "loss": 1.9643,
+      "mean_token_accuracy": 0.5566658496856689,
+      "step": 4160
+    },
+    {
+      "epoch": 0.36365221941222636,
+      "grad_norm": 3.4342923164367676,
+      "learning_rate": 4.201234475486063e-06,
+      "loss": 1.9389,
+      "mean_token_accuracy": 0.556360074877739,
+      "step": 4170
+    },
+    {
+      "epoch": 0.36452428708467777,
+      "grad_norm": 3.4406120777130127,
+      "learning_rate": 4.1517401396911725e-06,
+      "loss": 2.033,
+      "mean_token_accuracy": 0.5399951085448265,
+      "step": 4180
+    },
+    {
+      "epoch": 0.36539635475712917,
+      "grad_norm": 3.4231979846954346,
+      "learning_rate": 4.1024625871160325e-06,
+      "loss": 1.9199,
+      "mean_token_accuracy": 0.5610567629337311,
+      "step": 4190
+    },
+    {
+      "epoch": 0.3662684224295805,
+      "grad_norm": 3.2534894943237305,
+      "learning_rate": 4.053403644380321e-06,
+      "loss": 1.9269,
+      "mean_token_accuracy": 0.5566169306635856,
+      "step": 4200
+    },
+    {
+      "epoch": 0.3662684224295805,
+      "eval_loss": 1.9247615337371826,
+      "eval_mean_token_accuracy": 0.564203877990896,
+      "eval_runtime": 2.9184,
+      "eval_samples_per_second": 375.887,
+      "eval_steps_per_second": 94.229,
+      "step": 4200
+    },
+    {
+      "epoch": 0.3671404901020319,
+      "grad_norm": 3.5127272605895996,
+      "learning_rate": 4.004565130000277e-06,
+      "loss": 1.9657,
+      "mean_token_accuracy": 0.5480919778347015,
+      "step": 4210
+    },
+    {
+      "epoch": 0.3680125577744833,
+      "grad_norm": 3.278109073638916,
+      "learning_rate": 3.955948854321321e-06,
+      "loss": 2.0058,
+      "mean_token_accuracy": 0.5498776912689209,
+      "step": 4220
+    },
+    {
+      "epoch": 0.3688846254469347,
+      "grad_norm": 3.2456653118133545,
+      "learning_rate": 3.907556619450909e-06,
+      "loss": 1.9616,
+      "mean_token_accuracy": 0.5565557777881622,
+      "step": 4230
+    },
+    {
+      "epoch": 0.3697566931193861,
+      "grad_norm": 3.573273181915283,
+      "learning_rate": 3.859390219191775e-06,
+      "loss": 1.9816,
+      "mean_token_accuracy": 0.5487646847963333,
+      "step": 4240
+    },
+    {
+      "epoch": 0.37062876079183743,
+      "grad_norm": 3.2841413021087646,
+      "learning_rate": 3.8114514389754098e-06,
+      "loss": 1.9496,
+      "mean_token_accuracy": 0.5528742641210556,
+      "step": 4250
+    },
+    {
+      "epoch": 0.37150082846428883,
+      "grad_norm": 3.324763536453247,
+      "learning_rate": 3.7637420557958927e-06,
+      "loss": 1.9767,
+      "mean_token_accuracy": 0.5552592933177948,
+      "step": 4260
+    },
+    {
+      "epoch": 0.3723728961367402,
+      "grad_norm": 3.451409339904785,
+      "learning_rate": 3.7162638381440077e-06,
+      "loss": 1.9541,
+      "mean_token_accuracy": 0.5569593936204911,
+      "step": 4270
+    },
+    {
+      "epoch": 0.3732449638091916,
+      "grad_norm": 3.5858285427093506,
+      "learning_rate": 3.6690185459417107e-06,
+      "loss": 1.979,
+      "mean_token_accuracy": 0.5532656595110893,
+      "step": 4280
+    },
+    {
+      "epoch": 0.374117031481643,
+      "grad_norm": 3.5080277919769287,
+      "learning_rate": 3.622007930476865e-06,
+      "loss": 1.8988,
+      "mean_token_accuracy": 0.5693003922700882,
+      "step": 4290
+    },
+    {
+      "epoch": 0.37498909915409434,
+      "grad_norm": 3.3986520767211914,
+      "learning_rate": 3.575233734338356e-06,
+      "loss": 2.0099,
+      "mean_token_accuracy": 0.5445205435156822,
+      "step": 4300
+    },
+    {
+      "epoch": 0.37498909915409434,
+      "eval_loss": 1.9234017133712769,
+      "eval_mean_token_accuracy": 0.5648923686417666,
+      "eval_runtime": 2.9232,
+      "eval_samples_per_second": 375.268,
+      "eval_steps_per_second": 94.074,
+      "step": 4300
+    },
+    {
+      "epoch": 0.37586116682654575,
+      "grad_norm": 3.5340077877044678,
+      "learning_rate": 3.528697691351465e-06,
+      "loss": 2.0236,
+      "mean_token_accuracy": 0.541927595436573,
+      "step": 4310
+    },
+    {
+      "epoch": 0.37673323449899715,
+      "grad_norm": 3.5860188007354736,
+      "learning_rate": 3.4824015265136278e-06,
+      "loss": 2.0175,
+      "mean_token_accuracy": 0.5457436382770539,
+      "step": 4320
+    },
+    {
+      "epoch": 0.3776053021714485,
+      "grad_norm": 3.366842269897461,
+      "learning_rate": 3.436346955930472e-06,
+      "loss": 1.9546,
+      "mean_token_accuracy": 0.5558708429336547,
+      "step": 4330
+    },
+    {
+      "epoch": 0.3784773698438999,
+      "grad_norm": 3.4738569259643555,
+      "learning_rate": 3.3905356867522187e-06,
+      "loss": 2.0651,
+      "mean_token_accuracy": 0.5355063647031784,
+      "step": 4340
+    },
+    {
+      "epoch": 0.37934943751635125,
+      "grad_norm": 3.444784641265869,
+      "learning_rate": 3.344969417110391e-06,
+      "loss": 1.9937,
+      "mean_token_accuracy": 0.5502813145518303,
+      "step": 4350
+    },
+    {
+      "epoch": 0.38022150518880266,
+      "grad_norm": 3.4755406379699707,
+      "learning_rate": 3.29964983605487e-06,
+      "loss": 2.0533,
+      "mean_token_accuracy": 0.5411570399999619,
+      "step": 4360
+    },
+    {
+      "epoch": 0.38109357286125406,
+      "grad_norm": 3.371807098388672,
+      "learning_rate": 3.2545786234913e-06,
+      "loss": 1.9416,
+      "mean_token_accuracy": 0.558255872130394,
+      "step": 4370
+    },
+    {
+      "epoch": 0.3819656405337054,
+      "grad_norm": 3.4360344409942627,
+      "learning_rate": 3.2097574501187877e-06,
+      "loss": 1.9863,
+      "mean_token_accuracy": 0.552299416065216,
+      "step": 4380
+    },
+    {
+      "epoch": 0.3828377082061568,
+      "grad_norm": 3.42404842376709,
+      "learning_rate": 3.165187977368007e-06,
+      "loss": 1.9325,
+      "mean_token_accuracy": 0.5581947147846222,
+      "step": 4390
+    },
+    {
+      "epoch": 0.38370977587860816,
+      "grad_norm": 3.378279447555542,
+      "learning_rate": 3.120871857339582e-06,
+      "loss": 2.0528,
+      "mean_token_accuracy": 0.5382338553667069,
+      "step": 4400
+    },
+    {
+      "epoch": 0.38370977587860816,
+      "eval_loss": 1.9232269525527954,
+      "eval_mean_token_accuracy": 0.5646041637117213,
+      "eval_runtime": 2.924,
+      "eval_samples_per_second": 375.165,
+      "eval_steps_per_second": 94.048,
+      "step": 4400
+    },
+    {
+      "epoch": 0.38458184355105957,
+      "grad_norm": 3.5511367321014404,
+      "learning_rate": 3.0768107327428766e-06,
+      "loss": 1.9826,
+      "mean_token_accuracy": 0.5598214328289032,
+      "step": 4410
+    },
+    {
+      "epoch": 0.38545391122351097,
+      "grad_norm": 3.271724224090576,
+      "learning_rate": 3.033006236835071e-06,
+      "loss": 1.9475,
+      "mean_token_accuracy": 0.5505381613969803,
+      "step": 4420
+    },
+    {
+      "epoch": 0.3863259788959623,
+      "grad_norm": 3.6280934810638428,
+      "learning_rate": 2.9894599933606518e-06,
+      "loss": 1.9555,
+      "mean_token_accuracy": 0.5552348285913468,
+      "step": 4430
+    },
+    {
+      "epoch": 0.3871980465684137,
+      "grad_norm": 3.547792434692383,
+      "learning_rate": 2.9461736164911934e-06,
+      "loss": 1.8814,
+      "mean_token_accuracy": 0.5680773049592972,
+      "step": 4440
+    },
+    {
+      "epoch": 0.38807011424086507,
+      "grad_norm": 3.7517290115356445,
+      "learning_rate": 2.903148710765552e-06,
+      "loss": 1.9916,
+      "mean_token_accuracy": 0.5446917816996575,
+      "step": 4450
+    },
+    {
+      "epoch": 0.3889421819133165,
+      "grad_norm": 3.359938144683838,
+      "learning_rate": 2.8603868710303662e-06,
+      "loss": 1.9303,
+      "mean_token_accuracy": 0.5600293546915054,
+      "step": 4460
+    },
+    {
+      "epoch": 0.3898142495857679,
+      "grad_norm": 3.4032561779022217,
+      "learning_rate": 2.8178896823809465e-06,
+      "loss": 1.9941,
+      "mean_token_accuracy": 0.5448263198137283,
+      "step": 4470
+    },
+    {
+      "epoch": 0.39068631725821923,
+      "grad_norm": 3.420640468597412,
+      "learning_rate": 2.7756587201025297e-06,
+      "loss": 1.9851,
+      "mean_token_accuracy": 0.5504525452852249,
+      "step": 4480
+    },
+    {
+      "epoch": 0.39155838493067063,
+      "grad_norm": 3.4940996170043945,
+      "learning_rate": 2.7336955496118666e-06,
+      "loss": 2.0182,
+      "mean_token_accuracy": 0.5449119374155998,
+      "step": 4490
+    },
+    {
+      "epoch": 0.392430452603122,
+      "grad_norm": 3.4239795207977295,
+      "learning_rate": 2.692001726399215e-06,
+      "loss": 2.0,
+      "mean_token_accuracy": 0.5462695762515068,
+      "step": 4500
+    },
+    {
+      "epoch": 0.392430452603122,
+      "eval_loss": 1.9196184873580933,
+      "eval_mean_token_accuracy": 0.5653442435914819,
+      "eval_runtime": 2.9231,
+      "eval_samples_per_second": 375.29,
+      "eval_steps_per_second": 94.079,
+      "step": 4500
+    },
+    {
+      "epoch": 0.3933025202755734,
+      "grad_norm": 3.2954492568969727,
+      "learning_rate": 2.6505787959706607e-06,
+      "loss": 1.9041,
+      "mean_token_accuracy": 0.5592465728521347,
+      "step": 4510
+    },
+    {
+      "epoch": 0.3941745879480248,
+      "grad_norm": 3.3611555099487305,
+      "learning_rate": 2.609428293790852e-06,
+      "loss": 1.8998,
+      "mean_token_accuracy": 0.5617661386728287,
+      "step": 4520
+    },
+    {
+      "epoch": 0.39504665562047614,
+      "grad_norm": 3.3805503845214844,
+      "learning_rate": 2.5685517452260566e-06,
+      "loss": 2.0286,
+      "mean_token_accuracy": 0.5424902200698852,
+      "step": 4530
+    },
+    {
+      "epoch": 0.39591872329292754,
+      "grad_norm": 3.2891135215759277,
+      "learning_rate": 2.5279506654876473e-06,
+      "loss": 1.9389,
+      "mean_token_accuracy": 0.5557363063097001,
+      "step": 4540
+    },
+    {
+      "epoch": 0.3967907909653789,
+      "grad_norm": 3.500875473022461,
+      "learning_rate": 2.487626559575911e-06,
+      "loss": 1.9972,
+      "mean_token_accuracy": 0.5473703518509865,
+      "step": 4550
+    },
+    {
+      "epoch": 0.3976628586378303,
+      "grad_norm": 3.4249260425567627,
+      "learning_rate": 2.4475809222242775e-06,
+      "loss": 1.9999,
+      "mean_token_accuracy": 0.5520792543888092,
+      "step": 4560
+    },
+    {
+      "epoch": 0.3985349263102817,
+      "grad_norm": 3.4105777740478516,
+      "learning_rate": 2.4078152378439033e-06,
+      "loss": 1.9919,
+      "mean_token_accuracy": 0.5443615436553955,
+      "step": 4570
+    },
+    {
+      "epoch": 0.39940699398273305,
+      "grad_norm": 3.5384693145751953,
+      "learning_rate": 2.3683309804686604e-06,
+      "loss": 1.9395,
+      "mean_token_accuracy": 0.5591242611408234,
+      "step": 4580
+    },
+    {
+      "epoch": 0.40027906165518445,
+      "grad_norm": 3.435582160949707,
+      "learning_rate": 2.329129613700478e-06,
+      "loss": 2.0292,
+      "mean_token_accuracy": 0.5349804326891899,
+      "step": 4590
+    },
+    {
+      "epoch": 0.4011511293276358,
+      "grad_norm": 3.4797873497009277,
+      "learning_rate": 2.29021259065511e-06,
+      "loss": 1.9614,
+      "mean_token_accuracy": 0.5492661446332932,
+      "step": 4600
+    },
+    {
+      "epoch": 0.4011511293276358,
+      "eval_loss": 1.9179856777191162,
+      "eval_mean_token_accuracy": 0.5653495817834681,
+      "eval_runtime": 2.9245,
+      "eval_samples_per_second": 375.104,
+      "eval_steps_per_second": 94.032,
+      "step": 4600
+    },
+    {
+      "epoch": 0.4020231970000872,
+      "grad_norm": 3.3572614192962646,
+      "learning_rate": 2.251581353908252e-06,
+      "loss": 1.9783,
+      "mean_token_accuracy": 0.5447040110826492,
+      "step": 4610
+    },
+    {
+      "epoch": 0.4028952646725386,
+      "grad_norm": 3.5049843788146973,
+      "learning_rate": 2.2132373354420833e-06,
+      "loss": 1.9872,
+      "mean_token_accuracy": 0.5443126171827316,
+      "step": 4620
+    },
+    {
+      "epoch": 0.40376733234498996,
+      "grad_norm": 3.4441332817077637,
+      "learning_rate": 2.1751819565921774e-06,
+      "loss": 2.0432,
+      "mean_token_accuracy": 0.541340509057045,
+      "step": 4630
+    },
+    {
+      "epoch": 0.40463940001744136,
+      "grad_norm": 3.340221405029297,
+      "learning_rate": 2.137416627994814e-06,
+      "loss": 1.9943,
+      "mean_token_accuracy": 0.5530332699418068,
+      "step": 4640
+    },
+    {
+      "epoch": 0.4055114676898927,
+      "grad_norm": 3.2400784492492676,
+      "learning_rate": 2.0999427495347035e-06,
+      "loss": 2.01,
+      "mean_token_accuracy": 0.5498043030500412,
+      "step": 4650
+    },
+    {
+      "epoch": 0.4063835353623441,
+      "grad_norm": 3.424387216567993,
+      "learning_rate": 2.0627617102930753e-06,
+      "loss": 2.0281,
+      "mean_token_accuracy": 0.5404109537601471,
+      "step": 4660
+    },
+    {
+      "epoch": 0.4072556030347955,
+      "grad_norm": 3.28721284866333,
+      "learning_rate": 2.02587488849621e-06,
+      "loss": 1.9801,
+      "mean_token_accuracy": 0.5500489249825478,
+      "step": 4670
+    },
+    {
+      "epoch": 0.40812767070724687,
+      "grad_norm": 3.367316484451294,
+      "learning_rate": 1.989283651464329e-06,
+      "loss": 2.0021,
+      "mean_token_accuracy": 0.5436888486146927,
+      "step": 4680
+    },
+    {
+      "epoch": 0.4089997383796983,
+      "grad_norm": 3.267263412475586,
+      "learning_rate": 1.952989355560929e-06,
+      "loss": 1.9627,
+      "mean_token_accuracy": 0.5574241697788238,
+      "step": 4690
+    },
+    {
+      "epoch": 0.4098718060521496,
+      "grad_norm": 3.582045316696167,
+      "learning_rate": 1.9169933461424928e-06,
+      "loss": 1.9974,
+      "mean_token_accuracy": 0.5412426590919495,
+      "step": 4700
+    },
+    {
+      "epoch": 0.4098718060521496,
+      "eval_loss": 1.9162505865097046,
+      "eval_mean_token_accuracy": 0.5657605405287309,
+      "eval_runtime": 2.9237,
+      "eval_samples_per_second": 375.211,
+      "eval_steps_per_second": 94.059,
+      "step": 4700
+    },
+    {
+      "epoch": 0.410743873724601,
+      "grad_norm": 3.4188454151153564,
+      "learning_rate": 1.8812969575086272e-06,
+      "loss": 1.9726,
+      "mean_token_accuracy": 0.551504397392273,
+      "step": 4710
+    },
+    {
+      "epoch": 0.41161594139705243,
+      "grad_norm": 3.4676272869110107,
+      "learning_rate": 1.8459015128525937e-06,
+      "loss": 1.8693,
+      "mean_token_accuracy": 0.566059198975563,
+      "step": 4720
+    },
+    {
+      "epoch": 0.4124880090695038,
+      "grad_norm": 3.5259809494018555,
+      "learning_rate": 1.8108083242122764e-06,
+      "loss": 1.9644,
+      "mean_token_accuracy": 0.5478351294994355,
+      "step": 4730
+    },
+    {
+      "epoch": 0.4133600767419552,
+      "grad_norm": 3.391449451446533,
+      "learning_rate": 1.7760186924215239e-06,
+      "loss": 1.9035,
+      "mean_token_accuracy": 0.5650929600000382,
+      "step": 4740
+    },
+    {
+      "epoch": 0.41423214441440653,
+      "grad_norm": 3.447322130203247,
+      "learning_rate": 1.7415339070619586e-06,
+      "loss": 1.9984,
+      "mean_token_accuracy": 0.5452177107334137,
+      "step": 4750
+    },
+    {
+      "epoch": 0.41510421208685794,
+      "grad_norm": 3.4332034587860107,
+      "learning_rate": 1.7073552464151465e-06,
+      "loss": 2.0024,
+      "mean_token_accuracy": 0.5492172211408615,
+      "step": 4760
+    },
+    {
+      "epoch": 0.41597627975930934,
+      "grad_norm": 3.423743724822998,
+      "learning_rate": 1.6734839774152322e-06,
+      "loss": 2.0312,
+      "mean_token_accuracy": 0.5438967660069466,
+      "step": 4770
+    },
+    {
+      "epoch": 0.4168483474317607,
+      "grad_norm": 3.5208423137664795,
+      "learning_rate": 1.6399213556019732e-06,
+      "loss": 1.9309,
+      "mean_token_accuracy": 0.5570694729685783,
+      "step": 4780
+    },
+    {
+      "epoch": 0.4177204151042121,
+      "grad_norm": 3.5019121170043945,
+      "learning_rate": 1.6066686250741904e-06,
+      "loss": 2.0032,
+      "mean_token_accuracy": 0.5540484368801117,
+      "step": 4790
+    },
+    {
+      "epoch": 0.41859248277666344,
+      "grad_norm": 3.454352378845215,
+      "learning_rate": 1.573727018443667e-06,
+      "loss": 2.0285,
+      "mean_token_accuracy": 0.545413401722908,
+      "step": 4800
+    },
+    {
+      "epoch": 0.41859248277666344,
+      "eval_loss": 1.9150649309158325,
+      "eval_mean_token_accuracy": 0.5659366666186939,
+      "eval_runtime": 2.9235,
+      "eval_samples_per_second": 375.239,
+      "eval_steps_per_second": 94.066,
+      "step": 4800
+    },
+    {
+      "epoch": 0.41946455044911485,
+      "grad_norm": 3.410337448120117,
+      "learning_rate": 1.5410977567894403e-06,
+      "loss": 2.0063,
+      "mean_token_accuracy": 0.540398733317852,
+      "step": 4810
+    },
+    {
+      "epoch": 0.42033661812156625,
+      "grad_norm": 3.482410430908203,
+      "learning_rate": 1.5087820496125595e-06,
+      "loss": 1.9998,
+      "mean_token_accuracy": 0.5470034271478653,
+      "step": 4820
+    },
+    {
+      "epoch": 0.4212086857940176,
+      "grad_norm": 3.451167106628418,
+      "learning_rate": 1.4767810947912275e-06,
+      "loss": 1.9847,
+      "mean_token_accuracy": 0.5454134061932564,
+      "step": 4830
+    },
+    {
+      "epoch": 0.422080753466469,
+      "grad_norm": 3.5304558277130127,
+      "learning_rate": 1.4450960785364244e-06,
+      "loss": 2.0033,
+      "mean_token_accuracy": 0.5418909087777137,
+      "step": 4840
+    },
+    {
+      "epoch": 0.42295282113892035,
+      "grad_norm": 3.4220824241638184,
+      "learning_rate": 1.4137281753479092e-06,
+      "loss": 2.0273,
+      "mean_token_accuracy": 0.5440680056810379,
+      "step": 4850
+    },
+    {
+      "epoch": 0.42382488881137176,
+      "grad_norm": 3.5942747592926025,
+      "learning_rate": 1.3826785479707128e-06,
+      "loss": 2.0392,
+      "mean_token_accuracy": 0.542685903608799,
+      "step": 4860
+    },
+    {
+      "epoch": 0.42469695648382316,
+      "grad_norm": 3.4738378524780273,
+      "learning_rate": 1.3519483473520124e-06,
+      "loss": 2.0468,
+      "mean_token_accuracy": 0.535640898346901,
+      "step": 4870
+    },
+    {
+      "epoch": 0.4255690241562745,
+      "grad_norm": 3.283612012863159,
+      "learning_rate": 1.3215387125984813e-06,
+      "loss": 2.0134,
+      "mean_token_accuracy": 0.5466731935739517,
+      "step": 4880
+    },
+    {
+      "epoch": 0.4264410918287259,
+      "grad_norm": 3.293135166168213,
+      "learning_rate": 1.2914507709340596e-06,
+      "loss": 1.9607,
+      "mean_token_accuracy": 0.5518835619091987,
+      "step": 4890
+    },
+    {
+      "epoch": 0.42731315950117726,
+      "grad_norm": 3.3204004764556885,
+      "learning_rate": 1.2616856376581766e-06,
+      "loss": 2.0239,
+      "mean_token_accuracy": 0.5454011723399163,
+      "step": 4900
+    },
+    {
+      "epoch": 0.42731315950117726,
+      "eval_loss": 1.9135085344314575,
+      "eval_mean_token_accuracy": 0.5663013691251928,
+      "eval_runtime": 2.925,
+      "eval_samples_per_second": 375.045,
+      "eval_steps_per_second": 94.018,
+      "step": 4900
+    },
+    {
+      "epoch": 0.42818522717362867,
+      "grad_norm": 3.3719985485076904,
+      "learning_rate": 1.2322444161044e-06,
+      "loss": 2.0532,
+      "mean_token_accuracy": 0.5419398248195648,
+      "step": 4910
+    },
+    {
+      "epoch": 0.4290572948460801,
+      "grad_norm": 3.436063528060913,
+      "learning_rate": 1.2031281975995467e-06,
+      "loss": 1.9894,
+      "mean_token_accuracy": 0.5477617412805558,
+      "step": 4920
+    },
+    {
+      "epoch": 0.4299293625185314,
+      "grad_norm": 3.8871910572052,
+      "learning_rate": 1.1743380614232213e-06,
+      "loss": 2.0185,
+      "mean_token_accuracy": 0.5433830797672272,
+      "step": 4930
+    },
+    {
+      "epoch": 0.4308014301909828,
+      "grad_norm": 3.3078415393829346,
+      "learning_rate": 1.1458750747678105e-06,
+      "loss": 2.017,
+      "mean_token_accuracy": 0.5541952073574066,
+      "step": 4940
+    },
+    {
+      "epoch": 0.43167349786343423,
+      "grad_norm": 3.4217162132263184,
+      "learning_rate": 1.1177402926989345e-06,
+      "loss": 2.0053,
+      "mean_token_accuracy": 0.5484833657741547,
+      "step": 4950
+    },
+    {
+      "epoch": 0.4325455655358856,
+      "grad_norm": 3.2682292461395264,
+      "learning_rate": 1.0899347581163222e-06,
+      "loss": 2.0031,
+      "mean_token_accuracy": 0.5520058676600457,
+      "step": 4960
+    },
+    {
+      "epoch": 0.433417633208337,
+      "grad_norm": 3.4017956256866455,
+      "learning_rate": 1.0624595017151685e-06,
+      "loss": 2.0148,
+      "mean_token_accuracy": 0.5396159529685974,
+      "step": 4970
+    },
+    {
+      "epoch": 0.43428970088078833,
+      "grad_norm": 3.444763660430908,
+      "learning_rate": 1.0353155419479122e-06,
+      "loss": 2.0059,
+      "mean_token_accuracy": 0.5469911903142929,
+      "step": 4980
+    },
+    {
+      "epoch": 0.43516176855323974,
+      "grad_norm": 3.51785945892334,
+      "learning_rate": 1.0085038849865025e-06,
+      "loss": 2.0157,
+      "mean_token_accuracy": 0.5444104745984077,
+      "step": 4990
+    },
+    {
+      "epoch": 0.43603383622569114,
+      "grad_norm": 3.4006400108337402,
+      "learning_rate": 9.820255246850853e-07,
+      "loss": 1.9873,
+      "mean_token_accuracy": 0.5495841488242149,
+      "step": 5000
+    },
+    {
+      "epoch": 0.43603383622569114,
+      "eval_loss": 1.9125958681106567,
+      "eval_mean_token_accuracy": 0.5662017419121482,
+      "eval_runtime": 2.923,
+      "eval_samples_per_second": 375.294,
+      "eval_steps_per_second": 94.08,
+      "step": 5000
+    },
+    {
+      "epoch": 0.4369059038981425,
+      "grad_norm": 3.2100887298583984,
+      "learning_rate": 9.55881442543174e-07,
+      "loss": 2.0069,
+      "mean_token_accuracy": 0.545768091082573,
+      "step": 5010
+    },
+    {
+      "epoch": 0.4377779715705939,
+      "grad_norm": 3.3258049488067627,
+      "learning_rate": 9.30072607669259e-07,
+      "loss": 2.0494,
+      "mean_token_accuracy": 0.5419642895460128,
+      "step": 5020
+    },
+    {
+      "epoch": 0.43865003924304524,
+      "grad_norm": 3.4492013454437256,
+      "learning_rate": 9.045999767448988e-07,
+      "loss": 1.9828,
+      "mean_token_accuracy": 0.552849805355072,
+      "step": 5030
+    },
+    {
+      "epoch": 0.43952210691549665,
+      "grad_norm": 3.288825511932373,
+      "learning_rate": 8.794644939892361e-07,
+      "loss": 2.051,
+      "mean_token_accuracy": 0.5415239736437798,
+      "step": 5040
+    },
+    {
+      "epoch": 0.44039417458794805,
+      "grad_norm": 3.3537158966064453,
+      "learning_rate": 8.546670911240196e-07,
+      "loss": 2.0521,
+      "mean_token_accuracy": 0.5437133058905601,
+      "step": 5050
+    },
+    {
+      "epoch": 0.4412662422603994,
+      "grad_norm": 3.2526566982269287,
+      "learning_rate": 8.302086873390536e-07,
+      "loss": 2.018,
+      "mean_token_accuracy": 0.5421477481722832,
+      "step": 5060
+    },
+    {
+      "epoch": 0.4421383099328508,
+      "grad_norm": 3.7663211822509766,
+      "learning_rate": 8.060901892581241e-07,
+      "loss": 2.0251,
+      "mean_token_accuracy": 0.5435543030500412,
+      "step": 5070
+    },
+    {
+      "epoch": 0.44301037760530215,
+      "grad_norm": 3.2456564903259277,
+      "learning_rate": 7.82312490905407e-07,
+      "loss": 1.9813,
+      "mean_token_accuracy": 0.5500122308731079,
+      "step": 5080
+    },
+    {
+      "epoch": 0.44388244527775356,
+      "grad_norm": 3.383213758468628,
+      "learning_rate": 7.588764736723086e-07,
+      "loss": 2.0193,
+      "mean_token_accuracy": 0.5431139975786209,
+      "step": 5090
+    },
+    {
+      "epoch": 0.44475451295020496,
+      "grad_norm": 3.3814773559570312,
+      "learning_rate": 7.357830062848114e-07,
+      "loss": 1.9817,
+      "mean_token_accuracy": 0.547174659371376,
+      "step": 5100
+    },
+    {
+      "epoch": 0.44475451295020496,
+      "eval_loss": 1.9117127656936646,
+      "eval_mean_token_accuracy": 0.5664259019764987,
+      "eval_runtime": 2.9209,
+      "eval_samples_per_second": 375.571,
+      "eval_steps_per_second": 94.15,
+      "step": 5100
+    },
+    {
+      "epoch": 0.4456265806226563,
+      "grad_norm": 3.335890531539917,
+      "learning_rate": 7.130329447712581e-07,
+      "loss": 1.8672,
+      "mean_token_accuracy": 0.5681139975786209,
+      "step": 5110
+    },
+    {
+      "epoch": 0.4464986482951077,
+      "grad_norm": 3.3268182277679443,
+      "learning_rate": 6.906271324306335e-07,
+      "loss": 1.9851,
+      "mean_token_accuracy": 0.5408145874738693,
+      "step": 5120
+    },
+    {
+      "epoch": 0.44737071596755906,
+      "grad_norm": 3.679361343383789,
+      "learning_rate": 6.685663998012926e-07,
+      "loss": 2.0024,
+      "mean_token_accuracy": 0.5448752403259277,
+      "step": 5130
+    },
+    {
+      "epoch": 0.44824278364001047,
+      "grad_norm": 3.4542551040649414,
+      "learning_rate": 6.468515646301865e-07,
+      "loss": 1.9959,
+      "mean_token_accuracy": 0.5455724149942398,
+      "step": 5140
+    },
+    {
+      "epoch": 0.44911485131246187,
+      "grad_norm": 3.5810275077819824,
+      "learning_rate": 6.254834318425363e-07,
+      "loss": 2.0205,
+      "mean_token_accuracy": 0.5432485401630401,
+      "step": 5150
+    },
+    {
+      "epoch": 0.4499869189849132,
+      "grad_norm": 3.1930737495422363,
+      "learning_rate": 6.044627935120107e-07,
+      "loss": 1.9691,
+      "mean_token_accuracy": 0.5506726995110511,
+      "step": 5160
+    },
+    {
+      "epoch": 0.4508589866573646,
+      "grad_norm": 3.267045021057129,
+      "learning_rate": 5.837904288313545e-07,
+      "loss": 1.9781,
+      "mean_token_accuracy": 0.548813596367836,
+      "step": 5170
+    },
+    {
+      "epoch": 0.451731054329816,
+      "grad_norm": 3.275087356567383,
+      "learning_rate": 5.634671040835104e-07,
+      "loss": 2.0532,
+      "mean_token_accuracy": 0.5378424674272537,
+      "step": 5180
+    },
+    {
+      "epoch": 0.4526031220022674,
+      "grad_norm": 3.5153698921203613,
+      "learning_rate": 5.43493572613214e-07,
+      "loss": 2.0904,
+      "mean_token_accuracy": 0.5347602739930153,
+      "step": 5190
+    },
+    {
+      "epoch": 0.4534751896747188,
+      "grad_norm": 3.451738119125366,
+      "learning_rate": 5.238705747990669e-07,
+      "loss": 1.9553,
+      "mean_token_accuracy": 0.5565679997205735,
+      "step": 5200
+    },
+    {
+      "epoch": 0.4534751896747188,
+      "eval_loss": 1.9111390113830566,
+      "eval_mean_token_accuracy": 0.566555772044442,
+      "eval_runtime": 2.9256,
+      "eval_samples_per_second": 374.961,
+      "eval_steps_per_second": 93.996,
+      "step": 5200
+    },
+    {
+      "epoch": 0.45434725734717013,
+      "grad_norm": 3.4063291549682617,
+      "learning_rate": 5.045988380260935e-07,
+      "loss": 1.94,
+      "mean_token_accuracy": 0.5548067554831505,
+      "step": 5210
+    },
+    {
+      "epoch": 0.45521932501962153,
+      "grad_norm": 3.240532875061035,
+      "learning_rate": 4.856790766587815e-07,
+      "loss": 2.0801,
+      "mean_token_accuracy": 0.5411815106868744,
+      "step": 5220
+    },
+    {
+      "epoch": 0.4560913926920729,
+      "grad_norm": 3.362454414367676,
+      "learning_rate": 4.6711199201459833e-07,
+      "loss": 2.0323,
+      "mean_token_accuracy": 0.5411325857043267,
+      "step": 5230
+    },
+    {
+      "epoch": 0.4569634603645243,
+      "grad_norm": 3.218104362487793,
+      "learning_rate": 4.488982723379887e-07,
+      "loss": 1.9691,
+      "mean_token_accuracy": 0.5499755412340164,
+      "step": 5240
+    },
+    {
+      "epoch": 0.4578355280369757,
+      "grad_norm": 3.2326581478118896,
+      "learning_rate": 4.3103859277488056e-07,
+      "loss": 1.9656,
+      "mean_token_accuracy": 0.5533879607915878,
+      "step": 5250
+    },
+    {
+      "epoch": 0.45870759570942704,
+      "grad_norm": 3.5729854106903076,
+      "learning_rate": 4.1353361534763657e-07,
+      "loss": 1.9989,
+      "mean_token_accuracy": 0.545144322514534,
+      "step": 5260
+    },
+    {
+      "epoch": 0.45957966338187844,
+      "grad_norm": 3.513415813446045,
+      "learning_rate": 3.963839889305343e-07,
+      "loss": 2.0147,
+      "mean_token_accuracy": 0.5440557703375817,
+      "step": 5270
+    },
+    {
+      "epoch": 0.4604517310543298,
+      "grad_norm": 3.471275806427002,
+      "learning_rate": 3.7959034922569804e-07,
+      "loss": 2.0305,
+      "mean_token_accuracy": 0.5384907037019729,
+      "step": 5280
+    },
+    {
+      "epoch": 0.4613237987267812,
+      "grad_norm": 3.2812275886535645,
+      "learning_rate": 3.631533187395453e-07,
+      "loss": 2.0188,
+      "mean_token_accuracy": 0.544630628824234,
+      "step": 5290
+    },
+    {
+      "epoch": 0.4621958663992326,
+      "grad_norm": 3.3048923015594482,
+      "learning_rate": 3.470735067597053e-07,
+      "loss": 2.0139,
+      "mean_token_accuracy": 0.539114487171173,
+      "step": 5300
+    },
+    {
+      "epoch": 0.4621958663992326,
+      "eval_loss": 1.9108039140701294,
+      "eval_mean_token_accuracy": 0.5665237492864782,
+      "eval_runtime": 2.9253,
+      "eval_samples_per_second": 374.998,
+      "eval_steps_per_second": 94.006,
+      "step": 5300
+    },
+    {
+      "epoch": 0.46306793407168395,
+      "grad_norm": 3.2648355960845947,
+      "learning_rate": 3.313515093324393e-07,
+      "loss": 1.9608,
+      "mean_token_accuracy": 0.554134051501751,
+      "step": 5310
+    },
+    {
+      "epoch": 0.46394000174413536,
+      "grad_norm": 3.5620362758636475,
+      "learning_rate": 3.1598790924053936e-07,
+      "loss": 1.9765,
+      "mean_token_accuracy": 0.553681506216526,
+      "step": 5320
+    },
+    {
+      "epoch": 0.4648120694165867,
+      "grad_norm": 3.5021138191223145,
+      "learning_rate": 3.009832759817344e-07,
+      "loss": 2.0703,
+      "mean_token_accuracy": 0.5320817038416863,
+      "step": 5330
+    },
+    {
+      "epoch": 0.4656841370890381,
+      "grad_norm": 3.1931092739105225,
+      "learning_rate": 2.8633816574757166e-07,
+      "loss": 1.9752,
+      "mean_token_accuracy": 0.5508561670780182,
+      "step": 5340
+    },
+    {
+      "epoch": 0.4665562047614895,
+      "grad_norm": 3.376694679260254,
+      "learning_rate": 2.720531214028055e-07,
+      "loss": 1.9525,
+      "mean_token_accuracy": 0.5547088950872421,
+      "step": 5350
+    },
+    {
+      "epoch": 0.46742827243394086,
+      "grad_norm": 3.454848289489746,
+      "learning_rate": 2.5812867246527207e-07,
+      "loss": 1.965,
+      "mean_token_accuracy": 0.5496330693364143,
+      "step": 5360
+    },
+    {
+      "epoch": 0.46830034010639227,
+      "grad_norm": 3.4802231788635254,
+      "learning_rate": 2.445653350862609e-07,
+      "loss": 2.0356,
+      "mean_token_accuracy": 0.5389187902212143,
+      "step": 5370
+    },
+    {
+      "epoch": 0.4691724077788436,
+      "grad_norm": 3.1773571968078613,
+      "learning_rate": 2.3136361203138668e-07,
+      "loss": 2.0215,
+      "mean_token_accuracy": 0.5463918775320054,
+      "step": 5380
+    },
+    {
+      "epoch": 0.470044475451295,
+      "grad_norm": 3.5094268321990967,
+      "learning_rate": 2.1852399266194312e-07,
+      "loss": 2.0033,
+      "mean_token_accuracy": 0.5482632100582123,
+      "step": 5390
+    },
+    {
+      "epoch": 0.4709165431237464,
+      "grad_norm": 3.399948835372925,
+      "learning_rate": 2.0604695291677523e-07,
+      "loss": 1.9803,
+      "mean_token_accuracy": 0.5487035229802132,
+      "step": 5400
+    },
+    {
+      "epoch": 0.4709165431237464,
+      "eval_loss": 1.9105377197265625,
+      "eval_mean_token_accuracy": 0.5665611094778235,
+      "eval_runtime": 2.9298,
+      "eval_samples_per_second": 374.431,
+      "eval_steps_per_second": 93.864,
+      "step": 5400
+    },
+    {
+      "epoch": 0.47178861079619777,
+      "grad_norm": 3.342210292816162,
+      "learning_rate": 1.9393295529462674e-07,
+      "loss": 1.9609,
+      "mean_token_accuracy": 0.5476638942956924,
+      "step": 5410
+    },
+    {
+      "epoch": 0.4726606784686492,
+      "grad_norm": 3.237065553665161,
+      "learning_rate": 1.8218244883700386e-07,
+      "loss": 2.0066,
+      "mean_token_accuracy": 0.5420743629336358,
+      "step": 5420
+    },
+    {
+      "epoch": 0.4735327461411005,
+      "grad_norm": 3.1440958976745605,
+      "learning_rate": 1.7079586911152413e-07,
+      "loss": 1.9765,
+      "mean_token_accuracy": 0.54918053150177,
+      "step": 5430
+    },
+    {
+      "epoch": 0.47440481381355193,
+      "grad_norm": 3.388202428817749,
+      "learning_rate": 1.597736381957782e-07,
+      "loss": 1.9991,
+      "mean_token_accuracy": 0.5521159470081329,
+      "step": 5440
+    },
+    {
+      "epoch": 0.47527688148600333,
+      "grad_norm": 3.486203193664551,
+      "learning_rate": 1.4911616466167345e-07,
+      "loss": 2.0179,
+      "mean_token_accuracy": 0.540080726146698,
+      "step": 5450
+    },
+    {
+      "epoch": 0.4761489491584547,
+      "grad_norm": 3.573913812637329,
+      "learning_rate": 1.3882384356030066e-07,
+      "loss": 2.064,
+      "mean_token_accuracy": 0.5342710345983506,
+      "step": 5460
+    },
+    {
+      "epoch": 0.4770210168309061,
+      "grad_norm": 3.4833734035491943,
+      "learning_rate": 1.2889705640728445e-07,
+      "loss": 1.9714,
+      "mean_token_accuracy": 0.5572162434458733,
+      "step": 5470
+    },
+    {
+      "epoch": 0.47789308450335743,
+      "grad_norm": 3.512598752975464,
+      "learning_rate": 1.1933617116863805e-07,
+      "loss": 2.0608,
+      "mean_token_accuracy": 0.5376100778579712,
+      "step": 5480
+    },
+    {
+      "epoch": 0.47876515217580884,
+      "grad_norm": 3.4204320907592773,
+      "learning_rate": 1.1014154224713302e-07,
+      "loss": 2.0016,
+      "mean_token_accuracy": 0.5501100823283196,
+      "step": 5490
+    },
+    {
+      "epoch": 0.47963721984826024,
+      "grad_norm": 3.3977138996124268,
+      "learning_rate": 1.0131351046915094e-07,
+      "loss": 1.9461,
+      "mean_token_accuracy": 0.5514065578579903,
+      "step": 5500
+    },
+    {
+      "epoch": 0.47963721984826024,
+      "eval_loss": 1.9102822542190552,
+      "eval_mean_token_accuracy": 0.5665112960338593,
+      "eval_runtime": 2.9297,
+      "eval_samples_per_second": 374.436,
+      "eval_steps_per_second": 93.865,
+      "step": 5500
+    },
+    {
+      "epoch": 0.4805092875207116,
+      "grad_norm": 3.3802883625030518,
+      "learning_rate": 9.285240307206123e-08,
+      "loss": 2.0279,
+      "mean_token_accuracy": 0.5406678050756455,
+      "step": 5510
+    },
+    {
+      "epoch": 0.481381355193163,
+      "grad_norm": 3.4572670459747314,
+      "learning_rate": 8.475853369207753e-08,
+      "loss": 2.0322,
+      "mean_token_accuracy": 0.5403987318277359,
+      "step": 5520
+    },
+    {
+      "epoch": 0.48225342286561435,
+      "grad_norm": 3.4783124923706055,
+      "learning_rate": 7.703220235264708e-08,
+      "loss": 2.0105,
+      "mean_token_accuracy": 0.5460983395576477,
+      "step": 5530
+    },
+    {
+      "epoch": 0.48312549053806575,
+      "grad_norm": 3.3147389888763428,
+      "learning_rate": 6.967369545331615e-08,
+      "loss": 2.0254,
+      "mean_token_accuracy": 0.5439579218626023,
+      "step": 5540
+    },
+    {
+      "epoch": 0.48399755821051715,
+      "grad_norm": 3.2064905166625977,
+      "learning_rate": 6.26832857591242e-08,
+      "loss": 1.9975,
+      "mean_token_accuracy": 0.5437377646565438,
+      "step": 5550
+    },
+    {
+      "epoch": 0.4848696258829685,
+      "grad_norm": 3.2711522579193115,
+      "learning_rate": 5.606123239048522e-08,
+      "loss": 1.9938,
+      "mean_token_accuracy": 0.5489863157272339,
+      "step": 5560
+    },
+    {
+      "epoch": 0.4857416935554199,
+      "grad_norm": 3.4603748321533203,
+      "learning_rate": 4.9807780813586615e-08,
+      "loss": 1.9975,
+      "mean_token_accuracy": 0.5490826830267906,
+      "step": 5570
+    },
+    {
+      "epoch": 0.48661376122787126,
+      "grad_norm": 3.377051591873169,
+      "learning_rate": 4.392316283128861e-08,
+      "loss": 1.9271,
+      "mean_token_accuracy": 0.5597358107566833,
+      "step": 5580
+    },
+    {
+      "epoch": 0.48748582890032266,
+      "grad_norm": 3.5900418758392334,
+      "learning_rate": 3.840759657453452e-08,
+      "loss": 2.0507,
+      "mean_token_accuracy": 0.5380381584167481,
+      "step": 5590
+    },
+    {
+      "epoch": 0.48835789657277406,
+      "grad_norm": 3.386613368988037,
+      "learning_rate": 3.326128649426053e-08,
+      "loss": 2.0268,
+      "mean_token_accuracy": 0.5412793532013893,
+      "step": 5600
+    },
+    {
+      "epoch": 0.48835789657277406,
+      "eval_loss": 1.9102189540863037,
+      "eval_mean_token_accuracy": 0.5665611090443351,
+      "eval_runtime": 2.9221,
+      "eval_samples_per_second": 375.413,
+      "eval_steps_per_second": 94.11,
+      "step": 5600
+    },
+    {
+      "epoch": 0.4892299642452254,
+      "grad_norm": 3.6345722675323486,
+      "learning_rate": 2.8484423353822842e-08,
+      "loss": 1.9686,
+      "mean_token_accuracy": 0.5553326800465583,
+      "step": 5610
+    },
+    {
+      "epoch": 0.4901020319176768,
+      "grad_norm": 3.407338857650757,
+      "learning_rate": 2.4077184221920068e-08,
+      "loss": 1.9136,
+      "mean_token_accuracy": 0.5628913849592209,
+      "step": 5620
+    },
+    {
+      "epoch": 0.4909740995901282,
+      "grad_norm": 3.1782562732696533,
+      "learning_rate": 2.003973246603508e-08,
+      "loss": 1.9402,
+      "mean_token_accuracy": 0.5594667285680771,
+      "step": 5630
+    },
+    {
+      "epoch": 0.49184616726257957,
+      "grad_norm": 3.424807548522949,
+      "learning_rate": 1.637221774637765e-08,
+      "loss": 2.0295,
+      "mean_token_accuracy": 0.5384662494063377,
+      "step": 5640
+    },
+    {
+      "epoch": 0.492718234935031,
+      "grad_norm": 3.358490228652954,
+      "learning_rate": 1.3074776010334466e-08,
+      "loss": 1.956,
+      "mean_token_accuracy": 0.5518835619091987,
+      "step": 5650
+    },
+    {
+      "epoch": 0.4935903026074823,
+      "grad_norm": 3.3695414066314697,
+      "learning_rate": 1.0147529487432028e-08,
+      "loss": 1.9245,
+      "mean_token_accuracy": 0.5591976523399353,
+      "step": 5660
+    },
+    {
+      "epoch": 0.4944623702799337,
+      "grad_norm": 3.355546236038208,
+      "learning_rate": 7.590586684805834e-09,
+      "loss": 2.0699,
+      "mean_token_accuracy": 0.5332314103841782,
+      "step": 5670
+    },
+    {
+      "epoch": 0.49533443795238513,
+      "grad_norm": 3.2610630989074707,
+      "learning_rate": 5.4040423831802635e-09,
+      "loss": 2.0171,
+      "mean_token_accuracy": 0.544019079208374,
+      "step": 5680
+    },
+    {
+      "epoch": 0.4962065056248365,
+      "grad_norm": 3.4051926136016846,
+      "learning_rate": 3.587977633348061e-09,
+      "loss": 1.9788,
+      "mean_token_accuracy": 0.5509295523166656,
+      "step": 5690
+    },
+    {
+      "epoch": 0.4970785732972879,
+      "grad_norm": 3.3493635654449463,
+      "learning_rate": 2.1424597531749524e-09,
+      "loss": 2.042,
+      "mean_token_accuracy": 0.5401785746216774,
+      "step": 5700
+    },
+    {
+      "epoch": 0.4970785732972879,
+      "eval_loss": 1.9101980924606323,
+      "eval_mean_token_accuracy": 0.5665913531996987,
+      "eval_runtime": 2.922,
+      "eval_samples_per_second": 375.43,
+      "eval_steps_per_second": 94.114,
+      "step": 5700
+    },
+    {
+      "epoch": 0.49795064096973923,
+      "grad_norm": 3.290015697479248,
+      "learning_rate": 1.0675423250994244e-09,
+      "loss": 1.986,
+      "mean_token_accuracy": 0.5517857164144516,
+      "step": 5710
+    },
+    {
+      "epoch": 0.49882270864219064,
+      "grad_norm": 3.5052742958068848,
+      "learning_rate": 3.6326519414431327e-10,
+      "loss": 2.0027,
+      "mean_token_accuracy": 0.5462695732712746,
+      "step": 5720
+    },
+    {
+      "epoch": 0.49969477631464204,
+      "grad_norm": 3.4694395065307617,
+      "learning_rate": 2.965446644798142e-11,
+      "loss": 2.0181,
+      "mean_token_accuracy": 0.5399584159255028,
+      "step": 5730
+    },
+    {
+      "epoch": 0.5000436033836225,
+      "mean_token_accuracy": 0.5431445650756359,
+      "step": 5734,
+      "total_flos": 2.3971976183808e+16,
+      "train_loss": 1.9497717069673088,
+      "train_runtime": 714.9694,
+      "train_samples_per_second": 128.305,
+      "train_steps_per_second": 8.02
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 5734,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.3971976183808e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}