diff --git "a/checkpoint-30000/trainer_state.json" "b/checkpoint-30000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-30000/trainer_state.json"
@@ -0,0 +1,4503 @@
+{
+  "best_metric": 3.486082077026367,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M_low_2000_6910/checkpoint-30000",
+  "epoch": 3.234501347708895,
+  "eval_steps": 1000,
+  "global_step": 30000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005390835579514825,
+      "grad_norm": 3.8312079906463623,
+      "learning_rate": 0.000276,
+      "loss": 9.0233,
+      "step": 50
+    },
+    {
+      "epoch": 0.01078167115902965,
+      "grad_norm": 3.6952943801879883,
+      "learning_rate": 0.0005759999999999999,
+      "loss": 6.9873,
+      "step": 100
+    },
+    {
+      "epoch": 0.016172506738544475,
+      "grad_norm": 1.3713243007659912,
+      "learning_rate": 0.000599702104695089,
+      "loss": 6.5693,
+      "step": 150
+    },
+    {
+      "epoch": 0.0215633423180593,
+      "grad_norm": 1.7163493633270264,
+      "learning_rate": 0.0005993783054506205,
+      "loss": 6.2599,
+      "step": 200
+    },
+    {
+      "epoch": 0.026954177897574125,
+      "grad_norm": 2.214008331298828,
+      "learning_rate": 0.0005990545062061521,
+      "loss": 6.0915,
+      "step": 250
+    },
+    {
+      "epoch": 0.03234501347708895,
+      "grad_norm": 1.0381399393081665,
+      "learning_rate": 0.0005987307069616836,
+      "loss": 5.9541,
+      "step": 300
+    },
+    {
+      "epoch": 0.03773584905660377,
+      "grad_norm": 2.0930416584014893,
+      "learning_rate": 0.0005984069077172153,
+      "loss": 5.8795,
+      "step": 350
+    },
+    {
+      "epoch": 0.0431266846361186,
+      "grad_norm": 1.43650221824646,
+      "learning_rate": 0.0005980831084727469,
+      "loss": 5.8169,
+      "step": 400
+    },
+    {
+      "epoch": 0.04851752021563342,
+      "grad_norm": 1.5032957792282104,
+      "learning_rate": 0.0005977593092282784,
+      "loss": 5.7604,
+      "step": 450
+    },
+    {
+      "epoch": 0.05390835579514825,
+      "grad_norm": 1.0408196449279785,
+      "learning_rate": 0.00059743550998381,
+      "loss": 5.654,
+      "step": 500
+    },
+    {
+      "epoch": 0.05929919137466307,
+      "grad_norm": 0.941105306148529,
+      "learning_rate": 0.0005971117107393416,
+      "loss": 5.61,
+      "step": 550
+    },
+    {
+      "epoch": 0.0646900269541779,
+      "grad_norm": 0.9246237277984619,
+      "learning_rate": 0.0005967879114948732,
+      "loss": 5.5375,
+      "step": 600
+    },
+    {
+      "epoch": 0.07008086253369272,
+      "grad_norm": 1.4164918661117554,
+      "learning_rate": 0.0005964641122504047,
+      "loss": 5.4707,
+      "step": 650
+    },
+    {
+      "epoch": 0.07547169811320754,
+      "grad_norm": 1.0245190858840942,
+      "learning_rate": 0.0005961403130059363,
+      "loss": 5.4038,
+      "step": 700
+    },
+    {
+      "epoch": 0.08086253369272237,
+      "grad_norm": 1.3740367889404297,
+      "learning_rate": 0.0005958165137614678,
+      "loss": 5.3346,
+      "step": 750
+    },
+    {
+      "epoch": 0.0862533692722372,
+      "grad_norm": 0.9941785931587219,
+      "learning_rate": 0.0005954927145169995,
+      "loss": 5.2648,
+      "step": 800
+    },
+    {
+      "epoch": 0.09164420485175202,
+      "grad_norm": 1.0040581226348877,
+      "learning_rate": 0.0005951689152725309,
+      "loss": 5.2413,
+      "step": 850
+    },
+    {
+      "epoch": 0.09703504043126684,
+      "grad_norm": 0.9949366450309753,
+      "learning_rate": 0.0005948451160280626,
+      "loss": 5.1859,
+      "step": 900
+    },
+    {
+      "epoch": 0.10242587601078167,
+      "grad_norm": 1.3581753969192505,
+      "learning_rate": 0.0005945213167835941,
+      "loss": 5.1207,
+      "step": 950
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "grad_norm": 1.019547700881958,
+      "learning_rate": 0.0005941975175391257,
+      "loss": 5.1281,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "eval_accuracy": 0.22618532423271231,
+      "eval_loss": 5.033375263214111,
+      "eval_runtime": 184.0811,
+      "eval_samples_per_second": 97.843,
+      "eval_steps_per_second": 6.117,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11320754716981132,
+      "grad_norm": 1.211574673652649,
+      "learning_rate": 0.0005938737182946572,
+      "loss": 5.058,
+      "step": 1050
+    },
+    {
+      "epoch": 0.11859838274932614,
+      "grad_norm": 1.2203134298324585,
+      "learning_rate": 0.0005935499190501888,
+      "loss": 5.0235,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12398921832884097,
+      "grad_norm": 1.1273915767669678,
+      "learning_rate": 0.0005932261198057204,
+      "loss": 5.0256,
+      "step": 1150
+    },
+    {
+      "epoch": 0.1293800539083558,
+      "grad_norm": 0.9613862037658691,
+      "learning_rate": 0.000592902320561252,
+      "loss": 4.9834,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1347708894878706,
+      "grad_norm": 0.9953334927558899,
+      "learning_rate": 0.0005925785213167835,
+      "loss": 4.9593,
+      "step": 1250
+    },
+    {
+      "epoch": 0.14016172506738545,
+      "grad_norm": 0.9287815093994141,
+      "learning_rate": 0.0005922547220723151,
+      "loss": 4.914,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14555256064690028,
+      "grad_norm": 1.027787446975708,
+      "learning_rate": 0.0005919309228278468,
+      "loss": 4.8612,
+      "step": 1350
+    },
+    {
+      "epoch": 0.1509433962264151,
+      "grad_norm": 1.167672872543335,
+      "learning_rate": 0.0005916071235833783,
+      "loss": 4.8697,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15633423180592992,
+      "grad_norm": 0.9806638956069946,
+      "learning_rate": 0.0005912833243389097,
+      "loss": 4.8264,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16172506738544473,
+      "grad_norm": 1.000881552696228,
+      "learning_rate": 0.0005909595250944414,
+      "loss": 4.8298,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16711590296495957,
+      "grad_norm": 1.1669228076934814,
+      "learning_rate": 0.000590635725849973,
+      "loss": 4.8084,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1725067385444744,
+      "grad_norm": 0.9972322583198547,
+      "learning_rate": 0.0005903119266055045,
+      "loss": 4.7888,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1778975741239892,
+      "grad_norm": 0.9380745887756348,
+      "learning_rate": 0.0005899881273610361,
+      "loss": 4.7577,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18328840970350405,
+      "grad_norm": 1.014443039894104,
+      "learning_rate": 0.0005896643281165677,
+      "loss": 4.7411,
+      "step": 1700
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 1.024464726448059,
+      "learning_rate": 0.0005893405288720993,
+      "loss": 4.6881,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1940700808625337,
+      "grad_norm": 0.9571323990821838,
+      "learning_rate": 0.0005890167296276308,
+      "loss": 4.7224,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19946091644204852,
+      "grad_norm": 1.534552812576294,
+      "learning_rate": 0.0005886929303831624,
+      "loss": 4.6661,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20485175202156333,
+      "grad_norm": 0.9052402973175049,
+      "learning_rate": 0.0005883691311386939,
+      "loss": 4.6448,
+      "step": 1900
+    },
+    {
+      "epoch": 0.21024258760107817,
+      "grad_norm": 0.7421246767044067,
+      "learning_rate": 0.0005880453318942256,
+      "loss": 4.6249,
+      "step": 1950
+    },
+    {
+      "epoch": 0.215633423180593,
+      "grad_norm": 1.0020256042480469,
+      "learning_rate": 0.0005877215326497571,
+      "loss": 4.6047,
+      "step": 2000
+    },
+    {
+      "epoch": 0.215633423180593,
+      "eval_accuracy": 0.2674557111815013,
+      "eval_loss": 4.531125068664551,
+      "eval_runtime": 183.9893,
+      "eval_samples_per_second": 97.892,
+      "eval_steps_per_second": 6.12,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2210242587601078,
+      "grad_norm": 0.8753662109375,
+      "learning_rate": 0.0005873977334052887,
+      "loss": 4.59,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22641509433962265,
+      "grad_norm": 0.942737877368927,
+      "learning_rate": 0.0005870739341608202,
+      "loss": 4.572,
+      "step": 2100
+    },
+    {
+      "epoch": 0.23180592991913745,
+      "grad_norm": 0.830838143825531,
+      "learning_rate": 0.0005867501349163519,
+      "loss": 4.5431,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2371967654986523,
+      "grad_norm": 0.854256272315979,
+      "learning_rate": 0.0005864263356718833,
+      "loss": 4.5388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24258760107816713,
+      "grad_norm": 0.9155673384666443,
+      "learning_rate": 0.000586102536427415,
+      "loss": 4.5141,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24797843665768193,
+      "grad_norm": 0.7370288372039795,
+      "learning_rate": 0.0005857787371829465,
+      "loss": 4.5086,
+      "step": 2300
+    },
+    {
+      "epoch": 0.25336927223719674,
+      "grad_norm": 1.0621047019958496,
+      "learning_rate": 0.0005854549379384781,
+      "loss": 4.4894,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2587601078167116,
+      "grad_norm": 0.8664134740829468,
+      "learning_rate": 0.0005851311386940096,
+      "loss": 4.4697,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2641509433962264,
+      "grad_norm": 0.7284068465232849,
+      "learning_rate": 0.0005848073394495412,
+      "loss": 4.4574,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2695417789757412,
+      "grad_norm": 0.9439472556114197,
+      "learning_rate": 0.0005844835402050728,
+      "loss": 4.4345,
+      "step": 2500
+    },
+    {
+      "epoch": 0.2749326145552561,
+      "grad_norm": 0.8816312551498413,
+      "learning_rate": 0.0005841597409606044,
+      "loss": 4.4046,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2803234501347709,
+      "grad_norm": 0.92274010181427,
+      "learning_rate": 0.000583835941716136,
+      "loss": 4.4191,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 0.815833568572998,
+      "learning_rate": 0.0005835121424716675,
+      "loss": 4.4034,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29110512129380056,
+      "grad_norm": 0.76717609167099,
+      "learning_rate": 0.0005831883432271992,
+      "loss": 4.3846,
+      "step": 2700
+    },
+    {
+      "epoch": 0.29649595687331537,
+      "grad_norm": 0.8561816215515137,
+      "learning_rate": 0.0005828645439827307,
+      "loss": 4.3784,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3018867924528302,
+      "grad_norm": 0.8974009156227112,
+      "learning_rate": 0.0005825407447382622,
+      "loss": 4.3356,
+      "step": 2800
+    },
+    {
+      "epoch": 0.30727762803234504,
+      "grad_norm": 0.7354640364646912,
+      "learning_rate": 0.0005822169454937938,
+      "loss": 4.3588,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31266846361185985,
+      "grad_norm": 0.686019241809845,
+      "learning_rate": 0.0005818931462493254,
+      "loss": 4.3249,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31805929919137466,
+      "grad_norm": 0.8465150594711304,
+      "learning_rate": 0.0005815693470048569,
+      "loss": 4.3293,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "grad_norm": 0.8226708769798279,
+      "learning_rate": 0.0005812455477603885,
+      "loss": 4.3215,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "eval_accuracy": 0.29736046280045647,
+      "eval_loss": 4.2467756271362305,
+      "eval_runtime": 183.8152,
+      "eval_samples_per_second": 97.984,
+      "eval_steps_per_second": 6.126,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3288409703504043,
+      "grad_norm": 0.8238357901573181,
+      "learning_rate": 0.0005809217485159201,
+      "loss": 4.3216,
+      "step": 3050
+    },
+    {
+      "epoch": 0.33423180592991913,
+      "grad_norm": 0.8355266451835632,
+      "learning_rate": 0.0005805979492714517,
+      "loss": 4.2995,
+      "step": 3100
+    },
+    {
+      "epoch": 0.33962264150943394,
+      "grad_norm": 0.829166829586029,
+      "learning_rate": 0.0005802741500269832,
+      "loss": 4.28,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3450134770889488,
+      "grad_norm": 0.8682091236114502,
+      "learning_rate": 0.0005799503507825148,
+      "loss": 4.2822,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3504043126684636,
+      "grad_norm": 0.8935713171958923,
+      "learning_rate": 0.0005796265515380463,
+      "loss": 4.2735,
+      "step": 3250
+    },
+    {
+      "epoch": 0.3557951482479784,
+      "grad_norm": 0.9404758214950562,
+      "learning_rate": 0.000579302752293578,
+      "loss": 4.2796,
+      "step": 3300
+    },
+    {
+      "epoch": 0.3611859838274933,
+      "grad_norm": 0.8041167855262756,
+      "learning_rate": 0.0005789789530491095,
+      "loss": 4.2472,
+      "step": 3350
+    },
+    {
+      "epoch": 0.3665768194070081,
+      "grad_norm": 0.8318430781364441,
+      "learning_rate": 0.0005786551538046411,
+      "loss": 4.2381,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3719676549865229,
+      "grad_norm": 0.6364179849624634,
+      "learning_rate": 0.0005783313545601726,
+      "loss": 4.235,
+      "step": 3450
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.8213059306144714,
+      "learning_rate": 0.0005780075553157043,
+      "loss": 4.2304,
+      "step": 3500
+    },
+    {
+      "epoch": 0.38274932614555257,
+      "grad_norm": 0.7218520045280457,
+      "learning_rate": 0.0005776837560712357,
+      "loss": 4.227,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3881401617250674,
+      "grad_norm": 1.0270496606826782,
+      "learning_rate": 0.0005773599568267673,
+      "loss": 4.2181,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3935309973045822,
+      "grad_norm": 0.8176056742668152,
+      "learning_rate": 0.0005770361575822989,
+      "loss": 4.2063,
+      "step": 3650
+    },
+    {
+      "epoch": 0.39892183288409705,
+      "grad_norm": 0.7931913733482361,
+      "learning_rate": 0.0005767123583378305,
+      "loss": 4.1982,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40431266846361186,
+      "grad_norm": 0.8333051800727844,
+      "learning_rate": 0.000576388559093362,
+      "loss": 4.1965,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40970350404312667,
+      "grad_norm": 0.6684727668762207,
+      "learning_rate": 0.0005760647598488936,
+      "loss": 4.1616,
+      "step": 3800
+    },
+    {
+      "epoch": 0.41509433962264153,
+      "grad_norm": 0.6943632364273071,
+      "learning_rate": 0.0005757409606044253,
+      "loss": 4.1736,
+      "step": 3850
+    },
+    {
+      "epoch": 0.42048517520215634,
+      "grad_norm": 0.7040383815765381,
+      "learning_rate": 0.0005754171613599568,
+      "loss": 4.1912,
+      "step": 3900
+    },
+    {
+      "epoch": 0.42587601078167114,
+      "grad_norm": 1.0394502878189087,
+      "learning_rate": 0.0005750933621154884,
+      "loss": 4.1593,
+      "step": 3950
+    },
+    {
+      "epoch": 0.431266846361186,
+      "grad_norm": 0.6003571152687073,
+      "learning_rate": 0.0005747695628710199,
+      "loss": 4.1622,
+      "step": 4000
+    },
+    {
+      "epoch": 0.431266846361186,
+      "eval_accuracy": 0.3116602693657203,
+      "eval_loss": 4.093387603759766,
+      "eval_runtime": 183.8164,
+      "eval_samples_per_second": 97.984,
+      "eval_steps_per_second": 6.126,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4366576819407008,
+      "grad_norm": 0.7539233565330505,
+      "learning_rate": 0.0005744457636265515,
+      "loss": 4.1653,
+      "step": 4050
+    },
+    {
+      "epoch": 0.4420485175202156,
+      "grad_norm": 0.8013771772384644,
+      "learning_rate": 0.0005741219643820831,
+      "loss": 4.1607,
+      "step": 4100
+    },
+    {
+      "epoch": 0.4474393530997305,
+      "grad_norm": 0.627700924873352,
+      "learning_rate": 0.0005737981651376146,
+      "loss": 4.1414,
+      "step": 4150
+    },
+    {
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.6890294551849365,
+      "learning_rate": 0.0005734743658931462,
+      "loss": 4.151,
+      "step": 4200
+    },
+    {
+      "epoch": 0.4582210242587601,
+      "grad_norm": 0.6331459879875183,
+      "learning_rate": 0.0005731505666486778,
+      "loss": 4.1367,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4636118598382749,
+      "grad_norm": 0.8301665186882019,
+      "learning_rate": 0.0005728267674042093,
+      "loss": 4.1411,
+      "step": 4300
+    },
+    {
+      "epoch": 0.46900269541778977,
+      "grad_norm": 0.8348872065544128,
+      "learning_rate": 0.0005725029681597409,
+      "loss": 4.1303,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4743935309973046,
+      "grad_norm": 0.8137356638908386,
+      "learning_rate": 0.0005721791689152725,
+      "loss": 4.1206,
+      "step": 4400
+    },
+    {
+      "epoch": 0.4797843665768194,
+      "grad_norm": 0.7119181752204895,
+      "learning_rate": 0.0005718553696708041,
+      "loss": 4.1205,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48517520215633425,
+      "grad_norm": 0.6306732892990112,
+      "learning_rate": 0.0005715315704263356,
+      "loss": 4.114,
+      "step": 4500
+    },
+    {
+      "epoch": 0.49056603773584906,
+      "grad_norm": 0.6889007687568665,
+      "learning_rate": 0.0005712077711818672,
+      "loss": 4.1241,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49595687331536387,
+      "grad_norm": 0.7933549284934998,
+      "learning_rate": 0.0005708839719373987,
+      "loss": 4.1195,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5013477088948787,
+      "grad_norm": 0.6817076802253723,
+      "learning_rate": 0.0005705601726929304,
+      "loss": 4.0736,
+      "step": 4650
+    },
+    {
+      "epoch": 0.5067385444743935,
+      "grad_norm": 0.9549484848976135,
+      "learning_rate": 0.0005702363734484619,
+      "loss": 4.0785,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5121293800539084,
+      "grad_norm": 0.6651914715766907,
+      "learning_rate": 0.0005699125742039935,
+      "loss": 4.0853,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5175202156334232,
+      "grad_norm": 0.6444743871688843,
+      "learning_rate": 0.000569588774959525,
+      "loss": 4.0766,
+      "step": 4800
+    },
+    {
+      "epoch": 0.522911051212938,
+      "grad_norm": 0.7128944993019104,
+      "learning_rate": 0.0005692649757150567,
+      "loss": 4.0724,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5283018867924528,
+      "grad_norm": 0.8725901246070862,
+      "learning_rate": 0.0005689411764705881,
+      "loss": 4.0761,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5336927223719676,
+      "grad_norm": 0.77891606092453,
+      "learning_rate": 0.0005686173772261197,
+      "loss": 4.0657,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "grad_norm": 0.8567548394203186,
+      "learning_rate": 0.0005682935779816514,
+      "loss": 4.0649,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "eval_accuracy": 0.32061663556115577,
+      "eval_loss": 4.002915859222412,
+      "eval_runtime": 183.9111,
+      "eval_samples_per_second": 97.933,
+      "eval_steps_per_second": 6.123,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5444743935309974,
+      "grad_norm": 0.690135657787323,
+      "learning_rate": 0.0005679697787371829,
+      "loss": 4.0711,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5498652291105122,
+      "grad_norm": 0.6203147768974304,
+      "learning_rate": 0.0005676459794927145,
+      "loss": 4.0599,
+      "step": 5100
+    },
+    {
+      "epoch": 0.555256064690027,
+      "grad_norm": 0.6774439215660095,
+      "learning_rate": 0.000567322180248246,
+      "loss": 4.0524,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5606469002695418,
+      "grad_norm": 0.8506273627281189,
+      "learning_rate": 0.0005669983810037777,
+      "loss": 4.0627,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.6191375851631165,
+      "learning_rate": 0.0005666745817593092,
+      "loss": 4.0364,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 0.6032384634017944,
+      "learning_rate": 0.0005663507825148408,
+      "loss": 4.0343,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5768194070080862,
+      "grad_norm": 0.6567820310592651,
+      "learning_rate": 0.0005660269832703723,
+      "loss": 4.0251,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5822102425876011,
+      "grad_norm": 0.6662546396255493,
+      "learning_rate": 0.0005657031840259039,
+      "loss": 4.0298,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5876010781671159,
+      "grad_norm": 0.7775598764419556,
+      "learning_rate": 0.0005653793847814355,
+      "loss": 4.0309,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5929919137466307,
+      "grad_norm": 0.7273476123809814,
+      "learning_rate": 0.000565055585536967,
+      "loss": 4.0067,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5983827493261455,
+      "grad_norm": 0.6881582140922546,
+      "learning_rate": 0.0005647317862924986,
+      "loss": 4.0335,
+      "step": 5550
+    },
+    {
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.6695816516876221,
+      "learning_rate": 0.0005644079870480302,
+      "loss": 3.9784,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6091644204851752,
+      "grad_norm": 0.7060787081718445,
+      "learning_rate": 0.0005640841878035617,
+      "loss": 4.0152,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6145552560646901,
+      "grad_norm": 0.6417899131774902,
+      "learning_rate": 0.0005637603885590933,
+      "loss": 4.0129,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6199460916442049,
+      "grad_norm": 0.5996983647346497,
+      "learning_rate": 0.0005634365893146248,
+      "loss": 4.0148,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6253369272237197,
+      "grad_norm": 0.6481319665908813,
+      "learning_rate": 0.0005631127900701565,
+      "loss": 4.0051,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6307277628032345,
+      "grad_norm": 0.6492406129837036,
+      "learning_rate": 0.000562788990825688,
+      "loss": 3.9792,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6361185983827493,
+      "grad_norm": 0.6048398613929749,
+      "learning_rate": 0.0005624651915812196,
+      "loss": 3.9929,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6415094339622641,
+      "grad_norm": 0.6991953253746033,
+      "learning_rate": 0.0005621413923367511,
+      "loss": 3.9974,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "grad_norm": 0.6229884624481201,
+      "learning_rate": 0.0005618175930922828,
+      "loss": 4.003,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "eval_accuracy": 0.3281020589613588,
+      "eval_loss": 3.9165804386138916,
+      "eval_runtime": 183.9688,
+      "eval_samples_per_second": 97.902,
+      "eval_steps_per_second": 6.121,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6522911051212938,
+      "grad_norm": 0.6105901002883911,
+      "learning_rate": 0.0005614937938478143,
+      "loss": 3.9945,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6576819407008087,
+      "grad_norm": 0.7422035932540894,
+      "learning_rate": 0.0005611699946033459,
+      "loss": 3.9877,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6630727762803235,
+      "grad_norm": 0.5779343247413635,
+      "learning_rate": 0.0005608461953588774,
+      "loss": 3.9806,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6684636118598383,
+      "grad_norm": 0.6721993684768677,
+      "learning_rate": 0.000560522396114409,
+      "loss": 3.9741,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6738544474393531,
+      "grad_norm": 0.6363776326179504,
+      "learning_rate": 0.0005601985968699405,
+      "loss": 3.9571,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6792452830188679,
+      "grad_norm": 0.6985312104225159,
+      "learning_rate": 0.0005598747976254721,
+      "loss": 3.9717,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6846361185983828,
+      "grad_norm": 0.6108534932136536,
+      "learning_rate": 0.0005595509983810038,
+      "loss": 3.9454,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6900269541778976,
+      "grad_norm": 0.6570634841918945,
+      "learning_rate": 0.0005592271991365353,
+      "loss": 3.9575,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6954177897574124,
+      "grad_norm": 0.6376566886901855,
+      "learning_rate": 0.0005589033998920669,
+      "loss": 3.973,
+      "step": 6450
+    },
+    {
+      "epoch": 0.7008086253369272,
+      "grad_norm": 0.7813366651535034,
+      "learning_rate": 0.0005585796006475984,
+      "loss": 3.9435,
+      "step": 6500
+    },
+    {
+      "epoch": 0.706199460916442,
+      "grad_norm": 0.6727055907249451,
+      "learning_rate": 0.0005582558014031301,
+      "loss": 3.9684,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7115902964959568,
+      "grad_norm": 0.6897123456001282,
+      "learning_rate": 0.0005579320021586616,
+      "loss": 3.9519,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7169811320754716,
+      "grad_norm": 0.6443628072738647,
+      "learning_rate": 0.0005576082029141932,
+      "loss": 3.948,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7223719676549866,
+      "grad_norm": 0.529731810092926,
+      "learning_rate": 0.0005572844036697247,
+      "loss": 3.9334,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7277628032345014,
+      "grad_norm": 0.589066743850708,
+      "learning_rate": 0.0005569606044252563,
+      "loss": 3.939,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7331536388140162,
+      "grad_norm": 0.9202807545661926,
+      "learning_rate": 0.0005566368051807879,
+      "loss": 3.9555,
+      "step": 6800
+    },
+    {
+      "epoch": 0.738544474393531,
+      "grad_norm": 0.676599383354187,
+      "learning_rate": 0.0005563130059363194,
+      "loss": 3.9419,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7439353099730458,
+      "grad_norm": 0.5250130891799927,
+      "learning_rate": 0.000555989206691851,
+      "loss": 3.9408,
+      "step": 6900
+    },
+    {
+      "epoch": 0.7493261455525606,
+      "grad_norm": 0.6004669070243835,
+      "learning_rate": 0.0005556654074473826,
+      "loss": 3.9439,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.6935180425643921,
+      "learning_rate": 0.0005553416082029141,
+      "loss": 3.927,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "eval_accuracy": 0.3333785691522934,
+      "eval_loss": 3.8611528873443604,
+      "eval_runtime": 183.7687,
+      "eval_samples_per_second": 98.009,
+      "eval_steps_per_second": 6.127,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7601078167115903,
+      "grad_norm": 0.6420605182647705,
+      "learning_rate": 0.0005550178089584457,
+      "loss": 3.9192,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7654986522911051,
+      "grad_norm": 0.5632703900337219,
+      "learning_rate": 0.0005546940097139772,
+      "loss": 3.9194,
+      "step": 7100
+    },
+    {
+      "epoch": 0.77088948787062,
+      "grad_norm": 0.6754069328308105,
+      "learning_rate": 0.0005543702104695089,
+      "loss": 3.9238,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7762803234501348,
+      "grad_norm": 0.6374277472496033,
+      "learning_rate": 0.0005540464112250404,
+      "loss": 3.9201,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7816711590296496,
+      "grad_norm": 0.7086546421051025,
+      "learning_rate": 0.000553722611980572,
+      "loss": 3.9326,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7870619946091644,
+      "grad_norm": 0.5942779779434204,
+      "learning_rate": 0.0005533988127361035,
+      "loss": 3.9157,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7924528301886793,
+      "grad_norm": 0.5517537593841553,
+      "learning_rate": 0.0005530750134916352,
+      "loss": 3.933,
+      "step": 7350
+    },
+    {
+      "epoch": 0.7978436657681941,
+      "grad_norm": 0.6491912007331848,
+      "learning_rate": 0.0005527512142471668,
+      "loss": 3.9062,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8032345013477089,
+      "grad_norm": 0.6026391983032227,
+      "learning_rate": 0.0005524274150026982,
+      "loss": 3.908,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8086253369272237,
+      "grad_norm": 0.6502100825309753,
+      "learning_rate": 0.0005521036157582299,
+      "loss": 3.8975,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8140161725067385,
+      "grad_norm": 0.6446515917778015,
+      "learning_rate": 0.0005517798165137614,
+      "loss": 3.9019,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8194070080862533,
+      "grad_norm": 0.6300444602966309,
+      "learning_rate": 0.000551456017269293,
+      "loss": 3.8779,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8247978436657682,
+      "grad_norm": 0.6193793416023254,
+      "learning_rate": 0.0005511322180248245,
+      "loss": 3.8974,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8301886792452831,
+      "grad_norm": 0.6247860789299011,
+      "learning_rate": 0.0005508084187803562,
+      "loss": 3.9058,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8355795148247979,
+      "grad_norm": 0.5672491192817688,
+      "learning_rate": 0.0005504846195358877,
+      "loss": 3.9004,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8409703504043127,
+      "grad_norm": 0.6407556533813477,
+      "learning_rate": 0.0005501608202914193,
+      "loss": 3.8903,
+      "step": 7800
+    },
+    {
+      "epoch": 0.8463611859838275,
+      "grad_norm": 0.633316159248352,
+      "learning_rate": 0.0005498370210469508,
+      "loss": 3.8918,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8517520215633423,
+      "grad_norm": 0.6338444948196411,
+      "learning_rate": 0.0005495132218024824,
+      "loss": 3.8759,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.6238503456115723,
+      "learning_rate": 0.000549189422558014,
+      "loss": 3.8858,
+      "step": 7950
+    },
+    {
+      "epoch": 0.862533692722372,
+      "grad_norm": 0.6086992025375366,
+      "learning_rate": 0.0005488656233135456,
+      "loss": 3.8837,
+      "step": 8000
+    },
+    {
+      "epoch": 0.862533692722372,
+      "eval_accuracy": 0.3381041005491208,
+      "eval_loss": 3.8141894340515137,
+      "eval_runtime": 184.0519,
+      "eval_samples_per_second": 97.858,
+      "eval_steps_per_second": 6.118,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8679245283018868,
+      "grad_norm": 0.6189132928848267,
+      "learning_rate": 0.0005485418240690771,
+      "loss": 3.8654,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8733153638814016,
+      "grad_norm": 0.5865569710731506,
+      "learning_rate": 0.0005482180248246087,
+      "loss": 3.8813,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8787061994609164,
+      "grad_norm": 0.6706259250640869,
+      "learning_rate": 0.0005478942255801403,
+      "loss": 3.8778,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8840970350404312,
+      "grad_norm": 0.5675233602523804,
+      "learning_rate": 0.0005475704263356718,
+      "loss": 3.8673,
+      "step": 8200
+    },
+    {
+      "epoch": 0.889487870619946,
+      "grad_norm": 0.5570104122161865,
+      "learning_rate": 0.0005472466270912034,
+      "loss": 3.8772,
+      "step": 8250
+    },
+    {
+      "epoch": 0.894878706199461,
+      "grad_norm": 0.6178762912750244,
+      "learning_rate": 0.000546922827846735,
+      "loss": 3.8657,
+      "step": 8300
+    },
+    {
+      "epoch": 0.9002695417789758,
+      "grad_norm": 0.5492017269134521,
+      "learning_rate": 0.0005465990286022665,
+      "loss": 3.8649,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.6238774061203003,
+      "learning_rate": 0.0005462752293577981,
+      "loss": 3.8805,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9110512129380054,
+      "grad_norm": 0.6028736233711243,
+      "learning_rate": 0.0005459514301133296,
+      "loss": 3.8757,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9164420485175202,
+      "grad_norm": 0.5593979954719543,
+      "learning_rate": 0.0005456276308688613,
+      "loss": 3.8699,
+      "step": 8500
+    },
+    {
+      "epoch": 0.921832884097035,
+      "grad_norm": 0.535191535949707,
+      "learning_rate": 0.0005453038316243929,
+      "loss": 3.8642,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9272237196765498,
+      "grad_norm": 0.5982407927513123,
+      "learning_rate": 0.0005449800323799244,
+      "loss": 3.876,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9326145552560647,
+      "grad_norm": 0.5958526730537415,
+      "learning_rate": 0.000544656233135456,
+      "loss": 3.8524,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9380053908355795,
+      "grad_norm": 0.6209942698478699,
+      "learning_rate": 0.0005443324338909875,
+      "loss": 3.857,
+      "step": 8700
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.6151642799377441,
+      "learning_rate": 0.0005440086346465192,
+      "loss": 3.8689,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9487870619946092,
+      "grad_norm": 0.6302937269210815,
+      "learning_rate": 0.0005436848354020506,
+      "loss": 3.8544,
+      "step": 8800
+    },
+    {
+      "epoch": 0.954177897574124,
+      "grad_norm": 0.6161736249923706,
+      "learning_rate": 0.0005433610361575823,
+      "loss": 3.8582,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9595687331536388,
+      "grad_norm": 0.572306752204895,
+      "learning_rate": 0.0005430372369131138,
+      "loss": 3.8398,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9649595687331537,
+      "grad_norm": 0.6290417909622192,
+      "learning_rate": 0.0005427134376686454,
+      "loss": 3.8538,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "grad_norm": 0.6120606660842896,
+      "learning_rate": 0.0005423896384241769,
+      "loss": 3.8397,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "eval_accuracy": 0.3412786119723965,
+      "eval_loss": 3.7767457962036133,
+      "eval_runtime": 183.8892,
+      "eval_samples_per_second": 97.945,
+      "eval_steps_per_second": 6.123,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9757412398921833,
+      "grad_norm": 0.8714722394943237,
+      "learning_rate": 0.0005420658391797086,
+      "loss": 3.8333,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9811320754716981,
+      "grad_norm": 0.5733585953712463,
+      "learning_rate": 0.0005417420399352401,
+      "loss": 3.8474,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9865229110512129,
+      "grad_norm": 0.6671193242073059,
+      "learning_rate": 0.0005414182406907717,
+      "loss": 3.8378,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9919137466307277,
+      "grad_norm": 0.6461030840873718,
+      "learning_rate": 0.0005410944414463032,
+      "loss": 3.8529,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9973045822102425,
+      "grad_norm": 0.6000782251358032,
+      "learning_rate": 0.0005407706422018348,
+      "loss": 3.8554,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0026954177897573,
+      "grad_norm": 0.558596670627594,
+      "learning_rate": 0.0005404468429573664,
+      "loss": 3.7997,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0080862533692723,
+      "grad_norm": 0.6433195471763611,
+      "learning_rate": 0.000540123043712898,
+      "loss": 3.7679,
+      "step": 9350
+    },
+    {
+      "epoch": 1.013477088948787,
+      "grad_norm": 0.6428884267807007,
+      "learning_rate": 0.0005397992444684295,
+      "loss": 3.7523,
+      "step": 9400
+    },
+    {
+      "epoch": 1.0188679245283019,
+      "grad_norm": 0.6692826747894287,
+      "learning_rate": 0.0005394754452239611,
+      "loss": 3.7634,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0242587601078168,
+      "grad_norm": 0.5526914596557617,
+      "learning_rate": 0.0005391516459794927,
+      "loss": 3.777,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0296495956873315,
+      "grad_norm": 0.5531561374664307,
+      "learning_rate": 0.0005388278467350242,
+      "loss": 3.7854,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0350404312668464,
+      "grad_norm": 0.5635939240455627,
+      "learning_rate": 0.0005385040474905557,
+      "loss": 3.7791,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0404312668463611,
+      "grad_norm": 0.5495708584785461,
+      "learning_rate": 0.0005381802482460874,
+      "loss": 3.786,
+      "step": 9650
+    },
+    {
+      "epoch": 1.045822102425876,
+      "grad_norm": 0.5400436520576477,
+      "learning_rate": 0.000537856449001619,
+      "loss": 3.7857,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0512129380053907,
+      "grad_norm": 0.6219043731689453,
+      "learning_rate": 0.0005375326497571505,
+      "loss": 3.758,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0566037735849056,
+      "grad_norm": 0.6601648926734924,
+      "learning_rate": 0.000537208850512682,
+      "loss": 3.7657,
+      "step": 9800
+    },
+    {
+      "epoch": 1.0619946091644206,
+      "grad_norm": 0.5708723664283752,
+      "learning_rate": 0.0005368850512682137,
+      "loss": 3.7652,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0673854447439353,
+      "grad_norm": 0.5404535531997681,
+      "learning_rate": 0.0005365612520237453,
+      "loss": 3.7756,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0727762803234502,
+      "grad_norm": 0.6094656586647034,
+      "learning_rate": 0.0005362374527792768,
+      "loss": 3.7774,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "grad_norm": 0.6319515109062195,
+      "learning_rate": 0.0005359136535348084,
+      "loss": 3.7756,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "eval_accuracy": 0.34461186526476917,
+      "eval_loss": 3.7555184364318848,
+      "eval_runtime": 183.736,
+      "eval_samples_per_second": 98.026,
+      "eval_steps_per_second": 6.128,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0835579514824798,
+      "grad_norm": 0.5507042407989502,
+      "learning_rate": 0.0005355898542903399,
+      "loss": 3.754,
+      "step": 10050
+    },
+    {
+      "epoch": 1.0889487870619945,
+      "grad_norm": 0.6493269801139832,
+      "learning_rate": 0.0005352660550458716,
+      "loss": 3.7587,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0943396226415094,
+      "grad_norm": 0.6598133444786072,
+      "learning_rate": 0.000534942255801403,
+      "loss": 3.748,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0997304582210243,
+      "grad_norm": 0.6240821480751038,
+      "learning_rate": 0.0005346184565569347,
+      "loss": 3.762,
+      "step": 10200
+    },
+    {
+      "epoch": 1.105121293800539,
+      "grad_norm": 0.5528523325920105,
+      "learning_rate": 0.0005342946573124662,
+      "loss": 3.7801,
+      "step": 10250
+    },
+    {
+      "epoch": 1.110512129380054,
+      "grad_norm": 0.647997260093689,
+      "learning_rate": 0.0005339708580679978,
+      "loss": 3.7832,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1159029649595686,
+      "grad_norm": 0.5529667139053345,
+      "learning_rate": 0.0005336470588235293,
+      "loss": 3.7585,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1212938005390836,
+      "grad_norm": 0.628305971622467,
+      "learning_rate": 0.000533323259579061,
+      "loss": 3.7625,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1266846361185983,
+      "grad_norm": 0.5966048836708069,
+      "learning_rate": 0.0005329994603345925,
+      "loss": 3.7575,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1320754716981132,
+      "grad_norm": 0.6109882593154907,
+      "learning_rate": 0.0005326756610901241,
+      "loss": 3.7564,
+      "step": 10500
+    },
+    {
+      "epoch": 1.137466307277628,
+      "grad_norm": 0.5946037173271179,
+      "learning_rate": 0.0005323518618456556,
+      "loss": 3.7565,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.6087301969528198,
+      "learning_rate": 0.0005320280626011872,
+      "loss": 3.7684,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1482479784366577,
+      "grad_norm": 0.6136031150817871,
+      "learning_rate": 0.0005317042633567188,
+      "loss": 3.7575,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1536388140161726,
+      "grad_norm": 0.6112404465675354,
+      "learning_rate": 0.0005313804641122504,
+      "loss": 3.7598,
+      "step": 10700
+    },
+    {
+      "epoch": 1.1590296495956873,
+      "grad_norm": 0.5609838962554932,
+      "learning_rate": 0.0005310566648677819,
+      "loss": 3.7548,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1644204851752022,
+      "grad_norm": 0.5767527222633362,
+      "learning_rate": 0.0005307328656233135,
+      "loss": 3.7456,
+      "step": 10800
+    },
+    {
+      "epoch": 1.169811320754717,
+      "grad_norm": 0.6225987076759338,
+      "learning_rate": 0.000530409066378845,
+      "loss": 3.742,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1752021563342319,
+      "grad_norm": 1.1274323463439941,
+      "learning_rate": 0.0005300852671343766,
+      "loss": 3.7538,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1805929919137466,
+      "grad_norm": 0.6617433428764343,
+      "learning_rate": 0.0005297614678899081,
+      "loss": 3.7462,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "grad_norm": 0.5775931477546692,
+      "learning_rate": 0.0005294376686454398,
+      "loss": 3.7366,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "eval_accuracy": 0.34766566332968296,
+      "eval_loss": 3.719996690750122,
+      "eval_runtime": 183.7301,
+      "eval_samples_per_second": 98.03,
+      "eval_steps_per_second": 6.129,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1913746630727764,
+      "grad_norm": 0.5878674387931824,
+      "learning_rate": 0.0005291138694009714,
+      "loss": 3.7494,
+      "step": 11050
+    },
+    {
+      "epoch": 1.196765498652291,
+      "grad_norm": 0.7032240033149719,
+      "learning_rate": 0.0005287900701565029,
+      "loss": 3.7402,
+      "step": 11100
+    },
+    {
+      "epoch": 1.202156334231806,
+      "grad_norm": 0.6269480586051941,
+      "learning_rate": 0.0005284662709120345,
+      "loss": 3.7428,
+      "step": 11150
+    },
+    {
+      "epoch": 1.2075471698113207,
+      "grad_norm": 0.7135176062583923,
+      "learning_rate": 0.0005281424716675661,
+      "loss": 3.7423,
+      "step": 11200
+    },
+    {
+      "epoch": 1.2129380053908356,
+      "grad_norm": 0.5879312753677368,
+      "learning_rate": 0.0005278186724230977,
+      "loss": 3.7322,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2183288409703503,
+      "grad_norm": 0.5728159546852112,
+      "learning_rate": 0.0005274948731786292,
+      "loss": 3.7306,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2237196765498652,
+      "grad_norm": 0.6225752234458923,
+      "learning_rate": 0.0005271710739341608,
+      "loss": 3.7473,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2291105121293802,
+      "grad_norm": 0.5699577331542969,
+      "learning_rate": 0.0005268472746896923,
+      "loss": 3.7429,
+      "step": 11400
+    },
+    {
+      "epoch": 1.2345013477088949,
+      "grad_norm": 0.7263878583908081,
+      "learning_rate": 0.000526523475445224,
+      "loss": 3.7332,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2398921832884098,
+      "grad_norm": 0.5602506399154663,
+      "learning_rate": 0.0005261996762007554,
+      "loss": 3.7196,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2452830188679245,
+      "grad_norm": 0.615954577922821,
+      "learning_rate": 0.0005258758769562871,
+      "loss": 3.7403,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2506738544474394,
+      "grad_norm": 0.623807966709137,
+      "learning_rate": 0.0005255520777118186,
+      "loss": 3.7425,
+      "step": 11600
+    },
+    {
+      "epoch": 1.256064690026954,
+      "grad_norm": 0.6084052920341492,
+      "learning_rate": 0.0005252282784673502,
+      "loss": 3.7314,
+      "step": 11650
+    },
+    {
+      "epoch": 1.261455525606469,
+      "grad_norm": 0.6058493256568909,
+      "learning_rate": 0.0005249044792228817,
+      "loss": 3.7351,
+      "step": 11700
+    },
+    {
+      "epoch": 1.266846361185984,
+      "grad_norm": 0.5308002233505249,
+      "learning_rate": 0.0005245806799784133,
+      "loss": 3.73,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2722371967654986,
+      "grad_norm": 0.5932163000106812,
+      "learning_rate": 0.0005242568807339449,
+      "loss": 3.7266,
+      "step": 11800
+    },
+    {
+      "epoch": 1.2776280323450135,
+      "grad_norm": 0.6025761961936951,
+      "learning_rate": 0.0005239330814894765,
+      "loss": 3.7206,
+      "step": 11850
+    },
+    {
+      "epoch": 1.2830188679245282,
+      "grad_norm": 0.5808297395706177,
+      "learning_rate": 0.000523609282245008,
+      "loss": 3.7434,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2884097035040432,
+      "grad_norm": 0.6271538138389587,
+      "learning_rate": 0.0005232854830005396,
+      "loss": 3.7376,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "grad_norm": 0.5595883131027222,
+      "learning_rate": 0.0005229616837560712,
+      "loss": 3.733,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "eval_accuracy": 0.34978080909676745,
+      "eval_loss": 3.697939395904541,
+      "eval_runtime": 183.962,
+      "eval_samples_per_second": 97.906,
+      "eval_steps_per_second": 6.121,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2991913746630728,
+      "grad_norm": 0.5748785734176636,
+      "learning_rate": 0.0005226378845116028,
+      "loss": 3.7154,
+      "step": 12050
+    },
+    {
+      "epoch": 1.3045822102425877,
+      "grad_norm": 0.5758981704711914,
+      "learning_rate": 0.0005223140852671344,
+      "loss": 3.7283,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3099730458221024,
+      "grad_norm": 0.7004877924919128,
+      "learning_rate": 0.0005219902860226659,
+      "loss": 3.7372,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3153638814016173,
+      "grad_norm": 0.6041122078895569,
+      "learning_rate": 0.0005216664867781975,
+      "loss": 3.7387,
+      "step": 12200
+    },
+    {
+      "epoch": 1.320754716981132,
+      "grad_norm": 0.66448974609375,
+      "learning_rate": 0.000521342687533729,
+      "loss": 3.719,
+      "step": 12250
+    },
+    {
+      "epoch": 1.326145552560647,
+      "grad_norm": 0.5907305479049683,
+      "learning_rate": 0.0005210188882892606,
+      "loss": 3.7295,
+      "step": 12300
+    },
+    {
+      "epoch": 1.3315363881401616,
+      "grad_norm": 0.5952483415603638,
+      "learning_rate": 0.0005206950890447922,
+      "loss": 3.719,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3369272237196765,
+      "grad_norm": 0.5341550707817078,
+      "learning_rate": 0.0005203712898003238,
+      "loss": 3.733,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3423180592991915,
+      "grad_norm": 0.6329076886177063,
+      "learning_rate": 0.0005200474905558553,
+      "loss": 3.7221,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3477088948787062,
+      "grad_norm": 0.655746340751648,
+      "learning_rate": 0.0005197236913113869,
+      "loss": 3.7229,
+      "step": 12500
+    },
+    {
+      "epoch": 1.353099730458221,
+      "grad_norm": 0.5556245446205139,
+      "learning_rate": 0.0005193998920669184,
+      "loss": 3.7318,
+      "step": 12550
+    },
+    {
+      "epoch": 1.3584905660377358,
+      "grad_norm": 0.5874679684638977,
+      "learning_rate": 0.0005190760928224501,
+      "loss": 3.714,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3638814016172507,
+      "grad_norm": 0.572918713092804,
+      "learning_rate": 0.0005187522935779816,
+      "loss": 3.7238,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3692722371967654,
+      "grad_norm": 0.623197078704834,
+      "learning_rate": 0.0005184284943335132,
+      "loss": 3.736,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3746630727762803,
+      "grad_norm": 0.587779700756073,
+      "learning_rate": 0.0005181046950890447,
+      "loss": 3.7212,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3800539083557952,
+      "grad_norm": 0.6302971243858337,
+      "learning_rate": 0.0005177808958445764,
+      "loss": 3.713,
+      "step": 12800
+    },
+    {
+      "epoch": 1.38544474393531,
+      "grad_norm": 0.626609206199646,
+      "learning_rate": 0.0005174570966001078,
+      "loss": 3.7091,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3908355795148248,
+      "grad_norm": 0.6012193560600281,
+      "learning_rate": 0.0005171332973556395,
+      "loss": 3.7141,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3962264150943398,
+      "grad_norm": 0.5877785682678223,
+      "learning_rate": 0.000516809498111171,
+      "loss": 3.7119,
+      "step": 12950
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "grad_norm": 0.5771360993385315,
+      "learning_rate": 0.0005164856988667026,
+      "loss": 3.7094,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "eval_accuracy": 0.3517633983407183,
+      "eval_loss": 3.6760079860687256,
+      "eval_runtime": 183.7267,
+      "eval_samples_per_second": 98.031,
+      "eval_steps_per_second": 6.129,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4070080862533692,
+      "grad_norm": 0.6249587535858154,
+      "learning_rate": 0.0005161618996222341,
+      "loss": 3.7209,
+      "step": 13050
+    },
+    {
+      "epoch": 1.412398921832884,
+      "grad_norm": 0.5885133743286133,
+      "learning_rate": 0.0005158381003777657,
+      "loss": 3.7313,
+      "step": 13100
+    },
+    {
+      "epoch": 1.417789757412399,
+      "grad_norm": 0.5590161085128784,
+      "learning_rate": 0.0005155143011332973,
+      "loss": 3.71,
+      "step": 13150
+    },
+    {
+      "epoch": 1.4231805929919137,
+      "grad_norm": 0.53040611743927,
+      "learning_rate": 0.0005151905018888289,
+      "loss": 3.713,
+      "step": 13200
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.5803114175796509,
+      "learning_rate": 0.0005148731786292498,
+      "loss": 3.7096,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4339622641509435,
+      "grad_norm": 0.5958723425865173,
+      "learning_rate": 0.0005145493793847814,
+      "loss": 3.7077,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4393530997304582,
+      "grad_norm": 0.6058682203292847,
+      "learning_rate": 0.000514225580140313,
+      "loss": 3.6937,
+      "step": 13350
+    },
+    {
+      "epoch": 1.444743935309973,
+      "grad_norm": 0.5932038426399231,
+      "learning_rate": 0.0005139017808958445,
+      "loss": 3.7048,
+      "step": 13400
+    },
+    {
+      "epoch": 1.4501347708894878,
+      "grad_norm": 0.6021926403045654,
+      "learning_rate": 0.0005135779816513762,
+      "loss": 3.7134,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4555256064690028,
+      "grad_norm": 0.5810138583183289,
+      "learning_rate": 0.0005132541824069076,
+      "loss": 3.6991,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4609164420485174,
+      "grad_norm": 0.633111298084259,
+      "learning_rate": 0.0005129303831624393,
+      "loss": 3.7089,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4663072776280324,
+      "grad_norm": 0.5751838088035583,
+      "learning_rate": 0.0005126065839179708,
+      "loss": 3.7122,
+      "step": 13600
+    },
+    {
+      "epoch": 1.4716981132075473,
+      "grad_norm": 0.5811141133308411,
+      "learning_rate": 0.0005122827846735024,
+      "loss": 3.6844,
+      "step": 13650
+    },
+    {
+      "epoch": 1.477088948787062,
+      "grad_norm": 0.6062330007553101,
+      "learning_rate": 0.0005119589854290339,
+      "loss": 3.6816,
+      "step": 13700
+    },
+    {
+      "epoch": 1.482479784366577,
+      "grad_norm": 0.6124317049980164,
+      "learning_rate": 0.0005116351861845655,
+      "loss": 3.702,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4878706199460916,
+      "grad_norm": 0.6740068197250366,
+      "learning_rate": 0.0005113113869400971,
+      "loss": 3.7182,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4932614555256065,
+      "grad_norm": 0.595000684261322,
+      "learning_rate": 0.0005109875876956287,
+      "loss": 3.7092,
+      "step": 13850
+    },
+    {
+      "epoch": 1.4986522911051212,
+      "grad_norm": 0.5819430947303772,
+      "learning_rate": 0.0005106637884511602,
+      "loss": 3.7247,
+      "step": 13900
+    },
+    {
+      "epoch": 1.5040431266846361,
+      "grad_norm": 0.5515394806861877,
+      "learning_rate": 0.0005103399892066918,
+      "loss": 3.6982,
+      "step": 13950
+    },
+    {
+      "epoch": 1.509433962264151,
+      "grad_norm": 0.5956779718399048,
+      "learning_rate": 0.0005100161899622234,
+      "loss": 3.7,
+      "step": 14000
+    },
+    {
+      "epoch": 1.509433962264151,
+      "eval_accuracy": 0.35387289415763645,
+      "eval_loss": 3.656662940979004,
+      "eval_runtime": 183.8582,
+      "eval_samples_per_second": 97.961,
+      "eval_steps_per_second": 6.124,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5148247978436657,
+      "grad_norm": 0.5680575966835022,
+      "learning_rate": 0.000509692390717755,
+      "loss": 3.6909,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5202156334231804,
+      "grad_norm": 0.7383117079734802,
+      "learning_rate": 0.0005093685914732865,
+      "loss": 3.6859,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5256064690026954,
+      "grad_norm": 0.567956268787384,
+      "learning_rate": 0.0005090447922288181,
+      "loss": 3.6834,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5309973045822103,
+      "grad_norm": 0.5391273498535156,
+      "learning_rate": 0.0005087209929843496,
+      "loss": 3.717,
+      "step": 14200
+    },
+    {
+      "epoch": 1.536388140161725,
+      "grad_norm": 0.6262120008468628,
+      "learning_rate": 0.0005083971937398812,
+      "loss": 3.6904,
+      "step": 14250
+    },
+    {
+      "epoch": 1.54177897574124,
+      "grad_norm": 0.6266434192657471,
+      "learning_rate": 0.0005080733944954127,
+      "loss": 3.695,
+      "step": 14300
+    },
+    {
+      "epoch": 1.5471698113207548,
+      "grad_norm": 0.6252384185791016,
+      "learning_rate": 0.0005077495952509444,
+      "loss": 3.6769,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5525606469002695,
+      "grad_norm": 0.5647259950637817,
+      "learning_rate": 0.0005074257960064759,
+      "loss": 3.6894,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5579514824797842,
+      "grad_norm": 0.5937622785568237,
+      "learning_rate": 0.0005071019967620075,
+      "loss": 3.6954,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5633423180592994,
+      "grad_norm": 0.608908474445343,
+      "learning_rate": 0.000506778197517539,
+      "loss": 3.686,
+      "step": 14500
+    },
+    {
+      "epoch": 1.568733153638814,
+      "grad_norm": 0.6360867023468018,
+      "learning_rate": 0.0005064543982730707,
+      "loss": 3.6975,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5741239892183287,
+      "grad_norm": 0.5903249979019165,
+      "learning_rate": 0.0005061305990286023,
+      "loss": 3.6919,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5795148247978437,
+      "grad_norm": 0.6455665826797485,
+      "learning_rate": 0.0005058067997841338,
+      "loss": 3.6975,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5849056603773586,
+      "grad_norm": 0.6089785099029541,
+      "learning_rate": 0.0005054830005396654,
+      "loss": 3.6835,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5902964959568733,
+      "grad_norm": 0.6004166603088379,
+      "learning_rate": 0.0005051592012951969,
+      "loss": 3.6769,
+      "step": 14750
+    },
+    {
+      "epoch": 1.595687331536388,
+      "grad_norm": 0.6187482476234436,
+      "learning_rate": 0.0005048354020507286,
+      "loss": 3.696,
+      "step": 14800
+    },
+    {
+      "epoch": 1.6010781671159031,
+      "grad_norm": 0.5907650589942932,
+      "learning_rate": 0.00050451160280626,
+      "loss": 3.696,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6064690026954178,
+      "grad_norm": 0.5473130941390991,
+      "learning_rate": 0.0005041878035617917,
+      "loss": 3.6865,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6118598382749325,
+      "grad_norm": 0.5759040117263794,
+      "learning_rate": 0.0005038640043173232,
+      "loss": 3.6822,
+      "step": 14950
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "grad_norm": 0.7202991843223572,
+      "learning_rate": 0.0005035402050728548,
+      "loss": 3.6789,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "eval_accuracy": 0.35576888704999915,
+      "eval_loss": 3.6384365558624268,
+      "eval_runtime": 183.9946,
+      "eval_samples_per_second": 97.889,
+      "eval_steps_per_second": 6.12,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6226415094339623,
+      "grad_norm": 0.5561956167221069,
+      "learning_rate": 0.0005032164058283863,
+      "loss": 3.6811,
+      "step": 15050
+    },
+    {
+      "epoch": 1.628032345013477,
+      "grad_norm": 0.5614675879478455,
+      "learning_rate": 0.0005028926065839179,
+      "loss": 3.6759,
+      "step": 15100
+    },
+    {
+      "epoch": 1.633423180592992,
+      "grad_norm": 0.5612145066261292,
+      "learning_rate": 0.0005025688073394495,
+      "loss": 3.6769,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6388140161725069,
+      "grad_norm": 0.5823696851730347,
+      "learning_rate": 0.0005022450080949811,
+      "loss": 3.6828,
+      "step": 15200
+    },
+    {
+      "epoch": 1.6442048517520216,
+      "grad_norm": 0.6124740242958069,
+      "learning_rate": 0.0005019212088505126,
+      "loss": 3.6843,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6495956873315363,
+      "grad_norm": 0.5496434569358826,
+      "learning_rate": 0.0005015974096060442,
+      "loss": 3.6852,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6549865229110512,
+      "grad_norm": 0.534186601638794,
+      "learning_rate": 0.0005012736103615758,
+      "loss": 3.6721,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6603773584905661,
+      "grad_norm": 0.6018574237823486,
+      "learning_rate": 0.0005009498111171074,
+      "loss": 3.6751,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6657681940700808,
+      "grad_norm": 0.5834909081459045,
+      "learning_rate": 0.0005006260118726389,
+      "loss": 3.6759,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6711590296495957,
+      "grad_norm": 0.5618143677711487,
+      "learning_rate": 0.0005003022126281705,
+      "loss": 3.6761,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6765498652291106,
+      "grad_norm": 0.6277567148208618,
+      "learning_rate": 0.000499978413383702,
+      "loss": 3.6678,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6819407008086253,
+      "grad_norm": 0.5726228952407837,
+      "learning_rate": 0.0004996546141392336,
+      "loss": 3.6657,
+      "step": 15600
+    },
+    {
+      "epoch": 1.68733153638814,
+      "grad_norm": 0.5706265568733215,
+      "learning_rate": 0.0004993308148947651,
+      "loss": 3.6736,
+      "step": 15650
+    },
+    {
+      "epoch": 1.692722371967655,
+      "grad_norm": 0.5442909002304077,
+      "learning_rate": 0.0004990070156502968,
+      "loss": 3.663,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6981132075471699,
+      "grad_norm": 0.5733287334442139,
+      "learning_rate": 0.0004986832164058284,
+      "loss": 3.6602,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7035040431266846,
+      "grad_norm": 0.5900512337684631,
+      "learning_rate": 0.0004983594171613599,
+      "loss": 3.6734,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7088948787061995,
+      "grad_norm": 0.6177293658256531,
+      "learning_rate": 0.0004980356179168915,
+      "loss": 3.6791,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.6071768999099731,
+      "learning_rate": 0.000497711818672423,
+      "loss": 3.7025,
+      "step": 15900
+    },
+    {
+      "epoch": 1.719676549865229,
+      "grad_norm": 0.542489230632782,
+      "learning_rate": 0.0004973880194279547,
+      "loss": 3.6678,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "grad_norm": 0.6317324638366699,
+      "learning_rate": 0.0004970642201834862,
+      "loss": 3.6767,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "eval_accuracy": 0.35749940159422033,
+      "eval_loss": 3.616533041000366,
+      "eval_runtime": 183.6069,
+      "eval_samples_per_second": 98.095,
+      "eval_steps_per_second": 6.133,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7304582210242587,
+      "grad_norm": 0.5835158824920654,
+      "learning_rate": 0.0004967404209390178,
+      "loss": 3.6446,
+      "step": 16050
+    },
+    {
+      "epoch": 1.7358490566037736,
+      "grad_norm": 0.5380732417106628,
+      "learning_rate": 0.0004964166216945493,
+      "loss": 3.6536,
+      "step": 16100
+    },
+    {
+      "epoch": 1.7412398921832883,
+      "grad_norm": 0.5660726428031921,
+      "learning_rate": 0.000496092822450081,
+      "loss": 3.6682,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7466307277628033,
+      "grad_norm": 0.619783341884613,
+      "learning_rate": 0.0004957690232056125,
+      "loss": 3.6563,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7520215633423182,
+      "grad_norm": 0.5709429979324341,
+      "learning_rate": 0.0004954452239611441,
+      "loss": 3.6634,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7574123989218329,
+      "grad_norm": 0.5726304054260254,
+      "learning_rate": 0.0004951214247166756,
+      "loss": 3.6554,
+      "step": 16300
+    },
+    {
+      "epoch": 1.7628032345013476,
+      "grad_norm": 0.600409209728241,
+      "learning_rate": 0.0004947976254722072,
+      "loss": 3.6527,
+      "step": 16350
+    },
+    {
+      "epoch": 1.7681940700808625,
+      "grad_norm": 0.5817192792892456,
+      "learning_rate": 0.0004944738262277387,
+      "loss": 3.663,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7735849056603774,
+      "grad_norm": 0.5794048309326172,
+      "learning_rate": 0.0004941500269832703,
+      "loss": 3.6628,
+      "step": 16450
+    },
+    {
+      "epoch": 1.778975741239892,
+      "grad_norm": 0.6050107479095459,
+      "learning_rate": 0.0004938262277388019,
+      "loss": 3.6518,
+      "step": 16500
+    },
+    {
+      "epoch": 1.784366576819407,
+      "grad_norm": 0.638290524482727,
+      "learning_rate": 0.0004935024284943335,
+      "loss": 3.6464,
+      "step": 16550
+    },
+    {
+      "epoch": 1.789757412398922,
+      "grad_norm": 0.5926001667976379,
+      "learning_rate": 0.000493178629249865,
+      "loss": 3.654,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7951482479784366,
+      "grad_norm": 0.5781203508377075,
+      "learning_rate": 0.0004928548300053966,
+      "loss": 3.6613,
+      "step": 16650
+    },
+    {
+      "epoch": 1.8005390835579513,
+      "grad_norm": 0.5973802208900452,
+      "learning_rate": 0.0004925310307609282,
+      "loss": 3.6521,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8059299191374663,
+      "grad_norm": 0.5450361967086792,
+      "learning_rate": 0.0004922072315164598,
+      "loss": 3.6465,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8113207547169812,
+      "grad_norm": 0.5766150951385498,
+      "learning_rate": 0.0004918834322719913,
+      "loss": 3.651,
+      "step": 16800
+    },
+    {
+      "epoch": 1.8167115902964959,
+      "grad_norm": 0.5567212104797363,
+      "learning_rate": 0.0004915596330275229,
+      "loss": 3.6413,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8221024258760108,
+      "grad_norm": 0.5800772309303284,
+      "learning_rate": 0.0004912358337830544,
+      "loss": 3.6472,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8274932614555257,
+      "grad_norm": 0.5754286050796509,
+      "learning_rate": 0.000490912034538586,
+      "loss": 3.6532,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "grad_norm": 0.5063068270683289,
+      "learning_rate": 0.0004905882352941175,
+      "loss": 3.6596,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "eval_accuracy": 0.35870718709516614,
+      "eval_loss": 3.6030871868133545,
+      "eval_runtime": 184.2683,
+      "eval_samples_per_second": 97.743,
+      "eval_steps_per_second": 6.111,
+      "step": 17000
+    },
+    {
+      "epoch": 1.838274932614555,
+      "grad_norm": 0.6203033328056335,
+      "learning_rate": 0.0004902644360496492,
+      "loss": 3.6504,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8436657681940702,
+      "grad_norm": 0.5805009007453918,
+      "learning_rate": 0.0004899406368051808,
+      "loss": 3.6426,
+      "step": 17100
+    },
+    {
+      "epoch": 1.849056603773585,
+      "grad_norm": 0.5866754651069641,
+      "learning_rate": 0.0004896168375607123,
+      "loss": 3.6421,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8544474393530996,
+      "grad_norm": 0.578814685344696,
+      "learning_rate": 0.0004892930383162439,
+      "loss": 3.6404,
+      "step": 17200
+    },
+    {
+      "epoch": 1.8598382749326146,
+      "grad_norm": 0.628090500831604,
+      "learning_rate": 0.0004889692390717754,
+      "loss": 3.6365,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8652291105121295,
+      "grad_norm": 0.5448696613311768,
+      "learning_rate": 0.0004886454398273071,
+      "loss": 3.6613,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8706199460916442,
+      "grad_norm": 0.5808747410774231,
+      "learning_rate": 0.000488328116567728,
+      "loss": 3.6639,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8760107816711589,
+      "grad_norm": 0.5566965341567993,
+      "learning_rate": 0.0004880043173232595,
+      "loss": 3.6437,
+      "step": 17400
+    },
+    {
+      "epoch": 1.881401617250674,
+      "grad_norm": 0.5760877728462219,
+      "learning_rate": 0.0004876805180787911,
+      "loss": 3.6524,
+      "step": 17450
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.5915643572807312,
+      "learning_rate": 0.0004873567188343227,
+      "loss": 3.6365,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8921832884097034,
+      "grad_norm": 0.6473491191864014,
+      "learning_rate": 0.0004870329195898542,
+      "loss": 3.6513,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8975741239892183,
+      "grad_norm": 0.5361552238464355,
+      "learning_rate": 0.00048670912034538583,
+      "loss": 3.6392,
+      "step": 17600
+    },
+    {
+      "epoch": 1.9029649595687332,
+      "grad_norm": 0.5670789480209351,
+      "learning_rate": 0.0004863853211009174,
+      "loss": 3.6365,
+      "step": 17650
+    },
+    {
+      "epoch": 1.908355795148248,
+      "grad_norm": 0.5751020908355713,
+      "learning_rate": 0.000486061521856449,
+      "loss": 3.6501,
+      "step": 17700
+    },
+    {
+      "epoch": 1.9137466307277629,
+      "grad_norm": 0.6289040446281433,
+      "learning_rate": 0.00048573772261198054,
+      "loss": 3.639,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9191374663072778,
+      "grad_norm": 0.5996605157852173,
+      "learning_rate": 0.00048541392336751214,
+      "loss": 3.6535,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9245283018867925,
+      "grad_norm": 0.565588116645813,
+      "learning_rate": 0.0004850901241230437,
+      "loss": 3.6612,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9299191374663072,
+      "grad_norm": 0.5734473466873169,
+      "learning_rate": 0.0004847663248785753,
+      "loss": 3.6367,
+      "step": 17900
+    },
+    {
+      "epoch": 1.935309973045822,
+      "grad_norm": 0.5644897222518921,
+      "learning_rate": 0.0004844425256341068,
+      "loss": 3.6489,
+      "step": 17950
+    },
+    {
+      "epoch": 1.940700808625337,
+      "grad_norm": 0.6228716969490051,
+      "learning_rate": 0.00048411872638963834,
+      "loss": 3.6428,
+      "step": 18000
+    },
+    {
+      "epoch": 1.940700808625337,
+      "eval_accuracy": 0.3604395487384802,
+      "eval_loss": 3.586714744567871,
+      "eval_runtime": 184.0751,
+      "eval_samples_per_second": 97.846,
+      "eval_steps_per_second": 6.117,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9460916442048517,
+      "grad_norm": 0.5661214590072632,
+      "learning_rate": 0.00048379492714516995,
+      "loss": 3.6244,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9514824797843666,
+      "grad_norm": 0.5419253706932068,
+      "learning_rate": 0.0004834711279007015,
+      "loss": 3.6564,
+      "step": 18100
+    },
+    {
+      "epoch": 1.9568733153638815,
+      "grad_norm": 0.6449159383773804,
+      "learning_rate": 0.0004831473286562331,
+      "loss": 3.6534,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9622641509433962,
+      "grad_norm": 0.6221560835838318,
+      "learning_rate": 0.00048282352941176465,
+      "loss": 3.6442,
+      "step": 18200
+    },
+    {
+      "epoch": 1.967654986522911,
+      "grad_norm": 0.6196809411048889,
+      "learning_rate": 0.00048249973016729626,
+      "loss": 3.647,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9730458221024259,
+      "grad_norm": 0.535416841506958,
+      "learning_rate": 0.0004821759309228278,
+      "loss": 3.6259,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9784366576819408,
+      "grad_norm": 0.7311776280403137,
+      "learning_rate": 0.00048185213167835936,
+      "loss": 3.6501,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9838274932614555,
+      "grad_norm": 0.5222246646881104,
+      "learning_rate": 0.00048152833243389096,
+      "loss": 3.6261,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9892183288409704,
+      "grad_norm": 0.5267012715339661,
+      "learning_rate": 0.0004812045331894225,
+      "loss": 3.6294,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9946091644204853,
+      "grad_norm": 0.6064724326133728,
+      "learning_rate": 0.0004808807339449541,
+      "loss": 3.6201,
+      "step": 18500
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.1216500997543335,
+      "learning_rate": 0.0004805569347004856,
+      "loss": 3.6366,
+      "step": 18550
+    },
+    {
+      "epoch": 2.0053908355795147,
+      "grad_norm": 0.6476635336875916,
+      "learning_rate": 0.0004802331354560173,
+      "loss": 3.5507,
+      "step": 18600
+    },
+    {
+      "epoch": 2.01078167115903,
+      "grad_norm": 0.5348339676856995,
+      "learning_rate": 0.00047990933621154877,
+      "loss": 3.5432,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0161725067385445,
+      "grad_norm": 0.5870545506477356,
+      "learning_rate": 0.0004795855369670804,
+      "loss": 3.5513,
+      "step": 18700
+    },
+    {
+      "epoch": 2.0215633423180592,
+      "grad_norm": 0.5712009072303772,
+      "learning_rate": 0.0004792617377226119,
+      "loss": 3.5438,
+      "step": 18750
+    },
+    {
+      "epoch": 2.026954177897574,
+      "grad_norm": 0.5787289142608643,
+      "learning_rate": 0.0004789379384781435,
+      "loss": 3.5498,
+      "step": 18800
+    },
+    {
+      "epoch": 2.032345013477089,
+      "grad_norm": 0.5748060345649719,
+      "learning_rate": 0.0004786141392336751,
+      "loss": 3.5469,
+      "step": 18850
+    },
+    {
+      "epoch": 2.0377358490566038,
+      "grad_norm": 0.6476619243621826,
+      "learning_rate": 0.00047829033998920663,
+      "loss": 3.5605,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0431266846361185,
+      "grad_norm": 0.5768181681632996,
+      "learning_rate": 0.00047796654074473824,
+      "loss": 3.5566,
+      "step": 18950
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "grad_norm": 0.5920575857162476,
+      "learning_rate": 0.0004776427415002698,
+      "loss": 3.556,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "eval_accuracy": 0.3619965446208617,
+      "eval_loss": 3.5773985385894775,
+      "eval_runtime": 183.7894,
+      "eval_samples_per_second": 97.998,
+      "eval_steps_per_second": 6.127,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0539083557951483,
+      "grad_norm": 0.5745695233345032,
+      "learning_rate": 0.0004773189422558014,
+      "loss": 3.5666,
+      "step": 19050
+    },
+    {
+      "epoch": 2.059299191374663,
+      "grad_norm": 0.6024934649467468,
+      "learning_rate": 0.00047699514301133294,
+      "loss": 3.5589,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0646900269541777,
+      "grad_norm": 0.5652180314064026,
+      "learning_rate": 0.00047667134376686455,
+      "loss": 3.5538,
+      "step": 19150
+    },
+    {
+      "epoch": 2.070080862533693,
+      "grad_norm": 0.5560337901115417,
+      "learning_rate": 0.0004763475445223961,
+      "loss": 3.5739,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0754716981132075,
+      "grad_norm": 0.5556290149688721,
+      "learning_rate": 0.0004760237452779276,
+      "loss": 3.5626,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0808625336927222,
+      "grad_norm": 0.5928717851638794,
+      "learning_rate": 0.0004756999460334592,
+      "loss": 3.5542,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0862533692722374,
+      "grad_norm": 0.6069679856300354,
+      "learning_rate": 0.00047538262277388015,
+      "loss": 3.5717,
+      "step": 19350
+    },
+    {
+      "epoch": 2.091644204851752,
+      "grad_norm": 0.5808999538421631,
+      "learning_rate": 0.0004750588235294117,
+      "loss": 3.5625,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0970350404312668,
+      "grad_norm": 0.7181747555732727,
+      "learning_rate": 0.0004747350242849433,
+      "loss": 3.5455,
+      "step": 19450
+    },
+    {
+      "epoch": 2.1024258760107815,
+      "grad_norm": 0.5779094099998474,
+      "learning_rate": 0.00047441122504047486,
+      "loss": 3.5684,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1078167115902966,
+      "grad_norm": 0.5920312404632568,
+      "learning_rate": 0.0004740874257960064,
+      "loss": 3.5592,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1132075471698113,
+      "grad_norm": 0.5829529166221619,
+      "learning_rate": 0.000473763626551538,
+      "loss": 3.5564,
+      "step": 19600
+    },
+    {
+      "epoch": 2.118598382749326,
+      "grad_norm": 0.5921161770820618,
+      "learning_rate": 0.00047343982730706956,
+      "loss": 3.5592,
+      "step": 19650
+    },
+    {
+      "epoch": 2.123989218328841,
+      "grad_norm": 0.553687572479248,
+      "learning_rate": 0.00047311602806260117,
+      "loss": 3.5517,
+      "step": 19700
+    },
+    {
+      "epoch": 2.129380053908356,
+      "grad_norm": 0.6229360103607178,
+      "learning_rate": 0.0004727922288181327,
+      "loss": 3.5422,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1347708894878705,
+      "grad_norm": 0.6002780795097351,
+      "learning_rate": 0.0004724684295736643,
+      "loss": 3.5629,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1401617250673857,
+      "grad_norm": 0.5640039443969727,
+      "learning_rate": 0.0004721446303291959,
+      "loss": 3.5577,
+      "step": 19850
+    },
+    {
+      "epoch": 2.1455525606469004,
+      "grad_norm": 0.5708314180374146,
+      "learning_rate": 0.0004718208310847275,
+      "loss": 3.5596,
+      "step": 19900
+    },
+    {
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.5949370265007019,
+      "learning_rate": 0.000471497031840259,
+      "loss": 3.5509,
+      "step": 19950
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "grad_norm": 0.5819337964057922,
+      "learning_rate": 0.0004711732325957905,
+      "loss": 3.5538,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "eval_accuracy": 0.36339697169190255,
+      "eval_loss": 3.568969488143921,
+      "eval_runtime": 184.1987,
+      "eval_samples_per_second": 97.78,
+      "eval_steps_per_second": 6.113,
+      "step": 20000
+    },
+    {
+      "epoch": 2.161725067385445,
+      "grad_norm": 0.6424840092658997,
+      "learning_rate": 0.00047084943335132213,
+      "loss": 3.5584,
+      "step": 20050
+    },
+    {
+      "epoch": 2.1671159029649596,
+      "grad_norm": 0.5913828611373901,
+      "learning_rate": 0.0004705256341068537,
+      "loss": 3.5541,
+      "step": 20100
+    },
+    {
+      "epoch": 2.1725067385444743,
+      "grad_norm": 0.5962077379226685,
+      "learning_rate": 0.0004702018348623853,
+      "loss": 3.5632,
+      "step": 20150
+    },
+    {
+      "epoch": 2.177897574123989,
+      "grad_norm": 0.524539589881897,
+      "learning_rate": 0.00046987803561791684,
+      "loss": 3.5668,
+      "step": 20200
+    },
+    {
+      "epoch": 2.183288409703504,
+      "grad_norm": 0.5445467233657837,
+      "learning_rate": 0.00046955423637344844,
+      "loss": 3.5701,
+      "step": 20250
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 0.6836073398590088,
+      "learning_rate": 0.00046923043712898,
+      "loss": 3.5538,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1940700808625335,
+      "grad_norm": 0.5664141774177551,
+      "learning_rate": 0.0004689066378845116,
+      "loss": 3.5481,
+      "step": 20350
+    },
+    {
+      "epoch": 2.1994609164420487,
+      "grad_norm": 0.5738502740859985,
+      "learning_rate": 0.00046858283864004315,
+      "loss": 3.5596,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2048517520215634,
+      "grad_norm": 0.7970450520515442,
+      "learning_rate": 0.0004682590393955747,
+      "loss": 3.554,
+      "step": 20450
+    },
+    {
+      "epoch": 2.210242587601078,
+      "grad_norm": 0.535188615322113,
+      "learning_rate": 0.0004679352401511063,
+      "loss": 3.5712,
+      "step": 20500
+    },
+    {
+      "epoch": 2.215633423180593,
+      "grad_norm": 0.5921735763549805,
+      "learning_rate": 0.0004676114409066378,
+      "loss": 3.5654,
+      "step": 20550
+    },
+    {
+      "epoch": 2.221024258760108,
+      "grad_norm": 0.583526074886322,
+      "learning_rate": 0.00046728764166216946,
+      "loss": 3.563,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2264150943396226,
+      "grad_norm": 0.5687211155891418,
+      "learning_rate": 0.00046696384241770095,
+      "loss": 3.5718,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2318059299191373,
+      "grad_norm": 0.6883664131164551,
+      "learning_rate": 0.00046664004317323256,
+      "loss": 3.553,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2371967654986524,
+      "grad_norm": 0.5841453671455383,
+      "learning_rate": 0.0004663162439287641,
+      "loss": 3.5727,
+      "step": 20750
+    },
+    {
+      "epoch": 2.242587601078167,
+      "grad_norm": 0.5803972482681274,
+      "learning_rate": 0.00046599244468429566,
+      "loss": 3.5671,
+      "step": 20800
+    },
+    {
+      "epoch": 2.247978436657682,
+      "grad_norm": 0.605576753616333,
+      "learning_rate": 0.00046566864543982726,
+      "loss": 3.5697,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2533692722371965,
+      "grad_norm": 0.5493614077568054,
+      "learning_rate": 0.0004653448461953588,
+      "loss": 3.5583,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2587601078167117,
+      "grad_norm": 0.6035128831863403,
+      "learning_rate": 0.0004650210469508904,
+      "loss": 3.5588,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "grad_norm": 0.5737022161483765,
+      "learning_rate": 0.00046469724770642197,
+      "loss": 3.5574,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "eval_accuracy": 0.3639217651400465,
+      "eval_loss": 3.5595486164093018,
+      "eval_runtime": 183.5443,
+      "eval_samples_per_second": 98.129,
+      "eval_steps_per_second": 6.135,
+      "step": 21000
+    },
+    {
+      "epoch": 2.269541778975741,
+      "grad_norm": 0.62750244140625,
+      "learning_rate": 0.0004643734484619536,
+      "loss": 3.5534,
+      "step": 21050
+    },
+    {
+      "epoch": 2.274932614555256,
+      "grad_norm": 0.6026619076728821,
+      "learning_rate": 0.0004640496492174851,
+      "loss": 3.5427,
+      "step": 21100
+    },
+    {
+      "epoch": 2.280323450134771,
+      "grad_norm": 0.5918595790863037,
+      "learning_rate": 0.00046372584997301673,
+      "loss": 3.5706,
+      "step": 21150
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.5843772888183594,
+      "learning_rate": 0.0004634020507285483,
+      "loss": 3.55,
+      "step": 21200
+    },
+    {
+      "epoch": 2.2911051212938007,
+      "grad_norm": 0.5889182686805725,
+      "learning_rate": 0.0004630782514840798,
+      "loss": 3.5471,
+      "step": 21250
+    },
+    {
+      "epoch": 2.2964959568733154,
+      "grad_norm": 0.626357913017273,
+      "learning_rate": 0.0004627544522396114,
+      "loss": 3.558,
+      "step": 21300
+    },
+    {
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.5959623456001282,
+      "learning_rate": 0.00046243712898003233,
+      "loss": 3.5358,
+      "step": 21350
+    },
+    {
+      "epoch": 2.3072776280323453,
+      "grad_norm": 0.6014466881752014,
+      "learning_rate": 0.0004621133297355639,
+      "loss": 3.5723,
+      "step": 21400
+    },
+    {
+      "epoch": 2.31266846361186,
+      "grad_norm": 0.6045791506767273,
+      "learning_rate": 0.0004617895304910955,
+      "loss": 3.5327,
+      "step": 21450
+    },
+    {
+      "epoch": 2.3180592991913747,
+      "grad_norm": 0.5688396692276001,
+      "learning_rate": 0.00046146573124662704,
+      "loss": 3.5634,
+      "step": 21500
+    },
+    {
+      "epoch": 2.3234501347708894,
+      "grad_norm": 0.5908617973327637,
+      "learning_rate": 0.00046114193200215864,
+      "loss": 3.5617,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3288409703504045,
+      "grad_norm": 0.5954449772834778,
+      "learning_rate": 0.0004608181327576902,
+      "loss": 3.5373,
+      "step": 21600
+    },
+    {
+      "epoch": 2.334231805929919,
+      "grad_norm": 0.6025094389915466,
+      "learning_rate": 0.00046049433351322175,
+      "loss": 3.5556,
+      "step": 21650
+    },
+    {
+      "epoch": 2.339622641509434,
+      "grad_norm": 0.5918848514556885,
+      "learning_rate": 0.00046017053426875335,
+      "loss": 3.5713,
+      "step": 21700
+    },
+    {
+      "epoch": 2.3450134770889486,
+      "grad_norm": 0.6296148300170898,
+      "learning_rate": 0.0004598467350242849,
+      "loss": 3.5506,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3504043126684637,
+      "grad_norm": 0.5722049474716187,
+      "learning_rate": 0.0004595229357798165,
+      "loss": 3.5618,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3557951482479784,
+      "grad_norm": 0.6410542130470276,
+      "learning_rate": 0.00045919913653534806,
+      "loss": 3.5647,
+      "step": 21850
+    },
+    {
+      "epoch": 2.361185983827493,
+      "grad_norm": 0.5894669890403748,
+      "learning_rate": 0.00045887533729087966,
+      "loss": 3.5567,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3665768194070083,
+      "grad_norm": 0.5868241786956787,
+      "learning_rate": 0.00045855153804641116,
+      "loss": 3.5463,
+      "step": 21950
+    },
+    {
+      "epoch": 2.371967654986523,
+      "grad_norm": 0.5758694410324097,
+      "learning_rate": 0.0004582277388019427,
+      "loss": 3.542,
+      "step": 22000
+    },
+    {
+      "epoch": 2.371967654986523,
+      "eval_accuracy": 0.3651853982253289,
+      "eval_loss": 3.5488483905792236,
+      "eval_runtime": 184.0461,
+      "eval_samples_per_second": 97.861,
+      "eval_steps_per_second": 6.118,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 0.5851506590843201,
+      "learning_rate": 0.0004579039395574743,
+      "loss": 3.544,
+      "step": 22050
+    },
+    {
+      "epoch": 2.382749326145553,
+      "grad_norm": 0.5587830543518066,
+      "learning_rate": 0.00045758014031300586,
+      "loss": 3.5622,
+      "step": 22100
+    },
+    {
+      "epoch": 2.3881401617250675,
+      "grad_norm": 0.6127827167510986,
+      "learning_rate": 0.00045725634106853747,
+      "loss": 3.5548,
+      "step": 22150
+    },
+    {
+      "epoch": 2.393530997304582,
+      "grad_norm": 0.5660220980644226,
+      "learning_rate": 0.000456932541824069,
+      "loss": 3.5429,
+      "step": 22200
+    },
+    {
+      "epoch": 2.398921832884097,
+      "grad_norm": 0.8027902841567993,
+      "learning_rate": 0.0004566087425796006,
+      "loss": 3.5364,
+      "step": 22250
+    },
+    {
+      "epoch": 2.404312668463612,
+      "grad_norm": 0.6948328614234924,
+      "learning_rate": 0.0004562849433351322,
+      "loss": 3.5657,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4097035040431267,
+      "grad_norm": 0.5963208079338074,
+      "learning_rate": 0.0004559611440906638,
+      "loss": 3.5718,
+      "step": 22350
+    },
+    {
+      "epoch": 2.4150943396226414,
+      "grad_norm": 0.6638514995574951,
+      "learning_rate": 0.00045563734484619533,
+      "loss": 3.5562,
+      "step": 22400
+    },
+    {
+      "epoch": 2.420485175202156,
+      "grad_norm": 0.6002688407897949,
+      "learning_rate": 0.0004553135456017269,
+      "loss": 3.5648,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4258760107816713,
+      "grad_norm": 0.6028795838356018,
+      "learning_rate": 0.0004549897463572585,
+      "loss": 3.5599,
+      "step": 22500
+    },
+    {
+      "epoch": 2.431266846361186,
+      "grad_norm": 0.5684319138526917,
+      "learning_rate": 0.00045466594711279,
+      "loss": 3.5392,
+      "step": 22550
+    },
+    {
+      "epoch": 2.4366576819407006,
+      "grad_norm": 0.5978938341140747,
+      "learning_rate": 0.00045434214786832164,
+      "loss": 3.5577,
+      "step": 22600
+    },
+    {
+      "epoch": 2.442048517520216,
+      "grad_norm": 0.5744091272354126,
+      "learning_rate": 0.00045401834862385314,
+      "loss": 3.5539,
+      "step": 22650
+    },
+    {
+      "epoch": 2.4474393530997305,
+      "grad_norm": 0.615644097328186,
+      "learning_rate": 0.00045369454937938474,
+      "loss": 3.5605,
+      "step": 22700
+    },
+    {
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.585594892501831,
+      "learning_rate": 0.0004533707501349163,
+      "loss": 3.5482,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4582210242587603,
+      "grad_norm": 0.5654043555259705,
+      "learning_rate": 0.0004530469508904479,
+      "loss": 3.5457,
+      "step": 22800
+    },
+    {
+      "epoch": 2.463611859838275,
+      "grad_norm": 0.6015176773071289,
+      "learning_rate": 0.00045272315164597945,
+      "loss": 3.537,
+      "step": 22850
+    },
+    {
+      "epoch": 2.4690026954177897,
+      "grad_norm": 0.5832523107528687,
+      "learning_rate": 0.000452399352401511,
+      "loss": 3.55,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4743935309973044,
+      "grad_norm": 0.5774857997894287,
+      "learning_rate": 0.0004520755531570426,
+      "loss": 3.535,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "grad_norm": 0.5801119208335876,
+      "learning_rate": 0.00045175175391257415,
+      "loss": 3.5488,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "eval_accuracy": 0.3658704546829992,
+      "eval_loss": 3.538461685180664,
+      "eval_runtime": 183.5751,
+      "eval_samples_per_second": 98.112,
+      "eval_steps_per_second": 6.134,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4851752021563343,
+      "grad_norm": 0.6219615936279297,
+      "learning_rate": 0.00045142795466810576,
+      "loss": 3.5575,
+      "step": 23050
+    },
+    {
+      "epoch": 2.490566037735849,
+      "grad_norm": 0.5984650254249573,
+      "learning_rate": 0.0004511041554236373,
+      "loss": 3.5528,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4959568733153636,
+      "grad_norm": 0.6572667360305786,
+      "learning_rate": 0.0004507803561791689,
+      "loss": 3.5668,
+      "step": 23150
+    },
+    {
+      "epoch": 2.501347708894879,
+      "grad_norm": 0.584741473197937,
+      "learning_rate": 0.00045045655693470046,
+      "loss": 3.5567,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5067385444743935,
+      "grad_norm": 0.596365213394165,
+      "learning_rate": 0.00045013275769023207,
+      "loss": 3.5566,
+      "step": 23250
+    },
+    {
+      "epoch": 2.512129380053908,
+      "grad_norm": 0.5345525741577148,
+      "learning_rate": 0.0004498154344306529,
+      "loss": 3.5551,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5175202156334233,
+      "grad_norm": 0.5667235255241394,
+      "learning_rate": 0.0004494916351861845,
+      "loss": 3.5538,
+      "step": 23350
+    },
+    {
+      "epoch": 2.522911051212938,
+      "grad_norm": 0.6088711023330688,
+      "learning_rate": 0.00044916783594171607,
+      "loss": 3.5323,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5283018867924527,
+      "grad_norm": 0.617176353931427,
+      "learning_rate": 0.00044884403669724767,
+      "loss": 3.548,
+      "step": 23450
+    },
+    {
+      "epoch": 2.533692722371968,
+      "grad_norm": 0.6271255016326904,
+      "learning_rate": 0.0004485202374527792,
+      "loss": 3.5421,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5390835579514826,
+      "grad_norm": 0.5621574521064758,
+      "learning_rate": 0.00044819643820831083,
+      "loss": 3.545,
+      "step": 23550
+    },
+    {
+      "epoch": 2.5444743935309972,
+      "grad_norm": 0.6206479668617249,
+      "learning_rate": 0.0004478726389638424,
+      "loss": 3.5513,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5498652291105124,
+      "grad_norm": 0.543411374092102,
+      "learning_rate": 0.00044754883971937393,
+      "loss": 3.5373,
+      "step": 23650
+    },
+    {
+      "epoch": 2.555256064690027,
+      "grad_norm": 0.5803120732307434,
+      "learning_rate": 0.00044722504047490553,
+      "loss": 3.5272,
+      "step": 23700
+    },
+    {
+      "epoch": 2.560646900269542,
+      "grad_norm": 0.5784056186676025,
+      "learning_rate": 0.0004469012412304371,
+      "loss": 3.5345,
+      "step": 23750
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 0.5629208087921143,
+      "learning_rate": 0.0004465774419859687,
+      "loss": 3.5538,
+      "step": 23800
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.5571661591529846,
+      "learning_rate": 0.00044625364274150024,
+      "loss": 3.5402,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5768194070080863,
+      "grad_norm": 0.5853816866874695,
+      "learning_rate": 0.00044592984349703184,
+      "loss": 3.5423,
+      "step": 23900
+    },
+    {
+      "epoch": 2.582210242587601,
+      "grad_norm": 0.6156542301177979,
+      "learning_rate": 0.00044560604425256334,
+      "loss": 3.534,
+      "step": 23950
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "grad_norm": 0.5752986073493958,
+      "learning_rate": 0.000445282245008095,
+      "loss": 3.5386,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "eval_accuracy": 0.3673642145846727,
+      "eval_loss": 3.5285940170288086,
+      "eval_runtime": 183.9295,
+      "eval_samples_per_second": 97.923,
+      "eval_steps_per_second": 6.122,
+      "step": 24000
+    },
+    {
+      "epoch": 2.592991913746631,
+      "grad_norm": 0.5696158409118652,
+      "learning_rate": 0.0004449584457636265,
+      "loss": 3.5637,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5983827493261455,
+      "grad_norm": 0.5981115698814392,
+      "learning_rate": 0.00044463464651915805,
+      "loss": 3.5383,
+      "step": 24100
+    },
+    {
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.6076435446739197,
+      "learning_rate": 0.00044431084727468965,
+      "loss": 3.5499,
+      "step": 24150
+    },
+    {
+      "epoch": 2.6091644204851754,
+      "grad_norm": 0.573036253452301,
+      "learning_rate": 0.0004439870480302212,
+      "loss": 3.5361,
+      "step": 24200
+    },
+    {
+      "epoch": 2.61455525606469,
+      "grad_norm": 0.5571320652961731,
+      "learning_rate": 0.0004436632487857528,
+      "loss": 3.5534,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6199460916442048,
+      "grad_norm": 0.6245935559272766,
+      "learning_rate": 0.00044333944954128436,
+      "loss": 3.5447,
+      "step": 24300
+    },
+    {
+      "epoch": 2.62533692722372,
+      "grad_norm": 0.5823326706886292,
+      "learning_rate": 0.00044301565029681596,
+      "loss": 3.5386,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6307277628032346,
+      "grad_norm": 0.5950481295585632,
+      "learning_rate": 0.0004426918510523475,
+      "loss": 3.5422,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6361185983827493,
+      "grad_norm": 0.5968360304832458,
+      "learning_rate": 0.0004423680518078791,
+      "loss": 3.5412,
+      "step": 24450
+    },
+    {
+      "epoch": 2.641509433962264,
+      "grad_norm": 0.5554131865501404,
+      "learning_rate": 0.00044204425256341067,
+      "loss": 3.5503,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6469002695417787,
+      "grad_norm": 0.6161512732505798,
+      "learning_rate": 0.0004417204533189422,
+      "loss": 3.535,
+      "step": 24550
+    },
+    {
+      "epoch": 2.652291105121294,
+      "grad_norm": 0.6097846031188965,
+      "learning_rate": 0.0004413966540744738,
+      "loss": 3.5423,
+      "step": 24600
+    },
+    {
+      "epoch": 2.6576819407008085,
+      "grad_norm": 0.5795478224754333,
+      "learning_rate": 0.0004410728548300053,
+      "loss": 3.5383,
+      "step": 24650
+    },
+    {
+      "epoch": 2.6630727762803232,
+      "grad_norm": 0.5815718173980713,
+      "learning_rate": 0.0004407490555855369,
+      "loss": 3.5446,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6684636118598384,
+      "grad_norm": 0.6074739098548889,
+      "learning_rate": 0.0004404252563410685,
+      "loss": 3.5611,
+      "step": 24750
+    },
+    {
+      "epoch": 2.673854447439353,
+      "grad_norm": 0.585736095905304,
+      "learning_rate": 0.0004401014570966001,
+      "loss": 3.5542,
+      "step": 24800
+    },
+    {
+      "epoch": 2.6792452830188678,
+      "grad_norm": 0.5797416567802429,
+      "learning_rate": 0.00043977765785213163,
+      "loss": 3.5539,
+      "step": 24850
+    },
+    {
+      "epoch": 2.684636118598383,
+      "grad_norm": 0.6103894114494324,
+      "learning_rate": 0.0004394538586076632,
+      "loss": 3.5257,
+      "step": 24900
+    },
+    {
+      "epoch": 2.6900269541778976,
+      "grad_norm": 0.5875340700149536,
+      "learning_rate": 0.0004391300593631948,
+      "loss": 3.5451,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "grad_norm": 0.6124404072761536,
+      "learning_rate": 0.00043880626011872634,
+      "loss": 3.5235,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "eval_accuracy": 0.3682046446719177,
+      "eval_loss": 3.5184919834136963,
+      "eval_runtime": 183.9677,
+      "eval_samples_per_second": 97.903,
+      "eval_steps_per_second": 6.121,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7008086253369274,
+      "grad_norm": 0.5652927756309509,
+      "learning_rate": 0.00043848246087425794,
+      "loss": 3.5397,
+      "step": 25050
+    },
+    {
+      "epoch": 2.706199460916442,
+      "grad_norm": 0.6031397581100464,
+      "learning_rate": 0.0004381586616297895,
+      "loss": 3.5391,
+      "step": 25100
+    },
+    {
+      "epoch": 2.711590296495957,
+      "grad_norm": 0.5758184790611267,
+      "learning_rate": 0.0004378348623853211,
+      "loss": 3.5408,
+      "step": 25150
+    },
+    {
+      "epoch": 2.7169811320754715,
+      "grad_norm": 0.6211733818054199,
+      "learning_rate": 0.00043751106314085265,
+      "loss": 3.5427,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7223719676549867,
+      "grad_norm": 0.6134040951728821,
+      "learning_rate": 0.00043718726389638425,
+      "loss": 3.5337,
+      "step": 25250
+    },
+    {
+      "epoch": 2.7277628032345014,
+      "grad_norm": 0.6205156445503235,
+      "learning_rate": 0.00043686346465191575,
+      "loss": 3.5399,
+      "step": 25300
+    },
+    {
+      "epoch": 2.733153638814016,
+      "grad_norm": 0.610084593296051,
+      "learning_rate": 0.0004365396654074473,
+      "loss": 3.5354,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7385444743935308,
+      "grad_norm": 0.5997064113616943,
+      "learning_rate": 0.0004362158661629789,
+      "loss": 3.5302,
+      "step": 25400
+    },
+    {
+      "epoch": 2.743935309973046,
+      "grad_norm": 0.5822950005531311,
+      "learning_rate": 0.00043589206691851045,
+      "loss": 3.5462,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7493261455525606,
+      "grad_norm": 0.5831359624862671,
+      "learning_rate": 0.00043556826767404206,
+      "loss": 3.524,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.5996923446655273,
+      "learning_rate": 0.0004352444684295736,
+      "loss": 3.5168,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7601078167115904,
+      "grad_norm": 0.5952886939048767,
+      "learning_rate": 0.0004349206691851052,
+      "loss": 3.5214,
+      "step": 25600
+    },
+    {
+      "epoch": 2.765498652291105,
+      "grad_norm": 0.5794487595558167,
+      "learning_rate": 0.00043459686994063676,
+      "loss": 3.5289,
+      "step": 25650
+    },
+    {
+      "epoch": 2.77088948787062,
+      "grad_norm": 0.5602608323097229,
+      "learning_rate": 0.00043427307069616837,
+      "loss": 3.5167,
+      "step": 25700
+    },
+    {
+      "epoch": 2.776280323450135,
+      "grad_norm": 0.5887024998664856,
+      "learning_rate": 0.0004339492714516999,
+      "loss": 3.5206,
+      "step": 25750
+    },
+    {
+      "epoch": 2.7816711590296497,
+      "grad_norm": 0.5707184076309204,
+      "learning_rate": 0.00043362547220723147,
+      "loss": 3.5336,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7870619946091644,
+      "grad_norm": 0.5876629948616028,
+      "learning_rate": 0.0004333016729627631,
+      "loss": 3.5208,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7924528301886795,
+      "grad_norm": 0.5893009305000305,
+      "learning_rate": 0.0004329778737182946,
+      "loss": 3.5209,
+      "step": 25900
+    },
+    {
+      "epoch": 2.797843665768194,
+      "grad_norm": 0.5993021726608276,
+      "learning_rate": 0.00043265407447382623,
+      "loss": 3.5323,
+      "step": 25950
+    },
+    {
+      "epoch": 2.803234501347709,
+      "grad_norm": 0.6446514129638672,
+      "learning_rate": 0.0004323302752293577,
+      "loss": 3.5362,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803234501347709,
+      "eval_accuracy": 0.36895228519296913,
+      "eval_loss": 3.51017689704895,
+      "eval_runtime": 183.4946,
+      "eval_samples_per_second": 98.155,
+      "eval_steps_per_second": 6.136,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8086253369272236,
+      "grad_norm": 0.6374831199645996,
+      "learning_rate": 0.00043200647598488933,
+      "loss": 3.5372,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8140161725067383,
+      "grad_norm": 0.5802428722381592,
+      "learning_rate": 0.0004316826767404209,
+      "loss": 3.5188,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8194070080862534,
+      "grad_norm": 0.6247591972351074,
+      "learning_rate": 0.0004313588774959525,
+      "loss": 3.5266,
+      "step": 26150
+    },
+    {
+      "epoch": 2.824797843665768,
+      "grad_norm": 0.6217538714408875,
+      "learning_rate": 0.00043103507825148404,
+      "loss": 3.5239,
+      "step": 26200
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.6069292426109314,
+      "learning_rate": 0.0004307112790070156,
+      "loss": 3.5186,
+      "step": 26250
+    },
+    {
+      "epoch": 2.835579514824798,
+      "grad_norm": 0.5810019969940186,
+      "learning_rate": 0.0004303874797625472,
+      "loss": 3.5324,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8409703504043127,
+      "grad_norm": 0.5871881246566772,
+      "learning_rate": 0.00043006368051807874,
+      "loss": 3.537,
+      "step": 26350
+    },
+    {
+      "epoch": 2.8463611859838274,
+      "grad_norm": 0.5998797416687012,
+      "learning_rate": 0.00042973988127361035,
+      "loss": 3.5245,
+      "step": 26400
+    },
+    {
+      "epoch": 2.8517520215633425,
+      "grad_norm": 0.57145756483078,
+      "learning_rate": 0.0004294160820291419,
+      "loss": 3.5258,
+      "step": 26450
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.6319313049316406,
+      "learning_rate": 0.0004290922827846735,
+      "loss": 3.5229,
+      "step": 26500
+    },
+    {
+      "epoch": 2.862533692722372,
+      "grad_norm": 0.628434419631958,
+      "learning_rate": 0.00042876848354020505,
+      "loss": 3.5158,
+      "step": 26550
+    },
+    {
+      "epoch": 2.867924528301887,
+      "grad_norm": 0.5832825899124146,
+      "learning_rate": 0.00042844468429573655,
+      "loss": 3.5275,
+      "step": 26600
+    },
+    {
+      "epoch": 2.8733153638814017,
+      "grad_norm": 0.5889964699745178,
+      "learning_rate": 0.00042812088505126815,
+      "loss": 3.5215,
+      "step": 26650
+    },
+    {
+      "epoch": 2.8787061994609164,
+      "grad_norm": 0.5895435214042664,
+      "learning_rate": 0.0004277970858067997,
+      "loss": 3.5213,
+      "step": 26700
+    },
+    {
+      "epoch": 2.884097035040431,
+      "grad_norm": 0.5762864351272583,
+      "learning_rate": 0.0004274732865623313,
+      "loss": 3.5393,
+      "step": 26750
+    },
+    {
+      "epoch": 2.889487870619946,
+      "grad_norm": 0.6543927192687988,
+      "learning_rate": 0.00042714948731786286,
+      "loss": 3.527,
+      "step": 26800
+    },
+    {
+      "epoch": 2.894878706199461,
+      "grad_norm": 0.5806078910827637,
+      "learning_rate": 0.00042682568807339447,
+      "loss": 3.5189,
+      "step": 26850
+    },
+    {
+      "epoch": 2.9002695417789757,
+      "grad_norm": 0.5909567475318909,
+      "learning_rate": 0.000426501888828926,
+      "loss": 3.542,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.5814721584320068,
+      "learning_rate": 0.0004261780895844576,
+      "loss": 3.5356,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "grad_norm": 0.5588226318359375,
+      "learning_rate": 0.00042585429033998917,
+      "loss": 3.5198,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "eval_accuracy": 0.370136601670147,
+      "eval_loss": 3.5008394718170166,
+      "eval_runtime": 183.773,
+      "eval_samples_per_second": 98.007,
+      "eval_steps_per_second": 6.127,
+      "step": 27000
+    },
+    {
+      "epoch": 2.91644204851752,
+      "grad_norm": 0.6280841827392578,
+      "learning_rate": 0.0004255304910955207,
+      "loss": 3.5369,
+      "step": 27050
+    },
+    {
+      "epoch": 2.921832884097035,
+      "grad_norm": 0.5633023381233215,
+      "learning_rate": 0.0004252066918510523,
+      "loss": 3.5428,
+      "step": 27100
+    },
+    {
+      "epoch": 2.92722371967655,
+      "grad_norm": 0.6330850124359131,
+      "learning_rate": 0.0004248828926065839,
+      "loss": 3.5255,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9326145552560647,
+      "grad_norm": 0.5991080403327942,
+      "learning_rate": 0.0004245590933621155,
+      "loss": 3.5175,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9380053908355794,
+      "grad_norm": 0.5871846079826355,
+      "learning_rate": 0.00042423529411764703,
+      "loss": 3.53,
+      "step": 27250
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 0.6111835837364197,
+      "learning_rate": 0.00042391797085806793,
+      "loss": 3.5255,
+      "step": 27300
+    },
+    {
+      "epoch": 2.9487870619946093,
+      "grad_norm": 0.5987087488174438,
+      "learning_rate": 0.0004235941716135995,
+      "loss": 3.5191,
+      "step": 27350
+    },
+    {
+      "epoch": 2.954177897574124,
+      "grad_norm": 0.5940400958061218,
+      "learning_rate": 0.0004232703723691311,
+      "loss": 3.5237,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9595687331536387,
+      "grad_norm": 0.5561420321464539,
+      "learning_rate": 0.00042294657312466264,
+      "loss": 3.5192,
+      "step": 27450
+    },
+    {
+      "epoch": 2.964959568733154,
+      "grad_norm": 0.5664126873016357,
+      "learning_rate": 0.00042262277388019424,
+      "loss": 3.5263,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9703504043126685,
+      "grad_norm": 0.5555766820907593,
+      "learning_rate": 0.0004222989746357258,
+      "loss": 3.5287,
+      "step": 27550
+    },
+    {
+      "epoch": 2.975741239892183,
+      "grad_norm": 0.5817607641220093,
+      "learning_rate": 0.0004219751753912574,
+      "loss": 3.5281,
+      "step": 27600
+    },
+    {
+      "epoch": 2.981132075471698,
+      "grad_norm": 0.5847254991531372,
+      "learning_rate": 0.00042165137614678895,
+      "loss": 3.5301,
+      "step": 27650
+    },
+    {
+      "epoch": 2.986522911051213,
+      "grad_norm": 0.616742730140686,
+      "learning_rate": 0.00042132757690232055,
+      "loss": 3.5214,
+      "step": 27700
+    },
+    {
+      "epoch": 2.9919137466307277,
+      "grad_norm": 0.5673505663871765,
+      "learning_rate": 0.0004210037776578521,
+      "loss": 3.5315,
+      "step": 27750
+    },
+    {
+      "epoch": 2.9973045822102424,
+      "grad_norm": 0.6064647436141968,
+      "learning_rate": 0.00042067997841338365,
+      "loss": 3.5287,
+      "step": 27800
+    },
+    {
+      "epoch": 3.0026954177897576,
+      "grad_norm": 0.5815552473068237,
+      "learning_rate": 0.00042035617916891526,
+      "loss": 3.4745,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0080862533692723,
+      "grad_norm": 0.6101829409599304,
+      "learning_rate": 0.0004200388559093362,
+      "loss": 3.424,
+      "step": 27900
+    },
+    {
+      "epoch": 3.013477088948787,
+      "grad_norm": 0.6867734789848328,
+      "learning_rate": 0.0004197150566648677,
+      "loss": 3.4251,
+      "step": 27950
+    },
+    {
+      "epoch": 3.018867924528302,
+      "grad_norm": 0.5732295513153076,
+      "learning_rate": 0.00041939125742039936,
+      "loss": 3.4437,
+      "step": 28000
+    },
+    {
+      "epoch": 3.018867924528302,
+      "eval_accuracy": 0.37087022596867036,
+      "eval_loss": 3.500011682510376,
+      "eval_runtime": 183.9273,
+      "eval_samples_per_second": 97.925,
+      "eval_steps_per_second": 6.122,
+      "step": 28000
+    },
+    {
+      "epoch": 3.024258760107817,
+      "grad_norm": 0.634990930557251,
+      "learning_rate": 0.00041906745817593086,
+      "loss": 3.4381,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0296495956873315,
+      "grad_norm": 0.6349164247512817,
+      "learning_rate": 0.00041874365893146247,
+      "loss": 3.4456,
+      "step": 28100
+    },
+    {
+      "epoch": 3.035040431266846,
+      "grad_norm": 0.6146011352539062,
+      "learning_rate": 0.000418419859686994,
+      "loss": 3.4334,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0404312668463613,
+      "grad_norm": 0.605857253074646,
+      "learning_rate": 0.00041809606044252557,
+      "loss": 3.4403,
+      "step": 28200
+    },
+    {
+      "epoch": 3.045822102425876,
+      "grad_norm": 0.6127777099609375,
+      "learning_rate": 0.00041777226119805717,
+      "loss": 3.4213,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0512129380053907,
+      "grad_norm": 0.5787948369979858,
+      "learning_rate": 0.0004174484619535887,
+      "loss": 3.4423,
+      "step": 28300
+    },
+    {
+      "epoch": 3.056603773584906,
+      "grad_norm": 0.6266268491744995,
+      "learning_rate": 0.00041712466270912033,
+      "loss": 3.4176,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0619946091644206,
+      "grad_norm": 0.599656879901886,
+      "learning_rate": 0.0004168008634646519,
+      "loss": 3.446,
+      "step": 28400
+    },
+    {
+      "epoch": 3.0673854447439353,
+      "grad_norm": 0.6230536699295044,
+      "learning_rate": 0.0004164770642201835,
+      "loss": 3.4155,
+      "step": 28450
+    },
+    {
+      "epoch": 3.07277628032345,
+      "grad_norm": 0.6091195344924927,
+      "learning_rate": 0.00041615326497571503,
+      "loss": 3.4402,
+      "step": 28500
+    },
+    {
+      "epoch": 3.078167115902965,
+      "grad_norm": 0.6610572338104248,
+      "learning_rate": 0.0004158294657312466,
+      "loss": 3.4526,
+      "step": 28550
+    },
+    {
+      "epoch": 3.08355795148248,
+      "grad_norm": 0.6236415505409241,
+      "learning_rate": 0.0004155056664867782,
+      "loss": 3.4382,
+      "step": 28600
+    },
+    {
+      "epoch": 3.0889487870619945,
+      "grad_norm": 0.6569049954414368,
+      "learning_rate": 0.0004151818672423097,
+      "loss": 3.4493,
+      "step": 28650
+    },
+    {
+      "epoch": 3.0943396226415096,
+      "grad_norm": 0.5730276107788086,
+      "learning_rate": 0.0004148580679978413,
+      "loss": 3.4455,
+      "step": 28700
+    },
+    {
+      "epoch": 3.0997304582210243,
+      "grad_norm": 0.6144906282424927,
+      "learning_rate": 0.00041453426875337284,
+      "loss": 3.4714,
+      "step": 28750
+    },
+    {
+      "epoch": 3.105121293800539,
+      "grad_norm": 0.637140691280365,
+      "learning_rate": 0.00041421046950890445,
+      "loss": 3.4328,
+      "step": 28800
+    },
+    {
+      "epoch": 3.1105121293800537,
+      "grad_norm": 0.6305104494094849,
+      "learning_rate": 0.000413886670264436,
+      "loss": 3.4389,
+      "step": 28850
+    },
+    {
+      "epoch": 3.115902964959569,
+      "grad_norm": 0.5751136541366577,
+      "learning_rate": 0.0004135628710199676,
+      "loss": 3.4363,
+      "step": 28900
+    },
+    {
+      "epoch": 3.1212938005390836,
+      "grad_norm": 0.5962830781936646,
+      "learning_rate": 0.00041323907177549915,
+      "loss": 3.466,
+      "step": 28950
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "grad_norm": 0.6354585886001587,
+      "learning_rate": 0.0004129152725310307,
+      "loss": 3.4374,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "eval_accuracy": 0.3712109614248566,
+      "eval_loss": 3.4914464950561523,
+      "eval_runtime": 183.9704,
+      "eval_samples_per_second": 97.902,
+      "eval_steps_per_second": 6.121,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1320754716981134,
+      "grad_norm": 0.6473236680030823,
+      "learning_rate": 0.0004125914732865623,
+      "loss": 3.4544,
+      "step": 29050
+    },
+    {
+      "epoch": 3.137466307277628,
+      "grad_norm": 0.5821991562843323,
+      "learning_rate": 0.00041226767404209386,
+      "loss": 3.4583,
+      "step": 29100
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.6483465433120728,
+      "learning_rate": 0.00041194387479762546,
+      "loss": 3.4477,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1482479784366575,
+      "grad_norm": 0.5836642384529114,
+      "learning_rate": 0.000411620075553157,
+      "loss": 3.4566,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1536388140161726,
+      "grad_norm": 0.5946670174598694,
+      "learning_rate": 0.0004112962763086886,
+      "loss": 3.4603,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1590296495956873,
+      "grad_norm": 0.6464475393295288,
+      "learning_rate": 0.0004109724770642201,
+      "loss": 3.4636,
+      "step": 29300
+    },
+    {
+      "epoch": 3.164420485175202,
+      "grad_norm": 0.6841476559638977,
+      "learning_rate": 0.00041064867781975177,
+      "loss": 3.4534,
+      "step": 29350
+    },
+    {
+      "epoch": 3.169811320754717,
+      "grad_norm": 0.6808069348335266,
+      "learning_rate": 0.00041032487857528327,
+      "loss": 3.4451,
+      "step": 29400
+    },
+    {
+      "epoch": 3.175202156334232,
+      "grad_norm": 0.573727011680603,
+      "learning_rate": 0.0004100010793308148,
+      "loss": 3.4316,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1805929919137466,
+      "grad_norm": 0.6131392121315002,
+      "learning_rate": 0.0004096772800863464,
+      "loss": 3.4556,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1859838274932613,
+      "grad_norm": 0.5942975878715515,
+      "learning_rate": 0.000409353480841878,
+      "loss": 3.4407,
+      "step": 29550
+    },
+    {
+      "epoch": 3.1913746630727764,
+      "grad_norm": 0.5979737043380737,
+      "learning_rate": 0.0004090296815974096,
+      "loss": 3.4495,
+      "step": 29600
+    },
+    {
+      "epoch": 3.196765498652291,
+      "grad_norm": 0.647235631942749,
+      "learning_rate": 0.00040870588235294113,
+      "loss": 3.4671,
+      "step": 29650
+    },
+    {
+      "epoch": 3.202156334231806,
+      "grad_norm": 0.6004102230072021,
+      "learning_rate": 0.00040838208310847273,
+      "loss": 3.4479,
+      "step": 29700
+    },
+    {
+      "epoch": 3.207547169811321,
+      "grad_norm": 0.622308611869812,
+      "learning_rate": 0.0004080582838640043,
+      "loss": 3.4591,
+      "step": 29750
+    },
+    {
+      "epoch": 3.2129380053908356,
+      "grad_norm": 0.6031266450881958,
+      "learning_rate": 0.0004077344846195359,
+      "loss": 3.4676,
+      "step": 29800
+    },
+    {
+      "epoch": 3.2183288409703503,
+      "grad_norm": 0.6271257400512695,
+      "learning_rate": 0.00040741068537506744,
+      "loss": 3.4715,
+      "step": 29850
+    },
+    {
+      "epoch": 3.223719676549865,
+      "grad_norm": 0.6046510934829712,
+      "learning_rate": 0.000407086886130599,
+      "loss": 3.4383,
+      "step": 29900
+    },
+    {
+      "epoch": 3.22911051212938,
+      "grad_norm": 0.5502532720565796,
+      "learning_rate": 0.0004067630868861306,
+      "loss": 3.441,
+      "step": 29950
+    },
+    {
+      "epoch": 3.234501347708895,
+      "grad_norm": 0.5909226536750793,
+      "learning_rate": 0.0004064392876416621,
+      "loss": 3.4672,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234501347708895,
+      "eval_accuracy": 0.37218318746502055,
+      "eval_loss": 3.486082077026367,
+      "eval_runtime": 183.6809,
+      "eval_samples_per_second": 98.056,
+      "eval_steps_per_second": 6.13,
+      "step": 30000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 92750,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.508207538176e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}