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": 3.491060256958008,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M__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": 2.385856866836548,
+      "learning_rate": 0.000276,
+      "loss": 9.0223,
+      "step": 50
+    },
+    {
+      "epoch": 0.01078167115902965,
+      "grad_norm": 2.4063053131103516,
+      "learning_rate": 0.0005759999999999999,
+      "loss": 6.944,
+      "step": 100
+    },
+    {
+      "epoch": 0.016172506738544475,
+      "grad_norm": 1.2234184741973877,
+      "learning_rate": 0.000599702104695089,
+      "loss": 6.4879,
+      "step": 150
+    },
+    {
+      "epoch": 0.0215633423180593,
+      "grad_norm": 1.0247325897216797,
+      "learning_rate": 0.0005993783054506205,
+      "loss": 6.2173,
+      "step": 200
+    },
+    {
+      "epoch": 0.026954177897574125,
+      "grad_norm": 1.8562301397323608,
+      "learning_rate": 0.0005990545062061521,
+      "loss": 6.0821,
+      "step": 250
+    },
+    {
+      "epoch": 0.03234501347708895,
+      "grad_norm": 1.6646568775177002,
+      "learning_rate": 0.0005987307069616836,
+      "loss": 5.957,
+      "step": 300
+    },
+    {
+      "epoch": 0.03773584905660377,
+      "grad_norm": 0.8979423642158508,
+      "learning_rate": 0.0005984069077172153,
+      "loss": 5.8884,
+      "step": 350
+    },
+    {
+      "epoch": 0.0431266846361186,
+      "grad_norm": 1.5213276147842407,
+      "learning_rate": 0.0005980831084727469,
+      "loss": 5.8265,
+      "step": 400
+    },
+    {
+      "epoch": 0.04851752021563342,
+      "grad_norm": 1.5135802030563354,
+      "learning_rate": 0.0005977593092282784,
+      "loss": 5.7629,
+      "step": 450
+    },
+    {
+      "epoch": 0.05390835579514825,
+      "grad_norm": 1.3815815448760986,
+      "learning_rate": 0.00059743550998381,
+      "loss": 5.6521,
+      "step": 500
+    },
+    {
+      "epoch": 0.05929919137466307,
+      "grad_norm": 0.8758553266525269,
+      "learning_rate": 0.0005971117107393416,
+      "loss": 5.602,
+      "step": 550
+    },
+    {
+      "epoch": 0.0646900269541779,
+      "grad_norm": 1.5150190591812134,
+      "learning_rate": 0.0005967879114948732,
+      "loss": 5.5349,
+      "step": 600
+    },
+    {
+      "epoch": 0.07008086253369272,
+      "grad_norm": 1.2760694026947021,
+      "learning_rate": 0.0005964641122504047,
+      "loss": 5.4673,
+      "step": 650
+    },
+    {
+      "epoch": 0.07547169811320754,
+      "grad_norm": 1.3729991912841797,
+      "learning_rate": 0.0005961403130059363,
+      "loss": 5.3942,
+      "step": 700
+    },
+    {
+      "epoch": 0.08086253369272237,
+      "grad_norm": 1.3321388959884644,
+      "learning_rate": 0.0005958165137614678,
+      "loss": 5.3261,
+      "step": 750
+    },
+    {
+      "epoch": 0.0862533692722372,
+      "grad_norm": 1.0924150943756104,
+      "learning_rate": 0.0005954927145169995,
+      "loss": 5.2547,
+      "step": 800
+    },
+    {
+      "epoch": 0.09164420485175202,
+      "grad_norm": 1.0701996088027954,
+      "learning_rate": 0.0005951689152725309,
+      "loss": 5.2296,
+      "step": 850
+    },
+    {
+      "epoch": 0.09703504043126684,
+      "grad_norm": 1.1670644283294678,
+      "learning_rate": 0.0005948451160280626,
+      "loss": 5.1749,
+      "step": 900
+    },
+    {
+      "epoch": 0.10242587601078167,
+      "grad_norm": 0.833484411239624,
+      "learning_rate": 0.0005945213167835941,
+      "loss": 5.1112,
+      "step": 950
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "grad_norm": 1.1527302265167236,
+      "learning_rate": 0.0005941975175391257,
+      "loss": 5.1167,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "eval_accuracy": 0.22651139154904357,
+      "eval_loss": 5.03156852722168,
+      "eval_runtime": 184.3163,
+      "eval_samples_per_second": 97.718,
+      "eval_steps_per_second": 6.109,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11320754716981132,
+      "grad_norm": 0.9724206924438477,
+      "learning_rate": 0.0005938737182946572,
+      "loss": 5.0469,
+      "step": 1050
+    },
+    {
+      "epoch": 0.11859838274932614,
+      "grad_norm": 1.3919302225112915,
+      "learning_rate": 0.0005935499190501888,
+      "loss": 5.0082,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12398921832884097,
+      "grad_norm": 1.0906733274459839,
+      "learning_rate": 0.0005932261198057204,
+      "loss": 5.0148,
+      "step": 1150
+    },
+    {
+      "epoch": 0.1293800539083558,
+      "grad_norm": 0.9858213663101196,
+      "learning_rate": 0.000592902320561252,
+      "loss": 4.9728,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1347708894878706,
+      "grad_norm": 0.9534234404563904,
+      "learning_rate": 0.0005925785213167835,
+      "loss": 4.9499,
+      "step": 1250
+    },
+    {
+      "epoch": 0.14016172506738545,
+      "grad_norm": 0.9149211049079895,
+      "learning_rate": 0.0005922547220723151,
+      "loss": 4.9035,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14555256064690028,
+      "grad_norm": 1.0197161436080933,
+      "learning_rate": 0.0005919309228278468,
+      "loss": 4.8524,
+      "step": 1350
+    },
+    {
+      "epoch": 0.1509433962264151,
+      "grad_norm": 1.1303359270095825,
+      "learning_rate": 0.0005916071235833783,
+      "loss": 4.8606,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15633423180592992,
+      "grad_norm": 1.1484166383743286,
+      "learning_rate": 0.0005912833243389097,
+      "loss": 4.8134,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16172506738544473,
+      "grad_norm": 0.8566966652870178,
+      "learning_rate": 0.0005909595250944414,
+      "loss": 4.8164,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16711590296495957,
+      "grad_norm": 1.603456735610962,
+      "learning_rate": 0.000590635725849973,
+      "loss": 4.8012,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1725067385444744,
+      "grad_norm": 1.0490790605545044,
+      "learning_rate": 0.0005903119266055045,
+      "loss": 4.7774,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1778975741239892,
+      "grad_norm": 1.2788410186767578,
+      "learning_rate": 0.0005899881273610361,
+      "loss": 4.7472,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18328840970350405,
+      "grad_norm": 1.3455950021743774,
+      "learning_rate": 0.0005896643281165677,
+      "loss": 4.7321,
+      "step": 1700
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 1.0504180192947388,
+      "learning_rate": 0.0005893405288720993,
+      "loss": 4.6779,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1940700808625337,
+      "grad_norm": 0.8884590268135071,
+      "learning_rate": 0.0005890167296276308,
+      "loss": 4.7042,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19946091644204852,
+      "grad_norm": 1.235971212387085,
+      "learning_rate": 0.0005886929303831624,
+      "loss": 4.6545,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20485175202156333,
+      "grad_norm": 0.8869856595993042,
+      "learning_rate": 0.0005883691311386939,
+      "loss": 4.6304,
+      "step": 1900
+    },
+    {
+      "epoch": 0.21024258760107817,
+      "grad_norm": 0.8559747338294983,
+      "learning_rate": 0.0005880453318942256,
+      "loss": 4.6107,
+      "step": 1950
+    },
+    {
+      "epoch": 0.215633423180593,
+      "grad_norm": 0.8249241709709167,
+      "learning_rate": 0.0005877215326497571,
+      "loss": 4.5939,
+      "step": 2000
+    },
+    {
+      "epoch": 0.215633423180593,
+      "eval_accuracy": 0.27012444341200054,
+      "eval_loss": 4.512342929840088,
+      "eval_runtime": 183.3017,
+      "eval_samples_per_second": 98.259,
+      "eval_steps_per_second": 6.143,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2210242587601078,
+      "grad_norm": 1.0064095258712769,
+      "learning_rate": 0.0005873977334052887,
+      "loss": 4.5762,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22641509433962265,
+      "grad_norm": 0.8109007477760315,
+      "learning_rate": 0.0005870739341608202,
+      "loss": 4.5586,
+      "step": 2100
+    },
+    {
+      "epoch": 0.23180592991913745,
+      "grad_norm": 0.8962969183921814,
+      "learning_rate": 0.0005867501349163519,
+      "loss": 4.5323,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2371967654986523,
+      "grad_norm": 0.8573135733604431,
+      "learning_rate": 0.0005864263356718833,
+      "loss": 4.5247,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24258760107816713,
+      "grad_norm": 0.8410763740539551,
+      "learning_rate": 0.000586102536427415,
+      "loss": 4.5036,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24797843665768193,
+      "grad_norm": 0.7351256012916565,
+      "learning_rate": 0.0005857787371829465,
+      "loss": 4.4966,
+      "step": 2300
+    },
+    {
+      "epoch": 0.25336927223719674,
+      "grad_norm": 1.2595137357711792,
+      "learning_rate": 0.0005854549379384781,
+      "loss": 4.4755,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2587601078167116,
+      "grad_norm": 0.8755801320075989,
+      "learning_rate": 0.0005851311386940096,
+      "loss": 4.4587,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2641509433962264,
+      "grad_norm": 0.7965346574783325,
+      "learning_rate": 0.0005848073394495412,
+      "loss": 4.4459,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2695417789757412,
+      "grad_norm": 0.9087278842926025,
+      "learning_rate": 0.0005844835402050728,
+      "loss": 4.4255,
+      "step": 2500
+    },
+    {
+      "epoch": 0.2749326145552561,
+      "grad_norm": 0.754734218120575,
+      "learning_rate": 0.0005841597409606044,
+      "loss": 4.3955,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2803234501347709,
+      "grad_norm": 0.9010357856750488,
+      "learning_rate": 0.000583835941716136,
+      "loss": 4.4084,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 0.8619740009307861,
+      "learning_rate": 0.0005835121424716675,
+      "loss": 4.3899,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29110512129380056,
+      "grad_norm": 0.9053239226341248,
+      "learning_rate": 0.0005831883432271992,
+      "loss": 4.3775,
+      "step": 2700
+    },
+    {
+      "epoch": 0.29649595687331537,
+      "grad_norm": 0.8653854727745056,
+      "learning_rate": 0.0005828645439827307,
+      "loss": 4.3731,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3018867924528302,
+      "grad_norm": 0.7907358407974243,
+      "learning_rate": 0.0005825407447382622,
+      "loss": 4.3277,
+      "step": 2800
+    },
+    {
+      "epoch": 0.30727762803234504,
+      "grad_norm": 0.692861795425415,
+      "learning_rate": 0.0005822169454937938,
+      "loss": 4.3532,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31266846361185985,
+      "grad_norm": 0.7582786083221436,
+      "learning_rate": 0.0005818931462493254,
+      "loss": 4.3195,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31805929919137466,
+      "grad_norm": 0.8971874117851257,
+      "learning_rate": 0.0005815693470048569,
+      "loss": 4.3277,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "grad_norm": 0.7669051885604858,
+      "learning_rate": 0.0005812455477603885,
+      "loss": 4.3174,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "eval_accuracy": 0.29804573656390243,
+      "eval_loss": 4.239420413970947,
+      "eval_runtime": 184.1932,
+      "eval_samples_per_second": 97.783,
+      "eval_steps_per_second": 6.113,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3288409703504043,
+      "grad_norm": 0.6557288765907288,
+      "learning_rate": 0.0005809217485159201,
+      "loss": 4.3175,
+      "step": 3050
+    },
+    {
+      "epoch": 0.33423180592991913,
+      "grad_norm": 0.678659975528717,
+      "learning_rate": 0.0005805979492714517,
+      "loss": 4.2974,
+      "step": 3100
+    },
+    {
+      "epoch": 0.33962264150943394,
+      "grad_norm": 0.8476620316505432,
+      "learning_rate": 0.0005802741500269832,
+      "loss": 4.2748,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3450134770889488,
+      "grad_norm": 0.6797805428504944,
+      "learning_rate": 0.0005799503507825148,
+      "loss": 4.2811,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3504043126684636,
+      "grad_norm": 0.6718155145645142,
+      "learning_rate": 0.0005796265515380463,
+      "loss": 4.2701,
+      "step": 3250
+    },
+    {
+      "epoch": 0.3557951482479784,
+      "grad_norm": 0.9261127710342407,
+      "learning_rate": 0.000579302752293578,
+      "loss": 4.2727,
+      "step": 3300
+    },
+    {
+      "epoch": 0.3611859838274933,
+      "grad_norm": 0.8913077116012573,
+      "learning_rate": 0.0005789789530491095,
+      "loss": 4.2472,
+      "step": 3350
+    },
+    {
+      "epoch": 0.3665768194070081,
+      "grad_norm": 0.7431011199951172,
+      "learning_rate": 0.0005786551538046411,
+      "loss": 4.2378,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3719676549865229,
+      "grad_norm": 0.6269802451133728,
+      "learning_rate": 0.0005783313545601726,
+      "loss": 4.2305,
+      "step": 3450
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.8824395537376404,
+      "learning_rate": 0.0005780075553157043,
+      "loss": 4.2307,
+      "step": 3500
+    },
+    {
+      "epoch": 0.38274932614555257,
+      "grad_norm": 0.6562435626983643,
+      "learning_rate": 0.0005776837560712357,
+      "loss": 4.2244,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3881401617250674,
+      "grad_norm": 0.8994774222373962,
+      "learning_rate": 0.0005773599568267673,
+      "loss": 4.2167,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3935309973045822,
+      "grad_norm": 0.7905921339988708,
+      "learning_rate": 0.0005770361575822989,
+      "loss": 4.2048,
+      "step": 3650
+    },
+    {
+      "epoch": 0.39892183288409705,
+      "grad_norm": 0.7321567535400391,
+      "learning_rate": 0.0005767123583378305,
+      "loss": 4.1956,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40431266846361186,
+      "grad_norm": 0.697921872138977,
+      "learning_rate": 0.000576388559093362,
+      "loss": 4.1971,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40970350404312667,
+      "grad_norm": 0.6758502125740051,
+      "learning_rate": 0.0005760647598488936,
+      "loss": 4.1617,
+      "step": 3800
+    },
+    {
+      "epoch": 0.41509433962264153,
+      "grad_norm": 0.7216224670410156,
+      "learning_rate": 0.0005757409606044253,
+      "loss": 4.1734,
+      "step": 3850
+    },
+    {
+      "epoch": 0.42048517520215634,
+      "grad_norm": 0.7215520143508911,
+      "learning_rate": 0.0005754171613599568,
+      "loss": 4.1892,
+      "step": 3900
+    },
+    {
+      "epoch": 0.42587601078167114,
+      "grad_norm": 1.9430748224258423,
+      "learning_rate": 0.0005750933621154884,
+      "loss": 4.1593,
+      "step": 3950
+    },
+    {
+      "epoch": 0.431266846361186,
+      "grad_norm": 0.6709734797477722,
+      "learning_rate": 0.0005747695628710199,
+      "loss": 4.1616,
+      "step": 4000
+    },
+    {
+      "epoch": 0.431266846361186,
+      "eval_accuracy": 0.31166461548123287,
+      "eval_loss": 4.095009803771973,
+      "eval_runtime": 184.2427,
+      "eval_samples_per_second": 97.757,
+      "eval_steps_per_second": 6.112,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4366576819407008,
+      "grad_norm": 0.8147162795066833,
+      "learning_rate": 0.0005744457636265515,
+      "loss": 4.1673,
+      "step": 4050
+    },
+    {
+      "epoch": 0.4420485175202156,
+      "grad_norm": 0.6667271256446838,
+      "learning_rate": 0.0005741219643820831,
+      "loss": 4.1623,
+      "step": 4100
+    },
+    {
+      "epoch": 0.4474393530997305,
+      "grad_norm": 0.6467653512954712,
+      "learning_rate": 0.0005737981651376146,
+      "loss": 4.1408,
+      "step": 4150
+    },
+    {
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.6948724985122681,
+      "learning_rate": 0.0005734743658931462,
+      "loss": 4.1498,
+      "step": 4200
+    },
+    {
+      "epoch": 0.4582210242587601,
+      "grad_norm": 0.59138023853302,
+      "learning_rate": 0.0005731505666486778,
+      "loss": 4.1371,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4636118598382749,
+      "grad_norm": 0.8744353652000427,
+      "learning_rate": 0.0005728267674042093,
+      "loss": 4.1414,
+      "step": 4300
+    },
+    {
+      "epoch": 0.46900269541778977,
+      "grad_norm": 0.7583370804786682,
+      "learning_rate": 0.0005725029681597409,
+      "loss": 4.1251,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4743935309973046,
+      "grad_norm": 0.6753732562065125,
+      "learning_rate": 0.0005721791689152725,
+      "loss": 4.1139,
+      "step": 4400
+    },
+    {
+      "epoch": 0.4797843665768194,
+      "grad_norm": 0.7481412291526794,
+      "learning_rate": 0.0005718553696708041,
+      "loss": 4.1187,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48517520215633425,
+      "grad_norm": 0.6845588088035583,
+      "learning_rate": 0.0005715315704263356,
+      "loss": 4.1127,
+      "step": 4500
+    },
+    {
+      "epoch": 0.49056603773584906,
+      "grad_norm": 0.8330910205841064,
+      "learning_rate": 0.0005712077711818672,
+      "loss": 4.1234,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49595687331536387,
+      "grad_norm": 0.8156750202178955,
+      "learning_rate": 0.0005708839719373987,
+      "loss": 4.117,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5013477088948787,
+      "grad_norm": 0.6442606449127197,
+      "learning_rate": 0.0005705601726929304,
+      "loss": 4.0742,
+      "step": 4650
+    },
+    {
+      "epoch": 0.5067385444743935,
+      "grad_norm": 0.7438291907310486,
+      "learning_rate": 0.0005702363734484619,
+      "loss": 4.0767,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5121293800539084,
+      "grad_norm": 0.6073052883148193,
+      "learning_rate": 0.0005699125742039935,
+      "loss": 4.0814,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5175202156334232,
+      "grad_norm": 0.7255820631980896,
+      "learning_rate": 0.000569588774959525,
+      "loss": 4.0777,
+      "step": 4800
+    },
+    {
+      "epoch": 0.522911051212938,
+      "grad_norm": 0.6645416617393494,
+      "learning_rate": 0.0005692649757150567,
+      "loss": 4.0733,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5283018867924528,
+      "grad_norm": 0.7203142642974854,
+      "learning_rate": 0.0005689411764705881,
+      "loss": 4.0753,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5336927223719676,
+      "grad_norm": 0.6812610626220703,
+      "learning_rate": 0.0005686173772261197,
+      "loss": 4.0645,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "grad_norm": 0.7987631559371948,
+      "learning_rate": 0.0005682935779816514,
+      "loss": 4.0631,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "eval_accuracy": 0.32099018418946196,
+      "eval_loss": 3.9943337440490723,
+      "eval_runtime": 184.2999,
+      "eval_samples_per_second": 97.727,
+      "eval_steps_per_second": 6.11,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5444743935309974,
+      "grad_norm": 0.6587278246879578,
+      "learning_rate": 0.0005679697787371829,
+      "loss": 4.0704,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5498652291105122,
+      "grad_norm": 0.6426385641098022,
+      "learning_rate": 0.0005676459794927145,
+      "loss": 4.0577,
+      "step": 5100
+    },
+    {
+      "epoch": 0.555256064690027,
+      "grad_norm": 0.7008543014526367,
+      "learning_rate": 0.000567322180248246,
+      "loss": 4.0488,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5606469002695418,
+      "grad_norm": 0.6497129797935486,
+      "learning_rate": 0.0005669983810037777,
+      "loss": 4.0616,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.6053639650344849,
+      "learning_rate": 0.0005666745817593092,
+      "loss": 4.0349,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 0.5914347767829895,
+      "learning_rate": 0.0005663507825148408,
+      "loss": 4.035,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5768194070080862,
+      "grad_norm": 0.6652584671974182,
+      "learning_rate": 0.0005660269832703723,
+      "loss": 4.0262,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5822102425876011,
+      "grad_norm": 0.6383451223373413,
+      "learning_rate": 0.0005657031840259039,
+      "loss": 4.031,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5876010781671159,
+      "grad_norm": 0.6813966035842896,
+      "learning_rate": 0.0005653793847814355,
+      "loss": 4.033,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5929919137466307,
+      "grad_norm": 0.7794296145439148,
+      "learning_rate": 0.000565055585536967,
+      "loss": 4.0085,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5983827493261455,
+      "grad_norm": 0.6553782820701599,
+      "learning_rate": 0.0005647317862924986,
+      "loss": 4.0318,
+      "step": 5550
+    },
+    {
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.6803719401359558,
+      "learning_rate": 0.0005644079870480302,
+      "loss": 3.9794,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6091644204851752,
+      "grad_norm": 0.6251174211502075,
+      "learning_rate": 0.0005640841878035617,
+      "loss": 4.0158,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6145552560646901,
+      "grad_norm": 0.7014779448509216,
+      "learning_rate": 0.0005637603885590933,
+      "loss": 4.0177,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6199460916442049,
+      "grad_norm": 0.6140213012695312,
+      "learning_rate": 0.0005634365893146248,
+      "loss": 4.017,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6253369272237197,
+      "grad_norm": 0.6971213221549988,
+      "learning_rate": 0.0005631127900701565,
+      "loss": 4.0065,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6307277628032345,
+      "grad_norm": 0.6409980654716492,
+      "learning_rate": 0.000562788990825688,
+      "loss": 3.9797,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6361185983827493,
+      "grad_norm": 0.6414069533348083,
+      "learning_rate": 0.0005624651915812196,
+      "loss": 3.9951,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6415094339622641,
+      "grad_norm": 0.7267135381698608,
+      "learning_rate": 0.0005621413923367511,
+      "loss": 4.0005,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "grad_norm": 0.5980563759803772,
+      "learning_rate": 0.0005618175930922828,
+      "loss": 4.0063,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "eval_accuracy": 0.3277966356937123,
+      "eval_loss": 3.9200706481933594,
+      "eval_runtime": 184.1116,
+      "eval_samples_per_second": 97.827,
+      "eval_steps_per_second": 6.116,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6522911051212938,
+      "grad_norm": 0.7149510979652405,
+      "learning_rate": 0.0005614937938478143,
+      "loss": 3.9946,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6576819407008087,
+      "grad_norm": 0.691727340221405,
+      "learning_rate": 0.0005611699946033459,
+      "loss": 3.9905,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6630727762803235,
+      "grad_norm": 0.5647642612457275,
+      "learning_rate": 0.0005608461953588774,
+      "loss": 3.9812,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6684636118598383,
+      "grad_norm": 0.5823507308959961,
+      "learning_rate": 0.000560522396114409,
+      "loss": 3.9752,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6738544474393531,
+      "grad_norm": 0.6531707048416138,
+      "learning_rate": 0.0005601985968699405,
+      "loss": 3.9557,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6792452830188679,
+      "grad_norm": 0.6895961761474609,
+      "learning_rate": 0.0005598747976254721,
+      "loss": 3.9751,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6846361185983828,
+      "grad_norm": 0.5271945595741272,
+      "learning_rate": 0.0005595509983810038,
+      "loss": 3.9466,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6900269541778976,
+      "grad_norm": 0.5650665760040283,
+      "learning_rate": 0.0005592271991365353,
+      "loss": 3.9575,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6954177897574124,
+      "grad_norm": 0.629801332950592,
+      "learning_rate": 0.0005589033998920669,
+      "loss": 3.9759,
+      "step": 6450
+    },
+    {
+      "epoch": 0.7008086253369272,
+      "grad_norm": 0.630355954170227,
+      "learning_rate": 0.0005585796006475984,
+      "loss": 3.9447,
+      "step": 6500
+    },
+    {
+      "epoch": 0.706199460916442,
+      "grad_norm": 0.6709086894989014,
+      "learning_rate": 0.0005582558014031301,
+      "loss": 3.9707,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7115902964959568,
+      "grad_norm": 0.5377687811851501,
+      "learning_rate": 0.0005579320021586616,
+      "loss": 3.9539,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7169811320754716,
+      "grad_norm": 0.7604367733001709,
+      "learning_rate": 0.0005576082029141932,
+      "loss": 3.9489,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7223719676549866,
+      "grad_norm": 0.5272060632705688,
+      "learning_rate": 0.0005572844036697247,
+      "loss": 3.9361,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7277628032345014,
+      "grad_norm": 0.5893604159355164,
+      "learning_rate": 0.0005569606044252563,
+      "loss": 3.9425,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7331536388140162,
+      "grad_norm": 0.8032468557357788,
+      "learning_rate": 0.0005566368051807879,
+      "loss": 3.9585,
+      "step": 6800
+    },
+    {
+      "epoch": 0.738544474393531,
+      "grad_norm": 0.6482828855514526,
+      "learning_rate": 0.0005563130059363194,
+      "loss": 3.9433,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7439353099730458,
+      "grad_norm": 0.532512903213501,
+      "learning_rate": 0.000555989206691851,
+      "loss": 3.9431,
+      "step": 6900
+    },
+    {
+      "epoch": 0.7493261455525606,
+      "grad_norm": 0.5721356868743896,
+      "learning_rate": 0.0005556654074473826,
+      "loss": 3.9487,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.7197826504707336,
+      "learning_rate": 0.0005553416082029141,
+      "loss": 3.9296,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "eval_accuracy": 0.3329282029323024,
+      "eval_loss": 3.8656320571899414,
+      "eval_runtime": 184.4615,
+      "eval_samples_per_second": 97.641,
+      "eval_steps_per_second": 6.104,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7601078167115903,
+      "grad_norm": 0.6058018803596497,
+      "learning_rate": 0.0005550178089584457,
+      "loss": 3.9224,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7654986522911051,
+      "grad_norm": 0.5473665595054626,
+      "learning_rate": 0.0005546940097139772,
+      "loss": 3.9212,
+      "step": 7100
+    },
+    {
+      "epoch": 0.77088948787062,
+      "grad_norm": 0.6420671939849854,
+      "learning_rate": 0.0005543702104695089,
+      "loss": 3.9266,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7762803234501348,
+      "grad_norm": 0.5873612761497498,
+      "learning_rate": 0.0005540464112250404,
+      "loss": 3.9235,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7816711590296496,
+      "grad_norm": 0.6888593435287476,
+      "learning_rate": 0.000553722611980572,
+      "loss": 3.9344,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7870619946091644,
+      "grad_norm": 0.5773025155067444,
+      "learning_rate": 0.0005533988127361035,
+      "loss": 3.9178,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7924528301886793,
+      "grad_norm": 0.5834557414054871,
+      "learning_rate": 0.0005530750134916352,
+      "loss": 3.9313,
+      "step": 7350
+    },
+    {
+      "epoch": 0.7978436657681941,
+      "grad_norm": 0.7065815329551697,
+      "learning_rate": 0.0005527512142471668,
+      "loss": 3.9078,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8032345013477089,
+      "grad_norm": 0.5983999371528625,
+      "learning_rate": 0.0005524274150026982,
+      "loss": 3.9125,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8086253369272237,
+      "grad_norm": 0.6277061104774475,
+      "learning_rate": 0.0005521036157582299,
+      "loss": 3.9004,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8140161725067385,
+      "grad_norm": 0.7020032405853271,
+      "learning_rate": 0.0005517798165137614,
+      "loss": 3.9029,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8194070080862533,
+      "grad_norm": 0.6681187152862549,
+      "learning_rate": 0.000551456017269293,
+      "loss": 3.8801,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8247978436657682,
+      "grad_norm": 0.6278699636459351,
+      "learning_rate": 0.0005511322180248245,
+      "loss": 3.8967,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8301886792452831,
+      "grad_norm": 0.6532415747642517,
+      "learning_rate": 0.0005508084187803562,
+      "loss": 3.9083,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8355795148247979,
+      "grad_norm": 0.5810382962226868,
+      "learning_rate": 0.0005504846195358877,
+      "loss": 3.903,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8409703504043127,
+      "grad_norm": 0.6141226291656494,
+      "learning_rate": 0.0005501608202914193,
+      "loss": 3.8898,
+      "step": 7800
+    },
+    {
+      "epoch": 0.8463611859838275,
+      "grad_norm": 0.6052567362785339,
+      "learning_rate": 0.0005498370210469508,
+      "loss": 3.8946,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8517520215633423,
+      "grad_norm": 0.619564950466156,
+      "learning_rate": 0.0005495132218024824,
+      "loss": 3.8788,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.620516836643219,
+      "learning_rate": 0.000549189422558014,
+      "loss": 3.8888,
+      "step": 7950
+    },
+    {
+      "epoch": 0.862533692722372,
+      "grad_norm": 0.5579094290733337,
+      "learning_rate": 0.0005488656233135456,
+      "loss": 3.8882,
+      "step": 8000
+    },
+    {
+      "epoch": 0.862533692722372,
+      "eval_accuracy": 0.3372453081238352,
+      "eval_loss": 3.8190271854400635,
+      "eval_runtime": 183.5032,
+      "eval_samples_per_second": 98.151,
+      "eval_steps_per_second": 6.136,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8679245283018868,
+      "grad_norm": 0.5879581570625305,
+      "learning_rate": 0.0005485418240690771,
+      "loss": 3.8678,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8733153638814016,
+      "grad_norm": 0.6095612049102783,
+      "learning_rate": 0.0005482180248246087,
+      "loss": 3.8844,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8787061994609164,
+      "grad_norm": 0.5836382508277893,
+      "learning_rate": 0.0005478942255801403,
+      "loss": 3.8812,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8840970350404312,
+      "grad_norm": 0.5505040884017944,
+      "learning_rate": 0.0005475704263356718,
+      "loss": 3.8725,
+      "step": 8200
+    },
+    {
+      "epoch": 0.889487870619946,
+      "grad_norm": 0.5962595343589783,
+      "learning_rate": 0.0005472466270912034,
+      "loss": 3.8804,
+      "step": 8250
+    },
+    {
+      "epoch": 0.894878706199461,
+      "grad_norm": 0.6875723600387573,
+      "learning_rate": 0.000546922827846735,
+      "loss": 3.8692,
+      "step": 8300
+    },
+    {
+      "epoch": 0.9002695417789758,
+      "grad_norm": 0.5387910008430481,
+      "learning_rate": 0.0005465990286022665,
+      "loss": 3.8689,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.6388599276542664,
+      "learning_rate": 0.0005462752293577981,
+      "loss": 3.8823,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9110512129380054,
+      "grad_norm": 0.559060275554657,
+      "learning_rate": 0.0005459514301133296,
+      "loss": 3.8783,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9164420485175202,
+      "grad_norm": 0.5374552011489868,
+      "learning_rate": 0.0005456276308688613,
+      "loss": 3.8729,
+      "step": 8500
+    },
+    {
+      "epoch": 0.921832884097035,
+      "grad_norm": 0.5611122250556946,
+      "learning_rate": 0.0005453038316243929,
+      "loss": 3.8668,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9272237196765498,
+      "grad_norm": 0.5464676022529602,
+      "learning_rate": 0.0005449800323799244,
+      "loss": 3.8786,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9326145552560647,
+      "grad_norm": 0.6095228791236877,
+      "learning_rate": 0.000544656233135456,
+      "loss": 3.8565,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9380053908355795,
+      "grad_norm": 0.5888999104499817,
+      "learning_rate": 0.0005443324338909875,
+      "loss": 3.8615,
+      "step": 8700
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.5549721121788025,
+      "learning_rate": 0.0005440086346465192,
+      "loss": 3.8709,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9487870619946092,
+      "grad_norm": 0.650879979133606,
+      "learning_rate": 0.0005436848354020506,
+      "loss": 3.8519,
+      "step": 8800
+    },
+    {
+      "epoch": 0.954177897574124,
+      "grad_norm": 0.5871933102607727,
+      "learning_rate": 0.0005433610361575823,
+      "loss": 3.8592,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9595687331536388,
+      "grad_norm": 0.5539306402206421,
+      "learning_rate": 0.0005430372369131138,
+      "loss": 3.8421,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9649595687331537,
+      "grad_norm": 0.5857457518577576,
+      "learning_rate": 0.0005427134376686454,
+      "loss": 3.8551,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "grad_norm": 0.5824806690216064,
+      "learning_rate": 0.0005423896384241769,
+      "loss": 3.8432,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "eval_accuracy": 0.3406982968985794,
+      "eval_loss": 3.780831813812256,
+      "eval_runtime": 183.3642,
+      "eval_samples_per_second": 98.225,
+      "eval_steps_per_second": 6.141,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9757412398921833,
+      "grad_norm": 0.8111798167228699,
+      "learning_rate": 0.0005420658391797086,
+      "loss": 3.8374,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9811320754716981,
+      "grad_norm": 0.5787947177886963,
+      "learning_rate": 0.0005417420399352401,
+      "loss": 3.8494,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9865229110512129,
+      "grad_norm": 0.6467134356498718,
+      "learning_rate": 0.0005414182406907717,
+      "loss": 3.8425,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9919137466307277,
+      "grad_norm": 0.6325974464416504,
+      "learning_rate": 0.0005410944414463032,
+      "loss": 3.8555,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9973045822102425,
+      "grad_norm": 0.5801452994346619,
+      "learning_rate": 0.0005407706422018348,
+      "loss": 3.854,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0026954177897573,
+      "grad_norm": 0.621691107749939,
+      "learning_rate": 0.0005404468429573664,
+      "loss": 3.8021,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0080862533692723,
+      "grad_norm": 0.6182531118392944,
+      "learning_rate": 0.000540123043712898,
+      "loss": 3.771,
+      "step": 9350
+    },
+    {
+      "epoch": 1.013477088948787,
+      "grad_norm": 0.5976972579956055,
+      "learning_rate": 0.0005397992444684295,
+      "loss": 3.7536,
+      "step": 9400
+    },
+    {
+      "epoch": 1.0188679245283019,
+      "grad_norm": 0.6204501390457153,
+      "learning_rate": 0.0005394754452239611,
+      "loss": 3.7671,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0242587601078168,
+      "grad_norm": 0.532163679599762,
+      "learning_rate": 0.0005391516459794927,
+      "loss": 3.7787,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0296495956873315,
+      "grad_norm": 0.5977817177772522,
+      "learning_rate": 0.0005388278467350242,
+      "loss": 3.7871,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0350404312668464,
+      "grad_norm": 0.5936277508735657,
+      "learning_rate": 0.0005385040474905557,
+      "loss": 3.7799,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0404312668463611,
+      "grad_norm": 0.5855807662010193,
+      "learning_rate": 0.0005381802482460874,
+      "loss": 3.7879,
+      "step": 9650
+    },
+    {
+      "epoch": 1.045822102425876,
+      "grad_norm": 0.531672477722168,
+      "learning_rate": 0.000537856449001619,
+      "loss": 3.7905,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0512129380053907,
+      "grad_norm": 0.6275017261505127,
+      "learning_rate": 0.0005375326497571505,
+      "loss": 3.761,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0566037735849056,
+      "grad_norm": 0.6375147104263306,
+      "learning_rate": 0.000537208850512682,
+      "loss": 3.7682,
+      "step": 9800
+    },
+    {
+      "epoch": 1.0619946091644206,
+      "grad_norm": 0.5759090781211853,
+      "learning_rate": 0.0005368850512682137,
+      "loss": 3.7676,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0673854447439353,
+      "grad_norm": 0.5603178143501282,
+      "learning_rate": 0.0005365612520237453,
+      "loss": 3.779,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0727762803234502,
+      "grad_norm": 0.5935394763946533,
+      "learning_rate": 0.0005362374527792768,
+      "loss": 3.7782,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "grad_norm": 0.5596011281013489,
+      "learning_rate": 0.0005359136535348084,
+      "loss": 3.7782,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "eval_accuracy": 0.3441405290374299,
+      "eval_loss": 3.7568960189819336,
+      "eval_runtime": 183.0624,
+      "eval_samples_per_second": 98.387,
+      "eval_steps_per_second": 6.151,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0835579514824798,
+      "grad_norm": 0.5355740785598755,
+      "learning_rate": 0.0005355898542903399,
+      "loss": 3.7564,
+      "step": 10050
+    },
+    {
+      "epoch": 1.0889487870619945,
+      "grad_norm": 0.5919591188430786,
+      "learning_rate": 0.0005352660550458716,
+      "loss": 3.76,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0943396226415094,
+      "grad_norm": 0.6239626407623291,
+      "learning_rate": 0.000534942255801403,
+      "loss": 3.7471,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0997304582210243,
+      "grad_norm": 0.6291561722755432,
+      "learning_rate": 0.0005346184565569347,
+      "loss": 3.7655,
+      "step": 10200
+    },
+    {
+      "epoch": 1.105121293800539,
+      "grad_norm": 0.5858433842658997,
+      "learning_rate": 0.0005342946573124662,
+      "loss": 3.7832,
+      "step": 10250
+    },
+    {
+      "epoch": 1.110512129380054,
+      "grad_norm": 0.636974573135376,
+      "learning_rate": 0.0005339708580679978,
+      "loss": 3.7845,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1159029649595686,
+      "grad_norm": 0.5367472171783447,
+      "learning_rate": 0.0005336470588235293,
+      "loss": 3.76,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1212938005390836,
+      "grad_norm": 0.5671682357788086,
+      "learning_rate": 0.000533323259579061,
+      "loss": 3.7657,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1266846361185983,
+      "grad_norm": 0.5502925515174866,
+      "learning_rate": 0.0005329994603345925,
+      "loss": 3.7594,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1320754716981132,
+      "grad_norm": 0.5960683822631836,
+      "learning_rate": 0.0005326756610901241,
+      "loss": 3.7594,
+      "step": 10500
+    },
+    {
+      "epoch": 1.137466307277628,
+      "grad_norm": 0.5544126033782959,
+      "learning_rate": 0.0005323518618456556,
+      "loss": 3.7582,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.6969329714775085,
+      "learning_rate": 0.0005320280626011872,
+      "loss": 3.7705,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1482479784366577,
+      "grad_norm": 0.6323273181915283,
+      "learning_rate": 0.0005317042633567188,
+      "loss": 3.7581,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1536388140161726,
+      "grad_norm": 0.6453703045845032,
+      "learning_rate": 0.0005313804641122504,
+      "loss": 3.763,
+      "step": 10700
+    },
+    {
+      "epoch": 1.1590296495956873,
+      "grad_norm": 0.5417402386665344,
+      "learning_rate": 0.0005310566648677819,
+      "loss": 3.7579,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1644204851752022,
+      "grad_norm": 0.5967243313789368,
+      "learning_rate": 0.0005307328656233135,
+      "loss": 3.7471,
+      "step": 10800
+    },
+    {
+      "epoch": 1.169811320754717,
+      "grad_norm": 0.5780621767044067,
+      "learning_rate": 0.000530409066378845,
+      "loss": 3.7432,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1752021563342319,
+      "grad_norm": 1.0681740045547485,
+      "learning_rate": 0.0005300852671343766,
+      "loss": 3.759,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1805929919137466,
+      "grad_norm": 0.6521495580673218,
+      "learning_rate": 0.0005297614678899081,
+      "loss": 3.7513,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "grad_norm": 0.5670123100280762,
+      "learning_rate": 0.0005294376686454398,
+      "loss": 3.7405,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "eval_accuracy": 0.347109251891185,
+      "eval_loss": 3.7269060611724854,
+      "eval_runtime": 183.025,
+      "eval_samples_per_second": 98.407,
+      "eval_steps_per_second": 6.152,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1913746630727764,
+      "grad_norm": 0.5789473056793213,
+      "learning_rate": 0.0005291138694009714,
+      "loss": 3.7509,
+      "step": 11050
+    },
+    {
+      "epoch": 1.196765498652291,
+      "grad_norm": 0.6155396699905396,
+      "learning_rate": 0.0005287900701565029,
+      "loss": 3.7419,
+      "step": 11100
+    },
+    {
+      "epoch": 1.202156334231806,
+      "grad_norm": 0.6248864531517029,
+      "learning_rate": 0.0005284662709120345,
+      "loss": 3.7432,
+      "step": 11150
+    },
+    {
+      "epoch": 1.2075471698113207,
+      "grad_norm": 0.6895485520362854,
+      "learning_rate": 0.0005281424716675661,
+      "loss": 3.7428,
+      "step": 11200
+    },
+    {
+      "epoch": 1.2129380053908356,
+      "grad_norm": 0.6388883590698242,
+      "learning_rate": 0.0005278186724230977,
+      "loss": 3.7363,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2183288409703503,
+      "grad_norm": 0.5322048664093018,
+      "learning_rate": 0.0005274948731786292,
+      "loss": 3.7326,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2237196765498652,
+      "grad_norm": 0.6221327781677246,
+      "learning_rate": 0.0005271710739341608,
+      "loss": 3.7505,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2291105121293802,
+      "grad_norm": 0.6232865452766418,
+      "learning_rate": 0.0005268472746896923,
+      "loss": 3.7446,
+      "step": 11400
+    },
+    {
+      "epoch": 1.2345013477088949,
+      "grad_norm": 0.6101617813110352,
+      "learning_rate": 0.000526523475445224,
+      "loss": 3.7366,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2398921832884098,
+      "grad_norm": 0.5458541512489319,
+      "learning_rate": 0.0005261996762007554,
+      "loss": 3.7232,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2452830188679245,
+      "grad_norm": 0.5346110463142395,
+      "learning_rate": 0.0005258758769562871,
+      "loss": 3.7439,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2506738544474394,
+      "grad_norm": 0.5948981046676636,
+      "learning_rate": 0.0005255520777118186,
+      "loss": 3.7435,
+      "step": 11600
+    },
+    {
+      "epoch": 1.256064690026954,
+      "grad_norm": 0.6083604693412781,
+      "learning_rate": 0.0005252282784673502,
+      "loss": 3.7356,
+      "step": 11650
+    },
+    {
+      "epoch": 1.261455525606469,
+      "grad_norm": 0.6157535314559937,
+      "learning_rate": 0.0005249044792228817,
+      "loss": 3.7373,
+      "step": 11700
+    },
+    {
+      "epoch": 1.266846361185984,
+      "grad_norm": 0.5235822200775146,
+      "learning_rate": 0.0005245806799784133,
+      "loss": 3.7324,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2722371967654986,
+      "grad_norm": 0.6096104383468628,
+      "learning_rate": 0.0005242568807339449,
+      "loss": 3.7291,
+      "step": 11800
+    },
+    {
+      "epoch": 1.2776280323450135,
+      "grad_norm": 0.6221447587013245,
+      "learning_rate": 0.0005239330814894765,
+      "loss": 3.7236,
+      "step": 11850
+    },
+    {
+      "epoch": 1.2830188679245282,
+      "grad_norm": 0.5434000492095947,
+      "learning_rate": 0.000523609282245008,
+      "loss": 3.7454,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2884097035040432,
+      "grad_norm": 0.6391717195510864,
+      "learning_rate": 0.0005232854830005396,
+      "loss": 3.741,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "grad_norm": 0.5883181691169739,
+      "learning_rate": 0.0005229616837560712,
+      "loss": 3.7342,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "eval_accuracy": 0.3493015412086178,
+      "eval_loss": 3.702024221420288,
+      "eval_runtime": 183.3785,
+      "eval_samples_per_second": 98.218,
+      "eval_steps_per_second": 6.14,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2991913746630728,
+      "grad_norm": 0.5594079494476318,
+      "learning_rate": 0.0005226378845116028,
+      "loss": 3.7174,
+      "step": 12050
+    },
+    {
+      "epoch": 1.3045822102425877,
+      "grad_norm": 0.6001787781715393,
+      "learning_rate": 0.0005223140852671344,
+      "loss": 3.733,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3099730458221024,
+      "grad_norm": 0.623263955116272,
+      "learning_rate": 0.0005219902860226659,
+      "loss": 3.7411,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3153638814016173,
+      "grad_norm": 0.6152437329292297,
+      "learning_rate": 0.0005216664867781975,
+      "loss": 3.7405,
+      "step": 12200
+    },
+    {
+      "epoch": 1.320754716981132,
+      "grad_norm": 0.6849799752235413,
+      "learning_rate": 0.000521342687533729,
+      "loss": 3.7211,
+      "step": 12250
+    },
+    {
+      "epoch": 1.326145552560647,
+      "grad_norm": 0.5973236560821533,
+      "learning_rate": 0.0005210188882892606,
+      "loss": 3.7314,
+      "step": 12300
+    },
+    {
+      "epoch": 1.3315363881401616,
+      "grad_norm": 0.5735817551612854,
+      "learning_rate": 0.0005206950890447922,
+      "loss": 3.7194,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3369272237196765,
+      "grad_norm": 0.5044131278991699,
+      "learning_rate": 0.0005203712898003238,
+      "loss": 3.7348,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3423180592991915,
+      "grad_norm": 0.6093466281890869,
+      "learning_rate": 0.0005200474905558553,
+      "loss": 3.7264,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3477088948787062,
+      "grad_norm": 0.7663679122924805,
+      "learning_rate": 0.0005197236913113869,
+      "loss": 3.7262,
+      "step": 12500
+    },
+    {
+      "epoch": 1.353099730458221,
+      "grad_norm": 0.5749175548553467,
+      "learning_rate": 0.0005193998920669184,
+      "loss": 3.7343,
+      "step": 12550
+    },
+    {
+      "epoch": 1.3584905660377358,
+      "grad_norm": 0.6117640733718872,
+      "learning_rate": 0.0005190760928224501,
+      "loss": 3.7167,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3638814016172507,
+      "grad_norm": 0.5556171536445618,
+      "learning_rate": 0.0005187522935779816,
+      "loss": 3.7244,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3692722371967654,
+      "grad_norm": 0.59083092212677,
+      "learning_rate": 0.0005184284943335132,
+      "loss": 3.7366,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3746630727762803,
+      "grad_norm": 0.5685399174690247,
+      "learning_rate": 0.0005181046950890447,
+      "loss": 3.7232,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3800539083557952,
+      "grad_norm": 0.5734025239944458,
+      "learning_rate": 0.0005177808958445764,
+      "loss": 3.7146,
+      "step": 12800
+    },
+    {
+      "epoch": 1.38544474393531,
+      "grad_norm": 0.613262414932251,
+      "learning_rate": 0.0005174570966001078,
+      "loss": 3.7124,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3908355795148248,
+      "grad_norm": 0.6126392483711243,
+      "learning_rate": 0.0005171332973556395,
+      "loss": 3.7171,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3962264150943398,
+      "grad_norm": 0.5627440810203552,
+      "learning_rate": 0.000516809498111171,
+      "loss": 3.7154,
+      "step": 12950
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "grad_norm": 0.6159223318099976,
+      "learning_rate": 0.0005164856988667026,
+      "loss": 3.7098,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "eval_accuracy": 0.3513598615153753,
+      "eval_loss": 3.6815733909606934,
+      "eval_runtime": 182.9927,
+      "eval_samples_per_second": 98.425,
+      "eval_steps_per_second": 6.153,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4070080862533692,
+      "grad_norm": 0.5440373420715332,
+      "learning_rate": 0.0005161618996222341,
+      "loss": 3.7223,
+      "step": 13050
+    },
+    {
+      "epoch": 1.412398921832884,
+      "grad_norm": 0.6110854148864746,
+      "learning_rate": 0.0005158381003777657,
+      "loss": 3.7332,
+      "step": 13100
+    },
+    {
+      "epoch": 1.417789757412399,
+      "grad_norm": 0.634506106376648,
+      "learning_rate": 0.0005155143011332973,
+      "loss": 3.714,
+      "step": 13150
+    },
+    {
+      "epoch": 1.4231805929919137,
+      "grad_norm": 0.5525026321411133,
+      "learning_rate": 0.0005151905018888289,
+      "loss": 3.7169,
+      "step": 13200
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.5637493133544922,
+      "learning_rate": 0.0005148731786292498,
+      "loss": 3.712,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4339622641509435,
+      "grad_norm": 0.5339338779449463,
+      "learning_rate": 0.0005145493793847814,
+      "loss": 3.7101,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4393530997304582,
+      "grad_norm": 0.613170862197876,
+      "learning_rate": 0.000514225580140313,
+      "loss": 3.6965,
+      "step": 13350
+    },
+    {
+      "epoch": 1.444743935309973,
+      "grad_norm": 0.589502215385437,
+      "learning_rate": 0.0005139017808958445,
+      "loss": 3.7058,
+      "step": 13400
+    },
+    {
+      "epoch": 1.4501347708894878,
+      "grad_norm": 0.5917065143585205,
+      "learning_rate": 0.0005135779816513762,
+      "loss": 3.7144,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4555256064690028,
+      "grad_norm": 0.5540453195571899,
+      "learning_rate": 0.0005132541824069076,
+      "loss": 3.7001,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4609164420485174,
+      "grad_norm": 0.5671173930168152,
+      "learning_rate": 0.0005129303831624393,
+      "loss": 3.7114,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4663072776280324,
+      "grad_norm": 0.5867645740509033,
+      "learning_rate": 0.0005126065839179708,
+      "loss": 3.715,
+      "step": 13600
+    },
+    {
+      "epoch": 1.4716981132075473,
+      "grad_norm": 0.5680153369903564,
+      "learning_rate": 0.0005122827846735024,
+      "loss": 3.6852,
+      "step": 13650
+    },
+    {
+      "epoch": 1.477088948787062,
+      "grad_norm": 0.5795393586158752,
+      "learning_rate": 0.0005119589854290339,
+      "loss": 3.6843,
+      "step": 13700
+    },
+    {
+      "epoch": 1.482479784366577,
+      "grad_norm": 0.5783411860466003,
+      "learning_rate": 0.0005116351861845655,
+      "loss": 3.7033,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4878706199460916,
+      "grad_norm": 0.6578272581100464,
+      "learning_rate": 0.0005113113869400971,
+      "loss": 3.7184,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4932614555256065,
+      "grad_norm": 0.5720994472503662,
+      "learning_rate": 0.0005109875876956287,
+      "loss": 3.711,
+      "step": 13850
+    },
+    {
+      "epoch": 1.4986522911051212,
+      "grad_norm": 0.5809938311576843,
+      "learning_rate": 0.0005106637884511602,
+      "loss": 3.7267,
+      "step": 13900
+    },
+    {
+      "epoch": 1.5040431266846361,
+      "grad_norm": 0.5434926748275757,
+      "learning_rate": 0.0005103399892066918,
+      "loss": 3.7006,
+      "step": 13950
+    },
+    {
+      "epoch": 1.509433962264151,
+      "grad_norm": 0.6142950057983398,
+      "learning_rate": 0.0005100161899622234,
+      "loss": 3.7016,
+      "step": 14000
+    },
+    {
+      "epoch": 1.509433962264151,
+      "eval_accuracy": 0.35360506478917375,
+      "eval_loss": 3.658677816390991,
+      "eval_runtime": 183.8031,
+      "eval_samples_per_second": 97.991,
+      "eval_steps_per_second": 6.126,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5148247978436657,
+      "grad_norm": 0.5697357654571533,
+      "learning_rate": 0.000509692390717755,
+      "loss": 3.6912,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5202156334231804,
+      "grad_norm": 0.7325167059898376,
+      "learning_rate": 0.0005093685914732865,
+      "loss": 3.6894,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5256064690026954,
+      "grad_norm": 0.5748705863952637,
+      "learning_rate": 0.0005090447922288181,
+      "loss": 3.6853,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5309973045822103,
+      "grad_norm": 0.626307487487793,
+      "learning_rate": 0.0005087209929843496,
+      "loss": 3.7191,
+      "step": 14200
+    },
+    {
+      "epoch": 1.536388140161725,
+      "grad_norm": 0.6443688273429871,
+      "learning_rate": 0.0005083971937398812,
+      "loss": 3.6925,
+      "step": 14250
+    },
+    {
+      "epoch": 1.54177897574124,
+      "grad_norm": 0.6409396529197693,
+      "learning_rate": 0.0005080733944954127,
+      "loss": 3.6974,
+      "step": 14300
+    },
+    {
+      "epoch": 1.5471698113207548,
+      "grad_norm": 0.5955445766448975,
+      "learning_rate": 0.0005077495952509444,
+      "loss": 3.6789,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5525606469002695,
+      "grad_norm": 0.5558004379272461,
+      "learning_rate": 0.0005074257960064759,
+      "loss": 3.6912,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5579514824797842,
+      "grad_norm": 0.589004635810852,
+      "learning_rate": 0.0005071019967620075,
+      "loss": 3.6969,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5633423180592994,
+      "grad_norm": 0.6211824417114258,
+      "learning_rate": 0.000506778197517539,
+      "loss": 3.6912,
+      "step": 14500
+    },
+    {
+      "epoch": 1.568733153638814,
+      "grad_norm": 0.5962303280830383,
+      "learning_rate": 0.0005064543982730707,
+      "loss": 3.6995,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5741239892183287,
+      "grad_norm": 0.6277245283126831,
+      "learning_rate": 0.0005061305990286023,
+      "loss": 3.6931,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5795148247978437,
+      "grad_norm": 0.6255839467048645,
+      "learning_rate": 0.0005058067997841338,
+      "loss": 3.7005,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5849056603773586,
+      "grad_norm": 0.5719764828681946,
+      "learning_rate": 0.0005054830005396654,
+      "loss": 3.6871,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5902964959568733,
+      "grad_norm": 0.6281840801239014,
+      "learning_rate": 0.0005051592012951969,
+      "loss": 3.6767,
+      "step": 14750
+    },
+    {
+      "epoch": 1.595687331536388,
+      "grad_norm": 0.5663508772850037,
+      "learning_rate": 0.0005048354020507286,
+      "loss": 3.6987,
+      "step": 14800
+    },
+    {
+      "epoch": 1.6010781671159031,
+      "grad_norm": 0.5636190176010132,
+      "learning_rate": 0.00050451160280626,
+      "loss": 3.7,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6064690026954178,
+      "grad_norm": 0.5662776827812195,
+      "learning_rate": 0.0005041878035617917,
+      "loss": 3.6897,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6118598382749325,
+      "grad_norm": 0.5471853613853455,
+      "learning_rate": 0.0005038640043173232,
+      "loss": 3.6842,
+      "step": 14950
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "grad_norm": 0.704903781414032,
+      "learning_rate": 0.0005035402050728548,
+      "loss": 3.6781,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "eval_accuracy": 0.3554610734188207,
+      "eval_loss": 3.6408092975616455,
+      "eval_runtime": 183.2026,
+      "eval_samples_per_second": 98.312,
+      "eval_steps_per_second": 6.146,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6226415094339623,
+      "grad_norm": 0.5391875505447388,
+      "learning_rate": 0.0005032164058283863,
+      "loss": 3.6821,
+      "step": 15050
+    },
+    {
+      "epoch": 1.628032345013477,
+      "grad_norm": 0.5684933662414551,
+      "learning_rate": 0.0005028926065839179,
+      "loss": 3.6782,
+      "step": 15100
+    },
+    {
+      "epoch": 1.633423180592992,
+      "grad_norm": 0.5583056807518005,
+      "learning_rate": 0.0005025688073394495,
+      "loss": 3.6809,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6388140161725069,
+      "grad_norm": 0.5500176548957825,
+      "learning_rate": 0.0005022450080949811,
+      "loss": 3.6846,
+      "step": 15200
+    },
+    {
+      "epoch": 1.6442048517520216,
+      "grad_norm": 0.6347390413284302,
+      "learning_rate": 0.0005019212088505126,
+      "loss": 3.6879,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6495956873315363,
+      "grad_norm": 0.5185267925262451,
+      "learning_rate": 0.0005015974096060442,
+      "loss": 3.6893,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6549865229110512,
+      "grad_norm": 0.6127241253852844,
+      "learning_rate": 0.0005012736103615758,
+      "loss": 3.6717,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6603773584905661,
+      "grad_norm": 0.6135085821151733,
+      "learning_rate": 0.0005009498111171074,
+      "loss": 3.6772,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6657681940700808,
+      "grad_norm": 0.5636995434761047,
+      "learning_rate": 0.0005006260118726389,
+      "loss": 3.6764,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6711590296495957,
+      "grad_norm": 0.5827999114990234,
+      "learning_rate": 0.0005003022126281705,
+      "loss": 3.6785,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6765498652291106,
+      "grad_norm": 0.7018887400627136,
+      "learning_rate": 0.000499978413383702,
+      "loss": 3.6721,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6819407008086253,
+      "grad_norm": 0.564325749874115,
+      "learning_rate": 0.0004996546141392336,
+      "loss": 3.6671,
+      "step": 15600
+    },
+    {
+      "epoch": 1.68733153638814,
+      "grad_norm": 0.5548321008682251,
+      "learning_rate": 0.0004993308148947651,
+      "loss": 3.6766,
+      "step": 15650
+    },
+    {
+      "epoch": 1.692722371967655,
+      "grad_norm": 0.5533665418624878,
+      "learning_rate": 0.0004990070156502968,
+      "loss": 3.6657,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6981132075471699,
+      "grad_norm": 0.564487636089325,
+      "learning_rate": 0.0004986832164058284,
+      "loss": 3.6623,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7035040431266846,
+      "grad_norm": 0.6166607141494751,
+      "learning_rate": 0.0004983594171613599,
+      "loss": 3.676,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7088948787061995,
+      "grad_norm": 0.6325422525405884,
+      "learning_rate": 0.0004980356179168915,
+      "loss": 3.6815,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.6531565189361572,
+      "learning_rate": 0.000497711818672423,
+      "loss": 3.704,
+      "step": 15900
+    },
+    {
+      "epoch": 1.719676549865229,
+      "grad_norm": 0.5142133831977844,
+      "learning_rate": 0.0004973880194279547,
+      "loss": 3.6707,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "grad_norm": 0.5847429633140564,
+      "learning_rate": 0.0004970642201834862,
+      "loss": 3.6798,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "eval_accuracy": 0.35691093755381714,
+      "eval_loss": 3.620922088623047,
+      "eval_runtime": 183.1348,
+      "eval_samples_per_second": 98.348,
+      "eval_steps_per_second": 6.148,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7304582210242587,
+      "grad_norm": 0.5682497620582581,
+      "learning_rate": 0.0004967404209390178,
+      "loss": 3.6472,
+      "step": 16050
+    },
+    {
+      "epoch": 1.7358490566037736,
+      "grad_norm": 0.5600143671035767,
+      "learning_rate": 0.0004964166216945493,
+      "loss": 3.6561,
+      "step": 16100
+    },
+    {
+      "epoch": 1.7412398921832883,
+      "grad_norm": 0.5824739933013916,
+      "learning_rate": 0.000496092822450081,
+      "loss": 3.6703,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7466307277628033,
+      "grad_norm": 0.5835148096084595,
+      "learning_rate": 0.0004957690232056125,
+      "loss": 3.6583,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7520215633423182,
+      "grad_norm": 0.5758132338523865,
+      "learning_rate": 0.0004954452239611441,
+      "loss": 3.6676,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7574123989218329,
+      "grad_norm": 0.542614758014679,
+      "learning_rate": 0.0004951214247166756,
+      "loss": 3.6587,
+      "step": 16300
+    },
+    {
+      "epoch": 1.7628032345013476,
+      "grad_norm": 0.5760119557380676,
+      "learning_rate": 0.0004947976254722072,
+      "loss": 3.6519,
+      "step": 16350
+    },
+    {
+      "epoch": 1.7681940700808625,
+      "grad_norm": 0.5551959872245789,
+      "learning_rate": 0.0004944738262277387,
+      "loss": 3.6661,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7735849056603774,
+      "grad_norm": 0.5755851864814758,
+      "learning_rate": 0.0004941500269832703,
+      "loss": 3.6653,
+      "step": 16450
+    },
+    {
+      "epoch": 1.778975741239892,
+      "grad_norm": 0.5640000104904175,
+      "learning_rate": 0.0004938262277388019,
+      "loss": 3.6512,
+      "step": 16500
+    },
+    {
+      "epoch": 1.784366576819407,
+      "grad_norm": 0.5701029896736145,
+      "learning_rate": 0.0004935024284943335,
+      "loss": 3.647,
+      "step": 16550
+    },
+    {
+      "epoch": 1.789757412398922,
+      "grad_norm": 0.5458316802978516,
+      "learning_rate": 0.000493178629249865,
+      "loss": 3.6534,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7951482479784366,
+      "grad_norm": 0.5290755033493042,
+      "learning_rate": 0.0004928548300053966,
+      "loss": 3.6619,
+      "step": 16650
+    },
+    {
+      "epoch": 1.8005390835579513,
+      "grad_norm": 0.56656813621521,
+      "learning_rate": 0.0004925310307609282,
+      "loss": 3.6537,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8059299191374663,
+      "grad_norm": 0.544216513633728,
+      "learning_rate": 0.0004922072315164598,
+      "loss": 3.6468,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8113207547169812,
+      "grad_norm": 0.5542647242546082,
+      "learning_rate": 0.0004918834322719913,
+      "loss": 3.6546,
+      "step": 16800
+    },
+    {
+      "epoch": 1.8167115902964959,
+      "grad_norm": 0.5754142999649048,
+      "learning_rate": 0.0004915596330275229,
+      "loss": 3.6427,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8221024258760108,
+      "grad_norm": 0.5976019501686096,
+      "learning_rate": 0.0004912358337830544,
+      "loss": 3.6485,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8274932614555257,
+      "grad_norm": 0.5452532768249512,
+      "learning_rate": 0.000490912034538586,
+      "loss": 3.655,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "grad_norm": 0.5066123008728027,
+      "learning_rate": 0.0004905882352941175,
+      "loss": 3.6639,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "eval_accuracy": 0.3585854958608139,
+      "eval_loss": 3.6066739559173584,
+      "eval_runtime": 183.4798,
+      "eval_samples_per_second": 98.163,
+      "eval_steps_per_second": 6.137,
+      "step": 17000
+    },
+    {
+      "epoch": 1.838274932614555,
+      "grad_norm": 0.580244779586792,
+      "learning_rate": 0.0004902644360496492,
+      "loss": 3.6551,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8436657681940702,
+      "grad_norm": 0.5668180584907532,
+      "learning_rate": 0.0004899406368051808,
+      "loss": 3.6445,
+      "step": 17100
+    },
+    {
+      "epoch": 1.849056603773585,
+      "grad_norm": 0.5920048356056213,
+      "learning_rate": 0.0004896168375607123,
+      "loss": 3.6445,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8544474393530996,
+      "grad_norm": 0.5886459946632385,
+      "learning_rate": 0.0004892930383162439,
+      "loss": 3.6436,
+      "step": 17200
+    },
+    {
+      "epoch": 1.8598382749326146,
+      "grad_norm": 0.544743001461029,
+      "learning_rate": 0.0004889692390717754,
+      "loss": 3.6386,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8652291105121295,
+      "grad_norm": 0.5967076420783997,
+      "learning_rate": 0.0004886519158121964,
+      "loss": 3.6638,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8706199460916442,
+      "grad_norm": 0.575445294380188,
+      "learning_rate": 0.000488328116567728,
+      "loss": 3.6653,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8760107816711589,
+      "grad_norm": 0.5783690810203552,
+      "learning_rate": 0.0004880043173232595,
+      "loss": 3.6453,
+      "step": 17400
+    },
+    {
+      "epoch": 1.881401617250674,
+      "grad_norm": 0.5519454479217529,
+      "learning_rate": 0.0004876805180787911,
+      "loss": 3.6561,
+      "step": 17450
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.5622634887695312,
+      "learning_rate": 0.0004873567188343227,
+      "loss": 3.6404,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8921832884097034,
+      "grad_norm": 0.5808458924293518,
+      "learning_rate": 0.0004870329195898542,
+      "loss": 3.655,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8975741239892183,
+      "grad_norm": 0.5541762113571167,
+      "learning_rate": 0.00048670912034538583,
+      "loss": 3.6409,
+      "step": 17600
+    },
+    {
+      "epoch": 1.9029649595687332,
+      "grad_norm": 0.5877305865287781,
+      "learning_rate": 0.0004863853211009174,
+      "loss": 3.6382,
+      "step": 17650
+    },
+    {
+      "epoch": 1.908355795148248,
+      "grad_norm": 0.5746439695358276,
+      "learning_rate": 0.000486061521856449,
+      "loss": 3.6504,
+      "step": 17700
+    },
+    {
+      "epoch": 1.9137466307277629,
+      "grad_norm": 0.5914478898048401,
+      "learning_rate": 0.00048573772261198054,
+      "loss": 3.6397,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9191374663072778,
+      "grad_norm": 0.5980408191680908,
+      "learning_rate": 0.00048541392336751214,
+      "loss": 3.655,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9245283018867925,
+      "grad_norm": 0.5955798625946045,
+      "learning_rate": 0.0004850901241230437,
+      "loss": 3.6641,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9299191374663072,
+      "grad_norm": 0.6046438813209534,
+      "learning_rate": 0.0004847663248785753,
+      "loss": 3.639,
+      "step": 17900
+    },
+    {
+      "epoch": 1.935309973045822,
+      "grad_norm": 0.6008595824241638,
+      "learning_rate": 0.0004844425256341068,
+      "loss": 3.6477,
+      "step": 17950
+    },
+    {
+      "epoch": 1.940700808625337,
+      "grad_norm": 0.5790556073188782,
+      "learning_rate": 0.00048411872638963834,
+      "loss": 3.6461,
+      "step": 18000
+    },
+    {
+      "epoch": 1.940700808625337,
+      "eval_accuracy": 0.3601660694198512,
+      "eval_loss": 3.5904529094696045,
+      "eval_runtime": 183.0994,
+      "eval_samples_per_second": 98.367,
+      "eval_steps_per_second": 6.15,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9460916442048517,
+      "grad_norm": 0.5579012036323547,
+      "learning_rate": 0.00048379492714516995,
+      "loss": 3.6269,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9514824797843666,
+      "grad_norm": 0.5438567399978638,
+      "learning_rate": 0.0004834711279007015,
+      "loss": 3.6563,
+      "step": 18100
+    },
+    {
+      "epoch": 1.9568733153638815,
+      "grad_norm": 0.5683955550193787,
+      "learning_rate": 0.0004831473286562331,
+      "loss": 3.6579,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9622641509433962,
+      "grad_norm": 0.5797238349914551,
+      "learning_rate": 0.00048282352941176465,
+      "loss": 3.6462,
+      "step": 18200
+    },
+    {
+      "epoch": 1.967654986522911,
+      "grad_norm": 0.5775489211082458,
+      "learning_rate": 0.00048249973016729626,
+      "loss": 3.6476,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9730458221024259,
+      "grad_norm": 0.5357772707939148,
+      "learning_rate": 0.0004821759309228278,
+      "loss": 3.6281,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9784366576819408,
+      "grad_norm": 0.7332191467285156,
+      "learning_rate": 0.00048185213167835936,
+      "loss": 3.6523,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9838274932614555,
+      "grad_norm": 0.5149871110916138,
+      "learning_rate": 0.00048152833243389096,
+      "loss": 3.6294,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9892183288409704,
+      "grad_norm": 0.5842679142951965,
+      "learning_rate": 0.0004812045331894225,
+      "loss": 3.6318,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9946091644204853,
+      "grad_norm": 0.5898716449737549,
+      "learning_rate": 0.0004808807339449541,
+      "loss": 3.6217,
+      "step": 18500
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.104094386100769,
+      "learning_rate": 0.0004805569347004856,
+      "loss": 3.6377,
+      "step": 18550
+    },
+    {
+      "epoch": 2.0053908355795147,
+      "grad_norm": 0.5916197299957275,
+      "learning_rate": 0.0004802331354560173,
+      "loss": 3.552,
+      "step": 18600
+    },
+    {
+      "epoch": 2.01078167115903,
+      "grad_norm": 0.5588786005973816,
+      "learning_rate": 0.00047990933621154877,
+      "loss": 3.5459,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0161725067385445,
+      "grad_norm": 0.5954398512840271,
+      "learning_rate": 0.0004795855369670804,
+      "loss": 3.5515,
+      "step": 18700
+    },
+    {
+      "epoch": 2.0215633423180592,
+      "grad_norm": 0.5872916579246521,
+      "learning_rate": 0.0004792617377226119,
+      "loss": 3.5445,
+      "step": 18750
+    },
+    {
+      "epoch": 2.026954177897574,
+      "grad_norm": 0.5799959301948547,
+      "learning_rate": 0.0004789379384781435,
+      "loss": 3.5516,
+      "step": 18800
+    },
+    {
+      "epoch": 2.032345013477089,
+      "grad_norm": 0.5512281060218811,
+      "learning_rate": 0.0004786141392336751,
+      "loss": 3.5498,
+      "step": 18850
+    },
+    {
+      "epoch": 2.0377358490566038,
+      "grad_norm": 0.6156049370765686,
+      "learning_rate": 0.00047829033998920663,
+      "loss": 3.5626,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0431266846361185,
+      "grad_norm": 0.5738759636878967,
+      "learning_rate": 0.00047796654074473824,
+      "loss": 3.5591,
+      "step": 18950
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "grad_norm": 0.5836319923400879,
+      "learning_rate": 0.0004776427415002698,
+      "loss": 3.5609,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "eval_accuracy": 0.36152336129442963,
+      "eval_loss": 3.581784725189209,
+      "eval_runtime": 183.3107,
+      "eval_samples_per_second": 98.254,
+      "eval_steps_per_second": 6.143,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0539083557951483,
+      "grad_norm": 0.5089203119277954,
+      "learning_rate": 0.0004773189422558014,
+      "loss": 3.5693,
+      "step": 19050
+    },
+    {
+      "epoch": 2.059299191374663,
+      "grad_norm": 0.5695809721946716,
+      "learning_rate": 0.00047699514301133294,
+      "loss": 3.5613,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0646900269541777,
+      "grad_norm": 0.5608691573143005,
+      "learning_rate": 0.00047667134376686455,
+      "loss": 3.5554,
+      "step": 19150
+    },
+    {
+      "epoch": 2.070080862533693,
+      "grad_norm": 0.5364721417427063,
+      "learning_rate": 0.0004763475445223961,
+      "loss": 3.5751,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0754716981132075,
+      "grad_norm": 0.5620273351669312,
+      "learning_rate": 0.0004760237452779276,
+      "loss": 3.5632,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0808625336927222,
+      "grad_norm": 0.5621458888053894,
+      "learning_rate": 0.00047570642201834855,
+      "loss": 3.5559,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0862533692722374,
+      "grad_norm": 0.6203056573867798,
+      "learning_rate": 0.00047538262277388015,
+      "loss": 3.5737,
+      "step": 19350
+    },
+    {
+      "epoch": 2.091644204851752,
+      "grad_norm": 0.5451525449752808,
+      "learning_rate": 0.0004750588235294117,
+      "loss": 3.5642,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0970350404312668,
+      "grad_norm": 0.6249733567237854,
+      "learning_rate": 0.0004747350242849433,
+      "loss": 3.5484,
+      "step": 19450
+    },
+    {
+      "epoch": 2.1024258760107815,
+      "grad_norm": 0.5598724484443665,
+      "learning_rate": 0.00047441122504047486,
+      "loss": 3.5713,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1078167115902966,
+      "grad_norm": 0.5830626487731934,
+      "learning_rate": 0.0004740874257960064,
+      "loss": 3.562,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1132075471698113,
+      "grad_norm": 0.5715135931968689,
+      "learning_rate": 0.000473763626551538,
+      "loss": 3.5612,
+      "step": 19600
+    },
+    {
+      "epoch": 2.118598382749326,
+      "grad_norm": 0.6303309798240662,
+      "learning_rate": 0.00047343982730706956,
+      "loss": 3.5628,
+      "step": 19650
+    },
+    {
+      "epoch": 2.123989218328841,
+      "grad_norm": 0.5836952328681946,
+      "learning_rate": 0.00047311602806260117,
+      "loss": 3.5534,
+      "step": 19700
+    },
+    {
+      "epoch": 2.129380053908356,
+      "grad_norm": 0.6122373938560486,
+      "learning_rate": 0.0004727922288181327,
+      "loss": 3.5445,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1347708894878705,
+      "grad_norm": 0.5842995047569275,
+      "learning_rate": 0.0004724684295736643,
+      "loss": 3.566,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1401617250673857,
+      "grad_norm": 0.585098147392273,
+      "learning_rate": 0.0004721446303291959,
+      "loss": 3.5585,
+      "step": 19850
+    },
+    {
+      "epoch": 2.1455525606469004,
+      "grad_norm": 0.5444265604019165,
+      "learning_rate": 0.0004718208310847275,
+      "loss": 3.5622,
+      "step": 19900
+    },
+    {
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.5859130620956421,
+      "learning_rate": 0.000471497031840259,
+      "loss": 3.5544,
+      "step": 19950
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "grad_norm": 0.6213077306747437,
+      "learning_rate": 0.0004711732325957905,
+      "loss": 3.5583,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "eval_accuracy": 0.3627601571164219,
+      "eval_loss": 3.572903871536255,
+      "eval_runtime": 182.8426,
+      "eval_samples_per_second": 98.505,
+      "eval_steps_per_second": 6.158,
+      "step": 20000
+    },
+    {
+      "epoch": 2.161725067385445,
+      "grad_norm": 0.570515513420105,
+      "learning_rate": 0.00047084943335132213,
+      "loss": 3.5599,
+      "step": 20050
+    },
+    {
+      "epoch": 2.1671159029649596,
+      "grad_norm": 0.5801568031311035,
+      "learning_rate": 0.0004705256341068537,
+      "loss": 3.5563,
+      "step": 20100
+    },
+    {
+      "epoch": 2.1725067385444743,
+      "grad_norm": 0.5625453591346741,
+      "learning_rate": 0.0004702018348623853,
+      "loss": 3.5663,
+      "step": 20150
+    },
+    {
+      "epoch": 2.177897574123989,
+      "grad_norm": 0.5203284025192261,
+      "learning_rate": 0.00046987803561791684,
+      "loss": 3.5694,
+      "step": 20200
+    },
+    {
+      "epoch": 2.183288409703504,
+      "grad_norm": 0.5144570469856262,
+      "learning_rate": 0.00046955423637344844,
+      "loss": 3.5731,
+      "step": 20250
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 0.6565627455711365,
+      "learning_rate": 0.00046923043712898,
+      "loss": 3.5553,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1940700808625335,
+      "grad_norm": 0.5670105814933777,
+      "learning_rate": 0.0004689066378845116,
+      "loss": 3.5481,
+      "step": 20350
+    },
+    {
+      "epoch": 2.1994609164420487,
+      "grad_norm": 0.6016785502433777,
+      "learning_rate": 0.00046858283864004315,
+      "loss": 3.5625,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2048517520215634,
+      "grad_norm": 0.7234423756599426,
+      "learning_rate": 0.0004682590393955747,
+      "loss": 3.5573,
+      "step": 20450
+    },
+    {
+      "epoch": 2.210242587601078,
+      "grad_norm": 0.5361531972885132,
+      "learning_rate": 0.0004679352401511063,
+      "loss": 3.5735,
+      "step": 20500
+    },
+    {
+      "epoch": 2.215633423180593,
+      "grad_norm": 0.5746288299560547,
+      "learning_rate": 0.0004676114409066378,
+      "loss": 3.569,
+      "step": 20550
+    },
+    {
+      "epoch": 2.221024258760108,
+      "grad_norm": 0.5691829919815063,
+      "learning_rate": 0.00046728764166216946,
+      "loss": 3.564,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2264150943396226,
+      "grad_norm": 0.5995154976844788,
+      "learning_rate": 0.00046696384241770095,
+      "loss": 3.5719,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2318059299191373,
+      "grad_norm": 0.6250969767570496,
+      "learning_rate": 0.00046664004317323256,
+      "loss": 3.5569,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2371967654986524,
+      "grad_norm": 0.5564571619033813,
+      "learning_rate": 0.0004663162439287641,
+      "loss": 3.5747,
+      "step": 20750
+    },
+    {
+      "epoch": 2.242587601078167,
+      "grad_norm": 0.5897787809371948,
+      "learning_rate": 0.00046599244468429566,
+      "loss": 3.5692,
+      "step": 20800
+    },
+    {
+      "epoch": 2.247978436657682,
+      "grad_norm": 0.5856361985206604,
+      "learning_rate": 0.00046566864543982726,
+      "loss": 3.573,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2533692722371965,
+      "grad_norm": 0.5496835708618164,
+      "learning_rate": 0.0004653448461953588,
+      "loss": 3.5611,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2587601078167117,
+      "grad_norm": 0.6221925616264343,
+      "learning_rate": 0.0004650210469508904,
+      "loss": 3.5581,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "grad_norm": 0.5658513307571411,
+      "learning_rate": 0.00046469724770642197,
+      "loss": 3.5592,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "eval_accuracy": 0.3634160946001579,
+      "eval_loss": 3.564391851425171,
+      "eval_runtime": 183.2391,
+      "eval_samples_per_second": 98.292,
+      "eval_steps_per_second": 6.145,
+      "step": 21000
+    },
+    {
+      "epoch": 2.269541778975741,
+      "grad_norm": 0.5886167883872986,
+      "learning_rate": 0.0004643734484619536,
+      "loss": 3.5563,
+      "step": 21050
+    },
+    {
+      "epoch": 2.274932614555256,
+      "grad_norm": 0.6223853230476379,
+      "learning_rate": 0.0004640496492174851,
+      "loss": 3.545,
+      "step": 21100
+    },
+    {
+      "epoch": 2.280323450134771,
+      "grad_norm": 0.6089233756065369,
+      "learning_rate": 0.00046372584997301673,
+      "loss": 3.5734,
+      "step": 21150
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.5425137877464294,
+      "learning_rate": 0.0004634020507285483,
+      "loss": 3.5528,
+      "step": 21200
+    },
+    {
+      "epoch": 2.2911051212938007,
+      "grad_norm": 0.5784063935279846,
+      "learning_rate": 0.0004630782514840798,
+      "loss": 3.5513,
+      "step": 21250
+    },
+    {
+      "epoch": 2.2964959568733154,
+      "grad_norm": 0.5775195360183716,
+      "learning_rate": 0.0004627544522396114,
+      "loss": 3.5616,
+      "step": 21300
+    },
+    {
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.5808774828910828,
+      "learning_rate": 0.00046243712898003233,
+      "loss": 3.5398,
+      "step": 21350
+    },
+    {
+      "epoch": 2.3072776280323453,
+      "grad_norm": 0.5682329535484314,
+      "learning_rate": 0.0004621133297355639,
+      "loss": 3.5759,
+      "step": 21400
+    },
+    {
+      "epoch": 2.31266846361186,
+      "grad_norm": 0.6007602214813232,
+      "learning_rate": 0.0004617895304910955,
+      "loss": 3.5335,
+      "step": 21450
+    },
+    {
+      "epoch": 2.3180592991913747,
+      "grad_norm": 0.6135632991790771,
+      "learning_rate": 0.00046146573124662704,
+      "loss": 3.5656,
+      "step": 21500
+    },
+    {
+      "epoch": 2.3234501347708894,
+      "grad_norm": 0.6100749969482422,
+      "learning_rate": 0.00046114193200215864,
+      "loss": 3.5654,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3288409703504045,
+      "grad_norm": 0.5744963884353638,
+      "learning_rate": 0.0004608181327576902,
+      "loss": 3.5383,
+      "step": 21600
+    },
+    {
+      "epoch": 2.334231805929919,
+      "grad_norm": 0.6241092085838318,
+      "learning_rate": 0.00046049433351322175,
+      "loss": 3.5591,
+      "step": 21650
+    },
+    {
+      "epoch": 2.339622641509434,
+      "grad_norm": 0.5725473761558533,
+      "learning_rate": 0.00046017053426875335,
+      "loss": 3.5715,
+      "step": 21700
+    },
+    {
+      "epoch": 2.3450134770889486,
+      "grad_norm": 0.6116576790809631,
+      "learning_rate": 0.0004598467350242849,
+      "loss": 3.5538,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3504043126684637,
+      "grad_norm": 0.5640816688537598,
+      "learning_rate": 0.0004595229357798165,
+      "loss": 3.5645,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3557951482479784,
+      "grad_norm": 0.6325236558914185,
+      "learning_rate": 0.00045919913653534806,
+      "loss": 3.5661,
+      "step": 21850
+    },
+    {
+      "epoch": 2.361185983827493,
+      "grad_norm": 0.5464061498641968,
+      "learning_rate": 0.00045887533729087966,
+      "loss": 3.5576,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3665768194070083,
+      "grad_norm": 0.5492430329322815,
+      "learning_rate": 0.00045855153804641116,
+      "loss": 3.5492,
+      "step": 21950
+    },
+    {
+      "epoch": 2.371967654986523,
+      "grad_norm": 0.5400915145874023,
+      "learning_rate": 0.0004582277388019427,
+      "loss": 3.5447,
+      "step": 22000
+    },
+    {
+      "epoch": 2.371967654986523,
+      "eval_accuracy": 0.36474970014519287,
+      "eval_loss": 3.5518150329589844,
+      "eval_runtime": 183.1051,
+      "eval_samples_per_second": 98.364,
+      "eval_steps_per_second": 6.149,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 0.5935810804367065,
+      "learning_rate": 0.0004579039395574743,
+      "loss": 3.5459,
+      "step": 22050
+    },
+    {
+      "epoch": 2.382749326145553,
+      "grad_norm": 0.5556415319442749,
+      "learning_rate": 0.00045758014031300586,
+      "loss": 3.5638,
+      "step": 22100
+    },
+    {
+      "epoch": 2.3881401617250675,
+      "grad_norm": 0.5712567567825317,
+      "learning_rate": 0.0004572628170534268,
+      "loss": 3.5564,
+      "step": 22150
+    },
+    {
+      "epoch": 2.393530997304582,
+      "grad_norm": 0.5914583206176758,
+      "learning_rate": 0.0004569390178089584,
+      "loss": 3.5452,
+      "step": 22200
+    },
+    {
+      "epoch": 2.398921832884097,
+      "grad_norm": 0.7909786701202393,
+      "learning_rate": 0.00045661521856448997,
+      "loss": 3.539,
+      "step": 22250
+    },
+    {
+      "epoch": 2.404312668463612,
+      "grad_norm": 0.6220540404319763,
+      "learning_rate": 0.0004562914193200216,
+      "loss": 3.5651,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4097035040431267,
+      "grad_norm": 0.5887907147407532,
+      "learning_rate": 0.0004559676200755531,
+      "loss": 3.5751,
+      "step": 22350
+    },
+    {
+      "epoch": 2.4150943396226414,
+      "grad_norm": 0.6566967964172363,
+      "learning_rate": 0.0004556438208310847,
+      "loss": 3.5596,
+      "step": 22400
+    },
+    {
+      "epoch": 2.420485175202156,
+      "grad_norm": 0.5419936776161194,
+      "learning_rate": 0.0004553200215866163,
+      "loss": 3.5656,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4258760107816713,
+      "grad_norm": 0.5918455719947815,
+      "learning_rate": 0.00045499622234214783,
+      "loss": 3.5622,
+      "step": 22500
+    },
+    {
+      "epoch": 2.431266846361186,
+      "grad_norm": 0.5798980593681335,
+      "learning_rate": 0.00045467242309767944,
+      "loss": 3.5416,
+      "step": 22550
+    },
+    {
+      "epoch": 2.4366576819407006,
+      "grad_norm": 0.623996376991272,
+      "learning_rate": 0.00045434862385321093,
+      "loss": 3.5592,
+      "step": 22600
+    },
+    {
+      "epoch": 2.442048517520216,
+      "grad_norm": 0.550102710723877,
+      "learning_rate": 0.0004540248246087426,
+      "loss": 3.5557,
+      "step": 22650
+    },
+    {
+      "epoch": 2.4474393530997305,
+      "grad_norm": 0.6159896850585938,
+      "learning_rate": 0.0004537010253642741,
+      "loss": 3.5641,
+      "step": 22700
+    },
+    {
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.5841001272201538,
+      "learning_rate": 0.00045337722611980564,
+      "loss": 3.5525,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4582210242587603,
+      "grad_norm": 0.5647518038749695,
+      "learning_rate": 0.00045305342687533724,
+      "loss": 3.5495,
+      "step": 22800
+    },
+    {
+      "epoch": 2.463611859838275,
+      "grad_norm": 0.5946810245513916,
+      "learning_rate": 0.0004527296276308688,
+      "loss": 3.5387,
+      "step": 22850
+    },
+    {
+      "epoch": 2.4690026954177897,
+      "grad_norm": 0.5658169984817505,
+      "learning_rate": 0.0004524058283864004,
+      "loss": 3.5527,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4743935309973044,
+      "grad_norm": 0.6262547373771667,
+      "learning_rate": 0.00045208202914193195,
+      "loss": 3.5385,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "grad_norm": 0.5860013365745544,
+      "learning_rate": 0.00045175822989746355,
+      "loss": 3.5527,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "eval_accuracy": 0.36564021921372036,
+      "eval_loss": 3.5422823429107666,
+      "eval_runtime": 183.1023,
+      "eval_samples_per_second": 98.366,
+      "eval_steps_per_second": 6.15,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4851752021563343,
+      "grad_norm": 0.6349847912788391,
+      "learning_rate": 0.0004514344306529951,
+      "loss": 3.5606,
+      "step": 23050
+    },
+    {
+      "epoch": 2.490566037735849,
+      "grad_norm": 0.5899332761764526,
+      "learning_rate": 0.0004511106314085267,
+      "loss": 3.5561,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4959568733153636,
+      "grad_norm": 0.7006366848945618,
+      "learning_rate": 0.00045078683216405826,
+      "loss": 3.5688,
+      "step": 23150
+    },
+    {
+      "epoch": 2.501347708894879,
+      "grad_norm": 0.6054086685180664,
+      "learning_rate": 0.00045046303291958976,
+      "loss": 3.5605,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5067385444743935,
+      "grad_norm": 0.5730805397033691,
+      "learning_rate": 0.0004501392336751214,
+      "loss": 3.5609,
+      "step": 23250
+    },
+    {
+      "epoch": 2.512129380053908,
+      "grad_norm": 0.5215578675270081,
+      "learning_rate": 0.0004498154344306529,
+      "loss": 3.5586,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5175202156334233,
+      "grad_norm": 0.5702964067459106,
+      "learning_rate": 0.0004494916351861845,
+      "loss": 3.5549,
+      "step": 23350
+    },
+    {
+      "epoch": 2.522911051212938,
+      "grad_norm": 0.6069727540016174,
+      "learning_rate": 0.00044916783594171607,
+      "loss": 3.534,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5283018867924527,
+      "grad_norm": 0.5799237489700317,
+      "learning_rate": 0.00044884403669724767,
+      "loss": 3.5509,
+      "step": 23450
+    },
+    {
+      "epoch": 2.533692722371968,
+      "grad_norm": 0.6026593446731567,
+      "learning_rate": 0.0004485202374527792,
+      "loss": 3.5454,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5390835579514826,
+      "grad_norm": 0.557666003704071,
+      "learning_rate": 0.00044819643820831083,
+      "loss": 3.548,
+      "step": 23550
+    },
+    {
+      "epoch": 2.5444743935309972,
+      "grad_norm": 0.6157044768333435,
+      "learning_rate": 0.0004478726389638424,
+      "loss": 3.5527,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5498652291105124,
+      "grad_norm": 0.5705444812774658,
+      "learning_rate": 0.00044754883971937393,
+      "loss": 3.5384,
+      "step": 23650
+    },
+    {
+      "epoch": 2.555256064690027,
+      "grad_norm": 0.5903183221817017,
+      "learning_rate": 0.00044722504047490553,
+      "loss": 3.529,
+      "step": 23700
+    },
+    {
+      "epoch": 2.560646900269542,
+      "grad_norm": 0.5920628905296326,
+      "learning_rate": 0.0004469012412304371,
+      "loss": 3.5377,
+      "step": 23750
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 0.5568437576293945,
+      "learning_rate": 0.0004465774419859687,
+      "loss": 3.5558,
+      "step": 23800
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.5784709453582764,
+      "learning_rate": 0.00044625364274150024,
+      "loss": 3.5412,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5768194070080863,
+      "grad_norm": 0.5762706398963928,
+      "learning_rate": 0.00044592984349703184,
+      "loss": 3.5437,
+      "step": 23900
+    },
+    {
+      "epoch": 2.582210242587601,
+      "grad_norm": 0.6410912871360779,
+      "learning_rate": 0.00044560604425256334,
+      "loss": 3.5356,
+      "step": 23950
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "grad_norm": 0.579069197177887,
+      "learning_rate": 0.000445282245008095,
+      "loss": 3.541,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "eval_accuracy": 0.3668237751206835,
+      "eval_loss": 3.531728982925415,
+      "eval_runtime": 183.177,
+      "eval_samples_per_second": 98.326,
+      "eval_steps_per_second": 6.147,
+      "step": 24000
+    },
+    {
+      "epoch": 2.592991913746631,
+      "grad_norm": 0.5918624401092529,
+      "learning_rate": 0.0004449584457636265,
+      "loss": 3.5656,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5983827493261455,
+      "grad_norm": 0.6068958640098572,
+      "learning_rate": 0.00044463464651915805,
+      "loss": 3.5392,
+      "step": 24100
+    },
+    {
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.5975248217582703,
+      "learning_rate": 0.00044431084727468965,
+      "loss": 3.5544,
+      "step": 24150
+    },
+    {
+      "epoch": 2.6091644204851754,
+      "grad_norm": 0.5652947425842285,
+      "learning_rate": 0.0004439870480302212,
+      "loss": 3.5379,
+      "step": 24200
+    },
+    {
+      "epoch": 2.61455525606469,
+      "grad_norm": 0.5702681541442871,
+      "learning_rate": 0.0004436632487857528,
+      "loss": 3.5553,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6199460916442048,
+      "grad_norm": 0.5995779037475586,
+      "learning_rate": 0.00044333944954128436,
+      "loss": 3.5496,
+      "step": 24300
+    },
+    {
+      "epoch": 2.62533692722372,
+      "grad_norm": 0.5882319211959839,
+      "learning_rate": 0.00044301565029681596,
+      "loss": 3.541,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6307277628032346,
+      "grad_norm": 0.5981907844543457,
+      "learning_rate": 0.0004426918510523475,
+      "loss": 3.5461,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6361185983827493,
+      "grad_norm": 0.5818279385566711,
+      "learning_rate": 0.0004423680518078791,
+      "loss": 3.5445,
+      "step": 24450
+    },
+    {
+      "epoch": 2.641509433962264,
+      "grad_norm": 0.5762891173362732,
+      "learning_rate": 0.00044204425256341067,
+      "loss": 3.5515,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6469002695417787,
+      "grad_norm": 0.5954058766365051,
+      "learning_rate": 0.0004417204533189422,
+      "loss": 3.5369,
+      "step": 24550
+    },
+    {
+      "epoch": 2.652291105121294,
+      "grad_norm": 0.577364444732666,
+      "learning_rate": 0.0004413966540744738,
+      "loss": 3.544,
+      "step": 24600
+    },
+    {
+      "epoch": 2.6576819407008085,
+      "grad_norm": 0.5351988077163696,
+      "learning_rate": 0.0004410728548300053,
+      "loss": 3.5396,
+      "step": 24650
+    },
+    {
+      "epoch": 2.6630727762803232,
+      "grad_norm": 0.5411438345909119,
+      "learning_rate": 0.0004407490555855369,
+      "loss": 3.5456,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6684636118598384,
+      "grad_norm": 0.5812211632728577,
+      "learning_rate": 0.0004404252563410685,
+      "loss": 3.5628,
+      "step": 24750
+    },
+    {
+      "epoch": 2.673854447439353,
+      "grad_norm": 0.5671088099479675,
+      "learning_rate": 0.0004401014570966001,
+      "loss": 3.5564,
+      "step": 24800
+    },
+    {
+      "epoch": 2.6792452830188678,
+      "grad_norm": 0.5471763610839844,
+      "learning_rate": 0.00043977765785213163,
+      "loss": 3.558,
+      "step": 24850
+    },
+    {
+      "epoch": 2.684636118598383,
+      "grad_norm": 0.6299545764923096,
+      "learning_rate": 0.0004394538586076632,
+      "loss": 3.5286,
+      "step": 24900
+    },
+    {
+      "epoch": 2.6900269541778976,
+      "grad_norm": 0.584722101688385,
+      "learning_rate": 0.0004391300593631948,
+      "loss": 3.5491,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "grad_norm": 0.574306309223175,
+      "learning_rate": 0.00043880626011872634,
+      "loss": 3.5277,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "eval_accuracy": 0.3673932249057192,
+      "eval_loss": 3.5235400199890137,
+      "eval_runtime": 183.145,
+      "eval_samples_per_second": 98.343,
+      "eval_steps_per_second": 6.148,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7008086253369274,
+      "grad_norm": 0.5931011438369751,
+      "learning_rate": 0.00043848246087425794,
+      "loss": 3.5396,
+      "step": 25050
+    },
+    {
+      "epoch": 2.706199460916442,
+      "grad_norm": 0.5480889081954956,
+      "learning_rate": 0.0004381586616297895,
+      "loss": 3.5412,
+      "step": 25100
+    },
+    {
+      "epoch": 2.711590296495957,
+      "grad_norm": 0.5581277012825012,
+      "learning_rate": 0.0004378348623853211,
+      "loss": 3.5415,
+      "step": 25150
+    },
+    {
+      "epoch": 2.7169811320754715,
+      "grad_norm": 0.5917746424674988,
+      "learning_rate": 0.00043751106314085265,
+      "loss": 3.5425,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7223719676549867,
+      "grad_norm": 0.641676664352417,
+      "learning_rate": 0.00043718726389638425,
+      "loss": 3.5354,
+      "step": 25250
+    },
+    {
+      "epoch": 2.7277628032345014,
+      "grad_norm": 0.5899236798286438,
+      "learning_rate": 0.00043686346465191575,
+      "loss": 3.5433,
+      "step": 25300
+    },
+    {
+      "epoch": 2.733153638814016,
+      "grad_norm": 0.638559103012085,
+      "learning_rate": 0.0004365396654074473,
+      "loss": 3.5397,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7385444743935308,
+      "grad_norm": 0.5510205626487732,
+      "learning_rate": 0.0004362158661629789,
+      "loss": 3.5326,
+      "step": 25400
+    },
+    {
+      "epoch": 2.743935309973046,
+      "grad_norm": 0.6142581105232239,
+      "learning_rate": 0.00043589206691851045,
+      "loss": 3.55,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7493261455525606,
+      "grad_norm": 0.5508300065994263,
+      "learning_rate": 0.00043556826767404206,
+      "loss": 3.5277,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.647884726524353,
+      "learning_rate": 0.0004352444684295736,
+      "loss": 3.5193,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7601078167115904,
+      "grad_norm": 0.5603880882263184,
+      "learning_rate": 0.0004349206691851052,
+      "loss": 3.5248,
+      "step": 25600
+    },
+    {
+      "epoch": 2.765498652291105,
+      "grad_norm": 0.57041335105896,
+      "learning_rate": 0.00043459686994063676,
+      "loss": 3.5298,
+      "step": 25650
+    },
+    {
+      "epoch": 2.77088948787062,
+      "grad_norm": 0.5930079221725464,
+      "learning_rate": 0.00043427307069616837,
+      "loss": 3.5188,
+      "step": 25700
+    },
+    {
+      "epoch": 2.776280323450135,
+      "grad_norm": 0.5885521769523621,
+      "learning_rate": 0.0004339492714516999,
+      "loss": 3.5224,
+      "step": 25750
+    },
+    {
+      "epoch": 2.7816711590296497,
+      "grad_norm": 0.617006778717041,
+      "learning_rate": 0.00043362547220723147,
+      "loss": 3.5361,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7870619946091644,
+      "grad_norm": 0.5954247713088989,
+      "learning_rate": 0.0004333016729627631,
+      "loss": 3.5215,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7924528301886795,
+      "grad_norm": 0.5700495839118958,
+      "learning_rate": 0.0004329778737182946,
+      "loss": 3.5271,
+      "step": 25900
+    },
+    {
+      "epoch": 2.797843665768194,
+      "grad_norm": 0.5143706202507019,
+      "learning_rate": 0.00043265407447382623,
+      "loss": 3.5362,
+      "step": 25950
+    },
+    {
+      "epoch": 2.803234501347709,
+      "grad_norm": 0.6275197267532349,
+      "learning_rate": 0.0004323302752293577,
+      "loss": 3.5395,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803234501347709,
+      "eval_accuracy": 0.3686822827667502,
+      "eval_loss": 3.5142149925231934,
+      "eval_runtime": 183.2956,
+      "eval_samples_per_second": 98.262,
+      "eval_steps_per_second": 6.143,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8086253369272236,
+      "grad_norm": 0.6447458267211914,
+      "learning_rate": 0.00043200647598488933,
+      "loss": 3.5427,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8140161725067383,
+      "grad_norm": 0.5703551173210144,
+      "learning_rate": 0.0004316826767404209,
+      "loss": 3.5212,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8194070080862534,
+      "grad_norm": 0.6103937029838562,
+      "learning_rate": 0.00043136535348084183,
+      "loss": 3.53,
+      "step": 26150
+    },
+    {
+      "epoch": 2.824797843665768,
+      "grad_norm": 0.6109429001808167,
+      "learning_rate": 0.0004310415542363734,
+      "loss": 3.5268,
+      "step": 26200
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.616741418838501,
+      "learning_rate": 0.000430717754991905,
+      "loss": 3.5213,
+      "step": 26250
+    },
+    {
+      "epoch": 2.835579514824798,
+      "grad_norm": 0.5909930467605591,
+      "learning_rate": 0.00043039395574743654,
+      "loss": 3.5342,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8409703504043127,
+      "grad_norm": 0.6086243391036987,
+      "learning_rate": 0.00043007015650296814,
+      "loss": 3.54,
+      "step": 26350
+    },
+    {
+      "epoch": 2.8463611859838274,
+      "grad_norm": 0.6133831739425659,
+      "learning_rate": 0.0004297463572584997,
+      "loss": 3.5273,
+      "step": 26400
+    },
+    {
+      "epoch": 2.8517520215633425,
+      "grad_norm": 0.5661080479621887,
+      "learning_rate": 0.0004294225580140313,
+      "loss": 3.5292,
+      "step": 26450
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.623001217842102,
+      "learning_rate": 0.00042909875876956285,
+      "loss": 3.5253,
+      "step": 26500
+    },
+    {
+      "epoch": 2.862533692722372,
+      "grad_norm": 0.5956571102142334,
+      "learning_rate": 0.0004287749595250944,
+      "loss": 3.517,
+      "step": 26550
+    },
+    {
+      "epoch": 2.867924528301887,
+      "grad_norm": 0.6012107729911804,
+      "learning_rate": 0.000428451160280626,
+      "loss": 3.5283,
+      "step": 26600
+    },
+    {
+      "epoch": 2.8733153638814017,
+      "grad_norm": 0.5730655789375305,
+      "learning_rate": 0.0004281273610361575,
+      "loss": 3.5232,
+      "step": 26650
+    },
+    {
+      "epoch": 2.8787061994609164,
+      "grad_norm": 0.5700306296348572,
+      "learning_rate": 0.0004278035617916891,
+      "loss": 3.5222,
+      "step": 26700
+    },
+    {
+      "epoch": 2.884097035040431,
+      "grad_norm": 0.5891921520233154,
+      "learning_rate": 0.00042747976254722066,
+      "loss": 3.5415,
+      "step": 26750
+    },
+    {
+      "epoch": 2.889487870619946,
+      "grad_norm": 0.5979741215705872,
+      "learning_rate": 0.00042715596330275226,
+      "loss": 3.5262,
+      "step": 26800
+    },
+    {
+      "epoch": 2.894878706199461,
+      "grad_norm": 0.5651412010192871,
+      "learning_rate": 0.0004268321640582838,
+      "loss": 3.5226,
+      "step": 26850
+    },
+    {
+      "epoch": 2.9002695417789757,
+      "grad_norm": 0.6269994974136353,
+      "learning_rate": 0.0004265083648138154,
+      "loss": 3.5465,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.5697436928749084,
+      "learning_rate": 0.00042618456556934697,
+      "loss": 3.5394,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "grad_norm": 0.5543894171714783,
+      "learning_rate": 0.0004258607663248785,
+      "loss": 3.522,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "eval_accuracy": 0.3693975447272329,
+      "eval_loss": 3.504847764968872,
+      "eval_runtime": 183.0242,
+      "eval_samples_per_second": 98.408,
+      "eval_steps_per_second": 6.152,
+      "step": 27000
+    },
+    {
+      "epoch": 2.91644204851752,
+      "grad_norm": 0.6050317883491516,
+      "learning_rate": 0.0004255369670804101,
+      "loss": 3.5397,
+      "step": 27050
+    },
+    {
+      "epoch": 2.921832884097035,
+      "grad_norm": 0.538213312625885,
+      "learning_rate": 0.0004252131678359417,
+      "loss": 3.5442,
+      "step": 27100
+    },
+    {
+      "epoch": 2.92722371967655,
+      "grad_norm": 0.6090800762176514,
+      "learning_rate": 0.0004248893685914733,
+      "loss": 3.5282,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9326145552560647,
+      "grad_norm": 0.6525107026100159,
+      "learning_rate": 0.00042456556934700483,
+      "loss": 3.519,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9380053908355794,
+      "grad_norm": 0.5758708119392395,
+      "learning_rate": 0.00042424177010253643,
+      "loss": 3.5319,
+      "step": 27250
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 0.6072579026222229,
+      "learning_rate": 0.00042391797085806793,
+      "loss": 3.5283,
+      "step": 27300
+    },
+    {
+      "epoch": 2.9487870619946093,
+      "grad_norm": 0.5943818092346191,
+      "learning_rate": 0.0004235941716135995,
+      "loss": 3.5201,
+      "step": 27350
+    },
+    {
+      "epoch": 2.954177897574124,
+      "grad_norm": 0.5442832112312317,
+      "learning_rate": 0.0004232703723691311,
+      "loss": 3.5261,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9595687331536387,
+      "grad_norm": 0.5325363874435425,
+      "learning_rate": 0.00042294657312466264,
+      "loss": 3.5216,
+      "step": 27450
+    },
+    {
+      "epoch": 2.964959568733154,
+      "grad_norm": 0.5526369214057922,
+      "learning_rate": 0.00042262277388019424,
+      "loss": 3.5286,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9703504043126685,
+      "grad_norm": 0.5498561263084412,
+      "learning_rate": 0.0004222989746357258,
+      "loss": 3.5312,
+      "step": 27550
+    },
+    {
+      "epoch": 2.975741239892183,
+      "grad_norm": 0.5401663780212402,
+      "learning_rate": 0.0004219751753912574,
+      "loss": 3.5316,
+      "step": 27600
+    },
+    {
+      "epoch": 2.981132075471698,
+      "grad_norm": 0.5896415114402771,
+      "learning_rate": 0.00042165137614678895,
+      "loss": 3.5335,
+      "step": 27650
+    },
+    {
+      "epoch": 2.986522911051213,
+      "grad_norm": 0.6067506670951843,
+      "learning_rate": 0.00042132757690232055,
+      "loss": 3.5246,
+      "step": 27700
+    },
+    {
+      "epoch": 2.9919137466307277,
+      "grad_norm": 0.5816178321838379,
+      "learning_rate": 0.0004210037776578521,
+      "loss": 3.5349,
+      "step": 27750
+    },
+    {
+      "epoch": 2.9973045822102424,
+      "grad_norm": 0.6244997382164001,
+      "learning_rate": 0.00042067997841338365,
+      "loss": 3.5308,
+      "step": 27800
+    },
+    {
+      "epoch": 3.0026954177897576,
+      "grad_norm": 0.602767825126648,
+      "learning_rate": 0.00042035617916891526,
+      "loss": 3.4765,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0080862533692723,
+      "grad_norm": 0.6553490161895752,
+      "learning_rate": 0.0004200323799244468,
+      "loss": 3.4268,
+      "step": 27900
+    },
+    {
+      "epoch": 3.013477088948787,
+      "grad_norm": 0.6417362689971924,
+      "learning_rate": 0.0004197085806799784,
+      "loss": 3.4265,
+      "step": 27950
+    },
+    {
+      "epoch": 3.018867924528302,
+      "grad_norm": 0.6044764518737793,
+      "learning_rate": 0.0004193847814355099,
+      "loss": 3.4455,
+      "step": 28000
+    },
+    {
+      "epoch": 3.018867924528302,
+      "eval_accuracy": 0.37052916455382073,
+      "eval_loss": 3.5026047229766846,
+      "eval_runtime": 183.178,
+      "eval_samples_per_second": 98.325,
+      "eval_steps_per_second": 6.147,
+      "step": 28000
+    },
+    {
+      "epoch": 3.024258760107817,
+      "grad_norm": 0.6100830435752869,
+      "learning_rate": 0.0004190609821910415,
+      "loss": 3.4397,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0296495956873315,
+      "grad_norm": 0.5654774904251099,
+      "learning_rate": 0.00041873718294657306,
+      "loss": 3.4469,
+      "step": 28100
+    },
+    {
+      "epoch": 3.035040431266846,
+      "grad_norm": 0.5853551030158997,
+      "learning_rate": 0.000418419859686994,
+      "loss": 3.4362,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0404312668463613,
+      "grad_norm": 0.6212331652641296,
+      "learning_rate": 0.00041809606044252557,
+      "loss": 3.442,
+      "step": 28200
+    },
+    {
+      "epoch": 3.045822102425876,
+      "grad_norm": 0.5993689894676208,
+      "learning_rate": 0.00041777226119805717,
+      "loss": 3.4229,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0512129380053907,
+      "grad_norm": 0.6013529300689697,
+      "learning_rate": 0.0004174484619535887,
+      "loss": 3.4456,
+      "step": 28300
+    },
+    {
+      "epoch": 3.056603773584906,
+      "grad_norm": 0.599606990814209,
+      "learning_rate": 0.00041712466270912033,
+      "loss": 3.4189,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0619946091644206,
+      "grad_norm": 0.5922053456306458,
+      "learning_rate": 0.0004168008634646519,
+      "loss": 3.4486,
+      "step": 28400
+    },
+    {
+      "epoch": 3.0673854447439353,
+      "grad_norm": 0.6180664896965027,
+      "learning_rate": 0.0004164770642201835,
+      "loss": 3.4173,
+      "step": 28450
+    },
+    {
+      "epoch": 3.07277628032345,
+      "grad_norm": 0.6056652665138245,
+      "learning_rate": 0.00041615326497571503,
+      "loss": 3.4427,
+      "step": 28500
+    },
+    {
+      "epoch": 3.078167115902965,
+      "grad_norm": 0.6129570603370667,
+      "learning_rate": 0.0004158294657312466,
+      "loss": 3.4542,
+      "step": 28550
+    },
+    {
+      "epoch": 3.08355795148248,
+      "grad_norm": 0.6358380317687988,
+      "learning_rate": 0.0004155056664867782,
+      "loss": 3.4405,
+      "step": 28600
+    },
+    {
+      "epoch": 3.0889487870619945,
+      "grad_norm": 0.5838945508003235,
+      "learning_rate": 0.0004151818672423097,
+      "loss": 3.4519,
+      "step": 28650
+    },
+    {
+      "epoch": 3.0943396226415096,
+      "grad_norm": 0.5569682121276855,
+      "learning_rate": 0.0004148580679978413,
+      "loss": 3.4472,
+      "step": 28700
+    },
+    {
+      "epoch": 3.0997304582210243,
+      "grad_norm": 0.6249170899391174,
+      "learning_rate": 0.00041453426875337284,
+      "loss": 3.4719,
+      "step": 28750
+    },
+    {
+      "epoch": 3.105121293800539,
+      "grad_norm": 0.5985543727874756,
+      "learning_rate": 0.00041421046950890445,
+      "loss": 3.433,
+      "step": 28800
+    },
+    {
+      "epoch": 3.1105121293800537,
+      "grad_norm": 0.6323471665382385,
+      "learning_rate": 0.000413886670264436,
+      "loss": 3.4428,
+      "step": 28850
+    },
+    {
+      "epoch": 3.115902964959569,
+      "grad_norm": 0.5465901494026184,
+      "learning_rate": 0.0004135628710199676,
+      "loss": 3.4372,
+      "step": 28900
+    },
+    {
+      "epoch": 3.1212938005390836,
+      "grad_norm": 0.6596279144287109,
+      "learning_rate": 0.00041323907177549915,
+      "loss": 3.4682,
+      "step": 28950
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "grad_norm": 0.6128025054931641,
+      "learning_rate": 0.0004129152725310307,
+      "loss": 3.4399,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "eval_accuracy": 0.37060902442636434,
+      "eval_loss": 3.4956202507019043,
+      "eval_runtime": 183.1389,
+      "eval_samples_per_second": 98.346,
+      "eval_steps_per_second": 6.148,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1320754716981134,
+      "grad_norm": 0.621014416217804,
+      "learning_rate": 0.0004125914732865623,
+      "loss": 3.4579,
+      "step": 29050
+    },
+    {
+      "epoch": 3.137466307277628,
+      "grad_norm": 0.576972246170044,
+      "learning_rate": 0.00041226767404209386,
+      "loss": 3.4598,
+      "step": 29100
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.6160667538642883,
+      "learning_rate": 0.00041194387479762546,
+      "loss": 3.4504,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1482479784366575,
+      "grad_norm": 0.5812036395072937,
+      "learning_rate": 0.000411620075553157,
+      "loss": 3.4588,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1536388140161726,
+      "grad_norm": 0.5818783640861511,
+      "learning_rate": 0.0004112962763086886,
+      "loss": 3.4629,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1590296495956873,
+      "grad_norm": 0.6303808689117432,
+      "learning_rate": 0.0004109724770642201,
+      "loss": 3.4655,
+      "step": 29300
+    },
+    {
+      "epoch": 3.164420485175202,
+      "grad_norm": 0.6596410870552063,
+      "learning_rate": 0.00041064867781975177,
+      "loss": 3.4566,
+      "step": 29350
+    },
+    {
+      "epoch": 3.169811320754717,
+      "grad_norm": 0.581000030040741,
+      "learning_rate": 0.00041032487857528327,
+      "loss": 3.4488,
+      "step": 29400
+    },
+    {
+      "epoch": 3.175202156334232,
+      "grad_norm": 0.6004059314727783,
+      "learning_rate": 0.0004100010793308148,
+      "loss": 3.4341,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1805929919137466,
+      "grad_norm": 0.5809293985366821,
+      "learning_rate": 0.0004096772800863464,
+      "loss": 3.4601,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1859838274932613,
+      "grad_norm": 0.6131229400634766,
+      "learning_rate": 0.000409353480841878,
+      "loss": 3.4449,
+      "step": 29550
+    },
+    {
+      "epoch": 3.1913746630727764,
+      "grad_norm": 0.5700172185897827,
+      "learning_rate": 0.0004090296815974096,
+      "loss": 3.4522,
+      "step": 29600
+    },
+    {
+      "epoch": 3.196765498652291,
+      "grad_norm": 0.6279261112213135,
+      "learning_rate": 0.00040870588235294113,
+      "loss": 3.468,
+      "step": 29650
+    },
+    {
+      "epoch": 3.202156334231806,
+      "grad_norm": 0.6200495362281799,
+      "learning_rate": 0.00040838208310847273,
+      "loss": 3.4512,
+      "step": 29700
+    },
+    {
+      "epoch": 3.207547169811321,
+      "grad_norm": 0.5797140598297119,
+      "learning_rate": 0.0004080582838640043,
+      "loss": 3.4619,
+      "step": 29750
+    },
+    {
+      "epoch": 3.2129380053908356,
+      "grad_norm": 0.6226954460144043,
+      "learning_rate": 0.0004077344846195359,
+      "loss": 3.4709,
+      "step": 29800
+    },
+    {
+      "epoch": 3.2183288409703503,
+      "grad_norm": 0.5723651647567749,
+      "learning_rate": 0.00040741068537506744,
+      "loss": 3.4755,
+      "step": 29850
+    },
+    {
+      "epoch": 3.223719676549865,
+      "grad_norm": 0.5869478583335876,
+      "learning_rate": 0.000407086886130599,
+      "loss": 3.4414,
+      "step": 29900
+    },
+    {
+      "epoch": 3.22911051212938,
+      "grad_norm": 0.5735659599304199,
+      "learning_rate": 0.0004067630868861306,
+      "loss": 3.4457,
+      "step": 29950
+    },
+    {
+      "epoch": 3.234501347708895,
+      "grad_norm": 0.5913712978363037,
+      "learning_rate": 0.0004064392876416621,
+      "loss": 3.4688,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234501347708895,
+      "eval_accuracy": 0.37179301494487876,
+      "eval_loss": 3.491060256958008,
+      "eval_runtime": 183.4022,
+      "eval_samples_per_second": 98.205,
+      "eval_steps_per_second": 6.14,
+      "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
+}