diff --git "a/checkpoint-90000/trainer_state.json" "b/checkpoint-90000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-90000/trainer_state.json"
@@ -0,0 +1,13443 @@
+{
+  "best_metric": 3.3042373657226562,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M__6910/checkpoint-90000",
+  "epoch": 9.703504043126685,
+  "eval_steps": 1000,
+  "global_step": 90000,
+  "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
+    },
+    {
+      "epoch": 3.2398921832884096,
+      "grad_norm": 0.5873755812644958,
+      "learning_rate": 0.0004061154883971937,
+      "loss": 3.4555,
+      "step": 30050
+    },
+    {
+      "epoch": 3.2452830188679247,
+      "grad_norm": 0.5795559883117676,
+      "learning_rate": 0.00040579168915272525,
+      "loss": 3.4611,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2506738544474394,
+      "grad_norm": 0.5841328501701355,
+      "learning_rate": 0.0004054743658931462,
+      "loss": 3.468,
+      "step": 30150
+    },
+    {
+      "epoch": 3.256064690026954,
+      "grad_norm": 0.6134114265441895,
+      "learning_rate": 0.00040515056664867775,
+      "loss": 3.4516,
+      "step": 30200
+    },
+    {
+      "epoch": 3.2614555256064692,
+      "grad_norm": 0.6624637842178345,
+      "learning_rate": 0.00040482676740420935,
+      "loss": 3.4571,
+      "step": 30250
+    },
+    {
+      "epoch": 3.266846361185984,
+      "grad_norm": 0.61444491147995,
+      "learning_rate": 0.0004045029681597409,
+      "loss": 3.4429,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2722371967654986,
+      "grad_norm": 0.6057273745536804,
+      "learning_rate": 0.0004041791689152725,
+      "loss": 3.4574,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2776280323450133,
+      "grad_norm": 0.5823548436164856,
+      "learning_rate": 0.00040385536967080406,
+      "loss": 3.468,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2830188679245285,
+      "grad_norm": 0.5908448696136475,
+      "learning_rate": 0.00040353157042633567,
+      "loss": 3.4597,
+      "step": 30450
+    },
+    {
+      "epoch": 3.288409703504043,
+      "grad_norm": 0.6395969986915588,
+      "learning_rate": 0.0004032077711818672,
+      "loss": 3.4795,
+      "step": 30500
+    },
+    {
+      "epoch": 3.293800539083558,
+      "grad_norm": 0.5815283060073853,
+      "learning_rate": 0.0004028839719373988,
+      "loss": 3.4691,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2991913746630726,
+      "grad_norm": 0.5855560302734375,
+      "learning_rate": 0.00040256017269293037,
+      "loss": 3.4623,
+      "step": 30600
+    },
+    {
+      "epoch": 3.3045822102425877,
+      "grad_norm": 0.5978612899780273,
+      "learning_rate": 0.00040223637344846187,
+      "loss": 3.4608,
+      "step": 30650
+    },
+    {
+      "epoch": 3.3099730458221024,
+      "grad_norm": 0.635847270488739,
+      "learning_rate": 0.00040191257420399347,
+      "loss": 3.4523,
+      "step": 30700
+    },
+    {
+      "epoch": 3.315363881401617,
+      "grad_norm": 0.6061700582504272,
+      "learning_rate": 0.000401588774959525,
+      "loss": 3.4582,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3207547169811322,
+      "grad_norm": 0.6008456945419312,
+      "learning_rate": 0.00040126497571505663,
+      "loss": 3.46,
+      "step": 30800
+    },
+    {
+      "epoch": 3.326145552560647,
+      "grad_norm": 0.5615379214286804,
+      "learning_rate": 0.0004009411764705882,
+      "loss": 3.4519,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3315363881401616,
+      "grad_norm": 0.6048398017883301,
+      "learning_rate": 0.0004006173772261198,
+      "loss": 3.4741,
+      "step": 30900
+    },
+    {
+      "epoch": 3.3369272237196768,
+      "grad_norm": 0.5585691928863525,
+      "learning_rate": 0.00040029357798165133,
+      "loss": 3.4571,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "grad_norm": 0.5774768590927124,
+      "learning_rate": 0.00039996977873718294,
+      "loss": 3.4748,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "eval_accuracy": 0.37259172232320303,
+      "eval_loss": 3.483452320098877,
+      "eval_runtime": 183.6331,
+      "eval_samples_per_second": 98.081,
+      "eval_steps_per_second": 6.132,
+      "step": 31000
+    },
+    {
+      "epoch": 3.347708894878706,
+      "grad_norm": 0.6331956386566162,
+      "learning_rate": 0.0003996459794927145,
+      "loss": 3.4485,
+      "step": 31050
+    },
+    {
+      "epoch": 3.353099730458221,
+      "grad_norm": 0.6012781262397766,
+      "learning_rate": 0.00039932218024824604,
+      "loss": 3.4705,
+      "step": 31100
+    },
+    {
+      "epoch": 3.358490566037736,
+      "grad_norm": 0.6027830839157104,
+      "learning_rate": 0.00039899838100377764,
+      "loss": 3.4493,
+      "step": 31150
+    },
+    {
+      "epoch": 3.3638814016172507,
+      "grad_norm": 0.5623087286949158,
+      "learning_rate": 0.0003986745817593092,
+      "loss": 3.4515,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3692722371967654,
+      "grad_norm": 0.6207774877548218,
+      "learning_rate": 0.0003983507825148408,
+      "loss": 3.4675,
+      "step": 31250
+    },
+    {
+      "epoch": 3.37466307277628,
+      "grad_norm": 0.5530596375465393,
+      "learning_rate": 0.00039802698327037235,
+      "loss": 3.4544,
+      "step": 31300
+    },
+    {
+      "epoch": 3.3800539083557952,
+      "grad_norm": 0.600220799446106,
+      "learning_rate": 0.00039770318402590396,
+      "loss": 3.4418,
+      "step": 31350
+    },
+    {
+      "epoch": 3.38544474393531,
+      "grad_norm": 0.5809213519096375,
+      "learning_rate": 0.00039737938478143545,
+      "loss": 3.4483,
+      "step": 31400
+    },
+    {
+      "epoch": 3.3908355795148246,
+      "grad_norm": 0.5642971992492676,
+      "learning_rate": 0.000397055585536967,
+      "loss": 3.4688,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3962264150943398,
+      "grad_norm": 0.6100688576698303,
+      "learning_rate": 0.0003967317862924986,
+      "loss": 3.4687,
+      "step": 31500
+    },
+    {
+      "epoch": 3.4016172506738545,
+      "grad_norm": 0.6187987327575684,
+      "learning_rate": 0.00039641446303291956,
+      "loss": 3.455,
+      "step": 31550
+    },
+    {
+      "epoch": 3.407008086253369,
+      "grad_norm": 0.5841171145439148,
+      "learning_rate": 0.0003960906637884511,
+      "loss": 3.442,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4123989218328843,
+      "grad_norm": 0.6319450736045837,
+      "learning_rate": 0.0003957668645439827,
+      "loss": 3.4622,
+      "step": 31650
+    },
+    {
+      "epoch": 3.417789757412399,
+      "grad_norm": 0.6497536897659302,
+      "learning_rate": 0.00039544306529951426,
+      "loss": 3.4676,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4231805929919137,
+      "grad_norm": 0.567626416683197,
+      "learning_rate": 0.00039511926605504587,
+      "loss": 3.4645,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4285714285714284,
+      "grad_norm": 0.6872853636741638,
+      "learning_rate": 0.0003947954668105774,
+      "loss": 3.4514,
+      "step": 31800
+    },
+    {
+      "epoch": 3.4339622641509435,
+      "grad_norm": 0.5771532654762268,
+      "learning_rate": 0.00039447166756610897,
+      "loss": 3.4551,
+      "step": 31850
+    },
+    {
+      "epoch": 3.439353099730458,
+      "grad_norm": 0.6318898797035217,
+      "learning_rate": 0.0003941478683216406,
+      "loss": 3.4651,
+      "step": 31900
+    },
+    {
+      "epoch": 3.444743935309973,
+      "grad_norm": 0.5701496601104736,
+      "learning_rate": 0.0003938240690771721,
+      "loss": 3.4507,
+      "step": 31950
+    },
+    {
+      "epoch": 3.450134770889488,
+      "grad_norm": 0.6465626955032349,
+      "learning_rate": 0.00039350026983270373,
+      "loss": 3.4633,
+      "step": 32000
+    },
+    {
+      "epoch": 3.450134770889488,
+      "eval_accuracy": 0.37326808654984817,
+      "eval_loss": 3.476710319519043,
+      "eval_runtime": 183.4161,
+      "eval_samples_per_second": 98.197,
+      "eval_steps_per_second": 6.139,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4555256064690028,
+      "grad_norm": 0.5992223024368286,
+      "learning_rate": 0.00039317647058823523,
+      "loss": 3.4851,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4609164420485174,
+      "grad_norm": 0.6115779876708984,
+      "learning_rate": 0.00039285267134376683,
+      "loss": 3.4755,
+      "step": 32100
+    },
+    {
+      "epoch": 3.466307277628032,
+      "grad_norm": 0.5862306952476501,
+      "learning_rate": 0.0003925288720992984,
+      "loss": 3.4469,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4716981132075473,
+      "grad_norm": 0.5586680769920349,
+      "learning_rate": 0.00039220507285482993,
+      "loss": 3.4614,
+      "step": 32200
+    },
+    {
+      "epoch": 3.477088948787062,
+      "grad_norm": 0.5931172370910645,
+      "learning_rate": 0.00039188127361036154,
+      "loss": 3.4505,
+      "step": 32250
+    },
+    {
+      "epoch": 3.4824797843665767,
+      "grad_norm": 0.6547418236732483,
+      "learning_rate": 0.0003915574743658931,
+      "loss": 3.4545,
+      "step": 32300
+    },
+    {
+      "epoch": 3.487870619946092,
+      "grad_norm": 0.5955836176872253,
+      "learning_rate": 0.0003912336751214247,
+      "loss": 3.4467,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4932614555256065,
+      "grad_norm": 0.6000773906707764,
+      "learning_rate": 0.00039090987587695624,
+      "loss": 3.4527,
+      "step": 32400
+    },
+    {
+      "epoch": 3.498652291105121,
+      "grad_norm": 0.6324712038040161,
+      "learning_rate": 0.00039058607663248785,
+      "loss": 3.4604,
+      "step": 32450
+    },
+    {
+      "epoch": 3.5040431266846364,
+      "grad_norm": 0.6126226782798767,
+      "learning_rate": 0.0003902622773880194,
+      "loss": 3.4673,
+      "step": 32500
+    },
+    {
+      "epoch": 3.509433962264151,
+      "grad_norm": 0.5712622404098511,
+      "learning_rate": 0.000389938478143551,
+      "loss": 3.4675,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5148247978436657,
+      "grad_norm": 0.5935483574867249,
+      "learning_rate": 0.00038961467889908255,
+      "loss": 3.4556,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5202156334231804,
+      "grad_norm": 0.5569974184036255,
+      "learning_rate": 0.00038929087965461405,
+      "loss": 3.4606,
+      "step": 32650
+    },
+    {
+      "epoch": 3.525606469002695,
+      "grad_norm": 0.6109350919723511,
+      "learning_rate": 0.00038896708041014566,
+      "loss": 3.4762,
+      "step": 32700
+    },
+    {
+      "epoch": 3.5309973045822103,
+      "grad_norm": 0.6272343397140503,
+      "learning_rate": 0.0003886432811656772,
+      "loss": 3.4494,
+      "step": 32750
+    },
+    {
+      "epoch": 3.536388140161725,
+      "grad_norm": 0.5934761166572571,
+      "learning_rate": 0.0003883194819212088,
+      "loss": 3.4775,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5417789757412397,
+      "grad_norm": 0.5385004281997681,
+      "learning_rate": 0.00038799568267674036,
+      "loss": 3.4781,
+      "step": 32850
+    },
+    {
+      "epoch": 3.547169811320755,
+      "grad_norm": 0.5813574194908142,
+      "learning_rate": 0.00038767188343227197,
+      "loss": 3.4708,
+      "step": 32900
+    },
+    {
+      "epoch": 3.5525606469002695,
+      "grad_norm": 0.614341676235199,
+      "learning_rate": 0.0003873480841878035,
+      "loss": 3.4577,
+      "step": 32950
+    },
+    {
+      "epoch": 3.557951482479784,
+      "grad_norm": 0.6198434829711914,
+      "learning_rate": 0.0003870242849433351,
+      "loss": 3.4541,
+      "step": 33000
+    },
+    {
+      "epoch": 3.557951482479784,
+      "eval_accuracy": 0.3734748529953591,
+      "eval_loss": 3.4712295532226562,
+      "eval_runtime": 183.2674,
+      "eval_samples_per_second": 98.277,
+      "eval_steps_per_second": 6.144,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5633423180592994,
+      "grad_norm": 0.6407135725021362,
+      "learning_rate": 0.00038670048569886667,
+      "loss": 3.4659,
+      "step": 33050
+    },
+    {
+      "epoch": 3.568733153638814,
+      "grad_norm": 0.6006309986114502,
+      "learning_rate": 0.0003863766864543982,
+      "loss": 3.4664,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5741239892183287,
+      "grad_norm": 0.6235687136650085,
+      "learning_rate": 0.00038605288720992983,
+      "loss": 3.4704,
+      "step": 33150
+    },
+    {
+      "epoch": 3.579514824797844,
+      "grad_norm": 0.6336178183555603,
+      "learning_rate": 0.0003857290879654614,
+      "loss": 3.465,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5849056603773586,
+      "grad_norm": 0.571612536907196,
+      "learning_rate": 0.000385405288720993,
+      "loss": 3.4488,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5902964959568733,
+      "grad_norm": 0.6101464033126831,
+      "learning_rate": 0.00038508148947652453,
+      "loss": 3.4767,
+      "step": 33300
+    },
+    {
+      "epoch": 3.595687331536388,
+      "grad_norm": 0.6197885870933533,
+      "learning_rate": 0.00038475769023205614,
+      "loss": 3.4633,
+      "step": 33350
+    },
+    {
+      "epoch": 3.601078167115903,
+      "grad_norm": 0.6503936648368835,
+      "learning_rate": 0.00038443389098758763,
+      "loss": 3.4601,
+      "step": 33400
+    },
+    {
+      "epoch": 3.606469002695418,
+      "grad_norm": 0.6410123705863953,
+      "learning_rate": 0.00038411009174311924,
+      "loss": 3.4757,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6118598382749325,
+      "grad_norm": 0.5860520601272583,
+      "learning_rate": 0.0003837862924986508,
+      "loss": 3.4528,
+      "step": 33500
+    },
+    {
+      "epoch": 3.617250673854447,
+      "grad_norm": 0.6135208606719971,
+      "learning_rate": 0.00038346249325418234,
+      "loss": 3.4515,
+      "step": 33550
+    },
+    {
+      "epoch": 3.6226415094339623,
+      "grad_norm": 0.6425701379776001,
+      "learning_rate": 0.00038313869400971395,
+      "loss": 3.447,
+      "step": 33600
+    },
+    {
+      "epoch": 3.628032345013477,
+      "grad_norm": 0.6353036761283875,
+      "learning_rate": 0.0003828148947652455,
+      "loss": 3.4633,
+      "step": 33650
+    },
+    {
+      "epoch": 3.6334231805929917,
+      "grad_norm": 0.621608555316925,
+      "learning_rate": 0.0003824910955207771,
+      "loss": 3.4674,
+      "step": 33700
+    },
+    {
+      "epoch": 3.638814016172507,
+      "grad_norm": 0.6142898797988892,
+      "learning_rate": 0.00038216729627630865,
+      "loss": 3.4484,
+      "step": 33750
+    },
+    {
+      "epoch": 3.6442048517520216,
+      "grad_norm": 0.6292492151260376,
+      "learning_rate": 0.00038184349703184026,
+      "loss": 3.4527,
+      "step": 33800
+    },
+    {
+      "epoch": 3.6495956873315363,
+      "grad_norm": 0.6098408699035645,
+      "learning_rate": 0.0003815196977873718,
+      "loss": 3.4381,
+      "step": 33850
+    },
+    {
+      "epoch": 3.6549865229110514,
+      "grad_norm": 0.6431405544281006,
+      "learning_rate": 0.0003811958985429034,
+      "loss": 3.4306,
+      "step": 33900
+    },
+    {
+      "epoch": 3.660377358490566,
+      "grad_norm": 0.6275612711906433,
+      "learning_rate": 0.00038087209929843496,
+      "loss": 3.4561,
+      "step": 33950
+    },
+    {
+      "epoch": 3.665768194070081,
+      "grad_norm": 0.6220306754112244,
+      "learning_rate": 0.00038054830005396646,
+      "loss": 3.4513,
+      "step": 34000
+    },
+    {
+      "epoch": 3.665768194070081,
+      "eval_accuracy": 0.37447750184411116,
+      "eval_loss": 3.4639649391174316,
+      "eval_runtime": 183.3787,
+      "eval_samples_per_second": 98.218,
+      "eval_steps_per_second": 6.14,
+      "step": 34000
+    },
+    {
+      "epoch": 3.671159029649596,
+      "grad_norm": 0.6661257147789001,
+      "learning_rate": 0.00038022450080949806,
+      "loss": 3.456,
+      "step": 34050
+    },
+    {
+      "epoch": 3.6765498652291106,
+      "grad_norm": 0.6361551880836487,
+      "learning_rate": 0.0003799007015650296,
+      "loss": 3.4529,
+      "step": 34100
+    },
+    {
+      "epoch": 3.6819407008086253,
+      "grad_norm": 0.5787711143493652,
+      "learning_rate": 0.0003795769023205612,
+      "loss": 3.4494,
+      "step": 34150
+    },
+    {
+      "epoch": 3.68733153638814,
+      "grad_norm": 0.691253125667572,
+      "learning_rate": 0.00037925310307609277,
+      "loss": 3.4608,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6927223719676547,
+      "grad_norm": 0.6856685280799866,
+      "learning_rate": 0.0003789293038316244,
+      "loss": 3.4495,
+      "step": 34250
+    },
+    {
+      "epoch": 3.69811320754717,
+      "grad_norm": 0.6148468255996704,
+      "learning_rate": 0.0003786055045871559,
+      "loss": 3.455,
+      "step": 34300
+    },
+    {
+      "epoch": 3.7035040431266846,
+      "grad_norm": 0.6273311376571655,
+      "learning_rate": 0.0003782817053426875,
+      "loss": 3.4524,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7088948787061993,
+      "grad_norm": 0.6452623009681702,
+      "learning_rate": 0.0003779579060982191,
+      "loss": 3.4595,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7142857142857144,
+      "grad_norm": 0.6160526871681213,
+      "learning_rate": 0.00037763410685375063,
+      "loss": 3.4324,
+      "step": 34450
+    },
+    {
+      "epoch": 3.719676549865229,
+      "grad_norm": 0.6273506879806519,
+      "learning_rate": 0.00037731030760928223,
+      "loss": 3.473,
+      "step": 34500
+    },
+    {
+      "epoch": 3.725067385444744,
+      "grad_norm": 0.6229552626609802,
+      "learning_rate": 0.0003769865083648138,
+      "loss": 3.4495,
+      "step": 34550
+    },
+    {
+      "epoch": 3.730458221024259,
+      "grad_norm": 0.6476499438285828,
+      "learning_rate": 0.0003766627091203454,
+      "loss": 3.455,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7358490566037736,
+      "grad_norm": 0.603552520275116,
+      "learning_rate": 0.00037633890987587694,
+      "loss": 3.4588,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7412398921832883,
+      "grad_norm": 0.611316978931427,
+      "learning_rate": 0.00037601511063140855,
+      "loss": 3.4654,
+      "step": 34700
+    },
+    {
+      "epoch": 3.7466307277628035,
+      "grad_norm": 0.6224241852760315,
+      "learning_rate": 0.00037569131138694004,
+      "loss": 3.4704,
+      "step": 34750
+    },
+    {
+      "epoch": 3.752021563342318,
+      "grad_norm": 0.5943573117256165,
+      "learning_rate": 0.0003753675121424716,
+      "loss": 3.4547,
+      "step": 34800
+    },
+    {
+      "epoch": 3.757412398921833,
+      "grad_norm": 0.6184179782867432,
+      "learning_rate": 0.0003750437128980032,
+      "loss": 3.4624,
+      "step": 34850
+    },
+    {
+      "epoch": 3.7628032345013476,
+      "grad_norm": 0.6081331968307495,
+      "learning_rate": 0.00037471991365353475,
+      "loss": 3.4541,
+      "step": 34900
+    },
+    {
+      "epoch": 3.7681940700808623,
+      "grad_norm": 0.6248754262924194,
+      "learning_rate": 0.00037439611440906635,
+      "loss": 3.4487,
+      "step": 34950
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "grad_norm": 0.6937313079833984,
+      "learning_rate": 0.0003740723151645979,
+      "loss": 3.4707,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "eval_accuracy": 0.3750037077797967,
+      "eval_loss": 3.4571778774261475,
+      "eval_runtime": 183.6889,
+      "eval_samples_per_second": 98.052,
+      "eval_steps_per_second": 6.13,
+      "step": 35000
+    },
+    {
+      "epoch": 3.778975741239892,
+      "grad_norm": 0.6209198236465454,
+      "learning_rate": 0.0003737485159201295,
+      "loss": 3.4533,
+      "step": 35050
+    },
+    {
+      "epoch": 3.784366576819407,
+      "grad_norm": 0.6297261118888855,
+      "learning_rate": 0.00037342471667566106,
+      "loss": 3.4573,
+      "step": 35100
+    },
+    {
+      "epoch": 3.789757412398922,
+      "grad_norm": 0.6521171927452087,
+      "learning_rate": 0.00037310091743119266,
+      "loss": 3.4427,
+      "step": 35150
+    },
+    {
+      "epoch": 3.7951482479784366,
+      "grad_norm": 0.6195915341377258,
+      "learning_rate": 0.0003727771181867242,
+      "loss": 3.4488,
+      "step": 35200
+    },
+    {
+      "epoch": 3.8005390835579513,
+      "grad_norm": 0.6207437515258789,
+      "learning_rate": 0.00037245331894225576,
+      "loss": 3.455,
+      "step": 35250
+    },
+    {
+      "epoch": 3.8059299191374665,
+      "grad_norm": 0.6468321681022644,
+      "learning_rate": 0.00037212951969778737,
+      "loss": 3.4479,
+      "step": 35300
+    },
+    {
+      "epoch": 3.811320754716981,
+      "grad_norm": 0.6130130290985107,
+      "learning_rate": 0.00037180572045331887,
+      "loss": 3.469,
+      "step": 35350
+    },
+    {
+      "epoch": 3.816711590296496,
+      "grad_norm": 0.5929135084152222,
+      "learning_rate": 0.00037148192120885047,
+      "loss": 3.4505,
+      "step": 35400
+    },
+    {
+      "epoch": 3.822102425876011,
+      "grad_norm": 0.5916996002197266,
+      "learning_rate": 0.000371158121964382,
+      "loss": 3.4559,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8274932614555257,
+      "grad_norm": 0.6095645427703857,
+      "learning_rate": 0.0003708343227199136,
+      "loss": 3.4465,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8328840970350404,
+      "grad_norm": 0.561920702457428,
+      "learning_rate": 0.0003705105234754452,
+      "loss": 3.4501,
+      "step": 35550
+    },
+    {
+      "epoch": 3.838274932614555,
+      "grad_norm": 0.602867066860199,
+      "learning_rate": 0.0003701867242309768,
+      "loss": 3.4602,
+      "step": 35600
+    },
+    {
+      "epoch": 3.8436657681940702,
+      "grad_norm": 0.6007760763168335,
+      "learning_rate": 0.00036986292498650833,
+      "loss": 3.4449,
+      "step": 35650
+    },
+    {
+      "epoch": 3.849056603773585,
+      "grad_norm": 0.600691020488739,
+      "learning_rate": 0.0003695456017269293,
+      "loss": 3.4464,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8544474393530996,
+      "grad_norm": 0.6398903727531433,
+      "learning_rate": 0.00036922180248246083,
+      "loss": 3.4557,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8598382749326143,
+      "grad_norm": 0.6188865303993225,
+      "learning_rate": 0.00036889800323799244,
+      "loss": 3.4427,
+      "step": 35800
+    },
+    {
+      "epoch": 3.8652291105121295,
+      "grad_norm": 0.621609091758728,
+      "learning_rate": 0.000368574203993524,
+      "loss": 3.4525,
+      "step": 35850
+    },
+    {
+      "epoch": 3.870619946091644,
+      "grad_norm": 0.5898281931877136,
+      "learning_rate": 0.0003682504047490556,
+      "loss": 3.466,
+      "step": 35900
+    },
+    {
+      "epoch": 3.876010781671159,
+      "grad_norm": 0.6150808930397034,
+      "learning_rate": 0.00036792660550458714,
+      "loss": 3.4687,
+      "step": 35950
+    },
+    {
+      "epoch": 3.881401617250674,
+      "grad_norm": 0.6602946519851685,
+      "learning_rate": 0.00036760280626011864,
+      "loss": 3.4347,
+      "step": 36000
+    },
+    {
+      "epoch": 3.881401617250674,
+      "eval_accuracy": 0.3755471895246447,
+      "eval_loss": 3.4529879093170166,
+      "eval_runtime": 183.2643,
+      "eval_samples_per_second": 98.279,
+      "eval_steps_per_second": 6.144,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8867924528301887,
+      "grad_norm": 0.6273593306541443,
+      "learning_rate": 0.0003672790070156503,
+      "loss": 3.4721,
+      "step": 36050
+    },
+    {
+      "epoch": 3.8921832884097034,
+      "grad_norm": 0.5950552225112915,
+      "learning_rate": 0.0003669552077711818,
+      "loss": 3.4363,
+      "step": 36100
+    },
+    {
+      "epoch": 3.8975741239892185,
+      "grad_norm": 0.6331326961517334,
+      "learning_rate": 0.0003666314085267134,
+      "loss": 3.4437,
+      "step": 36150
+    },
+    {
+      "epoch": 3.9029649595687332,
+      "grad_norm": 0.624070405960083,
+      "learning_rate": 0.00036630760928224495,
+      "loss": 3.4499,
+      "step": 36200
+    },
+    {
+      "epoch": 3.908355795148248,
+      "grad_norm": 0.5709901452064514,
+      "learning_rate": 0.00036598381003777656,
+      "loss": 3.4563,
+      "step": 36250
+    },
+    {
+      "epoch": 3.913746630727763,
+      "grad_norm": 0.5700583457946777,
+      "learning_rate": 0.0003656600107933081,
+      "loss": 3.4594,
+      "step": 36300
+    },
+    {
+      "epoch": 3.9191374663072778,
+      "grad_norm": 0.607256293296814,
+      "learning_rate": 0.0003653362115488397,
+      "loss": 3.4724,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9245283018867925,
+      "grad_norm": 0.5817025303840637,
+      "learning_rate": 0.00036501241230437126,
+      "loss": 3.4393,
+      "step": 36400
+    },
+    {
+      "epoch": 3.929919137466307,
+      "grad_norm": 0.6234590411186218,
+      "learning_rate": 0.0003646886130599028,
+      "loss": 3.445,
+      "step": 36450
+    },
+    {
+      "epoch": 3.935309973045822,
+      "grad_norm": 0.5800814032554626,
+      "learning_rate": 0.0003643648138154344,
+      "loss": 3.4585,
+      "step": 36500
+    },
+    {
+      "epoch": 3.940700808625337,
+      "grad_norm": 0.5907812118530273,
+      "learning_rate": 0.00036404101457096597,
+      "loss": 3.4736,
+      "step": 36550
+    },
+    {
+      "epoch": 3.9460916442048517,
+      "grad_norm": 0.6417782306671143,
+      "learning_rate": 0.00036371721532649757,
+      "loss": 3.4261,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9514824797843664,
+      "grad_norm": 0.6414207816123962,
+      "learning_rate": 0.0003633934160820291,
+      "loss": 3.4434,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9568733153638815,
+      "grad_norm": 0.6781249642372131,
+      "learning_rate": 0.00036306961683756073,
+      "loss": 3.4484,
+      "step": 36700
+    },
+    {
+      "epoch": 3.9622641509433962,
+      "grad_norm": 0.6436617374420166,
+      "learning_rate": 0.0003627458175930922,
+      "loss": 3.4289,
+      "step": 36750
+    },
+    {
+      "epoch": 3.967654986522911,
+      "grad_norm": 0.5609354972839355,
+      "learning_rate": 0.0003624220183486238,
+      "loss": 3.4438,
+      "step": 36800
+    },
+    {
+      "epoch": 3.973045822102426,
+      "grad_norm": 0.5995511412620544,
+      "learning_rate": 0.0003620982191041554,
+      "loss": 3.4516,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9784366576819408,
+      "grad_norm": 0.6280034184455872,
+      "learning_rate": 0.00036177441985968693,
+      "loss": 3.45,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9838274932614555,
+      "grad_norm": 0.6208735704421997,
+      "learning_rate": 0.00036145062061521854,
+      "loss": 3.4302,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "grad_norm": 0.6470046639442444,
+      "learning_rate": 0.0003611268213707501,
+      "loss": 3.417,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "eval_accuracy": 0.3759129151450282,
+      "eval_loss": 3.4461848735809326,
+      "eval_runtime": 183.7064,
+      "eval_samples_per_second": 98.042,
+      "eval_steps_per_second": 6.129,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9946091644204853,
+      "grad_norm": 0.6202898621559143,
+      "learning_rate": 0.0003608030221262817,
+      "loss": 3.4704,
+      "step": 37050
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.2718347311019897,
+      "learning_rate": 0.00036047922288181324,
+      "loss": 3.4332,
+      "step": 37100
+    },
+    {
+      "epoch": 4.005390835579515,
+      "grad_norm": 0.6157583594322205,
+      "learning_rate": 0.00036015542363734485,
+      "loss": 3.3631,
+      "step": 37150
+    },
+    {
+      "epoch": 4.010781671159029,
+      "grad_norm": 0.607601523399353,
+      "learning_rate": 0.0003598316243928764,
+      "loss": 3.3591,
+      "step": 37200
+    },
+    {
+      "epoch": 4.0161725067385445,
+      "grad_norm": 0.6501696109771729,
+      "learning_rate": 0.00035950782514840795,
+      "loss": 3.355,
+      "step": 37250
+    },
+    {
+      "epoch": 4.02156334231806,
+      "grad_norm": 0.6216338872909546,
+      "learning_rate": 0.00035918402590393955,
+      "loss": 3.3606,
+      "step": 37300
+    },
+    {
+      "epoch": 4.026954177897574,
+      "grad_norm": 0.6452534198760986,
+      "learning_rate": 0.00035886022665947105,
+      "loss": 3.3414,
+      "step": 37350
+    },
+    {
+      "epoch": 4.032345013477089,
+      "grad_norm": 0.5642078518867493,
+      "learning_rate": 0.0003585364274150027,
+      "loss": 3.346,
+      "step": 37400
+    },
+    {
+      "epoch": 4.037735849056604,
+      "grad_norm": 0.6294705271720886,
+      "learning_rate": 0.0003582126281705342,
+      "loss": 3.3541,
+      "step": 37450
+    },
+    {
+      "epoch": 4.0431266846361185,
+      "grad_norm": 0.7128744721412659,
+      "learning_rate": 0.0003578888289260658,
+      "loss": 3.3548,
+      "step": 37500
+    },
+    {
+      "epoch": 4.048517520215634,
+      "grad_norm": 0.6176431179046631,
+      "learning_rate": 0.00035756502968159736,
+      "loss": 3.3643,
+      "step": 37550
+    },
+    {
+      "epoch": 4.053908355795148,
+      "grad_norm": 0.6452947854995728,
+      "learning_rate": 0.00035724123043712896,
+      "loss": 3.3591,
+      "step": 37600
+    },
+    {
+      "epoch": 4.059299191374663,
+      "grad_norm": 0.6863994598388672,
+      "learning_rate": 0.0003569174311926605,
+      "loss": 3.3658,
+      "step": 37650
+    },
+    {
+      "epoch": 4.064690026954178,
+      "grad_norm": 0.5968023538589478,
+      "learning_rate": 0.00035659363194819206,
+      "loss": 3.3681,
+      "step": 37700
+    },
+    {
+      "epoch": 4.070080862533692,
+      "grad_norm": 0.5858391523361206,
+      "learning_rate": 0.00035626983270372367,
+      "loss": 3.3728,
+      "step": 37750
+    },
+    {
+      "epoch": 4.0754716981132075,
+      "grad_norm": 0.6459882259368896,
+      "learning_rate": 0.0003559525094441446,
+      "loss": 3.3647,
+      "step": 37800
+    },
+    {
+      "epoch": 4.080862533692723,
+      "grad_norm": 0.6718738079071045,
+      "learning_rate": 0.00035562871019967617,
+      "loss": 3.3699,
+      "step": 37850
+    },
+    {
+      "epoch": 4.086253369272237,
+      "grad_norm": 0.6096557378768921,
+      "learning_rate": 0.0003553049109552078,
+      "loss": 3.3794,
+      "step": 37900
+    },
+    {
+      "epoch": 4.091644204851752,
+      "grad_norm": 0.6182774901390076,
+      "learning_rate": 0.00035498111171073933,
+      "loss": 3.3785,
+      "step": 37950
+    },
+    {
+      "epoch": 4.097035040431267,
+      "grad_norm": 0.6097174286842346,
+      "learning_rate": 0.0003546573124662708,
+      "loss": 3.3602,
+      "step": 38000
+    },
+    {
+      "epoch": 4.097035040431267,
+      "eval_accuracy": 0.3763951166611489,
+      "eval_loss": 3.448361396789551,
+      "eval_runtime": 183.0845,
+      "eval_samples_per_second": 98.375,
+      "eval_steps_per_second": 6.15,
+      "step": 38000
+    },
+    {
+      "epoch": 4.1024258760107815,
+      "grad_norm": 0.6500292420387268,
+      "learning_rate": 0.0003543335132218025,
+      "loss": 3.3779,
+      "step": 38050
+    },
+    {
+      "epoch": 4.107816711590297,
+      "grad_norm": 0.6115676164627075,
+      "learning_rate": 0.000354009713977334,
+      "loss": 3.3628,
+      "step": 38100
+    },
+    {
+      "epoch": 4.113207547169812,
+      "grad_norm": 0.6318033337593079,
+      "learning_rate": 0.0003536859147328656,
+      "loss": 3.377,
+      "step": 38150
+    },
+    {
+      "epoch": 4.118598382749326,
+      "grad_norm": 0.6544750332832336,
+      "learning_rate": 0.00035336211548839713,
+      "loss": 3.381,
+      "step": 38200
+    },
+    {
+      "epoch": 4.123989218328841,
+      "grad_norm": 0.8020888566970825,
+      "learning_rate": 0.00035303831624392874,
+      "loss": 3.3684,
+      "step": 38250
+    },
+    {
+      "epoch": 4.129380053908355,
+      "grad_norm": 0.6313894391059875,
+      "learning_rate": 0.0003527145169994603,
+      "loss": 3.3717,
+      "step": 38300
+    },
+    {
+      "epoch": 4.1347708894878705,
+      "grad_norm": 0.6045191287994385,
+      "learning_rate": 0.0003523907177549919,
+      "loss": 3.3555,
+      "step": 38350
+    },
+    {
+      "epoch": 4.140161725067386,
+      "grad_norm": 0.6183720231056213,
+      "learning_rate": 0.00035206691851052345,
+      "loss": 3.3793,
+      "step": 38400
+    },
+    {
+      "epoch": 4.1455525606469,
+      "grad_norm": 0.62325519323349,
+      "learning_rate": 0.000351743119266055,
+      "loss": 3.3754,
+      "step": 38450
+    },
+    {
+      "epoch": 4.150943396226415,
+      "grad_norm": 0.6222301125526428,
+      "learning_rate": 0.0003514193200215866,
+      "loss": 3.3679,
+      "step": 38500
+    },
+    {
+      "epoch": 4.15633423180593,
+      "grad_norm": 0.6233283877372742,
+      "learning_rate": 0.00035109552077711815,
+      "loss": 3.3677,
+      "step": 38550
+    },
+    {
+      "epoch": 4.1617250673854445,
+      "grad_norm": 0.6315566301345825,
+      "learning_rate": 0.00035077172153264976,
+      "loss": 3.3666,
+      "step": 38600
+    },
+    {
+      "epoch": 4.16711590296496,
+      "grad_norm": 0.6729894876480103,
+      "learning_rate": 0.0003504479222881813,
+      "loss": 3.3734,
+      "step": 38650
+    },
+    {
+      "epoch": 4.172506738544475,
+      "grad_norm": 0.62404465675354,
+      "learning_rate": 0.0003501241230437129,
+      "loss": 3.3882,
+      "step": 38700
+    },
+    {
+      "epoch": 4.177897574123989,
+      "grad_norm": 0.6314374208450317,
+      "learning_rate": 0.0003498003237992444,
+      "loss": 3.3631,
+      "step": 38750
+    },
+    {
+      "epoch": 4.183288409703504,
+      "grad_norm": 0.5996955037117004,
+      "learning_rate": 0.000349476524554776,
+      "loss": 3.379,
+      "step": 38800
+    },
+    {
+      "epoch": 4.188679245283019,
+      "grad_norm": 0.6228339076042175,
+      "learning_rate": 0.00034915272531030756,
+      "loss": 3.3869,
+      "step": 38850
+    },
+    {
+      "epoch": 4.1940700808625335,
+      "grad_norm": 0.6187623143196106,
+      "learning_rate": 0.0003488289260658391,
+      "loss": 3.4018,
+      "step": 38900
+    },
+    {
+      "epoch": 4.199460916442049,
+      "grad_norm": 0.68119215965271,
+      "learning_rate": 0.0003485051268213707,
+      "loss": 3.3689,
+      "step": 38950
+    },
+    {
+      "epoch": 4.204851752021563,
+      "grad_norm": 0.6336060762405396,
+      "learning_rate": 0.00034818132757690227,
+      "loss": 3.384,
+      "step": 39000
+    },
+    {
+      "epoch": 4.204851752021563,
+      "eval_accuracy": 0.37696337126441865,
+      "eval_loss": 3.4456894397735596,
+      "eval_runtime": 183.3606,
+      "eval_samples_per_second": 98.227,
+      "eval_steps_per_second": 6.141,
+      "step": 39000
+    },
+    {
+      "epoch": 4.210242587601078,
+      "grad_norm": 0.6565260887145996,
+      "learning_rate": 0.0003478575283324339,
+      "loss": 3.3906,
+      "step": 39050
+    },
+    {
+      "epoch": 4.215633423180593,
+      "grad_norm": 0.6345798969268799,
+      "learning_rate": 0.0003475337290879654,
+      "loss": 3.3934,
+      "step": 39100
+    },
+    {
+      "epoch": 4.2210242587601075,
+      "grad_norm": 0.628085732460022,
+      "learning_rate": 0.00034720992984349703,
+      "loss": 3.3796,
+      "step": 39150
+    },
+    {
+      "epoch": 4.226415094339623,
+      "grad_norm": 0.698582649230957,
+      "learning_rate": 0.0003468861305990286,
+      "loss": 3.371,
+      "step": 39200
+    },
+    {
+      "epoch": 4.231805929919138,
+      "grad_norm": 0.6223801970481873,
+      "learning_rate": 0.0003465623313545602,
+      "loss": 3.3729,
+      "step": 39250
+    },
+    {
+      "epoch": 4.237196765498652,
+      "grad_norm": 0.6224102973937988,
+      "learning_rate": 0.00034623853211009173,
+      "loss": 3.3776,
+      "step": 39300
+    },
+    {
+      "epoch": 4.242587601078167,
+      "grad_norm": 0.621496856212616,
+      "learning_rate": 0.00034591473286562323,
+      "loss": 3.3878,
+      "step": 39350
+    },
+    {
+      "epoch": 4.247978436657682,
+      "grad_norm": 0.6796809434890747,
+      "learning_rate": 0.0003455909336211549,
+      "loss": 3.3817,
+      "step": 39400
+    },
+    {
+      "epoch": 4.2533692722371965,
+      "grad_norm": 0.6903562545776367,
+      "learning_rate": 0.0003452671343766864,
+      "loss": 3.3942,
+      "step": 39450
+    },
+    {
+      "epoch": 4.258760107816712,
+      "grad_norm": 0.6326731443405151,
+      "learning_rate": 0.000344943335132218,
+      "loss": 3.3807,
+      "step": 39500
+    },
+    {
+      "epoch": 4.264150943396227,
+      "grad_norm": 0.6289458274841309,
+      "learning_rate": 0.00034461953588774954,
+      "loss": 3.3671,
+      "step": 39550
+    },
+    {
+      "epoch": 4.269541778975741,
+      "grad_norm": 0.6308515071868896,
+      "learning_rate": 0.00034429573664328115,
+      "loss": 3.3838,
+      "step": 39600
+    },
+    {
+      "epoch": 4.274932614555256,
+      "grad_norm": 0.5939794182777405,
+      "learning_rate": 0.0003439719373988127,
+      "loss": 3.3924,
+      "step": 39650
+    },
+    {
+      "epoch": 4.280323450134771,
+      "grad_norm": 0.6445115208625793,
+      "learning_rate": 0.00034364813815434425,
+      "loss": 3.3743,
+      "step": 39700
+    },
+    {
+      "epoch": 4.285714285714286,
+      "grad_norm": 0.6573401093482971,
+      "learning_rate": 0.00034332433890987585,
+      "loss": 3.3976,
+      "step": 39750
+    },
+    {
+      "epoch": 4.291105121293801,
+      "grad_norm": 0.637039840221405,
+      "learning_rate": 0.0003430005396654074,
+      "loss": 3.3844,
+      "step": 39800
+    },
+    {
+      "epoch": 4.296495956873315,
+      "grad_norm": 0.6390613913536072,
+      "learning_rate": 0.000342676740420939,
+      "loss": 3.385,
+      "step": 39850
+    },
+    {
+      "epoch": 4.30188679245283,
+      "grad_norm": 0.6816765666007996,
+      "learning_rate": 0.00034235294117647056,
+      "loss": 3.3858,
+      "step": 39900
+    },
+    {
+      "epoch": 4.307277628032345,
+      "grad_norm": 0.7338299751281738,
+      "learning_rate": 0.00034202914193200216,
+      "loss": 3.3951,
+      "step": 39950
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "grad_norm": 0.6921063661575317,
+      "learning_rate": 0.0003417118186724231,
+      "loss": 3.4014,
+      "step": 40000
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "eval_accuracy": 0.3773256199923921,
+      "eval_loss": 3.4401845932006836,
+      "eval_runtime": 183.335,
+      "eval_samples_per_second": 98.241,
+      "eval_steps_per_second": 6.142,
+      "step": 40000
+    },
+    {
+      "epoch": 4.318059299191375,
+      "grad_norm": 0.6596004366874695,
+      "learning_rate": 0.00034138801942795467,
+      "loss": 3.388,
+      "step": 40050
+    },
+    {
+      "epoch": 4.32345013477089,
+      "grad_norm": 0.6345239281654358,
+      "learning_rate": 0.00034106422018348616,
+      "loss": 3.3935,
+      "step": 40100
+    },
+    {
+      "epoch": 4.328840970350404,
+      "grad_norm": 0.630919873714447,
+      "learning_rate": 0.00034074042093901777,
+      "loss": 3.3751,
+      "step": 40150
+    },
+    {
+      "epoch": 4.334231805929919,
+      "grad_norm": 0.6389809846878052,
+      "learning_rate": 0.0003404166216945493,
+      "loss": 3.4071,
+      "step": 40200
+    },
+    {
+      "epoch": 4.339622641509434,
+      "grad_norm": 0.6863662004470825,
+      "learning_rate": 0.0003400928224500809,
+      "loss": 3.4102,
+      "step": 40250
+    },
+    {
+      "epoch": 4.345013477088949,
+      "grad_norm": 0.6372359991073608,
+      "learning_rate": 0.00033976902320561247,
+      "loss": 3.3826,
+      "step": 40300
+    },
+    {
+      "epoch": 4.350404312668464,
+      "grad_norm": 0.6467205882072449,
+      "learning_rate": 0.0003394452239611441,
+      "loss": 3.392,
+      "step": 40350
+    },
+    {
+      "epoch": 4.355795148247978,
+      "grad_norm": 0.6186798214912415,
+      "learning_rate": 0.00033912142471667563,
+      "loss": 3.3892,
+      "step": 40400
+    },
+    {
+      "epoch": 4.361185983827493,
+      "grad_norm": 0.6394029259681702,
+      "learning_rate": 0.00033879762547220723,
+      "loss": 3.3908,
+      "step": 40450
+    },
+    {
+      "epoch": 4.366576819407008,
+      "grad_norm": 0.6361256837844849,
+      "learning_rate": 0.0003384738262277388,
+      "loss": 3.3858,
+      "step": 40500
+    },
+    {
+      "epoch": 4.3719676549865225,
+      "grad_norm": 0.6550248265266418,
+      "learning_rate": 0.00033815002698327033,
+      "loss": 3.3896,
+      "step": 40550
+    },
+    {
+      "epoch": 4.377358490566038,
+      "grad_norm": 0.6077722311019897,
+      "learning_rate": 0.00033782622773880194,
+      "loss": 3.3989,
+      "step": 40600
+    },
+    {
+      "epoch": 4.382749326145553,
+      "grad_norm": 0.6376055479049683,
+      "learning_rate": 0.0003375024284943335,
+      "loss": 3.3889,
+      "step": 40650
+    },
+    {
+      "epoch": 4.388140161725067,
+      "grad_norm": 0.6313409209251404,
+      "learning_rate": 0.0003371786292498651,
+      "loss": 3.4015,
+      "step": 40700
+    },
+    {
+      "epoch": 4.393530997304582,
+      "grad_norm": 0.6659762263298035,
+      "learning_rate": 0.0003368548300053966,
+      "loss": 3.403,
+      "step": 40750
+    },
+    {
+      "epoch": 4.398921832884097,
+      "grad_norm": 0.6614190340042114,
+      "learning_rate": 0.0003365310307609282,
+      "loss": 3.3948,
+      "step": 40800
+    },
+    {
+      "epoch": 4.404312668463612,
+      "grad_norm": 0.6322287321090698,
+      "learning_rate": 0.00033620723151645975,
+      "loss": 3.383,
+      "step": 40850
+    },
+    {
+      "epoch": 4.409703504043127,
+      "grad_norm": 0.6170388460159302,
+      "learning_rate": 0.0003358834322719913,
+      "loss": 3.383,
+      "step": 40900
+    },
+    {
+      "epoch": 4.415094339622642,
+      "grad_norm": 0.896038830280304,
+      "learning_rate": 0.0003355596330275229,
+      "loss": 3.38,
+      "step": 40950
+    },
+    {
+      "epoch": 4.420485175202156,
+      "grad_norm": 0.6698582768440247,
+      "learning_rate": 0.00033523583378305445,
+      "loss": 3.4152,
+      "step": 41000
+    },
+    {
+      "epoch": 4.420485175202156,
+      "eval_accuracy": 0.37785747587824403,
+      "eval_loss": 3.4369399547576904,
+      "eval_runtime": 183.3239,
+      "eval_samples_per_second": 98.247,
+      "eval_steps_per_second": 6.142,
+      "step": 41000
+    },
+    {
+      "epoch": 4.425876010781671,
+      "grad_norm": 0.6632938981056213,
+      "learning_rate": 0.00033491203453858606,
+      "loss": 3.3909,
+      "step": 41050
+    },
+    {
+      "epoch": 4.431266846361186,
+      "grad_norm": 0.6375352144241333,
+      "learning_rate": 0.0003345882352941176,
+      "loss": 3.3986,
+      "step": 41100
+    },
+    {
+      "epoch": 4.436657681940701,
+      "grad_norm": 0.6697904467582703,
+      "learning_rate": 0.0003342644360496492,
+      "loss": 3.4054,
+      "step": 41150
+    },
+    {
+      "epoch": 4.442048517520216,
+      "grad_norm": 0.6245919466018677,
+      "learning_rate": 0.00033394063680518076,
+      "loss": 3.3906,
+      "step": 41200
+    },
+    {
+      "epoch": 4.44743935309973,
+      "grad_norm": 0.662598729133606,
+      "learning_rate": 0.00033361683756071237,
+      "loss": 3.4072,
+      "step": 41250
+    },
+    {
+      "epoch": 4.452830188679245,
+      "grad_norm": 0.6050485372543335,
+      "learning_rate": 0.0003332930383162439,
+      "loss": 3.3794,
+      "step": 41300
+    },
+    {
+      "epoch": 4.45822102425876,
+      "grad_norm": 0.6476341485977173,
+      "learning_rate": 0.0003329692390717754,
+      "loss": 3.385,
+      "step": 41350
+    },
+    {
+      "epoch": 4.463611859838275,
+      "grad_norm": 0.5851879715919495,
+      "learning_rate": 0.0003326454398273071,
+      "loss": 3.4083,
+      "step": 41400
+    },
+    {
+      "epoch": 4.46900269541779,
+      "grad_norm": 0.6802077293395996,
+      "learning_rate": 0.00033232164058283857,
+      "loss": 3.3837,
+      "step": 41450
+    },
+    {
+      "epoch": 4.474393530997305,
+      "grad_norm": 0.8038954734802246,
+      "learning_rate": 0.0003319978413383702,
+      "loss": 3.3939,
+      "step": 41500
+    },
+    {
+      "epoch": 4.479784366576819,
+      "grad_norm": 0.6305577158927917,
+      "learning_rate": 0.0003316740420939017,
+      "loss": 3.3885,
+      "step": 41550
+    },
+    {
+      "epoch": 4.485175202156334,
+      "grad_norm": 0.6377895474433899,
+      "learning_rate": 0.00033135024284943333,
+      "loss": 3.3988,
+      "step": 41600
+    },
+    {
+      "epoch": 4.490566037735849,
+      "grad_norm": 0.5902746319770813,
+      "learning_rate": 0.0003310264436049649,
+      "loss": 3.3829,
+      "step": 41650
+    },
+    {
+      "epoch": 4.495956873315364,
+      "grad_norm": 0.5873847603797913,
+      "learning_rate": 0.0003307026443604965,
+      "loss": 3.4045,
+      "step": 41700
+    },
+    {
+      "epoch": 4.501347708894879,
+      "grad_norm": 0.6426920890808105,
+      "learning_rate": 0.00033037884511602804,
+      "loss": 3.389,
+      "step": 41750
+    },
+    {
+      "epoch": 4.506738544474393,
+      "grad_norm": 0.7010065913200378,
+      "learning_rate": 0.0003300550458715596,
+      "loss": 3.3945,
+      "step": 41800
+    },
+    {
+      "epoch": 4.512129380053908,
+      "grad_norm": 0.6423835158348083,
+      "learning_rate": 0.0003297312466270912,
+      "loss": 3.383,
+      "step": 41850
+    },
+    {
+      "epoch": 4.517520215633423,
+      "grad_norm": 0.6504900455474854,
+      "learning_rate": 0.00032940744738262274,
+      "loss": 3.3921,
+      "step": 41900
+    },
+    {
+      "epoch": 4.5229110512129385,
+      "grad_norm": 0.6463404893875122,
+      "learning_rate": 0.00032908364813815435,
+      "loss": 3.3769,
+      "step": 41950
+    },
+    {
+      "epoch": 4.528301886792453,
+      "grad_norm": 0.6280781030654907,
+      "learning_rate": 0.0003287598488936859,
+      "loss": 3.3867,
+      "step": 42000
+    },
+    {
+      "epoch": 4.528301886792453,
+      "eval_accuracy": 0.37868834451136135,
+      "eval_loss": 3.430631399154663,
+      "eval_runtime": 183.346,
+      "eval_samples_per_second": 98.235,
+      "eval_steps_per_second": 6.141,
+      "step": 42000
+    },
+    {
+      "epoch": 4.533692722371968,
+      "grad_norm": 0.6555848717689514,
+      "learning_rate": 0.0003284360496492175,
+      "loss": 3.3855,
+      "step": 42050
+    },
+    {
+      "epoch": 4.539083557951482,
+      "grad_norm": 0.6349523067474365,
+      "learning_rate": 0.00032811872638963834,
+      "loss": 3.3933,
+      "step": 42100
+    },
+    {
+      "epoch": 4.544474393530997,
+      "grad_norm": 0.6653229594230652,
+      "learning_rate": 0.00032779492714516995,
+      "loss": 3.4217,
+      "step": 42150
+    },
+    {
+      "epoch": 4.549865229110512,
+      "grad_norm": 0.622232973575592,
+      "learning_rate": 0.0003274711279007015,
+      "loss": 3.3876,
+      "step": 42200
+    },
+    {
+      "epoch": 4.555256064690027,
+      "grad_norm": 0.6820586323738098,
+      "learning_rate": 0.0003271473286562331,
+      "loss": 3.3862,
+      "step": 42250
+    },
+    {
+      "epoch": 4.560646900269542,
+      "grad_norm": 0.6591930985450745,
+      "learning_rate": 0.00032682352941176466,
+      "loss": 3.39,
+      "step": 42300
+    },
+    {
+      "epoch": 4.566037735849057,
+      "grad_norm": 0.6004623174667358,
+      "learning_rate": 0.00032649973016729626,
+      "loss": 3.3902,
+      "step": 42350
+    },
+    {
+      "epoch": 4.571428571428571,
+      "grad_norm": 0.6339703798294067,
+      "learning_rate": 0.0003261759309228278,
+      "loss": 3.3813,
+      "step": 42400
+    },
+    {
+      "epoch": 4.576819407008086,
+      "grad_norm": 0.6326310038566589,
+      "learning_rate": 0.0003258521316783594,
+      "loss": 3.3934,
+      "step": 42450
+    },
+    {
+      "epoch": 4.5822102425876015,
+      "grad_norm": 0.6333923935890198,
+      "learning_rate": 0.00032552833243389097,
+      "loss": 3.3914,
+      "step": 42500
+    },
+    {
+      "epoch": 4.587601078167116,
+      "grad_norm": 0.6541619300842285,
+      "learning_rate": 0.0003252045331894225,
+      "loss": 3.3918,
+      "step": 42550
+    },
+    {
+      "epoch": 4.592991913746631,
+      "grad_norm": 0.6211109161376953,
+      "learning_rate": 0.0003248807339449541,
+      "loss": 3.3753,
+      "step": 42600
+    },
+    {
+      "epoch": 4.598382749326145,
+      "grad_norm": 0.7737997174263,
+      "learning_rate": 0.00032455693470048567,
+      "loss": 3.3756,
+      "step": 42650
+    },
+    {
+      "epoch": 4.60377358490566,
+      "grad_norm": 0.6622357368469238,
+      "learning_rate": 0.0003242331354560173,
+      "loss": 3.3862,
+      "step": 42700
+    },
+    {
+      "epoch": 4.609164420485175,
+      "grad_norm": 0.6532875299453735,
+      "learning_rate": 0.0003239093362115488,
+      "loss": 3.4049,
+      "step": 42750
+    },
+    {
+      "epoch": 4.6145552560646905,
+      "grad_norm": 0.6364518404006958,
+      "learning_rate": 0.00032358553696708043,
+      "loss": 3.3759,
+      "step": 42800
+    },
+    {
+      "epoch": 4.619946091644205,
+      "grad_norm": 0.6832470893859863,
+      "learning_rate": 0.00032326173772261193,
+      "loss": 3.3814,
+      "step": 42850
+    },
+    {
+      "epoch": 4.62533692722372,
+      "grad_norm": 0.677706241607666,
+      "learning_rate": 0.00032293793847814353,
+      "loss": 3.3801,
+      "step": 42900
+    },
+    {
+      "epoch": 4.630727762803234,
+      "grad_norm": 0.6813549995422363,
+      "learning_rate": 0.0003226141392336751,
+      "loss": 3.403,
+      "step": 42950
+    },
+    {
+      "epoch": 4.636118598382749,
+      "grad_norm": 0.7086654901504517,
+      "learning_rate": 0.00032229033998920663,
+      "loss": 3.4083,
+      "step": 43000
+    },
+    {
+      "epoch": 4.636118598382749,
+      "eval_accuracy": 0.37909992165040257,
+      "eval_loss": 3.4262819290161133,
+      "eval_runtime": 183.3323,
+      "eval_samples_per_second": 98.242,
+      "eval_steps_per_second": 6.142,
+      "step": 43000
+    },
+    {
+      "epoch": 4.6415094339622645,
+      "grad_norm": 0.6429668664932251,
+      "learning_rate": 0.00032196654074473824,
+      "loss": 3.3947,
+      "step": 43050
+    },
+    {
+      "epoch": 4.646900269541779,
+      "grad_norm": 0.610312283039093,
+      "learning_rate": 0.0003216427415002698,
+      "loss": 3.3979,
+      "step": 43100
+    },
+    {
+      "epoch": 4.652291105121294,
+      "grad_norm": 0.6420636177062988,
+      "learning_rate": 0.0003213189422558014,
+      "loss": 3.4103,
+      "step": 43150
+    },
+    {
+      "epoch": 4.657681940700809,
+      "grad_norm": 0.6642614603042603,
+      "learning_rate": 0.00032099514301133295,
+      "loss": 3.383,
+      "step": 43200
+    },
+    {
+      "epoch": 4.663072776280323,
+      "grad_norm": 0.622520923614502,
+      "learning_rate": 0.00032067134376686455,
+      "loss": 3.3825,
+      "step": 43250
+    },
+    {
+      "epoch": 4.668463611859838,
+      "grad_norm": 0.6294994950294495,
+      "learning_rate": 0.0003203475445223961,
+      "loss": 3.394,
+      "step": 43300
+    },
+    {
+      "epoch": 4.6738544474393535,
+      "grad_norm": 0.6598164439201355,
+      "learning_rate": 0.0003200237452779277,
+      "loss": 3.3937,
+      "step": 43350
+    },
+    {
+      "epoch": 4.679245283018868,
+      "grad_norm": 0.6587915420532227,
+      "learning_rate": 0.00031969994603345926,
+      "loss": 3.405,
+      "step": 43400
+    },
+    {
+      "epoch": 4.684636118598383,
+      "grad_norm": 0.6498952507972717,
+      "learning_rate": 0.00031937614678899075,
+      "loss": 3.3921,
+      "step": 43450
+    },
+    {
+      "epoch": 4.690026954177897,
+      "grad_norm": 0.7558854818344116,
+      "learning_rate": 0.00031905234754452236,
+      "loss": 3.4051,
+      "step": 43500
+    },
+    {
+      "epoch": 4.695417789757412,
+      "grad_norm": 0.6848349571228027,
+      "learning_rate": 0.0003187285483000539,
+      "loss": 3.3883,
+      "step": 43550
+    },
+    {
+      "epoch": 4.7008086253369274,
+      "grad_norm": 0.6309284567832947,
+      "learning_rate": 0.0003184047490555855,
+      "loss": 3.3793,
+      "step": 43600
+    },
+    {
+      "epoch": 4.706199460916442,
+      "grad_norm": 0.6401345729827881,
+      "learning_rate": 0.00031808094981111706,
+      "loss": 3.3948,
+      "step": 43650
+    },
+    {
+      "epoch": 4.711590296495957,
+      "grad_norm": 0.6336686611175537,
+      "learning_rate": 0.00031775715056664867,
+      "loss": 3.3802,
+      "step": 43700
+    },
+    {
+      "epoch": 4.716981132075472,
+      "grad_norm": 0.6566705703735352,
+      "learning_rate": 0.0003174333513221802,
+      "loss": 3.3997,
+      "step": 43750
+    },
+    {
+      "epoch": 4.722371967654986,
+      "grad_norm": 0.641469419002533,
+      "learning_rate": 0.00031710955207771177,
+      "loss": 3.3929,
+      "step": 43800
+    },
+    {
+      "epoch": 4.727762803234501,
+      "grad_norm": 0.6639798879623413,
+      "learning_rate": 0.0003167857528332434,
+      "loss": 3.3948,
+      "step": 43850
+    },
+    {
+      "epoch": 4.7331536388140165,
+      "grad_norm": 0.6445428729057312,
+      "learning_rate": 0.0003164619535887749,
+      "loss": 3.3984,
+      "step": 43900
+    },
+    {
+      "epoch": 4.738544474393531,
+      "grad_norm": 0.6511913537979126,
+      "learning_rate": 0.00031613815434430653,
+      "loss": 3.3889,
+      "step": 43950
+    },
+    {
+      "epoch": 4.743935309973046,
+      "grad_norm": 0.6922048330307007,
+      "learning_rate": 0.0003158143550998381,
+      "loss": 3.3929,
+      "step": 44000
+    },
+    {
+      "epoch": 4.743935309973046,
+      "eval_accuracy": 0.3794368542555153,
+      "eval_loss": 3.420168399810791,
+      "eval_runtime": 183.4637,
+      "eval_samples_per_second": 98.172,
+      "eval_steps_per_second": 6.137,
+      "step": 44000
+    },
+    {
+      "epoch": 4.74932614555256,
+      "grad_norm": 0.8792603015899658,
+      "learning_rate": 0.0003154905558553697,
+      "loss": 3.403,
+      "step": 44050
+    },
+    {
+      "epoch": 4.754716981132075,
+      "grad_norm": 0.676875650882721,
+      "learning_rate": 0.0003151667566109012,
+      "loss": 3.3966,
+      "step": 44100
+    },
+    {
+      "epoch": 4.7601078167115904,
+      "grad_norm": 0.675866425037384,
+      "learning_rate": 0.00031484943335132213,
+      "loss": 3.4062,
+      "step": 44150
+    },
+    {
+      "epoch": 4.765498652291106,
+      "grad_norm": 0.6132005453109741,
+      "learning_rate": 0.0003145256341068537,
+      "loss": 3.4029,
+      "step": 44200
+    },
+    {
+      "epoch": 4.77088948787062,
+      "grad_norm": 0.6132225394248962,
+      "learning_rate": 0.0003142018348623853,
+      "loss": 3.3648,
+      "step": 44250
+    },
+    {
+      "epoch": 4.776280323450135,
+      "grad_norm": 0.6344752907752991,
+      "learning_rate": 0.00031387803561791684,
+      "loss": 3.3899,
+      "step": 44300
+    },
+    {
+      "epoch": 4.781671159029649,
+      "grad_norm": 0.6366958022117615,
+      "learning_rate": 0.00031355423637344844,
+      "loss": 3.3942,
+      "step": 44350
+    },
+    {
+      "epoch": 4.787061994609164,
+      "grad_norm": 0.6339268088340759,
+      "learning_rate": 0.00031323043712898,
+      "loss": 3.3764,
+      "step": 44400
+    },
+    {
+      "epoch": 4.7924528301886795,
+      "grad_norm": 0.6333193182945251,
+      "learning_rate": 0.0003129066378845116,
+      "loss": 3.4021,
+      "step": 44450
+    },
+    {
+      "epoch": 4.797843665768194,
+      "grad_norm": 0.6859214305877686,
+      "learning_rate": 0.00031258283864004315,
+      "loss": 3.3991,
+      "step": 44500
+    },
+    {
+      "epoch": 4.803234501347709,
+      "grad_norm": 0.6658663153648376,
+      "learning_rate": 0.0003122590393955747,
+      "loss": 3.3747,
+      "step": 44550
+    },
+    {
+      "epoch": 4.808625336927224,
+      "grad_norm": 0.658751904964447,
+      "learning_rate": 0.0003119352401511063,
+      "loss": 3.3873,
+      "step": 44600
+    },
+    {
+      "epoch": 4.814016172506738,
+      "grad_norm": 0.6287225484848022,
+      "learning_rate": 0.00031161144090663786,
+      "loss": 3.3881,
+      "step": 44650
+    },
+    {
+      "epoch": 4.819407008086253,
+      "grad_norm": 0.6663537621498108,
+      "learning_rate": 0.00031128764166216946,
+      "loss": 3.389,
+      "step": 44700
+    },
+    {
+      "epoch": 4.824797843665769,
+      "grad_norm": 0.6953670382499695,
+      "learning_rate": 0.00031096384241770096,
+      "loss": 3.4036,
+      "step": 44750
+    },
+    {
+      "epoch": 4.830188679245283,
+      "grad_norm": 0.6840183734893799,
+      "learning_rate": 0.0003106400431732326,
+      "loss": 3.374,
+      "step": 44800
+    },
+    {
+      "epoch": 4.835579514824798,
+      "grad_norm": 0.6771771311759949,
+      "learning_rate": 0.0003103162439287641,
+      "loss": 3.3881,
+      "step": 44850
+    },
+    {
+      "epoch": 4.840970350404312,
+      "grad_norm": 0.6342368721961975,
+      "learning_rate": 0.0003099924446842957,
+      "loss": 3.3658,
+      "step": 44900
+    },
+    {
+      "epoch": 4.846361185983827,
+      "grad_norm": 0.6187852621078491,
+      "learning_rate": 0.00030966864543982727,
+      "loss": 3.4036,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "grad_norm": 0.6666272282600403,
+      "learning_rate": 0.0003093448461953588,
+      "loss": 3.389,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "eval_accuracy": 0.38026935268194983,
+      "eval_loss": 3.4122982025146484,
+      "eval_runtime": 183.2917,
+      "eval_samples_per_second": 98.264,
+      "eval_steps_per_second": 6.143,
+      "step": 45000
+    },
+    {
+      "epoch": 4.857142857142857,
+      "grad_norm": 0.7002741694450378,
+      "learning_rate": 0.0003090210469508904,
+      "loss": 3.4004,
+      "step": 45050
+    },
+    {
+      "epoch": 4.862533692722372,
+      "grad_norm": 0.648239016532898,
+      "learning_rate": 0.00030869724770642197,
+      "loss": 3.4038,
+      "step": 45100
+    },
+    {
+      "epoch": 4.867924528301887,
+      "grad_norm": 0.6664896607398987,
+      "learning_rate": 0.0003083734484619536,
+      "loss": 3.3746,
+      "step": 45150
+    },
+    {
+      "epoch": 4.873315363881401,
+      "grad_norm": 0.6408718824386597,
+      "learning_rate": 0.00030804964921748513,
+      "loss": 3.3832,
+      "step": 45200
+    },
+    {
+      "epoch": 4.878706199460916,
+      "grad_norm": 0.6719059944152832,
+      "learning_rate": 0.00030772584997301673,
+      "loss": 3.3973,
+      "step": 45250
+    },
+    {
+      "epoch": 4.884097035040432,
+      "grad_norm": 0.638828456401825,
+      "learning_rate": 0.0003074020507285483,
+      "loss": 3.4089,
+      "step": 45300
+    },
+    {
+      "epoch": 4.889487870619946,
+      "grad_norm": 0.6789811849594116,
+      "learning_rate": 0.0003070782514840799,
+      "loss": 3.3876,
+      "step": 45350
+    },
+    {
+      "epoch": 4.894878706199461,
+      "grad_norm": 0.6199337840080261,
+      "learning_rate": 0.00030675445223961144,
+      "loss": 3.3971,
+      "step": 45400
+    },
+    {
+      "epoch": 4.900269541778976,
+      "grad_norm": 0.6504912376403809,
+      "learning_rate": 0.00030643065299514294,
+      "loss": 3.3937,
+      "step": 45450
+    },
+    {
+      "epoch": 4.90566037735849,
+      "grad_norm": 0.675342857837677,
+      "learning_rate": 0.00030610685375067454,
+      "loss": 3.3929,
+      "step": 45500
+    },
+    {
+      "epoch": 4.9110512129380055,
+      "grad_norm": 0.6166520714759827,
+      "learning_rate": 0.0003057830545062061,
+      "loss": 3.3939,
+      "step": 45550
+    },
+    {
+      "epoch": 4.916442048517521,
+      "grad_norm": 0.6337319016456604,
+      "learning_rate": 0.0003054592552617377,
+      "loss": 3.3898,
+      "step": 45600
+    },
+    {
+      "epoch": 4.921832884097035,
+      "grad_norm": 0.6474262475967407,
+      "learning_rate": 0.00030513545601726925,
+      "loss": 3.3924,
+      "step": 45650
+    },
+    {
+      "epoch": 4.92722371967655,
+      "grad_norm": 0.687142014503479,
+      "learning_rate": 0.00030481165677280085,
+      "loss": 3.3675,
+      "step": 45700
+    },
+    {
+      "epoch": 4.932614555256064,
+      "grad_norm": 0.692780613899231,
+      "learning_rate": 0.0003044878575283324,
+      "loss": 3.3842,
+      "step": 45750
+    },
+    {
+      "epoch": 4.938005390835579,
+      "grad_norm": 0.6651083827018738,
+      "learning_rate": 0.000304164058283864,
+      "loss": 3.4037,
+      "step": 45800
+    },
+    {
+      "epoch": 4.943396226415095,
+      "grad_norm": 0.7113507390022278,
+      "learning_rate": 0.00030384025903939556,
+      "loss": 3.3832,
+      "step": 45850
+    },
+    {
+      "epoch": 4.948787061994609,
+      "grad_norm": 0.6864883303642273,
+      "learning_rate": 0.0003035164597949271,
+      "loss": 3.3896,
+      "step": 45900
+    },
+    {
+      "epoch": 4.954177897574124,
+      "grad_norm": 0.5930712223052979,
+      "learning_rate": 0.0003031926605504587,
+      "loss": 3.3812,
+      "step": 45950
+    },
+    {
+      "epoch": 4.959568733153639,
+      "grad_norm": 0.60821133852005,
+      "learning_rate": 0.00030286886130599026,
+      "loss": 3.3876,
+      "step": 46000
+    },
+    {
+      "epoch": 4.959568733153639,
+      "eval_accuracy": 0.380540441637047,
+      "eval_loss": 3.4126551151275635,
+      "eval_runtime": 183.4652,
+      "eval_samples_per_second": 98.171,
+      "eval_steps_per_second": 6.137,
+      "step": 46000
+    },
+    {
+      "epoch": 4.964959568733153,
+      "grad_norm": 0.6533971428871155,
+      "learning_rate": 0.00030254506206152187,
+      "loss": 3.398,
+      "step": 46050
+    },
+    {
+      "epoch": 4.9703504043126685,
+      "grad_norm": 0.633718729019165,
+      "learning_rate": 0.00030222126281705336,
+      "loss": 3.368,
+      "step": 46100
+    },
+    {
+      "epoch": 4.975741239892184,
+      "grad_norm": 0.6307101845741272,
+      "learning_rate": 0.0003019039395574743,
+      "loss": 3.3747,
+      "step": 46150
+    },
+    {
+      "epoch": 4.981132075471698,
+      "grad_norm": 0.6983139514923096,
+      "learning_rate": 0.00030158014031300587,
+      "loss": 3.3861,
+      "step": 46200
+    },
+    {
+      "epoch": 4.986522911051213,
+      "grad_norm": 0.6607715487480164,
+      "learning_rate": 0.00030125634106853747,
+      "loss": 3.4037,
+      "step": 46250
+    },
+    {
+      "epoch": 4.991913746630727,
+      "grad_norm": 0.6260371804237366,
+      "learning_rate": 0.000300932541824069,
+      "loss": 3.4017,
+      "step": 46300
+    },
+    {
+      "epoch": 4.997304582210242,
+      "grad_norm": 0.6400908827781677,
+      "learning_rate": 0.0003006087425796006,
+      "loss": 3.3789,
+      "step": 46350
+    },
+    {
+      "epoch": 5.002695417789758,
+      "grad_norm": 0.7368932366371155,
+      "learning_rate": 0.0003002849433351322,
+      "loss": 3.33,
+      "step": 46400
+    },
+    {
+      "epoch": 5.008086253369272,
+      "grad_norm": 0.7171549201011658,
+      "learning_rate": 0.00029996114409066373,
+      "loss": 3.3049,
+      "step": 46450
+    },
+    {
+      "epoch": 5.013477088948787,
+      "grad_norm": 0.6396157145500183,
+      "learning_rate": 0.00029963734484619533,
+      "loss": 3.307,
+      "step": 46500
+    },
+    {
+      "epoch": 5.018867924528302,
+      "grad_norm": 0.6665785312652588,
+      "learning_rate": 0.0002993135456017269,
+      "loss": 3.3074,
+      "step": 46550
+    },
+    {
+      "epoch": 5.024258760107816,
+      "grad_norm": 0.6311437487602234,
+      "learning_rate": 0.0002989897463572585,
+      "loss": 3.2912,
+      "step": 46600
+    },
+    {
+      "epoch": 5.0296495956873315,
+      "grad_norm": 0.6656121611595154,
+      "learning_rate": 0.00029866594711279004,
+      "loss": 3.2958,
+      "step": 46650
+    },
+    {
+      "epoch": 5.035040431266847,
+      "grad_norm": 0.6491976380348206,
+      "learning_rate": 0.00029834214786832164,
+      "loss": 3.2989,
+      "step": 46700
+    },
+    {
+      "epoch": 5.040431266846361,
+      "grad_norm": 0.6659978628158569,
+      "learning_rate": 0.0002980183486238532,
+      "loss": 3.3037,
+      "step": 46750
+    },
+    {
+      "epoch": 5.045822102425876,
+      "grad_norm": 0.7180138826370239,
+      "learning_rate": 0.00029769454937938474,
+      "loss": 3.3127,
+      "step": 46800
+    },
+    {
+      "epoch": 5.051212938005391,
+      "grad_norm": 0.6873078346252441,
+      "learning_rate": 0.0002973707501349163,
+      "loss": 3.3127,
+      "step": 46850
+    },
+    {
+      "epoch": 5.056603773584905,
+      "grad_norm": 0.6599047183990479,
+      "learning_rate": 0.0002970469508904479,
+      "loss": 3.3074,
+      "step": 46900
+    },
+    {
+      "epoch": 5.061994609164421,
+      "grad_norm": 0.6925784349441528,
+      "learning_rate": 0.00029672315164597945,
+      "loss": 3.3031,
+      "step": 46950
+    },
+    {
+      "epoch": 5.067385444743936,
+      "grad_norm": 0.6558621525764465,
+      "learning_rate": 0.00029639935240151105,
+      "loss": 3.3161,
+      "step": 47000
+    },
+    {
+      "epoch": 5.067385444743936,
+      "eval_accuracy": 0.38096636095727976,
+      "eval_loss": 3.4122540950775146,
+      "eval_runtime": 183.2457,
+      "eval_samples_per_second": 98.289,
+      "eval_steps_per_second": 6.145,
+      "step": 47000
+    },
+    {
+      "epoch": 5.07277628032345,
+      "grad_norm": 0.67520672082901,
+      "learning_rate": 0.0002960755531570426,
+      "loss": 3.3086,
+      "step": 47050
+    },
+    {
+      "epoch": 5.078167115902965,
+      "grad_norm": 0.6475718021392822,
+      "learning_rate": 0.0002957517539125742,
+      "loss": 3.3043,
+      "step": 47100
+    },
+    {
+      "epoch": 5.083557951482479,
+      "grad_norm": 0.6276121735572815,
+      "learning_rate": 0.00029542795466810576,
+      "loss": 3.3149,
+      "step": 47150
+    },
+    {
+      "epoch": 5.0889487870619945,
+      "grad_norm": 0.6585389375686646,
+      "learning_rate": 0.0002951041554236373,
+      "loss": 3.3141,
+      "step": 47200
+    },
+    {
+      "epoch": 5.09433962264151,
+      "grad_norm": 0.654176652431488,
+      "learning_rate": 0.00029478035617916886,
+      "loss": 3.2993,
+      "step": 47250
+    },
+    {
+      "epoch": 5.099730458221024,
+      "grad_norm": 0.6916609406471252,
+      "learning_rate": 0.00029445655693470047,
+      "loss": 3.3228,
+      "step": 47300
+    },
+    {
+      "epoch": 5.105121293800539,
+      "grad_norm": 0.6311585903167725,
+      "learning_rate": 0.000294132757690232,
+      "loss": 3.3229,
+      "step": 47350
+    },
+    {
+      "epoch": 5.110512129380054,
+      "grad_norm": 0.6877700090408325,
+      "learning_rate": 0.0002938089584457636,
+      "loss": 3.3043,
+      "step": 47400
+    },
+    {
+      "epoch": 5.115902964959568,
+      "grad_norm": 0.7259128093719482,
+      "learning_rate": 0.00029348515920129517,
+      "loss": 3.3085,
+      "step": 47450
+    },
+    {
+      "epoch": 5.121293800539084,
+      "grad_norm": 0.6702103018760681,
+      "learning_rate": 0.0002931613599568267,
+      "loss": 3.2989,
+      "step": 47500
+    },
+    {
+      "epoch": 5.126684636118599,
+      "grad_norm": 0.6595789790153503,
+      "learning_rate": 0.00029283756071235833,
+      "loss": 3.3049,
+      "step": 47550
+    },
+    {
+      "epoch": 5.132075471698113,
+      "grad_norm": 0.746311366558075,
+      "learning_rate": 0.0002925137614678899,
+      "loss": 3.2956,
+      "step": 47600
+    },
+    {
+      "epoch": 5.137466307277628,
+      "grad_norm": 0.714484691619873,
+      "learning_rate": 0.00029218996222342143,
+      "loss": 3.3097,
+      "step": 47650
+    },
+    {
+      "epoch": 5.142857142857143,
+      "grad_norm": 0.6939281821250916,
+      "learning_rate": 0.00029186616297895303,
+      "loss": 3.3248,
+      "step": 47700
+    },
+    {
+      "epoch": 5.1482479784366575,
+      "grad_norm": 0.7023195028305054,
+      "learning_rate": 0.0002915423637344846,
+      "loss": 3.3148,
+      "step": 47750
+    },
+    {
+      "epoch": 5.153638814016173,
+      "grad_norm": 0.633255660533905,
+      "learning_rate": 0.00029121856449001613,
+      "loss": 3.3001,
+      "step": 47800
+    },
+    {
+      "epoch": 5.159029649595688,
+      "grad_norm": 0.7074955105781555,
+      "learning_rate": 0.00029089476524554774,
+      "loss": 3.3097,
+      "step": 47850
+    },
+    {
+      "epoch": 5.164420485175202,
+      "grad_norm": 0.7120252251625061,
+      "learning_rate": 0.0002905709660010793,
+      "loss": 3.3325,
+      "step": 47900
+    },
+    {
+      "epoch": 5.169811320754717,
+      "grad_norm": 0.6911315321922302,
+      "learning_rate": 0.0002902471667566109,
+      "loss": 3.3164,
+      "step": 47950
+    },
+    {
+      "epoch": 5.175202156334231,
+      "grad_norm": 0.689180850982666,
+      "learning_rate": 0.00028992336751214245,
+      "loss": 3.3084,
+      "step": 48000
+    },
+    {
+      "epoch": 5.175202156334231,
+      "eval_accuracy": 0.3809488678423416,
+      "eval_loss": 3.4123737812042236,
+      "eval_runtime": 183.5311,
+      "eval_samples_per_second": 98.136,
+      "eval_steps_per_second": 6.135,
+      "step": 48000
+    },
+    {
+      "epoch": 5.180592991913747,
+      "grad_norm": 0.6789995431900024,
+      "learning_rate": 0.00028959956826767405,
+      "loss": 3.3138,
+      "step": 48050
+    },
+    {
+      "epoch": 5.185983827493262,
+      "grad_norm": 0.6542081832885742,
+      "learning_rate": 0.00028927576902320555,
+      "loss": 3.305,
+      "step": 48100
+    },
+    {
+      "epoch": 5.191374663072776,
+      "grad_norm": 0.6842049956321716,
+      "learning_rate": 0.00028895196977873715,
+      "loss": 3.3137,
+      "step": 48150
+    },
+    {
+      "epoch": 5.196765498652291,
+      "grad_norm": 0.6928696632385254,
+      "learning_rate": 0.0002886346465191581,
+      "loss": 3.3279,
+      "step": 48200
+    },
+    {
+      "epoch": 5.202156334231806,
+      "grad_norm": 0.6878511905670166,
+      "learning_rate": 0.00028831084727468965,
+      "loss": 3.32,
+      "step": 48250
+    },
+    {
+      "epoch": 5.2075471698113205,
+      "grad_norm": 0.7233080267906189,
+      "learning_rate": 0.00028798704803022126,
+      "loss": 3.3229,
+      "step": 48300
+    },
+    {
+      "epoch": 5.212938005390836,
+      "grad_norm": 0.7335208058357239,
+      "learning_rate": 0.0002876632487857528,
+      "loss": 3.3286,
+      "step": 48350
+    },
+    {
+      "epoch": 5.218328840970351,
+      "grad_norm": 0.7119630575180054,
+      "learning_rate": 0.0002873394495412844,
+      "loss": 3.3241,
+      "step": 48400
+    },
+    {
+      "epoch": 5.223719676549865,
+      "grad_norm": 0.6260629892349243,
+      "learning_rate": 0.0002870156502968159,
+      "loss": 3.3137,
+      "step": 48450
+    },
+    {
+      "epoch": 5.22911051212938,
+      "grad_norm": 0.6813670992851257,
+      "learning_rate": 0.0002866918510523475,
+      "loss": 3.3479,
+      "step": 48500
+    },
+    {
+      "epoch": 5.234501347708895,
+      "grad_norm": 0.6499662399291992,
+      "learning_rate": 0.00028636805180787907,
+      "loss": 3.3214,
+      "step": 48550
+    },
+    {
+      "epoch": 5.2398921832884096,
+      "grad_norm": 0.6558360457420349,
+      "learning_rate": 0.00028604425256341067,
+      "loss": 3.3194,
+      "step": 48600
+    },
+    {
+      "epoch": 5.245283018867925,
+      "grad_norm": 0.6462336778640747,
+      "learning_rate": 0.0002857204533189422,
+      "loss": 3.3194,
+      "step": 48650
+    },
+    {
+      "epoch": 5.250673854447439,
+      "grad_norm": 0.7324293851852417,
+      "learning_rate": 0.0002853966540744738,
+      "loss": 3.3235,
+      "step": 48700
+    },
+    {
+      "epoch": 5.256064690026954,
+      "grad_norm": 0.6889777779579163,
+      "learning_rate": 0.0002850728548300054,
+      "loss": 3.3417,
+      "step": 48750
+    },
+    {
+      "epoch": 5.261455525606469,
+      "grad_norm": 0.6718788743019104,
+      "learning_rate": 0.000284749055585537,
+      "loss": 3.3417,
+      "step": 48800
+    },
+    {
+      "epoch": 5.2668463611859835,
+      "grad_norm": 0.6515158414840698,
+      "learning_rate": 0.0002844252563410685,
+      "loss": 3.3182,
+      "step": 48850
+    },
+    {
+      "epoch": 5.272237196765499,
+      "grad_norm": 0.6929355263710022,
+      "learning_rate": 0.0002841014570966001,
+      "loss": 3.3162,
+      "step": 48900
+    },
+    {
+      "epoch": 5.277628032345014,
+      "grad_norm": 0.6629584431648254,
+      "learning_rate": 0.00028377765785213163,
+      "loss": 3.3398,
+      "step": 48950
+    },
+    {
+      "epoch": 5.283018867924528,
+      "grad_norm": 0.6952082514762878,
+      "learning_rate": 0.00028345385860766324,
+      "loss": 3.329,
+      "step": 49000
+    },
+    {
+      "epoch": 5.283018867924528,
+      "eval_accuracy": 0.3814610575554991,
+      "eval_loss": 3.408320665359497,
+      "eval_runtime": 183.4228,
+      "eval_samples_per_second": 98.194,
+      "eval_steps_per_second": 6.139,
+      "step": 49000
+    },
+    {
+      "epoch": 5.288409703504043,
+      "grad_norm": 0.6505891680717468,
+      "learning_rate": 0.0002831300593631948,
+      "loss": 3.3372,
+      "step": 49050
+    },
+    {
+      "epoch": 5.293800539083558,
+      "grad_norm": 0.6779699325561523,
+      "learning_rate": 0.0002828062601187264,
+      "loss": 3.3338,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2991913746630726,
+      "grad_norm": 0.6804792881011963,
+      "learning_rate": 0.00028248246087425794,
+      "loss": 3.3319,
+      "step": 49150
+    },
+    {
+      "epoch": 5.304582210242588,
+      "grad_norm": 0.7583218812942505,
+      "learning_rate": 0.0002821586616297895,
+      "loss": 3.3396,
+      "step": 49200
+    },
+    {
+      "epoch": 5.309973045822103,
+      "grad_norm": 0.6410008668899536,
+      "learning_rate": 0.0002818348623853211,
+      "loss": 3.3272,
+      "step": 49250
+    },
+    {
+      "epoch": 5.315363881401617,
+      "grad_norm": 0.7008423805236816,
+      "learning_rate": 0.00028151106314085265,
+      "loss": 3.3286,
+      "step": 49300
+    },
+    {
+      "epoch": 5.320754716981132,
+      "grad_norm": 0.7405087351799011,
+      "learning_rate": 0.0002811872638963842,
+      "loss": 3.34,
+      "step": 49350
+    },
+    {
+      "epoch": 5.3261455525606465,
+      "grad_norm": 0.7460556030273438,
+      "learning_rate": 0.0002808634646519158,
+      "loss": 3.3254,
+      "step": 49400
+    },
+    {
+      "epoch": 5.331536388140162,
+      "grad_norm": 0.6831420660018921,
+      "learning_rate": 0.00028053966540744736,
+      "loss": 3.3177,
+      "step": 49450
+    },
+    {
+      "epoch": 5.336927223719677,
+      "grad_norm": 0.6601572036743164,
+      "learning_rate": 0.0002802158661629789,
+      "loss": 3.3444,
+      "step": 49500
+    },
+    {
+      "epoch": 5.342318059299191,
+      "grad_norm": 0.667539119720459,
+      "learning_rate": 0.0002798920669185105,
+      "loss": 3.3185,
+      "step": 49550
+    },
+    {
+      "epoch": 5.347708894878706,
+      "grad_norm": 0.69802325963974,
+      "learning_rate": 0.00027956826767404206,
+      "loss": 3.3355,
+      "step": 49600
+    },
+    {
+      "epoch": 5.353099730458221,
+      "grad_norm": 0.668803334236145,
+      "learning_rate": 0.00027924446842957367,
+      "loss": 3.3321,
+      "step": 49650
+    },
+    {
+      "epoch": 5.3584905660377355,
+      "grad_norm": 0.6525208950042725,
+      "learning_rate": 0.0002789206691851052,
+      "loss": 3.3391,
+      "step": 49700
+    },
+    {
+      "epoch": 5.363881401617251,
+      "grad_norm": 0.7439292669296265,
+      "learning_rate": 0.00027859686994063677,
+      "loss": 3.3249,
+      "step": 49750
+    },
+    {
+      "epoch": 5.369272237196766,
+      "grad_norm": 0.669330894947052,
+      "learning_rate": 0.0002782795466810577,
+      "loss": 3.3276,
+      "step": 49800
+    },
+    {
+      "epoch": 5.37466307277628,
+      "grad_norm": 0.6631773710250854,
+      "learning_rate": 0.00027795574743658927,
+      "loss": 3.3419,
+      "step": 49850
+    },
+    {
+      "epoch": 5.380053908355795,
+      "grad_norm": 0.6949469447135925,
+      "learning_rate": 0.0002776319481921209,
+      "loss": 3.336,
+      "step": 49900
+    },
+    {
+      "epoch": 5.38544474393531,
+      "grad_norm": 0.6591998934745789,
+      "learning_rate": 0.0002773081489476524,
+      "loss": 3.3223,
+      "step": 49950
+    },
+    {
+      "epoch": 5.390835579514825,
+      "grad_norm": 0.6729640960693359,
+      "learning_rate": 0.00027698434970318403,
+      "loss": 3.3415,
+      "step": 50000
+    },
+    {
+      "epoch": 5.390835579514825,
+      "eval_accuracy": 0.3817132409081165,
+      "eval_loss": 3.4021902084350586,
+      "eval_runtime": 183.2074,
+      "eval_samples_per_second": 98.309,
+      "eval_steps_per_second": 6.146,
+      "step": 50000
+    },
+    {
+      "epoch": 5.39622641509434,
+      "grad_norm": 0.6849084496498108,
+      "learning_rate": 0.0002766605504587156,
+      "loss": 3.3519,
+      "step": 50050
+    },
+    {
+      "epoch": 5.401617250673855,
+      "grad_norm": 0.677198052406311,
+      "learning_rate": 0.00027633675121424713,
+      "loss": 3.3425,
+      "step": 50100
+    },
+    {
+      "epoch": 5.407008086253369,
+      "grad_norm": 0.692452609539032,
+      "learning_rate": 0.0002760129519697787,
+      "loss": 3.3341,
+      "step": 50150
+    },
+    {
+      "epoch": 5.412398921832884,
+      "grad_norm": 0.6756576895713806,
+      "learning_rate": 0.0002756891527253103,
+      "loss": 3.3153,
+      "step": 50200
+    },
+    {
+      "epoch": 5.4177897574123985,
+      "grad_norm": 0.6618334650993347,
+      "learning_rate": 0.00027536535348084184,
+      "loss": 3.346,
+      "step": 50250
+    },
+    {
+      "epoch": 5.423180592991914,
+      "grad_norm": 0.6220849752426147,
+      "learning_rate": 0.00027504155423637344,
+      "loss": 3.3392,
+      "step": 50300
+    },
+    {
+      "epoch": 5.428571428571429,
+      "grad_norm": 0.7351338267326355,
+      "learning_rate": 0.000274717754991905,
+      "loss": 3.3278,
+      "step": 50350
+    },
+    {
+      "epoch": 5.433962264150943,
+      "grad_norm": 0.7283994555473328,
+      "learning_rate": 0.0002743939557474366,
+      "loss": 3.3469,
+      "step": 50400
+    },
+    {
+      "epoch": 5.439353099730458,
+      "grad_norm": 0.729455828666687,
+      "learning_rate": 0.00027407015650296815,
+      "loss": 3.3467,
+      "step": 50450
+    },
+    {
+      "epoch": 5.444743935309973,
+      "grad_norm": 0.6562330722808838,
+      "learning_rate": 0.0002737463572584997,
+      "loss": 3.3311,
+      "step": 50500
+    },
+    {
+      "epoch": 5.450134770889488,
+      "grad_norm": 0.6799193024635315,
+      "learning_rate": 0.00027342255801403125,
+      "loss": 3.3278,
+      "step": 50550
+    },
+    {
+      "epoch": 5.455525606469003,
+      "grad_norm": 0.7393507957458496,
+      "learning_rate": 0.00027309875876956285,
+      "loss": 3.3434,
+      "step": 50600
+    },
+    {
+      "epoch": 5.460916442048518,
+      "grad_norm": 0.8206065893173218,
+      "learning_rate": 0.0002727749595250944,
+      "loss": 3.3369,
+      "step": 50650
+    },
+    {
+      "epoch": 5.466307277628032,
+      "grad_norm": 0.6812174916267395,
+      "learning_rate": 0.000272451160280626,
+      "loss": 3.3274,
+      "step": 50700
+    },
+    {
+      "epoch": 5.471698113207547,
+      "grad_norm": 0.6868910193443298,
+      "learning_rate": 0.00027212736103615756,
+      "loss": 3.3155,
+      "step": 50750
+    },
+    {
+      "epoch": 5.4770889487870615,
+      "grad_norm": 0.6723341941833496,
+      "learning_rate": 0.00027180356179168916,
+      "loss": 3.3438,
+      "step": 50800
+    },
+    {
+      "epoch": 5.482479784366577,
+      "grad_norm": 0.6819263100624084,
+      "learning_rate": 0.0002714797625472207,
+      "loss": 3.3283,
+      "step": 50850
+    },
+    {
+      "epoch": 5.487870619946092,
+      "grad_norm": 0.742080569267273,
+      "learning_rate": 0.00027115596330275226,
+      "loss": 3.3372,
+      "step": 50900
+    },
+    {
+      "epoch": 5.493261455525606,
+      "grad_norm": 0.6924917101860046,
+      "learning_rate": 0.0002708321640582838,
+      "loss": 3.3529,
+      "step": 50950
+    },
+    {
+      "epoch": 5.498652291105121,
+      "grad_norm": 0.7526167631149292,
+      "learning_rate": 0.0002705083648138154,
+      "loss": 3.3172,
+      "step": 51000
+    },
+    {
+      "epoch": 5.498652291105121,
+      "eval_accuracy": 0.3821663234503029,
+      "eval_loss": 3.3990259170532227,
+      "eval_runtime": 183.3129,
+      "eval_samples_per_second": 98.253,
+      "eval_steps_per_second": 6.143,
+      "step": 51000
+    },
+    {
+      "epoch": 5.504043126684636,
+      "grad_norm": 0.6703799962997437,
+      "learning_rate": 0.00027018456556934697,
+      "loss": 3.324,
+      "step": 51050
+    },
+    {
+      "epoch": 5.509433962264151,
+      "grad_norm": 0.6759470105171204,
+      "learning_rate": 0.0002698607663248786,
+      "loss": 3.3338,
+      "step": 51100
+    },
+    {
+      "epoch": 5.514824797843666,
+      "grad_norm": 0.7208074331283569,
+      "learning_rate": 0.0002695369670804101,
+      "loss": 3.3522,
+      "step": 51150
+    },
+    {
+      "epoch": 5.520215633423181,
+      "grad_norm": 0.6823359727859497,
+      "learning_rate": 0.0002692131678359417,
+      "loss": 3.3276,
+      "step": 51200
+    },
+    {
+      "epoch": 5.525606469002695,
+      "grad_norm": 0.7479263544082642,
+      "learning_rate": 0.0002688893685914733,
+      "loss": 3.3309,
+      "step": 51250
+    },
+    {
+      "epoch": 5.53099730458221,
+      "grad_norm": 0.6691256165504456,
+      "learning_rate": 0.00026856556934700483,
+      "loss": 3.3452,
+      "step": 51300
+    },
+    {
+      "epoch": 5.536388140161725,
+      "grad_norm": 0.6419733762741089,
+      "learning_rate": 0.0002682417701025364,
+      "loss": 3.3253,
+      "step": 51350
+    },
+    {
+      "epoch": 5.54177897574124,
+      "grad_norm": 0.6932953000068665,
+      "learning_rate": 0.000267917970858068,
+      "loss": 3.331,
+      "step": 51400
+    },
+    {
+      "epoch": 5.547169811320755,
+      "grad_norm": 0.7635331749916077,
+      "learning_rate": 0.00026759417161359954,
+      "loss": 3.3509,
+      "step": 51450
+    },
+    {
+      "epoch": 5.55256064690027,
+      "grad_norm": 0.680497407913208,
+      "learning_rate": 0.0002672703723691311,
+      "loss": 3.3369,
+      "step": 51500
+    },
+    {
+      "epoch": 5.557951482479784,
+      "grad_norm": 0.7138260006904602,
+      "learning_rate": 0.0002669465731246627,
+      "loss": 3.3277,
+      "step": 51550
+    },
+    {
+      "epoch": 5.563342318059299,
+      "grad_norm": 0.7187690734863281,
+      "learning_rate": 0.00026662277388019424,
+      "loss": 3.3546,
+      "step": 51600
+    },
+    {
+      "epoch": 5.568733153638814,
+      "grad_norm": 0.6561439633369446,
+      "learning_rate": 0.00026629897463572585,
+      "loss": 3.3507,
+      "step": 51650
+    },
+    {
+      "epoch": 5.574123989218329,
+      "grad_norm": 0.7170794010162354,
+      "learning_rate": 0.0002659751753912574,
+      "loss": 3.3347,
+      "step": 51700
+    },
+    {
+      "epoch": 5.579514824797844,
+      "grad_norm": 0.7045460939407349,
+      "learning_rate": 0.00026565137614678895,
+      "loss": 3.3274,
+      "step": 51750
+    },
+    {
+      "epoch": 5.584905660377358,
+      "grad_norm": 0.7216756939888,
+      "learning_rate": 0.0002653275769023205,
+      "loss": 3.3404,
+      "step": 51800
+    },
+    {
+      "epoch": 5.590296495956873,
+      "grad_norm": 0.7280001640319824,
+      "learning_rate": 0.0002650037776578521,
+      "loss": 3.3364,
+      "step": 51850
+    },
+    {
+      "epoch": 5.595687331536388,
+      "grad_norm": 0.7068616151809692,
+      "learning_rate": 0.00026467997841338366,
+      "loss": 3.3432,
+      "step": 51900
+    },
+    {
+      "epoch": 5.601078167115903,
+      "grad_norm": 0.699648916721344,
+      "learning_rate": 0.00026435617916891526,
+      "loss": 3.3336,
+      "step": 51950
+    },
+    {
+      "epoch": 5.606469002695418,
+      "grad_norm": 0.7143855690956116,
+      "learning_rate": 0.0002640323799244468,
+      "loss": 3.3287,
+      "step": 52000
+    },
+    {
+      "epoch": 5.606469002695418,
+      "eval_accuracy": 0.38255160659049303,
+      "eval_loss": 3.395909070968628,
+      "eval_runtime": 183.3498,
+      "eval_samples_per_second": 98.233,
+      "eval_steps_per_second": 6.141,
+      "step": 52000
+    },
+    {
+      "epoch": 5.611859838274933,
+      "grad_norm": 0.6716243028640747,
+      "learning_rate": 0.0002637085806799784,
+      "loss": 3.3521,
+      "step": 52050
+    },
+    {
+      "epoch": 5.617250673854447,
+      "grad_norm": 0.6960698366165161,
+      "learning_rate": 0.00026338478143550997,
+      "loss": 3.3314,
+      "step": 52100
+    },
+    {
+      "epoch": 5.622641509433962,
+      "grad_norm": 0.7130643725395203,
+      "learning_rate": 0.00026306098219104157,
+      "loss": 3.3342,
+      "step": 52150
+    },
+    {
+      "epoch": 5.628032345013477,
+      "grad_norm": 0.7162200808525085,
+      "learning_rate": 0.00026273718294657307,
+      "loss": 3.3302,
+      "step": 52200
+    },
+    {
+      "epoch": 5.633423180592992,
+      "grad_norm": 0.6905233263969421,
+      "learning_rate": 0.00026241338370210467,
+      "loss": 3.3313,
+      "step": 52250
+    },
+    {
+      "epoch": 5.638814016172507,
+      "grad_norm": 0.6718230247497559,
+      "learning_rate": 0.0002620895844576362,
+      "loss": 3.3348,
+      "step": 52300
+    },
+    {
+      "epoch": 5.644204851752022,
+      "grad_norm": 0.7265104651451111,
+      "learning_rate": 0.00026176578521316783,
+      "loss": 3.3412,
+      "step": 52350
+    },
+    {
+      "epoch": 5.649595687331536,
+      "grad_norm": 0.695438802242279,
+      "learning_rate": 0.0002614419859686994,
+      "loss": 3.3288,
+      "step": 52400
+    },
+    {
+      "epoch": 5.654986522911051,
+      "grad_norm": 0.7260077595710754,
+      "learning_rate": 0.000261118186724231,
+      "loss": 3.3566,
+      "step": 52450
+    },
+    {
+      "epoch": 5.660377358490566,
+      "grad_norm": 0.7230146527290344,
+      "learning_rate": 0.00026079438747976253,
+      "loss": 3.3331,
+      "step": 52500
+    },
+    {
+      "epoch": 5.665768194070081,
+      "grad_norm": 0.720332682132721,
+      "learning_rate": 0.0002604705882352941,
+      "loss": 3.3481,
+      "step": 52550
+    },
+    {
+      "epoch": 5.671159029649596,
+      "grad_norm": 0.6841501593589783,
+      "learning_rate": 0.00026014678899082563,
+      "loss": 3.3632,
+      "step": 52600
+    },
+    {
+      "epoch": 5.67654986522911,
+      "grad_norm": 0.7262223958969116,
+      "learning_rate": 0.00025982298974635724,
+      "loss": 3.333,
+      "step": 52650
+    },
+    {
+      "epoch": 5.681940700808625,
+      "grad_norm": 0.7236334085464478,
+      "learning_rate": 0.0002594991905018888,
+      "loss": 3.3233,
+      "step": 52700
+    },
+    {
+      "epoch": 5.6873315363881405,
+      "grad_norm": 0.7088338136672974,
+      "learning_rate": 0.0002591753912574204,
+      "loss": 3.3406,
+      "step": 52750
+    },
+    {
+      "epoch": 5.692722371967655,
+      "grad_norm": 0.6878891587257385,
+      "learning_rate": 0.00025885159201295195,
+      "loss": 3.3221,
+      "step": 52800
+    },
+    {
+      "epoch": 5.69811320754717,
+      "grad_norm": 0.6763601899147034,
+      "learning_rate": 0.0002585277927684835,
+      "loss": 3.3195,
+      "step": 52850
+    },
+    {
+      "epoch": 5.703504043126685,
+      "grad_norm": 0.7065742015838623,
+      "learning_rate": 0.0002582039935240151,
+      "loss": 3.3244,
+      "step": 52900
+    },
+    {
+      "epoch": 5.708894878706199,
+      "grad_norm": 0.7355487942695618,
+      "learning_rate": 0.00025788019427954665,
+      "loss": 3.328,
+      "step": 52950
+    },
+    {
+      "epoch": 5.714285714285714,
+      "grad_norm": 0.6837366819381714,
+      "learning_rate": 0.00025755639503507826,
+      "loss": 3.3407,
+      "step": 53000
+    },
+    {
+      "epoch": 5.714285714285714,
+      "eval_accuracy": 0.3830400013212191,
+      "eval_loss": 3.3904197216033936,
+      "eval_runtime": 183.2943,
+      "eval_samples_per_second": 98.263,
+      "eval_steps_per_second": 6.143,
+      "step": 53000
+    },
+    {
+      "epoch": 5.719676549865229,
+      "grad_norm": 0.679541826248169,
+      "learning_rate": 0.00025723907177549915,
+      "loss": 3.3307,
+      "step": 53050
+    },
+    {
+      "epoch": 5.725067385444744,
+      "grad_norm": 0.7455540895462036,
+      "learning_rate": 0.00025691527253103076,
+      "loss": 3.3419,
+      "step": 53100
+    },
+    {
+      "epoch": 5.730458221024259,
+      "grad_norm": 0.6685236096382141,
+      "learning_rate": 0.0002565914732865623,
+      "loss": 3.3354,
+      "step": 53150
+    },
+    {
+      "epoch": 5.735849056603773,
+      "grad_norm": 0.7152445912361145,
+      "learning_rate": 0.00025626767404209386,
+      "loss": 3.3457,
+      "step": 53200
+    },
+    {
+      "epoch": 5.741239892183288,
+      "grad_norm": 0.67889004945755,
+      "learning_rate": 0.00025594387479762546,
+      "loss": 3.3178,
+      "step": 53250
+    },
+    {
+      "epoch": 5.7466307277628035,
+      "grad_norm": 0.7162405252456665,
+      "learning_rate": 0.000255620075553157,
+      "loss": 3.3585,
+      "step": 53300
+    },
+    {
+      "epoch": 5.752021563342318,
+      "grad_norm": 0.6406696438789368,
+      "learning_rate": 0.00025529627630868857,
+      "loss": 3.3417,
+      "step": 53350
+    },
+    {
+      "epoch": 5.757412398921833,
+      "grad_norm": 0.7152397036552429,
+      "learning_rate": 0.00025497247706422017,
+      "loss": 3.3471,
+      "step": 53400
+    },
+    {
+      "epoch": 5.762803234501348,
+      "grad_norm": 0.6572598814964294,
+      "learning_rate": 0.0002546486778197517,
+      "loss": 3.3417,
+      "step": 53450
+    },
+    {
+      "epoch": 5.768194070080862,
+      "grad_norm": 0.6745120286941528,
+      "learning_rate": 0.00025432487857528327,
+      "loss": 3.3262,
+      "step": 53500
+    },
+    {
+      "epoch": 5.773584905660377,
+      "grad_norm": 0.708234429359436,
+      "learning_rate": 0.0002540010793308149,
+      "loss": 3.3263,
+      "step": 53550
+    },
+    {
+      "epoch": 5.7789757412398925,
+      "grad_norm": 0.7930406928062439,
+      "learning_rate": 0.0002536772800863464,
+      "loss": 3.3457,
+      "step": 53600
+    },
+    {
+      "epoch": 5.784366576819407,
+      "grad_norm": 0.682910144329071,
+      "learning_rate": 0.00025335348084187803,
+      "loss": 3.3295,
+      "step": 53650
+    },
+    {
+      "epoch": 5.789757412398922,
+      "grad_norm": 0.6747221350669861,
+      "learning_rate": 0.0002530296815974096,
+      "loss": 3.3507,
+      "step": 53700
+    },
+    {
+      "epoch": 5.795148247978437,
+      "grad_norm": 0.6941880583763123,
+      "learning_rate": 0.0002527058823529412,
+      "loss": 3.3379,
+      "step": 53750
+    },
+    {
+      "epoch": 5.800539083557951,
+      "grad_norm": 0.6580685973167419,
+      "learning_rate": 0.0002523820831084727,
+      "loss": 3.326,
+      "step": 53800
+    },
+    {
+      "epoch": 5.8059299191374665,
+      "grad_norm": 0.6960239410400391,
+      "learning_rate": 0.0002520582838640043,
+      "loss": 3.3307,
+      "step": 53850
+    },
+    {
+      "epoch": 5.811320754716981,
+      "grad_norm": 0.689068078994751,
+      "learning_rate": 0.00025173448461953584,
+      "loss": 3.3308,
+      "step": 53900
+    },
+    {
+      "epoch": 5.816711590296496,
+      "grad_norm": 0.7065226435661316,
+      "learning_rate": 0.00025141068537506744,
+      "loss": 3.33,
+      "step": 53950
+    },
+    {
+      "epoch": 5.822102425876011,
+      "grad_norm": 0.6940571069717407,
+      "learning_rate": 0.000251086886130599,
+      "loss": 3.3252,
+      "step": 54000
+    },
+    {
+      "epoch": 5.822102425876011,
+      "eval_accuracy": 0.3834316949817903,
+      "eval_loss": 3.38584303855896,
+      "eval_runtime": 183.5696,
+      "eval_samples_per_second": 98.115,
+      "eval_steps_per_second": 6.134,
+      "step": 54000
+    },
+    {
+      "epoch": 5.827493261455525,
+      "grad_norm": 0.6754686236381531,
+      "learning_rate": 0.0002507630868861306,
+      "loss": 3.3337,
+      "step": 54050
+    },
+    {
+      "epoch": 5.83288409703504,
+      "grad_norm": 0.7006556987762451,
+      "learning_rate": 0.00025043928764166215,
+      "loss": 3.3234,
+      "step": 54100
+    },
+    {
+      "epoch": 5.8382749326145555,
+      "grad_norm": 0.7364419102668762,
+      "learning_rate": 0.00025011548839719375,
+      "loss": 3.3379,
+      "step": 54150
+    },
+    {
+      "epoch": 5.84366576819407,
+      "grad_norm": 0.7036047577857971,
+      "learning_rate": 0.0002497916891527253,
+      "loss": 3.3288,
+      "step": 54200
+    },
+    {
+      "epoch": 5.849056603773585,
+      "grad_norm": 0.7184184789657593,
+      "learning_rate": 0.00024946788990825686,
+      "loss": 3.3293,
+      "step": 54250
+    },
+    {
+      "epoch": 5.8544474393531,
+      "grad_norm": 0.7846869826316833,
+      "learning_rate": 0.0002491440906637884,
+      "loss": 3.3292,
+      "step": 54300
+    },
+    {
+      "epoch": 5.859838274932614,
+      "grad_norm": 0.7330741882324219,
+      "learning_rate": 0.00024882029141932,
+      "loss": 3.3411,
+      "step": 54350
+    },
+    {
+      "epoch": 5.8652291105121295,
+      "grad_norm": 0.707969069480896,
+      "learning_rate": 0.00024849649217485156,
+      "loss": 3.327,
+      "step": 54400
+    },
+    {
+      "epoch": 5.870619946091644,
+      "grad_norm": 0.7619126439094543,
+      "learning_rate": 0.00024817269293038317,
+      "loss": 3.3445,
+      "step": 54450
+    },
+    {
+      "epoch": 5.876010781671159,
+      "grad_norm": 0.6992142200469971,
+      "learning_rate": 0.0002478488936859147,
+      "loss": 3.3452,
+      "step": 54500
+    },
+    {
+      "epoch": 5.881401617250674,
+      "grad_norm": 0.6992085576057434,
+      "learning_rate": 0.00024752509444144627,
+      "loss": 3.3183,
+      "step": 54550
+    },
+    {
+      "epoch": 5.886792452830189,
+      "grad_norm": 0.7042163610458374,
+      "learning_rate": 0.00024720129519697787,
+      "loss": 3.3434,
+      "step": 54600
+    },
+    {
+      "epoch": 5.892183288409703,
+      "grad_norm": 0.6926693916320801,
+      "learning_rate": 0.0002468774959525094,
+      "loss": 3.3261,
+      "step": 54650
+    },
+    {
+      "epoch": 5.8975741239892185,
+      "grad_norm": 0.707694411277771,
+      "learning_rate": 0.00024655369670804097,
+      "loss": 3.3333,
+      "step": 54700
+    },
+    {
+      "epoch": 5.902964959568733,
+      "grad_norm": 0.6961731910705566,
+      "learning_rate": 0.0002462298974635726,
+      "loss": 3.3429,
+      "step": 54750
+    },
+    {
+      "epoch": 5.908355795148248,
+      "grad_norm": 0.7367614507675171,
+      "learning_rate": 0.00024590609821910413,
+      "loss": 3.3355,
+      "step": 54800
+    },
+    {
+      "epoch": 5.913746630727763,
+      "grad_norm": 0.7213190793991089,
+      "learning_rate": 0.0002455822989746357,
+      "loss": 3.343,
+      "step": 54850
+    },
+    {
+      "epoch": 5.919137466307277,
+      "grad_norm": 0.7634758353233337,
+      "learning_rate": 0.0002452584997301673,
+      "loss": 3.3409,
+      "step": 54900
+    },
+    {
+      "epoch": 5.9245283018867925,
+      "grad_norm": 0.6998611092567444,
+      "learning_rate": 0.00024493470048569883,
+      "loss": 3.3457,
+      "step": 54950
+    },
+    {
+      "epoch": 5.929919137466308,
+      "grad_norm": 0.7190982699394226,
+      "learning_rate": 0.00024461090124123044,
+      "loss": 3.3267,
+      "step": 55000
+    },
+    {
+      "epoch": 5.929919137466308,
+      "eval_accuracy": 0.38390183602736355,
+      "eval_loss": 3.38206148147583,
+      "eval_runtime": 183.2415,
+      "eval_samples_per_second": 98.291,
+      "eval_steps_per_second": 6.145,
+      "step": 55000
+    },
+    {
+      "epoch": 5.935309973045822,
+      "grad_norm": 0.7191337943077087,
+      "learning_rate": 0.000244287101996762,
+      "loss": 3.3569,
+      "step": 55050
+    },
+    {
+      "epoch": 5.940700808625337,
+      "grad_norm": 0.740172803401947,
+      "learning_rate": 0.00024396330275229354,
+      "loss": 3.3212,
+      "step": 55100
+    },
+    {
+      "epoch": 5.946091644204852,
+      "grad_norm": 0.7168428301811218,
+      "learning_rate": 0.00024363950350782512,
+      "loss": 3.3283,
+      "step": 55150
+    },
+    {
+      "epoch": 5.951482479784366,
+      "grad_norm": 0.7316395044326782,
+      "learning_rate": 0.0002433157042633567,
+      "loss": 3.3247,
+      "step": 55200
+    },
+    {
+      "epoch": 5.9568733153638815,
+      "grad_norm": 0.7106702327728271,
+      "learning_rate": 0.00024299190501888827,
+      "loss": 3.3313,
+      "step": 55250
+    },
+    {
+      "epoch": 5.962264150943396,
+      "grad_norm": 0.6689465045928955,
+      "learning_rate": 0.00024266810577441985,
+      "loss": 3.3231,
+      "step": 55300
+    },
+    {
+      "epoch": 5.967654986522911,
+      "grad_norm": 0.738099992275238,
+      "learning_rate": 0.00024234430652995143,
+      "loss": 3.3268,
+      "step": 55350
+    },
+    {
+      "epoch": 5.973045822102426,
+      "grad_norm": 0.7050520181655884,
+      "learning_rate": 0.00024202050728548298,
+      "loss": 3.3283,
+      "step": 55400
+    },
+    {
+      "epoch": 5.97843665768194,
+      "grad_norm": 0.6874292492866516,
+      "learning_rate": 0.00024169670804101456,
+      "loss": 3.3466,
+      "step": 55450
+    },
+    {
+      "epoch": 5.9838274932614555,
+      "grad_norm": 0.7599024176597595,
+      "learning_rate": 0.0002413729087965461,
+      "loss": 3.3302,
+      "step": 55500
+    },
+    {
+      "epoch": 5.989218328840971,
+      "grad_norm": 0.7153570055961609,
+      "learning_rate": 0.00024104910955207768,
+      "loss": 3.3346,
+      "step": 55550
+    },
+    {
+      "epoch": 5.994609164420485,
+      "grad_norm": 0.7109009027481079,
+      "learning_rate": 0.00024072531030760926,
+      "loss": 3.3506,
+      "step": 55600
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 1.5732123851776123,
+      "learning_rate": 0.00024040151106314084,
+      "loss": 3.3502,
+      "step": 55650
+    },
+    {
+      "epoch": 6.005390835579515,
+      "grad_norm": 0.7368113398551941,
+      "learning_rate": 0.0002400777118186724,
+      "loss": 3.2415,
+      "step": 55700
+    },
+    {
+      "epoch": 6.010781671159029,
+      "grad_norm": 0.6908884644508362,
+      "learning_rate": 0.00023975391257420397,
+      "loss": 3.2472,
+      "step": 55750
+    },
+    {
+      "epoch": 6.0161725067385445,
+      "grad_norm": 0.7259300351142883,
+      "learning_rate": 0.00023943011332973555,
+      "loss": 3.2385,
+      "step": 55800
+    },
+    {
+      "epoch": 6.02156334231806,
+      "grad_norm": 0.7004247307777405,
+      "learning_rate": 0.00023910631408526712,
+      "loss": 3.2465,
+      "step": 55850
+    },
+    {
+      "epoch": 6.026954177897574,
+      "grad_norm": 0.6835771203041077,
+      "learning_rate": 0.0002387825148407987,
+      "loss": 3.2467,
+      "step": 55900
+    },
+    {
+      "epoch": 6.032345013477089,
+      "grad_norm": 0.7232950925827026,
+      "learning_rate": 0.00023845871559633025,
+      "loss": 3.2426,
+      "step": 55950
+    },
+    {
+      "epoch": 6.037735849056604,
+      "grad_norm": 0.6958901286125183,
+      "learning_rate": 0.0002381349163518618,
+      "loss": 3.2385,
+      "step": 56000
+    },
+    {
+      "epoch": 6.037735849056604,
+      "eval_accuracy": 0.383966593148501,
+      "eval_loss": 3.3858516216278076,
+      "eval_runtime": 183.2712,
+      "eval_samples_per_second": 98.275,
+      "eval_steps_per_second": 6.144,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0431266846361185,
+      "grad_norm": 0.7067846655845642,
+      "learning_rate": 0.00023781111710739338,
+      "loss": 3.2468,
+      "step": 56050
+    },
+    {
+      "epoch": 6.048517520215634,
+      "grad_norm": 0.8302700519561768,
+      "learning_rate": 0.00023748731786292496,
+      "loss": 3.2445,
+      "step": 56100
+    },
+    {
+      "epoch": 6.053908355795148,
+      "grad_norm": 0.7517115473747253,
+      "learning_rate": 0.00023716351861845654,
+      "loss": 3.261,
+      "step": 56150
+    },
+    {
+      "epoch": 6.059299191374663,
+      "grad_norm": 0.7173040509223938,
+      "learning_rate": 0.0002368397193739881,
+      "loss": 3.2433,
+      "step": 56200
+    },
+    {
+      "epoch": 6.064690026954178,
+      "grad_norm": 0.7438792586326599,
+      "learning_rate": 0.0002365159201295197,
+      "loss": 3.25,
+      "step": 56250
+    },
+    {
+      "epoch": 6.070080862533692,
+      "grad_norm": 0.6793874502182007,
+      "learning_rate": 0.00023619212088505127,
+      "loss": 3.2596,
+      "step": 56300
+    },
+    {
+      "epoch": 6.0754716981132075,
+      "grad_norm": 0.7372825145721436,
+      "learning_rate": 0.0002358683216405828,
+      "loss": 3.2565,
+      "step": 56350
+    },
+    {
+      "epoch": 6.080862533692723,
+      "grad_norm": 0.7043129801750183,
+      "learning_rate": 0.00023554452239611437,
+      "loss": 3.2768,
+      "step": 56400
+    },
+    {
+      "epoch": 6.086253369272237,
+      "grad_norm": 0.7136656641960144,
+      "learning_rate": 0.00023522072315164595,
+      "loss": 3.264,
+      "step": 56450
+    },
+    {
+      "epoch": 6.091644204851752,
+      "grad_norm": 0.729091227054596,
+      "learning_rate": 0.00023489692390717752,
+      "loss": 3.2463,
+      "step": 56500
+    },
+    {
+      "epoch": 6.097035040431267,
+      "grad_norm": 0.6749651432037354,
+      "learning_rate": 0.0002345731246627091,
+      "loss": 3.2743,
+      "step": 56550
+    },
+    {
+      "epoch": 6.1024258760107815,
+      "grad_norm": 0.7389279007911682,
+      "learning_rate": 0.00023425580140313005,
+      "loss": 3.2593,
+      "step": 56600
+    },
+    {
+      "epoch": 6.107816711590297,
+      "grad_norm": 0.7013686895370483,
+      "learning_rate": 0.00023393200215866163,
+      "loss": 3.242,
+      "step": 56650
+    },
+    {
+      "epoch": 6.113207547169812,
+      "grad_norm": 0.7369511127471924,
+      "learning_rate": 0.00023360820291419316,
+      "loss": 3.2733,
+      "step": 56700
+    },
+    {
+      "epoch": 6.118598382749326,
+      "grad_norm": 0.6994354724884033,
+      "learning_rate": 0.00023328440366972473,
+      "loss": 3.2821,
+      "step": 56750
+    },
+    {
+      "epoch": 6.123989218328841,
+      "grad_norm": 0.729927659034729,
+      "learning_rate": 0.0002329606044252563,
+      "loss": 3.2708,
+      "step": 56800
+    },
+    {
+      "epoch": 6.129380053908355,
+      "grad_norm": 0.7666236758232117,
+      "learning_rate": 0.0002326368051807879,
+      "loss": 3.2536,
+      "step": 56850
+    },
+    {
+      "epoch": 6.1347708894878705,
+      "grad_norm": 0.7106788158416748,
+      "learning_rate": 0.00023231300593631947,
+      "loss": 3.2691,
+      "step": 56900
+    },
+    {
+      "epoch": 6.140161725067386,
+      "grad_norm": 0.7282108068466187,
+      "learning_rate": 0.00023198920669185104,
+      "loss": 3.2538,
+      "step": 56950
+    },
+    {
+      "epoch": 6.1455525606469,
+      "grad_norm": 0.7123996019363403,
+      "learning_rate": 0.00023166540744738262,
+      "loss": 3.2695,
+      "step": 57000
+    },
+    {
+      "epoch": 6.1455525606469,
+      "eval_accuracy": 0.38406329421865587,
+      "eval_loss": 3.3840785026550293,
+      "eval_runtime": 183.6604,
+      "eval_samples_per_second": 98.067,
+      "eval_steps_per_second": 6.131,
+      "step": 57000
+    },
+    {
+      "epoch": 6.150943396226415,
+      "grad_norm": 0.6930298209190369,
+      "learning_rate": 0.0002313416082029142,
+      "loss": 3.2614,
+      "step": 57050
+    },
+    {
+      "epoch": 6.15633423180593,
+      "grad_norm": 0.7423911094665527,
+      "learning_rate": 0.00023101780895844572,
+      "loss": 3.2686,
+      "step": 57100
+    },
+    {
+      "epoch": 6.1617250673854445,
+      "grad_norm": 0.6822969913482666,
+      "learning_rate": 0.0002306940097139773,
+      "loss": 3.2835,
+      "step": 57150
+    },
+    {
+      "epoch": 6.16711590296496,
+      "grad_norm": 0.7284877300262451,
+      "learning_rate": 0.00023037021046950888,
+      "loss": 3.26,
+      "step": 57200
+    },
+    {
+      "epoch": 6.172506738544475,
+      "grad_norm": 0.7356117367744446,
+      "learning_rate": 0.00023004641122504046,
+      "loss": 3.2594,
+      "step": 57250
+    },
+    {
+      "epoch": 6.177897574123989,
+      "grad_norm": 0.7333625555038452,
+      "learning_rate": 0.00022972261198057203,
+      "loss": 3.2732,
+      "step": 57300
+    },
+    {
+      "epoch": 6.183288409703504,
+      "grad_norm": 0.6783888936042786,
+      "learning_rate": 0.0002293988127361036,
+      "loss": 3.2614,
+      "step": 57350
+    },
+    {
+      "epoch": 6.188679245283019,
+      "grad_norm": 0.710628867149353,
+      "learning_rate": 0.00022907501349163516,
+      "loss": 3.2687,
+      "step": 57400
+    },
+    {
+      "epoch": 6.1940700808625335,
+      "grad_norm": 0.7758549451828003,
+      "learning_rate": 0.00022875121424716674,
+      "loss": 3.265,
+      "step": 57450
+    },
+    {
+      "epoch": 6.199460916442049,
+      "grad_norm": 0.7403174042701721,
+      "learning_rate": 0.00022842741500269832,
+      "loss": 3.2574,
+      "step": 57500
+    },
+    {
+      "epoch": 6.204851752021563,
+      "grad_norm": 0.7618892192840576,
+      "learning_rate": 0.00022810361575822987,
+      "loss": 3.2807,
+      "step": 57550
+    },
+    {
+      "epoch": 6.210242587601078,
+      "grad_norm": 0.7250393033027649,
+      "learning_rate": 0.00022777981651376145,
+      "loss": 3.2661,
+      "step": 57600
+    },
+    {
+      "epoch": 6.215633423180593,
+      "grad_norm": 0.7694400548934937,
+      "learning_rate": 0.00022745601726929302,
+      "loss": 3.283,
+      "step": 57650
+    },
+    {
+      "epoch": 6.2210242587601075,
+      "grad_norm": 0.7738978862762451,
+      "learning_rate": 0.00022713221802482457,
+      "loss": 3.2729,
+      "step": 57700
+    },
+    {
+      "epoch": 6.226415094339623,
+      "grad_norm": 0.7094854116439819,
+      "learning_rate": 0.00022680841878035615,
+      "loss": 3.2647,
+      "step": 57750
+    },
+    {
+      "epoch": 6.231805929919138,
+      "grad_norm": 0.764233410358429,
+      "learning_rate": 0.00022648461953588773,
+      "loss": 3.2597,
+      "step": 57800
+    },
+    {
+      "epoch": 6.237196765498652,
+      "grad_norm": 0.7126579284667969,
+      "learning_rate": 0.0002261608202914193,
+      "loss": 3.2741,
+      "step": 57850
+    },
+    {
+      "epoch": 6.242587601078167,
+      "grad_norm": 0.7248375415802002,
+      "learning_rate": 0.00022583702104695088,
+      "loss": 3.2826,
+      "step": 57900
+    },
+    {
+      "epoch": 6.247978436657682,
+      "grad_norm": 0.7731271386146545,
+      "learning_rate": 0.00022551322180248246,
+      "loss": 3.2761,
+      "step": 57950
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "grad_norm": 0.7758564949035645,
+      "learning_rate": 0.00022518942255801399,
+      "loss": 3.2792,
+      "step": 58000
+    },
+    {
+      "epoch": 6.2533692722371965,
+      "eval_accuracy": 0.38461144803767994,
+      "eval_loss": 3.381291627883911,
+      "eval_runtime": 183.3406,
+      "eval_samples_per_second": 98.238,
+      "eval_steps_per_second": 6.142,
+      "step": 58000
+    },
+    {
+      "epoch": 6.258760107816712,
+      "grad_norm": 0.7030556797981262,
+      "learning_rate": 0.00022486562331354556,
+      "loss": 3.2632,
+      "step": 58050
+    },
+    {
+      "epoch": 6.264150943396227,
+      "grad_norm": 0.7208335399627686,
+      "learning_rate": 0.00022454182406907714,
+      "loss": 3.2758,
+      "step": 58100
+    },
+    {
+      "epoch": 6.269541778975741,
+      "grad_norm": 0.7208561301231384,
+      "learning_rate": 0.00022421802482460872,
+      "loss": 3.2702,
+      "step": 58150
+    },
+    {
+      "epoch": 6.274932614555256,
+      "grad_norm": 0.7258003950119019,
+      "learning_rate": 0.0002238942255801403,
+      "loss": 3.2921,
+      "step": 58200
+    },
+    {
+      "epoch": 6.280323450134771,
+      "grad_norm": 0.7505782246589661,
+      "learning_rate": 0.00022357042633567187,
+      "loss": 3.2719,
+      "step": 58250
+    },
+    {
+      "epoch": 6.285714285714286,
+      "grad_norm": 0.7109439373016357,
+      "learning_rate": 0.00022324662709120345,
+      "loss": 3.2642,
+      "step": 58300
+    },
+    {
+      "epoch": 6.291105121293801,
+      "grad_norm": 0.7011198401451111,
+      "learning_rate": 0.00022292282784673503,
+      "loss": 3.2867,
+      "step": 58350
+    },
+    {
+      "epoch": 6.296495956873315,
+      "grad_norm": 0.7072154879570007,
+      "learning_rate": 0.00022259902860226655,
+      "loss": 3.2703,
+      "step": 58400
+    },
+    {
+      "epoch": 6.30188679245283,
+      "grad_norm": 0.738429605960846,
+      "learning_rate": 0.00022227522935779813,
+      "loss": 3.2956,
+      "step": 58450
+    },
+    {
+      "epoch": 6.307277628032345,
+      "grad_norm": 0.7201331257820129,
+      "learning_rate": 0.0002219514301133297,
+      "loss": 3.2929,
+      "step": 58500
+    },
+    {
+      "epoch": 6.3126684636118595,
+      "grad_norm": 0.7926663160324097,
+      "learning_rate": 0.00022162763086886129,
+      "loss": 3.283,
+      "step": 58550
+    },
+    {
+      "epoch": 6.318059299191375,
+      "grad_norm": 0.7503251433372498,
+      "learning_rate": 0.00022130383162439286,
+      "loss": 3.2879,
+      "step": 58600
+    },
+    {
+      "epoch": 6.32345013477089,
+      "grad_norm": 0.695063591003418,
+      "learning_rate": 0.00022098003237992444,
+      "loss": 3.2842,
+      "step": 58650
+    },
+    {
+      "epoch": 6.328840970350404,
+      "grad_norm": 0.7448843717575073,
+      "learning_rate": 0.00022065623313545602,
+      "loss": 3.2834,
+      "step": 58700
+    },
+    {
+      "epoch": 6.334231805929919,
+      "grad_norm": 0.7408082485198975,
+      "learning_rate": 0.00022033243389098757,
+      "loss": 3.274,
+      "step": 58750
+    },
+    {
+      "epoch": 6.339622641509434,
+      "grad_norm": 0.767091691493988,
+      "learning_rate": 0.00022000863464651915,
+      "loss": 3.2815,
+      "step": 58800
+    },
+    {
+      "epoch": 6.345013477088949,
+      "grad_norm": 0.7382594347000122,
+      "learning_rate": 0.00021969131138694007,
+      "loss": 3.2942,
+      "step": 58850
+    },
+    {
+      "epoch": 6.350404312668464,
+      "grad_norm": 0.7217046618461609,
+      "learning_rate": 0.00021936751214247165,
+      "loss": 3.2801,
+      "step": 58900
+    },
+    {
+      "epoch": 6.355795148247978,
+      "grad_norm": 0.7197146415710449,
+      "learning_rate": 0.00021904371289800323,
+      "loss": 3.2772,
+      "step": 58950
+    },
+    {
+      "epoch": 6.361185983827493,
+      "grad_norm": 0.7674608826637268,
+      "learning_rate": 0.0002187199136535348,
+      "loss": 3.2941,
+      "step": 59000
+    },
+    {
+      "epoch": 6.361185983827493,
+      "eval_accuracy": 0.3850102041359591,
+      "eval_loss": 3.3767459392547607,
+      "eval_runtime": 183.2912,
+      "eval_samples_per_second": 98.264,
+      "eval_steps_per_second": 6.143,
+      "step": 59000
+    },
+    {
+      "epoch": 6.366576819407008,
+      "grad_norm": 0.7492465972900391,
+      "learning_rate": 0.00021839611440906638,
+      "loss": 3.2826,
+      "step": 59050
+    },
+    {
+      "epoch": 6.3719676549865225,
+      "grad_norm": 0.7502055168151855,
+      "learning_rate": 0.00021807231516459793,
+      "loss": 3.2785,
+      "step": 59100
+    },
+    {
+      "epoch": 6.377358490566038,
+      "grad_norm": 0.715414822101593,
+      "learning_rate": 0.00021774851592012948,
+      "loss": 3.2727,
+      "step": 59150
+    },
+    {
+      "epoch": 6.382749326145553,
+      "grad_norm": 0.7577951550483704,
+      "learning_rate": 0.00021742471667566106,
+      "loss": 3.2789,
+      "step": 59200
+    },
+    {
+      "epoch": 6.388140161725067,
+      "grad_norm": 0.7287423610687256,
+      "learning_rate": 0.00021710091743119264,
+      "loss": 3.282,
+      "step": 59250
+    },
+    {
+      "epoch": 6.393530997304582,
+      "grad_norm": 0.7242560982704163,
+      "learning_rate": 0.00021677711818672422,
+      "loss": 3.2832,
+      "step": 59300
+    },
+    {
+      "epoch": 6.398921832884097,
+      "grad_norm": 0.7242677211761475,
+      "learning_rate": 0.0002164533189422558,
+      "loss": 3.2799,
+      "step": 59350
+    },
+    {
+      "epoch": 6.404312668463612,
+      "grad_norm": 0.763663113117218,
+      "learning_rate": 0.00021612951969778734,
+      "loss": 3.3052,
+      "step": 59400
+    },
+    {
+      "epoch": 6.409703504043127,
+      "grad_norm": 0.7806388735771179,
+      "learning_rate": 0.00021580572045331892,
+      "loss": 3.2897,
+      "step": 59450
+    },
+    {
+      "epoch": 6.415094339622642,
+      "grad_norm": 0.7188853025436401,
+      "learning_rate": 0.0002154819212088505,
+      "loss": 3.2766,
+      "step": 59500
+    },
+    {
+      "epoch": 6.420485175202156,
+      "grad_norm": 0.8416811227798462,
+      "learning_rate": 0.00021515812196438208,
+      "loss": 3.275,
+      "step": 59550
+    },
+    {
+      "epoch": 6.425876010781671,
+      "grad_norm": 0.7588880658149719,
+      "learning_rate": 0.00021483432271991363,
+      "loss": 3.2964,
+      "step": 59600
+    },
+    {
+      "epoch": 6.431266846361186,
+      "grad_norm": 0.7670322060585022,
+      "learning_rate": 0.0002145105234754452,
+      "loss": 3.2883,
+      "step": 59650
+    },
+    {
+      "epoch": 6.436657681940701,
+      "grad_norm": 0.786287784576416,
+      "learning_rate": 0.00021418672423097676,
+      "loss": 3.2949,
+      "step": 59700
+    },
+    {
+      "epoch": 6.442048517520216,
+      "grad_norm": 0.7813776135444641,
+      "learning_rate": 0.00021386292498650833,
+      "loss": 3.296,
+      "step": 59750
+    },
+    {
+      "epoch": 6.44743935309973,
+      "grad_norm": 0.7637880444526672,
+      "learning_rate": 0.0002135391257420399,
+      "loss": 3.2848,
+      "step": 59800
+    },
+    {
+      "epoch": 6.452830188679245,
+      "grad_norm": 0.7501291036605835,
+      "learning_rate": 0.0002132153264975715,
+      "loss": 3.2685,
+      "step": 59850
+    },
+    {
+      "epoch": 6.45822102425876,
+      "grad_norm": 0.7429332137107849,
+      "learning_rate": 0.00021289152725310307,
+      "loss": 3.2826,
+      "step": 59900
+    },
+    {
+      "epoch": 6.463611859838275,
+      "grad_norm": 0.7317937016487122,
+      "learning_rate": 0.00021256772800863464,
+      "loss": 3.2717,
+      "step": 59950
+    },
+    {
+      "epoch": 6.46900269541779,
+      "grad_norm": 0.7276502847671509,
+      "learning_rate": 0.0002122439287641662,
+      "loss": 3.2853,
+      "step": 60000
+    },
+    {
+      "epoch": 6.46900269541779,
+      "eval_accuracy": 0.38552250250200437,
+      "eval_loss": 3.373230218887329,
+      "eval_runtime": 183.7089,
+      "eval_samples_per_second": 98.041,
+      "eval_steps_per_second": 6.129,
+      "step": 60000
+    },
+    {
+      "epoch": 6.474393530997305,
+      "grad_norm": 0.7643303275108337,
+      "learning_rate": 0.00021192012951969775,
+      "loss": 3.2783,
+      "step": 60050
+    },
+    {
+      "epoch": 6.479784366576819,
+      "grad_norm": 0.7558977007865906,
+      "learning_rate": 0.00021159633027522932,
+      "loss": 3.2947,
+      "step": 60100
+    },
+    {
+      "epoch": 6.485175202156334,
+      "grad_norm": 0.7364935874938965,
+      "learning_rate": 0.0002112725310307609,
+      "loss": 3.2702,
+      "step": 60150
+    },
+    {
+      "epoch": 6.490566037735849,
+      "grad_norm": 0.7122393250465393,
+      "learning_rate": 0.00021094873178629248,
+      "loss": 3.2841,
+      "step": 60200
+    },
+    {
+      "epoch": 6.495956873315364,
+      "grad_norm": 0.7319197058677673,
+      "learning_rate": 0.00021062493254182406,
+      "loss": 3.2804,
+      "step": 60250
+    },
+    {
+      "epoch": 6.501347708894879,
+      "grad_norm": 0.7408891320228577,
+      "learning_rate": 0.00021030113329735563,
+      "loss": 3.2762,
+      "step": 60300
+    },
+    {
+      "epoch": 6.506738544474393,
+      "grad_norm": 0.7808356285095215,
+      "learning_rate": 0.0002099773340528872,
+      "loss": 3.2618,
+      "step": 60350
+    },
+    {
+      "epoch": 6.512129380053908,
+      "grad_norm": 0.7603101134300232,
+      "learning_rate": 0.0002096535348084188,
+      "loss": 3.2922,
+      "step": 60400
+    },
+    {
+      "epoch": 6.517520215633423,
+      "grad_norm": 0.7897022366523743,
+      "learning_rate": 0.0002093297355639503,
+      "loss": 3.2805,
+      "step": 60450
+    },
+    {
+      "epoch": 6.5229110512129385,
+      "grad_norm": 0.723258912563324,
+      "learning_rate": 0.0002090059363194819,
+      "loss": 3.2764,
+      "step": 60500
+    },
+    {
+      "epoch": 6.528301886792453,
+      "grad_norm": 0.7086988687515259,
+      "learning_rate": 0.00020868213707501347,
+      "loss": 3.3005,
+      "step": 60550
+    },
+    {
+      "epoch": 6.533692722371968,
+      "grad_norm": 0.7663331031799316,
+      "learning_rate": 0.00020835833783054505,
+      "loss": 3.2727,
+      "step": 60600
+    },
+    {
+      "epoch": 6.539083557951482,
+      "grad_norm": 0.7294133901596069,
+      "learning_rate": 0.00020803453858607662,
+      "loss": 3.2892,
+      "step": 60650
+    },
+    {
+      "epoch": 6.544474393530997,
+      "grad_norm": 0.7393129467964172,
+      "learning_rate": 0.0002077107393416082,
+      "loss": 3.2864,
+      "step": 60700
+    },
+    {
+      "epoch": 6.549865229110512,
+      "grad_norm": 0.7316185235977173,
+      "learning_rate": 0.00020738694009713975,
+      "loss": 3.2902,
+      "step": 60750
+    },
+    {
+      "epoch": 6.555256064690027,
+      "grad_norm": 0.7547934055328369,
+      "learning_rate": 0.00020706314085267133,
+      "loss": 3.2853,
+      "step": 60800
+    },
+    {
+      "epoch": 6.560646900269542,
+      "grad_norm": 0.7477808594703674,
+      "learning_rate": 0.00020673934160820288,
+      "loss": 3.2746,
+      "step": 60850
+    },
+    {
+      "epoch": 6.566037735849057,
+      "grad_norm": 0.7314183115959167,
+      "learning_rate": 0.00020641554236373446,
+      "loss": 3.2715,
+      "step": 60900
+    },
+    {
+      "epoch": 6.571428571428571,
+      "grad_norm": 0.7928177714347839,
+      "learning_rate": 0.00020609174311926604,
+      "loss": 3.2892,
+      "step": 60950
+    },
+    {
+      "epoch": 6.576819407008086,
+      "grad_norm": 0.7356839776039124,
+      "learning_rate": 0.0002057679438747976,
+      "loss": 3.2766,
+      "step": 61000
+    },
+    {
+      "epoch": 6.576819407008086,
+      "eval_accuracy": 0.38595983037545767,
+      "eval_loss": 3.3684346675872803,
+      "eval_runtime": 183.2074,
+      "eval_samples_per_second": 98.309,
+      "eval_steps_per_second": 6.146,
+      "step": 61000
+    },
+    {
+      "epoch": 6.5822102425876015,
+      "grad_norm": 0.7141728401184082,
+      "learning_rate": 0.00020544414463032916,
+      "loss": 3.2728,
+      "step": 61050
+    },
+    {
+      "epoch": 6.587601078167116,
+      "grad_norm": 0.7390061020851135,
+      "learning_rate": 0.00020512034538586074,
+      "loss": 3.2807,
+      "step": 61100
+    },
+    {
+      "epoch": 6.592991913746631,
+      "grad_norm": 0.7230907082557678,
+      "learning_rate": 0.00020479654614139232,
+      "loss": 3.2976,
+      "step": 61150
+    },
+    {
+      "epoch": 6.598382749326145,
+      "grad_norm": 0.701111912727356,
+      "learning_rate": 0.0002044727468969239,
+      "loss": 3.294,
+      "step": 61200
+    },
+    {
+      "epoch": 6.60377358490566,
+      "grad_norm": 0.7841452956199646,
+      "learning_rate": 0.00020414894765245547,
+      "loss": 3.2851,
+      "step": 61250
+    },
+    {
+      "epoch": 6.609164420485175,
+      "grad_norm": 0.7554968595504761,
+      "learning_rate": 0.00020382514840798702,
+      "loss": 3.2778,
+      "step": 61300
+    },
+    {
+      "epoch": 6.6145552560646905,
+      "grad_norm": 0.7525690793991089,
+      "learning_rate": 0.0002035013491635186,
+      "loss": 3.29,
+      "step": 61350
+    },
+    {
+      "epoch": 6.619946091644205,
+      "grad_norm": 0.7529729008674622,
+      "learning_rate": 0.00020317754991905015,
+      "loss": 3.2886,
+      "step": 61400
+    },
+    {
+      "epoch": 6.62533692722372,
+      "grad_norm": 0.7427783608436584,
+      "learning_rate": 0.00020285375067458173,
+      "loss": 3.2861,
+      "step": 61450
+    },
+    {
+      "epoch": 6.630727762803234,
+      "grad_norm": 0.7345287203788757,
+      "learning_rate": 0.0002025299514301133,
+      "loss": 3.2816,
+      "step": 61500
+    },
+    {
+      "epoch": 6.636118598382749,
+      "grad_norm": 0.7362512946128845,
+      "learning_rate": 0.00020220615218564489,
+      "loss": 3.2793,
+      "step": 61550
+    },
+    {
+      "epoch": 6.6415094339622645,
+      "grad_norm": 0.7437974214553833,
+      "learning_rate": 0.00020188235294117646,
+      "loss": 3.308,
+      "step": 61600
+    },
+    {
+      "epoch": 6.646900269541779,
+      "grad_norm": 0.7193623781204224,
+      "learning_rate": 0.00020155855369670804,
+      "loss": 3.2643,
+      "step": 61650
+    },
+    {
+      "epoch": 6.652291105121294,
+      "grad_norm": 0.7621803879737854,
+      "learning_rate": 0.00020123475445223956,
+      "loss": 3.2824,
+      "step": 61700
+    },
+    {
+      "epoch": 6.657681940700809,
+      "grad_norm": 0.7773572206497192,
+      "learning_rate": 0.00020091095520777114,
+      "loss": 3.3017,
+      "step": 61750
+    },
+    {
+      "epoch": 6.663072776280323,
+      "grad_norm": 0.7783927917480469,
+      "learning_rate": 0.00020058715596330272,
+      "loss": 3.2839,
+      "step": 61800
+    },
+    {
+      "epoch": 6.668463611859838,
+      "grad_norm": 0.7530717253684998,
+      "learning_rate": 0.0002002633567188343,
+      "loss": 3.2978,
+      "step": 61850
+    },
+    {
+      "epoch": 6.6738544474393535,
+      "grad_norm": 0.7378366589546204,
+      "learning_rate": 0.00019993955747436588,
+      "loss": 3.2849,
+      "step": 61900
+    },
+    {
+      "epoch": 6.679245283018868,
+      "grad_norm": 0.7434169054031372,
+      "learning_rate": 0.00019961575822989745,
+      "loss": 3.2812,
+      "step": 61950
+    },
+    {
+      "epoch": 6.684636118598383,
+      "grad_norm": 0.7313494682312012,
+      "learning_rate": 0.00019929195898542903,
+      "loss": 3.2783,
+      "step": 62000
+    },
+    {
+      "epoch": 6.684636118598383,
+      "eval_accuracy": 0.38604544885105546,
+      "eval_loss": 3.3659772872924805,
+      "eval_runtime": 183.4736,
+      "eval_samples_per_second": 98.167,
+      "eval_steps_per_second": 6.137,
+      "step": 62000
+    },
+    {
+      "epoch": 6.690026954177897,
+      "grad_norm": 0.7627373337745667,
+      "learning_rate": 0.0001989681597409606,
+      "loss": 3.2761,
+      "step": 62050
+    },
+    {
+      "epoch": 6.695417789757412,
+      "grad_norm": 0.7412805557250977,
+      "learning_rate": 0.00019864436049649216,
+      "loss": 3.2823,
+      "step": 62100
+    },
+    {
+      "epoch": 6.7008086253369274,
+      "grad_norm": 0.7850884795188904,
+      "learning_rate": 0.0001983205612520237,
+      "loss": 3.2997,
+      "step": 62150
+    },
+    {
+      "epoch": 6.706199460916442,
+      "grad_norm": 0.7880491614341736,
+      "learning_rate": 0.0001979967620075553,
+      "loss": 3.3074,
+      "step": 62200
+    },
+    {
+      "epoch": 6.711590296495957,
+      "grad_norm": 0.7848995327949524,
+      "learning_rate": 0.00019767296276308686,
+      "loss": 3.2807,
+      "step": 62250
+    },
+    {
+      "epoch": 6.716981132075472,
+      "grad_norm": 0.7386139035224915,
+      "learning_rate": 0.00019734916351861844,
+      "loss": 3.2859,
+      "step": 62300
+    },
+    {
+      "epoch": 6.722371967654986,
+      "grad_norm": 0.7742098569869995,
+      "learning_rate": 0.00019702536427415002,
+      "loss": 3.2881,
+      "step": 62350
+    },
+    {
+      "epoch": 6.727762803234501,
+      "grad_norm": 0.7377965450286865,
+      "learning_rate": 0.0001967015650296816,
+      "loss": 3.2733,
+      "step": 62400
+    },
+    {
+      "epoch": 6.7331536388140165,
+      "grad_norm": 0.7602487802505493,
+      "learning_rate": 0.00019637776578521315,
+      "loss": 3.2771,
+      "step": 62450
+    },
+    {
+      "epoch": 6.738544474393531,
+      "grad_norm": 0.8373870253562927,
+      "learning_rate": 0.00019605396654074473,
+      "loss": 3.2859,
+      "step": 62500
+    },
+    {
+      "epoch": 6.743935309973046,
+      "grad_norm": 0.791016161441803,
+      "learning_rate": 0.0001957301672962763,
+      "loss": 3.286,
+      "step": 62550
+    },
+    {
+      "epoch": 6.74932614555256,
+      "grad_norm": 0.8201221227645874,
+      "learning_rate": 0.00019540636805180785,
+      "loss": 3.2789,
+      "step": 62600
+    },
+    {
+      "epoch": 6.754716981132075,
+      "grad_norm": 0.7430251240730286,
+      "learning_rate": 0.00019508256880733943,
+      "loss": 3.2898,
+      "step": 62650
+    },
+    {
+      "epoch": 6.7601078167115904,
+      "grad_norm": 0.7708198428153992,
+      "learning_rate": 0.000194758769562871,
+      "loss": 3.2805,
+      "step": 62700
+    },
+    {
+      "epoch": 6.765498652291106,
+      "grad_norm": 0.7176512479782104,
+      "learning_rate": 0.00019443497031840256,
+      "loss": 3.2982,
+      "step": 62750
+    },
+    {
+      "epoch": 6.77088948787062,
+      "grad_norm": 0.7348231077194214,
+      "learning_rate": 0.00019411117107393414,
+      "loss": 3.295,
+      "step": 62800
+    },
+    {
+      "epoch": 6.776280323450135,
+      "grad_norm": 0.8298662900924683,
+      "learning_rate": 0.00019378737182946572,
+      "loss": 3.286,
+      "step": 62850
+    },
+    {
+      "epoch": 6.781671159029649,
+      "grad_norm": 0.7712046504020691,
+      "learning_rate": 0.00019347004856988664,
+      "loss": 3.2835,
+      "step": 62900
+    },
+    {
+      "epoch": 6.787061994609164,
+      "grad_norm": 0.7421087622642517,
+      "learning_rate": 0.00019314624932541822,
+      "loss": 3.2838,
+      "step": 62950
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "grad_norm": 0.7551441788673401,
+      "learning_rate": 0.0001928224500809498,
+      "loss": 3.2875,
+      "step": 63000
+    },
+    {
+      "epoch": 6.7924528301886795,
+      "eval_accuracy": 0.3866129428841105,
+      "eval_loss": 3.3614447116851807,
+      "eval_runtime": 183.6202,
+      "eval_samples_per_second": 98.088,
+      "eval_steps_per_second": 6.132,
+      "step": 63000
+    },
+    {
+      "epoch": 6.797843665768194,
+      "grad_norm": 0.7673226594924927,
+      "learning_rate": 0.00019249865083648137,
+      "loss": 3.2742,
+      "step": 63050
+    },
+    {
+      "epoch": 6.803234501347709,
+      "grad_norm": 0.7840644121170044,
+      "learning_rate": 0.00019217485159201292,
+      "loss": 3.3004,
+      "step": 63100
+    },
+    {
+      "epoch": 6.808625336927224,
+      "grad_norm": 0.7589669227600098,
+      "learning_rate": 0.0001918510523475445,
+      "loss": 3.2869,
+      "step": 63150
+    },
+    {
+      "epoch": 6.814016172506738,
+      "grad_norm": 0.7413915395736694,
+      "learning_rate": 0.00019152725310307608,
+      "loss": 3.2838,
+      "step": 63200
+    },
+    {
+      "epoch": 6.819407008086253,
+      "grad_norm": 0.7459420561790466,
+      "learning_rate": 0.00019120345385860766,
+      "loss": 3.2921,
+      "step": 63250
+    },
+    {
+      "epoch": 6.824797843665769,
+      "grad_norm": 0.8218080997467041,
+      "learning_rate": 0.00019087965461413923,
+      "loss": 3.2928,
+      "step": 63300
+    },
+    {
+      "epoch": 6.830188679245283,
+      "grad_norm": 0.7318329215049744,
+      "learning_rate": 0.00019055585536967079,
+      "loss": 3.2823,
+      "step": 63350
+    },
+    {
+      "epoch": 6.835579514824798,
+      "grad_norm": 0.7867323756217957,
+      "learning_rate": 0.00019023205612520234,
+      "loss": 3.2785,
+      "step": 63400
+    },
+    {
+      "epoch": 6.840970350404312,
+      "grad_norm": 0.7573836445808411,
+      "learning_rate": 0.00018990825688073391,
+      "loss": 3.2806,
+      "step": 63450
+    },
+    {
+      "epoch": 6.846361185983827,
+      "grad_norm": 0.7248268127441406,
+      "learning_rate": 0.0001895844576362655,
+      "loss": 3.283,
+      "step": 63500
+    },
+    {
+      "epoch": 6.8517520215633425,
+      "grad_norm": 0.7397661805152893,
+      "learning_rate": 0.00018926065839179707,
+      "loss": 3.3005,
+      "step": 63550
+    },
+    {
+      "epoch": 6.857142857142857,
+      "grad_norm": 0.8155834078788757,
+      "learning_rate": 0.00018893685914732865,
+      "loss": 3.2861,
+      "step": 63600
+    },
+    {
+      "epoch": 6.862533692722372,
+      "grad_norm": 0.7723265290260315,
+      "learning_rate": 0.00018861305990286022,
+      "loss": 3.2891,
+      "step": 63650
+    },
+    {
+      "epoch": 6.867924528301887,
+      "grad_norm": 0.744189977645874,
+      "learning_rate": 0.0001882892606583918,
+      "loss": 3.2657,
+      "step": 63700
+    },
+    {
+      "epoch": 6.873315363881401,
+      "grad_norm": 0.7577908039093018,
+      "learning_rate": 0.00018796546141392333,
+      "loss": 3.2761,
+      "step": 63750
+    },
+    {
+      "epoch": 6.878706199460916,
+      "grad_norm": 0.7840666770935059,
+      "learning_rate": 0.0001876416621694549,
+      "loss": 3.2901,
+      "step": 63800
+    },
+    {
+      "epoch": 6.884097035040432,
+      "grad_norm": 0.7282495498657227,
+      "learning_rate": 0.00018731786292498648,
+      "loss": 3.261,
+      "step": 63850
+    },
+    {
+      "epoch": 6.889487870619946,
+      "grad_norm": 0.7487484812736511,
+      "learning_rate": 0.00018699406368051806,
+      "loss": 3.2759,
+      "step": 63900
+    },
+    {
+      "epoch": 6.894878706199461,
+      "grad_norm": 0.7612893581390381,
+      "learning_rate": 0.00018667026443604964,
+      "loss": 3.2937,
+      "step": 63950
+    },
+    {
+      "epoch": 6.900269541778976,
+      "grad_norm": 0.7528893351554871,
+      "learning_rate": 0.00018634646519158121,
+      "loss": 3.3016,
+      "step": 64000
+    },
+    {
+      "epoch": 6.900269541778976,
+      "eval_accuracy": 0.3869931193385734,
+      "eval_loss": 3.355668306350708,
+      "eval_runtime": 183.1773,
+      "eval_samples_per_second": 98.326,
+      "eval_steps_per_second": 6.147,
+      "step": 64000
+    },
+    {
+      "epoch": 6.90566037735849,
+      "grad_norm": 0.7915266752243042,
+      "learning_rate": 0.0001860226659471128,
+      "loss": 3.2852,
+      "step": 64050
+    },
+    {
+      "epoch": 6.9110512129380055,
+      "grad_norm": 0.7587608098983765,
+      "learning_rate": 0.00018569886670264434,
+      "loss": 3.2826,
+      "step": 64100
+    },
+    {
+      "epoch": 6.916442048517521,
+      "grad_norm": 0.7808550596237183,
+      "learning_rate": 0.00018537506745817592,
+      "loss": 3.2851,
+      "step": 64150
+    },
+    {
+      "epoch": 6.921832884097035,
+      "grad_norm": 0.7648224234580994,
+      "learning_rate": 0.00018505126821370747,
+      "loss": 3.3073,
+      "step": 64200
+    },
+    {
+      "epoch": 6.92722371967655,
+      "grad_norm": 0.7958652377128601,
+      "learning_rate": 0.00018472746896923905,
+      "loss": 3.2773,
+      "step": 64250
+    },
+    {
+      "epoch": 6.932614555256064,
+      "grad_norm": 0.774236798286438,
+      "learning_rate": 0.00018440366972477063,
+      "loss": 3.2858,
+      "step": 64300
+    },
+    {
+      "epoch": 6.938005390835579,
+      "grad_norm": 0.7416276335716248,
+      "learning_rate": 0.0001840798704803022,
+      "loss": 3.275,
+      "step": 64350
+    },
+    {
+      "epoch": 6.943396226415095,
+      "grad_norm": 0.7658240795135498,
+      "learning_rate": 0.00018375607123583378,
+      "loss": 3.2844,
+      "step": 64400
+    },
+    {
+      "epoch": 6.948787061994609,
+      "grad_norm": 0.7507860660552979,
+      "learning_rate": 0.00018343227199136533,
+      "loss": 3.2591,
+      "step": 64450
+    },
+    {
+      "epoch": 6.954177897574124,
+      "grad_norm": 0.7703077793121338,
+      "learning_rate": 0.0001831084727468969,
+      "loss": 3.3019,
+      "step": 64500
+    },
+    {
+      "epoch": 6.959568733153639,
+      "grad_norm": 0.7496325373649597,
+      "learning_rate": 0.0001827846735024285,
+      "loss": 3.2972,
+      "step": 64550
+    },
+    {
+      "epoch": 6.964959568733153,
+      "grad_norm": 0.7644689083099365,
+      "learning_rate": 0.00018246087425796004,
+      "loss": 3.2968,
+      "step": 64600
+    },
+    {
+      "epoch": 6.9703504043126685,
+      "grad_norm": 0.8026238679885864,
+      "learning_rate": 0.00018213707501349162,
+      "loss": 3.3095,
+      "step": 64650
+    },
+    {
+      "epoch": 6.975741239892184,
+      "grad_norm": 0.7411915063858032,
+      "learning_rate": 0.0001818132757690232,
+      "loss": 3.2999,
+      "step": 64700
+    },
+    {
+      "epoch": 6.981132075471698,
+      "grad_norm": 0.7806563377380371,
+      "learning_rate": 0.00018148947652455474,
+      "loss": 3.2871,
+      "step": 64750
+    },
+    {
+      "epoch": 6.986522911051213,
+      "grad_norm": 0.765687108039856,
+      "learning_rate": 0.00018116567728008632,
+      "loss": 3.2639,
+      "step": 64800
+    },
+    {
+      "epoch": 6.991913746630727,
+      "grad_norm": 0.7230010032653809,
+      "learning_rate": 0.0001808418780356179,
+      "loss": 3.2782,
+      "step": 64850
+    },
+    {
+      "epoch": 6.997304582210242,
+      "grad_norm": 0.7597923278808594,
+      "learning_rate": 0.00018052455477603885,
+      "loss": 3.2801,
+      "step": 64900
+    },
+    {
+      "epoch": 7.002695417789758,
+      "grad_norm": 0.779853105545044,
+      "learning_rate": 0.0001802007555315704,
+      "loss": 3.2323,
+      "step": 64950
+    },
+    {
+      "epoch": 7.008086253369272,
+      "grad_norm": 0.8532244563102722,
+      "learning_rate": 0.00017987695628710198,
+      "loss": 3.2097,
+      "step": 65000
+    },
+    {
+      "epoch": 7.008086253369272,
+      "eval_accuracy": 0.3869145633006835,
+      "eval_loss": 3.3596954345703125,
+      "eval_runtime": 183.2523,
+      "eval_samples_per_second": 98.285,
+      "eval_steps_per_second": 6.145,
+      "step": 65000
+    },
+    {
+      "epoch": 7.013477088948787,
+      "grad_norm": 0.7580965161323547,
+      "learning_rate": 0.00017955315704263356,
+      "loss": 3.1977,
+      "step": 65050
+    },
+    {
+      "epoch": 7.018867924528302,
+      "grad_norm": 0.8684035539627075,
+      "learning_rate": 0.0001792293577981651,
+      "loss": 3.2105,
+      "step": 65100
+    },
+    {
+      "epoch": 7.024258760107816,
+      "grad_norm": 0.799548864364624,
+      "learning_rate": 0.00017890555855369668,
+      "loss": 3.2211,
+      "step": 65150
+    },
+    {
+      "epoch": 7.0296495956873315,
+      "grad_norm": 0.8078487515449524,
+      "learning_rate": 0.00017858175930922826,
+      "loss": 3.2,
+      "step": 65200
+    },
+    {
+      "epoch": 7.035040431266847,
+      "grad_norm": 0.7694827318191528,
+      "learning_rate": 0.00017825796006475984,
+      "loss": 3.1929,
+      "step": 65250
+    },
+    {
+      "epoch": 7.040431266846361,
+      "grad_norm": 0.8291086554527283,
+      "learning_rate": 0.00017793416082029142,
+      "loss": 3.2101,
+      "step": 65300
+    },
+    {
+      "epoch": 7.045822102425876,
+      "grad_norm": 0.7600764036178589,
+      "learning_rate": 0.000177610361575823,
+      "loss": 3.1906,
+      "step": 65350
+    },
+    {
+      "epoch": 7.051212938005391,
+      "grad_norm": 0.8322035670280457,
+      "learning_rate": 0.00017728656233135452,
+      "loss": 3.2085,
+      "step": 65400
+    },
+    {
+      "epoch": 7.056603773584905,
+      "grad_norm": 0.7672620415687561,
+      "learning_rate": 0.0001769627630868861,
+      "loss": 3.209,
+      "step": 65450
+    },
+    {
+      "epoch": 7.061994609164421,
+      "grad_norm": 0.7917333841323853,
+      "learning_rate": 0.00017663896384241767,
+      "loss": 3.2056,
+      "step": 65500
+    },
+    {
+      "epoch": 7.067385444743936,
+      "grad_norm": 0.8109930753707886,
+      "learning_rate": 0.00017631516459794925,
+      "loss": 3.2173,
+      "step": 65550
+    },
+    {
+      "epoch": 7.07277628032345,
+      "grad_norm": 0.7816203832626343,
+      "learning_rate": 0.00017599136535348083,
+      "loss": 3.219,
+      "step": 65600
+    },
+    {
+      "epoch": 7.078167115902965,
+      "grad_norm": 0.7387610077857971,
+      "learning_rate": 0.0001756675661090124,
+      "loss": 3.2039,
+      "step": 65650
+    },
+    {
+      "epoch": 7.083557951482479,
+      "grad_norm": 0.7672150135040283,
+      "learning_rate": 0.00017534376686454398,
+      "loss": 3.2141,
+      "step": 65700
+    },
+    {
+      "epoch": 7.0889487870619945,
+      "grad_norm": 0.8220376968383789,
+      "learning_rate": 0.00017501996762007556,
+      "loss": 3.209,
+      "step": 65750
+    },
+    {
+      "epoch": 7.09433962264151,
+      "grad_norm": 0.7742305397987366,
+      "learning_rate": 0.00017469616837560709,
+      "loss": 3.2153,
+      "step": 65800
+    },
+    {
+      "epoch": 7.099730458221024,
+      "grad_norm": 0.7649239897727966,
+      "learning_rate": 0.00017437236913113866,
+      "loss": 3.1942,
+      "step": 65850
+    },
+    {
+      "epoch": 7.105121293800539,
+      "grad_norm": 0.7538276314735413,
+      "learning_rate": 0.00017404856988667024,
+      "loss": 3.2084,
+      "step": 65900
+    },
+    {
+      "epoch": 7.110512129380054,
+      "grad_norm": 0.7935147881507874,
+      "learning_rate": 0.00017372477064220182,
+      "loss": 3.1956,
+      "step": 65950
+    },
+    {
+      "epoch": 7.115902964959568,
+      "grad_norm": 0.7874711751937866,
+      "learning_rate": 0.0001734009713977334,
+      "loss": 3.2227,
+      "step": 66000
+    },
+    {
+      "epoch": 7.115902964959568,
+      "eval_accuracy": 0.38758962369267486,
+      "eval_loss": 3.360151767730713,
+      "eval_runtime": 183.4915,
+      "eval_samples_per_second": 98.157,
+      "eval_steps_per_second": 6.137,
+      "step": 66000
+    },
+    {
+      "epoch": 7.121293800539084,
+      "grad_norm": 0.7753985524177551,
+      "learning_rate": 0.00017307717215326497,
+      "loss": 3.2099,
+      "step": 66050
+    },
+    {
+      "epoch": 7.126684636118599,
+      "grad_norm": 0.7859460115432739,
+      "learning_rate": 0.00017275337290879655,
+      "loss": 3.2137,
+      "step": 66100
+    },
+    {
+      "epoch": 7.132075471698113,
+      "grad_norm": 0.8107000589370728,
+      "learning_rate": 0.0001724295736643281,
+      "loss": 3.2219,
+      "step": 66150
+    },
+    {
+      "epoch": 7.137466307277628,
+      "grad_norm": 0.7623615264892578,
+      "learning_rate": 0.00017210577441985968,
+      "loss": 3.207,
+      "step": 66200
+    },
+    {
+      "epoch": 7.142857142857143,
+      "grad_norm": 0.7917633056640625,
+      "learning_rate": 0.00017178197517539123,
+      "loss": 3.2271,
+      "step": 66250
+    },
+    {
+      "epoch": 7.1482479784366575,
+      "grad_norm": 0.795867919921875,
+      "learning_rate": 0.0001714581759309228,
+      "loss": 3.2088,
+      "step": 66300
+    },
+    {
+      "epoch": 7.153638814016173,
+      "grad_norm": 0.7880429625511169,
+      "learning_rate": 0.00017113437668645439,
+      "loss": 3.2149,
+      "step": 66350
+    },
+    {
+      "epoch": 7.159029649595688,
+      "grad_norm": 0.7899115085601807,
+      "learning_rate": 0.00017081057744198596,
+      "loss": 3.2151,
+      "step": 66400
+    },
+    {
+      "epoch": 7.164420485175202,
+      "grad_norm": 0.7710476517677307,
+      "learning_rate": 0.00017048677819751751,
+      "loss": 3.2245,
+      "step": 66450
+    },
+    {
+      "epoch": 7.169811320754717,
+      "grad_norm": 0.7943754196166992,
+      "learning_rate": 0.0001701629789530491,
+      "loss": 3.212,
+      "step": 66500
+    },
+    {
+      "epoch": 7.175202156334231,
+      "grad_norm": 0.8056839108467102,
+      "learning_rate": 0.00016983917970858067,
+      "loss": 3.2293,
+      "step": 66550
+    },
+    {
+      "epoch": 7.180592991913747,
+      "grad_norm": 0.7769573330879211,
+      "learning_rate": 0.00016951538046411225,
+      "loss": 3.2127,
+      "step": 66600
+    },
+    {
+      "epoch": 7.185983827493262,
+      "grad_norm": 0.7568739652633667,
+      "learning_rate": 0.0001691915812196438,
+      "loss": 3.2276,
+      "step": 66650
+    },
+    {
+      "epoch": 7.191374663072776,
+      "grad_norm": 0.8146957755088806,
+      "learning_rate": 0.00016886778197517538,
+      "loss": 3.2199,
+      "step": 66700
+    },
+    {
+      "epoch": 7.196765498652291,
+      "grad_norm": 0.762374222278595,
+      "learning_rate": 0.00016854398273070693,
+      "loss": 3.2275,
+      "step": 66750
+    },
+    {
+      "epoch": 7.202156334231806,
+      "grad_norm": 0.755624532699585,
+      "learning_rate": 0.0001682201834862385,
+      "loss": 3.203,
+      "step": 66800
+    },
+    {
+      "epoch": 7.2075471698113205,
+      "grad_norm": 0.8084667325019836,
+      "learning_rate": 0.00016789638424177008,
+      "loss": 3.2187,
+      "step": 66850
+    },
+    {
+      "epoch": 7.212938005390836,
+      "grad_norm": 0.7930940985679626,
+      "learning_rate": 0.00016757906098219103,
+      "loss": 3.238,
+      "step": 66900
+    },
+    {
+      "epoch": 7.218328840970351,
+      "grad_norm": 0.7964998483657837,
+      "learning_rate": 0.0001672552617377226,
+      "loss": 3.2275,
+      "step": 66950
+    },
+    {
+      "epoch": 7.223719676549865,
+      "grad_norm": 0.821081817150116,
+      "learning_rate": 0.00016693146249325416,
+      "loss": 3.2331,
+      "step": 67000
+    },
+    {
+      "epoch": 7.223719676549865,
+      "eval_accuracy": 0.3874663026650054,
+      "eval_loss": 3.3561620712280273,
+      "eval_runtime": 183.0972,
+      "eval_samples_per_second": 98.369,
+      "eval_steps_per_second": 6.15,
+      "step": 67000
+    },
+    {
+      "epoch": 7.22911051212938,
+      "grad_norm": 0.7554039359092712,
+      "learning_rate": 0.00016660766324878574,
+      "loss": 3.2252,
+      "step": 67050
+    },
+    {
+      "epoch": 7.234501347708895,
+      "grad_norm": 0.7950193285942078,
+      "learning_rate": 0.0001662838640043173,
+      "loss": 3.2301,
+      "step": 67100
+    },
+    {
+      "epoch": 7.2398921832884096,
+      "grad_norm": 0.7405107617378235,
+      "learning_rate": 0.00016596006475984887,
+      "loss": 3.2306,
+      "step": 67150
+    },
+    {
+      "epoch": 7.245283018867925,
+      "grad_norm": 0.8049962520599365,
+      "learning_rate": 0.00016563626551538045,
+      "loss": 3.2192,
+      "step": 67200
+    },
+    {
+      "epoch": 7.250673854447439,
+      "grad_norm": 0.7890737056732178,
+      "learning_rate": 0.00016531246627091202,
+      "loss": 3.2179,
+      "step": 67250
+    },
+    {
+      "epoch": 7.256064690026954,
+      "grad_norm": 0.7825189828872681,
+      "learning_rate": 0.0001649886670264436,
+      "loss": 3.2354,
+      "step": 67300
+    },
+    {
+      "epoch": 7.261455525606469,
+      "grad_norm": 0.7966928482055664,
+      "learning_rate": 0.00016466486778197518,
+      "loss": 3.2238,
+      "step": 67350
+    },
+    {
+      "epoch": 7.2668463611859835,
+      "grad_norm": 0.8031637072563171,
+      "learning_rate": 0.0001643410685375067,
+      "loss": 3.2312,
+      "step": 67400
+    },
+    {
+      "epoch": 7.272237196765499,
+      "grad_norm": 0.7795936465263367,
+      "learning_rate": 0.00016401726929303828,
+      "loss": 3.2421,
+      "step": 67450
+    },
+    {
+      "epoch": 7.277628032345014,
+      "grad_norm": 0.8232036828994751,
+      "learning_rate": 0.00016369347004856986,
+      "loss": 3.2339,
+      "step": 67500
+    },
+    {
+      "epoch": 7.283018867924528,
+      "grad_norm": 0.849933385848999,
+      "learning_rate": 0.00016336967080410143,
+      "loss": 3.2256,
+      "step": 67550
+    },
+    {
+      "epoch": 7.288409703504043,
+      "grad_norm": 0.7606790065765381,
+      "learning_rate": 0.000163045871559633,
+      "loss": 3.2491,
+      "step": 67600
+    },
+    {
+      "epoch": 7.293800539083558,
+      "grad_norm": 0.8448355197906494,
+      "learning_rate": 0.0001627220723151646,
+      "loss": 3.2314,
+      "step": 67650
+    },
+    {
+      "epoch": 7.2991913746630726,
+      "grad_norm": 0.7807988524436951,
+      "learning_rate": 0.00016239827307069617,
+      "loss": 3.2309,
+      "step": 67700
+    },
+    {
+      "epoch": 7.304582210242588,
+      "grad_norm": 0.7915521264076233,
+      "learning_rate": 0.00016207447382622775,
+      "loss": 3.2337,
+      "step": 67750
+    },
+    {
+      "epoch": 7.309973045822103,
+      "grad_norm": 0.7725659012794495,
+      "learning_rate": 0.00016175067458175932,
+      "loss": 3.2142,
+      "step": 67800
+    },
+    {
+      "epoch": 7.315363881401617,
+      "grad_norm": 0.8202341198921204,
+      "learning_rate": 0.00016142687533729085,
+      "loss": 3.2542,
+      "step": 67850
+    },
+    {
+      "epoch": 7.320754716981132,
+      "grad_norm": 0.8172433972358704,
+      "learning_rate": 0.00016110307609282242,
+      "loss": 3.2319,
+      "step": 67900
+    },
+    {
+      "epoch": 7.3261455525606465,
+      "grad_norm": 0.8626977801322937,
+      "learning_rate": 0.000160779276848354,
+      "loss": 3.2357,
+      "step": 67950
+    },
+    {
+      "epoch": 7.331536388140162,
+      "grad_norm": 0.8162325620651245,
+      "learning_rate": 0.00016045547760388558,
+      "loss": 3.2154,
+      "step": 68000
+    },
+    {
+      "epoch": 7.331536388140162,
+      "eval_accuracy": 0.3880209756572984,
+      "eval_loss": 3.354207754135132,
+      "eval_runtime": 184.4398,
+      "eval_samples_per_second": 97.652,
+      "eval_steps_per_second": 6.105,
+      "step": 68000
+    },
+    {
+      "epoch": 7.336927223719677,
+      "grad_norm": 0.839087188243866,
+      "learning_rate": 0.00016013167835941716,
+      "loss": 3.2344,
+      "step": 68050
+    },
+    {
+      "epoch": 7.342318059299191,
+      "grad_norm": 0.8025051355361938,
+      "learning_rate": 0.00015980787911494873,
+      "loss": 3.2419,
+      "step": 68100
+    },
+    {
+      "epoch": 7.347708894878706,
+      "grad_norm": 0.7863566279411316,
+      "learning_rate": 0.00015948407987048029,
+      "loss": 3.2234,
+      "step": 68150
+    },
+    {
+      "epoch": 7.353099730458221,
+      "grad_norm": 0.8169180154800415,
+      "learning_rate": 0.00015916028062601186,
+      "loss": 3.2446,
+      "step": 68200
+    },
+    {
+      "epoch": 7.3584905660377355,
+      "grad_norm": 0.8338359594345093,
+      "learning_rate": 0.00015883648138154344,
+      "loss": 3.2495,
+      "step": 68250
+    },
+    {
+      "epoch": 7.363881401617251,
+      "grad_norm": 0.804298996925354,
+      "learning_rate": 0.000158512682137075,
+      "loss": 3.2358,
+      "step": 68300
+    },
+    {
+      "epoch": 7.369272237196766,
+      "grad_norm": 0.7948055863380432,
+      "learning_rate": 0.00015818888289260657,
+      "loss": 3.236,
+      "step": 68350
+    },
+    {
+      "epoch": 7.37466307277628,
+      "grad_norm": 0.8031824827194214,
+      "learning_rate": 0.00015786508364813815,
+      "loss": 3.2276,
+      "step": 68400
+    },
+    {
+      "epoch": 7.380053908355795,
+      "grad_norm": 0.841023862361908,
+      "learning_rate": 0.0001575412844036697,
+      "loss": 3.212,
+      "step": 68450
+    },
+    {
+      "epoch": 7.38544474393531,
+      "grad_norm": 0.7824974656105042,
+      "learning_rate": 0.00015721748515920127,
+      "loss": 3.2158,
+      "step": 68500
+    },
+    {
+      "epoch": 7.390835579514825,
+      "grad_norm": 0.7631108164787292,
+      "learning_rate": 0.00015689368591473285,
+      "loss": 3.228,
+      "step": 68550
+    },
+    {
+      "epoch": 7.39622641509434,
+      "grad_norm": 0.7724801301956177,
+      "learning_rate": 0.00015656988667026443,
+      "loss": 3.244,
+      "step": 68600
+    },
+    {
+      "epoch": 7.401617250673855,
+      "grad_norm": 0.8289656639099121,
+      "learning_rate": 0.00015625256341068536,
+      "loss": 3.2166,
+      "step": 68650
+    },
+    {
+      "epoch": 7.407008086253369,
+      "grad_norm": 0.8198394179344177,
+      "learning_rate": 0.00015592876416621693,
+      "loss": 3.2208,
+      "step": 68700
+    },
+    {
+      "epoch": 7.412398921832884,
+      "grad_norm": 0.800442636013031,
+      "learning_rate": 0.0001556049649217485,
+      "loss": 3.2256,
+      "step": 68750
+    },
+    {
+      "epoch": 7.4177897574123985,
+      "grad_norm": 0.8246229887008667,
+      "learning_rate": 0.00015528116567728006,
+      "loss": 3.236,
+      "step": 68800
+    },
+    {
+      "epoch": 7.423180592991914,
+      "grad_norm": 0.7899819612503052,
+      "learning_rate": 0.00015495736643281164,
+      "loss": 3.2202,
+      "step": 68850
+    },
+    {
+      "epoch": 7.428571428571429,
+      "grad_norm": 0.8275824785232544,
+      "learning_rate": 0.00015463356718834322,
+      "loss": 3.2353,
+      "step": 68900
+    },
+    {
+      "epoch": 7.433962264150943,
+      "grad_norm": 0.8206462860107422,
+      "learning_rate": 0.0001543097679438748,
+      "loss": 3.2267,
+      "step": 68950
+    },
+    {
+      "epoch": 7.439353099730458,
+      "grad_norm": 0.801008403301239,
+      "learning_rate": 0.00015398596869940637,
+      "loss": 3.2493,
+      "step": 69000
+    },
+    {
+      "epoch": 7.439353099730458,
+      "eval_accuracy": 0.3881384294290258,
+      "eval_loss": 3.3499844074249268,
+      "eval_runtime": 184.0778,
+      "eval_samples_per_second": 97.845,
+      "eval_steps_per_second": 6.117,
+      "step": 69000
+    },
+    {
+      "epoch": 7.444743935309973,
+      "grad_norm": 0.8544893860816956,
+      "learning_rate": 0.00015366216945493792,
+      "loss": 3.2281,
+      "step": 69050
+    },
+    {
+      "epoch": 7.450134770889488,
+      "grad_norm": 0.7715708613395691,
+      "learning_rate": 0.00015333837021046947,
+      "loss": 3.2202,
+      "step": 69100
+    },
+    {
+      "epoch": 7.455525606469003,
+      "grad_norm": 0.8388634324073792,
+      "learning_rate": 0.00015301457096600105,
+      "loss": 3.2036,
+      "step": 69150
+    },
+    {
+      "epoch": 7.460916442048518,
+      "grad_norm": 0.8142986297607422,
+      "learning_rate": 0.00015269077172153263,
+      "loss": 3.2466,
+      "step": 69200
+    },
+    {
+      "epoch": 7.466307277628032,
+      "grad_norm": 0.7704258561134338,
+      "learning_rate": 0.0001523669724770642,
+      "loss": 3.2324,
+      "step": 69250
+    },
+    {
+      "epoch": 7.471698113207547,
+      "grad_norm": 0.7720257639884949,
+      "learning_rate": 0.00015204317323259578,
+      "loss": 3.2207,
+      "step": 69300
+    },
+    {
+      "epoch": 7.4770889487870615,
+      "grad_norm": 0.8082609176635742,
+      "learning_rate": 0.00015171937398812736,
+      "loss": 3.2417,
+      "step": 69350
+    },
+    {
+      "epoch": 7.482479784366577,
+      "grad_norm": 0.8061975240707397,
+      "learning_rate": 0.00015139557474365894,
+      "loss": 3.2422,
+      "step": 69400
+    },
+    {
+      "epoch": 7.487870619946092,
+      "grad_norm": 0.7687536478042603,
+      "learning_rate": 0.00015107177549919046,
+      "loss": 3.2369,
+      "step": 69450
+    },
+    {
+      "epoch": 7.493261455525606,
+      "grad_norm": 0.8298164010047913,
+      "learning_rate": 0.00015074797625472204,
+      "loss": 3.2256,
+      "step": 69500
+    },
+    {
+      "epoch": 7.498652291105121,
+      "grad_norm": 0.8489644527435303,
+      "learning_rate": 0.00015042417701025362,
+      "loss": 3.2311,
+      "step": 69550
+    },
+    {
+      "epoch": 7.504043126684636,
+      "grad_norm": 0.8438618183135986,
+      "learning_rate": 0.0001501003777657852,
+      "loss": 3.2258,
+      "step": 69600
+    },
+    {
+      "epoch": 7.509433962264151,
+      "grad_norm": 0.7734254598617554,
+      "learning_rate": 0.00014977657852131677,
+      "loss": 3.2298,
+      "step": 69650
+    },
+    {
+      "epoch": 7.514824797843666,
+      "grad_norm": 0.7958167195320129,
+      "learning_rate": 0.00014945277927684835,
+      "loss": 3.2405,
+      "step": 69700
+    },
+    {
+      "epoch": 7.520215633423181,
+      "grad_norm": 0.7930517196655273,
+      "learning_rate": 0.00014912898003237993,
+      "loss": 3.2311,
+      "step": 69750
+    },
+    {
+      "epoch": 7.525606469002695,
+      "grad_norm": 0.7839870452880859,
+      "learning_rate": 0.00014880518078791148,
+      "loss": 3.2469,
+      "step": 69800
+    },
+    {
+      "epoch": 7.53099730458221,
+      "grad_norm": 0.8096862435340881,
+      "learning_rate": 0.00014848138154344306,
+      "loss": 3.2198,
+      "step": 69850
+    },
+    {
+      "epoch": 7.536388140161725,
+      "grad_norm": 0.7993474006652832,
+      "learning_rate": 0.00014815758229897463,
+      "loss": 3.2288,
+      "step": 69900
+    },
+    {
+      "epoch": 7.54177897574124,
+      "grad_norm": 0.807650625705719,
+      "learning_rate": 0.0001478337830545062,
+      "loss": 3.2375,
+      "step": 69950
+    },
+    {
+      "epoch": 7.547169811320755,
+      "grad_norm": 0.8408615589141846,
+      "learning_rate": 0.00014750998381003776,
+      "loss": 3.2359,
+      "step": 70000
+    },
+    {
+      "epoch": 7.547169811320755,
+      "eval_accuracy": 0.38866626515802855,
+      "eval_loss": 3.346017360687256,
+      "eval_runtime": 183.2245,
+      "eval_samples_per_second": 98.3,
+      "eval_steps_per_second": 6.145,
+      "step": 70000
+    },
+    {
+      "epoch": 7.55256064690027,
+      "grad_norm": 0.8522095084190369,
+      "learning_rate": 0.00014718618456556934,
+      "loss": 3.2147,
+      "step": 70050
+    },
+    {
+      "epoch": 7.557951482479784,
+      "grad_norm": 0.7832167744636536,
+      "learning_rate": 0.00014686238532110092,
+      "loss": 3.2362,
+      "step": 70100
+    },
+    {
+      "epoch": 7.563342318059299,
+      "grad_norm": 0.8049901127815247,
+      "learning_rate": 0.00014653858607663247,
+      "loss": 3.2341,
+      "step": 70150
+    },
+    {
+      "epoch": 7.568733153638814,
+      "grad_norm": 0.7650406360626221,
+      "learning_rate": 0.00014621478683216405,
+      "loss": 3.2338,
+      "step": 70200
+    },
+    {
+      "epoch": 7.574123989218329,
+      "grad_norm": 0.7782807946205139,
+      "learning_rate": 0.00014589098758769562,
+      "loss": 3.2282,
+      "step": 70250
+    },
+    {
+      "epoch": 7.579514824797844,
+      "grad_norm": 0.786255955696106,
+      "learning_rate": 0.00014556718834322717,
+      "loss": 3.2427,
+      "step": 70300
+    },
+    {
+      "epoch": 7.584905660377358,
+      "grad_norm": 0.8183979988098145,
+      "learning_rate": 0.00014524338909875875,
+      "loss": 3.216,
+      "step": 70350
+    },
+    {
+      "epoch": 7.590296495956873,
+      "grad_norm": 0.7959901690483093,
+      "learning_rate": 0.00014491958985429033,
+      "loss": 3.222,
+      "step": 70400
+    },
+    {
+      "epoch": 7.595687331536388,
+      "grad_norm": 0.8111319541931152,
+      "learning_rate": 0.00014459579060982188,
+      "loss": 3.2284,
+      "step": 70450
+    },
+    {
+      "epoch": 7.601078167115903,
+      "grad_norm": 0.7604181170463562,
+      "learning_rate": 0.00014427199136535346,
+      "loss": 3.2437,
+      "step": 70500
+    },
+    {
+      "epoch": 7.606469002695418,
+      "grad_norm": 0.8370922207832336,
+      "learning_rate": 0.00014394819212088504,
+      "loss": 3.229,
+      "step": 70550
+    },
+    {
+      "epoch": 7.611859838274933,
+      "grad_norm": 0.7989806532859802,
+      "learning_rate": 0.0001436243928764166,
+      "loss": 3.2536,
+      "step": 70600
+    },
+    {
+      "epoch": 7.617250673854447,
+      "grad_norm": 0.8017361164093018,
+      "learning_rate": 0.00014330059363194816,
+      "loss": 3.2267,
+      "step": 70650
+    },
+    {
+      "epoch": 7.622641509433962,
+      "grad_norm": 0.8466594815254211,
+      "learning_rate": 0.00014297679438747974,
+      "loss": 3.2316,
+      "step": 70700
+    },
+    {
+      "epoch": 7.628032345013477,
+      "grad_norm": 0.8554249405860901,
+      "learning_rate": 0.00014265299514301132,
+      "loss": 3.2238,
+      "step": 70750
+    },
+    {
+      "epoch": 7.633423180592992,
+      "grad_norm": 0.8056179285049438,
+      "learning_rate": 0.0001423291958985429,
+      "loss": 3.2263,
+      "step": 70800
+    },
+    {
+      "epoch": 7.638814016172507,
+      "grad_norm": 0.8345566987991333,
+      "learning_rate": 0.00014200539665407445,
+      "loss": 3.2331,
+      "step": 70850
+    },
+    {
+      "epoch": 7.644204851752022,
+      "grad_norm": 0.8629750609397888,
+      "learning_rate": 0.00014168159740960602,
+      "loss": 3.2288,
+      "step": 70900
+    },
+    {
+      "epoch": 7.649595687331536,
+      "grad_norm": 0.8489576578140259,
+      "learning_rate": 0.0001413577981651376,
+      "loss": 3.2344,
+      "step": 70950
+    },
+    {
+      "epoch": 7.654986522911051,
+      "grad_norm": 0.8354818820953369,
+      "learning_rate": 0.00014103399892066918,
+      "loss": 3.236,
+      "step": 71000
+    },
+    {
+      "epoch": 7.654986522911051,
+      "eval_accuracy": 0.38923256400931766,
+      "eval_loss": 3.341547727584839,
+      "eval_runtime": 183.3377,
+      "eval_samples_per_second": 98.239,
+      "eval_steps_per_second": 6.142,
+      "step": 71000
+    },
+    {
+      "epoch": 7.660377358490566,
+      "grad_norm": 0.8664414882659912,
+      "learning_rate": 0.00014071019967620073,
+      "loss": 3.2377,
+      "step": 71050
+    },
+    {
+      "epoch": 7.665768194070081,
+      "grad_norm": 0.817379891872406,
+      "learning_rate": 0.0001403864004317323,
+      "loss": 3.2197,
+      "step": 71100
+    },
+    {
+      "epoch": 7.671159029649596,
+      "grad_norm": 0.8401409387588501,
+      "learning_rate": 0.00014006260118726389,
+      "loss": 3.2144,
+      "step": 71150
+    },
+    {
+      "epoch": 7.67654986522911,
+      "grad_norm": 0.7922741174697876,
+      "learning_rate": 0.00013973880194279546,
+      "loss": 3.2194,
+      "step": 71200
+    },
+    {
+      "epoch": 7.681940700808625,
+      "grad_norm": 0.7995992302894592,
+      "learning_rate": 0.00013941500269832704,
+      "loss": 3.2291,
+      "step": 71250
+    },
+    {
+      "epoch": 7.6873315363881405,
+      "grad_norm": 0.7867482304573059,
+      "learning_rate": 0.0001390912034538586,
+      "loss": 3.2516,
+      "step": 71300
+    },
+    {
+      "epoch": 7.692722371967655,
+      "grad_norm": 0.7929086685180664,
+      "learning_rate": 0.00013876740420939017,
+      "loss": 3.2399,
+      "step": 71350
+    },
+    {
+      "epoch": 7.69811320754717,
+      "grad_norm": 0.7987889051437378,
+      "learning_rate": 0.00013844360496492175,
+      "loss": 3.2341,
+      "step": 71400
+    },
+    {
+      "epoch": 7.703504043126685,
+      "grad_norm": 0.8319695591926575,
+      "learning_rate": 0.00013811980572045333,
+      "loss": 3.2248,
+      "step": 71450
+    },
+    {
+      "epoch": 7.708894878706199,
+      "grad_norm": 0.8619706034660339,
+      "learning_rate": 0.00013779600647598488,
+      "loss": 3.2284,
+      "step": 71500
+    },
+    {
+      "epoch": 7.714285714285714,
+      "grad_norm": 0.7763728499412537,
+      "learning_rate": 0.00013747220723151645,
+      "loss": 3.2544,
+      "step": 71550
+    },
+    {
+      "epoch": 7.719676549865229,
+      "grad_norm": 0.8085768222808838,
+      "learning_rate": 0.00013714840798704803,
+      "loss": 3.2416,
+      "step": 71600
+    },
+    {
+      "epoch": 7.725067385444744,
+      "grad_norm": 0.8040910363197327,
+      "learning_rate": 0.00013682460874257958,
+      "loss": 3.2494,
+      "step": 71650
+    },
+    {
+      "epoch": 7.730458221024259,
+      "grad_norm": 0.8536854386329651,
+      "learning_rate": 0.00013650080949811116,
+      "loss": 3.2476,
+      "step": 71700
+    },
+    {
+      "epoch": 7.735849056603773,
+      "grad_norm": 0.8234822154045105,
+      "learning_rate": 0.00013617701025364274,
+      "loss": 3.249,
+      "step": 71750
+    },
+    {
+      "epoch": 7.741239892183288,
+      "grad_norm": 0.8197305798530579,
+      "learning_rate": 0.0001358532110091743,
+      "loss": 3.2433,
+      "step": 71800
+    },
+    {
+      "epoch": 7.7466307277628035,
+      "grad_norm": 0.8008885383605957,
+      "learning_rate": 0.00013552941176470587,
+      "loss": 3.2561,
+      "step": 71850
+    },
+    {
+      "epoch": 7.752021563342318,
+      "grad_norm": 0.7963679432868958,
+      "learning_rate": 0.00013520561252023744,
+      "loss": 3.234,
+      "step": 71900
+    },
+    {
+      "epoch": 7.757412398921833,
+      "grad_norm": 0.8011940121650696,
+      "learning_rate": 0.00013488181327576902,
+      "loss": 3.2601,
+      "step": 71950
+    },
+    {
+      "epoch": 7.762803234501348,
+      "grad_norm": 0.7941238284111023,
+      "learning_rate": 0.00013455801403130057,
+      "loss": 3.2325,
+      "step": 72000
+    },
+    {
+      "epoch": 7.762803234501348,
+      "eval_accuracy": 0.38954396318579393,
+      "eval_loss": 3.339073896408081,
+      "eval_runtime": 183.6378,
+      "eval_samples_per_second": 98.079,
+      "eval_steps_per_second": 6.132,
+      "step": 72000
+    },
+    {
+      "epoch": 7.768194070080862,
+      "grad_norm": 0.7575421929359436,
+      "learning_rate": 0.00013423421478683215,
+      "loss": 3.2304,
+      "step": 72050
+    },
+    {
+      "epoch": 7.773584905660377,
+      "grad_norm": 0.7887769341468811,
+      "learning_rate": 0.00013391041554236373,
+      "loss": 3.2183,
+      "step": 72100
+    },
+    {
+      "epoch": 7.7789757412398925,
+      "grad_norm": 0.8231433033943176,
+      "learning_rate": 0.00013358661629789528,
+      "loss": 3.2349,
+      "step": 72150
+    },
+    {
+      "epoch": 7.784366576819407,
+      "grad_norm": 0.8212858438491821,
+      "learning_rate": 0.00013326281705342685,
+      "loss": 3.2358,
+      "step": 72200
+    },
+    {
+      "epoch": 7.789757412398922,
+      "grad_norm": 0.8329764008522034,
+      "learning_rate": 0.00013293901780895843,
+      "loss": 3.2337,
+      "step": 72250
+    },
+    {
+      "epoch": 7.795148247978437,
+      "grad_norm": 0.8348933458328247,
+      "learning_rate": 0.00013261521856449,
+      "loss": 3.229,
+      "step": 72300
+    },
+    {
+      "epoch": 7.800539083557951,
+      "grad_norm": 0.8081076741218567,
+      "learning_rate": 0.00013229141932002156,
+      "loss": 3.2333,
+      "step": 72350
+    },
+    {
+      "epoch": 7.8059299191374665,
+      "grad_norm": 0.7696288824081421,
+      "learning_rate": 0.00013196762007555314,
+      "loss": 3.236,
+      "step": 72400
+    },
+    {
+      "epoch": 7.811320754716981,
+      "grad_norm": 0.872471034526825,
+      "learning_rate": 0.00013164382083108472,
+      "loss": 3.2294,
+      "step": 72450
+    },
+    {
+      "epoch": 7.816711590296496,
+      "grad_norm": 0.8010265827178955,
+      "learning_rate": 0.0001313200215866163,
+      "loss": 3.2407,
+      "step": 72500
+    },
+    {
+      "epoch": 7.822102425876011,
+      "grad_norm": 0.8799073100090027,
+      "learning_rate": 0.00013099622234214784,
+      "loss": 3.2327,
+      "step": 72550
+    },
+    {
+      "epoch": 7.827493261455525,
+      "grad_norm": 0.8439316749572754,
+      "learning_rate": 0.00013067242309767942,
+      "loss": 3.2421,
+      "step": 72600
+    },
+    {
+      "epoch": 7.83288409703504,
+      "grad_norm": 0.7896981835365295,
+      "learning_rate": 0.00013035509983810037,
+      "loss": 3.2483,
+      "step": 72650
+    },
+    {
+      "epoch": 7.8382749326145555,
+      "grad_norm": 0.8126330971717834,
+      "learning_rate": 0.00013003130059363192,
+      "loss": 3.2446,
+      "step": 72700
+    },
+    {
+      "epoch": 7.84366576819407,
+      "grad_norm": 0.897280216217041,
+      "learning_rate": 0.0001297075013491635,
+      "loss": 3.2382,
+      "step": 72750
+    },
+    {
+      "epoch": 7.849056603773585,
+      "grad_norm": 0.9100159406661987,
+      "learning_rate": 0.00012938370210469508,
+      "loss": 3.2621,
+      "step": 72800
+    },
+    {
+      "epoch": 7.8544474393531,
+      "grad_norm": 0.821875274181366,
+      "learning_rate": 0.00012905990286022666,
+      "loss": 3.2318,
+      "step": 72850
+    },
+    {
+      "epoch": 7.859838274932614,
+      "grad_norm": 0.8124797344207764,
+      "learning_rate": 0.0001287361036157582,
+      "loss": 3.2374,
+      "step": 72900
+    },
+    {
+      "epoch": 7.8652291105121295,
+      "grad_norm": 0.7993474006652832,
+      "learning_rate": 0.00012841230437128979,
+      "loss": 3.2589,
+      "step": 72950
+    },
+    {
+      "epoch": 7.870619946091644,
+      "grad_norm": 0.8137656450271606,
+      "learning_rate": 0.00012808850512682136,
+      "loss": 3.2321,
+      "step": 73000
+    },
+    {
+      "epoch": 7.870619946091644,
+      "eval_accuracy": 0.3897977763317286,
+      "eval_loss": 3.334773540496826,
+      "eval_runtime": 183.1358,
+      "eval_samples_per_second": 98.348,
+      "eval_steps_per_second": 6.148,
+      "step": 73000
+    },
+    {
+      "epoch": 7.876010781671159,
+      "grad_norm": 0.8159238696098328,
+      "learning_rate": 0.00012776470588235294,
+      "loss": 3.2303,
+      "step": 73050
+    },
+    {
+      "epoch": 7.881401617250674,
+      "grad_norm": 0.8608067631721497,
+      "learning_rate": 0.0001274409066378845,
+      "loss": 3.2381,
+      "step": 73100
+    },
+    {
+      "epoch": 7.886792452830189,
+      "grad_norm": 0.8166864514350891,
+      "learning_rate": 0.00012711710739341607,
+      "loss": 3.2451,
+      "step": 73150
+    },
+    {
+      "epoch": 7.892183288409703,
+      "grad_norm": 0.7903938889503479,
+      "learning_rate": 0.00012679330814894765,
+      "loss": 3.2227,
+      "step": 73200
+    },
+    {
+      "epoch": 7.8975741239892185,
+      "grad_norm": 0.8242096900939941,
+      "learning_rate": 0.00012646950890447922,
+      "loss": 3.2402,
+      "step": 73250
+    },
+    {
+      "epoch": 7.902964959568733,
+      "grad_norm": 0.8333444595336914,
+      "learning_rate": 0.00012614570966001077,
+      "loss": 3.2401,
+      "step": 73300
+    },
+    {
+      "epoch": 7.908355795148248,
+      "grad_norm": 0.8446034789085388,
+      "learning_rate": 0.00012582191041554235,
+      "loss": 3.2446,
+      "step": 73350
+    },
+    {
+      "epoch": 7.913746630727763,
+      "grad_norm": 0.87595134973526,
+      "learning_rate": 0.00012549811117107393,
+      "loss": 3.2472,
+      "step": 73400
+    },
+    {
+      "epoch": 7.919137466307277,
+      "grad_norm": 0.8153154850006104,
+      "learning_rate": 0.0001251743119266055,
+      "loss": 3.229,
+      "step": 73450
+    },
+    {
+      "epoch": 7.9245283018867925,
+      "grad_norm": 0.8008875250816345,
+      "learning_rate": 0.00012485051268213706,
+      "loss": 3.2488,
+      "step": 73500
+    },
+    {
+      "epoch": 7.929919137466308,
+      "grad_norm": 0.8487942814826965,
+      "learning_rate": 0.00012452671343766864,
+      "loss": 3.2334,
+      "step": 73550
+    },
+    {
+      "epoch": 7.935309973045822,
+      "grad_norm": 0.8568927645683289,
+      "learning_rate": 0.00012420291419320021,
+      "loss": 3.2433,
+      "step": 73600
+    },
+    {
+      "epoch": 7.940700808625337,
+      "grad_norm": 0.8677870035171509,
+      "learning_rate": 0.0001238791149487318,
+      "loss": 3.2246,
+      "step": 73650
+    },
+    {
+      "epoch": 7.946091644204852,
+      "grad_norm": 0.8091750144958496,
+      "learning_rate": 0.00012355531570426334,
+      "loss": 3.2277,
+      "step": 73700
+    },
+    {
+      "epoch": 7.951482479784366,
+      "grad_norm": 0.7613544464111328,
+      "learning_rate": 0.00012323151645979492,
+      "loss": 3.2206,
+      "step": 73750
+    },
+    {
+      "epoch": 7.9568733153638815,
+      "grad_norm": 0.8073203563690186,
+      "learning_rate": 0.0001229077172153265,
+      "loss": 3.201,
+      "step": 73800
+    },
+    {
+      "epoch": 7.962264150943396,
+      "grad_norm": 0.8202934861183167,
+      "learning_rate": 0.00012258391797085805,
+      "loss": 3.2257,
+      "step": 73850
+    },
+    {
+      "epoch": 7.967654986522911,
+      "grad_norm": 0.8023872375488281,
+      "learning_rate": 0.00012226011872638963,
+      "loss": 3.2338,
+      "step": 73900
+    },
+    {
+      "epoch": 7.973045822102426,
+      "grad_norm": 0.834334135055542,
+      "learning_rate": 0.00012193631948192119,
+      "loss": 3.2538,
+      "step": 73950
+    },
+    {
+      "epoch": 7.97843665768194,
+      "grad_norm": 0.8463079333305359,
+      "learning_rate": 0.00012161252023745277,
+      "loss": 3.2455,
+      "step": 74000
+    },
+    {
+      "epoch": 7.97843665768194,
+      "eval_accuracy": 0.3900193195699823,
+      "eval_loss": 3.333192825317383,
+      "eval_runtime": 183.4669,
+      "eval_samples_per_second": 98.17,
+      "eval_steps_per_second": 6.137,
+      "step": 74000
+    },
+    {
+      "epoch": 7.9838274932614555,
+      "grad_norm": 0.8196555376052856,
+      "learning_rate": 0.00012129519697787372,
+      "loss": 3.2393,
+      "step": 74050
+    },
+    {
+      "epoch": 7.989218328840971,
+      "grad_norm": 0.8304194808006287,
+      "learning_rate": 0.00012097139773340527,
+      "loss": 3.2295,
+      "step": 74100
+    },
+    {
+      "epoch": 7.994609164420485,
+      "grad_norm": 0.915878176689148,
+      "learning_rate": 0.00012064759848893685,
+      "loss": 3.2217,
+      "step": 74150
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 1.9139153957366943,
+      "learning_rate": 0.00012032379924446843,
+      "loss": 3.2197,
+      "step": 74200
+    },
+    {
+      "epoch": 8.005390835579515,
+      "grad_norm": 0.7843438982963562,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 3.1613,
+      "step": 74250
+    },
+    {
+      "epoch": 8.01078167115903,
+      "grad_norm": 0.843238353729248,
+      "learning_rate": 0.00011967620075553155,
+      "loss": 3.1593,
+      "step": 74300
+    },
+    {
+      "epoch": 8.016172506738544,
+      "grad_norm": 0.8578649163246155,
+      "learning_rate": 0.00011935240151106313,
+      "loss": 3.1473,
+      "step": 74350
+    },
+    {
+      "epoch": 8.021563342318059,
+      "grad_norm": 0.8399001955986023,
+      "learning_rate": 0.00011902860226659471,
+      "loss": 3.1589,
+      "step": 74400
+    },
+    {
+      "epoch": 8.026954177897574,
+      "grad_norm": 0.7990881204605103,
+      "learning_rate": 0.00011870480302212627,
+      "loss": 3.1612,
+      "step": 74450
+    },
+    {
+      "epoch": 8.032345013477089,
+      "grad_norm": 0.8389109373092651,
+      "learning_rate": 0.00011838100377765784,
+      "loss": 3.1627,
+      "step": 74500
+    },
+    {
+      "epoch": 8.037735849056604,
+      "grad_norm": 0.7998147010803223,
+      "learning_rate": 0.00011805720453318941,
+      "loss": 3.1764,
+      "step": 74550
+    },
+    {
+      "epoch": 8.04312668463612,
+      "grad_norm": 0.7804146409034729,
+      "learning_rate": 0.00011773340528872098,
+      "loss": 3.1643,
+      "step": 74600
+    },
+    {
+      "epoch": 8.048517520215633,
+      "grad_norm": 0.8114757537841797,
+      "learning_rate": 0.00011740960604425256,
+      "loss": 3.165,
+      "step": 74650
+    },
+    {
+      "epoch": 8.053908355795148,
+      "grad_norm": 0.7970903515815735,
+      "learning_rate": 0.00011708580679978412,
+      "loss": 3.1518,
+      "step": 74700
+    },
+    {
+      "epoch": 8.059299191374663,
+      "grad_norm": 0.8087496161460876,
+      "learning_rate": 0.00011676200755531568,
+      "loss": 3.1721,
+      "step": 74750
+    },
+    {
+      "epoch": 8.064690026954178,
+      "grad_norm": 0.8303776979446411,
+      "learning_rate": 0.00011643820831084726,
+      "loss": 3.1657,
+      "step": 74800
+    },
+    {
+      "epoch": 8.070080862533693,
+      "grad_norm": 0.876771092414856,
+      "learning_rate": 0.00011611440906637884,
+      "loss": 3.1638,
+      "step": 74850
+    },
+    {
+      "epoch": 8.075471698113208,
+      "grad_norm": 0.8153787851333618,
+      "learning_rate": 0.00011579060982191042,
+      "loss": 3.1773,
+      "step": 74900
+    },
+    {
+      "epoch": 8.080862533692722,
+      "grad_norm": 0.832485020160675,
+      "learning_rate": 0.00011546681057744197,
+      "loss": 3.1585,
+      "step": 74950
+    },
+    {
+      "epoch": 8.086253369272237,
+      "grad_norm": 0.824671745300293,
+      "learning_rate": 0.00011514301133297355,
+      "loss": 3.1591,
+      "step": 75000
+    },
+    {
+      "epoch": 8.086253369272237,
+      "eval_accuracy": 0.3899736853571002,
+      "eval_loss": 3.337421178817749,
+      "eval_runtime": 183.0721,
+      "eval_samples_per_second": 98.382,
+      "eval_steps_per_second": 6.151,
+      "step": 75000
+    },
+    {
+      "epoch": 8.091644204851752,
+      "grad_norm": 0.8843544125556946,
+      "learning_rate": 0.00011481921208850512,
+      "loss": 3.1677,
+      "step": 75050
+    },
+    {
+      "epoch": 8.097035040431267,
+      "grad_norm": 0.793222188949585,
+      "learning_rate": 0.00011449541284403669,
+      "loss": 3.1573,
+      "step": 75100
+    },
+    {
+      "epoch": 8.102425876010782,
+      "grad_norm": 0.8197263479232788,
+      "learning_rate": 0.00011417161359956825,
+      "loss": 3.1709,
+      "step": 75150
+    },
+    {
+      "epoch": 8.107816711590296,
+      "grad_norm": 0.843536376953125,
+      "learning_rate": 0.00011384781435509983,
+      "loss": 3.1839,
+      "step": 75200
+    },
+    {
+      "epoch": 8.11320754716981,
+      "grad_norm": 0.8424277901649475,
+      "learning_rate": 0.0001135240151106314,
+      "loss": 3.1679,
+      "step": 75250
+    },
+    {
+      "epoch": 8.118598382749326,
+      "grad_norm": 0.8077847957611084,
+      "learning_rate": 0.00011320021586616297,
+      "loss": 3.1791,
+      "step": 75300
+    },
+    {
+      "epoch": 8.123989218328841,
+      "grad_norm": 0.8467236757278442,
+      "learning_rate": 0.00011287641662169454,
+      "loss": 3.1798,
+      "step": 75350
+    },
+    {
+      "epoch": 8.129380053908356,
+      "grad_norm": 0.7999237179756165,
+      "learning_rate": 0.0001125526173772261,
+      "loss": 3.1854,
+      "step": 75400
+    },
+    {
+      "epoch": 8.134770889487871,
+      "grad_norm": 0.8041212558746338,
+      "learning_rate": 0.00011222881813275768,
+      "loss": 3.1708,
+      "step": 75450
+    },
+    {
+      "epoch": 8.140161725067385,
+      "grad_norm": 0.8511107563972473,
+      "learning_rate": 0.00011190501888828925,
+      "loss": 3.1715,
+      "step": 75500
+    },
+    {
+      "epoch": 8.1455525606469,
+      "grad_norm": 0.7946009635925293,
+      "learning_rate": 0.00011158121964382083,
+      "loss": 3.1774,
+      "step": 75550
+    },
+    {
+      "epoch": 8.150943396226415,
+      "grad_norm": 0.8594131469726562,
+      "learning_rate": 0.00011125742039935238,
+      "loss": 3.1793,
+      "step": 75600
+    },
+    {
+      "epoch": 8.15633423180593,
+      "grad_norm": 0.8412104845046997,
+      "learning_rate": 0.00011093362115488396,
+      "loss": 3.1648,
+      "step": 75650
+    },
+    {
+      "epoch": 8.161725067385445,
+      "grad_norm": 0.8513756394386292,
+      "learning_rate": 0.00011060982191041554,
+      "loss": 3.1765,
+      "step": 75700
+    },
+    {
+      "epoch": 8.167115902964959,
+      "grad_norm": 0.8428555727005005,
+      "learning_rate": 0.00011028602266594712,
+      "loss": 3.1756,
+      "step": 75750
+    },
+    {
+      "epoch": 8.172506738544474,
+      "grad_norm": 0.8411218523979187,
+      "learning_rate": 0.00010996222342147867,
+      "loss": 3.1952,
+      "step": 75800
+    },
+    {
+      "epoch": 8.177897574123989,
+      "grad_norm": 0.8924553990364075,
+      "learning_rate": 0.00010963842417701024,
+      "loss": 3.198,
+      "step": 75850
+    },
+    {
+      "epoch": 8.183288409703504,
+      "grad_norm": 0.8435282111167908,
+      "learning_rate": 0.00010931462493254182,
+      "loss": 3.1619,
+      "step": 75900
+    },
+    {
+      "epoch": 8.18867924528302,
+      "grad_norm": 0.7845473289489746,
+      "learning_rate": 0.00010899082568807339,
+      "loss": 3.183,
+      "step": 75950
+    },
+    {
+      "epoch": 8.194070080862534,
+      "grad_norm": 0.828264594078064,
+      "learning_rate": 0.00010866702644360495,
+      "loss": 3.1871,
+      "step": 76000
+    },
+    {
+      "epoch": 8.194070080862534,
+      "eval_accuracy": 0.3901049380455801,
+      "eval_loss": 3.338113307952881,
+      "eval_runtime": 183.2538,
+      "eval_samples_per_second": 98.284,
+      "eval_steps_per_second": 6.144,
+      "step": 76000
+    },
+    {
+      "epoch": 8.199460916442048,
+      "grad_norm": 0.810897171497345,
+      "learning_rate": 0.00010834322719913653,
+      "loss": 3.1834,
+      "step": 76050
+    },
+    {
+      "epoch": 8.204851752021563,
+      "grad_norm": 0.8016768097877502,
+      "learning_rate": 0.00010801942795466809,
+      "loss": 3.1552,
+      "step": 76100
+    },
+    {
+      "epoch": 8.210242587601078,
+      "grad_norm": 0.7954621911048889,
+      "learning_rate": 0.00010769562871019967,
+      "loss": 3.1871,
+      "step": 76150
+    },
+    {
+      "epoch": 8.215633423180593,
+      "grad_norm": 0.8090617060661316,
+      "learning_rate": 0.00010737182946573123,
+      "loss": 3.1678,
+      "step": 76200
+    },
+    {
+      "epoch": 8.221024258760108,
+      "grad_norm": 0.8297533988952637,
+      "learning_rate": 0.0001070480302212628,
+      "loss": 3.1665,
+      "step": 76250
+    },
+    {
+      "epoch": 8.226415094339623,
+      "grad_norm": 0.8397152423858643,
+      "learning_rate": 0.00010672423097679438,
+      "loss": 3.1797,
+      "step": 76300
+    },
+    {
+      "epoch": 8.231805929919137,
+      "grad_norm": 0.873698890209198,
+      "learning_rate": 0.00010640043173232595,
+      "loss": 3.1729,
+      "step": 76350
+    },
+    {
+      "epoch": 8.237196765498652,
+      "grad_norm": 0.7904087901115417,
+      "learning_rate": 0.00010607663248785753,
+      "loss": 3.1668,
+      "step": 76400
+    },
+    {
+      "epoch": 8.242587601078167,
+      "grad_norm": 0.8790299892425537,
+      "learning_rate": 0.00010575283324338908,
+      "loss": 3.1892,
+      "step": 76450
+    },
+    {
+      "epoch": 8.247978436657682,
+      "grad_norm": 0.8573026657104492,
+      "learning_rate": 0.00010542903399892066,
+      "loss": 3.1899,
+      "step": 76500
+    },
+    {
+      "epoch": 8.253369272237197,
+      "grad_norm": 0.7959833741188049,
+      "learning_rate": 0.00010510523475445224,
+      "loss": 3.1932,
+      "step": 76550
+    },
+    {
+      "epoch": 8.25876010781671,
+      "grad_norm": 0.8418940305709839,
+      "learning_rate": 0.0001047814355099838,
+      "loss": 3.1814,
+      "step": 76600
+    },
+    {
+      "epoch": 8.264150943396226,
+      "grad_norm": 0.831942617893219,
+      "learning_rate": 0.00010445763626551537,
+      "loss": 3.1728,
+      "step": 76650
+    },
+    {
+      "epoch": 8.269541778975741,
+      "grad_norm": 0.8003145456314087,
+      "learning_rate": 0.00010413383702104694,
+      "loss": 3.1955,
+      "step": 76700
+    },
+    {
+      "epoch": 8.274932614555256,
+      "grad_norm": 0.803105354309082,
+      "learning_rate": 0.00010381003777657852,
+      "loss": 3.182,
+      "step": 76750
+    },
+    {
+      "epoch": 8.280323450134771,
+      "grad_norm": 0.7960680723190308,
+      "learning_rate": 0.00010348623853211008,
+      "loss": 3.1755,
+      "step": 76800
+    },
+    {
+      "epoch": 8.285714285714286,
+      "grad_norm": 0.820568323135376,
+      "learning_rate": 0.00010316243928764165,
+      "loss": 3.1804,
+      "step": 76850
+    },
+    {
+      "epoch": 8.2911051212938,
+      "grad_norm": 0.8628824353218079,
+      "learning_rate": 0.00010283864004317323,
+      "loss": 3.195,
+      "step": 76900
+    },
+    {
+      "epoch": 8.296495956873315,
+      "grad_norm": 0.83429354429245,
+      "learning_rate": 0.00010251484079870479,
+      "loss": 3.1789,
+      "step": 76950
+    },
+    {
+      "epoch": 8.30188679245283,
+      "grad_norm": 0.8758373856544495,
+      "learning_rate": 0.00010219104155423637,
+      "loss": 3.1747,
+      "step": 77000
+    },
+    {
+      "epoch": 8.30188679245283,
+      "eval_accuracy": 0.39049032983865806,
+      "eval_loss": 3.332897663116455,
+      "eval_runtime": 183.955,
+      "eval_samples_per_second": 97.91,
+      "eval_steps_per_second": 6.121,
+      "step": 77000
+    },
+    {
+      "epoch": 8.307277628032345,
+      "grad_norm": 0.7931265830993652,
+      "learning_rate": 0.00010186724230976793,
+      "loss": 3.1838,
+      "step": 77050
+    },
+    {
+      "epoch": 8.31266846361186,
+      "grad_norm": 0.8602927327156067,
+      "learning_rate": 0.0001015434430652995,
+      "loss": 3.1889,
+      "step": 77100
+    },
+    {
+      "epoch": 8.318059299191376,
+      "grad_norm": 0.8849713206291199,
+      "learning_rate": 0.00010121964382083107,
+      "loss": 3.1786,
+      "step": 77150
+    },
+    {
+      "epoch": 8.323450134770889,
+      "grad_norm": 0.8440691828727722,
+      "learning_rate": 0.00010089584457636265,
+      "loss": 3.1706,
+      "step": 77200
+    },
+    {
+      "epoch": 8.328840970350404,
+      "grad_norm": 0.7911861538887024,
+      "learning_rate": 0.00010057204533189423,
+      "loss": 3.1757,
+      "step": 77250
+    },
+    {
+      "epoch": 8.33423180592992,
+      "grad_norm": 0.8549502491950989,
+      "learning_rate": 0.00010024824608742578,
+      "loss": 3.1784,
+      "step": 77300
+    },
+    {
+      "epoch": 8.339622641509434,
+      "grad_norm": 0.8100817799568176,
+      "learning_rate": 9.992444684295736e-05,
+      "loss": 3.1701,
+      "step": 77350
+    },
+    {
+      "epoch": 8.34501347708895,
+      "grad_norm": 0.8498653769493103,
+      "learning_rate": 9.960064759848894e-05,
+      "loss": 3.2037,
+      "step": 77400
+    },
+    {
+      "epoch": 8.350404312668463,
+      "grad_norm": 0.8423439264297485,
+      "learning_rate": 9.92768483540205e-05,
+      "loss": 3.1777,
+      "step": 77450
+    },
+    {
+      "epoch": 8.355795148247978,
+      "grad_norm": 0.8109570145606995,
+      "learning_rate": 9.895304910955206e-05,
+      "loss": 3.1718,
+      "step": 77500
+    },
+    {
+      "epoch": 8.361185983827493,
+      "grad_norm": 0.8247761726379395,
+      "learning_rate": 9.862924986508364e-05,
+      "loss": 3.1931,
+      "step": 77550
+    },
+    {
+      "epoch": 8.366576819407008,
+      "grad_norm": 0.8085296750068665,
+      "learning_rate": 9.83054506206152e-05,
+      "loss": 3.1852,
+      "step": 77600
+    },
+    {
+      "epoch": 8.371967654986523,
+      "grad_norm": 0.8581119179725647,
+      "learning_rate": 9.798165137614678e-05,
+      "loss": 3.1816,
+      "step": 77650
+    },
+    {
+      "epoch": 8.377358490566039,
+      "grad_norm": 0.8424506187438965,
+      "learning_rate": 9.765785213167835e-05,
+      "loss": 3.1857,
+      "step": 77700
+    },
+    {
+      "epoch": 8.382749326145552,
+      "grad_norm": 0.799414336681366,
+      "learning_rate": 9.733405288720992e-05,
+      "loss": 3.1778,
+      "step": 77750
+    },
+    {
+      "epoch": 8.388140161725067,
+      "grad_norm": 0.8262041211128235,
+      "learning_rate": 9.701025364274149e-05,
+      "loss": 3.1836,
+      "step": 77800
+    },
+    {
+      "epoch": 8.393530997304582,
+      "grad_norm": 0.7753508687019348,
+      "learning_rate": 9.668645439827307e-05,
+      "loss": 3.1693,
+      "step": 77850
+    },
+    {
+      "epoch": 8.398921832884097,
+      "grad_norm": 0.8557844161987305,
+      "learning_rate": 9.636265515380464e-05,
+      "loss": 3.1769,
+      "step": 77900
+    },
+    {
+      "epoch": 8.404312668463612,
+      "grad_norm": 0.8551650047302246,
+      "learning_rate": 9.60388559093362e-05,
+      "loss": 3.1837,
+      "step": 77950
+    },
+    {
+      "epoch": 8.409703504043126,
+      "grad_norm": 0.8227559924125671,
+      "learning_rate": 9.571505666486777e-05,
+      "loss": 3.1932,
+      "step": 78000
+    },
+    {
+      "epoch": 8.409703504043126,
+      "eval_accuracy": 0.39094504217416165,
+      "eval_loss": 3.330205202102661,
+      "eval_runtime": 184.5807,
+      "eval_samples_per_second": 97.578,
+      "eval_steps_per_second": 6.1,
+      "step": 78000
+    },
+    {
+      "epoch": 8.415094339622641,
+      "grad_norm": 0.8216946125030518,
+      "learning_rate": 9.539773340528871e-05,
+      "loss": 3.172,
+      "step": 78050
+    },
+    {
+      "epoch": 8.420485175202156,
+      "grad_norm": 0.8431866765022278,
+      "learning_rate": 9.507393416082027e-05,
+      "loss": 3.1687,
+      "step": 78100
+    },
+    {
+      "epoch": 8.425876010781671,
+      "grad_norm": 0.8359267115592957,
+      "learning_rate": 9.475013491635185e-05,
+      "loss": 3.1923,
+      "step": 78150
+    },
+    {
+      "epoch": 8.431266846361186,
+      "grad_norm": 0.8680286407470703,
+      "learning_rate": 9.442633567188343e-05,
+      "loss": 3.2016,
+      "step": 78200
+    },
+    {
+      "epoch": 8.436657681940702,
+      "grad_norm": 0.8798090219497681,
+      "learning_rate": 9.410253642741498e-05,
+      "loss": 3.177,
+      "step": 78250
+    },
+    {
+      "epoch": 8.442048517520215,
+      "grad_norm": 0.8568891286849976,
+      "learning_rate": 9.377873718294656e-05,
+      "loss": 3.1711,
+      "step": 78300
+    },
+    {
+      "epoch": 8.44743935309973,
+      "grad_norm": 0.819851279258728,
+      "learning_rate": 9.345493793847814e-05,
+      "loss": 3.175,
+      "step": 78350
+    },
+    {
+      "epoch": 8.452830188679245,
+      "grad_norm": 0.8583675622940063,
+      "learning_rate": 9.313113869400971e-05,
+      "loss": 3.1698,
+      "step": 78400
+    },
+    {
+      "epoch": 8.45822102425876,
+      "grad_norm": 0.8463671207427979,
+      "learning_rate": 9.280733944954126e-05,
+      "loss": 3.1797,
+      "step": 78450
+    },
+    {
+      "epoch": 8.463611859838275,
+      "grad_norm": 0.8345748782157898,
+      "learning_rate": 9.248354020507284e-05,
+      "loss": 3.1865,
+      "step": 78500
+    },
+    {
+      "epoch": 8.46900269541779,
+      "grad_norm": 0.8341928720474243,
+      "learning_rate": 9.215974096060442e-05,
+      "loss": 3.1808,
+      "step": 78550
+    },
+    {
+      "epoch": 8.474393530997304,
+      "grad_norm": 0.8836248517036438,
+      "learning_rate": 9.1835941716136e-05,
+      "loss": 3.1896,
+      "step": 78600
+    },
+    {
+      "epoch": 8.479784366576819,
+      "grad_norm": 0.7918772101402283,
+      "learning_rate": 9.151861845655692e-05,
+      "loss": 3.175,
+      "step": 78650
+    },
+    {
+      "epoch": 8.485175202156334,
+      "grad_norm": 0.8527742624282837,
+      "learning_rate": 9.11948192120885e-05,
+      "loss": 3.1909,
+      "step": 78700
+    },
+    {
+      "epoch": 8.49056603773585,
+      "grad_norm": 0.8618723154067993,
+      "learning_rate": 9.087101996762008e-05,
+      "loss": 3.1871,
+      "step": 78750
+    },
+    {
+      "epoch": 8.495956873315365,
+      "grad_norm": 0.8747394680976868,
+      "learning_rate": 9.054722072315163e-05,
+      "loss": 3.1995,
+      "step": 78800
+    },
+    {
+      "epoch": 8.501347708894878,
+      "grad_norm": 0.8599660992622375,
+      "learning_rate": 9.02234214786832e-05,
+      "loss": 3.1712,
+      "step": 78850
+    },
+    {
+      "epoch": 8.506738544474393,
+      "grad_norm": 0.8503178358078003,
+      "learning_rate": 8.989962223421478e-05,
+      "loss": 3.1803,
+      "step": 78900
+    },
+    {
+      "epoch": 8.512129380053908,
+      "grad_norm": 0.7967650294303894,
+      "learning_rate": 8.957582298974635e-05,
+      "loss": 3.1807,
+      "step": 78950
+    },
+    {
+      "epoch": 8.517520215633423,
+      "grad_norm": 0.8530462980270386,
+      "learning_rate": 8.925202374527791e-05,
+      "loss": 3.1805,
+      "step": 79000
+    },
+    {
+      "epoch": 8.517520215633423,
+      "eval_accuracy": 0.3911156272080304,
+      "eval_loss": 3.3279731273651123,
+      "eval_runtime": 183.2589,
+      "eval_samples_per_second": 98.282,
+      "eval_steps_per_second": 6.144,
+      "step": 79000
+    },
+    {
+      "epoch": 8.522911051212938,
+      "grad_norm": 0.843241274356842,
+      "learning_rate": 8.892822450080949e-05,
+      "loss": 3.1917,
+      "step": 79050
+    },
+    {
+      "epoch": 8.528301886792454,
+      "grad_norm": 0.8443961143493652,
+      "learning_rate": 8.860442525634107e-05,
+      "loss": 3.1867,
+      "step": 79100
+    },
+    {
+      "epoch": 8.533692722371967,
+      "grad_norm": 0.8834829330444336,
+      "learning_rate": 8.828062601187263e-05,
+      "loss": 3.1753,
+      "step": 79150
+    },
+    {
+      "epoch": 8.539083557951482,
+      "grad_norm": 0.8263810276985168,
+      "learning_rate": 8.795682676740421e-05,
+      "loss": 3.19,
+      "step": 79200
+    },
+    {
+      "epoch": 8.544474393530997,
+      "grad_norm": 0.8687686324119568,
+      "learning_rate": 8.763302752293577e-05,
+      "loss": 3.1854,
+      "step": 79250
+    },
+    {
+      "epoch": 8.549865229110512,
+      "grad_norm": 0.8647416234016418,
+      "learning_rate": 8.730922827846734e-05,
+      "loss": 3.1802,
+      "step": 79300
+    },
+    {
+      "epoch": 8.555256064690028,
+      "grad_norm": 0.8912370204925537,
+      "learning_rate": 8.698542903399891e-05,
+      "loss": 3.204,
+      "step": 79350
+    },
+    {
+      "epoch": 8.560646900269543,
+      "grad_norm": 0.8945772647857666,
+      "learning_rate": 8.666162978953049e-05,
+      "loss": 3.1781,
+      "step": 79400
+    },
+    {
+      "epoch": 8.566037735849056,
+      "grad_norm": 0.8633682131767273,
+      "learning_rate": 8.633783054506204e-05,
+      "loss": 3.1935,
+      "step": 79450
+    },
+    {
+      "epoch": 8.571428571428571,
+      "grad_norm": 0.8997210264205933,
+      "learning_rate": 8.601403130059362e-05,
+      "loss": 3.1728,
+      "step": 79500
+    },
+    {
+      "epoch": 8.576819407008086,
+      "grad_norm": 0.8614071607589722,
+      "learning_rate": 8.56902320561252e-05,
+      "loss": 3.1743,
+      "step": 79550
+    },
+    {
+      "epoch": 8.582210242587601,
+      "grad_norm": 0.8510245084762573,
+      "learning_rate": 8.536643281165678e-05,
+      "loss": 3.1931,
+      "step": 79600
+    },
+    {
+      "epoch": 8.587601078167117,
+      "grad_norm": 0.8686354756355286,
+      "learning_rate": 8.504263356718833e-05,
+      "loss": 3.1733,
+      "step": 79650
+    },
+    {
+      "epoch": 8.59299191374663,
+      "grad_norm": 0.9134678840637207,
+      "learning_rate": 8.47188343227199e-05,
+      "loss": 3.1793,
+      "step": 79700
+    },
+    {
+      "epoch": 8.598382749326145,
+      "grad_norm": 0.8577871322631836,
+      "learning_rate": 8.439503507825148e-05,
+      "loss": 3.1808,
+      "step": 79750
+    },
+    {
+      "epoch": 8.60377358490566,
+      "grad_norm": 0.9031830430030823,
+      "learning_rate": 8.407123583378305e-05,
+      "loss": 3.1791,
+      "step": 79800
+    },
+    {
+      "epoch": 8.609164420485175,
+      "grad_norm": 0.8463714718818665,
+      "learning_rate": 8.374743658931461e-05,
+      "loss": 3.1771,
+      "step": 79850
+    },
+    {
+      "epoch": 8.61455525606469,
+      "grad_norm": 0.8088541626930237,
+      "learning_rate": 8.342363734484619e-05,
+      "loss": 3.1911,
+      "step": 79900
+    },
+    {
+      "epoch": 8.619946091644206,
+      "grad_norm": 0.8316529393196106,
+      "learning_rate": 8.309983810037775e-05,
+      "loss": 3.1954,
+      "step": 79950
+    },
+    {
+      "epoch": 8.625336927223719,
+      "grad_norm": 0.7918350100517273,
+      "learning_rate": 8.277603885590933e-05,
+      "loss": 3.1777,
+      "step": 80000
+    },
+    {
+      "epoch": 8.625336927223719,
+      "eval_accuracy": 0.3917196286113911,
+      "eval_loss": 3.324557304382324,
+      "eval_runtime": 184.1673,
+      "eval_samples_per_second": 97.797,
+      "eval_steps_per_second": 6.114,
+      "step": 80000
+    },
+    {
+      "epoch": 8.630727762803234,
+      "grad_norm": 0.8560460805892944,
+      "learning_rate": 8.245223961144091e-05,
+      "loss": 3.1742,
+      "step": 80050
+    },
+    {
+      "epoch": 8.63611859838275,
+      "grad_norm": 0.8520783185958862,
+      "learning_rate": 8.212844036697246e-05,
+      "loss": 3.1749,
+      "step": 80100
+    },
+    {
+      "epoch": 8.641509433962264,
+      "grad_norm": 0.8521518111228943,
+      "learning_rate": 8.180464112250404e-05,
+      "loss": 3.1987,
+      "step": 80150
+    },
+    {
+      "epoch": 8.64690026954178,
+      "grad_norm": 0.8230255842208862,
+      "learning_rate": 8.148084187803561e-05,
+      "loss": 3.1779,
+      "step": 80200
+    },
+    {
+      "epoch": 8.652291105121293,
+      "grad_norm": 0.8602603673934937,
+      "learning_rate": 8.115704263356719e-05,
+      "loss": 3.2016,
+      "step": 80250
+    },
+    {
+      "epoch": 8.657681940700808,
+      "grad_norm": 0.8230747580528259,
+      "learning_rate": 8.083324338909874e-05,
+      "loss": 3.1846,
+      "step": 80300
+    },
+    {
+      "epoch": 8.663072776280323,
+      "grad_norm": 0.8413875699043274,
+      "learning_rate": 8.050944414463032e-05,
+      "loss": 3.1856,
+      "step": 80350
+    },
+    {
+      "epoch": 8.668463611859838,
+      "grad_norm": 0.8389858603477478,
+      "learning_rate": 8.01856449001619e-05,
+      "loss": 3.1813,
+      "step": 80400
+    },
+    {
+      "epoch": 8.673854447439354,
+      "grad_norm": 0.8831660747528076,
+      "learning_rate": 7.986184565569347e-05,
+      "loss": 3.1783,
+      "step": 80450
+    },
+    {
+      "epoch": 8.679245283018869,
+      "grad_norm": 0.837313711643219,
+      "learning_rate": 7.953804641122503e-05,
+      "loss": 3.1781,
+      "step": 80500
+    },
+    {
+      "epoch": 8.684636118598382,
+      "grad_norm": 0.9414343237876892,
+      "learning_rate": 7.92142471667566e-05,
+      "loss": 3.1849,
+      "step": 80550
+    },
+    {
+      "epoch": 8.690026954177897,
+      "grad_norm": 0.8817822933197021,
+      "learning_rate": 7.889044792228818e-05,
+      "loss": 3.1773,
+      "step": 80600
+    },
+    {
+      "epoch": 8.695417789757412,
+      "grad_norm": 0.861007809638977,
+      "learning_rate": 7.856664867781974e-05,
+      "loss": 3.1899,
+      "step": 80650
+    },
+    {
+      "epoch": 8.700808625336927,
+      "grad_norm": 0.8453643321990967,
+      "learning_rate": 7.824284943335132e-05,
+      "loss": 3.1932,
+      "step": 80700
+    },
+    {
+      "epoch": 8.706199460916443,
+      "grad_norm": 0.8181665539741516,
+      "learning_rate": 7.791905018888289e-05,
+      "loss": 3.1835,
+      "step": 80750
+    },
+    {
+      "epoch": 8.711590296495956,
+      "grad_norm": 0.8547617793083191,
+      "learning_rate": 7.759525094441445e-05,
+      "loss": 3.154,
+      "step": 80800
+    },
+    {
+      "epoch": 8.716981132075471,
+      "grad_norm": 0.9098490476608276,
+      "learning_rate": 7.727145169994603e-05,
+      "loss": 3.1839,
+      "step": 80850
+    },
+    {
+      "epoch": 8.722371967654986,
+      "grad_norm": 0.8349794149398804,
+      "learning_rate": 7.69476524554776e-05,
+      "loss": 3.1758,
+      "step": 80900
+    },
+    {
+      "epoch": 8.727762803234501,
+      "grad_norm": 0.8465268015861511,
+      "learning_rate": 7.662385321100916e-05,
+      "loss": 3.1752,
+      "step": 80950
+    },
+    {
+      "epoch": 8.733153638814017,
+      "grad_norm": 0.8609336614608765,
+      "learning_rate": 7.630005396654073e-05,
+      "loss": 3.1899,
+      "step": 81000
+    },
+    {
+      "epoch": 8.733153638814017,
+      "eval_accuracy": 0.39213772492370125,
+      "eval_loss": 3.3192524909973145,
+      "eval_runtime": 184.5741,
+      "eval_samples_per_second": 97.581,
+      "eval_steps_per_second": 6.101,
+      "step": 81000
+    },
+    {
+      "epoch": 8.738544474393532,
+      "grad_norm": 0.8602554202079773,
+      "learning_rate": 7.597625472207231e-05,
+      "loss": 3.1996,
+      "step": 81050
+    },
+    {
+      "epoch": 8.743935309973045,
+      "grad_norm": 0.8530620336532593,
+      "learning_rate": 7.565245547760389e-05,
+      "loss": 3.2008,
+      "step": 81100
+    },
+    {
+      "epoch": 8.74932614555256,
+      "grad_norm": 0.8483010530471802,
+      "learning_rate": 7.532865623313544e-05,
+      "loss": 3.1748,
+      "step": 81150
+    },
+    {
+      "epoch": 8.754716981132075,
+      "grad_norm": 0.8151794672012329,
+      "learning_rate": 7.500485698866702e-05,
+      "loss": 3.1952,
+      "step": 81200
+    },
+    {
+      "epoch": 8.76010781671159,
+      "grad_norm": 0.8689604997634888,
+      "learning_rate": 7.46810577441986e-05,
+      "loss": 3.1793,
+      "step": 81250
+    },
+    {
+      "epoch": 8.765498652291106,
+      "grad_norm": 0.8660991191864014,
+      "learning_rate": 7.435725849973016e-05,
+      "loss": 3.1667,
+      "step": 81300
+    },
+    {
+      "epoch": 8.77088948787062,
+      "grad_norm": 0.8648094534873962,
+      "learning_rate": 7.403345925526174e-05,
+      "loss": 3.1908,
+      "step": 81350
+    },
+    {
+      "epoch": 8.776280323450134,
+      "grad_norm": 0.8741868734359741,
+      "learning_rate": 7.37096600107933e-05,
+      "loss": 3.2044,
+      "step": 81400
+    },
+    {
+      "epoch": 8.78167115902965,
+      "grad_norm": 0.8558143377304077,
+      "learning_rate": 7.338586076632488e-05,
+      "loss": 3.183,
+      "step": 81450
+    },
+    {
+      "epoch": 8.787061994609164,
+      "grad_norm": 0.8193835616111755,
+      "learning_rate": 7.306206152185644e-05,
+      "loss": 3.1864,
+      "step": 81500
+    },
+    {
+      "epoch": 8.79245283018868,
+      "grad_norm": 0.842094898223877,
+      "learning_rate": 7.273826227738801e-05,
+      "loss": 3.1863,
+      "step": 81550
+    },
+    {
+      "epoch": 8.797843665768195,
+      "grad_norm": 0.8272750377655029,
+      "learning_rate": 7.241446303291958e-05,
+      "loss": 3.1627,
+      "step": 81600
+    },
+    {
+      "epoch": 8.80323450134771,
+      "grad_norm": 0.8529045581817627,
+      "learning_rate": 7.209066378845115e-05,
+      "loss": 3.1933,
+      "step": 81650
+    },
+    {
+      "epoch": 8.808625336927223,
+      "grad_norm": 0.8626484870910645,
+      "learning_rate": 7.176686454398273e-05,
+      "loss": 3.1954,
+      "step": 81700
+    },
+    {
+      "epoch": 8.814016172506738,
+      "grad_norm": 0.8387072086334229,
+      "learning_rate": 7.144306529951429e-05,
+      "loss": 3.1865,
+      "step": 81750
+    },
+    {
+      "epoch": 8.819407008086253,
+      "grad_norm": 0.8420888185501099,
+      "learning_rate": 7.111926605504587e-05,
+      "loss": 3.1712,
+      "step": 81800
+    },
+    {
+      "epoch": 8.824797843665769,
+      "grad_norm": 0.8461816906929016,
+      "learning_rate": 7.079546681057743e-05,
+      "loss": 3.191,
+      "step": 81850
+    },
+    {
+      "epoch": 8.830188679245284,
+      "grad_norm": 0.8941715955734253,
+      "learning_rate": 7.047166756610901e-05,
+      "loss": 3.1918,
+      "step": 81900
+    },
+    {
+      "epoch": 8.835579514824797,
+      "grad_norm": 0.854226291179657,
+      "learning_rate": 7.014786832164057e-05,
+      "loss": 3.1781,
+      "step": 81950
+    },
+    {
+      "epoch": 8.840970350404312,
+      "grad_norm": 0.8555492758750916,
+      "learning_rate": 6.982406907717215e-05,
+      "loss": 3.196,
+      "step": 82000
+    },
+    {
+      "epoch": 8.840970350404312,
+      "eval_accuracy": 0.39242522046485834,
+      "eval_loss": 3.3160271644592285,
+      "eval_runtime": 183.5831,
+      "eval_samples_per_second": 98.108,
+      "eval_steps_per_second": 6.133,
+      "step": 82000
+    },
+    {
+      "epoch": 8.846361185983827,
+      "grad_norm": 0.8559656739234924,
+      "learning_rate": 6.950026983270372e-05,
+      "loss": 3.1871,
+      "step": 82050
+    },
+    {
+      "epoch": 8.851752021563343,
+      "grad_norm": 0.828024685382843,
+      "learning_rate": 6.91764705882353e-05,
+      "loss": 3.1773,
+      "step": 82100
+    },
+    {
+      "epoch": 8.857142857142858,
+      "grad_norm": 0.8505162000656128,
+      "learning_rate": 6.885267134376686e-05,
+      "loss": 3.1944,
+      "step": 82150
+    },
+    {
+      "epoch": 8.862533692722373,
+      "grad_norm": 0.8600515127182007,
+      "learning_rate": 6.852887209929844e-05,
+      "loss": 3.1814,
+      "step": 82200
+    },
+    {
+      "epoch": 8.867924528301886,
+      "grad_norm": 0.857338011264801,
+      "learning_rate": 6.820507285483e-05,
+      "loss": 3.1975,
+      "step": 82250
+    },
+    {
+      "epoch": 8.873315363881401,
+      "grad_norm": 0.846052885055542,
+      "learning_rate": 6.788127361036156e-05,
+      "loss": 3.1834,
+      "step": 82300
+    },
+    {
+      "epoch": 8.878706199460916,
+      "grad_norm": 0.8257965445518494,
+      "learning_rate": 6.755747436589314e-05,
+      "loss": 3.1911,
+      "step": 82350
+    },
+    {
+      "epoch": 8.884097035040432,
+      "grad_norm": 0.8405656814575195,
+      "learning_rate": 6.72336751214247e-05,
+      "loss": 3.1808,
+      "step": 82400
+    },
+    {
+      "epoch": 8.889487870619947,
+      "grad_norm": 0.8875756859779358,
+      "learning_rate": 6.690987587695628e-05,
+      "loss": 3.1942,
+      "step": 82450
+    },
+    {
+      "epoch": 8.89487870619946,
+      "grad_norm": 0.8436116576194763,
+      "learning_rate": 6.658607663248785e-05,
+      "loss": 3.1817,
+      "step": 82500
+    },
+    {
+      "epoch": 8.900269541778975,
+      "grad_norm": 0.8657233715057373,
+      "learning_rate": 6.626227738801942e-05,
+      "loss": 3.1703,
+      "step": 82550
+    },
+    {
+      "epoch": 8.90566037735849,
+      "grad_norm": 0.8531566262245178,
+      "learning_rate": 6.593847814355099e-05,
+      "loss": 3.1897,
+      "step": 82600
+    },
+    {
+      "epoch": 8.911051212938006,
+      "grad_norm": 0.8826576471328735,
+      "learning_rate": 6.561467889908257e-05,
+      "loss": 3.1804,
+      "step": 82650
+    },
+    {
+      "epoch": 8.91644204851752,
+      "grad_norm": 0.8785792589187622,
+      "learning_rate": 6.52973556395035e-05,
+      "loss": 3.1996,
+      "step": 82700
+    },
+    {
+      "epoch": 8.921832884097036,
+      "grad_norm": 0.8175323009490967,
+      "learning_rate": 6.497355639503507e-05,
+      "loss": 3.1739,
+      "step": 82750
+    },
+    {
+      "epoch": 8.92722371967655,
+      "grad_norm": 0.876568078994751,
+      "learning_rate": 6.464975715056663e-05,
+      "loss": 3.1764,
+      "step": 82800
+    },
+    {
+      "epoch": 8.932614555256064,
+      "grad_norm": 0.8088855743408203,
+      "learning_rate": 6.432595790609821e-05,
+      "loss": 3.1797,
+      "step": 82850
+    },
+    {
+      "epoch": 8.93800539083558,
+      "grad_norm": 0.8298810124397278,
+      "learning_rate": 6.400215866162979e-05,
+      "loss": 3.1778,
+      "step": 82900
+    },
+    {
+      "epoch": 8.943396226415095,
+      "grad_norm": 0.8599114418029785,
+      "learning_rate": 6.367835941716135e-05,
+      "loss": 3.1789,
+      "step": 82950
+    },
+    {
+      "epoch": 8.94878706199461,
+      "grad_norm": 0.8379719257354736,
+      "learning_rate": 6.335456017269293e-05,
+      "loss": 3.1743,
+      "step": 83000
+    },
+    {
+      "epoch": 8.94878706199461,
+      "eval_accuracy": 0.39262840136507143,
+      "eval_loss": 3.3144419193267822,
+      "eval_runtime": 183.2243,
+      "eval_samples_per_second": 98.3,
+      "eval_steps_per_second": 6.145,
+      "step": 83000
+    },
+    {
+      "epoch": 8.954177897574123,
+      "grad_norm": 0.8326112031936646,
+      "learning_rate": 6.30307609282245e-05,
+      "loss": 3.1714,
+      "step": 83050
+    },
+    {
+      "epoch": 8.959568733153638,
+      "grad_norm": 0.8796714544296265,
+      "learning_rate": 6.270696168375607e-05,
+      "loss": 3.1805,
+      "step": 83100
+    },
+    {
+      "epoch": 8.964959568733153,
+      "grad_norm": 0.8742213845252991,
+      "learning_rate": 6.238316243928764e-05,
+      "loss": 3.2045,
+      "step": 83150
+    },
+    {
+      "epoch": 8.970350404312669,
+      "grad_norm": 0.8367428183555603,
+      "learning_rate": 6.205936319481921e-05,
+      "loss": 3.184,
+      "step": 83200
+    },
+    {
+      "epoch": 8.975741239892184,
+      "grad_norm": 0.8559758067131042,
+      "learning_rate": 6.173556395035078e-05,
+      "loss": 3.1707,
+      "step": 83250
+    },
+    {
+      "epoch": 8.981132075471699,
+      "grad_norm": 0.8741273880004883,
+      "learning_rate": 6.141176470588236e-05,
+      "loss": 3.1795,
+      "step": 83300
+    },
+    {
+      "epoch": 8.986522911051212,
+      "grad_norm": 0.8743473887443542,
+      "learning_rate": 6.108796546141392e-05,
+      "loss": 3.1846,
+      "step": 83350
+    },
+    {
+      "epoch": 8.991913746630727,
+      "grad_norm": 0.8502634763717651,
+      "learning_rate": 6.076416621694549e-05,
+      "loss": 3.1702,
+      "step": 83400
+    },
+    {
+      "epoch": 8.997304582210242,
+      "grad_norm": 0.8954312801361084,
+      "learning_rate": 6.0440366972477055e-05,
+      "loss": 3.1986,
+      "step": 83450
+    },
+    {
+      "epoch": 9.002695417789758,
+      "grad_norm": 0.8895218372344971,
+      "learning_rate": 6.011656772800863e-05,
+      "loss": 3.1271,
+      "step": 83500
+    },
+    {
+      "epoch": 9.008086253369273,
+      "grad_norm": 0.8508468270301819,
+      "learning_rate": 5.9792768483540197e-05,
+      "loss": 3.1216,
+      "step": 83550
+    },
+    {
+      "epoch": 9.013477088948788,
+      "grad_norm": 0.8222100734710693,
+      "learning_rate": 5.946896923907177e-05,
+      "loss": 3.1388,
+      "step": 83600
+    },
+    {
+      "epoch": 9.018867924528301,
+      "grad_norm": 0.849051296710968,
+      "learning_rate": 5.9145169994603345e-05,
+      "loss": 3.1336,
+      "step": 83650
+    },
+    {
+      "epoch": 9.024258760107816,
+      "grad_norm": 0.8318443298339844,
+      "learning_rate": 5.882137075013491e-05,
+      "loss": 3.117,
+      "step": 83700
+    },
+    {
+      "epoch": 9.029649595687331,
+      "grad_norm": 0.8428478837013245,
+      "learning_rate": 5.849757150566649e-05,
+      "loss": 3.1147,
+      "step": 83750
+    },
+    {
+      "epoch": 9.035040431266847,
+      "grad_norm": 0.9040572643280029,
+      "learning_rate": 5.817377226119805e-05,
+      "loss": 3.1176,
+      "step": 83800
+    },
+    {
+      "epoch": 9.040431266846362,
+      "grad_norm": 0.8731981515884399,
+      "learning_rate": 5.784997301672963e-05,
+      "loss": 3.1063,
+      "step": 83850
+    },
+    {
+      "epoch": 9.045822102425875,
+      "grad_norm": 0.8788706660270691,
+      "learning_rate": 5.752617377226119e-05,
+      "loss": 3.1227,
+      "step": 83900
+    },
+    {
+      "epoch": 9.05121293800539,
+      "grad_norm": 0.8007208704948425,
+      "learning_rate": 5.7202374527792764e-05,
+      "loss": 3.1236,
+      "step": 83950
+    },
+    {
+      "epoch": 9.056603773584905,
+      "grad_norm": 0.861991286277771,
+      "learning_rate": 5.6878575283324335e-05,
+      "loss": 3.1092,
+      "step": 84000
+    },
+    {
+      "epoch": 9.056603773584905,
+      "eval_accuracy": 0.39257820373090113,
+      "eval_loss": 3.3174729347229004,
+      "eval_runtime": 183.6793,
+      "eval_samples_per_second": 98.057,
+      "eval_steps_per_second": 6.13,
+      "step": 84000
+    },
+    {
+      "epoch": 9.06199460916442,
+      "grad_norm": 0.8262946605682373,
+      "learning_rate": 5.6554776038855905e-05,
+      "loss": 3.1227,
+      "step": 84050
+    },
+    {
+      "epoch": 9.067385444743936,
+      "grad_norm": 0.8459504842758179,
+      "learning_rate": 5.623097679438747e-05,
+      "loss": 3.1484,
+      "step": 84100
+    },
+    {
+      "epoch": 9.07277628032345,
+      "grad_norm": 0.8339034914970398,
+      "learning_rate": 5.5913653534808415e-05,
+      "loss": 3.1279,
+      "step": 84150
+    },
+    {
+      "epoch": 9.078167115902964,
+      "grad_norm": 0.8436874747276306,
+      "learning_rate": 5.558985429033998e-05,
+      "loss": 3.1422,
+      "step": 84200
+    },
+    {
+      "epoch": 9.08355795148248,
+      "grad_norm": 0.8555903434753418,
+      "learning_rate": 5.526605504587156e-05,
+      "loss": 3.1369,
+      "step": 84250
+    },
+    {
+      "epoch": 9.088948787061994,
+      "grad_norm": 0.8775126934051514,
+      "learning_rate": 5.494225580140313e-05,
+      "loss": 3.1264,
+      "step": 84300
+    },
+    {
+      "epoch": 9.09433962264151,
+      "grad_norm": 0.8575963973999023,
+      "learning_rate": 5.461845655693469e-05,
+      "loss": 3.1215,
+      "step": 84350
+    },
+    {
+      "epoch": 9.099730458221025,
+      "grad_norm": 0.8739725351333618,
+      "learning_rate": 5.429465731246627e-05,
+      "loss": 3.1107,
+      "step": 84400
+    },
+    {
+      "epoch": 9.10512129380054,
+      "grad_norm": 0.8500518798828125,
+      "learning_rate": 5.3970858067997833e-05,
+      "loss": 3.1383,
+      "step": 84450
+    },
+    {
+      "epoch": 9.110512129380053,
+      "grad_norm": 0.9054665565490723,
+      "learning_rate": 5.364705882352941e-05,
+      "loss": 3.1242,
+      "step": 84500
+    },
+    {
+      "epoch": 9.115902964959568,
+      "grad_norm": 0.8518809080123901,
+      "learning_rate": 5.3323259579060975e-05,
+      "loss": 3.1319,
+      "step": 84550
+    },
+    {
+      "epoch": 9.121293800539084,
+      "grad_norm": 0.8315262198448181,
+      "learning_rate": 5.299946033459255e-05,
+      "loss": 3.1216,
+      "step": 84600
+    },
+    {
+      "epoch": 9.126684636118599,
+      "grad_norm": 0.8629969954490662,
+      "learning_rate": 5.267566109012412e-05,
+      "loss": 3.11,
+      "step": 84650
+    },
+    {
+      "epoch": 9.132075471698114,
+      "grad_norm": 0.8560936450958252,
+      "learning_rate": 5.235186184565569e-05,
+      "loss": 3.122,
+      "step": 84700
+    },
+    {
+      "epoch": 9.137466307277627,
+      "grad_norm": 0.9044983386993408,
+      "learning_rate": 5.202806260118726e-05,
+      "loss": 3.1238,
+      "step": 84750
+    },
+    {
+      "epoch": 9.142857142857142,
+      "grad_norm": 0.8578569293022156,
+      "learning_rate": 5.170426335671883e-05,
+      "loss": 3.1247,
+      "step": 84800
+    },
+    {
+      "epoch": 9.148247978436657,
+      "grad_norm": 0.88395756483078,
+      "learning_rate": 5.1380464112250394e-05,
+      "loss": 3.1364,
+      "step": 84850
+    },
+    {
+      "epoch": 9.153638814016173,
+      "grad_norm": 0.8741885423660278,
+      "learning_rate": 5.105666486778197e-05,
+      "loss": 3.1431,
+      "step": 84900
+    },
+    {
+      "epoch": 9.159029649595688,
+      "grad_norm": 0.8336004614830017,
+      "learning_rate": 5.0732865623313536e-05,
+      "loss": 3.1303,
+      "step": 84950
+    },
+    {
+      "epoch": 9.164420485175203,
+      "grad_norm": 0.9022056460380554,
+      "learning_rate": 5.040906637884511e-05,
+      "loss": 3.1468,
+      "step": 85000
+    },
+    {
+      "epoch": 9.164420485175203,
+      "eval_accuracy": 0.39289188461802155,
+      "eval_loss": 3.315161943435669,
+      "eval_runtime": 183.2552,
+      "eval_samples_per_second": 98.284,
+      "eval_steps_per_second": 6.144,
+      "step": 85000
+    },
+    {
+      "epoch": 9.169811320754716,
+      "grad_norm": 0.8745513558387756,
+      "learning_rate": 5.0085267134376684e-05,
+      "loss": 3.1347,
+      "step": 85050
+    },
+    {
+      "epoch": 9.175202156334231,
+      "grad_norm": 0.8498513102531433,
+      "learning_rate": 4.9761467889908255e-05,
+      "loss": 3.1333,
+      "step": 85100
+    },
+    {
+      "epoch": 9.180592991913747,
+      "grad_norm": 0.8566269278526306,
+      "learning_rate": 4.9437668645439826e-05,
+      "loss": 3.1396,
+      "step": 85150
+    },
+    {
+      "epoch": 9.185983827493262,
+      "grad_norm": 0.8552309274673462,
+      "learning_rate": 4.911386940097139e-05,
+      "loss": 3.1405,
+      "step": 85200
+    },
+    {
+      "epoch": 9.191374663072777,
+      "grad_norm": 0.8931257128715515,
+      "learning_rate": 4.879007015650297e-05,
+      "loss": 3.1182,
+      "step": 85250
+    },
+    {
+      "epoch": 9.19676549865229,
+      "grad_norm": 0.9051719307899475,
+      "learning_rate": 4.846627091203453e-05,
+      "loss": 3.1298,
+      "step": 85300
+    },
+    {
+      "epoch": 9.202156334231805,
+      "grad_norm": 0.8819602131843567,
+      "learning_rate": 4.814247166756611e-05,
+      "loss": 3.1231,
+      "step": 85350
+    },
+    {
+      "epoch": 9.20754716981132,
+      "grad_norm": 0.8713910579681396,
+      "learning_rate": 4.7818672423097674e-05,
+      "loss": 3.133,
+      "step": 85400
+    },
+    {
+      "epoch": 9.212938005390836,
+      "grad_norm": 0.8759667873382568,
+      "learning_rate": 4.7494873178629244e-05,
+      "loss": 3.1367,
+      "step": 85450
+    },
+    {
+      "epoch": 9.21832884097035,
+      "grad_norm": 0.8757563233375549,
+      "learning_rate": 4.7171073934160815e-05,
+      "loss": 3.1252,
+      "step": 85500
+    },
+    {
+      "epoch": 9.223719676549866,
+      "grad_norm": 0.882490336894989,
+      "learning_rate": 4.6847274689692386e-05,
+      "loss": 3.1112,
+      "step": 85550
+    },
+    {
+      "epoch": 9.22911051212938,
+      "grad_norm": 0.8384507298469543,
+      "learning_rate": 4.652347544522396e-05,
+      "loss": 3.1179,
+      "step": 85600
+    },
+    {
+      "epoch": 9.234501347708894,
+      "grad_norm": 0.8560865521430969,
+      "learning_rate": 4.619967620075553e-05,
+      "loss": 3.1353,
+      "step": 85650
+    },
+    {
+      "epoch": 9.23989218328841,
+      "grad_norm": 0.8397020101547241,
+      "learning_rate": 4.587587695628709e-05,
+      "loss": 3.1252,
+      "step": 85700
+    },
+    {
+      "epoch": 9.245283018867925,
+      "grad_norm": 0.905177116394043,
+      "learning_rate": 4.555207771181867e-05,
+      "loss": 3.1131,
+      "step": 85750
+    },
+    {
+      "epoch": 9.25067385444744,
+      "grad_norm": 0.8663109540939331,
+      "learning_rate": 4.522827846735024e-05,
+      "loss": 3.126,
+      "step": 85800
+    },
+    {
+      "epoch": 9.256064690026955,
+      "grad_norm": 0.8782655000686646,
+      "learning_rate": 4.490447922288181e-05,
+      "loss": 3.1507,
+      "step": 85850
+    },
+    {
+      "epoch": 9.261455525606468,
+      "grad_norm": 0.869422972202301,
+      "learning_rate": 4.458067997841338e-05,
+      "loss": 3.1346,
+      "step": 85900
+    },
+    {
+      "epoch": 9.266846361185983,
+      "grad_norm": 0.9452365636825562,
+      "learning_rate": 4.4256880733944947e-05,
+      "loss": 3.1364,
+      "step": 85950
+    },
+    {
+      "epoch": 9.272237196765499,
+      "grad_norm": 0.8705593347549438,
+      "learning_rate": 4.3933081489476524e-05,
+      "loss": 3.1336,
+      "step": 86000
+    },
+    {
+      "epoch": 9.272237196765499,
+      "eval_accuracy": 0.39303128627308753,
+      "eval_loss": 3.3147635459899902,
+      "eval_runtime": 183.47,
+      "eval_samples_per_second": 98.169,
+      "eval_steps_per_second": 6.137,
+      "step": 86000
+    },
+    {
+      "epoch": 9.277628032345014,
+      "grad_norm": 0.8607900142669678,
+      "learning_rate": 4.360928224500809e-05,
+      "loss": 3.1422,
+      "step": 86050
+    },
+    {
+      "epoch": 9.283018867924529,
+      "grad_norm": 0.8422678112983704,
+      "learning_rate": 4.3285483000539666e-05,
+      "loss": 3.1357,
+      "step": 86100
+    },
+    {
+      "epoch": 9.288409703504042,
+      "grad_norm": 0.8893538117408752,
+      "learning_rate": 4.296168375607123e-05,
+      "loss": 3.1485,
+      "step": 86150
+    },
+    {
+      "epoch": 9.293800539083557,
+      "grad_norm": 0.8745595812797546,
+      "learning_rate": 4.263788451160281e-05,
+      "loss": 3.1405,
+      "step": 86200
+    },
+    {
+      "epoch": 9.299191374663073,
+      "grad_norm": 0.8779528141021729,
+      "learning_rate": 4.231408526713437e-05,
+      "loss": 3.1441,
+      "step": 86250
+    },
+    {
+      "epoch": 9.304582210242588,
+      "grad_norm": 0.9104282855987549,
+      "learning_rate": 4.199028602266594e-05,
+      "loss": 3.1426,
+      "step": 86300
+    },
+    {
+      "epoch": 9.309973045822103,
+      "grad_norm": 0.8767886161804199,
+      "learning_rate": 4.1666486778197514e-05,
+      "loss": 3.1166,
+      "step": 86350
+    },
+    {
+      "epoch": 9.315363881401618,
+      "grad_norm": 0.8752766847610474,
+      "learning_rate": 4.1342687533729085e-05,
+      "loss": 3.1262,
+      "step": 86400
+    },
+    {
+      "epoch": 9.320754716981131,
+      "grad_norm": 0.8585779070854187,
+      "learning_rate": 4.101888828926065e-05,
+      "loss": 3.1362,
+      "step": 86450
+    },
+    {
+      "epoch": 9.326145552560646,
+      "grad_norm": 0.8988914489746094,
+      "learning_rate": 4.0695089044792226e-05,
+      "loss": 3.1335,
+      "step": 86500
+    },
+    {
+      "epoch": 9.331536388140162,
+      "grad_norm": 0.894045889377594,
+      "learning_rate": 4.037128980032379e-05,
+      "loss": 3.1418,
+      "step": 86550
+    },
+    {
+      "epoch": 9.336927223719677,
+      "grad_norm": 0.8583129644393921,
+      "learning_rate": 4.004749055585537e-05,
+      "loss": 3.1542,
+      "step": 86600
+    },
+    {
+      "epoch": 9.342318059299192,
+      "grad_norm": 0.8705652952194214,
+      "learning_rate": 3.972369131138694e-05,
+      "loss": 3.1218,
+      "step": 86650
+    },
+    {
+      "epoch": 9.347708894878707,
+      "grad_norm": 0.8325337767601013,
+      "learning_rate": 3.93998920669185e-05,
+      "loss": 3.1344,
+      "step": 86700
+    },
+    {
+      "epoch": 9.35309973045822,
+      "grad_norm": 0.8604037165641785,
+      "learning_rate": 3.907609282245008e-05,
+      "loss": 3.115,
+      "step": 86750
+    },
+    {
+      "epoch": 9.358490566037736,
+      "grad_norm": 0.8888213038444519,
+      "learning_rate": 3.8752293577981645e-05,
+      "loss": 3.118,
+      "step": 86800
+    },
+    {
+      "epoch": 9.36388140161725,
+      "grad_norm": 0.8469123244285583,
+      "learning_rate": 3.842849433351322e-05,
+      "loss": 3.1361,
+      "step": 86850
+    },
+    {
+      "epoch": 9.369272237196766,
+      "grad_norm": 0.9041824340820312,
+      "learning_rate": 3.810469508904479e-05,
+      "loss": 3.141,
+      "step": 86900
+    },
+    {
+      "epoch": 9.374663072776281,
+      "grad_norm": 0.8636651039123535,
+      "learning_rate": 3.7780895844576364e-05,
+      "loss": 3.1412,
+      "step": 86950
+    },
+    {
+      "epoch": 9.380053908355794,
+      "grad_norm": 0.8707324862480164,
+      "learning_rate": 3.745709660010793e-05,
+      "loss": 3.1193,
+      "step": 87000
+    },
+    {
+      "epoch": 9.380053908355794,
+      "eval_accuracy": 0.3932831436670415,
+      "eval_loss": 3.312223196029663,
+      "eval_runtime": 183.4336,
+      "eval_samples_per_second": 98.188,
+      "eval_steps_per_second": 6.138,
+      "step": 87000
+    },
+    {
+      "epoch": 9.38544474393531,
+      "grad_norm": 0.8810425996780396,
+      "learning_rate": 3.71332973556395e-05,
+      "loss": 3.141,
+      "step": 87050
+    },
+    {
+      "epoch": 9.390835579514825,
+      "grad_norm": 0.8390422463417053,
+      "learning_rate": 3.680949811117107e-05,
+      "loss": 3.1296,
+      "step": 87100
+    },
+    {
+      "epoch": 9.39622641509434,
+      "grad_norm": 0.8569521903991699,
+      "learning_rate": 3.648569886670264e-05,
+      "loss": 3.1333,
+      "step": 87150
+    },
+    {
+      "epoch": 9.401617250673855,
+      "grad_norm": 0.8592923879623413,
+      "learning_rate": 3.616189962223421e-05,
+      "loss": 3.1306,
+      "step": 87200
+    },
+    {
+      "epoch": 9.40700808625337,
+      "grad_norm": 0.8302024006843567,
+      "learning_rate": 3.583810037776578e-05,
+      "loss": 3.1448,
+      "step": 87250
+    },
+    {
+      "epoch": 9.412398921832883,
+      "grad_norm": 0.8882058262825012,
+      "learning_rate": 3.5514301133297354e-05,
+      "loss": 3.1514,
+      "step": 87300
+    },
+    {
+      "epoch": 9.417789757412399,
+      "grad_norm": 0.8802513480186462,
+      "learning_rate": 3.5190501888828925e-05,
+      "loss": 3.1292,
+      "step": 87350
+    },
+    {
+      "epoch": 9.423180592991914,
+      "grad_norm": 0.9002079367637634,
+      "learning_rate": 3.4866702644360496e-05,
+      "loss": 3.1277,
+      "step": 87400
+    },
+    {
+      "epoch": 9.428571428571429,
+      "grad_norm": 0.8838725090026855,
+      "learning_rate": 3.4542903399892067e-05,
+      "loss": 3.1346,
+      "step": 87450
+    },
+    {
+      "epoch": 9.433962264150944,
+      "grad_norm": 0.8865829706192017,
+      "learning_rate": 3.421910415542363e-05,
+      "loss": 3.1364,
+      "step": 87500
+    },
+    {
+      "epoch": 9.439353099730457,
+      "grad_norm": 0.834714412689209,
+      "learning_rate": 3.38953049109552e-05,
+      "loss": 3.1481,
+      "step": 87550
+    },
+    {
+      "epoch": 9.444743935309972,
+      "grad_norm": 0.9003492593765259,
+      "learning_rate": 3.357150566648677e-05,
+      "loss": 3.1328,
+      "step": 87600
+    },
+    {
+      "epoch": 9.450134770889488,
+      "grad_norm": 0.8663783073425293,
+      "learning_rate": 3.324770642201834e-05,
+      "loss": 3.1493,
+      "step": 87650
+    },
+    {
+      "epoch": 9.455525606469003,
+      "grad_norm": 0.8771715760231018,
+      "learning_rate": 3.2923907177549914e-05,
+      "loss": 3.1084,
+      "step": 87700
+    },
+    {
+      "epoch": 9.460916442048518,
+      "grad_norm": 0.844178318977356,
+      "learning_rate": 3.2600107933081485e-05,
+      "loss": 3.1292,
+      "step": 87750
+    },
+    {
+      "epoch": 9.466307277628033,
+      "grad_norm": 0.8180273771286011,
+      "learning_rate": 3.2276308688613056e-05,
+      "loss": 3.1236,
+      "step": 87800
+    },
+    {
+      "epoch": 9.471698113207546,
+      "grad_norm": 0.8939014077186584,
+      "learning_rate": 3.195250944414463e-05,
+      "loss": 3.1332,
+      "step": 87850
+    },
+    {
+      "epoch": 9.477088948787062,
+      "grad_norm": 0.9243705868721008,
+      "learning_rate": 3.16287101996762e-05,
+      "loss": 3.1442,
+      "step": 87900
+    },
+    {
+      "epoch": 9.482479784366577,
+      "grad_norm": 0.8558386564254761,
+      "learning_rate": 3.130491095520777e-05,
+      "loss": 3.1451,
+      "step": 87950
+    },
+    {
+      "epoch": 9.487870619946092,
+      "grad_norm": 0.8696820735931396,
+      "learning_rate": 3.098111171073934e-05,
+      "loss": 3.1383,
+      "step": 88000
+    },
+    {
+      "epoch": 9.487870619946092,
+      "eval_accuracy": 0.3937604558032105,
+      "eval_loss": 3.308690071105957,
+      "eval_runtime": 183.6085,
+      "eval_samples_per_second": 98.095,
+      "eval_steps_per_second": 6.133,
+      "step": 88000
+    },
+    {
+      "epoch": 9.493261455525607,
+      "grad_norm": 0.8751658797264099,
+      "learning_rate": 3.065731246627091e-05,
+      "loss": 3.1353,
+      "step": 88050
+    },
+    {
+      "epoch": 9.498652291105122,
+      "grad_norm": 0.8156877756118774,
+      "learning_rate": 3.033351322180248e-05,
+      "loss": 3.1331,
+      "step": 88100
+    },
+    {
+      "epoch": 9.504043126684635,
+      "grad_norm": 0.8443343639373779,
+      "learning_rate": 3.0016189962223416e-05,
+      "loss": 3.1318,
+      "step": 88150
+    },
+    {
+      "epoch": 9.50943396226415,
+      "grad_norm": 0.8381995558738708,
+      "learning_rate": 2.9692390717754987e-05,
+      "loss": 3.1284,
+      "step": 88200
+    },
+    {
+      "epoch": 9.514824797843666,
+      "grad_norm": 0.8590207695960999,
+      "learning_rate": 2.9368591473286558e-05,
+      "loss": 3.1363,
+      "step": 88250
+    },
+    {
+      "epoch": 9.520215633423181,
+      "grad_norm": 0.8370568752288818,
+      "learning_rate": 2.904479222881813e-05,
+      "loss": 3.1438,
+      "step": 88300
+    },
+    {
+      "epoch": 9.525606469002696,
+      "grad_norm": 0.8353069424629211,
+      "learning_rate": 2.87209929843497e-05,
+      "loss": 3.1433,
+      "step": 88350
+    },
+    {
+      "epoch": 9.530997304582211,
+      "grad_norm": 0.861653745174408,
+      "learning_rate": 2.8397193739881274e-05,
+      "loss": 3.1433,
+      "step": 88400
+    },
+    {
+      "epoch": 9.536388140161725,
+      "grad_norm": 0.8921690583229065,
+      "learning_rate": 2.8073394495412842e-05,
+      "loss": 3.1364,
+      "step": 88450
+    },
+    {
+      "epoch": 9.54177897574124,
+      "grad_norm": 0.8571642637252808,
+      "learning_rate": 2.7749595250944413e-05,
+      "loss": 3.1289,
+      "step": 88500
+    },
+    {
+      "epoch": 9.547169811320755,
+      "grad_norm": 0.8388140201568604,
+      "learning_rate": 2.7425796006475984e-05,
+      "loss": 3.1231,
+      "step": 88550
+    },
+    {
+      "epoch": 9.55256064690027,
+      "grad_norm": 0.8860104084014893,
+      "learning_rate": 2.7101996762007554e-05,
+      "loss": 3.1384,
+      "step": 88600
+    },
+    {
+      "epoch": 9.557951482479785,
+      "grad_norm": 0.8494702577590942,
+      "learning_rate": 2.6778197517539125e-05,
+      "loss": 3.1066,
+      "step": 88650
+    },
+    {
+      "epoch": 9.563342318059298,
+      "grad_norm": 0.8490155339241028,
+      "learning_rate": 2.6454398273070693e-05,
+      "loss": 3.1306,
+      "step": 88700
+    },
+    {
+      "epoch": 9.568733153638814,
+      "grad_norm": 0.8371010422706604,
+      "learning_rate": 2.6130599028602264e-05,
+      "loss": 3.1372,
+      "step": 88750
+    },
+    {
+      "epoch": 9.574123989218329,
+      "grad_norm": 0.8877601623535156,
+      "learning_rate": 2.5806799784133835e-05,
+      "loss": 3.1136,
+      "step": 88800
+    },
+    {
+      "epoch": 9.579514824797844,
+      "grad_norm": 0.9142347574234009,
+      "learning_rate": 2.5483000539665406e-05,
+      "loss": 3.1329,
+      "step": 88850
+    },
+    {
+      "epoch": 9.584905660377359,
+      "grad_norm": 0.863750159740448,
+      "learning_rate": 2.5159201295196976e-05,
+      "loss": 3.1539,
+      "step": 88900
+    },
+    {
+      "epoch": 9.590296495956874,
+      "grad_norm": 0.8566252589225769,
+      "learning_rate": 2.4835402050728544e-05,
+      "loss": 3.1405,
+      "step": 88950
+    },
+    {
+      "epoch": 9.595687331536388,
+      "grad_norm": 0.8429293632507324,
+      "learning_rate": 2.4511602806260115e-05,
+      "loss": 3.13,
+      "step": 89000
+    },
+    {
+      "epoch": 9.595687331536388,
+      "eval_accuracy": 0.3937995708428237,
+      "eval_loss": 3.306797742843628,
+      "eval_runtime": 183.3113,
+      "eval_samples_per_second": 98.254,
+      "eval_steps_per_second": 6.143,
+      "step": 89000
+    },
+    {
+      "epoch": 9.601078167115903,
+      "grad_norm": 0.7996606826782227,
+      "learning_rate": 2.4187803561791686e-05,
+      "loss": 3.1285,
+      "step": 89050
+    },
+    {
+      "epoch": 9.606469002695418,
+      "grad_norm": 0.863088846206665,
+      "learning_rate": 2.3864004317323257e-05,
+      "loss": 3.1334,
+      "step": 89100
+    },
+    {
+      "epoch": 9.611859838274933,
+      "grad_norm": 0.8993324041366577,
+      "learning_rate": 2.354020507285483e-05,
+      "loss": 3.157,
+      "step": 89150
+    },
+    {
+      "epoch": 9.617250673854448,
+      "grad_norm": 0.9314507246017456,
+      "learning_rate": 2.3216405828386402e-05,
+      "loss": 3.1221,
+      "step": 89200
+    },
+    {
+      "epoch": 9.622641509433961,
+      "grad_norm": 0.9125588536262512,
+      "learning_rate": 2.289260658391797e-05,
+      "loss": 3.143,
+      "step": 89250
+    },
+    {
+      "epoch": 9.628032345013477,
+      "grad_norm": 0.8934492468833923,
+      "learning_rate": 2.256880733944954e-05,
+      "loss": 3.1344,
+      "step": 89300
+    },
+    {
+      "epoch": 9.633423180592992,
+      "grad_norm": 0.8534916639328003,
+      "learning_rate": 2.224500809498111e-05,
+      "loss": 3.1308,
+      "step": 89350
+    },
+    {
+      "epoch": 9.638814016172507,
+      "grad_norm": 0.9043375253677368,
+      "learning_rate": 2.1921208850512682e-05,
+      "loss": 3.1314,
+      "step": 89400
+    },
+    {
+      "epoch": 9.644204851752022,
+      "grad_norm": 0.8547094464302063,
+      "learning_rate": 2.1597409606044253e-05,
+      "loss": 3.1342,
+      "step": 89450
+    },
+    {
+      "epoch": 9.649595687331537,
+      "grad_norm": 0.8704356551170349,
+      "learning_rate": 2.127361036157582e-05,
+      "loss": 3.1467,
+      "step": 89500
+    },
+    {
+      "epoch": 9.65498652291105,
+      "grad_norm": 0.9299810528755188,
+      "learning_rate": 2.094981111710739e-05,
+      "loss": 3.1332,
+      "step": 89550
+    },
+    {
+      "epoch": 9.660377358490566,
+      "grad_norm": 0.8665199279785156,
+      "learning_rate": 2.0626011872638962e-05,
+      "loss": 3.1221,
+      "step": 89600
+    },
+    {
+      "epoch": 9.66576819407008,
+      "grad_norm": 0.9026532769203186,
+      "learning_rate": 2.0302212628170533e-05,
+      "loss": 3.1229,
+      "step": 89650
+    },
+    {
+      "epoch": 9.671159029649596,
+      "grad_norm": 0.8790315985679626,
+      "learning_rate": 1.99784133837021e-05,
+      "loss": 3.1286,
+      "step": 89700
+    },
+    {
+      "epoch": 9.676549865229111,
+      "grad_norm": 0.8397992253303528,
+      "learning_rate": 1.965461413923367e-05,
+      "loss": 3.1425,
+      "step": 89750
+    },
+    {
+      "epoch": 9.681940700808624,
+      "grad_norm": 0.8655675053596497,
+      "learning_rate": 1.9330814894765242e-05,
+      "loss": 3.1238,
+      "step": 89800
+    },
+    {
+      "epoch": 9.68733153638814,
+      "grad_norm": 0.8382636904716492,
+      "learning_rate": 1.9007015650296813e-05,
+      "loss": 3.128,
+      "step": 89850
+    },
+    {
+      "epoch": 9.692722371967655,
+      "grad_norm": 0.8530244827270508,
+      "learning_rate": 1.8683216405828384e-05,
+      "loss": 3.1491,
+      "step": 89900
+    },
+    {
+      "epoch": 9.69811320754717,
+      "grad_norm": 0.8549965620040894,
+      "learning_rate": 1.8359417161359955e-05,
+      "loss": 3.149,
+      "step": 89950
+    },
+    {
+      "epoch": 9.703504043126685,
+      "grad_norm": 0.883962869644165,
+      "learning_rate": 1.8035617916891526e-05,
+      "loss": 3.135,
+      "step": 90000
+    },
+    {
+      "epoch": 9.703504043126685,
+      "eval_accuracy": 0.3941857232061164,
+      "eval_loss": 3.3042373657226562,
+      "eval_runtime": 183.4981,
+      "eval_samples_per_second": 98.154,
+      "eval_steps_per_second": 6.136,
+      "step": 90000
+    }
+  ],
+  "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": 7.524622614528e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}