diff --git "a/checkpoints/latest/trainer_state.json" "b/checkpoints/latest/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoints/latest/trainer_state.json"
@@ -0,0 +1,4454 @@
+{
+  "best_metric": 0.9587281193044483,
+  "best_model_checkpoint": "outputs/TypePredictor/checkpoints/checkpoint-25000",
+  "epoch": 3.9994047382832654,
+  "eval_steps": 500,
+  "global_step": 25196,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.007936822889797215,
+      "grad_norm": 50.20066833496094,
+      "learning_rate": 1.8253968253968257e-07,
+      "loss": 4.1606,
+      "step": 50
+    },
+    {
+      "epoch": 0.01587364577959443,
+      "grad_norm": 103.1044692993164,
+      "learning_rate": 3.80952380952381e-07,
+      "loss": 3.4782,
+      "step": 100
+    },
+    {
+      "epoch": 0.023810468669391644,
+      "grad_norm": 29.89408302307129,
+      "learning_rate": 5.793650793650794e-07,
+      "loss": 2.7518,
+      "step": 150
+    },
+    {
+      "epoch": 0.03174729155918886,
+      "grad_norm": 26.819448471069336,
+      "learning_rate": 7.738095238095239e-07,
+      "loss": 2.5176,
+      "step": 200
+    },
+    {
+      "epoch": 0.03968411444898607,
+      "grad_norm": 41.28379440307617,
+      "learning_rate": 9.722222222222224e-07,
+      "loss": 2.4074,
+      "step": 250
+    },
+    {
+      "epoch": 0.04762093733878329,
+      "grad_norm": 24.081151962280273,
+      "learning_rate": 1.1706349206349208e-06,
+      "loss": 2.2454,
+      "step": 300
+    },
+    {
+      "epoch": 0.0555577602285805,
+      "grad_norm": 27.1600341796875,
+      "learning_rate": 1.3690476190476193e-06,
+      "loss": 2.2193,
+      "step": 350
+    },
+    {
+      "epoch": 0.06349458311837772,
+      "grad_norm": 35.915611267089844,
+      "learning_rate": 1.5674603174603175e-06,
+      "loss": 2.0748,
+      "step": 400
+    },
+    {
+      "epoch": 0.07143140600817492,
+      "grad_norm": 25.79641342163086,
+      "learning_rate": 1.765873015873016e-06,
+      "loss": 1.9517,
+      "step": 450
+    },
+    {
+      "epoch": 0.07936822889797214,
+      "grad_norm": 32.5029182434082,
+      "learning_rate": 1.9642857142857144e-06,
+      "loss": 1.7744,
+      "step": 500
+    },
+    {
+      "epoch": 0.07936822889797214,
+      "eval_accuracy": 0.4996031746031746,
+      "eval_loss": 1.6034170389175415,
+      "eval_macro_f1": 0.16675213835131356,
+      "eval_macro_precision": 0.20112826287340668,
+      "eval_macro_recall": 0.1656798423386093,
+      "eval_micro_f1": 0.4996031746031746,
+      "eval_micro_precision": 0.4996031746031746,
+      "eval_micro_recall": 0.4996031746031746,
+      "eval_runtime": 50.3632,
+      "eval_samples_per_second": 250.183,
+      "eval_steps_per_second": 31.273,
+      "eval_weighted_f1": 0.4648136683073095,
+      "eval_weighted_precision": 0.4663299729027011,
+      "eval_weighted_recall": 0.4996031746031746,
+      "step": 500
+    },
+    {
+      "epoch": 0.08730505178776936,
+      "grad_norm": 32.64176940917969,
+      "learning_rate": 2.1626984126984128e-06,
+      "loss": 1.6452,
+      "step": 550
+    },
+    {
+      "epoch": 0.09524187467756658,
+      "grad_norm": 26.72295379638672,
+      "learning_rate": 2.361111111111111e-06,
+      "loss": 1.5402,
+      "step": 600
+    },
+    {
+      "epoch": 0.10317869756736378,
+      "grad_norm": 33.789302825927734,
+      "learning_rate": 2.5595238095238095e-06,
+      "loss": 1.4583,
+      "step": 650
+    },
+    {
+      "epoch": 0.111115520457161,
+      "grad_norm": 40.20470428466797,
+      "learning_rate": 2.7579365079365083e-06,
+      "loss": 1.338,
+      "step": 700
+    },
+    {
+      "epoch": 0.11905234334695822,
+      "grad_norm": 26.672527313232422,
+      "learning_rate": 2.9563492063492066e-06,
+      "loss": 1.2152,
+      "step": 750
+    },
+    {
+      "epoch": 0.12698916623675544,
+      "grad_norm": 26.061119079589844,
+      "learning_rate": 3.154761904761905e-06,
+      "loss": 1.1735,
+      "step": 800
+    },
+    {
+      "epoch": 0.13492598912655265,
+      "grad_norm": 49.5890007019043,
+      "learning_rate": 3.3531746031746034e-06,
+      "loss": 1.059,
+      "step": 850
+    },
+    {
+      "epoch": 0.14286281201634984,
+      "grad_norm": 34.52119827270508,
+      "learning_rate": 3.551587301587302e-06,
+      "loss": 0.9615,
+      "step": 900
+    },
+    {
+      "epoch": 0.15079963490614706,
+      "grad_norm": 41.126792907714844,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 1.0047,
+      "step": 950
+    },
+    {
+      "epoch": 0.15873645779594428,
+      "grad_norm": 38.969017028808594,
+      "learning_rate": 3.9484126984126985e-06,
+      "loss": 0.8757,
+      "step": 1000
+    },
+    {
+      "epoch": 0.15873645779594428,
+      "eval_accuracy": 0.7561111111111111,
+      "eval_loss": 0.8336577415466309,
+      "eval_macro_f1": 0.3742962053337755,
+      "eval_macro_precision": 0.4440270170632841,
+      "eval_macro_recall": 0.36141326244744065,
+      "eval_micro_f1": 0.7561111111111111,
+      "eval_micro_precision": 0.7561111111111111,
+      "eval_micro_recall": 0.7561111111111111,
+      "eval_runtime": 50.358,
+      "eval_samples_per_second": 250.208,
+      "eval_steps_per_second": 31.276,
+      "eval_weighted_f1": 0.7347115811510523,
+      "eval_weighted_precision": 0.739827037528709,
+      "eval_weighted_recall": 0.7561111111111111,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1666732806857415,
+      "grad_norm": 41.12884521484375,
+      "learning_rate": 4.146825396825397e-06,
+      "loss": 0.8476,
+      "step": 1050
+    },
+    {
+      "epoch": 0.17461010357553872,
+      "grad_norm": 20.529150009155273,
+      "learning_rate": 4.345238095238096e-06,
+      "loss": 0.799,
+      "step": 1100
+    },
+    {
+      "epoch": 0.18254692646533593,
+      "grad_norm": 31.87129783630371,
+      "learning_rate": 4.543650793650794e-06,
+      "loss": 0.8004,
+      "step": 1150
+    },
+    {
+      "epoch": 0.19048374935513315,
+      "grad_norm": 32.46973419189453,
+      "learning_rate": 4.742063492063492e-06,
+      "loss": 0.7508,
+      "step": 1200
+    },
+    {
+      "epoch": 0.19842057224493034,
+      "grad_norm": 23.948808670043945,
+      "learning_rate": 4.940476190476191e-06,
+      "loss": 0.6721,
+      "step": 1250
+    },
+    {
+      "epoch": 0.20635739513472756,
+      "grad_norm": 31.930994033813477,
+      "learning_rate": 5.138888888888889e-06,
+      "loss": 0.6309,
+      "step": 1300
+    },
+    {
+      "epoch": 0.21429421802452478,
+      "grad_norm": 22.865070343017578,
+      "learning_rate": 5.337301587301587e-06,
+      "loss": 0.6133,
+      "step": 1350
+    },
+    {
+      "epoch": 0.222231040914322,
+      "grad_norm": 27.975875854492188,
+      "learning_rate": 5.535714285714286e-06,
+      "loss": 0.6472,
+      "step": 1400
+    },
+    {
+      "epoch": 0.23016786380411922,
+      "grad_norm": 17.48501968383789,
+      "learning_rate": 5.734126984126984e-06,
+      "loss": 0.5141,
+      "step": 1450
+    },
+    {
+      "epoch": 0.23810468669391643,
+      "grad_norm": 8.121456146240234,
+      "learning_rate": 5.932539682539683e-06,
+      "loss": 0.6094,
+      "step": 1500
+    },
+    {
+      "epoch": 0.23810468669391643,
+      "eval_accuracy": 0.8311111111111111,
+      "eval_loss": 0.5750956535339355,
+      "eval_macro_f1": 0.5919220318807322,
+      "eval_macro_precision": 0.6615551279169473,
+      "eval_macro_recall": 0.5725095316587115,
+      "eval_micro_f1": 0.8311111111111111,
+      "eval_micro_precision": 0.8311111111111111,
+      "eval_micro_recall": 0.8311111111111111,
+      "eval_runtime": 50.1384,
+      "eval_samples_per_second": 251.304,
+      "eval_steps_per_second": 31.413,
+      "eval_weighted_f1": 0.8258660663106083,
+      "eval_weighted_precision": 0.8313369703842328,
+      "eval_weighted_recall": 0.8311111111111111,
+      "step": 1500
+    },
+    {
+      "epoch": 0.24604150958371365,
+      "grad_norm": 52.50892639160156,
+      "learning_rate": 6.130952380952382e-06,
+      "loss": 0.5746,
+      "step": 1550
+    },
+    {
+      "epoch": 0.25397833247351087,
+      "grad_norm": 21.188108444213867,
+      "learning_rate": 6.32936507936508e-06,
+      "loss": 0.6786,
+      "step": 1600
+    },
+    {
+      "epoch": 0.26191515536330806,
+      "grad_norm": 37.73038101196289,
+      "learning_rate": 6.5277777777777784e-06,
+      "loss": 0.5529,
+      "step": 1650
+    },
+    {
+      "epoch": 0.2698519782531053,
+      "grad_norm": 31.049955368041992,
+      "learning_rate": 6.726190476190477e-06,
+      "loss": 0.5307,
+      "step": 1700
+    },
+    {
+      "epoch": 0.2777888011429025,
+      "grad_norm": 13.330388069152832,
+      "learning_rate": 6.924603174603175e-06,
+      "loss": 0.5919,
+      "step": 1750
+    },
+    {
+      "epoch": 0.2857256240326997,
+      "grad_norm": 29.72938346862793,
+      "learning_rate": 7.123015873015874e-06,
+      "loss": 0.5222,
+      "step": 1800
+    },
+    {
+      "epoch": 0.29366244692249693,
+      "grad_norm": 20.42317008972168,
+      "learning_rate": 7.321428571428572e-06,
+      "loss": 0.4796,
+      "step": 1850
+    },
+    {
+      "epoch": 0.3015992698122941,
+      "grad_norm": 29.127174377441406,
+      "learning_rate": 7.519841269841271e-06,
+      "loss": 0.499,
+      "step": 1900
+    },
+    {
+      "epoch": 0.30953609270209137,
+      "grad_norm": 18.9819278717041,
+      "learning_rate": 7.718253968253969e-06,
+      "loss": 0.4897,
+      "step": 1950
+    },
+    {
+      "epoch": 0.31747291559188856,
+      "grad_norm": 34.057342529296875,
+      "learning_rate": 7.916666666666667e-06,
+      "loss": 0.4944,
+      "step": 2000
+    },
+    {
+      "epoch": 0.31747291559188856,
+      "eval_accuracy": 0.8685714285714285,
+      "eval_loss": 0.43678638339042664,
+      "eval_macro_f1": 0.704721130059806,
+      "eval_macro_precision": 0.8084736830738437,
+      "eval_macro_recall": 0.6569171685565485,
+      "eval_micro_f1": 0.8685714285714285,
+      "eval_micro_precision": 0.8685714285714285,
+      "eval_micro_recall": 0.8685714285714285,
+      "eval_runtime": 54.0525,
+      "eval_samples_per_second": 233.107,
+      "eval_steps_per_second": 29.138,
+      "eval_weighted_f1": 0.8650850638163624,
+      "eval_weighted_precision": 0.8704622381617023,
+      "eval_weighted_recall": 0.8685714285714285,
+      "step": 2000
+    },
+    {
+      "epoch": 0.3254097384816858,
+      "grad_norm": 32.23432159423828,
+      "learning_rate": 8.115079365079366e-06,
+      "loss": 0.4503,
+      "step": 2050
+    },
+    {
+      "epoch": 0.333346561371483,
+      "grad_norm": 19.20964813232422,
+      "learning_rate": 8.313492063492064e-06,
+      "loss": 0.4846,
+      "step": 2100
+    },
+    {
+      "epoch": 0.3412833842612802,
+      "grad_norm": 22.326303482055664,
+      "learning_rate": 8.511904761904762e-06,
+      "loss": 0.4545,
+      "step": 2150
+    },
+    {
+      "epoch": 0.34922020715107743,
+      "grad_norm": 39.84929275512695,
+      "learning_rate": 8.710317460317462e-06,
+      "loss": 0.4542,
+      "step": 2200
+    },
+    {
+      "epoch": 0.3571570300408746,
+      "grad_norm": 20.93601417541504,
+      "learning_rate": 8.90873015873016e-06,
+      "loss": 0.5178,
+      "step": 2250
+    },
+    {
+      "epoch": 0.36509385293067187,
+      "grad_norm": 8.917716979980469,
+      "learning_rate": 9.107142857142858e-06,
+      "loss": 0.3944,
+      "step": 2300
+    },
+    {
+      "epoch": 0.37303067582046906,
+      "grad_norm": 14.799666404724121,
+      "learning_rate": 9.305555555555557e-06,
+      "loss": 0.3669,
+      "step": 2350
+    },
+    {
+      "epoch": 0.3809674987102663,
+      "grad_norm": 25.97331428527832,
+      "learning_rate": 9.503968253968255e-06,
+      "loss": 0.3843,
+      "step": 2400
+    },
+    {
+      "epoch": 0.3889043216000635,
+      "grad_norm": 13.379817008972168,
+      "learning_rate": 9.702380952380953e-06,
+      "loss": 0.3901,
+      "step": 2450
+    },
+    {
+      "epoch": 0.3968411444898607,
+      "grad_norm": 18.588956832885742,
+      "learning_rate": 9.900793650793653e-06,
+      "loss": 0.3913,
+      "step": 2500
+    },
+    {
+      "epoch": 0.3968411444898607,
+      "eval_accuracy": 0.8876190476190476,
+      "eval_loss": 0.3863752782344818,
+      "eval_macro_f1": 0.7708534077145469,
+      "eval_macro_precision": 0.8424204092431431,
+      "eval_macro_recall": 0.7602036693520349,
+      "eval_micro_f1": 0.8876190476190476,
+      "eval_micro_precision": 0.8876190476190476,
+      "eval_micro_recall": 0.8876190476190476,
+      "eval_runtime": 58.1835,
+      "eval_samples_per_second": 216.556,
+      "eval_steps_per_second": 27.07,
+      "eval_weighted_f1": 0.8866826747603135,
+      "eval_weighted_precision": 0.894918351088462,
+      "eval_weighted_recall": 0.8876190476190476,
+      "step": 2500
+    },
+    {
+      "epoch": 0.40477796737965793,
+      "grad_norm": 26.919750213623047,
+      "learning_rate": 9.988975127888518e-06,
+      "loss": 0.3492,
+      "step": 2550
+    },
+    {
+      "epoch": 0.4127147902694551,
+      "grad_norm": 24.233909606933594,
+      "learning_rate": 9.96692538366555e-06,
+      "loss": 0.4253,
+      "step": 2600
+    },
+    {
+      "epoch": 0.42065161315925237,
+      "grad_norm": 8.280208587646484,
+      "learning_rate": 9.944875639442584e-06,
+      "loss": 0.3656,
+      "step": 2650
+    },
+    {
+      "epoch": 0.42858843604904956,
+      "grad_norm": 0.9590346813201904,
+      "learning_rate": 9.922825895219616e-06,
+      "loss": 0.4059,
+      "step": 2700
+    },
+    {
+      "epoch": 0.4365252589388468,
+      "grad_norm": 6.837221145629883,
+      "learning_rate": 9.90077615099665e-06,
+      "loss": 0.3804,
+      "step": 2750
+    },
+    {
+      "epoch": 0.444462081828644,
+      "grad_norm": 26.77887535095215,
+      "learning_rate": 9.878726406773682e-06,
+      "loss": 0.3876,
+      "step": 2800
+    },
+    {
+      "epoch": 0.4523989047184412,
+      "grad_norm": 1.809899091720581,
+      "learning_rate": 9.856676662550716e-06,
+      "loss": 0.3292,
+      "step": 2850
+    },
+    {
+      "epoch": 0.46033572760823843,
+      "grad_norm": 24.776517868041992,
+      "learning_rate": 9.83462691832775e-06,
+      "loss": 0.3909,
+      "step": 2900
+    },
+    {
+      "epoch": 0.4682725504980356,
+      "grad_norm": 33.20071792602539,
+      "learning_rate": 9.812577174104782e-06,
+      "loss": 0.3396,
+      "step": 2950
+    },
+    {
+      "epoch": 0.47620937338783287,
+      "grad_norm": 25.043102264404297,
+      "learning_rate": 9.790527429881814e-06,
+      "loss": 0.3553,
+      "step": 3000
+    },
+    {
+      "epoch": 0.47620937338783287,
+      "eval_accuracy": 0.9049206349206349,
+      "eval_loss": 0.3749255836009979,
+      "eval_macro_f1": 0.83146311466218,
+      "eval_macro_precision": 0.8822616279099216,
+      "eval_macro_recall": 0.8044703495421244,
+      "eval_micro_f1": 0.9049206349206349,
+      "eval_micro_precision": 0.9049206349206349,
+      "eval_micro_recall": 0.9049206349206349,
+      "eval_runtime": 51.2979,
+      "eval_samples_per_second": 245.624,
+      "eval_steps_per_second": 30.703,
+      "eval_weighted_f1": 0.9042368915226878,
+      "eval_weighted_precision": 0.9091398319969811,
+      "eval_weighted_recall": 0.9049206349206349,
+      "step": 3000
+    },
+    {
+      "epoch": 0.48414619627763006,
+      "grad_norm": 14.722782135009766,
+      "learning_rate": 9.768477685658848e-06,
+      "loss": 0.3143,
+      "step": 3050
+    },
+    {
+      "epoch": 0.4920830191674273,
+      "grad_norm": 14.118471145629883,
+      "learning_rate": 9.74642794143588e-06,
+      "loss": 0.398,
+      "step": 3100
+    },
+    {
+      "epoch": 0.5000198420572245,
+      "grad_norm": 12.885693550109863,
+      "learning_rate": 9.724378197212914e-06,
+      "loss": 0.278,
+      "step": 3150
+    },
+    {
+      "epoch": 0.5079566649470217,
+      "grad_norm": 11.429601669311523,
+      "learning_rate": 9.702328452989946e-06,
+      "loss": 0.3612,
+      "step": 3200
+    },
+    {
+      "epoch": 0.5158934878368189,
+      "grad_norm": 11.400711059570312,
+      "learning_rate": 9.68027870876698e-06,
+      "loss": 0.276,
+      "step": 3250
+    },
+    {
+      "epoch": 0.5238303107266161,
+      "grad_norm": 9.139787673950195,
+      "learning_rate": 9.658228964544012e-06,
+      "loss": 0.2994,
+      "step": 3300
+    },
+    {
+      "epoch": 0.5317671336164134,
+      "grad_norm": 22.870264053344727,
+      "learning_rate": 9.636179220321044e-06,
+      "loss": 0.3084,
+      "step": 3350
+    },
+    {
+      "epoch": 0.5397039565062106,
+      "grad_norm": 12.863465309143066,
+      "learning_rate": 9.61412947609808e-06,
+      "loss": 0.3557,
+      "step": 3400
+    },
+    {
+      "epoch": 0.5476407793960077,
+      "grad_norm": 22.37148094177246,
+      "learning_rate": 9.592079731875112e-06,
+      "loss": 0.3276,
+      "step": 3450
+    },
+    {
+      "epoch": 0.555577602285805,
+      "grad_norm": 11.454520225524902,
+      "learning_rate": 9.570029987652144e-06,
+      "loss": 0.308,
+      "step": 3500
+    },
+    {
+      "epoch": 0.555577602285805,
+      "eval_accuracy": 0.9016666666666666,
+      "eval_loss": 0.34068959951400757,
+      "eval_macro_f1": 0.852643844128256,
+      "eval_macro_precision": 0.8869002266347256,
+      "eval_macro_recall": 0.8344066000317897,
+      "eval_micro_f1": 0.9016666666666666,
+      "eval_micro_precision": 0.9016666666666666,
+      "eval_micro_recall": 0.9016666666666666,
+      "eval_runtime": 55.1712,
+      "eval_samples_per_second": 228.38,
+      "eval_steps_per_second": 28.547,
+      "eval_weighted_f1": 0.9056553567113477,
+      "eval_weighted_precision": 0.9166483950530401,
+      "eval_weighted_recall": 0.9016666666666666,
+      "step": 3500
+    },
+    {
+      "epoch": 0.5635144251756022,
+      "grad_norm": 10.13657283782959,
+      "learning_rate": 9.547980243429178e-06,
+      "loss": 0.2969,
+      "step": 3550
+    },
+    {
+      "epoch": 0.5714512480653994,
+      "grad_norm": 8.981364250183105,
+      "learning_rate": 9.52593049920621e-06,
+      "loss": 0.2571,
+      "step": 3600
+    },
+    {
+      "epoch": 0.5793880709551966,
+      "grad_norm": 16.016128540039062,
+      "learning_rate": 9.503880754983244e-06,
+      "loss": 0.3322,
+      "step": 3650
+    },
+    {
+      "epoch": 0.5873248938449939,
+      "grad_norm": 5.5154242515563965,
+      "learning_rate": 9.481831010760276e-06,
+      "loss": 0.3231,
+      "step": 3700
+    },
+    {
+      "epoch": 0.5952617167347911,
+      "grad_norm": 4.189787864685059,
+      "learning_rate": 9.45978126653731e-06,
+      "loss": 0.3301,
+      "step": 3750
+    },
+    {
+      "epoch": 0.6031985396245882,
+      "grad_norm": 9.305127143859863,
+      "learning_rate": 9.437731522314342e-06,
+      "loss": 0.2832,
+      "step": 3800
+    },
+    {
+      "epoch": 0.6111353625143855,
+      "grad_norm": 32.7320442199707,
+      "learning_rate": 9.415681778091374e-06,
+      "loss": 0.2736,
+      "step": 3850
+    },
+    {
+      "epoch": 0.6190721854041827,
+      "grad_norm": 12.206543922424316,
+      "learning_rate": 9.393632033868408e-06,
+      "loss": 0.2191,
+      "step": 3900
+    },
+    {
+      "epoch": 0.6270090082939799,
+      "grad_norm": 8.348316192626953,
+      "learning_rate": 9.371582289645441e-06,
+      "loss": 0.2834,
+      "step": 3950
+    },
+    {
+      "epoch": 0.6349458311837771,
+      "grad_norm": 12.099419593811035,
+      "learning_rate": 9.349532545422474e-06,
+      "loss": 0.2687,
+      "step": 4000
+    },
+    {
+      "epoch": 0.6349458311837771,
+      "eval_accuracy": 0.9234126984126985,
+      "eval_loss": 0.27018845081329346,
+      "eval_macro_f1": 0.8629184237562926,
+      "eval_macro_precision": 0.896103415544095,
+      "eval_macro_recall": 0.8427102453711095,
+      "eval_micro_f1": 0.9234126984126985,
+      "eval_micro_precision": 0.9234126984126985,
+      "eval_micro_recall": 0.9234126984126985,
+      "eval_runtime": 45.4117,
+      "eval_samples_per_second": 277.462,
+      "eval_steps_per_second": 34.683,
+      "eval_weighted_f1": 0.9225121978819937,
+      "eval_weighted_precision": 0.9237214610954837,
+      "eval_weighted_recall": 0.9234126984126985,
+      "step": 4000
+    },
+    {
+      "epoch": 0.6428826540735744,
+      "grad_norm": 6.140594959259033,
+      "learning_rate": 9.327482801199507e-06,
+      "loss": 0.2299,
+      "step": 4050
+    },
+    {
+      "epoch": 0.6508194769633716,
+      "grad_norm": 13.922382354736328,
+      "learning_rate": 9.30543305697654e-06,
+      "loss": 0.3092,
+      "step": 4100
+    },
+    {
+      "epoch": 0.6587562998531687,
+      "grad_norm": 21.830659866333008,
+      "learning_rate": 9.283383312753573e-06,
+      "loss": 0.2174,
+      "step": 4150
+    },
+    {
+      "epoch": 0.666693122742966,
+      "grad_norm": 5.917385101318359,
+      "learning_rate": 9.261333568530606e-06,
+      "loss": 0.3323,
+      "step": 4200
+    },
+    {
+      "epoch": 0.6746299456327632,
+      "grad_norm": 11.365561485290527,
+      "learning_rate": 9.23928382430764e-06,
+      "loss": 0.2597,
+      "step": 4250
+    },
+    {
+      "epoch": 0.6825667685225604,
+      "grad_norm": 9.686914443969727,
+      "learning_rate": 9.217234080084672e-06,
+      "loss": 0.2466,
+      "step": 4300
+    },
+    {
+      "epoch": 0.6905035914123576,
+      "grad_norm": 12.17849063873291,
+      "learning_rate": 9.195184335861704e-06,
+      "loss": 0.2752,
+      "step": 4350
+    },
+    {
+      "epoch": 0.6984404143021549,
+      "grad_norm": 16.642663955688477,
+      "learning_rate": 9.173134591638737e-06,
+      "loss": 0.2695,
+      "step": 4400
+    },
+    {
+      "epoch": 0.7063772371919521,
+      "grad_norm": 3.786797046661377,
+      "learning_rate": 9.151084847415771e-06,
+      "loss": 0.2697,
+      "step": 4450
+    },
+    {
+      "epoch": 0.7143140600817492,
+      "grad_norm": 13.046645164489746,
+      "learning_rate": 9.129035103192803e-06,
+      "loss": 0.3057,
+      "step": 4500
+    },
+    {
+      "epoch": 0.7143140600817492,
+      "eval_accuracy": 0.9284126984126985,
+      "eval_loss": 0.25139176845550537,
+      "eval_macro_f1": 0.8858543195103097,
+      "eval_macro_precision": 0.9001288293009205,
+      "eval_macro_recall": 0.8748161720063445,
+      "eval_micro_f1": 0.9284126984126985,
+      "eval_micro_precision": 0.9284126984126985,
+      "eval_micro_recall": 0.9284126984126985,
+      "eval_runtime": 65.3815,
+      "eval_samples_per_second": 192.715,
+      "eval_steps_per_second": 24.089,
+      "eval_weighted_f1": 0.9284553859947943,
+      "eval_weighted_precision": 0.9293969786750571,
+      "eval_weighted_recall": 0.9284126984126985,
+      "step": 4500
+    },
+    {
+      "epoch": 0.7222508829715465,
+      "grad_norm": 21.798290252685547,
+      "learning_rate": 9.106985358969837e-06,
+      "loss": 0.3534,
+      "step": 4550
+    },
+    {
+      "epoch": 0.7301877058613437,
+      "grad_norm": 1.7543187141418457,
+      "learning_rate": 9.08493561474687e-06,
+      "loss": 0.2373,
+      "step": 4600
+    },
+    {
+      "epoch": 0.7381245287511409,
+      "grad_norm": 17.317163467407227,
+      "learning_rate": 9.062885870523903e-06,
+      "loss": 0.2644,
+      "step": 4650
+    },
+    {
+      "epoch": 0.7460613516409381,
+      "grad_norm": 6.609444618225098,
+      "learning_rate": 9.040836126300935e-06,
+      "loss": 0.2586,
+      "step": 4700
+    },
+    {
+      "epoch": 0.7539981745307354,
+      "grad_norm": 6.7774176597595215,
+      "learning_rate": 9.01878638207797e-06,
+      "loss": 0.2228,
+      "step": 4750
+    },
+    {
+      "epoch": 0.7619349974205326,
+      "grad_norm": 2.356161117553711,
+      "learning_rate": 8.996736637855001e-06,
+      "loss": 0.2082,
+      "step": 4800
+    },
+    {
+      "epoch": 0.7698718203103297,
+      "grad_norm": 17.036968231201172,
+      "learning_rate": 8.974686893632034e-06,
+      "loss": 0.3077,
+      "step": 4850
+    },
+    {
+      "epoch": 0.777808643200127,
+      "grad_norm": 8.356477737426758,
+      "learning_rate": 8.952637149409067e-06,
+      "loss": 0.2918,
+      "step": 4900
+    },
+    {
+      "epoch": 0.7857454660899242,
+      "grad_norm": 2.3111681938171387,
+      "learning_rate": 8.930587405186101e-06,
+      "loss": 0.2871,
+      "step": 4950
+    },
+    {
+      "epoch": 0.7936822889797214,
+      "grad_norm": 9.850006103515625,
+      "learning_rate": 8.908537660963133e-06,
+      "loss": 0.2467,
+      "step": 5000
+    },
+    {
+      "epoch": 0.7936822889797214,
+      "eval_accuracy": 0.9337301587301587,
+      "eval_loss": 0.24514073133468628,
+      "eval_macro_f1": 0.8968955709213094,
+      "eval_macro_precision": 0.9239801593127872,
+      "eval_macro_recall": 0.8775447083795846,
+      "eval_micro_f1": 0.9337301587301587,
+      "eval_micro_precision": 0.9337301587301587,
+      "eval_micro_recall": 0.9337301587301587,
+      "eval_runtime": 55.1353,
+      "eval_samples_per_second": 228.529,
+      "eval_steps_per_second": 28.566,
+      "eval_weighted_f1": 0.9333867608778208,
+      "eval_weighted_precision": 0.9342365821570298,
+      "eval_weighted_recall": 0.9337301587301587,
+      "step": 5000
+    },
+    {
+      "epoch": 0.8016191118695186,
+      "grad_norm": 3.655590295791626,
+      "learning_rate": 8.886487916740166e-06,
+      "loss": 0.2314,
+      "step": 5050
+    },
+    {
+      "epoch": 0.8095559347593159,
+      "grad_norm": 1.2928906679153442,
+      "learning_rate": 8.8644381725172e-06,
+      "loss": 0.2448,
+      "step": 5100
+    },
+    {
+      "epoch": 0.8174927576491131,
+      "grad_norm": 1.8197067975997925,
+      "learning_rate": 8.842388428294233e-06,
+      "loss": 0.1976,
+      "step": 5150
+    },
+    {
+      "epoch": 0.8254295805389102,
+      "grad_norm": 0.5001784563064575,
+      "learning_rate": 8.820338684071265e-06,
+      "loss": 0.2934,
+      "step": 5200
+    },
+    {
+      "epoch": 0.8333664034287075,
+      "grad_norm": 1.4087576866149902,
+      "learning_rate": 8.7982889398483e-06,
+      "loss": 0.1564,
+      "step": 5250
+    },
+    {
+      "epoch": 0.8413032263185047,
+      "grad_norm": 16.14033317565918,
+      "learning_rate": 8.776239195625331e-06,
+      "loss": 0.2161,
+      "step": 5300
+    },
+    {
+      "epoch": 0.8492400492083019,
+      "grad_norm": 15.914708137512207,
+      "learning_rate": 8.754189451402363e-06,
+      "loss": 0.1935,
+      "step": 5350
+    },
+    {
+      "epoch": 0.8571768720980991,
+      "grad_norm": 0.31940773129463196,
+      "learning_rate": 8.732139707179397e-06,
+      "loss": 0.2073,
+      "step": 5400
+    },
+    {
+      "epoch": 0.8651136949878964,
+      "grad_norm": 15.5699462890625,
+      "learning_rate": 8.710089962956431e-06,
+      "loss": 0.2305,
+      "step": 5450
+    },
+    {
+      "epoch": 0.8730505178776936,
+      "grad_norm": 14.775138854980469,
+      "learning_rate": 8.688040218733463e-06,
+      "loss": 0.1914,
+      "step": 5500
+    },
+    {
+      "epoch": 0.8730505178776936,
+      "eval_accuracy": 0.9361904761904762,
+      "eval_loss": 0.2405300885438919,
+      "eval_macro_f1": 0.8852157589128977,
+      "eval_macro_precision": 0.8966852505493809,
+      "eval_macro_recall": 0.8879515092023682,
+      "eval_micro_f1": 0.9361904761904762,
+      "eval_micro_precision": 0.9361904761904762,
+      "eval_micro_recall": 0.9361904761904762,
+      "eval_runtime": 50.0976,
+      "eval_samples_per_second": 251.509,
+      "eval_steps_per_second": 31.439,
+      "eval_weighted_f1": 0.9373120231008413,
+      "eval_weighted_precision": 0.940336582906886,
+      "eval_weighted_recall": 0.9361904761904762,
+      "step": 5500
+    },
+    {
+      "epoch": 0.8809873407674907,
+      "grad_norm": 8.058653831481934,
+      "learning_rate": 8.665990474510495e-06,
+      "loss": 0.1996,
+      "step": 5550
+    },
+    {
+      "epoch": 0.888924163657288,
+      "grad_norm": 1.231372594833374,
+      "learning_rate": 8.64394073028753e-06,
+      "loss": 0.1763,
+      "step": 5600
+    },
+    {
+      "epoch": 0.8968609865470852,
+      "grad_norm": 25.968311309814453,
+      "learning_rate": 8.621890986064563e-06,
+      "loss": 0.219,
+      "step": 5650
+    },
+    {
+      "epoch": 0.9047978094368824,
+      "grad_norm": 23.313941955566406,
+      "learning_rate": 8.599841241841595e-06,
+      "loss": 0.2194,
+      "step": 5700
+    },
+    {
+      "epoch": 0.9127346323266796,
+      "grad_norm": 11.84118366241455,
+      "learning_rate": 8.577791497618629e-06,
+      "loss": 0.2463,
+      "step": 5750
+    },
+    {
+      "epoch": 0.9206714552164769,
+      "grad_norm": 18.897754669189453,
+      "learning_rate": 8.555741753395661e-06,
+      "loss": 0.2125,
+      "step": 5800
+    },
+    {
+      "epoch": 0.9286082781062741,
+      "grad_norm": 19.018964767456055,
+      "learning_rate": 8.533692009172693e-06,
+      "loss": 0.2493,
+      "step": 5850
+    },
+    {
+      "epoch": 0.9365451009960712,
+      "grad_norm": 2.0846264362335205,
+      "learning_rate": 8.511642264949727e-06,
+      "loss": 0.1844,
+      "step": 5900
+    },
+    {
+      "epoch": 0.9444819238858685,
+      "grad_norm": 10.483113288879395,
+      "learning_rate": 8.489592520726761e-06,
+      "loss": 0.2421,
+      "step": 5950
+    },
+    {
+      "epoch": 0.9524187467756657,
+      "grad_norm": 4.168255805969238,
+      "learning_rate": 8.467542776503793e-06,
+      "loss": 0.151,
+      "step": 6000
+    },
+    {
+      "epoch": 0.9524187467756657,
+      "eval_accuracy": 0.941984126984127,
+      "eval_loss": 0.2432594746351242,
+      "eval_macro_f1": 0.9103364146476316,
+      "eval_macro_precision": 0.9298312623198658,
+      "eval_macro_recall": 0.8954247781763811,
+      "eval_micro_f1": 0.941984126984127,
+      "eval_micro_precision": 0.941984126984127,
+      "eval_micro_recall": 0.941984126984127,
+      "eval_runtime": 49.9297,
+      "eval_samples_per_second": 252.355,
+      "eval_steps_per_second": 31.544,
+      "eval_weighted_f1": 0.9417305390461214,
+      "eval_weighted_precision": 0.9424591918065203,
+      "eval_weighted_recall": 0.941984126984127,
+      "step": 6000
+    },
+    {
+      "epoch": 0.9603555696654629,
+      "grad_norm": 16.349781036376953,
+      "learning_rate": 8.445493032280825e-06,
+      "loss": 0.2176,
+      "step": 6050
+    },
+    {
+      "epoch": 0.9682923925552601,
+      "grad_norm": 1.1434494256973267,
+      "learning_rate": 8.423443288057859e-06,
+      "loss": 0.2075,
+      "step": 6100
+    },
+    {
+      "epoch": 0.9762292154450574,
+      "grad_norm": 24.730289459228516,
+      "learning_rate": 8.401393543834893e-06,
+      "loss": 0.234,
+      "step": 6150
+    },
+    {
+      "epoch": 0.9841660383348546,
+      "grad_norm": 11.54680061340332,
+      "learning_rate": 8.379343799611925e-06,
+      "loss": 0.264,
+      "step": 6200
+    },
+    {
+      "epoch": 0.9921028612246517,
+      "grad_norm": 3.650730848312378,
+      "learning_rate": 8.357294055388959e-06,
+      "loss": 0.1762,
+      "step": 6250
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.1564973592758179,
+      "learning_rate": 8.33568530605045e-06,
+      "loss": 0.1916,
+      "step": 6300
+    },
+    {
+      "epoch": 1.0079368228897971,
+      "grad_norm": 21.63304328918457,
+      "learning_rate": 8.313635561827483e-06,
+      "loss": 0.0975,
+      "step": 6350
+    },
+    {
+      "epoch": 1.0158736457795945,
+      "grad_norm": 23.74657440185547,
+      "learning_rate": 8.291585817604516e-06,
+      "loss": 0.1956,
+      "step": 6400
+    },
+    {
+      "epoch": 1.0238104686693916,
+      "grad_norm": 0.026358341798186302,
+      "learning_rate": 8.26953607338155e-06,
+      "loss": 0.124,
+      "step": 6450
+    },
+    {
+      "epoch": 1.0317472915591888,
+      "grad_norm": 0.22795121371746063,
+      "learning_rate": 8.247486329158582e-06,
+      "loss": 0.1537,
+      "step": 6500
+    },
+    {
+      "epoch": 1.0317472915591888,
+      "eval_accuracy": 0.9456349206349206,
+      "eval_loss": 0.22848264873027802,
+      "eval_macro_f1": 0.909507315203273,
+      "eval_macro_precision": 0.9296991239988623,
+      "eval_macro_recall": 0.8961337328597196,
+      "eval_micro_f1": 0.9456349206349206,
+      "eval_micro_precision": 0.9456349206349206,
+      "eval_micro_recall": 0.9456349206349206,
+      "eval_runtime": 48.6038,
+      "eval_samples_per_second": 259.239,
+      "eval_steps_per_second": 32.405,
+      "eval_weighted_f1": 0.9454117579717388,
+      "eval_weighted_precision": 0.9459939034429095,
+      "eval_weighted_recall": 0.9456349206349206,
+      "step": 6500
+    },
+    {
+      "epoch": 1.0396841144489861,
+      "grad_norm": 0.07601787894964218,
+      "learning_rate": 8.225436584935616e-06,
+      "loss": 0.179,
+      "step": 6550
+    },
+    {
+      "epoch": 1.0476209373387833,
+      "grad_norm": 0.028876453638076782,
+      "learning_rate": 8.203386840712648e-06,
+      "loss": 0.1539,
+      "step": 6600
+    },
+    {
+      "epoch": 1.0555577602285804,
+      "grad_norm": 1.7003881931304932,
+      "learning_rate": 8.18133709648968e-06,
+      "loss": 0.106,
+      "step": 6650
+    },
+    {
+      "epoch": 1.0634945831183777,
+      "grad_norm": 0.29444923996925354,
+      "learning_rate": 8.159287352266714e-06,
+      "loss": 0.1515,
+      "step": 6700
+    },
+    {
+      "epoch": 1.0714314060081749,
+      "grad_norm": 22.709447860717773,
+      "learning_rate": 8.137237608043748e-06,
+      "loss": 0.178,
+      "step": 6750
+    },
+    {
+      "epoch": 1.0793682288979722,
+      "grad_norm": 20.404773712158203,
+      "learning_rate": 8.11518786382078e-06,
+      "loss": 0.1122,
+      "step": 6800
+    },
+    {
+      "epoch": 1.0873050517877694,
+      "grad_norm": 4.415476322174072,
+      "learning_rate": 8.093138119597813e-06,
+      "loss": 0.0948,
+      "step": 6850
+    },
+    {
+      "epoch": 1.0952418746775665,
+      "grad_norm": 22.34918975830078,
+      "learning_rate": 8.071088375374846e-06,
+      "loss": 0.2517,
+      "step": 6900
+    },
+    {
+      "epoch": 1.1031786975673639,
+      "grad_norm": 16.816791534423828,
+      "learning_rate": 8.04903863115188e-06,
+      "loss": 0.1379,
+      "step": 6950
+    },
+    {
+      "epoch": 1.111115520457161,
+      "grad_norm": 12.884002685546875,
+      "learning_rate": 8.026988886928912e-06,
+      "loss": 0.1574,
+      "step": 7000
+    },
+    {
+      "epoch": 1.111115520457161,
+      "eval_accuracy": 0.9484126984126984,
+      "eval_loss": 0.20768418908119202,
+      "eval_macro_f1": 0.9125337197557731,
+      "eval_macro_precision": 0.9190348121850224,
+      "eval_macro_recall": 0.9113443135094732,
+      "eval_micro_f1": 0.9484126984126984,
+      "eval_micro_precision": 0.9484126984126984,
+      "eval_micro_recall": 0.9484126984126984,
+      "eval_runtime": 51.0028,
+      "eval_samples_per_second": 247.045,
+      "eval_steps_per_second": 30.881,
+      "eval_weighted_f1": 0.9481753539970462,
+      "eval_weighted_precision": 0.9487731067127975,
+      "eval_weighted_recall": 0.9484126984126984,
+      "step": 7000
+    },
+    {
+      "epoch": 1.1190523433469581,
+      "grad_norm": 1.1013977527618408,
+      "learning_rate": 8.004939142705944e-06,
+      "loss": 0.1002,
+      "step": 7050
+    },
+    {
+      "epoch": 1.1269891662367555,
+      "grad_norm": 26.22823143005371,
+      "learning_rate": 7.982889398482978e-06,
+      "loss": 0.1197,
+      "step": 7100
+    },
+    {
+      "epoch": 1.1349259891265526,
+      "grad_norm": 4.3102593421936035,
+      "learning_rate": 7.96083965426001e-06,
+      "loss": 0.1526,
+      "step": 7150
+    },
+    {
+      "epoch": 1.1428628120163498,
+      "grad_norm": 0.07969073206186295,
+      "learning_rate": 7.938789910037044e-06,
+      "loss": 0.1292,
+      "step": 7200
+    },
+    {
+      "epoch": 1.1507996349061471,
+      "grad_norm": 16.321470260620117,
+      "learning_rate": 7.916740165814078e-06,
+      "loss": 0.1533,
+      "step": 7250
+    },
+    {
+      "epoch": 1.1587364577959443,
+      "grad_norm": 48.117523193359375,
+      "learning_rate": 7.89469042159111e-06,
+      "loss": 0.0941,
+      "step": 7300
+    },
+    {
+      "epoch": 1.1666732806857416,
+      "grad_norm": 0.04194047302007675,
+      "learning_rate": 7.872640677368142e-06,
+      "loss": 0.1223,
+      "step": 7350
+    },
+    {
+      "epoch": 1.1746101035755387,
+      "grad_norm": 2.087705612182617,
+      "learning_rate": 7.850590933145176e-06,
+      "loss": 0.1968,
+      "step": 7400
+    },
+    {
+      "epoch": 1.1825469264653359,
+      "grad_norm": 1.1391172409057617,
+      "learning_rate": 7.82854118892221e-06,
+      "loss": 0.1461,
+      "step": 7450
+    },
+    {
+      "epoch": 1.1904837493551332,
+      "grad_norm": 0.41210365295410156,
+      "learning_rate": 7.806491444699242e-06,
+      "loss": 0.1406,
+      "step": 7500
+    },
+    {
+      "epoch": 1.1904837493551332,
+      "eval_accuracy": 0.9442857142857143,
+      "eval_loss": 0.2635294198989868,
+      "eval_macro_f1": 0.9057940212609807,
+      "eval_macro_precision": 0.911842211214998,
+      "eval_macro_recall": 0.9069588500794394,
+      "eval_micro_f1": 0.9442857142857143,
+      "eval_micro_precision": 0.9442857142857143,
+      "eval_micro_recall": 0.9442857142857143,
+      "eval_runtime": 53.2935,
+      "eval_samples_per_second": 236.427,
+      "eval_steps_per_second": 29.553,
+      "eval_weighted_f1": 0.9443844599149406,
+      "eval_weighted_precision": 0.945921955299144,
+      "eval_weighted_recall": 0.9442857142857143,
+      "step": 7500
+    },
+    {
+      "epoch": 1.1984205722449304,
+      "grad_norm": 1.8356152772903442,
+      "learning_rate": 7.784441700476274e-06,
+      "loss": 0.1513,
+      "step": 7550
+    },
+    {
+      "epoch": 1.2063573951347275,
+      "grad_norm": 2.8009727001190186,
+      "learning_rate": 7.762391956253308e-06,
+      "loss": 0.0998,
+      "step": 7600
+    },
+    {
+      "epoch": 1.2142942180245249,
+      "grad_norm": 0.1467098593711853,
+      "learning_rate": 7.74034221203034e-06,
+      "loss": 0.1355,
+      "step": 7650
+    },
+    {
+      "epoch": 1.222231040914322,
+      "grad_norm": 0.005296769086271524,
+      "learning_rate": 7.718292467807374e-06,
+      "loss": 0.144,
+      "step": 7700
+    },
+    {
+      "epoch": 1.2301678638041191,
+      "grad_norm": 20.8825626373291,
+      "learning_rate": 7.696242723584408e-06,
+      "loss": 0.1296,
+      "step": 7750
+    },
+    {
+      "epoch": 1.2381046866939165,
+      "grad_norm": 11.227338790893555,
+      "learning_rate": 7.67419297936144e-06,
+      "loss": 0.1395,
+      "step": 7800
+    },
+    {
+      "epoch": 1.2460415095837136,
+      "grad_norm": 0.8737878203392029,
+      "learning_rate": 7.652143235138472e-06,
+      "loss": 0.1257,
+      "step": 7850
+    },
+    {
+      "epoch": 1.253978332473511,
+      "grad_norm": 8.432129859924316,
+      "learning_rate": 7.630093490915506e-06,
+      "loss": 0.1778,
+      "step": 7900
+    },
+    {
+      "epoch": 1.2619151553633081,
+      "grad_norm": 18.125516891479492,
+      "learning_rate": 7.608043746692539e-06,
+      "loss": 0.1222,
+      "step": 7950
+    },
+    {
+      "epoch": 1.2698519782531053,
+      "grad_norm": 0.770737886428833,
+      "learning_rate": 7.585994002469572e-06,
+      "loss": 0.147,
+      "step": 8000
+    },
+    {
+      "epoch": 1.2698519782531053,
+      "eval_accuracy": 0.9494444444444444,
+      "eval_loss": 0.21279270946979523,
+      "eval_macro_f1": 0.9101912478331972,
+      "eval_macro_precision": 0.9273715052159579,
+      "eval_macro_recall": 0.901402436657646,
+      "eval_micro_f1": 0.9494444444444444,
+      "eval_micro_precision": 0.9494444444444444,
+      "eval_micro_recall": 0.9494444444444444,
+      "eval_runtime": 48.6141,
+      "eval_samples_per_second": 259.184,
+      "eval_steps_per_second": 32.398,
+      "eval_weighted_f1": 0.9491690418479931,
+      "eval_weighted_precision": 0.9504200289531454,
+      "eval_weighted_recall": 0.9494444444444444,
+      "step": 8000
+    },
+    {
+      "epoch": 1.2777888011429024,
+      "grad_norm": 2.4760684967041016,
+      "learning_rate": 7.563944258246605e-06,
+      "loss": 0.134,
+      "step": 8050
+    },
+    {
+      "epoch": 1.2857256240326997,
+      "grad_norm": 5.634617328643799,
+      "learning_rate": 7.541894514023638e-06,
+      "loss": 0.1468,
+      "step": 8100
+    },
+    {
+      "epoch": 1.2936624469224969,
+      "grad_norm": 9.858975410461426,
+      "learning_rate": 7.51984476980067e-06,
+      "loss": 0.1531,
+      "step": 8150
+    },
+    {
+      "epoch": 1.3015992698122942,
+      "grad_norm": 1.8135714530944824,
+      "learning_rate": 7.497795025577704e-06,
+      "loss": 0.1038,
+      "step": 8200
+    },
+    {
+      "epoch": 1.3095360927020914,
+      "grad_norm": 1.2420053482055664,
+      "learning_rate": 7.475745281354737e-06,
+      "loss": 0.1224,
+      "step": 8250
+    },
+    {
+      "epoch": 1.3174729155918885,
+      "grad_norm": 0.03956647962331772,
+      "learning_rate": 7.45369553713177e-06,
+      "loss": 0.1161,
+      "step": 8300
+    },
+    {
+      "epoch": 1.3254097384816859,
+      "grad_norm": 18.216032028198242,
+      "learning_rate": 7.4320867877932624e-06,
+      "loss": 0.1121,
+      "step": 8350
+    },
+    {
+      "epoch": 1.333346561371483,
+      "grad_norm": 15.097914695739746,
+      "learning_rate": 7.4100370435702954e-06,
+      "loss": 0.1237,
+      "step": 8400
+    },
+    {
+      "epoch": 1.3412833842612801,
+      "grad_norm": 0.09787029772996902,
+      "learning_rate": 7.3879872993473276e-06,
+      "loss": 0.1129,
+      "step": 8450
+    },
+    {
+      "epoch": 1.3492202071510775,
+      "grad_norm": 28.132888793945312,
+      "learning_rate": 7.365937555124361e-06,
+      "loss": 0.0994,
+      "step": 8500
+    },
+    {
+      "epoch": 1.3492202071510775,
+      "eval_accuracy": 0.9530952380952381,
+      "eval_loss": 0.21746815741062164,
+      "eval_macro_f1": 0.9156584913591659,
+      "eval_macro_precision": 0.9241711528138504,
+      "eval_macro_recall": 0.9152950181018766,
+      "eval_micro_f1": 0.9530952380952381,
+      "eval_micro_precision": 0.9530952380952381,
+      "eval_micro_recall": 0.9530952380952381,
+      "eval_runtime": 49.6088,
+      "eval_samples_per_second": 253.987,
+      "eval_steps_per_second": 31.748,
+      "eval_weighted_f1": 0.9528707065770654,
+      "eval_weighted_precision": 0.9533906189807044,
+      "eval_weighted_recall": 0.9530952380952381,
+      "step": 8500
+    },
+    {
+      "epoch": 1.3571570300408746,
+      "grad_norm": 0.09891881048679352,
+      "learning_rate": 7.343887810901394e-06,
+      "loss": 0.1426,
+      "step": 8550
+    },
+    {
+      "epoch": 1.3650938529306718,
+      "grad_norm": 0.06244092062115669,
+      "learning_rate": 7.321838066678427e-06,
+      "loss": 0.1188,
+      "step": 8600
+    },
+    {
+      "epoch": 1.3730306758204691,
+      "grad_norm": 9.950495719909668,
+      "learning_rate": 7.29978832245546e-06,
+      "loss": 0.1068,
+      "step": 8650
+    },
+    {
+      "epoch": 1.3809674987102663,
+      "grad_norm": 21.47039031982422,
+      "learning_rate": 7.2777385782324925e-06,
+      "loss": 0.1311,
+      "step": 8700
+    },
+    {
+      "epoch": 1.3889043216000636,
+      "grad_norm": 0.21059951186180115,
+      "learning_rate": 7.255688834009526e-06,
+      "loss": 0.1359,
+      "step": 8750
+    },
+    {
+      "epoch": 1.3968411444898607,
+      "grad_norm": 0.07483835518360138,
+      "learning_rate": 7.233639089786559e-06,
+      "loss": 0.1312,
+      "step": 8800
+    },
+    {
+      "epoch": 1.4047779673796579,
+      "grad_norm": 7.540125370025635,
+      "learning_rate": 7.211589345563592e-06,
+      "loss": 0.1883,
+      "step": 8850
+    },
+    {
+      "epoch": 1.412714790269455,
+      "grad_norm": 0.05421961471438408,
+      "learning_rate": 7.1895396013406245e-06,
+      "loss": 0.1313,
+      "step": 8900
+    },
+    {
+      "epoch": 1.4206516131592524,
+      "grad_norm": 25.743249893188477,
+      "learning_rate": 7.1674898571176575e-06,
+      "loss": 0.1866,
+      "step": 8950
+    },
+    {
+      "epoch": 1.4285884360490495,
+      "grad_norm": 15.702727317810059,
+      "learning_rate": 7.145440112894691e-06,
+      "loss": 0.1323,
+      "step": 9000
+    },
+    {
+      "epoch": 1.4285884360490495,
+      "eval_accuracy": 0.9544444444444444,
+      "eval_loss": 0.21344760060310364,
+      "eval_macro_f1": 0.9316646003398935,
+      "eval_macro_precision": 0.9481060096209216,
+      "eval_macro_recall": 0.9195331693514776,
+      "eval_micro_f1": 0.9544444444444444,
+      "eval_micro_precision": 0.9544444444444444,
+      "eval_micro_recall": 0.9544444444444444,
+      "eval_runtime": 49.9732,
+      "eval_samples_per_second": 252.135,
+      "eval_steps_per_second": 31.517,
+      "eval_weighted_f1": 0.9541756739346082,
+      "eval_weighted_precision": 0.9548010972085054,
+      "eval_weighted_recall": 0.9544444444444444,
+      "step": 9000
+    },
+    {
+      "epoch": 1.4365252589388469,
+      "grad_norm": 12.405553817749023,
+      "learning_rate": 7.123390368671724e-06,
+      "loss": 0.1462,
+      "step": 9050
+    },
+    {
+      "epoch": 1.444462081828644,
+      "grad_norm": 36.17345428466797,
+      "learning_rate": 7.101340624448757e-06,
+      "loss": 0.1289,
+      "step": 9100
+    },
+    {
+      "epoch": 1.4523989047184411,
+      "grad_norm": 21.679637908935547,
+      "learning_rate": 7.0792908802257894e-06,
+      "loss": 0.1316,
+      "step": 9150
+    },
+    {
+      "epoch": 1.4603357276082385,
+      "grad_norm": 13.572071075439453,
+      "learning_rate": 7.057241136002822e-06,
+      "loss": 0.1104,
+      "step": 9200
+    },
+    {
+      "epoch": 1.4682725504980356,
+      "grad_norm": 13.087688446044922,
+      "learning_rate": 7.035191391779856e-06,
+      "loss": 0.135,
+      "step": 9250
+    },
+    {
+      "epoch": 1.476209373387833,
+      "grad_norm": 0.5103034377098083,
+      "learning_rate": 7.013141647556889e-06,
+      "loss": 0.094,
+      "step": 9300
+    },
+    {
+      "epoch": 1.4841461962776301,
+      "grad_norm": 0.2972245514392853,
+      "learning_rate": 6.991091903333922e-06,
+      "loss": 0.1333,
+      "step": 9350
+    },
+    {
+      "epoch": 1.4920830191674272,
+      "grad_norm": 19.705307006835938,
+      "learning_rate": 6.969042159110954e-06,
+      "loss": 0.1014,
+      "step": 9400
+    },
+    {
+      "epoch": 1.5000198420572244,
+      "grad_norm": 17.829219818115234,
+      "learning_rate": 6.946992414887987e-06,
+      "loss": 0.0851,
+      "step": 9450
+    },
+    {
+      "epoch": 1.5079566649470217,
+      "grad_norm": 17.99626922607422,
+      "learning_rate": 6.924942670665021e-06,
+      "loss": 0.2011,
+      "step": 9500
+    },
+    {
+      "epoch": 1.5079566649470217,
+      "eval_accuracy": 0.9568253968253968,
+      "eval_loss": 0.19153176248073578,
+      "eval_macro_f1": 0.921610631007778,
+      "eval_macro_precision": 0.9404398497869574,
+      "eval_macro_recall": 0.905748060531071,
+      "eval_micro_f1": 0.9568253968253968,
+      "eval_micro_precision": 0.9568253968253968,
+      "eval_micro_recall": 0.9568253968253968,
+      "eval_runtime": 49.0787,
+      "eval_samples_per_second": 256.731,
+      "eval_steps_per_second": 32.091,
+      "eval_weighted_f1": 0.9566293314445607,
+      "eval_weighted_precision": 0.9568515996032587,
+      "eval_weighted_recall": 0.9568253968253968,
+      "step": 9500
+    },
+    {
+      "epoch": 1.5158934878368189,
+      "grad_norm": 3.1612911224365234,
+      "learning_rate": 6.902892926442054e-06,
+      "loss": 0.1468,
+      "step": 9550
+    },
+    {
+      "epoch": 1.5238303107266162,
+      "grad_norm": 1.6461035013198853,
+      "learning_rate": 6.880843182219087e-06,
+      "loss": 0.088,
+      "step": 9600
+    },
+    {
+      "epoch": 1.5317671336164134,
+      "grad_norm": 0.029310036450624466,
+      "learning_rate": 6.858793437996119e-06,
+      "loss": 0.0723,
+      "step": 9650
+    },
+    {
+      "epoch": 1.5397039565062105,
+      "grad_norm": 0.6573673486709595,
+      "learning_rate": 6.836743693773152e-06,
+      "loss": 0.131,
+      "step": 9700
+    },
+    {
+      "epoch": 1.5476407793960076,
+      "grad_norm": 0.0239870622754097,
+      "learning_rate": 6.814693949550186e-06,
+      "loss": 0.0713,
+      "step": 9750
+    },
+    {
+      "epoch": 1.555577602285805,
+      "grad_norm": 0.05512286350131035,
+      "learning_rate": 6.792644205327219e-06,
+      "loss": 0.1083,
+      "step": 9800
+    },
+    {
+      "epoch": 1.5635144251756024,
+      "grad_norm": 8.742629051208496,
+      "learning_rate": 6.770594461104252e-06,
+      "loss": 0.125,
+      "step": 9850
+    },
+    {
+      "epoch": 1.5714512480653995,
+      "grad_norm": 0.23801352083683014,
+      "learning_rate": 6.748544716881284e-06,
+      "loss": 0.1548,
+      "step": 9900
+    },
+    {
+      "epoch": 1.5793880709551966,
+      "grad_norm": 0.06436382979154587,
+      "learning_rate": 6.726494972658317e-06,
+      "loss": 0.1131,
+      "step": 9950
+    },
+    {
+      "epoch": 1.5873248938449938,
+      "grad_norm": 0.07321972399950027,
+      "learning_rate": 6.704445228435351e-06,
+      "loss": 0.1171,
+      "step": 10000
+    },
+    {
+      "epoch": 1.5873248938449938,
+      "eval_accuracy": 0.959920634920635,
+      "eval_loss": 0.19577036798000336,
+      "eval_macro_f1": 0.9235742546691592,
+      "eval_macro_precision": 0.9273628410860382,
+      "eval_macro_recall": 0.9271792802963574,
+      "eval_micro_f1": 0.959920634920635,
+      "eval_micro_precision": 0.959920634920635,
+      "eval_micro_recall": 0.959920634920635,
+      "eval_runtime": 60.9394,
+      "eval_samples_per_second": 206.763,
+      "eval_steps_per_second": 25.845,
+      "eval_weighted_f1": 0.9599028886198321,
+      "eval_weighted_precision": 0.960305322421165,
+      "eval_weighted_recall": 0.959920634920635,
+      "step": 10000
+    },
+    {
+      "epoch": 1.5952617167347911,
+      "grad_norm": 8.259378433227539,
+      "learning_rate": 6.682395484212384e-06,
+      "loss": 0.0956,
+      "step": 10050
+    },
+    {
+      "epoch": 1.6031985396245882,
+      "grad_norm": 0.8749714493751526,
+      "learning_rate": 6.660345739989417e-06,
+      "loss": 0.1274,
+      "step": 10100
+    },
+    {
+      "epoch": 1.6111353625143856,
+      "grad_norm": 3.652017116546631,
+      "learning_rate": 6.638295995766449e-06,
+      "loss": 0.1042,
+      "step": 10150
+    },
+    {
+      "epoch": 1.6190721854041827,
+      "grad_norm": 1.1253262758255005,
+      "learning_rate": 6.616246251543482e-06,
+      "loss": 0.0914,
+      "step": 10200
+    },
+    {
+      "epoch": 1.6270090082939799,
+      "grad_norm": 0.6077162623405457,
+      "learning_rate": 6.594196507320516e-06,
+      "loss": 0.1225,
+      "step": 10250
+    },
+    {
+      "epoch": 1.634945831183777,
+      "grad_norm": 0.008563055656850338,
+      "learning_rate": 6.572146763097549e-06,
+      "loss": 0.1171,
+      "step": 10300
+    },
+    {
+      "epoch": 1.6428826540735744,
+      "grad_norm": 2.094672441482544,
+      "learning_rate": 6.550097018874582e-06,
+      "loss": 0.0798,
+      "step": 10350
+    },
+    {
+      "epoch": 1.6508194769633717,
+      "grad_norm": 0.12600164115428925,
+      "learning_rate": 6.528047274651614e-06,
+      "loss": 0.1071,
+      "step": 10400
+    },
+    {
+      "epoch": 1.6587562998531689,
+      "grad_norm": 0.3863985240459442,
+      "learning_rate": 6.505997530428647e-06,
+      "loss": 0.1485,
+      "step": 10450
+    },
+    {
+      "epoch": 1.666693122742966,
+      "grad_norm": 0.005566800944507122,
+      "learning_rate": 6.483947786205681e-06,
+      "loss": 0.1567,
+      "step": 10500
+    },
+    {
+      "epoch": 1.666693122742966,
+      "eval_accuracy": 0.960952380952381,
+      "eval_loss": 0.17503954470157623,
+      "eval_macro_f1": 0.9170932665272469,
+      "eval_macro_precision": 0.9116456444630479,
+      "eval_macro_recall": 0.9260931287646225,
+      "eval_micro_f1": 0.960952380952381,
+      "eval_micro_precision": 0.960952380952381,
+      "eval_micro_recall": 0.960952380952381,
+      "eval_runtime": 45.8116,
+      "eval_samples_per_second": 275.04,
+      "eval_steps_per_second": 34.38,
+      "eval_weighted_f1": 0.9609014768411174,
+      "eval_weighted_precision": 0.9611475161993005,
+      "eval_weighted_recall": 0.960952380952381,
+      "step": 10500
+    },
+    {
+      "epoch": 1.6746299456327631,
+      "grad_norm": 17.83533477783203,
+      "learning_rate": 6.461898041982714e-06,
+      "loss": 0.112,
+      "step": 10550
+    },
+    {
+      "epoch": 1.6825667685225603,
+      "grad_norm": 0.016817472875118256,
+      "learning_rate": 6.439848297759746e-06,
+      "loss": 0.1179,
+      "step": 10600
+    },
+    {
+      "epoch": 1.6905035914123576,
+      "grad_norm": 0.0496453195810318,
+      "learning_rate": 6.417798553536779e-06,
+      "loss": 0.1354,
+      "step": 10650
+    },
+    {
+      "epoch": 1.698440414302155,
+      "grad_norm": 24.511754989624023,
+      "learning_rate": 6.395748809313812e-06,
+      "loss": 0.1083,
+      "step": 10700
+    },
+    {
+      "epoch": 1.7063772371919521,
+      "grad_norm": 0.014569321647286415,
+      "learning_rate": 6.373699065090846e-06,
+      "loss": 0.0928,
+      "step": 10750
+    },
+    {
+      "epoch": 1.7143140600817492,
+      "grad_norm": 7.14270544052124,
+      "learning_rate": 6.351649320867879e-06,
+      "loss": 0.1354,
+      "step": 10800
+    },
+    {
+      "epoch": 1.7222508829715464,
+      "grad_norm": 0.039672911167144775,
+      "learning_rate": 6.329599576644911e-06,
+      "loss": 0.0879,
+      "step": 10850
+    },
+    {
+      "epoch": 1.7301877058613437,
+      "grad_norm": 0.11405997723340988,
+      "learning_rate": 6.307549832421944e-06,
+      "loss": 0.1525,
+      "step": 10900
+    },
+    {
+      "epoch": 1.7381245287511409,
+      "grad_norm": 1.720201015472412,
+      "learning_rate": 6.285500088198977e-06,
+      "loss": 0.1257,
+      "step": 10950
+    },
+    {
+      "epoch": 1.7460613516409382,
+      "grad_norm": 21.198429107666016,
+      "learning_rate": 6.263891338860469e-06,
+      "loss": 0.0848,
+      "step": 11000
+    },
+    {
+      "epoch": 1.7460613516409382,
+      "eval_accuracy": 0.9624603174603175,
+      "eval_loss": 0.17708104848861694,
+      "eval_macro_f1": 0.9319695940105357,
+      "eval_macro_precision": 0.9462998611668713,
+      "eval_macro_recall": 0.9234620410796542,
+      "eval_micro_f1": 0.9624603174603175,
+      "eval_micro_precision": 0.9624603174603175,
+      "eval_micro_recall": 0.9624603174603175,
+      "eval_runtime": 45.9838,
+      "eval_samples_per_second": 274.009,
+      "eval_steps_per_second": 34.251,
+      "eval_weighted_f1": 0.9623559428261603,
+      "eval_weighted_precision": 0.962723564261779,
+      "eval_weighted_recall": 0.9624603174603175,
+      "step": 11000
+    },
+    {
+      "epoch": 1.7539981745307354,
+      "grad_norm": 19.329214096069336,
+      "learning_rate": 6.241841594637502e-06,
+      "loss": 0.0969,
+      "step": 11050
+    },
+    {
+      "epoch": 1.7619349974205325,
+      "grad_norm": 0.17044760286808014,
+      "learning_rate": 6.219791850414536e-06,
+      "loss": 0.1143,
+      "step": 11100
+    },
+    {
+      "epoch": 1.7698718203103296,
+      "grad_norm": 20.753198623657227,
+      "learning_rate": 6.197742106191568e-06,
+      "loss": 0.1619,
+      "step": 11150
+    },
+    {
+      "epoch": 1.777808643200127,
+      "grad_norm": 0.05526827648282051,
+      "learning_rate": 6.175692361968601e-06,
+      "loss": 0.1297,
+      "step": 11200
+    },
+    {
+      "epoch": 1.7857454660899243,
+      "grad_norm": 13.288789749145508,
+      "learning_rate": 6.153642617745634e-06,
+      "loss": 0.0892,
+      "step": 11250
+    },
+    {
+      "epoch": 1.7936822889797215,
+      "grad_norm": 50.681602478027344,
+      "learning_rate": 6.131592873522668e-06,
+      "loss": 0.1218,
+      "step": 11300
+    },
+    {
+      "epoch": 1.8016191118695186,
+      "grad_norm": 18.851547241210938,
+      "learning_rate": 6.109543129299701e-06,
+      "loss": 0.1046,
+      "step": 11350
+    },
+    {
+      "epoch": 1.8095559347593158,
+      "grad_norm": 20.649171829223633,
+      "learning_rate": 6.087493385076733e-06,
+      "loss": 0.1481,
+      "step": 11400
+    },
+    {
+      "epoch": 1.817492757649113,
+      "grad_norm": 17.76620101928711,
+      "learning_rate": 6.065443640853766e-06,
+      "loss": 0.0807,
+      "step": 11450
+    },
+    {
+      "epoch": 1.8254295805389102,
+      "grad_norm": 9.238646507263184,
+      "learning_rate": 6.043393896630799e-06,
+      "loss": 0.0973,
+      "step": 11500
+    },
+    {
+      "epoch": 1.8254295805389102,
+      "eval_accuracy": 0.9632539682539683,
+      "eval_loss": 0.17640231549739838,
+      "eval_macro_f1": 0.9383726570780089,
+      "eval_macro_precision": 0.9361695510382104,
+      "eval_macro_recall": 0.9420156784780935,
+      "eval_micro_f1": 0.9632539682539683,
+      "eval_micro_precision": 0.9632539682539683,
+      "eval_micro_recall": 0.9632539682539683,
+      "eval_runtime": 46.3964,
+      "eval_samples_per_second": 271.573,
+      "eval_steps_per_second": 33.947,
+      "eval_weighted_f1": 0.9633197036349216,
+      "eval_weighted_precision": 0.9636434749852494,
+      "eval_weighted_recall": 0.9632539682539683,
+      "step": 11500
+    },
+    {
+      "epoch": 1.8333664034287076,
+      "grad_norm": 0.8048087954521179,
+      "learning_rate": 6.021344152407833e-06,
+      "loss": 0.1329,
+      "step": 11550
+    },
+    {
+      "epoch": 1.8413032263185047,
+      "grad_norm": 0.3991595506668091,
+      "learning_rate": 5.999294408184866e-06,
+      "loss": 0.1303,
+      "step": 11600
+    },
+    {
+      "epoch": 1.8492400492083019,
+      "grad_norm": 0.10322093963623047,
+      "learning_rate": 5.977244663961898e-06,
+      "loss": 0.1261,
+      "step": 11650
+    },
+    {
+      "epoch": 1.857176872098099,
+      "grad_norm": 7.725561141967773,
+      "learning_rate": 5.955194919738931e-06,
+      "loss": 0.0816,
+      "step": 11700
+    },
+    {
+      "epoch": 1.8651136949878964,
+      "grad_norm": 5.625209331512451,
+      "learning_rate": 5.933145175515964e-06,
+      "loss": 0.0861,
+      "step": 11750
+    },
+    {
+      "epoch": 1.8730505178776937,
+      "grad_norm": 0.045761458575725555,
+      "learning_rate": 5.911095431292998e-06,
+      "loss": 0.0718,
+      "step": 11800
+    },
+    {
+      "epoch": 1.8809873407674909,
+      "grad_norm": 15.094813346862793,
+      "learning_rate": 5.889045687070031e-06,
+      "loss": 0.1316,
+      "step": 11850
+    },
+    {
+      "epoch": 1.888924163657288,
+      "grad_norm": 0.025717562064528465,
+      "learning_rate": 5.866995942847063e-06,
+      "loss": 0.12,
+      "step": 11900
+    },
+    {
+      "epoch": 1.8968609865470851,
+      "grad_norm": 22.716604232788086,
+      "learning_rate": 5.844946198624096e-06,
+      "loss": 0.1371,
+      "step": 11950
+    },
+    {
+      "epoch": 1.9047978094368823,
+      "grad_norm": 20.683067321777344,
+      "learning_rate": 5.822896454401129e-06,
+      "loss": 0.1331,
+      "step": 12000
+    },
+    {
+      "epoch": 1.9047978094368823,
+      "eval_accuracy": 0.9657142857142857,
+      "eval_loss": 0.16351820528507233,
+      "eval_macro_f1": 0.9426202066832035,
+      "eval_macro_precision": 0.9507506384547523,
+      "eval_macro_recall": 0.9366009019904619,
+      "eval_micro_f1": 0.9657142857142857,
+      "eval_micro_precision": 0.9657142857142857,
+      "eval_micro_recall": 0.9657142857142857,
+      "eval_runtime": 46.8483,
+      "eval_samples_per_second": 268.953,
+      "eval_steps_per_second": 33.619,
+      "eval_weighted_f1": 0.9656532805046822,
+      "eval_weighted_precision": 0.9659512006173707,
+      "eval_weighted_recall": 0.9657142857142857,
+      "step": 12000
+    },
+    {
+      "epoch": 1.9127346323266796,
+      "grad_norm": 0.001627567340619862,
+      "learning_rate": 5.800846710178163e-06,
+      "loss": 0.1079,
+      "step": 12050
+    },
+    {
+      "epoch": 1.920671455216477,
+      "grad_norm": 21.03553581237793,
+      "learning_rate": 5.778796965955196e-06,
+      "loss": 0.0558,
+      "step": 12100
+    },
+    {
+      "epoch": 1.928608278106274,
+      "grad_norm": 13.464014053344727,
+      "learning_rate": 5.756747221732228e-06,
+      "loss": 0.0959,
+      "step": 12150
+    },
+    {
+      "epoch": 1.9365451009960712,
+      "grad_norm": 17.52626609802246,
+      "learning_rate": 5.734697477509261e-06,
+      "loss": 0.0883,
+      "step": 12200
+    },
+    {
+      "epoch": 1.9444819238858684,
+      "grad_norm": 0.12921009957790375,
+      "learning_rate": 5.712647733286294e-06,
+      "loss": 0.0769,
+      "step": 12250
+    },
+    {
+      "epoch": 1.9524187467756657,
+      "grad_norm": 0.0050261239521205425,
+      "learning_rate": 5.690597989063328e-06,
+      "loss": 0.1439,
+      "step": 12300
+    },
+    {
+      "epoch": 1.9603555696654629,
+      "grad_norm": 18.023502349853516,
+      "learning_rate": 5.668548244840361e-06,
+      "loss": 0.0804,
+      "step": 12350
+    },
+    {
+      "epoch": 1.9682923925552602,
+      "grad_norm": 0.03103369101881981,
+      "learning_rate": 5.646498500617393e-06,
+      "loss": 0.0833,
+      "step": 12400
+    },
+    {
+      "epoch": 1.9762292154450574,
+      "grad_norm": 30.99793243408203,
+      "learning_rate": 5.624448756394426e-06,
+      "loss": 0.1744,
+      "step": 12450
+    },
+    {
+      "epoch": 1.9841660383348545,
+      "grad_norm": 0.04181884601712227,
+      "learning_rate": 5.602399012171459e-06,
+      "loss": 0.0629,
+      "step": 12500
+    },
+    {
+      "epoch": 1.9841660383348545,
+      "eval_accuracy": 0.9681746031746031,
+      "eval_loss": 0.1587391495704651,
+      "eval_macro_f1": 0.9466181342033713,
+      "eval_macro_precision": 0.947442786667393,
+      "eval_macro_recall": 0.9465717167703007,
+      "eval_micro_f1": 0.9681746031746031,
+      "eval_micro_precision": 0.9681746031746031,
+      "eval_micro_recall": 0.9681746031746031,
+      "eval_runtime": 66.4354,
+      "eval_samples_per_second": 189.658,
+      "eval_steps_per_second": 23.707,
+      "eval_weighted_f1": 0.9681650956489429,
+      "eval_weighted_precision": 0.9683707150909681,
+      "eval_weighted_recall": 0.9681746031746031,
+      "step": 12500
+    },
+    {
+      "epoch": 1.9921028612246516,
+      "grad_norm": 11.374281883239746,
+      "learning_rate": 5.580349267948493e-06,
+      "loss": 0.0867,
+      "step": 12550
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.029758449643850327,
+      "learning_rate": 5.558299523725525e-06,
+      "loss": 0.109,
+      "step": 12600
+    },
+    {
+      "epoch": 2.007936822889797,
+      "grad_norm": 0.002391190268099308,
+      "learning_rate": 5.536249779502558e-06,
+      "loss": 0.0186,
+      "step": 12650
+    },
+    {
+      "epoch": 2.0158736457795943,
+      "grad_norm": 12.729046821594238,
+      "learning_rate": 5.514200035279591e-06,
+      "loss": 0.0258,
+      "step": 12700
+    },
+    {
+      "epoch": 2.023810468669392,
+      "grad_norm": 0.012360978871583939,
+      "learning_rate": 5.492150291056624e-06,
+      "loss": 0.0406,
+      "step": 12750
+    },
+    {
+      "epoch": 2.031747291559189,
+      "grad_norm": 2.931403160095215,
+      "learning_rate": 5.470100546833658e-06,
+      "loss": 0.0323,
+      "step": 12800
+    },
+    {
+      "epoch": 2.039684114448986,
+      "grad_norm": 0.34426018595695496,
+      "learning_rate": 5.44805080261069e-06,
+      "loss": 0.0493,
+      "step": 12850
+    },
+    {
+      "epoch": 2.0476209373387833,
+      "grad_norm": 1.6615216732025146,
+      "learning_rate": 5.426001058387723e-06,
+      "loss": 0.0227,
+      "step": 12900
+    },
+    {
+      "epoch": 2.0555577602285804,
+      "grad_norm": 17.994659423828125,
+      "learning_rate": 5.403951314164756e-06,
+      "loss": 0.0575,
+      "step": 12950
+    },
+    {
+      "epoch": 2.0634945831183775,
+      "grad_norm": 0.002307410817593336,
+      "learning_rate": 5.381901569941789e-06,
+      "loss": 0.0724,
+      "step": 13000
+    },
+    {
+      "epoch": 2.0634945831183775,
+      "eval_accuracy": 0.9689682539682539,
+      "eval_loss": 0.16660018265247345,
+      "eval_macro_f1": 0.9431082887845699,
+      "eval_macro_precision": 0.9435320164238983,
+      "eval_macro_recall": 0.9440342115841122,
+      "eval_micro_f1": 0.9689682539682539,
+      "eval_micro_precision": 0.9689682539682539,
+      "eval_micro_recall": 0.9689682539682539,
+      "eval_runtime": 55.6815,
+      "eval_samples_per_second": 226.287,
+      "eval_steps_per_second": 28.286,
+      "eval_weighted_f1": 0.9690350615547291,
+      "eval_weighted_precision": 0.9692930831154812,
+      "eval_weighted_recall": 0.9689682539682539,
+      "step": 13000
+    },
+    {
+      "epoch": 2.071431406008175,
+      "grad_norm": 1.1325668096542358,
+      "learning_rate": 5.359851825718823e-06,
+      "loss": 0.0317,
+      "step": 13050
+    },
+    {
+      "epoch": 2.0793682288979722,
+      "grad_norm": 10.751199722290039,
+      "learning_rate": 5.337802081495855e-06,
+      "loss": 0.0382,
+      "step": 13100
+    },
+    {
+      "epoch": 2.0873050517877694,
+      "grad_norm": 0.05518977344036102,
+      "learning_rate": 5.315752337272888e-06,
+      "loss": 0.032,
+      "step": 13150
+    },
+    {
+      "epoch": 2.0952418746775665,
+      "grad_norm": 0.05851821228861809,
+      "learning_rate": 5.293702593049921e-06,
+      "loss": 0.0347,
+      "step": 13200
+    },
+    {
+      "epoch": 2.1031786975673636,
+      "grad_norm": 0.0005113780498504639,
+      "learning_rate": 5.272093843711413e-06,
+      "loss": 0.0298,
+      "step": 13250
+    },
+    {
+      "epoch": 2.1111155204571608,
+      "grad_norm": 0.018820350989699364,
+      "learning_rate": 5.250044099488446e-06,
+      "loss": 0.0497,
+      "step": 13300
+    },
+    {
+      "epoch": 2.1190523433469584,
+      "grad_norm": 0.9959418177604675,
+      "learning_rate": 5.2279943552654785e-06,
+      "loss": 0.046,
+      "step": 13350
+    },
+    {
+      "epoch": 2.1269891662367555,
+      "grad_norm": 0.7051542401313782,
+      "learning_rate": 5.205944611042512e-06,
+      "loss": 0.04,
+      "step": 13400
+    },
+    {
+      "epoch": 2.1349259891265526,
+      "grad_norm": 0.0022753493394702673,
+      "learning_rate": 5.183894866819545e-06,
+      "loss": 0.0173,
+      "step": 13450
+    },
+    {
+      "epoch": 2.1428628120163498,
+      "grad_norm": 0.0039760335348546505,
+      "learning_rate": 5.161845122596578e-06,
+      "loss": 0.0582,
+      "step": 13500
+    },
+    {
+      "epoch": 2.1428628120163498,
+      "eval_accuracy": 0.9687301587301588,
+      "eval_loss": 0.18048988282680511,
+      "eval_macro_f1": 0.9422923846160581,
+      "eval_macro_precision": 0.9532870485294106,
+      "eval_macro_recall": 0.934424229646169,
+      "eval_micro_f1": 0.9687301587301588,
+      "eval_micro_precision": 0.9687301587301588,
+      "eval_micro_recall": 0.9687301587301588,
+      "eval_runtime": 48.6144,
+      "eval_samples_per_second": 259.182,
+      "eval_steps_per_second": 32.398,
+      "eval_weighted_f1": 0.9686287726395876,
+      "eval_weighted_precision": 0.9689670561915049,
+      "eval_weighted_recall": 0.9687301587301588,
+      "step": 13500
+    },
+    {
+      "epoch": 2.150799634906147,
+      "grad_norm": 0.096370629966259,
+      "learning_rate": 5.139795378373611e-06,
+      "loss": 0.0522,
+      "step": 13550
+    },
+    {
+      "epoch": 2.1587364577959445,
+      "grad_norm": 0.013626059517264366,
+      "learning_rate": 5.117745634150643e-06,
+      "loss": 0.0128,
+      "step": 13600
+    },
+    {
+      "epoch": 2.1666732806857416,
+      "grad_norm": 0.003456405596807599,
+      "learning_rate": 5.095695889927677e-06,
+      "loss": 0.0333,
+      "step": 13650
+    },
+    {
+      "epoch": 2.1746101035755387,
+      "grad_norm": 20.14434051513672,
+      "learning_rate": 5.07364614570471e-06,
+      "loss": 0.0572,
+      "step": 13700
+    },
+    {
+      "epoch": 2.182546926465336,
+      "grad_norm": 0.009079753421247005,
+      "learning_rate": 5.051596401481743e-06,
+      "loss": 0.0439,
+      "step": 13750
+    },
+    {
+      "epoch": 2.190483749355133,
+      "grad_norm": 0.017065979540348053,
+      "learning_rate": 5.029546657258776e-06,
+      "loss": 0.0405,
+      "step": 13800
+    },
+    {
+      "epoch": 2.19842057224493,
+      "grad_norm": 0.001399728120304644,
+      "learning_rate": 5.007496913035808e-06,
+      "loss": 0.0567,
+      "step": 13850
+    },
+    {
+      "epoch": 2.2063573951347277,
+      "grad_norm": 0.013922719284892082,
+      "learning_rate": 4.985447168812842e-06,
+      "loss": 0.0484,
+      "step": 13900
+    },
+    {
+      "epoch": 2.214294218024525,
+      "grad_norm": 0.0006722013931721449,
+      "learning_rate": 4.963397424589875e-06,
+      "loss": 0.0089,
+      "step": 13950
+    },
+    {
+      "epoch": 2.222231040914322,
+      "grad_norm": 16.31502342224121,
+      "learning_rate": 4.941347680366908e-06,
+      "loss": 0.0182,
+      "step": 14000
+    },
+    {
+      "epoch": 2.222231040914322,
+      "eval_accuracy": 0.9709523809523809,
+      "eval_loss": 0.17584019899368286,
+      "eval_macro_f1": 0.9548845957042732,
+      "eval_macro_precision": 0.9616225902344291,
+      "eval_macro_recall": 0.9490903584972133,
+      "eval_micro_f1": 0.9709523809523809,
+      "eval_micro_precision": 0.9709523809523809,
+      "eval_micro_recall": 0.9709523809523809,
+      "eval_runtime": 65.6446,
+      "eval_samples_per_second": 191.943,
+      "eval_steps_per_second": 23.993,
+      "eval_weighted_f1": 0.9709045086147281,
+      "eval_weighted_precision": 0.9709764613384323,
+      "eval_weighted_recall": 0.9709523809523809,
+      "step": 14000
+    },
+    {
+      "epoch": 2.230167863804119,
+      "grad_norm": 49.83951187133789,
+      "learning_rate": 4.919297936143941e-06,
+      "loss": 0.0383,
+      "step": 14050
+    },
+    {
+      "epoch": 2.2381046866939163,
+      "grad_norm": 0.010581075213849545,
+      "learning_rate": 4.897248191920974e-06,
+      "loss": 0.0527,
+      "step": 14100
+    },
+    {
+      "epoch": 2.246041509583714,
+      "grad_norm": 5.595621109008789,
+      "learning_rate": 4.875198447698007e-06,
+      "loss": 0.0361,
+      "step": 14150
+    },
+    {
+      "epoch": 2.253978332473511,
+      "grad_norm": 11.858242988586426,
+      "learning_rate": 4.85314870347504e-06,
+      "loss": 0.023,
+      "step": 14200
+    },
+    {
+      "epoch": 2.261915155363308,
+      "grad_norm": 28.24439239501953,
+      "learning_rate": 4.831098959252073e-06,
+      "loss": 0.029,
+      "step": 14250
+    },
+    {
+      "epoch": 2.2698519782531053,
+      "grad_norm": 0.001131124678067863,
+      "learning_rate": 4.809049215029106e-06,
+      "loss": 0.0522,
+      "step": 14300
+    },
+    {
+      "epoch": 2.2777888011429024,
+      "grad_norm": 14.565289497375488,
+      "learning_rate": 4.786999470806139e-06,
+      "loss": 0.0353,
+      "step": 14350
+    },
+    {
+      "epoch": 2.2857256240326995,
+      "grad_norm": 1.4213881492614746,
+      "learning_rate": 4.764949726583172e-06,
+      "loss": 0.0252,
+      "step": 14400
+    },
+    {
+      "epoch": 2.293662446922497,
+      "grad_norm": 0.011964501813054085,
+      "learning_rate": 4.742899982360205e-06,
+      "loss": 0.0572,
+      "step": 14450
+    },
+    {
+      "epoch": 2.3015992698122942,
+      "grad_norm": 12.1846342086792,
+      "learning_rate": 4.720850238137238e-06,
+      "loss": 0.0295,
+      "step": 14500
+    },
+    {
+      "epoch": 2.3015992698122942,
+      "eval_accuracy": 0.9695238095238096,
+      "eval_loss": 0.177631214261055,
+      "eval_macro_f1": 0.9506224965454739,
+      "eval_macro_precision": 0.9661964085054093,
+      "eval_macro_recall": 0.9377514378889877,
+      "eval_micro_f1": 0.9695238095238096,
+      "eval_micro_precision": 0.9695238095238096,
+      "eval_micro_recall": 0.9695238095238096,
+      "eval_runtime": 55.0712,
+      "eval_samples_per_second": 228.795,
+      "eval_steps_per_second": 28.599,
+      "eval_weighted_f1": 0.9694153409312327,
+      "eval_weighted_precision": 0.9696628969782775,
+      "eval_weighted_recall": 0.9695238095238096,
+      "step": 14500
+    },
+    {
+      "epoch": 2.3095360927020914,
+      "grad_norm": 0.0022249706089496613,
+      "learning_rate": 4.698800493914271e-06,
+      "loss": 0.051,
+      "step": 14550
+    },
+    {
+      "epoch": 2.3174729155918885,
+      "grad_norm": 0.13238926231861115,
+      "learning_rate": 4.676750749691304e-06,
+      "loss": 0.0247,
+      "step": 14600
+    },
+    {
+      "epoch": 2.3254097384816856,
+      "grad_norm": 0.06873074173927307,
+      "learning_rate": 4.654701005468337e-06,
+      "loss": 0.0406,
+      "step": 14650
+    },
+    {
+      "epoch": 2.333346561371483,
+      "grad_norm": 1.4111828804016113,
+      "learning_rate": 4.63265126124537e-06,
+      "loss": 0.0509,
+      "step": 14700
+    },
+    {
+      "epoch": 2.3412833842612804,
+      "grad_norm": 0.05572796240448952,
+      "learning_rate": 4.610601517022403e-06,
+      "loss": 0.0443,
+      "step": 14750
+    },
+    {
+      "epoch": 2.3492202071510775,
+      "grad_norm": 0.13843700289726257,
+      "learning_rate": 4.588551772799436e-06,
+      "loss": 0.0322,
+      "step": 14800
+    },
+    {
+      "epoch": 2.3571570300408746,
+      "grad_norm": 22.532489776611328,
+      "learning_rate": 4.566502028576469e-06,
+      "loss": 0.0404,
+      "step": 14850
+    },
+    {
+      "epoch": 2.3650938529306718,
+      "grad_norm": 0.40451693534851074,
+      "learning_rate": 4.544452284353502e-06,
+      "loss": 0.0384,
+      "step": 14900
+    },
+    {
+      "epoch": 2.373030675820469,
+      "grad_norm": 0.00039770189323462546,
+      "learning_rate": 4.522402540130535e-06,
+      "loss": 0.0446,
+      "step": 14950
+    },
+    {
+      "epoch": 2.3809674987102665,
+      "grad_norm": 0.19481758773326874,
+      "learning_rate": 4.500352795907568e-06,
+      "loss": 0.0105,
+      "step": 15000
+    },
+    {
+      "epoch": 2.3809674987102665,
+      "eval_accuracy": 0.97,
+      "eval_loss": 0.17878225445747375,
+      "eval_macro_f1": 0.952375532557884,
+      "eval_macro_precision": 0.9642924920360566,
+      "eval_macro_recall": 0.9429523998338022,
+      "eval_micro_f1": 0.97,
+      "eval_micro_precision": 0.97,
+      "eval_micro_recall": 0.97,
+      "eval_runtime": 51.0169,
+      "eval_samples_per_second": 246.977,
+      "eval_steps_per_second": 30.872,
+      "eval_weighted_f1": 0.9699429015929246,
+      "eval_weighted_precision": 0.9701562195790373,
+      "eval_weighted_recall": 0.97,
+      "step": 15000
+    },
+    {
+      "epoch": 2.3889043216000636,
+      "grad_norm": 0.19946858286857605,
+      "learning_rate": 4.478303051684601e-06,
+      "loss": 0.0317,
+      "step": 15050
+    },
+    {
+      "epoch": 2.3968411444898607,
+      "grad_norm": 0.13299234211444855,
+      "learning_rate": 4.456253307461634e-06,
+      "loss": 0.0504,
+      "step": 15100
+    },
+    {
+      "epoch": 2.404777967379658,
+      "grad_norm": 0.254424512386322,
+      "learning_rate": 4.434203563238667e-06,
+      "loss": 0.04,
+      "step": 15150
+    },
+    {
+      "epoch": 2.412714790269455,
+      "grad_norm": 35.3300666809082,
+      "learning_rate": 4.4121538190157e-06,
+      "loss": 0.05,
+      "step": 15200
+    },
+    {
+      "epoch": 2.4206516131592526,
+      "grad_norm": 0.015775412321090698,
+      "learning_rate": 4.390104074792733e-06,
+      "loss": 0.064,
+      "step": 15250
+    },
+    {
+      "epoch": 2.4285884360490497,
+      "grad_norm": 0.0011088565224781632,
+      "learning_rate": 4.368054330569766e-06,
+      "loss": 0.0429,
+      "step": 15300
+    },
+    {
+      "epoch": 2.436525258938847,
+      "grad_norm": 0.16335545480251312,
+      "learning_rate": 4.346004586346799e-06,
+      "loss": 0.0305,
+      "step": 15350
+    },
+    {
+      "epoch": 2.444462081828644,
+      "grad_norm": 0.9601665735244751,
+      "learning_rate": 4.323954842123832e-06,
+      "loss": 0.0174,
+      "step": 15400
+    },
+    {
+      "epoch": 2.452398904718441,
+      "grad_norm": 0.11341444402933121,
+      "learning_rate": 4.301905097900865e-06,
+      "loss": 0.0298,
+      "step": 15450
+    },
+    {
+      "epoch": 2.4603357276082383,
+      "grad_norm": 74.64165496826172,
+      "learning_rate": 4.279855353677898e-06,
+      "loss": 0.0212,
+      "step": 15500
+    },
+    {
+      "epoch": 2.4603357276082383,
+      "eval_accuracy": 0.9695238095238096,
+      "eval_loss": 0.17210516333580017,
+      "eval_macro_f1": 0.9462875984373381,
+      "eval_macro_precision": 0.9580763571099492,
+      "eval_macro_recall": 0.9378514622035224,
+      "eval_micro_f1": 0.9695238095238096,
+      "eval_micro_precision": 0.9695238095238096,
+      "eval_micro_recall": 0.9695238095238096,
+      "eval_runtime": 49.2932,
+      "eval_samples_per_second": 255.613,
+      "eval_steps_per_second": 31.952,
+      "eval_weighted_f1": 0.9694666934898747,
+      "eval_weighted_precision": 0.9696918541700538,
+      "eval_weighted_recall": 0.9695238095238096,
+      "step": 15500
+    },
+    {
+      "epoch": 2.4682725504980354,
+      "grad_norm": 2.2245137691497803,
+      "learning_rate": 4.257805609454931e-06,
+      "loss": 0.0631,
+      "step": 15550
+    },
+    {
+      "epoch": 2.476209373387833,
+      "grad_norm": 5.462784290313721,
+      "learning_rate": 4.235755865231964e-06,
+      "loss": 0.0206,
+      "step": 15600
+    },
+    {
+      "epoch": 2.48414619627763,
+      "grad_norm": 0.0205392986536026,
+      "learning_rate": 4.213706121008997e-06,
+      "loss": 0.0195,
+      "step": 15650
+    },
+    {
+      "epoch": 2.4920830191674272,
+      "grad_norm": 0.003773665986955166,
+      "learning_rate": 4.19165637678603e-06,
+      "loss": 0.0336,
+      "step": 15700
+    },
+    {
+      "epoch": 2.5000198420572244,
+      "grad_norm": 0.04200325533747673,
+      "learning_rate": 4.169606632563063e-06,
+      "loss": 0.0408,
+      "step": 15750
+    },
+    {
+      "epoch": 2.507956664947022,
+      "grad_norm": 23.36740493774414,
+      "learning_rate": 4.147556888340096e-06,
+      "loss": 0.0178,
+      "step": 15800
+    },
+    {
+      "epoch": 2.515893487836819,
+      "grad_norm": 19.962722778320312,
+      "learning_rate": 4.125507144117129e-06,
+      "loss": 0.0347,
+      "step": 15850
+    },
+    {
+      "epoch": 2.5238303107266162,
+      "grad_norm": 11.505073547363281,
+      "learning_rate": 4.103457399894162e-06,
+      "loss": 0.0125,
+      "step": 15900
+    },
+    {
+      "epoch": 2.5317671336164134,
+      "grad_norm": 45.94857406616211,
+      "learning_rate": 4.081407655671195e-06,
+      "loss": 0.0536,
+      "step": 15950
+    },
+    {
+      "epoch": 2.5397039565062105,
+      "grad_norm": 0.06325556337833405,
+      "learning_rate": 4.059357911448227e-06,
+      "loss": 0.0426,
+      "step": 16000
+    },
+    {
+      "epoch": 2.5397039565062105,
+      "eval_accuracy": 0.9714285714285714,
+      "eval_loss": 0.15938520431518555,
+      "eval_macro_f1": 0.951264365385694,
+      "eval_macro_precision": 0.9601219800413174,
+      "eval_macro_recall": 0.9443849087252134,
+      "eval_micro_f1": 0.9714285714285714,
+      "eval_micro_precision": 0.9714285714285714,
+      "eval_micro_recall": 0.9714285714285714,
+      "eval_runtime": 64.8873,
+      "eval_samples_per_second": 194.183,
+      "eval_steps_per_second": 24.273,
+      "eval_weighted_f1": 0.971362978431738,
+      "eval_weighted_precision": 0.9714994225414888,
+      "eval_weighted_recall": 0.9714285714285714,
+      "step": 16000
+    },
+    {
+      "epoch": 2.5476407793960076,
+      "grad_norm": 1.398796558380127,
+      "learning_rate": 4.037308167225261e-06,
+      "loss": 0.0193,
+      "step": 16050
+    },
+    {
+      "epoch": 2.5555776022858048,
+      "grad_norm": 0.5250974297523499,
+      "learning_rate": 4.015258423002294e-06,
+      "loss": 0.0216,
+      "step": 16100
+    },
+    {
+      "epoch": 2.5635144251756024,
+      "grad_norm": 0.027746569365262985,
+      "learning_rate": 3.993208678779327e-06,
+      "loss": 0.0255,
+      "step": 16150
+    },
+    {
+      "epoch": 2.5714512480653995,
+      "grad_norm": 0.02244049869477749,
+      "learning_rate": 3.97115893455636e-06,
+      "loss": 0.0382,
+      "step": 16200
+    },
+    {
+      "epoch": 2.5793880709551966,
+      "grad_norm": 0.029930179938673973,
+      "learning_rate": 3.949109190333392e-06,
+      "loss": 0.0241,
+      "step": 16250
+    },
+    {
+      "epoch": 2.5873248938449938,
+      "grad_norm": 2.274050712585449,
+      "learning_rate": 3.927059446110426e-06,
+      "loss": 0.042,
+      "step": 16300
+    },
+    {
+      "epoch": 2.5952617167347913,
+      "grad_norm": 0.0035721093881875277,
+      "learning_rate": 3.905009701887459e-06,
+      "loss": 0.0286,
+      "step": 16350
+    },
+    {
+      "epoch": 2.6031985396245885,
+      "grad_norm": 0.026004351675510406,
+      "learning_rate": 3.882959957664492e-06,
+      "loss": 0.0275,
+      "step": 16400
+    },
+    {
+      "epoch": 2.6111353625143856,
+      "grad_norm": 0.09917880594730377,
+      "learning_rate": 3.860910213441525e-06,
+      "loss": 0.0351,
+      "step": 16450
+    },
+    {
+      "epoch": 2.6190721854041827,
+      "grad_norm": 0.037691615521907806,
+      "learning_rate": 3.838860469218557e-06,
+      "loss": 0.0609,
+      "step": 16500
+    },
+    {
+      "epoch": 2.6190721854041827,
+      "eval_accuracy": 0.9723015873015873,
+      "eval_loss": 0.1643410474061966,
+      "eval_macro_f1": 0.9561822135615655,
+      "eval_macro_precision": 0.9665323672953172,
+      "eval_macro_recall": 0.947527505766003,
+      "eval_micro_f1": 0.9723015873015873,
+      "eval_micro_precision": 0.9723015873015873,
+      "eval_micro_recall": 0.9723015873015873,
+      "eval_runtime": 52.441,
+      "eval_samples_per_second": 240.27,
+      "eval_steps_per_second": 30.034,
+      "eval_weighted_f1": 0.9722751933392187,
+      "eval_weighted_precision": 0.9724025203500458,
+      "eval_weighted_recall": 0.9723015873015873,
+      "step": 16500
+    },
+    {
+      "epoch": 2.62700900829398,
+      "grad_norm": 0.0057956683449447155,
+      "learning_rate": 3.816810724995591e-06,
+      "loss": 0.0246,
+      "step": 16550
+    },
+    {
+      "epoch": 2.634945831183777,
+      "grad_norm": 0.0268319733440876,
+      "learning_rate": 3.794760980772623e-06,
+      "loss": 0.0188,
+      "step": 16600
+    },
+    {
+      "epoch": 2.642882654073574,
+      "grad_norm": 0.009871570393443108,
+      "learning_rate": 3.7727112365496566e-06,
+      "loss": 0.0171,
+      "step": 16650
+    },
+    {
+      "epoch": 2.6508194769633717,
+      "grad_norm": 0.0035246764309704304,
+      "learning_rate": 3.750661492326689e-06,
+      "loss": 0.0381,
+      "step": 16700
+    },
+    {
+      "epoch": 2.658756299853169,
+      "grad_norm": 0.0025346768088638783,
+      "learning_rate": 3.728611748103722e-06,
+      "loss": 0.0487,
+      "step": 16750
+    },
+    {
+      "epoch": 2.666693122742966,
+      "grad_norm": 23.958332061767578,
+      "learning_rate": 3.7065620038807556e-06,
+      "loss": 0.0326,
+      "step": 16800
+    },
+    {
+      "epoch": 2.674629945632763,
+      "grad_norm": 0.010150304064154625,
+      "learning_rate": 3.684512259657788e-06,
+      "loss": 0.0347,
+      "step": 16850
+    },
+    {
+      "epoch": 2.6825667685225603,
+      "grad_norm": 0.0012810941552743316,
+      "learning_rate": 3.6624625154348216e-06,
+      "loss": 0.0403,
+      "step": 16900
+    },
+    {
+      "epoch": 2.690503591412358,
+      "grad_norm": 24.288969039916992,
+      "learning_rate": 3.640412771211854e-06,
+      "loss": 0.013,
+      "step": 16950
+    },
+    {
+      "epoch": 2.698440414302155,
+      "grad_norm": 0.000993003137409687,
+      "learning_rate": 3.618363026988887e-06,
+      "loss": 0.0409,
+      "step": 17000
+    },
+    {
+      "epoch": 2.698440414302155,
+      "eval_accuracy": 0.973015873015873,
+      "eval_loss": 0.16805726289749146,
+      "eval_macro_f1": 0.9491028977586443,
+      "eval_macro_precision": 0.960433813735491,
+      "eval_macro_recall": 0.9406320914982489,
+      "eval_micro_f1": 0.973015873015873,
+      "eval_micro_precision": 0.973015873015873,
+      "eval_micro_recall": 0.973015873015873,
+      "eval_runtime": 50.2558,
+      "eval_samples_per_second": 250.717,
+      "eval_steps_per_second": 31.34,
+      "eval_weighted_f1": 0.9729576202329057,
+      "eval_weighted_precision": 0.9730875205196509,
+      "eval_weighted_recall": 0.973015873015873,
+      "step": 17000
+    },
+    {
+      "epoch": 2.706377237191952,
+      "grad_norm": 3.7704389095306396,
+      "learning_rate": 3.59631328276592e-06,
+      "loss": 0.0195,
+      "step": 17050
+    },
+    {
+      "epoch": 2.7143140600817492,
+      "grad_norm": 0.0006660755025222898,
+      "learning_rate": 3.574263538542953e-06,
+      "loss": 0.0353,
+      "step": 17100
+    },
+    {
+      "epoch": 2.7222508829715464,
+      "grad_norm": 0.3745381236076355,
+      "learning_rate": 3.5522137943199865e-06,
+      "loss": 0.0214,
+      "step": 17150
+    },
+    {
+      "epoch": 2.7301877058613435,
+      "grad_norm": 0.005227445159107447,
+      "learning_rate": 3.530164050097019e-06,
+      "loss": 0.0203,
+      "step": 17200
+    },
+    {
+      "epoch": 2.7381245287511407,
+      "grad_norm": 0.05709793046116829,
+      "learning_rate": 3.5085553007585115e-06,
+      "loss": 0.0289,
+      "step": 17250
+    },
+    {
+      "epoch": 2.7460613516409382,
+      "grad_norm": 0.3132238984107971,
+      "learning_rate": 3.4865055565355444e-06,
+      "loss": 0.0344,
+      "step": 17300
+    },
+    {
+      "epoch": 2.7539981745307354,
+      "grad_norm": 0.00023108968161977828,
+      "learning_rate": 3.464455812312578e-06,
+      "loss": 0.0299,
+      "step": 17350
+    },
+    {
+      "epoch": 2.7619349974205325,
+      "grad_norm": 0.0010186864528805017,
+      "learning_rate": 3.4424060680896104e-06,
+      "loss": 0.0156,
+      "step": 17400
+    },
+    {
+      "epoch": 2.7698718203103296,
+      "grad_norm": 0.4859501123428345,
+      "learning_rate": 3.420797318751103e-06,
+      "loss": 0.0178,
+      "step": 17450
+    },
+    {
+      "epoch": 2.777808643200127,
+      "grad_norm": 0.0007182751432992518,
+      "learning_rate": 3.398747574528136e-06,
+      "loss": 0.0526,
+      "step": 17500
+    },
+    {
+      "epoch": 2.777808643200127,
+      "eval_accuracy": 0.9734126984126984,
+      "eval_loss": 0.16671566665172577,
+      "eval_macro_f1": 0.9471381395941902,
+      "eval_macro_precision": 0.9549885122141759,
+      "eval_macro_recall": 0.9428154605456508,
+      "eval_micro_f1": 0.9734126984126984,
+      "eval_micro_precision": 0.9734126984126984,
+      "eval_micro_recall": 0.9734126984126984,
+      "eval_runtime": 64.8418,
+      "eval_samples_per_second": 194.319,
+      "eval_steps_per_second": 24.29,
+      "eval_weighted_f1": 0.9733825407985148,
+      "eval_weighted_precision": 0.973532406246267,
+      "eval_weighted_recall": 0.9734126984126984,
+      "step": 17500
+    },
+    {
+      "epoch": 2.7857454660899243,
+      "grad_norm": 0.03506314009428024,
+      "learning_rate": 3.3766978303051688e-06,
+      "loss": 0.0361,
+      "step": 17550
+    },
+    {
+      "epoch": 2.7936822889797215,
+      "grad_norm": 0.008137653581798077,
+      "learning_rate": 3.3546480860822018e-06,
+      "loss": 0.0407,
+      "step": 17600
+    },
+    {
+      "epoch": 2.8016191118695186,
+      "grad_norm": 3.514742136001587,
+      "learning_rate": 3.3325983418592343e-06,
+      "loss": 0.0298,
+      "step": 17650
+    },
+    {
+      "epoch": 2.8095559347593158,
+      "grad_norm": 0.0013472414575517178,
+      "learning_rate": 3.3105485976362678e-06,
+      "loss": 0.027,
+      "step": 17700
+    },
+    {
+      "epoch": 2.817492757649113,
+      "grad_norm": 0.00018067212658934295,
+      "learning_rate": 3.2884988534133007e-06,
+      "loss": 0.0243,
+      "step": 17750
+    },
+    {
+      "epoch": 2.82542958053891,
+      "grad_norm": 0.0013635717332363129,
+      "learning_rate": 3.2664491091903337e-06,
+      "loss": 0.0169,
+      "step": 17800
+    },
+    {
+      "epoch": 2.8333664034287076,
+      "grad_norm": 0.0013044564984738827,
+      "learning_rate": 3.2443993649673667e-06,
+      "loss": 0.0154,
+      "step": 17850
+    },
+    {
+      "epoch": 2.8413032263185047,
+      "grad_norm": 0.00033568262006156147,
+      "learning_rate": 3.2223496207443993e-06,
+      "loss": 0.0197,
+      "step": 17900
+    },
+    {
+      "epoch": 2.849240049208302,
+      "grad_norm": 0.00014405626279767603,
+      "learning_rate": 3.2002998765214327e-06,
+      "loss": 0.0181,
+      "step": 17950
+    },
+    {
+      "epoch": 2.857176872098099,
+      "grad_norm": 1.2447603940963745,
+      "learning_rate": 3.1782501322984653e-06,
+      "loss": 0.0657,
+      "step": 18000
+    },
+    {
+      "epoch": 2.857176872098099,
+      "eval_accuracy": 0.9738095238095238,
+      "eval_loss": 0.17004433274269104,
+      "eval_macro_f1": 0.9529234191380291,
+      "eval_macro_precision": 0.962677256883139,
+      "eval_macro_recall": 0.9454556144210745,
+      "eval_micro_f1": 0.9738095238095238,
+      "eval_micro_precision": 0.9738095238095238,
+      "eval_micro_recall": 0.9738095238095238,
+      "eval_runtime": 48.8808,
+      "eval_samples_per_second": 257.77,
+      "eval_steps_per_second": 32.221,
+      "eval_weighted_f1": 0.9737837762311287,
+      "eval_weighted_precision": 0.9739539079223564,
+      "eval_weighted_recall": 0.9738095238095238,
+      "step": 18000
+    },
+    {
+      "epoch": 2.8651136949878966,
+      "grad_norm": 0.04827316105365753,
+      "learning_rate": 3.1562003880754987e-06,
+      "loss": 0.0575,
+      "step": 18050
+    },
+    {
+      "epoch": 2.8730505178776937,
+      "grad_norm": 3.009031295776367,
+      "learning_rate": 3.1341506438525317e-06,
+      "loss": 0.009,
+      "step": 18100
+    },
+    {
+      "epoch": 2.880987340767491,
+      "grad_norm": 12.70515251159668,
+      "learning_rate": 3.1121008996295642e-06,
+      "loss": 0.0717,
+      "step": 18150
+    },
+    {
+      "epoch": 2.888924163657288,
+      "grad_norm": 0.008692610077559948,
+      "learning_rate": 3.0900511554065976e-06,
+      "loss": 0.0653,
+      "step": 18200
+    },
+    {
+      "epoch": 2.896860986547085,
+      "grad_norm": 0.0048131803050637245,
+      "learning_rate": 3.06800141118363e-06,
+      "loss": 0.0246,
+      "step": 18250
+    },
+    {
+      "epoch": 2.9047978094368823,
+      "grad_norm": 0.00026718704611994326,
+      "learning_rate": 3.0459516669606636e-06,
+      "loss": 0.0097,
+      "step": 18300
+    },
+    {
+      "epoch": 2.9127346323266794,
+      "grad_norm": 0.08372581005096436,
+      "learning_rate": 3.0239019227376966e-06,
+      "loss": 0.0618,
+      "step": 18350
+    },
+    {
+      "epoch": 2.920671455216477,
+      "grad_norm": 0.00336449034512043,
+      "learning_rate": 3.001852178514729e-06,
+      "loss": 0.0344,
+      "step": 18400
+    },
+    {
+      "epoch": 2.928608278106274,
+      "grad_norm": 0.0021333922632038593,
+      "learning_rate": 2.9798024342917626e-06,
+      "loss": 0.0296,
+      "step": 18450
+    },
+    {
+      "epoch": 2.9365451009960712,
+      "grad_norm": 0.002129336353391409,
+      "learning_rate": 2.957752690068795e-06,
+      "loss": 0.0056,
+      "step": 18500
+    },
+    {
+      "epoch": 2.9365451009960712,
+      "eval_accuracy": 0.975,
+      "eval_loss": 0.16028916835784912,
+      "eval_macro_f1": 0.9540199515604112,
+      "eval_macro_precision": 0.9647089557155375,
+      "eval_macro_recall": 0.945480105008737,
+      "eval_micro_f1": 0.975,
+      "eval_micro_precision": 0.975,
+      "eval_micro_recall": 0.975,
+      "eval_runtime": 60.6955,
+      "eval_samples_per_second": 207.594,
+      "eval_steps_per_second": 25.949,
+      "eval_weighted_f1": 0.9749518139909482,
+      "eval_weighted_precision": 0.9750584916764344,
+      "eval_weighted_recall": 0.975,
+      "step": 18500
+    },
+    {
+      "epoch": 2.9444819238858684,
+      "grad_norm": 0.045723918825387955,
+      "learning_rate": 2.9357029458458286e-06,
+      "loss": 0.0243,
+      "step": 18550
+    },
+    {
+      "epoch": 2.952418746775666,
+      "grad_norm": 0.007504200097173452,
+      "learning_rate": 2.9136532016228616e-06,
+      "loss": 0.0214,
+      "step": 18600
+    },
+    {
+      "epoch": 2.960355569665463,
+      "grad_norm": 0.0063537731766700745,
+      "learning_rate": 2.891603457399894e-06,
+      "loss": 0.0375,
+      "step": 18650
+    },
+    {
+      "epoch": 2.9682923925552602,
+      "grad_norm": 37.7386589050293,
+      "learning_rate": 2.8695537131769275e-06,
+      "loss": 0.0326,
+      "step": 18700
+    },
+    {
+      "epoch": 2.9762292154450574,
+      "grad_norm": 0.7210281491279602,
+      "learning_rate": 2.84750396895396e-06,
+      "loss": 0.0192,
+      "step": 18750
+    },
+    {
+      "epoch": 2.9841660383348545,
+      "grad_norm": 0.1253499835729599,
+      "learning_rate": 2.8254542247309935e-06,
+      "loss": 0.0159,
+      "step": 18800
+    },
+    {
+      "epoch": 2.9921028612246516,
+      "grad_norm": 0.18740814924240112,
+      "learning_rate": 2.803404480508026e-06,
+      "loss": 0.0409,
+      "step": 18850
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.6921190023422241,
+      "learning_rate": 2.7813547362850595e-06,
+      "loss": 0.0461,
+      "step": 18900
+    },
+    {
+      "epoch": 3.007936822889797,
+      "grad_norm": 0.004000192042440176,
+      "learning_rate": 2.7593049920620925e-06,
+      "loss": 0.0103,
+      "step": 18950
+    },
+    {
+      "epoch": 3.0158736457795943,
+      "grad_norm": 0.00037790765054523945,
+      "learning_rate": 2.737255247839125e-06,
+      "loss": 0.0117,
+      "step": 19000
+    },
+    {
+      "epoch": 3.0158736457795943,
+      "eval_accuracy": 0.9762698412698413,
+      "eval_loss": 0.14878861606121063,
+      "eval_macro_f1": 0.954347658252044,
+      "eval_macro_precision": 0.9607715131633574,
+      "eval_macro_recall": 0.9501403976665715,
+      "eval_micro_f1": 0.9762698412698413,
+      "eval_micro_precision": 0.9762698412698413,
+      "eval_micro_recall": 0.9762698412698413,
+      "eval_runtime": 51.7414,
+      "eval_samples_per_second": 243.519,
+      "eval_steps_per_second": 30.44,
+      "eval_weighted_f1": 0.9762667198823463,
+      "eval_weighted_precision": 0.9764195140172701,
+      "eval_weighted_recall": 0.9762698412698413,
+      "step": 19000
+    },
+    {
+      "epoch": 3.023810468669392,
+      "grad_norm": 0.00037838733987882733,
+      "learning_rate": 2.7152055036161585e-06,
+      "loss": 0.0051,
+      "step": 19050
+    },
+    {
+      "epoch": 3.031747291559189,
+      "grad_norm": 0.00012900579895358533,
+      "learning_rate": 2.693155759393191e-06,
+      "loss": 0.0028,
+      "step": 19100
+    },
+    {
+      "epoch": 3.039684114448986,
+      "grad_norm": 0.0025370146613568068,
+      "learning_rate": 2.6711060151702244e-06,
+      "loss": 0.0027,
+      "step": 19150
+    },
+    {
+      "epoch": 3.0476209373387833,
+      "grad_norm": 17.906938552856445,
+      "learning_rate": 2.6490562709472574e-06,
+      "loss": 0.0147,
+      "step": 19200
+    },
+    {
+      "epoch": 3.0555577602285804,
+      "grad_norm": 0.26221588253974915,
+      "learning_rate": 2.62700652672429e-06,
+      "loss": 0.0066,
+      "step": 19250
+    },
+    {
+      "epoch": 3.0634945831183775,
+      "grad_norm": 0.0025780112482607365,
+      "learning_rate": 2.6049567825013234e-06,
+      "loss": 0.0063,
+      "step": 19300
+    },
+    {
+      "epoch": 3.071431406008175,
+      "grad_norm": 0.16827242076396942,
+      "learning_rate": 2.582907038278356e-06,
+      "loss": 0.0019,
+      "step": 19350
+    },
+    {
+      "epoch": 3.0793682288979722,
+      "grad_norm": 0.018481239676475525,
+      "learning_rate": 2.5608572940553894e-06,
+      "loss": 0.0049,
+      "step": 19400
+    },
+    {
+      "epoch": 3.0873050517877694,
+      "grad_norm": 0.004198388196527958,
+      "learning_rate": 2.5388075498324224e-06,
+      "loss": 0.0073,
+      "step": 19450
+    },
+    {
+      "epoch": 3.0952418746775665,
+      "grad_norm": 0.00025698720128275454,
+      "learning_rate": 2.516757805609455e-06,
+      "loss": 0.0059,
+      "step": 19500
+    },
+    {
+      "epoch": 3.0952418746775665,
+      "eval_accuracy": 0.9756349206349206,
+      "eval_loss": 0.16310656070709229,
+      "eval_macro_f1": 0.957125861216084,
+      "eval_macro_precision": 0.9727967123798239,
+      "eval_macro_recall": 0.9442922894357393,
+      "eval_micro_f1": 0.9756349206349206,
+      "eval_micro_precision": 0.9756349206349206,
+      "eval_micro_recall": 0.9756349206349206,
+      "eval_runtime": 47.4743,
+      "eval_samples_per_second": 265.407,
+      "eval_steps_per_second": 33.176,
+      "eval_weighted_f1": 0.9755432972728447,
+      "eval_weighted_precision": 0.9756807849259415,
+      "eval_weighted_recall": 0.9756349206349206,
+      "step": 19500
+    },
+    {
+      "epoch": 3.1031786975673636,
+      "grad_norm": 0.009408715181052685,
+      "learning_rate": 2.4947080613864884e-06,
+      "loss": 0.0273,
+      "step": 19550
+    },
+    {
+      "epoch": 3.1111155204571608,
+      "grad_norm": 0.01176052913069725,
+      "learning_rate": 2.472658317163521e-06,
+      "loss": 0.01,
+      "step": 19600
+    },
+    {
+      "epoch": 3.1190523433469584,
+      "grad_norm": 0.0038633132353425026,
+      "learning_rate": 2.450608572940554e-06,
+      "loss": 0.0061,
+      "step": 19650
+    },
+    {
+      "epoch": 3.1269891662367555,
+      "grad_norm": 0.04556519538164139,
+      "learning_rate": 2.428558828717587e-06,
+      "loss": 0.0003,
+      "step": 19700
+    },
+    {
+      "epoch": 3.1349259891265526,
+      "grad_norm": 0.036106646060943604,
+      "learning_rate": 2.4065090844946203e-06,
+      "loss": 0.0156,
+      "step": 19750
+    },
+    {
+      "epoch": 3.1428628120163498,
+      "grad_norm": 0.2732146680355072,
+      "learning_rate": 2.3844593402716533e-06,
+      "loss": 0.0124,
+      "step": 19800
+    },
+    {
+      "epoch": 3.150799634906147,
+      "grad_norm": 0.002729196799919009,
+      "learning_rate": 2.362409596048686e-06,
+      "loss": 0.0159,
+      "step": 19850
+    },
+    {
+      "epoch": 3.1587364577959445,
+      "grad_norm": 0.008584374561905861,
+      "learning_rate": 2.340359851825719e-06,
+      "loss": 0.0096,
+      "step": 19900
+    },
+    {
+      "epoch": 3.1666732806857416,
+      "grad_norm": 0.12530040740966797,
+      "learning_rate": 2.318310107602752e-06,
+      "loss": 0.0003,
+      "step": 19950
+    },
+    {
+      "epoch": 3.1746101035755387,
+      "grad_norm": 0.0008394561009481549,
+      "learning_rate": 2.296260363379785e-06,
+      "loss": 0.0027,
+      "step": 20000
+    },
+    {
+      "epoch": 3.1746101035755387,
+      "eval_accuracy": 0.9767460317460317,
+      "eval_loss": 0.16326671838760376,
+      "eval_macro_f1": 0.9567534320462188,
+      "eval_macro_precision": 0.9698508452291146,
+      "eval_macro_recall": 0.9464068389671931,
+      "eval_micro_f1": 0.9767460317460317,
+      "eval_micro_precision": 0.9767460317460317,
+      "eval_micro_recall": 0.9767460317460317,
+      "eval_runtime": 65.9869,
+      "eval_samples_per_second": 190.947,
+      "eval_steps_per_second": 23.868,
+      "eval_weighted_f1": 0.9766894759901187,
+      "eval_weighted_precision": 0.9767954117631042,
+      "eval_weighted_recall": 0.9767460317460317,
+      "step": 20000
+    },
+    {
+      "epoch": 3.182546926465336,
+      "grad_norm": 0.00108958943746984,
+      "learning_rate": 2.2742106191568183e-06,
+      "loss": 0.0141,
+      "step": 20050
+    },
+    {
+      "epoch": 3.190483749355133,
+      "grad_norm": 1.8105076551437378,
+      "learning_rate": 2.252160874933851e-06,
+      "loss": 0.0161,
+      "step": 20100
+    },
+    {
+      "epoch": 3.19842057224493,
+      "grad_norm": 0.03791620582342148,
+      "learning_rate": 2.230111130710884e-06,
+      "loss": 0.03,
+      "step": 20150
+    },
+    {
+      "epoch": 3.2063573951347277,
+      "grad_norm": 0.05852150917053223,
+      "learning_rate": 2.208061386487917e-06,
+      "loss": 0.0118,
+      "step": 20200
+    },
+    {
+      "epoch": 3.214294218024525,
+      "grad_norm": 32.71089553833008,
+      "learning_rate": 2.1860116422649498e-06,
+      "loss": 0.0126,
+      "step": 20250
+    },
+    {
+      "epoch": 3.222231040914322,
+      "grad_norm": 0.0013255997328087687,
+      "learning_rate": 2.1639618980419828e-06,
+      "loss": 0.0003,
+      "step": 20300
+    },
+    {
+      "epoch": 3.230167863804119,
+      "grad_norm": 0.002787817968055606,
+      "learning_rate": 2.1419121538190158e-06,
+      "loss": 0.0034,
+      "step": 20350
+    },
+    {
+      "epoch": 3.2381046866939163,
+      "grad_norm": 0.002626606961712241,
+      "learning_rate": 2.1198624095960488e-06,
+      "loss": 0.0014,
+      "step": 20400
+    },
+    {
+      "epoch": 3.246041509583714,
+      "grad_norm": 0.0015965308994054794,
+      "learning_rate": 2.0978126653730817e-06,
+      "loss": 0.0074,
+      "step": 20450
+    },
+    {
+      "epoch": 3.253978332473511,
+      "grad_norm": 0.00022477486345451325,
+      "learning_rate": 2.0757629211501147e-06,
+      "loss": 0.0016,
+      "step": 20500
+    },
+    {
+      "epoch": 3.253978332473511,
+      "eval_accuracy": 0.9761111111111112,
+      "eval_loss": 0.1650867611169815,
+      "eval_macro_f1": 0.955655463952606,
+      "eval_macro_precision": 0.9677917142854108,
+      "eval_macro_recall": 0.9461865102151874,
+      "eval_micro_f1": 0.9761111111111112,
+      "eval_micro_precision": 0.9761111111111112,
+      "eval_micro_recall": 0.9761111111111112,
+      "eval_runtime": 63.0368,
+      "eval_samples_per_second": 199.883,
+      "eval_steps_per_second": 24.985,
+      "eval_weighted_f1": 0.9760636712967664,
+      "eval_weighted_precision": 0.9761579320691899,
+      "eval_weighted_recall": 0.9761111111111112,
+      "step": 20500
+    },
+    {
+      "epoch": 3.261915155363308,
+      "grad_norm": 0.015780743211507797,
+      "learning_rate": 2.0537131769271477e-06,
+      "loss": 0.002,
+      "step": 20550
+    },
+    {
+      "epoch": 3.2698519782531053,
+      "grad_norm": 0.0010949111310765147,
+      "learning_rate": 2.031663432704181e-06,
+      "loss": 0.0001,
+      "step": 20600
+    },
+    {
+      "epoch": 3.2777888011429024,
+      "grad_norm": 0.018619557842612267,
+      "learning_rate": 2.0096136884812137e-06,
+      "loss": 0.0091,
+      "step": 20650
+    },
+    {
+      "epoch": 3.2857256240326995,
+      "grad_norm": 0.00012928845535498112,
+      "learning_rate": 1.9875639442582467e-06,
+      "loss": 0.0048,
+      "step": 20700
+    },
+    {
+      "epoch": 3.293662446922497,
+      "grad_norm": 0.0020417433697730303,
+      "learning_rate": 1.9655142000352797e-06,
+      "loss": 0.0032,
+      "step": 20750
+    },
+    {
+      "epoch": 3.3015992698122942,
+      "grad_norm": 0.003012231085449457,
+      "learning_rate": 1.9434644558123127e-06,
+      "loss": 0.0089,
+      "step": 20800
+    },
+    {
+      "epoch": 3.3095360927020914,
+      "grad_norm": 0.0012670076685026288,
+      "learning_rate": 1.9214147115893457e-06,
+      "loss": 0.0027,
+      "step": 20850
+    },
+    {
+      "epoch": 3.3174729155918885,
+      "grad_norm": 0.0011639808071777225,
+      "learning_rate": 1.8993649673663786e-06,
+      "loss": 0.0015,
+      "step": 20900
+    },
+    {
+      "epoch": 3.3254097384816856,
+      "grad_norm": 0.0014048466691747308,
+      "learning_rate": 1.8773152231434116e-06,
+      "loss": 0.0096,
+      "step": 20950
+    },
+    {
+      "epoch": 3.333346561371483,
+      "grad_norm": 9.100540161132812,
+      "learning_rate": 1.8552654789204446e-06,
+      "loss": 0.0197,
+      "step": 21000
+    },
+    {
+      "epoch": 3.333346561371483,
+      "eval_accuracy": 0.9763492063492063,
+      "eval_loss": 0.17093592882156372,
+      "eval_macro_f1": 0.956419648478896,
+      "eval_macro_precision": 0.9707661139733746,
+      "eval_macro_recall": 0.9448057653586319,
+      "eval_micro_f1": 0.9763492063492063,
+      "eval_micro_precision": 0.9763492063492063,
+      "eval_micro_recall": 0.9763492063492063,
+      "eval_runtime": 52.3722,
+      "eval_samples_per_second": 240.586,
+      "eval_steps_per_second": 30.073,
+      "eval_weighted_f1": 0.9762812381183971,
+      "eval_weighted_precision": 0.976400813243511,
+      "eval_weighted_recall": 0.9763492063492063,
+      "step": 21000
+    },
+    {
+      "epoch": 3.3412833842612804,
+      "grad_norm": 0.00045143740135245025,
+      "learning_rate": 1.8332157346974776e-06,
+      "loss": 0.0037,
+      "step": 21050
+    },
+    {
+      "epoch": 3.3492202071510775,
+      "grad_norm": 0.00014390982687473297,
+      "learning_rate": 1.8111659904745108e-06,
+      "loss": 0.0096,
+      "step": 21100
+    },
+    {
+      "epoch": 3.3571570300408746,
+      "grad_norm": 4.516872650128789e-05,
+      "learning_rate": 1.7891162462515436e-06,
+      "loss": 0.0104,
+      "step": 21150
+    },
+    {
+      "epoch": 3.3650938529306718,
+      "grad_norm": 0.020564353093504906,
+      "learning_rate": 1.7670665020285766e-06,
+      "loss": 0.001,
+      "step": 21200
+    },
+    {
+      "epoch": 3.373030675820469,
+      "grad_norm": 0.00030001444974914193,
+      "learning_rate": 1.7450167578056096e-06,
+      "loss": 0.0168,
+      "step": 21250
+    },
+    {
+      "epoch": 3.3809674987102665,
+      "grad_norm": 0.19772754609584808,
+      "learning_rate": 1.7229670135826426e-06,
+      "loss": 0.0002,
+      "step": 21300
+    },
+    {
+      "epoch": 3.3889043216000636,
+      "grad_norm": 0.01970568671822548,
+      "learning_rate": 1.7009172693596758e-06,
+      "loss": 0.0309,
+      "step": 21350
+    },
+    {
+      "epoch": 3.3968411444898607,
+      "grad_norm": 0.00025925575755536556,
+      "learning_rate": 1.6788675251367083e-06,
+      "loss": 0.0073,
+      "step": 21400
+    },
+    {
+      "epoch": 3.404777967379658,
+      "grad_norm": 0.004676587413996458,
+      "learning_rate": 1.657258775798201e-06,
+      "loss": 0.0052,
+      "step": 21450
+    },
+    {
+      "epoch": 3.412714790269455,
+      "grad_norm": 0.00010141974780708551,
+      "learning_rate": 1.635209031575234e-06,
+      "loss": 0.0079,
+      "step": 21500
+    },
+    {
+      "epoch": 3.412714790269455,
+      "eval_accuracy": 0.9769047619047619,
+      "eval_loss": 0.16325406730175018,
+      "eval_macro_f1": 0.9557812168870482,
+      "eval_macro_precision": 0.9696293003435598,
+      "eval_macro_recall": 0.9445357837826058,
+      "eval_micro_f1": 0.9769047619047619,
+      "eval_micro_precision": 0.9769047619047619,
+      "eval_micro_recall": 0.9769047619047619,
+      "eval_runtime": 48.9392,
+      "eval_samples_per_second": 257.462,
+      "eval_steps_per_second": 32.183,
+      "eval_weighted_f1": 0.9768367289110993,
+      "eval_weighted_precision": 0.9769254523389815,
+      "eval_weighted_recall": 0.9769047619047619,
+      "step": 21500
+    },
+    {
+      "epoch": 3.4206516131592526,
+      "grad_norm": 6.527017103508115e-05,
+      "learning_rate": 1.613159287352267e-06,
+      "loss": 0.0001,
+      "step": 21550
+    },
+    {
+      "epoch": 3.4285884360490497,
+      "grad_norm": 0.00797638576477766,
+      "learning_rate": 1.5911095431292997e-06,
+      "loss": 0.0023,
+      "step": 21600
+    },
+    {
+      "epoch": 3.436525258938847,
+      "grad_norm": 0.0008530872291885316,
+      "learning_rate": 1.5690597989063327e-06,
+      "loss": 0.0007,
+      "step": 21650
+    },
+    {
+      "epoch": 3.444462081828644,
+      "grad_norm": 0.00018081249436363578,
+      "learning_rate": 1.5470100546833659e-06,
+      "loss": 0.0109,
+      "step": 21700
+    },
+    {
+      "epoch": 3.452398904718441,
+      "grad_norm": 5.029146996093914e-05,
+      "learning_rate": 1.5249603104603989e-06,
+      "loss": 0.0041,
+      "step": 21750
+    },
+    {
+      "epoch": 3.4603357276082383,
+      "grad_norm": 0.002522735157981515,
+      "learning_rate": 1.5029105662374318e-06,
+      "loss": 0.0006,
+      "step": 21800
+    },
+    {
+      "epoch": 3.4682725504980354,
+      "grad_norm": 2.8557164114317857e-05,
+      "learning_rate": 1.4808608220144646e-06,
+      "loss": 0.0031,
+      "step": 21850
+    },
+    {
+      "epoch": 3.476209373387833,
+      "grad_norm": 0.003235569689422846,
+      "learning_rate": 1.4588110777914976e-06,
+      "loss": 0.0185,
+      "step": 21900
+    },
+    {
+      "epoch": 3.48414619627763,
+      "grad_norm": 0.011313154362142086,
+      "learning_rate": 1.4367613335685306e-06,
+      "loss": 0.0145,
+      "step": 21950
+    },
+    {
+      "epoch": 3.4920830191674272,
+      "grad_norm": 3.0896131647750735e-05,
+      "learning_rate": 1.4147115893455638e-06,
+      "loss": 0.0231,
+      "step": 22000
+    },
+    {
+      "epoch": 3.4920830191674272,
+      "eval_accuracy": 0.9756349206349206,
+      "eval_loss": 0.16351020336151123,
+      "eval_macro_f1": 0.9551717284616833,
+      "eval_macro_precision": 0.9681364681283637,
+      "eval_macro_recall": 0.945098492018414,
+      "eval_micro_f1": 0.9756349206349206,
+      "eval_micro_precision": 0.9756349206349206,
+      "eval_micro_recall": 0.9756349206349206,
+      "eval_runtime": 48.5537,
+      "eval_samples_per_second": 259.506,
+      "eval_steps_per_second": 32.438,
+      "eval_weighted_f1": 0.9755767607894801,
+      "eval_weighted_precision": 0.9756980555491296,
+      "eval_weighted_recall": 0.9756349206349206,
+      "step": 22000
+    },
+    {
+      "epoch": 3.5000198420572244,
+      "grad_norm": 0.0014016503700986505,
+      "learning_rate": 1.3926618451225968e-06,
+      "loss": 0.0175,
+      "step": 22050
+    },
+    {
+      "epoch": 3.507956664947022,
+      "grad_norm": 0.005145555827766657,
+      "learning_rate": 1.3706121008996296e-06,
+      "loss": 0.0075,
+      "step": 22100
+    },
+    {
+      "epoch": 3.515893487836819,
+      "grad_norm": 1.581344986334443e-05,
+      "learning_rate": 1.3485623566766626e-06,
+      "loss": 0.0004,
+      "step": 22150
+    },
+    {
+      "epoch": 3.5238303107266162,
+      "grad_norm": 7.287700282176957e-05,
+      "learning_rate": 1.3265126124536956e-06,
+      "loss": 0.0008,
+      "step": 22200
+    },
+    {
+      "epoch": 3.5317671336164134,
+      "grad_norm": 0.002235785825178027,
+      "learning_rate": 1.3044628682307288e-06,
+      "loss": 0.0077,
+      "step": 22250
+    },
+    {
+      "epoch": 3.5397039565062105,
+      "grad_norm": 0.00019435022841207683,
+      "learning_rate": 1.2824131240077617e-06,
+      "loss": 0.0026,
+      "step": 22300
+    },
+    {
+      "epoch": 3.5476407793960076,
+      "grad_norm": 0.008783644996583462,
+      "learning_rate": 1.2603633797847947e-06,
+      "loss": 0.0031,
+      "step": 22350
+    },
+    {
+      "epoch": 3.5555776022858048,
+      "grad_norm": 8.886044815881178e-05,
+      "learning_rate": 1.2383136355618277e-06,
+      "loss": 0.0021,
+      "step": 22400
+    },
+    {
+      "epoch": 3.5635144251756024,
+      "grad_norm": 0.0015128047671169043,
+      "learning_rate": 1.2162638913388605e-06,
+      "loss": 0.0126,
+      "step": 22450
+    },
+    {
+      "epoch": 3.5714512480653995,
+      "grad_norm": 0.0013680335832759738,
+      "learning_rate": 1.1942141471158935e-06,
+      "loss": 0.0076,
+      "step": 22500
+    },
+    {
+      "epoch": 3.5714512480653995,
+      "eval_accuracy": 0.9765873015873016,
+      "eval_loss": 0.16308797895908356,
+      "eval_macro_f1": 0.9559072153039692,
+      "eval_macro_precision": 0.9680547193198462,
+      "eval_macro_recall": 0.9463389894990366,
+      "eval_micro_f1": 0.9765873015873016,
+      "eval_micro_precision": 0.9765873015873016,
+      "eval_micro_recall": 0.9765873015873016,
+      "eval_runtime": 49.1975,
+      "eval_samples_per_second": 256.11,
+      "eval_steps_per_second": 32.014,
+      "eval_weighted_f1": 0.9765228546965284,
+      "eval_weighted_precision": 0.9766302422786851,
+      "eval_weighted_recall": 0.9765873015873016,
+      "step": 22500
+    },
+    {
+      "epoch": 3.5793880709551966,
+      "grad_norm": 0.03631841763854027,
+      "learning_rate": 1.1721644028929267e-06,
+      "loss": 0.0081,
+      "step": 22550
+    },
+    {
+      "epoch": 3.5873248938449938,
+      "grad_norm": 0.0012924810871481895,
+      "learning_rate": 1.1501146586699595e-06,
+      "loss": 0.0049,
+      "step": 22600
+    },
+    {
+      "epoch": 3.5952617167347913,
+      "grad_norm": 0.006269298959523439,
+      "learning_rate": 1.1280649144469925e-06,
+      "loss": 0.0075,
+      "step": 22650
+    },
+    {
+      "epoch": 3.6031985396245885,
+      "grad_norm": 0.0009378429385833442,
+      "learning_rate": 1.1060151702240254e-06,
+      "loss": 0.0086,
+      "step": 22700
+    },
+    {
+      "epoch": 3.6111353625143856,
+      "grad_norm": 0.032317861914634705,
+      "learning_rate": 1.0839654260010584e-06,
+      "loss": 0.0023,
+      "step": 22750
+    },
+    {
+      "epoch": 3.6190721854041827,
+      "grad_norm": 0.001633369131013751,
+      "learning_rate": 1.0619156817780914e-06,
+      "loss": 0.0002,
+      "step": 22800
+    },
+    {
+      "epoch": 3.62700900829398,
+      "grad_norm": 6.64698745822534e-05,
+      "learning_rate": 1.0398659375551244e-06,
+      "loss": 0.0111,
+      "step": 22850
+    },
+    {
+      "epoch": 3.634945831183777,
+      "grad_norm": 14.95582389831543,
+      "learning_rate": 1.0178161933321574e-06,
+      "loss": 0.0045,
+      "step": 22900
+    },
+    {
+      "epoch": 3.642882654073574,
+      "grad_norm": 0.0011493790661916137,
+      "learning_rate": 9.957664491091904e-07,
+      "loss": 0.0025,
+      "step": 22950
+    },
+    {
+      "epoch": 3.6508194769633717,
+      "grad_norm": 0.04192957282066345,
+      "learning_rate": 9.737167048862234e-07,
+      "loss": 0.0032,
+      "step": 23000
+    },
+    {
+      "epoch": 3.6508194769633717,
+      "eval_accuracy": 0.9774603174603175,
+      "eval_loss": 0.16377675533294678,
+      "eval_macro_f1": 0.9572334679389451,
+      "eval_macro_precision": 0.9700362196030153,
+      "eval_macro_recall": 0.9471315195900166,
+      "eval_micro_f1": 0.9774603174603175,
+      "eval_micro_precision": 0.9774603174603175,
+      "eval_micro_recall": 0.9774603174603175,
+      "eval_runtime": 64.2461,
+      "eval_samples_per_second": 196.121,
+      "eval_steps_per_second": 24.515,
+      "eval_weighted_f1": 0.9773904798618511,
+      "eval_weighted_precision": 0.9775254966358302,
+      "eval_weighted_recall": 0.9774603174603175,
+      "step": 23000
+    },
+    {
+      "epoch": 3.658756299853169,
+      "grad_norm": 0.0003047911450266838,
+      "learning_rate": 9.516669606632564e-07,
+      "loss": 0.0003,
+      "step": 23050
+    },
+    {
+      "epoch": 3.666693122742966,
+      "grad_norm": 0.14043422043323517,
+      "learning_rate": 9.296172164402894e-07,
+      "loss": 0.0115,
+      "step": 23100
+    },
+    {
+      "epoch": 3.674629945632763,
+      "grad_norm": 0.007793279364705086,
+      "learning_rate": 9.075674722173224e-07,
+      "loss": 0.01,
+      "step": 23150
+    },
+    {
+      "epoch": 3.6825667685225603,
+      "grad_norm": 0.15494094789028168,
+      "learning_rate": 8.855177279943554e-07,
+      "loss": 0.0003,
+      "step": 23200
+    },
+    {
+      "epoch": 3.690503591412358,
+      "grad_norm": 0.00017790609854273498,
+      "learning_rate": 8.634679837713883e-07,
+      "loss": 0.0004,
+      "step": 23250
+    },
+    {
+      "epoch": 3.698440414302155,
+      "grad_norm": 0.006829131860285997,
+      "learning_rate": 8.414182395484213e-07,
+      "loss": 0.011,
+      "step": 23300
+    },
+    {
+      "epoch": 3.706377237191952,
+      "grad_norm": 0.0003464875335339457,
+      "learning_rate": 8.193684953254542e-07,
+      "loss": 0.0092,
+      "step": 23350
+    },
+    {
+      "epoch": 3.7143140600817492,
+      "grad_norm": 0.0005877360817976296,
+      "learning_rate": 7.973187511024873e-07,
+      "loss": 0.0,
+      "step": 23400
+    },
+    {
+      "epoch": 3.7222508829715464,
+      "grad_norm": 0.0003003095625899732,
+      "learning_rate": 7.752690068795203e-07,
+      "loss": 0.0075,
+      "step": 23450
+    },
+    {
+      "epoch": 3.7301877058613435,
+      "grad_norm": 0.010633031837642193,
+      "learning_rate": 7.532192626565532e-07,
+      "loss": 0.004,
+      "step": 23500
+    },
+    {
+      "epoch": 3.7301877058613435,
+      "eval_accuracy": 0.9770634920634921,
+      "eval_loss": 0.16450679302215576,
+      "eval_macro_f1": 0.957808445275978,
+      "eval_macro_precision": 0.9706746434129168,
+      "eval_macro_recall": 0.9475056428847263,
+      "eval_micro_f1": 0.9770634920634921,
+      "eval_micro_precision": 0.9770634920634921,
+      "eval_micro_recall": 0.9770634920634921,
+      "eval_runtime": 48.1744,
+      "eval_samples_per_second": 261.55,
+      "eval_steps_per_second": 32.694,
+      "eval_weighted_f1": 0.9769934673940778,
+      "eval_weighted_precision": 0.9770703311250897,
+      "eval_weighted_recall": 0.9770634920634921,
+      "step": 23500
+    },
+    {
+      "epoch": 3.7381245287511407,
+      "grad_norm": 2.1199164390563965,
+      "learning_rate": 7.311695184335863e-07,
+      "loss": 0.0016,
+      "step": 23550
+    },
+    {
+      "epoch": 3.7460613516409382,
+      "grad_norm": 0.0003738963569048792,
+      "learning_rate": 7.095607690950785e-07,
+      "loss": 0.0131,
+      "step": 23600
+    },
+    {
+      "epoch": 3.7539981745307354,
+      "grad_norm": 6.239602953428403e-05,
+      "learning_rate": 6.875110248721116e-07,
+      "loss": 0.0167,
+      "step": 23650
+    },
+    {
+      "epoch": 3.7619349974205325,
+      "grad_norm": 0.0012471129884943366,
+      "learning_rate": 6.654612806491445e-07,
+      "loss": 0.0,
+      "step": 23700
+    },
+    {
+      "epoch": 3.7698718203103296,
+      "grad_norm": 0.00014138520054984838,
+      "learning_rate": 6.438525313106368e-07,
+      "loss": 0.0133,
+      "step": 23750
+    },
+    {
+      "epoch": 3.777808643200127,
+      "grad_norm": 0.00015882054867688566,
+      "learning_rate": 6.218027870876699e-07,
+      "loss": 0.0043,
+      "step": 23800
+    },
+    {
+      "epoch": 3.7857454660899243,
+      "grad_norm": 0.0001894296583486721,
+      "learning_rate": 5.997530428647029e-07,
+      "loss": 0.0172,
+      "step": 23850
+    },
+    {
+      "epoch": 3.7936822889797215,
+      "grad_norm": 0.0012666372349485755,
+      "learning_rate": 5.777032986417358e-07,
+      "loss": 0.0005,
+      "step": 23900
+    },
+    {
+      "epoch": 3.8016191118695186,
+      "grad_norm": 0.0015349318273365498,
+      "learning_rate": 5.556535544187688e-07,
+      "loss": 0.0022,
+      "step": 23950
+    },
+    {
+      "epoch": 3.8095559347593158,
+      "grad_norm": 0.00015529831580352038,
+      "learning_rate": 5.336038101958018e-07,
+      "loss": 0.0082,
+      "step": 24000
+    },
+    {
+      "epoch": 3.8095559347593158,
+      "eval_accuracy": 0.9775396825396825,
+      "eval_loss": 0.16624021530151367,
+      "eval_macro_f1": 0.9583518371874029,
+      "eval_macro_precision": 0.9715119738925228,
+      "eval_macro_recall": 0.947699963206565,
+      "eval_micro_f1": 0.9775396825396825,
+      "eval_micro_precision": 0.9775396825396825,
+      "eval_micro_recall": 0.9775396825396825,
+      "eval_runtime": 46.101,
+      "eval_samples_per_second": 273.313,
+      "eval_steps_per_second": 34.164,
+      "eval_weighted_f1": 0.9774702480661045,
+      "eval_weighted_precision": 0.9775614944397367,
+      "eval_weighted_recall": 0.9775396825396825,
+      "step": 24000
+    },
+    {
+      "epoch": 3.817492757649113,
+      "grad_norm": 0.00563812954351306,
+      "learning_rate": 5.115540659728347e-07,
+      "loss": 0.0135,
+      "step": 24050
+    },
+    {
+      "epoch": 3.82542958053891,
+      "grad_norm": 0.0018910864600911736,
+      "learning_rate": 4.895043217498677e-07,
+      "loss": 0.0105,
+      "step": 24100
+    },
+    {
+      "epoch": 3.8333664034287076,
+      "grad_norm": 0.00015920153236947954,
+      "learning_rate": 4.674545775269007e-07,
+      "loss": 0.0214,
+      "step": 24150
+    },
+    {
+      "epoch": 3.8413032263185047,
+      "grad_norm": 0.0004170116735622287,
+      "learning_rate": 4.454048333039337e-07,
+      "loss": 0.0073,
+      "step": 24200
+    },
+    {
+      "epoch": 3.849240049208302,
+      "grad_norm": 0.00018354528583586216,
+      "learning_rate": 4.2335508908096674e-07,
+      "loss": 0.0049,
+      "step": 24250
+    },
+    {
+      "epoch": 3.857176872098099,
+      "grad_norm": 1.992350697517395,
+      "learning_rate": 4.013053448579997e-07,
+      "loss": 0.0079,
+      "step": 24300
+    },
+    {
+      "epoch": 3.8651136949878966,
+      "grad_norm": 0.0027675540186464787,
+      "learning_rate": 3.7925560063503267e-07,
+      "loss": 0.0032,
+      "step": 24350
+    },
+    {
+      "epoch": 3.8730505178776937,
+      "grad_norm": 0.017909307032823563,
+      "learning_rate": 3.572058564120656e-07,
+      "loss": 0.0009,
+      "step": 24400
+    },
+    {
+      "epoch": 3.880987340767491,
+      "grad_norm": 0.0002593659737613052,
+      "learning_rate": 3.3515611218909865e-07,
+      "loss": 0.0089,
+      "step": 24450
+    },
+    {
+      "epoch": 3.888924163657288,
+      "grad_norm": 0.03440633788704872,
+      "learning_rate": 3.1310636796613164e-07,
+      "loss": 0.0087,
+      "step": 24500
+    },
+    {
+      "epoch": 3.888924163657288,
+      "eval_accuracy": 0.9776984126984127,
+      "eval_loss": 0.16271832585334778,
+      "eval_macro_f1": 0.9581437912558748,
+      "eval_macro_precision": 0.9706729136929291,
+      "eval_macro_recall": 0.9480541459491892,
+      "eval_micro_f1": 0.9776984126984127,
+      "eval_micro_precision": 0.9776984126984127,
+      "eval_micro_recall": 0.9776984126984127,
+      "eval_runtime": 46.9842,
+      "eval_samples_per_second": 268.175,
+      "eval_steps_per_second": 33.522,
+      "eval_weighted_f1": 0.9776414544812937,
+      "eval_weighted_precision": 0.9777223020348919,
+      "eval_weighted_recall": 0.9776984126984127,
+      "step": 24500
+    },
+    {
+      "epoch": 3.896860986547085,
+      "grad_norm": 0.00048743950901553035,
+      "learning_rate": 2.9105662374316457e-07,
+      "loss": 0.0001,
+      "step": 24550
+    },
+    {
+      "epoch": 3.9047978094368823,
+      "grad_norm": 5.859128475189209,
+      "learning_rate": 2.690068795201976e-07,
+      "loss": 0.0005,
+      "step": 24600
+    },
+    {
+      "epoch": 3.9127346323266794,
+      "grad_norm": 0.07823960483074188,
+      "learning_rate": 2.4695713529723055e-07,
+      "loss": 0.0095,
+      "step": 24650
+    },
+    {
+      "epoch": 3.920671455216477,
+      "grad_norm": 0.01893731579184532,
+      "learning_rate": 2.2490739107426357e-07,
+      "loss": 0.0035,
+      "step": 24700
+    },
+    {
+      "epoch": 3.928608278106274,
+      "grad_norm": 0.00017635202675592154,
+      "learning_rate": 2.0285764685129653e-07,
+      "loss": 0.0054,
+      "step": 24750
+    },
+    {
+      "epoch": 3.9365451009960712,
+      "grad_norm": 0.14684739708900452,
+      "learning_rate": 1.8080790262832955e-07,
+      "loss": 0.0006,
+      "step": 24800
+    },
+    {
+      "epoch": 3.9444819238858684,
+      "grad_norm": 4.845940202358179e-05,
+      "learning_rate": 1.587581584053625e-07,
+      "loss": 0.0008,
+      "step": 24850
+    },
+    {
+      "epoch": 3.952418746775666,
+      "grad_norm": 0.0005116136162541807,
+      "learning_rate": 1.367084141823955e-07,
+      "loss": 0.0004,
+      "step": 24900
+    },
+    {
+      "epoch": 3.960355569665463,
+      "grad_norm": 0.00017137936083599925,
+      "learning_rate": 1.1465866995942849e-07,
+      "loss": 0.0131,
+      "step": 24950
+    },
+    {
+      "epoch": 3.9682923925552602,
+      "grad_norm": 33.685508728027344,
+      "learning_rate": 9.260892573646146e-08,
+      "loss": 0.0025,
+      "step": 25000
+    },
+    {
+      "epoch": 3.9682923925552602,
+      "eval_accuracy": 0.9778571428571429,
+      "eval_loss": 0.16299165785312653,
+      "eval_macro_f1": 0.9587281193044483,
+      "eval_macro_precision": 0.9725024753902682,
+      "eval_macro_recall": 0.9475659044801065,
+      "eval_micro_f1": 0.9778571428571429,
+      "eval_micro_precision": 0.9778571428571429,
+      "eval_micro_recall": 0.9778571428571429,
+      "eval_runtime": 45.6663,
+      "eval_samples_per_second": 275.914,
+      "eval_steps_per_second": 34.489,
+      "eval_weighted_f1": 0.9777906938936711,
+      "eval_weighted_precision": 0.977879248064854,
+      "eval_weighted_recall": 0.9778571428571429,
+      "step": 25000
+    },
+    {
+      "epoch": 3.9762292154450574,
+      "grad_norm": 0.0008726433152332902,
+      "learning_rate": 7.055918151349445e-08,
+      "loss": 0.0029,
+      "step": 25050
+    },
+    {
+      "epoch": 3.9841660383348545,
+      "grad_norm": 0.005452016368508339,
+      "learning_rate": 4.850943729052743e-08,
+      "loss": 0.0051,
+      "step": 25100
+    },
+    {
+      "epoch": 3.9921028612246516,
+      "grad_norm": 2.1614205837249756,
+      "learning_rate": 2.645969306756042e-08,
+      "loss": 0.021,
+      "step": 25150
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 25196,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 4,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}