diff --git "a/sample_rate=0.10/checkpoint-20800/trainer_state.json" "b/sample_rate=0.10/checkpoint-20800/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/sample_rate=0.10/checkpoint-20800/trainer_state.json"
@@ -0,0 +1,5034 @@
+{
+  "best_metric": 0.7087897186605305,
+  "best_model_checkpoint": "histv4_ftis_pretrain_tssp-smlm_0329/checkpoint-20800",
+  "epoch": 305.0004460966543,
+  "eval_steps": 100,
+  "global_step": 20800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0003717472118959108,
+      "grad_norm": 266.68756103515625,
+      "learning_rate": 7.434944237918216e-07,
+      "loss": 46.6745,
+      "step": 50
+    },
+    {
+      "epoch": 1.0002379182156134,
+      "grad_norm": 117.62641906738281,
+      "learning_rate": 1.4869888475836432e-06,
+      "loss": 38.9524,
+      "step": 100
+    },
+    {
+      "epoch": 1.0002379182156134,
+      "eval_accuracy": 0.050442267150417074,
+      "eval_loss": 25.470186233520508,
+      "eval_macro_f1": 0.025968145861855703,
+      "eval_runtime": 63.0308,
+      "eval_samples_per_second": 10.661,
+      "eval_steps_per_second": 2.665,
+      "step": 100
+    },
+    {
+      "epoch": 2.000104089219331,
+      "grad_norm": 205.3363494873047,
+      "learning_rate": 2.2304832713754648e-06,
+      "loss": 32.2356,
+      "step": 150
+    },
+    {
+      "epoch": 2.0004758364312267,
+      "grad_norm": 93.07455444335938,
+      "learning_rate": 2.9739776951672864e-06,
+      "loss": 23.6969,
+      "step": 200
+    },
+    {
+      "epoch": 2.0004758364312267,
+      "eval_accuracy": 0.06100102460059542,
+      "eval_loss": 14.61188793182373,
+      "eval_macro_f1": 0.02836623982308631,
+      "eval_runtime": 58.7907,
+      "eval_samples_per_second": 11.43,
+      "eval_steps_per_second": 2.858,
+      "step": 200
+    },
+    {
+      "epoch": 3.000342007434944,
+      "grad_norm": 26.653369903564453,
+      "learning_rate": 3.717472118959108e-06,
+      "loss": 15.7121,
+      "step": 250
+    },
+    {
+      "epoch": 4.000208178438662,
+      "grad_norm": 12.779120445251465,
+      "learning_rate": 4.4609665427509296e-06,
+      "loss": 11.4531,
+      "step": 300
+    },
+    {
+      "epoch": 4.000208178438662,
+      "eval_accuracy": 0.2494337794862716,
+      "eval_loss": 8.398870468139648,
+      "eval_macro_f1": 0.07975966315085589,
+      "eval_runtime": 57.7876,
+      "eval_samples_per_second": 11.629,
+      "eval_steps_per_second": 2.907,
+      "step": 300
+    },
+    {
+      "epoch": 5.00007434944238,
+      "grad_norm": 19.828210830688477,
+      "learning_rate": 5.2044609665427516e-06,
+      "loss": 9.2358,
+      "step": 350
+    },
+    {
+      "epoch": 5.000446096654275,
+      "grad_norm": 35.386112213134766,
+      "learning_rate": 5.947955390334573e-06,
+      "loss": 7.7719,
+      "step": 400
+    },
+    {
+      "epoch": 5.000446096654275,
+      "eval_accuracy": 0.5031238102652373,
+      "eval_loss": 6.939528465270996,
+      "eval_macro_f1": 0.13061780478609422,
+      "eval_runtime": 55.3045,
+      "eval_samples_per_second": 12.151,
+      "eval_steps_per_second": 3.038,
+      "step": 400
+    },
+    {
+      "epoch": 6.000312267657993,
+      "grad_norm": 8.993448257446289,
+      "learning_rate": 6.691449814126394e-06,
+      "loss": 7.3923,
+      "step": 450
+    },
+    {
+      "epoch": 7.00017843866171,
+      "grad_norm": 41.571659088134766,
+      "learning_rate": 7.434944237918216e-06,
+      "loss": 6.7875,
+      "step": 500
+    },
+    {
+      "epoch": 7.00017843866171,
+      "eval_accuracy": 0.552664406739702,
+      "eval_loss": 6.486644744873047,
+      "eval_macro_f1": 0.1470886256918198,
+      "eval_runtime": 54.4232,
+      "eval_samples_per_second": 12.348,
+      "eval_steps_per_second": 3.087,
+      "step": 500
+    },
+    {
+      "epoch": 8.000044609665428,
+      "grad_norm": 10.922592163085938,
+      "learning_rate": 8.178438661710038e-06,
+      "loss": 6.2835,
+      "step": 550
+    },
+    {
+      "epoch": 8.000416356877324,
+      "grad_norm": 8.742582321166992,
+      "learning_rate": 8.921933085501859e-06,
+      "loss": 6.2386,
+      "step": 600
+    },
+    {
+      "epoch": 8.000416356877324,
+      "eval_accuracy": 0.5771052858930745,
+      "eval_loss": 5.778635025024414,
+      "eval_macro_f1": 0.15409460900219704,
+      "eval_runtime": 51.6608,
+      "eval_samples_per_second": 13.008,
+      "eval_steps_per_second": 3.252,
+      "step": 600
+    },
+    {
+      "epoch": 9.00028252788104,
+      "grad_norm": 7.5611162185668945,
+      "learning_rate": 9.665427509293682e-06,
+      "loss": 5.7251,
+      "step": 650
+    },
+    {
+      "epoch": 10.00014869888476,
+      "grad_norm": 19.309194564819336,
+      "learning_rate": 1.0408921933085503e-05,
+      "loss": 5.8795,
+      "step": 700
+    },
+    {
+      "epoch": 10.00014869888476,
+      "eval_accuracy": 0.5914996103544202,
+      "eval_loss": 5.594468593597412,
+      "eval_macro_f1": 0.15595900294660686,
+      "eval_runtime": 50.9102,
+      "eval_samples_per_second": 13.2,
+      "eval_steps_per_second": 3.3,
+      "step": 700
+    },
+    {
+      "epoch": 11.000014869888476,
+      "grad_norm": 31.623323440551758,
+      "learning_rate": 1.1152416356877324e-05,
+      "loss": 5.5781,
+      "step": 750
+    },
+    {
+      "epoch": 11.000386617100371,
+      "grad_norm": 9.973241806030273,
+      "learning_rate": 1.1895910780669145e-05,
+      "loss": 5.4314,
+      "step": 800
+    },
+    {
+      "epoch": 11.000386617100371,
+      "eval_accuracy": 0.6039963812217861,
+      "eval_loss": 4.915599822998047,
+      "eval_macro_f1": 0.16604226857183932,
+      "eval_runtime": 49.9902,
+      "eval_samples_per_second": 13.443,
+      "eval_steps_per_second": 3.361,
+      "step": 800
+    },
+    {
+      "epoch": 12.000252788104088,
+      "grad_norm": 7.87252950668335,
+      "learning_rate": 1.2639405204460967e-05,
+      "loss": 5.0735,
+      "step": 850
+    },
+    {
+      "epoch": 13.000118959107807,
+      "grad_norm": 6.907707691192627,
+      "learning_rate": 1.3382899628252788e-05,
+      "loss": 5.0565,
+      "step": 900
+    },
+    {
+      "epoch": 13.000118959107807,
+      "eval_accuracy": 0.6114972362936331,
+      "eval_loss": 4.2589311599731445,
+      "eval_macro_f1": 0.1720376175468991,
+      "eval_runtime": 49.3118,
+      "eval_samples_per_second": 13.628,
+      "eval_steps_per_second": 3.407,
+      "step": 900
+    },
+    {
+      "epoch": 13.000490706319702,
+      "grad_norm": 6.981306076049805,
+      "learning_rate": 1.412639405204461e-05,
+      "loss": 4.8411,
+      "step": 950
+    },
+    {
+      "epoch": 14.00035687732342,
+      "grad_norm": 7.001990795135498,
+      "learning_rate": 1.4869888475836432e-05,
+      "loss": 4.6784,
+      "step": 1000
+    },
+    {
+      "epoch": 14.00035687732342,
+      "eval_accuracy": 0.6227607074266499,
+      "eval_loss": 4.0632429122924805,
+      "eval_macro_f1": 0.17299715055148895,
+      "eval_runtime": 48.9379,
+      "eval_samples_per_second": 13.732,
+      "eval_steps_per_second": 3.433,
+      "step": 1000
+    },
+    {
+      "epoch": 15.000223048327138,
+      "grad_norm": 8.357354164123535,
+      "learning_rate": 1.5613382899628255e-05,
+      "loss": 4.4646,
+      "step": 1050
+    },
+    {
+      "epoch": 16.000089219330857,
+      "grad_norm": 6.316117763519287,
+      "learning_rate": 1.6356877323420076e-05,
+      "loss": 4.3903,
+      "step": 1100
+    },
+    {
+      "epoch": 16.000089219330857,
+      "eval_accuracy": 0.6337044380919039,
+      "eval_loss": 3.4835870265960693,
+      "eval_macro_f1": 0.18789200970805336,
+      "eval_runtime": 50.0522,
+      "eval_samples_per_second": 13.426,
+      "eval_steps_per_second": 3.356,
+      "step": 1100
+    },
+    {
+      "epoch": 16.00046096654275,
+      "grad_norm": 14.071303367614746,
+      "learning_rate": 1.7100371747211897e-05,
+      "loss": 4.2677,
+      "step": 1150
+    },
+    {
+      "epoch": 17.000327137546467,
+      "grad_norm": 7.639101982116699,
+      "learning_rate": 1.7843866171003718e-05,
+      "loss": 3.7652,
+      "step": 1200
+    },
+    {
+      "epoch": 17.000327137546467,
+      "eval_accuracy": 0.6313526240053495,
+      "eval_loss": 3.313999652862549,
+      "eval_macro_f1": 0.19168345755706534,
+      "eval_runtime": 48.171,
+      "eval_samples_per_second": 13.95,
+      "eval_steps_per_second": 3.488,
+      "step": 1200
+    },
+    {
+      "epoch": 18.000193308550188,
+      "grad_norm": 6.102409839630127,
+      "learning_rate": 1.858736059479554e-05,
+      "loss": 3.7993,
+      "step": 1250
+    },
+    {
+      "epoch": 19.000059479553904,
+      "grad_norm": 5.737588882446289,
+      "learning_rate": 1.9330855018587364e-05,
+      "loss": 3.5554,
+      "step": 1300
+    },
+    {
+      "epoch": 19.000059479553904,
+      "eval_accuracy": 0.6518905281527871,
+      "eval_loss": 2.997008800506592,
+      "eval_macro_f1": 0.19591256463059686,
+      "eval_runtime": 48.2904,
+      "eval_samples_per_second": 13.916,
+      "eval_steps_per_second": 3.479,
+      "step": 1300
+    },
+    {
+      "epoch": 19.000431226765798,
+      "grad_norm": 8.27597713470459,
+      "learning_rate": 2.0074349442379185e-05,
+      "loss": 3.5706,
+      "step": 1350
+    },
+    {
+      "epoch": 20.00029739776952,
+      "grad_norm": 7.908275127410889,
+      "learning_rate": 2.0817843866171006e-05,
+      "loss": 3.2164,
+      "step": 1400
+    },
+    {
+      "epoch": 20.00029739776952,
+      "eval_accuracy": 0.6618917672038582,
+      "eval_loss": 2.7557449340820312,
+      "eval_macro_f1": 0.20366848025691103,
+      "eval_runtime": 48.1076,
+      "eval_samples_per_second": 13.969,
+      "eval_steps_per_second": 3.492,
+      "step": 1400
+    },
+    {
+      "epoch": 21.000163568773235,
+      "grad_norm": 10.48082447052002,
+      "learning_rate": 2.1561338289962827e-05,
+      "loss": 2.9457,
+      "step": 1450
+    },
+    {
+      "epoch": 22.000029739776952,
+      "grad_norm": 3.876420497894287,
+      "learning_rate": 2.230483271375465e-05,
+      "loss": 2.9022,
+      "step": 1500
+    },
+    {
+      "epoch": 22.000029739776952,
+      "eval_accuracy": 0.6745563484687298,
+      "eval_loss": 2.527372360229492,
+      "eval_macro_f1": 0.2157652452048655,
+      "eval_runtime": 48.1074,
+      "eval_samples_per_second": 13.969,
+      "eval_steps_per_second": 3.492,
+      "step": 1500
+    },
+    {
+      "epoch": 22.00040148698885,
+      "grad_norm": 7.9988932609558105,
+      "learning_rate": 2.304832713754647e-05,
+      "loss": 2.8802,
+      "step": 1550
+    },
+    {
+      "epoch": 23.000267657992566,
+      "grad_norm": 12.270583152770996,
+      "learning_rate": 2.379182156133829e-05,
+      "loss": 2.745,
+      "step": 1600
+    },
+    {
+      "epoch": 23.000267657992566,
+      "eval_accuracy": 0.6913894280781943,
+      "eval_loss": 2.355086088180542,
+      "eval_macro_f1": 0.23149517864191574,
+      "eval_runtime": 48.147,
+      "eval_samples_per_second": 13.957,
+      "eval_steps_per_second": 3.489,
+      "step": 1600
+    },
+    {
+      "epoch": 24.000133828996283,
+      "grad_norm": 5.4861040115356445,
+      "learning_rate": 2.4535315985130116e-05,
+      "loss": 2.6132,
+      "step": 1650
+    },
+    {
+      "epoch": 24.000505576208177,
+      "grad_norm": 6.245904445648193,
+      "learning_rate": 2.5278810408921933e-05,
+      "loss": 2.4721,
+      "step": 1700
+    },
+    {
+      "epoch": 24.000505576208177,
+      "eval_accuracy": 0.6931836196159022,
+      "eval_loss": 2.360691547393799,
+      "eval_macro_f1": 0.24161708403963666,
+      "eval_runtime": 46.7909,
+      "eval_samples_per_second": 14.362,
+      "eval_steps_per_second": 3.59,
+      "step": 1700
+    },
+    {
+      "epoch": 25.000371747211897,
+      "grad_norm": 3.8083865642547607,
+      "learning_rate": 2.6022304832713758e-05,
+      "loss": 2.4046,
+      "step": 1750
+    },
+    {
+      "epoch": 26.000237918215614,
+      "grad_norm": 10.2521333694458,
+      "learning_rate": 2.6765799256505576e-05,
+      "loss": 2.3065,
+      "step": 1800
+    },
+    {
+      "epoch": 26.000237918215614,
+      "eval_accuracy": 0.7109710905521758,
+      "eval_loss": 2.1845011711120605,
+      "eval_macro_f1": 0.25447139713367894,
+      "eval_runtime": 48.4023,
+      "eval_samples_per_second": 13.884,
+      "eval_steps_per_second": 3.471,
+      "step": 1800
+    },
+    {
+      "epoch": 27.00010408921933,
+      "grad_norm": 4.3791022300720215,
+      "learning_rate": 2.75092936802974e-05,
+      "loss": 2.328,
+      "step": 1850
+    },
+    {
+      "epoch": 27.000475836431228,
+      "grad_norm": 5.141542434692383,
+      "learning_rate": 2.825278810408922e-05,
+      "loss": 2.3132,
+      "step": 1900
+    },
+    {
+      "epoch": 27.000475836431228,
+      "eval_accuracy": 0.7144459720258327,
+      "eval_loss": 2.155479907989502,
+      "eval_macro_f1": 0.25474356578920043,
+      "eval_runtime": 48.1297,
+      "eval_samples_per_second": 13.962,
+      "eval_steps_per_second": 3.491,
+      "step": 1900
+    },
+    {
+      "epoch": 28.000342007434945,
+      "grad_norm": 7.797635078430176,
+      "learning_rate": 2.8996282527881043e-05,
+      "loss": 2.1353,
+      "step": 1950
+    },
+    {
+      "epoch": 29.000208178438662,
+      "grad_norm": 6.269540309906006,
+      "learning_rate": 2.9739776951672864e-05,
+      "loss": 2.069,
+      "step": 2000
+    },
+    {
+      "epoch": 29.000208178438662,
+      "eval_accuracy": 0.7228523113819296,
+      "eval_loss": 2.0353355407714844,
+      "eval_macro_f1": 0.2780559155369861,
+      "eval_runtime": 48.912,
+      "eval_samples_per_second": 13.739,
+      "eval_steps_per_second": 3.435,
+      "step": 2000
+    },
+    {
+      "epoch": 30.00007434944238,
+      "grad_norm": 7.6095099449157715,
+      "learning_rate": 3.0483271375464685e-05,
+      "loss": 2.0509,
+      "step": 2050
+    },
+    {
+      "epoch": 30.000446096654276,
+      "grad_norm": 7.309573650360107,
+      "learning_rate": 3.122676579925651e-05,
+      "loss": 2.065,
+      "step": 2100
+    },
+    {
+      "epoch": 30.000446096654276,
+      "eval_accuracy": 0.7238860947101218,
+      "eval_loss": 2.0333032608032227,
+      "eval_macro_f1": 0.2880652536535786,
+      "eval_runtime": 48.1719,
+      "eval_samples_per_second": 13.95,
+      "eval_steps_per_second": 3.488,
+      "step": 2100
+    },
+    {
+      "epoch": 31.000312267657993,
+      "grad_norm": 6.504179000854492,
+      "learning_rate": 3.1970260223048324e-05,
+      "loss": 1.9473,
+      "step": 2150
+    },
+    {
+      "epoch": 32.00017843866171,
+      "grad_norm": 7.720032215118408,
+      "learning_rate": 3.271375464684015e-05,
+      "loss": 1.9441,
+      "step": 2200
+    },
+    {
+      "epoch": 32.00017843866171,
+      "eval_accuracy": 0.7312337507081383,
+      "eval_loss": 2.0020694732666016,
+      "eval_macro_f1": 0.2978414990821893,
+      "eval_runtime": 48.306,
+      "eval_samples_per_second": 13.911,
+      "eval_steps_per_second": 3.478,
+      "step": 2200
+    },
+    {
+      "epoch": 33.00004460966543,
+      "grad_norm": 3.4699764251708984,
+      "learning_rate": 3.345724907063197e-05,
+      "loss": 1.8625,
+      "step": 2250
+    },
+    {
+      "epoch": 33.000416356877324,
+      "grad_norm": 7.346961975097656,
+      "learning_rate": 3.4200743494423794e-05,
+      "loss": 1.8603,
+      "step": 2300
+    },
+    {
+      "epoch": 33.000416356877324,
+      "eval_accuracy": 0.7178409171443214,
+      "eval_loss": 2.070431709289551,
+      "eval_macro_f1": 0.29655410586263253,
+      "eval_runtime": 48.3374,
+      "eval_samples_per_second": 13.902,
+      "eval_steps_per_second": 3.476,
+      "step": 2300
+    },
+    {
+      "epoch": 34.00028252788104,
+      "grad_norm": 7.2747015953063965,
+      "learning_rate": 3.4944237918215615e-05,
+      "loss": 1.8716,
+      "step": 2350
+    },
+    {
+      "epoch": 35.00014869888476,
+      "grad_norm": 27.447582244873047,
+      "learning_rate": 3.5687732342007436e-05,
+      "loss": 1.8589,
+      "step": 2400
+    },
+    {
+      "epoch": 35.00014869888476,
+      "eval_accuracy": 0.7422280762176736,
+      "eval_loss": 1.863403081893921,
+      "eval_macro_f1": 0.33096456369239546,
+      "eval_runtime": 47.8041,
+      "eval_samples_per_second": 14.057,
+      "eval_steps_per_second": 3.514,
+      "step": 2400
+    },
+    {
+      "epoch": 36.000014869888474,
+      "grad_norm": 4.940944194793701,
+      "learning_rate": 3.643122676579926e-05,
+      "loss": 1.7339,
+      "step": 2450
+    },
+    {
+      "epoch": 36.000386617100375,
+      "grad_norm": 4.48024845123291,
+      "learning_rate": 3.717472118959108e-05,
+      "loss": 1.6732,
+      "step": 2500
+    },
+    {
+      "epoch": 36.000386617100375,
+      "eval_accuracy": 0.7466635662811762,
+      "eval_loss": 1.8566985130310059,
+      "eval_macro_f1": 0.3421978183973209,
+      "eval_runtime": 48.6552,
+      "eval_samples_per_second": 13.811,
+      "eval_steps_per_second": 3.453,
+      "step": 2500
+    },
+    {
+      "epoch": 37.00025278810409,
+      "grad_norm": 5.992680072784424,
+      "learning_rate": 3.79182156133829e-05,
+      "loss": 1.6829,
+      "step": 2550
+    },
+    {
+      "epoch": 38.00011895910781,
+      "grad_norm": 3.3229565620422363,
+      "learning_rate": 3.866171003717473e-05,
+      "loss": 1.6328,
+      "step": 2600
+    },
+    {
+      "epoch": 38.00011895910781,
+      "eval_accuracy": 0.7481501704372262,
+      "eval_loss": 1.8362157344818115,
+      "eval_macro_f1": 0.3485086625465703,
+      "eval_runtime": 48.0664,
+      "eval_samples_per_second": 13.981,
+      "eval_steps_per_second": 3.495,
+      "step": 2600
+    },
+    {
+      "epoch": 38.0004907063197,
+      "grad_norm": 5.226800441741943,
+      "learning_rate": 3.940520446096654e-05,
+      "loss": 1.6336,
+      "step": 2650
+    },
+    {
+      "epoch": 39.00035687732342,
+      "grad_norm": 4.662332057952881,
+      "learning_rate": 4.014869888475837e-05,
+      "loss": 1.5936,
+      "step": 2700
+    },
+    {
+      "epoch": 39.00035687732342,
+      "eval_accuracy": 0.7517626786767897,
+      "eval_loss": 1.7911369800567627,
+      "eval_macro_f1": 0.37362545278695913,
+      "eval_runtime": 48.5579,
+      "eval_samples_per_second": 13.839,
+      "eval_steps_per_second": 3.46,
+      "step": 2700
+    },
+    {
+      "epoch": 40.000223048327136,
+      "grad_norm": 5.28360652923584,
+      "learning_rate": 4.0892193308550185e-05,
+      "loss": 1.5267,
+      "step": 2750
+    },
+    {
+      "epoch": 41.00008921933085,
+      "grad_norm": 13.149215698242188,
+      "learning_rate": 4.163568773234201e-05,
+      "loss": 1.5476,
+      "step": 2800
+    },
+    {
+      "epoch": 41.00008921933085,
+      "eval_accuracy": 0.743210771307319,
+      "eval_loss": 1.9220155477523804,
+      "eval_macro_f1": 0.3483976776394514,
+      "eval_runtime": 46.7958,
+      "eval_samples_per_second": 14.36,
+      "eval_steps_per_second": 3.59,
+      "step": 2800
+    },
+    {
+      "epoch": 41.000460966542754,
+      "grad_norm": 4.578529357910156,
+      "learning_rate": 4.237918215613383e-05,
+      "loss": 1.512,
+      "step": 2850
+    },
+    {
+      "epoch": 42.00032713754647,
+      "grad_norm": 5.931751728057861,
+      "learning_rate": 4.3122676579925655e-05,
+      "loss": 1.4547,
+      "step": 2900
+    },
+    {
+      "epoch": 42.00032713754647,
+      "eval_accuracy": 0.7573206038563081,
+      "eval_loss": 1.7732130289077759,
+      "eval_macro_f1": 0.3823146615563383,
+      "eval_runtime": 47.9976,
+      "eval_samples_per_second": 14.001,
+      "eval_steps_per_second": 3.5,
+      "step": 2900
+    },
+    {
+      "epoch": 43.00019330855019,
+      "grad_norm": 6.435998916625977,
+      "learning_rate": 4.3866171003717476e-05,
+      "loss": 1.4742,
+      "step": 2950
+    },
+    {
+      "epoch": 44.000059479553904,
+      "grad_norm": 5.018718242645264,
+      "learning_rate": 4.46096654275093e-05,
+      "loss": 1.4111,
+      "step": 3000
+    },
+    {
+      "epoch": 44.000059479553904,
+      "eval_accuracy": 0.7604534766737875,
+      "eval_loss": 1.8028564453125,
+      "eval_macro_f1": 0.3851260308276024,
+      "eval_runtime": 47.8675,
+      "eval_samples_per_second": 14.039,
+      "eval_steps_per_second": 3.51,
+      "step": 3000
+    },
+    {
+      "epoch": 44.0004312267658,
+      "grad_norm": 6.0326948165893555,
+      "learning_rate": 4.535315985130112e-05,
+      "loss": 1.4035,
+      "step": 3050
+    },
+    {
+      "epoch": 45.000297397769515,
+      "grad_norm": 6.028076171875,
+      "learning_rate": 4.609665427509294e-05,
+      "loss": 1.3497,
+      "step": 3100
+    },
+    {
+      "epoch": 45.000297397769515,
+      "eval_accuracy": 0.768372023656696,
+      "eval_loss": 1.7707154750823975,
+      "eval_macro_f1": 0.4061185384890694,
+      "eval_runtime": 47.5778,
+      "eval_samples_per_second": 14.124,
+      "eval_steps_per_second": 3.531,
+      "step": 3100
+    },
+    {
+      "epoch": 46.00016356877323,
+      "grad_norm": 4.857266426086426,
+      "learning_rate": 4.684014869888476e-05,
+      "loss": 1.3224,
+      "step": 3150
+    },
+    {
+      "epoch": 47.00002973977695,
+      "grad_norm": 4.493041038513184,
+      "learning_rate": 4.758364312267658e-05,
+      "loss": 1.3677,
+      "step": 3200
+    },
+    {
+      "epoch": 47.00002973977695,
+      "eval_accuracy": 0.7547429832358992,
+      "eval_loss": 1.8069781064987183,
+      "eval_macro_f1": 0.4303426673990079,
+      "eval_runtime": 47.9766,
+      "eval_samples_per_second": 14.007,
+      "eval_steps_per_second": 3.502,
+      "step": 3200
+    },
+    {
+      "epoch": 47.00040148698885,
+      "grad_norm": 5.508673191070557,
+      "learning_rate": 4.83271375464684e-05,
+      "loss": 1.2983,
+      "step": 3250
+    },
+    {
+      "epoch": 48.000267657992566,
+      "grad_norm": 2.9697608947753906,
+      "learning_rate": 4.907063197026023e-05,
+      "loss": 1.2558,
+      "step": 3300
+    },
+    {
+      "epoch": 48.000267657992566,
+      "eval_accuracy": 0.770120629172458,
+      "eval_loss": 1.7805372476577759,
+      "eval_macro_f1": 0.4152670510054953,
+      "eval_runtime": 47.6206,
+      "eval_samples_per_second": 14.112,
+      "eval_steps_per_second": 3.528,
+      "step": 3300
+    },
+    {
+      "epoch": 49.00013382899628,
+      "grad_norm": 3.1506409645080566,
+      "learning_rate": 4.9814126394052045e-05,
+      "loss": 1.2456,
+      "step": 3350
+    },
+    {
+      "epoch": 49.00050557620818,
+      "grad_norm": 2.9008400440216064,
+      "learning_rate": 5.0557620817843867e-05,
+      "loss": 1.2228,
+      "step": 3400
+    },
+    {
+      "epoch": 49.00050557620818,
+      "eval_accuracy": 0.7753736194741223,
+      "eval_loss": 1.6754295825958252,
+      "eval_macro_f1": 0.4497377753131999,
+      "eval_runtime": 46.8789,
+      "eval_samples_per_second": 14.335,
+      "eval_steps_per_second": 3.584,
+      "step": 3400
+    },
+    {
+      "epoch": 50.000371747211894,
+      "grad_norm": 14.068544387817383,
+      "learning_rate": 5.130111524163569e-05,
+      "loss": 1.2156,
+      "step": 3450
+    },
+    {
+      "epoch": 51.00023791821561,
+      "grad_norm": 7.377778053283691,
+      "learning_rate": 5.2044609665427516e-05,
+      "loss": 1.1783,
+      "step": 3500
+    },
+    {
+      "epoch": 51.00023791821561,
+      "eval_accuracy": 0.7714801943968466,
+      "eval_loss": 1.7035390138626099,
+      "eval_macro_f1": 0.43929205620995815,
+      "eval_runtime": 47.4365,
+      "eval_samples_per_second": 14.166,
+      "eval_steps_per_second": 3.542,
+      "step": 3500
+    },
+    {
+      "epoch": 52.00010408921933,
+      "grad_norm": 5.86587381362915,
+      "learning_rate": 5.278810408921934e-05,
+      "loss": 1.1385,
+      "step": 3550
+    },
+    {
+      "epoch": 52.00047583643123,
+      "grad_norm": 7.151806831359863,
+      "learning_rate": 5.353159851301115e-05,
+      "loss": 1.1714,
+      "step": 3600
+    },
+    {
+      "epoch": 52.00047583643123,
+      "eval_accuracy": 0.776484537461446,
+      "eval_loss": 1.6419715881347656,
+      "eval_macro_f1": 0.46826783464883714,
+      "eval_runtime": 47.8203,
+      "eval_samples_per_second": 14.053,
+      "eval_steps_per_second": 3.513,
+      "step": 3600
+    },
+    {
+      "epoch": 53.000342007434945,
+      "grad_norm": 3.089052677154541,
+      "learning_rate": 5.427509293680297e-05,
+      "loss": 1.1051,
+      "step": 3650
+    },
+    {
+      "epoch": 54.00020817843866,
+      "grad_norm": 4.17380428314209,
+      "learning_rate": 5.50185873605948e-05,
+      "loss": 1.1496,
+      "step": 3700
+    },
+    {
+      "epoch": 54.00020817843866,
+      "eval_accuracy": 0.783282215260751,
+      "eval_loss": 1.6009899377822876,
+      "eval_macro_f1": 0.4684935245173467,
+      "eval_runtime": 47.1333,
+      "eval_samples_per_second": 14.257,
+      "eval_steps_per_second": 3.564,
+      "step": 3700
+    },
+    {
+      "epoch": 55.00007434944238,
+      "grad_norm": 3.879401683807373,
+      "learning_rate": 5.576208178438662e-05,
+      "loss": 1.123,
+      "step": 3750
+    },
+    {
+      "epoch": 55.00044609665427,
+      "grad_norm": 5.521020889282227,
+      "learning_rate": 5.650557620817844e-05,
+      "loss": 1.0547,
+      "step": 3800
+    },
+    {
+      "epoch": 55.00044609665427,
+      "eval_accuracy": 0.7832706110213381,
+      "eval_loss": 1.6715576648712158,
+      "eval_macro_f1": 0.46885622369709506,
+      "eval_runtime": 47.4503,
+      "eval_samples_per_second": 14.162,
+      "eval_steps_per_second": 3.541,
+      "step": 3800
+    },
+    {
+      "epoch": 56.00031226765799,
+      "grad_norm": 6.454986095428467,
+      "learning_rate": 5.724907063197026e-05,
+      "loss": 1.0531,
+      "step": 3850
+    },
+    {
+      "epoch": 57.00017843866171,
+      "grad_norm": 9.67890453338623,
+      "learning_rate": 5.7992565055762085e-05,
+      "loss": 1.0515,
+      "step": 3900
+    },
+    {
+      "epoch": 57.00017843866171,
+      "eval_accuracy": 0.787819132865917,
+      "eval_loss": 1.6513065099716187,
+      "eval_macro_f1": 0.4838647951545401,
+      "eval_runtime": 47.7738,
+      "eval_samples_per_second": 14.066,
+      "eval_steps_per_second": 3.517,
+      "step": 3900
+    },
+    {
+      "epoch": 58.00004460966543,
+      "grad_norm": 9.32218074798584,
+      "learning_rate": 5.8736059479553906e-05,
+      "loss": 1.0074,
+      "step": 3950
+    },
+    {
+      "epoch": 58.000416356877324,
+      "grad_norm": 3.381666898727417,
+      "learning_rate": 5.947955390334573e-05,
+      "loss": 1.0019,
+      "step": 4000
+    },
+    {
+      "epoch": 58.000416356877324,
+      "eval_accuracy": 0.7864938662716643,
+      "eval_loss": 1.6623215675354004,
+      "eval_macro_f1": 0.4932068292621268,
+      "eval_runtime": 47.6569,
+      "eval_samples_per_second": 14.101,
+      "eval_steps_per_second": 3.525,
+      "step": 4000
+    },
+    {
+      "epoch": 59.00028252788104,
+      "grad_norm": 5.850093364715576,
+      "learning_rate": 6.0223048327137555e-05,
+      "loss": 1.0,
+      "step": 4050
+    },
+    {
+      "epoch": 60.00014869888476,
+      "grad_norm": 3.286019802093506,
+      "learning_rate": 6.096654275092937e-05,
+      "loss": 0.9517,
+      "step": 4100
+    },
+    {
+      "epoch": 60.00014869888476,
+      "eval_accuracy": 0.788711438830616,
+      "eval_loss": 1.5527822971343994,
+      "eval_macro_f1": 0.4998529689254287,
+      "eval_runtime": 47.1648,
+      "eval_samples_per_second": 14.248,
+      "eval_steps_per_second": 3.562,
+      "step": 4100
+    },
+    {
+      "epoch": 61.000014869888474,
+      "grad_norm": 5.813722133636475,
+      "learning_rate": 6.171003717472119e-05,
+      "loss": 0.95,
+      "step": 4150
+    },
+    {
+      "epoch": 61.000386617100375,
+      "grad_norm": 6.047854423522949,
+      "learning_rate": 6.245353159851302e-05,
+      "loss": 0.9249,
+      "step": 4200
+    },
+    {
+      "epoch": 61.000386617100375,
+      "eval_accuracy": 0.7949442146268098,
+      "eval_loss": 1.552842617034912,
+      "eval_macro_f1": 0.5062827822451884,
+      "eval_runtime": 47.7291,
+      "eval_samples_per_second": 14.079,
+      "eval_steps_per_second": 3.52,
+      "step": 4200
+    },
+    {
+      "epoch": 62.00025278810409,
+      "grad_norm": 7.298818588256836,
+      "learning_rate": 6.319702602230483e-05,
+      "loss": 0.9241,
+      "step": 4250
+    },
+    {
+      "epoch": 63.00011895910781,
+      "grad_norm": 3.9685351848602295,
+      "learning_rate": 6.394052044609665e-05,
+      "loss": 0.9072,
+      "step": 4300
+    },
+    {
+      "epoch": 63.00011895910781,
+      "eval_accuracy": 0.7972720697174538,
+      "eval_loss": 1.595497488975525,
+      "eval_macro_f1": 0.5229591985542493,
+      "eval_runtime": 47.7421,
+      "eval_samples_per_second": 14.076,
+      "eval_steps_per_second": 3.519,
+      "step": 4300
+    },
+    {
+      "epoch": 63.0004907063197,
+      "grad_norm": 6.513078689575195,
+      "learning_rate": 6.468401486988848e-05,
+      "loss": 0.9064,
+      "step": 4350
+    },
+    {
+      "epoch": 64.00035687732343,
+      "grad_norm": 9.337983131408691,
+      "learning_rate": 6.54275092936803e-05,
+      "loss": 0.8861,
+      "step": 4400
+    },
+    {
+      "epoch": 64.00035687732343,
+      "eval_accuracy": 0.8012236394427585,
+      "eval_loss": 1.577151894569397,
+      "eval_macro_f1": 0.5266623902892644,
+      "eval_runtime": 47.7981,
+      "eval_samples_per_second": 14.059,
+      "eval_steps_per_second": 3.515,
+      "step": 4400
+    },
+    {
+      "epoch": 65.00022304832714,
+      "grad_norm": 6.658684730529785,
+      "learning_rate": 6.617100371747212e-05,
+      "loss": 0.8829,
+      "step": 4450
+    },
+    {
+      "epoch": 66.00008921933086,
+      "grad_norm": 7.026724815368652,
+      "learning_rate": 6.691449814126395e-05,
+      "loss": 0.8687,
+      "step": 4500
+    },
+    {
+      "epoch": 66.00008921933086,
+      "eval_accuracy": 0.8062816088160649,
+      "eval_loss": 1.5187242031097412,
+      "eval_macro_f1": 0.5236433362876899,
+      "eval_runtime": 47.5759,
+      "eval_samples_per_second": 14.125,
+      "eval_steps_per_second": 3.531,
+      "step": 4500
+    },
+    {
+      "epoch": 66.00046096654275,
+      "grad_norm": 3.1086082458496094,
+      "learning_rate": 6.765799256505576e-05,
+      "loss": 0.8541,
+      "step": 4550
+    },
+    {
+      "epoch": 67.00032713754646,
+      "grad_norm": 2.039297342300415,
+      "learning_rate": 6.840148698884759e-05,
+      "loss": 0.8164,
+      "step": 4600
+    },
+    {
+      "epoch": 67.00032713754646,
+      "eval_accuracy": 0.8023215844672498,
+      "eval_loss": 1.605806589126587,
+      "eval_macro_f1": 0.5408551755484401,
+      "eval_runtime": 47.4302,
+      "eval_samples_per_second": 14.168,
+      "eval_steps_per_second": 3.542,
+      "step": 4600
+    },
+    {
+      "epoch": 68.00019330855018,
+      "grad_norm": 5.141190528869629,
+      "learning_rate": 6.91449814126394e-05,
+      "loss": 0.8411,
+      "step": 4650
+    },
+    {
+      "epoch": 69.0000594795539,
+      "grad_norm": 6.119557857513428,
+      "learning_rate": 6.988847583643123e-05,
+      "loss": 0.7923,
+      "step": 4700
+    },
+    {
+      "epoch": 69.0000594795539,
+      "eval_accuracy": 0.7978375042402981,
+      "eval_loss": 1.5691826343536377,
+      "eval_macro_f1": 0.541571467839333,
+      "eval_runtime": 47.0961,
+      "eval_samples_per_second": 14.269,
+      "eval_steps_per_second": 3.567,
+      "step": 4700
+    },
+    {
+      "epoch": 69.0004312267658,
+      "grad_norm": 5.380624294281006,
+      "learning_rate": 7.063197026022306e-05,
+      "loss": 0.8052,
+      "step": 4750
+    },
+    {
+      "epoch": 70.00029739776951,
+      "grad_norm": 7.781818389892578,
+      "learning_rate": 7.137546468401487e-05,
+      "loss": 0.8065,
+      "step": 4800
+    },
+    {
+      "epoch": 70.00029739776951,
+      "eval_accuracy": 0.7904748964659136,
+      "eval_loss": 1.6525987386703491,
+      "eval_macro_f1": 0.5321248674751812,
+      "eval_runtime": 47.8684,
+      "eval_samples_per_second": 14.038,
+      "eval_steps_per_second": 3.51,
+      "step": 4800
+    },
+    {
+      "epoch": 71.00016356877323,
+      "grad_norm": 2.671602249145508,
+      "learning_rate": 7.211895910780669e-05,
+      "loss": 0.7815,
+      "step": 4850
+    },
+    {
+      "epoch": 72.00002973977695,
+      "grad_norm": 4.526907444000244,
+      "learning_rate": 7.286245353159852e-05,
+      "loss": 0.771,
+      "step": 4900
+    },
+    {
+      "epoch": 72.00002973977695,
+      "eval_accuracy": 0.7965682882088077,
+      "eval_loss": 1.6163409948349,
+      "eval_macro_f1": 0.5339910056580364,
+      "eval_runtime": 47.2867,
+      "eval_samples_per_second": 14.211,
+      "eval_steps_per_second": 3.553,
+      "step": 4900
+    },
+    {
+      "epoch": 72.00040148698885,
+      "grad_norm": 4.758114337921143,
+      "learning_rate": 7.360594795539034e-05,
+      "loss": 0.7414,
+      "step": 4950
+    },
+    {
+      "epoch": 73.00026765799257,
+      "grad_norm": 4.144097805023193,
+      "learning_rate": 7.434944237918216e-05,
+      "loss": 0.7377,
+      "step": 5000
+    },
+    {
+      "epoch": 73.00026765799257,
+      "eval_accuracy": 0.8146328689405671,
+      "eval_loss": 1.5363017320632935,
+      "eval_macro_f1": 0.5563859432409435,
+      "eval_runtime": 47.0425,
+      "eval_samples_per_second": 14.285,
+      "eval_steps_per_second": 3.571,
+      "step": 5000
+    },
+    {
+      "epoch": 74.00013382899628,
+      "grad_norm": 4.5283284187316895,
+      "learning_rate": 7.509293680297399e-05,
+      "loss": 0.7373,
+      "step": 5050
+    },
+    {
+      "epoch": 74.00050557620818,
+      "grad_norm": 4.018579959869385,
+      "learning_rate": 7.58364312267658e-05,
+      "loss": 0.7194,
+      "step": 5100
+    },
+    {
+      "epoch": 74.00050557620818,
+      "eval_accuracy": 0.8134706343704267,
+      "eval_loss": 1.5270371437072754,
+      "eval_macro_f1": 0.5648332270298382,
+      "eval_runtime": 47.1218,
+      "eval_samples_per_second": 14.261,
+      "eval_steps_per_second": 3.565,
+      "step": 5100
+    },
+    {
+      "epoch": 75.0003717472119,
+      "grad_norm": 4.909033298492432,
+      "learning_rate": 7.657992565055763e-05,
+      "loss": 0.7112,
+      "step": 5150
+    },
+    {
+      "epoch": 76.00023791821562,
+      "grad_norm": 6.122696876525879,
+      "learning_rate": 7.732342007434946e-05,
+      "loss": 0.706,
+      "step": 5200
+    },
+    {
+      "epoch": 76.00023791821562,
+      "eval_accuracy": 0.8068405032797368,
+      "eval_loss": 1.5804245471954346,
+      "eval_macro_f1": 0.5562806947678465,
+      "eval_runtime": 47.508,
+      "eval_samples_per_second": 14.145,
+      "eval_steps_per_second": 3.536,
+      "step": 5200
+    },
+    {
+      "epoch": 77.00010408921933,
+      "grad_norm": 3.1820068359375,
+      "learning_rate": 7.806691449814127e-05,
+      "loss": 0.6735,
+      "step": 5250
+    },
+    {
+      "epoch": 77.00047583643122,
+      "grad_norm": 3.543308734893799,
+      "learning_rate": 7.881040892193308e-05,
+      "loss": 0.6827,
+      "step": 5300
+    },
+    {
+      "epoch": 77.00047583643122,
+      "eval_accuracy": 0.8132421110520636,
+      "eval_loss": 1.5398247241973877,
+      "eval_macro_f1": 0.5569486652835459,
+      "eval_runtime": 46.848,
+      "eval_samples_per_second": 14.344,
+      "eval_steps_per_second": 3.586,
+      "step": 5300
+    },
+    {
+      "epoch": 78.00034200743494,
+      "grad_norm": 4.693170547485352,
+      "learning_rate": 7.955390334572491e-05,
+      "loss": 0.6835,
+      "step": 5350
+    },
+    {
+      "epoch": 79.00020817843865,
+      "grad_norm": 2.5306339263916016,
+      "learning_rate": 8.029739776951674e-05,
+      "loss": 0.6852,
+      "step": 5400
+    },
+    {
+      "epoch": 79.00020817843865,
+      "eval_accuracy": 0.8227611279450846,
+      "eval_loss": 1.4755533933639526,
+      "eval_macro_f1": 0.5733192275425867,
+      "eval_runtime": 47.7076,
+      "eval_samples_per_second": 14.086,
+      "eval_steps_per_second": 3.521,
+      "step": 5400
+    },
+    {
+      "epoch": 80.00007434944239,
+      "grad_norm": 2.764157295227051,
+      "learning_rate": 8.104089219330855e-05,
+      "loss": 0.6602,
+      "step": 5450
+    },
+    {
+      "epoch": 80.00044609665427,
+      "grad_norm": 3.0805418491363525,
+      "learning_rate": 8.178438661710037e-05,
+      "loss": 0.6333,
+      "step": 5500
+    },
+    {
+      "epoch": 80.00044609665427,
+      "eval_accuracy": 0.8215874176654338,
+      "eval_loss": 1.4958947896957397,
+      "eval_macro_f1": 0.5765976404305969,
+      "eval_runtime": 47.5831,
+      "eval_samples_per_second": 14.123,
+      "eval_steps_per_second": 3.531,
+      "step": 5500
+    },
+    {
+      "epoch": 81.00031226765799,
+      "grad_norm": 3.5253844261169434,
+      "learning_rate": 8.25278810408922e-05,
+      "loss": 0.6417,
+      "step": 5550
+    },
+    {
+      "epoch": 82.0001784386617,
+      "grad_norm": 3.3636598587036133,
+      "learning_rate": 8.327137546468403e-05,
+      "loss": 0.6288,
+      "step": 5600
+    },
+    {
+      "epoch": 82.0001784386617,
+      "eval_accuracy": 0.8281630899930705,
+      "eval_loss": 1.4461908340454102,
+      "eval_macro_f1": 0.5842528545422506,
+      "eval_runtime": 47.5162,
+      "eval_samples_per_second": 14.143,
+      "eval_steps_per_second": 3.536,
+      "step": 5600
+    },
+    {
+      "epoch": 83.00004460966542,
+      "grad_norm": 3.808940887451172,
+      "learning_rate": 8.401486988847584e-05,
+      "loss": 0.6302,
+      "step": 5650
+    },
+    {
+      "epoch": 83.00041635687732,
+      "grad_norm": 2.76824951171875,
+      "learning_rate": 8.475836431226765e-05,
+      "loss": 0.6165,
+      "step": 5700
+    },
+    {
+      "epoch": 83.00041635687732,
+      "eval_accuracy": 0.8236607576629754,
+      "eval_loss": 1.5466784238815308,
+      "eval_macro_f1": 0.5788110277050088,
+      "eval_runtime": 46.9322,
+      "eval_samples_per_second": 14.319,
+      "eval_steps_per_second": 3.58,
+      "step": 5700
+    },
+    {
+      "epoch": 84.00028252788104,
+      "grad_norm": 3.550778388977051,
+      "learning_rate": 8.550185873605948e-05,
+      "loss": 0.6126,
+      "step": 5750
+    },
+    {
+      "epoch": 85.00014869888476,
+      "grad_norm": 4.159547805786133,
+      "learning_rate": 8.624535315985131e-05,
+      "loss": 0.6153,
+      "step": 5800
+    },
+    {
+      "epoch": 85.00014869888476,
+      "eval_accuracy": 0.8248911419166896,
+      "eval_loss": 1.5206892490386963,
+      "eval_macro_f1": 0.5832331531893454,
+      "eval_runtime": 47.5885,
+      "eval_samples_per_second": 14.121,
+      "eval_steps_per_second": 3.53,
+      "step": 5800
+    },
+    {
+      "epoch": 86.00001486988847,
+      "grad_norm": 3.1821999549865723,
+      "learning_rate": 8.698884758364312e-05,
+      "loss": 0.5915,
+      "step": 5850
+    },
+    {
+      "epoch": 86.00038661710038,
+      "grad_norm": 3.598517894744873,
+      "learning_rate": 8.773234200743495e-05,
+      "loss": 0.593,
+      "step": 5900
+    },
+    {
+      "epoch": 86.00038661710038,
+      "eval_accuracy": 0.8303127795000983,
+      "eval_loss": 1.5167983770370483,
+      "eval_macro_f1": 0.5917533935658699,
+      "eval_runtime": 47.842,
+      "eval_samples_per_second": 14.046,
+      "eval_steps_per_second": 3.512,
+      "step": 5900
+    },
+    {
+      "epoch": 87.00025278810409,
+      "grad_norm": 2.605105400085449,
+      "learning_rate": 8.847583643122677e-05,
+      "loss": 0.5804,
+      "step": 5950
+    },
+    {
+      "epoch": 88.00011895910781,
+      "grad_norm": 7.943371295928955,
+      "learning_rate": 8.92193308550186e-05,
+      "loss": 0.5669,
+      "step": 6000
+    },
+    {
+      "epoch": 88.00011895910781,
+      "eval_accuracy": 0.8296147233158495,
+      "eval_loss": 1.494034767150879,
+      "eval_macro_f1": 0.58373317591399,
+      "eval_runtime": 47.863,
+      "eval_samples_per_second": 14.04,
+      "eval_steps_per_second": 3.51,
+      "step": 6000
+    },
+    {
+      "epoch": 88.00049070631971,
+      "grad_norm": 2.271465301513672,
+      "learning_rate": 8.996282527881041e-05,
+      "loss": 0.5878,
+      "step": 6050
+    },
+    {
+      "epoch": 89.00035687732343,
+      "grad_norm": 4.07094669342041,
+      "learning_rate": 9.070631970260224e-05,
+      "loss": 0.5583,
+      "step": 6100
+    },
+    {
+      "epoch": 89.00035687732343,
+      "eval_accuracy": 0.8304963970843622,
+      "eval_loss": 1.4894771575927734,
+      "eval_macro_f1": 0.596375277813928,
+      "eval_runtime": 47.7329,
+      "eval_samples_per_second": 14.078,
+      "eval_steps_per_second": 3.52,
+      "step": 6100
+    },
+    {
+      "epoch": 90.00022304832714,
+      "grad_norm": 3.257551670074463,
+      "learning_rate": 9.144981412639405e-05,
+      "loss": 0.5567,
+      "step": 6150
+    },
+    {
+      "epoch": 91.00008921933086,
+      "grad_norm": 2.342348575592041,
+      "learning_rate": 9.219330855018588e-05,
+      "loss": 0.5479,
+      "step": 6200
+    },
+    {
+      "epoch": 91.00008921933086,
+      "eval_accuracy": 0.8359645780191465,
+      "eval_loss": 1.5507783889770508,
+      "eval_macro_f1": 0.5961247601926936,
+      "eval_runtime": 47.1119,
+      "eval_samples_per_second": 14.264,
+      "eval_steps_per_second": 3.566,
+      "step": 6200
+    },
+    {
+      "epoch": 91.00046096654275,
+      "grad_norm": 4.675436496734619,
+      "learning_rate": 9.29368029739777e-05,
+      "loss": 0.5373,
+      "step": 6250
+    },
+    {
+      "epoch": 92.00032713754646,
+      "grad_norm": 2.5292415618896484,
+      "learning_rate": 9.368029739776952e-05,
+      "loss": 0.5385,
+      "step": 6300
+    },
+    {
+      "epoch": 92.00032713754646,
+      "eval_accuracy": 0.8331121322684258,
+      "eval_loss": 1.4987480640411377,
+      "eval_macro_f1": 0.6056008732625726,
+      "eval_runtime": 47.6126,
+      "eval_samples_per_second": 14.114,
+      "eval_steps_per_second": 3.528,
+      "step": 6300
+    },
+    {
+      "epoch": 93.00019330855018,
+      "grad_norm": 4.040349006652832,
+      "learning_rate": 9.442379182156135e-05,
+      "loss": 0.5444,
+      "step": 6350
+    },
+    {
+      "epoch": 94.0000594795539,
+      "grad_norm": 4.418771266937256,
+      "learning_rate": 9.516728624535316e-05,
+      "loss": 0.5191,
+      "step": 6400
+    },
+    {
+      "epoch": 94.0000594795539,
+      "eval_accuracy": 0.8352555949405721,
+      "eval_loss": 1.5468343496322632,
+      "eval_macro_f1": 0.6039138175393013,
+      "eval_runtime": 48.0596,
+      "eval_samples_per_second": 13.983,
+      "eval_steps_per_second": 3.496,
+      "step": 6400
+    },
+    {
+      "epoch": 94.0004312267658,
+      "grad_norm": 3.186802864074707,
+      "learning_rate": 9.591078066914498e-05,
+      "loss": 0.5344,
+      "step": 6450
+    },
+    {
+      "epoch": 95.00029739776951,
+      "grad_norm": 3.07525372505188,
+      "learning_rate": 9.66542750929368e-05,
+      "loss": 0.5123,
+      "step": 6500
+    },
+    {
+      "epoch": 95.00029739776951,
+      "eval_accuracy": 0.8292436626419579,
+      "eval_loss": 1.5689940452575684,
+      "eval_macro_f1": 0.6086947888526043,
+      "eval_runtime": 47.3771,
+      "eval_samples_per_second": 14.184,
+      "eval_steps_per_second": 3.546,
+      "step": 6500
+    },
+    {
+      "epoch": 96.00016356877323,
+      "grad_norm": 4.3406243324279785,
+      "learning_rate": 9.739776951672863e-05,
+      "loss": 0.5028,
+      "step": 6550
+    },
+    {
+      "epoch": 97.00002973977695,
+      "grad_norm": 4.482454299926758,
+      "learning_rate": 9.814126394052046e-05,
+      "loss": 0.531,
+      "step": 6600
+    },
+    {
+      "epoch": 97.00002973977695,
+      "eval_accuracy": 0.8345079898078186,
+      "eval_loss": 1.422137975692749,
+      "eval_macro_f1": 0.6109642448055717,
+      "eval_runtime": 47.3199,
+      "eval_samples_per_second": 14.201,
+      "eval_steps_per_second": 3.55,
+      "step": 6600
+    },
+    {
+      "epoch": 97.00040148698885,
+      "grad_norm": 2.9254748821258545,
+      "learning_rate": 9.888475836431226e-05,
+      "loss": 0.4891,
+      "step": 6650
+    },
+    {
+      "epoch": 98.00026765799257,
+      "grad_norm": 2.5848145484924316,
+      "learning_rate": 9.962825278810409e-05,
+      "loss": 0.5194,
+      "step": 6700
+    },
+    {
+      "epoch": 98.00026765799257,
+      "eval_accuracy": 0.8365044824717158,
+      "eval_loss": 1.4362893104553223,
+      "eval_macro_f1": 0.6104741259771145,
+      "eval_runtime": 47.8303,
+      "eval_samples_per_second": 14.05,
+      "eval_steps_per_second": 3.512,
+      "step": 6700
+    },
+    {
+      "epoch": 99.00013382899628,
+      "grad_norm": 2.0179648399353027,
+      "learning_rate": 9.998043435726863e-05,
+      "loss": 0.4938,
+      "step": 6750
+    },
+    {
+      "epoch": 99.00050557620818,
+      "grad_norm": 3.029017210006714,
+      "learning_rate": 9.994130307180592e-05,
+      "loss": 0.4885,
+      "step": 6800
+    },
+    {
+      "epoch": 99.00050557620818,
+      "eval_accuracy": 0.8297226753963072,
+      "eval_loss": 1.5579314231872559,
+      "eval_macro_f1": 0.6109209923521035,
+      "eval_runtime": 47.6341,
+      "eval_samples_per_second": 14.108,
+      "eval_steps_per_second": 3.527,
+      "step": 6800
+    },
+    {
+      "epoch": 100.0003717472119,
+      "grad_norm": 10.595804214477539,
+      "learning_rate": 9.990217178634319e-05,
+      "loss": 0.4856,
+      "step": 6850
+    },
+    {
+      "epoch": 101.00023791821562,
+      "grad_norm": 3.8170125484466553,
+      "learning_rate": 9.986304050088046e-05,
+      "loss": 0.4672,
+      "step": 6900
+    },
+    {
+      "epoch": 101.00023791821562,
+      "eval_accuracy": 0.8405256629197011,
+      "eval_loss": 1.5172147750854492,
+      "eval_macro_f1": 0.6117042873240769,
+      "eval_runtime": 47.3002,
+      "eval_samples_per_second": 14.207,
+      "eval_steps_per_second": 3.552,
+      "step": 6900
+    },
+    {
+      "epoch": 102.00010408921933,
+      "grad_norm": 2.297795295715332,
+      "learning_rate": 9.982390921541773e-05,
+      "loss": 0.4861,
+      "step": 6950
+    },
+    {
+      "epoch": 102.00047583643122,
+      "grad_norm": 2.0339696407318115,
+      "learning_rate": 9.978477792995501e-05,
+      "loss": 0.4791,
+      "step": 7000
+    },
+    {
+      "epoch": 102.00047583643122,
+      "eval_accuracy": 0.8386383728766971,
+      "eval_loss": 1.5465196371078491,
+      "eval_macro_f1": 0.6121669524563613,
+      "eval_runtime": 47.7339,
+      "eval_samples_per_second": 14.078,
+      "eval_steps_per_second": 3.52,
+      "step": 7000
+    },
+    {
+      "epoch": 103.00034200743494,
+      "grad_norm": 1.891178846359253,
+      "learning_rate": 9.974564664449228e-05,
+      "loss": 0.459,
+      "step": 7050
+    },
+    {
+      "epoch": 104.00020817843865,
+      "grad_norm": 2.3269009590148926,
+      "learning_rate": 9.970651535902955e-05,
+      "loss": 0.4622,
+      "step": 7100
+    },
+    {
+      "epoch": 104.00020817843865,
+      "eval_accuracy": 0.8428495839253495,
+      "eval_loss": 1.5032823085784912,
+      "eval_macro_f1": 0.6198375498118172,
+      "eval_runtime": 47.9251,
+      "eval_samples_per_second": 14.022,
+      "eval_steps_per_second": 3.505,
+      "step": 7100
+    },
+    {
+      "epoch": 105.00007434944239,
+      "grad_norm": 2.6395773887634277,
+      "learning_rate": 9.966738407356682e-05,
+      "loss": 0.4773,
+      "step": 7150
+    },
+    {
+      "epoch": 105.00044609665427,
+      "grad_norm": 2.4456825256347656,
+      "learning_rate": 9.962825278810409e-05,
+      "loss": 0.4392,
+      "step": 7200
+    },
+    {
+      "epoch": 105.00044609665427,
+      "eval_accuracy": 0.8425608684033172,
+      "eval_loss": 1.506137490272522,
+      "eval_macro_f1": 0.620540991552123,
+      "eval_runtime": 48.1909,
+      "eval_samples_per_second": 13.945,
+      "eval_steps_per_second": 3.486,
+      "step": 7200
+    },
+    {
+      "epoch": 106.00031226765799,
+      "grad_norm": 2.2125296592712402,
+      "learning_rate": 9.958912150264137e-05,
+      "loss": 0.4427,
+      "step": 7250
+    },
+    {
+      "epoch": 107.0001784386617,
+      "grad_norm": 2.6064858436584473,
+      "learning_rate": 9.954999021717863e-05,
+      "loss": 0.4486,
+      "step": 7300
+    },
+    {
+      "epoch": 107.0001784386617,
+      "eval_accuracy": 0.8388894429121909,
+      "eval_loss": 1.5635102987289429,
+      "eval_macro_f1": 0.6163644154221748,
+      "eval_runtime": 48.2598,
+      "eval_samples_per_second": 13.925,
+      "eval_steps_per_second": 3.481,
+      "step": 7300
+    },
+    {
+      "epoch": 108.00004460966542,
+      "grad_norm": 1.7639355659484863,
+      "learning_rate": 9.951085893171591e-05,
+      "loss": 0.434,
+      "step": 7350
+    },
+    {
+      "epoch": 108.00041635687732,
+      "grad_norm": 1.9860466718673706,
+      "learning_rate": 9.947172764625318e-05,
+      "loss": 0.4339,
+      "step": 7400
+    },
+    {
+      "epoch": 108.00041635687732,
+      "eval_accuracy": 0.8519226590782587,
+      "eval_loss": 1.425977110862732,
+      "eval_macro_f1": 0.6347526514291026,
+      "eval_runtime": 48.0295,
+      "eval_samples_per_second": 13.991,
+      "eval_steps_per_second": 3.498,
+      "step": 7400
+    },
+    {
+      "epoch": 109.00028252788104,
+      "grad_norm": 7.6796441078186035,
+      "learning_rate": 9.943259636079045e-05,
+      "loss": 0.4346,
+      "step": 7450
+    },
+    {
+      "epoch": 110.00014869888476,
+      "grad_norm": 2.4183762073516846,
+      "learning_rate": 9.939346507532772e-05,
+      "loss": 0.4256,
+      "step": 7500
+    },
+    {
+      "epoch": 110.00014869888476,
+      "eval_accuracy": 0.844811035558245,
+      "eval_loss": 1.6024999618530273,
+      "eval_macro_f1": 0.6297670917886972,
+      "eval_runtime": 47.985,
+      "eval_samples_per_second": 14.004,
+      "eval_steps_per_second": 3.501,
+      "step": 7500
+    },
+    {
+      "epoch": 111.00001486988847,
+      "grad_norm": 3.8100974559783936,
+      "learning_rate": 9.935433378986501e-05,
+      "loss": 0.4296,
+      "step": 7550
+    },
+    {
+      "epoch": 111.00038661710038,
+      "grad_norm": 1.486348032951355,
+      "learning_rate": 9.931520250440226e-05,
+      "loss": 0.4162,
+      "step": 7600
+    },
+    {
+      "epoch": 111.00038661710038,
+      "eval_accuracy": 0.8440565173857372,
+      "eval_loss": 1.533053994178772,
+      "eval_macro_f1": 0.6327355016288086,
+      "eval_runtime": 48.0302,
+      "eval_samples_per_second": 13.991,
+      "eval_steps_per_second": 3.498,
+      "step": 7600
+    },
+    {
+      "epoch": 112.00025278810409,
+      "grad_norm": 3.1359190940856934,
+      "learning_rate": 9.927607121893955e-05,
+      "loss": 0.4187,
+      "step": 7650
+    },
+    {
+      "epoch": 113.00011895910781,
+      "grad_norm": 1.8909200429916382,
+      "learning_rate": 9.923693993347682e-05,
+      "loss": 0.413,
+      "step": 7700
+    },
+    {
+      "epoch": 113.00011895910781,
+      "eval_accuracy": 0.8493778775360123,
+      "eval_loss": 1.4858617782592773,
+      "eval_macro_f1": 0.6237896813023426,
+      "eval_runtime": 48.1079,
+      "eval_samples_per_second": 13.969,
+      "eval_steps_per_second": 3.492,
+      "step": 7700
+    },
+    {
+      "epoch": 113.00049070631971,
+      "grad_norm": 2.108914852142334,
+      "learning_rate": 9.919780864801409e-05,
+      "loss": 0.4126,
+      "step": 7750
+    },
+    {
+      "epoch": 114.00035687732343,
+      "grad_norm": 2.574831247329712,
+      "learning_rate": 9.915867736255137e-05,
+      "loss": 0.4146,
+      "step": 7800
+    },
+    {
+      "epoch": 114.00035687732343,
+      "eval_accuracy": 0.8482534627998802,
+      "eval_loss": 1.578758955001831,
+      "eval_macro_f1": 0.632708327751029,
+      "eval_runtime": 47.495,
+      "eval_samples_per_second": 14.149,
+      "eval_steps_per_second": 3.537,
+      "step": 7800
+    },
+    {
+      "epoch": 115.00022304832714,
+      "grad_norm": 2.461167335510254,
+      "learning_rate": 9.911954607708863e-05,
+      "loss": 0.4003,
+      "step": 7850
+    },
+    {
+      "epoch": 116.00008921933086,
+      "grad_norm": 2.769954204559326,
+      "learning_rate": 9.908041479162591e-05,
+      "loss": 0.4176,
+      "step": 7900
+    },
+    {
+      "epoch": 116.00008921933086,
+      "eval_accuracy": 0.8472196779845549,
+      "eval_loss": 1.5794305801391602,
+      "eval_macro_f1": 0.6313642185620171,
+      "eval_runtime": 47.7317,
+      "eval_samples_per_second": 14.079,
+      "eval_steps_per_second": 3.52,
+      "step": 7900
+    },
+    {
+      "epoch": 116.00046096654275,
+      "grad_norm": 1.5411442518234253,
+      "learning_rate": 9.904128350616318e-05,
+      "loss": 0.3905,
+      "step": 7950
+    },
+    {
+      "epoch": 117.00032713754646,
+      "grad_norm": 2.8972113132476807,
+      "learning_rate": 9.900215222070045e-05,
+      "loss": 0.4125,
+      "step": 8000
+    },
+    {
+      "epoch": 117.00032713754646,
+      "eval_accuracy": 0.847543988494726,
+      "eval_loss": 1.4954912662506104,
+      "eval_macro_f1": 0.6424840275505329,
+      "eval_runtime": 47.4476,
+      "eval_samples_per_second": 14.163,
+      "eval_steps_per_second": 3.541,
+      "step": 8000
+    },
+    {
+      "epoch": 118.00019330855018,
+      "grad_norm": 2.530663013458252,
+      "learning_rate": 9.896302093523772e-05,
+      "loss": 0.3915,
+      "step": 8050
+    },
+    {
+      "epoch": 119.0000594795539,
+      "grad_norm": 2.1395466327667236,
+      "learning_rate": 9.8923889649775e-05,
+      "loss": 0.389,
+      "step": 8100
+    },
+    {
+      "epoch": 119.0000594795539,
+      "eval_accuracy": 0.8531609291900277,
+      "eval_loss": 1.4969804286956787,
+      "eval_macro_f1": 0.640615164107896,
+      "eval_runtime": 47.2643,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 3.554,
+      "step": 8100
+    },
+    {
+      "epoch": 119.0004312267658,
+      "grad_norm": 8.651265144348145,
+      "learning_rate": 9.888475836431226e-05,
+      "loss": 0.391,
+      "step": 8150
+    },
+    {
+      "epoch": 120.00029739776951,
+      "grad_norm": 2.033094882965088,
+      "learning_rate": 9.884562707884955e-05,
+      "loss": 0.3935,
+      "step": 8200
+    },
+    {
+      "epoch": 120.00029739776951,
+      "eval_accuracy": 0.8519182980473882,
+      "eval_loss": 1.4476945400238037,
+      "eval_macro_f1": 0.6439031700059592,
+      "eval_runtime": 47.5822,
+      "eval_samples_per_second": 14.123,
+      "eval_steps_per_second": 3.531,
+      "step": 8200
+    },
+    {
+      "epoch": 121.00016356877323,
+      "grad_norm": 1.3807177543640137,
+      "learning_rate": 9.880649579338682e-05,
+      "loss": 0.3868,
+      "step": 8250
+    },
+    {
+      "epoch": 122.00002973977695,
+      "grad_norm": 1.9018504619598389,
+      "learning_rate": 9.876736450792409e-05,
+      "loss": 0.3701,
+      "step": 8300
+    },
+    {
+      "epoch": 122.00002973977695,
+      "eval_accuracy": 0.8542941516114884,
+      "eval_loss": 1.5301586389541626,
+      "eval_macro_f1": 0.6438060495518036,
+      "eval_runtime": 47.3252,
+      "eval_samples_per_second": 14.2,
+      "eval_steps_per_second": 3.55,
+      "step": 8300
+    },
+    {
+      "epoch": 122.00040148698885,
+      "grad_norm": 2.0158724784851074,
+      "learning_rate": 9.872823322246137e-05,
+      "loss": 0.3728,
+      "step": 8350
+    },
+    {
+      "epoch": 123.00026765799257,
+      "grad_norm": 1.53249990940094,
+      "learning_rate": 9.868910193699864e-05,
+      "loss": 0.3757,
+      "step": 8400
+    },
+    {
+      "epoch": 123.00026765799257,
+      "eval_accuracy": 0.8517502815811042,
+      "eval_loss": 1.489638328552246,
+      "eval_macro_f1": 0.6472024615679871,
+      "eval_runtime": 47.12,
+      "eval_samples_per_second": 14.261,
+      "eval_steps_per_second": 3.565,
+      "step": 8400
+    },
+    {
+      "epoch": 124.00013382899628,
+      "grad_norm": 1.1672303676605225,
+      "learning_rate": 9.864997065153591e-05,
+      "loss": 0.37,
+      "step": 8450
+    },
+    {
+      "epoch": 124.00050557620818,
+      "grad_norm": 1.6479486227035522,
+      "learning_rate": 9.861083936607318e-05,
+      "loss": 0.379,
+      "step": 8500
+    },
+    {
+      "epoch": 124.00050557620818,
+      "eval_accuracy": 0.8549860939703802,
+      "eval_loss": 1.4589107036590576,
+      "eval_macro_f1": 0.6522917447958335,
+      "eval_runtime": 47.0089,
+      "eval_samples_per_second": 14.295,
+      "eval_steps_per_second": 3.574,
+      "step": 8500
+    },
+    {
+      "epoch": 125.0003717472119,
+      "grad_norm": 2.498363971710205,
+      "learning_rate": 9.857170808061045e-05,
+      "loss": 0.3714,
+      "step": 8550
+    },
+    {
+      "epoch": 126.00023791821562,
+      "grad_norm": 1.800485610961914,
+      "learning_rate": 9.853257679514772e-05,
+      "loss": 0.3589,
+      "step": 8600
+    },
+    {
+      "epoch": 126.00023791821562,
+      "eval_accuracy": 0.8579483635541026,
+      "eval_loss": 1.5368930101394653,
+      "eval_macro_f1": 0.6484518382038933,
+      "eval_runtime": 47.2042,
+      "eval_samples_per_second": 14.236,
+      "eval_steps_per_second": 3.559,
+      "step": 8600
+    },
+    {
+      "epoch": 127.00010408921933,
+      "grad_norm": 2.102792739868164,
+      "learning_rate": 9.8493445509685e-05,
+      "loss": 0.3632,
+      "step": 8650
+    },
+    {
+      "epoch": 127.00047583643122,
+      "grad_norm": 1.4719001054763794,
+      "learning_rate": 9.845431422422226e-05,
+      "loss": 0.3608,
+      "step": 8700
+    },
+    {
+      "epoch": 127.00047583643122,
+      "eval_accuracy": 0.8541380281972378,
+      "eval_loss": 1.4789085388183594,
+      "eval_macro_f1": 0.6526740777626168,
+      "eval_runtime": 47.3365,
+      "eval_samples_per_second": 14.196,
+      "eval_steps_per_second": 3.549,
+      "step": 8700
+    },
+    {
+      "epoch": 128.00034200743494,
+      "grad_norm": 1.1043628454208374,
+      "learning_rate": 9.841518293875954e-05,
+      "loss": 0.3552,
+      "step": 8750
+    },
+    {
+      "epoch": 129.00020817843867,
+      "grad_norm": 1.163142442703247,
+      "learning_rate": 9.837605165329681e-05,
+      "loss": 0.3575,
+      "step": 8800
+    },
+    {
+      "epoch": 129.00020817843867,
+      "eval_accuracy": 0.8578308939153396,
+      "eval_loss": 1.567258358001709,
+      "eval_macro_f1": 0.6559202691799706,
+      "eval_runtime": 47.2359,
+      "eval_samples_per_second": 14.226,
+      "eval_steps_per_second": 3.557,
+      "step": 8800
+    },
+    {
+      "epoch": 130.00007434944237,
+      "grad_norm": 2.237764596939087,
+      "learning_rate": 9.833692036783408e-05,
+      "loss": 0.3559,
+      "step": 8850
+    },
+    {
+      "epoch": 130.0004460966543,
+      "grad_norm": 2.362916946411133,
+      "learning_rate": 9.829778908237135e-05,
+      "loss": 0.353,
+      "step": 8900
+    },
+    {
+      "epoch": 130.0004460966543,
+      "eval_accuracy": 0.8561993702494404,
+      "eval_loss": 1.4892395734786987,
+      "eval_macro_f1": 0.6546700309203295,
+      "eval_runtime": 47.1656,
+      "eval_samples_per_second": 14.248,
+      "eval_steps_per_second": 3.562,
+      "step": 8900
+    },
+    {
+      "epoch": 131.000312267658,
+      "grad_norm": 1.1094646453857422,
+      "learning_rate": 9.825865779690864e-05,
+      "loss": 0.3503,
+      "step": 8950
+    },
+    {
+      "epoch": 132.00017843866172,
+      "grad_norm": 0.9935618042945862,
+      "learning_rate": 9.821952651144591e-05,
+      "loss": 0.345,
+      "step": 9000
+    },
+    {
+      "epoch": 132.00017843866172,
+      "eval_accuracy": 0.8525096438318296,
+      "eval_loss": 1.518179178237915,
+      "eval_macro_f1": 0.6513330111299382,
+      "eval_runtime": 50.1526,
+      "eval_samples_per_second": 13.399,
+      "eval_steps_per_second": 3.35,
+      "step": 9000
+    },
+    {
+      "epoch": 133.00004460966542,
+      "grad_norm": 1.5325652360916138,
+      "learning_rate": 9.818039522598318e-05,
+      "loss": 0.3518,
+      "step": 9050
+    },
+    {
+      "epoch": 133.0004163568773,
+      "grad_norm": 1.7667288780212402,
+      "learning_rate": 9.814126394052046e-05,
+      "loss": 0.3454,
+      "step": 9100
+    },
+    {
+      "epoch": 133.0004163568773,
+      "eval_accuracy": 0.856074093035065,
+      "eval_loss": 1.4917337894439697,
+      "eval_macro_f1": 0.6523897388323083,
+      "eval_runtime": 50.7576,
+      "eval_samples_per_second": 13.239,
+      "eval_steps_per_second": 3.31,
+      "step": 9100
+    },
+    {
+      "epoch": 134.00028252788104,
+      "grad_norm": 0.9201573729515076,
+      "learning_rate": 9.810213265505772e-05,
+      "loss": 0.3505,
+      "step": 9150
+    },
+    {
+      "epoch": 135.00014869888477,
+      "grad_norm": 1.5801942348480225,
+      "learning_rate": 9.8063001369595e-05,
+      "loss": 0.3398,
+      "step": 9200
+    },
+    {
+      "epoch": 135.00014869888477,
+      "eval_accuracy": 0.8579610019182543,
+      "eval_loss": 1.500423789024353,
+      "eval_macro_f1": 0.6573863845637962,
+      "eval_runtime": 50.47,
+      "eval_samples_per_second": 13.315,
+      "eval_steps_per_second": 3.329,
+      "step": 9200
+    },
+    {
+      "epoch": 136.00001486988847,
+      "grad_norm": 1.5136662721633911,
+      "learning_rate": 9.802387008413226e-05,
+      "loss": 0.3379,
+      "step": 9250
+    },
+    {
+      "epoch": 136.00038661710036,
+      "grad_norm": 2.073296070098877,
+      "learning_rate": 9.798473879866954e-05,
+      "loss": 0.3424,
+      "step": 9300
+    },
+    {
+      "epoch": 136.00038661710036,
+      "eval_accuracy": 0.8589703018470214,
+      "eval_loss": 1.500227928161621,
+      "eval_macro_f1": 0.6582180779919148,
+      "eval_runtime": 50.2017,
+      "eval_samples_per_second": 13.386,
+      "eval_steps_per_second": 3.346,
+      "step": 9300
+    },
+    {
+      "epoch": 137.0002527881041,
+      "grad_norm": 2.330988883972168,
+      "learning_rate": 9.794560751320681e-05,
+      "loss": 0.3388,
+      "step": 9350
+    },
+    {
+      "epoch": 138.0001189591078,
+      "grad_norm": 1.1455012559890747,
+      "learning_rate": 9.790647622774408e-05,
+      "loss": 0.3444,
+      "step": 9400
+    },
+    {
+      "epoch": 138.0001189591078,
+      "eval_accuracy": 0.8580753729006343,
+      "eval_loss": 1.5448076725006104,
+      "eval_macro_f1": 0.659930109636158,
+      "eval_runtime": 49.8683,
+      "eval_samples_per_second": 13.475,
+      "eval_steps_per_second": 3.369,
+      "step": 9400
+    },
+    {
+      "epoch": 138.0004907063197,
+      "grad_norm": 2.7329256534576416,
+      "learning_rate": 9.786734494228135e-05,
+      "loss": 0.3312,
+      "step": 9450
+    },
+    {
+      "epoch": 139.0003568773234,
+      "grad_norm": 1.2961472272872925,
+      "learning_rate": 9.782821365681864e-05,
+      "loss": 0.3344,
+      "step": 9500
+    },
+    {
+      "epoch": 139.0003568773234,
+      "eval_accuracy": 0.8541455048920026,
+      "eval_loss": 1.523103952407837,
+      "eval_macro_f1": 0.6645437069408668,
+      "eval_runtime": 48.9673,
+      "eval_samples_per_second": 13.723,
+      "eval_steps_per_second": 3.431,
+      "step": 9500
+    },
+    {
+      "epoch": 140.00022304832714,
+      "grad_norm": 1.2091388702392578,
+      "learning_rate": 9.77890823713559e-05,
+      "loss": 0.3396,
+      "step": 9550
+    },
+    {
+      "epoch": 141.00008921933085,
+      "grad_norm": 1.0885772705078125,
+      "learning_rate": 9.774995108589318e-05,
+      "loss": 0.3341,
+      "step": 9600
+    },
+    {
+      "epoch": 141.00008921933085,
+      "eval_accuracy": 0.859411708208793,
+      "eval_loss": 1.5716527700424194,
+      "eval_macro_f1": 0.6649711044669897,
+      "eval_runtime": 48.314,
+      "eval_samples_per_second": 13.909,
+      "eval_steps_per_second": 3.477,
+      "step": 9600
+    },
+    {
+      "epoch": 141.00046096654276,
+      "grad_norm": 0.9991358518600464,
+      "learning_rate": 9.771081980043045e-05,
+      "loss": 0.3292,
+      "step": 9650
+    },
+    {
+      "epoch": 142.00032713754646,
+      "grad_norm": 0.9725756049156189,
+      "learning_rate": 9.767168851496772e-05,
+      "loss": 0.3259,
+      "step": 9700
+    },
+    {
+      "epoch": 142.00032713754646,
+      "eval_accuracy": 0.8582671230558392,
+      "eval_loss": 1.5597984790802002,
+      "eval_macro_f1": 0.66374123938933,
+      "eval_runtime": 48.6249,
+      "eval_samples_per_second": 13.82,
+      "eval_steps_per_second": 3.455,
+      "step": 9700
+    },
+    {
+      "epoch": 143.0001933085502,
+      "grad_norm": 1.9716778993606567,
+      "learning_rate": 9.7632557229505e-05,
+      "loss": 0.3214,
+      "step": 9750
+    },
+    {
+      "epoch": 144.0000594795539,
+      "grad_norm": 1.3841463327407837,
+      "learning_rate": 9.759342594404227e-05,
+      "loss": 0.3243,
+      "step": 9800
+    },
+    {
+      "epoch": 144.0000594795539,
+      "eval_accuracy": 0.8594564950163417,
+      "eval_loss": 1.600425362586975,
+      "eval_macro_f1": 0.6602272364365409,
+      "eval_runtime": 49.0882,
+      "eval_samples_per_second": 13.69,
+      "eval_steps_per_second": 3.422,
+      "step": 9800
+    },
+    {
+      "epoch": 144.0004312267658,
+      "grad_norm": 1.2652596235275269,
+      "learning_rate": 9.755429465857954e-05,
+      "loss": 0.3195,
+      "step": 9850
+    },
+    {
+      "epoch": 145.00029739776951,
+      "grad_norm": 0.9097074866294861,
+      "learning_rate": 9.751516337311681e-05,
+      "loss": 0.3236,
+      "step": 9900
+    },
+    {
+      "epoch": 145.00029739776951,
+      "eval_accuracy": 0.8547185372038479,
+      "eval_loss": 1.5480947494506836,
+      "eval_macro_f1": 0.6692649675278269,
+      "eval_runtime": 48.7778,
+      "eval_samples_per_second": 13.777,
+      "eval_steps_per_second": 3.444,
+      "step": 9900
+    },
+    {
+      "epoch": 146.00016356877325,
+      "grad_norm": 1.3369333744049072,
+      "learning_rate": 9.747603208765408e-05,
+      "loss": 0.3203,
+      "step": 9950
+    },
+    {
+      "epoch": 147.00002973977695,
+      "grad_norm": 0.7002597451210022,
+      "learning_rate": 9.743690080219135e-05,
+      "loss": 0.3212,
+      "step": 10000
+    },
+    {
+      "epoch": 147.00002973977695,
+      "eval_accuracy": 0.8636582594164792,
+      "eval_loss": 1.4696401357650757,
+      "eval_macro_f1": 0.669836854595993,
+      "eval_runtime": 49.8793,
+      "eval_samples_per_second": 13.473,
+      "eval_steps_per_second": 3.368,
+      "step": 10000
+    },
+    {
+      "epoch": 147.00040148698884,
+      "grad_norm": 0.966150164604187,
+      "learning_rate": 9.739776951672863e-05,
+      "loss": 0.314,
+      "step": 10050
+    },
+    {
+      "epoch": 148.00026765799257,
+      "grad_norm": 0.9162842631340027,
+      "learning_rate": 9.735863823126589e-05,
+      "loss": 0.3137,
+      "step": 10100
+    },
+    {
+      "epoch": 148.00026765799257,
+      "eval_accuracy": 0.8518328815305616,
+      "eval_loss": 1.57254958152771,
+      "eval_macro_f1": 0.6745871026457042,
+      "eval_runtime": 48.5751,
+      "eval_samples_per_second": 13.834,
+      "eval_steps_per_second": 3.459,
+      "step": 10100
+    },
+    {
+      "epoch": 149.00013382899627,
+      "grad_norm": 1.1828657388687134,
+      "learning_rate": 9.731950694580317e-05,
+      "loss": 0.3192,
+      "step": 10150
+    },
+    {
+      "epoch": 149.00050557620818,
+      "grad_norm": 1.726320505142212,
+      "learning_rate": 9.728037566034044e-05,
+      "loss": 0.3205,
+      "step": 10200
+    },
+    {
+      "epoch": 149.00050557620818,
+      "eval_accuracy": 0.8620538510229856,
+      "eval_loss": 1.4266021251678467,
+      "eval_macro_f1": 0.6685838494494656,
+      "eval_runtime": 48.1524,
+      "eval_samples_per_second": 13.956,
+      "eval_steps_per_second": 3.489,
+      "step": 10200
+    },
+    {
+      "epoch": 150.0003717472119,
+      "grad_norm": 1.412037968635559,
+      "learning_rate": 9.724124437487771e-05,
+      "loss": 0.3112,
+      "step": 10250
+    },
+    {
+      "epoch": 151.00023791821562,
+      "grad_norm": 1.0832695960998535,
+      "learning_rate": 9.7202113089415e-05,
+      "loss": 0.3102,
+      "step": 10300
+    },
+    {
+      "epoch": 151.00023791821562,
+      "eval_accuracy": 0.8641672940325766,
+      "eval_loss": 1.574267029762268,
+      "eval_macro_f1": 0.6709263512273164,
+      "eval_runtime": 48.5859,
+      "eval_samples_per_second": 13.831,
+      "eval_steps_per_second": 3.458,
+      "step": 10300
+    },
+    {
+      "epoch": 152.00010408921932,
+      "grad_norm": 0.9826604127883911,
+      "learning_rate": 9.716298180395227e-05,
+      "loss": 0.3152,
+      "step": 10350
+    },
+    {
+      "epoch": 152.00047583643124,
+      "grad_norm": 1.4301373958587646,
+      "learning_rate": 9.712385051848954e-05,
+      "loss": 0.3129,
+      "step": 10400
+    },
+    {
+      "epoch": 152.00047583643124,
+      "eval_accuracy": 0.8644154699933324,
+      "eval_loss": 1.5302149057388306,
+      "eval_macro_f1": 0.6709705649680309,
+      "eval_runtime": 48.0303,
+      "eval_samples_per_second": 13.991,
+      "eval_steps_per_second": 3.498,
+      "step": 10400
+    },
+    {
+      "epoch": 153.00034200743494,
+      "grad_norm": 1.2386184930801392,
+      "learning_rate": 9.708471923302681e-05,
+      "loss": 0.3126,
+      "step": 10450
+    },
+    {
+      "epoch": 154.00020817843867,
+      "grad_norm": 1.0621957778930664,
+      "learning_rate": 9.704558794756409e-05,
+      "loss": 0.3094,
+      "step": 10500
+    },
+    {
+      "epoch": 154.00020817843867,
+      "eval_accuracy": 0.8608746636189569,
+      "eval_loss": 1.5115866661071777,
+      "eval_macro_f1": 0.6730138894100086,
+      "eval_runtime": 47.7743,
+      "eval_samples_per_second": 14.066,
+      "eval_steps_per_second": 3.517,
+      "step": 10500
+    },
+    {
+      "epoch": 155.00007434944237,
+      "grad_norm": 1.3560012578964233,
+      "learning_rate": 9.700645666210135e-05,
+      "loss": 0.2996,
+      "step": 10550
+    },
+    {
+      "epoch": 155.0004460966543,
+      "grad_norm": 0.7811614274978638,
+      "learning_rate": 9.696732537663863e-05,
+      "loss": 0.3058,
+      "step": 10600
+    },
+    {
+      "epoch": 155.0004460966543,
+      "eval_accuracy": 0.860774923910523,
+      "eval_loss": 1.5835298299789429,
+      "eval_macro_f1": 0.6707021458940838,
+      "eval_runtime": 47.8347,
+      "eval_samples_per_second": 14.048,
+      "eval_steps_per_second": 3.512,
+      "step": 10600
+    },
+    {
+      "epoch": 156.000312267658,
+      "grad_norm": 1.2949461936950684,
+      "learning_rate": 9.692819409117589e-05,
+      "loss": 0.3079,
+      "step": 10650
+    },
+    {
+      "epoch": 157.00017843866172,
+      "grad_norm": 0.860791027545929,
+      "learning_rate": 9.688906280571317e-05,
+      "loss": 0.3063,
+      "step": 10700
+    },
+    {
+      "epoch": 157.00017843866172,
+      "eval_accuracy": 0.8623498068108009,
+      "eval_loss": 1.6110292673110962,
+      "eval_macro_f1": 0.6705726826881232,
+      "eval_runtime": 48.9902,
+      "eval_samples_per_second": 13.717,
+      "eval_steps_per_second": 3.429,
+      "step": 10700
+    },
+    {
+      "epoch": 158.00004460966542,
+      "grad_norm": 2.020859956741333,
+      "learning_rate": 9.684993152025044e-05,
+      "loss": 0.3018,
+      "step": 10750
+    },
+    {
+      "epoch": 158.0004163568773,
+      "grad_norm": 1.3179705142974854,
+      "learning_rate": 9.681080023478771e-05,
+      "loss": 0.2969,
+      "step": 10800
+    },
+    {
+      "epoch": 158.0004163568773,
+      "eval_accuracy": 0.8589260328139017,
+      "eval_loss": 1.658690333366394,
+      "eval_macro_f1": 0.6674403813053171,
+      "eval_runtime": 47.806,
+      "eval_samples_per_second": 14.057,
+      "eval_steps_per_second": 3.514,
+      "step": 10800
+    },
+    {
+      "epoch": 159.00028252788104,
+      "grad_norm": 1.1617735624313354,
+      "learning_rate": 9.6771668949325e-05,
+      "loss": 0.3034,
+      "step": 10850
+    },
+    {
+      "epoch": 160.00014869888477,
+      "grad_norm": 0.7395443320274353,
+      "learning_rate": 9.673253766386227e-05,
+      "loss": 0.3022,
+      "step": 10900
+    },
+    {
+      "epoch": 160.00014869888477,
+      "eval_accuracy": 0.8604493924135389,
+      "eval_loss": 1.7630372047424316,
+      "eval_macro_f1": 0.6697817878000056,
+      "eval_runtime": 48.5778,
+      "eval_samples_per_second": 13.833,
+      "eval_steps_per_second": 3.458,
+      "step": 10900
+    },
+    {
+      "epoch": 161.00001486988847,
+      "grad_norm": 1.5594608783721924,
+      "learning_rate": 9.669340637839954e-05,
+      "loss": 0.3029,
+      "step": 10950
+    },
+    {
+      "epoch": 161.00038661710036,
+      "grad_norm": 1.3082185983657837,
+      "learning_rate": 9.66542750929368e-05,
+      "loss": 0.2979,
+      "step": 11000
+    },
+    {
+      "epoch": 161.00038661710036,
+      "eval_accuracy": 0.8608462416731367,
+      "eval_loss": 1.5527905225753784,
+      "eval_macro_f1": 0.6727040248898062,
+      "eval_runtime": 47.7197,
+      "eval_samples_per_second": 14.082,
+      "eval_steps_per_second": 3.521,
+      "step": 11000
+    },
+    {
+      "epoch": 162.0002527881041,
+      "grad_norm": 0.7811280488967896,
+      "learning_rate": 9.661514380747409e-05,
+      "loss": 0.3009,
+      "step": 11050
+    },
+    {
+      "epoch": 163.0001189591078,
+      "grad_norm": 1.0631964206695557,
+      "learning_rate": 9.657601252201135e-05,
+      "loss": 0.3002,
+      "step": 11100
+    },
+    {
+      "epoch": 163.0001189591078,
+      "eval_accuracy": 0.8645934390800154,
+      "eval_loss": 1.5464845895767212,
+      "eval_macro_f1": 0.6708201336285708,
+      "eval_runtime": 47.8382,
+      "eval_samples_per_second": 14.047,
+      "eval_steps_per_second": 3.512,
+      "step": 11100
+    },
+    {
+      "epoch": 163.0004907063197,
+      "grad_norm": 2.354179620742798,
+      "learning_rate": 9.653688123654863e-05,
+      "loss": 0.3054,
+      "step": 11150
+    },
+    {
+      "epoch": 164.0003568773234,
+      "grad_norm": 0.7585428953170776,
+      "learning_rate": 9.64977499510859e-05,
+      "loss": 0.2979,
+      "step": 11200
+    },
+    {
+      "epoch": 164.0003568773234,
+      "eval_accuracy": 0.8666105756362361,
+      "eval_loss": 1.6061186790466309,
+      "eval_macro_f1": 0.6768795029988373,
+      "eval_runtime": 47.6396,
+      "eval_samples_per_second": 14.106,
+      "eval_steps_per_second": 3.526,
+      "step": 11200
+    },
+    {
+      "epoch": 165.00022304832714,
+      "grad_norm": 0.6630699634552002,
+      "learning_rate": 9.645861866562317e-05,
+      "loss": 0.29,
+      "step": 11250
+    },
+    {
+      "epoch": 166.00008921933085,
+      "grad_norm": 1.263273000717163,
+      "learning_rate": 9.641948738016044e-05,
+      "loss": 0.2908,
+      "step": 11300
+    },
+    {
+      "epoch": 166.00008921933085,
+      "eval_accuracy": 0.8643172326801727,
+      "eval_loss": 1.6028649806976318,
+      "eval_macro_f1": 0.6737691674037425,
+      "eval_runtime": 48.4294,
+      "eval_samples_per_second": 13.876,
+      "eval_steps_per_second": 3.469,
+      "step": 11300
+    },
+    {
+      "epoch": 166.00046096654276,
+      "grad_norm": 1.1215412616729736,
+      "learning_rate": 9.638035609469771e-05,
+      "loss": 0.2955,
+      "step": 11350
+    },
+    {
+      "epoch": 167.00032713754646,
+      "grad_norm": 0.9751584529876709,
+      "learning_rate": 9.634122480923498e-05,
+      "loss": 0.2912,
+      "step": 11400
+    },
+    {
+      "epoch": 167.00032713754646,
+      "eval_accuracy": 0.867831394314885,
+      "eval_loss": 1.5405371189117432,
+      "eval_macro_f1": 0.6776042669784504,
+      "eval_runtime": 47.677,
+      "eval_samples_per_second": 14.095,
+      "eval_steps_per_second": 3.524,
+      "step": 11400
+    },
+    {
+      "epoch": 168.0001933085502,
+      "grad_norm": 0.5147533416748047,
+      "learning_rate": 9.630209352377226e-05,
+      "loss": 0.3009,
+      "step": 11450
+    },
+    {
+      "epoch": 169.0000594795539,
+      "grad_norm": 0.7095732688903809,
+      "learning_rate": 9.626296223830953e-05,
+      "loss": 0.2898,
+      "step": 11500
+    },
+    {
+      "epoch": 169.0000594795539,
+      "eval_accuracy": 0.8701519012721739,
+      "eval_loss": 1.5441175699234009,
+      "eval_macro_f1": 0.6844898842204484,
+      "eval_runtime": 48.1832,
+      "eval_samples_per_second": 13.947,
+      "eval_steps_per_second": 3.487,
+      "step": 11500
+    },
+    {
+      "epoch": 169.0004312267658,
+      "grad_norm": 0.8257178068161011,
+      "learning_rate": 9.62238309528468e-05,
+      "loss": 0.2899,
+      "step": 11550
+    },
+    {
+      "epoch": 170.00029739776951,
+      "grad_norm": 1.3418561220169067,
+      "learning_rate": 9.618469966738409e-05,
+      "loss": 0.2907,
+      "step": 11600
+    },
+    {
+      "epoch": 170.00029739776951,
+      "eval_accuracy": 0.8683436627463533,
+      "eval_loss": 1.5258067846298218,
+      "eval_macro_f1": 0.6799004153156964,
+      "eval_runtime": 48.3802,
+      "eval_samples_per_second": 13.89,
+      "eval_steps_per_second": 3.472,
+      "step": 11600
+    },
+    {
+      "epoch": 171.00016356877325,
+      "grad_norm": 1.1356240510940552,
+      "learning_rate": 9.614556838192134e-05,
+      "loss": 0.2839,
+      "step": 11650
+    },
+    {
+      "epoch": 172.00002973977695,
+      "grad_norm": 2.4852797985076904,
+      "learning_rate": 9.610643709645863e-05,
+      "loss": 0.2893,
+      "step": 11700
+    },
+    {
+      "epoch": 172.00002973977695,
+      "eval_accuracy": 0.8644724453261277,
+      "eval_loss": 1.5444488525390625,
+      "eval_macro_f1": 0.6779267308558976,
+      "eval_runtime": 47.933,
+      "eval_samples_per_second": 14.02,
+      "eval_steps_per_second": 3.505,
+      "step": 11700
+    },
+    {
+      "epoch": 172.00040148698884,
+      "grad_norm": 0.7937450408935547,
+      "learning_rate": 9.60673058109959e-05,
+      "loss": 0.2875,
+      "step": 11750
+    },
+    {
+      "epoch": 173.00026765799257,
+      "grad_norm": 1.3148819208145142,
+      "learning_rate": 9.602817452553317e-05,
+      "loss": 0.2848,
+      "step": 11800
+    },
+    {
+      "epoch": 173.00026765799257,
+      "eval_accuracy": 0.8637417199497863,
+      "eval_loss": 1.5945580005645752,
+      "eval_macro_f1": 0.6752905458548558,
+      "eval_runtime": 48.2353,
+      "eval_samples_per_second": 13.932,
+      "eval_steps_per_second": 3.483,
+      "step": 11800
+    },
+    {
+      "epoch": 174.00013382899627,
+      "grad_norm": 1.375939965248108,
+      "learning_rate": 9.598904324007044e-05,
+      "loss": 0.2835,
+      "step": 11850
+    },
+    {
+      "epoch": 174.00050557620818,
+      "grad_norm": 1.0491206645965576,
+      "learning_rate": 9.594991195460772e-05,
+      "loss": 0.2864,
+      "step": 11900
+    },
+    {
+      "epoch": 174.00050557620818,
+      "eval_accuracy": 0.8652278114733596,
+      "eval_loss": 1.5291390419006348,
+      "eval_macro_f1": 0.6776906470006278,
+      "eval_runtime": 47.1132,
+      "eval_samples_per_second": 14.264,
+      "eval_steps_per_second": 3.566,
+      "step": 11900
+    },
+    {
+      "epoch": 175.0003717472119,
+      "grad_norm": 1.8452578783035278,
+      "learning_rate": 9.591078066914498e-05,
+      "loss": 0.2885,
+      "step": 11950
+    },
+    {
+      "epoch": 176.00023791821562,
+      "grad_norm": 0.6999953985214233,
+      "learning_rate": 9.587164938368226e-05,
+      "loss": 0.2896,
+      "step": 12000
+    },
+    {
+      "epoch": 176.00023791821562,
+      "eval_accuracy": 0.8645915023347069,
+      "eval_loss": 1.5473123788833618,
+      "eval_macro_f1": 0.6804909340063208,
+      "eval_runtime": 48.4286,
+      "eval_samples_per_second": 13.876,
+      "eval_steps_per_second": 3.469,
+      "step": 12000
+    },
+    {
+      "epoch": 177.00010408921932,
+      "grad_norm": 1.5631098747253418,
+      "learning_rate": 9.583251809821953e-05,
+      "loss": 0.2836,
+      "step": 12050
+    },
+    {
+      "epoch": 177.00047583643124,
+      "grad_norm": 1.2296806573867798,
+      "learning_rate": 9.57933868127568e-05,
+      "loss": 0.2889,
+      "step": 12100
+    },
+    {
+      "epoch": 177.00047583643124,
+      "eval_accuracy": 0.8661434049744006,
+      "eval_loss": 1.6414021253585815,
+      "eval_macro_f1": 0.6792450700800063,
+      "eval_runtime": 48.85,
+      "eval_samples_per_second": 13.756,
+      "eval_steps_per_second": 3.439,
+      "step": 12100
+    },
+    {
+      "epoch": 178.00034200743494,
+      "grad_norm": 0.6046510934829712,
+      "learning_rate": 9.575425552729409e-05,
+      "loss": 0.2849,
+      "step": 12150
+    },
+    {
+      "epoch": 179.00020817843867,
+      "grad_norm": 1.4595069885253906,
+      "learning_rate": 9.571512424183134e-05,
+      "loss": 0.2821,
+      "step": 12200
+    },
+    {
+      "epoch": 179.00020817843867,
+      "eval_accuracy": 0.8648826280775495,
+      "eval_loss": 1.6235281229019165,
+      "eval_macro_f1": 0.6799601201557365,
+      "eval_runtime": 48.1603,
+      "eval_samples_per_second": 13.953,
+      "eval_steps_per_second": 3.488,
+      "step": 12200
+    },
+    {
+      "epoch": 180.00007434944237,
+      "grad_norm": 0.7873971462249756,
+      "learning_rate": 9.567599295636863e-05,
+      "loss": 0.2828,
+      "step": 12250
+    },
+    {
+      "epoch": 180.0004460966543,
+      "grad_norm": 0.6876779794692993,
+      "learning_rate": 9.56368616709059e-05,
+      "loss": 0.2824,
+      "step": 12300
+    },
+    {
+      "epoch": 180.0004460966543,
+      "eval_accuracy": 0.8638875731209613,
+      "eval_loss": 1.5444902181625366,
+      "eval_macro_f1": 0.67770174790511,
+      "eval_runtime": 48.1412,
+      "eval_samples_per_second": 13.959,
+      "eval_steps_per_second": 3.49,
+      "step": 12300
+    },
+    {
+      "epoch": 181.000312267658,
+      "grad_norm": 0.6027410626411438,
+      "learning_rate": 9.559773038544317e-05,
+      "loss": 0.2794,
+      "step": 12350
+    },
+    {
+      "epoch": 182.00017843866172,
+      "grad_norm": 0.8014593720436096,
+      "learning_rate": 9.555859909998044e-05,
+      "loss": 0.2774,
+      "step": 12400
+    },
+    {
+      "epoch": 182.00017843866172,
+      "eval_accuracy": 0.8679265122778245,
+      "eval_loss": 1.644963264465332,
+      "eval_macro_f1": 0.6823268003968317,
+      "eval_runtime": 47.9941,
+      "eval_samples_per_second": 14.002,
+      "eval_steps_per_second": 3.5,
+      "step": 12400
+    },
+    {
+      "epoch": 183.00004460966542,
+      "grad_norm": 0.7129560112953186,
+      "learning_rate": 9.551946781451772e-05,
+      "loss": 0.2824,
+      "step": 12450
+    },
+    {
+      "epoch": 183.0004163568773,
+      "grad_norm": 0.695976197719574,
+      "learning_rate": 9.548033652905498e-05,
+      "loss": 0.2776,
+      "step": 12500
+    },
+    {
+      "epoch": 183.0004163568773,
+      "eval_accuracy": 0.8655064591347685,
+      "eval_loss": 1.7040234804153442,
+      "eval_macro_f1": 0.6769104108696231,
+      "eval_runtime": 47.9023,
+      "eval_samples_per_second": 14.029,
+      "eval_steps_per_second": 3.507,
+      "step": 12500
+    },
+    {
+      "epoch": 184.00028252788104,
+      "grad_norm": 0.7102384567260742,
+      "learning_rate": 9.544120524359226e-05,
+      "loss": 0.2812,
+      "step": 12550
+    },
+    {
+      "epoch": 185.00014869888477,
+      "grad_norm": 0.7373831868171692,
+      "learning_rate": 9.540207395812953e-05,
+      "loss": 0.2801,
+      "step": 12600
+    },
+    {
+      "epoch": 185.00014869888477,
+      "eval_accuracy": 0.86286967303127,
+      "eval_loss": 1.6474841833114624,
+      "eval_macro_f1": 0.6785408190937435,
+      "eval_runtime": 48.1272,
+      "eval_samples_per_second": 13.963,
+      "eval_steps_per_second": 3.491,
+      "step": 12600
+    },
+    {
+      "epoch": 186.00001486988847,
+      "grad_norm": 0.5094011425971985,
+      "learning_rate": 9.53629426726668e-05,
+      "loss": 0.2803,
+      "step": 12650
+    },
+    {
+      "epoch": 186.00038661710036,
+      "grad_norm": 0.6538822650909424,
+      "learning_rate": 9.532381138720407e-05,
+      "loss": 0.2745,
+      "step": 12700
+    },
+    {
+      "epoch": 186.00038661710036,
+      "eval_accuracy": 0.8650452479325981,
+      "eval_loss": 1.5569231510162354,
+      "eval_macro_f1": 0.677627388562232,
+      "eval_runtime": 48.1347,
+      "eval_samples_per_second": 13.961,
+      "eval_steps_per_second": 3.49,
+      "step": 12700
+    },
+    {
+      "epoch": 187.0002527881041,
+      "grad_norm": 0.5215421915054321,
+      "learning_rate": 9.528468010174134e-05,
+      "loss": 0.2805,
+      "step": 12750
+    },
+    {
+      "epoch": 188.0001189591078,
+      "grad_norm": 1.654314637184143,
+      "learning_rate": 9.524554881627862e-05,
+      "loss": 0.2791,
+      "step": 12800
+    },
+    {
+      "epoch": 188.0001189591078,
+      "eval_accuracy": 0.863827199940737,
+      "eval_loss": 1.655357003211975,
+      "eval_macro_f1": 0.6847853385351034,
+      "eval_runtime": 47.6816,
+      "eval_samples_per_second": 14.093,
+      "eval_steps_per_second": 3.523,
+      "step": 12800
+    },
+    {
+      "epoch": 188.0004907063197,
+      "grad_norm": 3.5391457080841064,
+      "learning_rate": 9.52064175308159e-05,
+      "loss": 0.2811,
+      "step": 12850
+    },
+    {
+      "epoch": 189.0003568773234,
+      "grad_norm": 0.7586809992790222,
+      "learning_rate": 9.516728624535316e-05,
+      "loss": 0.2753,
+      "step": 12900
+    },
+    {
+      "epoch": 189.0003568773234,
+      "eval_accuracy": 0.8674030325693027,
+      "eval_loss": 1.5796079635620117,
+      "eval_macro_f1": 0.6843978905909051,
+      "eval_runtime": 47.7157,
+      "eval_samples_per_second": 14.083,
+      "eval_steps_per_second": 3.521,
+      "step": 12900
+    },
+    {
+      "epoch": 190.00022304832714,
+      "grad_norm": 1.0755257606506348,
+      "learning_rate": 9.512815495989043e-05,
+      "loss": 0.2731,
+      "step": 12950
+    },
+    {
+      "epoch": 191.00008921933085,
+      "grad_norm": 0.6228829026222229,
+      "learning_rate": 9.508902367442772e-05,
+      "loss": 0.2753,
+      "step": 13000
+    },
+    {
+      "epoch": 191.00008921933085,
+      "eval_accuracy": 0.8665214023561504,
+      "eval_loss": 1.6560512781143188,
+      "eval_macro_f1": 0.6846732776298716,
+      "eval_runtime": 48.0681,
+      "eval_samples_per_second": 13.98,
+      "eval_steps_per_second": 3.495,
+      "step": 13000
+    },
+    {
+      "epoch": 191.00046096654276,
+      "grad_norm": 0.938022792339325,
+      "learning_rate": 9.504989238896497e-05,
+      "loss": 0.2709,
+      "step": 13050
+    },
+    {
+      "epoch": 192.00032713754646,
+      "grad_norm": 0.7752746939659119,
+      "learning_rate": 9.501076110350226e-05,
+      "loss": 0.2757,
+      "step": 13100
+    },
+    {
+      "epoch": 192.00032713754646,
+      "eval_accuracy": 0.8691043353883263,
+      "eval_loss": 1.572521448135376,
+      "eval_macro_f1": 0.6884560219258116,
+      "eval_runtime": 47.3838,
+      "eval_samples_per_second": 14.182,
+      "eval_steps_per_second": 3.546,
+      "step": 13100
+    },
+    {
+      "epoch": 193.0001933085502,
+      "grad_norm": 0.9790377616882324,
+      "learning_rate": 9.497162981803953e-05,
+      "loss": 0.2732,
+      "step": 13150
+    },
+    {
+      "epoch": 194.0000594795539,
+      "grad_norm": 1.399014949798584,
+      "learning_rate": 9.49324985325768e-05,
+      "loss": 0.2715,
+      "step": 13200
+    },
+    {
+      "epoch": 194.0000594795539,
+      "eval_accuracy": 0.8689936532336818,
+      "eval_loss": 1.643375039100647,
+      "eval_macro_f1": 0.6917083697924713,
+      "eval_runtime": 47.6035,
+      "eval_samples_per_second": 14.117,
+      "eval_steps_per_second": 3.529,
+      "step": 13200
+    },
+    {
+      "epoch": 194.0004312267658,
+      "grad_norm": 0.6019532084465027,
+      "learning_rate": 9.489336724711407e-05,
+      "loss": 0.2742,
+      "step": 13250
+    },
+    {
+      "epoch": 195.00029739776951,
+      "grad_norm": 0.8591573238372803,
+      "learning_rate": 9.485423596165135e-05,
+      "loss": 0.2743,
+      "step": 13300
+    },
+    {
+      "epoch": 195.00029739776951,
+      "eval_accuracy": 0.8696440189637991,
+      "eval_loss": 1.6730353832244873,
+      "eval_macro_f1": 0.6888831292939925,
+      "eval_runtime": 47.1438,
+      "eval_samples_per_second": 14.254,
+      "eval_steps_per_second": 3.564,
+      "step": 13300
+    },
+    {
+      "epoch": 196.00016356877325,
+      "grad_norm": 0.7350879907608032,
+      "learning_rate": 9.481510467618862e-05,
+      "loss": 0.2734,
+      "step": 13350
+    },
+    {
+      "epoch": 197.00002973977695,
+      "grad_norm": 0.5490862727165222,
+      "learning_rate": 9.477597339072589e-05,
+      "loss": 0.2685,
+      "step": 13400
+    },
+    {
+      "epoch": 197.00002973977695,
+      "eval_accuracy": 0.869489575266559,
+      "eval_loss": 1.5646990537643433,
+      "eval_macro_f1": 0.6848350674152234,
+      "eval_runtime": 47.4949,
+      "eval_samples_per_second": 14.149,
+      "eval_steps_per_second": 3.537,
+      "step": 13400
+    },
+    {
+      "epoch": 197.00040148698884,
+      "grad_norm": 0.7999712228775024,
+      "learning_rate": 9.473684210526316e-05,
+      "loss": 0.2687,
+      "step": 13450
+    },
+    {
+      "epoch": 198.00026765799257,
+      "grad_norm": 1.20505690574646,
+      "learning_rate": 9.469771081980043e-05,
+      "loss": 0.2697,
+      "step": 13500
+    },
+    {
+      "epoch": 198.00026765799257,
+      "eval_accuracy": 0.8686437550016163,
+      "eval_loss": 1.5947939157485962,
+      "eval_macro_f1": 0.6880418599230155,
+      "eval_runtime": 47.6512,
+      "eval_samples_per_second": 14.102,
+      "eval_steps_per_second": 3.526,
+      "step": 13500
+    },
+    {
+      "epoch": 199.00013382899627,
+      "grad_norm": 0.6366742849349976,
+      "learning_rate": 9.465857953433772e-05,
+      "loss": 0.2672,
+      "step": 13550
+    },
+    {
+      "epoch": 199.00050557620818,
+      "grad_norm": 34.58500289916992,
+      "learning_rate": 9.461944824887497e-05,
+      "loss": 0.2794,
+      "step": 13600
+    },
+    {
+      "epoch": 199.00050557620818,
+      "eval_accuracy": 0.8672534824226665,
+      "eval_loss": 1.5313513278961182,
+      "eval_macro_f1": 0.6841001571058899,
+      "eval_runtime": 45.9918,
+      "eval_samples_per_second": 14.611,
+      "eval_steps_per_second": 3.653,
+      "step": 13600
+    },
+    {
+      "epoch": 200.0003717472119,
+      "grad_norm": 1.101613163948059,
+      "learning_rate": 9.458031696341226e-05,
+      "loss": 0.2738,
+      "step": 13650
+    },
+    {
+      "epoch": 201.00023791821562,
+      "grad_norm": 0.7829649448394775,
+      "learning_rate": 9.454118567794953e-05,
+      "loss": 0.2706,
+      "step": 13700
+    },
+    {
+      "epoch": 201.00023791821562,
+      "eval_accuracy": 0.8678993615662772,
+      "eval_loss": 1.605455994606018,
+      "eval_macro_f1": 0.6885769863222782,
+      "eval_runtime": 47.403,
+      "eval_samples_per_second": 14.176,
+      "eval_steps_per_second": 3.544,
+      "step": 13700
+    },
+    {
+      "epoch": 202.00010408921932,
+      "grad_norm": 0.5832332968711853,
+      "learning_rate": 9.45020543924868e-05,
+      "loss": 0.2673,
+      "step": 13750
+    },
+    {
+      "epoch": 202.00047583643124,
+      "grad_norm": 0.7719833850860596,
+      "learning_rate": 9.446292310702407e-05,
+      "loss": 0.266,
+      "step": 13800
+    },
+    {
+      "epoch": 202.00047583643124,
+      "eval_accuracy": 0.8691502501234343,
+      "eval_loss": 1.5864771604537964,
+      "eval_macro_f1": 0.6932054967503338,
+      "eval_runtime": 47.7144,
+      "eval_samples_per_second": 14.084,
+      "eval_steps_per_second": 3.521,
+      "step": 13800
+    },
+    {
+      "epoch": 203.00034200743494,
+      "grad_norm": 0.8644407987594604,
+      "learning_rate": 9.442379182156135e-05,
+      "loss": 0.2679,
+      "step": 13850
+    },
+    {
+      "epoch": 204.00020817843867,
+      "grad_norm": 0.7103060483932495,
+      "learning_rate": 9.43846605360986e-05,
+      "loss": 0.2689,
+      "step": 13900
+    },
+    {
+      "epoch": 204.00020817843867,
+      "eval_accuracy": 0.8688131187854229,
+      "eval_loss": 1.6281613111495972,
+      "eval_macro_f1": 0.686904537208796,
+      "eval_runtime": 46.631,
+      "eval_samples_per_second": 14.411,
+      "eval_steps_per_second": 3.603,
+      "step": 13900
+    },
+    {
+      "epoch": 205.00007434944237,
+      "grad_norm": 1.3666083812713623,
+      "learning_rate": 9.434552925063589e-05,
+      "loss": 0.2655,
+      "step": 13950
+    },
+    {
+      "epoch": 205.0004460966543,
+      "grad_norm": 0.6814075708389282,
+      "learning_rate": 9.430639796517316e-05,
+      "loss": 0.2674,
+      "step": 14000
+    },
+    {
+      "epoch": 205.0004460966543,
+      "eval_accuracy": 0.8685227656776213,
+      "eval_loss": 1.5942567586898804,
+      "eval_macro_f1": 0.687446333114446,
+      "eval_runtime": 47.6458,
+      "eval_samples_per_second": 14.104,
+      "eval_steps_per_second": 3.526,
+      "step": 14000
+    },
+    {
+      "epoch": 206.000312267658,
+      "grad_norm": 0.704296886920929,
+      "learning_rate": 9.426726667971043e-05,
+      "loss": 0.2668,
+      "step": 14050
+    },
+    {
+      "epoch": 207.00017843866172,
+      "grad_norm": 0.8048647046089172,
+      "learning_rate": 9.422813539424771e-05,
+      "loss": 0.27,
+      "step": 14100
+    },
+    {
+      "epoch": 207.00017843866172,
+      "eval_accuracy": 0.8677061769198147,
+      "eval_loss": 1.6559548377990723,
+      "eval_macro_f1": 0.6895342478716917,
+      "eval_runtime": 46.3001,
+      "eval_samples_per_second": 14.514,
+      "eval_steps_per_second": 3.629,
+      "step": 14100
+    },
+    {
+      "epoch": 208.00004460966542,
+      "grad_norm": 1.0088293552398682,
+      "learning_rate": 9.418900410878497e-05,
+      "loss": 0.2692,
+      "step": 14150
+    },
+    {
+      "epoch": 208.0004163568773,
+      "grad_norm": 0.6096265912055969,
+      "learning_rate": 9.414987282332225e-05,
+      "loss": 0.264,
+      "step": 14200
+    },
+    {
+      "epoch": 208.0004163568773,
+      "eval_accuracy": 0.8624095986479511,
+      "eval_loss": 1.72756028175354,
+      "eval_macro_f1": 0.6788019065387273,
+      "eval_runtime": 47.1371,
+      "eval_samples_per_second": 14.256,
+      "eval_steps_per_second": 3.564,
+      "step": 14200
+    },
+    {
+      "epoch": 209.00028252788104,
+      "grad_norm": 0.8049959540367126,
+      "learning_rate": 9.411074153785952e-05,
+      "loss": 0.269,
+      "step": 14250
+    },
+    {
+      "epoch": 210.00014869888477,
+      "grad_norm": 0.405304878950119,
+      "learning_rate": 9.40716102523968e-05,
+      "loss": 0.2613,
+      "step": 14300
+    },
+    {
+      "epoch": 210.00014869888477,
+      "eval_accuracy": 0.8737539228529647,
+      "eval_loss": 1.6797696352005005,
+      "eval_macro_f1": 0.68707650553989,
+      "eval_runtime": 47.3163,
+      "eval_samples_per_second": 14.202,
+      "eval_steps_per_second": 3.551,
+      "step": 14300
+    },
+    {
+      "epoch": 211.00001486988847,
+      "grad_norm": 0.8155346512794495,
+      "learning_rate": 9.403247896693406e-05,
+      "loss": 0.2658,
+      "step": 14350
+    },
+    {
+      "epoch": 211.00038661710036,
+      "grad_norm": 1.3639463186264038,
+      "learning_rate": 9.399334768147135e-05,
+      "loss": 0.2623,
+      "step": 14400
+    },
+    {
+      "epoch": 211.00038661710036,
+      "eval_accuracy": 0.8712119696176184,
+      "eval_loss": 1.5848718881607056,
+      "eval_macro_f1": 0.6890734610343825,
+      "eval_runtime": 46.9022,
+      "eval_samples_per_second": 14.328,
+      "eval_steps_per_second": 3.582,
+      "step": 14400
+    },
+    {
+      "epoch": 212.0002527881041,
+      "grad_norm": 2.1215269565582275,
+      "learning_rate": 9.39542163960086e-05,
+      "loss": 0.2632,
+      "step": 14450
+    },
+    {
+      "epoch": 213.0001189591078,
+      "grad_norm": 2.304811477661133,
+      "learning_rate": 9.391508511054589e-05,
+      "loss": 0.2665,
+      "step": 14500
+    },
+    {
+      "epoch": 213.0001189591078,
+      "eval_accuracy": 0.8663827562423165,
+      "eval_loss": 1.670615553855896,
+      "eval_macro_f1": 0.6867730543695441,
+      "eval_runtime": 47.4401,
+      "eval_samples_per_second": 14.165,
+      "eval_steps_per_second": 3.541,
+      "step": 14500
+    },
+    {
+      "epoch": 213.0004907063197,
+      "grad_norm": 0.5944399833679199,
+      "learning_rate": 9.387595382508316e-05,
+      "loss": 0.2623,
+      "step": 14550
+    },
+    {
+      "epoch": 214.0003568773234,
+      "grad_norm": 0.8482255339622498,
+      "learning_rate": 9.383682253962043e-05,
+      "loss": 0.2624,
+      "step": 14600
+    },
+    {
+      "epoch": 214.0003568773234,
+      "eval_accuracy": 0.8698217804945703,
+      "eval_loss": 1.6686782836914062,
+      "eval_macro_f1": 0.6904881778891221,
+      "eval_runtime": 47.1229,
+      "eval_samples_per_second": 14.261,
+      "eval_steps_per_second": 3.565,
+      "step": 14600
+    },
+    {
+      "epoch": 215.00022304832714,
+      "grad_norm": 0.43856558203697205,
+      "learning_rate": 9.379769125415771e-05,
+      "loss": 0.2622,
+      "step": 14650
+    },
+    {
+      "epoch": 216.00008921933085,
+      "grad_norm": 0.8466907739639282,
+      "learning_rate": 9.375855996869498e-05,
+      "loss": 0.2579,
+      "step": 14700
+    },
+    {
+      "epoch": 216.00008921933085,
+      "eval_accuracy": 0.8698276830776899,
+      "eval_loss": 1.5432207584381104,
+      "eval_macro_f1": 0.693393345872542,
+      "eval_runtime": 47.3775,
+      "eval_samples_per_second": 14.184,
+      "eval_steps_per_second": 3.546,
+      "step": 14700
+    },
+    {
+      "epoch": 216.00046096654276,
+      "grad_norm": 1.5707125663757324,
+      "learning_rate": 9.371942868323225e-05,
+      "loss": 0.2628,
+      "step": 14750
+    },
+    {
+      "epoch": 217.00032713754646,
+      "grad_norm": 0.5471277832984924,
+      "learning_rate": 9.368029739776952e-05,
+      "loss": 0.2672,
+      "step": 14800
+    },
+    {
+      "epoch": 217.00032713754646,
+      "eval_accuracy": 0.8655290700380827,
+      "eval_loss": 1.6244316101074219,
+      "eval_macro_f1": 0.6921881782988438,
+      "eval_runtime": 47.6498,
+      "eval_samples_per_second": 14.103,
+      "eval_steps_per_second": 3.526,
+      "step": 14800
+    },
+    {
+      "epoch": 218.0001933085502,
+      "grad_norm": 0.6579570770263672,
+      "learning_rate": 9.364116611230679e-05,
+      "loss": 0.2581,
+      "step": 14850
+    },
+    {
+      "epoch": 219.0000594795539,
+      "grad_norm": 1.0227067470550537,
+      "learning_rate": 9.360203482684406e-05,
+      "loss": 0.2673,
+      "step": 14900
+    },
+    {
+      "epoch": 219.0000594795539,
+      "eval_accuracy": 0.8725138688744463,
+      "eval_loss": 1.5025845766067505,
+      "eval_macro_f1": 0.6950262696385097,
+      "eval_runtime": 47.57,
+      "eval_samples_per_second": 14.127,
+      "eval_steps_per_second": 3.532,
+      "step": 14900
+    },
+    {
+      "epoch": 219.0004312267658,
+      "grad_norm": 0.7601364254951477,
+      "learning_rate": 9.356290354138135e-05,
+      "loss": 0.2646,
+      "step": 14950
+    },
+    {
+      "epoch": 220.00029739776951,
+      "grad_norm": 1.6235822439193726,
+      "learning_rate": 9.35237722559186e-05,
+      "loss": 0.2657,
+      "step": 15000
+    },
+    {
+      "epoch": 220.00029739776951,
+      "eval_accuracy": 0.8726696606327754,
+      "eval_loss": 1.6585569381713867,
+      "eval_macro_f1": 0.6917753475326269,
+      "eval_runtime": 47.7138,
+      "eval_samples_per_second": 14.084,
+      "eval_steps_per_second": 3.521,
+      "step": 15000
+    },
+    {
+      "epoch": 221.00016356877325,
+      "grad_norm": 0.4802096486091614,
+      "learning_rate": 9.348464097045589e-05,
+      "loss": 0.2562,
+      "step": 15050
+    },
+    {
+      "epoch": 222.00002973977695,
+      "grad_norm": 1.0543347597122192,
+      "learning_rate": 9.344550968499316e-05,
+      "loss": 0.2598,
+      "step": 15100
+    },
+    {
+      "epoch": 222.00002973977695,
+      "eval_accuracy": 0.8727395565132572,
+      "eval_loss": 1.6446582078933716,
+      "eval_macro_f1": 0.6932911685254688,
+      "eval_runtime": 47.6711,
+      "eval_samples_per_second": 14.097,
+      "eval_steps_per_second": 3.524,
+      "step": 15100
+    },
+    {
+      "epoch": 222.00040148698884,
+      "grad_norm": 0.8065938949584961,
+      "learning_rate": 9.340637839953043e-05,
+      "loss": 0.2599,
+      "step": 15150
+    },
+    {
+      "epoch": 223.00026765799257,
+      "grad_norm": 0.8375520706176758,
+      "learning_rate": 9.33672471140677e-05,
+      "loss": 0.2621,
+      "step": 15200
+    },
+    {
+      "epoch": 223.00026765799257,
+      "eval_accuracy": 0.8728216475795783,
+      "eval_loss": 1.6157622337341309,
+      "eval_macro_f1": 0.6932958030431495,
+      "eval_runtime": 47.7404,
+      "eval_samples_per_second": 14.076,
+      "eval_steps_per_second": 3.519,
+      "step": 15200
+    },
+    {
+      "epoch": 224.00013382899627,
+      "grad_norm": 0.5836125612258911,
+      "learning_rate": 9.332811582860498e-05,
+      "loss": 0.2575,
+      "step": 15250
+    },
+    {
+      "epoch": 224.00050557620818,
+      "grad_norm": 1.7909495830535889,
+      "learning_rate": 9.328898454314225e-05,
+      "loss": 0.256,
+      "step": 15300
+    },
+    {
+      "epoch": 224.00050557620818,
+      "eval_accuracy": 0.8688527520082964,
+      "eval_loss": 1.7035082578659058,
+      "eval_macro_f1": 0.6917661335156897,
+      "eval_runtime": 46.9397,
+      "eval_samples_per_second": 14.316,
+      "eval_steps_per_second": 3.579,
+      "step": 15300
+    },
+    {
+      "epoch": 225.0003717472119,
+      "grad_norm": 0.9828186631202698,
+      "learning_rate": 9.324985325767952e-05,
+      "loss": 0.2583,
+      "step": 15350
+    },
+    {
+      "epoch": 226.00023791821562,
+      "grad_norm": 0.4985399544239044,
+      "learning_rate": 9.32107219722168e-05,
+      "loss": 0.2578,
+      "step": 15400
+    },
+    {
+      "epoch": 226.00023791821562,
+      "eval_accuracy": 0.8719250847337479,
+      "eval_loss": 1.5340908765792847,
+      "eval_macro_f1": 0.6947860278625747,
+      "eval_runtime": 47.7352,
+      "eval_samples_per_second": 14.078,
+      "eval_steps_per_second": 3.519,
+      "step": 15400
+    },
+    {
+      "epoch": 227.00010408921932,
+      "grad_norm": 0.8708937168121338,
+      "learning_rate": 9.317159068675406e-05,
+      "loss": 0.255,
+      "step": 15450
+    },
+    {
+      "epoch": 227.00047583643124,
+      "grad_norm": 1.0979160070419312,
+      "learning_rate": 9.313245940129134e-05,
+      "loss": 0.2552,
+      "step": 15500
+    },
+    {
+      "epoch": 227.00047583643124,
+      "eval_accuracy": 0.8702558461586529,
+      "eval_loss": 1.6572105884552002,
+      "eval_macro_f1": 0.6917563353825633,
+      "eval_runtime": 47.7184,
+      "eval_samples_per_second": 14.083,
+      "eval_steps_per_second": 3.521,
+      "step": 15500
+    },
+    {
+      "epoch": 228.00034200743494,
+      "grad_norm": 0.43901586532592773,
+      "learning_rate": 9.30933281158286e-05,
+      "loss": 0.2572,
+      "step": 15550
+    },
+    {
+      "epoch": 229.00020817843867,
+      "grad_norm": 0.9299412369728088,
+      "learning_rate": 9.305419683036588e-05,
+      "loss": 0.2551,
+      "step": 15600
+    },
+    {
+      "epoch": 229.00020817843867,
+      "eval_accuracy": 0.8717616878137655,
+      "eval_loss": 1.606767177581787,
+      "eval_macro_f1": 0.6934037548534984,
+      "eval_runtime": 47.3853,
+      "eval_samples_per_second": 14.182,
+      "eval_steps_per_second": 3.545,
+      "step": 15600
+    },
+    {
+      "epoch": 230.00007434944237,
+      "grad_norm": 0.5207256078720093,
+      "learning_rate": 9.301506554490315e-05,
+      "loss": 0.2576,
+      "step": 15650
+    },
+    {
+      "epoch": 230.0004460966543,
+      "grad_norm": 0.6443852782249451,
+      "learning_rate": 9.297593425944042e-05,
+      "loss": 0.263,
+      "step": 15700
+    },
+    {
+      "epoch": 230.0004460966543,
+      "eval_accuracy": 0.8718295420556332,
+      "eval_loss": 1.744031310081482,
+      "eval_macro_f1": 0.6939710912655463,
+      "eval_runtime": 47.6301,
+      "eval_samples_per_second": 14.109,
+      "eval_steps_per_second": 3.527,
+      "step": 15700
+    },
+    {
+      "epoch": 231.000312267658,
+      "grad_norm": 0.6609066724777222,
+      "learning_rate": 9.29368029739777e-05,
+      "loss": 0.2591,
+      "step": 15750
+    },
+    {
+      "epoch": 232.00017843866172,
+      "grad_norm": 0.614456295967102,
+      "learning_rate": 9.289767168851498e-05,
+      "loss": 0.2567,
+      "step": 15800
+    },
+    {
+      "epoch": 232.00017843866172,
+      "eval_accuracy": 0.8708844038546723,
+      "eval_loss": 1.5864840745925903,
+      "eval_macro_f1": 0.6954124817053439,
+      "eval_runtime": 47.5438,
+      "eval_samples_per_second": 14.134,
+      "eval_steps_per_second": 3.534,
+      "step": 15800
+    },
+    {
+      "epoch": 233.00004460966542,
+      "grad_norm": 0.9560025334358215,
+      "learning_rate": 9.285854040305225e-05,
+      "loss": 0.2558,
+      "step": 15850
+    },
+    {
+      "epoch": 233.0004163568773,
+      "grad_norm": 0.9449219703674316,
+      "learning_rate": 9.281940911758952e-05,
+      "loss": 0.2538,
+      "step": 15900
+    },
+    {
+      "epoch": 233.0004163568773,
+      "eval_accuracy": 0.8691010985986889,
+      "eval_loss": 1.6324706077575684,
+      "eval_macro_f1": 0.6932624958186178,
+      "eval_runtime": 47.7799,
+      "eval_samples_per_second": 14.064,
+      "eval_steps_per_second": 3.516,
+      "step": 15900
+    },
+    {
+      "epoch": 234.00028252788104,
+      "grad_norm": 0.6615203022956848,
+      "learning_rate": 9.278027783212679e-05,
+      "loss": 0.2562,
+      "step": 15950
+    },
+    {
+      "epoch": 235.00014869888477,
+      "grad_norm": 0.7397301197052002,
+      "learning_rate": 9.274114654666406e-05,
+      "loss": 0.2534,
+      "step": 16000
+    },
+    {
+      "epoch": 235.00014869888477,
+      "eval_accuracy": 0.8692960057739556,
+      "eval_loss": 1.5722757577896118,
+      "eval_macro_f1": 0.692964101120355,
+      "eval_runtime": 47.7627,
+      "eval_samples_per_second": 14.07,
+      "eval_steps_per_second": 3.517,
+      "step": 16000
+    },
+    {
+      "epoch": 236.00001486988847,
+      "grad_norm": 0.55063796043396,
+      "learning_rate": 9.270201526120134e-05,
+      "loss": 0.2542,
+      "step": 16050
+    },
+    {
+      "epoch": 236.00038661710036,
+      "grad_norm": 0.7421992421150208,
+      "learning_rate": 9.266288397573861e-05,
+      "loss": 0.254,
+      "step": 16100
+    },
+    {
+      "epoch": 236.00038661710036,
+      "eval_accuracy": 0.8703440842902799,
+      "eval_loss": 1.6220040321350098,
+      "eval_macro_f1": 0.6946437608735907,
+      "eval_runtime": 47.4085,
+      "eval_samples_per_second": 14.175,
+      "eval_steps_per_second": 3.544,
+      "step": 16100
+    },
+    {
+      "epoch": 237.0002527881041,
+      "grad_norm": 0.8659176230430603,
+      "learning_rate": 9.262375269027588e-05,
+      "loss": 0.2528,
+      "step": 16150
+    },
+    {
+      "epoch": 238.0001189591078,
+      "grad_norm": 0.7209817171096802,
+      "learning_rate": 9.258462140481315e-05,
+      "loss": 0.2509,
+      "step": 16200
+    },
+    {
+      "epoch": 238.0001189591078,
+      "eval_accuracy": 0.8701781168352621,
+      "eval_loss": 1.7125494480133057,
+      "eval_macro_f1": 0.6885334728232957,
+      "eval_runtime": 47.6343,
+      "eval_samples_per_second": 14.107,
+      "eval_steps_per_second": 3.527,
+      "step": 16200
+    },
+    {
+      "epoch": 238.0004907063197,
+      "grad_norm": 0.5732719898223877,
+      "learning_rate": 9.254549011935042e-05,
+      "loss": 0.2507,
+      "step": 16250
+    },
+    {
+      "epoch": 239.0003568773234,
+      "grad_norm": 0.7329777479171753,
+      "learning_rate": 9.250635883388769e-05,
+      "loss": 0.2513,
+      "step": 16300
+    },
+    {
+      "epoch": 239.0003568773234,
+      "eval_accuracy": 0.8709371142216245,
+      "eval_loss": 1.701737880706787,
+      "eval_macro_f1": 0.6957311110906366,
+      "eval_runtime": 47.2624,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 3.555,
+      "step": 16300
+    },
+    {
+      "epoch": 240.00022304832714,
+      "grad_norm": 0.6779870986938477,
+      "learning_rate": 9.246722754842498e-05,
+      "loss": 0.2555,
+      "step": 16350
+    },
+    {
+      "epoch": 241.00008921933085,
+      "grad_norm": 0.9598796963691711,
+      "learning_rate": 9.242809626296223e-05,
+      "loss": 0.2567,
+      "step": 16400
+    },
+    {
+      "epoch": 241.00008921933085,
+      "eval_accuracy": 0.8710340352265105,
+      "eval_loss": 1.658144235610962,
+      "eval_macro_f1": 0.6995813339320679,
+      "eval_runtime": 47.5149,
+      "eval_samples_per_second": 14.143,
+      "eval_steps_per_second": 3.536,
+      "step": 16400
+    },
+    {
+      "epoch": 241.00046096654276,
+      "grad_norm": 0.5498836040496826,
+      "learning_rate": 9.238896497749952e-05,
+      "loss": 0.2576,
+      "step": 16450
+    },
+    {
+      "epoch": 242.00032713754646,
+      "grad_norm": 1.8321760892868042,
+      "learning_rate": 9.234983369203679e-05,
+      "loss": 0.2579,
+      "step": 16500
+    },
+    {
+      "epoch": 242.00032713754646,
+      "eval_accuracy": 0.8703391367743829,
+      "eval_loss": 1.6368268728256226,
+      "eval_macro_f1": 0.6961760325639348,
+      "eval_runtime": 47.3632,
+      "eval_samples_per_second": 14.188,
+      "eval_steps_per_second": 3.547,
+      "step": 16500
+    },
+    {
+      "epoch": 243.0001933085502,
+      "grad_norm": 0.702711820602417,
+      "learning_rate": 9.231070240657406e-05,
+      "loss": 0.2579,
+      "step": 16550
+    },
+    {
+      "epoch": 244.0000594795539,
+      "grad_norm": 0.6555370092391968,
+      "learning_rate": 9.227157112111134e-05,
+      "loss": 0.2665,
+      "step": 16600
+    },
+    {
+      "epoch": 244.0000594795539,
+      "eval_accuracy": 0.86594513616592,
+      "eval_loss": 1.5685354471206665,
+      "eval_macro_f1": 0.6927762838387244,
+      "eval_runtime": 47.4037,
+      "eval_samples_per_second": 14.176,
+      "eval_steps_per_second": 3.544,
+      "step": 16600
+    },
+    {
+      "epoch": 244.0004312267658,
+      "grad_norm": 1.0492157936096191,
+      "learning_rate": 9.223243983564861e-05,
+      "loss": 0.2569,
+      "step": 16650
+    },
+    {
+      "epoch": 245.00029739776951,
+      "grad_norm": 0.9604139924049377,
+      "learning_rate": 9.219330855018588e-05,
+      "loss": 0.2625,
+      "step": 16700
+    },
+    {
+      "epoch": 245.00029739776951,
+      "eval_accuracy": 0.8723870633974999,
+      "eval_loss": 1.6895173788070679,
+      "eval_macro_f1": 0.6972317831696635,
+      "eval_runtime": 46.9034,
+      "eval_samples_per_second": 14.327,
+      "eval_steps_per_second": 3.582,
+      "step": 16700
+    },
+    {
+      "epoch": 246.00016356877325,
+      "grad_norm": 0.5593128204345703,
+      "learning_rate": 9.215417726472315e-05,
+      "loss": 0.2577,
+      "step": 16750
+    },
+    {
+      "epoch": 247.00002973977695,
+      "grad_norm": 0.554602861404419,
+      "learning_rate": 9.211504597926043e-05,
+      "loss": 0.2539,
+      "step": 16800
+    },
+    {
+      "epoch": 247.00002973977695,
+      "eval_accuracy": 0.8695801039488791,
+      "eval_loss": 1.5884020328521729,
+      "eval_macro_f1": 0.695830481158142,
+      "eval_runtime": 47.3906,
+      "eval_samples_per_second": 14.18,
+      "eval_steps_per_second": 3.545,
+      "step": 16800
+    },
+    {
+      "epoch": 247.00040148698884,
+      "grad_norm": 0.44518184661865234,
+      "learning_rate": 9.207591469379769e-05,
+      "loss": 0.25,
+      "step": 16850
+    },
+    {
+      "epoch": 248.00026765799257,
+      "grad_norm": 0.5679718852043152,
+      "learning_rate": 9.203678340833497e-05,
+      "loss": 0.252,
+      "step": 16900
+    },
+    {
+      "epoch": 248.00026765799257,
+      "eval_accuracy": 0.8730579371173861,
+      "eval_loss": 1.6707909107208252,
+      "eval_macro_f1": 0.6944050366018404,
+      "eval_runtime": 47.3149,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 3.551,
+      "step": 16900
+    },
+    {
+      "epoch": 249.00013382899627,
+      "grad_norm": 0.7690138220787048,
+      "learning_rate": 9.199765212287223e-05,
+      "loss": 0.2525,
+      "step": 16950
+    },
+    {
+      "epoch": 249.00050557620818,
+      "grad_norm": 0.6587095260620117,
+      "learning_rate": 9.195852083740951e-05,
+      "loss": 0.2476,
+      "step": 17000
+    },
+    {
+      "epoch": 249.00050557620818,
+      "eval_accuracy": 0.8728623895233069,
+      "eval_loss": 1.628631353378296,
+      "eval_macro_f1": 0.6983235583063505,
+      "eval_runtime": 46.8372,
+      "eval_samples_per_second": 14.348,
+      "eval_steps_per_second": 3.587,
+      "step": 17000
+    },
+    {
+      "epoch": 250.0003717472119,
+      "grad_norm": 0.5215486288070679,
+      "learning_rate": 9.191938955194678e-05,
+      "loss": 0.2457,
+      "step": 17050
+    },
+    {
+      "epoch": 251.00023791821562,
+      "grad_norm": 0.7685202956199646,
+      "learning_rate": 9.188025826648405e-05,
+      "loss": 0.2486,
+      "step": 17100
+    },
+    {
+      "epoch": 251.00023791821562,
+      "eval_accuracy": 0.875059372092214,
+      "eval_loss": 1.6925468444824219,
+      "eval_macro_f1": 0.6942957860515563,
+      "eval_runtime": 47.6893,
+      "eval_samples_per_second": 14.091,
+      "eval_steps_per_second": 3.523,
+      "step": 17100
+    },
+    {
+      "epoch": 252.00010408921932,
+      "grad_norm": 0.5681723356246948,
+      "learning_rate": 9.184112698102134e-05,
+      "loss": 0.2501,
+      "step": 17150
+    },
+    {
+      "epoch": 252.00047583643124,
+      "grad_norm": 0.766886293888092,
+      "learning_rate": 9.180199569555861e-05,
+      "loss": 0.2503,
+      "step": 17200
+    },
+    {
+      "epoch": 252.00047583643124,
+      "eval_accuracy": 0.8755406525042265,
+      "eval_loss": 1.5906827449798584,
+      "eval_macro_f1": 0.6991550920082847,
+      "eval_runtime": 47.5816,
+      "eval_samples_per_second": 14.123,
+      "eval_steps_per_second": 3.531,
+      "step": 17200
+    },
+    {
+      "epoch": 253.00034200743494,
+      "grad_norm": 1.2093149423599243,
+      "learning_rate": 9.176286441009588e-05,
+      "loss": 0.2485,
+      "step": 17250
+    },
+    {
+      "epoch": 254.00020817843867,
+      "grad_norm": 0.5527441501617432,
+      "learning_rate": 9.172373312463315e-05,
+      "loss": 0.2489,
+      "step": 17300
+    },
+    {
+      "epoch": 254.00020817843867,
+      "eval_accuracy": 0.8714815968328208,
+      "eval_loss": 1.7195326089859009,
+      "eval_macro_f1": 0.6970162423490259,
+      "eval_runtime": 47.4478,
+      "eval_samples_per_second": 14.163,
+      "eval_steps_per_second": 3.541,
+      "step": 17300
+    },
+    {
+      "epoch": 255.00007434944237,
+      "grad_norm": 0.45948362350463867,
+      "learning_rate": 9.168460183917043e-05,
+      "loss": 0.2439,
+      "step": 17350
+    },
+    {
+      "epoch": 255.0004460966543,
+      "grad_norm": 0.6823854446411133,
+      "learning_rate": 9.164547055370769e-05,
+      "loss": 0.2457,
+      "step": 17400
+    },
+    {
+      "epoch": 255.0004460966543,
+      "eval_accuracy": 0.8765321124358997,
+      "eval_loss": 1.636742353439331,
+      "eval_macro_f1": 0.699459918703319,
+      "eval_runtime": 47.7938,
+      "eval_samples_per_second": 14.06,
+      "eval_steps_per_second": 3.515,
+      "step": 17400
+    },
+    {
+      "epoch": 256.000312267658,
+      "grad_norm": 0.5651408433914185,
+      "learning_rate": 9.160633926824497e-05,
+      "loss": 0.2428,
+      "step": 17450
+    },
+    {
+      "epoch": 257.0001784386617,
+      "grad_norm": 0.7243895530700684,
+      "learning_rate": 9.156720798278224e-05,
+      "loss": 0.2517,
+      "step": 17500
+    },
+    {
+      "epoch": 257.0001784386617,
+      "eval_accuracy": 0.8649081285942019,
+      "eval_loss": 1.766264796257019,
+      "eval_macro_f1": 0.6946266382039408,
+      "eval_runtime": 47.9299,
+      "eval_samples_per_second": 14.02,
+      "eval_steps_per_second": 3.505,
+      "step": 17500
+    },
+    {
+      "epoch": 258.00004460966545,
+      "grad_norm": 0.6043080687522888,
+      "learning_rate": 9.152807669731951e-05,
+      "loss": 0.2468,
+      "step": 17550
+    },
+    {
+      "epoch": 258.00041635687734,
+      "grad_norm": 0.5234358906745911,
+      "learning_rate": 9.148894541185678e-05,
+      "loss": 0.2414,
+      "step": 17600
+    },
+    {
+      "epoch": 258.00041635687734,
+      "eval_accuracy": 0.8765574733951784,
+      "eval_loss": 1.6841180324554443,
+      "eval_macro_f1": 0.7006466360850203,
+      "eval_runtime": 48.1158,
+      "eval_samples_per_second": 13.966,
+      "eval_steps_per_second": 3.492,
+      "step": 17600
+    },
+    {
+      "epoch": 259.00028252788104,
+      "grad_norm": 0.5486935973167419,
+      "learning_rate": 9.144981412639405e-05,
+      "loss": 0.245,
+      "step": 17650
+    },
+    {
+      "epoch": 260.00014869888474,
+      "grad_norm": 0.5566163063049316,
+      "learning_rate": 9.141068284093132e-05,
+      "loss": 0.2462,
+      "step": 17700
+    },
+    {
+      "epoch": 260.00014869888474,
+      "eval_accuracy": 0.8747919852117518,
+      "eval_loss": 1.6303309202194214,
+      "eval_macro_f1": 0.7016985634720629,
+      "eval_runtime": 48.1791,
+      "eval_samples_per_second": 13.948,
+      "eval_steps_per_second": 3.487,
+      "step": 17700
+    },
+    {
+      "epoch": 261.0000148698885,
+      "grad_norm": 0.6687771677970886,
+      "learning_rate": 9.13715515554686e-05,
+      "loss": 0.2471,
+      "step": 17750
+    },
+    {
+      "epoch": 261.0003866171004,
+      "grad_norm": 0.799909234046936,
+      "learning_rate": 9.133242027000587e-05,
+      "loss": 0.2433,
+      "step": 17800
+    },
+    {
+      "epoch": 261.0003866171004,
+      "eval_accuracy": 0.8727416528292241,
+      "eval_loss": 1.6468901634216309,
+      "eval_macro_f1": 0.6941630525987106,
+      "eval_runtime": 47.9851,
+      "eval_samples_per_second": 14.004,
+      "eval_steps_per_second": 3.501,
+      "step": 17800
+    },
+    {
+      "epoch": 262.0002527881041,
+      "grad_norm": 0.40486371517181396,
+      "learning_rate": 9.129328898454315e-05,
+      "loss": 0.2451,
+      "step": 17850
+    },
+    {
+      "epoch": 263.0001189591078,
+      "grad_norm": 0.6486377120018005,
+      "learning_rate": 9.125415769908043e-05,
+      "loss": 0.2461,
+      "step": 17900
+    },
+    {
+      "epoch": 263.0001189591078,
+      "eval_accuracy": 0.871701417327654,
+      "eval_loss": 1.6379040479660034,
+      "eval_macro_f1": 0.6962308343355683,
+      "eval_runtime": 47.5217,
+      "eval_samples_per_second": 14.141,
+      "eval_steps_per_second": 3.535,
+      "step": 17900
+    },
+    {
+      "epoch": 263.0004907063197,
+      "grad_norm": 0.6154794096946716,
+      "learning_rate": 9.121502641361769e-05,
+      "loss": 0.2429,
+      "step": 17950
+    },
+    {
+      "epoch": 264.00035687732344,
+      "grad_norm": 0.868390679359436,
+      "learning_rate": 9.117589512815497e-05,
+      "loss": 0.245,
+      "step": 18000
+    },
+    {
+      "epoch": 264.00035687732344,
+      "eval_accuracy": 0.8723792366151837,
+      "eval_loss": 1.644251823425293,
+      "eval_macro_f1": 0.6975282371701368,
+      "eval_runtime": 47.5594,
+      "eval_samples_per_second": 14.13,
+      "eval_steps_per_second": 3.532,
+      "step": 18000
+    },
+    {
+      "epoch": 265.00022304832714,
+      "grad_norm": 0.9866790771484375,
+      "learning_rate": 9.113676384269224e-05,
+      "loss": 0.2421,
+      "step": 18050
+    },
+    {
+      "epoch": 266.00008921933085,
+      "grad_norm": 0.8919183611869812,
+      "learning_rate": 9.109763255722951e-05,
+      "loss": 0.2463,
+      "step": 18100
+    },
+    {
+      "epoch": 266.00008921933085,
+      "eval_accuracy": 0.876137634952197,
+      "eval_loss": 1.63772714138031,
+      "eval_macro_f1": 0.7015320633600683,
+      "eval_runtime": 47.5587,
+      "eval_samples_per_second": 14.13,
+      "eval_steps_per_second": 3.532,
+      "step": 18100
+    },
+    {
+      "epoch": 266.00046096654273,
+      "grad_norm": 0.6732456684112549,
+      "learning_rate": 9.105850127176678e-05,
+      "loss": 0.245,
+      "step": 18150
+    },
+    {
+      "epoch": 267.0003271375465,
+      "grad_norm": 0.9901390671730042,
+      "learning_rate": 9.101936998630406e-05,
+      "loss": 0.2433,
+      "step": 18200
+    },
+    {
+      "epoch": 267.0003271375465,
+      "eval_accuracy": 0.8707050401772733,
+      "eval_loss": 1.7547792196273804,
+      "eval_macro_f1": 0.6976431672619668,
+      "eval_runtime": 48.0258,
+      "eval_samples_per_second": 13.992,
+      "eval_steps_per_second": 3.498,
+      "step": 18200
+    },
+    {
+      "epoch": 268.0001933085502,
+      "grad_norm": 0.6878373622894287,
+      "learning_rate": 9.098023870084132e-05,
+      "loss": 0.2455,
+      "step": 18250
+    },
+    {
+      "epoch": 269.0000594795539,
+      "grad_norm": 0.7168475985527039,
+      "learning_rate": 9.09411074153786e-05,
+      "loss": 0.2398,
+      "step": 18300
+    },
+    {
+      "epoch": 269.0000594795539,
+      "eval_accuracy": 0.8749312429343421,
+      "eval_loss": 1.5549407005310059,
+      "eval_macro_f1": 0.6991405662687679,
+      "eval_runtime": 47.4439,
+      "eval_samples_per_second": 14.164,
+      "eval_steps_per_second": 3.541,
+      "step": 18300
+    },
+    {
+      "epoch": 269.0004312267658,
+      "grad_norm": 0.4068659245967865,
+      "learning_rate": 9.090197612991587e-05,
+      "loss": 0.2422,
+      "step": 18350
+    },
+    {
+      "epoch": 270.00029739776954,
+      "grad_norm": 0.5065603852272034,
+      "learning_rate": 9.086284484445314e-05,
+      "loss": 0.242,
+      "step": 18400
+    },
+    {
+      "epoch": 270.00029739776954,
+      "eval_accuracy": 0.8761036310202004,
+      "eval_loss": 1.6925346851348877,
+      "eval_macro_f1": 0.7020842188896902,
+      "eval_runtime": 47.8897,
+      "eval_samples_per_second": 14.032,
+      "eval_steps_per_second": 3.508,
+      "step": 18400
+    },
+    {
+      "epoch": 271.00016356877325,
+      "grad_norm": 0.6289934515953064,
+      "learning_rate": 9.082371355899041e-05,
+      "loss": 0.245,
+      "step": 18450
+    },
+    {
+      "epoch": 272.00002973977695,
+      "grad_norm": 0.5134521126747131,
+      "learning_rate": 9.078458227352768e-05,
+      "loss": 0.2429,
+      "step": 18500
+    },
+    {
+      "epoch": 272.00002973977695,
+      "eval_accuracy": 0.8727523810274236,
+      "eval_loss": 1.7503160238265991,
+      "eval_macro_f1": 0.6992435638111616,
+      "eval_runtime": 47.9819,
+      "eval_samples_per_second": 14.005,
+      "eval_steps_per_second": 3.501,
+      "step": 18500
+    },
+    {
+      "epoch": 272.00040148698884,
+      "grad_norm": 0.5185211300849915,
+      "learning_rate": 9.074545098806497e-05,
+      "loss": 0.2471,
+      "step": 18550
+    },
+    {
+      "epoch": 273.00026765799254,
+      "grad_norm": 0.6770671606063843,
+      "learning_rate": 9.070631970260224e-05,
+      "loss": 0.2428,
+      "step": 18600
+    },
+    {
+      "epoch": 273.00026765799254,
+      "eval_accuracy": 0.8747771988285471,
+      "eval_loss": 1.6137670278549194,
+      "eval_macro_f1": 0.7017125484073795,
+      "eval_runtime": 48.0779,
+      "eval_samples_per_second": 13.977,
+      "eval_steps_per_second": 3.494,
+      "step": 18600
+    },
+    {
+      "epoch": 274.0001338289963,
+      "grad_norm": 0.7311360836029053,
+      "learning_rate": 9.066718841713951e-05,
+      "loss": 0.242,
+      "step": 18650
+    },
+    {
+      "epoch": 274.0005055762082,
+      "grad_norm": 0.8545860052108765,
+      "learning_rate": 9.062805713167678e-05,
+      "loss": 0.2442,
+      "step": 18700
+    },
+    {
+      "epoch": 274.0005055762082,
+      "eval_accuracy": 0.875472191083495,
+      "eval_loss": 1.582758903503418,
+      "eval_macro_f1": 0.7032866342746882,
+      "eval_runtime": 47.9354,
+      "eval_samples_per_second": 14.019,
+      "eval_steps_per_second": 3.505,
+      "step": 18700
+    },
+    {
+      "epoch": 275.0003717472119,
+      "grad_norm": 0.5233895182609558,
+      "learning_rate": 9.058892584621406e-05,
+      "loss": 0.2428,
+      "step": 18750
+    },
+    {
+      "epoch": 276.0002379182156,
+      "grad_norm": 0.750148355960846,
+      "learning_rate": 9.054979456075132e-05,
+      "loss": 0.25,
+      "step": 18800
+    },
+    {
+      "epoch": 276.0002379182156,
+      "eval_accuracy": 0.8717148881461516,
+      "eval_loss": 1.7209230661392212,
+      "eval_macro_f1": 0.6960070669466294,
+      "eval_runtime": 47.9374,
+      "eval_samples_per_second": 14.018,
+      "eval_steps_per_second": 3.505,
+      "step": 18800
+    },
+    {
+      "epoch": 277.00010408921935,
+      "grad_norm": 0.6850331425666809,
+      "learning_rate": 9.05106632752886e-05,
+      "loss": 0.2445,
+      "step": 18850
+    },
+    {
+      "epoch": 277.00047583643124,
+      "grad_norm": 1.3163009881973267,
+      "learning_rate": 9.047153198982587e-05,
+      "loss": 0.2441,
+      "step": 18900
+    },
+    {
+      "epoch": 277.00047583643124,
+      "eval_accuracy": 0.8765440978351273,
+      "eval_loss": 1.6306798458099365,
+      "eval_macro_f1": 0.705625688490632,
+      "eval_runtime": 47.7345,
+      "eval_samples_per_second": 14.078,
+      "eval_steps_per_second": 3.519,
+      "step": 18900
+    },
+    {
+      "epoch": 278.00034200743494,
+      "grad_norm": 0.591773509979248,
+      "learning_rate": 9.043240070436314e-05,
+      "loss": 0.2416,
+      "step": 18950
+    },
+    {
+      "epoch": 279.00020817843864,
+      "grad_norm": 0.4475375711917877,
+      "learning_rate": 9.039326941890041e-05,
+      "loss": 0.2414,
+      "step": 19000
+    },
+    {
+      "epoch": 279.00020817843864,
+      "eval_accuracy": 0.8760741830973673,
+      "eval_loss": 1.6609206199645996,
+      "eval_macro_f1": 0.7028777819955179,
+      "eval_runtime": 48.2216,
+      "eval_samples_per_second": 13.936,
+      "eval_steps_per_second": 3.484,
+      "step": 19000
+    },
+    {
+      "epoch": 280.0000743494424,
+      "grad_norm": 0.6211313009262085,
+      "learning_rate": 9.035413813343768e-05,
+      "loss": 0.2366,
+      "step": 19050
+    },
+    {
+      "epoch": 280.0004460966543,
+      "grad_norm": 0.7641573548316956,
+      "learning_rate": 9.031500684797496e-05,
+      "loss": 0.2371,
+      "step": 19100
+    },
+    {
+      "epoch": 280.0004460966543,
+      "eval_accuracy": 0.8740981031647922,
+      "eval_loss": 1.7170416116714478,
+      "eval_macro_f1": 0.7018808537743648,
+      "eval_runtime": 48.2987,
+      "eval_samples_per_second": 13.913,
+      "eval_steps_per_second": 3.478,
+      "step": 19100
+    },
+    {
+      "epoch": 281.000312267658,
+      "grad_norm": 0.6113197207450867,
+      "learning_rate": 9.027587556251223e-05,
+      "loss": 0.2395,
+      "step": 19150
+    },
+    {
+      "epoch": 282.0001784386617,
+      "grad_norm": 0.7073433995246887,
+      "learning_rate": 9.02367442770495e-05,
+      "loss": 0.2385,
+      "step": 19200
+    },
+    {
+      "epoch": 282.0001784386617,
+      "eval_accuracy": 0.8743977079113086,
+      "eval_loss": 1.7015947103500366,
+      "eval_macro_f1": 0.7018147516994673,
+      "eval_runtime": 47.6214,
+      "eval_samples_per_second": 14.111,
+      "eval_steps_per_second": 3.528,
+      "step": 19200
+    },
+    {
+      "epoch": 283.00004460966545,
+      "grad_norm": 0.5074424147605896,
+      "learning_rate": 9.019761299158677e-05,
+      "loss": 0.2446,
+      "step": 19250
+    },
+    {
+      "epoch": 283.00041635687734,
+      "grad_norm": 0.6375361084938049,
+      "learning_rate": 9.015848170612406e-05,
+      "loss": 0.2385,
+      "step": 19300
+    },
+    {
+      "epoch": 283.00041635687734,
+      "eval_accuracy": 0.8740577076537963,
+      "eval_loss": 1.6805062294006348,
+      "eval_macro_f1": 0.7021029920673127,
+      "eval_runtime": 47.3156,
+      "eval_samples_per_second": 14.203,
+      "eval_steps_per_second": 3.551,
+      "step": 19300
+    },
+    {
+      "epoch": 284.00028252788104,
+      "grad_norm": 0.8452440500259399,
+      "learning_rate": 9.011935042066132e-05,
+      "loss": 0.2373,
+      "step": 19350
+    },
+    {
+      "epoch": 285.00014869888474,
+      "grad_norm": 0.6203405857086182,
+      "learning_rate": 9.00802191351986e-05,
+      "loss": 0.2422,
+      "step": 19400
+    },
+    {
+      "epoch": 285.00014869888474,
+      "eval_accuracy": 0.8719450365356796,
+      "eval_loss": 1.74236261844635,
+      "eval_macro_f1": 0.6972832445042872,
+      "eval_runtime": 47.5763,
+      "eval_samples_per_second": 14.125,
+      "eval_steps_per_second": 3.531,
+      "step": 19400
+    },
+    {
+      "epoch": 286.0000148698885,
+      "grad_norm": 0.6126111745834351,
+      "learning_rate": 9.004108784973587e-05,
+      "loss": 0.2386,
+      "step": 19450
+    },
+    {
+      "epoch": 286.0003866171004,
+      "grad_norm": 0.4864870011806488,
+      "learning_rate": 9.000195656427314e-05,
+      "loss": 0.2403,
+      "step": 19500
+    },
+    {
+      "epoch": 286.0003866171004,
+      "eval_accuracy": 0.8724967535271402,
+      "eval_loss": 1.7255445718765259,
+      "eval_macro_f1": 0.6975033096469039,
+      "eval_runtime": 47.0772,
+      "eval_samples_per_second": 14.274,
+      "eval_steps_per_second": 3.569,
+      "step": 19500
+    },
+    {
+      "epoch": 287.0002527881041,
+      "grad_norm": 1.169779658317566,
+      "learning_rate": 8.996282527881041e-05,
+      "loss": 0.2394,
+      "step": 19550
+    },
+    {
+      "epoch": 288.0001189591078,
+      "grad_norm": 0.7006259560585022,
+      "learning_rate": 8.992369399334769e-05,
+      "loss": 0.2439,
+      "step": 19600
+    },
+    {
+      "epoch": 288.0001189591078,
+      "eval_accuracy": 0.8719169307254028,
+      "eval_loss": 1.7644731998443604,
+      "eval_macro_f1": 0.6958908876410843,
+      "eval_runtime": 46.9308,
+      "eval_samples_per_second": 14.319,
+      "eval_steps_per_second": 3.58,
+      "step": 19600
+    },
+    {
+      "epoch": 288.0004907063197,
+      "grad_norm": 0.7097400426864624,
+      "learning_rate": 8.988456270788496e-05,
+      "loss": 0.2468,
+      "step": 19650
+    },
+    {
+      "epoch": 289.00035687732344,
+      "grad_norm": 0.7856957912445068,
+      "learning_rate": 8.984543142242223e-05,
+      "loss": 0.2414,
+      "step": 19700
+    },
+    {
+      "epoch": 289.00035687732344,
+      "eval_accuracy": 0.8756746554351618,
+      "eval_loss": 1.7827283143997192,
+      "eval_macro_f1": 0.7025011619593726,
+      "eval_runtime": 47.4623,
+      "eval_samples_per_second": 14.159,
+      "eval_steps_per_second": 3.54,
+      "step": 19700
+    },
+    {
+      "epoch": 290.00022304832714,
+      "grad_norm": 0.739522397518158,
+      "learning_rate": 8.98063001369595e-05,
+      "loss": 0.2357,
+      "step": 19750
+    },
+    {
+      "epoch": 291.00008921933085,
+      "grad_norm": 0.5940971970558167,
+      "learning_rate": 8.976716885149677e-05,
+      "loss": 0.2403,
+      "step": 19800
+    },
+    {
+      "epoch": 291.00008921933085,
+      "eval_accuracy": 0.8785574074303582,
+      "eval_loss": 1.5922495126724243,
+      "eval_macro_f1": 0.7042178263465622,
+      "eval_runtime": 47.3707,
+      "eval_samples_per_second": 14.186,
+      "eval_steps_per_second": 3.546,
+      "step": 19800
+    },
+    {
+      "epoch": 291.00046096654273,
+      "grad_norm": 0.4905942678451538,
+      "learning_rate": 8.972803756603406e-05,
+      "loss": 0.2342,
+      "step": 19850
+    },
+    {
+      "epoch": 292.0003271375465,
+      "grad_norm": 0.6634092926979065,
+      "learning_rate": 8.968890628057131e-05,
+      "loss": 0.233,
+      "step": 19900
+    },
+    {
+      "epoch": 292.0003271375465,
+      "eval_accuracy": 0.8748785909310688,
+      "eval_loss": 1.8774493932724,
+      "eval_macro_f1": 0.699269934433628,
+      "eval_runtime": 47.8531,
+      "eval_samples_per_second": 14.043,
+      "eval_steps_per_second": 3.511,
+      "step": 19900
+    },
+    {
+      "epoch": 293.0001933085502,
+      "grad_norm": 0.6518082618713379,
+      "learning_rate": 8.96497749951086e-05,
+      "loss": 0.2355,
+      "step": 19950
+    },
+    {
+      "epoch": 294.0000594795539,
+      "grad_norm": 0.6070118546485901,
+      "learning_rate": 8.961064370964587e-05,
+      "loss": 0.2357,
+      "step": 20000
+    },
+    {
+      "epoch": 294.0000594795539,
+      "eval_accuracy": 0.8755963691803673,
+      "eval_loss": 1.6461971998214722,
+      "eval_macro_f1": 0.6989364797944599,
+      "eval_runtime": 47.0949,
+      "eval_samples_per_second": 14.269,
+      "eval_steps_per_second": 3.567,
+      "step": 20000
+    },
+    {
+      "epoch": 294.0004312267658,
+      "grad_norm": 0.9214955568313599,
+      "learning_rate": 8.957151242418314e-05,
+      "loss": 0.2373,
+      "step": 20050
+    },
+    {
+      "epoch": 295.00029739776954,
+      "grad_norm": 0.7139339447021484,
+      "learning_rate": 8.95323811387204e-05,
+      "loss": 0.2468,
+      "step": 20100
+    },
+    {
+      "epoch": 295.00029739776954,
+      "eval_accuracy": 0.8752588183006401,
+      "eval_loss": 1.6570920944213867,
+      "eval_macro_f1": 0.7040971519065213,
+      "eval_runtime": 47.3271,
+      "eval_samples_per_second": 14.199,
+      "eval_steps_per_second": 3.55,
+      "step": 20100
+    },
+    {
+      "epoch": 296.00016356877325,
+      "grad_norm": 0.8615701794624329,
+      "learning_rate": 8.949324985325769e-05,
+      "loss": 0.2359,
+      "step": 20150
+    },
+    {
+      "epoch": 297.00002973977695,
+      "grad_norm": 0.5898351073265076,
+      "learning_rate": 8.945411856779495e-05,
+      "loss": 0.236,
+      "step": 20200
+    },
+    {
+      "epoch": 297.00002973977695,
+      "eval_accuracy": 0.8752191850777665,
+      "eval_loss": 1.6926116943359375,
+      "eval_macro_f1": 0.6999326241361957,
+      "eval_runtime": 47.1963,
+      "eval_samples_per_second": 14.238,
+      "eval_steps_per_second": 3.56,
+      "step": 20200
+    },
+    {
+      "epoch": 297.00040148698884,
+      "grad_norm": 0.7265653014183044,
+      "learning_rate": 8.941498728233223e-05,
+      "loss": 0.2431,
+      "step": 20250
+    },
+    {
+      "epoch": 298.00026765799254,
+      "grad_norm": 0.8821515440940857,
+      "learning_rate": 8.93758559968695e-05,
+      "loss": 0.2341,
+      "step": 20300
+    },
+    {
+      "epoch": 298.00026765799254,
+      "eval_accuracy": 0.8770282183880916,
+      "eval_loss": 1.7504407167434692,
+      "eval_macro_f1": 0.7051157111563411,
+      "eval_runtime": 47.5468,
+      "eval_samples_per_second": 14.133,
+      "eval_steps_per_second": 3.533,
+      "step": 20300
+    },
+    {
+      "epoch": 299.0001338289963,
+      "grad_norm": 0.8511101007461548,
+      "learning_rate": 8.933672471140677e-05,
+      "loss": 0.2356,
+      "step": 20350
+    },
+    {
+      "epoch": 299.0005055762082,
+      "grad_norm": 2.185602903366089,
+      "learning_rate": 8.929759342594405e-05,
+      "loss": 0.2359,
+      "step": 20400
+    },
+    {
+      "epoch": 299.0005055762082,
+      "eval_accuracy": 0.8763592902732993,
+      "eval_loss": 1.6859861612319946,
+      "eval_macro_f1": 0.700068017274899,
+      "eval_runtime": 47.108,
+      "eval_samples_per_second": 14.265,
+      "eval_steps_per_second": 3.566,
+      "step": 20400
+    },
+    {
+      "epoch": 300.0003717472119,
+      "grad_norm": 0.6592544317245483,
+      "learning_rate": 8.925846214048131e-05,
+      "loss": 0.2336,
+      "step": 20450
+    },
+    {
+      "epoch": 301.0002379182156,
+      "grad_norm": 1.0603123903274536,
+      "learning_rate": 8.92193308550186e-05,
+      "loss": 0.2377,
+      "step": 20500
+    },
+    {
+      "epoch": 301.0002379182156,
+      "eval_accuracy": 0.8745820471710939,
+      "eval_loss": 1.668921709060669,
+      "eval_macro_f1": 0.6997895576456115,
+      "eval_runtime": 47.3196,
+      "eval_samples_per_second": 14.201,
+      "eval_steps_per_second": 3.55,
+      "step": 20500
+    },
+    {
+      "epoch": 302.00010408921935,
+      "grad_norm": 0.744284451007843,
+      "learning_rate": 8.918019956955586e-05,
+      "loss": 0.2356,
+      "step": 20550
+    },
+    {
+      "epoch": 302.00047583643124,
+      "grad_norm": 0.5185156464576721,
+      "learning_rate": 8.914106828409313e-05,
+      "loss": 0.2336,
+      "step": 20600
+    },
+    {
+      "epoch": 302.00047583643124,
+      "eval_accuracy": 0.8787366418342881,
+      "eval_loss": 1.6256543397903442,
+      "eval_macro_f1": 0.7081033424785741,
+      "eval_runtime": 48.159,
+      "eval_samples_per_second": 13.954,
+      "eval_steps_per_second": 3.488,
+      "step": 20600
+    },
+    {
+      "epoch": 303.00034200743494,
+      "grad_norm": 0.48090434074401855,
+      "learning_rate": 8.91019369986304e-05,
+      "loss": 0.2319,
+      "step": 20650
+    },
+    {
+      "epoch": 304.00020817843864,
+      "grad_norm": 0.5701866149902344,
+      "learning_rate": 8.906280571316769e-05,
+      "loss": 0.2329,
+      "step": 20700
+    },
+    {
+      "epoch": 304.00020817843864,
+      "eval_accuracy": 0.8759103023549644,
+      "eval_loss": 1.680335283279419,
+      "eval_macro_f1": 0.7017996446225558,
+      "eval_runtime": 47.2644,
+      "eval_samples_per_second": 14.218,
+      "eval_steps_per_second": 3.554,
+      "step": 20700
+    },
+    {
+      "epoch": 305.0000743494424,
+      "grad_norm": 0.6104369759559631,
+      "learning_rate": 8.902367442770494e-05,
+      "loss": 0.2359,
+      "step": 20750
+    },
+    {
+      "epoch": 305.0004460966543,
+      "grad_norm": 0.6198025941848755,
+      "learning_rate": 8.898454314224223e-05,
+      "loss": 0.236,
+      "step": 20800
+    },
+    {
+      "epoch": 305.0004460966543,
+      "eval_accuracy": 0.8783429369856264,
+      "eval_loss": 1.7152197360992432,
+      "eval_macro_f1": 0.7087897186605305,
+      "eval_runtime": 47.7992,
+      "eval_samples_per_second": 14.059,
+      "eval_steps_per_second": 3.515,
+      "step": 20800
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 134500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 20,
+        "early_stopping_threshold": 0.0
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 0
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}