diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,2227 +1,3118 @@
 {
-  "best_metric": 0.4305761158466339,
-  "best_model_checkpoint": "./vit-base-beans/checkpoint-2680",
-  "epoch": 1.9734904270986746,
-  "global_step": 2680,
+  "best_metric": 0.3434308171272278,
+  "best_model_checkpoint": "./vit-base-beans/checkpoint-3760",
+  "epoch": 1.9810326659641728,
+  "global_step": 3760,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.01,
-      "learning_rate": 0.00019926362297496318,
-      "loss": 2.073,
+      "learning_rate": 0.00019947312961011594,
+      "loss": 2.0892,
       "step": 10
     },
     {
       "epoch": 0.01,
-      "learning_rate": 0.00019852724594992638,
-      "loss": 1.5903,
+      "learning_rate": 0.00019894625922023182,
+      "loss": 1.6471,
       "step": 20
     },
     {
       "epoch": 0.02,
-      "learning_rate": 0.00019779086892488955,
-      "loss": 1.4651,
+      "learning_rate": 0.00019841938883034775,
+      "loss": 1.407,
       "step": 30
     },
     {
-      "epoch": 0.03,
-      "learning_rate": 0.00019705449189985275,
-      "loss": 1.2028,
+      "epoch": 0.02,
+      "learning_rate": 0.00019789251844046365,
+      "loss": 1.2934,
       "step": 40
     },
     {
-      "epoch": 0.03,
-      "eval_accuracy": 0.5852670349907919,
-      "eval_loss": 1.2089080810546875,
-      "eval_runtime": 48.3873,
-      "eval_samples_per_second": 56.11,
-      "eval_steps_per_second": 7.027,
+      "epoch": 0.02,
+      "eval_accuracy": 0.554004214963119,
+      "eval_loss": 1.2484745979309082,
+      "eval_runtime": 40.9693,
+      "eval_samples_per_second": 92.655,
+      "eval_steps_per_second": 11.594,
       "step": 40
     },
     {
-      "epoch": 0.04,
-      "learning_rate": 0.00019631811487481592,
-      "loss": 1.1103,
+      "epoch": 0.03,
+      "learning_rate": 0.00019736564805057958,
+      "loss": 1.3421,
       "step": 50
     },
     {
-      "epoch": 0.04,
-      "learning_rate": 0.0001955817378497791,
-      "loss": 1.2374,
+      "epoch": 0.03,
+      "learning_rate": 0.00019683877766069548,
+      "loss": 1.1489,
       "step": 60
     },
     {
-      "epoch": 0.05,
-      "learning_rate": 0.00019484536082474229,
-      "loss": 1.1704,
+      "epoch": 0.04,
+      "learning_rate": 0.0001963119072708114,
+      "loss": 1.1359,
       "step": 70
     },
     {
-      "epoch": 0.06,
-      "learning_rate": 0.00019410898379970546,
-      "loss": 1.0396,
+      "epoch": 0.04,
+      "learning_rate": 0.00019578503688092732,
+      "loss": 1.0804,
       "step": 80
     },
     {
-      "epoch": 0.06,
-      "eval_accuracy": 0.6504604051565378,
-      "eval_loss": 1.0024482011795044,
-      "eval_runtime": 49.385,
-      "eval_samples_per_second": 54.976,
-      "eval_steps_per_second": 6.885,
+      "epoch": 0.04,
+      "eval_accuracy": 0.5076396206533192,
+      "eval_loss": 1.2709373235702515,
+      "eval_runtime": 41.0324,
+      "eval_samples_per_second": 92.512,
+      "eval_steps_per_second": 11.576,
       "step": 80
     },
     {
-      "epoch": 0.07,
-      "learning_rate": 0.00019337260677466863,
-      "loss": 0.9898,
+      "epoch": 0.05,
+      "learning_rate": 0.0001952581664910432,
+      "loss": 1.281,
       "step": 90
     },
     {
-      "epoch": 0.07,
-      "learning_rate": 0.00019263622974963183,
-      "loss": 0.9049,
+      "epoch": 0.05,
+      "learning_rate": 0.00019473129610115912,
+      "loss": 1.1561,
       "step": 100
     },
     {
-      "epoch": 0.08,
-      "learning_rate": 0.000191899852724595,
-      "loss": 1.0411,
+      "epoch": 0.06,
+      "learning_rate": 0.00019420442571127503,
+      "loss": 0.9883,
       "step": 110
     },
     {
-      "epoch": 0.09,
-      "learning_rate": 0.0001911634756995582,
-      "loss": 1.031,
+      "epoch": 0.06,
+      "learning_rate": 0.00019367755532139096,
+      "loss": 1.0225,
       "step": 120
     },
     {
-      "epoch": 0.09,
-      "eval_accuracy": 0.6088397790055249,
-      "eval_loss": 1.0570313930511475,
-      "eval_runtime": 52.8083,
-      "eval_samples_per_second": 51.412,
-      "eval_steps_per_second": 6.438,
+      "epoch": 0.06,
+      "eval_accuracy": 0.5835089567966281,
+      "eval_loss": 1.1064503192901611,
+      "eval_runtime": 41.5232,
+      "eval_samples_per_second": 91.419,
+      "eval_steps_per_second": 11.439,
       "step": 120
     },
     {
-      "epoch": 0.1,
-      "learning_rate": 0.00019042709867452136,
-      "loss": 0.9189,
+      "epoch": 0.07,
+      "learning_rate": 0.00019315068493150686,
+      "loss": 1.0751,
       "step": 130
     },
     {
-      "epoch": 0.1,
-      "learning_rate": 0.00018969072164948454,
-      "loss": 1.0785,
+      "epoch": 0.07,
+      "learning_rate": 0.00019262381454162277,
+      "loss": 1.2585,
       "step": 140
     },
     {
-      "epoch": 0.11,
-      "learning_rate": 0.00018895434462444773,
-      "loss": 0.9283,
+      "epoch": 0.08,
+      "learning_rate": 0.0001920969441517387,
+      "loss": 1.0325,
       "step": 150
     },
     {
-      "epoch": 0.12,
-      "learning_rate": 0.0001882179675994109,
-      "loss": 0.94,
+      "epoch": 0.08,
+      "learning_rate": 0.00019157007376185457,
+      "loss": 1.1999,
       "step": 160
     },
     {
-      "epoch": 0.12,
-      "eval_accuracy": 0.6386740331491713,
-      "eval_loss": 0.9433544278144836,
-      "eval_runtime": 51.6666,
-      "eval_samples_per_second": 52.548,
-      "eval_steps_per_second": 6.581,
+      "epoch": 0.08,
+      "eval_accuracy": 0.6240779768177028,
+      "eval_loss": 0.998985767364502,
+      "eval_runtime": 41.8244,
+      "eval_samples_per_second": 90.76,
+      "eval_steps_per_second": 11.357,
       "step": 160
     },
     {
-      "epoch": 0.13,
-      "learning_rate": 0.00018748159057437408,
-      "loss": 1.0078,
+      "epoch": 0.09,
+      "learning_rate": 0.0001910432033719705,
+      "loss": 1.0211,
       "step": 170
     },
     {
-      "epoch": 0.13,
-      "learning_rate": 0.00018674521354933727,
-      "loss": 0.9582,
+      "epoch": 0.09,
+      "learning_rate": 0.00019051633298208643,
+      "loss": 1.0205,
       "step": 180
     },
     {
-      "epoch": 0.14,
-      "learning_rate": 0.00018600883652430044,
-      "loss": 0.9966,
+      "epoch": 0.1,
+      "learning_rate": 0.0001899894625922023,
+      "loss": 1.0047,
       "step": 190
     },
     {
-      "epoch": 0.15,
-      "learning_rate": 0.00018527245949926364,
-      "loss": 0.8435,
+      "epoch": 0.11,
+      "learning_rate": 0.00018946259220231824,
+      "loss": 0.9969,
       "step": 200
     },
     {
-      "epoch": 0.15,
-      "eval_accuracy": 0.69097605893186,
-      "eval_loss": 0.8427824378013611,
-      "eval_runtime": 51.4559,
-      "eval_samples_per_second": 52.764,
-      "eval_steps_per_second": 6.608,
+      "epoch": 0.11,
+      "eval_accuracy": 0.6538461538461539,
+      "eval_loss": 0.9371430277824402,
+      "eval_runtime": 41.6306,
+      "eval_samples_per_second": 91.183,
+      "eval_steps_per_second": 11.41,
       "step": 200
     },
     {
-      "epoch": 0.15,
-      "learning_rate": 0.0001845360824742268,
-      "loss": 0.7854,
+      "epoch": 0.11,
+      "learning_rate": 0.00018893572181243414,
+      "loss": 1.025,
       "step": 210
     },
     {
-      "epoch": 0.16,
-      "learning_rate": 0.00018379970544918998,
-      "loss": 0.9786,
+      "epoch": 0.12,
+      "learning_rate": 0.00018840885142255007,
+      "loss": 1.0511,
       "step": 220
     },
     {
-      "epoch": 0.17,
-      "learning_rate": 0.00018306332842415318,
-      "loss": 0.736,
+      "epoch": 0.12,
+      "learning_rate": 0.00018788198103266598,
+      "loss": 0.8609,
       "step": 230
     },
     {
-      "epoch": 0.18,
-      "learning_rate": 0.00018232695139911635,
-      "loss": 0.851,
+      "epoch": 0.13,
+      "learning_rate": 0.00018735511064278188,
+      "loss": 0.8846,
       "step": 240
     },
     {
-      "epoch": 0.18,
-      "eval_accuracy": 0.6504604051565378,
-      "eval_loss": 0.9459872841835022,
-      "eval_runtime": 51.2583,
-      "eval_samples_per_second": 52.967,
-      "eval_steps_per_second": 6.633,
+      "epoch": 0.13,
+      "eval_accuracy": 0.6620126448893572,
+      "eval_loss": 0.9217966794967651,
+      "eval_runtime": 41.7497,
+      "eval_samples_per_second": 90.923,
+      "eval_steps_per_second": 11.377,
       "step": 240
     },
     {
-      "epoch": 0.18,
-      "learning_rate": 0.00018159057437407955,
-      "loss": 0.9344,
+      "epoch": 0.13,
+      "learning_rate": 0.0001868282402528978,
+      "loss": 1.0542,
       "step": 250
     },
     {
-      "epoch": 0.19,
-      "learning_rate": 0.00018085419734904272,
-      "loss": 0.8802,
+      "epoch": 0.14,
+      "learning_rate": 0.0001863013698630137,
+      "loss": 0.8937,
       "step": 260
     },
     {
-      "epoch": 0.2,
-      "learning_rate": 0.0001801178203240059,
-      "loss": 0.772,
+      "epoch": 0.14,
+      "learning_rate": 0.00018577449947312962,
+      "loss": 0.9493,
       "step": 270
     },
     {
-      "epoch": 0.21,
-      "learning_rate": 0.0001793814432989691,
-      "loss": 0.9324,
+      "epoch": 0.15,
+      "learning_rate": 0.00018524762908324552,
+      "loss": 0.9374,
       "step": 280
     },
     {
-      "epoch": 0.21,
-      "eval_accuracy": 0.6593001841620626,
-      "eval_loss": 0.9168590903282166,
-      "eval_runtime": 52.8324,
-      "eval_samples_per_second": 51.389,
-      "eval_steps_per_second": 6.435,
+      "epoch": 0.15,
+      "eval_accuracy": 0.6704425711275026,
+      "eval_loss": 0.8844485282897949,
+      "eval_runtime": 41.7229,
+      "eval_samples_per_second": 90.981,
+      "eval_steps_per_second": 11.385,
       "step": 280
     },
     {
-      "epoch": 0.21,
-      "learning_rate": 0.00017864506627393226,
-      "loss": 0.6804,
+      "epoch": 0.15,
+      "learning_rate": 0.00018472075869336145,
+      "loss": 0.93,
       "step": 290
     },
     {
-      "epoch": 0.22,
-      "learning_rate": 0.00017790868924889543,
-      "loss": 0.7877,
+      "epoch": 0.16,
+      "learning_rate": 0.00018419388830347736,
+      "loss": 0.9324,
       "step": 300
     },
     {
-      "epoch": 0.23,
-      "learning_rate": 0.00017717231222385863,
-      "loss": 0.9609,
+      "epoch": 0.16,
+      "learning_rate": 0.00018366701791359326,
+      "loss": 0.9806,
       "step": 310
     },
     {
-      "epoch": 0.24,
-      "learning_rate": 0.0001764359351988218,
-      "loss": 0.8963,
+      "epoch": 0.17,
+      "learning_rate": 0.0001831401475237092,
+      "loss": 0.8979,
       "step": 320
     },
     {
-      "epoch": 0.24,
-      "eval_accuracy": 0.7016574585635359,
-      "eval_loss": 0.7979860901832581,
-      "eval_runtime": 50.8385,
-      "eval_samples_per_second": 53.404,
-      "eval_steps_per_second": 6.688,
+      "epoch": 0.17,
+      "eval_accuracy": 0.6862486828240253,
+      "eval_loss": 0.8492920994758606,
+      "eval_runtime": 42.0264,
+      "eval_samples_per_second": 90.324,
+      "eval_steps_per_second": 11.302,
       "step": 320
     },
     {
-      "epoch": 0.24,
-      "learning_rate": 0.000175699558173785,
-      "loss": 0.8531,
+      "epoch": 0.17,
+      "learning_rate": 0.00018261327713382507,
+      "loss": 0.7995,
       "step": 330
     },
     {
-      "epoch": 0.25,
-      "learning_rate": 0.00017496318114874817,
-      "loss": 0.899,
+      "epoch": 0.18,
+      "learning_rate": 0.000182086406743941,
+      "loss": 0.8829,
       "step": 340
     },
     {
-      "epoch": 0.26,
-      "learning_rate": 0.00017422680412371134,
-      "loss": 0.816,
+      "epoch": 0.18,
+      "learning_rate": 0.00018155953635405693,
+      "loss": 0.8794,
       "step": 350
     },
     {
-      "epoch": 0.27,
-      "learning_rate": 0.00017349042709867454,
-      "loss": 0.7532,
+      "epoch": 0.19,
+      "learning_rate": 0.00018103266596417283,
+      "loss": 1.0743,
       "step": 360
     },
     {
-      "epoch": 0.27,
-      "eval_accuracy": 0.7031307550644568,
-      "eval_loss": 0.8063194155693054,
-      "eval_runtime": 50.4103,
-      "eval_samples_per_second": 53.858,
-      "eval_steps_per_second": 6.745,
+      "epoch": 0.19,
+      "eval_accuracy": 0.6680716543730242,
+      "eval_loss": 0.910214900970459,
+      "eval_runtime": 41.8529,
+      "eval_samples_per_second": 90.699,
+      "eval_steps_per_second": 11.349,
       "step": 360
     },
     {
-      "epoch": 0.27,
-      "learning_rate": 0.0001727540500736377,
-      "loss": 0.8984,
+      "epoch": 0.19,
+      "learning_rate": 0.00018050579557428873,
+      "loss": 0.9903,
       "step": 370
     },
     {
-      "epoch": 0.28,
-      "learning_rate": 0.0001720176730486009,
-      "loss": 0.843,
+      "epoch": 0.2,
+      "learning_rate": 0.00017997892518440464,
+      "loss": 0.9447,
       "step": 380
     },
     {
-      "epoch": 0.29,
-      "learning_rate": 0.00017128129602356408,
-      "loss": 0.7647,
+      "epoch": 0.21,
+      "learning_rate": 0.00017945205479452057,
+      "loss": 0.9659,
       "step": 390
     },
     {
-      "epoch": 0.29,
-      "learning_rate": 0.00017054491899852725,
-      "loss": 0.7937,
+      "epoch": 0.21,
+      "learning_rate": 0.00017892518440463647,
+      "loss": 0.8767,
       "step": 400
     },
     {
-      "epoch": 0.29,
-      "eval_accuracy": 0.7222836095764272,
-      "eval_loss": 0.7624691724777222,
-      "eval_runtime": 51.1567,
-      "eval_samples_per_second": 53.072,
-      "eval_steps_per_second": 6.646,
+      "epoch": 0.21,
+      "eval_accuracy": 0.6290832455216017,
+      "eval_loss": 0.9642680287361145,
+      "eval_runtime": 41.7862,
+      "eval_samples_per_second": 90.843,
+      "eval_steps_per_second": 11.367,
       "step": 400
     },
     {
-      "epoch": 0.3,
-      "learning_rate": 0.00016980854197349044,
-      "loss": 0.9007,
+      "epoch": 0.22,
+      "learning_rate": 0.00017839831401475237,
+      "loss": 1.0593,
       "step": 410
     },
     {
-      "epoch": 0.31,
-      "learning_rate": 0.00016907216494845361,
-      "loss": 0.7149,
+      "epoch": 0.22,
+      "learning_rate": 0.0001778714436248683,
+      "loss": 0.8566,
       "step": 420
     },
     {
-      "epoch": 0.32,
-      "learning_rate": 0.00016833578792341679,
-      "loss": 0.8868,
+      "epoch": 0.23,
+      "learning_rate": 0.0001773445732349842,
+      "loss": 0.8882,
       "step": 430
     },
     {
-      "epoch": 0.32,
-      "learning_rate": 0.00016759941089837998,
-      "loss": 0.8365,
+      "epoch": 0.23,
+      "learning_rate": 0.0001768177028451001,
+      "loss": 0.9381,
       "step": 440
     },
     {
-      "epoch": 0.32,
-      "eval_accuracy": 0.7134438305709024,
-      "eval_loss": 0.7824127078056335,
-      "eval_runtime": 50.4947,
-      "eval_samples_per_second": 53.768,
-      "eval_steps_per_second": 6.733,
+      "epoch": 0.23,
+      "eval_accuracy": 0.6383034773445733,
+      "eval_loss": 0.9608025550842285,
+      "eval_runtime": 41.7191,
+      "eval_samples_per_second": 90.99,
+      "eval_steps_per_second": 11.386,
       "step": 440
     },
     {
-      "epoch": 0.33,
-      "learning_rate": 0.00016686303387334315,
-      "loss": 0.7985,
+      "epoch": 0.24,
+      "learning_rate": 0.00017629083245521601,
+      "loss": 0.9577,
       "step": 450
     },
     {
-      "epoch": 0.34,
-      "learning_rate": 0.00016612665684830635,
-      "loss": 0.775,
+      "epoch": 0.24,
+      "learning_rate": 0.00017576396206533195,
+      "loss": 0.8785,
       "step": 460
     },
     {
-      "epoch": 0.35,
-      "learning_rate": 0.00016539027982326952,
-      "loss": 0.7912,
+      "epoch": 0.25,
+      "learning_rate": 0.00017523709167544785,
+      "loss": 0.7606,
       "step": 470
     },
     {
-      "epoch": 0.35,
-      "learning_rate": 0.0001646539027982327,
-      "loss": 0.8333,
+      "epoch": 0.25,
+      "learning_rate": 0.00017471022128556375,
+      "loss": 1.1457,
       "step": 480
     },
     {
-      "epoch": 0.35,
-      "eval_accuracy": 0.7263351749539595,
-      "eval_loss": 0.7473925948143005,
-      "eval_runtime": 51.4866,
-      "eval_samples_per_second": 52.732,
-      "eval_steps_per_second": 6.604,
+      "epoch": 0.25,
+      "eval_accuracy": 0.6654373024236038,
+      "eval_loss": 0.8898976445198059,
+      "eval_runtime": 41.808,
+      "eval_samples_per_second": 90.796,
+      "eval_steps_per_second": 11.361,
       "step": 480
     },
     {
-      "epoch": 0.36,
-      "learning_rate": 0.0001639175257731959,
-      "loss": 0.8501,
+      "epoch": 0.26,
+      "learning_rate": 0.00017418335089567968,
+      "loss": 0.9,
       "step": 490
     },
     {
-      "epoch": 0.37,
-      "learning_rate": 0.00016318114874815906,
-      "loss": 0.6558,
+      "epoch": 0.26,
+      "learning_rate": 0.00017365648050579559,
+      "loss": 0.7624,
       "step": 500
     },
     {
-      "epoch": 0.38,
-      "learning_rate": 0.00016244477172312223,
-      "loss": 0.5491,
+      "epoch": 0.27,
+      "learning_rate": 0.0001731296101159115,
+      "loss": 0.819,
       "step": 510
     },
     {
-      "epoch": 0.38,
-      "learning_rate": 0.00016170839469808543,
-      "loss": 0.6784,
+      "epoch": 0.27,
+      "learning_rate": 0.00017260273972602742,
+      "loss": 0.8516,
       "step": 520
     },
     {
-      "epoch": 0.38,
-      "eval_accuracy": 0.7351749539594844,
-      "eval_loss": 0.7210445404052734,
-      "eval_runtime": 51.0697,
-      "eval_samples_per_second": 53.163,
-      "eval_steps_per_second": 6.658,
+      "epoch": 0.27,
+      "eval_accuracy": 0.6791359325605901,
+      "eval_loss": 0.8746893405914307,
+      "eval_runtime": 41.7582,
+      "eval_samples_per_second": 90.904,
+      "eval_steps_per_second": 11.375,
       "step": 520
     },
     {
-      "epoch": 0.39,
-      "learning_rate": 0.0001609720176730486,
-      "loss": 0.6781,
+      "epoch": 0.28,
+      "learning_rate": 0.00017207586933614332,
+      "loss": 0.9307,
       "step": 530
     },
     {
-      "epoch": 0.4,
-      "learning_rate": 0.0001602356406480118,
-      "loss": 0.7243,
+      "epoch": 0.28,
+      "learning_rate": 0.00017154899894625923,
+      "loss": 0.8867,
       "step": 540
     },
     {
-      "epoch": 0.41,
-      "learning_rate": 0.00015949926362297497,
-      "loss": 0.664,
+      "epoch": 0.29,
+      "learning_rate": 0.00017102212855637513,
+      "loss": 0.8658,
       "step": 550
     },
     {
-      "epoch": 0.41,
-      "learning_rate": 0.00015876288659793814,
-      "loss": 0.7103,
+      "epoch": 0.3,
+      "learning_rate": 0.00017049525816649106,
+      "loss": 0.8935,
       "step": 560
     },
     {
-      "epoch": 0.41,
-      "eval_accuracy": 0.7204419889502762,
-      "eval_loss": 0.7556896805763245,
-      "eval_runtime": 51.9028,
-      "eval_samples_per_second": 52.309,
-      "eval_steps_per_second": 6.551,
+      "epoch": 0.3,
+      "eval_accuracy": 0.6699157007376185,
+      "eval_loss": 0.8650425672531128,
+      "eval_runtime": 42.0967,
+      "eval_samples_per_second": 90.173,
+      "eval_steps_per_second": 11.284,
       "step": 560
     },
     {
-      "epoch": 0.42,
-      "learning_rate": 0.00015802650957290134,
-      "loss": 0.7232,
+      "epoch": 0.3,
+      "learning_rate": 0.00016996838777660696,
+      "loss": 0.958,
       "step": 570
     },
     {
-      "epoch": 0.43,
-      "learning_rate": 0.0001572901325478645,
-      "loss": 0.7658,
+      "epoch": 0.31,
+      "learning_rate": 0.00016944151738672287,
+      "loss": 0.8805,
       "step": 580
     },
     {
-      "epoch": 0.43,
-      "learning_rate": 0.0001565537555228277,
-      "loss": 0.7853,
+      "epoch": 0.31,
+      "learning_rate": 0.0001689146469968388,
+      "loss": 0.9066,
       "step": 590
     },
     {
-      "epoch": 0.44,
-      "learning_rate": 0.00015581737849779088,
-      "loss": 0.6149,
+      "epoch": 0.32,
+      "learning_rate": 0.0001683877766069547,
+      "loss": 0.8468,
       "step": 600
     },
     {
-      "epoch": 0.44,
-      "eval_accuracy": 0.7484346224677716,
-      "eval_loss": 0.6908882260322571,
-      "eval_runtime": 50.7606,
-      "eval_samples_per_second": 53.486,
-      "eval_steps_per_second": 6.698,
+      "epoch": 0.32,
+      "eval_accuracy": 0.7268177028451,
+      "eval_loss": 0.7727631330490112,
+      "eval_runtime": 41.8106,
+      "eval_samples_per_second": 90.79,
+      "eval_steps_per_second": 11.361,
       "step": 600
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 0.00015508100147275405,
-      "loss": 0.7379,
+      "epoch": 0.32,
+      "learning_rate": 0.0001678609062170706,
+      "loss": 0.797,
       "step": 610
     },
     {
-      "epoch": 0.46,
-      "learning_rate": 0.00015434462444771725,
-      "loss": 0.7758,
+      "epoch": 0.33,
+      "learning_rate": 0.0001673340358271865,
+      "loss": 0.8564,
       "step": 620
     },
     {
-      "epoch": 0.46,
-      "learning_rate": 0.00015360824742268042,
-      "loss": 0.6333,
+      "epoch": 0.33,
+      "learning_rate": 0.00016680716543730244,
+      "loss": 0.7341,
       "step": 630
     },
     {
-      "epoch": 0.47,
-      "learning_rate": 0.0001528718703976436,
-      "loss": 0.5998,
+      "epoch": 0.34,
+      "learning_rate": 0.00016628029504741834,
+      "loss": 0.829,
       "step": 640
     },
     {
-      "epoch": 0.47,
-      "eval_accuracy": 0.7322283609576428,
-      "eval_loss": 0.7079024314880371,
-      "eval_runtime": 50.4304,
-      "eval_samples_per_second": 53.837,
-      "eval_steps_per_second": 6.742,
+      "epoch": 0.34,
+      "eval_accuracy": 0.6978398314014752,
+      "eval_loss": 0.8096127510070801,
+      "eval_runtime": 42.0874,
+      "eval_samples_per_second": 90.193,
+      "eval_steps_per_second": 11.286,
       "step": 640
     },
     {
-      "epoch": 0.48,
-      "learning_rate": 0.00015213549337260679,
-      "loss": 0.6485,
+      "epoch": 0.34,
+      "learning_rate": 0.00016575342465753425,
+      "loss": 0.7701,
       "step": 650
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 0.00015139911634756996,
-      "loss": 0.8297,
+      "epoch": 0.35,
+      "learning_rate": 0.00016522655426765018,
+      "loss": 0.763,
       "step": 660
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 0.00015066273932253315,
-      "loss": 0.8158,
+      "epoch": 0.35,
+      "learning_rate": 0.00016469968387776608,
+      "loss": 0.7975,
       "step": 670
     },
     {
-      "epoch": 0.5,
-      "learning_rate": 0.00014992636229749633,
-      "loss": 0.8814,
+      "epoch": 0.36,
+      "learning_rate": 0.00016417281348788198,
+      "loss": 0.7611,
       "step": 680
     },
     {
-      "epoch": 0.5,
-      "eval_accuracy": 0.6744014732965009,
-      "eval_loss": 0.922738254070282,
-      "eval_runtime": 50.6437,
-      "eval_samples_per_second": 53.61,
-      "eval_steps_per_second": 6.714,
+      "epoch": 0.36,
+      "eval_accuracy": 0.7168071654373024,
+      "eval_loss": 0.774563193321228,
+      "eval_runtime": 41.9532,
+      "eval_samples_per_second": 90.482,
+      "eval_steps_per_second": 11.322,
       "step": 680
     },
     {
-      "epoch": 0.51,
-      "learning_rate": 0.0001491899852724595,
-      "loss": 0.764,
+      "epoch": 0.36,
+      "learning_rate": 0.0001636459430979979,
+      "loss": 0.8112,
       "step": 690
     },
     {
-      "epoch": 0.52,
-      "learning_rate": 0.0001484536082474227,
-      "loss": 0.6825,
+      "epoch": 0.37,
+      "learning_rate": 0.00016311907270811382,
+      "loss": 0.7897,
       "step": 700
     },
     {
-      "epoch": 0.52,
-      "learning_rate": 0.00014771723122238586,
-      "loss": 0.6629,
+      "epoch": 0.37,
+      "learning_rate": 0.00016259220231822972,
+      "loss": 0.8332,
       "step": 710
     },
     {
-      "epoch": 0.53,
-      "learning_rate": 0.00014698085419734906,
-      "loss": 0.7405,
+      "epoch": 0.38,
+      "learning_rate": 0.00016206533192834562,
+      "loss": 0.8072,
       "step": 720
     },
     {
-      "epoch": 0.53,
-      "eval_accuracy": 0.7454880294659301,
-      "eval_loss": 0.6923764944076538,
-      "eval_runtime": 50.452,
-      "eval_samples_per_second": 53.814,
-      "eval_steps_per_second": 6.739,
+      "epoch": 0.38,
+      "eval_accuracy": 0.7067966280295047,
+      "eval_loss": 0.802047848701477,
+      "eval_runtime": 41.8602,
+      "eval_samples_per_second": 90.683,
+      "eval_steps_per_second": 11.347,
       "step": 720
     },
     {
-      "epoch": 0.54,
-      "learning_rate": 0.00014624447717231223,
-      "loss": 0.8144,
+      "epoch": 0.38,
+      "learning_rate": 0.00016153846153846155,
+      "loss": 0.7557,
       "step": 730
     },
     {
-      "epoch": 0.54,
-      "learning_rate": 0.0001455081001472754,
-      "loss": 0.5245,
+      "epoch": 0.39,
+      "learning_rate": 0.00016101159114857746,
+      "loss": 0.8949,
       "step": 740
     },
     {
-      "epoch": 0.55,
-      "learning_rate": 0.0001447717231222386,
-      "loss": 0.7705,
+      "epoch": 0.4,
+      "learning_rate": 0.00016048472075869336,
+      "loss": 0.7729,
       "step": 750
     },
     {
-      "epoch": 0.56,
-      "learning_rate": 0.00014403534609720177,
-      "loss": 0.7463,
+      "epoch": 0.4,
+      "learning_rate": 0.0001599578503688093,
+      "loss": 0.8402,
       "step": 760
     },
     {
-      "epoch": 0.56,
-      "eval_accuracy": 0.7182320441988951,
-      "eval_loss": 0.7563807368278503,
-      "eval_runtime": 50.5699,
-      "eval_samples_per_second": 53.688,
-      "eval_steps_per_second": 6.723,
+      "epoch": 0.4,
+      "eval_accuracy": 0.6983667017913593,
+      "eval_loss": 0.8181519508361816,
+      "eval_runtime": 42.0502,
+      "eval_samples_per_second": 90.273,
+      "eval_steps_per_second": 11.296,
       "step": 760
     },
     {
-      "epoch": 0.57,
-      "learning_rate": 0.00014329896907216494,
-      "loss": 0.7648,
+      "epoch": 0.41,
+      "learning_rate": 0.0001594309799789252,
+      "loss": 0.869,
       "step": 770
     },
     {
-      "epoch": 0.57,
-      "learning_rate": 0.00014256259204712814,
-      "loss": 0.6192,
+      "epoch": 0.41,
+      "learning_rate": 0.0001589041095890411,
+      "loss": 0.8356,
       "step": 780
     },
     {
-      "epoch": 0.58,
-      "learning_rate": 0.0001418262150220913,
-      "loss": 0.872,
+      "epoch": 0.42,
+      "learning_rate": 0.000158377239199157,
+      "loss": 0.9184,
       "step": 790
     },
     {
-      "epoch": 0.59,
-      "learning_rate": 0.0001410898379970545,
-      "loss": 0.684,
+      "epoch": 0.42,
+      "learning_rate": 0.00015785036880927293,
+      "loss": 0.6367,
       "step": 800
     },
     {
-      "epoch": 0.59,
-      "eval_accuracy": 0.7337016574585635,
-      "eval_loss": 0.7226435542106628,
-      "eval_runtime": 50.2491,
-      "eval_samples_per_second": 54.031,
-      "eval_steps_per_second": 6.766,
+      "epoch": 0.42,
+      "eval_accuracy": 0.7397260273972602,
+      "eval_loss": 0.7236610054969788,
+      "eval_runtime": 41.5948,
+      "eval_samples_per_second": 91.261,
+      "eval_steps_per_second": 11.42,
       "step": 800
     },
     {
-      "epoch": 0.6,
-      "learning_rate": 0.00014035346097201768,
-      "loss": 0.7713,
+      "epoch": 0.43,
+      "learning_rate": 0.00015732349841938883,
+      "loss": 0.7656,
       "step": 810
     },
     {
-      "epoch": 0.6,
-      "learning_rate": 0.00013961708394698085,
-      "loss": 0.6673,
+      "epoch": 0.43,
+      "learning_rate": 0.00015679662802950474,
+      "loss": 0.8507,
       "step": 820
     },
     {
-      "epoch": 0.61,
-      "learning_rate": 0.00013888070692194405,
-      "loss": 0.6792,
+      "epoch": 0.44,
+      "learning_rate": 0.00015626975763962067,
+      "loss": 0.7213,
       "step": 830
     },
     {
-      "epoch": 0.62,
-      "learning_rate": 0.00013814432989690722,
-      "loss": 0.6252,
+      "epoch": 0.44,
+      "learning_rate": 0.00015574288724973657,
+      "loss": 0.7094,
       "step": 840
     },
     {
-      "epoch": 0.62,
-      "eval_accuracy": 0.7524861878453039,
-      "eval_loss": 0.7024006247520447,
-      "eval_runtime": 50.1454,
-      "eval_samples_per_second": 54.143,
-      "eval_steps_per_second": 6.78,
+      "epoch": 0.44,
+      "eval_accuracy": 0.7336670179135932,
+      "eval_loss": 0.75114905834198,
+      "eval_runtime": 41.1659,
+      "eval_samples_per_second": 92.212,
+      "eval_steps_per_second": 11.539,
       "step": 840
     },
     {
-      "epoch": 0.63,
-      "learning_rate": 0.0001374079528718704,
-      "loss": 0.7425,
+      "epoch": 0.45,
+      "learning_rate": 0.00015521601685985248,
+      "loss": 0.7631,
       "step": 850
     },
     {
-      "epoch": 0.63,
-      "learning_rate": 0.0001366715758468336,
-      "loss": 0.7285,
+      "epoch": 0.45,
+      "learning_rate": 0.0001546891464699684,
+      "loss": 0.7045,
       "step": 860
     },
     {
-      "epoch": 0.64,
-      "learning_rate": 0.00013593519882179676,
-      "loss": 0.7498,
+      "epoch": 0.46,
+      "learning_rate": 0.0001541622760800843,
+      "loss": 0.7582,
       "step": 870
     },
     {
-      "epoch": 0.65,
-      "learning_rate": 0.00013519882179675996,
-      "loss": 0.754,
+      "epoch": 0.46,
+      "learning_rate": 0.0001536354056902002,
+      "loss": 0.8905,
       "step": 880
     },
     {
-      "epoch": 0.65,
-      "eval_accuracy": 0.7274401473296501,
-      "eval_loss": 0.6939438581466675,
-      "eval_runtime": 51.2872,
-      "eval_samples_per_second": 52.937,
-      "eval_steps_per_second": 6.629,
+      "epoch": 0.46,
+      "eval_accuracy": 0.7489462592202318,
+      "eval_loss": 0.6889460682868958,
+      "eval_runtime": 41.1832,
+      "eval_samples_per_second": 92.173,
+      "eval_steps_per_second": 11.534,
       "step": 880
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 0.00013446244477172313,
-      "loss": 0.6756,
+      "epoch": 0.47,
+      "learning_rate": 0.00015310853530031612,
+      "loss": 0.7429,
       "step": 890
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 0.0001337260677466863,
-      "loss": 0.636,
+      "epoch": 0.47,
+      "learning_rate": 0.00015258166491043205,
+      "loss": 0.7355,
       "step": 900
     },
     {
-      "epoch": 0.67,
-      "learning_rate": 0.0001329896907216495,
-      "loss": 0.5823,
+      "epoch": 0.48,
+      "learning_rate": 0.00015205479452054795,
+      "loss": 0.7525,
       "step": 910
     },
     {
-      "epoch": 0.68,
-      "learning_rate": 0.00013225331369661267,
-      "loss": 0.6416,
+      "epoch": 0.48,
+      "learning_rate": 0.00015152792413066385,
+      "loss": 0.8386,
       "step": 920
     },
     {
-      "epoch": 0.68,
-      "eval_accuracy": 0.7565377532228361,
-      "eval_loss": 0.6688955426216125,
-      "eval_runtime": 51.1119,
-      "eval_samples_per_second": 53.119,
-      "eval_steps_per_second": 6.652,
+      "epoch": 0.48,
+      "eval_accuracy": 0.7178609062170705,
+      "eval_loss": 0.7636825442314148,
+      "eval_runtime": 41.2051,
+      "eval_samples_per_second": 92.124,
+      "eval_steps_per_second": 11.528,
       "step": 920
     },
     {
-      "epoch": 0.68,
-      "learning_rate": 0.00013151693667157587,
-      "loss": 0.6125,
+      "epoch": 0.49,
+      "learning_rate": 0.00015100105374077978,
+      "loss": 0.7451,
       "step": 930
     },
     {
-      "epoch": 0.69,
-      "learning_rate": 0.00013078055964653904,
-      "loss": 0.687,
+      "epoch": 0.5,
+      "learning_rate": 0.0001504741833508957,
+      "loss": 0.6861,
       "step": 940
     },
     {
-      "epoch": 0.7,
-      "learning_rate": 0.0001300441826215022,
-      "loss": 0.7283,
+      "epoch": 0.5,
+      "learning_rate": 0.0001499473129610116,
+      "loss": 0.7938,
       "step": 950
     },
     {
-      "epoch": 0.71,
-      "learning_rate": 0.0001293078055964654,
-      "loss": 0.7123,
+      "epoch": 0.51,
+      "learning_rate": 0.0001494204425711275,
+      "loss": 0.7564,
       "step": 960
     },
     {
-      "epoch": 0.71,
-      "eval_accuracy": 0.7558011049723757,
-      "eval_loss": 0.6540634632110596,
-      "eval_runtime": 50.4305,
-      "eval_samples_per_second": 53.836,
-      "eval_steps_per_second": 6.742,
+      "epoch": 0.51,
+      "eval_accuracy": 0.7365648050579557,
+      "eval_loss": 0.7086778283119202,
+      "eval_runtime": 41.133,
+      "eval_samples_per_second": 92.286,
+      "eval_steps_per_second": 11.548,
       "step": 960
     },
     {
-      "epoch": 0.71,
-      "learning_rate": 0.00012857142857142858,
-      "loss": 0.6579,
+      "epoch": 0.51,
+      "learning_rate": 0.00014889357218124342,
+      "loss": 0.7699,
       "step": 970
     },
     {
-      "epoch": 0.72,
-      "learning_rate": 0.00012783505154639175,
-      "loss": 0.6744,
+      "epoch": 0.52,
+      "learning_rate": 0.00014836670179135933,
+      "loss": 0.754,
       "step": 980
     },
     {
-      "epoch": 0.73,
-      "learning_rate": 0.00012709867452135494,
-      "loss": 0.619,
+      "epoch": 0.52,
+      "learning_rate": 0.00014783983140147523,
+      "loss": 0.7126,
       "step": 990
     },
     {
-      "epoch": 0.74,
-      "learning_rate": 0.00012636229749631812,
-      "loss": 0.6599,
+      "epoch": 0.53,
+      "learning_rate": 0.00014731296101159116,
+      "loss": 0.7927,
       "step": 1000
     },
     {
-      "epoch": 0.74,
-      "eval_accuracy": 0.7683241252302025,
-      "eval_loss": 0.6279534101486206,
-      "eval_runtime": 50.5314,
-      "eval_samples_per_second": 53.729,
-      "eval_steps_per_second": 6.728,
+      "epoch": 0.53,
+      "eval_accuracy": 0.7307692307692307,
+      "eval_loss": 0.7288175225257874,
+      "eval_runtime": 40.9893,
+      "eval_samples_per_second": 92.609,
+      "eval_steps_per_second": 11.588,
       "step": 1000
     },
     {
-      "epoch": 0.74,
-      "learning_rate": 0.0001256259204712813,
-      "loss": 0.5407,
+      "epoch": 0.53,
+      "learning_rate": 0.00014678609062170707,
+      "loss": 0.8823,
       "step": 1010
     },
     {
-      "epoch": 0.75,
-      "learning_rate": 0.00012488954344624448,
-      "loss": 0.7339,
+      "epoch": 0.54,
+      "learning_rate": 0.00014625922023182297,
+      "loss": 0.6232,
       "step": 1020
     },
     {
-      "epoch": 0.76,
-      "learning_rate": 0.00012415316642120765,
-      "loss": 0.65,
+      "epoch": 0.54,
+      "learning_rate": 0.0001457323498419389,
+      "loss": 0.7342,
       "step": 1030
     },
     {
-      "epoch": 0.77,
-      "learning_rate": 0.00012341678939617085,
-      "loss": 0.7664,
+      "epoch": 0.55,
+      "learning_rate": 0.0001452054794520548,
+      "loss": 0.7456,
       "step": 1040
     },
     {
-      "epoch": 0.77,
-      "eval_accuracy": 0.7521178637200736,
-      "eval_loss": 0.6760364770889282,
-      "eval_runtime": 51.3273,
-      "eval_samples_per_second": 52.896,
-      "eval_steps_per_second": 6.624,
+      "epoch": 0.55,
+      "eval_accuracy": 0.7494731296101159,
+      "eval_loss": 0.7018606662750244,
+      "eval_runtime": 41.1744,
+      "eval_samples_per_second": 92.193,
+      "eval_steps_per_second": 11.536,
       "step": 1040
     },
     {
-      "epoch": 0.77,
-      "learning_rate": 0.00012268041237113402,
-      "loss": 0.667,
+      "epoch": 0.55,
+      "learning_rate": 0.00014467860906217073,
+      "loss": 0.7618,
       "step": 1050
     },
     {
-      "epoch": 0.78,
-      "learning_rate": 0.00012194403534609722,
-      "loss": 0.6393,
+      "epoch": 0.56,
+      "learning_rate": 0.0001441517386722866,
+      "loss": 0.6762,
       "step": 1060
     },
     {
-      "epoch": 0.79,
-      "learning_rate": 0.00012120765832106038,
-      "loss": 0.7258,
+      "epoch": 0.56,
+      "learning_rate": 0.00014362486828240254,
+      "loss": 0.7244,
       "step": 1070
     },
     {
-      "epoch": 0.8,
-      "learning_rate": 0.00012047128129602356,
-      "loss": 0.6898,
+      "epoch": 0.57,
+      "learning_rate": 0.00014309799789251844,
+      "loss": 0.6604,
       "step": 1080
     },
     {
-      "epoch": 0.8,
-      "eval_accuracy": 0.7815837937384899,
-      "eval_loss": 0.6094613075256348,
-      "eval_runtime": 50.3809,
-      "eval_samples_per_second": 53.89,
-      "eval_steps_per_second": 6.749,
+      "epoch": 0.57,
+      "eval_accuracy": 0.7386722866174921,
+      "eval_loss": 0.737483561038971,
+      "eval_runtime": 41.1001,
+      "eval_samples_per_second": 92.36,
+      "eval_steps_per_second": 11.557,
       "step": 1080
     },
     {
-      "epoch": 0.8,
-      "learning_rate": 0.00011973490427098675,
-      "loss": 0.6158,
+      "epoch": 0.57,
+      "learning_rate": 0.00014257112750263435,
+      "loss": 0.8146,
       "step": 1090
     },
     {
-      "epoch": 0.81,
-      "learning_rate": 0.00011899852724594993,
-      "loss": 0.5764,
+      "epoch": 0.58,
+      "learning_rate": 0.00014204425711275028,
+      "loss": 0.7946,
       "step": 1100
     },
     {
-      "epoch": 0.82,
-      "learning_rate": 0.00011826215022091312,
-      "loss": 0.6926,
+      "epoch": 0.58,
+      "learning_rate": 0.00014151738672286618,
+      "loss": 0.8021,
       "step": 1110
     },
     {
-      "epoch": 0.82,
-      "learning_rate": 0.0001175257731958763,
-      "loss": 0.693,
+      "epoch": 0.59,
+      "learning_rate": 0.0001409905163329821,
+      "loss": 0.6272,
       "step": 1120
     },
     {
-      "epoch": 0.82,
-      "eval_accuracy": 0.7543278084714549,
-      "eval_loss": 0.6533992290496826,
-      "eval_runtime": 51.5963,
-      "eval_samples_per_second": 52.62,
-      "eval_steps_per_second": 6.59,
+      "epoch": 0.59,
+      "eval_accuracy": 0.755795574288725,
+      "eval_loss": 0.6697063446044922,
+      "eval_runtime": 41.0445,
+      "eval_samples_per_second": 92.485,
+      "eval_steps_per_second": 11.573,
       "step": 1120
     },
     {
-      "epoch": 0.83,
-      "learning_rate": 0.00011678939617083948,
-      "loss": 0.7522,
+      "epoch": 0.6,
+      "learning_rate": 0.000140463645943098,
+      "loss": 0.7822,
       "step": 1130
     },
     {
-      "epoch": 0.84,
-      "learning_rate": 0.00011605301914580267,
-      "loss": 0.451,
+      "epoch": 0.6,
+      "learning_rate": 0.00013993677555321392,
+      "loss": 0.5635,
       "step": 1140
     },
     {
-      "epoch": 0.85,
-      "learning_rate": 0.00011531664212076583,
-      "loss": 0.564,
+      "epoch": 0.61,
+      "learning_rate": 0.00013940990516332982,
+      "loss": 0.8475,
       "step": 1150
     },
     {
-      "epoch": 0.85,
-      "learning_rate": 0.00011458026509572901,
-      "loss": 0.6738,
+      "epoch": 0.61,
+      "learning_rate": 0.00013888303477344572,
+      "loss": 0.6899,
       "step": 1160
     },
     {
-      "epoch": 0.85,
-      "eval_accuracy": 0.7826887661141805,
-      "eval_loss": 0.6036040186882019,
-      "eval_runtime": 50.9403,
-      "eval_samples_per_second": 53.298,
-      "eval_steps_per_second": 6.674,
+      "epoch": 0.61,
+      "eval_accuracy": 0.7510537407797682,
+      "eval_loss": 0.6927923560142517,
+      "eval_runtime": 41.0153,
+      "eval_samples_per_second": 92.551,
+      "eval_steps_per_second": 11.581,
       "step": 1160
     },
     {
-      "epoch": 0.86,
-      "learning_rate": 0.0001138438880706922,
-      "loss": 0.6272,
+      "epoch": 0.62,
+      "learning_rate": 0.00013835616438356166,
+      "loss": 0.6789,
       "step": 1170
     },
     {
-      "epoch": 0.87,
-      "learning_rate": 0.00011310751104565538,
-      "loss": 0.5538,
+      "epoch": 0.62,
+      "learning_rate": 0.00013782929399367756,
+      "loss": 0.7409,
       "step": 1180
     },
     {
-      "epoch": 0.88,
-      "learning_rate": 0.00011237113402061856,
-      "loss": 0.555,
+      "epoch": 0.63,
+      "learning_rate": 0.0001373024236037935,
+      "loss": 0.6629,
       "step": 1190
     },
     {
-      "epoch": 0.88,
-      "learning_rate": 0.00011163475699558175,
-      "loss": 0.6056,
+      "epoch": 0.63,
+      "learning_rate": 0.0001367755532139094,
+      "loss": 0.7612,
       "step": 1200
     },
     {
-      "epoch": 0.88,
-      "eval_accuracy": 0.7874769797421731,
-      "eval_loss": 0.584153950214386,
-      "eval_runtime": 51.1472,
-      "eval_samples_per_second": 53.082,
-      "eval_steps_per_second": 6.647,
+      "epoch": 0.63,
+      "eval_accuracy": 0.7726554267650158,
+      "eval_loss": 0.6404809355735779,
+      "eval_runtime": 40.9053,
+      "eval_samples_per_second": 92.8,
+      "eval_steps_per_second": 11.612,
       "step": 1200
     },
     {
-      "epoch": 0.89,
-      "learning_rate": 0.00011089837997054493,
-      "loss": 0.5609,
+      "epoch": 0.64,
+      "learning_rate": 0.0001362486828240253,
+      "loss": 0.727,
       "step": 1210
     },
     {
-      "epoch": 0.9,
-      "learning_rate": 0.00011016200294550812,
-      "loss": 0.6141,
+      "epoch": 0.64,
+      "learning_rate": 0.00013572181243414123,
+      "loss": 0.6921,
       "step": 1220
     },
     {
-      "epoch": 0.91,
-      "learning_rate": 0.0001094256259204713,
-      "loss": 0.6212,
+      "epoch": 0.65,
+      "learning_rate": 0.0001351949420442571,
+      "loss": 0.5298,
       "step": 1230
     },
     {
-      "epoch": 0.91,
-      "learning_rate": 0.00010868924889543446,
-      "loss": 0.746,
+      "epoch": 0.65,
+      "learning_rate": 0.00013466807165437303,
+      "loss": 0.5398,
       "step": 1240
     },
     {
-      "epoch": 0.91,
-      "eval_accuracy": 0.7675874769797422,
-      "eval_loss": 0.6469867825508118,
-      "eval_runtime": 50.9954,
-      "eval_samples_per_second": 53.24,
-      "eval_steps_per_second": 6.667,
+      "epoch": 0.65,
+      "eval_accuracy": 0.7673867228661749,
+      "eval_loss": 0.6329030394554138,
+      "eval_runtime": 41.1422,
+      "eval_samples_per_second": 92.265,
+      "eval_steps_per_second": 11.545,
       "step": 1240
     },
     {
-      "epoch": 0.92,
-      "learning_rate": 0.00010795287187039764,
-      "loss": 0.5153,
+      "epoch": 0.66,
+      "learning_rate": 0.00013414120126448894,
+      "loss": 0.6673,
       "step": 1250
     },
     {
-      "epoch": 0.93,
-      "learning_rate": 0.00010721649484536083,
-      "loss": 0.6638,
+      "epoch": 0.66,
+      "learning_rate": 0.00013361433087460484,
+      "loss": 0.682,
       "step": 1260
     },
     {
-      "epoch": 0.94,
-      "learning_rate": 0.00010648011782032401,
-      "loss": 0.6719,
+      "epoch": 0.67,
+      "learning_rate": 0.00013308746048472077,
+      "loss": 0.6457,
       "step": 1270
     },
     {
-      "epoch": 0.94,
-      "learning_rate": 0.0001057437407952872,
-      "loss": 0.7612,
+      "epoch": 0.67,
+      "learning_rate": 0.00013256059009483667,
+      "loss": 0.5942,
       "step": 1280
     },
     {
-      "epoch": 0.94,
-      "eval_accuracy": 0.7797421731123388,
-      "eval_loss": 0.613595724105835,
-      "eval_runtime": 51.0708,
-      "eval_samples_per_second": 53.161,
-      "eval_steps_per_second": 6.657,
+      "epoch": 0.67,
+      "eval_accuracy": 0.7481559536354057,
+      "eval_loss": 0.6839689016342163,
+      "eval_runtime": 41.0621,
+      "eval_samples_per_second": 92.445,
+      "eval_steps_per_second": 11.568,
       "step": 1280
     },
     {
-      "epoch": 0.95,
-      "learning_rate": 0.00010500736377025038,
-      "loss": 0.6873,
+      "epoch": 0.68,
+      "learning_rate": 0.0001320337197049526,
+      "loss": 0.5086,
       "step": 1290
     },
     {
-      "epoch": 0.96,
-      "learning_rate": 0.00010427098674521356,
-      "loss": 0.5534,
+      "epoch": 0.68,
+      "learning_rate": 0.00013150684931506848,
+      "loss": 0.6658,
       "step": 1300
     },
     {
-      "epoch": 0.96,
-      "learning_rate": 0.00010353460972017675,
-      "loss": 0.3849,
+      "epoch": 0.69,
+      "learning_rate": 0.0001309799789251844,
+      "loss": 0.7157,
       "step": 1310
     },
     {
-      "epoch": 0.97,
-      "learning_rate": 0.0001027982326951399,
-      "loss": 0.5939,
+      "epoch": 0.7,
+      "learning_rate": 0.00013045310853530031,
+      "loss": 0.5924,
       "step": 1320
     },
     {
-      "epoch": 0.97,
-      "eval_accuracy": 0.7952117863720074,
-      "eval_loss": 0.5709212422370911,
-      "eval_runtime": 52.4056,
-      "eval_samples_per_second": 51.807,
-      "eval_steps_per_second": 6.488,
+      "epoch": 0.7,
+      "eval_accuracy": 0.7795047418335089,
+      "eval_loss": 0.6245933175086975,
+      "eval_runtime": 40.9868,
+      "eval_samples_per_second": 92.615,
+      "eval_steps_per_second": 11.589,
       "step": 1320
     },
     {
-      "epoch": 0.98,
-      "learning_rate": 0.00010206185567010309,
-      "loss": 0.5831,
+      "epoch": 0.7,
+      "learning_rate": 0.00012992623814541622,
+      "loss": 0.6697,
       "step": 1330
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 0.00010132547864506627,
-      "loss": 0.6712,
+      "epoch": 0.71,
+      "learning_rate": 0.00012939936775553215,
+      "loss": 0.8167,
       "step": 1340
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 0.00010058910162002946,
-      "loss": 0.5368,
+      "epoch": 0.71,
+      "learning_rate": 0.00012887249736564805,
+      "loss": 0.641,
       "step": 1350
     },
     {
-      "epoch": 1.0,
-      "learning_rate": 9.985272459499264e-05,
-      "loss": 0.5586,
+      "epoch": 0.72,
+      "learning_rate": 0.00012834562697576398,
+      "loss": 0.7035,
       "step": 1360
     },
     {
-      "epoch": 1.0,
-      "eval_accuracy": 0.7701657458563536,
-      "eval_loss": 0.6024866104125977,
-      "eval_runtime": 50.9747,
-      "eval_samples_per_second": 53.262,
-      "eval_steps_per_second": 6.67,
+      "epoch": 0.72,
+      "eval_accuracy": 0.7710748155953635,
+      "eval_loss": 0.6422901153564453,
+      "eval_runtime": 41.0116,
+      "eval_samples_per_second": 92.559,
+      "eval_steps_per_second": 11.582,
       "step": 1360
     },
     {
-      "epoch": 1.01,
-      "learning_rate": 9.911634756995583e-05,
-      "loss": 0.4372,
+      "epoch": 0.72,
+      "learning_rate": 0.00012781875658587989,
+      "loss": 0.6054,
       "step": 1370
     },
     {
-      "epoch": 1.02,
-      "learning_rate": 9.8379970544919e-05,
-      "loss": 0.4076,
+      "epoch": 0.73,
+      "learning_rate": 0.0001272918861959958,
+      "loss": 0.5799,
       "step": 1380
     },
     {
-      "epoch": 1.02,
-      "learning_rate": 9.764359351988218e-05,
-      "loss": 0.3933,
+      "epoch": 0.73,
+      "learning_rate": 0.00012676501580611172,
+      "loss": 0.6575,
       "step": 1390
     },
     {
-      "epoch": 1.03,
-      "learning_rate": 9.690721649484537e-05,
-      "loss": 0.459,
+      "epoch": 0.74,
+      "learning_rate": 0.0001262381454162276,
+      "loss": 0.7114,
       "step": 1400
     },
     {
-      "epoch": 1.03,
-      "eval_accuracy": 0.7675874769797422,
-      "eval_loss": 0.6382798552513123,
-      "eval_runtime": 51.4385,
-      "eval_samples_per_second": 52.782,
-      "eval_steps_per_second": 6.61,
+      "epoch": 0.74,
+      "eval_accuracy": 0.7755532139093783,
+      "eval_loss": 0.6289725303649902,
+      "eval_runtime": 41.0218,
+      "eval_samples_per_second": 92.536,
+      "eval_steps_per_second": 11.579,
       "step": 1400
     },
     {
-      "epoch": 1.04,
-      "learning_rate": 9.617083946980855e-05,
-      "loss": 0.3769,
+      "epoch": 0.74,
+      "learning_rate": 0.00012571127502634353,
+      "loss": 0.6659,
       "step": 1410
     },
     {
-      "epoch": 1.05,
-      "learning_rate": 9.543446244477173e-05,
-      "loss": 0.411,
+      "epoch": 0.75,
+      "learning_rate": 0.00012518440463645943,
+      "loss": 0.5519,
       "step": 1420
     },
     {
-      "epoch": 1.05,
-      "learning_rate": 9.46980854197349e-05,
-      "loss": 0.4444,
+      "epoch": 0.75,
+      "learning_rate": 0.00012465753424657536,
+      "loss": 0.7071,
       "step": 1430
     },
     {
-      "epoch": 1.06,
-      "learning_rate": 9.396170839469809e-05,
-      "loss": 0.3771,
+      "epoch": 0.76,
+      "learning_rate": 0.00012413066385669126,
+      "loss": 0.477,
       "step": 1440
     },
     {
-      "epoch": 1.06,
-      "eval_accuracy": 0.7856353591160221,
-      "eval_loss": 0.5789190530776978,
-      "eval_runtime": 50.7253,
-      "eval_samples_per_second": 53.524,
-      "eval_steps_per_second": 6.703,
+      "epoch": 0.76,
+      "eval_accuracy": 0.7895152792413066,
+      "eval_loss": 0.5919615030288696,
+      "eval_runtime": 41.0316,
+      "eval_samples_per_second": 92.514,
+      "eval_steps_per_second": 11.576,
       "step": 1440
     },
     {
-      "epoch": 1.07,
-      "learning_rate": 9.322533136966127e-05,
-      "loss": 0.3943,
+      "epoch": 0.76,
+      "learning_rate": 0.00012360379346680717,
+      "loss": 0.5786,
       "step": 1450
     },
     {
-      "epoch": 1.08,
-      "learning_rate": 9.248895434462446e-05,
-      "loss": 0.5164,
+      "epoch": 0.77,
+      "learning_rate": 0.0001230769230769231,
+      "loss": 0.624,
       "step": 1460
     },
     {
-      "epoch": 1.08,
-      "learning_rate": 9.175257731958763e-05,
-      "loss": 0.3544,
+      "epoch": 0.77,
+      "learning_rate": 0.00012255005268703897,
+      "loss": 0.5879,
       "step": 1470
     },
     {
-      "epoch": 1.09,
-      "learning_rate": 9.101620029455081e-05,
-      "loss": 0.4272,
+      "epoch": 0.78,
+      "learning_rate": 0.0001220231822971549,
+      "loss": 0.6203,
       "step": 1480
     },
     {
-      "epoch": 1.09,
-      "eval_accuracy": 0.8044198895027624,
-      "eval_loss": 0.5364444851875305,
-      "eval_runtime": 52.5287,
-      "eval_samples_per_second": 51.686,
-      "eval_steps_per_second": 6.473,
+      "epoch": 0.78,
+      "eval_accuracy": 0.7881981032665965,
+      "eval_loss": 0.6011049747467041,
+      "eval_runtime": 41.3051,
+      "eval_samples_per_second": 91.902,
+      "eval_steps_per_second": 11.5,
       "step": 1480
     },
     {
-      "epoch": 1.1,
-      "learning_rate": 9.0279823269514e-05,
-      "loss": 0.3962,
+      "epoch": 0.79,
+      "learning_rate": 0.00012149631190727082,
+      "loss": 0.6142,
       "step": 1490
     },
     {
-      "epoch": 1.1,
-      "learning_rate": 8.954344624447718e-05,
-      "loss": 0.434,
+      "epoch": 0.79,
+      "learning_rate": 0.00012096944151738674,
+      "loss": 0.7426,
       "step": 1500
     },
     {
-      "epoch": 1.11,
-      "learning_rate": 8.880706921944035e-05,
-      "loss": 0.4332,
+      "epoch": 0.8,
+      "learning_rate": 0.00012044257112750264,
+      "loss": 0.5851,
       "step": 1510
     },
     {
-      "epoch": 1.12,
-      "learning_rate": 8.807069219440354e-05,
-      "loss": 0.3229,
+      "epoch": 0.8,
+      "learning_rate": 0.00011991570073761856,
+      "loss": 0.4557,
       "step": 1520
     },
     {
-      "epoch": 1.12,
-      "eval_accuracy": 0.8044198895027624,
-      "eval_loss": 0.5452917814254761,
-      "eval_runtime": 51.1816,
-      "eval_samples_per_second": 53.046,
-      "eval_steps_per_second": 6.643,
+      "epoch": 0.8,
+      "eval_accuracy": 0.773972602739726,
+      "eval_loss": 0.6244290471076965,
+      "eval_runtime": 41.0912,
+      "eval_samples_per_second": 92.38,
+      "eval_steps_per_second": 11.56,
       "step": 1520
     },
     {
-      "epoch": 1.13,
-      "learning_rate": 8.733431516936672e-05,
-      "loss": 0.4162,
+      "epoch": 0.81,
+      "learning_rate": 0.00011938883034773448,
+      "loss": 0.7113,
       "step": 1530
     },
     {
-      "epoch": 1.13,
-      "learning_rate": 8.65979381443299e-05,
-      "loss": 0.3704,
+      "epoch": 0.81,
+      "learning_rate": 0.00011886195995785037,
+      "loss": 0.4499,
       "step": 1540
     },
     {
-      "epoch": 1.14,
-      "learning_rate": 8.586156111929308e-05,
-      "loss": 0.6147,
+      "epoch": 0.82,
+      "learning_rate": 0.00011833508956796628,
+      "loss": 0.7979,
       "step": 1550
     },
     {
-      "epoch": 1.15,
-      "learning_rate": 8.512518409425626e-05,
-      "loss": 0.3736,
+      "epoch": 0.82,
+      "learning_rate": 0.0001178082191780822,
+      "loss": 0.6389,
       "step": 1560
     },
     {
-      "epoch": 1.15,
-      "eval_accuracy": 0.8092081031307551,
-      "eval_loss": 0.5403669476509094,
-      "eval_runtime": 50.5773,
-      "eval_samples_per_second": 53.68,
-      "eval_steps_per_second": 6.722,
+      "epoch": 0.82,
+      "eval_accuracy": 0.7968914646996839,
+      "eval_loss": 0.5629897117614746,
+      "eval_runtime": 41.1832,
+      "eval_samples_per_second": 92.174,
+      "eval_steps_per_second": 11.534,
       "step": 1560
     },
     {
-      "epoch": 1.16,
-      "learning_rate": 8.438880706921944e-05,
-      "loss": 0.4865,
+      "epoch": 0.83,
+      "learning_rate": 0.00011728134878819812,
+      "loss": 0.5088,
       "step": 1570
     },
     {
-      "epoch": 1.16,
-      "learning_rate": 8.365243004418263e-05,
-      "loss": 0.3929,
+      "epoch": 0.83,
+      "learning_rate": 0.00011675447839831402,
+      "loss": 0.5226,
       "step": 1580
     },
     {
-      "epoch": 1.17,
-      "learning_rate": 8.291605301914581e-05,
-      "loss": 0.3426,
+      "epoch": 0.84,
+      "learning_rate": 0.00011622760800842994,
+      "loss": 0.7143,
       "step": 1590
     },
     {
-      "epoch": 1.18,
-      "learning_rate": 8.217967599410898e-05,
-      "loss": 0.436,
+      "epoch": 0.84,
+      "learning_rate": 0.00011570073761854585,
+      "loss": 0.6855,
       "step": 1600
     },
     {
-      "epoch": 1.18,
-      "eval_accuracy": 0.8081031307550645,
-      "eval_loss": 0.5359194278717041,
-      "eval_runtime": 51.3206,
-      "eval_samples_per_second": 52.903,
-      "eval_steps_per_second": 6.625,
+      "epoch": 0.84,
+      "eval_accuracy": 0.7966280295047419,
+      "eval_loss": 0.570618748664856,
+      "eval_runtime": 41.1893,
+      "eval_samples_per_second": 92.16,
+      "eval_steps_per_second": 11.532,
       "step": 1600
     },
     {
-      "epoch": 1.19,
-      "learning_rate": 8.144329896907217e-05,
-      "loss": 0.4115,
+      "epoch": 0.85,
+      "learning_rate": 0.00011517386722866174,
+      "loss": 0.5177,
       "step": 1610
     },
     {
-      "epoch": 1.19,
-      "learning_rate": 8.070692194403535e-05,
-      "loss": 0.3784,
+      "epoch": 0.85,
+      "learning_rate": 0.00011464699683877766,
+      "loss": 0.6428,
       "step": 1620
     },
     {
-      "epoch": 1.2,
-      "learning_rate": 7.997054491899854e-05,
-      "loss": 0.3383,
+      "epoch": 0.86,
+      "learning_rate": 0.00011412012644889358,
+      "loss": 0.6207,
       "step": 1630
     },
     {
-      "epoch": 1.21,
-      "learning_rate": 7.923416789396171e-05,
-      "loss": 0.3721,
+      "epoch": 0.86,
+      "learning_rate": 0.00011359325605900948,
+      "loss": 0.5935,
       "step": 1640
     },
     {
-      "epoch": 1.21,
-      "eval_accuracy": 0.7815837937384899,
-      "eval_loss": 0.6128984689712524,
-      "eval_runtime": 51.0569,
-      "eval_samples_per_second": 53.176,
-      "eval_steps_per_second": 6.659,
+      "epoch": 0.86,
+      "eval_accuracy": 0.8071654373024236,
+      "eval_loss": 0.553897500038147,
+      "eval_runtime": 41.1107,
+      "eval_samples_per_second": 92.336,
+      "eval_steps_per_second": 11.554,
       "step": 1640
     },
     {
-      "epoch": 1.22,
-      "learning_rate": 7.849779086892489e-05,
-      "loss": 0.4844,
+      "epoch": 0.87,
+      "learning_rate": 0.0001130663856691254,
+      "loss": 0.5959,
       "step": 1650
     },
     {
-      "epoch": 1.22,
-      "learning_rate": 7.776141384388808e-05,
-      "loss": 0.4011,
+      "epoch": 0.87,
+      "learning_rate": 0.00011253951527924131,
+      "loss": 0.6435,
       "step": 1660
     },
     {
-      "epoch": 1.23,
-      "learning_rate": 7.702503681885126e-05,
-      "loss": 0.4711,
+      "epoch": 0.88,
+      "learning_rate": 0.00011201264488935723,
+      "loss": 0.6994,
       "step": 1670
     },
     {
-      "epoch": 1.24,
-      "learning_rate": 7.628865979381443e-05,
-      "loss": 0.3848,
+      "epoch": 0.89,
+      "learning_rate": 0.00011148577449947313,
+      "loss": 0.6779,
       "step": 1680
     },
     {
-      "epoch": 1.24,
-      "eval_accuracy": 0.7985267034990792,
-      "eval_loss": 0.5785150527954102,
-      "eval_runtime": 51.7772,
-      "eval_samples_per_second": 52.436,
-      "eval_steps_per_second": 6.567,
+      "epoch": 0.89,
+      "eval_accuracy": 0.7829293993677555,
+      "eval_loss": 0.5908519625663757,
+      "eval_runtime": 40.8752,
+      "eval_samples_per_second": 92.868,
+      "eval_steps_per_second": 11.621,
       "step": 1680
     },
     {
-      "epoch": 1.24,
-      "learning_rate": 7.555228276877762e-05,
-      "loss": 0.3771,
+      "epoch": 0.89,
+      "learning_rate": 0.00011095890410958905,
+      "loss": 0.6209,
       "step": 1690
     },
     {
-      "epoch": 1.25,
-      "learning_rate": 7.48159057437408e-05,
-      "loss": 0.5486,
+      "epoch": 0.9,
+      "learning_rate": 0.00011043203371970497,
+      "loss": 0.4842,
       "step": 1700
     },
     {
-      "epoch": 1.26,
-      "learning_rate": 7.407952871870398e-05,
-      "loss": 0.3795,
+      "epoch": 0.9,
+      "learning_rate": 0.00010990516332982086,
+      "loss": 0.6179,
       "step": 1710
     },
     {
-      "epoch": 1.27,
-      "learning_rate": 7.334315169366715e-05,
-      "loss": 0.393,
+      "epoch": 0.91,
+      "learning_rate": 0.00010937829293993678,
+      "loss": 0.5032,
       "step": 1720
     },
     {
-      "epoch": 1.27,
-      "eval_accuracy": 0.8047882136279927,
-      "eval_loss": 0.5534458756446838,
-      "eval_runtime": 50.9572,
-      "eval_samples_per_second": 53.28,
-      "eval_steps_per_second": 6.672,
+      "epoch": 0.91,
+      "eval_accuracy": 0.8174394099051633,
+      "eval_loss": 0.5368214249610901,
+      "eval_runtime": 40.9272,
+      "eval_samples_per_second": 92.75,
+      "eval_steps_per_second": 11.606,
       "step": 1720
     },
     {
-      "epoch": 1.27,
-      "learning_rate": 7.260677466863034e-05,
-      "loss": 0.4113,
+      "epoch": 0.91,
+      "learning_rate": 0.00010885142255005269,
+      "loss": 0.4929,
       "step": 1730
     },
     {
-      "epoch": 1.28,
-      "learning_rate": 7.187039764359352e-05,
-      "loss": 0.3919,
+      "epoch": 0.92,
+      "learning_rate": 0.00010832455216016861,
+      "loss": 0.5158,
       "step": 1740
     },
     {
-      "epoch": 1.29,
-      "learning_rate": 7.113402061855671e-05,
-      "loss": 0.5448,
+      "epoch": 0.92,
+      "learning_rate": 0.00010779768177028451,
+      "loss": 0.5903,
       "step": 1750
     },
     {
-      "epoch": 1.3,
-      "learning_rate": 7.039764359351989e-05,
-      "loss": 0.3676,
+      "epoch": 0.93,
+      "learning_rate": 0.00010727081138040043,
+      "loss": 0.5604,
       "step": 1760
     },
     {
-      "epoch": 1.3,
-      "eval_accuracy": 0.7952117863720074,
-      "eval_loss": 0.5770056247711182,
-      "eval_runtime": 50.6386,
-      "eval_samples_per_second": 53.615,
-      "eval_steps_per_second": 6.714,
+      "epoch": 0.93,
+      "eval_accuracy": 0.773709167544784,
+      "eval_loss": 0.6411083936691284,
+      "eval_runtime": 40.9594,
+      "eval_samples_per_second": 92.677,
+      "eval_steps_per_second": 11.597,
       "step": 1760
     },
     {
-      "epoch": 1.3,
-      "learning_rate": 6.966126656848306e-05,
-      "loss": 0.4213,
+      "epoch": 0.93,
+      "learning_rate": 0.00010674394099051635,
+      "loss": 0.7767,
       "step": 1770
     },
     {
-      "epoch": 1.31,
-      "learning_rate": 6.892488954344625e-05,
-      "loss": 0.5673,
+      "epoch": 0.94,
+      "learning_rate": 0.00010621707060063224,
+      "loss": 0.6115,
       "step": 1780
     },
     {
-      "epoch": 1.32,
-      "learning_rate": 6.818851251840943e-05,
-      "loss": 0.3463,
+      "epoch": 0.94,
+      "learning_rate": 0.00010569020021074815,
+      "loss": 0.49,
       "step": 1790
     },
     {
-      "epoch": 1.33,
-      "learning_rate": 6.745213549337262e-05,
-      "loss": 0.4531,
+      "epoch": 0.95,
+      "learning_rate": 0.00010516332982086407,
+      "loss": 0.5398,
       "step": 1800
     },
     {
-      "epoch": 1.33,
-      "eval_accuracy": 0.7922651933701658,
-      "eval_loss": 0.5833007097244263,
-      "eval_runtime": 51.1166,
-      "eval_samples_per_second": 53.114,
-      "eval_steps_per_second": 6.651,
+      "epoch": 0.95,
+      "eval_accuracy": 0.7642255005268704,
+      "eval_loss": 0.6474949717521667,
+      "eval_runtime": 41.1724,
+      "eval_samples_per_second": 92.198,
+      "eval_steps_per_second": 11.537,
       "step": 1800
     },
     {
-      "epoch": 1.33,
-      "learning_rate": 6.671575846833579e-05,
-      "loss": 0.3818,
+      "epoch": 0.95,
+      "learning_rate": 0.00010463645943097999,
+      "loss": 0.7346,
       "step": 1810
     },
     {
-      "epoch": 1.34,
-      "learning_rate": 6.597938144329897e-05,
-      "loss": 0.4865,
+      "epoch": 0.96,
+      "learning_rate": 0.00010410958904109589,
+      "loss": 0.5711,
       "step": 1820
     },
     {
-      "epoch": 1.35,
-      "learning_rate": 6.524300441826216e-05,
-      "loss": 0.4752,
+      "epoch": 0.96,
+      "learning_rate": 0.00010358271865121181,
+      "loss": 0.4447,
       "step": 1830
     },
     {
-      "epoch": 1.35,
-      "learning_rate": 6.450662739322534e-05,
-      "loss": 0.4353,
+      "epoch": 0.97,
+      "learning_rate": 0.00010305584826132772,
+      "loss": 0.5243,
       "step": 1840
     },
     {
-      "epoch": 1.35,
-      "eval_accuracy": 0.8136279926335175,
-      "eval_loss": 0.5133872628211975,
-      "eval_runtime": 52.3565,
-      "eval_samples_per_second": 51.856,
-      "eval_steps_per_second": 6.494,
+      "epoch": 0.97,
+      "eval_accuracy": 0.7829293993677555,
+      "eval_loss": 0.5976884365081787,
+      "eval_runtime": 41.1462,
+      "eval_samples_per_second": 92.256,
+      "eval_steps_per_second": 11.544,
       "step": 1840
     },
     {
-      "epoch": 1.36,
-      "learning_rate": 6.377025036818851e-05,
-      "loss": 0.3349,
+      "epoch": 0.97,
+      "learning_rate": 0.00010252897787144363,
+      "loss": 0.6316,
       "step": 1850
     },
     {
-      "epoch": 1.37,
-      "learning_rate": 6.30338733431517e-05,
-      "loss": 0.2902,
+      "epoch": 0.98,
+      "learning_rate": 0.00010200210748155954,
+      "loss": 0.4889,
       "step": 1860
     },
     {
-      "epoch": 1.38,
-      "learning_rate": 6.229749631811488e-05,
-      "loss": 0.5968,
+      "epoch": 0.99,
+      "learning_rate": 0.00010147523709167546,
+      "loss": 0.5548,
       "step": 1870
     },
     {
-      "epoch": 1.38,
-      "learning_rate": 6.156111929307806e-05,
-      "loss": 0.422,
+      "epoch": 0.99,
+      "learning_rate": 0.00010094836670179138,
+      "loss": 0.555,
       "step": 1880
     },
     {
-      "epoch": 1.38,
-      "eval_accuracy": 0.8106813996316758,
-      "eval_loss": 0.521941065788269,
-      "eval_runtime": 51.8454,
-      "eval_samples_per_second": 52.367,
-      "eval_steps_per_second": 6.558,
+      "epoch": 0.99,
+      "eval_accuracy": 0.8076923076923077,
+      "eval_loss": 0.5374019742012024,
+      "eval_runtime": 41.2791,
+      "eval_samples_per_second": 91.959,
+      "eval_steps_per_second": 11.507,
       "step": 1880
     },
     {
-      "epoch": 1.39,
-      "learning_rate": 6.0824742268041234e-05,
-      "loss": 0.3134,
+      "epoch": 1.0,
+      "learning_rate": 0.00010042149631190727,
+      "loss": 0.5253,
       "step": 1890
     },
     {
-      "epoch": 1.4,
-      "learning_rate": 6.008836524300442e-05,
-      "loss": 0.5106,
+      "epoch": 1.0,
+      "learning_rate": 9.989462592202319e-05,
+      "loss": 0.4858,
       "step": 1900
     },
     {
-      "epoch": 1.41,
-      "learning_rate": 5.93519882179676e-05,
-      "loss": 0.3672,
+      "epoch": 1.01,
+      "learning_rate": 9.936775553213909e-05,
+      "loss": 0.4391,
       "step": 1910
     },
     {
-      "epoch": 1.41,
-      "learning_rate": 5.861561119293079e-05,
-      "loss": 0.4875,
+      "epoch": 1.01,
+      "learning_rate": 9.884088514225502e-05,
+      "loss": 0.3991,
       "step": 1920
     },
     {
-      "epoch": 1.41,
-      "eval_accuracy": 0.823572744014733,
-      "eval_loss": 0.4959632158279419,
-      "eval_runtime": 51.2335,
-      "eval_samples_per_second": 52.993,
-      "eval_steps_per_second": 6.636,
+      "epoch": 1.01,
+      "eval_accuracy": 0.8079557428872497,
+      "eval_loss": 0.5598599910736084,
+      "eval_runtime": 41.1706,
+      "eval_samples_per_second": 92.202,
+      "eval_steps_per_second": 11.537,
       "step": 1920
     },
     {
-      "epoch": 1.42,
-      "learning_rate": 5.787923416789397e-05,
-      "loss": 0.3232,
+      "epoch": 1.02,
+      "learning_rate": 9.831401475237092e-05,
+      "loss": 0.4727,
       "step": 1930
     },
     {
-      "epoch": 1.43,
-      "learning_rate": 5.714285714285714e-05,
-      "loss": 0.3946,
+      "epoch": 1.02,
+      "learning_rate": 9.778714436248684e-05,
+      "loss": 0.4029,
       "step": 1940
     },
     {
-      "epoch": 1.44,
-      "learning_rate": 5.6406480117820326e-05,
-      "loss": 0.368,
+      "epoch": 1.03,
+      "learning_rate": 9.726027397260274e-05,
+      "loss": 0.4803,
       "step": 1950
     },
     {
-      "epoch": 1.44,
-      "learning_rate": 5.567010309278351e-05,
-      "loss": 0.2695,
+      "epoch": 1.03,
+      "learning_rate": 9.673340358271865e-05,
+      "loss": 0.418,
       "step": 1960
     },
     {
-      "epoch": 1.44,
-      "eval_accuracy": 0.8283609576427255,
-      "eval_loss": 0.4938356876373291,
-      "eval_runtime": 51.5366,
-      "eval_samples_per_second": 52.681,
-      "eval_steps_per_second": 6.597,
+      "epoch": 1.03,
+      "eval_accuracy": 0.8308746048472075,
+      "eval_loss": 0.5048983693122864,
+      "eval_runtime": 41.4141,
+      "eval_samples_per_second": 91.66,
+      "eval_steps_per_second": 11.47,
       "step": 1960
     },
     {
-      "epoch": 1.45,
-      "learning_rate": 5.4933726067746695e-05,
-      "loss": 0.3932,
+      "epoch": 1.04,
+      "learning_rate": 9.620653319283456e-05,
+      "loss": 0.3603,
       "step": 1970
     },
     {
-      "epoch": 1.46,
-      "learning_rate": 5.4197349042709865e-05,
-      "loss": 0.3572,
+      "epoch": 1.04,
+      "learning_rate": 9.567966280295048e-05,
+      "loss": 0.3883,
       "step": 1980
     },
     {
-      "epoch": 1.47,
-      "learning_rate": 5.346097201767305e-05,
-      "loss": 0.5021,
+      "epoch": 1.05,
+      "learning_rate": 9.51527924130664e-05,
+      "loss": 0.3467,
       "step": 1990
     },
     {
-      "epoch": 1.47,
-      "learning_rate": 5.2724594992636234e-05,
-      "loss": 0.3038,
+      "epoch": 1.05,
+      "learning_rate": 9.46259220231823e-05,
+      "loss": 0.4145,
       "step": 2000
     },
     {
-      "epoch": 1.47,
-      "eval_accuracy": 0.8276243093922652,
-      "eval_loss": 0.48587676882743835,
-      "eval_runtime": 52.2942,
-      "eval_samples_per_second": 51.918,
-      "eval_steps_per_second": 6.502,
+      "epoch": 1.05,
+      "eval_accuracy": 0.8116438356164384,
+      "eval_loss": 0.5375648140907288,
+      "eval_runtime": 41.1926,
+      "eval_samples_per_second": 92.152,
+      "eval_steps_per_second": 11.531,
       "step": 2000
     },
     {
-      "epoch": 1.48,
-      "learning_rate": 5.198821796759942e-05,
-      "loss": 0.3817,
+      "epoch": 1.06,
+      "learning_rate": 9.409905163329822e-05,
+      "loss": 0.4242,
       "step": 2010
     },
     {
-      "epoch": 1.49,
-      "learning_rate": 5.125184094256259e-05,
-      "loss": 0.3844,
+      "epoch": 1.06,
+      "learning_rate": 9.357218124341412e-05,
+      "loss": 0.3996,
       "step": 2020
     },
     {
-      "epoch": 1.49,
-      "learning_rate": 5.051546391752577e-05,
-      "loss": 0.3217,
+      "epoch": 1.07,
+      "learning_rate": 9.304531085353004e-05,
+      "loss": 0.2548,
       "step": 2030
     },
     {
-      "epoch": 1.5,
-      "learning_rate": 4.977908689248896e-05,
-      "loss": 0.4076,
+      "epoch": 1.07,
+      "learning_rate": 9.251844046364595e-05,
+      "loss": 0.4141,
       "step": 2040
     },
     {
-      "epoch": 1.5,
-      "eval_accuracy": 0.825414364640884,
-      "eval_loss": 0.4927995204925537,
-      "eval_runtime": 50.7363,
-      "eval_samples_per_second": 53.512,
-      "eval_steps_per_second": 6.701,
+      "epoch": 1.07,
+      "eval_accuracy": 0.8229715489989463,
+      "eval_loss": 0.5164880752563477,
+      "eval_runtime": 41.0208,
+      "eval_samples_per_second": 92.538,
+      "eval_steps_per_second": 11.579,
       "step": 2040
     },
     {
-      "epoch": 1.51,
-      "learning_rate": 4.9042709867452135e-05,
-      "loss": 0.431,
+      "epoch": 1.08,
+      "learning_rate": 9.199157007376186e-05,
+      "loss": 0.3818,
       "step": 2050
     },
     {
-      "epoch": 1.52,
-      "learning_rate": 4.830633284241532e-05,
+      "epoch": 1.09,
+      "learning_rate": 9.146469968387778e-05,
       "loss": 0.3252,
       "step": 2060
     },
     {
-      "epoch": 1.52,
-      "learning_rate": 4.75699558173785e-05,
-      "loss": 0.3145,
+      "epoch": 1.09,
+      "learning_rate": 9.093782929399368e-05,
+      "loss": 0.3749,
       "step": 2070
     },
     {
-      "epoch": 1.53,
-      "learning_rate": 4.683357879234168e-05,
-      "loss": 0.4613,
+      "epoch": 1.1,
+      "learning_rate": 9.041095890410958e-05,
+      "loss": 0.3729,
       "step": 2080
     },
     {
-      "epoch": 1.53,
-      "eval_accuracy": 0.8338858195211787,
-      "eval_loss": 0.4798433482646942,
-      "eval_runtime": 51.6374,
-      "eval_samples_per_second": 52.578,
-      "eval_steps_per_second": 6.584,
+      "epoch": 1.1,
+      "eval_accuracy": 0.8303477344573235,
+      "eval_loss": 0.5023428201675415,
+      "eval_runtime": 41.3017,
+      "eval_samples_per_second": 91.909,
+      "eval_steps_per_second": 11.501,
       "step": 2080
     },
     {
-      "epoch": 1.54,
-      "learning_rate": 4.6097201767304866e-05,
-      "loss": 0.3371,
+      "epoch": 1.1,
+      "learning_rate": 8.988408851422551e-05,
+      "loss": 0.4618,
       "step": 2090
     },
     {
-      "epoch": 1.55,
-      "learning_rate": 4.536082474226804e-05,
-      "loss": 0.5341,
+      "epoch": 1.11,
+      "learning_rate": 8.935721812434142e-05,
+      "loss": 0.3952,
       "step": 2100
     },
     {
-      "epoch": 1.55,
-      "learning_rate": 4.462444771723123e-05,
-      "loss": 0.3712,
+      "epoch": 1.11,
+      "learning_rate": 8.883034773445733e-05,
+      "loss": 0.3239,
       "step": 2110
     },
     {
-      "epoch": 1.56,
-      "learning_rate": 4.3888070692194405e-05,
-      "loss": 0.3672,
+      "epoch": 1.12,
+      "learning_rate": 8.830347734457324e-05,
+      "loss": 0.4594,
       "step": 2120
     },
     {
-      "epoch": 1.56,
-      "eval_accuracy": 0.8331491712707182,
-      "eval_loss": 0.47213244438171387,
-      "eval_runtime": 50.5283,
-      "eval_samples_per_second": 53.732,
-      "eval_steps_per_second": 6.729,
+      "epoch": 1.12,
+      "eval_accuracy": 0.8374604847207587,
+      "eval_loss": 0.4935886263847351,
+      "eval_runtime": 42.0509,
+      "eval_samples_per_second": 90.271,
+      "eval_steps_per_second": 11.296,
       "step": 2120
     },
     {
-      "epoch": 1.57,
-      "learning_rate": 4.315169366715759e-05,
-      "loss": 0.3377,
+      "epoch": 1.12,
+      "learning_rate": 8.777660695468915e-05,
+      "loss": 0.4036,
       "step": 2130
     },
     {
-      "epoch": 1.58,
-      "learning_rate": 4.241531664212077e-05,
-      "loss": 0.377,
+      "epoch": 1.13,
+      "learning_rate": 8.724973656480506e-05,
+      "loss": 0.2812,
       "step": 2140
     },
     {
-      "epoch": 1.58,
-      "learning_rate": 4.167893961708395e-05,
-      "loss": 0.3033,
+      "epoch": 1.13,
+      "learning_rate": 8.672286617492097e-05,
+      "loss": 0.3308,
       "step": 2150
     },
     {
-      "epoch": 1.59,
-      "learning_rate": 4.094256259204713e-05,
-      "loss": 0.4058,
+      "epoch": 1.14,
+      "learning_rate": 8.619599578503689e-05,
+      "loss": 0.3766,
       "step": 2160
     },
     {
-      "epoch": 1.59,
-      "eval_accuracy": 0.8320441988950277,
-      "eval_loss": 0.46750709414482117,
-      "eval_runtime": 51.3576,
-      "eval_samples_per_second": 52.865,
-      "eval_steps_per_second": 6.62,
+      "epoch": 1.14,
+      "eval_accuracy": 0.833245521601686,
+      "eval_loss": 0.5025116801261902,
+      "eval_runtime": 41.7352,
+      "eval_samples_per_second": 90.954,
+      "eval_steps_per_second": 11.381,
       "step": 2160
     },
     {
-      "epoch": 1.6,
-      "learning_rate": 4.020618556701031e-05,
-      "loss": 0.2632,
+      "epoch": 1.14,
+      "learning_rate": 8.56691253951528e-05,
+      "loss": 0.2861,
       "step": 2170
     },
     {
-      "epoch": 1.61,
-      "learning_rate": 3.946980854197349e-05,
-      "loss": 0.3504,
+      "epoch": 1.15,
+      "learning_rate": 8.514225500526871e-05,
+      "loss": 0.3525,
       "step": 2180
     },
     {
-      "epoch": 1.61,
-      "learning_rate": 3.8733431516936675e-05,
-      "loss": 0.2671,
+      "epoch": 1.15,
+      "learning_rate": 8.461538461538461e-05,
+      "loss": 0.4213,
       "step": 2190
     },
     {
-      "epoch": 1.62,
-      "learning_rate": 3.799705449189985e-05,
-      "loss": 0.2954,
+      "epoch": 1.16,
+      "learning_rate": 8.408851422550053e-05,
+      "loss": 0.3628,
       "step": 2200
     },
     {
-      "epoch": 1.62,
-      "eval_accuracy": 0.8368324125230202,
-      "eval_loss": 0.45633676648139954,
-      "eval_runtime": 51.1084,
-      "eval_samples_per_second": 53.122,
-      "eval_steps_per_second": 6.653,
+      "epoch": 1.16,
+      "eval_accuracy": 0.8285036880927292,
+      "eval_loss": 0.5210418105125427,
+      "eval_runtime": 41.7139,
+      "eval_samples_per_second": 91.001,
+      "eval_steps_per_second": 11.387,
       "step": 2200
     },
     {
-      "epoch": 1.63,
-      "learning_rate": 3.726067746686304e-05,
-      "loss": 0.3205,
+      "epoch": 1.16,
+      "learning_rate": 8.356164383561645e-05,
+      "loss": 0.5404,
       "step": 2210
     },
     {
-      "epoch": 1.63,
-      "learning_rate": 3.6524300441826214e-05,
-      "loss": 0.299,
+      "epoch": 1.17,
+      "learning_rate": 8.303477344573235e-05,
+      "loss": 0.383,
       "step": 2220
     },
     {
-      "epoch": 1.64,
-      "learning_rate": 3.57879234167894e-05,
-      "loss": 0.3119,
+      "epoch": 1.17,
+      "learning_rate": 8.250790305584827e-05,
+      "loss": 0.3332,
       "step": 2230
     },
     {
-      "epoch": 1.65,
-      "learning_rate": 3.5051546391752576e-05,
-      "loss": 0.3334,
+      "epoch": 1.18,
+      "learning_rate": 8.198103266596417e-05,
+      "loss": 0.4392,
       "step": 2240
     },
     {
-      "epoch": 1.65,
-      "eval_accuracy": 0.8324125230202578,
-      "eval_loss": 0.47254547476768494,
-      "eval_runtime": 50.5237,
-      "eval_samples_per_second": 53.737,
-      "eval_steps_per_second": 6.73,
+      "epoch": 1.18,
+      "eval_accuracy": 0.8387776606954689,
+      "eval_loss": 0.49416255950927734,
+      "eval_runtime": 41.5117,
+      "eval_samples_per_second": 91.444,
+      "eval_steps_per_second": 11.443,
       "step": 2240
     },
     {
-      "epoch": 1.66,
-      "learning_rate": 3.431516936671576e-05,
-      "loss": 0.3469,
+      "epoch": 1.19,
+      "learning_rate": 8.145416227608009e-05,
+      "loss": 0.3811,
       "step": 2250
     },
     {
-      "epoch": 1.66,
-      "learning_rate": 3.357879234167894e-05,
-      "loss": 0.2501,
+      "epoch": 1.19,
+      "learning_rate": 8.0927291886196e-05,
+      "loss": 0.3708,
       "step": 2260
     },
     {
-      "epoch": 1.67,
-      "learning_rate": 3.284241531664212e-05,
-      "loss": 0.3941,
+      "epoch": 1.2,
+      "learning_rate": 8.040042149631191e-05,
+      "loss": 0.5215,
       "step": 2270
     },
     {
-      "epoch": 1.68,
-      "learning_rate": 3.2106038291605307e-05,
-      "loss": 0.3151,
+      "epoch": 1.2,
+      "learning_rate": 7.987355110642783e-05,
+      "loss": 0.4257,
       "step": 2280
     },
     {
-      "epoch": 1.68,
-      "eval_accuracy": 0.8342541436464088,
-      "eval_loss": 0.49459975957870483,
-      "eval_runtime": 51.2803,
-      "eval_samples_per_second": 52.944,
-      "eval_steps_per_second": 6.63,
+      "epoch": 1.2,
+      "eval_accuracy": 0.8300842992623815,
+      "eval_loss": 0.49942925572395325,
+      "eval_runtime": 41.6212,
+      "eval_samples_per_second": 91.203,
+      "eval_steps_per_second": 11.412,
       "step": 2280
     },
     {
-      "epoch": 1.69,
-      "learning_rate": 3.1369661266568484e-05,
-      "loss": 0.3946,
+      "epoch": 1.21,
+      "learning_rate": 7.934668071654373e-05,
+      "loss": 0.4234,
       "step": 2290
     },
     {
-      "epoch": 1.69,
-      "learning_rate": 3.063328424153167e-05,
-      "loss": 0.3603,
+      "epoch": 1.21,
+      "learning_rate": 7.881981032665965e-05,
+      "loss": 0.3984,
       "step": 2300
     },
     {
-      "epoch": 1.7,
-      "learning_rate": 2.9896907216494846e-05,
-      "loss": 0.2634,
+      "epoch": 1.22,
+      "learning_rate": 7.829293993677555e-05,
+      "loss": 0.4514,
       "step": 2310
     },
     {
-      "epoch": 1.71,
-      "learning_rate": 2.916053019145803e-05,
-      "loss": 0.3322,
+      "epoch": 1.22,
+      "learning_rate": 7.776606954689147e-05,
+      "loss": 0.3442,
       "step": 2320
     },
     {
-      "epoch": 1.71,
-      "eval_accuracy": 0.8401473296500921,
-      "eval_loss": 0.4636726379394531,
-      "eval_runtime": 51.0603,
-      "eval_samples_per_second": 53.172,
-      "eval_steps_per_second": 6.659,
+      "epoch": 1.22,
+      "eval_accuracy": 0.8403582718651211,
+      "eval_loss": 0.4866703748703003,
+      "eval_runtime": 41.5299,
+      "eval_samples_per_second": 91.404,
+      "eval_steps_per_second": 11.438,
       "step": 2320
     },
     {
-      "epoch": 1.72,
-      "learning_rate": 2.8424153166421208e-05,
-      "loss": 0.3206,
+      "epoch": 1.23,
+      "learning_rate": 7.723919915700738e-05,
+      "loss": 0.3327,
       "step": 2330
     },
     {
-      "epoch": 1.72,
-      "learning_rate": 2.7687776141384392e-05,
-      "loss": 0.4357,
+      "epoch": 1.23,
+      "learning_rate": 7.671232876712329e-05,
+      "loss": 0.4527,
       "step": 2340
     },
     {
-      "epoch": 1.73,
-      "learning_rate": 2.695139911634757e-05,
-      "loss": 0.2878,
+      "epoch": 1.24,
+      "learning_rate": 7.61854583772392e-05,
+      "loss": 0.3862,
       "step": 2350
     },
     {
-      "epoch": 1.74,
-      "learning_rate": 2.6215022091310754e-05,
-      "loss": 0.4057,
+      "epoch": 1.24,
+      "learning_rate": 7.565858798735511e-05,
+      "loss": 0.4008,
       "step": 2360
     },
     {
-      "epoch": 1.74,
-      "eval_accuracy": 0.8379373848987108,
-      "eval_loss": 0.461891233921051,
-      "eval_runtime": 51.5789,
-      "eval_samples_per_second": 52.638,
-      "eval_steps_per_second": 6.592,
+      "epoch": 1.24,
+      "eval_accuracy": 0.8163856691253951,
+      "eval_loss": 0.5492302775382996,
+      "eval_runtime": 41.4705,
+      "eval_samples_per_second": 91.535,
+      "eval_steps_per_second": 11.454,
       "step": 2360
     },
     {
-      "epoch": 1.75,
-      "learning_rate": 2.547864506627393e-05,
-      "loss": 0.3853,
+      "epoch": 1.25,
+      "learning_rate": 7.513171759747102e-05,
+      "loss": 0.4259,
       "step": 2370
     },
     {
-      "epoch": 1.75,
-      "learning_rate": 2.4742268041237116e-05,
-      "loss": 0.3127,
+      "epoch": 1.25,
+      "learning_rate": 7.460484720758694e-05,
+      "loss": 0.3364,
       "step": 2380
     },
     {
-      "epoch": 1.76,
-      "learning_rate": 2.4005891016200297e-05,
-      "loss": 0.3005,
+      "epoch": 1.26,
+      "learning_rate": 7.407797681770284e-05,
+      "loss": 0.3355,
       "step": 2390
     },
     {
-      "epoch": 1.77,
-      "learning_rate": 2.3269513991163478e-05,
-      "loss": 0.3365,
+      "epoch": 1.26,
+      "learning_rate": 7.355110642781876e-05,
+      "loss": 0.3541,
       "step": 2400
     },
     {
-      "epoch": 1.77,
-      "eval_accuracy": 0.8372007366482505,
-      "eval_loss": 0.45343735814094543,
-      "eval_runtime": 52.5929,
-      "eval_samples_per_second": 51.623,
-      "eval_steps_per_second": 6.465,
+      "epoch": 1.26,
+      "eval_accuracy": 0.8335089567966281,
+      "eval_loss": 0.5003746151924133,
+      "eval_runtime": 41.6532,
+      "eval_samples_per_second": 91.134,
+      "eval_steps_per_second": 11.404,
       "step": 2400
     },
     {
-      "epoch": 1.77,
-      "learning_rate": 2.253313696612666e-05,
-      "loss": 0.285,
+      "epoch": 1.27,
+      "learning_rate": 7.302423603793467e-05,
+      "loss": 0.3951,
       "step": 2410
     },
     {
-      "epoch": 1.78,
-      "learning_rate": 2.179675994108984e-05,
-      "loss": 0.3114,
+      "epoch": 1.28,
+      "learning_rate": 7.249736564805058e-05,
+      "loss": 0.3318,
       "step": 2420
     },
     {
-      "epoch": 1.79,
-      "learning_rate": 2.106038291605302e-05,
-      "loss": 0.321,
+      "epoch": 1.28,
+      "learning_rate": 7.19704952581665e-05,
+      "loss": 0.3423,
       "step": 2430
     },
     {
-      "epoch": 1.8,
-      "learning_rate": 2.03240058910162e-05,
-      "loss": 0.2832,
+      "epoch": 1.29,
+      "learning_rate": 7.144362486828242e-05,
+      "loss": 0.3842,
       "step": 2440
     },
     {
-      "epoch": 1.8,
-      "eval_accuracy": 0.8397790055248618,
-      "eval_loss": 0.4500523805618286,
-      "eval_runtime": 51.02,
-      "eval_samples_per_second": 53.214,
-      "eval_steps_per_second": 6.664,
+      "epoch": 1.29,
+      "eval_accuracy": 0.8435194942044257,
+      "eval_loss": 0.46031510829925537,
+      "eval_runtime": 41.7276,
+      "eval_samples_per_second": 90.971,
+      "eval_steps_per_second": 11.383,
       "step": 2440
     },
     {
-      "epoch": 1.8,
-      "learning_rate": 1.9587628865979382e-05,
-      "loss": 0.3756,
+      "epoch": 1.29,
+      "learning_rate": 7.091675447839832e-05,
+      "loss": 0.4708,
       "step": 2450
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 1.8851251840942563e-05,
-      "loss": 0.2677,
+      "epoch": 1.3,
+      "learning_rate": 7.038988408851422e-05,
+      "loss": 0.2436,
       "step": 2460
     },
     {
-      "epoch": 1.82,
-      "learning_rate": 1.8114874815905744e-05,
-      "loss": 0.392,
+      "epoch": 1.3,
+      "learning_rate": 6.986301369863014e-05,
+      "loss": 0.5203,
       "step": 2470
     },
     {
-      "epoch": 1.83,
-      "learning_rate": 1.7378497790868925e-05,
-      "loss": 0.2512,
+      "epoch": 1.31,
+      "learning_rate": 6.933614330874604e-05,
+      "loss": 0.3398,
       "step": 2480
     },
     {
-      "epoch": 1.83,
-      "eval_accuracy": 0.8427255985267035,
-      "eval_loss": 0.44581884145736694,
-      "eval_runtime": 50.8126,
-      "eval_samples_per_second": 53.432,
-      "eval_steps_per_second": 6.691,
+      "epoch": 1.31,
+      "eval_accuracy": 0.8519494204425712,
+      "eval_loss": 0.44302287697792053,
+      "eval_runtime": 41.6802,
+      "eval_samples_per_second": 91.074,
+      "eval_steps_per_second": 11.396,
       "step": 2480
     },
     {
-      "epoch": 1.83,
-      "learning_rate": 1.6642120765832106e-05,
-      "loss": 0.3217,
+      "epoch": 1.31,
+      "learning_rate": 6.880927291886196e-05,
+      "loss": 0.2589,
       "step": 2490
     },
     {
-      "epoch": 1.84,
-      "learning_rate": 1.5905743740795287e-05,
-      "loss": 0.3821,
+      "epoch": 1.32,
+      "learning_rate": 6.828240252897788e-05,
+      "loss": 0.3447,
       "step": 2500
     },
     {
-      "epoch": 1.85,
-      "learning_rate": 1.5169366715758468e-05,
-      "loss": 0.2863,
+      "epoch": 1.32,
+      "learning_rate": 6.77555321390938e-05,
+      "loss": 0.2995,
       "step": 2510
     },
     {
-      "epoch": 1.86,
-      "learning_rate": 1.4432989690721649e-05,
-      "loss": 0.2347,
+      "epoch": 1.33,
+      "learning_rate": 6.72286617492097e-05,
+      "loss": 0.3823,
       "step": 2520
     },
     {
-      "epoch": 1.86,
-      "eval_accuracy": 0.8423572744014733,
-      "eval_loss": 0.44903260469436646,
-      "eval_runtime": 51.8254,
-      "eval_samples_per_second": 52.387,
-      "eval_steps_per_second": 6.56,
+      "epoch": 1.33,
+      "eval_accuracy": 0.8466807165437302,
+      "eval_loss": 0.45392054319381714,
+      "eval_runtime": 41.6585,
+      "eval_samples_per_second": 91.122,
+      "eval_steps_per_second": 11.402,
       "step": 2520
     },
     {
-      "epoch": 1.86,
-      "learning_rate": 1.3696612665684833e-05,
-      "loss": 0.1994,
+      "epoch": 1.33,
+      "learning_rate": 6.67017913593256e-05,
+      "loss": 0.3361,
       "step": 2530
     },
     {
-      "epoch": 1.87,
-      "learning_rate": 1.2960235640648014e-05,
-      "loss": 0.2408,
+      "epoch": 1.34,
+      "learning_rate": 6.617492096944152e-05,
+      "loss": 0.354,
       "step": 2540
     },
     {
-      "epoch": 1.88,
-      "learning_rate": 1.2223858615611193e-05,
-      "loss": 0.4364,
+      "epoch": 1.34,
+      "learning_rate": 6.564805057955743e-05,
+      "loss": 0.3558,
       "step": 2550
     },
     {
-      "epoch": 1.89,
-      "learning_rate": 1.1487481590574374e-05,
-      "loss": 0.1912,
+      "epoch": 1.35,
+      "learning_rate": 6.512118018967335e-05,
+      "loss": 0.452,
       "step": 2560
     },
     {
-      "epoch": 1.89,
-      "eval_accuracy": 0.8449355432780847,
-      "eval_loss": 0.44017237424850464,
-      "eval_runtime": 51.6194,
-      "eval_samples_per_second": 52.597,
-      "eval_steps_per_second": 6.587,
+      "epoch": 1.35,
+      "eval_accuracy": 0.8482613277133825,
+      "eval_loss": 0.4561702609062195,
+      "eval_runtime": 41.5457,
+      "eval_samples_per_second": 91.369,
+      "eval_steps_per_second": 11.433,
       "step": 2560
     },
     {
-      "epoch": 1.89,
-      "learning_rate": 1.0751104565537557e-05,
-      "loss": 0.3227,
+      "epoch": 1.35,
+      "learning_rate": 6.459430979978925e-05,
+      "loss": 0.3694,
       "step": 2570
     },
     {
-      "epoch": 1.9,
-      "learning_rate": 1.0014727540500737e-05,
-      "loss": 0.2769,
+      "epoch": 1.36,
+      "learning_rate": 6.406743940990516e-05,
+      "loss": 0.3642,
       "step": 2580
     },
     {
-      "epoch": 1.91,
-      "learning_rate": 9.278350515463918e-06,
-      "loss": 0.2485,
+      "epoch": 1.36,
+      "learning_rate": 6.354056902002108e-05,
+      "loss": 0.3627,
       "step": 2590
     },
     {
-      "epoch": 1.91,
-      "learning_rate": 8.5419734904271e-06,
-      "loss": 0.3679,
+      "epoch": 1.37,
+      "learning_rate": 6.301369863013699e-05,
+      "loss": 0.3121,
       "step": 2600
     },
     {
-      "epoch": 1.91,
-      "eval_accuracy": 0.8464088397790055,
-      "eval_loss": 0.4389094412326813,
-      "eval_runtime": 50.627,
-      "eval_samples_per_second": 53.628,
-      "eval_steps_per_second": 6.716,
+      "epoch": 1.37,
+      "eval_accuracy": 0.8548472075869337,
+      "eval_loss": 0.4360343813896179,
+      "eval_runtime": 41.6262,
+      "eval_samples_per_second": 91.193,
+      "eval_steps_per_second": 11.411,
       "step": 2600
     },
     {
-      "epoch": 1.92,
-      "learning_rate": 7.80559646539028e-06,
-      "loss": 0.3111,
+      "epoch": 1.38,
+      "learning_rate": 6.248682824025291e-05,
+      "loss": 0.3276,
       "step": 2610
     },
     {
-      "epoch": 1.93,
-      "learning_rate": 7.069219440353461e-06,
-      "loss": 0.33,
+      "epoch": 1.38,
+      "learning_rate": 6.195995785036881e-05,
+      "loss": 0.2538,
       "step": 2620
     },
     {
-      "epoch": 1.94,
-      "learning_rate": 6.332842415316642e-06,
-      "loss": 0.2817,
+      "epoch": 1.39,
+      "learning_rate": 6.143308746048473e-05,
+      "loss": 0.2862,
       "step": 2630
     },
     {
-      "epoch": 1.94,
-      "learning_rate": 5.596465390279824e-06,
-      "loss": 0.2953,
+      "epoch": 1.39,
+      "learning_rate": 6.090621707060063e-05,
+      "loss": 0.3032,
       "step": 2640
     },
     {
-      "epoch": 1.94,
-      "eval_accuracy": 0.8515653775322284,
-      "eval_loss": 0.4342978596687317,
-      "eval_runtime": 51.1052,
-      "eval_samples_per_second": 53.126,
-      "eval_steps_per_second": 6.653,
+      "epoch": 1.39,
+      "eval_accuracy": 0.8535300316122234,
+      "eval_loss": 0.4355548620223999,
+      "eval_runtime": 41.6726,
+      "eval_samples_per_second": 91.091,
+      "eval_steps_per_second": 11.398,
       "step": 2640
     },
     {
-      "epoch": 1.95,
-      "learning_rate": 4.860088365243005e-06,
-      "loss": 0.247,
+      "epoch": 1.4,
+      "learning_rate": 6.037934668071654e-05,
+      "loss": 0.3316,
       "step": 2650
     },
     {
-      "epoch": 1.96,
-      "learning_rate": 4.123711340206186e-06,
-      "loss": 0.3024,
+      "epoch": 1.4,
+      "learning_rate": 5.985247629083246e-05,
+      "loss": 0.3128,
       "step": 2660
     },
     {
-      "epoch": 1.97,
-      "learning_rate": 3.3873343151693666e-06,
-      "loss": 0.2232,
+      "epoch": 1.41,
+      "learning_rate": 5.932560590094837e-05,
+      "loss": 0.438,
       "step": 2670
     },
     {
-      "epoch": 1.97,
-      "learning_rate": 2.650957290132548e-06,
-      "loss": 0.2982,
+      "epoch": 1.41,
+      "learning_rate": 5.879873551106429e-05,
+      "loss": 0.361,
       "step": 2680
     },
     {
-      "epoch": 1.97,
-      "eval_accuracy": 0.8486187845303867,
-      "eval_loss": 0.4305761158466339,
-      "eval_runtime": 51.0559,
-      "eval_samples_per_second": 53.177,
-      "eval_steps_per_second": 6.659,
+      "epoch": 1.41,
+      "eval_accuracy": 0.8530031612223393,
+      "eval_loss": 0.4372209906578064,
+      "eval_runtime": 41.6096,
+      "eval_samples_per_second": 91.229,
+      "eval_steps_per_second": 11.416,
       "step": 2680
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.827186512118019e-05,
+      "loss": 0.4102,
+      "step": 2690
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.774499473129611e-05,
+      "loss": 0.3263,
+      "step": 2700
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.721812434141202e-05,
+      "loss": 0.2175,
+      "step": 2710
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.669125395152792e-05,
+      "loss": 0.5349,
+      "step": 2720
+    },
+    {
+      "epoch": 1.43,
+      "eval_accuracy": 0.8532665964172813,
+      "eval_loss": 0.4406832754611969,
+      "eval_runtime": 41.5463,
+      "eval_samples_per_second": 91.368,
+      "eval_steps_per_second": 11.433,
+      "step": 2720
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.616438356164384e-05,
+      "loss": 0.3344,
+      "step": 2730
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.563751317175975e-05,
+      "loss": 0.4089,
+      "step": 2740
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.5110642781875665e-05,
+      "loss": 0.4002,
+      "step": 2750
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.4583772391991575e-05,
+      "loss": 0.2898,
+      "step": 2760
+    },
+    {
+      "epoch": 1.45,
+      "eval_accuracy": 0.8606427818756586,
+      "eval_loss": 0.41055828332901,
+      "eval_runtime": 41.7124,
+      "eval_samples_per_second": 91.004,
+      "eval_steps_per_second": 11.388,
+      "step": 2760
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.405690200210748e-05,
+      "loss": 0.4475,
+      "step": 2770
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.3530031612223395e-05,
+      "loss": 0.3652,
+      "step": 2780
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.3003161222339306e-05,
+      "loss": 0.3678,
+      "step": 2790
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.247629083245522e-05,
+      "loss": 0.2751,
+      "step": 2800
+    },
+    {
+      "epoch": 1.48,
+      "eval_accuracy": 0.8648577449947313,
+      "eval_loss": 0.4102562963962555,
+      "eval_runtime": 41.6322,
+      "eval_samples_per_second": 91.179,
+      "eval_steps_per_second": 11.409,
+      "step": 2800
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.1949420442571126e-05,
+      "loss": 0.4149,
+      "step": 2810
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.142255005268704e-05,
+      "loss": 0.2141,
+      "step": 2820
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.089567966280295e-05,
+      "loss": 0.331,
+      "step": 2830
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.036880927291886e-05,
+      "loss": 0.2966,
+      "step": 2840
+    },
+    {
+      "epoch": 1.5,
+      "eval_accuracy": 0.8635405690200211,
+      "eval_loss": 0.4106617271900177,
+      "eval_runtime": 41.6805,
+      "eval_samples_per_second": 91.074,
+      "eval_steps_per_second": 11.396,
+      "step": 2840
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.984193888303478e-05,
+      "loss": 0.3219,
+      "step": 2850
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.9315068493150684e-05,
+      "loss": 0.3621,
+      "step": 2860
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.8788198103266594e-05,
+      "loss": 0.3827,
+      "step": 2870
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.826132771338251e-05,
+      "loss": 0.2146,
+      "step": 2880
+    },
+    {
+      "epoch": 1.52,
+      "eval_accuracy": 0.8619599578503688,
+      "eval_loss": 0.42587053775787354,
+      "eval_runtime": 41.6661,
+      "eval_samples_per_second": 91.105,
+      "eval_steps_per_second": 11.4,
+      "step": 2880
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.773445732349842e-05,
+      "loss": 0.2426,
+      "step": 2890
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.720758693361433e-05,
+      "loss": 0.2925,
+      "step": 2900
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.668071654373025e-05,
+      "loss": 0.308,
+      "step": 2910
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.615384615384616e-05,
+      "loss": 0.4042,
+      "step": 2920
+    },
+    {
+      "epoch": 1.54,
+      "eval_accuracy": 0.8714436248682824,
+      "eval_loss": 0.39998331665992737,
+      "eval_runtime": 41.7183,
+      "eval_samples_per_second": 90.991,
+      "eval_steps_per_second": 11.386,
+      "step": 2920
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.562697576396207e-05,
+      "loss": 0.4615,
+      "step": 2930
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.510010537407798e-05,
+      "loss": 0.3321,
+      "step": 2940
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.457323498419389e-05,
+      "loss": 0.2821,
+      "step": 2950
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.4046364594309806e-05,
+      "loss": 0.2715,
+      "step": 2960
+    },
+    {
+      "epoch": 1.56,
+      "eval_accuracy": 0.8695995785036881,
+      "eval_loss": 0.40408244729042053,
+      "eval_runtime": 41.2952,
+      "eval_samples_per_second": 91.923,
+      "eval_steps_per_second": 11.503,
+      "step": 2960
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.3519494204425716e-05,
+      "loss": 0.3723,
+      "step": 2970
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.2992623814541626e-05,
+      "loss": 0.2885,
+      "step": 2980
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.2465753424657536e-05,
+      "loss": 0.253,
+      "step": 2990
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.1938883034773446e-05,
+      "loss": 0.2795,
+      "step": 3000
+    },
+    {
+      "epoch": 1.58,
+      "eval_accuracy": 0.869072708113804,
+      "eval_loss": 0.4010108709335327,
+      "eval_runtime": 41.3963,
+      "eval_samples_per_second": 91.699,
+      "eval_steps_per_second": 11.474,
+      "step": 3000
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1412012644889356e-05,
+      "loss": 0.2582,
+      "step": 3010
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.088514225500527e-05,
+      "loss": 0.3652,
+      "step": 3020
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0358271865121184e-05,
+      "loss": 0.404,
+      "step": 3030
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.9831401475237094e-05,
+      "loss": 0.2104,
+      "step": 3040
+    },
+    {
+      "epoch": 1.6,
+      "eval_accuracy": 0.8756585879873551,
+      "eval_loss": 0.38390201330184937,
+      "eval_runtime": 41.4547,
+      "eval_samples_per_second": 91.57,
+      "eval_steps_per_second": 11.458,
+      "step": 3040
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9304531085353004e-05,
+      "loss": 0.2561,
+      "step": 3050
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.8777660695468914e-05,
+      "loss": 0.328,
+      "step": 3060
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8250790305584824e-05,
+      "loss": 0.2671,
+      "step": 3070
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.772391991570074e-05,
+      "loss": 0.2486,
+      "step": 3080
+    },
+    {
+      "epoch": 1.62,
+      "eval_accuracy": 0.8759220231822972,
+      "eval_loss": 0.38206353783607483,
+      "eval_runtime": 41.5068,
+      "eval_samples_per_second": 91.455,
+      "eval_steps_per_second": 11.444,
+      "step": 3080
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.719704952581665e-05,
+      "loss": 0.3262,
+      "step": 3090
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.667017913593256e-05,
+      "loss": 0.3114,
+      "step": 3100
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.614330874604848e-05,
+      "loss": 0.3031,
+      "step": 3110
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.561643835616438e-05,
+      "loss": 0.3005,
+      "step": 3120
+    },
+    {
+      "epoch": 1.64,
+      "eval_accuracy": 0.8695995785036881,
+      "eval_loss": 0.3960082530975342,
+      "eval_runtime": 41.7318,
+      "eval_samples_per_second": 90.962,
+      "eval_steps_per_second": 11.382,
+      "step": 3120
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.50895679662803e-05,
+      "loss": 0.3876,
+      "step": 3130
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.456269757639621e-05,
+      "loss": 0.3107,
+      "step": 3140
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.403582718651212e-05,
+      "loss": 0.2807,
+      "step": 3150
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.350895679662803e-05,
+      "loss": 0.2839,
+      "step": 3160
+    },
+    {
+      "epoch": 1.66,
+      "eval_accuracy": 0.8746048472075869,
+      "eval_loss": 0.38452914357185364,
+      "eval_runtime": 41.6884,
+      "eval_samples_per_second": 91.056,
+      "eval_steps_per_second": 11.394,
+      "step": 3160
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.2982086406743946e-05,
+      "loss": 0.3845,
+      "step": 3170
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.245521601685985e-05,
+      "loss": 0.2933,
+      "step": 3180
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.1928345626975767e-05,
+      "loss": 0.3032,
+      "step": 3190
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.140147523709168e-05,
+      "loss": 0.3101,
+      "step": 3200
+    },
+    {
+      "epoch": 1.69,
+      "eval_accuracy": 0.8777660695468915,
+      "eval_loss": 0.3834909200668335,
+      "eval_runtime": 41.5606,
+      "eval_samples_per_second": 91.337,
+      "eval_steps_per_second": 11.429,
+      "step": 3200
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.087460484720759e-05,
+      "loss": 0.3474,
+      "step": 3210
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.03477344573235e-05,
+      "loss": 0.3775,
+      "step": 3220
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.9820864067439414e-05,
+      "loss": 0.3458,
+      "step": 3230
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.929399367755532e-05,
+      "loss": 0.1596,
+      "step": 3240
+    },
+    {
+      "epoch": 1.71,
+      "eval_accuracy": 0.8822444678609063,
+      "eval_loss": 0.3663616478443146,
+      "eval_runtime": 41.6501,
+      "eval_samples_per_second": 91.14,
+      "eval_steps_per_second": 11.405,
+      "step": 3240
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.8767123287671234e-05,
+      "loss": 0.2665,
+      "step": 3250
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8240252897787145e-05,
+      "loss": 0.3029,
+      "step": 3260
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.7713382507903058e-05,
+      "loss": 0.3974,
+      "step": 3270
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7186512118018968e-05,
+      "loss": 0.3662,
+      "step": 3280
+    },
+    {
+      "epoch": 1.73,
+      "eval_accuracy": 0.8819810326659642,
+      "eval_loss": 0.3698595464229584,
+      "eval_runtime": 41.5938,
+      "eval_samples_per_second": 91.264,
+      "eval_steps_per_second": 11.42,
+      "step": 3280
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.6659641728134882e-05,
+      "loss": 0.304,
+      "step": 3290
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.613277133825079e-05,
+      "loss": 0.2496,
+      "step": 3300
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.5605900948366702e-05,
+      "loss": 0.3154,
+      "step": 3310
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5079030558482612e-05,
+      "loss": 0.406,
+      "step": 3320
+    },
+    {
+      "epoch": 1.75,
+      "eval_accuracy": 0.8788198103266597,
+      "eval_loss": 0.3647627532482147,
+      "eval_runtime": 41.6136,
+      "eval_samples_per_second": 91.22,
+      "eval_steps_per_second": 11.415,
+      "step": 3320
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4552160168598526e-05,
+      "loss": 0.2383,
+      "step": 3330
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.402528977871444e-05,
+      "loss": 0.2272,
+      "step": 3340
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3498419388830346e-05,
+      "loss": 0.3772,
+      "step": 3350
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.297154899894626e-05,
+      "loss": 0.3417,
+      "step": 3360
+    },
+    {
+      "epoch": 1.77,
+      "eval_accuracy": 0.8806638566912539,
+      "eval_loss": 0.37239569425582886,
+      "eval_runtime": 41.7989,
+      "eval_samples_per_second": 90.816,
+      "eval_steps_per_second": 11.364,
+      "step": 3360
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2444678609062173e-05,
+      "loss": 0.2708,
+      "step": 3370
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1917808219178083e-05,
+      "loss": 0.2455,
+      "step": 3380
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1390937829293994e-05,
+      "loss": 0.3778,
+      "step": 3390
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.0864067439409907e-05,
+      "loss": 0.2489,
+      "step": 3400
+    },
+    {
+      "epoch": 1.79,
+      "eval_accuracy": 0.8825079030558483,
+      "eval_loss": 0.361176073551178,
+      "eval_runtime": 41.7201,
+      "eval_samples_per_second": 90.987,
+      "eval_steps_per_second": 11.385,
+      "step": 3400
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0337197049525817e-05,
+      "loss": 0.3192,
+      "step": 3410
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.9810326659641728e-05,
+      "loss": 0.2914,
+      "step": 3420
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.928345626975764e-05,
+      "loss": 0.2141,
+      "step": 3430
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.8756585879873555e-05,
+      "loss": 0.2464,
+      "step": 3440
+    },
+    {
+      "epoch": 1.81,
+      "eval_accuracy": 0.8825079030558483,
+      "eval_loss": 0.36530667543411255,
+      "eval_runtime": 41.819,
+      "eval_samples_per_second": 90.772,
+      "eval_steps_per_second": 11.358,
+      "step": 3440
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.822971548998946e-05,
+      "loss": 0.1882,
+      "step": 3450
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.7702845100105375e-05,
+      "loss": 0.252,
+      "step": 3460
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.717597471022129e-05,
+      "loss": 0.3904,
+      "step": 3470
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.66491043203372e-05,
+      "loss": 0.2391,
+      "step": 3480
+    },
+    {
+      "epoch": 1.83,
+      "eval_accuracy": 0.881190727081138,
+      "eval_loss": 0.3640024662017822,
+      "eval_runtime": 41.67,
+      "eval_samples_per_second": 91.097,
+      "eval_steps_per_second": 11.399,
+      "step": 3480
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.612223393045311e-05,
+      "loss": 0.3707,
+      "step": 3490
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.5595363540569022e-05,
+      "loss": 0.3237,
+      "step": 3500
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5068493150684931e-05,
+      "loss": 0.2514,
+      "step": 3510
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4541622760800844e-05,
+      "loss": 0.4291,
+      "step": 3520
+    },
+    {
+      "epoch": 1.85,
+      "eval_accuracy": 0.8861959957850368,
+      "eval_loss": 0.3543952703475952,
+      "eval_runtime": 41.457,
+      "eval_samples_per_second": 91.565,
+      "eval_steps_per_second": 11.458,
+      "step": 3520
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4014752370916756e-05,
+      "loss": 0.2714,
+      "step": 3530
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3487881981032666e-05,
+      "loss": 0.4164,
+      "step": 3540
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2961011591148578e-05,
+      "loss": 0.2214,
+      "step": 3550
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2434141201264489e-05,
+      "loss": 0.259,
+      "step": 3560
+    },
+    {
+      "epoch": 1.88,
+      "eval_accuracy": 0.8896206533192834,
+      "eval_loss": 0.3500049412250519,
+      "eval_runtime": 41.6053,
+      "eval_samples_per_second": 91.238,
+      "eval_steps_per_second": 11.417,
+      "step": 3560
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1907270811380402e-05,
+      "loss": 0.252,
+      "step": 3570
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1380400421496312e-05,
+      "loss": 0.2847,
+      "step": 3580
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0853530031612224e-05,
+      "loss": 0.2444,
+      "step": 3590
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0326659641728136e-05,
+      "loss": 0.1871,
+      "step": 3600
+    },
+    {
+      "epoch": 1.9,
+      "eval_accuracy": 0.8864594309799789,
+      "eval_loss": 0.35397401452064514,
+      "eval_runtime": 41.3976,
+      "eval_samples_per_second": 91.696,
+      "eval_steps_per_second": 11.474,
+      "step": 3600
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.799789251844046e-06,
+      "loss": 0.1985,
+      "step": 3610
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.272918861959958e-06,
+      "loss": 0.2478,
+      "step": 3620
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.74604847207587e-06,
+      "loss": 0.2614,
+      "step": 3630
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.21917808219178e-06,
+      "loss": 0.2337,
+      "step": 3640
+    },
+    {
+      "epoch": 1.92,
+      "eval_accuracy": 0.8872497365648051,
+      "eval_loss": 0.3516286015510559,
+      "eval_runtime": 41.3744,
+      "eval_samples_per_second": 91.748,
+      "eval_steps_per_second": 11.481,
+      "step": 3640
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.692307692307694e-06,
+      "loss": 0.2885,
+      "step": 3650
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.165437302423604e-06,
+      "loss": 0.2682,
+      "step": 3660
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.638566912539515e-06,
+      "loss": 0.1558,
+      "step": 3670
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 6.1116965226554275e-06,
+      "loss": 0.2336,
+      "step": 3680
+    },
+    {
+      "epoch": 1.94,
+      "eval_accuracy": 0.8890937829293993,
+      "eval_loss": 0.3470406234264374,
+      "eval_runtime": 41.6183,
+      "eval_samples_per_second": 91.21,
+      "eval_steps_per_second": 11.413,
+      "step": 3680
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5848261327713385e-06,
+      "loss": 0.2797,
+      "step": 3690
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.05795574288725e-06,
+      "loss": 0.2202,
+      "step": 3700
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.531085353003161e-06,
+      "loss": 0.2526,
+      "step": 3710
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.004214963119072e-06,
+      "loss": 0.2401,
+      "step": 3720
+    },
+    {
+      "epoch": 1.96,
+      "eval_accuracy": 0.8901475237091675,
+      "eval_loss": 0.3446514308452606,
+      "eval_runtime": 41.3853,
+      "eval_samples_per_second": 91.723,
+      "eval_steps_per_second": 11.477,
+      "step": 3720
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.4773445732349843e-06,
+      "loss": 0.1877,
+      "step": 3730
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.9504741833508957e-06,
+      "loss": 0.163,
+      "step": 3740
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.423603793466807e-06,
+      "loss": 0.2202,
+      "step": 3750
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.8967334035827188e-06,
+      "loss": 0.2327,
+      "step": 3760
+    },
+    {
+      "epoch": 1.98,
+      "eval_accuracy": 0.8904109589041096,
+      "eval_loss": 0.3434308171272278,
+      "eval_runtime": 41.3779,
+      "eval_samples_per_second": 91.74,
+      "eval_steps_per_second": 11.48,
+      "step": 3760
     }
   ],
-  "max_steps": 2716,
+  "max_steps": 3796,
   "num_train_epochs": 2,
-  "total_flos": 3.322629783093584e+18,
+  "total_flos": 4.662174864046436e+18,
   "trial_name": null,
   "trial_params": null
 }