diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,3019 +1,2227 @@
 {
-  "best_metric": 1.2707958221435547,
-  "best_model_checkpoint": "./vit-base-beans/checkpoint-2600",
-  "epoch": 3.9608269858541894,
-  "global_step": 3640,
+  "best_metric": 0.4305761158466339,
+  "best_model_checkpoint": "./vit-base-beans/checkpoint-2680",
+  "epoch": 1.9734904270986746,
+  "global_step": 2680,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.01,
-      "learning_rate": 0.00019945593035908596,
-      "loss": 2.9646,
+      "learning_rate": 0.00019926362297496318,
+      "loss": 2.073,
       "step": 10
     },
     {
-      "epoch": 0.02,
-      "learning_rate": 0.00019891186071817194,
-      "loss": 2.8629,
+      "epoch": 0.01,
+      "learning_rate": 0.00019852724594992638,
+      "loss": 1.5903,
       "step": 20
     },
     {
-      "epoch": 0.03,
-      "learning_rate": 0.0001983677910772579,
-      "loss": 2.8738,
+      "epoch": 0.02,
+      "learning_rate": 0.00019779086892488955,
+      "loss": 1.4651,
       "step": 30
     },
     {
-      "epoch": 0.04,
-      "learning_rate": 0.00019782372143634387,
-      "loss": 2.7533,
+      "epoch": 0.03,
+      "learning_rate": 0.00019705449189985275,
+      "loss": 1.2028,
       "step": 40
     },
     {
-      "epoch": 0.04,
-      "eval_accuracy": 0.235038084874864,
-      "eval_loss": 2.7046968936920166,
-      "eval_runtime": 30.8039,
-      "eval_samples_per_second": 59.668,
-      "eval_steps_per_second": 7.467,
+      "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,
       "step": 40
     },
     {
-      "epoch": 0.05,
-      "learning_rate": 0.00019727965179542982,
-      "loss": 2.6376,
+      "epoch": 0.04,
+      "learning_rate": 0.00019631811487481592,
+      "loss": 1.1103,
       "step": 50
     },
     {
-      "epoch": 0.07,
-      "learning_rate": 0.0001967355821545158,
-      "loss": 2.7082,
+      "epoch": 0.04,
+      "learning_rate": 0.0001955817378497791,
+      "loss": 1.2374,
       "step": 60
     },
     {
-      "epoch": 0.08,
-      "learning_rate": 0.00019619151251360175,
-      "loss": 2.7173,
+      "epoch": 0.05,
+      "learning_rate": 0.00019484536082474229,
+      "loss": 1.1704,
       "step": 70
     },
     {
-      "epoch": 0.09,
-      "learning_rate": 0.0001956474428726877,
-      "loss": 2.6307,
+      "epoch": 0.06,
+      "learning_rate": 0.00019410898379970546,
+      "loss": 1.0396,
       "step": 80
     },
     {
-      "epoch": 0.09,
-      "eval_accuracy": 0.26550598476605003,
-      "eval_loss": 2.542313575744629,
-      "eval_runtime": 31.8198,
-      "eval_samples_per_second": 57.763,
-      "eval_steps_per_second": 7.228,
+      "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,
       "step": 80
     },
     {
-      "epoch": 0.1,
-      "learning_rate": 0.00019510337323177369,
-      "loss": 2.4765,
+      "epoch": 0.07,
+      "learning_rate": 0.00019337260677466863,
+      "loss": 0.9898,
       "step": 90
     },
     {
-      "epoch": 0.11,
-      "learning_rate": 0.00019455930359085964,
-      "loss": 2.4156,
+      "epoch": 0.07,
+      "learning_rate": 0.00019263622974963183,
+      "loss": 0.9049,
       "step": 100
     },
     {
-      "epoch": 0.12,
-      "learning_rate": 0.00019401523394994562,
-      "loss": 2.4868,
+      "epoch": 0.08,
+      "learning_rate": 0.000191899852724595,
+      "loss": 1.0411,
       "step": 110
     },
     {
-      "epoch": 0.13,
-      "learning_rate": 0.00019347116430903157,
-      "loss": 2.4875,
+      "epoch": 0.09,
+      "learning_rate": 0.0001911634756995582,
+      "loss": 1.031,
       "step": 120
     },
     {
-      "epoch": 0.13,
-      "eval_accuracy": 0.25081610446137104,
-      "eval_loss": 2.5568594932556152,
-      "eval_runtime": 34.0995,
-      "eval_samples_per_second": 53.901,
-      "eval_steps_per_second": 6.745,
+      "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,
       "step": 120
     },
     {
-      "epoch": 0.14,
-      "learning_rate": 0.00019292709466811755,
-      "loss": 2.5076,
+      "epoch": 0.1,
+      "learning_rate": 0.00019042709867452136,
+      "loss": 0.9189,
       "step": 130
     },
     {
-      "epoch": 0.15,
-      "learning_rate": 0.0001923830250272035,
-      "loss": 2.5709,
+      "epoch": 0.1,
+      "learning_rate": 0.00018969072164948454,
+      "loss": 1.0785,
       "step": 140
     },
     {
-      "epoch": 0.16,
-      "learning_rate": 0.00019183895538628948,
-      "loss": 2.4629,
+      "epoch": 0.11,
+      "learning_rate": 0.00018895434462444773,
+      "loss": 0.9283,
       "step": 150
     },
     {
-      "epoch": 0.17,
-      "learning_rate": 0.00019129488574537543,
-      "loss": 2.4422,
+      "epoch": 0.12,
+      "learning_rate": 0.0001882179675994109,
+      "loss": 0.94,
       "step": 160
     },
     {
-      "epoch": 0.17,
-      "eval_accuracy": 0.2867247007616975,
-      "eval_loss": 2.414120674133301,
-      "eval_runtime": 31.6836,
-      "eval_samples_per_second": 58.011,
-      "eval_steps_per_second": 7.259,
+      "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,
       "step": 160
     },
     {
-      "epoch": 0.18,
-      "learning_rate": 0.00019075081610446138,
-      "loss": 2.4772,
+      "epoch": 0.13,
+      "learning_rate": 0.00018748159057437408,
+      "loss": 1.0078,
       "step": 170
     },
     {
-      "epoch": 0.2,
-      "learning_rate": 0.00019020674646354733,
-      "loss": 2.3993,
+      "epoch": 0.13,
+      "learning_rate": 0.00018674521354933727,
+      "loss": 0.9582,
       "step": 180
     },
     {
-      "epoch": 0.21,
-      "learning_rate": 0.0001896626768226333,
-      "loss": 2.3927,
+      "epoch": 0.14,
+      "learning_rate": 0.00018600883652430044,
+      "loss": 0.9966,
       "step": 190
     },
     {
-      "epoch": 0.22,
-      "learning_rate": 0.00018911860718171927,
-      "loss": 2.3488,
+      "epoch": 0.15,
+      "learning_rate": 0.00018527245949926364,
+      "loss": 0.8435,
       "step": 200
     },
     {
-      "epoch": 0.22,
-      "eval_accuracy": 0.3133841131664853,
-      "eval_loss": 2.361457347869873,
-      "eval_runtime": 32.0343,
-      "eval_samples_per_second": 57.376,
-      "eval_steps_per_second": 7.18,
+      "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,
       "step": 200
     },
     {
-      "epoch": 0.23,
-      "learning_rate": 0.00018857453754080522,
-      "loss": 2.4865,
+      "epoch": 0.15,
+      "learning_rate": 0.0001845360824742268,
+      "loss": 0.7854,
       "step": 210
     },
     {
-      "epoch": 0.24,
-      "learning_rate": 0.0001880304678998912,
-      "loss": 2.287,
+      "epoch": 0.16,
+      "learning_rate": 0.00018379970544918998,
+      "loss": 0.9786,
       "step": 220
     },
     {
-      "epoch": 0.25,
-      "learning_rate": 0.00018748639825897715,
-      "loss": 2.3538,
+      "epoch": 0.17,
+      "learning_rate": 0.00018306332842415318,
+      "loss": 0.736,
       "step": 230
     },
     {
-      "epoch": 0.26,
-      "learning_rate": 0.0001869423286180631,
-      "loss": 2.4178,
+      "epoch": 0.18,
+      "learning_rate": 0.00018232695139911635,
+      "loss": 0.851,
       "step": 240
     },
     {
-      "epoch": 0.26,
-      "eval_accuracy": 0.3193688792165397,
-      "eval_loss": 2.2913949489593506,
-      "eval_runtime": 31.6054,
-      "eval_samples_per_second": 58.155,
-      "eval_steps_per_second": 7.277,
+      "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,
       "step": 240
     },
     {
-      "epoch": 0.27,
-      "learning_rate": 0.00018639825897714908,
-      "loss": 2.3083,
+      "epoch": 0.18,
+      "learning_rate": 0.00018159057437407955,
+      "loss": 0.9344,
       "step": 250
     },
     {
-      "epoch": 0.28,
-      "learning_rate": 0.00018585418933623503,
-      "loss": 2.1267,
+      "epoch": 0.19,
+      "learning_rate": 0.00018085419734904272,
+      "loss": 0.8802,
       "step": 260
     },
     {
-      "epoch": 0.29,
-      "learning_rate": 0.000185310119695321,
-      "loss": 2.3088,
+      "epoch": 0.2,
+      "learning_rate": 0.0001801178203240059,
+      "loss": 0.772,
       "step": 270
     },
     {
-      "epoch": 0.3,
-      "learning_rate": 0.00018476605005440696,
-      "loss": 2.1498,
+      "epoch": 0.21,
+      "learning_rate": 0.0001793814432989691,
+      "loss": 0.9324,
       "step": 280
     },
     {
-      "epoch": 0.3,
-      "eval_accuracy": 0.33133841131664854,
-      "eval_loss": 2.230421543121338,
-      "eval_runtime": 32.4123,
-      "eval_samples_per_second": 56.707,
-      "eval_steps_per_second": 7.096,
+      "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,
       "step": 280
     },
     {
-      "epoch": 0.32,
-      "learning_rate": 0.00018422198041349294,
-      "loss": 2.2239,
+      "epoch": 0.21,
+      "learning_rate": 0.00017864506627393226,
+      "loss": 0.6804,
       "step": 290
     },
     {
-      "epoch": 0.33,
-      "learning_rate": 0.0001836779107725789,
-      "loss": 2.2776,
+      "epoch": 0.22,
+      "learning_rate": 0.00017790868924889543,
+      "loss": 0.7877,
       "step": 300
     },
     {
-      "epoch": 0.34,
-      "learning_rate": 0.00018313384113166484,
-      "loss": 2.2717,
+      "epoch": 0.23,
+      "learning_rate": 0.00017717231222385863,
+      "loss": 0.9609,
       "step": 310
     },
     {
-      "epoch": 0.35,
-      "learning_rate": 0.00018258977149075082,
-      "loss": 2.2416,
+      "epoch": 0.24,
+      "learning_rate": 0.0001764359351988218,
+      "loss": 0.8963,
       "step": 320
     },
     {
-      "epoch": 0.35,
-      "eval_accuracy": 0.34766050054406966,
-      "eval_loss": 2.155261516571045,
-      "eval_runtime": 31.3536,
-      "eval_samples_per_second": 58.622,
-      "eval_steps_per_second": 7.336,
+      "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,
       "step": 320
     },
     {
-      "epoch": 0.36,
-      "learning_rate": 0.00018204570184983678,
-      "loss": 2.1923,
+      "epoch": 0.24,
+      "learning_rate": 0.000175699558173785,
+      "loss": 0.8531,
       "step": 330
     },
     {
-      "epoch": 0.37,
-      "learning_rate": 0.00018150163220892275,
-      "loss": 2.1404,
+      "epoch": 0.25,
+      "learning_rate": 0.00017496318114874817,
+      "loss": 0.899,
       "step": 340
     },
     {
-      "epoch": 0.38,
-      "learning_rate": 0.0001809575625680087,
-      "loss": 2.127,
+      "epoch": 0.26,
+      "learning_rate": 0.00017422680412371134,
+      "loss": 0.816,
       "step": 350
     },
     {
-      "epoch": 0.39,
-      "learning_rate": 0.00018041349292709469,
-      "loss": 2.1643,
+      "epoch": 0.27,
+      "learning_rate": 0.00017349042709867454,
+      "loss": 0.7532,
       "step": 360
     },
     {
-      "epoch": 0.39,
-      "eval_accuracy": 0.34494015233949943,
-      "eval_loss": 2.2201523780822754,
-      "eval_runtime": 31.5353,
-      "eval_samples_per_second": 58.284,
-      "eval_steps_per_second": 7.293,
+      "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,
       "step": 360
     },
     {
-      "epoch": 0.4,
-      "learning_rate": 0.00017986942328618064,
-      "loss": 2.3303,
+      "epoch": 0.27,
+      "learning_rate": 0.0001727540500736377,
+      "loss": 0.8984,
       "step": 370
     },
     {
-      "epoch": 0.41,
-      "learning_rate": 0.0001793253536452666,
-      "loss": 2.0957,
+      "epoch": 0.28,
+      "learning_rate": 0.0001720176730486009,
+      "loss": 0.843,
       "step": 380
     },
     {
-      "epoch": 0.42,
-      "learning_rate": 0.00017878128400435257,
-      "loss": 2.1675,
+      "epoch": 0.29,
+      "learning_rate": 0.00017128129602356408,
+      "loss": 0.7647,
       "step": 390
     },
     {
-      "epoch": 0.44,
-      "learning_rate": 0.00017823721436343852,
-      "loss": 2.1688,
+      "epoch": 0.29,
+      "learning_rate": 0.00017054491899852725,
+      "loss": 0.7937,
       "step": 400
     },
     {
-      "epoch": 0.44,
-      "eval_accuracy": 0.3715995647442873,
-      "eval_loss": 2.1544973850250244,
-      "eval_runtime": 31.7596,
-      "eval_samples_per_second": 57.872,
-      "eval_steps_per_second": 7.242,
+      "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,
       "step": 400
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 0.0001776931447225245,
-      "loss": 2.1298,
+      "epoch": 0.3,
+      "learning_rate": 0.00016980854197349044,
+      "loss": 0.9007,
       "step": 410
     },
     {
-      "epoch": 0.46,
-      "learning_rate": 0.00017714907508161045,
-      "loss": 2.1467,
+      "epoch": 0.31,
+      "learning_rate": 0.00016907216494845361,
+      "loss": 0.7149,
       "step": 420
     },
     {
-      "epoch": 0.47,
-      "learning_rate": 0.00017660500544069643,
-      "loss": 2.0672,
+      "epoch": 0.32,
+      "learning_rate": 0.00016833578792341679,
+      "loss": 0.8868,
       "step": 430
     },
     {
-      "epoch": 0.48,
-      "learning_rate": 0.00017606093579978238,
-      "loss": 2.0324,
+      "epoch": 0.32,
+      "learning_rate": 0.00016759941089837998,
+      "loss": 0.8365,
       "step": 440
     },
     {
-      "epoch": 0.48,
-      "eval_accuracy": 0.3928182807399347,
-      "eval_loss": 2.0655834674835205,
-      "eval_runtime": 31.6367,
-      "eval_samples_per_second": 58.097,
-      "eval_steps_per_second": 7.27,
+      "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,
       "step": 440
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 0.00017551686615886836,
-      "loss": 1.9756,
+      "epoch": 0.33,
+      "learning_rate": 0.00016686303387334315,
+      "loss": 0.7985,
       "step": 450
     },
     {
-      "epoch": 0.5,
-      "learning_rate": 0.0001749727965179543,
-      "loss": 2.1049,
+      "epoch": 0.34,
+      "learning_rate": 0.00016612665684830635,
+      "loss": 0.775,
       "step": 460
     },
     {
-      "epoch": 0.51,
-      "learning_rate": 0.00017442872687704027,
-      "loss": 2.0169,
+      "epoch": 0.35,
+      "learning_rate": 0.00016539027982326952,
+      "loss": 0.7912,
       "step": 470
     },
     {
-      "epoch": 0.52,
-      "learning_rate": 0.00017388465723612624,
-      "loss": 2.1648,
+      "epoch": 0.35,
+      "learning_rate": 0.0001646539027982327,
+      "loss": 0.8333,
       "step": 480
     },
     {
-      "epoch": 0.52,
-      "eval_accuracy": 0.3982589771490751,
-      "eval_loss": 2.052687644958496,
-      "eval_runtime": 31.623,
-      "eval_samples_per_second": 58.122,
-      "eval_steps_per_second": 7.273,
+      "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,
       "step": 480
     },
     {
-      "epoch": 0.53,
-      "learning_rate": 0.0001733405875952122,
-      "loss": 1.9753,
+      "epoch": 0.36,
+      "learning_rate": 0.0001639175257731959,
+      "loss": 0.8501,
       "step": 490
     },
     {
-      "epoch": 0.54,
-      "learning_rate": 0.00017279651795429818,
-      "loss": 2.2119,
+      "epoch": 0.37,
+      "learning_rate": 0.00016318114874815906,
+      "loss": 0.6558,
       "step": 500
     },
     {
-      "epoch": 0.55,
-      "learning_rate": 0.00017225244831338413,
-      "loss": 2.0171,
+      "epoch": 0.38,
+      "learning_rate": 0.00016244477172312223,
+      "loss": 0.5491,
       "step": 510
     },
     {
-      "epoch": 0.57,
-      "learning_rate": 0.0001717083786724701,
-      "loss": 2.0568,
+      "epoch": 0.38,
+      "learning_rate": 0.00016170839469808543,
+      "loss": 0.6784,
       "step": 520
     },
     {
-      "epoch": 0.57,
-      "eval_accuracy": 0.4009793253536453,
-      "eval_loss": 2.0108883380889893,
-      "eval_runtime": 31.345,
-      "eval_samples_per_second": 58.638,
-      "eval_steps_per_second": 7.338,
+      "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,
       "step": 520
     },
     {
-      "epoch": 0.58,
-      "learning_rate": 0.00017116430903155606,
-      "loss": 2.0011,
+      "epoch": 0.39,
+      "learning_rate": 0.0001609720176730486,
+      "loss": 0.6781,
       "step": 530
     },
     {
-      "epoch": 0.59,
-      "learning_rate": 0.000170620239390642,
-      "loss": 1.9614,
+      "epoch": 0.4,
+      "learning_rate": 0.0001602356406480118,
+      "loss": 0.7243,
       "step": 540
     },
     {
-      "epoch": 0.6,
-      "learning_rate": 0.000170076169749728,
-      "loss": 1.9669,
+      "epoch": 0.41,
+      "learning_rate": 0.00015949926362297497,
+      "loss": 0.664,
       "step": 550
     },
     {
-      "epoch": 0.61,
-      "learning_rate": 0.00016953210010881394,
-      "loss": 1.9287,
+      "epoch": 0.41,
+      "learning_rate": 0.00015876288659793814,
+      "loss": 0.7103,
       "step": 560
     },
     {
-      "epoch": 0.61,
-      "eval_accuracy": 0.3900979325353645,
-      "eval_loss": 2.06852126121521,
-      "eval_runtime": 31.417,
-      "eval_samples_per_second": 58.503,
-      "eval_steps_per_second": 7.321,
+      "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,
       "step": 560
     },
     {
-      "epoch": 0.62,
-      "learning_rate": 0.0001689880304678999,
-      "loss": 1.9279,
+      "epoch": 0.42,
+      "learning_rate": 0.00015802650957290134,
+      "loss": 0.7232,
       "step": 570
     },
     {
-      "epoch": 0.63,
-      "learning_rate": 0.00016844396082698585,
-      "loss": 2.0612,
+      "epoch": 0.43,
+      "learning_rate": 0.0001572901325478645,
+      "loss": 0.7658,
       "step": 580
     },
     {
-      "epoch": 0.64,
-      "learning_rate": 0.00016789989118607182,
-      "loss": 2.0225,
+      "epoch": 0.43,
+      "learning_rate": 0.0001565537555228277,
+      "loss": 0.7853,
       "step": 590
     },
     {
-      "epoch": 0.65,
-      "learning_rate": 0.00016735582154515778,
-      "loss": 2.0335,
+      "epoch": 0.44,
+      "learning_rate": 0.00015581737849779088,
+      "loss": 0.6149,
       "step": 600
     },
     {
-      "epoch": 0.65,
-      "eval_accuracy": 0.39336235038084877,
-      "eval_loss": 2.0253407955169678,
-      "eval_runtime": 31.882,
-      "eval_samples_per_second": 57.65,
-      "eval_steps_per_second": 7.214,
+      "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,
       "step": 600
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 0.00016681175190424373,
-      "loss": 2.2113,
+      "epoch": 0.45,
+      "learning_rate": 0.00015508100147275405,
+      "loss": 0.7379,
       "step": 610
     },
     {
-      "epoch": 0.67,
-      "learning_rate": 0.0001662676822633297,
-      "loss": 2.101,
+      "epoch": 0.46,
+      "learning_rate": 0.00015434462444771725,
+      "loss": 0.7758,
       "step": 620
     },
     {
-      "epoch": 0.69,
-      "learning_rate": 0.00016572361262241566,
-      "loss": 1.9211,
+      "epoch": 0.46,
+      "learning_rate": 0.00015360824742268042,
+      "loss": 0.6333,
       "step": 630
     },
     {
-      "epoch": 0.7,
-      "learning_rate": 0.00016517954298150164,
-      "loss": 2.1521,
+      "epoch": 0.47,
+      "learning_rate": 0.0001528718703976436,
+      "loss": 0.5998,
       "step": 640
     },
     {
-      "epoch": 0.7,
-      "eval_accuracy": 0.40696409140369966,
-      "eval_loss": 2.0242509841918945,
-      "eval_runtime": 31.6803,
-      "eval_samples_per_second": 58.017,
-      "eval_steps_per_second": 7.26,
+      "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,
       "step": 640
     },
     {
-      "epoch": 0.71,
-      "learning_rate": 0.0001646354733405876,
-      "loss": 1.9707,
+      "epoch": 0.48,
+      "learning_rate": 0.00015213549337260679,
+      "loss": 0.6485,
       "step": 650
     },
     {
-      "epoch": 0.72,
-      "learning_rate": 0.00016409140369967357,
-      "loss": 1.7482,
+      "epoch": 0.49,
+      "learning_rate": 0.00015139911634756996,
+      "loss": 0.8297,
       "step": 660
     },
     {
-      "epoch": 0.73,
-      "learning_rate": 0.00016354733405875952,
-      "loss": 2.0443,
+      "epoch": 0.49,
+      "learning_rate": 0.00015066273932253315,
+      "loss": 0.8158,
       "step": 670
     },
     {
-      "epoch": 0.74,
-      "learning_rate": 0.0001630032644178455,
-      "loss": 1.909,
+      "epoch": 0.5,
+      "learning_rate": 0.00014992636229749633,
+      "loss": 0.8814,
       "step": 680
     },
     {
-      "epoch": 0.74,
-      "eval_accuracy": 0.4363438520130577,
-      "eval_loss": 1.8839356899261475,
-      "eval_runtime": 31.4005,
-      "eval_samples_per_second": 58.534,
-      "eval_steps_per_second": 7.325,
+      "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,
       "step": 680
     },
     {
-      "epoch": 0.75,
-      "learning_rate": 0.00016245919477693145,
-      "loss": 1.9677,
+      "epoch": 0.51,
+      "learning_rate": 0.0001491899852724595,
+      "loss": 0.764,
       "step": 690
     },
     {
-      "epoch": 0.76,
-      "learning_rate": 0.0001619151251360174,
-      "loss": 1.9786,
+      "epoch": 0.52,
+      "learning_rate": 0.0001484536082474227,
+      "loss": 0.6825,
       "step": 700
     },
     {
-      "epoch": 0.77,
-      "learning_rate": 0.00016137105549510338,
-      "loss": 1.856,
+      "epoch": 0.52,
+      "learning_rate": 0.00014771723122238586,
+      "loss": 0.6629,
       "step": 710
     },
     {
-      "epoch": 0.78,
-      "learning_rate": 0.00016082698585418933,
-      "loss": 1.9212,
+      "epoch": 0.53,
+      "learning_rate": 0.00014698085419734906,
+      "loss": 0.7405,
       "step": 720
     },
     {
-      "epoch": 0.78,
-      "eval_accuracy": 0.4504896626768226,
-      "eval_loss": 1.8623088598251343,
-      "eval_runtime": 31.5361,
-      "eval_samples_per_second": 58.282,
-      "eval_steps_per_second": 7.293,
+      "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,
       "step": 720
     },
     {
-      "epoch": 0.79,
-      "learning_rate": 0.00016028291621327531,
-      "loss": 1.7232,
+      "epoch": 0.54,
+      "learning_rate": 0.00014624447717231223,
+      "loss": 0.8144,
       "step": 730
     },
     {
-      "epoch": 0.81,
-      "learning_rate": 0.00015973884657236127,
-      "loss": 1.9299,
+      "epoch": 0.54,
+      "learning_rate": 0.0001455081001472754,
+      "loss": 0.5245,
       "step": 740
     },
     {
-      "epoch": 0.82,
-      "learning_rate": 0.00015919477693144724,
-      "loss": 1.9622,
+      "epoch": 0.55,
+      "learning_rate": 0.0001447717231222386,
+      "loss": 0.7705,
       "step": 750
     },
     {
-      "epoch": 0.83,
-      "learning_rate": 0.0001586507072905332,
-      "loss": 1.9359,
+      "epoch": 0.56,
+      "learning_rate": 0.00014403534609720177,
+      "loss": 0.7463,
       "step": 760
     },
     {
-      "epoch": 0.83,
-      "eval_accuracy": 0.42546245919477693,
-      "eval_loss": 1.893409013748169,
-      "eval_runtime": 31.8615,
-      "eval_samples_per_second": 57.687,
-      "eval_steps_per_second": 7.219,
+      "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,
       "step": 760
     },
     {
-      "epoch": 0.84,
-      "learning_rate": 0.00015810663764961915,
-      "loss": 1.929,
+      "epoch": 0.57,
+      "learning_rate": 0.00014329896907216494,
+      "loss": 0.7648,
       "step": 770
     },
     {
-      "epoch": 0.85,
-      "learning_rate": 0.00015756256800870513,
-      "loss": 1.8717,
+      "epoch": 0.57,
+      "learning_rate": 0.00014256259204712814,
+      "loss": 0.6192,
       "step": 780
     },
     {
-      "epoch": 0.86,
-      "learning_rate": 0.00015701849836779108,
-      "loss": 1.9117,
+      "epoch": 0.58,
+      "learning_rate": 0.0001418262150220913,
+      "loss": 0.872,
       "step": 790
     },
     {
-      "epoch": 0.87,
-      "learning_rate": 0.00015647442872687706,
-      "loss": 1.8435,
+      "epoch": 0.59,
+      "learning_rate": 0.0001410898379970545,
+      "loss": 0.684,
       "step": 800
     },
     {
-      "epoch": 0.87,
-      "eval_accuracy": 0.4815016322089227,
-      "eval_loss": 1.765435814857483,
-      "eval_runtime": 31.2045,
-      "eval_samples_per_second": 58.902,
-      "eval_steps_per_second": 7.371,
+      "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,
       "step": 800
     },
     {
-      "epoch": 0.88,
-      "learning_rate": 0.000155930359085963,
-      "loss": 1.8022,
+      "epoch": 0.6,
+      "learning_rate": 0.00014035346097201768,
+      "loss": 0.7713,
       "step": 810
     },
     {
-      "epoch": 0.89,
-      "learning_rate": 0.000155386289445049,
-      "loss": 1.8581,
+      "epoch": 0.6,
+      "learning_rate": 0.00013961708394698085,
+      "loss": 0.6673,
       "step": 820
     },
     {
-      "epoch": 0.9,
-      "learning_rate": 0.00015484221980413494,
-      "loss": 2.1219,
+      "epoch": 0.61,
+      "learning_rate": 0.00013888070692194405,
+      "loss": 0.6792,
       "step": 830
     },
     {
-      "epoch": 0.91,
-      "learning_rate": 0.0001542981501632209,
-      "loss": 1.7116,
+      "epoch": 0.62,
+      "learning_rate": 0.00013814432989690722,
+      "loss": 0.6252,
       "step": 840
     },
     {
-      "epoch": 0.91,
-      "eval_accuracy": 0.47551686615886835,
-      "eval_loss": 1.7760648727416992,
-      "eval_runtime": 31.1164,
-      "eval_samples_per_second": 59.068,
-      "eval_steps_per_second": 7.392,
+      "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,
       "step": 840
     },
     {
-      "epoch": 0.92,
-      "learning_rate": 0.00015375408052230687,
-      "loss": 1.8721,
+      "epoch": 0.63,
+      "learning_rate": 0.0001374079528718704,
+      "loss": 0.7425,
       "step": 850
     },
     {
-      "epoch": 0.94,
-      "learning_rate": 0.00015321001088139282,
-      "loss": 1.8389,
+      "epoch": 0.63,
+      "learning_rate": 0.0001366715758468336,
+      "loss": 0.7285,
       "step": 860
     },
     {
-      "epoch": 0.95,
-      "learning_rate": 0.0001526659412404788,
-      "loss": 1.8014,
+      "epoch": 0.64,
+      "learning_rate": 0.00013593519882179676,
+      "loss": 0.7498,
       "step": 870
     },
     {
-      "epoch": 0.96,
-      "learning_rate": 0.00015212187159956476,
-      "loss": 1.7788,
+      "epoch": 0.65,
+      "learning_rate": 0.00013519882179675996,
+      "loss": 0.754,
       "step": 880
     },
     {
-      "epoch": 0.96,
-      "eval_accuracy": 0.49183895538628947,
-      "eval_loss": 1.7206392288208008,
-      "eval_runtime": 31.9363,
-      "eval_samples_per_second": 57.552,
-      "eval_steps_per_second": 7.202,
+      "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,
       "step": 880
     },
     {
-      "epoch": 0.97,
-      "learning_rate": 0.00015157780195865073,
-      "loss": 1.7437,
+      "epoch": 0.66,
+      "learning_rate": 0.00013446244477172313,
+      "loss": 0.6756,
       "step": 890
     },
     {
-      "epoch": 0.98,
-      "learning_rate": 0.00015103373231773669,
-      "loss": 1.7861,
+      "epoch": 0.66,
+      "learning_rate": 0.0001337260677466863,
+      "loss": 0.636,
       "step": 900
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 0.00015048966267682264,
-      "loss": 1.9204,
+      "epoch": 0.67,
+      "learning_rate": 0.0001329896907216495,
+      "loss": 0.5823,
       "step": 910
     },
     {
-      "epoch": 1.0,
-      "learning_rate": 0.00014994559303590862,
-      "loss": 1.8087,
+      "epoch": 0.68,
+      "learning_rate": 0.00013225331369661267,
+      "loss": 0.6416,
       "step": 920
     },
     {
-      "epoch": 1.0,
-      "eval_accuracy": 0.4956474428726877,
-      "eval_loss": 1.7031298875808716,
-      "eval_runtime": 31.4911,
-      "eval_samples_per_second": 58.366,
-      "eval_steps_per_second": 7.304,
+      "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,
       "step": 920
     },
     {
-      "epoch": 1.01,
-      "learning_rate": 0.00014940152339499457,
-      "loss": 1.3634,
+      "epoch": 0.68,
+      "learning_rate": 0.00013151693667157587,
+      "loss": 0.6125,
       "step": 930
     },
     {
-      "epoch": 1.02,
-      "learning_rate": 0.00014885745375408052,
-      "loss": 1.5091,
+      "epoch": 0.69,
+      "learning_rate": 0.00013078055964653904,
+      "loss": 0.687,
       "step": 940
     },
     {
-      "epoch": 1.03,
-      "learning_rate": 0.0001483133841131665,
-      "loss": 1.3875,
+      "epoch": 0.7,
+      "learning_rate": 0.0001300441826215022,
+      "loss": 0.7283,
       "step": 950
     },
     {
-      "epoch": 1.04,
-      "learning_rate": 0.00014776931447225245,
-      "loss": 1.352,
+      "epoch": 0.71,
+      "learning_rate": 0.0001293078055964654,
+      "loss": 0.7123,
       "step": 960
     },
     {
-      "epoch": 1.04,
-      "eval_accuracy": 0.500544069640914,
-      "eval_loss": 1.699517846107483,
-      "eval_runtime": 31.4103,
-      "eval_samples_per_second": 58.516,
-      "eval_steps_per_second": 7.322,
+      "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,
       "step": 960
     },
     {
-      "epoch": 1.06,
-      "learning_rate": 0.0001472252448313384,
-      "loss": 1.5501,
+      "epoch": 0.71,
+      "learning_rate": 0.00012857142857142858,
+      "loss": 0.6579,
       "step": 970
     },
     {
-      "epoch": 1.07,
-      "learning_rate": 0.00014668117519042438,
-      "loss": 1.2166,
+      "epoch": 0.72,
+      "learning_rate": 0.00012783505154639175,
+      "loss": 0.6744,
       "step": 980
     },
     {
-      "epoch": 1.08,
-      "learning_rate": 0.00014613710554951034,
-      "loss": 1.3054,
+      "epoch": 0.73,
+      "learning_rate": 0.00012709867452135494,
+      "loss": 0.619,
       "step": 990
     },
     {
-      "epoch": 1.09,
-      "learning_rate": 0.0001455930359085963,
-      "loss": 1.6001,
+      "epoch": 0.74,
+      "learning_rate": 0.00012636229749631812,
+      "loss": 0.6599,
       "step": 1000
     },
     {
-      "epoch": 1.09,
-      "eval_accuracy": 0.5119695321001088,
-      "eval_loss": 1.688511610031128,
-      "eval_runtime": 31.1036,
-      "eval_samples_per_second": 59.093,
-      "eval_steps_per_second": 7.395,
+      "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,
       "step": 1000
     },
     {
-      "epoch": 1.1,
-      "learning_rate": 0.00014504896626768227,
-      "loss": 1.427,
+      "epoch": 0.74,
+      "learning_rate": 0.0001256259204712813,
+      "loss": 0.5407,
       "step": 1010
     },
     {
-      "epoch": 1.11,
-      "learning_rate": 0.00014450489662676822,
-      "loss": 1.453,
+      "epoch": 0.75,
+      "learning_rate": 0.00012488954344624448,
+      "loss": 0.7339,
       "step": 1020
     },
     {
-      "epoch": 1.12,
-      "learning_rate": 0.0001439608269858542,
-      "loss": 1.5073,
+      "epoch": 0.76,
+      "learning_rate": 0.00012415316642120765,
+      "loss": 0.65,
       "step": 1030
     },
     {
-      "epoch": 1.13,
-      "learning_rate": 0.00014341675734494015,
-      "loss": 1.5872,
+      "epoch": 0.77,
+      "learning_rate": 0.00012341678939617085,
+      "loss": 0.7664,
       "step": 1040
     },
     {
-      "epoch": 1.13,
-      "eval_accuracy": 0.500544069640914,
-      "eval_loss": 1.7050400972366333,
-      "eval_runtime": 31.6643,
-      "eval_samples_per_second": 58.046,
-      "eval_steps_per_second": 7.264,
+      "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,
       "step": 1040
     },
     {
-      "epoch": 1.14,
-      "learning_rate": 0.00014287268770402613,
-      "loss": 1.4122,
+      "epoch": 0.77,
+      "learning_rate": 0.00012268041237113402,
+      "loss": 0.667,
       "step": 1050
     },
     {
-      "epoch": 1.15,
-      "learning_rate": 0.00014232861806311208,
-      "loss": 1.407,
+      "epoch": 0.78,
+      "learning_rate": 0.00012194403534609722,
+      "loss": 0.6393,
       "step": 1060
     },
     {
-      "epoch": 1.16,
-      "learning_rate": 0.00014178454842219803,
-      "loss": 1.4817,
+      "epoch": 0.79,
+      "learning_rate": 0.00012120765832106038,
+      "loss": 0.7258,
       "step": 1070
     },
     {
-      "epoch": 1.18,
-      "learning_rate": 0.000141240478781284,
-      "loss": 1.2903,
+      "epoch": 0.8,
+      "learning_rate": 0.00012047128129602356,
+      "loss": 0.6898,
       "step": 1080
     },
     {
-      "epoch": 1.18,
-      "eval_accuracy": 0.5070729053318824,
-      "eval_loss": 1.6926010847091675,
-      "eval_runtime": 31.355,
-      "eval_samples_per_second": 58.619,
-      "eval_steps_per_second": 7.335,
+      "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,
       "step": 1080
     },
     {
-      "epoch": 1.19,
-      "learning_rate": 0.00014069640914036996,
-      "loss": 1.4445,
+      "epoch": 0.8,
+      "learning_rate": 0.00011973490427098675,
+      "loss": 0.6158,
       "step": 1090
     },
     {
-      "epoch": 1.2,
-      "learning_rate": 0.00014015233949945594,
-      "loss": 1.3711,
+      "epoch": 0.81,
+      "learning_rate": 0.00011899852724594993,
+      "loss": 0.5764,
       "step": 1100
     },
     {
-      "epoch": 1.21,
-      "learning_rate": 0.0001396082698585419,
-      "loss": 1.3271,
+      "epoch": 0.82,
+      "learning_rate": 0.00011826215022091312,
+      "loss": 0.6926,
       "step": 1110
     },
     {
-      "epoch": 1.22,
-      "learning_rate": 0.00013906420021762787,
-      "loss": 1.3377,
+      "epoch": 0.82,
+      "learning_rate": 0.0001175257731958763,
+      "loss": 0.693,
       "step": 1120
     },
     {
-      "epoch": 1.22,
-      "eval_accuracy": 0.5184983677910773,
-      "eval_loss": 1.663212776184082,
-      "eval_runtime": 31.9094,
-      "eval_samples_per_second": 57.601,
-      "eval_steps_per_second": 7.208,
+      "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,
       "step": 1120
     },
     {
-      "epoch": 1.23,
-      "learning_rate": 0.00013852013057671382,
-      "loss": 1.3493,
+      "epoch": 0.83,
+      "learning_rate": 0.00011678939617083948,
+      "loss": 0.7522,
       "step": 1130
     },
     {
-      "epoch": 1.24,
-      "learning_rate": 0.00013797606093579978,
-      "loss": 1.2249,
+      "epoch": 0.84,
+      "learning_rate": 0.00011605301914580267,
+      "loss": 0.451,
       "step": 1140
     },
     {
-      "epoch": 1.25,
-      "learning_rate": 0.00013743199129488576,
-      "loss": 1.6033,
+      "epoch": 0.85,
+      "learning_rate": 0.00011531664212076583,
+      "loss": 0.564,
       "step": 1150
     },
     {
-      "epoch": 1.26,
-      "learning_rate": 0.0001368879216539717,
-      "loss": 1.3738,
+      "epoch": 0.85,
+      "learning_rate": 0.00011458026509572901,
+      "loss": 0.6738,
       "step": 1160
     },
     {
-      "epoch": 1.26,
-      "eval_accuracy": 0.5206746463547334,
-      "eval_loss": 1.6760975122451782,
-      "eval_runtime": 31.2925,
-      "eval_samples_per_second": 58.736,
-      "eval_steps_per_second": 7.35,
+      "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,
       "step": 1160
     },
     {
-      "epoch": 1.27,
-      "learning_rate": 0.0001363438520130577,
-      "loss": 1.5122,
+      "epoch": 0.86,
+      "learning_rate": 0.0001138438880706922,
+      "loss": 0.6272,
       "step": 1170
     },
     {
-      "epoch": 1.28,
-      "learning_rate": 0.00013579978237214364,
-      "loss": 1.3597,
+      "epoch": 0.87,
+      "learning_rate": 0.00011310751104565538,
+      "loss": 0.5538,
       "step": 1180
     },
     {
-      "epoch": 1.29,
-      "learning_rate": 0.00013525571273122962,
-      "loss": 1.2399,
+      "epoch": 0.88,
+      "learning_rate": 0.00011237113402061856,
+      "loss": 0.555,
       "step": 1190
     },
     {
-      "epoch": 1.31,
-      "learning_rate": 0.00013471164309031557,
-      "loss": 1.2835,
+      "epoch": 0.88,
+      "learning_rate": 0.00011163475699558175,
+      "loss": 0.6056,
       "step": 1200
     },
     {
-      "epoch": 1.31,
-      "eval_accuracy": 0.5375408052230686,
-      "eval_loss": 1.5999504327774048,
-      "eval_runtime": 31.108,
-      "eval_samples_per_second": 59.084,
-      "eval_steps_per_second": 7.394,
+      "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,
       "step": 1200
     },
     {
-      "epoch": 1.32,
-      "learning_rate": 0.00013416757344940155,
-      "loss": 1.6025,
+      "epoch": 0.89,
+      "learning_rate": 0.00011089837997054493,
+      "loss": 0.5609,
       "step": 1210
     },
     {
-      "epoch": 1.33,
-      "learning_rate": 0.0001336235038084875,
-      "loss": 1.3795,
+      "epoch": 0.9,
+      "learning_rate": 0.00011016200294550812,
+      "loss": 0.6141,
       "step": 1220
     },
     {
-      "epoch": 1.34,
-      "learning_rate": 0.00013307943416757345,
-      "loss": 1.3779,
+      "epoch": 0.91,
+      "learning_rate": 0.0001094256259204713,
+      "loss": 0.6212,
       "step": 1230
     },
     {
-      "epoch": 1.35,
-      "learning_rate": 0.00013253536452665943,
-      "loss": 1.2588,
+      "epoch": 0.91,
+      "learning_rate": 0.00010868924889543446,
+      "loss": 0.746,
       "step": 1240
     },
     {
-      "epoch": 1.35,
-      "eval_accuracy": 0.5239390642002176,
-      "eval_loss": 1.6404486894607544,
-      "eval_runtime": 31.2356,
-      "eval_samples_per_second": 58.843,
-      "eval_steps_per_second": 7.363,
+      "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,
       "step": 1240
     },
     {
-      "epoch": 1.36,
-      "learning_rate": 0.00013199129488574538,
-      "loss": 1.3196,
+      "epoch": 0.92,
+      "learning_rate": 0.00010795287187039764,
+      "loss": 0.5153,
       "step": 1250
     },
     {
-      "epoch": 1.37,
-      "learning_rate": 0.00013144722524483136,
-      "loss": 1.2977,
+      "epoch": 0.93,
+      "learning_rate": 0.00010721649484536083,
+      "loss": 0.6638,
       "step": 1260
     },
     {
-      "epoch": 1.38,
-      "learning_rate": 0.00013090315560391731,
-      "loss": 1.354,
+      "epoch": 0.94,
+      "learning_rate": 0.00010648011782032401,
+      "loss": 0.6719,
       "step": 1270
     },
     {
-      "epoch": 1.39,
-      "learning_rate": 0.0001303590859630033,
-      "loss": 1.198,
+      "epoch": 0.94,
+      "learning_rate": 0.0001057437407952872,
+      "loss": 0.7612,
       "step": 1280
     },
     {
-      "epoch": 1.39,
-      "eval_accuracy": 0.5457018498367792,
-      "eval_loss": 1.5461368560791016,
-      "eval_runtime": 31.2158,
-      "eval_samples_per_second": 58.88,
-      "eval_steps_per_second": 7.368,
+      "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,
       "step": 1280
     },
     {
-      "epoch": 1.4,
-      "learning_rate": 0.00012981501632208925,
-      "loss": 1.3318,
+      "epoch": 0.95,
+      "learning_rate": 0.00010500736377025038,
+      "loss": 0.6873,
       "step": 1290
     },
     {
-      "epoch": 1.41,
-      "learning_rate": 0.0001292709466811752,
-      "loss": 1.2417,
+      "epoch": 0.96,
+      "learning_rate": 0.00010427098674521356,
+      "loss": 0.5534,
       "step": 1300
     },
     {
-      "epoch": 1.43,
-      "learning_rate": 0.00012872687704026115,
-      "loss": 1.2636,
+      "epoch": 0.96,
+      "learning_rate": 0.00010353460972017675,
+      "loss": 0.3849,
       "step": 1310
     },
     {
-      "epoch": 1.44,
-      "learning_rate": 0.00012818280739934713,
-      "loss": 1.4038,
+      "epoch": 0.97,
+      "learning_rate": 0.0001027982326951399,
+      "loss": 0.5939,
       "step": 1320
     },
     {
-      "epoch": 1.44,
-      "eval_accuracy": 0.5337323177366703,
-      "eval_loss": 1.5740242004394531,
-      "eval_runtime": 31.2028,
-      "eval_samples_per_second": 58.905,
-      "eval_steps_per_second": 7.371,
+      "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,
       "step": 1320
     },
     {
-      "epoch": 1.45,
-      "learning_rate": 0.00012763873775843308,
-      "loss": 1.3172,
+      "epoch": 0.98,
+      "learning_rate": 0.00010206185567010309,
+      "loss": 0.5831,
       "step": 1330
     },
     {
-      "epoch": 1.46,
-      "learning_rate": 0.00012709466811751903,
-      "loss": 1.1899,
+      "epoch": 0.99,
+      "learning_rate": 0.00010132547864506627,
+      "loss": 0.6712,
       "step": 1340
     },
     {
-      "epoch": 1.47,
-      "learning_rate": 0.000126550598476605,
-      "loss": 1.2146,
+      "epoch": 0.99,
+      "learning_rate": 0.00010058910162002946,
+      "loss": 0.5368,
       "step": 1350
     },
     {
-      "epoch": 1.48,
-      "learning_rate": 0.00012600652883569096,
-      "loss": 1.2275,
+      "epoch": 1.0,
+      "learning_rate": 9.985272459499264e-05,
+      "loss": 0.5586,
       "step": 1360
     },
     {
-      "epoch": 1.48,
-      "eval_accuracy": 0.5516866158868335,
-      "eval_loss": 1.5276788473129272,
-      "eval_runtime": 31.2082,
-      "eval_samples_per_second": 58.895,
-      "eval_steps_per_second": 7.37,
+      "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,
       "step": 1360
     },
     {
-      "epoch": 1.49,
-      "learning_rate": 0.00012546245919477691,
-      "loss": 1.2164,
+      "epoch": 1.01,
+      "learning_rate": 9.911634756995583e-05,
+      "loss": 0.4372,
       "step": 1370
     },
     {
-      "epoch": 1.5,
-      "learning_rate": 0.0001249183895538629,
-      "loss": 1.1369,
+      "epoch": 1.02,
+      "learning_rate": 9.8379970544919e-05,
+      "loss": 0.4076,
       "step": 1380
     },
     {
-      "epoch": 1.51,
-      "learning_rate": 0.00012437431991294885,
-      "loss": 1.3417,
+      "epoch": 1.02,
+      "learning_rate": 9.764359351988218e-05,
+      "loss": 0.3933,
       "step": 1390
     },
     {
-      "epoch": 1.52,
-      "learning_rate": 0.00012383025027203483,
-      "loss": 1.2224,
+      "epoch": 1.03,
+      "learning_rate": 9.690721649484537e-05,
+      "loss": 0.459,
       "step": 1400
     },
     {
-      "epoch": 1.52,
-      "eval_accuracy": 0.5614798694232862,
-      "eval_loss": 1.4918007850646973,
-      "eval_runtime": 31.6631,
-      "eval_samples_per_second": 58.049,
-      "eval_steps_per_second": 7.264,
+      "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,
       "step": 1400
     },
     {
-      "epoch": 1.53,
-      "learning_rate": 0.00012328618063112078,
-      "loss": 1.208,
+      "epoch": 1.04,
+      "learning_rate": 9.617083946980855e-05,
+      "loss": 0.3769,
       "step": 1410
     },
     {
-      "epoch": 1.55,
-      "learning_rate": 0.00012274211099020676,
-      "loss": 1.4005,
+      "epoch": 1.05,
+      "learning_rate": 9.543446244477173e-05,
+      "loss": 0.411,
       "step": 1420
     },
     {
-      "epoch": 1.56,
-      "learning_rate": 0.0001221980413492927,
-      "loss": 1.2049,
+      "epoch": 1.05,
+      "learning_rate": 9.46980854197349e-05,
+      "loss": 0.4444,
       "step": 1430
     },
     {
-      "epoch": 1.57,
-      "learning_rate": 0.00012165397170837869,
-      "loss": 1.3221,
+      "epoch": 1.06,
+      "learning_rate": 9.396170839469809e-05,
+      "loss": 0.3771,
       "step": 1440
     },
     {
-      "epoch": 1.57,
-      "eval_accuracy": 0.558215451577802,
-      "eval_loss": 1.5159553289413452,
-      "eval_runtime": 31.668,
-      "eval_samples_per_second": 58.04,
-      "eval_steps_per_second": 7.263,
+      "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,
       "step": 1440
     },
     {
-      "epoch": 1.58,
-      "learning_rate": 0.00012110990206746464,
-      "loss": 1.0449,
+      "epoch": 1.07,
+      "learning_rate": 9.322533136966127e-05,
+      "loss": 0.3943,
       "step": 1450
     },
     {
-      "epoch": 1.59,
-      "learning_rate": 0.00012056583242655059,
-      "loss": 1.352,
+      "epoch": 1.08,
+      "learning_rate": 9.248895434462446e-05,
+      "loss": 0.5164,
       "step": 1460
     },
     {
-      "epoch": 1.6,
-      "learning_rate": 0.00012002176278563657,
-      "loss": 1.2259,
+      "epoch": 1.08,
+      "learning_rate": 9.175257731958763e-05,
+      "loss": 0.3544,
       "step": 1470
     },
     {
-      "epoch": 1.61,
-      "learning_rate": 0.00011947769314472252,
-      "loss": 1.1424,
+      "epoch": 1.09,
+      "learning_rate": 9.101620029455081e-05,
+      "loss": 0.4272,
       "step": 1480
     },
     {
-      "epoch": 1.61,
-      "eval_accuracy": 0.544069640914037,
-      "eval_loss": 1.553783893585205,
-      "eval_runtime": 31.3063,
-      "eval_samples_per_second": 58.71,
-      "eval_steps_per_second": 7.347,
+      "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,
       "step": 1480
     },
     {
-      "epoch": 1.62,
-      "learning_rate": 0.0001189336235038085,
-      "loss": 1.1163,
+      "epoch": 1.1,
+      "learning_rate": 9.0279823269514e-05,
+      "loss": 0.3962,
       "step": 1490
     },
     {
-      "epoch": 1.63,
-      "learning_rate": 0.00011838955386289445,
-      "loss": 1.0946,
+      "epoch": 1.1,
+      "learning_rate": 8.954344624447718e-05,
+      "loss": 0.434,
       "step": 1500
     },
     {
-      "epoch": 1.64,
-      "learning_rate": 0.00011784548422198043,
-      "loss": 1.083,
+      "epoch": 1.11,
+      "learning_rate": 8.880706921944035e-05,
+      "loss": 0.4332,
       "step": 1510
     },
     {
-      "epoch": 1.65,
-      "learning_rate": 0.00011730141458106638,
-      "loss": 1.2951,
+      "epoch": 1.12,
+      "learning_rate": 8.807069219440354e-05,
+      "loss": 0.3229,
       "step": 1520
     },
     {
-      "epoch": 1.65,
-      "eval_accuracy": 0.558215451577802,
-      "eval_loss": 1.4977946281433105,
-      "eval_runtime": 31.2346,
-      "eval_samples_per_second": 58.845,
-      "eval_steps_per_second": 7.364,
+      "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,
       "step": 1520
     },
     {
-      "epoch": 1.66,
-      "learning_rate": 0.00011675734494015234,
-      "loss": 1.2416,
+      "epoch": 1.13,
+      "learning_rate": 8.733431516936672e-05,
+      "loss": 0.4162,
       "step": 1530
     },
     {
-      "epoch": 1.68,
-      "learning_rate": 0.00011621327529923831,
-      "loss": 1.3838,
+      "epoch": 1.13,
+      "learning_rate": 8.65979381443299e-05,
+      "loss": 0.3704,
       "step": 1540
     },
     {
-      "epoch": 1.69,
-      "learning_rate": 0.00011566920565832427,
-      "loss": 1.2389,
+      "epoch": 1.14,
+      "learning_rate": 8.586156111929308e-05,
+      "loss": 0.6147,
       "step": 1550
     },
     {
-      "epoch": 1.7,
-      "learning_rate": 0.00011512513601741025,
-      "loss": 1.2609,
+      "epoch": 1.15,
+      "learning_rate": 8.512518409425626e-05,
+      "loss": 0.3736,
       "step": 1560
     },
     {
-      "epoch": 1.7,
-      "eval_accuracy": 0.5603917301414582,
-      "eval_loss": 1.4800572395324707,
-      "eval_runtime": 31.1581,
-      "eval_samples_per_second": 58.99,
-      "eval_steps_per_second": 7.382,
+      "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,
       "step": 1560
     },
     {
-      "epoch": 1.71,
-      "learning_rate": 0.0001145810663764962,
-      "loss": 1.2379,
+      "epoch": 1.16,
+      "learning_rate": 8.438880706921944e-05,
+      "loss": 0.4865,
       "step": 1570
     },
     {
-      "epoch": 1.72,
-      "learning_rate": 0.00011403699673558216,
-      "loss": 1.2098,
+      "epoch": 1.16,
+      "learning_rate": 8.365243004418263e-05,
+      "loss": 0.3929,
       "step": 1580
     },
     {
-      "epoch": 1.73,
-      "learning_rate": 0.00011349292709466813,
-      "loss": 0.9532,
+      "epoch": 1.17,
+      "learning_rate": 8.291605301914581e-05,
+      "loss": 0.3426,
       "step": 1590
     },
     {
-      "epoch": 1.74,
-      "learning_rate": 0.00011294885745375408,
-      "loss": 1.035,
+      "epoch": 1.18,
+      "learning_rate": 8.217967599410898e-05,
+      "loss": 0.436,
       "step": 1600
     },
     {
-      "epoch": 1.74,
-      "eval_accuracy": 0.5603917301414582,
-      "eval_loss": 1.4868272542953491,
-      "eval_runtime": 31.3011,
-      "eval_samples_per_second": 58.72,
-      "eval_steps_per_second": 7.348,
+      "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,
       "step": 1600
     },
     {
-      "epoch": 1.75,
-      "learning_rate": 0.00011240478781284005,
-      "loss": 1.2168,
+      "epoch": 1.19,
+      "learning_rate": 8.144329896907217e-05,
+      "loss": 0.4115,
       "step": 1610
     },
     {
-      "epoch": 1.76,
-      "learning_rate": 0.000111860718171926,
-      "loss": 1.2665,
+      "epoch": 1.19,
+      "learning_rate": 8.070692194403535e-05,
+      "loss": 0.3784,
       "step": 1620
     },
     {
-      "epoch": 1.77,
-      "learning_rate": 0.00011131664853101198,
-      "loss": 1.203,
+      "epoch": 1.2,
+      "learning_rate": 7.997054491899854e-05,
+      "loss": 0.3383,
       "step": 1630
     },
     {
-      "epoch": 1.78,
-      "learning_rate": 0.00011077257889009793,
-      "loss": 1.2621,
+      "epoch": 1.21,
+      "learning_rate": 7.923416789396171e-05,
+      "loss": 0.3721,
       "step": 1640
     },
     {
-      "epoch": 1.78,
-      "eval_accuracy": 0.5778019586507073,
-      "eval_loss": 1.439843773841858,
-      "eval_runtime": 31.2165,
-      "eval_samples_per_second": 58.879,
-      "eval_steps_per_second": 7.368,
+      "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,
       "step": 1640
     },
     {
-      "epoch": 1.8,
-      "learning_rate": 0.00011022850924918391,
-      "loss": 1.0592,
+      "epoch": 1.22,
+      "learning_rate": 7.849779086892489e-05,
+      "loss": 0.4844,
       "step": 1650
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 0.00010968443960826986,
-      "loss": 1.2805,
+      "epoch": 1.22,
+      "learning_rate": 7.776141384388808e-05,
+      "loss": 0.4011,
       "step": 1660
     },
     {
-      "epoch": 1.82,
-      "learning_rate": 0.00010914036996735584,
-      "loss": 1.24,
+      "epoch": 1.23,
+      "learning_rate": 7.702503681885126e-05,
+      "loss": 0.4711,
       "step": 1670
     },
     {
-      "epoch": 1.83,
-      "learning_rate": 0.00010859630032644179,
-      "loss": 1.3298,
+      "epoch": 1.24,
+      "learning_rate": 7.628865979381443e-05,
+      "loss": 0.3848,
       "step": 1680
     },
     {
-      "epoch": 1.83,
-      "eval_accuracy": 0.5756256800870512,
-      "eval_loss": 1.4398308992385864,
-      "eval_runtime": 31.3359,
-      "eval_samples_per_second": 58.655,
-      "eval_steps_per_second": 7.34,
+      "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,
       "step": 1680
     },
     {
-      "epoch": 1.84,
-      "learning_rate": 0.00010805223068552774,
-      "loss": 1.1844,
+      "epoch": 1.24,
+      "learning_rate": 7.555228276877762e-05,
+      "loss": 0.3771,
       "step": 1690
     },
     {
-      "epoch": 1.85,
-      "learning_rate": 0.00010750816104461372,
-      "loss": 1.1519,
+      "epoch": 1.25,
+      "learning_rate": 7.48159057437408e-05,
+      "loss": 0.5486,
       "step": 1700
     },
     {
-      "epoch": 1.86,
-      "learning_rate": 0.00010696409140369967,
-      "loss": 1.446,
+      "epoch": 1.26,
+      "learning_rate": 7.407952871870398e-05,
+      "loss": 0.3795,
       "step": 1710
     },
     {
-      "epoch": 1.87,
-      "learning_rate": 0.00010642002176278565,
-      "loss": 1.2588,
+      "epoch": 1.27,
+      "learning_rate": 7.334315169366715e-05,
+      "loss": 0.393,
       "step": 1720
     },
     {
-      "epoch": 1.87,
-      "eval_accuracy": 0.5652883569096845,
-      "eval_loss": 1.4510564804077148,
-      "eval_runtime": 31.3918,
-      "eval_samples_per_second": 58.55,
-      "eval_steps_per_second": 7.327,
+      "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,
       "step": 1720
     },
     {
-      "epoch": 1.88,
-      "learning_rate": 0.0001058759521218716,
-      "loss": 1.1114,
+      "epoch": 1.27,
+      "learning_rate": 7.260677466863034e-05,
+      "loss": 0.4113,
       "step": 1730
     },
     {
-      "epoch": 1.89,
-      "learning_rate": 0.00010533188248095758,
-      "loss": 1.0445,
+      "epoch": 1.28,
+      "learning_rate": 7.187039764359352e-05,
+      "loss": 0.3919,
       "step": 1740
     },
     {
-      "epoch": 1.9,
-      "learning_rate": 0.00010478781284004354,
-      "loss": 1.0959,
+      "epoch": 1.29,
+      "learning_rate": 7.113402061855671e-05,
+      "loss": 0.5448,
       "step": 1750
     },
     {
-      "epoch": 1.92,
-      "learning_rate": 0.00010424374319912949,
-      "loss": 1.2207,
+      "epoch": 1.3,
+      "learning_rate": 7.039764359351989e-05,
+      "loss": 0.3676,
       "step": 1760
     },
     {
-      "epoch": 1.92,
-      "eval_accuracy": 0.5935799782372143,
-      "eval_loss": 1.3912413120269775,
-      "eval_runtime": 32.052,
-      "eval_samples_per_second": 57.344,
-      "eval_steps_per_second": 7.176,
+      "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,
       "step": 1760
     },
     {
-      "epoch": 1.93,
-      "learning_rate": 0.00010369967355821547,
-      "loss": 1.2121,
+      "epoch": 1.3,
+      "learning_rate": 6.966126656848306e-05,
+      "loss": 0.4213,
       "step": 1770
     },
     {
-      "epoch": 1.94,
-      "learning_rate": 0.00010315560391730142,
-      "loss": 1.1847,
+      "epoch": 1.31,
+      "learning_rate": 6.892488954344625e-05,
+      "loss": 0.5673,
       "step": 1780
     },
     {
-      "epoch": 1.95,
-      "learning_rate": 0.00010261153427638738,
-      "loss": 1.1857,
+      "epoch": 1.32,
+      "learning_rate": 6.818851251840943e-05,
+      "loss": 0.3463,
       "step": 1790
     },
     {
-      "epoch": 1.96,
-      "learning_rate": 0.00010206746463547334,
-      "loss": 1.0643,
+      "epoch": 1.33,
+      "learning_rate": 6.745213549337262e-05,
+      "loss": 0.4531,
       "step": 1800
     },
     {
-      "epoch": 1.96,
-      "eval_accuracy": 0.6006528835690969,
-      "eval_loss": 1.390648365020752,
-      "eval_runtime": 31.7091,
-      "eval_samples_per_second": 57.964,
-      "eval_steps_per_second": 7.253,
+      "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,
       "step": 1800
     },
     {
-      "epoch": 1.97,
-      "learning_rate": 0.00010152339499455931,
-      "loss": 1.2481,
+      "epoch": 1.33,
+      "learning_rate": 6.671575846833579e-05,
+      "loss": 0.3818,
       "step": 1810
     },
     {
-      "epoch": 1.98,
-      "learning_rate": 0.00010097932535364527,
-      "loss": 1.2109,
+      "epoch": 1.34,
+      "learning_rate": 6.597938144329897e-05,
+      "loss": 0.4865,
       "step": 1820
     },
     {
-      "epoch": 1.99,
-      "learning_rate": 0.00010043525571273122,
-      "loss": 1.0191,
+      "epoch": 1.35,
+      "learning_rate": 6.524300441826216e-05,
+      "loss": 0.4752,
       "step": 1830
     },
     {
-      "epoch": 2.0,
-      "learning_rate": 9.98911860718172e-05,
-      "loss": 1.0464,
+      "epoch": 1.35,
+      "learning_rate": 6.450662739322534e-05,
+      "loss": 0.4353,
       "step": 1840
     },
     {
-      "epoch": 2.0,
-      "eval_accuracy": 0.5957562568008705,
-      "eval_loss": 1.3601454496383667,
-      "eval_runtime": 31.4007,
-      "eval_samples_per_second": 58.534,
-      "eval_steps_per_second": 7.325,
+      "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,
       "step": 1840
     },
     {
-      "epoch": 2.01,
-      "learning_rate": 9.934711643090316e-05,
-      "loss": 0.645,
+      "epoch": 1.36,
+      "learning_rate": 6.377025036818851e-05,
+      "loss": 0.3349,
       "step": 1850
     },
     {
-      "epoch": 2.02,
-      "learning_rate": 9.880304678998913e-05,
-      "loss": 0.6802,
+      "epoch": 1.37,
+      "learning_rate": 6.30338733431517e-05,
+      "loss": 0.2902,
       "step": 1860
     },
     {
-      "epoch": 2.03,
-      "learning_rate": 9.825897714907508e-05,
-      "loss": 0.6228,
+      "epoch": 1.38,
+      "learning_rate": 6.229749631811488e-05,
+      "loss": 0.5968,
       "step": 1870
     },
     {
-      "epoch": 2.05,
-      "learning_rate": 9.771490750816105e-05,
-      "loss": 0.6345,
+      "epoch": 1.38,
+      "learning_rate": 6.156111929307806e-05,
+      "loss": 0.422,
       "step": 1880
     },
     {
-      "epoch": 2.05,
-      "eval_accuracy": 0.5973884657236126,
-      "eval_loss": 1.3790967464447021,
-      "eval_runtime": 31.7311,
-      "eval_samples_per_second": 57.924,
-      "eval_steps_per_second": 7.248,
+      "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,
       "step": 1880
     },
     {
-      "epoch": 2.06,
-      "learning_rate": 9.717083786724701e-05,
-      "loss": 0.5766,
+      "epoch": 1.39,
+      "learning_rate": 6.0824742268041234e-05,
+      "loss": 0.3134,
       "step": 1890
     },
     {
-      "epoch": 2.07,
-      "learning_rate": 9.662676822633298e-05,
-      "loss": 0.5642,
+      "epoch": 1.4,
+      "learning_rate": 6.008836524300442e-05,
+      "loss": 0.5106,
       "step": 1900
     },
     {
-      "epoch": 2.08,
-      "learning_rate": 9.608269858541894e-05,
-      "loss": 0.5394,
+      "epoch": 1.41,
+      "learning_rate": 5.93519882179676e-05,
+      "loss": 0.3672,
       "step": 1910
     },
     {
-      "epoch": 2.09,
-      "learning_rate": 9.553862894450491e-05,
-      "loss": 0.5622,
+      "epoch": 1.41,
+      "learning_rate": 5.861561119293079e-05,
+      "loss": 0.4875,
       "step": 1920
     },
     {
-      "epoch": 2.09,
-      "eval_accuracy": 0.5957562568008705,
-      "eval_loss": 1.3800909519195557,
-      "eval_runtime": 31.768,
-      "eval_samples_per_second": 57.857,
-      "eval_steps_per_second": 7.24,
+      "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,
       "step": 1920
     },
     {
-      "epoch": 2.1,
-      "learning_rate": 9.499455930359087e-05,
-      "loss": 0.5452,
+      "epoch": 1.42,
+      "learning_rate": 5.787923416789397e-05,
+      "loss": 0.3232,
       "step": 1930
     },
     {
-      "epoch": 2.11,
-      "learning_rate": 9.445048966267683e-05,
-      "loss": 0.5394,
+      "epoch": 1.43,
+      "learning_rate": 5.714285714285714e-05,
+      "loss": 0.3946,
       "step": 1940
     },
     {
-      "epoch": 2.12,
-      "learning_rate": 9.390642002176279e-05,
-      "loss": 0.6138,
+      "epoch": 1.44,
+      "learning_rate": 5.6406480117820326e-05,
+      "loss": 0.368,
       "step": 1950
     },
     {
-      "epoch": 2.13,
-      "learning_rate": 9.336235038084876e-05,
-      "loss": 0.4926,
+      "epoch": 1.44,
+      "learning_rate": 5.567010309278351e-05,
+      "loss": 0.2695,
       "step": 1960
     },
     {
-      "epoch": 2.13,
-      "eval_accuracy": 0.6071817192600653,
-      "eval_loss": 1.3753130435943604,
-      "eval_runtime": 32.14,
-      "eval_samples_per_second": 57.187,
-      "eval_steps_per_second": 7.156,
+      "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,
       "step": 1960
     },
     {
-      "epoch": 2.14,
-      "learning_rate": 9.281828073993472e-05,
-      "loss": 0.5839,
+      "epoch": 1.45,
+      "learning_rate": 5.4933726067746695e-05,
+      "loss": 0.3932,
       "step": 1970
     },
     {
-      "epoch": 2.15,
-      "learning_rate": 9.227421109902067e-05,
-      "loss": 0.6893,
+      "epoch": 1.46,
+      "learning_rate": 5.4197349042709865e-05,
+      "loss": 0.3572,
       "step": 1980
     },
     {
-      "epoch": 2.17,
-      "learning_rate": 9.173014145810664e-05,
-      "loss": 0.5675,
+      "epoch": 1.47,
+      "learning_rate": 5.346097201767305e-05,
+      "loss": 0.5021,
       "step": 1990
     },
     {
-      "epoch": 2.18,
-      "learning_rate": 9.11860718171926e-05,
-      "loss": 0.6031,
+      "epoch": 1.47,
+      "learning_rate": 5.2724594992636234e-05,
+      "loss": 0.3038,
       "step": 2000
     },
     {
-      "epoch": 2.18,
-      "eval_accuracy": 0.5979325353645266,
-      "eval_loss": 1.3890224695205688,
-      "eval_runtime": 31.7313,
-      "eval_samples_per_second": 57.924,
-      "eval_steps_per_second": 7.248,
+      "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,
       "step": 2000
     },
     {
-      "epoch": 2.19,
-      "learning_rate": 9.064200217627857e-05,
-      "loss": 0.5095,
+      "epoch": 1.48,
+      "learning_rate": 5.198821796759942e-05,
+      "loss": 0.3817,
       "step": 2010
     },
     {
-      "epoch": 2.2,
-      "learning_rate": 9.009793253536452e-05,
-      "loss": 0.589,
+      "epoch": 1.49,
+      "learning_rate": 5.125184094256259e-05,
+      "loss": 0.3844,
       "step": 2020
     },
     {
-      "epoch": 2.21,
-      "learning_rate": 8.955386289445049e-05,
-      "loss": 0.667,
+      "epoch": 1.49,
+      "learning_rate": 5.051546391752577e-05,
+      "loss": 0.3217,
       "step": 2030
     },
     {
-      "epoch": 2.22,
-      "learning_rate": 8.900979325353645e-05,
-      "loss": 0.376,
+      "epoch": 1.5,
+      "learning_rate": 4.977908689248896e-05,
+      "loss": 0.4076,
       "step": 2040
     },
     {
-      "epoch": 2.22,
-      "eval_accuracy": 0.6191512513601741,
-      "eval_loss": 1.3675874471664429,
-      "eval_runtime": 31.4637,
-      "eval_samples_per_second": 58.417,
-      "eval_steps_per_second": 7.31,
+      "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,
       "step": 2040
     },
     {
-      "epoch": 2.23,
-      "learning_rate": 8.846572361262242e-05,
-      "loss": 0.5566,
+      "epoch": 1.51,
+      "learning_rate": 4.9042709867452135e-05,
+      "loss": 0.431,
       "step": 2050
     },
     {
-      "epoch": 2.24,
-      "learning_rate": 8.792165397170838e-05,
-      "loss": 0.5404,
+      "epoch": 1.52,
+      "learning_rate": 4.830633284241532e-05,
+      "loss": 0.3252,
       "step": 2060
     },
     {
-      "epoch": 2.25,
-      "learning_rate": 8.737758433079435e-05,
-      "loss": 0.7074,
+      "epoch": 1.52,
+      "learning_rate": 4.75699558173785e-05,
+      "loss": 0.3145,
       "step": 2070
     },
     {
-      "epoch": 2.26,
-      "learning_rate": 8.683351468988032e-05,
-      "loss": 0.5824,
+      "epoch": 1.53,
+      "learning_rate": 4.683357879234168e-05,
+      "loss": 0.4613,
       "step": 2080
     },
     {
-      "epoch": 2.26,
-      "eval_accuracy": 0.6066376496191512,
-      "eval_loss": 1.4152354001998901,
-      "eval_runtime": 31.6073,
-      "eval_samples_per_second": 58.151,
-      "eval_steps_per_second": 7.277,
+      "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,
       "step": 2080
     },
     {
-      "epoch": 2.27,
-      "learning_rate": 8.628944504896627e-05,
-      "loss": 0.5761,
+      "epoch": 1.54,
+      "learning_rate": 4.6097201767304866e-05,
+      "loss": 0.3371,
       "step": 2090
     },
     {
-      "epoch": 2.29,
-      "learning_rate": 8.574537540805223e-05,
-      "loss": 0.6037,
+      "epoch": 1.55,
+      "learning_rate": 4.536082474226804e-05,
+      "loss": 0.5341,
       "step": 2100
     },
     {
-      "epoch": 2.3,
-      "learning_rate": 8.52013057671382e-05,
-      "loss": 0.4611,
+      "epoch": 1.55,
+      "learning_rate": 4.462444771723123e-05,
+      "loss": 0.3712,
       "step": 2110
     },
     {
-      "epoch": 2.31,
-      "learning_rate": 8.465723612622416e-05,
-      "loss": 0.6753,
+      "epoch": 1.56,
+      "learning_rate": 4.3888070692194405e-05,
+      "loss": 0.3672,
       "step": 2120
     },
     {
-      "epoch": 2.31,
-      "eval_accuracy": 0.6099020674646355,
-      "eval_loss": 1.3810958862304688,
-      "eval_runtime": 31.5282,
-      "eval_samples_per_second": 58.297,
-      "eval_steps_per_second": 7.295,
+      "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,
       "step": 2120
     },
     {
-      "epoch": 2.32,
-      "learning_rate": 8.411316648531013e-05,
-      "loss": 0.5639,
+      "epoch": 1.57,
+      "learning_rate": 4.315169366715759e-05,
+      "loss": 0.3377,
       "step": 2130
     },
     {
-      "epoch": 2.33,
-      "learning_rate": 8.35690968443961e-05,
-      "loss": 0.5866,
+      "epoch": 1.58,
+      "learning_rate": 4.241531664212077e-05,
+      "loss": 0.377,
       "step": 2140
     },
     {
-      "epoch": 2.34,
-      "learning_rate": 8.302502720348205e-05,
-      "loss": 0.515,
+      "epoch": 1.58,
+      "learning_rate": 4.167893961708395e-05,
+      "loss": 0.3033,
       "step": 2150
     },
     {
-      "epoch": 2.35,
-      "learning_rate": 8.248095756256801e-05,
-      "loss": 0.5047,
+      "epoch": 1.59,
+      "learning_rate": 4.094256259204713e-05,
+      "loss": 0.4058,
       "step": 2160
     },
     {
-      "epoch": 2.35,
-      "eval_accuracy": 0.6082698585418934,
-      "eval_loss": 1.3635469675064087,
-      "eval_runtime": 31.4892,
-      "eval_samples_per_second": 58.369,
-      "eval_steps_per_second": 7.304,
+      "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,
       "step": 2160
     },
     {
-      "epoch": 2.36,
-      "learning_rate": 8.193688792165398e-05,
-      "loss": 0.5216,
+      "epoch": 1.6,
+      "learning_rate": 4.020618556701031e-05,
+      "loss": 0.2632,
       "step": 2170
     },
     {
-      "epoch": 2.37,
-      "learning_rate": 8.139281828073993e-05,
-      "loss": 0.4849,
+      "epoch": 1.61,
+      "learning_rate": 3.946980854197349e-05,
+      "loss": 0.3504,
       "step": 2180
     },
     {
-      "epoch": 2.38,
-      "learning_rate": 8.08487486398259e-05,
-      "loss": 0.561,
+      "epoch": 1.61,
+      "learning_rate": 3.8733431516936675e-05,
+      "loss": 0.2671,
       "step": 2190
     },
     {
-      "epoch": 2.39,
-      "learning_rate": 8.030467899891186e-05,
-      "loss": 0.4831,
+      "epoch": 1.62,
+      "learning_rate": 3.799705449189985e-05,
+      "loss": 0.2954,
       "step": 2200
     },
     {
-      "epoch": 2.39,
-      "eval_accuracy": 0.6039173014145811,
-      "eval_loss": 1.4076539278030396,
-      "eval_runtime": 32.0853,
-      "eval_samples_per_second": 57.285,
-      "eval_steps_per_second": 7.168,
+      "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,
       "step": 2200
     },
     {
-      "epoch": 2.4,
-      "learning_rate": 7.976060935799783e-05,
-      "loss": 0.5728,
+      "epoch": 1.63,
+      "learning_rate": 3.726067746686304e-05,
+      "loss": 0.3205,
       "step": 2210
     },
     {
-      "epoch": 2.42,
-      "learning_rate": 7.921653971708379e-05,
-      "loss": 0.444,
+      "epoch": 1.63,
+      "learning_rate": 3.6524300441826214e-05,
+      "loss": 0.299,
       "step": 2220
     },
     {
-      "epoch": 2.43,
-      "learning_rate": 7.867247007616976e-05,
-      "loss": 0.5329,
+      "epoch": 1.64,
+      "learning_rate": 3.57879234167894e-05,
+      "loss": 0.3119,
       "step": 2230
     },
     {
-      "epoch": 2.44,
-      "learning_rate": 7.812840043525572e-05,
-      "loss": 0.6054,
+      "epoch": 1.65,
+      "learning_rate": 3.5051546391752576e-05,
+      "loss": 0.3334,
       "step": 2240
     },
     {
-      "epoch": 2.44,
-      "eval_accuracy": 0.6175190424374319,
-      "eval_loss": 1.361071228981018,
-      "eval_runtime": 32.0081,
-      "eval_samples_per_second": 57.423,
-      "eval_steps_per_second": 7.186,
+      "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,
       "step": 2240
     },
     {
-      "epoch": 2.45,
-      "learning_rate": 7.758433079434167e-05,
-      "loss": 0.5624,
+      "epoch": 1.66,
+      "learning_rate": 3.431516936671576e-05,
+      "loss": 0.3469,
       "step": 2250
     },
     {
-      "epoch": 2.46,
-      "learning_rate": 7.704026115342764e-05,
-      "loss": 0.561,
+      "epoch": 1.66,
+      "learning_rate": 3.357879234167894e-05,
+      "loss": 0.2501,
       "step": 2260
     },
     {
-      "epoch": 2.47,
-      "learning_rate": 7.64961915125136e-05,
-      "loss": 0.5178,
+      "epoch": 1.67,
+      "learning_rate": 3.284241531664212e-05,
+      "loss": 0.3941,
       "step": 2270
     },
     {
-      "epoch": 2.48,
-      "learning_rate": 7.595212187159957e-05,
-      "loss": 0.5997,
+      "epoch": 1.68,
+      "learning_rate": 3.2106038291605307e-05,
+      "loss": 0.3151,
       "step": 2280
     },
     {
-      "epoch": 2.48,
-      "eval_accuracy": 0.6158868335146899,
-      "eval_loss": 1.3456101417541504,
-      "eval_runtime": 31.9844,
-      "eval_samples_per_second": 57.465,
-      "eval_steps_per_second": 7.191,
+      "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,
       "step": 2280
     },
     {
-      "epoch": 2.49,
-      "learning_rate": 7.540805223068554e-05,
-      "loss": 0.4876,
+      "epoch": 1.69,
+      "learning_rate": 3.1369661266568484e-05,
+      "loss": 0.3946,
       "step": 2290
     },
     {
-      "epoch": 2.5,
-      "learning_rate": 7.48639825897715e-05,
-      "loss": 0.5375,
+      "epoch": 1.69,
+      "learning_rate": 3.063328424153167e-05,
+      "loss": 0.3603,
       "step": 2300
     },
     {
-      "epoch": 2.51,
-      "learning_rate": 7.431991294885747e-05,
-      "loss": 0.5727,
+      "epoch": 1.7,
+      "learning_rate": 2.9896907216494846e-05,
+      "loss": 0.2634,
       "step": 2310
     },
     {
-      "epoch": 2.52,
-      "learning_rate": 7.377584330794342e-05,
-      "loss": 0.4648,
+      "epoch": 1.71,
+      "learning_rate": 2.916053019145803e-05,
+      "loss": 0.3322,
       "step": 2320
     },
     {
-      "epoch": 2.52,
-      "eval_accuracy": 0.6175190424374319,
-      "eval_loss": 1.3429198265075684,
-      "eval_runtime": 31.6074,
-      "eval_samples_per_second": 58.151,
-      "eval_steps_per_second": 7.277,
+      "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,
       "step": 2320
     },
     {
-      "epoch": 2.54,
-      "learning_rate": 7.323177366702938e-05,
-      "loss": 0.5728,
+      "epoch": 1.72,
+      "learning_rate": 2.8424153166421208e-05,
+      "loss": 0.3206,
       "step": 2330
     },
     {
-      "epoch": 2.55,
-      "learning_rate": 7.268770402611535e-05,
-      "loss": 0.4683,
+      "epoch": 1.72,
+      "learning_rate": 2.7687776141384392e-05,
+      "loss": 0.4357,
       "step": 2340
     },
     {
-      "epoch": 2.56,
-      "learning_rate": 7.21436343852013e-05,
-      "loss": 0.5095,
+      "epoch": 1.73,
+      "learning_rate": 2.695139911634757e-05,
+      "loss": 0.2878,
       "step": 2350
     },
     {
-      "epoch": 2.57,
-      "learning_rate": 7.159956474428727e-05,
-      "loss": 0.5885,
+      "epoch": 1.74,
+      "learning_rate": 2.6215022091310754e-05,
+      "loss": 0.4057,
       "step": 2360
     },
     {
-      "epoch": 2.57,
-      "eval_accuracy": 0.6218715995647442,
-      "eval_loss": 1.3531224727630615,
-      "eval_runtime": 31.6702,
-      "eval_samples_per_second": 58.036,
-      "eval_steps_per_second": 7.262,
+      "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,
       "step": 2360
     },
     {
-      "epoch": 2.58,
-      "learning_rate": 7.105549510337323e-05,
-      "loss": 0.483,
+      "epoch": 1.75,
+      "learning_rate": 2.547864506627393e-05,
+      "loss": 0.3853,
       "step": 2370
     },
     {
-      "epoch": 2.59,
-      "learning_rate": 7.05114254624592e-05,
-      "loss": 0.4248,
+      "epoch": 1.75,
+      "learning_rate": 2.4742268041237116e-05,
+      "loss": 0.3127,
       "step": 2380
     },
     {
-      "epoch": 2.6,
-      "learning_rate": 6.996735582154516e-05,
-      "loss": 0.4206,
+      "epoch": 1.76,
+      "learning_rate": 2.4005891016200297e-05,
+      "loss": 0.3005,
       "step": 2390
     },
     {
-      "epoch": 2.61,
-      "learning_rate": 6.942328618063112e-05,
-      "loss": 0.5405,
+      "epoch": 1.77,
+      "learning_rate": 2.3269513991163478e-05,
+      "loss": 0.3365,
       "step": 2400
     },
     {
-      "epoch": 2.61,
-      "eval_accuracy": 0.6273122959738846,
-      "eval_loss": 1.3313764333724976,
-      "eval_runtime": 31.9004,
-      "eval_samples_per_second": 57.617,
-      "eval_steps_per_second": 7.21,
+      "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,
       "step": 2400
     },
     {
-      "epoch": 2.62,
-      "learning_rate": 6.887921653971708e-05,
-      "loss": 0.4929,
+      "epoch": 1.77,
+      "learning_rate": 2.253313696612666e-05,
+      "loss": 0.285,
       "step": 2410
     },
     {
-      "epoch": 2.63,
-      "learning_rate": 6.833514689880305e-05,
-      "loss": 0.5165,
+      "epoch": 1.78,
+      "learning_rate": 2.179675994108984e-05,
+      "loss": 0.3114,
       "step": 2420
     },
     {
-      "epoch": 2.64,
-      "learning_rate": 6.779107725788901e-05,
-      "loss": 0.6003,
+      "epoch": 1.79,
+      "learning_rate": 2.106038291605302e-05,
+      "loss": 0.321,
       "step": 2430
     },
     {
-      "epoch": 2.66,
-      "learning_rate": 6.724700761697498e-05,
-      "loss": 0.5752,
+      "epoch": 1.8,
+      "learning_rate": 2.03240058910162e-05,
+      "loss": 0.2832,
       "step": 2440
     },
     {
-      "epoch": 2.66,
-      "eval_accuracy": 0.6343852013057671,
-      "eval_loss": 1.3247264623641968,
-      "eval_runtime": 31.5426,
-      "eval_samples_per_second": 58.27,
-      "eval_steps_per_second": 7.292,
+      "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,
       "step": 2440
     },
     {
-      "epoch": 2.67,
-      "learning_rate": 6.670293797606094e-05,
-      "loss": 0.4671,
+      "epoch": 1.8,
+      "learning_rate": 1.9587628865979382e-05,
+      "loss": 0.3756,
       "step": 2450
     },
     {
-      "epoch": 2.68,
-      "learning_rate": 6.615886833514691e-05,
-      "loss": 0.4499,
+      "epoch": 1.81,
+      "learning_rate": 1.8851251840942563e-05,
+      "loss": 0.2677,
       "step": 2460
     },
     {
-      "epoch": 2.69,
-      "learning_rate": 6.561479869423286e-05,
-      "loss": 0.5622,
+      "epoch": 1.82,
+      "learning_rate": 1.8114874815905744e-05,
+      "loss": 0.392,
       "step": 2470
     },
     {
-      "epoch": 2.7,
-      "learning_rate": 6.507072905331883e-05,
-      "loss": 0.5156,
+      "epoch": 1.83,
+      "learning_rate": 1.7378497790868925e-05,
+      "loss": 0.2512,
       "step": 2480
     },
     {
-      "epoch": 2.7,
-      "eval_accuracy": 0.6322089227421109,
-      "eval_loss": 1.3440065383911133,
-      "eval_runtime": 32.1302,
-      "eval_samples_per_second": 57.205,
-      "eval_steps_per_second": 7.158,
+      "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,
       "step": 2480
     },
     {
-      "epoch": 2.71,
-      "learning_rate": 6.452665941240479e-05,
-      "loss": 0.4567,
+      "epoch": 1.83,
+      "learning_rate": 1.6642120765832106e-05,
+      "loss": 0.3217,
       "step": 2490
     },
     {
-      "epoch": 2.72,
-      "learning_rate": 6.398258977149076e-05,
-      "loss": 0.6599,
+      "epoch": 1.84,
+      "learning_rate": 1.5905743740795287e-05,
+      "loss": 0.3821,
       "step": 2500
     },
     {
-      "epoch": 2.73,
-      "learning_rate": 6.343852013057672e-05,
-      "loss": 0.4837,
+      "epoch": 1.85,
+      "learning_rate": 1.5169366715758468e-05,
+      "loss": 0.2863,
       "step": 2510
     },
     {
-      "epoch": 2.74,
-      "learning_rate": 6.289445048966269e-05,
-      "loss": 0.5518,
+      "epoch": 1.86,
+      "learning_rate": 1.4432989690721649e-05,
+      "loss": 0.2347,
       "step": 2520
     },
     {
-      "epoch": 2.74,
-      "eval_accuracy": 0.6338411316648531,
-      "eval_loss": 1.315151333808899,
-      "eval_runtime": 31.8096,
-      "eval_samples_per_second": 57.781,
-      "eval_steps_per_second": 7.231,
+      "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,
       "step": 2520
     },
     {
-      "epoch": 2.75,
-      "learning_rate": 6.235038084874864e-05,
-      "loss": 0.4855,
+      "epoch": 1.86,
+      "learning_rate": 1.3696612665684833e-05,
+      "loss": 0.1994,
       "step": 2530
     },
     {
-      "epoch": 2.76,
-      "learning_rate": 6.18063112078346e-05,
-      "loss": 0.4465,
+      "epoch": 1.87,
+      "learning_rate": 1.2960235640648014e-05,
+      "loss": 0.2408,
       "step": 2540
     },
     {
-      "epoch": 2.77,
-      "learning_rate": 6.126224156692056e-05,
-      "loss": 0.5561,
+      "epoch": 1.88,
+      "learning_rate": 1.2223858615611193e-05,
+      "loss": 0.4364,
       "step": 2550
     },
     {
-      "epoch": 2.79,
-      "learning_rate": 6.071817192600653e-05,
-      "loss": 0.5928,
+      "epoch": 1.89,
+      "learning_rate": 1.1487481590574374e-05,
+      "loss": 0.1912,
       "step": 2560
     },
     {
-      "epoch": 2.79,
-      "eval_accuracy": 0.6398258977149075,
-      "eval_loss": 1.2906804084777832,
-      "eval_runtime": 31.6667,
-      "eval_samples_per_second": 58.042,
-      "eval_steps_per_second": 7.263,
+      "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,
       "step": 2560
     },
     {
-      "epoch": 2.8,
-      "learning_rate": 6.0174102285092495e-05,
-      "loss": 0.5153,
+      "epoch": 1.89,
+      "learning_rate": 1.0751104565537557e-05,
+      "loss": 0.3227,
       "step": 2570
     },
     {
-      "epoch": 2.81,
-      "learning_rate": 5.9630032644178454e-05,
-      "loss": 0.4992,
+      "epoch": 1.9,
+      "learning_rate": 1.0014727540500737e-05,
+      "loss": 0.2769,
       "step": 2580
     },
     {
-      "epoch": 2.82,
-      "learning_rate": 5.908596300326442e-05,
-      "loss": 0.5095,
+      "epoch": 1.91,
+      "learning_rate": 9.278350515463918e-06,
+      "loss": 0.2485,
       "step": 2590
     },
     {
-      "epoch": 2.83,
-      "learning_rate": 5.8541893362350385e-05,
-      "loss": 0.5283,
+      "epoch": 1.91,
+      "learning_rate": 8.5419734904271e-06,
+      "loss": 0.3679,
       "step": 2600
     },
     {
-      "epoch": 2.83,
-      "eval_accuracy": 0.6452665941240479,
-      "eval_loss": 1.2707958221435547,
-      "eval_runtime": 31.5986,
-      "eval_samples_per_second": 58.167,
-      "eval_steps_per_second": 7.279,
+      "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,
       "step": 2600
     },
     {
-      "epoch": 2.84,
-      "learning_rate": 5.799782372143635e-05,
-      "loss": 0.4745,
+      "epoch": 1.92,
+      "learning_rate": 7.80559646539028e-06,
+      "loss": 0.3111,
       "step": 2610
     },
     {
-      "epoch": 2.85,
-      "learning_rate": 5.7453754080522316e-05,
-      "loss": 0.503,
+      "epoch": 1.93,
+      "learning_rate": 7.069219440353461e-06,
+      "loss": 0.33,
       "step": 2620
     },
     {
-      "epoch": 2.86,
-      "learning_rate": 5.690968443960827e-05,
-      "loss": 0.3843,
+      "epoch": 1.94,
+      "learning_rate": 6.332842415316642e-06,
+      "loss": 0.2817,
       "step": 2630
     },
     {
-      "epoch": 2.87,
-      "learning_rate": 5.636561479869423e-05,
-      "loss": 0.5054,
+      "epoch": 1.94,
+      "learning_rate": 5.596465390279824e-06,
+      "loss": 0.2953,
       "step": 2640
     },
     {
-      "epoch": 2.87,
-      "eval_accuracy": 0.6332970620239391,
-      "eval_loss": 1.2805439233779907,
-      "eval_runtime": 31.8568,
-      "eval_samples_per_second": 57.696,
-      "eval_steps_per_second": 7.22,
+      "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,
       "step": 2640
     },
     {
-      "epoch": 2.88,
-      "learning_rate": 5.58215451577802e-05,
-      "loss": 0.4127,
+      "epoch": 1.95,
+      "learning_rate": 4.860088365243005e-06,
+      "loss": 0.247,
       "step": 2650
     },
     {
-      "epoch": 2.89,
-      "learning_rate": 5.5277475516866164e-05,
-      "loss": 0.4589,
+      "epoch": 1.96,
+      "learning_rate": 4.123711340206186e-06,
+      "loss": 0.3024,
       "step": 2660
     },
     {
-      "epoch": 2.91,
-      "learning_rate": 5.473340587595212e-05,
-      "loss": 0.6578,
+      "epoch": 1.97,
+      "learning_rate": 3.3873343151693666e-06,
+      "loss": 0.2232,
       "step": 2670
     },
     {
-      "epoch": 2.92,
-      "learning_rate": 5.418933623503809e-05,
-      "loss": 0.4513,
+      "epoch": 1.97,
+      "learning_rate": 2.650957290132548e-06,
+      "loss": 0.2982,
       "step": 2680
     },
     {
-      "epoch": 2.92,
-      "eval_accuracy": 0.6360174102285092,
-      "eval_loss": 1.304470419883728,
-      "eval_runtime": 32.1887,
-      "eval_samples_per_second": 57.101,
-      "eval_steps_per_second": 7.145,
+      "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,
       "step": 2680
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 5.3645266594124054e-05,
-      "loss": 0.5623,
-      "step": 2690
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 5.3101196953210006e-05,
-      "loss": 0.6832,
-      "step": 2700
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 5.255712731229597e-05,
-      "loss": 0.417,
-      "step": 2710
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 5.201305767138194e-05,
-      "loss": 0.4405,
-      "step": 2720
-    },
-    {
-      "epoch": 2.96,
-      "eval_accuracy": 0.6425462459194777,
-      "eval_loss": 1.3049399852752686,
-      "eval_runtime": 32.0538,
-      "eval_samples_per_second": 57.341,
-      "eval_steps_per_second": 7.175,
-      "step": 2720
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 5.14689880304679e-05,
-      "loss": 0.5936,
-      "step": 2730
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 5.092491838955387e-05,
-      "loss": 0.411,
-      "step": 2740
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 5.038084874863983e-05,
-      "loss": 0.4975,
-      "step": 2750
-    },
-    {
-      "epoch": 3.0,
-      "learning_rate": 4.983677910772579e-05,
-      "loss": 0.3695,
-      "step": 2760
-    },
-    {
-      "epoch": 3.0,
-      "eval_accuracy": 0.6398258977149075,
-      "eval_loss": 1.3069652318954468,
-      "eval_runtime": 31.9114,
-      "eval_samples_per_second": 57.597,
-      "eval_steps_per_second": 7.207,
-      "step": 2760
-    },
-    {
-      "epoch": 3.01,
-      "learning_rate": 4.929270946681175e-05,
-      "loss": 0.2096,
-      "step": 2770
-    },
-    {
-      "epoch": 3.03,
-      "learning_rate": 4.8748639825897716e-05,
-      "loss": 0.134,
-      "step": 2780
-    },
-    {
-      "epoch": 3.04,
-      "learning_rate": 4.820457018498368e-05,
-      "loss": 0.1749,
-      "step": 2790
-    },
-    {
-      "epoch": 3.05,
-      "learning_rate": 4.766050054406964e-05,
-      "loss": 0.2109,
-      "step": 2800
-    },
-    {
-      "epoch": 3.05,
-      "eval_accuracy": 0.6414581066376496,
-      "eval_loss": 1.3126190900802612,
-      "eval_runtime": 31.6295,
-      "eval_samples_per_second": 58.11,
-      "eval_steps_per_second": 7.272,
-      "step": 2800
-    },
-    {
-      "epoch": 3.06,
-      "learning_rate": 4.7116430903155606e-05,
-      "loss": 0.1672,
-      "step": 2810
-    },
-    {
-      "epoch": 3.07,
-      "learning_rate": 4.657236126224157e-05,
-      "loss": 0.1398,
-      "step": 2820
-    },
-    {
-      "epoch": 3.08,
-      "learning_rate": 4.602829162132753e-05,
-      "loss": 0.2211,
-      "step": 2830
-    },
-    {
-      "epoch": 3.09,
-      "learning_rate": 4.5484221980413496e-05,
-      "loss": 0.1899,
-      "step": 2840
-    },
-    {
-      "epoch": 3.09,
-      "eval_accuracy": 0.6517954298150164,
-      "eval_loss": 1.3004298210144043,
-      "eval_runtime": 31.8608,
-      "eval_samples_per_second": 57.688,
-      "eval_steps_per_second": 7.219,
-      "step": 2840
-    },
-    {
-      "epoch": 3.1,
-      "learning_rate": 4.494015233949946e-05,
-      "loss": 0.1822,
-      "step": 2850
-    },
-    {
-      "epoch": 3.11,
-      "learning_rate": 4.439608269858542e-05,
-      "loss": 0.1346,
-      "step": 2860
-    },
-    {
-      "epoch": 3.12,
-      "learning_rate": 4.385201305767138e-05,
-      "loss": 0.1895,
-      "step": 2870
-    },
-    {
-      "epoch": 3.13,
-      "learning_rate": 4.3307943416757344e-05,
-      "loss": 0.1338,
-      "step": 2880
-    },
-    {
-      "epoch": 3.13,
-      "eval_accuracy": 0.6507072905331882,
-      "eval_loss": 1.292977213859558,
-      "eval_runtime": 32.2424,
-      "eval_samples_per_second": 57.006,
-      "eval_steps_per_second": 7.133,
-      "step": 2880
-    },
-    {
-      "epoch": 3.14,
-      "learning_rate": 4.276387377584331e-05,
-      "loss": 0.1475,
-      "step": 2890
-    },
-    {
-      "epoch": 3.16,
-      "learning_rate": 4.2219804134929275e-05,
-      "loss": 0.1118,
-      "step": 2900
-    },
-    {
-      "epoch": 3.17,
-      "learning_rate": 4.1675734494015234e-05,
-      "loss": 0.2001,
-      "step": 2910
-    },
-    {
-      "epoch": 3.18,
-      "learning_rate": 4.11316648531012e-05,
-      "loss": 0.1681,
-      "step": 2920
-    },
-    {
-      "epoch": 3.18,
-      "eval_accuracy": 0.6452665941240479,
-      "eval_loss": 1.338887095451355,
-      "eval_runtime": 32.1012,
-      "eval_samples_per_second": 57.256,
-      "eval_steps_per_second": 7.165,
-      "step": 2920
-    },
-    {
-      "epoch": 3.19,
-      "learning_rate": 4.0587595212187165e-05,
-      "loss": 0.1207,
-      "step": 2930
-    },
-    {
-      "epoch": 3.2,
-      "learning_rate": 4.004352557127313e-05,
-      "loss": 0.158,
-      "step": 2940
-    },
-    {
-      "epoch": 3.21,
-      "learning_rate": 3.949945593035909e-05,
-      "loss": 0.1602,
-      "step": 2950
-    },
-    {
-      "epoch": 3.22,
-      "learning_rate": 3.895538628944505e-05,
-      "loss": 0.156,
-      "step": 2960
-    },
-    {
-      "epoch": 3.22,
-      "eval_accuracy": 0.6430903155603918,
-      "eval_loss": 1.3379881381988525,
-      "eval_runtime": 31.8038,
-      "eval_samples_per_second": 57.792,
-      "eval_steps_per_second": 7.232,
-      "step": 2960
-    },
-    {
-      "epoch": 3.23,
-      "learning_rate": 3.841131664853101e-05,
-      "loss": 0.2014,
-      "step": 2970
-    },
-    {
-      "epoch": 3.24,
-      "learning_rate": 3.786724700761698e-05,
-      "loss": 0.1704,
-      "step": 2980
-    },
-    {
-      "epoch": 3.25,
-      "learning_rate": 3.732317736670294e-05,
-      "loss": 0.2522,
-      "step": 2990
-    },
-    {
-      "epoch": 3.26,
-      "learning_rate": 3.67791077257889e-05,
-      "loss": 0.1717,
-      "step": 3000
-    },
-    {
-      "epoch": 3.26,
-      "eval_accuracy": 0.6371055495103374,
-      "eval_loss": 1.3541066646575928,
-      "eval_runtime": 32.1836,
-      "eval_samples_per_second": 57.11,
-      "eval_steps_per_second": 7.147,
-      "step": 3000
-    },
-    {
-      "epoch": 3.28,
-      "learning_rate": 3.623503808487487e-05,
-      "loss": 0.1593,
-      "step": 3010
-    },
-    {
-      "epoch": 3.29,
-      "learning_rate": 3.569096844396083e-05,
-      "loss": 0.1803,
-      "step": 3020
-    },
-    {
-      "epoch": 3.3,
-      "learning_rate": 3.514689880304679e-05,
-      "loss": 0.1687,
-      "step": 3030
-    },
-    {
-      "epoch": 3.31,
-      "learning_rate": 3.460282916213276e-05,
-      "loss": 0.1244,
-      "step": 3040
-    },
-    {
-      "epoch": 3.31,
-      "eval_accuracy": 0.6420021762785637,
-      "eval_loss": 1.3497653007507324,
-      "eval_runtime": 32.2339,
-      "eval_samples_per_second": 57.021,
-      "eval_steps_per_second": 7.135,
-      "step": 3040
-    },
-    {
-      "epoch": 3.32,
-      "learning_rate": 3.4058759521218717e-05,
-      "loss": 0.1832,
-      "step": 3050
-    },
-    {
-      "epoch": 3.33,
-      "learning_rate": 3.3514689880304675e-05,
-      "loss": 0.1061,
-      "step": 3060
-    },
-    {
-      "epoch": 3.34,
-      "learning_rate": 3.297062023939064e-05,
-      "loss": 0.1092,
-      "step": 3070
-    },
-    {
-      "epoch": 3.35,
-      "learning_rate": 3.2426550598476606e-05,
-      "loss": 0.1148,
-      "step": 3080
-    },
-    {
-      "epoch": 3.35,
-      "eval_accuracy": 0.6566920565832427,
-      "eval_loss": 1.3009717464447021,
-      "eval_runtime": 31.9059,
-      "eval_samples_per_second": 57.607,
-      "eval_steps_per_second": 7.209,
-      "step": 3080
-    },
-    {
-      "epoch": 3.36,
-      "learning_rate": 3.188248095756257e-05,
-      "loss": 0.1352,
-      "step": 3090
-    },
-    {
-      "epoch": 3.37,
-      "learning_rate": 3.133841131664853e-05,
-      "loss": 0.1482,
-      "step": 3100
-    },
-    {
-      "epoch": 3.38,
-      "learning_rate": 3.0794341675734496e-05,
-      "loss": 0.183,
-      "step": 3110
-    },
-    {
-      "epoch": 3.39,
-      "learning_rate": 3.0250272034820458e-05,
-      "loss": 0.1223,
-      "step": 3120
-    },
-    {
-      "epoch": 3.39,
-      "eval_accuracy": 0.6436343852013058,
-      "eval_loss": 1.3242275714874268,
-      "eval_runtime": 32.4153,
-      "eval_samples_per_second": 56.702,
-      "eval_steps_per_second": 7.095,
-      "step": 3120
-    },
-    {
-      "epoch": 3.41,
-      "learning_rate": 2.9706202393906423e-05,
-      "loss": 0.1757,
-      "step": 3130
-    },
-    {
-      "epoch": 3.42,
-      "learning_rate": 2.9162132752992382e-05,
-      "loss": 0.2056,
-      "step": 3140
-    },
-    {
-      "epoch": 3.43,
-      "learning_rate": 2.8618063112078348e-05,
-      "loss": 0.0992,
-      "step": 3150
-    },
-    {
-      "epoch": 3.44,
-      "learning_rate": 2.8073993471164313e-05,
-      "loss": 0.0822,
-      "step": 3160
-    },
-    {
-      "epoch": 3.44,
-      "eval_accuracy": 0.6485310119695321,
-      "eval_loss": 1.3289127349853516,
-      "eval_runtime": 32.0196,
-      "eval_samples_per_second": 57.402,
-      "eval_steps_per_second": 7.183,
-      "step": 3160
-    },
-    {
-      "epoch": 3.45,
-      "learning_rate": 2.7529923830250275e-05,
-      "loss": 0.1514,
-      "step": 3170
-    },
-    {
-      "epoch": 3.46,
-      "learning_rate": 2.6985854189336234e-05,
-      "loss": 0.1512,
-      "step": 3180
-    },
-    {
-      "epoch": 3.47,
-      "learning_rate": 2.64417845484222e-05,
-      "loss": 0.2447,
-      "step": 3190
-    },
-    {
-      "epoch": 3.48,
-      "learning_rate": 2.5897714907508165e-05,
-      "loss": 0.2122,
-      "step": 3200
-    },
-    {
-      "epoch": 3.48,
-      "eval_accuracy": 0.6452665941240479,
-      "eval_loss": 1.3312246799468994,
-      "eval_runtime": 31.8811,
-      "eval_samples_per_second": 57.652,
-      "eval_steps_per_second": 7.214,
-      "step": 3200
-    },
-    {
-      "epoch": 3.49,
-      "learning_rate": 2.5353645266594124e-05,
-      "loss": 0.245,
-      "step": 3210
-    },
-    {
-      "epoch": 3.5,
-      "learning_rate": 2.4809575625680086e-05,
-      "loss": 0.169,
-      "step": 3220
-    },
-    {
-      "epoch": 3.51,
-      "learning_rate": 2.426550598476605e-05,
-      "loss": 0.1088,
-      "step": 3230
-    },
-    {
-      "epoch": 3.53,
-      "learning_rate": 2.3721436343852013e-05,
-      "loss": 0.1107,
-      "step": 3240
-    },
-    {
-      "epoch": 3.53,
-      "eval_accuracy": 0.6539717083786725,
-      "eval_loss": 1.3308542966842651,
-      "eval_runtime": 31.9205,
-      "eval_samples_per_second": 57.581,
-      "eval_steps_per_second": 7.205,
-      "step": 3240
-    },
-    {
-      "epoch": 3.54,
-      "learning_rate": 2.317736670293798e-05,
-      "loss": 0.1956,
-      "step": 3250
-    },
-    {
-      "epoch": 3.55,
-      "learning_rate": 2.263329706202394e-05,
-      "loss": 0.2284,
-      "step": 3260
-    },
-    {
-      "epoch": 3.56,
-      "learning_rate": 2.2089227421109903e-05,
-      "loss": 0.2057,
-      "step": 3270
-    },
-    {
-      "epoch": 3.57,
-      "learning_rate": 2.1545157780195865e-05,
-      "loss": 0.1542,
-      "step": 3280
-    },
-    {
-      "epoch": 3.57,
-      "eval_accuracy": 0.6517954298150164,
-      "eval_loss": 1.3291780948638916,
-      "eval_runtime": 31.9281,
-      "eval_samples_per_second": 57.567,
-      "eval_steps_per_second": 7.204,
-      "step": 3280
-    },
-    {
-      "epoch": 3.58,
-      "learning_rate": 2.100108813928183e-05,
-      "loss": 0.2547,
-      "step": 3290
-    },
-    {
-      "epoch": 3.59,
-      "learning_rate": 2.0457018498367793e-05,
-      "loss": 0.1693,
-      "step": 3300
-    },
-    {
-      "epoch": 3.6,
-      "learning_rate": 1.9912948857453755e-05,
-      "loss": 0.1856,
-      "step": 3310
-    },
-    {
-      "epoch": 3.61,
-      "learning_rate": 1.9368879216539717e-05,
-      "loss": 0.1932,
-      "step": 3320
-    },
-    {
-      "epoch": 3.61,
-      "eval_accuracy": 0.6485310119695321,
-      "eval_loss": 1.3065351247787476,
-      "eval_runtime": 32.2802,
-      "eval_samples_per_second": 56.939,
-      "eval_steps_per_second": 7.125,
-      "step": 3320
-    },
-    {
-      "epoch": 3.62,
-      "learning_rate": 1.8824809575625682e-05,
-      "loss": 0.1828,
-      "step": 3330
-    },
-    {
-      "epoch": 3.63,
-      "learning_rate": 1.8280739934711644e-05,
-      "loss": 0.1613,
-      "step": 3340
-    },
-    {
-      "epoch": 3.65,
-      "learning_rate": 1.7736670293797607e-05,
-      "loss": 0.1393,
-      "step": 3350
-    },
-    {
-      "epoch": 3.66,
-      "learning_rate": 1.719260065288357e-05,
-      "loss": 0.2909,
-      "step": 3360
-    },
-    {
-      "epoch": 3.66,
-      "eval_accuracy": 0.6545157780195865,
-      "eval_loss": 1.309682846069336,
-      "eval_runtime": 31.9846,
-      "eval_samples_per_second": 57.465,
-      "eval_steps_per_second": 7.191,
-      "step": 3360
-    },
-    {
-      "epoch": 3.67,
-      "learning_rate": 1.6648531011969534e-05,
-      "loss": 0.1763,
-      "step": 3370
-    },
-    {
-      "epoch": 3.68,
-      "learning_rate": 1.6104461371055496e-05,
-      "loss": 0.1939,
-      "step": 3380
-    },
-    {
-      "epoch": 3.69,
-      "learning_rate": 1.5560391730141458e-05,
-      "loss": 0.2101,
-      "step": 3390
-    },
-    {
-      "epoch": 3.7,
-      "learning_rate": 1.5016322089227422e-05,
-      "loss": 0.1504,
-      "step": 3400
-    },
-    {
-      "epoch": 3.7,
-      "eval_accuracy": 0.6545157780195865,
-      "eval_loss": 1.3114700317382812,
-      "eval_runtime": 32.1163,
-      "eval_samples_per_second": 57.23,
-      "eval_steps_per_second": 7.161,
-      "step": 3400
-    },
-    {
-      "epoch": 3.71,
-      "learning_rate": 1.4472252448313384e-05,
-      "loss": 0.1275,
-      "step": 3410
-    },
-    {
-      "epoch": 3.72,
-      "learning_rate": 1.3928182807399348e-05,
-      "loss": 0.1111,
-      "step": 3420
-    },
-    {
-      "epoch": 3.73,
-      "learning_rate": 1.338411316648531e-05,
-      "loss": 0.1553,
-      "step": 3430
-    },
-    {
-      "epoch": 3.74,
-      "learning_rate": 1.2840043525571274e-05,
-      "loss": 0.1823,
-      "step": 3440
-    },
-    {
-      "epoch": 3.74,
-      "eval_accuracy": 0.6545157780195865,
-      "eval_loss": 1.320819616317749,
-      "eval_runtime": 31.7036,
-      "eval_samples_per_second": 57.975,
-      "eval_steps_per_second": 7.255,
-      "step": 3440
-    },
-    {
-      "epoch": 3.75,
-      "learning_rate": 1.2295973884657238e-05,
-      "loss": 0.1772,
-      "step": 3450
-    },
-    {
-      "epoch": 3.76,
-      "learning_rate": 1.17519042437432e-05,
-      "loss": 0.2327,
-      "step": 3460
-    },
-    {
-      "epoch": 3.78,
-      "learning_rate": 1.1207834602829162e-05,
-      "loss": 0.2037,
-      "step": 3470
-    },
-    {
-      "epoch": 3.79,
-      "learning_rate": 1.0663764961915126e-05,
-      "loss": 0.2477,
-      "step": 3480
-    },
-    {
-      "epoch": 3.79,
-      "eval_accuracy": 0.6523394994559304,
-      "eval_loss": 1.3207744359970093,
-      "eval_runtime": 31.8189,
-      "eval_samples_per_second": 57.765,
-      "eval_steps_per_second": 7.228,
-      "step": 3480
-    },
-    {
-      "epoch": 3.8,
-      "learning_rate": 1.0119695321001088e-05,
-      "loss": 0.1393,
-      "step": 3490
-    },
-    {
-      "epoch": 3.81,
-      "learning_rate": 9.575625680087052e-06,
-      "loss": 0.2626,
-      "step": 3500
-    },
-    {
-      "epoch": 3.82,
-      "learning_rate": 9.031556039173014e-06,
-      "loss": 0.1723,
-      "step": 3510
-    },
-    {
-      "epoch": 3.83,
-      "learning_rate": 8.487486398258977e-06,
-      "loss": 0.1264,
-      "step": 3520
-    },
-    {
-      "epoch": 3.83,
-      "eval_accuracy": 0.6539717083786725,
-      "eval_loss": 1.324898362159729,
-      "eval_runtime": 31.8821,
-      "eval_samples_per_second": 57.65,
-      "eval_steps_per_second": 7.214,
-      "step": 3520
-    },
-    {
-      "epoch": 3.84,
-      "learning_rate": 7.94341675734494e-06,
-      "loss": 0.1544,
-      "step": 3530
-    },
-    {
-      "epoch": 3.85,
-      "learning_rate": 7.399347116430903e-06,
-      "loss": 0.1696,
-      "step": 3540
-    },
-    {
-      "epoch": 3.86,
-      "learning_rate": 6.855277475516866e-06,
-      "loss": 0.2394,
-      "step": 3550
-    },
-    {
-      "epoch": 3.87,
-      "learning_rate": 6.311207834602829e-06,
-      "loss": 0.1003,
-      "step": 3560
-    },
-    {
-      "epoch": 3.87,
-      "eval_accuracy": 0.6599564744287268,
-      "eval_loss": 1.3238483667373657,
-      "eval_runtime": 32.249,
-      "eval_samples_per_second": 56.994,
-      "eval_steps_per_second": 7.132,
-      "step": 3560
-    },
-    {
-      "epoch": 3.88,
-      "learning_rate": 5.767138193688792e-06,
-      "loss": 0.2542,
-      "step": 3570
-    },
-    {
-      "epoch": 3.9,
-      "learning_rate": 5.223068552774756e-06,
-      "loss": 0.142,
-      "step": 3580
-    },
-    {
-      "epoch": 3.91,
-      "learning_rate": 4.678998911860719e-06,
-      "loss": 0.1282,
-      "step": 3590
-    },
-    {
-      "epoch": 3.92,
-      "learning_rate": 4.134929270946682e-06,
-      "loss": 0.1335,
-      "step": 3600
-    },
-    {
-      "epoch": 3.92,
-      "eval_accuracy": 0.6610446137105549,
-      "eval_loss": 1.3172351121902466,
-      "eval_runtime": 31.9012,
-      "eval_samples_per_second": 57.615,
-      "eval_steps_per_second": 7.21,
-      "step": 3600
-    },
-    {
-      "epoch": 3.93,
-      "learning_rate": 3.5908596300326447e-06,
-      "loss": 0.1675,
-      "step": 3610
-    },
-    {
-      "epoch": 3.94,
-      "learning_rate": 3.0467899891186073e-06,
-      "loss": 0.0947,
-      "step": 3620
-    },
-    {
-      "epoch": 3.95,
-      "learning_rate": 2.5027203482045706e-06,
-      "loss": 0.1827,
-      "step": 3630
-    },
-    {
-      "epoch": 3.96,
-      "learning_rate": 1.958650707290533e-06,
-      "loss": 0.2452,
-      "step": 3640
-    },
-    {
-      "epoch": 3.96,
-      "eval_accuracy": 0.6556039173014145,
-      "eval_loss": 1.3156225681304932,
-      "eval_runtime": 32.1002,
-      "eval_samples_per_second": 57.258,
-      "eval_steps_per_second": 7.165,
-      "step": 3640
     }
   ],
-  "max_steps": 3676,
-  "num_train_epochs": 4,
-  "total_flos": 4.5132044744502374e+18,
+  "max_steps": 2716,
+  "num_train_epochs": 2,
+  "total_flos": 3.322629783093584e+18,
   "trial_name": null,
   "trial_params": null
 }