diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,4817 +1,4583 @@
 {
-  "best_metric": 0.8525345622119815,
-  "best_model_checkpoint": "videomae-base-finetuned-subset\\checkpoint-5900",
-  "epoch": 59.00826210826211,
+  "best_metric": 0.7649769585253456,
+  "best_model_checkpoint": "videomae-base-finetuned-subset\\checkpoint-2016",
+  "epoch": 59.00780780780781,
   "eval_steps": 500,
-  "global_step": 7020,
+  "global_step": 6660,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.0,
-      "learning_rate": 7.122507122507123e-07,
-      "loss": 1.1155,
+      "learning_rate": 7.507507507507508e-07,
+      "loss": 1.7044,
       "step": 10
     },
     {
       "epoch": 0.0,
-      "learning_rate": 1.4245014245014246e-06,
-      "loss": 1.1405,
+      "learning_rate": 1.5015015015015015e-06,
+      "loss": 1.6581,
       "step": 20
     },
     {
       "epoch": 0.0,
-      "learning_rate": 2.136752136752137e-06,
-      "loss": 0.5289,
+      "learning_rate": 2.2522522522522524e-06,
+      "loss": 1.5572,
       "step": 30
     },
     {
       "epoch": 0.01,
-      "learning_rate": 2.8490028490028492e-06,
-      "loss": 0.8832,
+      "learning_rate": 3.003003003003003e-06,
+      "loss": 1.6749,
       "step": 40
     },
     {
       "epoch": 0.01,
-      "learning_rate": 3.5612535612535615e-06,
-      "loss": 1.0172,
+      "learning_rate": 3.753753753753754e-06,
+      "loss": 1.6511,
       "step": 50
     },
     {
       "epoch": 0.01,
-      "learning_rate": 4.273504273504274e-06,
-      "loss": 0.7363,
+      "learning_rate": 4.504504504504505e-06,
+      "loss": 1.6406,
       "step": 60
     },
     {
       "epoch": 0.01,
-      "learning_rate": 4.985754985754986e-06,
-      "loss": 0.8444,
+      "learning_rate": 5.255255255255255e-06,
+      "loss": 1.6205,
       "step": 70
     },
     {
       "epoch": 0.01,
-      "learning_rate": 5.6980056980056985e-06,
-      "loss": 0.8984,
+      "learning_rate": 6.006006006006006e-06,
+      "loss": 1.707,
       "step": 80
     },
     {
       "epoch": 0.01,
-      "learning_rate": 6.41025641025641e-06,
-      "loss": 1.085,
+      "learning_rate": 6.7567567567567575e-06,
+      "loss": 1.5589,
       "step": 90
     },
     {
-      "epoch": 0.01,
-      "learning_rate": 7.122507122507123e-06,
-      "loss": 0.7478,
+      "epoch": 0.02,
+      "learning_rate": 7.507507507507508e-06,
+      "loss": 1.6865,
       "step": 100
     },
     {
       "epoch": 0.02,
-      "learning_rate": 7.834757834757835e-06,
-      "loss": 0.9225,
+      "learning_rate": 8.258258258258259e-06,
+      "loss": 1.6074,
       "step": 110
     },
     {
       "epoch": 0.02,
-      "eval_accuracy": 0.7419354838709677,
-      "eval_loss": 0.7362823486328125,
-      "eval_runtime": 289.676,
-      "eval_samples_per_second": 0.749,
-      "eval_steps_per_second": 0.19,
-      "step": 118
+      "eval_accuracy": 0.3686635944700461,
+      "eval_loss": 1.5690008401870728,
+      "eval_runtime": 253.0884,
+      "eval_samples_per_second": 0.857,
+      "eval_steps_per_second": 0.217,
+      "step": 112
     },
     {
       "epoch": 1.0,
-      "learning_rate": 8.547008547008548e-06,
-      "loss": 0.8071,
+      "learning_rate": 9.00900900900901e-06,
+      "loss": 1.5636,
       "step": 120
     },
     {
       "epoch": 1.0,
-      "learning_rate": 9.259259259259259e-06,
-      "loss": 0.4488,
+      "learning_rate": 9.75975975975976e-06,
+      "loss": 1.6139,
       "step": 130
     },
     {
       "epoch": 1.0,
-      "learning_rate": 9.971509971509972e-06,
-      "loss": 1.23,
+      "learning_rate": 1.051051051051051e-05,
+      "loss": 1.591,
       "step": 140
     },
     {
-      "epoch": 1.0,
-      "learning_rate": 1.0683760683760684e-05,
-      "loss": 0.7655,
+      "epoch": 1.01,
+      "learning_rate": 1.1261261261261261e-05,
+      "loss": 1.5824,
       "step": 150
     },
     {
       "epoch": 1.01,
-      "learning_rate": 1.1396011396011397e-05,
-      "loss": 0.8764,
+      "learning_rate": 1.2012012012012012e-05,
+      "loss": 1.6132,
       "step": 160
     },
     {
       "epoch": 1.01,
-      "learning_rate": 1.2108262108262108e-05,
-      "loss": 0.5023,
+      "learning_rate": 1.2762762762762764e-05,
+      "loss": 1.6279,
       "step": 170
     },
     {
       "epoch": 1.01,
-      "learning_rate": 1.282051282051282e-05,
-      "loss": 0.5699,
+      "learning_rate": 1.3513513513513515e-05,
+      "loss": 1.6001,
       "step": 180
     },
     {
       "epoch": 1.01,
-      "learning_rate": 1.3532763532763535e-05,
-      "loss": 0.8942,
+      "learning_rate": 1.4264264264264266e-05,
+      "loss": 1.5873,
       "step": 190
     },
     {
       "epoch": 1.01,
-      "learning_rate": 1.4245014245014246e-05,
-      "loss": 0.7764,
+      "learning_rate": 1.5015015015015016e-05,
+      "loss": 1.59,
       "step": 200
     },
     {
       "epoch": 1.01,
-      "learning_rate": 1.4957264957264958e-05,
-      "loss": 0.8046,
+      "learning_rate": 1.5765765765765765e-05,
+      "loss": 1.5691,
       "step": 210
     },
     {
-      "epoch": 1.01,
-      "learning_rate": 1.566951566951567e-05,
-      "loss": 0.5461,
+      "epoch": 1.02,
+      "learning_rate": 1.6516516516516518e-05,
+      "loss": 1.6001,
       "step": 220
     },
     {
       "epoch": 1.02,
-      "learning_rate": 1.6381766381766382e-05,
-      "loss": 0.8357,
-      "step": 230
+      "eval_accuracy": 0.30414746543778803,
+      "eval_loss": 1.5782501697540283,
+      "eval_runtime": 245.0114,
+      "eval_samples_per_second": 0.886,
+      "eval_steps_per_second": 0.224,
+      "step": 224
     },
     {
-      "epoch": 1.02,
-      "eval_accuracy": 0.7004608294930875,
-      "eval_loss": 0.911893367767334,
-      "eval_runtime": 306.0217,
-      "eval_samples_per_second": 0.709,
-      "eval_steps_per_second": 0.18,
-      "step": 236
+      "epoch": 2.0,
+      "learning_rate": 1.7267267267267267e-05,
+      "loss": 1.5771,
+      "step": 230
     },
     {
       "epoch": 2.0,
-      "learning_rate": 1.7094017094017095e-05,
-      "loss": 0.575,
+      "learning_rate": 1.801801801801802e-05,
+      "loss": 1.5651,
       "step": 240
     },
     {
       "epoch": 2.0,
-      "learning_rate": 1.7806267806267805e-05,
-      "loss": 0.7452,
+      "learning_rate": 1.8768768768768768e-05,
+      "loss": 1.4911,
       "step": 250
     },
     {
-      "epoch": 2.0,
-      "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.7453,
+      "epoch": 2.01,
+      "learning_rate": 1.951951951951952e-05,
+      "loss": 1.6414,
       "step": 260
     },
     {
-      "epoch": 2.0,
-      "learning_rate": 1.923076923076923e-05,
-      "loss": 0.5861,
+      "epoch": 2.01,
+      "learning_rate": 2.0270270270270273e-05,
+      "loss": 1.5454,
       "step": 270
     },
     {
       "epoch": 2.01,
-      "learning_rate": 1.9943019943019945e-05,
-      "loss": 0.4319,
+      "learning_rate": 2.102102102102102e-05,
+      "loss": 1.4563,
       "step": 280
     },
     {
       "epoch": 2.01,
-      "learning_rate": 2.0655270655270654e-05,
-      "loss": 0.532,
+      "learning_rate": 2.1771771771771774e-05,
+      "loss": 1.4325,
       "step": 290
     },
     {
       "epoch": 2.01,
-      "learning_rate": 2.1367521367521368e-05,
-      "loss": 0.8728,
+      "learning_rate": 2.2522522522522523e-05,
+      "loss": 1.4738,
       "step": 300
     },
     {
       "epoch": 2.01,
-      "learning_rate": 2.207977207977208e-05,
-      "loss": 1.1116,
+      "learning_rate": 2.3273273273273275e-05,
+      "loss": 1.3045,
       "step": 310
     },
     {
       "epoch": 2.01,
-      "learning_rate": 2.2792022792022794e-05,
-      "loss": 0.9775,
+      "learning_rate": 2.4024024024024024e-05,
+      "loss": 1.2273,
       "step": 320
     },
     {
-      "epoch": 2.01,
-      "learning_rate": 2.3504273504273504e-05,
-      "loss": 1.1248,
+      "epoch": 2.02,
+      "learning_rate": 2.4774774774774777e-05,
+      "loss": 1.4193,
       "step": 330
     },
     {
-      "epoch": 2.01,
-      "learning_rate": 2.4216524216524217e-05,
-      "loss": 0.7516,
-      "step": 340
+      "epoch": 2.02,
+      "eval_accuracy": 0.3824884792626728,
+      "eval_loss": 1.4873727560043335,
+      "eval_runtime": 251.8516,
+      "eval_samples_per_second": 0.862,
+      "eval_steps_per_second": 0.218,
+      "step": 336
     },
     {
-      "epoch": 2.02,
-      "learning_rate": 2.492877492877493e-05,
-      "loss": 0.474,
-      "step": 350
+      "epoch": 3.0,
+      "learning_rate": 2.552552552552553e-05,
+      "loss": 1.5274,
+      "step": 340
     },
     {
-      "epoch": 2.02,
-      "eval_accuracy": 0.6820276497695853,
-      "eval_loss": 0.9698442220687866,
-      "eval_runtime": 313.4651,
-      "eval_samples_per_second": 0.692,
-      "eval_steps_per_second": 0.175,
-      "step": 354
+      "epoch": 3.0,
+      "learning_rate": 2.6276276276276278e-05,
+      "loss": 1.3287,
+      "step": 350
     },
     {
       "epoch": 3.0,
-      "learning_rate": 2.564102564102564e-05,
-      "loss": 0.999,
+      "learning_rate": 2.702702702702703e-05,
+      "loss": 1.4428,
       "step": 360
     },
     {
-      "epoch": 3.0,
-      "learning_rate": 2.6353276353276356e-05,
-      "loss": 0.758,
+      "epoch": 3.01,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 1.2924,
       "step": 370
     },
     {
-      "epoch": 3.0,
-      "learning_rate": 2.706552706552707e-05,
-      "loss": 0.6963,
+      "epoch": 3.01,
+      "learning_rate": 2.852852852852853e-05,
+      "loss": 1.204,
       "step": 380
     },
     {
       "epoch": 3.01,
-      "learning_rate": 2.777777777777778e-05,
-      "loss": 0.5529,
+      "learning_rate": 2.927927927927928e-05,
+      "loss": 1.1626,
       "step": 390
     },
     {
       "epoch": 3.01,
-      "learning_rate": 2.8490028490028492e-05,
-      "loss": 0.7653,
+      "learning_rate": 3.0030030030030033e-05,
+      "loss": 1.5428,
       "step": 400
     },
     {
       "epoch": 3.01,
-      "learning_rate": 2.9202279202279202e-05,
-      "loss": 0.8444,
+      "learning_rate": 3.078078078078078e-05,
+      "loss": 1.1494,
       "step": 410
     },
     {
       "epoch": 3.01,
-      "learning_rate": 2.9914529914529915e-05,
-      "loss": 0.6609,
+      "learning_rate": 3.153153153153153e-05,
+      "loss": 1.2899,
       "step": 420
     },
     {
       "epoch": 3.01,
-      "learning_rate": 3.0626780626780625e-05,
-      "loss": 0.5681,
+      "learning_rate": 3.2282282282282286e-05,
+      "loss": 1.4876,
       "step": 430
     },
     {
-      "epoch": 3.01,
-      "learning_rate": 3.133903133903134e-05,
-      "loss": 0.7641,
+      "epoch": 3.02,
+      "learning_rate": 3.3033033033033035e-05,
+      "loss": 1.398,
       "step": 440
     },
     {
-      "epoch": 3.01,
-      "learning_rate": 3.205128205128206e-05,
-      "loss": 0.7485,
+      "epoch": 3.02,
+      "eval_accuracy": 0.6405529953917051,
+      "eval_loss": 1.0196824073791504,
+      "eval_runtime": 247.8505,
+      "eval_samples_per_second": 0.876,
+      "eval_steps_per_second": 0.222,
+      "step": 448
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 3.3783783783783784e-05,
+      "loss": 1.3202,
       "step": 450
     },
     {
-      "epoch": 3.02,
-      "learning_rate": 3.2763532763532764e-05,
-      "loss": 0.7458,
+      "epoch": 4.0,
+      "learning_rate": 3.453453453453453e-05,
+      "loss": 1.145,
       "step": 460
     },
     {
-      "epoch": 3.02,
-      "learning_rate": 3.347578347578348e-05,
-      "loss": 0.7899,
+      "epoch": 4.0,
+      "learning_rate": 3.528528528528528e-05,
+      "loss": 1.2636,
       "step": 470
     },
-    {
-      "epoch": 3.02,
-      "eval_accuracy": 0.6774193548387096,
-      "eval_loss": 1.135077714920044,
-      "eval_runtime": 285.9321,
-      "eval_samples_per_second": 0.759,
-      "eval_steps_per_second": 0.192,
-      "step": 472
-    },
     {
       "epoch": 4.0,
-      "learning_rate": 3.418803418803419e-05,
-      "loss": 0.4874,
+      "learning_rate": 3.603603603603604e-05,
+      "loss": 1.2266,
       "step": 480
     },
     {
-      "epoch": 4.0,
-      "learning_rate": 3.4900284900284904e-05,
-      "loss": 0.6289,
+      "epoch": 4.01,
+      "learning_rate": 3.678678678678679e-05,
+      "loss": 1.6096,
       "step": 490
     },
     {
-      "epoch": 4.0,
-      "learning_rate": 3.561253561253561e-05,
-      "loss": 0.6611,
+      "epoch": 4.01,
+      "learning_rate": 3.7537537537537536e-05,
+      "loss": 1.1474,
       "step": 500
     },
     {
       "epoch": 4.01,
-      "learning_rate": 3.6324786324786323e-05,
-      "loss": 0.7506,
+      "learning_rate": 3.8288288288288285e-05,
+      "loss": 1.1879,
       "step": 510
     },
     {
       "epoch": 4.01,
-      "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.777,
+      "learning_rate": 3.903903903903904e-05,
+      "loss": 1.0165,
       "step": 520
     },
     {
       "epoch": 4.01,
-      "learning_rate": 3.774928774928775e-05,
-      "loss": 0.9948,
+      "learning_rate": 3.9789789789789796e-05,
+      "loss": 1.2291,
       "step": 530
     },
     {
       "epoch": 4.01,
-      "learning_rate": 3.846153846153846e-05,
-      "loss": 0.704,
+      "learning_rate": 4.0540540540540545e-05,
+      "loss": 1.2309,
       "step": 540
     },
     {
-      "epoch": 4.01,
-      "learning_rate": 3.9173789173789176e-05,
-      "loss": 0.8743,
+      "epoch": 4.02,
+      "learning_rate": 4.1291291291291294e-05,
+      "loss": 1.0638,
       "step": 550
     },
     {
-      "epoch": 4.01,
-      "learning_rate": 3.988603988603989e-05,
-      "loss": 0.6572,
+      "epoch": 4.02,
+      "learning_rate": 4.204204204204204e-05,
+      "loss": 1.2217,
       "step": 560
     },
     {
-      "epoch": 4.01,
-      "learning_rate": 4.05982905982906e-05,
-      "loss": 1.0614,
-      "step": 570
+      "epoch": 4.02,
+      "eval_accuracy": 0.391705069124424,
+      "eval_loss": 1.3385610580444336,
+      "eval_runtime": 242.8726,
+      "eval_samples_per_second": 0.893,
+      "eval_steps_per_second": 0.226,
+      "step": 560
     },
     {
-      "epoch": 4.02,
-      "learning_rate": 4.131054131054131e-05,
-      "loss": 0.6374,
-      "step": 580
+      "epoch": 5.0,
+      "learning_rate": 4.27927927927928e-05,
+      "loss": 1.4069,
+      "step": 570
     },
     {
-      "epoch": 4.02,
-      "learning_rate": 4.202279202279202e-05,
-      "loss": 0.9015,
-      "step": 590
+      "epoch": 5.0,
+      "learning_rate": 4.354354354354355e-05,
+      "loss": 1.26,
+      "step": 580
     },
     {
-      "epoch": 4.02,
-      "eval_accuracy": 0.4976958525345622,
-      "eval_loss": 1.3822749853134155,
-      "eval_runtime": 285.2486,
-      "eval_samples_per_second": 0.761,
-      "eval_steps_per_second": 0.193,
+      "epoch": 5.0,
+      "learning_rate": 4.42942942942943e-05,
+      "loss": 1.2424,
       "step": 590
     },
     {
-      "epoch": 5.0,
-      "learning_rate": 4.2735042735042735e-05,
-      "loss": 1.1476,
+      "epoch": 5.01,
+      "learning_rate": 4.5045045045045046e-05,
+      "loss": 1.2978,
       "step": 600
     },
     {
-      "epoch": 5.0,
-      "learning_rate": 4.344729344729345e-05,
-      "loss": 0.7482,
+      "epoch": 5.01,
+      "learning_rate": 4.57957957957958e-05,
+      "loss": 1.373,
       "step": 610
     },
     {
-      "epoch": 5.0,
-      "learning_rate": 4.415954415954416e-05,
-      "loss": 0.835,
+      "epoch": 5.01,
+      "learning_rate": 4.654654654654655e-05,
+      "loss": 1.2995,
       "step": 620
     },
     {
       "epoch": 5.01,
-      "learning_rate": 4.4871794871794874e-05,
-      "loss": 0.5413,
+      "learning_rate": 4.72972972972973e-05,
+      "loss": 1.053,
       "step": 630
     },
     {
       "epoch": 5.01,
-      "learning_rate": 4.558404558404559e-05,
-      "loss": 0.9714,
+      "learning_rate": 4.804804804804805e-05,
+      "loss": 1.1977,
       "step": 640
     },
     {
       "epoch": 5.01,
-      "learning_rate": 4.62962962962963e-05,
-      "loss": 0.4952,
+      "learning_rate": 4.87987987987988e-05,
+      "loss": 1.2698,
       "step": 650
     },
     {
-      "epoch": 5.01,
-      "learning_rate": 4.700854700854701e-05,
-      "loss": 1.0349,
+      "epoch": 5.02,
+      "learning_rate": 4.954954954954955e-05,
+      "loss": 1.1812,
       "step": 660
     },
-    {
-      "epoch": 5.01,
-      "learning_rate": 4.772079772079772e-05,
-      "loss": 0.8882,
-      "step": 670
-    },
-    {
-      "epoch": 5.01,
-      "learning_rate": 4.8433048433048433e-05,
-      "loss": 0.9244,
-      "step": 680
-    },
-    {
-      "epoch": 5.01,
-      "learning_rate": 4.9145299145299147e-05,
-      "loss": 1.1217,
-      "step": 690
-    },
     {
       "epoch": 5.02,
-      "learning_rate": 4.985754985754986e-05,
-      "loss": 0.7402,
-      "step": 700
+      "learning_rate": 4.9966633299966636e-05,
+      "loss": 1.2577,
+      "step": 670
     },
     {
       "epoch": 5.02,
-      "eval_accuracy": 0.695852534562212,
-      "eval_loss": 0.8660895824432373,
-      "eval_runtime": 289.3403,
-      "eval_samples_per_second": 0.75,
-      "eval_steps_per_second": 0.19,
-      "step": 708
-    },
-    {
-      "epoch": 6.0,
-      "learning_rate": 4.993668882557772e-05,
-      "loss": 0.6064,
-      "step": 710
+      "eval_accuracy": 0.5391705069124424,
+      "eval_loss": 1.2195504903793335,
+      "eval_runtime": 264.2835,
+      "eval_samples_per_second": 0.821,
+      "eval_steps_per_second": 0.208,
+      "step": 672
     },
     {
       "epoch": 6.0,
-      "learning_rate": 4.985754985754986e-05,
-      "loss": 0.6741,
-      "step": 720
+      "learning_rate": 4.988321654988322e-05,
+      "loss": 1.4038,
+      "step": 680
     },
     {
       "epoch": 6.0,
-      "learning_rate": 4.9778410889522e-05,
-      "loss": 0.7149,
-      "step": 730
+      "learning_rate": 4.97997997997998e-05,
+      "loss": 1.4382,
+      "step": 690
     },
     {
       "epoch": 6.0,
-      "learning_rate": 4.9699271921494144e-05,
-      "loss": 0.9701,
-      "step": 740
+      "learning_rate": 4.9716383049716386e-05,
+      "loss": 1.3709,
+      "step": 700
     },
     {
       "epoch": 6.01,
-      "learning_rate": 4.962013295346629e-05,
-      "loss": 0.6009,
-      "step": 750
+      "learning_rate": 4.963296629963297e-05,
+      "loss": 1.0241,
+      "step": 710
     },
     {
       "epoch": 6.01,
-      "learning_rate": 4.9540993985438435e-05,
-      "loss": 1.3047,
-      "step": 760
+      "learning_rate": 4.954954954954955e-05,
+      "loss": 1.5423,
+      "step": 720
     },
     {
       "epoch": 6.01,
-      "learning_rate": 4.946185501741058e-05,
-      "loss": 0.7303,
-      "step": 770
+      "learning_rate": 4.9466132799466136e-05,
+      "loss": 1.2711,
+      "step": 730
     },
     {
       "epoch": 6.01,
       "learning_rate": 4.938271604938271e-05,
-      "loss": 0.7054,
-      "step": 780
+      "loss": 1.4259,
+      "step": 740
     },
     {
       "epoch": 6.01,
-      "learning_rate": 4.930357708135486e-05,
-      "loss": 0.884,
-      "step": 790
+      "learning_rate": 4.92992992992993e-05,
+      "loss": 1.2684,
+      "step": 750
     },
     {
       "epoch": 6.01,
-      "learning_rate": 4.9224438113327004e-05,
-      "loss": 0.7226,
-      "step": 800
+      "learning_rate": 4.921588254921589e-05,
+      "loss": 0.9547,
+      "step": 760
     },
     {
       "epoch": 6.01,
-      "learning_rate": 4.9145299145299147e-05,
-      "loss": 0.8143,
-      "step": 810
+      "learning_rate": 4.9132465799132463e-05,
+      "loss": 1.1885,
+      "step": 770
     },
     {
       "epoch": 6.02,
-      "learning_rate": 4.906616017727129e-05,
-      "loss": 0.6343,
-      "step": 820
+      "learning_rate": 4.9049049049049054e-05,
+      "loss": 1.0121,
+      "step": 780
     },
     {
       "epoch": 6.02,
-      "eval_accuracy": 0.7004608294930875,
-      "eval_loss": 0.668888509273529,
-      "eval_runtime": 284.8469,
-      "eval_samples_per_second": 0.762,
-      "eval_steps_per_second": 0.193,
-      "step": 826
-    },
-    {
-      "epoch": 7.0,
-      "learning_rate": 4.898702120924343e-05,
-      "loss": 0.8134,
-      "step": 830
+      "eval_accuracy": 0.4792626728110599,
+      "eval_loss": 1.2319154739379883,
+      "eval_runtime": 261.8591,
+      "eval_samples_per_second": 0.829,
+      "eval_steps_per_second": 0.21,
+      "step": 784
     },
     {
       "epoch": 7.0,
-      "learning_rate": 4.890788224121557e-05,
-      "loss": 0.664,
-      "step": 840
+      "learning_rate": 4.896563229896564e-05,
+      "loss": 1.355,
+      "step": 790
     },
     {
       "epoch": 7.0,
-      "learning_rate": 4.882874327318772e-05,
-      "loss": 0.7448,
-      "step": 850
+      "learning_rate": 4.8882215548882214e-05,
+      "loss": 1.4279,
+      "step": 800
     },
     {
       "epoch": 7.0,
-      "learning_rate": 4.8749604305159865e-05,
-      "loss": 0.6677,
-      "step": 860
+      "learning_rate": 4.87987987987988e-05,
+      "loss": 1.0318,
+      "step": 810
     },
     {
       "epoch": 7.01,
-      "learning_rate": 4.867046533713201e-05,
-      "loss": 1.0314,
-      "step": 870
+      "learning_rate": 4.871538204871539e-05,
+      "loss": 1.1375,
+      "step": 820
     },
     {
       "epoch": 7.01,
-      "learning_rate": 4.859132636910415e-05,
-      "loss": 0.6049,
-      "step": 880
+      "learning_rate": 4.8631965298631964e-05,
+      "loss": 1.1721,
+      "step": 830
     },
     {
       "epoch": 7.01,
-      "learning_rate": 4.851218740107629e-05,
-      "loss": 1.3879,
-      "step": 890
+      "learning_rate": 4.854854854854855e-05,
+      "loss": 1.2527,
+      "step": 840
     },
     {
       "epoch": 7.01,
-      "learning_rate": 4.8433048433048433e-05,
-      "loss": 0.7445,
-      "step": 900
+      "learning_rate": 4.846513179846514e-05,
+      "loss": 1.0756,
+      "step": 850
     },
     {
       "epoch": 7.01,
-      "learning_rate": 4.835390946502058e-05,
-      "loss": 0.7087,
-      "step": 910
+      "learning_rate": 4.8381715048381715e-05,
+      "loss": 1.1513,
+      "step": 860
     },
     {
       "epoch": 7.01,
-      "learning_rate": 4.8274770496992725e-05,
-      "loss": 0.5025,
-      "step": 920
+      "learning_rate": 4.82982982982983e-05,
+      "loss": 1.0789,
+      "step": 870
     },
     {
       "epoch": 7.01,
-      "learning_rate": 4.819563152896486e-05,
-      "loss": 0.6759,
-      "step": 930
+      "learning_rate": 4.821488154821489e-05,
+      "loss": 1.2463,
+      "step": 880
     },
     {
       "epoch": 7.02,
-      "learning_rate": 4.811649256093701e-05,
-      "loss": 0.7427,
-      "step": 940
+      "learning_rate": 4.8131464798131465e-05,
+      "loss": 1.2485,
+      "step": 890
     },
     {
       "epoch": 7.02,
-      "eval_accuracy": 0.6728110599078341,
-      "eval_loss": 0.9108946919441223,
-      "eval_runtime": 320.8735,
-      "eval_samples_per_second": 0.676,
-      "eval_steps_per_second": 0.171,
-      "step": 944
+      "eval_accuracy": 0.7511520737327189,
+      "eval_loss": 0.8229795098304749,
+      "eval_runtime": 252.4438,
+      "eval_samples_per_second": 0.86,
+      "eval_steps_per_second": 0.218,
+      "step": 896
     },
     {
       "epoch": 8.0,
-      "learning_rate": 4.803735359290915e-05,
-      "loss": 0.9541,
-      "step": 950
+      "learning_rate": 4.804804804804805e-05,
+      "loss": 0.9738,
+      "step": 900
     },
     {
       "epoch": 8.0,
-      "learning_rate": 4.7958214624881294e-05,
-      "loss": 0.9822,
-      "step": 960
+      "learning_rate": 4.796463129796463e-05,
+      "loss": 0.7946,
+      "step": 910
     },
     {
       "epoch": 8.0,
-      "learning_rate": 4.787907565685344e-05,
-      "loss": 0.9233,
-      "step": 970
+      "learning_rate": 4.7881214547881215e-05,
+      "loss": 1.2819,
+      "step": 920
     },
     {
       "epoch": 8.01,
-      "learning_rate": 4.779993668882558e-05,
-      "loss": 0.9739,
-      "step": 980
+      "learning_rate": 4.77977977977978e-05,
+      "loss": 1.7354,
+      "step": 930
     },
     {
       "epoch": 8.01,
-      "learning_rate": 4.772079772079772e-05,
-      "loss": 0.966,
-      "step": 990
+      "learning_rate": 4.771438104771438e-05,
+      "loss": 1.161,
+      "step": 940
     },
     {
       "epoch": 8.01,
-      "learning_rate": 4.764165875276987e-05,
-      "loss": 0.8516,
-      "step": 1000
+      "learning_rate": 4.7630964297630966e-05,
+      "loss": 1.3311,
+      "step": 950
     },
     {
       "epoch": 8.01,
-      "learning_rate": 4.756251978474201e-05,
-      "loss": 0.6187,
-      "step": 1010
+      "learning_rate": 4.754754754754755e-05,
+      "loss": 0.9299,
+      "step": 960
     },
     {
       "epoch": 8.01,
-      "learning_rate": 4.7483380816714154e-05,
-      "loss": 1.2194,
-      "step": 1020
+      "learning_rate": 4.746413079746413e-05,
+      "loss": 0.9622,
+      "step": 970
     },
     {
       "epoch": 8.01,
-      "learning_rate": 4.7404241848686296e-05,
-      "loss": 1.1313,
-      "step": 1030
+      "learning_rate": 4.7380714047380716e-05,
+      "loss": 0.9557,
+      "step": 980
     },
     {
       "epoch": 8.01,
-      "learning_rate": 4.732510288065844e-05,
-      "loss": 0.7851,
-      "step": 1040
-    },
-    {
-      "epoch": 8.02,
-      "learning_rate": 4.724596391263058e-05,
-      "loss": 0.8549,
-      "step": 1050
+      "learning_rate": 4.72972972972973e-05,
+      "loss": 0.9241,
+      "step": 990
     },
     {
       "epoch": 8.02,
-      "learning_rate": 4.716682494460272e-05,
-      "loss": 0.5898,
-      "step": 1060
+      "learning_rate": 4.721388054721388e-05,
+      "loss": 1.025,
+      "step": 1000
     },
     {
       "epoch": 8.02,
-      "eval_accuracy": 0.5944700460829493,
-      "eval_loss": 1.0126854181289673,
-      "eval_runtime": 324.9051,
-      "eval_samples_per_second": 0.668,
-      "eval_steps_per_second": 0.169,
-      "step": 1062
+      "eval_accuracy": 0.6866359447004609,
+      "eval_loss": 0.8022767305374146,
+      "eval_runtime": 256.9331,
+      "eval_samples_per_second": 0.845,
+      "eval_steps_per_second": 0.214,
+      "step": 1008
     },
     {
       "epoch": 9.0,
-      "learning_rate": 4.708768597657487e-05,
-      "loss": 0.4566,
-      "step": 1070
+      "learning_rate": 4.7130463797130466e-05,
+      "loss": 1.6034,
+      "step": 1010
     },
     {
       "epoch": 9.0,
-      "learning_rate": 4.700854700854701e-05,
-      "loss": 0.9513,
-      "step": 1080
+      "learning_rate": 4.704704704704705e-05,
+      "loss": 1.0115,
+      "step": 1020
     },
     {
       "epoch": 9.0,
-      "learning_rate": 4.692940804051915e-05,
-      "loss": 0.3735,
-      "step": 1090
+      "learning_rate": 4.696363029696363e-05,
+      "loss": 1.337,
+      "step": 1030
     },
     {
-      "epoch": 9.01,
-      "learning_rate": 4.68502690724913e-05,
-      "loss": 0.7034,
-      "step": 1100
+      "epoch": 9.0,
+      "learning_rate": 4.688021354688022e-05,
+      "loss": 1.351,
+      "step": 1040
     },
     {
       "epoch": 9.01,
-      "learning_rate": 4.677113010446344e-05,
-      "loss": 0.6437,
-      "step": 1110
+      "learning_rate": 4.67967967967968e-05,
+      "loss": 1.0992,
+      "step": 1050
     },
     {
       "epoch": 9.01,
-      "learning_rate": 4.669199113643558e-05,
-      "loss": 0.8746,
-      "step": 1120
+      "learning_rate": 4.6713380046713384e-05,
+      "loss": 1.0828,
+      "step": 1060
     },
     {
       "epoch": 9.01,
-      "learning_rate": 4.6612852168407725e-05,
-      "loss": 0.6746,
-      "step": 1130
+      "learning_rate": 4.662996329662997e-05,
+      "loss": 1.0086,
+      "step": 1070
     },
     {
       "epoch": 9.01,
-      "learning_rate": 4.653371320037987e-05,
-      "loss": 0.9281,
-      "step": 1140
+      "learning_rate": 4.654654654654655e-05,
+      "loss": 0.8658,
+      "step": 1080
     },
     {
       "epoch": 9.01,
-      "learning_rate": 4.645457423235201e-05,
-      "loss": 0.9406,
-      "step": 1150
+      "learning_rate": 4.6463129796463134e-05,
+      "loss": 0.7355,
+      "step": 1090
     },
     {
       "epoch": 9.01,
-      "learning_rate": 4.637543526432416e-05,
-      "loss": 1.0328,
-      "step": 1160
+      "learning_rate": 4.637971304637971e-05,
+      "loss": 1.0191,
+      "step": 1100
     },
     {
       "epoch": 9.02,
       "learning_rate": 4.62962962962963e-05,
-      "loss": 0.8338,
-      "step": 1170
+      "loss": 1.0171,
+      "step": 1110
     },
     {
       "epoch": 9.02,
-      "learning_rate": 4.621715732826844e-05,
-      "loss": 0.6258,
-      "step": 1180
+      "learning_rate": 4.6212879546212884e-05,
+      "loss": 1.2952,
+      "step": 1120
     },
     {
       "epoch": 9.02,
-      "eval_accuracy": 0.7235023041474654,
-      "eval_loss": 0.7130948305130005,
-      "eval_runtime": 315.7938,
-      "eval_samples_per_second": 0.687,
-      "eval_steps_per_second": 0.174,
-      "step": 1180
+      "eval_accuracy": 0.6036866359447005,
+      "eval_loss": 0.9130178093910217,
+      "eval_runtime": 256.4943,
+      "eval_samples_per_second": 0.846,
+      "eval_steps_per_second": 0.214,
+      "step": 1120
     },
     {
       "epoch": 10.0,
-      "learning_rate": 4.6138018360240585e-05,
-      "loss": 0.635,
-      "step": 1190
+      "learning_rate": 4.612946279612946e-05,
+      "loss": 0.8975,
+      "step": 1130
     },
     {
       "epoch": 10.0,
-      "learning_rate": 4.605887939221273e-05,
-      "loss": 0.8311,
-      "step": 1200
+      "learning_rate": 4.604604604604605e-05,
+      "loss": 1.0215,
+      "step": 1140
     },
     {
       "epoch": 10.0,
-      "learning_rate": 4.597974042418487e-05,
-      "loss": 1.2343,
-      "step": 1210
+      "learning_rate": 4.5962629295962635e-05,
+      "loss": 0.9562,
+      "step": 1150
     },
     {
       "epoch": 10.01,
-      "learning_rate": 4.590060145615702e-05,
-      "loss": 0.8252,
-      "step": 1220
+      "learning_rate": 4.587921254587921e-05,
+      "loss": 1.2006,
+      "step": 1160
     },
     {
       "epoch": 10.01,
-      "learning_rate": 4.582146248812916e-05,
-      "loss": 0.7481,
-      "step": 1230
+      "learning_rate": 4.57957957957958e-05,
+      "loss": 1.2811,
+      "step": 1170
     },
     {
       "epoch": 10.01,
-      "learning_rate": 4.5742323520101296e-05,
-      "loss": 0.4203,
-      "step": 1240
+      "learning_rate": 4.5712379045712385e-05,
+      "loss": 1.0057,
+      "step": 1180
     },
     {
       "epoch": 10.01,
-      "learning_rate": 4.5663184552073445e-05,
-      "loss": 0.5636,
-      "step": 1250
+      "learning_rate": 4.562896229562896e-05,
+      "loss": 1.0403,
+      "step": 1190
     },
     {
       "epoch": 10.01,
-      "learning_rate": 4.558404558404559e-05,
-      "loss": 0.5062,
-      "step": 1260
+      "learning_rate": 4.5545545545545545e-05,
+      "loss": 0.8273,
+      "step": 1200
     },
     {
       "epoch": 10.01,
-      "learning_rate": 4.550490661601773e-05,
-      "loss": 0.5214,
-      "step": 1270
+      "learning_rate": 4.5462128795462135e-05,
+      "loss": 0.7807,
+      "step": 1210
     },
     {
-      "epoch": 10.01,
-      "learning_rate": 4.542576764798987e-05,
-      "loss": 0.6217,
-      "step": 1280
+      "epoch": 10.02,
+      "learning_rate": 4.537871204537871e-05,
+      "loss": 0.9059,
+      "step": 1220
     },
     {
       "epoch": 10.02,
-      "learning_rate": 4.5346628679962014e-05,
-      "loss": 0.9957,
-      "step": 1290
+      "learning_rate": 4.5295295295295295e-05,
+      "loss": 0.9499,
+      "step": 1230
     },
     {
       "epoch": 10.02,
-      "eval_accuracy": 0.6728110599078341,
-      "eval_loss": 0.9507045745849609,
-      "eval_runtime": 315.2197,
-      "eval_samples_per_second": 0.688,
-      "eval_steps_per_second": 0.174,
-      "step": 1298
+      "eval_accuracy": 0.6036866359447005,
+      "eval_loss": 1.0620718002319336,
+      "eval_runtime": 255.1418,
+      "eval_samples_per_second": 0.851,
+      "eval_steps_per_second": 0.216,
+      "step": 1232
     },
     {
       "epoch": 11.0,
-      "learning_rate": 4.5267489711934157e-05,
-      "loss": 0.5816,
-      "step": 1300
-    },
-    {
-      "epoch": 11.0,
-      "learning_rate": 4.51883507439063e-05,
-      "loss": 0.8868,
-      "step": 1310
+      "learning_rate": 4.5211878545211886e-05,
+      "loss": 1.0751,
+      "step": 1240
     },
     {
       "epoch": 11.0,
-      "learning_rate": 4.510921177587845e-05,
-      "loss": 0.7216,
-      "step": 1320
+      "learning_rate": 4.512846179512846e-05,
+      "loss": 1.034,
+      "step": 1250
     },
     {
       "epoch": 11.0,
-      "learning_rate": 4.503007280785059e-05,
-      "loss": 0.5962,
-      "step": 1330
+      "learning_rate": 4.5045045045045046e-05,
+      "loss": 0.6693,
+      "step": 1260
     },
     {
       "epoch": 11.01,
-      "learning_rate": 4.4950933839822725e-05,
-      "loss": 1.2384,
-      "step": 1340
+      "learning_rate": 4.4961628294961636e-05,
+      "loss": 1.0748,
+      "step": 1270
     },
     {
       "epoch": 11.01,
-      "learning_rate": 4.4871794871794874e-05,
-      "loss": 0.8794,
-      "step": 1350
+      "learning_rate": 4.487821154487821e-05,
+      "loss": 0.9789,
+      "step": 1280
     },
     {
       "epoch": 11.01,
-      "learning_rate": 4.479265590376702e-05,
-      "loss": 0.8364,
-      "step": 1360
+      "learning_rate": 4.4794794794794796e-05,
+      "loss": 0.8083,
+      "step": 1290
     },
     {
       "epoch": 11.01,
-      "learning_rate": 4.471351693573916e-05,
-      "loss": 0.8106,
-      "step": 1370
+      "learning_rate": 4.471137804471138e-05,
+      "loss": 0.9976,
+      "step": 1300
     },
     {
       "epoch": 11.01,
-      "learning_rate": 4.463437796771131e-05,
-      "loss": 0.7839,
-      "step": 1380
+      "learning_rate": 4.462796129462796e-05,
+      "loss": 0.7674,
+      "step": 1310
     },
     {
       "epoch": 11.01,
-      "learning_rate": 4.455523899968344e-05,
-      "loss": 0.7032,
-      "step": 1390
+      "learning_rate": 4.4544544544544546e-05,
+      "loss": 0.8705,
+      "step": 1320
     },
     {
       "epoch": 11.01,
-      "learning_rate": 4.4476100031655586e-05,
-      "loss": 0.9424,
-      "step": 1400
+      "learning_rate": 4.446112779446113e-05,
+      "loss": 1.195,
+      "step": 1330
     },
     {
       "epoch": 11.02,
-      "learning_rate": 4.4396961063627735e-05,
-      "loss": 0.401,
-      "step": 1410
+      "learning_rate": 4.437771104437771e-05,
+      "loss": 0.8805,
+      "step": 1340
     },
     {
       "epoch": 11.02,
-      "eval_accuracy": 0.7188940092165899,
-      "eval_loss": 0.6258705258369446,
-      "eval_runtime": 308.0396,
-      "eval_samples_per_second": 0.704,
-      "eval_steps_per_second": 0.179,
-      "step": 1416
-    },
-    {
-      "epoch": 12.0,
-      "learning_rate": 4.431782209559988e-05,
-      "loss": 0.4381,
-      "step": 1420
+      "eval_accuracy": 0.7050691244239631,
+      "eval_loss": 0.871336042881012,
+      "eval_runtime": 256.7263,
+      "eval_samples_per_second": 0.845,
+      "eval_steps_per_second": 0.214,
+      "step": 1344
     },
     {
       "epoch": 12.0,
-      "learning_rate": 4.423868312757202e-05,
-      "loss": 0.5723,
-      "step": 1430
+      "learning_rate": 4.42942942942943e-05,
+      "loss": 0.8859,
+      "step": 1350
     },
     {
       "epoch": 12.0,
-      "learning_rate": 4.415954415954416e-05,
-      "loss": 0.5368,
-      "step": 1440
+      "learning_rate": 4.421087754421088e-05,
+      "loss": 1.2024,
+      "step": 1360
     },
     {
       "epoch": 12.0,
-      "learning_rate": 4.4080405191516304e-05,
-      "loss": 1.065,
-      "step": 1450
+      "learning_rate": 4.4127460794127464e-05,
+      "loss": 0.9896,
+      "step": 1370
     },
     {
       "epoch": 12.01,
-      "learning_rate": 4.4001266223488446e-05,
-      "loss": 0.5411,
-      "step": 1460
+      "learning_rate": 4.404404404404405e-05,
+      "loss": 0.9208,
+      "step": 1380
     },
     {
       "epoch": 12.01,
-      "learning_rate": 4.3922127255460595e-05,
-      "loss": 1.1746,
-      "step": 1470
+      "learning_rate": 4.3960627293960624e-05,
+      "loss": 1.2375,
+      "step": 1390
     },
     {
       "epoch": 12.01,
-      "learning_rate": 4.384298828743274e-05,
-      "loss": 0.8506,
-      "step": 1480
+      "learning_rate": 4.3877210543877214e-05,
+      "loss": 0.9493,
+      "step": 1400
     },
     {
       "epoch": 12.01,
-      "learning_rate": 4.376384931940488e-05,
-      "loss": 0.9014,
-      "step": 1490
+      "learning_rate": 4.37937937937938e-05,
+      "loss": 0.9476,
+      "step": 1410
     },
     {
       "epoch": 12.01,
-      "learning_rate": 4.368471035137702e-05,
-      "loss": 0.7609,
-      "step": 1500
+      "learning_rate": 4.3710377043710374e-05,
+      "loss": 0.8137,
+      "step": 1420
     },
     {
       "epoch": 12.01,
-      "learning_rate": 4.3605571383349164e-05,
-      "loss": 0.6263,
-      "step": 1510
+      "learning_rate": 4.3626960293626964e-05,
+      "loss": 0.854,
+      "step": 1430
     },
     {
       "epoch": 12.01,
-      "learning_rate": 4.3526432415321306e-05,
-      "loss": 1.1741,
-      "step": 1520
+      "learning_rate": 4.354354354354355e-05,
+      "loss": 1.0682,
+      "step": 1440
     },
     {
       "epoch": 12.02,
-      "learning_rate": 4.344729344729345e-05,
-      "loss": 0.5422,
-      "step": 1530
+      "learning_rate": 4.3460126793460125e-05,
+      "loss": 1.2066,
+      "step": 1450
     },
     {
       "epoch": 12.02,
-      "eval_accuracy": 0.6774193548387096,
-      "eval_loss": 0.9453245997428894,
-      "eval_runtime": 323.3925,
-      "eval_samples_per_second": 0.671,
-      "eval_steps_per_second": 0.17,
-      "step": 1534
+      "eval_accuracy": 0.5852534562211982,
+      "eval_loss": 0.9363781809806824,
+      "eval_runtime": 263.3253,
+      "eval_samples_per_second": 0.824,
+      "eval_steps_per_second": 0.209,
+      "step": 1456
     },
     {
       "epoch": 13.0,
-      "learning_rate": 4.336815447926559e-05,
-      "loss": 0.607,
-      "step": 1540
+      "learning_rate": 4.3376710043376715e-05,
+      "loss": 0.7223,
+      "step": 1460
     },
     {
       "epoch": 13.0,
-      "learning_rate": 4.328901551123773e-05,
-      "loss": 0.7626,
-      "step": 1550
+      "learning_rate": 4.32932932932933e-05,
+      "loss": 0.7549,
+      "step": 1470
     },
     {
       "epoch": 13.0,
       "learning_rate": 4.3209876543209875e-05,
-      "loss": 0.4373,
-      "step": 1560
+      "loss": 0.9539,
+      "step": 1480
     },
     {
       "epoch": 13.01,
-      "learning_rate": 4.3130737575182024e-05,
-      "loss": 0.5643,
-      "step": 1570
+      "learning_rate": 4.312645979312646e-05,
+      "loss": 0.8501,
+      "step": 1490
     },
     {
       "epoch": 13.01,
-      "learning_rate": 4.3051598607154166e-05,
-      "loss": 0.7982,
-      "step": 1580
+      "learning_rate": 4.304304304304305e-05,
+      "loss": 0.9684,
+      "step": 1500
     },
     {
       "epoch": 13.01,
-      "learning_rate": 4.297245963912631e-05,
-      "loss": 0.971,
-      "step": 1590
+      "learning_rate": 4.2959626292959625e-05,
+      "loss": 1.3551,
+      "step": 1510
     },
     {
       "epoch": 13.01,
-      "learning_rate": 4.289332067109845e-05,
-      "loss": 0.9764,
-      "step": 1600
+      "learning_rate": 4.287620954287621e-05,
+      "loss": 0.9607,
+      "step": 1520
     },
     {
       "epoch": 13.01,
-      "learning_rate": 4.281418170307059e-05,
-      "loss": 0.8366,
-      "step": 1610
+      "learning_rate": 4.27927927927928e-05,
+      "loss": 0.8907,
+      "step": 1530
     },
     {
       "epoch": 13.01,
-      "learning_rate": 4.2735042735042735e-05,
-      "loss": 1.0652,
-      "step": 1620
+      "learning_rate": 4.2709376042709376e-05,
+      "loss": 1.2041,
+      "step": 1540
     },
     {
       "epoch": 13.01,
-      "learning_rate": 4.2655903767014884e-05,
-      "loss": 0.7619,
-      "step": 1630
-    },
-    {
-      "epoch": 13.02,
-      "learning_rate": 4.2576764798987026e-05,
-      "loss": 0.639,
-      "step": 1640
+      "learning_rate": 4.262595929262596e-05,
+      "loss": 0.7261,
+      "step": 1550
     },
     {
       "epoch": 13.02,
-      "learning_rate": 4.249762583095916e-05,
-      "loss": 0.6852,
-      "step": 1650
+      "learning_rate": 4.254254254254255e-05,
+      "loss": 0.9358,
+      "step": 1560
     },
     {
       "epoch": 13.02,
-      "eval_accuracy": 0.7004608294930875,
-      "eval_loss": 0.8649422526359558,
-      "eval_runtime": 301.7734,
-      "eval_samples_per_second": 0.719,
-      "eval_steps_per_second": 0.182,
-      "step": 1652
+      "eval_accuracy": 0.5852534562211982,
+      "eval_loss": 0.910686731338501,
+      "eval_runtime": 265.3055,
+      "eval_samples_per_second": 0.818,
+      "eval_steps_per_second": 0.207,
+      "step": 1568
     },
     {
       "epoch": 14.0,
-      "learning_rate": 4.241848686293131e-05,
-      "loss": 0.9047,
-      "step": 1660
+      "learning_rate": 4.2459125792459126e-05,
+      "loss": 0.9092,
+      "step": 1570
     },
     {
       "epoch": 14.0,
-      "learning_rate": 4.233934789490345e-05,
-      "loss": 1.1309,
-      "step": 1670
+      "learning_rate": 4.237570904237571e-05,
+      "loss": 0.8376,
+      "step": 1580
     },
     {
       "epoch": 14.0,
-      "learning_rate": 4.2260208926875595e-05,
-      "loss": 0.7564,
-      "step": 1680
+      "learning_rate": 4.229229229229229e-05,
+      "loss": 1.0121,
+      "step": 1590
     },
     {
-      "epoch": 14.01,
-      "learning_rate": 4.2181069958847744e-05,
-      "loss": 0.6608,
-      "step": 1690
+      "epoch": 14.0,
+      "learning_rate": 4.2208875542208876e-05,
+      "loss": 0.9361,
+      "step": 1600
     },
     {
       "epoch": 14.01,
-      "learning_rate": 4.210193099081988e-05,
-      "loss": 0.5688,
-      "step": 1700
+      "learning_rate": 4.212545879212546e-05,
+      "loss": 1.053,
+      "step": 1610
     },
     {
       "epoch": 14.01,
-      "learning_rate": 4.202279202279202e-05,
-      "loss": 0.6834,
-      "step": 1710
+      "learning_rate": 4.204204204204204e-05,
+      "loss": 0.9652,
+      "step": 1620
     },
     {
       "epoch": 14.01,
-      "learning_rate": 4.194365305476417e-05,
-      "loss": 0.7658,
-      "step": 1720
+      "learning_rate": 4.1958625291958627e-05,
+      "loss": 0.7609,
+      "step": 1630
     },
     {
       "epoch": 14.01,
-      "learning_rate": 4.186451408673631e-05,
-      "loss": 0.4818,
-      "step": 1730
+      "learning_rate": 4.187520854187521e-05,
+      "loss": 1.0871,
+      "step": 1640
     },
     {
       "epoch": 14.01,
-      "learning_rate": 4.1785375118708455e-05,
-      "loss": 0.6077,
-      "step": 1740
+      "learning_rate": 4.1791791791791793e-05,
+      "loss": 0.9353,
+      "step": 1650
     },
     {
       "epoch": 14.01,
-      "learning_rate": 4.17062361506806e-05,
-      "loss": 0.6641,
-      "step": 1750
+      "learning_rate": 4.170837504170838e-05,
+      "loss": 0.7793,
+      "step": 1660
     },
     {
       "epoch": 14.02,
-      "learning_rate": 4.162709718265274e-05,
-      "loss": 0.4068,
-      "step": 1760
+      "learning_rate": 4.162495829162496e-05,
+      "loss": 1.0073,
+      "step": 1670
     },
     {
       "epoch": 14.02,
-      "learning_rate": 4.154795821462488e-05,
-      "loss": 0.8469,
-      "step": 1770
+      "learning_rate": 4.1541541541541544e-05,
+      "loss": 0.9043,
+      "step": 1680
     },
     {
       "epoch": 14.02,
-      "eval_accuracy": 0.6912442396313364,
-      "eval_loss": 0.9379397630691528,
-      "eval_runtime": 304.2958,
-      "eval_samples_per_second": 0.713,
-      "eval_steps_per_second": 0.181,
-      "step": 1770
+      "eval_accuracy": 0.6359447004608295,
+      "eval_loss": 0.9146963953971863,
+      "eval_runtime": 259.9831,
+      "eval_samples_per_second": 0.835,
+      "eval_steps_per_second": 0.212,
+      "step": 1680
     },
     {
       "epoch": 15.0,
-      "learning_rate": 4.1468819246597024e-05,
-      "loss": 0.4658,
-      "step": 1780
+      "learning_rate": 4.145812479145813e-05,
+      "loss": 1.0818,
+      "step": 1690
     },
     {
       "epoch": 15.0,
-      "learning_rate": 4.138968027856917e-05,
-      "loss": 0.4626,
-      "step": 1790
+      "learning_rate": 4.137470804137471e-05,
+      "loss": 0.9315,
+      "step": 1700
     },
     {
       "epoch": 15.0,
-      "learning_rate": 4.131054131054131e-05,
-      "loss": 0.8248,
-      "step": 1800
-    },
-    {
-      "epoch": 15.01,
-      "learning_rate": 4.123140234251345e-05,
-      "loss": 0.883,
-      "step": 1810
+      "learning_rate": 4.1291291291291294e-05,
+      "loss": 0.8777,
+      "step": 1710
     },
     {
       "epoch": 15.01,
-      "learning_rate": 4.11522633744856e-05,
-      "loss": 0.4465,
-      "step": 1820
+      "learning_rate": 4.120787454120788e-05,
+      "loss": 1.0695,
+      "step": 1720
     },
     {
       "epoch": 15.01,
-      "learning_rate": 4.107312440645774e-05,
-      "loss": 0.696,
-      "step": 1830
+      "learning_rate": 4.112445779112446e-05,
+      "loss": 0.6343,
+      "step": 1730
     },
     {
       "epoch": 15.01,
-      "learning_rate": 4.0993985438429884e-05,
-      "loss": 0.852,
-      "step": 1840
+      "learning_rate": 4.1041041041041045e-05,
+      "loss": 1.0106,
+      "step": 1740
     },
     {
       "epoch": 15.01,
-      "learning_rate": 4.091484647040203e-05,
-      "loss": 0.7966,
-      "step": 1850
+      "learning_rate": 4.095762429095763e-05,
+      "loss": 0.9499,
+      "step": 1750
     },
     {
       "epoch": 15.01,
-      "learning_rate": 4.083570750237417e-05,
-      "loss": 0.6836,
-      "step": 1860
+      "learning_rate": 4.087420754087421e-05,
+      "loss": 0.9108,
+      "step": 1760
     },
     {
       "epoch": 15.01,
-      "learning_rate": 4.075656853434631e-05,
-      "loss": 0.8427,
-      "step": 1870
+      "learning_rate": 4.0790790790790795e-05,
+      "loss": 0.8515,
+      "step": 1770
     },
     {
       "epoch": 15.02,
-      "learning_rate": 4.067742956631846e-05,
-      "loss": 0.8492,
-      "step": 1880
+      "learning_rate": 4.070737404070737e-05,
+      "loss": 0.749,
+      "step": 1780
     },
     {
       "epoch": 15.02,
-      "eval_accuracy": 0.6451612903225806,
-      "eval_loss": 0.900291919708252,
-      "eval_runtime": 312.5208,
-      "eval_samples_per_second": 0.694,
-      "eval_steps_per_second": 0.176,
-      "step": 1888
+      "learning_rate": 4.062395729062396e-05,
+      "loss": 0.8383,
+      "step": 1790
     },
     {
-      "epoch": 16.0,
-      "learning_rate": 4.05982905982906e-05,
-      "loss": 0.9333,
-      "step": 1890
+      "epoch": 15.02,
+      "eval_accuracy": 0.6359447004608295,
+      "eval_loss": 0.945084810256958,
+      "eval_runtime": 259.6462,
+      "eval_samples_per_second": 0.836,
+      "eval_steps_per_second": 0.212,
+      "step": 1792
     },
     {
       "epoch": 16.0,
-      "learning_rate": 4.0519151630262745e-05,
-      "loss": 0.788,
-      "step": 1900
+      "learning_rate": 4.0540540540540545e-05,
+      "loss": 0.8241,
+      "step": 1800
     },
     {
       "epoch": 16.0,
-      "learning_rate": 4.044001266223489e-05,
-      "loss": 0.6258,
-      "step": 1910
+      "learning_rate": 4.045712379045712e-05,
+      "loss": 0.7361,
+      "step": 1810
     },
     {
       "epoch": 16.0,
-      "learning_rate": 4.036087369420703e-05,
-      "loss": 0.4548,
-      "step": 1920
+      "learning_rate": 4.037370704037371e-05,
+      "loss": 0.978,
+      "step": 1820
     },
     {
       "epoch": 16.01,
-      "learning_rate": 4.028173472617917e-05,
-      "loss": 0.7268,
-      "step": 1930
+      "learning_rate": 4.0290290290290296e-05,
+      "loss": 0.7197,
+      "step": 1830
     },
     {
       "epoch": 16.01,
-      "learning_rate": 4.020259575815132e-05,
-      "loss": 0.886,
-      "step": 1940
+      "learning_rate": 4.020687354020687e-05,
+      "loss": 1.0322,
+      "step": 1840
     },
     {
       "epoch": 16.01,
       "learning_rate": 4.012345679012346e-05,
-      "loss": 0.6221,
-      "step": 1950
+      "loss": 0.8089,
+      "step": 1850
     },
     {
       "epoch": 16.01,
-      "learning_rate": 4.00443178220956e-05,
-      "loss": 0.5802,
-      "step": 1960
+      "learning_rate": 4.0040040040040046e-05,
+      "loss": 0.8855,
+      "step": 1860
     },
     {
       "epoch": 16.01,
-      "learning_rate": 3.996517885406775e-05,
-      "loss": 0.6743,
-      "step": 1970
+      "learning_rate": 3.995662328995662e-05,
+      "loss": 0.9764,
+      "step": 1870
     },
     {
       "epoch": 16.01,
-      "learning_rate": 3.988603988603989e-05,
-      "loss": 0.6022,
-      "step": 1980
+      "learning_rate": 3.9873206539873206e-05,
+      "loss": 0.7759,
+      "step": 1880
     },
     {
       "epoch": 16.01,
-      "learning_rate": 3.980690091801203e-05,
-      "loss": 0.6691,
-      "step": 1990
+      "learning_rate": 3.9789789789789796e-05,
+      "loss": 0.5536,
+      "step": 1890
     },
     {
       "epoch": 16.02,
-      "learning_rate": 3.9727761949984174e-05,
-      "loss": 0.7633,
-      "step": 2000
+      "learning_rate": 3.970637303970637e-05,
+      "loss": 0.7482,
+      "step": 1900
     },
     {
       "epoch": 16.02,
-      "eval_accuracy": 0.7235023041474654,
-      "eval_loss": 0.7601491212844849,
-      "eval_runtime": 299.492,
-      "eval_samples_per_second": 0.725,
-      "eval_steps_per_second": 0.184,
-      "step": 2006
-    },
-    {
-      "epoch": 17.0,
-      "learning_rate": 3.9648622981956316e-05,
-      "loss": 0.7329,
-      "step": 2010
+      "eval_accuracy": 0.6221198156682027,
+      "eval_loss": 0.876456618309021,
+      "eval_runtime": 273.3147,
+      "eval_samples_per_second": 0.794,
+      "eval_steps_per_second": 0.201,
+      "step": 1904
     },
     {
       "epoch": 17.0,
-      "learning_rate": 3.956948401392846e-05,
-      "loss": 0.4407,
-      "step": 2020
+      "learning_rate": 3.9622956289622956e-05,
+      "loss": 0.6662,
+      "step": 1910
     },
     {
       "epoch": 17.0,
-      "learning_rate": 3.94903450459006e-05,
-      "loss": 0.3618,
-      "step": 2030
+      "learning_rate": 3.953953953953955e-05,
+      "loss": 0.857,
+      "step": 1920
     },
     {
       "epoch": 17.0,
-      "learning_rate": 3.941120607787275e-05,
-      "loss": 0.545,
-      "step": 2040
+      "learning_rate": 3.945612278945612e-05,
+      "loss": 0.8507,
+      "step": 1930
     },
     {
       "epoch": 17.01,
-      "learning_rate": 3.933206710984489e-05,
-      "loss": 0.6322,
-      "step": 2050
+      "learning_rate": 3.937270603937271e-05,
+      "loss": 1.1261,
+      "step": 1940
     },
     {
       "epoch": 17.01,
-      "learning_rate": 3.925292814181703e-05,
-      "loss": 0.7812,
-      "step": 2060
+      "learning_rate": 3.92892892892893e-05,
+      "loss": 0.9851,
+      "step": 1950
     },
     {
       "epoch": 17.01,
-      "learning_rate": 3.9173789173789176e-05,
-      "loss": 0.5198,
-      "step": 2070
+      "learning_rate": 3.9205872539205874e-05,
+      "loss": 0.5859,
+      "step": 1960
     },
     {
       "epoch": 17.01,
-      "learning_rate": 3.909465020576132e-05,
-      "loss": 0.563,
-      "step": 2080
+      "learning_rate": 3.912245578912246e-05,
+      "loss": 0.9539,
+      "step": 1970
     },
     {
       "epoch": 17.01,
-      "learning_rate": 3.901551123773346e-05,
-      "loss": 0.7545,
-      "step": 2090
+      "learning_rate": 3.903903903903904e-05,
+      "loss": 0.8663,
+      "step": 1980
     },
     {
       "epoch": 17.01,
-      "learning_rate": 3.893637226970561e-05,
-      "loss": 0.9906,
-      "step": 2100
+      "learning_rate": 3.8955622288955624e-05,
+      "loss": 0.9946,
+      "step": 1990
     },
     {
       "epoch": 17.01,
-      "learning_rate": 3.8857233301677745e-05,
-      "loss": 0.807,
-      "step": 2110
+      "learning_rate": 3.887220553887221e-05,
+      "loss": 0.835,
+      "step": 2000
     },
     {
       "epoch": 17.02,
-      "learning_rate": 3.877809433364989e-05,
-      "loss": 0.6063,
-      "step": 2120
+      "learning_rate": 3.878878878878879e-05,
+      "loss": 0.9547,
+      "step": 2010
     },
     {
       "epoch": 17.02,
-      "eval_accuracy": 0.7788018433179723,
-      "eval_loss": 0.6181166768074036,
-      "eval_runtime": 296.2425,
-      "eval_samples_per_second": 0.733,
-      "eval_steps_per_second": 0.186,
-      "step": 2124
+      "eval_accuracy": 0.7649769585253456,
+      "eval_loss": 0.7997832894325256,
+      "eval_runtime": 275.0933,
+      "eval_samples_per_second": 0.789,
+      "eval_steps_per_second": 0.2,
+      "step": 2016
     },
     {
       "epoch": 18.0,
-      "learning_rate": 3.8698955365622036e-05,
-      "loss": 0.8029,
-      "step": 2130
+      "learning_rate": 3.8705372038705374e-05,
+      "loss": 0.8025,
+      "step": 2020
     },
     {
       "epoch": 18.0,
-      "learning_rate": 3.861981639759418e-05,
-      "loss": 0.7383,
-      "step": 2140
+      "learning_rate": 3.862195528862196e-05,
+      "loss": 0.9637,
+      "step": 2030
     },
     {
       "epoch": 18.0,
-      "learning_rate": 3.854067742956632e-05,
-      "loss": 0.9192,
-      "step": 2150
+      "learning_rate": 3.8538538538538534e-05,
+      "loss": 1.1231,
+      "step": 2040
     },
     {
       "epoch": 18.01,
-      "learning_rate": 3.846153846153846e-05,
-      "loss": 0.8748,
-      "step": 2160
+      "learning_rate": 3.8455121788455125e-05,
+      "loss": 0.6693,
+      "step": 2050
     },
     {
       "epoch": 18.01,
-      "learning_rate": 3.8382399493510605e-05,
-      "loss": 0.5987,
-      "step": 2170
+      "learning_rate": 3.837170503837171e-05,
+      "loss": 1.0535,
+      "step": 2060
     },
     {
       "epoch": 18.01,
-      "learning_rate": 3.830326052548275e-05,
-      "loss": 0.7523,
-      "step": 2180
+      "learning_rate": 3.8288288288288285e-05,
+      "loss": 0.9,
+      "step": 2070
     },
     {
       "epoch": 18.01,
-      "learning_rate": 3.8224121557454896e-05,
-      "loss": 0.853,
-      "step": 2190
+      "learning_rate": 3.8204871538204875e-05,
+      "loss": 0.7855,
+      "step": 2080
     },
     {
       "epoch": 18.01,
-      "learning_rate": 3.814498258942704e-05,
-      "loss": 0.7287,
-      "step": 2200
+      "learning_rate": 3.812145478812146e-05,
+      "loss": 0.7349,
+      "step": 2090
     },
     {
       "epoch": 18.01,
-      "learning_rate": 3.806584362139918e-05,
-      "loss": 0.8661,
-      "step": 2210
+      "learning_rate": 3.8038038038038035e-05,
+      "loss": 0.9599,
+      "step": 2100
     },
     {
       "epoch": 18.01,
-      "learning_rate": 3.798670465337132e-05,
-      "loss": 0.6155,
-      "step": 2220
+      "learning_rate": 3.7954621287954625e-05,
+      "loss": 0.6589,
+      "step": 2110
     },
     {
       "epoch": 18.02,
-      "learning_rate": 3.7907565685343465e-05,
-      "loss": 0.792,
-      "step": 2230
+      "learning_rate": 3.787120453787121e-05,
+      "loss": 0.7028,
+      "step": 2120
     },
     {
       "epoch": 18.02,
-      "learning_rate": 3.782842671731561e-05,
-      "loss": 0.6436,
-      "step": 2240
+      "eval_accuracy": 0.6405529953917051,
+      "eval_loss": 0.9256826639175415,
+      "eval_runtime": 263.4393,
+      "eval_samples_per_second": 0.824,
+      "eval_steps_per_second": 0.209,
+      "step": 2128
     },
     {
-      "epoch": 18.02,
-      "eval_accuracy": 0.631336405529954,
-      "eval_loss": 0.9444882273674011,
-      "eval_runtime": 299.4212,
-      "eval_samples_per_second": 0.725,
-      "eval_steps_per_second": 0.184,
-      "step": 2242
+      "epoch": 19.0,
+      "learning_rate": 3.7787787787787786e-05,
+      "loss": 0.5654,
+      "step": 2130
     },
     {
       "epoch": 19.0,
-      "learning_rate": 3.774928774928775e-05,
-      "loss": 0.6049,
-      "step": 2250
+      "learning_rate": 3.7704371037704376e-05,
+      "loss": 0.8278,
+      "step": 2140
     },
     {
       "epoch": 19.0,
-      "learning_rate": 3.767014878125989e-05,
-      "loss": 0.6267,
-      "step": 2260
+      "learning_rate": 3.762095428762096e-05,
+      "loss": 1.0629,
+      "step": 2150
     },
     {
       "epoch": 19.0,
-      "learning_rate": 3.7591009813232034e-05,
-      "loss": 0.4215,
-      "step": 2270
+      "learning_rate": 3.7537537537537536e-05,
+      "loss": 1.026,
+      "step": 2160
     },
     {
       "epoch": 19.01,
-      "learning_rate": 3.7511870845204176e-05,
-      "loss": 0.9263,
-      "step": 2280
+      "learning_rate": 3.745412078745412e-05,
+      "loss": 0.8544,
+      "step": 2170
     },
     {
       "epoch": 19.01,
-      "learning_rate": 3.7432731877176325e-05,
-      "loss": 0.4688,
-      "step": 2290
+      "learning_rate": 3.737070403737071e-05,
+      "loss": 0.8669,
+      "step": 2180
     },
     {
       "epoch": 19.01,
-      "learning_rate": 3.735359290914847e-05,
-      "loss": 0.668,
-      "step": 2300
+      "learning_rate": 3.7287287287287286e-05,
+      "loss": 0.6044,
+      "step": 2190
     },
     {
       "epoch": 19.01,
-      "learning_rate": 3.727445394112061e-05,
-      "loss": 0.5778,
-      "step": 2310
-    },
-    {
-      "epoch": 19.01,
-      "learning_rate": 3.719531497309275e-05,
-      "loss": 0.4659,
-      "step": 2320
+      "learning_rate": 3.720387053720387e-05,
+      "loss": 0.8129,
+      "step": 2200
     },
     {
       "epoch": 19.01,
-      "learning_rate": 3.7116176005064894e-05,
-      "loss": 0.5316,
-      "step": 2330
+      "learning_rate": 3.712045378712046e-05,
+      "loss": 0.5283,
+      "step": 2210
     },
     {
       "epoch": 19.01,
       "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.6448,
-      "step": 2340
+      "loss": 1.0115,
+      "step": 2220
     },
     {
       "epoch": 19.02,
-      "learning_rate": 3.6957898069009186e-05,
-      "loss": 0.5125,
-      "step": 2350
+      "learning_rate": 3.695362028695362e-05,
+      "loss": 0.7743,
+      "step": 2230
     },
     {
       "epoch": 19.02,
-      "learning_rate": 3.687875910098133e-05,
-      "loss": 0.8931,
-      "step": 2360
+      "learning_rate": 3.687020353687021e-05,
+      "loss": 0.8659,
+      "step": 2240
     },
     {
       "epoch": 19.02,
-      "eval_accuracy": 0.728110599078341,
-      "eval_loss": 0.8515065908432007,
-      "eval_runtime": 305.633,
-      "eval_samples_per_second": 0.71,
-      "eval_steps_per_second": 0.18,
-      "step": 2360
+      "eval_accuracy": 0.5852534562211982,
+      "eval_loss": 1.0655064582824707,
+      "eval_runtime": 271.2554,
+      "eval_samples_per_second": 0.8,
+      "eval_steps_per_second": 0.203,
+      "step": 2240
     },
     {
       "epoch": 20.0,
-      "learning_rate": 3.679962013295346e-05,
-      "loss": 0.6621,
-      "step": 2370
+      "learning_rate": 3.678678678678679e-05,
+      "loss": 0.8819,
+      "step": 2250
     },
     {
       "epoch": 20.0,
-      "learning_rate": 3.672048116492561e-05,
-      "loss": 0.5259,
-      "step": 2380
+      "learning_rate": 3.670337003670337e-05,
+      "loss": 0.7077,
+      "step": 2260
     },
     {
       "epoch": 20.0,
-      "learning_rate": 3.6641342196897754e-05,
-      "loss": 0.3653,
-      "step": 2390
-    },
-    {
-      "epoch": 20.01,
-      "learning_rate": 3.65622032288699e-05,
-      "loss": 0.4086,
-      "step": 2400
+      "learning_rate": 3.6619953286619954e-05,
+      "loss": 0.737,
+      "step": 2270
     },
     {
       "epoch": 20.01,
-      "learning_rate": 3.6483064260842046e-05,
-      "loss": 0.4137,
-      "step": 2410
+      "learning_rate": 3.653653653653654e-05,
+      "loss": 0.8414,
+      "step": 2280
     },
     {
       "epoch": 20.01,
-      "learning_rate": 3.640392529281418e-05,
-      "loss": 1.0001,
-      "step": 2420
+      "learning_rate": 3.645311978645312e-05,
+      "loss": 0.8566,
+      "step": 2290
     },
     {
       "epoch": 20.01,
-      "learning_rate": 3.6324786324786323e-05,
-      "loss": 0.7551,
-      "step": 2430
+      "learning_rate": 3.6369703036369704e-05,
+      "loss": 0.8155,
+      "step": 2300
     },
     {
       "epoch": 20.01,
-      "learning_rate": 3.624564735675847e-05,
-      "loss": 0.7024,
-      "step": 2440
+      "learning_rate": 3.628628628628629e-05,
+      "loss": 0.6651,
+      "step": 2310
     },
     {
       "epoch": 20.01,
-      "learning_rate": 3.6166508388730615e-05,
-      "loss": 0.6067,
-      "step": 2450
+      "learning_rate": 3.620286953620287e-05,
+      "loss": 0.7734,
+      "step": 2320
     },
     {
       "epoch": 20.01,
-      "learning_rate": 3.608736942070276e-05,
-      "loss": 1.0071,
-      "step": 2460
+      "learning_rate": 3.6119452786119454e-05,
+      "loss": 0.9056,
+      "step": 2330
     },
     {
       "epoch": 20.02,
-      "learning_rate": 3.60082304526749e-05,
-      "loss": 0.8599,
-      "step": 2470
+      "learning_rate": 3.603603603603604e-05,
+      "loss": 0.9031,
+      "step": 2340
     },
     {
       "epoch": 20.02,
-      "eval_accuracy": 0.6359447004608295,
-      "eval_loss": 1.0786014795303345,
-      "eval_runtime": 314.2163,
-      "eval_samples_per_second": 0.691,
-      "eval_steps_per_second": 0.175,
-      "step": 2478
+      "learning_rate": 3.595261928595262e-05,
+      "loss": 0.5591,
+      "step": 2350
     },
     {
-      "epoch": 21.0,
-      "learning_rate": 3.592909148464704e-05,
-      "loss": 0.803,
-      "step": 2480
+      "epoch": 20.02,
+      "eval_accuracy": 0.576036866359447,
+      "eval_loss": 1.2793947458267212,
+      "eval_runtime": 265.9827,
+      "eval_samples_per_second": 0.816,
+      "eval_steps_per_second": 0.207,
+      "step": 2352
     },
     {
       "epoch": 21.0,
-      "learning_rate": 3.5849952516619184e-05,
-      "loss": 0.4828,
-      "step": 2490
+      "learning_rate": 3.5869202535869205e-05,
+      "loss": 0.6663,
+      "step": 2360
     },
     {
       "epoch": 21.0,
-      "learning_rate": 3.5770813548591326e-05,
-      "loss": 0.3992,
-      "step": 2500
+      "learning_rate": 3.578578578578579e-05,
+      "loss": 0.6847,
+      "step": 2370
     },
     {
       "epoch": 21.0,
-      "learning_rate": 3.5691674580563475e-05,
-      "loss": 0.62,
-      "step": 2510
+      "learning_rate": 3.570236903570237e-05,
+      "loss": 0.8432,
+      "step": 2380
     },
     {
       "epoch": 21.01,
-      "learning_rate": 3.561253561253561e-05,
-      "loss": 0.7068,
-      "step": 2520
+      "learning_rate": 3.5618952285618955e-05,
+      "loss": 0.6488,
+      "step": 2390
     },
     {
       "epoch": 21.01,
-      "learning_rate": 3.553339664450775e-05,
-      "loss": 0.3764,
-      "step": 2530
+      "learning_rate": 3.553553553553554e-05,
+      "loss": 0.9092,
+      "step": 2400
     },
     {
       "epoch": 21.01,
-      "learning_rate": 3.54542576764799e-05,
-      "loss": 0.7479,
-      "step": 2540
+      "learning_rate": 3.545211878545212e-05,
+      "loss": 0.8903,
+      "step": 2410
     },
     {
       "epoch": 21.01,
-      "learning_rate": 3.5375118708452044e-05,
-      "loss": 0.7272,
-      "step": 2550
+      "learning_rate": 3.5368702035368706e-05,
+      "loss": 0.4662,
+      "step": 2420
     },
     {
       "epoch": 21.01,
-      "learning_rate": 3.5295979740424186e-05,
-      "loss": 0.653,
-      "step": 2560
+      "learning_rate": 3.528528528528528e-05,
+      "loss": 0.9775,
+      "step": 2430
     },
     {
       "epoch": 21.01,
-      "learning_rate": 3.521684077239633e-05,
-      "loss": 0.5176,
-      "step": 2570
+      "learning_rate": 3.520186853520187e-05,
+      "loss": 0.973,
+      "step": 2440
     },
     {
       "epoch": 21.01,
-      "learning_rate": 3.513770180436847e-05,
-      "loss": 0.7597,
-      "step": 2580
+      "learning_rate": 3.5118451785118456e-05,
+      "loss": 0.5741,
+      "step": 2450
     },
     {
       "epoch": 21.02,
-      "learning_rate": 3.505856283634061e-05,
-      "loss": 0.5183,
-      "step": 2590
+      "learning_rate": 3.503503503503503e-05,
+      "loss": 0.8963,
+      "step": 2460
     },
     {
       "epoch": 21.02,
-      "eval_accuracy": 0.6866359447004609,
-      "eval_loss": 0.948082447052002,
-      "eval_runtime": 300.3013,
-      "eval_samples_per_second": 0.723,
-      "eval_steps_per_second": 0.183,
-      "step": 2596
-    },
-    {
-      "epoch": 22.0,
-      "learning_rate": 3.497942386831276e-05,
-      "loss": 0.7119,
-      "step": 2600
+      "eval_accuracy": 0.695852534562212,
+      "eval_loss": 1.0049302577972412,
+      "eval_runtime": 270.1661,
+      "eval_samples_per_second": 0.803,
+      "eval_steps_per_second": 0.204,
+      "step": 2464
     },
     {
       "epoch": 22.0,
-      "learning_rate": 3.4900284900284904e-05,
-      "loss": 0.6381,
-      "step": 2610
+      "learning_rate": 3.495161828495162e-05,
+      "loss": 0.7073,
+      "step": 2470
     },
     {
       "epoch": 22.0,
-      "learning_rate": 3.4821145932257046e-05,
-      "loss": 0.8518,
-      "step": 2620
+      "learning_rate": 3.4868201534868206e-05,
+      "loss": 0.5915,
+      "step": 2480
     },
     {
       "epoch": 22.0,
-      "learning_rate": 3.474200696422919e-05,
-      "loss": 0.9462,
-      "step": 2630
+      "learning_rate": 3.478478478478478e-05,
+      "loss": 0.3567,
+      "step": 2490
     },
     {
       "epoch": 22.01,
-      "learning_rate": 3.466286799620133e-05,
-      "loss": 0.6183,
-      "step": 2640
+      "learning_rate": 3.470136803470137e-05,
+      "loss": 0.7007,
+      "step": 2500
     },
     {
       "epoch": 22.01,
-      "learning_rate": 3.458372902817347e-05,
-      "loss": 0.7339,
-      "step": 2650
+      "learning_rate": 3.4617951284617957e-05,
+      "loss": 0.4253,
+      "step": 2510
     },
     {
       "epoch": 22.01,
-      "learning_rate": 3.450459006014562e-05,
-      "loss": 0.8466,
-      "step": 2660
+      "learning_rate": 3.453453453453453e-05,
+      "loss": 0.9297,
+      "step": 2520
     },
     {
       "epoch": 22.01,
-      "learning_rate": 3.4425451092117764e-05,
-      "loss": 0.6976,
-      "step": 2670
+      "learning_rate": 3.4451117784451123e-05,
+      "loss": 0.9365,
+      "step": 2530
     },
     {
       "epoch": 22.01,
-      "learning_rate": 3.43463121240899e-05,
-      "loss": 0.9784,
-      "step": 2680
+      "learning_rate": 3.436770103436771e-05,
+      "loss": 0.4607,
+      "step": 2540
     },
     {
       "epoch": 22.01,
-      "learning_rate": 3.426717315606205e-05,
-      "loss": 0.4932,
-      "step": 2690
+      "learning_rate": 3.4284284284284284e-05,
+      "loss": 0.8208,
+      "step": 2550
     },
     {
       "epoch": 22.01,
-      "learning_rate": 3.418803418803419e-05,
-      "loss": 0.5164,
-      "step": 2700
+      "learning_rate": 3.420086753420087e-05,
+      "loss": 0.8313,
+      "step": 2560
     },
     {
       "epoch": 22.02,
-      "learning_rate": 3.410889522000633e-05,
-      "loss": 0.7982,
-      "step": 2710
+      "learning_rate": 3.411745078411746e-05,
+      "loss": 0.9221,
+      "step": 2570
     },
     {
       "epoch": 22.02,
-      "eval_accuracy": 0.7235023041474654,
-      "eval_loss": 0.8364368677139282,
-      "eval_runtime": 306.1906,
-      "eval_samples_per_second": 0.709,
-      "eval_steps_per_second": 0.18,
-      "step": 2714
+      "eval_accuracy": 0.6082949308755761,
+      "eval_loss": 1.111289620399475,
+      "eval_runtime": 272.9644,
+      "eval_samples_per_second": 0.795,
+      "eval_steps_per_second": 0.201,
+      "step": 2576
     },
     {
       "epoch": 23.0,
-      "learning_rate": 3.4029756251978475e-05,
-      "loss": 0.6178,
-      "step": 2720
+      "learning_rate": 3.4034034034034034e-05,
+      "loss": 0.8386,
+      "step": 2580
     },
     {
       "epoch": 23.0,
       "learning_rate": 3.395061728395062e-05,
-      "loss": 0.5603,
-      "step": 2730
+      "loss": 0.549,
+      "step": 2590
     },
     {
       "epoch": 23.0,
-      "learning_rate": 3.387147831592276e-05,
-      "loss": 0.6577,
-      "step": 2740
+      "learning_rate": 3.386720053386721e-05,
+      "loss": 1.0528,
+      "step": 2600
     },
     {
       "epoch": 23.01,
-      "learning_rate": 3.37923393478949e-05,
-      "loss": 0.7942,
-      "step": 2750
+      "learning_rate": 3.3783783783783784e-05,
+      "loss": 0.6398,
+      "step": 2610
     },
     {
       "epoch": 23.01,
-      "learning_rate": 3.371320037986705e-05,
-      "loss": 0.4468,
-      "step": 2760
+      "learning_rate": 3.370036703370037e-05,
+      "loss": 0.7276,
+      "step": 2620
     },
     {
       "epoch": 23.01,
-      "learning_rate": 3.363406141183919e-05,
-      "loss": 0.6931,
-      "step": 2770
+      "learning_rate": 3.361695028361695e-05,
+      "loss": 0.8483,
+      "step": 2630
     },
     {
       "epoch": 23.01,
-      "learning_rate": 3.355492244381133e-05,
-      "loss": 0.5835,
-      "step": 2780
+      "learning_rate": 3.3533533533533535e-05,
+      "loss": 0.5737,
+      "step": 2640
     },
     {
       "epoch": 23.01,
-      "learning_rate": 3.347578347578348e-05,
-      "loss": 0.7099,
-      "step": 2790
+      "learning_rate": 3.345011678345012e-05,
+      "loss": 0.588,
+      "step": 2650
     },
     {
       "epoch": 23.01,
-      "learning_rate": 3.339664450775562e-05,
-      "loss": 0.6785,
-      "step": 2800
+      "learning_rate": 3.33667000333667e-05,
+      "loss": 0.6416,
+      "step": 2660
     },
     {
       "epoch": 23.01,
-      "learning_rate": 3.331750553972776e-05,
-      "loss": 0.3025,
-      "step": 2810
-    },
-    {
-      "epoch": 23.02,
-      "learning_rate": 3.323836657169991e-05,
-      "loss": 0.9246,
-      "step": 2820
+      "learning_rate": 3.3283283283283285e-05,
+      "loss": 0.7938,
+      "step": 2670
     },
     {
       "epoch": 23.02,
-      "learning_rate": 3.3159227603672046e-05,
-      "loss": 1.0003,
-      "step": 2830
+      "learning_rate": 3.319986653319987e-05,
+      "loss": 0.7154,
+      "step": 2680
     },
     {
       "epoch": 23.02,
-      "eval_accuracy": 0.7327188940092166,
-      "eval_loss": 0.7810962796211243,
-      "eval_runtime": 298.2745,
-      "eval_samples_per_second": 0.728,
-      "eval_steps_per_second": 0.184,
-      "step": 2832
+      "eval_accuracy": 0.6405529953917051,
+      "eval_loss": 0.9371297359466553,
+      "eval_runtime": 283.5798,
+      "eval_samples_per_second": 0.765,
+      "eval_steps_per_second": 0.194,
+      "step": 2688
     },
     {
       "epoch": 24.0,
-      "learning_rate": 3.308008863564419e-05,
-      "loss": 1.1583,
-      "step": 2840
+      "learning_rate": 3.311644978311645e-05,
+      "loss": 0.4492,
+      "step": 2690
     },
     {
       "epoch": 24.0,
-      "learning_rate": 3.300094966761634e-05,
-      "loss": 0.4923,
-      "step": 2850
+      "learning_rate": 3.3033033033033035e-05,
+      "loss": 0.6304,
+      "step": 2700
     },
     {
       "epoch": 24.0,
-      "learning_rate": 3.292181069958848e-05,
-      "loss": 0.4868,
-      "step": 2860
+      "learning_rate": 3.294961628294962e-05,
+      "loss": 0.7286,
+      "step": 2710
     },
     {
-      "epoch": 24.01,
-      "learning_rate": 3.284267173156062e-05,
-      "loss": 0.7352,
-      "step": 2870
+      "epoch": 24.0,
+      "learning_rate": 3.2866199532866195e-05,
+      "loss": 0.9193,
+      "step": 2720
     },
     {
       "epoch": 24.01,
-      "learning_rate": 3.2763532763532764e-05,
-      "loss": 0.8188,
-      "step": 2880
+      "learning_rate": 3.2782782782782786e-05,
+      "loss": 0.7511,
+      "step": 2730
     },
     {
       "epoch": 24.01,
-      "learning_rate": 3.268439379550491e-05,
-      "loss": 0.8439,
-      "step": 2890
+      "learning_rate": 3.269936603269937e-05,
+      "loss": 0.7076,
+      "step": 2740
     },
     {
       "epoch": 24.01,
-      "learning_rate": 3.260525482747705e-05,
-      "loss": 0.821,
-      "step": 2900
+      "learning_rate": 3.2615949282615946e-05,
+      "loss": 0.7735,
+      "step": 2750
     },
     {
       "epoch": 24.01,
-      "learning_rate": 3.25261158594492e-05,
-      "loss": 0.6092,
-      "step": 2910
+      "learning_rate": 3.2532532532532536e-05,
+      "loss": 0.9404,
+      "step": 2760
     },
     {
       "epoch": 24.01,
-      "learning_rate": 3.244697689142134e-05,
-      "loss": 0.7416,
-      "step": 2920
+      "learning_rate": 3.244911578244912e-05,
+      "loss": 0.9553,
+      "step": 2770
     },
     {
       "epoch": 24.01,
-      "learning_rate": 3.236783792339348e-05,
-      "loss": 0.4177,
-      "step": 2930
+      "learning_rate": 3.2365699032365696e-05,
+      "loss": 1.0067,
+      "step": 2780
     },
     {
       "epoch": 24.02,
-      "learning_rate": 3.2288698955365625e-05,
-      "loss": 0.5732,
-      "step": 2940
+      "learning_rate": 3.2282282282282286e-05,
+      "loss": 0.7712,
+      "step": 2790
     },
     {
       "epoch": 24.02,
-      "learning_rate": 3.220955998733777e-05,
-      "loss": 0.6666,
-      "step": 2950
+      "learning_rate": 3.219886553219887e-05,
+      "loss": 0.8795,
+      "step": 2800
     },
     {
       "epoch": 24.02,
-      "eval_accuracy": 0.7465437788018433,
-      "eval_loss": 0.7551702857017517,
-      "eval_runtime": 297.83,
-      "eval_samples_per_second": 0.729,
-      "eval_steps_per_second": 0.185,
-      "step": 2950
+      "eval_accuracy": 0.7235023041474654,
+      "eval_loss": 0.6837921738624573,
+      "eval_runtime": 264.0243,
+      "eval_samples_per_second": 0.822,
+      "eval_steps_per_second": 0.208,
+      "step": 2800
     },
     {
       "epoch": 25.0,
-      "learning_rate": 3.213042101930991e-05,
-      "loss": 0.5516,
-      "step": 2960
+      "learning_rate": 3.2115448782115447e-05,
+      "loss": 0.6295,
+      "step": 2810
     },
     {
       "epoch": 25.0,
-      "learning_rate": 3.205128205128206e-05,
-      "loss": 0.5643,
-      "step": 2970
+      "learning_rate": 3.203203203203203e-05,
+      "loss": 0.5106,
+      "step": 2820
     },
     {
       "epoch": 25.0,
-      "learning_rate": 3.1972143083254193e-05,
-      "loss": 0.7,
-      "step": 2980
-    },
-    {
-      "epoch": 25.01,
-      "learning_rate": 3.1893004115226336e-05,
-      "loss": 0.4949,
-      "step": 2990
+      "learning_rate": 3.194861528194862e-05,
+      "loss": 0.6515,
+      "step": 2830
     },
     {
       "epoch": 25.01,
-      "learning_rate": 3.181386514719848e-05,
-      "loss": 0.6588,
-      "step": 3000
+      "learning_rate": 3.18651985318652e-05,
+      "loss": 0.5182,
+      "step": 2840
     },
     {
       "epoch": 25.01,
-      "learning_rate": 3.173472617917063e-05,
-      "loss": 0.6374,
-      "step": 3010
+      "learning_rate": 3.178178178178178e-05,
+      "loss": 0.9393,
+      "step": 2850
     },
     {
       "epoch": 25.01,
-      "learning_rate": 3.165558721114277e-05,
-      "loss": 0.4516,
-      "step": 3020
+      "learning_rate": 3.169836503169837e-05,
+      "loss": 0.6204,
+      "step": 2860
     },
     {
       "epoch": 25.01,
-      "learning_rate": 3.157644824311491e-05,
-      "loss": 0.6673,
-      "step": 3030
+      "learning_rate": 3.161494828161495e-05,
+      "loss": 0.5407,
+      "step": 2870
     },
     {
       "epoch": 25.01,
-      "learning_rate": 3.1497309275087054e-05,
-      "loss": 0.6344,
-      "step": 3040
+      "learning_rate": 3.153153153153153e-05,
+      "loss": 0.8556,
+      "step": 2880
     },
     {
       "epoch": 25.01,
-      "learning_rate": 3.1418170307059196e-05,
-      "loss": 0.422,
-      "step": 3050
+      "learning_rate": 3.144811478144812e-05,
+      "loss": 0.6423,
+      "step": 2890
     },
     {
       "epoch": 25.02,
-      "learning_rate": 3.133903133903134e-05,
-      "loss": 0.8527,
-      "step": 3060
+      "learning_rate": 3.13646980313647e-05,
+      "loss": 0.8295,
+      "step": 2900
     },
     {
       "epoch": 25.02,
-      "eval_accuracy": 0.7188940092165899,
-      "eval_loss": 0.8201001286506653,
-      "eval_runtime": 289.214,
-      "eval_samples_per_second": 0.75,
-      "eval_steps_per_second": 0.19,
-      "step": 3068
+      "learning_rate": 3.128128128128128e-05,
+      "loss": 0.631,
+      "step": 2910
     },
     {
-      "epoch": 26.0,
-      "learning_rate": 3.125989237100349e-05,
-      "loss": 0.71,
-      "step": 3070
+      "epoch": 25.02,
+      "eval_accuracy": 0.6129032258064516,
+      "eval_loss": 1.209276556968689,
+      "eval_runtime": 285.1576,
+      "eval_samples_per_second": 0.761,
+      "eval_steps_per_second": 0.193,
+      "step": 2912
     },
     {
       "epoch": 26.0,
-      "learning_rate": 3.118075340297563e-05,
-      "loss": 0.7812,
-      "step": 3080
+      "learning_rate": 3.119786453119787e-05,
+      "loss": 0.7184,
+      "step": 2920
     },
     {
       "epoch": 26.0,
-      "learning_rate": 3.1101614434947765e-05,
-      "loss": 0.4559,
-      "step": 3090
+      "learning_rate": 3.111444778111445e-05,
+      "loss": 0.6305,
+      "step": 2930
     },
     {
       "epoch": 26.0,
-      "learning_rate": 3.1022475466919914e-05,
-      "loss": 0.5269,
-      "step": 3100
+      "learning_rate": 3.103103103103103e-05,
+      "loss": 0.7949,
+      "step": 2940
     },
     {
       "epoch": 26.01,
-      "learning_rate": 3.0943336498892056e-05,
-      "loss": 0.75,
-      "step": 3110
+      "learning_rate": 3.0947614280947615e-05,
+      "loss": 1.2476,
+      "step": 2950
     },
     {
       "epoch": 26.01,
       "learning_rate": 3.08641975308642e-05,
-      "loss": 0.7118,
-      "step": 3120
+      "loss": 0.7922,
+      "step": 2960
     },
     {
       "epoch": 26.01,
-      "learning_rate": 3.078505856283635e-05,
-      "loss": 0.7999,
-      "step": 3130
+      "learning_rate": 3.078078078078078e-05,
+      "loss": 0.581,
+      "step": 2970
     },
     {
       "epoch": 26.01,
-      "learning_rate": 3.070591959480848e-05,
-      "loss": 0.4535,
-      "step": 3140
+      "learning_rate": 3.0697364030697365e-05,
+      "loss": 0.6329,
+      "step": 2980
     },
     {
       "epoch": 26.01,
-      "learning_rate": 3.0626780626780625e-05,
-      "loss": 0.7403,
-      "step": 3150
+      "learning_rate": 3.061394728061395e-05,
+      "loss": 0.8019,
+      "step": 2990
     },
     {
       "epoch": 26.01,
-      "learning_rate": 3.0547641658752774e-05,
-      "loss": 0.7067,
-      "step": 3160
+      "learning_rate": 3.053053053053053e-05,
+      "loss": 0.7641,
+      "step": 3000
     },
     {
       "epoch": 26.01,
-      "learning_rate": 3.0468502690724916e-05,
+      "learning_rate": 3.044711378044712e-05,
       "loss": 0.6169,
-      "step": 3170
+      "step": 3010
     },
     {
       "epoch": 26.02,
-      "learning_rate": 3.0389363722697055e-05,
-      "loss": 0.4678,
-      "step": 3180
+      "learning_rate": 3.0363697030363696e-05,
+      "loss": 1.0489,
+      "step": 3020
     },
     {
       "epoch": 26.02,
-      "eval_accuracy": 0.695852534562212,
-      "eval_loss": 1.0259956121444702,
-      "eval_runtime": 293.3586,
-      "eval_samples_per_second": 0.74,
-      "eval_steps_per_second": 0.187,
-      "step": 3186
-    },
-    {
-      "epoch": 27.0,
-      "learning_rate": 3.0310224754669204e-05,
-      "loss": 0.4734,
-      "step": 3190
+      "eval_accuracy": 0.5483870967741935,
+      "eval_loss": 1.472022294998169,
+      "eval_runtime": 262.7206,
+      "eval_samples_per_second": 0.826,
+      "eval_steps_per_second": 0.209,
+      "step": 3024
     },
     {
       "epoch": 27.0,
-      "learning_rate": 3.0231085786641343e-05,
-      "loss": 0.6945,
-      "step": 3200
+      "learning_rate": 3.0280280280280282e-05,
+      "loss": 0.5064,
+      "step": 3030
     },
     {
       "epoch": 27.0,
-      "learning_rate": 3.0151946818613485e-05,
-      "loss": 0.667,
-      "step": 3210
+      "learning_rate": 3.0196863530196866e-05,
+      "loss": 0.493,
+      "step": 3040
     },
     {
       "epoch": 27.0,
-      "learning_rate": 3.0072807850585634e-05,
-      "loss": 0.6138,
-      "step": 3220
+      "learning_rate": 3.0113446780113446e-05,
+      "loss": 0.6668,
+      "step": 3050
     },
     {
       "epoch": 27.01,
-      "learning_rate": 2.9993668882557773e-05,
-      "loss": 0.7627,
-      "step": 3230
+      "learning_rate": 3.0030030030030033e-05,
+      "loss": 0.5836,
+      "step": 3060
     },
     {
       "epoch": 27.01,
-      "learning_rate": 2.9914529914529915e-05,
-      "loss": 0.8703,
-      "step": 3240
+      "learning_rate": 2.9946613279946616e-05,
+      "loss": 0.997,
+      "step": 3070
     },
     {
       "epoch": 27.01,
-      "learning_rate": 2.9835390946502057e-05,
-      "loss": 0.6306,
-      "step": 3250
+      "learning_rate": 2.9863196529863196e-05,
+      "loss": 0.8393,
+      "step": 3080
     },
     {
       "epoch": 27.01,
-      "learning_rate": 2.9756251978474203e-05,
-      "loss": 0.7293,
-      "step": 3260
+      "learning_rate": 2.9779779779779783e-05,
+      "loss": 0.609,
+      "step": 3090
     },
     {
       "epoch": 27.01,
-      "learning_rate": 2.9677113010446345e-05,
-      "loss": 0.5378,
-      "step": 3270
+      "learning_rate": 2.9696363029696367e-05,
+      "loss": 0.7111,
+      "step": 3100
     },
     {
       "epoch": 27.01,
-      "learning_rate": 2.9597974042418487e-05,
-      "loss": 0.9667,
-      "step": 3280
+      "learning_rate": 2.9612946279612947e-05,
+      "loss": 0.7491,
+      "step": 3110
     },
     {
       "epoch": 27.01,
-      "learning_rate": 2.9518835074390633e-05,
-      "loss": 0.4905,
-      "step": 3290
+      "learning_rate": 2.952952952952953e-05,
+      "loss": 0.69,
+      "step": 3120
     },
     {
       "epoch": 27.02,
-      "learning_rate": 2.9439696106362775e-05,
-      "loss": 0.7354,
-      "step": 3300
+      "learning_rate": 2.9446112779446117e-05,
+      "loss": 0.5881,
+      "step": 3130
     },
     {
       "epoch": 27.02,
-      "eval_accuracy": 0.6866359447004609,
-      "eval_loss": 0.8520135879516602,
-      "eval_runtime": 253.7054,
-      "eval_samples_per_second": 0.855,
-      "eval_steps_per_second": 0.217,
-      "step": 3304
+      "eval_accuracy": 0.631336405529954,
+      "eval_loss": 1.1905186176300049,
+      "eval_runtime": 281.4297,
+      "eval_samples_per_second": 0.771,
+      "eval_steps_per_second": 0.195,
+      "step": 3136
     },
     {
       "epoch": 28.0,
-      "learning_rate": 2.9360557138334914e-05,
-      "loss": 0.4119,
-      "step": 3310
+      "learning_rate": 2.9362696029362697e-05,
+      "loss": 1.0567,
+      "step": 3140
     },
     {
       "epoch": 28.0,
-      "learning_rate": 2.9281418170307063e-05,
-      "loss": 0.5532,
-      "step": 3320
+      "learning_rate": 2.927927927927928e-05,
+      "loss": 0.5978,
+      "step": 3150
     },
     {
       "epoch": 28.0,
-      "learning_rate": 2.9202279202279202e-05,
-      "loss": 0.5374,
-      "step": 3330
+      "learning_rate": 2.9195862529195867e-05,
+      "loss": 0.5434,
+      "step": 3160
     },
     {
       "epoch": 28.01,
-      "learning_rate": 2.9123140234251344e-05,
-      "loss": 0.4628,
-      "step": 3340
+      "learning_rate": 2.9112445779112447e-05,
+      "loss": 0.6441,
+      "step": 3170
     },
     {
       "epoch": 28.01,
-      "learning_rate": 2.9044001266223493e-05,
-      "loss": 0.6509,
-      "step": 3350
+      "learning_rate": 2.902902902902903e-05,
+      "loss": 0.8564,
+      "step": 3180
     },
     {
       "epoch": 28.01,
-      "learning_rate": 2.8964862298195632e-05,
-      "loss": 0.3157,
-      "step": 3360
+      "learning_rate": 2.8945612278945618e-05,
+      "loss": 0.7492,
+      "step": 3190
     },
     {
       "epoch": 28.01,
-      "learning_rate": 2.8885723330167774e-05,
-      "loss": 0.5705,
-      "step": 3370
+      "learning_rate": 2.8862195528862194e-05,
+      "loss": 0.7439,
+      "step": 3200
     },
     {
       "epoch": 28.01,
-      "learning_rate": 2.880658436213992e-05,
-      "loss": 0.5273,
-      "step": 3380
+      "learning_rate": 2.877877877877878e-05,
+      "loss": 0.731,
+      "step": 3210
     },
     {
       "epoch": 28.01,
-      "learning_rate": 2.8727445394112062e-05,
-      "loss": 0.4847,
-      "step": 3390
+      "learning_rate": 2.8695362028695365e-05,
+      "loss": 0.8052,
+      "step": 3220
     },
     {
       "epoch": 28.01,
-      "learning_rate": 2.8648306426084204e-05,
-      "loss": 0.5194,
-      "step": 3400
-    },
-    {
-      "epoch": 28.02,
-      "learning_rate": 2.856916745805635e-05,
-      "loss": 0.391,
-      "step": 3410
+      "learning_rate": 2.8611945278611945e-05,
+      "loss": 0.8463,
+      "step": 3230
     },
     {
       "epoch": 28.02,
-      "learning_rate": 2.8490028490028492e-05,
-      "loss": 1.1097,
-      "step": 3420
+      "learning_rate": 2.852852852852853e-05,
+      "loss": 0.7919,
+      "step": 3240
     },
     {
       "epoch": 28.02,
-      "eval_accuracy": 0.7327188940092166,
-      "eval_loss": 0.9238936901092529,
-      "eval_runtime": 306.5306,
-      "eval_samples_per_second": 0.708,
-      "eval_steps_per_second": 0.179,
-      "step": 3422
+      "eval_accuracy": 0.576036866359447,
+      "eval_loss": 1.1292119026184082,
+      "eval_runtime": 300.8957,
+      "eval_samples_per_second": 0.721,
+      "eval_steps_per_second": 0.183,
+      "step": 3248
     },
     {
       "epoch": 29.0,
-      "learning_rate": 2.8410889522000634e-05,
-      "loss": 0.7316,
-      "step": 3430
+      "learning_rate": 2.844511177844511e-05,
+      "loss": 0.9881,
+      "step": 3250
     },
     {
       "epoch": 29.0,
-      "learning_rate": 2.833175055397278e-05,
-      "loss": 0.5964,
-      "step": 3440
+      "learning_rate": 2.8361695028361695e-05,
+      "loss": 0.4171,
+      "step": 3260
     },
     {
       "epoch": 29.0,
-      "learning_rate": 2.8252611585944922e-05,
-      "loss": 0.4294,
-      "step": 3450
+      "learning_rate": 2.8278278278278282e-05,
+      "loss": 0.4612,
+      "step": 3270
     },
     {
-      "epoch": 29.01,
-      "learning_rate": 2.817347261791706e-05,
-      "loss": 0.5873,
-      "step": 3460
+      "epoch": 29.0,
+      "learning_rate": 2.8194861528194862e-05,
+      "loss": 0.6761,
+      "step": 3280
     },
     {
       "epoch": 29.01,
-      "learning_rate": 2.8094333649889203e-05,
-      "loss": 0.5362,
-      "step": 3470
+      "learning_rate": 2.8111444778111445e-05,
+      "loss": 0.4762,
+      "step": 3290
     },
     {
       "epoch": 29.01,
-      "learning_rate": 2.8015194681861352e-05,
-      "loss": 0.7654,
-      "step": 3480
+      "learning_rate": 2.8028028028028032e-05,
+      "loss": 0.6633,
+      "step": 3300
     },
     {
       "epoch": 29.01,
-      "learning_rate": 2.793605571383349e-05,
-      "loss": 0.6749,
-      "step": 3490
+      "learning_rate": 2.794461127794461e-05,
+      "loss": 0.5644,
+      "step": 3310
     },
     {
       "epoch": 29.01,
-      "learning_rate": 2.7856916745805633e-05,
-      "loss": 0.392,
-      "step": 3500
+      "learning_rate": 2.7861194527861196e-05,
+      "loss": 0.9361,
+      "step": 3320
     },
     {
       "epoch": 29.01,
       "learning_rate": 2.777777777777778e-05,
-      "loss": 0.9069,
-      "step": 3510
+      "loss": 0.8429,
+      "step": 3330
     },
     {
       "epoch": 29.01,
-      "learning_rate": 2.769863880974992e-05,
-      "loss": 0.4994,
-      "step": 3520
+      "learning_rate": 2.769436102769436e-05,
+      "loss": 0.6224,
+      "step": 3340
     },
     {
       "epoch": 29.02,
-      "learning_rate": 2.7619499841722064e-05,
-      "loss": 0.7977,
-      "step": 3530
+      "learning_rate": 2.7610944277610946e-05,
+      "loss": 0.8489,
+      "step": 3350
     },
     {
       "epoch": 29.02,
-      "learning_rate": 2.754036087369421e-05,
-      "loss": 0.6264,
-      "step": 3540
+      "learning_rate": 2.752752752752753e-05,
+      "loss": 0.9158,
+      "step": 3360
     },
     {
       "epoch": 29.02,
-      "eval_accuracy": 0.7557603686635944,
-      "eval_loss": 0.6894146800041199,
-      "eval_runtime": 290.6314,
-      "eval_samples_per_second": 0.747,
-      "eval_steps_per_second": 0.189,
-      "step": 3540
+      "eval_accuracy": 0.6359447004608295,
+      "eval_loss": 1.0214248895645142,
+      "eval_runtime": 281.3806,
+      "eval_samples_per_second": 0.771,
+      "eval_steps_per_second": 0.195,
+      "step": 3360
     },
     {
       "epoch": 30.0,
-      "learning_rate": 2.746122190566635e-05,
-      "loss": 1.0999,
-      "step": 3550
+      "learning_rate": 2.744411077744411e-05,
+      "loss": 0.6866,
+      "step": 3370
     },
     {
       "epoch": 30.0,
-      "learning_rate": 2.7382082937638494e-05,
-      "loss": 0.5372,
-      "step": 3560
+      "learning_rate": 2.7360694027360696e-05,
+      "loss": 0.8052,
+      "step": 3380
     },
     {
       "epoch": 30.0,
-      "learning_rate": 2.730294396961064e-05,
-      "loss": 0.595,
-      "step": 3570
-    },
-    {
-      "epoch": 30.01,
-      "learning_rate": 2.722380500158278e-05,
-      "loss": 0.9679,
-      "step": 3580
+      "learning_rate": 2.727727727727728e-05,
+      "loss": 0.6658,
+      "step": 3390
     },
     {
       "epoch": 30.01,
-      "learning_rate": 2.714466603355492e-05,
-      "loss": 0.9635,
-      "step": 3590
+      "learning_rate": 2.719386052719386e-05,
+      "loss": 0.6167,
+      "step": 3400
     },
     {
       "epoch": 30.01,
-      "learning_rate": 2.706552706552707e-05,
-      "loss": 0.3871,
-      "step": 3600
+      "learning_rate": 2.7110443777110443e-05,
+      "loss": 0.5109,
+      "step": 3410
     },
     {
       "epoch": 30.01,
-      "learning_rate": 2.698638809749921e-05,
-      "loss": 0.4446,
-      "step": 3610
+      "learning_rate": 2.702702702702703e-05,
+      "loss": 1.1015,
+      "step": 3420
     },
     {
       "epoch": 30.01,
-      "learning_rate": 2.690724912947135e-05,
-      "loss": 0.587,
-      "step": 3620
+      "learning_rate": 2.694361027694361e-05,
+      "loss": 0.7845,
+      "step": 3430
     },
     {
       "epoch": 30.01,
-      "learning_rate": 2.68281101614435e-05,
-      "loss": 0.5359,
-      "step": 3630
+      "learning_rate": 2.6860193526860194e-05,
+      "loss": 0.6872,
+      "step": 3440
     },
     {
       "epoch": 30.01,
-      "learning_rate": 2.6748971193415638e-05,
-      "loss": 0.7691,
-      "step": 3640
+      "learning_rate": 2.677677677677678e-05,
+      "loss": 0.4606,
+      "step": 3450
     },
     {
       "epoch": 30.02,
-      "learning_rate": 2.666983222538778e-05,
-      "loss": 0.3348,
-      "step": 3650
+      "learning_rate": 2.669336002669336e-05,
+      "loss": 0.7213,
+      "step": 3460
     },
     {
       "epoch": 30.02,
-      "eval_accuracy": 0.8064516129032258,
-      "eval_loss": 0.6229776740074158,
-      "eval_runtime": 293.7254,
-      "eval_samples_per_second": 0.739,
-      "eval_steps_per_second": 0.187,
-      "step": 3658
+      "learning_rate": 2.6609943276609944e-05,
+      "loss": 0.8319,
+      "step": 3470
     },
     {
-      "epoch": 31.0,
-      "learning_rate": 2.6590693257359926e-05,
-      "loss": 0.9398,
-      "step": 3660
+      "epoch": 30.02,
+      "eval_accuracy": 0.6682027649769585,
+      "eval_loss": 1.2861884832382202,
+      "eval_runtime": 276.1791,
+      "eval_samples_per_second": 0.786,
+      "eval_steps_per_second": 0.199,
+      "step": 3472
     },
     {
       "epoch": 31.0,
-      "learning_rate": 2.651155428933207e-05,
-      "loss": 0.6431,
-      "step": 3670
+      "learning_rate": 2.652652652652653e-05,
+      "loss": 0.7171,
+      "step": 3480
     },
     {
       "epoch": 31.0,
-      "learning_rate": 2.643241532130421e-05,
-      "loss": 0.3879,
-      "step": 3680
+      "learning_rate": 2.6443109776443108e-05,
+      "loss": 1.2605,
+      "step": 3490
     },
     {
       "epoch": 31.0,
-      "learning_rate": 2.6353276353276356e-05,
-      "loss": 0.4384,
-      "step": 3690
+      "learning_rate": 2.6359693026359694e-05,
+      "loss": 0.7113,
+      "step": 3500
     },
     {
       "epoch": 31.01,
-      "learning_rate": 2.62741373852485e-05,
-      "loss": 0.6234,
-      "step": 3700
+      "learning_rate": 2.6276276276276278e-05,
+      "loss": 0.4538,
+      "step": 3510
     },
     {
       "epoch": 31.01,
-      "learning_rate": 2.619499841722064e-05,
-      "loss": 0.7331,
-      "step": 3710
+      "learning_rate": 2.6192859526192858e-05,
+      "loss": 0.5572,
+      "step": 3520
     },
     {
       "epoch": 31.01,
-      "learning_rate": 2.611585944919278e-05,
-      "loss": 0.3598,
-      "step": 3720
+      "learning_rate": 2.6109442776109445e-05,
+      "loss": 0.7715,
+      "step": 3530
     },
     {
       "epoch": 31.01,
-      "learning_rate": 2.603672048116493e-05,
-      "loss": 0.6426,
-      "step": 3730
+      "learning_rate": 2.6026026026026028e-05,
+      "loss": 1.0429,
+      "step": 3540
     },
     {
       "epoch": 31.01,
-      "learning_rate": 2.595758151313707e-05,
-      "loss": 0.5889,
-      "step": 3740
+      "learning_rate": 2.5942609275942608e-05,
+      "loss": 0.7335,
+      "step": 3550
     },
     {
       "epoch": 31.01,
-      "learning_rate": 2.587844254510921e-05,
-      "loss": 0.6661,
-      "step": 3750
+      "learning_rate": 2.5859192525859195e-05,
+      "loss": 0.7139,
+      "step": 3560
     },
     {
       "epoch": 31.01,
-      "learning_rate": 2.579930357708136e-05,
-      "loss": 0.672,
-      "step": 3760
+      "learning_rate": 2.577577577577578e-05,
+      "loss": 0.422,
+      "step": 3570
     },
     {
       "epoch": 31.02,
-      "learning_rate": 2.5720164609053497e-05,
-      "loss": 0.5548,
-      "step": 3770
+      "learning_rate": 2.569235902569236e-05,
+      "loss": 0.6775,
+      "step": 3580
     },
     {
       "epoch": 31.02,
-      "eval_accuracy": 0.8202764976958525,
-      "eval_loss": 0.6430536508560181,
-      "eval_runtime": 296.7673,
-      "eval_samples_per_second": 0.731,
-      "eval_steps_per_second": 0.185,
-      "step": 3776
-    },
-    {
-      "epoch": 32.0,
-      "learning_rate": 2.564102564102564e-05,
-      "loss": 0.6612,
-      "step": 3780
+      "eval_accuracy": 0.6405529953917051,
+      "eval_loss": 1.0971248149871826,
+      "eval_runtime": 295.308,
+      "eval_samples_per_second": 0.735,
+      "eval_steps_per_second": 0.186,
+      "step": 3584
     },
     {
       "epoch": 32.0,
-      "learning_rate": 2.5561886672997785e-05,
-      "loss": 0.7013,
-      "step": 3790
+      "learning_rate": 2.5608942275608942e-05,
+      "loss": 0.3834,
+      "step": 3590
     },
     {
       "epoch": 32.0,
-      "learning_rate": 2.5482747704969927e-05,
-      "loss": 0.7166,
-      "step": 3800
+      "learning_rate": 2.552552552552553e-05,
+      "loss": 0.5938,
+      "step": 3600
     },
     {
       "epoch": 32.0,
-      "learning_rate": 2.540360873694207e-05,
-      "loss": 0.8458,
-      "step": 3810
+      "learning_rate": 2.544210877544211e-05,
+      "loss": 0.6666,
+      "step": 3610
     },
     {
       "epoch": 32.01,
-      "learning_rate": 2.5324469768914215e-05,
-      "loss": 0.7025,
-      "step": 3820
+      "learning_rate": 2.5358692025358692e-05,
+      "loss": 0.7156,
+      "step": 3620
     },
     {
       "epoch": 32.01,
-      "learning_rate": 2.5245330800886358e-05,
-      "loss": 0.4634,
-      "step": 3830
+      "learning_rate": 2.527527527527528e-05,
+      "loss": 0.816,
+      "step": 3630
     },
     {
       "epoch": 32.01,
-      "learning_rate": 2.51661918328585e-05,
-      "loss": 0.7509,
-      "step": 3840
+      "learning_rate": 2.519185852519186e-05,
+      "loss": 0.7641,
+      "step": 3640
     },
     {
       "epoch": 32.01,
-      "learning_rate": 2.5087052864830645e-05,
-      "loss": 0.5975,
-      "step": 3850
+      "learning_rate": 2.5108441775108443e-05,
+      "loss": 0.7681,
+      "step": 3650
     },
     {
       "epoch": 32.01,
-      "learning_rate": 2.5007913896802788e-05,
-      "loss": 0.6689,
-      "step": 3860
+      "learning_rate": 2.502502502502503e-05,
+      "loss": 0.9274,
+      "step": 3660
     },
     {
       "epoch": 32.01,
-      "learning_rate": 2.492877492877493e-05,
-      "loss": 0.4826,
-      "step": 3870
+      "learning_rate": 2.494160827494161e-05,
+      "loss": 0.8763,
+      "step": 3670
     },
     {
       "epoch": 32.01,
-      "learning_rate": 2.4849635960747072e-05,
-      "loss": 0.4976,
-      "step": 3880
+      "learning_rate": 2.4858191524858193e-05,
+      "loss": 0.5928,
+      "step": 3680
     },
     {
       "epoch": 32.02,
-      "learning_rate": 2.4770496992719218e-05,
-      "loss": 0.4242,
-      "step": 3890
+      "learning_rate": 2.4774774774774777e-05,
+      "loss": 0.7191,
+      "step": 3690
     },
     {
       "epoch": 32.02,
-      "eval_accuracy": 0.7050691244239631,
-      "eval_loss": 0.8081349730491638,
-      "eval_runtime": 298.374,
-      "eval_samples_per_second": 0.727,
-      "eval_steps_per_second": 0.184,
-      "step": 3894
+      "eval_accuracy": 0.6497695852534562,
+      "eval_loss": 1.026401400566101,
+      "eval_runtime": 280.0325,
+      "eval_samples_per_second": 0.775,
+      "eval_steps_per_second": 0.196,
+      "step": 3696
     },
     {
       "epoch": 33.0,
       "learning_rate": 2.4691358024691357e-05,
-      "loss": 0.3113,
-      "step": 3900
+      "loss": 0.9832,
+      "step": 3700
     },
     {
       "epoch": 33.0,
-      "learning_rate": 2.4612219056663502e-05,
-      "loss": 0.6925,
-      "step": 3910
+      "learning_rate": 2.4607941274607943e-05,
+      "loss": 0.7523,
+      "step": 3710
     },
     {
       "epoch": 33.0,
-      "learning_rate": 2.4533080088635644e-05,
-      "loss": 0.6335,
-      "step": 3920
+      "learning_rate": 2.4524524524524527e-05,
+      "loss": 0.6381,
+      "step": 3720
     },
     {
       "epoch": 33.01,
-      "learning_rate": 2.4453941120607787e-05,
-      "loss": 0.5331,
-      "step": 3930
+      "learning_rate": 2.4441107774441107e-05,
+      "loss": 0.7641,
+      "step": 3730
     },
     {
       "epoch": 33.01,
-      "learning_rate": 2.4374802152579932e-05,
-      "loss": 0.6092,
-      "step": 3940
+      "learning_rate": 2.4357691024357694e-05,
+      "loss": 0.5699,
+      "step": 3740
     },
     {
       "epoch": 33.01,
-      "learning_rate": 2.4295663184552074e-05,
-      "loss": 0.534,
-      "step": 3950
+      "learning_rate": 2.4274274274274274e-05,
+      "loss": 0.5342,
+      "step": 3750
     },
     {
       "epoch": 33.01,
-      "learning_rate": 2.4216524216524217e-05,
-      "loss": 0.4707,
-      "step": 3960
+      "learning_rate": 2.4190857524190857e-05,
+      "loss": 0.4933,
+      "step": 3760
     },
     {
       "epoch": 33.01,
-      "learning_rate": 2.4137385248496362e-05,
-      "loss": 0.5926,
-      "step": 3970
+      "learning_rate": 2.4107440774107444e-05,
+      "loss": 0.7027,
+      "step": 3770
     },
     {
       "epoch": 33.01,
-      "learning_rate": 2.4058246280468505e-05,
-      "loss": 0.2674,
-      "step": 3980
+      "learning_rate": 2.4024024024024024e-05,
+      "loss": 0.5812,
+      "step": 3780
     },
     {
       "epoch": 33.01,
-      "learning_rate": 2.3979107312440647e-05,
-      "loss": 0.9024,
-      "step": 3990
-    },
-    {
-      "epoch": 33.02,
-      "learning_rate": 2.389996834441279e-05,
-      "loss": 0.3402,
-      "step": 4000
+      "learning_rate": 2.3940607273940608e-05,
+      "loss": 0.5228,
+      "step": 3790
     },
     {
       "epoch": 33.02,
-      "learning_rate": 2.3820829376384935e-05,
-      "loss": 0.5805,
-      "step": 4010
+      "learning_rate": 2.385719052385719e-05,
+      "loss": 0.7662,
+      "step": 3800
     },
     {
       "epoch": 33.02,
-      "eval_accuracy": 0.8202764976958525,
-      "eval_loss": 0.5598491430282593,
-      "eval_runtime": 287.693,
-      "eval_samples_per_second": 0.754,
-      "eval_steps_per_second": 0.191,
-      "step": 4012
+      "eval_accuracy": 0.6405529953917051,
+      "eval_loss": 1.0589454174041748,
+      "eval_runtime": 290.7017,
+      "eval_samples_per_second": 0.746,
+      "eval_steps_per_second": 0.189,
+      "step": 3808
     },
     {
       "epoch": 34.0,
-      "learning_rate": 2.3741690408357077e-05,
-      "loss": 0.6835,
-      "step": 4020
+      "learning_rate": 2.3773773773773775e-05,
+      "loss": 0.5944,
+      "step": 3810
     },
     {
       "epoch": 34.0,
-      "learning_rate": 2.366255144032922e-05,
-      "loss": 0.6466,
-      "step": 4030
+      "learning_rate": 2.3690357023690358e-05,
+      "loss": 0.588,
+      "step": 3820
     },
     {
       "epoch": 34.0,
-      "learning_rate": 2.358341247230136e-05,
-      "loss": 0.3698,
-      "step": 4040
+      "learning_rate": 2.360694027360694e-05,
+      "loss": 0.6972,
+      "step": 3830
     },
     {
-      "epoch": 34.01,
-      "learning_rate": 2.3504273504273504e-05,
-      "loss": 0.4723,
-      "step": 4050
+      "epoch": 34.0,
+      "learning_rate": 2.3523523523523525e-05,
+      "loss": 0.544,
+      "step": 3840
     },
     {
       "epoch": 34.01,
-      "learning_rate": 2.342513453624565e-05,
-      "loss": 0.7078,
-      "step": 4060
+      "learning_rate": 2.344010677344011e-05,
+      "loss": 0.6615,
+      "step": 3850
     },
     {
       "epoch": 34.01,
-      "learning_rate": 2.334599556821779e-05,
-      "loss": 0.6876,
-      "step": 4070
+      "learning_rate": 2.3356690023356692e-05,
+      "loss": 0.5002,
+      "step": 3860
     },
     {
       "epoch": 34.01,
-      "learning_rate": 2.3266856600189934e-05,
-      "loss": 0.4859,
-      "step": 4080
+      "learning_rate": 2.3273273273273275e-05,
+      "loss": 0.5654,
+      "step": 3870
     },
     {
       "epoch": 34.01,
-      "learning_rate": 2.318771763216208e-05,
-      "loss": 0.5394,
-      "step": 4090
+      "learning_rate": 2.3189856523189855e-05,
+      "loss": 0.6728,
+      "step": 3880
     },
     {
       "epoch": 34.01,
-      "learning_rate": 2.310857866413422e-05,
-      "loss": 0.4636,
-      "step": 4100
+      "learning_rate": 2.3106439773106442e-05,
+      "loss": 0.5296,
+      "step": 3890
     },
     {
       "epoch": 34.01,
-      "learning_rate": 2.3029439696106364e-05,
-      "loss": 0.5758,
-      "step": 4110
+      "learning_rate": 2.3023023023023026e-05,
+      "loss": 0.4234,
+      "step": 3900
     },
     {
       "epoch": 34.02,
-      "learning_rate": 2.295030072807851e-05,
-      "loss": 0.4957,
-      "step": 4120
+      "learning_rate": 2.2939606272939606e-05,
+      "loss": 0.5872,
+      "step": 3910
     },
     {
       "epoch": 34.02,
-      "learning_rate": 2.2871161760050648e-05,
-      "loss": 0.7064,
-      "step": 4130
+      "learning_rate": 2.2856189522856192e-05,
+      "loss": 0.7313,
+      "step": 3920
     },
     {
       "epoch": 34.02,
-      "eval_accuracy": 0.7926267281105991,
-      "eval_loss": 0.7340723276138306,
-      "eval_runtime": 254.6894,
-      "eval_samples_per_second": 0.852,
-      "eval_steps_per_second": 0.216,
-      "step": 4130
+      "eval_accuracy": 0.5622119815668203,
+      "eval_loss": 1.5075870752334595,
+      "eval_runtime": 286.0606,
+      "eval_samples_per_second": 0.759,
+      "eval_steps_per_second": 0.192,
+      "step": 3920
     },
     {
       "epoch": 35.0,
-      "learning_rate": 2.2792022792022794e-05,
-      "loss": 0.4694,
-      "step": 4140
+      "learning_rate": 2.2772772772772773e-05,
+      "loss": 0.5481,
+      "step": 3930
     },
     {
       "epoch": 35.0,
-      "learning_rate": 2.2712883823994936e-05,
-      "loss": 0.6956,
-      "step": 4150
+      "learning_rate": 2.2689356022689356e-05,
+      "loss": 0.7286,
+      "step": 3940
     },
     {
       "epoch": 35.0,
-      "learning_rate": 2.2633744855967078e-05,
-      "loss": 0.4924,
-      "step": 4160
-    },
-    {
-      "epoch": 35.01,
-      "learning_rate": 2.2554605887939224e-05,
-      "loss": 0.7678,
-      "step": 4170
+      "learning_rate": 2.2605939272605943e-05,
+      "loss": 0.795,
+      "step": 3950
     },
     {
       "epoch": 35.01,
-      "learning_rate": 2.2475466919911363e-05,
-      "loss": 0.6094,
-      "step": 4180
+      "learning_rate": 2.2522522522522523e-05,
+      "loss": 0.7738,
+      "step": 3960
     },
     {
       "epoch": 35.01,
-      "learning_rate": 2.239632795188351e-05,
-      "loss": 0.3326,
-      "step": 4190
+      "learning_rate": 2.2439105772439106e-05,
+      "loss": 0.8864,
+      "step": 3970
     },
     {
       "epoch": 35.01,
-      "learning_rate": 2.2317188983855654e-05,
-      "loss": 0.623,
-      "step": 4200
+      "learning_rate": 2.235568902235569e-05,
+      "loss": 0.3864,
+      "step": 3980
     },
     {
       "epoch": 35.01,
-      "learning_rate": 2.2238050015827793e-05,
-      "loss": 0.6507,
-      "step": 4210
+      "learning_rate": 2.2272272272272273e-05,
+      "loss": 0.6556,
+      "step": 3990
     },
     {
       "epoch": 35.01,
-      "learning_rate": 2.215891104779994e-05,
-      "loss": 0.5673,
-      "step": 4220
+      "learning_rate": 2.2188855522188857e-05,
+      "loss": 0.9211,
+      "step": 4000
     },
     {
       "epoch": 35.01,
-      "learning_rate": 2.207977207977208e-05,
-      "loss": 0.9691,
-      "step": 4230
+      "learning_rate": 2.210543877210544e-05,
+      "loss": 0.6319,
+      "step": 4010
     },
     {
       "epoch": 35.02,
-      "learning_rate": 2.2000633111744223e-05,
-      "loss": 0.2534,
-      "step": 4240
+      "learning_rate": 2.2022022022022024e-05,
+      "loss": 0.6768,
+      "step": 4020
     },
     {
       "epoch": 35.02,
-      "eval_accuracy": 0.783410138248848,
-      "eval_loss": 0.6685347557067871,
-      "eval_runtime": 253.8074,
-      "eval_samples_per_second": 0.855,
-      "eval_steps_per_second": 0.217,
-      "step": 4248
+      "learning_rate": 2.1938605271938607e-05,
+      "loss": 0.7539,
+      "step": 4030
     },
     {
-      "epoch": 36.0,
-      "learning_rate": 2.192149414371637e-05,
-      "loss": 0.5305,
-      "step": 4250
+      "epoch": 35.02,
+      "eval_accuracy": 0.5898617511520737,
+      "eval_loss": 1.2265080213546753,
+      "eval_runtime": 266.6298,
+      "eval_samples_per_second": 0.814,
+      "eval_steps_per_second": 0.206,
+      "step": 4032
     },
     {
       "epoch": 36.0,
-      "learning_rate": 2.184235517568851e-05,
-      "loss": 0.3181,
-      "step": 4260
+      "learning_rate": 2.1855188521855187e-05,
+      "loss": 0.6197,
+      "step": 4040
     },
     {
       "epoch": 36.0,
-      "learning_rate": 2.1763216207660653e-05,
-      "loss": 0.2655,
-      "step": 4270
+      "learning_rate": 2.1771771771771774e-05,
+      "loss": 0.5972,
+      "step": 4050
     },
     {
       "epoch": 36.0,
-      "learning_rate": 2.1684077239632795e-05,
-      "loss": 0.546,
-      "step": 4280
+      "learning_rate": 2.1688355021688357e-05,
+      "loss": 0.9935,
+      "step": 4060
     },
     {
       "epoch": 36.01,
       "learning_rate": 2.1604938271604937e-05,
-      "loss": 0.5166,
-      "step": 4290
+      "loss": 0.6247,
+      "step": 4070
     },
     {
       "epoch": 36.01,
-      "learning_rate": 2.1525799303577083e-05,
-      "loss": 0.612,
-      "step": 4300
+      "learning_rate": 2.1521521521521524e-05,
+      "loss": 0.6728,
+      "step": 4080
     },
     {
       "epoch": 36.01,
-      "learning_rate": 2.1446660335549225e-05,
-      "loss": 1.0337,
-      "step": 4310
+      "learning_rate": 2.1438104771438104e-05,
+      "loss": 0.8943,
+      "step": 4090
     },
     {
       "epoch": 36.01,
-      "learning_rate": 2.1367521367521368e-05,
-      "loss": 0.4693,
-      "step": 4320
+      "learning_rate": 2.1354688021354688e-05,
+      "loss": 0.6704,
+      "step": 4100
     },
     {
       "epoch": 36.01,
-      "learning_rate": 2.1288382399493513e-05,
-      "loss": 0.535,
-      "step": 4330
+      "learning_rate": 2.1271271271271275e-05,
+      "loss": 0.8664,
+      "step": 4110
     },
     {
       "epoch": 36.01,
-      "learning_rate": 2.1209243431465655e-05,
-      "loss": 0.2098,
-      "step": 4340
+      "learning_rate": 2.1187854521187855e-05,
+      "loss": 0.8486,
+      "step": 4120
     },
     {
       "epoch": 36.01,
-      "learning_rate": 2.1130104463437798e-05,
-      "loss": 0.5005,
-      "step": 4350
+      "learning_rate": 2.1104437771104438e-05,
+      "loss": 0.5425,
+      "step": 4130
     },
     {
       "epoch": 36.02,
-      "learning_rate": 2.105096549540994e-05,
-      "loss": 0.7578,
-      "step": 4360
+      "learning_rate": 2.102102102102102e-05,
+      "loss": 0.571,
+      "step": 4140
     },
     {
       "epoch": 36.02,
-      "eval_accuracy": 0.7603686635944701,
-      "eval_loss": 0.7591729760169983,
-      "eval_runtime": 256.6972,
-      "eval_samples_per_second": 0.845,
-      "eval_steps_per_second": 0.214,
-      "step": 4366
-    },
-    {
-      "epoch": 37.0,
-      "learning_rate": 2.0971826527382085e-05,
-      "loss": 0.7535,
-      "step": 4370
+      "eval_accuracy": 0.6267281105990783,
+      "eval_loss": 1.1598420143127441,
+      "eval_runtime": 288.0962,
+      "eval_samples_per_second": 0.753,
+      "eval_steps_per_second": 0.191,
+      "step": 4144
     },
     {
       "epoch": 37.0,
-      "learning_rate": 2.0892687559354228e-05,
-      "loss": 0.5906,
-      "step": 4380
+      "learning_rate": 2.0937604270937605e-05,
+      "loss": 0.2915,
+      "step": 4150
     },
     {
       "epoch": 37.0,
-      "learning_rate": 2.081354859132637e-05,
-      "loss": 0.5422,
-      "step": 4390
+      "learning_rate": 2.085418752085419e-05,
+      "loss": 0.5187,
+      "step": 4160
     },
     {
       "epoch": 37.0,
-      "learning_rate": 2.0734409623298512e-05,
-      "loss": 0.4581,
-      "step": 4400
+      "learning_rate": 2.0770770770770772e-05,
+      "loss": 0.4894,
+      "step": 4170
     },
     {
       "epoch": 37.01,
-      "learning_rate": 2.0655270655270654e-05,
-      "loss": 0.4551,
-      "step": 4410
+      "learning_rate": 2.0687354020687355e-05,
+      "loss": 0.522,
+      "step": 4180
     },
     {
       "epoch": 37.01,
-      "learning_rate": 2.05761316872428e-05,
-      "loss": 0.528,
-      "step": 4420
+      "learning_rate": 2.060393727060394e-05,
+      "loss": 0.3428,
+      "step": 4190
     },
     {
       "epoch": 37.01,
-      "learning_rate": 2.0496992719214942e-05,
-      "loss": 0.2991,
-      "step": 4430
+      "learning_rate": 2.0520520520520522e-05,
+      "loss": 0.4749,
+      "step": 4200
     },
     {
       "epoch": 37.01,
-      "learning_rate": 2.0417853751187084e-05,
-      "loss": 0.8417,
-      "step": 4440
+      "learning_rate": 2.0437103770437106e-05,
+      "loss": 0.9899,
+      "step": 4210
     },
     {
       "epoch": 37.01,
-      "learning_rate": 2.033871478315923e-05,
-      "loss": 0.924,
-      "step": 4450
+      "learning_rate": 2.0353687020353686e-05,
+      "loss": 0.7875,
+      "step": 4220
     },
     {
       "epoch": 37.01,
-      "learning_rate": 2.0259575815131372e-05,
-      "loss": 0.5173,
-      "step": 4460
+      "learning_rate": 2.0270270270270273e-05,
+      "loss": 0.4509,
+      "step": 4230
     },
     {
       "epoch": 37.01,
-      "learning_rate": 2.0180436847103515e-05,
-      "loss": 0.3303,
-      "step": 4470
+      "learning_rate": 2.0186853520186856e-05,
+      "loss": 0.5523,
+      "step": 4240
     },
     {
       "epoch": 37.02,
-      "learning_rate": 2.010129787907566e-05,
-      "loss": 0.5822,
-      "step": 4480
+      "learning_rate": 2.0103436770103436e-05,
+      "loss": 0.3404,
+      "step": 4250
     },
     {
       "epoch": 37.02,
-      "eval_accuracy": 0.728110599078341,
-      "eval_loss": 0.9471691250801086,
-      "eval_runtime": 259.2305,
-      "eval_samples_per_second": 0.837,
-      "eval_steps_per_second": 0.212,
-      "step": 4484
+      "eval_accuracy": 0.6359447004608295,
+      "eval_loss": 1.0306791067123413,
+      "eval_runtime": 260.5195,
+      "eval_samples_per_second": 0.833,
+      "eval_steps_per_second": 0.211,
+      "step": 4256
     },
     {
       "epoch": 38.0,
-      "learning_rate": 2.00221589110478e-05,
-      "loss": 0.3202,
-      "step": 4490
+      "learning_rate": 2.0020020020020023e-05,
+      "loss": 0.6409,
+      "step": 4260
     },
     {
       "epoch": 38.0,
-      "learning_rate": 1.9943019943019945e-05,
-      "loss": 0.4554,
-      "step": 4500
+      "learning_rate": 1.9936603269936603e-05,
+      "loss": 0.9606,
+      "step": 4270
     },
     {
       "epoch": 38.0,
-      "learning_rate": 1.9863880974992087e-05,
-      "loss": 0.5358,
-      "step": 4510
+      "learning_rate": 1.9853186519853186e-05,
+      "loss": 0.5896,
+      "step": 4280
     },
     {
       "epoch": 38.01,
-      "learning_rate": 1.978474200696423e-05,
-      "loss": 0.3341,
-      "step": 4520
+      "learning_rate": 1.9769769769769773e-05,
+      "loss": 1.0073,
+      "step": 4290
     },
     {
       "epoch": 38.01,
-      "learning_rate": 1.9705603038936375e-05,
-      "loss": 1.1299,
-      "step": 4530
+      "learning_rate": 1.9686353019686353e-05,
+      "loss": 0.7469,
+      "step": 4300
     },
     {
       "epoch": 38.01,
-      "learning_rate": 1.9626464070908514e-05,
-      "loss": 0.4959,
-      "step": 4540
+      "learning_rate": 1.9602936269602937e-05,
+      "loss": 0.9638,
+      "step": 4310
     },
     {
       "epoch": 38.01,
-      "learning_rate": 1.954732510288066e-05,
-      "loss": 0.2519,
-      "step": 4550
+      "learning_rate": 1.951951951951952e-05,
+      "loss": 0.2893,
+      "step": 4320
     },
     {
       "epoch": 38.01,
-      "learning_rate": 1.9468186134852805e-05,
-      "loss": 0.9426,
-      "step": 4560
+      "learning_rate": 1.9436102769436104e-05,
+      "loss": 0.9141,
+      "step": 4330
     },
     {
       "epoch": 38.01,
-      "learning_rate": 1.9389047166824944e-05,
-      "loss": 0.3135,
-      "step": 4570
+      "learning_rate": 1.9352686019352687e-05,
+      "loss": 0.4846,
+      "step": 4340
     },
     {
       "epoch": 38.01,
-      "learning_rate": 1.930990819879709e-05,
-      "loss": 0.528,
-      "step": 4580
-    },
-    {
-      "epoch": 38.02,
-      "learning_rate": 1.923076923076923e-05,
-      "loss": 0.8759,
-      "step": 4590
+      "learning_rate": 1.9269269269269267e-05,
+      "loss": 0.2622,
+      "step": 4350
     },
     {
       "epoch": 38.02,
-      "learning_rate": 1.9151630262741374e-05,
-      "loss": 0.2939,
-      "step": 4600
+      "learning_rate": 1.9185852519185854e-05,
+      "loss": 0.5553,
+      "step": 4360
     },
     {
       "epoch": 38.02,
-      "eval_accuracy": 0.728110599078341,
-      "eval_loss": 0.8887839317321777,
-      "eval_runtime": 260.522,
-      "eval_samples_per_second": 0.833,
-      "eval_steps_per_second": 0.211,
-      "step": 4602
+      "eval_accuracy": 0.7235023041474654,
+      "eval_loss": 0.8180494904518127,
+      "eval_runtime": 302.1979,
+      "eval_samples_per_second": 0.718,
+      "eval_steps_per_second": 0.182,
+      "step": 4368
     },
     {
       "epoch": 39.0,
-      "learning_rate": 1.907249129471352e-05,
-      "loss": 0.7159,
-      "step": 4610
+      "learning_rate": 1.9102435769102438e-05,
+      "loss": 0.7832,
+      "step": 4370
     },
     {
       "epoch": 39.0,
-      "learning_rate": 1.899335232668566e-05,
-      "loss": 0.7495,
-      "step": 4620
+      "learning_rate": 1.9019019019019018e-05,
+      "loss": 0.7238,
+      "step": 4380
     },
     {
       "epoch": 39.0,
-      "learning_rate": 1.8914213358657804e-05,
-      "loss": 0.6435,
-      "step": 4630
+      "learning_rate": 1.8935602268935604e-05,
+      "loss": 0.2628,
+      "step": 4390
     },
     {
-      "epoch": 39.01,
-      "learning_rate": 1.8835074390629946e-05,
-      "loss": 0.5865,
-      "step": 4640
+      "epoch": 39.0,
+      "learning_rate": 1.8852185518852188e-05,
+      "loss": 0.7441,
+      "step": 4400
     },
     {
       "epoch": 39.01,
-      "learning_rate": 1.8755935422602088e-05,
-      "loss": 0.5622,
-      "step": 4650
+      "learning_rate": 1.8768768768768768e-05,
+      "loss": 0.4422,
+      "step": 4410
     },
     {
       "epoch": 39.01,
-      "learning_rate": 1.8676796454574234e-05,
-      "loss": 0.4081,
-      "step": 4660
+      "learning_rate": 1.8685352018685355e-05,
+      "loss": 0.5266,
+      "step": 4420
     },
     {
       "epoch": 39.01,
-      "learning_rate": 1.8597657486546376e-05,
-      "loss": 0.5133,
-      "step": 4670
+      "learning_rate": 1.8601935268601935e-05,
+      "loss": 0.5469,
+      "step": 4430
     },
     {
       "epoch": 39.01,
       "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.5643,
-      "step": 4680
+      "loss": 0.9752,
+      "step": 4440
     },
     {
       "epoch": 39.01,
-      "learning_rate": 1.8439379550490664e-05,
-      "loss": 0.5888,
-      "step": 4690
+      "learning_rate": 1.8435101768435105e-05,
+      "loss": 1.0334,
+      "step": 4450
     },
     {
       "epoch": 39.01,
-      "learning_rate": 1.8360240582462806e-05,
-      "loss": 0.7294,
-      "step": 4700
+      "learning_rate": 1.8351685018351685e-05,
+      "loss": 0.604,
+      "step": 4460
     },
     {
       "epoch": 39.02,
-      "learning_rate": 1.828110161443495e-05,
-      "loss": 0.4845,
-      "step": 4710
+      "learning_rate": 1.826826826826827e-05,
+      "loss": 0.7227,
+      "step": 4470
     },
     {
       "epoch": 39.02,
-      "learning_rate": 1.820196264640709e-05,
-      "loss": 0.4795,
-      "step": 4720
+      "learning_rate": 1.8184851518184852e-05,
+      "loss": 0.8499,
+      "step": 4480
     },
     {
       "epoch": 39.02,
-      "eval_accuracy": 0.663594470046083,
-      "eval_loss": 1.0767979621887207,
-      "eval_runtime": 260.495,
-      "eval_samples_per_second": 0.833,
-      "eval_steps_per_second": 0.211,
-      "step": 4720
+      "eval_accuracy": 0.6497695852534562,
+      "eval_loss": 1.0074414014816284,
+      "eval_runtime": 296.0859,
+      "eval_samples_per_second": 0.733,
+      "eval_steps_per_second": 0.186,
+      "step": 4480
     },
     {
       "epoch": 40.0,
-      "learning_rate": 1.8122823678379236e-05,
-      "loss": 0.3626,
-      "step": 4730
+      "learning_rate": 1.8101434768101436e-05,
+      "loss": 0.3984,
+      "step": 4490
     },
     {
       "epoch": 40.0,
-      "learning_rate": 1.804368471035138e-05,
-      "loss": 0.5294,
-      "step": 4740
+      "learning_rate": 1.801801801801802e-05,
+      "loss": 0.5711,
+      "step": 4500
     },
     {
       "epoch": 40.0,
-      "learning_rate": 1.796454574232352e-05,
-      "loss": 0.8263,
-      "step": 4750
-    },
-    {
-      "epoch": 40.01,
-      "learning_rate": 1.7885406774295663e-05,
-      "loss": 0.6514,
-      "step": 4760
+      "learning_rate": 1.7934601267934602e-05,
+      "loss": 0.5289,
+      "step": 4510
     },
     {
       "epoch": 40.01,
-      "learning_rate": 1.7806267806267805e-05,
-      "loss": 0.5644,
-      "step": 4770
+      "learning_rate": 1.7851184517851186e-05,
+      "loss": 0.7913,
+      "step": 4520
     },
     {
       "epoch": 40.01,
-      "learning_rate": 1.772712883823995e-05,
-      "loss": 0.5152,
-      "step": 4780
+      "learning_rate": 1.776776776776777e-05,
+      "loss": 0.6343,
+      "step": 4530
     },
     {
       "epoch": 40.01,
-      "learning_rate": 1.7647989870212093e-05,
-      "loss": 0.5133,
-      "step": 4790
+      "learning_rate": 1.7684351017684353e-05,
+      "loss": 0.5266,
+      "step": 4540
     },
     {
       "epoch": 40.01,
-      "learning_rate": 1.7568850902184235e-05,
-      "loss": 0.6261,
-      "step": 4800
+      "learning_rate": 1.7600934267600936e-05,
+      "loss": 0.5792,
+      "step": 4550
     },
     {
       "epoch": 40.01,
-      "learning_rate": 1.748971193415638e-05,
-      "loss": 0.8368,
-      "step": 4810
+      "learning_rate": 1.7517517517517516e-05,
+      "loss": 0.8629,
+      "step": 4560
     },
     {
       "epoch": 40.01,
-      "learning_rate": 1.7410572966128523e-05,
-      "loss": 0.5701,
-      "step": 4820
+      "learning_rate": 1.7434100767434103e-05,
+      "loss": 0.7761,
+      "step": 4570
     },
     {
       "epoch": 40.02,
-      "learning_rate": 1.7331433998100665e-05,
-      "loss": 0.4038,
-      "step": 4830
+      "learning_rate": 1.7350684017350687e-05,
+      "loss": 0.7784,
+      "step": 4580
     },
     {
       "epoch": 40.02,
-      "eval_accuracy": 0.8064516129032258,
-      "eval_loss": 0.6451985836029053,
-      "eval_runtime": 261.7369,
-      "eval_samples_per_second": 0.829,
-      "eval_steps_per_second": 0.21,
-      "step": 4838
+      "learning_rate": 1.7267267267267267e-05,
+      "loss": 0.5036,
+      "step": 4590
     },
     {
-      "epoch": 41.0,
-      "learning_rate": 1.725229503007281e-05,
-      "loss": 0.4527,
-      "step": 4840
+      "epoch": 40.02,
+      "eval_accuracy": 0.631336405529954,
+      "eval_loss": 1.1159977912902832,
+      "eval_runtime": 266.794,
+      "eval_samples_per_second": 0.813,
+      "eval_steps_per_second": 0.206,
+      "step": 4592
     },
     {
       "epoch": 41.0,
-      "learning_rate": 1.717315606204495e-05,
-      "loss": 0.9572,
-      "step": 4850
+      "learning_rate": 1.7183850517183853e-05,
+      "loss": 0.8591,
+      "step": 4600
     },
     {
       "epoch": 41.0,
-      "learning_rate": 1.7094017094017095e-05,
-      "loss": 0.5401,
-      "step": 4860
+      "learning_rate": 1.7100433767100434e-05,
+      "loss": 0.4294,
+      "step": 4610
     },
     {
       "epoch": 41.0,
-      "learning_rate": 1.7014878125989238e-05,
-      "loss": 0.3303,
-      "step": 4870
+      "learning_rate": 1.7017017017017017e-05,
+      "loss": 0.5119,
+      "step": 4620
     },
     {
       "epoch": 41.01,
-      "learning_rate": 1.693573915796138e-05,
-      "loss": 0.6231,
-      "step": 4880
+      "learning_rate": 1.6933600266933604e-05,
+      "loss": 0.3545,
+      "step": 4630
     },
     {
       "epoch": 41.01,
-      "learning_rate": 1.6856600189933525e-05,
-      "loss": 0.4519,
-      "step": 4890
+      "learning_rate": 1.6850183516850184e-05,
+      "loss": 0.7615,
+      "step": 4640
     },
     {
       "epoch": 41.01,
-      "learning_rate": 1.6777461221905664e-05,
-      "loss": 0.6937,
-      "step": 4900
+      "learning_rate": 1.6766766766766767e-05,
+      "loss": 0.1783,
+      "step": 4650
     },
     {
       "epoch": 41.01,
-      "learning_rate": 1.669832225387781e-05,
-      "loss": 0.4248,
-      "step": 4910
+      "learning_rate": 1.668335001668335e-05,
+      "loss": 0.6125,
+      "step": 4660
     },
     {
       "epoch": 41.01,
-      "learning_rate": 1.6619183285849956e-05,
-      "loss": 0.6655,
-      "step": 4920
+      "learning_rate": 1.6599933266599934e-05,
+      "loss": 1.0295,
+      "step": 4670
     },
     {
       "epoch": 41.01,
-      "learning_rate": 1.6540044317822094e-05,
-      "loss": 0.5651,
-      "step": 4930
+      "learning_rate": 1.6516516516516518e-05,
+      "loss": 0.6956,
+      "step": 4680
     },
     {
       "epoch": 41.01,
-      "learning_rate": 1.646090534979424e-05,
-      "loss": 0.6075,
-      "step": 4940
+      "learning_rate": 1.6433099766433098e-05,
+      "loss": 0.6383,
+      "step": 4690
     },
     {
       "epoch": 41.02,
-      "learning_rate": 1.6381766381766382e-05,
-      "loss": 0.8347,
-      "step": 4950
+      "learning_rate": 1.6349683016349685e-05,
+      "loss": 0.814,
+      "step": 4700
     },
     {
       "epoch": 41.02,
-      "eval_accuracy": 0.7926267281105991,
-      "eval_loss": 0.7040404677391052,
-      "eval_runtime": 257.2136,
-      "eval_samples_per_second": 0.844,
-      "eval_steps_per_second": 0.214,
-      "step": 4956
-    },
-    {
-      "epoch": 42.0,
-      "learning_rate": 1.6302627413738524e-05,
-      "loss": 0.7086,
-      "step": 4960
+      "eval_accuracy": 0.695852534562212,
+      "eval_loss": 0.903153657913208,
+      "eval_runtime": 298.2312,
+      "eval_samples_per_second": 0.728,
+      "eval_steps_per_second": 0.184,
+      "step": 4704
     },
     {
       "epoch": 42.0,
-      "learning_rate": 1.622348844571067e-05,
-      "loss": 0.6848,
-      "step": 4970
+      "learning_rate": 1.6266266266266268e-05,
+      "loss": 0.5097,
+      "step": 4710
     },
     {
       "epoch": 42.0,
-      "learning_rate": 1.6144349477682812e-05,
-      "loss": 0.718,
-      "step": 4980
+      "learning_rate": 1.6182849516182848e-05,
+      "loss": 0.4471,
+      "step": 4720
     },
     {
       "epoch": 42.0,
-      "learning_rate": 1.6065210509654955e-05,
-      "loss": 0.3821,
-      "step": 4990
+      "learning_rate": 1.6099432766099435e-05,
+      "loss": 0.4364,
+      "step": 4730
     },
     {
       "epoch": 42.01,
-      "learning_rate": 1.5986071541627097e-05,
-      "loss": 0.488,
-      "step": 5000
+      "learning_rate": 1.6016016016016015e-05,
+      "loss": 0.5641,
+      "step": 4740
     },
     {
       "epoch": 42.01,
-      "learning_rate": 1.590693257359924e-05,
-      "loss": 0.3441,
-      "step": 5010
+      "learning_rate": 1.59325992659326e-05,
+      "loss": 0.8213,
+      "step": 4750
     },
     {
       "epoch": 42.01,
-      "learning_rate": 1.5827793605571385e-05,
-      "loss": 0.5323,
-      "step": 5020
+      "learning_rate": 1.5849182515849185e-05,
+      "loss": 0.7006,
+      "step": 4760
     },
     {
       "epoch": 42.01,
-      "learning_rate": 1.5748654637543527e-05,
-      "loss": 0.3723,
-      "step": 5030
+      "learning_rate": 1.5765765765765765e-05,
+      "loss": 0.5187,
+      "step": 4770
     },
     {
       "epoch": 42.01,
-      "learning_rate": 1.566951566951567e-05,
-      "loss": 0.5536,
-      "step": 5040
+      "learning_rate": 1.568234901568235e-05,
+      "loss": 0.3012,
+      "step": 4780
     },
     {
       "epoch": 42.01,
-      "learning_rate": 1.5590376701487815e-05,
-      "loss": 0.2562,
-      "step": 5050
+      "learning_rate": 1.5598932265598936e-05,
+      "loss": 0.7188,
+      "step": 4790
     },
     {
       "epoch": 42.01,
-      "learning_rate": 1.5511237733459957e-05,
-      "loss": 0.4913,
-      "step": 5060
+      "learning_rate": 1.5515515515515516e-05,
+      "loss": 0.5198,
+      "step": 4800
     },
     {
       "epoch": 42.02,
       "learning_rate": 1.54320987654321e-05,
-      "loss": 0.4113,
-      "step": 5070
+      "loss": 0.7293,
+      "step": 4810
     },
     {
       "epoch": 42.02,
-      "eval_accuracy": 0.7373271889400922,
-      "eval_loss": 0.8011646866798401,
-      "eval_runtime": 252.0655,
-      "eval_samples_per_second": 0.861,
-      "eval_steps_per_second": 0.218,
-      "step": 5074
+      "eval_accuracy": 0.728110599078341,
+      "eval_loss": 0.9331218004226685,
+      "eval_runtime": 258.6829,
+      "eval_samples_per_second": 0.839,
+      "eval_steps_per_second": 0.213,
+      "step": 4816
     },
     {
       "epoch": 43.0,
-      "learning_rate": 1.535295979740424e-05,
-      "loss": 0.1942,
-      "step": 5080
+      "learning_rate": 1.5348682015348683e-05,
+      "loss": 0.4944,
+      "step": 4820
     },
     {
       "epoch": 43.0,
-      "learning_rate": 1.5273820829376387e-05,
-      "loss": 0.6847,
-      "step": 5090
+      "learning_rate": 1.5265265265265266e-05,
+      "loss": 0.4447,
+      "step": 4830
     },
     {
       "epoch": 43.0,
-      "learning_rate": 1.5194681861348528e-05,
-      "loss": 0.5127,
-      "step": 5100
+      "learning_rate": 1.5181848515181848e-05,
+      "loss": 0.3132,
+      "step": 4840
     },
     {
       "epoch": 43.01,
-      "learning_rate": 1.5115542893320671e-05,
-      "loss": 0.18,
-      "step": 5110
+      "learning_rate": 1.5098431765098433e-05,
+      "loss": 0.5882,
+      "step": 4850
     },
     {
       "epoch": 43.01,
-      "learning_rate": 1.5036403925292817e-05,
-      "loss": 0.6956,
-      "step": 5120
+      "learning_rate": 1.5015015015015016e-05,
+      "loss": 0.7151,
+      "step": 4860
     },
     {
       "epoch": 43.01,
-      "learning_rate": 1.4957264957264958e-05,
-      "loss": 0.7358,
-      "step": 5130
+      "learning_rate": 1.4931598264931598e-05,
+      "loss": 0.7703,
+      "step": 4870
     },
     {
       "epoch": 43.01,
-      "learning_rate": 1.4878125989237102e-05,
-      "loss": 0.7034,
-      "step": 5140
+      "learning_rate": 1.4848181514848183e-05,
+      "loss": 0.8505,
+      "step": 4880
     },
     {
       "epoch": 43.01,
-      "learning_rate": 1.4798987021209244e-05,
-      "loss": 0.5671,
-      "step": 5150
+      "learning_rate": 1.4764764764764765e-05,
+      "loss": 0.3957,
+      "step": 4890
     },
     {
       "epoch": 43.01,
-      "learning_rate": 1.4719848053181388e-05,
-      "loss": 0.276,
-      "step": 5160
+      "learning_rate": 1.4681348014681348e-05,
+      "loss": 0.5204,
+      "step": 4900
     },
     {
       "epoch": 43.01,
-      "learning_rate": 1.4640709085153532e-05,
-      "loss": 0.5131,
-      "step": 5170
-    },
-    {
-      "epoch": 43.02,
-      "learning_rate": 1.4561570117125672e-05,
-      "loss": 0.6712,
-      "step": 5180
+      "learning_rate": 1.4597931264597934e-05,
+      "loss": 0.4383,
+      "step": 4910
     },
     {
       "epoch": 43.02,
-      "learning_rate": 1.4482431149097816e-05,
-      "loss": 0.3681,
-      "step": 5190
+      "learning_rate": 1.4514514514514515e-05,
+      "loss": 0.4402,
+      "step": 4920
     },
     {
       "epoch": 43.02,
-      "eval_accuracy": 0.7880184331797235,
-      "eval_loss": 0.762200117111206,
-      "eval_runtime": 254.5668,
-      "eval_samples_per_second": 0.852,
-      "eval_steps_per_second": 0.216,
-      "step": 5192
+      "eval_accuracy": 0.5668202764976958,
+      "eval_loss": 1.4190495014190674,
+      "eval_runtime": 275.5299,
+      "eval_samples_per_second": 0.788,
+      "eval_steps_per_second": 0.2,
+      "step": 4928
     },
     {
       "epoch": 44.0,
-      "learning_rate": 1.440329218106996e-05,
-      "loss": 0.4388,
-      "step": 5200
+      "learning_rate": 1.4431097764431097e-05,
+      "loss": 0.2849,
+      "step": 4930
     },
     {
       "epoch": 44.0,
-      "learning_rate": 1.4324153213042102e-05,
-      "loss": 0.4731,
-      "step": 5210
+      "learning_rate": 1.4347681014347682e-05,
+      "loss": 0.5245,
+      "step": 4940
     },
     {
       "epoch": 44.0,
-      "learning_rate": 1.4245014245014246e-05,
-      "loss": 0.4798,
-      "step": 5220
+      "learning_rate": 1.4264264264264266e-05,
+      "loss": 0.4884,
+      "step": 4950
     },
     {
-      "epoch": 44.01,
-      "learning_rate": 1.416587527698639e-05,
-      "loss": 0.3783,
-      "step": 5230
+      "epoch": 44.0,
+      "learning_rate": 1.4180847514180847e-05,
+      "loss": 0.3824,
+      "step": 4960
     },
     {
       "epoch": 44.01,
-      "learning_rate": 1.408673630895853e-05,
-      "loss": 0.2336,
-      "step": 5240
+      "learning_rate": 1.4097430764097431e-05,
+      "loss": 0.5175,
+      "step": 4970
     },
     {
       "epoch": 44.01,
-      "learning_rate": 1.4007597340930676e-05,
-      "loss": 0.3797,
-      "step": 5250
+      "learning_rate": 1.4014014014014016e-05,
+      "loss": 0.5526,
+      "step": 4980
     },
     {
       "epoch": 44.01,
-      "learning_rate": 1.3928458372902817e-05,
-      "loss": 0.4423,
-      "step": 5260
+      "learning_rate": 1.3930597263930598e-05,
+      "loss": 0.4134,
+      "step": 4990
     },
     {
       "epoch": 44.01,
-      "learning_rate": 1.384931940487496e-05,
-      "loss": 0.5598,
-      "step": 5270
+      "learning_rate": 1.384718051384718e-05,
+      "loss": 0.4121,
+      "step": 5000
     },
     {
       "epoch": 44.01,
-      "learning_rate": 1.3770180436847105e-05,
-      "loss": 0.5773,
-      "step": 5280
+      "learning_rate": 1.3763763763763765e-05,
+      "loss": 0.7849,
+      "step": 5010
     },
     {
       "epoch": 44.01,
-      "learning_rate": 1.3691041468819247e-05,
-      "loss": 0.3113,
-      "step": 5290
+      "learning_rate": 1.3680347013680348e-05,
+      "loss": 0.7587,
+      "step": 5020
     },
     {
       "epoch": 44.02,
-      "learning_rate": 1.361190250079139e-05,
-      "loss": 0.2841,
-      "step": 5300
+      "learning_rate": 1.359693026359693e-05,
+      "loss": 0.4583,
+      "step": 5030
     },
     {
       "epoch": 44.02,
-      "learning_rate": 1.3532763532763535e-05,
-      "loss": 1.0092,
-      "step": 5310
+      "learning_rate": 1.3513513513513515e-05,
+      "loss": 0.4625,
+      "step": 5040
     },
     {
       "epoch": 44.02,
-      "eval_accuracy": 0.7880184331797235,
-      "eval_loss": 0.7931644916534424,
-      "eval_runtime": 259.7621,
-      "eval_samples_per_second": 0.835,
-      "eval_steps_per_second": 0.212,
-      "step": 5310
+      "eval_accuracy": 0.7004608294930875,
+      "eval_loss": 1.0268219709396362,
+      "eval_runtime": 299.0881,
+      "eval_samples_per_second": 0.726,
+      "eval_steps_per_second": 0.184,
+      "step": 5040
     },
     {
       "epoch": 45.0,
-      "learning_rate": 1.3453624564735675e-05,
-      "loss": 0.6327,
-      "step": 5320
+      "learning_rate": 1.3430096763430097e-05,
+      "loss": 0.4425,
+      "step": 5050
     },
     {
       "epoch": 45.0,
-      "learning_rate": 1.3374485596707819e-05,
-      "loss": 1.0639,
-      "step": 5330
+      "learning_rate": 1.334668001334668e-05,
+      "loss": 0.6364,
+      "step": 5060
     },
     {
       "epoch": 45.0,
-      "learning_rate": 1.3295346628679963e-05,
-      "loss": 0.2824,
-      "step": 5340
-    },
-    {
-      "epoch": 45.01,
-      "learning_rate": 1.3216207660652105e-05,
-      "loss": 0.3033,
-      "step": 5350
+      "learning_rate": 1.3263263263263265e-05,
+      "loss": 0.386,
+      "step": 5070
     },
     {
       "epoch": 45.01,
-      "learning_rate": 1.313706869262425e-05,
-      "loss": 0.5596,
-      "step": 5360
+      "learning_rate": 1.3179846513179847e-05,
+      "loss": 0.6462,
+      "step": 5080
     },
     {
       "epoch": 45.01,
-      "learning_rate": 1.305792972459639e-05,
-      "loss": 0.4285,
-      "step": 5370
+      "learning_rate": 1.3096429763096429e-05,
+      "loss": 0.8925,
+      "step": 5090
     },
     {
       "epoch": 45.01,
-      "learning_rate": 1.2978790756568535e-05,
-      "loss": 0.7668,
-      "step": 5380
+      "learning_rate": 1.3013013013013014e-05,
+      "loss": 0.6358,
+      "step": 5100
     },
     {
       "epoch": 45.01,
-      "learning_rate": 1.289965178854068e-05,
-      "loss": 0.6117,
-      "step": 5390
+      "learning_rate": 1.2929596262929598e-05,
+      "loss": 0.8605,
+      "step": 5110
     },
     {
       "epoch": 45.01,
-      "learning_rate": 1.282051282051282e-05,
-      "loss": 0.5597,
-      "step": 5400
+      "learning_rate": 1.284617951284618e-05,
+      "loss": 0.9282,
+      "step": 5120
     },
     {
       "epoch": 45.01,
-      "learning_rate": 1.2741373852484964e-05,
-      "loss": 0.2782,
-      "step": 5410
+      "learning_rate": 1.2762762762762764e-05,
+      "loss": 0.7836,
+      "step": 5130
     },
     {
       "epoch": 45.02,
-      "learning_rate": 1.2662234884457108e-05,
-      "loss": 0.321,
-      "step": 5420
+      "learning_rate": 1.2679346012679346e-05,
+      "loss": 0.4875,
+      "step": 5140
     },
     {
       "epoch": 45.02,
-      "eval_accuracy": 0.7373271889400922,
-      "eval_loss": 0.9068748354911804,
-      "eval_runtime": 259.5409,
-      "eval_samples_per_second": 0.836,
-      "eval_steps_per_second": 0.212,
-      "step": 5428
+      "learning_rate": 1.259592926259593e-05,
+      "loss": 0.2266,
+      "step": 5150
     },
     {
-      "epoch": 46.0,
-      "learning_rate": 1.258309591642925e-05,
-      "loss": 0.5204,
-      "step": 5430
+      "epoch": 45.02,
+      "eval_accuracy": 0.6405529953917051,
+      "eval_loss": 1.2808195352554321,
+      "eval_runtime": 257.9463,
+      "eval_samples_per_second": 0.841,
+      "eval_steps_per_second": 0.213,
+      "step": 5152
     },
     {
       "epoch": 46.0,
-      "learning_rate": 1.2503956948401394e-05,
-      "loss": 0.5057,
-      "step": 5440
+      "learning_rate": 1.2512512512512515e-05,
+      "loss": 0.6837,
+      "step": 5160
     },
     {
       "epoch": 46.0,
-      "learning_rate": 1.2424817980373536e-05,
-      "loss": 0.5511,
-      "step": 5450
+      "learning_rate": 1.2429095762429097e-05,
+      "loss": 0.3609,
+      "step": 5170
     },
     {
       "epoch": 46.0,
       "learning_rate": 1.2345679012345678e-05,
-      "loss": 0.6262,
-      "step": 5460
+      "loss": 0.7921,
+      "step": 5180
     },
     {
       "epoch": 46.01,
-      "learning_rate": 1.2266540044317822e-05,
-      "loss": 0.3925,
-      "step": 5470
+      "learning_rate": 1.2262262262262263e-05,
+      "loss": 1.2116,
+      "step": 5190
     },
     {
       "epoch": 46.01,
-      "learning_rate": 1.2187401076289966e-05,
-      "loss": 0.5744,
-      "step": 5480
+      "learning_rate": 1.2178845512178847e-05,
+      "loss": 0.6612,
+      "step": 5200
     },
     {
       "epoch": 46.01,
-      "learning_rate": 1.2108262108262108e-05,
-      "loss": 0.4216,
-      "step": 5490
+      "learning_rate": 1.2095428762095429e-05,
+      "loss": 0.2317,
+      "step": 5210
     },
     {
       "epoch": 46.01,
-      "learning_rate": 1.2029123140234252e-05,
-      "loss": 0.6228,
-      "step": 5500
+      "learning_rate": 1.2012012012012012e-05,
+      "loss": 0.4213,
+      "step": 5220
     },
     {
       "epoch": 46.01,
-      "learning_rate": 1.1949984172206395e-05,
-      "loss": 0.3394,
-      "step": 5510
+      "learning_rate": 1.1928595261928596e-05,
+      "loss": 0.8834,
+      "step": 5230
     },
     {
       "epoch": 46.01,
-      "learning_rate": 1.1870845204178538e-05,
-      "loss": 0.7777,
-      "step": 5520
+      "learning_rate": 1.1845178511845179e-05,
+      "loss": 0.8265,
+      "step": 5240
     },
     {
       "epoch": 46.01,
-      "learning_rate": 1.179170623615068e-05,
-      "loss": 0.3025,
-      "step": 5530
+      "learning_rate": 1.1761761761761762e-05,
+      "loss": 0.1315,
+      "step": 5250
     },
     {
       "epoch": 46.02,
-      "learning_rate": 1.1712567268122825e-05,
-      "loss": 0.399,
-      "step": 5540
+      "learning_rate": 1.1678345011678346e-05,
+      "loss": 0.7424,
+      "step": 5260
     },
     {
       "epoch": 46.02,
-      "eval_accuracy": 0.8110599078341014,
-      "eval_loss": 0.6439275741577148,
-      "eval_runtime": 263.8811,
-      "eval_samples_per_second": 0.822,
-      "eval_steps_per_second": 0.208,
-      "step": 5546
-    },
-    {
-      "epoch": 47.0,
-      "learning_rate": 1.1633428300094967e-05,
-      "loss": 0.6139,
-      "step": 5550
+      "eval_accuracy": 0.6497695852534562,
+      "eval_loss": 1.1821485757827759,
+      "eval_runtime": 277.4384,
+      "eval_samples_per_second": 0.782,
+      "eval_steps_per_second": 0.198,
+      "step": 5264
     },
     {
       "epoch": 47.0,
-      "learning_rate": 1.155428933206711e-05,
-      "loss": 0.4756,
-      "step": 5560
+      "learning_rate": 1.1594928261594928e-05,
+      "loss": 0.7674,
+      "step": 5270
     },
     {
       "epoch": 47.0,
-      "learning_rate": 1.1475150364039255e-05,
-      "loss": 0.4553,
-      "step": 5570
+      "learning_rate": 1.1511511511511513e-05,
+      "loss": 0.8093,
+      "step": 5280
     },
     {
       "epoch": 47.0,
-      "learning_rate": 1.1396011396011397e-05,
-      "loss": 0.4806,
-      "step": 5580
+      "learning_rate": 1.1428094761428096e-05,
+      "loss": 0.3749,
+      "step": 5290
     },
     {
       "epoch": 47.01,
-      "learning_rate": 1.1316872427983539e-05,
-      "loss": 0.6994,
-      "step": 5590
+      "learning_rate": 1.1344678011344678e-05,
+      "loss": 0.572,
+      "step": 5300
     },
     {
       "epoch": 47.01,
-      "learning_rate": 1.1237733459955681e-05,
-      "loss": 0.6582,
-      "step": 5600
+      "learning_rate": 1.1261261261261261e-05,
+      "loss": 0.4389,
+      "step": 5310
     },
     {
       "epoch": 47.01,
-      "learning_rate": 1.1158594491927827e-05,
-      "loss": 0.241,
-      "step": 5610
+      "learning_rate": 1.1177844511177845e-05,
+      "loss": 0.421,
+      "step": 5320
     },
     {
       "epoch": 47.01,
-      "learning_rate": 1.107945552389997e-05,
-      "loss": 0.6311,
-      "step": 5620
+      "learning_rate": 1.1094427761094428e-05,
+      "loss": 0.7683,
+      "step": 5330
     },
     {
       "epoch": 47.01,
-      "learning_rate": 1.1000316555872111e-05,
-      "loss": 0.8412,
-      "step": 5630
+      "learning_rate": 1.1011011011011012e-05,
+      "loss": 0.333,
+      "step": 5340
     },
     {
       "epoch": 47.01,
-      "learning_rate": 1.0921177587844255e-05,
-      "loss": 0.2699,
-      "step": 5640
+      "learning_rate": 1.0927594260927594e-05,
+      "loss": 0.8005,
+      "step": 5350
     },
     {
       "epoch": 47.01,
-      "learning_rate": 1.0842038619816398e-05,
-      "loss": 0.3901,
-      "step": 5650
+      "learning_rate": 1.0844177510844179e-05,
+      "loss": 0.5426,
+      "step": 5360
     },
     {
       "epoch": 47.02,
-      "learning_rate": 1.0762899651788542e-05,
-      "loss": 0.3699,
-      "step": 5660
+      "learning_rate": 1.0760760760760762e-05,
+      "loss": 0.4852,
+      "step": 5370
     },
     {
       "epoch": 47.02,
-      "eval_accuracy": 0.7695852534562212,
-      "eval_loss": 0.7740164399147034,
-      "eval_runtime": 266.667,
-      "eval_samples_per_second": 0.814,
-      "eval_steps_per_second": 0.206,
-      "step": 5664
+      "eval_accuracy": 0.6589861751152074,
+      "eval_loss": 1.2433668375015259,
+      "eval_runtime": 303.2594,
+      "eval_samples_per_second": 0.716,
+      "eval_steps_per_second": 0.181,
+      "step": 5376
     },
     {
       "epoch": 48.0,
-      "learning_rate": 1.0683760683760684e-05,
-      "loss": 0.8451,
-      "step": 5670
+      "learning_rate": 1.0677344010677344e-05,
+      "loss": 0.9861,
+      "step": 5380
     },
     {
       "epoch": 48.0,
-      "learning_rate": 1.0604621715732828e-05,
-      "loss": 0.5939,
-      "step": 5680
+      "learning_rate": 1.0593927260593927e-05,
+      "loss": 0.4832,
+      "step": 5390
     },
     {
       "epoch": 48.0,
-      "learning_rate": 1.052548274770497e-05,
-      "loss": 0.7695,
-      "step": 5690
+      "learning_rate": 1.051051051051051e-05,
+      "loss": 0.3998,
+      "step": 5400
     },
     {
       "epoch": 48.01,
-      "learning_rate": 1.0446343779677114e-05,
-      "loss": 0.5514,
-      "step": 5700
+      "learning_rate": 1.0427093760427094e-05,
+      "loss": 0.599,
+      "step": 5410
     },
     {
       "epoch": 48.01,
-      "learning_rate": 1.0367204811649256e-05,
-      "loss": 0.4196,
-      "step": 5710
+      "learning_rate": 1.0343677010343678e-05,
+      "loss": 0.6899,
+      "step": 5420
     },
     {
       "epoch": 48.01,
-      "learning_rate": 1.02880658436214e-05,
-      "loss": 0.2501,
-      "step": 5720
+      "learning_rate": 1.0260260260260261e-05,
+      "loss": 0.618,
+      "step": 5430
     },
     {
       "epoch": 48.01,
-      "learning_rate": 1.0208926875593542e-05,
-      "loss": 0.6282,
-      "step": 5730
+      "learning_rate": 1.0176843510176843e-05,
+      "loss": 0.5096,
+      "step": 5440
     },
     {
       "epoch": 48.01,
-      "learning_rate": 1.0129787907565686e-05,
-      "loss": 0.5634,
-      "step": 5740
+      "learning_rate": 1.0093426760093428e-05,
+      "loss": 0.379,
+      "step": 5450
     },
     {
       "epoch": 48.01,
-      "learning_rate": 1.005064893953783e-05,
-      "loss": 0.3623,
-      "step": 5750
+      "learning_rate": 1.0010010010010011e-05,
+      "loss": 0.5432,
+      "step": 5460
     },
     {
       "epoch": 48.01,
-      "learning_rate": 9.971509971509972e-06,
-      "loss": 0.4048,
-      "step": 5760
-    },
-    {
-      "epoch": 48.02,
-      "learning_rate": 9.892371003482115e-06,
-      "loss": 0.5855,
-      "step": 5770
+      "learning_rate": 9.926593259926593e-06,
+      "loss": 0.5489,
+      "step": 5470
     },
     {
       "epoch": 48.02,
-      "learning_rate": 9.813232035454257e-06,
-      "loss": 0.4297,
-      "step": 5780
+      "learning_rate": 9.843176509843177e-06,
+      "loss": 0.523,
+      "step": 5480
     },
     {
       "epoch": 48.02,
-      "eval_accuracy": 0.8248847926267281,
-      "eval_loss": 0.6810868978500366,
-      "eval_runtime": 261.851,
-      "eval_samples_per_second": 0.829,
-      "eval_steps_per_second": 0.21,
-      "step": 5782
+      "eval_accuracy": 0.6267281105990783,
+      "eval_loss": 1.2123322486877441,
+      "eval_runtime": 258.3979,
+      "eval_samples_per_second": 0.84,
+      "eval_steps_per_second": 0.213,
+      "step": 5488
     },
     {
       "epoch": 49.0,
-      "learning_rate": 9.734093067426402e-06,
-      "loss": 0.3992,
-      "step": 5790
+      "learning_rate": 9.75975975975976e-06,
+      "loss": 0.2441,
+      "step": 5490
     },
     {
       "epoch": 49.0,
-      "learning_rate": 9.654954099398545e-06,
-      "loss": 1.0214,
-      "step": 5800
+      "learning_rate": 9.676343009676344e-06,
+      "loss": 0.6196,
+      "step": 5500
     },
     {
       "epoch": 49.0,
-      "learning_rate": 9.575815131370687e-06,
-      "loss": 0.4853,
-      "step": 5810
+      "learning_rate": 9.592926259592927e-06,
+      "loss": 0.3021,
+      "step": 5510
     },
     {
-      "epoch": 49.01,
-      "learning_rate": 9.49667616334283e-06,
-      "loss": 0.2907,
-      "step": 5820
+      "epoch": 49.0,
+      "learning_rate": 9.509509509509509e-06,
+      "loss": 0.73,
+      "step": 5520
     },
     {
       "epoch": 49.01,
-      "learning_rate": 9.417537195314973e-06,
-      "loss": 0.4338,
-      "step": 5830
+      "learning_rate": 9.426092759426094e-06,
+      "loss": 0.5437,
+      "step": 5530
     },
     {
       "epoch": 49.01,
-      "learning_rate": 9.338398227287117e-06,
-      "loss": 0.3648,
-      "step": 5840
+      "learning_rate": 9.342676009342677e-06,
+      "loss": 0.3555,
+      "step": 5540
     },
     {
       "epoch": 49.01,
       "learning_rate": 9.259259259259259e-06,
-      "loss": 0.8034,
-      "step": 5850
+      "loss": 0.5324,
+      "step": 5550
     },
     {
       "epoch": 49.01,
-      "learning_rate": 9.180120291231403e-06,
-      "loss": 0.7394,
-      "step": 5860
+      "learning_rate": 9.175842509175843e-06,
+      "loss": 0.3838,
+      "step": 5560
     },
     {
       "epoch": 49.01,
-      "learning_rate": 9.100981323203545e-06,
-      "loss": 0.494,
-      "step": 5870
+      "learning_rate": 9.092425759092426e-06,
+      "loss": 0.5131,
+      "step": 5570
     },
     {
       "epoch": 49.01,
-      "learning_rate": 9.02184235517569e-06,
-      "loss": 0.9404,
-      "step": 5880
+      "learning_rate": 9.00900900900901e-06,
+      "loss": 0.5087,
+      "step": 5580
     },
     {
       "epoch": 49.02,
-      "learning_rate": 8.942703387147831e-06,
-      "loss": 0.4178,
-      "step": 5890
+      "learning_rate": 8.925592258925593e-06,
+      "loss": 0.5122,
+      "step": 5590
     },
     {
       "epoch": 49.02,
-      "learning_rate": 8.863564419119975e-06,
-      "loss": 0.2783,
-      "step": 5900
+      "learning_rate": 8.842175508842176e-06,
+      "loss": 0.8344,
+      "step": 5600
     },
     {
       "epoch": 49.02,
-      "eval_accuracy": 0.8525345622119815,
-      "eval_loss": 0.586846113204956,
-      "eval_runtime": 260.0845,
-      "eval_samples_per_second": 0.834,
-      "eval_steps_per_second": 0.211,
-      "step": 5900
+      "eval_accuracy": 0.663594470046083,
+      "eval_loss": 1.1888865232467651,
+      "eval_runtime": 295.3851,
+      "eval_samples_per_second": 0.735,
+      "eval_steps_per_second": 0.186,
+      "step": 5600
     },
     {
       "epoch": 50.0,
-      "learning_rate": 8.784425451092118e-06,
-      "loss": 0.4036,
-      "step": 5910
+      "learning_rate": 8.758758758758758e-06,
+      "loss": 0.6569,
+      "step": 5610
     },
     {
       "epoch": 50.0,
-      "learning_rate": 8.705286483064262e-06,
-      "loss": 0.6644,
-      "step": 5920
+      "learning_rate": 8.675342008675343e-06,
+      "loss": 0.4491,
+      "step": 5620
     },
     {
       "epoch": 50.0,
-      "learning_rate": 8.626147515036405e-06,
-      "loss": 0.2605,
-      "step": 5930
-    },
-    {
-      "epoch": 50.01,
-      "learning_rate": 8.547008547008548e-06,
-      "loss": 0.6677,
-      "step": 5940
+      "learning_rate": 8.591925258591927e-06,
+      "loss": 0.6111,
+      "step": 5630
     },
     {
       "epoch": 50.01,
-      "learning_rate": 8.46786957898069e-06,
-      "loss": 0.4203,
-      "step": 5950
+      "learning_rate": 8.508508508508508e-06,
+      "loss": 0.2911,
+      "step": 5640
     },
     {
       "epoch": 50.01,
-      "learning_rate": 8.388730610952832e-06,
-      "loss": 0.286,
-      "step": 5960
+      "learning_rate": 8.425091758425092e-06,
+      "loss": 0.6463,
+      "step": 5650
     },
     {
       "epoch": 50.01,
-      "learning_rate": 8.309591642924978e-06,
-      "loss": 0.5586,
-      "step": 5970
+      "learning_rate": 8.341675008341675e-06,
+      "loss": 0.3108,
+      "step": 5660
     },
     {
       "epoch": 50.01,
-      "learning_rate": 8.23045267489712e-06,
-      "loss": 0.5642,
-      "step": 5980
+      "learning_rate": 8.258258258258259e-06,
+      "loss": 0.4594,
+      "step": 5670
     },
     {
       "epoch": 50.01,
-      "learning_rate": 8.151313706869262e-06,
-      "loss": 0.583,
-      "step": 5990
+      "learning_rate": 8.174841508174842e-06,
+      "loss": 0.4389,
+      "step": 5680
     },
     {
       "epoch": 50.01,
-      "learning_rate": 8.072174738841406e-06,
-      "loss": 0.2579,
-      "step": 6000
+      "learning_rate": 8.091424758091424e-06,
+      "loss": 0.4616,
+      "step": 5690
     },
     {
       "epoch": 50.02,
-      "learning_rate": 7.993035770813548e-06,
-      "loss": 0.4946,
-      "step": 6010
+      "learning_rate": 8.008008008008007e-06,
+      "loss": 0.2669,
+      "step": 5700
     },
     {
       "epoch": 50.02,
-      "eval_accuracy": 0.7926267281105991,
-      "eval_loss": 0.673189103603363,
-      "eval_runtime": 263.3409,
-      "eval_samples_per_second": 0.824,
-      "eval_steps_per_second": 0.209,
-      "step": 6018
+      "learning_rate": 7.924591257924593e-06,
+      "loss": 0.6648,
+      "step": 5710
     },
     {
-      "epoch": 51.0,
-      "learning_rate": 7.913896802785692e-06,
-      "loss": 0.2915,
-      "step": 6020
+      "epoch": 50.02,
+      "eval_accuracy": 0.6405529953917051,
+      "eval_loss": 1.2328165769577026,
+      "eval_runtime": 294.1867,
+      "eval_samples_per_second": 0.738,
+      "eval_steps_per_second": 0.187,
+      "step": 5712
     },
     {
       "epoch": 51.0,
-      "learning_rate": 7.834757834757835e-06,
-      "loss": 0.4099,
-      "step": 6030
+      "learning_rate": 7.841174507841174e-06,
+      "loss": 0.5881,
+      "step": 5720
     },
     {
       "epoch": 51.0,
-      "learning_rate": 7.755618866729978e-06,
-      "loss": 0.3578,
-      "step": 6040
+      "learning_rate": 7.757757757757758e-06,
+      "loss": 0.658,
+      "step": 5730
     },
     {
       "epoch": 51.0,
-      "learning_rate": 7.67647989870212e-06,
-      "loss": 0.6813,
-      "step": 6050
+      "learning_rate": 7.674341007674341e-06,
+      "loss": 0.473,
+      "step": 5740
     },
     {
       "epoch": 51.01,
-      "learning_rate": 7.597340930674264e-06,
-      "loss": 0.7868,
-      "step": 6060
+      "learning_rate": 7.590924257590924e-06,
+      "loss": 0.561,
+      "step": 5750
     },
     {
       "epoch": 51.01,
-      "learning_rate": 7.5182019626464085e-06,
-      "loss": 0.3421,
-      "step": 6070
+      "learning_rate": 7.507507507507508e-06,
+      "loss": 0.3124,
+      "step": 5760
     },
     {
       "epoch": 51.01,
-      "learning_rate": 7.439062994618551e-06,
-      "loss": 0.3972,
-      "step": 6080
+      "learning_rate": 7.424090757424092e-06,
+      "loss": 0.6171,
+      "step": 5770
     },
     {
       "epoch": 51.01,
-      "learning_rate": 7.359924026590694e-06,
-      "loss": 0.4223,
-      "step": 6090
+      "learning_rate": 7.340674007340674e-06,
+      "loss": 0.2376,
+      "step": 5780
     },
     {
       "epoch": 51.01,
-      "learning_rate": 7.280785058562836e-06,
-      "loss": 0.4222,
-      "step": 6100
+      "learning_rate": 7.257257257257258e-06,
+      "loss": 0.5859,
+      "step": 5790
     },
     {
       "epoch": 51.01,
-      "learning_rate": 7.20164609053498e-06,
-      "loss": 0.5646,
-      "step": 6110
+      "learning_rate": 7.173840507173841e-06,
+      "loss": 0.3655,
+      "step": 5800
     },
     {
       "epoch": 51.01,
-      "learning_rate": 7.122507122507123e-06,
-      "loss": 0.5948,
-      "step": 6120
+      "learning_rate": 7.090423757090424e-06,
+      "loss": 0.5225,
+      "step": 5810
     },
     {
       "epoch": 51.02,
-      "learning_rate": 7.043368154479265e-06,
-      "loss": 0.3058,
-      "step": 6130
+      "learning_rate": 7.007007007007008e-06,
+      "loss": 0.6929,
+      "step": 5820
     },
     {
       "epoch": 51.02,
-      "eval_accuracy": 0.8341013824884793,
-      "eval_loss": 0.551148533821106,
-      "eval_runtime": 267.39,
-      "eval_samples_per_second": 0.812,
-      "eval_steps_per_second": 0.206,
-      "step": 6136
-    },
-    {
-      "epoch": 52.0,
-      "learning_rate": 6.964229186451408e-06,
-      "loss": 0.4489,
-      "step": 6140
+      "eval_accuracy": 0.6129032258064516,
+      "eval_loss": 1.3269143104553223,
+      "eval_runtime": 258.5914,
+      "eval_samples_per_second": 0.839,
+      "eval_steps_per_second": 0.213,
+      "step": 5824
     },
     {
       "epoch": 52.0,
-      "learning_rate": 6.885090218423552e-06,
-      "loss": 0.6678,
-      "step": 6150
+      "learning_rate": 6.92359025692359e-06,
+      "loss": 0.2531,
+      "step": 5830
     },
     {
       "epoch": 52.0,
-      "learning_rate": 6.805951250395695e-06,
-      "loss": 0.4425,
-      "step": 6160
+      "learning_rate": 6.840173506840174e-06,
+      "loss": 0.6214,
+      "step": 5840
     },
     {
       "epoch": 52.0,
-      "learning_rate": 6.726812282367838e-06,
-      "loss": 0.4,
-      "step": 6170
+      "learning_rate": 6.7567567567567575e-06,
+      "loss": 0.6634,
+      "step": 5850
     },
     {
       "epoch": 52.01,
-      "learning_rate": 6.6476733143399815e-06,
-      "loss": 0.6218,
-      "step": 6180
+      "learning_rate": 6.67334000667334e-06,
+      "loss": 0.5057,
+      "step": 5860
     },
     {
       "epoch": 52.01,
-      "learning_rate": 6.568534346312125e-06,
-      "loss": 0.4908,
-      "step": 6190
+      "learning_rate": 6.589923256589924e-06,
+      "loss": 0.4402,
+      "step": 5870
     },
     {
       "epoch": 52.01,
-      "learning_rate": 6.489395378284268e-06,
-      "loss": 0.4602,
-      "step": 6200
+      "learning_rate": 6.506506506506507e-06,
+      "loss": 0.9344,
+      "step": 5880
     },
     {
       "epoch": 52.01,
-      "learning_rate": 6.41025641025641e-06,
-      "loss": 0.3036,
-      "step": 6210
+      "learning_rate": 6.42308975642309e-06,
+      "loss": 0.1245,
+      "step": 5890
     },
     {
       "epoch": 52.01,
-      "learning_rate": 6.331117442228554e-06,
-      "loss": 0.229,
-      "step": 6220
+      "learning_rate": 6.339673006339673e-06,
+      "loss": 0.568,
+      "step": 5900
     },
     {
       "epoch": 52.01,
-      "learning_rate": 6.251978474200697e-06,
-      "loss": 0.5506,
-      "step": 6230
+      "learning_rate": 6.256256256256257e-06,
+      "loss": 0.4524,
+      "step": 5910
     },
     {
       "epoch": 52.01,
       "learning_rate": 6.172839506172839e-06,
-      "loss": 0.201,
-      "step": 6240
+      "loss": 0.4983,
+      "step": 5920
     },
     {
       "epoch": 52.02,
-      "learning_rate": 6.093700538144983e-06,
-      "loss": 0.1286,
-      "step": 6250
+      "learning_rate": 6.0894227560894234e-06,
+      "loss": 0.4253,
+      "step": 5930
     },
     {
       "epoch": 52.02,
-      "eval_accuracy": 0.8294930875576036,
-      "eval_loss": 0.5877251029014587,
-      "eval_runtime": 267.204,
-      "eval_samples_per_second": 0.812,
-      "eval_steps_per_second": 0.206,
-      "step": 6254
+      "eval_accuracy": 0.6820276497695853,
+      "eval_loss": 1.1884889602661133,
+      "eval_runtime": 278.2143,
+      "eval_samples_per_second": 0.78,
+      "eval_steps_per_second": 0.198,
+      "step": 5936
     },
     {
       "epoch": 53.0,
-      "learning_rate": 6.014561570117126e-06,
-      "loss": 0.5248,
-      "step": 6260
+      "learning_rate": 6.006006006006006e-06,
+      "loss": 0.5096,
+      "step": 5940
     },
     {
       "epoch": 53.0,
-      "learning_rate": 5.935422602089269e-06,
-      "loss": 0.5,
-      "step": 6270
+      "learning_rate": 5.9225892559225895e-06,
+      "loss": 0.6105,
+      "step": 5950
     },
     {
       "epoch": 53.0,
-      "learning_rate": 5.856283634061412e-06,
-      "loss": 0.484,
-      "step": 6280
+      "learning_rate": 5.839172505839173e-06,
+      "loss": 1.0454,
+      "step": 5960
     },
     {
       "epoch": 53.01,
-      "learning_rate": 5.777144666033555e-06,
-      "loss": 0.4989,
-      "step": 6290
+      "learning_rate": 5.755755755755756e-06,
+      "loss": 0.4144,
+      "step": 5970
     },
     {
       "epoch": 53.01,
-      "learning_rate": 5.6980056980056985e-06,
-      "loss": 0.71,
-      "step": 6300
+      "learning_rate": 5.672339005672339e-06,
+      "loss": 0.8673,
+      "step": 5980
     },
     {
       "epoch": 53.01,
-      "learning_rate": 5.618866729977841e-06,
-      "loss": 0.8,
-      "step": 6310
+      "learning_rate": 5.5889222555889224e-06,
+      "loss": 0.4563,
+      "step": 5990
     },
     {
       "epoch": 53.01,
-      "learning_rate": 5.539727761949985e-06,
-      "loss": 0.5235,
-      "step": 6320
+      "learning_rate": 5.505505505505506e-06,
+      "loss": 0.5437,
+      "step": 6000
     },
     {
       "epoch": 53.01,
-      "learning_rate": 5.460588793922128e-06,
-      "loss": 0.9647,
-      "step": 6330
+      "learning_rate": 5.422088755422089e-06,
+      "loss": 0.4049,
+      "step": 6010
     },
     {
       "epoch": 53.01,
-      "learning_rate": 5.381449825894271e-06,
-      "loss": 0.3382,
-      "step": 6340
+      "learning_rate": 5.338672005338672e-06,
+      "loss": 0.5225,
+      "step": 6020
     },
     {
       "epoch": 53.01,
-      "learning_rate": 5.302310857866414e-06,
-      "loss": 0.4008,
-      "step": 6350
-    },
-    {
-      "epoch": 53.02,
-      "learning_rate": 5.223171889838557e-06,
-      "loss": 0.5169,
-      "step": 6360
+      "learning_rate": 5.255255255255255e-06,
+      "loss": 0.4383,
+      "step": 6030
     },
     {
       "epoch": 53.02,
-      "learning_rate": 5.1440329218107e-06,
-      "loss": 0.2013,
-      "step": 6370
+      "learning_rate": 5.171838505171839e-06,
+      "loss": 0.7003,
+      "step": 6040
     },
     {
       "epoch": 53.02,
-      "eval_accuracy": 0.815668202764977,
-      "eval_loss": 0.6507552266120911,
-      "eval_runtime": 274.5228,
-      "eval_samples_per_second": 0.79,
-      "eval_steps_per_second": 0.2,
-      "step": 6372
+      "eval_accuracy": 0.7004608294930875,
+      "eval_loss": 1.1521508693695068,
+      "eval_runtime": 293.3646,
+      "eval_samples_per_second": 0.74,
+      "eval_steps_per_second": 0.187,
+      "step": 6048
     },
     {
       "epoch": 54.0,
-      "learning_rate": 5.064893953782843e-06,
-      "loss": 1.0097,
-      "step": 6380
+      "learning_rate": 5.0884217550884214e-06,
+      "loss": 0.6495,
+      "step": 6050
     },
     {
       "epoch": 54.0,
-      "learning_rate": 4.985754985754986e-06,
-      "loss": 0.2331,
-      "step": 6390
+      "learning_rate": 5.005005005005006e-06,
+      "loss": 0.5755,
+      "step": 6060
     },
     {
       "epoch": 54.0,
-      "learning_rate": 4.906616017727128e-06,
-      "loss": 0.388,
-      "step": 6400
+      "learning_rate": 4.921588254921588e-06,
+      "loss": 0.5131,
+      "step": 6070
     },
     {
-      "epoch": 54.01,
-      "learning_rate": 4.827477049699272e-06,
-      "loss": 0.4358,
-      "step": 6410
+      "epoch": 54.0,
+      "learning_rate": 4.838171504838172e-06,
+      "loss": 0.5332,
+      "step": 6080
     },
     {
       "epoch": 54.01,
-      "learning_rate": 4.748338081671415e-06,
-      "loss": 0.4098,
-      "step": 6420
+      "learning_rate": 4.754754754754754e-06,
+      "loss": 0.2762,
+      "step": 6090
     },
     {
       "epoch": 54.01,
-      "learning_rate": 4.6691991136435585e-06,
-      "loss": 0.3491,
-      "step": 6430
+      "learning_rate": 4.671338004671339e-06,
+      "loss": 0.383,
+      "step": 6100
     },
     {
       "epoch": 54.01,
-      "learning_rate": 4.5900601456157015e-06,
-      "loss": 0.2839,
-      "step": 6440
+      "learning_rate": 4.587921254587921e-06,
+      "loss": 0.6731,
+      "step": 6110
     },
     {
       "epoch": 54.01,
-      "learning_rate": 4.510921177587845e-06,
-      "loss": 0.1628,
-      "step": 6450
+      "learning_rate": 4.504504504504505e-06,
+      "loss": 0.4703,
+      "step": 6120
     },
     {
       "epoch": 54.01,
-      "learning_rate": 4.431782209559988e-06,
-      "loss": 1.0689,
-      "step": 6460
+      "learning_rate": 4.421087754421088e-06,
+      "loss": 0.3767,
+      "step": 6130
     },
     {
       "epoch": 54.01,
-      "learning_rate": 4.352643241532131e-06,
-      "loss": 0.6193,
-      "step": 6470
+      "learning_rate": 4.337671004337672e-06,
+      "loss": 0.5669,
+      "step": 6140
     },
     {
       "epoch": 54.02,
-      "learning_rate": 4.273504273504274e-06,
-      "loss": 0.7058,
-      "step": 6480
+      "learning_rate": 4.254254254254254e-06,
+      "loss": 0.4277,
+      "step": 6150
     },
     {
       "epoch": 54.02,
-      "learning_rate": 4.194365305476416e-06,
-      "loss": 0.2027,
-      "step": 6490
+      "learning_rate": 4.170837504170838e-06,
+      "loss": 0.4105,
+      "step": 6160
     },
     {
       "epoch": 54.02,
-      "eval_accuracy": 0.815668202764977,
-      "eval_loss": 0.6629670858383179,
-      "eval_runtime": 274.2142,
-      "eval_samples_per_second": 0.791,
-      "eval_steps_per_second": 0.201,
-      "step": 6490
+      "eval_accuracy": 0.7373271889400922,
+      "eval_loss": 1.0037211179733276,
+      "eval_runtime": 262.5907,
+      "eval_samples_per_second": 0.826,
+      "eval_steps_per_second": 0.209,
+      "step": 6160
     },
     {
       "epoch": 55.0,
-      "learning_rate": 4.11522633744856e-06,
-      "loss": 0.4227,
-      "step": 6500
+      "learning_rate": 4.087420754087421e-06,
+      "loss": 0.2836,
+      "step": 6170
     },
     {
       "epoch": 55.0,
-      "learning_rate": 4.036087369420703e-06,
-      "loss": 0.8382,
-      "step": 6510
+      "learning_rate": 4.004004004004004e-06,
+      "loss": 0.6498,
+      "step": 6180
     },
     {
       "epoch": 55.0,
-      "learning_rate": 3.956948401392846e-06,
-      "loss": 0.5188,
-      "step": 6520
-    },
-    {
-      "epoch": 55.01,
-      "learning_rate": 3.877809433364989e-06,
-      "loss": 0.4906,
-      "step": 6530
+      "learning_rate": 3.920587253920587e-06,
+      "loss": 0.5334,
+      "step": 6190
     },
     {
       "epoch": 55.01,
-      "learning_rate": 3.798670465337132e-06,
-      "loss": 0.4729,
-      "step": 6540
+      "learning_rate": 3.837170503837171e-06,
+      "loss": 0.7971,
+      "step": 6200
     },
     {
       "epoch": 55.01,
-      "learning_rate": 3.7195314973092754e-06,
-      "loss": 0.6123,
-      "step": 6550
+      "learning_rate": 3.753753753753754e-06,
+      "loss": 0.3278,
+      "step": 6210
     },
     {
       "epoch": 55.01,
-      "learning_rate": 3.640392529281418e-06,
-      "loss": 0.5144,
-      "step": 6560
+      "learning_rate": 3.670337003670337e-06,
+      "loss": 0.3237,
+      "step": 6220
     },
     {
       "epoch": 55.01,
-      "learning_rate": 3.5612535612535615e-06,
-      "loss": 0.863,
-      "step": 6570
+      "learning_rate": 3.5869202535869206e-06,
+      "loss": 0.7723,
+      "step": 6230
     },
     {
       "epoch": 55.01,
-      "learning_rate": 3.482114593225704e-06,
-      "loss": 0.4071,
-      "step": 6580
+      "learning_rate": 3.503503503503504e-06,
+      "loss": 0.6996,
+      "step": 6240
     },
     {
       "epoch": 55.01,
-      "learning_rate": 3.4029756251978477e-06,
-      "loss": 0.5755,
-      "step": 6590
+      "learning_rate": 3.420086753420087e-06,
+      "loss": 0.544,
+      "step": 6250
     },
     {
       "epoch": 55.02,
-      "learning_rate": 3.3238366571699908e-06,
-      "loss": 0.6267,
-      "step": 6600
+      "learning_rate": 3.33667000333667e-06,
+      "loss": 0.6693,
+      "step": 6260
     },
     {
       "epoch": 55.02,
-      "eval_accuracy": 0.8064516129032258,
-      "eval_loss": 0.7372620701789856,
-      "eval_runtime": 267.3033,
-      "eval_samples_per_second": 0.812,
-      "eval_steps_per_second": 0.206,
-      "step": 6608
+      "learning_rate": 3.2532532532532535e-06,
+      "loss": 0.5206,
+      "step": 6270
     },
     {
-      "epoch": 56.0,
-      "learning_rate": 3.244697689142134e-06,
-      "loss": 0.3038,
-      "step": 6610
+      "epoch": 55.02,
+      "eval_accuracy": 0.7188940092165899,
+      "eval_loss": 1.0913441181182861,
+      "eval_runtime": 299.8505,
+      "eval_samples_per_second": 0.724,
+      "eval_steps_per_second": 0.183,
+      "step": 6272
     },
     {
       "epoch": 56.0,
-      "learning_rate": 3.165558721114277e-06,
-      "loss": 0.6753,
-      "step": 6620
+      "learning_rate": 3.1698365031698365e-06,
+      "loss": 0.2842,
+      "step": 6280
     },
     {
       "epoch": 56.0,
       "learning_rate": 3.0864197530864196e-06,
-      "loss": 0.4761,
-      "step": 6630
+      "loss": 0.4048,
+      "step": 6290
     },
     {
       "epoch": 56.0,
-      "learning_rate": 3.007280785058563e-06,
-      "loss": 0.4697,
-      "step": 6640
+      "learning_rate": 3.003003003003003e-06,
+      "loss": 0.6278,
+      "step": 6300
     },
     {
       "epoch": 56.01,
-      "learning_rate": 2.928141817030706e-06,
-      "loss": 0.2215,
-      "step": 6650
+      "learning_rate": 2.9195862529195865e-06,
+      "loss": 0.4073,
+      "step": 6310
     },
     {
       "epoch": 56.01,
-      "learning_rate": 2.8490028490028492e-06,
-      "loss": 0.4558,
-      "step": 6660
+      "learning_rate": 2.8361695028361695e-06,
+      "loss": 0.7256,
+      "step": 6320
     },
     {
       "epoch": 56.01,
-      "learning_rate": 2.7698638809749923e-06,
-      "loss": 0.7401,
-      "step": 6670
+      "learning_rate": 2.752752752752753e-06,
+      "loss": 0.5157,
+      "step": 6330
     },
     {
       "epoch": 56.01,
-      "learning_rate": 2.6907249129471354e-06,
-      "loss": 0.5682,
-      "step": 6680
+      "learning_rate": 2.669336002669336e-06,
+      "loss": 0.205,
+      "step": 6340
     },
     {
       "epoch": 56.01,
-      "learning_rate": 2.6115859449192785e-06,
-      "loss": 0.5346,
-      "step": 6690
+      "learning_rate": 2.5859192525859194e-06,
+      "loss": 0.513,
+      "step": 6350
     },
     {
       "epoch": 56.01,
-      "learning_rate": 2.5324469768914215e-06,
-      "loss": 0.7019,
-      "step": 6700
+      "learning_rate": 2.502502502502503e-06,
+      "loss": 0.3176,
+      "step": 6360
     },
     {
       "epoch": 56.01,
-      "learning_rate": 2.453308008863564e-06,
-      "loss": 0.4813,
-      "step": 6710
+      "learning_rate": 2.419085752419086e-06,
+      "loss": 0.4901,
+      "step": 6370
     },
     {
       "epoch": 56.02,
-      "learning_rate": 2.3741690408357077e-06,
-      "loss": 0.4561,
-      "step": 6720
+      "learning_rate": 2.3356690023356693e-06,
+      "loss": 0.7129,
+      "step": 6380
     },
     {
       "epoch": 56.02,
-      "eval_accuracy": 0.8018433179723502,
-      "eval_loss": 0.7382919192314148,
-      "eval_runtime": 275.3448,
-      "eval_samples_per_second": 0.788,
-      "eval_steps_per_second": 0.2,
-      "step": 6726
-    },
-    {
-      "epoch": 57.0,
-      "learning_rate": 2.2950300728078508e-06,
-      "loss": 0.4686,
-      "step": 6730
+      "eval_accuracy": 0.6866359447004609,
+      "eval_loss": 1.108277678489685,
+      "eval_runtime": 292.9571,
+      "eval_samples_per_second": 0.741,
+      "eval_steps_per_second": 0.188,
+      "step": 6384
     },
     {
       "epoch": 57.0,
-      "learning_rate": 2.215891104779994e-06,
-      "loss": 0.3041,
-      "step": 6740
+      "learning_rate": 2.2522522522522524e-06,
+      "loss": 0.2624,
+      "step": 6390
     },
     {
       "epoch": 57.0,
-      "learning_rate": 2.136752136752137e-06,
-      "loss": 0.241,
-      "step": 6750
+      "learning_rate": 2.168835502168836e-06,
+      "loss": 0.0914,
+      "step": 6400
     },
     {
       "epoch": 57.0,
-      "learning_rate": 2.05761316872428e-06,
-      "loss": 0.4126,
-      "step": 6760
+      "learning_rate": 2.085418752085419e-06,
+      "loss": 0.684,
+      "step": 6410
     },
     {
       "epoch": 57.01,
-      "learning_rate": 1.978474200696423e-06,
-      "loss": 0.286,
-      "step": 6770
+      "learning_rate": 2.002002002002002e-06,
+      "loss": 0.3441,
+      "step": 6420
     },
     {
       "epoch": 57.01,
-      "learning_rate": 1.899335232668566e-06,
-      "loss": 0.582,
-      "step": 6780
+      "learning_rate": 1.9185852519185853e-06,
+      "loss": 0.4939,
+      "step": 6430
     },
     {
       "epoch": 57.01,
-      "learning_rate": 1.820196264640709e-06,
-      "loss": 0.3814,
-      "step": 6790
+      "learning_rate": 1.8351685018351686e-06,
+      "loss": 0.1774,
+      "step": 6440
     },
     {
       "epoch": 57.01,
-      "learning_rate": 1.741057296612852e-06,
-      "loss": 0.6788,
-      "step": 6800
+      "learning_rate": 1.751751751751752e-06,
+      "loss": 0.5289,
+      "step": 6450
     },
     {
       "epoch": 57.01,
-      "learning_rate": 1.6619183285849954e-06,
-      "loss": 0.44,
-      "step": 6810
+      "learning_rate": 1.668335001668335e-06,
+      "loss": 0.9437,
+      "step": 6460
     },
     {
       "epoch": 57.01,
-      "learning_rate": 1.5827793605571385e-06,
-      "loss": 0.4996,
-      "step": 6820
+      "learning_rate": 1.5849182515849183e-06,
+      "loss": 0.7129,
+      "step": 6470
     },
     {
       "epoch": 57.01,
-      "learning_rate": 1.5036403925292815e-06,
-      "loss": 0.2839,
-      "step": 6830
+      "learning_rate": 1.5015015015015015e-06,
+      "loss": 0.1014,
+      "step": 6480
     },
     {
       "epoch": 57.02,
-      "learning_rate": 1.4245014245014246e-06,
-      "loss": 0.7002,
-      "step": 6840
+      "learning_rate": 1.4180847514180847e-06,
+      "loss": 0.4772,
+      "step": 6490
     },
     {
       "epoch": 57.02,
-      "eval_accuracy": 0.8110599078341014,
-      "eval_loss": 0.7072679400444031,
-      "eval_runtime": 266.757,
-      "eval_samples_per_second": 0.813,
-      "eval_steps_per_second": 0.206,
-      "step": 6844
+      "eval_accuracy": 0.7142857142857143,
+      "eval_loss": 1.1276459693908691,
+      "eval_runtime": 263.52,
+      "eval_samples_per_second": 0.823,
+      "eval_steps_per_second": 0.209,
+      "step": 6496
     },
     {
       "epoch": 58.0,
-      "learning_rate": 1.3453624564735677e-06,
-      "loss": 0.215,
-      "step": 6850
+      "learning_rate": 1.334668001334668e-06,
+      "loss": 0.5678,
+      "step": 6500
     },
     {
       "epoch": 58.0,
-      "learning_rate": 1.2662234884457108e-06,
-      "loss": 0.6569,
-      "step": 6860
+      "learning_rate": 1.2512512512512514e-06,
+      "loss": 0.2869,
+      "step": 6510
     },
     {
       "epoch": 58.0,
-      "learning_rate": 1.1870845204178538e-06,
-      "loss": 0.6116,
-      "step": 6870
+      "learning_rate": 1.1678345011678347e-06,
+      "loss": 0.9173,
+      "step": 6520
     },
     {
       "epoch": 58.01,
-      "learning_rate": 1.107945552389997e-06,
-      "loss": 0.3416,
-      "step": 6880
+      "learning_rate": 1.084417751084418e-06,
+      "loss": 0.5002,
+      "step": 6530
     },
     {
       "epoch": 58.01,
-      "learning_rate": 1.02880658436214e-06,
-      "loss": 0.571,
-      "step": 6890
+      "learning_rate": 1.001001001001001e-06,
+      "loss": 0.4222,
+      "step": 6540
     },
     {
       "epoch": 58.01,
-      "learning_rate": 9.49667616334283e-07,
-      "loss": 0.6238,
-      "step": 6900
+      "learning_rate": 9.175842509175843e-07,
+      "loss": 0.2793,
+      "step": 6550
     },
     {
       "epoch": 58.01,
-      "learning_rate": 8.70528648306426e-07,
-      "loss": 0.346,
-      "step": 6910
+      "learning_rate": 8.341675008341675e-07,
+      "loss": 0.2781,
+      "step": 6560
     },
     {
       "epoch": 58.01,
-      "learning_rate": 7.913896802785692e-07,
-      "loss": 0.4527,
-      "step": 6920
+      "learning_rate": 7.507507507507508e-07,
+      "loss": 0.4779,
+      "step": 6570
     },
     {
       "epoch": 58.01,
-      "learning_rate": 7.122507122507123e-07,
-      "loss": 0.5342,
-      "step": 6930
+      "learning_rate": 6.67334000667334e-07,
+      "loss": 0.4709,
+      "step": 6580
     },
     {
       "epoch": 58.01,
-      "learning_rate": 6.331117442228554e-07,
-      "loss": 0.523,
-      "step": 6940
-    },
-    {
-      "epoch": 58.02,
-      "learning_rate": 5.539727761949985e-07,
-      "loss": 0.2861,
-      "step": 6950
+      "learning_rate": 5.839172505839173e-07,
+      "loss": 0.5879,
+      "step": 6590
     },
     {
       "epoch": 58.02,
-      "learning_rate": 4.748338081671415e-07,
-      "loss": 0.1823,
-      "step": 6960
+      "learning_rate": 5.005005005005005e-07,
+      "loss": 0.4822,
+      "step": 6600
     },
     {
       "epoch": 58.02,
-      "eval_accuracy": 0.8202764976958525,
-      "eval_loss": 0.6870871186256409,
-      "eval_runtime": 266.4451,
-      "eval_samples_per_second": 0.814,
-      "eval_steps_per_second": 0.206,
-      "step": 6962
+      "eval_accuracy": 0.7235023041474654,
+      "eval_loss": 1.0919580459594727,
+      "eval_runtime": 304.8665,
+      "eval_samples_per_second": 0.712,
+      "eval_steps_per_second": 0.18,
+      "step": 6608
     },
     {
       "epoch": 59.0,
-      "learning_rate": 3.956948401392846e-07,
-      "loss": 0.5552,
-      "step": 6970
+      "learning_rate": 4.1708375041708376e-07,
+      "loss": 0.601,
+      "step": 6610
     },
     {
       "epoch": 59.0,
-      "learning_rate": 3.165558721114277e-07,
-      "loss": 0.4444,
-      "step": 6980
+      "learning_rate": 3.33667000333667e-07,
+      "loss": 0.5131,
+      "step": 6620
     },
     {
       "epoch": 59.0,
-      "learning_rate": 2.3741690408357074e-07,
-      "loss": 0.4752,
-      "step": 6990
+      "learning_rate": 2.5025025025025023e-07,
+      "loss": 0.175,
+      "step": 6630
     },
     {
-      "epoch": 59.01,
-      "learning_rate": 1.5827793605571385e-07,
-      "loss": 0.176,
-      "step": 7000
+      "epoch": 59.0,
+      "learning_rate": 1.668335001668335e-07,
+      "loss": 0.4028,
+      "step": 6640
     },
     {
       "epoch": 59.01,
-      "learning_rate": 7.913896802785692e-08,
-      "loss": 0.8408,
-      "step": 7010
+      "learning_rate": 8.341675008341675e-08,
+      "loss": 0.5526,
+      "step": 6650
     },
     {
       "epoch": 59.01,
       "learning_rate": 0.0,
-      "loss": 0.2439,
-      "step": 7020
-    },
-    {
-      "epoch": 59.01,
-      "eval_accuracy": 0.8202764976958525,
-      "eval_loss": 0.690118670463562,
-      "eval_runtime": 307.1421,
-      "eval_samples_per_second": 0.707,
-      "eval_steps_per_second": 0.179,
-      "step": 7020
-    },
-    {
-      "epoch": 59.01,
-      "step": 7020,
-      "total_flos": 3.476987046291161e+19,
-      "train_loss": 0.6254725841715125,
-      "train_runtime": 65940.1985,
-      "train_samples_per_second": 0.426,
-      "train_steps_per_second": 0.106
+      "loss": 0.6307,
+      "step": 6660
     },
     {
       "epoch": 59.01,
-      "eval_accuracy": 0.6990740740740741,
-      "eval_loss": 1.074745774269104,
-      "eval_runtime": 265.3059,
-      "eval_samples_per_second": 0.814,
-      "eval_steps_per_second": 0.204,
-      "step": 7020
+      "eval_accuracy": 0.7188940092165899,
+      "eval_loss": 1.098652958869934,
+      "eval_runtime": 289.552,
+      "eval_samples_per_second": 0.749,
+      "eval_steps_per_second": 0.19,
+      "step": 6660
     },
     {
       "epoch": 59.01,
-      "eval_accuracy": 0.6990740740740741,
-      "eval_loss": 1.074745774269104,
-      "eval_runtime": 283.9645,
-      "eval_samples_per_second": 0.761,
-      "eval_steps_per_second": 0.19,
-      "step": 7020
+      "step": 6660,
+      "total_flos": 3.3122529003141366e+19,
+      "train_loss": 0.7877646436920395,
+      "train_runtime": 57566.1475,
+      "train_samples_per_second": 0.463,
+      "train_steps_per_second": 0.116
     },
     {
       "epoch": 59.01,
-      "eval_accuracy": 0.6990740740740741,
-      "eval_loss": 1.0747456550598145,
-      "eval_runtime": 273.0941,
-      "eval_samples_per_second": 0.791,
-      "eval_steps_per_second": 0.198,
-      "step": 7020
+      "eval_accuracy": 0.6296296296296297,
+      "eval_loss": 0.8929917812347412,
+      "eval_runtime": 260.8949,
+      "eval_samples_per_second": 0.828,
+      "eval_steps_per_second": 0.207,
+      "step": 6660
     },
     {
       "epoch": 59.01,
-      "eval_accuracy": 0.6990740740740741,
-      "eval_loss": 1.074745774269104,
-      "eval_runtime": 320.3996,
-      "eval_samples_per_second": 0.674,
-      "eval_steps_per_second": 0.169,
-      "step": 7020
+      "eval_accuracy": 0.6296296296296297,
+      "eval_loss": 0.8929917812347412,
+      "eval_runtime": 274.034,
+      "eval_samples_per_second": 0.788,
+      "eval_steps_per_second": 0.197,
+      "step": 6660
     }
   ],
   "logging_steps": 10,
-  "max_steps": 7020,
+  "max_steps": 6660,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 9223372036854775807,
   "save_steps": 500,
-  "total_flos": 3.476987046291161e+19,
+  "total_flos": 3.3122529003141366e+19,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null