frostsolutions/frost-vision-v2-google_vit-base-patch16-224-v2024-11-09

README.md

@@ -26,16 +26,16 @@ model-index:
     metrics:
     - name: Accuracy
       type: accuracy
-      value: 0.9415492957746479
+      value: 0.9411971830985916
     - name: F1
       type: f1
-      value: 0.8501805054151624
+      value: 0.8485947416137806
     - name: Precision
       type: precision
-      value: 0.8517179023508138
+      value: 0.8540145985401459
     - name: Recall
       type: recall
-      value: 0.8486486486486486
+      value: 0.8432432432432433
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -45,11 +45,11 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/vit-base-patch16-224](https://huggingface.co/google/vit-base-patch16-224) on the webdataset dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1914
-- Accuracy: 0.9415
-- F1: 0.8502
-- Precision: 0.8517
-- Recall: 0.8486
+- Loss: 0.1716
+- Accuracy: 0.9412
+- F1: 0.8486
+- Precision: 0.8540
+- Recall: 0.8432
 
 ## Model description
 

all_results.json

@@ -1,16 +1,16 @@
 {
     "epoch": 30.0,
-    "eval_accuracy": 0.9316901408450704,
-    "eval_f1": 0.8267857142857142,
-    "eval_loss": 0.17685964703559875,
-    "eval_precision": 0.8312387791741472,
-    "eval_recall": 0.822380106571936,
-    "eval_runtime": 2.5564,
-    "eval_samples_per_second": 111.094,
-    "eval_steps_per_second": 14.082,
+    "eval_accuracy": 0.9411971830985916,
+    "eval_f1": 0.8485947416137806,
+    "eval_loss": 0.17164453864097595,
+    "eval_precision": 0.8540145985401459,
+    "eval_recall": 0.8432432432432433,
+    "eval_runtime": 2.933,
+    "eval_samples_per_second": 96.828,
+    "eval_steps_per_second": 12.274,
     "total_flos": 2.6364665116584346e+18,
-    "train_loss": 0.08667128345636135,
-    "train_runtime": 602.5463,
-    "train_samples_per_second": 56.46,
-    "train_steps_per_second": 3.535
+    "train_loss": 0.08832058483656023,
+    "train_runtime": 625.5786,
+    "train_samples_per_second": 54.382,
+    "train_steps_per_second": 3.405
 }

eval_results.json

@@ -1,11 +1,11 @@
 {
     "epoch": 30.0,
-    "eval_accuracy": 0.9316901408450704,
-    "eval_f1": 0.8267857142857142,
-    "eval_loss": 0.17685964703559875,
-    "eval_precision": 0.8312387791741472,
-    "eval_recall": 0.822380106571936,
-    "eval_runtime": 2.5564,
-    "eval_samples_per_second": 111.094,
-    "eval_steps_per_second": 14.082
+    "eval_accuracy": 0.9411971830985916,
+    "eval_f1": 0.8485947416137806,
+    "eval_loss": 0.17164453864097595,
+    "eval_precision": 0.8540145985401459,
+    "eval_recall": 0.8432432432432433,
+    "eval_runtime": 2.933,
+    "eval_samples_per_second": 96.828,
+    "eval_steps_per_second": 12.274
 }

runs/Nov09_22-53-35_b320e0693068/events.out.tfevents.1731193489.b320e0693068.1890.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:905de748efc443f64bc059569c134b6689c8123bb54dec4f3498b732cae60f05
+size 560

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 30.0,
     "total_flos": 2.6364665116584346e+18,
-    "train_loss": 0.08667128345636135,
-    "train_runtime": 602.5463,
-    "train_samples_per_second": 56.46,
-    "train_steps_per_second": 3.535
+    "train_loss": 0.08832058483656023,
+    "train_runtime": 625.5786,
+    "train_samples_per_second": 54.382,
+    "train_steps_per_second": 3.405
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.17685964703559875,
-  "best_model_checkpoint": "frost-vision-v2-google_vit-base-patch16-224-v2024-11-09/checkpoint-500",
+  "best_metric": 0.17164453864097595,
+  "best_model_checkpoint": "frost-vision-v2-google_vit-base-patch16-224-v2024-11-09/checkpoint-1000",
   "epoch": 30.0,
   "eval_steps": 100,
   "global_step": 2130,
@@ -10,1755 +10,1755 @@
   "log_history": [
     {
       "epoch": 0.14084507042253522,
-      "grad_norm": 2.3482229709625244,
+      "grad_norm": 1.7827035188674927,
       "learning_rate": 9.389671361502347e-06,
-      "loss": 0.7612,
+      "loss": 0.6939,
       "step": 10
     },
     {
       "epoch": 0.28169014084507044,
-      "grad_norm": 1.601408839225769,
+      "grad_norm": 1.6746207475662231,
       "learning_rate": 1.8779342723004694e-05,
-      "loss": 0.6382,
+      "loss": 0.6183,
       "step": 20
     },
     {
       "epoch": 0.4225352112676056,
-      "grad_norm": 1.162585735321045,
+      "grad_norm": 1.225101351737976,
       "learning_rate": 2.8169014084507046e-05,
-      "loss": 0.4766,
+      "loss": 0.4717,
       "step": 30
     },
     {
       "epoch": 0.5633802816901409,
-      "grad_norm": 0.8410656452178955,
+      "grad_norm": 1.1545004844665527,
       "learning_rate": 3.755868544600939e-05,
-      "loss": 0.4136,
+      "loss": 0.4007,
       "step": 40
     },
     {
       "epoch": 0.704225352112676,
-      "grad_norm": 0.8442389965057373,
+      "grad_norm": 1.03401517868042,
       "learning_rate": 4.694835680751174e-05,
-      "loss": 0.3483,
+      "loss": 0.3307,
       "step": 50
     },
     {
       "epoch": 0.8450704225352113,
-      "grad_norm": 0.8858609795570374,
+      "grad_norm": 1.1851330995559692,
       "learning_rate": 5.633802816901409e-05,
-      "loss": 0.312,
+      "loss": 0.3059,
       "step": 60
     },
     {
       "epoch": 0.9859154929577465,
-      "grad_norm": 0.911562442779541,
+      "grad_norm": 1.1718066930770874,
       "learning_rate": 6.572769953051644e-05,
-      "loss": 0.2769,
+      "loss": 0.307,
       "step": 70
     },
     {
       "epoch": 1.1267605633802817,
-      "grad_norm": 1.1327377557754517,
+      "grad_norm": 1.0279697179794312,
       "learning_rate": 7.511737089201878e-05,
-      "loss": 0.2412,
+      "loss": 0.247,
       "step": 80
     },
     {
       "epoch": 1.267605633802817,
-      "grad_norm": 0.9264270663261414,
+      "grad_norm": 0.9676464796066284,
       "learning_rate": 8.450704225352113e-05,
-      "loss": 0.2406,
+      "loss": 0.2487,
       "step": 90
     },
     {
       "epoch": 1.408450704225352,
-      "grad_norm": 0.6789488792419434,
+      "grad_norm": 1.0950922966003418,
       "learning_rate": 9.389671361502347e-05,
-      "loss": 0.2067,
+      "loss": 0.2398,
       "step": 100
     },
     {
       "epoch": 1.408450704225352,
-      "eval_accuracy": 0.9154929577464789,
-      "eval_f1": 0.7735849056603774,
-      "eval_loss": 0.2228715419769287,
-      "eval_precision": 0.8249496981891348,
-      "eval_recall": 0.7282415630550622,
-      "eval_runtime": 2.1991,
-      "eval_samples_per_second": 129.141,
-      "eval_steps_per_second": 16.37,
+      "eval_accuracy": 0.9214788732394367,
+      "eval_f1": 0.783284742468416,
+      "eval_loss": 0.20955488085746765,
+      "eval_precision": 0.8502109704641351,
+      "eval_recall": 0.7261261261261261,
+      "eval_runtime": 2.9059,
+      "eval_samples_per_second": 97.732,
+      "eval_steps_per_second": 12.389,
       "step": 100
     },
     {
       "epoch": 1.5492957746478875,
-      "grad_norm": 0.8904222846031189,
+      "grad_norm": 0.7596128582954407,
       "learning_rate": 0.00010328638497652582,
-      "loss": 0.2286,
+      "loss": 0.2263,
       "step": 110
     },
     {
       "epoch": 1.6901408450704225,
-      "grad_norm": 0.6201413869857788,
+      "grad_norm": 0.9664576649665833,
       "learning_rate": 0.00011267605633802819,
-      "loss": 0.2127,
+      "loss": 0.2153,
       "step": 120
     },
     {
       "epoch": 1.8309859154929577,
-      "grad_norm": 1.2106386423110962,
+      "grad_norm": 1.1117621660232544,
       "learning_rate": 0.00012206572769953053,
-      "loss": 0.2518,
+      "loss": 0.2372,
       "step": 130
     },
     {
       "epoch": 1.971830985915493,
-      "grad_norm": 0.9419664144515991,
+      "grad_norm": 0.805321216583252,
       "learning_rate": 0.00013145539906103288,
-      "loss": 0.2157,
+      "loss": 0.2102,
       "step": 140
     },
     {
       "epoch": 2.112676056338028,
-      "grad_norm": 0.7283549308776855,
+      "grad_norm": 1.1108847856521606,
       "learning_rate": 0.00014084507042253522,
-      "loss": 0.2119,
+      "loss": 0.2271,
       "step": 150
     },
     {
       "epoch": 2.2535211267605635,
-      "grad_norm": 0.7335492372512817,
+      "grad_norm": 0.7654628753662109,
       "learning_rate": 0.00015023474178403755,
-      "loss": 0.1762,
+      "loss": 0.1863,
       "step": 160
     },
     {
       "epoch": 2.3943661971830985,
-      "grad_norm": 1.075255036354065,
+      "grad_norm": 0.8311775326728821,
       "learning_rate": 0.00015962441314553992,
-      "loss": 0.1844,
+      "loss": 0.1649,
       "step": 170
     },
     {
       "epoch": 2.535211267605634,
-      "grad_norm": 0.7910892367362976,
+      "grad_norm": 0.9679107069969177,
       "learning_rate": 0.00016901408450704225,
-      "loss": 0.1904,
+      "loss": 0.19,
       "step": 180
     },
     {
       "epoch": 2.676056338028169,
-      "grad_norm": 0.9743360280990601,
+      "grad_norm": 1.0160714387893677,
       "learning_rate": 0.00017840375586854461,
-      "loss": 0.1968,
+      "loss": 0.2102,
       "step": 190
     },
     {
       "epoch": 2.816901408450704,
-      "grad_norm": 0.6024548411369324,
+      "grad_norm": 0.6983464956283569,
       "learning_rate": 0.00018779342723004695,
-      "loss": 0.1989,
+      "loss": 0.1746,
       "step": 200
     },
     {
       "epoch": 2.816901408450704,
-      "eval_accuracy": 0.9102112676056338,
-      "eval_f1": 0.7649769585253456,
-      "eval_loss": 0.2252175509929657,
-      "eval_precision": 0.7950191570881227,
-      "eval_recall": 0.7371225577264654,
-      "eval_runtime": 3.0527,
-      "eval_samples_per_second": 93.033,
-      "eval_steps_per_second": 11.793,
+      "eval_accuracy": 0.9369718309859155,
+      "eval_f1": 0.8362305580969808,
+      "eval_loss": 0.16762028634548187,
+      "eval_precision": 0.8494423791821561,
+      "eval_recall": 0.8234234234234235,
+      "eval_runtime": 2.0426,
+      "eval_samples_per_second": 139.041,
+      "eval_steps_per_second": 17.625,
       "step": 200
     },
     {
       "epoch": 2.9577464788732395,
-      "grad_norm": 0.7431610226631165,
+      "grad_norm": 0.7636697888374329,
       "learning_rate": 0.0001971830985915493,
-      "loss": 0.1922,
+      "loss": 0.2038,
       "step": 210
     },
     {
       "epoch": 3.0985915492957745,
-      "grad_norm": 0.8809910416603088,
+      "grad_norm": 0.7589529156684875,
       "learning_rate": 0.00019926969222743872,
-      "loss": 0.2013,
+      "loss": 0.2047,
       "step": 220
     },
     {
       "epoch": 3.23943661971831,
-      "grad_norm": 0.7516496181488037,
+      "grad_norm": 0.785400927066803,
       "learning_rate": 0.00019822639540949402,
-      "loss": 0.175,
+      "loss": 0.1831,
       "step": 230
     },
     {
       "epoch": 3.380281690140845,
-      "grad_norm": 1.8546067476272583,
+      "grad_norm": 0.9033864736557007,
       "learning_rate": 0.0001971830985915493,
-      "loss": 0.1817,
+      "loss": 0.1743,
       "step": 240
     },
     {
       "epoch": 3.52112676056338,
-      "grad_norm": 0.7087880969047546,
+      "grad_norm": 1.0347667932510376,
       "learning_rate": 0.0001961398017736046,
-      "loss": 0.1663,
+      "loss": 0.1691,
       "step": 250
     },
     {
       "epoch": 3.6619718309859155,
-      "grad_norm": 0.890684962272644,
+      "grad_norm": 0.9231723546981812,
       "learning_rate": 0.0001950965049556599,
-      "loss": 0.1782,
+      "loss": 0.1649,
       "step": 260
     },
     {
       "epoch": 3.802816901408451,
-      "grad_norm": 1.163325548171997,
+      "grad_norm": 1.205672264099121,
       "learning_rate": 0.0001940532081377152,
-      "loss": 0.1868,
+      "loss": 0.2018,
       "step": 270
     },
     {
       "epoch": 3.943661971830986,
-      "grad_norm": 0.7395078539848328,
+      "grad_norm": 0.6101861000061035,
       "learning_rate": 0.0001930099113197705,
-      "loss": 0.176,
+      "loss": 0.1805,
       "step": 280
     },
     {
       "epoch": 4.084507042253521,
-      "grad_norm": 1.1575692892074585,
+      "grad_norm": 0.8451747298240662,
       "learning_rate": 0.0001919666145018258,
-      "loss": 0.1603,
+      "loss": 0.1479,
       "step": 290
     },
     {
       "epoch": 4.225352112676056,
-      "grad_norm": 0.7146375179290771,
+      "grad_norm": 1.2933802604675293,
       "learning_rate": 0.0001909233176838811,
-      "loss": 0.1364,
+      "loss": 0.1316,
       "step": 300
     },
     {
       "epoch": 4.225352112676056,
-      "eval_accuracy": 0.9267605633802817,
-      "eval_f1": 0.8162544169611308,
-      "eval_loss": 0.18342211842536926,
-      "eval_precision": 0.81195079086116,
-      "eval_recall": 0.8206039076376554,
-      "eval_runtime": 2.1207,
-      "eval_samples_per_second": 133.919,
-      "eval_steps_per_second": 16.976,
+      "eval_accuracy": 0.928169014084507,
+      "eval_f1": 0.8125,
+      "eval_loss": 0.17500653862953186,
+      "eval_precision": 0.8292682926829268,
+      "eval_recall": 0.7963963963963964,
+      "eval_runtime": 2.0499,
+      "eval_samples_per_second": 138.543,
+      "eval_steps_per_second": 17.562,
       "step": 300
     },
     {
       "epoch": 4.366197183098592,
-      "grad_norm": 1.2778205871582031,
+      "grad_norm": 1.1605525016784668,
       "learning_rate": 0.00018988002086593636,
-      "loss": 0.1571,
+      "loss": 0.1595,
       "step": 310
     },
     {
       "epoch": 4.507042253521127,
-      "grad_norm": 1.1578363180160522,
+      "grad_norm": 0.7514552474021912,
       "learning_rate": 0.00018883672404799165,
-      "loss": 0.1552,
+      "loss": 0.1455,
       "step": 320
     },
     {
       "epoch": 4.647887323943662,
-      "grad_norm": 1.1817420721054077,
+      "grad_norm": 0.9141556620597839,
       "learning_rate": 0.00018779342723004695,
-      "loss": 0.1755,
+      "loss": 0.1737,
       "step": 330
     },
     {
       "epoch": 4.788732394366197,
-      "grad_norm": 1.138913869857788,
+      "grad_norm": 0.7212592363357544,
       "learning_rate": 0.00018675013041210225,
-      "loss": 0.1485,
+      "loss": 0.1464,
       "step": 340
     },
     {
       "epoch": 4.929577464788732,
-      "grad_norm": 1.0575881004333496,
+      "grad_norm": 1.1275196075439453,
       "learning_rate": 0.00018570683359415754,
-      "loss": 0.1777,
+      "loss": 0.2212,
       "step": 350
     },
     {
       "epoch": 5.070422535211268,
-      "grad_norm": 0.7797971367835999,
+      "grad_norm": 0.8327043652534485,
       "learning_rate": 0.00018466353677621284,
-      "loss": 0.1672,
+      "loss": 0.144,
       "step": 360
     },
     {
       "epoch": 5.211267605633803,
-      "grad_norm": 0.8628588914871216,
+      "grad_norm": 0.8596853017807007,
       "learning_rate": 0.00018362023995826813,
-      "loss": 0.1398,
+      "loss": 0.1347,
       "step": 370
     },
     {
       "epoch": 5.352112676056338,
-      "grad_norm": 0.7360591888427734,
+      "grad_norm": 0.8283455967903137,
       "learning_rate": 0.00018257694314032343,
-      "loss": 0.1481,
+      "loss": 0.1558,
       "step": 380
     },
     {
       "epoch": 5.492957746478873,
-      "grad_norm": 1.0168707370758057,
+      "grad_norm": 0.7472031116485596,
       "learning_rate": 0.00018153364632237873,
-      "loss": 0.1362,
+      "loss": 0.1336,
       "step": 390
     },
     {
       "epoch": 5.633802816901408,
-      "grad_norm": 0.850728452205658,
+      "grad_norm": 0.8499220013618469,
       "learning_rate": 0.00018049034950443402,
-      "loss": 0.1368,
+      "loss": 0.1305,
       "step": 400
     },
     {
       "epoch": 5.633802816901408,
-      "eval_accuracy": 0.9267605633802817,
-      "eval_f1": 0.7980582524271844,
-      "eval_loss": 0.18743538856506348,
-      "eval_precision": 0.880085653104925,
-      "eval_recall": 0.7300177619893428,
-      "eval_runtime": 2.1497,
-      "eval_samples_per_second": 132.11,
-      "eval_steps_per_second": 16.746,
+      "eval_accuracy": 0.9341549295774648,
+      "eval_f1": 0.8270120259019427,
+      "eval_loss": 0.16712401807308197,
+      "eval_precision": 0.8498098859315589,
+      "eval_recall": 0.8054054054054054,
+      "eval_runtime": 4.4386,
+      "eval_samples_per_second": 63.984,
+      "eval_steps_per_second": 8.111,
       "step": 400
     },
     {
       "epoch": 5.774647887323944,
-      "grad_norm": 0.4130128026008606,
+      "grad_norm": 0.8095440864562988,
       "learning_rate": 0.00017944705268648932,
-      "loss": 0.1455,
+      "loss": 0.1625,
       "step": 410
     },
     {
       "epoch": 5.915492957746479,
-      "grad_norm": 0.896547794342041,
+      "grad_norm": 0.7673202753067017,
       "learning_rate": 0.00017840375586854461,
-      "loss": 0.135,
+      "loss": 0.1327,
       "step": 420
     },
     {
       "epoch": 6.056338028169014,
-      "grad_norm": 0.6682694554328918,
+      "grad_norm": 1.032628059387207,
       "learning_rate": 0.0001773604590505999,
-      "loss": 0.1199,
+      "loss": 0.1508,
       "step": 430
     },
     {
       "epoch": 6.197183098591549,
-      "grad_norm": 0.6317305564880371,
+      "grad_norm": 0.7713437676429749,
       "learning_rate": 0.0001763171622326552,
-      "loss": 0.1043,
+      "loss": 0.1286,
       "step": 440
     },
     {
       "epoch": 6.338028169014084,
-      "grad_norm": 1.0829790830612183,
+      "grad_norm": 0.5040615200996399,
       "learning_rate": 0.0001752738654147105,
-      "loss": 0.1347,
+      "loss": 0.1253,
       "step": 450
     },
     {
       "epoch": 6.47887323943662,
-      "grad_norm": 1.4727702140808105,
+      "grad_norm": 0.8310524225234985,
       "learning_rate": 0.0001742305685967658,
-      "loss": 0.1244,
+      "loss": 0.1268,
       "step": 460
     },
     {
       "epoch": 6.619718309859155,
-      "grad_norm": 1.014894962310791,
+      "grad_norm": 0.9123954176902771,
       "learning_rate": 0.00017318727177882107,
-      "loss": 0.1377,
+      "loss": 0.1325,
       "step": 470
     },
     {
       "epoch": 6.76056338028169,
-      "grad_norm": 0.9650211334228516,
+      "grad_norm": 0.5047963857650757,
       "learning_rate": 0.00017214397496087636,
-      "loss": 0.1443,
+      "loss": 0.1225,
       "step": 480
     },
     {
       "epoch": 6.901408450704225,
-      "grad_norm": 0.796774685382843,
+      "grad_norm": 0.5679495930671692,
       "learning_rate": 0.00017110067814293166,
-      "loss": 0.1165,
+      "loss": 0.1156,
       "step": 490
     },
     {
       "epoch": 7.042253521126761,
-      "grad_norm": 0.5631269216537476,
+      "grad_norm": 0.6002945899963379,
       "learning_rate": 0.00017005738132498696,
-      "loss": 0.1197,
+      "loss": 0.1119,
       "step": 500
     },
     {
       "epoch": 7.042253521126761,
       "eval_accuracy": 0.9316901408450704,
-      "eval_f1": 0.8267857142857142,
-      "eval_loss": 0.17685964703559875,
-      "eval_precision": 0.8312387791741472,
-      "eval_recall": 0.822380106571936,
-      "eval_runtime": 2.1178,
-      "eval_samples_per_second": 134.103,
-      "eval_steps_per_second": 16.999,
+      "eval_f1": 0.8239564428312159,
+      "eval_loss": 0.17465578019618988,
+      "eval_precision": 0.829981718464351,
+      "eval_recall": 0.818018018018018,
+      "eval_runtime": 2.9059,
+      "eval_samples_per_second": 97.733,
+      "eval_steps_per_second": 12.389,
       "step": 500
     },
     {
       "epoch": 7.183098591549296,
-      "grad_norm": 0.40783101320266724,
+      "grad_norm": 0.7332452535629272,
       "learning_rate": 0.00016901408450704225,
-      "loss": 0.0913,
+      "loss": 0.1237,
       "step": 510
     },
     {
       "epoch": 7.323943661971831,
-      "grad_norm": 0.536299467086792,
+      "grad_norm": 0.5455431342124939,
       "learning_rate": 0.00016797078768909755,
-      "loss": 0.0983,
+      "loss": 0.1333,
       "step": 520
     },
     {
       "epoch": 7.464788732394366,
-      "grad_norm": 0.587491512298584,
+      "grad_norm": 0.7222278714179993,
       "learning_rate": 0.00016692749087115284,
-      "loss": 0.116,
+      "loss": 0.1126,
       "step": 530
     },
     {
       "epoch": 7.605633802816901,
-      "grad_norm": 0.5410102009773254,
+      "grad_norm": 0.6804931163787842,
       "learning_rate": 0.00016588419405320814,
-      "loss": 0.1075,
+      "loss": 0.1242,
       "step": 540
     },
     {
       "epoch": 7.746478873239437,
-      "grad_norm": 0.5919251441955566,
+      "grad_norm": 0.6327610015869141,
       "learning_rate": 0.00016484089723526344,
-      "loss": 0.1132,
+      "loss": 0.1079,
       "step": 550
     },
     {
       "epoch": 7.887323943661972,
-      "grad_norm": 0.632838785648346,
+      "grad_norm": 0.7840393781661987,
       "learning_rate": 0.00016379760041731873,
-      "loss": 0.0963,
+      "loss": 0.0932,
       "step": 560
     },
     {
       "epoch": 8.028169014084508,
-      "grad_norm": 0.3347577750682831,
+      "grad_norm": 0.9229692220687866,
       "learning_rate": 0.00016275430359937403,
-      "loss": 0.0852,
+      "loss": 0.1163,
       "step": 570
     },
     {
       "epoch": 8.169014084507042,
-      "grad_norm": 0.4444897472858429,
+      "grad_norm": 0.7363346815109253,
       "learning_rate": 0.00016171100678142932,
-      "loss": 0.0998,
+      "loss": 0.116,
       "step": 580
     },
     {
       "epoch": 8.309859154929578,
-      "grad_norm": 0.8959450125694275,
+      "grad_norm": 0.6889787912368774,
       "learning_rate": 0.00016066770996348462,
-      "loss": 0.0983,
+      "loss": 0.1043,
       "step": 590
     },
     {
       "epoch": 8.450704225352112,
-      "grad_norm": 0.8289921283721924,
+      "grad_norm": 0.43690377473831177,
       "learning_rate": 0.00015962441314553992,
-      "loss": 0.099,
+      "loss": 0.0913,
       "step": 600
     },
     {
       "epoch": 8.450704225352112,
-      "eval_accuracy": 0.9313380281690141,
-      "eval_f1": 0.8189415041782729,
-      "eval_loss": 0.18414147198200226,
-      "eval_precision": 0.857976653696498,
-      "eval_recall": 0.783303730017762,
-      "eval_runtime": 2.7312,
-      "eval_samples_per_second": 103.983,
-      "eval_steps_per_second": 13.181,
+      "eval_accuracy": 0.9415492957746479,
+      "eval_f1": 0.8504504504504504,
+      "eval_loss": 0.15154092013835907,
+      "eval_precision": 0.8504504504504504,
+      "eval_recall": 0.8504504504504504,
+      "eval_runtime": 2.0762,
+      "eval_samples_per_second": 136.789,
+      "eval_steps_per_second": 17.339,
       "step": 600
     },
     {
       "epoch": 8.591549295774648,
-      "grad_norm": 0.8808104395866394,
+      "grad_norm": 0.6768948435783386,
       "learning_rate": 0.0001585811163275952,
-      "loss": 0.1169,
+      "loss": 0.1107,
       "step": 610
     },
     {
       "epoch": 8.732394366197184,
-      "grad_norm": 0.23073981702327728,
+      "grad_norm": 0.308359295129776,
       "learning_rate": 0.0001575378195096505,
-      "loss": 0.0751,
+      "loss": 0.0979,
       "step": 620
     },
     {
       "epoch": 8.873239436619718,
-      "grad_norm": 0.6636078953742981,
+      "grad_norm": 0.825429379940033,
       "learning_rate": 0.0001564945226917058,
-      "loss": 0.0955,
+      "loss": 0.1135,
       "step": 630
     },
     {
       "epoch": 9.014084507042254,
-      "grad_norm": 0.6324308514595032,
+      "grad_norm": 0.5222116708755493,
       "learning_rate": 0.00015545122587376107,
-      "loss": 0.1028,
+      "loss": 0.1038,
       "step": 640
     },
     {
       "epoch": 9.154929577464788,
-      "grad_norm": 0.6154828667640686,
+      "grad_norm": 0.36319026350975037,
       "learning_rate": 0.00015440792905581637,
-      "loss": 0.084,
+      "loss": 0.0911,
       "step": 650
     },
     {
       "epoch": 9.295774647887324,
-      "grad_norm": 0.5773690342903137,
+      "grad_norm": 0.7786665558815002,
       "learning_rate": 0.00015336463223787167,
-      "loss": 0.0929,
+      "loss": 0.0948,
       "step": 660
     },
     {
       "epoch": 9.43661971830986,
-      "grad_norm": 0.3276579976081848,
+      "grad_norm": 0.462139368057251,
       "learning_rate": 0.00015232133541992696,
-      "loss": 0.081,
+      "loss": 0.0854,
       "step": 670
     },
     {
       "epoch": 9.577464788732394,
-      "grad_norm": 0.6751773357391357,
+      "grad_norm": 0.5465391278266907,
       "learning_rate": 0.00015127803860198226,
-      "loss": 0.0798,
+      "loss": 0.069,
       "step": 680
     },
     {
       "epoch": 9.71830985915493,
-      "grad_norm": 0.41989338397979736,
+      "grad_norm": 0.4502487778663635,
       "learning_rate": 0.00015023474178403755,
-      "loss": 0.0862,
+      "loss": 0.0854,
       "step": 690
     },
     {
       "epoch": 9.859154929577464,
-      "grad_norm": 1.3621422052383423,
+      "grad_norm": 0.7790795564651489,
       "learning_rate": 0.00014919144496609285,
-      "loss": 0.0748,
+      "loss": 0.0964,
       "step": 700
     },
     {
       "epoch": 9.859154929577464,
-      "eval_accuracy": 0.9359154929577465,
-      "eval_f1": 0.8366247755834829,
-      "eval_loss": 0.17389163374900818,
-      "eval_precision": 0.8457350272232305,
-      "eval_recall": 0.827708703374778,
-      "eval_runtime": 2.2415,
-      "eval_samples_per_second": 126.7,
-      "eval_steps_per_second": 16.061,
+      "eval_accuracy": 0.9376760563380282,
+      "eval_f1": 0.8418230563002681,
+      "eval_loss": 0.16803742945194244,
+      "eval_precision": 0.8351063829787234,
+      "eval_recall": 0.8486486486486486,
+      "eval_runtime": 2.0298,
+      "eval_samples_per_second": 139.919,
+      "eval_steps_per_second": 17.736,
       "step": 700
     },
     {
       "epoch": 10.0,
-      "grad_norm": 0.49502134323120117,
+      "grad_norm": 0.28997573256492615,
       "learning_rate": 0.00014814814814814815,
-      "loss": 0.0718,
+      "loss": 0.0819,
       "step": 710
     },
     {
       "epoch": 10.140845070422536,
-      "grad_norm": 1.2556084394454956,
+      "grad_norm": 0.49839815497398376,
       "learning_rate": 0.00014710485133020344,
-      "loss": 0.0715,
+      "loss": 0.0639,
       "step": 720
     },
     {
       "epoch": 10.28169014084507,
-      "grad_norm": 0.5221930742263794,
+      "grad_norm": 0.6128919720649719,
       "learning_rate": 0.00014606155451225874,
-      "loss": 0.0863,
+      "loss": 0.0783,
       "step": 730
     },
     {
       "epoch": 10.422535211267606,
-      "grad_norm": 0.9766262173652649,
+      "grad_norm": 0.7849855422973633,
       "learning_rate": 0.00014501825769431403,
-      "loss": 0.0791,
+      "loss": 0.0762,
       "step": 740
     },
     {
       "epoch": 10.56338028169014,
-      "grad_norm": 0.447653204202652,
+      "grad_norm": 0.3956754803657532,
       "learning_rate": 0.00014397496087636933,
-      "loss": 0.068,
+      "loss": 0.082,
       "step": 750
     },
     {
       "epoch": 10.704225352112676,
-      "grad_norm": 0.7212210893630981,
+      "grad_norm": 0.9166898131370544,
       "learning_rate": 0.00014293166405842463,
-      "loss": 0.0755,
+      "loss": 0.0856,
       "step": 760
     },
     {
       "epoch": 10.845070422535212,
-      "grad_norm": 0.4740016460418701,
+      "grad_norm": 0.7876815795898438,
       "learning_rate": 0.00014188836724047992,
-      "loss": 0.082,
+      "loss": 0.0822,
       "step": 770
     },
     {
       "epoch": 10.985915492957746,
-      "grad_norm": 0.8718104362487793,
+      "grad_norm": 0.41309139132499695,
       "learning_rate": 0.00014084507042253522,
-      "loss": 0.0845,
+      "loss": 0.0822,
       "step": 780
     },
     {
       "epoch": 11.126760563380282,
-      "grad_norm": 0.6141758561134338,
+      "grad_norm": 0.5138004422187805,
       "learning_rate": 0.00013980177360459051,
-      "loss": 0.0637,
+      "loss": 0.0666,
       "step": 790
     },
     {
       "epoch": 11.267605633802816,
-      "grad_norm": 1.284001350402832,
+      "grad_norm": 0.6155378222465515,
       "learning_rate": 0.00013875847678664578,
-      "loss": 0.0706,
+      "loss": 0.0659,
       "step": 800
     },
     {
       "epoch": 11.267605633802816,
-      "eval_accuracy": 0.9373239436619718,
-      "eval_f1": 0.8399280575539568,
-      "eval_loss": 0.17618876695632935,
-      "eval_precision": 0.8506375227686703,
-      "eval_recall": 0.8294849023090586,
-      "eval_runtime": 2.1627,
-      "eval_samples_per_second": 131.319,
-      "eval_steps_per_second": 16.646,
+      "eval_accuracy": 0.9274647887323944,
+      "eval_f1": 0.8144144144144144,
+      "eval_loss": 0.1891278326511383,
+      "eval_precision": 0.8144144144144144,
+      "eval_recall": 0.8144144144144144,
+      "eval_runtime": 2.8812,
+      "eval_samples_per_second": 98.569,
+      "eval_steps_per_second": 12.495,
       "step": 800
     },
     {
       "epoch": 11.408450704225352,
-      "grad_norm": 0.4294152855873108,
+      "grad_norm": 0.7656810283660889,
       "learning_rate": 0.00013771517996870108,
-      "loss": 0.0707,
+      "loss": 0.0947,
       "step": 810
     },
     {
       "epoch": 11.549295774647888,
-      "grad_norm": 0.28866302967071533,
+      "grad_norm": 0.4831065237522125,
       "learning_rate": 0.00013667188315075638,
-      "loss": 0.057,
+      "loss": 0.0836,
       "step": 820
     },
     {
       "epoch": 11.690140845070422,
-      "grad_norm": 0.7050051689147949,
+      "grad_norm": 0.7897108793258667,
       "learning_rate": 0.0001356285863328117,
-      "loss": 0.0771,
+      "loss": 0.0679,
       "step": 830
     },
     {
       "epoch": 11.830985915492958,
-      "grad_norm": 0.3648245632648468,
+      "grad_norm": 0.6471636891365051,
       "learning_rate": 0.000134585289514867,
-      "loss": 0.0578,
+      "loss": 0.1061,
       "step": 840
     },
     {
       "epoch": 11.971830985915492,
-      "grad_norm": 0.9216500520706177,
+      "grad_norm": 0.6389226317405701,
       "learning_rate": 0.0001335419926969223,
-      "loss": 0.0724,
+      "loss": 0.0735,
       "step": 850
     },
     {
       "epoch": 12.112676056338028,
-      "grad_norm": 0.3451094925403595,
+      "grad_norm": 0.4462195336818695,
       "learning_rate": 0.0001324986958789776,
-      "loss": 0.0756,
+      "loss": 0.0827,
       "step": 860
     },
     {
       "epoch": 12.253521126760564,
-      "grad_norm": 0.45862048864364624,
+      "grad_norm": 0.42086052894592285,
       "learning_rate": 0.00013145539906103288,
-      "loss": 0.057,
+      "loss": 0.0662,
       "step": 870
     },
     {
       "epoch": 12.394366197183098,
-      "grad_norm": 0.562519371509552,
+      "grad_norm": 0.7582985162734985,
       "learning_rate": 0.00013041210224308818,
-      "loss": 0.0761,
+      "loss": 0.0877,
       "step": 880
     },
     {
       "epoch": 12.535211267605634,
-      "grad_norm": 0.3094978928565979,
+      "grad_norm": 0.6410930156707764,
       "learning_rate": 0.00012936880542514348,
-      "loss": 0.0614,
+      "loss": 0.0659,
       "step": 890
     },
     {
       "epoch": 12.676056338028168,
-      "grad_norm": 0.9596508145332336,
+      "grad_norm": 1.0315440893173218,
       "learning_rate": 0.00012832550860719877,
-      "loss": 0.0865,
+      "loss": 0.0706,
       "step": 900
     },
     {
       "epoch": 12.676056338028168,
-      "eval_accuracy": 0.9408450704225352,
-      "eval_f1": 0.8486486486486486,
-      "eval_loss": 0.17658686637878418,
-      "eval_precision": 0.8610603290676416,
-      "eval_recall": 0.8365896980461812,
-      "eval_runtime": 2.0967,
-      "eval_samples_per_second": 135.451,
-      "eval_steps_per_second": 17.17,
+      "eval_accuracy": 0.9320422535211268,
+      "eval_f1": 0.8234217749313815,
+      "eval_loss": 0.17875343561172485,
+      "eval_precision": 0.8364312267657993,
+      "eval_recall": 0.8108108108108109,
+      "eval_runtime": 5.3131,
+      "eval_samples_per_second": 53.453,
+      "eval_steps_per_second": 6.776,
       "step": 900
     },
     {
       "epoch": 12.816901408450704,
-      "grad_norm": 0.4770239293575287,
+      "grad_norm": 0.7598835825920105,
       "learning_rate": 0.00012728221178925407,
-      "loss": 0.0657,
+      "loss": 0.0675,
       "step": 910
     },
     {
       "epoch": 12.95774647887324,
-      "grad_norm": 0.4350703954696655,
+      "grad_norm": 0.425029993057251,
       "learning_rate": 0.00012623891497130936,
-      "loss": 0.0628,
+      "loss": 0.0691,
       "step": 920
     },
     {
       "epoch": 13.098591549295774,
-      "grad_norm": 0.4072926342487335,
+      "grad_norm": 0.39021769165992737,
       "learning_rate": 0.00012519561815336466,
-      "loss": 0.05,
+      "loss": 0.0645,
       "step": 930
     },
     {
       "epoch": 13.23943661971831,
-      "grad_norm": 1.0825406312942505,
+      "grad_norm": 0.6953460574150085,
       "learning_rate": 0.00012415232133541993,
-      "loss": 0.0659,
+      "loss": 0.0695,
       "step": 940
     },
     {
       "epoch": 13.380281690140846,
-      "grad_norm": 1.0538973808288574,
+      "grad_norm": 0.8735091090202332,
       "learning_rate": 0.00012310902451747523,
-      "loss": 0.0512,
+      "loss": 0.0523,
       "step": 950
     },
     {
       "epoch": 13.52112676056338,
-      "grad_norm": 0.6860778331756592,
+      "grad_norm": 0.9906308650970459,
       "learning_rate": 0.00012206572769953053,
-      "loss": 0.0548,
+      "loss": 0.0591,
       "step": 960
     },
     {
       "epoch": 13.661971830985916,
-      "grad_norm": 0.8300151228904724,
+      "grad_norm": 0.7933890223503113,
       "learning_rate": 0.00012102243088158583,
-      "loss": 0.0725,
+      "loss": 0.0614,
       "step": 970
     },
     {
       "epoch": 13.80281690140845,
-      "grad_norm": 0.38458070158958435,
+      "grad_norm": 0.4813683331012726,
       "learning_rate": 0.00011997913406364111,
-      "loss": 0.0655,
+      "loss": 0.056,
       "step": 980
     },
     {
       "epoch": 13.943661971830986,
-      "grad_norm": 0.745091438293457,
+      "grad_norm": 0.4669698476791382,
       "learning_rate": 0.00011893583724569641,
-      "loss": 0.0636,
+      "loss": 0.0583,
       "step": 990
     },
     {
       "epoch": 14.084507042253522,
-      "grad_norm": 0.8269743919372559,
+      "grad_norm": 0.6160480976104736,
       "learning_rate": 0.0001178925404277517,
-      "loss": 0.061,
+      "loss": 0.069,
       "step": 1000
     },
     {
       "epoch": 14.084507042253522,
-      "eval_accuracy": 0.9380281690140845,
-      "eval_f1": 0.8445229681978799,
-      "eval_loss": 0.18520130217075348,
-      "eval_precision": 0.8400702987697716,
-      "eval_recall": 0.8490230905861457,
-      "eval_runtime": 3.0489,
-      "eval_samples_per_second": 93.147,
-      "eval_steps_per_second": 11.807,
+      "eval_accuracy": 0.9411971830985916,
+      "eval_f1": 0.8485947416137806,
+      "eval_loss": 0.17164453864097595,
+      "eval_precision": 0.8540145985401459,
+      "eval_recall": 0.8432432432432433,
+      "eval_runtime": 3.5722,
+      "eval_samples_per_second": 79.503,
+      "eval_steps_per_second": 10.078,
       "step": 1000
     },
     {
       "epoch": 14.225352112676056,
-      "grad_norm": 0.2855857312679291,
+      "grad_norm": 0.3227427899837494,
       "learning_rate": 0.000116849243609807,
-      "loss": 0.0564,
+      "loss": 0.0677,
       "step": 1010
     },
     {
       "epoch": 14.366197183098592,
-      "grad_norm": 0.4285324811935425,
+      "grad_norm": 0.6062753796577454,
       "learning_rate": 0.0001158059467918623,
-      "loss": 0.0682,
+      "loss": 0.0551,
       "step": 1020
     },
     {
       "epoch": 14.507042253521126,
-      "grad_norm": 0.45068418979644775,
+      "grad_norm": 0.37269070744514465,
       "learning_rate": 0.0001147626499739176,
-      "loss": 0.0619,
+      "loss": 0.0722,
       "step": 1030
     },
     {
       "epoch": 14.647887323943662,
-      "grad_norm": 0.4685414433479309,
+      "grad_norm": 0.5302610397338867,
       "learning_rate": 0.00011371935315597289,
-      "loss": 0.0556,
+      "loss": 0.069,
       "step": 1040
     },
     {
       "epoch": 14.788732394366198,
-      "grad_norm": 0.2533360719680786,
+      "grad_norm": 0.3961109519004822,
       "learning_rate": 0.00011267605633802819,
-      "loss": 0.0513,
+      "loss": 0.0508,
       "step": 1050
     },
     {
       "epoch": 14.929577464788732,
-      "grad_norm": 0.45339080691337585,
+      "grad_norm": 0.5787925720214844,
       "learning_rate": 0.00011163275952008347,
-      "loss": 0.0507,
+      "loss": 0.0513,
       "step": 1060
     },
     {
       "epoch": 15.070422535211268,
-      "grad_norm": 0.4735059142112732,
+      "grad_norm": 0.4569196403026581,
       "learning_rate": 0.00011058946270213876,
-      "loss": 0.0642,
+      "loss": 0.0504,
       "step": 1070
     },
     {
       "epoch": 15.211267605633802,
-      "grad_norm": 0.2932089567184448,
+      "grad_norm": 0.21465392410755157,
       "learning_rate": 0.00010954616588419406,
-      "loss": 0.0412,
+      "loss": 0.0588,
       "step": 1080
     },
     {
       "epoch": 15.352112676056338,
-      "grad_norm": 0.262364000082016,
+      "grad_norm": 0.35163068771362305,
       "learning_rate": 0.00010850286906624936,
-      "loss": 0.0561,
+      "loss": 0.0537,
       "step": 1090
     },
     {
       "epoch": 15.492957746478874,
-      "grad_norm": 0.4154641032218933,
+      "grad_norm": 0.43309277296066284,
       "learning_rate": 0.00010745957224830465,
-      "loss": 0.0449,
+      "loss": 0.0543,
       "step": 1100
     },
     {
       "epoch": 15.492957746478874,
-      "eval_accuracy": 0.9401408450704225,
-      "eval_f1": 0.8482142857142857,
-      "eval_loss": 0.179457426071167,
-      "eval_precision": 0.8527827648114902,
-      "eval_recall": 0.8436944937833037,
-      "eval_runtime": 2.2185,
-      "eval_samples_per_second": 128.015,
-      "eval_steps_per_second": 16.227,
+      "eval_accuracy": 0.9362676056338028,
+      "eval_f1": 0.8340971585701191,
+      "eval_loss": 0.18467187881469727,
+      "eval_precision": 0.8488805970149254,
+      "eval_recall": 0.8198198198198198,
+      "eval_runtime": 3.083,
+      "eval_samples_per_second": 92.117,
+      "eval_steps_per_second": 11.677,
       "step": 1100
     },
     {
       "epoch": 15.633802816901408,
-      "grad_norm": 0.44636908173561096,
+      "grad_norm": 0.3860970139503479,
       "learning_rate": 0.00010641627543035995,
-      "loss": 0.0414,
+      "loss": 0.0508,
       "step": 1110
     },
     {
       "epoch": 15.774647887323944,
-      "grad_norm": 0.2414419800043106,
+      "grad_norm": 0.3090944290161133,
       "learning_rate": 0.00010537297861241524,
-      "loss": 0.0474,
+      "loss": 0.0505,
       "step": 1120
     },
     {
       "epoch": 15.915492957746478,
-      "grad_norm": 0.3585289716720581,
+      "grad_norm": 0.2525678277015686,
       "learning_rate": 0.00010432968179447054,
-      "loss": 0.0422,
+      "loss": 0.0587,
       "step": 1130
     },
     {
       "epoch": 16.056338028169016,
-      "grad_norm": 0.3467848598957062,
+      "grad_norm": 0.7144160866737366,
       "learning_rate": 0.00010328638497652582,
-      "loss": 0.0512,
+      "loss": 0.0667,
       "step": 1140
     },
     {
       "epoch": 16.197183098591548,
-      "grad_norm": 0.3701542019844055,
+      "grad_norm": 0.30514514446258545,
       "learning_rate": 0.00010224308815858112,
-      "loss": 0.0329,
+      "loss": 0.0479,
       "step": 1150
     },
     {
       "epoch": 16.338028169014084,
-      "grad_norm": 0.4284435510635376,
+      "grad_norm": 0.3105055093765259,
       "learning_rate": 0.00010119979134063642,
-      "loss": 0.0475,
+      "loss": 0.0382,
       "step": 1160
     },
     {
       "epoch": 16.47887323943662,
-      "grad_norm": 0.6009432673454285,
+      "grad_norm": 0.20530584454536438,
       "learning_rate": 0.00010015649452269171,
-      "loss": 0.0467,
+      "loss": 0.044,
       "step": 1170
     },
     {
       "epoch": 16.619718309859156,
-      "grad_norm": 0.45662951469421387,
+      "grad_norm": 0.31594333052635193,
       "learning_rate": 9.911319770474701e-05,
-      "loss": 0.0721,
+      "loss": 0.0418,
       "step": 1180
     },
     {
       "epoch": 16.760563380281692,
-      "grad_norm": 0.617853581905365,
+      "grad_norm": 0.8795649409294128,
       "learning_rate": 9.80699008868023e-05,
-      "loss": 0.0603,
+      "loss": 0.0567,
       "step": 1190
     },
     {
       "epoch": 16.901408450704224,
-      "grad_norm": 0.34839749336242676,
+      "grad_norm": 0.655049741268158,
       "learning_rate": 9.70266040688576e-05,
-      "loss": 0.0488,
+      "loss": 0.0515,
       "step": 1200
     },
     {
       "epoch": 16.901408450704224,
-      "eval_accuracy": 0.9309859154929577,
-      "eval_f1": 0.8253119429590018,
-      "eval_loss": 0.20648732781410217,
-      "eval_precision": 0.8282647584973166,
-      "eval_recall": 0.822380106571936,
-      "eval_runtime": 2.4518,
-      "eval_samples_per_second": 115.834,
-      "eval_steps_per_second": 14.683,
+      "eval_accuracy": 0.9408450704225352,
+      "eval_f1": 0.8469945355191257,
+      "eval_loss": 0.17406770586967468,
+      "eval_precision": 0.856353591160221,
+      "eval_recall": 0.8378378378378378,
+      "eval_runtime": 6.0044,
+      "eval_samples_per_second": 47.299,
+      "eval_steps_per_second": 5.996,
       "step": 1200
     },
     {
       "epoch": 17.04225352112676,
-      "grad_norm": 1.235846996307373,
+      "grad_norm": 0.5397161841392517,
       "learning_rate": 9.59833072509129e-05,
-      "loss": 0.0605,
+      "loss": 0.0345,
       "step": 1210
     },
     {
       "epoch": 17.183098591549296,
-      "grad_norm": 0.442062646150589,
+      "grad_norm": 0.4117436408996582,
       "learning_rate": 9.494001043296818e-05,
-      "loss": 0.0478,
+      "loss": 0.0343,
       "step": 1220
     },
     {
       "epoch": 17.323943661971832,
-      "grad_norm": 0.1856844127178192,
+      "grad_norm": 0.6277669072151184,
       "learning_rate": 9.389671361502347e-05,
       "loss": 0.0447,
       "step": 1230
     },
     {
       "epoch": 17.464788732394368,
-      "grad_norm": 0.3410162627696991,
+      "grad_norm": 0.46144264936447144,
       "learning_rate": 9.285341679707877e-05,
-      "loss": 0.0419,
+      "loss": 0.0593,
       "step": 1240
     },
     {
       "epoch": 17.6056338028169,
-      "grad_norm": 0.34299418330192566,
+      "grad_norm": 0.25574731826782227,
       "learning_rate": 9.181011997913407e-05,
-      "loss": 0.0534,
+      "loss": 0.0556,
       "step": 1250
     },
     {
       "epoch": 17.746478873239436,
-      "grad_norm": 0.7993258833885193,
+      "grad_norm": 0.6915749311447144,
       "learning_rate": 9.076682316118936e-05,
-      "loss": 0.036,
+      "loss": 0.05,
       "step": 1260
     },
     {
       "epoch": 17.887323943661972,
-      "grad_norm": 0.8087213635444641,
+      "grad_norm": 1.0586333274841309,
       "learning_rate": 8.972352634324466e-05,
-      "loss": 0.0569,
+      "loss": 0.058,
       "step": 1270
     },
     {
       "epoch": 18.028169014084508,
-      "grad_norm": 0.48548343777656555,
+      "grad_norm": 0.4597637951374054,
       "learning_rate": 8.868022952529996e-05,
-      "loss": 0.0295,
+      "loss": 0.0559,
       "step": 1280
     },
     {
       "epoch": 18.169014084507044,
-      "grad_norm": 0.6621302962303162,
+      "grad_norm": 0.33155715465545654,
       "learning_rate": 8.763693270735525e-05,
-      "loss": 0.0356,
+      "loss": 0.0466,
       "step": 1290
     },
     {
       "epoch": 18.309859154929576,
-      "grad_norm": 0.19855746626853943,
+      "grad_norm": 1.1297364234924316,
       "learning_rate": 8.659363588941053e-05,
-      "loss": 0.0483,
+      "loss": 0.0489,
       "step": 1300
     },
     {
       "epoch": 18.309859154929576,
-      "eval_accuracy": 0.9376760563380282,
-      "eval_f1": 0.8426666666666667,
-      "eval_loss": 0.197713240981102,
-      "eval_precision": 0.8434163701067615,
-      "eval_recall": 0.8419182948490231,
-      "eval_runtime": 3.0583,
-      "eval_samples_per_second": 92.863,
-      "eval_steps_per_second": 11.771,
+      "eval_accuracy": 0.9461267605633803,
+      "eval_f1": 0.8620378719567178,
+      "eval_loss": 0.17934167385101318,
+      "eval_precision": 0.8628158844765343,
+      "eval_recall": 0.8612612612612612,
+      "eval_runtime": 4.5996,
+      "eval_samples_per_second": 61.744,
+      "eval_steps_per_second": 7.827,
       "step": 1300
     },
     {
       "epoch": 18.450704225352112,
-      "grad_norm": 0.43651339411735535,
+      "grad_norm": 0.4485849142074585,
       "learning_rate": 8.555033907146583e-05,
-      "loss": 0.0245,
+      "loss": 0.0346,
       "step": 1310
     },
     {
       "epoch": 18.591549295774648,
-      "grad_norm": 0.5284798741340637,
+      "grad_norm": 0.43537047505378723,
       "learning_rate": 8.450704225352113e-05,
-      "loss": 0.0493,
+      "loss": 0.0475,
       "step": 1320
     },
     {
       "epoch": 18.732394366197184,
-      "grad_norm": 0.49052557349205017,
+      "grad_norm": 1.3907153606414795,
       "learning_rate": 8.346374543557642e-05,
-      "loss": 0.0291,
+      "loss": 0.053,
       "step": 1330
     },
     {
       "epoch": 18.87323943661972,
-      "grad_norm": 0.3060341477394104,
+      "grad_norm": 0.093208909034729,
       "learning_rate": 8.242044861763172e-05,
-      "loss": 0.0435,
+      "loss": 0.0261,
       "step": 1340
     },
     {
       "epoch": 19.014084507042252,
-      "grad_norm": 0.5241355895996094,
+      "grad_norm": 0.5435425043106079,
       "learning_rate": 8.137715179968701e-05,
-      "loss": 0.0439,
+      "loss": 0.0443,
       "step": 1350
     },
     {
       "epoch": 19.154929577464788,
-      "grad_norm": 0.5462380647659302,
+      "grad_norm": 0.1869378536939621,
       "learning_rate": 8.033385498174231e-05,
-      "loss": 0.0442,
+      "loss": 0.0519,
       "step": 1360
     },
     {
       "epoch": 19.295774647887324,
-      "grad_norm": 0.8547139167785645,
+      "grad_norm": 0.4740372896194458,
       "learning_rate": 7.92905581637976e-05,
-      "loss": 0.0412,
+      "loss": 0.0421,
       "step": 1370
     },
     {
       "epoch": 19.43661971830986,
-      "grad_norm": 0.5264360308647156,
+      "grad_norm": 0.30357396602630615,
       "learning_rate": 7.82472613458529e-05,
-      "loss": 0.041,
+      "loss": 0.0352,
       "step": 1380
     },
     {
       "epoch": 19.577464788732396,
-      "grad_norm": 0.804766833782196,
+      "grad_norm": 0.33824267983436584,
       "learning_rate": 7.720396452790818e-05,
-      "loss": 0.0336,
+      "loss": 0.0382,
       "step": 1390
     },
     {
       "epoch": 19.718309859154928,
-      "grad_norm": 0.5560759902000427,
+      "grad_norm": 0.4455321133136749,
       "learning_rate": 7.616066770996348e-05,
-      "loss": 0.0317,
+      "loss": 0.0339,
       "step": 1400
     },
     {
       "epoch": 19.718309859154928,
-      "eval_accuracy": 0.9369718309859155,
-      "eval_f1": 0.8394618834080717,
-      "eval_loss": 0.20056433975696564,
-      "eval_precision": 0.8478260869565217,
-      "eval_recall": 0.8312611012433393,
-      "eval_runtime": 2.2231,
-      "eval_samples_per_second": 127.75,
-      "eval_steps_per_second": 16.194,
+      "eval_accuracy": 0.9443661971830986,
+      "eval_f1": 0.8568840579710145,
+      "eval_loss": 0.18061767518520355,
+      "eval_precision": 0.8615664845173042,
+      "eval_recall": 0.8522522522522522,
+      "eval_runtime": 5.3442,
+      "eval_samples_per_second": 53.142,
+      "eval_steps_per_second": 6.736,
       "step": 1400
     },
     {
       "epoch": 19.859154929577464,
-      "grad_norm": 0.4625890254974365,
+      "grad_norm": 0.35027405619621277,
       "learning_rate": 7.511737089201878e-05,
-      "loss": 0.045,
+      "loss": 0.0372,
       "step": 1410
     },
     {
       "epoch": 20.0,
-      "grad_norm": 0.5293967127799988,
+      "grad_norm": 0.6206357479095459,
       "learning_rate": 7.407407407407407e-05,
-      "loss": 0.0329,
+      "loss": 0.0517,
       "step": 1420
     },
     {
       "epoch": 20.140845070422536,
-      "grad_norm": 0.27396202087402344,
+      "grad_norm": 0.403296560049057,
       "learning_rate": 7.303077725612937e-05,
-      "loss": 0.0281,
+      "loss": 0.0337,
       "step": 1430
     },
     {
       "epoch": 20.281690140845072,
-      "grad_norm": 1.0957276821136475,
+      "grad_norm": 0.11548905074596405,
       "learning_rate": 7.198748043818467e-05,
-      "loss": 0.0397,
+      "loss": 0.026,
       "step": 1440
     },
     {
       "epoch": 20.422535211267604,
-      "grad_norm": 0.5030433535575867,
+      "grad_norm": 0.21008001267910004,
       "learning_rate": 7.094418362023996e-05,
-      "loss": 0.0384,
+      "loss": 0.045,
       "step": 1450
     },
     {
       "epoch": 20.56338028169014,
-      "grad_norm": 0.05735601857304573,
+      "grad_norm": 0.537829577922821,
       "learning_rate": 6.990088680229526e-05,
-      "loss": 0.028,
+      "loss": 0.0616,
       "step": 1460
     },
     {
       "epoch": 20.704225352112676,
-      "grad_norm": 0.4034988284111023,
+      "grad_norm": 0.7347171306610107,
       "learning_rate": 6.885758998435054e-05,
-      "loss": 0.0387,
+      "loss": 0.0405,
       "step": 1470
     },
     {
       "epoch": 20.845070422535212,
-      "grad_norm": 0.6823456287384033,
+      "grad_norm": 0.26924213767051697,
       "learning_rate": 6.781429316640585e-05,
-      "loss": 0.0334,
+      "loss": 0.0305,
       "step": 1480
     },
     {
       "epoch": 20.985915492957748,
-      "grad_norm": 0.36791157722473145,
+      "grad_norm": 0.20689117908477783,
       "learning_rate": 6.677099634846115e-05,
-      "loss": 0.0413,
+      "loss": 0.0396,
       "step": 1490
     },
     {
       "epoch": 21.12676056338028,
-      "grad_norm": 0.3275921642780304,
+      "grad_norm": 0.9461146593093872,
       "learning_rate": 6.572769953051644e-05,
-      "loss": 0.0411,
+      "loss": 0.0409,
       "step": 1500
     },
     {
       "epoch": 21.12676056338028,
-      "eval_accuracy": 0.9362676056338028,
-      "eval_f1": 0.8367899008115419,
-      "eval_loss": 0.2068268060684204,
-      "eval_precision": 0.8498168498168498,
-      "eval_recall": 0.8241563055062167,
-      "eval_runtime": 2.1811,
-      "eval_samples_per_second": 130.209,
-      "eval_steps_per_second": 16.505,
+      "eval_accuracy": 0.9440140845070423,
+      "eval_f1": 0.8568856885688569,
+      "eval_loss": 0.1784217804670334,
+      "eval_precision": 0.8561151079136691,
+      "eval_recall": 0.8576576576576577,
+      "eval_runtime": 4.9417,
+      "eval_samples_per_second": 57.471,
+      "eval_steps_per_second": 7.285,
       "step": 1500
     },
     {
       "epoch": 21.267605633802816,
-      "grad_norm": 0.7097957134246826,
+      "grad_norm": 0.32038381695747375,
       "learning_rate": 6.468440271257174e-05,
-      "loss": 0.0382,
+      "loss": 0.0378,
       "step": 1510
     },
     {
       "epoch": 21.408450704225352,
-      "grad_norm": 0.36959874629974365,
+      "grad_norm": 0.14526519179344177,
       "learning_rate": 6.364110589462703e-05,
-      "loss": 0.0364,
+      "loss": 0.0325,
       "step": 1520
     },
     {
       "epoch": 21.549295774647888,
-      "grad_norm": 1.0485373735427856,
+      "grad_norm": 0.3381955921649933,
       "learning_rate": 6.259780907668233e-05,
-      "loss": 0.0345,
+      "loss": 0.0397,
       "step": 1530
     },
     {
       "epoch": 21.690140845070424,
-      "grad_norm": 0.35940033197402954,
+      "grad_norm": 0.622305691242218,
       "learning_rate": 6.155451225873761e-05,
-      "loss": 0.0542,
+      "loss": 0.0515,
       "step": 1540
     },
     {
       "epoch": 21.830985915492956,
-      "grad_norm": 0.34279265999794006,
+      "grad_norm": 0.5114145874977112,
       "learning_rate": 6.0511215440792915e-05,
-      "loss": 0.039,
+      "loss": 0.0353,
       "step": 1550
     },
     {
       "epoch": 21.971830985915492,
-      "grad_norm": 0.4195212721824646,
+      "grad_norm": 0.2956676781177521,
       "learning_rate": 5.9467918622848205e-05,
-      "loss": 0.0477,
+      "loss": 0.0391,
       "step": 1560
     },
     {
       "epoch": 22.112676056338028,
-      "grad_norm": 0.4848310351371765,
+      "grad_norm": 0.44752395153045654,
       "learning_rate": 5.84246218049035e-05,
-      "loss": 0.0413,
+      "loss": 0.037,
       "step": 1570
     },
     {
       "epoch": 22.253521126760564,
-      "grad_norm": 0.31897225975990295,
+      "grad_norm": 0.633104681968689,
       "learning_rate": 5.73813249869588e-05,
-      "loss": 0.0425,
+      "loss": 0.0316,
       "step": 1580
     },
     {
       "epoch": 22.3943661971831,
-      "grad_norm": 0.7236142158508301,
+      "grad_norm": 0.16630919277668,
       "learning_rate": 5.633802816901409e-05,
-      "loss": 0.0467,
+      "loss": 0.028,
       "step": 1590
     },
     {
       "epoch": 22.535211267605632,
-      "grad_norm": 0.5363371968269348,
+      "grad_norm": 0.24732300639152527,
       "learning_rate": 5.529473135106938e-05,
-      "loss": 0.0512,
+      "loss": 0.0275,
       "step": 1600
     },
     {
       "epoch": 22.535211267605632,
-      "eval_accuracy": 0.9390845070422535,
-      "eval_f1": 0.8445642407906558,
-      "eval_loss": 0.20564186573028564,
-      "eval_precision": 0.8545454545454545,
-      "eval_recall": 0.8348134991119005,
-      "eval_runtime": 3.113,
-      "eval_samples_per_second": 91.231,
-      "eval_steps_per_second": 11.564,
+      "eval_accuracy": 0.9436619718309859,
+      "eval_f1": 0.8548094373865699,
+      "eval_loss": 0.18389663100242615,
+      "eval_precision": 0.8610603290676416,
+      "eval_recall": 0.8486486486486486,
+      "eval_runtime": 2.0239,
+      "eval_samples_per_second": 140.322,
+      "eval_steps_per_second": 17.787,
       "step": 1600
     },
     {
       "epoch": 22.676056338028168,
-      "grad_norm": 0.33758053183555603,
+      "grad_norm": 0.7257568836212158,
       "learning_rate": 5.425143453312468e-05,
-      "loss": 0.0357,
+      "loss": 0.025,
       "step": 1610
     },
     {
       "epoch": 22.816901408450704,
-      "grad_norm": 0.27382269501686096,
+      "grad_norm": 0.22960269451141357,
       "learning_rate": 5.3208137715179974e-05,
-      "loss": 0.0313,
+      "loss": 0.0197,
       "step": 1620
     },
     {
       "epoch": 22.95774647887324,
-      "grad_norm": 0.3563465476036072,
+      "grad_norm": 0.5778278112411499,
       "learning_rate": 5.216484089723527e-05,
-      "loss": 0.0297,
+      "loss": 0.0304,
       "step": 1630
     },
     {
       "epoch": 23.098591549295776,
-      "grad_norm": 0.2593757212162018,
+      "grad_norm": 0.2570374608039856,
       "learning_rate": 5.112154407929056e-05,
-      "loss": 0.0258,
+      "loss": 0.0314,
       "step": 1640
     },
     {
       "epoch": 23.239436619718308,
-      "grad_norm": 0.28875812888145447,
+      "grad_norm": 0.9027012586593628,
       "learning_rate": 5.0078247261345856e-05,
-      "loss": 0.0272,
+      "loss": 0.0323,
       "step": 1650
     },
     {
       "epoch": 23.380281690140844,
-      "grad_norm": 0.7140405178070068,
+      "grad_norm": 0.14204645156860352,
       "learning_rate": 4.903495044340115e-05,
-      "loss": 0.0237,
+      "loss": 0.0248,
       "step": 1660
     },
     {
       "epoch": 23.52112676056338,
-      "grad_norm": 0.19595526158809662,
+      "grad_norm": 1.2641935348510742,
       "learning_rate": 4.799165362545645e-05,
-      "loss": 0.0215,
+      "loss": 0.0279,
       "step": 1670
     },
     {
       "epoch": 23.661971830985916,
-      "grad_norm": 0.3630121052265167,
+      "grad_norm": 0.4698958098888397,
       "learning_rate": 4.694835680751174e-05,
-      "loss": 0.0276,
+      "loss": 0.0414,
       "step": 1680
     },
     {
       "epoch": 23.802816901408452,
-      "grad_norm": 0.531328558921814,
+      "grad_norm": 0.35781243443489075,
       "learning_rate": 4.5905059989567033e-05,
-      "loss": 0.0289,
+      "loss": 0.0297,
       "step": 1690
     },
     {
       "epoch": 23.943661971830984,
-      "grad_norm": 0.039958756417036057,
+      "grad_norm": 0.21168813109397888,
       "learning_rate": 4.486176317162233e-05,
-      "loss": 0.0329,
+      "loss": 0.0231,
       "step": 1700
     },
     {
       "epoch": 23.943661971830984,
-      "eval_accuracy": 0.9338028169014084,
-      "eval_f1": 0.8294010889292196,
-      "eval_loss": 0.21274171769618988,
-      "eval_precision": 0.8478664192949907,
-      "eval_recall": 0.8117229129662522,
-      "eval_runtime": 2.4843,
-      "eval_samples_per_second": 114.317,
-      "eval_steps_per_second": 14.491,
+      "eval_accuracy": 0.9415492957746479,
+      "eval_f1": 0.847985347985348,
+      "eval_loss": 0.18647325038909912,
+      "eval_precision": 0.8621973929236499,
+      "eval_recall": 0.8342342342342343,
+      "eval_runtime": 2.0446,
+      "eval_samples_per_second": 138.905,
+      "eval_steps_per_second": 17.608,
       "step": 1700
     },
     {
       "epoch": 24.08450704225352,
-      "grad_norm": 0.15843306481838226,
+      "grad_norm": 0.10770827531814575,
       "learning_rate": 4.3818466353677626e-05,
-      "loss": 0.0217,
+      "loss": 0.0287,
       "step": 1710
     },
     {
       "epoch": 24.225352112676056,
-      "grad_norm": 0.12869732081890106,
+      "grad_norm": 0.4609077274799347,
       "learning_rate": 4.2775169535732915e-05,
-      "loss": 0.0227,
+      "loss": 0.0233,
       "step": 1720
     },
     {
       "epoch": 24.366197183098592,
-      "grad_norm": 0.08577102422714233,
+      "grad_norm": 0.25236785411834717,
       "learning_rate": 4.173187271778821e-05,
-      "loss": 0.0249,
+      "loss": 0.0217,
       "step": 1730
     },
     {
       "epoch": 24.507042253521128,
-      "grad_norm": 0.2362302988767624,
+      "grad_norm": 0.5101929903030396,
       "learning_rate": 4.068857589984351e-05,
-      "loss": 0.0286,
+      "loss": 0.034,
       "step": 1740
     },
     {
       "epoch": 24.647887323943664,
-      "grad_norm": 0.24027810990810394,
+      "grad_norm": 0.47603026032447815,
       "learning_rate": 3.96452790818988e-05,
-      "loss": 0.0248,
+      "loss": 0.0259,
       "step": 1750
     },
     {
       "epoch": 24.788732394366196,
-      "grad_norm": 0.790462076663971,
+      "grad_norm": 0.27863284945487976,
       "learning_rate": 3.860198226395409e-05,
-      "loss": 0.0263,
+      "loss": 0.0238,
       "step": 1760
     },
     {
       "epoch": 24.929577464788732,
-      "grad_norm": 0.16294124722480774,
+      "grad_norm": 0.3325282633304596,
       "learning_rate": 3.755868544600939e-05,
-      "loss": 0.0442,
+      "loss": 0.0361,
       "step": 1770
     },
     {
       "epoch": 25.070422535211268,
-      "grad_norm": 0.4505746066570282,
+      "grad_norm": 0.05602555721998215,
       "learning_rate": 3.6515388628064685e-05,
-      "loss": 0.0256,
+      "loss": 0.0257,
       "step": 1780
     },
     {
       "epoch": 25.211267605633804,
-      "grad_norm": 0.650375485420227,
+      "grad_norm": 0.447354257106781,
       "learning_rate": 3.547209181011998e-05,
-      "loss": 0.0228,
+      "loss": 0.0223,
       "step": 1790
     },
     {
       "epoch": 25.352112676056336,
-      "grad_norm": 0.1388479620218277,
+      "grad_norm": 0.6506637334823608,
       "learning_rate": 3.442879499217527e-05,
-      "loss": 0.0197,
+      "loss": 0.0204,
       "step": 1800
     },
     {
       "epoch": 25.352112676056336,
-      "eval_accuracy": 0.9334507042253521,
-      "eval_f1": 0.828649138712602,
-      "eval_loss": 0.2122490257024765,
-      "eval_precision": 0.8462962962962963,
-      "eval_recall": 0.8117229129662522,
-      "eval_runtime": 2.1705,
-      "eval_samples_per_second": 130.848,
-      "eval_steps_per_second": 16.586,
+      "eval_accuracy": 0.9404929577464789,
+      "eval_f1": 0.8481581311769991,
+      "eval_loss": 0.18843932449817657,
+      "eval_precision": 0.8458781362007168,
+      "eval_recall": 0.8504504504504504,
+      "eval_runtime": 2.5356,
+      "eval_samples_per_second": 112.004,
+      "eval_steps_per_second": 14.198,
       "step": 1800
     },
     {
       "epoch": 25.492957746478872,
-      "grad_norm": 0.27917221188545227,
+      "grad_norm": 0.920504629611969,
       "learning_rate": 3.338549817423057e-05,
-      "loss": 0.0169,
+      "loss": 0.0348,
       "step": 1810
     },
     {
       "epoch": 25.633802816901408,
-      "grad_norm": 0.3449484407901764,
+      "grad_norm": 0.6100931763648987,
       "learning_rate": 3.234220135628587e-05,
-      "loss": 0.0363,
+      "loss": 0.0364,
       "step": 1820
     },
     {
       "epoch": 25.774647887323944,
-      "grad_norm": 0.1575915664434433,
+      "grad_norm": 0.3321876525878906,
       "learning_rate": 3.1298904538341165e-05,
-      "loss": 0.0397,
+      "loss": 0.0237,
       "step": 1830
     },
     {
       "epoch": 25.91549295774648,
-      "grad_norm": 0.2630869150161743,
+      "grad_norm": 0.14392893016338348,
       "learning_rate": 3.0255607720396458e-05,
-      "loss": 0.0257,
+      "loss": 0.0292,
       "step": 1840
     },
     {
       "epoch": 26.056338028169016,
-      "grad_norm": 0.36075156927108765,
+      "grad_norm": 0.2463379055261612,
       "learning_rate": 2.921231090245175e-05,
-      "loss": 0.0248,
+      "loss": 0.0227,
       "step": 1850
     },
     {
       "epoch": 26.197183098591548,
-      "grad_norm": 0.1405423879623413,
+      "grad_norm": 0.13013498485088348,
       "learning_rate": 2.8169014084507046e-05,
-      "loss": 0.0248,
+      "loss": 0.0238,
       "step": 1860
     },
     {
       "epoch": 26.338028169014084,
-      "grad_norm": 0.23486283421516418,
+      "grad_norm": 0.11781849712133408,
       "learning_rate": 2.712571726656234e-05,
-      "loss": 0.0229,
+      "loss": 0.0348,
       "step": 1870
     },
     {
       "epoch": 26.47887323943662,
-      "grad_norm": 0.47251322865486145,
+      "grad_norm": 0.19677963852882385,
       "learning_rate": 2.6082420448617635e-05,
-      "loss": 0.0186,
+      "loss": 0.0222,
       "step": 1880
     },
     {
       "epoch": 26.619718309859156,
-      "grad_norm": 0.14794085919857025,
+      "grad_norm": 0.32697793841362,
       "learning_rate": 2.5039123630672928e-05,
-      "loss": 0.0214,
+      "loss": 0.0178,
       "step": 1890
     },
     {
       "epoch": 26.760563380281692,
-      "grad_norm": 0.1104566678404808,
+      "grad_norm": 0.32760560512542725,
       "learning_rate": 2.3995826812728224e-05,
-      "loss": 0.0316,
+      "loss": 0.0245,
       "step": 1900
     },
     {
       "epoch": 26.760563380281692,
-      "eval_accuracy": 0.9373239436619718,
-      "eval_f1": 0.8399280575539568,
-      "eval_loss": 0.20504821836948395,
-      "eval_precision": 0.8506375227686703,
-      "eval_recall": 0.8294849023090586,
-      "eval_runtime": 2.1762,
-      "eval_samples_per_second": 130.502,
-      "eval_steps_per_second": 16.543,
+      "eval_accuracy": 0.9376760563380282,
+      "eval_f1": 0.8409703504043127,
+      "eval_loss": 0.19348740577697754,
+      "eval_precision": 0.8387096774193549,
+      "eval_recall": 0.8432432432432433,
+      "eval_runtime": 3.7052,
+      "eval_samples_per_second": 76.649,
+      "eval_steps_per_second": 9.716,
       "step": 1900
     },
     {
       "epoch": 26.901408450704224,
-      "grad_norm": 0.35611000657081604,
+      "grad_norm": 0.23730838298797607,
       "learning_rate": 2.2952529994783517e-05,
-      "loss": 0.028,
+      "loss": 0.0246,
       "step": 1910
     },
     {
       "epoch": 27.04225352112676,
-      "grad_norm": 0.31670650839805603,
+      "grad_norm": 1.0503803491592407,
       "learning_rate": 2.1909233176838813e-05,
-      "loss": 0.0186,
+      "loss": 0.0235,
       "step": 1920
     },
     {
       "epoch": 27.183098591549296,
-      "grad_norm": 0.17706644535064697,
+      "grad_norm": 0.8191851377487183,
       "learning_rate": 2.0865936358894105e-05,
-      "loss": 0.0153,
+      "loss": 0.0158,
       "step": 1930
     },
     {
       "epoch": 27.323943661971832,
-      "grad_norm": 0.3245967626571655,
+      "grad_norm": 0.2159469872713089,
       "learning_rate": 1.98226395409494e-05,
-      "loss": 0.0189,
+      "loss": 0.0123,
       "step": 1940
     },
     {
       "epoch": 27.464788732394368,
-      "grad_norm": 0.24404743313789368,
+      "grad_norm": 0.11513730883598328,
       "learning_rate": 1.8779342723004694e-05,
-      "loss": 0.016,
+      "loss": 0.0202,
       "step": 1950
     },
     {
       "epoch": 27.6056338028169,
-      "grad_norm": 0.5631425380706787,
+      "grad_norm": 0.43825531005859375,
       "learning_rate": 1.773604590505999e-05,
-      "loss": 0.0347,
+      "loss": 0.0189,
       "step": 1960
     },
     {
       "epoch": 27.746478873239436,
-      "grad_norm": 0.18322208523750305,
+      "grad_norm": 0.6861417293548584,
       "learning_rate": 1.6692749087115286e-05,
-      "loss": 0.0224,
+      "loss": 0.0293,
       "step": 1970
     },
     {
       "epoch": 27.887323943661972,
-      "grad_norm": 0.26571351289749146,
+      "grad_norm": 0.4493369162082672,
       "learning_rate": 1.5649452269170582e-05,
-      "loss": 0.0136,
+      "loss": 0.0169,
       "step": 1980
     },
     {
       "epoch": 28.028169014084508,
-      "grad_norm": 0.1630268394947052,
+      "grad_norm": 0.07595561444759369,
       "learning_rate": 1.4606155451225875e-05,
-      "loss": 0.0234,
+      "loss": 0.0284,
       "step": 1990
     },
     {
       "epoch": 28.169014084507044,
-      "grad_norm": 0.20336656272411346,
+      "grad_norm": 0.29891517758369446,
       "learning_rate": 1.356285863328117e-05,
-      "loss": 0.0133,
+      "loss": 0.0202,
       "step": 2000
     },
     {
       "epoch": 28.169014084507044,
-      "eval_accuracy": 0.9408450704225352,
-      "eval_f1": 0.8494623655913979,
-      "eval_loss": 0.20187228918075562,
-      "eval_precision": 0.8571428571428571,
-      "eval_recall": 0.8419182948490231,
-      "eval_runtime": 2.7007,
-      "eval_samples_per_second": 105.158,
-      "eval_steps_per_second": 13.33,
+      "eval_accuracy": 0.9394366197183098,
+      "eval_f1": 0.8456014362657092,
+      "eval_loss": 0.18883755803108215,
+      "eval_precision": 0.8425760286225402,
+      "eval_recall": 0.8486486486486486,
+      "eval_runtime": 2.034,
+      "eval_samples_per_second": 139.628,
+      "eval_steps_per_second": 17.699,
       "step": 2000
     },
     {
       "epoch": 28.309859154929576,
-      "grad_norm": 0.20230424404144287,
+      "grad_norm": 0.08975574374198914,
       "learning_rate": 1.2519561815336464e-05,
-      "loss": 0.0206,
+      "loss": 0.0133,
       "step": 2010
     },
     {
       "epoch": 28.450704225352112,
-      "grad_norm": 0.5370287299156189,
+      "grad_norm": 0.14954060316085815,
       "learning_rate": 1.1476264997391758e-05,
-      "loss": 0.0327,
+      "loss": 0.0155,
       "step": 2020
     },
     {
       "epoch": 28.591549295774648,
-      "grad_norm": 0.2412065863609314,
+      "grad_norm": 0.32284414768218994,
       "learning_rate": 1.0432968179447053e-05,
-      "loss": 0.0101,
+      "loss": 0.0213,
       "step": 2030
     },
     {
       "epoch": 28.732394366197184,
-      "grad_norm": 0.49697521328926086,
+      "grad_norm": 0.3323783278465271,
       "learning_rate": 9.389671361502347e-06,
-      "loss": 0.0193,
+      "loss": 0.0186,
       "step": 2040
     },
     {
       "epoch": 28.87323943661972,
-      "grad_norm": 0.17303723096847534,
+      "grad_norm": 0.22159124910831451,
       "learning_rate": 8.346374543557643e-06,
-      "loss": 0.0167,
+      "loss": 0.0298,
       "step": 2050
     },
     {
       "epoch": 29.014084507042252,
-      "grad_norm": 0.311748206615448,
+      "grad_norm": 0.13430196046829224,
       "learning_rate": 7.303077725612938e-06,
-      "loss": 0.0166,
+      "loss": 0.0156,
       "step": 2060
     },
     {
       "epoch": 29.154929577464788,
-      "grad_norm": 0.1465146541595459,
+      "grad_norm": 0.19941723346710205,
       "learning_rate": 6.259780907668232e-06,
-      "loss": 0.0252,
+      "loss": 0.0144,
       "step": 2070
     },
     {
       "epoch": 29.295774647887324,
-      "grad_norm": 0.1742701530456543,
+      "grad_norm": 0.049271322786808014,
       "learning_rate": 5.216484089723526e-06,
-      "loss": 0.0249,
+      "loss": 0.0243,
       "step": 2080
     },
     {
       "epoch": 29.43661971830986,
-      "grad_norm": 0.12084958702325821,
+      "grad_norm": 0.5687671899795532,
       "learning_rate": 4.173187271778822e-06,
-      "loss": 0.0219,
+      "loss": 0.0143,
       "step": 2090
     },
     {
       "epoch": 29.577464788732396,
-      "grad_norm": 0.1655401885509491,
+      "grad_norm": 0.10554461926221848,
       "learning_rate": 3.129890453834116e-06,
-      "loss": 0.0181,
+      "loss": 0.0187,
       "step": 2100
     },
     {
       "epoch": 29.577464788732396,
-      "eval_accuracy": 0.9401408450704225,
-      "eval_f1": 0.8473967684021544,
-      "eval_loss": 0.20311439037322998,
-      "eval_precision": 0.8566243194192378,
-      "eval_recall": 0.8383658969804618,
-      "eval_runtime": 2.8967,
-      "eval_samples_per_second": 98.042,
-      "eval_steps_per_second": 12.428,
+      "eval_accuracy": 0.9415492957746479,
+      "eval_f1": 0.8501805054151624,
+      "eval_loss": 0.19139410555362701,
+      "eval_precision": 0.8517179023508138,
+      "eval_recall": 0.8486486486486486,
+      "eval_runtime": 2.1198,
+      "eval_samples_per_second": 133.977,
+      "eval_steps_per_second": 16.983,
       "step": 2100
     },
     {
       "epoch": 29.718309859154928,
-      "grad_norm": 0.2921566963195801,
+      "grad_norm": 0.6691368818283081,
       "learning_rate": 2.086593635889411e-06,
-      "loss": 0.0208,
+      "loss": 0.0132,
       "step": 2110
     },
     {
       "epoch": 29.859154929577464,
-      "grad_norm": 0.12771935760974884,
+      "grad_norm": 0.6403772234916687,
       "learning_rate": 1.0432968179447054e-06,
-      "loss": 0.014,
+      "loss": 0.0282,
       "step": 2120
     },
     {
       "epoch": 30.0,
-      "grad_norm": 0.040317874401807785,
+      "grad_norm": 0.040005020797252655,
       "learning_rate": 0.0,
-      "loss": 0.019,
+      "loss": 0.0234,
       "step": 2130
     },
     {
       "epoch": 30.0,
       "step": 2130,
       "total_flos": 2.6364665116584346e+18,
-      "train_loss": 0.08667128345636135,
-      "train_runtime": 602.5463,
-      "train_samples_per_second": 56.46,
-      "train_steps_per_second": 3.535
+      "train_loss": 0.08832058483656023,
+      "train_runtime": 625.5786,
+      "train_samples_per_second": 54.382,
+      "train_steps_per_second": 3.405
     }
   ],
   "logging_steps": 10,