End of training

README.md

@@ -22,7 +22,7 @@ model-index:
     metrics:
     - name: Accuracy
       type: accuracy
-      value: 0.8169014084507042
+      value: 0.8450704225352113
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -32,8 +32,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [facebook/deit-base-distilled-patch16-224](https://huggingface.co/facebook/deit-base-distilled-patch16-224) on the imagefolder dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.5748
-- Accuracy: 0.8169
+- Loss: 0.5367
+- Accuracy: 0.8451
 
 ## Model description
 

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 92.3076923076923,
-    "eval_accuracy": 0.8169014084507042,
-    "eval_loss": 0.7217584848403931,
-    "eval_runtime": 0.9525,
-    "eval_samples_per_second": 74.537,
-    "eval_steps_per_second": 3.149,
+    "eval_accuracy": 0.8450704225352113,
+    "eval_loss": 0.5367295145988464,
+    "eval_runtime": 0.9601,
+    "eval_samples_per_second": 73.951,
+    "eval_steps_per_second": 3.125,
     "total_flos": 2.8402872494292173e+18,
-    "train_loss": 0.23628523468971252,
-    "train_runtime": 1593.1864,
-    "train_samples_per_second": 24.919,
+    "train_loss": 0.25812069336573285,
+    "train_runtime": 1594.5895,
+    "train_samples_per_second": 24.897,
     "train_steps_per_second": 0.188
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 92.3076923076923,
-    "eval_accuracy": 0.8169014084507042,
-    "eval_loss": 0.7217584848403931,
-    "eval_runtime": 0.9525,
-    "eval_samples_per_second": 74.537,
-    "eval_steps_per_second": 3.149
+    "eval_accuracy": 0.8450704225352113,
+    "eval_loss": 0.5367295145988464,
+    "eval_runtime": 0.9601,
+    "eval_samples_per_second": 73.951,
+    "eval_steps_per_second": 3.125
 }

runs/May22_07-20-17_defc3061f818/events.out.tfevents.1716364021.defc3061f818.1801.6

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2bb43c9d51362498731268ebde1ae01eb483e439b1be7494ec618e47185698c
+size 411

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 92.3076923076923,
     "total_flos": 2.8402872494292173e+18,
-    "train_loss": 0.23628523468971252,
-    "train_runtime": 1593.1864,
-    "train_samples_per_second": 24.919,
+    "train_loss": 0.25812069336573285,
+    "train_runtime": 1594.5895,
+    "train_samples_per_second": 24.897,
     "train_steps_per_second": 0.188
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.8169014084507042,
-  "best_model_checkpoint": "deit-base-distilled-patch16-224-65-fold4/checkpoint-91",
+  "best_metric": 0.8450704225352113,
+  "best_model_checkpoint": "deit-base-distilled-patch16-224-65-fold4/checkpoint-182",
   "epoch": 92.3076923076923,
   "eval_steps": 500,
   "global_step": 300,
@@ -10,1067 +10,1067 @@
   "log_history": [
     {
       "epoch": 0.9230769230769231,
-      "eval_accuracy": 0.4647887323943662,
-      "eval_loss": 0.7266488075256348,
-      "eval_runtime": 0.8434,
-      "eval_samples_per_second": 84.184,
-      "eval_steps_per_second": 3.557,
+      "eval_accuracy": 0.5211267605633803,
+      "eval_loss": 0.7144731879234314,
+      "eval_runtime": 1.0188,
+      "eval_samples_per_second": 69.692,
+      "eval_steps_per_second": 2.945,
       "step": 3
     },
     {
       "epoch": 1.8461538461538463,
-      "eval_accuracy": 0.5211267605633803,
-      "eval_loss": 0.8115941286087036,
-      "eval_runtime": 0.8919,
-      "eval_samples_per_second": 79.608,
-      "eval_steps_per_second": 3.364,
+      "eval_accuracy": 0.5070422535211268,
+      "eval_loss": 0.7082186341285706,
+      "eval_runtime": 0.9339,
+      "eval_samples_per_second": 76.023,
+      "eval_steps_per_second": 3.212,
       "step": 6
     },
     {
       "epoch": 2.769230769230769,
-      "eval_accuracy": 0.4647887323943662,
-      "eval_loss": 0.7081143856048584,
-      "eval_runtime": 0.8712,
-      "eval_samples_per_second": 81.496,
-      "eval_steps_per_second": 3.443,
+      "eval_accuracy": 0.6056338028169014,
+      "eval_loss": 0.6889063715934753,
+      "eval_runtime": 0.9191,
+      "eval_samples_per_second": 77.245,
+      "eval_steps_per_second": 3.264,
       "step": 9
     },
     {
       "epoch": 3.076923076923077,
-      "grad_norm": 2.1958322525024414,
+      "grad_norm": 3.284947633743286,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.7173,
+      "loss": 0.6878,
       "step": 10
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.5633802816901409,
-      "eval_loss": 0.6644501090049744,
-      "eval_runtime": 0.8954,
-      "eval_samples_per_second": 79.298,
-      "eval_steps_per_second": 3.351,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.6703165173530579,
+      "eval_runtime": 0.9363,
+      "eval_samples_per_second": 75.83,
+      "eval_steps_per_second": 3.204,
       "step": 13
     },
     {
       "epoch": 4.923076923076923,
-      "eval_accuracy": 0.5915492957746479,
-      "eval_loss": 0.6440630555152893,
-      "eval_runtime": 0.8941,
-      "eval_samples_per_second": 79.413,
-      "eval_steps_per_second": 3.355,
+      "eval_accuracy": 0.676056338028169,
+      "eval_loss": 0.6556025743484497,
+      "eval_runtime": 0.9344,
+      "eval_samples_per_second": 75.984,
+      "eval_steps_per_second": 3.211,
       "step": 16
     },
     {
       "epoch": 5.846153846153846,
-      "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.6400186419487,
-      "eval_runtime": 0.933,
-      "eval_samples_per_second": 76.096,
-      "eval_steps_per_second": 3.215,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.642956554889679,
+      "eval_runtime": 0.9474,
+      "eval_samples_per_second": 74.941,
+      "eval_steps_per_second": 3.167,
       "step": 19
     },
     {
       "epoch": 6.153846153846154,
-      "grad_norm": 2.7608325481414795,
+      "grad_norm": 5.985997676849365,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.6351,
+      "loss": 0.6203,
       "step": 20
     },
     {
       "epoch": 6.769230769230769,
-      "eval_accuracy": 0.6619718309859155,
-      "eval_loss": 0.6054678559303284,
-      "eval_runtime": 0.9387,
-      "eval_samples_per_second": 75.638,
-      "eval_steps_per_second": 3.196,
+      "eval_accuracy": 0.676056338028169,
+      "eval_loss": 0.6249927282333374,
+      "eval_runtime": 0.9539,
+      "eval_samples_per_second": 74.435,
+      "eval_steps_per_second": 3.145,
       "step": 22
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.5352112676056338,
-      "eval_loss": 0.7769902944564819,
-      "eval_runtime": 0.9329,
-      "eval_samples_per_second": 76.108,
-      "eval_steps_per_second": 3.216,
+      "eval_accuracy": 0.6197183098591549,
+      "eval_loss": 0.7463871240615845,
+      "eval_runtime": 0.9471,
+      "eval_samples_per_second": 74.963,
+      "eval_steps_per_second": 3.167,
       "step": 26
     },
     {
       "epoch": 8.923076923076923,
-      "eval_accuracy": 0.6901408450704225,
-      "eval_loss": 0.6259447932243347,
-      "eval_runtime": 0.924,
-      "eval_samples_per_second": 76.842,
-      "eval_steps_per_second": 3.247,
+      "eval_accuracy": 0.6056338028169014,
+      "eval_loss": 0.6646994352340698,
+      "eval_runtime": 0.9245,
+      "eval_samples_per_second": 76.797,
+      "eval_steps_per_second": 3.245,
       "step": 29
     },
     {
       "epoch": 9.23076923076923,
-      "grad_norm": 5.699825286865234,
+      "grad_norm": 7.882161617279053,
       "learning_rate": 5e-05,
-      "loss": 0.5434,
+      "loss": 0.5703,
       "step": 30
     },
     {
       "epoch": 9.846153846153847,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.5889422297477722,
-      "eval_runtime": 0.9148,
-      "eval_samples_per_second": 77.612,
-      "eval_steps_per_second": 3.279,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.6096706390380859,
+      "eval_runtime": 0.9619,
+      "eval_samples_per_second": 73.809,
+      "eval_steps_per_second": 3.119,
       "step": 32
     },
     {
       "epoch": 10.76923076923077,
-      "eval_accuracy": 0.647887323943662,
-      "eval_loss": 0.7283326387405396,
-      "eval_runtime": 0.9062,
-      "eval_samples_per_second": 78.347,
-      "eval_steps_per_second": 3.31,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.6261224746704102,
+      "eval_runtime": 0.9396,
+      "eval_samples_per_second": 75.56,
+      "eval_steps_per_second": 3.193,
       "step": 35
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.647887323943662,
-      "eval_loss": 0.6897829174995422,
-      "eval_runtime": 0.9341,
-      "eval_samples_per_second": 76.013,
-      "eval_steps_per_second": 3.212,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.5925809741020203,
+      "eval_runtime": 0.9277,
+      "eval_samples_per_second": 76.537,
+      "eval_steps_per_second": 3.234,
       "step": 39
     },
     {
       "epoch": 12.307692307692308,
-      "grad_norm": 2.502607583999634,
+      "grad_norm": 5.672287940979004,
       "learning_rate": 4.814814814814815e-05,
-      "loss": 0.4861,
+      "loss": 0.5281,
       "step": 40
     },
     {
       "epoch": 12.923076923076923,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.6428806781768799,
-      "eval_runtime": 0.9382,
-      "eval_samples_per_second": 75.679,
-      "eval_steps_per_second": 3.198,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.5370251536369324,
+      "eval_runtime": 0.9419,
+      "eval_samples_per_second": 75.378,
+      "eval_steps_per_second": 3.185,
       "step": 42
     },
     {
       "epoch": 13.846153846153847,
-      "eval_accuracy": 0.6619718309859155,
-      "eval_loss": 0.691542387008667,
-      "eval_runtime": 0.932,
-      "eval_samples_per_second": 76.179,
-      "eval_steps_per_second": 3.219,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.5637533664703369,
+      "eval_runtime": 0.9264,
+      "eval_samples_per_second": 76.638,
+      "eval_steps_per_second": 3.238,
       "step": 45
     },
     {
       "epoch": 14.76923076923077,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.5702247619628906,
-      "eval_runtime": 0.9242,
-      "eval_samples_per_second": 76.822,
-      "eval_steps_per_second": 3.246,
+      "eval_accuracy": 0.6056338028169014,
+      "eval_loss": 0.7175151109695435,
+      "eval_runtime": 0.9684,
+      "eval_samples_per_second": 73.32,
+      "eval_steps_per_second": 3.098,
       "step": 48
     },
     {
       "epoch": 15.384615384615385,
-      "grad_norm": 5.8594255447387695,
+      "grad_norm": 7.855106830596924,
       "learning_rate": 4.62962962962963e-05,
-      "loss": 0.4285,
+      "loss": 0.4616,
       "step": 50
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.6356058120727539,
-      "eval_runtime": 0.919,
-      "eval_samples_per_second": 77.258,
-      "eval_steps_per_second": 3.264,
+      "eval_accuracy": 0.5774647887323944,
+      "eval_loss": 0.8917320966720581,
+      "eval_runtime": 0.9446,
+      "eval_samples_per_second": 75.161,
+      "eval_steps_per_second": 3.176,
       "step": 52
     },
     {
       "epoch": 16.923076923076923,
       "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.6981013417243958,
-      "eval_runtime": 0.9247,
-      "eval_samples_per_second": 76.785,
-      "eval_steps_per_second": 3.244,
+      "eval_loss": 0.6760779023170471,
+      "eval_runtime": 0.9228,
+      "eval_samples_per_second": 76.943,
+      "eval_steps_per_second": 3.251,
       "step": 55
     },
     {
       "epoch": 17.846153846153847,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.5218324661254883,
-      "eval_runtime": 0.917,
-      "eval_samples_per_second": 77.425,
-      "eval_steps_per_second": 3.271,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.5606212019920349,
+      "eval_runtime": 0.9447,
+      "eval_samples_per_second": 75.157,
+      "eval_steps_per_second": 3.176,
       "step": 58
     },
     {
       "epoch": 18.46153846153846,
-      "grad_norm": 6.547220706939697,
+      "grad_norm": 5.475687503814697,
       "learning_rate": 4.4444444444444447e-05,
-      "loss": 0.3781,
+      "loss": 0.4943,
       "step": 60
     },
     {
       "epoch": 18.76923076923077,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.534016489982605,
-      "eval_runtime": 0.914,
-      "eval_samples_per_second": 77.678,
-      "eval_steps_per_second": 3.282,
+      "eval_accuracy": 0.6338028169014085,
+      "eval_loss": 0.696312427520752,
+      "eval_runtime": 0.9291,
+      "eval_samples_per_second": 76.418,
+      "eval_steps_per_second": 3.229,
       "step": 61
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.7611135244369507,
-      "eval_runtime": 0.9206,
-      "eval_samples_per_second": 77.126,
-      "eval_steps_per_second": 3.259,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.6461552977561951,
+      "eval_runtime": 0.9441,
+      "eval_samples_per_second": 75.207,
+      "eval_steps_per_second": 3.178,
       "step": 65
     },
     {
       "epoch": 20.923076923076923,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.5939193964004517,
-      "eval_runtime": 0.9415,
-      "eval_samples_per_second": 75.408,
-      "eval_steps_per_second": 3.186,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.6246035695075989,
+      "eval_runtime": 0.9435,
+      "eval_samples_per_second": 75.251,
+      "eval_steps_per_second": 3.18,
       "step": 68
     },
     {
       "epoch": 21.53846153846154,
-      "grad_norm": 4.307927131652832,
+      "grad_norm": 4.991897106170654,
       "learning_rate": 4.259259259259259e-05,
-      "loss": 0.3516,
+      "loss": 0.4058,
       "step": 70
     },
     {
       "epoch": 21.846153846153847,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.61859130859375,
-      "eval_runtime": 0.9317,
-      "eval_samples_per_second": 76.201,
-      "eval_steps_per_second": 3.22,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.7336124181747437,
+      "eval_runtime": 0.9352,
+      "eval_samples_per_second": 75.921,
+      "eval_steps_per_second": 3.208,
       "step": 71
     },
     {
       "epoch": 22.76923076923077,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.712211549282074,
-      "eval_runtime": 0.9272,
-      "eval_samples_per_second": 76.571,
-      "eval_steps_per_second": 3.235,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.6270064115524292,
+      "eval_runtime": 0.9384,
+      "eval_samples_per_second": 75.658,
+      "eval_steps_per_second": 3.197,
       "step": 74
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5930981040000916,
-      "eval_runtime": 0.9231,
-      "eval_samples_per_second": 76.916,
-      "eval_steps_per_second": 3.25,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.6096646189689636,
+      "eval_runtime": 0.941,
+      "eval_samples_per_second": 75.453,
+      "eval_steps_per_second": 3.188,
       "step": 78
     },
     {
       "epoch": 24.615384615384617,
-      "grad_norm": 5.136722087860107,
+      "grad_norm": 4.380879878997803,
       "learning_rate": 4.074074074074074e-05,
-      "loss": 0.296,
+      "loss": 0.3577,
       "step": 80
     },
     {
       "epoch": 24.923076923076923,
-      "eval_accuracy": 0.6901408450704225,
-      "eval_loss": 0.6304548978805542,
-      "eval_runtime": 0.9177,
-      "eval_samples_per_second": 77.366,
-      "eval_steps_per_second": 3.269,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.6699813008308411,
+      "eval_runtime": 0.9378,
+      "eval_samples_per_second": 75.708,
+      "eval_steps_per_second": 3.199,
       "step": 81
     },
     {
       "epoch": 25.846153846153847,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.8947206735610962,
-      "eval_runtime": 0.9322,
-      "eval_samples_per_second": 76.164,
-      "eval_steps_per_second": 3.218,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.6676053404808044,
+      "eval_runtime": 0.9357,
+      "eval_samples_per_second": 75.876,
+      "eval_steps_per_second": 3.206,
       "step": 84
     },
     {
       "epoch": 26.76923076923077,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.6216529011726379,
-      "eval_runtime": 0.9208,
-      "eval_samples_per_second": 77.108,
-      "eval_steps_per_second": 3.258,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5475191473960876,
+      "eval_runtime": 0.9411,
+      "eval_samples_per_second": 75.444,
+      "eval_steps_per_second": 3.188,
       "step": 87
     },
     {
       "epoch": 27.692307692307693,
-      "grad_norm": 2.791210651397705,
+      "grad_norm": 2.8975579738616943,
       "learning_rate": 3.888888888888889e-05,
-      "loss": 0.2741,
+      "loss": 0.2988,
       "step": 90
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.7217584848403931,
-      "eval_runtime": 0.9386,
-      "eval_samples_per_second": 75.642,
-      "eval_steps_per_second": 3.196,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5382919311523438,
+      "eval_runtime": 0.9412,
+      "eval_samples_per_second": 75.435,
+      "eval_steps_per_second": 3.187,
       "step": 91
     },
     {
       "epoch": 28.923076923076923,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.6687091588973999,
-      "eval_runtime": 0.9304,
-      "eval_samples_per_second": 76.314,
-      "eval_steps_per_second": 3.225,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.5533848404884338,
+      "eval_runtime": 0.9288,
+      "eval_samples_per_second": 76.444,
+      "eval_steps_per_second": 3.23,
       "step": 94
     },
     {
       "epoch": 29.846153846153847,
       "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.6647565960884094,
-      "eval_runtime": 0.9206,
-      "eval_samples_per_second": 77.123,
-      "eval_steps_per_second": 3.259,
+      "eval_loss": 0.5842180252075195,
+      "eval_runtime": 0.9277,
+      "eval_samples_per_second": 76.533,
+      "eval_steps_per_second": 3.234,
       "step": 97
     },
     {
       "epoch": 30.76923076923077,
-      "grad_norm": 4.038685321807861,
+      "grad_norm": 3.0538382530212402,
       "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.2559,
+      "loss": 0.2595,
       "step": 100
     },
     {
       "epoch": 30.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.64328533411026,
-      "eval_runtime": 0.9253,
-      "eval_samples_per_second": 76.729,
-      "eval_steps_per_second": 3.242,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5964645743370056,
+      "eval_runtime": 0.9103,
+      "eval_samples_per_second": 77.994,
+      "eval_steps_per_second": 3.296,
       "step": 100
     },
     {
       "epoch": 32.0,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.6673524975776672,
-      "eval_runtime": 0.9299,
-      "eval_samples_per_second": 76.356,
-      "eval_steps_per_second": 3.226,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.6219843626022339,
+      "eval_runtime": 0.9431,
+      "eval_samples_per_second": 75.281,
+      "eval_steps_per_second": 3.181,
       "step": 104
     },
     {
       "epoch": 32.92307692307692,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6642715334892273,
-      "eval_runtime": 0.9221,
-      "eval_samples_per_second": 76.995,
-      "eval_steps_per_second": 3.253,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.6026638150215149,
+      "eval_runtime": 0.9477,
+      "eval_samples_per_second": 74.916,
+      "eval_steps_per_second": 3.165,
       "step": 107
     },
     {
       "epoch": 33.84615384615385,
-      "grad_norm": 2.9196064472198486,
+      "grad_norm": 3.94685959815979,
       "learning_rate": 3.518518518518519e-05,
-      "loss": 0.2001,
+      "loss": 0.2422,
       "step": 110
     },
     {
       "epoch": 33.84615384615385,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6247330904006958,
-      "eval_runtime": 0.923,
-      "eval_samples_per_second": 76.922,
-      "eval_steps_per_second": 3.25,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.6368895769119263,
+      "eval_runtime": 0.9216,
+      "eval_samples_per_second": 77.044,
+      "eval_steps_per_second": 3.255,
       "step": 110
     },
     {
       "epoch": 34.76923076923077,
-      "eval_accuracy": 0.6901408450704225,
-      "eval_loss": 0.634434163570404,
-      "eval_runtime": 0.9207,
-      "eval_samples_per_second": 77.116,
-      "eval_steps_per_second": 3.258,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.6033357381820679,
+      "eval_runtime": 0.9401,
+      "eval_samples_per_second": 75.526,
+      "eval_steps_per_second": 3.191,
       "step": 113
     },
     {
       "epoch": 36.0,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7071972489356995,
-      "eval_runtime": 0.9162,
-      "eval_samples_per_second": 77.491,
-      "eval_steps_per_second": 3.274,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.6912439465522766,
+      "eval_runtime": 0.9337,
+      "eval_samples_per_second": 76.043,
+      "eval_steps_per_second": 3.213,
       "step": 117
     },
     {
       "epoch": 36.92307692307692,
-      "grad_norm": 2.7139010429382324,
+      "grad_norm": 1.8390007019042969,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.1728,
+      "loss": 0.1927,
       "step": 120
     },
     {
       "epoch": 36.92307692307692,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.7145668864250183,
-      "eval_runtime": 0.9546,
-      "eval_samples_per_second": 74.38,
-      "eval_steps_per_second": 3.143,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.6582470536231995,
+      "eval_runtime": 0.9394,
+      "eval_samples_per_second": 75.581,
+      "eval_steps_per_second": 3.194,
       "step": 120
     },
     {
       "epoch": 37.84615384615385,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.8212233781814575,
-      "eval_runtime": 0.9206,
-      "eval_samples_per_second": 77.122,
-      "eval_steps_per_second": 3.259,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.6320064067840576,
+      "eval_runtime": 0.9433,
+      "eval_samples_per_second": 75.271,
+      "eval_steps_per_second": 3.18,
       "step": 123
     },
     {
       "epoch": 38.76923076923077,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.7901431322097778,
-      "eval_runtime": 0.9408,
-      "eval_samples_per_second": 75.465,
-      "eval_steps_per_second": 3.189,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.753158450126648,
+      "eval_runtime": 0.9223,
+      "eval_samples_per_second": 76.983,
+      "eval_steps_per_second": 3.253,
       "step": 126
     },
     {
       "epoch": 40.0,
-      "grad_norm": 3.7214014530181885,
+      "grad_norm": 8.522945404052734,
       "learning_rate": 3.148148148148148e-05,
-      "loss": 0.2109,
+      "loss": 0.2399,
       "step": 130
     },
     {
       "epoch": 40.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.8235028982162476,
-      "eval_runtime": 0.9237,
-      "eval_samples_per_second": 76.863,
-      "eval_steps_per_second": 3.248,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7909286618232727,
+      "eval_runtime": 0.9385,
+      "eval_samples_per_second": 75.651,
+      "eval_steps_per_second": 3.197,
       "step": 130
     },
     {
       "epoch": 40.92307692307692,
-      "eval_accuracy": 0.6901408450704225,
-      "eval_loss": 0.9196304678916931,
-      "eval_runtime": 0.9201,
-      "eval_samples_per_second": 77.162,
-      "eval_steps_per_second": 3.26,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.6808232665061951,
+      "eval_runtime": 0.929,
+      "eval_samples_per_second": 76.424,
+      "eval_steps_per_second": 3.229,
       "step": 133
     },
     {
       "epoch": 41.84615384615385,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7758485078811646,
-      "eval_runtime": 0.918,
-      "eval_samples_per_second": 77.345,
-      "eval_steps_per_second": 3.268,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5815951824188232,
+      "eval_runtime": 0.9333,
+      "eval_samples_per_second": 76.072,
+      "eval_steps_per_second": 3.214,
       "step": 136
     },
     {
       "epoch": 42.76923076923077,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.7692318558692932,
-      "eval_runtime": 0.9271,
-      "eval_samples_per_second": 76.58,
-      "eval_steps_per_second": 3.236,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5473891496658325,
+      "eval_runtime": 0.9353,
+      "eval_samples_per_second": 75.912,
+      "eval_steps_per_second": 3.208,
       "step": 139
     },
     {
       "epoch": 43.07692307692308,
-      "grad_norm": 2.981480121612549,
+      "grad_norm": 3.9758527278900146,
       "learning_rate": 2.962962962962963e-05,
-      "loss": 0.1634,
+      "loss": 0.2218,
       "step": 140
     },
     {
       "epoch": 44.0,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.830990731716156,
-      "eval_runtime": 0.9143,
-      "eval_samples_per_second": 77.651,
-      "eval_steps_per_second": 3.281,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.6309952139854431,
+      "eval_runtime": 0.9488,
+      "eval_samples_per_second": 74.835,
+      "eval_steps_per_second": 3.162,
       "step": 143
     },
     {
       "epoch": 44.92307692307692,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.7550302147865295,
-      "eval_runtime": 0.9175,
-      "eval_samples_per_second": 77.387,
-      "eval_steps_per_second": 3.27,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.6452985405921936,
+      "eval_runtime": 0.9466,
+      "eval_samples_per_second": 75.001,
+      "eval_steps_per_second": 3.169,
       "step": 146
     },
     {
       "epoch": 45.84615384615385,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.7645807862281799,
-      "eval_runtime": 0.9201,
-      "eval_samples_per_second": 77.165,
-      "eval_steps_per_second": 3.26,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.616996169090271,
+      "eval_runtime": 0.934,
+      "eval_samples_per_second": 76.018,
+      "eval_steps_per_second": 3.212,
       "step": 149
     },
     {
       "epoch": 46.15384615384615,
-      "grad_norm": 2.219327926635742,
+      "grad_norm": 3.51767635345459,
       "learning_rate": 2.777777777777778e-05,
-      "loss": 0.148,
+      "loss": 0.1817,
       "step": 150
     },
     {
       "epoch": 46.76923076923077,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7207580804824829,
-      "eval_runtime": 0.9248,
-      "eval_samples_per_second": 76.77,
-      "eval_steps_per_second": 3.244,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.603383481502533,
+      "eval_runtime": 0.9385,
+      "eval_samples_per_second": 75.653,
+      "eval_steps_per_second": 3.197,
       "step": 152
     },
     {
       "epoch": 48.0,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7324273586273193,
-      "eval_runtime": 0.9287,
-      "eval_samples_per_second": 76.454,
-      "eval_steps_per_second": 3.23,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.6349979043006897,
+      "eval_runtime": 0.9472,
+      "eval_samples_per_second": 74.957,
+      "eval_steps_per_second": 3.167,
       "step": 156
     },
     {
       "epoch": 48.92307692307692,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7855945229530334,
-      "eval_runtime": 0.923,
-      "eval_samples_per_second": 76.922,
-      "eval_steps_per_second": 3.25,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.6026537418365479,
+      "eval_runtime": 0.9329,
+      "eval_samples_per_second": 76.103,
+      "eval_steps_per_second": 3.216,
       "step": 159
     },
     {
       "epoch": 49.23076923076923,
-      "grad_norm": 3.3428738117218018,
+      "grad_norm": 2.8314857482910156,
       "learning_rate": 2.5925925925925925e-05,
-      "loss": 0.1568,
+      "loss": 0.1483,
       "step": 160
     },
     {
       "epoch": 49.84615384615385,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.8032997250556946,
-      "eval_runtime": 0.9319,
-      "eval_samples_per_second": 76.191,
-      "eval_steps_per_second": 3.219,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.559914767742157,
+      "eval_runtime": 0.9519,
+      "eval_samples_per_second": 74.59,
+      "eval_steps_per_second": 3.152,
       "step": 162
     },
     {
       "epoch": 50.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.9006530046463013,
-      "eval_runtime": 0.9271,
-      "eval_samples_per_second": 76.582,
-      "eval_steps_per_second": 3.236,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5817492604255676,
+      "eval_runtime": 0.9277,
+      "eval_samples_per_second": 76.537,
+      "eval_steps_per_second": 3.234,
       "step": 165
     },
     {
       "epoch": 52.0,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.817884624004364,
-      "eval_runtime": 0.926,
-      "eval_samples_per_second": 76.671,
-      "eval_steps_per_second": 3.24,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.6085928082466125,
+      "eval_runtime": 0.9365,
+      "eval_samples_per_second": 75.816,
+      "eval_steps_per_second": 3.203,
       "step": 169
     },
     {
       "epoch": 52.30769230769231,
-      "grad_norm": 2.3081116676330566,
+      "grad_norm": 4.343232154846191,
       "learning_rate": 2.4074074074074074e-05,
-      "loss": 0.1659,
+      "loss": 0.1668,
       "step": 170
     },
     {
       "epoch": 52.92307692307692,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7775102853775024,
-      "eval_runtime": 0.9343,
-      "eval_samples_per_second": 75.993,
-      "eval_steps_per_second": 3.211,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5743552446365356,
+      "eval_runtime": 0.9412,
+      "eval_samples_per_second": 75.439,
+      "eval_steps_per_second": 3.188,
       "step": 172
     },
     {
       "epoch": 53.84615384615385,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7214329838752747,
-      "eval_runtime": 0.9316,
-      "eval_samples_per_second": 76.215,
-      "eval_steps_per_second": 3.22,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.605885922908783,
+      "eval_runtime": 0.9418,
+      "eval_samples_per_second": 75.387,
+      "eval_steps_per_second": 3.185,
       "step": 175
     },
     {
       "epoch": 54.76923076923077,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.7385321259498596,
-      "eval_runtime": 0.9261,
-      "eval_samples_per_second": 76.67,
-      "eval_steps_per_second": 3.24,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.6454694867134094,
+      "eval_runtime": 0.9437,
+      "eval_samples_per_second": 75.239,
+      "eval_steps_per_second": 3.179,
       "step": 178
     },
     {
       "epoch": 55.38461538461539,
-      "grad_norm": 2.9334683418273926,
+      "grad_norm": 4.7387166023254395,
       "learning_rate": 2.2222222222222223e-05,
-      "loss": 0.1352,
+      "loss": 0.1372,
       "step": 180
     },
     {
       "epoch": 56.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.7433763742446899,
-      "eval_runtime": 0.9297,
-      "eval_samples_per_second": 76.371,
-      "eval_steps_per_second": 3.227,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5367295145988464,
+      "eval_runtime": 0.9301,
+      "eval_samples_per_second": 76.34,
+      "eval_steps_per_second": 3.226,
       "step": 182
     },
     {
       "epoch": 56.92307692307692,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.8971463441848755,
-      "eval_runtime": 0.9262,
-      "eval_samples_per_second": 76.656,
-      "eval_steps_per_second": 3.239,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5615168213844299,
+      "eval_runtime": 0.9333,
+      "eval_samples_per_second": 76.074,
+      "eval_steps_per_second": 3.214,
       "step": 185
     },
     {
       "epoch": 57.84615384615385,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.7821467518806458,
-      "eval_runtime": 0.928,
-      "eval_samples_per_second": 76.509,
-      "eval_steps_per_second": 3.233,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.6377549767494202,
+      "eval_runtime": 0.9271,
+      "eval_samples_per_second": 76.581,
+      "eval_steps_per_second": 3.236,
       "step": 188
     },
     {
       "epoch": 58.46153846153846,
-      "grad_norm": 4.111489295959473,
+      "grad_norm": 3.1127660274505615,
       "learning_rate": 2.037037037037037e-05,
-      "loss": 0.1309,
+      "loss": 0.1485,
       "step": 190
     },
     {
       "epoch": 58.76923076923077,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.7896379828453064,
-      "eval_runtime": 0.9233,
-      "eval_samples_per_second": 76.899,
-      "eval_steps_per_second": 3.249,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5687428116798401,
+      "eval_runtime": 0.9342,
+      "eval_samples_per_second": 75.999,
+      "eval_steps_per_second": 3.211,
       "step": 191
     },
     {
       "epoch": 60.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.8339643478393555,
-      "eval_runtime": 0.933,
-      "eval_samples_per_second": 76.101,
-      "eval_steps_per_second": 3.216,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.48974475264549255,
+      "eval_runtime": 0.9499,
+      "eval_samples_per_second": 74.743,
+      "eval_steps_per_second": 3.158,
       "step": 195
     },
     {
       "epoch": 60.92307692307692,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.8154428005218506,
-      "eval_runtime": 0.9292,
-      "eval_samples_per_second": 76.406,
-      "eval_steps_per_second": 3.228,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.43841129541397095,
+      "eval_runtime": 0.9329,
+      "eval_samples_per_second": 76.104,
+      "eval_steps_per_second": 3.216,
       "step": 198
     },
     {
       "epoch": 61.53846153846154,
-      "grad_norm": 4.092026233673096,
+      "grad_norm": 2.5722365379333496,
       "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.1201,
+      "loss": 0.1426,
       "step": 200
     },
     {
       "epoch": 61.84615384615385,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.8184639811515808,
-      "eval_runtime": 0.9373,
-      "eval_samples_per_second": 75.748,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5086758732795715,
+      "eval_runtime": 0.9371,
+      "eval_samples_per_second": 75.766,
       "eval_steps_per_second": 3.201,
       "step": 201
     },
     {
       "epoch": 62.76923076923077,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.9639940857887268,
-      "eval_runtime": 0.9235,
-      "eval_samples_per_second": 76.878,
-      "eval_steps_per_second": 3.248,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.47570785880088806,
+      "eval_runtime": 0.935,
+      "eval_samples_per_second": 75.939,
+      "eval_steps_per_second": 3.209,
       "step": 204
     },
     {
       "epoch": 64.0,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.8484686613082886,
-      "eval_runtime": 0.9218,
-      "eval_samples_per_second": 77.024,
-      "eval_steps_per_second": 3.255,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.4372512102127075,
+      "eval_runtime": 0.9455,
+      "eval_samples_per_second": 75.095,
+      "eval_steps_per_second": 3.173,
       "step": 208
     },
     {
       "epoch": 64.61538461538461,
-      "grad_norm": 2.0266196727752686,
+      "grad_norm": 3.0248568058013916,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.1291,
+      "loss": 0.1333,
       "step": 210
     },
     {
       "epoch": 64.92307692307692,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.8807466626167297,
-      "eval_runtime": 0.9258,
-      "eval_samples_per_second": 76.694,
-      "eval_steps_per_second": 3.241,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.4512348473072052,
+      "eval_runtime": 0.9662,
+      "eval_samples_per_second": 73.486,
+      "eval_steps_per_second": 3.105,
       "step": 211
     },
     {
       "epoch": 65.84615384615384,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.8652527332305908,
-      "eval_runtime": 0.9568,
-      "eval_samples_per_second": 74.203,
-      "eval_steps_per_second": 3.135,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.4619483947753906,
+      "eval_runtime": 0.9227,
+      "eval_samples_per_second": 76.949,
+      "eval_steps_per_second": 3.251,
       "step": 214
     },
     {
       "epoch": 66.76923076923077,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.8744374513626099,
-      "eval_runtime": 0.9208,
-      "eval_samples_per_second": 77.107,
-      "eval_steps_per_second": 3.258,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5519948601722717,
+      "eval_runtime": 0.9387,
+      "eval_samples_per_second": 75.634,
+      "eval_steps_per_second": 3.196,
       "step": 217
     },
     {
       "epoch": 67.6923076923077,
-      "grad_norm": 2.0726959705352783,
+      "grad_norm": 3.7715258598327637,
       "learning_rate": 1.4814814814814815e-05,
-      "loss": 0.124,
+      "loss": 0.1306,
       "step": 220
     },
     {
       "epoch": 68.0,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.8723464012145996,
-      "eval_runtime": 0.9161,
-      "eval_samples_per_second": 77.501,
-      "eval_steps_per_second": 3.275,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5160856246948242,
+      "eval_runtime": 0.9414,
+      "eval_samples_per_second": 75.417,
+      "eval_steps_per_second": 3.187,
       "step": 221
     },
     {
       "epoch": 68.92307692307692,
       "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.8948094248771667,
-      "eval_runtime": 0.93,
-      "eval_samples_per_second": 76.342,
-      "eval_steps_per_second": 3.226,
+      "eval_loss": 0.5180492401123047,
+      "eval_runtime": 0.9311,
+      "eval_samples_per_second": 76.252,
+      "eval_steps_per_second": 3.222,
       "step": 224
     },
     {
       "epoch": 69.84615384615384,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.9777162671089172,
-      "eval_runtime": 0.9269,
-      "eval_samples_per_second": 76.603,
-      "eval_steps_per_second": 3.237,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5777754783630371,
+      "eval_runtime": 0.9362,
+      "eval_samples_per_second": 75.843,
+      "eval_steps_per_second": 3.205,
       "step": 227
     },
     {
       "epoch": 70.76923076923077,
-      "grad_norm": 3.592405319213867,
+      "grad_norm": 2.9611454010009766,
       "learning_rate": 1.2962962962962962e-05,
-      "loss": 0.1262,
+      "loss": 0.1327,
       "step": 230
     },
     {
       "epoch": 70.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.940915048122406,
-      "eval_runtime": 0.9252,
-      "eval_samples_per_second": 76.743,
-      "eval_steps_per_second": 3.243,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.593269407749176,
+      "eval_runtime": 0.9395,
+      "eval_samples_per_second": 75.57,
+      "eval_steps_per_second": 3.193,
       "step": 230
     },
     {
       "epoch": 72.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.9617937207221985,
-      "eval_runtime": 0.9316,
-      "eval_samples_per_second": 76.211,
-      "eval_steps_per_second": 3.22,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5222498178482056,
+      "eval_runtime": 0.9354,
+      "eval_samples_per_second": 75.903,
+      "eval_steps_per_second": 3.207,
       "step": 234
     },
     {
       "epoch": 72.92307692307692,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.9641876220703125,
-      "eval_runtime": 0.9327,
-      "eval_samples_per_second": 76.122,
-      "eval_steps_per_second": 3.216,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5103858709335327,
+      "eval_runtime": 0.941,
+      "eval_samples_per_second": 75.455,
+      "eval_steps_per_second": 3.188,
       "step": 237
     },
     {
       "epoch": 73.84615384615384,
-      "grad_norm": 1.9709769487380981,
+      "grad_norm": 2.8741390705108643,
       "learning_rate": 1.1111111111111112e-05,
-      "loss": 0.1036,
+      "loss": 0.1171,
       "step": 240
     },
     {
       "epoch": 73.84615384615384,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.9737982153892517,
-      "eval_runtime": 0.9281,
-      "eval_samples_per_second": 76.5,
-      "eval_steps_per_second": 3.232,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5024046301841736,
+      "eval_runtime": 0.943,
+      "eval_samples_per_second": 75.292,
+      "eval_steps_per_second": 3.181,
       "step": 240
     },
     {
       "epoch": 74.76923076923077,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.9787779450416565,
-      "eval_runtime": 0.9363,
-      "eval_samples_per_second": 75.831,
-      "eval_steps_per_second": 3.204,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.50595623254776,
+      "eval_runtime": 0.944,
+      "eval_samples_per_second": 75.213,
+      "eval_steps_per_second": 3.178,
       "step": 243
     },
     {
       "epoch": 76.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 1.0114330053329468,
-      "eval_runtime": 0.9271,
-      "eval_samples_per_second": 76.582,
-      "eval_steps_per_second": 3.236,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.5267102718353271,
+      "eval_runtime": 0.937,
+      "eval_samples_per_second": 75.772,
+      "eval_steps_per_second": 3.202,
       "step": 247
     },
     {
       "epoch": 76.92307692307692,
-      "grad_norm": 3.0993545055389404,
+      "grad_norm": 3.9980766773223877,
       "learning_rate": 9.259259259259259e-06,
-      "loss": 0.1183,
+      "loss": 0.1227,
       "step": 250
     },
     {
       "epoch": 76.92307692307692,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 1.0004260540008545,
-      "eval_runtime": 0.9321,
-      "eval_samples_per_second": 76.171,
-      "eval_steps_per_second": 3.218,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.4774617552757263,
+      "eval_runtime": 0.9456,
+      "eval_samples_per_second": 75.082,
+      "eval_steps_per_second": 3.172,
       "step": 250
     },
     {
       "epoch": 77.84615384615384,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 1.0407198667526245,
-      "eval_runtime": 0.9175,
-      "eval_samples_per_second": 77.387,
-      "eval_steps_per_second": 3.27,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5019908547401428,
+      "eval_runtime": 0.9328,
+      "eval_samples_per_second": 76.116,
+      "eval_steps_per_second": 3.216,
       "step": 253
     },
     {
       "epoch": 78.76923076923077,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 1.1509737968444824,
-      "eval_runtime": 0.9303,
-      "eval_samples_per_second": 76.319,
-      "eval_steps_per_second": 3.225,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.5243141651153564,
+      "eval_runtime": 0.9369,
+      "eval_samples_per_second": 75.782,
+      "eval_steps_per_second": 3.202,
       "step": 256
     },
     {
       "epoch": 80.0,
-      "grad_norm": 5.949638843536377,
+      "grad_norm": 5.15329647064209,
       "learning_rate": 7.4074074074074075e-06,
-      "loss": 0.0981,
+      "loss": 0.1304,
       "step": 260
     },
     {
       "epoch": 80.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 1.0718269348144531,
-      "eval_runtime": 0.9185,
-      "eval_samples_per_second": 77.304,
-      "eval_steps_per_second": 3.266,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.6195448637008667,
+      "eval_runtime": 0.9685,
+      "eval_samples_per_second": 73.307,
+      "eval_steps_per_second": 3.097,
       "step": 260
     },
     {
       "epoch": 80.92307692307692,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.9988247752189636,
-      "eval_runtime": 0.933,
-      "eval_samples_per_second": 76.098,
-      "eval_steps_per_second": 3.215,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.5740300416946411,
+      "eval_runtime": 0.9291,
+      "eval_samples_per_second": 76.417,
+      "eval_steps_per_second": 3.229,
       "step": 263
     },
     {
       "epoch": 81.84615384615384,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 1.0053763389587402,
-      "eval_runtime": 0.9278,
-      "eval_samples_per_second": 76.524,
-      "eval_steps_per_second": 3.233,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5652071237564087,
+      "eval_runtime": 0.9331,
+      "eval_samples_per_second": 76.091,
+      "eval_steps_per_second": 3.215,
       "step": 266
     },
     {
       "epoch": 82.76923076923077,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.9896395206451416,
-      "eval_runtime": 0.9224,
-      "eval_samples_per_second": 76.975,
-      "eval_steps_per_second": 3.252,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5749574303627014,
+      "eval_runtime": 0.9389,
+      "eval_samples_per_second": 75.618,
+      "eval_steps_per_second": 3.195,
       "step": 269
     },
     {
       "epoch": 83.07692307692308,
-      "grad_norm": 2.5808982849121094,
+      "grad_norm": 2.7778007984161377,
       "learning_rate": 5.555555555555556e-06,
-      "loss": 0.106,
+      "loss": 0.1152,
       "step": 270
     },
     {
       "epoch": 84.0,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.9850640892982483,
-      "eval_runtime": 0.9326,
-      "eval_samples_per_second": 76.134,
-      "eval_steps_per_second": 3.217,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5829400420188904,
+      "eval_runtime": 0.9675,
+      "eval_samples_per_second": 73.381,
+      "eval_steps_per_second": 3.101,
       "step": 273
     },
     {
       "epoch": 84.92307692307692,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.9769949913024902,
-      "eval_runtime": 0.9194,
-      "eval_samples_per_second": 77.228,
-      "eval_steps_per_second": 3.263,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5853610038757324,
+      "eval_runtime": 0.9581,
+      "eval_samples_per_second": 74.104,
+      "eval_steps_per_second": 3.131,
       "step": 276
     },
     {
       "epoch": 85.84615384615384,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.9622512459754944,
-      "eval_runtime": 0.9188,
-      "eval_samples_per_second": 77.276,
-      "eval_steps_per_second": 3.265,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5853627324104309,
+      "eval_runtime": 0.9385,
+      "eval_samples_per_second": 75.649,
+      "eval_steps_per_second": 3.196,
       "step": 279
     },
     {
       "epoch": 86.15384615384616,
-      "grad_norm": 3.007988214492798,
+      "grad_norm": 2.607625722885132,
       "learning_rate": 3.7037037037037037e-06,
-      "loss": 0.114,
+      "loss": 0.1069,
       "step": 280
     },
     {
       "epoch": 86.76923076923077,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.9664495587348938,
-      "eval_runtime": 0.9429,
-      "eval_samples_per_second": 75.301,
-      "eval_steps_per_second": 3.182,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5825924277305603,
+      "eval_runtime": 0.9377,
+      "eval_samples_per_second": 75.72,
+      "eval_steps_per_second": 3.199,
       "step": 282
     },
     {
       "epoch": 88.0,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.9780421853065491,
-      "eval_runtime": 0.9309,
-      "eval_samples_per_second": 76.269,
-      "eval_steps_per_second": 3.223,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5838791728019714,
+      "eval_runtime": 0.9668,
+      "eval_samples_per_second": 73.437,
+      "eval_steps_per_second": 3.103,
       "step": 286
     },
     {
       "epoch": 88.92307692307692,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.9670152068138123,
-      "eval_runtime": 0.927,
-      "eval_samples_per_second": 76.593,
-      "eval_steps_per_second": 3.236,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5792059898376465,
+      "eval_runtime": 0.9421,
+      "eval_samples_per_second": 75.36,
+      "eval_steps_per_second": 3.184,
       "step": 289
     },
     {
       "epoch": 89.23076923076923,
-      "grad_norm": 2.9891533851623535,
+      "grad_norm": 3.399799108505249,
       "learning_rate": 1.8518518518518519e-06,
-      "loss": 0.1157,
+      "loss": 0.122,
       "step": 290
     },
     {
       "epoch": 89.84615384615384,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.9586439728736877,
-      "eval_runtime": 0.9232,
-      "eval_samples_per_second": 76.909,
-      "eval_steps_per_second": 3.25,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5755288600921631,
+      "eval_runtime": 0.9561,
+      "eval_samples_per_second": 74.263,
+      "eval_steps_per_second": 3.138,
       "step": 292
     },
     {
       "epoch": 90.76923076923077,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.9586858749389648,
-      "eval_runtime": 0.9402,
-      "eval_samples_per_second": 75.515,
-      "eval_steps_per_second": 3.191,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5751092433929443,
+      "eval_runtime": 0.9263,
+      "eval_samples_per_second": 76.65,
+      "eval_steps_per_second": 3.239,
       "step": 295
     },
     {
       "epoch": 92.0,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.9610524773597717,
-      "eval_runtime": 0.9383,
-      "eval_samples_per_second": 75.669,
-      "eval_steps_per_second": 3.197,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5747934579849243,
+      "eval_runtime": 0.9473,
+      "eval_samples_per_second": 74.952,
+      "eval_steps_per_second": 3.167,
       "step": 299
     },
     {
       "epoch": 92.3076923076923,
-      "grad_norm": 2.5594146251678467,
+      "grad_norm": 2.51318621635437,
       "learning_rate": 0.0,
-      "loss": 0.0834,
+      "loss": 0.1268,
       "step": 300
     },
     {
       "epoch": 92.3076923076923,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.9612475037574768,
-      "eval_runtime": 0.9284,
-      "eval_samples_per_second": 76.472,
-      "eval_steps_per_second": 3.231,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5748143196105957,
+      "eval_runtime": 0.9381,
+      "eval_samples_per_second": 75.682,
+      "eval_steps_per_second": 3.198,
       "step": 300
     },
     {
       "epoch": 92.3076923076923,
       "step": 300,
       "total_flos": 2.8402872494292173e+18,
-      "train_loss": 0.23628523468971252,
-      "train_runtime": 1593.1864,
-      "train_samples_per_second": 24.919,
+      "train_loss": 0.25812069336573285,
+      "train_runtime": 1594.5895,
+      "train_samples_per_second": 24.897,
       "train_steps_per_second": 0.188
     },
     {
       "epoch": 92.3076923076923,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.7217584848403931,
-      "eval_runtime": 0.9525,
-      "eval_samples_per_second": 74.537,
-      "eval_steps_per_second": 3.149,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5367295145988464,
+      "eval_runtime": 0.9601,
+      "eval_samples_per_second": 73.951,
+      "eval_steps_per_second": 3.125,
       "step": 300
     }
   ],