End of training

README.md

@@ -22,7 +22,7 @@ model-index:
     metrics:
     - name: Accuracy
       type: accuracy
-      value: 0.704225352112676
+      value: 0.8169014084507042
 ---
 
 <!-- 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: 0.9612
-- Accuracy: 0.7042
+- Loss: 0.7218
+- Accuracy: 0.8169
 
 ## Model description
 

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 92.3076923076923,
-    "eval_accuracy": 0.8732394366197183,
-    "eval_loss": 0.48321232199668884,
-    "eval_runtime": 1.0328,
-    "eval_samples_per_second": 68.746,
-    "eval_steps_per_second": 2.905,
+    "eval_accuracy": 0.8169014084507042,
+    "eval_loss": 0.7217584848403931,
+    "eval_runtime": 0.9525,
+    "eval_samples_per_second": 74.537,
+    "eval_steps_per_second": 3.149,
     "total_flos": 2.8402872494292173e+18,
-    "train_loss": 0.24676756938298544,
-    "train_runtime": 1640.323,
-    "train_samples_per_second": 24.203,
-    "train_steps_per_second": 0.183
+    "train_loss": 0.23628523468971252,
+    "train_runtime": 1593.1864,
+    "train_samples_per_second": 24.919,
+    "train_steps_per_second": 0.188
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 92.3076923076923,
-    "eval_accuracy": 0.8732394366197183,
-    "eval_loss": 0.48321232199668884,
-    "eval_runtime": 1.0328,
-    "eval_samples_per_second": 68.746,
-    "eval_steps_per_second": 2.905
+    "eval_accuracy": 0.8169014084507042,
+    "eval_loss": 0.7217584848403931,
+    "eval_runtime": 0.9525,
+    "eval_samples_per_second": 74.537,
+    "eval_steps_per_second": 3.149
 }

runs/May22_06-21-30_defc3061f818/events.out.tfevents.1716360498.defc3061f818.1801.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7321ac633b801e66970e19a71bde34301a6ea70a92c1242102044a029121f77e
+size 411

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 92.3076923076923,
     "total_flos": 2.8402872494292173e+18,
-    "train_loss": 0.24676756938298544,
-    "train_runtime": 1640.323,
-    "train_samples_per_second": 24.203,
-    "train_steps_per_second": 0.183
+    "train_loss": 0.23628523468971252,
+    "train_runtime": 1593.1864,
+    "train_samples_per_second": 24.919,
+    "train_steps_per_second": 0.188
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.8732394366197183,
-  "best_model_checkpoint": "deit-base-distilled-patch16-224-65-fold4/checkpoint-188",
+  "best_metric": 0.8169014084507042,
+  "best_model_checkpoint": "deit-base-distilled-patch16-224-65-fold4/checkpoint-91",
   "epoch": 92.3076923076923,
   "eval_steps": 500,
   "global_step": 300,
@@ -10,1067 +10,1067 @@
   "log_history": [
     {
       "epoch": 0.9230769230769231,
-      "eval_accuracy": 0.5492957746478874,
-      "eval_loss": 0.7218557596206665,
-      "eval_runtime": 0.9745,
-      "eval_samples_per_second": 72.855,
-      "eval_steps_per_second": 3.078,
+      "eval_accuracy": 0.4647887323943662,
+      "eval_loss": 0.7266488075256348,
+      "eval_runtime": 0.8434,
+      "eval_samples_per_second": 84.184,
+      "eval_steps_per_second": 3.557,
       "step": 3
     },
     {
       "epoch": 1.8461538461538463,
-      "eval_accuracy": 0.5492957746478874,
-      "eval_loss": 0.686266303062439,
-      "eval_runtime": 0.9669,
-      "eval_samples_per_second": 73.432,
-      "eval_steps_per_second": 3.103,
+      "eval_accuracy": 0.5211267605633803,
+      "eval_loss": 0.8115941286087036,
+      "eval_runtime": 0.8919,
+      "eval_samples_per_second": 79.608,
+      "eval_steps_per_second": 3.364,
       "step": 6
     },
     {
       "epoch": 2.769230769230769,
-      "eval_accuracy": 0.5492957746478874,
-      "eval_loss": 0.6578378081321716,
-      "eval_runtime": 0.9657,
-      "eval_samples_per_second": 73.525,
-      "eval_steps_per_second": 3.107,
+      "eval_accuracy": 0.4647887323943662,
+      "eval_loss": 0.7081143856048584,
+      "eval_runtime": 0.8712,
+      "eval_samples_per_second": 81.496,
+      "eval_steps_per_second": 3.443,
       "step": 9
     },
     {
       "epoch": 3.076923076923077,
-      "grad_norm": 3.424938678741455,
+      "grad_norm": 2.1958322525024414,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.6825,
+      "loss": 0.7173,
       "step": 10
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.6338028169014085,
-      "eval_loss": 0.6292961239814758,
-      "eval_runtime": 0.9785,
-      "eval_samples_per_second": 72.558,
-      "eval_steps_per_second": 3.066,
+      "eval_accuracy": 0.5633802816901409,
+      "eval_loss": 0.6644501090049744,
+      "eval_runtime": 0.8954,
+      "eval_samples_per_second": 79.298,
+      "eval_steps_per_second": 3.351,
       "step": 13
     },
     {
       "epoch": 4.923076923076923,
-      "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.6186050176620483,
-      "eval_runtime": 0.987,
-      "eval_samples_per_second": 71.936,
-      "eval_steps_per_second": 3.04,
+      "eval_accuracy": 0.5915492957746479,
+      "eval_loss": 0.6440630555152893,
+      "eval_runtime": 0.8941,
+      "eval_samples_per_second": 79.413,
+      "eval_steps_per_second": 3.355,
       "step": 16
     },
     {
       "epoch": 5.846153846153846,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.6135374903678894,
-      "eval_runtime": 1.0113,
-      "eval_samples_per_second": 70.204,
-      "eval_steps_per_second": 2.966,
+      "eval_accuracy": 0.676056338028169,
+      "eval_loss": 0.6400186419487,
+      "eval_runtime": 0.933,
+      "eval_samples_per_second": 76.096,
+      "eval_steps_per_second": 3.215,
       "step": 19
     },
     {
       "epoch": 6.153846153846154,
-      "grad_norm": 4.245552062988281,
+      "grad_norm": 2.7608325481414795,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.6206,
+      "loss": 0.6351,
       "step": 20
     },
     {
       "epoch": 6.769230769230769,
-      "eval_accuracy": 0.647887323943662,
-      "eval_loss": 0.6162843704223633,
-      "eval_runtime": 0.9805,
-      "eval_samples_per_second": 72.413,
-      "eval_steps_per_second": 3.06,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.6054678559303284,
+      "eval_runtime": 0.9387,
+      "eval_samples_per_second": 75.638,
+      "eval_steps_per_second": 3.196,
       "step": 22
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.647887323943662,
-      "eval_loss": 0.6350468397140503,
-      "eval_runtime": 0.9932,
-      "eval_samples_per_second": 71.487,
-      "eval_steps_per_second": 3.021,
+      "eval_accuracy": 0.5352112676056338,
+      "eval_loss": 0.7769902944564819,
+      "eval_runtime": 0.9329,
+      "eval_samples_per_second": 76.108,
+      "eval_steps_per_second": 3.216,
       "step": 26
     },
     {
       "epoch": 8.923076923076923,
       "eval_accuracy": 0.6901408450704225,
-      "eval_loss": 0.6078002452850342,
-      "eval_runtime": 0.9899,
-      "eval_samples_per_second": 71.723,
-      "eval_steps_per_second": 3.031,
+      "eval_loss": 0.6259447932243347,
+      "eval_runtime": 0.924,
+      "eval_samples_per_second": 76.842,
+      "eval_steps_per_second": 3.247,
       "step": 29
     },
     {
       "epoch": 9.23076923076923,
-      "grad_norm": 3.0329267978668213,
+      "grad_norm": 5.699825286865234,
       "learning_rate": 5e-05,
-      "loss": 0.5728,
+      "loss": 0.5434,
       "step": 30
     },
     {
       "epoch": 9.846153846153847,
-      "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.6872609257698059,
-      "eval_runtime": 0.9874,
-      "eval_samples_per_second": 71.907,
-      "eval_steps_per_second": 3.038,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.5889422297477722,
+      "eval_runtime": 0.9148,
+      "eval_samples_per_second": 77.612,
+      "eval_steps_per_second": 3.279,
       "step": 32
     },
     {
       "epoch": 10.76923076923077,
-      "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.6771360635757446,
-      "eval_runtime": 0.984,
-      "eval_samples_per_second": 72.152,
-      "eval_steps_per_second": 3.049,
+      "eval_accuracy": 0.647887323943662,
+      "eval_loss": 0.7283326387405396,
+      "eval_runtime": 0.9062,
+      "eval_samples_per_second": 78.347,
+      "eval_steps_per_second": 3.31,
       "step": 35
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.6619718309859155,
-      "eval_loss": 0.5912012457847595,
-      "eval_runtime": 0.9919,
-      "eval_samples_per_second": 71.581,
-      "eval_steps_per_second": 3.025,
+      "eval_accuracy": 0.647887323943662,
+      "eval_loss": 0.6897829174995422,
+      "eval_runtime": 0.9341,
+      "eval_samples_per_second": 76.013,
+      "eval_steps_per_second": 3.212,
       "step": 39
     },
     {
       "epoch": 12.307692307692308,
-      "grad_norm": 5.794826030731201,
+      "grad_norm": 2.502607583999634,
       "learning_rate": 4.814814814814815e-05,
-      "loss": 0.5329,
+      "loss": 0.4861,
       "step": 40
     },
     {
       "epoch": 12.923076923076923,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.5523556470870972,
-      "eval_runtime": 0.9889,
-      "eval_samples_per_second": 71.799,
-      "eval_steps_per_second": 3.034,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.6428806781768799,
+      "eval_runtime": 0.9382,
+      "eval_samples_per_second": 75.679,
+      "eval_steps_per_second": 3.198,
       "step": 42
     },
     {
       "epoch": 13.846153846153847,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.5922580361366272,
-      "eval_runtime": 0.9948,
-      "eval_samples_per_second": 71.37,
-      "eval_steps_per_second": 3.016,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.691542387008667,
+      "eval_runtime": 0.932,
+      "eval_samples_per_second": 76.179,
+      "eval_steps_per_second": 3.219,
       "step": 45
     },
     {
       "epoch": 14.76923076923077,
-      "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.6649972200393677,
-      "eval_runtime": 0.9954,
-      "eval_samples_per_second": 71.326,
-      "eval_steps_per_second": 3.014,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.5702247619628906,
+      "eval_runtime": 0.9242,
+      "eval_samples_per_second": 76.822,
+      "eval_steps_per_second": 3.246,
       "step": 48
     },
     {
       "epoch": 15.384615384615385,
-      "grad_norm": 4.871856689453125,
+      "grad_norm": 5.8594255447387695,
       "learning_rate": 4.62962962962963e-05,
-      "loss": 0.4279,
+      "loss": 0.4285,
       "step": 50
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.5182512402534485,
-      "eval_runtime": 0.9953,
-      "eval_samples_per_second": 71.338,
-      "eval_steps_per_second": 3.014,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.6356058120727539,
+      "eval_runtime": 0.919,
+      "eval_samples_per_second": 77.258,
+      "eval_steps_per_second": 3.264,
       "step": 52
     },
     {
       "epoch": 16.923076923076923,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.4760647714138031,
-      "eval_runtime": 0.9935,
-      "eval_samples_per_second": 71.467,
-      "eval_steps_per_second": 3.02,
+      "eval_accuracy": 0.676056338028169,
+      "eval_loss": 0.6981013417243958,
+      "eval_runtime": 0.9247,
+      "eval_samples_per_second": 76.785,
+      "eval_steps_per_second": 3.244,
       "step": 55
     },
     {
       "epoch": 17.846153846153847,
       "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.558963418006897,
-      "eval_runtime": 0.9911,
-      "eval_samples_per_second": 71.637,
-      "eval_steps_per_second": 3.027,
+      "eval_loss": 0.5218324661254883,
+      "eval_runtime": 0.917,
+      "eval_samples_per_second": 77.425,
+      "eval_steps_per_second": 3.271,
       "step": 58
     },
     {
       "epoch": 18.46153846153846,
-      "grad_norm": 4.168473243713379,
+      "grad_norm": 6.547220706939697,
       "learning_rate": 4.4444444444444447e-05,
-      "loss": 0.4055,
+      "loss": 0.3781,
       "step": 60
     },
     {
       "epoch": 18.76923076923077,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.5320470333099365,
-      "eval_runtime": 1.0156,
-      "eval_samples_per_second": 69.91,
-      "eval_steps_per_second": 2.954,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.534016489982605,
+      "eval_runtime": 0.914,
+      "eval_samples_per_second": 77.678,
+      "eval_steps_per_second": 3.282,
       "step": 61
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.6604605913162231,
-      "eval_runtime": 0.9924,
-      "eval_samples_per_second": 71.546,
-      "eval_steps_per_second": 3.023,
+      "eval_accuracy": 0.676056338028169,
+      "eval_loss": 0.7611135244369507,
+      "eval_runtime": 0.9206,
+      "eval_samples_per_second": 77.126,
+      "eval_steps_per_second": 3.259,
       "step": 65
     },
     {
       "epoch": 20.923076923076923,
       "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.5820997357368469,
-      "eval_runtime": 0.9824,
-      "eval_samples_per_second": 72.271,
-      "eval_steps_per_second": 3.054,
+      "eval_loss": 0.5939193964004517,
+      "eval_runtime": 0.9415,
+      "eval_samples_per_second": 75.408,
+      "eval_steps_per_second": 3.186,
       "step": 68
     },
     {
       "epoch": 21.53846153846154,
-      "grad_norm": 5.980050086975098,
+      "grad_norm": 4.307927131652832,
       "learning_rate": 4.259259259259259e-05,
-      "loss": 0.3478,
+      "loss": 0.3516,
       "step": 70
     },
     {
       "epoch": 21.846153846153847,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.5588834881782532,
-      "eval_runtime": 0.9895,
-      "eval_samples_per_second": 71.75,
-      "eval_steps_per_second": 3.032,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.61859130859375,
+      "eval_runtime": 0.9317,
+      "eval_samples_per_second": 76.201,
+      "eval_steps_per_second": 3.22,
       "step": 71
     },
     {
       "epoch": 22.76923076923077,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6247106194496155,
-      "eval_runtime": 0.9919,
-      "eval_samples_per_second": 71.578,
-      "eval_steps_per_second": 3.024,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.712211549282074,
+      "eval_runtime": 0.9272,
+      "eval_samples_per_second": 76.571,
+      "eval_steps_per_second": 3.235,
       "step": 74
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.6619718309859155,
-      "eval_loss": 0.7006160020828247,
-      "eval_runtime": 1.0036,
-      "eval_samples_per_second": 70.747,
-      "eval_steps_per_second": 2.989,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5930981040000916,
+      "eval_runtime": 0.9231,
+      "eval_samples_per_second": 76.916,
+      "eval_steps_per_second": 3.25,
       "step": 78
     },
     {
       "epoch": 24.615384615384617,
-      "grad_norm": 3.0665063858032227,
+      "grad_norm": 5.136722087860107,
       "learning_rate": 4.074074074074074e-05,
-      "loss": 0.3769,
+      "loss": 0.296,
       "step": 80
     },
     {
       "epoch": 24.923076923076923,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.7575426697731018,
-      "eval_runtime": 0.9915,
-      "eval_samples_per_second": 71.611,
-      "eval_steps_per_second": 3.026,
+      "eval_accuracy": 0.6901408450704225,
+      "eval_loss": 0.6304548978805542,
+      "eval_runtime": 0.9177,
+      "eval_samples_per_second": 77.366,
+      "eval_steps_per_second": 3.269,
       "step": 81
     },
     {
       "epoch": 25.846153846153847,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.5367358922958374,
-      "eval_runtime": 0.991,
-      "eval_samples_per_second": 71.646,
-      "eval_steps_per_second": 3.027,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.8947206735610962,
+      "eval_runtime": 0.9322,
+      "eval_samples_per_second": 76.164,
+      "eval_steps_per_second": 3.218,
       "step": 84
     },
     {
       "epoch": 26.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.533458948135376,
-      "eval_runtime": 0.9946,
-      "eval_samples_per_second": 71.388,
-      "eval_steps_per_second": 3.016,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.6216529011726379,
+      "eval_runtime": 0.9208,
+      "eval_samples_per_second": 77.108,
+      "eval_steps_per_second": 3.258,
       "step": 87
     },
     {
       "epoch": 27.692307692307693,
-      "grad_norm": 5.900879859924316,
+      "grad_norm": 2.791210651397705,
       "learning_rate": 3.888888888888889e-05,
-      "loss": 0.2957,
+      "loss": 0.2741,
       "step": 90
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.5913785696029663,
-      "eval_runtime": 1.0059,
-      "eval_samples_per_second": 70.582,
-      "eval_steps_per_second": 2.982,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.7217584848403931,
+      "eval_runtime": 0.9386,
+      "eval_samples_per_second": 75.642,
+      "eval_steps_per_second": 3.196,
       "step": 91
     },
     {
       "epoch": 28.923076923076923,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6780489683151245,
-      "eval_runtime": 0.9943,
-      "eval_samples_per_second": 71.404,
-      "eval_steps_per_second": 3.017,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.6687091588973999,
+      "eval_runtime": 0.9304,
+      "eval_samples_per_second": 76.314,
+      "eval_steps_per_second": 3.225,
       "step": 94
     },
     {
       "epoch": 29.846153846153847,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.534493625164032,
-      "eval_runtime": 0.9982,
-      "eval_samples_per_second": 71.129,
-      "eval_steps_per_second": 3.005,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.6647565960884094,
+      "eval_runtime": 0.9206,
+      "eval_samples_per_second": 77.123,
+      "eval_steps_per_second": 3.259,
       "step": 97
     },
     {
       "epoch": 30.76923076923077,
-      "grad_norm": 3.5530407428741455,
+      "grad_norm": 4.038685321807861,
       "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.2463,
+      "loss": 0.2559,
       "step": 100
     },
     {
       "epoch": 30.76923076923077,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.613211989402771,
-      "eval_runtime": 0.9924,
-      "eval_samples_per_second": 71.54,
-      "eval_steps_per_second": 3.023,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.64328533411026,
+      "eval_runtime": 0.9253,
+      "eval_samples_per_second": 76.729,
+      "eval_steps_per_second": 3.242,
       "step": 100
     },
     {
       "epoch": 32.0,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5757858157157898,
-      "eval_runtime": 0.9882,
-      "eval_samples_per_second": 71.845,
-      "eval_steps_per_second": 3.036,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.6673524975776672,
+      "eval_runtime": 0.9299,
+      "eval_samples_per_second": 76.356,
+      "eval_steps_per_second": 3.226,
       "step": 104
     },
     {
       "epoch": 32.92307692307692,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.7235603332519531,
-      "eval_runtime": 0.9977,
-      "eval_samples_per_second": 71.165,
-      "eval_steps_per_second": 3.007,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.6642715334892273,
+      "eval_runtime": 0.9221,
+      "eval_samples_per_second": 76.995,
+      "eval_steps_per_second": 3.253,
       "step": 107
     },
     {
       "epoch": 33.84615384615385,
-      "grad_norm": 4.502363204956055,
+      "grad_norm": 2.9196064472198486,
       "learning_rate": 3.518518518518519e-05,
-      "loss": 0.2323,
+      "loss": 0.2001,
       "step": 110
     },
     {
       "epoch": 33.84615384615385,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5247332453727722,
-      "eval_runtime": 1.0185,
-      "eval_samples_per_second": 69.714,
-      "eval_steps_per_second": 2.946,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.6247330904006958,
+      "eval_runtime": 0.923,
+      "eval_samples_per_second": 76.922,
+      "eval_steps_per_second": 3.25,
       "step": 110
     },
     {
       "epoch": 34.76923076923077,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.6018014550209045,
-      "eval_runtime": 0.994,
-      "eval_samples_per_second": 71.429,
-      "eval_steps_per_second": 3.018,
+      "eval_accuracy": 0.6901408450704225,
+      "eval_loss": 0.634434163570404,
+      "eval_runtime": 0.9207,
+      "eval_samples_per_second": 77.116,
+      "eval_steps_per_second": 3.258,
       "step": 113
     },
     {
       "epoch": 36.0,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5366302728652954,
-      "eval_runtime": 0.9895,
-      "eval_samples_per_second": 71.751,
-      "eval_steps_per_second": 3.032,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7071972489356995,
+      "eval_runtime": 0.9162,
+      "eval_samples_per_second": 77.491,
+      "eval_steps_per_second": 3.274,
       "step": 117
     },
     {
       "epoch": 36.92307692307692,
-      "grad_norm": 1.8001881837844849,
+      "grad_norm": 2.7139010429382324,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.1921,
+      "loss": 0.1728,
       "step": 120
     },
     {
       "epoch": 36.92307692307692,
       "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6314247846603394,
-      "eval_runtime": 0.9859,
-      "eval_samples_per_second": 72.013,
-      "eval_steps_per_second": 3.043,
+      "eval_loss": 0.7145668864250183,
+      "eval_runtime": 0.9546,
+      "eval_samples_per_second": 74.38,
+      "eval_steps_per_second": 3.143,
       "step": 120
     },
     {
       "epoch": 37.84615384615385,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.5763274431228638,
-      "eval_runtime": 0.9986,
-      "eval_samples_per_second": 71.096,
-      "eval_steps_per_second": 3.004,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.8212233781814575,
+      "eval_runtime": 0.9206,
+      "eval_samples_per_second": 77.122,
+      "eval_steps_per_second": 3.259,
       "step": 123
     },
     {
       "epoch": 38.76923076923077,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5573564171791077,
-      "eval_runtime": 0.9999,
-      "eval_samples_per_second": 71.009,
-      "eval_steps_per_second": 3.0,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.7901431322097778,
+      "eval_runtime": 0.9408,
+      "eval_samples_per_second": 75.465,
+      "eval_steps_per_second": 3.189,
       "step": 126
     },
     {
       "epoch": 40.0,
-      "grad_norm": 2.9866695404052734,
+      "grad_norm": 3.7214014530181885,
       "learning_rate": 3.148148148148148e-05,
-      "loss": 0.1686,
+      "loss": 0.2109,
       "step": 130
     },
     {
       "epoch": 40.0,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.6260572075843811,
-      "eval_runtime": 1.0023,
-      "eval_samples_per_second": 70.835,
-      "eval_steps_per_second": 2.993,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.8235028982162476,
+      "eval_runtime": 0.9237,
+      "eval_samples_per_second": 76.863,
+      "eval_steps_per_second": 3.248,
       "step": 130
     },
     {
       "epoch": 40.92307692307692,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.6524580121040344,
-      "eval_runtime": 0.9907,
-      "eval_samples_per_second": 71.669,
-      "eval_steps_per_second": 3.028,
+      "eval_accuracy": 0.6901408450704225,
+      "eval_loss": 0.9196304678916931,
+      "eval_runtime": 0.9201,
+      "eval_samples_per_second": 77.162,
+      "eval_steps_per_second": 3.26,
       "step": 133
     },
     {
       "epoch": 41.84615384615385,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5726364254951477,
-      "eval_runtime": 0.9944,
-      "eval_samples_per_second": 71.399,
-      "eval_steps_per_second": 3.017,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7758485078811646,
+      "eval_runtime": 0.918,
+      "eval_samples_per_second": 77.345,
+      "eval_steps_per_second": 3.268,
       "step": 136
     },
     {
       "epoch": 42.76923076923077,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.8199837803840637,
-      "eval_runtime": 0.9956,
-      "eval_samples_per_second": 71.317,
-      "eval_steps_per_second": 3.013,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.7692318558692932,
+      "eval_runtime": 0.9271,
+      "eval_samples_per_second": 76.58,
+      "eval_steps_per_second": 3.236,
       "step": 139
     },
     {
       "epoch": 43.07692307692308,
-      "grad_norm": 8.961894035339355,
+      "grad_norm": 2.981480121612549,
       "learning_rate": 2.962962962962963e-05,
-      "loss": 0.2073,
+      "loss": 0.1634,
       "step": 140
     },
     {
       "epoch": 44.0,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.4798416793346405,
-      "eval_runtime": 0.9976,
-      "eval_samples_per_second": 71.172,
-      "eval_steps_per_second": 3.007,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.830990731716156,
+      "eval_runtime": 0.9143,
+      "eval_samples_per_second": 77.651,
+      "eval_steps_per_second": 3.281,
       "step": 143
     },
     {
       "epoch": 44.92307692307692,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5341726541519165,
-      "eval_runtime": 0.9976,
-      "eval_samples_per_second": 71.169,
-      "eval_steps_per_second": 3.007,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.7550302147865295,
+      "eval_runtime": 0.9175,
+      "eval_samples_per_second": 77.387,
+      "eval_steps_per_second": 3.27,
       "step": 146
     },
     {
       "epoch": 45.84615384615385,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.48342809081077576,
-      "eval_runtime": 0.9955,
-      "eval_samples_per_second": 71.324,
-      "eval_steps_per_second": 3.014,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.7645807862281799,
+      "eval_runtime": 0.9201,
+      "eval_samples_per_second": 77.165,
+      "eval_steps_per_second": 3.26,
       "step": 149
     },
     {
       "epoch": 46.15384615384615,
-      "grad_norm": 6.036099433898926,
+      "grad_norm": 2.219327926635742,
       "learning_rate": 2.777777777777778e-05,
-      "loss": 0.1702,
+      "loss": 0.148,
       "step": 150
     },
     {
       "epoch": 46.76923076923077,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6100845336914062,
-      "eval_runtime": 1.0014,
-      "eval_samples_per_second": 70.903,
-      "eval_steps_per_second": 2.996,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7207580804824829,
+      "eval_runtime": 0.9248,
+      "eval_samples_per_second": 76.77,
+      "eval_steps_per_second": 3.244,
       "step": 152
     },
     {
       "epoch": 48.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.47786733508110046,
-      "eval_runtime": 0.9939,
-      "eval_samples_per_second": 71.433,
-      "eval_steps_per_second": 3.018,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7324273586273193,
+      "eval_runtime": 0.9287,
+      "eval_samples_per_second": 76.454,
+      "eval_steps_per_second": 3.23,
       "step": 156
     },
     {
       "epoch": 48.92307692307692,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5047819018363953,
-      "eval_runtime": 0.9956,
-      "eval_samples_per_second": 71.313,
-      "eval_steps_per_second": 3.013,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7855945229530334,
+      "eval_runtime": 0.923,
+      "eval_samples_per_second": 76.922,
+      "eval_steps_per_second": 3.25,
       "step": 159
     },
     {
       "epoch": 49.23076923076923,
-      "grad_norm": 3.7059566974639893,
+      "grad_norm": 3.3428738117218018,
       "learning_rate": 2.5925925925925925e-05,
-      "loss": 0.153,
+      "loss": 0.1568,
       "step": 160
     },
     {
       "epoch": 49.84615384615385,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6297626495361328,
-      "eval_runtime": 0.9916,
-      "eval_samples_per_second": 71.6,
-      "eval_steps_per_second": 3.025,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.8032997250556946,
+      "eval_runtime": 0.9319,
+      "eval_samples_per_second": 76.191,
+      "eval_steps_per_second": 3.219,
       "step": 162
     },
     {
       "epoch": 50.76923076923077,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.5995147824287415,
-      "eval_runtime": 0.9924,
-      "eval_samples_per_second": 71.541,
-      "eval_steps_per_second": 3.023,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.9006530046463013,
+      "eval_runtime": 0.9271,
+      "eval_samples_per_second": 76.582,
+      "eval_steps_per_second": 3.236,
       "step": 165
     },
     {
       "epoch": 52.0,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.6475109457969666,
-      "eval_runtime": 0.9899,
-      "eval_samples_per_second": 71.724,
-      "eval_steps_per_second": 3.031,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.817884624004364,
+      "eval_runtime": 0.926,
+      "eval_samples_per_second": 76.671,
+      "eval_steps_per_second": 3.24,
       "step": 169
     },
     {
       "epoch": 52.30769230769231,
-      "grad_norm": 2.797576427459717,
+      "grad_norm": 2.3081116676330566,
       "learning_rate": 2.4074074074074074e-05,
-      "loss": 0.1508,
+      "loss": 0.1659,
       "step": 170
     },
     {
       "epoch": 52.92307692307692,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.48875874280929565,
-      "eval_runtime": 1.0078,
-      "eval_samples_per_second": 70.449,
-      "eval_steps_per_second": 2.977,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7775102853775024,
+      "eval_runtime": 0.9343,
+      "eval_samples_per_second": 75.993,
+      "eval_steps_per_second": 3.211,
       "step": 172
     },
     {
       "epoch": 53.84615384615385,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.4953955411911011,
-      "eval_runtime": 0.9916,
-      "eval_samples_per_second": 71.604,
-      "eval_steps_per_second": 3.026,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7214329838752747,
+      "eval_runtime": 0.9316,
+      "eval_samples_per_second": 76.215,
+      "eval_steps_per_second": 3.22,
       "step": 175
     },
     {
       "epoch": 54.76923076923077,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.43896353244781494,
-      "eval_runtime": 0.9963,
-      "eval_samples_per_second": 71.26,
-      "eval_steps_per_second": 3.011,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.7385321259498596,
+      "eval_runtime": 0.9261,
+      "eval_samples_per_second": 76.67,
+      "eval_steps_per_second": 3.24,
       "step": 178
     },
     {
       "epoch": 55.38461538461539,
-      "grad_norm": 2.899003744125366,
+      "grad_norm": 2.9334683418273926,
       "learning_rate": 2.2222222222222223e-05,
-      "loss": 0.1293,
+      "loss": 0.1352,
       "step": 180
     },
     {
       "epoch": 56.0,
-      "eval_accuracy": 0.8591549295774648,
-      "eval_loss": 0.477837473154068,
-      "eval_runtime": 0.9883,
-      "eval_samples_per_second": 71.838,
-      "eval_steps_per_second": 3.035,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.7433763742446899,
+      "eval_runtime": 0.9297,
+      "eval_samples_per_second": 76.371,
+      "eval_steps_per_second": 3.227,
       "step": 182
     },
     {
       "epoch": 56.92307692307692,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.4887799024581909,
-      "eval_runtime": 0.9913,
-      "eval_samples_per_second": 71.622,
-      "eval_steps_per_second": 3.026,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.8971463441848755,
+      "eval_runtime": 0.9262,
+      "eval_samples_per_second": 76.656,
+      "eval_steps_per_second": 3.239,
       "step": 185
     },
     {
       "epoch": 57.84615384615385,
-      "eval_accuracy": 0.8732394366197183,
-      "eval_loss": 0.48321232199668884,
-      "eval_runtime": 0.9931,
-      "eval_samples_per_second": 71.494,
-      "eval_steps_per_second": 3.021,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.7821467518806458,
+      "eval_runtime": 0.928,
+      "eval_samples_per_second": 76.509,
+      "eval_steps_per_second": 3.233,
       "step": 188
     },
     {
       "epoch": 58.46153846153846,
-      "grad_norm": 2.5362701416015625,
+      "grad_norm": 4.111489295959473,
       "learning_rate": 2.037037037037037e-05,
-      "loss": 0.1489,
+      "loss": 0.1309,
       "step": 190
     },
     {
       "epoch": 58.76923076923077,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.527686595916748,
-      "eval_runtime": 0.981,
-      "eval_samples_per_second": 72.378,
-      "eval_steps_per_second": 3.058,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.7896379828453064,
+      "eval_runtime": 0.9233,
+      "eval_samples_per_second": 76.899,
+      "eval_steps_per_second": 3.249,
       "step": 191
     },
     {
       "epoch": 60.0,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.6216662526130676,
-      "eval_runtime": 0.9828,
-      "eval_samples_per_second": 72.243,
-      "eval_steps_per_second": 3.053,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.8339643478393555,
+      "eval_runtime": 0.933,
+      "eval_samples_per_second": 76.101,
+      "eval_steps_per_second": 3.216,
       "step": 195
     },
     {
       "epoch": 60.92307692307692,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.6089593768119812,
-      "eval_runtime": 0.9925,
-      "eval_samples_per_second": 71.539,
-      "eval_steps_per_second": 3.023,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.8154428005218506,
+      "eval_runtime": 0.9292,
+      "eval_samples_per_second": 76.406,
+      "eval_steps_per_second": 3.228,
       "step": 198
     },
     {
       "epoch": 61.53846153846154,
-      "grad_norm": 2.5179524421691895,
+      "grad_norm": 4.092026233673096,
       "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.1487,
+      "loss": 0.1201,
       "step": 200
     },
     {
       "epoch": 61.84615384615385,
-      "eval_accuracy": 0.8450704225352113,
-      "eval_loss": 0.5424482226371765,
-      "eval_runtime": 0.9877,
-      "eval_samples_per_second": 71.886,
-      "eval_steps_per_second": 3.037,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.8184639811515808,
+      "eval_runtime": 0.9373,
+      "eval_samples_per_second": 75.748,
+      "eval_steps_per_second": 3.201,
       "step": 201
     },
     {
       "epoch": 62.76923076923077,
-      "eval_accuracy": 0.8450704225352113,
-      "eval_loss": 0.5569549202919006,
-      "eval_runtime": 0.9873,
-      "eval_samples_per_second": 71.913,
-      "eval_steps_per_second": 3.039,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.9639940857887268,
+      "eval_runtime": 0.9235,
+      "eval_samples_per_second": 76.878,
+      "eval_steps_per_second": 3.248,
       "step": 204
     },
     {
       "epoch": 64.0,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.7247959971427917,
-      "eval_runtime": 0.9969,
-      "eval_samples_per_second": 71.22,
-      "eval_steps_per_second": 3.009,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.8484686613082886,
+      "eval_runtime": 0.9218,
+      "eval_samples_per_second": 77.024,
+      "eval_steps_per_second": 3.255,
       "step": 208
     },
     {
       "epoch": 64.61538461538461,
-      "grad_norm": 3.1306471824645996,
+      "grad_norm": 2.0266196727752686,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.1456,
+      "loss": 0.1291,
       "step": 210
     },
     {
       "epoch": 64.92307692307692,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5840876698493958,
-      "eval_runtime": 1.0357,
-      "eval_samples_per_second": 68.554,
-      "eval_steps_per_second": 2.897,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.8807466626167297,
+      "eval_runtime": 0.9258,
+      "eval_samples_per_second": 76.694,
+      "eval_steps_per_second": 3.241,
       "step": 211
     },
     {
       "epoch": 65.84615384615384,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5905009508132935,
-      "eval_runtime": 1.1058,
-      "eval_samples_per_second": 64.208,
-      "eval_steps_per_second": 2.713,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.8652527332305908,
+      "eval_runtime": 0.9568,
+      "eval_samples_per_second": 74.203,
+      "eval_steps_per_second": 3.135,
       "step": 214
     },
     {
       "epoch": 66.76923076923077,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5608553290367126,
-      "eval_runtime": 1.0367,
-      "eval_samples_per_second": 68.484,
-      "eval_steps_per_second": 2.894,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.8744374513626099,
+      "eval_runtime": 0.9208,
+      "eval_samples_per_second": 77.107,
+      "eval_steps_per_second": 3.258,
       "step": 217
     },
     {
       "epoch": 67.6923076923077,
-      "grad_norm": 2.4782862663269043,
+      "grad_norm": 2.0726959705352783,
       "learning_rate": 1.4814814814814815e-05,
-      "loss": 0.1284,
+      "loss": 0.124,
       "step": 220
     },
     {
       "epoch": 68.0,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.546994149684906,
-      "eval_runtime": 0.9751,
-      "eval_samples_per_second": 72.81,
-      "eval_steps_per_second": 3.076,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.8723464012145996,
+      "eval_runtime": 0.9161,
+      "eval_samples_per_second": 77.501,
+      "eval_steps_per_second": 3.275,
       "step": 221
     },
     {
       "epoch": 68.92307692307692,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5472754240036011,
-      "eval_runtime": 1.0047,
-      "eval_samples_per_second": 70.665,
-      "eval_steps_per_second": 2.986,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.8948094248771667,
+      "eval_runtime": 0.93,
+      "eval_samples_per_second": 76.342,
+      "eval_steps_per_second": 3.226,
       "step": 224
     },
     {
       "epoch": 69.84615384615384,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5813232660293579,
-      "eval_runtime": 0.9977,
-      "eval_samples_per_second": 71.162,
-      "eval_steps_per_second": 3.007,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.9777162671089172,
+      "eval_runtime": 0.9269,
+      "eval_samples_per_second": 76.603,
+      "eval_steps_per_second": 3.237,
       "step": 227
     },
     {
       "epoch": 70.76923076923077,
-      "grad_norm": 1.8887909650802612,
+      "grad_norm": 3.592405319213867,
       "learning_rate": 1.2962962962962962e-05,
-      "loss": 0.1225,
+      "loss": 0.1262,
       "step": 230
     },
     {
       "epoch": 70.76923076923077,
-      "eval_accuracy": 0.8450704225352113,
-      "eval_loss": 0.5682988166809082,
-      "eval_runtime": 0.9949,
-      "eval_samples_per_second": 71.367,
-      "eval_steps_per_second": 3.016,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.940915048122406,
+      "eval_runtime": 0.9252,
+      "eval_samples_per_second": 76.743,
+      "eval_steps_per_second": 3.243,
       "step": 230
     },
     {
       "epoch": 72.0,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5581419467926025,
-      "eval_runtime": 0.9914,
-      "eval_samples_per_second": 71.615,
-      "eval_steps_per_second": 3.026,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.9617937207221985,
+      "eval_runtime": 0.9316,
+      "eval_samples_per_second": 76.211,
+      "eval_steps_per_second": 3.22,
       "step": 234
     },
     {
       "epoch": 72.92307692307692,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5717195272445679,
-      "eval_runtime": 0.976,
-      "eval_samples_per_second": 72.745,
-      "eval_steps_per_second": 3.074,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.9641876220703125,
+      "eval_runtime": 0.9327,
+      "eval_samples_per_second": 76.122,
+      "eval_steps_per_second": 3.216,
       "step": 237
     },
     {
       "epoch": 73.84615384615384,
-      "grad_norm": 2.5398831367492676,
+      "grad_norm": 1.9709769487380981,
       "learning_rate": 1.1111111111111112e-05,
-      "loss": 0.1233,
+      "loss": 0.1036,
       "step": 240
     },
     {
       "epoch": 73.84615384615384,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.6054385304450989,
-      "eval_runtime": 0.9717,
-      "eval_samples_per_second": 73.067,
-      "eval_steps_per_second": 3.087,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.9737982153892517,
+      "eval_runtime": 0.9281,
+      "eval_samples_per_second": 76.5,
+      "eval_steps_per_second": 3.232,
       "step": 240
     },
     {
       "epoch": 74.76923076923077,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5910421013832092,
-      "eval_runtime": 0.9862,
-      "eval_samples_per_second": 71.995,
-      "eval_steps_per_second": 3.042,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.9787779450416565,
+      "eval_runtime": 0.9363,
+      "eval_samples_per_second": 75.831,
+      "eval_steps_per_second": 3.204,
       "step": 243
     },
     {
       "epoch": 76.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.570685863494873,
-      "eval_runtime": 0.9929,
-      "eval_samples_per_second": 71.507,
-      "eval_steps_per_second": 3.021,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 1.0114330053329468,
+      "eval_runtime": 0.9271,
+      "eval_samples_per_second": 76.582,
+      "eval_steps_per_second": 3.236,
       "step": 247
     },
     {
       "epoch": 76.92307692307692,
-      "grad_norm": 2.284491539001465,
+      "grad_norm": 3.0993545055389404,
       "learning_rate": 9.259259259259259e-06,
-      "loss": 0.1234,
+      "loss": 0.1183,
       "step": 250
     },
     {
       "epoch": 76.92307692307692,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5732807517051697,
-      "eval_runtime": 0.9877,
-      "eval_samples_per_second": 71.884,
-      "eval_steps_per_second": 3.037,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 1.0004260540008545,
+      "eval_runtime": 0.9321,
+      "eval_samples_per_second": 76.171,
+      "eval_steps_per_second": 3.218,
       "step": 250
     },
     {
       "epoch": 77.84615384615384,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5748334527015686,
-      "eval_runtime": 0.9897,
-      "eval_samples_per_second": 71.74,
-      "eval_steps_per_second": 3.031,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 1.0407198667526245,
+      "eval_runtime": 0.9175,
+      "eval_samples_per_second": 77.387,
+      "eval_steps_per_second": 3.27,
       "step": 253
     },
     {
       "epoch": 78.76923076923077,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5722648501396179,
-      "eval_runtime": 0.9949,
-      "eval_samples_per_second": 71.363,
-      "eval_steps_per_second": 3.015,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 1.1509737968444824,
+      "eval_runtime": 0.9303,
+      "eval_samples_per_second": 76.319,
+      "eval_steps_per_second": 3.225,
       "step": 256
     },
     {
       "epoch": 80.0,
-      "grad_norm": 4.123226165771484,
+      "grad_norm": 5.949638843536377,
       "learning_rate": 7.4074074074074075e-06,
-      "loss": 0.1219,
+      "loss": 0.0981,
       "step": 260
     },
     {
       "epoch": 80.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5503106713294983,
-      "eval_runtime": 1.0099,
-      "eval_samples_per_second": 70.307,
-      "eval_steps_per_second": 2.971,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 1.0718269348144531,
+      "eval_runtime": 0.9185,
+      "eval_samples_per_second": 77.304,
+      "eval_steps_per_second": 3.266,
       "step": 260
     },
     {
       "epoch": 80.92307692307692,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5531720519065857,
-      "eval_runtime": 0.978,
-      "eval_samples_per_second": 72.6,
-      "eval_steps_per_second": 3.068,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.9988247752189636,
+      "eval_runtime": 0.933,
+      "eval_samples_per_second": 76.098,
+      "eval_steps_per_second": 3.215,
       "step": 263
     },
     {
       "epoch": 81.84615384615384,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.5827966332435608,
-      "eval_runtime": 0.9983,
-      "eval_samples_per_second": 71.119,
-      "eval_steps_per_second": 3.005,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 1.0053763389587402,
+      "eval_runtime": 0.9278,
+      "eval_samples_per_second": 76.524,
+      "eval_steps_per_second": 3.233,
       "step": 266
     },
     {
       "epoch": 82.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.6061961650848389,
-      "eval_runtime": 0.98,
-      "eval_samples_per_second": 72.449,
-      "eval_steps_per_second": 3.061,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.9896395206451416,
+      "eval_runtime": 0.9224,
+      "eval_samples_per_second": 76.975,
+      "eval_steps_per_second": 3.252,
       "step": 269
     },
     {
       "epoch": 83.07692307692308,
-      "grad_norm": 2.491278886795044,
+      "grad_norm": 2.5808982849121094,
       "learning_rate": 5.555555555555556e-06,
-      "loss": 0.1075,
+      "loss": 0.106,
       "step": 270
     },
     {
       "epoch": 84.0,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5751623511314392,
-      "eval_runtime": 1.0114,
-      "eval_samples_per_second": 70.2,
-      "eval_steps_per_second": 2.966,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.9850640892982483,
+      "eval_runtime": 0.9326,
+      "eval_samples_per_second": 76.134,
+      "eval_steps_per_second": 3.217,
       "step": 273
     },
     {
       "epoch": 84.92307692307692,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5748041868209839,
-      "eval_runtime": 0.989,
-      "eval_samples_per_second": 71.791,
-      "eval_steps_per_second": 3.033,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.9769949913024902,
+      "eval_runtime": 0.9194,
+      "eval_samples_per_second": 77.228,
+      "eval_steps_per_second": 3.263,
       "step": 276
     },
     {
       "epoch": 85.84615384615384,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5776336789131165,
-      "eval_runtime": 0.9954,
-      "eval_samples_per_second": 71.325,
-      "eval_steps_per_second": 3.014,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.9622512459754944,
+      "eval_runtime": 0.9188,
+      "eval_samples_per_second": 77.276,
+      "eval_steps_per_second": 3.265,
       "step": 279
     },
     {
       "epoch": 86.15384615384616,
-      "grad_norm": 2.633176565170288,
+      "grad_norm": 3.007988214492798,
       "learning_rate": 3.7037037037037037e-06,
-      "loss": 0.1013,
+      "loss": 0.114,
       "step": 280
     },
     {
       "epoch": 86.76923076923077,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5844343304634094,
-      "eval_runtime": 0.9859,
-      "eval_samples_per_second": 72.012,
-      "eval_steps_per_second": 3.043,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.9664495587348938,
+      "eval_runtime": 0.9429,
+      "eval_samples_per_second": 75.301,
+      "eval_steps_per_second": 3.182,
       "step": 282
     },
     {
       "epoch": 88.0,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5929856896400452,
-      "eval_runtime": 1.0276,
-      "eval_samples_per_second": 69.094,
-      "eval_steps_per_second": 2.919,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.9780421853065491,
+      "eval_runtime": 0.9309,
+      "eval_samples_per_second": 76.269,
+      "eval_steps_per_second": 3.223,
       "step": 286
     },
     {
       "epoch": 88.92307692307692,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.6019977331161499,
-      "eval_runtime": 0.9928,
-      "eval_samples_per_second": 71.514,
-      "eval_steps_per_second": 3.022,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.9670152068138123,
+      "eval_runtime": 0.927,
+      "eval_samples_per_second": 76.593,
+      "eval_steps_per_second": 3.236,
       "step": 289
     },
     {
       "epoch": 89.23076923076923,
-      "grad_norm": 3.9098527431488037,
+      "grad_norm": 2.9891533851623535,
       "learning_rate": 1.8518518518518519e-06,
-      "loss": 0.1092,
+      "loss": 0.1157,
       "step": 290
     },
     {
       "epoch": 89.84615384615384,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.6055042147636414,
-      "eval_runtime": 1.004,
-      "eval_samples_per_second": 70.718,
-      "eval_steps_per_second": 2.988,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.9586439728736877,
+      "eval_runtime": 0.9232,
+      "eval_samples_per_second": 76.909,
+      "eval_steps_per_second": 3.25,
       "step": 292
     },
     {
       "epoch": 90.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.6074590086936951,
-      "eval_runtime": 0.9928,
-      "eval_samples_per_second": 71.517,
-      "eval_steps_per_second": 3.022,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.9586858749389648,
+      "eval_runtime": 0.9402,
+      "eval_samples_per_second": 75.515,
+      "eval_steps_per_second": 3.191,
       "step": 295
     },
     {
       "epoch": 92.0,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.607978880405426,
-      "eval_runtime": 0.9917,
-      "eval_samples_per_second": 71.596,
-      "eval_steps_per_second": 3.025,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.9610524773597717,
+      "eval_runtime": 0.9383,
+      "eval_samples_per_second": 75.669,
+      "eval_steps_per_second": 3.197,
       "step": 299
     },
     {
       "epoch": 92.3076923076923,
-      "grad_norm": 2.8199360370635986,
+      "grad_norm": 2.5594146251678467,
       "learning_rate": 0.0,
-      "loss": 0.1096,
+      "loss": 0.0834,
       "step": 300
     },
     {
       "epoch": 92.3076923076923,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.6081295013427734,
-      "eval_runtime": 1.0121,
-      "eval_samples_per_second": 70.154,
-      "eval_steps_per_second": 2.964,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.9612475037574768,
+      "eval_runtime": 0.9284,
+      "eval_samples_per_second": 76.472,
+      "eval_steps_per_second": 3.231,
       "step": 300
     },
     {
       "epoch": 92.3076923076923,
       "step": 300,
       "total_flos": 2.8402872494292173e+18,
-      "train_loss": 0.24676756938298544,
-      "train_runtime": 1640.323,
-      "train_samples_per_second": 24.203,
-      "train_steps_per_second": 0.183
+      "train_loss": 0.23628523468971252,
+      "train_runtime": 1593.1864,
+      "train_samples_per_second": 24.919,
+      "train_steps_per_second": 0.188
     },
     {
       "epoch": 92.3076923076923,
-      "eval_accuracy": 0.8732394366197183,
-      "eval_loss": 0.48321232199668884,
-      "eval_runtime": 1.0328,
-      "eval_samples_per_second": 68.746,
-      "eval_steps_per_second": 2.905,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.7217584848403931,
+      "eval_runtime": 0.9525,
+      "eval_samples_per_second": 74.537,
+      "eval_steps_per_second": 3.149,
       "step": 300
     }
   ],