End of training

README.md

@@ -22,7 +22,7 @@ model-index:
     metrics:
     - name: Accuracy
       type: accuracy
-      value: 0.8309859154929577
+      value: 0.8732394366197183
 ---
 
 <!-- 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: 0.5407
-- Accuracy: 0.8310
+- Loss: 0.5156
+- Accuracy: 0.8732
 
 ## Model description
 

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 92.3076923076923,
-    "eval_accuracy": 0.8591549295774648,
-    "eval_loss": 0.4150727093219757,
-    "eval_runtime": 1.0438,
-    "eval_samples_per_second": 68.024,
-    "eval_steps_per_second": 2.874,
+    "eval_accuracy": 0.8732394366197183,
+    "eval_loss": 0.5155981183052063,
+    "eval_runtime": 0.9923,
+    "eval_samples_per_second": 71.551,
+    "eval_steps_per_second": 3.023,
     "total_flos": 2.8402872494292173e+18,
-    "train_loss": 0.2401055637995402,
-    "train_runtime": 1630.2519,
-    "train_samples_per_second": 24.352,
-    "train_steps_per_second": 0.184
+    "train_loss": 0.23062967757383981,
+    "train_runtime": 1594.1954,
+    "train_samples_per_second": 24.903,
+    "train_steps_per_second": 0.188
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 92.3076923076923,
-    "eval_accuracy": 0.8591549295774648,
-    "eval_loss": 0.4150727093219757,
-    "eval_runtime": 1.0438,
-    "eval_samples_per_second": 68.024,
-    "eval_steps_per_second": 2.874
+    "eval_accuracy": 0.8732394366197183,
+    "eval_loss": 0.5155981183052063,
+    "eval_runtime": 0.9923,
+    "eval_samples_per_second": 71.551,
+    "eval_steps_per_second": 3.023
 }

runs/May22_06-48-30_defc3061f818/events.out.tfevents.1716362115.defc3061f818.1801.3

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e6d9732eebf20325eae0dc4739c7b06e2aa75a942783a23261dea29edcdb7e9
+size 411

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 92.3076923076923,
     "total_flos": 2.8402872494292173e+18,
-    "train_loss": 0.2401055637995402,
-    "train_runtime": 1630.2519,
-    "train_samples_per_second": 24.352,
-    "train_steps_per_second": 0.184
+    "train_loss": 0.23062967757383981,
+    "train_runtime": 1594.1954,
+    "train_samples_per_second": 24.903,
+    "train_steps_per_second": 0.188
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.8591549295774648,
-  "best_model_checkpoint": "deit-base-distilled-patch16-224-65-fold5/checkpoint-123",
+  "best_metric": 0.8732394366197183,
+  "best_model_checkpoint": "deit-base-distilled-patch16-224-65-fold5/checkpoint-260",
   "epoch": 92.3076923076923,
   "eval_steps": 500,
   "global_step": 300,
@@ -11,1066 +11,1066 @@
     {
       "epoch": 0.9230769230769231,
       "eval_accuracy": 0.49295774647887325,
-      "eval_loss": 0.7462348937988281,
-      "eval_runtime": 1.0051,
-      "eval_samples_per_second": 70.638,
-      "eval_steps_per_second": 2.985,
+      "eval_loss": 0.7424866557121277,
+      "eval_runtime": 0.9428,
+      "eval_samples_per_second": 75.31,
+      "eval_steps_per_second": 3.182,
       "step": 3
     },
     {
       "epoch": 1.8461538461538463,
-      "eval_accuracy": 0.5774647887323944,
-      "eval_loss": 0.6782689094543457,
-      "eval_runtime": 0.9584,
-      "eval_samples_per_second": 74.079,
-      "eval_steps_per_second": 3.13,
+      "eval_accuracy": 0.5633802816901409,
+      "eval_loss": 0.7193098664283752,
+      "eval_runtime": 0.9623,
+      "eval_samples_per_second": 73.782,
+      "eval_steps_per_second": 3.118,
       "step": 6
     },
     {
       "epoch": 2.769230769230769,
-      "eval_accuracy": 0.6338028169014085,
-      "eval_loss": 0.6407648921012878,
-      "eval_runtime": 0.9595,
-      "eval_samples_per_second": 73.994,
-      "eval_steps_per_second": 3.127,
+      "eval_accuracy": 0.5915492957746479,
+      "eval_loss": 0.6808025240898132,
+      "eval_runtime": 0.9647,
+      "eval_samples_per_second": 73.601,
+      "eval_steps_per_second": 3.11,
       "step": 9
     },
     {
       "epoch": 3.076923076923077,
-      "grad_norm": 2.5982041358947754,
+      "grad_norm": 6.528295040130615,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.7011,
+      "loss": 0.7309,
       "step": 10
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.6078053712844849,
-      "eval_runtime": 0.9526,
-      "eval_samples_per_second": 74.534,
-      "eval_steps_per_second": 3.149,
+      "eval_accuracy": 0.5915492957746479,
+      "eval_loss": 0.6252564787864685,
+      "eval_runtime": 0.9206,
+      "eval_samples_per_second": 77.124,
+      "eval_steps_per_second": 3.259,
       "step": 13
     },
     {
       "epoch": 4.923076923076923,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.5911686420440674,
-      "eval_runtime": 0.964,
-      "eval_samples_per_second": 73.654,
-      "eval_steps_per_second": 3.112,
+      "eval_accuracy": 0.676056338028169,
+      "eval_loss": 0.6021668910980225,
+      "eval_runtime": 0.9277,
+      "eval_samples_per_second": 76.533,
+      "eval_steps_per_second": 3.234,
       "step": 16
     },
     {
       "epoch": 5.846153846153846,
-      "eval_accuracy": 0.5633802816901409,
-      "eval_loss": 0.6206040978431702,
-      "eval_runtime": 0.9732,
-      "eval_samples_per_second": 72.955,
-      "eval_steps_per_second": 3.083,
+      "eval_accuracy": 0.647887323943662,
+      "eval_loss": 0.5588602423667908,
+      "eval_runtime": 0.9212,
+      "eval_samples_per_second": 77.074,
+      "eval_steps_per_second": 3.257,
       "step": 19
     },
     {
       "epoch": 6.153846153846154,
-      "grad_norm": 7.104850769042969,
+      "grad_norm": 2.3774571418762207,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.6334,
+      "loss": 0.6449,
       "step": 20
     },
     {
       "epoch": 6.769230769230769,
-      "eval_accuracy": 0.7323943661971831,
-      "eval_loss": 0.5421347618103027,
-      "eval_runtime": 0.9755,
-      "eval_samples_per_second": 72.781,
-      "eval_steps_per_second": 3.075,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.5559040904045105,
+      "eval_runtime": 0.942,
+      "eval_samples_per_second": 75.374,
+      "eval_steps_per_second": 3.185,
       "step": 22
     },
     {
       "epoch": 8.0,
       "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.5170249938964844,
-      "eval_runtime": 0.9786,
-      "eval_samples_per_second": 72.551,
-      "eval_steps_per_second": 3.066,
+      "eval_loss": 0.49099108576774597,
+      "eval_runtime": 0.9333,
+      "eval_samples_per_second": 76.075,
+      "eval_steps_per_second": 3.214,
       "step": 26
     },
     {
       "epoch": 8.923076923076923,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.5054832100868225,
-      "eval_runtime": 0.9817,
-      "eval_samples_per_second": 72.327,
-      "eval_steps_per_second": 3.056,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.4996400773525238,
+      "eval_runtime": 0.9274,
+      "eval_samples_per_second": 76.562,
+      "eval_steps_per_second": 3.235,
       "step": 29
     },
     {
       "epoch": 9.23076923076923,
-      "grad_norm": 3.090980052947998,
+      "grad_norm": 4.510417461395264,
       "learning_rate": 5e-05,
-      "loss": 0.5715,
+      "loss": 0.5494,
       "step": 30
     },
     {
       "epoch": 9.846153846153847,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.5054039359092712,
-      "eval_runtime": 0.9759,
-      "eval_samples_per_second": 72.751,
-      "eval_steps_per_second": 3.074,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.490305095911026,
+      "eval_runtime": 0.9249,
+      "eval_samples_per_second": 76.765,
+      "eval_steps_per_second": 3.244,
       "step": 32
     },
     {
       "epoch": 10.76923076923077,
-      "eval_accuracy": 0.676056338028169,
-      "eval_loss": 0.5726912617683411,
-      "eval_runtime": 0.98,
-      "eval_samples_per_second": 72.451,
-      "eval_steps_per_second": 3.061,
+      "eval_accuracy": 0.6619718309859155,
+      "eval_loss": 0.7331178188323975,
+      "eval_runtime": 0.925,
+      "eval_samples_per_second": 76.755,
+      "eval_steps_per_second": 3.243,
       "step": 35
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.45958590507507324,
-      "eval_runtime": 0.9845,
-      "eval_samples_per_second": 72.119,
-      "eval_steps_per_second": 3.047,
+      "eval_accuracy": 0.6901408450704225,
+      "eval_loss": 0.5053162574768066,
+      "eval_runtime": 0.9365,
+      "eval_samples_per_second": 75.815,
+      "eval_steps_per_second": 3.203,
       "step": 39
     },
     {
       "epoch": 12.307692307692308,
-      "grad_norm": 4.1016669273376465,
+      "grad_norm": 8.283596992492676,
       "learning_rate": 4.814814814814815e-05,
-      "loss": 0.512,
+      "loss": 0.4793,
       "step": 40
     },
     {
       "epoch": 12.923076923076923,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.4308609664440155,
-      "eval_runtime": 0.9866,
-      "eval_samples_per_second": 71.961,
-      "eval_steps_per_second": 3.041,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.4781084656715393,
+      "eval_runtime": 0.9253,
+      "eval_samples_per_second": 76.728,
+      "eval_steps_per_second": 3.242,
       "step": 42
     },
     {
       "epoch": 13.846153846153847,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.4322933256626129,
-      "eval_runtime": 0.9762,
-      "eval_samples_per_second": 72.729,
-      "eval_steps_per_second": 3.073,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.49969470500946045,
+      "eval_runtime": 0.9696,
+      "eval_samples_per_second": 73.226,
+      "eval_steps_per_second": 3.094,
       "step": 45
     },
     {
       "epoch": 14.76923076923077,
-      "eval_accuracy": 0.704225352112676,
-      "eval_loss": 0.49735942482948303,
-      "eval_runtime": 0.9782,
-      "eval_samples_per_second": 72.58,
-      "eval_steps_per_second": 3.067,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.5196781158447266,
+      "eval_runtime": 0.9285,
+      "eval_samples_per_second": 76.466,
+      "eval_steps_per_second": 3.231,
       "step": 48
     },
     {
       "epoch": 15.384615384615385,
-      "grad_norm": 3.7396647930145264,
+      "grad_norm": 2.751551389694214,
       "learning_rate": 4.62962962962963e-05,
-      "loss": 0.4498,
+      "loss": 0.4327,
       "step": 50
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.45005080103874207,
-      "eval_runtime": 0.9874,
-      "eval_samples_per_second": 71.903,
-      "eval_steps_per_second": 3.038,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.5338966846466064,
+      "eval_runtime": 0.9263,
+      "eval_samples_per_second": 76.647,
+      "eval_steps_per_second": 3.239,
       "step": 52
     },
     {
       "epoch": 16.923076923076923,
       "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.4345299005508423,
-      "eval_runtime": 0.9895,
-      "eval_samples_per_second": 71.753,
-      "eval_steps_per_second": 3.032,
+      "eval_loss": 0.4475107491016388,
+      "eval_runtime": 0.9361,
+      "eval_samples_per_second": 75.843,
+      "eval_steps_per_second": 3.205,
       "step": 55
     },
     {
       "epoch": 17.846153846153847,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.4991867244243622,
-      "eval_runtime": 0.9918,
-      "eval_samples_per_second": 71.586,
-      "eval_steps_per_second": 3.025,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.48082780838012695,
+      "eval_runtime": 0.9281,
+      "eval_samples_per_second": 76.496,
+      "eval_steps_per_second": 3.232,
       "step": 58
     },
     {
       "epoch": 18.46153846153846,
-      "grad_norm": 10.223017692565918,
+      "grad_norm": 4.935845375061035,
       "learning_rate": 4.4444444444444447e-05,
-      "loss": 0.3786,
+      "loss": 0.3747,
       "step": 60
     },
     {
       "epoch": 18.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.45762118697166443,
-      "eval_runtime": 0.9899,
-      "eval_samples_per_second": 71.727,
-      "eval_steps_per_second": 3.031,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.48682263493537903,
+      "eval_runtime": 0.9333,
+      "eval_samples_per_second": 76.07,
+      "eval_steps_per_second": 3.214,
       "step": 61
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.47582072019577026,
-      "eval_runtime": 0.9908,
-      "eval_samples_per_second": 71.656,
-      "eval_steps_per_second": 3.028,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.6205573081970215,
+      "eval_runtime": 0.9305,
+      "eval_samples_per_second": 76.299,
+      "eval_steps_per_second": 3.224,
       "step": 65
     },
     {
       "epoch": 20.923076923076923,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.6087139248847961,
-      "eval_runtime": 0.9941,
-      "eval_samples_per_second": 71.422,
-      "eval_steps_per_second": 3.018,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.5270622968673706,
+      "eval_runtime": 0.9317,
+      "eval_samples_per_second": 76.209,
+      "eval_steps_per_second": 3.22,
       "step": 68
     },
     {
       "epoch": 21.53846153846154,
-      "grad_norm": 8.329760551452637,
+      "grad_norm": 7.510834217071533,
       "learning_rate": 4.259259259259259e-05,
-      "loss": 0.3595,
+      "loss": 0.3474,
       "step": 70
     },
     {
       "epoch": 21.846153846153847,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5352088809013367,
-      "eval_runtime": 1.0035,
-      "eval_samples_per_second": 70.753,
-      "eval_steps_per_second": 2.99,
+      "eval_accuracy": 0.6901408450704225,
+      "eval_loss": 0.5227216482162476,
+      "eval_runtime": 0.9233,
+      "eval_samples_per_second": 76.898,
+      "eval_steps_per_second": 3.249,
       "step": 71
     },
     {
       "epoch": 22.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.4859572947025299,
-      "eval_runtime": 0.9997,
-      "eval_samples_per_second": 71.022,
-      "eval_steps_per_second": 3.001,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.5078049302101135,
+      "eval_runtime": 0.9339,
+      "eval_samples_per_second": 76.025,
+      "eval_steps_per_second": 3.212,
       "step": 74
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.46237364411354065,
-      "eval_runtime": 0.9896,
-      "eval_samples_per_second": 71.747,
-      "eval_steps_per_second": 3.032,
+      "eval_accuracy": 0.6901408450704225,
+      "eval_loss": 0.5842347741127014,
+      "eval_runtime": 0.9296,
+      "eval_samples_per_second": 76.378,
+      "eval_steps_per_second": 3.227,
       "step": 78
     },
     {
       "epoch": 24.615384615384617,
-      "grad_norm": 3.3443713188171387,
+      "grad_norm": 2.7830913066864014,
       "learning_rate": 4.074074074074074e-05,
-      "loss": 0.3018,
+      "loss": 0.267,
       "step": 80
     },
     {
       "epoch": 24.923076923076923,
-      "eval_accuracy": 0.7464788732394366,
-      "eval_loss": 0.5939740538597107,
-      "eval_runtime": 0.9934,
-      "eval_samples_per_second": 71.475,
-      "eval_steps_per_second": 3.02,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.601521372795105,
+      "eval_runtime": 0.9344,
+      "eval_samples_per_second": 75.981,
+      "eval_steps_per_second": 3.21,
       "step": 81
     },
     {
       "epoch": 25.846153846153847,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.48636069893836975,
-      "eval_runtime": 0.992,
-      "eval_samples_per_second": 71.573,
-      "eval_steps_per_second": 3.024,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.653270423412323,
+      "eval_runtime": 0.9381,
+      "eval_samples_per_second": 75.688,
+      "eval_steps_per_second": 3.198,
       "step": 84
     },
     {
       "epoch": 26.76923076923077,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.4621841013431549,
-      "eval_runtime": 0.9942,
-      "eval_samples_per_second": 71.413,
-      "eval_steps_per_second": 3.017,
+      "eval_accuracy": 0.7323943661971831,
+      "eval_loss": 0.5763739943504333,
+      "eval_runtime": 0.9321,
+      "eval_samples_per_second": 76.173,
+      "eval_steps_per_second": 3.219,
       "step": 87
     },
     {
       "epoch": 27.692307692307693,
-      "grad_norm": 6.436526298522949,
+      "grad_norm": 6.375743865966797,
       "learning_rate": 3.888888888888889e-05,
-      "loss": 0.2634,
+      "loss": 0.2333,
       "step": 90
     },
     {
       "epoch": 28.0,
       "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.4574315845966339,
-      "eval_runtime": 0.9876,
-      "eval_samples_per_second": 71.895,
-      "eval_steps_per_second": 3.038,
+      "eval_loss": 0.4862464666366577,
+      "eval_runtime": 0.9495,
+      "eval_samples_per_second": 74.776,
+      "eval_steps_per_second": 3.16,
       "step": 91
     },
     {
       "epoch": 28.923076923076923,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.4833536148071289,
-      "eval_runtime": 0.989,
-      "eval_samples_per_second": 71.792,
-      "eval_steps_per_second": 3.033,
+      "eval_accuracy": 0.7183098591549296,
+      "eval_loss": 0.6232538223266602,
+      "eval_runtime": 0.9407,
+      "eval_samples_per_second": 75.474,
+      "eval_steps_per_second": 3.189,
       "step": 94
     },
     {
       "epoch": 29.846153846153847,
-      "eval_accuracy": 0.7183098591549296,
-      "eval_loss": 0.7057759165763855,
-      "eval_runtime": 0.986,
-      "eval_samples_per_second": 72.009,
-      "eval_steps_per_second": 3.043,
+      "eval_accuracy": 0.7464788732394366,
+      "eval_loss": 0.754891037940979,
+      "eval_runtime": 0.9293,
+      "eval_samples_per_second": 76.398,
+      "eval_steps_per_second": 3.228,
       "step": 97
     },
     {
       "epoch": 30.76923076923077,
-      "grad_norm": 2.784860610961914,
+      "grad_norm": 2.766845226287842,
       "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.2705,
+      "loss": 0.2635,
       "step": 100
     },
     {
       "epoch": 30.76923076923077,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.6007413268089294,
-      "eval_runtime": 0.9852,
-      "eval_samples_per_second": 72.067,
-      "eval_steps_per_second": 3.045,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.48901528120040894,
+      "eval_runtime": 0.94,
+      "eval_samples_per_second": 75.533,
+      "eval_steps_per_second": 3.192,
       "step": 100
     },
     {
       "epoch": 32.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5075341463088989,
-      "eval_runtime": 0.9871,
-      "eval_samples_per_second": 71.927,
-      "eval_steps_per_second": 3.039,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5615856647491455,
+      "eval_runtime": 0.9614,
+      "eval_samples_per_second": 73.849,
+      "eval_steps_per_second": 3.12,
       "step": 104
     },
     {
       "epoch": 32.92307692307692,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.48067355155944824,
-      "eval_runtime": 0.9866,
-      "eval_samples_per_second": 71.961,
-      "eval_steps_per_second": 3.041,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.550128698348999,
+      "eval_runtime": 0.9431,
+      "eval_samples_per_second": 75.286,
+      "eval_steps_per_second": 3.181,
       "step": 107
     },
     {
       "epoch": 33.84615384615385,
-      "grad_norm": 2.4923012256622314,
+      "grad_norm": 4.919862747192383,
       "learning_rate": 3.518518518518519e-05,
-      "loss": 0.2222,
+      "loss": 0.192,
       "step": 110
     },
     {
       "epoch": 33.84615384615385,
       "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5054088234901428,
-      "eval_runtime": 0.9831,
-      "eval_samples_per_second": 72.222,
-      "eval_steps_per_second": 3.052,
+      "eval_loss": 0.4845168888568878,
+      "eval_runtime": 0.9359,
+      "eval_samples_per_second": 75.862,
+      "eval_steps_per_second": 3.205,
       "step": 110
     },
     {
       "epoch": 34.76923076923077,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.4999428987503052,
-      "eval_runtime": 0.9863,
-      "eval_samples_per_second": 71.989,
-      "eval_steps_per_second": 3.042,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.511615514755249,
+      "eval_runtime": 0.9324,
+      "eval_samples_per_second": 76.147,
+      "eval_steps_per_second": 3.217,
       "step": 113
     },
     {
       "epoch": 36.0,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.48362308740615845,
-      "eval_runtime": 0.9842,
-      "eval_samples_per_second": 72.137,
-      "eval_steps_per_second": 3.048,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5016549825668335,
+      "eval_runtime": 0.9526,
+      "eval_samples_per_second": 74.529,
+      "eval_steps_per_second": 3.149,
       "step": 117
     },
     {
       "epoch": 36.92307692307692,
-      "grad_norm": 5.310535907745361,
+      "grad_norm": 2.7866628170013428,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.1945,
+      "loss": 0.1763,
       "step": 120
     },
     {
       "epoch": 36.92307692307692,
-      "eval_accuracy": 0.7605633802816901,
-      "eval_loss": 0.5453178882598877,
-      "eval_runtime": 0.9863,
-      "eval_samples_per_second": 71.986,
-      "eval_steps_per_second": 3.042,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.47983619570732117,
+      "eval_runtime": 0.9561,
+      "eval_samples_per_second": 74.258,
+      "eval_steps_per_second": 3.138,
       "step": 120
     },
     {
       "epoch": 37.84615384615385,
-      "eval_accuracy": 0.8591549295774648,
-      "eval_loss": 0.4150727093219757,
-      "eval_runtime": 0.9838,
-      "eval_samples_per_second": 72.169,
-      "eval_steps_per_second": 3.049,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.5327603220939636,
+      "eval_runtime": 0.9331,
+      "eval_samples_per_second": 76.093,
+      "eval_steps_per_second": 3.215,
       "step": 123
     },
     {
       "epoch": 38.76923076923077,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.44067808985710144,
-      "eval_runtime": 0.9899,
-      "eval_samples_per_second": 71.725,
-      "eval_steps_per_second": 3.031,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.6393086910247803,
+      "eval_runtime": 0.9362,
+      "eval_samples_per_second": 75.837,
+      "eval_steps_per_second": 3.204,
       "step": 126
     },
     {
       "epoch": 40.0,
-      "grad_norm": 4.0999979972839355,
+      "grad_norm": 3.5983285903930664,
       "learning_rate": 3.148148148148148e-05,
-      "loss": 0.2075,
+      "loss": 0.172,
       "step": 130
     },
     {
       "epoch": 40.0,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.7065132856369019,
-      "eval_runtime": 1.0099,
-      "eval_samples_per_second": 70.307,
-      "eval_steps_per_second": 2.971,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5480958223342896,
+      "eval_runtime": 0.932,
+      "eval_samples_per_second": 76.18,
+      "eval_steps_per_second": 3.219,
       "step": 130
     },
     {
       "epoch": 40.92307692307692,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.4591623842716217,
-      "eval_runtime": 0.9906,
-      "eval_samples_per_second": 71.675,
-      "eval_steps_per_second": 3.029,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5867339372634888,
+      "eval_runtime": 0.9467,
+      "eval_samples_per_second": 74.998,
+      "eval_steps_per_second": 3.169,
       "step": 133
     },
     {
       "epoch": 41.84615384615385,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5610753893852234,
-      "eval_runtime": 0.9828,
-      "eval_samples_per_second": 72.239,
-      "eval_steps_per_second": 3.052,
+      "eval_accuracy": 0.704225352112676,
+      "eval_loss": 0.9222853779792786,
+      "eval_runtime": 0.9412,
+      "eval_samples_per_second": 75.436,
+      "eval_steps_per_second": 3.187,
       "step": 136
     },
     {
       "epoch": 42.76923076923077,
       "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.49726197123527527,
-      "eval_runtime": 0.993,
-      "eval_samples_per_second": 71.5,
-      "eval_steps_per_second": 3.021,
+      "eval_loss": 0.6261908411979675,
+      "eval_runtime": 0.9446,
+      "eval_samples_per_second": 75.163,
+      "eval_steps_per_second": 3.176,
       "step": 139
     },
     {
       "epoch": 43.07692307692308,
-      "grad_norm": 4.051938056945801,
+      "grad_norm": 6.744588851928711,
       "learning_rate": 2.962962962962963e-05,
-      "loss": 0.1845,
+      "loss": 0.1832,
       "step": 140
     },
     {
       "epoch": 44.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.4929070770740509,
-      "eval_runtime": 0.9861,
-      "eval_samples_per_second": 71.999,
-      "eval_steps_per_second": 3.042,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.6090897917747498,
+      "eval_runtime": 0.9537,
+      "eval_samples_per_second": 74.447,
+      "eval_steps_per_second": 3.146,
       "step": 143
     },
     {
       "epoch": 44.92307692307692,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5009937882423401,
-      "eval_runtime": 0.9872,
-      "eval_samples_per_second": 71.918,
-      "eval_steps_per_second": 3.039,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.5836533308029175,
+      "eval_runtime": 0.936,
+      "eval_samples_per_second": 75.851,
+      "eval_steps_per_second": 3.205,
       "step": 146
     },
     {
       "epoch": 45.84615384615385,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5320884585380554,
-      "eval_runtime": 0.9793,
-      "eval_samples_per_second": 72.499,
-      "eval_steps_per_second": 3.063,
+      "eval_accuracy": 0.7605633802816901,
+      "eval_loss": 0.5465303063392639,
+      "eval_runtime": 0.9256,
+      "eval_samples_per_second": 76.703,
+      "eval_steps_per_second": 3.241,
       "step": 149
     },
     {
       "epoch": 46.15384615384615,
-      "grad_norm": 3.662769317626953,
+      "grad_norm": 5.820557594299316,
       "learning_rate": 2.777777777777778e-05,
-      "loss": 0.1399,
+      "loss": 0.1641,
       "step": 150
     },
     {
       "epoch": 46.76923076923077,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5591984391212463,
-      "eval_runtime": 0.9835,
-      "eval_samples_per_second": 72.192,
-      "eval_steps_per_second": 3.05,
+      "eval_accuracy": 0.7746478873239436,
+      "eval_loss": 0.6745445132255554,
+      "eval_runtime": 0.9266,
+      "eval_samples_per_second": 76.622,
+      "eval_steps_per_second": 3.238,
       "step": 152
     },
     {
       "epoch": 48.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5693766474723816,
-      "eval_runtime": 0.9878,
-      "eval_samples_per_second": 71.874,
-      "eval_steps_per_second": 3.037,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5397518873214722,
+      "eval_runtime": 0.9297,
+      "eval_samples_per_second": 76.369,
+      "eval_steps_per_second": 3.227,
       "step": 156
     },
     {
       "epoch": 48.92307692307692,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5988211035728455,
-      "eval_runtime": 1.0243,
-      "eval_samples_per_second": 69.318,
-      "eval_steps_per_second": 2.929,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5387323498725891,
+      "eval_runtime": 0.943,
+      "eval_samples_per_second": 75.293,
+      "eval_steps_per_second": 3.181,
       "step": 159
     },
     {
       "epoch": 49.23076923076923,
-      "grad_norm": 3.793342113494873,
+      "grad_norm": 1.1889840364456177,
       "learning_rate": 2.5925925925925925e-05,
-      "loss": 0.1575,
+      "loss": 0.1366,
       "step": 160
     },
     {
       "epoch": 49.84615384615385,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5647031664848328,
-      "eval_runtime": 0.9806,
-      "eval_samples_per_second": 72.406,
-      "eval_steps_per_second": 3.059,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5737282037734985,
+      "eval_runtime": 0.9304,
+      "eval_samples_per_second": 76.315,
+      "eval_steps_per_second": 3.225,
       "step": 162
     },
     {
       "epoch": 50.76923076923077,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5079817175865173,
-      "eval_runtime": 0.9853,
-      "eval_samples_per_second": 72.059,
-      "eval_steps_per_second": 3.045,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5254746675491333,
+      "eval_runtime": 0.9335,
+      "eval_samples_per_second": 76.061,
+      "eval_steps_per_second": 3.214,
       "step": 165
     },
     {
       "epoch": 52.0,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.560810923576355,
-      "eval_runtime": 0.9736,
-      "eval_samples_per_second": 72.929,
-      "eval_steps_per_second": 3.081,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.6485558152198792,
+      "eval_runtime": 0.9328,
+      "eval_samples_per_second": 76.117,
+      "eval_steps_per_second": 3.216,
       "step": 169
     },
     {
       "epoch": 52.30769230769231,
-      "grad_norm": 2.4892890453338623,
+      "grad_norm": 3.566678285598755,
       "learning_rate": 2.4074074074074074e-05,
-      "loss": 0.1628,
+      "loss": 0.149,
       "step": 170
     },
     {
       "epoch": 52.92307692307692,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5770761370658875,
-      "eval_runtime": 1.0087,
-      "eval_samples_per_second": 70.386,
-      "eval_steps_per_second": 2.974,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5404361486434937,
+      "eval_runtime": 0.9413,
+      "eval_samples_per_second": 75.426,
+      "eval_steps_per_second": 3.187,
       "step": 172
     },
     {
       "epoch": 53.84615384615385,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5863744616508484,
-      "eval_runtime": 0.9881,
-      "eval_samples_per_second": 71.858,
-      "eval_steps_per_second": 3.036,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5654813051223755,
+      "eval_runtime": 0.9373,
+      "eval_samples_per_second": 75.746,
+      "eval_steps_per_second": 3.201,
       "step": 175
     },
     {
       "epoch": 54.76923076923077,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.6020362973213196,
-      "eval_runtime": 0.9958,
-      "eval_samples_per_second": 71.297,
-      "eval_steps_per_second": 3.013,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.6121301054954529,
+      "eval_runtime": 0.9398,
+      "eval_samples_per_second": 75.547,
+      "eval_steps_per_second": 3.192,
       "step": 178
     },
     {
       "epoch": 55.38461538461539,
-      "grad_norm": 3.317295789718628,
+      "grad_norm": 3.30234694480896,
       "learning_rate": 2.2222222222222223e-05,
-      "loss": 0.1247,
+      "loss": 0.1196,
       "step": 180
     },
     {
       "epoch": 56.0,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.6511205434799194,
-      "eval_runtime": 0.9852,
-      "eval_samples_per_second": 72.069,
-      "eval_steps_per_second": 3.045,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.6182261109352112,
+      "eval_runtime": 0.9409,
+      "eval_samples_per_second": 75.462,
+      "eval_steps_per_second": 3.189,
       "step": 182
     },
     {
       "epoch": 56.92307692307692,
-      "eval_accuracy": 0.7746478873239436,
-      "eval_loss": 0.60261470079422,
-      "eval_runtime": 0.9823,
-      "eval_samples_per_second": 72.283,
-      "eval_steps_per_second": 3.054,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.6174789071083069,
+      "eval_runtime": 0.944,
+      "eval_samples_per_second": 75.21,
+      "eval_steps_per_second": 3.178,
       "step": 185
     },
     {
       "epoch": 57.84615384615385,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.6173571348190308,
-      "eval_runtime": 0.9893,
-      "eval_samples_per_second": 71.768,
-      "eval_steps_per_second": 3.032,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5920690894126892,
+      "eval_runtime": 0.9418,
+      "eval_samples_per_second": 75.387,
+      "eval_steps_per_second": 3.185,
       "step": 188
     },
     {
       "epoch": 58.46153846153846,
-      "grad_norm": 2.5961012840270996,
+      "grad_norm": 2.1697278022766113,
       "learning_rate": 2.037037037037037e-05,
-      "loss": 0.1343,
+      "loss": 0.1202,
       "step": 190
     },
     {
       "epoch": 58.76923076923077,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.647166907787323,
-      "eval_runtime": 0.9806,
-      "eval_samples_per_second": 72.401,
-      "eval_steps_per_second": 3.059,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5953019261360168,
+      "eval_runtime": 0.9459,
+      "eval_samples_per_second": 75.059,
+      "eval_steps_per_second": 3.172,
       "step": 191
     },
     {
       "epoch": 60.0,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.6184616684913635,
-      "eval_runtime": 0.9804,
-      "eval_samples_per_second": 72.418,
-      "eval_steps_per_second": 3.06,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.6065444350242615,
+      "eval_runtime": 0.9408,
+      "eval_samples_per_second": 75.471,
+      "eval_steps_per_second": 3.189,
       "step": 195
     },
     {
       "epoch": 60.92307692307692,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5780375003814697,
-      "eval_runtime": 0.9922,
-      "eval_samples_per_second": 71.561,
-      "eval_steps_per_second": 3.024,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5448285341262817,
+      "eval_runtime": 0.9326,
+      "eval_samples_per_second": 76.133,
+      "eval_steps_per_second": 3.217,
       "step": 198
     },
     {
       "epoch": 61.53846153846154,
-      "grad_norm": 2.727761745452881,
+      "grad_norm": 3.645998001098633,
       "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.1404,
+      "loss": 0.1289,
       "step": 200
     },
     {
       "epoch": 61.84615384615385,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5088328123092651,
-      "eval_runtime": 0.9897,
-      "eval_samples_per_second": 71.741,
-      "eval_steps_per_second": 3.031,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5258444547653198,
+      "eval_runtime": 0.9405,
+      "eval_samples_per_second": 75.491,
+      "eval_steps_per_second": 3.19,
       "step": 201
     },
     {
       "epoch": 62.76923076923077,
       "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.47978872060775757,
-      "eval_runtime": 0.9923,
-      "eval_samples_per_second": 71.554,
-      "eval_steps_per_second": 3.023,
+      "eval_loss": 0.5439636707305908,
+      "eval_runtime": 0.9352,
+      "eval_samples_per_second": 75.923,
+      "eval_steps_per_second": 3.208,
       "step": 204
     },
     {
       "epoch": 64.0,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.518284022808075,
-      "eval_runtime": 0.987,
-      "eval_samples_per_second": 71.938,
-      "eval_steps_per_second": 3.04,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.6082146763801575,
+      "eval_runtime": 0.9371,
+      "eval_samples_per_second": 75.764,
+      "eval_steps_per_second": 3.201,
       "step": 208
     },
     {
       "epoch": 64.61538461538461,
-      "grad_norm": 3.1862618923187256,
+      "grad_norm": 3.992480993270874,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.1199,
+      "loss": 0.1262,
       "step": 210
     },
     {
       "epoch": 64.92307692307692,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.49685367941856384,
-      "eval_runtime": 0.9844,
-      "eval_samples_per_second": 72.125,
-      "eval_steps_per_second": 3.048,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.6357871890068054,
+      "eval_runtime": 0.9342,
+      "eval_samples_per_second": 76.002,
+      "eval_steps_per_second": 3.211,
       "step": 211
     },
     {
       "epoch": 65.84615384615384,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.4816421866416931,
-      "eval_runtime": 0.9837,
-      "eval_samples_per_second": 72.175,
-      "eval_steps_per_second": 3.05,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5981662273406982,
+      "eval_runtime": 0.9361,
+      "eval_samples_per_second": 75.846,
+      "eval_steps_per_second": 3.205,
       "step": 214
     },
     {
       "epoch": 66.76923076923077,
-      "eval_accuracy": 0.8591549295774648,
-      "eval_loss": 0.47663062810897827,
-      "eval_runtime": 1.0005,
-      "eval_samples_per_second": 70.964,
-      "eval_steps_per_second": 2.998,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5850167870521545,
+      "eval_runtime": 0.965,
+      "eval_samples_per_second": 73.577,
+      "eval_steps_per_second": 3.109,
       "step": 217
     },
     {
       "epoch": 67.6923076923077,
-      "grad_norm": 3.8146462440490723,
+      "grad_norm": 2.3649914264678955,
       "learning_rate": 1.4814814814814815e-05,
-      "loss": 0.1321,
+      "loss": 0.124,
       "step": 220
     },
     {
       "epoch": 68.0,
-      "eval_accuracy": 0.8450704225352113,
-      "eval_loss": 0.4813210368156433,
-      "eval_runtime": 0.9887,
-      "eval_samples_per_second": 71.814,
-      "eval_steps_per_second": 3.034,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5732671022415161,
+      "eval_runtime": 0.9312,
+      "eval_samples_per_second": 76.246,
+      "eval_steps_per_second": 3.222,
       "step": 221
     },
     {
       "epoch": 68.92307692307692,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5145000219345093,
-      "eval_runtime": 1.007,
-      "eval_samples_per_second": 70.51,
-      "eval_steps_per_second": 2.979,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5631324648857117,
+      "eval_runtime": 0.9307,
+      "eval_samples_per_second": 76.283,
+      "eval_steps_per_second": 3.223,
       "step": 224
     },
     {
       "epoch": 69.84615384615384,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5145190954208374,
-      "eval_runtime": 0.9767,
-      "eval_samples_per_second": 72.694,
-      "eval_steps_per_second": 3.072,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5375062227249146,
+      "eval_runtime": 0.9261,
+      "eval_samples_per_second": 76.662,
+      "eval_steps_per_second": 3.239,
       "step": 227
     },
     {
       "epoch": 70.76923076923077,
-      "grad_norm": 4.591117858886719,
+      "grad_norm": 2.5271434783935547,
       "learning_rate": 1.2962962962962962e-05,
-      "loss": 0.1183,
+      "loss": 0.1208,
       "step": 230
     },
     {
       "epoch": 70.76923076923077,
       "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.6660017371177673,
-      "eval_runtime": 0.9893,
-      "eval_samples_per_second": 71.768,
-      "eval_steps_per_second": 3.032,
+      "eval_loss": 0.5157544016838074,
+      "eval_runtime": 0.9385,
+      "eval_samples_per_second": 75.656,
+      "eval_steps_per_second": 3.197,
       "step": 230
     },
     {
       "epoch": 72.0,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5571023225784302,
-      "eval_runtime": 0.9834,
-      "eval_samples_per_second": 72.201,
-      "eval_steps_per_second": 3.051,
+      "eval_accuracy": 0.8169014084507042,
+      "eval_loss": 0.5430620908737183,
+      "eval_runtime": 0.9444,
+      "eval_samples_per_second": 75.177,
+      "eval_steps_per_second": 3.176,
       "step": 234
     },
     {
       "epoch": 72.92307692307692,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5238744020462036,
-      "eval_runtime": 1.0028,
-      "eval_samples_per_second": 70.8,
-      "eval_steps_per_second": 2.992,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5099484920501709,
+      "eval_runtime": 0.9479,
+      "eval_samples_per_second": 74.905,
+      "eval_steps_per_second": 3.165,
       "step": 237
     },
     {
       "epoch": 73.84615384615384,
-      "grad_norm": 3.053189277648926,
+      "grad_norm": 2.769530773162842,
       "learning_rate": 1.1111111111111112e-05,
-      "loss": 0.1239,
+      "loss": 0.1126,
       "step": 240
     },
     {
       "epoch": 73.84615384615384,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5199792981147766,
-      "eval_runtime": 0.9798,
-      "eval_samples_per_second": 72.462,
-      "eval_steps_per_second": 3.062,
+      "eval_accuracy": 0.7887323943661971,
+      "eval_loss": 0.5803037285804749,
+      "eval_runtime": 0.9361,
+      "eval_samples_per_second": 75.848,
+      "eval_steps_per_second": 3.205,
       "step": 240
     },
     {
       "epoch": 74.76923076923077,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5257282257080078,
-      "eval_runtime": 1.0082,
-      "eval_samples_per_second": 70.42,
-      "eval_steps_per_second": 2.975,
+      "eval_accuracy": 0.8028169014084507,
+      "eval_loss": 0.5416265726089478,
+      "eval_runtime": 0.9298,
+      "eval_samples_per_second": 76.361,
+      "eval_steps_per_second": 3.227,
       "step": 243
     },
     {
       "epoch": 76.0,
-      "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.5306513905525208,
-      "eval_runtime": 0.9844,
-      "eval_samples_per_second": 72.127,
-      "eval_steps_per_second": 3.048,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5834996104240417,
+      "eval_runtime": 0.9267,
+      "eval_samples_per_second": 76.616,
+      "eval_steps_per_second": 3.237,
       "step": 247
     },
     {
       "epoch": 76.92307692307692,
-      "grad_norm": 3.1847753524780273,
+      "grad_norm": 5.331729412078857,
       "learning_rate": 9.259259259259259e-06,
-      "loss": 0.1105,
+      "loss": 0.1089,
       "step": 250
     },
     {
       "epoch": 76.92307692307692,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5484855771064758,
-      "eval_runtime": 0.9865,
-      "eval_samples_per_second": 71.972,
-      "eval_steps_per_second": 3.041,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5923494100570679,
+      "eval_runtime": 0.9331,
+      "eval_samples_per_second": 76.087,
+      "eval_steps_per_second": 3.215,
       "step": 250
     },
     {
       "epoch": 77.84615384615384,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5527692437171936,
-      "eval_runtime": 0.9821,
-      "eval_samples_per_second": 72.297,
-      "eval_steps_per_second": 3.055,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5228320360183716,
+      "eval_runtime": 0.9263,
+      "eval_samples_per_second": 76.652,
+      "eval_steps_per_second": 3.239,
       "step": 253
     },
     {
       "epoch": 78.76923076923077,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5314151644706726,
-      "eval_runtime": 0.9715,
-      "eval_samples_per_second": 73.082,
-      "eval_steps_per_second": 3.088,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5466788411140442,
+      "eval_runtime": 0.9292,
+      "eval_samples_per_second": 76.41,
+      "eval_steps_per_second": 3.229,
       "step": 256
     },
     {
       "epoch": 80.0,
-      "grad_norm": 2.466536521911621,
+      "grad_norm": 3.553982973098755,
       "learning_rate": 7.4074074074074075e-06,
-      "loss": 0.1008,
+      "loss": 0.0965,
       "step": 260
     },
     {
       "epoch": 80.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.6122465133666992,
-      "eval_runtime": 0.9778,
-      "eval_samples_per_second": 72.613,
-      "eval_steps_per_second": 3.068,
+      "eval_accuracy": 0.8732394366197183,
+      "eval_loss": 0.5155981183052063,
+      "eval_runtime": 0.9345,
+      "eval_samples_per_second": 75.975,
+      "eval_steps_per_second": 3.21,
       "step": 260
     },
     {
       "epoch": 80.92307692307692,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5795039534568787,
-      "eval_runtime": 0.9924,
-      "eval_samples_per_second": 71.546,
-      "eval_steps_per_second": 3.023,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5082412958145142,
+      "eval_runtime": 0.9291,
+      "eval_samples_per_second": 76.42,
+      "eval_steps_per_second": 3.229,
       "step": 263
     },
     {
       "epoch": 81.84615384615384,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5279837250709534,
-      "eval_runtime": 0.9966,
-      "eval_samples_per_second": 71.242,
-      "eval_steps_per_second": 3.01,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5070865154266357,
+      "eval_runtime": 0.9317,
+      "eval_samples_per_second": 76.206,
+      "eval_steps_per_second": 3.22,
       "step": 266
     },
     {
       "epoch": 82.76923076923077,
-      "eval_accuracy": 0.8028169014084507,
-      "eval_loss": 0.5277493596076965,
-      "eval_runtime": 0.9864,
-      "eval_samples_per_second": 71.981,
-      "eval_steps_per_second": 3.041,
+      "eval_accuracy": 0.8591549295774648,
+      "eval_loss": 0.5070091485977173,
+      "eval_runtime": 0.932,
+      "eval_samples_per_second": 76.176,
+      "eval_steps_per_second": 3.219,
       "step": 269
     },
     {
       "epoch": 83.07692307692308,
-      "grad_norm": 3.0015382766723633,
+      "grad_norm": 1.3318876028060913,
       "learning_rate": 5.555555555555556e-06,
-      "loss": 0.1023,
+      "loss": 0.0947,
       "step": 270
     },
     {
       "epoch": 84.0,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5573487877845764,
-      "eval_runtime": 0.9858,
-      "eval_samples_per_second": 72.021,
-      "eval_steps_per_second": 3.043,
+      "eval_accuracy": 0.8591549295774648,
+      "eval_loss": 0.5268462300300598,
+      "eval_runtime": 0.9334,
+      "eval_samples_per_second": 76.068,
+      "eval_steps_per_second": 3.214,
       "step": 273
     },
     {
       "epoch": 84.92307692307692,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.5539447069168091,
-      "eval_runtime": 0.9998,
-      "eval_samples_per_second": 71.015,
-      "eval_steps_per_second": 3.001,
+      "eval_accuracy": 0.8591549295774648,
+      "eval_loss": 0.528256356716156,
+      "eval_runtime": 0.938,
+      "eval_samples_per_second": 75.693,
+      "eval_steps_per_second": 3.198,
       "step": 276
     },
     {
       "epoch": 85.84615384615384,
-      "eval_accuracy": 0.7887323943661971,
-      "eval_loss": 0.5272416472434998,
-      "eval_runtime": 0.9912,
-      "eval_samples_per_second": 71.631,
-      "eval_steps_per_second": 3.027,
+      "eval_accuracy": 0.8450704225352113,
+      "eval_loss": 0.5261198282241821,
+      "eval_runtime": 0.9373,
+      "eval_samples_per_second": 75.752,
+      "eval_steps_per_second": 3.201,
       "step": 279
     },
     {
       "epoch": 86.15384615384616,
-      "grad_norm": 1.397282600402832,
+      "grad_norm": 1.2073172330856323,
       "learning_rate": 3.7037037037037037e-06,
-      "loss": 0.089,
+      "loss": 0.0751,
       "step": 280
     },
     {
       "epoch": 86.76923076923077,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.508567750453949,
-      "eval_runtime": 1.0068,
-      "eval_samples_per_second": 70.518,
-      "eval_steps_per_second": 2.98,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.528601884841919,
+      "eval_runtime": 0.9192,
+      "eval_samples_per_second": 77.242,
+      "eval_steps_per_second": 3.264,
       "step": 282
     },
     {
       "epoch": 88.0,
       "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.4971120059490204,
-      "eval_runtime": 0.9989,
-      "eval_samples_per_second": 71.081,
-      "eval_steps_per_second": 3.003,
+      "eval_loss": 0.5414741039276123,
+      "eval_runtime": 0.9317,
+      "eval_samples_per_second": 76.207,
+      "eval_steps_per_second": 3.22,
       "step": 286
     },
     {
       "epoch": 88.92307692307692,
       "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.4941565692424774,
-      "eval_runtime": 1.0024,
-      "eval_samples_per_second": 70.83,
-      "eval_steps_per_second": 2.993,
+      "eval_loss": 0.5511464476585388,
+      "eval_runtime": 0.9273,
+      "eval_samples_per_second": 76.567,
+      "eval_steps_per_second": 3.235,
       "step": 289
     },
     {
       "epoch": 89.23076923076923,
-      "grad_norm": 2.480825185775757,
+      "grad_norm": 1.936850905418396,
       "learning_rate": 1.8518518518518519e-06,
-      "loss": 0.0919,
+      "loss": 0.0912,
       "step": 290
     },
     {
       "epoch": 89.84615384615384,
       "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.49278023838996887,
-      "eval_runtime": 0.9826,
-      "eval_samples_per_second": 72.255,
-      "eval_steps_per_second": 3.053,
+      "eval_loss": 0.5541730523109436,
+      "eval_runtime": 0.9344,
+      "eval_samples_per_second": 75.984,
+      "eval_steps_per_second": 3.211,
       "step": 292
     },
     {
       "epoch": 90.76923076923077,
-      "eval_accuracy": 0.8169014084507042,
-      "eval_loss": 0.4940688908100128,
-      "eval_runtime": 1.0126,
-      "eval_samples_per_second": 70.117,
-      "eval_steps_per_second": 2.963,
+      "eval_accuracy": 0.8309859154929577,
+      "eval_loss": 0.5464221239089966,
+      "eval_runtime": 0.9256,
+      "eval_samples_per_second": 76.711,
+      "eval_steps_per_second": 3.241,
       "step": 295
     },
     {
       "epoch": 92.0,
       "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.496449738740921,
-      "eval_runtime": 1.0016,
-      "eval_samples_per_second": 70.887,
-      "eval_steps_per_second": 2.995,
+      "eval_loss": 0.5409799218177795,
+      "eval_runtime": 0.9348,
+      "eval_samples_per_second": 75.953,
+      "eval_steps_per_second": 3.209,
       "step": 299
     },
     {
       "epoch": 92.3076923076923,
-      "grad_norm": 2.8444457054138184,
+      "grad_norm": 2.42252516746521,
       "learning_rate": 0.0,
-      "loss": 0.1044,
+      "loss": 0.104,
       "step": 300
     },
     {
       "epoch": 92.3076923076923,
       "eval_accuracy": 0.8309859154929577,
-      "eval_loss": 0.49665746092796326,
-      "eval_runtime": 1.0024,
-      "eval_samples_per_second": 70.829,
-      "eval_steps_per_second": 2.993,
+      "eval_loss": 0.540703535079956,
+      "eval_runtime": 0.9779,
+      "eval_samples_per_second": 72.607,
+      "eval_steps_per_second": 3.068,
       "step": 300
     },
     {
       "epoch": 92.3076923076923,
       "step": 300,
       "total_flos": 2.8402872494292173e+18,
-      "train_loss": 0.2401055637995402,
-      "train_runtime": 1630.2519,
-      "train_samples_per_second": 24.352,
-      "train_steps_per_second": 0.184
+      "train_loss": 0.23062967757383981,
+      "train_runtime": 1594.1954,
+      "train_samples_per_second": 24.903,
+      "train_steps_per_second": 0.188
     },
     {
       "epoch": 92.3076923076923,
-      "eval_accuracy": 0.8591549295774648,
-      "eval_loss": 0.4150727093219757,
-      "eval_runtime": 1.0438,
-      "eval_samples_per_second": 68.024,
-      "eval_steps_per_second": 2.874,
+      "eval_accuracy": 0.8732394366197183,
+      "eval_loss": 0.5155981183052063,
+      "eval_runtime": 0.9923,
+      "eval_samples_per_second": 71.551,
+      "eval_steps_per_second": 3.023,
       "step": 300
     }
   ],