🍻 cheers

README.md

@@ -2,6 +2,7 @@
 license: apache-2.0
 base_model: microsoft/swin-tiny-patch4-window7-224
 tags:
+- image-classification
 - generated_from_trainer
 metrics:
 - accuracy
@@ -15,9 +16,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 # swin-tiny-patch4-window7-224-finetuned_ASL_Isolated_Swin_dataset2
 
-This model is a fine-tuned version of [microsoft/swin-tiny-patch4-window7-224](https://huggingface.co/microsoft/swin-tiny-patch4-window7-224) on an unknown dataset.
+This model is a fine-tuned version of [microsoft/swin-tiny-patch4-window7-224](https://huggingface.co/microsoft/swin-tiny-patch4-window7-224) on the ASL_Isolated_Swin_dataset dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.0909
+- Loss: 0.1269
 - Accuracy: 0.9769
 
 ## Model description

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 20.0,
-    "eval_accuracy": 0.9846153846153847,
-    "eval_loss": 0.05584708973765373,
-    "eval_runtime": 2.0579,
-    "eval_samples_per_second": 126.342,
-    "eval_steps_per_second": 16.036,
+    "eval_accuracy": 0.9769230769230769,
+    "eval_loss": 0.12689848244190216,
+    "eval_runtime": 1.9212,
+    "eval_samples_per_second": 135.332,
+    "eval_steps_per_second": 17.177,
     "total_flos": 7.30261234607063e+17,
-    "train_loss": 0.5342570722103119,
-    "train_runtime": 409.3983,
-    "train_samples_per_second": 71.715,
-    "train_steps_per_second": 4.494
+    "train_loss": 0.5132673807766126,
+    "train_runtime": 395.7262,
+    "train_samples_per_second": 74.193,
+    "train_steps_per_second": 4.65
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
-    "eval_accuracy": 0.9846153846153847,
-    "eval_loss": 0.05584708973765373,
-    "eval_runtime": 2.0579,
-    "eval_samples_per_second": 126.342,
-    "eval_steps_per_second": 16.036
+    "eval_accuracy": 0.9769230769230769,
+    "eval_loss": 0.12689848244190216,
+    "eval_runtime": 1.9212,
+    "eval_samples_per_second": 135.332,
+    "eval_steps_per_second": 17.177
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
     "total_flos": 7.30261234607063e+17,
-    "train_loss": 0.5342570722103119,
-    "train_runtime": 409.3983,
-    "train_samples_per_second": 71.715,
-    "train_steps_per_second": 4.494
+    "train_loss": 0.5132673807766126,
+    "train_runtime": 395.7262,
+    "train_samples_per_second": 74.193,
+    "train_steps_per_second": 4.65
 }

trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.9846153846153847,
+  "best_metric": 0.9769230769230769,
   "best_model_checkpoint": "swin-tiny-patch4-window7-224-finetuned_ASL_Isolated_Swin_dataset2/checkpoint-1700",
   "epoch": 20.0,
   "eval_steps": 100,
@@ -10,1278 +10,1278 @@
   "log_history": [
     {
       "epoch": 0.11,
-      "learning_rate": 0.00019891304347826087,
-      "loss": 3.3365,
+      "learning_rate": 0.0001991304347826087,
+      "loss": 3.3413,
       "step": 10
     },
     {
       "epoch": 0.22,
-      "learning_rate": 0.00019782608695652175,
-      "loss": 3.2188,
+      "learning_rate": 0.00019804347826086956,
+      "loss": 3.1778,
       "step": 20
     },
     {
       "epoch": 0.33,
-      "learning_rate": 0.00019673913043478263,
-      "loss": 3.1756,
+      "learning_rate": 0.00019695652173913044,
+      "loss": 3.1311,
       "step": 30
     },
     {
       "epoch": 0.43,
-      "learning_rate": 0.0001956521739130435,
-      "loss": 2.9023,
+      "learning_rate": 0.0001958695652173913,
+      "loss": 2.9107,
       "step": 40
     },
     {
       "epoch": 0.54,
-      "learning_rate": 0.00019467391304347825,
-      "loss": 2.5517,
+      "learning_rate": 0.00019478260869565218,
+      "loss": 2.3763,
       "step": 50
     },
     {
       "epoch": 0.65,
-      "learning_rate": 0.00019358695652173916,
-      "loss": 2.4131,
+      "learning_rate": 0.00019369565217391307,
+      "loss": 2.0376,
       "step": 60
     },
     {
       "epoch": 0.76,
-      "learning_rate": 0.00019250000000000002,
-      "loss": 2.0957,
+      "learning_rate": 0.00019260869565217392,
+      "loss": 1.9444,
       "step": 70
     },
     {
       "epoch": 0.87,
-      "learning_rate": 0.00019141304347826087,
-      "loss": 1.8856,
+      "learning_rate": 0.00019152173913043478,
+      "loss": 1.8087,
       "step": 80
     },
     {
       "epoch": 0.98,
-      "learning_rate": 0.00019032608695652176,
-      "loss": 1.7217,
+      "learning_rate": 0.00019043478260869566,
+      "loss": 1.6835,
       "step": 90
     },
     {
       "epoch": 1.09,
-      "learning_rate": 0.0001892391304347826,
-      "loss": 1.6758,
+      "learning_rate": 0.00018934782608695655,
+      "loss": 1.5439,
       "step": 100
     },
     {
       "epoch": 1.09,
-      "eval_accuracy": 0.5769230769230769,
-      "eval_loss": 1.320558786392212,
-      "eval_runtime": 2.846,
-      "eval_samples_per_second": 91.356,
-      "eval_steps_per_second": 11.595,
+      "eval_accuracy": 0.5538461538461539,
+      "eval_loss": 1.4188454151153564,
+      "eval_runtime": 1.8531,
+      "eval_samples_per_second": 140.308,
+      "eval_steps_per_second": 17.808,
       "step": 100
     },
     {
       "epoch": 1.2,
-      "learning_rate": 0.0001881521739130435,
-      "loss": 1.4695,
+      "learning_rate": 0.0001882608695652174,
+      "loss": 1.3483,
       "step": 110
     },
     {
       "epoch": 1.3,
-      "learning_rate": 0.00018706521739130435,
-      "loss": 1.311,
+      "learning_rate": 0.00018717391304347826,
+      "loss": 1.3837,
       "step": 120
     },
     {
       "epoch": 1.41,
-      "learning_rate": 0.00018597826086956523,
-      "loss": 1.2741,
+      "learning_rate": 0.00018608695652173914,
+      "loss": 1.2886,
       "step": 130
     },
     {
       "epoch": 1.52,
-      "learning_rate": 0.0001848913043478261,
-      "loss": 1.3444,
+      "learning_rate": 0.00018500000000000002,
+      "loss": 1.1915,
       "step": 140
     },
     {
       "epoch": 1.63,
-      "learning_rate": 0.00018380434782608697,
-      "loss": 1.0445,
+      "learning_rate": 0.00018391304347826088,
+      "loss": 1.1142,
       "step": 150
     },
     {
       "epoch": 1.74,
       "learning_rate": 0.00018282608695652174,
-      "loss": 1.0316,
+      "loss": 0.9908,
       "step": 160
     },
     {
       "epoch": 1.85,
       "learning_rate": 0.00018173913043478262,
-      "loss": 1.0877,
+      "loss": 0.9563,
       "step": 170
     },
     {
       "epoch": 1.96,
       "learning_rate": 0.00018065217391304348,
-      "loss": 1.1561,
+      "loss": 0.9761,
       "step": 180
     },
     {
       "epoch": 2.07,
       "learning_rate": 0.00017956521739130436,
-      "loss": 1.0726,
+      "loss": 1.0636,
       "step": 190
     },
     {
       "epoch": 2.17,
       "learning_rate": 0.00017847826086956522,
-      "loss": 1.0717,
+      "loss": 0.8646,
       "step": 200
     },
     {
       "epoch": 2.17,
-      "eval_accuracy": 0.8153846153846154,
-      "eval_loss": 0.6482219696044922,
-      "eval_runtime": 1.825,
-      "eval_samples_per_second": 142.465,
-      "eval_steps_per_second": 18.082,
+      "eval_accuracy": 0.8884615384615384,
+      "eval_loss": 0.45421722531318665,
+      "eval_runtime": 2.2052,
+      "eval_samples_per_second": 117.904,
+      "eval_steps_per_second": 14.965,
       "step": 200
     },
     {
       "epoch": 2.28,
       "learning_rate": 0.0001773913043478261,
-      "loss": 0.744,
+      "loss": 0.7407,
       "step": 210
     },
     {
       "epoch": 2.39,
       "learning_rate": 0.00017630434782608695,
-      "loss": 0.9959,
+      "loss": 0.8695,
       "step": 220
     },
     {
       "epoch": 2.5,
       "learning_rate": 0.00017521739130434784,
-      "loss": 0.9342,
+      "loss": 0.7102,
       "step": 230
     },
     {
       "epoch": 2.61,
       "learning_rate": 0.00017413043478260872,
-      "loss": 0.8275,
+      "loss": 0.9596,
       "step": 240
     },
     {
       "epoch": 2.72,
       "learning_rate": 0.00017304347826086958,
-      "loss": 0.8388,
+      "loss": 0.8462,
       "step": 250
     },
     {
       "epoch": 2.83,
       "learning_rate": 0.00017195652173913043,
-      "loss": 0.8996,
+      "loss": 0.8556,
       "step": 260
     },
     {
       "epoch": 2.93,
       "learning_rate": 0.00017086956521739132,
-      "loss": 0.6616,
+      "loss": 0.664,
       "step": 270
     },
     {
       "epoch": 3.04,
       "learning_rate": 0.0001697826086956522,
-      "loss": 0.7714,
+      "loss": 0.8349,
       "step": 280
     },
     {
       "epoch": 3.15,
       "learning_rate": 0.00016869565217391306,
-      "loss": 0.7296,
+      "loss": 0.6526,
       "step": 290
     },
     {
       "epoch": 3.26,
       "learning_rate": 0.0001676086956521739,
-      "loss": 0.627,
+      "loss": 0.5485,
       "step": 300
     },
     {
       "epoch": 3.26,
-      "eval_accuracy": 0.8653846153846154,
-      "eval_loss": 0.4485549330711365,
-      "eval_runtime": 2.5537,
-      "eval_samples_per_second": 101.814,
-      "eval_steps_per_second": 12.923,
+      "eval_accuracy": 0.8538461538461538,
+      "eval_loss": 0.4102769196033478,
+      "eval_runtime": 1.7696,
+      "eval_samples_per_second": 146.93,
+      "eval_steps_per_second": 18.649,
       "step": 300
     },
     {
       "epoch": 3.37,
-      "learning_rate": 0.0001666304347826087,
-      "loss": 0.8732,
+      "learning_rate": 0.0001665217391304348,
+      "loss": 0.701,
       "step": 310
     },
     {
       "epoch": 3.48,
-      "learning_rate": 0.00016554347826086958,
-      "loss": 0.8933,
+      "learning_rate": 0.00016543478260869568,
+      "loss": 0.725,
       "step": 320
     },
     {
       "epoch": 3.59,
-      "learning_rate": 0.00016445652173913044,
-      "loss": 0.8468,
+      "learning_rate": 0.00016434782608695653,
+      "loss": 0.7088,
       "step": 330
     },
     {
       "epoch": 3.7,
-      "learning_rate": 0.0001633695652173913,
-      "loss": 0.6468,
+      "learning_rate": 0.0001632608695652174,
+      "loss": 0.5813,
       "step": 340
     },
     {
       "epoch": 3.8,
-      "learning_rate": 0.00016228260869565218,
-      "loss": 0.7266,
+      "learning_rate": 0.00016217391304347827,
+      "loss": 0.7197,
       "step": 350
     },
     {
       "epoch": 3.91,
-      "learning_rate": 0.00016119565217391306,
-      "loss": 0.8522,
+      "learning_rate": 0.00016108695652173913,
+      "loss": 0.7112,
       "step": 360
     },
     {
       "epoch": 4.02,
-      "learning_rate": 0.00016010869565217392,
-      "loss": 0.5888,
+      "learning_rate": 0.00016,
+      "loss": 0.6083,
       "step": 370
     },
     {
       "epoch": 4.13,
-      "learning_rate": 0.00015902173913043478,
-      "loss": 0.685,
+      "learning_rate": 0.00015891304347826087,
+      "loss": 0.6653,
       "step": 380
     },
     {
       "epoch": 4.24,
-      "learning_rate": 0.00015793478260869566,
-      "loss": 0.6937,
+      "learning_rate": 0.00015782608695652175,
+      "loss": 0.6319,
       "step": 390
     },
     {
       "epoch": 4.35,
-      "learning_rate": 0.00015684782608695654,
-      "loss": 0.5397,
+      "learning_rate": 0.0001567391304347826,
+      "loss": 0.5082,
       "step": 400
     },
     {
       "epoch": 4.35,
-      "eval_accuracy": 0.8923076923076924,
-      "eval_loss": 0.2977767884731293,
-      "eval_runtime": 1.8479,
-      "eval_samples_per_second": 140.701,
-      "eval_steps_per_second": 17.858,
+      "eval_accuracy": 0.8961538461538462,
+      "eval_loss": 0.29254817962646484,
+      "eval_runtime": 2.5312,
+      "eval_samples_per_second": 102.718,
+      "eval_steps_per_second": 13.037,
       "step": 400
     },
     {
       "epoch": 4.46,
-      "learning_rate": 0.0001557608695652174,
-      "loss": 0.6058,
+      "learning_rate": 0.00015565217391304346,
+      "loss": 0.6872,
       "step": 410
     },
     {
       "epoch": 4.57,
-      "learning_rate": 0.00015467391304347828,
-      "loss": 0.553,
+      "learning_rate": 0.00015456521739130437,
+      "loss": 0.5159,
       "step": 420
     },
     {
       "epoch": 4.67,
-      "learning_rate": 0.00015358695652173914,
-      "loss": 0.4621,
+      "learning_rate": 0.00015347826086956523,
+      "loss": 0.4834,
       "step": 430
     },
     {
       "epoch": 4.78,
-      "learning_rate": 0.0001525,
-      "loss": 0.5002,
+      "learning_rate": 0.0001523913043478261,
+      "loss": 0.4467,
       "step": 440
     },
     {
       "epoch": 4.89,
-      "learning_rate": 0.00015141304347826088,
-      "loss": 0.4833,
+      "learning_rate": 0.00015130434782608694,
+      "loss": 0.5091,
       "step": 450
     },
     {
       "epoch": 5.0,
-      "learning_rate": 0.00015032608695652176,
-      "loss": 0.5952,
+      "learning_rate": 0.00015021739130434785,
+      "loss": 0.701,
       "step": 460
     },
     {
       "epoch": 5.11,
-      "learning_rate": 0.00014923913043478262,
-      "loss": 0.5049,
+      "learning_rate": 0.0001491304347826087,
+      "loss": 0.5314,
       "step": 470
     },
     {
       "epoch": 5.22,
-      "learning_rate": 0.00014815217391304347,
-      "loss": 0.5219,
+      "learning_rate": 0.00014804347826086957,
+      "loss": 0.5484,
       "step": 480
     },
     {
       "epoch": 5.33,
-      "learning_rate": 0.00014706521739130436,
-      "loss": 0.6222,
+      "learning_rate": 0.00014695652173913045,
+      "loss": 0.4856,
       "step": 490
     },
     {
       "epoch": 5.43,
-      "learning_rate": 0.00014597826086956524,
-      "loss": 0.537,
+      "learning_rate": 0.0001458695652173913,
+      "loss": 0.5302,
       "step": 500
     },
     {
       "epoch": 5.43,
-      "eval_accuracy": 0.9423076923076923,
-      "eval_loss": 0.151279479265213,
-      "eval_runtime": 1.8783,
-      "eval_samples_per_second": 138.42,
-      "eval_steps_per_second": 17.569,
+      "eval_accuracy": 0.926923076923077,
+      "eval_loss": 0.24707049131393433,
+      "eval_runtime": 1.8365,
+      "eval_samples_per_second": 141.57,
+      "eval_steps_per_second": 17.969,
       "step": 500
     },
     {
       "epoch": 5.54,
-      "learning_rate": 0.0001448913043478261,
-      "loss": 0.498,
+      "learning_rate": 0.0001447826086956522,
+      "loss": 0.4876,
       "step": 510
     },
     {
       "epoch": 5.65,
-      "learning_rate": 0.00014380434782608695,
-      "loss": 0.5532,
+      "learning_rate": 0.00014369565217391304,
+      "loss": 0.6267,
       "step": 520
     },
     {
       "epoch": 5.76,
-      "learning_rate": 0.00014271739130434783,
-      "loss": 0.5434,
+      "learning_rate": 0.00014260869565217393,
+      "loss": 0.5022,
       "step": 530
     },
     {
       "epoch": 5.87,
-      "learning_rate": 0.00014163043478260872,
-      "loss": 0.6535,
+      "learning_rate": 0.00014152173913043478,
+      "loss": 0.6259,
       "step": 540
     },
     {
       "epoch": 5.98,
-      "learning_rate": 0.00014054347826086957,
-      "loss": 0.4965,
+      "learning_rate": 0.00014043478260869567,
+      "loss": 0.588,
       "step": 550
     },
     {
       "epoch": 6.09,
-      "learning_rate": 0.00013945652173913043,
-      "loss": 0.6712,
+      "learning_rate": 0.00013934782608695652,
+      "loss": 0.6801,
       "step": 560
     },
     {
       "epoch": 6.2,
-      "learning_rate": 0.0001383695652173913,
-      "loss": 0.5683,
+      "learning_rate": 0.0001382608695652174,
+      "loss": 0.6062,
       "step": 570
     },
     {
       "epoch": 6.3,
-      "learning_rate": 0.00013728260869565217,
-      "loss": 0.3377,
+      "learning_rate": 0.00013717391304347826,
+      "loss": 0.3015,
       "step": 580
     },
     {
       "epoch": 6.41,
-      "learning_rate": 0.00013619565217391305,
-      "loss": 0.482,
+      "learning_rate": 0.00013608695652173912,
+      "loss": 0.5071,
       "step": 590
     },
     {
       "epoch": 6.52,
-      "learning_rate": 0.00013510869565217394,
-      "loss": 0.3766,
+      "learning_rate": 0.00013500000000000003,
+      "loss": 0.4072,
       "step": 600
     },
     {
       "epoch": 6.52,
-      "eval_accuracy": 0.8846153846153846,
-      "eval_loss": 0.4736884534358978,
-      "eval_runtime": 1.8513,
-      "eval_samples_per_second": 140.44,
-      "eval_steps_per_second": 17.825,
+      "eval_accuracy": 0.9230769230769231,
+      "eval_loss": 0.2676173746585846,
+      "eval_runtime": 2.5959,
+      "eval_samples_per_second": 100.157,
+      "eval_steps_per_second": 12.712,
       "step": 600
     },
     {
       "epoch": 6.63,
-      "learning_rate": 0.0001340217391304348,
-      "loss": 0.5068,
+      "learning_rate": 0.00013391304347826088,
+      "loss": 0.451,
       "step": 610
     },
     {
       "epoch": 6.74,
-      "learning_rate": 0.00013293478260869565,
-      "loss": 0.3019,
+      "learning_rate": 0.00013282608695652174,
+      "loss": 0.3797,
       "step": 620
     },
     {
       "epoch": 6.85,
-      "learning_rate": 0.00013184782608695653,
-      "loss": 0.4852,
+      "learning_rate": 0.0001317391304347826,
+      "loss": 0.415,
       "step": 630
     },
     {
       "epoch": 6.96,
-      "learning_rate": 0.00013076086956521741,
-      "loss": 0.55,
+      "learning_rate": 0.00013065217391304348,
+      "loss": 0.4253,
       "step": 640
     },
     {
       "epoch": 7.07,
-      "learning_rate": 0.00012967391304347827,
-      "loss": 0.4122,
+      "learning_rate": 0.00012956521739130436,
+      "loss": 0.4366,
       "step": 650
     },
     {
       "epoch": 7.17,
-      "learning_rate": 0.00012858695652173913,
-      "loss": 0.3566,
+      "learning_rate": 0.00012847826086956522,
+      "loss": 0.3499,
       "step": 660
     },
     {
       "epoch": 7.28,
-      "learning_rate": 0.0001275,
-      "loss": 0.4279,
+      "learning_rate": 0.0001273913043478261,
+      "loss": 0.4012,
       "step": 670
     },
     {
       "epoch": 7.39,
-      "learning_rate": 0.0001264130434782609,
-      "loss": 0.3295,
+      "learning_rate": 0.00012630434782608696,
+      "loss": 0.414,
       "step": 680
     },
     {
       "epoch": 7.5,
-      "learning_rate": 0.00012532608695652175,
-      "loss": 0.5113,
+      "learning_rate": 0.00012521739130434784,
+      "loss": 0.5052,
       "step": 690
     },
     {
       "epoch": 7.61,
-      "learning_rate": 0.0001242391304347826,
-      "loss": 0.3994,
+      "learning_rate": 0.0001241304347826087,
+      "loss": 0.4424,
       "step": 700
     },
     {
       "epoch": 7.61,
-      "eval_accuracy": 0.9115384615384615,
-      "eval_loss": 0.3060305714607239,
-      "eval_runtime": 1.8463,
-      "eval_samples_per_second": 140.821,
-      "eval_steps_per_second": 17.873,
+      "eval_accuracy": 0.9038461538461539,
+      "eval_loss": 0.4149817228317261,
+      "eval_runtime": 1.8169,
+      "eval_samples_per_second": 143.1,
+      "eval_steps_per_second": 18.163,
       "step": 700
     },
     {
       "epoch": 7.72,
-      "learning_rate": 0.0001231521739130435,
-      "loss": 0.5572,
+      "learning_rate": 0.00012304347826086958,
+      "loss": 0.5183,
       "step": 710
     },
     {
       "epoch": 7.83,
-      "learning_rate": 0.00012206521739130434,
-      "loss": 0.2986,
+      "learning_rate": 0.00012195652173913044,
+      "loss": 0.2976,
       "step": 720
     },
     {
       "epoch": 7.93,
-      "learning_rate": 0.00012097826086956523,
-      "loss": 0.5146,
+      "learning_rate": 0.00012086956521739131,
+      "loss": 0.5184,
       "step": 730
     },
     {
       "epoch": 8.04,
-      "learning_rate": 0.0001198913043478261,
-      "loss": 0.4029,
+      "learning_rate": 0.00011978260869565219,
+      "loss": 0.4194,
       "step": 740
     },
     {
       "epoch": 8.15,
-      "learning_rate": 0.00011880434782608695,
-      "loss": 0.3901,
+      "learning_rate": 0.00011869565217391305,
+      "loss": 0.3032,
       "step": 750
     },
     {
       "epoch": 8.26,
-      "learning_rate": 0.00011771739130434782,
-      "loss": 0.5112,
+      "learning_rate": 0.00011760869565217392,
+      "loss": 0.5633,
       "step": 760
     },
     {
       "epoch": 8.37,
-      "learning_rate": 0.0001166304347826087,
-      "loss": 0.2704,
+      "learning_rate": 0.00011652173913043479,
+      "loss": 0.3037,
       "step": 770
     },
     {
       "epoch": 8.48,
-      "learning_rate": 0.00011554347826086958,
-      "loss": 0.3839,
+      "learning_rate": 0.00011543478260869567,
+      "loss": 0.3617,
       "step": 780
     },
     {
       "epoch": 8.59,
-      "learning_rate": 0.00011445652173913045,
-      "loss": 0.4988,
+      "learning_rate": 0.00011434782608695654,
+      "loss": 0.4318,
       "step": 790
     },
     {
       "epoch": 8.7,
-      "learning_rate": 0.0001133695652173913,
-      "loss": 0.2139,
+      "learning_rate": 0.0001132608695652174,
+      "loss": 0.3409,
       "step": 800
     },
     {
       "epoch": 8.7,
-      "eval_accuracy": 0.9576923076923077,
-      "eval_loss": 0.13448870182037354,
-      "eval_runtime": 1.9169,
-      "eval_samples_per_second": 135.638,
-      "eval_steps_per_second": 17.216,
+      "eval_accuracy": 0.9538461538461539,
+      "eval_loss": 0.19218704104423523,
+      "eval_runtime": 2.6334,
+      "eval_samples_per_second": 98.732,
+      "eval_steps_per_second": 12.531,
       "step": 800
     },
     {
       "epoch": 8.8,
-      "learning_rate": 0.00011228260869565217,
-      "loss": 0.2857,
+      "learning_rate": 0.00011217391304347826,
+      "loss": 0.2645,
       "step": 810
     },
     {
       "epoch": 8.91,
-      "learning_rate": 0.00011119565217391305,
-      "loss": 0.2416,
+      "learning_rate": 0.00011108695652173912,
+      "loss": 0.2775,
       "step": 820
     },
     {
       "epoch": 9.02,
-      "learning_rate": 0.00011010869565217392,
-      "loss": 0.4803,
+      "learning_rate": 0.00011000000000000002,
+      "loss": 0.4254,
       "step": 830
     },
     {
       "epoch": 9.13,
-      "learning_rate": 0.00010902173913043478,
-      "loss": 0.4416,
+      "learning_rate": 0.00010891304347826087,
+      "loss": 0.3383,
       "step": 840
     },
     {
       "epoch": 9.24,
-      "learning_rate": 0.00010793478260869565,
-      "loss": 0.3271,
+      "learning_rate": 0.00010782608695652174,
+      "loss": 0.3443,
       "step": 850
     },
     {
       "epoch": 9.35,
-      "learning_rate": 0.00010684782608695653,
-      "loss": 0.2261,
+      "learning_rate": 0.00010673913043478261,
+      "loss": 0.2455,
       "step": 860
     },
     {
       "epoch": 9.46,
-      "learning_rate": 0.0001057608695652174,
-      "loss": 0.2063,
+      "learning_rate": 0.00010565217391304347,
+      "loss": 0.2475,
       "step": 870
     },
     {
       "epoch": 9.57,
-      "learning_rate": 0.00010467391304347827,
-      "loss": 0.4865,
+      "learning_rate": 0.00010456521739130437,
+      "loss": 0.4743,
       "step": 880
     },
     {
       "epoch": 9.67,
-      "learning_rate": 0.00010358695652173913,
-      "loss": 0.4903,
+      "learning_rate": 0.00010347826086956522,
+      "loss": 0.4935,
       "step": 890
     },
     {
       "epoch": 9.78,
-      "learning_rate": 0.0001025,
-      "loss": 0.2995,
+      "learning_rate": 0.00010239130434782609,
+      "loss": 0.3046,
       "step": 900
     },
     {
       "epoch": 9.78,
-      "eval_accuracy": 0.95,
-      "eval_loss": 0.155814990401268,
-      "eval_runtime": 1.8528,
-      "eval_samples_per_second": 140.328,
-      "eval_steps_per_second": 17.811,
+      "eval_accuracy": 0.9461538461538461,
+      "eval_loss": 0.1916845738887787,
+      "eval_runtime": 1.7809,
+      "eval_samples_per_second": 145.992,
+      "eval_steps_per_second": 18.53,
       "step": 900
     },
     {
       "epoch": 9.89,
-      "learning_rate": 0.00010141304347826088,
-      "loss": 0.2847,
+      "learning_rate": 0.00010130434782608695,
+      "loss": 0.2834,
       "step": 910
     },
     {
       "epoch": 10.0,
-      "learning_rate": 0.00010032608695652175,
-      "loss": 0.2913,
+      "learning_rate": 0.00010021739130434784,
+      "loss": 0.4026,
       "step": 920
     },
     {
       "epoch": 10.11,
-      "learning_rate": 9.923913043478261e-05,
-      "loss": 0.2779,
+      "learning_rate": 9.91304347826087e-05,
+      "loss": 0.3004,
       "step": 930
     },
     {
       "epoch": 10.22,
-      "learning_rate": 9.815217391304349e-05,
-      "loss": 0.2205,
+      "learning_rate": 9.804347826086957e-05,
+      "loss": 0.2161,
       "step": 940
     },
     {
       "epoch": 10.33,
-      "learning_rate": 9.706521739130435e-05,
-      "loss": 0.405,
+      "learning_rate": 9.695652173913044e-05,
+      "loss": 0.3859,
       "step": 950
     },
     {
       "epoch": 10.43,
-      "learning_rate": 9.597826086956522e-05,
-      "loss": 0.3613,
+      "learning_rate": 9.586956521739131e-05,
+      "loss": 0.2911,
       "step": 960
     },
     {
       "epoch": 10.54,
-      "learning_rate": 9.48913043478261e-05,
-      "loss": 0.2381,
+      "learning_rate": 9.478260869565218e-05,
+      "loss": 0.2693,
       "step": 970
     },
     {
       "epoch": 10.65,
-      "learning_rate": 9.380434782608696e-05,
-      "loss": 0.224,
+      "learning_rate": 9.369565217391305e-05,
+      "loss": 0.2366,
       "step": 980
     },
     {
       "epoch": 10.76,
-      "learning_rate": 9.271739130434784e-05,
-      "loss": 0.2544,
+      "learning_rate": 9.260869565217392e-05,
+      "loss": 0.3161,
       "step": 990
     },
     {
       "epoch": 10.87,
-      "learning_rate": 9.16304347826087e-05,
-      "loss": 0.2835,
+      "learning_rate": 9.152173913043479e-05,
+      "loss": 0.2911,
       "step": 1000
     },
     {
       "epoch": 10.87,
-      "eval_accuracy": 0.9730769230769231,
-      "eval_loss": 0.09426813572645187,
-      "eval_runtime": 2.2351,
-      "eval_samples_per_second": 116.327,
-      "eval_steps_per_second": 14.765,
+      "eval_accuracy": 0.9423076923076923,
+      "eval_loss": 0.2271503061056137,
+      "eval_runtime": 2.5147,
+      "eval_samples_per_second": 103.391,
+      "eval_steps_per_second": 13.123,
       "step": 1000
     },
     {
       "epoch": 10.98,
-      "learning_rate": 9.054347826086958e-05,
-      "loss": 0.4605,
+      "learning_rate": 9.043478260869566e-05,
+      "loss": 0.4276,
       "step": 1010
     },
     {
       "epoch": 11.09,
-      "learning_rate": 8.945652173913043e-05,
-      "loss": 0.2259,
+      "learning_rate": 8.934782608695653e-05,
+      "loss": 0.245,
       "step": 1020
     },
     {
       "epoch": 11.2,
-      "learning_rate": 8.83695652173913e-05,
-      "loss": 0.2241,
+      "learning_rate": 8.82608695652174e-05,
+      "loss": 0.2376,
       "step": 1030
     },
     {
       "epoch": 11.3,
-      "learning_rate": 8.728260869565217e-05,
-      "loss": 0.2173,
+      "learning_rate": 8.717391304347827e-05,
+      "loss": 0.1843,
       "step": 1040
     },
     {
       "epoch": 11.41,
-      "learning_rate": 8.619565217391304e-05,
-      "loss": 0.3629,
+      "learning_rate": 8.608695652173914e-05,
+      "loss": 0.3411,
       "step": 1050
     },
     {
       "epoch": 11.52,
-      "learning_rate": 8.510869565217393e-05,
-      "loss": 0.6198,
+      "learning_rate": 8.5e-05,
+      "loss": 0.5081,
       "step": 1060
     },
     {
       "epoch": 11.63,
-      "learning_rate": 8.402173913043478e-05,
-      "loss": 0.5369,
+      "learning_rate": 8.391304347826088e-05,
+      "loss": 0.2889,
       "step": 1070
     },
     {
       "epoch": 11.74,
-      "learning_rate": 8.293478260869567e-05,
-      "loss": 0.304,
+      "learning_rate": 8.282608695652175e-05,
+      "loss": 0.2588,
       "step": 1080
     },
     {
       "epoch": 11.85,
-      "learning_rate": 8.184782608695652e-05,
-      "loss": 0.2756,
+      "learning_rate": 8.173913043478262e-05,
+      "loss": 0.1457,
       "step": 1090
     },
     {
       "epoch": 11.96,
       "learning_rate": 8.076086956521739e-05,
-      "loss": 0.3089,
+      "loss": 0.269,
       "step": 1100
     },
     {
       "epoch": 11.96,
-      "eval_accuracy": 0.9576923076923077,
-      "eval_loss": 0.09128668904304504,
-      "eval_runtime": 1.8693,
-      "eval_samples_per_second": 139.093,
-      "eval_steps_per_second": 17.654,
+      "eval_accuracy": 0.9692307692307692,
+      "eval_loss": 0.07221826165914536,
+      "eval_runtime": 1.7687,
+      "eval_samples_per_second": 146.998,
+      "eval_steps_per_second": 18.657,
       "step": 1100
     },
     {
       "epoch": 12.07,
       "learning_rate": 7.967391304347826e-05,
-      "loss": 0.3491,
+      "loss": 0.3124,
       "step": 1110
     },
     {
       "epoch": 12.17,
       "learning_rate": 7.858695652173913e-05,
-      "loss": 0.1677,
+      "loss": 0.1558,
       "step": 1120
     },
     {
       "epoch": 12.28,
       "learning_rate": 7.75e-05,
-      "loss": 0.2795,
+      "loss": 0.3082,
       "step": 1130
     },
     {
       "epoch": 12.39,
       "learning_rate": 7.641304347826087e-05,
-      "loss": 0.2714,
+      "loss": 0.2492,
       "step": 1140
     },
     {
       "epoch": 12.5,
       "learning_rate": 7.532608695652175e-05,
-      "loss": 0.302,
+      "loss": 0.3847,
       "step": 1150
     },
     {
       "epoch": 12.61,
       "learning_rate": 7.423913043478261e-05,
-      "loss": 0.3564,
+      "loss": 0.3452,
       "step": 1160
     },
     {
       "epoch": 12.72,
-      "learning_rate": 7.32608695652174e-05,
-      "loss": 0.3085,
+      "learning_rate": 7.315217391304349e-05,
+      "loss": 0.3171,
       "step": 1170
     },
     {
       "epoch": 12.83,
-      "learning_rate": 7.217391304347827e-05,
-      "loss": 0.2515,
+      "learning_rate": 7.206521739130435e-05,
+      "loss": 0.3295,
       "step": 1180
     },
     {
       "epoch": 12.93,
-      "learning_rate": 7.108695652173914e-05,
-      "loss": 0.3075,
+      "learning_rate": 7.097826086956522e-05,
+      "loss": 0.3387,
       "step": 1190
     },
     {
       "epoch": 13.04,
-      "learning_rate": 7e-05,
-      "loss": 0.3632,
+      "learning_rate": 6.989130434782609e-05,
+      "loss": 0.3709,
       "step": 1200
     },
     {
       "epoch": 13.04,
-      "eval_accuracy": 0.9692307692307692,
-      "eval_loss": 0.08875690400600433,
-      "eval_runtime": 2.7533,
-      "eval_samples_per_second": 94.431,
-      "eval_steps_per_second": 11.986,
+      "eval_accuracy": 0.9653846153846154,
+      "eval_loss": 0.14726483821868896,
+      "eval_runtime": 2.5212,
+      "eval_samples_per_second": 103.126,
+      "eval_steps_per_second": 13.089,
       "step": 1200
     },
     {
       "epoch": 13.15,
-      "learning_rate": 6.891304347826088e-05,
-      "loss": 0.3888,
+      "learning_rate": 6.880434782608696e-05,
+      "loss": 0.3647,
       "step": 1210
     },
     {
       "epoch": 13.26,
-      "learning_rate": 6.782608695652173e-05,
-      "loss": 0.1683,
+      "learning_rate": 6.771739130434783e-05,
+      "loss": 0.137,
       "step": 1220
     },
     {
       "epoch": 13.37,
-      "learning_rate": 6.673913043478262e-05,
-      "loss": 0.3672,
+      "learning_rate": 6.66304347826087e-05,
+      "loss": 0.3562,
       "step": 1230
     },
     {
       "epoch": 13.48,
-      "learning_rate": 6.565217391304349e-05,
-      "loss": 0.335,
+      "learning_rate": 6.554347826086957e-05,
+      "loss": 0.323,
       "step": 1240
     },
     {
       "epoch": 13.59,
-      "learning_rate": 6.456521739130436e-05,
-      "loss": 0.1861,
+      "learning_rate": 6.445652173913044e-05,
+      "loss": 0.2173,
       "step": 1250
     },
     {
       "epoch": 13.7,
-      "learning_rate": 6.347826086956523e-05,
-      "loss": 0.3386,
+      "learning_rate": 6.33695652173913e-05,
+      "loss": 0.315,
       "step": 1260
     },
     {
       "epoch": 13.8,
-      "learning_rate": 6.239130434782608e-05,
-      "loss": 0.2222,
+      "learning_rate": 6.228260869565218e-05,
+      "loss": 0.274,
       "step": 1270
     },
     {
       "epoch": 13.91,
-      "learning_rate": 6.130434782608696e-05,
-      "loss": 0.1714,
+      "learning_rate": 6.119565217391304e-05,
+      "loss": 0.1757,
       "step": 1280
     },
     {
       "epoch": 14.02,
-      "learning_rate": 6.021739130434783e-05,
-      "loss": 0.2699,
+      "learning_rate": 6.010869565217392e-05,
+      "loss": 0.1968,
       "step": 1290
     },
     {
       "epoch": 14.13,
-      "learning_rate": 5.9130434782608704e-05,
-      "loss": 0.327,
+      "learning_rate": 5.9021739130434784e-05,
+      "loss": 0.3443,
       "step": 1300
     },
     {
       "epoch": 14.13,
-      "eval_accuracy": 0.9807692307692307,
-      "eval_loss": 0.10383553802967072,
-      "eval_runtime": 1.9077,
-      "eval_samples_per_second": 136.293,
-      "eval_steps_per_second": 17.299,
+      "eval_accuracy": 0.9615384615384616,
+      "eval_loss": 0.15447035431861877,
+      "eval_runtime": 1.7932,
+      "eval_samples_per_second": 144.988,
+      "eval_steps_per_second": 18.402,
       "step": 1300
     },
     {
       "epoch": 14.24,
-      "learning_rate": 5.804347826086957e-05,
-      "loss": 0.224,
+      "learning_rate": 5.793478260869566e-05,
+      "loss": 0.225,
       "step": 1310
     },
     {
       "epoch": 14.35,
-      "learning_rate": 5.695652173913044e-05,
-      "loss": 0.378,
+      "learning_rate": 5.6847826086956524e-05,
+      "loss": 0.336,
       "step": 1320
     },
     {
       "epoch": 14.46,
-      "learning_rate": 5.5869565217391306e-05,
-      "loss": 0.2503,
+      "learning_rate": 5.5760869565217386e-05,
+      "loss": 0.2073,
       "step": 1330
     },
     {
       "epoch": 14.57,
-      "learning_rate": 5.478260869565217e-05,
-      "loss": 0.1633,
+      "learning_rate": 5.467391304347826e-05,
+      "loss": 0.2144,
       "step": 1340
     },
     {
       "epoch": 14.67,
-      "learning_rate": 5.3695652173913046e-05,
-      "loss": 0.2973,
+      "learning_rate": 5.3586956521739126e-05,
+      "loss": 0.3399,
       "step": 1350
     },
     {
       "epoch": 14.78,
-      "learning_rate": 5.260869565217391e-05,
-      "loss": 0.1887,
+      "learning_rate": 5.25e-05,
+      "loss": 0.2516,
       "step": 1360
     },
     {
       "epoch": 14.89,
-      "learning_rate": 5.1521739130434785e-05,
-      "loss": 0.2953,
+      "learning_rate": 5.141304347826087e-05,
+      "loss": 0.2494,
       "step": 1370
     },
     {
       "epoch": 15.0,
-      "learning_rate": 5.0434782608695655e-05,
-      "loss": 0.248,
+      "learning_rate": 5.032608695652175e-05,
+      "loss": 0.2213,
       "step": 1380
     },
     {
       "epoch": 15.11,
-      "learning_rate": 4.9347826086956524e-05,
-      "loss": 0.152,
+      "learning_rate": 4.923913043478261e-05,
+      "loss": 0.1834,
       "step": 1390
     },
     {
       "epoch": 15.22,
-      "learning_rate": 4.8260869565217394e-05,
-      "loss": 0.313,
+      "learning_rate": 4.815217391304348e-05,
+      "loss": 0.187,
       "step": 1400
     },
     {
       "epoch": 15.22,
       "eval_accuracy": 0.9730769230769231,
-      "eval_loss": 0.09758734703063965,
-      "eval_runtime": 2.57,
-      "eval_samples_per_second": 101.167,
-      "eval_steps_per_second": 12.84,
+      "eval_loss": 0.10598226636648178,
+      "eval_runtime": 2.4502,
+      "eval_samples_per_second": 106.116,
+      "eval_steps_per_second": 13.469,
       "step": 1400
     },
     {
       "epoch": 15.33,
-      "learning_rate": 4.7173913043478264e-05,
-      "loss": 0.2301,
+      "learning_rate": 4.706521739130435e-05,
+      "loss": 0.1387,
       "step": 1410
     },
     {
       "epoch": 15.43,
-      "learning_rate": 4.608695652173913e-05,
-      "loss": 0.1671,
+      "learning_rate": 4.597826086956522e-05,
+      "loss": 0.1653,
       "step": 1420
     },
     {
       "epoch": 15.54,
-      "learning_rate": 4.5e-05,
-      "loss": 0.2399,
+      "learning_rate": 4.489130434782609e-05,
+      "loss": 0.2012,
       "step": 1430
     },
     {
       "epoch": 15.65,
-      "learning_rate": 4.391304347826087e-05,
-      "loss": 0.278,
+      "learning_rate": 4.380434782608696e-05,
+      "loss": 0.224,
       "step": 1440
     },
     {
       "epoch": 15.76,
-      "learning_rate": 4.282608695652174e-05,
-      "loss": 0.3062,
+      "learning_rate": 4.271739130434783e-05,
+      "loss": 0.3341,
       "step": 1450
     },
     {
       "epoch": 15.87,
-      "learning_rate": 4.1739130434782605e-05,
-      "loss": 0.3704,
+      "learning_rate": 4.16304347826087e-05,
+      "loss": 0.3279,
       "step": 1460
     },
     {
       "epoch": 15.98,
-      "learning_rate": 4.065217391304348e-05,
-      "loss": 0.231,
+      "learning_rate": 4.054347826086957e-05,
+      "loss": 0.2218,
       "step": 1470
     },
     {
       "epoch": 16.09,
-      "learning_rate": 3.956521739130435e-05,
-      "loss": 0.2339,
+      "learning_rate": 3.945652173913044e-05,
+      "loss": 0.2184,
       "step": 1480
     },
     {
       "epoch": 16.2,
-      "learning_rate": 3.847826086956522e-05,
-      "loss": 0.2079,
+      "learning_rate": 3.836956521739131e-05,
+      "loss": 0.2069,
       "step": 1490
     },
     {
       "epoch": 16.3,
-      "learning_rate": 3.739130434782609e-05,
-      "loss": 0.1752,
+      "learning_rate": 3.728260869565218e-05,
+      "loss": 0.1879,
       "step": 1500
     },
     {
       "epoch": 16.3,
-      "eval_accuracy": 0.9807692307692307,
-      "eval_loss": 0.050352372229099274,
-      "eval_runtime": 1.9714,
-      "eval_samples_per_second": 131.889,
-      "eval_steps_per_second": 16.74,
+      "eval_accuracy": 0.9692307692307692,
+      "eval_loss": 0.11237060278654099,
+      "eval_runtime": 1.8389,
+      "eval_samples_per_second": 141.393,
+      "eval_steps_per_second": 17.946,
       "step": 1500
     },
     {
       "epoch": 16.41,
-      "learning_rate": 3.630434782608696e-05,
-      "loss": 0.1359,
+      "learning_rate": 3.619565217391305e-05,
+      "loss": 0.1226,
       "step": 1510
     },
     {
       "epoch": 16.52,
-      "learning_rate": 3.521739130434783e-05,
-      "loss": 0.1186,
+      "learning_rate": 3.510869565217392e-05,
+      "loss": 0.1394,
       "step": 1520
     },
     {
       "epoch": 16.63,
-      "learning_rate": 3.413043478260869e-05,
-      "loss": 0.2223,
+      "learning_rate": 3.402173913043478e-05,
+      "loss": 0.2203,
       "step": 1530
     },
     {
       "epoch": 16.74,
-      "learning_rate": 3.304347826086956e-05,
-      "loss": 0.3185,
+      "learning_rate": 3.293478260869565e-05,
+      "loss": 0.3049,
       "step": 1540
     },
     {
       "epoch": 16.85,
-      "learning_rate": 3.195652173913043e-05,
-      "loss": 0.1554,
+      "learning_rate": 3.1847826086956526e-05,
+      "loss": 0.1622,
       "step": 1550
     },
     {
       "epoch": 16.96,
-      "learning_rate": 3.086956521739131e-05,
-      "loss": 0.2177,
+      "learning_rate": 3.0760869565217395e-05,
+      "loss": 0.2375,
       "step": 1560
     },
     {
       "epoch": 17.07,
-      "learning_rate": 2.9782608695652175e-05,
-      "loss": 0.167,
+      "learning_rate": 2.967391304347826e-05,
+      "loss": 0.123,
       "step": 1570
     },
     {
       "epoch": 17.17,
-      "learning_rate": 2.8695652173913044e-05,
-      "loss": 0.216,
+      "learning_rate": 2.8586956521739135e-05,
+      "loss": 0.1922,
       "step": 1580
     },
     {
       "epoch": 17.28,
-      "learning_rate": 2.7608695652173917e-05,
-      "loss": 0.2664,
+      "learning_rate": 2.7500000000000004e-05,
+      "loss": 0.2794,
       "step": 1590
     },
     {
       "epoch": 17.39,
-      "learning_rate": 2.6521739130434787e-05,
-      "loss": 0.2397,
+      "learning_rate": 2.6413043478260867e-05,
+      "loss": 0.2183,
       "step": 1600
     },
     {
       "epoch": 17.39,
-      "eval_accuracy": 0.9807692307692307,
-      "eval_loss": 0.06116783991456032,
-      "eval_runtime": 1.9643,
-      "eval_samples_per_second": 132.362,
-      "eval_steps_per_second": 16.8,
+      "eval_accuracy": 0.9615384615384616,
+      "eval_loss": 0.13765688240528107,
+      "eval_runtime": 2.3734,
+      "eval_samples_per_second": 109.546,
+      "eval_steps_per_second": 13.904,
       "step": 1600
     },
     {
       "epoch": 17.5,
-      "learning_rate": 2.543478260869565e-05,
-      "loss": 0.1432,
+      "learning_rate": 2.5326086956521737e-05,
+      "loss": 0.1748,
       "step": 1610
     },
     {
       "epoch": 17.61,
-      "learning_rate": 2.4347826086956523e-05,
-      "loss": 0.2176,
+      "learning_rate": 2.423913043478261e-05,
+      "loss": 0.2382,
       "step": 1620
     },
     {
       "epoch": 17.72,
-      "learning_rate": 2.3260869565217393e-05,
-      "loss": 0.296,
+      "learning_rate": 2.315217391304348e-05,
+      "loss": 0.2712,
       "step": 1630
     },
     {
       "epoch": 17.83,
-      "learning_rate": 2.2173913043478262e-05,
-      "loss": 0.2308,
+      "learning_rate": 2.206521739130435e-05,
+      "loss": 0.2208,
       "step": 1640
     },
     {
       "epoch": 17.93,
-      "learning_rate": 2.1086956521739132e-05,
-      "loss": 0.1676,
+      "learning_rate": 2.097826086956522e-05,
+      "loss": 0.1639,
       "step": 1650
     },
     {
       "epoch": 18.04,
-      "learning_rate": 2e-05,
-      "loss": 0.1274,
+      "learning_rate": 1.9891304347826085e-05,
+      "loss": 0.1006,
       "step": 1660
     },
     {
       "epoch": 18.15,
-      "learning_rate": 1.8913043478260868e-05,
-      "loss": 0.179,
+      "learning_rate": 1.8804347826086958e-05,
+      "loss": 0.1697,
       "step": 1670
     },
     {
       "epoch": 18.26,
-      "learning_rate": 1.782608695652174e-05,
-      "loss": 0.1569,
+      "learning_rate": 1.7717391304347828e-05,
+      "loss": 0.1967,
       "step": 1680
     },
     {
       "epoch": 18.37,
-      "learning_rate": 1.673913043478261e-05,
+      "learning_rate": 1.6630434782608698e-05,
       "loss": 0.1897,
       "step": 1690
     },
     {
       "epoch": 18.48,
-      "learning_rate": 1.565217391304348e-05,
-      "loss": 0.1348,
+      "learning_rate": 1.5543478260869564e-05,
+      "loss": 0.1478,
       "step": 1700
     },
     {
       "epoch": 18.48,
-      "eval_accuracy": 0.9846153846153847,
-      "eval_loss": 0.05584708973765373,
-      "eval_runtime": 1.9494,
-      "eval_samples_per_second": 133.373,
-      "eval_steps_per_second": 16.928,
+      "eval_accuracy": 0.9769230769230769,
+      "eval_loss": 0.12689848244190216,
+      "eval_runtime": 1.8016,
+      "eval_samples_per_second": 144.314,
+      "eval_steps_per_second": 18.317,
       "step": 1700
     },
     {
       "epoch": 18.59,
-      "learning_rate": 1.4565217391304348e-05,
-      "loss": 0.1662,
+      "learning_rate": 1.4456521739130435e-05,
+      "loss": 0.1515,
       "step": 1710
     },
     {
       "epoch": 18.7,
-      "learning_rate": 1.3478260869565218e-05,
-      "loss": 0.1644,
+      "learning_rate": 1.3369565217391305e-05,
+      "loss": 0.2222,
       "step": 1720
     },
     {
       "epoch": 18.8,
-      "learning_rate": 1.2391304347826088e-05,
-      "loss": 0.2025,
+      "learning_rate": 1.2282608695652175e-05,
+      "loss": 0.201,
       "step": 1730
     },
     {
       "epoch": 18.91,
-      "learning_rate": 1.1304347826086957e-05,
-      "loss": 0.1745,
+      "learning_rate": 1.1195652173913044e-05,
+      "loss": 0.1389,
       "step": 1740
     },
     {
       "epoch": 19.02,
-      "learning_rate": 1.0217391304347827e-05,
-      "loss": 0.1786,
+      "learning_rate": 1.0108695652173914e-05,
+      "loss": 0.1495,
       "step": 1750
     },
     {
       "epoch": 19.13,
-      "learning_rate": 9.130434782608697e-06,
-      "loss": 0.2791,
+      "learning_rate": 9.021739130434784e-06,
+      "loss": 0.2616,
       "step": 1760
     },
     {
       "epoch": 19.24,
-      "learning_rate": 8.043478260869565e-06,
-      "loss": 0.2056,
+      "learning_rate": 7.934782608695653e-06,
+      "loss": 0.1864,
       "step": 1770
     },
     {
       "epoch": 19.35,
-      "learning_rate": 6.956521739130435e-06,
-      "loss": 0.2616,
+      "learning_rate": 6.847826086956521e-06,
+      "loss": 0.1395,
       "step": 1780
     },
     {
       "epoch": 19.46,
-      "learning_rate": 5.869565217391305e-06,
-      "loss": 0.1488,
+      "learning_rate": 5.760869565217392e-06,
+      "loss": 0.1486,
       "step": 1790
     },
     {
       "epoch": 19.57,
-      "learning_rate": 4.782608695652174e-06,
-      "loss": 0.2842,
+      "learning_rate": 4.673913043478261e-06,
+      "loss": 0.1944,
       "step": 1800
     },
     {
       "epoch": 19.57,
       "eval_accuracy": 0.9769230769230769,
-      "eval_loss": 0.05035410821437836,
-      "eval_runtime": 1.9009,
-      "eval_samples_per_second": 136.776,
-      "eval_steps_per_second": 17.36,
+      "eval_loss": 0.09090477973222733,
+      "eval_runtime": 2.3489,
+      "eval_samples_per_second": 110.691,
+      "eval_steps_per_second": 14.049,
       "step": 1800
     },
     {
       "epoch": 19.67,
-      "learning_rate": 3.695652173913044e-06,
-      "loss": 0.1487,
+      "learning_rate": 3.5869565217391305e-06,
+      "loss": 0.1359,
       "step": 1810
     },
     {
       "epoch": 19.78,
-      "learning_rate": 2.608695652173913e-06,
-      "loss": 0.1818,
+      "learning_rate": 2.5e-06,
+      "loss": 0.143,
       "step": 1820
     },
     {
       "epoch": 19.89,
-      "learning_rate": 1.5217391304347827e-06,
-      "loss": 0.1469,
+      "learning_rate": 1.4130434782608697e-06,
+      "loss": 0.1368,
       "step": 1830
     },
     {
       "epoch": 20.0,
-      "learning_rate": 4.347826086956522e-07,
-      "loss": 0.213,
+      "learning_rate": 3.260869565217391e-07,
+      "loss": 0.162,
       "step": 1840
     },
     {
       "epoch": 20.0,
       "step": 1840,
       "total_flos": 7.30261234607063e+17,
-      "train_loss": 0.5342570722103119,
-      "train_runtime": 409.3983,
-      "train_samples_per_second": 71.715,
-      "train_steps_per_second": 4.494
+      "train_loss": 0.5132673807766126,
+      "train_runtime": 395.7262,
+      "train_samples_per_second": 74.193,
+      "train_steps_per_second": 4.65
     }
   ],
   "logging_steps": 10,