Training in progress, epoch 0

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 9.98,
-    "eval_accuracy": 0.9405358686257562,
-    "eval_loss": 0.3180586099624634,
-    "eval_runtime": 38.4742,
-    "eval_samples_per_second": 150.361,
-    "eval_steps_per_second": 4.704,
+    "eval_accuracy": 0.9676750216076059,
+    "eval_loss": 0.17719660699367523,
+    "eval_runtime": 39.5245,
+    "eval_samples_per_second": 146.365,
+    "eval_steps_per_second": 4.579,
     "total_flos": 1.3140619208067262e+19,
-    "train_loss": 1.0548507860728673,
-    "train_runtime": 4291.8856,
-    "train_samples_per_second": 121.301,
-    "train_steps_per_second": 0.946
+    "train_loss": 0.6519044913681857,
+    "train_runtime": 4257.1599,
+    "train_samples_per_second": 122.29,
+    "train_steps_per_second": 0.954
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 9.98,
-    "eval_accuracy": 0.9405358686257562,
-    "eval_loss": 0.3180586099624634,
-    "eval_runtime": 38.4742,
-    "eval_samples_per_second": 150.361,
-    "eval_steps_per_second": 4.704
+    "eval_accuracy": 0.9676750216076059,
+    "eval_loss": 0.17719660699367523,
+    "eval_runtime": 39.5245,
+    "eval_samples_per_second": 146.365,
+    "eval_steps_per_second": 4.579
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:98a51a57fc178dbdb002a5ebb4e723eb886ab624410b0670278008dde8af634d
+oid sha256:83a41aa2fe7fb29377d5a39f74ecb8c77f954f774989e488dc7341910f1d415f
 size 112253100

runs/Dec14_22-06-10_bc148a6f3d4e/events.out.tfevents.1702600361.bc148a6f3d4e.163484.3

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32b4e4b7659b50fe66b8615b452a3a76ff415578879db6cf25f51dbaf082cfe0
+size 36401

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 9.98,
     "total_flos": 1.3140619208067262e+19,
-    "train_loss": 1.0548507860728673,
-    "train_runtime": 4291.8856,
-    "train_samples_per_second": 121.301,
-    "train_steps_per_second": 0.946
+    "train_loss": 0.6519044913681857,
+    "train_runtime": 4257.1599,
+    "train_samples_per_second": 122.29,
+    "train_steps_per_second": 0.954
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.9405358686257562,
-  "best_model_checkpoint": "swin-tiny-patch4-window7-224-finetuned-eurosat/checkpoint-3660",
+  "best_metric": 0.9676750216076059,
+  "best_model_checkpoint": "swin-tiny-patch4-window7-224-finetuned-eurosat/checkpoint-4060",
   "epoch": 9.981561155500922,
   "eval_steps": 500,
   "global_step": 4060,
@@ -11,2537 +11,2537 @@
     {
       "epoch": 0.02,
       "learning_rate": 1.2315270935960593e-06,
-      "loss": 1.8272,
+      "loss": 0.8581,
       "step": 10
     },
     {
       "epoch": 0.05,
       "learning_rate": 2.4630541871921186e-06,
-      "loss": 1.7835,
+      "loss": 0.7471,
       "step": 20
     },
     {
       "epoch": 0.07,
       "learning_rate": 3.6945812807881777e-06,
-      "loss": 1.751,
+      "loss": 0.7525,
       "step": 30
     },
     {
       "epoch": 0.1,
       "learning_rate": 4.926108374384237e-06,
-      "loss": 1.7539,
+      "loss": 0.7179,
       "step": 40
     },
     {
       "epoch": 0.12,
       "learning_rate": 6.157635467980296e-06,
-      "loss": 1.7125,
+      "loss": 0.769,
       "step": 50
     },
     {
       "epoch": 0.15,
       "learning_rate": 7.3891625615763555e-06,
-      "loss": 1.6708,
+      "loss": 0.6952,
       "step": 60
     },
     {
       "epoch": 0.17,
       "learning_rate": 8.620689655172414e-06,
-      "loss": 1.7727,
+      "loss": 0.8408,
       "step": 70
     },
     {
       "epoch": 0.2,
       "learning_rate": 9.852216748768475e-06,
-      "loss": 1.6377,
+      "loss": 0.7396,
       "step": 80
     },
     {
       "epoch": 0.22,
       "learning_rate": 1.1083743842364533e-05,
-      "loss": 1.8024,
+      "loss": 0.8311,
       "step": 90
     },
     {
       "epoch": 0.25,
       "learning_rate": 1.2315270935960592e-05,
-      "loss": 1.7476,
+      "loss": 0.7802,
       "step": 100
     },
     {
       "epoch": 0.27,
       "learning_rate": 1.354679802955665e-05,
-      "loss": 1.6509,
+      "loss": 0.7188,
       "step": 110
     },
     {
       "epoch": 0.3,
       "learning_rate": 1.4778325123152711e-05,
-      "loss": 1.7052,
+      "loss": 0.7815,
       "step": 120
     },
     {
       "epoch": 0.32,
       "learning_rate": 1.6009852216748768e-05,
-      "loss": 1.7003,
+      "loss": 0.7817,
       "step": 130
     },
     {
       "epoch": 0.34,
       "learning_rate": 1.7241379310344828e-05,
-      "loss": 1.7424,
+      "loss": 0.7546,
       "step": 140
     },
     {
       "epoch": 0.37,
       "learning_rate": 1.847290640394089e-05,
-      "loss": 1.7223,
+      "loss": 0.7266,
       "step": 150
     },
     {
       "epoch": 0.39,
       "learning_rate": 1.970443349753695e-05,
-      "loss": 1.6932,
+      "loss": 0.7962,
       "step": 160
     },
     {
       "epoch": 0.42,
       "learning_rate": 2.0935960591133006e-05,
-      "loss": 1.6393,
+      "loss": 0.7822,
       "step": 170
     },
     {
       "epoch": 0.44,
       "learning_rate": 2.2167487684729066e-05,
-      "loss": 1.7023,
+      "loss": 0.8589,
       "step": 180
     },
     {
       "epoch": 0.47,
       "learning_rate": 2.3399014778325123e-05,
-      "loss": 1.6219,
+      "loss": 0.7298,
       "step": 190
     },
     {
       "epoch": 0.49,
       "learning_rate": 2.4630541871921184e-05,
-      "loss": 1.7031,
+      "loss": 0.8412,
       "step": 200
     },
     {
       "epoch": 0.52,
       "learning_rate": 2.5862068965517244e-05,
-      "loss": 1.7309,
+      "loss": 0.8004,
       "step": 210
     },
     {
       "epoch": 0.54,
       "learning_rate": 2.70935960591133e-05,
-      "loss": 1.7251,
+      "loss": 0.7673,
       "step": 220
     },
     {
       "epoch": 0.57,
       "learning_rate": 2.8325123152709358e-05,
-      "loss": 1.7039,
+      "loss": 0.7293,
       "step": 230
     },
     {
       "epoch": 0.59,
       "learning_rate": 2.9556650246305422e-05,
-      "loss": 1.6753,
+      "loss": 0.8035,
       "step": 240
     },
     {
       "epoch": 0.61,
       "learning_rate": 3.078817733990148e-05,
-      "loss": 1.7149,
+      "loss": 0.7597,
       "step": 250
     },
     {
       "epoch": 0.64,
       "learning_rate": 3.2019704433497536e-05,
-      "loss": 1.7101,
+      "loss": 0.8599,
       "step": 260
     },
     {
       "epoch": 0.66,
       "learning_rate": 3.3251231527093596e-05,
-      "loss": 1.6208,
+      "loss": 0.7831,
       "step": 270
     },
     {
       "epoch": 0.69,
       "learning_rate": 3.4482758620689657e-05,
-      "loss": 1.5855,
+      "loss": 0.7714,
       "step": 280
     },
     {
       "epoch": 0.71,
       "learning_rate": 3.571428571428572e-05,
-      "loss": 1.741,
+      "loss": 0.8696,
       "step": 290
     },
     {
       "epoch": 0.74,
       "learning_rate": 3.694581280788178e-05,
-      "loss": 1.6925,
+      "loss": 0.8704,
       "step": 300
     },
     {
       "epoch": 0.76,
       "learning_rate": 3.817733990147783e-05,
-      "loss": 1.6607,
+      "loss": 0.7956,
       "step": 310
     },
     {
       "epoch": 0.79,
       "learning_rate": 3.94088669950739e-05,
-      "loss": 1.5157,
+      "loss": 0.758,
       "step": 320
     },
     {
       "epoch": 0.81,
       "learning_rate": 4.064039408866995e-05,
-      "loss": 1.5652,
+      "loss": 0.7565,
       "step": 330
     },
     {
       "epoch": 0.84,
       "learning_rate": 4.187192118226601e-05,
-      "loss": 1.6547,
+      "loss": 0.8304,
       "step": 340
     },
     {
       "epoch": 0.86,
       "learning_rate": 4.3103448275862066e-05,
-      "loss": 1.5618,
+      "loss": 0.8016,
       "step": 350
     },
     {
       "epoch": 0.89,
       "learning_rate": 4.433497536945813e-05,
-      "loss": 1.6555,
+      "loss": 0.8437,
       "step": 360
     },
     {
       "epoch": 0.91,
       "learning_rate": 4.5566502463054186e-05,
-      "loss": 1.5538,
+      "loss": 0.8352,
       "step": 370
     },
     {
       "epoch": 0.93,
       "learning_rate": 4.679802955665025e-05,
-      "loss": 1.5496,
+      "loss": 0.7455,
       "step": 380
     },
     {
       "epoch": 0.96,
       "learning_rate": 4.802955665024631e-05,
-      "loss": 1.5248,
+      "loss": 0.7252,
       "step": 390
     },
     {
       "epoch": 0.98,
       "learning_rate": 4.926108374384237e-05,
-      "loss": 1.5645,
+      "loss": 0.7629,
       "step": 400
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.7904926534140018,
-      "eval_loss": 1.0879656076431274,
-      "eval_runtime": 27.6968,
-      "eval_samples_per_second": 208.869,
-      "eval_steps_per_second": 6.535,
+      "eval_accuracy": 0.9312013828867761,
+      "eval_loss": 0.3326094150543213,
+      "eval_runtime": 27.2379,
+      "eval_samples_per_second": 212.388,
+      "eval_steps_per_second": 6.645,
       "step": 406
     },
     {
       "epoch": 1.01,
       "learning_rate": 4.9945265462506846e-05,
-      "loss": 1.5933,
+      "loss": 0.836,
       "step": 410
     },
     {
       "epoch": 1.03,
       "learning_rate": 4.980842911877395e-05,
-      "loss": 1.5284,
+      "loss": 0.7864,
       "step": 420
     },
     {
       "epoch": 1.06,
       "learning_rate": 4.9671592775041054e-05,
-      "loss": 1.5239,
+      "loss": 0.8633,
       "step": 430
     },
     {
       "epoch": 1.08,
       "learning_rate": 4.953475643130816e-05,
-      "loss": 1.4781,
+      "loss": 0.8069,
       "step": 440
     },
     {
       "epoch": 1.11,
       "learning_rate": 4.939792008757526e-05,
-      "loss": 1.468,
+      "loss": 0.7711,
       "step": 450
     },
     {
       "epoch": 1.13,
       "learning_rate": 4.926108374384237e-05,
-      "loss": 1.5165,
+      "loss": 0.7662,
       "step": 460
     },
     {
       "epoch": 1.16,
       "learning_rate": 4.912424740010947e-05,
-      "loss": 1.5641,
+      "loss": 0.7585,
       "step": 470
     },
     {
       "epoch": 1.18,
       "learning_rate": 4.8987411056376576e-05,
-      "loss": 1.53,
+      "loss": 0.8457,
       "step": 480
     },
     {
       "epoch": 1.2,
       "learning_rate": 4.885057471264368e-05,
-      "loss": 1.4946,
+      "loss": 0.8136,
       "step": 490
     },
     {
       "epoch": 1.23,
       "learning_rate": 4.8713738368910785e-05,
-      "loss": 1.5378,
+      "loss": 0.8722,
       "step": 500
     },
     {
       "epoch": 1.25,
       "learning_rate": 4.857690202517789e-05,
-      "loss": 1.4228,
+      "loss": 0.7583,
       "step": 510
     },
     {
       "epoch": 1.28,
       "learning_rate": 4.8440065681444994e-05,
-      "loss": 1.4621,
+      "loss": 0.8012,
       "step": 520
     },
     {
       "epoch": 1.3,
       "learning_rate": 4.83032293377121e-05,
-      "loss": 1.5405,
+      "loss": 0.8165,
       "step": 530
     },
     {
       "epoch": 1.33,
       "learning_rate": 4.81663929939792e-05,
-      "loss": 1.4338,
+      "loss": 0.7361,
       "step": 540
     },
     {
       "epoch": 1.35,
       "learning_rate": 4.802955665024631e-05,
-      "loss": 1.3429,
+      "loss": 0.7531,
       "step": 550
     },
     {
       "epoch": 1.38,
       "learning_rate": 4.789272030651341e-05,
-      "loss": 1.4912,
+      "loss": 0.823,
       "step": 560
     },
     {
       "epoch": 1.4,
       "learning_rate": 4.7755883962780516e-05,
-      "loss": 1.4859,
+      "loss": 0.8301,
       "step": 570
     },
     {
       "epoch": 1.43,
       "learning_rate": 4.761904761904762e-05,
-      "loss": 1.3696,
+      "loss": 0.7367,
       "step": 580
     },
     {
       "epoch": 1.45,
       "learning_rate": 4.7482211275314725e-05,
-      "loss": 1.3818,
+      "loss": 0.8218,
       "step": 590
     },
     {
       "epoch": 1.48,
       "learning_rate": 4.734537493158183e-05,
-      "loss": 1.395,
+      "loss": 0.8237,
       "step": 600
     },
     {
       "epoch": 1.5,
       "learning_rate": 4.7208538587848934e-05,
-      "loss": 1.5301,
+      "loss": 0.8238,
       "step": 610
     },
     {
       "epoch": 1.52,
       "learning_rate": 4.707170224411604e-05,
-      "loss": 1.3008,
+      "loss": 0.7757,
       "step": 620
     },
     {
       "epoch": 1.55,
       "learning_rate": 4.693486590038315e-05,
-      "loss": 1.3089,
+      "loss": 0.8156,
       "step": 630
     },
     {
       "epoch": 1.57,
       "learning_rate": 4.679802955665025e-05,
-      "loss": 1.4767,
+      "loss": 0.8355,
       "step": 640
     },
     {
       "epoch": 1.6,
       "learning_rate": 4.666119321291735e-05,
-      "loss": 1.3715,
+      "loss": 0.7258,
       "step": 650
     },
     {
       "epoch": 1.62,
       "learning_rate": 4.652435686918446e-05,
-      "loss": 1.3359,
+      "loss": 0.7613,
       "step": 660
     },
     {
       "epoch": 1.65,
       "learning_rate": 4.638752052545156e-05,
-      "loss": 1.38,
+      "loss": 0.7365,
       "step": 670
     },
     {
       "epoch": 1.67,
       "learning_rate": 4.6250684181718664e-05,
-      "loss": 1.4252,
+      "loss": 0.7489,
       "step": 680
     },
     {
       "epoch": 1.7,
       "learning_rate": 4.611384783798577e-05,
-      "loss": 1.2751,
+      "loss": 0.781,
       "step": 690
     },
     {
       "epoch": 1.72,
       "learning_rate": 4.597701149425287e-05,
-      "loss": 1.4558,
+      "loss": 0.8505,
       "step": 700
     },
     {
       "epoch": 1.75,
       "learning_rate": 4.5840175150519984e-05,
-      "loss": 1.3744,
+      "loss": 0.7674,
       "step": 710
     },
     {
       "epoch": 1.77,
       "learning_rate": 4.570333880678708e-05,
-      "loss": 1.3676,
+      "loss": 0.8244,
       "step": 720
     },
     {
       "epoch": 1.79,
       "learning_rate": 4.5566502463054186e-05,
-      "loss": 1.3031,
+      "loss": 0.8095,
       "step": 730
     },
     {
       "epoch": 1.82,
       "learning_rate": 4.54296661193213e-05,
-      "loss": 1.346,
+      "loss": 0.8051,
       "step": 740
     },
     {
       "epoch": 1.84,
       "learning_rate": 4.5292829775588395e-05,
-      "loss": 1.337,
+      "loss": 0.7743,
       "step": 750
     },
     {
       "epoch": 1.87,
       "learning_rate": 4.5155993431855506e-05,
-      "loss": 1.318,
+      "loss": 0.8334,
       "step": 760
     },
     {
       "epoch": 1.89,
       "learning_rate": 4.501915708812261e-05,
-      "loss": 1.4677,
+      "loss": 0.829,
       "step": 770
     },
     {
       "epoch": 1.92,
       "learning_rate": 4.488232074438971e-05,
-      "loss": 1.2769,
+      "loss": 0.788,
       "step": 780
     },
     {
       "epoch": 1.94,
       "learning_rate": 4.474548440065682e-05,
-      "loss": 1.2428,
+      "loss": 0.692,
       "step": 790
     },
     {
       "epoch": 1.97,
       "learning_rate": 4.460864805692392e-05,
-      "loss": 1.2961,
+      "loss": 0.7329,
       "step": 800
     },
     {
       "epoch": 1.99,
       "learning_rate": 4.447181171319103e-05,
-      "loss": 1.3126,
+      "loss": 0.8118,
       "step": 810
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.8399308556611927,
-      "eval_loss": 0.7987398505210876,
-      "eval_runtime": 27.7317,
-      "eval_samples_per_second": 208.606,
-      "eval_steps_per_second": 6.527,
+      "eval_accuracy": 0.9450302506482282,
+      "eval_loss": 0.29584750533103943,
+      "eval_runtime": 27.0638,
+      "eval_samples_per_second": 213.754,
+      "eval_steps_per_second": 6.688,
       "step": 813
     },
     {
       "epoch": 2.02,
       "learning_rate": 4.433497536945813e-05,
-      "loss": 1.271,
+      "loss": 0.7741,
       "step": 820
     },
     {
       "epoch": 2.04,
       "learning_rate": 4.419813902572523e-05,
-      "loss": 1.3539,
+      "loss": 0.814,
       "step": 830
     },
     {
       "epoch": 2.07,
       "learning_rate": 4.406130268199234e-05,
-      "loss": 1.2938,
+      "loss": 0.8301,
       "step": 840
     },
     {
       "epoch": 2.09,
       "learning_rate": 4.3924466338259446e-05,
-      "loss": 1.2817,
+      "loss": 0.789,
       "step": 850
     },
     {
       "epoch": 2.11,
       "learning_rate": 4.3787629994526544e-05,
-      "loss": 1.2808,
+      "loss": 0.7549,
       "step": 860
     },
     {
       "epoch": 2.14,
       "learning_rate": 4.3650793650793655e-05,
-      "loss": 1.2517,
+      "loss": 0.6721,
       "step": 870
     },
     {
       "epoch": 2.16,
       "learning_rate": 4.351395730706076e-05,
-      "loss": 1.2371,
+      "loss": 0.7396,
       "step": 880
     },
     {
       "epoch": 2.19,
       "learning_rate": 4.3377120963327864e-05,
-      "loss": 1.277,
+      "loss": 0.6782,
       "step": 890
     },
     {
       "epoch": 2.21,
       "learning_rate": 4.324028461959497e-05,
-      "loss": 1.2152,
+      "loss": 0.7289,
       "step": 900
     },
     {
       "epoch": 2.24,
       "learning_rate": 4.3103448275862066e-05,
-      "loss": 1.1733,
+      "loss": 0.712,
       "step": 910
     },
     {
       "epoch": 2.26,
       "learning_rate": 4.296661193212918e-05,
-      "loss": 1.3563,
+      "loss": 0.8466,
       "step": 920
     },
     {
       "epoch": 2.29,
       "learning_rate": 4.282977558839628e-05,
-      "loss": 1.4114,
+      "loss": 0.8502,
       "step": 930
     },
     {
       "epoch": 2.31,
       "learning_rate": 4.2692939244663386e-05,
-      "loss": 1.1794,
+      "loss": 0.6847,
       "step": 940
     },
     {
       "epoch": 2.34,
       "learning_rate": 4.255610290093049e-05,
-      "loss": 1.2454,
+      "loss": 0.8097,
       "step": 950
     },
     {
       "epoch": 2.36,
       "learning_rate": 4.2419266557197594e-05,
-      "loss": 1.2245,
+      "loss": 0.709,
       "step": 960
     },
     {
       "epoch": 2.38,
       "learning_rate": 4.22824302134647e-05,
-      "loss": 1.2197,
+      "loss": 0.665,
       "step": 970
     },
     {
       "epoch": 2.41,
       "learning_rate": 4.21455938697318e-05,
-      "loss": 1.3025,
+      "loss": 0.6726,
       "step": 980
     },
     {
       "epoch": 2.43,
       "learning_rate": 4.200875752599891e-05,
-      "loss": 1.1717,
+      "loss": 0.7391,
       "step": 990
     },
     {
       "epoch": 2.46,
       "learning_rate": 4.187192118226601e-05,
-      "loss": 1.2077,
+      "loss": 0.7706,
       "step": 1000
     },
     {
       "epoch": 2.48,
       "learning_rate": 4.1735084838533116e-05,
-      "loss": 1.3137,
+      "loss": 0.8454,
       "step": 1010
     },
     {
       "epoch": 2.51,
       "learning_rate": 4.159824849480022e-05,
-      "loss": 1.2348,
+      "loss": 0.7661,
       "step": 1020
     },
     {
       "epoch": 2.53,
       "learning_rate": 4.1461412151067325e-05,
-      "loss": 1.2611,
+      "loss": 0.7454,
       "step": 1030
     },
     {
       "epoch": 2.56,
       "learning_rate": 4.132457580733443e-05,
-      "loss": 1.0912,
+      "loss": 0.6551,
       "step": 1040
     },
     {
       "epoch": 2.58,
       "learning_rate": 4.1187739463601534e-05,
-      "loss": 1.3145,
+      "loss": 0.7453,
       "step": 1050
     },
     {
       "epoch": 2.61,
       "learning_rate": 4.105090311986864e-05,
-      "loss": 1.2088,
+      "loss": 0.7144,
       "step": 1060
     },
     {
       "epoch": 2.63,
       "learning_rate": 4.091406677613574e-05,
-      "loss": 1.1283,
+      "loss": 0.734,
       "step": 1070
     },
     {
       "epoch": 2.66,
       "learning_rate": 4.077723043240285e-05,
-      "loss": 1.1388,
+      "loss": 0.7234,
       "step": 1080
     },
     {
       "epoch": 2.68,
       "learning_rate": 4.064039408866995e-05,
-      "loss": 1.2748,
+      "loss": 0.7665,
       "step": 1090
     },
     {
       "epoch": 2.7,
       "learning_rate": 4.050355774493706e-05,
-      "loss": 1.0486,
+      "loss": 0.6977,
       "step": 1100
     },
     {
       "epoch": 2.73,
       "learning_rate": 4.036672140120416e-05,
-      "loss": 1.1514,
+      "loss": 0.7377,
       "step": 1110
     },
     {
       "epoch": 2.75,
       "learning_rate": 4.0229885057471265e-05,
-      "loss": 1.2419,
+      "loss": 0.7405,
       "step": 1120
     },
     {
       "epoch": 2.78,
       "learning_rate": 4.009304871373837e-05,
-      "loss": 1.3025,
+      "loss": 0.7858,
       "step": 1130
     },
     {
       "epoch": 2.8,
       "learning_rate": 3.9956212370005474e-05,
-      "loss": 1.1627,
+      "loss": 0.6955,
       "step": 1140
     },
     {
       "epoch": 2.83,
       "learning_rate": 3.981937602627258e-05,
-      "loss": 1.2263,
+      "loss": 0.8183,
       "step": 1150
     },
     {
       "epoch": 2.85,
       "learning_rate": 3.968253968253968e-05,
-      "loss": 1.1918,
+      "loss": 0.7564,
       "step": 1160
     },
     {
       "epoch": 2.88,
       "learning_rate": 3.954570333880679e-05,
-      "loss": 1.2337,
+      "loss": 0.7048,
       "step": 1170
     },
     {
       "epoch": 2.9,
       "learning_rate": 3.94088669950739e-05,
-      "loss": 1.325,
+      "loss": 0.7414,
       "step": 1180
     },
     {
       "epoch": 2.93,
       "learning_rate": 3.9272030651340996e-05,
-      "loss": 1.1631,
+      "loss": 0.7146,
       "step": 1190
     },
     {
       "epoch": 2.95,
       "learning_rate": 3.91351943076081e-05,
-      "loss": 1.2506,
+      "loss": 0.7579,
       "step": 1200
     },
     {
       "epoch": 2.97,
       "learning_rate": 3.899835796387521e-05,
-      "loss": 1.2831,
+      "loss": 0.7648,
       "step": 1210
     },
     {
       "epoch": 3.0,
       "learning_rate": 3.886152162014231e-05,
-      "loss": 1.2335,
+      "loss": 0.7189,
       "step": 1220
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.8741573033707866,
-      "eval_loss": 0.6398221254348755,
-      "eval_runtime": 27.2608,
-      "eval_samples_per_second": 212.209,
-      "eval_steps_per_second": 6.64,
+      "eval_accuracy": 0.9515989628349178,
+      "eval_loss": 0.25021520256996155,
+      "eval_runtime": 28.0749,
+      "eval_samples_per_second": 206.056,
+      "eval_steps_per_second": 6.447,
       "step": 1220
     },
     {
       "epoch": 3.02,
       "learning_rate": 3.872468527640942e-05,
-      "loss": 1.1144,
+      "loss": 0.748,
       "step": 1230
     },
     {
       "epoch": 3.05,
       "learning_rate": 3.858784893267652e-05,
-      "loss": 1.1746,
+      "loss": 0.7554,
       "step": 1240
     },
     {
       "epoch": 3.07,
       "learning_rate": 3.845101258894362e-05,
-      "loss": 1.1363,
+      "loss": 0.7256,
       "step": 1250
     },
     {
       "epoch": 3.1,
       "learning_rate": 3.831417624521073e-05,
-      "loss": 1.1003,
+      "loss": 0.6269,
       "step": 1260
     },
     {
       "epoch": 3.12,
       "learning_rate": 3.817733990147783e-05,
-      "loss": 1.1743,
+      "loss": 0.6848,
       "step": 1270
     },
     {
       "epoch": 3.15,
       "learning_rate": 3.8040503557744935e-05,
-      "loss": 1.1832,
+      "loss": 0.7929,
       "step": 1280
     },
     {
       "epoch": 3.17,
       "learning_rate": 3.7903667214012047e-05,
-      "loss": 1.1495,
+      "loss": 0.7091,
       "step": 1290
     },
     {
       "epoch": 3.2,
       "learning_rate": 3.7766830870279144e-05,
-      "loss": 1.1282,
+      "loss": 0.7314,
       "step": 1300
     },
     {
       "epoch": 3.22,
       "learning_rate": 3.7629994526546255e-05,
-      "loss": 1.1273,
+      "loss": 0.6781,
       "step": 1310
     },
     {
       "epoch": 3.25,
       "learning_rate": 3.749315818281336e-05,
-      "loss": 1.0586,
+      "loss": 0.6571,
       "step": 1320
     },
     {
       "epoch": 3.27,
       "learning_rate": 3.735632183908046e-05,
-      "loss": 1.1353,
+      "loss": 0.6719,
       "step": 1330
     },
     {
       "epoch": 3.29,
       "learning_rate": 3.721948549534757e-05,
-      "loss": 1.0796,
+      "loss": 0.7128,
       "step": 1340
     },
     {
       "epoch": 3.32,
       "learning_rate": 3.7082649151614666e-05,
-      "loss": 1.1132,
+      "loss": 0.6949,
       "step": 1350
     },
     {
       "epoch": 3.34,
       "learning_rate": 3.694581280788178e-05,
-      "loss": 1.0474,
+      "loss": 0.6583,
       "step": 1360
     },
     {
       "epoch": 3.37,
       "learning_rate": 3.680897646414888e-05,
-      "loss": 1.1392,
+      "loss": 0.7373,
       "step": 1370
     },
     {
       "epoch": 3.39,
       "learning_rate": 3.667214012041598e-05,
-      "loss": 1.1334,
+      "loss": 0.707,
       "step": 1380
     },
     {
       "epoch": 3.42,
       "learning_rate": 3.653530377668309e-05,
-      "loss": 1.0506,
+      "loss": 0.715,
       "step": 1390
     },
     {
       "epoch": 3.44,
       "learning_rate": 3.6398467432950195e-05,
-      "loss": 1.1107,
+      "loss": 0.6595,
       "step": 1400
     },
     {
       "epoch": 3.47,
       "learning_rate": 3.62616310892173e-05,
-      "loss": 1.1643,
+      "loss": 0.6768,
       "step": 1410
     },
     {
       "epoch": 3.49,
       "learning_rate": 3.6124794745484404e-05,
-      "loss": 1.0167,
+      "loss": 0.6416,
       "step": 1420
     },
     {
       "epoch": 3.52,
       "learning_rate": 3.598795840175151e-05,
-      "loss": 1.2377,
+      "loss": 0.8161,
       "step": 1430
     },
     {
       "epoch": 3.54,
       "learning_rate": 3.585112205801861e-05,
-      "loss": 1.0392,
+      "loss": 0.7023,
       "step": 1440
     },
     {
       "epoch": 3.56,
       "learning_rate": 3.571428571428572e-05,
-      "loss": 1.1304,
+      "loss": 0.6853,
       "step": 1450
     },
     {
       "epoch": 3.59,
       "learning_rate": 3.5577449370552815e-05,
-      "loss": 1.2185,
+      "loss": 0.7589,
       "step": 1460
     },
     {
       "epoch": 3.61,
       "learning_rate": 3.5440613026819926e-05,
-      "loss": 1.0848,
+      "loss": 0.6967,
       "step": 1470
     },
     {
       "epoch": 3.64,
       "learning_rate": 3.530377668308703e-05,
-      "loss": 1.1142,
+      "loss": 0.7275,
       "step": 1480
     },
     {
       "epoch": 3.66,
       "learning_rate": 3.5166940339354135e-05,
-      "loss": 1.0458,
+      "loss": 0.7026,
       "step": 1490
     },
     {
       "epoch": 3.69,
       "learning_rate": 3.503010399562124e-05,
-      "loss": 1.041,
+      "loss": 0.6894,
       "step": 1500
     },
     {
       "epoch": 3.71,
       "learning_rate": 3.489326765188834e-05,
-      "loss": 1.119,
+      "loss": 0.6976,
       "step": 1510
     },
     {
       "epoch": 3.74,
       "learning_rate": 3.475643130815545e-05,
-      "loss": 1.0259,
+      "loss": 0.6837,
       "step": 1520
     },
     {
       "epoch": 3.76,
       "learning_rate": 3.461959496442255e-05,
-      "loss": 1.14,
+      "loss": 0.6921,
       "step": 1530
     },
     {
       "epoch": 3.79,
       "learning_rate": 3.4482758620689657e-05,
-      "loss": 0.994,
+      "loss": 0.5885,
       "step": 1540
     },
     {
       "epoch": 3.81,
       "learning_rate": 3.434592227695676e-05,
-      "loss": 1.011,
+      "loss": 0.7013,
       "step": 1550
     },
     {
       "epoch": 3.84,
       "learning_rate": 3.4209085933223865e-05,
-      "loss": 1.0745,
+      "loss": 0.6954,
       "step": 1560
     },
     {
       "epoch": 3.86,
       "learning_rate": 3.407224958949097e-05,
-      "loss": 1.0777,
+      "loss": 0.7609,
       "step": 1570
     },
     {
       "epoch": 3.88,
       "learning_rate": 3.3935413245758074e-05,
-      "loss": 1.1397,
+      "loss": 0.7052,
       "step": 1580
     },
     {
       "epoch": 3.91,
       "learning_rate": 3.379857690202518e-05,
-      "loss": 1.0727,
+      "loss": 0.6929,
       "step": 1590
     },
     {
       "epoch": 3.93,
       "learning_rate": 3.366174055829228e-05,
-      "loss": 1.0662,
+      "loss": 0.6814,
       "step": 1600
     },
     {
       "epoch": 3.96,
       "learning_rate": 3.352490421455939e-05,
-      "loss": 1.0208,
+      "loss": 0.7084,
       "step": 1610
     },
     {
       "epoch": 3.98,
       "learning_rate": 3.338806787082649e-05,
-      "loss": 1.1244,
+      "loss": 0.7529,
       "step": 1620
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.9040622299049266,
-      "eval_loss": 0.497685045003891,
-      "eval_runtime": 27.8668,
-      "eval_samples_per_second": 207.595,
-      "eval_steps_per_second": 6.495,
+      "eval_accuracy": 0.9566119273984443,
+      "eval_loss": 0.2300640344619751,
+      "eval_runtime": 27.261,
+      "eval_samples_per_second": 212.208,
+      "eval_steps_per_second": 6.64,
       "step": 1627
     },
     {
       "epoch": 4.01,
       "learning_rate": 3.3251231527093596e-05,
-      "loss": 0.9674,
+      "loss": 0.6286,
       "step": 1630
     },
     {
       "epoch": 4.03,
       "learning_rate": 3.31143951833607e-05,
-      "loss": 1.0643,
+      "loss": 0.7495,
       "step": 1640
     },
     {
       "epoch": 4.06,
       "learning_rate": 3.297755883962781e-05,
-      "loss": 1.0174,
+      "loss": 0.7085,
       "step": 1650
     },
     {
       "epoch": 4.08,
       "learning_rate": 3.284072249589491e-05,
-      "loss": 1.0732,
+      "loss": 0.72,
       "step": 1660
     },
     {
       "epoch": 4.11,
       "learning_rate": 3.2703886152162014e-05,
-      "loss": 1.0587,
+      "loss": 0.6716,
       "step": 1670
     },
     {
       "epoch": 4.13,
       "learning_rate": 3.256704980842912e-05,
-      "loss": 1.0531,
+      "loss": 0.641,
       "step": 1680
     },
     {
       "epoch": 4.15,
       "learning_rate": 3.243021346469622e-05,
-      "loss": 1.0491,
+      "loss": 0.6338,
       "step": 1690
     },
     {
       "epoch": 4.18,
       "learning_rate": 3.2293377120963334e-05,
-      "loss": 0.9433,
+      "loss": 0.5716,
       "step": 1700
     },
     {
       "epoch": 4.2,
       "learning_rate": 3.215654077723043e-05,
-      "loss": 0.9851,
+      "loss": 0.6536,
       "step": 1710
     },
     {
       "epoch": 4.23,
       "learning_rate": 3.2019704433497536e-05,
-      "loss": 1.0559,
+      "loss": 0.6454,
       "step": 1720
     },
     {
       "epoch": 4.25,
       "learning_rate": 3.188286808976465e-05,
-      "loss": 0.9604,
+      "loss": 0.6392,
       "step": 1730
     },
     {
       "epoch": 4.28,
       "learning_rate": 3.1746031746031745e-05,
-      "loss": 0.9938,
+      "loss": 0.6275,
       "step": 1740
     },
     {
       "epoch": 4.3,
       "learning_rate": 3.160919540229885e-05,
-      "loss": 1.0587,
+      "loss": 0.6634,
       "step": 1750
     },
     {
       "epoch": 4.33,
       "learning_rate": 3.147235905856596e-05,
-      "loss": 0.9771,
+      "loss": 0.6603,
       "step": 1760
     },
     {
       "epoch": 4.35,
       "learning_rate": 3.133552271483306e-05,
-      "loss": 1.0232,
+      "loss": 0.704,
       "step": 1770
     },
     {
       "epoch": 4.38,
       "learning_rate": 3.119868637110017e-05,
-      "loss": 1.014,
+      "loss": 0.7157,
       "step": 1780
     },
     {
       "epoch": 4.4,
       "learning_rate": 3.1061850027367273e-05,
-      "loss": 0.9797,
+      "loss": 0.6142,
       "step": 1790
     },
     {
       "epoch": 4.43,
       "learning_rate": 3.092501368363437e-05,
-      "loss": 0.9898,
+      "loss": 0.5772,
       "step": 1800
     },
     {
       "epoch": 4.45,
       "learning_rate": 3.078817733990148e-05,
-      "loss": 1.0044,
+      "loss": 0.6924,
       "step": 1810
     },
     {
       "epoch": 4.47,
       "learning_rate": 3.065134099616858e-05,
-      "loss": 1.0723,
+      "loss": 0.6931,
       "step": 1820
     },
     {
       "epoch": 4.5,
       "learning_rate": 3.0514504652435688e-05,
-      "loss": 1.0545,
+      "loss": 0.6155,
       "step": 1830
     },
     {
       "epoch": 4.52,
       "learning_rate": 3.0377668308702795e-05,
-      "loss": 0.9979,
+      "loss": 0.6732,
       "step": 1840
     },
     {
       "epoch": 4.55,
       "learning_rate": 3.0240831964969896e-05,
-      "loss": 0.9971,
+      "loss": 0.6537,
       "step": 1850
     },
     {
       "epoch": 4.57,
       "learning_rate": 3.0103995621237e-05,
-      "loss": 1.0341,
+      "loss": 0.6704,
       "step": 1860
     },
     {
       "epoch": 4.6,
       "learning_rate": 2.996715927750411e-05,
-      "loss": 0.9853,
+      "loss": 0.6187,
       "step": 1870
     },
     {
       "epoch": 4.62,
       "learning_rate": 2.983032293377121e-05,
-      "loss": 1.0434,
+      "loss": 0.6287,
       "step": 1880
     },
     {
       "epoch": 4.65,
       "learning_rate": 2.9693486590038317e-05,
-      "loss": 0.94,
+      "loss": 0.5647,
       "step": 1890
     },
     {
       "epoch": 4.67,
       "learning_rate": 2.9556650246305422e-05,
-      "loss": 1.0002,
+      "loss": 0.685,
       "step": 1900
     },
     {
       "epoch": 4.7,
       "learning_rate": 2.9419813902572523e-05,
-      "loss": 0.9252,
+      "loss": 0.5978,
       "step": 1910
     },
     {
       "epoch": 4.72,
       "learning_rate": 2.928297755883963e-05,
-      "loss": 0.8899,
+      "loss": 0.608,
       "step": 1920
     },
     {
       "epoch": 4.74,
       "learning_rate": 2.914614121510673e-05,
-      "loss": 1.0344,
+      "loss": 0.6599,
       "step": 1930
     },
     {
       "epoch": 4.77,
       "learning_rate": 2.900930487137384e-05,
-      "loss": 1.008,
+      "loss": 0.6674,
       "step": 1940
     },
     {
       "epoch": 4.79,
       "learning_rate": 2.8872468527640944e-05,
-      "loss": 0.9935,
+      "loss": 0.6663,
       "step": 1950
     },
     {
       "epoch": 4.82,
       "learning_rate": 2.8735632183908045e-05,
-      "loss": 1.0432,
+      "loss": 0.6524,
       "step": 1960
     },
     {
       "epoch": 4.84,
       "learning_rate": 2.8598795840175153e-05,
-      "loss": 0.9235,
+      "loss": 0.6612,
       "step": 1970
     },
     {
       "epoch": 4.87,
       "learning_rate": 2.8461959496442257e-05,
-      "loss": 0.9564,
+      "loss": 0.7027,
       "step": 1980
     },
     {
       "epoch": 4.89,
       "learning_rate": 2.8325123152709358e-05,
-      "loss": 0.9873,
+      "loss": 0.6129,
       "step": 1990
     },
     {
       "epoch": 4.92,
       "learning_rate": 2.8188286808976466e-05,
-      "loss": 1.0216,
+      "loss": 0.7099,
       "step": 2000
     },
     {
       "epoch": 4.94,
       "learning_rate": 2.8051450465243574e-05,
-      "loss": 0.9,
+      "loss": 0.5781,
       "step": 2010
     },
     {
       "epoch": 4.97,
       "learning_rate": 2.7914614121510675e-05,
-      "loss": 0.9814,
+      "loss": 0.6978,
       "step": 2020
     },
     {
       "epoch": 4.99,
       "learning_rate": 2.777777777777778e-05,
-      "loss": 0.9785,
+      "loss": 0.6746,
       "step": 2030
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.9139152981849611,
-      "eval_loss": 0.4472927749156952,
-      "eval_runtime": 27.773,
-      "eval_samples_per_second": 208.296,
-      "eval_steps_per_second": 6.517,
+      "eval_accuracy": 0.957476231633535,
+      "eval_loss": 0.21456871926784515,
+      "eval_runtime": 27.8285,
+      "eval_samples_per_second": 207.881,
+      "eval_steps_per_second": 6.504,
       "step": 2033
     },
     {
       "epoch": 5.02,
       "learning_rate": 2.764094143404488e-05,
-      "loss": 0.97,
+      "loss": 0.6939,
       "step": 2040
     },
     {
       "epoch": 5.04,
       "learning_rate": 2.7504105090311988e-05,
-      "loss": 0.9339,
+      "loss": 0.5992,
       "step": 2050
     },
     {
       "epoch": 5.06,
       "learning_rate": 2.7367268746579096e-05,
-      "loss": 0.8636,
+      "loss": 0.6252,
       "step": 2060
     },
     {
       "epoch": 5.09,
       "learning_rate": 2.7230432402846197e-05,
-      "loss": 0.8525,
+      "loss": 0.6173,
       "step": 2070
     },
     {
       "epoch": 5.11,
       "learning_rate": 2.70935960591133e-05,
-      "loss": 0.9281,
+      "loss": 0.6292,
       "step": 2080
     },
     {
       "epoch": 5.14,
       "learning_rate": 2.695675971538041e-05,
-      "loss": 0.9573,
+      "loss": 0.6444,
       "step": 2090
     },
     {
       "epoch": 5.16,
       "learning_rate": 2.681992337164751e-05,
-      "loss": 1.0129,
+      "loss": 0.6745,
       "step": 2100
     },
     {
       "epoch": 5.19,
       "learning_rate": 2.6683087027914618e-05,
-      "loss": 0.9568,
+      "loss": 0.6157,
       "step": 2110
     },
     {
       "epoch": 5.21,
       "learning_rate": 2.6546250684181722e-05,
-      "loss": 0.8739,
+      "loss": 0.5988,
       "step": 2120
     },
     {
       "epoch": 5.24,
       "learning_rate": 2.6409414340448823e-05,
-      "loss": 1.0142,
+      "loss": 0.6707,
       "step": 2130
     },
     {
       "epoch": 5.26,
       "learning_rate": 2.627257799671593e-05,
-      "loss": 0.8615,
+      "loss": 0.5951,
       "step": 2140
     },
     {
       "epoch": 5.29,
       "learning_rate": 2.6135741652983032e-05,
-      "loss": 0.9213,
+      "loss": 0.5863,
       "step": 2150
     },
     {
       "epoch": 5.31,
       "learning_rate": 2.5998905309250136e-05,
-      "loss": 0.9751,
+      "loss": 0.6182,
       "step": 2160
     },
     {
       "epoch": 5.33,
       "learning_rate": 2.5862068965517244e-05,
-      "loss": 0.9144,
+      "loss": 0.6163,
       "step": 2170
     },
     {
       "epoch": 5.36,
       "learning_rate": 2.5725232621784345e-05,
-      "loss": 0.8717,
+      "loss": 0.6525,
       "step": 2180
     },
     {
       "epoch": 5.38,
       "learning_rate": 2.5588396278051453e-05,
-      "loss": 1.0529,
+      "loss": 0.6545,
       "step": 2190
     },
     {
       "epoch": 5.41,
       "learning_rate": 2.5451559934318557e-05,
-      "loss": 0.9158,
+      "loss": 0.6883,
       "step": 2200
     },
     {
       "epoch": 5.43,
       "learning_rate": 2.531472359058566e-05,
-      "loss": 0.9584,
+      "loss": 0.6644,
       "step": 2210
     },
     {
       "epoch": 5.46,
       "learning_rate": 2.5177887246852766e-05,
-      "loss": 0.9806,
+      "loss": 0.641,
       "step": 2220
     },
     {
       "epoch": 5.48,
       "learning_rate": 2.5041050903119874e-05,
-      "loss": 0.8941,
+      "loss": 0.6029,
       "step": 2230
     },
     {
       "epoch": 5.51,
       "learning_rate": 2.4904214559386975e-05,
-      "loss": 0.9625,
+      "loss": 0.6049,
       "step": 2240
     },
     {
       "epoch": 5.53,
       "learning_rate": 2.476737821565408e-05,
-      "loss": 0.9277,
+      "loss": 0.6282,
       "step": 2250
     },
     {
       "epoch": 5.56,
       "learning_rate": 2.4630541871921184e-05,
-      "loss": 0.9645,
+      "loss": 0.5836,
       "step": 2260
     },
     {
       "epoch": 5.58,
       "learning_rate": 2.4493705528188288e-05,
-      "loss": 0.9162,
+      "loss": 0.6288,
       "step": 2270
     },
     {
       "epoch": 5.61,
       "learning_rate": 2.4356869184455393e-05,
-      "loss": 0.8772,
+      "loss": 0.5648,
       "step": 2280
     },
     {
       "epoch": 5.63,
       "learning_rate": 2.4220032840722497e-05,
-      "loss": 0.9556,
+      "loss": 0.5689,
       "step": 2290
     },
     {
       "epoch": 5.65,
       "learning_rate": 2.40831964969896e-05,
-      "loss": 0.885,
+      "loss": 0.6038,
       "step": 2300
     },
     {
       "epoch": 5.68,
       "learning_rate": 2.3946360153256706e-05,
-      "loss": 0.7871,
+      "loss": 0.5866,
       "step": 2310
     },
     {
       "epoch": 5.7,
       "learning_rate": 2.380952380952381e-05,
-      "loss": 0.8639,
+      "loss": 0.6283,
       "step": 2320
     },
     {
       "epoch": 5.73,
       "learning_rate": 2.3672687465790915e-05,
-      "loss": 0.9446,
+      "loss": 0.6445,
       "step": 2330
     },
     {
       "epoch": 5.75,
       "learning_rate": 2.353585112205802e-05,
-      "loss": 0.8905,
+      "loss": 0.6545,
       "step": 2340
     },
     {
       "epoch": 5.78,
       "learning_rate": 2.3399014778325123e-05,
-      "loss": 0.9502,
+      "loss": 0.6394,
       "step": 2350
     },
     {
       "epoch": 5.8,
       "learning_rate": 2.326217843459223e-05,
-      "loss": 0.9842,
+      "loss": 0.5928,
       "step": 2360
     },
     {
       "epoch": 5.83,
       "learning_rate": 2.3125342090859332e-05,
-      "loss": 0.8827,
+      "loss": 0.6121,
       "step": 2370
     },
     {
       "epoch": 5.85,
       "learning_rate": 2.2988505747126437e-05,
-      "loss": 0.8861,
+      "loss": 0.6055,
       "step": 2380
     },
     {
       "epoch": 5.88,
       "learning_rate": 2.285166940339354e-05,
-      "loss": 0.9701,
+      "loss": 0.5902,
       "step": 2390
     },
     {
       "epoch": 5.9,
       "learning_rate": 2.271483305966065e-05,
-      "loss": 0.8456,
+      "loss": 0.5764,
       "step": 2400
     },
     {
       "epoch": 5.93,
       "learning_rate": 2.2577996715927753e-05,
-      "loss": 0.8858,
+      "loss": 0.5722,
       "step": 2410
     },
     {
       "epoch": 5.95,
       "learning_rate": 2.2441160372194854e-05,
-      "loss": 0.8431,
+      "loss": 0.5832,
       "step": 2420
     },
     {
       "epoch": 5.97,
       "learning_rate": 2.230432402846196e-05,
-      "loss": 0.9466,
+      "loss": 0.5632,
       "step": 2430
     },
     {
       "epoch": 6.0,
       "learning_rate": 2.2167487684729066e-05,
-      "loss": 0.9625,
+      "loss": 0.546,
       "step": 2440
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.9199654278305963,
-      "eval_loss": 0.3929709494113922,
-      "eval_runtime": 27.3929,
-      "eval_samples_per_second": 211.186,
-      "eval_steps_per_second": 6.608,
+      "eval_accuracy": 0.9609334485738981,
+      "eval_loss": 0.2027408331632614,
+      "eval_runtime": 27.561,
+      "eval_samples_per_second": 209.898,
+      "eval_steps_per_second": 6.567,
       "step": 2440
     },
     {
       "epoch": 6.02,
       "learning_rate": 2.203065134099617e-05,
-      "loss": 0.886,
+      "loss": 0.5745,
       "step": 2450
     },
     {
       "epoch": 6.05,
       "learning_rate": 2.1893814997263272e-05,
-      "loss": 0.8862,
+      "loss": 0.5959,
       "step": 2460
     },
     {
       "epoch": 6.07,
       "learning_rate": 2.175697865353038e-05,
-      "loss": 0.8938,
+      "loss": 0.6355,
       "step": 2470
     },
     {
       "epoch": 6.1,
       "learning_rate": 2.1620142309797484e-05,
-      "loss": 0.8165,
+      "loss": 0.5944,
       "step": 2480
     },
     {
       "epoch": 6.12,
       "learning_rate": 2.148330596606459e-05,
-      "loss": 0.8117,
+      "loss": 0.5942,
       "step": 2490
     },
     {
       "epoch": 6.15,
       "learning_rate": 2.1346469622331693e-05,
-      "loss": 0.8538,
+      "loss": 0.5985,
       "step": 2500
     },
     {
       "epoch": 6.17,
       "learning_rate": 2.1209633278598797e-05,
-      "loss": 0.8646,
+      "loss": 0.5785,
       "step": 2510
     },
     {
       "epoch": 6.2,
       "learning_rate": 2.10727969348659e-05,
-      "loss": 0.8506,
+      "loss": 0.5717,
       "step": 2520
     },
     {
       "epoch": 6.22,
       "learning_rate": 2.0935960591133006e-05,
-      "loss": 0.9332,
+      "loss": 0.6025,
       "step": 2530
     },
     {
       "epoch": 6.24,
       "learning_rate": 2.079912424740011e-05,
-      "loss": 0.7804,
+      "loss": 0.5668,
       "step": 2540
     },
     {
       "epoch": 6.27,
       "learning_rate": 2.0662287903667215e-05,
-      "loss": 0.8503,
+      "loss": 0.5703,
       "step": 2550
     },
     {
       "epoch": 6.29,
       "learning_rate": 2.052545155993432e-05,
-      "loss": 0.8818,
+      "loss": 0.6261,
       "step": 2560
     },
     {
       "epoch": 6.32,
       "learning_rate": 2.0388615216201424e-05,
-      "loss": 0.7721,
+      "loss": 0.4868,
       "step": 2570
     },
     {
       "epoch": 6.34,
       "learning_rate": 2.025177887246853e-05,
-      "loss": 0.8895,
+      "loss": 0.6238,
       "step": 2580
     },
     {
       "epoch": 6.37,
       "learning_rate": 2.0114942528735632e-05,
-      "loss": 0.8575,
+      "loss": 0.5533,
       "step": 2590
     },
     {
       "epoch": 6.39,
       "learning_rate": 1.9978106185002737e-05,
-      "loss": 0.8596,
+      "loss": 0.5589,
       "step": 2600
     },
     {
       "epoch": 6.42,
       "learning_rate": 1.984126984126984e-05,
-      "loss": 0.8394,
+      "loss": 0.6414,
       "step": 2610
     },
     {
       "epoch": 6.44,
       "learning_rate": 1.970443349753695e-05,
-      "loss": 0.8782,
+      "loss": 0.5837,
       "step": 2620
     },
     {
       "epoch": 6.47,
       "learning_rate": 1.956759715380405e-05,
-      "loss": 0.7369,
+      "loss": 0.5744,
       "step": 2630
     },
     {
       "epoch": 6.49,
       "learning_rate": 1.9430760810071154e-05,
-      "loss": 0.8684,
+      "loss": 0.5979,
       "step": 2640
     },
     {
       "epoch": 6.52,
       "learning_rate": 1.929392446633826e-05,
-      "loss": 0.8494,
+      "loss": 0.5642,
       "step": 2650
     },
     {
       "epoch": 6.54,
       "learning_rate": 1.9157088122605367e-05,
-      "loss": 0.764,
+      "loss": 0.508,
       "step": 2660
     },
     {
       "epoch": 6.56,
       "learning_rate": 1.9020251778872468e-05,
-      "loss": 0.855,
+      "loss": 0.6201,
       "step": 2670
     },
     {
       "epoch": 6.59,
       "learning_rate": 1.8883415435139572e-05,
-      "loss": 0.8766,
+      "loss": 0.6078,
       "step": 2680
     },
     {
       "epoch": 6.61,
       "learning_rate": 1.874657909140668e-05,
-      "loss": 0.8737,
+      "loss": 0.6139,
       "step": 2690
     },
     {
       "epoch": 6.64,
       "learning_rate": 1.8609742747673784e-05,
-      "loss": 0.8441,
+      "loss": 0.5885,
       "step": 2700
     },
     {
       "epoch": 6.66,
       "learning_rate": 1.847290640394089e-05,
-      "loss": 0.8257,
+      "loss": 0.5809,
       "step": 2710
     },
     {
       "epoch": 6.69,
       "learning_rate": 1.833607006020799e-05,
-      "loss": 0.8367,
+      "loss": 0.5503,
       "step": 2720
     },
     {
       "epoch": 6.71,
       "learning_rate": 1.8199233716475097e-05,
-      "loss": 0.8218,
+      "loss": 0.5675,
       "step": 2730
     },
     {
       "epoch": 6.74,
       "learning_rate": 1.8062397372742202e-05,
-      "loss": 0.7645,
+      "loss": 0.554,
       "step": 2740
     },
     {
       "epoch": 6.76,
       "learning_rate": 1.7925561029009306e-05,
-      "loss": 0.8305,
+      "loss": 0.6173,
       "step": 2750
     },
     {
       "epoch": 6.79,
       "learning_rate": 1.7788724685276407e-05,
-      "loss": 0.8306,
+      "loss": 0.5588,
       "step": 2760
     },
     {
       "epoch": 6.81,
       "learning_rate": 1.7651888341543515e-05,
-      "loss": 0.9052,
+      "loss": 0.5736,
       "step": 2770
     },
     {
       "epoch": 6.83,
       "learning_rate": 1.751505199781062e-05,
-      "loss": 0.7477,
+      "loss": 0.5748,
       "step": 2780
     },
     {
       "epoch": 6.86,
       "learning_rate": 1.7378215654077724e-05,
-      "loss": 0.8458,
+      "loss": 0.6228,
       "step": 2790
     },
     {
       "epoch": 6.88,
       "learning_rate": 1.7241379310344828e-05,
-      "loss": 0.901,
+      "loss": 0.635,
       "step": 2800
     },
     {
       "epoch": 6.91,
       "learning_rate": 1.7104542966611933e-05,
-      "loss": 0.8307,
+      "loss": 0.5662,
       "step": 2810
     },
     {
       "epoch": 6.93,
       "learning_rate": 1.6967706622879037e-05,
-      "loss": 0.8201,
+      "loss": 0.6354,
       "step": 2820
     },
     {
       "epoch": 6.96,
       "learning_rate": 1.683087027914614e-05,
-      "loss": 0.8514,
+      "loss": 0.5635,
       "step": 2830
     },
     {
       "epoch": 6.98,
       "learning_rate": 1.6694033935413246e-05,
-      "loss": 0.8414,
+      "loss": 0.5983,
       "step": 2840
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.9327571305099395,
-      "eval_loss": 0.3537824749946594,
-      "eval_runtime": 27.9648,
-      "eval_samples_per_second": 206.867,
-      "eval_steps_per_second": 6.472,
+      "eval_accuracy": 0.9640449438202248,
+      "eval_loss": 0.19191542267799377,
+      "eval_runtime": 27.0395,
+      "eval_samples_per_second": 213.946,
+      "eval_steps_per_second": 6.694,
       "step": 2847
     },
     {
       "epoch": 7.01,
       "learning_rate": 1.655719759168035e-05,
-      "loss": 0.9236,
+      "loss": 0.5661,
       "step": 2850
     },
     {
       "epoch": 7.03,
       "learning_rate": 1.6420361247947455e-05,
-      "loss": 0.8118,
+      "loss": 0.5409,
       "step": 2860
     },
     {
       "epoch": 7.06,
       "learning_rate": 1.628352490421456e-05,
-      "loss": 0.8715,
+      "loss": 0.5863,
       "step": 2870
     },
     {
       "epoch": 7.08,
       "learning_rate": 1.6146688560481667e-05,
-      "loss": 0.8181,
+      "loss": 0.5587,
       "step": 2880
     },
     {
       "epoch": 7.11,
       "learning_rate": 1.6009852216748768e-05,
-      "loss": 0.8577,
+      "loss": 0.5921,
       "step": 2890
     },
     {
       "epoch": 7.13,
       "learning_rate": 1.5873015873015872e-05,
-      "loss": 0.7951,
+      "loss": 0.5591,
       "step": 2900
     },
     {
       "epoch": 7.15,
       "learning_rate": 1.573617952928298e-05,
-      "loss": 0.8414,
+      "loss": 0.6061,
       "step": 2910
     },
     {
       "epoch": 7.18,
       "learning_rate": 1.5599343185550085e-05,
-      "loss": 0.8405,
+      "loss": 0.5408,
       "step": 2920
     },
     {
       "epoch": 7.2,
       "learning_rate": 1.5462506841817186e-05,
-      "loss": 0.8375,
+      "loss": 0.5585,
       "step": 2930
     },
     {
       "epoch": 7.23,
       "learning_rate": 1.532567049808429e-05,
-      "loss": 0.8256,
+      "loss": 0.5551,
       "step": 2940
     },
     {
       "epoch": 7.25,
       "learning_rate": 1.5188834154351398e-05,
-      "loss": 0.9106,
+      "loss": 0.5976,
       "step": 2950
     },
     {
       "epoch": 7.28,
       "learning_rate": 1.50519978106185e-05,
-      "loss": 0.7944,
+      "loss": 0.5359,
       "step": 2960
     },
     {
       "epoch": 7.3,
       "learning_rate": 1.4915161466885605e-05,
-      "loss": 0.8647,
+      "loss": 0.5213,
       "step": 2970
     },
     {
       "epoch": 7.33,
       "learning_rate": 1.4778325123152711e-05,
-      "loss": 0.7921,
+      "loss": 0.552,
       "step": 2980
     },
     {
       "epoch": 7.35,
       "learning_rate": 1.4641488779419815e-05,
-      "loss": 0.7978,
+      "loss": 0.5372,
       "step": 2990
     },
     {
       "epoch": 7.38,
       "learning_rate": 1.450465243568692e-05,
-      "loss": 0.8136,
+      "loss": 0.61,
       "step": 3000
     },
     {
       "epoch": 7.4,
       "learning_rate": 1.4367816091954022e-05,
-      "loss": 0.8278,
+      "loss": 0.5742,
       "step": 3010
     },
     {
       "epoch": 7.42,
       "learning_rate": 1.4230979748221129e-05,
-      "loss": 0.8061,
+      "loss": 0.5856,
       "step": 3020
     },
     {
       "epoch": 7.45,
       "learning_rate": 1.4094143404488233e-05,
-      "loss": 0.7277,
+      "loss": 0.5193,
       "step": 3030
     },
     {
       "epoch": 7.47,
       "learning_rate": 1.3957307060755337e-05,
-      "loss": 0.809,
+      "loss": 0.5345,
       "step": 3040
     },
     {
       "epoch": 7.5,
       "learning_rate": 1.382047071702244e-05,
-      "loss": 0.8359,
+      "loss": 0.5558,
       "step": 3050
     },
     {
       "epoch": 7.52,
       "learning_rate": 1.3683634373289548e-05,
-      "loss": 0.8107,
+      "loss": 0.58,
       "step": 3060
     },
     {
       "epoch": 7.55,
       "learning_rate": 1.354679802955665e-05,
-      "loss": 0.8016,
+      "loss": 0.5425,
       "step": 3070
     },
     {
       "epoch": 7.57,
       "learning_rate": 1.3409961685823755e-05,
-      "loss": 0.8512,
+      "loss": 0.6588,
       "step": 3080
     },
     {
       "epoch": 7.6,
       "learning_rate": 1.3273125342090861e-05,
-      "loss": 0.7948,
+      "loss": 0.5137,
       "step": 3090
     },
     {
       "epoch": 7.62,
       "learning_rate": 1.3136288998357965e-05,
-      "loss": 0.8081,
+      "loss": 0.5975,
       "step": 3100
     },
     {
       "epoch": 7.65,
       "learning_rate": 1.2999452654625068e-05,
-      "loss": 0.854,
+      "loss": 0.5728,
       "step": 3110
     },
     {
       "epoch": 7.67,
       "learning_rate": 1.2862616310892173e-05,
-      "loss": 0.8504,
+      "loss": 0.5112,
       "step": 3120
     },
     {
       "epoch": 7.7,
       "learning_rate": 1.2725779967159279e-05,
-      "loss": 0.824,
+      "loss": 0.5435,
       "step": 3130
     },
     {
       "epoch": 7.72,
       "learning_rate": 1.2588943623426383e-05,
-      "loss": 0.8119,
+      "loss": 0.5792,
       "step": 3140
     },
     {
       "epoch": 7.74,
       "learning_rate": 1.2452107279693487e-05,
-      "loss": 0.8304,
+      "loss": 0.5114,
       "step": 3150
     },
     {
       "epoch": 7.77,
       "learning_rate": 1.2315270935960592e-05,
-      "loss": 0.7737,
+      "loss": 0.5976,
       "step": 3160
     },
     {
       "epoch": 7.79,
       "learning_rate": 1.2178434592227696e-05,
-      "loss": 0.8032,
+      "loss": 0.5955,
       "step": 3170
     },
     {
       "epoch": 7.82,
       "learning_rate": 1.20415982484948e-05,
-      "loss": 0.8448,
+      "loss": 0.5476,
       "step": 3180
     },
     {
       "epoch": 7.84,
       "learning_rate": 1.1904761904761905e-05,
-      "loss": 0.755,
+      "loss": 0.5329,
       "step": 3190
     },
     {
       "epoch": 7.87,
       "learning_rate": 1.176792556102901e-05,
-      "loss": 0.812,
+      "loss": 0.5765,
       "step": 3200
     },
     {
       "epoch": 7.89,
       "learning_rate": 1.1631089217296116e-05,
-      "loss": 0.8194,
+      "loss": 0.6101,
       "step": 3210
     },
     {
       "epoch": 7.92,
       "learning_rate": 1.1494252873563218e-05,
-      "loss": 0.8348,
+      "loss": 0.673,
       "step": 3220
     },
     {
       "epoch": 7.94,
       "learning_rate": 1.1357416529830324e-05,
-      "loss": 0.7492,
+      "loss": 0.4893,
       "step": 3230
     },
     {
       "epoch": 7.97,
       "learning_rate": 1.1220580186097427e-05,
-      "loss": 0.8227,
+      "loss": 0.4964,
       "step": 3240
     },
     {
       "epoch": 7.99,
       "learning_rate": 1.1083743842364533e-05,
-      "loss": 0.8336,
+      "loss": 0.5653,
       "step": 3250
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.9334485738980121,
-      "eval_loss": 0.3332645893096924,
-      "eval_runtime": 29.4073,
-      "eval_samples_per_second": 196.72,
-      "eval_steps_per_second": 6.155,
+      "eval_accuracy": 0.9652549697493518,
+      "eval_loss": 0.18624082207679749,
+      "eval_runtime": 27.8689,
+      "eval_samples_per_second": 207.579,
+      "eval_steps_per_second": 6.495,
       "step": 3254
     },
     {
       "epoch": 8.01,
       "learning_rate": 1.0946907498631636e-05,
-      "loss": 0.7665,
+      "loss": 0.5571,
       "step": 3260
     },
     {
       "epoch": 8.04,
       "learning_rate": 1.0810071154898742e-05,
-      "loss": 0.807,
+      "loss": 0.5585,
       "step": 3270
     },
     {
       "epoch": 8.06,
       "learning_rate": 1.0673234811165846e-05,
-      "loss": 0.8521,
+      "loss": 0.5709,
       "step": 3280
     },
     {
       "epoch": 8.09,
       "learning_rate": 1.053639846743295e-05,
-      "loss": 0.7569,
+      "loss": 0.5892,
       "step": 3290
     },
     {
       "epoch": 8.11,
       "learning_rate": 1.0399562123700055e-05,
-      "loss": 0.7465,
+      "loss": 0.5463,
       "step": 3300
     },
     {
       "epoch": 8.14,
       "learning_rate": 1.026272577996716e-05,
-      "loss": 0.8432,
+      "loss": 0.5466,
       "step": 3310
     },
     {
       "epoch": 8.16,
       "learning_rate": 1.0125889436234266e-05,
-      "loss": 0.7659,
+      "loss": 0.5939,
       "step": 3320
     },
     {
       "epoch": 8.19,
       "learning_rate": 9.989053092501368e-06,
-      "loss": 0.7682,
+      "loss": 0.5314,
       "step": 3330
     },
     {
       "epoch": 8.21,
       "learning_rate": 9.852216748768475e-06,
-      "loss": 0.7285,
+      "loss": 0.4986,
       "step": 3340
     },
     {
       "epoch": 8.24,
       "learning_rate": 9.715380405035577e-06,
-      "loss": 0.8721,
+      "loss": 0.622,
       "step": 3350
     },
     {
       "epoch": 8.26,
       "learning_rate": 9.578544061302683e-06,
-      "loss": 0.761,
+      "loss": 0.5836,
       "step": 3360
     },
     {
       "epoch": 8.29,
       "learning_rate": 9.441707717569786e-06,
-      "loss": 0.7812,
+      "loss": 0.5582,
       "step": 3370
     },
     {
       "epoch": 8.31,
       "learning_rate": 9.304871373836892e-06,
-      "loss": 0.8125,
+      "loss": 0.5248,
       "step": 3380
     },
     {
       "epoch": 8.33,
       "learning_rate": 9.168035030103995e-06,
-      "loss": 0.8344,
+      "loss": 0.5676,
       "step": 3390
     },
     {
       "epoch": 8.36,
       "learning_rate": 9.031198686371101e-06,
-      "loss": 0.7884,
+      "loss": 0.5659,
       "step": 3400
     },
     {
       "epoch": 8.38,
       "learning_rate": 8.894362342638204e-06,
-      "loss": 0.7692,
+      "loss": 0.5762,
       "step": 3410
     },
     {
       "epoch": 8.41,
       "learning_rate": 8.75752599890531e-06,
-      "loss": 0.875,
+      "loss": 0.5365,
       "step": 3420
     },
     {
       "epoch": 8.43,
       "learning_rate": 8.620689655172414e-06,
-      "loss": 0.7938,
+      "loss": 0.5394,
       "step": 3430
     },
     {
       "epoch": 8.46,
       "learning_rate": 8.483853311439519e-06,
-      "loss": 0.7749,
+      "loss": 0.5327,
       "step": 3440
     },
     {
       "epoch": 8.48,
       "learning_rate": 8.347016967706623e-06,
-      "loss": 0.7628,
+      "loss": 0.5503,
       "step": 3450
     },
     {
       "epoch": 8.51,
       "learning_rate": 8.210180623973727e-06,
-      "loss": 0.8007,
+      "loss": 0.4698,
       "step": 3460
     },
     {
       "epoch": 8.53,
       "learning_rate": 8.073344280240833e-06,
-      "loss": 0.7281,
+      "loss": 0.5264,
       "step": 3470
     },
     {
       "epoch": 8.56,
       "learning_rate": 7.936507936507936e-06,
-      "loss": 0.7633,
+      "loss": 0.5127,
       "step": 3480
     },
     {
       "epoch": 8.58,
       "learning_rate": 7.799671592775042e-06,
-      "loss": 0.8202,
+      "loss": 0.5655,
       "step": 3490
     },
     {
       "epoch": 8.6,
       "learning_rate": 7.662835249042145e-06,
-      "loss": 0.8123,
+      "loss": 0.5611,
       "step": 3500
     },
     {
       "epoch": 8.63,
       "learning_rate": 7.52599890530925e-06,
-      "loss": 0.8223,
+      "loss": 0.5306,
       "step": 3510
     },
     {
       "epoch": 8.65,
       "learning_rate": 7.3891625615763555e-06,
-      "loss": 0.8456,
+      "loss": 0.531,
       "step": 3520
     },
     {
       "epoch": 8.68,
       "learning_rate": 7.25232621784346e-06,
-      "loss": 0.752,
+      "loss": 0.5006,
       "step": 3530
     },
     {
       "epoch": 8.7,
       "learning_rate": 7.115489874110564e-06,
-      "loss": 0.7942,
+      "loss": 0.5036,
       "step": 3540
     },
     {
       "epoch": 8.73,
       "learning_rate": 6.978653530377669e-06,
-      "loss": 0.7635,
+      "loss": 0.5519,
       "step": 3550
     },
     {
       "epoch": 8.75,
       "learning_rate": 6.841817186644774e-06,
-      "loss": 0.7014,
+      "loss": 0.5233,
       "step": 3560
     },
     {
       "epoch": 8.78,
       "learning_rate": 6.7049808429118775e-06,
-      "loss": 0.7572,
+      "loss": 0.567,
       "step": 3570
     },
     {
       "epoch": 8.8,
       "learning_rate": 6.568144499178983e-06,
-      "loss": 0.7816,
+      "loss": 0.4994,
       "step": 3580
     },
     {
       "epoch": 8.83,
       "learning_rate": 6.431308155446086e-06,
-      "loss": 0.7336,
+      "loss": 0.5869,
       "step": 3590
     },
     {
       "epoch": 8.85,
       "learning_rate": 6.2944718117131915e-06,
-      "loss": 0.7665,
+      "loss": 0.5738,
       "step": 3600
     },
     {
       "epoch": 8.88,
       "learning_rate": 6.157635467980296e-06,
-      "loss": 0.7669,
+      "loss": 0.4842,
       "step": 3610
     },
     {
       "epoch": 8.9,
       "learning_rate": 6.0207991242474e-06,
-      "loss": 0.7587,
+      "loss": 0.5135,
       "step": 3620
     },
     {
       "epoch": 8.92,
       "learning_rate": 5.883962780514505e-06,
-      "loss": 0.8309,
+      "loss": 0.5603,
       "step": 3630
     },
     {
       "epoch": 8.95,
       "learning_rate": 5.747126436781609e-06,
-      "loss": 0.7201,
+      "loss": 0.5068,
       "step": 3640
     },
     {
       "epoch": 8.97,
       "learning_rate": 5.6102900930487136e-06,
-      "loss": 0.736,
+      "loss": 0.5566,
       "step": 3650
     },
     {
       "epoch": 9.0,
       "learning_rate": 5.473453749315818e-06,
-      "loss": 0.758,
+      "loss": 0.5361,
       "step": 3660
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.9405358686257562,
-      "eval_loss": 0.3180586099624634,
-      "eval_runtime": 28.3456,
-      "eval_samples_per_second": 204.088,
-      "eval_steps_per_second": 6.385,
+      "eval_accuracy": 0.9659464131374244,
+      "eval_loss": 0.18145236372947693,
+      "eval_runtime": 28.6712,
+      "eval_samples_per_second": 201.771,
+      "eval_steps_per_second": 6.313,
       "step": 3660
     },
     {
       "epoch": 9.02,
       "learning_rate": 5.336617405582923e-06,
-      "loss": 0.6856,
+      "loss": 0.5322,
       "step": 3670
     },
     {
       "epoch": 9.05,
       "learning_rate": 5.199781061850028e-06,
-      "loss": 0.8134,
+      "loss": 0.5509,
       "step": 3680
     },
     {
       "epoch": 9.07,
       "learning_rate": 5.062944718117133e-06,
-      "loss": 0.8841,
+      "loss": 0.5806,
       "step": 3690
     },
     {
       "epoch": 9.1,
       "learning_rate": 4.926108374384237e-06,
-      "loss": 0.7623,
+      "loss": 0.4942,
       "step": 3700
     },
     {
       "epoch": 9.12,
       "learning_rate": 4.789272030651342e-06,
-      "loss": 0.7488,
+      "loss": 0.5086,
       "step": 3710
     },
     {
       "epoch": 9.15,
       "learning_rate": 4.652435686918446e-06,
-      "loss": 0.7433,
+      "loss": 0.5668,
       "step": 3720
     },
     {
       "epoch": 9.17,
       "learning_rate": 4.5155993431855505e-06,
-      "loss": 0.7512,
+      "loss": 0.5554,
       "step": 3730
     },
     {
       "epoch": 9.19,
       "learning_rate": 4.378762999452655e-06,
-      "loss": 0.7725,
+      "loss": 0.5122,
       "step": 3740
     },
     {
       "epoch": 9.22,
       "learning_rate": 4.241926655719759e-06,
-      "loss": 0.7567,
+      "loss": 0.5301,
       "step": 3750
     },
     {
       "epoch": 9.24,
       "learning_rate": 4.105090311986864e-06,
-      "loss": 0.7213,
+      "loss": 0.5617,
       "step": 3760
     },
     {
       "epoch": 9.27,
       "learning_rate": 3.968253968253968e-06,
-      "loss": 0.75,
+      "loss": 0.5423,
       "step": 3770
     },
     {
       "epoch": 9.29,
       "learning_rate": 3.8314176245210725e-06,
-      "loss": 0.7999,
+      "loss": 0.6171,
       "step": 3780
     },
     {
       "epoch": 9.32,
       "learning_rate": 3.6945812807881777e-06,
-      "loss": 0.7489,
+      "loss": 0.498,
       "step": 3790
     },
     {
       "epoch": 9.34,
       "learning_rate": 3.557744937055282e-06,
-      "loss": 0.7591,
+      "loss": 0.5354,
       "step": 3800
     },
     {
       "epoch": 9.37,
       "learning_rate": 3.420908593322387e-06,
-      "loss": 0.769,
+      "loss": 0.5846,
       "step": 3810
     },
     {
       "epoch": 9.39,
       "learning_rate": 3.2840722495894914e-06,
-      "loss": 0.7171,
+      "loss": 0.502,
       "step": 3820
     },
     {
       "epoch": 9.42,
       "learning_rate": 3.1472359058565958e-06,
-      "loss": 0.7158,
+      "loss": 0.5138,
       "step": 3830
     },
     {
       "epoch": 9.44,
       "learning_rate": 3.0103995621237e-06,
-      "loss": 0.7132,
+      "loss": 0.5014,
       "step": 3840
     },
     {
       "epoch": 9.47,
       "learning_rate": 2.8735632183908046e-06,
-      "loss": 0.7236,
+      "loss": 0.4884,
       "step": 3850
     },
     {
       "epoch": 9.49,
       "learning_rate": 2.736726874657909e-06,
-      "loss": 0.7361,
+      "loss": 0.6063,
       "step": 3860
     },
     {
       "epoch": 9.51,
       "learning_rate": 2.599890530925014e-06,
-      "loss": 0.6774,
+      "loss": 0.5386,
       "step": 3870
     },
     {
       "epoch": 9.54,
       "learning_rate": 2.4630541871921186e-06,
-      "loss": 0.8082,
+      "loss": 0.5255,
       "step": 3880
     },
     {
       "epoch": 9.56,
       "learning_rate": 2.326217843459223e-06,
-      "loss": 0.776,
+      "loss": 0.5019,
       "step": 3890
     },
     {
       "epoch": 9.59,
       "learning_rate": 2.1893814997263274e-06,
-      "loss": 0.7655,
+      "loss": 0.503,
       "step": 3900
     },
     {
       "epoch": 9.61,
       "learning_rate": 2.052545155993432e-06,
-      "loss": 0.7834,
+      "loss": 0.5699,
       "step": 3910
     },
     {
       "epoch": 9.64,
       "learning_rate": 1.9157088122605362e-06,
-      "loss": 0.7236,
+      "loss": 0.4565,
       "step": 3920
     },
     {
       "epoch": 9.66,
       "learning_rate": 1.778872468527641e-06,
-      "loss": 0.6974,
+      "loss": 0.5098,
       "step": 3930
     },
     {
       "epoch": 9.69,
       "learning_rate": 1.6420361247947457e-06,
-      "loss": 0.73,
+      "loss": 0.5032,
       "step": 3940
     },
     {
       "epoch": 9.71,
       "learning_rate": 1.50519978106185e-06,
-      "loss": 0.6768,
+      "loss": 0.4713,
       "step": 3950
     },
     {
       "epoch": 9.74,
       "learning_rate": 1.3683634373289545e-06,
-      "loss": 0.7841,
+      "loss": 0.4554,
       "step": 3960
     },
     {
       "epoch": 9.76,
       "learning_rate": 1.2315270935960593e-06,
-      "loss": 0.759,
+      "loss": 0.4773,
       "step": 3970
     },
     {
       "epoch": 9.78,
       "learning_rate": 1.0946907498631637e-06,
-      "loss": 0.7244,
+      "loss": 0.5536,
       "step": 3980
     },
     {
       "epoch": 9.81,
       "learning_rate": 9.578544061302681e-07,
-      "loss": 0.7326,
+      "loss": 0.4913,
       "step": 3990
     },
     {
       "epoch": 9.83,
       "learning_rate": 8.210180623973728e-07,
-      "loss": 0.7484,
+      "loss": 0.5538,
       "step": 4000
     },
     {
       "epoch": 9.86,
       "learning_rate": 6.841817186644772e-07,
-      "loss": 0.7589,
+      "loss": 0.4838,
       "step": 4010
     },
     {
       "epoch": 9.88,
       "learning_rate": 5.473453749315819e-07,
-      "loss": 0.7169,
+      "loss": 0.569,
       "step": 4020
     },
     {
       "epoch": 9.91,
       "learning_rate": 4.105090311986864e-07,
-      "loss": 0.6762,
+      "loss": 0.5847,
       "step": 4030
     },
     {
       "epoch": 9.93,
       "learning_rate": 2.7367268746579093e-07,
-      "loss": 0.7896,
+      "loss": 0.5507,
       "step": 4040
     },
     {
       "epoch": 9.96,
       "learning_rate": 1.3683634373289546e-07,
-      "loss": 0.8115,
+      "loss": 0.5834,
       "step": 4050
     },
     {
       "epoch": 9.98,
       "learning_rate": 0.0,
-      "loss": 0.7343,
+      "loss": 0.5017,
       "step": 4060
     },
     {
       "epoch": 9.98,
-      "eval_accuracy": 0.9403630077787382,
-      "eval_loss": 0.31005963683128357,
-      "eval_runtime": 28.1514,
-      "eval_samples_per_second": 205.496,
-      "eval_steps_per_second": 6.43,
+      "eval_accuracy": 0.9676750216076059,
+      "eval_loss": 0.17719660699367523,
+      "eval_runtime": 28.9267,
+      "eval_samples_per_second": 199.989,
+      "eval_steps_per_second": 6.257,
       "step": 4060
     },
     {
       "epoch": 9.98,
       "step": 4060,
       "total_flos": 1.3140619208067262e+19,
-      "train_loss": 1.0548507860728673,
-      "train_runtime": 4291.8856,
-      "train_samples_per_second": 121.301,
-      "train_steps_per_second": 0.946
+      "train_loss": 0.6519044913681857,
+      "train_runtime": 4257.1599,
+      "train_samples_per_second": 122.29,
+      "train_steps_per_second": 0.954
     }
   ],
   "logging_steps": 10,