End of training

README.md

@@ -2,6 +2,8 @@
 license: apache-2.0
 base_model: google/vit-base-patch16-224-in21k
 tags:
+- image-classification
+- vision
 - generated_from_trainer
 datasets:
 - imagefolder
@@ -22,7 +24,7 @@ model-index:
     metrics:
     - name: Accuracy
       type: accuracy
-      value: 0.5538461538461539
+      value: 0.5461538461538461
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -32,8 +34,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the imagefolder dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.8622
-- Accuracy: 0.5538
+- Loss: 2.8599
+- Accuracy: 0.5462
 
 ## Model description
 

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 100.0,
-    "eval_accuracy": 0.15306122448979592,
-    "eval_loss": 3.643535614013672,
-    "eval_runtime": 1.6218,
-    "eval_samples_per_second": 60.427,
-    "eval_steps_per_second": 1.233,
-    "train_loss": 3.289549721309117,
-    "train_runtime": 4059.8356,
-    "train_samples_per_second": 13.597,
-    "train_steps_per_second": 0.172
+    "eval_accuracy": 0.5461538461538461,
+    "eval_loss": 2.859877347946167,
+    "eval_runtime": 2.5644,
+    "eval_samples_per_second": 50.694,
+    "eval_steps_per_second": 0.78,
+    "train_loss": 2.6316724504743303,
+    "train_runtime": 4344.3701,
+    "train_samples_per_second": 11.97,
+    "train_steps_per_second": 0.161
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 100.0,
-    "eval_accuracy": 0.15306122448979592,
-    "eval_loss": 3.643535614013672,
-    "eval_runtime": 1.6218,
-    "eval_samples_per_second": 60.427,
-    "eval_steps_per_second": 1.233
+    "eval_accuracy": 0.5461538461538461,
+    "eval_loss": 2.859877347946167,
+    "eval_runtime": 2.5644,
+    "eval_samples_per_second": 50.694,
+    "eval_steps_per_second": 0.78
 }

runs/Jan17_18-13-33_d06676088071/events.out.tfevents.1705519839.d06676088071.1503.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d63a6efa2ac983c8a4ef9eed4231219ac5348c0a7d1f235e72d3287697daddb4
+size 411

train_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 100.0,
-    "train_loss": 3.289549721309117,
-    "train_runtime": 4059.8356,
-    "train_samples_per_second": 13.597,
-    "train_steps_per_second": 0.172
+    "train_loss": 2.6316724504743303,
+    "train_runtime": 4344.3701,
+    "train_samples_per_second": 11.97,
+    "train_steps_per_second": 0.161
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 3.643535614013672,
-  "best_model_checkpoint": "./drive/MyDrive/repositories/torch_example_image-classification/outputs_letter3/checkpoint-700",
+  "best_metric": 2.859877347946167,
+  "best_model_checkpoint": "./drive/MyDrive/repositories/torch_example_image-classification/outputs_letter3/checkpoint-644",
   "epoch": 100.0,
   "eval_steps": 500,
   "global_step": 700,
@@ -10,1332 +10,1332 @@
   "log_history": [
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.030612244897959183,
-      "eval_loss": 3.945223808288574,
-      "eval_runtime": 35.9416,
-      "eval_samples_per_second": 2.727,
-      "eval_steps_per_second": 0.056,
+      "eval_accuracy": 0.015384615384615385,
+      "eval_loss": 3.9449453353881836,
+      "eval_runtime": 2.0983,
+      "eval_samples_per_second": 61.955,
+      "eval_steps_per_second": 0.953,
       "step": 7
     },
     {
       "epoch": 1.43,
       "learning_rate": 1.9714285714285718e-05,
-      "loss": 3.9498,
+      "loss": 3.9333,
       "step": 10
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.05102040816326531,
-      "eval_loss": 3.943814277648926,
-      "eval_runtime": 1.6003,
-      "eval_samples_per_second": 61.237,
-      "eval_steps_per_second": 1.25,
+      "eval_accuracy": 0.023076923076923078,
+      "eval_loss": 3.9366512298583984,
+      "eval_runtime": 2.2606,
+      "eval_samples_per_second": 57.506,
+      "eval_steps_per_second": 0.885,
       "step": 14
     },
     {
       "epoch": 2.86,
       "learning_rate": 1.942857142857143e-05,
-      "loss": 3.9413,
+      "loss": 3.8939,
       "step": 20
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.04081632653061224,
-      "eval_loss": 3.9436657428741455,
-      "eval_runtime": 1.5584,
-      "eval_samples_per_second": 62.886,
-      "eval_steps_per_second": 1.283,
+      "eval_accuracy": 0.03076923076923077,
+      "eval_loss": 3.9279532432556152,
+      "eval_runtime": 2.3275,
+      "eval_samples_per_second": 55.855,
+      "eval_steps_per_second": 0.859,
       "step": 21
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.04081632653061224,
-      "eval_loss": 3.9431352615356445,
-      "eval_runtime": 1.6475,
-      "eval_samples_per_second": 59.483,
-      "eval_steps_per_second": 1.214,
+      "eval_accuracy": 0.046153846153846156,
+      "eval_loss": 3.9167158603668213,
+      "eval_runtime": 2.2361,
+      "eval_samples_per_second": 58.137,
+      "eval_steps_per_second": 0.894,
       "step": 28
     },
     {
       "epoch": 4.29,
       "learning_rate": 1.9142857142857146e-05,
-      "loss": 3.9255,
+      "loss": 3.8562,
       "step": 30
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.04081632653061224,
-      "eval_loss": 3.9424338340759277,
-      "eval_runtime": 1.5527,
-      "eval_samples_per_second": 63.114,
-      "eval_steps_per_second": 1.288,
+      "eval_accuracy": 0.06923076923076923,
+      "eval_loss": 3.9032633304595947,
+      "eval_runtime": 2.4923,
+      "eval_samples_per_second": 52.161,
+      "eval_steps_per_second": 0.802,
       "step": 35
     },
     {
       "epoch": 5.71,
       "learning_rate": 1.885714285714286e-05,
-      "loss": 3.9132,
+      "loss": 3.8008,
       "step": 40
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.030612244897959183,
-      "eval_loss": 3.9400839805603027,
-      "eval_runtime": 1.6728,
-      "eval_samples_per_second": 58.585,
-      "eval_steps_per_second": 1.196,
+      "eval_accuracy": 0.07692307692307693,
+      "eval_loss": 3.8873634338378906,
+      "eval_runtime": 2.5262,
+      "eval_samples_per_second": 51.461,
+      "eval_steps_per_second": 0.792,
       "step": 42
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.030612244897959183,
-      "eval_loss": 3.937299966812134,
-      "eval_runtime": 1.6864,
-      "eval_samples_per_second": 58.111,
-      "eval_steps_per_second": 1.186,
+      "eval_accuracy": 0.1076923076923077,
+      "eval_loss": 3.867002010345459,
+      "eval_runtime": 2.5898,
+      "eval_samples_per_second": 50.196,
+      "eval_steps_per_second": 0.772,
       "step": 49
     },
     {
       "epoch": 7.14,
       "learning_rate": 1.8571428571428575e-05,
-      "loss": 3.8913,
+      "loss": 3.7555,
       "step": 50
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.02040816326530612,
-      "eval_loss": 3.9351377487182617,
-      "eval_runtime": 1.7475,
-      "eval_samples_per_second": 56.08,
-      "eval_steps_per_second": 1.144,
+      "eval_accuracy": 0.1,
+      "eval_loss": 3.8494887351989746,
+      "eval_runtime": 2.6867,
+      "eval_samples_per_second": 48.387,
+      "eval_steps_per_second": 0.744,
       "step": 56
     },
     {
       "epoch": 8.57,
       "learning_rate": 1.8285714285714288e-05,
-      "loss": 3.8685,
+      "loss": 3.6917,
       "step": 60
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.02040816326530612,
-      "eval_loss": 3.931184768676758,
-      "eval_runtime": 1.7165,
-      "eval_samples_per_second": 57.092,
-      "eval_steps_per_second": 1.165,
+      "eval_accuracy": 0.11538461538461539,
+      "eval_loss": 3.8304965496063232,
+      "eval_runtime": 2.4963,
+      "eval_samples_per_second": 52.077,
+      "eval_steps_per_second": 0.801,
       "step": 63
     },
     {
       "epoch": 10.0,
       "learning_rate": 1.8e-05,
-      "loss": 3.8413,
+      "loss": 3.6372,
       "step": 70
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.030612244897959183,
-      "eval_loss": 3.9259071350097656,
-      "eval_runtime": 1.7086,
-      "eval_samples_per_second": 57.357,
-      "eval_steps_per_second": 1.171,
+      "eval_accuracy": 0.13846153846153847,
+      "eval_loss": 3.8138267993927,
+      "eval_runtime": 2.3942,
+      "eval_samples_per_second": 54.298,
+      "eval_steps_per_second": 0.835,
       "step": 70
     },
     {
       "epoch": 11.0,
-      "eval_accuracy": 0.030612244897959183,
-      "eval_loss": 3.9218833446502686,
-      "eval_runtime": 1.6006,
-      "eval_samples_per_second": 61.228,
-      "eval_steps_per_second": 1.25,
+      "eval_accuracy": 0.12307692307692308,
+      "eval_loss": 3.796628952026367,
+      "eval_runtime": 2.3929,
+      "eval_samples_per_second": 54.328,
+      "eval_steps_per_second": 0.836,
       "step": 77
     },
     {
       "epoch": 11.43,
       "learning_rate": 1.7714285714285717e-05,
-      "loss": 3.8163,
+      "loss": 3.5846,
       "step": 80
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.02040816326530612,
-      "eval_loss": 3.9182682037353516,
-      "eval_runtime": 1.8155,
-      "eval_samples_per_second": 53.98,
-      "eval_steps_per_second": 1.102,
+      "eval_accuracy": 0.15384615384615385,
+      "eval_loss": 3.7766666412353516,
+      "eval_runtime": 2.4605,
+      "eval_samples_per_second": 52.835,
+      "eval_steps_per_second": 0.813,
       "step": 84
     },
     {
       "epoch": 12.86,
       "learning_rate": 1.742857142857143e-05,
-      "loss": 3.7912,
+      "loss": 3.5047,
       "step": 90
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.04081632653061224,
-      "eval_loss": 3.9151253700256348,
-      "eval_runtime": 1.704,
-      "eval_samples_per_second": 57.511,
-      "eval_steps_per_second": 1.174,
+      "eval_accuracy": 0.23076923076923078,
+      "eval_loss": 3.751601219177246,
+      "eval_runtime": 2.5029,
+      "eval_samples_per_second": 51.94,
+      "eval_steps_per_second": 0.799,
       "step": 91
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.030612244897959183,
-      "eval_loss": 3.911611795425415,
-      "eval_runtime": 1.7037,
-      "eval_samples_per_second": 57.521,
-      "eval_steps_per_second": 1.174,
+      "eval_accuracy": 0.23846153846153847,
+      "eval_loss": 3.7278525829315186,
+      "eval_runtime": 2.6158,
+      "eval_samples_per_second": 49.699,
+      "eval_steps_per_second": 0.765,
       "step": 98
     },
     {
       "epoch": 14.29,
       "learning_rate": 1.7142857142857142e-05,
-      "loss": 3.7616,
+      "loss": 3.4547,
       "step": 100
     },
     {
       "epoch": 15.0,
-      "eval_accuracy": 0.04081632653061224,
-      "eval_loss": 3.9074392318725586,
-      "eval_runtime": 1.6801,
-      "eval_samples_per_second": 58.331,
-      "eval_steps_per_second": 1.19,
+      "eval_accuracy": 0.23846153846153847,
+      "eval_loss": 3.7031078338623047,
+      "eval_runtime": 2.5457,
+      "eval_samples_per_second": 51.066,
+      "eval_steps_per_second": 0.786,
       "step": 105
     },
     {
       "epoch": 15.71,
       "learning_rate": 1.6857142857142858e-05,
-      "loss": 3.734,
+      "loss": 3.3796,
       "step": 110
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.04081632653061224,
-      "eval_loss": 3.9029488563537598,
-      "eval_runtime": 1.7659,
-      "eval_samples_per_second": 55.496,
-      "eval_steps_per_second": 1.133,
+      "eval_accuracy": 0.2692307692307692,
+      "eval_loss": 3.672494649887085,
+      "eval_runtime": 2.368,
+      "eval_samples_per_second": 54.899,
+      "eval_steps_per_second": 0.845,
       "step": 112
     },
     {
       "epoch": 17.0,
-      "eval_accuracy": 0.061224489795918366,
-      "eval_loss": 3.8968818187713623,
-      "eval_runtime": 1.6451,
-      "eval_samples_per_second": 59.57,
-      "eval_steps_per_second": 1.216,
+      "eval_accuracy": 0.27692307692307694,
+      "eval_loss": 3.646197557449341,
+      "eval_runtime": 2.3752,
+      "eval_samples_per_second": 54.731,
+      "eval_steps_per_second": 0.842,
       "step": 119
     },
     {
       "epoch": 17.14,
       "learning_rate": 1.6571428571428574e-05,
-      "loss": 3.7014,
+      "loss": 3.3283,
       "step": 120
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.07142857142857142,
-      "eval_loss": 3.890695095062256,
-      "eval_runtime": 1.6901,
-      "eval_samples_per_second": 57.983,
-      "eval_steps_per_second": 1.183,
+      "eval_accuracy": 0.2923076923076923,
+      "eval_loss": 3.622591495513916,
+      "eval_runtime": 2.3748,
+      "eval_samples_per_second": 54.741,
+      "eval_steps_per_second": 0.842,
       "step": 126
     },
     {
       "epoch": 18.57,
       "learning_rate": 1.6285714285714287e-05,
-      "loss": 3.6707,
+      "loss": 3.2728,
       "step": 130
     },
     {
       "epoch": 19.0,
-      "eval_accuracy": 0.07142857142857142,
-      "eval_loss": 3.8844714164733887,
-      "eval_runtime": 1.6505,
-      "eval_samples_per_second": 59.374,
-      "eval_steps_per_second": 1.212,
+      "eval_accuracy": 0.2846153846153846,
+      "eval_loss": 3.6022486686706543,
+      "eval_runtime": 2.4033,
+      "eval_samples_per_second": 54.091,
+      "eval_steps_per_second": 0.832,
       "step": 133
     },
     {
       "epoch": 20.0,
       "learning_rate": 1.6000000000000003e-05,
-      "loss": 3.6307,
+      "loss": 3.2229,
       "step": 140
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.08163265306122448,
-      "eval_loss": 3.877913475036621,
-      "eval_runtime": 1.8041,
-      "eval_samples_per_second": 54.322,
-      "eval_steps_per_second": 1.109,
+      "eval_accuracy": 0.27692307692307694,
+      "eval_loss": 3.5929768085479736,
+      "eval_runtime": 2.4687,
+      "eval_samples_per_second": 52.66,
+      "eval_steps_per_second": 0.81,
       "step": 140
     },
     {
       "epoch": 21.0,
-      "eval_accuracy": 0.08163265306122448,
-      "eval_loss": 3.8703930377960205,
-      "eval_runtime": 1.7575,
-      "eval_samples_per_second": 55.762,
-      "eval_steps_per_second": 1.138,
+      "eval_accuracy": 0.33076923076923076,
+      "eval_loss": 3.574838876724243,
+      "eval_runtime": 2.5603,
+      "eval_samples_per_second": 50.775,
+      "eval_steps_per_second": 0.781,
       "step": 147
     },
     {
       "epoch": 21.43,
       "learning_rate": 1.5714285714285715e-05,
-      "loss": 3.596,
+      "loss": 3.1514,
       "step": 150
     },
     {
       "epoch": 22.0,
-      "eval_accuracy": 0.09183673469387756,
-      "eval_loss": 3.8646252155303955,
-      "eval_runtime": 1.6598,
-      "eval_samples_per_second": 59.042,
-      "eval_steps_per_second": 1.205,
+      "eval_accuracy": 0.3384615384615385,
+      "eval_loss": 3.5403976440429688,
+      "eval_runtime": 2.3284,
+      "eval_samples_per_second": 55.833,
+      "eval_steps_per_second": 0.859,
       "step": 154
     },
     {
       "epoch": 22.86,
       "learning_rate": 1.542857142857143e-05,
-      "loss": 3.5875,
+      "loss": 3.1179,
       "step": 160
     },
     {
       "epoch": 23.0,
-      "eval_accuracy": 0.09183673469387756,
-      "eval_loss": 3.8603618144989014,
-      "eval_runtime": 1.6878,
-      "eval_samples_per_second": 58.065,
-      "eval_steps_per_second": 1.185,
+      "eval_accuracy": 0.3384615384615385,
+      "eval_loss": 3.514573335647583,
+      "eval_runtime": 2.3901,
+      "eval_samples_per_second": 54.39,
+      "eval_steps_per_second": 0.837,
       "step": 161
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.09183673469387756,
-      "eval_loss": 3.8561482429504395,
-      "eval_runtime": 1.6677,
-      "eval_samples_per_second": 58.764,
-      "eval_steps_per_second": 1.199,
+      "eval_accuracy": 0.34615384615384615,
+      "eval_loss": 3.491649866104126,
+      "eval_runtime": 2.3495,
+      "eval_samples_per_second": 55.331,
+      "eval_steps_per_second": 0.851,
       "step": 168
     },
     {
       "epoch": 24.29,
       "learning_rate": 1.5142857142857144e-05,
-      "loss": 3.5532,
+      "loss": 3.0559,
       "step": 170
     },
     {
       "epoch": 25.0,
-      "eval_accuracy": 0.09183673469387756,
-      "eval_loss": 3.8509910106658936,
-      "eval_runtime": 1.7363,
-      "eval_samples_per_second": 56.441,
-      "eval_steps_per_second": 1.152,
+      "eval_accuracy": 0.3384615384615385,
+      "eval_loss": 3.473316192626953,
+      "eval_runtime": 2.387,
+      "eval_samples_per_second": 54.461,
+      "eval_steps_per_second": 0.838,
       "step": 175
     },
     {
       "epoch": 25.71,
       "learning_rate": 1.4857142857142858e-05,
-      "loss": 3.5374,
+      "loss": 3.0051,
       "step": 180
     },
     {
       "epoch": 26.0,
-      "eval_accuracy": 0.09183673469387756,
-      "eval_loss": 3.844221353530884,
-      "eval_runtime": 1.7776,
-      "eval_samples_per_second": 55.131,
-      "eval_steps_per_second": 1.125,
+      "eval_accuracy": 0.36153846153846153,
+      "eval_loss": 3.453972339630127,
+      "eval_runtime": 2.5367,
+      "eval_samples_per_second": 51.249,
+      "eval_steps_per_second": 0.788,
       "step": 182
     },
     {
       "epoch": 27.0,
-      "eval_accuracy": 0.10204081632653061,
-      "eval_loss": 3.83986496925354,
-      "eval_runtime": 1.6514,
-      "eval_samples_per_second": 59.342,
-      "eval_steps_per_second": 1.211,
+      "eval_accuracy": 0.36923076923076925,
+      "eval_loss": 3.4498963356018066,
+      "eval_runtime": 2.5601,
+      "eval_samples_per_second": 50.779,
+      "eval_steps_per_second": 0.781,
       "step": 189
     },
     {
       "epoch": 27.14,
       "learning_rate": 1.4571428571428573e-05,
-      "loss": 3.51,
+      "loss": 2.9775,
       "step": 190
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.11224489795918367,
-      "eval_loss": 3.8350086212158203,
-      "eval_runtime": 1.6793,
-      "eval_samples_per_second": 58.357,
-      "eval_steps_per_second": 1.191,
+      "eval_accuracy": 0.3769230769230769,
+      "eval_loss": 3.435525417327881,
+      "eval_runtime": 2.3663,
+      "eval_samples_per_second": 54.939,
+      "eval_steps_per_second": 0.845,
       "step": 196
     },
     {
       "epoch": 28.57,
       "learning_rate": 1.4285714285714287e-05,
-      "loss": 3.4842,
+      "loss": 2.9277,
       "step": 200
     },
     {
       "epoch": 29.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.8296377658843994,
-      "eval_runtime": 1.7517,
-      "eval_samples_per_second": 55.946,
-      "eval_steps_per_second": 1.142,
+      "eval_accuracy": 0.38461538461538464,
+      "eval_loss": 3.4166250228881836,
+      "eval_runtime": 2.3557,
+      "eval_samples_per_second": 55.185,
+      "eval_steps_per_second": 0.849,
       "step": 203
     },
     {
       "epoch": 30.0,
       "learning_rate": 1.4e-05,
-      "loss": 3.4495,
+      "loss": 2.9066,
       "step": 210
     },
     {
       "epoch": 30.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.8243255615234375,
-      "eval_runtime": 1.7837,
-      "eval_samples_per_second": 54.942,
-      "eval_steps_per_second": 1.121,
+      "eval_accuracy": 0.4,
+      "eval_loss": 3.40069842338562,
+      "eval_runtime": 2.3914,
+      "eval_samples_per_second": 54.362,
+      "eval_steps_per_second": 0.836,
       "step": 210
     },
     {
       "epoch": 31.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.8213043212890625,
-      "eval_runtime": 1.6545,
-      "eval_samples_per_second": 59.233,
-      "eval_steps_per_second": 1.209,
+      "eval_accuracy": 0.36923076923076925,
+      "eval_loss": 3.3825573921203613,
+      "eval_runtime": 2.5239,
+      "eval_samples_per_second": 51.509,
+      "eval_steps_per_second": 0.792,
       "step": 217
     },
     {
       "epoch": 31.43,
       "learning_rate": 1.3714285714285716e-05,
-      "loss": 3.4155,
+      "loss": 2.8464,
       "step": 220
     },
     {
       "epoch": 32.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.815812110900879,
-      "eval_runtime": 1.6721,
-      "eval_samples_per_second": 58.61,
-      "eval_steps_per_second": 1.196,
+      "eval_accuracy": 0.4076923076923077,
+      "eval_loss": 3.369760274887085,
+      "eval_runtime": 2.3953,
+      "eval_samples_per_second": 54.273,
+      "eval_steps_per_second": 0.835,
       "step": 224
     },
     {
       "epoch": 32.86,
       "learning_rate": 1.3428571428571429e-05,
-      "loss": 3.4257,
+      "loss": 2.8044,
       "step": 230
     },
     {
       "epoch": 33.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.8117594718933105,
-      "eval_runtime": 1.6581,
-      "eval_samples_per_second": 59.102,
-      "eval_steps_per_second": 1.206,
+      "eval_accuracy": 0.4076923076923077,
+      "eval_loss": 3.350893497467041,
+      "eval_runtime": 2.5141,
+      "eval_samples_per_second": 51.708,
+      "eval_steps_per_second": 0.796,
       "step": 231
     },
     {
       "epoch": 34.0,
-      "eval_accuracy": 0.1326530612244898,
-      "eval_loss": 3.8061439990997314,
-      "eval_runtime": 1.6272,
-      "eval_samples_per_second": 60.227,
-      "eval_steps_per_second": 1.229,
+      "eval_accuracy": 0.3769230769230769,
+      "eval_loss": 3.324298620223999,
+      "eval_runtime": 2.5107,
+      "eval_samples_per_second": 51.778,
+      "eval_steps_per_second": 0.797,
       "step": 238
     },
     {
       "epoch": 34.29,
       "learning_rate": 1.3142857142857145e-05,
-      "loss": 3.395,
+      "loss": 2.7699,
       "step": 240
     },
     {
       "epoch": 35.0,
-      "eval_accuracy": 0.1326530612244898,
-      "eval_loss": 3.8029837608337402,
-      "eval_runtime": 1.6449,
-      "eval_samples_per_second": 59.577,
-      "eval_steps_per_second": 1.216,
+      "eval_accuracy": 0.3923076923076923,
+      "eval_loss": 3.320146322250366,
+      "eval_runtime": 2.3695,
+      "eval_samples_per_second": 54.864,
+      "eval_steps_per_second": 0.844,
       "step": 245
     },
     {
       "epoch": 35.71,
       "learning_rate": 1.2857142857142859e-05,
-      "loss": 3.3693,
+      "loss": 2.7251,
       "step": 250
     },
     {
       "epoch": 36.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.795672655105591,
-      "eval_runtime": 1.6624,
-      "eval_samples_per_second": 58.953,
-      "eval_steps_per_second": 1.203,
+      "eval_accuracy": 0.4,
+      "eval_loss": 3.301337242126465,
+      "eval_runtime": 2.3948,
+      "eval_samples_per_second": 54.285,
+      "eval_steps_per_second": 0.835,
       "step": 252
     },
     {
       "epoch": 37.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.790419578552246,
-      "eval_runtime": 1.6455,
-      "eval_samples_per_second": 59.556,
-      "eval_steps_per_second": 1.215,
+      "eval_accuracy": 0.4230769230769231,
+      "eval_loss": 3.2935874462127686,
+      "eval_runtime": 2.3921,
+      "eval_samples_per_second": 54.345,
+      "eval_steps_per_second": 0.836,
       "step": 259
     },
     {
       "epoch": 37.14,
       "learning_rate": 1.2571428571428572e-05,
-      "loss": 3.35,
+      "loss": 2.6915,
       "step": 260
     },
     {
       "epoch": 38.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.783411741256714,
-      "eval_runtime": 1.6562,
-      "eval_samples_per_second": 59.172,
-      "eval_steps_per_second": 1.208,
+      "eval_accuracy": 0.45384615384615384,
+      "eval_loss": 3.2827184200286865,
+      "eval_runtime": 2.4835,
+      "eval_samples_per_second": 52.345,
+      "eval_steps_per_second": 0.805,
       "step": 266
     },
     {
       "epoch": 38.57,
       "learning_rate": 1.2285714285714288e-05,
-      "loss": 3.3453,
+      "loss": 2.6527,
       "step": 270
     },
     {
       "epoch": 39.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.778721332550049,
-      "eval_runtime": 1.7932,
-      "eval_samples_per_second": 54.651,
-      "eval_steps_per_second": 1.115,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.262709379196167,
+      "eval_runtime": 2.5052,
+      "eval_samples_per_second": 51.892,
+      "eval_steps_per_second": 0.798,
       "step": 273
     },
     {
       "epoch": 40.0,
       "learning_rate": 1.2e-05,
-      "loss": 3.2977,
+      "loss": 2.6541,
       "step": 280
     },
     {
       "epoch": 40.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.772735357284546,
-      "eval_runtime": 1.6336,
-      "eval_samples_per_second": 59.991,
-      "eval_steps_per_second": 1.224,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.2581355571746826,
+      "eval_runtime": 2.5207,
+      "eval_samples_per_second": 51.573,
+      "eval_steps_per_second": 0.793,
       "step": 280
     },
     {
       "epoch": 41.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.7680680751800537,
-      "eval_runtime": 1.7171,
-      "eval_samples_per_second": 57.072,
-      "eval_steps_per_second": 1.165,
+      "eval_accuracy": 0.4230769230769231,
+      "eval_loss": 3.2342007160186768,
+      "eval_runtime": 2.363,
+      "eval_samples_per_second": 55.016,
+      "eval_steps_per_second": 0.846,
       "step": 287
     },
     {
       "epoch": 41.43,
       "learning_rate": 1.1714285714285716e-05,
-      "loss": 3.2875,
+      "loss": 2.5968,
       "step": 290
     },
     {
       "epoch": 42.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.762765407562256,
-      "eval_runtime": 1.6284,
-      "eval_samples_per_second": 60.181,
-      "eval_steps_per_second": 1.228,
+      "eval_accuracy": 0.43846153846153846,
+      "eval_loss": 3.2211241722106934,
+      "eval_runtime": 2.5033,
+      "eval_samples_per_second": 51.932,
+      "eval_steps_per_second": 0.799,
       "step": 294
     },
     {
       "epoch": 42.86,
       "learning_rate": 1.1428571428571429e-05,
-      "loss": 3.2504,
+      "loss": 2.573,
       "step": 300
     },
     {
       "epoch": 43.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.758150100708008,
-      "eval_runtime": 1.6731,
-      "eval_samples_per_second": 58.575,
-      "eval_steps_per_second": 1.195,
+      "eval_accuracy": 0.4076923076923077,
+      "eval_loss": 3.212242603302002,
+      "eval_runtime": 2.4013,
+      "eval_samples_per_second": 54.137,
+      "eval_steps_per_second": 0.833,
       "step": 301
     },
     {
       "epoch": 44.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.7527263164520264,
-      "eval_runtime": 1.7657,
-      "eval_samples_per_second": 55.502,
-      "eval_steps_per_second": 1.133,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.2259228229522705,
+      "eval_runtime": 2.5062,
+      "eval_samples_per_second": 51.871,
+      "eval_steps_per_second": 0.798,
       "step": 308
     },
     {
       "epoch": 44.29,
       "learning_rate": 1.1142857142857143e-05,
-      "loss": 3.2772,
+      "loss": 2.554,
       "step": 310
     },
     {
       "epoch": 45.0,
-      "eval_accuracy": 0.12244897959183673,
-      "eval_loss": 3.749258041381836,
-      "eval_runtime": 1.6842,
-      "eval_samples_per_second": 58.188,
-      "eval_steps_per_second": 1.188,
+      "eval_accuracy": 0.4307692307692308,
+      "eval_loss": 3.2271060943603516,
+      "eval_runtime": 2.357,
+      "eval_samples_per_second": 55.154,
+      "eval_steps_per_second": 0.849,
       "step": 315
     },
     {
       "epoch": 45.71,
       "learning_rate": 1.0857142857142858e-05,
-      "loss": 3.2353,
+      "loss": 2.5222,
       "step": 320
     },
     {
       "epoch": 46.0,
-      "eval_accuracy": 0.11224489795918367,
-      "eval_loss": 3.7462167739868164,
-      "eval_runtime": 1.6536,
-      "eval_samples_per_second": 59.264,
-      "eval_steps_per_second": 1.209,
+      "eval_accuracy": 0.4461538461538462,
+      "eval_loss": 3.2208163738250732,
+      "eval_runtime": 2.7528,
+      "eval_samples_per_second": 47.225,
+      "eval_steps_per_second": 0.727,
       "step": 322
     },
     {
       "epoch": 47.0,
-      "eval_accuracy": 0.1326530612244898,
-      "eval_loss": 3.743089437484741,
-      "eval_runtime": 1.639,
-      "eval_samples_per_second": 59.793,
-      "eval_steps_per_second": 1.22,
+      "eval_accuracy": 0.4461538461538462,
+      "eval_loss": 3.2139155864715576,
+      "eval_runtime": 2.3673,
+      "eval_samples_per_second": 54.915,
+      "eval_steps_per_second": 0.845,
       "step": 329
     },
     {
       "epoch": 47.14,
       "learning_rate": 1.0571428571428572e-05,
-      "loss": 3.2198,
+      "loss": 2.5085,
       "step": 330
     },
     {
       "epoch": 48.0,
-      "eval_accuracy": 0.1326530612244898,
-      "eval_loss": 3.7392406463623047,
-      "eval_runtime": 1.6659,
-      "eval_samples_per_second": 58.827,
-      "eval_steps_per_second": 1.201,
+      "eval_accuracy": 0.45384615384615384,
+      "eval_loss": 3.204026460647583,
+      "eval_runtime": 2.3343,
+      "eval_samples_per_second": 55.691,
+      "eval_steps_per_second": 0.857,
       "step": 336
     },
     {
       "epoch": 48.57,
       "learning_rate": 1.0285714285714285e-05,
-      "loss": 3.204,
+      "loss": 2.4593,
       "step": 340
     },
     {
       "epoch": 49.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.73702073097229,
-      "eval_runtime": 1.7576,
-      "eval_samples_per_second": 55.758,
-      "eval_steps_per_second": 1.138,
+      "eval_accuracy": 0.49230769230769234,
+      "eval_loss": 3.205333948135376,
+      "eval_runtime": 2.3891,
+      "eval_samples_per_second": 54.414,
+      "eval_steps_per_second": 0.837,
       "step": 343
     },
     {
       "epoch": 50.0,
       "learning_rate": 1e-05,
-      "loss": 3.1762,
+      "loss": 2.4585,
       "step": 350
     },
     {
       "epoch": 50.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.733855962753296,
-      "eval_runtime": 1.6472,
-      "eval_samples_per_second": 59.494,
-      "eval_steps_per_second": 1.214,
+      "eval_accuracy": 0.47692307692307695,
+      "eval_loss": 3.1822495460510254,
+      "eval_runtime": 2.3305,
+      "eval_samples_per_second": 55.781,
+      "eval_steps_per_second": 0.858,
       "step": 350
     },
     {
       "epoch": 51.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.730581283569336,
-      "eval_runtime": 1.6737,
-      "eval_samples_per_second": 58.554,
-      "eval_steps_per_second": 1.195,
+      "eval_accuracy": 0.46923076923076923,
+      "eval_loss": 3.1696622371673584,
+      "eval_runtime": 2.3113,
+      "eval_samples_per_second": 56.246,
+      "eval_steps_per_second": 0.865,
       "step": 357
     },
     {
       "epoch": 51.43,
       "learning_rate": 9.714285714285715e-06,
-      "loss": 3.1741,
+      "loss": 2.4228,
       "step": 360
     },
     {
       "epoch": 52.0,
-      "eval_accuracy": 0.16326530612244897,
-      "eval_loss": 3.7267162799835205,
-      "eval_runtime": 1.669,
-      "eval_samples_per_second": 58.718,
-      "eval_steps_per_second": 1.198,
+      "eval_accuracy": 0.46923076923076923,
+      "eval_loss": 3.1589303016662598,
+      "eval_runtime": 2.5283,
+      "eval_samples_per_second": 51.419,
+      "eval_steps_per_second": 0.791,
       "step": 364
     },
     {
       "epoch": 52.86,
       "learning_rate": 9.42857142857143e-06,
-      "loss": 3.1757,
+      "loss": 2.3954,
       "step": 370
     },
     {
       "epoch": 53.0,
-      "eval_accuracy": 0.16326530612244897,
-      "eval_loss": 3.7221927642822266,
-      "eval_runtime": 1.6595,
-      "eval_samples_per_second": 59.054,
-      "eval_steps_per_second": 1.205,
+      "eval_accuracy": 0.47692307692307695,
+      "eval_loss": 3.1375324726104736,
+      "eval_runtime": 2.3577,
+      "eval_samples_per_second": 55.139,
+      "eval_steps_per_second": 0.848,
       "step": 371
     },
     {
       "epoch": 54.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.717994451522827,
-      "eval_runtime": 1.6603,
-      "eval_samples_per_second": 59.027,
-      "eval_steps_per_second": 1.205,
+      "eval_accuracy": 0.45384615384615384,
+      "eval_loss": 3.1092171669006348,
+      "eval_runtime": 2.3799,
+      "eval_samples_per_second": 54.625,
+      "eval_steps_per_second": 0.84,
       "step": 378
     },
     {
       "epoch": 54.29,
       "learning_rate": 9.142857142857144e-06,
-      "loss": 3.1492,
+      "loss": 2.3641,
       "step": 380
     },
     {
       "epoch": 55.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.7148733139038086,
-      "eval_runtime": 1.6452,
-      "eval_samples_per_second": 59.566,
-      "eval_steps_per_second": 1.216,
+      "eval_accuracy": 0.47692307692307695,
+      "eval_loss": 3.099919319152832,
+      "eval_runtime": 2.3488,
+      "eval_samples_per_second": 55.348,
+      "eval_steps_per_second": 0.852,
       "step": 385
     },
     {
       "epoch": 55.71,
       "learning_rate": 8.857142857142858e-06,
-      "loss": 3.1442,
+      "loss": 2.3651,
       "step": 390
     },
     {
       "epoch": 56.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.7107248306274414,
-      "eval_runtime": 1.6947,
-      "eval_samples_per_second": 57.827,
-      "eval_steps_per_second": 1.18,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.0859622955322266,
+      "eval_runtime": 2.5127,
+      "eval_samples_per_second": 51.737,
+      "eval_steps_per_second": 0.796,
       "step": 392
     },
     {
       "epoch": 57.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.7084951400756836,
-      "eval_runtime": 1.6087,
-      "eval_samples_per_second": 60.919,
-      "eval_steps_per_second": 1.243,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.0813159942626953,
+      "eval_runtime": 2.449,
+      "eval_samples_per_second": 53.082,
+      "eval_steps_per_second": 0.817,
       "step": 399
     },
     {
       "epoch": 57.14,
       "learning_rate": 8.571428571428571e-06,
-      "loss": 3.1174,
+      "loss": 2.3182,
       "step": 400
     },
     {
       "epoch": 58.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.705909013748169,
-      "eval_runtime": 1.7081,
-      "eval_samples_per_second": 57.372,
-      "eval_steps_per_second": 1.171,
+      "eval_accuracy": 0.49230769230769234,
+      "eval_loss": 3.0692226886749268,
+      "eval_runtime": 2.3556,
+      "eval_samples_per_second": 55.188,
+      "eval_steps_per_second": 0.849,
       "step": 406
     },
     {
       "epoch": 58.57,
       "learning_rate": 8.285714285714287e-06,
-      "loss": 3.0962,
+      "loss": 2.3029,
       "step": 410
     },
     {
       "epoch": 59.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.7031126022338867,
-      "eval_runtime": 1.5815,
-      "eval_samples_per_second": 61.967,
-      "eval_steps_per_second": 1.265,
+      "eval_accuracy": 0.4846153846153846,
+      "eval_loss": 3.061020612716675,
+      "eval_runtime": 2.4546,
+      "eval_samples_per_second": 52.961,
+      "eval_steps_per_second": 0.815,
       "step": 413
     },
     {
       "epoch": 60.0,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 3.1237,
+      "loss": 2.2988,
       "step": 420
     },
     {
       "epoch": 60.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.701946258544922,
-      "eval_runtime": 1.7026,
-      "eval_samples_per_second": 57.559,
-      "eval_steps_per_second": 1.175,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.0626943111419678,
+      "eval_runtime": 2.3415,
+      "eval_samples_per_second": 55.521,
+      "eval_steps_per_second": 0.854,
       "step": 420
     },
     {
       "epoch": 61.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6996471881866455,
-      "eval_runtime": 1.6975,
-      "eval_samples_per_second": 57.732,
-      "eval_steps_per_second": 1.178,
+      "eval_accuracy": 0.46923076923076923,
+      "eval_loss": 3.051997184753418,
+      "eval_runtime": 2.492,
+      "eval_samples_per_second": 52.167,
+      "eval_steps_per_second": 0.803,
       "step": 427
     },
     {
       "epoch": 61.43,
       "learning_rate": 7.714285714285716e-06,
-      "loss": 3.1229,
+      "loss": 2.2865,
       "step": 430
     },
     {
       "epoch": 62.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6955974102020264,
-      "eval_runtime": 1.6826,
-      "eval_samples_per_second": 58.242,
-      "eval_steps_per_second": 1.189,
+      "eval_accuracy": 0.45384615384615384,
+      "eval_loss": 3.039459705352783,
+      "eval_runtime": 2.4542,
+      "eval_samples_per_second": 52.97,
+      "eval_steps_per_second": 0.815,
       "step": 434
     },
     {
       "epoch": 62.86,
       "learning_rate": 7.428571428571429e-06,
-      "loss": 3.0946,
+      "loss": 2.2623,
       "step": 440
     },
     {
       "epoch": 63.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.692981719970703,
-      "eval_runtime": 1.7683,
-      "eval_samples_per_second": 55.42,
-      "eval_steps_per_second": 1.131,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.0356502532958984,
+      "eval_runtime": 2.3341,
+      "eval_samples_per_second": 55.697,
+      "eval_steps_per_second": 0.857,
       "step": 441
     },
     {
       "epoch": 64.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6915957927703857,
-      "eval_runtime": 1.6752,
-      "eval_samples_per_second": 58.501,
-      "eval_steps_per_second": 1.194,
+      "eval_accuracy": 0.46153846153846156,
+      "eval_loss": 3.0332510471343994,
+      "eval_runtime": 2.3535,
+      "eval_samples_per_second": 55.236,
+      "eval_steps_per_second": 0.85,
       "step": 448
     },
     {
       "epoch": 64.29,
       "learning_rate": 7.1428571428571436e-06,
-      "loss": 3.0861,
+      "loss": 2.2252,
       "step": 450
     },
     {
       "epoch": 65.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6893138885498047,
-      "eval_runtime": 1.6711,
-      "eval_samples_per_second": 58.644,
-      "eval_steps_per_second": 1.197,
+      "eval_accuracy": 0.47692307692307695,
+      "eval_loss": 3.022916316986084,
+      "eval_runtime": 2.5362,
+      "eval_samples_per_second": 51.258,
+      "eval_steps_per_second": 0.789,
       "step": 455
     },
     {
       "epoch": 65.71,
       "learning_rate": 6.857142857142858e-06,
-      "loss": 3.0406,
+      "loss": 2.2339,
       "step": 460
     },
     {
       "epoch": 66.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6859352588653564,
-      "eval_runtime": 1.7238,
-      "eval_samples_per_second": 56.85,
-      "eval_steps_per_second": 1.16,
+      "eval_accuracy": 0.47692307692307695,
+      "eval_loss": 3.020322799682617,
+      "eval_runtime": 2.3519,
+      "eval_samples_per_second": 55.275,
+      "eval_steps_per_second": 0.85,
       "step": 462
     },
     {
       "epoch": 67.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6839077472686768,
-      "eval_runtime": 1.6549,
-      "eval_samples_per_second": 59.218,
-      "eval_steps_per_second": 1.209,
+      "eval_accuracy": 0.49230769230769234,
+      "eval_loss": 3.0076119899749756,
+      "eval_runtime": 2.359,
+      "eval_samples_per_second": 55.107,
+      "eval_steps_per_second": 0.848,
       "step": 469
     },
     {
       "epoch": 67.14,
       "learning_rate": 6.571428571428572e-06,
-      "loss": 3.077,
+      "loss": 2.2017,
       "step": 470
     },
     {
       "epoch": 68.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6815552711486816,
-      "eval_runtime": 1.673,
-      "eval_samples_per_second": 58.578,
-      "eval_steps_per_second": 1.195,
+      "eval_accuracy": 0.4846153846153846,
+      "eval_loss": 2.9876248836517334,
+      "eval_runtime": 2.372,
+      "eval_samples_per_second": 54.805,
+      "eval_steps_per_second": 0.843,
       "step": 476
     },
     {
       "epoch": 68.57,
       "learning_rate": 6.285714285714286e-06,
-      "loss": 3.0555,
+      "loss": 2.1972,
       "step": 480
     },
     {
       "epoch": 69.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.678163766860962,
-      "eval_runtime": 1.6518,
-      "eval_samples_per_second": 59.329,
-      "eval_steps_per_second": 1.211,
+      "eval_accuracy": 0.49230769230769234,
+      "eval_loss": 2.971572160720825,
+      "eval_runtime": 2.3479,
+      "eval_samples_per_second": 55.368,
+      "eval_steps_per_second": 0.852,
       "step": 483
     },
     {
       "epoch": 70.0,
       "learning_rate": 6e-06,
-      "loss": 3.035,
+      "loss": 2.1964,
       "step": 490
     },
     {
       "epoch": 70.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6762583255767822,
-      "eval_runtime": 1.6591,
-      "eval_samples_per_second": 59.069,
-      "eval_steps_per_second": 1.205,
+      "eval_accuracy": 0.5,
+      "eval_loss": 2.96323823928833,
+      "eval_runtime": 2.4351,
+      "eval_samples_per_second": 53.387,
+      "eval_steps_per_second": 0.821,
       "step": 490
     },
     {
       "epoch": 71.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.672853469848633,
-      "eval_runtime": 1.6495,
-      "eval_samples_per_second": 59.412,
-      "eval_steps_per_second": 1.212,
+      "eval_accuracy": 0.49230769230769234,
+      "eval_loss": 2.9597153663635254,
+      "eval_runtime": 2.4929,
+      "eval_samples_per_second": 52.149,
+      "eval_steps_per_second": 0.802,
       "step": 497
     },
     {
       "epoch": 71.43,
       "learning_rate": 5.7142857142857145e-06,
-      "loss": 3.0246,
+      "loss": 2.1775,
       "step": 500
     },
     {
       "epoch": 72.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6719117164611816,
-      "eval_runtime": 1.6901,
-      "eval_samples_per_second": 57.986,
-      "eval_steps_per_second": 1.183,
+      "eval_accuracy": 0.5,
+      "eval_loss": 2.9580914974212646,
+      "eval_runtime": 2.5027,
+      "eval_samples_per_second": 51.943,
+      "eval_steps_per_second": 0.799,
       "step": 504
     },
     {
       "epoch": 72.86,
       "learning_rate": 5.428571428571429e-06,
-      "loss": 3.0282,
+      "loss": 2.1619,
       "step": 510
     },
     {
       "epoch": 73.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.670848846435547,
-      "eval_runtime": 1.7293,
-      "eval_samples_per_second": 56.669,
-      "eval_steps_per_second": 1.157,
+      "eval_accuracy": 0.5076923076923077,
+      "eval_loss": 2.9516122341156006,
+      "eval_runtime": 2.4828,
+      "eval_samples_per_second": 52.361,
+      "eval_steps_per_second": 0.806,
       "step": 511
     },
     {
       "epoch": 74.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.6683461666107178,
-      "eval_runtime": 1.6959,
-      "eval_samples_per_second": 57.785,
-      "eval_steps_per_second": 1.179,
+      "eval_accuracy": 0.5153846153846153,
+      "eval_loss": 2.935607433319092,
+      "eval_runtime": 2.3758,
+      "eval_samples_per_second": 54.719,
+      "eval_steps_per_second": 0.842,
       "step": 518
     },
     {
       "epoch": 74.29,
       "learning_rate": 5.142857142857142e-06,
-      "loss": 3.0293,
+      "loss": 2.1633,
       "step": 520
     },
     {
       "epoch": 75.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.665170669555664,
-      "eval_runtime": 1.6438,
-      "eval_samples_per_second": 59.617,
-      "eval_steps_per_second": 1.217,
+      "eval_accuracy": 0.5076923076923077,
+      "eval_loss": 2.928561210632324,
+      "eval_runtime": 2.3327,
+      "eval_samples_per_second": 55.729,
+      "eval_steps_per_second": 0.857,
       "step": 525
     },
     {
       "epoch": 75.71,
       "learning_rate": 4.857142857142858e-06,
-      "loss": 2.9893,
+      "loss": 2.1207,
       "step": 530
     },
     {
       "epoch": 76.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.66402268409729,
-      "eval_runtime": 1.6852,
-      "eval_samples_per_second": 58.155,
-      "eval_steps_per_second": 1.187,
+      "eval_accuracy": 0.5153846153846153,
+      "eval_loss": 2.9266109466552734,
+      "eval_runtime": 2.3813,
+      "eval_samples_per_second": 54.591,
+      "eval_steps_per_second": 0.84,
       "step": 532
     },
     {
       "epoch": 77.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.6634998321533203,
-      "eval_runtime": 1.7537,
-      "eval_samples_per_second": 55.882,
-      "eval_steps_per_second": 1.14,
+      "eval_accuracy": 0.5230769230769231,
+      "eval_loss": 2.920504093170166,
+      "eval_runtime": 2.4777,
+      "eval_samples_per_second": 52.468,
+      "eval_steps_per_second": 0.807,
       "step": 539
     },
     {
       "epoch": 77.14,
       "learning_rate": 4.571428571428572e-06,
-      "loss": 2.9888,
+      "loss": 2.1353,
       "step": 540
     },
     {
       "epoch": 78.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.6618170738220215,
-      "eval_runtime": 1.7774,
-      "eval_samples_per_second": 55.135,
-      "eval_steps_per_second": 1.125,
+      "eval_accuracy": 0.5153846153846153,
+      "eval_loss": 2.9131386280059814,
+      "eval_runtime": 2.3865,
+      "eval_samples_per_second": 54.473,
+      "eval_steps_per_second": 0.838,
       "step": 546
     },
     {
       "epoch": 78.57,
       "learning_rate": 4.2857142857142855e-06,
-      "loss": 2.9833,
+      "loss": 2.1075,
       "step": 550
     },
     {
       "epoch": 79.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.659451484680176,
-      "eval_runtime": 1.6474,
-      "eval_samples_per_second": 59.487,
-      "eval_steps_per_second": 1.214,
+      "eval_accuracy": 0.5230769230769231,
+      "eval_loss": 2.907454252243042,
+      "eval_runtime": 2.3713,
+      "eval_samples_per_second": 54.823,
+      "eval_steps_per_second": 0.843,
       "step": 553
     },
     {
       "epoch": 80.0,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 2.9739,
+      "loss": 2.1025,
       "step": 560
     },
     {
       "epoch": 80.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.6578376293182373,
-      "eval_runtime": 1.6531,
-      "eval_samples_per_second": 59.283,
-      "eval_steps_per_second": 1.21,
+      "eval_accuracy": 0.5230769230769231,
+      "eval_loss": 2.907280921936035,
+      "eval_runtime": 2.3736,
+      "eval_samples_per_second": 54.768,
+      "eval_steps_per_second": 0.843,
       "step": 560
     },
     {
       "epoch": 81.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.656160593032837,
-      "eval_runtime": 1.7409,
-      "eval_samples_per_second": 56.294,
-      "eval_steps_per_second": 1.149,
+      "eval_accuracy": 0.5153846153846153,
+      "eval_loss": 2.9174439907073975,
+      "eval_runtime": 2.3602,
+      "eval_samples_per_second": 55.079,
+      "eval_steps_per_second": 0.847,
       "step": 567
     },
     {
       "epoch": 81.43,
       "learning_rate": 3.7142857142857146e-06,
-      "loss": 2.9513,
+      "loss": 2.1031,
       "step": 570
     },
     {
       "epoch": 82.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.655242681503296,
-      "eval_runtime": 1.6811,
-      "eval_samples_per_second": 58.294,
-      "eval_steps_per_second": 1.19,
+      "eval_accuracy": 0.5307692307692308,
+      "eval_loss": 2.913137197494507,
+      "eval_runtime": 2.4712,
+      "eval_samples_per_second": 52.605,
+      "eval_steps_per_second": 0.809,
       "step": 574
     },
     {
       "epoch": 82.86,
       "learning_rate": 3.428571428571429e-06,
-      "loss": 2.9503,
+      "loss": 2.0932,
       "step": 580
     },
     {
       "epoch": 83.0,
-      "eval_accuracy": 0.14285714285714285,
-      "eval_loss": 3.6539218425750732,
-      "eval_runtime": 1.7574,
-      "eval_samples_per_second": 55.764,
-      "eval_steps_per_second": 1.138,
+      "eval_accuracy": 0.5307692307692308,
+      "eval_loss": 2.909245491027832,
+      "eval_runtime": 2.3787,
+      "eval_samples_per_second": 54.651,
+      "eval_steps_per_second": 0.841,
       "step": 581
     },
     {
       "epoch": 84.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.653193950653076,
-      "eval_runtime": 1.6302,
-      "eval_samples_per_second": 60.114,
-      "eval_steps_per_second": 1.227,
+      "eval_accuracy": 0.5307692307692308,
+      "eval_loss": 2.8977668285369873,
+      "eval_runtime": 4.0301,
+      "eval_samples_per_second": 32.257,
+      "eval_steps_per_second": 0.496,
       "step": 588
     },
     {
       "epoch": 84.29,
       "learning_rate": 3.142857142857143e-06,
-      "loss": 2.9792,
+      "loss": 2.0861,
       "step": 590
     },
     {
       "epoch": 85.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.651691436767578,
-      "eval_runtime": 1.7277,
-      "eval_samples_per_second": 56.724,
-      "eval_steps_per_second": 1.158,
+      "eval_accuracy": 0.5307692307692308,
+      "eval_loss": 2.8871328830718994,
+      "eval_runtime": 3.7091,
+      "eval_samples_per_second": 35.049,
+      "eval_steps_per_second": 0.539,
       "step": 595
     },
     {
       "epoch": 85.71,
       "learning_rate": 2.8571428571428573e-06,
-      "loss": 2.9561,
+      "loss": 2.0478,
       "step": 600
     },
     {
       "epoch": 86.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6496691703796387,
-      "eval_runtime": 1.6772,
-      "eval_samples_per_second": 58.432,
-      "eval_steps_per_second": 1.192,
+      "eval_accuracy": 0.5384615384615384,
+      "eval_loss": 2.882897138595581,
+      "eval_runtime": 4.3893,
+      "eval_samples_per_second": 29.617,
+      "eval_steps_per_second": 0.456,
       "step": 602
     },
     {
       "epoch": 87.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6485908031463623,
-      "eval_runtime": 1.664,
-      "eval_samples_per_second": 58.893,
-      "eval_steps_per_second": 1.202,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.880449056625366,
+      "eval_runtime": 2.9082,
+      "eval_samples_per_second": 44.701,
+      "eval_steps_per_second": 0.688,
       "step": 609
     },
     {
       "epoch": 87.14,
       "learning_rate": 2.571428571428571e-06,
-      "loss": 2.964,
+      "loss": 2.0815,
       "step": 610
     },
     {
       "epoch": 88.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.647573709487915,
-      "eval_runtime": 1.7778,
-      "eval_samples_per_second": 55.124,
-      "eval_steps_per_second": 1.125,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.872459888458252,
+      "eval_runtime": 2.8921,
+      "eval_samples_per_second": 44.95,
+      "eval_steps_per_second": 0.692,
       "step": 616
     },
     {
       "epoch": 88.57,
       "learning_rate": 2.285714285714286e-06,
-      "loss": 2.9665,
+      "loss": 2.0756,
       "step": 620
     },
     {
       "epoch": 89.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6470184326171875,
-      "eval_runtime": 1.6689,
-      "eval_samples_per_second": 58.722,
-      "eval_steps_per_second": 1.198,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.8694419860839844,
+      "eval_runtime": 3.9643,
+      "eval_samples_per_second": 32.793,
+      "eval_steps_per_second": 0.505,
       "step": 623
     },
     {
       "epoch": 90.0,
       "learning_rate": 2.0000000000000003e-06,
-      "loss": 2.9439,
+      "loss": 2.065,
       "step": 630
     },
     {
       "epoch": 90.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.646164894104004,
-      "eval_runtime": 2.2025,
-      "eval_samples_per_second": 44.495,
-      "eval_steps_per_second": 0.908,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.866473913192749,
+      "eval_runtime": 3.1035,
+      "eval_samples_per_second": 41.888,
+      "eval_steps_per_second": 0.644,
       "step": 630
     },
     {
       "epoch": 91.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6452953815460205,
-      "eval_runtime": 1.6826,
-      "eval_samples_per_second": 58.242,
-      "eval_steps_per_second": 1.189,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.861470937728882,
+      "eval_runtime": 3.1207,
+      "eval_samples_per_second": 41.657,
+      "eval_steps_per_second": 0.641,
       "step": 637
     },
     {
       "epoch": 91.43,
       "learning_rate": 1.7142857142857145e-06,
-      "loss": 2.9369,
+      "loss": 2.0572,
       "step": 640
     },
     {
       "epoch": 92.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.645128011703491,
-      "eval_runtime": 1.646,
-      "eval_samples_per_second": 59.539,
-      "eval_steps_per_second": 1.215,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.859877347946167,
+      "eval_runtime": 2.6425,
+      "eval_samples_per_second": 49.197,
+      "eval_steps_per_second": 0.757,
       "step": 644
     },
     {
       "epoch": 92.86,
       "learning_rate": 1.4285714285714286e-06,
-      "loss": 2.9619,
+      "loss": 2.0358,
       "step": 650
     },
     {
       "epoch": 93.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.64505934715271,
-      "eval_runtime": 1.7663,
-      "eval_samples_per_second": 55.483,
-      "eval_steps_per_second": 1.132,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.8619918823242188,
+      "eval_runtime": 2.4174,
+      "eval_samples_per_second": 53.776,
+      "eval_steps_per_second": 0.827,
       "step": 651
     },
     {
       "epoch": 94.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6448814868927,
-      "eval_runtime": 1.6466,
-      "eval_samples_per_second": 59.516,
-      "eval_steps_per_second": 1.215,
+      "eval_accuracy": 0.5461538461538461,
+      "eval_loss": 2.862945079803467,
+      "eval_runtime": 2.4216,
+      "eval_samples_per_second": 53.684,
+      "eval_steps_per_second": 0.826,
       "step": 658
     },
     {
       "epoch": 94.29,
       "learning_rate": 1.142857142857143e-06,
-      "loss": 2.955,
+      "loss": 2.0663,
       "step": 660
     },
     {
       "epoch": 95.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6444060802459717,
-      "eval_runtime": 1.6863,
-      "eval_samples_per_second": 58.115,
-      "eval_steps_per_second": 1.186,
+      "eval_accuracy": 0.5538461538461539,
+      "eval_loss": 2.862516403198242,
+      "eval_runtime": 2.3749,
+      "eval_samples_per_second": 54.739,
+      "eval_steps_per_second": 0.842,
       "step": 665
     },
     {
       "epoch": 95.71,
       "learning_rate": 8.571428571428572e-07,
-      "loss": 2.9323,
+      "loss": 2.0353,
       "step": 670
     },
     {
       "epoch": 96.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6440815925598145,
-      "eval_runtime": 1.626,
-      "eval_samples_per_second": 60.271,
-      "eval_steps_per_second": 1.23,
+      "eval_accuracy": 0.5538461538461539,
+      "eval_loss": 2.8627583980560303,
+      "eval_runtime": 2.3527,
+      "eval_samples_per_second": 55.257,
+      "eval_steps_per_second": 0.85,
       "step": 672
     },
     {
       "epoch": 97.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6438138484954834,
-      "eval_runtime": 1.7724,
-      "eval_samples_per_second": 55.292,
-      "eval_steps_per_second": 1.128,
+      "eval_accuracy": 0.5538461538461539,
+      "eval_loss": 2.8628671169281006,
+      "eval_runtime": 2.5748,
+      "eval_samples_per_second": 50.49,
+      "eval_steps_per_second": 0.777,
       "step": 679
     },
     {
       "epoch": 97.14,
       "learning_rate": 5.714285714285715e-07,
-      "loss": 2.9466,
+      "loss": 2.0506,
       "step": 680
     },
     {
       "epoch": 98.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.643672466278076,
-      "eval_runtime": 1.7638,
-      "eval_samples_per_second": 55.56,
-      "eval_steps_per_second": 1.134,
+      "eval_accuracy": 0.5538461538461539,
+      "eval_loss": 2.862170457839966,
+      "eval_runtime": 2.3987,
+      "eval_samples_per_second": 54.196,
+      "eval_steps_per_second": 0.834,
       "step": 686
     },
     {
       "epoch": 98.57,
       "learning_rate": 2.8571428571428575e-07,
-      "loss": 2.945,
+      "loss": 2.0494,
       "step": 690
     },
     {
       "epoch": 99.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.6435706615448,
-      "eval_runtime": 1.6937,
-      "eval_samples_per_second": 57.862,
-      "eval_steps_per_second": 1.181,
+      "eval_accuracy": 0.5538461538461539,
+      "eval_loss": 2.8621606826782227,
+      "eval_runtime": 2.676,
+      "eval_samples_per_second": 48.58,
+      "eval_steps_per_second": 0.747,
       "step": 693
     },
     {
       "epoch": 100.0,
       "learning_rate": 0.0,
-      "loss": 2.9665,
+      "loss": 2.0566,
       "step": 700
     },
     {
       "epoch": 100.0,
-      "eval_accuracy": 0.15306122448979592,
-      "eval_loss": 3.643535614013672,
-      "eval_runtime": 1.66,
-      "eval_samples_per_second": 59.037,
-      "eval_steps_per_second": 1.205,
+      "eval_accuracy": 0.5538461538461539,
+      "eval_loss": 2.862189292907715,
+      "eval_runtime": 2.8132,
+      "eval_samples_per_second": 46.211,
+      "eval_steps_per_second": 0.711,
       "step": 700
     },
     {
       "epoch": 100.0,
       "step": 700,
-      "total_flos": 4.2794747466153984e+18,
-      "train_loss": 3.289549721309117,
-      "train_runtime": 4059.8356,
-      "train_samples_per_second": 13.597,
-      "train_steps_per_second": 0.172
+      "total_flos": 4.031389254057984e+18,
+      "train_loss": 2.6316724504743303,
+      "train_runtime": 4344.3701,
+      "train_samples_per_second": 11.97,
+      "train_steps_per_second": 0.161
     }
   ],
   "logging_steps": 10,
@@ -1343,7 +1343,7 @@
   "num_input_tokens_seen": 0,
   "num_train_epochs": 100,
   "save_steps": 500,
-  "total_flos": 4.2794747466153984e+18,
+  "total_flos": 4.031389254057984e+18,
   "train_batch_size": 80,
   "trial_name": null,
   "trial_params": null